From ca4ea8b9fbe64554fc707b020e866c721b70d35d Mon Sep 17 00:00:00 2001
From: Lee Jackson <130007945+Imagineer99@users.noreply.github.com>
Date: Thu, 2 Apr 2026 03:57:10 +0100
Subject: [PATCH 001/216] studio: align composer/code, unify fonts, and remove
 tool collapse jitter (#4763)

- Add min-w-0 guards to thread/message/markdown containers to prevent
  content overflow past the composer width
- Unify chat typography from Hellix/Space Grotesk to the sans stack,
  keeping monospace for code blocks and inline code
- Restructure desktop navbar right-side controls with shrink-0 wrappers
  for consistent spacing across HoverCard roots
- Soften tool-call label styling (font-medium + text-foreground/85
  instead of bold)
- Add responsive code block sizing via @container queries
- Add horizontal scrolling for wide code blocks within the thread column
- Scope list-item code block alignment CSS to .aui-thread-root
- Preserve useScrollLock in tool-fallback and tool-group collapsibles
- Fall back to bg-background on ViewportFooter when hideComposer is true
- Widen inline code monospace selector to cover th, blockquote, and
  heading elements
- Remove unused @fontsource-variable/space-grotesk import
---
 .../components/assistant-ui/markdown-text.tsx |   2 +-
 .../src/components/assistant-ui/thread.tsx    |  40 +++--
 .../components/assistant-ui/tool-fallback.tsx |   6 +-
 studio/frontend/src/components/navbar.tsx     | 164 +++++++++---------
 .../src/features/chat/shared-composer.tsx     |   2 +-
 studio/frontend/src/index.css                 |  72 ++++++--
 6 files changed, 175 insertions(+), 111 deletions(-)

diff --git a/studio/frontend/src/components/assistant-ui/markdown-text.tsx b/studio/frontend/src/components/assistant-ui/markdown-text.tsx
index 5e84b9175e..17a6f8f054 100644
--- a/studio/frontend/src/components/assistant-ui/markdown-text.tsx
+++ b/studio/frontend/src/components/assistant-ui/markdown-text.tsx
@@ -403,7 +403,7 @@ const MarkdownTextImpl = () => {
   }
 
   return (
-    <div data-status={status.type}>
+    <div data-status={status.type} className="min-w-0 max-w-full">
       <Streamdown
         mode="streaming"
         isAnimating={status.type === "running"}
diff --git a/studio/frontend/src/components/assistant-ui/thread.tsx b/studio/frontend/src/components/assistant-ui/thread.tsx
index aae027c74b..5170f6ff69 100644
--- a/studio/frontend/src/components/assistant-ui/thread.tsx
+++ b/studio/frontend/src/components/assistant-ui/thread.tsx
@@ -73,7 +73,7 @@ export const Thread: FC<{ hideComposer?: boolean; hideWelcome?: boolean }> = ({
       }}
     >
       <ThreadPrimitive.Viewport
-        className="aui-thread-viewport relative flex flex-1 flex-col overflow-x-auto overflow-y-scroll scroll-smooth px-4 pt-4"
+        className="aui-thread-viewport relative flex min-w-0 flex-1 flex-col overflow-x-auto overflow-y-scroll scroll-smooth px-4 pt-4"
       >
         {!hideWelcome && (
           <AuiIf condition={({ thread }) => thread.isEmpty}>
@@ -89,7 +89,13 @@ export const Thread: FC<{ hideComposer?: boolean; hideWelcome?: boolean }> = ({
           }}
         />
 
-        <ThreadPrimitive.ViewportFooter className="aui-thread-viewport-footer sticky bottom-0 z-20 mt-auto flex w-full flex-col gap-4 overflow-visible bg-background pb-4 md:pb-4">
+        <ThreadPrimitive.ViewportFooter className={cn(
+          "aui-thread-viewport-footer sticky bottom-0 z-20 mt-auto flex w-full flex-col gap-4 overflow-visible pb-4 md:pb-4",
+          hideComposer ? "bg-background" : "relative bg-transparent",
+        )}>
+          {!hideComposer && (
+            <div className="pointer-events-none absolute inset-x-0 bottom-0 h-4 bg-background" aria-hidden />
+          )}
           <ThreadScrollToBottom />
           <AuiIf condition={({ thread }) => !thread.isEmpty}>
             {!hideComposer && <ComposerAnimated />}
@@ -220,14 +226,20 @@ const GeneratingSpinner: FC = () => {
 
 const ComposerAnimated: FC = () => {
   return (
-    <motion.div
-      layout={true}
-      layoutId="composer"
-      transition={{ type: "spring", bounce: 0.15, duration: 0.5 }}
-      className="mx-auto w-full max-w-(--thread-max-width)"
-    >
-      <Composer />
-    </motion.div>
+    <div className="relative mx-auto min-w-0 w-full max-w-(--thread-max-width)">
+      <div
+        className="pointer-events-none absolute inset-x-0 top-1/2 bottom-0 z-0 bg-background"
+        aria-hidden
+      />
+      <motion.div
+        layout={true}
+        layoutId="composer"
+        transition={{ type: "spring", bounce: 0.15, duration: 0.5 }}
+        className="relative z-10 w-full"
+      >
+        <Composer />
+      </motion.div>
+    </div>
   );
 };
 
@@ -262,7 +274,7 @@ const Composer: FC = () => {
         <ToolStatusDisplay />
         <ComposerPrimitive.Input
           placeholder="Send a message..."
-          className="aui-composer-input mb-1 max-h-32 min-h-12 w-full resize-none bg-transparent px-4 pt-2 pb-3 text-sm outline-none placeholder:text-muted-foreground focus-visible:ring-0"
+          className="aui-composer-input mb-1 max-h-32 min-h-12 w-full resize-none bg-transparent pl-5 pr-4 pt-2 pb-3 text-sm outline-none placeholder:text-muted-foreground focus-visible:ring-0"
           rows={1}
           autoFocus={true}
           aria-label="Message input"
@@ -574,10 +586,10 @@ const GeneratingIndicator: FC = () => {
 const AssistantMessage: FC = () => {
   return (
     <MessagePrimitive.Root
-      className="aui-assistant-message-root fade-in slide-in-from-bottom-1 relative mx-auto w-full max-w-(--thread-max-width) animate-in py-3 duration-150"
+      className="aui-assistant-message-root fade-in slide-in-from-bottom-1 relative mx-auto min-w-0 w-full max-w-(--thread-max-width) animate-in py-3 duration-150"
       data-role="assistant"
     >
-      <div className="aui-assistant-message-content wrap-break-word px-2 text-foreground leading-relaxed">
+      <div className="aui-assistant-message-content wrap-break-word min-w-0 text-foreground leading-relaxed">
         <GeneratingIndicator />
         <MessagePrimitive.Parts
           components={{
@@ -600,7 +612,7 @@ const AssistantMessage: FC = () => {
         <MessageError />
       </div>
 
-      <div className="aui-assistant-message-footer mt-1 ml-2 flex">
+      <div className="aui-assistant-message-footer mt-1 flex">
         <BranchPicker />
         <AssistantActionBar />
       </div>
diff --git a/studio/frontend/src/components/assistant-ui/tool-fallback.tsx b/studio/frontend/src/components/assistant-ui/tool-fallback.tsx
index c965749f74..82a5b17e04 100644
--- a/studio/frontend/src/components/assistant-ui/tool-fallback.tsx
+++ b/studio/frontend/src/components/assistant-ui/tool-fallback.tsx
@@ -156,12 +156,12 @@ function ToolFallbackTrigger({
       <span
         data-slot="tool-fallback-trigger-label"
         className={cn(
-          "aui-tool-fallback-trigger-label-wrapper relative inline-block grow text-left leading-none",
+          "aui-tool-fallback-trigger-label-wrapper relative inline-block grow text-left leading-none text-muted-foreground",
           isCancelled && "text-muted-foreground line-through",
         )}
       >
         <span>
-          {label}: <b>{toolName}</b>
+          {label}: <span className="font-medium text-foreground/85">{toolName}</span>
         </span>
         {isRunning && (
           <span
@@ -169,7 +169,7 @@ function ToolFallbackTrigger({
             data-slot="tool-fallback-trigger-shimmer"
             className="aui-tool-fallback-trigger-shimmer shimmer pointer-events-none absolute inset-0 motion-reduce:animate-none"
           >
-            {label}: <b>{toolName}</b>
+            {label}: <span className="font-medium text-foreground/85">{toolName}</span>
           </span>
         )}
       </span>
diff --git a/studio/frontend/src/components/navbar.tsx b/studio/frontend/src/components/navbar.tsx
index 63189a0d5d..121c559db8 100644
--- a/studio/frontend/src/components/navbar.tsx
+++ b/studio/frontend/src/components/navbar.tsx
@@ -372,90 +372,94 @@ export function Navbar() {
           })}
         </nav>
 
-        {/* Right: docs/tour desktop */}
-        <div className="hidden items-center justify-self-end gap-2 md:flex">
-          <AnimatedThemeToggler
-            className="flex h-9 w-9 items-center justify-center rounded-md text-muted-foreground transition-colors hover:bg-accent hover:text-foreground [&_svg]:size-4"
-            title="Toggle theme"
-            aria-label="Toggle theme"
-          />
-          <HoverCard openDelay={200} closeDelay={100}>
-            <HoverCardTrigger asChild={true}>
-              <a
-                href="https://unsloth.ai/docs"
-                target="_blank"
-                rel="noopener noreferrer"
-                className="flex items-center gap-1.5 text-sm font-medium text-emerald-600 hover:text-emerald-700 transition-colors"
-              >
-                <HugeiconsIcon icon={Book03Icon} className="size-4" />
-                Learn more
-              </a>
-            </HoverCardTrigger>
-            <HoverCardContent align="end" className="w-80 p-0">
-              <a
-                href="https://unsloth.ai/docs"
-                target="_blank"
-                rel="noopener noreferrer"
-                className="group/card flex flex-col gap-1 p-4 no-underline"
-              >
-                <p className="text-sm font-semibold font-heading">
-                  Unsloth Documentation
-                </p>
-                <p className="text-xs text-muted-foreground leading-relaxed">
-                  Guides on fine-tuning LLMs 2x faster with 70% less memory.
-                  Covers LoRA, QLoRA, data formatting, and deployment.
-                </p>
-                <span className="mt-1 flex items-center gap-1 text-xs font-medium text-emerald-600 group-hover/card:underline">
-                  Visit docs
-                  <HugeiconsIcon icon={ArrowRight01Icon} className="size-3" />
-                </span>
-              </a>
-            </HoverCardContent>
-          </HoverCard>
-
-          <button
-            type="button"
-            onClick={tourId ? openTour : undefined}
-            className={cn(
-              "flex h-9 items-center gap-1.5 rounded-md px-3 text-muted-foreground transition-colors hover:bg-accent hover:text-foreground",
-              !tourId && "invisible pointer-events-none",
-            )}
-            title="Tour"
-            aria-hidden={!tourId}
-            tabIndex={tourId ? 0 : -1}
-          >
-            <HugeiconsIcon icon={CursorInfo02Icon} className="size-4" />
-            <span className="text-sm font-medium">Tour</span>
-          </button>
-
-          <HoverCard openDelay={200} closeDelay={100}>
-            <HoverCardTrigger asChild={true}>
+        {/* Right: docs/tour desktop — one wrapper per control so flex gap is even (HoverCard roots can confuse flex spacing). */}
+        <div className="hidden items-center justify-self-end gap-0 md:flex">
+          <div className="flex shrink-0 items-center">
+            <AnimatedThemeToggler
+              className="flex h-9 w-9 items-center justify-center rounded-md text-muted-foreground transition-colors hover:bg-accent hover:text-foreground [&_svg]:size-4"
+              title="Toggle theme"
+              aria-label="Toggle theme"
+            />
+          </div>
+          <div className="flex shrink-0 items-center">
+            <HoverCard openDelay={200} closeDelay={100}>
+              <HoverCardTrigger asChild={true}>
+                <a
+                  href="https://unsloth.ai/docs"
+                  target="_blank"
+                  rel="noopener noreferrer"
+                  className="inline-flex h-9 items-center gap-1.5 rounded-md px-3 text-sm font-medium text-emerald-600 transition-colors hover:bg-accent hover:text-emerald-700 dark:hover:text-emerald-400"
+                >
+                  <HugeiconsIcon icon={Book03Icon} className="size-4" />
+                  Learn more
+                </a>
+              </HoverCardTrigger>
+              <HoverCardContent align="end" className="w-80 p-0">
+                <a
+                  href="https://unsloth.ai/docs"
+                  target="_blank"
+                  rel="noopener noreferrer"
+                  className="group/card flex flex-col gap-1 p-4 no-underline"
+                >
+                  <p className="text-sm font-semibold font-heading">
+                    Unsloth Documentation
+                  </p>
+                  <p className="text-xs text-muted-foreground leading-relaxed">
+                    Guides on fine-tuning LLMs 2x faster with 70% less memory.
+                    Covers LoRA, QLoRA, data formatting, and deployment.
+                  </p>
+                  <span className="mt-1 flex items-center gap-1 text-xs font-medium text-emerald-600 group-hover/card:underline">
+                    Visit docs
+                    <HugeiconsIcon icon={ArrowRight01Icon} className="size-3" />
+                  </span>
+                </a>
+              </HoverCardContent>
+            </HoverCard>
+          </div>
+          {tourId ? (
+            <div className="flex shrink-0 items-center">
               <button
                 type="button"
-                className="flex h-9 items-center gap-1.5 rounded-md px-3 text-sm font-medium text-muted-foreground transition-colors hover:bg-accent hover:text-foreground"
-                aria-label="How to update Unsloth Studio"
+                onClick={openTour}
+                className="flex h-9 items-center gap-1.5 rounded-md px-3 text-muted-foreground transition-colors hover:bg-accent hover:text-foreground"
+                title="Tour"
               >
-                <HugeiconsIcon icon={ArrowReloadHorizontalIcon} className="size-4" />
-                Update
+                <HugeiconsIcon icon={CursorInfo02Icon} className="size-4" />
+                <span className="text-sm font-medium">Tour</span>
               </button>
-            </HoverCardTrigger>
-            <HoverCardContent align="end" className="w-[22.5rem] p-0">
-              <UpdateStudioInstructions
-                className="p-4"
-                defaultShell={defaultUpdateShell}
-              />
-            </HoverCardContent>
-          </HoverCard>
-
-          <button
-            type="button"
-            onClick={() => setShutdownOpen(true)}
-            className="-mr-1.5 flex h-9 w-9 items-center justify-center rounded-md text-muted-foreground transition-colors hover:bg-accent hover:text-foreground"
-            title="Shut down Unsloth Studio server"
-            aria-label="Shut down Unsloth Studio server"
-          >
-            <HugeiconsIcon icon={Cancel01Icon} className="size-5" />
-          </button>
+            </div>
+          ) : null}
+          <div className="flex shrink-0 items-center">
+            <HoverCard openDelay={200} closeDelay={100}>
+              <HoverCardTrigger asChild={true}>
+                <button
+                  type="button"
+                  className="flex h-9 items-center gap-1.5 rounded-md px-3 text-sm font-medium text-muted-foreground transition-colors hover:bg-accent hover:text-foreground"
+                  aria-label="How to update Unsloth Studio"
+                >
+                  <HugeiconsIcon icon={ArrowReloadHorizontalIcon} className="size-4" />
+                  Update
+                </button>
+              </HoverCardTrigger>
+              <HoverCardContent align="end" className="w-[22.5rem] p-0">
+                <UpdateStudioInstructions
+                  className="p-4"
+                  defaultShell={defaultUpdateShell}
+                />
+              </HoverCardContent>
+            </HoverCard>
+          </div>
+          <div className="flex shrink-0 items-center">
+            <button
+              type="button"
+              onClick={() => setShutdownOpen(true)}
+              className="flex h-9 w-9 items-center justify-center rounded-md text-muted-foreground transition-colors hover:bg-accent hover:text-foreground"
+              title="Shut down Unsloth Studio server"
+              aria-label="Shut down Unsloth Studio server"
+            >
+              <HugeiconsIcon icon={Cancel01Icon} className="size-5" />
+            </button>
+          </div>
         </div>
 
         {/* Right: mobile */}
diff --git a/studio/frontend/src/features/chat/shared-composer.tsx b/studio/frontend/src/features/chat/shared-composer.tsx
index 59b0880add..ac01f77381 100644
--- a/studio/frontend/src/features/chat/shared-composer.tsx
+++ b/studio/frontend/src/features/chat/shared-composer.tsx
@@ -473,7 +473,7 @@ export function SharedComposer({
         onChange={(e) => setText(e.target.value)}
         onKeyDown={onKeyDown}
         placeholder="Send to both models..."
-        className="mb-1 max-h-32 min-h-14 w-full resize-none bg-transparent px-4 pt-2 pb-3 text-sm outline-none placeholder:text-muted-foreground"
+        className="mb-1 max-h-32 min-h-14 w-full resize-none bg-transparent pl-5 pr-4 pt-2 pb-3 text-sm outline-none placeholder:text-muted-foreground"
         rows={1}
       />
       <div className="relative mx-2 mb-2 flex items-center justify-between">
diff --git a/studio/frontend/src/index.css b/studio/frontend/src/index.css
index 8dc159a002..45696b2274 100644
--- a/studio/frontend/src/index.css
+++ b/studio/frontend/src/index.css
@@ -6,21 +6,11 @@
 @import "shadcn/tailwind.css";
 @import "streamdown/styles.css";
 @import "@fontsource-variable/figtree";
-@import "@fontsource-variable/space-grotesk";
 @import "@fontsource-variable/inter";
 @import "tw-shimmer";
 @plugin "@toolwind/corner-shape";
 @source "../node_modules/streamdown/dist/*.js";
 
-@font-face {
-  font-family: "Hellix";
-  src: url("/fonts/Hellix-SemiBold.woff2") format("woff2"),
-    url("/fonts/Hellix-SemiBold.woff") format("woff");
-  font-weight: 600;
-  font-style: normal;
-  font-display: swap;
-}
-
 @custom-variant dark (&:is(.dark *));
 
 :root {
@@ -67,7 +57,7 @@
   --sidebar-ring: oklch(0.6929 0.1396 166.5513);
   --destructive-foreground: oklch(1 0 0);
   --font-sans: "Inter Variable", ui-sans-serif, sans-serif, system-ui;
-  --font-heading: "Hellix", "Space Grotesk Variable", ui-sans-serif, sans-serif;
+  --font-heading: var(--font-sans);
   --font-serif: Source Serif 4, serif;
   --font-mono: JetBrains Mono, monospace;
   --shadow-color: hsl(0 0% 0%);
@@ -159,7 +149,7 @@
 
 @theme inline {
   --font-sans: "Inter Variable", ui-sans-serif, sans-serif, system-ui;
-  --font-heading: "Hellix", "Space Grotesk Variable", ui-sans-serif, sans-serif;
+  --font-heading: var(--font-sans);
   --color-sidebar-ring: var(--sidebar-ring);
   --color-sidebar-border: var(--sidebar-border);
   --color-sidebar-accent-foreground: var(--sidebar-accent-foreground);
@@ -343,10 +333,68 @@
   [data-streamdown="code-block"] {
     gap: 0;
     padding: 0.5rem;
+    /* Wide lines must scroll inside the thread column, not widen past the composer (flex min-width:auto). */
+    max-width: 100%;
+    min-width: 0;
+    overflow-x: auto;
   }
   [data-streamdown="code-block-header"] {
     padding-left: 0.75rem;
   }
+
+  /* Chat thread: code slightly smaller by default; step up when the thread column is wide. */
+  .aui-thread-root [data-streamdown="code-block"] {
+    font-size: 0.8125rem;
+    line-height: 1.55;
+  }
+
+  .aui-thread-root [data-streamdown="code-block-header"] {
+    font-size: 0.6875rem;
+  }
+
+  @container (min-width: 36rem) {
+    .aui-thread-root [data-streamdown="code-block"] {
+      font-size: 0.875rem;
+    }
+
+    .aui-thread-root [data-streamdown="code-block-header"] {
+      font-size: 0.75rem;
+    }
+  }
+
+  /* Chat: use the app sans stack for UI + prose. */
+  .aui-thread-root {
+    font-family: var(--font-sans);
+  }
+
+  .aui-thread-root .font-medium,
+  .aui-thread-root .font-semibold,
+  .aui-thread-root .font-bold,
+  .aui-thread-root .font-heading {
+    font-family: var(--font-sans);
+  }
+
+  .aui-thread-root :where(h1, h2, h3, h4, h5, h6) {
+    font-family: var(--font-sans);
+  }
+
+  /* Keep monospace for code fences and inline code (not KaTeX). */
+  .aui-thread-root [data-streamdown="code-block"] pre,
+  .aui-thread-root [data-streamdown="code-block"] code {
+    font-family: var(--font-mono), ui-monospace, monospace;
+  }
+
+  .aui-thread-root :where(p, li, td, th, blockquote, h1, h2, h3, h4, h5, h6) code {
+    font-family: var(--font-mono), ui-monospace, monospace;
+  }
+
+  /* Align fenced code blocks with the main chat column even when nested in lists. */
+  .aui-thread-root [data-streamdown="list-item"] > [data-streamdown="code-block"],
+  .aui-thread-root [data-streamdown="list-item"] [data-streamdown="code-block"] {
+    margin-left: -1.25rem;
+    width: calc(100% + 1.25rem);
+    max-width: calc(100% + 1.25rem);
+  }
 }
 
 /* Minimal scrollbar — thumb only, no track */

From ba8081fc96f8253bbeae846961e1c5812491e25b Mon Sep 17 00:00:00 2001
From: AdamPlatin123 <154926636+AdamPlatin123@users.noreply.github.com>
Date: Thu, 2 Apr 2026 11:24:48 +0800
Subject: [PATCH 002/216] fix(chat): correct loading text for cached models
 during inference (#4764)

Distinguish between actual network downloads and GPU memory loading for cached LoRA adapters in Studio chat.

- Add isCachedLora detection for local LoRA adapter paths using comprehensive cross-platform regex (Unix, Windows, UNC, WSL, tilde)
- Thread isCachedLora through loadInfo to chat-page inline status for proper 3-way distinction (cached / local LoRA / downloading)
- Skip download progress polling for cached LoRA models (no useless /download-progress API calls)
- Fix initial toast state to use isCachedLoad consistently instead of only checking isDownloaded
- Fix cancelLoading toast to not mention background downloads for cached/local loads
- Keep download-specific text ("Downloading model..." / "Download complete") inside the download-only polling block
---
 .../frontend/src/features/chat/chat-page.tsx  |  6 +++--
 .../chat/hooks/use-chat-model-runtime.ts      | 23 +++++++++++++------
 2 files changed, 20 insertions(+), 9 deletions(-)

diff --git a/studio/frontend/src/features/chat/chat-page.tsx b/studio/frontend/src/features/chat/chat-page.tsx
index 8d0a9649b4..08450c7ec7 100644
--- a/studio/frontend/src/features/chat/chat-page.tsx
+++ b/studio/frontend/src/features/chat/chat-page.tsx
@@ -866,14 +866,16 @@ export function ChatPage(): ReactElement {
                   label={
                     loadProgress?.phase === "starting"
                       ? "Starting model…"
-                      : loadingModel.isDownloaded
+                      : loadingModel.isDownloaded || loadingModel.isCachedLora
                         ? "Loading model…"
                         : "Downloading model…"
                   }
                   title={
                     loadingModel.isDownloaded
                       ? `Loading ${loadingModel.displayName} from cache.`
-                      : `Loading ${loadingModel.displayName}. This may include downloading.`
+                      : loadingModel.isCachedLora
+                        ? `Loading ${loadingModel.displayName} into memory.`
+                        : `Loading ${loadingModel.displayName}. This may include downloading.`
                   }
                   progressPercent={loadProgress?.percent}
                   progressLabel={loadProgress?.label}
diff --git a/studio/frontend/src/features/chat/hooks/use-chat-model-runtime.ts b/studio/frontend/src/features/chat/hooks/use-chat-model-runtime.ts
index fc0e392596..8fcd22d0f8 100644
--- a/studio/frontend/src/features/chat/hooks/use-chat-model-runtime.ts
+++ b/studio/frontend/src/features/chat/hooks/use-chat-model-runtime.ts
@@ -168,6 +168,7 @@ export function useChatModelRuntime() {
     id: string;
     displayName: string;
     isDownloaded?: boolean;
+    isCachedLora?: boolean;
   } | null>(null);
   const [loadToastDismissed, setLoadToastDismissed] = useState(false);
   const [loadProgress, setLoadProgress] = useState<{
@@ -294,8 +295,11 @@ export function useChatModelRuntime() {
     setLoadToastDismissedState(false);
     clearCheckpoint();
     if (tid != null) toast.dismiss(tid);
+    const isCachedOrLocal = model.isDownloaded || model.isCachedLora;
     toast.info("Stopped loading model", {
-      description: "The current download may still finish in the background.",
+      description: isCachedOrLocal
+        ? undefined
+        : "The current download may still finish in the background.",
     });
     // Fire-and-forget: tell backend to stop, don't block UI
     unloadModel({ model_path: model.id }).catch(() => {});
@@ -339,20 +343,24 @@ export function useChatModelRuntime() {
         : undefined;
       const previousIsLora =
         previousModel?.isLora ?? (previousLora ? true : false);
+      // Covers Unix absolute (/), relative (./  ../), tilde (~/), Windows drive (C:\), UNC (\\server)
+      const isLocal = /^(\/|\.{1,2}[\\\/]|~[\\\/]|[A-Za-z]:[\\\/]|\\\\)/.test(modelId);
+      const isCachedLora = isLora && isLocal;
       const loadingDescription = [
         currentCheckpoint ? "Switching models." : null,
         extraLoadingDescription ?? null,
         isDownloaded ? "Loading cached model into memory." : null,
+        !isDownloaded && isCachedLora ? "Loading trained model into memory." : null,
       ]
         .filter(Boolean)
         .join(" ");
       setModelsError(null);
       setLoadToastDismissedState(false);
-      const loadInfo = { id: modelId, displayName, isDownloaded };
+      const loadInfo = { id: modelId, displayName, isDownloaded, isCachedLora };
       setLoadingModel(loadInfo);
       useChatRuntimeStore.getState().setModelLoading(true);
       setLoadProgress(
-        isDownloaded
+        isDownloaded || isCachedLora
           ? { percent: null, label: null, phase: "starting" }
           : { percent: 0, label: "Preparing download", phase: "downloading" },
       );
@@ -485,15 +493,16 @@ export function useChatModelRuntime() {
           }
         }
 
-        const toastTitle = isDownloaded ? "Starting model…" : "Downloading model…";
+        const isCachedLoad = isDownloaded || isCachedLora;
+        const toastTitle = isCachedLoad ? "Starting model…" : "Downloading model…";
         const toastId = toast(
           null,
           {
             description: renderLoadDescription(
               toastTitle,
               loadingDescription,
-              isDownloaded ? null : 0,
-              isDownloaded ? null : "Preparing download",
+              isCachedLoad ? null : 0,
+              isCachedLoad ? null : "Preparing download",
               cancelLoading,
             ),
             duration: Infinity,
@@ -511,7 +520,7 @@ export function useChatModelRuntime() {
 
         // Poll download progress for non-cached models (GGUF and non-GGUF)
         let progressInterval: ReturnType<typeof setInterval> | null = null;
-        if (!isDownloaded) {
+        if (!isDownloaded && !isCachedLora) {
           const expectedBytes =
             typeof selection !== "string" ? selection.expectedBytes ?? 0 : 0;
           let hasShownProgress = false;

From 752cef329906d860188cda3fdaf7becaa2413940 Mon Sep 17 00:00:00 2001
From: mateeaaaaaaa <voiderxa@proton.me>
Date: Thu, 2 Apr 2026 10:10:43 +0300
Subject: [PATCH 003/216] fix(security): shell injection in GGML export
 conversion (#4768)

* Fix shell injection in GGML conversion paths

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

* Remove test file from security fix PR

Move test_save_shell_injection.py to a separate PR to keep this PR focused on the security fix itself.

---------

Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com>
Co-authored-by: Daniel Han <danielhanchen@gmail.com>
---
 unsloth/save.py | 26 ++++++++++++++++----------
 1 file changed, 16 insertions(+), 10 deletions(-)

diff --git a/unsloth/save.py b/unsloth/save.py
index 178178980d..3c318fab02 100644
--- a/unsloth/save.py
+++ b/unsloth/save.py
@@ -2533,12 +2533,15 @@ def unsloth_convert_lora_to_ggml_and_push_to_hub(
     )
     print(f"The output file will be {output_file}")
 
-    command = f"python3 llama.cpp/convert-lora-to-ggml.py {lora_directory_push} {output_file} llama"
-
     try:
         with subprocess.Popen(
-            command,
-            shell = True,
+            [
+                sys.executable,
+                "llama.cpp/convert-lora-to-ggml.py",
+                lora_directory_push,
+                output_file,
+                "llama",
+            ],
             stdout = subprocess.PIPE,
             stderr = subprocess.PIPE,
             bufsize = 1,
@@ -2550,7 +2553,7 @@ def unsloth_convert_lora_to_ggml_and_push_to_hub(
                 print(line, end = "", flush = True)
             sp.wait()
             if sp.returncode != 0:
-                raise subprocess.CalledProcessError(sp.returncode, command)
+                raise subprocess.CalledProcessError(sp.returncode, sp.args)
     except subprocess.CalledProcessError as e:
         print(f"Error: Conversion failed with return code {e.returncode}")
         return
@@ -2612,12 +2615,15 @@ def unsloth_convert_lora_to_ggml_and_save_locally(
     )
     print(f"The output file will be {output_file}")
 
-    command = f"python3 llama.cpp/convert-lora-to-ggml.py {save_directory} {output_file} llama"
-
     try:
         with subprocess.Popen(
-            command,
-            shell = True,
+            [
+                sys.executable,
+                "llama.cpp/convert-lora-to-ggml.py",
+                save_directory,
+                output_file,
+                "llama",
+            ],
             stdout = subprocess.PIPE,
             stderr = subprocess.PIPE,
             bufsize = 1,
@@ -2629,7 +2635,7 @@ def unsloth_convert_lora_to_ggml_and_save_locally(
                 print(line, end = "", flush = True)
             sp.wait()
             if sp.returncode != 0:
-                raise subprocess.CalledProcessError(sp.returncode, command)
+                raise subprocess.CalledProcessError(sp.returncode, sp.args)
     except subprocess.CalledProcessError as e:
         print(f"Error: Conversion failed with return code {e.returncode}")
         return

From dc0729aadfb6aa891468d1432394db59a183eca1 Mon Sep 17 00:00:00 2001
From: Daniel Han <danielhanchen@gmail.com>
Date: Thu, 2 Apr 2026 00:10:47 -0700
Subject: [PATCH 004/216] Add regression test for shell injection fix in GGML
 conversion (#4773)

AST-based test ensures subprocess.Popen calls in GGML conversion functions
use argv lists instead of shell=True. Companion to PR #4768.
---
 tests/saving/test_save_shell_injection.py | 74 +++++++++++++++++++++++
 1 file changed, 74 insertions(+)
 create mode 100644 tests/saving/test_save_shell_injection.py

diff --git a/tests/saving/test_save_shell_injection.py b/tests/saving/test_save_shell_injection.py
new file mode 100644
index 0000000000..c6c2c8fe15
--- /dev/null
+++ b/tests/saving/test_save_shell_injection.py
@@ -0,0 +1,74 @@
+from __future__ import annotations
+
+import ast
+from pathlib import Path
+
+
+SAVE_PY = Path(__file__).resolve().parents[2] / "unsloth" / "save.py"
+
+
+def _function_calls(source: str, function_name: str) -> list[ast.Call]:
+    tree = ast.parse(source, filename = str(SAVE_PY))
+    for node in tree.body:
+        if isinstance(node, ast.FunctionDef) and node.name == function_name:
+            return [child for child in ast.walk(node) if isinstance(child, ast.Call)]
+    raise AssertionError(f"Function {function_name} not found in save.py")
+
+
+def _assert_safe_ggml_calls(calls: list[ast.Call]) -> None:
+    popen_calls = []
+    for call in calls:
+        if isinstance(call.func, ast.Attribute) and call.func.attr == "Popen":
+            if (
+                isinstance(call.func.value, ast.Name)
+                and call.func.value.id == "subprocess"
+            ):
+                popen_calls.append(call)
+
+    assert popen_calls, "Expected at least one subprocess.Popen call"
+
+    ggml_calls = []
+    for call in popen_calls:
+        if not call.args:
+            continue
+        argv = call.args[0]
+        if isinstance(argv, ast.List) and len(argv.elts) >= 2:
+            second_arg = argv.elts[1]
+            if (
+                isinstance(second_arg, ast.Constant)
+                and second_arg.value == "llama.cpp/convert-lora-to-ggml.py"
+            ):
+                ggml_calls.append(call)
+
+    assert ggml_calls, "Expected the GGML conversion subprocess call"
+
+    for call in ggml_calls:
+        shell_kwargs = [
+            keyword
+            for keyword in call.keywords
+            if keyword.arg == "shell"
+            and isinstance(keyword.value, ast.Constant)
+            and keyword.value.value is True
+        ]
+        assert not shell_kwargs, "subprocess.Popen must not use shell=True"
+
+        assert call.args, "subprocess.Popen must receive argv as a positional argument"
+        argv = call.args[0]
+        assert isinstance(
+            argv, ast.List
+        ), "subprocess.Popen must be called with an argv list"
+        assert len(argv.elts) == 5, "GGML conversion argv should have five elements"
+
+        second_arg = argv.elts[1]
+        assert isinstance(second_arg, ast.Constant)
+        assert second_arg.value == "llama.cpp/convert-lora-to-ggml.py"
+
+
+def test_ggml_conversion_paths_do_not_use_shell() -> None:
+    source = SAVE_PY.read_text(encoding = "utf-8")
+    for function_name in (
+        "unsloth_convert_lora_to_ggml_and_push_to_hub",
+        "unsloth_convert_lora_to_ggml_and_save_locally",
+    ):
+        calls = _function_calls(source, function_name)
+        _assert_safe_ggml_calls(calls)

From e4d149923097f25d804177d53e2f2964e7bb1ca7 Mon Sep 17 00:00:00 2001
From: Daniel Han <danielhanchen@gmail.com>
Date: Thu, 2 Apr 2026 02:11:07 -0700
Subject: [PATCH 005/216] fix(studio): prevent small models from stalling on
 tool-calling tasks (#4769)

* fix(studio): prevent small models from stalling on tool-calling tasks

Small GGUF models (< 9B params) in "Think, Search, Code" mode would
often describe what they planned to do ("Let me create this dashboard")
and then stop generating without ever calling a tool.

Three changes:

1. Simplify web_tips for small models: remove the "fetch its full content
   by calling web_search with the url parameter" guidance for models < 9B.
   This multi-step instruction causes small models to plan elaborate
   search-then-fetch-then-code sequences they cannot reliably execute.

2. Add "always call tools directly" imperative to the system prompt nudge
   so models act immediately instead of narrating their intentions.

3. Add plan-without-action re-prompt in the agentic loop: when the model
   emits planning text (matching patterns like "let me", "I'll", etc.)
   without calling any tool, inject a nudge asking it to call the tool
   and continue the loop. Capped at 2 re-prompts per request.

Benchmarked with Qwen3.5-4B-GGUF (N=5 trials per variant):
- Baseline: 40% of requests had any tool call
- Combined fix: 100% of requests had at least one tool call

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

---------

Co-authored-by: Daniel Han <danielhanchen@users.noreply.github.com>
Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com>
---
 studio/backend/core/inference/llama_cpp.py  | 172 ++++++++++++++------
 studio/backend/routes/inference.py          |  37 ++++-
 studio/backend/utils/models/__init__.py     |   2 +
 studio/backend/utils/models/model_config.py |  31 ++++
 4 files changed, 189 insertions(+), 53 deletions(-)

diff --git a/studio/backend/core/inference/llama_cpp.py b/studio/backend/core/inference/llama_cpp.py
index 5f974b0af3..1b7fe548e4 100644
--- a/studio/backend/core/inference/llama_cpp.py
+++ b/studio/backend/core/inference/llama_cpp.py
@@ -27,6 +27,52 @@ import httpx
 
 logger = get_logger(__name__)
 
+# ── Pre-compiled patterns for plan-without-action re-prompt ──
+# Forward-looking intent signals that indicate the model is
+# describing what it *will* do rather than giving a final answer.
+_INTENT_SIGNAL = re.compile(
+    r"(?i)("
+    # Direct intent: "I'll ...", "I will ...", "Let me ...", "I am going to ..."
+    # Handles both straight and curly apostrophes.
+    # Excludes "I can", "I should", "I want to", "let's" which
+    # appear frequently in direct answers / explanations.
+    r"\b(i['\u2019](ll|m going to|m gonna)|i am (going to|gonna)|i will|i shall|let me|allow me)\b"
+    r"|"
+    # Step/plan framing: "First ...", "Step 1:", "Here's my plan"
+    r"\b(?:first\b|step \d+:?|here['\u2019]?s (?:my |the |a )?(?:plan|approach))"
+    r"|"
+    # "Now I" / "Next I" patterns
+    r"\b(?:now i|next i)\b"
+    r")"
+)
+_MAX_REPROMPTS = 1
+_REPROMPT_MAX_CHARS = 500
+
+# ── Pre-compiled patterns for GGUF shard detection ───────────
+_SHARD_FULL_RE = re.compile(r"^(.*)-(\d{5})-of-(\d{5})\.gguf$")
+_SHARD_RE = re.compile(r"^(.*)-\d{5}-of-\d{5}\.gguf$")
+
+# Model size extraction (shared with routes/inference.py)
+from utils.models import extract_model_size_b as _extract_model_size_b
+
+# ── Pre-compiled patterns for tool XML stripping ─────────────
+_TOOL_CLOSED_PATS = [
+    re.compile(r"<tool_call>.*?</tool_call>", re.DOTALL),
+    re.compile(r"<function=\w+>.*?</function>", re.DOTALL),
+]
+_TOOL_ALL_PATS = _TOOL_CLOSED_PATS + [
+    re.compile(r"<tool_call>.*$", re.DOTALL),
+    re.compile(r"<function=\w+>.*$", re.DOTALL),
+]
+
+# ── Pre-compiled patterns for tool-call XML parsing ──────────
+_TC_JSON_START_RE = re.compile(r"<tool_call>\s*\{")
+_TC_FUNC_START_RE = re.compile(r"<function=(\w+)>\s*")
+_TC_END_TAG_RE = re.compile(r"</tool_call>")
+_TC_FUNC_CLOSE_RE = re.compile(r"\s*</function>\s*$")
+_TC_PARAM_START_RE = re.compile(r"<parameter=(\w+)>\s*")
+_TC_PARAM_CLOSE_RE = re.compile(r"\s*</parameter>\s*$")
+
 
 class LlamaCppBackend:
     """
@@ -242,14 +288,11 @@ class LlamaCppBackend:
     @staticmethod
     def _get_gguf_size_bytes(model_path: str) -> int:
         """Get total GGUF size in bytes, including split shards."""
-        import re
-
         main = Path(model_path)
         total = main.stat().st_size
 
         # Check for split shards (e.g., model-00001-of-00003.gguf)
-        shard_pat = re.compile(r"^(.*)-(\d{5})-of-(\d{5})\.gguf$")
-        m = shard_pat.match(main.name)
+        m = _SHARD_FULL_RE.match(main.name)
         if m:
             prefix, _, num_total = m.group(1), m.group(2), m.group(3)
             sibling_pat = re.compile(
@@ -539,8 +582,6 @@ class LlamaCppBackend:
 
         Returns (first_shard_filename, total_size_bytes) or None if nothing fits.
         """
-        import re
-
         try:
             from huggingface_hub import get_paths_info, list_repo_files
 
@@ -556,10 +597,9 @@ class LlamaCppBackend:
             size_map = {p.path: (p.size or 0) for p in path_infos}
 
             # Group files by variant: shards share a prefix before -NNNNN-of-NNNNN
-            shard_pat = re.compile(r"^(.*)-\d{5}-of-\d{5}\.gguf$")
             variants: dict[str, list[str]] = {}
             for f in gguf_files:
-                m = shard_pat.match(f)
+                m = _SHARD_RE.match(f)
                 key = m.group(1) if m else f
                 variants.setdefault(key, []).append(f)
 
@@ -810,7 +850,6 @@ class LlamaCppBackend:
         gguf_extra_shards: list[str] = []
         if hf_variant:
             try:
-                import re
                 from huggingface_hub import list_repo_files
 
                 files = list_repo_files(hf_repo, token = hf_token)
@@ -825,11 +864,10 @@ class LlamaCppBackend:
                 )
                 if gguf_files:
                     gguf_filename = gguf_files[0]
-                    shard_pat = re.compile(r"^(.*)-\d{5}-of-(\d{5})\.gguf$")
-                    m = shard_pat.match(gguf_filename)
+                    m = _SHARD_FULL_RE.match(gguf_filename)
                     if m:
                         prefix = m.group(1)
-                        total = m.group(2)
+                        total = m.group(3)
                         sibling_pat = re.compile(
                             r"^"
                             + re.escape(prefix)
@@ -886,10 +924,7 @@ class LlamaCppBackend:
                             f"falling back to {fallback_file} ({fallback_size / (1024**3):.1f} GB)"
                         )
                         gguf_filename = fallback_file
-                        import re as _re
-
-                        _shard_pat = _re.compile(r"^(.*)-\d{5}-of-\d{5}\.gguf$")
-                        _m = _shard_pat.match(gguf_filename)
+                        _m = _SHARD_RE.match(gguf_filename)
                         _prefix = _m.group(1) if _m else None
                         if _prefix:
                             gguf_extra_shards = sorted(
@@ -1292,17 +1327,12 @@ class LlamaCppBackend:
             # Qwen3.5 models below 9B (0.8B, 2B, 4B) disable thinking by default.
             # Only 9B and larger enable thinking.
             if self._supports_reasoning:
-                import re
-
                 thinking_default = True
                 mid = (model_identifier or "").lower()
                 if "qwen3.5" in mid:
-                    # Extract size like "0.8b", "4b", "35b" etc.
-                    size_match = re.search(r"(\d+\.?\d*)\s*b", mid)
-                    if size_match:
-                        size_val = float(size_match.group(1))
-                        if size_val < 9:
-                            thinking_default = False
+                    size_val = _extract_model_size_b(mid)
+                    if size_val is not None and size_val < 9:
+                        thinking_default = False
                 self._reasoning_default = thinking_default
                 cmd.extend(
                     [
@@ -1775,13 +1805,11 @@ class LlamaCppBackend:
         Closing tags (</tool_call>, </function>, </parameter>) are all optional
         since models frequently omit them.
         """
-        import re
-
         tool_calls = []
 
         # Pattern 1: JSON inside <tool_call> tags.
         # Use balanced-brace extraction that skips braces inside JSON strings.
-        for m in re.finditer(r"<tool_call>\s*\{", content):
+        for m in _TC_JSON_START_RE.finditer(content):
             brace_start = m.end() - 1  # position of the opening {
             depth, i = 0, brace_start
             in_string = False
@@ -1831,7 +1859,7 @@ class LlamaCppBackend:
             # boundaries.  We avoid using </function> as a boundary because
             # code parameter values can contain that literal string.
             # After extracting, we trim a trailing </function> if present.
-            func_starts = list(re.finditer(r"<function=(\w+)>\s*", content))
+            func_starts = list(_TC_FUNC_START_RE.finditer(content))
             for idx, fm in enumerate(func_starts):
                 func_name = fm.group(1)
                 body_start = fm.end()
@@ -1841,7 +1869,7 @@ class LlamaCppBackend:
                     if idx + 1 < len(func_starts)
                     else len(content)
                 )
-                end_tag = re.search(r"</tool_call>", content[body_start:])
+                end_tag = _TC_END_TAG_RE.search(content[body_start:])
                 if end_tag:
                     body_end = body_start + end_tag.start()
                 else:
@@ -1849,20 +1877,20 @@ class LlamaCppBackend:
                 body_end = min(body_end, next_func)
                 body = content[body_start:body_end]
                 # Trim trailing </function> if present (it's the real closing tag)
-                body = re.sub(r"\s*</function>\s*$", "", body)
+                body = _TC_FUNC_CLOSE_RE.sub("", body)
 
                 # Step 2: Extract parameters from body.
                 # For single-parameter functions (the common case: code, command,
                 # query), use body end as the only boundary to avoid false matches
                 # on </parameter> inside code strings.
                 arguments = {}
-                param_starts = list(re.finditer(r"<parameter=(\w+)>\s*", body))
+                param_starts = list(_TC_PARAM_START_RE.finditer(body))
                 if len(param_starts) == 1:
                     # Single parameter: value is everything from after the tag
                     # to end of body, trimming any trailing </parameter>.
                     pm = param_starts[0]
                     val = body[pm.end() :]
-                    val = re.sub(r"\s*</parameter>\s*$", "", val)
+                    val = _TC_PARAM_CLOSE_RE.sub("", val)
                     arguments[pm.group(1)] = val.strip()
                 else:
                     for pidx, pm in enumerate(param_starts):
@@ -1876,7 +1904,7 @@ class LlamaCppBackend:
                         )
                         val = body[val_start:next_param]
                         # Trim trailing </parameter> if present
-                        val = re.sub(r"\s*</parameter>\s*$", "", val)
+                        val = _TC_PARAM_CLOSE_RE.sub("", val)
                         arguments[param_name] = val.strip()
 
                 tc = {
@@ -2249,22 +2277,10 @@ class LlamaCppBackend:
         _accumulated_predicted_ms = 0.0
         _accumulated_predicted_n = 0
 
-        # ── Shared patterns for stripping tool XML from streamed content ──
-        import re as _re_tool
-
-        _TOOL_CLOSED_PATTERNS = [
-            _re_tool.compile(r"<tool_call>.*?</tool_call>", _re_tool.DOTALL),
-            _re_tool.compile(r"<function=\w+>.*?</function>", _re_tool.DOTALL),
-        ]
-        _TOOL_ALL_PATTERNS = _TOOL_CLOSED_PATTERNS + [
-            _re_tool.compile(r"<tool_call>.*$", _re_tool.DOTALL),
-            _re_tool.compile(r"<function=\w+>.*$", _re_tool.DOTALL),
-        ]
-
         def _strip_tool_markup(text: str, *, final: bool = False) -> str:
             if not auto_heal_tool_calls:
                 return text
-            patterns = _TOOL_ALL_PATTERNS if final else _TOOL_CLOSED_PATTERNS
+            patterns = _TOOL_ALL_PATS if final else _TOOL_CLOSED_PATS
             for pat in patterns:
                 text = pat.sub("", text)
             return text.strip() if final else text
@@ -2284,7 +2300,19 @@ class LlamaCppBackend:
         # identical call succeeded).
         _tool_call_history: list[tuple[str, bool]] = []  # (key, failed)
 
-        for iteration in range(max_tool_iterations):
+        # ── Re-prompt on plan-without-action ─────────────────
+        # When the model describes what it intends to do (forward-looking
+        # language) without actually calling a tool, re-prompt once.
+        # Only triggers on responses that signal intent/planning -- a
+        # direct answer like "4" or "Hello!" will not match.
+        # Pattern is compiled once at module level (_INTENT_SIGNAL).
+        _reprompt_count = 0
+
+        # Reserve extra iterations for re-prompts so they don't
+        # consume the caller's tool-call budget.  Only add the
+        # extra slot when tool iterations are actually allowed.
+        _extra = _MAX_REPROMPTS if max_tool_iterations > 0 else 0
+        for iteration in range(max_tool_iterations + _extra):
             if cancel_event is not None and cancel_event.is_set():
                 return
 
@@ -2595,6 +2623,56 @@ class LlamaCppBackend:
                             content_accum,
                         )
                     if not _safety_tc:
+                        # ── Re-prompt on plan-without-action ──
+                        # If the model described what it intends to do
+                        # (forward-looking language) without calling any
+                        # tool, nudge it to act.  Only fires once per
+                        # request and only on short responses that
+                        # contain intent signals -- a direct answer
+                        # like "4" or "Hello!" won't trigger this.
+                        # Use content if available, otherwise fall back
+                        # to reasoning text (reasoning-only stalls).
+                        _stripped = content_accum.strip()
+                        if not _stripped:
+                            _stripped = reasoning_accum.strip()
+                        if (
+                            tools
+                            and _reprompt_count < _MAX_REPROMPTS
+                            and 0 < len(_stripped) < _REPROMPT_MAX_CHARS
+                            and _INTENT_SIGNAL.search(_stripped)
+                        ):
+                            _reprompt_count += 1
+                            logger.info(
+                                f"Re-prompt {_reprompt_count}/{_MAX_REPROMPTS}: "
+                                f"model responded without calling tools "
+                                f"({len(_stripped)} chars)"
+                            )
+                            conversation.append(
+                                {
+                                    "role": "assistant",
+                                    "content": _stripped,
+                                }
+                            )
+                            conversation.append(
+                                {
+                                    "role": "user",
+                                    "content": (
+                                        "Please use the available tools to complete "
+                                        "the task instead of describing what to do."
+                                    ),
+                                }
+                            )
+                            # Accumulate tokens and timing from this iteration
+                            _fu_r = _iter_usage or {}
+                            _accumulated_completion_tokens += _fu_r.get(
+                                "completion_tokens", 0
+                            )
+                            _it_r = _iter_timings or {}
+                            _accumulated_predicted_ms += _it_r.get("predicted_ms", 0)
+                            _accumulated_predicted_n += _it_r.get("predicted_n", 0)
+                            yield {"type": "status", "text": ""}
+                            continue
+
                         # Content was already streamed.  Yield metadata.
                         yield {"type": "status", "text": ""}
                         _fu = _iter_usage or {}
diff --git a/studio/backend/routes/inference.py b/studio/backend/routes/inference.py
index 99a52dfe2d..0c1d7e37f9 100644
--- a/studio/backend/routes/inference.py
+++ b/studio/backend/routes/inference.py
@@ -21,6 +21,9 @@ import threading
 
 import re as _re
 
+# Model size extraction (shared with core/inference/llama_cpp.py)
+from utils.models import extract_model_size_b as _extract_model_size_b
+
 
 def _friendly_error(exc: Exception) -> str:
     """Extract a user-friendly message from known llama-server errors."""
@@ -90,6 +93,12 @@ from datetime import date as _date
 
 router = APIRouter()
 
+# Appended to tool-use nudge to discourage plan-without-action
+_TOOL_ACTION_NUDGE = (
+    " Always call tools directly."
+    " Never describe what you plan to do -- just call the tool immediately."
+)
+
 # Regex for stripping leaked tool-call XML from assistant messages/stream
 _TOOL_XML_RE = _re.compile(
     r"<tool_call>.*?</tool_call>|<function=\w+>.*?</function>",
@@ -1095,12 +1104,20 @@ async def openai_chat_completions(
 
             _date_line = f"The current date is {_date.today().isoformat()}."
 
-            _web_tips = (
-                "When you search and find a relevant URL in the results, "
-                "fetch its full content by calling web_search with the url parameter. "
-                "Do not repeat the same search query. If a search returns "
-                "no useful results, try rephrasing or fetching a result URL directly."
-            )
+            # Small models (<9B) struggle with multi-step search plans,
+            # so simplify the web tips to avoid plan-then-stall behavior.
+            _model_size_b = _extract_model_size_b(model_name)
+            _is_small_model = _model_size_b is not None and _model_size_b < 9
+
+            if _is_small_model:
+                _web_tips = "Do not repeat the same search query."
+            else:
+                _web_tips = (
+                    "When you search and find a relevant URL in the results, "
+                    "fetch its full content by calling web_search with the url parameter. "
+                    "Do not repeat the same search query. If a search returns "
+                    "no useful results, try rephrasing or fetching a result URL directly."
+                )
             _code_tips = (
                 "Use code execution for math, calculations, data processing, "
                 "or to parse and analyze information from tool results."
@@ -1132,6 +1149,7 @@ async def openai_chat_completions(
                 _nudge = ""
 
             if _nudge:
+                _nudge += _TOOL_ACTION_NUDGE
                 # Append nudge to system prompt (preserve user's prompt)
                 if system_prompt:
                     system_prompt = system_prompt.rstrip() + "\n\n" + _nudge
@@ -1208,7 +1226,14 @@ async def openai_chat_completions(
                             break
 
                         if event["type"] == "status":
+                            # Empty status marks an iteration boundary
+                            # in the GGUF tool loop (e.g. after a
+                            # re-prompt).  Reset the cumulative cursor
+                            # so the next assistant turn streams cleanly.
+                            if not event["text"]:
+                                prev_text = ""
                             # Emit tool status as a custom SSE event
+                            # (including empty ones to clear UI badges)
                             status_data = json.dumps(
                                 {
                                     "type": "tool_status",
diff --git a/studio/backend/utils/models/__init__.py b/studio/backend/utils/models/__init__.py
index 82236d8013..a81682d8b7 100644
--- a/studio/backend/utils/models/__init__.py
+++ b/studio/backend/utils/models/__init__.py
@@ -19,6 +19,7 @@ from .model_config import (
     get_base_model_from_lora,
     load_model_config,
     list_gguf_variants,
+    extract_model_size_b,
     MODEL_NAME_MAPPING,
     UI_STATUS_INDICATORS,
 )
@@ -38,6 +39,7 @@ __all__ = [
     "get_base_model_from_lora",
     "load_model_config",
     "list_gguf_variants",
+    "extract_model_size_b",
     "MODEL_NAME_MAPPING",
     "UI_STATUS_INDICATORS",
     "scan_checkpoints",
diff --git a/studio/backend/utils/models/model_config.py b/studio/backend/utils/models/model_config.py
index f7d9b33542..df1058abf6 100644
--- a/studio/backend/utils/models/model_config.py
+++ b/studio/backend/utils/models/model_config.py
@@ -31,6 +31,37 @@ import yaml
 
 logger = get_logger(__name__)
 
+# ── Model size extraction ────────────────────────────────────
+import re as _re
+
+_MODEL_SIZE_RE = _re.compile(
+    r"(?:^|[-_/])(\d+\.?\d*)\s*([bm])(?:$|[-_/])", _re.IGNORECASE
+)
+# MoE active-parameter pattern: matches "A3B", "A3.5B", etc.
+_ACTIVE_SIZE_RE = _re.compile(
+    r"(?:^|[-_/])a(\d+\.?\d*)\s*([bm])(?:$|[-_/])", _re.IGNORECASE
+)
+
+
+def extract_model_size_b(model_id: str) -> float | None:
+    """Extract model size in billions from a model identifier.
+
+    Prefers MoE active-parameter notation (e.g. ``A3B`` in
+    ``Qwen3.5-35B-A3B``) over the total parameter count.
+    Handles both ``B`` (billions) and ``M`` (millions) suffixes.
+    """
+    mid = (model_id or "").lower()
+    active = _ACTIVE_SIZE_RE.search(mid)
+    if active:
+        val = float(active.group(1))
+        return val / 1000.0 if active.group(2).lower() == "m" else val
+    size = _MODEL_SIZE_RE.search(mid)
+    if not size:
+        return None
+    val = float(size.group(1))
+    return val / 1000.0 if size.group(2).lower() == "m" else val
+
+
 # Model name mapping: maps all equivalent model names to their canonical YAML config file
 # Format: "canonical_model_name.yaml": [list of all equivalent model names]
 # Based on the model mapper provided - canonical filename is based on the first model name in the mapper

From 31d6aeb1975c418267547e0baa4ed698564d9940 Mon Sep 17 00:00:00 2001
From: Michael Han <107991372+shimmyshimmer@users.noreply.github.com>
Date: Thu, 2 Apr 2026 02:58:21 -0700
Subject: [PATCH 006/216] Unsloth new logo

---
 images/unsloth logo only.png | Bin 57176 -> 60610 bytes
 1 file changed, 0 insertions(+), 0 deletions(-)

diff --git a/images/unsloth logo only.png b/images/unsloth logo only.png
index 92340eef059b9b687cf92aea9fe109106906eff3..adaafee48dc80f38582cad3df0b56b0b3e7d7302 100644
GIT binary patch
literal 60610
zcmYgY2RPO3`&ZdBdykNjU6RNV;n+K}SF%^O;~cVOg`{jw_KIwdJrZT_l@W(yI~?2p
z>;2XH{;!M6b-6m8=leYOeSgM%e~Z*q2a^yo5Mp6rktiw3X=7pCI>N%bA&ZX-{KjZj
zfd%+M;Hqfgj)f)2as7S6Cs*1F3v1IuNlr%BCv7tw&)?|xSa)WzfUi*nvyBD?8>dC6
zRC|Us6(QAkeru{@Lp?i{KEbN0r`i45dVLwzNeCPKJDe|f?7%fmoDOk2-lpry`jwo0
zHQh=6m#$VZ7*i7^{bA_5`%m}JuHwXD^!x%|GrCyISfWO%t3vIIyW+JI?)L%>)Ac51
zG>xu2A>KXfi>^MzYzFrCisuPJWn<L6cgPbyzfp(Vk%Y-|baitcZxq|Jy4U$(4)M+<
z2i3TO_Z}n;)iv}Dxr@c|Zqvcrga;f&1=LuUJupA_PA<Oss@laC3LDUPp*=2SFI6Y>
zCqSK6-(6J8dmhf94t`{#a1k(z+<HrvINmvt>gQu|^@t4S(`Yx3^cZ2{Cwb@|q~~ew
z^AB$a5_)`^=6P8%z0FwQ!y-~9Heu~|IM_VL^v9l{j{H)LO)({!;+vvK^uomRv8JKv
zAunBVCa)XFg0h09#{3Bb0`r^cdbcJ=9i2SBWqO9u<5LhpxR?dJQdwH%uttdeF32QG
zo!|B?UnVKQ>0JibA9_B2C@nr$))D9$qcGh1QdBl5#GD5^hEryE@L}PkMz954*!7pL
z3=fqokE#kCZaT~5Q!LXK{WNJ*)%0&_u0DsFpGVJvvj_FCP!9Ks%h$b*Uf8MQw@$xu
zu*#Yoc%#M5zez6>HTMeFl`b${ltIm1XD%Xl4iVc&-p!e=N0c~(9iwowJ&?2Fh0Xug
z`sgpiAV5bbn?KgJhN+)3Eq{EQ@c7X1;vnh6+bc(u?F@6PELK!XX$&d~(W1I%VIi$*
zS22U^`vu1GDzDL%_`#(UGqe0yE`LO(d`t#DL<OOx|K}m-!j#@j>^x(;&9=8wTvpJ~
zH$P%`baEkuwR+>YZfbhcWE1+5kcjBaLN;p7ize_^6~m7V63_9_%cs9?hjftmrDt&L
z5_bq+f9bTp%R9h3-^IE4e(6d^nCd*IZ|VVVE4WJ3i185>g!R*TdwVl6u&tsSOa#Af
zC$>!*>p!=0J$r(Os`ewv9eeQi@62Nd$tW*#54BXcJDz|Yz(ywGOqVpRkRRg;kP1v(
zuK@pMUm#6mjO@1LE%Cg(obQDb>!kZQ=U~P|T&?8)ZKgxGf0yevCmh7;4+|*B%h_mB
z2V-8DwJr=fIvHzjLao9`iOwi;yPV;JXDqtXBU>Klw<t(O>7(0C5_M32&!!hUce}BP
z#RuA~W>|Ven7M$tZ+-a4h%vZp{?V&nU%m**7;e@ydKhl4a|OG}C50NA-T(6t9mFgF
zY+tJTrkwDxgN1JTG)+w(TUTHrds%3X?mms06Ju=sgp07yckpyl6Q)Yb9dj*x@^|O;
zU94_1E+X{Nn3_o^1uP+8!4|$Fn8+s!i*jvXL51LX;i746-;h33^O&P>Th;_Er1pPt
zLKjBQb$kx1<26~>NnLnhlRjfvTmSgES^AeR`t(2v@WOJwk0j}N{z(A%Bbu8gmVZCy
z=JgoTFlRbogT-=l*Wzn+eAj1XbBpqfBabybecgvC#NEq4)gYby5c?<k=^oY$x%<u)
z2!ATyTK8<;+4D_~&-SjsIMj8tH-B?PIbMCzgV#@3R_cBtQShqz6UTIUoLNV{q{$<T
zWY{ctt}0)vE2<~YM2#&h%&>tcL)<X7D=jP@bMVAdO!&-u1QPl8IBRIkipQFM@B3tU
zHbC^hz+sEPXV?rk#R{jUc|6e`bMP^Hp)z5E*7c*mpLtKzrg>YvHi{SnA_(tD7Bga`
zrG{jN5%F&lu_13T@hJ_nGTNgUs~EQGEO#jXp1x``eyA+|Er%x1junuSU09~pa0C##
zpcbSA^PC2<YI=l6wtUF7Tk7v8I)XLDLriOl5G<G=S8;#f^!8FMcXaB3&yZD3I^iiN
zcsM+Bf_~M5Gn;K*{5`)gUT19Srs*||^x*H!lNumCv??%#v*s3=R@04=&J!rKzFrDn
z)^GfO*HKmdB{_?5g*>?;;y6dWa#((Qy-I642oG502bMF7mX5T1`N-D)Uqp-Vjrr|G
z*JGW4)&MK-aw{d^#B(RlbT__yVPyvbV$UyqdtvOucjfDFnEP*U1m|zC8X%A~AkW3H
zynzg!>zdEk4h_9v0}?p4w6tXY$4g;Tex>0E%N?@62cgQlwC~4u!~5js<-&xnZv9(?
zbS^&8*^Ob#+L&avIV+J*v87eIvK(5EB>xm23%XZ4bqE&k+L1j|y2ZZ#6)@>sY@#!M
zU>%R7)wgy4(9mG@ta~<C?oj-@5Pja7q_IS)*TD4mjjpcwOzk(JP#WMA8tpjO7gN4(
z7;ptwOg8@g=fpbO9v<Qirbnm#n1?_VP(f@)f%JzV59_=SPM~H$gl|+aI8}Iq4?Rxi
zko%CKElk$PS{NTPoC6u$u@fpu7069@J7DfkOWgn^F*_Y8yJEUPM^$HSlK;yv^1NwA
z#RTk904x*ChmjWJ8|LFk^YLM8se9Fax~S--*VR7*4H5L7xRe{jG`hBjp)jStX-ALU
zrWw(Hc+)g<#|5j*sYVR<EkeUSG!)7JWW!@z#O0&$jfQ78bg3Z$qqvcqG;z0#Oh~J6
zjFH4XUw%vlZ@(imxgjKR;VgcuSrnC5_g-cY%%bLN=6~~Py+DJ&V#6_CwyBF8Vm{h1
zhFq=MvmlwcV1jPQv8bKLoti;4DHz85cPG6<rY%CYql|(MTec9pm(XVp6&P^LmoFLs
zPI;wkxzyF<RT}1x*}g?c7G?GAwAtq~qh?LM`1)C3j&kO9!UT{ClLgNPQDE`g{dF;y
z*w7`ZLTsrRjd^7cRn<}M(q{%(-z%uL9x>jrt11}Z7_u$uRCK4dm$Ha{sB58|M4i}Q
zE$oJp{Sy~j)akL?fjw&%l!2+@ewp}05F7z7BP$@Y%*|S_zR2X)ntznluA4)(zqXOo
z6Q8M=T7H_EPb+376vPl2^b0G}$4#WM3=6t)jD-VT;DP!}?p5qTR+by$CgXnIj1bXY
zY4{wUB|^{B!&>b}sh~HObH}mur4Gp}I+RGSW{my87b&JcC!_GN?uqlgWLS1~&E<;a
zPE?Ay1uK9XlHl@@1*g)o($5Uri@0N*E?yXO*n}Rxdldos!a9bHQ_Dn>_g6#)_vbtK
zer|;t42akMOAJeXMO9}Zy>EnN-^TET!g8|W7Cw9&E^iVz#ja*>bI5Irjwd0Zb1CQ+
z?Dp%&e{LR9m=Fs36UJ2J0c^X*08=yiG2RW1LI~LhhDu=><zLt*4jr%EQ;{DJsvg}7
z53)U!sHmx>H1l$7XVA~Ty`Q1U#x=`vb(iqe@$Q8=3aKNHTwbQcLEM*Roq}2VgXSh(
z8Dxa|&$OS!ew>I<z%Zty3>s_2Ont!J+<0J>R!{Ted9^_k``@a%lE9KoiEy6K!4kBu
z5j!%cdr%Sz1puuP09oRwUVq-eN}-}WZ$Bp&ybqGoBIOPb^u)K2q{p{=%i`YpDA62{
zpRm*mTm`|*J<*Xv#~@Gi+lo)v&`qzdkf+9j`SctKwC(ue?ZV^Ftvo!Xj>_mFrwY2M
zbOgF}$E^46{RM<Xq9z!tZ1&(nzrsioT{p3j#e!vRy3^qLPH}0GSTclB@@>D+mR*N_
zdg`9q8ckitn}~PGSe3Z%g?g;e$};6a5~uudG%pI7gyPxQUS2X|$?w0dXs9oultp#I
zp1_LGuoAKKhu$JBVTwuDiZR(xfVZ3XZ+Ke$iHcXYXtc!mk>BhivWd}i3*JpB&`#Oq
zQ*%)6eDVWbp(_)8BGCF|pwqmmE!B{ZwMT|C_fjROMeDR}e26>$2Sy(pcO!q!h8GF1
zjcmAF=^0fX)O&8OenGbx?zd?wMKNNZTIoammxs_3N+C$^k3;ZVLxH=uxa$#~^u?uO
zM#1i3$_yh)fr%iC-#u@}U(kQFne;%ic4=n&L&Qhg_!HYFAba~vrrd#Q5!0p8`eJxo
z{v4mPhH_&Z-Fk2Plk_YNfm06sWuxTUu^xX7k;U7}-v}dIG;}N!tUQ^I?#H%*@)Ve+
zMR6qLy&}%17-L_>{E31@vhz2AD&-^TvzE2cX@FUqe-w5Mxefzx@lZ3s>mDeVk7UoX
zQ8qntDi$txe^BF>CUT=Zo;{)Y9dPX!;*wa9780WaW@XM)V@AA&T2Dh8Xnf;))cHvi
z8%>Lpu)DBBUJgF960yKc4}ElCiCnhl&K^(uC?Fg6<7Va`oC;GHXCkfk^sd4VZleUS
zW5JwG*f<0lI)gfCEqK&FWLr?w^kg^u8T;7fR?XOPSuwSgjEm|Q@bLC;-LuOL<kuaX
zL2xXY?jS~~Q6pWMwNV;%<O$_ituKIUG)o`)L0N@rT)wY;9(sCzL&URFkxBP4jr9GP
z?)RrYDIoCfpWESj6e4@aL!wa%)N(ZmpfZHV{k!30?WEJ^cb6}W450B-w)awHZOnKg
zG^Qt(AI>OSyrq>DPp}ud8|f$i-n6V-|Lw7C!<<bKT;pyE$BaJy%dT5@^Owk~V>7ZQ
zgy-aBPfyQa#AgXT-v%O0eM)gTQd3-?Uih59n29z4M?Y^W<Nv#pR|O#)Yex$WPs)gx
z%jRAB>bIL6uX9sj3w8P2x=<W`pLeTXcP}jHc($2bsD{5w+lEn8_9q|5euC^Tn=-h*
zj`S%A_0rZXx!OlP5mc_4q|xNW=868k#ho|Tm%9B`gSVSN|KhghoMzlJuxhtPj8%n0
zM_q`4f|gj@zqf*`R^Fr5?Wchy-ZC|fd`7SRp*@5BkORdSyNh$~LU$OJpf7)^dE3mx
zvHid(**sEqut#<8rA&EAroI+4^c4T%_6CxAIbK$rFVwfP&@?BIRhb#O)cMqs0T$MO
zqbrKOu`J%aa{QM#N;l0SCD4Gx%Cek+$o~&=&@(uf;{m))Ut#=FTHyW9o7Yh}UJ@@V
z^l03`JDyW1j$)1S47aJylec+iibkb^7O|E%yZvBnXhiyZ1XN@L`8yz756p`A5H8{8
z+h|~o^Ad=L=;8dIJc?<pilP*s9}V9hew)b0y*T<k@}r2Q$EG07@t?GK;M2bNg|$__
zl-a2qVC}Wp(L8+Dg-LD5UUX;l`ItJt=h~UP)J#a$)nqe>CSJn;Y}N@wL*=2TRcE?|
z=y>3uJNx*!<W@)I`zdo4mE+S52Fa>oO<{si)ba1$K#w5MYZXo%HqeduCxR!{3F&B*
zg4i$OW$ZM&^jNL2!?&0aX7r!9d?gI%C=awp8Bv&QI6!_UvQ&ZH>>QkJ58GB<<&Shj
zLX#OvO}R7X_HUUwnUFu~Qi=G+S58~YdSO02(X}n4$qda29B7UORY;jGVTYWbzRdK>
zM9=D26LDHUittoy)&D4xZNVw{MYWsW>Xi)R)<)|P<nJ{E)p>hV;|8~-r<TT)^4$6^
zdz}%Dsl?qqs3Q9sa3lp`_?~j#{U^1ayfr6jd+#BSC4_R0lw?iiJ<)qJ!}-H}J^2hX
z+$rfxN1X-GC$d6Ud=o?53hNJuPCXLwd*f*12D!_)rOQ|N0|5wuW&~*G1uZ@9$#kK!
zh{0B?#=#R+@9z4*I~lODGRg2?hPa-pa3}=;-12hP(P{H(q|SbObC>gse)$MO%j~(4
zNs-hBtw;5*3N<46Q^f2C!B=ndBf3Yd8}gb2Dq>-~TBrTwr&Q!?RCX)-Err<76_KR`
z@$_STLZsYKX>=<l8B|WSO8`l$i8ZY+b=*KEnuU7+`J>M@^|^e&A$iyD9)`;gt`Ib3
z7?l}2$&x(k#Hunth8vt^mD%-|=MzUCZU1Be___x`cOJYK&K2CO2Bby!E4Z6<jILi!
z^UO|LfS6EwpFp47XM-sh5RkR9<3sro+!*q#NLeE{C{_^2>v5Xh#i&&R4x;JjAb@gr
zZ>|H#pG3zQr{Z&wVAco<0<EVjevFP7UC>AuCP=2|T7EAg*%emN`qy!ILrxezI5aYc
zB0DZzy;OP6bC++^XtU;`m_Nx|FGZj+3;U$Mnfz+H4huDBGE{ZeDdCr^m62u%z>QFe
zh-p~W+2<qA;6nHgrO3cLuT1_^N*w>U-+W$EnccaJ50*I9wMwkbR35aU%-vy}h+E(t
z(_Q*JB)O!YG*V%5@$F+BZI4FetMY%Gg$`lWY(T-iPr1sOSd;o;c|Xw|)~l7!4oZNP
zUu9M|)Kre<B*q5Pl9gT2l+Wnpxiut@?2goyGe%@-+$F*@1TQ=)N|V?ZT*!nnL;Yti
zCWd-ylk~oDTjQ>j$AV5y@_!0PNDsJ@;y@`3Nf^4LyQz~#+~z0KtvoSRWYZL4@h|^}
z`X6OS){)!0$4xy+mWD81LP+7v`U}11S|X*R!#iu9&fQSdnpJQk5yPI@1`HkPiH=8w
zmcWyDgkCWt?vsFwWHNW@7Td>$@>8+sU74Zi*p+VqVzwK_t_hwDut%0+3KOnd|3-nJ
zKuhrqD?Zb>qf=TqMb78zI##0t_1t2O_Uq{7xf8T_n`P>lIb*uAPUEZrh%$np=7<X<
z<z4p;e5iy>CV#6H#wOx5H+Y?r0~JYr@P2t%9UFMw<{ea$>26PLI$E96nt+(jKE_7O
za^s`hF#^UA%g|!Q@2BBv|Ce*3$<n1E7tP5(6h<O>?kFo<C$UbI;Z{B%OX6P(?P!}L
ziit2KXABFTwMT_;4Wve<F4)XbCezgjt*kthQu)JZp<+E}{!{dxa0I}7-=BhJH2U%B
z6qym)=d@+9u)x7vcA|r#<4f$Q*t#M+T3tG|CBfMFKVFdPkq1Ts<n23a15{IeI5rQ#
zbv0EojPa5I4#LrdV2q^2$|yKo%s)!7h5O#@GH?nxARC~9sGO9+HF%L$vX4O$W_Flf
z@-rAWIj$x+bK`L?WBx4dZ-?aLhTjH*`A;k9Dt10@xkAU(KTZiN%+lav%gsqe{jFoY
zs%&&2h<tZX9;#^o&L(dr!TqlH@chwOjJN@O=v>Ens?1Av@yl)WfQtzQ?SAD5G|0eo
z8%Q>S@11+R*men7adXK)oKv;*lAoZ1Oq<jR;xd7SQxw0w_Y#UD>9Qf#&fDg0Azj}<
zQr6`7AWp|S$f}?wEJOCf_Wg9Tno^WKar1Y$ws~I<6rTL$U-SWNSmAvqioh}WYCbNa
zIRgAlnFP+)Ez*kRBXFuOzL33%hi1PT%RX?{LJS7EC4LoW$AT7XZV>wAVrg(EB%$&l
zX)8R<Pq^NT?!9Zc1bH$;YsgQrwK5cdtzl_7b3cFIY#Rh3_z}Ti{MeGFEai(0zku27
z&wsZt!rIJ|e6V3!i@JQ}04KXAU)`QDymRBcv8M9W!@WxV#TNgja~HEi^V8yqPr%w4
zZ_PGg=-wc`Y?H)J$PsWyX9*lvZ>u0(p2U?vRQm0S>|cobCOm0-XKW<or`OHG;Y%SC
z>WL11I9^=X>SpEDS$f!LDQE3ja+l@5`-z@G$s~VWQ^JyVs1YmDep;+Id5dnD=r}+a
ziGIj)<FeL?`m*_m?+(SXENF^fNNW&af9@Pe%AH<*q`MY4lLrZ(NOS>01y2v0DLBny
zpb@))fx_51*mL;;?6cK>Q(7v0zp3bRq5aMMIngOq*~tTd-7N6Evg*mJSb9~~__tnO
zx-~!+T@Npd04K-0;P!LEGn&S&s)nKzBURjGx%XHa&ZZ$fj|7-p&l-$Jhq&kOnQ0hv
zBq^%J4g&xWO09zY{;0a@>mgu}mR|%XyDLAlK1O-<U6QLMZRqS`@ty5MBmqvB7a}{W
zt}Kc3W0eHmLkU~zoi|NpoVy(H{#~Et_W5SPgT)+MAdadLT&eD8=HLL`?I5Uv8u_EM
z3)r3Q`t6KAym^?CFmyG%J-c3_^d~`5w#0QnZ&?hkOBa+ER=?aKFnh9XbVYJGQ~AU`
zHhJZ>v|R1k8$!Rs@gYpiVBbvS=pg0?v7+;{i8cDDsD$HXGMrxoIKS|Zf8qNql7(hX
z*qRT~14fAX!;f9Z62bIbF*HB3SOr+{YJAgnas8LnDv&V+<Mg4EaRX#>TudRYafBBu
zCOiZ(<sqgo(rFIi;HC1h#$S^n<T_F`m_GqOB5m?BLl9_M_<XH)W6=%S2!Vn%CTyyc
z#?A^%%<F<+v-PVg{?ipA6ovZZ18a}BV1OwYIkRpAUdTL-1DHd8E#J5bZFZa*uAIDk
z(X?l1R<fCjuhDBo(+Hmy+H}~Xn~S$M(68>y_oO<A6z38A$h#-Xz$g4&H^vY_0FO>$
zcgRt22~r<!r6sg1?s@YN!Te&?v4<jD?U<wF^)z6D+|0%z#(ki`zz2_GB|nl0Ay?(G
z(X^X}1`PhR<D^U7*bARFs55VOCRj3MYDr5O_=9Xvw(=qX?|-yuy2b(GL9YON3?K5k
zwY<3Kq`ueuJbZd)I{x4yQ$HXqH}=T3=$aot#g-ZZ$8sa2pxW&RlX0L%iPKSX*a?oJ
zQ(k_7l?Hjb#_gkba>~&Bff&ISaKx#?Gn{S%dtrD3Ib=MhExL$H^~-2Z?~6CtUuMb4
zPZm6!+RnC6!VAqel5mL>=<ao5-`paLgfti@1c7iE97qJ6QzB8_Thx@8*OE{Ey^zEP
zl8m0UlV?e}`>zVH9}=9M+4B_(X*b@wl+dSjHaesit8e^T1Rv0!zA)cFCIvl7tvY)t
zllgreG;*M&&cw;M+{N-Jl(OaA&+~XTrz21X-WB^i=A*WjQ0KV?Nun^KNapj@8qsv3
zq+USef)JE!UK4AXuM|I|K+Rb6HLEtrJ1K9@<{8~Pd;a68g7n?BiZh@)Qd#$JI<DKH
zD$V9K9WH2f1Mgjfr&mqaOITH24x#<4(2>|weONSUizJPjpx#-7F8S&4@e06LzQJC1
zK8ZO_o@wLl^6IP<1gR^3HO(ZrIn@SYqVPmrYHEJ&)WOccZjf9FHywVf>)W^xD%e8%
zwJ24%K8K|zE-jWaCBLI&3_f`N>&qh8Oe)e7u|=o<_~si;#V0o3+3~xG&y|^8_m@}x
z0>f*2=HN%cUA8aX{szOv;0Wqzc@NYJV18~?^w04jfZLT~nlGvF<CSkQMAG$QKW>nE
zx1cN*1T%Ua0(y)9$*Vye{Vne`RhCJeoP1WzkDKSNjX4MpMiTf8G44k7tte<%DRw%r
zWy{(;t)7n{$0e%qO}7<(Ay*_taN0}lIXNJ~(n8l#T-P#L<p-&lgLt=5?pl4WH~hRb
z;;y9FD@5rQKpc6t3e~i`|MlO~C4NV<f!ux#7`^F$?sgC{zLuGKcB70L#pPvE6!+pr
z7b^FyJWi$PH;{M>&CabC$g|IL7&EYtnfgStpvW`}H5%R2PQK>b0G@JIhrLXwq4M>{
z^+%HWgm56ViseJ~;&^HJQdFsPxSHEuyzxWO!&RSuL89AtokQB9ai8JT_(F!vqk&YZ
z$+RsDo%ShrVkW&fzVm|5!Qx%jQq9v*UbB1LZf17f<na9E?L35*kyh{0-vW65sh?rO
z=RU02U<>}uD;=}v?hQlcA%+G@>zHB?`KnXbHn?21Q3Ie5>Wu@sV1TBAYnba<d^6%N
zcPlRU-=h7@cmlTr_4%nHYYE#-W47+0DgfV9wPut+|8(nKDSLIZvH0N(?e1oS%b@WX
z9pqu(J$sEoTAfgOaP$nzmF+a+*ec<%fAb5z|CdpU!c^!fiLEQU8nKM<MXpV!^sTdx
z(Rl+!0!f;D4iHJVk?(O9`lF-8)vVA{3BN9=9SXVZ-~b;GeH@x;8Z6jiH2T4aghsWn
zphHfD*{hU3)WkkTq%eo{1N&%eeozRqJBG2sx{t<jzKOoHv=oU%29B(U^eoE=(}heq
zDmXa_<f(oNO81_pe_-=-)4Op2Q)M|pw(~3|TbAyW{B@*E{hd!S;$E|{b=-#TJBNPR
zBkM0_|1vc~KfiRqPyY#cf}l6-@1%|Phyg!74=8Dcnj)#QQpY138X;ONV<BtL2D#@G
z>tjKWWu@OqoT>`^e!uEd=@8h2@X93e%2b(EzgUcOOnJ*(%n<S}Q)`Te{j7`YHvX@(
zotB=ZuBZyzA$ER~1}Z-NGOG3rZ2uZ2*zscB(X>OwPV-qF^MhhP&OE(Gkub~(0b9!}
z>!euLyPHQ#)dGFyZ)m(@XIK^ft6(3B-V)v?Y$0f+>|nP9AaI>i>9nL`TmlTwCf(f#
z#@=H65D-Wc_rNRlbC87J$-;`R-^q=H(B>Q}kkgq&ydN2_gfDrRB-PX69^Tmv?=t#G
z=thslD>X_|77?X~<h~SJE-e>gpbNB)|Iw7qi&|%spEVahC3k)eJ|k0uT+tJmwkHCD
z;iWwEb2(Hz8mpZrGbngP&xcyvVI{T>I{apdivr_^09ms97r7L>r_h!S<a$Hx)HL1G
zPvFe}GpEAY;2fBJCnfQ(qp?gPEwgACdJ13|l0*3@aOM-S$6Fa}aefWWKi1@G7lQAl
zlY<{Qg>$5E5s8>Z)3_J1QP!?yKu9&D4yooLJ5uq^nPh@>ziyVisF01A&~-F>d=vlF
zYC>t&9{zx5zf)=In@Obs(rt-g>;9-R?m28}zTz9={PfGer9r&Sd{+=_1s8Vf`}0Xj
zV?@e5m99{lM0)$*+n#1b9ZwztU<_VAW8^DpwepsW-s6$g;1GlO{UX5;M!XKO74r63
zrgAe7(Db`#xw>$<v%#yK>B4imFfwwhcr0cAvbs$DDdWlHH+EC*AqV=$%E;<GuYf)M
z#?koO8y>%$=PaWGS!h*fnQ?<<R7?Kbm7F5V&wqXakS{JkGiwb#dyNRMN=i$bha5#?
zLX>@dI&S(;RTTaLyaYJ<qPASc#Eg}-FnX|u=_cp>o{@+n57caW`w&!p%Qo$?msqS_
z3L$j#wjiq}0%Bxj<alRKHOxyenFxVE$nt2f-iRqKD0r%`ug^c${Eo?EP^O@`SjCqW
z;S7j`o7m?S%g;_uRkRLd_F1xwf~`4spKZ<6skh*hpoqoYmiV<CG%j(`$)SI>U}s@r
z$vEU<-s;Bt%rAJmU@ZM-xlbCc3FfOxnY7Ou0866T4l^}AWQDbbS}CaDNB*XPqRXYJ
zIvHx2{gi^O!0H{@99)ZatLxwHI@dB5>y=Uv<yMs&e8puHbG#k(;}vC)Y3V!3OCq5r
zw*EzJney#SZ*i&slM9@2FQ*$-xT7ieEQXOe?9!yxa_@fSVmROHUTP+U1e!bG!Qdt#
zc>>gvHfX94+3b%p*@h|(e=PF6LEDE6A4&4eYQ3sT=?oTjb*DbARRbmv!(mwYMzKTG
zQu_kyPIYM)pqi=qx0=hdSpAuy{P$T?Z{zoHalaq1{@RGw?0WlbEUstD%4RVuOz9gA
zmBH0Uae3$dR_%i2VtZ&#mvN!n>3aV1UJonn#q?QlG>+L+0~MLmF->MwIanC^th$gP
zNg`OWL!Z3`FWVNo<#YDjgq<+#+~<kuS6VMe29Axb^@s3Kj=w~cG9RAL_v`S?%cVw4
zQ0_54ByIo_Z<^l{_eyZI`@5;+<(JJ{@gGd>J<-a0LWX6LJW}#+m$jDFTkuk6=O#;?
zsbCN7G?=D^>7QBmCfsdLy1c+F^fdV%49?HTYInW7zX%k7&QMxY@q$BI&)TISH6pi(
zje$4jSe_pYKUO!qzbQ?%1u@%D#AQBWktWd)WsR^`!Fj?Gr`T)rFHLX@%4VBxkZ2|e
z?{Ej;H#^~4{>^BlY$(hY4%sTmSc~5RoxfUdv_0&ena<@f4wL|dopgkRUwwI#Vj~$|
z#rLT<<hT!hvOHpJ$!O9azl&RVyxq97dx|PwPSQ7R3&J7NB;9O4mF0wGc3x;249*2P
zBqQRks`+VMwB4<tqHy*!&g~9-O+6fCI*Iur0r5sV@VWJOV>il3$o-LJmVXbEn~u;E
zXi^l8K4;FocWocJ`K7~pBT^ca3*IpT3~Z;-h{zbs%KC6c;uNX;ni(qYS3fsFlsXX#
zPt98@;yp8O{60#$b5Q}i@Y8a??0+y2A$s=OYGbre{PK9YYXN#Rg(tcBrQ==gNrn?i
znc=<|O-<>WUTwVMwuIZcWXVdS85xl^y>C<byxRxzx9?1oYyXhKT-PdIm;Jp9^*NWH
z;>*!#F3A?kIG4sr(xl1mpGLMMa;By~){2>(l=-&6^Kz7vaxSG)X|1UayA&ID%40L|
zdAW$>1fb}sr?D$S-qTv}2TLvwNIUG{-tFeczk<njC#~WLWHf?Gjqe1v%(%2jc|mLE
zygDj+SfZ9=B}u=&`oS<NMrM~5y~7~Xeb^`6nW+zbE0AuasXvDwr5C69LX?G{k`ueu
za|w4u*|D}6_e|rX;Ko+3YMjpg-=;%sule-+x|deK8D%{8z5%+-p@gvlVVVzWX%dnU
zTm1dOns-NM{3{<E{ay6BJbWKLj(DoU#2VWO=VRZw1N2);K7a%-?IPt?Icnu+w>hfO
zhP=~JASRlC##=1>uN^zyaB%%<fuxD?XlYnp2HqlVZNp>be|7TRU^L(I^8BQ3y8$H(
zIda@YVCL(7e=rJ3y)v(Zof`9hgpdoS!Pq3Eq$Za(LiC0^=hrIPn8IwLNG6Vx&E3J0
z(F*X}M^yDfA>Md@TAS~0`lZL7&SVZ+*8bq-wa}14NixlqJ3LIuUK3lRE6NY>Sq{2E
z<ad&BiGz5}_f~{ZhBfLNh8+F9A~MMIV}-oVmV71}xpG5ZLH}?XpuNkh>v+r3jcF0Q
z)Q7kH)L5Wst<9<tb@8<{_uXFZOyXT?0gdMKeHCNKR`sM?5A)tg<|UO2;&42$wgPH>
zYdh1Ee}3|eKl;ADl{-dn&vZla7o0dNt4TU=%;?c_*w3AiEbf+@8!iL?GRmb<UES=d
z>AFYC{FYaIiSw@Fy12d3XA=Rz&Mo-^TTsEf1b(@4N(In3`pLsjG!bL1U{T*CIjmuP
zKVN?N0}d6rR)^08k6GmRc$J0h<le*F2E_WySR!e?t_TFvzRU!glHG6TxK-~yAVn*E
zzS|Kv&Mb&NI$B!K&+Lu|vqT1yN+=q$@PHpWh3Dd#g(Y%r#vZXSI<jyRr(Di#Bw>s8
zt{g3m9gX>zUp!jkdj0LcViI|Lekl!<<zzV3H74o-!&GGzANw^Kg&a+SSp6Iyt2V|F
zo`Re)dsS!egFve7i1I9rSx^iSZ~q?W9pt*S^7=e-JB;sFPt3^b+tD-nu^59E(oy$I
zBs{L0fICf7j1&|4tAk9MY6x<F8nP70jawyn!t$i$jMv9DZbY;EQFd+*Zn^^B{KV^b
z&Dnz7mm9`m8js2;+B0I^l>78wRs2!P{jS|v6RC$1bDlu4Qye2Y1Gpu&gQp3#ZufmK
zW5!%1CwCC&K8Qj|<MzoA($8`lS_!n*1)~BCR{F^OD0ZvfArs-~HlzaYA*`^2D`d&N
zF*7=fYa5h>Jg`nS2}2xB6S*{>tZJGKTlCSI`M>%4Q-TPxS%GH?JNbR16mofNvUefV
z(%G!0Gc;Xq(6n><TJZ;2_~Mr-w<sI>=>|zk?Z4R>7OCk3K(?*un(U8%`U^wr+fZ-<
z=kwq6kY_TpwU><YgwSok@9XAa+FY*9ErWwIB~E>Cu`a-2KI|!sG21FbVs=ptwH5%~
zXsTJ$sL=7$885$4T{;6h+j1L_!Vq8JSYsco{^*G0&fuwJr4cesJM@FAJwNtn$y`SB
zo@LEVk$MlZs{9bLd&7!vDTBhDt#`_<vlpflzx%g+N=1jp6$Na}FI~=wqf8Yoz(kMO
zGwUbK>ynOp;uH;C`4s5FjceV;SH?Fpk*nOOPk5?HSPQ9-HJ$h39Ak0QiaS1>&te6<
z-=(HK$>k&|^$tZVFB5El5V`I5az|N@<f+yB8+_t<e<AF?s%i|JqpL)^K{+PVM5Dd*
zYpi@NoP>Ga^kY>+gb2HGvTHeGT;2Gu>W2&N%6|>$&umnr%=eUrRW#vWT&iogUW%l|
z4cU#2%{yCo^UYV5u2beFD<ykn0yfBVX6TcoW{n~kzj7vBwtj331|AnDuiHs^XODGk
zMCFTb`sB)X_fE|WS-||_U`giL*8vb;4w|aeW`6(>J-iJ>?|3b0j30PPpW9@9T_ifK
zm>MEYRN$UJNJwhE!)CkFW_HVAaD0s5>lFVq*Xz;0Vv_wGdBc>s#C02q)}Hc+qm)E)
zRJW+ODkEo>%=!3LgdJwD0`PX9ecC|AH^iD|2*2aN<Ya)jHsI>V?*^CLTB!+ty)o!n
zs@1{tZTKd>YdBtta@#1(&nYhE<1W9O$c&mOr_BqYHFkZoEBY;Lq^G}=7SA|Pg!*uD
z^ec&93{8VJHO2$~?EJ<2*jq^>TF%ca?*DC0*N}hEu6j*I?K%!6XSWV#iIVn4SLG*H
zx3;*qKxxmXp?MOg(HBT7-qY_@5bG3}+^EX;_t2rA>avXkp<XSFaXbD`Vi_L`6tA7n
zc`sNTZBEr)U7opL`rV&80F;bHcQpMq=v6Yb*v!%KKS3scG?gSYYN(arwo`fZT1W}s
zQ+!X+m2%z$a|`W<h?wl-{p(^sm>sgOvMs3IUf%O1DIz~g;G6>{kV-VkRhXT89A-A!
z4&h{Ld4O##Hbo~?5;04(cBc65!BQCO6ifA2heL_3{mIScicC*`y@3;;2$DnBlLbvu
zW@r=XJ<ylHmbwLMXA{PJisz$o+=aSJPTeMl>*-ImX)rI3Ca2%D9P7h7hDf5i+7<l%
zxQuW%CQ^;9ZBKvVhxmYf>vTn!=o_n$c<|FVxSN)6X(6>aF0To2G0LgX*E0fCR{;Dx
zzQY#f_xnO32A5H4&J`DWJh8d#syi}Tl5y}{-z4<lr(~RbB4ItCP(vZjGh5r<B?j;g
zytDLhgP6pMTn9dn1cj;&zQ@ScBfKx)nsKhCkAmwW`_ca@YhugNexUu50TjI<I`(IJ
z#J~i|-t!NeOw&YNjWrKnY&)_7joLb(el!PxdLK(G4fZ`e+&euz7N5GvpxEu|ltIJZ
zj=H#dMk#JZHKqr#TF1jQWY>?E<9f`lPA8h0KBzip!7z=Rl|*_aFNSs|{SX8TEIuNn
zOx3q)jvOgh88~+%wG%tCS!h{~%DUPGP;#YGxyxTw{-&s0foP=R*#k}0Ya5w)WzY8t
zmwIT?eYr)T56>8!`Ocq}Xr2%CyTbq|B((zh>wIeHmBRuJs6YQ4c*!X$AMcGTe)-SG
zB<>pum^0&<nHdqApG@bNma88#sNR*>-+jqZ+L4%rfR=@d_d!>S&1;qed4-eH^-8Lz
z??2a(g%3@=>S-}{_(g_oF3RxO{vR(*RRORv*x5hT8>>sKf3}ns2vZ`kZhkas=uKZF
z#yb2xfzzaTVx2LI-PF#6G<D%2z>oosRLK-aa0tVOHUMV)Df@7FK{5A=(@^T*qI-7V
zS{q!FH4gB03?k*;VkH`|=t#WNxx5q36co-16EEIl-Gq`J(LH81Vset5)iH#1+E#er
zLoR#@y`|<&T0$T*e~l6fV9vgAJhrA*llULs-BOHh&HlNrQA=L!Ef+_l$E^~6Q&oa6
zDG@bkku6X?D<aKHXc%Qhe$tHUvtrBt1+IO{{)2wk0%*LE{RGkA#eAIHIY^TZXpR_l
znXcj$D(6TH_$htmKMNqmMeG9y&m3w~9eL&&amO6$4E}AEFjqPlgp7sIMOMQ_h;jYT
zK$Yt2Qbe1NEik}?7yNRt2pCtkEh*E~UpTd?;g=D95(q<o19*Ci$)}uh*J}OIp&lvp
zZ(CR;GLFz!DIU!lOLtMaxz#E#!x~MQTDyaJ{4@0t+1ch-GA2?mi<wg1ZDZ)+SlxPF
zYUSy>=H@ke;SDugH}B+^hOkN2)?{YJz2OgcE8dI8PX9fn5W=sa-@j_e-}z^@EO%lV
z_EG)kSbNYmO+RWnjdSDd2(IC5mV96NTGm)kBlwRwq}Dm&WRicw4qtm@mhrn)Ee=Pz
zyH0(UI)%oXK@X{q<{>VC<JQ!Yp=xow%z72EaIx)u5j#`rMFpC^^2a_m^*EawR7@!7
z5V|WX<@|(VIzr$`?B5+l11}MT=be&YFqZxnw+x6kEuOn^Z|ReOG;u*J=tI~7w-ZPZ
z`PI&B!utY={nSI@Da$iCIp_-PexK3e7pw9Hkvh-tXQxnD6pcXp4Ll~coHSKmH&kiu
z5+J&Eyk|42;chb}R=-@UBKxV7zfI(XoX+9}zZkOaAgcMjbWe>qp>D0<uOEQ*#(cLE
z=q{}WU0116ZD0<|VURJ4X4TV~d~0h-!83k_GGYB}3tX<)qj3j`rwwK`dsJbAcw+Z-
zXk2aESr--K<DX0k?8afDtgHEKyD$Fq{%RPbbL+z8ctCLEg6b3=Sap(QrokvQm`i@^
zHz`r>(}Ia$T3MrtdT46I^oSfac<YZ6!5uwP&!1R_DWyp6v^Qsu+m{p<RXmk%8m#1G
zPfZ-&P6`6m@Aw;k&M&IJ*7x^xq^8!Rb4s-7kZ8y+($l?_GU;v}o#!nATXlIbAJ^YW
z`b*j2EF0dlQSRz(bt$hL-KOOch?@6l`xiG#&>qXV<J86ue+_e`7tyPRSlsNH*XWjL
z%raqO7^C-a@f46i7N6i!GEFv{U?K3K4*l|7@bm*9Y}L^1h1|E^hxxwkA!ZBt9qTpQ
zyllZv(jrYNfn=+ysaeIyztru(p*0Tlp7-Tmri)UD^FDfPMCD!Oo8GQi{(fWYSvu(x
zxLuc@5z)PS_v*&~vVfcys;}-Bl|RkkI06Qq=`{&BQ_Q9t`2Bo=F~>z>PjogvKmm9+
zkX>~JS!BI-7)}&?f>%N*@it(-^ZjIyY+P9L-mK`c+*|7Mg;CR2yY-9on0LLUrteWL
zu38$JQxBV%Bheou6A;bSaoHh@7=JHRC7;2OMCJM~SCVsuNCntd%p57OgkAua8-e{I
z{OIt-yh`<<mRmZ|*DZo4_k`i<f?3t+5}D?}LB1@Jr}%9J9G9o_g8e1VowvTmu4C#=
z+`6cJE)FLS&I#|^ckW=8TW^tt^!@%WzP(xT)%AC5IK0E+nfOM683pf20%y*V#)i|2
zdQVEpl|Q)~eG{QRLHsNETk@Ck(h?SVv2**t&I5>(1HAh1)7en|wn-*c_COIZA>&-7
zXBO`z1YTDf1#{}Hh@CFb@4_D;JAi#`Vd|w0M|J7|y}W<2FLE#ZyHG6X6bb9mQtDAB
zNgNt>piQS)xjZW3rs^J#^=cYH$QUO%Y#IFZqdS<>_3(mCHAqj!9VdDXaV8~ZY=GgT
zx$K<j(;bK*d}}!&6rzo)r}9KsjGh4Zvu6uimGCnd<RfV<STTS-p%CEnn{nH!x+VOz
zfp_^JXY_1DLmDA0#16Ly_Q4TfKg5dZf5NZ}StP~j#7kkv<g&#)Bc;tqG^`LbbSS4f
z;@|#Xf_z+5HL+H!D|8NdwAKz{B9)e{KG!73tZ|7Lt;-kgR-risY>%Z$+YNthW$2Pq
zq?VsGGnDgox#(aN_y|X<O@rMb^eeMFSoL=4_xmz4-?)^GNL4dI%vRlK7Z&l^?EaXm
zi=6=a9wu|<IHLRkd2k#F8;6u8Q4_^&O!gO~!<Fa6hsiyme>>*%KW#-B+~fLuZN`_s
z(vJe2-<P6v-g)1?GQ3!<vVvj8VZ+;J<3r*1$A^00Xh8JUxsq{>4e^a^|DL)9c*(=N
z)^8T5cb^K`xnA{OqBa0AV(d6wMew`-T8FM>2C(#f5a?J8XEoiN<0Vjk{QEr&wRHtr
z#*Z#>mTm006WfwdP@6wsRE}33W;irIMv&Osx&C*6j&f}c;5o!A%9D(TVa)ebUloK7
zY%vZpqPfxVofcq-z$;UeBU5{@r{^0=o{N+M2QjQ>HXp0l7Iz75xA%0x6HKqd%hMgT
z@mCi@pea}&n<DQyfNn;o=Y&Z(4iU|-ACDF~mAKP6ov&@gw+M&3Z#oulFM%j=V`u`V
z7d4M(P4JaO>qA}yKpt}71Ru{<lSFMAWlf+$zx=0|1G{*lQ3L(299z_VfU%2LFQFa8
z)pDpU5G|;}MRh0+B&$q`pbBE)9pACapp9vzeC=TtJIP(nO>}Zh!|8Y^E@xD^pgt$G
z?7p$N$zR;y2Oo+6m^E`a)9Z*AM^pJO1nv93`zl`cCgg0I@yaq&Tml!oY$JVUFfFnc
zof^|P4hxlMW)yCyyn!DcVg%Ft__!!)@-n1-*8}sb`Ylbrn;KOQ>EjiLYQrI14S@y8
zKkQb=me}uB5nqvgC4Fj4*F`h%ieL5f56T5^aOMPT_ar*44J><rku_(PSm1oQ@-};J
z1;3^7t?;D&gtfPS@)=e6KA9rtc%}(a=e;__o{|#z;HUU4mN>jz{OWtD!WK8VG&1<v
z4;r^>QeT`FX9jFlPiXS<+y@#w$}tB6(yTo!5Po|Hhmlxc*Kv_a)I}p;)U=h1MH)hU
zwUq@Z9{WnSU$V`UO}cuWiS;u)B9M}jBH;4-FOA+#nWq7-GsOUJNjNhUeUa{ZDJ)BZ
z-G0Y^n8hwGiT`^d4%AZ4l=qT4r&%KeB*r6BEp2ejSJroU#JR!o@Q1~amt)dMY&{3p
zQpoDo15iG9roZ;*{Ng0H<!%c!a;TuJENsl4y01VB9`cS=(yOE{vyALLfC6}-VL5G?
zuv6X*6siJfmE<50%()>Gs4lzw^bbR-E&~9z`=32??Wx3Al-X0!_`Ofi$V|6P)XNBO
zpBx{0<NYwTA~>KVVm$FneyE%j(Vx+xS9KJ`IQcj0@>-gYE(jqAb?-pa{BVXBGc%Fp
zjbhxna~SGX=<v1<;pvJe@M!*)Pg!_BYo&NVhgn{~9HrciVmNQv^s((;TlSFHi!wXE
zX&=PC70&EMceD@?uvytBIO*KD6~gxOF1o4ZY+I5lB~JSC#e4kQ2WtbgOBttK$rOO$
zMRR$!aJ2+fSry6E59@e~&7NlPz8#s;c{+zT`X%rEd7~g5%~<ov60RRc-ZT{1*UfFe
zkoZgdiS^MEQXYl^hgB^lQ`3qW+q~<T5??trsF>D8(5`h)UdWLrfzvh>bVUaP2Y18J
zmB$*}V4(#rLlkd~KpbV+vDR-{+QP9(!hDis^^{<sTloOQN^y8mpH547509;_?)7iw
zbClUtyMax5Xu8)l?^;gAi#pdkn_iNHz0y}F8USFjRG^l7;(D(A8gdSWtkL6>fdIjc
z!4Zq$jorRTNbA-{2&Y|V1*_c~{TIppyzl~zCVj#%IiqhlPN^YTw3!0ic7Mr4fw0lH
zNGZ5$FVCit<TQ+tI5bZN2N*4BnycZ;hMJUY8NY?0mE&2}Usq_1#hH3F#z}7W&8Vlz
zMM!yhv<JRx-<v+o2&CnRPCEYG`ItU)AlVESnV$(`=nBcnPBwl`0O+eb(Ni!qoVCsI
zX8<O+O#y|+h`1}GVJZG8l9Kx^=8n#f28L{kAV!xak>!u6c?0s#p5K!2M^wxd$jPZ`
z`JE)x(GW%|m0g_}h43}2=UZ9)-8-AN|JUro=4{@PkI^=#3`z$?rOMhC$r97%2NF7j
zn<1meFRpxZh4uCN^EbdkwLjosWk;Lj$@@YQ<ZMw_FTP}q6GVEfRmX~@Njlp^KeQ1O
zI{z*7;Dh1{bJ19y={{!GJx;M!h;t=E`gXzC4A9S@VBfy=Cc{eNADxBK^)^$U^+T2B
zq{>S48%ha&tdo55lT^z}D0S_(q`h<&4|Lz;4I2$kV?}B4Gdia4p8pRXV-bxU@G~F7
zv3bD#y}v<1nME;QT^b3VhVve}-MDskhS2IFTpwO#&c*=%pdnARK75#DcftNp`d;$O
zwuQxmtTy8qmD9*^ozKm54_7<U>mz&Jj4r<l^E2+%WhEQ5z@Xz+y!S7zy9d)g6p&li
z3s6kV*jT>R*;Z}OqXmDC%MXO_-v*35*SBUmC{}{B8IFXgXYvUaAH0-%q$wkmTSq~P
zU2zZD$z4;ugvc9Ul;1HJ{lj)Tn1g=`mjT|7U2A)Db3@J7@7gyW0>e<k<qj&{fHV;`
zr8$9y`Wg@Rz%PKoQ*Ou18gc{+xl;Tg&ktW67e>Jhac--s4EA^W#%?%o+pd{sR}}4X
z0X$Y#B-`uNg~k0(ayv)T=Zj=Z5}yi5UdldQooKoA)LZ6KO9{C~v<Qyar-e^V@fpSB
z8st8=+Y2uz08j9HBZ%uB@&VCqn~1IqD+iN8E2?X)Z2ZM~=>7jGWB?O7sNro-PY+r>
zF!>L=500CzY)XWmlK9Wn8;JY&Ev0BSjN3RulLz<9u7F|ITJGAusQeY30M`a~@UBy{
zX@3D@R3G<F(q8>6WCvkuXD5^Giyi{I0j=(5BxwB+_4orIDMjuCsoSm1)HngAJQEZ>
zWDu9%i*0`wOey3&>xQTN#-SXccg*p5*~ZFdDa?8&`w?BB!801a_}&89i6NUtOF(sm
z#;}w)PK79u{K(nPfJ{H=Z++lrI~>=qIZ<X3_0VD5e5~mCWR)EyFb9AaPr&i4RHZPT
zS75Z=qdkNmCx+Me-JtA5?_h#bi0QLafqViRznJy?@lvA---B0sema=snwlCA<<uz4
zhlYW7ZhVNJ9N<{0a>+S`QtRHy>dMMW+kvQF5;R2K#2fhloZ01x8joBKj-`|^MW%(t
zGXJ)TpC%abC4z1vU%aC|d7wlDRj1E;F}SDP^T5}=w}lVwz1$tc?#lD5h2CR!Sj6P{
z^SjLxD+|;Os@7l7Cg0XPZiO2T#+A4r>EB(y(_<gw*Ks9t*MG7`u#HZGX0lfuv)^)2
zU7m>J&zTN_sVw<HXMRU|Qf+Omh<KPgyL&Dz5j4ULZTcs}&jBe3zLT98I$;VS`D!x|
zW_aBslNSY11APxt7XRHgkR67ZnR)4u5g#Z^r&-1qS4Xuik$OI(^nv$rM*9OU4_LLz
zo|=6q{ODNNVP|Etync9KQOIxyB+0qGF8*PkA9-fnw|XO3ri4Ud&w2*eapH;CYOIDz
zeRXpqFn+PWRBiEfYH)BcP&?Sr5eD>SAJ44<Hbl4v=!7$WroSf=U+t%sO+t2xJq!qw
ze_CamUkBbUvoRt+1063;b<oL6CZAt#CV{Y!^BQq|LVs}L-_E)U^w6lw_xgldqUY!9
z9^2KttApRA1c(MBsrWA}Ed?>FZelI5M<2CC^IQjAs3Mu}p*S?MfW-c~FrasTR3PGm
zDvB9s!Y?%@#Ku2%UoF^g1hFpst`7=QIJvyc!J6-)HH&_o-PmZ<l*tN3lV|wv_m)h{
zjbLLj_k4=A9_B{R$lm=5M0xGUm6w$riyIg4k-0N_;8ofOGZ7be6c94hI%sClBfZ0E
z4yhF@JIqo7yR!5b82ZzMj>6elY_?Oc-H&|`&a*pD;XU(*R%cT2+;0qdyk_0N{>O{q
zMQ2OWJ52|24pJ$un&X!X9wTbv>z@;~%bp3Yzqa?PA^t7*soBwM-q$A0clA^IW3wOU
z>ly(RTj#@Wd!2K8=oBG2R>-O5$R6LcXc1ul4cdH}2<}Rf!pFX4IiAdMM4V{qINXV&
zM*L*2lhW6h_^r6k6qx{adf<?^c#=obStOk`NHdb=WcbN$>n#I~55!3)2WwICVL7HR
z*&)d6@V829uLI^(a43XUlF{QvE(ANDb5)X8FGTFLBWdWrD}Od8U}mHuu>NSwJR>S2
zFPqVP_Wsk|`k>0x?WP0uirXGcBfe+XPFYGyKkiBgbBscY+=0KS{@7ZxF5kzPJY~Ye
z;u}S!wZ0~A4KctV`|Bu4`R`Ee_onL!nT4^Sn$pi=KH0Wa4=DDci7pPy)YHW6Axy(h
z%iZnIVbTD}sR=?ji}q;*={n_@u8!%OQLk@bVFIqWVYhp;rPB+zSLRozWO<)@+!UpH
z*>jF6Yb3Q_1?B|?#4W14=U!||`*BYio5<3Qgl8x#wVybeO{^&&h0K}y=X+boqiD2P
z?tnMz0iDkdyli%7#cuzDI`ZJ|hK|faB<E&oYHDaW(C1gtboTJ{T)Z}+%Y{6xvP%kN
zBTscTsgiHr;!bl7@|j!5_F*%iS9Q*gCi0WE-OsGN_l>CFb`FSIblv0fWG%{_MIzh}
z@IK@S7(0l)DnH(IdB3HQSoa?Kj!Y|)P{M1SE^nWb@J0NtjkeWjl=jIpow;P~NqwVx
z+{L+h+cQ4}@Vl9=O|iI0?_8t?(rKZV6)|jLCe|0#i0YZ!;Rpf=?&E07-R@1i!4QMp
zG7;Iqz-{BtLbb&+8Yc{G8|kGTt1h6I;B5kq#}#Nk97%r2=zF?B8gRbM%Av1$g8jo7
z==YXsa){b_VCKB@5t;00%#SCA5q=sx`#7cXuTfqLHnUkbQSnL-PT2%8URDM+4kek0
z&DeqV4Emp%k10Gwr!2ym<q8Kk#)_jFq|p~@iQiuKOAJ<7J(Rt9Dim-MrAVIJOSI3U
z$>Fi2){@G(4=6b8J&0LB;?5#*r1yrxvkZJ)h1?QeJ(ZNlYL}|1Vrr^Sc00)u1w5P&
zf$;@G5e>@lu72HIiu{GEOAPz1-+-fa+w`s~yF@7|@E?pI3<$10eMdn03g`v`(=q?Y
z(pg4D*?nzX3F!s_m97D#K|;EFfI&*SQ<Riex^w96P(YBDknSEHN<iuE&iC+N?+3rQ
zbcr+fea_z3{$1C8rp+&@p`XHZkq^Z`!00gg<QIO1Pjm>l%Kk`>G`~vXySTSR`8VYs
zm@K1HLSeGQeV6)}B<1{M*tB<sdwIS&vEyUnp2L%*F!o$s55=h4g#%wSzGE;Gv1s<|
zEqgXp_A)@#(XmFsjNU0b+4mIr`I!@?a>kFr2)>VD3GW(F0)j+%GIHk($|@NS@M~qb
zH*Z3=e*<RhJU;Td20Z*$(Yj-mzwK%`$95y)sSn3JP%VFp-s}c6I&WaMLq~)9*)-q!
zjmwaRVMn*!pH(8WV7z{IdM8eVM;{3pPUTi3W27+Ol_v@UT|vqrzB<;aVE*qH<Tfyc
zCs0GiyvfaTyQ=x`3jD>w^<uwsU?B>Fd07rR8~c(16CvzoH%+U)r&TU@zSrY&D9zlq
zRM}*qQR-y-5})4QO|E_Vn+L8PEl0&0;XVjE?0vPqWUwf46$d}WNoh-PR$SJ}yeAIt
zL*v?W6E&G4K3o3`x5O;e(iGTROp|OS>BBVvF9E!}O*F~s9MZ9k@s)1=G)jB>B}Qy~
z3nD;$4}I)hIr_Vv?)ZF3kT~@n>*-bV_4c=*-?#TS2kI<itrMaTyH;YXTj)%w@5ck*
z$o!H2n;`Zifaax!#s+y>O`gz-#_fK??eU<DE=^+6VF6ykZKeuQ8vNhUALmht;Z#jc
zFQ_P;Z90OT@&1dS`xCh{)?L1M-cJ1f1rDJl*V-KZ&6v}0ZA9x}+^I4!)lBZ^SBvB>
z`7~3tIK4fx_BQow{r+W{2+0Ix%P{KM+iBj5ooYk$iss{iyo-`iCSsNTVDGkP#1vQH
zN`6#%B~|*?wYD0JZ$EI7vHru5RlPJP>Z6$&zi2Z}D=&nfa&_e5x1P8e8^jox=l=L%
z6-?(sD7CJm+4-JDZFYkF*Dh_d7~)o!;h&Oelu%EJM~8WiT2~axJ%QXSnnf(jziQ^O
z6LVI>X$*`fN)&{$ntGB&=87;`%S*|klGQ&@qBPABUCc47_-ZR=*3_riajMKWXFK6b
znoGtZiqx>C9|fFUu+wxZd9mxR^Uq;>$6cW*AG8-K-T%hCsa0x-s;K?^YNpbJ&~l2m
z&<7W>6Q?J=`jY!oE^nW?Ja+$`Ybl<vK~Wn544`JrJLnmi{UuB^|C^bC(5}FWRlA2$
zd1RIpYQ?EG!KkER=~GbJpN91`yN|1Ja^dj$QL%?t-NAD^V|W6BkYdy1IZNe|8dK1z
zGnxen1lw40__}P2?-^EUXnhNM8Z#ZBB5S4m>W2p<PX;yrIT2L2%6QhWF;MiIRklwZ
z=YRS3$Y~icqSG5WN{$%OYsf$%m>e0AJ=}ppha{C4L6x~hB_x*zx!@TU4sZ&n7TOdJ
z-%eu2H?PtOZ&6Yvn5nji2>n7TnZEkMwpNb8M)UscB>TZz$6ay4#uYKSn5O<{fjR@h
z6oX~2Ng!|Ln^nI%yUTeFv!b}=amEaMl2H+vI0b`etP74!*l9H>62!*;_sYrNvVO15
z0)=TleDe9&HB6x&HtP;|*JpWW1XF`!{nQ4<p?G&{f|(;v9h0|D`#Gbpw}5t4CBq-{
z7F0<$6jIN;YTAPQEl-*Gva7jE?psqzKk)8c-=2)FbYO8PzqMStuF{!pB#)1aMDH9^
zBt&BFc8Wvs@y5h4smD{Y&DeB0{arE{(a-e3rxzOy;;(!bU8u2>iJN9$6t)V;S|dJM
ze^u=MV)0q-?-{v6+$`hZU+)bZvzj<LpLbjVFAd4pSf^m2yIA$P06ftYS4MgmrTw1<
zb@Q)xhn;v33tu2Xgf2Z7I$)$K`lo2{P9>Y@->SA5o<av%$(KSAaI9=AI+Z}}JxCRb
zq4ti+fMDn97L|G5O#ic`&#Z^;b~(mWKV~UnJW*@cj$*o!2AbPiQd*ah5#Mj`PfGx=
zkSyW|AUH#jIkAzAv2=Hzci{h5r^wUVZF;=gf2Yc>hSHLkcSr<#?4q8t>xUxa3FCJM
zzWRBi)En~O7jXg8dnt#5)e;{GV9bh!*&BMNNa>L6%0}d4dA4M|U*ziSp?7?^yWsFp
z=a4GDgjDA}gNYEH@hXVuOScRGHBRP6+ey(1xeQ@J4yxpAnJdZ02v2fMc{Os(D=yAX
zCJiIpJi}l`L<vdh-0<rYh05N$d@!NuQ||DE){<?MZ|#Xw8{TE%vuOiN^EZG8<2W22
z2Uw7Ex^$(;i9x=5w``)f(~;7diy7m*F_!RSoIMzgZ)8a?edb3EyZm<5=h%ta_ov0=
z2Vz&N3F@!=8WR{{B@dQWIbo#N$Ev-%X>ln!HZn0wM8DV!?TluGu4`fn1uFPay~g*Q
z&`-K%x+te-c`Pthx+tB8xGb1jvwW{@;@)WuMKZF$^5#zd>6<BAnDH;Tz^dG(-o=^;
zPi4bw7Cr`4-){0qY`nIw)4A^z6kJM<C`r!!O;-z{sMrW>l6SlonZ&#SAuJA_iRd>`
z@ZfAjD|=&1i+yx*sZVD0y(r~E^36=|&G_KT@-v^Ky}@d!EPDL>fTyn}YYcie)S~(H
zP^Rj}ln=CxeBK@Qih@Q3UoiSc!@wZwk(BEWwtcY#>hA9bx6Z@gDqGXeMFV6b=w258
z-icBklHT4~Yw-CnKW)L!jhxHrDg43yyq7Ndk<1vtDZ43Z>eX<w97RdpQg<mNV=^|Z
z)DGq67NOXWrNSfnip0T`=@Q?^0&0`dLnV$D&4x{$H8d6CgkmHM8#tZIh)P0lCxMjo
z&x16r8KwA|>=)ZwK;iu_?vo*ey+DfjdYpCsPj`@~#+=B)v=yy*oLlyxQC^kdYaFrX
zG?5vxpCu)9lt~0iOibrC#%5>3_!mDTuMMQ}+5G_B*}s(-G4{=s|E8u2UHyuE%Wb)A
zXvLmzx6$UONuEq7ae(>(sEQb625-g_9M%84OR2fRHv4;IT3ScCpc*3IyAdh^iyk|=
z?G?MP+BF#JH)(Ej%Ju#@?@<3~fGhKEujv~^+l;ohhZvnYLE>4S6pD&(Tq!rq@D`|j
z`}KAjrBbcTYxg&)m=n9em;bz&_>mvA2Ba&C=5NJfgMwvN@S0py#k^map*;={`|=t%
zh?&pB0LijbzzL7@1+Ev}cPc4*adSTU(P?G5MVw2qZyBZzNC=M!s0pC7+0+}p+*Y)c
zFZm=Wp7o)88(Yq@v|zqYj?OGgrzU@WD6_L0^h4Wng_LHmbyR>Wl-o>FhtTc!{_Ymc
z3NUrdq!823pvp885MNq~AZ5x3S#1#x^R`NnS1HqDY>U{d6|e2dXpcizX&U!Hd=zk}
zVD|=pyu<o~fOuh(gM$OvhtJ-NL?A+^NR!|uvtP&v$KK7o^V?(r7>Rz5!z<+JQ7X%i
ziMXJzS}KkfWxriX@(4#4?Lfc$lcJm1+D&l;x>8dCaoQblwsCy+JgxLSV;xtF#op1X
z&`%x5XiZUb5U}8jtuo2YacO!S6z<*qw3Rj8NV9ag(#Lpr7O#lX;v}~Dvq5lIw*FT=
zv@0t}1jeqmHS$prOXT791dEok?@R3V0J-r&%f)k3w+1>dHu(GUPwZYnMlRzrGH0E}
zPQ4WFf7H2sww>b?Zs(Wo-DWahK9kC|K@c>@{Ft_JCFzm>d6PG}?o9V8YNor}qXyML
zGZ4u4aRwAVro3Xd&B?Ft&9?3_<Cb~_poDnbEoC%&_wFt(*x!iw6m5R3#w%goH{jsl
z<Sk^3TL-3br4N$_xrz3<+>XoHRk$MgtD8VWYVZ-=0CS<+Gz<Kim4ioNm~3s+k(Xtn
zNEHqTUV#=5zgOFeu)f_Ve9l1Z4-Z@ahpoaG7lSIdeKSgZJ?!Y)>H9{x|ESXM?%Vgl
z99V9Qcjf3d-7W~-dW$g5kVEor`9>rWQx65%ALp)I$+y11&*FN5F`d6cn@pVgG`tk`
z%GXZ-y}Uhg!A{O<H%PCH$J}{+h~Swkv3joP03RQpZn<3Pp)hh*7Bj3UR7)x-udOcd
z*K`yTuTWsYN6A{B25{7x>K94;<J4DWdh?r&7F`ig)XG~<XH(h{-@kcF)OMI-_DJWQ
zg;07zf_dmQwmyFb<yv?PD?_`1`$Ev;S>L^g6HLyTY~Yr<iNu)^*wSt7c1m#WU_iky
zj%t5$W}28!L8Gvp;Fg_XN~M_YmTf=*2JQH%#GCJ9sCM?tFDEGuO(kW3AICxU=9MCs
z9`xsoxv&J!Y>e=6=*QM#jBo~}q`%;n9c5urZTXXWs^&Ql00FhwUcEZ|REWBKFT|NI
zXEGwJhD_tr<o+bVIck-nQ|FdjSZG3ZPZp|ZC!8Hf9nOoQo#sh$Hq+RFTq_@A{)5rS
zbg?9eXfDgMd6O5xf%_zLjH?qhsz9Q!TZ8<U>H~1+=ta@*A(4zt%tsCD8Tq5*74FJW
zQG#hC?=P|T6sWE`6<-%rXB7PS&5h3(Rj#S}fsTVx*p(k6%A(H$OKnykd}EvNXvm4A
z5))l<gDiNTZBKdpnVi@FknQUkGYduj4C4xkfGVU4<7W5r0&WQ^*K*>)IW>+yhK||3
zAgB$_4rj^6V=}w`0Ff?h8;bF}HKTc9<grFFj`Ii)<~ab55wYvNxuxPXQ2SW34uA&q
zx+kG@3QTE^7jxHQrEuFn1~-2Wj`*qrA`RHs*#7rz;ZN&3M{)MQVs{brte|6qCR?})
zDHKnY<*ChDI?DZ+^Z&s!@k>uvj*4jhc<nQ%$i6Yv%a`T(P%c$j@SANihR`b2)A%u3
zUn;i>RjTr0ms1VG`c;3=Ih*3Mw~vYg5#z_nV(9414m@xNj24quT3%CG=dFT<6VhkL
z>v!*c3)&MUk0l*v9Jg;aJr8=?d5Wa&3aW_tS*>v^-Rx{v3crbxkLV7E9V-_S-=<`%
z?TOXB_-Z!I$t*wKo>lsF1JE~la`dR|-M3>yITZ^#c%f`wV{5CZoe>$g#yi!#gw(&m
zI4`B$UZJ8|=ZPezJ~L;OSxWCDVd+<c^}b?Z0+QJsr%4COxL|R0wFX(I#$6kjbke3F
zF$e9xfk&6WugCmpDpdb+UzbXG>+i))w>z`X0}h&tUU@s|VS>_y?G>!4qAV}oS7GA9
zkrjWQ*Y{HIt*Jky;!!5`@k)pA)Si>r^~~93fp~|f0`tXV7KSF>ai#SyCYvAiuJ&_x
z9ECLuY3(Djh(Wckd!QP1gJWQ)#td^|x>a+MmN`wDAjA6@jAPB4v%5qj2{=M85sBW+
z?meGP_5a)0^+D6n7G;v$JZduIdwfpNIGPXrQ}CC-%+0B8xU%&c^Qyy~J8W(x`*xK{
z(1t=T&ugMU!Bp^k#_ZD|TYc#BcU0boPl4tsL315KSS|*XEv}5ly}J3X9o^Tm(XW{G
zg!cjeIrAt<co`BN&nnBS1*4^kY3QzkIllof&)c-m`O%nGSss@EiFAWkwU0)O-RC)E
zg&|ojb{kv|vZAS6X0nF$qvn3M`ryniDKD34%%9xqJoN183T%;`*nN)MDU@<Q)o7_S
z6Oc2_@xrdR%iX^Z`usHRJyFyVEu^=9I)Uy|vAyz7t5aYz1{Crnf82j7kuKaxGES+)
z^4`(!W<G}zxrCIxkgUjmD7J4k(re8!nD+eYL#H$o3LGhI<jE(m?cE%KOABO?TJ-A~
zk$?D}BJB247pAfhdRi|hv73Ld!c55+KrJ6)Eg}RETrLcyaTB53I2ja~;DrzTbMx~u
z;j|tfy@3QJsbe+J!Gc1JL;p01^>`{MkoJ3x!*)+R$7X%PqLLna+pQBpcX2$b6ARVQ
z!O+M1=z@L+Oj+W92#!{=;w;ay&s1p6qMD#^oBe4O5yXA4-?%~93_wv+jnXj#x6sED
zOPN^xKNCOwkL{>}Co(zDEgJ(F@k5k|`L)%8F;X2T0#Pl}#vHLSwFj2ih$)Jfo^R7Y
z+t^l*PG_yM{}m=161KhV&a1B{t!%q*f-o_=13aL#!~8ypU5WV4wrYMK<RALkAh_GC
z8v=awTUHI4r(54F#<W7It{(jl)}rw}7KEp9TmyCk()WEn^!zpIuBIoAoV9tB5gUfs
znZzd)%3n05PPR_dSycu8p&aGD#0LHV?Kuo=p|AqHv(^3t_3g1pK^djZ!En`J1^Efm
zxtS`n?vs}T7(^85yb0L{C3lY_Kaq<+JZnA&^_878*<#_0uv&qvEeM|UFK9t!Jx5NC
zw!N|)*D~IJ1|P34XAf_@CI3wJ6H+)9dM+y(5FT?iO7(*Od|?kzI+h&yL)tT9SLaTo
zRSK!lC|qLYd5Sg^;(oV_C)AvZJXR^;L|dlf*(TS~!!qVUs`Oh0JkxQa{?QDc<q=Qs
zo~zq#J@yRWoA%WL&uXU)2M3SRVjSW}q3qD#Erykl1nRHq@_ioJNYnv$!j{ha;-Wa3
z9g+!7L8}{L4~?8!><D*iSE7O1A53DAl5HOfFXx@Q2;@R>b#H*)mg6J7`6`Vu#%8%L
z_FkjkGw57GrP4u^+Gk>M@@l6?Z^xioD6<6C-v0?0(Vu72?Ym{?QTb$_yV~50Go`Ls
zm_(8iosHkX>n8|Lf61i(a>>i0r7oo24g(jj>)4KE+0Q&8+%gi@caQ-UzC&If5!Y=^
zg1fxI;tZ#7$om{3)TSs)A7-{FTIK+PjZ+5aeuiY~C3B~qrB){t*_1MAMn&(o2}jnA
zp3gRQDch9XqxLur_@N{BEI6kEeA(n5)OcejB1)053R`exq}8H0WXvdiFHMeoB+C6U
zal(!s)dQyays#PXP)|a;Zo)GZ3ln|6yE7dyPSkR^Ey43>%pS5_zS_v<C@B#GdVSYM
zh>*yt5`yS7OXK;I!d_UTtmsd<_}l~ZG!*jz(g8x+wwOm$C-09nWIYS+D#<|ZlBK++
zwUOygi8%}&9Of)&qtp@>@J~`Vy(TW}H{Mn6#1n`9j$r#BXF7Ua;Cp(!C*6?kds;d-
zeMWK`rhdJ8>PvpKjS?`!AeH`>o0XMh6M34-97%q^g?P4@_{Txp10^^rZF@ZV%D!rK
zR`IG?%N~GXKT%$$YStziPo{CP{1XcXwZN|$s}N6UNA`c@=s}#Pxp&IF5EzN27GG}f
zLDWLhz=J3g@Ne6HdU7Fstz~LA7A4LZ8yo&8?5e8Sa;S*L%b^u+u3!I3V&nnBYg@!Y
z=c9=@DA3o=sU3KZ<M0siBMMA&^psoXXL4fK>H}P-?`7la{$a{sVkLf1(KREE@?KB1
z>e?t)8?*A;|I@hqQ*q;UkUtVtw{oTBF|vyOgrHdBW<05M#Vl~^$dF@XG9`>k(0G8N
zUwwb<;;U?aDn8LU$Zej_L6SF~xvRk5+acb#0!>;74**5qj4COP#9-WuUZ4kBglWat
zByQcIZUuhR>_fB$+OGEN2l^JM3Li<wHA%<W3dI!=1uKu^b3Y%joDg>y=McikQ;29U
zs-D;tlsRW6@)yD}N1iHEZ-R+#RZ4Y2U<@7Y4PUe12j{NHp%YAmvR-`w;t(>PF*Eg1
zY*P6lbt5-I{2)%tIZ!ZnW9ZP(y(XU8+TMO~?AE7EPh}6(UNs_@W0iz`-gf1-UyAaG
zq-7K@_L#!qbpDOIPY^v-+;g$JEog)mp_aNjO*3)Hek=Zl;i*8U1O$h4^CGaiu&;z?
ziJ|7$oH+4MCYW47h)0D_hJFZWf9KW?CGu4Po}PuZ%|>FDoP`a-k(272WUuL7&!Juv
zAJe&(R~AsRbox!Rd5WBzWV)1vryR0c|Cx<8okm|G+Uhj>Ru`d2XJA-Dg>X~&u7Tsp
zEY6&csNMvJ)9CD*EwUVh4&kL^^O0=Z3iA|mwwl<QvKoh%$=gb54Ts9-5>yPt{xHnE
z-j(NKxEbN#<Jf%bU)>H7QcBPZ^50&d`uz8)Ns=Bq1eu9+f9;9|6}A${7p5>kpJOkS
z78wWY*m8GhkYrmaIZ<w<!9!*VqA_CN6JO&+ELY)=j{yK=RK|JJ$_XveAg)jJ?-yju
z>jra4#RNxE!_OuXl#Ywd9(P$-$;h%Txq00mc*b-ke-^}Pj%QRy<((ma&zDPbJ(Y*a
zWvh<$O3P`we)j&btG^H^<g+}~fGtx+3PdE0uR{l|cUoi+#*fnm?l`6FQI<cHXf(uq
z{I^z%+mb#?5nT4Za*h-rFDAcpdF3`^;*q(<6f&VKt90h++~ImHxp}e2T95;vOUcct
zz^dUX7-27I=jtQgp`P)CkH47X=^)c;v7}9F5WZsLHj7jT@DpPs>CtdqfPyK^WsbG7
z*-6JwPQw!Gu62WGK)+z9nSgG>U7qC`Z@>8XskhacT424VkT_Q<O}?hk5{DM60Jm6C
zb*1EI6@qSi4_!zG<+RCas@%qrSq*9=(y3JzsZzN%=H>UD&7Yh-5fv$O%wi<tRDC&k
zliW3pbd$~AZ!<PtC<dwDKZy_cP(*n)8+(8_))()^8&-?QP;bqhhBLmcXjl%1vmr5e
zevSNij4G%y86VAM87?VPtx@n=x;}J*RAWqzBczk?j|N#LrFQ?d0Oyc6w7CM<tG;_0
zZV+8Jt0S^r#h<8Dyd_(Wm%KH^4k{~S8MiSLklR82`sDsNxo!1~3zB~X*dCb0Gc<`F
zA5AUbitcI_Rb}U)NbQjLN<YBJ5L0ZLEf?2Nyo+Z$ndbN$t|ohC9A<kpetHl>!Wh9)
zl=tgyv1{?Hf8-oMw?Mxq0uCnL9;NV!3p9#z(~N?C>x|@nAN}(0@9R~`=oi4uHSfS$
zh%U=MS4hk@zt{~R^tb-M%oo9V)1e^tj$_#y(`$gfpSXn@S)^Pw##mCQmA9+r=Ii35
zyZ*9{f;alYKkBBII8E<hW~&lnMI0u>fxbI%45nVlQ%@au3TM$cbXxP$QmIKd6RXZm
zISmmbD>e=nPZPK@<}4kW$_V}hY*?+!pw$~{3qqWb({-nSKl?F4+mI5}h;G^@HpJA_
z+S#Yp`)fP+TfL~u<{jP_QI^?i5O}sx$I}UtL;AvHrWcz8)%2qP9~D~7mUADAVR6Pe
z|3;PVW?$rD$0ubp1h-UZQ3$~#XPcUy#`E-1B!5_r;5Y=omVur6`5u=O&Quw0f%ZuX
zmzh)2GbMs-;DA#?tTRJJj}3^D6=q^Mb&SVe;A@(<)01tI(Cmq>rTydlgPuFNL&#Lo
zd9|9z><AHw)3YhV*~WoBQ@RS=b_;nyUQE>OEzY)X&7Gmn;j$kK%>t%fn%V~zdSE$;
z;M5vo0;W4lP)M&XN=G-7A)aMg!RpjSAJ6s>VVm!6L$Z(2o<=qx?A1gfF~!*BMg+rW
zCt-453S)2I3B142eGT&A)pVP^0?vFr>2?#*RQqC=PHgMOGfl3n6;0bLU{Z(!ya+Lz
z54mvP@Q(@aMTXbc!Xwz5N|v~Y7?tq}+|qQYwLMsV=Ey%Oyj<%bPxZwvEzh!7<U=Tj
z-EZ#hDtl#58B3Ipp!-3pd{nqlS`edY5`ljYJv72S-;9|JJiYd1(H#C~OdU|?ePHC5
zDSS<5<T}cGq)ha4^?gUl&+Y|K5+>B|!3I8x{@gi!TKjL<3Ij*w^vKcCMtmdv>Wv+x
ztqo^lV~NLv&K?MDd`wTrC|r&ivqliyIX&|spWI(DqC!596HBS2)D;UvGE}ZcSVa=9
ze{=s(DlS$MAefs+?q)lG0>~h~wN>;p>q5j8U5qfdS$d0o{#nhbyJ?Hdmc~CmaUXu>
z@&sE__az^9t}Ag<zY<XtZbU@N&`AkpG-W;+!HqbCe&R48X79Dn5f<&H*@YtEpQlC=
zneF67YO$of4)qH~*DE*LZP)J%>)R}c(`d?<0PW!|fjd!sBaqL&DbjCF{1!1<l*aUD
zLSl)*B};<s=-5JDE8L=_GwSYq9(u;cS!y3BY;oKVx>vQuKpP}8?I?A}neSn%Yoru%
zztdfBnl5YaC->2$q)q#JJf80AzvaI3efvr(nj6l|JNYK1jzkk#cKpiz8W-lXV(7KU
zS3tFYKOtXA|8?T)(7JT-<tmk8__`I*)#w|M(7MD?;#}pu?2<vLs6`YbjV@mZh~Rag
zadv2xrm|J=@7#EI7DLpDo%AU5l<}=<NVY^fpWP*gv%Rmww(+s8PlI+!r2Sa{-FUa1
zc;-g|(^6bfFp+6?NdMD$@@_fKvBc%efY&i9su7v1GZ6DEg)84-fq_b!!00Gx9U+eZ
z7375i3hFlB*;MT}UKT<1eH;`_L%HnCPn4YP8H=wi02*cC_uKrMS`;vL0g8CV_rH+8
zZclKuUYcSD8*`?N--cT9IyU?Z{$4Lf^h8BPrH|w)oEJZpQ5ET5tj?;!_aWv>kBa;Q
z$L}ZtcLXMn>M2u2<NAv@5>1v2fZ%w{pR3_C$|FcSUTuFxqr_kT*-Pcu45=L8{W`F^
zgd>*7<-ne!(I|OwHNsb%>6VrfJE*-uCEF6qkDgeYghVVbxKZ%QE;U(wd`H`+;!<@I
z?AKt{^cPcpS2Iv^156G1oleY~Ha;rfQFXfF-om64Y`a3}5XG6-skm;>K5_bdEnW2G
zEHxswm~HlOp1VQg$rnj^9c23V<Z5D3Zv0v=d?q9SmEcsjc%itFA3xd;S5Tt|j-`1H
zQ&cPG5)9lxKISE739`<nmnzgW4^?FR20uFk#}#I7KQS)L%HXsIS>oTsYiVwamwj2H
zR|5ATdW`+6F<X<x`S8Bt#-(-a7diSM3l3;q3R@KaFeCYimAAFCWhi_!5gU?YOw`>>
zFX>w_V65dfA-HlT55;143PT9obF>)!1GGUHR3vdSL5904zi&`Hx9LHHdJ+XQb^S{6
z>E^_>{c4lX!e100`UG)*u@c;TXfXM{W$9J;9BhHzLI($x+k`FH<7I910(fnti;2n{
z?g4Yot@@q2k%o%G#r$WVchNIa!hzTQa;12w@oT2s(I%mBbaLy%(vDKI(=*5*@eJk?
zPro)=YxgBKRCOwbl05PVZsrtJ-MR=2VEQs5Zr@avmJ0xn929IjxrHL>Cd_p7-}pF<
zMicW{+5qyVbinpQzc8D2>EdYvyjnHbUFwR4OtEPV=!FR|<i<x7YNUWU+1&G>6uUQP
zoEE`!o%vEi4B?=N@gQ$XNW0cR7*oQj%$#I0j^Zeqv-C3;!3*1Fwjd!MT#csfBc>@V
ztg2G4;mCOFUU6C8pKSN=F}Rxk_^HL(+D}sN-1&Ego?C^F;Q&sp4z?G{duL>f+Oxkw
zTjD!XKR-VH)k$tbo}?)_?b@oO3y&o?#b5NDPYcuEUI7?inQKpt+5Fs`8g^&Xy`J`1
zKLLi99^{+x-dt_o0&<wGqC)O0D>uu4W`KdNE;H>M=#KGKMb0#bvP8W4Zd5ybA!13h
zDSJ5dwknoNv*JVy=1y{z7=?O^fJcL%a+52h)(BZML`lvJA~hZq+fh$=P*dL6$oS|_
zvG&NkRIiS;O39N{-}2@{gjrwv=Ik?^roXfA^9yNe9iyGLhi!|h2#+~_Yd40RvnL7P
zx&9)5ZM%CVLL4n}KGTeT15Aq>-swY!7Qh_qF<W6rcA2xut5;kxy)sugk&&%Mh8)x-
zqbQM`=VJhTd(*w+cCIt)u6g{MQbGUaq-ov)Il9Ei1X@04or~Vy6-ERrVw%8yE<-AR
z2s_Q5c&1P@vCBTCHW4PruClz!tLkK17#7^lnc`Bd%vSK{gL_+PE?i5YJX(I{eYw^V
zsya(>SgMA^7LRu6q+F0so=%b@b6rky<v8JLlV`9xDgYFT9BPi$jnX4dT?x5hMat?1
zy)ViXih@9F4N(G~`H1ZJEd~+eVZqvy+>=Caf#rm!bcLwNM4C?Wy%`x>sjQo?!c3fG
zS|cfcH8rX2BebV|OHMAmy-H+eSa)`?*fxp69)5teJp@y{@^cb}9vcyOVz+AON{S~h
zG0)D<OlN_H3WSKDdmzcZL^K8TzvzSCr`XhGtCGd<_&|{J;C=9kvIJghUd@hd_U@Kg
zn+qG1mfPf{Rt9A3PGaPNra5q>g{&sWN4JxGE4p5!gK36Ge<LB(dGw9ypocGKI}I=|
zUwm30N_{mvI<X~0Wnh6huA4hHb|j5^g_vTR-pJ=|my~!JOwK}lIfzznu^$cv3GcXh
zgzsHTu7%I?sjvqi7GnqP`?dMG)7I}27bUiKA1x+|iAJDg4KTdOa2_E)6au`1iq{f2
zVl%kCz6=m!ZdGJ(aNo_;U3ng<Olu*?2)o579=4oc5%yZlx3ZUW8K%{E6_&*4$70^9
z{N_|<1XM3Pd2fB-T^6>gQlS@yMAj3a3^ZE%$TQU=%C>>nW{lj5y~^tR;lJ;?5IO5(
z$n;p(z{e6JOH5>o7aMf8!0pk*e;M#~>qWrG4>DU?0NI4dq7N4hMAA8B{s(7!lr<C+
z$MfZ|(Efcr27{$H`SPQubeYVy&~@4VSAfd37_WoCqFTW63nt<(Z_@5lV*tol$4vR=
z7@A#GH8|dS)<Nzk9`qEubCt;UsK|2DYXx*G!x1U{{BR8U^*~KYxk4AvFHt;ytwYUt
zmMU;X2oDe(g`gIxq$efs;fC;96@7&3rHfAC2OGjHN-rr{BrsvpxH6=t7%^-%RfxF7
zS24ckXlCn4SnI>-Ar^y4xtfpuX*2|Pm@NU?PKPqHb*i5AqiER?b^;~>#5NXQ`;AtM
z8E#lO9!LfGvpPYtW(>mZ;)`?BG*UNBt?Qo$*uo`_2u#d}GEB&de#FL=%$1gx|4(GE
zZyJ^q7ZnBb*?dT9{&C3vuE(MlTY9R(*i{O376Zp;iglLl$b~vj*^7k96}932l!IoE
zQ;YJ8cZ$_8i22}N6QraTQGR6e+C>lkOgy#T2zQG?VCac5m2%zG?3HH=hC~4it<dJf
z?S`1-cPk+YdgPZ*#E*PzNKH2I^-J{JEPY<V8Lv+gdP4k7MMIP`RoUiX{_nnQ#4o0D
zF7pSF?i4-zdb3c|-%xrpaUO1(ii+J?LT6E}8F>!`mxr)E_fn~Z(MJU4(x9OK6L%ZM
z$f}f@Mz9=X=Yj^K(cbe+RV%}~L2PK>Y8<)1i>RIokf&Vs$tu_1?~I@a-RwG*PH~Fe
za-j!QjT$wPPyM-{K?v^wf3*ydo!_9131h8jF#WxCrj*Q163p(>At_ddpwoThp<89z
z{iQw?bDK{1l~=iC8n5p5kM?!fj+DH+oa(2BBNO{SOY%vZL`~9<{dliP1!rIMtrB!{
zUgg5`Sm1gpEpd+|2OJsBCSN|-xA6tVx75X^pYIByLdIi>JwP&GI20mC4lFawqS1yM
zpUTqW=|lES=`Jd@LDJ*<xuezCtSzjlxt=wi!U$#8)%%9-zd(?K&W1bY=wibu-2fbi
zvGfIKV2WgOkAA}Cy7roSrI1}w;fy{_w6_DAK`bo<_XP?0i1Pj86gMhzIQ{vGzhM?p
z&Upg(gab63+lxZ)-aEETIeweb2tc{VeY^Vwe9X_=jqy9fyiwkim6fTDsJ)j7HW|Rj
zkWevj9X3$Iu*E-D4E&u;KUDfMvkJ|#X!9Sa{{tSQTdr2YppQcjt4@=N0NyUH+G-o<
ziJI+qe?_Khu4ofp@Rtd5U(LnVj@tE9O5qo2&u5>m$0Z@5z8JyWDlz6{MIW>1z!0wB
z(t*8R&-pl4aOeBaV24v*0S`Is+IJQe%(ZWRW5c4Y8TN#t23Hm-S=w`G&A$Km!}I=I
zFCzSO-Hcvw;_Of1wUi=LvCyrzWQ>>&tT8vfqWISD;l`+hWB`)mCfWHkT-bcz^JTLc
z-79!_<#>;@2*rhu*-oR4QF2+d=qF`iMXHzJ4q8m;_#$KVmls6u2WQxP1p2G-W8@WY
zZAOkl1St0Lut?copI5+N|1tddF*pDm)X8&Y+Z%izD&3h%0$DPfB0ozH`;FnH^ZVe4
zghjeAvZB5E_(kAr&Wj^&DfOGAr02vh%H9{Jm78aMUJm;XL$Y~IK7V%<<JWbuL|UYP
zEVz`=`LR(dPc%Vygp?vZ&q+uxn2>1r3jCM=8_rEN;GLr)x-#96EJ37qi9SYyT9&9?
zUjtEvHu9`bpjKa}(W&r|@G}r#)%I8Enhf{HH@samA=*DBf|(hbGDN&=I?#)8`ANSj
z8eIF%cZ++g1WC^9?kl_v-rkrbM^~qEp*4^eOO6eF<W|%owIs#vk78g&sQ3HsZ9DFF
z0P-*_z||nmyq<$S`wfuxHzhu^4WCxNSSFVyJp>Y|iVI5RE!w{+ZdszjMXIJTcse(&
zpo<u1MfEurh@QE?OGM+sPVmt?ex&@o4f^yCB6&u{1;sZ|z)LA$=zSyu=ed$+<&cdn
zh_wvhJ>toT;w3NAKSI4-;d_>O4=V|@*ZV-?A8H>Oj-~(><#@?DH@o1LJ?iI|=?9wL
zsr5^~zaY$qtG3zUZNTocEX_<o{~fl+5w;f{9sRWS*VTB8FTH-umUPdWlSM711+4jW
zLg~?xt2lzfqP^cb-#j83>|!&y<2=@i3x4p6^7@$QLQU@$(ZU~ARoO<#*1^o~oBUwS
z;G@l~CyDKSU>vK(LZT2rZzgh>29=&-vahB|n(ID~sHmuDzLf>!W6ClSZct_#K#E3l
z3AeGg=ptmKzV&*#YxmF~Y#CD@rL&aNKDHyw?o_KR-9ujPZ6iOBDOA*a@Hh*8oSM4P
z(vr|mRJ%TcASBEoF3XaxpWvSIY$wk;zzEXrNJ^E6>ni1rgAswrOcDJBZ^vMvWQ^QN
zPcG*5_4wl^5W>@=J(+{JCqHEa^zwsXM<E_Bm^)Wr+M>y}(P>u8j1Pm4Uq|{(iDV!$
zKkU>=T6A=P;hT1)EEIaJJgi~4^Iy5g><1qoA8_DuDxYhnusK=-6eWMkDPF&rM}n=7
zBTUz^hc*O9<m>6y>8kY2qZ9Q%`$xc_;Wk9<md35k(5QhB`I<{H@dd~}OcY%>p>J32
z$ktn7Giuc*doFT!I%%4zIh4XBRl_hKi%1x}o2#`W`cT}P3%M6ZJp^bY-TA=)Yb9}Q
z^4f2=d57flR#b!cRf|($ifk>8pL8>urVKy;w-X4y^%wnB;}w(>vzRk3|3f@PD1qU*
zL#R0>XDKb%WQR?09i6NE14NiW5&vNH4<NR#&*SBP-cn1cz&u$!^Wa2h<nPoOj02Ga
z)ks3}p2j&#Cug@z@zjT$Wxaaaz!nLL4#P{Au!p`eXv~`sitJgQX-6tBj%)yKkwb8C
z8R)ZH4(|Z(2<y-(rih)Yh|A>B@gPUUVmp{$x8UzTD&LbMudEV++du*w%{IRS?MZ!W
zsYgImGc=W3rrC}6>hNsZ2*lhbjv)|%8uXJW{U9R(#hY@oplez&Az#RIE#)gF2t&Dg
zgQIMMV~cGGUM%GLCexH#nPZ#f4BwspFZLkKEYE>*p~&sz2T<Kt%Dz>w{L;u&UjHEl
zY|Osmg4M?D{_&nvWWoKo*B=gCTt*9x@<!BjD2z^{o}JYU=4b~Nw6L={?;DBvjm@)|
zcNwmOue3*w%13jow_xsa{=+XeCMIf^JUBZ^G$TGRa%{l!vKm0lk29f4ZWwa-VuIaq
zwZ-87^7kIKm44}yQcIWX(=D3=(nW$7nY)ZMeNeDM;4vc+`YQePX1*D-Db;_*frVxE
zs>&)wx60-f@9Wt)<}Jt=aa{v)M_3SVV)8CEMP;}4FBZKjZG%KD2GrLj1i3s<H=FH~
zhlmGp((p$^mXK<-m8U4eOj0@srjFfI%reg?4dOkts;oIn2EFNCkzCNRY#&`01%0k^
zy~?Ruxwuisjs}t>O*~e=qwmQ};0<^n19J>SGzxFx8?|uZJF*2}0$fH5RHT|Hd<m0%
za`kdGt*J@q(fQQ?9wQmvhsSa)hR5!|*Cqcm-U5%7&4JOs8ma8u?p1MA5WRA9`NLZ5
zI`9*mCN{+_n;Ni}dTcCGFA(W01F|z4JJOrr$rv7VBM<KX<K_?}a*j_lsF(}Y^)}xx
zdtUKtzo~kAQgtcc#W2>6((m@@d(>wo)MMPoctfXF1I!w*rcz&;yqvy#9Bf9JM)eQk
zaUYbf9YOzDa`6shT&+kw|LeGMyN!`pR1~@1QenPM^o}NyID1==(^~S?8iyaV1~DG6
z$<R22Y(HflJ_4a~L`<>>(F4%9D82RFovC^>HDQ{5dCXMBqzVecA`R~?*boMJlO<s`
zlSPF-Y+4s?#-C!?wC)La%&1!0{LF-PudJ9hi9RU7<e=%W@Y|Gb%Dc)qHH!ZvD<=z;
zJ2|>NZ@e4P+pgBrWzv<+$r#l)Cs%o0{b-hw!FLP2{h*vB+-Q9xL*y4UvOdTU7mTyp
zY~GrLzMJG_VTx`1D4O#H6TFl`Li$nP9WRR+@v$RMN>}NB$T~{?tRl3PNWyFXNQSd+
zdx>oL>b`qr^AJxN0vkOwt+JhB8LuuLQ&+vtWsv(<zvzkezln?aFCf?F3V#h0NZrOk
zbc-%T9sxNa00<X(V?&v5wauB;nnW9!U`2s1ay0y1VoMGVg-rh2H1&Tg3?`V!U`a%_
zTjd|E)laKFNH1y_K@TpFG??47Xg>la#b+_mZT|epux^-h=r8p_GC+-W%stgfpAX#T
zVRLS?z^s8aO94)l>bnz+R4e+L)Vlfth+2Jvbyg!8VH#hVT2h_5cTRGF9W##GtiH6c
zr&ZOlhYY5FOOhS>m68A*m8+=<WAN8|&Yb|izflgXKNl6N-j{%v-m!1|q+QZy4a}7M
z!u!tiWFn+rkYOAGpHui7Ng&4MZIUYcq0Psf1KgT8n6vv$16{7{em@v}JrjMnb&j)9
zQ??w<$~<pf8+k)b9QtxirrX(Md$SM@mVj`L*P34R4xgmS&)D&yKMR1M9oxK-?xpJF
z+zuSUC_zLDzh7OF1w^nW{$CL1);Ca4r%5ff=&kZa4cygVxUo(w&{H^;X!5yyO8**@
zTldljGn3p(aM@TCom^W8l~?=-njycQiFpGXj9ivmf!p+ROAuZp(*2lS2rz5)B)suE
zM<02A$ZdEaxB4xY4Yv@kz6ub=7XC1m?YGeV6U7tG@IRAyo1k*eb`3uUvptT;FZ)2G
zDTxBvjxRwnFg$em065Ca-p*dt{X~t|q7v-MXFdeWU6LQ7@RZm#Zz}V(3l<{o)Mc+2
z$j&??5h>DT>Wzq*4cN@2xgK>+G$CuS$%GW67jbRd{q>H_6?xdc{ag!fqSNtmPEz1i
z8gVWzdLi!J^VM%KLR65X1@AAK*Qm&~Cl6+p^N(|0VM_z6b0a_={tFgyI)xOqcTkxv
zHQ10Lvj$j$BCBhw?)c;H<oVfW^ZxhX4y}*SP<pE-dbLWl3$7LrIj2TP{Hb{H-2Ctu
zyq)2p7vMBtS9MBDPp<;S3?}lgbjMcR40x>F9;mdQ)@FYsZd#y`_;3lfQG|nhJ23(K
z5-uIntz&6L_gh~OXg(Js`j~;2<+M!Q&I={muu*%b{&oCCM5s&Lk7T>;?v*5CA0N>h
z4kCCsbnk@{>cC&-dQYuja%GOg?7LknTI4A2_li|}x?DNb66}iC8WN&ng@$v|Ph){v
zA;vLZl%w3LmgKlzP4SxiJ-2eB$(xkcC)C$7`7OUzj7tL5pHlF;CbBqnA7CZ2lkw_B
z@@gYb(Q!@aq;mCs@%=u9RJf@jG6&}FxmxmF56h%&_ymqDbUEZ&$Cx?1+_r)jpcZSj
zRzLhIK5}PXBBy>M`4b??g)NRCW+c5uqy!<OKWkPkpnoakd6W&7b5Xvf{ExK9BT*08
z7N;n1?plN7bl@P!+w$ZOP;D6g$%r$)Gw1B#R7MH^GO>|-J+cwu6h1=)6fEn-I-g;_
zX}J9};!I#poo*YMKWBEXMtEB{FuaCk1!{v`oFbAlXi~Wdoc0LufF2Zu%Dhk?0cVph
zTKO^|OE`W4J`0Ym6Yysd8I-Dg+J#TZ_edtceEjju@^PFaU#^+1#97<8bLNqg49B=E
z$jzT+7Oc!aNOxPRk2Y+Da*<OTCG(bG*cx+_b@!8Cc!vK;cB(N9Zqi!CvTZ!;CguJ5
z?E}DL#Mao-x;P6lT_<VsMOX|5(bke^5^T8Nl4i<+slWqR2smhT?+={Ds}}Qc6@^sz
z6)jcirF5y4F{d(%P^Jj&GIO90*bkGF)Tauwz)9m<m#$Sq2;B@-VkzW{3!wNxQmW9J
z=6=lYb>#l<(E~jZ1%fW+dRxQu6PMlXWL1OJOMKijg4OCfZ?HKgVSf5C+r(?Gn13pN
zXE<H~-HUr90aawGD5<Gov_Ab1cpMGj8nN~sKtcF4slrd<#ZM+%%e;<Q=c57{MoI6o
z&wF}gfgmO5F)!+R!MyHxZJbjf4UTZT#X9I3?Unt8uq~nlR04BQZe>A<7ox<N<uMqK
zftpkZCnw>eCdvfXZQ5L?MC!ROg^i;*<2k(t5;05fH98E}AMOtxr03oyyxf!yljduD
zMH-=yz!-M!x1}>Sg%q$jBWFK29RU)jcsilp*heXPY&}5=GcQdEZk!0tubiBdazjOG
zO*QF=QEz~WRd08}IQS_qO-a&U8FnU{5Jg8vj)5?5-|R6i>3K&-M>%G3hDdM-DjHI)
zq?kqb@0|BWl0O!jq>OyY`P@U|xrCph3a%I#cjudg%5GUosml0te`FRBj%_1gfEo+@
zV!QY2mJ;MyhA;bOvZ6UC4s+C}{gRXtBVIL9+nwfu*>!^0dDcuf#Flcn-u303%(r}K
zY6VrJ`S%~j-eXT9`d<;9uTgKG%rZg*fkdVNQZ^XP1N%9H{ysWQ_f{e4We94RZn*45
zfD_TVIc`QJgyRTgK1cMm=;4G~)O}b8Pf!(7fTOHO<VOo(0;p9=lw8ose6dcY92<0N
zIn{?CRm{CR(1)6=<oJ5(B!GXeIKaH$=V72-yqYGEbAoS%mU$T`q$|a=!Ds_>n*&<w
ztvT6?odJ|QbPy%n0o<p6R~7&%$P<h8P4&p-hG;{2J{nlR<P8VD>}<w?8rTAEcm#P7
z75E~79H%KX3|FO^nc_$()4R?_(T)9n{UulGdrio;E0|v`gIqw4^9s&R6mF|#pt;HP
zqX~{U&e!1-%8LGJAFOrxhv#=GwaCN(cudKuB(9p8n$UwA@3Zg8%z^)y$(c8>gj5A=
zN8wl>Dw$@$;=3peuP3Zqfl6$VZ?wPVc8T^=h*j-nMkaMZIltf<T?=WYITWLtF^j9?
z*m7L}&X2@PTk+X3a|#$Je9UHhgV5uGVf_wTDF2A&J!VrAMk|~F_;#>ERoV5l6l9C-
z&_v}V+wJrdw26L~i)uvjZ4-N5$bqH~CBuG};kjLkO2F?DsgWFJD#836L=%a!H?GF#
z?IH8rQWGgOzPWDP$Rc92D>VO4e#7#L+ho|v$E7Mqo{xNgohbZVL(f-E4h?~lx;ozF
z+xk9Jqm;^ncP-t8&CB_$G4BJK?Sv>cj}CZNle~XzReygoz2n*hde*wi4_)*>Ju_Bi
z5TY4#vQQ=)0_@L+5;skerI&@gE201nHo12LK@Y~Daq3>An35FRejL0VGOJ-d>IXb^
z+P8VmMD&VgF{+vh<4UH>ED=JU%qn22FOj_)pABO8_{U)KUb>f9CY9~xI%-y>%J}DN
z@!`-{PJJ|}9<LOC6?!f_N`X0NJ-KS~h)J027(ll5e)L+9g>=&7o%NSjfFuJVWq9>>
zxBmcD&A5KdN1*><?CL)p{y(u7y}5lNVVjAp+LAbr?sQs6!hyWV5m1Wjs1qa+xJd)7
z$2N<ez$~@laDxqlAhlB$E>yzOHc%s<bq=awB(Uyql~-_CZWgYWw8Y-TKeLQtH~aO=
z`#;RJ2yiLLBc3FDvu&JO#z_QUh+&Nfu(Vs{FkbTv>Q33oMJ*CTU1KOZ`}D?&BggQn
z9^KzMHsa<7HL`R?mr*5knxWS^7G15cq($OJiIUABLK5Uii0KBf&Vrgj^sy-|Dc(?!
zEE-cY=UezB$3T<71byx|73+`~?R{l7wrX9Z@K+<4I{?!}PuW+<{EOno)9&zFkP50|
zuC{|185wC7TLi|tuaD*<+e7RD)MP_X<}C0aHH$v$TusOPxi$ipU0o7}uZ7GD?fpWN
z+G6R)!Yagw8P^n_C}s5#fwT8bK~Ez*KR{DBpBAD|#i{?IAglu7+S0oPp>O!#joJ(g
z{@$%Ji4pAiBP^@o%-6q$W`p0UzO{9w-(6Y%94FTAupck|nQ~{-%wN>wJAJz`SFSF1
zI6|>;DwA8!2zNOv!w<`WUM|~0h~^G-&}?~Wje(0y+i^ZU<W}ggk;fmnc-|_-o~JB6
z|Eyn)K7;nc97#6rZLMGyOtu{~%h~*2Wj7et&}H@>z}-i(S_Mgn<i4=(Z~-|Q5MsEi
zjQTYoG?*NF<yrc^`GbwHs1}}U>5o*jOU%C-?;_&d1AlSO2IbZY&5;OX30Hv?q6F-1
zMT6wLIF+<(`6R;lk<ggEm(oqXLZAo<(Fo3?Y`dPS`~eif6`&}y%hy0vVzJooa>g;J
z3mr=d%F6j)Os1>0F4f_>xpPFmFwk^iQrz_h=2)gpi=!ngPYI;&FX+Daga^KHU!&bj
zj4jp9t(vtLI2q#M^0{`o-E;IKa_tLH880TH)yvMy20b9_V+N8Ng=IKqqR-6U`FAla
z5Jgh;QS$*btPo|jhy}_R#wm`b&AetA69xPv9^R^kN!7O`h(VJ_wm(7Piq|54CEj+E
zxj|gr(`lQ781C~fbk37O(@+f^;sm9@QykY%AYZ&h@)gh@;VKQht8Z5u!%O6x7e-Pc
z;q!S{gE?`O5LY8qbHiO+2F(y4SbKuu?L<^a>_d3n{b4W35t8Q&{5|MLQcT{qAClqy
zI7a(M95#Gkh(9GkM?(6w!RAd2BMl0l@ZBdKaM^F|ZgYyf@BZsDyl^5pKV6K=cdp9w
z*EaMh>>EfGbS|=vkt<{b8JCHDYPk@`Fl>+99pcoylG{g@*`K)nh3x-|H?3%(yyT1|
za1|NX-vME*6DaH!KgX<&K}#3&k(`5<cU(Nx1X09cG=`!t2m=$<ydOd9N$)?jUmmR-
zf$~SgTW)<j<16E5>*FtshX7T8erk+Wb9*F0v_8^Vr`Zphqh}~R=B;rFN?DZHvX7dO
zF^s#NRS{L{n1G#KM{G~1o(Q~RF_yKUv*eAEvEv`sB6VLEo|T3s(6gMzKifBcBJz~O
zR_4yh=y>7a=h#@kNa4wZ6G%Y}P5omQBT`50>Jaorb^Vdit%6G`5&xWx?8JjUj2F5^
zf&%@=i~iTl><t!6p@HZ=r;^!lukJg=+^m#|Y}<u`a(bXml5CpZyujgQq$|1^{-O0Q
ze|q+PQjZ7sX=CGd*|)$F2%~wkCi)3Oa6c1V3rgz)X$IZ3DNTvyIuL5{eaT#wvnvlM
z2fRI2qWHVabz4Q4F-b7k7~z7_2~pknmw;Bv*6B0a`u=l-kerYhKX-T7UjL>PO@y3@
z-LRPqm@(@5i0SDjOdrF$je%7`#=|~s)pLg}9&#08=*}Gt;^+s1xVTkV)Yj2FEY7sA
ze|nzTj%y_`8V>j@N2=oy{zx@-KR!e0t!KA_zH_kPLwu1a?ld>US<I&zqb=?#44;$p
z?xxe+J_S}=LXi&lZxo46Ev7y+&;6&3{0v?LtJgA0XyiMHOynySC7kP{oTe4<1j*F;
z*uo_87Hw-S?VaoUs&H0@_(1AdoP<5oQa0JU14nM-zy5}Ah6;$RTE;^^z2G@*lJ^mj
z!5&aY0N=7v71&V%7h3g%u9-FlcD-Q({#=Ds*dRxf^6Y0RSM=ier?%~KL~BOgN2td*
zQ`5;Of?<T*NnSv44_`Ng&eA<%H)hjY{E<z#aR-F8l0`#Egx4;u6245tnmRSs1W6o5
zxV7RNuHY{ij#EDL(Zvu~zNcY&(+7|Beqml23<!qM@4TL)n}3qHxk{uH^win)8#w>n
zc^-9>>dx?>=6Be8RCm~i)qYmE<a6>8S>PjiS0-qaWl<6$ODDmQ<;xYHe=ejD@bPzo
z`V5SP--kvm+@Mj?^#VR1jTJT|g(pHRN8zfuE2msO7DlY1tY-V3X4uiGAGGNy=dI!b
zfiJbau?=+mDdR&Iu}s)3i)e@U35oVQV>>>PP+rlw2gX7Y#8`gQG2b0=o$_aVWP|Y<
zF}fMl%tb5-EUHCWnz_C{PSQ84Ogbplm%v>hF8u|_`iK@*Y*@9LITzsa1-}JVJIH^I
z?PID^F74#)S7G~LF)lkM8h7=YZDC<S68|*-m9o8$!&2ngYwWxmef)~@TU#li!bgFj
zd1X(9a7FJg&8px?F5a+p2^yRejZ_+qEv7`7=Mf$!>tPHpogCx`Ia7uC>s;$%Ug|F4
z|6=n<yLt(h&lX4+HRJt=Z;8T5nE;FM>OW|xAVg`c2qTaoo-u0pAoKk%Vw^{#LhkIR
zxv=&~ho~Tu1e>vSu3_V<y<({~C`<`$RCveLH}%-A<Pw!zt%DP7L&YxtB7br!)Y?Rg
z^3~x{5f}@jOO(Lx4~DuN*Zug0T3KK58y+t7$#3}^rAl{9#q&X_xOfzw0=lf_*1m>I
z&vA((nyN(WNT&&LE^5P{**`|V^yKua(qQ!<^vp8JHb2KDI|Zp4LL6#Rn3J(#Won3B
zux;h_rkTc`Gjr`=-ckhB-KHFws#-9Tg_N2t9}nl|j8?!Ej|Da{Yz7bI`MGlLA=<5r
zVKy~>e9&FKUim)Kb_Oa5XHyw2k)5ou1r~-^$+J0<a8KL(W7*xbkwA{OZ!Gj<r1y+1
zZ60GVGh8MV^0e&!e>q0D=_Zd3q&qT?u%=Hfb9s06XHZ`@w?%lU3ok!^ghh!>)PNqd
z;$Q)>FyVpbH`yM;LgK^+;JaYn-hD)DYs?^e^wWL}=r<q9;<X|w-}2LuVM=1DoZgqr
zJvW2r>1)jsCu;ltT_!ZjJA<TEHi#IOQvDxMZygp@xU~-(G$P$d3(|-vQiF7e3|+z?
zC84C`z<|;{A}!q|DGdWkBS?2j$<Q^!z`)42Iqx~Y@4C47f6v~}de*wv9cgOXRKBSN
z3yCSA({L@SZs*TA<#lf|!0@kSj;p!?#eY&h@Pp(X5Q!?9LOhINj0}@Yp9Y2w1F!w!
zbwn50pGZrQ1NKXVH~PoB#M=stGibsK>Z=<Jq1VeXd{`Es0mhx*YByX?&dC8-nY*xn
zYmEw(d=g9Wkl5s^Q%STw4>XD3N&8`qB@mLRQ&g6LUT*`;*g{DW_omP~9?1>LrT&E^
zkxr&rz{Y;oNxxAEG0T_a2KX<oi78+5CB@NEeelQ(kt?)I=+76?Zqe+0NF$l+eyhwu
zPq>goc^5D-^cmx7>}UYtMSA(hzf_9A&wQQBtHbssf0KKdi+=l=ZH4~^SdG%yNb`<T
zPlq6OeGC^t9!9aSs$HiExI^Y6xbvJZh{=;Fe!#C@`n}>wwDvT}i;K;I3+~ujS{(4^
z=nw0D`>GwrrwkY)OIv#VDk#p&Q>Ytdj{B@S2oTSNd484t%^Da2mLZQcx+d{FT`U0?
zJ~w{N<QX7zJ!0@(A}DuE%g!5c=piZzRlLV_JhH>&;pJT_>{ul}_TApIy1C_85iV3w
zeWPqAL^06x7)C)EqZY4$*FarS`imGVR#2G8xpT3BngG)W0zBkX;9CIL))_=Q=q!qQ
z31oFyD;eADv{hXVmG1Me%MarE-*KdFCPH&p)^)CYVE&Idc39QR!Qb-X5m@`M7aq7K
zRB|I*#db>ijKGsSGzyswRgamft$zeLy!zdlp#rf;3$hqq%i?Le(aQlCZsTlgr|84m
zu#^EFBTeX|2bpaFVFDYzFVFyj_7d?Y4JC|9#3>|jI0ns^=?)^^RW=hih6QQf(<6eK
zY59Gxf1F$VHrr}S`w_~4oq>|b@y%POCq>V`aAab&HuxMx6df*zut&;a9{yr|c~Y0}
z2lNS5`SvEi1oK&x-~})Kd2%84nV8B}^m9pZ%X=np;3m`l<&1wUI?hYq@t(&IT{2==
zKSyXMr>O|%8TYv&M@Wx?wS8_0Osz)C_!~7Nz1Cyl+E>19{VboW!Z#n*X7<k_eZJn)
zjscjkQ*@eM4x`#e<W(~QS*o>`klK_N&2GRKUz9N^V5Ih5`_L|o@!M_3+)HQ*Q=%>m
zm`$vj6w^L#m}<QZ*R-%751v=n^(ruZGXLm<!EbGXo4w%pjxbzWev-P>tmGdvNhp`I
z0so8gu|I#x9&0n#k{PQx5EniDQ$9T8c{WKP{93KeF352|?Sxh;F<zcWa9c@`QS4W#
zdt$9{G!$E2Y~XMC4e6j*KOCYhT0$|UGxo~il>1k3T?hw0-wQT2#YT<@<Vt^PbrTCM
zYV6^&XJ-;EuQ;%mB+wSYtON@57}KRIDuB3Tnn^UjOs6hAcPn9-%)-P=J5D=Ow*E28
zeQlC@rh8vg$3ya~(sG5LZoJQWb&XoRm4pRgKu62oZVnONU-sG`%B_C2X!{M2C54D-
zwjV1?sleu|s)0nTcFWP-%X@-6s_#@gbuAJ;*OnJ~_%r4&hddni@)5I<VFA1)OkA?Z
zzg@OoC+ca*gK}y=0qh9_)W-+XUhg5S!Rl}`LiIKCXT2}OG^ddvv1aS)Lxa(B6dw&;
zI%|i28_&)KQJU)Mrw#_CD|t|?mKdifdp}oNlLOtRpj<m%c$5c3FT}g)jDdFvgX5j>
z5)O;e2`>5ScO+l3f5m@MC`8Mn`N2`tdSlj@DIgHQoM<tqWx11nu1>SsjGIP}0EOrm
z-EL7AEzNyTAoka0ZB%Iu*vy9kn}(*@);y8`pdq;l6EmtWNtv@vF+T!U#-xsgi9;1>
z<I3e@!URXoC;NPtD)pXoySbc2(@(7osS6#-4)GwRe2yrU_g)R_>v3p@z6m%H<4@b*
z#lHg77xpU(R#`hOrE%MUzL|?o$TS4t)eCIFy9&Y_2R6k>sl43<q#xdphFyOZt0DT<
zDTA`cj~&K7yPgpqHzsB0|K9(IZyGww^O%hmeY<cLF8oP!6VsjS%Fa#aCd&|`w7BV+
zA5DQgLC+R^WNC&I!5)Wow~2ja%M_%@%Mv4eM7Hwg$K+>5p8QhqHFou8go@v|8A)p^
zqGfkV^1CswB>i|qLobF4XqIkJCF|Z2k*=!L8@C9$LRn1<q#qiPUv^4vWdyGd0TUz~
zC|pMM+nCZE%wA&$PH3_p@U@6SkPeByD4%K{eed-Cx5g}C&`%c}f7$mLH6|9@_eB;i
zCB$=wuaoJeppth3+S)vQ=RF_M#>J@D$}`qz_5nohhxe4-@s&TR4+-Q(JZ|>L28K{c
zKNcQY@koU^26w1zTF)MU*AOq0CaS1?o?|a82kfeJKa_dC6?A?LPmvJ$$p9;yRngQQ
zjHXzntkH@;OVtKbt)2Mz;;Jdm#|CRWxhOBHt;X5Kgj1^@@HJE*_Z5az>BJ5a#Hn^V
zrG_dUsruT*04I2fD|fn!$-i5`gVM)MZ%)7dsuVu=UKjwyY;T<VKLT=vFy6d;?O(Nv
zPeawoluHI2Zc-h)U4w^?`&~Z%k|HCxt3fQqmP1VJerUw(F|2%>K5W1a278^7c@K&G
zti{){b&4tN6X7Sc|27J*Q;6+<yE3`J3S67J!;4n?UtS6iIJUZXi@nC6%utPm4<Foc
z_PhYk(hM28i9our!A+{p%=u;5(N97j00CCsVOMr79gZ1CDD5Ct;J1WmD&W)Q*G_19
z<dGDm3Q^og!lm01QaE7`po2QtS56<foj${8&SH;7<xDNb8cN{ye>TjHlcl#3Itdf*
zIrFwG3VZmOx%w=*UdS9_5o>Ph{iL}foVZUBB}PivP2!M{epsiCA_G(AZR@*G&2R5$
zihOHt=hS-33UvtqF{fJx31ARkdcW)|wf2)=0w#7Mk2M>#XBe;*dF$_<s~+-IVB1cI
z1<srk@oZka++<+`oI9%@9~T_H`PF-4QK(ZkYSRt4U*u|VIY<Rsr~0p=iDQ>}$>eNX
zTOVE5rw)nSIe)C*Sx5Q%xr``rm5YO(Lne&-t8s;dwF6iH$Q^z>ZI<Ak39MUJZw3aT
zU9os4C_h@VIBhWZg+&Za#UUKvRQeic3(|5v^#zy<H7&kWjQ|l}(ZGNf;1Ew53(#W6
z*saw$J_Dxat_<ie#8uyL_gyM2qB3I)_3`hGDhAy)MLw)((%c<eElUy~)fH!9LguZR
zXlf}-Pecxmb;yUTl7|D4mVd|ySIO~L0j?z({wfLnDD(X&r!g<Jh>3<$+JaO`?#IGx
zBX#z9rD$<N4&}4ZGJNh#0%<~lo_~_&m3Z=1N_3~C&LUfh&Se5u05E%;kbQ+0O|R!*
z;Gn?B%iU%?`Q##~ov?o?T_c(GMoXt8X?DZd?U+bJ{J3|I-y|o7T;tgA0~tm;mk4*W
zHP9`^Z0E#O0b8-xHp%~hU1ykS&Ij>-vDPQb>@;<E&)MHuuDo|bc}w~|o6^@W{i~EV
zehfZ|{gFQH7_uj;O|4+3WKpD%$@;o=;dnF*zsUiL1x+Rbv7^=5AQ;pT-m&DhK>YX|
zz`w0UwH<C{UfJyOytK#yJnUQu&_F`M#!M0?Is*NbkVRJ3DsqvoR2Acc;50oX$1C+G
z_)1aTi@+E8_K%JbFjM)`K=)QXT%+Rl6cblj9W=F`fT7-0Z0Vy`ekjCjI--MV$K+*8
zdWam~i9Lx-B$VoO_%pezya>zfiz@oK9QyqL`gs__^#bOKkL#2fXTJYaB$Q%2q8B}C
z&u!oNh2na?I;5K7f_q1i6W>A%Ir^>sP2<<UsTQ|>_wthZnY0Ng`!|zfHx5H?{KIT;
zG63Z9NIAdlhJWOvDyqq$BpfOJmFDIQ8hEZ{lLO%d_BsNxW_iOGIk~yFZ<wCS-08Jj
zhE<SQcS}IzNsZec6gvGfOnb9%BF;B)y#0-!q@28KBl>GpnSdjInG_*aP58p@+*>gQ
z`DoMG58cl-`yL589qO38_KQl<s98RH`$6cyLV~;_tp!eDu$^~?Qce{9^5O{$z^q<R
z`?+2n_mOMN>Pc4sFsS>B-BC@EOc}tU%k2leuOPhk`J>Lsi6ycqx0hq~&!&F+$$&UE
zWFc(+|304Yo&Z_GAhxM#LZy&4u0Op8^E~3l{G2+zv2>;t1X5AMs>(9H1|jOz@3=!a
zjaCyB)<j~b)GQyz2`8KvJ!;b$&sPx3jf^`hiAb(va@%N?W%ZDT0zp}kLQ4;3|L!Ol
z{fK7ei-0&$u#){53;Qz_x;l0nB(z_uBoE~{dCEa1E+te%_`9T0KPv5Dqr|_~7gF2A
z=FFRgN-IQC5&JjW$nW2Yo_`d7#1MDLsAtwsnboGze~})i*plH?`kL>>GQh6<02ieZ
zs5f<Zs<5JyDBB3oi=wg)JhBW)R=#iN97qNNE4+{ug=SVZnvk3kiTX$$xsuJn2|sh%
zJHXV}4ut9=z0PuKoRxlEK4~69Q2t&uBAbrlXQ6ivve6x(*ZUTZ2$BlM_X;P&RHbLT
znMZMdPH+*^SAF@`<nF~CR?P)zM~55Drlsgj>`t&LTz!NMa5lc$-`i30Q>CAFnhz>0
zpSC<csXup~<M(gnuJ!M*5;An24f?$>KQ_Jnv1{{_oLcEIM~tYUmA~=>K>{xRSf12a
zl~fdgo8UG|${uA}WAE@7+3f^H&_RP)p>F1+nH*v&;|?6#Pl?_=bTcWQ>9P^e8@ExY
z?+GG7+%^rQkPm16sl{8DDLm<UDY~5dGX}zS=9*d)Sj<S+kX{W=zunrfU9HzjurT-O
zvALywD<>Ka$|ry(5c@gqD!q*x_jvmpUBPDm>QTMY6+ndQCNBA!i!-aWiL`g<%gf^9
zp&84Mi3hk7q+Q}vw+rvX+5vYYzF*tF4C-hnjN!G&XYXgVznDacuPQ!80-3X&!0jz9
z7)LH<i@TwCK?1ruvj*sCHb}!1NXM754QC}@Gt=*Sv=rxxoz=&LG{<4zmtpX-bqo0%
z0~dBi#Lnt?OE5R49nsWU&*d|6^eV0Z(o{&zz>8VnH-Bv18}qGif|!i4mW;0KG*jGx
ziiE6?Vf>wP*e9DW`obBW!_5q`ZX}@dJpupkKKOryOygYWMEbl34XroBe%m`8hD_|s
z8%IS*1w%xXT6=6ldaM*e=VCOzGZN&pirVLJxM?Yv3qp~xQpfqu;!JPEnBEi&hV|RF
zEWH_x-d(HBI>_IJ9zU;(8!HKiR8i|-Z*{u*EZRG{^o0Rt(*e70-_JJCzTSK0(UjpD
ze&;BZbMv-Ma`9=n_ExfJrE2GWC#JmygU*asmU}H65lhd)YwU{g`@D>-+X}U=7=Fda
zd(@+Ae;ASAR{77jKis5OkmQTHj~G!Wk}2^T-@uQ^lOLjvpLj<2Y`uX5*Sz{&Aj;K*
z3fG$S`Sco4NAQ9y0I}nvvxCH^8v;rutcl6seaR_^hLV#mUvMx_bpUSHB08pY6>n-`
z#eGm{?etZaa1VQWrQy9N3~d1`D&xrQ18{m2*#7BPKgHZphrGXr5BHWu?-NS*{;6+>
zi%4(KU^mH)OxwO)uR?0tVtRD`R~az86vjo_n1+$yj<od!)9aL^(jGH~8@_eB@ryF~
zqws>yEph3Er_0H=53gURi{3Bzx~{8OK4Ez>>5%%|Ryu}Nf|ZcXonaY=5}2J|$er2F
zopH#Wa<F{1t8CgPS}<YHRlnnrxwBG#;>qW1th#+RHt!Q_2mxP6NF~Q?i&JvRMvkL5
zlig3w@)z<QZ;cOeHw4$=DR$0Q^^FBrRtkG~dy~rDY80?K<Bhv^CCb${PAnXSKEAEh
z7~a)M`msS)*ZNa$mApBeZh~o-0(J4naB^Cp!~099jB3ETb|gMyX#D4*XAyv51wCNQ
z`aN%-e4kWOQ%{z6pSi1q_A7Vc&{NZX?HHasfb;jj^1!|^K_b2-CpqOt({Xf9dqHX0
z^$lQPnD;<5>4lCrlOyjrU`72xX!L@zaACqlUXo#*ED2!Wn`lBO4mohb?EwE4-|ziR
z$ttPUw|O6$tTXE^)#6F}mh>hhp43AjFv6AwA|b3eyBW$9(R*+GK6K5Ifi~In3OdI8
zyvRX1=Go@;(*o?&pm3|^^N_bu`kza9Q_z|>fx)n)nRDDEZahvB*sS!Z6A-ug54Y*^
zGfv$vDtlU#Ac_SPTes`=^j?QMNFRjoExb~-K7Aw(X9@;Lb8o~M#-`>tOxvam4U~5l
zZ7H>^%sKi?4?yxI?`(6y=u_@9uKHt-%<0IO$f|rAZ2Z6{<VuJeWgqs<qZoGO<-776
z$Wj-csSa;%OipV#^AZrHC&hB)v#8KtOpRV^w`T@@&G^h8OXT=e)X+rKY0p?QTgo7&
zkoeBmH|$>?9Dk4PC!cBY3(8ZwS)qom5GeEaGp4tR+Hm#y&$!;G&~6;HL;f&S9^Te#
z(Iy=OkX?O2ntiSjCU><!AO=9$P_riIP@#;mXymEnQi|zYc}WFUH4iNXfMmm{qrMIV
zn7{$Vl-A$k(L_RAY}bA1m{-(#FFKUP2Q&xA3t#pet^rGmWN$sCuOZf{W<a1}CLrHa
zew**$+R6H(7?Ryco2+55DZ^$iO$?W)=nX=b=Ov?hRFycB4wm%axqSQaR+afacF*G6
z`9Q>S><r_MA)nG=eiZXH42VHLWB~F=B8n2U^%b^plccigb+g$*K8#g@zJ%d1sS&a}
zx{Zzb_KPC{XHVi!lE~Exv`ZC=68)sMa}zb5wmjJNfKI~ge@t#mzq!nOOTr<SSEEkk
zubwaeV=H)OKd5+y56?fehb6LeP(hl%mMh*T^bs7ALUurfFCFPih5mKPdB-BcQMnuB
zc8_shx!eB{Lx>y&1!>_#f+%l3pGHD4k?gCD&x#%2XNa>$*c>nnu{#r6A5D)!fxK86
zy1poz$B|In4#^XqIHu5n^}gtDlu<EbE&Ak-?)}Sq$*+2~D!BMP6e8^|TNIl2awXuh
zB7#5u#el8XV20!erLSC!pmFO`S=l7Z+R)mtjge8{h};YB`a~;W7}350Xbfnwd9h}j
z(OhxQ6(W4Fzl%z^qJ2FTI+XSAs1MQ_nSDa<&O{HeQ=LzuBSLH{IVSW?`_0ngz$DY6
zbMgtpb!H!aKh3iRtj}N}H-ySIqh?W68Dvhd(wUI$ur48A*uc~KDG5W&VocA~aLSzx
zo^{tnxt0H(QBgatWjT?$O{`%k@8jhXAPu;;+$0}ZgelD>S60ln`bvRT4!0DcF{%;)
z;T;t~=T6?N*%W0LGz97Y2oHGkXz8==_u5yO4%^M*rnAeLM;0vM;x2G+4Ax}6#%4K*
z{4OWe&PIDSnMbYi{=dmPkr#Xu4SmC{hmktgy}1qTtOf@0)Zag@z}YW^N0mNv?#og|
z0tu~6>|nY-wG*&TSSK2^=FbLhgvaD3kh__*n-gu>+py{i{?0NTy1?D0%;kLo;+RzA
z@IipwtX@(=-S>)#M3oQtf<(k&`rYB393j?YShWy{>Yb(cGzXUkJX}1ntf9n!bc|B1
z)G@Ig90Y-=AywOM#jHs^XsuJqN;#>IR7E9Zo|>E&hcCARjHw9f(VM)4Lt@iM>xG4d
z_?elR#C-iUk;V5}^}Aj=lx&pC?>`*~5uD*o)L)B#SA{1<S2E%zoNFyN)87|Z$}#Ws
z_osPQJKLoGoz#Qxb+&_aHQGX_4NeQyp1X#}l8ykwKTuB@f4emXq^D2#Dq&imNi~IU
zUV{`iJ(;EQJNwgK_1yclOHY7yvvz3+(n3n7w%cv{r3T!!EuqV*y^Ty|w^Wm7AekI}
z{MdQ%dCd_w_#k(MRg9rvL;#_R%BnDLHT2KghUMMYKEbfy4q_P|E3}D-tH0<71(bfY
zTRDojR~sV+EMhvts=Or8zdBw&k^k9nb`W<7pyM#}8A8szeC?G82W`sNEScxW*v6W-
zyF3O;`<qJsR2Yq7e~!o8#6q$QuCAx|eQZ*w9}mkMiTstT!qUtLLaALn*V=I-eah?J
z%Ah9S&EoSeOjTZP*=-K>VBsoif&1<E0G8H7r6MJMl-0(ao5TK0te=@k`k=`g^0x1L
zVP369#Q9w7;im_7!zxY)g90xq?o`FwwHeVi4B9P{X`_*9{%ioIgdFQ{qxj?Qk;Qo&
zr3O=)9psdK9F*;20ukXn2u&qd-PsIHg$u|3JEokFruyrVX*ws#A`5f_@?dZy#d#u>
z?Vd{9k?}8>xD=2w#?c!~aQ@U%p9#${GZ~_8-P>s|)_d>7W{mam;Y%~!Na#hCSs-8?
zWXTW){PB5T{roY1RvBlVUJ=XKqh;8<+92FIY5XqqD}%W~I5ktznoGrN3d12EI#QD%
z!Y&Su18Tg$=P^1{uY3Y`b%%E5g93-7=!)TKX;WSjO)W>|J=a|V%zyJ+#-plBnNk9;
z3E~_wqH2Ou+55ZYreVQyf>$CtWlV3z0dMTR0P+Ll(}xF{z8<}%mu5tMW$IYK@KYH&
z_k!aIhlaGUxOMNV?!@hQ<W&=K!kf}+B5RqB5pwu~P%n4Eufcds$@AE$)@asKcP3aj
z1DNm6CbHE`ku7?jw_vkR86-p_l+rhTTccB&g8lI!HSAFR2O*yn+pA{!P7wbN7aDb7
z9Fo;7t2_$>Ri&y%3$BKs{93dw-+U|i$h#3fkn=M!{uP59v@YUDLQXLXW>Po!WI*Gw
zNb$Nr{>fd+UBR9I=F7QrvXFGlW|xdeSpfm8xE@n;Kic}Ay41UVd(YqQ@6I;-E{ZCF
zEk3Xa(t3u|T`Zokub!F;*6xP~r#EytKh3YM@??2CBW|s!U*LTqYrOcKEPGz!4P@)>
z(C%VTGDKR%FfJnExq9#`*=IoJOKcCJ?NBcRg?H5@)3wC)!%4%*TvHb%Q7p9K(u~2W
zvbQW)_|*}GwnFL4#AI>(1Q~bX!v4!Q{YFxt`eqKr>Y`qi$J<9h6^mMETyc2z_MlDr
z?cp;*G?z-mLjUtP-Tmz+wA6<RKD?1WuZl{fh~0hfybus3%@xSQ#P7Q6TNHEMW=XrH
zt%6lfl;)3euu!XxE@_i{@gyL=q)oItzXQnn>h+=d$r~@cvz>|HRA8e--)}Ml7U5EP
zT(+mjrqHt7;ywvu!J`oW5^?0?c3*Cu7)CDV;zWKZOJORQcPB{)>Z_x&o4|#3ZxL4A
z{ojw=wT1b)JmCp!H;#AAh%eKU-VRCt;c1?HezyYcGs6k#0`oejy=|a$r6ES`o!n<v
zxfJs<vL9DSk@pxJDu2dJ#R4Lu>bkzL`bw=3f8iJTs-ZP^>Y?A&+<5H$xKEyWH~twN
zlm3I24J5d$Y!m)tzg(_wiqt}}zL}t&;CsK+XEp(f`?*sS!7lfkC|alDG+N2rZD11o
zP(qTa7$98LD32GToeF}rh(7;Cd($!;x31vm=w5a&)y-^nq9s<MXx~j+UpI%7$RD1z
zgxWuh7a;u|V&6uw@I`bmI&Hqk89wuilv04y!_ku-A;h`2j_K=-uZBB=mHs+pmK<kg
zP>F=J<vC|{w`Mr9QJ%%C$r4{`i|X1PqJ779Jyh^y88;*!xVFmHNd1V2gHJ9`d=g?^
zu?dl52~w}>nGK^%l?lk0r7xq;ck_@;$}xhZ;3IS=1@IB?i~7*$|4XPFa?M?syPGoe
zq<n6^sidQ3ZhDrw<JJ+oKEc#HTib!b?CXYy{Pszci>lo)$8u>Z4(=joaKLjMD?3Yw
zyq!|vFj0a^%8^InZh}{?r-XN-RhXyDzwd+a$K`KbV@ZY0Q3U^L*0?4Ojddy{h;_1$
zj6YWoSavDvVR~SU<Hr_0tN4H=<-py5tFK>~#-x2p0Fe&BR(I<nlep(AEk1ViOF+c>
zU@^EGZBRaA)iaMqBDeADpgse@h1#av<19$RR*`l%VI^$1!rGD9-BjxZ;#pR&COsBj
zq3e(Z8M4wx4wG8H<g*G=4O*gSR}_#GV?%NwdpzUAz@BIF$60zBAm!-?_lwn9i$is{
zU=Wv$)Cj!4lqC3Ey@8WjBR01@D<zog7@R8TDWP*TBEYXWJ;57&vGE8<x;G*LbmVsb
z%L7XkI6D|N`u5b*Bn@sWwC7}g#sv%r*H0K@w3#Jb{xNx1VeeSF94z^MczBfLus1Ny
zbwa0v{iv$@=OJY%d^Uqm+bo_7lb*3(B4+p}=(u$)rslh7xs>#$>tv^=&gK_<AP4iX
z-ZGU`@p&~<PZBeExy7HilG*%;d(+<O#&O(Mzu%#uy8n-zaO@^^PHFzqx-|a*{QVVO
z;L`C?ThTh0K@PGv(R93C<BTH^G;JVZZg5s??BZKyiypuWme2l7B0Rpa$$CO8e0JRM
zfmlJ7uC!PUI3~iCyLHEdI+iavEh>}X<}<2#`%jrH^;J1#sWRatP{|(9=SLg1ZX)R$
zV;0bnZLz|6Jup<GwfK>&@&Ya(8;D)EisI!FQ5${vQN4;XDB!9@r_&*k=Z<|k=LEa*
zJi<M`uRjQ)som{=XUzeVcJDN?a!zpXzcoNCpYTV?zxhQn<C_cs!(d1UP-}~sY(+42
z&L2G1e@r*NdHwyiqcYvOk~_3LMywJXAMoX@#?1jm!id%}4}gCsGf3ic{daC&10<!_
z#KcCq-mqXhm@PcscXu8*`BXUxwI9qrI$(aSL+@ZxG@Gx#d6&7>^k_6gulwvteM$bm
zoH<dz#j_320-2z(W;^y>m29%Nz}}?GpequVX;I1509%(+z<_fd*e3P8nRUC|K1Z&@
zALr##y0l+X8amC!u=3Wk5qG^8>+P}A*Y7ji<<nq$+7V5ptBeP+bfoWKF;VI^ZwFg?
z&^N~_1&9mD@gHzKMKc{QYwySu&Rt&pvrm@-Py@a_MZWRFeA4<tQha4L>~cv7{mUx7
z)PG~f%~=p4C=@~<8}zj#hfN3W#+4qrnrjcECqfAz6GHuKQ^5w{Dow<e{eRPppO{r+
zJS1TO7J9(qxI7@@u6hrz+qI2i1-h79Y_eCDx&7xW=60po%}-461U^;wZQ6J^G71ej
ztP##s=LgO&@|C|AvbIKjJr|vBZ5aj3iW4TZqA0#BhwxZGL~&E;wOw<BgsOi0Qc)>9
z<dhMFIRib}2LS<6k7UrD!`*<NY#5;UmnmQV%2rp^6=@-duR0q3d1a=U`$h9L_nvdh
zc9DGZx{Dl)3F5aNxK{6ruE310@6m4WzvU{7O}I71bAvs@UXS)D!4zWs#p@vvC?6UK
z-4SsfgudfjF;P23C#C$$F?0^aY5c<%lu%oJ5K|$%zA_;nRjUKhpNuy2%9Dc?WMy%p
z>bi9g^8;vU3h75i-h8wmQN@G7GLeeliYrbXSYS{NnC>jmE1M+~95-5wP1hI|Fh7wr
z{ci$bqQIv&w*XUWFd(V`k9cK%0&MvKKL6G&ua6p^^Xt!+Sy`seZchuprsPu)k(U>-
zARVOiqdEgtas5)mAIW8He(>H_sgN%?Gun<JHo04R7UozqAPOdY)w!M^IsXhS(Vp!`
z{MJyS5O#GV1yg19wV9vyjM@RJW~hm5O3Lfx_3o-{8uVfUe+8hC{`u`IkAMBjNDuC$
zM&SBkO{2{K6(Cr>dlz6>7OiFx{)_zB`?2cpZ~8kKoG2&<NWDpQai4l(kJbvrWDc(@
z(Tni4{VfxT;?8AjHm7fX)igJlCn1d7Cd+|(YSa`X{R_6*SmNQS>klVK1}QJ?1zJNo
za(x1!WyQs&=Mat^Kdwp)NCN>!<<kpec&OVtX+)@UixV$39BsFbe{7vTBszP~iPAXE
z+HvQXuZkHLcyOc_=}QM#4T}PQnv#D!T4nC1oJIgYVV^JMnNrI4Een7<wpF&tZJHK_
zJO6v=Hq|>q!ISWviV8gd<T)7O_SgglK`V^Tq@UBJI^*X9h(|e@XbT6wgpkQT%aC;-
zmyx<WkAVaGE%2Y4r755c;IEVbm_VRYBXiHBrVKJ5{XNpU{i-oFk-@rD`J)ow$5)64
z=RPZ^)z=Zn@ypv|$L{3lPetp84O`1iN8<uHPfe@6mYDQYf9HjPbnN-bj@NLTNqm^j
zdtkb)JUR$H?ddTQLWQvuP7G6~u^?KtZ-G5Gl|DRG#gmOhmkc5@97XWj&Gf3&>cJEk
zE8>Tf6RmvcEG3wT5aLKlZ@b5hS6wu9=!olHf>vj$O1WsylXOjmp$RJAtZjT)84;{3
z!oM~L{H%W$JUKX%6!?E-lA-8c6T=VOT@y<n5V6O(wXP^{$2of7Uj6EcuiHtH_hXme
zwGpA-e8gB932mHz3-rp21WGl;VKSMp9`F@d*W5)Ox7MtV3XHxl8#?zfH#Y}_vLYdW
z<^c|+(vDYy48rzt0Cp;&M9d(6%$QG%Xk)9T-Hv%Uy^fG|`QrKN4iC0em3~eZ2QZ9&
z0vr6~<HtMSxLijm`~Fnt!|C^zrKINp@HXE{kD2~7s>HuL3Bs{qy?r{ikhgv2F6*Dy
z1rw*LijBae=FI5)f+^ued-y5u%M>;z_+WTdx3+5|x(CPnlx;u-5u>X4D19(EfgSj_
zeM(cvt8=*H`Exh6=~B-SufBrRft$gDDW-PXawp`tI`wxA6BH_x7=Nz0zayrMwEda2
zm<}?}h8FsO>2v@8oBOVF72$oO!?9S;X^B&zJpr`TQZglAAEwRWQeW9^vb&;KGH62O
zfY~-I@K#psRB^~jsSDMF7iUnYJr(}iEj%<96a~a>GC=+3$!i`&%}dW8C)9jD-gCa|
z;*>%Lc7}vJX%AqUDsmKO5U|RF$OJqWjJZ>eRuRnfBj!gDa6Bj0Fe<iAb8Oi@xiPx{
zbO1&EF^Drcq7US91k}ZvwDp^l(U7VM(%fa<O%Oja$r*jz;qHhK7XItWgq(=HNX>y4
zFHCrjooPQ`xTHm&)_yiBS|(gg`F&)?cdid+(`aIe9v>{l*nHVSD|4ba_d%RHz-PB?
znmHCxZODZO`v4Q-fip+%l<`yk2HQmye=7yg)Cx<AWmg7(5+qrpp*NGGCBxh2W6Mip
zo8)-C&t5ZU`qG{$73W{(w3K~q4DCxNx!CN5A1y{*yqLuMmWT(F(#f2i0HLFPF~QuM
zQx=d9<x=-Pib$jF#+6S})ixk^9=><kcl_DqcSW-myTI7I(M!v=-eP;0Z|b5&f#yag
zee<3@n#`DRou5sMN`4<-A{z;HL`;v52kB`^;sard4(}0>dJSERTyZ_$%)ROJCZ~33
zbOM1ZHFjJz<3O)zUVwAtm%V?Uy^+enW`CQoV-T+@avNF~0AaXKpzM*Aw_W>@zGC;w
z&{1N=uDw|`xJA}HSy`4CtKIz`zO%i=+648>*TF(op(1%GmAl*sEh6lgp;!-d==|5O
z0Oz@ukM&i?BNsiDt?#s`ZQXs7(f<XMGFtoqBX+Ns-#cARy_k8BSj}t3is6hNXrtEe
zKdE$j?l#eCzd%)bEF4;Qso(!}Wb#Z{`mqQpHp0WGGXD8v3S4@jaGp=Q%)pDwiPaC$
ziB~;@=246WWDNX39*S1_waoHD3k?^4Maftil@d(w@<o)U*Ei33H-7-r@Ew_;xR19=
zc5a=QMd$9%`$=OB4XAv6wjKr&8L`Jause*SBzwiF0G3@z$}^8@sN7vB(xThZ8Ovhv
zoKacawGQ*CAwmF#qP?G6L9vwR)4CNMc@fnuu1=8EXL?E$NrFss7-%{rOUYlx6g(oy
zv!fiTpf<VUCv%{d*gnJOxQm0xac9zfQ76w-p$ZCEL+X^2SHz~R#lO+obghb+7RcQH
z?%6b=GO8Lk*u7h`p1iyL-)`^ZZ!-|!vj@mMTHo9hGQY~)hJQ9xj~(S53cOd6@tXqn
z?T9c?4XsguNFr23iPpn|q%mJfs+AQ*o2ux7L`J+xQ<va!y}BX44^>Y@TcO+pg9{B)
zO@A(DYI5zEdyOQN&2d~-tm=7?MY=owxfBg8tw$H_Tb5P5#GJNfakHeB%=bv-0e7?8
zesJbfdZDhO3Lz^N<(7v!a%_<~g9xPF*;ja!VKW`b*`GeLmvps~2E|zq;Q-tLY`kF{
zNZf4>q_RFFbxqlCv{)ShN9U$w9)|j)gd#c_MELhYOUOf&dG0DN^C^JymbZ?hrX%Gu
z@es+JsQOfyw}Lx6j;mg+J?-42Q0z-e?0ZM#<&)lZqW?B+T?OI8r>%Z|0oE0jm!t$g
zwL@^MbotxmUo||F<_FiZ@FtF#mwVf$YzpXgJ?CLX`+Qs{$A+Gi&CDM{QuKB%R~9M0
zopJUv;z48ScA?U_Ngq3YKfjZrwRrdLDq%Poc||tdZ#qCIW`W0Wl6bvwmTR!Dkj8Dd
zxZBe_722%YMmR0G&-waKYjX!pm1=Q0;i|R`=MVeQ+}6HGmv)P&&_TmUa^kcf#GfkO
z8x*^?fZyvB+rOf(2!<2;nM?{`YX&iTN}y+kL{Nm!y{IgpvSa-z<*R6pYbOvH?rgLt
zCr*?pm*RXFeMPp4!r}--zbTOlcA|BF8;%`sT_OroJmMs}ON<SJYBEY2+K7)nj~cfV
z>v=iVAe_kosg;EASdIk!7vi#9^N{xYa{@iHZ+JGW$Pp^L>M9v%WLw&X&#%B(e{Hj1
zn$dP_9<sOekO%qX^*Ueq-|tITZ}L#3&atUjhmg^tepq@H{e2gaIQwPR%V~I5`?OoC
z&J<Dk&v83uz`lgE_4QHbbfJ&Oy0@!O27`%1sB$^|n#4amu0sx4x~I+5LBLz}(bv%?
zH)k-OnhF}^IgIWYkz_^4-p8Y$3G(~hD|GJv4>Q~3$s9u<D@gZ2A7GQl8^fh)RULQN
zg)#9FFbhhN;^hQaxbimb_8Fy-pKpxrsNRN54h4!N<r41`n|FK4>$!)|mj(~cqL{F1
zJFo&2dvjKn?g4@u3Bb!i0cu26?}KUedg0VXJ~Wxdk6>j?SPwwnG(eLJo~n)GLj7Oc
zbuWQQ%Ee{3Ht>0E`zrgrFC4^|TB?mdt;8A`(P57DQJ3$Y1%#<i53A1ZGY9nxYtF-L
zl5%?|KLCTpVSHW?gz+bNSl8^v0l^&z68$ztV&;v;jYlGLtjW|X<dNS|mQ!rRc-05b
zpX(p;f9gi_XFgURojBWO<C_l>p`pB3Rg(t&qRHtbr^1R4v5~yd12eOx!eoR~YuH+(
zq(ra7ZO@^rRL{*>^$k)j#TnR%jr?p^E}AG}l@{}H+nZP)Ni@jOgFd>yZ+vX?vMzEA
z+5*m}y<~dQ9Zl?W;chk?4MGF`lD%~bLoQrtn^p^hl3<|<RBn~iK#=2wr1DzJ{4zWy
z@F&7+v#%`wiBhLl4uO~`u4LzcZKF~uy&!Xe?Q-&Ll03~}|NFLRzP+zmID6Od4Dd;`
z1lM`h!KRGgEd^`+nF_kfP-90mnLlqqn+q0C+uilA45dNMX<)60u^yce$+&;mE)Eod
zKk~j0<$dh!SYd3-%M!8FkW>k^d7lTc&S${XBC_1|nF3W8fzdO&uc*ouTJwTvUX<|=
z@P9j7cUdRz^m;~YA!r^}t)lQyd3&?Y%{NkI$~Jzb5toq>l>KO#kECki7v)~W5Bf0Q
z6I+<ssGj(u03OV2^3sYD>m!7y<%J1qfP-YAvUHh7Dm%6xe%Kd43YPW00w60}<k?5I
zgIl(QO=HFek^8;y;QDq@1s|jZf=cW5FAG57(0O3mOfYSlo&0|%pnDPuxfD>d;f@(Y
zL-JW-x#F9uE!_erKoIk~#7}c+FxhqAdcpH@@~$6V8%8x6d!lggA?HzT>wjs}Mn!A9
z(udc9es*^rnn02+4iWSs(~DeB%JRzPZ65r&<;c|}{!5HVm|cvsDOM^j#x6siMPiZR
z@4e^mrSjQhhIA<L?q}!_$}3U4&arh!CSbCE&(zY=+L9;6&-LVNv*N?SH6pM(x+LP;
zx1F+I5krT4kx}&V@$n*D>#kUO%<`d@A5Xlw??6Jw<_4zUi55$@MS$e?w>>XIgR+|8
z*F_Cx__tZ`jws$#u>R3mpkobxS|H{yo95A$08uzu^F6X0U9t!HCz{GRHnSOSZax?K
zj>;cxQbTeb)6Rhx8l>*`nI5cr`FU2@Arq1=LMF*WL0WBK!@uGCXKtcdY6vQ<y#(S4
z?Cdwrn8S|}hAz=Xy_A&MnB;r!M3*!)$bt)?#p2FJ59z}s-US47kRU*;|CabrAruNk
z92=<<dktgTo`E@Lg1G(MD8%TF7o;^q<Vx)CJxb0c^`rTT?C$cB020gZtrQ<nr}~9p
zs>1bM#2I~nQ-@vJ^jwbBt%pLOWM79KiM;z)xsxxaAc7Y4e+E~?@mRQ&ke5C-T*1ug
z0><R;QBLfIhQE4#`jkz9sJs#0pyp^0$<LR2>5bF_XO4y-DXoUpm!6;pjZ1^-Q3V*n
zNJi)qRPOyxHgMa_DQZBw&E7oIxfR}ZbQLnR>xigZ8GOfH6~op^DucR9j_|tsPLg@W
zXUX%ClQXLtQlud~pd-nXP`$+{5JDo$@t=pi#5&ePE>Q!q6)ruvji?A%@ycL<;5%vS
zVxz~!?jlyGbqtJ%4Vxm;FZ^lHaA}<%HP9EUd%phby~26ErzgGdKV#%(4QIJneifT_
z$;E|Eemlp9Gu>Re<CI-?rM{!;4F(@8Svd3j%YKt*ScmTKsAHzqkbZdQ!A85%sJVL_
zNULV%-Co}1-dIgUJCzk9foK>l6cN6hCZ;Ht93VICC8yEKM&5P0x-Hsfe3JjBP3>gU
zGoI{@C>ctYKGX2QONYBFKD9B_g+dQe<6c#v0n_tR&FxGIY``b~ZJPXgj|K|7Q{Y*@
z_@fsOE{d}kXIq#t2?Dht?Vo>>4ew`lIK6XPS0Uy|=hbg%!K+H+fqeB7cNWU*axN4T
z6cKS>z}3@WH>|B?qop@hnM4|Gahd;w`q<>$#(|Id3)B<L@`avud)-c3S0oAg<DN~H
zjM9Ru&R&|&&Vd-K<RkWKA=JK_u&UG9@jqWmoK0vzlQl?EQLe08wJ-i^&<*;ir=biO
zAfjUi?;%KfqirE?k@*NzzTu_LJVF|hHe>1ImIOI45<<H@g1CnT7#CvhriKwqC-kr%
zjXEE2xP<xaNtlIK{C<UglK%Ve>^)on9|r3_`T+Ia^dQthw+gNr=>EUAt-jHLURWA#
ztqhs8sEI;omm0VX*}1cso16B^G;Va@+82&qdI#cJe*Zq=J>;A8kAGzB@On1bmWF#1
zbkdA0Fi@6@30etDx_-_gGCUjYT>c9Y0n_ina3_u&xt3Y4GPb_+`JB?l?!;bg69=kk
z&@xy@lSm64<)c0GcdmR2L77?&8+}R8*9VOT#yM(D`pEn@GE0tfRF}A!nOGyF@>;^G
z5C0~prbUYshE!#~JDw(ZXHcoUK0Ii3{S&Xotg(d=eBY~8PXP5jB%tHsBA5tQFj?Lq
z7@Bo+azj49%o}sgl}4&l{rjEDTSL#3IW1B}muwL&$<2omkvi3BJB<T#u{2$pUM>>j
zIO>+zP_1u&FTgVb^h;+jyBe|zxTW|dX8vRIhdX?w-imDIU0uNo&$4u0D@yj~@>XC<
zLne8t55g-j`CgYblD>$n;i&SHV;O{#a>bcr9LTABD;^Webt3G9ao=xIj7(Y3QHCf5
zn*FW+13w!b+-IkZpl_TfpE&5U+(Gysafm4MXNIL|<xe<qtdk%87%$RjPI&`zXr{m4
z5)QFAOmPhyeIfI|mzM(H@|!IMTn%j8c1@ZT%yeq#p~3%*+|%~BMD0pQYK9)A^EA9B
zlL@@!@xMT2e%$SAkm0Jw&646aWsB*sH8W3-A<+e(V=}iLoNF(neua?XM;|-txc@By
zxE!!5i6I%#J4LzKLaJ)MhdW{Nmre)7{T^PFZ9<GTsHc1lN@u<!`8<Oh;{w<Lzs&LS
z^%~bUJ|WC;9P}Ydu?8-1S~I&_clkIsMH0$rr&ehB_Pqz;g92H?8K-Y~X8ffuWSWW8
z<5ygqmIQObzuJh-zo9)Ze~<01udXC2fd6wu`<02ijuf}NMy`FDjpuu5*P@1Ic}pA9
z*=jUTDIKI^eKYFmKS|ZJMa`gD;q)0hYge*%eSf<Tt0qNiM=j202mOXEcVQ<rks3IL
zE|i}TTFjO6^o#?l!u<nYwRN-<Ye1u^-cD|*RkXm%kES==Ql*8$EL04(Ri`fN%!((<
z9$37c+qBadM+}YamR<Fj9(u09bdd>8nE1&)1(CjEKetCHjyLmssdM+Bi2Gs);dV<~
zV%~;Iqm>EA`pH4h&y3iycTc`x9Tem$6y%%;ao;jg`5hAf`#OW!Lpu+0Fw+8jXrn?a
zi&6vdZ0Y2K^m1l_R7ItNx}2*eMC1C(6o{asR$6$J)S$5A>;tOzc4&Zy!2RCn!}z|^
zDbJco)hB43T|~&HLuvTo`fJQdQ%!T!x1`P=y`uXdl0N4BVxv51!$P>;!8+PwpbVkb
z_Qmq0*I1turZkaiuYJAu61l$PpQpg49^g(71MrIG@PMXvVc^I$F7bvy5cJl$jkP8y
ztoiKETQe0EYu#@jV?XAu7_q7C^PYa!JrF6}44A(s`@d%r7}T<E+=ec5{puNSZl?3I
zEN;In_+1~1xsxj0JY(P&&$5aE2U1m)`qRlHG8wKwAOozxxR;zBGGQMhB|{FB2KR6L
zXm?utO?C2sl_SgM+wHK%kGu^jNlmQ{Za-OYuE~&Yb7z51jlI*8%6?nxbG||ts;J`e
zgM4dBy?CD`uPVIdOtrQY!^_AeloSIvE?l>&(z69uf`j~?d|4Z)fE2ozOA{vu1_@S$
z?@k+O^V3b0Pwk>7>*>x+6y!{D<&gJq-!6O&k;MPG6Pu+x2w~#$zf%&U3ysXTx7zuw
zM8u`{<siT4A@;+ofmh&PIMPD|ao~p;dI~SX6ssv3>~Wy5&wKqs@+pyLF@4bjXs%ND
z+4n)*$vLactn)|SspO;cVX=GCU$Dn-+UzqkzxW894NUtz_*fijTv5qv(DnqV@mXLS
zAiC$9hTON&UAil!(B)SousjX91~}v;0w|P{P$ASw>%13%+5*FK*ery6Gxn)LZVINb
zU=eJHMAr{pgUd%PqDH@O{<ju(aS#Pvw;&o%%iL>d@a=?xW{n)D4hQW%1h$%{cA&iz
z+%4Wo94Bna1&u7D4MO6$H<?upkyVBZxJh*aM7!1sT`Bw!K7W`$=l(R1Ekv>O`XKY6
zOnYn4k7uj98VE@$6G~|5<bvB*oGZ*+@Yr-mkF%>qmyOMC%~j^&a_}C^V*DVpMBxz0
z2sN((_g{$~LxN7~x}z8SKAh?-z&^gHU*(;ijwOaMsYu0UEhx_K=eKbPVbXUVIy5fq
zNHu?T?vO=Y)-Qq+phb(+|GC87ed_!&s}JvKIaNCASN5J-JguSUuivq7sdSc6?RI31
zG>i__2x>Q6=mqaB$5ds59uRNgbg?}vs(~g>m?!NPaZ{Iq;%k}c$d_&7mC}4@2h#S~
znkmnG6KMR|@kg15M_q}RO*oJaf<8Cph3cSqVnMm9jD)FWaRKD8$S6pTF)oE@fcjI)
z^rep%NTUt@cPqRU6Df<@!9D!`Z!RXp=EoV~qblL`ZbNR=1V3tGejh`P#RVM;8zGA$
z|J!nHj8mq}I_$N#nhDAh8by{nEnb6cO+0VO#Y!r5s~G9C>B@YPJ97Rp0y)fq9;O`|
zyeEYecDR2)t;3tNk?WFJzIXdEFScQm0ab?NzhDf8J9?0h%-$)kiZ1w;y$bC67~Mi6
zUAFq=!DD;dr4k(I3)Vgk!d9%8ztEx~B6Pg;?@y2kwnhV{u+!Ih{H1)icm7~~@=_@Y
zqHw-}d9rEtgb|ei3UYo>oJ{&C<=2kqT-D}3D_SfN>f79Yp!*k@-y!n9(_b!MzK~37
zB+_ZHQD(Roy7U&OKBsxcb)iXjy7;APa|>s8jW5@6>W1RqLi?P5qQRDs;65!J)HB4X
zBf3}mA~8B6FmdlHVEJ}|Wz9of9)1&AU62-VEP?>vooid92Ewrl&VSY=+V;3~rUEmz
zuc`8Pdt%2$;0{~;QuOG%_?g?j`LbwwePsRQ?v8(3LO@fe7C-u!1;H|SXxD%7w-$M?
zr5Z-an|$)TxxRE*QTG5%)GwZvT^}AWeI@h1$7nA43thI}*WcMwh=vs!zG4h>8=Ri+
zl^BH9`YJ*0#*4*2S~a(G?uzzymaj%MepU#4b@LEqfu5qvYA5K0L5k88aevw@`p%=f
zC`$Yxbu2D+-ypRCe<W^ZZrcQ_WE=44Bytx8^eXROt0Vl)LtckK%21|j$_Jj~xDtE|
zynhP0aH_-ndQblS0OQ;lK1@57amSl#Z20e}H@5nURpTXo;(zCp5SEu$KU6Dp0=*p&
zK2xSgRcW)NNv7!6CuxgvRl-6ec5?wyiyhksBY!&fmq|MK6YwS^mp`D!K-ikT63mxr
ztntn|o_|t6)yyLP?M2a#)csjlmaK6zy(FK$XKUs_7o9z;F2{2knS7})c)AVTHFZnT
zlj&il1u#*DdkY6>2v+BL7F$*Q2LtvjX9V2zAVFZSsBtlRa`3R2TROex2jeYg`ohgS
zl?(IwTf#z^1chwPW;wM%Tu;cq9CgfPq)g#|OO(K%^Y;TXvn}FHt8GIT{2KZo(WYf_
zKoMo!lfhXP+Mn5<d?a_+j$c)}9QDRXg)<I=Z6W=O<m~`gRZrLtuobPSPC=g#B~4-!
z$l>p7j!?Z)waH{g{Ai60&w7wfAvoj@Sr5bpQn<39R+vxgHRquMdqEL)o8YcdMRjyR
zKpq-GxZZqVyny%um6OWK$TxDrSl0d03Kb?8^KVNV$T^e}&A6{I=$o-9$eA#ZGqF8H
zgEcY0;zB^a{)NE#cK*M*q;MdYbyz<#<85IgoDBK&?#EMVzuXq1*3zbye>MVI>shM$
zPp2{A?$<l5a*@}b2ZnfP`v+-mc%lD*<HVb-I5I{qcZ<Pxqlfrhrx4`WLB1N!G4FbO
zlpM=8yYM3_US2NN-V4%s@;=_rAs=29ehEYnV78AN{3M6pifOnaEZ1P%D2>0FqzebC
zkr+;lyj3y?Vp%*hU3#g2XXzo0;>v9la{vaWu!){XT1xCIgHI{-o><(IIj`(R#q301
zn7?*zM+V9NyVc(X@8~{@$tFDNn?b5VUe{DjmPy#`q532Yr#at2D4MSiF#&h3FZZ^>
z+ynRiB8%&nc>f~N!Qqy^$IjRiQ?|yHOZNvkwoT8cYL50yDS6`+MW_#*8;5d7$;;7a
z8-g&iLN->_?<mHFWa5A+lBzNkal8@Ge0l~vLQ(r|PYL3T=V6>IE=|#(5Px>tty*OB
z-PD6yJqiTH8n5Z=7DqIiXKLjnRpt1OCZ?w!0M}#3cF*g>|GU6BjOy{P(+7LLczm;<
zuG|i~aIZn7X;fg#+YhgcY)u{x@j*Kp`*Zqd4`u!=d?rz71FUa?5XTKy`5;xf`hC!2
z)vD|vwJ@N;f<LsVs8@`rT(vH@TeFy%Ntz@V3p2H)|2+kKcMDXv>6+nXXekC4vz`1h
zm~g@;Jh5N?Z6WKCle@`&Kwo0Fqr5g10IeU^nm?eXg?b*Z0k;c)faDu(<B}uagTmt$
zf>+*IalKo|?pEq;ap(FVAYgUqv-AfE*q5cFhrB5D`0R)1OJZEz!hfdqkk10(#+9xg
zo{#{c!hc_l_MQg(ks}vfemWv_angc(A#2`zjhM*(aX90wSUnMuG5vg~;qsjTY7CK3
zalSeW>+1HSO-@n!0iT(lbN*;IJ$>}t8$Y=2wqgO~?S{|?{na=D^gz95*%C-r%J8=D
zb{*DS>5pwzz`P(#zka6AZ5)wRoZ>1o<t%4vD*;@8Eu`F=%foNS+_0j!+fj3MWuwv6
z(W8SOUKt`2KKZ1~y9YiDfvjYCY5u3ZCu(llt7ob@ln%Dwr+>*B*K*U4gwK~wR`g7v
z3rUL`wp{uU{RpD;quz<^P64LIp7Nv4=JIiidP$&k^)uV8U4~umaFwRaL49%Df+O*w
z(kngQheGS>OP(PP)SsN<@He&rb~xunz;ji$CEo9-6NdQ15)F}xGvMrNYyf}}FJ)YD
z&vM6$9crs8epH$5-EIH-Fz=5)w)t~KT<8pxA!Vm0vYhi1ru)3F4-9@l1dk><k01RX
zE)*gN;!r>l1gD>*?dg%PZ?Q%Fr|uS$%w44$k-SW`D;mHpXstx0c)2JZ?)f^%rRp@u
zk$gZM<07SnQ!CgS{`3)_ONqKLB-7{6ww(V}xxcb9uk@>7b-=VW(gbR-pGAQk3&1vM
zWUF`8dc4T0>{aDt`#rvUcG<KKf;dP(3RCF(njicg2D-x+pb$|(<K`HtE&1X&E8$?h
z*Xd6}L37SkRmJ;xKtKEc4suVyoJ0P^;Xlo0W9VrOm7itR)Oq9XVEHxS3!9-2!Whno
z+FzPJ*RmG9RcIH*{UMW(%9PKS`Oj44pkh;2HzSWqRr3V>-TZnl-TIW7Z`3X4VR5lr
z(~Z^6wxj=FQD+_x)&Ko*qDZ1b6dEnE(}zm3Or#_+rHp-KYb;qJ)R?8DREo-0mYF1k
zkZoicYqEVrwz2QaFf(Q>voMC=?GxXhf9G+>`=0lCpYuA;bMD-PGK5+7lIBft!7ufG
z1a4pGegIEtNFYpd3#3JVbmg)&!L%=*Ic^+OWrBf7OJul#owMZ5k>-K&)0L+l4SL?z
zhfj9*9u#_7cWN4D96B&i?rD^tj*jn$L_!8l@{45EPcqWLySKA+{hhJ%2i~*rz;PCf
z2oi}<W*VgpuN9>uH!|h<eQr#_+$@aeW8e!vB00E5efYwqb5~lp<9YcNp8{9z#i74W
z8=#Mr^jnmUMt0{MMDD)vJeU|V=8_J)6=MF9-U$aI>jU(jD8&A$FD`!MUQYJ|GWUJ=
zAYQ2OrX^0WA<~X+l0*LNg&Q)<x9*d_Npp{aWiy<|>cx56-NwU@RKldmAH*%<t}2Qz
zd6d}Ab7DFT4n5cQ7UdcO@lk#VBDO}NHE-AQ%P&FfYOCQKLY{-Fxv3ra<ezE~lw{El
z8ntn3ZC|cLX<mm!WHe7l&7&UNn+;^I6GDzsL`AGf`D_{P6zV{H=7{>D^@KR`;xH&;
zx}j$cO*+1UL_$z|A@_-68Rio4+QLoJE%yRaQVwzy83swezKHS|cLz-nL>eMH7r4$U
z8y-(nPK;F4&m%QgVR;w9kDT9Mao3ucd4d+m!9ONLk}n*^S$mT|AHUFwupEogT8W`D
z1tiv*Q7hX-?O;e=)ri*uBc--EpkPX7rc9>1kNl7mSJ#;YoC$~0DM$;UDelThKVomS
z<)*NK<?7)uD2w#(Z-{|~{D`O)-DGgVB{e~r;9zDJs&W6`#hgbps&OErDlrE_$s~T#
z8?pxNCdj}k9#~kobx9v&;S({CVK)OLh~XFbP(qS{lo&!3T^-@B@`Le*|Gf*)Fv@pF
z7fz!!=U|Z>!<Yq)5+g7bvGo2wI<zm~!D(H!`)dT8x}+Hf-8ZW#>7o%X2ztE_q$TXS
zGteSH0_$KF8k|FHQ2B)t@*CtThY{{{NBOGKWw<|FgM>Ed51%Qq?dH=WE03gf)bdrF
z#daw>ceT_O{Bt9VXv)@1J5n-3eu?Tv;Gpfti<HX`0Phuv&R(upCTKLK^@f5A(qNs-
zJ3Y%l{OXm4lC3BG=YO0OD%0Ebr1m-~X2Re@0Dui}*D7!&kytg>--s`xA0`Jp&PT0G
zbx<Yn2i6QdC;0oa*1_V|y||N|rGg%136UZ_oXD=y3Cw*z6I6xJxvtrT{3aPL8*J?s
zg<hJy^bRyC#u$R|&s_3hEuH^jye?;~egAFH4!ulR-SK(}Va;8tyIXE%u@;ld)k>)a
z$^?t51Xo+ILSZhkyG?v;To2V9mT$i>9W2GKDYvh^gX51d|N2=qc|bW^2zY#RIOr_^
zNCBj-#_Ia<|LgWajwWncU><uW(taeLvi{@Xt?{12g=OVvpxf^xUMBe=w#1`OVWEt#
zdJ?s>6h{zn4)OO0T#-am<pNi_vR6MfE$VxvITrU?SS?O2?CcU9duFg!6{(<2q>IDF
z3<r!AIrCORRla4Q^igu~y5b3q0q(-yi*#d>2|(je!#geWbCv^MgzcTa$g5+ml61)<
z(L>wY1}`kXV)iu5t1JPECg6RcL1?>9ZgDV&JFrlOkNV3I(=P4f2WqsGMZV1!fHqZ~
z2`@VknrfM2fDR0rln7R4=-cdj_hPlYJhUADIzijFew0`hbvwW0T;NKxNy{YQC5x!@
zt1!aD6OOQKX5m?NW8M37y2ZhED)xt^G9N4KAkzETX^0uB*!@C#hglCwkrrU@298HM
zy&(8j5OXNT^y!;gVs{2E-#$z_9r9NDV)B5hi%*c%(}}hNYt8yAYkm(3R#$<dwn%y9
z_o)9mzRGXlpT9`ot?J7eW_i@CLArzLaF~wE!Vj3#_g^o`WrJpAMz?l2$(2RVl=E2;
z5>WGgdR))SyiF@%ytndOu3cWfE=h)aE0xIdd=ZT{4NsMs0}9=S6wpU$L{ml0a-zcG
zDc3`UVwg`tVxKk0l<;}jj7)e=*hl?{d$VRUX2SdM#;Fm1Z9;Yj=hVL41iu{wDBr{S
z#vuD78%JBi%&z`YBV1+55Ev)iGzcMyYF-2p{>SGJCb!>W1*0>Hs~$?vyebp_WA2UK
zfGYY@x&v>dk>)>6;aUl;Ssd^KR>ES%g6#LA08m|(*3eE2W`<<aSTsv&Jc4<nv-Z2I
z0f3h&jNckkj-Rd)uApw!vMz##Yihd&$PW?&claJYHBp8j<QpXB5|5=Aos^ZIFiUs(
z5&GQ^(4Oy*f7WLXN*`(ckWM)*qh(w@Xw)0xKb#@R)kF}mUKJ30HzX+y_q7w~9c4x<
zxTE8{bBIyf0JE-ZAt;@POPOdH2t5_?Z&6?5V&y~cT=r0`yP9+D6yOWTz8=U{C=5Zs
zr7_)WLL()i3hRXU?BMnu8GtmVjh=Y-fV=a|AKPB77X5T4#F><)q{bdr#S(R&%B$>E
z*!$-*u19)GdG6-{Xb*Cge$kHVt_~m~-+aik3HfPQl|U-5ijOM~=m|^!uX~~FdsIoB
zgv4_NBaQr3^lnyHI6A(mXnVbj>Dr{U5#Or4HIVqo^fGZ$nFNO)!r<HFi}$p(k$V@&
z3bSa#5L0js<gl*q6?TXA^j6@3BPxVis&Zc9(vcHA!S_61r{SQs-Vi=fTS@jf(^T(3
z4k^9<N_;vO*2?We=t{?XitIEHI?_!nABI_z(t45dE7-3G`M=&jEVASNdGpT%-WF-H
zjUx;S`s9&{vr2(nCFK1J!T-sbtyvx9$tlLvwr?-o!?S`nP78)PsY;rO%*2*Sa<1kk
zYVBQ)kyi1Jwv9dWplfC!CU@28e2B)z_eF_y^G)G(J28L4751wfZIgVEZcr3@{rHpA
z9#PtBzU5aB;3%nG*><#@kFVBZEa7|G=@vCb=5p>)cgT~_^)Gv!)H~Cne_&ThC5Nb3
z(Fwu|TF(z;MM~F|i)*gY-rJpYrY}m6n|J{67QG)(rFYVYobLJz18BP_d_ztGQ3mzq
z$!yvJKW-cM(rcvH)rsbCSfmbop#k=#JDd0+V|Bc(hY?{nV<fSl;n5~p8*^L#VN4}C
zuYDX+7fNRkmIV`sBk{vdaexfQiDCTTywA0!MPE3g6aadq%zPIHI(M$8bn!U%W?x}|
z)aMgBM~%Gto`Jm@C4tvIiz7g=eb&@;7n*|I0^gVGdiD4lIabbPsAL=a+O_#?gqph~
z;hR${p{h4--L`@h(-?E0{m@p+EJ_kfOY0q~&=L$&NQg}5oSIb)|4;MsDrn<jZc;z<
zZ}aXifb1f_7>YbGg4&Q7DdxCW)>nw4gCHY@0$)0vvUJuwd6A1~1giQ9GC2&Ci;v73
z?%;C(jSddhi9Inthv%O$ErD6j9j_%WGnUTA2qEPxaU_@LcGMUvjt$92`trcfkr9zP
z*q&)d(A}^3S@V_u{@%l(rb04?+Oz-gPKD|W*dhEjAav{Zq+o;eX%(#vohddv2QO2e
zUd{CyRN{=hp6(GlW*hk|Z;0F<T9(p2+e0*<GY07=lWXDL@K5uED=WM}4NGlJOe1ba
zl`p6DKg?w>UITI;sC24TK-}<SHbay)yQnvdn7Vn^q4iP9%h}$z>n4T|16147jA6Th
zR@e!~<TB&8z+q_1<@flGFSJ_PY5xEp^WB77v+{EQHRn!QtTpSbaRG9`>~OmJ6m2=S
z4Ruu24W;T%ztV<!^+?7=rYoNE;k8lxk8X-lZq4&atB`YMc3izxc*aTKrzr#GRy5)}
zE(@=IUMejyR7!Y#UAgNRn-vQ9cBn!%)&Z^^Kb#3CNAv*mE1bsygM8RQ-4~5gWKyQl
z6@2Oy<SqH=0lHf2Eiz?_)eZ8v1|_2)1DwaPKJ`Vy#RF;&a`j`O2{#;^tb0~{OLc<)
z5gU#U@ZFFzYL5;5+fI!F8A{zQwe|asWyt~=veicJpv50sBV6wVafzKu99x^~%$$+U
zR_39~^yM>^rGdMs2tNPHaf?-z!C>(Ij|$-91;Z94Q;lTm%B3F;!MzwW%+B`UR^_=7
zh@t$jN&)_p_ghyk=Td_IwfP=lLUGriCp-QHw1T(Pw(+Osk%p2p42DIfmK8}M=TC`j
zsJv1Xws>bZp$h!a{BON!>ri*fIM<d|j^(@xD7de6@YPdsj#SQX&lK1%4ae%|aU3@W
ztTa1vEHb>sOV87~dKIfn#}JCx{v2ThRo$pj)~ZJ%lL{0h(Z4+I+bgf!v%xxe7SMKz
z2w@*fuH-j9dF{SpdGtib_3Y&vd(cA9JcZq1%Iizb3BI+Wm5Saq7MbI=CljgXlkdoq
zS7pvZq21B(LcnpW;}!#G8>rL(DE7IDrmYJ}RrpWtO<h^c14zXnV9^|#3kPi}=>eOL
z!7b>~pDkXsrw{tgpzT7dM2IoN<YNUW_#@2uK!K52pGn-V*;5`t`&M>IzFfUDL(=f5
zAVzmNVSzZ&%vR=gddd8Kmmr8}1dW+;_v&?N9Df51-IG>^Xyq5anOooG0?zI;@g`HQ
z0j-5TpYE|U(l+hzf`amFG<^5Z7@S(Wc3H&6RUz(I$Cde@j~!bm{;pshPiWg^1;Hl*
zKDiPh-#)4q_mT1I)5viNv$DDn=UP)>{*j4PD<K$yKvgV}N}_H=Qk^V2)F+%!YXCH%
zFdF;T4pmSh{GOasgp`aos<B<B(2rmwm;N)0u1^HU9{)U04*jUixY1y6HI6u&Z}8=f
zs4DWIrd%A}d3O0+btue=MAY;YCla59k}?5CqRc&=THTH306q3Xpzj?+jA`7WCJhRL
zJ1F>hbK-IpX8#~9I<~J-iV9=tmtA9W+<Qaceh?;h)dn782h1@!?2h5mN!0k+jef71
zK~Fh~StR-a9nWc=ugdNv<1}dgt1puy{v`@{HKhF8Sc(#WQ2i$6YE8dGpLd+@zRSHb
zpGJENAx!lN12Ff)^bY3ynfoE$eNFI!vVv5j#S_yobg7#>OnoWT;^YYA<PZqZ=uk!M
zT3dg>Rui<~=Q4qql25BDaZp7G?fG?Me`wYaxVP|Y^{5R@SIK8Q(n8ZgLy4`4(U=%q
z2&Ol4s-7j)IU<{eF6H1Ozs(xQF-f>AqkO+;(`tn14;JYl@{E(HRaQbaO8b7n_-A@p
zE#sWKct^qd_+3I`eMGX~N^1D-i@CXG5``pYxd$!zV!<10BMooz32}%uDaLBejm#~c
zU!UGJNd+>fpZ7Kok@3+RHA(9#CwwntD$4ASdi7s!IO;jL(KXSs7L#!|%%0>=_&nAs
zpU}G3ex*8KL3qR2Uj%3*x`dWK`v$&1GVxh)1prBPOx!v<kcWFwY?*O>_o#`=8!yxi
zKv1+@*g5s#>U`QEWL9^p)?o1AJL*p!&webjZ@3IqULpHCO19kzyheiO=IcK1fE-cC
zS&jk9y-#S&cA4EML|pzx%pRQ<sAU1&Jl%W7=PX_F51sqvC$$T$?UQ*={Jr@4=P9!w
z`)u~oE-r$h>g#Jwcg~5*4+{Pr-}3!yTugT8F>}n_1K{7!K(^;^kx||6KqV6)e3F+m
zp3X8L<KjXjTeND;W$|Kx=6bXGru1=sgEfZinTyqB_>3x8KCwqfL0(^p(G?EM|2GMk
zz$v_nQA&N_U~lrb<6Y3;6Qo;BTnO6u=AdfjAAauLpF+ucZ#_Gwr*7$H;0M(xp(1{#
z9VFvd=`S@(L+s6w&MZ}f0}Er!k56@zDg+%xQ^}X#f#VZEk*KMatMgXggi4Kd4;bvn
zhOU4#8#a4FM&-gsd7Gyk_ctx@jy(?i)RWN8vuCZX;}ii@S0+D=Ozc*mRVN=>^YmeT
z-8$Vyr#t6{s4SSHgX3n9jA7L0KU-_n!ob|J8;!j{(2Uy0v7%q{FY5n#0BRd&okgUN
z&!E^29W^P20#=z%KGNjw4;c@ly0IF8OE-izzU-|Q5G7-u&fbSiJa*<qX=QIj#=@<I
zNNegHKCb#nKYQ28&nG~0((la0P#2r8uA`~Ugw6o4${cn}sS5rs4wWq~PbuYLmjSS0
zOfbN$0VeiroUi=J2zB*Zp1-)e?{_<2Bl@&~tg*t1tbo%%l_ZXoT%Q63h1lp*boBXi
z)$0t4?I3N<3d&4pnZ%Hbn4P0U@cy8M!Gl8H-XsC_4nJ3|g!<;(wsB#gtUanU*Blho
z4CR;tk~Do30yOYYONd_7R!BMC4lfM}L2C#;2?0FoxJtDN>`(ezOFNKVT2-9bvlP;;
z!j7>*%3fC}tR&tYi=3s^Q>n5@waJOdEcM`FgUZX&<25(p5dGj$QO7BLzINXPdj2<|
zK-WW6Q`3wAMKwu(Zgb4lDe2e1D_)kFyXUab*Tlj<qeE1Lo!k-og)P`Ed0cQu_h4I<
zyPc@Lqn*K%<G%{_{a30U#kFx~yqs5Ud7`kx3j9v*#gUHvk}G;u*Zt1i<1h~m^<<Cd
z#XiwE<TEsl@1B(Ao;^~OP#ZmF*rcYB3icQ;?~%}>cL!aa3JD2m3^%aV8HM!0Wu3eL
z)iT2p+mZoIi#5yol(W)sZp~&YkYo2m@itzjoN2QlR}Uz0+)b2j`5oYDr$8GZF#fFM
z+2x1wHLsz2La4rz*vorE?gw{>T-2*mbnzT>1z6yIQS`U9H@8DWrK2W=F;S<?e!?9y
zUT}e5Sl+pPqK(C#IitqTR~Hh!)*Z`i&PTe|6^Q*Q0eI;VPwy@~Re9<z6ZFn!&2EWN
z;@d;~q@k&*wR(*6?ha;SUBoW;Y&_7@^u4Q?N(-)VZ3max`7SXHFJYkR|MN2)!8VIM
zDQKEDFwdZJ@85I5F8NvA+3nzhg4HJ(&KrtRySq*}MH1qXz(aBY!!%Z0hk^>XuK2`S
zWb;OxI0B`+Ek1P6-H8nB<oV440NfmEeB-LZOk8FRL-MB*e4BpEEjB)O%h_-Xt-3Wo
z{qF0OLUar*!L%7$=?7(?2SgChCief;S?UtRK<7+~B!J4b0S&$@N~sY>Q7M@hdZev7
zy9Znqt1|XI)g4u+4^~<>&4TP&k^jw$Nc{Y!RTxkfkcz5UA?|kZH1**3e}j#qykqdz
z220V#_pTbuL;`g3x3`%}^_)O$bh_%E#;3NXYehz^lMU!z4tyYL)l&kwZTu-q&mi4W
zd}s<0P^e&CO5g_pMdogx0nd2myxwyaJ!ik%?;$&MLqTc2%ZPcBk9Bf(C6j=8XPF15
z1MTlEj=L{F={`rFZAi}sw1RvDN`0BX0UZWx;SPFdNqXGujAp;d<Z~_a`Iyomn~~2$
zWKPCEN@Rf)T;_Ig$~d}kgLazmcye9+Q9?qbJnLJ#C-!4s`%jNxJHLXPOo&WE3Tq8v
zxs+)&A2X`etm^r{OV#7#1P6}Rjb<(M7(mtIeRU|yA50XP9AFf#f#ei#3P(Tm@sxKz
zV)#2*o>>&dN-DrfU`@8vS-68Tdf&(2o)7upaust<p?$Mj?bj)y7t*1qAz*@1f&F+h
zF*eSwVZTe_XGl&9=(8o*TSE4KxBY?wa6F^|P`j#0n&Av6F&{z{nF+wv+?~4a)rgXB
z$)z_nUt&(bKS?Q4UM`wKgKbCsME`&L*b|5&=eY|D1D!3$d2fkwkOz0Z<jPwD;20>y
z{D8s6H4&brJN6Hl_D<SrtLW<fwO0L;TNKgHZa9jPj`4d=2LoQz6@fB{<ZswmGNqjk
z2FJC}fReyF#t{Pj=9msv%ja)Sl(T<X(iFQhWLMMYE{*;y8n*{%iCh2qxBFjmxgRZa
zS_(Ov?+Us$Jzd>%Pb3Y0tC1<z#O%vx*4@72Pxl-o+kA7Vh!GXu$_xipod}dgq)@|#
z5Mb@x7-{>_ZVDa@Ynq|Px;x+Kr*Y++#B^}i>WD1J6V1P3NYHeGW1!l)mN^WaOyPsm
z7Xk`y=}-dg?kh5_&*_goT5K^=-+Tr5QIiyP8b7wxjeUMna7Qshc4dvriV(yUI;c`*
zm^!i&z%`Gl!zpAlaziFSfwgCgY`ab$EWe2RONWirljMYNAD$sH?OwB1QffzmxSk}!
zNsc-!&dIp~>~Xxal!kqaM3uWz0+`bE*FfQKyp>l{V3I(@DBAT3p3>K+2e`%kTd93c
z?#d<E&WSU}Az%H__L)S~P4$?PYcUeYzvsH?w@;b2xmD^Wr%H>>!;{r{Q!oFRILBL&
z&4cxn`e_Dq^Ixkxl^@t18fh9^H2{9~6qxuk;ayJxj5jL^ENzcC=OVBZ27q{Q9HkU*
zo=y7oHQk5v4f109p7#6~KC1K1unZ=@FPH;`g)6WNjeB2mkN)u-p)pBnEY~$DhiuCu
z^W$$360XXxpBZhDPYlp!wezkc%53J30QR$1WCjAx+6I_q-M45XA?wz>(&{x4rPlM5
zu47wSanAsi^Q*?bWg0VwBm@sEbE5&Q^wSwfycgjD#ssn<ymS!0_hpo)RQo@|t}i&_
zTS+Y9T4<@Qs6gIw2+s)r7zk*bfUi?#-vjydTGuJ3ls2~`H=AZ`8@ATxpd~|$lhb=p
z#d<t3zb0z!Wrd?Mmv=Xyc>$oJtvW44o{Qz`H3q`oo5P(mEhA8N+e!Nw`5{kTzVx$2
z$ystmE6<1>5CZfm%f~nu3!LPD{fedFgvVBavS1ttiS}5UzOws`rL8q4wATbN6I0@8
zs~b*y4=$G3tPMHi#B|Pyb9QD}<|OiL)M(~{{>|vhgbKATHM&(BY#pHqAEpI}udz%!
zp?Te*f=IAkuX0cYLcOY#I44c4ob@fBp!lXYg)nmJ^(>rH{Grv&`S8TX)`}-z&4yaW
z0kdr*z^QM?<`W&-S=AzA&}U<62I8FKfLjjp&{ulWr_bKjYrmI*yE=GQGwZbKF8bf}
z5un=Jse|R?`<c>ze?|?Mti<i`7keaXYDoXq9_?VO$ZUKB*jVmE8zhrbGBIE7<fn}R
zyQf>u#qQmY-g4B5V-^ajA$haMlA0?s2Wg`{3qepG<hg^Zp`sj_=ANf67-1ko%JYow
z8haMgkUIzr7;CT%{nLEX-<ZO`L!<u^b)8G=ZHo>Muo3&l7cBGWnLUJocn}-TGjt#l
zIs2_W#=%~ZX?!Egs(OEob~<%sb=m=q${$8rwG<xl%B1+qR~OH&j_iJz>~42T_|V+8
zBkcY0J527nYcUEwCL+_w+|k~aQG{{%Fk=Fo;Rs$fQV)b_cgAVxAQhQzCjkY!w$i$U
zIS5S4l<{DN#%fho(=`V0j`J(!VE4q6cU7jRi*!aAL$4~6PJC<EG)MDX_)n!@+3}Vk
zCoCs)C(}~-18ioDD|g|i1*QdWSxPz0!XDtS^{hWtg$!^drvu-7^ni=#SRSBJk1%N$
zX2i(u8lnBpIKB@z<CmY_)^YOq(O9Z6yN71~OL8P0PZhPxU~JA)XF%mxR_|l-dOrZX
z_sp__t2<>^7D&ucCr#nW*H{|nOYc`dAZF<|23gE%_Lueyw)F((Cs9S#fDkqc;vY|t
zPU$_4cw!(!)JZUhkBpFt@#Wl+>>;?+E>Pgx#_*#CL=SC{e5pWQQ4;WZOQQ^-5g2Oh
zv8b*%%Hn04@ZiQ>qk3Jh)w3TEhLNv^7IdEktH+?UKt<o9fuY<xU8nr8$ZI)+%|VD^
zHG@0<fLv{R!3!maM^BW7T8uWk7h>xEd?g^#bB4*-SPn4ABkyht(j&Q^=jQ2U#J5k4
zbd!n)a($(Ny8GI`U6>DrUy0T%FJE9%#3Q65W;1W2){DTjXbjryWK>$SQffh8#p6c_
zv8~n4I!+HbfXf&**8oP#Ze2Ap-xyrCd&pB;x46mcWoeu#qg-gp1N@m@vM|mwx*hg^
DzlO!b

literal 57176
zcmYg%byySJ`!}E{5)x7(p)>-cTR>t+N=ps|L>eSU*MoG5bcu|v(KWikQ6n}QrD240
z<2&_vzQ6an_-lW#vvbaU-=8`W8n0gw;nUz_VPO#|DavbMVcjvs!orroyNCINu|<Lq
z3+pMClDv$LXXZ}U{Y-LQWbY{YOJ=Isz*DgZW>V}s=8V)RF4%UM-RptvboXI1HM@C=
zN9!%2;Nesf!qk_Vd_R;)hE@p$Kui0Sx(gL1A<%8;mxr#uns?pwT1&ij7X&7^eD<4g
z$jVLHD5Ir=Z%L*P1FF`**G(6OPtC9MegS)u*GRothO#5T@#n&_mp0K8{V!+Zz**pc
zaGrz>vg{hMU+2wEfEaD!uLObaeMjeVcMN3$tc9??VI|t7w;2%JOa?L%6$!i7(!9L$
zwO8;h3-?&y36?O{>~pp5{lHs&nkXB1;ItNzeq`#Y00e<sYuufjWw$;g$bD=4@z_{}
zOQ63$+wPFYB8zojPas(=I`^0Dx7)<sHNFEsuU);GJ(U2K;icX4?fEiP*SyEIve?io
z!xDprnmxK?3&VyX+V*|03Hw!Y1$H+tVK=WTK(l9Gf%iHR0D31D&aD0Nfv)}F@(^Q(
zj?DW7Rtf~})8}`+vCQqT-Eha~UW(Xzy^gazfru^ym7X=88kv+ElHEU97f{n~K-3Qi
zUlMbNcO~L5FH%}};8@RI>#JrETN9YaR8_q3m)Z}gle-iajiM)($-`q%Bb7RIDct=S
zC2UYxyfP1@%DiiW4}rVH_(cNdgx^w!nqmIB0((=`Rks+)5tiR**s$M`s%rHpQ@y1`
zB<b(7!rt+*y_yIHMk~qk@{PU;q-GAp#=3|1&p(0F3Rr9YBQJo6{Q3EAkeh*gJlKvY
zNciPME%s$l_yIs3QbafI(OO^zrHt}}VrSbK?i@`Q47<MH@!|)L^KNHdEsyu>4v9u^
z+Aumh-`fZ`TK3(Un*(Oc^^B;Aiibmczzwge%Ywccxee~8oPerdHS4?aVK$SnJD4Df
zhanZ`e&wq~M#>zEIC-%Gl1pVi@?^hKjwqZrRaSVS-qQCUJaZ@Sh{Z8y4vck)%;D}F
z!$rq<*ml7|PZd`(fho>h=*=FNn^#Tu9M~DnH<R;8<-e&YxpB*IoyJl|wNgd(E*f36
zGpzf%%&!p%nin?gBj&a5mUnT)S8u8dUc}U<p}LgVn+mAK<&l}tVs79B%WN+s>i3RP
zvr2pzPPtz^3HF+@?ltA)2#p0BKMWdr)?2BH7fntgfi_nOC&ek!(jw$06_?I?vg<+y
z3d|@{<WgyIgQ5?~)M9ZgqkbBc+suZI)t;HBpN<sdb1r=W{}g*-eG>lI!>(l0xfJa}
zrj~q!$h-^Tf1R*iYFrvTz_|kh@HDRfSwyOA7?bJ4#t6XEcv)(ym-V%>`!wZG!Jq<-
zs8{XvT>BhP`yTw$dbbCPQ}jvunPvRcB!u9htegquuL`F-5zJd&w-B|d^bUHO9(p$!
zWkVG73EvHLD?Kztq{2uVjBJcJ8uEIW<zHCPDYm>0qGBZwK55Y^M^|T8NX{miN9L?o
z68*hjsjA(}DxHq{|4gD`FHEl#RpzV9<X`Rpf~VPVq92h+4qrT@7n9qB5CB9Dkh?oZ
zv$_1^iVPf~jtD;K_te||#(lak!&?QdeSmOy<;`((+k|;lF`qRL^l~-BX6ZG|nJH%R
zw0us8Xape^02wt9^&aMSZX47hY=ZuK8Vo;b<?d|2psGgw%}8((!9P_v%Wt3cwBZyb
z;nD?RKK$%h3zvM(6mvuZN~?@j&xP8BP)?XcKRZ1Ry^wCb9Cp@&3KWXHom~z7C(Qer
zw`|2P0uY8S#N)YW{kBFf$0oY_>G!#($`;N<RaBO(Xjh;k32KW1r{(ZyN&gYmz(X(w
zf_)b!!M**~4<UC?72|=aVh_kw%9FCmWzcPsWh;8N>8GaUIR{mi+DiczgG6z{saE6r
z!vSf9Uxx%l%2!-LZfTY0=D&BeV9cg|nX?uKJuYAIkq=x6A?e2pf9Q_()FlZ*!<LUu
zpF#5~GuEHNxZn%Gxwv#npGN1YS)VqkpV#|C9LIhu+378<`(B1npU`bs;zfr`^n7qH
z)t^2U^16oH^u6E%XPJ=3z`<Z}OLw)_vmP1J?~U9i#w0lPhCC`Dszhr&*0Pl>M*{&(
zlqFO<*8;QEAHxBK&h7^rfOO;2(q{#&6n6G4R^1IzUkp4$O&AJ@w@xn0wGMNxq$XXJ
zjT!1=tqyws^u?TsvL3zX$b&5u*kGKjff9#J{^_my%UZX}Se_#C?jHD0%rENyK*JV6
z4{y7p-+xtvHl(x*mn1i8F#GoF;fr4?7jDh%{x#$uR46YvjnU^{|8J8hmpRK|P{Fbl
zk4LhZ+)qzq$H+>-z9W$UcG=$gq54t6eCiVSz2pn#m^b-PXWw-e#MVrK5kV;y8BZLq
z{rB5d6Vw~fb|N;PMmpWwo*}_A^)k1e%m3C^;n!Zh7Y01t$O8bRR1qsoKKMK;aC(HX
z)*cKxEz<L!0?%l#rjLo66r&@f82iJGrjZB9UbeILWM-u%=@*mLb6;lcp?inhG3hje
z=4ej|uC_nb!r_Mh#uv9PR=okLLT0Wj``r_pkAqR3%7`wF`LwG(7G-QAln?ezK!H1-
z^^OfHAp$@t+?SRn1Z+90`lx)Hs9J?xq}mcs+Tnza2vpdkW~x-gdRxfB5ezc<-_~%5
z9waG)^c0Z?YqH_mOdVdb;mGydbRSSkiZwo5TxtpiKv(gb)BQ3k%Q*;DbKO`vMT;e%
z1~WV|FNuu=34f-TZ*rjy87NKvv}1B{T&6aDJGy-X6mTSj4H!reOtO17m}Pv;1l5Z?
zFhLaPpkYJ#gTRcIA(<Bi%&?t}()L)Jc59*-6f}cx1G<KigqbXYu8L6^VGXUO!_Q_}
zy8pM)PkT~0>`AC}!U<9<w_(A<7aEE3Ph%-?s1%-xi_P@hZ-`Y_ck5hZU4&)%!QIMN
z79bX8NU&IpE4kW-*pYl~)S6I3&xLrn21{hVO%;ow+{zNF_4MF|)5**B<BOIBPL$oD
z7rT?8OcdbJ|9()suT;Ay{#O<u4R64Mr;@fTcOi_`YivU(#x?*W;{ghBj#gY2I?iTf
z?8(u5wyqG|$iyhm<`uB+Kg6Ea#0BVEE_m`@kh{e?rQzo|Io~_33EJeeEpIKu?DwQx
zhwkaXW9xLVf$Cy*HN)_9+#2<GV=4=N!JhgeQXe`fJ21bvcAjqVgq6-zC7jReOV#<o
z=Uut<(tBwB%WM)lO=CBsUXyMqyR-QvVxGMz4qMy(?EWmKp)|OYdDn7VWys%ZR&{;K
z!(^IP9ebvU5|YJp*1k*qtdND$8>Hu}iq8>hHU&0-=sz>)pJURI(IQ;v7dO$qXkR}t
zuOhZU3xBzpdqy>yc;b|iQlULk;1v#t_5Cw_wKrFj&JU|&)!U))3Gud8lI4Az^Y-v*
zmtf5GR5x{Ng2F$6TWYMeA>OYGp89XtH0>d)-e7NhxFJWdw3}WgS50>lMFli}n28YK
ziZ7hIT(iqaihlcFLr4%==JW!kkY8vqYTD@cQvsijkl>`?g;U0BY{RbMfZ;V={vT`P
zd`b;axN)IEzdfWM_%#@WT4R~Dzfwbq(~rapHE&j%O6z(SujH9=VyYQx_-o(hm+{q4
zyNs2InM0o$uF1PmJb$ncepY71g;sA8=IR@}CWXH>uxqTJ^aaMCG;<5!)r1Eda|$8n
zPZ@h+NS_(XTgDemg0~n)bD?nCB2pbso8%{+Lf#O3L8IFH=_=xxIt_UMfrqRpaPK?B
ziY_W2`I4m={W&C|915Rb<nQzLZGsYP`_rLenQ{AuDBLjD$m|rjP|_l{-LCO6=<^cE
zN-DR!^D!)v=`3~T5D^th)1Co?{_?9hy1u9M_|JM%X<4l>M_Pg=QwF8Tkw;&GL6q3H
zT*2&QB$4MT@9DA^V1NPhj^uULL7d^Xa>;MZ<)fd$EtXAk$Af&mW5R%UXwQsIm&QKj
zJoqWV=HVPOQuu4m0W!uZz<p*LO>bIg*ZrOZcV*zec`y>eQvrh#8ud?%E`)S|NS}GH
z%7Uog#<>+6;ZlGe<XEW3+bN?x)zTFipBF?*8@P!%?Q>sdhFDyus_CI{3H-}fgm!%#
z7W%Z4>0Jgh4s0kZ@htv#*4j2<2iD1Q@`Ksn7IUI~xVVX_f=$SoLa?Xgb4JPIwa*u!
znSF{luj^w|5(rj~h@Q4gKDj;@Fhw40OO@fs#L!rN6PI#)P{C3BF&!?dBpPBy#}GBD
ztHxJ5y2LMH6Yya^%u6k+Y^5AEJ#XTt4$>l|>&cqn3El)ehnv`(TrcexH!Z7k4-3lM
zlRnRk9QZwx2otPU<yYZOkvZvy8>>3(lj*!pXwTUtN7C>9&*V4O&FQi|ZQ}UTBtAr2
z_r~QGJpVRll&k9xv8)fzSUC_1v#{(JU#!=uakOfcJylE*M^dGqeyo3g^FW2Tc}I<d
zsOl@^E$zj>lYG#275-U)9hf}XX7rksCYg0mgVA^#E$=oY&ib4&U-Ow`IR1gRyu5e<
zLc!PfbCKTsf(HM@r-?@iW-+M)nkhY@r$yoGPgw;s4nETA{zE3F#RRftzOS4&fLzrl
z6@QH=NbTlZfV|}k3pY}&85yMMt^N|mJ&r3TYZ>!;2I5k!I9X;fjK+tbRN=QYD(VFP
z2HFG_c?bZ?l(H(R2<lmU+f?dUbUm?1L5pkgXG9H3S5sanPl-zf!vM3U*?2t`vf&Oj
z7IYN@&y6OP7e?AF6AbzbHBoDVBlMP98}<~S*efC})LLa6rVU3Jtcf@Gl=<rS&&lgq
zThCg>v~x*&UsR#|brm&v|JPy!y;eED;81snFB+m)xHP!~z@X$^R?LtUg%I#qp*>xf
zWDR)=Un<DMkOzGNreIJ#bsD2J*-FqPI1PE=k0~)m#dMUYsC@d1ZCDvXIOPU53GV!H
z-dVP?unn8f1yTM}tN_b{0XEKr)Pkw09Z-j@msWsmtf${JdH+`0_1D3kl&NA3wUsD#
zbIxa>I@lCcO5@^wX&ghJHc|Q1PKmFA8+nw8<xBF8<<t=r@$xCN6{Q?ngt~~eStI2w
z6a_;9bMKfx*D&K#vLIYx1cfX02CKJuy_jc&uz<H273koxN`JUPy+}P+S-KWPaOMv>
zwBtcgP?<QB7CENzJqLPify%%9x=JjWATCP=QWN~UvU%neo3LWT{*0|4<gM|;iu_A1
zBlU{l83i8fgMB<+(G284+9oQg@kc}9fS6(b{asZ4PQODYt6I<VLqRy7ARPRG#INMT
zK^kOJR2>d(h)t>}UArtoO}p<-c?=r332~&KdaW!k#TxrB6D_T}(l8d%@&B$2?ag<B
zhZGGKaZ!1}d_bMp_exvANJX6<`pzKN$ean`(4^}{(s1Hjl8+AkT4gIt^vHuwCQzjm
z!PAy`2tQ@#k9hMUF=~SW1fNl!QiLonrmPv&_yoxE^VndRF?2@ID{c}K-UNh~X^M5w
z@YNW%lb5e>#{(>YPtoV9&w2u_3z7RsY3ISV%+P;pwfUXAZ*jq6|81C#h+I=W=k3G}
zoML}8Ns%vFT75@FKIGi)?(#)ANu4byWvg=Y4JyBY`m*ym>>V~%m{0_ZBS+|=aI$lH
zH~-4p*I<x_d>m*@G)hFod7at2>|xL%L<x!xf|(9SyW1cw@-S28i@7xgNK6YwLH^8E
zUaeRAWYpvWWjzODd@c9xZ*@q@Qkcq=tza;e!OIy{!M8CV#GZW6-&uo0!e7vnE|-Lp
zkk-fQG3q9-3<ULlfmzkXo>5BfArGYFvvboFo?{4_rz>R*rOcMLTkjH!MF@*{GC>!o
zjC>-csg}oR6R{qx5KJpFunOznYQ&Z;zEG}%a>}DUy#mVb_5_5GNW9O55>!S3uG}AL
za?c5m2Yf_=4YbzFAa~NiWfT7MeKzCrWsK@IL+UlgcdkzBSHkS+cuW_9iqZJW7CRq@
zO^2jC*DgG6-oJB6pOvsecl?i1vh5C}aD34DVX4k1AC5eT_nghCZZ{!)_Ff@)8o|*;
zQ6b}s`6t~POhRbS=@g|a*Jz~Y6!@Ytme?4@5<xGGJgBsHl5X><g=1iXP%rN)4B9iL
z-V!0EiK0lfe&>16vFC(ZyxoxM+ZRLq^9OHqze`JGGIicsehMpd4AR2Sew=d{`C$;j
z-HW%<8sFbEVs$)p-y`<r3(nx86ejya<6!2B>NWFC@}z&bXzZ_a9M7}nf4z+(H|Vpj
z9F<9EXizA<mWSJuiuAA8J?wXh)MA0+wlc&d`#FidnX6V#3}^H!oyrg~uR(@8CV}-A
zDh(J6o<|;AwmTR{=@LCT@jH)jAKXt;bDLSLdLciN-0yGXHC_Fy`O<*{M<FgVD$sfL
zKaRk@dC?RBcv+G~7I}_8D83lOC%{0%5YlLhr)f(Cy0E3Ih-7IF|3Kpuu^9Zxe3SJr
zNG98IC5fYhmmo^=$TBSW+}KOy9h5@oDayksbJf9;Z-ne+!^WT2vOkw=h!n2{l#ewS
zO;l={Hq<*m67Hv|s&^K+ZEzQRUvW;l&tn=5MC(Ch609`7a)g@TX;Rr}U&-D)O3AD}
z{t2f~QLtrpKOBtw@>jH2dO&``A40Sogzt>~s4AS^jxO0K=FJ!kx+OtpmoYJ&Myy6I
zJ5P%^(J_C>%V?zm0Uci7e5X9aGePC!Q(5N4!2pUuxJp2Eb2SnKru(Z2;*vTXGr&&y
zn1GeABJx}#k5?&&hR5ZOA-7Ww8m3yL(xXH%8Q>o}x;o2fKdlS@U4B)A;|I4#M=APE
zUb!x<Whga&ZF~i{8AR>LlTncOSnn@LBySUgJ{0f&X2m&~X>lI4A#Ng9k9?Qe7}<}*
z*D8773HQvmX{g3ChBh6nlyfCMaz<z2rgZb7NI(O*oE`#K6eHZExroLl1SeQ7waI6|
z{e`~D8Bi!QYSwa=ce&TmRoRsS*kj0Z+Zt)XES3-Tq+a+hR^B_i=6%vI`b&|@_0(e!
z04p2CCvaF<Z|9Nx-o;6HFqNu;6a62Tjx9Vt<b<tWa<O6O7ZBs;L*X2~V>THmj~5Er
zzZX`}J<tA>l|t(`9M|UveFzA$nF6y=I8?vA^DP%B`qj^LX9jlPy$HSP-6}Bro;%d0
zeYoEyyZvh+M~Ra;_N!)Egs{+j+qF9CmZjcJ_v21*ssnc4f4r$5vCC_wZ?=zO$uQUV
zwn`fb`9#BFl+N(Ve)v-qGpg&HdS+(Ky2zfMAoOSC%X!LwQpqC<dDiXP0ZQabL9br_
zsA;@>{L6J4Wog7Fh9%xaV*B~ySK8gkv^nt;^ylD|_3tn7jsPYTAN2bjK1?BfzF(U#
z)jV$rHz)0+FK=E6`SFTnoLcDWYTVdy2B?JK0H1yFP|0g6P;{4lM<xGv#15H9`KFOF
z2kW5LXc1aW@4Bj;ge*1r55n=oC75vfdvlHGIYK+pu(FEr826da^Ic`=AQ0i;)xm30
zHzmGAMn|Jy2j*LE?J}7F)91M?)<%XQ019rmkAyH};%o!N?9T02J&$|I5`Gdws}@lF
zGHO|?5T2#G2&%^m3%E}7mm1l(Ig**1{Q12kYER7J$ZE;yTL>QYtN*At*LL1893@l?
zZn157wy#d%ekt%zGj5NwJ>5R<eLb}284PeF38`&yq^R4}TEqcyZo3IY-{TGyig{Ay
zeZ)#w>=4(MRIWiDu^IYUu&phRwwE!|wwFXsTUtkKU`3^^qi59A7TE=t$1Gv&@PS@7
znNeGT4BxV~3!D7^sA%%lYJG(<!?d`Gqe<p#PzuK+8BsbnW8AH@0Tx>gxTSjJ{RUJF
zU()qz+?v#oLhq}kjo5Z1ywb-{WM{VIiEtl{Tjvy{h`cwxXtXqyj^@b|#Ms!^0IA5R
zsFirbcKb-7<v?PF{#TsItH)l#6te}PKFR8hmqYdhy#KcbDbrK(Zh5!CWQ9!|%y<<y
zS;*EP5+{Q@lHssfDZ62x!Z5?;lzSI>@EQV_>>cA4r%VR77^pYcl|rYBEv;_6oyKXv
z2SR1g+_mcgS|%#}HeWGe`{}@M-;^gOCpQJs1*{&n-JBY6nY2&?fxzinn}pk%+FBhd
z^Wz%zY^hCD1;aKPo$v)0l#<Z=u(v-O@M9B|OjEhCy(bZHG=2y0!pM-;^+NYwO<SwS
zrGGccJhctnvJV491>X5kOtL&=To+5Mg-U{gTNn_!nmT1hFzwbiztV+ME2rM-P$qrl
z$4J!Ag(RPdIQD3(ONL|z+&(_q7)6C3E81@(SyYmr(!@X@5ZO4lo6|Wb*mSk=9kq!N
z>IZcut=?wGn+VaXof=ILXxE#bS<%CB=5aS#d}}!M@ykTYY~x?qoXI?rKOvm?aMBfp
zp+C9_?+>#W-@*0A&ro2od=~LZaPbX{HQld8gBJtfiqJR;wL8K7y5)%aS~591iLZWX
z&F2HGcE~arF3O33wqLhbo`-{MU`-*9J>xR{s;<V^^tFv@OL*SdFf4Yu#Y-R=04kA|
zx;fUqK3|Dzygq~@pdbm9fP;Zn?5?du<@sk4OBSk0Y4A*Bewgtgj}6~{0P~Z^_`|Gm
z{3&rj7%F)nSrY#}sb%GCI{^zL>tnxbO)L>tP>Pt7U2PN;hkFhNqj<ejO(&kxFksLL
zY`NNN_P%`jAh%e%aJhNEZ6enYjjBNPQqu+}v6N3(q+O?<idcRlciWqPOl9Jkvyn+t
zmwIbpt{eYSCmA$xco3BWcoJ}7F>X%z_U|S=d;7=~19rnSP!!pB0_kvjS_>Ax1%rLJ
z!JrQ$ntNc-I;>KQP$F437p%k*$wV4$vp~>Xza11wtIom9=hjc>vB?pVOm80f>5#KN
zQ$FP~GqwS<cNX6}{(-vMh4&s`^6=A`Er%;ceH#RyWu3tfEh_x9FWIaA1z1XtqRt^t
z5XTrEVH2T*0e0kau3qX`qxdh+czMXtp6w9Mw%2@XUumFx4&HC=Gfbx2>OWT4kqQg4
zu~~H{aq5{*mFTi_a3FFG-US}OdD_1(`t0|<V(<$#sCmD|NGK(ryaXOJ(Qd^Y+0lz$
zsuR`~Xjg2a{2WqJ>uNN@LNC)9O>2Yo-`NOOPWb{S2Xx+JBv^SWwVXOE6J)@;exz9?
z3}EAjhdfx3IPU*Sx$E{Ze4KEmLwnnlnGBI#0cxhKu^R0gN)tdpfCQ6MQ#tkZln!(C
zy=O}SaTRTs%2awfQNTXR)z^wb#>Xc9N5dlejURFVU3kiOg2M)s*?Pwu=$y}ggjq~i
zntj=5dAoq$YV~*}kVFV5DgFX<(~#>9@|)^*F{Ax&q2yK88EsD26Lb1Z7*V3hxu(@r
z4>wRGq~qs26z!tU=k~}}rhKK?%UM(Y?OQbo+Swe2OH}P}xY08)yw8KU!14jm&k!n;
zj^nKf@DF)&+s+`|1FylLaBsVKeWqLjcJslC0cF<8pgii5Xc{5>1U7YK`9`a!#|&f3
zIsApHDGX}~YCtot%gsw;!BEk<zvhROmb9ot+$lp`vxjkaG>-nOal`^wj(v_yzkop)
zEvys)pr)rMz9Y&%glTD_%xZj-lg?_!sHx&Ig$5bvdU!p1oc#k;1sq{g_m7N*48A|H
zO?-BRKrc1((3q|4f8x}VBY~Ul>P)66gZt)_U{|HQUL3vThuez{i#W;Cmko{!zMw}X
zXX&lWZ1XkN!cxX<ccmSTW2l!}jK+jTUNm19wKF_2s&lWI3^pKs9Lmw5|DWupzqfr~
zwV8VYc~G)Bd#L@LC$Ld!2@Pw}BD~S>KdW$zQd)TR)d|cQK-G0WkzVoS4W?^}JGS4r
zx61GVn_BKBG4Lg+2Im{1)=rE;Y}&fIA$q2M6mt#@StdTo_?T|dlSLra*9`aXnC(-`
zX}up3SnqzrqC(tstT0}hM)vl93Uy+7g(Ai}u;&Oh`r=TGo|VHwoi~lBWt*p7z@Utz
zoda(`W2F!lg}Ut$qrO4{V1uGtJZWFes8m#(^sPP>NRctGiv;BFFCL?r&1PgJOO)($
zwpX6vL&zZYqMWlq#gFQS&u2|N_4opQZachh(@qr<RK(TZxc9|X8st9@cG8zxe%<c!
zw>?BNCBL2c)Uc<)2+A?&H-dmZr}e$TdLQ2zike$thJD53DU|3?)$eZx_6H%sh=9Qt
zE7fhyX6TYPMLfknVRw~sHjxV+OEDgcUiYPMjuT8OfI<evc_f#h{rb3f@TraBa6kfP
zO1-$;%%v80V>fJYEHBg!?Y$4DTFGpxsKygO)Gmn(U%F3XFa+r7RDk9&XW}K#|C9{<
z9@Oc?{-GsZwmFE3EMMhP>l=|z*h4fpMj+mtTHejFOT0~A{<u!$lp(BBXjcom#i&h4
z24b#-!nU6eWl89nCfmN_OaOxhkr!UOcc;n`xovIqAMX)$#Y*gjgWP9unO?<C=9+K4
zd}Ru=p)0t(q}ww(=2$Pg6f3`>xEoWY{>JobdKx3})gu*UqC;5z{|j~bYN$~H^S?{L
z*9%y&9?flOB%p}3w5r%MZIn=@y$r|he_4P6ZkS$u&*O#ATxK%0;!*OMhS5|$x&0g=
zhgrdeMLGcE?<M{|(gSZOlmOrF*4EZ`tjglmf}2H0AojG!l3&$s{Tu|sShvZT=HlD#
z-UMj-a_tN#xM_R3c0qwN8>&Y9dp>ICUsh<)k{JBorN8m;^-)em9vFQQJ41`pbbNe`
z5(Z(s>FhNafMphFSM+6byZp>e++pd(*}N;vt8cPc8+;~vB)XON=w5`QWiFHxba}cP
z<-F2owliG?!?VNm;}&8A4gdK{8q|=_{fb|wlKykG!c0fQ_s!vNva5IE^rUDDu<!q;
z5z|Gn(p;Rj{vOlrJKGPxYhz6v%C059IN*x-`Ar^_()2neHs74iDY8boc0Wp>nfR3r
zv;OnUnd?+l0ciWJ_fUq2x(BCV+#`iV&S)3_>{DUsXptV2<+1p9<BJdAbjB*%Q+rFt
z^Rke~d8$F6((|TdeM>aQ#!fikaA6=RAcJ2$(qE|H(T(|k!l{L(LhfSIsMZk$#%O)O
z5!b-ct!a{qF)<m_Xd$4IaIIb~)@|4jc-6%7Lzw?`g#*LFTP8esGWSF;(xG7?bf(4%
zO%xw5FP~3#;9|H@Puy_gh;I2(icbIh&z<|`%RLdIdyOkvQrnOdPUjuA(d8rEc0qM)
z<d~>kZ<9x?&AIK5P+mgO|7%Uh2*>@!{RwfY%1!<z!j9hcC^sP0yPQWa-Fo}qVN^VF
z>$Q0EOoX4P!X#K0{W-FufX3AK0GCz$OY}x|z@li__xN(-60~wX;7MX~vUwPtdj~F+
zaUh>XU!LGca7>2tAZp_0_hqr+YO$Jn_iiwzE3JOx-ERNC6_9s+*_R}F4@&v=8mr(=
zxOfDBy^E6$LGjD!kx_rfifyDym1(n-XcK%aF_Un~|H|Qd*D0X;`r>FJ(w=2*#O{f-
z^cvsmmDavh{9`evrMA{Zrk%h$566bZ4D0P5d%E-7@)M`{3J?*X$_I*%>~LtFP~BAT
zA0{~Z_LG=6-~SJ<^S<Hw_IAjJxWIyLA4n7^Ml&YF$69Bd!ZlAOtMoWU2JOj%0FGXv
zYCAWPk{@0+OR+ypeeb!E<*_{yvfp|}Psw9KEB(4)8`jdt21xm3ttS-$VB6l_KE|Vd
z?D`4hceb~1GTLgY=-GNYch=~=IHx~V$=5#u_-H)~0mrpfO~ArTji7jyFA+Tdhu$UN
z&H@bh0=*jRQbehhYhX_ZXVr@n7@&e&t`G)bM;@d*4;@<b_0?|0u1GkAU4C{6(Ccl~
zy#5i!a6Bnef;MdR2KzQ$2oN5u)v1Cu>K7fdP_AH58}GXTl7;WVmm#<|^DP|6aF3Lm
zKqb?bG6^POs#>B9dL!@eKsL^;uz&7S|6eUFJjovbM!U&HdIR5i$Z^bKHlI!IPM;&G
z$pag?Glh~sEk{(jCWmX<`xtkd@c_~#D|4e^(bNJK6T5Gou5PL(6*l$OsAMtt6_O+&
zR;&zqht>TFy`m>K2&SXX{o-gy;9x{zUq6^ia$sw0%d0wULSK3g-hW#2+JrJ-!&vpj
zArV<u?yEQ?dao$H&a}v5&aoV}obK-fX=AQyGrtID=M;lrMoqt_Vby^Li@Cqv(Ur+Y
zJ`$MuM<prfsEp*%zw#O6tKfzZ$jDPkmSFD!8mv=5Zk?>K+H<uQSPW|fV`*izy%i~F
zT)6(=0mAapyU&}EGd*dcGTgI2HeY`{-b^ta&{)ZI9YJW%4=A)%Rq+8ne}0hvRCcJM
zLo4M@CAWvsx?lkbgxD>MO)Wi7#I|5VX1Y|3XTEV*pRmX#V|T)uAF^w-sN+1jDb9x7
z_s6*CW?xpJj^{D(1_o!EZ|+Me!&utW=aE0icC{L+kw(wxcdNMtv~bE9ztQgoC2WqD
z!OahE6K{HY!&?|E=LbPy%gc+oV->Kv{kVzf#{wUgBiNRw^h|p%PtXoQ1OP<EJJ})M
zhxh-n1ZbZX#sbNz)>vSx2O)Q7Zq7Cla&e)!X8ZN6A6$PqPuI1T(vxy}25)Nk{{*B^
zFT)u-L@)n1{)8*k7u5ySIR9u7T3!g!c5q;vM_q)a&}8K+s7_}x*KfhRn_W&Dm_jQw
z!Nn1}F3LaAShnv-y<~<pn__SY^x?Nda%BeGUg+0a$MeX9V95^AqJ<vP2YpkI%vsCg
z{6sD%&81)Y(rL(dElCfOlhb+Te#dTh8QOWZGCeW}VSM>c(3I4Y<&mwEglj5ui|Eat
z=mvFP2hD|YD^w-MfWn4mZ?d-Mw!0jyrFD-EnDRc#O>7D*R`^`G3og^B$vxU6>CxqQ
z0@9>8p+e}_*;2%cF5^i~X;7LBV!VZ59r5BY3}_I^(cOfotYg|+pAgD8cMc(qxteW2
zeUS5G=uWc)t2@IYE#upG$cTdzV0)@6aRP((e&4V%cAGGcDX>ySv=z<BUEYB<)+**!
z7y#(0<fCSY%FlafxD6A;_?6;Ne#2-2pZy@exzkfi7Int0Ij8pCoG|*5vwcQRFX=80
z3g@*X_s!aQU!wam`2bg%31j(`q1ie+KJd0qrgPupR8eO}|I015rMWK)7O!^Zt6hJT
zeu7xHr*p<hg#2$g@?|{Qn9i)2&&^m&7Z_D;M`p30MOc0X#z6Q|bmWC1U#$e?Whwe|
zu4TZEAjurc{pSrI#I0^ZU7F5yV?<4ik)jp~uh=|f>X=;HhOK|YfWUCfQZX)Kfl5BZ
z9({{+Nt80dZ3(c1bENq4=-F=O_`h3`&RJjE{Z15v@ZZy*tY4~ma)gdR;21lVvzLSt
z!4mOMyTc&%ff4!6K<u5L0Y*d$I_&9F*vd**-#-$6n9|kT>QWJ!QWVM}Q8Ym%sYYLr
zIvsdGA5VevyMEJq0-hEPb){oDHBFB9G2##83~NVg1c<^(rt+H!VTrkavkkltK9}O%
zVME|f4+)0ab?_l~C&}fU5J?2c13nuuB3n)4P*?B0U;&@#mpaVtPR0$6bo4$ue64-x
zwG`8btNM7vo#cIY6zNY~Xu2WrEF-5@U3EEQ46~{@sOz=#bUY^&Ek=*+s5VAm$aolP
zQQSF|?85wSL95pErkraf;F!g^BWPf0Fh0_kDgyw!$Y(onT40J#WX}6J1yXGrBzTJ{
zx{3TTG&Ib8a|10Pdjq5Q{Eltxf3b$)Y9G})?($9lTru3NpWiBksV+R^f@y^^zP(TT
zA<(g4Ic!{!^mg9wt^(3yuz%ttihUcor>v>@h8fs#w~*nF>oZ||KyE6fMYtw9orC3m
zT!KEtL_IRiQ@ue{v}G$d^6;XKnV+y&YLhtxs^4PmTT<icterE4SC|c_3$<xK&{leK
z9BG29#24KjXN*b`&i23L2qk8S?Pmy>M?nXKj-WDETkglWDw=LrvvVUa#&S-h{@r~+
z3EE!0mk3~@hTK|aDLUwR&m;_R#!meh<3AVmRv8f?fvF7i{^w?@Y5ce<DF7laSkUxP
zA=Ba~d28odPRXKE^k!}x64=OKn56T4MCK|-&Bk)^iz`3G!ItMa{d8;Qu*e_R#%52y
zCuI}{C~8=zskIRLDxrf|k?0*WLE$RV=NqHeKAGc%Ie$8(6Nqh913$U;LoXnfUg-p#
z9`rm?|IO#8(5=^zrEW}bw_T3vTIL<nCa5VvKVH7)oQ%Fm(la@zYyXnO4SsSN#6q^E
z<+n%rievo=?af~kc;*MmMRg<qlUJ}1vbMwm=x7s`&nXuBOQB)r!a3_8RK8d^VAzh9
zY{mRgTn!`o8)hy#P<!jts$QPFIPvrT5kmDdF5vo(N;200OKtkB_@av2;7Z!hueB3L
zBNFOPLf^v<=&mfjeZM&|tdwOQuf7~*kvQ4d{7Zj_X@I^L%6vK#ch$!NUt|0*Oj1Zg
zKE)_9M;nv^reltOPXXeK`tZ>%GN)}m^;O(AaoQy>3BqDEN8HypS?sMlliKRN#M0N_
z*+X4A#TO4poFmN=4tl9fzhjE9FU8b9?f4S~!u&c1WNO?#ycr}GKp^W@B);46AoARN
zL+($@OQ%TA_;N?ebI4_HBpv|2#D95}jnFrkRiM<1xY3tSq0Na$*&<QN4hu+*8(8LF
z`-;c;wPn2db@`d%&O&`k#c0RsV1o0EWs1SZchVS_NYx92jT3E?4Z#sR!UV5ilFgZ#
z!o5(@58q#!_Ftj(oWAdc)Zp(gAZktZOn>#bKPmK+1r>{Ehq68n{$|ksgEmxp$)8jT
zZSXjBkb^zyX|&^wU37al;iG%Q+pcLNt)ZoJIVnFsYZ0mog`}|Tj(rAfexbnwXBmM~
zK8vTuf_;ZnZFI&T+00<Tg?U(VmM=(DnUC3(sKKr4EAd@bzF==UzdUM}ywa^4w4~GG
z3)NI!nYh@?A3cD%9H29G0EXT^o7MT!xPQ>MMAvk{k8xjDvP-eKh<{WGn7XpcpSx~7
znu}5YOfJ_1Ej66@0N{ea#d>+$vQ$7RdXYH^;Fjl^pOIjj-@+_k;GMZMAjsKkoby$z
z`$negLk!B?d0u^u1Yf8z1%LeHVpIUoL-DDNvSs^1M9gnbP^Jm=8MGv$yG;Tmk;?dc
zS&_9xFIw%}>1=85QlI_$dvCv2HPL;mAe%CHBS5qwb<0d-o%QME9^-RCwCC^_b86S3
zU&?P#ETg>d@p_}Y_k*y796q-#t-kjhS7Q5t`GzyP0~~%+)LK&LKo_UT`F_B!QjsN-
zcK_B5Uqv^~V|0)S69EqntGr)tywB^mh`)J3*Pi^*6v<u<GSD$h{^P0D8M&OLgiy#B
zCq{AxYef2El5=_!v=))p68UJ2ow<4u<<FwS@ijj#HnQW+7JWt;^GKdTlJ-11#J1N2
zLxFmGGQ}krSthl@D;`Ut(NB8GZ~ZU?qXcS4UU_h?{8u7G8FM-Ik<VT=CtAL;dl~i=
z1DMr{ILx180uXRf%!xO(3W<x<*58TU7R1Zf^!Ne1`bsWW{g7?tt~5ue4C7mkyGhsD
zpaRGHRHS{YtQdKD(kD>z*Yy!b@pBy)qaQbWGD(?2fo$fGpWAlrkF%Qd`db!HV%3uA
zFpZZ6kb`Y?Q4_1)3MF`mi8th09+1lkEf<8>C$p?IenkO_xSpT-&ZY2}#$pBxg}U@j
zt^ZM`qYl9zRAhF2qs-WqI{)tW`a*O))#7L06Q{3LRS4Hu%&hK^a`2(b<d2xt&F1>1
zlKUEY#MSesBXa1_+P3XLREp6wxi*_tk2EoHF&PT;Q|lx>@}WAY%GQ}K1$W<tqzCfK
zn`LFUVF`%_3p>oN(j4HU|Di2M+s@T<-y@_Bbl;u5pdMzn=#8Xk{MkiNWj|A$XW=ye
zp6gPVt>V=XXC$Q6&gli%gA(*d&k<xRTZXg3K9MLykatTrW4@E@1xA_M!npQjXqQJ0
zwI4o2xl_*-SdBM$W5$aoMyEWu+Hb!i(;v(vnpz8##9yx;vQgERhbhF+gk(6iJ<j$!
zeT*Tq4|Q~PEw59D|G}uPKQP$j1Y9s`AJ9l&%*_m;Gt3&&{2K^Me*_49Qdw}HRg3V-
zW~{Gq3fyZiJDd+yQR1sr7~WYPN;by}Kpwn7Bv11gcri@3Wmhy@U_e)w3Qz0mocN+m
zTFRz!awv6xdbI|s4nyyk)9s+0AdiKR1Mi53yjCNbd5(V30Rv)3nDM77{zc0oZvREz
zQ)pRnrMwaX$_+i&HhnGmCrVJz^wv4;6B7vQQ)7D^kAzHa2BS)zieZI;DICGu4>mL(
zt^N?iKjLr|{4*h0#U{Gaq1|$;W7d7T%CR)h>G!V04JNSl=bHQ^&Bw>YLCY>*%)lzA
zZA;5FjEth%{I1N0Qu(Nb?WuoDHFb7+aMB`E-D+``K53m_b!cd_`T9^=QALHv0#r~d
z@0Bfs=^-v<xoKL1p1!6I#1jqZ#}QR0GvTwKrsDaV4*9@N8Bh{l9nNGc>Q<ZZ{U7E@
z9lo{a1GTFwf;NmUmk{0SK?vN-$E!W-o85=E&236*!g$+*$KAxz`WsJz&Ru-IC?O~!
z<d4I%R0U~zUrX!;cYeITQj{Cpg|S;IDjck>UuwoL+{All4rfV-lzZ{IuMcMw5|}8<
z9h<*ZGooFx!~@)V8!m!?4ZO)k?2>u*JEwZ~sh0@l``Gs=&Hmgkax&yVO3fxe(&j#P
zQ5aWHi+l=M2tFh1YnjT<C^y7%VUcrO!7t%jzV-IBp?xc!M>^=6itz>M+cYvzet!Pu
zSk6cKPTPrs7{^pT_Z|Jj#KcWW_`u8H<dYqnaz9rx6W-`o$y^4X4}H6WRHZR62b%Gf
z9G`#+(9{0w#);N_+U$#!n=)#bZ};pX_4F#JMf51I=_BJf<S;Imc$W7h_kLbtx8k*4
z+F!C6_(4rV_3r1@L^(CXJwSS<)-RHZ&d>D+s-!x1vuK=MU$m`%;6ghOz?wzLPJHkG
zI;jZwQ9tK+M7MUx1pCaboh{M&#%2Ll_vLw`S@%7+?MbCyGll95gpc{}avRo;z0#ij
zK9(bgsJ2wO5dyZe<0s=gwVou?-+W`ebN?X*+r0QEG7>*AC1vHHpJ8;{x66EMJU>;^
z$6f0Hc<r-kM$PNM@4wi{YBkE{)%nI;J`Z}y{=C91+}HW|cLu`d^&}28ZPx|h-cPy!
z7nessUnQJ=8q4hm*3E|6WEu6C1n2WYYVktRZ}74W=Wbz(O0X{Hs7Mf_>^n@xQ;~YL
za6<pYs~Y^Ae3<bLAsJ&By>}5no;?yMIk3t0fTv+ox1=k|n4x%)(`f91)@wDsaSw)e
zoAU{xr7qUwJ2B<F3vE3$yy0SBNP;Ao979V+4Gz}0TXyV#nr@1_6#XY4XN07w%kFO^
zG!gym2Ze$|CpQ-B!F>)9=Wt&!8Uxx_BC;4K&V|N`Dxh;4wgXY`_TRmHtK%n<2wbph
zEkZ^>Bud|uM~=eb>kFeUCtl(Z6}*2U9<iBYx=!aNB9|Mz`Cf8{$VXgb*QmZH#USIf
z;|{TYasplkzhxwVu-J^RHQ?gmgUa7J?n+<G5&1W&6Z#N@0K-sm4H1U7n^HI}T;O3*
zaYDoR<^X>hv)LgvdLc{hqd8HyEe)%uVSj1zPyZ{hm_a}1aC+7yxEK>OUzsc05yHc}
zuG<*MV|}vv;#Hl=YfYZI9RAfK)z9V_c=7^RSHIwHq?#=S5_#XaQw_K8D&^3xe2eXW
zbAsx4r7SegDi`1K=(T94GhIGN3Zs4h76dikyg>%*F}u!tqszVVi>b&j#Jx*p#Q2t0
z+kPTR>kn{9k}Sg%=8zPeZ;tFGM*k%IWy~C^?W3f&%-@rLt*Z$yZG5T)Oqq98p?nqS
z=hOTl661N!_hh;GHu<YkfR59HQJ-c{i&VzHCPP@sXD)89_8o!-%SYf2s{ESkC$quw
zD((I^AGY#Sw(DMEgl~C_x3XMktTqe*M&U42VSI#X=YBfpZX%JdP$*#Sal*~bUi1DQ
z?928{5lhEotIT-Brm40JLXT|0L`Oc#`F+EW=^RU!qf}MW(9K955(qddDsD!J48+y)
zcM6J<AFw?561*ukL|Gj9D(7vCyFqZAkRi*f&UbP#sT{DxaHApq`}*}MaTKoN^~{%w
zZjz0=GD=GgJZ<MV9`nv&++AxM*u)47%-N*eg<KY3Tp&uG_G|tTu?<?k$N9AHA4$gq
zS}KaRoi9hh<dzMrvhoWb7T^GD20~l?u1>!gCFtjSS4!V(sprVgRX;e<ygJJMmFadb
zi_?}5sy13SB$!1mC!$2*U>Xqk@h2Hq*WY!M@5J-i<&}ib&tyR~_zc`GFnI}iJylaE
zUeY<g2ZA-KDwTWuJ%!)87XNfOD=A?0`zxpyga3I>ML6Z#xS%F(SDF>wxjzF8a|Zot
zA`8JISrQvV{n6Vc*Rr8cb+Un^oYa100|U$yMHd27FE7?GO&XO!xgkIv!qPCT0%+<v
z;@Z>(aWRGq=I|1;%&yg2dTlLY+`?pbuqhYp?=q<|)7}JC*L<%$36mf+&Pd_{Y80b`
z8mzU@u&;Y4mR=1>|EOjDovC`g@w3SJ!-2N>ir@1!weidCHcXY>VGr88*TwMb{FRQL
z-o~U=wEe<O-m#|lDn?u{o}23qU-ms)@X&Af_fziA?;E|jIwP(nN^{C|pSAJMuJH{q
zro|Mhfg~du_wnFyk#*WZ4WUr)7?=uTIXn{<<5qO`9~GDR{>k0ine@ecvU!mbq&r&Q
z0HSLf`Be^-!Z>TLv<=IYr=rE|uTluD#sp<^?auOrl;$M*x6_L`>}7*yiae-_Ul!La
z(8eY4(6bj4(=cjq6uW3*JXm{gsy>NAp*cb^CW*-Df`=&_hDjcX<y%lvQCa-DIYV8g
z;f8qE4P*Qk52Yc!w%bc|-ope3l~Uf~e!N7VM$ZD3K46M`Eogy(yqoC1IY;;~^og^K
zHbh2ZEZKbX6Fz%032X{n%n^DE%t$n62f~&)B83sDqR8**?$+9dTNfsh*FU>2)f}s5
z0q>sne)_sTP|<Q2xra>Q79f(=kcSs{@;`XRkCP_h0dQaRvT7ZyA)N9~L*B^bNWEuL
zh*tC(As6ylPqT8<M6=MkQf$v8bImH+t_HgMA?jLyo8Nj&YvF6Uky;U2#8C}<THpRY
zl}nZFp}$a;&p8dp2LH1Sds~iH#uL*uo+)1eouO0Bt<~dov}x_4;Jifz<r_xzo(Jah
zE^u!BD)ZjtBM*`{2X>F1_KwAEGO~XIQXame%SJE1J^%T``)CRK&b>`d-be++_2BTQ
zapB><{@2VkY91e<DjC8KT3-ax630&8oy*wd>qsX%WqPfr!75K6%w)Zmejaj%>Y1bH
z<W3kgub(oWQs@!!Vu<!%Ns2?hOf6Z6+y71>_LWSe@L=*@gq}OX(g2XpCL5_U7!N(C
zOtzR|GG9Lp<bR#s{383~oKv~ckdMq(XhGcz-*5mp?Q(vr?g{hVqPD1==_BltY{re=
z*3(&nn)k!wxEQ^rd%1wd<Qh|R99cik4_4&17J&7;_vzd|0eVKHfBF8-S9m!OKTr;n
zJekx6UwQHL!*7Kgj@kDso@fgQ%fKFn-nxIU+UV&+)PpUsTGq!t%+Qi2kNgd%+Gt_w
zt)+B^G58dFHC)R)QH$#RBOOZTWtCc+ZwHpnYm6g9*^!(4*$iX;b~;o|PXM2Cu%`@~
zrW{xvVGKwwM6-vJvERuQtj0`AMm9WJP*s4a{nOUHe74Uq?WXL0@zKIvwbO;QiX}=+
z2(|Z=@9usk*5IalU{ZlLq2TVk#mvrcT{9e6nxBqk=RVgiGMdd9CnjOSwY6OOo`2Gw
z38SL=!)P+Az<X})7<G}Lj2Q#|F}`ktUPwwNru6J=b~fKgj+H}@M-DkU(0r8?Z{&v%
zT@;HPAoN}L>-f}l+0ieO3Bp9lIqY+o^2>5@;^nDd^tLk&xH)OuZ@W5Xrx&sw0dF1w
zC5KwrqbM!pL4_SwpE1v&tT7q9E)_PiEq$L}=bQh<147M(7^nn%I8IC#kq7i(5E91g
z+EQ(r0t009wuD7Nm}#Bm3vVxni=O=U*PPno<?ZwB=qhB_A6RAkxf`;QT@1v(qA!-9
zx(vyDBtS=z$3BhW3V9+y(9%-)MX9*e2(9`GnHKAuPD3Yh>NfLPMvwR+cobUA`Pz9j
z)`vYd%=p(|^1Hb5R;Z$xB(@$#z*-fX5|UPEEr1&Za+9HS2Otk@Ylj(OgwH0c%4j4S
zc6Pix>a7usEf?z<k85H*axtSMRnP|W{wVEfu}_n%gM-6YcFL*Fcu`w?MZ5rv1ql-{
z2gt4SuFh`)QIUrBsIKM_J`_;T*WU7iJ#2s~3h;-L9@6+ESQ4!T*4LynQe)O>P&;T7
zhI!?{fc6xdeLq6=`Mi|tc^?O;0kyEy^0lw-x6!60;dKHofkgezaNS-l{esO7Y7@~+
zDGZJgi9Gm-k&Z|v3e+ZWS0wgZK4K<Nr?7B}L9=`LUgTy92LI<46ch}OgzwxILnAYJ
zIJT3cJ`MlEq-R#eYA{g)gb|U8L&c}$90kMi{`O6NVhWDfhdgwMicHYI1<Q=-iJ*a<
zGHh6HrtSM54xwg%1K|}7BH>J9k{?yXJ;M?@M?^$)%B59Sc1~v<p`rI$oy#A~96s4n
zzF%TSBjL_<kBEkR2_A$?;P$M}|HixZVf0=6CY;aQ5|3p*buh{(#7&jzp<4&u6}>42
zHM{?22#&91f<g$N58SrumcDshoT5yj+;H8^a2xpB?%kpvm}Ghq6Tc<Ns3*f;C%O_X
z!0Y2~+eeYMiix|7Z8Q!G&o$CpYN$yZm)geUXCQMB=kx24QD$Ea@H7p<lc{vVZNJu@
z07!X-7yqQ)Soc4;NYJuG(MLTpdt}hx90@2|9-VIp4JVm~LhsC`==!wX*O8FZ%&RP4
zY5kfgYzw*0{&jsn?!{p^pti_AjPZjl9dXlD{0GMPO<YXAU<H#=>OohuLz^F9$h+=-
zKluAU_g$~?%>K&)Yzt{OX~E@BOqdC`*dmm824G)Q`Kqx`4Cl<k=zgvys4rLb(|6~_
zNtKu`8UHhGxfOLM5&~;_sSwhUDlExsG1GNT4kAM!>m;x|C!Ss6{+sk#WuiQwn1gt_
zR{^VxG4J@~6S$@)ojWhVy{MU<CtNsC2OAcJDG!xP$`=^z<b3JZt(DjtU#DjZ&F|PF
zWi@$VW(7Ci`Gn(_@4Da*B3Cs>`Mp1VcuIPZ(tgciWlXU(KX~%9w4&{hwgC=kepmOH
ze>lN*1Y;Gx0CjwR>;JR!{sBf7uylCMk1}>8!c+u<o|#(qEGDV^+1Eb<DZc1a+YdV&
zr!$6_{0j~TRk)T3Xub))P1TDu`{Zx>=Gl5C1TW0by1GS~5jNyURDlluQAhW=?S4h;
zDGjpBz(R7DFaq$X1Z89Uek3#EBe4|qZyA1ZE>erM<^ga`QO7HSC-OzS{!mCb-Easb
z{Fs{m9p3FhTmW-<lP!J`u<J96ig5*L@n_0jlbGU!NpQXlH39%nUEcgYzI&DIbiIid
z(mLi0h@^YW+B~*TkWXs=%Cdje^6pnUtmj4LLbn&j58g`N2;jbb5V#-HZ>&t7qk4nm
zPuz?}ZEc(pyBYypTA@LO?ZuQ}+4Dn(#aAwP`oNHWbF3wtugBprVjrYJ?zOVr#EgER
zS`oVG)^ltTnF<Z*FvKd29zAfulGd=ij|H1Kr|obWzL6D{Q!c{4(?OzK1ToTB@o4^b
zep7H2KGkOx=HEbto$7SGdo^^PHB!`^MRP%;u*Q>~{DNrg&^X4?vJwh^4kKf#uRw>x
z$o5UN5?vH+j9AksK}}dw2AUYuJZ8z66=#j%we(eKQhwsXj&8Rew#~cs-5Bos#MWYs
zS9GVcyAq>eY=iCg+kg4L*^u4JCs)cBd@TagoBR=%b#FPoH(QCx1*7ph+sh~FFbH}7
z?2BvjM|HY=%r&)P4Bk}o7kIX_t&PQ5?ae<j#-l{NB#tZDZf~LoYnrCQ%73BJ9O_2%
zcsyqCK2`y?Cr(F;;As-U*cJantP{S+mUv^l*1V3#8`LPY5Z2P_Mcuc13o&O_9#|4I
z=Awj(;jQ-rx5G@ae_A}seBxJsudvafWUhE@=_^QXHs!$rmIIzwwwVkf<zq~H025LC
zm*>YG9;|^V>=!n8ml~ROZJ|QCAgUMIm^Yg^rx%!HbH~?hcO|Et+l$eFZIN3HWBsPf
zP8ToFCd4g;E~{}9h_L5#UX~xqmS*@(NgTN#F0(d;Ld9@1lG$3Dk0OhncUrwq#<fXm
zPSTzo07)AyV#WHvss^=qSM|Fw8v(cXN@;`*kd`WLxZ;V5(J}SKbhiJGrn8KyqHUw_
z0i?U6K}w`MC8QhV0Mdt+?rxNl?v`$(yFt3UySuyj#&@mnAB#W4%z5U%_rCU?=dJ@I
z-Z5wf0R?#5?+8x|s2y4h{Fr!U@6LlTZn3YQtL$_R0>fV$BH`EDb6UCMd&5}-$}&;h
zLFmv8W4}1IVT^P=sC;t{J};t-B%bEL;Ya3#8Ha`w)#HI$jKqxbKa+&t?IKVjB_lQw
zMpB0RFblprFTA)H5bYzj^fZf>N8=*ZpB%xcZiHTe31fZTJuT&HD1ZvyYYpoY0*&K!
zpo3hgO4f{P2QF>917h?^m<xzyIJLh0aSi%709CrRg~>6(_kP@xLSFgZDzMXMO6Iu~
z6Y?U)u!`fqG+m<dNrrgqcwY2002WF$-&3k}lXQ6j1-uxx{aG{cOm)Z?90>=c5dr%}
z6@mWWneM_>Jq3$`%<if)IWF-tTtYww%UcpzpM;)nu<oYt$slQW)Va6*A#U=b03#p~
zJFe5Anw)_Noz*@LaX<f60)z~;oFcrj11C=M5#%~3<>AqfG4<?XfsB*v+ADR}#OK#3
z8&D>+vbSm=)<MeV9hCHJD71x~r50wYrrtSpz%I(|<wBG(chz<O`45){Ddt%{Zlr<6
zlh=S(i^2DhtON|5%iH3wlbIU5i%(8PO6(dP3*x^Y*1Q(4B~+DyPqsJr^2{FP*%#__
z2!fg7fdzg&MwV&9X0>`CFA-8)APlkn3)=uZzRs_w6t{-i$Ha{D9DItke^Tg8mxogX
zFZc@9t8h!s4Tqu`&s=!xMuf*GQ~f$dVrU|U3ysMp4_P0aCQ}a_pjpIMDvX2*v)Xi1
zE*%k<AlyXc{rHIvMPI5+(}yeDWg&m^*RNKy=?^3(f9h=nj}62eOh7y=Lq21n;deWR
zTX5c{4j$~6ae@*^2M*-<rjL2?7sh|*4kU25$e#b&tgL>U8kc@A&#HwpM3`44-ps*^
z%Ss-1+&RK%@KLf?Wg?F@+hQoGN{SV3y5@p^U}q@ts8(@A)o}ozqnVrb$)^G>MZSMf
z)3$?+5PURkRn!m^{*I@}e<7CHJE8xj*7o{Cv$xFJwf#tLAW#^gsYw)3fXR#nWu%6m
zO|L*T+2-EaFgsYr``E)_g2rVY7BPCsnMoAgjG}}o5w6vAtgnrS6`04%LPk?4y@j*2
zdlEUaZ=<e8Imx+FSU<$sOL*`L9Y3E#?o_y)J~nJ|aHX36bJe!D48lb&21Q=eV#EC;
z2a0oqN;0RC^8_q!N}hByv$)h>+Ly1fIeN3k71K@{x_w9ED5&BGfe<C;&3{AyfWsjx
zxaKWNCCgC)ycMwxM;fE)>+};nSy6HEWR5cckW3{3ph9W=>LhFuo>u%^rn*=vf6w@x
zdWN5D{n_UZQCA>rmVzuUfwZD*Mrh_gQWnVDpRU0tsFT=EB!QP!%@LG&4LYw-8aq5q
zk9!*z*wqO1*!{ul&2?=vz#7ORC<wpXMU}aK;ZtZ&-Pfkh>;HO-NFMlZ<tx&I`O*TV
zs{LY?z$C6B>$S#Q0M1bp2mC}HK`(U9Ve?0Ufia;sezS17wl8KjMqUcx&I1y!%MN2b
zfF4=?8TK)N;v-<|Fbe|YJ#+)bf**zR9K?fhaA=Bk=AuK`G1(K4vzmV09wKrVihs5o
zZ%NhkAZ{4|4$fO*D-HH2N{AeLK_`oKf4X4|_l?BiH|4dOnC7ie5L5qDJ2E9`7}^#B
z^{R7ZlpC}XwV~NvVWvNQ$FJyZ?Z$B?60qv$BEqIP_znlwLlU7<H_EH;D#Ebi3(b<t
z!VCr})32xiFu-GT*@y<&IrSLhL*`fyq`|F#J@>8cOKX#L>&>^7+ZKSqzR)PwZ6^in
z>lvGVo0bcS7T{)-9@_`Vk=dg_sqTSYyf7ar{BSZM2wulEsedq8vscw`q4N%yn{*=u
z$I?Pkh5%EWI!c|xbaC+ou1S%)RFf1U$F0XICFa;6vGCMc!meYAsT!^_**2n&<3b&C
z_y?|%yQWKlnaAI2X!=bqs>=2EZGTOO_?o#NhS%(BN1{z}pN}Dse>6?G{gr`=WfnAE
zS2<&`5XQ1>Q4e&y{ERW!SDB-p^AW}}B^fjyZ?0N(2D2F<D{vP*C;;o%#keT?#(5Wd
zJp%KK4l}c4fkIjZaCRIX_!|{Z04rOr$ur;@AIt?wlgOk#&$YkeF5)`MusZrpnzi~n
zh}R%UIuUNa%@Y(KiDXLQS$ez|pIFsC>)}wA8s|@g@01OsAfYk7$FJ>`*C6Mei7vfa
zg4LsQ%#McEL{Q360B+7`UY5jNi6Z^tXlAp<YsPyRZ%VyKX16zP8~Ry}(#ZP%a+ZIU
zQDTG3T#w~o+`r9cd1=({A%X`t03i|F{+kzsSZ<Q;1_x-evV6&9I?adHKe?A}VKnQm
znUFK37T#X(+P~X~#Gp~d6+T?5s_Dd^M*QSVd=I_mc@=IlqAi_^;8AzDYhpqW?wn^O
zql<F_SuXXelm%*qSn{J9M@{)tZREI0GTDWr$)%7hznLc}?L?P<6>9h#&Wwr_fR)oe
z#S81G@X%2zQGMe1Q}9j^fjp+JZoDTEt~3!2r>}Gi82<hm^lw;^kSU8LI_VAodRYil
zl3CQ3_Cv03+}!Yj&*!2m1_ESRsTWE)5(j4fFmNacj&R!_dF;3QbGfg9@n1?f*XTPo
z)|a<CpSK|WuE0DFWP!yUA6rlURvX9}N%(;Z#LpV{RN&km$gWw8L4PU}beD56W=XYX
z2^d`s)DRZLVojO99qm|V9)zl|maAAWMM;lD$bnxDfW3)O&tevtd`^hJ7=V&^$4Qoj
z)BT1(&QIAMcrQZ6eRTf?*tSO!PqK%sM>kadrE*i~hYcm(jsIb5+?WEGf6?+jGrGlk
zn@B|LkNeT(Dw*z?RaI<bj726hz2QVxmyg?6*ANnq<H}ErH#XhaR6Q7I6juKnK>&x~
zuRl<i*c@6#=Cqi-fVyQUXIw0nM~`r29SN(xdUhN@*;*rSjy+ND;J&|uzGs5pjYgV-
z=3#%!CrN4t|C!`Zbr57@9ot)*kP>!mRa~HG<0vi2Xkt3U5q)#gsrbd@18`}c!<Z^(
zwW@B&W7RYhz|SeEDL1IY=XYh1h$KzBoS-FCR#x8P4Ln{QD9Pdd^0NA^OuOiA04Y_9
z>Sq_)=V3K}HtLBTmh@2*U;;~Dnth)L*#08E@K-FBgrcYgoQIl9s;lxEBelA6S;`2g
zJ@5q4#Vnae7h}*?HEjzOLt~P!i^JecCe}xJ*5n+FSDm#X)56nV*0K3e0-NBzfs=6@
zodg_VheofGm7eia`=t4AkTNl#?x%n2)8R~6&EVB=5jxgN<}Uq`bu%NM<9^iT@qqxP
zg}Zslu|(*5#tLBdYiX4j_@LJJkBndQx+8hdq+N(S0-Ue%k+)7qv#6p$XfxtW!TCV?
zDa)`gqU`j?Kqg_cupeAwvjv<q^EGSwV-2HLo!nc@-;q;6ig$d8wc|(aTz$_4iQh#X
z-AZw(dscX6qVzAZs+4?&_iy6RrsH^Qh0^$c9OhnD@_2c&%lA+!CkIHJ(FWRW@sW2s
zq~yO8m(88A8IHU2Eq0}rr}Xo!{tVs!X*c&TSZ3m2j74fx7d>`;y7D;*(GObOH74(`
zQtM=s)6wN4-@w;3dbaF_t%%R(0MvMp0b}3nSp4(^JolZZqTinPgqo!C+H^jjHc&dR
zdAR^_F46kif?wY-{t5Lf;xQDKY{F}j?OO-PfjyykN1YSl05O|wFr@ZJI+DmV0t&9+
zQ`6>=Y{*Wa_iyihmIs&>ljpAck31x!BOL^NVGxyR$$e~)5Ok){YPdvufjMPBVbBG>
z5q~E~+<a64u)U()?R4;0L9O8PzteL?(4}JuK*n9xMOk;eL8<rbDbM;7n;e+>ft>_}
zJd|eSW4Yx73(qg6!cN)hA;y`|at=*@h&014ZHg%7FV9{}l@ljQ1?~qTF3S!e1ya1`
zKQ%9(%B4RLLwyfe^(-!BcWVK~#dL~&EuNb}WZvUXnJW%s0!^9M%0m(pIsXI=bbD`4
zz-GDLp=$dmpW+W#fvjK2IIwJuT7)M!jlMc2kDL77JrzHSL9z^SBQk}ZQPQVodMu-+
z4{?m)vSVA;8&U`y(CM$VQv=!L@pPq$ICE>;h%xuavzoTYBT+(2?y>h4ipolctt596
zSE0_Z^TTvB7q(OsaZu=9Yp73!mBVa}NQ?aPF-@uaLaI7_bfYDPlHzT+6>osA;y6&g
zq3pO&XU9usUUqCm@!<+^ySIS#edK7Y8$egD`6%k+*h0>~>N1BcN1bQdU4Mv0Witd=
zP1Q-muRaxz?l$D9M^XJ0TfyvO{K|M~`d>tj!`MeV09Grw{q?3iiD$y+WzTIdEoctn
zBcO*vTMI(VBL>%u$>*lZC$y$8?wQvNxob}d$Gw-V2~U;BHKRv11gO1;uSUi>7oRtE
z-u>2iO$of`qw7xgf?Y=k?dCPS@KJ(qBQ_MIxG4>xi1m3?;2yXftkNY~`{7!eBklAr
z{!HvAtyZ6*R#s2m%hj`xa%2d`;Tww)r~HHj4sBh?yE5Xxji0IcUF`gFNdxbGVgr{e
zMn`_lg|g!83jZ_*u&1V~;9xZAM-Qb^MRo4Yg-syX;v~S9wq#b276@#4lx<<w-F|$Z
zZ%>x+YHL?TxaJzye<IsX$#WYUC8#jF03oF<z{TlXd%fi*c2c?Y@ZIPF0$;aFTY#7&
z<L#U2q*pf*K>Ff{ZKHBEQ!uEZz2~<d5{Z|c`*^;hRqAeg$o$*6pQUV=b8>%`JADj3
zGq87wkk)ZJ(>0eIpDuMhCW5PEuU%e$sgiu;M}z=UhYk5jWajT%nlN|`7BcHxf}5+#
zlXevSXrQ{m8zRrLHWvP>`Ic9g3ssjB*YU&X-jYL7VT5T>$f)=n$5$<vQHpp}+RjI{
z_spc}S)o3JJmV-o>JIN?b$GiM)nq$|`Q>&BVUyavp7zp7c<+s2$|rY^IE+ng`Ke8x
zP-#!A_)wr0E9$(M>9WjNvqJ9&V0@c=s=E0%z5tQQ`WaFIV6c3@5l?4yxba?N$f<vN
zGeOIGAnBZqTv)bN-z3fHJ)kiQtbo0rPF6hjqxJiMFA`Ne^OVO-tdD4nRRrW9!A!w1
z-YbQ-n|SaOJxa{^yjSeL&0&N%Lf>}U@gwh0?!G+2uYcHDTF^tK_H*z=s?}NvlWa1T
zc*iKWkynM%aRNGJ5gU9U4L6v>ueE)67ks_M#=lk+7~Ho+Ypu9|nL!vSi@M>R72T9V
zd&<&X=T`m3O7N|~l|jbjrjGTtk8-3V-tLFp3E3?<2d5#5z;}GJzL`L?VO|Lwc4ZX;
z-9b6lJF*FmQj*`6J3zyl#rubCYoHAR8Yp<_Pg733eE&&dQajmSwn{U}(>GhortKl$
zI3G}LZB2g~cK^csAx~`|01S6%0P5!!I}&KV{|VZ;6Om?p(W1QPH?IS68pXi2|8Mja
z1anRs;e`aSTx_Fi9aB;FZS9F)Krr;pk%->Wy^#kNnk>Z(t`%`0P@iB{#sFu__7ZD4
zuUoK4Y2Ip=BLRohN4S<`5@HCQzw`H(fj`vB^plh%0&G*B+iJtc+8p(wi)diE)5pxz
z=~05kI~NHm18u@iS#fp>@MjpF>HD;ScRDG_E8cYvi&9r7`qxRaoSl*s8`d(=?AbE|
z7Ea*3cWy(!;A(<3#MH9uWD_LJWwI0Ze^NAUd+R7nlx6B_xG>p7Uq@K1wz$Dg@FyL*
zF*WT|ym<cVBlVCgr_D~ub<A=vcPu%sMx1d~l216Ko5zXT&sOQ~>WiEcpRZr@wx~>?
zvCHM1QO~1nB2HoGGOZ~IaloxgR(wzUEuUcgYnPU5UK`C3mXgM72rORLJ|~JAA+4uN
zU=os;Pt7TJwk|uHWAaiy#*cmM$;u2=Ga+HO35+N;LL~zH8yRjBGr^)sK|km%>n(>n
zO}M?cILtJ63ehK_xdsWj-AtR5F*ZT^NZEG911xrmrfZLsqTP-ze~f%P{9xu%5{7Kq
z>R={$hxa-uq5G3669NAFlMHY2;iU1WO0d}WuMWqxaX~FWYS#U&u+D1fFuH=w=b7~+
zHgk6{M<RT%xe(8b6!0wF0E&_P52qKEf7{D0(Mpt`yZ?&U-AckuT!K_YaIc=^2F|v+
zb#{0ey!dvDsoJ-Il+A6k5sI%h^o{#SbS*JVwDcSOcvH*CCe!rVMq(vwP<gDUQ0o&Y
zCqjigo-|~J&ltJZpDeiu&lWyyJ+7KghwNaR5Gh3JS1m}13rw&C4siSl=cuk0X|dxj
zVgb7(51(g|M_9A9`3}BA4%e(mWUG0z+l{4=7d>J*YTB;c4KbR%5SHF`loCkmx9wtm
zj4H@A$lriXwj_4n&oVplJ3bkhO(*131nA;_7#YCM7f4P?MWGP$jI+Nm;TA@m{e>r+
zPNjy|-|mlgU#kJoX2dn0x7Jto_`>8^gluo~f^>Jfe8r3i05->1vaAk=$j&u6S-hwh
zZF0*<e7nclaXLDX@X%P>f32rWJZ$2J)lr%rNr;8TpCWecDJ1GQG7ppXi>8qB;zC1o
zTl0WzkwzWLs-(3(jTJ@<77g!L6I3$EZ3rBI+q!zzeqYulABf=nA%H`j`BUw&ea>wp
zVkA7uk5!yNvLlr0ghZ%g(#y6xwxhNX4%4O>py1JSK6@zFs;J(ON?fe-7aA@d&Itv1
z6|hnlcm~F~OTyt$HS|xW%YBprMd@W0YChebu)1HHt}qy#umAnwZtQlk8OnV4{&^sC
zD;jwd*urzh3YwZU@d8xHYHc?}JONR?0kErY5X=~)(HZV>L;xw}9mH)1x0N)K!okVO
z>9O{<+Gj`a$Mm}!w;X2TPi%+1<BPXRm5xwcpC}lHfsV`DI8Ss6LROp3-`Nn$roT9B
zk%VXt$&bBr!4g|Z0#hvLwm-pvn(!=b2(q3|?7zpW<OLU5-e`>^hm*(;by4gxgx!Yk
zz1%6fna}CMq*k5eT?k>vMkI2pXy3fnjHVI1aqMLSbVg`Yu2hc-%7z$wFo4d9gK4Km
zB`S(ShRe3-f<iU&Br78%nV_6i)gmr62oMMR7x-wM{Y9aK^OBw48@DUXX-?JIZxsW+
z=y!f;dpbpCG?%0e!m+#hneG0Mw`FR*2CB#Sa6|)9gv%kVUo|v>0akLK_D4a6H{bWM
z^ze(VK92$jT*;JXqj@}&Oc5`8a9id(X+@4H%z>L2!7IpgqKck_T$wB~cMl5!1bAvm
z;zdg#C%!^Zl>49Oeu;4AowI3};M4`_d?67V4sbfA6j@Ys*v^d_d#GVRW{#x^cN}El
z{qd%Q6z);|p+)YKJ1_aRVLgt>ejq32mp`?$PQsWvwqeeVvrf{0)fd7PTk4r)Mb(UH
zHTqR71iTG!32SlX&;-6w1OB=Sf)7T-4#TWJlwnmrb&5uFF@Ne5NBgA`8bF0fZ_f16
zPZ_z7zNkQLnwG$uK1uLTfgk?OJLNJ>{|hFYFCk(NS0i`r?>}ox$)@`^kPdG?M`G~~
zy|>paeKkrF*<jC6Q{qBSY9IYVxlRt%TXx0Ti~dyb)utCjDN?uLJzfDnhHK3>Ow^fO
zAn5_sE5sh&+r2Ps>ZVl}gdS{F*<(%$UetF6vI!!nh8~YBOZ|7dgULS+qyEJWrDMT1
zFJC&tZ!xEbw~GqWCv^y?HUP)#I?P>FRdoZ<mS%J?{A}m-gMhTC82VV|w{wN9gU7G_
zEb&$XaGxXcjH%>S%!#>s2(Q=Dma%kbkA)~Sc^JPZK?h92q9H4J=E4w&<=&kn#|uOC
z2AcJ~)=a3RTqwHtY<zo<pRn54I}o@aqE0-TQC_2N<#`Z?ET2}6!h2cnb0KCHv@=k$
zr8<FI=j4g8nCLd=@q<z~;`(?&x!jI#-4_PD3O_TUl<`O90nk=14hk~zTLglci<un^
z0qbK5mHDT^Y=?OEM2)x_1~l1_F-b5JAH0jC*uD~ThNO&Psh1*9lhag&{Yf>ja*15I
z|Eu4;js5m>A^G_>#vNX6M5j-Rxcy+bBX?E&t@EjY9Jt-A!Xe+zFwLR_rIFk&X|9CV
zKAC*l;fFZrm1L8@sk0)5S+<~zK9)9+j#XatA75@al*l}GE_aC@!gtbiR7N|Ndf3o^
zOhF<D7<O^tUkekCLz&xbb##@Keyq4Ct|Wa=)NvM0oh^MqMa$7u3!x2E0LnmI^GU#>
zm+Ih)c<CcoVBwB9^0y<Qz<e7S#Q*sAz|uFmo{sUByhhhC&(f1PPy{N}*o9RIivIC}
zha@1vLF`H?I~wTiHk)LKZ)wbNY7C}SZmk3FoMPPNanI%TtI3zmIJqfp<!CV62*~X1
z^5`@<9%O$$ioOs9NIpmXb+}}R7>tDW^OvWise;H=lq5MiG@<A7IFt`^{VZFtih|M}
zkoM<GS=&3Ioa72ML$hLIB31*j!K>^VQK2*8{6{BVlv{y>lGoPBFSI+hPJUX`{C$T_
z+fJQfXh;si@l{1{FsvxCOvtEr<t!g8zO}!7@QNc(y-$!}T*B&luoj|Z>!q4!?6$Cz
zC)J+&Bf30CN`4wTz;qUv^jhL`q4Li3QC^m-bjxo%YuGw0Q{;AI!GD;#a1NUn%K3O6
zGmXz_(k9}7^TwaT<qOs?qqr}#dn!4%Biw8IH*zX8Mon;&4Vks7RJ;9~&j1MC-C(zM
zy{f2!9M8mIgZ1Tr4ajrrAh;ufX@*y=tjC=`PNdQ~rnRQv7Enhhvz#KZ)wQ%}mp|B(
zh(^oyjVaxhYza&um%c^!NfICL@ZLn}r=T5jaxwVlh@(R7(%>5S*<HmJr7p1LT%5XY
zsoM|;co$9_x)AVvZ)4%lbsL2`cPIVLF5V`HT2$MKK-n5U;Pc7gyjUqHfyrI*gL##<
z(P{sL9L%Vb&O$bWX54i@(8n@n?q?i=5xtd;w6Z=mQk7WC8%&G^MfIVGYz7eTu79C(
z!sf>bV0|$<QZVpOkFChwC*-`&+aLR;WtB&y5S8YcG|u{|l~G5^77iD}Lh&84`K!BF
zSL(d&@$gTRDS4vkVWiV>Wgj2%3I~Obl(1AD8<)l+d8KDH!)Rye5mJ*=43M0i*Er15
z0=m=nzW-8VmAdNdOZgP@Ma1&=MMKZ8TX}ZtLFqW-yVR9@8k8(~{y53KXIrUF3sZzA
zJ^LVZ5iH%ua1Pq!K~D^wk3AIGsqkVt)+z&0f9a9Naaw@zR8xPZBMX6$nGa55cAU67
z3eoPfOQaYACuhC`HKxRkIu#W_)<;!V=cK-<JpD|L+Rt?Zh(YusNj!0cv=5Y#M$ln(
ztvd-?j&`pTl4J)IVWR_#*>iE<2@k@khdghp^TAlb6ciNcwYnPN&TgShl!PDYh`+I-
zOgZMBCK=KE@;KNEq;|cWPP9ERxPDC5JHGi$`FKa@%Z*AIBNAZEubxf!e-=Rf(cjTp
zad(0<yYZFM|K*ykO|KuePw=Bx=i^d=q%OrmO@tz>%;zxI<r45BDd*J4M@PSBz(X>z
zgu~g%o)(d*CgL!SVwz;5ZILUiw4W=}QC>~Yc5FUL&_v?R!*xDgUT_i+hmn?R6#i)e
zz;JzWw8|_%e(}XYBzs2MHE+uxsR+0nl~m>eMB*?2%SjX05H^tF<k?b&{8bs{#5wQ>
zVI1&rBk)F>codYP@+id&9^?{0{d4_BD6d{XB=5fRW;Y1rsRQ-xuhTa06PX#<+?VoX
zr&$>Ekzib=e)uAgn^P{-MW92Aw*vhnH!VuJQeW$O@Mg>f9=2mO<h_R90YaGkP)!a1
z^zrOx`hi#5jE553yT`%-4*1Y4m2P_Z0Z7kw{A+i~F>Ev-A#-F9|MEiY;TQ8Q*=!<!
z$^tu^o~itJf+?vpwo3;kmhu^6lH=!Yr7oHMSh-`CN|a96rvd03a(cvSE?$m4N2%6d
zi!w<?#qvF$$w_EREc%TyGQt<jzy9j>U^4p4C-`ZmC-2_6)Fnf*DjD4zjLEgVE4fP;
zFl!T*O%I&68Zt{@ok;*KyD8Ku2DFl1<!3KoPIHenCXkuP%^!bzt(Qy<B(9D?=Swh$
z8X6qbp5r!UeQx=?s7c+5(`8$-MS{W^XBD6}LSn%hIDocrxFaG@syH-p_xqLf1Z6!V
zFc@JW;mkN6e}|StB*poWYd~I6o%Xx1CLK$oykU~SA;oxI+>h<5eKy)qoQ?YMRRVM>
z*Otk#ln`K0Aa&Rs9(d&M+y&BOvpG`2ANqQf6?kEN9)@%$Q$~Pj<%TA_iHKQ%%ZgKC
z`gb0WTT5-Xo%n|B7+LMUy=*7XaRlOEuksY;N3Sxv)fj;o<=G2qFv^T+;&@HPWEy^;
zG)a9Mt~``isIyN*yUR7|1{q3yH!~D5-ncfES|PF->DATwayz=2(1LuILFY#fJEdIV
zCB64;yY4Z>$#9mlEjiv)Z9jR|jFF0I-T%>!f1}Ej#OI?vJs<ulza}gJ>Jtvpg1v2S
zX*)$wQx6Dr0qjh=FBX|hJO_}`wv*ZR;n|Kd_ZI+X&WgxtG{c`enE&uWd_t0K(>A+k
z5}DnM-A2SZ9P+ivx_PoVtO-BglmvYV9CR(B92LyC#t+=eI#pI|c35sKQ1k3dK-G(p
zIx8&53&yo_BRBDd2*<~N=8B<&V<w-9;X*Z)tYGNA8!{I|;mF;>s!JCu<*6#x8i2B5
zTQJ$pjs10H4xpnWM9~E}N;QjE8m}j1{|)q;S7SnRs5DJ3XJtfRf$?MNPU+L2{e(DP
zmdn|yE5OrU0NK{VZ}&zJVTIV~sAQhYeoDLR2}whs#sDpE4u59+=4@@PX<T@DoFC$|
zl9wk%s@?~mhA#d-7V&wF`vIBI7zin9s9U+^L!kAaY+N5^QMz*F>Uz39QdV6TUa*s8
zUROV%I7CdL=0s@IloXCs@w-JD%?8zwSuc+8fOB^y>Idk#Y&!DaOwRb%mkZ07yd@`3
zo9<-T{lBv$VVUo*T%%=u#E8KVkwW0C8Ta^r2o=Mer3M3M*pa`u{*wsKa0_Cmn|Sg^
zkbUi}&CO<a#Pqw0b8h_Xoo@?W4sA?Jdi!YWp80G09qUb*w*Lkh0&UWH;z-)30G3zM
zoLRee{|f%gNm=HAe+@ep2I&mHBb|V_F)IzyUi5}LMo0JhKA=UW!CLA`NqiL<pLd$m
zA@F@+v4MDF%>2!i^=9f+<C^KMZr*u_!~3p9QLyi7H=+Sax+rk-k&JNq7%kO^{FWt@
z*eva$^3`Sr;t83TO~UMXnjcgZ?s=>~JI?6DnSK&E7q?f<joDS1rZ&V-diLq#;^uDt
zxQe(@u?~aIj{%q@YQNJQO@YA1utpQ$D9Tm|2yJwHm+N&f2T-JQz<~P;aMOJ)bHNd_
z0CqYr{=P}Y3JRH-y!#a(%#9)l;pDfos~A`<kx}^4Q-LD>-nhGVpC1YMbjh-~>IMk6
z;<)3gr2-!EQdD3!>75<V`^X0eP=yI>t57@0w+pn60#g49_q3ItNrzguyK%2lDr~P>
z7$ZoWv^bkH_<06Z8EqoqLgsTf_5miPL<NC%9vw`#l>vg72eLjw?Ax*oTcr=OET*5}
zsR6<YuOk-bd%Ih;I<HTzLYJP_Q4<#rf_Om;CR2y%hPzFda;(>=Q~g*ZHXdbKNrA!I
zlNYp`u051>5^oWleLut~aEr}A70jUJpKI)@<9`~EBRPaBP;PtVg#(aXGiu_?e;&>t
z6?;Zc|Mfqg`AQo!%?L&npj@T?w(tJXNbVXFvSUkOnGGyUe_xB;u6E;%JRqI1>crsX
zKZ#Fgae(Dfd-S^V2WU7V`j@=E0}p)2-n!DLlcEi56m$XlZ-8vZGk*Q3^ko23|7D>;
zj2e@5;wG{VQiDh$x%{78Y<jF<6+ZpOp&ZY`6H4Ly7+ei6SGrAW;|3ey9&`9PkFR{N
zPCupoU2yTgYHx>S<Fv<VNkM?sVBH@@F1l_a^h6*l#*o0i_3^l}v#LvbXgh=&W$rvA
zxX;UxUium<0c<daCxVecA3^fDZoAK~{6HK}$H;?O7MDa9ZJfGhglnO%@>f-thzyFr
zN*R|}U$aP}dO_)t?4qnZ%{@i5ly|3aW<7KV8pu{Ffa>H1#`g|}e*;}c_=hMp8sV$-
zmm}J}bV2av;_UT5bT~g(Raaj^=bwIJZb3SyTLFl(y59jM>{FiY?lj;&j81}<WcRU7
zfOfdD(>ZuDPY~cKSt+njR)k???Eyzi;6#pOHE9WbdCR0vlwI_7Co*$KvPt@Y2u%ng
z&tzvF;&i{^XM`DV;~GRp)Y1bLTk60Nj}U#Pao9=2Td)}N_VR3xHj66uTa?m(zFYA3
zg{VRKq>4KX(D6G^0jPt>_%sl8XH<umworfZbK-*CPIl&PavKX(Bk=3fm-u_O7xZD<
z3Ke!j$oZwkT~PK^<24a@*uM{ddT=%-0a8XNA}iN}^yM9ZC_YHB-|Bz5*&j`r)z-H9
z-V&^(H57xM$f%Vt_o|0Z_)%rk`QpXxqUU+nghFbIzZT#RQ$}!x)WlDY%#l%cD;fBj
zAR;mktcz56{%~In_9y0G(TM3Q%7-c;J=&Uc0}ph{sPl`uAn=0B&JSWd!#ybYN*_>%
zi}C!KHz&x|$3xq_6wx$Ph%v`-P!%4&COs1(3WirkaP)Ej?239@PAbl~)dT=EOhQF<
zMw+@XQbREjj@Q(xv;2|FZc=xIVd}3y91-@SYuyBhKKdJ-0qJ9XeSlzf`lt@fVjF-{
zHsYomGeG4~hQ_lFYPf5AS_l4yp8yXP{Fmf*a3=?d49EO9VfMZV{oScTSKIwnb6)#B
zHh-&s`PQJiK<>g3foHJ(t^;HJ&}))baMq@dfoT^Lnpe^F`=FydSv>G7lDXct^YvmZ
z-4{eO10QJL7pP2bP#DPf38d#qaJ{!~)7YLkF?4VE<;d=>uD-=d{fkF^Hgsf~3N*Q=
zRbd<+;hyJ+#p-_ZFs3B&U(jYn7n-3%KCshup0)#^mIe!yIp7KpH=3af0FSA20IaX4
z_eCY)i|aY@KqQ|*&90!r+`#mcU-oBjy#>Vl-*;O;Akj#`6v0sma{9~HeTcEXLjBn@
zV5v7o%JbCc^@{A9!GTJp42mqA4w?yytkuh|CwY8H$Rm+N&*2et)Czq)mgt=mEtd0w
z4DnWxe1xb2DIKj4tZE=<9K;0-i!_uZv{Bs|v4!X&3F=t!M6j<SK|>@-+#1@>(<T`|
z`z4IK>wQ8pjweI=lQ|}lrPV+=uZ8cAL3jnfCi>G2@o^(c`t-|XV!J;wQ-g2MC?yNf
zoAx`tAQ)TlUHR={YiGk9z^tLKx^AI<%b$aYAcz&1UG2(kVe6&>$n2fv2X<(N(euNV
zW^T*n2F!o<#h8G^edmq?W*a>4Y|l8l#Y|@*pDh19Dmy<HI1s+?tM9gRyimOXSZpRj
zcDsbu=tIO_Q&3kZC!c1BS?i4}*N}ylAd_E$IPLDzb=NLMp2!G>meP+d-3P|0*ADg^
zjpHc!PL*>mLLPKf$&n5d^{5kYqoow6-L5}FPB)<N5kv~0ItG%gRI=4jhR}tW{agTZ
zupbRUPiJpSY63K?jHH7z&b510rG`Z-iaCCL6VFCO>IWmZ+JF;rP>+3mfw=l5YpWJe
z;*FLw7-O1M9)SlK3V_de5A25X)&l>HF2LOE>!e%u9|!^f4b<iqpum&26q`i6-J3NF
zNNW8IaQJ}Ji1>7ivAviL_@cY#8?fYf<izzs)UmqW57fwCG=K<E?|T`ne9U!c=Vd!d
z=MqRKq}++=JYGvT@u_OGt!%iD*mnm?XC2P0PE71c&=i#sy3O%`h8Ge*Ur1H9@=(8o
zt1K=|vGd_r9)E=Sp5BxU&7k0XwWRY6>U<7Ckm%3LWFx}}0zw<p4w|7yL+}FvQ6Dq9
zsSZ2f@TOl1vVN#-7HKM06~XyaG;xI2*?jcgH^D?Ps^pg80o&Zt(#T8Zq=<c#$8Tg1
z2?;cz`UsHrqweM7I`-x#i?s}-uK;Gl`Fz*@cB73BJq6fQc>DBzFQz=-*B_6{O^h>u
zL6!4}@oA>q8x}}OF^{#CF{}qfl6yWdPW_hY?DbdW1wXbi+0+lwvBo7j#gWKZgpKG{
z&R%_y4*1kR=kTr-CKru?e<KMMbO+Nh<{I~+!@nqKs_(vEX?zJ*Ks6QMf!Xi$omG|7
z>AvGcz<peV<<u=TQPQOar{XA?7jvCUsGRjivrQG=Ty0#Q-2E`8_A&i3ZcFb(UyGqY
z>b_6@Z;BFFDF7pQY+%Ush7ojS7L^K9A#fmYA#fJ_wnwt8UvVyZ#y;}^Tugl+>fSci
z2mTA_tAu<mw=P^ubSTDuAJl=+HUIMlx{n^(8n8&}P3B6k28UZEk$(#VOE6S+SBVPE
zeALS1nn>}Rf^8c&hA6Xblty8Rrv9O}S}WBJdnrFEVzlvQdh3}XLBo$iPN@G$%<oD7
z+Qw7W^qk;s&owYZML+ED4TA@*qeI#3VF-Ex-2j3eCu>k38#w%H>>BfXLAX2--OorO
zi%2+&o|la7c|H2xQx$06IZ6U*AnIUCj{qwkRW-G~-yi&AZn18WH=?!yfafoX%kp0^
z26C`vVy*j)c>@pw+Q{@nGpKyU|7ky`<MQwm@8`P&eS=^_!7LIh%%83RhNTZ+=LTLs
z6$Nf^_yHetI9YTLm5kkwe37q(H^bP$v6rQBl5s(KHpCQ$BX31k7#BF7T|jVal^r=-
z!j&CEyS~AkgG(y4PMQ=4-@Z$As5bm8aqqA>bYZ-Kx<{BgR-nsyA$O9?lC>iNX=AXc
zpfJa(A-88H6OA`HZ<VVy9=e0(sWZL0`uF2u1-T3D&jnyK(Tfm(=w*Pn)S|q}m@D?`
zS_OvY<U=CH&{ff**1T50#(N%^5`h4}L{{VTe>i@gTaxSkar2^aqK+P5!TeZpUJB=B
zYO9{4`qd%$xt>id%X+1$(iew#mfv>0!`TD(+|_n=9x(O0w7-P?l|-PQ+Zefj{jmIa
zkos6Vet793#yUWcS5H3`eHj|zlE5`Mo~Qojnvu1Jl$)ukct97UQG6KNf2VvAB&+?k
zKGD3d8o}x+@!~0-ZGjBjhsP!H#@CU<Q1EWz4n})!1qxarYB4r38ZK|bN?q}5d+gJd
zw>FqSrekhW#?ztf^;NXtddX#@Yq7^o06i_OuIaNd(iH53TQZNGLG#%VuzaBbPAZFB
zIzK_ad@3CFM>)Oda~@v;!{v!Q4`7>8H(Lcl#2PWV$`3pO_-{VIw>mA#TW8DGck=(+
zFdR#?>LX4J%gkjRJ<Eo2#`AEtu$SW|Ig{q95LJqVsu(9O+PrkYjx7$Y)?PZd9wgL(
zNIbK7pj?&qeh73qbS_cX^|G=J5;j*c{8CAB`?;YXYF^i(n~;=N1_aud7K+4ajm`aL
z6~$I3=>u;JyjEN&-<20q)m#>1Il(QayN^~{#wePpK(4Q`k;>8IW!QWb(nMU7Dod?5
zC-J<6)-1g9W5-IS!+iTRS?ntdDrV4gLTJUi3t)LNeo=6pTwCpo;XjrXtsg?BQs6tF
za{jxZt1j4RI+5LS>khcfwm_3i^%j%K28Q8r_-CNsQUCe=;v6VU{0qz-&?&p0JQ(f8
zS{)u<?$$DR6CmbtJ$lkZ?<*|~7^B=udAxK^X#)klbjqHLmnX3h_c{~_D~Q9|kZlp@
z+BBF!9l|F`=C>Fi?Ma2DkMR=ZT;e(WH}+p(+?x|8CPubcFNDM6I%3xM+%il(-d-hK
zh=Wp2ao71w;Lr1Bi+z{J)<7LO#?N=gf8sYh58TE1bVZUpd2+l_m-*h<@mWt)lcrRT
zn)yc~J4tiwv2SHoY`ah^^hF=2xb~nCfZPYj56iKy!V}{Ms_dpMhy~7sCkV6!vSeTd
zuyed2(iZ=7{M%&*#eZ!|8p(yqz;JP?Noqss`4I1IEmb38NVl;x2dkit1Wb#q-Q&is
z5m=&2g!fmq`QMZX<vVao+E87+@cUlMoq5m#{N@->Dy+Z<{xn0_Zku|%bcL^l_Q+p6
zxYlnr@TekTTSh$t+S$|Ld>x=6HW=kIm!R1ni?kj4o?u_L-^}v5Dzw#YE&e8kBL`H&
zHmmOtM|=|SM&Hb>(W++1c^EaS{815Jr!rpO0awLd<1|Ovl|s<<T<@POvp#PaY24?F
z7zE6m`X7QZ0wc3h1AQ0jZFFb3zXf-aFZ`=tIdI?Ef$R`1ZdCJkmM#&0)}YTpO&;HM
z7nnK$v%TUpZ=d~c%)@U=z*}7A!K;_6WX@1;)Gv~a#QUMXq#CAQ*1qf2jh?WN&vykD
zwQu2TW7-;>bQKJ1SjOV=GyfC`3(UpWd?0=-Ti$Kdmgn=xlXDMZ^4B-e`k?PzXx_*2
zcNOVp_xz0?t`QVRm+{9RrJ#QOO>DBhiaAGvQ6`~I;V#ZkB6@$x!`kcLZjZELizLbn
z&~W{mi_Fc(Z1VtJGiU+t<&P1F$E<(tnTI4QTGsYxeh(-@i`s1Svpbb*^jgEuliw`f
z$ydsvTt!5<EZ~h{iR&oZcpCqHmj4i7^^OLG?SXu~ifH{}Ne8q(O3^O#bRL^FDV7_v
zs64}80oA~M4XwD)!qws5P-nj9JF9HHhDM1ZIJ&a);RaqX^F>B*x(VlS#K+{froWW{
zDF)9Nf*y&;JpL;j=(jeMXjwfWYx|$qdc_aH+liV=y_(n-*8*<gJq9HiMZas!XXx&a
zbl(QQe~s_bt}4~ow`YX3B%$V!a$9Ss%EB!*@e2xC&a>=(a!2gqN~oRnh=cuU<JbcJ
z5}Z}bAZ1ZmuD1I9dvI2(*aVzmaUDaOed~Q)H}6_#$DfnO2C1W->n_%LE)9Pt!>!jY
z5KC-_#UD(*Vo_pv4}!=W!FD-_e<oqwkX<H!0=5lZr~5_gQ{dfCT&<uB@R5P<Ts}y)
z%+H^BT^HP<EPdtqxk;=t3$5u?=%r4?pT;u7N^@6hHDJdEKN*4u)D%Rjli_zJl})*Q
z_b`39+_BvX=dL0j#!t-%4S~`Tncs`IdcSz6PFZ4VRWZw~<8WKBGX<<%`$$#gLyZu-
ziP!_`XIgJ-yqBF%+QRwCyiSBdysu7Q5KRQ2);~gKU6E12tLD=SCLpE+%@$ZtENFe)
z*;Ckg=c?`JSKjo<EcTl@ic$|3G-95b#n0l!wl`aO1F=wgeH0{JI#5k6M6UhPQJu_i
zZJ8TGJLHX1^!@!C&atELLyw5$`Qr=nCs_bRL!bcU4U#zN_uE5rEj!z4=X8Bm1XjjH
zdzhrVy~`~H9E3Qm{A+{1dkn+5qz|9)We)jMWMFlnbPO6p5=(uM*;d#N1ZgriI5`uN
zXifr8C(p#!XShb_F3;ZDTH2wXZePJ)%-%!KUVzSpARix0lf~^~L&}9NFFAzt&G+U&
z=hQ+I3_HK;nObV-3z=cO5ljVykJ?+vitv%H?%Z%)YU|PUh<KEb$viWkRmDAZ+IHGA
zeyV7F4{$qRAw|0KV88k?%~S>P)Oo7|Z|$ZLs!z?_*#Xuy71l|m1uU}R@|fV0<vqYx
zX$%k)-KX%FVYOU59x?Q=;^7iCbvppZU3rOdwL_{o`CuS@qOvZY>ld}%x^2^l>s`3L
zV1#%Qxbi*FdWQIF?=8PS40GtcTrTW#&*HELb+N1ThTf~yz4x>K-?v1*B1+`6sMIT-
zww!xoTp{w_vD+q!Ho=$sVEeC1$AyY8+G<fUSNE4@;{Hpg%QJ1s&B<pM8TcFNrPsM3
z?dW}=4h_#^@ya}V6Csjw=T%tWr;v)XR#&&}QsR8c3p;2C(gpeS)0GlZF36}eF<s85
zz>%(?(ItSs$ZR|+M6{gl4*ztu)g#eqJunea+`1H6?6}_F`l*^5=uz*7bFb?8#j~cd
zrpmzThyW$^#K#ReK0EL7y$#k99h7ByGZwOj1W>x53@<YX6cULX7v|R6K5ofO%{lDG
zEU7xS4!>ocX=I<Aj51S!xmG$vu`Xu-b%q)JAgNQ4ZEfr8`?%2^T}?~ubv}#GeJ0o}
z8`sQH9r0DPS9V*fb`W}~`e1uW(U$J4zpJ!4*{MCMSrB-;Ih3Ue#`S=<sX-D?f2Csx
z{cIzVaoI|zOP1yj&pY&u^_lRXMGQ|^D(tqp-(DtG?RJqR2M9Nr`r`VPSCI1nbJKHX
zc?OeKVDCN$8sIvBgAoqk%n?GZX<!Qg%oP8hve?n&q|wHJ%EGt$L@U#K@=u$%VRSEI
zBJgt?A*zg?a)<P0Bp<n)9aWv0(mOVm&e=H<EKPA*y#yxq4a%>MR*C+s(3|cbatQGf
z5vSOpbc`>v*`g1m*_S*=zQq~YP#83NWnZ%mU0qgvAOR7p`@3A_A>Fq+lQF|G?X=XI
zqd|Kl*rqOAt72k(Fcwf1yN{VkZ0_+sRN93ZrI)!)QPjym3jtr8$4{$BnZKg3NTeHX
zd-9RgC~JG_pD9^cF}46QISWwP*032uzs}wMAm2kG$t!&gMdA9NBec;nt#q}?b?fmE
zPrTVaAF5gZ#c=V{pPN9Y$fuDED@NAq;0PCu+2<(@iex#Ty-jCKM_ft8a|q~>D@cW@
zDF=bNpzA?gZfI?IDt$s#<Hy>_B{QYwJues?QXD|@@&w&cL6NY0APyupQS#hm$oT|e
z==FR;$)BBXJBek;m$F$~s4BGG<=yS(qNgPW&hxHnk{#$@#H|!e$c%lBwHiOp#s9$|
zE&%3Dz}ALg&}oVrfnD-)Dc&AXV`OAxal177$LtVH*VWazmt{>;2aYZQrxNe=@qrFO
zpe~<YJ08ybT>NJa*rmB0mi}RrTrGp_=UzmJ%1`_jLbngSZ%#EZl&ZNEmgn`<^B3OF
zU>UqV+49S(CK-&*zGM)~kHl9idGnntrdn~qU5+!ar$RZ%(&{_izn_e>Lh}{Nmqjc6
zMH4JKSB07O?ezvAlEK!JuC9P1oAeYg;K_l<;KIqxT|Fs}!v0sv4WN3yJGRU@Y%kqc
zI}|9Khw45<i)2OZ@ezU%@7v5oM<*^qP$xrhNirH4qwYipmeI4i7v=34TZtSoU19(1
zx$?dbo?h*QT3Ky{AyH&~-I1h593x;}7?ha+#~i64g)?q4^m?j>v4XVnEKTe*z1aD5
z<8fj3vvDWf_i-)*nfVccmBebv4m~sn@cTNc6SblLMUWi2=vY;9RRFjj+w);35eVy`
z|I-T(NhmXz#b8l%Ho43r1@8QT-k%;9olq^Vu9`<q=TBf9-TFV%6Bk_j_8P0&;JH{L
z2<hk4wOmN~B*GhocEp;jqpx9>jVZm_eduXO>2cB-!_>y$!ZqnzS|uk4pLA;sN>4ku
z{%MKrAx&HA9CbjP>%O@8_%9D=xa1lSNpW4XI*ybg$>?)*sY*C8Zw%_j3h)`l|D`WY
zF)#^eB=BlFPPp__-LFhzV@l~z1h9-5uua`3Ogc>+fRTK1<eueix{820z@z&E*4o-S
z5w!y65Bf0Y*DffXRqJGu>|={T$62;|Q^wM~$?DYajK7}Zh_QTeCF%J(g3p%t<$kCn
z^A&bGG%ae8U8N7n%vfka`6BimtvJtfeRO;+CaF8Z;njtry^>RyqxrU9nN*Avun}JS
z%8nF-LI4fy?w-<jf2K8?o&T$OKCjMrK|$l)#%KVV<@FY-j7c7oR{VGJvb(PVHYRZC
ztif+Ywrk&;JHEJk{Ky&3e`Btr`4}4Du@q)+s-U)b6YxVm8}AQ(VpVyTL0MV;Gi=ou
zkN@3z;(}isN+2OAA#W~Jg+F`~vpc>p&+|QIk^K#y3x2lNUyp`&Af_Ww`y56G)A~*<
ziTpla+JwS7^|f<!EUp74@D~hxOP#Y>rsOMJ6hok|SG^R@b3*;rN9dJ!u-wRwg3pp^
z^fb-1Npn6A6PRez{_-BM2H*YDFrKdqgaQC!67TWuO!w<2u{?yQ2H-6@9bb-syqigV
z<=Bw?2}%Bj3lY4<8LMOAS(9Da+@)!2LF_hZKGdlKUPC<IVi(8q>`~jtr`z99#J4*o
zeOO2JzerWM`$9XpLdrc{V-{v`YsR$hrNKL+SmcP|+top<bk(jG9|DX6Y0kTQA~Kcu
zf58MPxWr+XsPd~DIin4|*Q$r<$aF4|`>;p<!I(qBEW9*U#H*=~zHvhJ4t1aU|17|i
zQ=xoKMKxRdlwnxv*+dRJ2PbE|?U@WEH%o_OH?4Ah1#;&<rXC(o#}Sj*b^Qm4^*41x
zL(0^Ez!&W`FOzq>8y1_mRg;j3i^k>NWC+4g8s8JVjC{E+jD?|mx|hyu-G?m*_dfJF
z=p%Z;&;vuk)Xs8bm{q@VYE!!U)7r!oaVU5}-z=c~tk!e=J8xWz5%w#E5H9F_+?{pI
zB}T91TbvK3ttdRbc=zQhHq?3_+kD~&PF&C}ysryrJqk%Eff=+;VTsM`=LPlp`C+c5
zrCPw7CiV(9ifs_>h^bKsu1PQtj4o9uh#Mr%_^SIc>u8F12+=B!b`GmSodFIg%t89?
zeptj}I{qQy^&Y@$jK~<Y17l2T$sJcb&RT76X4I<BF`*#9qcx4L8B0{MlN@4OhS!;x
zagH1mZQ<kl7^id*Juw&HFx=6w{)W`N-}}-u$|}Kcy=)>dn%L9laC`)g1Mb!xZ{Fu3
znkn{NZ{`OP7*bjK_&38hspK;nK7^pFYff+<)+=d4fT(&;y|}GDcr~j<td0`tO^|e8
zXq)cB#W{pVF<o6^otny<W^D~^=_7Dm-Nd22;Xr+=%M-whGg~kd>Sw#gtzY;P7l6Vl
zKDlL2EniUVD_|<ZZvh3B<RtjBKdkZ7&$mvzUmjVd7~k=S&@zx|x4LttP6P<yF+)&*
z43Pm8<w6YUQgb&(xuI*|z9crr=D!Z&3^Y*^>Y&lFlU*bF=W+F+RoQ0xyw9C%Nc5~V
z#I!7+5x_*KPry@)AFczd6%1XdH}UeZ%8tV|Lh`WNW&sz9SP+I`KdatqW)5Ua3Myfq
znv28<YlHXA0j)ED8i+u*MdNe0L7(1|MU5rjzfy)VGMvG5yn*z}k$7hhh^==M{<gVF
z?;RvHlrx|6I^|*fJ91#!D$ZmhX@tJ<sS0Tq!3ThyNC>B1+n0c|oqOyxLmOZNVg2CI
zzr&=~TP{FfxCZ-V>?T+2)q7Gz`vxH6vH21JO(CC|9Ca{yfzrQGI3B~V(zLQTm>;+O
z+l|-|$-J&HLqqyrOKH+wJN<DP#)kNKK6lIwJ<%@6B3;yWV}t{COqHP>6Mfe_BF%5m
zN|VzNh{=vf=!pGjwsX-H+sFEQu)f`6({GF2$b)C^@A?O~eYxRm-135u15L-jAs3r0
zMQeX7{mBukLF?2Yo(2?Wx`Fl=uP7fPxkWzz7cs;8sHM~)BOZIEPG<PBji+`_Gd^a~
zX93W%YA?<C{p&Imi1}NT)*bm9)}|21Rui#K@*FoxJFT42WR16AtwHkJ1>zO%pW4dr
zQl@5w7fBBRFiy=bXQMLbm#XsesQ++K+i_v|%2-3k;P1mpQ2*j4o<KV%wc9|*&g#0f
zKpWTohYeUA)`vy}!+pZjma9=qh$f&?F2D}qO;BUkryb!Oed&imD;wRvHsu(s3to~O
z?~Lhebl9uv!0^mCe*_v>GVAKNx`Kz`AUo4@PC(ahW%_60l(6$1t~ulYfnbC-Z!{}q
zIw-PPCa#r_N~HZ|W^lv@F<N;J7BUsl2H>!ok^>0sE}v6W0%w*qyiRmtr{->ylPGbl
zlWng%?EFH6EbDB+kwxP2#j(f%l#;d7sM?bSrqDtU$KNNJ`|v-?hK(K-g>t$yV7oj!
z$sI39#J$#P>QpI;u8X7Z>*!mXqyK|jfK&?{0+-KbV)|d3{0~;^tZ(~Lb#ViNJ2_6z
zp}%79EsisWXSkFW^uJuDi<KN13M<jk?$C(}^%80mJsXoaO)H%P8JxY=Q@~?G^nN+g
z1>%D0o*_pmK+%Q6bO^x#BqWBneQy(F^6rCw!Qr8~GZ~)>PbkvUK1K`1dyIq%ZOk*#
zesP|zqe6u=K2~_Er<3|1v#I>b8}jklnVG*Kxb*7f2`I!3B+GbE)(^L4wKqvuUJ-b5
zP(lfm5j@bqYt}R()0Vf41^XIcyaDc}YHo1oM#SnD_urC7r@(<+{4Y626stN+nA=xP
zVrz*l?Y)|%SfVRE6ESnSTJ#lxtf941)``Af#OeIri9?GYBztPqQr`v)yKP4OcZ;5N
zwL`6|lJ?tqA;yjf?kLQu;hytF%j9T05BUmRs5FJT{RG|z)&7bf`=;w{&|aK&m#015
zAuB<)7wn~0Zx}=EE&Gb%a*e%n2i_@FWWi5P+}T_UpbV)%wQqxtA@lV5dT?ILw2CrA
zer8L>u+vH2eb#cN0BO}#XA7kL!RCsX6?-+SWn+vg^+^gA<-%%dcR3@n9G4I5idja^
zv@|&h4K(!D2&~a>i?l`f27cYy)6POpIAe_`atwYN=k1OW^;cl<xY3077W8?0T3dpM
zaw9yI=?ie=8hjtfc%a2>H3GyCWXItD5%rY;QT@-?ba!_t4NIr=8c2zBEgjMbOG(Er
zrGO|Qf`ot)Qc5kgbhngrEGz<&((zn>`u#tze24emozHyc%sF#r^4%z|QD5$S571+9
z_@WO7m8!pn)a}-;q|IK8Bh7eP5GOu!?e9SJ4VEa*tF$wlQ13V}8iW(Bzkdk}KlPwa
zV!dq@-|OJ0si5#%Tu%3xCq#~{9dMVgz5@;`Hu8P&F($wMS>qais9fyrFO_F!k6&8}
zYK;UZg+1KxN1i@d``WvRQuN#051nboHjz28xRD6|?e-f><gC4=;(HcqtrUJ+(30}_
zw8!Mmb`0TVmVuqs1SUQU@`b1OBw%Vw1ekyxTfAt!KC#2_Ys00-@!{Y`>sKo)G9aHv
zM$T2&lcX`>H9QFvu_|Z*Rm}adP4L00niu`?BlCb|<%P`*AtDMbAkS@Sp5=5JxT~jY
zG72}m;++?y7{#Tlj>S?djUL6|hrl^oLf1x;8+0I%lCwi8ltXyAh#F97(h!F5IX6{d
zV(!)}!JCtx_UY*spnFh+ZNZ6QO3t&EKFv6t^}Z;R5A{}<2g}Yk<kvVD3_Xx?wHFA`
zirBfkVN0QU*vitc^cRhVdO^+Q=jA~(t+{@9;T-|(%b{2H-!xU`fLzG)-1tPFjnm4_
zw_s$3B(&`*R->0mckBgrmDJMo-F@Sfjwvm+uq3OWU-&0gQzJr`(}T{Ny1{1-<QOxr
zt(r8Ht0IEsM7BQ+S~bGtSo9v<F38LabGKkfy$Jox?0Nnqe`zM>QB?l7Y0)}qVyBSn
zkaL528>hWciyd<{lYab`KNYOs;SRULc_-EVorHF8MD6EhO_C70VOEnLSWyrBF5`Fo
z<k&l?+P@CVCt%X1*_04ygGPE%h5?>>9!Esv)OEExRkYM58xx%0sq262>x?*NlGme9
z1&JMNh<)D~W=MKp*DgsjcLe2!jLni4?YDqJIaYIra$F=&9=pvlo!_6^3ht{TZ0~;}
zyKv(kiuo&YMfvdFfhM~WxH{(eCNIP?VTpSmy$i!k#jM$=)DojvOHV90c>$Trcepwf
zw|bP(Xw$eqP#YZJZnZc)t}ov=e%ALaGfZcCpYQhr1<Y@XdeZlPI=S(;UcFx`UtNCL
z3a;IJTzYl`-JsWgA)I-(mm|f(Vx&irRBTrKNRRGD5j|`*z+_3s<O!y70=}pdGtqq7
zH|VV=jU+H!BuL+~>cX@kb(!=GDc%-5aS117?$$}Xj&9ceW1(mPjDX|r29@iE7do4y
z#1y(Q)Fc*$;kuPQ3)dH55W0uP;(r@?+d1+BfHb6lW0ymivsl^Pk4p)>7`DCLfZa*|
z@&Iwu{>3-MWb(St@D6u0K%EesW<aS6pp@R69dAB%>TYr}B1jjYm$fgk*SAqeZT7B1
zJKI#;G2`1s@5|{s(<;t&z3T?SzX@kKFkGWFL|w33Lo5e@R!xICifocd$t=d;gT{e~
zo%eHy`ne_wIjK?saYni3aJSrG4k%sm(xx{^^M0)4SzK>?!<!AIx<@(ufFUhBL}ylK
z@!$aPHe6TkluI%-lfB#m&Z+Ie$sfnaV8FC$ztjpPb8OLv-E{cwXfo}cb)aJ1eQ?TG
z(uaPaJsgp%uYiP4r~vm#0A<NkImA`zQMv_3@GvmNMDl0yc{{4Z;ee}r?Mai?7jtZ4
zW*H{#n9N}o5S}<w>t`l(3EcE}C;v{bvw%4pyYb8e2@&f>&jrWnD*7Z5@(ph$K6xu0
z5_`w(>6h2YYnwS7pVk?%pP4mbi0AV>y~Sn=hmWkTnRXoC%{|#<o~hk<NELS<&=wU?
z^-%-WSLu1%v|jG^p3&iZu^JDlL67<6MiRK)I?-cvfXxpp&=26ZT>&Ob1_P~*FgYzw
z0dMa<EJoc7(b%ZhfNh-39~9_4XGEz|Vn3R?xs3}-e!F+G?lTw2nnyICTV-#Eq(OWP
z+O0`%C_|^q11NAA(C`d~3qPkkRcv%)a?|<70I1e_v3R_&sFk`^UWQF~MZ8pHY{Y0}
zNOr``3WONSnK07hnGyKQXACgretJjIo%>QXGbDu0i!q7+N~}ii!r6BB;Zq(i70u%p
z&g0iQP9y|K^}Byw(lD}GI&sVSceX|tZ`G;h5r0M7xZ=)O{{xfNIu}Ea;wr#?$<~HV
z_G<dcxf(t^S^<z2mq}{Vm45_wpLzjRR0rUwwtX?5-=N_hUT)iH3`0OH>#jKlx8=Lh
zcI4>{IIrk;WbT_N0iKXHd}Ni2&mX0=oNp@MzDsSF&(pLJbvV4PA+~<H_E6W5u~vxB
z_H+Ig4<Dxi;V0s=L-|Gmf0CREaVISW$8Z{a9>#J5OQK?CUA*yYjs_R;w_8{kT@MsM
zH<mv!gcFQkQHvC7&T^_JH9&Y@|L^Stf*fEKots}<*1>22qnXRpsq05LE04WEyy#*B
z0YR-k#73qogHS^;MJ#f>1~90fq^A=jSqD7mc(+Bom6UVbQHF^bN11c=*?vM|Q1cuM
zBeaFiv~WFgU5%x8SLjZsWpfx%2Z{R4%Q7g3Dp&y8U!lhnr1gW_B}5_>SBcj&9EDzI
zPi5`JV#f{0?7Ru{9o1;D#ZOo;?N{3&Vwk?;4`j_wei!pV#n5Y&4w2d6Y75@u8JfI}
zj?TUVhcT8Jplh8&0XZD+hJQa-N$?7>Cb{{|xd#*U8l@2$!)|rrYjmd8gGQzM+#6(5
ze6ilN>s~vH_wXe{TRWW=iWqnN5ZI<DwawhmBlJ87IEXfYSly}8WQ-f`9?(p*kYfSV
z?hY`=k5n<BhAfIU*Ad0jW9ZjT+?-89$d+WmxU<RFH{7XiCL#KM9G;J#5`O!}U5@*O
zQ+d!A;A%ZrUn5J)HyhY?o+NW^O43p&&zPIm?j3&+nD1gClb4XL7>^EYpeGUO6DGAA
z*ZyLbmP=ZJB(2;2U-Z3JcleaG!HYgv_tDtQd@t~uc{2@4rL^>oWtibx<EA*lBNMLJ
z4$pKD<Y!|GaEd?`yGy`f+Xv`rH3$TvsvCX@CrtR$$lcy<uRpZ-23ST^iA(PgcjzVx
zG4~Hd2pDwfaFbLwiO#t{hlHp7I%H`|<c=}Hi{E?i*dFWQ(|z^pO@+)|TT#bJ1qQk;
z(uMZPQDFn%AE{Ac3tUAmsZm8e>})V4nx;HbAWiABN?i#P0gUFe5tlA&V&(Q@2sSC9
z`qNygZ2NN#=TXkQBF^%{Jnro9A}AKqo9BMbrrIhJtrZ|fn4_>W-b8}^qlL^tw|VAO
z=XimU_vA2@VEp{bJ+A!>8o%by{?k3gRp@iI^wsCtf3nYAs$#CU)WPm@o8^Vq-AY%q
zeEnl`LY8SSY-bIu&aghP1pq&02LOUM8c5Qv^P*lSRleK-TqL6W7yFt+bd(~N0Gm&f
zq-ntV(9wU?n_~Da;%7<em9)Y4(Rj5Sb2qeiP-4Xd?g={JQzNSOWj#6Ac>gPRmasB<
zT#|LQ52urwU%hZ#=%G?}w;p(iPXY%wv4Rncmy88VgAa!@+K0Q76Jm49e>H%!)^4{r
z>pj)t`;;qq3G>AiRN8h6JSE1Adi~6z*o}@+Fm>G_>dZt-yS@9q;P>uK8}K(wX>PMT
z5NnjYx|+TTF6(sR!8^-(N*odBs@I8$IwoYzgHz2)NY4Lv>my;G;};G-jad${IJSSI
zrr*-^&nb|}2_{Y>QN(}k)zv=TwDHQTuk7jmk0&F)A-{kN6b4vmb$ovMQJjq#uqxah
zX``wReye_$K?Q|lMaIxh(*Z0*r3A($W1Esp6>fp)I^P99F`Lf@bNtfKh6Uc-0e8kt
zMCCg5z<sgc2ggBizIXNcpS|klWO_8Cp{DRI`u=#vY?n$rreOwz{J@u=ld!g4uq2vh
ztM<37`sq9F<XZ2?O2-T~u}MEPzE-QU8bFQ<ogF2WEFaa8c*aPEFK&q%aF^!#=S%F2
zmZ(aWfVqmmdgNfc)NYXWB!;$>yrl0q2dyfBzDWs}Pm9{aNt<BQg})+(hEGG1O2Yr*
zL$6DKcau2Pvf9zWK8Z7IlX_WDuhj09Xlo$#&TMn*-mg|*ZU|5v-b4OTogfM81yBBH
ztAU1H&S{?A?Cg`!L+V?=XqxY3Yb!95v;1vBNAL{QP3@pSmc=gxjIy*8pTIy`hpz7K
zLwvBo#){Hcmv8pm>oyx8^ur84|F~BQGCtUM@pz@`<NCe+zTMy$QB3=eNxKbp{Ju1H
zSu$$@4ipwWN4t8u;!g0~hu^+E_4$H;IW5t1C#^-}r$ljHg)Ra<^`j9J25a(B52rF6
zFc^43SA(X1{J<I-8ro^r>RhFG(iynkOr|(4e;g~oQzq6MkFG5IibV>ceYV2{-vDt2
z2k)jd-Y8~*4SOm1$wAXY>5FS(krIDza0kJ;SqFd&3i!3T@y3(z_vL;bFDDD4JY{J(
zY&S%}qJFosY+!{D1t56qgVioTZH~9$XJZhM_XF5c{eDR@@NoU5R6zFn$2DIdEw0b6
zp+5bA{l^_eJhu@O<2tm9%bNipytmduH-K)r6OO-@>;k2~z1|Cx;1Fu0>cGm=G&1y+
zBAIiQo%gwo8f8iVYf5&<svzHl=X=ed2W+irjkLm3I5po=2msK`<Vml0O~LQ@p03N>
zaHz7GPCPC|fIWUNH(^0KR}<v`&)j{Ut!gaE1WQYEjIDjh$>i>^$6FL0U;TK&NT>Kf
zi`1@`o_N-pGloge$Yr2NTO{Ost3KWMi$w(W$;4I4{SJnX2j>q^x;&2Ia!#QUBo>HY
z62_7LvJGo0c`$#-QcI9Xq50;Xpxn>%ZxM}c6CF=rpp}vo&mcg$=GD*tDZ{7ziet5|
z^;-DPqRIj1GaraJo??j$M}EL*)a2AuwQndea$!16|8{Qjlc$w*>rcRyFZz#9ywDdB
zs;WUF7NZcrW?WQs5ZFDklXO&TY?c1COw=_l&*NC?H?t_k$-Jv9$y|`tjBk=+p8h8$
z>@-;wSm8G3FM!B3%_B70X-Ou8JT%KEafmmCe~wo3%g`61p#GH>4ob#&*<eQUdE9M!
zEQ0#rlmrR=^<u?)epl@@p#dWNWpw-p*4UI0o@GEzz=v?5S75r!C+QW;Np_32A>f6e
zH_Degtpju-^PO${yI5Avge?Y-F7hM@DLZ{5MgGPUsw8WG?5X*1D==wUMIMG1?`>Py
z@{xTK1YdWf1yqa%Xvn1BBYt)p-5x*%MY{RHpRFL}@rV-_^m;?4wU4lguEs6G$#^)L
zxQ-Do=45u!rvYeF0+vF`xTdxNgJM0%k+2&dBR}4|QS>f=NG=jEgeCE4T_anXzVi%y
zh!DoZZ^V7pJ3{H}p6A8j$Bf|JX*g{Wb=zAJwO+QS{&1f?jl}rj<2;lZ?13|fKT?aA
zIQoqP`Ecb0BEXqzW`Gz~N%}ko6VmOA_tY7h6>yv8_08bbJqS~yN3<)(XJ_nT6AX0)
zyf4DSBrRqig((ucIbLfqRUIN(MSi`AARxD?;LV^YazU6ancR-O1>EhHQXUXXKhM=G
zSOjS!!h76A1SIbMulO>k>YqKwS_yp>XdSIb%V^$*&uT;?Yd5NU)*7Ni?v}0mkU5u!
zB|=qdp&FEcom{m}c4v3&X>w4s*US;%>Hl^@npvMPz^m_mfb;KDukrBgm8Fg+QNv1!
zfG)VXCxtw_Fh=#%HyhXzC<H9e%=RC?xB|MJ)hOc%%j~0zquHFSJi(so_L35pnUERF
z%Do%>_R5fxC=W=ZL%r!BYBj`&A;kt^5x{y&9Dy=Buq^kA#kbEu!8xh(8NTtSTsou)
zdoQn|VA2$8s5p)kVngh2FW1$w3mM|Y!G_wJz>*aX)3>ZVqr367;%F)lV@Wz}sZ?oe
z?GY`vZ`|@+&@F_~w8?GX^qO**O<y}MbAfpKH>e`N9OUthzN}3q_9(LAI63S8!m24t
z#9+Ygj`HqJdIEi9KFOEw{<H%n-BX~5gZIYBs%t-&qYJO_qowpGc1NimbGzzjy5bYa
z4KT_a1t^A-QZaPiiNoa7&@|az?Sz$poA|~zZH2~Q|7$FUnoy-i#gC!FTIJ@XyvL_x
zCrE~CfCwZ7l&Wh)EP;I=c}bG~nCpxXHihHA`%^(wIu0=<kG#`-ZA$>^gVk<#vy!;@
zErdLoxm^ZZpM~P9Y=R(K7*t-gMCPOR^%g8|+c`_uN1+Yvj`y3;M?{aS4vge&dW9tH
zr%~(e$+mMJCZ`>Clo8tS#==Zp2bGNOek?;n0f}{fcxI95K#}2wU(GkBQCK<s$qSp$
zaDr9Qpa$rWcLUV?JkMxaSH|Y&*?|biQZ_6bM{5~_hq2NCUzdOUNp=?K@1aaBYg!Q$
zb52~m-pElVe7zqgaE_!lWAezpm78zvsq(VDomXk&f6j9@{&RNUW|vFdnaKiitz<eu
z^OS*g?`4vIg6QdZPmSg>`@knvP5G|0&riBmVm8cV%a?RE6nJGI0Bqj}HqK_g-h}m%
zdRG=4{6V%Dasm6xO;Z<*OV?xoOJ-j#Xhn+!Sr$tl@YA}!=-`ZD#l>g!lLe;K=5W{;
z5rb^P8!o0E1}5G$_A&rc#Dl`hwsn%JM<}l?I42gWbG4+_Rst*NVnCGa74eoJYDA}k
z$k5C7o_!J=VQ7O{<@ZXdeiqT!j(jY_l;wE`<6z37yq^fmB@>yzpV0>*LYB)!5XASM
z)HDmxPXnl(t^ZY!araA}p$z>9nvN?k-P55Za>0A3T)rrQYuuS~ee$!l;TYsd-7K*@
z%*SaAZomQ{d*B1v20WkiuLLa)S`L@`U-8f1JD$nk75NaCXePtXW-8BU)U$(%!arT4
z`!IK^g*IO3@@EAs{KDaIcvIajt~lEJ&jv9sBqqN7u`h~CA&_128AXDy=fKT+rWv2|
z#?E$0Nl9e$$wH3m^l&4&6o;VwTU-IhoHW~15^r?-2Vvdy%9oX?_Rk*n#C$nc>*5x-
zj09mi+Zv^#52|~5V|K1BGFPDQrUUOc^^X%%B9U6ck`tud_#_LZ0YYSBCf~gp%}DEr
zCrBoW-=h5DR7jmozB$fM)k#iizdx8X77@u(_J6f*-<Mk`uR%1CXiX|-4So~a?+`Pr
z2Nt;ZFMCn;-)b>pWU>bfKkcyHOx^irv|&c=8QDmu6K+bny}Z0Q*Vdwl*LIQ}R6X8S
z+0%D<s?EIIz~1WM5cXIbmlFNzh3}9B+G>HeoTZu_N_v^jzEPTy8CmRG`fd${BeGl=
zQ*%IPb=J<PJ);jVjA-a5xBavBCIB!tTLu_;p|g5Xu|4|MGhAlt?1Ud!5ZN6jDL~or
z{!+$kRN8BF<xzVCK06+y(O&|>&kQ9@7hn;;s5E(!?yF2ZM1jf7;c>L7=@7?NEXv|n
zKnpKGlHRa1<Op_Nw$rfXbLKCMHQR<eU}YHU@IC5dR^bu($TW6{z}tG;Z?{hRiFid2
z?f`ep<Rl+cIT?#cXLRAR2QQ7AD7$udR#=+Kdj4`D)35d^Y{1B%Zj#&z%uK-E;E5Fa
zdl+d8o3cfge;HSNx}A|YeB++BJRv`BzWB8SPlv5i@TzR&1b|d_S?E<&BmaoWgNG*U
zo)Pp$Aw77~_oz=%=F>Y+jw_o_QF%{%PdeKzqInx{Eze06T6SFVY&^ugJ`4KIkh9IN
z$u3EOOsjEDmcqh(!oTHQsW<{$g3}5ou<oOh^icRdgOWF}%&+Z_zuph_B`Po*W>ULB
z*PpSJA+3(yOTf7i3CNdKjk&zmpYyh}1(C}@%7b$S-Q*4NpoE*6V$`I&oJ{dFcRbrN
zaGj0B0Xs`?Q$+~Yx0g7oukM+9j25b9-m^I@10Cq&xH34v#bV>cJ3cdHW{DztB=0mN
zUrRDI24r%ogy?tdc=9|6cDR|}GnV68F^qYW>ixW4FLC+Rw}d$qo~_5zBgxHw)%<KJ
z$0qUQ{$-ft@5|-3R@J*#x{qGG7gOCIs!>Mv)+krbWw6Ccw7t0Rk}JmQBvGT{jfke{
zdF5m@99>`1?%v3Xx$qUoYd!z5uX1~<(O~%J%)><x#>-L&mdu_v!|`F<SkJhM5M^g2
zxroDdb4hu*qoSp3oe%*juyX6q0*|vE<(m~mZE)_uPPRUc(eT-yNiwo~I~Y<@QY1R#
zB}9`DM2BA^iCxHxLy9#5pZBs9Fk!T~n#4<SFm_@<A16M~$34yPx3tur?eD1(G_LRA
z9bFSdszfscf*bC^pPsYP4{wNLCg58UpfyS}LsZ1~_B0YC!P)pO>LFHWHycx%Qn|(f
zuSN=8sSUTd&jH$*v$(oT^P-+p>wOi&?<Im@31HNQM*~FiQC!~Fvg8*nf14t{e6%=L
z*WHTwo=WF*>cW=S&`CBh;N-+O`qYw!*BekFs_A}W{Z9*!F5LHdn@R_r^84ziXC4m?
zywqPF1JA(!H);F+#OC|74nR`|h6rpl<ft*f`Qo*=DzXx1b4OwN&w|cBlOB*2!E_|?
z4r36ogqlGJm2V_PxNR$33EBMh;prJ#-hHu2Zu@3k&z28M>`wzd0FX?ZveRR~x{sAs
zVMaV)^d-t(pbLXQ)!=nM=B!0IOWpvq<}R8jW~UIw83<2jNO3!fY><Qa;gfI*rs}=I
zAO*;s?Mz`d%i|Yu+X#lYSyOKNF#2eakg!|V4TvR!<rSkdJk#;S$5~-W_bt<uj0`A4
z(f22_kJE_RhSz7Hwd2tY;eSM)@xCK<XrUj^9>?_`iZ;T%XZ5+8`GfoJy^V;N)RJE6
z)EhddM`Y~iL({ge%EQsGx5L64{qHdiFFBeFW7K~0C2nf)WAm8<$`Yv+k-4u2IDx)p
zvs$)LeV{{QBCrAs8_p8eSJ|EdErlB(oW#-nH9e(Dv{9eZM#ANWYW9~$*KF!Menowt
zloa)28^9V1j0n~w;m#BE?`CvDt|ciCt6*Kur<GZBfY2{s59S-Qq@SI*!+at^7CK^1
z&^}d0+lBZouXoMniFd()b7!kD1sJ4h@)<_L$$F$RfQ~bt*s<O+F`SgH<aL^(ALDR3
zAEUi!x|SLVSc)Nl@fs3$pM@%}0Mm@)MPHeH$Hmk};nI`$7eBH*TTOpbg|E;Rt>9Vd
zd=h-WhN3IMR670JM&wGZ`GI5gdHvh@?HAwzWmbJ!TijWPFd5}0)}*btv|+h&MUk8x
zJt2ch;FC&mlHA!d53~p_&9v2-z1q;;|HQ83s6&tUV!evL=xvc2M4TujBvOVVx8?RG
z2#>GfTQ3L44&cn>1xPJbfTdyHKa!-Xc#H(je`d;+oqkb*VK~Q#1kU~G;W6VV0=5Dp
z@2o;YAngb3i((`+s{XFCNv_rhhm*-#pBRttbJ<AzR20HYCe-2*R}zvlHO(Es^R{gs
z&s|SmcbUCtrRl@%uo=l^*5w{rOli}g<&Hw2!sDpa3^#wJ3y^A0N%?%P_Uf)2N#bUx
zeV8ZpqkWm#ZoO&`fKO*BF2P3h&qhaoG}#UxGggEF*MV)!07^?7dg9lIQPgk@l>;o5
zdj@7FP$xFr1b6<z07?NqFEBxrj2CN2kHY@FLV*->0|;XuyQ4+_>=pa)V-r7_&qd5T
zGljfv6U8wrpnad7*RGpYIg@!_s-w7iE*EPZI*|#r3-`8tfH_c4lcs>GJ#q9~n<sQd
zl)!%F6}jD>(^t&PV-EV(hiUAplbBo@1UHR)TV9cp`r3Fx&;OWSMF9hNFEQ0Q#|~r*
zm~txMCDvnyhRNzY?hy>qGbu55VmjO?sxd@G&_sfI{COsi_#{lg3f&)^PJA$1-Ep0l
zFARr>MnwS#MZ}pDD-*6El`SH{W;yM~AA?m{CTH2sDlsn-UB)L5sWQT|V&G2_g%idO
z(o+NFEt&RMnxRkYyMu0jiBBTX2+*b*=+hbs+L)!Bjf(dujA@X!T@DxNpX2!ZwQ60y
z9>Fw%m9}9zcgPT!G!CCM0!lSk`~vyQi)<XeML?SFB1<XIR2N<yw$fgK*4i>ln|Yei
zTQMHIWdsC|*aGje{GsVft*+yLzd#8<8(V%{eQbc5O#>gt&<om{&9N#gUgdrjxLY&b
z4mbw&Wif2eRD<b%CP$IOOUoPR=x*vU^ABWmYPdI`ruz;sJK2l69|Vx@YR-YDK0b)`
zSql@Y;vy5}w2UqoVPqzb0oI@-|Kz9=-@AJb1dh(BG4|-Prg8Lad-cFuu~xw!E=c+!
z#_qLbE$+Z@xL7Ok^aP75L?UcT=8n@)hWyBD3v@J}la(=O6oqFVnwTzP-w@0ai};(@
zfn&EG9I+GH9Ci!S@SWVI^{JE-qPz=5O+7V4I@0W+ZmYK`y?!U+tEH~8VkZsbW6`Bu
z^tIa~1<UIrQ@L%eAC#+k1-ot01U&!k7F)daaxK&Qtit+o_~C5+#Cbb$T3fd7l&Pru
zk8fFd#1&RYG27NjEBGp!LLS2-;+w8<;%$}KNVS#7+ySp4BsF6F9MYX@FUs}v+zo_j
znCXJ?Ha05#vN0={8=Qe_i`WVkW9Zo2s>Uu-r5Yd+Gl}A5T-|=J#lw**3Fl$1P*gL9
zUPtb=glxZe*56nsfH)I2Nfz&j8lw;6Q#>^p^BJnMjy6;OwPM~9b)Kt9EuQV%YrYBH
z^`#(U)l1y7R^o=Msu_pQOiI@c6vw1#MGE@Qke39*74O(<=P0p$0DKP500XW(V7a|$
z@R?T{JOH?(HgGLx)_#xBPS_k@z{;0lC;i6eM(u@jr2e=f%N)MIx{az3Z$-x~<nvjj
zkjply=_MRU&~fB5t~M2dl}B$+VS)Bd6T4J4Og7YM?Q&q_4CRrmS#DwGG*%G5HTZs;
zM68A@57gPwHZO#rdh40<L%lo~JXeD0gcw)rFnWOCB{eXT`Zy^1qd%OOwIYD!vXN@1
zL7sdvmdCD!*l8p*sC)WNi%)p+{4@$5;-Qmn{v>^Ar`!z?_j-}$1NH-dz6fox^gSz%
zMk~*ykk8{^jtXj5HIiN4%o*m5az`Ka@v!-E1IZ&uXP~b7dm|2Xd+HT8-o{6{S2g>M
z4eaenW8H&mRMxl{_n^OUsmU(6qu3kfb@3|Ck2b%4YeyjDd!7q!yr76BQ#>@0-8%j{
z+X4qR;N*W#9vO|VqyZBn%2HH2NE2e;g4qgQlOoy``0lXO*7AmH-tSqvCU{C<>b6|P
zC-R-NSNx)p<zRmtX)j^4$nsJM*DLme2pi`}ZbA&yx{v4mQkUKOT2mddKD#fQ47(<`
zw8)2}Yx9XuT30&LzrWpS7$t-_CylEC(m`7GoS|Ls$n@zl5oLlmF+!WaXSx4gMFOc<
zJ}tU``cG!9No+4@)$#^@T(a)Oa>LJG5$exitnj93!+K(}GbSVue!Kl382VS2)eq#r
zyy5H$mNht^zRm-l2f1u$r?(U}q>PgBXxx>4=H@oYC&tpWaN;#QA`;_Xx77C1PbODg
zEU58*?n7BB>4ndt<n08SAy$m1a^u7><}_lc=~sjrNtvF@xOc5=bljZtFYDs1Z@A`M
zC|&&_zCh}hfd0I9(Y*^Isp-tCVFcDdahHsTm7cK%TZVz(9MrtxS@YEl%eXu_=BI8Q
zlN6>$d$$|+SCvuMUo8)8c{^Z(>cqi+fBJ%WLymoJIOyYdLuOC%Jz2CCMz3&^g0&$L
z6jj5Y(*kOh`UEAlw7QqhEwm_eoY6FuT>3{M*tkuPe0fnbp4PxZhNCp~xqBt*cY^h4
z_GR^D)-pHV)d!zpIX8?$#nXB0?db|9-6-zlo<v#JMA=_B8wQMpFg%<G>oz?$m{y#9
zlhxq#j8+GX3`1`|a60WU<uEZKWh;8?e&GY#JPViD2!<^hO|kEi&WY)csVpTzqS&(0
zz=f!Z{if?d%(pbSy+WIH!ta2UQG1d}q$F57%0ytX(ycAhzeqbv6fz+xW71EL$WnyB
zj(u=x>s75E{v&^A)MxW3M3hPXzyW3sDv!B-BHf_-Jqf&Z{&6<7Qn$f#&C3UhPX+^f
zeLyel`|LAJPVWRSnY?fC<qhDU;y)Y<i9t#P<q%D3%uu)Bm`KXVbPT?eusNts9W8$3
z`T5uPuFOx|Sdtv}vKBnSk5}lcJ7~qzUTc!v;bH+ZB?@goVRtBv(P}0VYLaLB_F5o&
zk2JBt83EqGoxS#6|4C>Qv~*tgtL>WUy29SMc&?zk8vVoeuXn+uS{NPTV3rh?gzlGY
z70+l=;_lu}AyhATqdh<`#31E9o~~~=|IA~QyCcm<#an3<g#+SsL+M$fg`}e0mLE98
z{Hsq5BhpyCx5i#m-^`*`*FcyL7%k)1JX%}I5BbFgPtbGQ8*Q<ejmWt1tS@S?ICv5$
z8%v}ta|nLl2?o*)f3oXvQEQ{9HPb|;vvW%Q1WlFhC5vUa@B&<SL&&T$3tanJ`{I}a
zLchKF@v>G9-FmC%H@PEaaB_hlzs>>?BlLKng%3Q_zAqJOd~LkS)qL|26$ouofAszS
zyL^`h%d$bych+KuVnPxxAHJJQjqTyc)RL^Q__o|!OwFsM^9ehu+#(23-C&W4T_Sm~
z)hA*3f$tyGd%&-|NqqOTbmYDNwb1vk+vx$#CL#pCeNmgge}DOfbz3*G<)yy2dpJu#
zjb*GaKYY=<XxfVL!9to=Z=B;94`EZ;S|w8RwME>Ui(L}`ofmhHD}L<|TGhO-MSkB4
z4G$3K9~!2TXe?tm7Zp*Jd_x-eBTslMImY!qPIR2!{SV>K%5&0r`?p|=qp5z!UZ$fC
zC;rdT;sYd%$y<)jd9g9+cI#pp*im8zTQ$isd~2Gq6fw;fB#2e9*hh90NnN>ei$4(F
z2l=N(Pp-VZsdg*o^9C3T@-_$|hyiy%1x-IgM1$x(|J#8wC>Nj$jyvoK0|4z@Zc;P&
z#21&qGPHEdcmdaMZ;C!~uT_F&!k>4h#yqFY+-#<aQK<|<l3*s&w#jT<X<7Z!-99|s
zcCiAzTH-OE@Fv=qm2B%P>y<m2aP(`*8&U*?#)gmy5pnwMGpuw|vfeg9uHUY4z@}<h
zTY$7<0|_xpQj-EIK<X(qm}+`SDf~Su9Q3gunU^th3+9ZcZzK!v=h}1!r3cQ-Z|{ju
z<w;;jXpnR%h`f5xY3IUu;%)QJ<|WLv(@WMH+LaHpD+)b7Pk(HIWJXB#E%f6eyh+EN
z`J=0q;&H70g{^)>2fg|)<D|>5$*(VNA)Dd(Zp`1uJcj`7c$wU^p$$P#<uB~zWy`tt
z{SVkSjXH|C<^@c6{y*^`=N6*T<$b=T3;`h%mBT^T8!tgUR4h${jF$Pwx=u0-(b<b(
zGWB6-F`jTq={n@}(lD&MC~Z(l=%^M~eZ!sqw`gZ}MJ%DA_k~R-pWNF4dT^rmgD#@E
z<S5CeERYPkT|;@S0)-3l?!^-nqzQr8dWpB(->?w*I8Tu$I2<WuW}0KH@9N_=8;Q%V
zb^S9zWUb0%9vhLx4f8lSpe2V)^&~F;i{QYLjQuW*vc1OB4`tAD#)&Cky*L}uQB;ir
z)}s-1-p@oZMjVdVl@MlX;n5VGzzTNE0a64~)<Jp!glGBiC!EXi#fn`fS=}#h*({l7
z<R;bLQo%A|lu@MA1#Tf0@FQBW2qBV{Si^Weq##!c=^dkX&J!=jYs-h?`Epuv_M{GE
zk2_xD?hvBAx3{Q93yTcR_tLzY49KjUKj=JzI~N!-M_09g3?<%>JIqvX_!kR3WW(pS
ziw@Rw)ePQh547VAJJ{N+dL{{wDp(>i%ZEf0qnf0^3nu;l7v$T=^}`@O1(NS32Nmeq
ziSeW8a<z|Zb4?37?~(!)`?G}b1Nc`vjng$KeGf7mF&IVe+1E*h3qT@{^yk`_pLMEp
z+K1El$?CPwR<o4GcHr72rrTmLc7Km)aP0~wVHJ68#<z-MC4{T0&<71>O**Ki^9Xx3
zCXt&sRvk7y8Dxki;R(-6;E9yXG|!~gq#M2s7<Xb1oBA=%GK2;3$mPa&mc^EuM-C8L
zbcKU9ccsUu+vxx5mb?wc#>w6l%vzawpNun?gdyf$=kkFMO&=+lP~3(3egkY^9FDro
zHti6*)<XUXOPmqu2N~UYns3>E?$9&`$>TWP$46Q>_md2WW|vhq?~LwDt|oJD1j0Jf
zjvVSC4~z4%Y#IF0gyO~TOA8KqBPj}~`0Zr9^G0FK=Bu^C`os^8o4jfaD7i-_NPHzs
zqGFQvdqjss8I!=8uC6Y9`R6%W43}-Z!b93$9z>Q~=#w<(4s3J1^><VF-}B$qd!#qE
zteLV1-8}jo@^I1j!RJTQBcp@^CphtmB7oeSId33q8bxMll%LnWAiNgL+@`sLbO=u;
zyI0_{{Cx%kU1T7ZT-`2;BlLh}nOl#9)IR~{9!-1a8MuHA4j?ujWFN?QhV7{cT=>pD
z0ot4FQQ8FYghJuyz*Y)a(%^$%@4Uq1Pn2F@^+z<d1m{aRYxp({6`%2*@>OjSok-U0
z>{*qal5<&w>o6u_ez{aJmzYJF@E>e(|1DrPo0n`6I&*rB|I4W4PF<j|W(cGUFjmr!
z%ToZxd7FgClG6<j$JKZPszODyW3cTMmtzgA?vQ|wePI$&l3ZM5i#xxoMeI`Vz}*zB
zDJ}CeC?9(!l*%L`8J@WT+%ZUtxzoGVA}`?@s*?WBny-A$F@tbsbvWI?CAR9g%0#F+
z;Na@O>Sds(+d|rqHXC<@<u9Y;(Za@gkSjXB`d?{<D$$d*;?P+(yH(tX@R<JKixsVg
z-%sw>i#|&npXR>h)FVdq6zYK?Dw!A}lqGFXmY&Zh6Nz&DAcvPMu%g1pn?P*jN!-yw
zY;4)(m0}!nZi}@@At78h0%J`)sRm`Ol}daTL{@HL(RrEhQ0yC-J8OYk*1tv_x;9+L
zuf#F!ruOXzaKU`9y6k=)7ACY@L61L^%ec>vqQgN8pT3RC@y%|<*L0!!aho{E(cX>P
ze=n5pD~8ZbVwSC#=3D;>(2D!TFwTQ5JX~wc2R)l#-7vV}z;kUL(gpCl-OvOLl1z*?
z92C9>nedV>mu*B;vO-^TE$)gPw4-djx~z%t?+hf(O%R$^;#s9*I@x@EtNWH=RopQH
zP9@H%p+e8>Ei|+%t}+lVP}(Xu#3>kt28w(n6e+pj)nX5HG5LMB#=G7q1@7&VyG(!i
zavA}WYaS&=*c-c*MuEKl^(RwHHL$Ffg_MI&7YD%Et~Z;vBm!b?b3d^7#Av9V9~}BT
zfHrZ#Tw@Q_K0NDHlasYLetO5tXd46lahjno3bag1#H^D}V#e8`1s{M7JQQ~dw#hM(
zb^9I13FY)!Ig(QPX7X!|YSe*5hc1r^U^S6xER0Fzskk=pM!D4_4YcbVeH&B&@?`ua
zL7;nAJn9W4VyfJ%p$j*@DR*GkNq01rv<bN>uby<d0Lmoz{|{~G$JeRzvo_0TTJTyq
zVBae?I_zBofvWM`)RYmrTBNZ4Bp|}j*g1_}LnnjK=HcT7zhW0z33&npyd>allWiI)
zlS3ruR|DPXXUik%_u=A#3Se0qk`gYXCmFal@w-Y;lpflv*w24JLt?$MtDl&3roBaw
z5kZlw=R8S`n7IP+8+4EXoYr>NWwWHAdGke&b57Ig3EEIF&3zaP;+%s6W<(rrZU6Kd
zeekc=_$Bb}2*&*5ZYsN9^Kmo?0#fq8;_JVeqB=oc@AZ0a8`x&#%sC*;hc|z$QWbmv
z0hDJH{o<nHBICnOe3cqu`9-!{HaZ8G?L&H>3KpLx77WKU#(v7<CQqTx0aD80@%dL8
z_cRE<Sc2{-DP^azh$saO2uT>AOag-fInC=x5m0>lI8oA-7sVHuHKk6$@-Myz38sO5
z@U<mc**CF6QFGIiD|re3pZ85crZ-VCB3~%`E+oFru6swM@i}Mu5C<}wMox!7Vp{sK
zd>uIu&jyiMBj)!Cw<@1B39{s%MY@{J9inZDJ4e;U*E^C6--3-kd^Y*o#AORlzhm7I
z{Vg`tmaqSoWHd#S(6D1xqp-{rLZHjPb3LxTnUKmIFRo@-L;8RY1v^)p{_)$j<vMsU
z=wlivtF;YN#U*jHP@bjyBpl=wZa6Ck`M<ix@BvJu;!P--HG>@e&&!K!-sl(p>`~3u
zz&G9dfwi;k{jg`?RKObdSvYIcI-5vS<O);3b*_oOEZ6J^NlP*{0cm{-s~~m(7gpKK
z{(3B?z2kYFm}EmdwyApA1SM}0_M0y%B!b1FLpJ>+6Y82teE@z67MWy57-$ns_ToVh
zMB7spmmckp_L9ubI<kD_RMO2ojc38r&h!>VVP{i<Ejc}UasS=l4JCYvR&r{vv0LAR
zkepNYYot|bX36CdB;v9*0UN?9F2CpySlXjmm1v430$_X~&C#zgb)x_hP4@2O95kOx
z;TJ?jFS-%pXrW*}Owb>@KXYCZE5$eK)g_yCU}|YAZDL|phpENytX^$k``Jv{@F|=x
z?Wy5cp{IjO8lMdBK@T{4%eP=8)sp7Pc;&7g-K0J(SI-9T9c>2J8EFXb=#KjS#7+Ns
zp6L3w`FH<^8EyShsvw=4{JG=TD^Z0?|AIZAx+b#MqqFX_;=g@qN2j@!K@sVJgDg!X
z1ZznyOs(Zd`hi2xro{N)aBRJ1x3Pm-n_h;Lr?geSpT&8RVpo#TzZ9<UdL$&&PkL?_
zPy0A2*Kv4ob+oG=*Ra%BP`E8sV1f5EvbeSd)FJn^JfR0@1zIjuI(Ib1v4<U>%zHYO
zpAe3DF(>W-MPu$m5YiBemPcJ_`W`S+h8_J^;om<rwjMTaF%h+%|9H{;#<ROL)6Bf$
zVUJcr$$)8$`z8xAYp(X*i#VIt7l`%kooGYv#?BD=u#gA*`Gl6%*doC5C!Iuuj4J|V
zdLTtLO})*DSD!<ScP{)qc#*yvA0%NyY#L%`V*9ydiUdTN`+8%T+6OtPLMkPpa0j~`
z03Ti*1vfzEO$O5|88rs3wLi{z8RB{}=&W3eiO3=|ey2LkjbL5Ijd*?eFHyYJc86C4
zS0Nn&{UT=$!`LN_N#_pQOBQP1<HT3Hd(0KJlsA1`gXC<7gDNjt_Mc2pg7>LxpC2+^
z#wAY2jVXy*>^d~)7i~Mh+)e7h4^8t-Q=WD9Fk><!H7Pqbf-5Ne`}PrBj(*Ne1$xec
zl3;Jd;0KV?R-R@YXOzTbMw>&<q{9<Qn$g09Logviddl5W{QS*x&yaf|{ods0afNU}
zzCSII{ZZdV?phnq_>_uit-mSmzCis><<Mx?-i&O!kg~xH#s<9yD;wMQOJd!(Y>TdB
zC71mUv*F0nvP*;;v0$yK<;I$V*kUjVD`<i5z)puAUt}Yo>jK%bW$8!&ms2oRp%-D5
zd_!^vD5X+gFjT?3VwFS!{fyPU8)uOAAM9f(U>rQZd`^7dHCv|KafCaokGuv;l+7`b
zk5%Ss%8V1+_<$D<E0izo*I)8m2@hc-))xV!(O+0mYCr0G*pQxqf*m^b<5|T&-;H`L
zavAEV$^CyMd$|BX%V8H**R2gB(ge%4Rpyrh=zHVkGmOuE<#88?KlkG_O&w={ywukO
znIOhW+$j|YZHdi5R@Pl?{Ljs0=aBX-BGaQrE*m>P`gYS)8g}+L6BQ#v+Lfc$g2Vfm
zF7dC!dt<20E2E=cp;SnPgujzM8m!6DoEYo`<3Y+~v}DbV>fwZD-cQosxqNO<>%K^x
zXgPCR{~ouNULT%PFN|%ZaZ%hLYw17$q2I1uRA%3&N}>q<#|oVbikuPk?|Q$%Mz&us
zHEh}fs=3Xw2_*r5mO)^!d~G_*;hLpj1wF``Vs?gau!u~0MUo}_js#DH46dtDUVF4t
zr%`Ee$Jlxp8x|0$3C)VyHs-W3L<t9_aJHO!E2jaGE*<gMf*zDHyV<qUk0Je0Mr&|i
zT$5NX_FXeXp}vC@@iUWYN-Jl->%BftaIG#D_d?f}x4eZ8dfY;f<Y;SCKpcv>PHd^N
z|6kY}%KZ+erA9>K1z&w`c^*xPj2M}LE?Qab-7bb)t7YF)%-o(+4}rWJ2s-hVLuw*3
zhV@CRnntoVyV?>U(=Q!DnfAzHiY|Xx2nb5vaYLp`bS0nq_2n@+#&K=ae&u+H$=cl`
zvH$q}C^w5R?=#b83%r+AJ%irhCUnzCf|W5q7hg6~2&md6xsErLBZ|6fmb9|=r;Kx6
zy-&wQG`&DH<!T;bR`^AN_z0T#yrTYg!}%M7S4wF{QC!b$3EvO592H}JWJiC5mb6H<
zD)v>o;R}mFL4wMw3XWapypi<gpKR~4#8ZLEEbsytK{IT9<z8AY;ik5|Sps5c*|N$r
zf9xhMqg%Sga-#P~4OxoT-jdB@bGXU26b?wL_B0k;Pk>4d_NsP?Fg8ooBX*B^Bax@6
z|LR5<fNu&Mrcv;XBGRt_r|){z!#;SBF)>N^2SQe?tJU4@2+t!X!V^*Zr|K%$IgGB5
zjxdgJvYo%y!wJrl{@{D5h~2I5UB5W-s=V_0ZJIE1U1?_sWVi=fN$-7_5r}D{vu}Iq
z{7`QCq*zCkYix;*-BZP=m#9ayO=Cr+J4P_3U5_eXG5C%fnGKowwONUQ0H5u8p=@Au
zk%3vgjq@*$gPyy()Z1bjq`r&Og|l3!FEatOOy=g&q4ykC&DVND_M73g?^%IOwk8fy
zpt_Y_AXXGT8r#yCgH>BbrB<^Neh;>i11@)J!+){lT|FZQr18B5$W<{B$<>)yvtOrn
zJT$-9tz|Wj`LcOGovBdX+kx}_Y{?}eD``9N@;t6aaaLC*s8w`jluzn=UYI<V6F0u9
zc6~pN;6#;oxUq?bsC{x*n;=!nM%d%R`1W0mPxYFGrsb(m%|0_9HC6g+YMLurpiTO!
z4)u1L<XB7#7)naEVx18lFjucL7Z3N@tKI20OXpV~$G&<;gDPmiBIuN96t2aC7?9WX
zz1PYCgv(s}|FUJjw)9y)K5--*WXgi<RLra`<Sr?rz`$UeetcimfWP@XKIUe057ue`
zE4h=7s#;a>9|+p=gMt3%6lx>z6=!f~J0*&s^KlJBfzS)&8HEGHcu+l1NPVXWwwyI%
zji!qC+V>{WkX}Q9^^gx^Q}4z%K=1<isj^HH(}xFcOtx|}mbe|E&JVrd7Pc8_7GfWJ
zG;!Rl)$%59xv^w$2DdT|U+}d{W_}d*{rV~7diWE2(>?tzd<j*4D|ElFj&aK<Zi=7<
z1ymLIqbHYVF@M7Fs=@nt<9`!8ud4-PkgtRo#)^DEdzO@h$eBJKERzn_+s2~PYvaPy
zpS`rE(^|lp4uNo}($Bmh?Q$P|N8Y_>R|D5A#L8l1CTLch)GG*{A$!WwdH*nP#SJ!e
zc;C*%yBFHt-b79^Z72Iwpg@lcM!~6xZ7363WQb`@_fqT;G5f?gNqrgL`M22~2ZR$-
zjQEfVZz5j@%6#n$sgI)$yVS`-T@m0rD(+SYdC|~MgwSa}RGV|^yCLlR>4zEdUwG;E
zV_CJG`kFo&X~E(R=iU_3^tkAX3PaznTG?u7PfC?<7gB-!(NtpxI{V@SlvUp;%Iv`s
zD-_G?0bw(Ah|2G3F!FoSEP`FQodw=%8^N$b6;c?cG40-wTA?1lAV-Ih%jZmwv~aOI
z(2ftfy&9B;rJkAv>@Jxs<u*yLJhkt?-x$nLa+ycheJV=ngQ|?C4rG;Eq!4G1q6}RK
z8r0}%h%8Mn8flzyJO6d~)MrtcR)vCM$5wRKqdiGm%=Y+YEpm{c4R;w*mTQ&L@!}t4
z+Vej@uDK`OvvOFFxA6lB=4#4XBQPz`T-L7fJA;tA%`ta$JG^Vv^bCUe?0<WT$?6w`
zlS)|J89aG#Y*Eb8grgESYf=0S<!5@-JPyB8RP4CZw3S>uUIrp_n|cXa8>dQ=8o51r
zKgQm4)LkGB19?5KKZKXRJZ<>w0AFqR+lo{#aPvkQ2dV(JO*$ChNg(?yU@tXSJlGoF
z`jCq(of$>d-Pa9))H)2cfO-%>nm(v^RD8x2K?c^#Ugn$p;Kr9Fj3Z?9kYR;_E=7~*
zyqZ#JiFksdK@;v_YDOZRmIZps_DtT2rQJlC9zCw>V41?F@+zGsz6CT?I4H`LR0qnq
zUO<yChKA@VuxV~!LWXI;lLssn^#QFk9bBr@4-iM5rd8u44*Nv>8$^Fe-`DD<Z_U34
z1EpfMhjEHvc(%{x8un7dC&R5pRMk$9saBofObcl9OOqxEirRj|{_&~AeoBYT%N}1g
zUQZIUsk}$>IG>8}!~)dt2QYX)LPw)Tiz^0L%7rXMn!fPCC5reQKjWn_+N#4Hizp6F
z0<~*$g}!7a8k^Nnkvv$<b3xc=zO#N-$Y}N$K6alAVLj|`rxdVVBVsr9nI0;)@FhOF
zeDGBHXu=?Njyo*PVxc@a<&<9V-}g2zt$s9=UfC)7XiFO7-d@EBS@8BRgziZB2$qU=
zi>t>iRYz`~Cc;WXhK3vdFc)6`qz*2qd2LCl9V4+CmHS@L0dB|eE8SMr*aWL8Is9w#
z>?L;!=>zBdetiO5+1V5VRG?+;xy8c*g*4`$3VKRI9di4jjEHsd4ij-oD_Q>&e7lD&
zN_jPfa+VlSIaX0#HwU10CmHjCJEIFDKm;?p;vNGY>0Dou{@$!{>mqx=mgp!D!*QS8
z1^kVh1Tw=#q&~?0mfpjZnQ{J9Tlei^Ls`Rh-av~%`KW_pxqXvWe6To(ilh*i=7{7&
zjHlOKT}jtmakZJ}aW|E(#!bEJ%c{Y|)g-)HRdP0*6P$L-v}9ku6zVzO`N5i}QD-6I
z!<JqmuR-yu8tkN;u@mmB_b#tHhgp+{Dq(Zz`q?qCH_p@9<u4&Nya*E6cfm)eDB})~
zygd$`%N;;e3^hZ@>IUkS*XJ?+GK6<i!0#Ml8~~4UgpgIhK5x?2HUz_Mf;LI(=jHld
z(w9ZM-p1T$KO(nAT!nk=bpxzr$k0Q#pusIYSzyRV5hJYw*Y-m?8P`+GvB0@>GEt#P
zTDYsC$!sphmt@i=8Y836d?#`cp>tir=fM9<p4ST`^tP0(0O!QEUQnG>ka7lz_29wk
zXsQ<Pn}m%a*3+BbX(iXPiUh|9kc__|JvwF22kVc*!55PpUA=G;2@0RrAi)%007y<l
zK$}Qkm$1eZ<@MjowC?O1?w{))<Z04jF)UhyrF_BHTJpZ0NO-Z}wNjOxv~VjN8%^#u
z;3ts|v<R~50CSRKuVJm}geAcS>qBdYDe@vy!kFgoqAcl57%RG+c!oCodLMrK!U6-}
zw)N)q?2d<dwhA8IjXs|=ktqymZp@)swIgV~E*6p0N8PT@cwUZA=yPz>)Qj{I04L4k
zYZ<Hag3y7DG%ht`5I-EILSH|R=Z}tj+3HkC_I)1)vi+~zDeJfCt!|Asw9U{YC-iSq
z@k27*NLNcYEcfWYu<)4*0~BVa=N{7`=e21Jba;@dHB$`|)@|>!Xv{SswrWlg67yv$
z`%j8-sH~hAg*=RF;|hdn!3pS4j+)yY_u-h9w&CJFooexnx5IEo*J}MPYLW+93jTGH
z+8VNN@$L}w=>u_JZ^4zVA#D0>`AE#O>Ed8RyJiSz6fg(Z;6o$UoQ6jog2IS_lThEO
zyD}h1_5IDitP^Y~8ca)!n8)aR5~UnyVHGz1iUZpOy7O)7>}^9jdBDw-aty+^g&LzB
zw(G>4J8<);`_M0=UqXJdqEk0B%k3?5BxN<{yomW=6GK5+M_!+0r7%?5$q&$lR^r&$
zR#tP#SvorB=T}mJ=}1TBK||8ZCUBwDkxkoc3#Pqu@B2f<CU0_bv_6dk11K!FH}5MF
zOChc0V;{)}1{R2@5YY5FpuE`^6nm?5mpcZudGcSjEEm9N(J;5o7|3MB=+?2U={5@K
zXz*83T85doCXoEPWZ~0joj<LD0@I#b+bQXiDw=g|rTOn^WBe$v2Qb0=KX@S6F`uss
z^uSLwp5eM>EYphXkymYM!X;8q{kQU)bl+^a_ff#DxmB5>s=hh$aGFs#7Si)fIQ!0y
zKd@@K?Ji~iV%3Of!-OE7IOqB3R!`8nCJdw!2dI{h;$j~A;!?5h8QOm4n>r*?QNPIL
z_q+Hj47<-2$R~2<GD6ki=9-2<Qt-xmCZMas>bUu~<YUR{+;2l}szs5@A3r{Ij?yAn
zwg+oEM-flba}(}Ch@MDwHcGO(7{G>(hy<r7?AL5lnc?S8x}t?k_cU$5Z%C8X7`+?-
z&p)nry0SLTLoZ=YPY(~TOQeDsb8_@1e#Sd!@+O6@rHZ*4eJML7KzbF?zfeuzi3fnu
z|JtDM`&khjbyOX;FlWHv5%2gP-g|A4N*$ak0qsljIO<!wkS9i#aU|ZmsjD7{!~gL8
z=+sF$55OefL)_hg@=*^c9MnI}>-gNMXaPa$mR#X3`@RP&eO%Vy&8ovkyq9Vsh455T
zv-0$Xg=!6AaIoCRyQFo3cz3qv3vJ0V>WA@pBpB9C3nzxZgBvOa(B~S(CAGn`<NQE^
zm)__>Y1XY{LOhhCBFo*DaZwl$VCqnVBmwXLx^?Puee^Adig_;N``$`im>FMo1TVG*
z|6fzr9oEG1wL!%zpkg5cDj-EEfdr5yMUbLE1f>}SLX;xXivcO2TrN!#=}BmUgoJJ=
zVt`Oox-=z-G=VEUARvT*wB(0){hn|C+}W9V_dRE3-r3pboG<eQSIY#}XW?&8ZZ2#{
zO773cn`*zNM2f=LG4Cv2=lwuO80isgPt71wc?x72+#xpeYEkWb+Vb&Ww+3xan+BTz
z4N)VzBVLlp!01g>N<h{`ly-r9TK~|ltiR3F>lzDo>tbFNFXFjkv$uRCF^q{D5DLm<
zzPIs0+mQpl<;#4+Y#7R=<^ppXNhdjBiR`!8b@~1~wi2hZs-J?ieR<~uSwa3QCI#v=
z!u*~#kGs2VrI@;^)M0sY<SCIM(8a%wA2n)uR`E{xfq{;JGzRZpe!}3!v0=S%araoq
zA_87}*7Dfd@`#IeShPEuP4lE?61%4An1h#^QvcX%O-;?$SA*+<>*iVij0G}!yRX08
zemXG-FA%-ixZq_A#pap4OZ%?K_vA<H&qi*q-Qjs*Tc#~NC-kQf_Tvk)<pDQqwd+>r
z)LU~yayc;!4~+kO&+27S7}IXq4AnJX+E|g>flVyUVvR?{#RRG0c@^N~Z5T@<)aOLH
z^*rng$}y@f#E^Xod+2;)B;$#64M(HH;KtlwncbV^Hxmu}hiNaF3IyO{4avxmPXG@P
zgYj;8l_n7(7o(BvZ2MQLs3`n-vb`0T@jD9%>-Rk4<p>*kdw!3gU%!f{_eeByur%6A
zmL@G!?N@)5z`qb|ylS==0&497N4|Ib4B>74fw#+9x}YmqI=teoo?9F>eW$$J@s{25
zMc=gftP+Gy8M;0;=-K*>Fa@W}4=M|9$`=DbeDGt^=P5|%xp>@{GP~N1%`*34oDRj^
zC3^2>WJx>`C_OlU1)YiFE|G?HZ|MH@3n<%Z9az3k!dS+_+0F%147Ij&PkkV?I#G_{
zO>p1ps(`R)Rryj=%U9Y=3{puy%%?*;lSp@+RcCEwJ}U>EYzzFEF><66pOKp=ey3?S
z1XREoB8zyq;@BBHGkxC#i{UdZLtdA-a%W?9t+VS|gX8d)eVho+6ny417+ie$vP_!Q
z{hQ>^@}|~yv1uwfgQQ*<SgUelZZULc4ON%GM=ORAAXUW}L=k?yE6lfUEB`1*Zy@aJ
z0rqgiSu;B0;!>QOI&GB$xj8&o)bsLG%%-aM)~gFKaQ|Lfb0e~i$HjoIPljB6zNnC3
zQH4N$!ge>9RiWKnCyDVWgl?Yz18*4*Q7rH}K2u*|W06%2loVNc3vVxbWTm0P&1LUR
z(yzA_vB%%8zpZHj;Bk`1mzbSu_<$J;9ckEnC7V!szhPsBFLV!k4^w%Ly8r*!Vht9w
zr!iv$l`lAbK(Tm|+S<S|D!@Z{d?aMUFGyujKJUTQGIMU&IHmooN&Hoe4Asf=vVL7D
z)yY3icFqVS=#`C&+1LqN-84TA6R}fhy&G9r;{`8RM|_&&a?p!=Xo8L3#1>cPex18h
zob_b)NpOUIr*>7xIUEFvNu6`e!kuGJxjeV|eYAm*^v3N=eYuN+vk^_9-9Md5<og_s
zIH5-|=$W$2Mieoeo`Mu7reB74@cjb2fJsg&YfA;~2nnz|VB@PL9w1pNp2MG%H$Q4!
zo$RDDdjv+*;4CE7h|dM)_2_%Ih1$4p{J#U}h{sr=b^9!Ug!R34C%NuXt*VJ!^PO8=
zAChNOLHV6+F`I#+vz-f3p}XAxQ$>v0yhrf93uKz=*R&)%TI!l6I(%$wD7()5moo5C
zb7#Gm-2f->ZTkU)51<c$gQ1j28$ehI1`E2`7YjaMH8z=l%V;d8Gs};6f^wefq`r+!
zl-o<{7kyakgCa34`VCW7yw_^)Sbek|pssRKi5o6qW1H4HeNu1=5}2&7tj8HJ8#gno
z>EIHr{*@01t|W4K6wX7-msNFYB1Qxis7}w30F>tuQ@LK3kIaVAY~=j=9Ic%$iRyL(
z+yyaPOfty4@e)kredj%c=A<&iWK!@su-wG@|4|8Wn=UzxhutT3xyx^@a_q{=pF}UR
zi!+A87<j(SI~f|m-+OQ$pGH^1KCTD8rxN(@x!U5s6rChT6!N90ywa}=@BvS0NY=~e
zh8^qiR*y(tEjBMxo7=`jS$(()dB4W;c_Vtk_N+X-QE%F9;{9PSPb=1^V)*sd`9=9C
zMRzJOVX4v(n&AUe!h-0*_)nXJuu;n-&>5%WuOWOKeZA#X7%=Sk!vi|QEh2-Ar_ArE
z6OJ0l-wys9Ed1h$uBlu*KM(c-BZ1HOvHDbsO68YXy5CBr5%kA!%dNUf0)Mnxt!vZ;
zm2aKTQl9;_D4u7R?H|y^T_@2e?s~$Et3f)tTgs%c6c4vIk>hGZZ%IWnDOSa?e4AbE
zT~nZf!(kfbQy<_O^{&l%i*SetjN;A^Xt#yjcrH^EC2U3G(x^R3C7PM`8{n)#uv8{X
zc;HCV0ou>N^qIwf#9=`ty|sf2FemeOyE84q@Etzmeu?pxeWTyPLoL)=<>)n_uAK+<
zH7Ohp4F*72aJ!2k&7K620F2Ck9bdIip3a(r_W|?2`DHxuH355#f<;#LgZ0VF+^kOx
zNJ68F)|;v)G>5jj^Abow)46_c`=t#RPM7!+yNpkjoreq#Ml3mg!)F-chPowT(Q+p&
zC=ol@MV8-=ESZNetKIm^Pw$pFN-M#oR*T69jO#<7pw`T@C?xKboXy;r@R2ujol%)t
zxw^gCh)o9kL&EY;=xn*m!ja*2KLLe&t*dl~wM|>|oF(<LjwRCy91;({^TuRF5WxHS
z2YjYbUdzN=Gt7i^Y2a?$Ik7MqzO5!AQHTN()L36$cFqZKA82-ivc4o{tL~3{#0^Sk
zDyU*Y3PPzjhL4~N2eb(&^n}8}SSn93YFv1OKz^a)qy@DY&!R$|gXg}wN#!42sez5v
z(<d)Lt9y5UG&^|JtW$Rj<K1G1^ys#ReWW!4Dvtj2LqnFEXxl(&lK=e275%J7>u>sx
zxnf@g=KGPOIs)-CM4pAdLrUoiwHo}aRzvh+&|F7dFl)2u1aR<GJxgT-gIRaPfJQZ7
z`2nkG9+~dL)2L1ih+gdok=;+;Zb#&HIG~cv?DT?Cx=W03DQ|NCjjJf}d*rg}VqFl<
zT60}dFRWn+>`abt*xv2(6&ooNX=}S;2qp9CLaNXgR&+8g?ia@V;JG59)mxi2>t~$%
ztyGGCaRxj?S-4#y4--c{{Yi`2Rs}CpgKzPyge?~pFoTyma~i)F@9ORhfy^o03<{RC
zpFS+)Nh0Z!Ce!g7kB9KRo*N?~zX)6a)FA~1c8$(X1l=V9cX?gBQb2UZX!Dub$6blB
zC)DZv`&l+rS5}?~Hk0hpd(Zqw&{ROYf0SKW&l2&9#XYGQ+ndFDfq~21L*3&KnUS62
zCAyO{&sUO7Ur+Lv;R8+su{f<=W?ywq*hZL;Rp3Men4hYj!B2D9&yefaYKQ~DeVQyS
zY1=^A2DzUnkBY*A5J?L#N=u4EU3Dsz$gEkN;?OD&d93u;8pl4~@Sw}0@P7bHozgI(
z^-!DNF#N)cOIZEvXX6@+=QR(tq45|0j`hn59x3mGYi+KqHIITM+S9E(;~t71YHN0A
zsJYm$?|a5k0H16$X-3NFcWTeStM0Z2|He1!-Q*JEDVKzc^-TI+6O<s<PGSC}w!z0(
z@79Q!Tu~7vW3FaZKtl?`5Xk5dg%Ljd8%_W+tG(jqxWW?J>kv|x;0G@GA(E1LR_Kn6
zG{tv+<%Hj-DO4xpVahboSLWMoStSfmjd4?i!BAp?iAVbS3NGZD)}v{a7f8=X;;!~w
z4}Ow$cGe4_KcoR1DeomezDSeVS^h9s5ER|-ymND>=Slprp=&0qaoZYpI<qa}owpMF
z8eY$A!CCIDu&2!8yOM?68I%29dJWyN&}IW@vUgOxE+uaBG4h-%D|z%kgoci(Wv)hp
z?|QwKPFce*<k<5kzkKOp(Po!ml$ITzb~Vqwe}%$CV_l}oyfZRq)Asg;cy;NGfj*<)
z{L^R#wJrEVYV#x9W4l4l&z*GfdI*BNSGj&$c&asGy*W>_eSCDAiqe9G4o*4;QR;Ty
zKo;@MvQm+TRLmpaX+;dTF)a?FL=dN`%>nz?n<SNK3VE1BP{`X6JKxbBqSp9%wyi_^
zRBAUQ=YB@kZyzs4MCFJsH7fvQV?CbxV<ZK`s3}i>MhVzuJ^{EKdWU9cH<3TG`ug?O
zyHhjg!VvT5vvV^eHsA-{0;cu%R=JY2A>P6ozQV;yD)Axsr~D_rT@$eu@9e2FtK5cN
z`vUR5TenT&X`lqsL%OgsBcBEsjTh8HtU|k+`16sL5K70HZ^h*Z{k-ISM7qAB&P1ON
z2zFrVOM+6Oa16@d2}3fImCm;37lQ?LdE2}f?yZk=0n4L&^@NrRvc4*;Qc#Vb6^p0N
z6x7nHyrG)Nfk|baD^?pzCMe`qA5dV+SGBLcIi;|N?s}yy=vgJ*9?j<<Ix;Xe80<Lu
z1$ysm-SkF#{Ke~E%l4CyMwqTDpxjc$U}XIHe=fm%ngG|2(zqI?(pa-C8!xKIcv(J!
zr}X}t!o@qFGlZ`3P_{mj5d`OBGFoedi4S9RucR>v5UWzO@m@lhYQ<h^3iv~Hb?5-l
zQn(hf!hF-jZ-1rC{^1E>c)i~Jp8}d5iY38_ODA>e^*VE%-N&NF^VF+7Y$_0a5it|N
zn=c|nV6HClE0gk5)$0y^GO#wUR@X227o&*4^<*PMT&S!^5yV`Y#zj(6r4qw`Z#2Xm
zlFEW`B=3?cs-;G@r!6j&^wP#73A9%~{`PIo-jXUOBrdvQmNRdyzfpxn?>T#uW!(5?
z=A!RV`M=_}$9OqTUYQ0;lHa3Rl|k1Pl5Vc$Plmc&3GvKz8_e|j8vJpRh5WI!x3i7j
zv5$V21~}1^+FY}w*aU#y{)fkw0wQlvUg;~zn36hFxM^|5;Ho`&`_)&8za@qE@hysM
z$drtP$<~5QAE?gC^b=>i(A6Wre-JcL92v3ur7$wxR4x;}wWDT|Gpx3nIoIPUbajop
z7K^QId0Z0~t*sSMvHei4uM#4$PX5F<Aq+I<ZmqBz4|DgLBRO>~+g3e#I@c?tdxR1E
znkbObSEdNyb`yc+F1=;i)FGnep4+^_7|9YwjBt%wrj)eZ7lVHziZHF!$cgQ~!SjC<
zuLXz7(Q`}1rHFJJ!#+?6BK^hVu3C4&?ozs6$ZpYPS9=h%v($^`+e~m*9Z?W>h*9>u
z9`tx>mZ7(fA`3{6Z`aPcIV&azUz1QlbjaAd2YpJKZ8f_vo}I4Q)Mtw`|GRE%XLIR3
z+5Q9mFaP@0plbbH3!f9_0f-w5xB?ZR;E0}#3p@+asws&<ya4a3#nW3i1m5PPNkR*W
z=NQG2h%E);o_4yTL?&X(`3d5F91fJfSMji%z<n@$2x9VUN^&&mkZ4y!=xP9&x7lop
zvQCfAdx#}tsjD`RU`20Rkb2_XmD!owV0u0eYx>rsxtKRS{u%UwGo~dH!eZRlwij|&
zEiP5>{L9?id}U2WY<-b^ORFUCDNHTZWi+NrGwh@*EjekTUCR4WUd{+Tk=%ef0YF?3
zY%R=~R;mB4_zG;>A3)^(t<S<8OV^YNsesV9s)R+!m;{^_sq2dIcHk0~_|$i4ZeY_R
z+b^othLLU#pxsT_NC<n}K`0Ab{XA<bX43fX#fI~inC9r<*SE-42&<YFE92*t4V5qn
zoph8zMLH#c6tERkzhw>`iV@?LpGquKN5&LHZD~ElZ2;Xi;F~_6<x+ZSHitBd_LTgX
zzhBR&NAO%m$Cj9IraYp%W>r_^dc(JJe{>bmxV}SFo6M*=mL~P^ek*YKdUr4ZZ0jw>
zBOJS8YvY8nk3@Z^0?x^SmF|-6=~B`tF>M&oz;wLujo*?aIUG5_<I^XX&&hXn8t8n?
zy4T>oW7|{rv%+ynA_|TlQcH^=&~9b6XMfZnLAr+BH3B4l+eX3IJ=H3mL){mvObNg&
zJa2nug{^qt63I>dwM4VA#oJthGoDH5<{QxZBwoAXrR>P_P_vf)bBs7tIOYsE0&(!E
zf$h>h1eN^QdHO?nT|`<^#^VxMJa|f6rUZF|a?lW)0sC}`&!-u^gn*)@tSc7VR`O8-
zCs28*NGA+PVOj7`e~?S?=O~b7d=}eqDR$nP$}@FDrilOOe{L4eerU_}c52OhtExM3
z!T6hz`Z(%%y6^LlM9}YE6!yzyDmzjd!N#4r#%YsyIN8Bx&VZuwyEy*BT4uaQdo}DY
zWWHYz)p&@|g$_+$oLyX3_8F8LgaAH<|Mx<PXY?A+3wtJkDr2EV*WS99C8Wq09PCcn
z*aZGYDhVkURdS};#a%ac34HS6IJ?-N!kNSS2$PSX?ZvyldWn$)|3C7J$#EWOlMY<a
zYx-)F`>XkHP?Co`pi{{fk-F{5Lfzue@BU$opDnSht%;XZ!xeqMMY71_Jm=xx6$or=
z0|BXV0VniYl1Kiua`3dcOD|~jXPKGDWVrWl!`v0wC)C)i1!7vntM|LQcI~M?)?q7N
V?zYKCmJ7e`8S0zpk#rtJ{0}|tFD3v0


From e2fd946fe12e7defab5f36973628d3bc3ff388c2 Mon Sep 17 00:00:00 2001
From: Michael Han <107991372+shimmyshimmer@users.noreply.github.com>
Date: Thu, 2 Apr 2026 03:00:10 -0700
Subject: [PATCH 007/216] Add files via upload

---
 images/unsloth new logo.png | Bin 60119 -> 60610 bytes
 1 file changed, 0 insertions(+), 0 deletions(-)

diff --git a/images/unsloth new logo.png b/images/unsloth new logo.png
index 20dac04f388d8eec89066d81362359994e855786..adaafee48dc80f38582cad3df0b56b0b3e7d7302 100644
GIT binary patch
literal 60610
zcmYgY2RPO3`&ZdBdykNjU6RNV;n+K}SF%^O;~cVOg`{jw_KIwdJrZT_l@W(yI~?2p
z>;2XH{;!M6b-6m8=leYOeSgM%e~Z*q2a^yo5Mp6rktiw3X=7pCI>N%bA&ZX-{KjZj
zfd%+M;Hqfgj)f)2as7S6Cs*1F3v1IuNlr%BCv7tw&)?|xSa)WzfUi*nvyBD?8>dC6
zRC|Us6(QAkeru{@Lp?i{KEbN0r`i45dVLwzNeCPKJDe|f?7%fmoDOk2-lpry`jwo0
zHQh=6m#$VZ7*i7^{bA_5`%m}JuHwXD^!x%|GrCyISfWO%t3vIIyW+JI?)L%>)Ac51
zG>xu2A>KXfi>^MzYzFrCisuPJWn<L6cgPbyzfp(Vk%Y-|baitcZxq|Jy4U$(4)M+<
z2i3TO_Z}n;)iv}Dxr@c|Zqvcrga;f&1=LuUJupA_PA<Oss@laC3LDUPp*=2SFI6Y>
zCqSK6-(6J8dmhf94t`{#a1k(z+<HrvINmvt>gQu|^@t4S(`Yx3^cZ2{Cwb@|q~~ew
z^AB$a5_)`^=6P8%z0FwQ!y-~9Heu~|IM_VL^v9l{j{H)LO)({!;+vvK^uomRv8JKv
zAunBVCa)XFg0h09#{3Bb0`r^cdbcJ=9i2SBWqO9u<5LhpxR?dJQdwH%uttdeF32QG
zo!|B?UnVKQ>0JibA9_B2C@nr$))D9$qcGh1QdBl5#GD5^hEryE@L}PkMz954*!7pL
z3=fqokE#kCZaT~5Q!LXK{WNJ*)%0&_u0DsFpGVJvvj_FCP!9Ks%h$b*Uf8MQw@$xu
zu*#Yoc%#M5zez6>HTMeFl`b${ltIm1XD%Xl4iVc&-p!e=N0c~(9iwowJ&?2Fh0Xug
z`sgpiAV5bbn?KgJhN+)3Eq{EQ@c7X1;vnh6+bc(u?F@6PELK!XX$&d~(W1I%VIi$*
zS22U^`vu1GDzDL%_`#(UGqe0yE`LO(d`t#DL<OOx|K}m-!j#@j>^x(;&9=8wTvpJ~
zH$P%`baEkuwR+>YZfbhcWE1+5kcjBaLN;p7ize_^6~m7V63_9_%cs9?hjftmrDt&L
z5_bq+f9bTp%R9h3-^IE4e(6d^nCd*IZ|VVVE4WJ3i185>g!R*TdwVl6u&tsSOa#Af
zC$>!*>p!=0J$r(Os`ewv9eeQi@62Nd$tW*#54BXcJDz|Yz(ywGOqVpRkRRg;kP1v(
zuK@pMUm#6mjO@1LE%Cg(obQDb>!kZQ=U~P|T&?8)ZKgxGf0yevCmh7;4+|*B%h_mB
z2V-8DwJr=fIvHzjLao9`iOwi;yPV;JXDqtXBU>Klw<t(O>7(0C5_M32&!!hUce}BP
z#RuA~W>|Ven7M$tZ+-a4h%vZp{?V&nU%m**7;e@ydKhl4a|OG}C50NA-T(6t9mFgF
zY+tJTrkwDxgN1JTG)+w(TUTHrds%3X?mms06Ju=sgp07yckpyl6Q)Yb9dj*x@^|O;
zU94_1E+X{Nn3_o^1uP+8!4|$Fn8+s!i*jvXL51LX;i746-;h33^O&P>Th;_Er1pPt
zLKjBQb$kx1<26~>NnLnhlRjfvTmSgES^AeR`t(2v@WOJwk0j}N{z(A%Bbu8gmVZCy
z=JgoTFlRbogT-=l*Wzn+eAj1XbBpqfBabybecgvC#NEq4)gYby5c?<k=^oY$x%<u)
z2!ATyTK8<;+4D_~&-SjsIMj8tH-B?PIbMCzgV#@3R_cBtQShqz6UTIUoLNV{q{$<T
zWY{ctt}0)vE2<~YM2#&h%&>tcL)<X7D=jP@bMVAdO!&-u1QPl8IBRIkipQFM@B3tU
zHbC^hz+sEPXV?rk#R{jUc|6e`bMP^Hp)z5E*7c*mpLtKzrg>YvHi{SnA_(tD7Bga`
zrG{jN5%F&lu_13T@hJ_nGTNgUs~EQGEO#jXp1x``eyA+|Er%x1junuSU09~pa0C##
zpcbSA^PC2<YI=l6wtUF7Tk7v8I)XLDLriOl5G<G=S8;#f^!8FMcXaB3&yZD3I^iiN
zcsM+Bf_~M5Gn;K*{5`)gUT19Srs*||^x*H!lNumCv??%#v*s3=R@04=&J!rKzFrDn
z)^GfO*HKmdB{_?5g*>?;;y6dWa#((Qy-I642oG502bMF7mX5T1`N-D)Uqp-Vjrr|G
z*JGW4)&MK-aw{d^#B(RlbT__yVPyvbV$UyqdtvOucjfDFnEP*U1m|zC8X%A~AkW3H
zynzg!>zdEk4h_9v0}?p4w6tXY$4g;Tex>0E%N?@62cgQlwC~4u!~5js<-&xnZv9(?
zbS^&8*^Ob#+L&avIV+J*v87eIvK(5EB>xm23%XZ4bqE&k+L1j|y2ZZ#6)@>sY@#!M
zU>%R7)wgy4(9mG@ta~<C?oj-@5Pja7q_IS)*TD4mjjpcwOzk(JP#WMA8tpjO7gN4(
z7;ptwOg8@g=fpbO9v<Qirbnm#n1?_VP(f@)f%JzV59_=SPM~H$gl|+aI8}Iq4?Rxi
zko%CKElk$PS{NTPoC6u$u@fpu7069@J7DfkOWgn^F*_Y8yJEUPM^$HSlK;yv^1NwA
z#RTk904x*ChmjWJ8|LFk^YLM8se9Fax~S--*VR7*4H5L7xRe{jG`hBjp)jStX-ALU
zrWw(Hc+)g<#|5j*sYVR<EkeUSG!)7JWW!@z#O0&$jfQ78bg3Z$qqvcqG;z0#Oh~J6
zjFH4XUw%vlZ@(imxgjKR;VgcuSrnC5_g-cY%%bLN=6~~Py+DJ&V#6_CwyBF8Vm{h1
zhFq=MvmlwcV1jPQv8bKLoti;4DHz85cPG6<rY%CYql|(MTec9pm(XVp6&P^LmoFLs
zPI;wkxzyF<RT}1x*}g?c7G?GAwAtq~qh?LM`1)C3j&kO9!UT{ClLgNPQDE`g{dF;y
z*w7`ZLTsrRjd^7cRn<}M(q{%(-z%uL9x>jrt11}Z7_u$uRCK4dm$Ha{sB58|M4i}Q
zE$oJp{Sy~j)akL?fjw&%l!2+@ewp}05F7z7BP$@Y%*|S_zR2X)ntznluA4)(zqXOo
z6Q8M=T7H_EPb+376vPl2^b0G}$4#WM3=6t)jD-VT;DP!}?p5qTR+by$CgXnIj1bXY
zY4{wUB|^{B!&>b}sh~HObH}mur4Gp}I+RGSW{my87b&JcC!_GN?uqlgWLS1~&E<;a
zPE?Ay1uK9XlHl@@1*g)o($5Uri@0N*E?yXO*n}Rxdldos!a9bHQ_Dn>_g6#)_vbtK
zer|;t42akMOAJeXMO9}Zy>EnN-^TET!g8|W7Cw9&E^iVz#ja*>bI5Irjwd0Zb1CQ+
z?Dp%&e{LR9m=Fs36UJ2J0c^X*08=yiG2RW1LI~LhhDu=><zLt*4jr%EQ;{DJsvg}7
z53)U!sHmx>H1l$7XVA~Ty`Q1U#x=`vb(iqe@$Q8=3aKNHTwbQcLEM*Roq}2VgXSh(
z8Dxa|&$OS!ew>I<z%Zty3>s_2Ont!J+<0J>R!{Ted9^_k``@a%lE9KoiEy6K!4kBu
z5j!%cdr%Sz1puuP09oRwUVq-eN}-}WZ$Bp&ybqGoBIOPb^u)K2q{p{=%i`YpDA62{
zpRm*mTm`|*J<*Xv#~@Gi+lo)v&`qzdkf+9j`SctKwC(ue?ZV^Ftvo!Xj>_mFrwY2M
zbOgF}$E^46{RM<Xq9z!tZ1&(nzrsioT{p3j#e!vRy3^qLPH}0GSTclB@@>D+mR*N_
zdg`9q8ckitn}~PGSe3Z%g?g;e$};6a5~uudG%pI7gyPxQUS2X|$?w0dXs9oultp#I
zp1_LGuoAKKhu$JBVTwuDiZR(xfVZ3XZ+Ke$iHcXYXtc!mk>BhivWd}i3*JpB&`#Oq
zQ*%)6eDVWbp(_)8BGCF|pwqmmE!B{ZwMT|C_fjROMeDR}e26>$2Sy(pcO!q!h8GF1
zjcmAF=^0fX)O&8OenGbx?zd?wMKNNZTIoammxs_3N+C$^k3;ZVLxH=uxa$#~^u?uO
zM#1i3$_yh)fr%iC-#u@}U(kQFne;%ic4=n&L&Qhg_!HYFAba~vrrd#Q5!0p8`eJxo
z{v4mPhH_&Z-Fk2Plk_YNfm06sWuxTUu^xX7k;U7}-v}dIG;}N!tUQ^I?#H%*@)Ve+
zMR6qLy&}%17-L_>{E31@vhz2AD&-^TvzE2cX@FUqe-w5Mxefzx@lZ3s>mDeVk7UoX
zQ8qntDi$txe^BF>CUT=Zo;{)Y9dPX!;*wa9780WaW@XM)V@AA&T2Dh8Xnf;))cHvi
z8%>Lpu)DBBUJgF960yKc4}ElCiCnhl&K^(uC?Fg6<7Va`oC;GHXCkfk^sd4VZleUS
zW5JwG*f<0lI)gfCEqK&FWLr?w^kg^u8T;7fR?XOPSuwSgjEm|Q@bLC;-LuOL<kuaX
zL2xXY?jS~~Q6pWMwNV;%<O$_ituKIUG)o`)L0N@rT)wY;9(sCzL&URFkxBP4jr9GP
z?)RrYDIoCfpWESj6e4@aL!wa%)N(ZmpfZHV{k!30?WEJ^cb6}W450B-w)awHZOnKg
zG^Qt(AI>OSyrq>DPp}ud8|f$i-n6V-|Lw7C!<<bKT;pyE$BaJy%dT5@^Owk~V>7ZQ
zgy-aBPfyQa#AgXT-v%O0eM)gTQd3-?Uih59n29z4M?Y^W<Nv#pR|O#)Yex$WPs)gx
z%jRAB>bIL6uX9sj3w8P2x=<W`pLeTXcP}jHc($2bsD{5w+lEn8_9q|5euC^Tn=-h*
zj`S%A_0rZXx!OlP5mc_4q|xNW=868k#ho|Tm%9B`gSVSN|KhghoMzlJuxhtPj8%n0
zM_q`4f|gj@zqf*`R^Fr5?Wchy-ZC|fd`7SRp*@5BkORdSyNh$~LU$OJpf7)^dE3mx
zvHid(**sEqut#<8rA&EAroI+4^c4T%_6CxAIbK$rFVwfP&@?BIRhb#O)cMqs0T$MO
zqbrKOu`J%aa{QM#N;l0SCD4Gx%Cek+$o~&=&@(uf;{m))Ut#=FTHyW9o7Yh}UJ@@V
z^l03`JDyW1j$)1S47aJylec+iibkb^7O|E%yZvBnXhiyZ1XN@L`8yz756p`A5H8{8
z+h|~o^Ad=L=;8dIJc?<pilP*s9}V9hew)b0y*T<k@}r2Q$EG07@t?GK;M2bNg|$__
zl-a2qVC}Wp(L8+Dg-LD5UUX;l`ItJt=h~UP)J#a$)nqe>CSJn;Y}N@wL*=2TRcE?|
z=y>3uJNx*!<W@)I`zdo4mE+S52Fa>oO<{si)ba1$K#w5MYZXo%HqeduCxR!{3F&B*
zg4i$OW$ZM&^jNL2!?&0aX7r!9d?gI%C=awp8Bv&QI6!_UvQ&ZH>>QkJ58GB<<&Shj
zLX#OvO}R7X_HUUwnUFu~Qi=G+S58~YdSO02(X}n4$qda29B7UORY;jGVTYWbzRdK>
zM9=D26LDHUittoy)&D4xZNVw{MYWsW>Xi)R)<)|P<nJ{E)p>hV;|8~-r<TT)^4$6^
zdz}%Dsl?qqs3Q9sa3lp`_?~j#{U^1ayfr6jd+#BSC4_R0lw?iiJ<)qJ!}-H}J^2hX
z+$rfxN1X-GC$d6Ud=o?53hNJuPCXLwd*f*12D!_)rOQ|N0|5wuW&~*G1uZ@9$#kK!
zh{0B?#=#R+@9z4*I~lODGRg2?hPa-pa3}=;-12hP(P{H(q|SbObC>gse)$MO%j~(4
zNs-hBtw;5*3N<46Q^f2C!B=ndBf3Yd8}gb2Dq>-~TBrTwr&Q!?RCX)-Err<76_KR`
z@$_STLZsYKX>=<l8B|WSO8`l$i8ZY+b=*KEnuU7+`J>M@^|^e&A$iyD9)`;gt`Ib3
z7?l}2$&x(k#Hunth8vt^mD%-|=MzUCZU1Be___x`cOJYK&K2CO2Bby!E4Z6<jILi!
z^UO|LfS6EwpFp47XM-sh5RkR9<3sro+!*q#NLeE{C{_^2>v5Xh#i&&R4x;JjAb@gr
zZ>|H#pG3zQr{Z&wVAco<0<EVjevFP7UC>AuCP=2|T7EAg*%emN`qy!ILrxezI5aYc
zB0DZzy;OP6bC++^XtU;`m_Nx|FGZj+3;U$Mnfz+H4huDBGE{ZeDdCr^m62u%z>QFe
zh-p~W+2<qA;6nHgrO3cLuT1_^N*w>U-+W$EnccaJ50*I9wMwkbR35aU%-vy}h+E(t
z(_Q*JB)O!YG*V%5@$F+BZI4FetMY%Gg$`lWY(T-iPr1sOSd;o;c|Xw|)~l7!4oZNP
zUu9M|)Kre<B*q5Pl9gT2l+Wnpxiut@?2goyGe%@-+$F*@1TQ=)N|V?ZT*!nnL;Yti
zCWd-ylk~oDTjQ>j$AV5y@_!0PNDsJ@;y@`3Nf^4LyQz~#+~z0KtvoSRWYZL4@h|^}
z`X6OS){)!0$4xy+mWD81LP+7v`U}11S|X*R!#iu9&fQSdnpJQk5yPI@1`HkPiH=8w
zmcWyDgkCWt?vsFwWHNW@7Td>$@>8+sU74Zi*p+VqVzwK_t_hwDut%0+3KOnd|3-nJ
zKuhrqD?Zb>qf=TqMb78zI##0t_1t2O_Uq{7xf8T_n`P>lIb*uAPUEZrh%$np=7<X<
z<z4p;e5iy>CV#6H#wOx5H+Y?r0~JYr@P2t%9UFMw<{ea$>26PLI$E96nt+(jKE_7O
za^s`hF#^UA%g|!Q@2BBv|Ce*3$<n1E7tP5(6h<O>?kFo<C$UbI;Z{B%OX6P(?P!}L
ziit2KXABFTwMT_;4Wve<F4)XbCezgjt*kthQu)JZp<+E}{!{dxa0I}7-=BhJH2U%B
z6qym)=d@+9u)x7vcA|r#<4f$Q*t#M+T3tG|CBfMFKVFdPkq1Ts<n23a15{IeI5rQ#
zbv0EojPa5I4#LrdV2q^2$|yKo%s)!7h5O#@GH?nxARC~9sGO9+HF%L$vX4O$W_Flf
z@-rAWIj$x+bK`L?WBx4dZ-?aLhTjH*`A;k9Dt10@xkAU(KTZiN%+lav%gsqe{jFoY
zs%&&2h<tZX9;#^o&L(dr!TqlH@chwOjJN@O=v>Ens?1Av@yl)WfQtzQ?SAD5G|0eo
z8%Q>S@11+R*men7adXK)oKv;*lAoZ1Oq<jR;xd7SQxw0w_Y#UD>9Qf#&fDg0Azj}<
zQr6`7AWp|S$f}?wEJOCf_Wg9Tno^WKar1Y$ws~I<6rTL$U-SWNSmAvqioh}WYCbNa
zIRgAlnFP+)Ez*kRBXFuOzL33%hi1PT%RX?{LJS7EC4LoW$AT7XZV>wAVrg(EB%$&l
zX)8R<Pq^NT?!9Zc1bH$;YsgQrwK5cdtzl_7b3cFIY#Rh3_z}Ti{MeGFEai(0zku27
z&wsZt!rIJ|e6V3!i@JQ}04KXAU)`QDymRBcv8M9W!@WxV#TNgja~HEi^V8yqPr%w4
zZ_PGg=-wc`Y?H)J$PsWyX9*lvZ>u0(p2U?vRQm0S>|cobCOm0-XKW<or`OHG;Y%SC
z>WL11I9^=X>SpEDS$f!LDQE3ja+l@5`-z@G$s~VWQ^JyVs1YmDep;+Id5dnD=r}+a
ziGIj)<FeL?`m*_m?+(SXENF^fNNW&af9@Pe%AH<*q`MY4lLrZ(NOS>01y2v0DLBny
zpb@))fx_51*mL;;?6cK>Q(7v0zp3bRq5aMMIngOq*~tTd-7N6Evg*mJSb9~~__tnO
zx-~!+T@Npd04K-0;P!LEGn&S&s)nKzBURjGx%XHa&ZZ$fj|7-p&l-$Jhq&kOnQ0hv
zBq^%J4g&xWO09zY{;0a@>mgu}mR|%XyDLAlK1O-<U6QLMZRqS`@ty5MBmqvB7a}{W
zt}Kc3W0eHmLkU~zoi|NpoVy(H{#~Et_W5SPgT)+MAdadLT&eD8=HLL`?I5Uv8u_EM
z3)r3Q`t6KAym^?CFmyG%J-c3_^d~`5w#0QnZ&?hkOBa+ER=?aKFnh9XbVYJGQ~AU`
zHhJZ>v|R1k8$!Rs@gYpiVBbvS=pg0?v7+;{i8cDDsD$HXGMrxoIKS|Zf8qNql7(hX
z*qRT~14fAX!;f9Z62bIbF*HB3SOr+{YJAgnas8LnDv&V+<Mg4EaRX#>TudRYafBBu
zCOiZ(<sqgo(rFIi;HC1h#$S^n<T_F`m_GqOB5m?BLl9_M_<XH)W6=%S2!Vn%CTyyc
z#?A^%%<F<+v-PVg{?ipA6ovZZ18a}BV1OwYIkRpAUdTL-1DHd8E#J5bZFZa*uAIDk
z(X?l1R<fCjuhDBo(+Hmy+H}~Xn~S$M(68>y_oO<A6z38A$h#-Xz$g4&H^vY_0FO>$
zcgRt22~r<!r6sg1?s@YN!Te&?v4<jD?U<wF^)z6D+|0%z#(ki`zz2_GB|nl0Ay?(G
z(X^X}1`PhR<D^U7*bARFs55VOCRj3MYDr5O_=9Xvw(=qX?|-yuy2b(GL9YON3?K5k
zwY<3Kq`ueuJbZd)I{x4yQ$HXqH}=T3=$aot#g-ZZ$8sa2pxW&RlX0L%iPKSX*a?oJ
zQ(k_7l?Hjb#_gkba>~&Bff&ISaKx#?Gn{S%dtrD3Ib=MhExL$H^~-2Z?~6CtUuMb4
zPZm6!+RnC6!VAqel5mL>=<ao5-`paLgfti@1c7iE97qJ6QzB8_Thx@8*OE{Ey^zEP
zl8m0UlV?e}`>zVH9}=9M+4B_(X*b@wl+dSjHaesit8e^T1Rv0!zA)cFCIvl7tvY)t
zllgreG;*M&&cw;M+{N-Jl(OaA&+~XTrz21X-WB^i=A*WjQ0KV?Nun^KNapj@8qsv3
zq+USef)JE!UK4AXuM|I|K+Rb6HLEtrJ1K9@<{8~Pd;a68g7n?BiZh@)Qd#$JI<DKH
zD$V9K9WH2f1Mgjfr&mqaOITH24x#<4(2>|weONSUizJPjpx#-7F8S&4@e06LzQJC1
zK8ZO_o@wLl^6IP<1gR^3HO(ZrIn@SYqVPmrYHEJ&)WOccZjf9FHywVf>)W^xD%e8%
zwJ24%K8K|zE-jWaCBLI&3_f`N>&qh8Oe)e7u|=o<_~si;#V0o3+3~xG&y|^8_m@}x
z0>f*2=HN%cUA8aX{szOv;0Wqzc@NYJV18~?^w04jfZLT~nlGvF<CSkQMAG$QKW>nE
zx1cN*1T%Ua0(y)9$*Vye{Vne`RhCJeoP1WzkDKSNjX4MpMiTf8G44k7tte<%DRw%r
zWy{(;t)7n{$0e%qO}7<(Ay*_taN0}lIXNJ~(n8l#T-P#L<p-&lgLt=5?pl4WH~hRb
z;;y9FD@5rQKpc6t3e~i`|MlO~C4NV<f!ux#7`^F$?sgC{zLuGKcB70L#pPvE6!+pr
z7b^FyJWi$PH;{M>&CabC$g|IL7&EYtnfgStpvW`}H5%R2PQK>b0G@JIhrLXwq4M>{
z^+%HWgm56ViseJ~;&^HJQdFsPxSHEuyzxWO!&RSuL89AtokQB9ai8JT_(F!vqk&YZ
z$+RsDo%ShrVkW&fzVm|5!Qx%jQq9v*UbB1LZf17f<na9E?L35*kyh{0-vW65sh?rO
z=RU02U<>}uD;=}v?hQlcA%+G@>zHB?`KnXbHn?21Q3Ie5>Wu@sV1TBAYnba<d^6%N
zcPlRU-=h7@cmlTr_4%nHYYE#-W47+0DgfV9wPut+|8(nKDSLIZvH0N(?e1oS%b@WX
z9pqu(J$sEoTAfgOaP$nzmF+a+*ec<%fAb5z|CdpU!c^!fiLEQU8nKM<MXpV!^sTdx
z(Rl+!0!f;D4iHJVk?(O9`lF-8)vVA{3BN9=9SXVZ-~b;GeH@x;8Z6jiH2T4aghsWn
zphHfD*{hU3)WkkTq%eo{1N&%eeozRqJBG2sx{t<jzKOoHv=oU%29B(U^eoE=(}heq
zDmXa_<f(oNO81_pe_-=-)4Op2Q)M|pw(~3|TbAyW{B@*E{hd!S;$E|{b=-#TJBNPR
zBkM0_|1vc~KfiRqPyY#cf}l6-@1%|Phyg!74=8Dcnj)#QQpY138X;ONV<BtL2D#@G
z>tjKWWu@OqoT>`^e!uEd=@8h2@X93e%2b(EzgUcOOnJ*(%n<S}Q)`Te{j7`YHvX@(
zotB=ZuBZyzA$ER~1}Z-NGOG3rZ2uZ2*zscB(X>OwPV-qF^MhhP&OE(Gkub~(0b9!}
z>!euLyPHQ#)dGFyZ)m(@XIK^ft6(3B-V)v?Y$0f+>|nP9AaI>i>9nL`TmlTwCf(f#
z#@=H65D-Wc_rNRlbC87J$-;`R-^q=H(B>Q}kkgq&ydN2_gfDrRB-PX69^Tmv?=t#G
z=thslD>X_|77?X~<h~SJE-e>gpbNB)|Iw7qi&|%spEVahC3k)eJ|k0uT+tJmwkHCD
z;iWwEb2(Hz8mpZrGbngP&xcyvVI{T>I{apdivr_^09ms97r7L>r_h!S<a$Hx)HL1G
zPvFe}GpEAY;2fBJCnfQ(qp?gPEwgACdJ13|l0*3@aOM-S$6Fa}aefWWKi1@G7lQAl
zlY<{Qg>$5E5s8>Z)3_J1QP!?yKu9&D4yooLJ5uq^nPh@>ziyVisF01A&~-F>d=vlF
zYC>t&9{zx5zf)=In@Obs(rt-g>;9-R?m28}zTz9={PfGer9r&Sd{+=_1s8Vf`}0Xj
zV?@e5m99{lM0)$*+n#1b9ZwztU<_VAW8^DpwepsW-s6$g;1GlO{UX5;M!XKO74r63
zrgAe7(Db`#xw>$<v%#yK>B4imFfwwhcr0cAvbs$DDdWlHH+EC*AqV=$%E;<GuYf)M
z#?koO8y>%$=PaWGS!h*fnQ?<<R7?Kbm7F5V&wqXakS{JkGiwb#dyNRMN=i$bha5#?
zLX>@dI&S(;RTTaLyaYJ<qPASc#Eg}-FnX|u=_cp>o{@+n57caW`w&!p%Qo$?msqS_
z3L$j#wjiq}0%Bxj<alRKHOxyenFxVE$nt2f-iRqKD0r%`ug^c${Eo?EP^O@`SjCqW
z;S7j`o7m?S%g;_uRkRLd_F1xwf~`4spKZ<6skh*hpoqoYmiV<CG%j(`$)SI>U}s@r
z$vEU<-s;Bt%rAJmU@ZM-xlbCc3FfOxnY7Ou0866T4l^}AWQDbbS}CaDNB*XPqRXYJ
zIvHx2{gi^O!0H{@99)ZatLxwHI@dB5>y=Uv<yMs&e8puHbG#k(;}vC)Y3V!3OCq5r
zw*EzJney#SZ*i&slM9@2FQ*$-xT7ieEQXOe?9!yxa_@fSVmROHUTP+U1e!bG!Qdt#
zc>>gvHfX94+3b%p*@h|(e=PF6LEDE6A4&4eYQ3sT=?oTjb*DbARRbmv!(mwYMzKTG
zQu_kyPIYM)pqi=qx0=hdSpAuy{P$T?Z{zoHalaq1{@RGw?0WlbEUstD%4RVuOz9gA
zmBH0Uae3$dR_%i2VtZ&#mvN!n>3aV1UJonn#q?QlG>+L+0~MLmF->MwIanC^th$gP
zNg`OWL!Z3`FWVNo<#YDjgq<+#+~<kuS6VMe29Axb^@s3Kj=w~cG9RAL_v`S?%cVw4
zQ0_54ByIo_Z<^l{_eyZI`@5;+<(JJ{@gGd>J<-a0LWX6LJW}#+m$jDFTkuk6=O#;?
zsbCN7G?=D^>7QBmCfsdLy1c+F^fdV%49?HTYInW7zX%k7&QMxY@q$BI&)TISH6pi(
zje$4jSe_pYKUO!qzbQ?%1u@%D#AQBWktWd)WsR^`!Fj?Gr`T)rFHLX@%4VBxkZ2|e
z?{Ej;H#^~4{>^BlY$(hY4%sTmSc~5RoxfUdv_0&ena<@f4wL|dopgkRUwwI#Vj~$|
z#rLT<<hT!hvOHpJ$!O9azl&RVyxq97dx|PwPSQ7R3&J7NB;9O4mF0wGc3x;249*2P
zBqQRks`+VMwB4<tqHy*!&g~9-O+6fCI*Iur0r5sV@VWJOV>il3$o-LJmVXbEn~u;E
zXi^l8K4;FocWocJ`K7~pBT^ca3*IpT3~Z;-h{zbs%KC6c;uNX;ni(qYS3fsFlsXX#
zPt98@;yp8O{60#$b5Q}i@Y8a??0+y2A$s=OYGbre{PK9YYXN#Rg(tcBrQ==gNrn?i
znc=<|O-<>WUTwVMwuIZcWXVdS85xl^y>C<byxRxzx9?1oYyXhKT-PdIm;Jp9^*NWH
z;>*!#F3A?kIG4sr(xl1mpGLMMa;By~){2>(l=-&6^Kz7vaxSG)X|1UayA&ID%40L|
zdAW$>1fb}sr?D$S-qTv}2TLvwNIUG{-tFeczk<njC#~WLWHf?Gjqe1v%(%2jc|mLE
zygDj+SfZ9=B}u=&`oS<NMrM~5y~7~Xeb^`6nW+zbE0AuasXvDwr5C69LX?G{k`ueu
za|w4u*|D}6_e|rX;Ko+3YMjpg-=;%sule-+x|deK8D%{8z5%+-p@gvlVVVzWX%dnU
zTm1dOns-NM{3{<E{ay6BJbWKLj(DoU#2VWO=VRZw1N2);K7a%-?IPt?Icnu+w>hfO
zhP=~JASRlC##=1>uN^zyaB%%<fuxD?XlYnp2HqlVZNp>be|7TRU^L(I^8BQ3y8$H(
zIda@YVCL(7e=rJ3y)v(Zof`9hgpdoS!Pq3Eq$Za(LiC0^=hrIPn8IwLNG6Vx&E3J0
z(F*X}M^yDfA>Md@TAS~0`lZL7&SVZ+*8bq-wa}14NixlqJ3LIuUK3lRE6NY>Sq{2E
z<ad&BiGz5}_f~{ZhBfLNh8+F9A~MMIV}-oVmV71}xpG5ZLH}?XpuNkh>v+r3jcF0Q
z)Q7kH)L5Wst<9<tb@8<{_uXFZOyXT?0gdMKeHCNKR`sM?5A)tg<|UO2;&42$wgPH>
zYdh1Ee}3|eKl;ADl{-dn&vZla7o0dNt4TU=%;?c_*w3AiEbf+@8!iL?GRmb<UES=d
z>AFYC{FYaIiSw@Fy12d3XA=Rz&Mo-^TTsEf1b(@4N(In3`pLsjG!bL1U{T*CIjmuP
zKVN?N0}d6rR)^08k6GmRc$J0h<le*F2E_WySR!e?t_TFvzRU!glHG6TxK-~yAVn*E
zzS|Kv&Mb&NI$B!K&+Lu|vqT1yN+=q$@PHpWh3Dd#g(Y%r#vZXSI<jyRr(Di#Bw>s8
zt{g3m9gX>zUp!jkdj0LcViI|Lekl!<<zzV3H74o-!&GGzANw^Kg&a+SSp6Iyt2V|F
zo`Re)dsS!egFve7i1I9rSx^iSZ~q?W9pt*S^7=e-JB;sFPt3^b+tD-nu^59E(oy$I
zBs{L0fICf7j1&|4tAk9MY6x<F8nP70jawyn!t$i$jMv9DZbY;EQFd+*Zn^^B{KV^b
z&Dnz7mm9`m8js2;+B0I^l>78wRs2!P{jS|v6RC$1bDlu4Qye2Y1Gpu&gQp3#ZufmK
zW5!%1CwCC&K8Qj|<MzoA($8`lS_!n*1)~BCR{F^OD0ZvfArs-~HlzaYA*`^2D`d&N
zF*7=fYa5h>Jg`nS2}2xB6S*{>tZJGKTlCSI`M>%4Q-TPxS%GH?JNbR16mofNvUefV
z(%G!0Gc;Xq(6n><TJZ;2_~Mr-w<sI>=>|zk?Z4R>7OCk3K(?*un(U8%`U^wr+fZ-<
z=kwq6kY_TpwU><YgwSok@9XAa+FY*9ErWwIB~E>Cu`a-2KI|!sG21FbVs=ptwH5%~
zXsTJ$sL=7$885$4T{;6h+j1L_!Vq8JSYsco{^*G0&fuwJr4cesJM@FAJwNtn$y`SB
zo@LEVk$MlZs{9bLd&7!vDTBhDt#`_<vlpflzx%g+N=1jp6$Na}FI~=wqf8Yoz(kMO
zGwUbK>ynOp;uH;C`4s5FjceV;SH?Fpk*nOOPk5?HSPQ9-HJ$h39Ak0QiaS1>&te6<
z-=(HK$>k&|^$tZVFB5El5V`I5az|N@<f+yB8+_t<e<AF?s%i|JqpL)^K{+PVM5Dd*
zYpi@NoP>Ga^kY>+gb2HGvTHeGT;2Gu>W2&N%6|>$&umnr%=eUrRW#vWT&iogUW%l|
z4cU#2%{yCo^UYV5u2beFD<ykn0yfBVX6TcoW{n~kzj7vBwtj331|AnDuiHs^XODGk
zMCFTb`sB)X_fE|WS-||_U`giL*8vb;4w|aeW`6(>J-iJ>?|3b0j30PPpW9@9T_ifK
zm>MEYRN$UJNJwhE!)CkFW_HVAaD0s5>lFVq*Xz;0Vv_wGdBc>s#C02q)}Hc+qm)E)
zRJW+ODkEo>%=!3LgdJwD0`PX9ecC|AH^iD|2*2aN<Ya)jHsI>V?*^CLTB!+ty)o!n
zs@1{tZTKd>YdBtta@#1(&nYhE<1W9O$c&mOr_BqYHFkZoEBY;Lq^G}=7SA|Pg!*uD
z^ec&93{8VJHO2$~?EJ<2*jq^>TF%ca?*DC0*N}hEu6j*I?K%!6XSWV#iIVn4SLG*H
zx3;*qKxxmXp?MOg(HBT7-qY_@5bG3}+^EX;_t2rA>avXkp<XSFaXbD`Vi_L`6tA7n
zc`sNTZBEr)U7opL`rV&80F;bHcQpMq=v6Yb*v!%KKS3scG?gSYYN(arwo`fZT1W}s
zQ+!X+m2%z$a|`W<h?wl-{p(^sm>sgOvMs3IUf%O1DIz~g;G6>{kV-VkRhXT89A-A!
z4&h{Ld4O##Hbo~?5;04(cBc65!BQCO6ifA2heL_3{mIScicC*`y@3;;2$DnBlLbvu
zW@r=XJ<ylHmbwLMXA{PJisz$o+=aSJPTeMl>*-ImX)rI3Ca2%D9P7h7hDf5i+7<l%
zxQuW%CQ^;9ZBKvVhxmYf>vTn!=o_n$c<|FVxSN)6X(6>aF0To2G0LgX*E0fCR{;Dx
zzQY#f_xnO32A5H4&J`DWJh8d#syi}Tl5y}{-z4<lr(~RbB4ItCP(vZjGh5r<B?j;g
zytDLhgP6pMTn9dn1cj;&zQ@ScBfKx)nsKhCkAmwW`_ca@YhugNexUu50TjI<I`(IJ
z#J~i|-t!NeOw&YNjWrKnY&)_7joLb(el!PxdLK(G4fZ`e+&euz7N5GvpxEu|ltIJZ
zj=H#dMk#JZHKqr#TF1jQWY>?E<9f`lPA8h0KBzip!7z=Rl|*_aFNSs|{SX8TEIuNn
zOx3q)jvOgh88~+%wG%tCS!h{~%DUPGP;#YGxyxTw{-&s0foP=R*#k}0Ya5w)WzY8t
zmwIT?eYr)T56>8!`Ocq}Xr2%CyTbq|B((zh>wIeHmBRuJs6YQ4c*!X$AMcGTe)-SG
zB<>pum^0&<nHdqApG@bNma88#sNR*>-+jqZ+L4%rfR=@d_d!>S&1;qed4-eH^-8Lz
z??2a(g%3@=>S-}{_(g_oF3RxO{vR(*RRORv*x5hT8>>sKf3}ns2vZ`kZhkas=uKZF
z#yb2xfzzaTVx2LI-PF#6G<D%2z>oosRLK-aa0tVOHUMV)Df@7FK{5A=(@^T*qI-7V
zS{q!FH4gB03?k*;VkH`|=t#WNxx5q36co-16EEIl-Gq`J(LH81Vset5)iH#1+E#er
zLoR#@y`|<&T0$T*e~l6fV9vgAJhrA*llULs-BOHh&HlNrQA=L!Ef+_l$E^~6Q&oa6
zDG@bkku6X?D<aKHXc%Qhe$tHUvtrBt1+IO{{)2wk0%*LE{RGkA#eAIHIY^TZXpR_l
znXcj$D(6TH_$htmKMNqmMeG9y&m3w~9eL&&amO6$4E}AEFjqPlgp7sIMOMQ_h;jYT
zK$Yt2Qbe1NEik}?7yNRt2pCtkEh*E~UpTd?;g=D95(q<o19*Ci$)}uh*J}OIp&lvp
zZ(CR;GLFz!DIU!lOLtMaxz#E#!x~MQTDyaJ{4@0t+1ch-GA2?mi<wg1ZDZ)+SlxPF
zYUSy>=H@ke;SDugH}B+^hOkN2)?{YJz2OgcE8dI8PX9fn5W=sa-@j_e-}z^@EO%lV
z_EG)kSbNYmO+RWnjdSDd2(IC5mV96NTGm)kBlwRwq}Dm&WRicw4qtm@mhrn)Ee=Pz
zyH0(UI)%oXK@X{q<{>VC<JQ!Yp=xow%z72EaIx)u5j#`rMFpC^^2a_m^*EawR7@!7
z5V|WX<@|(VIzr$`?B5+l11}MT=be&YFqZxnw+x6kEuOn^Z|ReOG;u*J=tI~7w-ZPZ
z`PI&B!utY={nSI@Da$iCIp_-PexK3e7pw9Hkvh-tXQxnD6pcXp4Ll~coHSKmH&kiu
z5+J&Eyk|42;chb}R=-@UBKxV7zfI(XoX+9}zZkOaAgcMjbWe>qp>D0<uOEQ*#(cLE
z=q{}WU0116ZD0<|VURJ4X4TV~d~0h-!83k_GGYB}3tX<)qj3j`rwwK`dsJbAcw+Z-
zXk2aESr--K<DX0k?8afDtgHEKyD$Fq{%RPbbL+z8ctCLEg6b3=Sap(QrokvQm`i@^
zHz`r>(}Ia$T3MrtdT46I^oSfac<YZ6!5uwP&!1R_DWyp6v^Qsu+m{p<RXmk%8m#1G
zPfZ-&P6`6m@Aw;k&M&IJ*7x^xq^8!Rb4s-7kZ8y+($l?_GU;v}o#!nATXlIbAJ^YW
z`b*j2EF0dlQSRz(bt$hL-KOOch?@6l`xiG#&>qXV<J86ue+_e`7tyPRSlsNH*XWjL
z%raqO7^C-a@f46i7N6i!GEFv{U?K3K4*l|7@bm*9Y}L^1h1|E^hxxwkA!ZBt9qTpQ
zyllZv(jrYNfn=+ysaeIyztru(p*0Tlp7-Tmri)UD^FDfPMCD!Oo8GQi{(fWYSvu(x
zxLuc@5z)PS_v*&~vVfcys;}-Bl|RkkI06Qq=`{&BQ_Q9t`2Bo=F~>z>PjogvKmm9+
zkX>~JS!BI-7)}&?f>%N*@it(-^ZjIyY+P9L-mK`c+*|7Mg;CR2yY-9on0LLUrteWL
zu38$JQxBV%Bheou6A;bSaoHh@7=JHRC7;2OMCJM~SCVsuNCntd%p57OgkAua8-e{I
z{OIt-yh`<<mRmZ|*DZo4_k`i<f?3t+5}D?}LB1@Jr}%9J9G9o_g8e1VowvTmu4C#=
z+`6cJE)FLS&I#|^ckW=8TW^tt^!@%WzP(xT)%AC5IK0E+nfOM683pf20%y*V#)i|2
zdQVEpl|Q)~eG{QRLHsNETk@Ck(h?SVv2**t&I5>(1HAh1)7en|wn-*c_COIZA>&-7
zXBO`z1YTDf1#{}Hh@CFb@4_D;JAi#`Vd|w0M|J7|y}W<2FLE#ZyHG6X6bb9mQtDAB
zNgNt>piQS)xjZW3rs^J#^=cYH$QUO%Y#IFZqdS<>_3(mCHAqj!9VdDXaV8~ZY=GgT
zx$K<j(;bK*d}}!&6rzo)r}9KsjGh4Zvu6uimGCnd<RfV<STTS-p%CEnn{nH!x+VOz
zfp_^JXY_1DLmDA0#16Ly_Q4TfKg5dZf5NZ}StP~j#7kkv<g&#)Bc;tqG^`LbbSS4f
z;@|#Xf_z+5HL+H!D|8NdwAKz{B9)e{KG!73tZ|7Lt;-kgR-risY>%Z$+YNthW$2Pq
zq?VsGGnDgox#(aN_y|X<O@rMb^eeMFSoL=4_xmz4-?)^GNL4dI%vRlK7Z&l^?EaXm
zi=6=a9wu|<IHLRkd2k#F8;6u8Q4_^&O!gO~!<Fa6hsiyme>>*%KW#-B+~fLuZN`_s
z(vJe2-<P6v-g)1?GQ3!<vVvj8VZ+;J<3r*1$A^00Xh8JUxsq{>4e^a^|DL)9c*(=N
z)^8T5cb^K`xnA{OqBa0AV(d6wMew`-T8FM>2C(#f5a?J8XEoiN<0Vjk{QEr&wRHtr
z#*Z#>mTm006WfwdP@6wsRE}33W;irIMv&Osx&C*6j&f}c;5o!A%9D(TVa)ebUloK7
zY%vZpqPfxVofcq-z$;UeBU5{@r{^0=o{N+M2QjQ>HXp0l7Iz75xA%0x6HKqd%hMgT
z@mCi@pea}&n<DQyfNn;o=Y&Z(4iU|-ACDF~mAKP6ov&@gw+M&3Z#oulFM%j=V`u`V
z7d4M(P4JaO>qA}yKpt}71Ru{<lSFMAWlf+$zx=0|1G{*lQ3L(299z_VfU%2LFQFa8
z)pDpU5G|;}MRh0+B&$q`pbBE)9pACapp9vzeC=TtJIP(nO>}Zh!|8Y^E@xD^pgt$G
z?7p$N$zR;y2Oo+6m^E`a)9Z*AM^pJO1nv93`zl`cCgg0I@yaq&Tml!oY$JVUFfFnc
zof^|P4hxlMW)yCyyn!DcVg%Ft__!!)@-n1-*8}sb`Ylbrn;KOQ>EjiLYQrI14S@y8
zKkQb=me}uB5nqvgC4Fj4*F`h%ieL5f56T5^aOMPT_ar*44J><rku_(PSm1oQ@-};J
z1;3^7t?;D&gtfPS@)=e6KA9rtc%}(a=e;__o{|#z;HUU4mN>jz{OWtD!WK8VG&1<v
z4;r^>QeT`FX9jFlPiXS<+y@#w$}tB6(yTo!5Po|Hhmlxc*Kv_a)I}p;)U=h1MH)hU
zwUq@Z9{WnSU$V`UO}cuWiS;u)B9M}jBH;4-FOA+#nWq7-GsOUJNjNhUeUa{ZDJ)BZ
z-G0Y^n8hwGiT`^d4%AZ4l=qT4r&%KeB*r6BEp2ejSJroU#JR!o@Q1~amt)dMY&{3p
zQpoDo15iG9roZ;*{Ng0H<!%c!a;TuJENsl4y01VB9`cS=(yOE{vyALLfC6}-VL5G?
zuv6X*6siJfmE<50%()>Gs4lzw^bbR-E&~9z`=32??Wx3Al-X0!_`Ofi$V|6P)XNBO
zpBx{0<NYwTA~>KVVm$FneyE%j(Vx+xS9KJ`IQcj0@>-gYE(jqAb?-pa{BVXBGc%Fp
zjbhxna~SGX=<v1<;pvJe@M!*)Pg!_BYo&NVhgn{~9HrciVmNQv^s((;TlSFHi!wXE
zX&=PC70&EMceD@?uvytBIO*KD6~gxOF1o4ZY+I5lB~JSC#e4kQ2WtbgOBttK$rOO$
zMRR$!aJ2+fSry6E59@e~&7NlPz8#s;c{+zT`X%rEd7~g5%~<ov60RRc-ZT{1*UfFe
zkoZgdiS^MEQXYl^hgB^lQ`3qW+q~<T5??trsF>D8(5`h)UdWLrfzvh>bVUaP2Y18J
zmB$*}V4(#rLlkd~KpbV+vDR-{+QP9(!hDis^^{<sTloOQN^y8mpH547509;_?)7iw
zbClUtyMax5Xu8)l?^;gAi#pdkn_iNHz0y}F8USFjRG^l7;(D(A8gdSWtkL6>fdIjc
z!4Zq$jorRTNbA-{2&Y|V1*_c~{TIppyzl~zCVj#%IiqhlPN^YTw3!0ic7Mr4fw0lH
zNGZ5$FVCit<TQ+tI5bZN2N*4BnycZ;hMJUY8NY?0mE&2}Usq_1#hH3F#z}7W&8Vlz
zMM!yhv<JRx-<v+o2&CnRPCEYG`ItU)AlVESnV$(`=nBcnPBwl`0O+eb(Ni!qoVCsI
zX8<O+O#y|+h`1}GVJZG8l9Kx^=8n#f28L{kAV!xak>!u6c?0s#p5K!2M^wxd$jPZ`
z`JE)x(GW%|m0g_}h43}2=UZ9)-8-AN|JUro=4{@PkI^=#3`z$?rOMhC$r97%2NF7j
zn<1meFRpxZh4uCN^EbdkwLjosWk;Lj$@@YQ<ZMw_FTP}q6GVEfRmX~@Njlp^KeQ1O
zI{z*7;Dh1{bJ19y={{!GJx;M!h;t=E`gXzC4A9S@VBfy=Cc{eNADxBK^)^$U^+T2B
zq{>S48%ha&tdo55lT^z}D0S_(q`h<&4|Lz;4I2$kV?}B4Gdia4p8pRXV-bxU@G~F7
zv3bD#y}v<1nME;QT^b3VhVve}-MDskhS2IFTpwO#&c*=%pdnARK75#DcftNp`d;$O
zwuQxmtTy8qmD9*^ozKm54_7<U>mz&Jj4r<l^E2+%WhEQ5z@Xz+y!S7zy9d)g6p&li
z3s6kV*jT>R*;Z}OqXmDC%MXO_-v*35*SBUmC{}{B8IFXgXYvUaAH0-%q$wkmTSq~P
zU2zZD$z4;ugvc9Ul;1HJ{lj)Tn1g=`mjT|7U2A)Db3@J7@7gyW0>e<k<qj&{fHV;`
zr8$9y`Wg@Rz%PKoQ*Ou18gc{+xl;Tg&ktW67e>Jhac--s4EA^W#%?%o+pd{sR}}4X
z0X$Y#B-`uNg~k0(ayv)T=Zj=Z5}yi5UdldQooKoA)LZ6KO9{C~v<Qyar-e^V@fpSB
z8st8=+Y2uz08j9HBZ%uB@&VCqn~1IqD+iN8E2?X)Z2ZM~=>7jGWB?O7sNro-PY+r>
zF!>L=500CzY)XWmlK9Wn8;JY&Ev0BSjN3RulLz<9u7F|ITJGAusQeY30M`a~@UBy{
zX@3D@R3G<F(q8>6WCvkuXD5^Giyi{I0j=(5BxwB+_4orIDMjuCsoSm1)HngAJQEZ>
zWDu9%i*0`wOey3&>xQTN#-SXccg*p5*~ZFdDa?8&`w?BB!801a_}&89i6NUtOF(sm
z#;}w)PK79u{K(nPfJ{H=Z++lrI~>=qIZ<X3_0VD5e5~mCWR)EyFb9AaPr&i4RHZPT
zS75Z=qdkNmCx+Me-JtA5?_h#bi0QLafqViRznJy?@lvA---B0sema=snwlCA<<uz4
zhlYW7ZhVNJ9N<{0a>+S`QtRHy>dMMW+kvQF5;R2K#2fhloZ01x8joBKj-`|^MW%(t
zGXJ)TpC%abC4z1vU%aC|d7wlDRj1E;F}SDP^T5}=w}lVwz1$tc?#lD5h2CR!Sj6P{
z^SjLxD+|;Os@7l7Cg0XPZiO2T#+A4r>EB(y(_<gw*Ks9t*MG7`u#HZGX0lfuv)^)2
zU7m>J&zTN_sVw<HXMRU|Qf+Omh<KPgyL&Dz5j4ULZTcs}&jBe3zLT98I$;VS`D!x|
zW_aBslNSY11APxt7XRHgkR67ZnR)4u5g#Z^r&-1qS4Xuik$OI(^nv$rM*9OU4_LLz
zo|=6q{ODNNVP|Etync9KQOIxyB+0qGF8*PkA9-fnw|XO3ri4Ud&w2*eapH;CYOIDz
zeRXpqFn+PWRBiEfYH)BcP&?Sr5eD>SAJ44<Hbl4v=!7$WroSf=U+t%sO+t2xJq!qw
ze_CamUkBbUvoRt+1063;b<oL6CZAt#CV{Y!^BQq|LVs}L-_E)U^w6lw_xgldqUY!9
z9^2KttApRA1c(MBsrWA}Ed?>FZelI5M<2CC^IQjAs3Mu}p*S?MfW-c~FrasTR3PGm
zDvB9s!Y?%@#Ku2%UoF^g1hFpst`7=QIJvyc!J6-)HH&_o-PmZ<l*tN3lV|wv_m)h{
zjbLLj_k4=A9_B{R$lm=5M0xGUm6w$riyIg4k-0N_;8ofOGZ7be6c94hI%sClBfZ0E
z4yhF@JIqo7yR!5b82ZzMj>6elY_?Oc-H&|`&a*pD;XU(*R%cT2+;0qdyk_0N{>O{q
zMQ2OWJ52|24pJ$un&X!X9wTbv>z@;~%bp3Yzqa?PA^t7*soBwM-q$A0clA^IW3wOU
z>ly(RTj#@Wd!2K8=oBG2R>-O5$R6LcXc1ul4cdH}2<}Rf!pFX4IiAdMM4V{qINXV&
zM*L*2lhW6h_^r6k6qx{adf<?^c#=obStOk`NHdb=WcbN$>n#I~55!3)2WwICVL7HR
z*&)d6@V829uLI^(a43XUlF{QvE(ANDb5)X8FGTFLBWdWrD}Od8U}mHuu>NSwJR>S2
zFPqVP_Wsk|`k>0x?WP0uirXGcBfe+XPFYGyKkiBgbBscY+=0KS{@7ZxF5kzPJY~Ye
z;u}S!wZ0~A4KctV`|Bu4`R`Ee_onL!nT4^Sn$pi=KH0Wa4=DDci7pPy)YHW6Axy(h
z%iZnIVbTD}sR=?ji}q;*={n_@u8!%OQLk@bVFIqWVYhp;rPB+zSLRozWO<)@+!UpH
z*>jF6Yb3Q_1?B|?#4W14=U!||`*BYio5<3Qgl8x#wVybeO{^&&h0K}y=X+boqiD2P
z?tnMz0iDkdyli%7#cuzDI`ZJ|hK|faB<E&oYHDaW(C1gtboTJ{T)Z}+%Y{6xvP%kN
zBTscTsgiHr;!bl7@|j!5_F*%iS9Q*gCi0WE-OsGN_l>CFb`FSIblv0fWG%{_MIzh}
z@IK@S7(0l)DnH(IdB3HQSoa?Kj!Y|)P{M1SE^nWb@J0NtjkeWjl=jIpow;P~NqwVx
z+{L+h+cQ4}@Vl9=O|iI0?_8t?(rKZV6)|jLCe|0#i0YZ!;Rpf=?&E07-R@1i!4QMp
zG7;Iqz-{BtLbb&+8Yc{G8|kGTt1h6I;B5kq#}#Nk97%r2=zF?B8gRbM%Av1$g8jo7
z==YXsa){b_VCKB@5t;00%#SCA5q=sx`#7cXuTfqLHnUkbQSnL-PT2%8URDM+4kek0
z&DeqV4Emp%k10Gwr!2ym<q8Kk#)_jFq|p~@iQiuKOAJ<7J(Rt9Dim-MrAVIJOSI3U
z$>Fi2){@G(4=6b8J&0LB;?5#*r1yrxvkZJ)h1?QeJ(ZNlYL}|1Vrr^Sc00)u1w5P&
zf$;@G5e>@lu72HIiu{GEOAPz1-+-fa+w`s~yF@7|@E?pI3<$10eMdn03g`v`(=q?Y
z(pg4D*?nzX3F!s_m97D#K|;EFfI&*SQ<Riex^w96P(YBDknSEHN<iuE&iC+N?+3rQ
zbcr+fea_z3{$1C8rp+&@p`XHZkq^Z`!00gg<QIO1Pjm>l%Kk`>G`~vXySTSR`8VYs
zm@K1HLSeGQeV6)}B<1{M*tB<sdwIS&vEyUnp2L%*F!o$s55=h4g#%wSzGE;Gv1s<|
zEqgXp_A)@#(XmFsjNU0b+4mIr`I!@?a>kFr2)>VD3GW(F0)j+%GIHk($|@NS@M~qb
zH*Z3=e*<RhJU;Td20Z*$(Yj-mzwK%`$95y)sSn3JP%VFp-s}c6I&WaMLq~)9*)-q!
zjmwaRVMn*!pH(8WV7z{IdM8eVM;{3pPUTi3W27+Ol_v@UT|vqrzB<;aVE*qH<Tfyc
zCs0GiyvfaTyQ=x`3jD>w^<uwsU?B>Fd07rR8~c(16CvzoH%+U)r&TU@zSrY&D9zlq
zRM}*qQR-y-5})4QO|E_Vn+L8PEl0&0;XVjE?0vPqWUwf46$d}WNoh-PR$SJ}yeAIt
zL*v?W6E&G4K3o3`x5O;e(iGTROp|OS>BBVvF9E!}O*F~s9MZ9k@s)1=G)jB>B}Qy~
z3nD;$4}I)hIr_Vv?)ZF3kT~@n>*-bV_4c=*-?#TS2kI<itrMaTyH;YXTj)%w@5ck*
z$o!H2n;`Zifaax!#s+y>O`gz-#_fK??eU<DE=^+6VF6ykZKeuQ8vNhUALmht;Z#jc
zFQ_P;Z90OT@&1dS`xCh{)?L1M-cJ1f1rDJl*V-KZ&6v}0ZA9x}+^I4!)lBZ^SBvB>
z`7~3tIK4fx_BQow{r+W{2+0Ix%P{KM+iBj5ooYk$iss{iyo-`iCSsNTVDGkP#1vQH
zN`6#%B~|*?wYD0JZ$EI7vHru5RlPJP>Z6$&zi2Z}D=&nfa&_e5x1P8e8^jox=l=L%
z6-?(sD7CJm+4-JDZFYkF*Dh_d7~)o!;h&Oelu%EJM~8WiT2~axJ%QXSnnf(jziQ^O
z6LVI>X$*`fN)&{$ntGB&=87;`%S*|klGQ&@qBPABUCc47_-ZR=*3_riajMKWXFK6b
znoGtZiqx>C9|fFUu+wxZd9mxR^Uq;>$6cW*AG8-K-T%hCsa0x-s;K?^YNpbJ&~l2m
z&<7W>6Q?J=`jY!oE^nW?Ja+$`Ybl<vK~Wn544`JrJLnmi{UuB^|C^bC(5}FWRlA2$
zd1RIpYQ?EG!KkER=~GbJpN91`yN|1Ja^dj$QL%?t-NAD^V|W6BkYdy1IZNe|8dK1z
zGnxen1lw40__}P2?-^EUXnhNM8Z#ZBB5S4m>W2p<PX;yrIT2L2%6QhWF;MiIRklwZ
z=YRS3$Y~icqSG5WN{$%OYsf$%m>e0AJ=}ppha{C4L6x~hB_x*zx!@TU4sZ&n7TOdJ
z-%eu2H?PtOZ&6Yvn5nji2>n7TnZEkMwpNb8M)UscB>TZz$6ay4#uYKSn5O<{fjR@h
z6oX~2Ng!|Ln^nI%yUTeFv!b}=amEaMl2H+vI0b`etP74!*l9H>62!*;_sYrNvVO15
z0)=TleDe9&HB6x&HtP;|*JpWW1XF`!{nQ4<p?G&{f|(;v9h0|D`#Gbpw}5t4CBq-{
z7F0<$6jIN;YTAPQEl-*Gva7jE?psqzKk)8c-=2)FbYO8PzqMStuF{!pB#)1aMDH9^
zBt&BFc8Wvs@y5h4smD{Y&DeB0{arE{(a-e3rxzOy;;(!bU8u2>iJN9$6t)V;S|dJM
ze^u=MV)0q-?-{v6+$`hZU+)bZvzj<LpLbjVFAd4pSf^m2yIA$P06ftYS4MgmrTw1<
zb@Q)xhn;v33tu2Xgf2Z7I$)$K`lo2{P9>Y@->SA5o<av%$(KSAaI9=AI+Z}}JxCRb
zq4ti+fMDn97L|G5O#ic`&#Z^;b~(mWKV~UnJW*@cj$*o!2AbPiQd*ah5#Mj`PfGx=
zkSyW|AUH#jIkAzAv2=Hzci{h5r^wUVZF;=gf2Yc>hSHLkcSr<#?4q8t>xUxa3FCJM
zzWRBi)En~O7jXg8dnt#5)e;{GV9bh!*&BMNNa>L6%0}d4dA4M|U*ziSp?7?^yWsFp
z=a4GDgjDA}gNYEH@hXVuOScRGHBRP6+ey(1xeQ@J4yxpAnJdZ02v2fMc{Os(D=yAX
zCJiIpJi}l`L<vdh-0<rYh05N$d@!NuQ||DE){<?MZ|#Xw8{TE%vuOiN^EZG8<2W22
z2Uw7Ex^$(;i9x=5w``)f(~;7diy7m*F_!RSoIMzgZ)8a?edb3EyZm<5=h%ta_ov0=
z2Vz&N3F@!=8WR{{B@dQWIbo#N$Ev-%X>ln!HZn0wM8DV!?TluGu4`fn1uFPay~g*Q
z&`-K%x+te-c`Pthx+tB8xGb1jvwW{@;@)WuMKZF$^5#zd>6<BAnDH;Tz^dG(-o=^;
zPi4bw7Cr`4-){0qY`nIw)4A^z6kJM<C`r!!O;-z{sMrW>l6SlonZ&#SAuJA_iRd>`
z@ZfAjD|=&1i+yx*sZVD0y(r~E^36=|&G_KT@-v^Ky}@d!EPDL>fTyn}YYcie)S~(H
zP^Rj}ln=CxeBK@Qih@Q3UoiSc!@wZwk(BEWwtcY#>hA9bx6Z@gDqGXeMFV6b=w258
z-icBklHT4~Yw-CnKW)L!jhxHrDg43yyq7Ndk<1vtDZ43Z>eX<w97RdpQg<mNV=^|Z
z)DGq67NOXWrNSfnip0T`=@Q?^0&0`dLnV$D&4x{$H8d6CgkmHM8#tZIh)P0lCxMjo
z&x16r8KwA|>=)ZwK;iu_?vo*ey+DfjdYpCsPj`@~#+=B)v=yy*oLlyxQC^kdYaFrX
zG?5vxpCu)9lt~0iOibrC#%5>3_!mDTuMMQ}+5G_B*}s(-G4{=s|E8u2UHyuE%Wb)A
zXvLmzx6$UONuEq7ae(>(sEQb625-g_9M%84OR2fRHv4;IT3ScCpc*3IyAdh^iyk|=
z?G?MP+BF#JH)(Ej%Ju#@?@<3~fGhKEujv~^+l;ohhZvnYLE>4S6pD&(Tq!rq@D`|j
z`}KAjrBbcTYxg&)m=n9em;bz&_>mvA2Ba&C=5NJfgMwvN@S0py#k^map*;={`|=t%
zh?&pB0LijbzzL7@1+Ev}cPc4*adSTU(P?G5MVw2qZyBZzNC=M!s0pC7+0+}p+*Y)c
zFZm=Wp7o)88(Yq@v|zqYj?OGgrzU@WD6_L0^h4Wng_LHmbyR>Wl-o>FhtTc!{_Ymc
z3NUrdq!823pvp885MNq~AZ5x3S#1#x^R`NnS1HqDY>U{d6|e2dXpcizX&U!Hd=zk}
zVD|=pyu<o~fOuh(gM$OvhtJ-NL?A+^NR!|uvtP&v$KK7o^V?(r7>Rz5!z<+JQ7X%i
ziMXJzS}KkfWxriX@(4#4?Lfc$lcJm1+D&l;x>8dCaoQblwsCy+JgxLSV;xtF#op1X
z&`%x5XiZUb5U}8jtuo2YacO!S6z<*qw3Rj8NV9ag(#Lpr7O#lX;v}~Dvq5lIw*FT=
zv@0t}1jeqmHS$prOXT791dEok?@R3V0J-r&%f)k3w+1>dHu(GUPwZYnMlRzrGH0E}
zPQ4WFf7H2sww>b?Zs(Wo-DWahK9kC|K@c>@{Ft_JCFzm>d6PG}?o9V8YNor}qXyML
zGZ4u4aRwAVro3Xd&B?Ft&9?3_<Cb~_poDnbEoC%&_wFt(*x!iw6m5R3#w%goH{jsl
z<Sk^3TL-3br4N$_xrz3<+>XoHRk$MgtD8VWYVZ-=0CS<+Gz<Kim4ioNm~3s+k(Xtn
zNEHqTUV#=5zgOFeu)f_Ve9l1Z4-Z@ahpoaG7lSIdeKSgZJ?!Y)>H9{x|ESXM?%Vgl
z99V9Qcjf3d-7W~-dW$g5kVEor`9>rWQx65%ALp)I$+y11&*FN5F`d6cn@pVgG`tk`
z%GXZ-y}Uhg!A{O<H%PCH$J}{+h~Swkv3joP03RQpZn<3Pp)hh*7Bj3UR7)x-udOcd
z*K`yTuTWsYN6A{B25{7x>K94;<J4DWdh?r&7F`ig)XG~<XH(h{-@kcF)OMI-_DJWQ
zg;07zf_dmQwmyFb<yv?PD?_`1`$Ev;S>L^g6HLyTY~Yr<iNu)^*wSt7c1m#WU_iky
zj%t5$W}28!L8Gvp;Fg_XN~M_YmTf=*2JQH%#GCJ9sCM?tFDEGuO(kW3AICxU=9MCs
z9`xsoxv&J!Y>e=6=*QM#jBo~}q`%;n9c5urZTXXWs^&Ql00FhwUcEZ|REWBKFT|NI
zXEGwJhD_tr<o+bVIck-nQ|FdjSZG3ZPZp|ZC!8Hf9nOoQo#sh$Hq+RFTq_@A{)5rS
zbg?9eXfDgMd6O5xf%_zLjH?qhsz9Q!TZ8<U>H~1+=ta@*A(4zt%tsCD8Tq5*74FJW
zQG#hC?=P|T6sWE`6<-%rXB7PS&5h3(Rj#S}fsTVx*p(k6%A(H$OKnykd}EvNXvm4A
z5))l<gDiNTZBKdpnVi@FknQUkGYduj4C4xkfGVU4<7W5r0&WQ^*K*>)IW>+yhK||3
zAgB$_4rj^6V=}w`0Ff?h8;bF}HKTc9<grFFj`Ii)<~ab55wYvNxuxPXQ2SW34uA&q
zx+kG@3QTE^7jxHQrEuFn1~-2Wj`*qrA`RHs*#7rz;ZN&3M{)MQVs{brte|6qCR?})
zDHKnY<*ChDI?DZ+^Z&s!@k>uvj*4jhc<nQ%$i6Yv%a`T(P%c$j@SANihR`b2)A%u3
zUn;i>RjTr0ms1VG`c;3=Ih*3Mw~vYg5#z_nV(9414m@xNj24quT3%CG=dFT<6VhkL
z>v!*c3)&MUk0l*v9Jg;aJr8=?d5Wa&3aW_tS*>v^-Rx{v3crbxkLV7E9V-_S-=<`%
z?TOXB_-Z!I$t*wKo>lsF1JE~la`dR|-M3>yITZ^#c%f`wV{5CZoe>$g#yi!#gw(&m
zI4`B$UZJ8|=ZPezJ~L;OSxWCDVd+<c^}b?Z0+QJsr%4COxL|R0wFX(I#$6kjbke3F
zF$e9xfk&6WugCmpDpdb+UzbXG>+i))w>z`X0}h&tUU@s|VS>_y?G>!4qAV}oS7GA9
zkrjWQ*Y{HIt*Jky;!!5`@k)pA)Si>r^~~93fp~|f0`tXV7KSF>ai#SyCYvAiuJ&_x
z9ECLuY3(Djh(Wckd!QP1gJWQ)#td^|x>a+MmN`wDAjA6@jAPB4v%5qj2{=M85sBW+
z?meGP_5a)0^+D6n7G;v$JZduIdwfpNIGPXrQ}CC-%+0B8xU%&c^Qyy~J8W(x`*xK{
z(1t=T&ugMU!Bp^k#_ZD|TYc#BcU0boPl4tsL315KSS|*XEv}5ly}J3X9o^Tm(XW{G
zg!cjeIrAt<co`BN&nnBS1*4^kY3QzkIllof&)c-m`O%nGSss@EiFAWkwU0)O-RC)E
zg&|ojb{kv|vZAS6X0nF$qvn3M`ryniDKD34%%9xqJoN183T%;`*nN)MDU@<Q)o7_S
z6Oc2_@xrdR%iX^Z`usHRJyFyVEu^=9I)Uy|vAyz7t5aYz1{Crnf82j7kuKaxGES+)
z^4`(!W<G}zxrCIxkgUjmD7J4k(re8!nD+eYL#H$o3LGhI<jE(m?cE%KOABO?TJ-A~
zk$?D}BJB247pAfhdRi|hv73Ld!c55+KrJ6)Eg}RETrLcyaTB53I2ja~;DrzTbMx~u
z;j|tfy@3QJsbe+J!Gc1JL;p01^>`{MkoJ3x!*)+R$7X%PqLLna+pQBpcX2$b6ARVQ
z!O+M1=z@L+Oj+W92#!{=;w;ay&s1p6qMD#^oBe4O5yXA4-?%~93_wv+jnXj#x6sED
zOPN^xKNCOwkL{>}Co(zDEgJ(F@k5k|`L)%8F;X2T0#Pl}#vHLSwFj2ih$)Jfo^R7Y
z+t^l*PG_yM{}m=161KhV&a1B{t!%q*f-o_=13aL#!~8ypU5WV4wrYMK<RALkAh_GC
z8v=awTUHI4r(54F#<W7It{(jl)}rw}7KEp9TmyCk()WEn^!zpIuBIoAoV9tB5gUfs
znZzd)%3n05PPR_dSycu8p&aGD#0LHV?Kuo=p|AqHv(^3t_3g1pK^djZ!En`J1^Efm
zxtS`n?vs}T7(^85yb0L{C3lY_Kaq<+JZnA&^_878*<#_0uv&qvEeM|UFK9t!Jx5NC
zw!N|)*D~IJ1|P34XAf_@CI3wJ6H+)9dM+y(5FT?iO7(*Od|?kzI+h&yL)tT9SLaTo
zRSK!lC|qLYd5Sg^;(oV_C)AvZJXR^;L|dlf*(TS~!!qVUs`Oh0JkxQa{?QDc<q=Qs
zo~zq#J@yRWoA%WL&uXU)2M3SRVjSW}q3qD#Erykl1nRHq@_ioJNYnv$!j{ha;-Wa3
z9g+!7L8}{L4~?8!><D*iSE7O1A53DAl5HOfFXx@Q2;@R>b#H*)mg6J7`6`Vu#%8%L
z_FkjkGw57GrP4u^+Gk>M@@l6?Z^xioD6<6C-v0?0(Vu72?Ym{?QTb$_yV~50Go`Ls
zm_(8iosHkX>n8|Lf61i(a>>i0r7oo24g(jj>)4KE+0Q&8+%gi@caQ-UzC&If5!Y=^
zg1fxI;tZ#7$om{3)TSs)A7-{FTIK+PjZ+5aeuiY~C3B~qrB){t*_1MAMn&(o2}jnA
zp3gRQDch9XqxLur_@N{BEI6kEeA(n5)OcejB1)053R`exq}8H0WXvdiFHMeoB+C6U
zal(!s)dQyays#PXP)|a;Zo)GZ3ln|6yE7dyPSkR^Ey43>%pS5_zS_v<C@B#GdVSYM
zh>*yt5`yS7OXK;I!d_UTtmsd<_}l~ZG!*jz(g8x+wwOm$C-09nWIYS+D#<|ZlBK++
zwUOygi8%}&9Of)&qtp@>@J~`Vy(TW}H{Mn6#1n`9j$r#BXF7Ua;Cp(!C*6?kds;d-
zeMWK`rhdJ8>PvpKjS?`!AeH`>o0XMh6M34-97%q^g?P4@_{Txp10^^rZF@ZV%D!rK
zR`IG?%N~GXKT%$$YStziPo{CP{1XcXwZN|$s}N6UNA`c@=s}#Pxp&IF5EzN27GG}f
zLDWLhz=J3g@Ne6HdU7Fstz~LA7A4LZ8yo&8?5e8Sa;S*L%b^u+u3!I3V&nnBYg@!Y
z=c9=@DA3o=sU3KZ<M0siBMMA&^psoXXL4fK>H}P-?`7la{$a{sVkLf1(KREE@?KB1
z>e?t)8?*A;|I@hqQ*q;UkUtVtw{oTBF|vyOgrHdBW<05M#Vl~^$dF@XG9`>k(0G8N
zUwwb<;;U?aDn8LU$Zej_L6SF~xvRk5+acb#0!>;74**5qj4COP#9-WuUZ4kBglWat
zByQcIZUuhR>_fB$+OGEN2l^JM3Li<wHA%<W3dI!=1uKu^b3Y%joDg>y=McikQ;29U
zs-D;tlsRW6@)yD}N1iHEZ-R+#RZ4Y2U<@7Y4PUe12j{NHp%YAmvR-`w;t(>PF*Eg1
zY*P6lbt5-I{2)%tIZ!ZnW9ZP(y(XU8+TMO~?AE7EPh}6(UNs_@W0iz`-gf1-UyAaG
zq-7K@_L#!qbpDOIPY^v-+;g$JEog)mp_aNjO*3)Hek=Zl;i*8U1O$h4^CGaiu&;z?
ziJ|7$oH+4MCYW47h)0D_hJFZWf9KW?CGu4Po}PuZ%|>FDoP`a-k(272WUuL7&!Juv
zAJe&(R~AsRbox!Rd5WBzWV)1vryR0c|Cx<8okm|G+Uhj>Ru`d2XJA-Dg>X~&u7Tsp
zEY6&csNMvJ)9CD*EwUVh4&kL^^O0=Z3iA|mwwl<QvKoh%$=gb54Ts9-5>yPt{xHnE
z-j(NKxEbN#<Jf%bU)>H7QcBPZ^50&d`uz8)Ns=Bq1eu9+f9;9|6}A${7p5>kpJOkS
z78wWY*m8GhkYrmaIZ<w<!9!*VqA_CN6JO&+ELY)=j{yK=RK|JJ$_XveAg)jJ?-yju
z>jra4#RNxE!_OuXl#Ywd9(P$-$;h%Txq00mc*b-ke-^}Pj%QRy<((ma&zDPbJ(Y*a
zWvh<$O3P`we)j&btG^H^<g+}~fGtx+3PdE0uR{l|cUoi+#*fnm?l`6FQI<cHXf(uq
z{I^z%+mb#?5nT4Za*h-rFDAcpdF3`^;*q(<6f&VKt90h++~ImHxp}e2T95;vOUcct
zz^dUX7-27I=jtQgp`P)CkH47X=^)c;v7}9F5WZsLHj7jT@DpPs>CtdqfPyK^WsbG7
z*-6JwPQw!Gu62WGK)+z9nSgG>U7qC`Z@>8XskhacT424VkT_Q<O}?hk5{DM60Jm6C
zb*1EI6@qSi4_!zG<+RCas@%qrSq*9=(y3JzsZzN%=H>UD&7Yh-5fv$O%wi<tRDC&k
zliW3pbd$~AZ!<PtC<dwDKZy_cP(*n)8+(8_))()^8&-?QP;bqhhBLmcXjl%1vmr5e
zevSNij4G%y86VAM87?VPtx@n=x;}J*RAWqzBczk?j|N#LrFQ?d0Oyc6w7CM<tG;_0
zZV+8Jt0S^r#h<8Dyd_(Wm%KH^4k{~S8MiSLklR82`sDsNxo!1~3zB~X*dCb0Gc<`F
zA5AUbitcI_Rb}U)NbQjLN<YBJ5L0ZLEf?2Nyo+Z$ndbN$t|ohC9A<kpetHl>!Wh9)
zl=tgyv1{?Hf8-oMw?Mxq0uCnL9;NV!3p9#z(~N?C>x|@nAN}(0@9R~`=oi4uHSfS$
zh%U=MS4hk@zt{~R^tb-M%oo9V)1e^tj$_#y(`$gfpSXn@S)^Pw##mCQmA9+r=Ii35
zyZ*9{f;alYKkBBII8E<hW~&lnMI0u>fxbI%45nVlQ%@au3TM$cbXxP$QmIKd6RXZm
zISmmbD>e=nPZPK@<}4kW$_V}hY*?+!pw$~{3qqWb({-nSKl?F4+mI5}h;G^@HpJA_
z+S#Yp`)fP+TfL~u<{jP_QI^?i5O}sx$I}UtL;AvHrWcz8)%2qP9~D~7mUADAVR6Pe
z|3;PVW?$rD$0ubp1h-UZQ3$~#XPcUy#`E-1B!5_r;5Y=omVur6`5u=O&Quw0f%ZuX
zmzh)2GbMs-;DA#?tTRJJj}3^D6=q^Mb&SVe;A@(<)01tI(Cmq>rTydlgPuFNL&#Lo
zd9|9z><AHw)3YhV*~WoBQ@RS=b_;nyUQE>OEzY)X&7Gmn;j$kK%>t%fn%V~zdSE$;
z;M5vo0;W4lP)M&XN=G-7A)aMg!RpjSAJ6s>VVm!6L$Z(2o<=qx?A1gfF~!*BMg+rW
zCt-453S)2I3B142eGT&A)pVP^0?vFr>2?#*RQqC=PHgMOGfl3n6;0bLU{Z(!ya+Lz
z54mvP@Q(@aMTXbc!Xwz5N|v~Y7?tq}+|qQYwLMsV=Ey%Oyj<%bPxZwvEzh!7<U=Tj
z-EZ#hDtl#58B3Ipp!-3pd{nqlS`edY5`ljYJv72S-;9|JJiYd1(H#C~OdU|?ePHC5
zDSS<5<T}cGq)ha4^?gUl&+Y|K5+>B|!3I8x{@gi!TKjL<3Ij*w^vKcCMtmdv>Wv+x
ztqo^lV~NLv&K?MDd`wTrC|r&ivqliyIX&|spWI(DqC!596HBS2)D;UvGE}ZcSVa=9
ze{=s(DlS$MAefs+?q)lG0>~h~wN>;p>q5j8U5qfdS$d0o{#nhbyJ?Hdmc~CmaUXu>
z@&sE__az^9t}Ag<zY<XtZbU@N&`AkpG-W;+!HqbCe&R48X79Dn5f<&H*@YtEpQlC=
zneF67YO$of4)qH~*DE*LZP)J%>)R}c(`d?<0PW!|fjd!sBaqL&DbjCF{1!1<l*aUD
zLSl)*B};<s=-5JDE8L=_GwSYq9(u;cS!y3BY;oKVx>vQuKpP}8?I?A}neSn%Yoru%
zztdfBnl5YaC->2$q)q#JJf80AzvaI3efvr(nj6l|JNYK1jzkk#cKpiz8W-lXV(7KU
zS3tFYKOtXA|8?T)(7JT-<tmk8__`I*)#w|M(7MD?;#}pu?2<vLs6`YbjV@mZh~Rag
zadv2xrm|J=@7#EI7DLpDo%AU5l<}=<NVY^fpWP*gv%Rmww(+s8PlI+!r2Sa{-FUa1
zc;-g|(^6bfFp+6?NdMD$@@_fKvBc%efY&i9su7v1GZ6DEg)84-fq_b!!00Gx9U+eZ
z7375i3hFlB*;MT}UKT<1eH;`_L%HnCPn4YP8H=wi02*cC_uKrMS`;vL0g8CV_rH+8
zZclKuUYcSD8*`?N--cT9IyU?Z{$4Lf^h8BPrH|w)oEJZpQ5ET5tj?;!_aWv>kBa;Q
z$L}ZtcLXMn>M2u2<NAv@5>1v2fZ%w{pR3_C$|FcSUTuFxqr_kT*-Pcu45=L8{W`F^
zgd>*7<-ne!(I|OwHNsb%>6VrfJE*-uCEF6qkDgeYghVVbxKZ%QE;U(wd`H`+;!<@I
z?AKt{^cPcpS2Iv^156G1oleY~Ha;rfQFXfF-om64Y`a3}5XG6-skm;>K5_bdEnW2G
zEHxswm~HlOp1VQg$rnj^9c23V<Z5D3Zv0v=d?q9SmEcsjc%itFA3xd;S5Tt|j-`1H
zQ&cPG5)9lxKISE739`<nmnzgW4^?FR20uFk#}#I7KQS)L%HXsIS>oTsYiVwamwj2H
zR|5ATdW`+6F<X<x`S8Bt#-(-a7diSM3l3;q3R@KaFeCYimAAFCWhi_!5gU?YOw`>>
zFX>w_V65dfA-HlT55;143PT9obF>)!1GGUHR3vdSL5904zi&`Hx9LHHdJ+XQb^S{6
z>E^_>{c4lX!e100`UG)*u@c;TXfXM{W$9J;9BhHzLI($x+k`FH<7I910(fnti;2n{
z?g4Yot@@q2k%o%G#r$WVchNIa!hzTQa;12w@oT2s(I%mBbaLy%(vDKI(=*5*@eJk?
zPro)=YxgBKRCOwbl05PVZsrtJ-MR=2VEQs5Zr@avmJ0xn929IjxrHL>Cd_p7-}pF<
zMicW{+5qyVbinpQzc8D2>EdYvyjnHbUFwR4OtEPV=!FR|<i<x7YNUWU+1&G>6uUQP
zoEE`!o%vEi4B?=N@gQ$XNW0cR7*oQj%$#I0j^Zeqv-C3;!3*1Fwjd!MT#csfBc>@V
ztg2G4;mCOFUU6C8pKSN=F}Rxk_^HL(+D}sN-1&Ego?C^F;Q&sp4z?G{duL>f+Oxkw
zTjD!XKR-VH)k$tbo}?)_?b@oO3y&o?#b5NDPYcuEUI7?inQKpt+5Fs`8g^&Xy`J`1
zKLLi99^{+x-dt_o0&<wGqC)O0D>uu4W`KdNE;H>M=#KGKMb0#bvP8W4Zd5ybA!13h
zDSJ5dwknoNv*JVy=1y{z7=?O^fJcL%a+52h)(BZML`lvJA~hZq+fh$=P*dL6$oS|_
zvG&NkRIiS;O39N{-}2@{gjrwv=Ik?^roXfA^9yNe9iyGLhi!|h2#+~_Yd40RvnL7P
zx&9)5ZM%CVLL4n}KGTeT15Aq>-swY!7Qh_qF<W6rcA2xut5;kxy)sugk&&%Mh8)x-
zqbQM`=VJhTd(*w+cCIt)u6g{MQbGUaq-ov)Il9Ei1X@04or~Vy6-ERrVw%8yE<-AR
z2s_Q5c&1P@vCBTCHW4PruClz!tLkK17#7^lnc`Bd%vSK{gL_+PE?i5YJX(I{eYw^V
zsya(>SgMA^7LRu6q+F0so=%b@b6rky<v8JLlV`9xDgYFT9BPi$jnX4dT?x5hMat?1
zy)ViXih@9F4N(G~`H1ZJEd~+eVZqvy+>=Caf#rm!bcLwNM4C?Wy%`x>sjQo?!c3fG
zS|cfcH8rX2BebV|OHMAmy-H+eSa)`?*fxp69)5teJp@y{@^cb}9vcyOVz+AON{S~h
zG0)D<OlN_H3WSKDdmzcZL^K8TzvzSCr`XhGtCGd<_&|{J;C=9kvIJghUd@hd_U@Kg
zn+qG1mfPf{Rt9A3PGaPNra5q>g{&sWN4JxGE4p5!gK36Ge<LB(dGw9ypocGKI}I=|
zUwm30N_{mvI<X~0Wnh6huA4hHb|j5^g_vTR-pJ=|my~!JOwK}lIfzznu^$cv3GcXh
zgzsHTu7%I?sjvqi7GnqP`?dMG)7I}27bUiKA1x+|iAJDg4KTdOa2_E)6au`1iq{f2
zVl%kCz6=m!ZdGJ(aNo_;U3ng<Olu*?2)o579=4oc5%yZlx3ZUW8K%{E6_&*4$70^9
z{N_|<1XM3Pd2fB-T^6>gQlS@yMAj3a3^ZE%$TQU=%C>>nW{lj5y~^tR;lJ;?5IO5(
z$n;p(z{e6JOH5>o7aMf8!0pk*e;M#~>qWrG4>DU?0NI4dq7N4hMAA8B{s(7!lr<C+
z$MfZ|(Efcr27{$H`SPQubeYVy&~@4VSAfd37_WoCqFTW63nt<(Z_@5lV*tol$4vR=
z7@A#GH8|dS)<Nzk9`qEubCt;UsK|2DYXx*G!x1U{{BR8U^*~KYxk4AvFHt;ytwYUt
zmMU;X2oDe(g`gIxq$efs;fC;96@7&3rHfAC2OGjHN-rr{BrsvpxH6=t7%^-%RfxF7
zS24ckXlCn4SnI>-Ar^y4xtfpuX*2|Pm@NU?PKPqHb*i5AqiER?b^;~>#5NXQ`;AtM
z8E#lO9!LfGvpPYtW(>mZ;)`?BG*UNBt?Qo$*uo`_2u#d}GEB&de#FL=%$1gx|4(GE
zZyJ^q7ZnBb*?dT9{&C3vuE(MlTY9R(*i{O376Zp;iglLl$b~vj*^7k96}932l!IoE
zQ;YJ8cZ$_8i22}N6QraTQGR6e+C>lkOgy#T2zQG?VCac5m2%zG?3HH=hC~4it<dJf
z?S`1-cPk+YdgPZ*#E*PzNKH2I^-J{JEPY<V8Lv+gdP4k7MMIP`RoUiX{_nnQ#4o0D
zF7pSF?i4-zdb3c|-%xrpaUO1(ii+J?LT6E}8F>!`mxr)E_fn~Z(MJU4(x9OK6L%ZM
z$f}f@Mz9=X=Yj^K(cbe+RV%}~L2PK>Y8<)1i>RIokf&Vs$tu_1?~I@a-RwG*PH~Fe
za-j!QjT$wPPyM-{K?v^wf3*ydo!_9131h8jF#WxCrj*Q163p(>At_ddpwoThp<89z
z{iQw?bDK{1l~=iC8n5p5kM?!fj+DH+oa(2BBNO{SOY%vZL`~9<{dliP1!rIMtrB!{
zUgg5`Sm1gpEpd+|2OJsBCSN|-xA6tVx75X^pYIByLdIi>JwP&GI20mC4lFawqS1yM
zpUTqW=|lES=`Jd@LDJ*<xuezCtSzjlxt=wi!U$#8)%%9-zd(?K&W1bY=wibu-2fbi
zvGfIKV2WgOkAA}Cy7roSrI1}w;fy{_w6_DAK`bo<_XP?0i1Pj86gMhzIQ{vGzhM?p
z&Upg(gab63+lxZ)-aEETIeweb2tc{VeY^Vwe9X_=jqy9fyiwkim6fTDsJ)j7HW|Rj
zkWevj9X3$Iu*E-D4E&u;KUDfMvkJ|#X!9Sa{{tSQTdr2YppQcjt4@=N0NyUH+G-o<
ziJI+qe?_Khu4ofp@Rtd5U(LnVj@tE9O5qo2&u5>m$0Z@5z8JyWDlz6{MIW>1z!0wB
z(t*8R&-pl4aOeBaV24v*0S`Is+IJQe%(ZWRW5c4Y8TN#t23Hm-S=w`G&A$Km!}I=I
zFCzSO-Hcvw;_Of1wUi=LvCyrzWQ>>&tT8vfqWISD;l`+hWB`)mCfWHkT-bcz^JTLc
z-79!_<#>;@2*rhu*-oR4QF2+d=qF`iMXHzJ4q8m;_#$KVmls6u2WQxP1p2G-W8@WY
zZAOkl1St0Lut?copI5+N|1tddF*pDm)X8&Y+Z%izD&3h%0$DPfB0ozH`;FnH^ZVe4
zghjeAvZB5E_(kAr&Wj^&DfOGAr02vh%H9{Jm78aMUJm;XL$Y~IK7V%<<JWbuL|UYP
zEVz`=`LR(dPc%Vygp?vZ&q+uxn2>1r3jCM=8_rEN;GLr)x-#96EJ37qi9SYyT9&9?
zUjtEvHu9`bpjKa}(W&r|@G}r#)%I8Enhf{HH@samA=*DBf|(hbGDN&=I?#)8`ANSj
z8eIF%cZ++g1WC^9?kl_v-rkrbM^~qEp*4^eOO6eF<W|%owIs#vk78g&sQ3HsZ9DFF
z0P-*_z||nmyq<$S`wfuxHzhu^4WCxNSSFVyJp>Y|iVI5RE!w{+ZdszjMXIJTcse(&
zpo<u1MfEurh@QE?OGM+sPVmt?ex&@o4f^yCB6&u{1;sZ|z)LA$=zSyu=ed$+<&cdn
zh_wvhJ>toT;w3NAKSI4-;d_>O4=V|@*ZV-?A8H>Oj-~(><#@?DH@o1LJ?iI|=?9wL
zsr5^~zaY$qtG3zUZNTocEX_<o{~fl+5w;f{9sRWS*VTB8FTH-umUPdWlSM711+4jW
zLg~?xt2lzfqP^cb-#j83>|!&y<2=@i3x4p6^7@$QLQU@$(ZU~ARoO<#*1^o~oBUwS
z;G@l~CyDKSU>vK(LZT2rZzgh>29=&-vahB|n(ID~sHmuDzLf>!W6ClSZct_#K#E3l
z3AeGg=ptmKzV&*#YxmF~Y#CD@rL&aNKDHyw?o_KR-9ujPZ6iOBDOA*a@Hh*8oSM4P
z(vr|mRJ%TcASBEoF3XaxpWvSIY$wk;zzEXrNJ^E6>ni1rgAswrOcDJBZ^vMvWQ^QN
zPcG*5_4wl^5W>@=J(+{JCqHEa^zwsXM<E_Bm^)Wr+M>y}(P>u8j1Pm4Uq|{(iDV!$
zKkU>=T6A=P;hT1)EEIaJJgi~4^Iy5g><1qoA8_DuDxYhnusK=-6eWMkDPF&rM}n=7
zBTUz^hc*O9<m>6y>8kY2qZ9Q%`$xc_;Wk9<md35k(5QhB`I<{H@dd~}OcY%>p>J32
z$ktn7Giuc*doFT!I%%4zIh4XBRl_hKi%1x}o2#`W`cT}P3%M6ZJp^bY-TA=)Yb9}Q
z^4f2=d57flR#b!cRf|($ifk>8pL8>urVKy;w-X4y^%wnB;}w(>vzRk3|3f@PD1qU*
zL#R0>XDKb%WQR?09i6NE14NiW5&vNH4<NR#&*SBP-cn1cz&u$!^Wa2h<nPoOj02Ga
z)ks3}p2j&#Cug@z@zjT$Wxaaaz!nLL4#P{Au!p`eXv~`sitJgQX-6tBj%)yKkwb8C
z8R)ZH4(|Z(2<y-(rih)Yh|A>B@gPUUVmp{$x8UzTD&LbMudEV++du*w%{IRS?MZ!W
zsYgImGc=W3rrC}6>hNsZ2*lhbjv)|%8uXJW{U9R(#hY@oplez&Az#RIE#)gF2t&Dg
zgQIMMV~cGGUM%GLCexH#nPZ#f4BwspFZLkKEYE>*p~&sz2T<Kt%Dz>w{L;u&UjHEl
zY|Osmg4M?D{_&nvWWoKo*B=gCTt*9x@<!BjD2z^{o}JYU=4b~Nw6L={?;DBvjm@)|
zcNwmOue3*w%13jow_xsa{=+XeCMIf^JUBZ^G$TGRa%{l!vKm0lk29f4ZWwa-VuIaq
zwZ-87^7kIKm44}yQcIWX(=D3=(nW$7nY)ZMeNeDM;4vc+`YQePX1*D-Db;_*frVxE
zs>&)wx60-f@9Wt)<}Jt=aa{v)M_3SVV)8CEMP;}4FBZKjZG%KD2GrLj1i3s<H=FH~
zhlmGp((p$^mXK<-m8U4eOj0@srjFfI%reg?4dOkts;oIn2EFNCkzCNRY#&`01%0k^
zy~?Ruxwuisjs}t>O*~e=qwmQ};0<^n19J>SGzxFx8?|uZJF*2}0$fH5RHT|Hd<m0%
za`kdGt*J@q(fQQ?9wQmvhsSa)hR5!|*Cqcm-U5%7&4JOs8ma8u?p1MA5WRA9`NLZ5
zI`9*mCN{+_n;Ni}dTcCGFA(W01F|z4JJOrr$rv7VBM<KX<K_?}a*j_lsF(}Y^)}xx
zdtUKtzo~kAQgtcc#W2>6((m@@d(>wo)MMPoctfXF1I!w*rcz&;yqvy#9Bf9JM)eQk
zaUYbf9YOzDa`6shT&+kw|LeGMyN!`pR1~@1QenPM^o}NyID1==(^~S?8iyaV1~DG6
z$<R22Y(HflJ_4a~L`<>>(F4%9D82RFovC^>HDQ{5dCXMBqzVecA`R~?*boMJlO<s`
zlSPF-Y+4s?#-C!?wC)La%&1!0{LF-PudJ9hi9RU7<e=%W@Y|Gb%Dc)qHH!ZvD<=z;
zJ2|>NZ@e4P+pgBrWzv<+$r#l)Cs%o0{b-hw!FLP2{h*vB+-Q9xL*y4UvOdTU7mTyp
zY~GrLzMJG_VTx`1D4O#H6TFl`Li$nP9WRR+@v$RMN>}NB$T~{?tRl3PNWyFXNQSd+
zdx>oL>b`qr^AJxN0vkOwt+JhB8LuuLQ&+vtWsv(<zvzkezln?aFCf?F3V#h0NZrOk
zbc-%T9sxNa00<X(V?&v5wauB;nnW9!U`2s1ay0y1VoMGVg-rh2H1&Tg3?`V!U`a%_
zTjd|E)laKFNH1y_K@TpFG??47Xg>la#b+_mZT|epux^-h=r8p_GC+-W%stgfpAX#T
zVRLS?z^s8aO94)l>bnz+R4e+L)Vlfth+2Jvbyg!8VH#hVT2h_5cTRGF9W##GtiH6c
zr&ZOlhYY5FOOhS>m68A*m8+=<WAN8|&Yb|izflgXKNl6N-j{%v-m!1|q+QZy4a}7M
z!u!tiWFn+rkYOAGpHui7Ng&4MZIUYcq0Psf1KgT8n6vv$16{7{em@v}JrjMnb&j)9
zQ??w<$~<pf8+k)b9QtxirrX(Md$SM@mVj`L*P34R4xgmS&)D&yKMR1M9oxK-?xpJF
z+zuSUC_zLDzh7OF1w^nW{$CL1);Ca4r%5ff=&kZa4cygVxUo(w&{H^;X!5yyO8**@
zTldljGn3p(aM@TCom^W8l~?=-njycQiFpGXj9ivmf!p+ROAuZp(*2lS2rz5)B)suE
zM<02A$ZdEaxB4xY4Yv@kz6ub=7XC1m?YGeV6U7tG@IRAyo1k*eb`3uUvptT;FZ)2G
zDTxBvjxRwnFg$em065Ca-p*dt{X~t|q7v-MXFdeWU6LQ7@RZm#Zz}V(3l<{o)Mc+2
z$j&??5h>DT>Wzq*4cN@2xgK>+G$CuS$%GW67jbRd{q>H_6?xdc{ag!fqSNtmPEz1i
z8gVWzdLi!J^VM%KLR65X1@AAK*Qm&~Cl6+p^N(|0VM_z6b0a_={tFgyI)xOqcTkxv
zHQ10Lvj$j$BCBhw?)c;H<oVfW^ZxhX4y}*SP<pE-dbLWl3$7LrIj2TP{Hb{H-2Ctu
zyq)2p7vMBtS9MBDPp<;S3?}lgbjMcR40x>F9;mdQ)@FYsZd#y`_;3lfQG|nhJ23(K
z5-uIntz&6L_gh~OXg(Js`j~;2<+M!Q&I={muu*%b{&oCCM5s&Lk7T>;?v*5CA0N>h
z4kCCsbnk@{>cC&-dQYuja%GOg?7LknTI4A2_li|}x?DNb66}iC8WN&ng@$v|Ph){v
zA;vLZl%w3LmgKlzP4SxiJ-2eB$(xkcC)C$7`7OUzj7tL5pHlF;CbBqnA7CZ2lkw_B
z@@gYb(Q!@aq;mCs@%=u9RJf@jG6&}FxmxmF56h%&_ymqDbUEZ&$Cx?1+_r)jpcZSj
zRzLhIK5}PXBBy>M`4b??g)NRCW+c5uqy!<OKWkPkpnoakd6W&7b5Xvf{ExK9BT*08
z7N;n1?plN7bl@P!+w$ZOP;D6g$%r$)Gw1B#R7MH^GO>|-J+cwu6h1=)6fEn-I-g;_
zX}J9};!I#poo*YMKWBEXMtEB{FuaCk1!{v`oFbAlXi~Wdoc0LufF2Zu%Dhk?0cVph
zTKO^|OE`W4J`0Ym6Yysd8I-Dg+J#TZ_edtceEjju@^PFaU#^+1#97<8bLNqg49B=E
z$jzT+7Oc!aNOxPRk2Y+Da*<OTCG(bG*cx+_b@!8Cc!vK;cB(N9Zqi!CvTZ!;CguJ5
z?E}DL#Mao-x;P6lT_<VsMOX|5(bke^5^T8Nl4i<+slWqR2smhT?+={Ds}}Qc6@^sz
z6)jcirF5y4F{d(%P^Jj&GIO90*bkGF)Tauwz)9m<m#$Sq2;B@-VkzW{3!wNxQmW9J
z=6=lYb>#l<(E~jZ1%fW+dRxQu6PMlXWL1OJOMKijg4OCfZ?HKgVSf5C+r(?Gn13pN
zXE<H~-HUr90aawGD5<Gov_Ab1cpMGj8nN~sKtcF4slrd<#ZM+%%e;<Q=c57{MoI6o
z&wF}gfgmO5F)!+R!MyHxZJbjf4UTZT#X9I3?Unt8uq~nlR04BQZe>A<7ox<N<uMqK
zftpkZCnw>eCdvfXZQ5L?MC!ROg^i;*<2k(t5;05fH98E}AMOtxr03oyyxf!yljduD
zMH-=yz!-M!x1}>Sg%q$jBWFK29RU)jcsilp*heXPY&}5=GcQdEZk!0tubiBdazjOG
zO*QF=QEz~WRd08}IQS_qO-a&U8FnU{5Jg8vj)5?5-|R6i>3K&-M>%G3hDdM-DjHI)
zq?kqb@0|BWl0O!jq>OyY`P@U|xrCph3a%I#cjudg%5GUosml0te`FRBj%_1gfEo+@
zV!QY2mJ;MyhA;bOvZ6UC4s+C}{gRXtBVIL9+nwfu*>!^0dDcuf#Flcn-u303%(r}K
zY6VrJ`S%~j-eXT9`d<;9uTgKG%rZg*fkdVNQZ^XP1N%9H{ysWQ_f{e4We94RZn*45
zfD_TVIc`QJgyRTgK1cMm=;4G~)O}b8Pf!(7fTOHO<VOo(0;p9=lw8ose6dcY92<0N
zIn{?CRm{CR(1)6=<oJ5(B!GXeIKaH$=V72-yqYGEbAoS%mU$T`q$|a=!Ds_>n*&<w
ztvT6?odJ|QbPy%n0o<p6R~7&%$P<h8P4&p-hG;{2J{nlR<P8VD>}<w?8rTAEcm#P7
z75E~79H%KX3|FO^nc_$()4R?_(T)9n{UulGdrio;E0|v`gIqw4^9s&R6mF|#pt;HP
zqX~{U&e!1-%8LGJAFOrxhv#=GwaCN(cudKuB(9p8n$UwA@3Zg8%z^)y$(c8>gj5A=
zN8wl>Dw$@$;=3peuP3Zqfl6$VZ?wPVc8T^=h*j-nMkaMZIltf<T?=WYITWLtF^j9?
z*m7L}&X2@PTk+X3a|#$Je9UHhgV5uGVf_wTDF2A&J!VrAMk|~F_;#>ERoV5l6l9C-
z&_v}V+wJrdw26L~i)uvjZ4-N5$bqH~CBuG};kjLkO2F?DsgWFJD#836L=%a!H?GF#
z?IH8rQWGgOzPWDP$Rc92D>VO4e#7#L+ho|v$E7Mqo{xNgohbZVL(f-E4h?~lx;ozF
z+xk9Jqm;^ncP-t8&CB_$G4BJK?Sv>cj}CZNle~XzReygoz2n*hde*wi4_)*>Ju_Bi
z5TY4#vQQ=)0_@L+5;skerI&@gE201nHo12LK@Y~Daq3>An35FRejL0VGOJ-d>IXb^
z+P8VmMD&VgF{+vh<4UH>ED=JU%qn22FOj_)pABO8_{U)KUb>f9CY9~xI%-y>%J}DN
z@!`-{PJJ|}9<LOC6?!f_N`X0NJ-KS~h)J027(ll5e)L+9g>=&7o%NSjfFuJVWq9>>
zxBmcD&A5KdN1*><?CL)p{y(u7y}5lNVVjAp+LAbr?sQs6!hyWV5m1Wjs1qa+xJd)7
z$2N<ez$~@laDxqlAhlB$E>yzOHc%s<bq=awB(Uyql~-_CZWgYWw8Y-TKeLQtH~aO=
z`#;RJ2yiLLBc3FDvu&JO#z_QUh+&Nfu(Vs{FkbTv>Q33oMJ*CTU1KOZ`}D?&BggQn
z9^KzMHsa<7HL`R?mr*5knxWS^7G15cq($OJiIUABLK5Uii0KBf&Vrgj^sy-|Dc(?!
zEE-cY=UezB$3T<71byx|73+`~?R{l7wrX9Z@K+<4I{?!}PuW+<{EOno)9&zFkP50|
zuC{|185wC7TLi|tuaD*<+e7RD)MP_X<}C0aHH$v$TusOPxi$ipU0o7}uZ7GD?fpWN
z+G6R)!Yagw8P^n_C}s5#fwT8bK~Ez*KR{DBpBAD|#i{?IAglu7+S0oPp>O!#joJ(g
z{@$%Ji4pAiBP^@o%-6q$W`p0UzO{9w-(6Y%94FTAupck|nQ~{-%wN>wJAJz`SFSF1
zI6|>;DwA8!2zNOv!w<`WUM|~0h~^G-&}?~Wje(0y+i^ZU<W}ggk;fmnc-|_-o~JB6
z|Eyn)K7;nc97#6rZLMGyOtu{~%h~*2Wj7et&}H@>z}-i(S_Mgn<i4=(Z~-|Q5MsEi
zjQTYoG?*NF<yrc^`GbwHs1}}U>5o*jOU%C-?;_&d1AlSO2IbZY&5;OX30Hv?q6F-1
zMT6wLIF+<(`6R;lk<ggEm(oqXLZAo<(Fo3?Y`dPS`~eif6`&}y%hy0vVzJooa>g;J
z3mr=d%F6j)Os1>0F4f_>xpPFmFwk^iQrz_h=2)gpi=!ngPYI;&FX+Daga^KHU!&bj
zj4jp9t(vtLI2q#M^0{`o-E;IKa_tLH880TH)yvMy20b9_V+N8Ng=IKqqR-6U`FAla
z5Jgh;QS$*btPo|jhy}_R#wm`b&AetA69xPv9^R^kN!7O`h(VJ_wm(7Piq|54CEj+E
zxj|gr(`lQ781C~fbk37O(@+f^;sm9@QykY%AYZ&h@)gh@;VKQht8Z5u!%O6x7e-Pc
z;q!S{gE?`O5LY8qbHiO+2F(y4SbKuu?L<^a>_d3n{b4W35t8Q&{5|MLQcT{qAClqy
zI7a(M95#Gkh(9GkM?(6w!RAd2BMl0l@ZBdKaM^F|ZgYyf@BZsDyl^5pKV6K=cdp9w
z*EaMh>>EfGbS|=vkt<{b8JCHDYPk@`Fl>+99pcoylG{g@*`K)nh3x-|H?3%(yyT1|
za1|NX-vME*6DaH!KgX<&K}#3&k(`5<cU(Nx1X09cG=`!t2m=$<ydOd9N$)?jUmmR-
zf$~SgTW)<j<16E5>*FtshX7T8erk+Wb9*F0v_8^Vr`Zphqh}~R=B;rFN?DZHvX7dO
zF^s#NRS{L{n1G#KM{G~1o(Q~RF_yKUv*eAEvEv`sB6VLEo|T3s(6gMzKifBcBJz~O
zR_4yh=y>7a=h#@kNa4wZ6G%Y}P5omQBT`50>Jaorb^Vdit%6G`5&xWx?8JjUj2F5^
zf&%@=i~iTl><t!6p@HZ=r;^!lukJg=+^m#|Y}<u`a(bXml5CpZyujgQq$|1^{-O0Q
ze|q+PQjZ7sX=CGd*|)$F2%~wkCi)3Oa6c1V3rgz)X$IZ3DNTvyIuL5{eaT#wvnvlM
z2fRI2qWHVabz4Q4F-b7k7~z7_2~pknmw;Bv*6B0a`u=l-kerYhKX-T7UjL>PO@y3@
z-LRPqm@(@5i0SDjOdrF$je%7`#=|~s)pLg}9&#08=*}Gt;^+s1xVTkV)Yj2FEY7sA
ze|nzTj%y_`8V>j@N2=oy{zx@-KR!e0t!KA_zH_kPLwu1a?ld>US<I&zqb=?#44;$p
z?xxe+J_S}=LXi&lZxo46Ev7y+&;6&3{0v?LtJgA0XyiMHOynySC7kP{oTe4<1j*F;
z*uo_87Hw-S?VaoUs&H0@_(1AdoP<5oQa0JU14nM-zy5}Ah6;$RTE;^^z2G@*lJ^mj
z!5&aY0N=7v71&V%7h3g%u9-FlcD-Q({#=Ds*dRxf^6Y0RSM=ier?%~KL~BOgN2td*
zQ`5;Of?<T*NnSv44_`Ng&eA<%H)hjY{E<z#aR-F8l0`#Egx4;u6245tnmRSs1W6o5
zxV7RNuHY{ij#EDL(Zvu~zNcY&(+7|Beqml23<!qM@4TL)n}3qHxk{uH^win)8#w>n
zc^-9>>dx?>=6Be8RCm~i)qYmE<a6>8S>PjiS0-qaWl<6$ODDmQ<;xYHe=ejD@bPzo
z`V5SP--kvm+@Mj?^#VR1jTJT|g(pHRN8zfuE2msO7DlY1tY-V3X4uiGAGGNy=dI!b
zfiJbau?=+mDdR&Iu}s)3i)e@U35oVQV>>>PP+rlw2gX7Y#8`gQG2b0=o$_aVWP|Y<
zF}fMl%tb5-EUHCWnz_C{PSQ84Ogbplm%v>hF8u|_`iK@*Y*@9LITzsa1-}JVJIH^I
z?PID^F74#)S7G~LF)lkM8h7=YZDC<S68|*-m9o8$!&2ngYwWxmef)~@TU#li!bgFj
zd1X(9a7FJg&8px?F5a+p2^yRejZ_+qEv7`7=Mf$!>tPHpogCx`Ia7uC>s;$%Ug|F4
z|6=n<yLt(h&lX4+HRJt=Z;8T5nE;FM>OW|xAVg`c2qTaoo-u0pAoKk%Vw^{#LhkIR
zxv=&~ho~Tu1e>vSu3_V<y<({~C`<`$RCveLH}%-A<Pw!zt%DP7L&YxtB7br!)Y?Rg
z^3~x{5f}@jOO(Lx4~DuN*Zug0T3KK58y+t7$#3}^rAl{9#q&X_xOfzw0=lf_*1m>I
z&vA((nyN(WNT&&LE^5P{**`|V^yKua(qQ!<^vp8JHb2KDI|Zp4LL6#Rn3J(#Won3B
zux;h_rkTc`Gjr`=-ckhB-KHFws#-9Tg_N2t9}nl|j8?!Ej|Da{Yz7bI`MGlLA=<5r
zVKy~>e9&FKUim)Kb_Oa5XHyw2k)5ou1r~-^$+J0<a8KL(W7*xbkwA{OZ!Gj<r1y+1
zZ60GVGh8MV^0e&!e>q0D=_Zd3q&qT?u%=Hfb9s06XHZ`@w?%lU3ok!^ghh!>)PNqd
z;$Q)>FyVpbH`yM;LgK^+;JaYn-hD)DYs?^e^wWL}=r<q9;<X|w-}2LuVM=1DoZgqr
zJvW2r>1)jsCu;ltT_!ZjJA<TEHi#IOQvDxMZygp@xU~-(G$P$d3(|-vQiF7e3|+z?
zC84C`z<|;{A}!q|DGdWkBS?2j$<Q^!z`)42Iqx~Y@4C47f6v~}de*wv9cgOXRKBSN
z3yCSA({L@SZs*TA<#lf|!0@kSj;p!?#eY&h@Pp(X5Q!?9LOhINj0}@Yp9Y2w1F!w!
zbwn50pGZrQ1NKXVH~PoB#M=stGibsK>Z=<Jq1VeXd{`Es0mhx*YByX?&dC8-nY*xn
zYmEw(d=g9Wkl5s^Q%STw4>XD3N&8`qB@mLRQ&g6LUT*`;*g{DW_omP~9?1>LrT&E^
zkxr&rz{Y;oNxxAEG0T_a2KX<oi78+5CB@NEeelQ(kt?)I=+76?Zqe+0NF$l+eyhwu
zPq>goc^5D-^cmx7>}UYtMSA(hzf_9A&wQQBtHbssf0KKdi+=l=ZH4~^SdG%yNb`<T
zPlq6OeGC^t9!9aSs$HiExI^Y6xbvJZh{=;Fe!#C@`n}>wwDvT}i;K;I3+~ujS{(4^
z=nw0D`>GwrrwkY)OIv#VDk#p&Q>Ytdj{B@S2oTSNd484t%^Da2mLZQcx+d{FT`U0?
zJ~w{N<QX7zJ!0@(A}DuE%g!5c=piZzRlLV_JhH>&;pJT_>{ul}_TApIy1C_85iV3w
zeWPqAL^06x7)C)EqZY4$*FarS`imGVR#2G8xpT3BngG)W0zBkX;9CIL))_=Q=q!qQ
z31oFyD;eADv{hXVmG1Me%MarE-*KdFCPH&p)^)CYVE&Idc39QR!Qb-X5m@`M7aq7K
zRB|I*#db>ijKGsSGzyswRgamft$zeLy!zdlp#rf;3$hqq%i?Le(aQlCZsTlgr|84m
zu#^EFBTeX|2bpaFVFDYzFVFyj_7d?Y4JC|9#3>|jI0ns^=?)^^RW=hih6QQf(<6eK
zY59Gxf1F$VHrr}S`w_~4oq>|b@y%POCq>V`aAab&HuxMx6df*zut&;a9{yr|c~Y0}
z2lNS5`SvEi1oK&x-~})Kd2%84nV8B}^m9pZ%X=np;3m`l<&1wUI?hYq@t(&IT{2==
zKSyXMr>O|%8TYv&M@Wx?wS8_0Osz)C_!~7Nz1Cyl+E>19{VboW!Z#n*X7<k_eZJn)
zjscjkQ*@eM4x`#e<W(~QS*o>`klK_N&2GRKUz9N^V5Ih5`_L|o@!M_3+)HQ*Q=%>m
zm`$vj6w^L#m}<QZ*R-%751v=n^(ruZGXLm<!EbGXo4w%pjxbzWev-P>tmGdvNhp`I
z0so8gu|I#x9&0n#k{PQx5EniDQ$9T8c{WKP{93KeF352|?Sxh;F<zcWa9c@`QS4W#
zdt$9{G!$E2Y~XMC4e6j*KOCYhT0$|UGxo~il>1k3T?hw0-wQT2#YT<@<Vt^PbrTCM
zYV6^&XJ-;EuQ;%mB+wSYtON@57}KRIDuB3Tnn^UjOs6hAcPn9-%)-P=J5D=Ow*E28
zeQlC@rh8vg$3ya~(sG5LZoJQWb&XoRm4pRgKu62oZVnONU-sG`%B_C2X!{M2C54D-
zwjV1?sleu|s)0nTcFWP-%X@-6s_#@gbuAJ;*OnJ~_%r4&hddni@)5I<VFA1)OkA?Z
zzg@OoC+ca*gK}y=0qh9_)W-+XUhg5S!Rl}`LiIKCXT2}OG^ddvv1aS)Lxa(B6dw&;
zI%|i28_&)KQJU)Mrw#_CD|t|?mKdifdp}oNlLOtRpj<m%c$5c3FT}g)jDdFvgX5j>
z5)O;e2`>5ScO+l3f5m@MC`8Mn`N2`tdSlj@DIgHQoM<tqWx11nu1>SsjGIP}0EOrm
z-EL7AEzNyTAoka0ZB%Iu*vy9kn}(*@);y8`pdq;l6EmtWNtv@vF+T!U#-xsgi9;1>
z<I3e@!URXoC;NPtD)pXoySbc2(@(7osS6#-4)GwRe2yrU_g)R_>v3p@z6m%H<4@b*
z#lHg77xpU(R#`hOrE%MUzL|?o$TS4t)eCIFy9&Y_2R6k>sl43<q#xdphFyOZt0DT<
zDTA`cj~&K7yPgpqHzsB0|K9(IZyGww^O%hmeY<cLF8oP!6VsjS%Fa#aCd&|`w7BV+
zA5DQgLC+R^WNC&I!5)Wow~2ja%M_%@%Mv4eM7Hwg$K+>5p8QhqHFou8go@v|8A)p^
zqGfkV^1CswB>i|qLobF4XqIkJCF|Z2k*=!L8@C9$LRn1<q#qiPUv^4vWdyGd0TUz~
zC|pMM+nCZE%wA&$PH3_p@U@6SkPeByD4%K{eed-Cx5g}C&`%c}f7$mLH6|9@_eB;i
zCB$=wuaoJeppth3+S)vQ=RF_M#>J@D$}`qz_5nohhxe4-@s&TR4+-Q(JZ|>L28K{c
zKNcQY@koU^26w1zTF)MU*AOq0CaS1?o?|a82kfeJKa_dC6?A?LPmvJ$$p9;yRngQQ
zjHXzntkH@;OVtKbt)2Mz;;Jdm#|CRWxhOBHt;X5Kgj1^@@HJE*_Z5az>BJ5a#Hn^V
zrG_dUsruT*04I2fD|fn!$-i5`gVM)MZ%)7dsuVu=UKjwyY;T<VKLT=vFy6d;?O(Nv
zPeawoluHI2Zc-h)U4w^?`&~Z%k|HCxt3fQqmP1VJerUw(F|2%>K5W1a278^7c@K&G
zti{){b&4tN6X7Sc|27J*Q;6+<yE3`J3S67J!;4n?UtS6iIJUZXi@nC6%utPm4<Foc
z_PhYk(hM28i9our!A+{p%=u;5(N97j00CCsVOMr79gZ1CDD5Ct;J1WmD&W)Q*G_19
z<dGDm3Q^og!lm01QaE7`po2QtS56<foj${8&SH;7<xDNb8cN{ye>TjHlcl#3Itdf*
zIrFwG3VZmOx%w=*UdS9_5o>Ph{iL}foVZUBB}PivP2!M{epsiCA_G(AZR@*G&2R5$
zihOHt=hS-33UvtqF{fJx31ARkdcW)|wf2)=0w#7Mk2M>#XBe;*dF$_<s~+-IVB1cI
z1<srk@oZka++<+`oI9%@9~T_H`PF-4QK(ZkYSRt4U*u|VIY<Rsr~0p=iDQ>}$>eNX
zTOVE5rw)nSIe)C*Sx5Q%xr``rm5YO(Lne&-t8s;dwF6iH$Q^z>ZI<Ak39MUJZw3aT
zU9os4C_h@VIBhWZg+&Za#UUKvRQeic3(|5v^#zy<H7&kWjQ|l}(ZGNf;1Ew53(#W6
z*saw$J_Dxat_<ie#8uyL_gyM2qB3I)_3`hGDhAy)MLw)((%c<eElUy~)fH!9LguZR
zXlf}-Pecxmb;yUTl7|D4mVd|ySIO~L0j?z({wfLnDD(X&r!g<Jh>3<$+JaO`?#IGx
zBX#z9rD$<N4&}4ZGJNh#0%<~lo_~_&m3Z=1N_3~C&LUfh&Se5u05E%;kbQ+0O|R!*
z;Gn?B%iU%?`Q##~ov?o?T_c(GMoXt8X?DZd?U+bJ{J3|I-y|o7T;tgA0~tm;mk4*W
zHP9`^Z0E#O0b8-xHp%~hU1ykS&Ij>-vDPQb>@;<E&)MHuuDo|bc}w~|o6^@W{i~EV
zehfZ|{gFQH7_uj;O|4+3WKpD%$@;o=;dnF*zsUiL1x+Rbv7^=5AQ;pT-m&DhK>YX|
zz`w0UwH<C{UfJyOytK#yJnUQu&_F`M#!M0?Is*NbkVRJ3DsqvoR2Acc;50oX$1C+G
z_)1aTi@+E8_K%JbFjM)`K=)QXT%+Rl6cblj9W=F`fT7-0Z0Vy`ekjCjI--MV$K+*8
zdWam~i9Lx-B$VoO_%pezya>zfiz@oK9QyqL`gs__^#bOKkL#2fXTJYaB$Q%2q8B}C
z&u!oNh2na?I;5K7f_q1i6W>A%Ir^>sP2<<UsTQ|>_wthZnY0Ng`!|zfHx5H?{KIT;
zG63Z9NIAdlhJWOvDyqq$BpfOJmFDIQ8hEZ{lLO%d_BsNxW_iOGIk~yFZ<wCS-08Jj
zhE<SQcS}IzNsZec6gvGfOnb9%BF;B)y#0-!q@28KBl>GpnSdjInG_*aP58p@+*>gQ
z`DoMG58cl-`yL589qO38_KQl<s98RH`$6cyLV~;_tp!eDu$^~?Qce{9^5O{$z^q<R
z`?+2n_mOMN>Pc4sFsS>B-BC@EOc}tU%k2leuOPhk`J>Lsi6ycqx0hq~&!&F+$$&UE
zWFc(+|304Yo&Z_GAhxM#LZy&4u0Op8^E~3l{G2+zv2>;t1X5AMs>(9H1|jOz@3=!a
zjaCyB)<j~b)GQyz2`8KvJ!;b$&sPx3jf^`hiAb(va@%N?W%ZDT0zp}kLQ4;3|L!Ol
z{fK7ei-0&$u#){53;Qz_x;l0nB(z_uBoE~{dCEa1E+te%_`9T0KPv5Dqr|_~7gF2A
z=FFRgN-IQC5&JjW$nW2Yo_`d7#1MDLsAtwsnboGze~})i*plH?`kL>>GQh6<02ieZ
zs5f<Zs<5JyDBB3oi=wg)JhBW)R=#iN97qNNE4+{ug=SVZnvk3kiTX$$xsuJn2|sh%
zJHXV}4ut9=z0PuKoRxlEK4~69Q2t&uBAbrlXQ6ivve6x(*ZUTZ2$BlM_X;P&RHbLT
znMZMdPH+*^SAF@`<nF~CR?P)zM~55Drlsgj>`t&LTz!NMa5lc$-`i30Q>CAFnhz>0
zpSC<csXup~<M(gnuJ!M*5;An24f?$>KQ_Jnv1{{_oLcEIM~tYUmA~=>K>{xRSf12a
zl~fdgo8UG|${uA}WAE@7+3f^H&_RP)p>F1+nH*v&;|?6#Pl?_=bTcWQ>9P^e8@ExY
z?+GG7+%^rQkPm16sl{8DDLm<UDY~5dGX}zS=9*d)Sj<S+kX{W=zunrfU9HzjurT-O
zvALywD<>Ka$|ry(5c@gqD!q*x_jvmpUBPDm>QTMY6+ndQCNBA!i!-aWiL`g<%gf^9
zp&84Mi3hk7q+Q}vw+rvX+5vYYzF*tF4C-hnjN!G&XYXgVznDacuPQ!80-3X&!0jz9
z7)LH<i@TwCK?1ruvj*sCHb}!1NXM754QC}@Gt=*Sv=rxxoz=&LG{<4zmtpX-bqo0%
z0~dBi#Lnt?OE5R49nsWU&*d|6^eV0Z(o{&zz>8VnH-Bv18}qGif|!i4mW;0KG*jGx
ziiE6?Vf>wP*e9DW`obBW!_5q`ZX}@dJpupkKKOryOygYWMEbl34XroBe%m`8hD_|s
z8%IS*1w%xXT6=6ldaM*e=VCOzGZN&pirVLJxM?Yv3qp~xQpfqu;!JPEnBEi&hV|RF
zEWH_x-d(HBI>_IJ9zU;(8!HKiR8i|-Z*{u*EZRG{^o0Rt(*e70-_JJCzTSK0(UjpD
ze&;BZbMv-Ma`9=n_ExfJrE2GWC#JmygU*asmU}H65lhd)YwU{g`@D>-+X}U=7=Fda
zd(@+Ae;ASAR{77jKis5OkmQTHj~G!Wk}2^T-@uQ^lOLjvpLj<2Y`uX5*Sz{&Aj;K*
z3fG$S`Sco4NAQ9y0I}nvvxCH^8v;rutcl6seaR_^hLV#mUvMx_bpUSHB08pY6>n-`
z#eGm{?etZaa1VQWrQy9N3~d1`D&xrQ18{m2*#7BPKgHZphrGXr5BHWu?-NS*{;6+>
zi%4(KU^mH)OxwO)uR?0tVtRD`R~az86vjo_n1+$yj<od!)9aL^(jGH~8@_eB@ryF~
zqws>yEph3Er_0H=53gURi{3Bzx~{8OK4Ez>>5%%|Ryu}Nf|ZcXonaY=5}2J|$er2F
zopH#Wa<F{1t8CgPS}<YHRlnnrxwBG#;>qW1th#+RHt!Q_2mxP6NF~Q?i&JvRMvkL5
zlig3w@)z<QZ;cOeHw4$=DR$0Q^^FBrRtkG~dy~rDY80?K<Bhv^CCb${PAnXSKEAEh
z7~a)M`msS)*ZNa$mApBeZh~o-0(J4naB^Cp!~099jB3ETb|gMyX#D4*XAyv51wCNQ
z`aN%-e4kWOQ%{z6pSi1q_A7Vc&{NZX?HHasfb;jj^1!|^K_b2-CpqOt({Xf9dqHX0
z^$lQPnD;<5>4lCrlOyjrU`72xX!L@zaACqlUXo#*ED2!Wn`lBO4mohb?EwE4-|ziR
z$ttPUw|O6$tTXE^)#6F}mh>hhp43AjFv6AwA|b3eyBW$9(R*+GK6K5Ifi~In3OdI8
zyvRX1=Go@;(*o?&pm3|^^N_bu`kza9Q_z|>fx)n)nRDDEZahvB*sS!Z6A-ug54Y*^
zGfv$vDtlU#Ac_SPTes`=^j?QMNFRjoExb~-K7Aw(X9@;Lb8o~M#-`>tOxvam4U~5l
zZ7H>^%sKi?4?yxI?`(6y=u_@9uKHt-%<0IO$f|rAZ2Z6{<VuJeWgqs<qZoGO<-776
z$Wj-csSa;%OipV#^AZrHC&hB)v#8KtOpRV^w`T@@&G^h8OXT=e)X+rKY0p?QTgo7&
zkoeBmH|$>?9Dk4PC!cBY3(8ZwS)qom5GeEaGp4tR+Hm#y&$!;G&~6;HL;f&S9^Te#
z(Iy=OkX?O2ntiSjCU><!AO=9$P_riIP@#;mXymEnQi|zYc}WFUH4iNXfMmm{qrMIV
zn7{$Vl-A$k(L_RAY}bA1m{-(#FFKUP2Q&xA3t#pet^rGmWN$sCuOZf{W<a1}CLrHa
zew**$+R6H(7?Ryco2+55DZ^$iO$?W)=nX=b=Ov?hRFycB4wm%axqSQaR+afacF*G6
z`9Q>S><r_MA)nG=eiZXH42VHLWB~F=B8n2U^%b^plccigb+g$*K8#g@zJ%d1sS&a}
zx{Zzb_KPC{XHVi!lE~Exv`ZC=68)sMa}zb5wmjJNfKI~ge@t#mzq!nOOTr<SSEEkk
zubwaeV=H)OKd5+y56?fehb6LeP(hl%mMh*T^bs7ALUurfFCFPih5mKPdB-BcQMnuB
zc8_shx!eB{Lx>y&1!>_#f+%l3pGHD4k?gCD&x#%2XNa>$*c>nnu{#r6A5D)!fxK86
zy1poz$B|In4#^XqIHu5n^}gtDlu<EbE&Ak-?)}Sq$*+2~D!BMP6e8^|TNIl2awXuh
zB7#5u#el8XV20!erLSC!pmFO`S=l7Z+R)mtjge8{h};YB`a~;W7}350Xbfnwd9h}j
z(OhxQ6(W4Fzl%z^qJ2FTI+XSAs1MQ_nSDa<&O{HeQ=LzuBSLH{IVSW?`_0ngz$DY6
zbMgtpb!H!aKh3iRtj}N}H-ySIqh?W68Dvhd(wUI$ur48A*uc~KDG5W&VocA~aLSzx
zo^{tnxt0H(QBgatWjT?$O{`%k@8jhXAPu;;+$0}ZgelD>S60ln`bvRT4!0DcF{%;)
z;T;t~=T6?N*%W0LGz97Y2oHGkXz8==_u5yO4%^M*rnAeLM;0vM;x2G+4Ax}6#%4K*
z{4OWe&PIDSnMbYi{=dmPkr#Xu4SmC{hmktgy}1qTtOf@0)Zag@z}YW^N0mNv?#og|
z0tu~6>|nY-wG*&TSSK2^=FbLhgvaD3kh__*n-gu>+py{i{?0NTy1?D0%;kLo;+RzA
z@IipwtX@(=-S>)#M3oQtf<(k&`rYB393j?YShWy{>Yb(cGzXUkJX}1ntf9n!bc|B1
z)G@Ig90Y-=AywOM#jHs^XsuJqN;#>IR7E9Zo|>E&hcCARjHw9f(VM)4Lt@iM>xG4d
z_?elR#C-iUk;V5}^}Aj=lx&pC?>`*~5uD*o)L)B#SA{1<S2E%zoNFyN)87|Z$}#Ws
z_osPQJKLoGoz#Qxb+&_aHQGX_4NeQyp1X#}l8ykwKTuB@f4emXq^D2#Dq&imNi~IU
zUV{`iJ(;EQJNwgK_1yclOHY7yvvz3+(n3n7w%cv{r3T!!EuqV*y^Ty|w^Wm7AekI}
z{MdQ%dCd_w_#k(MRg9rvL;#_R%BnDLHT2KghUMMYKEbfy4q_P|E3}D-tH0<71(bfY
zTRDojR~sV+EMhvts=Or8zdBw&k^k9nb`W<7pyM#}8A8szeC?G82W`sNEScxW*v6W-
zyF3O;`<qJsR2Yq7e~!o8#6q$QuCAx|eQZ*w9}mkMiTstT!qUtLLaALn*V=I-eah?J
z%Ah9S&EoSeOjTZP*=-K>VBsoif&1<E0G8H7r6MJMl-0(ao5TK0te=@k`k=`g^0x1L
zVP369#Q9w7;im_7!zxY)g90xq?o`FwwHeVi4B9P{X`_*9{%ioIgdFQ{qxj?Qk;Qo&
zr3O=)9psdK9F*;20ukXn2u&qd-PsIHg$u|3JEokFruyrVX*ws#A`5f_@?dZy#d#u>
z?Vd{9k?}8>xD=2w#?c!~aQ@U%p9#${GZ~_8-P>s|)_d>7W{mam;Y%~!Na#hCSs-8?
zWXTW){PB5T{roY1RvBlVUJ=XKqh;8<+92FIY5XqqD}%W~I5ktznoGrN3d12EI#QD%
z!Y&Su18Tg$=P^1{uY3Y`b%%E5g93-7=!)TKX;WSjO)W>|J=a|V%zyJ+#-plBnNk9;
z3E~_wqH2Ou+55ZYreVQyf>$CtWlV3z0dMTR0P+Ll(}xF{z8<}%mu5tMW$IYK@KYH&
z_k!aIhlaGUxOMNV?!@hQ<W&=K!kf}+B5RqB5pwu~P%n4Eufcds$@AE$)@asKcP3aj
z1DNm6CbHE`ku7?jw_vkR86-p_l+rhTTccB&g8lI!HSAFR2O*yn+pA{!P7wbN7aDb7
z9Fo;7t2_$>Ri&y%3$BKs{93dw-+U|i$h#3fkn=M!{uP59v@YUDLQXLXW>Po!WI*Gw
zNb$Nr{>fd+UBR9I=F7QrvXFGlW|xdeSpfm8xE@n;Kic}Ay41UVd(YqQ@6I;-E{ZCF
zEk3Xa(t3u|T`Zokub!F;*6xP~r#EytKh3YM@??2CBW|s!U*LTqYrOcKEPGz!4P@)>
z(C%VTGDKR%FfJnExq9#`*=IoJOKcCJ?NBcRg?H5@)3wC)!%4%*TvHb%Q7p9K(u~2W
zvbQW)_|*}GwnFL4#AI>(1Q~bX!v4!Q{YFxt`eqKr>Y`qi$J<9h6^mMETyc2z_MlDr
z?cp;*G?z-mLjUtP-Tmz+wA6<RKD?1WuZl{fh~0hfybus3%@xSQ#P7Q6TNHEMW=XrH
zt%6lfl;)3euu!XxE@_i{@gyL=q)oItzXQnn>h+=d$r~@cvz>|HRA8e--)}Ml7U5EP
zT(+mjrqHt7;ywvu!J`oW5^?0?c3*Cu7)CDV;zWKZOJORQcPB{)>Z_x&o4|#3ZxL4A
z{ojw=wT1b)JmCp!H;#AAh%eKU-VRCt;c1?HezyYcGs6k#0`oejy=|a$r6ES`o!n<v
zxfJs<vL9DSk@pxJDu2dJ#R4Lu>bkzL`bw=3f8iJTs-ZP^>Y?A&+<5H$xKEyWH~twN
zlm3I24J5d$Y!m)tzg(_wiqt}}zL}t&;CsK+XEp(f`?*sS!7lfkC|alDG+N2rZD11o
zP(qTa7$98LD32GToeF}rh(7;Cd($!;x31vm=w5a&)y-^nq9s<MXx~j+UpI%7$RD1z
zgxWuh7a;u|V&6uw@I`bmI&Hqk89wuilv04y!_ku-A;h`2j_K=-uZBB=mHs+pmK<kg
zP>F=J<vC|{w`Mr9QJ%%C$r4{`i|X1PqJ779Jyh^y88;*!xVFmHNd1V2gHJ9`d=g?^
zu?dl52~w}>nGK^%l?lk0r7xq;ck_@;$}xhZ;3IS=1@IB?i~7*$|4XPFa?M?syPGoe
zq<n6^sidQ3ZhDrw<JJ+oKEc#HTib!b?CXYy{Pszci>lo)$8u>Z4(=joaKLjMD?3Yw
zyq!|vFj0a^%8^InZh}{?r-XN-RhXyDzwd+a$K`KbV@ZY0Q3U^L*0?4Ojddy{h;_1$
zj6YWoSavDvVR~SU<Hr_0tN4H=<-py5tFK>~#-x2p0Fe&BR(I<nlep(AEk1ViOF+c>
zU@^EGZBRaA)iaMqBDeADpgse@h1#av<19$RR*`l%VI^$1!rGD9-BjxZ;#pR&COsBj
zq3e(Z8M4wx4wG8H<g*G=4O*gSR}_#GV?%NwdpzUAz@BIF$60zBAm!-?_lwn9i$is{
zU=Wv$)Cj!4lqC3Ey@8WjBR01@D<zog7@R8TDWP*TBEYXWJ;57&vGE8<x;G*LbmVsb
z%L7XkI6D|N`u5b*Bn@sWwC7}g#sv%r*H0K@w3#Jb{xNx1VeeSF94z^MczBfLus1Ny
zbwa0v{iv$@=OJY%d^Uqm+bo_7lb*3(B4+p}=(u$)rslh7xs>#$>tv^=&gK_<AP4iX
z-ZGU`@p&~<PZBeExy7HilG*%;d(+<O#&O(Mzu%#uy8n-zaO@^^PHFzqx-|a*{QVVO
z;L`C?ThTh0K@PGv(R93C<BTH^G;JVZZg5s??BZKyiypuWme2l7B0Rpa$$CO8e0JRM
zfmlJ7uC!PUI3~iCyLHEdI+iavEh>}X<}<2#`%jrH^;J1#sWRatP{|(9=SLg1ZX)R$
zV;0bnZLz|6Jup<GwfK>&@&Ya(8;D)EisI!FQ5${vQN4;XDB!9@r_&*k=Z<|k=LEa*
zJi<M`uRjQ)som{=XUzeVcJDN?a!zpXzcoNCpYTV?zxhQn<C_cs!(d1UP-}~sY(+42
z&L2G1e@r*NdHwyiqcYvOk~_3LMywJXAMoX@#?1jm!id%}4}gCsGf3ic{daC&10<!_
z#KcCq-mqXhm@PcscXu8*`BXUxwI9qrI$(aSL+@ZxG@Gx#d6&7>^k_6gulwvteM$bm
zoH<dz#j_320-2z(W;^y>m29%Nz}}?GpequVX;I1509%(+z<_fd*e3P8nRUC|K1Z&@
zALr##y0l+X8amC!u=3Wk5qG^8>+P}A*Y7ji<<nq$+7V5ptBeP+bfoWKF;VI^ZwFg?
z&^N~_1&9mD@gHzKMKc{QYwySu&Rt&pvrm@-Py@a_MZWRFeA4<tQha4L>~cv7{mUx7
z)PG~f%~=p4C=@~<8}zj#hfN3W#+4qrnrjcECqfAz6GHuKQ^5w{Dow<e{eRPppO{r+
zJS1TO7J9(qxI7@@u6hrz+qI2i1-h79Y_eCDx&7xW=60po%}-461U^;wZQ6J^G71ej
ztP##s=LgO&@|C|AvbIKjJr|vBZ5aj3iW4TZqA0#BhwxZGL~&E;wOw<BgsOi0Qc)>9
z<dhMFIRib}2LS<6k7UrD!`*<NY#5;UmnmQV%2rp^6=@-duR0q3d1a=U`$h9L_nvdh
zc9DGZx{Dl)3F5aNxK{6ruE310@6m4WzvU{7O}I71bAvs@UXS)D!4zWs#p@vvC?6UK
z-4SsfgudfjF;P23C#C$$F?0^aY5c<%lu%oJ5K|$%zA_;nRjUKhpNuy2%9Dc?WMy%p
z>bi9g^8;vU3h75i-h8wmQN@G7GLeeliYrbXSYS{NnC>jmE1M+~95-5wP1hI|Fh7wr
z{ci$bqQIv&w*XUWFd(V`k9cK%0&MvKKL6G&ua6p^^Xt!+Sy`seZchuprsPu)k(U>-
zARVOiqdEgtas5)mAIW8He(>H_sgN%?Gun<JHo04R7UozqAPOdY)w!M^IsXhS(Vp!`
z{MJyS5O#GV1yg19wV9vyjM@RJW~hm5O3Lfx_3o-{8uVfUe+8hC{`u`IkAMBjNDuC$
zM&SBkO{2{K6(Cr>dlz6>7OiFx{)_zB`?2cpZ~8kKoG2&<NWDpQai4l(kJbvrWDc(@
z(Tni4{VfxT;?8AjHm7fX)igJlCn1d7Cd+|(YSa`X{R_6*SmNQS>klVK1}QJ?1zJNo
za(x1!WyQs&=Mat^Kdwp)NCN>!<<kpec&OVtX+)@UixV$39BsFbe{7vTBszP~iPAXE
z+HvQXuZkHLcyOc_=}QM#4T}PQnv#D!T4nC1oJIgYVV^JMnNrI4Een7<wpF&tZJHK_
zJO6v=Hq|>q!ISWviV8gd<T)7O_SgglK`V^Tq@UBJI^*X9h(|e@XbT6wgpkQT%aC;-
zmyx<WkAVaGE%2Y4r755c;IEVbm_VRYBXiHBrVKJ5{XNpU{i-oFk-@rD`J)ow$5)64
z=RPZ^)z=Zn@ypv|$L{3lPetp84O`1iN8<uHPfe@6mYDQYf9HjPbnN-bj@NLTNqm^j
zdtkb)JUR$H?ddTQLWQvuP7G6~u^?KtZ-G5Gl|DRG#gmOhmkc5@97XWj&Gf3&>cJEk
zE8>Tf6RmvcEG3wT5aLKlZ@b5hS6wu9=!olHf>vj$O1WsylXOjmp$RJAtZjT)84;{3
z!oM~L{H%W$JUKX%6!?E-lA-8c6T=VOT@y<n5V6O(wXP^{$2of7Uj6EcuiHtH_hXme
zwGpA-e8gB932mHz3-rp21WGl;VKSMp9`F@d*W5)Ox7MtV3XHxl8#?zfH#Y}_vLYdW
z<^c|+(vDYy48rzt0Cp;&M9d(6%$QG%Xk)9T-Hv%Uy^fG|`QrKN4iC0em3~eZ2QZ9&
z0vr6~<HtMSxLijm`~Fnt!|C^zrKINp@HXE{kD2~7s>HuL3Bs{qy?r{ikhgv2F6*Dy
z1rw*LijBae=FI5)f+^ued-y5u%M>;z_+WTdx3+5|x(CPnlx;u-5u>X4D19(EfgSj_
zeM(cvt8=*H`Exh6=~B-SufBrRft$gDDW-PXawp`tI`wxA6BH_x7=Nz0zayrMwEda2
zm<}?}h8FsO>2v@8oBOVF72$oO!?9S;X^B&zJpr`TQZglAAEwRWQeW9^vb&;KGH62O
zfY~-I@K#psRB^~jsSDMF7iUnYJr(}iEj%<96a~a>GC=+3$!i`&%}dW8C)9jD-gCa|
z;*>%Lc7}vJX%AqUDsmKO5U|RF$OJqWjJZ>eRuRnfBj!gDa6Bj0Fe<iAb8Oi@xiPx{
zbO1&EF^Drcq7US91k}ZvwDp^l(U7VM(%fa<O%Oja$r*jz;qHhK7XItWgq(=HNX>y4
zFHCrjooPQ`xTHm&)_yiBS|(gg`F&)?cdid+(`aIe9v>{l*nHVSD|4ba_d%RHz-PB?
znmHCxZODZO`v4Q-fip+%l<`yk2HQmye=7yg)Cx<AWmg7(5+qrpp*NGGCBxh2W6Mip
zo8)-C&t5ZU`qG{$73W{(w3K~q4DCxNx!CN5A1y{*yqLuMmWT(F(#f2i0HLFPF~QuM
zQx=d9<x=-Pib$jF#+6S})ixk^9=><kcl_DqcSW-myTI7I(M!v=-eP;0Z|b5&f#yag
zee<3@n#`DRou5sMN`4<-A{z;HL`;v52kB`^;sard4(}0>dJSERTyZ_$%)ROJCZ~33
zbOM1ZHFjJz<3O)zUVwAtm%V?Uy^+enW`CQoV-T+@avNF~0AaXKpzM*Aw_W>@zGC;w
z&{1N=uDw|`xJA}HSy`4CtKIz`zO%i=+648>*TF(op(1%GmAl*sEh6lgp;!-d==|5O
z0Oz@ukM&i?BNsiDt?#s`ZQXs7(f<XMGFtoqBX+Ns-#cARy_k8BSj}t3is6hNXrtEe
zKdE$j?l#eCzd%)bEF4;Qso(!}Wb#Z{`mqQpHp0WGGXD8v3S4@jaGp=Q%)pDwiPaC$
ziB~;@=246WWDNX39*S1_waoHD3k?^4Maftil@d(w@<o)U*Ei33H-7-r@Ew_;xR19=
zc5a=QMd$9%`$=OB4XAv6wjKr&8L`Jause*SBzwiF0G3@z$}^8@sN7vB(xThZ8Ovhv
zoKacawGQ*CAwmF#qP?G6L9vwR)4CNMc@fnuu1=8EXL?E$NrFss7-%{rOUYlx6g(oy
zv!fiTpf<VUCv%{d*gnJOxQm0xac9zfQ76w-p$ZCEL+X^2SHz~R#lO+obghb+7RcQH
z?%6b=GO8Lk*u7h`p1iyL-)`^ZZ!-|!vj@mMTHo9hGQY~)hJQ9xj~(S53cOd6@tXqn
z?T9c?4XsguNFr23iPpn|q%mJfs+AQ*o2ux7L`J+xQ<va!y}BX44^>Y@TcO+pg9{B)
zO@A(DYI5zEdyOQN&2d~-tm=7?MY=owxfBg8tw$H_Tb5P5#GJNfakHeB%=bv-0e7?8
zesJbfdZDhO3Lz^N<(7v!a%_<~g9xPF*;ja!VKW`b*`GeLmvps~2E|zq;Q-tLY`kF{
zNZf4>q_RFFbxqlCv{)ShN9U$w9)|j)gd#c_MELhYOUOf&dG0DN^C^JymbZ?hrX%Gu
z@es+JsQOfyw}Lx6j;mg+J?-42Q0z-e?0ZM#<&)lZqW?B+T?OI8r>%Z|0oE0jm!t$g
zwL@^MbotxmUo||F<_FiZ@FtF#mwVf$YzpXgJ?CLX`+Qs{$A+Gi&CDM{QuKB%R~9M0
zopJUv;z48ScA?U_Ngq3YKfjZrwRrdLDq%Poc||tdZ#qCIW`W0Wl6bvwmTR!Dkj8Dd
zxZBe_722%YMmR0G&-waKYjX!pm1=Q0;i|R`=MVeQ+}6HGmv)P&&_TmUa^kcf#GfkO
z8x*^?fZyvB+rOf(2!<2;nM?{`YX&iTN}y+kL{Nm!y{IgpvSa-z<*R6pYbOvH?rgLt
zCr*?pm*RXFeMPp4!r}--zbTOlcA|BF8;%`sT_OroJmMs}ON<SJYBEY2+K7)nj~cfV
z>v=iVAe_kosg;EASdIk!7vi#9^N{xYa{@iHZ+JGW$Pp^L>M9v%WLw&X&#%B(e{Hj1
zn$dP_9<sOekO%qX^*Ueq-|tITZ}L#3&atUjhmg^tepq@H{e2gaIQwPR%V~I5`?OoC
z&J<Dk&v83uz`lgE_4QHbbfJ&Oy0@!O27`%1sB$^|n#4amu0sx4x~I+5LBLz}(bv%?
zH)k-OnhF}^IgIWYkz_^4-p8Y$3G(~hD|GJv4>Q~3$s9u<D@gZ2A7GQl8^fh)RULQN
zg)#9FFbhhN;^hQaxbimb_8Fy-pKpxrsNRN54h4!N<r41`n|FK4>$!)|mj(~cqL{F1
zJFo&2dvjKn?g4@u3Bb!i0cu26?}KUedg0VXJ~Wxdk6>j?SPwwnG(eLJo~n)GLj7Oc
zbuWQQ%Ee{3Ht>0E`zrgrFC4^|TB?mdt;8A`(P57DQJ3$Y1%#<i53A1ZGY9nxYtF-L
zl5%?|KLCTpVSHW?gz+bNSl8^v0l^&z68$ztV&;v;jYlGLtjW|X<dNS|mQ!rRc-05b
zpX(p;f9gi_XFgURojBWO<C_l>p`pB3Rg(t&qRHtbr^1R4v5~yd12eOx!eoR~YuH+(
zq(ra7ZO@^rRL{*>^$k)j#TnR%jr?p^E}AG}l@{}H+nZP)Ni@jOgFd>yZ+vX?vMzEA
z+5*m}y<~dQ9Zl?W;chk?4MGF`lD%~bLoQrtn^p^hl3<|<RBn~iK#=2wr1DzJ{4zWy
z@F&7+v#%`wiBhLl4uO~`u4LzcZKF~uy&!Xe?Q-&Ll03~}|NFLRzP+zmID6Od4Dd;`
z1lM`h!KRGgEd^`+nF_kfP-90mnLlqqn+q0C+uilA45dNMX<)60u^yce$+&;mE)Eod
zKk~j0<$dh!SYd3-%M!8FkW>k^d7lTc&S${XBC_1|nF3W8fzdO&uc*ouTJwTvUX<|=
z@P9j7cUdRz^m;~YA!r^}t)lQyd3&?Y%{NkI$~Jzb5toq>l>KO#kECki7v)~W5Bf0Q
z6I+<ssGj(u03OV2^3sYD>m!7y<%J1qfP-YAvUHh7Dm%6xe%Kd43YPW00w60}<k?5I
zgIl(QO=HFek^8;y;QDq@1s|jZf=cW5FAG57(0O3mOfYSlo&0|%pnDPuxfD>d;f@(Y
zL-JW-x#F9uE!_erKoIk~#7}c+FxhqAdcpH@@~$6V8%8x6d!lggA?HzT>wjs}Mn!A9
z(udc9es*^rnn02+4iWSs(~DeB%JRzPZ65r&<;c|}{!5HVm|cvsDOM^j#x6siMPiZR
z@4e^mrSjQhhIA<L?q}!_$}3U4&arh!CSbCE&(zY=+L9;6&-LVNv*N?SH6pM(x+LP;
zx1F+I5krT4kx}&V@$n*D>#kUO%<`d@A5Xlw??6Jw<_4zUi55$@MS$e?w>>XIgR+|8
z*F_Cx__tZ`jws$#u>R3mpkobxS|H{yo95A$08uzu^F6X0U9t!HCz{GRHnSOSZax?K
zj>;cxQbTeb)6Rhx8l>*`nI5cr`FU2@Arq1=LMF*WL0WBK!@uGCXKtcdY6vQ<y#(S4
z?Cdwrn8S|}hAz=Xy_A&MnB;r!M3*!)$bt)?#p2FJ59z}s-US47kRU*;|CabrAruNk
z92=<<dktgTo`E@Lg1G(MD8%TF7o;^q<Vx)CJxb0c^`rTT?C$cB020gZtrQ<nr}~9p
zs>1bM#2I~nQ-@vJ^jwbBt%pLOWM79KiM;z)xsxxaAc7Y4e+E~?@mRQ&ke5C-T*1ug
z0><R;QBLfIhQE4#`jkz9sJs#0pyp^0$<LR2>5bF_XO4y-DXoUpm!6;pjZ1^-Q3V*n
zNJi)qRPOyxHgMa_DQZBw&E7oIxfR}ZbQLnR>xigZ8GOfH6~op^DucR9j_|tsPLg@W
zXUX%ClQXLtQlud~pd-nXP`$+{5JDo$@t=pi#5&ePE>Q!q6)ruvji?A%@ycL<;5%vS
zVxz~!?jlyGbqtJ%4Vxm;FZ^lHaA}<%HP9EUd%phby~26ErzgGdKV#%(4QIJneifT_
z$;E|Eemlp9Gu>Re<CI-?rM{!;4F(@8Svd3j%YKt*ScmTKsAHzqkbZdQ!A85%sJVL_
zNULV%-Co}1-dIgUJCzk9foK>l6cN6hCZ;Ht93VICC8yEKM&5P0x-Hsfe3JjBP3>gU
zGoI{@C>ctYKGX2QONYBFKD9B_g+dQe<6c#v0n_tR&FxGIY``b~ZJPXgj|K|7Q{Y*@
z_@fsOE{d}kXIq#t2?Dht?Vo>>4ew`lIK6XPS0Uy|=hbg%!K+H+fqeB7cNWU*axN4T
z6cKS>z}3@WH>|B?qop@hnM4|Gahd;w`q<>$#(|Id3)B<L@`avud)-c3S0oAg<DN~H
zjM9Ru&R&|&&Vd-K<RkWKA=JK_u&UG9@jqWmoK0vzlQl?EQLe08wJ-i^&<*;ir=biO
zAfjUi?;%KfqirE?k@*NzzTu_LJVF|hHe>1ImIOI45<<H@g1CnT7#CvhriKwqC-kr%
zjXEE2xP<xaNtlIK{C<UglK%Ve>^)on9|r3_`T+Ia^dQthw+gNr=>EUAt-jHLURWA#
ztqhs8sEI;omm0VX*}1cso16B^G;Va@+82&qdI#cJe*Zq=J>;A8kAGzB@On1bmWF#1
zbkdA0Fi@6@30etDx_-_gGCUjYT>c9Y0n_ina3_u&xt3Y4GPb_+`JB?l?!;bg69=kk
z&@xy@lSm64<)c0GcdmR2L77?&8+}R8*9VOT#yM(D`pEn@GE0tfRF}A!nOGyF@>;^G
z5C0~prbUYshE!#~JDw(ZXHcoUK0Ii3{S&Xotg(d=eBY~8PXP5jB%tHsBA5tQFj?Lq
z7@Bo+azj49%o}sgl}4&l{rjEDTSL#3IW1B}muwL&$<2omkvi3BJB<T#u{2$pUM>>j
zIO>+zP_1u&FTgVb^h;+jyBe|zxTW|dX8vRIhdX?w-imDIU0uNo&$4u0D@yj~@>XC<
zLne8t55g-j`CgYblD>$n;i&SHV;O{#a>bcr9LTABD;^Webt3G9ao=xIj7(Y3QHCf5
zn*FW+13w!b+-IkZpl_TfpE&5U+(Gysafm4MXNIL|<xe<qtdk%87%$RjPI&`zXr{m4
z5)QFAOmPhyeIfI|mzM(H@|!IMTn%j8c1@ZT%yeq#p~3%*+|%~BMD0pQYK9)A^EA9B
zlL@@!@xMT2e%$SAkm0Jw&646aWsB*sH8W3-A<+e(V=}iLoNF(neua?XM;|-txc@By
zxE!!5i6I%#J4LzKLaJ)MhdW{Nmre)7{T^PFZ9<GTsHc1lN@u<!`8<Oh;{w<Lzs&LS
z^%~bUJ|WC;9P}Ydu?8-1S~I&_clkIsMH0$rr&ehB_Pqz;g92H?8K-Y~X8ffuWSWW8
z<5ygqmIQObzuJh-zo9)Ze~<01udXC2fd6wu`<02ijuf}NMy`FDjpuu5*P@1Ic}pA9
z*=jUTDIKI^eKYFmKS|ZJMa`gD;q)0hYge*%eSf<Tt0qNiM=j202mOXEcVQ<rks3IL
zE|i}TTFjO6^o#?l!u<nYwRN-<Ye1u^-cD|*RkXm%kES==Ql*8$EL04(Ri`fN%!((<
z9$37c+qBadM+}YamR<Fj9(u09bdd>8nE1&)1(CjEKetCHjyLmssdM+Bi2Gs);dV<~
zV%~;Iqm>EA`pH4h&y3iycTc`x9Tem$6y%%;ao;jg`5hAf`#OW!Lpu+0Fw+8jXrn?a
zi&6vdZ0Y2K^m1l_R7ItNx}2*eMC1C(6o{asR$6$J)S$5A>;tOzc4&Zy!2RCn!}z|^
zDbJco)hB43T|~&HLuvTo`fJQdQ%!T!x1`P=y`uXdl0N4BVxv51!$P>;!8+PwpbVkb
z_Qmq0*I1turZkaiuYJAu61l$PpQpg49^g(71MrIG@PMXvVc^I$F7bvy5cJl$jkP8y
ztoiKETQe0EYu#@jV?XAu7_q7C^PYa!JrF6}44A(s`@d%r7}T<E+=ec5{puNSZl?3I
zEN;In_+1~1xsxj0JY(P&&$5aE2U1m)`qRlHG8wKwAOozxxR;zBGGQMhB|{FB2KR6L
zXm?utO?C2sl_SgM+wHK%kGu^jNlmQ{Za-OYuE~&Yb7z51jlI*8%6?nxbG||ts;J`e
zgM4dBy?CD`uPVIdOtrQY!^_AeloSIvE?l>&(z69uf`j~?d|4Z)fE2ozOA{vu1_@S$
z?@k+O^V3b0Pwk>7>*>x+6y!{D<&gJq-!6O&k;MPG6Pu+x2w~#$zf%&U3ysXTx7zuw
zM8u`{<siT4A@;+ofmh&PIMPD|ao~p;dI~SX6ssv3>~Wy5&wKqs@+pyLF@4bjXs%ND
z+4n)*$vLactn)|SspO;cVX=GCU$Dn-+UzqkzxW894NUtz_*fijTv5qv(DnqV@mXLS
zAiC$9hTON&UAil!(B)SousjX91~}v;0w|P{P$ASw>%13%+5*FK*ery6Gxn)LZVINb
zU=eJHMAr{pgUd%PqDH@O{<ju(aS#Pvw;&o%%iL>d@a=?xW{n)D4hQW%1h$%{cA&iz
z+%4Wo94Bna1&u7D4MO6$H<?upkyVBZxJh*aM7!1sT`Bw!K7W`$=l(R1Ekv>O`XKY6
zOnYn4k7uj98VE@$6G~|5<bvB*oGZ*+@Yr-mkF%>qmyOMC%~j^&a_}C^V*DVpMBxz0
z2sN((_g{$~LxN7~x}z8SKAh?-z&^gHU*(;ijwOaMsYu0UEhx_K=eKbPVbXUVIy5fq
zNHu?T?vO=Y)-Qq+phb(+|GC87ed_!&s}JvKIaNCASN5J-JguSUuivq7sdSc6?RI31
zG>i__2x>Q6=mqaB$5ds59uRNgbg?}vs(~g>m?!NPaZ{Iq;%k}c$d_&7mC}4@2h#S~
znkmnG6KMR|@kg15M_q}RO*oJaf<8Cph3cSqVnMm9jD)FWaRKD8$S6pTF)oE@fcjI)
z^rep%NTUt@cPqRU6Df<@!9D!`Z!RXp=EoV~qblL`ZbNR=1V3tGejh`P#RVM;8zGA$
z|J!nHj8mq}I_$N#nhDAh8by{nEnb6cO+0VO#Y!r5s~G9C>B@YPJ97Rp0y)fq9;O`|
zyeEYecDR2)t;3tNk?WFJzIXdEFScQm0ab?NzhDf8J9?0h%-$)kiZ1w;y$bC67~Mi6
zUAFq=!DD;dr4k(I3)Vgk!d9%8ztEx~B6Pg;?@y2kwnhV{u+!Ih{H1)icm7~~@=_@Y
zqHw-}d9rEtgb|ei3UYo>oJ{&C<=2kqT-D}3D_SfN>f79Yp!*k@-y!n9(_b!MzK~37
zB+_ZHQD(Roy7U&OKBsxcb)iXjy7;APa|>s8jW5@6>W1RqLi?P5qQRDs;65!J)HB4X
zBf3}mA~8B6FmdlHVEJ}|Wz9of9)1&AU62-VEP?>vooid92Ewrl&VSY=+V;3~rUEmz
zuc`8Pdt%2$;0{~;QuOG%_?g?j`LbwwePsRQ?v8(3LO@fe7C-u!1;H|SXxD%7w-$M?
zr5Z-an|$)TxxRE*QTG5%)GwZvT^}AWeI@h1$7nA43thI}*WcMwh=vs!zG4h>8=Ri+
zl^BH9`YJ*0#*4*2S~a(G?uzzymaj%MepU#4b@LEqfu5qvYA5K0L5k88aevw@`p%=f
zC`$Yxbu2D+-ypRCe<W^ZZrcQ_WE=44Bytx8^eXROt0Vl)LtckK%21|j$_Jj~xDtE|
zynhP0aH_-ndQblS0OQ;lK1@57amSl#Z20e}H@5nURpTXo;(zCp5SEu$KU6Dp0=*p&
zK2xSgRcW)NNv7!6CuxgvRl-6ec5?wyiyhksBY!&fmq|MK6YwS^mp`D!K-ikT63mxr
ztntn|o_|t6)yyLP?M2a#)csjlmaK6zy(FK$XKUs_7o9z;F2{2knS7})c)AVTHFZnT
zlj&il1u#*DdkY6>2v+BL7F$*Q2LtvjX9V2zAVFZSsBtlRa`3R2TROex2jeYg`ohgS
zl?(IwTf#z^1chwPW;wM%Tu;cq9CgfPq)g#|OO(K%^Y;TXvn}FHt8GIT{2KZo(WYf_
zKoMo!lfhXP+Mn5<d?a_+j$c)}9QDRXg)<I=Z6W=O<m~`gRZrLtuobPSPC=g#B~4-!
z$l>p7j!?Z)waH{g{Ai60&w7wfAvoj@Sr5bpQn<39R+vxgHRquMdqEL)o8YcdMRjyR
zKpq-GxZZqVyny%um6OWK$TxDrSl0d03Kb?8^KVNV$T^e}&A6{I=$o-9$eA#ZGqF8H
zgEcY0;zB^a{)NE#cK*M*q;MdYbyz<#<85IgoDBK&?#EMVzuXq1*3zbye>MVI>shM$
zPp2{A?$<l5a*@}b2ZnfP`v+-mc%lD*<HVb-I5I{qcZ<Pxqlfrhrx4`WLB1N!G4FbO
zlpM=8yYM3_US2NN-V4%s@;=_rAs=29ehEYnV78AN{3M6pifOnaEZ1P%D2>0FqzebC
zkr+;lyj3y?Vp%*hU3#g2XXzo0;>v9la{vaWu!){XT1xCIgHI{-o><(IIj`(R#q301
zn7?*zM+V9NyVc(X@8~{@$tFDNn?b5VUe{DjmPy#`q532Yr#at2D4MSiF#&h3FZZ^>
z+ynRiB8%&nc>f~N!Qqy^$IjRiQ?|yHOZNvkwoT8cYL50yDS6`+MW_#*8;5d7$;;7a
z8-g&iLN->_?<mHFWa5A+lBzNkal8@Ge0l~vLQ(r|PYL3T=V6>IE=|#(5Px>tty*OB
z-PD6yJqiTH8n5Z=7DqIiXKLjnRpt1OCZ?w!0M}#3cF*g>|GU6BjOy{P(+7LLczm;<
zuG|i~aIZn7X;fg#+YhgcY)u{x@j*Kp`*Zqd4`u!=d?rz71FUa?5XTKy`5;xf`hC!2
z)vD|vwJ@N;f<LsVs8@`rT(vH@TeFy%Ntz@V3p2H)|2+kKcMDXv>6+nXXekC4vz`1h
zm~g@;Jh5N?Z6WKCle@`&Kwo0Fqr5g10IeU^nm?eXg?b*Z0k;c)faDu(<B}uagTmt$
zf>+*IalKo|?pEq;ap(FVAYgUqv-AfE*q5cFhrB5D`0R)1OJZEz!hfdqkk10(#+9xg
zo{#{c!hc_l_MQg(ks}vfemWv_angc(A#2`zjhM*(aX90wSUnMuG5vg~;qsjTY7CK3
zalSeW>+1HSO-@n!0iT(lbN*;IJ$>}t8$Y=2wqgO~?S{|?{na=D^gz95*%C-r%J8=D
zb{*DS>5pwzz`P(#zka6AZ5)wRoZ>1o<t%4vD*;@8Eu`F=%foNS+_0j!+fj3MWuwv6
z(W8SOUKt`2KKZ1~y9YiDfvjYCY5u3ZCu(llt7ob@ln%Dwr+>*B*K*U4gwK~wR`g7v
z3rUL`wp{uU{RpD;quz<^P64LIp7Nv4=JIiidP$&k^)uV8U4~umaFwRaL49%Df+O*w
z(kngQheGS>OP(PP)SsN<@He&rb~xunz;ji$CEo9-6NdQ15)F}xGvMrNYyf}}FJ)YD
z&vM6$9crs8epH$5-EIH-Fz=5)w)t~KT<8pxA!Vm0vYhi1ru)3F4-9@l1dk><k01RX
zE)*gN;!r>l1gD>*?dg%PZ?Q%Fr|uS$%w44$k-SW`D;mHpXstx0c)2JZ?)f^%rRp@u
zk$gZM<07SnQ!CgS{`3)_ONqKLB-7{6ww(V}xxcb9uk@>7b-=VW(gbR-pGAQk3&1vM
zWUF`8dc4T0>{aDt`#rvUcG<KKf;dP(3RCF(njicg2D-x+pb$|(<K`HtE&1X&E8$?h
z*Xd6}L37SkRmJ;xKtKEc4suVyoJ0P^;Xlo0W9VrOm7itR)Oq9XVEHxS3!9-2!Whno
z+FzPJ*RmG9RcIH*{UMW(%9PKS`Oj44pkh;2HzSWqRr3V>-TZnl-TIW7Z`3X4VR5lr
z(~Z^6wxj=FQD+_x)&Ko*qDZ1b6dEnE(}zm3Or#_+rHp-KYb;qJ)R?8DREo-0mYF1k
zkZoicYqEVrwz2QaFf(Q>voMC=?GxXhf9G+>`=0lCpYuA;bMD-PGK5+7lIBft!7ufG
z1a4pGegIEtNFYpd3#3JVbmg)&!L%=*Ic^+OWrBf7OJul#owMZ5k>-K&)0L+l4SL?z
zhfj9*9u#_7cWN4D96B&i?rD^tj*jn$L_!8l@{45EPcqWLySKA+{hhJ%2i~*rz;PCf
z2oi}<W*VgpuN9>uH!|h<eQr#_+$@aeW8e!vB00E5efYwqb5~lp<9YcNp8{9z#i74W
z8=#Mr^jnmUMt0{MMDD)vJeU|V=8_J)6=MF9-U$aI>jU(jD8&A$FD`!MUQYJ|GWUJ=
zAYQ2OrX^0WA<~X+l0*LNg&Q)<x9*d_Npp{aWiy<|>cx56-NwU@RKldmAH*%<t}2Qz
zd6d}Ab7DFT4n5cQ7UdcO@lk#VBDO}NHE-AQ%P&FfYOCQKLY{-Fxv3ra<ezE~lw{El
z8ntn3ZC|cLX<mm!WHe7l&7&UNn+;^I6GDzsL`AGf`D_{P6zV{H=7{>D^@KR`;xH&;
zx}j$cO*+1UL_$z|A@_-68Rio4+QLoJE%yRaQVwzy83swezKHS|cLz-nL>eMH7r4$U
z8y-(nPK;F4&m%QgVR;w9kDT9Mao3ucd4d+m!9ONLk}n*^S$mT|AHUFwupEogT8W`D
z1tiv*Q7hX-?O;e=)ri*uBc--EpkPX7rc9>1kNl7mSJ#;YoC$~0DM$;UDelThKVomS
z<)*NK<?7)uD2w#(Z-{|~{D`O)-DGgVB{e~r;9zDJs&W6`#hgbps&OErDlrE_$s~T#
z8?pxNCdj}k9#~kobx9v&;S({CVK)OLh~XFbP(qS{lo&!3T^-@B@`Le*|Gf*)Fv@pF
z7fz!!=U|Z>!<Yq)5+g7bvGo2wI<zm~!D(H!`)dT8x}+Hf-8ZW#>7o%X2ztE_q$TXS
zGteSH0_$KF8k|FHQ2B)t@*CtThY{{{NBOGKWw<|FgM>Ed51%Qq?dH=WE03gf)bdrF
z#daw>ceT_O{Bt9VXv)@1J5n-3eu?Tv;Gpfti<HX`0Phuv&R(upCTKLK^@f5A(qNs-
zJ3Y%l{OXm4lC3BG=YO0OD%0Ebr1m-~X2Re@0Dui}*D7!&kytg>--s`xA0`Jp&PT0G
zbx<Yn2i6QdC;0oa*1_V|y||N|rGg%136UZ_oXD=y3Cw*z6I6xJxvtrT{3aPL8*J?s
zg<hJy^bRyC#u$R|&s_3hEuH^jye?;~egAFH4!ulR-SK(}Va;8tyIXE%u@;ld)k>)a
z$^?t51Xo+ILSZhkyG?v;To2V9mT$i>9W2GKDYvh^gX51d|N2=qc|bW^2zY#RIOr_^
zNCBj-#_Ia<|LgWajwWncU><uW(taeLvi{@Xt?{12g=OVvpxf^xUMBe=w#1`OVWEt#
zdJ?s>6h{zn4)OO0T#-am<pNi_vR6MfE$VxvITrU?SS?O2?CcU9duFg!6{(<2q>IDF
z3<r!AIrCORRla4Q^igu~y5b3q0q(-yi*#d>2|(je!#geWbCv^MgzcTa$g5+ml61)<
z(L>wY1}`kXV)iu5t1JPECg6RcL1?>9ZgDV&JFrlOkNV3I(=P4f2WqsGMZV1!fHqZ~
z2`@VknrfM2fDR0rln7R4=-cdj_hPlYJhUADIzijFew0`hbvwW0T;NKxNy{YQC5x!@
zt1!aD6OOQKX5m?NW8M37y2ZhED)xt^G9N4KAkzETX^0uB*!@C#hglCwkrrU@298HM
zy&(8j5OXNT^y!;gVs{2E-#$z_9r9NDV)B5hi%*c%(}}hNYt8yAYkm(3R#$<dwn%y9
z_o)9mzRGXlpT9`ot?J7eW_i@CLArzLaF~wE!Vj3#_g^o`WrJpAMz?l2$(2RVl=E2;
z5>WGgdR))SyiF@%ytndOu3cWfE=h)aE0xIdd=ZT{4NsMs0}9=S6wpU$L{ml0a-zcG
zDc3`UVwg`tVxKk0l<;}jj7)e=*hl?{d$VRUX2SdM#;Fm1Z9;Yj=hVL41iu{wDBr{S
z#vuD78%JBi%&z`YBV1+55Ev)iGzcMyYF-2p{>SGJCb!>W1*0>Hs~$?vyebp_WA2UK
zfGYY@x&v>dk>)>6;aUl;Ssd^KR>ES%g6#LA08m|(*3eE2W`<<aSTsv&Jc4<nv-Z2I
z0f3h&jNckkj-Rd)uApw!vMz##Yihd&$PW?&claJYHBp8j<QpXB5|5=Aos^ZIFiUs(
z5&GQ^(4Oy*f7WLXN*`(ckWM)*qh(w@Xw)0xKb#@R)kF}mUKJ30HzX+y_q7w~9c4x<
zxTE8{bBIyf0JE-ZAt;@POPOdH2t5_?Z&6?5V&y~cT=r0`yP9+D6yOWTz8=U{C=5Zs
zr7_)WLL()i3hRXU?BMnu8GtmVjh=Y-fV=a|AKPB77X5T4#F><)q{bdr#S(R&%B$>E
z*!$-*u19)GdG6-{Xb*Cge$kHVt_~m~-+aik3HfPQl|U-5ijOM~=m|^!uX~~FdsIoB
zgv4_NBaQr3^lnyHI6A(mXnVbj>Dr{U5#Or4HIVqo^fGZ$nFNO)!r<HFi}$p(k$V@&
z3bSa#5L0js<gl*q6?TXA^j6@3BPxVis&Zc9(vcHA!S_61r{SQs-Vi=fTS@jf(^T(3
z4k^9<N_;vO*2?We=t{?XitIEHI?_!nABI_z(t45dE7-3G`M=&jEVASNdGpT%-WF-H
zjUx;S`s9&{vr2(nCFK1J!T-sbtyvx9$tlLvwr?-o!?S`nP78)PsY;rO%*2*Sa<1kk
zYVBQ)kyi1Jwv9dWplfC!CU@28e2B)z_eF_y^G)G(J28L4751wfZIgVEZcr3@{rHpA
z9#PtBzU5aB;3%nG*><#@kFVBZEa7|G=@vCb=5p>)cgT~_^)Gv!)H~Cne_&ThC5Nb3
z(Fwu|TF(z;MM~F|i)*gY-rJpYrY}m6n|J{67QG)(rFYVYobLJz18BP_d_ztGQ3mzq
z$!yvJKW-cM(rcvH)rsbCSfmbop#k=#JDd0+V|Bc(hY?{nV<fSl;n5~p8*^L#VN4}C
zuYDX+7fNRkmIV`sBk{vdaexfQiDCTTywA0!MPE3g6aadq%zPIHI(M$8bn!U%W?x}|
z)aMgBM~%Gto`Jm@C4tvIiz7g=eb&@;7n*|I0^gVGdiD4lIabbPsAL=a+O_#?gqph~
z;hR${p{h4--L`@h(-?E0{m@p+EJ_kfOY0q~&=L$&NQg}5oSIb)|4;MsDrn<jZc;z<
zZ}aXifb1f_7>YbGg4&Q7DdxCW)>nw4gCHY@0$)0vvUJuwd6A1~1giQ9GC2&Ci;v73
z?%;C(jSddhi9Inthv%O$ErD6j9j_%WGnUTA2qEPxaU_@LcGMUvjt$92`trcfkr9zP
z*q&)d(A}^3S@V_u{@%l(rb04?+Oz-gPKD|W*dhEjAav{Zq+o;eX%(#vohddv2QO2e
zUd{CyRN{=hp6(GlW*hk|Z;0F<T9(p2+e0*<GY07=lWXDL@K5uED=WM}4NGlJOe1ba
zl`p6DKg?w>UITI;sC24TK-}<SHbay)yQnvdn7Vn^q4iP9%h}$z>n4T|16147jA6Th
zR@e!~<TB&8z+q_1<@flGFSJ_PY5xEp^WB77v+{EQHRn!QtTpSbaRG9`>~OmJ6m2=S
z4Ruu24W;T%ztV<!^+?7=rYoNE;k8lxk8X-lZq4&atB`YMc3izxc*aTKrzr#GRy5)}
zE(@=IUMejyR7!Y#UAgNRn-vQ9cBn!%)&Z^^Kb#3CNAv*mE1bsygM8RQ-4~5gWKyQl
z6@2Oy<SqH=0lHf2Eiz?_)eZ8v1|_2)1DwaPKJ`Vy#RF;&a`j`O2{#;^tb0~{OLc<)
z5gU#U@ZFFzYL5;5+fI!F8A{zQwe|asWyt~=veicJpv50sBV6wVafzKu99x^~%$$+U
zR_39~^yM>^rGdMs2tNPHaf?-z!C>(Ij|$-91;Z94Q;lTm%B3F;!MzwW%+B`UR^_=7
zh@t$jN&)_p_ghyk=Td_IwfP=lLUGriCp-QHw1T(Pw(+Osk%p2p42DIfmK8}M=TC`j
zsJv1Xws>bZp$h!a{BON!>ri*fIM<d|j^(@xD7de6@YPdsj#SQX&lK1%4ae%|aU3@W
ztTa1vEHb>sOV87~dKIfn#}JCx{v2ThRo$pj)~ZJ%lL{0h(Z4+I+bgf!v%xxe7SMKz
z2w@*fuH-j9dF{SpdGtib_3Y&vd(cA9JcZq1%Iizb3BI+Wm5Saq7MbI=CljgXlkdoq
zS7pvZq21B(LcnpW;}!#G8>rL(DE7IDrmYJ}RrpWtO<h^c14zXnV9^|#3kPi}=>eOL
z!7b>~pDkXsrw{tgpzT7dM2IoN<YNUW_#@2uK!K52pGn-V*;5`t`&M>IzFfUDL(=f5
zAVzmNVSzZ&%vR=gddd8Kmmr8}1dW+;_v&?N9Df51-IG>^Xyq5anOooG0?zI;@g`HQ
z0j-5TpYE|U(l+hzf`amFG<^5Z7@S(Wc3H&6RUz(I$Cde@j~!bm{;pshPiWg^1;Hl*
zKDiPh-#)4q_mT1I)5viNv$DDn=UP)>{*j4PD<K$yKvgV}N}_H=Qk^V2)F+%!YXCH%
zFdF;T4pmSh{GOasgp`aos<B<B(2rmwm;N)0u1^HU9{)U04*jUixY1y6HI6u&Z}8=f
zs4DWIrd%A}d3O0+btue=MAY;YCla59k}?5CqRc&=THTH306q3Xpzj?+jA`7WCJhRL
zJ1F>hbK-IpX8#~9I<~J-iV9=tmtA9W+<Qaceh?;h)dn782h1@!?2h5mN!0k+jef71
zK~Fh~StR-a9nWc=ugdNv<1}dgt1puy{v`@{HKhF8Sc(#WQ2i$6YE8dGpLd+@zRSHb
zpGJENAx!lN12Ff)^bY3ynfoE$eNFI!vVv5j#S_yobg7#>OnoWT;^YYA<PZqZ=uk!M
zT3dg>Rui<~=Q4qql25BDaZp7G?fG?Me`wYaxVP|Y^{5R@SIK8Q(n8ZgLy4`4(U=%q
z2&Ol4s-7j)IU<{eF6H1Ozs(xQF-f>AqkO+;(`tn14;JYl@{E(HRaQbaO8b7n_-A@p
zE#sWKct^qd_+3I`eMGX~N^1D-i@CXG5``pYxd$!zV!<10BMooz32}%uDaLBejm#~c
zU!UGJNd+>fpZ7Kok@3+RHA(9#CwwntD$4ASdi7s!IO;jL(KXSs7L#!|%%0>=_&nAs
zpU}G3ex*8KL3qR2Uj%3*x`dWK`v$&1GVxh)1prBPOx!v<kcWFwY?*O>_o#`=8!yxi
zKv1+@*g5s#>U`QEWL9^p)?o1AJL*p!&webjZ@3IqULpHCO19kzyheiO=IcK1fE-cC
zS&jk9y-#S&cA4EML|pzx%pRQ<sAU1&Jl%W7=PX_F51sqvC$$T$?UQ*={Jr@4=P9!w
z`)u~oE-r$h>g#Jwcg~5*4+{Pr-}3!yTugT8F>}n_1K{7!K(^;^kx||6KqV6)e3F+m
zp3X8L<KjXjTeND;W$|Kx=6bXGru1=sgEfZinTyqB_>3x8KCwqfL0(^p(G?EM|2GMk
zz$v_nQA&N_U~lrb<6Y3;6Qo;BTnO6u=AdfjAAauLpF+ucZ#_Gwr*7$H;0M(xp(1{#
z9VFvd=`S@(L+s6w&MZ}f0}Er!k56@zDg+%xQ^}X#f#VZEk*KMatMgXggi4Kd4;bvn
zhOU4#8#a4FM&-gsd7Gyk_ctx@jy(?i)RWN8vuCZX;}ii@S0+D=Ozc*mRVN=>^YmeT
z-8$Vyr#t6{s4SSHgX3n9jA7L0KU-_n!ob|J8;!j{(2Uy0v7%q{FY5n#0BRd&okgUN
z&!E^29W^P20#=z%KGNjw4;c@ly0IF8OE-izzU-|Q5G7-u&fbSiJa*<qX=QIj#=@<I
zNNegHKCb#nKYQ28&nG~0((la0P#2r8uA`~Ugw6o4${cn}sS5rs4wWq~PbuYLmjSS0
zOfbN$0VeiroUi=J2zB*Zp1-)e?{_<2Bl@&~tg*t1tbo%%l_ZXoT%Q63h1lp*boBXi
z)$0t4?I3N<3d&4pnZ%Hbn4P0U@cy8M!Gl8H-XsC_4nJ3|g!<;(wsB#gtUanU*Blho
z4CR;tk~Do30yOYYONd_7R!BMC4lfM}L2C#;2?0FoxJtDN>`(ezOFNKVT2-9bvlP;;
z!j7>*%3fC}tR&tYi=3s^Q>n5@waJOdEcM`FgUZX&<25(p5dGj$QO7BLzINXPdj2<|
zK-WW6Q`3wAMKwu(Zgb4lDe2e1D_)kFyXUab*Tlj<qeE1Lo!k-og)P`Ed0cQu_h4I<
zyPc@Lqn*K%<G%{_{a30U#kFx~yqs5Ud7`kx3j9v*#gUHvk}G;u*Zt1i<1h~m^<<Cd
z#XiwE<TEsl@1B(Ao;^~OP#ZmF*rcYB3icQ;?~%}>cL!aa3JD2m3^%aV8HM!0Wu3eL
z)iT2p+mZoIi#5yol(W)sZp~&YkYo2m@itzjoN2QlR}Uz0+)b2j`5oYDr$8GZF#fFM
z+2x1wHLsz2La4rz*vorE?gw{>T-2*mbnzT>1z6yIQS`U9H@8DWrK2W=F;S<?e!?9y
zUT}e5Sl+pPqK(C#IitqTR~Hh!)*Z`i&PTe|6^Q*Q0eI;VPwy@~Re9<z6ZFn!&2EWN
z;@d;~q@k&*wR(*6?ha;SUBoW;Y&_7@^u4Q?N(-)VZ3max`7SXHFJYkR|MN2)!8VIM
zDQKEDFwdZJ@85I5F8NvA+3nzhg4HJ(&KrtRySq*}MH1qXz(aBY!!%Z0hk^>XuK2`S
zWb;OxI0B`+Ek1P6-H8nB<oV440NfmEeB-LZOk8FRL-MB*e4BpEEjB)O%h_-Xt-3Wo
z{qF0OLUar*!L%7$=?7(?2SgChCief;S?UtRK<7+~B!J4b0S&$@N~sY>Q7M@hdZev7
zy9Znqt1|XI)g4u+4^~<>&4TP&k^jw$Nc{Y!RTxkfkcz5UA?|kZH1**3e}j#qykqdz
z220V#_pTbuL;`g3x3`%}^_)O$bh_%E#;3NXYehz^lMU!z4tyYL)l&kwZTu-q&mi4W
zd}s<0P^e&CO5g_pMdogx0nd2myxwyaJ!ik%?;$&MLqTc2%ZPcBk9Bf(C6j=8XPF15
z1MTlEj=L{F={`rFZAi}sw1RvDN`0BX0UZWx;SPFdNqXGujAp;d<Z~_a`Iyomn~~2$
zWKPCEN@Rf)T;_Ig$~d}kgLazmcye9+Q9?qbJnLJ#C-!4s`%jNxJHLXPOo&WE3Tq8v
zxs+)&A2X`etm^r{OV#7#1P6}Rjb<(M7(mtIeRU|yA50XP9AFf#f#ei#3P(Tm@sxKz
zV)#2*o>>&dN-DrfU`@8vS-68Tdf&(2o)7upaust<p?$Mj?bj)y7t*1qAz*@1f&F+h
zF*eSwVZTe_XGl&9=(8o*TSE4KxBY?wa6F^|P`j#0n&Av6F&{z{nF+wv+?~4a)rgXB
z$)z_nUt&(bKS?Q4UM`wKgKbCsME`&L*b|5&=eY|D1D!3$d2fkwkOz0Z<jPwD;20>y
z{D8s6H4&brJN6Hl_D<SrtLW<fwO0L;TNKgHZa9jPj`4d=2LoQz6@fB{<ZswmGNqjk
z2FJC}fReyF#t{Pj=9msv%ja)Sl(T<X(iFQhWLMMYE{*;y8n*{%iCh2qxBFjmxgRZa
zS_(Ov?+Us$Jzd>%Pb3Y0tC1<z#O%vx*4@72Pxl-o+kA7Vh!GXu$_xipod}dgq)@|#
z5Mb@x7-{>_ZVDa@Ynq|Px;x+Kr*Y++#B^}i>WD1J6V1P3NYHeGW1!l)mN^WaOyPsm
z7Xk`y=}-dg?kh5_&*_goT5K^=-+Tr5QIiyP8b7wxjeUMna7Qshc4dvriV(yUI;c`*
zm^!i&z%`Gl!zpAlaziFSfwgCgY`ab$EWe2RONWirljMYNAD$sH?OwB1QffzmxSk}!
zNsc-!&dIp~>~Xxal!kqaM3uWz0+`bE*FfQKyp>l{V3I(@DBAT3p3>K+2e`%kTd93c
z?#d<E&WSU}Az%H__L)S~P4$?PYcUeYzvsH?w@;b2xmD^Wr%H>>!;{r{Q!oFRILBL&
z&4cxn`e_Dq^Ixkxl^@t18fh9^H2{9~6qxuk;ayJxj5jL^ENzcC=OVBZ27q{Q9HkU*
zo=y7oHQk5v4f109p7#6~KC1K1unZ=@FPH;`g)6WNjeB2mkN)u-p)pBnEY~$DhiuCu
z^W$$360XXxpBZhDPYlp!wezkc%53J30QR$1WCjAx+6I_q-M45XA?wz>(&{x4rPlM5
zu47wSanAsi^Q*?bWg0VwBm@sEbE5&Q^wSwfycgjD#ssn<ymS!0_hpo)RQo@|t}i&_
zTS+Y9T4<@Qs6gIw2+s)r7zk*bfUi?#-vjydTGuJ3ls2~`H=AZ`8@ATxpd~|$lhb=p
z#d<t3zb0z!Wrd?Mmv=Xyc>$oJtvW44o{Qz`H3q`oo5P(mEhA8N+e!Nw`5{kTzVx$2
z$ystmE6<1>5CZfm%f~nu3!LPD{fedFgvVBavS1ttiS}5UzOws`rL8q4wATbN6I0@8
zs~b*y4=$G3tPMHi#B|Pyb9QD}<|OiL)M(~{{>|vhgbKATHM&(BY#pHqAEpI}udz%!
zp?Te*f=IAkuX0cYLcOY#I44c4ob@fBp!lXYg)nmJ^(>rH{Grv&`S8TX)`}-z&4yaW
z0kdr*z^QM?<`W&-S=AzA&}U<62I8FKfLjjp&{ulWr_bKjYrmI*yE=GQGwZbKF8bf}
z5un=Jse|R?`<c>ze?|?Mti<i`7keaXYDoXq9_?VO$ZUKB*jVmE8zhrbGBIE7<fn}R
zyQf>u#qQmY-g4B5V-^ajA$haMlA0?s2Wg`{3qepG<hg^Zp`sj_=ANf67-1ko%JYow
z8haMgkUIzr7;CT%{nLEX-<ZO`L!<u^b)8G=ZHo>Muo3&l7cBGWnLUJocn}-TGjt#l
zIs2_W#=%~ZX?!Egs(OEob~<%sb=m=q${$8rwG<xl%B1+qR~OH&j_iJz>~42T_|V+8
zBkcY0J527nYcUEwCL+_w+|k~aQG{{%Fk=Fo;Rs$fQV)b_cgAVxAQhQzCjkY!w$i$U
zIS5S4l<{DN#%fho(=`V0j`J(!VE4q6cU7jRi*!aAL$4~6PJC<EG)MDX_)n!@+3}Vk
zCoCs)C(}~-18ioDD|g|i1*QdWSxPz0!XDtS^{hWtg$!^drvu-7^ni=#SRSBJk1%N$
zX2i(u8lnBpIKB@z<CmY_)^YOq(O9Z6yN71~OL8P0PZhPxU~JA)XF%mxR_|l-dOrZX
z_sp__t2<>^7D&ucCr#nW*H{|nOYc`dAZF<|23gE%_Lueyw)F((Cs9S#fDkqc;vY|t
zPU$_4cw!(!)JZUhkBpFt@#Wl+>>;?+E>Pgx#_*#CL=SC{e5pWQQ4;WZOQQ^-5g2Oh
zv8b*%%Hn04@ZiQ>qk3Jh)w3TEhLNv^7IdEktH+?UKt<o9fuY<xU8nr8$ZI)+%|VD^
zHG@0<fLv{R!3!maM^BW7T8uWk7h>xEd?g^#bB4*-SPn4ABkyht(j&Q^=jQ2U#J5k4
zbd!n)a($(Ny8GI`U6>DrUy0T%FJE9%#3Q65W;1W2){DTjXbjryWK>$SQffh8#p6c_
zv8~n4I!+HbfXf&**8oP#Ze2Ap-xyrCd&pB;x46mcWoeu#qg-gp1N@m@vM|mwx*hg^
DzlO!b

literal 60119
zcmYIPcRW@9|G!p3*&|tHkL)5@DSL0Sg(92m8Oh#-%7}<;+2gv1l)bm>y7tAr)^%}>
z-;qAQ?;joy?mf<VkJtP49Iw}j*49+INqn0a1OnYuQI^*Qf$%UO5Kh4LYrrSzjsXfF
z&>fJ9yqvyo=2jMACN=u_LUuCIan8Ms39N#OvBGH&qin9?EE=|t`9*scdt2k4@cHw5
z!ro(30nuO|d>wpzg6|_(8d7A!65`x)Fnu08J^gWddVc!cZ_^{7xmvHax?VB|wI7(x
z)xYBLCJFxw=&?ko2*`X6BZ<R}qg~qQg`c!>{_*A=OHcc~z}s44ci)mfQSQ0(yin1z
z5)JwU+QJ=zpFjKBxQ(xU(~2RK7$gaDBGB%xzY*=jjZ|sTzW${GJE65d>Iq>Rq`v&;
z>(LafNO90q(ebfp=s0KsRJ0s@x9S}{TR!!Ey)g^1)veG)WX>7R`#p8w;=6B?@D=jT
zOF*aiR^PE}A|S{4X}q`cj4=U}JR<WQLH9!OKyf%iw;xj|W&XW9|9Ugdy=#@%+i)?1
z1ukl!y9rO+wKDS<Lc{OZ<Gl{{xx6L$lixB6kn3lzIDG^=@w1a!%EsDrsoEPitwch9
z&eh8?s$5<5<caPH1Yf8($aK(}y&bsE7AwJa0!U~)Xf}~e>F<q{<&&NSu{68pG?tuD
zJtyjtu6oWMx-?%;1QK~@ayi?(Z{<H2g7!fzH(}-=S+}ET3TlkJ+B?av^ZTIgCetB`
zt4F~j(Jsjs#gY+_4{Ps2;n^T%LLrJadXOf8_JG{qWxLCMcZwAD03?W$aq~-jA-`8{
zgO*1A!fCsUoS@DL^LLp$SC3<V_<8=xg&oMU8WN0<bI&n}iaZy0(OKXdCk($@P<Feo
zrpvMfFs1C}u~4|o@diF@5;Wk6$>zGcLq6$<-h~Qi5U*&}Ps@TEnVS9-=aiyt1LUC_
zB=mRI*Hzqa&np};dt>)iQh;ZeQR6(E<wq!eJF)uv2=XU4p5emof(CJ?`tPce0=Edq
z0SOSewVJ}HuI8(JM}A5c2POlO^{k;}h{1NgAP(gNWx`w>uGaPX-F8;~1L(CK7Eq4o
z0VNf1#Z&oH6;L^j(B1zgLJsHN2hEYfZXWh)-Qz}5Mc%8=3BDG3_q)uktBE8<{a|ck
z0%hY2_2njna3hy;oIq>1M<k)a|GT8<2WguuNb2U7=%*|Y?YVBjXGmPXNu2f1h;X^9
zxkmG_?p+L;hbAVUK70$8IVaaTs{qa6TS4?$Zv2~!S=CILfyEat+14)&iGm3}g)MQa
z=r{j1r1OC+sG;55E~gWR8nfGO4Dx9<o&C4%%6U5kp?o-jgQB#iQw-8|2iJDkLyxHs
z|BWBtHpX;NYGp#4T~Ww`h2asXp~oBt@|2ep^e-ef?A_?*n@59k<SAmX>z*Q^&7cFV
zTD<=@65kPYDhsOEf2~9Yd@T%WI|F?y=)w872fClgKxVkGHGP&S`^SSp%YUx9Nr$rV
zY!O}EoTu}CJ!uoPELIucdRgX&I{uDO=;EKiq^q%!Q-<b%PI20tNcC;4mSse!K@DVW
z>Hm$KJj~i9=rzPtkRsB08S0mSyK_A>_>Z^MzyF1U9C1#W+qf9j*Q}PI5_@wzM=Uy1
z|ArOCb2iuH@Xb_-f)MzUo)omh7~1~FP3XVpau|X#NUd-PrKD~(l9ACv96@~rmH%d(
z^o4$FLT^$x=N%y%CU=yiO%Ft>HHP<ZCz2k*Nq_*D`khKCrNV(69T{oCNJ2saI_va=
zoquh(gea?YialY>tt%j+(LfhDKx1dp6w%FVYsfT3FGqW4MLpA!raw79vnO6Oz@C(p
z+zc==9dmj8`?tKJ$)jbawDjZx=3dpo$KL+${L8u72o;`5aQF4e$tj8vi&l@HX+CRQ
z|704P*T<NxeHA#k)l_XPI4ylu#0h=!Gt<VvFch<!y?!2lm!{l(`nfm4Io@lk#z&m@
z_GXp&JYwvONds-~7VF8sJ^JiZfIzj~;qVvy3mNa)`~K-8tph85eXdyo%P(wm?{Tr%
zNf2OzU(kJ#;$i{sJ<1Evyel&D@s0f7n2wSRJ)sP>1U2A4mL}|Cj+>j!e?Aj;25Pxx
zr7!QDjW3_KwRt#Z1)jU@Z5{OVrkL24qNiy|Sv!fsm}A$Y8eHh(v49J&+KTR#{Zgkv
zABW?Qso8I!Ofkb+E`D9>>6T$Qepy4pj+Ql+*jCQJ`^Fw|MA7ygWVoA8_zn)eMl<??
zkqkx;>Y|u-hbxuzM$`s)Kld{)XyPv^du*T=t{RKUT)jU<1r|tM2%|WhiN+b+G&@;w
zit-=yJXvq`_8UAKDV2)j-3nV&u)Ae3MT1>Q=XL~fAN`G8WrX(9;Uc8p)8%Ftm;iWI
zR7w*?o8^Zzc8aH`NHs?vSm6vY$EwkoQEpXj7avSI9D<^r{eWIn+KiW%ex2;I;MFvT
z3qB(%gH7IBttQQ}GrNX`&8%Trbcg8KNSdGPcZ<Xc@ET=@Xm9-u+N5PgfoN9(x73H8
ze&zAG4dS`La%zIIy#l0_@uz;&(~`WZvuZErKXUgW_%vhk-dAf>`NTNd-=3hQj5-i}
zYf_7R{VVo754t0f9-_ZcE!((3e`~agkud$B<1L)I2ZDO@Jm+s_WxtEUd&uzkfmXRJ
zHRhVW`e)&5Z5N<}*xw1?xcSS)?#<NR+<zrzW8F{dG;g{_v;V~3#1@!R2XvhYCE{D)
z$_sY+b0-Kr6SSk&E>3k?4olmd&Q*>+eX15<W7ds~LtM?r)|Nli64$u1dXuCJoDkz6
zI)J;=@LD$Zx8}E7@y=zuk7S$u!zs~deht;)LPz<aPo`szdONdPc20Ub;S}f>JV%DD
z{$HsBXY(xf&9Enj^Qne`RBHZ9Q37XfZYo*^Wea#+GJn&Iira0_DQ#P<RPwg=T=Vw~
zv8NV+c%h#+D#*wjwCdd^4FfG2X+2BCpC{(eHrq9cZ9|pE*xm%9p?0r%j-U%qcE+3U
z>w(Jed_7a7$J^7EdqMl{{b8426EYG2<DeO&&id7i@Vi#Z6TM4RJY0I@+#y<1Yq-8+
z+Y5K@vNE4w>MA<8J~(P`mpSFUdkxMgtdPm44m84sLsdX!MHs(1Tiq{?-ZBY%3{jS8
zH|0<*&ru6E6#flgI~&jKZYcPBFs`1(Tr<fUm;g3(9cePcD!mFRhK6G5D>~rjv-(Ft
zmTxWO!LBvTE|${jx;t4;&Myzj6W&%Z-<G>cqBDPP=G*F@?j4{g28va|Im7QdN|4^^
zb$KdO+Y4G5EF8MZZ$v$eAg*uzx#0J-7$B?<2zHo5b19&;`8Gj61zMVeIAmsaMG>7!
zIq!lf{56V$L<(T_7l@V$s5rsGoi0mMiMT&_F+cQ=zZcDs<ck$pr$D%yKgN2h;>*|?
zD(cC<5WC(xpEWR1fFM$P=8h_l66|O-rS|4`I0~LLY-zFMS>2x7VuOu>ab;u_q+dUo
zLJ32>6d|?dU+e?yQ*mdAaLzmgAS%&+3WD5%E2<tA<rki04F~>Bcqxya2;4G%r|C;M
z@bDn{zV&C^fI0K<G8dSu*_)U<ALYH0wt)V?FF_OsMYQBpKxJR$7`OWGQwLT@gSbp*
zk53R~Kzx0q-k~|c;p8+ExFNU|ws3g6Bxc7>*zVuxNozrxxI2EuS=H~UE*D81Z|8Ps
z3pdi_$;i-5oKIavXNhTAZr*nbqP^E;Zj<sWSuW0NBFFrn%C_^6CE}{>;>De=qeiap
zd0_M;kZ9smuh)Y-ZGPZi@IK^EourQ%$y$GNC?yV0IT!!Hf_g#q2VHn8aNzakz~t3m
zTO5fZ4!(uhY>Kp09xH9Va(Jz@GO7zTa^>jBdgV-4TmkhlRZP9n%5!v5wf8x6hxupg
zf@-xStiuAkBo0>m8s7X;CgLi%qHRn-W+2#xzJv-7+u!9Ph4Fysg{NxYKIUD3GzJl5
zFC5@`i=n&%Hnxut(=!9SCd|MC_U&ORINLVYQmO7n`(G@52q^f-<mSGBCk*aZUVfpw
ze03rk6b0-A33eXyW`285ez{TrB0uFV$vhg~dZkykl?`pUWuOsoV|I#QVMahohb*g7
z+&{hPh6TNgqh$4ZYbtQSS8gaYs(8yiZ1pb*a+I|QS0PP5^!aj+i!A*xD^#@k=2Oyt
z$0XWvnevbqbfJ0Z`TpSVV#(q)sHJjf&i(#)k>|nBYz`qeJ)H0#MSfi=iNqWBXTbr|
z)s2Zf8I!W@xV*FW6GZ&P<}?eNC3rzt;tm46G+F<Yq8u5*ogm2YHtrzE^Cbc-_vUJ6
z;IWQ4_{j9^Hul<viVU$Oe^C9e*M;!+1+|iH7c+jwhNijSy>m`m{ZS%!^I?)JC79?A
z$*~!Mq@qY2!#5kox9lDtC8m1^c2lP<ev{J^D$j3DY$*JFbp+izEmNPdaF#XrG@6N~
z1ij#-X5T(fMw9F=CKy;;1bWoh#U@oimG52NklhkIE_^HM<j@_18OJJamI3FbLb_~L
z+iW(3waJS+|KB}#l}8kJQg!E=%>-1sAd;(+)Jn<~CzEyeex@?eS!+2~@AX8YO6mnS
z%{d_il!wVX<4^gb%H(UKLEbY)t`B%%^-v+x*^NXM2y@QvNPLdP)Mom=_WR`UUTl1K
z1Q>%k{QbY-fB3WBuJy<r<<TgB<0dG5xLsOY(g6w14ma=`-*&(_Ij_;QY91qkCMe!a
zcw1u;XTLhCU1X1ubp4YF6^CZW*8}>+A}Auv-MW*$-<q$Q<J;Nj{<k!tW|bDCuvXhg
z^Z>iU$=pWTD5GIWql=?=s9i)_fqgjN+X?R~8nh7kkC|?%i$@zJ^;G;mcWWlR54i{q
zy1#wS6rw<fx)Zmm?I&8)5T=H?av<UdVv|0mB`lzC$ChAO!bm?pQ`XXYNqu$LJMldJ
zoG^d}*?XszIo(58Loccl&tzyPoVu7~2ixm@H8t#v<GqmCK2!>$L02Y+p}0)U&wIN&
zj9g#pwi&H6XsP?w({osYWe6-;%EYnvH%4>WkheUjIX@ik#=`Xe9h6(c#ISS_RP`NI
z(mnY3br1^YcXW7AG!KA&4X)c#fekjb;$@eCYQ_oS&~<FK0swcgwnEVDQ_C~z1<SGe
z#E9G?apU^)jDyWL!523<ZNoI$>-TwBQMsZ$kwA#0>SWHp(YZ>Ilw|jDAJVqHj1Cq{
zk(*1z-=PUjaIX{;Ucs81?hYDQS}Tn*pCBIa2~n-+I5^YgS3q4Q`s>g(EAo!bZVKSw
zXOaiMc{=Bx78Y~2D*bN?KS<ixL9(2CD$1|j7pU6IFoW+48p~)B%Pm8V>IL+kl4n*h
z16ywnHM$oGFTqIN_SJoi&G@q{PYz7$Q^-M{^bF6DcM|Ch$hPq9Nt8HI;{p6CjZ9S&
zwY4L2h*Vf(u>zo0e4+cKha4}Q?5843jRK>`1H69~)n=iv-7`igzFYlYdp>ah*hlxS
z0&3rACEv)_>L2b^!&2eNbCgVt_J2q7`d9eRkT-G!S@d03QC0QgLFiMEK0ExQvc7xH
zO&AHLPqk;SJz>1TXS?3s$yr#V-Dk##j4%(qs$<vhL2)o1kV8^fYksTMfAIX)G2$%Y
zkLnmUK4LfH(Mrjr;3l;n{j&xym~G~C&|Au*xPP;_wRbc059kx+HamM>gO+DKUj{QF
zmXa#JYhK!M75<e;wh(}7!zZ-Al08ta*Una@hdn}eorQ$S>&v#cu!vd+-l05r!I9i=
zfSV;I<i;b$<7W~wI)TLhHq{Pb37YQ(!kO48y3Wubm-z6rD5B4Y$Y1<?t(~7v5E*rK
zv84ds%{WcX2pC&}lZ_MX9VNHx;w)VQWP^T53kRZeg=)Ttz95yQ#nQYN(fAktDG<(2
zNcH>fA+(`u58Bt>!)LQTyUi9?R2CLU^tAdfEIf&xQ0g~wWer{GHX^IIB~A(dvtB}$
z))D74C%$z{eOn~d;pB>-8#c^M!v)!U=#|;sWog@-s@8wCx60E<JQZ|0q0n+NW`7x1
zwtXYbRm;_9$jJ5J<a%u#$$$}yW!X0C8+A?d@t3#)o8mb1x8Drj5BdxClnRFprpx^5
zZ{@V8-334FuHmgS0c&ai5MXyKLL$SUHlb|$J_IZO<?GChz^7IKsnCfNTxk_D#tA49
za?Fja7^#z;D?;xy1+p?@)gRKrch4b!v4Qm9R+f1^<*SNw4X2`~`bP9rmeN*RqlZCR
za`@9}MUzC&Qn@9qlEYZ&TK(q_nOp$eEyB<noqdc2NU0>Fws8WNVph~hrqs8Mt0*?U
z!#`DsYoD+-3_K!)EVOsYN;a=&72NE8_Cu@6%asYpuL|I3P~CP(q#SYZTa(!cUIhB@
zaUL+O<T-M`NUzBCaiPej_e!ZBVkY<p;35||ZVFOl{w=XsQJj`rgjfoLv@v8`e=p>*
zvm;dy`jL(!cjzSnNvg0IP%0U0!Z?@Q6K9!r_=qs5?T@$juFbOyCVFN1X*@*&_^c}C
zm;aWFgZ^eERareB0^Q3=-1Sky*XhXpVUeC-MO!)p_rC>uorH@L?AkQ$u5Mt#waTU*
zOaL<0ZA<;~pXRb5@w7$^fCk`~ObT=5L`WxI%cdTZc_Ae+%vMjHRlPqvH}MYHc+5R8
z@bub9QhIVkdqlz1{K#9R1i^a7$Jk&$j`ft(LJOnGk)rGpSLY|5rhEv~&%2&yn)gbl
z!(6>xI^MlAM@U<<BXU{z=A+fe?7<Y%8}tp1L&*;dX0>y!%LE-=#T@Y%?kzGaKPgi0
zmY9I=Q5Q_1)<CY)S&A>~74|#}gxE_nx;n1~41{CGm8yUvod>m-5Y=Js%MU$evsqZ3
z!5}*N7}d8&Sb`o=+bCPVjN{&l8QD7gMjBxHc^K}Idq;HE>NQ^Vx7OQ<&PU7@u2_|E
z>m4S+3(sXg){pXFTEF^*hgOsV>knw@>FKHe4eHbBGVKw1R}8Vsi_F2_`TTGyv1oVe
z8_UkzHx{P?Sa)m%^A1gGXq7TXYLTm`OkO%pdR?=wj-<<ILAw62LSnBcvD;&JGWCr)
zLWzxM-2I5qlr6LHWS<1*fX$tSc#_?r8Pu5Vn~P_Abm&%0N-W<=T(`j)sV>|?Ugp2J
zq0BQHOssEDP;!?wuW@1AQqoTv4zWE8Y-#jT&^<<gd*$<Mv+!C^LjMx(FHaRYHIZWY
zN?!D#;cEV$5`6L<ylWTtV``5^6H!I?gU?jdC+xxWYhEl7p9wV;ze$Ew#9t*oa!NT(
z!kt%AH{DvIu{a=FT<#bkP#%ixk(Fmf0>UUp)v@{a3F7Q4dIC7;OC^&6YqE_nNgk^M
z|83^_M#~VzL<)f7b0utzusn0vc(=r)k>3VIoV;ShZdVsky~*hWVB=x)?DbQdp~e4S
z;Vpcj+ikym>CwKFn32I?(ono5D@)ddqgs?HmS3j)Xxh-R85V?Yg(y_lnh&-5i?*Y*
z9VH!DG7?^+P#Vts#qZ>T6~%fyh=;f;zVe20+3<fyJ>U;;*PSM3G?^herk7Yx$%;&P
zCZYI~NpDr(we91+r9rn%pr&4=XwN(NVY?Emq-ZMcRT7nt9mHW=i6;5=2^dX1Zi5Wz
zAshU6)I#$rL!epJJ0R)lE=fD=9_p=Qh<xdJR7_FnMsYGskMdkCqItc!3|2Qy7<6Xm
z-k!xaTPp{@wEb3CLPYzl^?oznj}Z~=B0k@>k<3`%m9Bm@x9LPwK}{vb&42Q4gUxRp
z-mnV(8p66Bw!fz#dFc`8A}(?_p3Xf)D6|4(+X>{~_d(x}78b*qN~UHsFETBgCDQGP
zgESR6AH7`3)zc(=Z@U7I8VSxAX=J%QYx4dJD&u2#>A_L3T5VljUO|CE^k8Xe=@$ii
zJySl-Yz%8_$uwhCkUIOfbTZ7hb-gPyHi`RAL?tOs<(-LeCNg*{s0{-;j72BNhbv7A
zw$08O$*Ye&<=ToLC<jk?E9OC^2o~-&q4XV_<7!^p<`-D2uC+d_(R0mbyQxub$bsxm
zKd2P9(s{8qkTuopBe=J}ucKd65!VUtimmckQFWTBA^Q07qy4K_y0V@g9`A?JbHWM9
zepOok)HX4>?H1V;a)CVj(BL-wIS$%<tWlrbU<b41K%&d`{+Y2LJ`f}M=ArLLcefWP
zWg6>?({C)hhld|rtXV?Jw!?PEBWGs@5{d?U`*iu&It69(iueXLA6z^BIQ@I@_YVox
z;lipaii*|~-a6;mceRd_ymuis-K0*?mS;r$EiF<s-z$oWw4Xy?-Y^s3IIv^2g0;QL
zc;v-f-h|+#rltMF9I5WBzBR@`rRop=ob?&*ZI<^(&oo5d-B6%>OR~M$fa1vNAp5UX
zBmb@)gv58#xz{Z|k&I=-3CsjdtX5L!Tuhx1s7%ZVP`3KRzGzAdmwqAX@z@hCZ;)(S
z`4l^AtZQTxSyv~9-v8{{+Rj(ek|&QHDNxhV(_3OzN?#m&oJiEz*q9*`^vI=gh4y`V
zx+(L9248C%p5t2A%EPW02s$T<#6Q6r|6g?DH74Z{!$ytcz3;JMQW7rgd9PyyOC}cG
z>npYVEWvU$7K(Dr3?@_b5R@m#-)H?+eV9dt=*tAviiIC`Br>NvPfIPpCehvR-d*E3
zr1(~ZIThFp*sAItYW8uRbNhKS4LgQ-rHA79wpzcBc~GsB@-fyjjcixu*Y$rGNbcYg
zB&iAyt-7^r3}kukDZzKMiNw8|yzUF1{9K3_K2CV&ViDGNM*4%$*=w1q?WknaD1qtY
z->O>O!{*nXtFFoXj@^GBKru@uds?a5XtaH({<g_n*p(fD#@GdLW&5%!#^1JUa^-h+
zE*5x0P05bj#dGUN3=VUY;9W9#V*meW*xi#xFC<6REtqjM_9>1T<5y>;__Baax7nVG
z7M*ch2j`bYV*xp%H1kwsXT#=%u#9p%cPn=vQ=7&6KeQC#E|cZvQ{VvcAJx?c_brI$
zoEKV|5@pXHwqlWz_lzqaM+Z)k?|SEu95f%Hjmk{yc#s9+vIWJmBO0*06&CnPGM9~S
zw+L$gzdZOPY9NTBZOfj_?OrGN?F$x_!`G+t{P9QF9zpHdky}1INA1Um2ZNGBGssdT
zn^2QYBQGeODp$-a<u;1K6^%w4H@Mwh9mtAJV3N$s&&TiT>f+|(`}ygCW2Muy2nV#J
zq{LwH(RUX?W`BwxKEu+s6ZGbX-*oI6%(SdBnSiUZ_EvOIifXAlv0iG`E8vy>U@%+@
z;3WeW8|E#-C;G1kH`jx*|4Wmg#)eEHtI}o-i?JvLIWPotq$C1~1=7xL{Ys9tL6AWv
z7SXS_)6a<;=IosQA?;1y!TmU1-tLY0NNahWXY$4KY6Si~&$)DbzH43#NTOi9f)M=<
zv(A$yugw;00nOZqsRbwAr_Y~*5*O~Q$WZIJ9=43{nQ&@l#}TpwYS;d%ey)kW2Dn@w
zicyt6%6r~WXvqAce%-atru9V5yS)B)c46Cp#LGQn6K9ULUoO6`fdyn0O^utls`xcJ
zZuOGPmlUHD07OfF+D09i9^A-{MFgzUeiw6I!_dc)ZeUa7N0=yeD3dP^XI*!9yIJ_&
z2<S}p^r!}&Y!nt17w;islR;#LZG@>WQ16Ragsed2;6oEezvblh$w3PT;@w%n(&?BB
z7vm^DNpTxB8~&N+vp3n0M+y!hQbMn9{}2%c(;U#ZUuhc(E5XoYkfD;J+gE<1EXYm$
zzW6O?=v*D6ma|#r4&bT)WVLKt5%3{m&Izpnf)jveji-`B2f=dX1cpU2`(M}4(BM^d
z<(S%zLO*7e)dviRpfgVsVq)T*VI`KuP&`s?=!G^wwV6dH`@1s{@kOrjuf(+HM86K=
zn~iOPY<2%(!0)}gp$@~hm+WJ(sW>gnW$|5MEZ(l&3r4{Xd?vTgs84@m5u&3b!rwhv
zZuD)w{gZ_)PV-)59rmL@>#if?#i4Dd$NL|Olt8)2w?*~)qxSS?XO9Q{$g3pmhwd1(
zxOTiHvQEsZ1pHjH-6gECq0z?glax1cHTR{24<0fa&M3t<UJr1Nxk?##Nj-=|#q)%}
z3{zrK-d*U4)VG#6&4ibv2j>%SJiFwTRELT$*(uAm`|_}WHMC;Mli;(T^ivp6NUGKQ
zPH`C$lvyDM!%D&dhfYrOO$Ns+T~sWpG2r>N{t1e8_p>u?d2r<EX`s}_>HaNo7sjU7
z8+lvTG<UZLG!SPh$&f^#RO9P;;IN)$w1r-X{ax60wO^#wv~5B@pMvG1(;6=LLQPZ9
z=ByTLo9RI@<LL3UUIkVH*;_g+(G%VP6+%omE2;m?$lFVjM(ZAE%7d>Thnz047&pEC
z?tgFWcOZKZ`b~EAeltb(!OqBtZXC1p?MOzC>!gxfLh7IHt0-jrl&iBCuOVbX8pm^q
ziFva4{nra{uLC|m+hJlUA_SbjSIOz=SxfP-|G@nEjyNbJ@t{rd=^O^X68Ew1;@rd4
zT(5=2)R+g~BySPvES&q=n#@k`jwkZ+KD{{AY^Yhlw=I16c5GIArnmpo12uFKuquNF
zSc#0NBjpps_UV0-Ux?Q8Mk|Up&-C@*GE4gx<@vKIC3FIZBeBY+n^jP-MwaVil!+ch
zFB)r9cdu*h!{!qx#1>W~5WZU>r{sHvDgAO~&9SyIgPww%;+A2+_<$1hoU$vsq4H)`
zz+|9f1KLeNy5{-Z<FTJguYW#{lUy#DRM=dh-{rel82Z|A8(k0eSEK31?))S(KnDzZ
zcc|2paXw-pu?%XsK`(+o`c31fa5v*u`Cx#0fP;?0?Rudmr=H_Z4}=Yd^;we7H1N+K
zz}zyS&fFwhf6K2)%6sbhUher_XE>l$m0#}c;*@-K;uDL-1uS{FwOudC06*u-ZVSjJ
z5hGYoAM3r~@?r?<#cDm)^mqFXAJMXqPo?f^UP_Rxd^OH%unnzd;>tVFJVKx=>izqi
zKH3Yp!{z%jgz>tWd@_yV-gP*dhIGgp4s?8bH~$XJoRW2VaIZ!B<o<!+`f@z9x{dj7
zNyI%v477xEFKVn!(PS<J$I63UoSe_ZTwt#9RIjwOWN$j$_&;BNKT{tfTZ<LS)D|<Q
z86tUyK)m8uW$}gK+%sz+fAHV|()n}jfi~-*;^4=!pId&_Yo*A|oLJ7Uit!zG>E`c9
z1RCxvm$?95%|0XqPZkNyPQeYSIYKFAh0T>Z(N6s%rf@Zo9&7EMatzi6r-kABhwe{w
z;(|Kac4Uh&{`L{C$@8Gi>i8zJew?63`7ako&Z*qARkU9J86au0ho*3-?ZCpx;M?^J
z-Yiddl8=4&#tHoeXqT6lVZi6Vye7>fzb48&%9oxRorz;e4@oE#=dT+d5T0(x?>=w%
zfx<sQU3mcP5-N;6$;Ki1onSMfLBh%qc0*>zv`GVaMX#(L?r-Hwmim(d4GsA^H#1Oe
zrejE|$;}ccdq_{G%W(=YKH^Es5b23>a|T0OnW4>63!Ak<Z~Ct*WoglL@}b_F6B_8>
zvd7D*2D@O*tVaZ3AC!w_1cOmv0y^Lo9C5@Ne*g(_nIg*29?>HT2qL)hIG!-V2thXU
z4krneJnn+=Cl~2)FI#hK>5T-=$Y&X934T9`op9v3HWSg{(s;eDu9_0<5hAZ+1o$o>
zu5>a*!z70z(U$}IA5f6<6Gwm>|7p%rK1P(Kxv~gghfm}w(TeZizc0GpN)1i*_CLd7
z$W*_?R9-YZzj^LDZUkuZTm9yVJkp*dy8}g|I+>&GN-$CXtCPlP`zUlRi+M;nD&YH8
zt@4brBpB`sJFnf}dA4GL$$rR%m}McGH*)<D=DJ)GSzepf3+^YSfA+A+F~n>N3wr;;
zr&1cRXED>@{#ErRon?@-a~Z~cP>ZV*JZ^Kc)|V|FFmRa^<gG8xFh&G73M_*S4b!&4
zF0UNB0GMfb{s&w9Be>X2b?r@q=TNfD#_>PW(<v8xO^Et&bt+$f!v^5^%BZ}=ZJ65l
z@Cz^2+xgP(j<|=I%}I23l*F?Hd$TSMjaXz<E#&fYEC?7F7);eUGn}8U$u7;hLb{;f
zfcd4RK*wf%^mt^?M49RCR`p;pb&x|2SKc@&y~D;jAk<p@x&Zi-vn$KT?ZxQZ*jl_i
z*d7}0_C!$}{zt>qJ0+W^H+qgWm5>}h0nyEJA-Z?Bkph-`%)FlueaU(+oy$!$&rWMa
z)xylcy(X4~3XA5sjD(fT)Omw%>pAv7rM;yY)oOaY2^7c)f{I%GB?mwA+tg(ai=1@#
zrK!+fBV~+Q=}jk`e}n>Wm$<DBT9&*z`Tp3?TL#Jy^hniX@5aBJL{7=M3A&N?_F3Vz
zk&W~H(Nvu&l0g@fEt<nT3@%;YXpB9^Mj*U<PRBD{jY<I%{ln3+&~E_y>9BjIW^=Tx
zo3E~RbC@5~-MX|>FP9old073TzXKI=5fBK#p}oDmHh6301^Y;CBEx*mU;xt5vcj@3
zIOxRAn)N(>qR^67g?o#@LFu7Y8_4eS^>;SC*WlR{qe2BWf<MAbAxTkAj>X+s46C`d
zwFdKX(8<5L_3c|ZOAy4}1C*uSn{KvXG{<vIrK{(Jx6e#mmbca4UI9Gpi2DsTyFcz&
z&)v1>!YD=1?ZXrY-!hvmbZHMGXh0N~0Q&FU7d8}-IWOTxF+_IA^N*s2#!aaICjSRJ
z_uqFTrxM`lE7Vr6CUNZT9ov7*5Z`ZR6x^XVx;L*E{WRA9i>W5g=>ES%k@U!m?2&=Q
zf{J@cHa1Parpc_qdF9@g6LeGh9-j_b*)4H$^bW;*YX#67!XQO=hrrSWu)W;;F2ddP
z*Uo<rRMI8S12iAts8TKGC!5CAFEr!>(canBCk_Xh=fPjrQa&~{J?i@`PFDSLP|?C7
zTXnLYZBNqH79KTlrl0n^<x@Wb#r7<j4V&visEHCW9%mUM35*H2vPC(&kO!xztXs;3
zo#3LdHF>^ARX=>;+uM9Q3M-FL%+t{mk(+=6Uzmslq^u^yELf4&h1sKl(vj5y_pZaA
zl!>78<29S})8nGj(i_iOMvYhk0Mfj>KPb!h2I9SKVqgHNtQ?vMW*YfU2h?igyi}kh
zkYYPQVv!$0bzkC!{|$raUBg$5bu49}6o!lA>tyRUKS2v<i~hXNaeUptDAKP_Yicua
z?^?w;7BO2xg(L9Q*)Q;<9r(i%)p3yh)eX=^vrAPj^N-9t{%~BfngSoFie=E3Zoms)
zrMeFFOu=nEv6==nB(wkeT{G`#JLqAAx9{t|Xzsq61)aM(epdTWw(LtWCrH5d`MWev
zI^<&zSzDcV9?>+0P(Ie;GNjg$`VstQZx);4Tee-32i-r}kXD3ag1V<;rs`{IV8EgK
zDUF#6Z}eOnG0Mq5Js8z0s;Ds9-UApgF0-wsKWFf$<BdxPXyR8>K2tiU5kv7O1~+{b
z7RCQiWt|XK<u<wbwjl=>c48~p#kX0!v+lo87k*y0{p4VLQ+F*$9`cMh7^o+tzfgMg
zVg(KW+|I0wO|r|=_gmA|go2RMA2CyEB(-k3TUoT27Z@JI)3wX|1gI@x>kfQsIyyo#
zEf+=3fpuX<bXmW#b~Q&Z**$f`t?!3C<!j%y9{;ObGJYWp6`ViECb{Pgo-rV6PV?my
zmz#2A+V`4NF?XLCwPXB%iYQdvgay#;hZ5goLe_S^y1PY#*-0F54u7*H7BL@E4&UVB
zhbCCnQzh49GpB!aGTZ!$YX_W!Xdqk8nzR->kOkKtIR_fl$~Y+bI>{NeJUnH$wR=$v
z#duI93yNO`sXOx@MyFRErDxSAa{~hds~=h2aTXRjL&bk;43bTHLRdnnQnvd7&B5*b
zTYzS7?SQscK-qphMUVf)W<ZU|!YAtkkQ}PAx%q^g)R@W#*P6YzI{^GoAY}G^Y+xw9
z*GK3X?xNe)qvcv#I_xP#zDIqySY<!VN&^nr?V@uoE4(jBmKZ04^EIXij``BDjBR4_
zt?;PL!B|`b1C#`Q!!r1o18|m@n&$!fASKG~0>gg3<rZdY8L8d?KNlg4-RmL%x?!Zi
z?o}ZWUpyVu$f+aeNRR@U(*zZ!Rn;tpr-S@#QbBhjP($xjt-*y@f!xo})v^#CKt1?n
z$lJOD013RRoB1gKIYWTyJgv<|uodNBaz7}hd}tcLpaF?B#s`0F2pviudL9t1e@e|s
z>UNB57VlSj>PZ#5dDY!lsN%D^nM9<NL5K5Dc3N|zkjAmH-<+{eNgo^KUqj+br_Q>z
zP2p|7Trbw-O5=c`IQ>KvCMLU!YG%B@>Eh`c;A?F(Cw6c>u9CvwgICMJh??)|Zuf@;
zXt^($Qa*z;>5<ZIKeMRGyRcq*cDt<!Q`as)C|fxOVOM-Q4Dd*!FYri`<Kt_6YoAB_
zDE%G(@8rYptTlUiE5Bt!4&snE<a^@|k4?m+*WE*_wPRqZE+{SpUVWYCMjqOt3aL>x
zS+KgPd#sWE6zkzUvt}l6(98FW<LsBsV+$_)nCaMx^DG}|JMa_T|5<#Na&Oqh`6UOU
zwwG#+-1OIjNwcpecV`w$nb~%xGN=q5oY~s0PIOJp&h)KIUs2uL!;>!~n+DuE&sb8o
z^@H_yMFGss_y+rY42ZnTJ|ab+inaFTBTxrLJIhev32C$&B{pMh9tRT%Yz2W11(CEn
zQw5O5d`ru~-L5Hst>jLTvL8tI&O~qhl;0x!#5mMJeY##{`-d{qAMw4|+~Urj$PYK&
zWAZuSz*o8nU3K65TN;8sCN@mAuKk6N>qOzPPl6nZ;K8h+6n|d-x->X?(m3AbK^Kf%
zv%^qyc4V*;arO2|r*XT*lF34tW0!)9KP_M~(kl5dS#2KW_^s0SWlHE7Y~PRhWvZ~1
z)$|6a55;DpP>>5|K(~toA1xA^HXoNTv+Xv^5VEiMDj&4AHVz9t$U!MxLBut-Ie$s}
zCo~upy4ny|tg62FwgCnYsKx88<%SIO8Ih6bxB`RtbZaN-L`<s(A`oQBEZO2|$vF4N
zqE}L-Jss|+8SV2BRzmhZOpAgh$hHNrqayYPB(>~(rWP?lq>sBxGxHf@dseSWHJ@>z
zec%*F9#>?)6X!_`G?Eo5iSzKm#{uJ~j-nQ`O@i}NM&AfsoC)K<kiHLd9h=b<?87@j
zn4behJ7-YgYlg6tQc|2mbKX45AoTcDjYH~p<ND2#dcH&JJ7fLKqq5~83Gyt_FiZc9
zf~hJSr8Z|Zf-`!ohP>R1J+ygH;A~=%v&#Y1QKm2JUrJ%^QEci%sK5<BH%dWgslVI=
z3~-&%Is23_*L#TmK(yiCxUf%i6x~@M+#7eqRQ5|#>`ux>THESOhIaFh?_BIe%ED5$
zWU82|l&AyIFP>!Q@dW$s_lvvt^4EXK&29I{ksnh#IB$BBnj#QpIsb!B5NWKabH$xA
zIlIf6!&#}_bM&`S`4wlIvuV%rVT3#5rUkIsopC|{mR<4eM>UAe#(Wqr8&I}DA9@Nz
z&TA(vVk;o$Gg=uBlfztJ`n5fL<yvz?LGSdgVFI0nZ~H`WbenALtWbU-0b)AISgO&6
z=dkYne~dko(|(?*w1I`_&UaqhAj3vl1C5(<wRN7y8u;s-O?a*TIV-1u=<G@C3%WQy
zWI}a7f}-ZT#>#XrW#ZNvu>kbAk#hlzkg7dWv(hmH!7>UV<z&t2<cg1vFD5((@F;$D
zbu|?0Yl2zi9#AdLXVJgE-Oli#WQsQNuU-8xejTaPUMgzre(~A%k(~$8`N@QxnCWZ;
z5~z4ON;VG%yArp`*gO{J8RBWiD`)V%wI^DAFVPQB+YK0K!0wD=1&=$ei#{`L*t{lj
zE5;0;{Fq(lS+h6)fHcPJ(mmt#aPE53HScaVeTUs>i`n<pEM*O11}^-cV!6Y8#uV3}
zMaF@W%|32@!8d`LUJ)G0+8ci=cyRW0+fXM^&k|g|5G1EaTR!iahj3(x0%V=)aY4aK
z*0WCNcLw;Fk)@&GyJR+{JiyA=ffl<1_Sj9AhQ(|8Mn;5l{mqdq;nopvHs#nUWyO&l
zTc6^o(SPHu(_w&Ev6_$#SqNIsRxea*IegLTKLx0Y&taFXU}8LX$PzY;6KtB0bUVJA
z2*l3x%beUFKSt6*$@J+hFCM~Jy2Jze>_6IH&-;_gwUixz)B(@11Rq&bQ&Y<ae*WGT
zMnF;t0Gz4P2?TFY|0Cz0PQBT5kyro1?OU^H<ZMS1;*%gm^P666t~DL)nA0w?-%+$b
z>L;}O3k7<6(7|%e6vxsuf6!IpD8_$uvibGqNLjs_H?!u3Sl@)O?0%o9FA9ulIcva+
zc0SV*MYSBh{Ia%n%1o5-2*K?ox$<652(aR6Bf4Vg2ulw&x7f1iUJPc(Bj?N3KwXbK
ztWjsX72}?i!LOxh!z+ATe-%qh3O9Ti<J$q+q{lS)(`_U0f&7>Rl#)ziu_(wwRhc(p
zOvSvJJOva*deJxFGDCxr_%ktke#TUy|9;MB?5p`JhL~5DI>`OZ-Vod`=G7;P*mw(;
z)BLtpB_*2lrKA)KoJ7oJkt3W;W<7r9r{a6*j2y*Y1d7mqfX5$=h=9YNUphNVEJ62w
zC9)*u<*^=Y&rZH18L5y?(#o7tI>qRHQ*WPZ^wcpid3Dm1d7%EmVGq8_@hXL1XSm19
zk4w$81+vNnr@?G&Y#^jHtuNR6?I6H}cGDrJT~*1M_Z=v^q*bWpBQI1C&j@EMo8jl&
zpsqh2C!c|=@6cuPRPJVpJ@_NchrnWB{QS#6H!4Ti%18Ih%ShHtT^V*O3TZk26xX?i
zVV@8ta)A1=@gNlL`~4u3vbD8^4Lv36$`H13JDh&G)P_f<KUh;!b4PUa!dXE2JYB&2
z_Czpwh+&yY0jEB@#zKsaf?Y%9;Q7()-XN6=b$|Ye<hblokHr$u$F<0{)BX=IiEa)r
z>aBQ)4wY=jVDn>U2FsyQ=KTTVXWF`#jkiP6ldUrQ_k$#pgO8O?-pxkOScaUTc|BZ}
zDgncqh5MCjojNyQy%fE6jc)VVo-v^5lw8fQNy^G9^FLI3^VrV`au(=Fc8e&WfrP{b
z+9I)h=1$Cm5C>?i!LraZV@R;HMewm*ru#4YG^N)%3wXPJD<(jPuE5;{sigZ0^FV`T
z$pozLN3O*JSVsD)0+mF5RG<g9=;+|$9HmcsA6QmYPr0J_eRqfU*J#TxkqDXBmS9v}
zU_Dz|*N+j=dx-`uDr8bSq1)V-I=TdVvKF^KQV^?Jkp8|TTRNclZ@uswi%JBF9>qq=
zCT9)qOYfCbcjOhHF2)MBY7b+LMP7|@U(%)Dy}eH(Ntt%wv4X!|hh*jz7Za1v3EbdD
zMx=LVy-CiIjRN}X-2<fF*!f{8D-6mzHq5VK6_E7cE4SC$EyXll+_bDCp&{o++uY;N
zR!-BVv0cq|zs8JEe)>ReaKagqS^eA*GP&~=@SdwzU~2N%oQu;!I|Zy&@NNDQ2+O*J
zG-bT(W@+{NA?dS&(GNo(fRVw{b6B_f^H90O0UNjP1Y?cwcJFUmUfjMonS|(@%E<8^
ziHnObU82c!u*4sLIhRjxdXxVtuY9g8=P(x&;K+l#*h)mx7|&Nyh3~0t{2Q<4@E`LB
z7d9p6X$lNP{hJ<~C!dbU5<KkTufq&97lr7yM^N<;+ITp~F2BeK*zR%^7~Ys~_6fl7
zY1^GXPTWLAvLO`KFO}LuWW$n>7eYU0P5#C;%D0PmY6+inAf;<@vi%V}Io%4&vWHVP
zui%TLTCxUfg2NPVLZ+-Et!HW+(x`+(nC{c-t;sMMDKV(mSX|6*8k7jJRitV;wrF$Z
zT@`WNJyq|xyROCTH`vnmi=~sWxLiyZF0}NHMp{f(STHzaj+Xeb<4@}u&15Q^L%^<(
z6EXX`?+dNHOZ`56^~vw<pV~oGaAMb0Ry=lR7!<Z^gBs90$8U6Z(7o-p@V4vGJ&8<`
zw17w`;BD~+Fds{VwXXz}>ef^h+d99*-63|WMIQ(@Yti$$nlB)x-P;N1?f^@E`zeJx
z5{$91tEzm$idj3j%|wU&{DRLU=S6Ga?HzCQ?fdn0C#P9em@D#1lp5M3m1IIs^u^~#
zW<*zW{nsYH#0p;wu?x5;QZd7^OZ<ifO0aL%tjVsAlkEO%WJHmX16cXzdDpyl%6$9Y
zLk<LA^2P(li6bHp(H!u!$QyacIEq=Gnv6m0hWPyJ)V=3wo5?tM*L=YT!!LfmyD8^d
z6>@R@a5?Fz+oZWqZhk)dvUJ*Epc13|+l_r)t;Z&`768Gb=(mWXO_63zePp<7XS=u;
zOy%ZQ*w%HIErJRV(2or~@mfpf5j5z3@9K2o_&1{)dERsejmU3D2=Y#zD9t<|Km_KY
zWjbcWXUgN@jwGwKe)3@P2MbJDJ`f8&#B4RFX7m13-3whg`Lj_Qyk_>ad(XlmS%Rko
zpSZ4F1IX!zZ?2qXotDAC;x3(G7P`lcEX$Z<YcamPY5a2hf>czfR&053Wys10qvEj>
z?Iin_n&{1mvd=3eb|UhpN075^S1OsT=Gt1Ckn{Z<Jb_8cid&-q`nokLlLO>py0i@|
zIe9`9@U@ig$~g#w;g=}j@kHi-Pgy{CSR11Ke{%bOHXNH(X`ci*GP*(YdH{=P4)svX
znb^80K_C~6RH7;lI1bcd%-g0mj`rJMosRj9cdipy-5r-!Maiqg<VBqwxTkW8UARz0
zL3dL@tgYV7%I&hd-D;goNExJVkq|WiQUy}$-hS)3kwSJo?~QM!vMJL>zjkcc{l&m<
z#DzkX*1XPFa));(+k=U6kTap*>I_*Hq!s>)pQSPj7Y<~>v-2Zc`ZJV|vx3IHMJjb*
znYfHLF?hg;%u%NtGyW6kmE`V+*cb`~>xc=#+*ND694cNIjS;GjVyq^OP;w(ZEPl%r
z>Nzq@5fKsAS(1gxb)Me3&@LGv&sFvDU@|240b;ylcxZ@xgqg_~_U4xGb9wEGWVnCG
z$!Ly#&V{@}6Q~6kG4Iz-VjgOVI~97Ji6eig14Un%+}_<r9*zYnZvT}=QFw7aMg!uV
zs&gy?UEtJ}8-a$66S4#g1r)W}QO#O4<~dIQSu&|XTtl!k&JgVnv_6Yv8Ccos(35VC
zb?@Eccy)2KlxTxJ+)jfYTLq#|7VwCuRoLj%+7>R(eGWD!1v{&0p2^RBuAQV(RZ}w{
zJ_jlTF-wV<EXLV7XH#Wt+J_GjR9aFx=zX#@fP=6A=BBQ}A=QUVYjsOoJ54W??;7J<
zQd;vn9ULQh7QK;@ll~VLak>ZplpW>pn$jmQFn%RSoZnlZyKb?h*z26<=-T(a8Qx3I
z7ew58rh=nE+{&qo(pbdsO-#0y2`j7s0^e2Q@t_R0lv3wS3p(mCaqCd^ORd12cEaMB
zfYO+iX*t$C5sEw|PR{(H1)y6)7^wT)V3wMCNwhl?U_x>&_Fz|b{Y&^0=*#Gn3&4)m
zOXW1AqS{r})6;8nA{y=oiUD#9q)g$%Lv@vOW#Y&D$~q{yg4Oyu?Xkc8iR4@oJx=C4
zcad}xPVMDs-@cg6Is~D&lfA&c7+2?)O%^|?1MzD>)Gt?_zt;J<BJ2XR*<2hChG^^Q
zeGL|;L^6k|wAb?~v#Sd7BlBy`TVe-^s5oJ1rY`Njli3WYE%P>hT<G4kbh<g>35|Q~
zJl~h(zA;nl_#1mq;ra?%cSy^L4!)puUjTCI@xcALg17%Qz><8JnbMYgFix0PS4+Pz
zCi374?jOc{*A&;Y);+r?R#3v$1{4EsS4ak$4Afq%3L=7-C-YHHsT?L15#WlQ>9i2}
zg?<0W^_f0?3uh9wv)bdhgTyM`n3~q`n>rOC>LCm3KBDsWF%XNHE)HC{WqQvU<fu!q
z(yASoW2ZZj1;zjArGlNEowAbe#Iu(YBh3~wJ!gCA5d9|?fE?+{l<=f{-bYXsYX*1&
z#nRaZ-ES!r%nj0d5`GO4)3MGG9~~N6RjnS<g)}-+k@6u|^o+Gc<fTGPtB{OWFMm=d
z1iOHdlMA!=JY*$44(v{jGX#yooDHfl=yAt%7tw7WSKU%AhVPNR03}E_fCZl~yFAs&
zaInzUAWn-Olm=RPKm4xe+akPq`)vTSQOV$!7Q4K?T5gN%=_1)f%$%n)z_el!1hd(i
zBiMqZ?9?ZkPI$Uhu!}RuMdaC`TBhH_?{;^ncnzs-ZI?Rt+TR%3sR||gKYSmC@^k-$
z?5g&5J=mT9PeC%aX;*yy{$Aye2;33j^W4-X>w5*1;RE1TkW~OZ*qi8Kf9K2Q`qF!n
zwfi8z$*Gf4RPib=F8*;a1iS_@$!f4upphw`qDMe6LMdGzYfHokbo$)I{HlMxK9t)4
zlsR`gskDmjz17TOwRDg%TR!8EJ}vv*|GPtuLpmI)%iZc<sJnA!cP3LOw%y(%J^qdM
z)>V3b(pCxa9#KNBA?Mbq_{H~YXLrZ3Cwu0<HXi`L)V5XsrX5l|4H@@*ODXvR5Nmwy
zk)Y>L)iKBP`dDhEW}rgQ!AqGQbqWw9!w;x!)(;rAtnoWKr-1Mz1E54eRaF%ptWPIs
z;q)v+TYhh+qx_bMeMR--`zEs=E2h;a>O})EFI<<i4X#QAQFv1cbLnbzUO(cFTCzTS
zjc?%P!yF?ZEBT>H<k(n$doI#@x;xhIp%sm=%WkCTEPqqg$eRpeu<Ah>l8mqljxp*$
z{`n@apEO2h6dm*qO7fa{I%Z~!`qUN(@~ESPYa+vYKgnc^dQNuZs6*cOTo@ko^ZmkM
zD%(`=Y%anGO`7W`{%6OL<0+(U7}>ES$u!d_T%8=dfa?CK?s_=TjU>hMMWYv!I5z9x
zT-Z+=TDYG5!sD)NOp0&ff4u;rBy?zLZ3D9U=YLW-j2k@)oGpWjc#cfo-DG$tz7RNC
zGoLjEB6n?~rFhLMwbuVRa!~fdOhdUZTbhAE+~vSjEc-j)mm8qBMPBsXqNbry*pA9)
z$~~GX@C>uP*~vEYZU?Z~tLLiwYF=lKI@UN~yDN!TCALb)YUmawox7oRv?s_bY)CyS
z&}J~`)%qm02~cEf*IAjxqKe0m!ABM7)&S=HiZIlO{t)K7KWeY6$W>o~g?2NvJf+oJ
zdft=RU7v0fvCq%Xe+2M)Foj!l@`)Qc*)dSf4<G)j6fpS%;BFja47YE;C$q$_)c=rP
z=U!q|VTfyi51D)A=x87q%nzp#6ckj#Am<HsB`7*|Lp5hcy^wPftifn)3ds%?uB^8}
z_W&R4cE0^WZY1+<bkKWGc7u%qD_B9gMN9+<onjCR?k-U9DGFkg32fL%`5N1E*_c=S
z>F~?Ap1BO{(^owL*$l9~N3Hca(m6waq*o|xFy{ELd%%t_Uk@n{dZrtTjcBeriF#E=
zJ{fbSTvsPbgC5cn@SWHO-om_1%Ds``(ioA}%>}iK4wIap86Z|UZiR)B?3`>?EH$oW
z>pihJ9Ziw@g3*b!r?3`CB{x)_2Q>nsLF;HaRZIWDNox?cIQXD2(<5fDv^zH4?DSA%
z_l~UHj}GGRZGsA@=W;JzaFWU#S+(r-@cAxA(Cq+i93?<IX{RPDm0k+Lb50l0krB7o
zlQ<a$V41rL;4}iSS>dXNLdyeYBt#A9kvqH_LS|a;(gnmw9C~xIWRTe5Ya`GdiT=nZ
zTR>A(BlH3OEhU#-O!$vU3K9IW5hSPG`+*AWvg9+)IlDRA(&ak8{TC081s1CJj@bx&
z?cipdVI;)FV-Ho|eT*}I_Kg0r(Z*#jK#0ILy85?7YN?V$`kr`|^Q;)UzS46|b8Kv^
z2=XB&hL}b_8A90jXCd57@K59Zsrm09iG{cL;kM&F0a`M#9Liid-INWfm|U|-wMtE>
zV>TuSA0dSjAADIAn7wnt3CE9{>qRZ4EE&H+Da--2oy4Q@qm_QP?Z0m)wxdfhGHc=z
zpd6yGADMyb0^4*uN6=ubGg>JU^aT3nz%m+FZOr1_*bZ#SgWe0gW9|CZEKqqcxG`FY
zCYhIi0balhVQkj#x2$vexZ)AX<b4Yuf;6>fBWzfn><h>8Mn#RJ-*5W>byOxm34e1*
zVJ{VxI+x3`xny3YV<@KLBK@5n%}#hKdg3#syDN2VVI>#-tLI0X6_8)0&#K4X=Gz8F
zgxiK?`NE?!r2Kw-Vow`vw>^_gC2>BsNZX}qo&{VFCUZz!ZXlsr+6Q&R-nXS@zp2)e
zqcap@X%{a|6#)D$d%z^>ZW4(x1|19|P?x3p?zT60OT2t3orI=!PC63<f~`lmg5ZOc
zPs+rEf7(6_!c|f^zhknaLf{XV=1%7M8V=yW_@Vh3-<*NM{>Th*FZHQ+Ke-cet2?pm
zjfusP9t<cCT3t@WAuXKS5YT3G3J^RUBcru_<9vBA;jWqzL!+?-dQ8uK?+;fA6j_!1
zCan~gV%rdI!MY|wPeDdc_(O0n7jdi$0o-Xkf9TgYHE!>tQl@YHDe&zsHH(!d0>BH8
zD-UgwCZ=!g>PWIDD$+8H)o-ml0S*S!Hmm!rbNyn|X1S+RQ1Ws+Axme@`2kmV9eGbx
z_s^r=PE`Fua0#c|r5C5<SP=yaT259P(@1*6lj6?ltfFwl^@tT$;Yb?0Rec2X6|wo>
zzf_G)eo3XZ8>59wQU6ENS#U-9y<J$OyIX4LMi3E@bZ8W5=@5RjbPe4lU4nEsf^-bs
zDxK0H-QDluzt;N!TuWx2InRCXeeHedqdrb?SB>RcCIu48Le(CqO;r?{7NHnb?<_fE
z39_z!+j<8Ss}`1=l9eCkx_vhxq@x`0k`J5{a?s;G?06FU4Xf4+fe2b&g1{*7v?0Ga
zY&o5{<-(s6U`BKQau4tS*?gn4{&K(}59HFbM)kS}%WBhQ3PQ%_{uG77m5IH{D!N*2
zk`+|lD!s;*+vBKN2I?|C$0m5nSzRovRkgTDdmfzl;jtJ~TQKM;L;l2#nfD@~zVHQL
zR__AoEV)@)_T9npNasf`G{i~7txQe-Z4sGm>RM%FF{ppBKaHXuoTP@;&sI$>I_O-M
zjjr~SzQ;bldNfoh?CnuIB0hHso3ai5WgHU>>}ofYUUEe19*;Q@Tq?yCn`w?8x`SHr
zjU(~!?G73@LE>zp|N64+vE16=j^n3{s%pHDYun9a<r6BL=JRlkPkbRQ>+;Ls@kY~`
zlXmk6oWv@uj||RuRw0UI@+2Y9>2#6{>8Gjs4Ny!b%#Pm+EaZ(ngwcPUtblSK^f3*Q
zaTk{#+1eBmE_&xI63b3}ukFPgnfvp@_yK-D@$Y&6{hwKPGQCgKA79P8E_s<-TVpjF
zctPlLWVyaD+(7zh6u6rAc%5SzrB1)b^y|JVT}KVLQa4S~c&l8M=80!d!xBI%Ku_40
zj4i4B--r2{x_3O^<>F;~;LTzahdef6eG$fSd;Oc2VE3p*y0<7yFAyq(QqH2{=fvrA
zmFxBlr=j&jA5zRRilN_hzxEM|JzV9c^2FA7g$;(?E3~Y3(}cyy*9Co|G%eEyBWVK}
zV{r-t@|TU&Wp1DAi150&bt$G#nQp7-iLG?gRdL}oHYPM*;Me~p)^UsV3@HH#?M-G)
z>@8#S<!|wun!0gJTYBX~=KflIulwCdoxkp?4|?z+SkC!T*u<^$sS?1S5U!_+=TBs4
zgErH)G&9ZnU&>cyJ#E+D1ReiVywv9+=o#iBJ`?Fj1|t1Dp};x9uNa7&LkC$o^_{eA
z%?Fj$ygGkkL`?aYMeX(F1{^wnXZ5t9tuC{S9wmximw^m?5E2G?R>Gf{Bpjg;^og&O
zk5?lm*+1G(e4^XF<ooZ}N0*Xh18+et@KXDPyxy$Dr1St5>Jon)6N9>dt7Mr-ihN$K
zRenb(Q`^Jg<Ac(@Ug&Qo{hd?1%^2UoSQ+ZDrH709k{|z-#a^aCk2dDCc$?lXw8@Tq
zlRg-aV|94!I<N_5z3SaQ&G0@y!f;NmGiWucS={(p_8OT+0?c&D_))t;XvitE1$~UW
z_{vP8oOv-t)AZX=fkn4mf{-exXhI%At0(?o&sWbkdfwL?{lfwW!_-Q$NRPh-mefe+
ztHc-UjjLi_{><_Cl3l;ldJ6I^$=E6rlpHN4O8L=Qo5^HyF6&-gIcj6|YszWiTQ~QI
zK9xI--8!(MJe%kjf&{Mf_A)Q_-rLEl?_+YeeB5tKyZnf;%Y+Fj=>EV`+@JBhKg^e?
zg*cWw1ro<7Pe57aKks$9KqlthQq9^!$s5Z`uE)FG&f+cmb<`48-f)Q6-gK$$|8a7{
z7X*qx9^U3^#|W4d%*acc6U|OJ-b=iV+gfB=6O#@x`40JR!{o;SCm~@)NjcV4(D3_v
z@ZZzqJ5kYPiH$LKAyAI-R~?1!J#&bUunODJa{uxK`O;PrR*&-|`{aN%Be3Z5N2@z-
ziS|W)DKAVZj!Le%<z}Z;U8r$B+N}GKWAylw5=mEj6K9G=TpcHAOW4;?Pr*kAj<}n2
zdD*JZg3*wI8f0`r^y}ut8XHzPYm#i-H(AVIV2-BL)N@v!KYUbG63O4G3S9_6lFHWU
zW3O5DcB}rn7TzQuFWcPrcZYk&Xt~rDvsd$wm*9NvTZ@bE5&n6`E=;bkK70<7U9&=s
z?`D7rMY4Q?jq`KfjZ;}}7T9H_-r&cI_Qg5lMY4@iD_2TsS0J5So$q`MEUWoiRJi<C
zf9hW+;K9aSmEtn=nq7)t-8COHu^Zv2tElCBZ95K&kpA?-d+7sQtI~m06G_V!S)xZU
zoGOK}z|mH~WdGLU&l-R~TBR;k1TpFH@vVYP%;W`vp%P9}is1vBDCprv89^Y@{g|h{
zybb1#8qY`BR?xywP)SV(j8l}H#SN|(uN!M@+ds$8mWfkuKqbaH8MPs--{VMF7bWF3
z;^j|WVnYwtqQ#=X%ZrY+7jUw@9Y71FZhYtg|2*Id*EFtoC&$RC&tcCwXeR=@LN_CR
z6n3AQOdjeX8HYZ=6>nVto^pD+XnBZk1_^8wYK@>$B{Zi#o;v`h-jca3Myt7U&kO}C
zna1E0F(2<2d$B)dM@1|$NpbaU2*_9yTmdsynv@$BY!WHkE6eLcE&Z%AYC8PZXmrNm
z)h8_jQ}Eb<Id)K!M$t;RwGt?UB14HL8^)4y(Z0F#-nX*#*x@K+r4Nf9hw45JV~2M@
z22@{Fp^lG`++KN%Fu3cS%8*t1d~6d@7#bLiW_yezozvMCK85;8T6<ON`sU_eV$lJq
zgsm{3r_Fw|Q(8?Z6>P+tKvf3w%QbUkbRU4pTV!2ePPb$<rq7Sh-5DVD_A0I*lO-Kv
z4>C<NrhPEo_Ta3do0~$%AI!jR)Dup8+!z3*b1q&_<r>}RIca3sKMm!?qQ86>Z@Q;3
zsd;_7agVW(;<5R}vSih!D`R-yt{r_Ka~kIC%(_!qLv*0u=J&htXMR5By#|hXW&O9?
z)1^|7J`s-#H9nWPFWge6H)q>~M@1$MzKMEW@2rvIUT+ggvtIotOOaH!>zmQ?RWWVB
zv1VT+BB7F6A`Es_?*9HAc94jid^ZK7O--Jx885AtiEZ1xOLw#hSZCzy@u{)m%1x)F
zU*pF*vILPo9!L8y{C=rLN1nGD5!KgeZDJt33p|I-h9bIEq&^xSPnG{Vq0ipgi?NVL
zmGw*p*PaU_5)FZLB8|3{FuO9~9#1~E-4Qqq)mE6X&=ehOj0${$&iqf5;=A<Q`deRF
zBFzo|4H#H0E>2QRm9Qj9Rwrm^rqh_*ia{0<9r%tFh}uj3lW>7CMeWQF{~3hrl;YM~
zECv^7pOc%e*Kc|xGw>3O#$PxLbh8$w`GbK}YA@785*n_{XxUNw5udxs<d6{&vNsHT
zVRbZ)#G{F#EN-EP80rfp(QA{j7tx9u{k0a=rngXEzYKL)cPmTbs>z#bz~PSUSz0M|
z06illlE_o3yY2TY<fV*CYM3&v8#6&LvCDZm%9r*Q_fxf!_YjD&?aBIp8<=_3jnWPA
z9{=UZ#XrZ8W(`dEwQV;lj#THillM)IWk#;!ay0`<*3rjJCtX;p062R<7^(Wi1XnqV
z#4~Dx+t<T_rOsumX<K*mgOsc3GUZnBqZ9f*F+r(;IFfK9%)-g3o^Be4fPXPIEv8ki
z?l1Kqx)l+{-92~OC-(4iJ4#^(HB^snQMj_3Z1k87d7l0t9%9K|wxIeM<erO&JVU(B
zNAi|EGLEeIM6k)e*;1vDn7Fh7M&bBn1j>`h;J%q&2)WFr#ZM5o@H{~t@-ADU-<0?K
zDJUpdH%+XFA&SL>kY_U@eAx2&I+eN+a&W&%`+0JX>l?pSrry%0p->Syw!D8~HMNS;
z-Z$IKPpPe_$Mi4r4ft)h3x_HPxGO*jlnXCv^)!>~&WN=tl8h-%@*=40ceblWX*3~_
z6MBuOpmw%Me}Kv`^vMQKsTA9w)fofsWgy}SZEYp;9DZ<1!fW%Y#G&GhR)DrbWeA8Z
za(SJ1CSor2a|ZmvTjhkC)QVTxh>P<V)EKym#>dBjgJ?E0_ybK2yh*wHPOmpJb>MYx
z*h#5=&M~|Bm-^9wSeW0GHN&+)D)fN$1U*cN8qDy0x$*T2E;@r!*$kH4$eRRI+j|bP
zrpfu%-TKj?a*gnK|KUJr90<+Di5!(H!$ikhL#g<zB!6TZxi%-bSOhglTLY-^Nc5ns
zOy&n7eAhp^$Tk{)<2zn<9kxQ^F-2bPAN)3s<Z0~i#lr>xsxsQ@aS8s2KPfeT{=&VQ
zw|S|AU7%iGUY$|Ow0SwzXW#W&Vf<5-29->_d8NK4>_EiEUF(fhjc0)jq5|j_EXlNd
ze0+H}a*<D3#&=gUG^YE--jh|4@`?(L_)q`sUyY5AjR@E0o-7H}&t|4X1r^o9Tmpn0
zhSfYJ2|_?d>IEV52el=>6-dviNhj@O42-L%Cv5*&;_$AHZL9iqs+=%^c?{N&*3bX4
zT@M#ukQ2Pjv#BEz#y-D!J#dVF={FS1jJEsM;@Z;n-s>FJfty|wqR~xxwUJHu;9FI>
zh)Sr8TmOvAU<?sC<zHgw#t_4(SqTa@1(9wnKh;SOKFfREyVbF9R;`#VU!v6bh<4MV
zq~c3ICk398X4>Lh6}6-DX#>r8+OhhFn0GiG_%bj2fMr$mN;a$?wd=|EX~L5Y_DwPz
z&Ry6reXf8}(BB<MPQ_at6N@r4pcyKld}ySqntmw?=v_IiJi}e^#;Yj#7^09DQ8{2z
zdA6G@=X}it9AHcU{DG`XA=5|3&*Dr$rq#MFWhzCQ@Xt>X#d8q?bQ1?L@|+75ysLlT
zZKDE2!o(Q1Xtk1ztKu&y>0-sT*W~Ui2tT~>TiaFQHt+VTd^1<|W2P)}9+lwSywVr9
z{Eh4NfNPt%C>B)YtaEXLd-FW`P>?)@j4TT|1DZ=HvM@leUly1VD%-dPo{JCy_v*5<
zsD&1HJ~lQsH8D?iK%8>0S1=Q-_9bLIj#+6jaqLYH#B8|&aK33EsE9dIt@|FRh1uHY
zpK8C4(~}0Nd{U}L#FGF*|KeKpp$5oHyi?Wre(n-5H(<!V9MZO6^ZT}1;TOhE5FffU
zTG)$eT}N#1^HPIsz_zU0$o5XQJ>DON?&X@;h8-n9%>K(fk0ak0h>BZ_!Q$kMqhosR
zb1>iN2DY{r%@=r;UPjr@GFfgiOycWzgpTm5Rm{mlgQge9&0nDy7R|6<FZ+<3-+zY`
z)uiMRTspY7NU7)pO%#<KR9kcv6kU2m4u~e3Y*3J(kE;M3XVm5GsJKdPcb`p?11G-a
zwiS1*8Gc;`CLU`JbC6XYn+<KA7~JiKfE@prcbAhji??_M333L629-p47Mjpi;+hEE
zU*2~E%o5Ta?`|cNL*q=?zoE~+zNFFrE)^)OpL?!dCoq}&dotL`b%4;T8-x&JuAsdA
zdF&B2m6(a(J;4A$uFSJ3%PURw>YI}SLSTVnioaE5tA^C){O9#*7e2e?^^)3!iaY?v
zOEKElEQ^(MNuiB4U6Ey#eG71gM0D0!GF~mnv@8Kp?4eT13(~(CCMP}B?dgj8a#LAn
z%*+e~p|x^3mt&CVs}Vv9`%EH~l)(ZKb0-h`HPEOby}9S<ggdgN|KC_T0j$dSPnuwX
zPSWSw@yv&HRc8%&k^epxk4b+_;N_{F>w+_W#EvXRfo_>x?I_TA^z&^{B)J%RY20W*
zcC@*j5^XBv@bctUVyh98b&sQ1L%Z7v6FwMv+$|=x7RExkw#gNTn338;@#)owq<4Zs
z4Kyvi&0F_R^k_4&PxR<kH^4mI1vnX^`0RMJ8ns4TwdF}HC8mV<ezT^UFb9~*$EoRh
zTLWi8dpOWi6Pq5la-&hymo=HF8Gdrb#wrRGyA!^LHVoB{f}N|Fm&3nUhe6T+3RnCk
zJA7(x<|EIHFi@9pAk1JNWBnVLa@u-I9-<MoB$z4daW@7hS{x}HyjtY@ZJ4-mRFHHE
zGOXhc%En~dqV9(Y=bz5DHoBm>3(v^lo5en3GPhdTA~P>RiqF3kGdmzt-nh3dxvgQ#
z%ngLL@yr)VmBq;Zd6LXR(LHb-VmVloRqW5eXQsh<i1YSHe0(F{FzhPzdDyL;sIEw5
z?dReVTc%jh<q)<Q(Id{iLte{V><_)@&J?y7V_>8Xa9>i8j`d^)>ODT!=-y9GEfa#5
z7>op*_!?n%!UIyIdPMj@-b-ViokAwn&0^XEkfyhKat}t>oC5b1R5~Y+_n4Y9Pe01U
zD6|OISp}MN!z;*LufWxa6hX?Z;ls`=hmQ??7A^&4%BVSfU!@luFryueL$x!5-+v`*
zU#BLWvov0)q<JXJ+Wi9SY3VHaOPARb=fysqIJ#O|3GW2CNzO_mRlMtV76N;NB#u7)
z(NBjYv&S2y^*Kr*QT_76KXj!7?P<9kAhdiVAOoc7EspzSMSk|b$O^#0XfoGrq1}!m
zzz-);=@N%KZ3KLJ8wkR1s;oXYkT)vbAOHLE8*SHLg>*ZaMJnqHs$TUe$TSQ5s+^_h
zcJU`(1HBHNI^%E|w(oy`s?1<S=Mfm=6WG78oGAJsDwZd^wOyDQHU2>ud&%=O^5!l<
zQejjaV>bwL&;!|Ux)rX{mhzDN$mGQgQo3C@!dR#Oeye`Aecb^eKFN<U$bk?Z@}>}B
zGdONU2_JmSSsajGcF<$!z-<iMP^!a^{DKYAPu*H(>G>Sg)ZHKOltRsf7cW>T`aL9h
z-wXrBdI-|%67!WH<wcAmC&@hJmj*gZh=G3q(WQ4n7{LyOnJ_W0`fqIW({0Tqu01``
zQP@&~S90_ZS9kgwX<&A>!E(*Lek5<U3Oix3kO9`pv@}b*h2GnewFc5h-*VK6xuc-U
z<<=>6cd=447s0q%m5izHrG8^h`|~=k#OI%u*$PMj_YEepd);L@jGD6u1?<@#J~})S
zc^8+{Zr5qZQnrYva5XJpRU2mTbuRk6hexZ`zg8KEPmmmQVziQlgv1%|CD10GzW}Nh
z+YQsK5jpu({Kf~`B;Swl`aHJQ_<n{k&&wVRa*r_2!Sxzllif?xnIA^zhhd2vk>)1B
z{g;(!kg7sxbx$(VClg}>jHNItrm_C>AR=j|;{d0-7;Oq(R?><e(~!(o0AMsUMa5r!
zdKI&mZOGZeDn5N?5<Y)uex{av5eiGrEV`^27>|G@hZVVQONq=$wGM4AjgZCIQWKxk
zCa%#d`!bWr21;@54Gs>fF&K=wcS?zVWC)KjSN3oh8${j)V<_o(Du#n!okI_Xx}6dP
zb+vE;@-rt+ds~0z<2TfwoIfLA<hTPYKXTsIKa<jGd|gMVNa1a$OjGs)_e@qDlm3!{
ze`~xjI=o0h{)f|YC=nzyFDjHMGcwpb*7BqSvM-trg<q~BT;q`7bHl$>ExO}XwwjnQ
zvO(k-`5<xqfuU|mjM!qS&gko{^ir4Isibn<r`&{M3>Tr9`n)y=;4BUOfFa+zN%MFm
z<J`)f`4|heRw2lPh%#ud!<&kYmZYjb)ZB){wlYGDFyPtsi=21o_78SwYODsZR0WAv
zlcu9&t=GqTVa!A5d7&A3X2Ai4_Zc%Z-AuFwZmU(m3cH=aa&>mDU0b8g7%j~7)^_jr
zmv~!3P$dX}DD_c7o0Vci9)dX)^ee!eX%ydNOa;UZGFw>15}a!@Nc~|`keV79nKQ0R
zp9zNnurMi|v$s#yPbm&vZjwkT4@#e#D8$vNA3h}q6^sisMnHSzPR?<xQKnc%5Esr_
zhioyFjHP@_<8$}YcsKd72rcU!uE;Cnzds~7TzOM`Hq%|DorG6gUFV`C{oe_B%G#8g
z7yKBy)e()jJ>MlA-4DIYxcbP|D3%^?s_0U?$8Q~giVBL1*1$=_(FZHgfHZx8*2fcv
zO)2P*;WuSNetN4I9)PFn<4<1#1Oh)+(p-Yr7|fz4!8^1Ykc}^PORo=_57~TFXiYqL
zNA2eA`-<;lB?XNo@O+vI-^aRbnK<NdisQP?oDH-A9&cw=ok@t{XI5SwG_p4ReD^g;
zf_M_<vy_pzUpRcd#I9XsCC)RYgx4R4e>o0n+Dhr|!e2?{7Kudc#F~h=Oc6wWH3_dG
z9~*q16BGww)9{t0t>cvw!7gxZ{uG_o{Q-Z>oPLRrSCFeyG}hyzLfyJe{Q^nVc|@f?
z-0brLT(g38{YJ(gX$%=cAUo;3UEN-1TD1a(sLGq~UwfQ`^6rNkGvC-7&-4)ecH6I=
zwgW`9jL8DJJagDQv&$h5c5^DF&GpW2U<`v|si?}uufP;tH`DgHmYWbI58sx;kRwg5
zY5d6TzyRS^wme3YW6&Er%9{-FtWt=4hJQU)F8*J=DalC~bB5ioC3*m?iS9A5ONiDm
zv?@dBfTY}Z^CO|t6{u0_KYys8yRLKwm3#}!=ckj+ysKDfo=k?HU10b0Q7Gmu5zZ#3
zn7R+^D5s`}{CjPq_slH!yu(mHrP2Vf7z01rMCb;5p|+-yyd<5A0cdwm3aqik3h(Ik
z<E${Qo8VwcWajLG&X@;XWFP@B=*4|$TZ2A65B1G*N~nN5hrCD5L-Qv>Rt}dXy?RE^
zmgR)Ta7H9AD=A4pAlusLHFW8*nBw|x0W`^{FQhEYwr=8nZ#KP2O4$_IeA{QMMjFs3
zL9pTH!%aVB`Rx2q)(8i&A#UxbmW2^_M<ZzmriYw<OxEkvVfzi#fV9F4n57gVO2dD@
zr`2<b@RkQ%7iN04bW~Tp5h0gS1<S(WFh*oCS!Hf=E#DMsAHxL+C>85Es6H?s9rEp<
zQC!DBc82l<pM00MMa8!?!Eo~n#luqih^gcbIcSg-k6@fngUETS?rfa#T)!*l*ipcY
zoFh&3KYYbGwMo3+;cDkBgP9L?61HT$3Q}51*m-qW8tZXu5M2B^W<LO-gT*BCje95(
zy3p(uow)&-iO<Uw-w&cjzC@<bz1%lfehzIvnl7?vU~nsHrcWT~mfdfnkd^l8m*O|t
zuZ>ora|D;6^Qsyc@McVJ!QGwoPzwqhW#1Hi#>lB0tey5z)zg;KKkP8}mps0vSN!bV
zXePzw_Av$jh-*Wta8$Hcat=d2M+r(#)XU$Y**K(7_Z8v|BN*&Dq>v07EqA@2clq2^
zU-|zmKwkW6vnsapxqjvA{27sN#Ue4=r)xBiZ&YrBz80vU__iLgYlNT6*dqosZtSry
zsJAy>S2Q1BnA+0qvCpowW2X7U+N+j?#8O(9X^P?zNorUoh*L_FS%ye_+R#*&w+Ztg
zv)Qn+F(W;fsINQiVv4*@R~B^0#Js7>?Y#q584HYD){G-x9d+z{c~F-F?^2&NRrsJG
zcZ+z3bXx*1qECV3-214ND|woBrwaEV>V_fb0=c3%P}4D4O%Vf5NtKdNC8p7xk$bMl
zAJC7L<E^}_?LX#g?5%)_@MX<gTu^<)^U!07@9z+A$1}nOo`#Ji9mrBV<r1m28CiLd
z(Z^pl$DODM8BeKphrg4p9<D|eWGHFGaY1;O8@sLeyf)h=a?a;QAd4zz#ye42p|iN{
z?~qW*h{wO=ao8HMNIx-}ocbqJd&qc~N_j}k?9Yqo>0`tOaO#t^+5miq8Jqf5(T}iz
zElmRx3^*Z>1}+qbDkFM&T$zVcC9Q2V-Sk@5tWv#S2n`gzr0asF0Y!$7=$HcONI?Qg
z%zWeqySE?+z{$W%K`1EUDfPD?#>GSF0zNU-74OTdplZ0~Gsftn-)IyGe4R9*=(t!F
zAOKsQL#0O~_(YF|Ee@yhM~JDDh|nE0%a~fpo4%_S5k8^zYl3mZm`{AJM}!LFY9qFk
z!|-*<nzMgn2NqKD)JEKXB_n;F6!OqdYT#MLlYM*O+NP7x@f|7Ht$~F-F_B&K#x+{t
zMSSZ{fSqwVuc`c~j;8tFVuzYKJ|-?V?YweWq8u$_+dzJ#?!?Z{PyKNdmjF?zpQ~rX
zy)ebv0G83K<2CwLQl3N%M6m~UqfigtvHrsM-<jlp`yb>lO^Vl(Ot7K;20=Yd^#w%Q
z&d#D~9Cmj9QHdd86I*FwPgIblOKY#knLhK}HyNYSsOuo{Qzv5ARwA~v*$gp?yw=&u
zYSm)&bDw|x*P1?UJp`AA%d&?SenciR8fzScfJT;0=Z_Hm>>JjKF{1O_m^1bGb-7KS
zA%gag1*h@%tDVU7u|-nc#MZ8OoMFIqlX^L_%)h*X#6{gQ$Z41MvPgyF%^&^y1O*CD
z*3NIk!!<y0JKYP30By0}QEbz5e|FV5Z)r{&u{P@3)+Rg)w2M*9oJ?MINQiBz2&7?R
z67T2AMB_LuxWtI~QKIPB9Y_!Rd@0?I<{9$<-9T<DF$V8xeNToA9(a#9g3U7ArI9SZ
z*|qaKPDto%W_38zJE1*~#X9$E;1)%qcque)(MNR0@nUC(UF!2&pYOC^J2g-IX>yG|
z92DclG?*WiR0f6y_jf--RzXeX(1JjwK&=lR>ptF$xiD%~#4X-IKMvU|^O#V5;5@02
z4Q4Os#TZh_K5`M?#at0zm_YAZy9H3*VIU*20rfQvug7jc#>ef9&-sNcE1imx6xz9Q
z!SCcG+$oomRk4{SSev5<h6krB#x+a&N{weQD>K=P;Ql7SiX!TjPTMw8g?d0I@0`q8
zQTFJ+c<k4brrUL*8yh0(^K+b?5=z4k!Dvr>7o6k=r~-7`T-wN^5qH#S@dgc}CZ>Lr
zMIduSch_h;Y+pur_?b)U@bO=A7XK;1JC76MWct;y%EhcpR^cx7a;O7@o4KK73mf$@
zHEz(wW5!+^FzmRbp%_S$H5eMuohaktW`~nV8~I_T8Q-od@D%y&bD)@rWy!SaP%@I8
zZxCL6BP|dhyX)vdgNj}ZaMEX|{pX&q7?I?C{UU$&vwgV5%i+1(aOqtUXlu(-(x2B8
zpUId!mnR8o8ZP6LqL<#jiGBPh(8fCS(0Qbk`jBn?9FACYA8b^rI;0l-{7}?)t#!Ae
zSz{x%XOkEqDNFBQ#rn@G+{zJP@yJ{U*>q$Q(dM&rNwSPOe~A&tPCio`qgm`HzGmr+
zz6Crh{@e+Ez6=Ha*oaZJpM)<@P%9Frq`X`#G^s=Mc29MpMYj1;ehvCMAX$zONyMA%
z5g?9egp8gacvw~$_7v<pa7!x|t+cUH*B$>Z);^+@xTX}Sl<)VKl`>a@e4^l`7yp(S
z853>i5%Q7nGa6}mOS-IgLqLq8`=b;3h01wC84~1ku{XP#))m7-6IiJflLpj&_4Jp1
z4$rZgP|-yr6sMWK?t1E?@(J-X9k7FjHQdlSr7e+e)lfsClCf24=H@3cil5Xd?XA0m
zT!AMD9cR-=zpdW7R?bb8ZAdoF(tI^Yu}3Q(C|YCgwNAVcEpH{<HVz5Mjd8A@9mMG*
zIZxPBSD5&IBEqcA%}f;VOtdc1=$oN<vF7K`pR;7fD;xTrU4UK8VAPdJ`UhE!Ow$|f
zpV==k5ElUgSc#|aeQ$4xoPa5m?y8VILo2S<e!20L?+==j6{yG3`5h97?wd7a6V~HJ
zf_FK(r>BY}LMUDS1zG#=b-tz^BB%Mqhs}4CE%KQr_V~}bvb1;Wqqpp;?qA!}1)lXv
z1l4_aG~Age&IrJ?Z9ON^a~hS>8qN+}iT1gp7Xfqs0cO;vr@v-(b#Ql8I7Mz;)8_r`
z5SKaTd;xAjZMukdrf}uI9ryO)Yb;f2o~&@8GumVJgX89CmkQ<Q)8x@_PeQj?o+0X)
z)an6jE60@M&!b*YzCtoeQz47bjP=joB@bK3BH|tUM`Tx)$k-7c@%ng|*Ln*jfuE#H
zK~1dKt6qXL=Vn*77f7|4)J;ZJb}N!Mpj(9vP9<@ge&{I!?YMuY$^;c&me$4KA+$fG
zi=o|r0Iz-mi>o)}`XI5{**X=;Qdf8SUBb76p)%FOGt*;go0B1TV(yQqS5#5G)NXU1
zWp7`CQAv1KkH$;V1RmCaqX;&hjv#>!0CBLPR4EDXzlyEEc}&+v2^f;_=Mt+GAKBpG
znLPdc;@1TAQK+OqG%dQ&ZbA5f>b$n;L<HtP^U29(Kb!quYu~W#X@Y3PC!C(zS7Ps0
z4M!00L&*95Llf{8Q~n4I#g~u>%)wh;jWQSdqR)CB^o&mw`j{T$PlhN=1{If|$Mrr!
zY}&8;`?iof*cjTHN#xnK&NnL|I5ZNRkyu*l%Jb`#e&>9eKirDrfV_Yp5dY-~CQ2B{
zY2K8$KOQ!2Qm>r)&sm*~h2WFOUsy(F^?2nEar-|n=$E+saQPWJFu_zCGfALsDL*gJ
zy&4-eRVqI38toxmM^A+l`-{c5$zx(ms+i<Zv+X7kuk{J}d7mk&mlu^inL&fc2){cS
z+8*|K`1uJZXS--9a3FIh__6iSOX<ub=m&02As-5zgES{q;A_PbeYX|lm8{Hn3A;NH
zGhDv{A{IT<vEbV*zwTHQnqZs?aN+66_~xSI$KHmW3?nwlD!mW>D6=|2I2<y_X^)2u
zYdCX0kM~kHqJ!N)=}_mRF%h%D#_Sk+Q$L3`cT8his5p(%JE8c7_U>C*;>4Y$1B2f5
ztTdIyVUcIT3u~hTRD3K1svr_3SzIXlo7!{aLDZ2xvU2Ma!HUu>Mm!XBJUJOLZw}Pf
zt12Am@JqXAh^@5HR;0{H&iZ*3v^zn^&6{|b`bgy>w}XD@t!J6_kU_~R^S$U2Q4p9U
z4h2JR_I#XS>*O=wWPQ=(dPto@AK|$3D}`?xm;{mSQ&*os7l;!W21bC+Ad{@?fS&F!
z$eOeJo}aL$fkw#5T%86tF)O$1i;YK01!BbQi0nHAb$F35Ell=ZE1Tha8Ft23wg__*
z6YD7$Vs^{oLSSMVto!*U1!&1c;uxZ7llx%S7OvRdy>7?F&cVk!1^atbXbG*bAt3nB
z)MgZTWF3HLJa%N%g^|G=?tIQ@iZ}BwNXtlc%C(8|6S9G(lFRi0JMvQaw2qRo&#ooA
zMYz#7NP9c);Glnn2q$xcw4Wen_fn#!z8t_3%I3V#g;Kv_wQD1AdIkjch;#)?tG?GW
z(oij7*KPBdZfy3azsFWKsek2$hV-qLRvn&@^u@bf$mm4FB4#WK-=#1P*^Q`uMPTb5
z2hoI-mhgE1{`H46#@f15LED(}{kuOpvSJJQ<&6a&RwgGm3zp3y&hJ>SPUO6N&NnHD
z-nCjD`2FHhw8+Z125@O^>*Zo|Ni@TsJ|0ggn~Me4CB4KyljY^*eq7x0T9AAdS&ej>
zCd+22U{+Lq#)I1l<l8kuCiB13%q#HL&^eEkW^p9C&bK}fch&iI;%ZOjYQ%@i7%v00
z>j0K21lPAIwtGT#wguX*BOO^KKyCP)B4r$=clId`_1iaxa*q+O73M?qG;K3Qg^>+T
ziOwlD!eN45+#iO_$a<eK`Pk)F)=_ve$0&#J3SjAK>LgR?x*{WEdE6(UN>iDnAcZZF
z#9?(2;dr}SG6@tYb&3#B^P@~&X7x5SPa>!+wB$4v+1yUm!x>+k50PKLi@XJEHd#Yy
znL*6+t*Ks|zgr2_hbZ#cb}ygD-p6O&1!j0%%vK+QSg(B_Rr0xh5<9XOMvgvjnJ)oy
zN>XNXo{H*`qm&mN>s3^aKG%5mqCVj&93~%17K~1VHTHN>-$uMi@7AIwZ0`CR2+b=s
zhlLUE%Rg3p9kgU*`)?FWHQ?2|`Dw*3iu_UsyT&hfG&c|CBIeduRSO#6Pm+JcsCSJC
zCJf3?v(@Omou91Xv`$RF+F)lFzDWPv0l;(^d@Y_?zn<mX0P3U=EXnc{?U|s!;B4k|
zbfQ=0Q%X(f?x;)ZMn@3bi=pKbOR0X;LRNF!#A2z$Tr4<Cu8ZLv>U=I}Y(S2-MO?}^
zHOzXA)%8g+T}VCcb7jM7c<LLrt=MR%d&?~=Y7+{P3NcHIkMvs<*V=4ahTlSC0~>}p
z4a-$v9(j#2kq!U$@t*l@GS0=Gpx_5?iH@}mVk4b_NV(o<ZqsIlN4NwIGx+Y?!dOBz
z7ig;KslCpWjdF4+NIUV+{#<+i-7#XIpu+i~-y(Lo;F?)^{qo}}AezSjK*0*zxn;C0
z-w)S{PX5cw6%RL`&2maXqXn6wiVSOqwr@?aLWJjn&ckds<;G+MEruf$OwO5J7b!fh
z<t|U1taL}Y)aBSs-7n>-qU9OQagnKK{`)ra3D@3$uT2zjCwH{14iDD5TTH*7{;Fz-
zt2s1d*X6^TmB6whZ$@1i%}V|EAVXBLW&Re0|0yRvtzT_-a1R((z0E!=@Ul-@vl4-w
zWr5<7tYw?+tu4IROAzoLPt??tda8XX%IseDTN=FuysZ{i+SAa4oPVKig#qWS666#T
zZ}3}dXmI9a>^a5~MHX|_2T8*FR8<W4zf^xEt!jM(&`sj?1zkGq0Yi`cwVAPHEN9(}
z|HE0K7<wjU!dG=Z;P|~~z;bT|nyFIOP-(vYfkd)#lJHuBQ4<~~zdG=V;4(;cIXokq
z-e1PqhAl2<KI77Tbp2~1pd7$F{*9|?3u5#Z<>?g^o-SU|_5gzco{639s4tRIMtHv_
zes`A)$(~?aHl=Gh@Vg(5_L5EWejT!H-&b<JK-6a_SlVA5o?-dkfxazLo7+2o=QT;R
z)74cRMl_dB`3~;yU@o{&4qx5q$CPW!>k%GhDQkj3_-_sXHP*uMKK9%aUEHUKrrW#a
z)ADV|U$&REzrd!r4fxR3H%qxrSUGJJf@@#-pazdcfJV`AgG^$0^O2u+i2NaEdX>7e
zRac=Ab?AGuQ@o-z)|5d~SC+qjYrQIj&2{pGi3ow%GF0a++E^GNbNHDJ>oeLtcJ<pR
z7pJG-o(TA9Kq6aJ3}X;tPi*1WOlFvN)=!-+O+Yf}eujQClY6;WqF%c72%tmVdNSr~
z$@h^*)U^8}7Wa6YTRh8&Uk*?_k(ckH(L@Jv7kwVRRz$zH0mcz;_uX8k*~C~Ot3mOf
zJFq~&JUr_bWL1OV){d{`&lKB+ADGQ<%G{}<lG%19nOh;}g4;KBn>RhBeRn@83^z>+
z*nN3hXzsUUoIqe{6cA)V*X$^jq@3Dn{2o3aC_WA>Op<|At1B|(0gxk;dTqdo=S0p_
z(FG2kpjvP;hrCi#TBq(ceAKV?y&!)ReB(j38*B8o`g}@f^DRxH+5}jYGNVLqw!#~`
z06O|Fr;_W}R+_J)C|97}EkG~zM|kn+So3=Ko`g@HB}c34fQg?fxd_hA&X9Kq0$4eQ
zF{r!m<T@0%TN!>OYm|M8<ulOwb@E(ycxyDb2j1kTeY=6P3jfgr+qBqsK5wj?MQVvC
zVlI^M<&e4Gvx$lI;M)Ob4P%wt@haOQEch4|MItsan9_0<q)(Vpyy#M&_|{l{5_Ld@
ziNr=;0K@c^gASP!rQRNU(~+8R1H9dgLw4j(iqW@^pGZp??ZsNET)|Urm!W<9Ql!yS
zG+)6^Ez%!xAJi*JrMoc-?)~V3R$gVHkl_)_P%ucBrYqVfaq2M5P2U4NEKdx?;3|gv
zXqUWi_Yp;Sj}XNvp}H@!IaWbuDPHZp9V7KC^N+7U`~=OT1;zf5|GlaiYPn^-5N$$R
z#iEF|ZT@oU_}th@A|uWpzAVjMw?qRXM9Vnx0(O5#>cNKP9>meFYDY?9_iCY@3E}U*
z@z6QQZ1`hx;UR7*({J_Pw>_L!_V0gwiFAM41oBfboOM1dRf1hhHV3hWD(1<)Rrko9
zoNhw)f1_da^|`R$Z7e-<3#`N~^Wt}a!>+9X&uf$P@DM@^8m;G3u?*dyLU~t^VX$4W
zx#-n$fJJvGe$LzH!`ZI3@L21-vBwo$pR6wnox~12dQhJ$5WO|XvxdA|0f<JSrh(UF
z>xYU!>KmiHn_MIWi);O|y%|~dzuz%KC*li>1P(140}y-imd&I#63>Mvch??;#)>M-
z|LKEvQNO;jRP>up98)$B({;3t9N^JIho0MW_3ne-DWZTyys`cx$wT*qT!}kPf`)%<
zFDPnVe7^i^>G5}bDSzs(|E19#R68N+(RYh<@nq9ME)z9F5d4N#j3OsD+66rp9AF*`
zyz>8rTj3KD3rz7_O@+$7{xFmHoRr^=NV<EqdWegVdBRwGusG|#12Ab5Kh1JEi}*|-
z4YbDu7kDKss^aLrG^%iwc5&P^Z~S~er7ocUO9S>cq@P3l-!uIoc?50;J)_!@cbkZX
zfAH8Mw5LXc^Bju<lM?+8zu1{se_Afzm;D*&rGaF?rL48N3>En7(OeO$Pqr^i%B4$o
z#Y<i=*Kvs9t@HB88`MsX0^fEU6wKi78$YeSnq_G?oS&6>Ng3^;`mTj#GXU;h@_^Ja
zj6_Z+@%)fJ*hum-ws%hp=ikz=$_xHBrYbn_`wRNBGfkTDUXzf++T#0JcI*!mJeBWo
z@8tcrsC0KSJq`q8o1&Va9t*<r!S#Q?2vZ2Vl!cOcgkwwPb#Atox^liSH1b80S)+Rw
zthUEu=rdyV2zJukeq-_PqM>)-Dk65LSvMa$qm7=u+M%M^72%weL3oV0*B9tQDcDQT
z4T#TEYh{OtJa?V%UqiQdRwkZTvWMPfjh{TFN1hunbxly<RlYf-^0_<i{EAD?9s0{y
zyoGFS+Stz(t9mF0>CbmmDyNTGI%j~h<UhH;Z}p9SH$si0OmOL~L(Hd0{DWZC+my3a
z^>dSEA>w8V&9AJrC9k;u%FdFxLi(MGN4IG^Shyj5l`SXoJ@uxg<#hzH7-?xE-l(KJ
z=2{`4$aBBwhL%CiENStslE`_&7Sym$DUV_w9lU!TU?3*I&bF5hn$VkgmfuxsTYt_@
z7@MmY*&dJ$N!X`K-};qjanM`Inx7V`U}3YVrCC{1_mg|NxI*-Hc8|!DJ=L~VsAaJg
z`J-moV#rFu9oW+Nj!bzK%L5R$!L+R7<SF5fId&Oau7M-qh}ABGT|<pSfCcWXxz8lU
zgZDqjqR{;=h0M%n99--GFKO(*FHk3IOm2byhn{}%Fz&RagFE_pXj8?C%}0&;!?Cgz
zYXFfK6q${xBbjy|?>yv5E~9ejr!)R*=K~fHp7DR6^+J3_Jeoy7@#rpm&SJb^6r54A
zpg71V4|xvcc7g;D(h_Z(wwdtr?G)c}NKYj5h%91O!nV)T%g<(eCmqIL=pA9=pF>M%
z>l>G?aEzDukPU{u`)9Utc+I`v`PmG6BXww%o^>mS=u{qBt?;G`N|jJY*M*65$Ku#{
zfARMnzo-Yu$3Vu<W&8<RvivST@JolTkUzuCrIqM4fpds(`~SI-q#R%&Q{_|iEi38h
z_j0CW>kxe0uQeF!In5}v;G@aH6>w`4k<G<ylIB4BSo-iY^Iq}IrF4*FHz9270Ysk(
zN4@f<>Ah$q^8l^ZNjB_l+`%HIa&*{yDEiVI+#yT@s$P`9P+7r}`-4zJul4Vl8<zc{
zI>oWN2VeQ2^(7FszR%-US1bx5FBvsf<Sd4-<5un}PgIpM(vo%B9ao^Z?yxik@_nVk
zs$0GDz48238{7fj_0Y`}7KCH+SGXc)G5xe<Ogviz!7Fsyh)E8MUC-2FvP>Wty%Si3
zwCu;6e!1w@JqwOfrJ<ZKHJ?MpuZOQ0h0XQHrzbM{u4(QMzo(`S*4cPpftI+fy!a0}
zV6Egn5~};rE{D2rXZWfH7zW{9zVAdK{$xeI{mD#FvUc%GjR|fIdeFwgzX0&d|3tDX
z2`da#MWeT0y7>=Rkw)lDZQx+MSsZrmn09zb{JyT2d${g@_|3br<;xNYe~0)7vFZxV
z50PG)L^oH9NP5iCvTLBAj`9lG@FT4-TDlZjfi@Z6e_!op#6a3ANX~j>`Hq57z}=*t
z%Dx*vM~z!HX7X}Ofjh{Ui2`Nj(dY4Y$&Z_ST0I|6`jSp}6Prrt)zJp8BC>DB5^be#
z65i;Qwj1f|-O=csnr}s;<;dD9>5x<!$Ht>RDeDBq^S!Ks0|4rg?5~f6kdk6dK8M0J
zmvm}+OEnBJ+yK`z2DrsN9W@jY5w5+H+76u&>wFzjwjDYAiPZf8S`c2ZyLRM7OxRp4
zYP9{)tK6KEl7U1``ZU*N4Fy&rSst>_#*Jf>MLjH|iiOU*yi^qZnCt^OFF>Hpi{a0v
z#B-W&MSdjy5)mGEuZ@oIMILf-<}T9#Q|f2S=ue+t*U=?NK%RIMf))<2$gmHxSgME|
zJzV{s2l=%pQIa#Zi}Y8pKk~<>yRc!*hp8zIWhH7;yBu+B8rToV8&y+@1Zt5eT+qZF
z+o}7f307~hoE8xerz9)7ZE-JGn0wneHEC6sdL^lPRBj=sC4b>?IyqU&dRH<-qqP)4
zpv|#)<<!&f^(kuLs(iM={xi!E7s-SZ#36)2Be+;PrhrCzN^SyPYS?yvfoa-!g&him
zdCSOzu=JsIwXK)^{DaX=&$bm;0W9H6z0(x!?AJXaTrqu9;+CT1{wvp&VwZ|O(l9qN
zeq$rwP>P)+x{K?O8s2sW^w<DUaolQ1pUVS;8P@QziIK?s@?8t%Qz5JCL6K%L-^?ah
zb=VE<?#}*DV!d*(DqOP$UK)D$1BEg62IpsgkuO5+(%nA+$jJcDB-UIh%KkB-l)aes
z43DxmDGK=LYqmxRrH*uhggl&xF%T#IqDsIO#oCtBjsbEheIjB+mPt#vM0BZOxxMv%
zYkyItG}ZS`Z=(8-V+37XsT`JJpc9Z$&p^x;YG~C&LKYTM1rFWQO#j=1;PmQf*sW&<
zvg7Nr9Mk>xv#(*&>+`ym?<c<J{G~@0lW8X%s}wEu&Tz=RA9TDxcs5T7I_AAgYA-C2
z+Dh^M3&Y8|yh_ug@EjhAA}sGA>%~FQ8jY&ee!@+7(y<8=LZ$F)&ae^PF2&cS2duw+
zwHnX?X<V=r-F!@H1B)Tj#jIbf(}>}c;8yp<Gp*R$pxU#j4|`PQ)NA=XdD$VG^LC=P
zwzknXUrZ1#_0r~nu1uL-USgx~ey-lMS3R^<-*#^!aZCPRjea`nw(~WlfyU;O9Opn3
zhw|YIFHP0+R-M15=x#cInr<{9JU?K)YGl3J(Y<nw2;Mshs7OnSR>30E4C--(pHkg`
zC?b2EkByBD1pDbnTR!ueXB0U_l`K$_;Szk*6{#+|5|q(WlZH6^4qaPY2+MlDOYkD6
zKkc;ZPhG!4puqi|4r)@Q;y5i@u9B#BvBtZ4;^2V*S<JXq(NsZ6scM_bxzPE5D5umZ
z=ekDtVx>WwEagt4Up&Q!G+A)t?;ApDaM=b=I&6^D2^dA|_hktR6GQGF|Gt5pnGbJ|
z@;`UJlmLC!1&p^p8S|DmE7+~<B<SZ_vTc*2rv9e9w$U~p`ts#V0V2*Dab4MD$^>d7
zJ39Fd1jU-CVDi?}?lHrJM)Qp`hhKLBeXJh~NjZQD#eDkuH1qLo%|@3Yu3q5sgua~J
z`y)_{B&254cKz@B_)~Iw=A<{0l60n8&G7D5xqJ$n4AO3XLnW&-e^E9nw$&c7A17;i
zLqS-bn|_(jW_lBiDK+RunFE2Io}l++C|r6}c}NDJur|jgli6De7M7zgDqWH>ZT^rN
zrIGr#CY;wKf2gr=*{pE9blhWlvbk`%IqXJK<w#vQ+7w&yl}a1*-Jj#FlJcYWdiZ(y
z$7I~98eseckBtqWLqCslVTlp|G|yUVYCw$!Pfb(b-E_+3@x0O}A2eiYd1~aQQh!xD
z0bMi5O^SIpGQ<u4-hB&3*7k#_Q+Y!Z4L1f8od83#;YkvJVk#$rvzn!fKZl2kExPoE
z+G?sSY|kvt=!EvHh3N8%L9<fQ$uLCpfb&;LSd>$^(AeVk+r!Aj|7QVUSJ@Gk2AiI=
zR9MVpx-*6*$?ps0W%S7;XXqFB3`*&mS@lH2Qx~%<g?IK_RZgl%GhbB}oXfTDb^ol;
zS49>wU2rWa<$i^nxi>zCg508V^ppM;y4G|)95&^uuFO3PZ0UEl(P3z!5R=dk*;IO;
zveF5(kEk4fz09r%DfZ?$I_;&jUEf26!PkC<!wy+eD}rL{UgT3>y3oqnQ)3g}X@L8b
z0VBnw<BK?rM)tA2ee=l(+iBt%3F9cBl@MPBLKn8F$LQ~8e<^!NBC+C8x7?2DbLKLp
z6wjzhhiV}Qr&9X}W2u<VqS0(c0$^S(up4jN%87vLiLTMS!R?a8%0WLr*Vm%jkneEy
z@LUaJhDqD-6a~RN*#0*>v`tL8d^k1Z467;Yhw~551jDuZvPvm;5U}JkKuW??i!?`O
ztLkfn)<hR7g10N(SQKz`@tw=nkfMsIk7ffN!$^5Dq_}P0B+D?qC)47&u4^XC;W$9B
zTvPU;dCgP({f#|D<Yp_^?_N~U&l(}&ow2ylXpU6&A!)@saN`Ndl|vWDHmGY#?)Er%
z^{kHvzqjmIx##{)8`kGihHt+Vx04s#;GsrM|KlpQf;d?zbR6K161qoSv+1|1q#r=I
zwn5xii%xry&UhQz2#)mV<9eOX^yxr^%CJcUKkIQ+N;q57)0||G{yMD**4Kgj?HT{Y
zkRbJo4(3!*FWBr<5NstkN=W|7YaV=`uYkDgFDxTVyA_-23GCIe55oIUAP7;4fJM{j
z%WKtlZ-GH0X-M3uuBpeE)?xJPlN_RU=x^}b5;!tv`gQ$!mSJw%lFaw1OK7N?>xBU{
zZ*-9i@vVeAPnN5Q5MLTFhaqt|G#L)Q5`l=gZ+Q~<->2|0PKdNOK5yGcdu>mYSxlfv
zRaw$&h*`|`3&%?!`{t8wUack{Sk_cN|I(X(hlfQY+VT(X1~!K92CvKJCF0TN4Y{4m
z!2=rr5hfw^E}Dc=z}ZE|x4o1ynN=!97h!j9_$eo=<B>}vziJKsn;tGjMA<q0XW%T}
zu_AbG!0GysCV}+Vv98tr+Z0?n9rixADpF$|UOGKj3Y@ps#=g8|i-R9%H?En*uEX4r
zSGRw3;SiFxUU5osnhj@&{BGot4VIApw%(C;SX`K~8X;1^!s8Zc-a^B3g=8N_IVc)a
z-S#vOej(g2+@GtJoas`4n8zFwBm}^|WpO>63~`eNaBe(f>LR;Cx$*!z4Q5C5&7-6O
z4?u#J*w1Vq0XYDX7af#~rWV;swW>hhzk3bb4P_dd3vp!s_fG+y?(64%geE>4r?}#%
z>@6W~xP}zJcuYHAAZYk7T9$BYz2lJhdpR4psI9Z`C;3Xt`h&|zJM5B%@Y|_43-Z1(
zC2~woFeO)0EVb*R`{o95kX{%79lvJ)wE-XfM1JWhLY<wH@l~i7Y(;<hEOGz6D$jAI
zl7{8Tw80$k&s}2E)irIui)9woD_K2!Ut7-fm_PcJ6nS5yAJJT}?&|FKm`ihCRJN**
zOxpUC-jSO6?Zj!v#h#z^0N`R00W7DkzzdKa!7oVIN+m>jx|}QB@sD7(bJ~nqrkAAF
z7cyFZ%wfOTpJvI<tw;2okyssH&3RFvsJPS|C~hP2cqzmU*;u#-w==!+2Dm_0YRyj8
zHC)bpp@l|f_-+|VK)1e~6mix7G^i0y^SQO==LZ3SNE%?2Raa@WU6?jpTJwCS-OBUR
zYU|W{bGFdQh}))5Y1I0!L%0?lT&FGY+@G8^E0uYLqugtd=~6y}EP#gty8hfoq0^~4
z5>kEl-1rFB=L%8J`Fyfws%w_tTKVu*v<Q{hF8>*9$ry7*GA&w)bN3|trc<i%dy!^=
zOBh_Z@7{5D$1y6x8P8nu(l4X-LXXNblTM|fH?~-vwq>M|0gWfi2izPIO4*G9`F4f#
z9$X7kgU;TQ^5Q`t^rCbA!Vi@iZGXX+XTi($ayPZe&zc(Aj^9|G(1%|g3OYyy?IU>2
z(w&o7d@v}H(2v|hvbad`PL4EKyA6=F60uaJ!EY_xV}0)xzViFCm@cW)`kTnSrSV^A
z*6)uQ_$;*Nnid_sza<(H)fc>k;*EW__W<&}lp4ae<(t{noc~6$B=|X#KFUOEaE6Z1
zOA$sZn;k;)rCO&yL#aT6d?AU~K51Q9t$I1-(94J4%+={CM+?x_aR;!WeyG+bK0kT5
z>V5e4(k^1o)igU=O_gl%eYz0;bDb`p4nh;`T8wU#w={W&h~1)tigx;evj?E5N=k+b
zdP(<jcXzkP39ksaSCTx&&3MON;I!&?7>w<Vq(0eT<3z*A>!lBma(LcL{^lsROI+}e
z88$Yy6F6!yxcyi<MNU(At4BwUpR3Pzsc(5f=5jR_BU$gIL(lFm-j=WDC<?`AC+<^9
z>F`u$%BQv1=alu3D1Kl@hmsoN1yG_0@W_ANLdL4--15<u1IaHXp#Jy-?e2lc&WxH+
zfUt)Jlmye{!U;&zBi-j6W9@&IIloh8y^@%o{QgMt&7WJ7<8a#QW$H+|&xKQ3QYPXU
zY3wLM*UR8yM!XJ8Hp@)hP_HY=WfIfnLe<D8-i?r2U|QdU(?`aJwm;B1+5u;7InD3#
zIa8r>ZqiSSl>Xz`>^0<zoasx;?R(3w%;w4PF31nY{HnU2k)pF(`xMbd&Sd-2Y;xH-
zsG}iI+5emeyn|+*O-U*Z{#1RV5y|JuPd$A9vtxA4@_#ThiWuaNmG7SjuCE2dF9618
zD84&t>S%6bRw`QQ&bNA%f@P7nbkxT5h5Y{d>%)tb%H3hDO3Lw7=j17uyfXE8L?oQ@
z{pzRy!Doh{&KuXq3n<^tFLpRHH^ns(`w&SjJaxZmBW}-sSx{(#JoR~%)aBu{E%ftt
z3Z_)Gw~KhHetfO<;iJDw#TS$wUu}RLA!^+$d}+S+E^kHjz<%+*vP_Er%y}7&?@P^%
zPPHkf9kX_MN2Xge=2j9RZ}4k<QwAO%F2yLG_ThHvS2X?v(JMc~C1Q%C6k_$;oVHb9
z0PpQKcvGG;Ehg=)Y_Kgww6jCke{?;Bwqg_XYkxd@fA)=zJpA7A_Pq?>Z>=_!mgHIG
z^RKh1AH0%0K{@&G{}FZF(NzEczhx(ctn94HCS_dV8kyN4ZbDqMvU0Cg_9mOGYh1Fj
z=e<-$g{*5`T(Vtz+>49)z59MYpYQJ+=lGY`dA*;n=kxJ=jF@f**U=vW4i)Rm)}rQK
z^6L1lYPVHGk3qvz*PeJz9l&)&AqUtn8e;E;i?A21e1YZ+t9TtDmCvE3+_Wk8Ixkbs
z(mi|mEpcr;#HCU@IQK|6m2#9H@nn}xr2BG+E%lPMw^Unmb=fy1v8*EwiSR0fIo~JQ
z?&Kf#P64$8o`Fx9+QuL()HmM@d){bx1GpjG*e-Vv=rnV4QooeUOt}OUPP0uGL=XKb
zpdMtZBs>E@`N<R*X)1d=3RfrLpz=cyzxo$SGG4P~Qmlp=fGwwgJ^6CWW4yw`f?TR3
zUqI&-_sh<Ui0x6awPid1amPvvjXvx>)`OLj`3Pfk4@m8vWzQQMMh>V|wN&KW3Cy|a
zw%#^qQjzPaM;>=`@K=f3cxZR`JD8JK^RrfR;AL^_zB-AZtYuAvKH1_SoB{Qz;4if`
zWx(6vCFf=1RqeM2s{l+b{!ZSoZ*nP?vqbkz^y!0dd@3b6mF*mhaJmwPIMjMyBfd%<
zGLF5~ewqC#47W&81nltm)Mv#d1XOgg{&v}s;@}nl?00Jgtd)L@69p>Pq{000w0!f;
zydSH7($pkWP!v%jW^M970bR7~=aKt;y+nr}4$tV{#=d2M%S(>beA2JrR^<{oauT_s
z_ORWcSvlypzh%p72mLE@bKTEU@Oax7mj^G`7r$-x_3Z2IrYARRm(soG)i@fPnwK^F
zIuj(mUF^M=V~E&d<F6NMXc8pc7a7eO_8oZoB0L4EXP8?MTcRJ<TAV?6xSibKNqW89
zsWKi2eeUOE``{yPUpUp`+RYYUiaCMi;Kfrz?dea#mkLGB^y`{;<mV6i*E+#;4xSCF
ze}12<Q{(~W2@N_Ly9}WZcW+T^lsf4g0j|ST6z9TkzX5~;-55hpkLTvm!pC7*({*XM
zsB;Q8jU&;@?dTA3%ZUf<Z-8i^37bP;B{zFiE8f1VN&@x=8Dq&zeLz3h1lD+%!7{n5
zC32L-a{i0=)2oYu)}aixn(WL78hRpqA^oD(YOiPW{fOiAVT!{{b@B^Oqa1&J5J`);
zm<%BJLkRU4aAo^#ry4HTn47G-RzeUw=rpIdyn&PGkf21&2humA89b3%^|3Q~{`=n-
zp0ucMeKvRFSj%bXBk;78(5&e8lm-lH7o=BiNl95L!vl;kF>QXNF~<tBCvHbeMk!9H
zoAyH~QWr;3-cmZ~+hM=Yk~`3nC-LpF`R<;>fR^9sM~cYuAP;pEfBb#bB7XkY-@Q1i
znir@g-^6}4_+Zs~8&Egg|LYi7(ehcB(%(z<7?9Vy5({^z=d&n6Qk-nrnP1LE&zai?
zsguP6O~SLrf;qhpUpl2-^k%s4zd88Ga@04SRSLS{8^Bri6F@BiYv<!dwDV4q5pBAk
z>7)+NLy6ltgBL1={YxU#VJh?Yl;fHC@Xg;VD0E|n%^KX2vK3I)^|lu-AZG9E#|xw9
z0;01@(+7DEE{dG6VjfyVPD|~cC<0D9webbX--0}5sVyg_oJ+d0^R8R;c>whINLb5#
zV@z_YQT6yR95ZIsXQGZu?GoR1bZg|=8r&!2amnIrqx)&8kB+76x#Wg?+)KrB?rh+q
zVTSqZl;n9KX<dF^As+VZ%#5#(icQ3t);}WYi;YW3$94PSzwsK+ga6>mtK)k#!YBi?
zxq;38t8-S_MD@3>TJ_&zE)}QsN$Fl(I`sFY<~H@Ud&6<RW*g_hM5uI$#rT{o$BYvY
zTvIfdRD)c_sfqW)$zI<Qd8?`c;wMpco+5lePf=W;MD+9#oR~E(2F@wzmeJ}Kzf}y9
zv4s>}4g^iWo7N2;EHfp25B55{;Sg$RjB`8-%jDG{F|@pwgLVpWi$g6m#VejZWRE=@
zVSB~@idzo#J!0KIZ`_0Jt|qQ`|7wK)S9-|cm}b4MZUl#eo&lTRT$YV{x@`Wl&E(p<
zHF_hN%5GP$vC6e@Dm|ylnpB#&VyxF!&CJ;U>ESJM!UwHC6=s3}nP2MTvs*o0);y8V
zGGKt<+`G)~g6$t*^%Y6!D;n}v!K>*r+U#J*6eRID3`-QIhD%k+iJ_05w*t<EK;6E(
zomu-hu{o%292XP*=G`^+cJ(WbMQ?^Sejm}G=GcpVtHhpOgjUl#y3a>kXPGS35od-V
z(2?t&E21V#;j(Tb8zQgiJ9pB6^K4T}v!K?J4VTbl_ojX~;eH(nMJ(9@jhP%7>h`WA
zJ>O18t+OPm&b>t-v3OU%@9>Rvf3ja6<*+#UMf*O5y~#H}Z}NOwn^o>UpkMA06-_})
z<~8VgqjfFWv=T!umK>W?aKxT0%S4V7Z>rXKmE<pe=bOs`rnSVsjzzUZ{cF+IK2z$)
z{`Va;B(3TE!+GTs*Fz7SD?@U3g{$8N>6u^XSX~4BTu*mmb^=A;szB|zXMx@*iM#Z3
zBj0Pfoiyz_TQ%>~07+97m4)AT_(&DwxIqs(QR@$8?;B0tuiOjRY9TvTIoSA(9=JEX
zBzvaf_JGuSL#Kd`xon#CT!SrmA~-~dmwMqn1;?4LRd^u5=F6;&v_}wao-FIc;o7ho
zrEvzsO1}rCS;m@e{hZj!){$eCg*NBf0o|*d5sy}<^x;!4{SHzrSF{}80h_d{hx&%}
z%7Oh=u_p-5pJCuk6O9g$(acjfN?%Wj&J2UWTb~eTCHK=sF09K0l%EcLSZhy@;-c~Q
zo!_{SY|s&so_ggA`IkUo6Kh&Gpz9aBtEL(GSBmF+idwzAy-fekM-5wbq1JmopF4H!
zRT<Ce0^lE01d#r7u2VT0ua>^`-2BUgK|7w42<4OGHO-Ey<gkFS0A}|peenrwTaDG9
zBuodobpwT7cly3TN)ToVHRFA#RrpgePoir>ON3x8zN+ShI647p`JMi{zdS-1V7$Cu
zKdOJw()#j>mq`3dfdhRg4}ib#StPg#`W>?f>10U!&^oJi=<D~>O>(}=!BPv**nOCY
z$1&5{F6g3!iE1h1eQD?&2b&oq=#Qe2wNxQ@SaiS-rWqJ}hb-kYbz)`j0!UjhB_&9D
z02jMtm2_N@^2ek{dtHwI$gBvW*Gqx2Bghm?UpRO&8Am?LN4&c`tHLfkqEcQ!0fIgI
z`n3E#OR8@D=M#XnlE92u2Ut;zIDv)>+MjIvmYG`tB&W5$_X;`M@ZDSCK&F|cn2Q|{
zz|*VgWtS#TDZ2`Va0YuJ8C2j&le@Jkmt2YSOU#D*0=kvs>vU(Bm-Su!h-+LNWn24w
z*)C}#ryCvnA7~7BRMqeE5nslLhXQ=`;avEQ?YcEhC%dUq<B(tMomzhYz~dHH!y14G
z!z%AwDX_U`2(kaJDl)yeT4VfC>4(hY5#Db3S_S&bPZL<cayjYCe2>`5vL48{cU5{l
zn;6#`d=Eh|N=&v+T$JTK{NC(vcUav0mU~KV@mTg{G7+_AE5pF6Hz}>%bdhwoo-~l`
zi?$snPVadCzBPJEnDE&KWjtrwc=F!JMLcJJ7QbyV-!=+}wp)SC9JD8vtsaQW({MkP
zXpWx4)cuq%cjkMl>!9Lz<No?rr@z?4b;A*7vgbG0?tQDb^_p+E@GADm=j?dT^p(lW
zc-6)gfSJnob8ZZz72-+uLyZgYfv%-%t_3bzk72cM<~OAbP*H2Q5?jL&G$AIasFB68
zk79x-pp(WVBP#aMc07GhxTm@_jA|aR#ky<MwAH6#&7$!Ks5^5b6Q(-egp0%qaTz5D
zD&byGX+f`}O0-u|+H;8&V5~2AW#y)x9nU4&D)-j}4q?(IhF_+&rp~pwmj|-wh&B>2
z?YBZ?+f%p4)*~yll-NccyYw`UB!jfU#ATP($AcMAQpp>G;z~={gB-riEyB@r&ey@8
z?@~Q>-hVXsW3Hvm-6%7))<c14lXIDMXHXjC$nC#-@f;we#eunB*>n9`U<kQ3#{TR1
zZ5nJS!un?I`0mtSQdB_tlxg%|{L7w)hZa?inK-?;s@>iM81%(U2x}b03cqs=0lexa
z4!c<(k&R)svmT<spT_}!{qzvPEk=NJBYtAT&3pl=uc49Rujsm)^2c#;v!n#XXWHcx
z)d(`@on*Ehn-+SDFv`xKKlLG1Ray~YtiA}=q-X|{<}4VkMn3U3klMsdm?fT^SHx`K
zQ5*P24@&f9i66Dpu>xZHn^2+?A+!`<_lK0Om8UTW@p}I9`r*pCN%S`4Ecf_MSnG>O
zmjU`c)gq*E_@ue<;b@5y%R5LvKUqgRn_<8Mr~DavFbM7^()}diaKheRIrZnq+4Rf{
zS=+~s*Df56<DLA~l<SZUR@$JNQgJH#<EH>T|2%jfFvqu652cqX9=cA=vh~gAmy?q5
z?c;#XXL27V_NFv4u?B#3xE8xM-^V*-cLM*gc}N6%0*ug64v%(TZI;K3V$v<QsiZ_7
zYlIF?12O}{Q4lYyKdD<hkeA53#U?4iAqc3+^@@En%dBq{WM+-x{LVirLUbZ73-yEV
zYRqw(%^Ie1C<pOZH*0pC0xyZUlV6}1E8uCX0SMdsKx`+>IX@2X*A%!baBE$R*jNWb
zb9u=5KgyGJgf5zH5}Az5Jq+5?jkOG`OQzK(`B6in?BX67gQggvWK5>+2!gg8YOx`I
zC)%C*HLAo0GuF5Ie(I97#)gPUcdI1jGw^&vifn*j<`1r`68PigjjWd6u_6o7QUS`v
zb8_k^!y7yU=NL=>8J`VFlbb@~{3!o$*j*A0MGku+Z|$4C=Oh3l;_c`7dWB_;wHc7L
zH0N>yz&Kw+mhC(8ygNnfbAT=h0YEm@9r|)JFV)mXaY>&kCcOOT9w)>dydKVUw`O|U
zJFnE`Jpgca1S}nL0T)mQX43AL7Jv=}e(()guHGGoY-~}ReyVm_jydnWz<iY_<ZgB+
z+z&wTQtdScB>Y~Et!Dv24N2fbuqda+=Ej#5+4>C^Ut?K_{G8JIOL^B)--Ka6pp;Q!
zkf%>`<Bia>`K8$5Bh}AQlu)FW>}?IN1q!-@#Rw*ELZH4NtA&O&NcU*J;+y#u)DvyJ
z1UfiOIl(kUuIcTC8BOiKAz;uKR;BKH`%PUN4+y_oiDs}j-y;schgX+5w&%in@W8LP
z01>m)7x5M*z+P%~S0!#w{-YX79*U`}diz{54&QKPte0^iP*YG3D(tFPxMCl%`$xwp
z9Bcf@Sw{$JSDSP17GQd=ncX~>;*1nrY?0*@#iVQ}!b^kPnHW_Rd2{2w52ltVN5n`-
z#%WmU2I=EE+`HUmY75evgVk=a$Q8vehGW$g6L#KK02yL+6FgUq?;i(bDZv{f6H$OV
zd10Y}iM}9$chtRYOHeG@@VCib;fOEuW{>9&mYW1`Z1nn=yym#A&J-U+Ut!r0y#mnL
z`oHX2`~g;&YQO+~&0mSNUv)JCnEEv-PXYVQG}BO@;z2yHu2T2=3LvUFf5ju`z=Ep?
z7<|ZUH~*5eQbvC@T?J{*bn~-3P4B$PI(H}XH;`o*eG0HrrKI;}I`=~97~XL=H+D;E
zJa7TE?t4g9Egx{2D?jUKdqSL%1Rwe$S=zglK&jt^L=-&~7PJgsEmdf)6B=~NeUuy5
zWixEBTTAVu1vZ?8XYM?rTpHrXDz}+#Dj&WI0J7w0ay6MnoE*Z~;ZJX!q8<YAE-0<C
z*XRA=uG;3C?Y1#6q@N-`0c>&Y*hkYB30?9$<$f8g#ge)cl3r?nl*Wt?>$`2`V5Tye
z+_v98Dq27iWc)|2CY8r5vyunNK;@B9t1@<1o_4T3x71CyzL!523%msF84cpFlM9!2
zM22G^Dc4z{gg2G4V=Cp?Y^Z|^W58H#Jlfr(M7GS#QjN2oHy5GX|KkOHA4{H<$iA$n
z#D!|N@yYZ+C4ZJ06#bqNrjJo<yL`cBmg4lZdjWL+<#4<I$!p2`Q898eSy@@7D3tY6
z#YO42N0)aPLDOs3-c|88Kbr=?+g1~BOqu>Jn@!F0uw&PtljZb!02tb1J=(ho`jZ#*
zyXAPmSY=DH7NQY<Q$)LJH?x-(AO>$Pg3_~1$lH;uU7z7m9mnTiE?m>aN(76vjeYZ^
zBIo24-h+qDNwilVbE=ICkV7*Tf-m$+dO)EJN5c-(o$c(a-u9guHLl@h&|NhUYt2<;
z&f)c=d0IYfLt6Yb;?dr+=lW(sE+B_s8pNn%kA^4&3hkUpw7wlq=4_5Yfd0yQhyt~e
zBjF^ojPs>#!A3*=m>Y&pe~j;DOhBLU>eg6S5kf`rGyy+Fqpup005Bj5tF4&EUv;eS
z?d4<Ssl(AwuhQ@P*h-V4!FO#~k{1*9EFtDpl0QT(M&!AT-%pq;O8<xF6u{Uk=d$fR
zJAU-F?opCiUV(`Q=~B@^4DXC)J$Hg3*ANb?!I7O^*5b!6D<>C|{08n){*C$x1UYvP
z!|3M3gB>b0biBj|iJIX!GGLR*C2mO@J-uodgo#8kofyq1Y(-4(IRczkYB2Sl!nbz`
zwyY($GuMj;4`_RT7mjx+QB9mxuGJ<$iez&t878w9o?u&ZS7Zb97U6zvnFvB}sbOJg
zsYXilRmat7_cpyE=m{Bgign<qi}zhT!_U8dcwRs##|_jV{qspWFfR^xZuXuQt1IcY
z^N?>9UqWx-b!NCa5^`2N88cpA*ahh5Hw2g?&tznqjEp^F0$rfY3@2~yiV+34Cc1BP
z(Ol&2yMZ^t^xD^hUMNQNGVv?jcaMUcqfN(oqu7!HysbtxtQEW9w{xyD!0hRZj7b3u
z-7Lj>t)0E<VmM$%2fyC4QC(|lXkrsRk-{is7T`p)9<T;~9Zvf;gm-7J+$b0(alLZJ
za}~agn}sHDWiNq^_i3B&JLo5>MP2zSp-b<z{QKbQR|(n9tJT)uSXv^^*CQH;BAsDJ
z4hMfUgKS9dR~k#+xX%&kN!z<;<ZtQk9;^<sh(1gl56j}g*f7(~DF3=jt0to0{(R{#
zJyq1~T_jWf4eIR^j(e`cuHn~j?rt6M5vcWh{zjFTqSRsIU&3hAIpxW3@}<9n6qTBn
zl~*I(1a*O+#Q<7a{x7p5w-xUc-A_j-a7}D1guP5~DS2qF-frnq6%)|%A%i*9Jnard
zZgk<1pH0kq*<N_5K~(M`n>8k;x3$!O4SKv@F(Ivr;aBZ=Wyjbc)AU$Jr|ymNu)1m*
zvwiIk+7Uo0i<`E|SfffllY#EtcSCi*kg=V4f*+rB8pkg3v0r{J^-jEySgR9e><pTi
zl?ILm_z}wwI@WX-Utv75pWICNSX>wrklXRg1$#Ph`VsqNXNcEsosN;NV0&->$S?dm
zr4h*?&tIJ3HDE;-X%YzDIjAWqOodIzK34S<XAX9eNTaID^uNMfxug6ncx^7;rs3s7
z_1RR*i2HyteAI7jr#5Da>(s6ZP+x~A)~ch(_>k6xZ$l!6j;R=-*=|C^X;ulyo*r`R
zSdL3dRWk`}!^h1@J|5J*dr$kB2wApTj!}UFx}y`%Yc@Y$8n{l%;tQ08K6RX7G4$h|
zOnyF^q{1_OK=n3H5<$%P89B}m6Gsp@vY~bd8qN&R<AIWv7_Re)Afb`)NR~einA7lY
z3DEkf5`L$)%0`5bOxX6nTmUYm)|jpLUu>9<Moj1sHa`9@TU*fL|2;8XbBZr6#?^L@
zBXy{mFcu9yJiP>VEjS&ko_Lkb&9{{z@j+E*De^dj;Vt0c0=<_jbJY40|E5<Q<`l*w
z^y!cWktAT2Me*4F`Q9_;)w^JUu>MtbuGy+j@wMEa++3w({;L3M!*4B}^hw%q07`i%
z^x@?tYXC&a?d5ke+#%k#t9LLS*!}iM@jVWqbR%JO^3@{rtTEXMIKh#5-t3Hpr{kZ?
z7px6hnh@KsGUYwMy8~K^-?jW3C3AtjW6`#;4+-U5;>f<7<d7u<hDqri8(;?eBK>c>
zf$^gYX!l&r6m9$hg!QbU!RVkTPIX4E$+c)a&CGdcZD<;o#84xLAl@3d*u8g4XDL~u
zP~%!5m0eT*_G~NF=lIpST4vW0zw7#6fnsF~{6zD3*h||yq|*>iWdvAGeKc#oy=T{=
zm=Ertm5M(#dvl7q%|0AGsP8%W$|AY}rm4xL+G$m+?^4N<C(V9&gb~vRbeHm^slh-&
zzO<7kb5r_FVp)&q0#6t!bsC53KU+Pv?r$-?e1#FrVfjak`(o7WhkiQ!14Itb%=zD8
zt6k@v?YwfXUr)o+{EU<UxQyJK9THm1n{$5B+6u$M7JOs~^Iz8{84U_&i8B#Bt~F!?
z{H!LW*ZB=(Xp(lWR##YL??H23q=2xkF26fffTg04@nCnohR*%-K-H^PBZ2SH@AbnA
z8-^p}i?2iy7gzFwBYT7hYw+vaM;f<mIzERfLlx6+RGmS;W^yz51Eme0eNSe%r$jI*
zer{$)HHR~2g(sMv_;Io0RWkIXS=sgYG1jk*s(yV}uM}PNQ-&kui%r%_8H0aQq|y}P
zP=8SJzc=wJgS9r8Mv0&X5Yr#OfT+++j4-6KY{T1q=mN$UCLiTHJ4Fht{pLjLonE_X
zye}vu4Q+k7?TP3(7O*?&zq9)pFw?qiv;;6Ly?O2F*Fp2am%(|1s(i%bE4Zem0MdY?
zIKZ?<ep7Aso(@?LR>AcItX<jdmmX}^P^IZ97UVvC1#E?)i2-5n&cj*Wl2FtNWF3j?
z_Dfusy!tA8Mz6K$7DP=Ol(^5fWsB^p=0DGR)Bn19B6bbsZ&R|AZAfaK`l3HR)tl1j
zlhzi{ultqd$aEDxrZT5BYaS-Humtl0Ie^!^RmAxYR$SKaviP6ZNV-WmP@cX>0$eDS
zoVRSVgyM+(GH|udWZPCpP&ETBn_^;594-Y$+$^~I@QGX|th!`rdUUV2vx$TJdJOzO
z7m~GylXAQOj9Hg8<b|F*=^O=nX;*T(KE0Rs048<riu3&YdO{6^)))N5hsLaS?pQ7{
ztM%OH3bH1F2A-Tyd6$fOoqgNdZ`}(4w9INVCTB&gf8fiGUVUFORVPCla@=H)P=PJE
zm$6fWGgODud6IVFQ|~KTKY(S7e)HmaF>+J(j1V0qz_7jR8AYlUhA{~xj*^G`c^l&N
z!yP{Svu#U1fOIHL)^Hm?!0x9gSU%BiAy+*<qoA11kGC5SIDPo`gXZ|Cj;9iz&LOg&
z=&e!N&Wm?|x204R1H;y)qc}u|##52f4}&S3c)&;4a!cbt--RTu_(Az)=VbaA9e4tu
zcOqd|LxtVXvtFMSHK(ISmIL5@3jaNyaFKV&11k7!iPv=3WU&ilOp`~A_Ck7r1To98
zVIxA@Jw%P%x&48N*XAcG46C^PjE1ANBareltSW65sC^{rmyp&ZOO@okvc|do<JHON
za12>qyHMI-;RZ{)pf-((&$`@)?Od<>v=uwsEr*{&C}Q4U=xC;@1aOiMfR2wV3sA)C
zTL<tyym^gN-lX32AY={@wM}e#6Czm+GU`L<3E3X9KVLh`Ay)S)NG}JN2wD+Bda_?*
ztn`mxxU@Ad2Ub6*uQvvxZeB@5;T_^oX=5v5Y1RgL<NSn`fWojGtqudV`Q&w&N9*Dp
z27mC}wB>YZaf5ooxZK7ttofIaf<mPeN~S+qHSZAPKhzZDfqlJ1iJvNj>lfLSV9S$W
ze#uEN+2`V`gZLZyltBx?;^^;NdUf&Vp}*U1X=fH(>gK_q0AjzjG4)~*>muU!vMj{p
z9PO&rIO^h(HuN6A_#>3BFc}dI+qQ39{PXOqx!35b$I#%al7V<Y-6SBn8x3rK)$pVz
z6c<v4EB=joVriUnh{ubswUSj2BhsF}ioY1X<60|A>%R7@mHbdR;u#45kj;0^-l#tY
z>^}7AK*#w`uaFu}=NI;xfrNY5=~hz=HrGGFv=LB6NG1l4hxYU@3?>$Qb((`l480OM
zPZTsbNqdQA`NBxplSGN(DtrxCRb@*idV1qpNv)w3Yzf)b0@o$W)1Hoxa3Ew2h4RuD
z<lKM8yuI*Bxk{|>V|9A3CABwZ;41+H9Uno0&;3WVx+l1i;8v8Zo{?_IiGo71Mob&+
z^F;N-YB{4l<L@cZ3?b4u(m^i(D}mw1k0FKP=AO&Mu}WpIbFkO;?`<|55>5PPdcF3m
zZ*S-Do!AhLx)yT(J}Io>yzidKXGa>H#H$x1!FO5s<SWOoi4lX;G&#1;NDwebqb0(?
z`ZA}?8y%7vfCb+f&YzItA81A^h`L!B+`--(LFwQM0j95>int+nj?xPu(r|`#X;8p#
z(urztMyuzZYwO;DT<AA*<+pg40b7HTbsZrvSLY1iuR8(KgKv`Ko?mo?<8D59?~Wpv
zWdZEX;V(9|K**LnkN^Qa*<VIC_DYk@IT=#S{|Ew{LLb}N6*_%C0_b}q!2VI4O;iQG
zEtlGe^Ld_@-ye<X!}zJK*+uc)_Rw`?o3u!R4|sK%!^Z(mn3ChzR4<d`d4{OA2bCo>
zG-9Vfx6Poq&Sj18n*besXs~4}{^vz3=m^Fhc&Fiq%vstCXAgDuz&0C{c(=c$ho@u9
zFGk=!3u$DPjD>maq?vMy@02t6!9#7L{l;_$xhjwD>~bno{M*|sy7Q4#cuzMqXZV#*
zJAmfKL4MnfW)4x#($4-?l$uL~uC;puY;cB)BDc(Ql_r;d-5s*BJC8|H7y=%Zd>b;>
z&E>`Av6%G=y5AzTG*2gJdLzSg4;S2pj3EHf8FNDv0UUs7#z{+(0ng{-t7XRkN!siC
zDLIeCUx~_Jwc$<35kKB^=Ms4RuV{D}m|MU*UXc)RRk|4Dk*D&>)8kvE3k<{Fv+ZIp
zGM6$--oV(=V(7ROf3?I7?<1p9E4R)5&llTDCn~(7-(IqV#W7t*HNOD$D8zkh&(hD6
zirp$IoyggyEvP>HsA&QrR<JDCl(lK*!R+;AvQZmWU7oBLh6@jCQ>`u6EetRyM>(wP
zq#&kGtHWHqg=v@Qgi?C$QQhVnJ*iN<Sl9edo3kJ6<fu~iBQQD|qS`3KKwDm1C@!9B
zz|o*Id(!PnDyj299QjxT3qNo#*9*FM<b(=Xfv2H$Xlw50Ja{2RTeb5G>b2TU0}*E`
zJO22|*J|Wr{EOE8(e$+vw~`jISHKeRud+=(`cyTP!I=mU5lY^L1$cf8No#U1Vwi<R
zeZDTlgYP%-OV|=As<P3DN9ty(0_{&8eTwS0#-`CP0)1--0)1tLb;gvglsa|cHpF7_
z$CzwGKXnXH&<2=LQ?f+Kx*KyMMtPi{BO-S;b=S%LED#{K#_D84k_q->L;8`!ZPc1a
zbtHU-2XAF%S;|Hrhz?-JM?SXwLXKLvZArC$VL56etZCTYW8(t{n3aT7R^iGTaofQ=
zf=j3?{-9)~8^r%!H0#{u@N43f$_rdQ!wK}mCk<~x+l27zy>>nYcR-@Ylj><_?>p5n
z3ZMedUY5GGVP?-YbGxIG!U#9xPo0pV_AD-od*=2uifymoK4?z#aHsrKT&Y)*+`$N>
zv$9kX6F{U=`ZuI^IY9cx$#;W5KJYoPsW2G|tk`lbp1y(B<Gxo}<kGWcU<a4?jR+M6
zg1tUyWos^F_-kyKCYrtkOlbZpiG)LM`;&vL&WZsOP3|USVEF=wOI!;RbLjc?P^?O0
zdBDNtMW{_e|Em12@{OL)2_ocvUmCa?fS^m{p;Nn8y(SYR0Xu#5FcfMt;tEZd<U5io
zwz+!4%j@wS-Gk6i69Wg>>1hCPQDfOpk>op1lMW0x(!lo96Nr(m$ricE%}tuCW1u7V
zFS$&H9)IqBO8n=p-F(wE1UO#^iE`_PQdm`Tdb3Tr==lt#Ol+4_%V+U_GW|`ZgHe!d
zT)iM4RA;@uTd=5`l#u2>prO#VVuFl39CRUN{mee_U?|(svf`o-t8z{c<X_u<6fqpE
zSap(a?UPC08RnnMi*Cd^-}#$jQwj)=$lQ9<D>>GP>K>z)A%Vi}6u;YNy=vTU<69Y2
zG-l5M>BSEA09Ov&y-hm#HJ(-=QmagJF!%6;1u}x9NjK7zDnkx(T$**YMMlHA&v8u0
zWPrg1Wqr+4y}l@|X_qZoy{HA3EUh3Gg!Q^oHT6-7ZrR`y2LH_Gc~bpbMrE-QKa!!x
zB89%F*v+qwD);B*njU-hX-GDQ-18NVp`U}^9H}2J&mA$Dcw%4|eytd%<T<)CM%j*P
zv|$EP3w>>p;7wZSL$6)D$IkJ|v;Ph!t?bEXYaSg7T>XLwPM58G?YDFxEiuYfAeW0C
zY>%H<qacZC@vX<hv!_F!dT=l=%%@(s*VNNJ7O~UrNSIM;>umGDTd$d2w*#PSYn&bd
z3^MQ!X!*{w-CIRw0sU+}U|hgb&&gf!Cpz%-WaAKv7Z*<D)pn939Tcs5b><HYKruT>
z?LlLv=EJmdGdx$^B`dcC^|EfQ7<A*;lEZXLQHA_{yJ9;1eHb{uDtEe}_pir0ckzo^
z@14cd8Tw>@CReYMd)DUW>VACyO^H}Wj-;vNR+L918gPhidVS87;y-$E+Nz87uP9lu
z3fOe;f+XP7x|(>5=CF6Wlq_*7n_u-HU4%Fc{QmNo?D+6nK}ebJAlCx~4wsGNs%g^P
z*b3TOKeLe~p74=~gJl2o&>3O_3SgP3!Hc_d-?G<mmP|+Q7sK`PmBn(^b+GrC?x}@|
zIM(Czt!_U8WLi<#97dBY{CELrG%1D1aX&m<=J4<XUDE63RR!87Do?fmw~cy$us00E
z0|`2B(U1+3(mQG_1q7v`FgL<=EPS3-dy=SI#S!#3|DYAnHK^a3!Zv+zu0U?q(wxt4
zdaf^uqY-J1_X;Q4H;p}_;C*L5&6FXC(g|do1=~gaBv^x-k~X<f4uZ>uh={lxV9Bs)
zbc|#+_aCD-C|O7HoF$9YnR5Q1JIwM+5%9vFZ&+n{zoL^nYrKbF+myJvfB2j5MBn-{
z_?!nnTJD&3fU+){-oXE*Wj@DM_MllA2EY8bF)8Wute(64Sa%8CVK_PW*hSMyV<*p2
zGVkZ{S{}nBpYLrEJWaWi_;cB;)5DI**L3ophx#VSmZmbbr5l&H%64zROj8DO=fPOe
zm8S*Io-KrvRgVVf%pu23q#QeJ=FVl<r}PXqnl!i7Gc{fA{1vLnc&_xnQ&-x~J1^uO
z9cL68?HRvk9o*|R0Y67q1T;~dJ!uy<tkyjLZi?w6^ku;}o1F@fNn_}1s-R*Tuc<2i
z%IYQ>Khq^ERb@Mkov_JXe>T~7c_r71eto?^`0`8d-qxBE@@KrhlD2q!K-L0*3hVWa
z2S3qfu7a@fMw@f)(=Zd<0Vl}Fz=hI(+TdiJHe36Sp1~Rs!Pg1QiYnl_@-aDEQex%W
zAf#QbPR^hCZRPiF1M0v;1TL>aT-4HI6+38CY^>>K;Qi{z%%6VmuaAV2i=180Rs47*
zhu(+#hRy2+Ir3xapETpnD`Rj7q?!Mw@FoM7hEB(#rmhzg(Vns%aCS9uO<n}uQHY10
z(y?ke{y9{|54VXo)HiV~NWY*mDJ%w@rNjsJ7fM(YSb%S+$N$A3%mLfXUSNLp(7s%r
zcO)5Vj6v%4!hsnfb`k(YEPY!IGJE`_oINsD3(z>H2qD_TrFuTvj$#W+7<h9U`xpYS
zB?eea{Jd$Y*k#`c6LzK*bezfoM@a-u#AwIFU}vli1ffRHDki8a{xc+WG6sezF|)N$
z!0CtYG#ra*y}IYJC6xu*u)U363k5$KCf#NxTt8e5a%eJ!KC!EQSK@J_YJCe+=V<Wj
z{!RqlM0HUeC`$!WUoki<E;1A4WYUC2&?$E{WW`rgp8--Cs-{y}m)ef(7xgy;F7Msn
zYF$Np%^!NixD9BB$RXBhD>+~2P_p>zV5~$4Fd1cD>wVh7c7FuOo%L7u;y9@&A7GE|
zDuU&~{u=jZI2ev%#TE!8nANHL6w<~G?nvo#&{A*Y=~3Sdq+?6`QhOA#2JxtGaglc3
z7?$O+tN*GabIa*4@iV5$cd)`ISRY|iS!r&(vz0_r)!U!BdhlTz-m+3R|Ib^-M7)JC
zKj%t(-mP>T87YXtD&$AkaSxubC^bH}>24~!VZ3|3&Z1YeSa^<$o7N>r+Zo<`uwe*;
zZ%H|9kvH!A$R}|fHgR_7EyAvPF&*s#?qUhU0_jgxHzukF8D3?_vOdz}g7eXGKlpJy
zRdPL;+x+RCY_!nA3mGMM*4ggV*3%EGuG>daS&+!mggXpQyff0;_Bi^J{Z}KAoyhl(
zbozDD&%A!PdOJqLzY5HGjmRO6MEE(m|ESFKOVJf7Qo^QsxxRohb)y$Qj&(crCw`X6
zhveP$RF*UJU^#Lop^`UH>A*w2en16Nh8_ckTK^fcTV*`wnQi(n3JZ!tUJEhI+PG+1
zqduYS4`$aP5b^vHdrgPN?p3DL>jbpfvZ47-!_rUPEG|o8(#y3D=NS+>?XfUT_QXi?
zot5k_AQuI<FI?4?1`Wf?X8x*HZlKPLTUHq4XsgQ^U`{V`Ww|=&+{N<m+9p=z$+m)<
z{8r7}MZS@$7-Q>Og8lXIvj$6?h2e>-q4??G#-%y!@8;?#2Q<r(0AAt!){9$8BAh#8
zCjP8A0fjJ{K>5nU8Urhl6A{};&G2vLWPv@g<9#G}O}p27c`P<8gGH7oI=}}@$bt|1
z@e~ZN>4lHv2L|ky%(3v{w`2+3m{AIc(%%2CmyR_Q`6=H5xKjiflbXW(Qu(^sJ{(k^
zRh%(XraHL^IaYa52ciYI75p#F-VgjZ1yQD~`m-qua~NPb{7g!1acQD~w>v2|90lvG
zCropx5-ev-XLNo6u2x`RBZx3ihUmtjP7MPJH=C*sOOP2{1#;b}=k7?2H9L}wGj$-J
zJ4>rs9;unWu?4{ua<TPMu7Ezl$=}HLqul70=f=kHdoc%%O=&avK&-QikmUlK>=6LR
zXCb&mH$oW$vvl7k$ea;<ihecWf&W_+GmKN&;@o1==>qGtshpo$XaoT{@V8gC&H>&m
ztS!GO=0tU5XRGYE>z@M%|NerXfmR5Cuxv`b-U9KMJor;;Hb?)|_0LPpwZr3RKN*ij
zxn&FFpvXH$!tkTdgGcze^uc0iNK<~9cHYVJxBS07SlacE#QgHGqu<XExiaT|i^aYZ
zmH8|k7P`h1?kl(P%c5KckF5s-F`kd%X%Exg57V^b93Cw6bx^Gb=M9Y>G9H(YPwChx
z0pxSfj-zOd6#_lVM0jMuQH_~Err>~D2C8gCOIuDq@;$HDT{N&TgeQg14U1uC8vd3-
zb@0aYw513qKfAO)B`w=z$Glbm$PnVB7l0N<!SZ6$VwQODA_<9(wuApcNxrg@TGw(n
zi^1EUZ-5V}s-P}=If<9HwqzkA3_G9~jnmYZtIChk%O&#B3bPHfURfg^p?Rs4JZo~d
zrc0%BN_x%vk=e>Y$?|c5XM#WAZd(xoo3h2{vu`iUA^RF<lF7$>J(-X0blB*5pyeX4
z^71LTy;?`q*vij}^E5;FS5f6YkvWM$9jzpBV*)Kjfai$@3H=n^RBC{AUK2I+7@y!k
z-~cf4ow8jCu8-*LoUOpR^Z8h;_$JLP)!cD5ucTJ6C5#t8ZCO@XoQfS<5t|jU#L}`2
z&9QzH*2PT6?VIod#l)Npm0l`DAvuWn2eS})FT>!b4yK8=ept4^rKk>$m{?FK)p#zR
z`{&ZR_f*8&>O#Q{{ADJhO$^(UxtvD)gJ2?e{evy(XII^(>Vj~x0W*iR001cxFmfh~
z`sAEcx;$v18UJabf3+dDJ$pD$&#~*kn3Rw|7pD+R!pOdef<@x!R>kuZSON<uH3nj5
zVlXd@ZERYXgchh!IGgbywbHEkW58-z#Kx{0$ev%0a%$JZY=|6%Gosf6^O&B`N_FT&
zrM<B*gn4#IBhWSw4Gl_vB{B%H&LHN7&r6i1sL}|aZT7Rd(}0wR9tWfAzjywQKVL2m
zP7qOA2il{stfiXLe+vEWzd$Qkq?}fDW=9V!K8|k_?>-NYzlU{Mfk4$ydpgr*Qvt8f
z+9g#MvSgGSjJCv@bPX955Iz*ygTqK08f425i?}o({*k1K@w*IP0*S55+kJ9y%yfxF
zC~E4fP05O5*Ja*)BpthsdBSurQ}agly%CO`K=6i~Yky_P;PnquZZ?e#bx#lMG?qZN
z(LQ5J$_2=!^7V5`EBTHWnd@a5e(p&yAwxP}n1%R`Al{TwDal8C123&Xqj;0At{Rma
zPiHtM7<-UKkUx#T$Bu21qvf2;!#@S1l$LR2D9#wze*x4I>bEY(I>oL&^b1_=>R^S5
z1tG0C+@GWDFt4*ygVc6i{m*n7m^M8+m1&`2E$^CU`b>J#8+X%F4Q$4y7X);rJ-JlK
zx+{Aa#=GfluB1n-?{!(fYTe$w?S&btT(9faSx5QhQA)Z{6Ds6kO0V-I$bvpBP~9<w
zP@jf+^%y+4c9JFWPMYsH{AFt8fjxT%Jwdx3&=2X$5{xj&_hl7WKeieJwgnKhc{=-r
z-*=O=Z!F$~vikK8)zN<CK>^ypYwqI0EV|9JNpXl@VJj0i!?aPsg;aK4q#>i>Hwgcr
zMU?G&;I@?3$^5SKzpNQb1vaXp&WQO39@`!3HPtID4HgZ;_!u`g^L^OC(bZi7;DJ9K
ziW2PLU=uJ^=7;EPyqnYNV@VhVd$YyGZ{O6<uD|8Q0_W{C{aj|@eW=}~WE-;(6NxIJ
z3FLiR%o@jWZD)jF)7Vk>V4i{(-|h0E8m5CmrmolJGBN}eZsiyv^BBVY?5X|jFKQJk
zVkVBZa1Yf^j~EcMz%)$dHzqvjp0RpsbL$=Gs-Z_v|Np+r;2rwm<YKPU{4S-c;8iyA
z3$jtMxN0|#f-(=~hdR?)XPi#l)#M_07g4e~Kg*7<g2V}_DMYSt9|MG$NOFOyVfqgJ
zl|(ztr8FWMu9oOY%`zpoKk)(2($A*798+)KbmUI3N`fm{RwEb6#g1RWwL5f2Sy20x
zlxXL0^NLc<t#<zpl23OWTol<-4ZfTChI0h=$o!P<CySey2MX=DAUL49N1lDAiDfL}
z7XDy%KMGcDL$LS!YM$rW`cF}n;Ofi|Y_}~nzpfGme|ei>*1CKrD=SyobHGR*kLnFG
zegkT5u+9)U7KnxoM0Yvr<t;{5vkm*9zHxoAnK4Kf+0^vDH^f^RQie2h%v$I+RpBz2
z3`Kc;o+v!{qAOmtt^pg7RenLJk)xo*2nQ&(h2`cFx%=?yc~UN`#t*iS9(`TLi5gXB
zEfrsF@@F_VuxL0GDU=Sb7?o2;N6v2;!ix^R-lrQTxf3MWC(9{hF{>;D{RQ-OM6RnL
z?0;7(_>4!Gle4JVfVJ>iHd#YwX+g<qp&V(a<IS`;O-mJMrj_*GXxD*+!bgkm*gsg{
zaG=n@qpT-mV5eT;KF(WuaIb4a1Kal20g7pMDd~w@_<qaNO=yMJ`4cnRz%N?AnrbiJ
z-4mC-yn9nCTDVp0XK`tSmD0#BV2WAzYn^cIct6mLE=Xi0?i=cTbV8$_ljj7z!Dwgh
zY#B}kqf^7xSn7+n_+L#g<B5rKy362AUPGl44S&hUa4!3eOoPg=g2xh@oa5ECskx3S
zNwu(gB8tFGd~`0ZvH|wIxbo^dWydcNXeQ!{UeLAC9wYFmCHYXZiMUm3`$#p;2@=5s
zA4XqGf`Qk<UZy69{^I^SpqSWjQa;`|khPPzm{GPLooFC1`htmj{qTj30)GSgrJ+Xy
zLFdl-Y4IdLC)Q@-4&jpc;S%Xs8kC^%vAXZxk5L)XJK8(9@WpR)l|DkWQ_%Fb1|UFP
zX!q~ZCs8_btt$VT+0X(=AMELZBHEO!iesC@ue~`{an|3sEE?$G_)_Hfsd=6UHM|3M
zdYJZU0{q9@baEs9*j|2N(ZeolEwY4H=3VO@glt<}PV5NHo>-TBS2+pR*oLBScyVi*
zzBa3(JRx&5<AtAPKhru;jHK2p(1g~+3H3HpNS+M1J;<<nVj`LCvw`E`H`Hys@%pW9
z5%u%dF~g9`5}<;8JVgOo&_dE({D0f%4E6;Ye&;y`j2M~d_U9V#;maYKxa-ca--Ugv
zLx45HV4e4ecM$BzC9Tit-+X-V?X?X#B^j&V58+=KAQ{_i3k^ZHRWxPjN`7aO3+T<@
z@@x_?<*Fb7eX#PsT!0t4$&AG;3-_9OmGFlQ%jttE7_}k0z+Z}aKdK7T5unsZA|-wj
zpGUjfu_M<EeRY;rqhZkZ>DRx^T4Vay;lE-HpxRDs_)&T!xquO%9u+d&Ig}GN%rlrw
z*85iX@9X%ZD-(Q(<S}OsZ`RD+Jq6dh4bj9?JHtTWL9&B9q^4QF_oVy{La}J2iD8~n
z>IyIy_7@f5T(qgtzdz)@1|bS_uX(U(D?OL0ttk4*4{Y*h2INngw{7yYN)dBXy~#=b
zyjVsvJ+#1nL&a*MDNlD3drs5LF0G{a$d0#ULaVPSMlIi0(6|a=b69Pzdr+$$pi$N!
zcO_6z&%5R-!7b9UcupY-##S&Wgg?iO{qGO|7&{o{w2|+;B>?!??{4vs4%OLYl{y;g
zLeS^ic89j3x9wuSG7>(#%L}50*W+S;POo~?6D>8MAxKKNBJLwF-n($<5-b4u<Me2^
z0~WPu$Ze1zOL!4FC%8Kx<8MmI5kf8r3|tG!Q1EDSqY1n`&BntcUH3Kvyzr<!Gd#~|
z8s~tVi*ay_BI%|@Hudk@V?gTOuYN>hT%zF}5W}6Os$BJdpb6t*|GY>pYS`%4>qpJ(
zxq4uQ(3;)SeCjiSfsM8KDZ1gTpTrNBf1P!THwfW9k#uI5WW7npzVl5G@kJ4i{+O)+
z2g)hf!or4rZ-u1zc$sAdJbm<eB4^ibp&@kPy*OexYpf5%3!el~-+q<g(B*gHql@0W
z)g@Tu;WvLr<XAb^C6u%XR_4byl-6`Qlqgb+${8udRhK68j$o>5#B?yzWUc9?=|^`&
z@ZSYzEHMA@bR6dS`LdY65V(RZP~IgN&ZR5DG{a)wmz?cS#H@uT<T8AT1}0@O(au&<
z#hSmM6~@J$ZOO3llex?0D~c7r<_=F6acrSoKGYQ{Ke@4B9oM{}s>B<nFtWc$Wxbod
zev+j%SS@`eNLOrheEZtMh*W#shwbF>CvpS(?|-;N2OE6<O%(Ap%+tP8;&%cL5LAC9
zs&>xKNEQV{V*kH<v#h@4g9*l}(U&w+h%*e^*jD9_d|N}wOrCySu<)AwB3DrvonKo0
zecb2TsFf@+t6?zu)8Gw4_^0&e%9V5IF>FMM7g0L2jUm<i!#1NvJ<ioph;O@<fy-#0
zcO)z_U6%N&aN8z2NVnReFIn}T0PU_l1g1XOr`66@W?*^L?GnD0e!iOg-K4Qp4pbfw
z5HM&_&ME2FlSo$y9=eZFzeGC!efDkr52P5jILB>itY&t!z#bfeaM_rwT)JWJ@ehvc
zQyw=`KDGU)*7df+2;S<G*{80QaO)+G>XOl(^VnS0<VV>Irzp$Q;iaEw;PzsO-t@~+
ztzfgbcA3FZGXH*wN$D&8U!<^%g8~cTpC$@PtdtV)9u9;}(I~RQ@}!6lXZ*v{Sv2rA
z0GTJ8Ssg9AoB%MMi)A#au-KZ<DzYsK24_cGW6Hz*!YKd$gySxZ)i5%k7{|9mfr@ty
z_EMd^Qzqkl7-*--!Fl@m&+w_<od6HhE#!QJiFq}SB@ZfKX%PL0D>tc(RPrf(#jo{K
z@A<hUY1HRC$c#ErG0ou}i@EXMz(mTp&kvqpP={f9f?piQe^w7tJsr`RGE6Gs?Hah@
z6Mo+yE8MT_sb~AmBmjUWIH8Ro)va@VX+9G4m=QiBNOWnF=Gy=6t93>P!*G=CcHo1q
zb1gnnMr3OQ3Zac??EUKIw?Qs{HmeC$hkL46b$b)k>l-~kn9KF`S*G!>NV5$@uOVsw
zM8|GI^KJw&)Z}4D@&rprS%d=%!_y~Eqi`F~QC0eqBhuJuf$QE(qguXwa%sHkpwIx-
zqQ;FL*he0NOz4r>fWH|Qg^%~p`OLDQTzr&2xNVxZ6)CQcx*2Pv9zc#jgH+-OKM|mk
z!#s1c03c^gZ2NzIyX-?6hB$sivF*0@gD$1!50O=Od!uhhp?(MA!cDy}rwPW!KTrV$
z9aqkNP@ZZ{j5+F6I1p;(AadIHfeHG8!8MWHEQT83uAVj#k&nKkOTJq_m0!q~-G|-6
zd3`Z>5`UcL63EwQZ!kVpj@q!(uIGBL2)+MHt;p9$a9|anIlQPuT^|NMyc9nZW5tbE
zSc&g-;DDZ&9V`emuS%SESPZP85%&ngVr2fmMr8fcQx&jz7%F@D`cWKMCLyvX#m*rW
zT-njPH8utrdgS?AC1*Xwt-QGkuoovQp6`RRJ5a&(ewPC%-e^&3@0a`HZp6B|wqj*H
zeS&FM886<>><<b2bjO-*4}z?3(z$i_GsN-!cgNzLzPA_je1|5&nwV1~hV(y%`)Q#D
zi@XZ}u+!sMp(u9!5bMvD+0=ia%UeB;$*ww1EA`OodL8;^W`<j*v+wk{oI3E|cJJ!q
zH$Fag1F)MdU~vM}wJQwgkQOtuM<#FSYza%zCx@d3xkB9wmGhMwts=)Du+kKKisO&A
zqd#~}R)8!|%8dIoh3h-6vO#<+Ytuu1u2)a<G-KU4@3GOo!yHMzXpZCOG5$bxlSVRc
zD>UKArZQrJcK>||53tH)nyaXj>uLk>JiEV+l%6U(3Llbl*`T!p2^l8*A(z1z+T{o=
z$kx$yFqZ-MdJAEr*9N?NIr#Peza~GMqSne@Z$5tjK$mh1oKJo-n^DUualgZh>)?vZ
zZ=h{<$6s0pY#+uWa$kp`m`>Tgl0lLMVH)+I<VNi5uix=*Kyvuc&lZM0eN5t2F}`gy
zAfp{dXFO`am2P2z{^WV*g{>`DgpJ11mMZRJ@kBTJ&46xiuc#enisQ&QtyFGfL()kF
zx(BmdH_!V~6nhBwFx)XS1}4|Bj(-~@R*l`K4+tJD-RA(-l9W^e8J(mh^0Wsw?dr9<
zL4W_G`0#FL2&>=XIOnJbuwmR!*hCJP4Gn680(zX_>H4G1@myY7r}V}d8-xRGnzjr>
zJMr#HnSt^c$A%_jm))^(KZ^!JIIBVQZ-R(`rN@5=?eOxSHo#YDy~CWWBxIPUzfL%S
zal>_nfc9Zv9EoN!+!5TI{jihsVfX*fl6+*hNdRF4D>~&fTv=@Dl(>GZmpNhO!z079
z%lqt>^0cctHW8oSA@=9nE6<P8mXMWJ*r|&Kylqn$&5ZKBPPDY-_Gg=#(i+85IH=Ef
zGjP=P@96%jT7N$K^1DHX=oaSpmc|t;hV~aIf9*%b>2BL8KECwJIgi)|4A_VL7G)iH
z9*SuibPuw66V@xze9f%dn2M#K_z8NXJ75Zqu=y=&f|+*N$N|v2wSy=>p;rd~9&rYC
z9{-ve{_)`w7lv^d9$xlvz|MC_6GkY`Lu=6m$qvF8{cnbf0Y>gYN_M_|y%gA$*gXC0
zhNCQo%3ipnaAv;Ubhxo^<xl$U#zWGFUMha)YK281{B_m&?DHKvGbRX2<t@H-z8|@b
zh;0}?Kx4--0ej*#@(~b?f!*N9`tIZ@ZjtqOdD@u_=U#$#nX8$Xl+}U4f++;qB$oPm
ze)a_zAhk1Bl>UcL#9XRrQeyB<@jU9|Y!R}f800~iM-Qc&M%;e*y8{KMOWwS_q4ay&
zbXHnvLU(&f%64cz?3?H=eH_JQ;gIMBGTDo_7(6KGD3lTynf(-RGBBHT(%8JY&+`4N
zI#^}-&3z|D-mEK}TF-gHL%&}U)RlVlLXb^JDSf*N3(s-fLXB6MwjEVf<(5@pO$p}@
z^-Rsp&3CY4CyVAgmOI(xYHhB+Hf$=)-mlqF>}>p5PPsC}9Q15#ZvAKv$$)SBV?jLL
z{)BEPJVRH5&K}>!-~DV(zsWyMHP?JXbYIG`{?zdRx<+AOlOGqC$usHOGv(bg=biOy
z>YUK{02j+0G4T?<*Y*PE_;$S6VQ+nels2uo<A@Qmuxo!w6{t{(cZ6>STvnR?g|7R9
zlfxWv-^eJclbMt2a}jcLvzi#{@vbXY@oy+V53hb8jfKNn{&0ayDK1>-w6C#UB7D6@
zWH)SXa!)<%eUFns%qg36OV}BZ6wto~vZB~Y2fgU#g@RAj=pi!Y=F#3ipQO=$Ebn1X
zd5^Rw7F_P7p?8)}u;WiAINy9t+<u4N`H*M2(snfO$PDNHz_DGUXZ|Q*%=tu&c(2P5
zj@$B9ApSxz5=W!`6=+tXrgieXzj>^lt{vcm7F0W?H9a&BkqL5cIksx&u73(ZR}Vn*
zjQEQeE(DK0_uLjOydPHhByGnn?Jx|Q5wOGTINyYIL%HpUw*E?rRyw}>S)7g#SZ^&k
zxmtiL>N`02dE98jEwmmm=eh7SAY=-E$u%W{;D44w+wyDV%>UL?{v5+OXa-BM4#9C2
z1gv<X{4Tdw6(7H{i951YcKJ9l`+q%sdpwi<|G)e0pi=Gb$epqdq=-?Wgr&M^6e=uf
zMv~+(G-R`>By~I8E$nt&vO6N?kQ}##P%YFP#;`fFu}#dzPQHuJ@AuyxkJt6O_Ikfg
z&*$rXJ+AjPYy~R>Mu^&^{3XbC(~fB^e}-vbVD@wRxHx9hg8bWCLH<0(xC?5~te&f-
zqN4J$q5XYy8qquNfk|HB0F69Ez#&IZ9_xU*t)upJxf^0~O-+3lbOD2ZS_HXtp!TqN
zlYI_mT}2vf`yg>Ilr%z*>fHFGxpgAfBXS4%eJJ5J`X_-}+PzNNh(klu5o6-Iq{X&Z
z)*d@gey3$l`k?#dIT7q_Y*i#UlhPd&p&D)0YE?X?8)CH^M4D}Fc!v@{I_qmgmU@2X
zza_jdEzW%FQvij;<~RYAKHYfdCS@O_l<};Iys)fI?%*{LHZ?y)z?R=5dfrUDV=WEw
zAFY#CZe1X285dv{V99ObZuJ~}c#`tWN+gBjI@+&lQE8H2G{SFRq#P%#?|!4{t)9^2
zzZR%+TRJl+lRO1fU2#n{gFyddfe0P?V#Bd~DQA45ueAY(S+;)Y9x4!b>*~;Ed7~fN
zP2YQAP5>(zjs>9b20)C=?mo9DV`AG%@CTwJmcnn_)Er}1(B=g=VP&4xKVF;gQ__+t
zGYeUyTmlX1{3`f-!*yD<jzn4PF2oPTd4Y;QsfzO_jRH!1=`E2hiFNjxQ#v;8{afTy
zBjJINGoz~ZvoR%$!xU-y{>AV;QznmXP}x3<A2xd6BFZAX1-R-RlKv2nGDK^(m>Kac
zd0sCX7x7VmY!V<F5zS?$9d?<%b>}18Y|-#+-L5st=$|jMrH|v1lIM_&0|V~RTK-bo
zp1m}u#d(j-i-c3vwg{%w&p;>Si`S(e6iMU#r-MJ`+oSr89IyH=VXWrb2a|TRY1=Te
zsxPYBXvS=dNQ?quv>s?)5rsh)#a_`ZVJ#hZhFVMF$$$x1_DoZx2sY2Dco7kPni6tI
zz${tHA%izB3?G-d4)h$ZlVK0*WrL_!3qY>3@m`c;y`(#?7!Cf$1*M(;p0hXXU>S4q
z9{6B{{3+3lF?8zb8@-`u5gXXkF(5v=FA!9FKD`{X;H4h5Zr^N-Dl%iL<{m{&2xC0|
z(H?tWI5U>8f314jc)!z(sH0y6+@E#Sq9on4)84mPKeQXEK5;q!9-$*Id9-o6Jm4rM
z1dP4j;P}{Q@wv252gvo!YwvkvbsiJqhzUVpTuN8HS0ti^Yuwo9;GW$mH-}U~P}nU=
zK6k(VP9bB5dPgN+(BQYqWA4QkDczs)tocPyNF~Lh+PBKpl4%$fpfKAr?!`Wm<%<}w
zKvx}OuanWzOu2P8Ci7vuOzZtyZF}2JVl;xnP5SI(<d@E$NK6+p5qZyIejp(pWq>+X
zYpmg-*8-&i2;RKt*UXCE@5#)F-yWF7nURt<E%<aVVb5V6z>U9)cW{dg3pXq@lU6<X
zhnc-(E5wIZ5?A6d%&K+91!_|NTi<%zx!E|<RJhMP@)C(EzE(TE*x(a)yiCf~e~j_y
z2jC~AC5z34WMgJe5VumTzcSZ2(=10#+!<l`;keby^?sU9c_H7JP65>L-@pQAf!=~u
zLjL7BTy=#u_=~7F?gZBx+-enmh=&W8{ANwbXSFpVD!q#l+RR>OD-VRhISd#!>!{Nu
z75}qR=tn~?&?|I!BpR*Qy}HWd5GTguR8p?K+M@1__*#V`Qy7sPtz(%LG6*H@r;Bes
zhCe7dvN(S!I=c4)sGfLvE^(*sJ#8B(q5h(JKSpHzd7(QJ_|SKgl-IN#r3eY%^^kWJ
zkvS`Ul@p6({Kif$yWhQ}z0rJZQrsFjl^k<g<d;{k-BWv{=grfB2V_<*<=8;-!LA~w
z|5gI;`Sqp&Coc^3C2X^}XA7>RxFKXS%<W^9^-qjc&?NvP(X}0Qz%^SI2A7Yaqav=u
z$M4o@0sk1I{gkV&8{RBMp2m)}45bGEu(#;alA8l>5YIclXpY2Y3+6*Y+U{6&u%KU?
zq~ld~#BNCd0Rvq_4n9ipf5vba$dNW4u~_yxyEy-01nkdbe!p0r`n!&FsHkv0X7hs2
ziRqOXpLG=Ag|)TWuWO%5FXOEAIe(g*O3R)9js0)?Tp$`Vy#wC8fJV@F6vuu&e`6Y>
zOpW!GB&(>~tP7w5h3%V)gWw`&-C>l-1~*2V0|gX!n%MVO{0NJTTfDpU;Ed)>X3(eX
z>Bj<~cT9Z<aQCMM`Vx{Ar7_1tdzaTNU~j|r2onJ!2(%JBjRpGK7z951Rz5B^-3&D#
z1J1De*Q1&X6&z=dbff6TUj-oPo?rF-2Rudi_)0Qa{~9A0fyq-_%=8aFc3yqGtaOQa
zUN@PWC>|_8JM8rczZmBB7)~ZpMieshs2&_yWIfKpva8PmYsw;~M=vrAgBYg8(<fW}
z?_|$n?$_3%z@vr2AUIj@ir~VjU}wjQ9XjhjMSKva%EEEU#rC#?vEn+y7V(4`F3r>i
z!kCj94XD?#p|O(ELTXyBN9=yb72c6@X5z7&9A>d4Y^%7(1gChnzSVep^FtuNDSCQM
z6?ZgFOpG&GbRyk~0@YZ4#4HR~*kJoG!3d3+L)5)&<V2rtU*H)L-_875;;Y|t)=2`E
z$fWj&w4pe3KWx;(y)wcjT=Uy}|6%_${8LN>ybd;#0pu5TxZdpjPjV0BcYbGt3m2-a
zYlafV_1B?SQRX-aCwx(xlU^H#o)+uK1GXf2Ys;j{6F4(FY?Hk)vNEYKQ)-nr3hCd+
z^u!MVeBc;{_u%Dg58nMO62@1|?5MH;|48atPY&h0eow>XKleyUs+}LU^v{86L@>Uc
zD*;m>5-)39JE6nMIizCB8#p%Vx!0pK@L(^3VA!4Poi2K8dNScoCQXSdVrl9*`8`qh
zD*JgPSpxw>C$E`r9cHytsaNP#-9y)HUY}9}-+V(AEN+uGvK*)Rqcl7x?kROGRUEfA
z@?G*K@pztKfT|Fjfc;HaxAg9wz<g3yi55s+qYiU-aYqwI>y>7<9@eeYQGG4Znw404
z-(HaTrFJxeR~PXcJLTkO`}5d<E{B;7^?wB>B;46E_*)%%Ej&6GTPfwL?&9LfaZgan
zbh|`~#tE-N?8B_qjQ+tdzOOTQzz?nL-P+3D4akb;r77+%?~?$NqrBYA<dvHiHvPVM
zR8R!?DpFFM5`t-i78*DMQ<qze=}JV+_ZPBmyYGJ5WE^VR+y8eJ?2Vv^&5=porRQK+
zr(jP7XGVQ#)1AAvTMgB{iOepH5}&FIXKiB}_QSttjnrRd>v!P|n#5JBl^bZG%?X9X
zaTRF^_SIE6lUBxk*JHf@OfR=VD<q=4<SD#XwVUeh<SlpK4-SdC@t_7!l#%BpvEXEO
zFd?z-J}`9c;7`!qVup#my04Ub2+T74nlaU9$@eQ21ZDWHLHKT@keK>Ky}E8gj!_Q*
z?G?)&rV>LnN%#_`uJYCC&ciFb;N|!ecWZ3xsrtf@P2!#+9ZV^yZG6YOcRz9fuWa!S
zT%J_nMmSXu^$`S+l>I}n8TmfDYr)#u*<w9-qMaT1w+q@^IvSBN)ohdaY05-ReL5=e
zv$#7ymjRvB+@d}g_?>sMXCBQ-+^8pl>=srShx4TrZ;E>%VNpj?BE<}Nbk#mM3d0I)
z`?ehFvW@K%uEfoAC}8h{Vj>`s_I6)YFoOIS-Q7QU?;)2;02|H@zfHvkvI9|fe_oeV
zzu*@E&PmIi2+RJpd`CFEb}cw$%R+*Cl0cKIgJO;RSzCw+(wfRtnBPMVPKx_zJd*N5
z^y<zNb>JgEAAzqwCK$GLh!|D<YJLR>8?Jtry0}OEJf_zu`4>E?bTyKa-1&Q3gAvzS
zRkg`)l=yko>m$hhP4sIr5j+qbXN8>gs0_Rq-h<kR?Z7O=PF`MSS29>(EoRx8S89<S
zd~+T$KQV*|ghypeyng!HCjWU3I|69hQ_;{zIHWmU(V=#U@}7ka)E9+UQc2%^8KJPB
zFOeJto&LB-*PzcdDt`|*ei!PTrP3OwGr&+6Y1t;)s0P1y0jIb^-=%E$FVyfHX-K4(
ztsFf8;81yWE@%g>FM!H3??}J!D3`_dy~H>{>|Qu2)a?t6VqEBOWp+L!XruZ@B5RNJ
zG2ddPuXv9!hc?LkQaUpEb#@rPaZ0B%h%};^7u)<Y!~Q(x>K?9UgKWv@J3y~Xy~%QA
zptydbPKSJv-dLSI60$`cx=#o`#Gxxm<>v+N8w4rPRt^B8l3C~>!|V-{BjSp+(N0pX
zoO&e~?orvnKUs1fd)u7h29+J_>MvbR5uwQU{Lnq8KyzD!QAgS{I3|ObY`E8)3PAzd
zm@$>Ue=2|2s%QYzYZGbOl}UVm#lx)tAnw7(O0fU05xwZuSH2)Ml>on!)kMBoTD422
zri7Nt??rt%<h?g7O4$9;byksMG+3PbDy-Bl>nwUDeM4mM0o%&u!XN;a?pKIkCq4EA
zUJ?Alkp9%6H|XRplN48O;HU+h>=@lH-^xA(@_&toN&hB8TKj7Rx;F9lx+a;&-uBrR
z0GjH_TdC@<Q@^^my<nxxKYrgE+JF^Q=gfsDtv#VLc#xB@XRp1%JLl#q27wTLNr0NV
zUn^Z49-b)z$)ZsGNPfcKXdY}0IY5{_fRkNtL?Gxid=AcKCNoiJ#=H}0`c9%VFPcVv
zf`#wCsV;M9W~rhfFq%N!UAK~wZ9o;fESHX83v;dNR|~xIzB64-XV9`8(Wd<bJ*|UN
zkceT*q`7g#{{wXMQn-yYUu(IXCH)LT^@eqWD=`Zg!v?tMdZ+wa(6?+7v!5&~>4%~G
zH1hl%0~qN)YJI$_D=fgFB)26G>B-Oi`N^19fV2tS=ewElH-NsW5+i3Nq>kjQCalqi
zG;Iu?nAsvX$VySx@i}^HU|Hq*)~lK>E;6Gr(EBtYBTm5B5=SYBvDc-yirtIB|2YBI
z%})=K`wHqDx>=ysj&s<ai-OEF)9HTq$w?)>3+^0WOf0BCfEP2{(?)$e(yI%}S3Z!i
z#vjmJZ<EV2Ni{nPp_*LUM3;IDh7q?-2R+@S0^iO5>lePOtpBq95!6nS2{~2tawoso
zUx!R33i_S(U4O=V&&znR)x}iv_qJwx>ur;(9JVU=Fd&ueoh+i!cg-rX2jvXq$_e9I
z{wE#gp0@C~TM_*cGBKJvo-zNYnGJ~Ti?C%j+h^D_{R+gxRRIX){H=~yg~>9XHLcUy
zgGNUY!xv)Fns*b7!TKEWhDF-0qD3zg@McM?yS=1$7{pnlD(HU6Yo3X8J;5xQmR?!v
zP_5u~j@sdeeC_8Fi+X)W1?3J)9UjcrMa;=J;K5xpJO!$1)2oVB%8r>`VIBzmAiump
zM+6g=%Yv)pyP<7G$}z#KnzU)PwZ7Vob)W+;(cO(hhCSS_%OW9yJxjlJJBJuzyF=xp
zJROW|R!csvw8Znq7jUJ{B291c4Vr{u;G)u_PGemg9J>+{u`^>j%)M=Iz{6t3q?eh2
zP--o53-f504$vtxwSAuPw_NpX-@LG(N4#kv(Q2}bWZI>_So&zCrM^pt;aHY_`4I=3
z$mO2m!w>|J^IP}z?yK^0JH^$39`N42eCnxsTV)e-+s&=K)kH3+D@0rpJ1*%_raciA
zkz+0NzToA7=3<oBbe3?q?WF8Z;(UE@5=NwCFj&Ly4ov3p9e%D5#h&;E^LO>?V)|+b
zp3l>8EaT^(fLl%Sb<yH8=B4uz_zU5O7h2A3sbSrq{+9w!6?mz@O0tN5k5+6;$Kr=3
z*%6l}{p+HzZig%xChVuuu;<9Z=RBroKta3JmAk}l@k~=}yO#6_JcoTeFuN?srs4|S
zO$3x-c4kh*rb-H(sPrre_6%EzDcx72R@t02nx;M3^jyg=A7wMHH3-gkPg$Vrj*SV<
zL^=~5M3EGVsTlL!RpoG}4{NDT-B6LO`d+8qQ<#rtOE}q#WG>dAwmeL^Ex=3e*I8;w
z{BE!ORu{Akq2H5!zk>NaVU95jzot<VpD6fy7+ZXfY^V1>P(8x3;M4?<DBb=LA$FE+
zy)1Iw{dDA3C%K$<b^G*0o^;`hF6A94>r_e$zq|!C=n@4jj`&Mf8cJ9nqR!W}spWr(
ze-Ak^-oWA!l2fN<woF$>$}cgvF)}m!TIsF%`fD2(d@j##_%+C%E$;Q4RqOozGElFP
zeEF$rI3E>56#r30y%w*cvdIo|!P<Eiy4$thWy2D=QPC7XFByA5nrg`daG&=)mRed*
zlaCZIGmr5#!}d)2P`?o~`Mf4W_W`0hC@b&XU0{tT!Qe|3VMDJ|G%M{g^SGJjw=#j1
zXhb0?FlwUo4zKJYsa)DTT027Z-NT}91c(P4rardA7!U~omvQJ+ec82LbLe%l_qZVO
zL<hJqGl|cjegl3F>2!#ubRGqN6%R^-31*!qvhScs=~O4*jJ|d`pX#vV&unU7#fQMG
zX=PD$T9A6TQLDA7+-dCZ7?+BAqk+{_d+|(1{z<#74E|+poHQcMYC`$XY+cSF+{Dud
zcfE@>=Hgvs<)s3F`z5AiK$ce<sCXEzG-CAU)o?u-Asa!&v)h^D2>1<>z!IC|jv`<{
zeizkqDqVaLtdxjWkYN*v35UO9i93H{HF7D7CM$@1Dli^bk|s3kMC#=m4)ohv)Zb^=
zYs)K%2Inp6pF5eHBlSKZJ?Qu7moF`wFaK*5F>_@wa^#hk$G;t{7sz&U|Fl3)3?hJ`
z`1h0L_3UTJAE*0~iVp^TSZU*Y;U8R@O?#gc3FolnV@+jXTyiPggDbSe=#uH}C@L}Y
z4ey=f<!X>keq00}l|!Va3R4d3V_i6^w&as7Dhle~DpO5M;v{(|b3{w6LPRUzKNqYf
zzazjbY~e@#=aYwENRcTlith%?r9Ai<HRT#ycW>vQUFwqc6V$GWo;Y#p;Cf|d1>ewm
zDGY@~TpIPuS`23En_DWE$+##<A?5>yXAfWP!^yE}Jay2A!f0_>3@UgSl70Ww54Vc@
zwTtx?T>&TQfZxBS4WV6zYxQPGhsQ{V|NpqJa{j&&3sQvRewO1qH2G4WgwyVHU2WLu
z!qJ?y!|W?ACT!?TQ{PhapAufUdvW*~z`}F?_xQ0b;xtwCu0~`xuW%FAc#LG)3*Be?
zQzO1gxjU?~`IE1<68A<sqs;TI2K3buKGd+5al*n!z^<Kzg%i?$1Z%Jwe0e26;WY(_
zX6dW-==HHZHV^vHgmK%rG|I}7nE@3-2IR)SqHB=qhy-I%T(puAr^kX%xuECcX>&z-
zOJ3DbpdiHMJG{+9hW&k&p=lh|`~&Mk?|<xe^IBps+%uuZhE6!HhbWZ{zJzZUdzFHU
zSYnCPpoj@|nT2i=9b6d+!pepU7l6b2IQs|iMlCn_HsgW7*#i#bk?qp*r|4mp?A97`
zxz4i7f#}Nrm01@H3jrGy@B4{9FaR}b$<pQI;Y#+*dO1{<2Wzd5(w|8Jx=P5mOsHS&
zhOw2o9=m9YqLp<SIaOVh9Dt~vE~ahcqm5d=*`e@6v#pCx`K~S%mb)HfyZ_SZUiR20
zbh#3_x)bL;Zix$r-23?<_0WL0LBqRlit0$8Fj-o6jd<)IJg*4Oxf+%9`j!SjllxYC
zOdV*ug8BD`|0Oc)C4V4n5l^I6e|l4_>8;w9-fh?~I}4a#_H;N&fOmWAZM9NG8fxMi
z`GWXocyYk%Mmawg<1djxzwt^_rC=PAm!WvQk%3sCFnA*!afOX!y>r;U(Z8!!<5YUb
zYCi{D^VL#vwi>)nsq~;&ZOQ%VoSN<>mufhtmN33pxMAGoz-o8u`dUvks}4?f**`#d
ztv?ZN0i(z1SVrHaBF{Oyu%U%hP|Ftgqp!ruAM(x0OnR7G1EUw|40aralUu6y1;b`M
zdVQCNC@%CfGr|vkU|@bB;Unxr3d^)b-=pwOpH#4mK)M%>={ts2YODAk4zJ<5ZxrPx
zeU+=+G_Gn=Y-8=`_}hl6c;EexJnyng2~%VE!K$l1IcI;3KAQ*E;7VR1HS-lhryAhH
zK{e?NaOD1Ab^Vv6_qQ(=o{J>uN#A1O&MVvC=0=8KOS~so!<nx-^LNcYP*^naj1AVv
zgi&-0ey1`~=qk~A^y@K%#zxHn><xNimZBjy6zo~2F6sk+_pJ?*IR8Xne|w8PjV9~O
z>szX){u<vswG(KNj)T1btyHNAvUwm$-3~<}`gU<Y-YXzbn^Mik#OMC~Me1z09+2fL
zbc6aW{xM(Wdzj=>!S1TKF{+*U=`s*f1?4hC#uhHPB(GCmYqVV#riI$tfDOFtJWiP~
zOcloAJWP4Izbj=QofpxN<@BH7P1fVs2O>L(0|Y#^=b}=g@=+W0eLM%hZ6+@6tNd=j
zKf-t%R(amDQt~jw7eBQ8fyln)J~z9KZ#%smBKQZHhylG>y6{U=!`)PQx(_{58<ScU
za&m)s!84OjcBjX`k*|vN&Fq;v90%5iAGbO%ma-zEx;-kESeGxq{Y$%mTnEEO8aGh1
zl$vwY!Dr7Hb|_D#<H|oXBR`I(PFBkvXL!0{F~u*1-Z<7V&f8Ax@n5P5DZGcv;gD!O
zzvor2u}XE{^NKLTF~Ps0m%lobXTgkyfSrO|qkI`_gLu@t#~`}wkdT7?-@Q#LTNB?)
z_vd0Q{@}MJ^Ip!%gKQLPZ&t0+DCnqhT884bJMb3w4<wuQr_WI@O0+T#{qo|ezf3<Z
z10hs>PG{DqpBJhLoAGw2JR$?aJkHbo^)%uvd;%6zFTTj69o1A;<d^bp9&Pzg2>!D4
zi&CoM#MOdCALOjT<&5pTyuMLBCIJ|ZQLoY^e&9cq;*TlrDMDj)E+Sj>()jrX@k_xm
z2r2x;Ji%=B?*AT-xQL<>0rsfJX3?SB`1!OGu+Ph0kpBO9bO-kU1Q?d92x5(K&Q=o~
zUJ&(l<j>S3>X)P&fBB_l;1EvH7%PdslQ1YeFpxJanuPO2lLYXB?^>l;1CU&U5t?=_
z?|6i*-NxT`&kprMZ3%=&ZuI??iUp~p+H|FueeCE2@6iVF^@a)^Ocha(D*Rntu>uIm
zTLgquVU=N5bk6<c<4@2+|4btOBI+$snKeh1=VGd;T;Bn>SbK#IzE}&PZ*w@|dak<l
zEvk@Mq>ag}v&cTNQF%!gj*L_M;AK|fF<Cb6b`NXCgys?vX+i^B$v?;%;^i!u@}gI?
z5J-d_(|`lN=__56bui1Y(X!`j0ud2<VEHuvmUtF1o}i3}r=0~y^qD<ws9aPg(KK2`
z<7A`d(=Y9e=nNtv1({Ctt)?+9nrBA*3a&SotuqMX#N+s3zp*@`|JwdY?YwohEi549
zKy+N4g+X;7$amI3xlqpwNj2M$QXR{&5@0ZToHuTS*ld#+oE+lSw7d@2{t{VIhj~*M
za36nSfN{A>7ELB5@)z3dnBVHQD2p^Kzz9*JjyXqvO^mEM$qLT}bM*v@xhF(h?R_}p
k_raXWj`fv?Im>JM6Rn>$X}XY<rl~;wa=1`(&hzpA0R~M^w*UYD


From b20efc370ac62d9e2317166ae3d552017b665702 Mon Sep 17 00:00:00 2001
From: Daniel Han <danielhanchen@gmail.com>
Date: Thu, 2 Apr 2026 04:45:09 -0700
Subject: [PATCH 008/216] Add regression tests for custom llama prebuilt
 installer (#4772)

Expand test coverage for install_llama_prebuilt.py:
- Add tests for source build plan resolution with custom repos
- Add tests for branch/commit/PR ref matching and normalization
- Add tests for manifest checksum validation
- Add tests for Windows CUDA upstream asset name patterns
- Update capsys checks to capture stderr after log() redirect
---
 .../test_install_llama_prebuilt_logic.py      |   6 +-
 tests/studio/install/test_pr4562_bugfixes.py  |  82 +++-
 tests/studio/install/test_selection_logic.py  | 384 +++++++++++++++++-
 3 files changed, 445 insertions(+), 27 deletions(-)

diff --git a/tests/studio/install/test_install_llama_prebuilt_logic.py b/tests/studio/install/test_install_llama_prebuilt_logic.py
index cfbbdeb66c..79dab30129 100644
--- a/tests/studio/install/test_install_llama_prebuilt_logic.py
+++ b/tests/studio/install/test_install_llama_prebuilt_logic.py
@@ -509,7 +509,8 @@ def test_activate_install_tree_restores_existing_install_after_activation_failur
     assert not staging_dir.exists()
     assert not (tmp_path / ".staging").exists()
 
-    output = capsys.readouterr().out
+    captured = capsys.readouterr()
+    output = captured.out + captured.err
     assert "moving existing install to rollback path" in output
     assert "restored previous install from rollback path" in output
 
@@ -571,7 +572,8 @@ def test_activate_install_tree_cleans_all_paths_when_rollback_restore_fails(
     assert not staging_dir.exists()
     assert not (tmp_path / ".staging").exists()
 
-    output = capsys.readouterr().out
+    captured = capsys.readouterr()
+    output = captured.out + captured.err
     assert "rollback after failed activation also failed: restore failed" in output
     assert (
         "cleaning staging, install, and rollback paths before source build fallback"
diff --git a/tests/studio/install/test_pr4562_bugfixes.py b/tests/studio/install/test_pr4562_bugfixes.py
index 32787aee34..f5a423d130 100644
--- a/tests/studio/install/test_pr4562_bugfixes.py
+++ b/tests/studio/install/test_pr4562_bugfixes.py
@@ -6,7 +6,7 @@ Tests cover:
   - Bug 2: Source-build fallback ignores pinned tag (both .sh and .ps1)
   - Bug 3: Unix fallback deletes install before checking prerequisites
   - Bug 4: Linux LD_LIBRARY_PATH missing build/bin
-  - "latest" tag resolution fallback chain (helper -> raw)
+  - "latest" tag resolution fallback chain (helper only)
   - Cross-platform binary_env (Linux, macOS, Windows)
   - Edge cases: malformed JSON, empty responses, env overrides
 
@@ -296,6 +296,56 @@ class TestResolveRequestedLlamaTag:
 
         assert resolve_requested_llama_tag("latest", "unslothai/llama.cpp") == "b8999"
 
+    def test_latest_with_published_release_tag_passes_pin_through(
+        self, monkeypatch: pytest.MonkeyPatch
+    ):
+        captured = {}
+
+        def fake_resolve(requested_tag, published_repo, published_release_tag = ""):
+            captured["requested_tag"] = requested_tag
+            captured["published_repo"] = published_repo
+            captured["published_release_tag"] = published_release_tag
+            return MOD.ResolvedPublishedRelease(
+                bundle = PublishedReleaseBundle(
+                    repo = published_repo,
+                    release_tag = published_release_tag,
+                    upstream_tag = "b9001",
+                    assets = {},
+                    manifest_asset_name = "llama-prebuilt-manifest.json",
+                    artifacts = [],
+                    selection_log = [],
+                ),
+                checksums = ApprovedReleaseChecksums(
+                    repo = published_repo,
+                    release_tag = published_release_tag,
+                    upstream_tag = "b9001",
+                    artifacts = {
+                        source_archive_logical_name("b9001"): ApprovedArtifactHash(
+                            asset_name = source_archive_logical_name("b9001"),
+                            sha256 = "a" * 64,
+                            repo = "ggml-org/llama.cpp",
+                            kind = "upstream-source",
+                        )
+                    },
+                ),
+            )
+
+        monkeypatch.setattr(MOD, "resolve_published_release", fake_resolve)
+
+        assert (
+            resolve_requested_llama_tag(
+                "latest",
+                "unslothai/llama.cpp",
+                "llama-prebuilt-main",
+            )
+            == "b9001"
+        )
+        assert captured == {
+            "requested_tag": "latest",
+            "published_repo": "unslothai/llama.cpp",
+            "published_release_tag": "llama-prebuilt-main",
+        }
+
 
 # =========================================================================
 # TEST GROUP C: setup.sh logic (bash subprocess tests)
@@ -586,18 +636,29 @@ class TestSourceCodePatterns:
         content = SETUP_SH.read_text()
         assert "_CLONE_ARGS=(git clone --depth 1)" in content
         assert (
-            '_CLONE_ARGS+=(--branch "$_RESOLVED_LLAMA_TAG")' in content
-        ), "_CLONE_ARGS should be extended with --branch $_RESOLVED_LLAMA_TAG"
+            '_CLONE_ARGS+=(--branch "$_RESOLVED_SOURCE_REF")' in content
+        ), "_CLONE_ARGS should be extended with --branch $_RESOLVED_SOURCE_REF"
         # Verify the guard: --branch is only used when tag is not "latest"
         assert (
-            '_RESOLVED_LLAMA_TAG" != "latest"' in content
+            '_RESOLVED_SOURCE_REF" != "latest"' in content
         ), "Should guard against literal 'latest' tag"
 
+    def test_setup_sh_source_build_uses_helper_resolution(self):
+        """Shell source fallback should consult the helper for repo/ref planning."""
+        content = SETUP_SH.read_text()
+        assert "--resolve-source-build" in content
+        assert "--output-format json" in content
+        assert "_RESOLVED_SOURCE_URL" in content
+        assert "_RESOLVED_SOURCE_REF_KIND" in content
+        assert "_RESOLVED_SOURCE_REF" in content
+
     def test_setup_sh_latest_resolution_uses_helper_only(self):
         """Shell fallback should rely on helper output, not raw GitHub API tag_name."""
         content = SETUP_SH.read_text()
         assert "--resolve-install-tag" in content
         assert "--resolve-llama-tag" in content
+        assert 'tail -n 1 "$_RESOLVE_LLAMA_LOG"' not in content
+        assert "json.load" in content
         assert "_HELPER_RELEASE_REPO}/releases/latest" not in content
         assert "ggml-org/llama.cpp/releases/latest" not in content
 
@@ -677,7 +738,7 @@ class TestSourceCodePatterns:
     def test_setup_ps1_clone_uses_branch_tag(self):
         """PS1 clone should use --branch with the resolved tag."""
         content = SETUP_PS1.read_text()
-        assert "--branch" in content and "$ResolvedLlamaTag" in content
+        assert "--branch" in content and "$ResolvedSourceRef" in content
         # The old commented-out line should be gone
         assert "# git clone --depth 1 --branch" not in content
 
@@ -703,9 +764,20 @@ class TestSourceCodePatterns:
         content = SETUP_PS1.read_text()
         assert "--resolve-install-tag" in content
         assert "--resolve-llama-tag" in content
+        assert '--output-format", "json"' in content
+        assert "ConvertFrom-Json" in content
         assert "$HelperReleaseRepo/releases/latest" not in content
         assert "ggml-org/llama.cpp/releases/latest" not in content
 
+    def test_setup_ps1_source_build_uses_helper_resolution(self):
+        """PS1 source fallback should consult the helper for repo/ref planning."""
+        content = SETUP_PS1.read_text()
+        assert "--resolve-source-build" in content
+        assert '--output-format", "json"' in content
+        assert "$ResolvedSourceUrl" in content
+        assert "$ResolvedSourceRefKind" in content
+        assert "$ResolvedSourceRef" in content
+
     def test_binary_env_linux_has_binary_parent(self):
         """The Linux branch of binary_env should include binary_path.parent."""
         content = MODULE_PATH.read_text()
diff --git a/tests/studio/install/test_selection_logic.py b/tests/studio/install/test_selection_logic.py
index d7bea4bb5c..b8de0e221f 100644
--- a/tests/studio/install/test_selection_logic.py
+++ b/tests/studio/install/test_selection_logic.py
@@ -58,7 +58,21 @@ resolve_requested_install_tag = INSTALL_LLAMA_PREBUILT.resolve_requested_install
 resolve_install_attempts = INSTALL_LLAMA_PREBUILT.resolve_install_attempts
 resolve_install_release_plans = INSTALL_LLAMA_PREBUILT.resolve_install_release_plans
 resolve_published_release = INSTALL_LLAMA_PREBUILT.resolve_published_release
+resolve_source_build_plan = INSTALL_LLAMA_PREBUILT.resolve_source_build_plan
+validated_checksums_for_bundle = INSTALL_LLAMA_PREBUILT.validated_checksums_for_bundle
+parse_approved_release_checksums = (
+    INSTALL_LLAMA_PREBUILT.parse_approved_release_checksums
+)
+published_release_matches_request = (
+    INSTALL_LLAMA_PREBUILT.published_release_matches_request
+)
+exact_source_archive_logical_name = (
+    INSTALL_LLAMA_PREBUILT.exact_source_archive_logical_name
+)
 source_archive_logical_name = INSTALL_LLAMA_PREBUILT.source_archive_logical_name
+windows_cuda_upstream_asset_names = (
+    INSTALL_LLAMA_PREBUILT.windows_cuda_upstream_asset_names
+)
 env_int = INSTALL_LLAMA_PREBUILT.env_int
 
 
@@ -110,6 +124,13 @@ def make_release(artifacts, **overrides):
         repo = "unslothai/llama.cpp",
         release_tag = "v1.0",
         upstream_tag = "b8508",
+        source_repo = None,
+        source_repo_url = None,
+        source_ref_kind = None,
+        requested_source_ref = None,
+        resolved_source_ref = None,
+        source_commit = None,
+        source_commit_short = None,
         assets = {a.asset_name: f"https://example.com/{a.asset_name}" for a in artifacts},
         manifest_asset_name = "llama-prebuilt-manifest.json",
         artifacts = artifacts,
@@ -124,7 +145,13 @@ def make_checksums(asset_names):
         repo = "unslothai/llama.cpp",
         release_tag = "v1.0",
         upstream_tag = "b8508",
+        source_repo = None,
+        source_repo_url = None,
+        source_ref_kind = None,
+        requested_source_ref = None,
+        resolved_source_ref = None,
         source_commit = None,
+        source_commit_short = None,
         artifacts = {
             name: ApprovedArtifactHash(
                 asset_name = name,
@@ -142,29 +169,56 @@ def make_checksums_with_source(
     *,
     release_tag = "v1.0",
     upstream_tag = "b8508",
+    source_repo = None,
+    source_repo_url = None,
+    source_ref_kind = None,
+    requested_source_ref = None,
+    resolved_source_ref = None,
+    source_commit = None,
 ):
+    artifacts = {
+        **{
+            name: ApprovedArtifactHash(
+                asset_name = name,
+                sha256 = "a" * 64,
+                repo = "unslothai/llama.cpp",
+                kind = "prebuilt",
+            )
+            for name in asset_names
+        },
+        source_archive_logical_name(upstream_tag): ApprovedArtifactHash(
+            asset_name = source_archive_logical_name(upstream_tag),
+            sha256 = "b" * 64,
+            repo = "ggml-org/llama.cpp",
+            kind = "upstream-source",
+        ),
+    }
+    normalized_source_commit = (
+        source_commit.lower() if isinstance(source_commit, str) else None
+    )
+    if normalized_source_commit:
+        artifacts[exact_source_archive_logical_name(normalized_source_commit)] = (
+            ApprovedArtifactHash(
+                asset_name = exact_source_archive_logical_name(normalized_source_commit),
+                sha256 = "c" * 64,
+                repo = source_repo or "example/custom-llama.cpp",
+                kind = "exact-source",
+            )
+        )
     return ApprovedReleaseChecksums(
         repo = "unslothai/llama.cpp",
         release_tag = release_tag,
         upstream_tag = upstream_tag,
-        source_commit = None,
-        artifacts = {
-            **{
-                name: ApprovedArtifactHash(
-                    asset_name = name,
-                    sha256 = "a" * 64,
-                    repo = "unslothai/llama.cpp",
-                    kind = "prebuilt",
-                )
-                for name in asset_names
-            },
-            source_archive_logical_name(upstream_tag): ApprovedArtifactHash(
-                asset_name = source_archive_logical_name(upstream_tag),
-                sha256 = "b" * 64,
-                repo = "ggml-org/llama.cpp",
-                kind = "upstream-source",
-            ),
-        },
+        source_repo = source_repo,
+        source_repo_url = source_repo_url,
+        source_ref_kind = source_ref_kind,
+        requested_source_ref = requested_source_ref,
+        resolved_source_ref = resolved_source_ref,
+        source_commit = normalized_source_commit,
+        source_commit_short = normalized_source_commit[:7]
+        if normalized_source_commit
+        else None,
+        artifacts = artifacts,
     )
 
 
@@ -432,6 +486,31 @@ class TestApplyApprovedHashes:
         assert len(result) == 1
         assert result[0].name == "a.tar.gz"
 
+    def test_upstream_asset_can_match_compatibility_tag_name(self):
+        choice = AssetChoice(
+            repo = UPSTREAM_REPO,
+            tag = "main",
+            name = "llama-main-bin-macos-arm64.tar.gz",
+            url = "https://x/llama-main-bin-macos-arm64.tar.gz",
+            source_label = "upstream",
+        )
+        checksums = ApprovedReleaseChecksums(
+            repo = "unslothai/llama.cpp",
+            release_tag = "r1",
+            upstream_tag = "b9000",
+            artifacts = {
+                "llama-b9000-bin-macos-arm64.tar.gz": ApprovedArtifactHash(
+                    asset_name = "llama-b9000-bin-macos-arm64.tar.gz",
+                    sha256 = "a" * 64,
+                    repo = UPSTREAM_REPO,
+                    kind = "macos-arm64-upstream",
+                )
+            },
+        )
+
+        result = apply_approved_hashes([choice], checksums)
+        assert result[0].expected_sha256 == "a" * 64
+
     def test_none_approved(self):
         c1 = self._choice("missing.tar.gz")
         checksums = make_checksums(["other.tar.gz"])
@@ -536,6 +615,259 @@ class TestPublishedReleaseResolution:
                 "unslothai/llama.cpp",
             )
 
+    def test_request_matches_requested_source_ref(self, monkeypatch):
+        release = make_release(
+            [],
+            release_tag = "release-main",
+            upstream_tag = "b9000",
+            requested_source_ref = "main",
+            resolved_source_ref = "refs/heads/main",
+        )
+        monkeypatch.setattr(
+            INSTALL_LLAMA_PREBUILT,
+            "iter_published_release_bundles",
+            lambda repo, published_release_tag = "": iter([release]),
+        )
+        monkeypatch.setattr(
+            INSTALL_LLAMA_PREBUILT,
+            "load_approved_release_checksums",
+            lambda repo, release_tag: make_checksums_with_source(
+                [],
+                release_tag = release_tag,
+                upstream_tag = "b9000",
+                requested_source_ref = "main",
+                resolved_source_ref = "refs/heads/main",
+            ),
+        )
+
+        resolved = resolve_published_release("main", "unslothai/llama.cpp")
+        assert resolved.bundle.release_tag == "release-main"
+
+    def test_request_matches_source_commit(self, monkeypatch):
+        commit = "a" * 40
+        release = make_release(
+            [],
+            release_tag = "release-commit",
+            upstream_tag = "b9000",
+            source_commit = commit,
+        )
+        monkeypatch.setattr(
+            INSTALL_LLAMA_PREBUILT,
+            "iter_published_release_bundles",
+            lambda repo, published_release_tag = "": iter([release]),
+        )
+        monkeypatch.setattr(
+            INSTALL_LLAMA_PREBUILT,
+            "load_approved_release_checksums",
+            lambda repo, release_tag: make_checksums_with_source(
+                [],
+                release_tag = release_tag,
+                upstream_tag = "b9000",
+                source_commit = commit,
+            ),
+        )
+
+        resolved = resolve_published_release(commit, "unslothai/llama.cpp")
+        assert resolved.bundle.release_tag == "release-commit"
+
+
+class TestSourceBuildPlanResolution:
+    def test_matches_request_by_non_tag_provenance(self):
+        bundle = make_release(
+            [],
+            requested_source_ref = "main",
+            resolved_source_ref = "refs/heads/main",
+            source_commit = "a" * 40,
+        )
+        assert published_release_matches_request(bundle, "main") is True
+        assert published_release_matches_request(bundle, "refs/heads/main") is True
+        assert published_release_matches_request(bundle, "a" * 12) is True
+        assert published_release_matches_request(bundle, "a" * 40) is True
+
+    def test_matches_pull_ref_aliases(self):
+        bundle = make_release(
+            [],
+            requested_source_ref = "refs/pull/123/head",
+            resolved_source_ref = "pull/123/head",
+        )
+        assert published_release_matches_request(bundle, "refs/pull/123/head") is True
+        assert published_release_matches_request(bundle, "pull/123/head") is True
+
+    def test_prefers_exact_source_commit_when_available(self, monkeypatch):
+        commit = "a" * 40
+        resolved = INSTALL_LLAMA_PREBUILT.ResolvedPublishedRelease(
+            bundle = make_release(
+                [],
+                release_tag = "release-main",
+                upstream_tag = "b9000",
+                source_repo = "example/custom-llama.cpp",
+                source_repo_url = "https://github.com/example/custom-llama.cpp",
+                source_ref_kind = "branch",
+                requested_source_ref = "main",
+                resolved_source_ref = "refs/heads/main",
+                source_commit = commit,
+            ),
+            checksums = make_checksums_with_source(
+                [],
+                release_tag = "release-main",
+                upstream_tag = "b9000",
+                source_repo = "example/custom-llama.cpp",
+                source_repo_url = "https://github.com/example/custom-llama.cpp",
+                source_ref_kind = "branch",
+                requested_source_ref = "main",
+                resolved_source_ref = "refs/heads/main",
+                source_commit = commit,
+            ),
+        )
+        monkeypatch.setattr(
+            INSTALL_LLAMA_PREBUILT,
+            "resolve_published_release",
+            lambda requested_tag, published_repo, published_release_tag = "": resolved,
+        )
+
+        plan = resolve_source_build_plan("main", "unslothai/llama.cpp")
+        assert plan.source_url == "https://github.com/example/custom-llama.cpp"
+        assert plan.source_ref_kind == "commit"
+        assert plan.source_ref == commit
+        assert plan.compatibility_upstream_tag == "b9000"
+
+    def test_uses_branch_provenance_without_exact_source_hash(self, monkeypatch):
+        resolved = INSTALL_LLAMA_PREBUILT.ResolvedPublishedRelease(
+            bundle = make_release(
+                [],
+                release_tag = "release-main",
+                upstream_tag = "b9000",
+                source_repo = "example/custom-llama.cpp",
+                source_repo_url = "https://github.com/example/custom-llama.cpp",
+                source_ref_kind = "branch",
+                requested_source_ref = "main",
+                resolved_source_ref = "main",
+            ),
+            checksums = make_checksums_with_source(
+                [],
+                release_tag = "release-main",
+                upstream_tag = "b9000",
+                source_repo = "example/custom-llama.cpp",
+                source_repo_url = "https://github.com/example/custom-llama.cpp",
+                source_ref_kind = "branch",
+                requested_source_ref = "main",
+                resolved_source_ref = "main",
+                source_commit = None,
+            ),
+        )
+        monkeypatch.setattr(
+            INSTALL_LLAMA_PREBUILT,
+            "resolve_published_release",
+            lambda requested_tag, published_repo, published_release_tag = "": resolved,
+        )
+
+        plan = resolve_source_build_plan("main", "unslothai/llama.cpp")
+        assert plan.source_url == "https://github.com/example/custom-llama.cpp"
+        assert plan.source_ref_kind == "branch"
+        assert plan.source_ref == "main"
+        assert plan.compatibility_upstream_tag == "b9000"
+
+    def test_direct_main_request_without_published_release_uses_branch_kind(
+        self, monkeypatch
+    ):
+        monkeypatch.setattr(
+            INSTALL_LLAMA_PREBUILT,
+            "resolve_published_release",
+            lambda requested_tag, published_repo, published_release_tag = "": (
+                _ for _ in ()
+            ).throw(PrebuiltFallback("missing")),
+        )
+
+        plan = resolve_source_build_plan("main", "unslothai/llama.cpp")
+        assert plan.source_url == "https://github.com/ggml-org/llama.cpp"
+        assert plan.source_ref_kind == "branch"
+        assert plan.source_ref == "main"
+
+
+class TestParseApprovedReleaseChecksums:
+    def test_rejects_wrong_component(self):
+        with pytest.raises(RuntimeError, match = "did not describe llama.cpp"):
+            parse_approved_release_checksums(
+                "repo/test",
+                "r1",
+                {
+                    "schema_version": 1,
+                    "component": "other",
+                    "release_tag": "r1",
+                    "upstream_tag": "b8508",
+                    "artifacts": {},
+                },
+            )
+
+    def test_rejects_mismatched_release_tag(self):
+        with pytest.raises(RuntimeError, match = "did not match pinned release tag"):
+            parse_approved_release_checksums(
+                "repo/test",
+                "r1",
+                {
+                    "schema_version": 1,
+                    "component": "llama.cpp",
+                    "release_tag": "r2",
+                    "upstream_tag": "b8508",
+                    "artifacts": {},
+                },
+            )
+
+    def test_rejects_bad_sha256(self):
+        with pytest.raises(RuntimeError, match = "valid sha256"):
+            parse_approved_release_checksums(
+                "repo/test",
+                "r1",
+                {
+                    "schema_version": 1,
+                    "component": "llama.cpp",
+                    "release_tag": "r1",
+                    "upstream_tag": "b8508",
+                    "artifacts": {
+                        "asset.tar.gz": {
+                            "sha256": "bad-digest",
+                        }
+                    },
+                },
+            )
+
+    def test_rejects_unsupported_schema_version(self):
+        with pytest.raises(RuntimeError, match = "schema_version=2 is unsupported"):
+            parse_approved_release_checksums(
+                "repo/test",
+                "r1",
+                {
+                    "schema_version": 2,
+                    "component": "llama.cpp",
+                    "release_tag": "r1",
+                    "upstream_tag": "b8508",
+                    "artifacts": {},
+                },
+            )
+
+
+class TestValidatedChecksumsForBundle:
+    def test_rejects_manifest_checksum_mismatch(self, monkeypatch):
+        bundle = make_release([], release_tag = "r1", upstream_tag = "b8508")
+        bundle.manifest_sha256 = "a" * 64
+        checksums = make_checksums_with_source(
+            [], release_tag = "r1", upstream_tag = "b8508"
+        )
+        checksums.artifacts[bundle.manifest_asset_name] = ApprovedArtifactHash(
+            asset_name = bundle.manifest_asset_name,
+            sha256 = "b" * 64,
+            repo = "unslothai/llama.cpp",
+            kind = "published-manifest",
+        )
+        monkeypatch.setattr(
+            INSTALL_LLAMA_PREBUILT,
+            "load_approved_release_checksums",
+            lambda repo, release_tag: checksums,
+        )
+
+        with pytest.raises(PrebuiltFallback, match = "manifest checksum"):
+            validated_checksums_for_bundle("unslothai/llama.cpp", bundle)
+
 
 # ===========================================================================
 # K. linux_cuda_choice_from_release -- core selection
@@ -1360,10 +1692,13 @@ class TestResolveInstallReleasePlans:
 class TestWindowsCudaAttempts:
     TAG = "b8508"
 
-    def _upstream(self, *runtime_versions):
+    def _upstream(self, *runtime_versions, current_names: bool = False):
         assets = {}
         for rv in runtime_versions:
-            name = f"llama-{self.TAG}-bin-win-cuda-{rv}-x64.zip"
+            if current_names:
+                name = f"cudart-llama-bin-win-cuda-{rv}-x64.zip"
+            else:
+                name = f"llama-{self.TAG}-bin-win-cuda-{rv}-x64.zip"
             assets[name] = f"https://example.com/{name}"
         return assets
 
@@ -1428,6 +1763,15 @@ class TestWindowsCudaAttempts:
         result = windows_cuda_attempts(host, self.TAG, assets, None)
         assert len(result) == 2
 
+    def test_current_upstream_names_are_supported(self, monkeypatch):
+        mock_windows_runtime(monkeypatch, ["cuda13", "cuda12"])
+        host = make_host(system = "Windows", machine = "AMD64", driver_cuda_version = (13, 1))
+        assets = self._upstream("13.1", "12.4", current_names = True)
+        result = windows_cuda_attempts(host, self.TAG, assets, None)
+        assert len(result) == 2
+        assert result[0].name == "cudart-llama-bin-win-cuda-13.1-x64.zip"
+        assert result[1].name == "cudart-llama-bin-win-cuda-12.4-x64.zip"
+
 
 # ===========================================================================
 # O. resolve_upstream_asset_choice -- platform routing

From 1ce8a8e7cdb37f5ab613a8f1468dd0c3b92fb5aa Mon Sep 17 00:00:00 2001
From: DoubleMathew <mmathew23@gmail.com>
Date: Thu, 2 Apr 2026 06:52:26 -0500
Subject: [PATCH 009/216] Feat/custom llama prebuilt (#4771)

* update logic to incorporate custom prebuilt installs

* bug fixes

* update for review comments

* fix tags

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

* Separate test changes from main PR

Move test file changes out of this PR to keep the diff focused on
the install_llama_prebuilt.py and setup script changes. Test updates
will be submitted in a follow-up PR.

* Fix branch ref normalization and harden JSON parsing

- Add checkout_friendly_ref() to strip refs/heads/ prefix from branch
  refs before emitting them in SourceBuildPlan. git clone --branch does
  not accept fully qualified refs like refs/heads/main.
- Apply normalization in source_build_plan_for_release() and the
  direct-ref fallback in resolve_source_build_plan().
- Allow validated_checksums_for_bundle() to accept releases that carry
  only an exact-commit source archive without the legacy upstream-tag
  source tarball.
- Add 2>/dev/null || true guards to all inline python -c JSON parsing
  in setup.sh so a malformed payload does not abort the script under
  set -e.

* Fix Windows CUDA asset ordering and tag ref normalization

- Reorder windows_cuda_upstream_asset_names to prefer the main binary
  archive (llama-{tag}-bin-win-cuda-*) over the cudart sidecar archive
  (cudart-llama-bin-win-cuda-*). The cudart ZIP only contains CUDA
  runtime DLLs, not llama-server or llama-quantize binaries.
- Extend checkout_friendly_ref to also strip refs/tags/ prefix for tag
  refs, matching the refs/heads/ handling for branch refs.

* Simplify JSON parsing consistency in setup.sh

Use json.load(sys.stdin) consistently for all inline JSON parsing
in setup.sh, instead of the more complex json.loads(raw) pattern
on the install-tag resolution path. The 2>/dev/null || true guard
already handles empty/malformed input gracefully.

* Fix source build plan fallback for commit ref kind in PR #4771

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

---------

Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com>
Co-authored-by: Daniel Han <daniel@unsloth.ai>
Co-authored-by: Daniel Han <danielhanchen@gmail.com>
---
 studio/install_llama_prebuilt.py | 710 +++++++++++++++++++++++++++----
 studio/setup.ps1                 | 145 ++++++-
 studio/setup.sh                  | 102 ++++-
 3 files changed, 857 insertions(+), 100 deletions(-)

diff --git a/studio/install_llama_prebuilt.py b/studio/install_llama_prebuilt.py
index 0f96cf34eb..76d22984d7 100755
--- a/studio/install_llama_prebuilt.py
+++ b/studio/install_llama_prebuilt.py
@@ -28,7 +28,7 @@ import urllib.parse
 import urllib.request
 import zipfile
 from contextlib import contextmanager
-from dataclasses import dataclass
+from dataclasses import dataclass, field
 
 try:
     from filelock import FileLock, Timeout as FileLockTimeout
@@ -59,7 +59,6 @@ def env_int(name: str, default: int, *, minimum: int | None = None) -> int:
     return value
 
 
-APPROVED_PREBUILT_LLAMA_TAG = "b8508"
 DEFAULT_LLAMA_TAG = os.environ.get("UNSLOTH_LLAMA_TAG", "latest")
 DEFAULT_PUBLISHED_REPO = os.environ.get(
     "UNSLOTH_LLAMA_RELEASE_REPO", "unslothai/llama.cpp"
@@ -151,10 +150,18 @@ class PublishedReleaseBundle:
     repo: str
     release_tag: str
     upstream_tag: str
-    assets: dict[str, str]
-    manifest_asset_name: str
-    artifacts: list[PublishedLlamaArtifact]
-    selection_log: list[str]
+    manifest_sha256: str | None = None
+    source_repo: str | None = None
+    source_repo_url: str | None = None
+    source_ref_kind: str | None = None
+    requested_source_ref: str | None = None
+    resolved_source_ref: str | None = None
+    source_commit: str | None = None
+    source_commit_short: str | None = None
+    assets: dict[str, str] = field(default_factory = dict)
+    manifest_asset_name: str = DEFAULT_PUBLISHED_MANIFEST_ASSET
+    artifacts: list[PublishedLlamaArtifact] = field(default_factory = list)
+    selection_log: list[str] = field(default_factory = list)
 
 
 @dataclass
@@ -188,8 +195,14 @@ class ApprovedReleaseChecksums:
     repo: str
     release_tag: str
     upstream_tag: str
-    source_commit: str | None
-    artifacts: dict[str, ApprovedArtifactHash]
+    source_repo: str | None = None
+    source_repo_url: str | None = None
+    source_ref_kind: str | None = None
+    requested_source_ref: str | None = None
+    resolved_source_ref: str | None = None
+    source_commit: str | None = None
+    source_commit_short: str | None = None
+    artifacts: dict[str, ApprovedArtifactHash] = field(default_factory = dict)
 
 
 @dataclass(frozen = True)
@@ -198,6 +211,19 @@ class ResolvedPublishedRelease:
     checksums: ApprovedReleaseChecksums
 
 
+@dataclass(frozen = True)
+class SourceBuildPlan:
+    source_url: str
+    source_ref: str
+    source_ref_kind: str
+    compatibility_upstream_tag: str
+    source_repo: str | None = None
+    source_repo_url: str | None = None
+    requested_source_ref: str | None = None
+    resolved_source_ref: str | None = None
+    source_commit: str | None = None
+
+
 @dataclass(frozen = True)
 class InstallReleasePlan:
     requested_tag: str
@@ -270,7 +296,7 @@ def is_busy_lock_error(exc: BaseException) -> bool:
 
 
 def log(message: str) -> None:
-    print(f"[llama-prebuilt] {message}")
+    print(f"[llama-prebuilt] {message}", file = sys.stderr)
 
 
 def log_lines(lines: Iterable[str]) -> None:
@@ -358,6 +384,10 @@ def source_archive_logical_name(upstream_tag: str) -> str:
     return f"llama.cpp-source-{upstream_tag}.tar.gz"
 
 
+def exact_source_archive_logical_name(source_commit: str) -> str:
+    return f"llama.cpp-source-commit-{source_commit}.tar.gz"
+
+
 def sha256_file(path: Path) -> str:
     digest = hashlib.sha256()
     with path.open("rb") as handle:
@@ -366,6 +396,10 @@ def sha256_file(path: Path) -> str:
     return digest.hexdigest()
 
 
+def sha256_bytes(data: bytes) -> str:
+    return hashlib.sha256(data).hexdigest()
+
+
 def normalize_sha256_digest(value: str | None) -> str | None:
     if not isinstance(value, str) or not value:
         return None
@@ -377,6 +411,156 @@ def normalize_sha256_digest(value: str | None) -> str | None:
     return lowered
 
 
+def normalize_source_ref_kind(value: str | None) -> str | None:
+    if not isinstance(value, str):
+        return None
+    normalized = value.strip().lower()
+    if normalized in {"tag", "branch", "pull", "commit", "custom"}:
+        return normalized
+    return None
+
+
+def normalize_source_commit(value: str | None) -> str | None:
+    if not isinstance(value, str):
+        return None
+    normalized = value.strip().lower()
+    if len(normalized) < 7 or len(normalized) > 40:
+        return None
+    if any(ch not in "0123456789abcdef" for ch in normalized):
+        return None
+    return normalized
+
+
+def validate_schema_version(payload: dict[str, Any], *, label: str) -> None:
+    schema_version = payload.get("schema_version")
+    if schema_version is None:
+        return
+    try:
+        normalized = int(schema_version)
+    except (TypeError, ValueError) as exc:
+        raise RuntimeError(f"{label} schema_version was not an integer") from exc
+    if normalized != 1:
+        raise RuntimeError(f"{label} schema_version={normalized} is unsupported")
+
+
+def repo_slug_from_source(value: str | None) -> str | None:
+    if not isinstance(value, str):
+        return None
+    normalized = value.strip()
+    if not normalized:
+        return None
+    normalized = normalized.removesuffix(".git")
+    if normalized.startswith("https://github.com/"):
+        slug = normalized[len("https://github.com/") :]
+    elif normalized.startswith("http://github.com/"):
+        slug = normalized[len("http://github.com/") :]
+    elif normalized.startswith("git@github.com:"):
+        slug = normalized[len("git@github.com:") :]
+    else:
+        slug = normalized
+    slug = slug.strip("/")
+    parts = slug.split("/")
+    if len(parts) != 2 or not all(parts):
+        return None
+    return f"{parts[0]}/{parts[1]}"
+
+
+def source_url_from_repo_slug(repo_slug: str | None) -> str | None:
+    if not isinstance(repo_slug, str) or not repo_slug:
+        return None
+    return f"https://github.com/{repo_slug}"
+
+
+def source_repo_clone_url(repo: str | None, repo_url: str | None) -> str | None:
+    if isinstance(repo_url, str) and repo_url.strip():
+        return repo_url.strip().removesuffix(".git")
+    return source_url_from_repo_slug(repo_slug_from_source(repo))
+
+
+def infer_source_ref_kind(ref: str | None) -> str:
+    if not isinstance(ref, str):
+        return "tag"
+    normalized = ref.strip()
+    lowered = normalized.lower()
+    if not normalized:
+        return "tag"
+    if lowered.startswith("refs/pull/") or lowered.startswith("pull/"):
+        return "pull"
+    if (
+        lowered.startswith("refs/heads/")
+        or lowered in {"main", "master", "head"}
+        or lowered.startswith("origin/")
+    ):
+        return "branch"
+    normalized_commit = normalize_source_commit(normalized)
+    if normalized_commit is not None:
+        return "commit"
+    return "tag"
+
+
+def normalized_ref_aliases(ref: str | None) -> set[str]:
+    if not isinstance(ref, str):
+        return set()
+    normalized = ref.strip()
+    if not normalized:
+        return set()
+    aliases = {normalized}
+    lowered = normalized.lower()
+    commit = normalize_source_commit(normalized)
+    if commit is not None:
+        aliases.add(commit)
+    if lowered.startswith("refs/heads/"):
+        aliases.add(normalized.split("/", 2)[2])
+    elif "/" not in normalized and infer_source_ref_kind(normalized) == "branch":
+        aliases.add(f"refs/heads/{normalized}")
+    if lowered.startswith("refs/pull/"):
+        aliases.add(normalized.removeprefix("refs/"))
+    elif lowered.startswith("pull/"):
+        aliases.add(f"refs/{normalized}")
+    return aliases
+
+
+def refs_match(candidate_ref: str | None, requested_ref: str | None) -> bool:
+    candidate_aliases = normalized_ref_aliases(candidate_ref)
+    requested_aliases = normalized_ref_aliases(requested_ref)
+    if not candidate_aliases or not requested_aliases:
+        return False
+    if candidate_aliases & requested_aliases:
+        return True
+    candidate_commit = normalize_source_commit(candidate_ref)
+    requested_commit = normalize_source_commit(requested_ref)
+    if candidate_commit and requested_commit:
+        return candidate_commit.startswith(
+            requested_commit
+        ) or requested_commit.startswith(candidate_commit)
+    return False
+
+
+def checkout_friendly_ref(ref_kind: str | None, ref: str | None) -> str | None:
+    """Normalize a source ref to a form that ``git clone --branch`` accepts.
+
+    Fully qualified branch refs like ``refs/heads/main`` are stripped to
+    ``main``; tag refs like ``refs/tags/b8508`` are stripped to ``b8508``.
+    Pull refs like ``refs/pull/123/head`` are left as-is since they are
+    always fetched explicitly rather than cloned with ``--branch``.
+    """
+    if not isinstance(ref, str) or not ref:
+        return ref
+    lowered = ref.lower()
+    if ref_kind == "branch" and lowered.startswith("refs/heads/"):
+        return ref.split("/", 2)[2]
+    if ref_kind == "tag" and lowered.startswith("refs/tags/"):
+        return ref.split("/", 2)[2]
+    return ref
+
+
+def windows_cuda_upstream_asset_names(llama_tag: str, runtime: str) -> list[str]:
+    return [
+        f"llama-{llama_tag}-bin-win-cuda-{runtime}-x64.zip",
+        f"cudart-llama-bin-win-cuda-{runtime}-x64.zip",
+    ]
+
+
 def format_byte_count(num_bytes: float) -> str:
     units = ["B", "KiB", "MiB", "GiB", "TiB"]
     value = float(num_bytes)
@@ -537,13 +721,21 @@ def download_bytes(
 
 
 def fetch_json(url: str) -> Any:
-    data = download_bytes(
-        url,
-        timeout = 30,
-        headers = github_api_headers(url)
-        if is_github_api_url(url)
-        else auth_headers(url),
-    )
+    try:
+        data = download_bytes(
+            url,
+            timeout = 30,
+            headers = github_api_headers(url)
+            if is_github_api_url(url)
+            else auth_headers(url),
+        )
+    except urllib.error.HTTPError as exc:
+        if exc.code == 403 and is_github_api_url(url):
+            hint = ""
+            if not (os.environ.get("GH_TOKEN") or os.environ.get("GITHUB_TOKEN")):
+                hint = "; set GH_TOKEN or GITHUB_TOKEN to avoid GitHub API rate limits"
+            raise RuntimeError(f"GitHub API returned 403 for {url}{hint}") from exc
+        raise
     if not data:
         raise RuntimeError(f"downloaded empty JSON payload from {url}")
     try:
@@ -648,6 +840,14 @@ def upstream_source_archive_urls(tag: str) -> list[str]:
     ]
 
 
+def commit_source_archive_urls(repo: str, source_commit: str) -> list[str]:
+    encoded_commit = urllib.parse.quote(source_commit, safe = "")
+    return [
+        f"https://codeload.github.com/{repo}/tar.gz/{encoded_commit}",
+        f"https://github.com/{repo}/archive/{encoded_commit}.tar.gz",
+    ]
+
+
 def github_release_assets(repo: str, tag: str) -> dict[str, str]:
     payload = fetch_json(
         f"https://api.github.com/repos/{repo}/releases/tags/{urllib.parse.quote(tag, safe = '')}"
@@ -984,13 +1184,35 @@ def parse_published_release_bundle(
 
     # Mixed repos are filtered by an explicit release-side manifest rather than
     # by release tag or asset filename conventions.
-    manifest_payload = fetch_json(manifest_url)
+    manifest_bytes = download_bytes(
+        manifest_url,
+        timeout = 30,
+        headers = auth_headers(manifest_url),
+    )
+    manifest_sha256 = sha256_bytes(manifest_bytes)
+    try:
+        manifest_payload = json.loads(manifest_bytes.decode("utf-8"))
+    except (UnicodeDecodeError, json.JSONDecodeError) as exc:
+        raise RuntimeError(
+            f"published manifest {DEFAULT_PUBLISHED_MANIFEST_ASSET} was not valid JSON"
+        ) from exc
     if not isinstance(manifest_payload, dict):
         raise RuntimeError(
             f"published manifest {DEFAULT_PUBLISHED_MANIFEST_ASSET} was not a JSON object"
         )
+    validate_schema_version(
+        manifest_payload,
+        label = f"published manifest {DEFAULT_PUBLISHED_MANIFEST_ASSET} in {repo}@{release_tag}",
+    )
     component = manifest_payload.get("component")
     upstream_tag = manifest_payload.get("upstream_tag")
+    source_repo = manifest_payload.get("source_repo")
+    source_repo_url = manifest_payload.get("source_repo_url")
+    source_ref_kind = normalize_source_ref_kind(manifest_payload.get("source_ref_kind"))
+    requested_source_ref = manifest_payload.get("requested_source_ref")
+    resolved_source_ref = manifest_payload.get("resolved_source_ref")
+    source_commit = normalize_source_commit(manifest_payload.get("source_commit"))
+    source_commit_short = manifest_payload.get("source_commit_short")
     if component != "llama.cpp":
         return None
     if not isinstance(upstream_tag, str) or not upstream_tag:
@@ -1021,10 +1243,32 @@ def parse_published_release_bundle(
         f"published_release: manifest={DEFAULT_PUBLISHED_MANIFEST_ASSET}",
         f"published_release: upstream_tag={upstream_tag}",
     ]
+    if isinstance(source_repo, str) and source_repo:
+        selection_log.append(f"published_release: source_repo={source_repo}")
+    if source_commit:
+        selection_log.append(f"published_release: source_commit={source_commit}")
     return PublishedReleaseBundle(
         repo = repo,
         release_tag = release_tag,
         upstream_tag = upstream_tag,
+        manifest_sha256 = manifest_sha256,
+        source_repo = source_repo
+        if isinstance(source_repo, str) and source_repo
+        else None,
+        source_repo_url = source_repo_url
+        if isinstance(source_repo_url, str) and source_repo_url
+        else None,
+        source_ref_kind = source_ref_kind,
+        requested_source_ref = requested_source_ref
+        if isinstance(requested_source_ref, str) and requested_source_ref
+        else None,
+        resolved_source_ref = resolved_source_ref
+        if isinstance(resolved_source_ref, str) and resolved_source_ref
+        else None,
+        source_commit = source_commit,
+        source_commit_short = source_commit_short
+        if isinstance(source_commit_short, str) and source_commit_short
+        else None,
         assets = assets,
         manifest_asset_name = DEFAULT_PUBLISHED_MANIFEST_ASSET,
         artifacts = artifacts,
@@ -1041,6 +1285,10 @@ def parse_approved_release_checksums(
         raise RuntimeError(
             f"published checksum asset {DEFAULT_PUBLISHED_SHA256_ASSET} was not a JSON object"
         )
+    validate_schema_version(
+        payload,
+        label = f"published checksum asset {DEFAULT_PUBLISHED_SHA256_ASSET}",
+    )
     if payload.get("component") != "llama.cpp":
         raise RuntimeError(
             f"published checksum asset {DEFAULT_PUBLISHED_SHA256_ASSET} did not describe llama.cpp"
@@ -1090,13 +1338,33 @@ def parse_approved_release_checksums(
             kind = kind_value if isinstance(kind_value, str) and kind_value else None,
         )
 
-    source_commit = payload.get("source_commit")
+    source_commit = normalize_source_commit(payload.get("source_commit"))
+    source_commit_short = payload.get("source_commit_short")
+    source_repo = payload.get("source_repo")
+    source_repo_url = payload.get("source_repo_url")
+    source_ref_kind = normalize_source_ref_kind(payload.get("source_ref_kind"))
+    requested_source_ref = payload.get("requested_source_ref")
+    resolved_source_ref = payload.get("resolved_source_ref")
     return ApprovedReleaseChecksums(
         repo = repo,
         release_tag = release_tag,
         upstream_tag = upstream_tag,
-        source_commit = source_commit
-        if isinstance(source_commit, str) and source_commit
+        source_repo = source_repo
+        if isinstance(source_repo, str) and source_repo
+        else None,
+        source_repo_url = source_repo_url
+        if isinstance(source_repo_url, str) and source_repo_url
+        else None,
+        source_ref_kind = source_ref_kind,
+        requested_source_ref = requested_source_ref
+        if isinstance(requested_source_ref, str) and requested_source_ref
+        else None,
+        resolved_source_ref = resolved_source_ref
+        if isinstance(resolved_source_ref, str) and resolved_source_ref
+        else None,
+        source_commit = source_commit,
+        source_commit_short = source_commit_short
+        if isinstance(source_commit_short, str) and source_commit_short
         else None,
         artifacts = artifacts,
     )
@@ -1390,10 +1658,36 @@ def validated_checksums_for_bundle(
     repo: str, bundle: PublishedReleaseBundle
 ) -> ApprovedReleaseChecksums:
     checksums = load_approved_release_checksums(repo, bundle.release_tag)
-    require_approved_source_hash(checksums, bundle.upstream_tag)
+    manifest_hash = checksums.artifacts.get(bundle.manifest_asset_name)
+    if manifest_hash is not None and bundle.manifest_sha256 is not None:
+        if manifest_hash.sha256 != bundle.manifest_sha256:
+            raise PrebuiltFallback(
+                "published manifest checksum did not match the approved checksum asset"
+            )
+    # Accept bundles that carry only an exact-commit source archive
+    # (e.g. llama.cpp-source-commit-<sha>.tar.gz) without requiring the
+    # legacy llama.cpp-source-<upstream_tag>.tar.gz entry.
+    if exact_source_archive_hash(checksums) is None:
+        require_approved_source_hash(checksums, bundle.upstream_tag)
     return checksums
 
 
+def published_release_matches_request(
+    bundle: PublishedReleaseBundle, requested_ref: str
+) -> bool:
+    if requested_ref == "latest":
+        return True
+    for candidate in (
+        bundle.upstream_tag,
+        bundle.requested_source_ref,
+        bundle.resolved_source_ref,
+        bundle.source_commit,
+    ):
+        if refs_match(candidate, requested_ref):
+            return True
+    return False
+
+
 def resolve_published_release(
     requested_tag: str | None,
     published_repo: str,
@@ -1404,10 +1698,7 @@ def resolve_published_release(
 
     if published_release_tag:
         bundle = pinned_published_release_bundle(repo, published_release_tag)
-        if (
-            normalized_requested != "latest"
-            and bundle.upstream_tag != normalized_requested
-        ):
+        if not published_release_matches_request(bundle, normalized_requested):
             raise PrebuiltFallback(
                 "published release "
                 f"{repo}@{published_release_tag} targeted upstream tag {bundle.upstream_tag}, "
@@ -1420,10 +1711,7 @@ def resolve_published_release(
 
     skipped_invalid = 0
     for bundle in iter_published_release_bundles(repo):
-        if (
-            normalized_requested != "latest"
-            and bundle.upstream_tag != normalized_requested
-        ):
+        if not published_release_matches_request(bundle, normalized_requested):
             continue
         try:
             checksums = validated_checksums_for_bundle(repo, bundle)
@@ -1460,10 +1748,7 @@ def iter_resolved_published_releases(
 
     if published_release_tag:
         bundle = pinned_published_release_bundle(repo, published_release_tag)
-        if (
-            normalized_requested != "latest"
-            and bundle.upstream_tag != normalized_requested
-        ):
+        if not published_release_matches_request(bundle, normalized_requested):
             raise PrebuiltFallback(
                 "published release "
                 f"{repo}@{published_release_tag} targeted upstream tag {bundle.upstream_tag}, "
@@ -1479,10 +1764,7 @@ def iter_resolved_published_releases(
     skipped_invalid = 0
     yielded_valid = False
     for bundle in iter_published_release_bundles(repo):
-        if (
-            normalized_requested != "latest"
-            and bundle.upstream_tag != normalized_requested
-        ):
+        if not published_release_matches_request(bundle, normalized_requested):
             continue
         matched_any = True
         try:
@@ -1520,6 +1802,7 @@ def iter_resolved_published_releases(
 def resolve_requested_llama_tag(
     requested_tag: str | None,
     published_repo: str = "",
+    published_release_tag: str = "",
 ) -> str:
     """Resolve a llama.cpp tag for source-build fallback.
 
@@ -1547,6 +1830,7 @@ def resolve_requested_llama_tag(
             return resolve_published_release(
                 "latest",
                 published_repo,
+                published_release_tag,
             ).bundle.upstream_tag
         except Exception:
             pass
@@ -1566,6 +1850,125 @@ def resolve_requested_install_tag(
     ).bundle.upstream_tag
 
 
+def exact_source_archive_hash(
+    checksums: ApprovedReleaseChecksums,
+) -> ApprovedArtifactHash | None:
+    if not checksums.source_commit:
+        return None
+    return checksums.artifacts.get(
+        exact_source_archive_logical_name(checksums.source_commit)
+    )
+
+
+def source_clone_url_from_checksums(checksums: ApprovedReleaseChecksums) -> str | None:
+    return source_repo_clone_url(checksums.source_repo, checksums.source_repo_url)
+
+
+def source_build_plan_for_release(
+    release: ResolvedPublishedRelease,
+) -> SourceBuildPlan:
+    checksums = release.checksums
+    exact_source = exact_source_archive_hash(checksums)
+    source_repo = checksums.source_repo or release.bundle.source_repo
+    source_repo_url = checksums.source_repo_url or release.bundle.source_repo_url
+    requested_source_ref = (
+        checksums.requested_source_ref or release.bundle.requested_source_ref
+    )
+    resolved_source_ref = (
+        checksums.resolved_source_ref or release.bundle.resolved_source_ref
+    )
+    source_commit = checksums.source_commit or release.bundle.source_commit
+    source_ref_kind = checksums.source_ref_kind or release.bundle.source_ref_kind
+    source_url = source_repo_clone_url(source_repo, source_repo_url)
+    if exact_source is not None and source_url and source_commit:
+        return SourceBuildPlan(
+            source_url = source_url,
+            source_ref = source_commit,
+            source_ref_kind = "commit",
+            compatibility_upstream_tag = release.bundle.upstream_tag,
+            source_repo = source_repo,
+            source_repo_url = source_repo_url,
+            requested_source_ref = requested_source_ref,
+            resolved_source_ref = resolved_source_ref,
+            source_commit = source_commit,
+        )
+    source_ref = checkout_friendly_ref(
+        source_ref_kind, resolved_source_ref or requested_source_ref
+    )
+    if (
+        source_url
+        and source_ref
+        and source_ref_kind in {"tag", "branch", "pull", "commit"}
+    ):
+        return SourceBuildPlan(
+            source_url = source_url,
+            source_ref = source_ref,
+            source_ref_kind = source_ref_kind,
+            compatibility_upstream_tag = release.bundle.upstream_tag,
+            source_repo = source_repo,
+            source_repo_url = source_repo_url,
+            requested_source_ref = requested_source_ref,
+            resolved_source_ref = resolved_source_ref,
+            source_commit = source_commit,
+        )
+    return SourceBuildPlan(
+        source_url = source_url_from_repo_slug(UPSTREAM_REPO)
+        or "https://github.com/ggml-org/llama.cpp",
+        source_ref = release.bundle.upstream_tag,
+        source_ref_kind = "tag",
+        compatibility_upstream_tag = release.bundle.upstream_tag,
+        source_repo = source_repo,
+        source_repo_url = source_repo_url,
+        requested_source_ref = requested_source_ref,
+        resolved_source_ref = resolved_source_ref,
+        source_commit = source_commit,
+    )
+
+
+def resolve_source_build_plan(
+    requested_tag: str | None,
+    published_repo: str,
+    published_release_tag: str = "",
+) -> SourceBuildPlan:
+    normalized_requested = normalized_requested_llama_tag(requested_tag)
+    if normalized_requested != "latest":
+        try:
+            release = resolve_published_release(
+                normalized_requested,
+                published_repo,
+                published_release_tag,
+            )
+            return source_build_plan_for_release(release)
+        except Exception:
+            pass
+        inferred_kind = infer_source_ref_kind(normalized_requested)
+        return SourceBuildPlan(
+            source_url = "https://github.com/ggml-org/llama.cpp",
+            source_ref = checkout_friendly_ref(inferred_kind, normalized_requested)
+            or normalized_requested,
+            source_ref_kind = inferred_kind,
+            compatibility_upstream_tag = normalized_requested,
+        )
+
+    if published_repo:
+        try:
+            release = resolve_published_release(
+                "latest",
+                published_repo,
+                published_release_tag,
+            )
+            return source_build_plan_for_release(release)
+        except Exception:
+            pass
+    latest_tag = latest_upstream_release_tag()
+    return SourceBuildPlan(
+        source_url = "https://github.com/ggml-org/llama.cpp",
+        source_ref = latest_tag,
+        source_ref_kind = "tag",
+        compatibility_upstream_tag = latest_tag,
+    )
+
+
 def run_capture(
     command: list[str],
     *,
@@ -1883,25 +2286,31 @@ def windows_cuda_attempts(
     attempts: list[AssetChoice] = []
     for runtime_line in runtime_order:
         runtime = runtime_by_line[runtime_line]
-        upstream_name = f"llama-{llama_tag}-bin-win-cuda-{runtime}-x64.zip"
-        asset_url = upstream_assets.get(upstream_name)
-        if not asset_url:
+        selected_name = None
+        asset_url = None
+        for candidate_name in windows_cuda_upstream_asset_names(llama_tag, runtime):
+            asset_url = upstream_assets.get(candidate_name)
+            if asset_url:
+                selected_name = candidate_name
+                break
+        if not asset_url or not selected_name:
             selection_log.append(
-                f"windows_cuda_selection: skip missing asset {upstream_name}"
+                "windows_cuda_selection: skip missing assets "
+                + ",".join(windows_cuda_upstream_asset_names(llama_tag, runtime))
             )
             continue
         attempts.append(
             AssetChoice(
                 repo = UPSTREAM_REPO,
                 tag = llama_tag,
-                name = upstream_name,
+                name = selected_name,
                 url = asset_url,
                 source_label = "upstream",
                 install_kind = "windows-cuda",
                 runtime_line = runtime_line,
                 selection_log = list(selection_log)
                 + [
-                    f"windows_cuda_selection: selected {upstream_name} runtime={runtime}"
+                    f"windows_cuda_selection: selected {selected_name} runtime={runtime}"
                 ],
             )
         )
@@ -2353,18 +2762,26 @@ def copy_directory_contents(source_dir: Path, destination: Path) -> None:
 
 
 def hydrate_source_tree(
-    upstream_tag: str,
+    source_ref: str,
     install_dir: Path,
     work_dir: Path,
     *,
+    source_repo: str = UPSTREAM_REPO,
     expected_sha256: str,
+    source_label: str | None = None,
+    exact_source: bool = False,
 ) -> None:
-    archive_path = work_dir / f"llama.cpp-source-{upstream_tag}.tar.gz"
-    source_urls = upstream_source_archive_urls(upstream_tag)
+    archive_path = work_dir / f"llama.cpp-source-{source_ref}.tar.gz"
+    source_urls = (
+        commit_source_archive_urls(source_repo, source_ref)
+        if exact_source
+        else upstream_source_archive_urls(source_ref)
+    )
+    label = source_label or f"llama.cpp source tree for {source_ref}"
     extract_dir = Path(tempfile.mkdtemp(prefix = "source-extract-", dir = work_dir))
 
     try:
-        log(f"downloading llama.cpp source tree for upstream tag {upstream_tag}")
+        log(f"downloading {label}")
         last_exc: Exception | None = None
         downloaded = False
         for index, source_url in enumerate(source_urls):
@@ -2377,7 +2794,7 @@ def hydrate_source_tree(
                     source_url,
                     archive_path,
                     expected_sha256 = expected_sha256,
-                    label = f"llama.cpp source tree for {upstream_tag}",
+                    label = label,
                 )
                 downloaded = True
                 break
@@ -2410,9 +2827,7 @@ def hydrate_source_tree(
     except PrebuiltFallback:
         raise
     except Exception as exc:
-        raise PrebuiltFallback(
-            f"failed to hydrate upstream llama.cpp source tree for {upstream_tag}: {exc}"
-        ) from exc
+        raise PrebuiltFallback(f"failed to hydrate {label}: {exc}") from exc
     finally:
         remove_tree(extract_dir)
 
@@ -3451,10 +3866,32 @@ def apply_approved_hashes(
     attempts: Iterable[AssetChoice],
     checksums: ApprovedReleaseChecksums,
 ) -> list[AssetChoice]:
+    def approved_hash_for_attempt(attempt: AssetChoice) -> ApprovedArtifactHash | None:
+        approved = checksums.artifacts.get(attempt.name)
+        if approved is not None:
+            return approved
+        if (
+            isinstance(attempt.tag, str)
+            and attempt.tag
+            and attempt.tag != checksums.upstream_tag
+            and attempt.name.startswith("llama-")
+        ):
+            legacy_prefix = f"llama-{attempt.tag}-"
+            compatibility_prefix = f"llama-{checksums.upstream_tag}-"
+            compatibility_name = (
+                attempt.name.replace(legacy_prefix, compatibility_prefix, 1)
+                if attempt.name.startswith(legacy_prefix)
+                else attempt.name
+            )
+            approved = checksums.artifacts.get(compatibility_name)
+            if approved is not None:
+                return approved
+        return None
+
     approved_attempts: list[AssetChoice] = []
     missing_assets: list[str] = []
     for attempt in attempts:
-        approved = checksums.artifacts.get(attempt.name)
+        approved = approved_hash_for_attempt(attempt)
         if approved is None:
             missing_assets.append(attempt.name)
             continue
@@ -3481,6 +3918,39 @@ def require_approved_source_hash(
     return approved_source
 
 
+def preferred_source_archive(
+    checksums: ApprovedReleaseChecksums, llama_tag: str
+) -> tuple[str, str, ApprovedArtifactHash, bool]:
+    exact_source = exact_source_archive_hash(checksums)
+    exact_repo = repo_slug_from_source(checksums.source_repo) or repo_slug_from_source(
+        checksums.source_repo_url
+    )
+    if exact_source is not None and exact_repo and checksums.source_commit:
+        return (
+            exact_repo,
+            checksums.source_commit,
+            exact_source,
+            True,
+        )
+    legacy = require_approved_source_hash(checksums, llama_tag)
+    return (
+        UPSTREAM_REPO,
+        llama_tag,
+        legacy,
+        False,
+    )
+
+
+def selected_source_archive_metadata(
+    checksums: ApprovedReleaseChecksums,
+    llama_tag: str,
+) -> tuple[str, str | None]:
+    _source_repo, _source_ref, source_archive, _exact_source = preferred_source_archive(
+        checksums, llama_tag
+    )
+    return source_archive.asset_name, source_archive.sha256
+
+
 def resolve_install_attempts(
     llama_tag: str,
     host: HostInfo,
@@ -3583,10 +4053,10 @@ def write_prebuilt_metadata(
     approved_checksums: ApprovedReleaseChecksums,
     prebuilt_fallback_used: bool,
 ) -> None:
-    source_archive = approved_checksums.artifacts.get(
-        source_archive_logical_name(llama_tag)
+    source_asset_name, source_sha256 = selected_source_archive_metadata(
+        approved_checksums,
+        llama_tag,
     )
-    source_sha256 = source_archive.sha256 if source_archive is not None else None
     fingerprint_payload = {
         "published_repo": approved_checksums.repo,
         "release_tag": release_tag,
@@ -3594,6 +4064,7 @@ def write_prebuilt_metadata(
         "asset": choice.name,
         "asset_sha256": choice.expected_sha256,
         "source": choice.source_label,
+        "source_asset": source_asset_name,
         "source_sha256": source_sha256,
         "runtime_line": choice.runtime_line,
         "bundle_profile": choice.bundle_profile,
@@ -3612,8 +4083,15 @@ def write_prebuilt_metadata(
         "asset": choice.name,
         "asset_sha256": choice.expected_sha256,
         "source": choice.source_label,
+        "source_asset": source_asset_name,
         "source_sha256": source_sha256,
         "source_commit": approved_checksums.source_commit,
+        "source_commit_short": approved_checksums.source_commit_short,
+        "source_repo": approved_checksums.source_repo,
+        "source_repo_url": approved_checksums.source_repo_url,
+        "source_ref_kind": approved_checksums.source_ref_kind,
+        "requested_source_ref": approved_checksums.requested_source_ref,
+        "resolved_source_ref": approved_checksums.resolved_source_ref,
         "bundle_profile": choice.bundle_profile,
         "runtime_line": choice.runtime_line,
         "coverage_class": choice.coverage_class,
@@ -3635,10 +4113,10 @@ def expected_install_fingerprint(
 ) -> str | None:
     if not choice.expected_sha256:
         return None
-    source_archive = approved_checksums.artifacts.get(
-        source_archive_logical_name(llama_tag)
+    source_asset_name, source_sha256 = selected_source_archive_metadata(
+        approved_checksums,
+        llama_tag,
     )
-    source_sha256 = source_archive.sha256 if source_archive is not None else None
     payload = {
         "published_repo": approved_checksums.repo,
         "release_tag": release_tag,
@@ -3646,6 +4124,7 @@ def expected_install_fingerprint(
         "asset": choice.name,
         "asset_sha256": choice.expected_sha256,
         "source": choice.source_label,
+        "source_asset": source_asset_name,
         "source_sha256": source_sha256,
         "runtime_line": choice.runtime_line,
         "bundle_profile": choice.bundle_profile,
@@ -3817,19 +4296,27 @@ def validate_prebuilt_choice(
     prebuilt_fallback_used: bool,
     quantized_path: Path,
 ) -> tuple[Path, Path]:
-    source_archive = approved_checksums.artifacts.get(
-        source_archive_logical_name(llama_tag)
+    source_repo, source_ref, source_archive, exact_source = preferred_source_archive(
+        approved_checksums, llama_tag
     )
-    if source_archive is None:
-        raise PrebuiltFallback(
-            f"approved checksum asset did not contain source archive {source_archive_logical_name(llama_tag)}"
+    if exact_source:
+        log(
+            f"hydrating exact llama.cpp source for {source_repo}@{source_ref} into {install_dir}"
         )
-    log(f"hydrating upstream llama.cpp source for {llama_tag} into {install_dir}")
+    else:
+        log(f"hydrating upstream llama.cpp source for {llama_tag} into {install_dir}")
     hydrate_source_tree(
-        llama_tag,
+        source_ref,
         install_dir,
         work_dir,
+        source_repo = source_repo,
         expected_sha256 = source_archive.sha256,
+        source_label = (
+            f"llama.cpp source tree for {source_repo}@{source_ref}"
+            if exact_source
+            else f"llama.cpp source tree for {llama_tag}"
+        ),
+        exact_source = exact_source,
     )
     log(f"overlaying prebuilt bundle {choice.name} into {install_dir}")
     server_path, quantize_path = install_from_archives(
@@ -4087,30 +4574,103 @@ def parse_args() -> argparse.Namespace:
             "selected by the current published-release policy."
         ),
     )
+    resolve_group.add_argument(
+        "--resolve-source-build",
+        nargs = "?",
+        const = "latest",
+        help = ("Resolve the source-build fallback plan."),
+    )
+    parser.add_argument(
+        "--output-format",
+        choices = ("plain", "json"),
+        default = "plain",
+        help = "Resolver output format. Defaults to plain.",
+    )
     return parser.parse_args()
 
 
+def emit_resolver_output(payload: dict[str, Any], *, output_format: str) -> None:
+    if output_format == "json":
+        print(json.dumps(payload, sort_keys = True))
+        return
+    if "llama_tag" in payload:
+        print(payload["llama_tag"])
+        return
+    if {
+        "source_url",
+        "source_ref_kind",
+        "source_ref",
+    }.issubset(payload):
+        print(
+            "\t".join(
+                (
+                    str(payload["source_url"]),
+                    str(payload["source_ref_kind"]),
+                    str(payload["source_ref"]),
+                )
+            )
+        )
+        return
+    print(json.dumps(payload, sort_keys = True))
+
+
 def main() -> int:
     args = parse_args()
     if args.resolve_llama_tag is not None:
-        # Pass published_repo so the resolver prefers the Unsloth release tag
-        # (tested/approved) over the upstream ggml-org bleeding-edge tag.
-        print(resolve_requested_llama_tag(args.resolve_llama_tag, args.published_repo))
+        resolved = resolve_requested_llama_tag(
+            args.resolve_llama_tag,
+            args.published_repo,
+            args.published_release_tag or "",
+        )
+        emit_resolver_output(
+            {
+                "requested_tag": normalized_requested_llama_tag(args.resolve_llama_tag),
+                "llama_tag": resolved,
+            },
+            output_format = args.output_format,
+        )
         return EXIT_SUCCESS
 
     if args.resolve_install_tag is not None:
-        print(
-            resolve_requested_install_tag(
-                args.resolve_install_tag,
-                args.published_release_tag or "",
-                args.published_repo,
-            )
+        resolved = resolve_requested_install_tag(
+            args.resolve_install_tag,
+            args.published_release_tag or "",
+            args.published_repo,
+        )
+        emit_resolver_output(
+            {
+                "requested_tag": normalized_requested_llama_tag(
+                    args.resolve_install_tag
+                ),
+                "llama_tag": resolved,
+            },
+            output_format = args.output_format,
+        )
+        return EXIT_SUCCESS
+
+    if args.resolve_source_build is not None:
+        plan = resolve_source_build_plan(
+            args.resolve_source_build,
+            args.published_repo,
+            args.published_release_tag or "",
+        )
+        emit_resolver_output(
+            {
+                "requested_tag": normalized_requested_llama_tag(
+                    args.resolve_source_build
+                ),
+                "source_url": plan.source_url,
+                "source_ref_kind": plan.source_ref_kind,
+                "source_ref": plan.source_ref,
+                "compatibility_upstream_tag": plan.compatibility_upstream_tag,
+            },
+            output_format = args.output_format,
         )
         return EXIT_SUCCESS
 
     if not args.install_dir:
         raise SystemExit(
-            "install_llama_prebuilt.py: --install-dir is required unless --resolve-llama-tag or --resolve-install-tag is used"
+            "install_llama_prebuilt.py: --install-dir is required unless --resolve-llama-tag, --resolve-install-tag, or --resolve-source-build is used"
         )
     install_prebuilt(
         install_dir = Path(args.install_dir).expanduser().resolve(),
diff --git a/studio/setup.ps1 b/studio/setup.ps1
index f4ad42d615..1784b66ee9 100644
--- a/studio/setup.ps1
+++ b/studio/setup.ps1
@@ -1604,6 +1604,9 @@ $LlamaPr = if ($env:UNSLOTH_LLAMA_PR) { $env:UNSLOTH_LLAMA_PR.Trim() } else { ""
 $LlamaPrForce = if ($env:UNSLOTH_LLAMA_PR_FORCE) { $env:UNSLOTH_LLAMA_PR_FORCE.Trim() } else { $DefaultLlamaPrForce }
 $LlamaSource = if ($env:UNSLOTH_LLAMA_SOURCE) { $env:UNSLOTH_LLAMA_SOURCE.Trim() } else { $DefaultLlamaSource }
 if ($LlamaSource.EndsWith('.git')) { $LlamaSource = $LlamaSource.Substring(0, $LlamaSource.Length - 4) }
+$ResolvedSourceUrl = $LlamaSource
+$ResolvedSourceRef = $RequestedLlamaTag
+$ResolvedSourceRefKind = "tag"
 
 if ($LlamaSource -ne "https://github.com/ggml-org/llama.cpp") {
     step "llama.cpp" "custom source: $LlamaSource -- forcing source build" "Yellow"
@@ -1623,6 +1626,9 @@ if ($LlamaPr) {
     }
     step "llama.cpp" "UNSLOTH_LLAMA_PR=$LlamaPr -- will build from PR head" "Yellow"
     $ResolvedLlamaTag = "pr-$LlamaPr"
+    $ResolvedSourceUrl = $LlamaSource
+    $ResolvedSourceRef = "pr-$LlamaPr"
+    $ResolvedSourceRefKind = "pull"
     $NeedLlamaSourceBuild = $true
     $SkipPrebuiltInstall = $true
 } elseif ($SkipPrebuiltInstall) {
@@ -1630,12 +1636,16 @@ if ($LlamaPr) {
     # prebuilt release resolution. When building from a custom fork, the fork
     # may not carry upstream bNNNN tags.
     if ($LlamaSource -eq "https://github.com/ggml-org/llama.cpp") {
-        $resolveTagArgs = @("--resolve-llama-tag", $RequestedLlamaTag, "--published-repo", $HelperReleaseRepo)
+        $resolveTagArgs = @("--resolve-llama-tag", $RequestedLlamaTag, "--published-repo", $HelperReleaseRepo, "--output-format", "json")
         if ($env:UNSLOTH_LLAMA_RELEASE_TAG) { $resolveTagArgs += @("--published-release-tag", $env:UNSLOTH_LLAMA_RELEASE_TAG) }
         $fallbackOutput = & python "$PSScriptRoot\install_llama_prebuilt.py" @resolveTagArgs 2>$null
         $fallbackExit = $LASTEXITCODE
         $ResolvedLlamaTag = if ($fallbackExit -eq 0 -and $fallbackOutput) {
-            ($fallbackOutput | Select-Object -Last 1).ToString().Trim()
+            try {
+                (($fallbackOutput | Out-String) | ConvertFrom-Json).llama_tag
+            } catch {
+                $RequestedLlamaTag
+            }
         } else {
             $RequestedLlamaTag
         }
@@ -1643,30 +1653,42 @@ if ($LlamaPr) {
         $ResolvedLlamaTag = $RequestedLlamaTag
     }
 } else {
-    $resolveInstallArgs = @("--resolve-install-tag", $RequestedLlamaTag, "--published-repo", $HelperReleaseRepo)
+    $resolveInstallArgs = @("--resolve-install-tag", $RequestedLlamaTag, "--published-repo", $HelperReleaseRepo, "--output-format", "json")
     if ($env:UNSLOTH_LLAMA_RELEASE_TAG) { $resolveInstallArgs += @("--published-release-tag", $env:UNSLOTH_LLAMA_RELEASE_TAG) }
-    $resolveOutput = & python "$PSScriptRoot\install_llama_prebuilt.py" @resolveInstallArgs 2>&1
+    $resolveErrorLog = New-TemporaryFile
+    $resolveOutput = & python "$PSScriptRoot\install_llama_prebuilt.py" @resolveInstallArgs 2>$resolveErrorLog
     $resolveExit = $LASTEXITCODE
-    $ResolvedLlamaTag = if ($resolveOutput) { ($resolveOutput | Select-Object -Last 1).ToString().Trim() } else { "" }
+    $ResolvedLlamaTag = if ($resolveOutput) {
+        try {
+            (($resolveOutput | Out-String) | ConvertFrom-Json).llama_tag
+        } catch {
+            ""
+        }
+    } else { "" }
     if ($resolveExit -ne 0 -or [string]::IsNullOrWhiteSpace($ResolvedLlamaTag)) {
         Write-Host ""
         substep "Failed to resolve a published llama.cpp release via $HelperReleaseRepo" "Yellow"
-        Write-LlamaFailureLog -Output ($resolveOutput | Out-String)
+        Write-LlamaFailureLog -Output (Get-Content -Raw $resolveErrorLog)
         # Resolve the llama.cpp tag for source-build fallback. Pass --published-repo
         # so the resolver prefers the latest usable Unsloth-published upstream tag
         # before falling back to the bleeding-edge ggml-org/llama.cpp tag.
-        $resolveFallbackArgs = @("--resolve-llama-tag", $RequestedLlamaTag, "--published-repo", $HelperReleaseRepo)
+        $resolveFallbackArgs = @("--resolve-llama-tag", $RequestedLlamaTag, "--published-repo", $HelperReleaseRepo, "--output-format", "json")
         if ($env:UNSLOTH_LLAMA_RELEASE_TAG) { $resolveFallbackArgs += @("--published-release-tag", $env:UNSLOTH_LLAMA_RELEASE_TAG) }
         $fallbackOutput = & python "$PSScriptRoot\install_llama_prebuilt.py" @resolveFallbackArgs 2>$null
         $fallbackExit = $LASTEXITCODE
         $ResolvedLlamaTag = if ($fallbackExit -eq 0 -and $fallbackOutput) {
-            ($fallbackOutput | Select-Object -Last 1).ToString().Trim()
+            try {
+                (($fallbackOutput | Out-String) | ConvertFrom-Json).llama_tag
+            } catch {
+                $RequestedLlamaTag
+            }
         } else {
             $RequestedLlamaTag
         }
         $NeedLlamaSourceBuild = $true
         $SkipPrebuiltInstall = $true
     }
+    Remove-Item $resolveErrorLog -Force -ErrorAction SilentlyContinue
 }
 
 Write-Host ""
@@ -1875,14 +1897,34 @@ if (-not $NeedLlamaSourceBuild) {
         [Environment]::SetEnvironmentVariable('CudaToolkitDir', "$CudaToolkitRoot\", 'Process')
     }
 
+    if (-not $LlamaPr) {
+        if ($LlamaSource -eq "https://github.com/ggml-org/llama.cpp") {
+            $resolveSourceArgs = @("--resolve-source-build", $RequestedLlamaTag, "--published-repo", $HelperReleaseRepo, "--output-format", "json")
+            if ($env:UNSLOTH_LLAMA_RELEASE_TAG) { $resolveSourceArgs += @("--published-release-tag", $env:UNSLOTH_LLAMA_RELEASE_TAG) }
+            $sourcePlanOutput = & python "$PSScriptRoot\install_llama_prebuilt.py" @resolveSourceArgs 2>$null
+            $sourcePlanExit = $LASTEXITCODE
+            if ($sourcePlanExit -eq 0 -and $sourcePlanOutput) {
+                try {
+                    $sourcePlan = ($sourcePlanOutput | Out-String) | ConvertFrom-Json
+                    $ResolvedSourceUrl = $sourcePlan.source_url
+                    $ResolvedSourceRefKind = $sourcePlan.source_ref_kind
+                    $ResolvedSourceRef = $sourcePlan.source_ref
+                } catch {
+                }
+            }
+        }
+        if ([string]::IsNullOrWhiteSpace($ResolvedSourceUrl)) { $ResolvedSourceUrl = $LlamaSource }
+        if ([string]::IsNullOrWhiteSpace($ResolvedSourceRef)) { $ResolvedSourceRef = $ResolvedLlamaTag }
+    }
+
     # -- Step A: Clone or pull llama.cpp --
 
-    $UseConcreteRef = ($ResolvedLlamaTag -ne "latest" -and -not [string]::IsNullOrWhiteSpace($ResolvedLlamaTag))
+    $UseConcreteRef = ($ResolvedSourceRef -ne "latest" -and -not [string]::IsNullOrWhiteSpace($ResolvedSourceRef))
 
     if (Test-Path (Join-Path $LlamaCppDir ".git")) {
-        Write-Host "   Syncing llama.cpp to $ResolvedLlamaTag..." -ForegroundColor Gray
+        Write-Host "   Syncing llama.cpp to $ResolvedSourceRef..." -ForegroundColor Gray
         # Always sync the remote URL so switching between default/fork sources works
-        Invoke-SetupCommand -AlwaysQuiet { git -C $LlamaCppDir remote set-url origin "$LlamaSource.git" } | Out-Null
+        Invoke-SetupCommand -AlwaysQuiet { git -C $LlamaCppDir remote set-url origin "$ResolvedSourceUrl.git" } | Out-Null
         if ($LlamaPr) {
             $gitFetchExit = Invoke-SetupCommand -AlwaysQuiet { git -C $LlamaCppDir fetch --depth 1 origin "pull/$LlamaPr/head" }
             if ($gitFetchExit -ne 0) {
@@ -1897,8 +1939,34 @@ if (-not $NeedLlamaSourceBuild) {
                     Invoke-SetupCommand -AlwaysQuiet { git -C $LlamaCppDir clean -fdx } | Out-Null
                 }
             }
+        } elseif ($ResolvedSourceRefKind -eq "pull") {
+            $gitFetchExit = Invoke-SetupCommand -AlwaysQuiet { git -C $LlamaCppDir fetch --depth 1 origin $ResolvedSourceRef }
+            if ($gitFetchExit -ne 0) {
+                substep "git fetch failed -- using existing source" "Yellow"
+            } else {
+                $gitCheckoutExit = Invoke-SetupCommand -AlwaysQuiet { git -C $LlamaCppDir checkout -B unsloth-llama-build FETCH_HEAD }
+                if ($gitCheckoutExit -ne 0) {
+                    $BuildOk = $false
+                    $FailedStep = "git checkout"
+                } else {
+                    Invoke-SetupCommand -AlwaysQuiet { git -C $LlamaCppDir clean -fdx } | Out-Null
+                }
+            }
+        } elseif ($ResolvedSourceRefKind -eq "commit") {
+            $gitFetchExit = Invoke-SetupCommand -AlwaysQuiet { git -C $LlamaCppDir fetch --depth 1 origin $ResolvedSourceRef }
+            if ($gitFetchExit -ne 0) {
+                substep "git fetch failed -- using existing source" "Yellow"
+            } else {
+                $gitCheckoutExit = Invoke-SetupCommand -AlwaysQuiet { git -C $LlamaCppDir checkout -B unsloth-llama-build FETCH_HEAD }
+                if ($gitCheckoutExit -ne 0) {
+                    $BuildOk = $false
+                    $FailedStep = "git checkout"
+                } else {
+                    Invoke-SetupCommand -AlwaysQuiet { git -C $LlamaCppDir clean -fdx } | Out-Null
+                }
+            }
         } elseif ($UseConcreteRef) {
-            $gitFetchExit = Invoke-SetupCommand -AlwaysQuiet { git -C $LlamaCppDir fetch --depth 1 origin $ResolvedLlamaTag }
+            $gitFetchExit = Invoke-SetupCommand -AlwaysQuiet { git -C $LlamaCppDir fetch --depth 1 origin $ResolvedSourceRef }
             if ($gitFetchExit -ne 0) {
                 substep "git fetch failed -- using existing source" "Yellow"
             } else {
@@ -1925,8 +1993,9 @@ if (-not $NeedLlamaSourceBuild) {
             }
         }
     } else {
-        Write-Host "   Cloning llama.cpp @ $ResolvedLlamaTag..." -ForegroundColor Gray
+        Write-Host "   Cloning llama.cpp @ $ResolvedSourceRef..." -ForegroundColor Gray
         $buildTmp = "$LlamaCppDir.build.$PID"
+        $null = New-Item -ItemType Directory -Force -Path (Split-Path $LlamaCppDir -Parent)
         if (Test-Path $buildTmp) { Remove-Item -Recurse -Force $buildTmp }
         if ($LlamaPr) {
             $cloneExit = Invoke-SetupCommand -AlwaysQuiet { git clone --depth 1 "$LlamaSource.git" $buildTmp }
@@ -1951,12 +2020,58 @@ if (-not $NeedLlamaSourceBuild) {
                     if (Test-Path $buildTmp) { Remove-Item -Recurse -Force $buildTmp }
                 }
             }
+        } elseif ($ResolvedSourceRefKind -eq "pull") {
+            $cloneExit = Invoke-SetupCommand -AlwaysQuiet { git clone --depth 1 "$ResolvedSourceUrl.git" $buildTmp }
+            if ($cloneExit -ne 0) {
+                $BuildOk = $false
+                $FailedStep = "git clone"
+                if (Test-Path $buildTmp) { Remove-Item -Recurse -Force $buildTmp }
+            }
+            if ($BuildOk) {
+                $fetchExit = Invoke-SetupCommand -AlwaysQuiet { git -C $buildTmp fetch --depth 1 origin $ResolvedSourceRef }
+                if ($fetchExit -ne 0) {
+                    $BuildOk = $false
+                    $FailedStep = "git fetch source PR ref"
+                    if (Test-Path $buildTmp) { Remove-Item -Recurse -Force $buildTmp }
+                }
+            }
+            if ($BuildOk) {
+                $checkoutExit = Invoke-SetupCommand -AlwaysQuiet { git -C $buildTmp checkout -B unsloth-llama-build FETCH_HEAD }
+                if ($checkoutExit -ne 0) {
+                    $BuildOk = $false
+                    $FailedStep = "git checkout source PR ref"
+                    if (Test-Path $buildTmp) { Remove-Item -Recurse -Force $buildTmp }
+                }
+            }
+        } elseif ($ResolvedSourceRefKind -eq "commit") {
+            $cloneExit = Invoke-SetupCommand -AlwaysQuiet { git clone --depth 1 "$ResolvedSourceUrl.git" $buildTmp }
+            if ($cloneExit -ne 0) {
+                $BuildOk = $false
+                $FailedStep = "git clone"
+                if (Test-Path $buildTmp) { Remove-Item -Recurse -Force $buildTmp }
+            }
+            if ($BuildOk) {
+                $fetchExit = Invoke-SetupCommand -AlwaysQuiet { git -C $buildTmp fetch --depth 1 origin $ResolvedSourceRef }
+                if ($fetchExit -ne 0) {
+                    $BuildOk = $false
+                    $FailedStep = "git fetch source commit"
+                    if (Test-Path $buildTmp) { Remove-Item -Recurse -Force $buildTmp }
+                }
+            }
+            if ($BuildOk) {
+                $checkoutExit = Invoke-SetupCommand -AlwaysQuiet { git -C $buildTmp checkout -B unsloth-llama-build FETCH_HEAD }
+                if ($checkoutExit -ne 0) {
+                    $BuildOk = $false
+                    $FailedStep = "git checkout source commit"
+                    if (Test-Path $buildTmp) { Remove-Item -Recurse -Force $buildTmp }
+                }
+            }
         } else {
             $cloneArgs = @("clone", "--depth", "1")
             if ($UseConcreteRef) {
-                $cloneArgs += @("--branch", $ResolvedLlamaTag)
+                $cloneArgs += @("--branch", $ResolvedSourceRef)
             }
-            $cloneArgs += @("$LlamaSource.git", $buildTmp)
+            $cloneArgs += @("$ResolvedSourceUrl.git", $buildTmp)
             $cloneExit = Invoke-SetupCommand -AlwaysQuiet { git @cloneArgs }
             if ($cloneExit -ne 0) {
                 $BuildOk = $false
diff --git a/studio/setup.sh b/studio/setup.sh
index 36926733e1..26903b120a 100755
--- a/studio/setup.sh
+++ b/studio/setup.sh
@@ -485,6 +485,9 @@ _LLAMA_PR="${UNSLOTH_LLAMA_PR:-}"
 _LLAMA_PR_FORCE="${UNSLOTH_LLAMA_PR_FORCE:-${_DEFAULT_LLAMA_PR_FORCE}}"
 _LLAMA_SOURCE="${UNSLOTH_LLAMA_SOURCE:-${_DEFAULT_LLAMA_SOURCE}}"
 _LLAMA_SOURCE="${_LLAMA_SOURCE%.git}"  # normalize: strip trailing .git
+_RESOLVED_SOURCE_URL="$_LLAMA_SOURCE"
+_RESOLVED_SOURCE_REF="$_REQUESTED_LLAMA_TAG"
+_RESOLVED_SOURCE_REF_KIND="tag"
 
 # Non-default source URL forces source build (fork has different code than prebuilt).
 if [ "$_LLAMA_SOURCE" != "https://github.com/ggml-org/llama.cpp" ]; then
@@ -507,6 +510,9 @@ if [ -n "$_LLAMA_PR" ]; then
     fi
     step "llama.cpp" "UNSLOTH_LLAMA_PR=$_LLAMA_PR -- will build from PR head" "$C_WARN"
     _RESOLVED_LLAMA_TAG="pr-$_LLAMA_PR"
+    _RESOLVED_SOURCE_URL="$_LLAMA_SOURCE"
+    _RESOLVED_SOURCE_REF="pr-$_LLAMA_PR"
+    _RESOLVED_SOURCE_REF_KIND="pull"
     _NEED_LLAMA_SOURCE_BUILD=true
     _SKIP_PREBUILT_INSTALL=true
 elif [ "${_SKIP_PREBUILT_INSTALL:-false}" = true ]; then
@@ -516,14 +522,22 @@ elif [ "${_SKIP_PREBUILT_INSTALL:-false}" = true ]; then
     # only when the source is the default ggml-org repo.
     if [ "$_LLAMA_SOURCE" = "https://github.com/ggml-org/llama.cpp" ]; then
         _RESOLVE_TAG_ARGS=(--resolve-llama-tag "$_REQUESTED_LLAMA_TAG" --published-repo "$_HELPER_RELEASE_REPO")
+        _RESOLVE_TAG_ARGS+=(--output-format json)
         if [ -n "${UNSLOTH_LLAMA_RELEASE_TAG:-}" ]; then
             _RESOLVE_TAG_ARGS+=(--published-release-tag "$UNSLOTH_LLAMA_RELEASE_TAG")
         fi
         set +e
-        _RESOLVED_LLAMA_TAG="$(python "$SCRIPT_DIR/install_llama_prebuilt.py" "${_RESOLVE_TAG_ARGS[@]}" 2>/dev/null)"
+        _RESOLVE_TAG_JSON="$(python "$SCRIPT_DIR/install_llama_prebuilt.py" "${_RESOLVE_TAG_ARGS[@]}" 2>/dev/null)"
         _RESOLVE_UPSTREAM_STATUS=$?
         set -e
-        if [ "$_RESOLVE_UPSTREAM_STATUS" -ne 0 ] || [ -z "$_RESOLVED_LLAMA_TAG" ]; then
+        if [ "$_RESOLVE_UPSTREAM_STATUS" -eq 0 ] && [ -n "${_RESOLVE_TAG_JSON:-}" ]; then
+            _RESOLVED_LLAMA_TAG="$(
+                printf '%s' "$_RESOLVE_TAG_JSON" | python -c 'import json,sys; print(json.load(sys.stdin).get("llama_tag",""))' 2>/dev/null || true
+            )"
+        else
+            _RESOLVED_LLAMA_TAG=""
+        fi
+        if [ -z "$_RESOLVED_LLAMA_TAG" ]; then
             _RESOLVED_LLAMA_TAG="$_REQUESTED_LLAMA_TAG"
         fi
     else
@@ -531,17 +545,22 @@ elif [ "${_SKIP_PREBUILT_INSTALL:-false}" = true ]; then
     fi
 else
     _RESOLVE_INSTALL_ARGS=(--resolve-install-tag "$_REQUESTED_LLAMA_TAG" --published-repo "$_HELPER_RELEASE_REPO")
+    _RESOLVE_INSTALL_ARGS+=(--output-format json)
     if [ -n "${UNSLOTH_LLAMA_RELEASE_TAG:-}" ]; then
         _RESOLVE_INSTALL_ARGS+=(--published-release-tag "$UNSLOTH_LLAMA_RELEASE_TAG")
     fi
     _RESOLVE_LLAMA_LOG="$(mktemp)"
     set +e
-    python "$SCRIPT_DIR/install_llama_prebuilt.py" \
-        "${_RESOLVE_INSTALL_ARGS[@]}" >"$_RESOLVE_LLAMA_LOG" 2>&1
+    _RESOLVE_INSTALL_JSON="$(
+        python "$SCRIPT_DIR/install_llama_prebuilt.py" \
+            "${_RESOLVE_INSTALL_ARGS[@]}" 2>"$_RESOLVE_LLAMA_LOG"
+    )"
     _RESOLVE_LLAMA_STATUS=$?
     set -e
     if [ "$_RESOLVE_LLAMA_STATUS" -eq 0 ]; then
-        _RESOLVED_LLAMA_TAG="$(tail -n 1 "$_RESOLVE_LLAMA_LOG" | tr -d '\r')"
+        _RESOLVED_LLAMA_TAG="$(
+            printf '%s' "${_RESOLVE_INSTALL_JSON:-}" | python -c 'import json,sys; print(json.load(sys.stdin).get("llama_tag",""))' 2>/dev/null || true
+        )"
     else
         _RESOLVED_LLAMA_TAG=""
     fi
@@ -553,13 +572,21 @@ else
         # so the resolver prefers the latest usable Unsloth-published upstream tag
         # before falling back to the bleeding-edge ggml-org/llama.cpp tag.
         _RESOLVE_FALLBACK_ARGS=(--resolve-llama-tag "$_REQUESTED_LLAMA_TAG" --published-repo "$_HELPER_RELEASE_REPO")
+        _RESOLVE_FALLBACK_ARGS+=(--output-format json)
         if [ -n "${UNSLOTH_LLAMA_RELEASE_TAG:-}" ]; then
             _RESOLVE_FALLBACK_ARGS+=(--published-release-tag "$UNSLOTH_LLAMA_RELEASE_TAG")
         fi
-        _RESOLVED_LLAMA_TAG="$(python "$SCRIPT_DIR/install_llama_prebuilt.py" "${_RESOLVE_FALLBACK_ARGS[@]}" 2>/dev/null)"
+        _RESOLVE_FALLBACK_JSON="$(python "$SCRIPT_DIR/install_llama_prebuilt.py" "${_RESOLVE_FALLBACK_ARGS[@]}" 2>/dev/null)"
         _RESOLVE_UPSTREAM_STATUS=$?
         set -e
-        if [ "$_RESOLVE_UPSTREAM_STATUS" -ne 0 ] || [ -z "$_RESOLVED_LLAMA_TAG" ]; then
+        if [ "$_RESOLVE_UPSTREAM_STATUS" -eq 0 ] && [ -n "${_RESOLVE_FALLBACK_JSON:-}" ]; then
+            _RESOLVED_LLAMA_TAG="$(
+                printf '%s' "$_RESOLVE_FALLBACK_JSON" | python -c 'import json,sys; print(json.load(sys.stdin).get("llama_tag",""))' 2>/dev/null || true
+            )"
+        else
+            _RESOLVED_LLAMA_TAG=""
+        fi
+        if [ -z "$_RESOLVED_LLAMA_TAG" ]; then
             _RESOLVED_LLAMA_TAG="$_REQUESTED_LLAMA_TAG"
         fi
         _NEED_LLAMA_SOURCE_BUILD=true
@@ -701,7 +728,40 @@ else
         step "llama.cpp" "skipped (git not found)" "$C_WARN"
         [ -f "$LLAMA_SERVER_BIN" ] || _LLAMA_CPP_DEGRADED=true
     else
+        if [ -z "$_LLAMA_PR" ]; then
+            if [ "$_LLAMA_SOURCE" = "https://github.com/ggml-org/llama.cpp" ]; then
+                _RESOLVE_SOURCE_ARGS=(--resolve-source-build "$_REQUESTED_LLAMA_TAG" --published-repo "$_HELPER_RELEASE_REPO")
+                _RESOLVE_SOURCE_ARGS+=(--output-format json)
+                if [ -n "${UNSLOTH_LLAMA_RELEASE_TAG:-}" ]; then
+                    _RESOLVE_SOURCE_ARGS+=(--published-release-tag "$UNSLOTH_LLAMA_RELEASE_TAG")
+                fi
+                set +e
+                _SOURCE_BUILD_PLAN="$(python "$SCRIPT_DIR/install_llama_prebuilt.py" "${_RESOLVE_SOURCE_ARGS[@]}" 2>/dev/null)"
+                _RESOLVE_SOURCE_STATUS=$?
+                set -e
+                if [ "$_RESOLVE_SOURCE_STATUS" -eq 0 ] && [ -n "$_SOURCE_BUILD_PLAN" ]; then
+                    _RESOLVED_SOURCE_URL="$(
+                        printf '%s' "$_SOURCE_BUILD_PLAN" | python -c 'import json,sys; print(json.load(sys.stdin).get("source_url",""))' 2>/dev/null || true
+                    )"
+                    _RESOLVED_SOURCE_REF_KIND="$(
+                        printf '%s' "$_SOURCE_BUILD_PLAN" | python -c 'import json,sys; print(json.load(sys.stdin).get("source_ref_kind",""))' 2>/dev/null || true
+                    )"
+                    _RESOLVED_SOURCE_REF="$(
+                        printf '%s' "$_SOURCE_BUILD_PLAN" | python -c 'import json,sys; print(json.load(sys.stdin).get("source_ref",""))' 2>/dev/null || true
+                    )"
+                fi
+            fi
+            if [ -z "$_RESOLVED_SOURCE_URL" ]; then
+                _RESOLVED_SOURCE_URL="$_LLAMA_SOURCE"
+            fi
+            if [ -z "$_RESOLVED_SOURCE_REF" ]; then
+                _RESOLVED_SOURCE_REF="$_RESOLVED_LLAMA_TAG"
+            fi
+        fi
+        verbose_substep "source build repo: $_RESOLVED_SOURCE_URL"
+        verbose_substep "source build ref: ${_RESOLVED_SOURCE_REF:-latest} (${_RESOLVED_SOURCE_REF_KIND})"
         BUILD_OK=true
+        mkdir -p "$(dirname "$LLAMA_CPP_DIR")"
         _BUILD_TMP="${LLAMA_CPP_DIR}.build.$$"
         rm -rf "$_BUILD_TMP"
         if [ -n "$_LLAMA_PR" ]; then
@@ -715,12 +775,34 @@ else
                 run_quiet_no_exit "checkout PR #$_LLAMA_PR" \
                     git -C "$_BUILD_TMP" checkout "pr-$_LLAMA_PR" || BUILD_OK=false
             fi
+        elif [ "$_RESOLVED_SOURCE_REF_KIND" = "pull" ] && [ -n "$_RESOLVED_SOURCE_REF" ]; then
+            run_quiet_no_exit "clone llama.cpp" \
+                git clone --depth 1 "${_RESOLVED_SOURCE_URL}.git" "$_BUILD_TMP" || BUILD_OK=false
+            if [ "$BUILD_OK" = true ]; then
+                run_quiet_no_exit "fetch source PR ref" \
+                    git -C "$_BUILD_TMP" fetch --depth 1 origin "$_RESOLVED_SOURCE_REF" || BUILD_OK=false
+            fi
+            if [ "$BUILD_OK" = true ]; then
+                run_quiet_no_exit "checkout source PR ref" \
+                    git -C "$_BUILD_TMP" checkout -B unsloth-llama-build FETCH_HEAD || BUILD_OK=false
+            fi
+        elif [ "$_RESOLVED_SOURCE_REF_KIND" = "commit" ] && [ -n "$_RESOLVED_SOURCE_REF" ]; then
+            run_quiet_no_exit "clone llama.cpp" \
+                git clone --depth 1 "${_RESOLVED_SOURCE_URL}.git" "$_BUILD_TMP" || BUILD_OK=false
+            if [ "$BUILD_OK" = true ]; then
+                run_quiet_no_exit "fetch source commit" \
+                    git -C "$_BUILD_TMP" fetch --depth 1 origin "$_RESOLVED_SOURCE_REF" || BUILD_OK=false
+            fi
+            if [ "$BUILD_OK" = true ]; then
+                run_quiet_no_exit "checkout source commit" \
+                    git -C "$_BUILD_TMP" checkout -B unsloth-llama-build FETCH_HEAD || BUILD_OK=false
+            fi
         else
             _CLONE_ARGS=(git clone --depth 1)
-            if [ "$_RESOLVED_LLAMA_TAG" != "latest" ] && [ -n "$_RESOLVED_LLAMA_TAG" ]; then
-                _CLONE_ARGS+=(--branch "$_RESOLVED_LLAMA_TAG")
+            if [ "$_RESOLVED_SOURCE_REF" != "latest" ] && [ -n "$_RESOLVED_SOURCE_REF" ]; then
+                _CLONE_ARGS+=(--branch "$_RESOLVED_SOURCE_REF")
             fi
-            _CLONE_ARGS+=("${_LLAMA_SOURCE}.git" "$_BUILD_TMP")
+            _CLONE_ARGS+=("${_RESOLVED_SOURCE_URL}.git" "$_BUILD_TMP")
             run_quiet_no_exit "clone llama.cpp" \
                 "${_CLONE_ARGS[@]}" || BUILD_OK=false
         fi

From 5a5f1a4f342b058cf5d161513148be33a4d2f7aa Mon Sep 17 00:00:00 2001
From: Lee Jackson <130007945+Imagineer99@users.noreply.github.com>
Date: Thu, 2 Apr 2026 13:04:23 +0100
Subject: [PATCH 010/216] studio: fix chat font changes leaking outside chat
 page (#4775)

* fix(frontend): scope sans font overrides to chat thread only

* fix(frontend): use font-sans fallback for heading stack and simplify chat font rules

---------

Co-authored-by: Daniel Han <danielhanchen@gmail.com>
---
 studio/frontend/src/index.css | 26 +++++++++++++-------------
 1 file changed, 13 insertions(+), 13 deletions(-)

diff --git a/studio/frontend/src/index.css b/studio/frontend/src/index.css
index 45696b2274..a30b4ca287 100644
--- a/studio/frontend/src/index.css
+++ b/studio/frontend/src/index.css
@@ -6,6 +6,7 @@
 @import "shadcn/tailwind.css";
 @import "streamdown/styles.css";
 @import "@fontsource-variable/figtree";
+@import "@fontsource-variable/space-grotesk";
 @import "@fontsource-variable/inter";
 @import "tw-shimmer";
 @plugin "@toolwind/corner-shape";
@@ -13,6 +14,15 @@
 
 @custom-variant dark (&:is(.dark *));
 
+@font-face {
+  font-family: "Hellix";
+  src: url("/fonts/Hellix-SemiBold.woff2") format("woff2"),
+    url("/fonts/Hellix-SemiBold.woff") format("woff");
+  font-weight: 600;
+  font-style: normal;
+  font-display: swap;
+}
+
 :root {
   /* Animation timing */
   --duration-micro: 100ms;
@@ -57,7 +67,7 @@
   --sidebar-ring: oklch(0.6929 0.1396 166.5513);
   --destructive-foreground: oklch(1 0 0);
   --font-sans: "Inter Variable", ui-sans-serif, sans-serif, system-ui;
-  --font-heading: var(--font-sans);
+  --font-heading: "Hellix", "Space Grotesk Variable", var(--font-sans);
   --font-serif: Source Serif 4, serif;
   --font-mono: JetBrains Mono, monospace;
   --shadow-color: hsl(0 0% 0%);
@@ -149,7 +159,7 @@
 
 @theme inline {
   --font-sans: "Inter Variable", ui-sans-serif, sans-serif, system-ui;
-  --font-heading: var(--font-sans);
+  --font-heading: "Hellix", "Space Grotesk Variable", var(--font-sans);
   --color-sidebar-ring: var(--sidebar-ring);
   --color-sidebar-border: var(--sidebar-border);
   --color-sidebar-accent-foreground: var(--sidebar-accent-foreground);
@@ -364,17 +374,7 @@
 
   /* Chat: use the app sans stack for UI + prose. */
   .aui-thread-root {
-    font-family: var(--font-sans);
-  }
-
-  .aui-thread-root .font-medium,
-  .aui-thread-root .font-semibold,
-  .aui-thread-root .font-bold,
-  .aui-thread-root .font-heading {
-    font-family: var(--font-sans);
-  }
-
-  .aui-thread-root :where(h1, h2, h3, h4, h5, h6) {
+    --font-heading: var(--font-sans);
     font-family: var(--font-sans);
   }
 

From c8d311a053711348b6b5ca09e2ab810e9bf0ceb7 Mon Sep 17 00:00:00 2001
From: Daniel Han <danielhanchen@gmail.com>
Date: Thu, 2 Apr 2026 05:08:16 -0700
Subject: [PATCH 011/216] feat(studio): display images from Python tool
 execution in chat UI (#4778)

* feat(studio): display images from Python tool execution in chat UI

When the model calls the Python tool to create a matplotlib plot or
other image file, the image now displays inline in the chat output
instead of being invisible to the user.

Backend:
- Detect new image files (png/jpg/gif/webp/bmp) after Python subprocess
  completes by diffing os.listdir before/after execution
- Append __IMAGES__ sentinel to tool result for frontend consumption
- Strip sentinel before injecting result into LLM context (role: tool)
  so the model never sees file paths
- Add GET /sandbox/{session_id}/{filename} endpoint with JWT auth
  (header or query param), path traversal protection, extension
  allowlist, realpath containment check, and nosniff header

Frontend:
- Parse __IMAGES__ sentinel in tool_end SSE events, create structured
  result with text/images/sessionId
- Render <img> tags in Python tool UI pointing at the sandbox endpoint

Also fixes a bug where SyntaxError in user code was misreported as
"unsafe code detected" instead of showing the actual Python traceback.
The _check_code_safety function now lets SyntaxError pass through to
the subprocess for a proper error message.

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

* fix(studio): improve SVG detection and strip XML preamble

Handle <?xml ...?> declarations before <svg> tags in code fences,
strip XML declaration from SVGs before data URI rendering, and
update the sloth suggestion prompt to request showing code.

* fix(studio): persist parentId so retries survive reload

The append() handler was destructuring only { message } from
ExportedMessageRepositoryItem and discarding parentId. When loading
a saved thread, load() used ExportedMessageRepository.fromArray()
which chains all messages sequentially, flattening retry branches
into a linear list.

Now append() writes parentId to the MessageRecord, and load()
reconstructs the tree when parentIds are present. Old threads
without parentId fall back to the existing fromArray() behavior.

* fix(studio): address review findings for image display and retry persistence

Image detection:
- Use mtime comparison instead of filename-only diff so overwritten
  files (e.g. plt.savefig("chart.png") called twice) are detected

Sentinel parsing:
- Use rsplit/lastIndexOf instead of split/indexOf so user code that
  prints __IMAGES__: does not collide with the backend sentinel

Mixed legacy/new threads:
- For old messages without a stored parentId, infer sequential parent
  from the previous message instead of null, preventing multiple roots

Sandbox endpoint:
- Change Cache-Control from "public, max-age=3600" to "private,
  no-store" since these are authenticated responses

---------

Co-authored-by: Daniel Han <danielhanchen@users.noreply.github.com>
Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com>
---
 studio/backend/core/inference/llama_cpp.py    |  7 +-
 studio/backend/core/inference/tools.py        | 45 ++++++++-
 studio/backend/routes/inference.py            | 91 ++++++++++++++++++-
 .../components/assistant-ui/markdown-text.tsx | 13 ++-
 .../src/components/assistant-ui/thread.tsx    |  2 +-
 .../assistant-ui/tool-ui-python.tsx           | 56 ++++++++++--
 .../src/features/chat/api/chat-adapter.ts     | 18 +++-
 .../src/features/chat/runtime-provider.tsx    | 29 +++++-
 studio/frontend/src/features/chat/types.ts    |  1 +
 9 files changed, 245 insertions(+), 17 deletions(-)

diff --git a/studio/backend/core/inference/llama_cpp.py b/studio/backend/core/inference/llama_cpp.py
index 1b7fe548e4..80dda61b02 100644
--- a/studio/backend/core/inference/llama_cpp.py
+++ b/studio/backend/core/inference/llama_cpp.py
@@ -2905,10 +2905,15 @@ class LlamaCppBackend:
                         _error_prefixes
                     )
                     _tool_call_history.append((_tc_key, _is_error))
+                    # Strip image sentinel before feeding result to the LLM
+                    # (the full result with sentinel is still yielded via
+                    # tool_end so the frontend can extract image paths).
                     _result_content = result
+                    if "\n__IMAGES__:" in _result_content:
+                        _result_content = _result_content.rsplit("\n__IMAGES__:", 1)[0]
                     if _is_error:
                         _result_content = (
-                            result + "\n\nThe tool call encountered an issue. "
+                            _result_content + "\n\nThe tool call encountered an issue. "
                             "Please try a different approach or rephrase your request."
                         )
 
diff --git a/studio/backend/core/inference/tools.py b/studio/backend/core/inference/tools.py
index d425daa49d..b23372b766 100644
--- a/studio/backend/core/inference/tools.py
+++ b/studio/backend/core/inference/tools.py
@@ -26,6 +26,10 @@ from loggers import get_logger
 logger = get_logger(__name__)
 
 _EXEC_TIMEOUT = 300  # 5 minutes
+
+# Strict raster-image allowlist for sandbox file serving.
+# No .svg (XSS risk via embedded scripts), no .html, no .pdf.
+_IMAGE_EXTS = frozenset({".png", ".jpg", ".jpeg", ".gif", ".webp", ".bmp"})
 _MAX_OUTPUT_CHARS = 8000  # truncate long output
 _BASH_BLOCKED_WORDS = {"rm", "sudo", "dd", "chmod", "mkfs", "shutdown", "reboot"}
 
@@ -591,6 +595,12 @@ def _check_code_safety(code: str) -> str | None:
     """
     safe, info = _check_signal_escape_patterns(code)
     if not safe:
+        # SyntaxError from ast.parse -- let these through so the subprocess
+        # produces a normal Python traceback instead of a misleading
+        # "unsafe code detected" message.
+        if info.get("error"):
+            return None
+
         reasons = [
             item.get("description", "") for item in info.get("signal_tampering", [])
         ]
@@ -634,6 +644,17 @@ def _python_exec(
 
     tmp_path = None
     workdir = _get_workdir(session_id)
+    # Snapshot image mtimes so we detect both new and overwritten files.
+    _before: dict[str, int] = {}
+    if os.path.isdir(workdir):
+        for _name in os.listdir(workdir):
+            if os.path.splitext(_name)[1].lower() in _IMAGE_EXTS:
+                _p = os.path.join(workdir, _name)
+                if os.path.isfile(_p):
+                    try:
+                        _before[_name] = os.stat(_p).st_mtime_ns
+                    except OSError:
+                        pass
     try:
         fd, tmp_path = tempfile.mkstemp(
             suffix = ".py", prefix = "studio_exec_", dir = workdir
@@ -669,7 +690,29 @@ def _python_exec(
         result = output or ""
         if proc.returncode != 0:
             result = f"Exit code {proc.returncode}:\n{result}"
-        return _truncate(result) if result.strip() else "(no output)"
+        result = _truncate(result) if result.strip() else "(no output)"
+
+        # Detect new or overwritten image files and append sentinel for frontend
+        if session_id and os.path.isdir(workdir):
+            new_images = []
+            for _name in os.listdir(workdir):
+                if os.path.splitext(_name)[1].lower() not in _IMAGE_EXTS:
+                    continue
+                _p = os.path.join(workdir, _name)
+                if not os.path.isfile(_p):
+                    continue
+                try:
+                    _mtime = os.stat(_p).st_mtime_ns
+                except OSError:
+                    continue
+                if _name not in _before or _mtime != _before[_name]:
+                    new_images.append(_name)
+            if new_images:
+                import json as _json
+
+                result += f"\n__IMAGES__:{_json.dumps(sorted(new_images))}"
+
+        return result
 
     except Exception as e:
         return f"Execution error: {e}"
diff --git a/studio/backend/routes/inference.py b/studio/backend/routes/inference.py
index 0c1d7e37f9..8f058bd0f2 100644
--- a/studio/backend/routes/inference.py
+++ b/studio/backend/routes/inference.py
@@ -5,11 +5,12 @@
 Inference API routes for model loading and text generation.
 """
 
+import os
 import sys
 import time
 import uuid
 from pathlib import Path
-from fastapi import APIRouter, Depends, HTTPException, Request
+from fastapi import APIRouter, Depends, HTTPException, Request, status
 from fastapi.responses import StreamingResponse, JSONResponse
 from typing import Optional
 import json
@@ -1680,6 +1681,94 @@ async def openai_chat_completions(
             raise HTTPException(status_code = 500, detail = str(e))
 
 
+# =====================================================================
+# Sandbox file serving  (/sandbox/{session_id}/{filename})
+# =====================================================================
+
+_SANDBOX_MEDIA_TYPES = {
+    ".png": "image/png",
+    ".jpg": "image/jpeg",
+    ".jpeg": "image/jpeg",
+    ".gif": "image/gif",
+    ".webp": "image/webp",
+    ".bmp": "image/bmp",
+}
+
+
+@router.get("/sandbox/{session_id}/{filename}")
+async def serve_sandbox_file(
+    session_id: str,
+    filename: str,
+    request: Request,
+    token: Optional[str] = None,
+):
+    """
+    Serve image files created by Python tool execution.
+
+    Accepts auth via Authorization header OR ?token= query param
+    (needed because <img src> cannot send custom headers).
+    """
+    from fastapi.responses import FileResponse
+
+    # ── Authentication (header or query param) ──────────────────
+    auth_header = request.headers.get("authorization")
+    if auth_header and auth_header.lower().startswith("bearer "):
+        jwt_token = auth_header[7:]
+    elif token:
+        jwt_token = token
+    else:
+        raise HTTPException(
+            status_code = status.HTTP_401_UNAUTHORIZED,
+            detail = "Missing authentication token",
+        )
+    from fastapi.security import HTTPAuthorizationCredentials
+
+    creds = HTTPAuthorizationCredentials(scheme = "Bearer", credentials = jwt_token)
+    await get_current_subject(creds)
+
+    # ── Filename sanitization ───────────────────────────────────
+    safe_filename = os.path.basename(filename)
+    if not safe_filename or safe_filename in (".", ".."):
+        raise HTTPException(status_code = 404, detail = "Not found")
+
+    # ── Extension allowlist ─────────────────────────────────────
+    ext = os.path.splitext(safe_filename)[1].lower()
+    media_type = _SANDBOX_MEDIA_TYPES.get(ext)
+    if not media_type:
+        raise HTTPException(
+            status_code = status.HTTP_403_FORBIDDEN,
+            detail = "File type not allowed",
+        )
+
+    # ── Path containment check ──────────────────────────────────
+    home = os.path.expanduser("~")
+    sandbox_root = os.path.realpath(os.path.join(home, "studio_sandbox"))
+    safe_session = os.path.basename(session_id.replace("..", ""))
+    if not safe_session:
+        raise HTTPException(status_code = 404, detail = "Not found")
+
+    file_path = os.path.realpath(
+        os.path.join(sandbox_root, safe_session, safe_filename)
+    )
+    if not file_path.startswith(sandbox_root + os.sep):
+        raise HTTPException(
+            status_code = status.HTTP_403_FORBIDDEN,
+            detail = "Access denied",
+        )
+
+    if not os.path.isfile(file_path):
+        raise HTTPException(status_code = 404, detail = "Not found")
+
+    return FileResponse(
+        path = file_path,
+        media_type = media_type,
+        headers = {
+            "Cache-Control": "private, no-store",
+            "X-Content-Type-Options": "nosniff",
+        },
+    )
+
+
 # =====================================================================
 # OpenAI-Compatible Models Listing  (/models → /v1/models)
 # =====================================================================
diff --git a/studio/frontend/src/components/assistant-ui/markdown-text.tsx b/studio/frontend/src/components/assistant-ui/markdown-text.tsx
index 17a6f8f054..91ef78fcf9 100644
--- a/studio/frontend/src/components/assistant-ui/markdown-text.tsx
+++ b/studio/frontend/src/components/assistant-ui/markdown-text.tsx
@@ -100,20 +100,27 @@ function getCodeFilename(language: string | null) {
 function isSvgFence(codeFence: CodeFence): boolean {
   const lang = codeFence.language?.toLowerCase() ?? "";
   if (lang === "svg") return true;
-  if ((lang === "xml" || lang === "html") && codeFence.source.trimStart().startsWith("<svg")) return true;
+  if (lang === "xml" || lang === "html") {
+    const trimmed = codeFence.source.trimStart();
+    // Match <svg directly or <?xml ...?> followed by <svg
+    if (trimmed.startsWith("<svg")) return true;
+    if (trimmed.startsWith("<?xml") && trimmed.includes("<svg")) return true;
+  }
   return false;
 }
 
 function isHtmlFence(codeFence: CodeFence): boolean {
   const lang = codeFence.language?.toLowerCase() ?? "";
-  return lang === "html" && !codeFence.source.trimStart().startsWith("<svg");
+  return lang === "html" && !isSvgFence(codeFence);
 }
 
 const UNSAFE_SVG_RE = /<script[\s>]|on\w+\s*=|javascript:|<foreignObject[\s>]|<iframe[\s>]|<embed[\s>]|<object[\s>]/i;
 
 function sanitizeSvg(source: string): string | null {
   if (UNSAFE_SVG_RE.test(source)) return null;
-  return source;
+  // Strip XML declaration (<?xml ...?>) -- not needed for data URI
+  // rendering and can cause issues with some renderers.
+  return source.replace(/^\s*<\?xml[^?]*\?>\s*/i, "");
 }
 
 function SvgPreview({ source }: { source: string }) {
diff --git a/studio/frontend/src/components/assistant-ui/thread.tsx b/studio/frontend/src/components/assistant-ui/thread.tsx
index 5170f6ff69..4db0eda0ad 100644
--- a/studio/frontend/src/components/assistant-ui/thread.tsx
+++ b/studio/frontend/src/components/assistant-ui/thread.tsx
@@ -124,7 +124,7 @@ const SUGGESTION_TOOLS: Record<string, Array<"thinking" | "search" | "code">> =
   "How do you fine-tune an audio model with Unsloth?": ["thinking", "search"],
   "Create a live weather dashboard in HTML using no API key. Show me the code": ["thinking", "code", "search"],
   "Solve the integral of x·sin(x), and verify it step by step": ["thinking", "code"],
-  "Draw an SVG of a cute sloth": ["thinking", "code", "search"],
+  "Draw an SVG of a cute sloth & show the code": ["thinking", "code", "search"],
 };
 
 const toolIconMap = {
diff --git a/studio/frontend/src/components/assistant-ui/tool-ui-python.tsx b/studio/frontend/src/components/assistant-ui/tool-ui-python.tsx
index a510ed0d9e..6aa590ae11 100644
--- a/studio/frontend/src/components/assistant-ui/tool-ui-python.tsx
+++ b/studio/frontend/src/components/assistant-ui/tool-ui-python.tsx
@@ -4,6 +4,7 @@
 "use client";
 
 import { copyToClipboard } from "@/lib/copy-to-clipboard";
+import { getAuthToken } from "@/features/auth/session";
 import type { ToolCallMessagePartComponent } from "@assistant-ui/react";
 import { code as codePlugin } from "@streamdown/code";
 import { CheckIcon, CodeIcon, CopyIcon, LoaderIcon } from "lucide-react";
@@ -15,6 +16,12 @@ import {
   ToolFallbackTrigger,
 } from "./tool-fallback";
 
+interface StructuredResult {
+  text: string;
+  images: string[];
+  sessionId: string;
+}
+
 const MAX_DISPLAY = 10_000;
 const COPY_RESET_MS = 2000;
 const SHIKI_THEME = ["github-light", "github-dark"] as ["github-light", "github-dark"];
@@ -84,6 +91,16 @@ function HighlightedCode({ code: source, language }: { code: string; language: s
   );
 }
 
+function isStructuredResult(val: unknown): val is StructuredResult {
+  return (
+    typeof val === "object" &&
+    val !== null &&
+    "text" in val &&
+    "images" in val &&
+    "sessionId" in val
+  );
+}
+
 const PythonToolUIImpl: ToolCallMessagePartComponent = ({
   args,
   result,
@@ -92,12 +109,24 @@ const PythonToolUIImpl: ToolCallMessagePartComponent = ({
   const code = (args as { code?: string })?.code ?? "";
   const firstLine = code.split("\n")[0]?.slice(0, 60) ?? "";
   const isRunning = status?.type === "running";
-  const output =
-    typeof result === "string"
-      ? result
-      : result
-        ? JSON.stringify(result, null, 2)
-        : "";
+
+  let output: string;
+  let images: string[] = [];
+  let sessionId = "";
+
+  if (isStructuredResult(result)) {
+    output = result.text;
+    images = result.images;
+    sessionId = result.sessionId;
+  } else if (typeof result === "string") {
+    output = result;
+  } else if (result) {
+    output = JSON.stringify(result, null, 2);
+  } else {
+    output = "";
+  }
+
+  const authToken = getAuthToken();
 
   return (
     <ToolFallbackRoot>
@@ -133,6 +162,21 @@ const PythonToolUIImpl: ToolCallMessagePartComponent = ({
               </pre>
             </div>
           ) : null}
+
+          {/* Images from Python tool execution */}
+          {images.length > 0 && sessionId && (
+            <div className="mt-2 flex flex-col gap-2">
+              {images.map((filename) => (
+                <img
+                  key={filename}
+                  src={`/api/inference/sandbox/${encodeURIComponent(sessionId)}/${encodeURIComponent(filename)}${authToken ? `?token=${encodeURIComponent(authToken)}` : ""}`}
+                  alt={filename}
+                  loading="lazy"
+                  className="max-w-full rounded border border-border"
+                />
+              ))}
+            </div>
+          )}
         </div>
       </ToolFallbackContent>
     </ToolFallbackRoot>
diff --git a/studio/frontend/src/features/chat/api/chat-adapter.ts b/studio/frontend/src/features/chat/api/chat-adapter.ts
index 2b8a259930..e287daf33a 100644
--- a/studio/frontend/src/features/chat/api/chat-adapter.ts
+++ b/studio/frontend/src/features/chat/api/chat-adapter.ts
@@ -635,7 +635,23 @@ export function createOpenAIStreamAdapter(): ChatModelAdapter {
                 toolCallParts[toolCallParts.length - 1]?.toolCallId || "";
               const idx = toolCallParts.findIndex((p) => p.toolCallId === id);
               if (idx !== -1) {
-                toolCallParts[idx] = { ...toolCallParts[idx], result: toolEvent.result as string };
+                const rawResult = (toolEvent.result as string) ?? "";
+                const imgMarker = "\n__IMAGES__:";
+                const imgIdx = rawResult.lastIndexOf(imgMarker);
+                let parsedResult: string | { text: string; images: string[]; sessionId: string };
+                if (imgIdx !== -1) {
+                  const text = rawResult.slice(0, imgIdx);
+                  const sessionId = unstable_threadId || "";
+                  try {
+                    const images = JSON.parse(rawResult.slice(imgIdx + imgMarker.length)) as string[];
+                    parsedResult = { text, images, sessionId };
+                  } catch {
+                    parsedResult = rawResult;
+                  }
+                } else {
+                  parsedResult = rawResult;
+                }
+                toolCallParts[idx] = { ...toolCallParts[idx], result: parsedResult };
               }
             }
             // Yield cumulative state so tool UI updates (tools first, text after)
diff --git a/studio/frontend/src/features/chat/runtime-provider.tsx b/studio/frontend/src/features/chat/runtime-provider.tsx
index 404271f896..02f792b509 100644
--- a/studio/frontend/src/features/chat/runtime-provider.tsx
+++ b/studio/frontend/src/features/chat/runtime-provider.tsx
@@ -47,9 +47,9 @@ const DEFAULT_SUGGESTIONS = [
     prompt: "Solve the integral of x·sin(x), and verify it step by step",
   },
   {
-    title: "Draw an SVG of a cute sloth",
+    title: "Draw an SVG of a cute sloth & show the code",
     label: "SVG sloth",
-    prompt: "Draw an SVG of a cute sloth",
+    prompt: "Draw an SVG of a cute sloth & show the code",
   },
 ];
 
@@ -569,10 +569,32 @@ function ThreadHistoryProvider({
           store.setContextUsage(savedUsage);
         }
 
+        // If any message has a stored parentId, reconstruct the tree
+        // so retries/regenerations load as branches instead of being
+        // unrolled into a flat list.  For mixed legacy/new threads
+        // (old messages without parentId + new messages with), infer
+        // sequential parents for old messages to preserve the chain.
+        // Fall back to fromArray for fully legacy threads.
+        const hasParentIds = msgs.some((m) => "parentId" in m);
+        if (hasParentIds) {
+          let previousId: string | null = null;
+          return {
+            messages: msgs.map((m) => {
+              const parentId = "parentId" in m
+                ? (m.parentId ?? null)
+                : previousId;
+              previousId = m.id;
+              return {
+                parentId,
+                message: toThreadMessage(m),
+              };
+            }),
+          };
+        }
         return ExportedMessageRepository.fromArray(msgs.map(toThreadMessage));
       },
 
-      async append({ message }: ExportedMessageRepositoryItem) {
+      async append({ parentId, message }: ExportedMessageRepositoryItem) {
         const { remoteId } = await aui.threadListItem().initialize();
         const content = cloneContent(message.content);
         const attachments =
@@ -586,6 +608,7 @@ function ThreadHistoryProvider({
         await db.messages.put({
           id: message.id,
           threadId: remoteId,
+          parentId: parentId ?? null,
           role: message.role,
           content,
           ...(attachments.length > 0 && { attachments }),
diff --git a/studio/frontend/src/features/chat/types.ts b/studio/frontend/src/features/chat/types.ts
index 11e53d76d3..1f370b6ac1 100644
--- a/studio/frontend/src/features/chat/types.ts
+++ b/studio/frontend/src/features/chat/types.ts
@@ -20,6 +20,7 @@ export interface ThreadRecord {
 export interface MessageRecord {
   id: string;
   threadId: string;
+  parentId?: string | null;
   role: import("@assistant-ui/react").ThreadMessage["role"];
   content: import("@assistant-ui/react").ThreadMessage["content"];
   attachments?: import("@assistant-ui/react").ThreadMessage["attachments"];

From 3b613eb1e8ff71a0fca02b83cdfa06a09c864c86 Mon Sep 17 00:00:00 2001
From: Roland Tannous <115670425+rolandtannous@users.noreply.github.com>
Date: Thu, 2 Apr 2026 18:57:47 +0400
Subject: [PATCH 012/216] ui improvement (#4781)

* ui

* ui

* ui
---
 install_gemma4_mlx.sh | 215 ++++++++++++++++++++++++++++++++++++++++++
 1 file changed, 215 insertions(+)
 create mode 100755 install_gemma4_mlx.sh

diff --git a/install_gemma4_mlx.sh b/install_gemma4_mlx.sh
new file mode 100755
index 0000000000..1f5e7c2d9e
--- /dev/null
+++ b/install_gemma4_mlx.sh
@@ -0,0 +1,215 @@
+#!/bin/bash
+set -e
+
+# ============================================================
+# Gemma 4 MLX — One-command setup + inference
+#
+# Usage:
+#   bash install_gemma4_mlx.sh [--venv-dir DIR]
+#
+# This script:
+#   1. Creates a Python virtual environment
+#   2. Installs uv, mlx, mlx-lm, transformers
+#   3. Downloads gemma4.py and gemma4_text.py from unsloth repo
+#   4. Installs them into mlx-lm's models directory
+# ============================================================
+
+# ── Output style (inspired by unsloth/install.sh) ─────────────
+RULE=""
+_rule_i=0
+while [ "$_rule_i" -lt 52 ]; do
+    RULE="${RULE}─"
+    _rule_i=$((_rule_i + 1))
+done
+
+if [ -n "${NO_COLOR:-}" ]; then
+    C_TITLE= C_DIM= C_OK= C_WARN= C_ERR= C_RST=
+elif [ -t 1 ] || [ -n "${FORCE_COLOR:-}" ]; then
+    _ESC="$(printf '\033')"
+    C_TITLE="${_ESC}[38;5;117m"
+    C_DIM="${_ESC}[38;5;245m"
+    C_OK="${_ESC}[38;5;108m"
+    C_WARN="${_ESC}[38;5;136m"
+    C_ERR="${_ESC}[91m"
+    C_RST="${_ESC}[0m"
+else
+    C_TITLE= C_DIM= C_OK= C_WARN= C_ERR= C_RST=
+fi
+
+step()    { printf "  ${C_DIM}%-18.18s${C_RST}${3:-$C_OK}%s${C_RST}\n" "$1" "$2"; }
+substep() { printf "  ${C_DIM}%-18s${2:-$C_DIM}%s${C_RST}\n" "" "$1"; }
+fail()    { step "error" "$1" "$C_ERR"; exit 1; }
+
+# ── Parse flags ───────────────────────────────────────────────
+VENV_DIR=""
+_next_is_venv=false
+
+for arg in "$@"; do
+    if [ "$_next_is_venv" = true ]; then
+        VENV_DIR="$arg"
+        _next_is_venv=false
+        continue
+    fi
+    case "$arg" in
+        --venv-dir)  _next_is_venv=true ;;
+    esac
+done
+
+# Default venv location
+if [ -z "$VENV_DIR" ]; then
+    VENV_DIR="$HOME/.unsloth/unsloth_gemma4_mlx"
+fi
+
+# ── Banner ────────────────────────────────────────────────────
+echo ""
+printf "  ${C_TITLE}%s${C_RST}\n" "💎 Gemma 4 MLX Installer"
+printf "  ${C_DIM}%s${C_RST}\n" "$RULE"
+echo ""
+
+# ── Platform check ────────────────────────────────────────────
+if [ "$(uname)" != "Darwin" ]; then
+    fail "MLX requires macOS with Apple Silicon. Detected: $(uname)"
+fi
+
+_ARCH=$(uname -m)
+if [ "$_ARCH" != "arm64" ]; then
+    step "warning" "Apple Silicon recommended (detected: $_ARCH)" "$C_WARN"
+fi
+
+step "platform" "macOS ($_ARCH)"
+
+# ── Detect Python ─────────────────────────────────────────────
+PYTHON=""
+for _candidate in python3.12 python3.11 python3.13 python3; do
+    if command -v "$_candidate" >/dev/null 2>&1; then
+        PYTHON="$_candidate"
+        break
+    fi
+done
+
+if [ -z "$PYTHON" ]; then
+    fail "Python 3 not found. Install via: brew install python@3.12"
+fi
+
+_PY_VERSION=$("$PYTHON" -c "import sys; print(f'{sys.version_info.major}.{sys.version_info.minor}.{sys.version_info.micro}')")
+step "python" "$PYTHON ($_PY_VERSION)"
+
+# ── Create virtual environment ────────────────────────────────
+if [ -x "$VENV_DIR/bin/python" ]; then
+    step "venv" "using existing environment"
+    substep "$VENV_DIR"
+else
+    step "venv" "creating virtual environment"
+    substep "$VENV_DIR"
+    mkdir -p "$(dirname "$VENV_DIR")"
+    "$PYTHON" -m venv "$VENV_DIR"
+fi
+
+# ── Install uv ───────────────────────────────────────────────
+if ! command -v uv >/dev/null 2>&1; then
+    step "uv" "installing uv package manager..."
+    _uv_tmp=$(mktemp)
+    curl -LsSf "https://astral.sh/uv/install.sh" -o "$_uv_tmp"
+    sh "$_uv_tmp" </dev/null >/dev/null 2>&1
+    rm -f "$_uv_tmp"
+    if [ -f "$HOME/.local/bin/env" ]; then
+        . "$HOME/.local/bin/env"
+    fi
+    export PATH="$HOME/.local/bin:$PATH"
+    substep "done"
+else
+    step "uv" "found $(uv --version 2>/dev/null || echo 'uv')"
+fi
+
+_VENV_PY="$VENV_DIR/bin/python"
+
+# ── Repo config ──────────────────────────────────────────────
+BRANCH="fix/ui-fix"
+REPO_URL="https://raw.githubusercontent.com/unslothai/unsloth/refs/heads/${BRANCH}"
+
+# ── Install dependencies ──────────────────────────────────────
+step "install" "installing mlx, mlx-lm..."
+uv pip install --python "$_VENV_PY" -q mlx mlx-lm 2>/dev/null
+substep "done"
+
+TRANSFORMERS_WHL="transformers-5.5.0.dev0-py3-none-any.whl"
+TRANSFORMERS_GH="git+https://github.com/huggingface/transformers.git@v5.5-release"
+
+step "install" "installing transformers>=5.5.0..."
+if uv pip install --python "$_VENV_PY" -q "$TRANSFORMERS_GH" 2>/dev/null; then
+    substep "installed from huggingface/transformers v5.5-release"
+elif uv pip install --python "$_VENV_PY" -q "transformers>=5.5.0" 2>/dev/null; then
+    substep "installed from PyPI"
+else
+    substep "not on PyPI, trying unsloth branch..."
+    _whl_tmp=$(mktemp -d)/"${TRANSFORMERS_WHL}"
+    if curl -fsSL "${REPO_URL}/${TRANSFORMERS_WHL}" -o "$_whl_tmp" 2>/dev/null && \
+       uv pip install --python "$_VENV_PY" -q "$_whl_tmp" 2>/dev/null; then
+        substep "installed from branch ${BRANCH}"
+    elif [ -f "./${TRANSFORMERS_WHL}" ]; then
+        substep "using local ./${TRANSFORMERS_WHL}"
+        uv pip install --python "$_VENV_PY" -q "./${TRANSFORMERS_WHL}"
+    else
+        rm -f "$_whl_tmp" 2>/dev/null
+        step "install" "skipping transformers — could not find >=5.5.0" "$C_WARN"
+        substep "tried: huggingface/transformers v5.5-release, PyPI, branch ${BRANCH}, local ./${TRANSFORMERS_WHL}"
+    fi
+    rm -f "$_whl_tmp" 2>/dev/null
+fi
+
+# ── Find mlx-lm models directory ─────────────────────────────
+MLX_MODELS=$("$_VENV_PY" -c "import mlx_lm; print(mlx_lm.__path__[0])")/models
+step "models dir" "$MLX_MODELS"
+
+# ── Download and install Gemma 4 model files ──────────────────
+
+step "download" "installing Gemma 4 model files..."
+
+_install_model_file() {
+    _fname="$1"
+    if curl -fsSL "${REPO_URL}/unsloth/models/${_fname}" -o "${MLX_MODELS}/${_fname}" 2>/dev/null; then
+        substep "downloaded ${_fname} from branch ${BRANCH}"
+    elif [ -f "./${_fname}" ]; then
+        substep "using local ./${_fname}"
+        cp "./${_fname}" "${MLX_MODELS}/${_fname}"
+    else
+        fail "Could not install ${_fname}. Tried:
+                    1) ${REPO_URL}/unsloth/models/${_fname}
+                    2) Local file ./${_fname}
+
+                    To fix, download the file manually and place it in the current directory,
+                    then re-run this script."
+    fi
+}
+
+_install_model_file "gemma4.py"
+_install_model_file "gemma4_text.py"
+
+# Verify files were installed correctly
+if "$_VENV_PY" -c "from mlx_lm.models.gemma4_text import ProportionalRoPE" 2>/dev/null; then
+    substep "model files verified"
+else
+    fail "Model files installed but verification failed (ProportionalRoPE import error).
+                    Try manually from: https://github.com/unslothai/unsloth/tree/feature/${BRANCH}"
+fi
+
+# ── Done ──────────────────────────────────────────────────────
+echo ""
+printf "  ${C_TITLE}%s${C_RST}\n" "Gemma 4 MLX installed!"
+printf "  ${C_DIM}%s${C_RST}\n" "$RULE"
+echo ""
+step "available models" "unsloth/gemma-4-E2B-it-MLX-4bit (/BF16)"
+substep "unsloth/gemma-4-E4B-it-MLX-4bit (/BF16)"
+echo ""
+step "venv activate" "source ${VENV_DIR}/bin/activate"
+echo ""
+step "quick start" "python -m mlx_lm chat --model unsloth/gemma-4-E2B-it-MLX-4bit --max-tokens 200"
+echo ""
+step "python API" "from mlx_lm import load, generate"
+substep "model, tokenizer = load('unsloth/gemma-4-E2B-it-MLX-4bit')"
+substep "messages = [{'role': 'user', 'content': 'Hello!'}]"
+substep "prompt = tokenizer.apply_chat_template(messages, tokenize=False, add_generation_prompt=True)"
+substep "print(generate(model, tokenizer, prompt=prompt, max_tokens=200))"
+echo ""
+printf "  ${C_DIM}%s${C_RST}\n" "$RULE"
+echo ""

From f9c4b08726939fb415f23ae4ed356120dbf796b0 Mon Sep 17 00:00:00 2001
From: Daniel Han <danielhanchen@gmail.com>
Date: Thu, 2 Apr 2026 08:05:55 -0700
Subject: [PATCH 013/216] UI Changes (#4782)

* UI Changes

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

* Remove unrelated test file

---------

Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com>
---
 studio/backend/core/inference/defaults.py     |  20 ++++
 studio/backend/core/inference/llama_cpp.py    |   2 +
 .../backend/requirements/extras-no-deps.txt   |   2 +-
 .../requirements/single-env/constraints.txt   |   4 +-
 studio/backend/utils/transformers_version.py  |   6 +-
 studio/frontend/src/config/training.ts        |   4 +
 unsloth/chat_templates.py                     | 108 ++++++++++++++++++
 unsloth/models/_utils.py                      |  21 ++++
 unsloth/models/loader.py                      |  14 +++
 unsloth/models/vision.py                      |  67 ++++++++++-
 unsloth/ollama_template_mappers.py            |  25 ++++
 11 files changed, 267 insertions(+), 6 deletions(-)

diff --git a/studio/backend/core/inference/defaults.py b/studio/backend/core/inference/defaults.py
index d5e9ca2e97..f3026dddaf 100644
--- a/studio/backend/core/inference/defaults.py
+++ b/studio/backend/core/inference/defaults.py
@@ -6,6 +6,14 @@
 import utils.hardware.hardware as hw
 
 DEFAULT_MODELS_GGUF = [
+    "unsloth/gemma-4-E2B-it-GGUF",
+    "unsloth/gemma-4-E4B-it-GGUF",
+    "unsloth/gemma-4-31B-it-GGUF",
+    "unsloth/gemma-4-26B-A4B-it-GGUF",
+    "unsloth/Qwen3.5-4B-GGUF",
+    "unsloth/Qwen3.5-9B-GGUF",
+    "unsloth/Qwen3.5-35B-A3B-GGUF",
+    "unsloth/Qwen3.5-0.8B-GGUF",
     "unsloth/Llama-3.2-1B-Instruct-GGUF",
     "unsloth/Llama-3.2-3B-Instruct-GGUF",
     "unsloth/Llama-3.1-8B-Instruct-GGUF",
@@ -15,6 +23,18 @@ DEFAULT_MODELS_GGUF = [
 ]
 
 DEFAULT_MODELS_STANDARD = [
+    "unsloth/gemma-4-E2B-it-GGUF",
+    "unsloth/gemma-4-E4B-it-GGUF",
+    "unsloth/gemma-4-31B-it-GGUF",
+    "unsloth/gemma-4-26B-A4B-it-GGUF",
+    "unsloth/Qwen3.5-4B-GGUF",
+    "unsloth/Qwen3.5-9B-GGUF",
+    "unsloth/Qwen3.5-35B-A3B-GGUF",
+    "unsloth/Qwen3.5-0.8B-GGUF",
+    "unsloth/gemma-4-E2B-it",
+    "unsloth/gemma-4-E4B-it",
+    "unsloth/gemma-4-31B-it",
+    "unsloth/gemma-4-26B-A4B-it",
     "unsloth/Qwen3-4B-Instruct-2507",
     "unsloth/Meta-Llama-3.1-8B-Instruct-bnb-4bit",
     "unsloth/Mistral-Nemo-Instruct-2407-bnb-4bit",
diff --git a/studio/backend/core/inference/llama_cpp.py b/studio/backend/core/inference/llama_cpp.py
index 80dda61b02..b436f4dc6d 100644
--- a/studio/backend/core/inference/llama_cpp.py
+++ b/studio/backend/core/inference/llama_cpp.py
@@ -810,7 +810,9 @@ class LlamaCppBackend:
                 # Detect tool calling support from chat template
                 tool_markers = [
                     "{%- if tools %}",
+                    "{%- if tools -%}",
                     "{% if tools %}",
+                    "{% if tools -%}",
                     '"role" == "tool"',
                     "'role' == 'tool'",
                     'message.role == "tool"',
diff --git a/studio/backend/requirements/extras-no-deps.txt b/studio/backend/requirements/extras-no-deps.txt
index 9934bacd24..325eb494fc 100644
--- a/studio/backend/requirements/extras-no-deps.txt
+++ b/studio/backend/requirements/extras-no-deps.txt
@@ -11,6 +11,6 @@ git+https://github.com/meta-pytorch/OpenEnv.git
 # executorch>=1.0.1               # 41.5 MB - no imports in unsloth/zoo/studio
 torch-c-dlpack-ext
 sentence_transformers==5.2.0
-transformers==4.57.6
+transformers>=4.57.6
 pytorch_tokenizers
 kernels
diff --git a/studio/backend/requirements/single-env/constraints.txt b/studio/backend/requirements/single-env/constraints.txt
index 156f78567e..37391e739e 100644
--- a/studio/backend/requirements/single-env/constraints.txt
+++ b/studio/backend/requirements/single-env/constraints.txt
@@ -1,8 +1,8 @@
 # Single-env pins for unsloth + studio + data-designer
 # Keep compatible with unsloth transformers bounds.
-transformers==4.57.6
+transformers>=4.57.6
 trl==0.23.1
-huggingface-hub==0.36.2
+huggingface-hub>=0.36.2
 
 # Studio stack
 datasets==4.3.0
diff --git a/studio/backend/utils/transformers_version.py b/studio/backend/utils/transformers_version.py
index d8724de723..9db4b3ce4b 100644
--- a/studio/backend/utils/transformers_version.py
+++ b/studio/backend/utils/transformers_version.py
@@ -47,6 +47,8 @@ TRANSFORMERS_5_MODEL_SUBSTRINGS: tuple[str, ...] = (
     "qwen3.5",  # Qwen3.5 family (35B-A3B, etc.)
     "qwen3-next",  # Qwen3-Next and variants
     "tiny_qwen3_moe",  # imdatta0/tiny_qwen3_moe_2.8B_0.7B
+    "gemma-4",  # Gemma-4 (E2B-it, E4B-it, 31B-it, 26B-A4B-it)
+    "gemma4",  # Gemma-4 alternate naming
 )
 
 # Tokenizer classes that only exist in transformers>=5.x
@@ -58,7 +60,7 @@ _TRANSFORMERS_5_TOKENIZER_CLASSES: set[str] = {
 _tokenizer_class_cache: dict[str, bool] = {}
 
 # Versions
-TRANSFORMERS_5_VERSION = "5.3.0"
+TRANSFORMERS_5_VERSION = "5.5.0.dev0"
 TRANSFORMERS_DEFAULT_VERSION = "4.57.6"
 
 # Pre-installed directory for transformers 5.x — created by setup.sh / setup.ps1
@@ -258,7 +260,7 @@ def _purge_modules() -> int:
 
 _VENV_T5_PACKAGES = (
     f"transformers=={TRANSFORMERS_5_VERSION}",
-    "huggingface_hub==1.7.1",
+    "huggingface_hub==1.8.0",
     "hf_xet==1.4.2",
     "tiktoken",
 )
diff --git a/studio/frontend/src/config/training.ts b/studio/frontend/src/config/training.ts
index 42fccde552..913d612838 100644
--- a/studio/frontend/src/config/training.ts
+++ b/studio/frontend/src/config/training.ts
@@ -141,6 +141,10 @@ export const MODEL_TYPE_TO_HF_TASK: Record<ModelType, PipelineType> = {
 
 
 export const PRIORITY_TRAINING_MODELS: readonly string[] = [
+  "unsloth/gemma-4-E2B-it",
+  "unsloth/gemma-4-E4B-it",
+  "unsloth/gemma-4-31B-it",
+  "unsloth/gemma-4-26B-A4B-it",
   "unsloth/Qwen3.5-2B",
   "unsloth/Qwen3.5-9B",
   "unsloth/gpt-oss-20b",
diff --git a/unsloth/chat_templates.py b/unsloth/chat_templates.py
index 35eb871529..71f91cc828 100644
--- a/unsloth/chat_templates.py
+++ b/unsloth/chat_templates.py
@@ -863,6 +863,114 @@ DEFAULT_SYSTEM_MESSAGE["gemma-3n"] = None # No system message in Gemma-3n
 CHAT_TEMPLATES["gemma3n"] = (gemma3n_template, gemma3n_template_eos_token, False, gemma3n_ollama,)
 DEFAULT_SYSTEM_MESSAGE["gemma3n"] = None # No system message in Gemma-3n
 
+# =========================================== Gemma-4
+# Gemma-4 uses <|turn>role\n...<turn|>\n format
+gemma4_template = \
+"""{%- if messages[0]['role'] == 'system' -%}
+    {%- set first_user_prefix = messages[0]['content'] + '\n\n' -%}
+    {%- set loop_messages = messages[1:] -%}
+{%- else -%}
+    {%- set first_user_prefix = "" -%}
+    {%- set loop_messages = messages -%}
+{%- endif -%}
+{%- for message in loop_messages -%}
+    {%- if (message['role'] == 'user') != (loop.index0 % 2 == 0) -%}
+        {{ raise_exception("Conversation roles must alternate user/assistant/user/assistant/...") }}
+    {%- endif -%}
+    {%- if (message['role'] == 'assistant') -%}
+        {%- set role = "model" -%}
+    {%- else -%}
+        {%- set role = message['role'] -%}
+    {%- endif -%}
+    {{ '<|turn>' + role + '\n' + (first_user_prefix if loop.first else "") }}
+    {%- if message['content'] is string -%}
+        {{ message['content'] | trim }}
+    {%- elif message['content'] is iterable -%}
+        {%- for item in message['content'] -%}
+            {%- if item['type'] == 'audio' -%}
+                {{ '<|audio|>' }}
+            {%- elif item['type'] == 'image' -%}
+                {{ '<|image|>' }}
+            {%- elif item['type'] == 'video' -%}
+                {{ '<|video|>' }}
+            {%- elif item['type'] == 'text' -%}
+                {{ item['text'] | trim }}
+            {%- endif -%}
+        {%- endfor -%}
+    {%- else -%}
+        {{ raise_exception("Invalid content type") }}
+    {%- endif -%}
+    {{ '<turn|>\n' }}
+{%- endfor -%}
+{%- if add_generation_prompt -%}
+    {{'<|turn>model\n'}}
+{%- endif -%}
+"""
+
+try:
+    gemma4_ollama = _ollama_template("gemma-4")
+except KeyError:
+    gemma4_ollama = ""
+gemma4_template_eos_token = "<turn|>"
+CHAT_TEMPLATES["gemma-4"] = (gemma4_template, gemma4_template_eos_token, False, gemma4_ollama,)
+DEFAULT_SYSTEM_MESSAGE["gemma-4"] = None
+
+CHAT_TEMPLATES["gemma4"] = (gemma4_template, gemma4_template_eos_token, False, gemma4_ollama,)
+DEFAULT_SYSTEM_MESSAGE["gemma4"] = None
+
+# Gemma-4 with empty thought channel (required for larger models like 31B, 26B-A4B)
+# Injects <|channel>thought\n<channel|> at the start of each model response during training
+gemma4_thinking_template = \
+"""{%- if messages[0]['role'] == 'system' -%}
+    {%- set first_user_prefix = messages[0]['content'] + '\n\n' -%}
+    {%- set loop_messages = messages[1:] -%}
+{%- else -%}
+    {%- set first_user_prefix = "" -%}
+    {%- set loop_messages = messages -%}
+{%- endif -%}
+{%- for message in loop_messages -%}
+    {%- if (message['role'] == 'user') != (loop.index0 % 2 == 0) -%}
+        {{ raise_exception("Conversation roles must alternate user/assistant/user/assistant/...") }}
+    {%- endif -%}
+    {%- if (message['role'] == 'assistant') -%}
+        {%- set role = "model" -%}
+    {%- else -%}
+        {%- set role = message['role'] -%}
+    {%- endif -%}
+    {{ '<|turn>' + role + '\n' + (first_user_prefix if loop.first else "") }}
+    {%- if role == "model" -%}
+        {{ '<|channel>thought\n<channel|>' }}
+    {%- endif -%}
+    {%- if message['content'] is string -%}
+        {{ message['content'] | trim }}
+    {%- elif message['content'] is iterable -%}
+        {%- for item in message['content'] -%}
+            {%- if item['type'] == 'audio' -%}
+                {{ '<|audio|>' }}
+            {%- elif item['type'] == 'image' -%}
+                {{ '<|image|>' }}
+            {%- elif item['type'] == 'video' -%}
+                {{ '<|video|>' }}
+            {%- elif item['type'] == 'text' -%}
+                {{ item['text'] | trim }}
+            {%- endif -%}
+        {%- endfor -%}
+    {%- else -%}
+        {{ raise_exception("Invalid content type") }}
+    {%- endif -%}
+    {{ '<turn|>\n' }}
+{%- endfor -%}
+{%- if add_generation_prompt -%}
+    {{'<|turn>model\n'}}
+{%- endif -%}
+"""
+
+CHAT_TEMPLATES["gemma-4-thinking"] = (gemma4_thinking_template, gemma4_template_eos_token, False, gemma4_ollama,)
+DEFAULT_SYSTEM_MESSAGE["gemma-4-thinking"] = None
+
+CHAT_TEMPLATES["gemma4-thinking"] = (gemma4_thinking_template, gemma4_template_eos_token, False, gemma4_ollama,)
+DEFAULT_SYSTEM_MESSAGE["gemma4-thinking"] = None
+
 # =========================================== GPT-OSS
 # Obtained via
 # print(tokenizer.chat_template.replace("}\n", "####").replace("\n", "\\n").replace("####", "}\n"))
diff --git a/unsloth/models/_utils.py b/unsloth/models/_utils.py
index df6462eade..1c803a4972 100644
--- a/unsloth/models/_utils.py
+++ b/unsloth/models/_utils.py
@@ -537,6 +537,15 @@ try:
 except:
     pass
 
+# Gemma4 It is strongly recommended to train Gemma4 models with the `eager`
+try:
+    from transformers.models.gemma4.modeling_gemma4 import logger as gemma4_logger
+
+    gemma4_logger.addFilter(HideLoggingMessage("strongly recommended"))
+    del gemma4_logger
+except:
+    pass
+
 # Xet Storage is enabled for this repo, but the 'hf_xet' package is not installed.
 try:
     from huggingface_hub.file_download import logger as hub_logger
@@ -1930,6 +1939,18 @@ def _unsloth_pre_compute_loss(self, model, inputs, *args, **kwargs):
             _has_ccm = _mod is not None and hasattr(_mod, "create_causal_mask_mapping")
             if _has_ccm and _inner.training:
                 inputs["token_type_ids"] = torch.zeros_like(inputs["input_ids"])
+    # Gemma4 uses mm_token_type_ids (not token_type_ids) for VLM masking
+    if "mm_token_type_ids" not in inputs and "input_ids" in inputs:
+        _inner = model
+        for _attr in ("base_model", "model", "model"):
+            _inner = getattr(_inner, _attr, _inner)
+        if getattr(getattr(_inner, "config", None), "model_type", "") in ("gemma4",):
+            import sys as _sys
+
+            _mod = _sys.modules.get(type(_inner).__module__)
+            _has_ccm = _mod is not None and hasattr(_mod, "create_causal_mask_mapping")
+            if _has_ccm and _inner.training:
+                inputs["mm_token_type_ids"] = torch.zeros_like(inputs["input_ids"])
 
     outputs = self._old_compute_loss(model, inputs, *args, **kwargs)
     return outputs
diff --git a/unsloth/models/loader.py b/unsloth/models/loader.py
index 9b7b1b02c5..297da73183 100644
--- a/unsloth/models/loader.py
+++ b/unsloth/models/loader.py
@@ -78,6 +78,7 @@ SUPPORTS_QWEN3_MOE = transformers_version >= Version("4.50.3")
 SUPPORTS_FALCON_H1 = transformers_version >= Version("4.53.0")
 SUPPORTS_GEMMA3N = transformers_version >= Version("4.53.0")
 SUPPORTS_GPTOSS = transformers_version >= Version("4.55.0")
+SUPPORTS_GEMMA4 = transformers_version >= Version("5.5.0.dev0")
 # Transformers v5 meta-device loading corrupts non-persistent buffers (inv_freq).
 # See _fix_rope_inv_freq() below for details.
 _NEEDS_ROPE_FIX = transformers_version >= Version("5.0.0")
@@ -107,6 +108,8 @@ FORCE_FLOAT32 = [
     "gemma3n",
     "gpt_oss",
     "qwen3_5",  # Qwen3.5 GDN layers produce NaN grad norms in float16 training
+    "gemma4,",  # Add comma bc gemma4 will match gemma4_text
+    "gemma4_text",
 ]
 
 global DISABLE_COMPILE_MODEL_NAMES
@@ -1130,6 +1133,17 @@ class FastModel(FastBaseModel):
             raise RuntimeError(
                 "Unsloth: Qwen 2.5 only works on transformers >= 4.49.0." + LATEST
             )
+        # Gemma 4 must be before Gemma 3N and Gemma 3
+        elif "gemma4" in model_types_all:
+            if not SUPPORTS_GEMMA4:
+                raise RuntimeError(
+                    "Unsloth: Gemma 4 requires transformers >= 5.5.0" + LATEST
+                )
+            os.environ["UNSLOTH_DISABLE_STATIC_GENERATION"] = "1"
+            os.environ["UNSLOTH_HIGH_PRECISION_LAYERNORM"] = "1"
+            # Disable flex_attention for Gemma-4: flex compile overhead is 2.7x slower
+            # than SDPA. Our attention patch ensures Q/K/V dtype alignment for SDPA.
+            os.environ["UNSLOTH_ENABLE_FLEX_ATTENTION"] = "0"
         # Gemma 3N must be before Gemma 3
         elif "gemma3n" in model_types_all:
             if transformers_version < Version("4.53.0"):
diff --git a/unsloth/models/vision.py b/unsloth/models/vision.py
index ceff1c4393..5abeb3a81a 100644
--- a/unsloth/models/vision.py
+++ b/unsloth/models/vision.py
@@ -216,6 +216,10 @@ def unsloth_base_fast_generate(
         kwargs["pixel_values"] = kwargs["pixel_values"].to(dtype)
     except:
         pass
+    try:
+        kwargs["pixel_values_videos"] = kwargs["pixel_values_videos"].to(dtype)
+    except:
+        pass
 
     # Mixed precision autocast
     if os.environ.get("UNSLOTH_FORCE_FLOAT32", "0") == "1":
@@ -1029,6 +1033,15 @@ class FastBaseModel:
                     f"Unsloth: Warning - VLM processor fallback returned None for model_type={model_type_arch}",
                     file = sys.stderr,
                 )
+        # Backwards compat: if processor has no chat_template (e.g. old saves without
+        # chat_template.jinja) but the inner tokenizer does, copy it to the processor.
+        if (
+            hasattr(tokenizer, "tokenizer")
+            and getattr(tokenizer, "chat_template", None) is None
+            and getattr(tokenizer.tokenizer, "chat_template", None) is not None
+        ):
+            tokenizer.chat_template = tokenizer.tokenizer.chat_template
+
         if hasattr(tokenizer, "tokenizer"):
             __tokenizer = tokenizer.tokenizer
             # Add padding side as well
@@ -1285,7 +1298,59 @@ class FastBaseModel:
             model,
             use_gradient_checkpointing = use_gradient_checkpointing,
         )
+        # Gemma4 ClippableLinear wraps nn.Linear -- PEFT can't inject LoRA on it directly.
+        # Monkey-patch PEFT to target the inner .linear child instead.
+        _clippable_linear_cls = None
+        try:
+            from transformers.models.gemma4.modeling_gemma4 import (
+                Gemma4ClippableLinear as _clippable_linear_cls,
+            )
+        except ImportError:
+            pass
+        if _clippable_linear_cls is not None:
+            from peft.tuners.lora.model import LoraModel as _LoraModel
+
+            _original_car = _LoraModel._create_and_replace
+
+            def _patched_car(
+                self,
+                peft_config,
+                adapter_name,
+                target,
+                target_name,
+                parent,
+                current_key = None,
+                **kwargs,
+            ):
+                if isinstance(target, _clippable_linear_cls):
+                    return _original_car(
+                        self,
+                        peft_config,
+                        adapter_name,
+                        target.linear,
+                        "linear",
+                        target,
+                        current_key = current_key,
+                        **kwargs,
+                    )
+                return _original_car(
+                    self,
+                    peft_config,
+                    adapter_name,
+                    target,
+                    target_name,
+                    parent,
+                    current_key = current_key,
+                    **kwargs,
+                )
+
+            _LoraModel._create_and_replace = _patched_car
+
         model = _get_peft_model(model, lora_config)
+
+        # Restore original PEFT method
+        if _clippable_linear_cls is not None:
+            _LoraModel._create_and_replace = _original_car
         # Apply QAT + LoRA if specified
         if qat_scheme is not None:
             print("Unsloth: Applying QAT to mitigate quantization degradation")
@@ -1383,7 +1448,7 @@ class FastBaseModel:
         # after this point, so we intercept gradient_checkpointing_enable
         # to always force use_reentrant=True for Gemma3N.
         _model_type = getattr(getattr(model, "config", None), "model_type", "") or ""
-        if "gemma3n" in _model_type.lower():
+        if "gemma3n" in _model_type.lower() or "gemma4" in _model_type.lower():
             _original_gc_enable = model.gradient_checkpointing_enable
 
             def _gc_enable_reentrant(**kwargs):
diff --git a/unsloth/ollama_template_mappers.py b/unsloth/ollama_template_mappers.py
index 1bf77461d9..728b08813a 100644
--- a/unsloth/ollama_template_mappers.py
+++ b/unsloth/ollama_template_mappers.py
@@ -1199,6 +1199,21 @@ TEMPLATE """{{- range $i, $_ := .Messages }}
 OLLAMA_TEMPLATES["gemma-3n"] = gemma3n_ollama
 OLLAMA_TEMPLATES["gemma3n"] = gemma3n_ollama
 
+# =========================================== Gemma-4
+gemma4_ollama = '''
+FROM {__FILE_LOCATION__}
+TEMPLATE """{{- range $i, $_ := .Messages }}
+{{- $last := eq (len (slice $.Messages $i)) 1 }}
+<|turn>{{ .Role }}
+{{ .Content }}{{ if not $last }}<turn|>
+{{ end }}
+{{- end }}<turn|>
+<|turn>model
+"""
+'''
+OLLAMA_TEMPLATES["gemma-4"] = gemma4_ollama
+OLLAMA_TEMPLATES["gemma4"] = gemma4_ollama
+
 # =========================================== GPT-OSS
 
 # Ollama from https://ollama.com/library/gpt-oss:latest/blobs/fa6710a93d78
@@ -1961,6 +1976,16 @@ OLLAMA_TEMPLATE_TO_MODEL_MAPPER = {
         "google/medgemma-27b-text-it",
         "unsloth/medgemma-27b-text-it-bnb-4bit",
     ),
+    "gemma4": (
+        "unsloth/gemma-4-E2B-it",
+        "unsloth/gemma-4-E2B",
+        "unsloth/gemma-4-E4B-it",
+        "unsloth/gemma-4-E4B",
+        "unsloth/gemma-4-31B-it",
+        "unsloth/gemma-4-31B",
+        "unsloth/gemma-4-26B-A4B-it",
+        "unsloth/gemma-4-26B-A4B",
+    ),
     "gemma3n": (
         "unsloth/gemma-3n-E4B-it-unsloth-bnb-4bit",
         "unsloth/gemma-3n-E4B-it",

From 4f9986ecb991c51b26c206ffcf33f9ff23291d7b Mon Sep 17 00:00:00 2001
From: Daniel Han <danielhanchen@gmail.com>
Date: Thu, 2 Apr 2026 08:59:02 -0700
Subject: [PATCH 014/216] fix(studio): improve tool-calling re-prompt for small
 models (#4783)

Small GGUF models (<9B) frequently generate full code or lengthy
explanations instead of calling tools, bypassing the existing
plan-without-action re-prompt mechanism. Three issues:

1. _REPROMPT_MAX_CHARS=500 was too low -- models that output full
   HTML/code responses (often 1000+ chars) never triggered the
   re-prompt at all, since it only fires on short responses.

2. _MAX_REPROMPTS=1 gave the model only one chance to comply.
   Small models often need 2-3 nudges before switching from
   text generation to tool calling.

3. The re-prompt text ("Please use the available tools...") was
   too polite for small models to follow reliably.

4. Tool-calling detection missed chat templates using Jinja
   whitespace-trimming syntax ({%- if tools -%}) since only
   ({%- if tools %}) and ({% if tools %}) were checked.

Changes:
- Raise _REPROMPT_MAX_CHARS from 500 to 2000 so longer responses
  (code blocks, multi-paragraph plans) still trigger re-prompts
- Raise _MAX_REPROMPTS from 1 to 3 for more retry budget
- Use direct, imperative re-prompt language that small models
  follow more reliably ("STOP. You MUST call a tool NOW.")
- Strengthen the system prompt tool nudge to explicitly forbid
  outputting code blocks (redirect to the python tool instead)
- Add Jinja whitespace-trimmed variants to the tool_markers
  list so all template styles are detected correctly
---
 studio/backend/core/inference/llama_cpp.py | 9 +++++----
 studio/backend/routes/inference.py         | 4 +++-
 2 files changed, 8 insertions(+), 5 deletions(-)

diff --git a/studio/backend/core/inference/llama_cpp.py b/studio/backend/core/inference/llama_cpp.py
index b436f4dc6d..44c700bf3d 100644
--- a/studio/backend/core/inference/llama_cpp.py
+++ b/studio/backend/core/inference/llama_cpp.py
@@ -45,8 +45,8 @@ _INTENT_SIGNAL = re.compile(
     r"\b(?:now i|next i)\b"
     r")"
 )
-_MAX_REPROMPTS = 1
-_REPROMPT_MAX_CHARS = 500
+_MAX_REPROMPTS = 3
+_REPROMPT_MAX_CHARS = 2000
 
 # ── Pre-compiled patterns for GGUF shard detection ───────────
 _SHARD_FULL_RE = re.compile(r"^(.*)-(\d{5})-of-(\d{5})\.gguf$")
@@ -2659,8 +2659,9 @@ class LlamaCppBackend:
                                 {
                                     "role": "user",
                                     "content": (
-                                        "Please use the available tools to complete "
-                                        "the task instead of describing what to do."
+                                        "STOP. Do NOT write code or explain. "
+                                        "You MUST call a tool NOW. "
+                                        "Call web_search or python immediately."
                                     ),
                                 }
                             )
diff --git a/studio/backend/routes/inference.py b/studio/backend/routes/inference.py
index 8f058bd0f2..ced24c1d5f 100644
--- a/studio/backend/routes/inference.py
+++ b/studio/backend/routes/inference.py
@@ -96,8 +96,10 @@ router = APIRouter()
 
 # Appended to tool-use nudge to discourage plan-without-action
 _TOOL_ACTION_NUDGE = (
-    " Always call tools directly."
+    " IMPORTANT: Always call tools directly -- never write code yourself."
     " Never describe what you plan to do -- just call the tool immediately."
+    " For any code request, call the python tool. For any factual question, call web_search."
+    " Do NOT output code blocks -- use the python tool instead."
 )
 
 # Regex for stripping leaked tool-call XML from assistant messages/stream

From f1c3b9caa94e4eebe0218ee4d80436dd6e9cb92a Mon Sep 17 00:00:00 2001
From: Daniel Han <danielhanchen@gmail.com>
Date: Thu, 2 Apr 2026 08:59:21 -0700
Subject: [PATCH 015/216] Pin Gemma-4 transformers requirement to 5.5.0 stable
 (#4784)

Gemma-4 support landed in transformers main
(huggingface/transformers#45192). Update the version pin from
5.5.0.dev0 to 5.5.0 across loader, Studio version switcher,
and the MLX installer. Also fix install_gemma4_mlx.sh which
referenced a non-existent v5.5-release branch -- pin it to
the correct commit (91b1ab1) instead.
---
 install_gemma4_mlx.sh                        | 4 ++--
 studio/backend/utils/transformers_version.py | 2 +-
 unsloth/models/loader.py                     | 2 +-
 3 files changed, 4 insertions(+), 4 deletions(-)

diff --git a/install_gemma4_mlx.sh b/install_gemma4_mlx.sh
index 1f5e7c2d9e..71575846a5 100755
--- a/install_gemma4_mlx.sh
+++ b/install_gemma4_mlx.sh
@@ -132,8 +132,8 @@ step "install" "installing mlx, mlx-lm..."
 uv pip install --python "$_VENV_PY" -q mlx mlx-lm 2>/dev/null
 substep "done"
 
-TRANSFORMERS_WHL="transformers-5.5.0.dev0-py3-none-any.whl"
-TRANSFORMERS_GH="git+https://github.com/huggingface/transformers.git@v5.5-release"
+TRANSFORMERS_WHL="transformers-5.5.0-py3-none-any.whl"
+TRANSFORMERS_GH="git+https://github.com/huggingface/transformers.git@91b1ab1fdfa81a552644a92fbe3e8d88de40e167"
 
 step "install" "installing transformers>=5.5.0..."
 if uv pip install --python "$_VENV_PY" -q "$TRANSFORMERS_GH" 2>/dev/null; then
diff --git a/studio/backend/utils/transformers_version.py b/studio/backend/utils/transformers_version.py
index 9db4b3ce4b..07e4a5c000 100644
--- a/studio/backend/utils/transformers_version.py
+++ b/studio/backend/utils/transformers_version.py
@@ -60,7 +60,7 @@ _TRANSFORMERS_5_TOKENIZER_CLASSES: set[str] = {
 _tokenizer_class_cache: dict[str, bool] = {}
 
 # Versions
-TRANSFORMERS_5_VERSION = "5.5.0.dev0"
+TRANSFORMERS_5_VERSION = "5.5.0"
 TRANSFORMERS_DEFAULT_VERSION = "4.57.6"
 
 # Pre-installed directory for transformers 5.x — created by setup.sh / setup.ps1
diff --git a/unsloth/models/loader.py b/unsloth/models/loader.py
index 297da73183..a811d3fb75 100644
--- a/unsloth/models/loader.py
+++ b/unsloth/models/loader.py
@@ -78,7 +78,7 @@ SUPPORTS_QWEN3_MOE = transformers_version >= Version("4.50.3")
 SUPPORTS_FALCON_H1 = transformers_version >= Version("4.53.0")
 SUPPORTS_GEMMA3N = transformers_version >= Version("4.53.0")
 SUPPORTS_GPTOSS = transformers_version >= Version("4.55.0")
-SUPPORTS_GEMMA4 = transformers_version >= Version("5.5.0.dev0")
+SUPPORTS_GEMMA4 = transformers_version >= Version("5.5.0")
 # Transformers v5 meta-device loading corrupts non-persistent buffers (inv_freq).
 # See _fix_rope_inv_freq() below for details.
 _NEEDS_ROPE_FIX = transformers_version >= Version("5.0.0")

From a3535572499194909386318322ce18edabcdf0c1 Mon Sep 17 00:00:00 2001
From: Daniel Han <danielhanchen@gmail.com>
Date: Thu, 2 Apr 2026 09:03:00 -0700
Subject: [PATCH 016/216] Force llama.cpp to always use mainline ggml-org
 (#4785)

Hardcode the release repo to ggml-org/llama.cpp and remove the
UNSLOTH_LLAMA_RELEASE_REPO and UNSLOTH_LLAMA_SOURCE env var overrides
so that all users always build/download from mainline llama.cpp.
---
 studio/install_llama_prebuilt.py | 6 +++---
 studio/setup.ps1                 | 8 ++++++--
 studio/setup.sh                  | 8 ++++++--
 3 files changed, 15 insertions(+), 7 deletions(-)

diff --git a/studio/install_llama_prebuilt.py b/studio/install_llama_prebuilt.py
index 76d22984d7..898606514e 100755
--- a/studio/install_llama_prebuilt.py
+++ b/studio/install_llama_prebuilt.py
@@ -60,9 +60,9 @@ def env_int(name: str, default: int, *, minimum: int | None = None) -> int:
 
 
 DEFAULT_LLAMA_TAG = os.environ.get("UNSLOTH_LLAMA_TAG", "latest")
-DEFAULT_PUBLISHED_REPO = os.environ.get(
-    "UNSLOTH_LLAMA_RELEASE_REPO", "unslothai/llama.cpp"
-)
+# Force all installs to use mainline llama.cpp from ggml-org.
+# Previously: DEFAULT_PUBLISHED_REPO = os.environ.get("UNSLOTH_LLAMA_RELEASE_REPO", "unslothai/llama.cpp")
+DEFAULT_PUBLISHED_REPO = "ggml-org/llama.cpp"
 DEFAULT_PUBLISHED_TAG = os.environ.get("UNSLOTH_LLAMA_RELEASE_TAG")
 DEFAULT_PUBLISHED_MANIFEST_ASSET = os.environ.get(
     "UNSLOTH_LLAMA_RELEASE_MANIFEST_ASSET", "llama-prebuilt-manifest.json"
diff --git a/studio/setup.ps1 b/studio/setup.ps1
index 1784b66ee9..5e29d1876a 100644
--- a/studio/setup.ps1
+++ b/studio/setup.ps1
@@ -1598,11 +1598,15 @@ $LlamaCppDir = Join-Path $UnslothHome "llama.cpp"
 $NeedLlamaSourceBuild = $false
 $SkipPrebuiltInstall = $false
 $RequestedLlamaTag = if ($env:UNSLOTH_LLAMA_TAG) { $env:UNSLOTH_LLAMA_TAG } else { "latest" }
-$HelperReleaseRepo = if ($env:UNSLOTH_LLAMA_RELEASE_REPO) { $env:UNSLOTH_LLAMA_RELEASE_REPO } else { "unslothai/llama.cpp" }
+# Force all installs to use mainline llama.cpp from ggml-org.
+# Previously: $HelperReleaseRepo = if ($env:UNSLOTH_LLAMA_RELEASE_REPO) { $env:UNSLOTH_LLAMA_RELEASE_REPO } else { "unslothai/llama.cpp" }
+$HelperReleaseRepo = "ggml-org/llama.cpp"
 $LlamaPr = if ($env:UNSLOTH_LLAMA_PR) { $env:UNSLOTH_LLAMA_PR.Trim() } else { "" }
 
 $LlamaPrForce = if ($env:UNSLOTH_LLAMA_PR_FORCE) { $env:UNSLOTH_LLAMA_PR_FORCE.Trim() } else { $DefaultLlamaPrForce }
-$LlamaSource = if ($env:UNSLOTH_LLAMA_SOURCE) { $env:UNSLOTH_LLAMA_SOURCE.Trim() } else { $DefaultLlamaSource }
+# Force mainline source -- no env var override.
+# Previously: $LlamaSource = if ($env:UNSLOTH_LLAMA_SOURCE) { $env:UNSLOTH_LLAMA_SOURCE.Trim() } else { $DefaultLlamaSource }
+$LlamaSource = $DefaultLlamaSource
 if ($LlamaSource.EndsWith('.git')) { $LlamaSource = $LlamaSource.Substring(0, $LlamaSource.Length - 4) }
 $ResolvedSourceUrl = $LlamaSource
 $ResolvedSourceRef = $RequestedLlamaTag
diff --git a/studio/setup.sh b/studio/setup.sh
index 26903b120a..39b186ab09 100755
--- a/studio/setup.sh
+++ b/studio/setup.sh
@@ -479,11 +479,15 @@ _NEED_LLAMA_SOURCE_BUILD=false
 _LLAMA_CPP_DEGRADED=false
 _LLAMA_FORCE_COMPILE="${UNSLOTH_LLAMA_FORCE_COMPILE:-0}"
 _REQUESTED_LLAMA_TAG="${UNSLOTH_LLAMA_TAG:-latest}"
-_HELPER_RELEASE_REPO="${UNSLOTH_LLAMA_RELEASE_REPO:-unslothai/llama.cpp}"
+# Force all installs to use mainline llama.cpp from ggml-org.
+# Previously: _HELPER_RELEASE_REPO="${UNSLOTH_LLAMA_RELEASE_REPO:-unslothai/llama.cpp}"
+_HELPER_RELEASE_REPO="ggml-org/llama.cpp"
 _LLAMA_PR="${UNSLOTH_LLAMA_PR:-}"
 
 _LLAMA_PR_FORCE="${UNSLOTH_LLAMA_PR_FORCE:-${_DEFAULT_LLAMA_PR_FORCE}}"
-_LLAMA_SOURCE="${UNSLOTH_LLAMA_SOURCE:-${_DEFAULT_LLAMA_SOURCE}}"
+# Force mainline source -- no env var override.
+# Previously: _LLAMA_SOURCE="${UNSLOTH_LLAMA_SOURCE:-${_DEFAULT_LLAMA_SOURCE}}"
+_LLAMA_SOURCE="${_DEFAULT_LLAMA_SOURCE}"
 _LLAMA_SOURCE="${_LLAMA_SOURCE%.git}"  # normalize: strip trailing .git
 _RESOLVED_SOURCE_URL="$_LLAMA_SOURCE"
 _RESOLVED_SOURCE_REF="$_REQUESTED_LLAMA_TAG"

From a241c58d841fed0c697e0239fce863a87423c6d2 Mon Sep 17 00:00:00 2001
From: Daniel Han <danielhanchen@gmail.com>
Date: Thu, 2 Apr 2026 09:10:02 -0700
Subject: [PATCH 017/216] Use transformers v5.5-release branch and pin to 5.5.0
 (#4786)

The v5.5-release branch now exists on huggingface/transformers.
Use transformers==5.5.0 for all install paths and
git+transformers.git@v5.5-release for the MLX installer.

Also bumps huggingface_hub from 1.7.1 to 1.8.0 in setup.sh and
setup.ps1 to stay consistent.
---
 install_gemma4_mlx.sh | 2 +-
 studio/setup.ps1      | 2 +-
 studio/setup.sh       | 4 ++--
 3 files changed, 4 insertions(+), 4 deletions(-)

diff --git a/install_gemma4_mlx.sh b/install_gemma4_mlx.sh
index 71575846a5..128b0d5bca 100755
--- a/install_gemma4_mlx.sh
+++ b/install_gemma4_mlx.sh
@@ -133,7 +133,7 @@ uv pip install --python "$_VENV_PY" -q mlx mlx-lm 2>/dev/null
 substep "done"
 
 TRANSFORMERS_WHL="transformers-5.5.0-py3-none-any.whl"
-TRANSFORMERS_GH="git+https://github.com/huggingface/transformers.git@91b1ab1fdfa81a552644a92fbe3e8d88de40e167"
+TRANSFORMERS_GH="git+https://github.com/huggingface/transformers.git@v5.5-release"
 
 step "install" "installing transformers>=5.5.0..."
 if uv pip install --python "$_VENV_PY" -q "$TRANSFORMERS_GH" 2>/dev/null; then
diff --git a/studio/setup.ps1 b/studio/setup.ps1
index 5e29d1876a..130b66c7df 100644
--- a/studio/setup.ps1
+++ b/studio/setup.ps1
@@ -1551,7 +1551,7 @@ if (Test-Path $VenvT5Dir) { Remove-Item -Recurse -Force $VenvT5Dir }
 New-Item -ItemType Directory -Path $VenvT5Dir -Force | Out-Null
 $prevEAP_t5 = $ErrorActionPreference
 $ErrorActionPreference = "Continue"
-foreach ($pkg in @("transformers==5.3.0", "huggingface_hub==1.7.1", "hf_xet==1.4.2")) {
+foreach ($pkg in @("transformers==5.5.0", "huggingface_hub==1.8.0", "hf_xet==1.4.2")) {
     if ($script:UnslothVerbose) {
         Fast-Install --target $VenvT5Dir --no-deps $pkg
         $t5PkgExit = $LASTEXITCODE
diff --git a/studio/setup.sh b/studio/setup.sh
index 39b186ab09..9096c3bb19 100755
--- a/studio/setup.sh
+++ b/studio/setup.sh
@@ -459,8 +459,8 @@ if [ "$_SKIP_PYTHON_DEPS" = false ]; then
     # at runtime (slow, ~10-15s), we pre-install into a separate directory.
     # The training subprocess just prepends .venv_t5/ to sys.path -- instant switch.
     mkdir -p "$VENV_T5_DIR"
-    run_quiet "install transformers 5.x" fast_install --target "$VENV_T5_DIR" --no-deps "transformers==5.3.0"
-    run_quiet "install huggingface_hub for t5" fast_install --target "$VENV_T5_DIR" --no-deps "huggingface_hub==1.7.1"
+    run_quiet "install transformers 5.x" fast_install --target "$VENV_T5_DIR" --no-deps "transformers==5.5.0"
+    run_quiet "install huggingface_hub for t5" fast_install --target "$VENV_T5_DIR" --no-deps "huggingface_hub==1.8.0"
     run_quiet "install hf_xet for t5" fast_install --target "$VENV_T5_DIR" --no-deps "hf_xet==1.4.2"
     run_quiet "install tiktoken for t5" fast_install --target "$VENV_T5_DIR" "tiktoken"
     step "transformers" "5.x pre-installed"

From 2af53bf9a65d1b96a8e10c8d67feaefd1a2c67d7 Mon Sep 17 00:00:00 2001
From: Daniel Han <danielhanchen@gmail.com>
Date: Thu, 2 Apr 2026 09:21:30 -0700
Subject: [PATCH 018/216] Pin transformers and huggingface-hub in main Studio
 venv (#4788)

Revert the >= loosening from f9c4b08 back to exact pins.
Using transformers>=4.57.6 allows pip to install 5.x into the main
Studio venv, which breaks huggingface_hub imports
(is_offline_mode removed in newer hub versions).

The main venv must stay on transformers==4.57.6 and
huggingface-hub==0.36.2. The 5.x version lives only in .venv_t5/
and is dynamically switched via sys.path at runtime.
---
 studio/backend/requirements/extras-no-deps.txt         | 2 +-
 studio/backend/requirements/single-env/constraints.txt | 4 ++--
 2 files changed, 3 insertions(+), 3 deletions(-)

diff --git a/studio/backend/requirements/extras-no-deps.txt b/studio/backend/requirements/extras-no-deps.txt
index 325eb494fc..9934bacd24 100644
--- a/studio/backend/requirements/extras-no-deps.txt
+++ b/studio/backend/requirements/extras-no-deps.txt
@@ -11,6 +11,6 @@ git+https://github.com/meta-pytorch/OpenEnv.git
 # executorch>=1.0.1               # 41.5 MB - no imports in unsloth/zoo/studio
 torch-c-dlpack-ext
 sentence_transformers==5.2.0
-transformers>=4.57.6
+transformers==4.57.6
 pytorch_tokenizers
 kernels
diff --git a/studio/backend/requirements/single-env/constraints.txt b/studio/backend/requirements/single-env/constraints.txt
index 37391e739e..156f78567e 100644
--- a/studio/backend/requirements/single-env/constraints.txt
+++ b/studio/backend/requirements/single-env/constraints.txt
@@ -1,8 +1,8 @@
 # Single-env pins for unsloth + studio + data-designer
 # Keep compatible with unsloth transformers bounds.
-transformers>=4.57.6
+transformers==4.57.6
 trl==0.23.1
-huggingface-hub>=0.36.2
+huggingface-hub==0.36.2
 
 # Studio stack
 datasets==4.3.0

From 1ce83c40aa6087445703d009d904d793b76b2c30 Mon Sep 17 00:00:00 2001
From: Daniel Han <danielhanchen@gmail.com>
Date: Thu, 2 Apr 2026 09:45:56 -0700
Subject: [PATCH 019/216] fix(studio): build llama.cpp from master instead of
 latest release tag (#4790)

The latest ggml-org/llama.cpp release (b8635) does not include Gemma 4
support (ggml-org/llama.cpp#21309 merged after the release was cut).
This causes `llama-server` to fail with "unknown model architecture:
gemma4" when loading Gemma 4 GGUFs.

Temporarily default _DEFAULT_LLAMA_TAG to "master" so all new installs
build from the llama.cpp master branch which includes Gemma 4 support.
Once a new upstream release is cut with Gemma 4, this can be reverted
back to "latest".

Changes:
- setup.sh: add _DEFAULT_LLAMA_TAG="master" maintainer default
- setup.ps1: add $DefaultLlamaTag="master" maintainer default
- install_llama_prebuilt.py: change DEFAULT_LLAMA_TAG fallback to "master"

Users can still override via UNSLOTH_LLAMA_TAG env var.
---
 studio/install_llama_prebuilt.py | 2 +-
 studio/setup.ps1                 | 3 ++-
 studio/setup.sh                  | 5 ++++-
 3 files changed, 7 insertions(+), 3 deletions(-)

diff --git a/studio/install_llama_prebuilt.py b/studio/install_llama_prebuilt.py
index 898606514e..61e57614a8 100755
--- a/studio/install_llama_prebuilt.py
+++ b/studio/install_llama_prebuilt.py
@@ -59,7 +59,7 @@ def env_int(name: str, default: int, *, minimum: int | None = None) -> int:
     return value
 
 
-DEFAULT_LLAMA_TAG = os.environ.get("UNSLOTH_LLAMA_TAG", "latest")
+DEFAULT_LLAMA_TAG = os.environ.get("UNSLOTH_LLAMA_TAG", "master")
 # Force all installs to use mainline llama.cpp from ggml-org.
 # Previously: DEFAULT_PUBLISHED_REPO = os.environ.get("UNSLOTH_LLAMA_RELEASE_REPO", "unslothai/llama.cpp")
 DEFAULT_PUBLISHED_REPO = "ggml-org/llama.cpp"
diff --git a/studio/setup.ps1 b/studio/setup.ps1
index 130b66c7df..396a04486f 100644
--- a/studio/setup.ps1
+++ b/studio/setup.ps1
@@ -29,6 +29,7 @@ $PackageDir = Split-Path -Parent $ScriptDir
 # --------------------------------------------------------------------------
 $DefaultLlamaPrForce = ""
 $DefaultLlamaSource = "https://github.com/ggml-org/llama.cpp"
+$DefaultLlamaTag = "master"
 
 # Verbose can be enabled either by CLI flag or by UNSLOTH_VERBOSE=1.
 $script:UnslothVerbose = ($env:UNSLOTH_VERBOSE -eq '1')
@@ -1597,7 +1598,7 @@ if (-not (Test-Path $UnslothHome)) { New-Item -ItemType Directory -Force $Unslot
 $LlamaCppDir = Join-Path $UnslothHome "llama.cpp"
 $NeedLlamaSourceBuild = $false
 $SkipPrebuiltInstall = $false
-$RequestedLlamaTag = if ($env:UNSLOTH_LLAMA_TAG) { $env:UNSLOTH_LLAMA_TAG } else { "latest" }
+$RequestedLlamaTag = if ($env:UNSLOTH_LLAMA_TAG) { $env:UNSLOTH_LLAMA_TAG } else { $DefaultLlamaTag }
 # Force all installs to use mainline llama.cpp from ggml-org.
 # Previously: $HelperReleaseRepo = if ($env:UNSLOTH_LLAMA_RELEASE_REPO) { $env:UNSLOTH_LLAMA_RELEASE_REPO } else { "unslothai/llama.cpp" }
 $HelperReleaseRepo = "ggml-org/llama.cpp"
diff --git a/studio/setup.sh b/studio/setup.sh
index 9096c3bb19..116dc12e08 100755
--- a/studio/setup.sh
+++ b/studio/setup.sh
@@ -14,9 +14,12 @@ RULE=$(printf '\342\224\200%.0s' {1..52})
 #
 #   _DEFAULT_LLAMA_PR_FORCE : PR number to build by default ("" = normal path)
 #   _DEFAULT_LLAMA_SOURCE   : git clone URL for source builds
+#   _DEFAULT_LLAMA_TAG      : llama.cpp ref to build ("latest" = newest release,
+#                             "master" = bleeding-edge, "bNNNN" = specific tag)
 # ──────────────────────────────────────────────────────────────────────────
 _DEFAULT_LLAMA_PR_FORCE=""
 _DEFAULT_LLAMA_SOURCE="https://github.com/ggml-org/llama.cpp"
+_DEFAULT_LLAMA_TAG="master"
 
 # ── Colors (same palette as startup_banner / install_python_stack) ──
 if [ -n "${NO_COLOR:-}" ]; then
@@ -478,7 +481,7 @@ LLAMA_SERVER_BIN="$LLAMA_CPP_DIR/build/bin/llama-server"
 _NEED_LLAMA_SOURCE_BUILD=false
 _LLAMA_CPP_DEGRADED=false
 _LLAMA_FORCE_COMPILE="${UNSLOTH_LLAMA_FORCE_COMPILE:-0}"
-_REQUESTED_LLAMA_TAG="${UNSLOTH_LLAMA_TAG:-latest}"
+_REQUESTED_LLAMA_TAG="${UNSLOTH_LLAMA_TAG:-${_DEFAULT_LLAMA_TAG}}"
 # Force all installs to use mainline llama.cpp from ggml-org.
 # Previously: _HELPER_RELEASE_REPO="${UNSLOTH_LLAMA_RELEASE_REPO:-unslothai/llama.cpp}"
 _HELPER_RELEASE_REPO="ggml-org/llama.cpp"

From 0446d4668957827e7af1aa8f6719df720d9b6457 Mon Sep 17 00:00:00 2001
From: Roland Tannous <115670425+rolandtannous@users.noreply.github.com>
Date: Thu, 2 Apr 2026 21:04:42 +0400
Subject: [PATCH 020/216] fixed name (#4791)

---
 install_gemma4_mlx.sh | 8 ++++----
 1 file changed, 4 insertions(+), 4 deletions(-)

diff --git a/install_gemma4_mlx.sh b/install_gemma4_mlx.sh
index 128b0d5bca..e06339e204 100755
--- a/install_gemma4_mlx.sh
+++ b/install_gemma4_mlx.sh
@@ -198,15 +198,15 @@ echo ""
 printf "  ${C_TITLE}%s${C_RST}\n" "Gemma 4 MLX installed!"
 printf "  ${C_DIM}%s${C_RST}\n" "$RULE"
 echo ""
-step "available models" "unsloth/gemma-4-E2B-it-MLX-4bit (/BF16)"
-substep "unsloth/gemma-4-E4B-it-MLX-4bit (/BF16)"
+step "available models" "unsloth/gemma-4-E2B-it-UD-MLX-4bit (/BF16)"
+substep "unsloth/gemma-4-E4B-it-UD-MLX-4bit (/BF16)"
 echo ""
 step "venv activate" "source ${VENV_DIR}/bin/activate"
 echo ""
-step "quick start" "python -m mlx_lm chat --model unsloth/gemma-4-E2B-it-MLX-4bit --max-tokens 200"
+step "quick start" "python -m mlx_lm chat --model unsloth/gemma-4-E2B-it-UD-MLX-4bit --max-tokens 200"
 echo ""
 step "python API" "from mlx_lm import load, generate"
-substep "model, tokenizer = load('unsloth/gemma-4-E2B-it-MLX-4bit')"
+substep "model, tokenizer = load('unsloth/gemma-4-E2B-it-UD-MLX-4bit')"
 substep "messages = [{'role': 'user', 'content': 'Hello!'}]"
 substep "prompt = tokenizer.apply_chat_template(messages, tokenize=False, add_generation_prompt=True)"
 substep "print(generate(model, tokenizer, prompt=prompt, max_tokens=200))"

From 7023e2a4ff0a8fa968aed2f55a0abd81a28feb55 Mon Sep 17 00:00:00 2001
From: Daniel Han <danielhanchen@gmail.com>
Date: Thu, 2 Apr 2026 10:46:53 -0700
Subject: [PATCH 021/216] fix(studio): prioritize curated defaults over HF
 download ranking in Recommended (#4792)

The model list merge order was `top_gguf + top_hub + static_models`,
which meant the HF download-ranked models always came first. New models
like Gemma 4 have low download counts and were not in the HF top-40,
so they got buried after 80 other models despite being at the top of
the curated static defaults in defaults.py.

Flip the merge to `static_models + top_gguf + top_hub` so editorial
picks (new model launches, promoted models) always appear first in the
Recommended section, with HF popularity backfilling after.

Co-authored-by: Daniel Han <danielhanchen@users.noreply.github.com>
---
 studio/backend/core/inference/orchestrator.py | 5 +++--
 1 file changed, 3 insertions(+), 2 deletions(-)

diff --git a/studio/backend/core/inference/orchestrator.py b/studio/backend/core/inference/orchestrator.py
index f293a0dcd8..cb5d9da34a 100644
--- a/studio/backend/core/inference/orchestrator.py
+++ b/studio/backend/core/inference/orchestrator.py
@@ -109,12 +109,13 @@ class InferenceOrchestrator:
         self._top_models_ready.wait(timeout = 5)
         top_gguf = self._top_gguf_cache or []
         top_hub = self._top_hub_cache or []
-        # GGUFs first, then hub models, then static fallbacks.
+        # Curated static defaults first (editorial picks like new models),
+        # then HF download-ranked models to backfill.
         # Send extras so the frontend still has 4 per category
         # after removing already-downloaded models.
         result: list[str] = []
         seen: set[str] = set()
-        for m in top_gguf + top_hub + self._static_models:
+        for m in self._static_models + top_gguf + top_hub:
             if m not in seen:
                 result.append(m)
                 seen.add(m)

From 7ae9b7f45f966abf5a8dde752509015e5ace3931 Mon Sep 17 00:00:00 2001
From: DoubleMathew <mmathew23@gmail.com>
Date: Thu, 2 Apr 2026 13:43:46 -0500
Subject: [PATCH 022/216] fix windows llama.cpp compile from source issue
 (#4793)

* fix windows llama.cpp compile from source issue

* undo local repo usage

* fix llama.cpp install

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

* fix windows

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

* fix: route resolve-source-build call through Invoke-LlamaHelper

The --resolve-source-build call at the source-build resolution path
was still calling install_llama_prebuilt.py directly instead of going
through Invoke-LlamaHelper. On PS7+ with ErrorActionPreference=Stop,
stderr from the 422 response (when tag is "master") would trigger a
terminating NativeCommandError and crash setup.

* fix: suppress stderr error records from Invoke-LlamaHelper

ErrorActionPreference=Continue prevents termination but PowerShell
still displays stderr lines as visible ErrorRecord objects. Capture
all output via 2>&1 and split stdout from stderr manually so that
stderr lines never appear on the console. When StderrPath is given
the stderr content is written to that file for diagnostics.

* fix: always rebuild llama.cpp on Windows when tag is master

When the requested llama.cpp tag is "master" (a moving target), skip
the "already built" early exit so the build path runs and syncs to
the latest commit. Without this, existing llama-server binaries from
an older build (e.g. b8635 which lacks Gemma 4 support) are reused
and model loading fails.

Pinned tags (e.g. b8635) still skip the rebuild when the binary
already exists, since the tag is immutable.

---------

Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com>
Co-authored-by: Daniel Han <danielhanchen@users.noreply.github.com>
---
 studio/install_llama_prebuilt.py              |  46 ++++++-
 studio/setup.ps1                              | 125 +++++++++++++----
 studio/setup.sh                               |  13 +-
 .../install/test_llama_pr_force_and_source.py | 129 +++++++++---------
 tests/studio/install/test_pr4562_bugfixes.py  |  28 ++++
 tests/studio/install/test_selection_logic.py  |  50 +++++++
 6 files changed, 296 insertions(+), 95 deletions(-)

diff --git a/studio/install_llama_prebuilt.py b/studio/install_llama_prebuilt.py
index 61e57614a8..13382d3c16 100755
--- a/studio/install_llama_prebuilt.py
+++ b/studio/install_llama_prebuilt.py
@@ -14,6 +14,7 @@ import json
 import os
 import platform
 import random
+import re
 import shutil
 import site
 import socket
@@ -561,6 +562,33 @@ def windows_cuda_upstream_asset_names(llama_tag: str, runtime: str) -> list[str]
     ]
 
 
+def windows_cuda_asset_aliases(
+    asset_name: str,
+    *,
+    compatibility_tag: str | None = None,
+) -> list[str]:
+    aliases: list[str] = []
+    legacy_match = re.fullmatch(
+        r"llama-(?P<tag>[^/]+)-bin-win-cuda-(?P<runtime>\d+\.\d+)-x64\.zip",
+        asset_name,
+    )
+    if legacy_match:
+        runtime = legacy_match.group("runtime")
+        aliases.append(f"cudart-llama-bin-win-cuda-{runtime}-x64.zip")
+        if compatibility_tag:
+            aliases.append(f"llama-{compatibility_tag}-bin-win-cuda-{runtime}-x64.zip")
+        return aliases
+
+    current_match = re.fullmatch(
+        r"cudart-llama-bin-win-cuda-(?P<runtime>\d+\.\d+)-x64\.zip",
+        asset_name,
+    )
+    if current_match and compatibility_tag:
+        runtime = current_match.group("runtime")
+        aliases.append(f"llama-{compatibility_tag}-bin-win-cuda-{runtime}-x64.zip")
+    return aliases
+
+
 def format_byte_count(num_bytes: float) -> str:
     units = ["B", "KiB", "MiB", "GiB", "TiB"]
     value = float(num_bytes)
@@ -3867,9 +3895,7 @@ def apply_approved_hashes(
     checksums: ApprovedReleaseChecksums,
 ) -> list[AssetChoice]:
     def approved_hash_for_attempt(attempt: AssetChoice) -> ApprovedArtifactHash | None:
-        approved = checksums.artifacts.get(attempt.name)
-        if approved is not None:
-            return approved
+        candidate_names = [attempt.name]
         if (
             isinstance(attempt.tag, str)
             and attempt.tag
@@ -3883,7 +3909,19 @@ def apply_approved_hashes(
                 if attempt.name.startswith(legacy_prefix)
                 else attempt.name
             )
-            approved = checksums.artifacts.get(compatibility_name)
+            candidate_names.append(compatibility_name)
+        candidate_names.extend(
+            windows_cuda_asset_aliases(
+                attempt.name,
+                compatibility_tag = checksums.upstream_tag,
+            )
+        )
+        seen_names: set[str] = set()
+        for candidate_name in candidate_names:
+            if candidate_name in seen_names:
+                continue
+            seen_names.add(candidate_name)
+            approved = checksums.artifacts.get(candidate_name)
             if approved is not None:
                 return approved
         return None
diff --git a/studio/setup.ps1 b/studio/setup.ps1
index 396a04486f..382de89b9f 100644
--- a/studio/setup.ps1
+++ b/studio/setup.ps1
@@ -71,6 +71,12 @@ function Refresh-Environment {
     $env:Path = "$machinePath;$userPath"
 }
 
+# PowerShell 5.1 compatibility helper: avoid relying on New-TemporaryFile.
+function New-UnslothTemporaryFile {
+    $tempPath = [System.IO.Path]::GetTempFileName()
+    return Get-Item -LiteralPath $tempPath
+}
+
 # Find nvcc on PATH, CUDA_PATH, or standard toolkit dirs.
 # Returns the path to nvcc.exe, or $null if not found.
 function Find-Nvcc {
@@ -1599,20 +1605,67 @@ $LlamaCppDir = Join-Path $UnslothHome "llama.cpp"
 $NeedLlamaSourceBuild = $false
 $SkipPrebuiltInstall = $false
 $RequestedLlamaTag = if ($env:UNSLOTH_LLAMA_TAG) { $env:UNSLOTH_LLAMA_TAG } else { $DefaultLlamaTag }
-# Force all installs to use mainline llama.cpp from ggml-org.
-# Previously: $HelperReleaseRepo = if ($env:UNSLOTH_LLAMA_RELEASE_REPO) { $env:UNSLOTH_LLAMA_RELEASE_REPO } else { "unslothai/llama.cpp" }
 $HelperReleaseRepo = "ggml-org/llama.cpp"
 $LlamaPr = if ($env:UNSLOTH_LLAMA_PR) { $env:UNSLOTH_LLAMA_PR.Trim() } else { "" }
 
 $LlamaPrForce = if ($env:UNSLOTH_LLAMA_PR_FORCE) { $env:UNSLOTH_LLAMA_PR_FORCE.Trim() } else { $DefaultLlamaPrForce }
-# Force mainline source -- no env var override.
-# Previously: $LlamaSource = if ($env:UNSLOTH_LLAMA_SOURCE) { $env:UNSLOTH_LLAMA_SOURCE.Trim() } else { $DefaultLlamaSource }
 $LlamaSource = $DefaultLlamaSource
 if ($LlamaSource.EndsWith('.git')) { $LlamaSource = $LlamaSource.Substring(0, $LlamaSource.Length - 4) }
 $ResolvedSourceUrl = $LlamaSource
 $ResolvedSourceRef = $RequestedLlamaTag
 $ResolvedSourceRefKind = "tag"
 
+if ($env:UNSLOTH_LLAMA_FORCE_COMPILE -eq "1") {
+    $NeedLlamaSourceBuild = $true
+    $SkipPrebuiltInstall = $true
+}
+
+function Invoke-LlamaHelper {
+    param(
+        [string[]]$Arguments,
+        [string]$StderrPath = $null
+    )
+
+    $previousErrorActionPreference = $ErrorActionPreference
+    $previousNativeErrorPreference = $null
+    $restoreNativeErrorPreference = $false
+    $ErrorActionPreference = "Continue"
+    if ($PSVersionTable.PSVersion.Major -ge 7) {
+        $previousNativeErrorPreference = $PSNativeCommandUseErrorActionPreference
+        $PSNativeCommandUseErrorActionPreference = $false
+        $restoreNativeErrorPreference = $true
+    }
+
+    try {
+        # Capture all output (stdout + stderr) so that PowerShell does not
+        # convert stderr lines into visible ErrorRecord objects.  Separate
+        # stdout from stderr afterwards.
+        $allOutput = & python "$PSScriptRoot\install_llama_prebuilt.py" @Arguments 2>&1
+        $exitCode = $LASTEXITCODE
+        $stdoutLines = @()
+        $stderrLines = @()
+        foreach ($line in $allOutput) {
+            if ($line -is [System.Management.Automation.ErrorRecord]) {
+                $stderrLines += $line.ToString()
+            } else {
+                $stdoutLines += $line
+            }
+        }
+        if ($StderrPath -and $stderrLines.Count -gt 0) {
+            $stderrLines | Out-File -FilePath $StderrPath -Encoding utf8
+        }
+        return [pscustomobject]@{
+            Output = $stdoutLines
+            ExitCode = $exitCode
+        }
+    } finally {
+        $ErrorActionPreference = $previousErrorActionPreference
+        if ($restoreNativeErrorPreference) {
+            $PSNativeCommandUseErrorActionPreference = $previousNativeErrorPreference
+        }
+    }
+}
+
 if ($LlamaSource -ne "https://github.com/ggml-org/llama.cpp") {
     step "llama.cpp" "custom source: $LlamaSource -- forcing source build" "Yellow"
     $NeedLlamaSourceBuild = $true
@@ -1640,11 +1693,14 @@ if ($LlamaPr) {
     # Custom source or other override already forced source build; skip the
     # prebuilt release resolution. When building from a custom fork, the fork
     # may not carry upstream bNNNN tags.
-    if ($LlamaSource -eq "https://github.com/ggml-org/llama.cpp") {
+    if ($env:UNSLOTH_LLAMA_FORCE_COMPILE -eq "1") {
+        $ResolvedLlamaTag = $RequestedLlamaTag
+    } elseif ($LlamaSource -eq "https://github.com/ggml-org/llama.cpp") {
         $resolveTagArgs = @("--resolve-llama-tag", $RequestedLlamaTag, "--published-repo", $HelperReleaseRepo, "--output-format", "json")
         if ($env:UNSLOTH_LLAMA_RELEASE_TAG) { $resolveTagArgs += @("--published-release-tag", $env:UNSLOTH_LLAMA_RELEASE_TAG) }
-        $fallbackOutput = & python "$PSScriptRoot\install_llama_prebuilt.py" @resolveTagArgs 2>$null
-        $fallbackExit = $LASTEXITCODE
+        $fallbackResult = Invoke-LlamaHelper -Arguments $resolveTagArgs
+        $fallbackOutput = $fallbackResult.Output
+        $fallbackExit = $fallbackResult.ExitCode
         $ResolvedLlamaTag = if ($fallbackExit -eq 0 -and $fallbackOutput) {
             try {
                 (($fallbackOutput | Out-String) | ConvertFrom-Json).llama_tag
@@ -1660,9 +1716,10 @@ if ($LlamaPr) {
 } else {
     $resolveInstallArgs = @("--resolve-install-tag", $RequestedLlamaTag, "--published-repo", $HelperReleaseRepo, "--output-format", "json")
     if ($env:UNSLOTH_LLAMA_RELEASE_TAG) { $resolveInstallArgs += @("--published-release-tag", $env:UNSLOTH_LLAMA_RELEASE_TAG) }
-    $resolveErrorLog = New-TemporaryFile
-    $resolveOutput = & python "$PSScriptRoot\install_llama_prebuilt.py" @resolveInstallArgs 2>$resolveErrorLog
-    $resolveExit = $LASTEXITCODE
+    $resolveErrorLog = New-UnslothTemporaryFile
+    $resolveResult = Invoke-LlamaHelper -Arguments $resolveInstallArgs -StderrPath $resolveErrorLog
+    $resolveOutput = $resolveResult.Output
+    $resolveExit = $resolveResult.ExitCode
     $ResolvedLlamaTag = if ($resolveOutput) {
         try {
             (($resolveOutput | Out-String) | ConvertFrom-Json).llama_tag
@@ -1679,8 +1736,9 @@ if ($LlamaPr) {
         # before falling back to the bleeding-edge ggml-org/llama.cpp tag.
         $resolveFallbackArgs = @("--resolve-llama-tag", $RequestedLlamaTag, "--published-repo", $HelperReleaseRepo, "--output-format", "json")
         if ($env:UNSLOTH_LLAMA_RELEASE_TAG) { $resolveFallbackArgs += @("--published-release-tag", $env:UNSLOTH_LLAMA_RELEASE_TAG) }
-        $fallbackOutput = & python "$PSScriptRoot\install_llama_prebuilt.py" @resolveFallbackArgs 2>$null
-        $fallbackExit = $LASTEXITCODE
+        $fallbackResult = Invoke-LlamaHelper -Arguments $resolveFallbackArgs
+        $fallbackOutput = $fallbackResult.Output
+        $fallbackExit = $fallbackResult.ExitCode
         $ResolvedLlamaTag = if ($fallbackExit -eq 0 -and $fallbackOutput) {
             try {
                 (($fallbackOutput | Out-String) | ConvertFrom-Json).llama_tag
@@ -1723,16 +1781,29 @@ if ($env:UNSLOTH_LLAMA_FORCE_COMPILE -eq "1") {
         }
         $prevEAPPrebuilt = $ErrorActionPreference
         $ErrorActionPreference = "Continue"
-        if ($script:UnslothVerbose) {
-            # Show live output in verbose mode while still capturing for error log
-            $prebuiltLog = Join-Path $env:TEMP "unsloth-prebuilt-$PID.log"
-            & python @prebuiltArgs 2>&1 | Tee-Object -FilePath $prebuiltLog | Out-Host
-            $prebuiltExit = $LASTEXITCODE
-            $prebuiltOutput = if (Test-Path $prebuiltLog) { Get-Content $prebuiltLog -Raw } else { "" }
-            Remove-Item $prebuiltLog -ErrorAction SilentlyContinue
-        } else {
-            $prebuiltOutput = & python @prebuiltArgs 2>&1 | Out-String
-            $prebuiltExit = $LASTEXITCODE
+        $previousNativeErrorPreference = $null
+        $restoreNativeErrorPreference = $false
+        if ($PSVersionTable.PSVersion.Major -ge 7) {
+            $previousNativeErrorPreference = $PSNativeCommandUseErrorActionPreference
+            $PSNativeCommandUseErrorActionPreference = $false
+            $restoreNativeErrorPreference = $true
+        }
+        try {
+            if ($script:UnslothVerbose) {
+                # Show live output in verbose mode while still capturing for error log
+                $prebuiltLog = Join-Path $env:TEMP "unsloth-prebuilt-$PID.log"
+                & python @prebuiltArgs 2>&1 | Tee-Object -FilePath $prebuiltLog | Out-Host
+                $prebuiltExit = $LASTEXITCODE
+                $prebuiltOutput = if (Test-Path $prebuiltLog) { Get-Content $prebuiltLog -Raw } else { "" }
+                Remove-Item $prebuiltLog -ErrorAction SilentlyContinue
+            } else {
+                $prebuiltOutput = & python @prebuiltArgs 2>&1 | Out-String
+                $prebuiltExit = $LASTEXITCODE
+            }
+        } finally {
+            if ($restoreNativeErrorPreference) {
+                $PSNativeCommandUseErrorActionPreference = $previousNativeErrorPreference
+            }
         }
         $ErrorActionPreference = $prevEAPPrebuilt
 
@@ -1847,7 +1918,10 @@ if (Test-Path $LlamaServerBin) {
 if (-not $NeedLlamaSourceBuild) {
     Write-Host ""
     step "llama.cpp" "prebuilt (validated)"
-} elseif ((Test-Path $LlamaServerBin) -and -not $NeedRebuild) {
+} elseif ((Test-Path $LlamaServerBin) -and -not $NeedRebuild -and $RequestedLlamaTag -ne "master") {
+    # Skip rebuild only for pinned tags (e.g. b8635).  When the requested
+    # tag is "master" (a moving target), always rebuild so the binary picks
+    # up new model architecture support (e.g. Gemma 4).
     Write-Host ""
     step "llama.cpp" "already built"
 } elseif (-not $HasCmakeForBuild) {
@@ -1906,8 +1980,9 @@ if (-not $NeedLlamaSourceBuild) {
         if ($LlamaSource -eq "https://github.com/ggml-org/llama.cpp") {
             $resolveSourceArgs = @("--resolve-source-build", $RequestedLlamaTag, "--published-repo", $HelperReleaseRepo, "--output-format", "json")
             if ($env:UNSLOTH_LLAMA_RELEASE_TAG) { $resolveSourceArgs += @("--published-release-tag", $env:UNSLOTH_LLAMA_RELEASE_TAG) }
-            $sourcePlanOutput = & python "$PSScriptRoot\install_llama_prebuilt.py" @resolveSourceArgs 2>$null
-            $sourcePlanExit = $LASTEXITCODE
+            $sourcePlanResult = Invoke-LlamaHelper -Arguments $resolveSourceArgs
+            $sourcePlanOutput = $sourcePlanResult.Output
+            $sourcePlanExit = $sourcePlanResult.ExitCode
             if ($sourcePlanExit -eq 0 -and $sourcePlanOutput) {
                 try {
                     $sourcePlan = ($sourcePlanOutput | Out-String) | ConvertFrom-Json
diff --git a/studio/setup.sh b/studio/setup.sh
index 116dc12e08..76fd79e7a2 100755
--- a/studio/setup.sh
+++ b/studio/setup.sh
@@ -483,19 +483,22 @@ _LLAMA_CPP_DEGRADED=false
 _LLAMA_FORCE_COMPILE="${UNSLOTH_LLAMA_FORCE_COMPILE:-0}"
 _REQUESTED_LLAMA_TAG="${UNSLOTH_LLAMA_TAG:-${_DEFAULT_LLAMA_TAG}}"
 # Force all installs to use mainline llama.cpp from ggml-org.
-# Previously: _HELPER_RELEASE_REPO="${UNSLOTH_LLAMA_RELEASE_REPO:-unslothai/llama.cpp}"
 _HELPER_RELEASE_REPO="ggml-org/llama.cpp"
 _LLAMA_PR="${UNSLOTH_LLAMA_PR:-}"
 
 _LLAMA_PR_FORCE="${UNSLOTH_LLAMA_PR_FORCE:-${_DEFAULT_LLAMA_PR_FORCE}}"
-# Force mainline source -- no env var override.
-# Previously: _LLAMA_SOURCE="${UNSLOTH_LLAMA_SOURCE:-${_DEFAULT_LLAMA_SOURCE}}"
+# Force mainline source -- no env var override for now.
 _LLAMA_SOURCE="${_DEFAULT_LLAMA_SOURCE}"
 _LLAMA_SOURCE="${_LLAMA_SOURCE%.git}"  # normalize: strip trailing .git
 _RESOLVED_SOURCE_URL="$_LLAMA_SOURCE"
 _RESOLVED_SOURCE_REF="$_REQUESTED_LLAMA_TAG"
 _RESOLVED_SOURCE_REF_KIND="tag"
 
+if [ "$_LLAMA_FORCE_COMPILE" = "1" ]; then
+    _NEED_LLAMA_SOURCE_BUILD=true
+    _SKIP_PREBUILT_INSTALL=true
+fi
+
 # Non-default source URL forces source build (fork has different code than prebuilt).
 if [ "$_LLAMA_SOURCE" != "https://github.com/ggml-org/llama.cpp" ]; then
     step "llama.cpp" "custom source: $_LLAMA_SOURCE -- forcing source build" "$C_WARN"
@@ -527,7 +530,9 @@ elif [ "${_SKIP_PREBUILT_INSTALL:-false}" = true ]; then
     # the prebuilt release resolution entirely. When building from a custom
     # fork, the fork may not carry upstream bNNNN tags, so resolve the tag
     # only when the source is the default ggml-org repo.
-    if [ "$_LLAMA_SOURCE" = "https://github.com/ggml-org/llama.cpp" ]; then
+    if [ "$_LLAMA_FORCE_COMPILE" = "1" ]; then
+        _RESOLVED_LLAMA_TAG="$_REQUESTED_LLAMA_TAG"
+    elif [ "$_LLAMA_SOURCE" = "https://github.com/ggml-org/llama.cpp" ]; then
         _RESOLVE_TAG_ARGS=(--resolve-llama-tag "$_REQUESTED_LLAMA_TAG" --published-repo "$_HELPER_RELEASE_REPO")
         _RESOLVE_TAG_ARGS+=(--output-format json)
         if [ -n "${UNSLOTH_LLAMA_RELEASE_TAG:-}" ]; then
diff --git a/tests/studio/install/test_llama_pr_force_and_source.py b/tests/studio/install/test_llama_pr_force_and_source.py
index 44cbda4c71..114680f458 100644
--- a/tests/studio/install/test_llama_pr_force_and_source.py
+++ b/tests/studio/install/test_llama_pr_force_and_source.py
@@ -1,11 +1,11 @@
 """
-Tests for UNSLOTH_LLAMA_PR_FORCE and UNSLOTH_LLAMA_SOURCE in setup.sh / setup.ps1.
+Tests for the current llama.cpp wrapper policy in setup.sh / setup.ps1.
 
 Tests cover:
   - Bash subprocess: PR_FORCE promotion, user-override, zero/empty/invalid ignored
-  - Bash subprocess: custom source URL forces build, clone URL uses variable
-  - Static source checks: defaults present, clone URLs parameterized
-  - PowerShell subprocess: PR_FORCE promotion, user-override parity
+  - Bash subprocess: source remains pinned to ggml-org even if env source is set
+  - Static source checks: mainline repo/source are hardcoded for now
+  - PowerShell subprocess: PR_FORCE promotion and fixed-source parity
 
 Run: pytest tests/studio/install/test_llama_pr_force_and_source.py -v
 """
@@ -109,7 +109,7 @@ def make_mock_git(tmp_path: Path, *, fail_on: str = "") -> tuple[Path, Path]:
 
 
 # =========================================================================
-# Bash fragment that exercises PR_FORCE and _LLAMA_SOURCE resolution
+# Bash fragment that exercises PR_FORCE and fixed _LLAMA_SOURCE resolution
 # =========================================================================
 def _bash_resolution_fragment(
     llama_pr: str = "",
@@ -125,7 +125,8 @@ def _bash_resolution_fragment(
         _DEFAULT_LLAMA_SOURCE={shlex.quote(default_source)}
 
         _LLAMA_PR_FORCE={shlex.quote(llama_pr_force) if llama_pr_force else '"$_DEFAULT_LLAMA_PR_FORCE"'}
-        _LLAMA_SOURCE={shlex.quote(llama_source) if llama_source else '"$_DEFAULT_LLAMA_SOURCE"'}
+        export UNSLOTH_LLAMA_SOURCE={shlex.quote(llama_source) if llama_source else '""'}
+        _LLAMA_SOURCE="$_DEFAULT_LLAMA_SOURCE"
         _LLAMA_SOURCE="${{_LLAMA_SOURCE%.git}}"
 
         _NEED_LLAMA_SOURCE_BUILD=false
@@ -225,20 +226,10 @@ class TestBashPrForcePromotion:
 
 
 # =========================================================================
-# TEST GROUP B: Bash custom source URL (subprocess)
+# TEST GROUP B: Bash fixed mainline source (subprocess)
 # =========================================================================
-class TestBashCustomSource:
-    """Custom _LLAMA_SOURCE forces source build."""
-
-    def test_custom_source_forces_build(self):
-        script = _bash_resolution_fragment(
-            llama_source = "https://github.com/unslothai/llama.cpp",
-        )
-        r = run_bash(script)
-        assert r.returncode == 0
-        assert "NEED_SOURCE=true" in r.stdout
-        assert "SKIP_PREBUILT=true" in r.stdout
-        assert "custom source:" in r.stdout
+class TestBashFixedMainlineSource:
+    """Source remains pinned to ggml-org while the temporary policy is active."""
 
     def test_default_source_no_force(self):
         script = _bash_resolution_fragment()
@@ -248,32 +239,23 @@ class TestBashCustomSource:
         assert "SKIP_PREBUILT=false" in r.stdout
         assert "custom source:" not in r.stdout
 
-    def test_trailing_git_stripped(self):
+    def test_env_source_override_is_ignored(self):
         script = _bash_resolution_fragment(
             llama_source = "https://github.com/unslothai/llama.cpp.git",
         )
         r = run_bash(script)
         assert r.returncode == 0
-        assert "LLAMA_SOURCE=https://github.com/unslothai/llama.cpp" in r.stdout
-        assert "NEED_SOURCE=true" in r.stdout
+        assert "LLAMA_SOURCE=https://github.com/ggml-org/llama.cpp" in r.stdout
+        assert "NEED_SOURCE=false" in r.stdout
+        assert "SKIP_PREBUILT=false" in r.stdout
 
-    def test_baked_in_source_forces_build(self):
+    def test_baked_in_source_stays_mainline(self):
         script = _bash_resolution_fragment(
-            default_source = "https://github.com/unslothai/llama.cpp",
+            default_source = "https://github.com/ggml-org/llama.cpp",
         )
         r = run_bash(script)
         assert r.returncode == 0
-        assert "NEED_SOURCE=true" in r.stdout
-
-    def test_env_source_overrides_baked_in(self):
-        """User UNSLOTH_LLAMA_SOURCE overrides _DEFAULT_LLAMA_SOURCE."""
-        script = _bash_resolution_fragment(
-            llama_source = "https://github.com/custom/llama.cpp",
-            default_source = "https://github.com/unslothai/llama.cpp",
-        )
-        r = run_bash(script)
-        assert r.returncode == 0
-        assert "LLAMA_SOURCE=https://github.com/custom/llama.cpp" in r.stdout
+        assert "LLAMA_SOURCE=https://github.com/ggml-org/llama.cpp" in r.stdout
 
 
 # =========================================================================
@@ -386,7 +368,7 @@ class TestBashCloneUrlParameterized:
 # TEST GROUP D: Static source patterns -- setup.sh
 # =========================================================================
 class TestSourcePatternsSh:
-    """Verify setup.sh has the new defaults and parameterized clone URLs."""
+    """Verify setup.sh keeps the temporary mainline-only llama.cpp policy."""
 
     @pytest.fixture(autouse = True)
     def _load_source(self):
@@ -404,8 +386,21 @@ class TestSourcePatternsSh:
     def test_has_pr_force_env_read(self):
         assert "UNSLOTH_LLAMA_PR_FORCE" in self.content
 
-    def test_has_source_env_read(self):
-        assert "UNSLOTH_LLAMA_SOURCE" in self.content
+    def test_source_env_override_removed(self):
+        assert "UNSLOTH_LLAMA_SOURCE:-${_DEFAULT_LLAMA_SOURCE}" not in self.content
+        assert '_LLAMA_SOURCE="${_DEFAULT_LLAMA_SOURCE}"' in self.content
+
+    def test_release_repo_override_removed(self):
+        assert "UNSLOTH_LLAMA_RELEASE_REPO:-unslothai/llama.cpp" not in self.content
+        assert '_HELPER_RELEASE_REPO="ggml-org/llama.cpp"' in self.content
+
+    def test_force_compile_skips_prebuilt_resolution_early(self):
+        assert 'if [ "$_LLAMA_FORCE_COMPILE" = "1" ]; then' in self.content
+        assert "_SKIP_PREBUILT_INSTALL=true" in self.content
+
+    def test_force_compile_uses_requested_tag_without_helper(self):
+        assert 'if [ "$_LLAMA_FORCE_COMPILE" = "1" ]; then' in self.content
+        assert '_RESOLVED_LLAMA_TAG="$_REQUESTED_LLAMA_TAG"' in self.content
 
     def test_pr_force_resolution_block(self):
         assert '_LLAMA_PR="$_LLAMA_PR_FORCE"' in self.content
@@ -425,16 +420,13 @@ class TestSourcePatternsSh:
         assert "ggml-org/llama.cpp.git" not in pr_block
 
     def test_clone_urls_parameterized_tag_path(self):
-        """Non-PR clone path uses ${_LLAMA_SOURCE}.git, not hardcoded URL."""
+        """Non-PR clone path uses the resolved source URL, not a hardcoded URL."""
         # Find the non-PR clone line (after _CLONE_ARGS)
         idx = self.content.index("_CLONE_ARGS=(git clone --depth 1)")
         block = self.content[idx : idx + 400]
-        assert '"${_LLAMA_SOURCE}.git"' in block
+        assert '"${_RESOLVED_SOURCE_URL}.git"' in block
         assert "ggml-org/llama.cpp.git" not in block
 
-    def test_custom_source_forces_build(self):
-        assert "custom source: $_LLAMA_SOURCE -- forcing source build" in self.content
-
     def test_no_hardcoded_clone_urls(self):
         """No remaining hardcoded ggml-org clone URLs in clone commands."""
         lines = self.content.splitlines()
@@ -449,7 +441,7 @@ class TestSourcePatternsSh:
 # TEST GROUP E: Static source patterns -- setup.ps1
 # =========================================================================
 class TestSourcePatternsPs1:
-    """Verify setup.ps1 has the new defaults and parameterized clone URLs."""
+    """Verify setup.ps1 keeps the temporary mainline-only llama.cpp policy."""
 
     @pytest.fixture(autouse = True)
     def _load_source(self):
@@ -467,8 +459,24 @@ class TestSourcePatternsPs1:
     def test_has_pr_force_env_read(self):
         assert "$env:UNSLOTH_LLAMA_PR_FORCE" in self.content
 
-    def test_has_source_env_read(self):
-        assert "$env:UNSLOTH_LLAMA_SOURCE" in self.content
+    def test_source_env_override_removed(self):
+        assert "$LlamaSource = if ($env:UNSLOTH_LLAMA_SOURCE)" not in self.content
+        assert "$LlamaSource = $DefaultLlamaSource" in self.content
+
+    def test_release_repo_override_removed(self):
+        assert (
+            "$HelperReleaseRepo = if ($env:UNSLOTH_LLAMA_RELEASE_REPO)"
+            not in self.content
+        )
+        assert '$HelperReleaseRepo = "ggml-org/llama.cpp"' in self.content
+
+    def test_force_compile_skips_prebuilt_resolution_early(self):
+        assert 'if ($env:UNSLOTH_LLAMA_FORCE_COMPILE -eq "1") {' in self.content
+        assert "$SkipPrebuiltInstall = $true" in self.content
+
+    def test_force_compile_uses_requested_tag_without_helper(self):
+        assert 'if ($env:UNSLOTH_LLAMA_FORCE_COMPILE -eq "1") {' in self.content
+        assert "$ResolvedLlamaTag = $RequestedLlamaTag" in self.content
 
     def test_pr_force_promotion_block(self):
         assert "$LlamaPr = $LlamaPrForce" in self.content
@@ -487,15 +495,12 @@ class TestSourcePatternsPs1:
         assert "ggml-org/llama.cpp.git" not in pr_block
 
     def test_clone_urls_parameterized_tag_path(self):
-        """Non-PR clone path uses $LlamaSource.git, not hardcoded URL."""
+        """Non-PR clone path uses the resolved source URL, not a hardcoded URL."""
         clone_args_idx = self.content.index('$cloneArgs = @("clone"')
         block = self.content[clone_args_idx : clone_args_idx + 400]
-        assert '"$LlamaSource.git"' in block
+        assert '"$ResolvedSourceUrl.git"' in block
         assert "ggml-org/llama.cpp.git" not in block
 
-    def test_custom_source_forces_build(self):
-        assert "custom source: $LlamaSource -- forcing source build" in self.content
-
     def test_no_hardcoded_clone_urls(self):
         """No remaining hardcoded ggml-org clone URLs in clone commands."""
         lines = self.content.splitlines()
@@ -511,7 +516,7 @@ class TestSourcePatternsPs1:
 # =========================================================================
 @requires_pwsh
 class TestPwshPrForcePromotion:
-    """PR_FORCE promotion and source URL logic via pwsh subprocess."""
+    """PR_FORCE promotion and fixed-source logic via pwsh subprocess."""
 
     FRAGMENT_TEMPLATE = textwrap.dedent("""\
         function step($a, $b, $c) { Write-Output "step:$a`:$b" }
@@ -521,7 +526,7 @@ class TestPwshPrForcePromotion:
 
         $LlamaPr = if ($env:UNSLOTH_LLAMA_PR) { $env:UNSLOTH_LLAMA_PR.Trim() } else { "" }
         $LlamaPrForce = if ($env:UNSLOTH_LLAMA_PR_FORCE) { $env:UNSLOTH_LLAMA_PR_FORCE.Trim() } else { $DefaultLlamaPrForce }
-        $LlamaSource = if ($env:UNSLOTH_LLAMA_SOURCE) { $env:UNSLOTH_LLAMA_SOURCE.Trim() } else { $DefaultLlamaSource }
+        $LlamaSource = $DefaultLlamaSource
         if ($LlamaSource.EndsWith('.git')) { $LlamaSource = $LlamaSource.Substring(0, $LlamaSource.Length - 4) }
 
         $NeedLlamaSourceBuild = $false
@@ -561,7 +566,6 @@ class TestPwshPrForcePromotion:
         # Ensure env vars are unset by default
         run_env["UNSLOTH_LLAMA_PR"] = ""
         run_env["UNSLOTH_LLAMA_PR_FORCE"] = ""
-        run_env["UNSLOTH_LLAMA_SOURCE"] = ""
         if env:
             run_env.update(env)
         return run_pwsh(script, env = run_env)
@@ -599,15 +603,16 @@ class TestPwshPrForcePromotion:
         assert r.returncode == 0
         assert "baked-in PR_FORCE" not in r.stdout
 
-    def test_custom_source_forces_build(self):
+    def test_env_source_override_is_ignored(self):
         r = self._run(
             env = {
                 "UNSLOTH_LLAMA_SOURCE": "https://github.com/unslothai/llama.cpp",
             }
         )
         assert r.returncode == 0
-        assert "NEED_SOURCE=True" in r.stdout
-        assert "SKIP_PREBUILT=True" in r.stdout
+        assert "LLAMA_SOURCE=https://github.com/ggml-org/llama.cpp" in r.stdout
+        assert "NEED_SOURCE=False" in r.stdout
+        assert "SKIP_PREBUILT=False" in r.stdout
 
     def test_default_source_no_force(self):
         r = self._run()
@@ -615,16 +620,16 @@ class TestPwshPrForcePromotion:
         assert "NEED_SOURCE=False" in r.stdout
         assert "SKIP_PREBUILT=False" in r.stdout
 
-    def test_trailing_git_stripped(self):
+    def test_trailing_git_override_is_ignored(self):
         r = self._run(
             env = {
                 "UNSLOTH_LLAMA_SOURCE": "https://github.com/unslothai/llama.cpp.git",
             }
         )
         assert r.returncode == 0
-        assert "LLAMA_SOURCE=https://github.com/unslothai/llama.cpp" in r.stdout
+        assert "LLAMA_SOURCE=https://github.com/ggml-org/llama.cpp" in r.stdout
 
-    def test_baked_in_source_forces_build(self):
-        r = self._run(default_source = "https://github.com/unslothai/llama.cpp")
+    def test_baked_in_source_stays_mainline(self):
+        r = self._run(default_source = "https://github.com/ggml-org/llama.cpp")
         assert r.returncode == 0
-        assert "NEED_SOURCE=True" in r.stdout
+        assert "LLAMA_SOURCE=https://github.com/ggml-org/llama.cpp" in r.stdout
diff --git a/tests/studio/install/test_pr4562_bugfixes.py b/tests/studio/install/test_pr4562_bugfixes.py
index f5a423d130..7fa654d845 100644
--- a/tests/studio/install/test_pr4562_bugfixes.py
+++ b/tests/studio/install/test_pr4562_bugfixes.py
@@ -778,6 +778,34 @@ class TestSourceCodePatterns:
         assert "$ResolvedSourceRefKind" in content
         assert "$ResolvedSourceRef" in content
 
+    def test_setup_ps1_prebuilt_install_disables_native_error_abort(self):
+        """PS1 prebuilt install should not abort setup on helper stderr."""
+        content = SETUP_PS1.read_text()
+        install_idx = content.index("& python @prebuiltArgs 2>&1")
+        block = content[max(0, install_idx - 800) : install_idx + 800]
+        assert "$PSNativeCommandUseErrorActionPreference = $false" in block
+        assert "$restoreNativeErrorPreference = $true" in block
+        assert (
+            "$PSNativeCommandUseErrorActionPreference = $previousNativeErrorPreference"
+            in block
+        )
+
+    def test_setup_ps1_helper_disables_error_action_abort(self):
+        """Helper resolution should suppress terminating NativeCommandError on PS 5.1."""
+        content = SETUP_PS1.read_text()
+        helper_idx = content.index("function Invoke-LlamaHelper")
+        block = content[helper_idx : helper_idx + 1200]
+        assert "$previousErrorActionPreference = $ErrorActionPreference" in block
+        assert '$ErrorActionPreference = "Continue"' in block
+        assert "$ErrorActionPreference = $previousErrorActionPreference" in block
+
+    def test_setup_ps1_uses_local_tempfile_helper(self):
+        """PS1 should not depend on New-TemporaryFile being available."""
+        content = SETUP_PS1.read_text()
+        assert "function New-UnslothTemporaryFile" in content
+        assert "$resolveErrorLog = New-UnslothTemporaryFile" in content
+        assert "$resolveErrorLog = New-TemporaryFile" not in content
+
     def test_binary_env_linux_has_binary_parent(self):
         """The Linux branch of binary_env should include binary_path.parent."""
         content = MODULE_PATH.read_text()
diff --git a/tests/studio/install/test_selection_logic.py b/tests/studio/install/test_selection_logic.py
index b8de0e221f..6a2e367dc8 100644
--- a/tests/studio/install/test_selection_logic.py
+++ b/tests/studio/install/test_selection_logic.py
@@ -511,6 +511,56 @@ class TestApplyApprovedHashes:
         result = apply_approved_hashes([choice], checksums)
         assert result[0].expected_sha256 == "a" * 64
 
+    def test_windows_cuda_legacy_choice_can_match_current_upstream_name(self):
+        choice = AssetChoice(
+            repo = UPSTREAM_REPO,
+            tag = "b9000",
+            name = "llama-b9000-bin-win-cuda-13.1-x64.zip",
+            url = "https://x/llama-b9000-bin-win-cuda-13.1-x64.zip",
+            source_label = "upstream",
+        )
+        checksums = ApprovedReleaseChecksums(
+            repo = "unslothai/llama.cpp",
+            release_tag = "r1",
+            upstream_tag = "b9000",
+            artifacts = {
+                "cudart-llama-bin-win-cuda-13.1-x64.zip": ApprovedArtifactHash(
+                    asset_name = "cudart-llama-bin-win-cuda-13.1-x64.zip",
+                    sha256 = "b" * 64,
+                    repo = UPSTREAM_REPO,
+                    kind = "windows-cuda-upstream",
+                )
+            },
+        )
+
+        result = apply_approved_hashes([choice], checksums)
+        assert result[0].expected_sha256 == "b" * 64
+
+    def test_windows_cuda_current_choice_can_match_legacy_compatibility_name(self):
+        choice = AssetChoice(
+            repo = UPSTREAM_REPO,
+            tag = "main",
+            name = "cudart-llama-bin-win-cuda-13.1-x64.zip",
+            url = "https://x/cudart-llama-bin-win-cuda-13.1-x64.zip",
+            source_label = "upstream",
+        )
+        checksums = ApprovedReleaseChecksums(
+            repo = "unslothai/llama.cpp",
+            release_tag = "r1",
+            upstream_tag = "b9000",
+            artifacts = {
+                "llama-b9000-bin-win-cuda-13.1-x64.zip": ApprovedArtifactHash(
+                    asset_name = "llama-b9000-bin-win-cuda-13.1-x64.zip",
+                    sha256 = "c" * 64,
+                    repo = UPSTREAM_REPO,
+                    kind = "windows-cuda-upstream",
+                )
+            },
+        )
+
+        result = apply_approved_hashes([choice], checksums)
+        assert result[0].expected_sha256 == "c" * 64
+
     def test_none_approved(self):
         c1 = self._choice("missing.tar.gz")
         checksums = make_checksums(["other.tar.gz"])

From 8d1712b4ea50cdbf36c1db3f22c7e2c23eadfedf Mon Sep 17 00:00:00 2001
From: Daniel Han <danielhanchen@gmail.com>
Date: Thu, 2 Apr 2026 11:43:53 -0700
Subject: [PATCH 023/216] fix(studio): pin llama.cpp to b8637 release (Gemma 4
 support) (#4796)

ggml-org/llama.cpp b8637 includes Gemma 4 support (ggml-org/llama.cpp#21309).
Revert the temporary "master" default back to a pinned release tag.

This eliminates the HTTP 422 errors from the prebuilt resolver (which
could not find a release matching "master"), avoids unnecessary source
builds, and restores prebuilt binary downloads on all platforms.

Co-authored-by: Daniel Han <danielhanchen@users.noreply.github.com>
---
 studio/install_llama_prebuilt.py | 2 +-
 studio/setup.ps1                 | 2 +-
 studio/setup.sh                  | 2 +-
 3 files changed, 3 insertions(+), 3 deletions(-)

diff --git a/studio/install_llama_prebuilt.py b/studio/install_llama_prebuilt.py
index 13382d3c16..9fe5b55e9f 100755
--- a/studio/install_llama_prebuilt.py
+++ b/studio/install_llama_prebuilt.py
@@ -60,7 +60,7 @@ def env_int(name: str, default: int, *, minimum: int | None = None) -> int:
     return value
 
 
-DEFAULT_LLAMA_TAG = os.environ.get("UNSLOTH_LLAMA_TAG", "master")
+DEFAULT_LLAMA_TAG = os.environ.get("UNSLOTH_LLAMA_TAG", "b8637")
 # Force all installs to use mainline llama.cpp from ggml-org.
 # Previously: DEFAULT_PUBLISHED_REPO = os.environ.get("UNSLOTH_LLAMA_RELEASE_REPO", "unslothai/llama.cpp")
 DEFAULT_PUBLISHED_REPO = "ggml-org/llama.cpp"
diff --git a/studio/setup.ps1 b/studio/setup.ps1
index 382de89b9f..47d7f0ee55 100644
--- a/studio/setup.ps1
+++ b/studio/setup.ps1
@@ -29,7 +29,7 @@ $PackageDir = Split-Path -Parent $ScriptDir
 # --------------------------------------------------------------------------
 $DefaultLlamaPrForce = ""
 $DefaultLlamaSource = "https://github.com/ggml-org/llama.cpp"
-$DefaultLlamaTag = "master"
+$DefaultLlamaTag = "b8637"
 
 # Verbose can be enabled either by CLI flag or by UNSLOTH_VERBOSE=1.
 $script:UnslothVerbose = ($env:UNSLOTH_VERBOSE -eq '1')
diff --git a/studio/setup.sh b/studio/setup.sh
index 76fd79e7a2..4e7bd366ef 100755
--- a/studio/setup.sh
+++ b/studio/setup.sh
@@ -19,7 +19,7 @@ RULE=$(printf '\342\224\200%.0s' {1..52})
 # ──────────────────────────────────────────────────────────────────────────
 _DEFAULT_LLAMA_PR_FORCE=""
 _DEFAULT_LLAMA_SOURCE="https://github.com/ggml-org/llama.cpp"
-_DEFAULT_LLAMA_TAG="master"
+_DEFAULT_LLAMA_TAG="b8637"
 
 # ── Colors (same palette as startup_banner / install_python_stack) ──
 if [ -n "${NO_COLOR:-}" ]; then

From 401621618b7ff548fd04c2ea5b2aca05afd936ac Mon Sep 17 00:00:00 2001
From: Daniel Han <danielhanchen@gmail.com>
Date: Thu, 2 Apr 2026 11:44:26 -0700
Subject: [PATCH 024/216] fix(studio): don't set trust_remote_code for Gemma 4
 training (#4795)

Gemma 4 is a native transformers 5.5 model and does not need
trust_remote_code=True. The auto-enable logic (added for NemotronH)
was catching all transformers 5.x models, including Gemma 4.

When trust_remote_code=True, unsloth_compile_transformers() returns
early without running the compiler. This disables the fused cross
entropy patch, causing logged training loss to be inflated by the
gradient_accumulation_steps factor.

Exclude models matching "gemma-4" or "gemma4" from the auto-enable
so the compiler runs and applies fused cross entropy correctly.
---
 studio/backend/core/training/worker.py | 7 ++++++-
 1 file changed, 6 insertions(+), 1 deletion(-)

diff --git a/studio/backend/core/training/worker.py b/studio/backend/core/training/worker.py
index 699cfe74f7..0454eada89 100644
--- a/studio/backend/core/training/worker.py
+++ b/studio/backend/core/training/worker.py
@@ -390,11 +390,16 @@ def run_training_process(
     # Some newer architectures (e.g. NemotronH) have config parsing bugs in
     # transformers that require trust_remote_code=True as a workaround.
     # Only auto-enable for unsloth/* prefixed models (trusted source).
+    # Exclude Gemma 4 since it is a native transformers 5.5 model and
+    # trust_remote_code=True would bypass the compiler (disabling fused CE).
     from utils.transformers_version import needs_transformers_5
 
+    _lowered = model_name.lower()
+    _is_native_t5 = any(x in _lowered for x in ("gemma-4", "gemma4"))
     if (
         needs_transformers_5(model_name)
-        and model_name.lower().startswith("unsloth/")
+        and _lowered.startswith("unsloth/")
+        and not _is_native_t5
         and not config.get("trust_remote_code", False)
     ):
         config["trust_remote_code"] = True

From 934478ae317d337bc89887eb445df533de42b40b Mon Sep 17 00:00:00 2001
From: Daniel Han <danielhanchen@gmail.com>
Date: Thu, 2 Apr 2026 11:52:37 -0700
Subject: [PATCH 025/216] fix(studio): revert llama.cpp default tag to latest
 (#4797)

* fix(studio): revert llama.cpp default tag to latest

The latest ggml-org/llama.cpp release (b8637) now includes Gemma 4
support. Revert the temporary "b8637" pin from #4796 to "latest" so
the prebuilt resolver always picks the newest release automatically
without needing manual tag bumps.

* docs: add comment explaining latest vs master for llama.cpp tag

Document in all three files why "latest" is preferred over "master"
and when "master" should be used as a temporary override.

---------

Co-authored-by: Daniel Han <danielhanchen@users.noreply.github.com>
---
 studio/install_llama_prebuilt.py | 6 +++++-
 studio/setup.ps1                 | 6 +++++-
 studio/setup.sh                  | 7 ++++++-
 3 files changed, 16 insertions(+), 3 deletions(-)

diff --git a/studio/install_llama_prebuilt.py b/studio/install_llama_prebuilt.py
index 9fe5b55e9f..1b02729649 100755
--- a/studio/install_llama_prebuilt.py
+++ b/studio/install_llama_prebuilt.py
@@ -60,7 +60,11 @@ def env_int(name: str, default: int, *, minimum: int | None = None) -> int:
     return value
 
 
-DEFAULT_LLAMA_TAG = os.environ.get("UNSLOTH_LLAMA_TAG", "b8637")
+# Prefer "latest" over "master" -- "master" bypasses the prebuilt resolver
+# (no matching GitHub release), forces a source build, and causes HTTP 422
+# errors. Only use "master" temporarily when the latest release is missing
+# support for a new model architecture.
+DEFAULT_LLAMA_TAG = os.environ.get("UNSLOTH_LLAMA_TAG", "latest")
 # Force all installs to use mainline llama.cpp from ggml-org.
 # Previously: DEFAULT_PUBLISHED_REPO = os.environ.get("UNSLOTH_LLAMA_RELEASE_REPO", "unslothai/llama.cpp")
 DEFAULT_PUBLISHED_REPO = "ggml-org/llama.cpp"
diff --git a/studio/setup.ps1 b/studio/setup.ps1
index 47d7f0ee55..d478319098 100644
--- a/studio/setup.ps1
+++ b/studio/setup.ps1
@@ -27,9 +27,13 @@ $PackageDir = Split-Path -Parent $ScriptDir
 #  Change these in the GitHub-hosted script so users get updated defaults.
 #  User env vars always override these baked-in values.
 # --------------------------------------------------------------------------
+# Prefer "latest" over "master" -- "master" bypasses the prebuilt resolver
+# (no matching GitHub release), forces a source build, and causes HTTP 422
+# errors. Only use "master" temporarily when the latest release is missing
+# support for a new model architecture.
 $DefaultLlamaPrForce = ""
 $DefaultLlamaSource = "https://github.com/ggml-org/llama.cpp"
-$DefaultLlamaTag = "b8637"
+$DefaultLlamaTag = "latest"
 
 # Verbose can be enabled either by CLI flag or by UNSLOTH_VERBOSE=1.
 $script:UnslothVerbose = ($env:UNSLOTH_VERBOSE -eq '1')
diff --git a/studio/setup.sh b/studio/setup.sh
index 4e7bd366ef..d9ce73661f 100755
--- a/studio/setup.sh
+++ b/studio/setup.sh
@@ -16,10 +16,15 @@ RULE=$(printf '\342\224\200%.0s' {1..52})
 #   _DEFAULT_LLAMA_SOURCE   : git clone URL for source builds
 #   _DEFAULT_LLAMA_TAG      : llama.cpp ref to build ("latest" = newest release,
 #                             "master" = bleeding-edge, "bNNNN" = specific tag)
+#                             Prefer "latest" over "master" -- "master" bypasses
+#                             the prebuilt resolver (no matching GitHub release),
+#                             forces a source build, and causes HTTP 422 errors.
+#                             Only use "master" temporarily when the latest release
+#                             is missing support for a new model architecture.
 # ──────────────────────────────────────────────────────────────────────────
 _DEFAULT_LLAMA_PR_FORCE=""
 _DEFAULT_LLAMA_SOURCE="https://github.com/ggml-org/llama.cpp"
-_DEFAULT_LLAMA_TAG="b8637"
+_DEFAULT_LLAMA_TAG="latest"
 
 # ── Colors (same palette as startup_banner / install_python_stack) ──
 if [ -n "${NO_COLOR:-}" ]; then

From 8ffd5826f23d487999bee55d7e6345378c03c225 Mon Sep 17 00:00:00 2001
From: Daniel Han <danielhanchen@gmail.com>
Date: Thu, 2 Apr 2026 11:59:37 -0700
Subject: [PATCH 026/216] Gemma-4

---
 pyproject.toml           | 8 ++++----
 unsloth/models/_utils.py | 2 +-
 2 files changed, 5 insertions(+), 5 deletions(-)

diff --git a/pyproject.toml b/pyproject.toml
index b876c19467..e391b4df3d 100644
--- a/pyproject.toml
+++ b/pyproject.toml
@@ -82,13 +82,13 @@ huggingfacenotorch = [
     "huggingface_hub>=0.34.0",
     "hf_transfer",
     "diffusers",
-    "transformers>=4.51.3,!=4.52.0,!=4.52.1,!=4.52.2,!=4.52.3,!=4.53.0,!=4.54.0,!=4.55.0,!=4.55.1,!=4.57.0,!=4.57.4,!=4.57.5,!=5.0.0,!=5.1.0,<=5.3.0",
+    "transformers>=4.51.3,!=4.52.0,!=4.52.1,!=4.52.2,!=4.52.3,!=4.53.0,!=4.54.0,!=4.55.0,!=4.55.1,!=4.57.0,!=4.57.4,!=4.57.5,!=5.0.0,!=5.1.0,<=5.5.0",
     "trl>=0.18.2,!=0.19.0,<=0.24.0",
     "sentence-transformers",
 ]
 huggingface = [
     "unsloth[huggingfacenotorch]",
-    "unsloth_zoo>=2026.3.7",
+    "unsloth_zoo>=2026.4.2",
     "torchvision",
     "unsloth[triton]",
 ]
@@ -578,10 +578,10 @@ colab-ampere-torch220 = [
     "flash-attn>=2.6.3 ; ('linux' in sys_platform)",
 ]
 colab-new = [
-    "unsloth_zoo>=2026.3.7",
+    "unsloth_zoo>=2026.4.2",
     "packaging",
     "tyro",
-    "transformers>=4.51.3,!=4.52.0,!=4.52.1,!=4.52.2,!=4.52.3,!=4.53.0,!=4.54.0,!=4.55.0,!=4.55.1,!=4.57.0,!=4.57.4,!=4.57.5,!=5.0.0,!=5.1.0,<=5.3.0",
+    "transformers>=4.51.3,!=4.52.0,!=4.52.1,!=4.52.2,!=4.52.3,!=4.53.0,!=4.54.0,!=4.55.0,!=4.55.1,!=4.57.0,!=4.57.4,!=4.57.5,!=5.0.0,!=5.1.0,<=5.5.0",
     "datasets>=3.4.1,!=4.0.*,!=4.1.0,<4.4.0",
     "sentencepiece>=0.2.0",
     "tqdm",
diff --git a/unsloth/models/_utils.py b/unsloth/models/_utils.py
index 1c803a4972..28526056ba 100644
--- a/unsloth/models/_utils.py
+++ b/unsloth/models/_utils.py
@@ -12,7 +12,7 @@
 # See the License for the specific language governing permissions and
 # limitations under the License.
 
-__version__ = "2026.3.18"
+__version__ = "2026.4.1"
 
 __all__ = [
     "SUPPORTS_BFLOAT16",

From e553a8ad0bc9645fdad3159c8f49f4340ab144a1 Mon Sep 17 00:00:00 2001
From: Daniel Han <danielhanchen@gmail.com>
Date: Thu, 2 Apr 2026 12:18:11 -0700
Subject: [PATCH 027/216] fix(studio): suppress fatal error when prebuilt
 manifest is missing (#4799)

When DEFAULT_PUBLISHED_REPO is ggml-org/llama.cpp, the prebuilt
resolver raises PrebuiltFallback because ggml-org releases do not
include a llama-prebuilt-manifest.json asset. This was caught by the
generic Exception handler and printed as "fatal helper error" to
stderr, which triggers NativeCommandError on PowerShell.

Catch PrebuiltFallback separately in the top-level __main__ handler
and exit with EXIT_FALLBACK (code 2) instead of EXIT_ERROR (code 1).
The message is still logged but without the "fatal helper error"
prefix. The shell scripts already handle non-zero exits and fall
back to source builds.

Co-authored-by: Daniel Han <danielhanchen@users.noreply.github.com>
---
 studio/install_llama_prebuilt.py | 6 ++++++
 1 file changed, 6 insertions(+)

diff --git a/studio/install_llama_prebuilt.py b/studio/install_llama_prebuilt.py
index 1b02729649..5e8fe314b4 100755
--- a/studio/install_llama_prebuilt.py
+++ b/studio/install_llama_prebuilt.py
@@ -4733,6 +4733,12 @@ if __name__ == "__main__":
             f"fatal helper busy conflict: {textwrap.shorten(str(exc), width = 400, placeholder = '...')}"
         )
         raise SystemExit(EXIT_BUSY)
+    except PrebuiltFallback as exc:
+        # Expected when the published repo (e.g. ggml-org/llama.cpp) has no
+        # prebuilt manifest.  Exit quietly with EXIT_FALLBACK so the caller
+        # falls back to source build without a noisy "fatal helper error".
+        log(textwrap.shorten(str(exc), width = 400, placeholder = "..."))
+        raise SystemExit(EXIT_FALLBACK)
     except Exception as exc:
         message = textwrap.shorten(str(exc), width = 400, placeholder = "...")
         log(f"fatal helper error: {message}")

From f91ef8f9b048938dd22bb37e820855d7a97f7871 Mon Sep 17 00:00:00 2001
From: Roland Tannous <115670425+rolandtannous@users.noreply.github.com>
Date: Fri, 3 Apr 2026 02:56:01 +0400
Subject: [PATCH 028/216] fix(studio): lazy-import transformers in model_config
 to fix 5.x version switch (#4806)
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

* fix(studio): lazy-import AutoConfig in model_config.py to fix transformers 5.x version switch

Move `from transformers import AutoConfig` from module level to inside
load_model_config() where it is actually used.

model_config.py is transitively imported at module load time via:
  core/inference/__init__ → llama_cpp → utils.models → model_config

In inference subprocesses (mp.spawn), this chain runs before
_activate_transformers_version() can prepend .venv_t5/ to sys.path.
The eager import caches transformers 4.57.6 in sys.modules, and the
subsequent sys.path change has no effect — Python always checks
sys.modules before sys.path.

Making the import lazy ensures transformers is not loaded until after
version activation, so the subprocess picks up the correct version.

* fix(studio): also lazy-import extract_model_size_b in llama_cpp.py

Belt-and-suspenders: make the import that originally triggered the
chain lazy as well, so future module-level AutoConfig additions in
utils.models cannot reintroduce the problem.

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

---------

Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com>
---
 studio/backend/core/inference/llama_cpp.py  | 10 ++++++++--
 studio/backend/utils/models/model_config.py |  2 +-
 2 files changed, 9 insertions(+), 3 deletions(-)

diff --git a/studio/backend/core/inference/llama_cpp.py b/studio/backend/core/inference/llama_cpp.py
index 44c700bf3d..d752500bf4 100644
--- a/studio/backend/core/inference/llama_cpp.py
+++ b/studio/backend/core/inference/llama_cpp.py
@@ -52,8 +52,14 @@ _REPROMPT_MAX_CHARS = 2000
 _SHARD_FULL_RE = re.compile(r"^(.*)-(\d{5})-of-(\d{5})\.gguf$")
 _SHARD_RE = re.compile(r"^(.*)-\d{5}-of-\d{5}\.gguf$")
 
-# Model size extraction (shared with routes/inference.py)
-from utils.models import extract_model_size_b as _extract_model_size_b
+
+# Model size extraction — lazy import to avoid pulling in transformers
+# at module level.  See PR description for the full explanation.
+def _extract_model_size_b(model_id: str):
+    from utils.models import extract_model_size_b
+
+    return extract_model_size_b(model_id)
+
 
 # ── Pre-compiled patterns for tool XML stripping ─────────────
 _TOOL_CLOSED_PATS = [
diff --git a/studio/backend/utils/models/model_config.py b/studio/backend/utils/models/model_config.py
index df1058abf6..6cffc534aa 100644
--- a/studio/backend/utils/models/model_config.py
+++ b/studio/backend/utils/models/model_config.py
@@ -5,7 +5,6 @@
 Model and LoRA configuration handling
 """
 
-from transformers import AutoConfig
 from dataclasses import dataclass
 from typing import Optional, Dict, Any
 from utils.paths import (
@@ -422,6 +421,7 @@ def load_model_config(
     """
     Load model config with optional authentication control.
     """
+    from transformers import AutoConfig
 
     if token:
         # Explicit token provided - use it

From 6644a771b4c5c0b13f9e06c2e459c4d82a204c6b Mon Sep 17 00:00:00 2001
From: Roland Tannous <115670425+rolandtannous@users.noreply.github.com>
Date: Fri, 3 Apr 2026 04:03:54 +0400
Subject: [PATCH 029/216] fix: patch PEFT for Gemma4ClippableLinear in loader
 checkpoint path (fixes export) (#4807)

* fix: patch PEFT for Gemma4ClippableLinear in loader checkpoint path

The same Gemma4ClippableLinear monkey-patch that exists in vision.py
for training is needed in loader.py for loading existing checkpoints
(used by export and inference).

Gemma4ClippableLinear wraps nn.Linear but does not subclass it, so
PEFT's LoRA injection fails with "Target module not supported".
The patch redirects PEFT to target the inner .linear child instead.

Applied only to the vision model PeftModel.from_pretrained path.
Temporary fix until PEFT adds native support (peft#3129).

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

* fix: wrap ClippableLinear patch in try/finally to always restore

Ensures _create_and_replace is restored even if PeftModel.from_pretrained
raises, preventing leaked global state across subsequent model loads.

---------

Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com>
---
 unsloth/models/loader.py | 74 +++++++++++++++++++++++++++++++++++-----
 1 file changed, 66 insertions(+), 8 deletions(-)

diff --git a/unsloth/models/loader.py b/unsloth/models/loader.py
index a811d3fb75..cfe3acb656 100644
--- a/unsloth/models/loader.py
+++ b/unsloth/models/loader.py
@@ -1533,14 +1533,72 @@ class FastModel(FastBaseModel):
         if is_peft:
             # From https://github.com/huggingface/peft/issues/184
             # Now add PEFT adapters
-            model = PeftModel.from_pretrained(
-                model,
-                old_model_name,
-                token = token,
-                revision = revision,
-                is_trainable = True,
-                trust_remote_code = trust_remote_code,
-            )
+
+            # Gemma4 ClippableLinear wraps nn.Linear -- PEFT can't inject LoRA
+            # on it directly.  Monkey-patch PEFT to target the inner .linear
+            # child instead (same patch as vision.py training path).
+            # See https://github.com/huggingface/peft/issues/3129
+            _clippable_linear_cls = None
+            try:
+                from transformers.models.gemma4.modeling_gemma4 import (
+                    Gemma4ClippableLinear as _clippable_linear_cls,
+                )
+            except ImportError:
+                pass
+
+            if _clippable_linear_cls is not None:
+                from peft.tuners.lora.model import LoraModel as _LoraModel
+
+                _original_car = _LoraModel._create_and_replace
+
+                def _patched_car(
+                    self,
+                    peft_config,
+                    adapter_name,
+                    target,
+                    target_name,
+                    parent,
+                    current_key = None,
+                    **kwargs,
+                ):
+                    if isinstance(target, _clippable_linear_cls):
+                        return _original_car(
+                            self,
+                            peft_config,
+                            adapter_name,
+                            target.linear,
+                            "linear",
+                            target,
+                            current_key = current_key,
+                            **kwargs,
+                        )
+                    return _original_car(
+                        self,
+                        peft_config,
+                        adapter_name,
+                        target,
+                        target_name,
+                        parent,
+                        current_key = current_key,
+                        **kwargs,
+                    )
+
+                _LoraModel._create_and_replace = _patched_car
+
+            try:
+                model = PeftModel.from_pretrained(
+                    model,
+                    old_model_name,
+                    token = token,
+                    revision = revision,
+                    is_trainable = True,
+                    trust_remote_code = trust_remote_code,
+                )
+            finally:
+                # Always restore original PEFT method, even if loading fails
+                if _clippable_linear_cls is not None:
+                    _LoraModel._create_and_replace = _original_car
+
             # Patch it as well!
             model = FastBaseModel.post_patch_model(
                 model, use_gradient_checkpointing, trust_remote_code = trust_remote_code

From a7e69641171467f49487979b1ce29d03aaacf2ca Mon Sep 17 00:00:00 2001
From: Manan Shah <52329525+Manan17@users.noreply.github.com>
Date: Fri, 3 Apr 2026 00:03:35 -0500
Subject: [PATCH 030/216] Fix/gemma4 install script (#4815)

* transformer 5.5.0 has now been released

* fallback for python < 3.10
:
---
 install_gemma4_mlx.sh | 24 ++++++------------------
 1 file changed, 6 insertions(+), 18 deletions(-)

diff --git a/install_gemma4_mlx.sh b/install_gemma4_mlx.sh
index e06339e204..b653af9154 100755
--- a/install_gemma4_mlx.sh
+++ b/install_gemma4_mlx.sh
@@ -132,29 +132,17 @@ step "install" "installing mlx, mlx-lm..."
 uv pip install --python "$_VENV_PY" -q mlx mlx-lm 2>/dev/null
 substep "done"
 
-TRANSFORMERS_WHL="transformers-5.5.0-py3-none-any.whl"
-TRANSFORMERS_GH="git+https://github.com/huggingface/transformers.git@v5.5-release"
-
 step "install" "installing transformers>=5.5.0..."
-if uv pip install --python "$_VENV_PY" -q "$TRANSFORMERS_GH" 2>/dev/null; then
-    substep "installed from huggingface/transformers v5.5-release"
-elif uv pip install --python "$_VENV_PY" -q "transformers>=5.5.0" 2>/dev/null; then
+if uv pip install --python "$_VENV_PY" -q "transformers>=5.5.0" 2>/dev/null; then
     substep "installed from PyPI"
 else
-    substep "not on PyPI, trying unsloth branch..."
-    _whl_tmp=$(mktemp -d)/"${TRANSFORMERS_WHL}"
-    if curl -fsSL "${REPO_URL}/${TRANSFORMERS_WHL}" -o "$_whl_tmp" 2>/dev/null && \
-       uv pip install --python "$_VENV_PY" -q "$_whl_tmp" 2>/dev/null; then
-        substep "installed from branch ${BRANCH}"
-    elif [ -f "./${TRANSFORMERS_WHL}" ]; then
-        substep "using local ./${TRANSFORMERS_WHL}"
-        uv pip install --python "$_VENV_PY" -q "./${TRANSFORMERS_WHL}"
+    substep "PyPI install failed (Python <3.10?), trying GitHub..."
+    if uv pip install --python "$_VENV_PY" -q "git+https://github.com/huggingface/transformers.git@v5.5-release" 2>/dev/null; then
+        substep "installed from huggingface/transformers v5.5-release"
     else
-        rm -f "$_whl_tmp" 2>/dev/null
-        step "install" "skipping transformers — could not find >=5.5.0" "$C_WARN"
-        substep "tried: huggingface/transformers v5.5-release, PyPI, branch ${BRANCH}, local ./${TRANSFORMERS_WHL}"
+        step "warning" "could not install transformers>=5.5.0" "$C_WARN"
+        substep "tried: PyPI, huggingface/transformers v5.5-release"
     fi
-    rm -f "$_whl_tmp" 2>/dev/null
 fi
 
 # ── Find mlx-lm models directory ─────────────────────────────

From c1685b945902393e863a1ccddc347aa5229769df Mon Sep 17 00:00:00 2001
From: Michael Han <107991372+shimmyshimmer@users.noreply.github.com>
Date: Thu, 2 Apr 2026 22:54:03 -0700
Subject: [PATCH 031/216] Gemma 4 update.md

---
 README.md | 8 +++++---
 1 file changed, 5 insertions(+), 3 deletions(-)

diff --git a/README.md b/README.md
index 26a578656c..7046a2af7c 100644
--- a/README.md
+++ b/README.md
@@ -109,18 +109,19 @@ You can use the same Docker image as Unsloth Studio.
 For RTX 50x, B200, 6000 GPUs: `uv pip install unsloth --torch-backend=auto`. Read our guides for: [Blackwell](https://unsloth.ai/docs/blog/fine-tuning-llms-with-blackwell-rtx-50-series-and-unsloth) and [DGX Spark](https://unsloth.ai/docs/blog/fine-tuning-llms-with-nvidia-dgx-spark-and-unsloth). <br>
 To install Unsloth on **AMD** and **Intel** GPUs, follow our [AMD Guide](https://unsloth.ai/docs/get-started/install/amd) and [Intel Guide](https://unsloth.ai/docs/get-started/install/intel).
 
-## ✨ Free Notebooks
+## 📒 Free Notebooks
 
-Train for free with our notebooks. Read our [guide](https://unsloth.ai/docs/get-started/fine-tuning-llms-guide). Add dataset, run, then deploy your trained model.
+Train for free with our notebooks. You can use our new [free Unsloth Studio notebook](https://colab.research.google.com/github/unslothai/unsloth/blob/main/studio/Unsloth_Studio_Colab.ipynb) to run and train models for free in a web UI.
+Read our [guide](https://unsloth.ai/docs/get-started/fine-tuning-llms-guide). Add dataset, run, then deploy your trained model.
 
 | Model | Free Notebooks | Performance | Memory use |
 |-----------|---------|--------|----------|
+| **Gemma 4 (E2B)**      | [▶️ Start for free](https://colab.research.google.com/github/unslothai/notebooks/blob/main/nb/Gemma4_(E2B)-Vision.ipynb)               | 1.5x faster | 50% less |
 | **Qwen3.5 (4B)**      | [▶️ Start for free](https://colab.research.google.com/github/unslothai/notebooks/blob/main/nb/Qwen3_5_(4B)_Vision.ipynb)               | 1.5x faster | 60% less |
 | **gpt-oss (20B)**      | [▶️ Start for free](https://colab.research.google.com/github/unslothai/notebooks/blob/main/nb/gpt-oss-(20B)-Fine-tuning.ipynb)               | 2x faster | 70% less |
 | **Qwen3.5 GSPO**      | [▶️ Start for free](https://colab.research.google.com/github/unslothai/notebooks/blob/main/nb/Qwen3_5_(4B)_Vision_GRPO.ipynb)               | 2x faster | 70% less |
 | **gpt-oss (20B): GRPO**      | [▶️ Start for free](https://colab.research.google.com/github/unslothai/notebooks/blob/main/nb/gpt-oss-(20B)-GRPO.ipynb)               | 2x faster | 80% less |
 | **Qwen3: Advanced GRPO**      | [▶️ Start for free](https://colab.research.google.com/github/unslothai/notebooks/blob/main/nb/Qwen3_(4B)-GRPO.ipynb)               | 2x faster | 70% less |
-| **Gemma 3 (4B) Vision** | [▶️ Start for free](https://colab.research.google.com/github/unslothai/notebooks/blob/main/nb/Gemma3_(4B)-Vision.ipynb)               | 1.7x faster | 60% less |
 | **embeddinggemma (300M)**    | [▶️ Start for free](https://colab.research.google.com/github/unslothai/notebooks/blob/main/nb/EmbeddingGemma_(300M).ipynb)               | 2x faster | 20% less |
 | **Mistral Ministral 3 (3B)**      | [▶️ Start for free](https://colab.research.google.com/github/unslothai/notebooks/blob/main/nb/Ministral_3_VL_(3B)_Vision.ipynb)               | 1.5x faster | 60% less |
 | **Llama 3.1 (8B) Alpaca**      | [▶️ Start for free](https://colab.research.google.com/github/unslothai/notebooks/blob/main/nb/Llama3.1_(8B)-Alpaca.ipynb)               | 2x faster | 70% less |
@@ -132,6 +133,7 @@ Train for free with our notebooks. Read our [guide](https://unsloth.ai/docs/get-
 - See detailed documentation for Unsloth [here](https://unsloth.ai/docs)
 
 ## 🦥 Unsloth News
+- **Gemma 4**: Run and train Google’s new models directly in Unsloth Studio! [Blog](https://unsloth.ai/docs/models/gemma-4)
 - **Introducing Unsloth Studio**: our new web UI for running and training LLMs. [Blog](https://unsloth.ai/docs/new/studio)
 - **Qwen3.5** - 0.8B, 2B, 4B, 9B, 27B, 35-A3B, 112B-A10B are now supported. [Guide + notebooks](https://unsloth.ai/docs/models/qwen3.5/fine-tune)
 - Train **MoE LLMs 12x faster** with 35% less VRAM - DeepSeek, GLM, Qwen and gpt-oss. [Blog](https://unsloth.ai/docs/new/faster-moe)

From ac562bac668347b70017dce7540fad1dd96989bc Mon Sep 17 00:00:00 2001
From: DoubleMathew <mmathew23@gmail.com>
Date: Fri, 3 Apr 2026 02:34:20 -0500
Subject: [PATCH 032/216] Fix/llama.cppbuilding (#4804)

* Simplify llama.cpp install logic

* print release tag

* Retry failed json decode

* don't pull all ggml releases

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

* Remove test file changes from main PR

Test changes for test_pr4562_bugfixes.py will be submitted in a separate PR to keep this PR focused on the install path simplification.

* Fix setup.sh executable bit and direct tag lookup for pinned releases

- Restore setup.sh file mode to 100755 (was accidentally changed to 100644)
- Add direct GitHub API tag lookup in iter_release_payloads_by_time for
  non-latest requested tags (e.g. b7879) instead of relying on paginated
  release scans that may miss older releases beyond the 5-page limit
- Update stale DEFAULT_PUBLISHED_REPO comment to match new value

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

* Fix force-compile default ref and remove dead code in setup.ps1

- Change FORCE_COMPILE_DEFAULT_REF from "main" to "master" in all three
  files (install_llama_prebuilt.py, setup.sh, setup.ps1) since
  ggml-org/llama.cpp uses "master" as its default branch, not "main".
  Using "main" would cause git clone --branch to fail when
  UNSLOTH_LLAMA_FORCE_COMPILE=1 with UNSLOTH_LLAMA_TAG=latest.
- Remove dead if ($SkipPrebuiltInstall) block inside the else branch of
  setup.ps1 that could never be reached (the outer elseif already
  handles $SkipPrebuiltInstall=true).
- Maintain setup.sh executable bit (100755).

* Improve iter_release_payloads_by_time error handling for direct tag lookup

When a pinned release tag is not found (HTTP 404), fall through to the
paginated release scan instead of silently returning empty results.
Non-404 errors (network failures, rate limits) are propagated to the
caller so users get actionable error messages.

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

---------

Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com>
Co-authored-by: Daniel Han <danielhanchen@gmail.com>
---
 studio/install_llama_prebuilt.py | 571 ++++++++++++++++++++++++++++---
 studio/setup.ps1                 | 160 ++++-----
 studio/setup.sh                  | 272 +++++++--------
 3 files changed, 711 insertions(+), 292 deletions(-)

diff --git a/studio/install_llama_prebuilt.py b/studio/install_llama_prebuilt.py
index 5e8fe314b4..8d06c7d0e1 100755
--- a/studio/install_llama_prebuilt.py
+++ b/studio/install_llama_prebuilt.py
@@ -65,9 +65,10 @@ def env_int(name: str, default: int, *, minimum: int | None = None) -> int:
 # errors. Only use "master" temporarily when the latest release is missing
 # support for a new model architecture.
 DEFAULT_LLAMA_TAG = os.environ.get("UNSLOTH_LLAMA_TAG", "latest")
-# Force all installs to use mainline llama.cpp from ggml-org.
-# Previously: DEFAULT_PUBLISHED_REPO = os.environ.get("UNSLOTH_LLAMA_RELEASE_REPO", "unslothai/llama.cpp")
-DEFAULT_PUBLISHED_REPO = "ggml-org/llama.cpp"
+# Default published repo for prebuilt release resolution. Linux uses
+# Unsloth prebuilts; setup.sh/setup.ps1 pass --published-repo explicitly
+# for macOS/Windows to override with ggml-org/llama.cpp when needed.
+DEFAULT_PUBLISHED_REPO = "unslothai/llama.cpp"
 DEFAULT_PUBLISHED_TAG = os.environ.get("UNSLOTH_LLAMA_RELEASE_TAG")
 DEFAULT_PUBLISHED_MANIFEST_ASSET = os.environ.get(
     "UNSLOTH_LLAMA_RELEASE_MANIFEST_ASSET", "llama-prebuilt-manifest.json"
@@ -89,6 +90,12 @@ GITHUB_AUTH_HOSTS = {"api.github.com", "github.com"}
 RETRYABLE_HTTP_STATUS = {408, 429, 500, 502, 503, 504}
 HTTP_FETCH_ATTEMPTS = 4
 HTTP_FETCH_BASE_DELAY_SECONDS = 0.75
+JSON_FETCH_ATTEMPTS = 3
+DEFAULT_GITHUB_RELEASE_SCAN_MAX_PAGES = env_int(
+    "UNSLOTH_LLAMA_GITHUB_RELEASE_SCAN_MAX_PAGES",
+    5,
+    minimum = 1,
+)
 SERVER_PORT_BIND_ATTEMPTS = 3
 SERVER_BIND_RETRY_WINDOW_SECONDS = 5.0
 TTY_PROGRESS_START_DELAY_SECONDS = 0.5
@@ -97,6 +104,58 @@ DEFAULT_MAX_PREBUILT_RELEASE_FALLBACKS = env_int(
     2,
     minimum = 1,
 )
+FORCE_COMPILE_DEFAULT_REF = os.environ.get("UNSLOTH_LLAMA_FORCE_COMPILE_REF", "master")
+
+DIRECT_LINUX_BUNDLE_PROFILES: dict[str, dict[str, Any]] = {
+    "cuda12-older": {
+        "runtime_line": "cuda12",
+        "coverage_class": "older",
+        "supported_sms": ["70", "75", "80", "86", "89"],
+        "min_sm": 70,
+        "max_sm": 89,
+        "rank": 10,
+    },
+    "cuda12-newer": {
+        "runtime_line": "cuda12",
+        "coverage_class": "newer",
+        "supported_sms": ["86", "89", "90", "100", "120"],
+        "min_sm": 86,
+        "max_sm": 120,
+        "rank": 20,
+    },
+    "cuda12-portable": {
+        "runtime_line": "cuda12",
+        "coverage_class": "portable",
+        "supported_sms": ["70", "75", "80", "86", "89", "90", "100", "120"],
+        "min_sm": 70,
+        "max_sm": 120,
+        "rank": 30,
+    },
+    "cuda13-older": {
+        "runtime_line": "cuda13",
+        "coverage_class": "older",
+        "supported_sms": ["75", "80", "86", "89"],
+        "min_sm": 75,
+        "max_sm": 89,
+        "rank": 40,
+    },
+    "cuda13-newer": {
+        "runtime_line": "cuda13",
+        "coverage_class": "newer",
+        "supported_sms": ["86", "89", "90", "100", "120"],
+        "min_sm": 86,
+        "max_sm": 120,
+        "rank": 50,
+    },
+    "cuda13-portable": {
+        "runtime_line": "cuda13",
+        "coverage_class": "portable",
+        "supported_sms": ["75", "80", "86", "89", "90", "100", "120"],
+        "min_sm": 75,
+        "max_sm": 120,
+        "rank": 60,
+    },
+}
 
 
 @dataclass
@@ -753,32 +812,48 @@ def download_bytes(
 
 
 def fetch_json(url: str) -> Any:
-    try:
-        data = download_bytes(
-            url,
-            timeout = 30,
-            headers = github_api_headers(url)
-            if is_github_api_url(url)
-            else auth_headers(url),
-        )
-    except urllib.error.HTTPError as exc:
-        if exc.code == 403 and is_github_api_url(url):
-            hint = ""
-            if not (os.environ.get("GH_TOKEN") or os.environ.get("GITHUB_TOKEN")):
-                hint = "; set GH_TOKEN or GITHUB_TOKEN to avoid GitHub API rate limits"
-            raise RuntimeError(f"GitHub API returned 403 for {url}{hint}") from exc
-        raise
-    if not data:
-        raise RuntimeError(f"downloaded empty JSON payload from {url}")
-    try:
-        payload = json.loads(data.decode("utf-8"))
-    except (UnicodeDecodeError, json.JSONDecodeError) as exc:
-        raise RuntimeError(f"downloaded invalid JSON from {url}: {exc}") from exc
-    if not isinstance(payload, dict) and not isinstance(payload, list):
-        raise RuntimeError(
-            f"downloaded unexpected JSON type from {url}: {type(payload).__name__}"
-        )
-    return payload
+    attempts = JSON_FETCH_ATTEMPTS if is_github_api_url(url) else 1
+    last_decode_exc: Exception | None = None
+    for attempt in range(1, attempts + 1):
+        try:
+            data = download_bytes(
+                url,
+                timeout = 30,
+                headers = github_api_headers(url)
+                if is_github_api_url(url)
+                else auth_headers(url),
+            )
+        except urllib.error.HTTPError as exc:
+            if exc.code == 403 and is_github_api_url(url):
+                hint = ""
+                if not (os.environ.get("GH_TOKEN") or os.environ.get("GITHUB_TOKEN")):
+                    hint = (
+                        "; set GH_TOKEN or GITHUB_TOKEN to avoid GitHub API rate limits"
+                    )
+                raise RuntimeError(f"GitHub API returned 403 for {url}{hint}") from exc
+            raise
+        if not data:
+            last_decode_exc = RuntimeError(f"downloaded empty JSON payload from {url}")
+        else:
+            try:
+                payload = json.loads(data.decode("utf-8"))
+            except (UnicodeDecodeError, json.JSONDecodeError) as exc:
+                last_decode_exc = RuntimeError(
+                    f"downloaded invalid JSON from {url}: {exc}"
+                )
+            else:
+                if not isinstance(payload, dict) and not isinstance(payload, list):
+                    raise RuntimeError(
+                        f"downloaded unexpected JSON type from {url}: {type(payload).__name__}"
+                    )
+                return payload
+        if attempt >= attempts:
+            assert last_decode_exc is not None
+            raise last_decode_exc
+        log(f"json fetch failed ({attempt}/{attempts}) for {url}; retrying")
+        sleep_backoff(attempt)
+    assert last_decode_exc is not None
+    raise last_decode_exc
 
 
 def download_file(url: str, destination: Path) -> None:
@@ -838,12 +913,16 @@ def download_file_verified(
     url: str,
     destination: Path,
     *,
-    expected_sha256: str,
+    expected_sha256: str | None,
     label: str,
 ) -> None:
     normalized_expected = normalize_sha256_digest(expected_sha256)
     if not normalized_expected:
-        raise PrebuiltFallback(f"{label} did not have a valid approved sha256")
+        download_file(url, destination)
+        log(
+            f"downloaded {label} without a published sha256; relying on install validation"
+        )
+        return
 
     for attempt in range(1, 3):
         download_file(url, destination)
@@ -898,7 +977,12 @@ def github_release(repo: str, tag: str) -> dict[str, Any]:
     return payload
 
 
-def github_releases(repo: str, *, per_page: int = 100) -> list[dict[str, Any]]:
+def github_releases(
+    repo: str,
+    *,
+    per_page: int = 100,
+    max_pages: int = 0,
+) -> list[dict[str, Any]]:
     releases: list[dict[str, Any]] = []
     page = 1
     while True:
@@ -912,6 +996,8 @@ def github_releases(repo: str, *, per_page: int = 100) -> list[dict[str, Any]]:
         if len(payload) < per_page:
             break
         page += 1
+        if max_pages > 0 and page > max_pages:
+            break
     return releases
 
 
@@ -925,6 +1011,372 @@ def latest_upstream_release_tag() -> str:
     return tag
 
 
+def is_release_tag_like(value: str | None) -> bool:
+    return isinstance(value, str) and bool(re.fullmatch(r"b\d+", value.strip()))
+
+
+def release_time_sort_key(release: dict[str, Any]) -> tuple[str, int]:
+    published_at = release.get("published_at")
+    created_at = release.get("created_at")
+    release_id = release.get("id")
+    timestamp = (
+        published_at
+        if isinstance(published_at, str) and published_at
+        else created_at
+        if isinstance(created_at, str) and created_at
+        else ""
+    )
+    try:
+        normalized_id = int(release_id)
+    except (TypeError, ValueError):
+        normalized_id = 0
+    return (timestamp, normalized_id)
+
+
+def iter_release_payloads_by_time(
+    repo: str,
+    published_release_tag: str = "",
+    requested_tag: str = "",
+) -> Iterable[dict[str, Any]]:
+    if published_release_tag:
+        yield github_release(repo, published_release_tag)
+        return
+
+    if (
+        requested_tag
+        and requested_tag != "latest"
+        and is_release_tag_like(requested_tag)
+    ):
+        try:
+            yield github_release(repo, requested_tag)
+            return
+        except urllib.error.HTTPError as exc:
+            if exc.code == 404:
+                log(
+                    f"release tag {requested_tag} not found in {repo}; scanning recent releases"
+                )
+            else:
+                raise
+        except Exception:
+            raise
+
+    releases = [
+        release
+        for release in github_releases(
+            repo, max_pages = DEFAULT_GITHUB_RELEASE_SCAN_MAX_PAGES
+        )
+        if isinstance(release, dict)
+        and not release.get("draft")
+        and not release.get("prerelease")
+    ]
+    releases.sort(key = release_time_sort_key, reverse = True)
+    for release in releases:
+        yield release
+
+
+def direct_release_matches_request(
+    *, release_tag: str, llama_tag: str, requested_tag: str
+) -> bool:
+    if requested_tag == "latest":
+        return True
+    for candidate in (release_tag, llama_tag):
+        if refs_match(candidate, requested_tag):
+            return True
+    return False
+
+
+def synthetic_checksums_for_release(
+    repo: str, release_tag: str, upstream_tag: str
+) -> ApprovedReleaseChecksums:
+    return ApprovedReleaseChecksums(
+        repo = repo,
+        release_tag = release_tag,
+        upstream_tag = upstream_tag,
+        artifacts = {},
+    )
+
+
+def parse_direct_linux_release_bundle(
+    repo: str, release: dict[str, Any]
+) -> PublishedReleaseBundle | None:
+    release_tag = release.get("tag_name")
+    if not isinstance(release_tag, str) or not release_tag:
+        return None
+
+    assets = release_asset_map(release)
+    artifacts: list[PublishedLlamaArtifact] = []
+    inferred_labels: list[str] = []
+
+    linux_asset_re = re.compile(
+        r"^app-(?P<label>.+)-(?P<target>linux-x64(?:-cpu)?|linux-x64-(?:cuda12|cuda13)-(?:older|newer|portable))\.tar\.gz$"
+    )
+    for asset_name in sorted(assets):
+        match = linux_asset_re.fullmatch(asset_name)
+        if not match:
+            continue
+        inferred_labels.append(match.group("label"))
+        target = match.group("target")
+        if target in {"linux-x64", "linux-x64-cpu"}:
+            artifacts.append(
+                PublishedLlamaArtifact(
+                    asset_name = asset_name,
+                    install_kind = "linux-cpu",
+                    runtime_line = None,
+                    coverage_class = None,
+                    supported_sms = [],
+                    min_sm = None,
+                    max_sm = None,
+                    bundle_profile = None,
+                    rank = 1000,
+                )
+            )
+            continue
+
+        bundle_profile = target.removeprefix("linux-x64-")
+        profile = DIRECT_LINUX_BUNDLE_PROFILES.get(bundle_profile)
+        if profile is None:
+            continue
+        artifacts.append(
+            PublishedLlamaArtifact(
+                asset_name = asset_name,
+                install_kind = "linux-cuda",
+                runtime_line = str(profile["runtime_line"]),
+                coverage_class = str(profile["coverage_class"]),
+                supported_sms = [str(value) for value in profile["supported_sms"]],
+                min_sm = int(profile["min_sm"]),
+                max_sm = int(profile["max_sm"]),
+                bundle_profile = bundle_profile,
+                rank = int(profile["rank"]),
+            )
+        )
+
+    if not artifacts:
+        return None
+
+    upstream_tag = (
+        release_tag
+        if is_release_tag_like(release_tag)
+        else inferred_labels[0]
+        if len(set(inferred_labels)) == 1 and inferred_labels
+        else release_tag
+    )
+    selection_log = [
+        f"published_release: repo={repo}",
+        f"published_release: tag={release_tag}",
+        f"published_release: upstream_tag={upstream_tag}",
+        "published_release: direct_asset_scan=linux",
+    ]
+    return PublishedReleaseBundle(
+        repo = repo,
+        release_tag = release_tag,
+        upstream_tag = upstream_tag,
+        assets = assets,
+        manifest_asset_name = DEFAULT_PUBLISHED_MANIFEST_ASSET,
+        artifacts = artifacts,
+        selection_log = selection_log,
+    )
+
+
+def direct_linux_release_plan(
+    release: dict[str, Any],
+    host: HostInfo,
+    repo: str,
+    requested_tag: str,
+) -> InstallReleasePlan | None:
+    bundle = parse_direct_linux_release_bundle(repo, release)
+    if bundle is None:
+        return None
+    if not direct_release_matches_request(
+        release_tag = bundle.release_tag,
+        llama_tag = bundle.upstream_tag,
+        requested_tag = requested_tag,
+    ):
+        return None
+
+    attempts: list[AssetChoice] = []
+    if host.has_usable_nvidia:
+        selection = linux_cuda_choice_from_release(host, bundle)
+        if selection is not None:
+            attempts.extend(selection.attempts)
+    cpu_choice = published_asset_choice_for_kind(bundle, "linux-cpu")
+    if cpu_choice is not None:
+        attempts.append(cpu_choice)
+    if not attempts:
+        raise PrebuiltFallback("no compatible Linux prebuilt asset was found")
+    return InstallReleasePlan(
+        requested_tag = requested_tag,
+        llama_tag = bundle.upstream_tag,
+        release_tag = bundle.release_tag,
+        attempts = attempts,
+        approved_checksums = synthetic_checksums_for_release(
+            repo,
+            bundle.release_tag,
+            bundle.upstream_tag,
+        ),
+    )
+
+
+def direct_upstream_release_plan(
+    release: dict[str, Any],
+    host: HostInfo,
+    repo: str,
+    requested_tag: str,
+) -> InstallReleasePlan | None:
+    release_tag = release.get("tag_name")
+    if not isinstance(release_tag, str) or not release_tag:
+        return None
+    if not direct_release_matches_request(
+        release_tag = release_tag,
+        llama_tag = release_tag,
+        requested_tag = requested_tag,
+    ):
+        return None
+
+    assets = release_asset_map(release)
+    attempts: list[AssetChoice] = []
+    if host.is_windows and host.is_x86_64:
+        if host.has_usable_nvidia:
+            torch_preference = detect_torch_cuda_runtime_preference(host)
+            attempts.extend(
+                windows_cuda_attempts(
+                    host,
+                    release_tag,
+                    assets,
+                    torch_preference.runtime_line,
+                    torch_preference.selection_log,
+                )
+            )
+        cpu_asset = f"llama-{release_tag}-bin-win-cpu-x64.zip"
+        cpu_url = assets.get(cpu_asset)
+        if cpu_url:
+            attempts.append(
+                AssetChoice(
+                    repo = repo,
+                    tag = release_tag,
+                    name = cpu_asset,
+                    url = cpu_url,
+                    source_label = "upstream",
+                    install_kind = "windows-cpu",
+                )
+            )
+    elif host.is_macos and host.is_arm64:
+        asset_name = f"llama-{release_tag}-bin-macos-arm64.tar.gz"
+        asset_url = assets.get(asset_name)
+        if asset_url:
+            attempts.append(
+                AssetChoice(
+                    repo = repo,
+                    tag = release_tag,
+                    name = asset_name,
+                    url = asset_url,
+                    source_label = "upstream",
+                    install_kind = "macos-arm64",
+                )
+            )
+    elif host.is_macos and host.is_x86_64:
+        asset_name = f"llama-{release_tag}-bin-macos-x64.tar.gz"
+        asset_url = assets.get(asset_name)
+        if asset_url:
+            attempts.append(
+                AssetChoice(
+                    repo = repo,
+                    tag = release_tag,
+                    name = asset_name,
+                    url = asset_url,
+                    source_label = "upstream",
+                    install_kind = "macos-x64",
+                )
+            )
+    elif host.is_linux and host.is_x86_64 and not host.has_usable_nvidia:
+        asset_name = f"llama-{release_tag}-bin-ubuntu-x64.tar.gz"
+        asset_url = assets.get(asset_name)
+        if asset_url:
+            attempts.append(
+                AssetChoice(
+                    repo = repo,
+                    tag = release_tag,
+                    name = asset_name,
+                    url = asset_url,
+                    source_label = "upstream",
+                    install_kind = "linux-cpu",
+                )
+            )
+    if not attempts:
+        raise PrebuiltFallback("no compatible upstream prebuilt asset was found")
+    return InstallReleasePlan(
+        requested_tag = requested_tag,
+        llama_tag = release_tag,
+        release_tag = release_tag,
+        attempts = attempts,
+        approved_checksums = synthetic_checksums_for_release(
+            repo,
+            release_tag,
+            release_tag,
+        ),
+    )
+
+
+def resolve_simple_install_release_plans(
+    llama_tag: str,
+    host: HostInfo,
+    published_repo: str,
+    published_release_tag: str,
+    *,
+    max_release_fallbacks: int = DEFAULT_MAX_PREBUILT_RELEASE_FALLBACKS,
+) -> tuple[str, list[InstallReleasePlan]]:
+    repo = published_repo or DEFAULT_PUBLISHED_REPO
+    requested_tag = normalized_requested_llama_tag(llama_tag)
+    allow_older_release_fallback = (
+        requested_tag == "latest" and not published_release_tag
+    )
+    release_limit = max(1, max_release_fallbacks)
+    plans: list[InstallReleasePlan] = []
+    last_error: PrebuiltFallback | None = None
+
+    try:
+        releases = iter_release_payloads_by_time(
+            repo, published_release_tag, requested_tag
+        )
+        for release in releases:
+            try:
+                if host.is_linux and repo == "unslothai/llama.cpp":
+                    plan = direct_linux_release_plan(release, host, repo, requested_tag)
+                else:
+                    plan = direct_upstream_release_plan(
+                        release, host, repo, requested_tag
+                    )
+                if plan is None:
+                    continue
+            except PrebuiltFallback as exc:
+                last_error = exc
+                if not allow_older_release_fallback:
+                    raise
+                release_tag = release.get("tag_name") or "unknown"
+                log(
+                    "published release skipped for install planning: "
+                    f"{repo}@{release_tag} ({exc})"
+                )
+                continue
+
+            plans.append(plan)
+            if not allow_older_release_fallback or len(plans) >= release_limit:
+                break
+    except PrebuiltFallback:
+        raise
+    except Exception as exc:
+        raise PrebuiltFallback(
+            f"failed to inspect published releases in {repo}: {exc}"
+        ) from exc
+
+    if plans:
+        return requested_tag, plans
+    if last_error is not None:
+        raise last_error
+    raise PrebuiltFallback(
+        f"no installable published llama.cpp releases were found in {repo}"
+    )
+
+
 def normalized_requested_llama_tag(requested_tag: str | None) -> str:
     if isinstance(requested_tag, str):
         normalized = requested_tag.strip()
@@ -1435,7 +1887,7 @@ def iter_published_release_bundles(
     releases = (
         [github_release(repo, published_release_tag)]
         if published_release_tag
-        else github_releases(repo)
+        else github_releases(repo, max_pages = DEFAULT_GITHUB_RELEASE_SCAN_MAX_PAGES)
     )
     for release in releases:
         if not published_release_tag and (
@@ -1669,7 +2121,9 @@ def latest_published_linux_cuda_tag(host: HostInfo, published_repo: str) -> str
 
 
 def iter_upstream_releases() -> Iterable[dict[str, Any]]:
-    for release in github_releases(UPSTREAM_REPO):
+    for release in github_releases(
+        UPSTREAM_REPO, max_pages = DEFAULT_GITHUB_RELEASE_SCAN_MAX_PAGES
+    ):
         if release.get("draft") or release.get("prerelease"):
             continue
         yield release
@@ -2799,7 +3253,7 @@ def hydrate_source_tree(
     work_dir: Path,
     *,
     source_repo: str = UPSTREAM_REPO,
-    expected_sha256: str,
+    expected_sha256: str | None,
     source_label: str | None = None,
     exact_source: bool = False,
 ) -> None:
@@ -3231,10 +3685,6 @@ def install_from_archives(
 ) -> tuple[Path, Path]:
     main_archive = work_dir / choice.name
     log(f"downloading {choice.name} from {choice.source_label} release")
-    if not choice.expected_sha256:
-        raise PrebuiltFallback(
-            f"approved checksum was missing for selected asset {choice.name}"
-        )
     download_file_verified(
         choice.url,
         main_archive,
@@ -3962,7 +4412,7 @@ def require_approved_source_hash(
 
 def preferred_source_archive(
     checksums: ApprovedReleaseChecksums, llama_tag: str
-) -> tuple[str, str, ApprovedArtifactHash, bool]:
+) -> tuple[str, str, ApprovedArtifactHash | None, bool]:
     exact_source = exact_source_archive_hash(checksums)
     exact_repo = repo_slug_from_source(checksums.source_repo) or repo_slug_from_source(
         checksums.source_repo_url
@@ -3974,7 +4424,7 @@ def preferred_source_archive(
             exact_source,
             True,
         )
-    legacy = require_approved_source_hash(checksums, llama_tag)
+    legacy = checksums.artifacts.get(source_archive_logical_name(llama_tag))
     return (
         UPSTREAM_REPO,
         llama_tag,
@@ -3990,6 +4440,8 @@ def selected_source_archive_metadata(
     _source_repo, _source_ref, source_archive, _exact_source = preferred_source_archive(
         checksums, llama_tag
     )
+    if source_archive is None:
+        return source_archive_logical_name(llama_tag), None
     return source_archive.asset_name, source_archive.sha256
 
 
@@ -4153,8 +4605,6 @@ def expected_install_fingerprint(
     choice: AssetChoice,
     approved_checksums: ApprovedReleaseChecksums,
 ) -> str | None:
-    if not choice.expected_sha256:
-        return None
     source_asset_name, source_sha256 = selected_source_archive_metadata(
         approved_checksums,
         llama_tag,
@@ -4352,7 +4802,7 @@ def validate_prebuilt_choice(
         install_dir,
         work_dir,
         source_repo = source_repo,
-        expected_sha256 = source_archive.sha256,
+        expected_sha256 = source_archive.sha256 if source_archive is not None else None,
         source_label = (
             f"llama.cpp source tree for {source_repo}@{source_ref}"
             if exact_source
@@ -4477,7 +4927,12 @@ def validate_prebuilt_attempts(
 
 
 def install_prebuilt(
-    install_dir: Path, llama_tag: str, published_repo: str, published_release_tag: str
+    install_dir: Path,
+    llama_tag: str,
+    published_repo: str,
+    published_release_tag: str,
+    *,
+    simple_policy: bool = False,
 ) -> None:
     host = detect_host()
     choice: AssetChoice | None = None
@@ -4491,12 +4946,20 @@ def install_prebuilt(
                 log(
                     f"no existing llama.cpp install detected at {install_dir}; performing fresh prebuilt install"
                 )
-            requested_tag, release_plans = resolve_install_release_plans(
-                llama_tag,
-                host,
-                published_repo,
-                published_release_tag,
-            )
+            if simple_policy:
+                requested_tag, release_plans = resolve_simple_install_release_plans(
+                    llama_tag,
+                    host,
+                    published_repo,
+                    published_release_tag,
+                )
+            else:
+                requested_tag, release_plans = resolve_install_release_plans(
+                    llama_tag,
+                    host,
+                    published_repo,
+                    published_release_tag,
+                )
             if release_plans and existing_install_matches_plan(
                 install_dir, host, release_plans[0]
             ):
@@ -4600,6 +5063,11 @@ def parse_args() -> argparse.Namespace:
             "until a usable published llama.cpp release bundle is found."
         ),
     )
+    parser.add_argument(
+        "--simple-policy",
+        action = "store_true",
+        help = "Use the simplified platform-specific prebuilt selection policy.",
+    )
     resolve_group = parser.add_mutually_exclusive_group()
     resolve_group.add_argument(
         "--resolve-llama-tag",
@@ -4719,6 +5187,7 @@ def main() -> int:
         llama_tag = args.llama_tag,
         published_repo = args.published_repo,
         published_release_tag = args.published_release_tag or "",
+        simple_policy = args.simple_policy,
     )
     return EXIT_SUCCESS
 
diff --git a/studio/setup.ps1 b/studio/setup.ps1
index d478319098..60dfce0661 100644
--- a/studio/setup.ps1
+++ b/studio/setup.ps1
@@ -34,6 +34,7 @@ $PackageDir = Split-Path -Parent $ScriptDir
 $DefaultLlamaPrForce = ""
 $DefaultLlamaSource = "https://github.com/ggml-org/llama.cpp"
 $DefaultLlamaTag = "latest"
+$DefaultLlamaForceCompileRef = "master"
 
 # Verbose can be enabled either by CLI flag or by UNSLOTH_VERBOSE=1.
 $script:UnslothVerbose = ($env:UNSLOTH_VERBOSE -eq '1')
@@ -81,6 +82,31 @@ function New-UnslothTemporaryFile {
     return Get-Item -LiteralPath $tempPath
 }
 
+function Get-InstalledLlamaPrebuiltRelease {
+    param([string]$InstallDir)
+
+    $metadataPath = Join-Path $InstallDir "UNSLOTH_PREBUILT_INFO.json"
+    if (-not (Test-Path $metadataPath)) {
+        return $null
+    }
+
+    try {
+        $payload = Get-Content $metadataPath -Raw | ConvertFrom-Json
+    } catch {
+        return $null
+    }
+
+    if (-not $payload.published_repo -or -not $payload.release_tag) {
+        return $null
+    }
+
+    $message = "installed release: $($payload.published_repo)@$($payload.release_tag)"
+    if ($payload.tag -and $payload.tag -ne $payload.release_tag) {
+        $message += " (tag $($payload.tag))"
+    }
+    return $message
+}
+
 # Find nvcc on PATH, CUDA_PATH, or standard toolkit dirs.
 # Returns the path to nvcc.exe, or $null if not found.
 function Find-Nvcc {
@@ -133,7 +159,9 @@ function Find-Nvcc {
 
     # 3. Scan standard toolkit directory
     if (Test-Path $toolkitBase) {
-        $latest = Get-ChildItem -Directory $toolkitBase | Sort-Object Name | Select-Object -Last 1
+        $latest = Get-ChildItem -Directory $toolkitBase | Where-Object {
+            $_.Name -match '^v(\d+)\.(\d+)'
+        } | Sort-Object { [version]($_.Name -replace '^v','') } -Descending | Select-Object -First 1
         if ($latest -and (Test-Path (Join-Path $latest.FullName 'bin\nvcc.exe'))) {
             return (Join-Path $latest.FullName 'bin\nvcc.exe')
         }
@@ -1618,6 +1646,7 @@ if ($LlamaSource.EndsWith('.git')) { $LlamaSource = $LlamaSource.Substring(0, $L
 $ResolvedSourceUrl = $LlamaSource
 $ResolvedSourceRef = $RequestedLlamaTag
 $ResolvedSourceRefKind = "tag"
+$ResolvedLlamaTag = $RequestedLlamaTag
 
 if ($env:UNSLOTH_LLAMA_FORCE_COMPILE -eq "1") {
     $NeedLlamaSourceBuild = $true
@@ -1693,92 +1722,27 @@ if ($LlamaPr) {
     $ResolvedSourceRefKind = "pull"
     $NeedLlamaSourceBuild = $true
     $SkipPrebuiltInstall = $true
-} elseif ($SkipPrebuiltInstall) {
-    # Custom source or other override already forced source build; skip the
-    # prebuilt release resolution. When building from a custom fork, the fork
-    # may not carry upstream bNNNN tags.
-    if ($env:UNSLOTH_LLAMA_FORCE_COMPILE -eq "1") {
-        $ResolvedLlamaTag = $RequestedLlamaTag
-    } elseif ($LlamaSource -eq "https://github.com/ggml-org/llama.cpp") {
-        $resolveTagArgs = @("--resolve-llama-tag", $RequestedLlamaTag, "--published-repo", $HelperReleaseRepo, "--output-format", "json")
-        if ($env:UNSLOTH_LLAMA_RELEASE_TAG) { $resolveTagArgs += @("--published-release-tag", $env:UNSLOTH_LLAMA_RELEASE_TAG) }
-        $fallbackResult = Invoke-LlamaHelper -Arguments $resolveTagArgs
-        $fallbackOutput = $fallbackResult.Output
-        $fallbackExit = $fallbackResult.ExitCode
-        $ResolvedLlamaTag = if ($fallbackExit -eq 0 -and $fallbackOutput) {
-            try {
-                (($fallbackOutput | Out-String) | ConvertFrom-Json).llama_tag
-            } catch {
-                $RequestedLlamaTag
-            }
-        } else {
-            $RequestedLlamaTag
-        }
-    } else {
-        $ResolvedLlamaTag = $RequestedLlamaTag
-    }
-} else {
-    $resolveInstallArgs = @("--resolve-install-tag", $RequestedLlamaTag, "--published-repo", $HelperReleaseRepo, "--output-format", "json")
-    if ($env:UNSLOTH_LLAMA_RELEASE_TAG) { $resolveInstallArgs += @("--published-release-tag", $env:UNSLOTH_LLAMA_RELEASE_TAG) }
-    $resolveErrorLog = New-UnslothTemporaryFile
-    $resolveResult = Invoke-LlamaHelper -Arguments $resolveInstallArgs -StderrPath $resolveErrorLog
-    $resolveOutput = $resolveResult.Output
-    $resolveExit = $resolveResult.ExitCode
-    $ResolvedLlamaTag = if ($resolveOutput) {
-        try {
-            (($resolveOutput | Out-String) | ConvertFrom-Json).llama_tag
-        } catch {
-            ""
-        }
-    } else { "" }
-    if ($resolveExit -ne 0 -or [string]::IsNullOrWhiteSpace($ResolvedLlamaTag)) {
-        Write-Host ""
-        substep "Failed to resolve a published llama.cpp release via $HelperReleaseRepo" "Yellow"
-        Write-LlamaFailureLog -Output (Get-Content -Raw $resolveErrorLog)
-        # Resolve the llama.cpp tag for source-build fallback. Pass --published-repo
-        # so the resolver prefers the latest usable Unsloth-published upstream tag
-        # before falling back to the bleeding-edge ggml-org/llama.cpp tag.
-        $resolveFallbackArgs = @("--resolve-llama-tag", $RequestedLlamaTag, "--published-repo", $HelperReleaseRepo, "--output-format", "json")
-        if ($env:UNSLOTH_LLAMA_RELEASE_TAG) { $resolveFallbackArgs += @("--published-release-tag", $env:UNSLOTH_LLAMA_RELEASE_TAG) }
-        $fallbackResult = Invoke-LlamaHelper -Arguments $resolveFallbackArgs
-        $fallbackOutput = $fallbackResult.Output
-        $fallbackExit = $fallbackResult.ExitCode
-        $ResolvedLlamaTag = if ($fallbackExit -eq 0 -and $fallbackOutput) {
-            try {
-                (($fallbackOutput | Out-String) | ConvertFrom-Json).llama_tag
-            } catch {
-                $RequestedLlamaTag
-            }
-        } else {
-            $RequestedLlamaTag
-        }
-        $NeedLlamaSourceBuild = $true
-        $SkipPrebuiltInstall = $true
-    }
-    Remove-Item $resolveErrorLog -Force -ErrorAction SilentlyContinue
 }
 
-Write-Host ""
-substep "Resolved llama.cpp release tag: $ResolvedLlamaTag"
-
 if ($env:UNSLOTH_LLAMA_FORCE_COMPILE -eq "1") {
     Write-Host ""
     substep "UNSLOTH_LLAMA_FORCE_COMPILE=1 -- skipping prebuilt llama.cpp install" "Yellow"
     $NeedLlamaSourceBuild = $true
+} elseif ($SkipPrebuiltInstall) {
+    Write-Host ""
+    substep "Skipping prebuilt install -- falling back to source build" "Yellow"
 } else {
     Write-Host ""
     substep "installing prebuilt llama.cpp bundle (preferred path)..."
     if (Test-Path $LlamaCppDir) {
         substep "Existing llama.cpp install detected -- validating staged prebuilt update before replacement"
     }
-    if ($SkipPrebuiltInstall) {
-        substep "Skipping prebuilt install because prebuilt tag resolution failed -- falling back to source build" "Yellow"
-    } else {
-        $prebuiltArgs = @(
+    $prebuiltArgs = @(
             "$PSScriptRoot\install_llama_prebuilt.py",
             "--install-dir", $LlamaCppDir,
             "--llama-tag", $RequestedLlamaTag,
-            "--published-repo", $HelperReleaseRepo
+            "--published-repo", $HelperReleaseRepo,
+            "--simple-policy"
         )
         if ($env:UNSLOTH_LLAMA_RELEASE_TAG) {
             $prebuiltArgs += @("--published-release-tag", $env:UNSLOTH_LLAMA_RELEASE_TAG)
@@ -1817,6 +1781,10 @@ if ($env:UNSLOTH_LLAMA_FORCE_COMPILE -eq "1") {
             } else {
                 step "llama.cpp" "prebuilt installed and validated"
             }
+            $installedRelease = Get-InstalledLlamaPrebuiltRelease -InstallDir $LlamaCppDir
+            if ($installedRelease) {
+                substep $installedRelease
+            }
         } elseif ($prebuiltExit -eq 3) {
             step "llama.cpp" "install blocked by active llama.cpp process" "Yellow"
             Write-LlamaFailureLog -Output $prebuiltOutput
@@ -1834,7 +1802,6 @@ if ($env:UNSLOTH_LLAMA_FORCE_COMPILE -eq "1") {
             substep "Prebuilt llama.cpp path unavailable or failed validation -- falling back to source build" "Yellow"
             $NeedLlamaSourceBuild = $true
         }
-    }
 }
 
 # ==========================================================================
@@ -1981,24 +1948,43 @@ if (-not $NeedLlamaSourceBuild) {
     }
 
     if (-not $LlamaPr) {
-        if ($LlamaSource -eq "https://github.com/ggml-org/llama.cpp") {
-            $resolveSourceArgs = @("--resolve-source-build", $RequestedLlamaTag, "--published-repo", $HelperReleaseRepo, "--output-format", "json")
-            if ($env:UNSLOTH_LLAMA_RELEASE_TAG) { $resolveSourceArgs += @("--published-release-tag", $env:UNSLOTH_LLAMA_RELEASE_TAG) }
-            $sourcePlanResult = Invoke-LlamaHelper -Arguments $resolveSourceArgs
-            $sourcePlanOutput = $sourcePlanResult.Output
-            $sourcePlanExit = $sourcePlanResult.ExitCode
-            if ($sourcePlanExit -eq 0 -and $sourcePlanOutput) {
-                try {
-                    $sourcePlan = ($sourcePlanOutput | Out-String) | ConvertFrom-Json
-                    $ResolvedSourceUrl = $sourcePlan.source_url
-                    $ResolvedSourceRefKind = $sourcePlan.source_ref_kind
-                    $ResolvedSourceRef = $sourcePlan.source_ref
-                } catch {
+        $ResolvedSourceUrl = $LlamaSource
+        if ($env:UNSLOTH_LLAMA_FORCE_COMPILE -eq "1") {
+            if ($RequestedLlamaTag -eq "latest") {
+                $ResolvedSourceRef = if ($env:UNSLOTH_LLAMA_FORCE_COMPILE_REF) {
+                    $env:UNSLOTH_LLAMA_FORCE_COMPILE_REF
+                } else {
+                    $DefaultLlamaForceCompileRef
                 }
+                $ResolvedSourceRefKind = "branch"
+            } else {
+                $ResolvedSourceRef = $RequestedLlamaTag
+                $ResolvedSourceRefKind = "tag"
             }
+        } elseif ($RequestedLlamaTag -eq "latest") {
+            $resolveTagArgs = @("--resolve-llama-tag", "latest", "--published-repo", "ggml-org/llama.cpp", "--output-format", "json")
+            $resolveTagResult = Invoke-LlamaHelper -Arguments $resolveTagArgs
+            $resolveTagOutput = $resolveTagResult.Output
+            $resolveTagExit = $resolveTagResult.ExitCode
+            if ($resolveTagExit -eq 0 -and $resolveTagOutput) {
+                try {
+                    $ResolvedSourceRef = (($resolveTagOutput | Out-String) | ConvertFrom-Json).llama_tag
+                } catch {
+                    $ResolvedSourceRef = ""
+                }
+            } else {
+                $ResolvedSourceRef = ""
+            }
+            if ([string]::IsNullOrWhiteSpace($ResolvedSourceRef)) {
+                $ResolvedSourceRef = "latest"
+            }
+            $ResolvedSourceRefKind = "tag"
+        } else {
+            $ResolvedSourceRef = $RequestedLlamaTag
+            $ResolvedSourceRefKind = "tag"
         }
         if ([string]::IsNullOrWhiteSpace($ResolvedSourceUrl)) { $ResolvedSourceUrl = $LlamaSource }
-        if ([string]::IsNullOrWhiteSpace($ResolvedSourceRef)) { $ResolvedSourceRef = $ResolvedLlamaTag }
+        if ([string]::IsNullOrWhiteSpace($ResolvedSourceRef)) { $ResolvedSourceRef = $RequestedLlamaTag }
     }
 
     # -- Step A: Clone or pull llama.cpp --
diff --git a/studio/setup.sh b/studio/setup.sh
index d9ce73661f..e3ff2da35c 100755
--- a/studio/setup.sh
+++ b/studio/setup.sh
@@ -25,6 +25,7 @@ RULE=$(printf '\342\224\200%.0s' {1..52})
 _DEFAULT_LLAMA_PR_FORCE=""
 _DEFAULT_LLAMA_SOURCE="https://github.com/ggml-org/llama.cpp"
 _DEFAULT_LLAMA_TAG="latest"
+_DEFAULT_LLAMA_FORCE_COMPILE_REF="master"
 
 # ── Colors (same palette as startup_banner / install_python_stack) ──
 if [ -n "${NO_COLOR:-}" ]; then
@@ -121,6 +122,45 @@ print_llama_error_log() {
     tail -n 120 "$log_file" | sed 's/^/   | /' >&2
 }
 
+installed_llama_prebuilt_release() {
+    local install_dir=${1:-}
+    local metadata_path="$install_dir/UNSLOTH_PREBUILT_INFO.json"
+    [ -f "$metadata_path" ] || return 0
+    python - "$metadata_path" <<'PY' 2>/dev/null || true
+import json
+import sys
+from pathlib import Path
+
+try:
+    payload = json.loads(Path(sys.argv[1]).read_text(encoding="utf-8"))
+except Exception:
+    raise SystemExit(0)
+
+if not isinstance(payload, dict):
+    raise SystemExit(0)
+
+repo = str(payload.get("published_repo") or "").strip()
+release_tag = str(payload.get("release_tag") or "").strip()
+llama_tag = str(payload.get("tag") or "").strip()
+if not repo or not release_tag:
+    raise SystemExit(0)
+
+message = f"installed release: {repo}@{release_tag}"
+if llama_tag and llama_tag != release_tag:
+    message += f" (tag {llama_tag})"
+print(message)
+PY
+}
+
+print_installed_llama_prebuilt_release() {
+    local install_dir=${1:-}
+    local installed_release
+    installed_release="$(installed_llama_prebuilt_release "$install_dir")"
+    if [ -n "$installed_release" ]; then
+        substep "$installed_release"
+    fi
+}
+
 # ── Banner ──
 echo ""
 printf "  ${C_TITLE}%s${C_RST}\n" "🦥 Unsloth Studio Setup"
@@ -487,30 +527,27 @@ _NEED_LLAMA_SOURCE_BUILD=false
 _LLAMA_CPP_DEGRADED=false
 _LLAMA_FORCE_COMPILE="${UNSLOTH_LLAMA_FORCE_COMPILE:-0}"
 _REQUESTED_LLAMA_TAG="${UNSLOTH_LLAMA_TAG:-${_DEFAULT_LLAMA_TAG}}"
-# Force all installs to use mainline llama.cpp from ggml-org.
-_HELPER_RELEASE_REPO="ggml-org/llama.cpp"
+_HOST_SYSTEM="$(uname -s 2>/dev/null || true)"
+if [ "$_HOST_SYSTEM" = "Darwin" ]; then
+    _HELPER_RELEASE_REPO="ggml-org/llama.cpp"
+else
+    _HELPER_RELEASE_REPO="unslothai/llama.cpp"
+fi
 _LLAMA_PR="${UNSLOTH_LLAMA_PR:-}"
-
+_SKIP_PREBUILT_INSTALL=false
 _LLAMA_PR_FORCE="${UNSLOTH_LLAMA_PR_FORCE:-${_DEFAULT_LLAMA_PR_FORCE}}"
-# Force mainline source -- no env var override for now.
 _LLAMA_SOURCE="${_DEFAULT_LLAMA_SOURCE}"
 _LLAMA_SOURCE="${_LLAMA_SOURCE%.git}"  # normalize: strip trailing .git
 _RESOLVED_SOURCE_URL="$_LLAMA_SOURCE"
 _RESOLVED_SOURCE_REF="$_REQUESTED_LLAMA_TAG"
 _RESOLVED_SOURCE_REF_KIND="tag"
+_RESOLVED_LLAMA_TAG="$_REQUESTED_LLAMA_TAG"
 
 if [ "$_LLAMA_FORCE_COMPILE" = "1" ]; then
     _NEED_LLAMA_SOURCE_BUILD=true
     _SKIP_PREBUILT_INSTALL=true
 fi
 
-# Non-default source URL forces source build (fork has different code than prebuilt).
-if [ "$_LLAMA_SOURCE" != "https://github.com/ggml-org/llama.cpp" ]; then
-    step "llama.cpp" "custom source: $_LLAMA_SOURCE -- forcing source build" "$C_WARN"
-    _NEED_LLAMA_SOURCE_BUILD=true
-    _SKIP_PREBUILT_INSTALL=true
-fi
-
 # Baked-in PR_FORCE promotes to _LLAMA_PR when user hasn't set one.
 if [ -z "$_LLAMA_PR" ] && [ -n "$_LLAMA_PR_FORCE" ] && \
    [[ "$_LLAMA_PR_FORCE" =~ ^[0-9]+$ ]] && [ "$_LLAMA_PR_FORCE" -gt 0 ]; then
@@ -530,149 +567,68 @@ if [ -n "$_LLAMA_PR" ]; then
     _RESOLVED_SOURCE_REF_KIND="pull"
     _NEED_LLAMA_SOURCE_BUILD=true
     _SKIP_PREBUILT_INSTALL=true
-elif [ "${_SKIP_PREBUILT_INSTALL:-false}" = true ]; then
-    # Custom source or other override already forced source build; skip
-    # the prebuilt release resolution entirely. When building from a custom
-    # fork, the fork may not carry upstream bNNNN tags, so resolve the tag
-    # only when the source is the default ggml-org repo.
-    if [ "$_LLAMA_FORCE_COMPILE" = "1" ]; then
-        _RESOLVED_LLAMA_TAG="$_REQUESTED_LLAMA_TAG"
-    elif [ "$_LLAMA_SOURCE" = "https://github.com/ggml-org/llama.cpp" ]; then
-        _RESOLVE_TAG_ARGS=(--resolve-llama-tag "$_REQUESTED_LLAMA_TAG" --published-repo "$_HELPER_RELEASE_REPO")
-        _RESOLVE_TAG_ARGS+=(--output-format json)
-        if [ -n "${UNSLOTH_LLAMA_RELEASE_TAG:-}" ]; then
-            _RESOLVE_TAG_ARGS+=(--published-release-tag "$UNSLOTH_LLAMA_RELEASE_TAG")
-        fi
-        set +e
-        _RESOLVE_TAG_JSON="$(python "$SCRIPT_DIR/install_llama_prebuilt.py" "${_RESOLVE_TAG_ARGS[@]}" 2>/dev/null)"
-        _RESOLVE_UPSTREAM_STATUS=$?
-        set -e
-        if [ "$_RESOLVE_UPSTREAM_STATUS" -eq 0 ] && [ -n "${_RESOLVE_TAG_JSON:-}" ]; then
-            _RESOLVED_LLAMA_TAG="$(
-                printf '%s' "$_RESOLVE_TAG_JSON" | python -c 'import json,sys; print(json.load(sys.stdin).get("llama_tag",""))' 2>/dev/null || true
-            )"
-        else
-            _RESOLVED_LLAMA_TAG=""
-        fi
-        if [ -z "$_RESOLVED_LLAMA_TAG" ]; then
-            _RESOLVED_LLAMA_TAG="$_REQUESTED_LLAMA_TAG"
-        fi
-    else
-        _RESOLVED_LLAMA_TAG="$_REQUESTED_LLAMA_TAG"
-    fi
-else
-    _RESOLVE_INSTALL_ARGS=(--resolve-install-tag "$_REQUESTED_LLAMA_TAG" --published-repo "$_HELPER_RELEASE_REPO")
-    _RESOLVE_INSTALL_ARGS+=(--output-format json)
-    if [ -n "${UNSLOTH_LLAMA_RELEASE_TAG:-}" ]; then
-        _RESOLVE_INSTALL_ARGS+=(--published-release-tag "$UNSLOTH_LLAMA_RELEASE_TAG")
-    fi
-    _RESOLVE_LLAMA_LOG="$(mktemp)"
-    set +e
-    _RESOLVE_INSTALL_JSON="$(
-        python "$SCRIPT_DIR/install_llama_prebuilt.py" \
-            "${_RESOLVE_INSTALL_ARGS[@]}" 2>"$_RESOLVE_LLAMA_LOG"
-    )"
-    _RESOLVE_LLAMA_STATUS=$?
-    set -e
-    if [ "$_RESOLVE_LLAMA_STATUS" -eq 0 ]; then
-        _RESOLVED_LLAMA_TAG="$(
-            printf '%s' "${_RESOLVE_INSTALL_JSON:-}" | python -c 'import json,sys; print(json.load(sys.stdin).get("llama_tag",""))' 2>/dev/null || true
-        )"
-    else
-        _RESOLVED_LLAMA_TAG=""
-    fi
-    if [ -z "$_RESOLVED_LLAMA_TAG" ]; then
-        step "llama.cpp" "failed to resolve a published llama.cpp release via $_HELPER_RELEASE_REPO" "$C_WARN"
-        print_llama_error_log "$_RESOLVE_LLAMA_LOG"
-        set +e
-        # Resolve the llama.cpp tag for source-build fallback. Pass --published-repo
-        # so the resolver prefers the latest usable Unsloth-published upstream tag
-        # before falling back to the bleeding-edge ggml-org/llama.cpp tag.
-        _RESOLVE_FALLBACK_ARGS=(--resolve-llama-tag "$_REQUESTED_LLAMA_TAG" --published-repo "$_HELPER_RELEASE_REPO")
-        _RESOLVE_FALLBACK_ARGS+=(--output-format json)
-        if [ -n "${UNSLOTH_LLAMA_RELEASE_TAG:-}" ]; then
-            _RESOLVE_FALLBACK_ARGS+=(--published-release-tag "$UNSLOTH_LLAMA_RELEASE_TAG")
-        fi
-        _RESOLVE_FALLBACK_JSON="$(python "$SCRIPT_DIR/install_llama_prebuilt.py" "${_RESOLVE_FALLBACK_ARGS[@]}" 2>/dev/null)"
-        _RESOLVE_UPSTREAM_STATUS=$?
-        set -e
-        if [ "$_RESOLVE_UPSTREAM_STATUS" -eq 0 ] && [ -n "${_RESOLVE_FALLBACK_JSON:-}" ]; then
-            _RESOLVED_LLAMA_TAG="$(
-                printf '%s' "$_RESOLVE_FALLBACK_JSON" | python -c 'import json,sys; print(json.load(sys.stdin).get("llama_tag",""))' 2>/dev/null || true
-            )"
-        else
-            _RESOLVED_LLAMA_TAG=""
-        fi
-        if [ -z "$_RESOLVED_LLAMA_TAG" ]; then
-            _RESOLVED_LLAMA_TAG="$_REQUESTED_LLAMA_TAG"
-        fi
-        _NEED_LLAMA_SOURCE_BUILD=true
-        _SKIP_PREBUILT_INSTALL=true
-    fi
-    rm -f "$_RESOLVE_LLAMA_LOG"
 fi
 
-substep "resolved llama.cpp tag: $_RESOLVED_LLAMA_TAG"
 verbose_substep "requested llama.cpp tag: $_REQUESTED_LLAMA_TAG (repo: $_HELPER_RELEASE_REPO)"
 
 if [ "$_LLAMA_FORCE_COMPILE" = "1" ]; then
     step "llama.cpp" "UNSLOTH_LLAMA_FORCE_COMPILE=1 -- skipping prebuilt" "$C_WARN"
     _NEED_LLAMA_SOURCE_BUILD=true
+elif [ "${_SKIP_PREBUILT_INSTALL:-false}" = true ]; then
+    substep "prebuilt install skipped -- falling back to source build"
 else
     substep "installing prebuilt llama.cpp..."
     if [ -d "$LLAMA_CPP_DIR" ]; then
         substep "existing install detected -- validating update"
     fi
-    if [ "${_SKIP_PREBUILT_INSTALL:-false}" = true ]; then
-        substep "prebuilt tag resolution failed -- falling back to source build"
+    _PREBUILT_CMD=(
+        python "$SCRIPT_DIR/install_llama_prebuilt.py"
+        --install-dir "$LLAMA_CPP_DIR"
+        --llama-tag "$_REQUESTED_LLAMA_TAG"
+        --published-repo "$_HELPER_RELEASE_REPO"
+        --simple-policy
+    )
+    if [ -n "${UNSLOTH_LLAMA_RELEASE_TAG:-}" ]; then
+        _PREBUILT_CMD+=(--published-release-tag "$UNSLOTH_LLAMA_RELEASE_TAG")
+    fi
+    _PREBUILT_LOG="$(mktemp)"
+    set +e
+    if _is_verbose; then
+        "${_PREBUILT_CMD[@]}" 2>&1 | tee "$_PREBUILT_LOG"
+        _PREBUILT_STATUS=${PIPESTATUS[0]}
     else
-        _PREBUILT_CMD=(
-            python "$SCRIPT_DIR/install_llama_prebuilt.py"
-            --install-dir "$LLAMA_CPP_DIR"
-            --llama-tag "$_REQUESTED_LLAMA_TAG"
-            --published-repo "$_HELPER_RELEASE_REPO"
-        )
-        if [ -n "${UNSLOTH_LLAMA_RELEASE_TAG:-}" ]; then
-            _PREBUILT_CMD+=(--published-release-tag "$UNSLOTH_LLAMA_RELEASE_TAG")
-        fi
-        _PREBUILT_LOG="$(mktemp)"
-        set +e
-        if _is_verbose; then
-            "${_PREBUILT_CMD[@]}" 2>&1 | tee "$_PREBUILT_LOG"
-            _PREBUILT_STATUS=${PIPESTATUS[0]}
-        else
-            "${_PREBUILT_CMD[@]}" >"$_PREBUILT_LOG" 2>&1
-            _PREBUILT_STATUS=$?
-        fi
-        set -e
+        "${_PREBUILT_CMD[@]}" >"$_PREBUILT_LOG" 2>&1
+        _PREBUILT_STATUS=$?
+    fi
+    set -e
 
-        if [ "$_PREBUILT_STATUS" -eq 0 ]; then
-            if grep -Fq "already matches" "$_PREBUILT_LOG"; then
-                step "llama.cpp" "prebuilt up to date and validated"
-            else
-                step "llama.cpp" "prebuilt installed and validated"
-            fi
-            verbose_substep "llama.cpp install dir: $LLAMA_CPP_DIR"
-            rm -f "$_PREBUILT_LOG"
-        elif [ "$_PREBUILT_STATUS" -eq 3 ]; then
-            step "llama.cpp" "install blocked by active llama.cpp process" "$C_WARN"
-            print_llama_error_log "$_PREBUILT_LOG"
-            rm -f "$_PREBUILT_LOG"
-            if [ -d "$LLAMA_CPP_DIR" ]; then
-                substep "existing install was restored"
-            fi
-            substep "close Studio or other llama.cpp users and retry"
-            exit 3
+    if [ "$_PREBUILT_STATUS" -eq 0 ]; then
+        if grep -Fq "already matches" "$_PREBUILT_LOG"; then
+            step "llama.cpp" "prebuilt up to date and validated"
         else
-            step "llama.cpp" "prebuilt install failed (continuing)" "$C_WARN"
-            print_llama_error_log "$_PREBUILT_LOG"
-            rm -f "$_PREBUILT_LOG"
-            if [ -d "$LLAMA_CPP_DIR" ]; then
-                substep "prebuilt update failed; existing install restored"
-            fi
-            substep "falling back to source build"
-            _NEED_LLAMA_SOURCE_BUILD=true
+            step "llama.cpp" "prebuilt installed and validated"
         fi
+        print_installed_llama_prebuilt_release "$LLAMA_CPP_DIR"
+        verbose_substep "llama.cpp install dir: $LLAMA_CPP_DIR"
+        rm -f "$_PREBUILT_LOG"
+    elif [ "$_PREBUILT_STATUS" -eq 3 ]; then
+        step "llama.cpp" "install blocked by active llama.cpp process" "$C_WARN"
+        print_llama_error_log "$_PREBUILT_LOG"
+        rm -f "$_PREBUILT_LOG"
+        if [ -d "$LLAMA_CPP_DIR" ]; then
+            substep "existing install was restored"
+        fi
+        substep "close Studio or other llama.cpp users and retry"
+        exit 3
+    else
+        step "llama.cpp" "prebuilt install failed (continuing)" "$C_WARN"
+        print_llama_error_log "$_PREBUILT_LOG"
+        rm -f "$_PREBUILT_LOG"
+        if [ -d "$LLAMA_CPP_DIR" ]; then
+            substep "prebuilt update failed; existing install restored"
+        fi
+        substep "falling back to source build"
+        _NEED_LLAMA_SOURCE_BUILD=true
     fi
 fi
 
@@ -746,33 +702,41 @@ else
         [ -f "$LLAMA_SERVER_BIN" ] || _LLAMA_CPP_DEGRADED=true
     else
         if [ -z "$_LLAMA_PR" ]; then
-            if [ "$_LLAMA_SOURCE" = "https://github.com/ggml-org/llama.cpp" ]; then
-                _RESOLVE_SOURCE_ARGS=(--resolve-source-build "$_REQUESTED_LLAMA_TAG" --published-repo "$_HELPER_RELEASE_REPO")
-                _RESOLVE_SOURCE_ARGS+=(--output-format json)
-                if [ -n "${UNSLOTH_LLAMA_RELEASE_TAG:-}" ]; then
-                    _RESOLVE_SOURCE_ARGS+=(--published-release-tag "$UNSLOTH_LLAMA_RELEASE_TAG")
+            _RESOLVED_SOURCE_URL="$_LLAMA_SOURCE"
+            if [ "$_LLAMA_FORCE_COMPILE" = "1" ]; then
+                if [ "$_REQUESTED_LLAMA_TAG" = "latest" ]; then
+                    _RESOLVED_SOURCE_REF="${UNSLOTH_LLAMA_FORCE_COMPILE_REF:-${_DEFAULT_LLAMA_FORCE_COMPILE_REF}}"
+                    _RESOLVED_SOURCE_REF_KIND="branch"
+                else
+                    _RESOLVED_SOURCE_REF="$_REQUESTED_LLAMA_TAG"
+                    _RESOLVED_SOURCE_REF_KIND="tag"
                 fi
+            elif [ "$_REQUESTED_LLAMA_TAG" = "latest" ]; then
+                _RESOLVE_TAG_ARGS=(--resolve-llama-tag latest --published-repo "ggml-org/llama.cpp" --output-format json)
                 set +e
-                _SOURCE_BUILD_PLAN="$(python "$SCRIPT_DIR/install_llama_prebuilt.py" "${_RESOLVE_SOURCE_ARGS[@]}" 2>/dev/null)"
-                _RESOLVE_SOURCE_STATUS=$?
+                _RESOLVE_TAG_JSON="$(python "$SCRIPT_DIR/install_llama_prebuilt.py" "${_RESOLVE_TAG_ARGS[@]}" 2>/dev/null)"
+                _RESOLVE_TAG_STATUS=$?
                 set -e
-                if [ "$_RESOLVE_SOURCE_STATUS" -eq 0 ] && [ -n "$_SOURCE_BUILD_PLAN" ]; then
-                    _RESOLVED_SOURCE_URL="$(
-                        printf '%s' "$_SOURCE_BUILD_PLAN" | python -c 'import json,sys; print(json.load(sys.stdin).get("source_url",""))' 2>/dev/null || true
-                    )"
-                    _RESOLVED_SOURCE_REF_KIND="$(
-                        printf '%s' "$_SOURCE_BUILD_PLAN" | python -c 'import json,sys; print(json.load(sys.stdin).get("source_ref_kind",""))' 2>/dev/null || true
-                    )"
+                if [ "$_RESOLVE_TAG_STATUS" -eq 0 ] && [ -n "${_RESOLVE_TAG_JSON:-}" ]; then
                     _RESOLVED_SOURCE_REF="$(
-                        printf '%s' "$_SOURCE_BUILD_PLAN" | python -c 'import json,sys; print(json.load(sys.stdin).get("source_ref",""))' 2>/dev/null || true
+                        printf '%s' "$_RESOLVE_TAG_JSON" | python -c 'import json,sys; print(json.load(sys.stdin).get("llama_tag",""))' 2>/dev/null || true
                     )"
+                else
+                    _RESOLVED_SOURCE_REF=""
                 fi
+                if [ -z "$_RESOLVED_SOURCE_REF" ]; then
+                    _RESOLVED_SOURCE_REF="latest"
+                fi
+                _RESOLVED_SOURCE_REF_KIND="tag"
+            else
+                _RESOLVED_SOURCE_REF="$_REQUESTED_LLAMA_TAG"
+                _RESOLVED_SOURCE_REF_KIND="tag"
             fi
             if [ -z "$_RESOLVED_SOURCE_URL" ]; then
                 _RESOLVED_SOURCE_URL="$_LLAMA_SOURCE"
             fi
             if [ -z "$_RESOLVED_SOURCE_REF" ]; then
-                _RESOLVED_SOURCE_REF="$_RESOLVED_LLAMA_TAG"
+                _RESOLVED_SOURCE_REF="$_REQUESTED_LLAMA_TAG"
             fi
         fi
         verbose_substep "source build repo: $_RESOLVED_SOURCE_URL"

From 8981e6c8044aeed2c6debeeed6a79d98f3b08873 Mon Sep 17 00:00:00 2001
From: Daniel Han <danielhanchen@gmail.com>
Date: Fri, 3 Apr 2026 04:06:14 -0700
Subject: [PATCH 033/216] Update test_pr4562_bugfixes.py for simplified install
 policy (#4817)

- Add TestFetchJsonRetries for JSON retry logic and max_pages
- Update TestSourceCodePatterns for simplified --simple-policy flow
- Add tests for installed prebuilt release reporting
- Add test for CUDA toolkit version-sorted nvcc discovery
- Remove assertions for removed --resolve-install-tag / --resolve-source-build paths
---
 tests/studio/install/test_pr4562_bugfixes.py | 116 +++++++++++++++----
 1 file changed, 95 insertions(+), 21 deletions(-)

diff --git a/tests/studio/install/test_pr4562_bugfixes.py b/tests/studio/install/test_pr4562_bugfixes.py
index 7fa654d845..4a309036d6 100644
--- a/tests/studio/install/test_pr4562_bugfixes.py
+++ b/tests/studio/install/test_pr4562_bugfixes.py
@@ -14,10 +14,12 @@ Run: pytest tests/studio/install/test_pr4562_bugfixes.py -v
 """
 
 import importlib.util
+import json
 import os
 import subprocess
 import sys
 import textwrap
+import urllib.parse
 from pathlib import Path
 
 import pytest
@@ -347,6 +349,47 @@ class TestResolveRequestedLlamaTag:
         }
 
 
+class TestFetchJsonRetries:
+    def test_fetch_json_retries_invalid_github_api_json(
+        self, monkeypatch: pytest.MonkeyPatch
+    ):
+        calls = {"count": 0}
+
+        def fake_download_bytes(url, **kwargs):
+            calls["count"] += 1
+            if calls["count"] == 1:
+                return b'{"incomplete":"payload'
+            return json.dumps([{"tag_name": "b8635"}]).encode("utf-8")
+
+        monkeypatch.setattr(MOD, "download_bytes", fake_download_bytes)
+        monkeypatch.setattr(MOD, "sleep_backoff", lambda _attempt: None)
+
+        payload = MOD.fetch_json(
+            "https://api.github.com/repos/ggml-org/llama.cpp/releases?per_page=100&page=1"
+        )
+
+        assert isinstance(payload, list)
+        assert payload[0]["tag_name"] == "b8635"
+        assert calls["count"] == 2
+
+    def test_github_releases_honors_max_pages(self, monkeypatch: pytest.MonkeyPatch):
+        seen_pages: list[int] = []
+
+        def fake_fetch_json(url: str):
+            parsed = urllib.parse.urlparse(url)
+            params = urllib.parse.parse_qs(parsed.query)
+            page = int(params["page"][0])
+            seen_pages.append(page)
+            return [{"tag_name": f"b{page:04d}"} for _ in range(100)]
+
+        monkeypatch.setattr(MOD, "fetch_json", fake_fetch_json)
+
+        releases = MOD.github_releases("ggml-org/llama.cpp", max_pages = 2)
+
+        assert seen_pages == [1, 2]
+        assert len(releases) == 200
+
+
 # =========================================================================
 # TEST GROUP C: setup.sh logic (bash subprocess tests)
 # =========================================================================
@@ -643,25 +686,35 @@ class TestSourceCodePatterns:
             '_RESOLVED_SOURCE_REF" != "latest"' in content
         ), "Should guard against literal 'latest' tag"
 
-    def test_setup_sh_source_build_uses_helper_resolution(self):
-        """Shell source fallback should consult the helper for repo/ref planning."""
+    def test_setup_sh_source_build_uses_helper_latest_tag_only(self):
+        """Shell source fallback should only use helper latest-tag resolution."""
         content = SETUP_SH.read_text()
-        assert "--resolve-source-build" in content
+        assert "--resolve-source-build" not in content
+        assert "--resolve-install-tag" not in content
+        assert (
+            '--resolve-llama-tag latest --published-repo "ggml-org/llama.cpp"'
+            in content
+        )
         assert "--output-format json" in content
         assert "_RESOLVED_SOURCE_URL" in content
         assert "_RESOLVED_SOURCE_REF_KIND" in content
         assert "_RESOLVED_SOURCE_REF" in content
 
-    def test_setup_sh_latest_resolution_uses_helper_only(self):
-        """Shell fallback should rely on helper output, not raw GitHub API tag_name."""
+    def test_setup_sh_prebuilt_install_uses_simple_policy_only(self):
+        """Shell prebuilt path should use the simplified helper install entrypoint."""
         content = SETUP_SH.read_text()
-        assert "--resolve-install-tag" in content
-        assert "--resolve-llama-tag" in content
-        assert 'tail -n 1 "$_RESOLVE_LLAMA_LOG"' not in content
-        assert "json.load" in content
+        assert "--simple-policy" in content
+        assert "--resolve-install-tag" not in content
         assert "_HELPER_RELEASE_REPO}/releases/latest" not in content
         assert "ggml-org/llama.cpp/releases/latest" not in content
 
+    def test_setup_sh_reports_installed_prebuilt_release(self):
+        """Shell wrapper should report the installed prebuilt release from metadata."""
+        content = SETUP_SH.read_text()
+        assert "UNSLOTH_PREBUILT_INFO.json" in content
+        assert "installed release:" in content
+        assert 'print_installed_llama_prebuilt_release "$LLAMA_CPP_DIR"' in content
+
     def test_setup_sh_macos_arm64_uses_metal_flags(self):
         """Apple Silicon source builds should explicitly enable Metal like upstream."""
         content = SETUP_SH.read_text()
@@ -759,20 +812,34 @@ class TestSourceCodePatterns:
                         f"Found 'git pull' in llama.cpp build section at line {i+1}"
                     )
 
-    def test_setup_ps1_latest_resolution_uses_helper_only(self):
-        """PS1 fallback should rely on helper output, not raw GitHub API tag_name."""
+    def test_setup_ps1_prebuilt_install_uses_simple_policy_only(self):
+        """PS1 prebuilt path should use the simplified helper install entrypoint."""
         content = SETUP_PS1.read_text()
-        assert "--resolve-install-tag" in content
-        assert "--resolve-llama-tag" in content
-        assert '--output-format", "json"' in content
-        assert "ConvertFrom-Json" in content
+        assert '"--simple-policy"' in content
+        assert "--resolve-install-tag" not in content
         assert "$HelperReleaseRepo/releases/latest" not in content
         assert "ggml-org/llama.cpp/releases/latest" not in content
 
-    def test_setup_ps1_source_build_uses_helper_resolution(self):
-        """PS1 source fallback should consult the helper for repo/ref planning."""
+    def test_setup_ps1_reports_installed_prebuilt_release(self):
+        """PS1 wrapper should report the installed prebuilt release from metadata."""
         content = SETUP_PS1.read_text()
-        assert "--resolve-source-build" in content
+        assert "Get-InstalledLlamaPrebuiltRelease" in content
+        assert "UNSLOTH_PREBUILT_INFO.json" in content
+        assert "installed release:" in content
+        assert (
+            "$installedRelease = Get-InstalledLlamaPrebuiltRelease -InstallDir $LlamaCppDir"
+            in content
+        )
+
+    def test_setup_ps1_source_build_uses_helper_latest_tag_only(self):
+        """PS1 source fallback should only use helper latest-tag resolution."""
+        content = SETUP_PS1.read_text()
+        assert "--resolve-source-build" not in content
+        assert "--resolve-install-tag" not in content
+        assert (
+            '"--resolve-llama-tag", "latest", "--published-repo", "ggml-org/llama.cpp"'
+            in content
+        )
         assert '--output-format", "json"' in content
         assert "$ResolvedSourceUrl" in content
         assert "$ResolvedSourceRefKind" in content
@@ -794,18 +861,25 @@ class TestSourceCodePatterns:
         """Helper resolution should suppress terminating NativeCommandError on PS 5.1."""
         content = SETUP_PS1.read_text()
         helper_idx = content.index("function Invoke-LlamaHelper")
-        block = content[helper_idx : helper_idx + 1200]
+        block = content[helper_idx : helper_idx + 2200]
         assert "$previousErrorActionPreference = $ErrorActionPreference" in block
         assert '$ErrorActionPreference = "Continue"' in block
         assert "$ErrorActionPreference = $previousErrorActionPreference" in block
 
     def test_setup_ps1_uses_local_tempfile_helper(self):
-        """PS1 should not depend on New-TemporaryFile being available."""
+        """PS1 should not depend on New-TemporaryFile being available anywhere."""
         content = SETUP_PS1.read_text()
         assert "function New-UnslothTemporaryFile" in content
-        assert "$resolveErrorLog = New-UnslothTemporaryFile" in content
         assert "$resolveErrorLog = New-TemporaryFile" not in content
 
+    def test_setup_ps1_find_nvcc_uses_version_sort_for_latest_toolkit(self):
+        """The unconstrained nvcc fallback should not sort toolkit dirs lexicographically."""
+        content = SETUP_PS1.read_text()
+        assert "Sort-Object Name | Select-Object -Last 1" not in content
+        assert (
+            "Sort-Object { [version]($_.Name -replace '^v','') } -Descending" in content
+        )
+
     def test_binary_env_linux_has_binary_parent(self):
         """The Linux branch of binary_env should include binary_path.parent."""
         content = MODULE_PATH.read_text()

From 5b7c0615f31bf9611ba9ce2b6a28c30ff6744d42 Mon Sep 17 00:00:00 2001
From: Wasim Yousef Said <wasimysdev@gmail.com>
Date: Fri, 3 Apr 2026 14:03:27 +0200
Subject: [PATCH 034/216] feat(studio): differentiate web search and URL fetch
 in chat tool UI (#4802)

Differentiate web_search query searches from URL fetches in the Studio chat UI.

Backend (llama_cpp.py):
- Emit "Reading: hostname" for URL fetches and "Searching: query" for query searches in SSE status events
- Only show hostname for valid http/https URLs; schemeless/non-http URLs get "Reading page..." generic fallback
- Strip www. prefix for consistency with the frontend

Frontend (tool-ui-web-search.tsx):
- Tool card shows "Read hostname" / "Reading hostname..." for URL fetches
- Shows "Searched query" / "Searching for query..." for query searches
- Uses new URL() with protocol check; falls back to "Read page" / "Reading page..." for non-http URLs
---
 studio/backend/core/inference/llama_cpp.py    | 16 +++++++++--
 .../assistant-ui/tool-ui-web-search.tsx       | 27 +++++++++++++++++--
 2 files changed, 39 insertions(+), 4 deletions(-)

diff --git a/studio/backend/core/inference/llama_cpp.py b/studio/backend/core/inference/llama_cpp.py
index d752500bf4..4da62e23c3 100644
--- a/studio/backend/core/inference/llama_cpp.py
+++ b/studio/backend/core/inference/llama_cpp.py
@@ -22,6 +22,7 @@ import threading
 import time
 from pathlib import Path
 from typing import Generator, Optional
+from urllib.parse import urlparse
 
 import httpx
 
@@ -2270,7 +2271,7 @@ class LlamaCppBackend:
         Agentic loop: let the model call tools, execute them, and continue.
 
         Yields dicts with:
-          {"type": "status", "text": "Searching: ..."}   -- tool status updates
+          {"type": "status", "text": "Searching: ..."/"Reading: ..."}   -- tool status updates
           {"type": "content", "text": "token"}            -- streamed content tokens (cumulative)
           {"type": "reasoning", "text": "token"}          -- streamed reasoning tokens (cumulative)
         """
@@ -2837,7 +2838,18 @@ class LlamaCppBackend:
                         arguments = raw_args
 
                     if tool_name == "web_search":
-                        status_text = f"Searching: {arguments.get('query', '')}"
+                        _ws_url = (arguments.get("url") or "").strip()
+                        if _ws_url:
+                            _parsed = urlparse(_ws_url)
+                            if _parsed.scheme in ("http", "https") and _parsed.hostname:
+                                _ws_host = _parsed.hostname
+                                if _ws_host.startswith("www."):
+                                    _ws_host = _ws_host[4:]
+                                status_text = f"Reading: {_ws_host}"
+                            else:
+                                status_text = "Reading page..."
+                        else:
+                            status_text = f"Searching: {arguments.get('query', '')}"
                     elif tool_name == "python":
                         preview = (
                             (arguments.get("code") or "").strip().split("\n")[0][:60]
diff --git a/studio/frontend/src/components/assistant-ui/tool-ui-web-search.tsx b/studio/frontend/src/components/assistant-ui/tool-ui-web-search.tsx
index d3a86846de..da4c0cbbb9 100644
--- a/studio/frontend/src/components/assistant-ui/tool-ui-web-search.tsx
+++ b/studio/frontend/src/components/assistant-ui/tool-ui-web-search.tsx
@@ -52,6 +52,18 @@ const WebSearchToolUIImpl: ToolCallMessagePartComponent = ({
   status,
 }) => {
   const query = (args as { query?: string })?.query ?? "";
+  const url = ((args as { url?: string })?.url ?? "").trim();
+  const isUrlFetch = !!url;
+  const displayDomain = (() => {
+    if (!url) return "";
+    try {
+      const parsed = new URL(url);
+      if (parsed.protocol !== "http:" && parsed.protocol !== "https:") return "";
+      return parsed.hostname.replace(/^www\./, "");
+    } catch {
+      return "";
+    }
+  })();
   const isRunning = status?.type === "running";
   const sources = result
     ? parseSearchResults(
@@ -75,7 +87,13 @@ const WebSearchToolUIImpl: ToolCallMessagePartComponent = ({
   return (
     <ToolFallbackRoot open={open} onOpenChange={setOpen}>
       <ToolFallbackTrigger
-        toolName={query ? `Searched "${query}"` : "Web Search"}
+        toolName={
+          isUrlFetch
+            ? displayDomain ? `Read ${displayDomain}` : "Read page"
+            : query
+              ? `Searched "${query}"`
+              : "Web Search"
+        }
         status={status}
         icon={GlobeIcon}
       />
@@ -83,7 +101,12 @@ const WebSearchToolUIImpl: ToolCallMessagePartComponent = ({
         {isRunning ? (
           <div className="flex items-center gap-2 text-sm text-muted-foreground">
             <LoaderIcon className="size-3.5 animate-spin" />
-            <span>Searching for &ldquo;{query}&rdquo;&hellip;</span>
+            <span>
+              {isUrlFetch
+                ? <>Reading {displayDomain || "page"}&hellip;</>
+                : <>Searching for &ldquo;{query}&rdquo;&hellip;</>
+              }
+            </span>
           </div>
         ) : sources.length > 0 ? (
           <div className="flex flex-wrap gap-1.5">

From 50dede11cc805db7584356183890d49d0b9f7c03 Mon Sep 17 00:00:00 2001
From: Wasim Yousef Said <wasimysdev@gmail.com>
Date: Fri, 3 Apr 2026 14:37:26 +0200
Subject: [PATCH 035/216] Allow non-LLM recipes to run and move Data tab first
 in executions (#4805)

* feat: allow non-LLM recipes to run without provider block

* feat: reorder execution tabs and add generation-aware data tab empty state

* fix: add accessibility attrs to data tab spinner and use literal ellipsis

* fix(studio): use shared spinner, stub provider, and hide unused LLM metrics

Backend: inject stub model provider for sampler-only recipes so
DataDesigner init does not reject empty provider lists.

Frontend: use shared Spinner component, hide LLM columns metric
and model usage card when recipe has no LLM columns.

* Fix tab reset and terminal auto-scroll regressions for PR #4805

Reset detailTab to "data" when switching between executions so
the Data tab default is applied consistently, not only on first
mount. Also add detailTab to the terminal scroll effect deps so
auto-scroll-to-bottom fires when the user opens the Overview tab
after landing on Data.

* Guard terminal scroll reset to only fire on Overview tab

The previous scroll effect ran on every tab switch, which could
reset the user's manual scroll position if they scrolled up in
the terminal and briefly switched tabs. Now the scroll-to-bottom
and sticky-bottom reset only fires when navigating to the
Overview tab.

* Use None for stub provider api_key instead of literal string

The stub ModelProvider that satisfies the DataDesigner registry
for non-LLM recipes should not carry a fake credential string.
Using None avoids sending an Authorization header if the provider
is ever inadvertently invoked.

---------

Co-authored-by: Daniel Han <danielhanchen@gmail.com>
---
 studio/backend/core/data_recipe/service.py    | 22 ++++--
 .../executions/execution-data-tab.tsx         | 17 +++-
 .../executions/execution-overview-tab.tsx     | 78 ++++++++++---------
 .../components/executions/executions-view.tsx | 12 ++-
 4 files changed, 81 insertions(+), 48 deletions(-)

diff --git a/studio/backend/core/data_recipe/service.py b/studio/backend/core/data_recipe/service.py
index 550358ae61..7fff36aefd 100644
--- a/studio/backend/core/data_recipe/service.py
+++ b/studio/backend/core/data_recipe/service.py
@@ -167,12 +167,7 @@ def _validate_recipe_runtime_support(
     recipe: dict[str, Any],
     model_providers: list[Any],
 ) -> None:
-    if not _recipe_has_llm_columns(recipe):
-        raise ValueError(
-            "Recipe Studio currently requires at least one AI generation step."
-        )
-
-    if not model_providers:
+    if _recipe_has_llm_columns(recipe) and not model_providers:
         raise ValueError("Add a Provider connection block before running this recipe.")
 
 
@@ -266,6 +261,21 @@ def create_data_designer(
     model_providers = build_model_providers(recipe)
     _validate_recipe_runtime_support(recipe, model_providers)
 
+    # DataDesigner requires at least one model provider in its registry even
+    # when the pipeline contains no LLM columns.  Supply a lightweight stub
+    # so sampler/expression-only recipes can run without a real provider.
+    if not model_providers:
+        from data_designer.config.models import ModelProvider
+
+        model_providers = [
+            ModelProvider(
+                name = "_unused",
+                endpoint = "http://localhost",
+                provider_type = "openai",
+                api_key = None,
+            )
+        ]
+
     return DataDesigner(
         artifact_path = artifact_path,
         model_providers = model_providers,
diff --git a/studio/frontend/src/features/recipe-studio/components/executions/execution-data-tab.tsx b/studio/frontend/src/features/recipe-studio/components/executions/execution-data-tab.tsx
index dc6baddf02..a8e12ac1ca 100644
--- a/studio/frontend/src/features/recipe-studio/components/executions/execution-data-tab.tsx
+++ b/studio/frontend/src/features/recipe-studio/components/executions/execution-data-tab.tsx
@@ -12,6 +12,7 @@ import {
   DropdownMenuLabel,
   DropdownMenuTrigger,
 } from "@/components/ui/dropdown-menu";
+import { Spinner } from "@/components/ui/spinner";
 import { cn } from "@/lib/utils";
 import { isExecutionInProgress } from "../../executions/execution-helpers";
 import type { RecipeExecutionRecord } from "../../execution-types";
@@ -121,7 +122,21 @@ export function ExecutionDataTab({
         </div>
       </div>
       {execution.dataset.length === 0 ? (
-        <p className="text-xs text-muted-foreground">No rows returned.</p>
+        isExecutionInProgress(execution.status) ? (
+          <div className="flex flex-col items-center justify-center gap-3 py-12 text-center">
+            <Spinner className="size-5" />
+            <div className="space-y-1">
+              <p className="text-sm font-medium text-muted-foreground">
+                Generating data…
+              </p>
+              <p className="text-xs text-muted-foreground">
+                Check the Overview tab for live terminal logs.
+              </p>
+            </div>
+          </div>
+        ) : (
+          <p className="text-xs text-muted-foreground">No rows returned.</p>
+        )
       ) : tableColumns.length === 0 ? (
         <p className="text-xs text-muted-foreground">
           All columns hidden. Use Columns to show at least one.
diff --git a/studio/frontend/src/features/recipe-studio/components/executions/execution-overview-tab.tsx b/studio/frontend/src/features/recipe-studio/components/executions/execution-overview-tab.tsx
index 06b4b20629..8e2e710299 100644
--- a/studio/frontend/src/features/recipe-studio/components/executions/execution-overview-tab.tsx
+++ b/studio/frontend/src/features/recipe-studio/components/executions/execution-overview-tab.tsx
@@ -116,10 +116,12 @@ export function ExecutionOverviewTab({
                 />
               </div>
               <div className="space-y-1.5 text-xs">
-                <p className="flex items-center justify-between gap-3">
-                  <span className="text-muted-foreground">LLM columns</span>
-                  <span className="font-semibold">{formatMetricValue(llmColumnCount)}</span>
-                </p>
+                {llmColumnCount > 0 && (
+                  <p className="flex items-center justify-between gap-3">
+                    <span className="text-muted-foreground">LLM columns</span>
+                    <span className="font-semibold">{formatMetricValue(llmColumnCount)}</span>
+                  </p>
+                )}
                 <p className="flex items-center justify-between gap-3">
                   <span className="text-muted-foreground">Null rate</span>
                   <span className="font-semibold">{nullRate?.toFixed(1) ?? "--"}%</span>
@@ -164,40 +166,42 @@ export function ExecutionOverviewTab({
               </div>
             </div>
           </div>
-          <div className="rounded-xl border border-border/60 bg-card/55 p-3">
-            <div className="mb-2 flex items-center justify-between">
-              <p className="text-xs text-muted-foreground">Model usage</p>
-              <HugeiconsIcon icon={Flag02Icon} className="size-4 text-muted-foreground" />
-            </div>
-            {modelUsageRows.length === 0 ? (
-              <p className="text-xs text-muted-foreground">No model usage yet.</p>
-            ) : (
-              <div className="overflow-hidden rounded-lg border border-border/60 bg-card/50">
-                <Table>
-                  <TableHeader>
-                    <TableRow>
-                      <TableHead>Model</TableHead>
-                      <TableHead className="text-right">Input</TableHead>
-                      <TableHead className="text-right">Output</TableHead>
-                    </TableRow>
-                  </TableHeader>
-                  <TableBody>
-                    {modelUsageRows.map((usage) => (
-                      <TableRow key={usage.model}>
-                        <TableCell className="max-w-[320px] truncate">{usage.model}</TableCell>
-                        <TableCell className="text-right">
-                          {formatMetricValue(usage.input)}
-                        </TableCell>
-                        <TableCell className="text-right">
-                          {formatMetricValue(usage.output)}
-                        </TableCell>
-                      </TableRow>
-                    ))}
-                  </TableBody>
-                </Table>
+          {(llmColumnCount > 0 || modelUsageRows.length > 0) && (
+            <div className="rounded-xl border border-border/60 bg-card/55 p-3">
+              <div className="mb-2 flex items-center justify-between">
+                <p className="text-xs text-muted-foreground">Model usage</p>
+                <HugeiconsIcon icon={Flag02Icon} className="size-4 text-muted-foreground" />
               </div>
-            )}
-          </div>
+              {modelUsageRows.length === 0 ? (
+                <p className="text-xs text-muted-foreground">No model usage yet.</p>
+              ) : (
+                <div className="overflow-hidden rounded-lg border border-border/60 bg-card/50">
+                  <Table>
+                    <TableHeader>
+                      <TableRow>
+                        <TableHead>Model</TableHead>
+                        <TableHead className="text-right">Input</TableHead>
+                        <TableHead className="text-right">Output</TableHead>
+                      </TableRow>
+                    </TableHeader>
+                    <TableBody>
+                      {modelUsageRows.map((usage) => (
+                        <TableRow key={usage.model}>
+                          <TableCell className="max-w-[320px] truncate">{usage.model}</TableCell>
+                          <TableCell className="text-right">
+                            {formatMetricValue(usage.input)}
+                          </TableCell>
+                          <TableCell className="text-right">
+                            {formatMetricValue(usage.output)}
+                          </TableCell>
+                        </TableRow>
+                      ))}
+                    </TableBody>
+                  </Table>
+                </div>
+              )}
+            </div>
+          )}
         </div>
       )}
       <div className="overflow-hidden rounded-xl corner-squircle border">
diff --git a/studio/frontend/src/features/recipe-studio/components/executions/executions-view.tsx b/studio/frontend/src/features/recipe-studio/components/executions/executions-view.tsx
index f30cf61012..f442d8f98e 100644
--- a/studio/frontend/src/features/recipe-studio/components/executions/executions-view.tsx
+++ b/studio/frontend/src/features/recipe-studio/components/executions/executions-view.tsx
@@ -59,7 +59,7 @@ export function ExecutionsView({
     typeof value === "number" && Number.isFinite(value)
       ? `${value.toLocaleString()} s`
       : "--";
-  const [detailTab, setDetailTab] = useState("overview");
+  const [detailTab, setDetailTab] = useState("data");
   const [hiddenDatasetColumnsByExecution, setHiddenDatasetColumnsByExecution] = useState<
     Record<string, string[]>
   >({});
@@ -341,12 +341,16 @@ export function ExecutionsView({
   }, [selectedExecution]);
 
   useEffect(() => {
-    if (!terminalRef.current) {
+    setDetailTab("data");
+  }, [selectedExecution?.id]);
+
+  useEffect(() => {
+    if (detailTab !== "overview" || !terminalRef.current) {
       return;
     }
     shouldStickTerminalToBottomRef.current = true;
     terminalRef.current.scrollTop = terminalRef.current.scrollHeight;
-  }, [selectedExecution?.id]);
+  }, [detailTab, selectedExecution?.id]);
 
   useEffect(() => {
     if (!terminalRef.current) {
@@ -440,9 +444,9 @@ export function ExecutionsView({
             <Tabs value={detailTab} onValueChange={setDetailTab}>
               <div className="flex items-center justify-between gap-2">
                 <TabsList className="border border-border/60 bg-card/40">
+                  <TabsTrigger value="data">Data</TabsTrigger>
                   <TabsTrigger value="overview">Overview</TabsTrigger>
                   <TabsTrigger value="columns">Columns</TabsTrigger>
-                  <TabsTrigger value="data">Data</TabsTrigger>
                   <TabsTrigger value="raw">Raw</TabsTrigger>
                 </TabsList>
                 <div className="flex items-center gap-2">

From a29b4e23fd92c804ce2c46ddc4acf73d756ec823 Mon Sep 17 00:00:00 2001
From: Lee Jackson <130007945+Imagineer99@users.noreply.github.com>
Date: Fri, 3 Apr 2026 13:48:24 +0100
Subject: [PATCH 036/216] studio: reuse HF cached repo casing to prevent
 duplicate downloads (#4822)

* fix(studio): reuse HF cached repo casing to prevent duplicate downloads

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

* Move cache case resolution tests to separate PR

Tests for resolve_cached_repo_id_case and get_model_config case resolution
belong in their own PR to keep this change focused on the runtime fix.

* fix(studio): debug-log HF_HUB_CACHE fallback in path_utils

* Fix stale memoization in resolve_cached_repo_id_case

- Check exact-case path before memo to ensure a newly-appeared exact
  match always wins over a previously memoized variant
- Validate memoized entries still exist on disk before returning them
  to prevent stale results when cache dirs are deleted/recreated

* Minor cleanups for cache case resolution

- Use .is_dir() instead of .exists() for exact-case cache check
  (cache entries are always directories)
- Remove redundant fallback in _detect_audio_from_tokenizer since
  get_cache_path already handles case resolution and returns None
  when the model is not cached

---------

Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com>
Co-authored-by: Daniel Han <danielhanchen@gmail.com>
---
 studio/backend/core/inference/worker.py     |   3 +
 studio/backend/routes/models.py             |  15 ++-
 studio/backend/utils/models/model_config.py |  31 +++--
 studio/backend/utils/paths/__init__.py      |  13 ++-
 studio/backend/utils/paths/path_utils.py    | 118 +++++++++++++++++++-
 unsloth/models/loader.py                    |  23 ++--
 unsloth/models/loader_utils.py              |   2 +-
 7 files changed, 180 insertions(+), 25 deletions(-)

diff --git a/studio/backend/core/inference/worker.py b/studio/backend/core/inference/worker.py
index e2513f43de..7f7291a56d 100644
--- a/studio/backend/core/inference/worker.py
+++ b/studio/backend/core/inference/worker.py
@@ -145,6 +145,8 @@ def _get_hf_download_state(
         blobs_dirs: list[Path] = []
 
         if model_names:
+            from utils.paths import resolve_cached_repo_id_case
+
             for name in model_names:
                 if not name:
                     continue
@@ -154,6 +156,7 @@ def _get_hf_download_state(
                 # relative paths, and Windows paths.
                 if name.startswith(("/", ".", "~")) or "\\" in name:
                     continue
+                name = resolve_cached_repo_id_case(name)
                 # HF cache dir format: models--org--name (slashes -> --)
                 cache_dir_name = "models--" + name.replace("/", "--")
                 blobs_dir = cache / cache_dir_name / "blobs"
diff --git a/studio/backend/routes/models.py b/studio/backend/routes/models.py
index 445cf0e7f4..1e31a91e26 100644
--- a/studio/backend/routes/models.py
+++ b/studio/backend/routes/models.py
@@ -49,8 +49,10 @@ try:
     )
     from core.inference import get_inference_backend
     from utils.paths import (
+        is_local_path,
         outputs_root,
         exports_root,
+        resolve_cached_repo_id_case,
         resolve_output_dir,
         resolve_export_dir,
     )
@@ -77,8 +79,10 @@ except ImportError:
     )
     from core.inference import get_inference_backend
     from utils.paths import (
+        is_local_path,
         outputs_root,
         exports_root,
+        resolve_cached_repo_id_case,
         resolve_output_dir,
         resolve_export_dir,
     )
@@ -597,10 +601,15 @@ async def get_model_config(
     This endpoint wraps the backend load_model_defaults function.
     """
     try:
-        from utils.models.model_config import is_local_path
-
         if not is_local_path(model_name):
-            model_name = model_name.lower()
+            resolved = resolve_cached_repo_id_case(model_name)
+            if resolved != model_name:
+                logger.info(
+                    "Using cached repo_id casing '%s' for requested '%s'",
+                    resolved,
+                    model_name,
+                )
+            model_name = resolved
 
         logger.info(f"Getting model config for: {model_name}")
         from utils.models.model_config import detect_audio_type
diff --git a/studio/backend/utils/models/model_config.py b/studio/backend/utils/models/model_config.py
index 6cffc534aa..8c25da4f43 100644
--- a/studio/backend/utils/models/model_config.py
+++ b/studio/backend/utils/models/model_config.py
@@ -11,6 +11,8 @@ from utils.paths import (
     normalize_path,
     is_local_path,
     is_model_cached,
+    get_cache_path,
+    resolve_cached_repo_id_case,
     outputs_root,
     exports_root,
     resolve_output_dir,
@@ -711,12 +713,8 @@ def _detect_audio_from_tokenizer(
 
     # 1) Check local HF cache first (works for gated/offline models)
     try:
-        from huggingface_hub.constants import HF_HUB_CACHE
-
-        cache_dir = Path(HF_HUB_CACHE)
-        repo_dir_name = f"models--{model_name.replace('/', '--')}"
-        repo_dir = cache_dir / repo_dir_name
-        if repo_dir.exists():
+        repo_dir = get_cache_path(model_name)
+        if repo_dir is not None and repo_dir.exists():
             snapshots_dir = repo_dir / "snapshots"
             if snapshots_dir.exists():
                 for snapshot in snapshots_dir.iterdir():
@@ -1627,11 +1625,18 @@ class ModelConfig:
             identifier = f"unsloth/{identifier}"
             path = identifier
 
-        # Enforce lowercase for remote Hugging Face identifiers to prevent cache duplication
-        # Hugging Face Hub APIs are case-insensitive remotely, but case-sensitive locally (repo_folder_name).
+        # Preserve requested casing, but if a case-variant already exists in local HF cache,
+        # reuse that exact repo_id spelling to avoid one-time re-downloads after #2592.
         if not is_local:
-            identifier = identifier.lower()
-            path = path.lower()
+            resolved_identifier = resolve_cached_repo_id_case(identifier)
+            if resolved_identifier != identifier:
+                logger.info(
+                    "Using cached repo_id casing '%s' for requested '%s'",
+                    resolved_identifier,
+                    identifier,
+                )
+                identifier = resolved_identifier
+                path = resolved_identifier
 
         # Auto-detect GGUF models (check before LoRA/vision detection)
         if is_local:
@@ -1852,6 +1857,12 @@ class ModelConfig:
             identifier = f"unsloth/{identifier}"
             path = identifier
 
+        if not is_local:
+            resolved_identifier = resolve_cached_repo_id_case(identifier)
+            if resolved_identifier != identifier:
+                identifier = resolved_identifier
+                path = resolved_identifier
+
         # --- Logic for Base Model and Vision Detection ---
         base_model = None
         is_vision = False
diff --git a/studio/backend/utils/paths/__init__.py b/studio/backend/utils/paths/__init__.py
index 44a7c8e287..11709ae56e 100644
--- a/studio/backend/utils/paths/__init__.py
+++ b/studio/backend/utils/paths/__init__.py
@@ -5,7 +5,15 @@
 Path utilities for model and dataset handling
 """
 
-from .path_utils import normalize_path, is_local_path, is_model_cached, get_cache_path
+from .path_utils import (
+    normalize_path,
+    is_local_path,
+    is_model_cached,
+    get_cache_path,
+    resolve_cached_repo_id_case,
+    get_cache_case_resolution_stats,
+    reset_cache_case_resolution_state,
+)
 from .storage_roots import (
     studio_root,
     assets_root,
@@ -40,6 +48,9 @@ __all__ = [
     "is_local_path",
     "is_model_cached",
     "get_cache_path",
+    "resolve_cached_repo_id_case",
+    "get_cache_case_resolution_stats",
+    "reset_cache_case_resolution_state",
     "studio_root",
     "assets_root",
     "datasets_root",
diff --git a/studio/backend/utils/paths/path_utils.py b/studio/backend/utils/paths/path_utils.py
index b38db18286..9ef9a2dd92 100644
--- a/studio/backend/utils/paths/path_utils.py
+++ b/studio/backend/utils/paths/path_utils.py
@@ -14,6 +14,20 @@ from loggers import get_logger
 
 logger = get_logger(__name__)
 
+# Per-process cache to avoid repeated cache-dir scans for the same identifier.
+_CACHE_CASE_RESOLUTION_MEMO: dict[str, str] = {}
+
+# Lightweight instrumentation counters for operational visibility.
+_CACHE_CASE_RESOLUTION_STATS: dict[str, int] = {
+    "calls": 0,
+    "memo_hits": 0,
+    "exact_hits": 0,
+    "variant_hits": 0,
+    "tie_breaks": 0,
+    "fallbacks": 0,
+    "errors": 0,
+}
+
 
 def _is_wsl() -> bool:
     """Detect if we are running inside WSL (Windows Subsystem for Linux)."""
@@ -94,8 +108,9 @@ def is_local_path(path: str) -> bool:
 
 def get_cache_path(model_name: str) -> Optional[Path]:
     """Get HuggingFace cache path for a model if it exists."""
-    cache_dir = Path.home() / ".cache" / "huggingface" / "hub"
-    model_cache_name = model_name.replace("/", "--")
+    cache_dir = _hf_hub_cache_dir()
+    resolved_name = resolve_cached_repo_id_case(model_name)
+    model_cache_name = resolved_name.replace("/", "--")
     model_cache_path = cache_dir / f"models--{model_cache_name}"
 
     return model_cache_path if model_cache_path.exists() else None
@@ -113,3 +128,102 @@ def is_model_cached(model_name: str) -> bool:
             return True
 
     return False
+
+
+def _hf_hub_cache_dir() -> Path:
+    """Return HF cache root honoring HF_HUB_CACHE when available."""
+    try:
+        from huggingface_hub.constants import HF_HUB_CACHE
+
+        return Path(HF_HUB_CACHE)
+    except Exception as exc:
+        logger.debug(
+            "Could not read huggingface_hub HF_HUB_CACHE, using default hub path: %s",
+            exc,
+        )
+        return Path.home() / ".cache" / "huggingface" / "hub"
+
+
+def resolve_cached_repo_id_case(model_name: str, use_memo: bool = True) -> str:
+    """Resolve repo_id to the exact casing already present in local HF cache.
+
+    Policy: prefer the requested/canonical repo_id, but if a case-variant already
+    exists in local HF cache, reuse that exact cached spelling. This avoids
+    duplicate downloads while preserving user intent whenever possible.
+    """
+    _CACHE_CASE_RESOLUTION_STATS["calls"] += 1
+
+    if not model_name or "/" not in model_name:
+        _CACHE_CASE_RESOLUTION_STATS["fallbacks"] += 1
+        return model_name
+
+    cache_dir = _hf_hub_cache_dir()
+    if not cache_dir.exists():
+        _CACHE_CASE_RESOLUTION_STATS["fallbacks"] += 1
+        return model_name
+
+    expected_dir = f"models--{model_name.replace('/', '--')}"
+
+    # Always check the exact-case path first so a newly-appeared exact match
+    # wins over any previously memoized variant.
+    exact_path = cache_dir / expected_dir
+    if exact_path.is_dir():
+        if use_memo:
+            _CACHE_CASE_RESOLUTION_MEMO[model_name] = model_name
+        _CACHE_CASE_RESOLUTION_STATS["exact_hits"] += 1
+        return model_name
+
+    # Validate memoized entries still exist on disk before returning them.
+    # This prevents stale results when cache dirs are deleted/recreated.
+    if use_memo:
+        cached = _CACHE_CASE_RESOLUTION_MEMO.get(model_name)
+        if cached is not None:
+            cached_path = cache_dir / f"models--{cached.replace('/', '--')}"
+            if cached_path.is_dir():
+                _CACHE_CASE_RESOLUTION_STATS["memo_hits"] += 1
+                return cached
+            # Stale entry -- drop it and re-scan below.
+            _CACHE_CASE_RESOLUTION_MEMO.pop(model_name, None)
+
+    expected_lower = expected_dir.lower()
+    try:
+        candidates: list[str] = []
+        for entry in cache_dir.iterdir():
+            if not entry.is_dir():
+                continue
+            if entry.name.lower() != expected_lower:
+                continue
+            if not entry.name.startswith("models--"):
+                continue
+            repo_part = entry.name[len("models--") :]
+            if not repo_part:
+                continue
+            candidates.append(repo_part.replace("--", "/"))
+
+        if candidates:
+            # Deterministic tie-break if multiple case variants coexist.
+            resolved = sorted(candidates)[0]
+            if len(candidates) > 1:
+                _CACHE_CASE_RESOLUTION_STATS["tie_breaks"] += 1
+            _CACHE_CASE_RESOLUTION_STATS["variant_hits"] += 1
+            if use_memo:
+                _CACHE_CASE_RESOLUTION_MEMO[model_name] = resolved
+            return resolved
+    except Exception as exc:
+        _CACHE_CASE_RESOLUTION_STATS["errors"] += 1
+        logger.debug(f"Could not resolve cached repo_id case for '{model_name}': {exc}")
+
+    _CACHE_CASE_RESOLUTION_STATS["fallbacks"] += 1
+    return model_name
+
+
+def get_cache_case_resolution_stats() -> dict[str, int]:
+    """Return a copy of case-resolution instrumentation counters."""
+    return dict(_CACHE_CASE_RESOLUTION_STATS)
+
+
+def reset_cache_case_resolution_state() -> None:
+    """Clear resolver memo and counters (primarily for tests)."""
+    _CACHE_CASE_RESOLUTION_MEMO.clear()
+    for key in _CACHE_CASE_RESOLUTION_STATS:
+        _CACHE_CASE_RESOLUTION_STATS[key] = 0
diff --git a/unsloth/models/loader.py b/unsloth/models/loader.py
index cfe3acb656..a6cb3eb529 100644
--- a/unsloth/models/loader.py
+++ b/unsloth/models/loader.py
@@ -114,6 +114,17 @@ FORCE_FLOAT32 = [
 
 global DISABLE_COMPILE_MODEL_NAMES
 # Must be alphabetically sorted for each entry
+
+
+def _strip_unsloth_bnb_4bit_suffix(model_name: str) -> str:
+    """Remove Unsloth 4bit suffixes without lowercasing (HF cache dirs are case-sensitive)."""
+    s = model_name
+    for suffix in ("-unsloth-bnb-4bit", "-bnb-4bit"):
+        if len(s) >= len(suffix) and s.lower().endswith(suffix.lower()):
+            s = s[: -len(suffix)]
+    return s
+
+
 DISABLE_COMPILE_MODEL_NAMES = [
     "aya_vision",
     "modernbert",
@@ -404,8 +415,7 @@ class FastLanguageModel(FastLlamaModel):
         if not ALLOW_PREQUANTIZED_MODELS and model_name.lower().endswith(
             ("-unsloth-bnb-4bit", "-bnb-4bit")
         ):
-            model_name = model_name.lower().removesuffix("-unsloth-bnb-4bit")
-            model_name = model_name.lower().removesuffix("-bnb-4bit")
+            model_name = _strip_unsloth_bnb_4bit_suffix(model_name)
         # Change -BF16 to all False for 4bit, 8bit etc
         if model_name.lower().endswith("-bf16"):
             load_in_4bit = False
@@ -551,8 +561,7 @@ class FastLanguageModel(FastLlamaModel):
             if not ALLOW_PREQUANTIZED_MODELS and model_name.lower().endswith(
                 ("-unsloth-bnb-4bit", "-bnb-4bit")
             ):
-                model_name = model_name.lower().removesuffix("-unsloth-bnb-4bit")
-                model_name = model_name.lower().removesuffix("-bnb-4bit")
+                model_name = _strip_unsloth_bnb_4bit_suffix(model_name)
             # Change -BF16 to all False for 4bit, 8bit etc
             if model_name.lower().endswith("-bf16"):
                 load_in_4bit = False
@@ -1019,8 +1028,7 @@ class FastModel(FastBaseModel):
         if not ALLOW_PREQUANTIZED_MODELS and model_name.lower().endswith(
             ("-unsloth-bnb-4bit", "-bnb-4bit")
         ):
-            model_name = model_name.lower().removesuffix("-unsloth-bnb-4bit")
-            model_name = model_name.lower().removesuffix("-bnb-4bit")
+            model_name = _strip_unsloth_bnb_4bit_suffix(model_name)
         # Change -BF16 to all False for 4bit, 8bit etc
         if model_name.lower().endswith("-bf16"):
             load_in_4bit = False
@@ -1320,8 +1328,7 @@ class FastModel(FastBaseModel):
             if not ALLOW_PREQUANTIZED_MODELS and model_name.lower().endswith(
                 ("-unsloth-bnb-4bit", "-bnb-4bit")
             ):
-                model_name = model_name.lower().removesuffix("-unsloth-bnb-4bit")
-                model_name = model_name.lower().removesuffix("-bnb-4bit")
+                model_name = _strip_unsloth_bnb_4bit_suffix(model_name)
             # Change -BF16 to all False for 4bit, 8bit etc
             if model_name.lower().endswith("-bf16"):
                 load_in_4bit = False
diff --git a/unsloth/models/loader_utils.py b/unsloth/models/loader_utils.py
index cf5af983a6..99da5f799e 100644
--- a/unsloth/models/loader_utils.py
+++ b/unsloth/models/loader_utils.py
@@ -162,7 +162,7 @@ def __get_model_name(
         # Support returning original full -bnb-4bit name if specified specifically
         # since we'll map it to the dynamic version instead
         if lower_model_name.endswith("-bnb-4bit"):
-            return lower_model_name
+            return model_name
 
         new_model_name = FLOAT_TO_INT_MAPPER[lower_model_name]
         # logger.warning_once(

From c027ec192ef5e4e69f4c02973081f5b3da67aa20 Mon Sep 17 00:00:00 2001
From: Neodon <82944+neodon@users.noreply.github.com>
Date: Fri, 3 Apr 2026 13:44:22 -0500
Subject: [PATCH 037/216] fix(studio): ensure first chat tool call starts in
 session sandbox (#4810)

Fixes #4809

On a new Studio chat, the first tool call could start before the frontend
initializes the thread ID. That meant the first request could go out without
a session_id, so the backend started the tool in the shared sandbox root
instead of the chat's session sandbox.

Frontend:
- Eagerly initialize the thread when switching to a new chat
- Resolve the thread ID once at request time and keep it stable through
  async model-load waits
- Disable ActiveThreadSync during new-chat initialization to prevent
  stale thread IDs from being written back
- Add error handling for thread initialization failures
- Clear activeThreadId on all compare-mode entry paths to prevent
  cross-session leakage
- Fix exitCompare to restore context usage from the saved view
- Coerce falsy thread IDs to undefined for consistent backend/frontend
  fallback behavior
- Use _default as the image sessionId fallback to match the backend

Backend:
- Use ~/studio_sandbox/_default when a request arrives without a session_id
---
 studio/backend/core/inference/tools.py        |  5 ++--
 .../src/features/chat/api/chat-adapter.ts     | 16 ++++++----
 .../frontend/src/features/chat/chat-page.tsx  | 15 ++++++++--
 .../src/features/chat/runtime-provider.tsx    | 30 +++++++++++++++++--
 4 files changed, 54 insertions(+), 12 deletions(-)

diff --git a/studio/backend/core/inference/tools.py b/studio/backend/core/inference/tools.py
index b23372b766..86c22ce25a 100644
--- a/studio/backend/core/inference/tools.py
+++ b/studio/backend/core/inference/tools.py
@@ -34,7 +34,8 @@ _MAX_OUTPUT_CHARS = 8000  # truncate long output
 _BASH_BLOCKED_WORDS = {"rm", "sudo", "dd", "chmod", "mkfs", "shutdown", "reboot"}
 
 # Per-session working directories so each chat thread gets its own sandbox.
-# Falls back to a shared ~/studio_sandbox/ for API callers without a session_id.
+# Falls back to a shared ~/studio_sandbox/_default for API callers without a
+# session_id.
 _workdirs: dict[str, str] = {}
 
 
@@ -55,7 +56,7 @@ def _get_workdir(session_id: str | None = None) -> str:
             if not os.path.realpath(workdir).startswith(os.path.realpath(sandbox_root)):
                 workdir = os.path.join(sandbox_root, "_invalid")
         else:
-            workdir = sandbox_root
+            workdir = os.path.join(sandbox_root, "_default")
         os.makedirs(workdir, exist_ok = True)
         _workdirs[key] = workdir
     return _workdirs[key]
diff --git a/studio/frontend/src/features/chat/api/chat-adapter.ts b/studio/frontend/src/features/chat/api/chat-adapter.ts
index e287daf33a..3d1bce2905 100644
--- a/studio/frontend/src/features/chat/api/chat-adapter.ts
+++ b/studio/frontend/src/features/chat/api/chat-adapter.ts
@@ -421,6 +421,10 @@ export function createOpenAIStreamAdapter(): ChatModelAdapter {
   return {
     async *run({ messages, abortSignal, unstable_threadId }) {
       let runtime = useChatRuntimeStore.getState();
+      // Capture the thread ID once at the start so it stays stable even if
+      // the user switches chats while waiting for model load / auto-load.
+      const resolvedThreadId =
+        (unstable_threadId ?? runtime.activeThreadId) || undefined;
 
       // Wait for in-progress model load to finish before inferring
       if (runtime.modelLoading) {
@@ -473,14 +477,14 @@ export function createOpenAIStreamAdapter(): ChatModelAdapter {
         }
         runtime.clearPendingAudio();
       }
-      const useAdapter = await resolveUseAdapter(unstable_threadId);
+      const useAdapter = await resolveUseAdapter(resolvedThreadId);
 
       // ── Audio model path (non-streaming) ─────────────────────
       const activeModel = runtime.models.find(
         (m) => m.id === params.checkpoint,
       );
       if (activeModel?.isAudio && !activeModel?.hasAudioInput) {
-        const threadKey = unstable_threadId || "__default";
+        const threadKey = resolvedThreadId || "__default";
         runtime.setThreadRunning(threadKey, true);
         try {
           yield {
@@ -527,7 +531,7 @@ export function createOpenAIStreamAdapter(): ChatModelAdapter {
         return;
       }
 
-      const threadKey = unstable_threadId || "__default";
+      const threadKey = resolvedThreadId || "__default";
       let waitingFirstChunk = true;
       let firstTokenSettled = false;
       const streamStartTime = Date.now();
@@ -600,7 +604,7 @@ export function createOpenAIStreamAdapter(): ChatModelAdapter {
                     const mins = useChatRuntimeStore.getState().toolCallTimeout;
                     return mins >= 9999 ? 9999 : mins * 60;
                   })(),
-                  session_id: unstable_threadId || undefined,
+                  session_id: resolvedThreadId,
                 }
               : {}),
           },
@@ -641,7 +645,9 @@ export function createOpenAIStreamAdapter(): ChatModelAdapter {
                 let parsedResult: string | { text: string; images: string[]; sessionId: string };
                 if (imgIdx !== -1) {
                   const text = rawResult.slice(0, imgIdx);
-                  const sessionId = unstable_threadId || "";
+                  // Fall back to "_default" to match the backend sandbox directory
+                  // used when no session_id is provided (see tools.py _get_workdir).
+                  const sessionId = resolvedThreadId || "_default";
                   try {
                     const images = JSON.parse(rawResult.slice(imgIdx + imgMarker.length)) as string[];
                     parsedResult = { text, images, sessionId };
diff --git a/studio/frontend/src/features/chat/chat-page.tsx b/studio/frontend/src/features/chat/chat-page.tsx
index 08450c7ec7..1dbff145ee 100644
--- a/studio/frontend/src/features/chat/chat-page.tsx
+++ b/studio/frontend/src/features/chat/chat-page.tsx
@@ -592,6 +592,9 @@ export function ChatPage(): ReactElement {
   }, []);
   const handleNewCompare = useCallback(() => {
     setView({ mode: "compare", pairId: crypto.randomUUID() });
+    // Clear activeThreadId so compare panes do not inherit the single-chat
+    // thread ID as a fallback for session_id routing.
+    useChatRuntimeStore.getState().setActiveThreadId(null);
     useChatRuntimeStore.getState().setContextUsage(null);
   }, []);
 
@@ -619,6 +622,9 @@ export function ChatPage(): ReactElement {
   const enterCompare = useCallback(() => {
     setViewBeforeCompare((prev) => prev ?? view);
     setView({ mode: "compare", pairId: crypto.randomUUID() });
+    // Clear activeThreadId so compare panes do not inherit the single-chat
+    // thread ID as a fallback for session_id routing.
+    useChatRuntimeStore.getState().setActiveThreadId(null);
     useChatRuntimeStore.getState().setContextUsage(null);
   }, [view]);
 
@@ -626,9 +632,13 @@ export function ChatPage(): ReactElement {
     if (!viewBeforeCompare) return;
     setView(viewBeforeCompare);
     setViewBeforeCompare(null);
-    // Restore context usage from the active thread's last assistant message
+    // Restore context usage from the active thread's last assistant message.
+    // Use the thread ID from the saved view rather than the store, because
+    // activeThreadId may have been cleared on compare entry.
     const store = useChatRuntimeStore.getState();
-    const threadId = store.activeThreadId;
+    const threadId =
+      ("threadId" in viewBeforeCompare ? viewBeforeCompare.threadId : null) ??
+      store.activeThreadId;
     if (threadId) {
       void db.messages
         .where("threadId")
@@ -735,6 +745,7 @@ export function ChatPage(): ReactElement {
           await selectModelRef.current({ id: targetLora.id, isLora: true });
           if (canceled) return;
           setView({ mode: "compare", pairId: crypto.randomUUID() });
+          useChatRuntimeStore.getState().setActiveThreadId(null);
           useChatRuntimeStore.getState().setContextUsage(null);
           clearHandoff();
           console.info("[chat-handoff] loaded lora + opened compare");
diff --git a/studio/frontend/src/features/chat/runtime-provider.tsx b/studio/frontend/src/features/chat/runtime-provider.tsx
index 02f792b509..7e6cd8e1dd 100644
--- a/studio/frontend/src/features/chat/runtime-provider.tsx
+++ b/studio/frontend/src/features/chat/runtime-provider.tsx
@@ -679,9 +679,33 @@ function ThreadNewChatSwitch({
   const isLoading = useAuiState(({ threads }) => threads.isLoading);
 
   useEffect(() => {
-    if (!isLoading) {
-      aui.threads().switchToNewThread();
+    if (isLoading) {
+      return;
     }
+
+    let cancelled = false;
+    // Clear immediately so the adapter never picks up a stale thread ID
+    // from a previous chat while we initialize the new one.
+    useChatRuntimeStore.getState().setActiveThreadId(null);
+
+    void (async () => {
+      try {
+        aui.threads().switchToNewThread();
+        const { remoteId } = await aui.threadListItem().initialize();
+        if (!cancelled) {
+          useChatRuntimeStore.getState().setActiveThreadId(remoteId);
+        }
+      } catch (error) {
+        if (!cancelled) {
+          useChatRuntimeStore.getState().setActiveThreadId(null);
+        }
+        console.error("Failed to initialize new chat thread", error);
+      }
+    })();
+
+    return () => {
+      cancelled = true;
+    };
   }, [aui, isLoading, nonce]);
 
   return null;
@@ -730,7 +754,7 @@ export function ChatRuntimeProvider({
 
   return (
     <AssistantRuntimeProvider runtime={runtime} aui={aui}>
-      <ActiveThreadSync enabled={modelType === "base" && !pairId} />
+      <ActiveThreadSync enabled={modelType === "base" && !pairId && !newThreadNonce} />
       {initialThreadId && <ThreadAutoSwitch threadId={initialThreadId} />}
       {!initialThreadId && newThreadNonce && (
         <ThreadNewChatSwitch nonce={newThreadNonce} />

From 2c73ab7871849d50f85399d0cf8a453ce55c1baa Mon Sep 17 00:00:00 2001
From: Daniel Han <danielhanchen@gmail.com>
Date: Fri, 3 Apr 2026 13:33:42 -0700
Subject: [PATCH 038/216] fix(studio): harden sandbox security for terminal and
 python tools (#4827)

* fix(studio): harden sandbox security for terminal and python tools

The existing command blocklist used naive str.split() which is trivially
bypassable via quoting, full paths, nested shells, variable expansion,
and cross-tool pivoting through Python os.system/subprocess. Fixes #4818.

Changes:
- Replace str.split() blocklist with shlex.split() + os.path.basename()
  tokenization and regex scanning at shell command boundaries
- Add sanitized subprocess environment (_build_safe_env) that strips
  credentials (HF_TOKEN, WANDB_API_KEY, GH_TOKEN, AWS_*, etc.) and
  restricts PATH to /usr/local/bin:/usr/bin:/bin
- Add PR_SET_NO_NEW_PRIVS via prctl on Linux so sudo/su/pkexec fail
  at the kernel level regardless of how they are invoked
- Add RLIMIT_NPROC (256) and RLIMIT_FSIZE (100MB) to prevent fork
  bombs and disk filling attacks
- Extend AST safety checker to detect os.system(), os.popen(),
  subprocess.run/Popen/call/check_output, os.exec*, os.spawn* calls
  containing blocked commands or dynamic (non-literal) arguments
- Add cross-platform support: cmd.exe on Windows, bash on Unix;
  CREATE_NO_WINDOW flag on Windows, preexec_fn on Unix
- Expand blocklist from 7 to 14 commands: add su, chown, passwd,
  mount, umount, fdisk, kill, killall, pkill
- Apply all layers to both _bash_exec and _python_exec

Zero measurable performance overhead -- shlex parsing and a single
prctl syscall per subprocess fork.

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

* Fix review findings: exception_catching dead code, false positives, process substitution

- Include exception_catching reasons in _check_code_safety so bare
  except-in-loop timeout evasion is actually blocked (was computed in
  _check_signal_escape_patterns but never read by the caller)
- Remove base.split() inner loop that caused false positives on quoted
  text arguments containing blocked words (e.g. echo "kill this process")
- Add targeted nested shell detection for bash/sh/zsh -c arguments
  instead, which catches bash -c 'sudo whoami' without false positives
- Add <() process substitution to the regex character class so
  diff <(rm -rf /path) is also caught
- Fix error message to say "unsafe patterns" instead of specifically
  mentioning signal manipulation when other categories trigger

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

* Address review feedback: regex paths, keyword args, list element scanning

- Regex now matches blocked commands after optional path prefix at shell
  boundaries (catches ls; /usr/bin/sudo and similar)
- Nested shell detection uses os.path.basename so bash -c "/bin/rm" is
  caught
- AST checker now inspects keyword arguments (not just positional) so
  subprocess.run(args="sudo ...", shell=True) is detected
- List elements in subprocess calls are now checked via
  _find_blocked_commands for consistency (catches subprocess.run(["bash",
  "-c", "rm -rf /"]))
- Dynamic argument check uses _is_safe_literal that validates list
  contents are all string literals

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

* Fix nested shell scan to only check the script body, not positional args

bash -c 'script' arg0 arg1 -- only tokens[i+1] is the script body;
subsequent tokens are $0, $1 positional parameters passed to the script
and are not executed as shell commands. Scanning all remaining tokens
caused false positives.

* Add subshell parentheses to regex command boundary detection

(sudo whoami) was not caught because ( was not in the regex character
class for shell command boundaries. Add ( to the set alongside ;, &,
|, backtick, newline.

* Address high-priority review findings from 7 parallel reviewers

- Track from-imports of dangerous functions (from os import system,
  from subprocess import run as r, etc.) via shell_exec_aliases dict
  so bare-name calls are detected by the AST checker
- Include the active Python interpreter and virtualenv directories
  in the sanitized PATH so pip, uv, and Studio packages remain
  accessible in the sandbox
- Add Windows-specific blocked commands (rmdir, takeown, icacls,
  runas, powershell, pwsh) only on win32 platform
- Add os.posix_spawn and os.posix_spawnp to _SHELL_EXEC_FUNCS
- Handle tuple literals same as list literals in AST argument
  inspection (both _extract_strings_from_list and _is_safe_literal)

* Fix false positive on check=True kwargs and recursive nested shell scanning

- Only inspect command-carrying keyword arguments (args, command,
  executable, path, file) in the AST checker, not control flags like
  check=True, text=True, capture_output=True which are booleans and
  were incorrectly flagged as non-literal dynamic arguments
- Replace split() in nested shell detection with recursive call to
  _find_blocked_commands so that quoted commands (bash -c '"sudo"
  whoami') and semicolons (bash -c "sudo;ls") within nested shells
  are properly detected through the full shlex + regex pipeline

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

* Move preexec_fn imports to module level and use find_library for libc

Addresses two Gemini review findings:

1. preexec_fn thread safety: _sandbox_preexec previously imported ctypes
   and resource inside the function body, which runs between fork() and
   exec() in the child process. In a multi-threaded server, this could
   deadlock if the import machinery locks were held by another thread at
   fork time. Now all imports and the libc handle are resolved once at
   module load time, so _sandbox_preexec only calls C-level functions
   (prctl, setrlimit) with no Python import activity.

2. Hardcoded libc.so.6 path: replaced with ctypes.util.find_library("c")
   which works on glibc (libc.so.6), musl (libc.musl-*.so.1), and other
   Linux distributions where libc has a different soname.

* Apply Gemini style suggestions: combined regex, dict.fromkeys, constant hoisting

- Combine per-word regex loop into a single re.findall with alternation
  pattern, avoiding repeated regex compilation and searching
- Replace manual dedup loop with dict.fromkeys for PATH entries
- Hoist _CMD_KWARGS frozenset out of visit_Call to avoid recreating it
  on every AST node visit

* Add cmd /c nested shell detection for Windows parity

The nested shell scan only checked for Unix shells (bash -c, sh -c, etc).
Add cmd /c and cmd.exe /c detection so that Windows nested shell
invocations are also recursively scanned for blocked commands. The token
scan already catches blocked commands at any position, so this is
defense-in-depth for consistency across platforms.

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

* Handle combined shell flags (-lc, -xc) and interleaved flags (--login -c)

The nested shell scan only matched token == "-c" with the immediately
preceding token being a shell name. This missed:
- Combined flags: bash -lc 'rm ...' (-lc ends with c, is a valid
  combined flag meaning -l -c)
- Interleaved flags: bash --login -c 'sudo ...' (--login sits between
  bash and -c)

Now matches any short flag ending in 'c' (e.g. -lc, -xc, -ic) and
walks backwards past intermediate flags to find the shell binary.

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

* Fix /bin/bash bypass, remove RLIMIT_NPROC, reduce AST false positives

Addresses three high-consensus findings from 20-reviewer pass:

1. /bin/bash -c 'sudo whoami' bypassed nested shell scan because the
   backwards flag-skip logic treated paths starting with / as flags.
   Now only skips tokens starting with - as Unix flags; on Windows
   only skips short /X flags (not /bin/bash style paths). [9/20]

2. RLIMIT_NPROC=256 caused subprocess.run to fail with EAGAIN because
   Linux enforces NPROC per real UID, not per process tree. Removed
   RLIMIT_NPROC entirely; RLIMIT_FSIZE and PR_SET_NO_NEW_PRIVS remain
   as the primary resource and privilege controls. [5/20]

3. AST checker rejected safe dynamic subprocess usage like
   cmd=["git","status"]; subprocess.run(cmd) as shell_escape_dynamic.
   Now only flags dynamic args for shell-string functions (os.system,
   os.popen, subprocess.getoutput, etc.) or when shell=True is
   explicitly set. List-based subprocess calls with shell=False (the
   default) do not pass through a shell and are not flagged. [12/20]

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

* Handle Windows drive letter paths and .exe extensions in command detection

Gemini review found that Windows absolute paths (C:\Windows\System32\
shutdown.exe) and executable extensions (.exe, .com, .bat, .cmd) were
not handled:

- Token scan now strips .exe/.com/.bat/.cmd extensions before checking
  the blocklist, so sudo.exe matches sudo, shutdown.bat matches shutdown
- Regex pattern now includes optional Windows drive letter prefix
  ([a-zA-Z]:[/\\]) and optional executable extension suffix, so commands
  after shell metacharacters with full Windows paths are also caught

* Handle **kwargs dict expansion, non-literal shell=, and except Exception false positive

Addresses three findings from second 20-reviewer pass:

1. **kwargs dict expansion (9/20): subprocess.run(**{"args": "rm ...",
   "shell": True}) bypassed the AST checker because **kwargs were
   treated as opaque. Now expands literal dict **kwargs to inspect
   their keys, and flags opaque **kwargs (variable dicts) as unsafe.

2. Non-literal shell= values (7/20): shell=variable was treated as
   shell=False (safe). Now any shell= value that is not literally
   False is treated as potentially True (conservative default).

3. except Exception false positive (1/20): except Exception in a loop
   was flagged as timeout evasion, but Exception does not catch
   SystemExit or KeyboardInterrupt which are used for timeout
   enforcement. Narrowed to only flag except BaseException and
   except TimeoutError in loops.

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

---------

Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com>
---
 studio/backend/core/inference/tools.py | 481 ++++++++++++++++++++++++-
 1 file changed, 466 insertions(+), 15 deletions(-)

diff --git a/studio/backend/core/inference/tools.py b/studio/backend/core/inference/tools.py
index 86c22ce25a..87cc933d4b 100644
--- a/studio/backend/core/inference/tools.py
+++ b/studio/backend/core/inference/tools.py
@@ -14,6 +14,8 @@ import os
 os.environ["UNSLOTH_IS_PRESENT"] = "1"
 
 import random
+import re
+import shlex
 import ssl
 import subprocess
 import sys
@@ -27,11 +29,235 @@ logger = get_logger(__name__)
 
 _EXEC_TIMEOUT = 300  # 5 minutes
 
+# Pre-import modules used in _sandbox_preexec at module level so that
+# the preexec_fn closure does not trigger the import machinery in the
+# forked child (which can deadlock in multi-threaded servers).
+_libc = None
+if sys.platform == "linux":
+    try:
+        import ctypes
+        import ctypes.util
+
+        _libc_name = ctypes.util.find_library("c")
+        if _libc_name:
+            _libc = ctypes.CDLL(_libc_name, use_errno = True)
+    except (OSError, AttributeError):
+        pass
+
+_resource = None
+if sys.platform != "win32":
+    try:
+        import resource as _resource
+    except ImportError:
+        pass
+
 # Strict raster-image allowlist for sandbox file serving.
 # No .svg (XSS risk via embedded scripts), no .html, no .pdf.
 _IMAGE_EXTS = frozenset({".png", ".jpg", ".jpeg", ".gif", ".webp", ".bmp"})
 _MAX_OUTPUT_CHARS = 8000  # truncate long output
-_BASH_BLOCKED_WORDS = {"rm", "sudo", "dd", "chmod", "mkfs", "shutdown", "reboot"}
+_BLOCKED_COMMANDS_COMMON = frozenset(
+    {
+        "rm",
+        "sudo",
+        "su",
+        "dd",
+        "chmod",
+        "chown",
+        "mkfs",
+        "shutdown",
+        "reboot",
+        "passwd",
+        "mount",
+        "umount",
+        "fdisk",
+        "kill",
+        "killall",
+        "pkill",
+    }
+)
+_BLOCKED_COMMANDS_WIN = frozenset(
+    {
+        "rmdir",
+        "takeown",
+        "icacls",
+        "runas",
+        "powershell",
+        "pwsh",
+    }
+)
+_BLOCKED_COMMANDS = (
+    _BLOCKED_COMMANDS_COMMON | _BLOCKED_COMMANDS_WIN
+    if sys.platform == "win32"
+    else _BLOCKED_COMMANDS_COMMON
+)
+
+
+def _find_blocked_commands(command: str) -> set[str]:
+    """Detect blocked commands using shlex tokenization and regex scanning.
+
+    Catches: full paths (/usr/bin/sudo), quoted strings ("sudo"),
+    split-quotes (su""do), backslash escapes (\\rm), and command-position
+    words after ;, |, &&, $().
+    """
+    blocked = set()
+
+    # 1. shlex tokenization (handles quotes, escapes, concatenation)
+    try:
+        tokens = (
+            shlex.split(command)
+            if sys.platform != "win32"
+            else shlex.split(command, posix = False)
+        )
+    except ValueError:
+        tokens = command.split()
+
+    for token in tokens:
+        base = os.path.basename(token).lower()
+        # Strip common Windows executable extensions so that
+        # runas.exe, shutdown.bat, etc. match the blocklist.
+        stem, ext = os.path.splitext(base)
+        if ext in {".exe", ".com", ".bat", ".cmd"}:
+            base = stem
+        if base in _BLOCKED_COMMANDS:
+            blocked.add(base)
+
+    # 2. Regex: catch blocked words at shell command boundaries
+    #    (semicolons, pipes, &&, ||, backticks, $(), <(), subshells, newlines)
+    #    Uses a single combined pattern for all blocked words.
+    #    Handles optional Unix path prefix (/usr/bin/) and Windows drive
+    #    letter prefix (C:\Windows\...\).
+    lowered = command.lower()
+    if _BLOCKED_COMMANDS:
+        words_alt = "|".join(re.escape(w) for w in sorted(_BLOCKED_COMMANDS))
+        pattern = (
+            rf"(?:^|[;&|`\n(]\s*|[$]\(\s*|<\(\s*)"
+            rf"(?:[\w./\\-]*/|[a-zA-Z]:[/\\][\w./\\-]*)?"
+            rf"({words_alt})(?:\.(?:exe|com|bat|cmd))?\b"
+        )
+        blocked.update(re.findall(pattern, lowered))
+
+    # 3. Check for nested shell invocations (bash -c 'sudo whoami',
+    #    bash -lc '...', bash --login -c '...', cmd /c '...').
+    #    When a -c or /c flag is found, look backwards for a shell name
+    #    (skipping intermediate flags like --login, -l, -x) and recursively
+    #    scan the nested command string.
+    _SHELLS = {"bash", "sh", "zsh", "dash", "ksh", "csh", "tcsh", "fish"}
+    _SHELLS_WIN = {"cmd", "cmd.exe"}
+    for i, token in enumerate(tokens):
+        tok_lower = token.lower()
+        # Match -c exactly, or combined flags ending in c (e.g. -lc, -xc)
+        is_unix_c = tok_lower == "-c" or (
+            tok_lower.startswith("-")
+            and tok_lower.endswith("c")
+            and not tok_lower.startswith("--")
+        )
+        is_win_c = tok_lower == "/c"
+        if not (is_unix_c or is_win_c) or i < 1 or i + 1 >= len(tokens):
+            continue
+        # Look backwards past any flags to find the shell binary.
+        # On Unix, flags start with - (skip those). On Windows, flags
+        # start with / but so do absolute paths, so only skip short
+        # single-char /X flags (not /bin/bash style paths).
+        for j in range(i - 1, -1, -1):
+            prev = tokens[j]
+            if prev.startswith("-"):
+                continue  # skip Unix flags like --login, -l
+            if is_win_c and prev.startswith("/") and len(prev) <= 3:
+                continue  # skip Windows flags like /s, /q (not /bin/bash)
+            prev_base = os.path.basename(prev).lower()
+            if is_unix_c and prev_base in _SHELLS:
+                blocked |= _find_blocked_commands(tokens[i + 1])
+            elif is_win_c and prev_base in _SHELLS_WIN:
+                blocked |= _find_blocked_commands(tokens[i + 1])
+            break  # stop at first non-flag token
+
+    return blocked
+
+
+def _build_safe_env(workdir: str) -> dict[str, str]:
+    """Build a minimal, credential-free environment for sandboxed subprocesses.
+
+    Strips HF_TOKEN, WANDB_API_KEY, AWS_*, GH_TOKEN, LD_PRELOAD, DYLD_*, etc.
+    Preserves the active Python interpreter and virtualenv directories in PATH
+    so that pip, uv, and packages installed in the Studio runtime remain
+    accessible.
+    """
+    # Start with the directory containing the running Python interpreter
+    # so that subprocess calls to 'python', 'pip', etc. resolve to the
+    # same environment the Studio server is running in.
+    exe_dir = os.path.dirname(sys.executable)
+    path_entries = [exe_dir] if exe_dir else []
+
+    # If a virtualenv is active, include its bin/Scripts directory.
+    venv = os.environ.get("VIRTUAL_ENV")
+    if venv:
+        venv_bin = os.path.join(venv, "Scripts" if sys.platform == "win32" else "bin")
+        if venv_bin not in path_entries:
+            path_entries.append(venv_bin)
+
+    if sys.platform == "win32":
+        sysroot = os.environ.get("SystemRoot", r"C:\Windows")
+        path_entries.extend([os.path.join(sysroot, "System32"), sysroot])
+    else:
+        path_entries.extend(["/usr/local/bin", "/usr/bin", "/bin"])
+
+    # Deduplicate while preserving order
+    deduped = list(dict.fromkeys(p for p in path_entries if p))
+
+    env = {
+        "PATH": os.pathsep.join(deduped),
+        "HOME": workdir,
+        "TMPDIR": workdir,
+        "LANG": os.environ.get("LANG", "C.UTF-8"),
+        "TERM": "dumb",
+        "PYTHONIOENCODING": "utf-8",
+    }
+    if venv:
+        env["VIRTUAL_ENV"] = venv
+    # Windows needs SystemRoot for Python/subprocess to work
+    if sys.platform == "win32":
+        env["SystemRoot"] = os.environ.get("SystemRoot", r"C:\Windows")
+    return env
+
+
+def _sandbox_preexec():
+    """Pre-exec hook: drop privilege escalation ability and set resource limits.
+
+    On Linux, applies PR_SET_NO_NEW_PRIVS so sudo/su/pkexec fail at the
+    kernel level. On Linux and macOS, sets RLIMIT_FSIZE.
+    No-op on Windows (use creationflags instead).
+
+    Note: RLIMIT_NPROC is intentionally NOT set because Linux enforces it
+    per real UID, not per process tree, so it would starve the Studio
+    server and other sessions sharing the same user account.
+
+    All modules and handles are resolved at import time (module level) so
+    this function does not trigger Python imports in the forked child,
+    avoiding potential deadlocks in multi-threaded servers.
+    """
+    if _libc is not None:
+        try:
+            # PR_SET_NO_NEW_PRIVS = 38, arg2 = 1 (enable)
+            _libc.prctl(38, 1, 0, 0, 0)
+        except (OSError, AttributeError):
+            pass  # Not available (container, old kernel, etc.)
+
+    if _resource is not None:
+        try:
+            # Limit file size to 100MB (prevents disk filling)
+            _resource.setrlimit(
+                _resource.RLIMIT_FSIZE, (100 * 1024 * 1024, 100 * 1024 * 1024)
+            )
+        except (ValueError, OSError):
+            pass
+
+
+def _get_shell_cmd(command: str) -> list[str]:
+    """Return the platform-appropriate shell invocation for a command string."""
+    if sys.platform == "win32":
+        return ["cmd", "/c", command]
+    return ["bash", "-c", command]
+
 
 # Per-session working directories so each chat thread gets its own sandbox.
 # Falls back to a shared ~/studio_sandbox/_default for API callers without a
@@ -429,6 +655,7 @@ def _check_signal_escape_patterns(code: str):
 
     signal_tampering = []
     exception_catching = []
+    shell_escapes = []
     warnings = []
 
     def _ast_name_matches(node, names):
@@ -446,10 +673,84 @@ def _check_signal_escape_patterns(code: str):
             return full_name in names
         return False
 
+    # Dangerous os/subprocess functions that can execute shell commands
+    _SHELL_EXEC_FUNCS = frozenset(
+        {
+            "os.system",
+            "os.popen",
+            "os.popen2",
+            "os.popen3",
+            "os.popen4",
+            "os.execl",
+            "os.execle",
+            "os.execlp",
+            "os.execlpe",
+            "os.execv",
+            "os.execve",
+            "os.execvp",
+            "os.execvpe",
+            "os.spawnl",
+            "os.spawnle",
+            "os.spawnlp",
+            "os.spawnlpe",
+            "os.spawnv",
+            "os.spawnve",
+            "os.spawnvp",
+            "os.spawnvpe",
+            "os.posix_spawn",
+            "os.posix_spawnp",
+            "subprocess.run",
+            "subprocess.call",
+            "subprocess.check_call",
+            "subprocess.check_output",
+            "subprocess.Popen",
+            "subprocess.getoutput",
+            "subprocess.getstatusoutput",
+        }
+    )
+
+    def _extract_string_from_node(node):
+        """Extract a plain string value from an AST node, if it is a constant."""
+        if isinstance(node, ast.Constant) and isinstance(node.value, str):
+            return node.value
+        return None
+
+    def _extract_strings_from_list(node):
+        """Extract string elements from an AST List or Tuple node."""
+        if isinstance(node, (ast.List, ast.Tuple)):
+            parts = []
+            for elt in node.elts:
+                s = _extract_string_from_node(elt)
+                if s is not None:
+                    parts.append(s)
+            return parts
+        return []
+
+    # Keyword argument names that carry command content (as opposed to
+    # control flags like check=True, text=True, capture_output=True).
+    _CMD_KWARGS = frozenset({"args", "command", "executable", "path", "file"})
+
+    def _check_args_for_blocked(args_nodes):
+        """Check if any call arguments contain blocked commands."""
+        found = set()
+        for arg in args_nodes:
+            s = _extract_string_from_node(arg)
+            if s is not None:
+                found |= _find_blocked_commands(s)
+            strs = _extract_strings_from_list(arg)
+            for s in strs:
+                found |= _find_blocked_commands(s)
+        return found
+
     class SignalEscapeVisitor(ast.NodeVisitor):
         def __init__(self):
             self.imports_signal = False
             self.signal_aliases = {"signal"}
+            self.os_aliases = {"os"}
+            self.subprocess_aliases = {"subprocess"}
+            # Maps bare function names to their fully-qualified form
+            # for from-import tracking (e.g. "system" -> "os.system")
+            self.shell_exec_aliases: dict[str, str] = {}
             self.loop_depth = 0
 
         def visit_Import(self, node):
@@ -458,6 +759,10 @@ def _check_signal_escape_patterns(code: str):
                     self.imports_signal = True
                     if alias.asname:
                         self.signal_aliases.add(alias.asname)
+                elif alias.name == "os":
+                    self.os_aliases.add(alias.asname or "os")
+                elif alias.name == "subprocess":
+                    self.subprocess_aliases.add(alias.asname or "subprocess")
             self.generic_visit(node)
 
         def visit_ImportFrom(self, node):
@@ -475,6 +780,16 @@ def _check_signal_escape_patterns(code: str):
                         "alarm",
                     ):
                         self.signal_aliases.add(alias.asname or alias.name)
+            elif node.module in ("os", "subprocess"):
+                if node.module == "os":
+                    self.os_aliases.add("os")
+                else:
+                    self.subprocess_aliases.add("subprocess")
+                # Track from-imports of dangerous functions
+                for alias in node.names:
+                    fq = f"{node.module}.{alias.name}"
+                    if fq in _SHELL_EXEC_FUNCS:
+                        self.shell_exec_aliases[alias.asname or alias.name] = fq
             self.generic_visit(node)
 
         def visit_While(self, node):
@@ -539,6 +854,111 @@ def _check_signal_escape_patterns(code: str):
                             "description": "Modifies signal mask (may block SIGALRM)",
                         }
                     )
+
+            # --- Shell escape detection ---
+            # Resolve the fully qualified function name for os.*/subprocess.*
+            shell_func = None
+            if isinstance(func, ast.Attribute):
+                if isinstance(func.value, ast.Name):
+                    if func.value.id in self.os_aliases:
+                        shell_func = f"os.{func.attr}"
+                    elif func.value.id in self.subprocess_aliases:
+                        shell_func = f"subprocess.{func.attr}"
+            elif isinstance(func, ast.Name):
+                # Check from-import aliases: from os import system; system(...)
+                shell_func = self.shell_exec_aliases.get(func.id)
+
+            if shell_func and shell_func in _SHELL_EXEC_FUNCS:
+                # Expand **kwargs dicts to inspect their keys
+                expanded_kwargs: dict[str, ast.AST] = {}
+                has_opaque_kwargs = False
+                for kw in node.keywords:
+                    if kw.arg is not None:
+                        expanded_kwargs[kw.arg] = kw.value
+                    elif isinstance(kw.value, ast.Dict):
+                        for k, v in zip(kw.value.keys, kw.value.values):
+                            key = _extract_string_from_node(k) if k else None
+                            if key is not None:
+                                expanded_kwargs[key] = v
+                    else:
+                        has_opaque_kwargs = True
+
+                cmd_kw_values = [
+                    v for k, v in expanded_kwargs.items() if k in _CMD_KWARGS
+                ]
+                all_call_args = list(node.args) + cmd_kw_values
+                blocked_in_args = _check_args_for_blocked(all_call_args)
+
+                if has_opaque_kwargs:
+                    # Can't inspect dynamic **kwargs -- flag as unsafe
+                    shell_escapes.append(
+                        {
+                            "type": "shell_escape_dynamic",
+                            "line": node.lineno,
+                            "description": (
+                                f"{shell_func}() called with dynamic **kwargs"
+                            ),
+                        }
+                    )
+                elif blocked_in_args:
+                    shell_escapes.append(
+                        {
+                            "type": "shell_escape",
+                            "line": node.lineno,
+                            "description": (
+                                f"{shell_func}() invokes blocked command(s): "
+                                f"{', '.join(sorted(blocked_in_args))}"
+                            ),
+                        }
+                    )
+                else:
+                    # Only flag dynamic args for functions that interpret
+                    # strings as shell commands, or when shell= might be
+                    # enabled.  Treat any non-literal-False shell= value
+                    # as potentially True (conservative).
+                    _STRING_SHELL_FUNCS = frozenset(
+                        {
+                            "os.system",
+                            "os.popen",
+                            "os.popen2",
+                            "os.popen3",
+                            "os.popen4",
+                            "subprocess.getoutput",
+                            "subprocess.getstatusoutput",
+                        }
+                    )
+                    shell_node = expanded_kwargs.get("shell")
+                    shell_safe = shell_node is None or (
+                        isinstance(shell_node, ast.Constant)
+                        and shell_node.value is False
+                    )
+                    if shell_func in _STRING_SHELL_FUNCS or not shell_safe:
+
+                        def _is_safe_literal(n):
+                            if _extract_string_from_node(n) is not None:
+                                return True
+                            if isinstance(n, (ast.List, ast.Tuple)):
+                                return all(
+                                    _extract_string_from_node(e) is not None
+                                    for e in n.elts
+                                )
+                            return False
+
+                        has_non_literal = any(
+                            not _is_safe_literal(a) for a in all_call_args
+                        )
+                        if has_non_literal:
+                            shell_escapes.append(
+                                {
+                                    "type": "shell_escape_dynamic",
+                                    "line": node.lineno,
+                                    "description": (
+                                        f"{shell_func}() called with non-literal "
+                                        f"shell command (potential shell escape)"
+                                    ),
+                                }
+                            )
+
             self.generic_visit(node)
 
         def visit_ExceptHandler(self, node):
@@ -554,7 +974,12 @@ def _check_signal_escape_patterns(code: str):
                     }
                 )
             elif isinstance(node.type, ast.Name):
-                if node.type.id in ("TimeoutError", "BaseException", "Exception"):
+                # Only flag BaseException and TimeoutError, NOT Exception.
+                # except Exception does not catch SystemExit or
+                # KeyboardInterrupt, so it cannot suppress timeout
+                # enforcement.  Flagging Exception causes false positives
+                # on normal error-handling patterns.
+                if node.type.id in ("TimeoutError", "BaseException"):
                     exception_catching.append(
                         {
                             "type": f"catches_{node.type.id}_in_loop",
@@ -565,7 +990,7 @@ def _check_signal_escape_patterns(code: str):
             elif isinstance(node.type, ast.Tuple):
                 for elt in node.type.elts:
                     if isinstance(elt, ast.Name):
-                        if elt.id in ("TimeoutError", "BaseException", "Exception"):
+                        if elt.id in ("TimeoutError", "BaseException"):
                             exception_catching.append(
                                 {
                                     "type": f"catches_{elt.id}_in_loop",
@@ -581,10 +1006,15 @@ def _check_signal_escape_patterns(code: str):
     if visitor.imports_signal and not signal_tampering:
         warnings.append("Code imports 'signal' module - review manually for safety")
 
-    is_safe = len(signal_tampering) == 0 and len(exception_catching) == 0
+    is_safe = (
+        len(signal_tampering) == 0
+        and len(exception_catching) == 0
+        and len(shell_escapes) == 0
+    )
     return is_safe, {
         "signal_tampering": signal_tampering,
         "exception_catching": exception_catching,
+        "shell_escapes": shell_escapes,
         "warnings": warnings,
     }
 
@@ -605,10 +1035,18 @@ def _check_code_safety(code: str) -> str | None:
         reasons = [
             item.get("description", "") for item in info.get("signal_tampering", [])
         ]
-        return (
-            f"Error: unsafe code detected ({'; '.join(reasons)}). "
-            f"Please remove signal manipulation from your code."
-        )
+        shell_reasons = [
+            item.get("description", "") for item in info.get("shell_escapes", [])
+        ]
+        exception_reasons = [
+            item.get("description", "") for item in info.get("exception_catching", [])
+        ]
+        all_reasons = [r for r in reasons + shell_reasons + exception_reasons if r]
+        if all_reasons:
+            return (
+                f"Error: unsafe code detected ({'; '.join(all_reasons)}). "
+                f"Please remove unsafe patterns from your code."
+            )
 
     return None
 
@@ -663,13 +1101,20 @@ def _python_exec(
         with os.fdopen(fd, "w") as f:
             f.write(code)
 
-        proc = subprocess.Popen(
-            [sys.executable, tmp_path],
+        safe_env = _build_safe_env(workdir)
+        popen_kwargs = dict(
             stdout = subprocess.PIPE,
             stderr = subprocess.STDOUT,
             text = True,
             cwd = workdir,
+            env = safe_env,
         )
+        if sys.platform != "win32":
+            popen_kwargs["preexec_fn"] = _sandbox_preexec
+        else:
+            popen_kwargs["creationflags"] = subprocess.CREATE_NO_WINDOW
+
+        proc = subprocess.Popen([sys.executable, tmp_path], **popen_kwargs)
 
         # Spawn cancel watcher if we have a cancel event
         if cancel_event is not None:
@@ -735,21 +1180,27 @@ def _bash_exec(
     if not command or not command.strip():
         return "No command provided."
 
-    # Block dangerous commands
-    tokens = set(command.lower().split())
-    blocked = tokens & _BASH_BLOCKED_WORDS
+    # Block dangerous commands (shlex + regex based)
+    blocked = _find_blocked_commands(command)
     if blocked:
         return f"Blocked command(s) for safety: {', '.join(sorted(blocked))}"
 
     try:
         workdir = _get_workdir(session_id)
-        proc = subprocess.Popen(
-            ["bash", "-c", command],
+        safe_env = _build_safe_env(workdir)
+        popen_kwargs = dict(
             stdout = subprocess.PIPE,
             stderr = subprocess.STDOUT,
             text = True,
             cwd = workdir,
+            env = safe_env,
         )
+        if sys.platform != "win32":
+            popen_kwargs["preexec_fn"] = _sandbox_preexec
+        else:
+            popen_kwargs["creationflags"] = subprocess.CREATE_NO_WINDOW
+
+        proc = subprocess.Popen(_get_shell_cmd(command), **popen_kwargs)
 
         if cancel_event is not None:
             watcher = threading.Thread(

From a32b871f0e2e2f94cb88fec81596eff0796f1254 Mon Sep 17 00:00:00 2001
From: Daniel Han <danielhanchen@gmail.com>
Date: Fri, 3 Apr 2026 13:56:59 -0700
Subject: [PATCH 039/216] studio: add speculative decoding support (ngram-mod,
 on by default) (#4836)

* studio: add speculative decoding support (ngram-mod, on by default)

Enable n-gram speculative decoding for GGUF models in Unsloth Studio.
Uses llama.cpp's ngram-mod mode which gives 10-40% faster generation
with zero VRAM cost via a 4MB fixed hash table that auto-resets on
low acceptance rates.

Backend:
- Add speculative_type field to LoadRequest, LoadResponse, and
  InferenceStatusResponse pydantic models
- Add speculative_type parameter to LlamaCppBackend.load_model()
  with allowlist validation (ngram-simple, ngram-mod)
- Pass --spec-type, --spec-ngram-size-n 16, --draft-max 24 flags
  to llama-server when ngram-mod is active
- Default to ngram-mod for non-vision GGUF models server-side
- Silently skip speculative decoding for vision models (unsupported
  in llama.cpp server-context.cpp)

Frontend:
- Add speculative_type to TS API types
- Add speculativeType/loadedSpeculativeType to chat runtime store
  with default value of "ngram-mod"
- Add On/Off toggle in Model settings section (GGUF only, hidden
  for vision models), included in dirty check for Apply/Reset
- Wire speculative_type through model load request and response
- Restore speculative type state on page refresh/reconnect

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

* fix: remove server-side speculative decoding override

The backend was overriding speculative_type=None to "ngram-mod" for
non-vision GGUF models, which prevented users from disabling spec
decoding via the UI toggle. The frontend store already defaults to
"ngram-mod", so the backend fallback was redundant and blocked the
explicit "Off" setting.

* fix: use recommended ngram-mod params from llama.cpp docs

Update speculative decoding params to match the recommended values
from llama.cpp docs (docs/speculative.md):
  --spec-ngram-size-n 24 (was 16, docs say small n not recommended)
  --draft-min 48 (was 0)
  --draft-max 64 (was 24, docs note MoEs need long drafts)

Also fix comment: ngram-mod uses ~16 MB (4M entries * 4 bytes),
not 4 MB.

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

* add benchmark table and references to speculative decoding comment

Include speedup numbers from llama.cpp PRs #18471 and #19164 as an
inline comment so future readers understand the expected gains.

---------

Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com>
---
 studio/backend/core/inference/llama_cpp.py    | 47 +++++++++++++++++++
 studio/backend/models/inference.py            | 12 +++++
 studio/backend/routes/inference.py            |  5 ++
 .../src/features/chat/chat-settings-sheet.tsx | 39 ++++++++++++++-
 .../chat/hooks/use-chat-model-runtime.ts      |  9 +++-
 .../chat/stores/chat-runtime-store.ts         |  8 ++++
 .../frontend/src/features/chat/types/api.ts   |  3 ++
 7 files changed, 121 insertions(+), 2 deletions(-)

diff --git a/studio/backend/core/inference/llama_cpp.py b/studio/backend/core/inference/llama_cpp.py
index 4da62e23c3..c84ac640df 100644
--- a/studio/backend/core/inference/llama_cpp.py
+++ b/studio/backend/core/inference/llama_cpp.py
@@ -109,6 +109,7 @@ class LlamaCppBackend:
         self._supports_tools: bool = False
         self._cache_type_kv: Optional[str] = None
         self._reasoning_default: bool = True
+        self._speculative_type: Optional[str] = None
         # KV-cache estimation fields (populated by _read_gguf_metadata)
         self._n_layers: Optional[int] = None
         self._n_kv_heads: Optional[int] = None
@@ -198,6 +199,10 @@ class LlamaCppBackend:
     def cache_type_kv(self) -> Optional[str]:
         return self._cache_type_kv
 
+    @property
+    def speculative_type(self) -> Optional[str]:
+        return self._speculative_type
+
     # ── Binary discovery ──────────────────────────────────────────
 
     @staticmethod
@@ -1055,6 +1060,7 @@ class LlamaCppBackend:
         n_ctx: int = 4096,
         chat_template_override: Optional[str] = None,
         cache_type_kv: Optional[str] = None,
+        speculative_type: Optional[str] = None,
         n_threads: Optional[int] = None,
         n_gpu_layers: Optional[int] = None,  # Accepted for caller compat, unused
     ) -> bool:
@@ -1315,6 +1321,46 @@ class LlamaCppBackend:
             else:
                 self._cache_type_kv = None
 
+            # Speculative decoding (n-gram self-speculation, zero VRAM cost)
+            # ngram-mod: ~16 MB shared hash pool, constant memory/complexity,
+            # variable draft lengths.  Helps most when the model repeats
+            # existing text (code refactoring, summarization, reasoning).
+            # For general chat with low repetition, overhead is ~5 ms.
+            #
+            # Benchmarks from llama.cpp PRs #18471, #19164:
+            #   Scenario                        | Without | With    | Speedup
+            #   gpt-oss-120b code refactor      | 181 t/s | 446 t/s | 2.5x
+            #   Qwen3-235B offloaded            |  12 t/s |  21 t/s | 1.8x
+            #   gpt-oss-120b repeat (92% accept)| 181 t/s | 814 t/s | 4.5x
+            #
+            # Params from llama.cpp docs (docs/speculative.md):
+            #   --spec-ngram-size-n 24  (small n not recommended)
+            #   --draft-min 48 --draft-max 64 (MoEs need long drafts;
+            #     dense models can reduce these)
+            # ref: https://github.com/ggml-org/llama.cpp/blob/master/docs/speculative.md
+            # ref: https://github.com/ggml-org/llama.cpp/pull/19164
+            # ref: https://github.com/ggml-org/llama.cpp/pull/18471
+            _valid_spec_types = {"ngram-simple", "ngram-mod"}
+            if speculative_type and speculative_type in _valid_spec_types:
+                if not is_vision:  # spec decoding disabled for vision models
+                    cmd.extend(["--spec-type", speculative_type])
+                    if speculative_type == "ngram-mod":
+                        cmd.extend(
+                            [
+                                "--spec-ngram-size-n",
+                                "24",
+                                "--draft-min",
+                                "48",
+                                "--draft-max",
+                                "64",
+                            ]
+                        )
+                    self._speculative_type = speculative_type
+                else:
+                    self._speculative_type = None
+            else:
+                self._speculative_type = None
+
             # Apply custom chat template override if provided
             if chat_template_override:
                 import tempfile
@@ -1553,6 +1599,7 @@ class LlamaCppBackend:
             self._reasoning_always_on = False
             self._supports_tools = False
             self._cache_type_kv = None
+            self._speculative_type = None
             self._n_layers = None
             self._n_kv_heads = None
             self._n_heads = None
diff --git a/studio/backend/models/inference.py b/studio/backend/models/inference.py
index 3094df4169..cf08ecbc12 100644
--- a/studio/backend/models/inference.py
+++ b/studio/backend/models/inference.py
@@ -48,6 +48,10 @@ class LoadRequest(BaseModel):
         None,
         description = "Physical GPU indices to use, for example [0, 1]. Omit or pass [] to use automatic selection. Explicit gpu_ids are unsupported when the parent CUDA_VISIBLE_DEVICES uses UUID/MIG entries. Not supported for GGUF models.",
     )
+    speculative_type: Optional[str] = Field(
+        None,
+        description = "Speculative decoding mode for GGUF models (e.g. 'ngram-simple', 'ngram-mod'). Ignored for non-GGUF and vision models.",
+    )
 
 
 class UnloadRequest(BaseModel):
@@ -163,6 +167,10 @@ class LoadResponse(BaseModel):
         None,
         description = "Jinja2 chat template string (from GGUF metadata or tokenizer)",
     )
+    speculative_type: Optional[str] = Field(
+        None,
+        description = "Active speculative decoding mode (e.g. 'ngram-simple', 'ngram-mod'), or None if disabled",
+    )
 
 
 class UnloadResponse(BaseModel):
@@ -225,6 +233,10 @@ class InferenceStatusResponse(BaseModel):
         None,
         description = "Model's native context length from GGUF metadata (not capped by VRAM)",
     )
+    speculative_type: Optional[str] = Field(
+        None,
+        description = "Active speculative decoding mode (e.g. 'ngram-simple', 'ngram-mod'), or None if disabled",
+    )
 
 
 # =====================================================================
diff --git a/studio/backend/routes/inference.py b/studio/backend/routes/inference.py
index ced24c1d5f..30ff7da49c 100644
--- a/studio/backend/routes/inference.py
+++ b/studio/backend/routes/inference.py
@@ -179,6 +179,7 @@ async def load_model(
                     supports_reasoning = llama_backend.supports_reasoning,
                     reasoning_always_on = llama_backend.reasoning_always_on,
                     chat_template = llama_backend.chat_template,
+                    speculative_type = llama_backend.speculative_type,
                 )
         else:
             if (
@@ -263,6 +264,7 @@ async def load_model(
                     n_ctx = request.max_seq_length,
                     chat_template_override = request.chat_template_override,
                     cache_type_kv = request.cache_type_kv,
+                    speculative_type = request.speculative_type,
                 )
             else:
                 # Local mode: llama-server loads via -m <path>
@@ -275,6 +277,7 @@ async def load_model(
                     n_ctx = request.max_seq_length,
                     chat_template_override = request.chat_template_override,
                     cache_type_kv = request.cache_type_kv,
+                    speculative_type = request.speculative_type,
                 )
 
             if not success:
@@ -317,6 +320,7 @@ async def load_model(
                 supports_tools = llama_backend.supports_tools,
                 cache_type_kv = llama_backend.cache_type_kv,
                 chat_template = llama_backend.chat_template,
+                speculative_type = llama_backend.speculative_type,
             )
 
         # ── Standard path: load via Unsloth/transformers ──────────
@@ -652,6 +656,7 @@ async def get_status(
                 context_length = llama_backend.context_length,
                 max_context_length = llama_backend.max_context_length,
                 native_context_length = llama_backend.native_context_length,
+                speculative_type = llama_backend.speculative_type,
             )
 
         # Otherwise, report Unsloth backend status
diff --git a/studio/frontend/src/features/chat/chat-settings-sheet.tsx b/studio/frontend/src/features/chat/chat-settings-sheet.tsx
index e5b0814343..550df2bf7c 100644
--- a/studio/frontend/src/features/chat/chat-settings-sheet.tsx
+++ b/studio/frontend/src/features/chat/chat-settings-sheet.tsx
@@ -280,6 +280,15 @@ export function ChatSettingsPanel({
 }: ChatSettingsPanelProps) {
   const isMobile = useIsMobile();
   const isGguf = useChatRuntimeStore((s) => s.activeGgufVariant) != null;
+  const speculativeType = useChatRuntimeStore((s) => s.speculativeType);
+  const setSpeculativeType = useChatRuntimeStore((s) => s.setSpeculativeType);
+  const loadedSpeculativeType = useChatRuntimeStore(
+    (s) => s.loadedSpeculativeType,
+  );
+  const currentModels = useChatRuntimeStore((s) => s.models);
+  const currentCheckpoint = params.checkpoint;
+  const currentModelIsVision =
+    currentModels.find((m) => m.id === currentCheckpoint)?.isVision ?? false;
   const ggufContextLength = useChatRuntimeStore((s) => s.ggufContextLength);
   const ggufMaxContextLength = useChatRuntimeStore(
     (s) => s.ggufMaxContextLength,
@@ -299,7 +308,8 @@ export function ChatSettingsPanel({
   const ctxMaxValue = ggufNativeContextLength ?? ggufContextLength ?? null;
   const kvDirty = kvCacheDtype !== loadedKvCacheDtype;
   const ctxDirty = customContextLength !== null;
-  const modelSettingsDirty = kvDirty || ctxDirty;
+  const specDirty = speculativeType !== loadedSpeculativeType;
+  const modelSettingsDirty = kvDirty || ctxDirty || specDirty;
   const [customPresets, setCustomPresets] = useState<Preset[]>(() =>
     loadSavedCustomPresets(),
   );
@@ -580,6 +590,32 @@ export function ChatSettingsPanel({
                     </SelectContent>
                   </Select>
                 </div>
+                {!currentModelIsVision && (
+                  <div className="flex items-center justify-between gap-3">
+                    <div className="min-w-0">
+                      <div className="text-xs font-medium">
+                        Speculative Decoding
+                      </div>
+                      <div className="text-[11px] text-muted-foreground">
+                        Speed up generation with no VRAM cost.
+                      </div>
+                    </div>
+                    <Select
+                      value={speculativeType ?? "off"}
+                      onValueChange={(v) => {
+                        setSpeculativeType(v === "off" ? null : v);
+                      }}
+                    >
+                      <SelectTrigger className="h-7 w-[120px] text-xs">
+                        <SelectValue />
+                      </SelectTrigger>
+                      <SelectContent>
+                        <SelectItem value="ngram-mod">On</SelectItem>
+                        <SelectItem value="off">Off</SelectItem>
+                      </SelectContent>
+                    </Select>
+                  </div>
+                )}
                 {modelSettingsDirty && (
                   <div className="flex flex-wrap gap-1.5 pt-1">
                     <button
@@ -594,6 +630,7 @@ export function ChatSettingsPanel({
                       onClick={() => {
                         setCustomContextLength(null);
                         setKvCacheDtype(loadedKvCacheDtype);
+                        setSpeculativeType(loadedSpeculativeType);
                       }}
                       className="rounded-md border px-2.5 py-1 text-[11px] font-medium text-muted-foreground transition-colors hover:bg-accent"
                     >
diff --git a/studio/frontend/src/features/chat/hooks/use-chat-model-runtime.ts b/studio/frontend/src/features/chat/hooks/use-chat-model-runtime.ts
index 8fcd22d0f8..50b0200a56 100644
--- a/studio/frontend/src/features/chat/hooks/use-chat-model-runtime.ts
+++ b/studio/frontend/src/features/chat/hooks/use-chat-model-runtime.ts
@@ -250,6 +250,7 @@ export function useChatModelRuntime() {
         const ggufNativeContextLength = statusRes.is_gguf
           ? (statusRes.native_context_length ?? null)
           : null;
+        const currentSpecType = statusRes.speculative_type ?? null;
         useChatRuntimeStore.setState({
           supportsReasoning,
           reasoningAlwaysOn,
@@ -257,6 +258,8 @@ export function useChatModelRuntime() {
           ggufContextLength: currentGgufContextLength,
           ggufMaxContextLength,
           ggufNativeContextLength,
+          speculativeType: currentSpecType,
+          loadedSpeculativeType: currentSpecType,
         });
 
         // Set reasoning default for Qwen3.5 small models
@@ -393,7 +396,7 @@ export function useChatModelRuntime() {
               previousWasUnloaded = true;
             }
 
-            const { chatTemplateOverride, kvCacheDtype, customContextLength, ggufContextLength } = useChatRuntimeStore.getState();
+            const { chatTemplateOverride, kvCacheDtype, customContextLength, ggufContextLength, speculativeType } = useChatRuntimeStore.getState();
             // GGUF: use custom context length, or 0 = model's native context
             // Non-GGUF: use the Max Seq Length slider value
             const effectiveMaxSeqLength = customContextLength != null
@@ -409,6 +412,7 @@ export function useChatModelRuntime() {
               trust_remote_code: paramsBeforeLoad.trustRemoteCode ?? false,
               chat_template_override: chatTemplateOverride,
               cache_type_kv: kvCacheDtype,
+              speculative_type: speculativeType,
             });
 
             // If cancelled while loading, don't update UI to show
@@ -431,6 +435,7 @@ export function useChatModelRuntime() {
               }
             }
             const loadedKv = loadResponse.cache_type_kv ?? null;
+            const loadedSpec = loadResponse.speculative_type ?? null;
             const nativeCtx = loadResponse.is_gguf
               ? (loadResponse.context_length ?? 131072)
               : null;
@@ -457,6 +462,8 @@ export function useChatModelRuntime() {
               codeToolsEnabled: loadResponse.supports_tools ?? false,
               kvCacheDtype: loadedKv,
               loadedKvCacheDtype: loadedKv,
+              speculativeType: loadedSpec,
+              loadedSpeculativeType: loadedSpec,
               customContextLength: keepCustomCtx,
               defaultChatTemplate: loadResponse.chat_template ?? null,
               chatTemplateOverride: null,
diff --git a/studio/frontend/src/features/chat/stores/chat-runtime-store.ts b/studio/frontend/src/features/chat/stores/chat-runtime-store.ts
index 48abaf7580..c86226ce4e 100644
--- a/studio/frontend/src/features/chat/stores/chat-runtime-store.ts
+++ b/studio/frontend/src/features/chat/stores/chat-runtime-store.ts
@@ -165,6 +165,8 @@ type ChatRuntimeStore = {
   toolCallTimeout: number;
   kvCacheDtype: string | null;
   loadedKvCacheDtype: string | null;
+  speculativeType: string | null;
+  loadedSpeculativeType: string | null;
   customContextLength: number | null;
   defaultChatTemplate: string | null;
   chatTemplateOverride: string | null;
@@ -198,6 +200,7 @@ type ChatRuntimeStore = {
   setMaxToolCallsPerMessage: (value: number) => void;
   setToolCallTimeout: (value: number) => void;
   setKvCacheDtype: (dtype: string | null) => void;
+  setSpeculativeType: (type: string | null) => void;
   setCustomContextLength: (v: number | null) => void;
   setChatTemplateOverride: (template: string | null) => void;
   setPendingAudio: (base64: string, name: string) => void;
@@ -230,6 +233,8 @@ export const useChatRuntimeStore = create<ChatRuntimeStore>((set) => ({
   toolCallTimeout: loadInt(TOOL_CALL_TIMEOUT_KEY, 5),
   kvCacheDtype: null,
   loadedKvCacheDtype: null,
+  speculativeType: "ngram-mod",
+  loadedSpeculativeType: null,
   customContextLength: null,
   defaultChatTemplate: null,
   chatTemplateOverride: null,
@@ -302,6 +307,8 @@ export const useChatRuntimeStore = create<ChatRuntimeStore>((set) => ({
       toolStatus: null,
       kvCacheDtype: null,
       loadedKvCacheDtype: null,
+      speculativeType: "ngram-mod",
+      loadedSpeculativeType: null,
       customContextLength: null,
       defaultChatTemplate: null,
       chatTemplateOverride: null,
@@ -327,6 +334,7 @@ export const useChatRuntimeStore = create<ChatRuntimeStore>((set) => ({
       return { toolCallTimeout };
     }),
   setKvCacheDtype: (kvCacheDtype) => set({ kvCacheDtype }),
+  setSpeculativeType: (speculativeType) => set({ speculativeType }),
   setCustomContextLength: (customContextLength) => set({ customContextLength }),
   setChatTemplateOverride: (chatTemplateOverride) => set({ chatTemplateOverride }),
   setPendingAudio: (base64, name) =>
diff --git a/studio/frontend/src/features/chat/types/api.ts b/studio/frontend/src/features/chat/types/api.ts
index 8f0839615f..90489222c3 100644
--- a/studio/frontend/src/features/chat/types/api.ts
+++ b/studio/frontend/src/features/chat/types/api.ts
@@ -41,6 +41,7 @@ export interface LoadModelRequest {
   trust_remote_code?: boolean;
   chat_template_override?: string | null;
   cache_type_kv?: string | null;
+  speculative_type?: string | null;
 }
 
 export interface ValidateModelResponse {
@@ -93,6 +94,7 @@ export interface LoadModelResponse {
   supports_tools?: boolean;
   cache_type_kv?: string | null;
   chat_template?: string | null;
+  speculative_type?: string | null;
 }
 
 export interface UnloadModelRequest {
@@ -123,6 +125,7 @@ export interface InferenceStatusResponse {
   context_length?: number | null;
   max_context_length?: number | null;
   native_context_length?: number | null;
+  speculative_type?: string | null;
 }
 
 export interface AudioGenerationResponse {

From 4f65cc94bc365965178a0e42b58e64bf7a9c7065 Mon Sep 17 00:00:00 2001
From: Daniel Han <danielhanchen@gmail.com>
Date: Fri, 3 Apr 2026 13:57:15 -0700
Subject: [PATCH 040/216] Add Gemma 4 model sampling defaults (#4838)

Add per-model YAML configs and MODEL_NAME_MAPPING entries for all 8
Gemma 4 models (4 instruct + 4 base):
- gemma-4-31B-it / gemma-4-31B
- gemma-4-26B-A4B-it / gemma-4-26B-A4B
- gemma-4-E2B-it / gemma-4-E2B
- gemma-4-E4B-it / gemma-4-E4B

GGUF variants (only for -it models) resolve via the gemma-4 family
entry in inference_defaults.json.

Sampling defaults: temperature=1.0, top_p=0.95, top_k=64, min_p=0.0,
no repetition or presence penalty. Matches gemma-3n and gemma-3.
---
 .../assets/configs/inference_defaults.json    | 10 +++-
 .../gemma/unsloth_gemma-4-26B-A4B-it.yaml     | 47 +++++++++++++++++++
 .../gemma/unsloth_gemma-4-26B-A4B.yaml        | 47 +++++++++++++++++++
 .../gemma/unsloth_gemma-4-31B-it.yaml         | 47 +++++++++++++++++++
 .../gemma/unsloth_gemma-4-31B.yaml            | 47 +++++++++++++++++++
 .../gemma/unsloth_gemma-4-E2B-it.yaml         | 47 +++++++++++++++++++
 .../gemma/unsloth_gemma-4-E2B.yaml            | 47 +++++++++++++++++++
 .../gemma/unsloth_gemma-4-E4B-it.yaml         | 47 +++++++++++++++++++
 .../gemma/unsloth_gemma-4-E4B.yaml            | 47 +++++++++++++++++++
 studio/backend/utils/models/model_config.py   | 32 +++++++++++++
 10 files changed, 417 insertions(+), 1 deletion(-)
 create mode 100644 studio/backend/assets/configs/model_defaults/gemma/unsloth_gemma-4-26B-A4B-it.yaml
 create mode 100644 studio/backend/assets/configs/model_defaults/gemma/unsloth_gemma-4-26B-A4B.yaml
 create mode 100644 studio/backend/assets/configs/model_defaults/gemma/unsloth_gemma-4-31B-it.yaml
 create mode 100644 studio/backend/assets/configs/model_defaults/gemma/unsloth_gemma-4-31B.yaml
 create mode 100644 studio/backend/assets/configs/model_defaults/gemma/unsloth_gemma-4-E2B-it.yaml
 create mode 100644 studio/backend/assets/configs/model_defaults/gemma/unsloth_gemma-4-E2B.yaml
 create mode 100644 studio/backend/assets/configs/model_defaults/gemma/unsloth_gemma-4-E4B-it.yaml
 create mode 100644 studio/backend/assets/configs/model_defaults/gemma/unsloth_gemma-4-E4B.yaml

diff --git a/studio/backend/assets/configs/inference_defaults.json b/studio/backend/assets/configs/inference_defaults.json
index f520041add..1b4b5381e8 100644
--- a/studio/backend/assets/configs/inference_defaults.json
+++ b/studio/backend/assets/configs/inference_defaults.json
@@ -93,6 +93,14 @@
       "min_p": 0.0,
       "repetition_penalty": 1.0
     },
+    "gemma-4": {
+      "temperature": 1.0,
+      "top_p": 0.95,
+      "top_k": 64,
+      "min_p": 0.0,
+      "repetition_penalty": 1.0,
+      "presence_penalty": 0.0
+    },
     "gemma-3n": {
       "temperature": 1.0,
       "top_p": 0.95,
@@ -366,7 +374,7 @@
     "qwen2.5-coder", "qwen2.5-vl", "qwen2.5-omni", "qwen2.5-math", "qwen2.5",
     "qwen2-vl", "qwen2",
     "qwq",
-    "gemma-3n", "gemma-3", "medgemma", "gemma-2",
+    "gemma-4", "gemma-3n", "gemma-3", "medgemma", "gemma-2",
     "llama-4", "llama-3.3", "llama-3.2", "llama-3.1", "llama-3",
     "phi-4", "phi-3",
     "mistral-nemo", "mistral-small", "mistral-large", "magistral", "ministral",
diff --git a/studio/backend/assets/configs/model_defaults/gemma/unsloth_gemma-4-26B-A4B-it.yaml b/studio/backend/assets/configs/model_defaults/gemma/unsloth_gemma-4-26B-A4B-it.yaml
new file mode 100644
index 0000000000..c80506d9f5
--- /dev/null
+++ b/studio/backend/assets/configs/model_defaults/gemma/unsloth_gemma-4-26B-A4B-it.yaml
@@ -0,0 +1,47 @@
+# Model defaults for unsloth/gemma-4-26B-A4B-it
+# Also applies to: google/gemma-4-26B-A4B-it, unsloth/gemma-4-26B-A4B-it-GGUF
+
+training:
+  trust_remote_code: false
+  max_seq_length: 2048
+  num_epochs: 0
+  learning_rate: 2e-4
+  batch_size: 2
+  gradient_accumulation_steps: 4
+  warmup_steps: 5
+  max_steps: 30
+  save_steps: 30
+  weight_decay: 0.001
+  random_seed: 3407
+  packing: false
+  train_on_completions: true
+  gradient_checkpointing: "unsloth"
+  optim: "adamw_8bit"
+  lr_scheduler_type: "linear"
+
+lora:
+  lora_r: 8
+  lora_alpha: 8
+  lora_dropout: 0.0
+  target_modules:
+    - "all-linear"
+  use_rslora: false
+  use_loftq: false
+  finetune_vision_layers: true
+  finetune_language_layers: true
+  finetune_attention_modules: true
+  finetune_mlp_modules: true
+
+logging:
+  enable_wandb: false
+  wandb_project: "llm-finetuning"
+  enable_tensorboard: false
+  tensorboard_dir: "runs"
+  log_frequency: 10
+
+inference:
+  trust_remote_code: false
+  temperature: 1.0
+  top_p: 0.95
+  top_k: 64
+  min_p: 0.0
diff --git a/studio/backend/assets/configs/model_defaults/gemma/unsloth_gemma-4-26B-A4B.yaml b/studio/backend/assets/configs/model_defaults/gemma/unsloth_gemma-4-26B-A4B.yaml
new file mode 100644
index 0000000000..9e579be503
--- /dev/null
+++ b/studio/backend/assets/configs/model_defaults/gemma/unsloth_gemma-4-26B-A4B.yaml
@@ -0,0 +1,47 @@
+# Model defaults for unsloth/gemma-4-26B-A4B (base/pretrained)
+# Also applies to: google/gemma-4-26B-A4B
+
+training:
+  trust_remote_code: false
+  max_seq_length: 2048
+  num_epochs: 0
+  learning_rate: 2e-4
+  batch_size: 2
+  gradient_accumulation_steps: 4
+  warmup_steps: 5
+  max_steps: 30
+  save_steps: 30
+  weight_decay: 0.001
+  random_seed: 3407
+  packing: false
+  train_on_completions: true
+  gradient_checkpointing: "unsloth"
+  optim: "adamw_8bit"
+  lr_scheduler_type: "linear"
+
+lora:
+  lora_r: 8
+  lora_alpha: 8
+  lora_dropout: 0.0
+  target_modules:
+    - "all-linear"
+  use_rslora: false
+  use_loftq: false
+  finetune_vision_layers: true
+  finetune_language_layers: true
+  finetune_attention_modules: true
+  finetune_mlp_modules: true
+
+logging:
+  enable_wandb: false
+  wandb_project: "llm-finetuning"
+  enable_tensorboard: false
+  tensorboard_dir: "runs"
+  log_frequency: 10
+
+inference:
+  trust_remote_code: false
+  temperature: 1.0
+  top_p: 0.95
+  top_k: 64
+  min_p: 0.0
diff --git a/studio/backend/assets/configs/model_defaults/gemma/unsloth_gemma-4-31B-it.yaml b/studio/backend/assets/configs/model_defaults/gemma/unsloth_gemma-4-31B-it.yaml
new file mode 100644
index 0000000000..cec4ea95e1
--- /dev/null
+++ b/studio/backend/assets/configs/model_defaults/gemma/unsloth_gemma-4-31B-it.yaml
@@ -0,0 +1,47 @@
+# Model defaults for unsloth/gemma-4-31B-it
+# Also applies to: google/gemma-4-31B-it, unsloth/gemma-4-31B-it-GGUF
+
+training:
+  trust_remote_code: false
+  max_seq_length: 2048
+  num_epochs: 0
+  learning_rate: 2e-4
+  batch_size: 2
+  gradient_accumulation_steps: 4
+  warmup_steps: 5
+  max_steps: 30
+  save_steps: 30
+  weight_decay: 0.001
+  random_seed: 3407
+  packing: false
+  train_on_completions: true
+  gradient_checkpointing: "unsloth"
+  optim: "adamw_8bit"
+  lr_scheduler_type: "linear"
+
+lora:
+  lora_r: 8
+  lora_alpha: 8
+  lora_dropout: 0.0
+  target_modules:
+    - "all-linear"
+  use_rslora: false
+  use_loftq: false
+  finetune_vision_layers: true
+  finetune_language_layers: true
+  finetune_attention_modules: true
+  finetune_mlp_modules: true
+
+logging:
+  enable_wandb: false
+  wandb_project: "llm-finetuning"
+  enable_tensorboard: false
+  tensorboard_dir: "runs"
+  log_frequency: 10
+
+inference:
+  trust_remote_code: false
+  temperature: 1.0
+  top_p: 0.95
+  top_k: 64
+  min_p: 0.0
diff --git a/studio/backend/assets/configs/model_defaults/gemma/unsloth_gemma-4-31B.yaml b/studio/backend/assets/configs/model_defaults/gemma/unsloth_gemma-4-31B.yaml
new file mode 100644
index 0000000000..717cdd5e63
--- /dev/null
+++ b/studio/backend/assets/configs/model_defaults/gemma/unsloth_gemma-4-31B.yaml
@@ -0,0 +1,47 @@
+# Model defaults for unsloth/gemma-4-31B (base/pretrained)
+# Also applies to: google/gemma-4-31B
+
+training:
+  trust_remote_code: false
+  max_seq_length: 2048
+  num_epochs: 0
+  learning_rate: 2e-4
+  batch_size: 2
+  gradient_accumulation_steps: 4
+  warmup_steps: 5
+  max_steps: 30
+  save_steps: 30
+  weight_decay: 0.001
+  random_seed: 3407
+  packing: false
+  train_on_completions: true
+  gradient_checkpointing: "unsloth"
+  optim: "adamw_8bit"
+  lr_scheduler_type: "linear"
+
+lora:
+  lora_r: 8
+  lora_alpha: 8
+  lora_dropout: 0.0
+  target_modules:
+    - "all-linear"
+  use_rslora: false
+  use_loftq: false
+  finetune_vision_layers: true
+  finetune_language_layers: true
+  finetune_attention_modules: true
+  finetune_mlp_modules: true
+
+logging:
+  enable_wandb: false
+  wandb_project: "llm-finetuning"
+  enable_tensorboard: false
+  tensorboard_dir: "runs"
+  log_frequency: 10
+
+inference:
+  trust_remote_code: false
+  temperature: 1.0
+  top_p: 0.95
+  top_k: 64
+  min_p: 0.0
diff --git a/studio/backend/assets/configs/model_defaults/gemma/unsloth_gemma-4-E2B-it.yaml b/studio/backend/assets/configs/model_defaults/gemma/unsloth_gemma-4-E2B-it.yaml
new file mode 100644
index 0000000000..43e3d78a23
--- /dev/null
+++ b/studio/backend/assets/configs/model_defaults/gemma/unsloth_gemma-4-E2B-it.yaml
@@ -0,0 +1,47 @@
+# Model defaults for unsloth/gemma-4-E2B-it
+# Also applies to: google/gemma-4-E2B-it, unsloth/gemma-4-E2B-it-GGUF
+
+training:
+  trust_remote_code: false
+  max_seq_length: 2048
+  num_epochs: 0
+  learning_rate: 2e-4
+  batch_size: 2
+  gradient_accumulation_steps: 4
+  warmup_steps: 5
+  max_steps: 30
+  save_steps: 30
+  weight_decay: 0.001
+  random_seed: 3407
+  packing: false
+  train_on_completions: true
+  gradient_checkpointing: "unsloth"
+  optim: "adamw_8bit"
+  lr_scheduler_type: "linear"
+
+lora:
+  lora_r: 8
+  lora_alpha: 8
+  lora_dropout: 0.0
+  target_modules:
+    - "all-linear"
+  use_rslora: false
+  use_loftq: false
+  finetune_vision_layers: true
+  finetune_language_layers: true
+  finetune_attention_modules: true
+  finetune_mlp_modules: true
+
+logging:
+  enable_wandb: false
+  wandb_project: "llm-finetuning"
+  enable_tensorboard: false
+  tensorboard_dir: "runs"
+  log_frequency: 10
+
+inference:
+  trust_remote_code: false
+  temperature: 1.0
+  top_p: 0.95
+  top_k: 64
+  min_p: 0.0
diff --git a/studio/backend/assets/configs/model_defaults/gemma/unsloth_gemma-4-E2B.yaml b/studio/backend/assets/configs/model_defaults/gemma/unsloth_gemma-4-E2B.yaml
new file mode 100644
index 0000000000..bd86cef751
--- /dev/null
+++ b/studio/backend/assets/configs/model_defaults/gemma/unsloth_gemma-4-E2B.yaml
@@ -0,0 +1,47 @@
+# Model defaults for unsloth/gemma-4-E2B (base/pretrained)
+# Also applies to: google/gemma-4-E2B
+
+training:
+  trust_remote_code: false
+  max_seq_length: 2048
+  num_epochs: 0
+  learning_rate: 2e-4
+  batch_size: 2
+  gradient_accumulation_steps: 4
+  warmup_steps: 5
+  max_steps: 30
+  save_steps: 30
+  weight_decay: 0.001
+  random_seed: 3407
+  packing: false
+  train_on_completions: true
+  gradient_checkpointing: "unsloth"
+  optim: "adamw_8bit"
+  lr_scheduler_type: "linear"
+
+lora:
+  lora_r: 8
+  lora_alpha: 8
+  lora_dropout: 0.0
+  target_modules:
+    - "all-linear"
+  use_rslora: false
+  use_loftq: false
+  finetune_vision_layers: true
+  finetune_language_layers: true
+  finetune_attention_modules: true
+  finetune_mlp_modules: true
+
+logging:
+  enable_wandb: false
+  wandb_project: "llm-finetuning"
+  enable_tensorboard: false
+  tensorboard_dir: "runs"
+  log_frequency: 10
+
+inference:
+  trust_remote_code: false
+  temperature: 1.0
+  top_p: 0.95
+  top_k: 64
+  min_p: 0.0
diff --git a/studio/backend/assets/configs/model_defaults/gemma/unsloth_gemma-4-E4B-it.yaml b/studio/backend/assets/configs/model_defaults/gemma/unsloth_gemma-4-E4B-it.yaml
new file mode 100644
index 0000000000..a8ef51836b
--- /dev/null
+++ b/studio/backend/assets/configs/model_defaults/gemma/unsloth_gemma-4-E4B-it.yaml
@@ -0,0 +1,47 @@
+# Model defaults for unsloth/gemma-4-E4B-it
+# Also applies to: google/gemma-4-E4B-it, unsloth/gemma-4-E4B-it-GGUF
+
+training:
+  trust_remote_code: false
+  max_seq_length: 2048
+  num_epochs: 0
+  learning_rate: 2e-4
+  batch_size: 2
+  gradient_accumulation_steps: 4
+  warmup_steps: 5
+  max_steps: 30
+  save_steps: 30
+  weight_decay: 0.001
+  random_seed: 3407
+  packing: false
+  train_on_completions: true
+  gradient_checkpointing: "unsloth"
+  optim: "adamw_8bit"
+  lr_scheduler_type: "linear"
+
+lora:
+  lora_r: 8
+  lora_alpha: 8
+  lora_dropout: 0.0
+  target_modules:
+    - "all-linear"
+  use_rslora: false
+  use_loftq: false
+  finetune_vision_layers: true
+  finetune_language_layers: true
+  finetune_attention_modules: true
+  finetune_mlp_modules: true
+
+logging:
+  enable_wandb: false
+  wandb_project: "llm-finetuning"
+  enable_tensorboard: false
+  tensorboard_dir: "runs"
+  log_frequency: 10
+
+inference:
+  trust_remote_code: false
+  temperature: 1.0
+  top_p: 0.95
+  top_k: 64
+  min_p: 0.0
diff --git a/studio/backend/assets/configs/model_defaults/gemma/unsloth_gemma-4-E4B.yaml b/studio/backend/assets/configs/model_defaults/gemma/unsloth_gemma-4-E4B.yaml
new file mode 100644
index 0000000000..740cc99df5
--- /dev/null
+++ b/studio/backend/assets/configs/model_defaults/gemma/unsloth_gemma-4-E4B.yaml
@@ -0,0 +1,47 @@
+# Model defaults for unsloth/gemma-4-E4B (base/pretrained)
+# Also applies to: google/gemma-4-E4B
+
+training:
+  trust_remote_code: false
+  max_seq_length: 2048
+  num_epochs: 0
+  learning_rate: 2e-4
+  batch_size: 2
+  gradient_accumulation_steps: 4
+  warmup_steps: 5
+  max_steps: 30
+  save_steps: 30
+  weight_decay: 0.001
+  random_seed: 3407
+  packing: false
+  train_on_completions: true
+  gradient_checkpointing: "unsloth"
+  optim: "adamw_8bit"
+  lr_scheduler_type: "linear"
+
+lora:
+  lora_r: 8
+  lora_alpha: 8
+  lora_dropout: 0.0
+  target_modules:
+    - "all-linear"
+  use_rslora: false
+  use_loftq: false
+  finetune_vision_layers: true
+  finetune_language_layers: true
+  finetune_attention_modules: true
+  finetune_mlp_modules: true
+
+logging:
+  enable_wandb: false
+  wandb_project: "llm-finetuning"
+  enable_tensorboard: false
+  tensorboard_dir: "runs"
+  log_frequency: 10
+
+inference:
+  trust_remote_code: false
+  temperature: 1.0
+  top_p: 0.95
+  top_k: 64
+  min_p: 0.0
diff --git a/studio/backend/utils/models/model_config.py b/studio/backend/utils/models/model_config.py
index 8c25da4f43..f97ea993eb 100644
--- a/studio/backend/utils/models/model_config.py
+++ b/studio/backend/utils/models/model_config.py
@@ -159,6 +159,38 @@ MODEL_NAME_MAPPING = {
         "unsloth/gemma-3n-E4B-unsloth-bnb-4bit",
         "google/gemma-3n-E4B",
     ],
+    "unsloth_gemma-4-31B-it.yaml": [
+        "unsloth/gemma-4-31B-it",
+        "google/gemma-4-31B-it",
+    ],
+    "unsloth_gemma-4-26B-A4B-it.yaml": [
+        "unsloth/gemma-4-26B-A4B-it",
+        "google/gemma-4-26B-A4B-it",
+    ],
+    "unsloth_gemma-4-E2B-it.yaml": [
+        "unsloth/gemma-4-E2B-it",
+        "google/gemma-4-E2B-it",
+    ],
+    "unsloth_gemma-4-E4B-it.yaml": [
+        "unsloth/gemma-4-E4B-it",
+        "google/gemma-4-E4B-it",
+    ],
+    "unsloth_gemma-4-31B.yaml": [
+        "unsloth/gemma-4-31B",
+        "google/gemma-4-31B",
+    ],
+    "unsloth_gemma-4-26B-A4B.yaml": [
+        "unsloth/gemma-4-26B-A4B",
+        "google/gemma-4-26B-A4B",
+    ],
+    "unsloth_gemma-4-E2B.yaml": [
+        "unsloth/gemma-4-E2B",
+        "google/gemma-4-E2B",
+    ],
+    "unsloth_gemma-4-E4B.yaml": [
+        "unsloth/gemma-4-E4B",
+        "google/gemma-4-E4B",
+    ],
     "unsloth_gpt-oss-20b.yaml": [
         "openai/gpt-oss-20b",
         "unsloth/gpt-oss-20b-unsloth-bnb-4bit",

From 4020a70a93a466f5ea67ffc3b0390db5d3aac21e Mon Sep 17 00:00:00 2001
From: Daniel Han <danielhanchen@gmail.com>
Date: Fri, 3 Apr 2026 13:58:26 -0700
Subject: [PATCH 041/216] Add tests for cache case resolution (from PR #4822)
 (#4823)

Tests for resolve_cached_repo_id_case and get_model_config case
resolution, separated from the runtime changes in PR #4822.
---
 .../tests/test_cache_case_resolution.py       | 120 ++++++++++++++++++
 ...models_get_model_config_case_resolution.py |  81 ++++++++++++
 2 files changed, 201 insertions(+)
 create mode 100644 studio/backend/tests/test_cache_case_resolution.py
 create mode 100644 studio/backend/tests/test_models_get_model_config_case_resolution.py

diff --git a/studio/backend/tests/test_cache_case_resolution.py b/studio/backend/tests/test_cache_case_resolution.py
new file mode 100644
index 0000000000..60963b4f7d
--- /dev/null
+++ b/studio/backend/tests/test_cache_case_resolution.py
@@ -0,0 +1,120 @@
+# SPDX-License-Identifier: AGPL-3.0-only
+# Copyright 2026-present the Unsloth AI Inc. team. All rights reserved. See /studio/LICENSE.AGPL-3.0
+
+from pathlib import Path
+import sys
+import types
+
+# Keep this test runnable in lightweight environments where optional logging
+# deps are not installed.
+if "structlog" not in sys.modules:
+
+    class _DummyLogger:
+        def __getattr__(self, _name):
+            return lambda *args, **kwargs: None
+
+    sys.modules["structlog"] = types.SimpleNamespace(
+        BoundLogger = _DummyLogger,
+        get_logger = lambda *args, **kwargs: _DummyLogger(),
+    )
+
+from utils.paths.path_utils import (
+    resolve_cached_repo_id_case,
+    get_cache_case_resolution_stats,
+    reset_cache_case_resolution_state,
+)
+import utils.paths.path_utils as path_utils
+
+
+def _mk_cache_repo(cache_root: Path, repo_id: str) -> Path:
+    repo_dir = cache_root / f"models--{repo_id.replace('/', '--')}"
+    repo_dir.mkdir(parents = True, exist_ok = True)
+    return repo_dir
+
+
+def test_resolve_cached_repo_id_case_exact_hit(tmp_path, monkeypatch):
+    reset_cache_case_resolution_state()
+    _mk_cache_repo(tmp_path, "Org/Model")
+    monkeypatch.setattr(path_utils, "_hf_hub_cache_dir", lambda: tmp_path)
+
+    resolved = resolve_cached_repo_id_case("Org/Model")
+
+    assert resolved == "Org/Model"
+    stats = get_cache_case_resolution_stats()
+    assert stats["calls"] == 1
+    assert stats["exact_hits"] == 1
+    assert stats["variant_hits"] == 0
+
+
+def test_resolve_cached_repo_id_case_variant_hit(tmp_path, monkeypatch):
+    reset_cache_case_resolution_state()
+    _mk_cache_repo(tmp_path, "Org/Model")
+    monkeypatch.setattr(path_utils, "_hf_hub_cache_dir", lambda: tmp_path)
+
+    resolved = resolve_cached_repo_id_case("org/model")
+
+    assert resolved == "Org/Model"
+    stats = get_cache_case_resolution_stats()
+    assert stats["variant_hits"] == 1
+    assert stats["tie_breaks"] == 0
+
+
+def test_resolve_cached_repo_id_case_tie_break_deterministic(tmp_path, monkeypatch):
+    reset_cache_case_resolution_state()
+    _mk_cache_repo(tmp_path, "Org/Model")
+    _mk_cache_repo(tmp_path, "org/model")
+    monkeypatch.setattr(path_utils, "_hf_hub_cache_dir", lambda: tmp_path)
+
+    resolved = resolve_cached_repo_id_case("oRg/mOdEl")
+
+    # Deterministic rule: lexical sort of candidate repo ids.
+    assert resolved == "Org/Model"
+    stats = get_cache_case_resolution_stats()
+    assert stats["variant_hits"] == 1
+    assert stats["tie_breaks"] == 1
+
+
+def test_resolve_cached_repo_id_case_no_cache_fallback(tmp_path, monkeypatch):
+    reset_cache_case_resolution_state()
+    monkeypatch.setattr(path_utils, "_hf_hub_cache_dir", lambda: tmp_path)
+
+    resolved = resolve_cached_repo_id_case("Org/Missing")
+
+    assert resolved == "Org/Missing"
+    stats = get_cache_case_resolution_stats()
+    assert stats["fallbacks"] == 1
+    assert stats["variant_hits"] == 0
+    assert stats["exact_hits"] == 0
+
+
+def test_resolve_cached_repo_id_case_memoization(tmp_path, monkeypatch):
+    reset_cache_case_resolution_state()
+    _mk_cache_repo(tmp_path, "Org/Model")
+    monkeypatch.setattr(path_utils, "_hf_hub_cache_dir", lambda: tmp_path)
+
+    first = resolve_cached_repo_id_case("org/model")
+    second = resolve_cached_repo_id_case("org/model")
+
+    assert first == "Org/Model"
+    assert second == "Org/Model"
+    stats = get_cache_case_resolution_stats()
+    assert stats["calls"] == 2
+    assert stats["variant_hits"] == 1
+    assert stats["memo_hits"] == 1
+
+
+def test_resolve_cached_repo_id_case_late_cache_population(tmp_path, monkeypatch):
+    """Regression guard: memoized fallback should not hide a later cache variant."""
+    reset_cache_case_resolution_state()
+    monkeypatch.setattr(path_utils, "_hf_hub_cache_dir", lambda: tmp_path)
+
+    first = resolve_cached_repo_id_case("org/model")
+    assert first == "org/model"
+
+    # Simulate cache being populated after first miss (e.g. another code path/download).
+    _mk_cache_repo(tmp_path, "Org/Model")
+
+    second = resolve_cached_repo_id_case("org/model")
+
+    # Desired behavior: second lookup should pick up the now-existing variant.
+    assert second == "Org/Model"
diff --git a/studio/backend/tests/test_models_get_model_config_case_resolution.py b/studio/backend/tests/test_models_get_model_config_case_resolution.py
new file mode 100644
index 0000000000..3481e29948
--- /dev/null
+++ b/studio/backend/tests/test_models_get_model_config_case_resolution.py
@@ -0,0 +1,81 @@
+# SPDX-License-Identifier: AGPL-3.0-only
+# Copyright 2026-present the Unsloth AI Inc. team. All rights reserved. See /studio/LICENSE.AGPL-3.0
+
+import asyncio
+import sys
+import types
+
+# Keep this test runnable in lightweight environments where optional logging
+# deps are not installed.
+if "structlog" not in sys.modules:
+
+    class _DummyLogger:
+        def __getattr__(self, _name):
+            return lambda *args, **kwargs: None
+
+    sys.modules["structlog"] = types.SimpleNamespace(
+        BoundLogger = _DummyLogger,
+        get_logger = lambda *args, **kwargs: _DummyLogger(),
+    )
+
+import routes.models as models_route
+import utils.models.model_config as model_config_module
+
+
+def test_get_model_config_resolves_cached_case_before_model_checks(monkeypatch):
+    calls: dict[str, str] = {}
+
+    class _DummyModelConfig:
+        is_lora = False
+        base_model = None
+
+    def _record_load(model_name):
+        calls["load_model_defaults"] = model_name
+        return {}
+
+    def _record_vision(model_name, hf_token = None):
+        calls["is_vision_model"] = model_name
+        return False
+
+    def _record_embedding(model_name, hf_token = None):
+        calls["is_embedding_model"] = model_name
+        return False
+
+    def _record_audio(model_name, hf_token = None):
+        calls["detect_audio_type"] = model_name
+        return None
+
+    def _record_from_identifier(cls, model_name):
+        calls["from_identifier"] = model_name
+        return _DummyModelConfig()
+
+    monkeypatch.setattr(models_route, "is_local_path", lambda _: False)
+    monkeypatch.setattr(
+        models_route, "resolve_cached_repo_id_case", lambda _: "Org/Model"
+    )
+    monkeypatch.setattr(models_route, "load_model_defaults", _record_load)
+    monkeypatch.setattr(models_route, "is_vision_model", _record_vision)
+    monkeypatch.setattr(models_route, "is_embedding_model", _record_embedding)
+    monkeypatch.setattr(model_config_module, "detect_audio_type", _record_audio)
+    monkeypatch.setattr(
+        models_route.ModelConfig,
+        "from_identifier",
+        classmethod(_record_from_identifier),
+    )
+    monkeypatch.setattr(models_route, "_get_max_position_embeddings", lambda _: 4096)
+    monkeypatch.setattr(models_route, "_get_model_size_bytes", lambda *_args, **_kw: 0)
+
+    result = asyncio.run(
+        models_route.get_model_config(
+            model_name = "org/model",
+            hf_token = None,
+            current_subject = "test-subject",
+        )
+    )
+
+    assert result.model_name == "Org/Model"
+    assert calls["load_model_defaults"] == "Org/Model"
+    assert calls["is_vision_model"] == "Org/Model"
+    assert calls["is_embedding_model"] == "Org/Model"
+    assert calls["detect_audio_type"] == "Org/Model"
+    assert calls["from_identifier"] == "Org/Model"

From 170c4b9b99c6f7bee2fe8cf3a965fddc3eea0fd9 Mon Sep 17 00:00:00 2001
From: Daniel Han <danielhanchen@gmail.com>
Date: Fri, 3 Apr 2026 15:02:14 -0700
Subject: [PATCH 042/216] Update _utils.py

---
 unsloth/models/_utils.py | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/unsloth/models/_utils.py b/unsloth/models/_utils.py
index 28526056ba..dcb7334417 100644
--- a/unsloth/models/_utils.py
+++ b/unsloth/models/_utils.py
@@ -12,7 +12,7 @@
 # See the License for the specific language governing permissions and
 # limitations under the License.
 
-__version__ = "2026.4.1"
+__version__ = "2026.4.2"
 
 __all__ = [
     "SUPPORTS_BFLOAT16",

From 610086744766a052cb799b17538b1238c7931659 Mon Sep 17 00:00:00 2001
From: Daniel Han <danielhanchen@gmail.com>
Date: Fri, 3 Apr 2026 15:14:28 -0700
Subject: [PATCH 043/216] Bump minimum unsloth version to 2026.4.2 in install
 scripts (#4842)

---
 install.ps1 | 10 +++++-----
 install.sh  | 10 +++++-----
 2 files changed, 10 insertions(+), 10 deletions(-)

diff --git a/install.ps1 b/install.ps1
index ead4e7368d..5ddb42ea7e 100644
--- a/install.ps1
+++ b/install.ps1
@@ -819,7 +819,7 @@ shell.Run cmd, 0, False
         if ($SkipTorch) {
             # No-torch: install unsloth + unsloth-zoo with --no-deps, then
             # runtime deps (typer, safetensors, transformers, etc.) with --no-deps.
-            $baseInstallExit = Invoke-InstallCommand { uv pip install --python $VenvPython --no-deps --reinstall-package unsloth --reinstall-package unsloth-zoo "unsloth>=2026.3.18" unsloth-zoo }
+            $baseInstallExit = Invoke-InstallCommand { uv pip install --python $VenvPython --no-deps --reinstall-package unsloth --reinstall-package unsloth-zoo "unsloth>=2026.4.2" unsloth-zoo }
             if ($baseInstallExit -eq 0) {
                 $NoTorchReq = Find-NoTorchRuntimeFile
                 if ($NoTorchReq) {
@@ -827,7 +827,7 @@ shell.Run cmd, 0, False
                 }
             }
         } else {
-            $baseInstallExit = Invoke-InstallCommand { uv pip install --python $VenvPython --reinstall-package unsloth --reinstall-package unsloth-zoo "unsloth>=2026.3.18" unsloth-zoo }
+            $baseInstallExit = Invoke-InstallCommand { uv pip install --python $VenvPython --reinstall-package unsloth --reinstall-package unsloth-zoo "unsloth>=2026.4.2" unsloth-zoo }
         }
         if ($baseInstallExit -ne 0) {
             Write-Host "[ERROR] Failed to install unsloth (exit code $baseInstallExit)" -ForegroundColor Red
@@ -857,7 +857,7 @@ shell.Run cmd, 0, False
         if ($SkipTorch) {
             # No-torch: install unsloth + unsloth-zoo with --no-deps, then
             # runtime deps (typer, safetensors, transformers, etc.) with --no-deps.
-            $baseInstallExit = Invoke-InstallCommand { uv pip install --python $VenvPython --no-deps --upgrade-package unsloth --upgrade-package unsloth-zoo "unsloth>=2026.3.18" unsloth-zoo }
+            $baseInstallExit = Invoke-InstallCommand { uv pip install --python $VenvPython --no-deps --upgrade-package unsloth --upgrade-package unsloth-zoo "unsloth>=2026.4.2" unsloth-zoo }
             if ($baseInstallExit -eq 0) {
                 $NoTorchReq = Find-NoTorchRuntimeFile
                 if ($NoTorchReq) {
@@ -865,7 +865,7 @@ shell.Run cmd, 0, False
                 }
             }
         } elseif ($StudioLocalInstall) {
-            $baseInstallExit = Invoke-InstallCommand { uv pip install --python $VenvPython --upgrade-package unsloth "unsloth>=2026.3.18" unsloth-zoo }
+            $baseInstallExit = Invoke-InstallCommand { uv pip install --python $VenvPython --upgrade-package unsloth "unsloth>=2026.4.2" unsloth-zoo }
         } else {
             $baseInstallExit = Invoke-InstallCommand { uv pip install --python $VenvPython --upgrade-package unsloth "$PackageName" }
         }
@@ -886,7 +886,7 @@ shell.Run cmd, 0, False
         # Fallback: GPU detection failed to produce a URL -- let uv resolve torch
         substep "installing unsloth (this may take a few minutes)..."
         if ($StudioLocalInstall) {
-            $baseInstallExit = Invoke-InstallCommand { uv pip install --python $VenvPython unsloth-zoo "unsloth>=2026.3.18" --torch-backend=auto }
+            $baseInstallExit = Invoke-InstallCommand { uv pip install --python $VenvPython unsloth-zoo "unsloth>=2026.4.2" --torch-backend=auto }
             if ($baseInstallExit -ne 0) {
                 Write-Host "[ERROR] Failed to install unsloth (exit code $baseInstallExit)" -ForegroundColor Red
                 return
diff --git a/install.sh b/install.sh
index 2d9a368555..053f334d2b 100755
--- a/install.sh
+++ b/install.sh
@@ -1040,7 +1040,7 @@ if [ "$_MIGRATED" = true ]; then
         # to prevent transitive torch resolution.
         run_install_cmd "install unsloth (migrated no-torch)" uv pip install --python "$_VENV_PY" --no-deps \
             --reinstall-package unsloth --reinstall-package unsloth-zoo \
-            "unsloth>=2026.3.18" unsloth-zoo
+            "unsloth>=2026.4.2" unsloth-zoo
         _NO_TORCH_RT="$(_find_no_torch_runtime)"
         if [ -n "$_NO_TORCH_RT" ]; then
             run_install_cmd "install no-torch runtime deps" uv pip install --python "$_VENV_PY" --no-deps -r "$_NO_TORCH_RT"
@@ -1048,7 +1048,7 @@ if [ "$_MIGRATED" = true ]; then
     else
         run_install_cmd "install unsloth (migrated)" uv pip install --python "$_VENV_PY" \
             --reinstall-package unsloth --reinstall-package unsloth-zoo \
-            "unsloth>=2026.3.18" unsloth-zoo
+            "unsloth>=2026.4.2" unsloth-zoo
     fi
     if [ "$STUDIO_LOCAL_INSTALL" = true ]; then
         substep "overlaying local repo (editable)..."
@@ -1070,7 +1070,7 @@ elif [ -n "$TORCH_INDEX_URL" ]; then
         # runtime deps (typer, safetensors, transformers, etc.) with --no-deps.
         run_install_cmd "install unsloth (no-torch)" uv pip install --python "$_VENV_PY" --no-deps \
             --upgrade-package unsloth --upgrade-package unsloth-zoo \
-            "unsloth>=2026.3.18" unsloth-zoo
+            "unsloth>=2026.4.2" unsloth-zoo
         _NO_TORCH_RT="$(_find_no_torch_runtime)"
         if [ -n "$_NO_TORCH_RT" ]; then
             run_install_cmd "install no-torch runtime deps" uv pip install --python "$_VENV_PY" --no-deps -r "$_NO_TORCH_RT"
@@ -1081,7 +1081,7 @@ elif [ -n "$TORCH_INDEX_URL" ]; then
         fi
     elif [ "$STUDIO_LOCAL_INSTALL" = true ]; then
         run_install_cmd "install unsloth (local)" uv pip install --python "$_VENV_PY" \
-            --upgrade-package unsloth "unsloth>=2026.3.18" unsloth-zoo
+            --upgrade-package unsloth "unsloth>=2026.4.2" unsloth-zoo
         substep "overlaying local repo (editable)..."
         run_install_cmd "overlay local repo" uv pip install --python "$_VENV_PY" -e "$_REPO_ROOT" --no-deps
     else
@@ -1092,7 +1092,7 @@ else
     # Fallback: GPU detection failed to produce a URL -- let uv resolve torch
     substep "installing unsloth (this may take a few minutes)..."
     if [ "$STUDIO_LOCAL_INSTALL" = true ]; then
-        run_install_cmd "install unsloth (auto torch backend)" uv pip install --python "$_VENV_PY" unsloth-zoo "unsloth>=2026.3.18" --torch-backend=auto
+        run_install_cmd "install unsloth (auto torch backend)" uv pip install --python "$_VENV_PY" unsloth-zoo "unsloth>=2026.4.2" --torch-backend=auto
         substep "overlaying local repo (editable)..."
         run_install_cmd "overlay local repo" uv pip install --python "$_VENV_PY" -e "$_REPO_ROOT" --no-deps
     else

From 68965988cfb626f8de5a8e336484ea3abcec2d00 Mon Sep 17 00:00:00 2001
From: Leo Borcherding <borchborchmail@gmail.com>
Date: Sun, 5 Apr 2026 23:57:45 -0500
Subject: [PATCH 044/216] Fix/studio colab button message: Add fallback message
 for Colab Studio button when proxy URL fails (#4866)

* Add fallback message for Colab Studio button when localhost link doesn't work

* Make fallback message darker grey for better readability

* Make fallback message bold for better visibility

---------

Co-authored-by: LeoBorcherding <LeoBorcherding@users.noreply.github.com>
---
 studio/backend/colab.py | 5 ++++-
 1 file changed, 4 insertions(+), 1 deletion(-)

diff --git a/studio/backend/colab.py b/studio/backend/colab.py
index efd0e10bdb..7336f8a532 100644
--- a/studio/backend/colab.py
+++ b/studio/backend/colab.py
@@ -66,7 +66,10 @@ def show_link(port: int = 8888):
             <svg xmlns="http://www.w3.org/2000/svg" width="18" height="18" viewBox="0 0 24 24" fill="white"><polygon points="5,3 19,12 5,21"/></svg>
             Open Unsloth Studio
         </a>
-        <p style="color: #333333; margin: 16px 0 0 0; font-size: 13px; font-family: monospace;">
+        <p style="color: #333333; margin: 12px 0 0 0; font-size: 14px; font-weight: bold;">
+            If the link doesn't work, you can scroll down to view the UI generated directly in Colab.
+        </p>
+        <p style="color: #333333; margin: 16px 0 0 0; font-size: 13px; font-family: monospace; font-weight: bold;">
             {short_url}
         </p>
     </div>

From 278f4629962358268674de0a102f7537c2b0ab3a Mon Sep 17 00:00:00 2001
From: Roland Tannous <115670425+rolandtannous@users.noreply.github.com>
Date: Mon, 6 Apr 2026 17:41:20 +0400
Subject: [PATCH 045/216] [Studio][Optimization]Add vision detection cache to
 is_vision_model() (#4853)

* Add vision detection cache to is_vision_model() to avoid redundant subprocess spawns

is_vision_model() is called 4-5 times per training run for the same model
with zero caching. For transformers 5.x models, each call spawns a full
subprocess (~6s each). This adds a module-level _vision_detection_cache dict
following the same pattern as the existing _audio_detection_cache used by
detect_audio_type(). The function is refactored into a thin cache wrapper
around _is_vision_model_uncached(), saving ~12s per training run.

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

* Include hf_token in vision cache key for gated model correctness

Cache key is now (model_name, hf_token) instead of just model_name.
This prevents stale False results when an unauthenticated probe for a
gated model is followed by an authenticated call.

* Remove test file from main PR - will be submitted separately

* Fix vision cache: normalize model names and skip caching transient failures

- Normalize model names in cache key using resolve_cached_repo_id_case()
  to avoid duplicate entries for different casings of the same HF repo
  (aligns with case normalization from #4822)
- Return None instead of False on transient failures (network errors,
  subprocess timeouts, HF API issues) so the cache layer can distinguish
  "definitely not a vision model" from "failed to check"
- Only cache definitive True/False results; transient failures are retried
  on the next call instead of being permanently locked in as False

* Refine failure handling: cache deterministic failures, guard normalization

- Subprocess non-zero exit, JSON errors, and general exceptions return
  False (deterministic, cached) instead of None (retryable). Only
  subprocess.TimeoutExpired returns None since timeouts are transient.
- Wrap cache key normalization in try/except so resolve_cached_repo_id_case
  or normalize_path failures fall back to raw model_name instead of
  crashing callers.

* Harden vision detection cache: fix transient failure handling, thread safety, token security

- All subprocess failure paths now return None (transient) instead of False,
  preventing permanent misclassification of VLMs after temporary HF/auth/network errors
- Use SHA256 fingerprint for hf_token in cache key instead of raw bearer token
- Add threading.Lock with double-checked locking to prevent thundering herd
  of concurrent subprocess spawns for the same uncached model
- Distinguish permanent failures (RepositoryNotFoundError, GatedRepoError,
  ValueError) from transient ones in _is_vision_model_uncached
- Pass resolved/normalized model name to detection (not just cache key)
- Log normalization fallback at debug level instead of silent swallow
- Thread hf_token through callers in routes/models.py and trainer.py
  that previously omitted it

* Refine lock strategy and token fingerprint

- Move detection computation outside the lock to avoid serializing
  long-running subprocess spawns (60s timeout) and HF API calls across
  all concurrent model checks. Lock is now only held for cache writes.
- Use full SHA256 digest for token fingerprint instead of truncated
  16-char prefix to eliminate collision risk.

* Fix huggingface_hub import fallback and use atomic cache read

- Add fallback import path for RepositoryNotFoundError/GatedRepoError
  from huggingface_hub.utils (older hub versions) when .errors is
  not available
- Use sentinel-based dict.get() for single atomic cache read instead
  of two-step in/[] pattern (future-proof for no-GIL runtimes)

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

---------

Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com>
Co-authored-by: Daniel Han <danielhanchen@gmail.com>
---
 studio/backend/core/training/trainer.py     |  12 ++-
 studio/backend/routes/models.py             |   2 +-
 studio/backend/utils/models/model_config.py | 112 ++++++++++++++++++--
 3 files changed, 115 insertions(+), 11 deletions(-)

diff --git a/studio/backend/core/training/trainer.py b/studio/backend/core/training/trainer.py
index ab1825d94a..77cbda6b45 100644
--- a/studio/backend/core/training/trainer.py
+++ b/studio/backend/core/training/trainer.py
@@ -190,7 +190,11 @@ class UnslothTrainer:
             self._cuda_audio_used = False
 
         # --- Detect VLM ---
-        vision = is_vision_model(model_name) if not self.is_audio else False
+        vision = (
+            is_vision_model(model_name, hf_token = hf_token)
+            if not self.is_audio
+            else False
+        )
         self.is_vlm = not self.is_audio_vlm and vision and is_dataset_image
 
         logger.info(
@@ -558,7 +562,11 @@ class UnslothTrainer:
                 self._cuda_audio_used = False
 
             # VLM: vision model with image dataset (mutually exclusive with audio paths)
-            vision = is_vision_model(model_name) if not self.is_audio else False
+            vision = (
+                is_vision_model(model_name, hf_token = hf_token)
+                if not self.is_audio
+                else False
+            )
             self.is_vlm = not self.is_audio_vlm and vision and is_dataset_image
             self.model_name = model_name
             self.max_seq_length = max_seq_length
diff --git a/studio/backend/routes/models.py b/studio/backend/routes/models.py
index 1e31a91e26..ea385436b8 100644
--- a/studio/backend/routes/models.py
+++ b/studio/backend/routes/models.py
@@ -618,7 +618,7 @@ async def get_model_config(
         config_dict = load_model_defaults(model_name)
 
         # Detect model capabilities (pass HF token for gated models)
-        is_vision = is_vision_model(model_name)
+        is_vision = is_vision_model(model_name, hf_token = hf_token)
         is_embedding = is_embedding_model(model_name, hf_token = hf_token)
         audio_type = detect_audio_type(model_name, hf_token = hf_token)
 
diff --git a/studio/backend/utils/models/model_config.py b/studio/backend/utils/models/model_config.py
index f97ea993eb..9d72846ca3 100644
--- a/studio/backend/utils/models/model_config.py
+++ b/studio/backend/utils/models/model_config.py
@@ -26,7 +26,9 @@ import subprocess
 import sys
 from pathlib import Path
 from typing import List, Tuple
+import hashlib
 import json
+import threading
 import yaml
 
 
@@ -548,12 +550,17 @@ except Exception as exc:
 
 def _is_vision_model_subprocess(
     model_name: str, hf_token: Optional[str] = None
-) -> bool:
+) -> Optional[bool]:
     """Run is_vision_model check in a subprocess with transformers 5.x.
 
     Same pattern as training/inference workers: spawn a clean subprocess
     with .venv_t5/ prepended to sys.path so AutoConfig recognizes newer
     architectures (glm4_moe_lite, etc.).
+
+    Returns True/False for definitive results, or None for transient failures
+    (timeouts, subprocess errors) so callers can decide whether to cache
+    the result. Subprocess failures are treated as transient because they
+    can be caused by temporary HF/auth/network issues.
     """
     token_arg = hf_token or ""
 
@@ -580,7 +587,7 @@ def _is_vision_model_subprocess(
                 model_name,
                 stderr or result.stdout.strip(),
             )
-            return False
+            return None
 
         data = json.loads(result.stdout.strip())
         if "error" in data:
@@ -589,7 +596,7 @@ def _is_vision_model_subprocess(
                 model_name,
                 data["error"],
             )
-            return False
+            return None
 
         is_vlm = data["is_vision"]
         logger.info(
@@ -604,10 +611,28 @@ def _is_vision_model_subprocess(
 
     except subprocess.TimeoutExpired:
         logger.warning("Vision check subprocess timed out for '%s'", model_name)
-        return False
+        return None
     except Exception as exc:
         logger.warning("Vision check subprocess failed for '%s': %s", model_name, exc)
-        return False
+        return None
+
+
+def _token_fingerprint(token: Optional[str]) -> Optional[str]:
+    """Return a SHA256 digest of the token for use as a cache key.
+
+    Avoids storing the raw bearer token in process memory as a dict key.
+    """
+    if token is None:
+        return None
+    return hashlib.sha256(token.encode("utf-8")).hexdigest()
+
+
+# Cache vision detection results per session to avoid repeated subprocess spawns.
+# Keyed by (normalized_model_name, token_fingerprint) to handle gated models correctly.
+# Only definitive results (True/False from successful detection) are cached;
+# transient failures (network errors, timeouts) are NOT cached so they can be retried.
+_vision_detection_cache: Dict[Tuple[str, Optional[str]], bool] = {}
+_vision_cache_lock = threading.Lock()
 
 
 def is_vision_model(model_name: str, hf_token: Optional[str] = None) -> bool:
@@ -616,13 +641,66 @@ def is_vision_model(model_name: str, hf_token: Optional[str] = None) -> bool:
     Works for fine-tuned models since they inherit the base architecture.
 
     For models that require transformers 5.x (e.g. GLM-4.7-Flash), the check
-    runs in a subprocess with .venv_t5/ activated — same pattern as the
+    runs in a subprocess with .venv_t5/ activated -- same pattern as the
     training and inference workers.
 
+    Results are cached per (model_name, token_fingerprint) for the lifetime of
+    the process to avoid repeated subprocess spawns and HuggingFace API calls.
+    Transient failures are not cached so they can be retried on the next call.
+
     Args:
         model_name: Model identifier (HF repo or local path)
         hf_token: Optional HF token for accessing gated/private models
     """
+    # Normalize model name for cache key to avoid duplicate entries for
+    # different casings of the same HF repo (e.g. "Org/Model" vs "org/model").
+    try:
+        if is_local_path(model_name):
+            resolved_name = normalize_path(model_name)
+        else:
+            resolved_name = resolve_cached_repo_id_case(model_name)
+    except Exception as exc:
+        logger.debug(
+            "Could not normalize model name '%s' for cache key: %s",
+            model_name,
+            exc,
+        )
+        resolved_name = model_name
+    cache_key = (resolved_name, _token_fingerprint(hf_token))
+
+    # Lock-free fast path for cache hits. Uses a sentinel to distinguish
+    # "key not found" from "value is False" in a single atomic dict.get() call.
+    _MISS = object()
+    cached = _vision_detection_cache.get(cache_key, _MISS)
+    if cached is not _MISS:
+        return cached
+
+    # Compute outside the lock to avoid serializing long-running detection
+    # (subprocess spawns with 60s timeout, HF API calls) across all models.
+    # The tradeoff: two concurrent calls for the same uncached model may
+    # both run detection, but they produce the same result and the second
+    # write is a benign no-op.
+    result = _is_vision_model_uncached(resolved_name, hf_token)
+    # Only cache definitive results; None means a transient failure occurred
+    # and we should retry on the next call instead of locking in a wrong answer.
+    if result is not None:
+        with _vision_cache_lock:
+            _vision_detection_cache[cache_key] = result
+        return result
+    return False
+
+
+def _is_vision_model_uncached(
+    model_name: str, hf_token: Optional[str] = None
+) -> Optional[bool]:
+    """Uncached vision model detection -- called by is_vision_model().
+
+    Returns True/False for definitive results, or None when detection failed
+    due to a transient error (network, timeout, subprocess failure) so the
+    caller knows not to cache the result.
+
+    Do not call directly; use is_vision_model() instead.
+    """
     # Models that need transformers 5.x must be checked in a subprocess
     # because AutoConfig in the main process (transformers 4.57.x) doesn't
     # recognize their architectures.
@@ -630,7 +708,7 @@ def is_vision_model(model_name: str, hf_token: Optional[str] = None) -> bool:
 
     if needs_transformers_5(model_name):
         logger.info(
-            "Model '%s' needs transformers 5.x — checking vision via subprocess",
+            "Model '%s' needs transformers 5.x -- checking vision via subprocess",
             model_name,
         )
         return _is_vision_model_subprocess(model_name, hf_token = hf_token)
@@ -681,7 +759,25 @@ def is_vision_model(model_name: str, hf_token: Optional[str] = None) -> bool:
 
     except Exception as e:
         logger.warning(f"Could not determine if {model_name} is vision model: {e}")
-        return False
+        # Permanent failures (model not found, gated, bad config) should be
+        # cached as False. Transient failures (network, timeout) should not.
+        try:
+            from huggingface_hub.errors import RepositoryNotFoundError, GatedRepoError
+        except ImportError:
+            try:
+                from huggingface_hub.utils import (
+                    RepositoryNotFoundError,
+                    GatedRepoError,
+                )
+            except ImportError:
+                RepositoryNotFoundError = GatedRepoError = None
+        if RepositoryNotFoundError is not None and isinstance(
+            e, (RepositoryNotFoundError, GatedRepoError)
+        ):
+            return False
+        if isinstance(e, (ValueError, json.JSONDecodeError)):
+            return False
+        return None
 
 
 VALID_AUDIO_TYPES = ("snac", "csm", "bicodec", "dac", "whisper", "audio_vlm")

From ab65b47c7312b0ac18e78a708180d08c75fed423 Mon Sep 17 00:00:00 2001
From: Daniel Han <danielhanchen@gmail.com>
Date: Mon, 6 Apr 2026 06:41:40 -0700
Subject: [PATCH 046/216] Add tests for is_vision_model() caching behaviour
 (#4855)

* Add tests for is_vision_model() caching behaviour

* Fix review feedback: remove dead helper, fix exception test

- Remove unused _make_config() helper function (dead code)
- Fix test_exception_result_cached to actually exercise the exception path
  by mocking load_model_config to raise OSError instead of using
  side_effect=[False] which only tested normal False returns

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

* Use strict mock specs so tests exercise intended detection paths

Use MagicMock(spec=[]) for all config mocks so hasattr() only returns
True for explicitly set attributes. Without this, MagicMock defaults
make all hasattr checks truthy, allowing tests to pass via unintended
detection paths (e.g. img_processor instead of vision_config).

---------

Co-authored-by: Roland Tannous <rolandtannous@gravityq.ai>
Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com>
---
 studio/backend/tests/test_vision_cache.py | 238 ++++++++++++++++++++++
 1 file changed, 238 insertions(+)
 create mode 100644 studio/backend/tests/test_vision_cache.py

diff --git a/studio/backend/tests/test_vision_cache.py b/studio/backend/tests/test_vision_cache.py
new file mode 100644
index 0000000000..fae1e95311
--- /dev/null
+++ b/studio/backend/tests/test_vision_cache.py
@@ -0,0 +1,238 @@
+# SPDX-License-Identifier: AGPL-3.0-only
+# Copyright 2026-present the Unsloth AI Inc. team. All rights reserved. See /studio/LICENSE.AGPL-3.0
+
+"""Tests for is_vision_model() caching behaviour.
+
+The vision detection cache (``_vision_detection_cache``) mirrors the existing
+``_audio_detection_cache`` pattern used by ``detect_audio_type()``.  These
+tests verify that:
+
+* Repeated calls for the same model hit the cache (no redundant work).
+* Different models each trigger their own detection.
+* Both True and False results are cached.
+* The subprocess path (transformers 5.x models) is also cached.
+* Exceptions that fall back to False are cached.
+"""
+
+import sys
+import types as _types
+from pathlib import Path
+from unittest.mock import patch, MagicMock
+
+import pytest
+
+# ---------------------------------------------------------------------------
+# sys.path + logger stub — same pattern as the rest of the test suite
+# ---------------------------------------------------------------------------
+_BACKEND_DIR = str(Path(__file__).resolve().parent.parent)
+if _BACKEND_DIR not in sys.path:
+    sys.path.insert(0, _BACKEND_DIR)
+
+_loggers_stub = _types.ModuleType("loggers")
+_loggers_stub.get_logger = lambda name: __import__("logging").getLogger(name)
+sys.modules.setdefault("loggers", _loggers_stub)
+
+from utils.models.model_config import (
+    is_vision_model,
+    _is_vision_model_uncached,
+    _vision_detection_cache,
+)
+
+
+# ---------------------------------------------------------------------------
+# Helpers
+# ---------------------------------------------------------------------------
+
+
+@pytest.fixture(autouse = True)
+def _clear_vision_cache():
+    """Ensure every test starts with a fresh cache."""
+    _vision_detection_cache.clear()
+    yield
+    _vision_detection_cache.clear()
+
+
+# ---------------------------------------------------------------------------
+# Cache hit / miss tests
+# ---------------------------------------------------------------------------
+
+
+class TestVisionCacheHitMiss:
+    """Verify the cache prevents redundant detection calls."""
+
+    @patch("utils.models.model_config._is_vision_model_uncached", return_value = True)
+    def test_second_call_uses_cache(self, mock_uncached):
+        """Calling is_vision_model() twice for the same model should invoke
+        the uncached function only once."""
+        assert is_vision_model("org/my-vlm") is True
+        assert is_vision_model("org/my-vlm") is True
+        mock_uncached.assert_called_once_with("org/my-vlm", None)
+
+    @patch("utils.models.model_config._is_vision_model_uncached", return_value = False)
+    def test_different_models_each_detected(self, mock_uncached):
+        """Different model names should each trigger detection."""
+        is_vision_model("model-a")
+        is_vision_model("model-b")
+        assert mock_uncached.call_count == 2
+
+    @patch("utils.models.model_config._is_vision_model_uncached", return_value = True)
+    def test_cache_returns_correct_value(self, mock_uncached):
+        """The cached value must match what _is_vision_model_uncached returned."""
+        first = is_vision_model("org/vlm")
+        second = is_vision_model("org/vlm")
+        assert first is True
+        assert second is True
+
+
+class TestVisionCacheStoresFalse:
+    """Non-VLM results (False) must also be cached to avoid re-detection."""
+
+    @patch("utils.models.model_config._is_vision_model_uncached", return_value = False)
+    def test_false_result_cached(self, mock_uncached):
+        assert is_vision_model("org/text-only") is False
+        assert is_vision_model("org/text-only") is False
+        mock_uncached.assert_called_once()
+        assert _vision_detection_cache[("org/text-only", None)] is False
+
+
+# ---------------------------------------------------------------------------
+# Subprocess path (transformers 5.x) caching
+# ---------------------------------------------------------------------------
+
+
+class TestVisionCacheSubprocessPath:
+    """Models needing transformers 5.x go through _is_vision_model_subprocess.
+    The cache should prevent the subprocess from being spawned more than once
+    per model per process."""
+
+    @patch("utils.models.model_config._is_vision_model_subprocess", return_value = True)
+    @patch("utils.transformers_version.needs_transformers_5", return_value = True)
+    def test_subprocess_called_once_with_cache(self, mock_needs_t5, mock_subprocess):
+        """Subprocess should only fire on the first call; second is cached."""
+        # First call: goes through uncached → subprocess
+        assert is_vision_model("unsloth/Qwen3.5-2B") is True
+        # Second call: cache hit, no subprocess
+        assert is_vision_model("unsloth/Qwen3.5-2B") is True
+
+        mock_subprocess.assert_called_once()
+        assert _vision_detection_cache[("unsloth/Qwen3.5-2B", None)] is True
+
+
+# ---------------------------------------------------------------------------
+# Exception handling — cache the False fallback
+# ---------------------------------------------------------------------------
+
+
+class TestVisionCacheOnException:
+    """When detection raises an exception, _is_vision_model_uncached catches
+    it and returns False.  That False must be cached so subsequent calls don't
+    retry and fail again."""
+
+    @patch(
+        "utils.models.model_config.load_model_config",
+        side_effect = OSError("network down"),
+    )
+    @patch("utils.transformers_version.needs_transformers_5", return_value = False)
+    def test_exception_result_cached(self, mock_needs_t5, mock_load_config):
+        """A real exception inside _is_vision_model_uncached should be caught,
+        return False, and that False should be cached for subsequent calls."""
+        # First call: load_model_config raises → except branch → False
+        assert is_vision_model("broken/model") is False
+        # Second call: cache hit, load_model_config not called again
+        assert is_vision_model("broken/model") is False
+        mock_load_config.assert_called_once()
+
+
+# ---------------------------------------------------------------------------
+# Direct detection path (non-transformers-5 models) caching
+# ---------------------------------------------------------------------------
+
+
+class TestVisionCacheDirectPath:
+    """For models that do NOT need transformers 5.x, the detection goes through
+    load_model_config directly. The cache must work the same way."""
+
+    @patch("utils.transformers_version.needs_transformers_5", return_value = False)
+    @patch("utils.models.model_config.load_model_config")
+    def test_direct_vlm_detection_cached(self, mock_load_config, mock_needs_t5):
+        """A standard VLM detected via architecture suffix should be cached."""
+        cfg = MagicMock(spec = [])  # strict: only explicitly set attrs exist
+        cfg.model_type = "gemma3"
+        cfg.architectures = ["Gemma3ForConditionalGeneration"]
+        mock_load_config.return_value = cfg
+
+        assert is_vision_model("google/gemma-3-4b-it") is True
+        assert is_vision_model("google/gemma-3-4b-it") is True
+        # load_model_config should only be called once
+        mock_load_config.assert_called_once()
+
+    @patch("utils.transformers_version.needs_transformers_5", return_value = False)
+    @patch("utils.models.model_config.load_model_config")
+    def test_direct_non_vlm_detection_cached(self, mock_load_config, mock_needs_t5):
+        """A standard text model (no VLM indicators) should cache False."""
+        cfg = MagicMock(spec = [])  # spec=[] means no attributes at all
+        cfg.model_type = "llama"
+        cfg.architectures = ["LlamaForCausalLM"]
+        mock_load_config.return_value = cfg
+
+        # LlamaForCausalLM doesn't end with VLM suffixes, no vision_config, etc.
+        assert is_vision_model("meta-llama/Llama-3-8B") is False
+        assert is_vision_model("meta-llama/Llama-3-8B") is False
+        mock_load_config.assert_called_once()
+
+    @patch("utils.transformers_version.needs_transformers_5", return_value = False)
+    @patch("utils.models.model_config.load_model_config")
+    def test_vision_config_attr_detected_and_cached(
+        self, mock_load_config, mock_needs_t5
+    ):
+        """Models with vision_config (LLaVA, Qwen2-VL, etc.) should be cached as True."""
+        cfg = MagicMock(spec = [])  # strict: only explicitly set attrs exist
+        cfg.model_type = "qwen2_vl"
+        cfg.architectures = ["Qwen2VLForCausalLM"]  # Doesn't match VLM suffixes
+        cfg.vision_config = {"hidden_size": 1024}
+        mock_load_config.return_value = cfg
+
+        assert is_vision_model("Qwen/Qwen2-VL-7B") is True
+        assert is_vision_model("Qwen/Qwen2-VL-7B") is True
+        mock_load_config.assert_called_once()
+
+    @patch("utils.transformers_version.needs_transformers_5", return_value = False)
+    @patch("utils.models.model_config.load_model_config")
+    def test_audio_model_excluded_and_cached(self, mock_load_config, mock_needs_t5):
+        """Audio-only models (csm, whisper) with ForConditionalGeneration
+        should be excluded from VLM detection and cached as False."""
+        cfg = MagicMock(spec = [])  # strict: only explicitly set attrs exist
+        cfg.model_type = "whisper"
+        cfg.architectures = ["WhisperForConditionalGeneration"]
+        mock_load_config.return_value = cfg
+
+        assert is_vision_model("openai/whisper-large-v3") is False
+        assert is_vision_model("openai/whisper-large-v3") is False
+        mock_load_config.assert_called_once()
+
+
+# ---------------------------------------------------------------------------
+# hf_token handling
+# ---------------------------------------------------------------------------
+
+
+class TestVisionCacheTokenHandling:
+    """The cache is keyed on (model_name, hf_token).
+    Different tokens for the same model should trigger separate detections
+    to handle gated models correctly."""
+
+    @patch("utils.models.model_config._is_vision_model_uncached", return_value = True)
+    def test_different_tokens_trigger_new_detection(self, mock_uncached):
+        """Calls with different tokens should trigger separate detections to
+        handle gated models correctly (e.g. unauthenticated probe → False,
+        then authenticated call should re-check)."""
+        assert is_vision_model("gated/model", hf_token = "token-a") is True
+        assert is_vision_model("gated/model", hf_token = "token-b") is True
+        assert mock_uncached.call_count == 2
+
+    @patch("utils.models.model_config._is_vision_model_uncached", return_value = True)
+    def test_same_token_uses_cache(self, mock_uncached):
+        """Repeated calls with identical model + token should hit cache."""
+        assert is_vision_model("gated/model", hf_token = "token-a") is True
+        assert is_vision_model("gated/model", hf_token = "token-a") is True
+        mock_uncached.assert_called_once()

From 07b6fcc344adff761f2130ac8434bac594a13c68 Mon Sep 17 00:00:00 2001
From: Daniel Han <danielhanchen@gmail.com>
Date: Mon, 6 Apr 2026 07:33:28 -0700
Subject: [PATCH 047/216] Remove Gemma-4 from FORCE_FLOAT32 (#4875)

Gemma-4 does not need FORCE_FLOAT32. Testing shows that both float16 and
bfloat16 work correctly without the forced float32 override:

- Inference: identical outputs for float16 and bfloat16 (greedy decoding)
- Training (100 steps, 4-bit LoRA, SFT on FineTome-100k):
  - float16 final loss: 3.048
  - bfloat16 final loss: 3.065
  - Losses converge to within 0.02 by step 60
  - Grad norms healthy and comparable for both dtypes

The FORCE_FLOAT32 path was actually causing training divergence. With
it enabled, the compiled float32 run diverged at step ~28 with grad norms
collapsing to near zero and loss plateauing at ~12.4. Without it, both
dtypes train normally.

This enables float16 on Tesla T4 and other GPUs without bfloat16 support.
---
 unsloth/models/loader.py | 2 --
 1 file changed, 2 deletions(-)

diff --git a/unsloth/models/loader.py b/unsloth/models/loader.py
index a6cb3eb529..5827711d04 100644
--- a/unsloth/models/loader.py
+++ b/unsloth/models/loader.py
@@ -108,8 +108,6 @@ FORCE_FLOAT32 = [
     "gemma3n",
     "gpt_oss",
     "qwen3_5",  # Qwen3.5 GDN layers produce NaN grad norms in float16 training
-    "gemma4,",  # Add comma bc gemma4 will match gemma4_text
-    "gemma4_text",
 ]
 
 global DISABLE_COMPILE_MODEL_NAMES

From 0835f0a61bad20f286507ffeaee3831944ec782c Mon Sep 17 00:00:00 2001
From: Roland Tannous <115670425+rolandtannous@users.noreply.github.com>
Date: Mon, 6 Apr 2026 18:46:39 +0400
Subject: [PATCH 048/216] fix: skip redundant HfFileSystem().glob() calls in
 loader.py (#4852)

* fix: skip redundant HfFileSystem().glob() calls in loader.py

Guard the SUPPORTS_LLAMA32 glob blocks with `is_model and is_peft` so
the HfFileSystem HTTP call is only made when both configs could actually
exist. This prevents indefinite hangs on slow/unreliable networks since
the glob result is redundant when either AutoConfig or PeftConfig
already failed to load.

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

* Remove test file from main PR - moved to separate PR

Tests for the glob skip guard belong in their own PR to keep
the loader change minimal and reviewable.

* Harden HfFileSystem glob: fix Windows path splitting, add try/except

- Use str.rsplit("/", 1) instead of os.path.split to extract filenames
  from HfFileSystem paths. HfFileSystem always returns POSIX-style paths,
  but os.path.split uses the OS separator, so on Windows the entire path
  was returned as the "filename" and the config name comparison always
  failed.
- Wrap the HfFileSystem().glob() call in try/except to gracefully handle
  network failures (offline mode, timeouts, unreachable Hub). On failure
  both_exist stays False, which is the safe default.

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

* Remove redundant HfFileSystem().glob() call for remote repos

When is_model and is_peft are both True, AutoConfig and PeftConfig
have already loaded successfully, proving both config.json and
adapter_config.json exist. The HfFileSystem network call to re-verify
this was redundant and could cause hangs on slow networks.

Replace the glob + try/except block with a direct both_exist = True
assignment.

* Remove unused HfFileSystem import

HfFileSystem was only used for the glob() calls that were replaced
with direct both_exist = True assignments in the previous commit.

---------

Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com>
Co-authored-by: Daniel Han <danielhanchen@gmail.com>
---
 unsloth/models/loader.py | 28 ++++++++++------------------
 1 file changed, 10 insertions(+), 18 deletions(-)

diff --git a/unsloth/models/loader.py b/unsloth/models/loader.py
index 5827711d04..df97c5c7df 100644
--- a/unsloth/models/loader.py
+++ b/unsloth/models/loader.py
@@ -49,7 +49,6 @@ except:
     except:
         # For older versions of huggingface_hub
         from huggingface_hub.utils._token import get_token
-from huggingface_hub import HfFileSystem
 import importlib.util
 from ..device_type import (
     is_hip,
@@ -506,7 +505,7 @@ class FastLanguageModel(FastLlamaModel):
             model_type = model_types
 
         # New transformers need to check manually.
-        if SUPPORTS_LLAMA32:
+        if SUPPORTS_LLAMA32 and is_model and is_peft:
             # Check if folder exists locally
             if os.path.isdir(model_name):
                 exist_adapter_config = os.path.exists(
@@ -515,14 +514,10 @@ class FastLanguageModel(FastLlamaModel):
                 exist_config = os.path.exists(os.path.join(model_name, "config.json"))
                 both_exist = exist_adapter_config and exist_config
             else:
-                # Because HfFileSystem assumes linux paths, we need to set the path with forward slashes, even on Windows.
-                files = HfFileSystem(token = token).glob(f"{model_name}/*.json")
-                files = list(os.path.split(x)[-1] for x in files)
-                if (
-                    sum(x == "adapter_config.json" or x == "config.json" for x in files)
-                    >= 2
-                ):
-                    both_exist = True
+                # Both AutoConfig and PeftConfig loaded successfully from this
+                # remote repo, so both config.json and adapter_config.json
+                # definitely exist -- no need for an extra HfFileSystem network call.
+                both_exist = True
 
         if not is_model and not is_peft:
             error = autoconfig_error if autoconfig_error is not None else peft_error
@@ -1280,7 +1275,7 @@ class FastModel(FastBaseModel):
             os.environ["UNSLOTH_DISABLE_STATIC_GENERATION"] = "1"
 
         # New transformers need to check manually.
-        if SUPPORTS_LLAMA32:
+        if SUPPORTS_LLAMA32 and is_model and is_peft:
             # Check if folder exists locally
             if os.path.isdir(model_name):
                 exist_adapter_config = os.path.exists(
@@ -1289,13 +1284,10 @@ class FastModel(FastBaseModel):
                 exist_config = os.path.exists(os.path.join(model_name, "config.json"))
                 both_exist = exist_adapter_config and exist_config
             else:
-                files = HfFileSystem(token = token).glob(f"{model_name}/*.json")
-                files = list(os.path.split(x)[-1] for x in files)
-                if (
-                    sum(x == "adapter_config.json" or x == "config.json" for x in files)
-                    >= 2
-                ):
-                    both_exist = True
+                # Both AutoConfig and PeftConfig loaded successfully from this
+                # remote repo, so both config.json and adapter_config.json
+                # definitely exist -- no need for an extra HfFileSystem network call.
+                both_exist = True
 
         if not is_model and not is_peft:
             error = autoconfig_error if autoconfig_error is not None else peft_error

From aa4c6010e189c3391bed238110cf46007b32a7d0 Mon Sep 17 00:00:00 2001
From: JYYYYYT <18164100443@163.com>
Date: Mon, 6 Apr 2026 23:31:07 +0800
Subject: [PATCH 049/216] fix(studio): custom folder scan fails to find GGUF
 variants when pointing directly at a model directory (#4860)

Fix custom folder scanning when pointing directly at a model directory.

When a user adds a custom scan folder that points directly at a model
directory (e.g. /path/to/gemma-4-e2b-it-gguf/ containing config.json
and gemma-4-E2B-it-BF16.gguf), the model list previously showed
individual .gguf files as separate entries instead of recognizing the
directory as a single model. Clicking any entry showed "No GGUF
variants found" because list_local_gguf_variants received a file path
and immediately returned empty.

Changes:
- Add _is_model_directory() helper that detects directories with both
  config metadata and actual model weight files (excludes mmproj GGUFs
  and non-weight .bin files like tokenizer.bin)
- _scan_models_dir: detect self-model and return single directory entry
- _scan_lmstudio_dir: surface model directories directly instead of
  descending into them as publisher folders; handle both root and child
  model directories
- Add _resolve_gguf_dir() helper for GGUF path resolution that only
  falls back to parent directory when parent has model metadata
- list_local_gguf_variants / _find_local_gguf_by_variant: use resolver
  so .gguf file paths inside model directories work correctly
---
 studio/backend/routes/models.py             | 96 +++++++++++++++++++++
 studio/backend/utils/models/model_config.py | 28 +++++-
 2 files changed, 120 insertions(+), 4 deletions(-)

diff --git a/studio/backend/routes/models.py b/studio/backend/routes/models.py
index ea385436b8..3f361ca5eb 100644
--- a/studio/backend/routes/models.py
+++ b/studio/backend/routes/models.py
@@ -138,6 +138,47 @@ def _resolve_hf_cache_dir() -> Path:
         return Path.home() / ".cache" / "huggingface" / "hub"
 
 
+def _is_model_directory(d: Path) -> bool:
+    """Return ``True`` when *d* looks like a model directory.
+
+    A model directory must have **both** a config file (``config.json`` or
+    ``adapter_config.json``) **and** actual model weight files.  Both
+    conditions are required: a bare directory with only loose ``.gguf``
+    files (no config) might be a mixed collection, and a ``config.json``
+    alone (no weights) is not a model directory.
+
+    Excludes ``mmproj`` GGUF files (vision projectors) and non-weight
+    ``.bin`` files (``tokenizer.bin``, ``vocab.bin``, etc.) from the
+    weight check to avoid false positives.
+    """
+
+    def _is_weight_file(f: Path) -> bool:
+        suffix = f.suffix.lower()
+        if suffix == ".safetensors":
+            return True
+        if suffix == ".gguf":
+            return "mmproj" not in f.name.lower()
+        if suffix == ".bin":
+            name = f.name.lower()
+            return (
+                name.startswith("pytorch_model")
+                or name.startswith("model")
+                or name.startswith("adapter_model")
+                or name.startswith("consolidated")
+            )
+        return False
+
+    try:
+        has_config = (d / "config.json").exists() or (
+            d / "adapter_config.json"
+        ).exists()
+        if not has_config:
+            return False
+        return any(_is_weight_file(f) for f in d.iterdir() if f.is_file())
+    except OSError:
+        return False
+
+
 def _scan_models_dir(
     models_dir: Path,
     *,
@@ -146,6 +187,23 @@ def _scan_models_dir(
     if not models_dir.exists() or not models_dir.is_dir():
         return []
 
+    _is_self_model = _is_model_directory(models_dir)
+
+    if _is_self_model:
+        try:
+            updated_at = models_dir.stat().st_mtime
+        except OSError:
+            updated_at = None
+        return [
+            LocalModelInfo(
+                id = str(models_dir),
+                display_name = models_dir.name,
+                path = str(models_dir),
+                source = "models_dir",
+                updated_at = updated_at,
+            ),
+        ]
+
     found: List[LocalModelInfo] = []
     for child in models_dir.iterdir():
         if limit is not None and len(found) >= limit:
@@ -243,6 +301,25 @@ def _scan_lmstudio_dir(lm_dir: Path) -> List[LocalModelInfo]:
     if not lm_dir.exists() or not lm_dir.is_dir():
         return []
 
+    # If the directory itself is a model directory (has config AND weight
+    # files), it is not an LM Studio publisher structure -- return it as a
+    # single model entry.  We cannot skip it silently because this function
+    # is the only scanner called for default LM Studio roots.
+    if _is_model_directory(lm_dir):
+        try:
+            updated_at = lm_dir.stat().st_mtime
+        except OSError:
+            updated_at = None
+        return [
+            LocalModelInfo(
+                id = str(lm_dir),
+                display_name = lm_dir.name,
+                path = str(lm_dir),
+                source = "lmstudio",
+                updated_at = updated_at,
+            ),
+        ]
+
     found: List[LocalModelInfo] = []
     for child in lm_dir.iterdir():
         try:
@@ -263,6 +340,25 @@ def _scan_lmstudio_dir(lm_dir: Path) -> List[LocalModelInfo]:
                     )
                 continue
 
+            # If the child directory itself looks like a model directory
+            # (has config AND weight files), surface it directly instead
+            # of descending into it as a publisher.
+            if _is_model_directory(child):
+                try:
+                    updated_at = child.stat().st_mtime
+                except OSError:
+                    updated_at = None
+                found.append(
+                    LocalModelInfo(
+                        id = str(child),
+                        display_name = child.name,
+                        path = str(child),
+                        source = "lmstudio",
+                        updated_at = updated_at,
+                    ),
+                )
+                continue
+
             # child is a publisher directory -- scan its sub-directories
             for model_dir in child.iterdir():
                 try:
diff --git a/studio/backend/utils/models/model_config.py b/studio/backend/utils/models/model_config.py
index 9d72846ca3..61226e52cb 100644
--- a/studio/backend/utils/models/model_config.py
+++ b/studio/backend/utils/models/model_config.py
@@ -1130,6 +1130,26 @@ def list_gguf_variants(
     return variants, has_vision
 
 
+def _resolve_gguf_dir(p: Path) -> Optional[Path]:
+    """Resolve a path to the directory containing GGUF variants.
+
+    If *p* is already a directory, returns it directly.  If *p* is a ``.gguf``
+    file whose parent directory has model metadata (``config.json`` or
+    ``adapter_config.json``), returns the parent -- all GGUFs in that
+    directory belong to the same model.  Returns ``None`` for loose standalone
+    GGUFs (no config) to avoid cross-wiring unrelated models.
+    """
+    if p.is_dir():
+        return p
+    if p.is_file() and p.suffix.lower() == ".gguf":
+        parent = p.parent
+        if (parent / "config.json").exists() or (
+            parent / "adapter_config.json"
+        ).exists():
+            return parent
+    return None
+
+
 def list_local_gguf_variants(
     directory: str,
 ) -> tuple[list[GgufVariantInfo], bool]:
@@ -1142,8 +1162,8 @@ def list_local_gguf_variants(
     Returns:
         (variants, has_vision): list of non-mmproj GGUF variants + vision flag.
     """
-    p = Path(directory)
-    if not p.is_dir():
+    p = _resolve_gguf_dir(Path(directory))
+    if p is None:
         return [], False
 
     quant_totals: dict[str, int] = {}
@@ -1183,8 +1203,8 @@ def _find_local_gguf_by_variant(directory: str, variant: str) -> Optional[str]:
 
     Returns the resolved absolute path, or ``None`` if no match.
     """
-    p = Path(directory)
-    if not p.is_dir():
+    p = _resolve_gguf_dir(Path(directory))
+    if p is None:
         return None
 
     matches = sorted(

From 723bfb236305eb4ae4d0fdbf724fbe801b71da66 Mon Sep 17 00:00:00 2001
From: Daniel Han <danielhanchen@gmail.com>
Date: Mon, 6 Apr 2026 08:54:36 -0700
Subject: [PATCH 050/216] Add unit tests for HfFileSystem glob skip guard
 (#4854)

Tests verifying that HfFileSystem().glob() is correctly skipped when
is_model or is_peft is False, matching the guard added in PR #4852.
---
 tests/test_loader_glob_skip.py | 152 +++++++++++++++++++++++++++++++++
 1 file changed, 152 insertions(+)
 create mode 100644 tests/test_loader_glob_skip.py

diff --git a/tests/test_loader_glob_skip.py b/tests/test_loader_glob_skip.py
new file mode 100644
index 0000000000..d461e23cf8
--- /dev/null
+++ b/tests/test_loader_glob_skip.py
@@ -0,0 +1,152 @@
+"""Tests that HfFileSystem().glob() is skipped when is_model or is_peft is False.
+
+The glob calls in FastLanguageModel.from_pretrained and FastModel.from_pretrained
+exist solely to detect repos with both config.json and adapter_config.json. When
+either AutoConfig or PeftConfig fails to load, the glob cannot find both files,
+so calling it is redundant and risks hanging on slow networks.
+"""
+
+import os
+import unittest
+from unittest.mock import MagicMock, patch
+
+
+class TestGlobSkippedWhenNotBothConfigs(unittest.TestCase):
+    """Verify HfFileSystem.glob is not called when is_model or is_peft is False."""
+
+    def _run_both_exist_block(
+        self, is_model, is_peft, supports_llama32, model_name, is_local_dir = False
+    ):
+        """Simulate the both_exist detection block from loader.py.
+
+        This mirrors the exact logic at lines 500-517 / 1276-1292 of loader.py.
+        Returns (both_exist, glob_called).
+        """
+        from unittest.mock import MagicMock
+
+        both_exist = (is_model and is_peft) and not supports_llama32
+        glob_mock = MagicMock(
+            return_value = [
+                f"{model_name}/config.json",
+                f"{model_name}/adapter_config.json",
+            ]
+        )
+
+        # This mirrors the guarded block in loader.py
+        if supports_llama32 and is_model and is_peft:
+            if is_local_dir:
+                # Local path branch — would use os.path.exists in real code
+                both_exist = True  # simulate both files present locally
+            else:
+                files = glob_mock(f"{model_name}/*.json")
+                files = list(os.path.split(x)[-1] for x in files)
+                if (
+                    sum(x == "adapter_config.json" or x == "config.json" for x in files)
+                    >= 2
+                ):
+                    both_exist = True
+
+        return both_exist, glob_mock.called
+
+    # --- Cases where glob should NOT be called ---
+
+    def test_glob_skipped_when_is_model_false(self):
+        both_exist, glob_called = self._run_both_exist_block(
+            is_model = False,
+            is_peft = True,
+            supports_llama32 = True,
+            model_name = "org/some-adapter",
+        )
+        self.assertFalse(glob_called, "glob should not be called when is_model=False")
+        self.assertFalse(both_exist)
+
+    def test_glob_skipped_when_is_peft_false(self):
+        both_exist, glob_called = self._run_both_exist_block(
+            is_model = True,
+            is_peft = False,
+            supports_llama32 = True,
+            model_name = "org/some-model",
+        )
+        self.assertFalse(glob_called, "glob should not be called when is_peft=False")
+        self.assertFalse(both_exist)
+
+    def test_glob_skipped_when_both_false(self):
+        both_exist, glob_called = self._run_both_exist_block(
+            is_model = False,
+            is_peft = False,
+            supports_llama32 = True,
+            model_name = "org/bad-repo",
+        )
+        self.assertFalse(glob_called, "glob should not be called when both are False")
+        self.assertFalse(both_exist)
+
+    def test_glob_skipped_when_supports_llama32_false(self):
+        both_exist, glob_called = self._run_both_exist_block(
+            is_model = True,
+            is_peft = True,
+            supports_llama32 = False,
+            model_name = "org/some-model",
+        )
+        self.assertFalse(
+            glob_called, "glob should not be called when SUPPORTS_LLAMA32=False"
+        )
+        # both_exist is set by the old-style check: (is_model and is_peft) and not SUPPORTS_LLAMA32
+        self.assertTrue(both_exist)
+
+    # --- Cases where glob SHOULD be called ---
+
+    def test_glob_called_when_both_true_and_supports_llama32(self):
+        both_exist, glob_called = self._run_both_exist_block(
+            is_model = True,
+            is_peft = True,
+            supports_llama32 = True,
+            model_name = "org/mixed-repo",
+        )
+        self.assertTrue(
+            glob_called, "glob should be called when is_model and is_peft are both True"
+        )
+        self.assertTrue(both_exist)
+
+    def test_local_dir_skips_glob(self):
+        both_exist, glob_called = self._run_both_exist_block(
+            is_model = True,
+            is_peft = True,
+            supports_llama32 = True,
+            model_name = "/local/path/to/model",
+            is_local_dir = True,
+        )
+        self.assertFalse(glob_called, "glob should not be called for local directories")
+        self.assertTrue(both_exist)
+
+
+class TestLoaderSourceHasGuard(unittest.TestCase):
+    """Verify the actual loader.py source code has the is_model/is_peft guard."""
+
+    def test_loader_source_has_guard(self):
+        """Check that both SUPPORTS_LLAMA32 checks in loader.py include is_model and is_peft."""
+        loader_path = os.path.join(
+            os.path.dirname(__file__), os.pardir, "unsloth", "models", "loader.py"
+        )
+        with open(loader_path) as f:
+            source = f.read()
+
+        # Find all lines with the SUPPORTS_LLAMA32 check near glob usage
+        lines = source.splitlines()
+        guard_lines = [
+            line.strip()
+            for line in lines
+            if "SUPPORTS_LLAMA32" in line and "if " in line and "is_model" in line
+        ]
+        # There should be exactly 2 guarded checks (one per from_pretrained method)
+        self.assertEqual(
+            len(guard_lines),
+            2,
+            f"Expected 2 guarded SUPPORTS_LLAMA32 checks with is_model/is_peft, found {len(guard_lines)}: {guard_lines}",
+        )
+        for line in guard_lines:
+            self.assertIn("is_model", line)
+            self.assertIn("is_peft", line)
+
+
+if __name__ == "__main__":
+    unittest.main()

From 4c83e3540ec878db055d35583fa83a3415acc957 Mon Sep 17 00:00:00 2001
From: Daniel Han <danielhanchen@gmail.com>
Date: Mon, 6 Apr 2026 09:20:17 -0700
Subject: [PATCH 051/216] Update

---
 pyproject.toml           | 4 ++--
 unsloth/models/_utils.py | 2 +-
 2 files changed, 3 insertions(+), 3 deletions(-)

diff --git a/pyproject.toml b/pyproject.toml
index e391b4df3d..50bdf58b95 100644
--- a/pyproject.toml
+++ b/pyproject.toml
@@ -88,7 +88,7 @@ huggingfacenotorch = [
 ]
 huggingface = [
     "unsloth[huggingfacenotorch]",
-    "unsloth_zoo>=2026.4.2",
+    "unsloth_zoo>=2026.4.3",
     "torchvision",
     "unsloth[triton]",
 ]
@@ -578,7 +578,7 @@ colab-ampere-torch220 = [
     "flash-attn>=2.6.3 ; ('linux' in sys_platform)",
 ]
 colab-new = [
-    "unsloth_zoo>=2026.4.2",
+    "unsloth_zoo>=2026.4.3",
     "packaging",
     "tyro",
     "transformers>=4.51.3,!=4.52.0,!=4.52.1,!=4.52.2,!=4.52.3,!=4.53.0,!=4.54.0,!=4.55.0,!=4.55.1,!=4.57.0,!=4.57.4,!=4.57.5,!=5.0.0,!=5.1.0,<=5.5.0",
diff --git a/unsloth/models/_utils.py b/unsloth/models/_utils.py
index dcb7334417..9ee2ade3db 100644
--- a/unsloth/models/_utils.py
+++ b/unsloth/models/_utils.py
@@ -12,7 +12,7 @@
 # See the License for the specific language governing permissions and
 # limitations under the License.
 
-__version__ = "2026.4.2"
+__version__ = "2026.4.3"
 
 __all__ = [
     "SUPPORTS_BFLOAT16",

From 8c89b84bb678659139e2530cc56ca291583828c7 Mon Sep 17 00:00:00 2001
From: Lee Jackson <130007945+Imagineer99@users.noreply.github.com>
Date: Mon, 6 Apr 2026 17:32:54 +0100
Subject: [PATCH 052/216] Studio: Fix empty chat threads on navigation and
 stabilize new chat flow (#4872)

* fix(chat): prevent implicit empty thread creation and stabilize new-chat flow

* fix(chat): harden compare thread sync and simplify sidebar thread query

* fix(chat): harden new-thread state sync and isolate compare active thread updates

* fix(chat): stabilize new-thread state sync and prevent compare/session bleed

* Fix thread restoration, handleNewThread guard, sidebar filter, and delete flow

- Remove __LOCALID_ filter from getInitialSingleChatView: in this
  Dexie-backed adapter, AUI's __LOCALID_ prefixed IDs ARE the real
  persistent thread IDs stored by initialize(). Filtering them out
  breaks thread restoration on navigation.

- Simplify handleNewThread to synchronous: the async Dexie message
  check is redundant (persistence is already deferred to first append)
  and strands users on legacy empty threads. Use a simple guard that
  checks the store's activeThreadId to detect unsent drafts.

- Add message-count filter to sidebar: filter threads to only show
  those with at least one message, hiding legacy empty threads.

- Add store-based sidebar highlighting fallback: use activeThreadId
  from the store when view.threadId is not set (nonce-backed chats).

- Fix handleDelete to call onNewThread() instead of onSelect(), and
  clear activeThreadId, so the runtime properly resets after deleting
  the active thread.

* Fix handleDelete nonce path and restore __LOCALID_ filter

handleDelete was calling onNewThread() after clearing activeThreadId,
but the handleNewThread guard sees !view.threadId && !activeThreadId
and returns early, leaving the UI stuck on the deleted thread.
Fix by directly calling onSelect with a new nonce instead.

Restore __LOCALID_ filter in getInitialSingleChatView to prevent
restoring unpersisted AUI local thread IDs on navigation. Without
this filter, navigating away from /chat before sending a message
would restore a non-existent thread that Dexie cannot fetch.

---------

Co-authored-by: Daniel Han <danielhanchen@gmail.com>
---
 .../frontend/src/features/chat/chat-page.tsx  | 35 +++++++++--
 .../src/features/chat/runtime-provider.tsx    | 61 +++++++++++--------
 .../src/features/chat/thread-sidebar.tsx      | 22 +++++--
 3 files changed, 80 insertions(+), 38 deletions(-)

diff --git a/studio/frontend/src/features/chat/chat-page.tsx b/studio/frontend/src/features/chat/chat-page.tsx
index 1dbff145ee..cf1ba11d7b 100644
--- a/studio/frontend/src/features/chat/chat-page.tsx
+++ b/studio/frontend/src/features/chat/chat-page.tsx
@@ -225,6 +225,7 @@ const LoraCompareContent = memo(function LoraCompareContent({
                 modelType="base"
                 pairId={pairId}
                 initialThreadId={baseThreadId}
+                syncActiveThreadId={false}
               >
                 <RegisterCompareHandle name="base" />
                 <Thread hideComposer={true} hideWelcome={true} />
@@ -242,6 +243,7 @@ const LoraCompareContent = memo(function LoraCompareContent({
                 modelType="lora"
                 pairId={pairId}
                 initialThreadId={loraThreadId}
+                syncActiveThreadId={false}
               >
                 <RegisterCompareHandle name="lora" />
                 <Thread hideComposer={true} hideWelcome={true} />
@@ -343,6 +345,7 @@ const GeneralCompareContent = memo(function GeneralCompareContent({
                 modelType="model1"
                 pairId={pairId}
                 initialThreadId={model1ThreadId}
+                syncActiveThreadId={false}
               >
                 <RegisterCompareHandle name="model1" />
                 <Thread hideComposer={true} hideWelcome={true} />
@@ -376,6 +379,7 @@ const GeneralCompareContent = memo(function GeneralCompareContent({
                 modelType="model2"
                 pairId={pairId}
                 initialThreadId={model2ThreadId}
+                syncActiveThreadId={false}
               >
                 <RegisterCompareHandle name="model2" />
                 <Thread hideComposer={true} hideWelcome={true} />
@@ -479,11 +483,19 @@ function TopBarActions({
   );
 }
 
+function getInitialSingleChatView(): ChatView {
+  const id = useChatRuntimeStore.getState().activeThreadId;
+  if (typeof id === "string" && id.length > 0 && !id.startsWith("__LOCALID_")) {
+    return { mode: "single", threadId: id };
+  }
+  return { mode: "single" };
+}
+
 export function ChatPage(): ReactElement {
-  const [view, setView] = useState<ChatView>({
-    mode: "single",
-    newThreadNonce: crypto.randomUUID(),
-  });
+  // Do not set newThreadNonce here: each /chat mount would run ThreadNewChatSwitch
+  // and create spurious threads when navigating (e.g. Recipes / Export). New Chat
+  // explicitly sets a nonce in handleNewThread.
+  const [view, setView] = useState<ChatView>(getInitialSingleChatView);
   const [settingsOpen, setSettingsOpen] = useState(false);
   const [modelSelectorOpen, setModelSelectorOpen] = useState(false);
   const [modelSelectorLocked, setModelSelectorLocked] = useState(false);
@@ -587,9 +599,20 @@ export function ChatPage(): ReactElement {
     void ejectModel();
   }, [ejectModel]);
   const handleNewThread = useCallback(() => {
+    // Skip if we are already on a fresh unsaved draft with no messages sent.
+    // Once the user sends a message, append() sets activeThreadId in the store,
+    // so we check the store to know whether the current draft has been sent.
+    if (
+      view.mode === "single" &&
+      !view.threadId &&
+      !useChatRuntimeStore.getState().activeThreadId
+    ) {
+      return;
+    }
+
     useChatRuntimeStore.getState().setActiveThreadId(null);
     setView({ mode: "single", newThreadNonce: crypto.randomUUID() });
-  }, []);
+  }, [view]);
   const handleNewCompare = useCallback(() => {
     setView({ mode: "compare", pairId: crypto.randomUUID() });
     // Clear activeThreadId so compare panes do not inherit the single-chat
@@ -922,7 +945,7 @@ export function ChatPage(): ReactElement {
 
           {view.mode === "single" ? (
             <SingleContent
-              key={view.threadId ?? view.newThreadNonce ?? "new"}
+              key={view.threadId ?? "single"}
               threadId={view.threadId}
               newThreadNonce={view.newThreadNonce}
             />
diff --git a/studio/frontend/src/features/chat/runtime-provider.tsx b/studio/frontend/src/features/chat/runtime-provider.tsx
index 7e6cd8e1dd..024543edb9 100644
--- a/studio/frontend/src/features/chat/runtime-provider.tsx
+++ b/studio/frontend/src/features/chat/runtime-provider.tsx
@@ -596,6 +596,15 @@ function ThreadHistoryProvider({
 
       async append({ parentId, message }: ExportedMessageRepositoryItem) {
         const { remoteId } = await aui.threadListItem().initialize();
+        // Keep single-chat runtime state in sync once a new chat is first
+        // persisted. Compare panes intentionally do not write global activeThreadId.
+        const thread = await db.threads.get(remoteId);
+        if (thread?.modelType === "base" && !thread.pairId) {
+          const store = useChatRuntimeStore.getState();
+          if (store.activeThreadId !== remoteId) {
+            store.setActiveThreadId(remoteId);
+          }
+        }
         const content = cloneContent(message.content);
         const attachments =
           message.role === "user" ? cloneAttachments(message.attachments) : [];
@@ -658,7 +667,11 @@ function useRuntimeHook(): ReturnType<typeof useLocalRuntime> {
 
 function ThreadAutoSwitch({
   threadId,
-}: { threadId: string }): ReactElement | null {
+  syncActiveThreadId = true,
+}: {
+  threadId: string;
+  syncActiveThreadId?: boolean;
+}): ReactElement | null {
   const aui = useAui();
   const isLoading = useAuiState(({ threads }) => threads.isLoading);
   const mainThreadId = useAuiState(({ threads }) => threads.mainThreadId);
@@ -669,6 +682,13 @@ function ThreadAutoSwitch({
     }
   }, [aui, isLoading, mainThreadId, threadId]);
 
+  useEffect(() => {
+    if (!syncActiveThreadId || isLoading || mainThreadId !== threadId) {
+      return;
+    }
+    useChatRuntimeStore.getState().setActiveThreadId(threadId);
+  }, [isLoading, mainThreadId, syncActiveThreadId, threadId]);
+
   return null;
 }
 
@@ -682,30 +702,10 @@ function ThreadNewChatSwitch({
     if (isLoading) {
       return;
     }
-
-    let cancelled = false;
-    // Clear immediately so the adapter never picks up a stale thread ID
-    // from a previous chat while we initialize the new one.
+    // Switch to a fresh local thread without persisting it yet.
+    // Persistence still happens on first message append.
+    void aui.threads().switchToNewThread();
     useChatRuntimeStore.getState().setActiveThreadId(null);
-
-    void (async () => {
-      try {
-        aui.threads().switchToNewThread();
-        const { remoteId } = await aui.threadListItem().initialize();
-        if (!cancelled) {
-          useChatRuntimeStore.getState().setActiveThreadId(remoteId);
-        }
-      } catch (error) {
-        if (!cancelled) {
-          useChatRuntimeStore.getState().setActiveThreadId(null);
-        }
-        console.error("Failed to initialize new chat thread", error);
-      }
-    })();
-
-    return () => {
-      cancelled = true;
-    };
   }, [aui, isLoading, nonce]);
 
   return null;
@@ -733,12 +733,14 @@ export function ChatRuntimeProvider({
   pairId,
   initialThreadId,
   newThreadNonce,
+  syncActiveThreadId = true,
 }: {
   children: ReactNode;
   modelType?: ModelType;
   pairId?: string;
   initialThreadId?: string;
   newThreadNonce?: string;
+  syncActiveThreadId?: boolean;
 }): ReactElement {
   const runtime = useRemoteThreadListRuntime({
     runtimeHook: useRuntimeHook,
@@ -754,8 +756,15 @@ export function ChatRuntimeProvider({
 
   return (
     <AssistantRuntimeProvider runtime={runtime} aui={aui}>
-      <ActiveThreadSync enabled={modelType === "base" && !pairId && !newThreadNonce} />
-      {initialThreadId && <ThreadAutoSwitch threadId={initialThreadId} />}
+      <ActiveThreadSync
+        enabled={modelType === "base" && !pairId && !newThreadNonce && !initialThreadId}
+      />
+      {initialThreadId && (
+        <ThreadAutoSwitch
+          threadId={initialThreadId}
+          syncActiveThreadId={syncActiveThreadId}
+        />
+      )}
       {!initialThreadId && newThreadNonce && (
         <ThreadNewChatSwitch nonce={newThreadNonce} />
       )}
diff --git a/studio/frontend/src/features/chat/thread-sidebar.tsx b/studio/frontend/src/features/chat/thread-sidebar.tsx
index ba97d2ee6e..62246cdad5 100644
--- a/studio/frontend/src/features/chat/thread-sidebar.tsx
+++ b/studio/frontend/src/features/chat/thread-sidebar.tsx
@@ -22,6 +22,7 @@ import {
 } from "@hugeicons/core-free-icons";
 import { HugeiconsIcon } from "@hugeicons/react";
 import { db, useLiveQuery } from "./db";
+import { useChatRuntimeStore } from "./stores/chat-runtime-store";
 import type { ChatView, ThreadRecord } from "./types";
 
 interface SidebarItem {
@@ -76,12 +77,17 @@ export function ThreadSidebar({
   onNewCompare: () => void;
   showCompare: boolean;
 }) {
-  const allThreads = useLiveQuery(
-    () => db.threads.orderBy("createdAt").reverse().toArray(),
-    [],
-  );
+  const allThreads = useLiveQuery(async () => {
+    const threadIdsWithMessage = new Set(
+      (await db.messages.orderBy("threadId").uniqueKeys()) as string[],
+    );
+    const rows = await db.threads.orderBy("createdAt").reverse().toArray();
+    return rows.filter((t) => !t.archived && threadIdsWithMessage.has(t.id));
+  }, []);
   const items = groupThreads(allThreads ?? []);
-  const activeId = view.mode === "single" ? view.threadId : view.pairId;
+  const storeThreadId = useChatRuntimeStore((s) => s.activeThreadId);
+  const activeId =
+    view.mode === "single" ? (view.threadId ?? storeThreadId) : view.pairId;
 
   function viewForItem(item: SidebarItem): ChatView {
     return item.type === "single"
@@ -101,7 +107,11 @@ export function ThreadSidebar({
       }
     }
     if (activeId === item.id) {
-      onSelect({ mode: "single" });
+      // Directly set a new view with a nonce rather than going through
+      // onNewThread(), which may return early if the guard sees no
+      // threadId and no activeThreadId (after we just cleared it).
+      useChatRuntimeStore.getState().setActiveThreadId(null);
+      onSelect({ mode: "single", newThreadNonce: crypto.randomUUID() });
     }
   }
 

From b295daf9323d2782fe7d1286b2e1bde89dc6cd29 Mon Sep 17 00:00:00 2001
From: Daniel Han <danielhanchen@gmail.com>
Date: Mon, 6 Apr 2026 09:39:06 -0700
Subject: [PATCH 053/216] Update _utils.py

---
 unsloth/models/_utils.py | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/unsloth/models/_utils.py b/unsloth/models/_utils.py
index 9ee2ade3db..d8fe92739a 100644
--- a/unsloth/models/_utils.py
+++ b/unsloth/models/_utils.py
@@ -12,7 +12,7 @@
 # See the License for the specific language governing permissions and
 # limitations under the License.
 
-__version__ = "2026.4.3"
+__version__ = "2026.4.4"
 
 __all__ = [
     "SUPPORTS_BFLOAT16",

From 1d8160376e169d13c386b7ef4bc1fdc8f855de68 Mon Sep 17 00:00:00 2001
From: Daniel Han <danielhanchen@gmail.com>
Date: Mon, 6 Apr 2026 09:46:35 -0700
Subject: [PATCH 054/216] Bump minimum unsloth version to 2026.4.4 in install
 scripts (#4876)

---
 install.ps1 | 10 +++++-----
 install.sh  | 10 +++++-----
 2 files changed, 10 insertions(+), 10 deletions(-)

diff --git a/install.ps1 b/install.ps1
index 5ddb42ea7e..a2acd6c4ea 100644
--- a/install.ps1
+++ b/install.ps1
@@ -819,7 +819,7 @@ shell.Run cmd, 0, False
         if ($SkipTorch) {
             # No-torch: install unsloth + unsloth-zoo with --no-deps, then
             # runtime deps (typer, safetensors, transformers, etc.) with --no-deps.
-            $baseInstallExit = Invoke-InstallCommand { uv pip install --python $VenvPython --no-deps --reinstall-package unsloth --reinstall-package unsloth-zoo "unsloth>=2026.4.2" unsloth-zoo }
+            $baseInstallExit = Invoke-InstallCommand { uv pip install --python $VenvPython --no-deps --reinstall-package unsloth --reinstall-package unsloth-zoo "unsloth>=2026.4.4" unsloth-zoo }
             if ($baseInstallExit -eq 0) {
                 $NoTorchReq = Find-NoTorchRuntimeFile
                 if ($NoTorchReq) {
@@ -827,7 +827,7 @@ shell.Run cmd, 0, False
                 }
             }
         } else {
-            $baseInstallExit = Invoke-InstallCommand { uv pip install --python $VenvPython --reinstall-package unsloth --reinstall-package unsloth-zoo "unsloth>=2026.4.2" unsloth-zoo }
+            $baseInstallExit = Invoke-InstallCommand { uv pip install --python $VenvPython --reinstall-package unsloth --reinstall-package unsloth-zoo "unsloth>=2026.4.4" unsloth-zoo }
         }
         if ($baseInstallExit -ne 0) {
             Write-Host "[ERROR] Failed to install unsloth (exit code $baseInstallExit)" -ForegroundColor Red
@@ -857,7 +857,7 @@ shell.Run cmd, 0, False
         if ($SkipTorch) {
             # No-torch: install unsloth + unsloth-zoo with --no-deps, then
             # runtime deps (typer, safetensors, transformers, etc.) with --no-deps.
-            $baseInstallExit = Invoke-InstallCommand { uv pip install --python $VenvPython --no-deps --upgrade-package unsloth --upgrade-package unsloth-zoo "unsloth>=2026.4.2" unsloth-zoo }
+            $baseInstallExit = Invoke-InstallCommand { uv pip install --python $VenvPython --no-deps --upgrade-package unsloth --upgrade-package unsloth-zoo "unsloth>=2026.4.4" unsloth-zoo }
             if ($baseInstallExit -eq 0) {
                 $NoTorchReq = Find-NoTorchRuntimeFile
                 if ($NoTorchReq) {
@@ -865,7 +865,7 @@ shell.Run cmd, 0, False
                 }
             }
         } elseif ($StudioLocalInstall) {
-            $baseInstallExit = Invoke-InstallCommand { uv pip install --python $VenvPython --upgrade-package unsloth "unsloth>=2026.4.2" unsloth-zoo }
+            $baseInstallExit = Invoke-InstallCommand { uv pip install --python $VenvPython --upgrade-package unsloth "unsloth>=2026.4.4" unsloth-zoo }
         } else {
             $baseInstallExit = Invoke-InstallCommand { uv pip install --python $VenvPython --upgrade-package unsloth "$PackageName" }
         }
@@ -886,7 +886,7 @@ shell.Run cmd, 0, False
         # Fallback: GPU detection failed to produce a URL -- let uv resolve torch
         substep "installing unsloth (this may take a few minutes)..."
         if ($StudioLocalInstall) {
-            $baseInstallExit = Invoke-InstallCommand { uv pip install --python $VenvPython unsloth-zoo "unsloth>=2026.4.2" --torch-backend=auto }
+            $baseInstallExit = Invoke-InstallCommand { uv pip install --python $VenvPython unsloth-zoo "unsloth>=2026.4.4" --torch-backend=auto }
             if ($baseInstallExit -ne 0) {
                 Write-Host "[ERROR] Failed to install unsloth (exit code $baseInstallExit)" -ForegroundColor Red
                 return
diff --git a/install.sh b/install.sh
index 053f334d2b..ea53ecc6d6 100755
--- a/install.sh
+++ b/install.sh
@@ -1040,7 +1040,7 @@ if [ "$_MIGRATED" = true ]; then
         # to prevent transitive torch resolution.
         run_install_cmd "install unsloth (migrated no-torch)" uv pip install --python "$_VENV_PY" --no-deps \
             --reinstall-package unsloth --reinstall-package unsloth-zoo \
-            "unsloth>=2026.4.2" unsloth-zoo
+            "unsloth>=2026.4.4" unsloth-zoo
         _NO_TORCH_RT="$(_find_no_torch_runtime)"
         if [ -n "$_NO_TORCH_RT" ]; then
             run_install_cmd "install no-torch runtime deps" uv pip install --python "$_VENV_PY" --no-deps -r "$_NO_TORCH_RT"
@@ -1048,7 +1048,7 @@ if [ "$_MIGRATED" = true ]; then
     else
         run_install_cmd "install unsloth (migrated)" uv pip install --python "$_VENV_PY" \
             --reinstall-package unsloth --reinstall-package unsloth-zoo \
-            "unsloth>=2026.4.2" unsloth-zoo
+            "unsloth>=2026.4.4" unsloth-zoo
     fi
     if [ "$STUDIO_LOCAL_INSTALL" = true ]; then
         substep "overlaying local repo (editable)..."
@@ -1070,7 +1070,7 @@ elif [ -n "$TORCH_INDEX_URL" ]; then
         # runtime deps (typer, safetensors, transformers, etc.) with --no-deps.
         run_install_cmd "install unsloth (no-torch)" uv pip install --python "$_VENV_PY" --no-deps \
             --upgrade-package unsloth --upgrade-package unsloth-zoo \
-            "unsloth>=2026.4.2" unsloth-zoo
+            "unsloth>=2026.4.4" unsloth-zoo
         _NO_TORCH_RT="$(_find_no_torch_runtime)"
         if [ -n "$_NO_TORCH_RT" ]; then
             run_install_cmd "install no-torch runtime deps" uv pip install --python "$_VENV_PY" --no-deps -r "$_NO_TORCH_RT"
@@ -1081,7 +1081,7 @@ elif [ -n "$TORCH_INDEX_URL" ]; then
         fi
     elif [ "$STUDIO_LOCAL_INSTALL" = true ]; then
         run_install_cmd "install unsloth (local)" uv pip install --python "$_VENV_PY" \
-            --upgrade-package unsloth "unsloth>=2026.4.2" unsloth-zoo
+            --upgrade-package unsloth "unsloth>=2026.4.4" unsloth-zoo
         substep "overlaying local repo (editable)..."
         run_install_cmd "overlay local repo" uv pip install --python "$_VENV_PY" -e "$_REPO_ROOT" --no-deps
     else
@@ -1092,7 +1092,7 @@ else
     # Fallback: GPU detection failed to produce a URL -- let uv resolve torch
     substep "installing unsloth (this may take a few minutes)..."
     if [ "$STUDIO_LOCAL_INSTALL" = true ]; then
-        run_install_cmd "install unsloth (auto torch backend)" uv pip install --python "$_VENV_PY" unsloth-zoo "unsloth>=2026.4.2" --torch-backend=auto
+        run_install_cmd "install unsloth (auto torch backend)" uv pip install --python "$_VENV_PY" unsloth-zoo "unsloth>=2026.4.4" --torch-backend=auto
         substep "overlaying local repo (editable)..."
         run_install_cmd "overlay local repo" uv pip install --python "$_VENV_PY" -e "$_REPO_ROOT" --no-deps
     else

From f801e59c29db7b4028297ef987af6bfdaa464500 Mon Sep 17 00:00:00 2001
From: Roland Tannous <115670425+rolandtannous@users.noreply.github.com>
Date: Tue, 7 Apr 2026 20:05:01 +0400
Subject: [PATCH 055/216] split venv_t5 into tiered 5.3.0/5.5.0 and fix
 trust_remote_code (#4878)

* split venv_t5 into venv_t5_530 and venv_t5_550 for tiered transformers 5.x support

* fix bfloat16 crash on T4 for FORCE_FLOAT32 models and disable trust_remote_code auto-enable for native t5 models

* revert FORCE_FLOAT32 dtype change

* restrict trust_remote_code auto-enable to Nemotron models only

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

* use config.json model_type for tier detection, add unsloth/nvidia namespace guard

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

* Revert "[pre-commit.ci] auto fixes from pre-commit.com hooks"

This reverts commit fb43d468e25379f28dd2477e6c24dd60cf55c099.

* Revert "use config.json model_type for tier detection, add unsloth/nvidia namespace guard"

This reverts commit fc49ae24531780a658049e6c238f146960f466b6.

* add unsloth/nvidia namespace guard to Nemotron trust_remote_code auto-enable

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

* reorder tier checks: all substring matches before config.json fetches

* extract shared activate_transformers_for_subprocess into transformers_version.py

* narrow Nemotron trust_remote_code to nemotron_h/nemotron-3-nano, add to export worker

* clean venv_t5 dirs before re-install in setup.sh, clarify version alias comment

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

* run venv_t5 migration outside deps fast-path gate in both setup scripts

---------

Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com>
---
 studio/backend/core/export/worker.py          |  41 +--
 studio/backend/core/inference/worker.py       |  44 +--
 studio/backend/core/training/worker.py        |  51 +--
 .../tests/test_transformers_version.py        | 148 ++++++++
 studio/backend/utils/models/model_config.py   |   5 +-
 studio/backend/utils/transformers_version.py  | 340 ++++++++++++++----
 studio/setup.ps1                              | 135 ++++---
 studio/setup.sh                               |  53 ++-
 8 files changed, 601 insertions(+), 216 deletions(-)

diff --git a/studio/backend/core/export/worker.py b/studio/backend/core/export/worker.py
index 6af6ff1193..3f3dc955fa 100644
--- a/studio/backend/core/export/worker.py
+++ b/studio/backend/core/export/worker.py
@@ -30,37 +30,15 @@ logger = get_logger(__name__)
 
 
 def _activate_transformers_version(model_name: str) -> None:
-    """Activate the correct transformers version BEFORE any ML imports.
-
-    If the model needs transformers 5.x, prepend the pre-installed .venv_t5/
-    directory to sys.path. Otherwise do nothing (default 4.57.x in .venv/).
-    """
+    """Activate the correct transformers version BEFORE any ML imports."""
     # Ensure backend is on path for utils imports
     backend_path = str(Path(__file__).resolve().parent.parent.parent)
     if backend_path not in sys.path:
         sys.path.insert(0, backend_path)
 
-    from utils.transformers_version import (
-        needs_transformers_5,
-        _resolve_base_model,
-        _ensure_venv_t5_exists,
-        _VENV_T5_DIR,
-    )
+    from utils.transformers_version import activate_transformers_for_subprocess
 
-    resolved = _resolve_base_model(model_name)
-    if needs_transformers_5(resolved):
-        if not _ensure_venv_t5_exists():
-            raise RuntimeError(
-                f"Cannot activate transformers 5.x: .venv_t5 missing at {_VENV_T5_DIR}"
-            )
-        if _VENV_T5_DIR not in sys.path:
-            sys.path.insert(0, _VENV_T5_DIR)
-        logger.info("Activated transformers 5.x from %s", _VENV_T5_DIR)
-        # Propagate to child subprocesses (e.g. GGUF converter)
-        _pp = os.environ.get("PYTHONPATH", "")
-        os.environ["PYTHONPATH"] = _VENV_T5_DIR + (os.pathsep + _pp if _pp else "")
-    else:
-        logger.info("Using default transformers (4.57.x) for %s", model_name)
+    activate_transformers_for_subprocess(model_name)
 
 
 def _send_response(resp_queue: Any, response: dict) -> None:
@@ -78,6 +56,19 @@ def _handle_load(backend, cmd: dict, resp_queue: Any) -> None:
     load_in_4bit = cmd.get("load_in_4bit", True)
     trust_remote_code = cmd.get("trust_remote_code", False)
 
+    # Auto-enable trust_remote_code for NemotronH/Nano models.
+    if not trust_remote_code:
+        _NEMOTRON_TRUST_SUBSTRINGS = ("nemotron_h", "nemotron-h", "nemotron-3-nano")
+        _cp_lower = checkpoint_path.lower()
+        if any(sub in _cp_lower for sub in _NEMOTRON_TRUST_SUBSTRINGS) and (
+            _cp_lower.startswith("unsloth/") or _cp_lower.startswith("nvidia/")
+        ):
+            trust_remote_code = True
+            logger.info(
+                "Auto-enabled trust_remote_code for Nemotron model: %s",
+                checkpoint_path,
+            )
+
     try:
         _send_response(
             resp_queue,
diff --git a/studio/backend/core/inference/worker.py b/studio/backend/core/inference/worker.py
index 7f7291a56d..fbcce276ba 100644
--- a/studio/backend/core/inference/worker.py
+++ b/studio/backend/core/inference/worker.py
@@ -34,37 +34,15 @@ from utils.hardware import apply_gpu_ids
 
 
 def _activate_transformers_version(model_name: str) -> None:
-    """Activate the correct transformers version BEFORE any ML imports.
-
-    If the model needs transformers 5.x, prepend the pre-installed .venv_t5/
-    directory to sys.path. Otherwise do nothing (default 4.57.x in .venv/).
-    """
+    """Activate the correct transformers version BEFORE any ML imports."""
     # Ensure backend is on path for utils imports
     backend_path = str(Path(__file__).resolve().parent.parent.parent)
     if backend_path not in sys.path:
         sys.path.insert(0, backend_path)
 
-    from utils.transformers_version import (
-        needs_transformers_5,
-        _resolve_base_model,
-        _ensure_venv_t5_exists,
-        _VENV_T5_DIR,
-    )
+    from utils.transformers_version import activate_transformers_for_subprocess
 
-    resolved = _resolve_base_model(model_name)
-    if needs_transformers_5(resolved):
-        if not _ensure_venv_t5_exists():
-            raise RuntimeError(
-                f"Cannot activate transformers 5.x: .venv_t5 missing at {_VENV_T5_DIR}"
-            )
-        if _VENV_T5_DIR not in sys.path:
-            sys.path.insert(0, _VENV_T5_DIR)
-        logger.info("Activated transformers 5.x from %s", _VENV_T5_DIR)
-        # Propagate to child subprocesses (e.g. GGUF converter)
-        _pp = os.environ.get("PYTHONPATH", "")
-        os.environ["PYTHONPATH"] = _VENV_T5_DIR + (os.pathsep + _pp if _pp else "")
-    else:
-        logger.info("Using default transformers (4.57.x) for %s", model_name)
+    activate_transformers_for_subprocess(model_name)
 
 
 def _decode_image(image_base64: str):
@@ -309,19 +287,21 @@ def _handle_load(backend, config: dict, resp_queue: Any) -> None:
                 except Exception as e:
                     logger.warning("Could not read adapter_config.json: %s", e)
 
-        # Auto-enable trust_remote_code for unsloth/* transformers 5.x models
-        # (matches the training worker logic in core/training/worker.py)
+        # Auto-enable trust_remote_code for NemotronH/Nano models only.
+        # NemotronH has config parsing bugs requiring trust_remote_code=True.
+        # Other transformers 5.x models are native and do NOT need it.
+        # NOTE: Must NOT match Llama-Nemotron (standard Llama architecture).
+        _NEMOTRON_TRUST_SUBSTRINGS = ("nemotron_h", "nemotron-h", "nemotron-3-nano")
         trust_remote_code = config.get("trust_remote_code", False)
         if not trust_remote_code:
-            from utils.transformers_version import needs_transformers_5
-
             model_name = config["model_name"]
-            if needs_transformers_5(model_name) and model_name.lower().startswith(
-                "unsloth/"
+            _mn_lower = model_name.lower()
+            if any(sub in _mn_lower for sub in _NEMOTRON_TRUST_SUBSTRINGS) and (
+                _mn_lower.startswith("unsloth/") or _mn_lower.startswith("nvidia/")
             ):
                 trust_remote_code = True
                 logger.info(
-                    "Auto-enabled trust_remote_code for unsloth/* transformers 5.x model: %s",
+                    "Auto-enabled trust_remote_code for Nemotron model: %s",
                     model_name,
                 )
 
diff --git a/studio/backend/core/training/worker.py b/studio/backend/core/training/worker.py
index 0454eada89..a461972eca 100644
--- a/studio/backend/core/training/worker.py
+++ b/studio/backend/core/training/worker.py
@@ -306,37 +306,15 @@ def _ensure_mamba_ssm(event_queue: Any, model_name: str) -> None:
 
 
 def _activate_transformers_version(model_name: str) -> None:
-    """Activate the correct transformers version BEFORE any ML imports.
-
-    If the model needs transformers 5.x, prepend the pre-installed .venv_t5/
-    directory to sys.path. Otherwise do nothing (default 4.57.x in .venv/).
-    """
+    """Activate the correct transformers version BEFORE any ML imports."""
     # Ensure backend is on path for utils imports
     backend_path = str(Path(__file__).resolve().parent.parent.parent)
     if backend_path not in sys.path:
         sys.path.insert(0, backend_path)
 
-    from utils.transformers_version import (
-        needs_transformers_5,
-        _resolve_base_model,
-        _ensure_venv_t5_exists,
-        _VENV_T5_DIR,
-    )
+    from utils.transformers_version import activate_transformers_for_subprocess
 
-    resolved = _resolve_base_model(model_name)
-    if needs_transformers_5(resolved):
-        if not _ensure_venv_t5_exists():
-            raise RuntimeError(
-                f"Cannot activate transformers 5.x: .venv_t5 missing at {_VENV_T5_DIR}"
-            )
-        if _VENV_T5_DIR not in sys.path:
-            sys.path.insert(0, _VENV_T5_DIR)
-        logger.info("Activated transformers 5.x from %s", _VENV_T5_DIR)
-        # Propagate to child subprocesses (e.g. GGUF converter)
-        _pp = os.environ.get("PYTHONPATH", "")
-        os.environ["PYTHONPATH"] = _VENV_T5_DIR + (os.pathsep + _pp if _pp else "")
-    else:
-        logger.info("Using default transformers (4.57.x) for %s", model_name)
+    activate_transformers_for_subprocess(model_name)
 
 
 def run_training_process(
@@ -386,25 +364,22 @@ def run_training_process(
         )
         return
 
-    # ── 1a. Auto-enable trust_remote_code for unsloth/* transformers 5.x models ──
-    # Some newer architectures (e.g. NemotronH) have config parsing bugs in
-    # transformers that require trust_remote_code=True as a workaround.
-    # Only auto-enable for unsloth/* prefixed models (trusted source).
-    # Exclude Gemma 4 since it is a native transformers 5.5 model and
-    # trust_remote_code=True would bypass the compiler (disabling fused CE).
-    from utils.transformers_version import needs_transformers_5
-
+    # ── 1a. Auto-enable trust_remote_code for NemotronH/Nano models ──
+    # NemotronH has config parsing bugs in transformers that require
+    # trust_remote_code=True as a workaround. Other transformers 5.x models
+    # (Qwen3.5, Gemma 4, etc.) are native and do NOT need it — enabling it
+    # bypasses the compiler (disabling fused CE).
+    # NOTE: Must NOT match Llama-Nemotron (standard Llama architecture).
+    _NEMOTRON_TRUST_SUBSTRINGS = ("nemotron_h", "nemotron-h", "nemotron-3-nano")
     _lowered = model_name.lower()
-    _is_native_t5 = any(x in _lowered for x in ("gemma-4", "gemma4"))
     if (
-        needs_transformers_5(model_name)
-        and _lowered.startswith("unsloth/")
-        and not _is_native_t5
+        any(sub in _lowered for sub in _NEMOTRON_TRUST_SUBSTRINGS)
+        and (_lowered.startswith("unsloth/") or _lowered.startswith("nvidia/"))
         and not config.get("trust_remote_code", False)
     ):
         config["trust_remote_code"] = True
         logger.info(
-            "Auto-enabled trust_remote_code for unsloth/* transformers 5.x model: %s",
+            "Auto-enabled trust_remote_code for Nemotron model: %s",
             model_name,
         )
 
diff --git a/studio/backend/tests/test_transformers_version.py b/studio/backend/tests/test_transformers_version.py
index f3dae537c7..c031c2fea3 100644
--- a/studio/backend/tests/test_transformers_version.py
+++ b/studio/backend/tests/test_transformers_version.py
@@ -31,8 +31,11 @@ sys.modules.setdefault("loggers", _loggers_stub)
 from utils.transformers_version import (
     _resolve_base_model,
     _check_tokenizer_config_needs_v5,
+    _check_config_needs_550,
     _tokenizer_class_cache,
+    _config_needs_550_cache,
     needs_transformers_5,
+    get_transformers_tier,
 )
 
 
@@ -188,3 +191,148 @@ class TestNeedsTransformers5:
         # We test the full resolution chain here:
         resolved = _resolve_base_model(str(tmp_path))
         assert needs_transformers_5(resolved) is True
+
+
+# ---------------------------------------------------------------------------
+# _check_config_needs_550 — config.json architecture/model_type check
+# ---------------------------------------------------------------------------
+
+
+class TestCheckConfigNeeds550:
+    """Tests for _check_config_needs_550() local config.json checks."""
+
+    def setup_method(self):
+        _config_needs_550_cache.clear()
+
+    def test_gemma4_architecture(self, tmp_path: Path):
+        """config.json with Gemma4ForConditionalGeneration should return True."""
+        cfg = {
+            "architectures": ["Gemma4ForConditionalGeneration"],
+            "model_type": "gemma4",
+        }
+        (tmp_path / "config.json").write_text(json.dumps(cfg))
+
+        assert _check_config_needs_550(str(tmp_path)) is True
+
+    def test_gemma4_model_type_only(self, tmp_path: Path):
+        """config.json with model_type=gemma4 (no architectures) should return True."""
+        cfg = {"model_type": "gemma4"}
+        (tmp_path / "config.json").write_text(json.dumps(cfg))
+
+        assert _check_config_needs_550(str(tmp_path)) is True
+
+    def test_llama_architecture(self, tmp_path: Path):
+        """config.json with LlamaForCausalLM should return False."""
+        cfg = {"architectures": ["LlamaForCausalLM"], "model_type": "llama"}
+        (tmp_path / "config.json").write_text(json.dumps(cfg))
+
+        assert _check_config_needs_550(str(tmp_path)) is False
+
+    def test_no_config_json(self, tmp_path: Path):
+        """Missing config.json should return False (fail-open)."""
+        # Patch network call to avoid real fetch
+        with patch("urllib.request.urlopen") as mock_urlopen:
+            mock_urlopen.side_effect = Exception("no network")
+            assert _check_config_needs_550(str(tmp_path)) is False
+
+    def test_result_is_cached(self, tmp_path: Path):
+        """Subsequent calls should use the cache."""
+        cfg = {"architectures": ["Gemma4ForConditionalGeneration"]}
+        (tmp_path / "config.json").write_text(json.dumps(cfg))
+
+        key = str(tmp_path)
+        _check_config_needs_550(key)
+        assert key in _config_needs_550_cache
+        assert _config_needs_550_cache[key] is True
+
+    def test_local_file_skips_network(self, tmp_path: Path):
+        """When local config.json exists, no network request should be made."""
+        cfg = {"architectures": ["LlamaForCausalLM"]}
+        (tmp_path / "config.json").write_text(json.dumps(cfg))
+
+        with patch("urllib.request.urlopen") as mock_urlopen:
+            _check_config_needs_550(str(tmp_path))
+            mock_urlopen.assert_not_called()
+
+
+# ---------------------------------------------------------------------------
+# get_transformers_tier — tier detection
+# ---------------------------------------------------------------------------
+
+
+class TestGetTransformersTier:
+    """Tests for get_transformers_tier() tiered version detection."""
+
+    def setup_method(self):
+        _tokenizer_class_cache.clear()
+        _config_needs_550_cache.clear()
+
+    def test_gemma4_substring_returns_550(self):
+        assert get_transformers_tier("google/gemma-4-E2B-it") == "550"
+
+    def test_gemma4_alt_substring_returns_550(self):
+        assert get_transformers_tier("unsloth/gemma4-E4B-it") == "550"
+
+    def test_gemma4_config_json_returns_550(self, tmp_path: Path):
+        """Local checkpoint with Gemma4 architecture → 550."""
+        cfg = {
+            "architectures": ["Gemma4ForConditionalGeneration"],
+            "model_type": "gemma4",
+        }
+        (tmp_path / "config.json").write_text(json.dumps(cfg))
+
+        assert get_transformers_tier(str(tmp_path)) == "550"
+
+    def test_qwen35_returns_530(self):
+        with patch(
+            "utils.transformers_version._check_config_needs_550",
+            return_value = False,
+        ):
+            assert get_transformers_tier("Qwen/Qwen3.5-9B") == "530"
+
+    def test_ministral_returns_530(self):
+        with patch(
+            "utils.transformers_version._check_config_needs_550",
+            return_value = False,
+        ):
+            assert (
+                get_transformers_tier("mistralai/Ministral-3-8B-Instruct-2512") == "530"
+            )
+
+    def test_llama_returns_default(self):
+        with (
+            patch(
+                "utils.transformers_version._check_config_needs_550",
+                return_value = False,
+            ),
+            patch(
+                "utils.transformers_version._check_tokenizer_config_needs_v5",
+                return_value = False,
+            ),
+        ):
+            assert get_transformers_tier("meta-llama/Llama-3-8B") == "default"
+
+    def test_550_checked_before_530(self):
+        """Ensure 5.5.0 is checked first — a model matching both should get 550."""
+        # This shouldn't happen in practice, but verifies priority
+        assert get_transformers_tier("gemma-4-model") == "550"
+
+    def test_needs_transformers_5_compat(self):
+        """needs_transformers_5 should return True for both 530 and 550 models."""
+        assert needs_transformers_5("google/gemma-4-E2B-it") is True
+        with patch(
+            "utils.transformers_version._check_config_needs_550",
+            return_value = False,
+        ):
+            assert needs_transformers_5("Qwen/Qwen3.5-9B") is True
+        with (
+            patch(
+                "utils.transformers_version._check_config_needs_550",
+                return_value = False,
+            ),
+            patch(
+                "utils.transformers_version._check_tokenizer_config_needs_v5",
+                return_value = False,
+            ),
+        ):
+            assert needs_transformers_5("meta-llama/Llama-3-8B") is False
diff --git a/studio/backend/utils/models/model_config.py b/studio/backend/utils/models/model_config.py
index 61226e52cb..5be0b183d5 100644
--- a/studio/backend/utils/models/model_config.py
+++ b/studio/backend/utils/models/model_config.py
@@ -493,8 +493,9 @@ _VLM_MODEL_TYPES = {
     "minicpmv",
 }
 
-# Pre-computed .venv_t5 path and backend dir for subprocess version switching.
-_VENV_T5_DIR = str(Path.home() / ".unsloth" / "studio" / ".venv_t5")
+# Pre-computed .venv_t5 paths and backend dir for subprocess version switching.
+# Vision check uses 5.5.0 (newest, recognizes all architectures).
+_VENV_T5_DIR = str(Path.home() / ".unsloth" / "studio" / ".venv_t5_550")
 _BACKEND_DIR = str(Path(__file__).resolve().parent.parent.parent)
 
 # Inline script executed in a subprocess with transformers 5.x activated.
diff --git a/studio/backend/utils/transformers_version.py b/studio/backend/utils/transformers_version.py
index 07e4a5c000..0c13b5455b 100644
--- a/studio/backend/utils/transformers_version.py
+++ b/studio/backend/utils/transformers_version.py
@@ -5,20 +5,25 @@
 Automatic transformers version switching.
 
 Some newer model architectures (Ministral-3, GLM-4.7-Flash, Qwen3-30B-A3B MoE,
-tiny_qwen3_moe) require transformers>=5.3.0, while everything else needs the
-default 4.57.x that ships with Unsloth.
+tiny_qwen3_moe) require transformers>=5.3.0, while Gemma 4 models require
+transformers>=5.5.0.  Everything else needs the default 4.57.x that ships
+with Unsloth.
+
+Two separate target directories are maintained:
+  - .venv_t5_530/  — transformers 5.3.0 (Ministral-3, GLM, Qwen3 MoE, etc.)
+  - .venv_t5_550/  — transformers 5.5.0 (Gemma 4)
 
 When loading a LoRA adapter with a custom name, we resolve the base model from
 ``adapter_config.json`` and check *that* against the model list.
 
 Strategy:
   Training and inference run in subprocesses that activate the correct version
-  via sys.path (prepending .venv_t5/ for 5.x models). See:
+  via sys.path (prepending the appropriate .venv_t5_*/ directory). See:
     - core/training/worker.py
     - core/inference/worker.py
 
   For export (still in-process), ensure_transformers_version() does a lightweight
-  sys.path swap using the same .venv_t5/ directory pre-installed by setup.sh.
+  sys.path swap using the same directories pre-installed by setup.sh.
 """
 
 import importlib
@@ -39,7 +44,7 @@ logger = get_logger(__name__)
 # ---------------------------------------------------------------------------
 
 # Lowercase substrings — if ANY appears anywhere in the lowered model name,
-# we need transformers 5.x.
+# we need transformers 5.3.0.
 TRANSFORMERS_5_MODEL_SUBSTRINGS: tuple[str, ...] = (
     "ministral-3-",  # Ministral-3-{3,8,14}B-{Instruct,Reasoning,Base}-2512
     "glm-4.7-flash",  # GLM-4.7-Flash
@@ -47,10 +52,23 @@ TRANSFORMERS_5_MODEL_SUBSTRINGS: tuple[str, ...] = (
     "qwen3.5",  # Qwen3.5 family (35B-A3B, etc.)
     "qwen3-next",  # Qwen3-Next and variants
     "tiny_qwen3_moe",  # imdatta0/tiny_qwen3_moe_2.8B_0.7B
+)
+
+# Lowercase substrings for models that require transformers 5.5.0 (checked first).
+TRANSFORMERS_550_MODEL_SUBSTRINGS: tuple[str, ...] = (
     "gemma-4",  # Gemma-4 (E2B-it, E4B-it, 31B-it, 26B-A4B-it)
     "gemma4",  # Gemma-4 alternate naming
 )
 
+# Architecture classes / model_type values that require transformers 5.5.0.
+# Checked via config.json (local or HuggingFace).
+_TRANSFORMERS_550_ARCHITECTURES: set[str] = {
+    "Gemma4ForConditionalGeneration",
+}
+_TRANSFORMERS_550_MODEL_TYPES: set[str] = {
+    "gemma4",
+}
+
 # Tokenizer classes that only exist in transformers>=5.x
 _TRANSFORMERS_5_TOKENIZER_CLASSES: set[str] = {
     "TokenizersBackend",
@@ -59,12 +77,61 @@ _TRANSFORMERS_5_TOKENIZER_CLASSES: set[str] = {
 # Cache for dynamic tokenizer_config.json lookups to avoid repeated fetches
 _tokenizer_class_cache: dict[str, bool] = {}
 
-# Versions
-TRANSFORMERS_5_VERSION = "5.5.0"
-TRANSFORMERS_DEFAULT_VERSION = "4.57.6"
+# Cache for dynamic config.json lookups (architecture/model_type checks)
+_config_needs_550_cache: dict[str, bool] = {}
 
-# Pre-installed directory for transformers 5.x — created by setup.sh / setup.ps1
-_VENV_T5_DIR = str(Path.home() / ".unsloth" / "studio" / ".venv_t5")
+# Versions
+TRANSFORMERS_550_VERSION = "5.5.0"
+TRANSFORMERS_530_VERSION = "5.3.0"
+TRANSFORMERS_DEFAULT_VERSION = "4.57.6"
+# Backwards-compat alias — points to 5.5.0 (the highest 5.x tier).
+# Consumers should prefer TRANSFORMERS_530_VERSION / TRANSFORMERS_550_VERSION.
+TRANSFORMERS_5_VERSION = TRANSFORMERS_550_VERSION
+
+# Pre-installed directories — created by setup.sh / setup.ps1
+_VENV_T5_530_DIR = str(Path.home() / ".unsloth" / "studio" / ".venv_t5_530")
+_VENV_T5_550_DIR = str(Path.home() / ".unsloth" / "studio" / ".venv_t5_550")
+# Backwards-compat alias
+_VENV_T5_DIR = _VENV_T5_550_DIR
+
+
+def activate_transformers_for_subprocess(model_name: str) -> None:
+    """Activate the correct transformers version in a subprocess worker.
+
+    Call this BEFORE any ML imports. Resolves LoRA adapters to their base
+    model, determines the required tier, and prepends the appropriate
+    ``.venv_t5_*`` directory to ``sys.path``.  Also propagates the path
+    via ``PYTHONPATH`` for child processes (e.g. GGUF converter).
+
+    Used by training, inference, and export workers.
+    """
+    resolved = _resolve_base_model(model_name)
+    tier = get_transformers_tier(resolved)
+
+    if tier == "550":
+        if not _ensure_venv_t5_550_exists():
+            raise RuntimeError(
+                f"Cannot activate transformers 5.5.0: "
+                f".venv_t5_550 missing at {_VENV_T5_550_DIR}"
+            )
+        if _VENV_T5_550_DIR not in sys.path:
+            sys.path.insert(0, _VENV_T5_550_DIR)
+        logger.info("Activated transformers 5.5.0 from %s", _VENV_T5_550_DIR)
+        _pp = os.environ.get("PYTHONPATH", "")
+        os.environ["PYTHONPATH"] = _VENV_T5_550_DIR + (os.pathsep + _pp if _pp else "")
+    elif tier == "530":
+        if not _ensure_venv_t5_530_exists():
+            raise RuntimeError(
+                f"Cannot activate transformers 5.3.0: "
+                f".venv_t5_530 missing at {_VENV_T5_530_DIR}"
+            )
+        if _VENV_T5_530_DIR not in sys.path:
+            sys.path.insert(0, _VENV_T5_530_DIR)
+        logger.info("Activated transformers 5.3.0 from %s", _VENV_T5_530_DIR)
+        _pp = os.environ.get("PYTHONPATH", "")
+        os.environ["PYTHONPATH"] = _VENV_T5_530_DIR + (os.pathsep + _pp if _pp else "")
+    else:
+        logger.info("Using default transformers (4.57.x) for %s", model_name)
 
 
 def _resolve_base_model(model_name: str) -> str:
@@ -192,18 +259,103 @@ def _check_tokenizer_config_needs_v5(model_name: str) -> bool:
         return False
 
 
-def needs_transformers_5(model_name: str) -> bool:
-    """Return True if *model_name* belongs to an architecture that requires
-    ``transformers>=5.3.0``.
+def _check_config_needs_550(model_name: str) -> bool:
+    """Check ``config.json`` for architectures or model_type that require
+    transformers 5.5.0 (e.g. Gemma 4).
 
-    First checks the hardcoded substring list for known models, then
-    dynamically fetches ``tokenizer_config.json`` from HuggingFace to check
-    if the tokenizer_class (e.g. ``TokenizersBackend``) requires v5.
+    Checks locally first, then falls back to fetching from HuggingFace.
+    Results are cached in ``_config_needs_550_cache``.
+    Returns False on any error (fail-open to lower tier).
+    """
+    if model_name in _config_needs_550_cache:
+        return _config_needs_550_cache[model_name]
+
+    def _check_cfg(cfg: dict) -> bool:
+        archs = cfg.get("architectures", [])
+        if any(a in _TRANSFORMERS_550_ARCHITECTURES for a in archs):
+            return True
+        if cfg.get("model_type") in _TRANSFORMERS_550_MODEL_TYPES:
+            return True
+        return False
+
+    # --- Check local config.json first ------------------------------------
+    local_path = Path(model_name)
+    local_cfg = local_path / "config.json"
+    if local_cfg.is_file():
+        try:
+            with open(local_cfg) as f:
+                cfg = json.load(f)
+            result = _check_cfg(cfg)
+            if result:
+                logger.info(
+                    "Local config.json check: %s needs transformers 5.5.0 "
+                    "(architectures=%s, model_type=%s)",
+                    model_name,
+                    cfg.get("architectures", []),
+                    cfg.get("model_type"),
+                )
+            _config_needs_550_cache[model_name] = result
+            return result
+        except Exception as exc:
+            logger.debug("Could not read %s: %s", local_cfg, exc)
+
+    # --- Fall back to fetching from HuggingFace ---------------------------
+    import urllib.request
+
+    url = f"https://huggingface.co/{model_name}/raw/main/config.json"
+    try:
+        req = urllib.request.Request(url, headers = {"User-Agent": "unsloth-studio"})
+        with urllib.request.urlopen(req, timeout = 10) as resp:
+            cfg = json.loads(resp.read().decode())
+        result = _check_cfg(cfg)
+        if result:
+            logger.info(
+                "Dynamic config.json check: %s needs transformers 5.5.0 "
+                "(architectures=%s, model_type=%s)",
+                model_name,
+                cfg.get("architectures", []),
+                cfg.get("model_type"),
+            )
+        _config_needs_550_cache[model_name] = result
+        return result
+    except Exception as exc:
+        logger.debug("Could not fetch config.json for '%s': %s", model_name, exc)
+        _config_needs_550_cache[model_name] = False
+        return False
+
+
+def get_transformers_tier(model_name: str) -> str:
+    """Return the transformers tier required for *model_name*.
+
+    Returns ``"550"`` for models needing transformers 5.5.0 (e.g. Gemma 4),
+    ``"530"`` for models needing transformers 5.3.0 (e.g. Ministral-3, Qwen3 MoE),
+    or ``"default"`` for everything else (4.57.x).
+
+    The 5.5.0 check runs first, then 5.3.0.
     """
     lowered = model_name.lower()
+
+    # --- Fast substring checks (no I/O) ------------------------------------
+    if any(sub in lowered for sub in TRANSFORMERS_550_MODEL_SUBSTRINGS):
+        return "550"
     if any(sub in lowered for sub in TRANSFORMERS_5_MODEL_SUBSTRINGS):
-        return True
-    return _check_tokenizer_config_needs_v5(model_name)
+        return "530"
+
+    # --- Slow config fallbacks (local file first, then network) -----------
+    if _check_config_needs_550(model_name):
+        return "550"
+    if _check_tokenizer_config_needs_v5(model_name):
+        return "530"
+
+    return "default"
+
+
+def needs_transformers_5(model_name: str) -> bool:
+    """Return True if *model_name* requires any transformers 5.x version.
+
+    Convenience wrapper around :func:`get_transformers_tier`.
+    """
+    return get_transformers_tier(model_name) != "default"
 
 
 # ---------------------------------------------------------------------------
@@ -258,27 +410,36 @@ def _purge_modules() -> int:
     return len(to_remove)
 
 
-_VENV_T5_PACKAGES = (
-    f"transformers=={TRANSFORMERS_5_VERSION}",
+_VENV_T5_530_PACKAGES = (
+    f"transformers=={TRANSFORMERS_530_VERSION}",
     "huggingface_hub==1.8.0",
     "hf_xet==1.4.2",
     "tiktoken",
 )
 
+_VENV_T5_550_PACKAGES = (
+    f"transformers=={TRANSFORMERS_550_VERSION}",
+    "huggingface_hub==1.8.0",
+    "hf_xet==1.4.2",
+    "tiktoken",
+)
 
-def _venv_t5_is_valid() -> bool:
-    """Return True if .venv_t5/ has all required packages at the correct versions."""
-    if not os.path.isdir(_VENV_T5_DIR) or not os.listdir(_VENV_T5_DIR):
+# Backwards-compat alias
+_VENV_T5_PACKAGES = _VENV_T5_550_PACKAGES
+
+
+def _venv_dir_is_valid(venv_dir: str, packages: tuple[str, ...]) -> bool:
+    """Return True if *venv_dir* has all *packages* at the correct versions."""
+    if not os.path.isdir(venv_dir) or not os.listdir(venv_dir):
         return False
-    # Check that the key package directories exist AND match the required version
-    for pkg_spec in _VENV_T5_PACKAGES:
+    for pkg_spec in packages:
         parts = pkg_spec.split("==")
         pkg_name = parts[0]
         pkg_version = parts[1] if len(parts) > 1 else None
         pkg_name_norm = pkg_name.replace("-", "_")
         # Check directory exists
         if not any(
-            (Path(_VENV_T5_DIR) / d).is_dir()
+            (Path(venv_dir) / d).is_dir()
             for d in (pkg_name_norm, pkg_name_norm.replace("_", "-"))
         ):
             return False
@@ -287,7 +448,7 @@ def _venv_t5_is_valid() -> bool:
             continue
         # Check version via .dist-info metadata
         dist_info_found = False
-        for di in Path(_VENV_T5_DIR).glob(f"{pkg_name_norm}-*.dist-info"):
+        for di in Path(venv_dir).glob(f"{pkg_name_norm}-*.dist-info"):
             metadata = di / "METADATA"
             if not metadata.is_file():
                 continue
@@ -296,7 +457,8 @@ def _venv_t5_is_valid() -> bool:
                     installed_ver = line.split(":", 1)[1].strip()
                     if installed_ver != pkg_version:
                         logger.info(
-                            ".venv_t5 has %s==%s but need %s",
+                            "%s has %s==%s but need %s",
+                            venv_dir,
                             pkg_name,
                             installed_ver,
                             pkg_version,
@@ -311,8 +473,13 @@ def _venv_t5_is_valid() -> bool:
     return True
 
 
-def _install_to_venv_t5(pkg: str) -> bool:
-    """Install a single package into .venv_t5/, preferring uv then pip."""
+def _venv_t5_is_valid() -> bool:
+    """Backwards-compat: check the 5.5.0 venv."""
+    return _venv_dir_is_valid(_VENV_T5_550_DIR, _VENV_T5_550_PACKAGES)
+
+
+def _install_to_dir(pkg: str, target_dir: str) -> bool:
+    """Install a single package into *target_dir*, preferring uv then pip."""
     # Try uv first (faster) if already on PATH -- do NOT install uv at runtime
     if shutil.which("uv"):
         result = subprocess.run(
@@ -323,7 +490,7 @@ def _install_to_venv_t5(pkg: str) -> bool:
                 "--python",
                 sys.executable,
                 "--target",
-                _VENV_T5_DIR,
+                target_dir,
                 "--no-deps",
                 "--upgrade",
                 pkg,
@@ -344,7 +511,7 @@ def _install_to_venv_t5(pkg: str) -> bool:
             "pip",
             "install",
             "--target",
-            _VENV_T5_DIR,
+            target_dir,
             "--no-deps",
             "--upgrade",
             pkg,
@@ -359,47 +526,62 @@ def _install_to_venv_t5(pkg: str) -> bool:
     return True
 
 
-def _ensure_venv_t5_exists() -> bool:
-    """Ensure .venv_t5/ exists with all required packages. Install if missing."""
-    if _venv_t5_is_valid():
+def _ensure_venv_dir(venv_dir: str, packages: tuple[str, ...], label: str) -> bool:
+    """Ensure *venv_dir* exists with all *packages*. Install if missing."""
+    if _venv_dir_is_valid(venv_dir, packages):
         return True
 
     logger.warning(
-        ".venv_t5 not found or incomplete at %s -- installing at runtime", _VENV_T5_DIR
+        "%s not found or incomplete at %s -- installing at runtime", label, venv_dir
     )
-    shutil.rmtree(_VENV_T5_DIR, ignore_errors = True)
-    os.makedirs(_VENV_T5_DIR, exist_ok = True)
-    for pkg in _VENV_T5_PACKAGES:
-        if not _install_to_venv_t5(pkg):
+    shutil.rmtree(venv_dir, ignore_errors = True)
+    os.makedirs(venv_dir, exist_ok = True)
+    for pkg in packages:
+        if not _install_to_dir(pkg, venv_dir):
             return False
-    logger.info("Installed transformers 5.x to %s", _VENV_T5_DIR)
+    logger.info("Installed %s to %s", label, venv_dir)
     return True
 
 
-def _activate_5x() -> None:
-    """Prepend .venv_t5/ to sys.path, purge stale modules, reimport."""
-    if not _ensure_venv_t5_exists():
-        raise RuntimeError(
-            f"Cannot activate transformers 5.x: .venv_t5 missing at {_VENV_T5_DIR}"
-        )
+def _ensure_venv_t5_530_exists() -> bool:
+    """Ensure .venv_t5_530/ exists with transformers 5.3.0."""
+    return _ensure_venv_dir(
+        _VENV_T5_530_DIR, _VENV_T5_530_PACKAGES, "transformers 5.3.0"
+    )
 
-    if _VENV_T5_DIR not in sys.path:
-        sys.path.insert(0, _VENV_T5_DIR)
-        logger.info("Prepended %s to sys.path", _VENV_T5_DIR)
+
+def _ensure_venv_t5_550_exists() -> bool:
+    """Ensure .venv_t5_550/ exists with transformers 5.5.0."""
+    return _ensure_venv_dir(
+        _VENV_T5_550_DIR, _VENV_T5_550_PACKAGES, "transformers 5.5.0"
+    )
+
+
+def _ensure_venv_t5_exists() -> bool:
+    """Backwards-compat: ensure the 5.5.0 venv exists."""
+    return _ensure_venv_t5_550_exists()
+
+
+def _activate_venv(venv_dir: str, label: str) -> None:
+    """Prepend *venv_dir* to sys.path, purge stale modules, reimport."""
+    if venv_dir not in sys.path:
+        sys.path.insert(0, venv_dir)
+        logger.info("Prepended %s to sys.path", venv_dir)
 
     count = _purge_modules()
     logger.info("Purged %d cached modules", count)
 
     import transformers
 
-    logger.info("Loaded transformers %s", transformers.__version__)
+    logger.info("Loaded transformers %s (%s)", transformers.__version__, label)
 
 
 def _deactivate_5x() -> None:
-    """Remove .venv_t5/ from sys.path, purge stale modules, reimport."""
-    while _VENV_T5_DIR in sys.path:
-        sys.path.remove(_VENV_T5_DIR)
-    logger.info("Removed %s from sys.path", _VENV_T5_DIR)
+    """Remove all .venv_t5_*/ dirs from sys.path, purge stale modules, reimport."""
+    for d in (_VENV_T5_530_DIR, _VENV_T5_550_DIR):
+        while d in sys.path:
+            sys.path.remove(d)
+    logger.info("Removed venv_t5 dirs from sys.path")
 
     count = _purge_modules()
     logger.info("Purged %d cached modules", count)
@@ -412,9 +594,10 @@ def _deactivate_5x() -> None:
 def ensure_transformers_version(model_name: str) -> None:
     """Ensure the correct ``transformers`` version is active for *model_name*.
 
-    Uses sys.path with .venv_t5/ (pre-installed by setup.sh):
-      • Need 5.x → prepend .venv_t5/ to sys.path, purge modules.
-      • Need 4.x → remove .venv_t5/ from sys.path, purge modules.
+    Uses sys.path with .venv_t5_530/ or .venv_t5_550/ (pre-installed by setup.sh):
+      • Need 5.5.0 → prepend .venv_t5_550/ to sys.path, purge modules.
+      • Need 5.3.0 → prepend .venv_t5_530/ to sys.path, purge modules.
+      • Need 4.x  → remove all .venv_t5_*/ from sys.path, purge modules.
 
     For LoRA adapters with custom names, the base model is resolved from
     ``adapter_config.json`` before checking.
@@ -424,8 +607,21 @@ def ensure_transformers_version(model_name: str) -> None:
     """
     # Resolve LoRA adapters to their base model for accurate detection
     resolved = _resolve_base_model(model_name)
-    want_5 = needs_transformers_5(resolved)
-    target_version = TRANSFORMERS_5_VERSION if want_5 else TRANSFORMERS_DEFAULT_VERSION
+    tier = get_transformers_tier(resolved)
+
+    if tier == "550":
+        target_version = TRANSFORMERS_550_VERSION
+        venv_dir = _VENV_T5_550_DIR
+        ensure_fn = _ensure_venv_t5_550_exists
+    elif tier == "530":
+        target_version = TRANSFORMERS_530_VERSION
+        venv_dir = _VENV_T5_530_DIR
+        ensure_fn = _ensure_venv_t5_530_exists
+    else:
+        target_version = TRANSFORMERS_DEFAULT_VERSION
+        venv_dir = None
+        ensure_fn = None
+
     target_major = int(target_version.split(".")[0])
 
     # Check what's actually loaded in memory
@@ -441,8 +637,17 @@ def ensure_transformers_version(model_name: str) -> None:
 
     # --- Already correct? ---------------------------------------------------
     if in_memory is not None:
+        if in_memory == target_version:
+            logger.info(
+                "transformers %s already loaded — correct for '%s'",
+                in_memory,
+                model_name,
+            )
+            return
+        # Different 5.x → need to switch (e.g. 5.3.0 loaded but need 5.5.0)
         in_memory_major = int(in_memory.split(".")[0])
-        if in_memory_major == target_major:
+        if in_memory_major == target_major and venv_dir is None:
+            # Both are default (4.x) — close enough
             logger.info(
                 "transformers %s already loaded — correct for '%s'",
                 in_memory,
@@ -451,9 +656,16 @@ def ensure_transformers_version(model_name: str) -> None:
             return
 
     # --- Switch version -----------------------------------------------------
-    if want_5:
-        logger.info("Activating transformers %s via .venv_t5…", TRANSFORMERS_5_VERSION)
-        _activate_5x()
+    if venv_dir is not None:
+        # First remove any other 5.x venv from sys.path
+        _deactivate_5x()
+        if not ensure_fn():
+            raise RuntimeError(
+                f"Cannot activate transformers {target_version}: "
+                f"venv missing at {venv_dir}"
+            )
+        logger.info("Activating transformers %s…", target_version)
+        _activate_venv(venv_dir, f"transformers {target_version}")
     else:
         logger.info(
             "Reverting to default transformers %s…", TRANSFORMERS_DEFAULT_VERSION
diff --git a/studio/setup.ps1 b/studio/setup.ps1
index 60dfce0661..c3a8cd71ca 100644
--- a/studio/setup.ps1
+++ b/studio/setup.ps1
@@ -1579,55 +1579,104 @@ if ($stackExit -ne 0) {
     exit 1
 }
 
-# ── Pre-install transformers 5.x into .venv_t5/ ──
-# Models like GLM-4.7-Flash need transformers>=5.3.0. Instead of pip-installing
-# at runtime (slow, ~10-15s), we pre-install into a separate directory.
-# The training subprocess just prepends .venv_t5/ to sys.path -- instant switch.
-Write-Host ""
-substep "pre-installing transformers 5.x for newer model support..."
-$VenvT5Dir = Join-Path $env:USERPROFILE ".unsloth\studio\.venv_t5"
-if (Test-Path $VenvT5Dir) { Remove-Item -Recurse -Force $VenvT5Dir }
-New-Item -ItemType Directory -Path $VenvT5Dir -Force | Out-Null
-$prevEAP_t5 = $ErrorActionPreference
-$ErrorActionPreference = "Continue"
-foreach ($pkg in @("transformers==5.5.0", "huggingface_hub==1.8.0", "hf_xet==1.4.2")) {
-    if ($script:UnslothVerbose) {
-        Fast-Install --target $VenvT5Dir --no-deps $pkg
-        $t5PkgExit = $LASTEXITCODE
-        $output = ""
-    } else {
-        $output = Fast-Install --target $VenvT5Dir --no-deps $pkg | Out-String
-        $t5PkgExit = $LASTEXITCODE
-    }
-    if ($t5PkgExit -ne 0) {
-        Write-Host "[FAIL] Could not install $pkg into .venv_t5/" -ForegroundColor Red
-        Write-Host $output -ForegroundColor Red
-        $ErrorActionPreference = $prevEAP_t5
-        exit 1
-    }
-}
-# tiktoken is needed by Qwen-family tokenizers -- install with deps since
-# regex/requests may be missing on Windows
-if ($script:UnslothVerbose) {
-    Fast-Install --target $VenvT5Dir tiktoken
-    $tiktokenInstallExit = $LASTEXITCODE
-    $output = ""
-} else {
-    $output = Fast-Install --target $VenvT5Dir tiktoken | Out-String
-    $tiktokenInstallExit = $LASTEXITCODE
-}
-if ($tiktokenInstallExit -ne 0) {
-    substep "Could not install tiktoken into .venv_t5/ -- Qwen tokenizers may fail" "Yellow"
-}
-$ErrorActionPreference = $prevEAP_t5
-step "transformers" "5.x pre-installed"
-
 } else {
     step "python" "dependencies up to date"
     # Restore ErrorActionPreference (was lowered for pip/python section)
     $ErrorActionPreference = $prevEAP
 }
 
+# ── Pre-install transformers 5.x into .venv_t5_530/ and .venv_t5_550/ ──
+# Runs outside the deps fast-path gate so that upgrades from the legacy
+# single .venv_t5 are always migrated to the tiered layout.
+$VenvT5_530Dir = Join-Path $env:USERPROFILE ".unsloth\studio\.venv_t5_530"
+$VenvT5_550Dir = Join-Path $env:USERPROFILE ".unsloth\studio\.venv_t5_550"
+$VenvT5Legacy = Join-Path $env:USERPROFILE ".unsloth\studio\.venv_t5"
+
+$_NeedT5Install = $false
+if (Test-Path $VenvT5Legacy) {
+    Remove-Item -Recurse -Force $VenvT5Legacy
+    $_NeedT5Install = $true
+}
+if (-not (Test-Path $VenvT5_530Dir)) { $_NeedT5Install = $true }
+if (-not (Test-Path $VenvT5_550Dir)) { $_NeedT5Install = $true }
+# Also reinstall when python deps were updated
+if (-not $SkipPythonDeps) { $_NeedT5Install = $true }
+
+if ($_NeedT5Install) {
+Write-Host ""
+
+$prevEAP_t5 = $ErrorActionPreference
+$ErrorActionPreference = "Continue"
+
+# --- .venv_t5_530 (transformers 5.3.0) ---
+substep "pre-installing transformers 5.3.0 for newer model support..."
+if (Test-Path $VenvT5_530Dir) { Remove-Item -Recurse -Force $VenvT5_530Dir }
+New-Item -ItemType Directory -Path $VenvT5_530Dir -Force | Out-Null
+foreach ($pkg in @("transformers==5.3.0", "huggingface_hub==1.8.0", "hf_xet==1.4.2")) {
+    if ($script:UnslothVerbose) {
+        Fast-Install --target $VenvT5_530Dir --no-deps $pkg
+        $t5PkgExit = $LASTEXITCODE
+        $output = ""
+    } else {
+        $output = Fast-Install --target $VenvT5_530Dir --no-deps $pkg | Out-String
+        $t5PkgExit = $LASTEXITCODE
+    }
+    if ($t5PkgExit -ne 0) {
+        Write-Host "[FAIL] Could not install $pkg into .venv_t5_530/" -ForegroundColor Red
+        Write-Host $output -ForegroundColor Red
+        $ErrorActionPreference = $prevEAP_t5
+        exit 1
+    }
+}
+if ($script:UnslothVerbose) {
+    Fast-Install --target $VenvT5_530Dir tiktoken
+    $tiktokenInstallExit = $LASTEXITCODE
+    $output = ""
+} else {
+    $output = Fast-Install --target $VenvT5_530Dir tiktoken | Out-String
+    $tiktokenInstallExit = $LASTEXITCODE
+}
+if ($tiktokenInstallExit -ne 0) {
+    substep "Could not install tiktoken into .venv_t5_530/ -- Qwen tokenizers may fail" "Yellow"
+}
+step "transformers" "5.3.0 pre-installed"
+
+# --- .venv_t5_550 (transformers 5.5.0) ---
+substep "pre-installing transformers 5.5.0 for Gemma 4 support..."
+if (Test-Path $VenvT5_550Dir) { Remove-Item -Recurse -Force $VenvT5_550Dir }
+New-Item -ItemType Directory -Path $VenvT5_550Dir -Force | Out-Null
+foreach ($pkg in @("transformers==5.5.0", "huggingface_hub==1.8.0", "hf_xet==1.4.2")) {
+    if ($script:UnslothVerbose) {
+        Fast-Install --target $VenvT5_550Dir --no-deps $pkg
+        $t5PkgExit = $LASTEXITCODE
+        $output = ""
+    } else {
+        $output = Fast-Install --target $VenvT5_550Dir --no-deps $pkg | Out-String
+        $t5PkgExit = $LASTEXITCODE
+    }
+    if ($t5PkgExit -ne 0) {
+        Write-Host "[FAIL] Could not install $pkg into .venv_t5_550/" -ForegroundColor Red
+        Write-Host $output -ForegroundColor Red
+        $ErrorActionPreference = $prevEAP_t5
+        exit 1
+    }
+}
+if ($script:UnslothVerbose) {
+    Fast-Install --target $VenvT5_550Dir tiktoken
+    $tiktokenInstallExit = $LASTEXITCODE
+    $output = ""
+} else {
+    $output = Fast-Install --target $VenvT5_550Dir tiktoken | Out-String
+    $tiktokenInstallExit = $LASTEXITCODE
+}
+if ($tiktokenInstallExit -ne 0) {
+    substep "Could not install tiktoken into .venv_t5_550/ -- Qwen tokenizers may fail" "Yellow"
+}
+$ErrorActionPreference = $prevEAP_t5
+step "transformers" "5.5.0 pre-installed"
+
+} # end $_NeedT5Install
+
 # ==========================================================================
 #  PHASE 3.4: Prefer prebuilt llama.cpp bundles before source build
 # ==========================================================================
diff --git a/studio/setup.sh b/studio/setup.sh
index e3ff2da35c..eb89e583ca 100755
--- a/studio/setup.sh
+++ b/studio/setup.sh
@@ -394,11 +394,14 @@ fi
 # ── Python venv + deps ──
 STUDIO_HOME="$HOME/.unsloth/studio"
 VENV_DIR="$STUDIO_HOME/unsloth_studio"
-VENV_T5_DIR="$STUDIO_HOME/.venv_t5"
+VENV_T5_530_DIR="$STUDIO_HOME/.venv_t5_530"
+VENV_T5_550_DIR="$STUDIO_HOME/.venv_t5_550"
 
 [ -d "$REPO_ROOT/.venv" ] && rm -rf "$REPO_ROOT/.venv"
 [ -d "$REPO_ROOT/.venv_overlay" ] && rm -rf "$REPO_ROOT/.venv_overlay"
 [ -d "$REPO_ROOT/.venv_t5" ] && rm -rf "$REPO_ROOT/.venv_t5"
+[ -d "$REPO_ROOT/.venv_t5_530" ] && rm -rf "$REPO_ROOT/.venv_t5_530"
+[ -d "$REPO_ROOT/.venv_t5_550" ] && rm -rf "$REPO_ROOT/.venv_t5_550"
 # Note: do NOT delete $STUDIO_HOME/.venv here — install.sh handles migration
 
 _COLAB_NO_VENV=false
@@ -501,21 +504,47 @@ fi
 
 if [ "$_SKIP_PYTHON_DEPS" = false ]; then
     install_python_stack
-
-    # ── 6b. Pre-install transformers 5.x into .venv_t5/ ──
-    # Models like GLM-4.7-Flash need transformers>=5.3.0. Instead of pip-installing
-    # at runtime (slow, ~10-15s), we pre-install into a separate directory.
-    # The training subprocess just prepends .venv_t5/ to sys.path -- instant switch.
-    mkdir -p "$VENV_T5_DIR"
-    run_quiet "install transformers 5.x" fast_install --target "$VENV_T5_DIR" --no-deps "transformers==5.5.0"
-    run_quiet "install huggingface_hub for t5" fast_install --target "$VENV_T5_DIR" --no-deps "huggingface_hub==1.8.0"
-    run_quiet "install hf_xet for t5" fast_install --target "$VENV_T5_DIR" --no-deps "hf_xet==1.4.2"
-    run_quiet "install tiktoken for t5" fast_install --target "$VENV_T5_DIR" "tiktoken"
-    step "transformers" "5.x pre-installed"
 else
     step "python" "dependencies up to date"
     verbose_substep "python deps check: installed=$_PKG_NAME@${INSTALLED_VER:-unknown} latest=${LATEST_VER:-unknown}"
 fi
+
+# ── 6b. Pre-install transformers 5.x into .venv_t5_530/ and .venv_t5_550/ ──
+# Models like GLM-4.7-Flash, Qwen3 MoE need transformers>=5.3.0.
+# Gemma 4 models need transformers>=5.5.0.
+# Pre-install into separate directories to avoid runtime pip overhead.
+# The training subprocess prepends the appropriate dir to sys.path.
+#
+# Runs outside the _SKIP_PYTHON_DEPS gate so that upgrades from legacy
+# single .venv_t5 are always migrated to the tiered layout.
+_NEED_T5_INSTALL=false
+if [ -d "$STUDIO_HOME/.venv_t5" ]; then
+    # Legacy layout — migrate
+    rm -rf "$STUDIO_HOME/.venv_t5"
+    _NEED_T5_INSTALL=true
+fi
+[ ! -d "$VENV_T5_530_DIR" ] && _NEED_T5_INSTALL=true
+[ ! -d "$VENV_T5_550_DIR" ] && _NEED_T5_INSTALL=true
+# Also reinstall when python deps were updated (packages may need rebuild)
+[ "$_SKIP_PYTHON_DEPS" = false ] && _NEED_T5_INSTALL=true
+
+if [ "$_NEED_T5_INSTALL" = true ]; then
+    [ -d "$VENV_T5_530_DIR" ] && rm -rf "$VENV_T5_530_DIR"
+    mkdir -p "$VENV_T5_530_DIR"
+    run_quiet "install transformers 5.3.0" fast_install --target "$VENV_T5_530_DIR" --no-deps "transformers==5.3.0"
+    run_quiet "install huggingface_hub for t5_530" fast_install --target "$VENV_T5_530_DIR" --no-deps "huggingface_hub==1.8.0"
+    run_quiet "install hf_xet for t5_530" fast_install --target "$VENV_T5_530_DIR" --no-deps "hf_xet==1.4.2"
+    run_quiet "install tiktoken for t5_530" fast_install --target "$VENV_T5_530_DIR" "tiktoken"
+    step "transformers" "5.3.0 pre-installed"
+
+    [ -d "$VENV_T5_550_DIR" ] && rm -rf "$VENV_T5_550_DIR"
+    mkdir -p "$VENV_T5_550_DIR"
+    run_quiet "install transformers 5.5.0" fast_install --target "$VENV_T5_550_DIR" --no-deps "transformers==5.5.0"
+    run_quiet "install huggingface_hub for t5_550" fast_install --target "$VENV_T5_550_DIR" --no-deps "huggingface_hub==1.8.0"
+    run_quiet "install hf_xet for t5_550" fast_install --target "$VENV_T5_550_DIR" --no-deps "hf_xet==1.4.2"
+    run_quiet "install tiktoken for t5_550" fast_install --target "$VENV_T5_550_DIR" "tiktoken"
+    step "transformers" "5.5.0 pre-installed"
+fi
 fi
 
 # ── 7. Prefer prebuilt llama.cpp bundles before any source build path ──

From c2184af07985b30eecdd41480d52dfafefc1e2ec Mon Sep 17 00:00:00 2001
From: "pre-commit-ci[bot]"
 <66853113+pre-commit-ci[bot]@users.noreply.github.com>
Date: Tue, 7 Apr 2026 22:50:48 -0700
Subject: [PATCH 056/216] [pre-commit.ci] pre-commit autoupdate (#4879)
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

updates:
- [github.com/astral-sh/ruff-pre-commit: v0.15.8 → v0.15.9](https://github.com/astral-sh/ruff-pre-commit/compare/v0.15.8...v0.15.9)

Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com>
---
 .pre-commit-config.yaml | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/.pre-commit-config.yaml b/.pre-commit-config.yaml
index f3b5987a9c..e41c37d209 100644
--- a/.pre-commit-config.yaml
+++ b/.pre-commit-config.yaml
@@ -1,6 +1,6 @@
 repos:
   - repo: https://github.com/astral-sh/ruff-pre-commit
-    rev: v0.15.8
+    rev: v0.15.9
     hooks:
       - id: ruff
         args:

From 67e9db4921e16562497dfd5c873e8bd01970906b Mon Sep 17 00:00:00 2001
From: "dependabot[bot]" <49699333+dependabot[bot]@users.noreply.github.com>
Date: Wed, 8 Apr 2026 03:35:33 -0700
Subject: [PATCH 057/216] build(deps): bump oxc-parser (#4776)

Bumps the npm-oxc-validator group in /studio/backend/core/data_recipe/oxc-validator with 1 update: [oxc-parser](https://github.com/oxc-project/oxc/tree/HEAD/napi/parser).


Updates `oxc-parser` from 0.121.0 to 0.123.0
- [Release notes](https://github.com/oxc-project/oxc/releases)
- [Changelog](https://github.com/oxc-project/oxc/blob/main/napi/parser/CHANGELOG.md)
- [Commits](https://github.com/oxc-project/oxc/commits/crates_v0.123.0/napi/parser)

---
updated-dependencies:
- dependency-name: oxc-parser
  dependency-version: 0.123.0
  dependency-type: direct:production
  update-type: version-update:semver-minor
  dependency-group: npm-oxc-validator
...

Signed-off-by: dependabot[bot] <support@github.com>
Co-authored-by: dependabot[bot] <49699333+dependabot[bot]@users.noreply.github.com>
---
 studio/backend/core/data_recipe/oxc-validator/package.json | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/studio/backend/core/data_recipe/oxc-validator/package.json b/studio/backend/core/data_recipe/oxc-validator/package.json
index d1c765a2e1..111ae2b257 100644
--- a/studio/backend/core/data_recipe/oxc-validator/package.json
+++ b/studio/backend/core/data_recipe/oxc-validator/package.json
@@ -4,7 +4,7 @@
   "version": "0.0.1",
   "type": "module",
   "dependencies": {
-    "oxc-parser": "^0.121.0",
+    "oxc-parser": "^0.123.0",
     "oxlint": "^1.51.0"
   }
 }

From 0087515d5cd5ffd0eb469c15c01a7d0aac3a0040 Mon Sep 17 00:00:00 2001
From: "dependabot[bot]" <49699333+dependabot[bot]@users.noreply.github.com>
Date: Wed, 8 Apr 2026 03:35:40 -0700
Subject: [PATCH 058/216] build(deps): bump oxc-parser (#4776)

Bumps the npm-oxc-validator group in /studio/backend/core/data_recipe/oxc-validator with 1 update: [oxc-parser](https://github.com/oxc-project/oxc/tree/HEAD/napi/parser).


Updates `oxc-parser` from 0.121.0 to 0.123.0
- [Release notes](https://github.com/oxc-project/oxc/releases)
- [Changelog](https://github.com/oxc-project/oxc/blob/main/napi/parser/CHANGELOG.md)
- [Commits](https://github.com/oxc-project/oxc/commits/crates_v0.123.0/napi/parser)

---
updated-dependencies:
- dependency-name: oxc-parser
  dependency-version: 0.123.0
  dependency-type: direct:production
  update-type: version-update:semver-minor
  dependency-group: npm-oxc-validator
...

Signed-off-by: dependabot[bot] <support@github.com>
Co-authored-by: dependabot[bot] <49699333+dependabot[bot]@users.noreply.github.com>

From c3d2d58046ebf5d706d038173ddf661436c4e861 Mon Sep 17 00:00:00 2001
From: Daniel Han <danielhanchen@gmail.com>
Date: Wed, 8 Apr 2026 03:39:50 -0700
Subject: [PATCH 059/216] Update dependabot.yml (#4915)

---
 .github/dependabot.yml | 13 -------------
 1 file changed, 13 deletions(-)

diff --git a/.github/dependabot.yml b/.github/dependabot.yml
index a06cb1d114..4a0bfa70f1 100644
--- a/.github/dependabot.yml
+++ b/.github/dependabot.yml
@@ -9,19 +9,6 @@ updates:
       actions:
         patterns: ["*"]
 
-  - package-ecosystem: "pip"
-    directories:
-      - "/"
-      - "/studio/backend/plugins/data-designer-unstructured-seed"
-      - "/studio/backend/requirements"
-      - "/unsloth/kernels/moe"
-    schedule:
-      interval: "weekly"
-    open-pull-requests-limit: 10
-    groups:
-      pip:
-        patterns: ["*"]
-
   - package-ecosystem: "bun"
     directory: "/studio/frontend"
     schedule:

From 8e977445d43350e3bc74b67544a5fbdbde29bb52 Mon Sep 17 00:00:00 2001
From: Wasim Yousef Said <wasimysdev@gmail.com>
Date: Wed, 8 Apr 2026 12:48:22 +0200
Subject: [PATCH 060/216] Let recipes use the model loaded in Chat (#4840)

* feat: inject local model provider into recipe jobs via JWT

* feat: auto-generate JWT for local model providers in recipes

* feat: add is_local flag to model provider config types and utils

* fix(studio): skip endpoint validation for local providers

* feat(studio): add local/external model source toggle to provider dialog

* feat(studio): thread localProviderNames through model config dialog chain

* feat(studio): show 'Local model (Chat)' label for local model_provider configs

* fix: hardcode loopback for local endpoint, clear stale creds on toggle

* fix: document TOCTOU/JWT rotation, add deferred import comments, fix is_local serialization

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

* fix(studio): clear stale local model state on provider toggle and validation

* fix(studio): override empty local endpoint in validation and skip model gate for unused providers

* fix(studio): resolve loopback port from app.state, clear stale local provider fields, sync model id on toggle

Address review feedback on the local-model-provider flow:

- Backend (jobs.py): _resolve_local_v1_endpoint now reads the actual bound
  port from app.state.server_port (set in run.py after binding) instead of
  parsing it out of request.base_url, which is wrong behind any reverse
  proxy or non-default port. The two duplicated urlparse blocks are gone.
- Backend (jobs.py): defensively pop api_key_env, extra_headers, extra_body
  from local providers so a previously external provider that flipped to
  local cannot leak invalid JSON or rogue auth headers into the local /v1
  call. Also dedupe the post-loop assignment and tighten the local-name
  intersection so empty names cannot match.
- Backend (jobs.py): hoist datetime and urllib.parse imports to the top
  import block for consistency with the rest of the file.
- Backend (run.py): expose the bound port on app.state.server_port after
  the uvicorn server is constructed.
- Frontend (model-provider-dialog.tsx): clear extra_headers and extra_body
  when toggling to local mode. Hidden inputs would otherwise keep stale
  JSON blocking validate/run.
- Frontend (model-config-dialog.tsx): factor the local-aware provider
  selection logic into applyProviderChange and call it from both
  onValueChange and onBlur, so manually typing a provider name and tabbing
  away keeps the model field consistent.
- Frontend (recipe-studio.ts store): handle both directions of the
  is_local toggle in the cascade. external -> local now backfills
  model: "local" on already-linked model_configs so they pass validation
  immediately, mirroring the existing local -> external clear path.
- Frontend (validate.ts + build-payload.ts): thread localProviderNames
  into validateModelConfigProviders and skip the "model is required"
  check for local-linked configs. Local providers do not need a real
  model id since the inference endpoint uses the loaded Chat model.

* fix(studio): narrow store cascade types, sync model placeholder on graph relink and node removal, harden ephemeral port path

Loop 2 review fixes:

- recipe-studio.ts: type-narrow next.is_local by also checking
  next.kind === "model_provider". TS otherwise raised TS2339 because
  next was typed as the union NodeConfig after the spread. The behavior
  is unchanged but the code now compiles cleanly.
- model-config-dialog.tsx: convert the lastProviderRef / providerInputRef
  ref-during-render pattern (pre-existing react-hooks/refs lint error)
  to a useEffect that syncs providerInputRef from config.provider. The
  combobox blur path still uses applyProviderChange and remains stable.
- recipe-graph-connection.ts: when a graph drag links a model_provider
  to a model_config, mirror the dialog applyProviderChange behavior:
  fill model: "local" if the new provider is local and the model field
  is blank, clear model when relinking from a local placeholder to an
  external provider, otherwise leave the model alone.
- reference-sync.ts: when a referenced provider node is removed, clear
  the synthetic model: "local" placeholder along with the provider
  field, so a future relink to an external provider does not pass
  validation with a stale value that fails at runtime.
- run.py: only publish app.state.server_port when the bound port is a
  real positive integer; for ephemeral binds (port==0) leave it unset
  and let request handlers fall back to request.base_url.
- jobs.py: _resolve_local_v1_endpoint also falls back when
  app.state.server_port is non-positive, and uses `is None` instead of
  the truthy fallback so a literal 0 is handled correctly.

* fix(studio): strict is_local check, narrow loaded-model gate to LLM-reachable configs, add scope-server port fallback

Loop 3 review fixes:

- jobs.py, validate.py: require `is_local is True` instead of truthy
  check. Malformed payloads such as is_local: "false" or is_local: 1
  would otherwise be treated as local and silently rewritten to the
  loopback endpoint.
- jobs.py: _resolve_local_v1_endpoint now tries request.scope["server"]
  (the actual uvicorn-assigned (host, port) tuple) as a second
  resolution step before falling back to parsing request.base_url.
  This covers direct-uvicorn startup paths and ephemeral binds that
  never publish app.state.server_port.
- jobs.py: new _used_llm_model_aliases helper collects the set of
  model_aliases that an LLM column actually references, and the
  "Chat model loaded" gate is now only triggered when a local
  provider is reachable from that set. Orphan model_config nodes on
  the canvas no longer block unrelated recipe runs.

* fix(studio): force skip_health_check on local-linked configs, skip JSON parsing for local providers, local-aware inline editor

Loop 4 review fixes:

- jobs.py: after rewriting local providers, also force
  skip_health_check: true on any model_config linked to a local
  provider. The /v1/models endpoint only advertises the real loaded
  model id, so data_designer's default model-availability health check
  would otherwise fail against the placeholder "local" id before the
  first chat completion call. The inference route already ignores the
  model id in chat completions, so skipping the check is safe.
- builders-model.ts: buildModelProvider now short-circuits for local
  providers and emits only { name, endpoint: "", provider_type, is_local }
  without running parseJsonObject on the hidden extra_headers/extra_body
  inputs. Imported or hydrated recipes with stale invalid JSON in those
  fields no longer block client-side validate/run.
- inline-model.tsx: the model_config branch now accepts an optional
  localProviderNames prop and mirrors the dialog applyProviderChange
  behavior. Changing provider to/from a local one auto-fills or clears
  the "local" placeholder consistently with the other edit paths.
- recipe-graph-node.tsx: derive localProviderNames from the store via
  useMemo (stable identity) and pass it through renderNodeBody to
  <InlineModel>. Hooks order is preserved by declaring them above the
  early return for markdown_note nodes.
- run.py: minor comment tweak - loop 3 already added the scope-server
  fallback path, note that in the comment.

---------

Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com>
Co-authored-by: danielhanchen <info@unsloth.ai>
---
 studio/backend/routes/data_recipe/jobs.py     | 164 ++++++++++++-
 studio/backend/routes/data_recipe/validate.py |  16 ++
 studio/backend/run.py                         |   8 +
 .../recipe-studio/blocks/render-dialog.tsx    |   2 +
 .../components/inline/inline-model.tsx        |  39 +++-
 .../components/recipe-graph-node.tsx          |  28 ++-
 .../recipe-studio/dialogs/config-dialog.tsx   |   3 +
 .../dialogs/models/model-config-dialog.tsx    |  38 ++-
 .../dialogs/models/model-provider-dialog.tsx  | 218 ++++++++++++------
 .../recipe-studio/recipe-studio-page.tsx      |   1 +
 .../stores/helpers/reference-sync.ts          |  10 +-
 .../recipe-studio/stores/recipe-studio.ts     |  28 +++
 .../src/features/recipe-studio/types/index.ts |   2 +
 .../recipe-studio/utils/config-factories.ts   |   2 +
 .../utils/graph/recipe-graph-connection.ts    |  12 +-
 .../utils/import/parsers/model-parser.ts      |   2 +
 .../utils/payload/build-payload.ts            |   5 +
 .../utils/payload/builders-model.ts           |  20 ++
 .../recipe-studio/utils/payload/validate.ts   |   9 +-
 .../recipe-studio/utils/recipe-studio-view.ts |   6 +
 20 files changed, 518 insertions(+), 95 deletions(-)

diff --git a/studio/backend/routes/data_recipe/jobs.py b/studio/backend/routes/data_recipe/jobs.py
index 1d5eceee03..00546b47a4 100644
--- a/studio/backend/routes/data_recipe/jobs.py
+++ b/studio/backend/routes/data_recipe/jobs.py
@@ -5,7 +5,9 @@
 
 from __future__ import annotations
 
+from datetime import timedelta
 from typing import Any
+from urllib.parse import urlparse
 
 from fastapi import APIRouter, HTTPException, Query, Request
 from fastapi.responses import JSONResponse, StreamingResponse
@@ -26,6 +28,161 @@ from models.data_recipe import (
 router = APIRouter()
 
 
+def _resolve_local_v1_endpoint(request: Request) -> str:
+    """Return the loopback /v1 URL for the actual backend listen port.
+
+    Resolution order:
+      1. ``app.state.server_port`` - explicitly published by run.py after
+         the uvicorn server has bound. This is the most reliable source
+         because it survives reverse proxies, TLS terminators and tunnels.
+      2. ``request.scope["server"]`` - the real (host, port) tuple uvicorn
+         sets when the request is dispatched. Used when Studio is started
+         outside ``run_server`` (e.g. ``uvicorn studio.backend.main:app``).
+      3. ``request.base_url`` parsed - last resort for test fixtures that
+         do not route through a live uvicorn server.
+    """
+    port: Any = getattr(request.app.state, "server_port", None)
+    if not isinstance(port, int) or port <= 0:
+        server = request.scope.get("server")
+        if (
+            isinstance(server, tuple)
+            and len(server) >= 2
+            and isinstance(server[1], int)
+            and server[1] > 0
+        ):
+            port = server[1]
+        else:
+            parsed = urlparse(str(request.base_url))
+            port = parsed.port if parsed.port is not None else 8888
+    return f"http://127.0.0.1:{int(port)}/v1"
+
+
+def _used_llm_model_aliases(recipe: dict[str, Any]) -> set[str]:
+    """Return the set of model_aliases that are actually referenced by an
+    LLM column. Used to narrow the "Chat model loaded" gate so that orphan
+    model_config nodes on the canvas do not block unrelated recipe runs.
+
+    The ``llm-`` prefix matches the existing convention in
+    ``core/data_recipe/service.py::_recipe_has_llm_columns`` and covers all
+    LLM column types emitted by the frontend (llm-text, llm-code,
+    llm-structured, llm-judge).
+    """
+    aliases: set[str] = set()
+    for column in recipe.get("columns", []):
+        if not isinstance(column, dict):
+            continue
+        column_type = column.get("column_type")
+        if not isinstance(column_type, str) or not column_type.startswith("llm-"):
+            continue
+        alias = column.get("model_alias")
+        if isinstance(alias, str) and alias:
+            aliases.add(alias)
+    return aliases
+
+
+def _inject_local_providers(recipe: dict[str, Any], request: Request) -> None:
+    """
+    Mutate recipe dict in-place: for any provider with is_local=True,
+    generate a JWT and fill in the endpoint pointing at this server.
+    """
+    providers = recipe.get("model_providers")
+    if not providers:
+        return
+
+    # Collect local providers and pop is_local from ALL dicts unconditionally.
+    # Strict `is True` guard so malformed payloads (is_local: 1,
+    # is_local: "true") do not accidentally trigger the loopback rewrite.
+    local_indices: list[int] = []
+    for i, provider in enumerate(providers):
+        if not isinstance(provider, dict):
+            continue
+        is_local = provider.pop("is_local", None)
+        if is_local is True:
+            local_indices.append(i)
+
+    if not local_indices:
+        return
+
+    endpoint = _resolve_local_v1_endpoint(request)
+
+    # Only gate on model-loaded if a local provider is actually reachable
+    # from an LLM column through a model_config. Orphan model_config nodes
+    # that reference a local provider but that no LLM column uses should
+    # not block runs; the recipe would never call /v1 for them.
+    local_names = {
+        providers[i].get("name") for i in local_indices if providers[i].get("name")
+    }
+    used_aliases = _used_llm_model_aliases(recipe)
+    referenced_providers = {
+        mc.get("provider")
+        for mc in recipe.get("model_configs", [])
+        if (
+            isinstance(mc, dict)
+            and mc.get("provider")
+            and mc.get("alias") in used_aliases
+        )
+    }
+
+    token = ""
+    if local_names & referenced_providers:
+        # Verify a model is loaded.
+        # NOTE: This is a point-in-time check (TOCTOU). The model could be unloaded
+        # or swapped after this check but before the recipe subprocess calls /v1.
+        # The inference endpoint returns a clear 400 in that case.
+        #
+        # Imports are deferred to avoid circular dependencies with inference modules.
+        from routes.inference import get_llama_cpp_backend
+        from core.inference import get_inference_backend
+
+        llama = get_llama_cpp_backend()
+        model_loaded = llama.is_loaded
+        if not model_loaded:
+            backend = get_inference_backend()
+            model_loaded = bool(backend.active_model_name)
+        if not model_loaded:
+            raise ValueError(
+                "No model loaded in Chat. Load a model first, then run the recipe."
+            )
+
+        from auth.authentication import (
+            create_access_token,
+        )  # deferred: avoids circular import
+
+        # Uses the "unsloth" admin subject. If the user changes their password,
+        # the JWT secret rotates and this token becomes invalid mid-run.
+        # Acceptable for v1 - recipes typically finish well within one session.
+        token = create_access_token(
+            subject = "unsloth",
+            expires_delta = timedelta(hours = 24),
+        )
+
+    # Defensively strip any stale "external"-only fields the frontend may
+    # have left on the dict (extra_headers/extra_body/api_key_env). The UI
+    # hides these inputs in local mode but the payload builder still serializes
+    # them, so a previously external provider that flipped to local can carry
+    # invalid JSON or rogue auth headers into the local /v1 call.
+    for i in local_indices:
+        providers[i]["endpoint"] = endpoint
+        providers[i]["api_key"] = token
+        providers[i]["provider_type"] = "openai"
+        providers[i].pop("api_key_env", None)
+        providers[i].pop("extra_headers", None)
+        providers[i].pop("extra_body", None)
+
+    # Force skip_health_check on any model_config that references a local
+    # provider. The local /v1/models endpoint only lists the real loaded
+    # model (e.g. "unsloth/llama-3.2-1b") and not the placeholder "local"
+    # that the recipe sends as the model id, so data_designer's pre-flight
+    # health check would otherwise fail before the first completion call.
+    # The backend route ignores the model id field in chat completions, so
+    # skipping the check is safe.
+    for mc in recipe.get("model_configs", []):
+        if not isinstance(mc, dict):
+            continue
+        if mc.get("provider") in local_names:
+            mc["skip_health_check"] = True
+
+
 def _normalize_run_name(value: Any) -> str | None:
     if value is None:
         return None
@@ -40,7 +197,7 @@ def _normalize_run_name(value: Any) -> str | None:
 
 
 @router.post("/jobs", response_class = JSONResponse, response_model = JobCreateResponse)
-def create_job(payload: RecipePayload):
+def create_job(payload: RecipePayload, request: Request):
     recipe = payload.recipe
     if not recipe.get("columns"):
         raise HTTPException(status_code = 400, detail = "Recipe must include columns.")
@@ -67,6 +224,11 @@ def create_job(payload: RecipePayload):
                 status_code = 400, detail = f"invalid run_config: {exc}"
             ) from exc
 
+    try:
+        _inject_local_providers(recipe, request)
+    except ValueError as exc:
+        raise HTTPException(status_code = 400, detail = str(exc)) from exc
+
     mgr = get_job_manager()
     try:
         job_id = mgr.start(recipe = recipe, run = run)
diff --git a/studio/backend/routes/data_recipe/validate.py b/studio/backend/routes/data_recipe/validate.py
index a793a3b172..555e3eaa06 100644
--- a/studio/backend/routes/data_recipe/validate.py
+++ b/studio/backend/routes/data_recipe/validate.py
@@ -68,6 +68,20 @@ def _collect_validation_errors(recipe: dict[str, Any]) -> list[ValidateError]:
     return errors
 
 
+def _patch_local_providers(recipe: dict[str, Any]) -> None:
+    """Strip is_local and fill a dummy endpoint so validation doesn't choke.
+
+    Uses a strict `is True` check to match _inject_local_providers in
+    jobs.py - malformed payloads with truthy but non-boolean is_local
+    values should not be treated as local.
+    """
+    for provider in recipe.get("model_providers", []):
+        if not isinstance(provider, dict):
+            continue
+        if provider.pop("is_local", None) is True:
+            provider["endpoint"] = "http://127.0.0.1"
+
+
 @router.post("/validate", response_model = ValidateResponse)
 def validate(payload: RecipePayload) -> ValidateResponse:
     recipe = payload.recipe
@@ -77,6 +91,8 @@ def validate(payload: RecipePayload) -> ValidateResponse:
             errors = [ValidateError(message = "Recipe must include columns.")],
         )
 
+    _patch_local_providers(recipe)
+
     try:
         validate_recipe(recipe)
     except RuntimeError as exc:
diff --git a/studio/backend/run.py b/studio/backend/run.py
index 9c3622988e..86c1194661 100644
--- a/studio/backend/run.py
+++ b/studio/backend/run.py
@@ -324,6 +324,14 @@ def run_server(
     _server = uvicorn.Server(config)
     _shutdown_event = Event()
 
+    # Expose the actual bound port so request-handling code can build
+    # loopback URLs that point at the real backend, not whatever port a
+    # reverse proxy or tunnel exposed in the request URL. Only publish
+    # an explicit value when we know the concrete port; for ephemeral
+    # binds (port==0) leave it unset and let request handlers fall back
+    # to the ASGI request scope or request.base_url.
+    app.state.server_port = port if port and port > 0 else None
+
     # Run server in a daemon thread
     def _run():
         asyncio.run(_server.serve())
diff --git a/studio/frontend/src/features/recipe-studio/blocks/render-dialog.tsx b/studio/frontend/src/features/recipe-studio/blocks/render-dialog.tsx
index 10fcaa489d..92f72dfff1 100644
--- a/studio/frontend/src/features/recipe-studio/blocks/render-dialog.tsx
+++ b/studio/frontend/src/features/recipe-studio/blocks/render-dialog.tsx
@@ -28,6 +28,7 @@ export function renderBlockDialog(
   categoryOptions: SamplerConfig[],
   modelConfigAliases: string[],
   modelProviderOptions: string[],
+  localProviderNames: Set<string>,
   toolProfileAliases: string[],
   datetimeOptions: string[],
   onUpdate: (id: string, patch: Partial<NodeConfig>) => void,
@@ -109,6 +110,7 @@ export function renderBlockDialog(
         <ModelConfigDialog
           config={config}
           providerOptions={modelProviderOptions}
+          localProviderNames={localProviderNames}
           onUpdate={update}
         />
       ) : null;
diff --git a/studio/frontend/src/features/recipe-studio/components/inline/inline-model.tsx b/studio/frontend/src/features/recipe-studio/components/inline/inline-model.tsx
index d3ca22b00b..16e99f4fae 100644
--- a/studio/frontend/src/features/recipe-studio/components/inline/inline-model.tsx
+++ b/studio/frontend/src/features/recipe-studio/components/inline/inline-model.tsx
@@ -10,11 +10,21 @@ type InlineModelPatch = Partial<ModelProviderConfig> | Partial<ModelConfig>;
 
 type InlineModelProps = {
   config: ModelProviderConfig | ModelConfig;
+  localProviderNames?: Set<string>;
   onUpdate: (patch: InlineModelPatch) => void;
 };
 
 export function InlineModel(props: InlineModelProps): ReactElement {
   if (props.config.kind === "model_provider") {
+    if (props.config.is_local) {
+      return (
+        <div className="flex items-center gap-2 px-1 py-0.5">
+          <span className="text-xs font-medium text-muted-foreground">
+            Local model (Chat)
+          </span>
+        </div>
+      );
+    }
     return (
       <div className="grid gap-3 sm:grid-cols-2">
         <InlineField label="Endpoint">
@@ -42,21 +52,40 @@ export function InlineModel(props: InlineModelProps): ReactElement {
     );
   }
 
+  // model_config branch - mirror the local-aware provider sync from the
+  // dialog path so inline edits do not leave stale "local" placeholders
+  // on external providers and fill the placeholder when switching to local.
+  const localNames = props.localProviderNames ?? new Set<string>();
+  const modelConfig = props.config;
+  const handleProviderChange = (nextProvider: string) => {
+    const isLocal = localNames.has(nextProvider);
+    if (isLocal && !modelConfig.model.trim()) {
+      props.onUpdate({ provider: nextProvider, model: "local" });
+      return;
+    }
+    if (!isLocal && modelConfig.model === "local") {
+      props.onUpdate({ provider: nextProvider, model: "" });
+      return;
+    }
+    props.onUpdate({ provider: nextProvider });
+  };
+  const isLinkedToLocal = localNames.has(modelConfig.provider);
+
   return (
     <div className="grid gap-3 sm:grid-cols-2">
       <InlineField label="Provider">
         <Input
           className="nodrag h-8 w-full text-xs"
           placeholder="provider alias"
-          value={props.config.provider}
-          onChange={(event) => props.onUpdate({ provider: event.target.value })}
+          value={modelConfig.provider}
+          onChange={(event) => handleProviderChange(event.target.value)}
         />
       </InlineField>
       <InlineField label="Model">
         <Input
           className="nodrag h-8 w-full text-xs"
-          placeholder="gpt-4o-mini"
-          value={props.config.model}
+          placeholder={isLinkedToLocal ? "local" : "gpt-4o-mini"}
+          value={modelConfig.model}
           onChange={(event) => props.onUpdate({ model: event.target.value })}
         />
       </InlineField>
@@ -65,7 +94,7 @@ export function InlineModel(props: InlineModelProps): ReactElement {
           className="nodrag h-8 w-full text-xs"
           type="number"
           placeholder="0.7"
-          value={props.config.inference_temperature ?? ""}
+          value={modelConfig.inference_temperature ?? ""}
           onChange={(event) =>
             props.onUpdate({
               // biome-ignore lint/style/useNamingConvention: api schema
diff --git a/studio/frontend/src/features/recipe-studio/components/recipe-graph-node.tsx b/studio/frontend/src/features/recipe-studio/components/recipe-graph-node.tsx
index 0fcf202190..8afc4e26bf 100644
--- a/studio/frontend/src/features/recipe-studio/components/recipe-graph-node.tsx
+++ b/studio/frontend/src/features/recipe-studio/components/recipe-graph-node.tsx
@@ -30,7 +30,7 @@ import {
   Position,
   useUpdateNodeInternals,
 } from "@xyflow/react";
-import { type ReactElement, memo, useEffect } from "react";
+import { type ReactElement, memo, useEffect, useMemo } from "react";
 import {
   MAX_NODE_WIDTH,
   MAX_NOTE_NODE_WIDTH,
@@ -287,6 +287,7 @@ function renderNodeBody(
   config: NodeConfig | undefined,
   summary: string,
   updateConfig: (id: string, patch: Partial<NodeConfig>) => void,
+  localProviderNames: Set<string>,
 ): ReactElement {
   if (config?.kind === "markdown_note") {
     return <MarkdownPreview markdown={config.markdown} />;
@@ -300,7 +301,13 @@ function renderNodeBody(
       return <InlineSampler config={config} onUpdate={onUpdate} />;
     }
     if (config.kind === "model_provider" || config.kind === "model_config") {
-      return <InlineModel config={config} onUpdate={onUpdate} />;
+      return (
+        <InlineModel
+          config={config}
+          localProviderNames={localProviderNames}
+          onUpdate={onUpdate}
+        />
+      );
     }
     if (config.kind === "llm") {
       return <InlineLlm config={config} onUpdate={onUpdate} />;
@@ -355,6 +362,16 @@ function RecipeGraphNodeBase({
   const config = useRecipeStudioStore((state) => state.configs[id]);
   const openConfig = useRecipeStudioStore((state) => state.openConfig);
   const updateConfig = useRecipeStudioStore((state) => state.updateConfig);
+  const allConfigs = useRecipeStudioStore((state) => state.configs);
+  const localProviderNames = useMemo(() => {
+    const names = new Set<string>();
+    for (const cfg of Object.values(allConfigs)) {
+      if (cfg.kind === "model_provider" && cfg.is_local === true) {
+        names.add(cfg.name);
+      }
+    }
+    return names;
+  }, [allConfigs]);
   const llmAuxVisible = useRecipeStudioStore(
     (state) => state.llmAuxVisibility[id] ?? false,
   );
@@ -418,7 +435,12 @@ function RecipeGraphNodeBase({
     data.kind === "tool_config" ||
     data.kind === "validator";
   const summary = getConfigSummary(config);
-  const nodeBody = renderNodeBody(config, summary, updateConfig);
+  const nodeBody = renderNodeBody(
+    config,
+    summary,
+    updateConfig,
+    localProviderNames,
+  );
   const canShowLlmAux =
     config?.kind === "llm" &&
     (Boolean(config.prompt.trim()) ||
diff --git a/studio/frontend/src/features/recipe-studio/dialogs/config-dialog.tsx b/studio/frontend/src/features/recipe-studio/dialogs/config-dialog.tsx
index 6457993c35..c62169956d 100644
--- a/studio/frontend/src/features/recipe-studio/dialogs/config-dialog.tsx
+++ b/studio/frontend/src/features/recipe-studio/dialogs/config-dialog.tsx
@@ -18,6 +18,7 @@ type ConfigDialogProps = {
   categoryOptions: SamplerConfig[];
   modelConfigAliases: string[];
   modelProviderOptions: string[];
+  localProviderNames: Set<string>;
   toolProfileAliases: string[];
   datetimeOptions: string[];
   onUpdate: (id: string, patch: Partial<NodeConfig>) => void;
@@ -32,6 +33,7 @@ export function ConfigDialog({
   categoryOptions,
   modelConfigAliases,
   modelProviderOptions,
+  localProviderNames,
   toolProfileAliases,
   datetimeOptions,
   onUpdate,
@@ -101,6 +103,7 @@ export function ConfigDialog({
                 categoryOptions,
                 modelConfigAliases,
                 modelProviderOptions,
+                localProviderNames,
                 toolProfileAliases,
                 datetimeOptions,
                 onUpdate,
diff --git a/studio/frontend/src/features/recipe-studio/dialogs/models/model-config-dialog.tsx b/studio/frontend/src/features/recipe-studio/dialogs/models/model-config-dialog.tsx
index 3192dc8fd9..368ae08acb 100644
--- a/studio/frontend/src/features/recipe-studio/dialogs/models/model-config-dialog.tsx
+++ b/studio/frontend/src/features/recipe-studio/dialogs/models/model-config-dialog.tsx
@@ -17,7 +17,7 @@ import {
 } from "@/components/ui/combobox";
 import { Input } from "@/components/ui/input";
 import { Textarea } from "@/components/ui/textarea";
-import { type ReactElement, useRef, useState } from "react";
+import { type ReactElement, useEffect, useRef, useState } from "react";
 import type { ModelConfig } from "../../types";
 import { CollapsibleSectionTriggerButton } from "../shared/collapsible-section-trigger";
 import { FieldLabel } from "../shared/field-label";
@@ -26,14 +26,17 @@ import { NameField } from "../shared/name-field";
 type ModelConfigDialogProps = {
   config: ModelConfig;
   providerOptions: string[];
+  localProviderNames: Set<string>;
   onUpdate: (patch: Partial<ModelConfig>) => void;
 };
 
 export function ModelConfigDialog({
   config,
   providerOptions,
+  localProviderNames,
   onUpdate,
 }: ModelConfigDialogProps): ReactElement {
+  const isLinkedToLocal = localProviderNames.has(config.provider);
   const [optionalOpen, setOptionalOpen] = useState(false);
   const modelId = `${config.id}-model`;
   const providerId = `${config.id}-provider`;
@@ -44,11 +47,13 @@ export function ModelConfigDialog({
   const extraBodyId = `${config.id}-inference-extra-body`;
   const providerAnchorRef = useRef<HTMLDivElement>(null);
   const providerInputRef = useRef(config.provider);
-  const lastProviderRef = useRef(config.provider);
-  if (lastProviderRef.current !== config.provider) {
-    lastProviderRef.current = config.provider;
+  // Sync providerInputRef with the current provider value. Updating a ref in
+  // an effect (vs reading/writing it during render) satisfies the
+  // react-hooks/refs rule and keeps the combobox blur path stable across
+  // re-renders.
+  useEffect(() => {
     providerInputRef.current = config.provider;
-  }
+  }, [config.provider]);
   const updateField = <K extends keyof ModelConfig>(
     key: K,
     value: ModelConfig[K],
@@ -56,6 +61,21 @@ export function ModelConfigDialog({
     onUpdate({ [key]: value } as Partial<ModelConfig>);
   };
 
+  // Apply provider selection while keeping the local-provider model autofill
+  // consistent across both dropdown selection and free-typed + blur input.
+  const applyProviderChange = (selectedProvider: string) => {
+    const isLocal = localProviderNames.has(selectedProvider);
+    if (isLocal && !config.model.trim()) {
+      onUpdate({ provider: selectedProvider, model: "local" });
+      return;
+    }
+    if (!isLocal && config.model === "local") {
+      onUpdate({ provider: selectedProvider, model: "" });
+      return;
+    }
+    updateField("provider", selectedProvider);
+  };
+
   return (
     <div className="space-y-4">
       <NameField
@@ -84,7 +104,7 @@ export function ModelConfigDialog({
             filteredItems={providerOptions}
             filter={null}
             value={config.provider || null}
-            onValueChange={(value) => updateField("provider", value ?? "")}
+            onValueChange={(value) => applyProviderChange(value ?? "")}
             onInputValueChange={(value) => {
               providerInputRef.current = value;
             }}
@@ -98,7 +118,7 @@ export function ModelConfigDialog({
               onBlur={() => {
                 const next = providerInputRef.current;
                 if (next !== config.provider) {
-                  updateField("provider", next);
+                  applyProviderChange(next);
                 }
               }}
             />
@@ -124,12 +144,12 @@ export function ModelConfigDialog({
         <FieldLabel
           label="Model ID"
           htmlFor={modelId}
-          hint="The exact model name sent to the connection."
+          hint={isLinkedToLocal ? "Uses the model loaded in Chat. Any value works here." : "The exact model name sent to the connection."}
         />
         <Input
           id={modelId}
           className="nodrag"
-          placeholder="gpt-4o-mini"
+          placeholder={isLinkedToLocal ? "local" : "gpt-4o-mini"}
           value={config.model}
           onChange={(event) => updateField("model", event.target.value)}
         />
diff --git a/studio/frontend/src/features/recipe-studio/dialogs/models/model-provider-dialog.tsx b/studio/frontend/src/features/recipe-studio/dialogs/models/model-provider-dialog.tsx
index 897fc7ee36..ef7366a225 100644
--- a/studio/frontend/src/features/recipe-studio/dialogs/models/model-provider-dialog.tsx
+++ b/studio/frontend/src/features/recipe-studio/dialogs/models/model-provider-dialog.tsx
@@ -24,6 +24,7 @@ export function ModelProviderDialog({
   onUpdate,
 }: ModelProviderDialogProps): ReactElement {
   const [optionalOpen, setOptionalOpen] = useState(false);
+  const isLocal = config.is_local ?? false;
   const endpointId = `${config.id}-endpoint`;
   const apiKeyEnvId = `${config.id}-api-key-env`;
   const apiKeyId = `${config.id}-api-key`;
@@ -43,94 +44,163 @@ export function ModelProviderDialog({
         value={config.name}
         onChange={(value) => onUpdate({ name: value })}
       />
-      <div className="rounded-2xl border border-border/60 bg-muted/10 px-4 py-3">
-        <p className="text-sm font-semibold text-foreground">
-          Start with the endpoint you want this model to use
-        </p>
-        <p className="mt-1 text-xs text-muted-foreground">
-          Most connections only need an endpoint. Add an API key if that
-          service requires one.
-        </p>
-      </div>
+
+      {/* Model source toggle */}
       <div className="grid gap-1.5">
-        <FieldLabel
-          label="Endpoint"
-          htmlFor={endpointId}
-          hint="Base URL for the model service or gateway."
-        />
-        <Input
-          id={endpointId}
-          className="nodrag"
-          placeholder="https://..."
-          value={config.endpoint}
-          onChange={(event) => updateField("endpoint", event.target.value)}
-        />
+        <p className="text-sm font-semibold text-foreground">Model source</p>
+        <div className="grid grid-cols-2 gap-2">
+          <button
+            type="button"
+            className={`rounded-xl border px-4 py-3 text-left transition-colors ${
+              isLocal
+                ? "border-primary/40 bg-primary/5"
+                : "border-border/60 bg-muted/10 hover:border-border"
+            }`}
+            onClick={() =>
+              onUpdate({
+                is_local: true,
+                endpoint: "",
+                api_key: "",
+                api_key_env: "",
+                extra_headers: "",
+                extra_body: "",
+              })
+            }
+          >
+            <p className="text-sm font-semibold text-foreground">
+              Local model
+            </p>
+            <p className="mt-0.5 text-xs text-muted-foreground">
+              Use the model loaded in the Chat tab
+            </p>
+          </button>
+          <button
+            type="button"
+            className={`rounded-xl border px-4 py-3 text-left transition-colors ${
+              !isLocal
+                ? "border-primary/40 bg-primary/5"
+                : "border-border/60 bg-muted/10 hover:border-border"
+            }`}
+            onClick={() => onUpdate({ is_local: false })}
+          >
+            <p className="text-sm font-semibold text-foreground">
+              External endpoint
+            </p>
+            <p className="mt-0.5 text-xs text-muted-foreground">
+              Connect to an API like OpenAI, Together, or a custom server
+            </p>
+          </button>
+        </div>
       </div>
-      <div className="grid gap-1.5">
-        <FieldLabel
-          label="API key (optional)"
-          htmlFor={apiKeyId}
-          hint="Paste a key here, or use an environment variable below."
-        />
-        <Input
-          id={apiKeyId}
-          className="nodrag"
-          value={config.api_key ?? ""}
-          onChange={(event) => updateField("api_key", event.target.value)}
-        />
-      </div>
-      <Collapsible open={optionalOpen} onOpenChange={setOptionalOpen}>
-        <CollapsibleTrigger asChild={true}>
-          <CollapsibleSectionTriggerButton
-            label="Advanced request overrides"
-            open={optionalOpen}
-          />
-        </CollapsibleTrigger>
-        <CollapsibleContent className="mt-3 space-y-4">
+
+      {isLocal ? (
+        <div className="rounded-2xl border border-border/60 bg-muted/10 px-4 py-3">
+          <p className="text-sm font-semibold text-foreground">
+            Ready to go
+          </p>
+          <p className="mt-1 text-xs text-muted-foreground">
+            Recipes will use whatever model is loaded in the Chat tab when you
+            hit run. No endpoint or API key needed.
+          </p>
+        </div>
+      ) : (
+        <>
+          <div className="rounded-2xl border border-border/60 bg-muted/10 px-4 py-3">
+            <p className="text-sm font-semibold text-foreground">
+              Start with the endpoint you want this model to use
+            </p>
+            <p className="mt-1 text-xs text-muted-foreground">
+              Most connections only need an endpoint. Add an API key if that
+              service requires one.
+            </p>
+          </div>
           <div className="grid gap-1.5">
             <FieldLabel
-              label="API key environment variable"
-              htmlFor={apiKeyEnvId}
-              hint="Name of the environment variable that stores the key."
+              label="Endpoint"
+              htmlFor={endpointId}
+              hint="Base URL for the model service or gateway."
             />
             <Input
-              id={apiKeyEnvId}
+              id={endpointId}
               className="nodrag"
-              placeholder="OPENAI_API_KEY"
-              value={config.api_key_env ?? ""}
-              onChange={(event) => updateField("api_key_env", event.target.value)}
+              placeholder="https://..."
+              value={config.endpoint}
+              onChange={(event) => updateField("endpoint", event.target.value)}
             />
           </div>
           <div className="grid gap-1.5">
             <FieldLabel
-              label="Extra headers (JSON)"
-              htmlFor={extraHeadersId}
-              hint="Optional headers to send with every request."
+              label="API key (optional)"
+              htmlFor={apiKeyId}
+              hint="Paste a key here, or use an environment variable below."
             />
-            <Textarea
-              id={extraHeadersId}
-              className="corner-squircle nodrag"
-              placeholder='{"X-Header": "value"}'
-              value={config.extra_headers ?? ""}
-              onChange={(event) => updateField("extra_headers", event.target.value)}
+            <Input
+              id={apiKeyId}
+              className="nodrag"
+              value={config.api_key ?? ""}
+              onChange={(event) => updateField("api_key", event.target.value)}
             />
           </div>
-          <div className="grid gap-1.5">
-            <FieldLabel
-              label="Extra body (JSON)"
-              htmlFor={extraBodyId}
-              hint="Optional request fields to send every time."
-            />
-            <Textarea
-              id={extraBodyId}
-              className="corner-squircle nodrag"
-              placeholder='{"key": "value"}'
-              value={config.extra_body ?? ""}
-              onChange={(event) => updateField("extra_body", event.target.value)}
-            />
-          </div>
-        </CollapsibleContent>
-      </Collapsible>
+          <Collapsible open={optionalOpen} onOpenChange={setOptionalOpen}>
+            <CollapsibleTrigger asChild={true}>
+              <CollapsibleSectionTriggerButton
+                label="Advanced request overrides"
+                open={optionalOpen}
+              />
+            </CollapsibleTrigger>
+            <CollapsibleContent className="mt-3 space-y-4">
+              <div className="grid gap-1.5">
+                <FieldLabel
+                  label="API key environment variable"
+                  htmlFor={apiKeyEnvId}
+                  hint="Name of the environment variable that stores the key."
+                />
+                <Input
+                  id={apiKeyEnvId}
+                  className="nodrag"
+                  placeholder="OPENAI_API_KEY"
+                  value={config.api_key_env ?? ""}
+                  onChange={(event) =>
+                    updateField("api_key_env", event.target.value)
+                  }
+                />
+              </div>
+              <div className="grid gap-1.5">
+                <FieldLabel
+                  label="Extra headers (JSON)"
+                  htmlFor={extraHeadersId}
+                  hint="Optional headers to send with every request."
+                />
+                <Textarea
+                  id={extraHeadersId}
+                  className="corner-squircle nodrag"
+                  placeholder='{"X-Header": "value"}'
+                  value={config.extra_headers ?? ""}
+                  onChange={(event) =>
+                    updateField("extra_headers", event.target.value)
+                  }
+                />
+              </div>
+              <div className="grid gap-1.5">
+                <FieldLabel
+                  label="Extra body (JSON)"
+                  htmlFor={extraBodyId}
+                  hint="Optional request fields to send every time."
+                />
+                <Textarea
+                  id={extraBodyId}
+                  className="corner-squircle nodrag"
+                  placeholder='{"key": "value"}'
+                  value={config.extra_body ?? ""}
+                  onChange={(event) =>
+                    updateField("extra_body", event.target.value)
+                  }
+                />
+              </div>
+            </CollapsibleContent>
+          </Collapsible>
+        </>
+      )}
     </div>
   );
 }
diff --git a/studio/frontend/src/features/recipe-studio/recipe-studio-page.tsx b/studio/frontend/src/features/recipe-studio/recipe-studio-page.tsx
index c922293c02..586d5d8ad8 100644
--- a/studio/frontend/src/features/recipe-studio/recipe-studio-page.tsx
+++ b/studio/frontend/src/features/recipe-studio/recipe-studio-page.tsx
@@ -776,6 +776,7 @@ export function RecipeStudioPage({
         categoryOptions={dialogOptions.categoryOptions}
         modelConfigAliases={dialogOptions.modelConfigAliases}
         modelProviderOptions={dialogOptions.modelProviderOptions}
+        localProviderNames={dialogOptions.localProviderNames}
         toolProfileAliases={dialogOptions.toolProfileAliases}
         datetimeOptions={dialogOptions.datetimeOptions}
         onUpdate={updateConfig}
diff --git a/studio/frontend/src/features/recipe-studio/stores/helpers/reference-sync.ts b/studio/frontend/src/features/recipe-studio/stores/helpers/reference-sync.ts
index 3030bcd8ae..56634b84c8 100644
--- a/studio/frontend/src/features/recipe-studio/stores/helpers/reference-sync.ts
+++ b/studio/frontend/src/features/recipe-studio/stores/helpers/reference-sync.ts
@@ -137,7 +137,15 @@ export function applyRemovalToConfig(
   }
   if (config.kind === "model_config" && config.provider === ref) {
     const base = next as ModelConfig;
-    next = { ...base, provider: "" };
+    // Clear the synthetic "local" placeholder when the provider that was
+    // a local provider is removed; otherwise the stale placeholder would
+    // pass validation against a future external provider and then fail
+    // at runtime against a real API ("model not found").
+    next = {
+      ...base,
+      provider: "",
+      model: base.model === "local" ? "" : base.model,
+    };
   }
   if (config.kind === "llm" && config.model_alias === ref) {
     const base = next as LlmConfig;
diff --git a/studio/frontend/src/features/recipe-studio/stores/recipe-studio.ts b/studio/frontend/src/features/recipe-studio/stores/recipe-studio.ts
index 7416efa457..5b16bde4b0 100644
--- a/studio/frontend/src/features/recipe-studio/stores/recipe-studio.ts
+++ b/studio/frontend/src/features/recipe-studio/stores/recipe-studio.ts
@@ -736,6 +736,34 @@ export const useRecipeStudioStore = create<RecipeStudioState>((set, get) => ({
         configs = applyRenameToConfigs(configs, oldName, newName);
       }
 
+      // When a provider toggles between local and external, keep already
+      // linked model_config nodes in sync. applyRenameToConfigs above has
+      // already propagated any name change, so providerName here is the
+      // post-rename value.
+      if (current.kind === "model_provider" && next.kind === "model_provider") {
+        const prevIsLocal = current.is_local === true;
+        const nextIsLocal = next.is_local === true;
+        if (prevIsLocal !== nextIsLocal) {
+          const providerName = next.name;
+          for (const [cfgId, cfg] of Object.entries(configs)) {
+            if (cfg.kind !== "model_config" || cfg.provider !== providerName) {
+              continue;
+            }
+            if (nextIsLocal && !cfg.model.trim()) {
+              // external -> local: auto fill the placeholder model id so the
+              // config does not fail "model is required" validation.
+              configs = { ...configs, [cfgId]: { ...cfg, model: "local" } };
+              continue;
+            }
+            if (!nextIsLocal && cfg.model === "local") {
+              // local -> external: clear the placeholder so the user picks a
+              // real model id for the new external endpoint.
+              configs = { ...configs, [cfgId]: { ...cfg, model: "" } };
+            }
+          }
+        }
+      }
+
       return { configs, nodes, edges };
     };
     set(applyUpdate);
diff --git a/studio/frontend/src/features/recipe-studio/types/index.ts b/studio/frontend/src/features/recipe-studio/types/index.ts
index 1fb0f36f51..cd1127a4e2 100644
--- a/studio/frontend/src/features/recipe-studio/types/index.ts
+++ b/studio/frontend/src/features/recipe-studio/types/index.ts
@@ -252,6 +252,8 @@ export type ModelProviderConfig = {
   extra_headers?: string;
   // biome-ignore lint/style/useNamingConvention: api schema
   extra_body?: string;
+  // biome-ignore lint/style/useNamingConvention: api schema
+  is_local?: boolean;
 };
 
 export type ModelConfig = {
diff --git a/studio/frontend/src/features/recipe-studio/utils/config-factories.ts b/studio/frontend/src/features/recipe-studio/utils/config-factories.ts
index 142779757d..76fc2c38ee 100644
--- a/studio/frontend/src/features/recipe-studio/utils/config-factories.ts
+++ b/studio/frontend/src/features/recipe-studio/utils/config-factories.ts
@@ -240,6 +240,8 @@ export function makeModelProviderConfig(
     extra_headers: "",
     // biome-ignore lint/style/useNamingConvention: api schema
     extra_body: "",
+    // biome-ignore lint/style/useNamingConvention: api schema
+    is_local: false,
   };
 }
 
diff --git a/studio/frontend/src/features/recipe-studio/utils/graph/recipe-graph-connection.ts b/studio/frontend/src/features/recipe-studio/utils/graph/recipe-graph-connection.ts
index 32310cd8e0..4fa70b3e3a 100644
--- a/studio/frontend/src/features/recipe-studio/utils/graph/recipe-graph-connection.ts
+++ b/studio/frontend/src/features/recipe-studio/utils/graph/recipe-graph-connection.ts
@@ -386,7 +386,17 @@ export function applyRecipeConnection(
     nextBaseEdges,
   );
   if (source.kind === "model_provider" && target.kind === "model_config") {
-    const next = { ...target, provider: source.name };
+    // Keep the model_config.model field in sync with provider mode when the
+    // link is changed via graph drag (the model-config dialog path has its
+    // own applyProviderChange helper that does the same thing).
+    const isSourceLocal = source.is_local === true;
+    let nextModel = target.model;
+    if (isSourceLocal && !nextModel.trim()) {
+      nextModel = "local";
+    } else if (!isSourceLocal && nextModel === "local") {
+      nextModel = "";
+    }
+    const next = { ...target, provider: source.name, model: nextModel };
     return { edges: nextEdges, configs: { ...configs, [target.id]: next } };
   }
   if (source.kind === "model_config" && target.kind === "llm") {
diff --git a/studio/frontend/src/features/recipe-studio/utils/import/parsers/model-parser.ts b/studio/frontend/src/features/recipe-studio/utils/import/parsers/model-parser.ts
index 8935c65675..15ecf39a7b 100644
--- a/studio/frontend/src/features/recipe-studio/utils/import/parsers/model-parser.ts
+++ b/studio/frontend/src/features/recipe-studio/utils/import/parsers/model-parser.ts
@@ -35,6 +35,8 @@ export function parseModelProvider(
     extra_body: isRecord(provider.extra_body)
       ? JSON.stringify(provider.extra_body, null, 2)
       : "",
+    // biome-ignore lint/style/useNamingConvention: api schema
+    is_local: provider.is_local === true,
   };
 }
 
diff --git a/studio/frontend/src/features/recipe-studio/utils/payload/build-payload.ts b/studio/frontend/src/features/recipe-studio/utils/payload/build-payload.ts
index 57160140d9..34c3c34274 100644
--- a/studio/frontend/src/features/recipe-studio/utils/payload/build-payload.ts
+++ b/studio/frontend/src/features/recipe-studio/utils/payload/build-payload.ts
@@ -104,6 +104,7 @@ export function buildRecipePayload(
   const columns: Record<string, unknown>[] = [];
   const modelAliases = new Set<string>();
   const modelProviderNames = new Set<string>();
+  const localProviderNames = new Set<string>();
   const modelProviders: Record<string, unknown>[] = [];
   const mcpProviders: Record<string, unknown>[] = [];
   const modelConfigs: Record<string, unknown>[] = [];
@@ -200,6 +201,9 @@ export function buildRecipePayload(
     }
     if (config.kind === "model_provider") {
       modelProviderNames.add(config.name);
+      if (config.is_local) {
+        localProviderNames.add(config.name);
+      }
       modelProviders.push(buildModelProvider(config, errors));
       modelProviderConfigs.push(config);
       continue;
@@ -240,6 +244,7 @@ export function buildRecipePayload(
     modelConfigConfigs,
     modelAliases,
     modelProviderNames,
+    localProviderNames,
     errors,
   );
   validateUsedProviders(modelProviderConfigs, modelConfigConfigs, errors);
diff --git a/studio/frontend/src/features/recipe-studio/utils/payload/builders-model.ts b/studio/frontend/src/features/recipe-studio/utils/payload/builders-model.ts
index 507da6d32d..14e0faa5cc 100644
--- a/studio/frontend/src/features/recipe-studio/utils/payload/builders-model.ts
+++ b/studio/frontend/src/features/recipe-studio/utils/payload/builders-model.ts
@@ -8,6 +8,26 @@ export function buildModelProvider(
   config: ModelProviderConfig,
   errors: string[],
 ): Record<string, unknown> {
+  // Local providers do not use any of the advanced request overrides -
+  // the backend overrides endpoint/api_key/provider_type and strips the
+  // extra fields in _inject_local_providers. Skip parsing the hidden
+  // JSON inputs here so imported or hydrated recipes with stale extra
+  // headers/body cannot block the client-side validation step.
+  if (config.is_local === true) {
+    return {
+      name: config.name,
+      endpoint: "",
+      // biome-ignore lint/style/useNamingConvention: api schema
+      provider_type: "openai",
+      // biome-ignore lint/style/useNamingConvention: api schema
+      extra_headers: {},
+      // biome-ignore lint/style/useNamingConvention: api schema
+      extra_body: {},
+      // biome-ignore lint/style/useNamingConvention: api schema
+      is_local: true,
+    };
+  }
+
   const extraHeaders = parseJsonObject(
     config.extra_headers,
     `Provider ${config.name} extra_headers`,
diff --git a/studio/frontend/src/features/recipe-studio/utils/payload/validate.ts b/studio/frontend/src/features/recipe-studio/utils/payload/validate.ts
index d73762006c..7e72e8d919 100644
--- a/studio/frontend/src/features/recipe-studio/utils/payload/validate.ts
+++ b/studio/frontend/src/features/recipe-studio/utils/payload/validate.ts
@@ -84,12 +84,16 @@ export function validateModelConfigProviders(
   modelConfigConfigs: ModelConfig[],
   modelAliases: Set<string>,
   modelProviderNames: Set<string>,
+  localProviderNames: Set<string>,
   errors: string[],
 ): void {
   for (const config of modelConfigConfigs) {
     const provider = config.provider.trim();
     const alias = config.name;
-    if (modelAliases.has(alias) && !config.model.trim()) {
+    const isLocal = localProviderNames.has(provider);
+    // Local providers do not require a real model id - the loaded Chat
+    // model is used regardless of what gets sent in the payload.
+    if (!isLocal && modelAliases.has(alias) && !config.model.trim()) {
       errors.push(`Model config ${alias}: model is required.`);
     }
     if (provider && !modelProviderNames.has(provider)) {
@@ -110,6 +114,9 @@ export function validateUsedProviders(
     if (!usedProviders.has(provider.name)) {
       continue;
     }
+    if (provider.is_local) {
+      continue;
+    }
     if (!provider.endpoint.trim()) {
       errors.push(`Model provider ${provider.name}: endpoint is required.`);
     }
diff --git a/studio/frontend/src/features/recipe-studio/utils/recipe-studio-view.ts b/studio/frontend/src/features/recipe-studio/utils/recipe-studio-view.ts
index 5fe4b86118..9de896abea 100644
--- a/studio/frontend/src/features/recipe-studio/utils/recipe-studio-view.ts
+++ b/studio/frontend/src/features/recipe-studio/utils/recipe-studio-view.ts
@@ -7,6 +7,7 @@ export type DialogOptions = {
   categoryOptions: SamplerConfig[];
   modelConfigAliases: string[];
   modelProviderOptions: string[];
+  localProviderNames: Set<string>;
   toolProfileAliases: string[];
   datetimeOptions: string[];
 };
@@ -15,6 +16,7 @@ export function buildDialogOptions(configList: NodeConfig[]): DialogOptions {
   const categoryOptions: SamplerConfig[] = [];
   const modelConfigAliases: string[] = [];
   const modelProviderOptions: string[] = [];
+  const localProviderNames = new Set<string>();
   const toolProfileAliases: string[] = [];
   const datetimeOptions: string[] = [];
 
@@ -34,6 +36,9 @@ export function buildDialogOptions(configList: NodeConfig[]): DialogOptions {
     }
     if (config.kind === "model_provider") {
       modelProviderOptions.push(config.name);
+      if (config.is_local) {
+        localProviderNames.add(config.name);
+      }
       continue;
     }
     if (config.kind === "tool_config") {
@@ -45,6 +50,7 @@ export function buildDialogOptions(configList: NodeConfig[]): DialogOptions {
     categoryOptions,
     modelConfigAliases,
     modelProviderOptions,
+    localProviderNames,
     toolProfileAliases,
     datetimeOptions,
   };

From 5fa8683b27333111657e1976166fabce7e4abed5 Mon Sep 17 00:00:00 2001
From: "dependabot[bot]" <49699333+dependabot[bot]@users.noreply.github.com>
Date: Wed, 8 Apr 2026 04:34:33 -0700
Subject: [PATCH 061/216] build(deps): bump the bun-frontend group across 1
 directory with 16 updates (#4586)

* build(deps): bump the bun-frontend group across 1 directory with 16 updates

Bumps the bun-frontend group with 16 updates in the /studio/frontend directory:

| Package | From | To |
| --- | --- | --- |
| [@dagrejs/dagre](https://github.com/dagrejs/dagre) | `2.0.4` | `3.0.0` |
| [@dagrejs/graphlib](https://github.com/dagrejs/graphlib) | `3.0.4` | `4.0.1` |
| @hugeicons/core-free-icons | `3.3.0` | `4.0.0` |
| [@streamdown/cjk](https://github.com/vercel/streamdown/tree/HEAD/packages/streamdown-cjk) | `1.0.2` | `1.0.3` |
| [@streamdown/code](https://github.com/vercel/streamdown/tree/HEAD/packages/streamdown-code) | `1.0.2` | `1.1.1` |
| [lucide-react](https://github.com/lucide-icons/lucide/tree/HEAD/packages/lucide-react) | `0.577.0` | `1.6.0` |
| [recharts](https://github.com/recharts/recharts) | `3.7.0` | `3.8.0` |
| [shadcn](https://github.com/shadcn-ui/ui/tree/HEAD/packages/shadcn) | `3.8.5` | `4.1.0` |
| [streamdown](https://github.com/vercel/streamdown/tree/HEAD/packages/streamdown) | `2.3.0` | `2.5.0` |
| [@biomejs/biome](https://github.com/biomejs/biome/tree/HEAD/packages/@biomejs/biome) | `1.9.4` | `2.4.8` |
| [@eslint/js](https://github.com/eslint/eslint/tree/HEAD/packages/js) | `9.39.4` | `10.0.1` |
| [@types/node](https://github.com/DefinitelyTyped/DefinitelyTyped/tree/HEAD/types/node) | `24.12.0` | `25.5.0` |
| [eslint](https://github.com/eslint/eslint) | `9.39.4` | `10.1.0` |
| [eslint-plugin-react-refresh](https://github.com/ArnaudBarre/eslint-plugin-react-refresh) | `0.4.26` | `0.5.2` |
| [globals](https://github.com/sindresorhus/globals) | `16.5.0` | `17.4.0` |
| [typescript](https://github.com/microsoft/TypeScript) | `5.9.3` | `6.0.2` |



Updates `@dagrejs/dagre` from 2.0.4 to 3.0.0
- [Release notes](https://github.com/dagrejs/dagre/releases)
- [Changelog](https://github.com/dagrejs/dagre/blob/master/changelog.md)
- [Commits](https://github.com/dagrejs/dagre/compare/v2.0.4...v3.0.0)

Updates `@dagrejs/graphlib` from 3.0.4 to 4.0.1
- [Release notes](https://github.com/dagrejs/graphlib/releases)
- [Changelog](https://github.com/dagrejs/graphlib/blob/master/changelog.md)
- [Commits](https://github.com/dagrejs/graphlib/compare/v3.0.4...v4.0.1)

Updates `@hugeicons/core-free-icons` from 3.3.0 to 4.0.0

Updates `@streamdown/cjk` from 1.0.2 to 1.0.3
- [Release notes](https://github.com/vercel/streamdown/releases)
- [Changelog](https://github.com/vercel/streamdown/blob/main/packages/streamdown-cjk/CHANGELOG.md)
- [Commits](https://github.com/vercel/streamdown/commits/@streamdown/cjk@1.0.3/packages/streamdown-cjk)

Updates `@streamdown/code` from 1.0.2 to 1.1.1
- [Release notes](https://github.com/vercel/streamdown/releases)
- [Changelog](https://github.com/vercel/streamdown/blob/main/packages/streamdown-code/CHANGELOG.md)
- [Commits](https://github.com/vercel/streamdown/commits/@streamdown/code@1.1.1/packages/streamdown-code)

Updates `lucide-react` from 0.577.0 to 1.6.0
- [Release notes](https://github.com/lucide-icons/lucide/releases)
- [Commits](https://github.com/lucide-icons/lucide/commits/1.6.0/packages/lucide-react)

Updates `recharts` from 3.7.0 to 3.8.0
- [Release notes](https://github.com/recharts/recharts/releases)
- [Changelog](https://github.com/recharts/recharts/blob/main/CHANGELOG.md)
- [Commits](https://github.com/recharts/recharts/compare/v3.7.0...v3.8.0)

Updates `shadcn` from 3.8.5 to 4.1.0
- [Release notes](https://github.com/shadcn-ui/ui/releases)
- [Changelog](https://github.com/shadcn-ui/ui/blob/main/packages/shadcn/CHANGELOG.md)
- [Commits](https://github.com/shadcn-ui/ui/commits/shadcn@4.1.0/packages/shadcn)

Updates `streamdown` from 2.3.0 to 2.5.0
- [Release notes](https://github.com/vercel/streamdown/releases)
- [Changelog](https://github.com/vercel/streamdown/blob/main/packages/streamdown/CHANGELOG.md)
- [Commits](https://github.com/vercel/streamdown/commits/streamdown@2.5.0/packages/streamdown)

Updates `@biomejs/biome` from 1.9.4 to 2.4.8
- [Release notes](https://github.com/biomejs/biome/releases)
- [Changelog](https://github.com/biomejs/biome/blob/main/packages/@biomejs/biome/CHANGELOG.md)
- [Commits](https://github.com/biomejs/biome/commits/@biomejs/biome@2.4.8/packages/@biomejs/biome)

Updates `@eslint/js` from 9.39.4 to 10.0.1
- [Release notes](https://github.com/eslint/eslint/releases)
- [Commits](https://github.com/eslint/eslint/commits/v10.0.1/packages/js)

Updates `@types/node` from 24.12.0 to 25.5.0
- [Release notes](https://github.com/DefinitelyTyped/DefinitelyTyped/releases)
- [Commits](https://github.com/DefinitelyTyped/DefinitelyTyped/commits/HEAD/types/node)

Updates `eslint` from 9.39.4 to 10.1.0
- [Release notes](https://github.com/eslint/eslint/releases)
- [Commits](https://github.com/eslint/eslint/compare/v9.39.4...v10.1.0)

Updates `eslint-plugin-react-refresh` from 0.4.26 to 0.5.2
- [Release notes](https://github.com/ArnaudBarre/eslint-plugin-react-refresh/releases)
- [Changelog](https://github.com/ArnaudBarre/eslint-plugin-react-refresh/blob/main/CHANGELOG.md)
- [Commits](https://github.com/ArnaudBarre/eslint-plugin-react-refresh/compare/v0.4.26...v0.5.2)

Updates `globals` from 16.5.0 to 17.4.0
- [Release notes](https://github.com/sindresorhus/globals/releases)
- [Commits](https://github.com/sindresorhus/globals/compare/v16.5.0...v17.4.0)

Updates `typescript` from 5.9.3 to 6.0.2
- [Release notes](https://github.com/microsoft/TypeScript/releases)
- [Commits](https://github.com/microsoft/TypeScript/compare/v5.9.3...v6.0.2)

---
updated-dependencies:
- dependency-name: "@dagrejs/dagre"
  dependency-version: 3.0.0
  dependency-type: direct:production
  update-type: version-update:semver-major
  dependency-group: bun-frontend
- dependency-name: "@dagrejs/graphlib"
  dependency-version: 4.0.1
  dependency-type: direct:production
  update-type: version-update:semver-major
  dependency-group: bun-frontend
- dependency-name: "@hugeicons/core-free-icons"
  dependency-version: 4.0.0
  dependency-type: direct:production
  update-type: version-update:semver-major
  dependency-group: bun-frontend
- dependency-name: "@streamdown/cjk"
  dependency-version: 1.0.3
  dependency-type: direct:production
  update-type: version-update:semver-patch
  dependency-group: bun-frontend
- dependency-name: "@streamdown/code"
  dependency-version: 1.1.1
  dependency-type: direct:production
  update-type: version-update:semver-minor
  dependency-group: bun-frontend
- dependency-name: lucide-react
  dependency-version: 1.6.0
  dependency-type: direct:production
  update-type: version-update:semver-major
  dependency-group: bun-frontend
- dependency-name: recharts
  dependency-version: 3.8.0
  dependency-type: direct:production
  update-type: version-update:semver-minor
  dependency-group: bun-frontend
- dependency-name: shadcn
  dependency-version: 4.1.0
  dependency-type: direct:production
  update-type: version-update:semver-major
  dependency-group: bun-frontend
- dependency-name: streamdown
  dependency-version: 2.5.0
  dependency-type: direct:production
  update-type: version-update:semver-minor
  dependency-group: bun-frontend
- dependency-name: "@biomejs/biome"
  dependency-version: 2.4.8
  dependency-type: direct:development
  update-type: version-update:semver-major
  dependency-group: bun-frontend
- dependency-name: "@eslint/js"
  dependency-version: 10.0.1
  dependency-type: direct:development
  update-type: version-update:semver-major
  dependency-group: bun-frontend
- dependency-name: "@types/node"
  dependency-version: 25.5.0
  dependency-type: direct:development
  update-type: version-update:semver-major
  dependency-group: bun-frontend
- dependency-name: eslint
  dependency-version: 10.1.0
  dependency-type: direct:development
  update-type: version-update:semver-major
  dependency-group: bun-frontend
- dependency-name: eslint-plugin-react-refresh
  dependency-version: 0.5.2
  dependency-type: direct:development
  update-type: version-update:semver-minor
  dependency-group: bun-frontend
- dependency-name: globals
  dependency-version: 17.4.0
  dependency-type: direct:development
  update-type: version-update:semver-major
  dependency-group: bun-frontend
- dependency-name: typescript
  dependency-version: 6.0.2
  dependency-type: direct:development
  update-type: version-update:semver-major
  dependency-group: bun-frontend
...

Signed-off-by: dependabot[bot] <support@github.com>

* Revert dagrejs upgrades

Keep @dagrejs/dagre at ^2.0.4 and @dagrejs/graphlib at ^3.0.4.

* Revert biome, eslint, typescript, and recharts upgrades

These upgrades break studio/frontend locally:

- @biomejs/biome 2.4.10 fails to parse the existing biome.json
  (files.ignore and organizeImports keys removed in v2; schema
  version mismatch).
- typescript 6.0.2 emits TS5101 on tsconfig.app.json baseUrl
  ("Option 'baseUrl' is deprecated and will stop functioning in
  TypeScript 7.0"), so tsc -b exits 2.
- eslint 10.2.0 conflicts with eslint-plugin-react-hooks@7.0.1,
  which peers on eslint ^9; npm install fails with ERESOLVE.
- recharts 3.8.1 widened LegendPayload.dataKey to include a
  function type, which breaks the React key={item.dataKey} usage
  in src/components/ui/chart.tsx (TS2322).

Hold these at their current pinned versions until the upstream
peer deps and config migrations are ready.

---------

Signed-off-by: dependabot[bot] <support@github.com>
Co-authored-by: dependabot[bot] <49699333+dependabot[bot]@users.noreply.github.com>
Co-authored-by: Daniel Han <danielhanchen@gmail.com>
---
 studio/frontend/package.json | 18 +++++++++---------
 1 file changed, 9 insertions(+), 9 deletions(-)

diff --git a/studio/frontend/package.json b/studio/frontend/package.json
index d9bab4de7e..ffb3c65719 100644
--- a/studio/frontend/package.json
+++ b/studio/frontend/package.json
@@ -25,7 +25,7 @@
     "@fontsource-variable/figtree": "^5.2.10",
     "@fontsource-variable/inter": "^5.2.8",
     "@fontsource-variable/space-grotesk": "^5.2.10",
-    "@hugeicons/core-free-icons": "^3.1.1",
+    "@hugeicons/core-free-icons": "^4.1.1",
     "@hugeicons/react": "^1.1.5",
     "@huggingface/hub": "^2.9.0",
     "@langchain/core": "^1.1.27",
@@ -34,8 +34,8 @@
     "@radix-ui/react-select": "^2.2.6",
     "@radix-ui/react-separator": "^1.1.8",
     "@radix-ui/react-slot": "^1.2.4",
-    "@streamdown/cjk": "1.0.2",
-    "@streamdown/code": "1.0.2",
+    "@streamdown/cjk": "1.0.3",
+    "@streamdown/code": "1.1.1",
     "@streamdown/math": "1.0.2",
     "@streamdown/mermaid": "1.0.2",
     "@tailwindcss/vite": "^4.2.2",
@@ -53,7 +53,7 @@
     "dexie": "^4.3.0",
     "js-yaml": "^4.1.1",
     "katex": "^0.16.28",
-    "lucide-react": "^0.577.0",
+    "lucide-react": "^1.7.0",
     "mammoth": "^1.11.0",
     "motion": "^12.34.0",
     "next": "^16.1.6",
@@ -65,9 +65,9 @@
     "react-resizable-panels": "^4.6.4",
     "recharts": "3.7.0",
     "remark-gfm": "^4.0.1",
-    "shadcn": "^3.8.4",
+    "shadcn": "^4.2.0",
     "sonner": "^2.0.7",
-    "streamdown": "2.3.0",
+    "streamdown": "2.5.0",
     "tailwind-merge": "^3.4.0",
     "tailwindcss": "^4.1.18",
     "tw-animate-css": "^1.4.0",
@@ -79,14 +79,14 @@
     "@biomejs/biome": "^1.9.4",
     "@eslint/js": "^9.39.1",
     "@types/js-yaml": "^4.0.9",
-    "@types/node": "^24.10.1",
+    "@types/node": "^25.5.2",
     "@types/react": "^19.2.5",
     "@types/react-dom": "^19.2.3",
     "@vitejs/plugin-react": "^6.0.1",
     "eslint": "^9.39.1",
     "eslint-plugin-react-hooks": "^7.0.1",
-    "eslint-plugin-react-refresh": "^0.4.26",
-    "globals": "^16.5.0",
+    "eslint-plugin-react-refresh": "^0.5.2",
+    "globals": "^17.4.0",
     "typescript": "~5.9.3",
     "typescript-eslint": "^8.55.0",
     "vite": "^8.0.1"

From da2ef6dce68202dee7bc0e534a8b6780e27c365a Mon Sep 17 00:00:00 2001
From: Daniel Han <danielhanchen@gmail.com>
Date: Thu, 9 Apr 2026 00:07:25 -0700
Subject: [PATCH 062/216] Only run ldconfig CUDA-linking recovery when we have
 permission (#4930)

* Only run ldconfig CUDA-linking recovery when we have permission

When `import unsloth` runs on a non-root environment (shared HPC,
locked-down container, CI runner, etc.) the CUDA-linking recovery path
shells out to `os.system("ldconfig /usr/lib64-nvidia")`, which fails
loudly with "Permission denied". It's especially noisy for users who
don't even have bitsandbytes installed - they're doing 16bit or full
finetuning and the line immediately above told them "16bit and full
finetuning works!". The reason the recovery runs at all in that case
is that `bnb.functional.lib.cdequantize_blockwise_fp32` raises
AttributeError on `bnb is None`, the bare `except:` swallows it, and
the code drops into the recovery unconditionally.

Fix: gate the recovery body on `os.geteuid() == 0`. When we don't
have permission to run ldconfig, silently skip the recovery. When we
do, the recovery runs UNCHANGED - same `os.system()` calls, same
reload + retry, same warnings. `libcuda_dirs()` is used by both triton
and bitsandbytes, so we still want to run the recovery whenever we
have permission, regardless of whether bnb is installed.

For non-root users who DO have bitsandbytes installed and broken,
emit a single remediation warning telling them how to fix it manually
(`sudo ldconfig /usr/lib64-nvidia`). This preserves the diagnostic
guidance from the original code without the Permission denied noise.

Scope:
- Only the `DEVICE_TYPE == "cuda"` branch is touched.
- The `hip` (AMD ROCm) and `xpu` (Intel) branches are unchanged.
- On a real CUDA box running as root, behavior is byte-identical to
  main: same os.system() calls, same reload, same retry, same warnings.
  AST-verified by /tmp/verify_minimal/verify.py.
- `hasattr(os, "geteuid")` guards against Windows where `os.geteuid`
  doesn't exist.

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

---------

Co-authored-by: Daniel Han <info@unsloth.ai>
Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com>
---
 unsloth/__init__.py | 107 +++++++++++++++++++++++++++-----------------
 1 file changed, 65 insertions(+), 42 deletions(-)

diff --git a/unsloth/__init__.py b/unsloth/__init__.py
index f922a9fa44..73a9ce8fef 100644
--- a/unsloth/__init__.py
+++ b/unsloth/__init__.py
@@ -247,53 +247,76 @@ if DEVICE_TYPE == "cuda":
         cdequantize_blockwise_fp32 = bnb.functional.lib.cdequantize_blockwise_fp32
         libcuda_dirs()
     except:
-        warnings.warn("Unsloth: Running `ldconfig /usr/lib64-nvidia` to link CUDA.")
+        # Only run the ldconfig recovery when we can actually run
+        # ldconfig (root). On non-root environments (shared HPC,
+        # locked-down containers, CI runners, etc.) the recovery would
+        # shell out to `ldconfig` and fail with "Permission denied",
+        # which is especially noisy for users who don't even have
+        # bitsandbytes installed and are just doing 16bit/full
+        # finetuning. libcuda_dirs() is used by both triton and bnb,
+        # so we still run the recovery whenever we're root, regardless
+        # of whether bnb is installed.
+        if hasattr(os, "geteuid") and os.geteuid() == 0:
+            warnings.warn("Unsloth: Running `ldconfig /usr/lib64-nvidia` to link CUDA.")
 
-        if os.path.exists("/usr/lib64-nvidia"):
-            os.system("ldconfig /usr/lib64-nvidia")
-        elif os.path.exists("/usr/local"):
-            # Sometimes bitsandbytes cannot be linked properly in Runpod for example
-            possible_cudas = (
-                subprocess.check_output(["ls", "-al", "/usr/local"])
-                .decode("utf-8")
-                .split("\n")
-            )
-            find_cuda = re.compile(r"[\s](cuda\-[\d\.]{2,})$")
-            possible_cudas = [find_cuda.search(x) for x in possible_cudas]
-            possible_cudas = [x.group(1) for x in possible_cudas if x is not None]
+            if os.path.exists("/usr/lib64-nvidia"):
+                os.system("ldconfig /usr/lib64-nvidia")
+            elif os.path.exists("/usr/local"):
+                # Sometimes bitsandbytes cannot be linked properly in Runpod for example
+                possible_cudas = (
+                    subprocess.check_output(["ls", "-al", "/usr/local"])
+                    .decode("utf-8")
+                    .split("\n")
+                )
+                find_cuda = re.compile(r"[\s](cuda\-[\d\.]{2,})$")
+                possible_cudas = [find_cuda.search(x) for x in possible_cudas]
+                possible_cudas = [x.group(1) for x in possible_cudas if x is not None]
 
-            # Try linking cuda folder, or everything in local
-            if len(possible_cudas) == 0:
-                os.system("ldconfig /usr/local/")
-            else:
-                find_number = re.compile(r"([\d\.]{2,})")
-                latest_cuda = np.argsort(
-                    [float(find_number.search(x).group(1)) for x in possible_cudas]
-                )[::-1][0]
-                latest_cuda = possible_cudas[latest_cuda]
-                os.system(f"ldconfig /usr/local/{latest_cuda}")
-                del find_number, latest_cuda
-            del possible_cudas, find_cuda
+                # Try linking cuda folder, or everything in local
+                if len(possible_cudas) == 0:
+                    os.system("ldconfig /usr/local/")
+                else:
+                    find_number = re.compile(r"([\d\.]{2,})")
+                    latest_cuda = np.argsort(
+                        [float(find_number.search(x).group(1)) for x in possible_cudas]
+                    )[::-1][0]
+                    latest_cuda = possible_cudas[latest_cuda]
+                    os.system(f"ldconfig /usr/local/{latest_cuda}")
+                    del find_number, latest_cuda
+                del possible_cudas, find_cuda
 
-        if bnb is not None:
-            importlib.reload(bnb)
-        importlib.reload(triton)
-        try:
-            libcuda_dirs = lambda: None
-            if Version(triton.__version__) >= Version("3.0.0"):
-                try:
-                    from triton.backends.nvidia.driver import libcuda_dirs
-                except:
-                    pass
-            else:
-                from triton.common.build import libcuda_dirs
-            cdequantize_blockwise_fp32 = bnb.functional.lib.cdequantize_blockwise_fp32
-            libcuda_dirs()
-        except:
+            if bnb is not None:
+                importlib.reload(bnb)
+            importlib.reload(triton)
+            try:
+                libcuda_dirs = lambda: None
+                if Version(triton.__version__) >= Version("3.0.0"):
+                    try:
+                        from triton.backends.nvidia.driver import libcuda_dirs
+                    except:
+                        pass
+                else:
+                    from triton.common.build import libcuda_dirs
+                cdequantize_blockwise_fp32 = (
+                    bnb.functional.lib.cdequantize_blockwise_fp32
+                )
+                libcuda_dirs()
+            except:
+                warnings.warn(
+                    "Unsloth: CUDA is not linked properly.\n"
+                    "Try running `python -m bitsandbytes` then `python -m xformers.info`\n"
+                    "We tried running `ldconfig /usr/lib64-nvidia` ourselves, but it didn't work.\n"
+                    "You need to run in your terminal `sudo ldconfig /usr/lib64-nvidia` yourself, then import Unsloth.\n"
+                    "Also try `sudo ldconfig /usr/local/cuda-xx.x` - find the latest cuda version.\n"
+                    "Unsloth will still run for now, but maybe it might crash - let's hope it works!"
+                )
+        elif bnb is not None:
+            # Non-root + bnb installed: we can't run ldconfig ourselves,
+            # but bnb is going to crash later when the user actually uses
+            # 4bit quantization - tell them how to fix it manually so
+            # they're not surprised by an opaque error down the road.
             warnings.warn(
                 "Unsloth: CUDA is not linked properly.\n"
-                "Try running `python -m bitsandbytes` then `python -m xformers.info`\n"
-                "We tried running `ldconfig /usr/lib64-nvidia` ourselves, but it didn't work.\n"
                 "You need to run in your terminal `sudo ldconfig /usr/lib64-nvidia` yourself, then import Unsloth.\n"
                 "Also try `sudo ldconfig /usr/local/cuda-xx.x` - find the latest cuda version.\n"
                 "Unsloth will still run for now, but maybe it might crash - let's hope it works!"

From 7aa442289bac98998cf4062bab604683659729f8 Mon Sep 17 00:00:00 2001
From: cheehook <chee.hoo.kok@intel.com>
Date: Thu, 9 Apr 2026 19:38:44 +0800
Subject: [PATCH 063/216] Fix Mistral DPO/preference training crash on
 non-xformers platforms (e.g. Intel XPU) (#4889)

* Fix Mistral training crash when xformers is unavailable

* Fix/adjust Mistral DPO training crash fix for PR #4889

- Clarify comment in MistralForCausalLM_fast_forward: the DPO embed-masking
  block runs BEFORE attention_mask is nulled out, and it is the consumer that
  requires a 2D mask.
- Add defensive attention_mask.ndim == 2 guard to the LlamaModel_fast_forward
  DPO embed-masking block so it self-protects if a 4D mask ever reaches it.

---------

Co-authored-by: Daniel Han <danielhanchen@gmail.com>
---
 unsloth/models/llama.py   |  1 +
 unsloth/models/mistral.py | 10 ++++++++++
 2 files changed, 11 insertions(+)

diff --git a/unsloth/models/llama.py b/unsloth/models/llama.py
index 2f61913550..031cabcd39 100644
--- a/unsloth/models/llama.py
+++ b/unsloth/models/llama.py
@@ -1030,6 +1030,7 @@ def LlamaModel_fast_forward(
     if (
         getattr(self, "_has_no_labels", False) is True
         and (attention_mask is not None)
+        and attention_mask.ndim == 2
         and (past_key_values is None)
         and (not train_embed_tokens)
         and self.training
diff --git a/unsloth/models/mistral.py b/unsloth/models/mistral.py
index 83e9ab9486..d42e906604 100644
--- a/unsloth/models/mistral.py
+++ b/unsloth/models/mistral.py
@@ -188,6 +188,16 @@ def MistralForCausalLM_fast_forward(
 
             # If attention_mask exists, it will be handled in the attention forward
 
+        elif self.training:
+            # During training, LlamaModel_fast_forward's DPO embed-masking
+            # block requires a 2D attention_mask (it does
+            # inputs_embeds *= attention_mask.unsqueeze(0).transpose(0, 1).transpose(1, 2)).
+            # Afterwards, LlamaModel_fast_forward sets attention_mask=None
+            # before the attention layers anyway, so leaving the 2D mask
+            # untouched here is safe and avoids converting to 4D (which would
+            # crash the DPO block).
+            pass
+
         else:
             # Not using xformers - need to create attention masks
             if (

From ad5972492da5358950083a0de9f0ebc757ce0b4f Mon Sep 17 00:00:00 2001
From: kiankyars <69437137+kiankyars@users.noreply.github.com>
Date: Thu, 9 Apr 2026 04:45:43 -0700
Subject: [PATCH 064/216] Fix raw text paragraph break normalization (#4884)

* Fix raw text paragraph break normalization

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

* Normalize horizontal whitespace before stripping non-ASCII and collapse leftover doubles

Run the [^\S\n]+ horizontal-whitespace collapse before the non-ASCII strip
so that Unicode whitespace (\u00A0, \u202F, \u2009, \u3000, \v, \f, etc.)
becomes a single ASCII space instead of being deleted outright. The prior
ordering silently merged adjacent words on HTML/PDF/OCR-sourced text:
"hello\u00a0world" used to produce "helloworld" after this PR; it now
produces "hello world".

Also drop \t from the allow-list since the horizontal-whitespace collapse
already normalizes tabs to a single space, and add a targeted [ ]{2,} pass
right after the non-ASCII strip so that a non-whitespace non-ASCII character
sitting between two spaces ("word1 (c) word2") does not leave an interior
double space. Without this extra pass, clean_text was not idempotent on
such inputs: the first call produced "word1  word2" and only the second
call collapsed it to "word1 word2". Fuzz testing over 10000 random inputs
now satisfies the idempotence invariant in every case.

* Add regression tests for Unicode/control whitespace and non-ASCII edge cases

Cover:
- Unicode horizontal whitespace separators (NBSP, narrow NBSP, thin space,
  en/em space, ideographic space, vertical tab, form feed) normalizing to
  a single ASCII space instead of being deleted.
- Mixed paragraph + Unicode whitespace realistic input ("Section\u00a01\r\n\r\nBody\ftext\u202Fhere").
- Tab collapsing and space trimming around newlines.
- Non-whitespace non-ASCII characters (copyright, accented letters, emoji)
  sitting between spaces: must not leave an interior double space, and
  clean_text must be idempotent on these inputs.
- Non-ASCII characters adjacent to a newline: stripping must not leave
  stray leading or trailing spaces on the neighbouring line, and must not
  swallow an adjacent paragraph break.

---------

Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com>
Co-authored-by: Daniel Han <danielhanchen@gmail.com>
---
 tests/test_raw_text.py       | 71 ++++++++++++++++++++++++++++++++++++
 unsloth/dataprep/raw_text.py |  6 ++-
 2 files changed, 75 insertions(+), 2 deletions(-)

diff --git a/tests/test_raw_text.py b/tests/test_raw_text.py
index 9f2e8cda4e..604326976f 100644
--- a/tests/test_raw_text.py
+++ b/tests/test_raw_text.py
@@ -149,6 +149,77 @@ def test_raw_text_loader():
         preprocessor = TextPreprocessor()
         clean_text = preprocessor.clean_text("  messy   text  \n\n\n  ")
         assert "messy text" in clean_text, "Should clean text properly"
+        paragraph_text = preprocessor.clean_text("Line 1\r\n\r\n\r\nLine 2")
+        assert (
+            paragraph_text == "Line 1\n\nLine 2"
+        ), "Should preserve paragraph breaks while normalizing newlines"
+
+        # Non-ASCII horizontal whitespace separators (NBSP, thin space,
+        # ideographic space, narrow NBSP, em space, vertical tab, form feed)
+        # should be normalized to a single ASCII space, not deleted, otherwise
+        # adjacent words get silently fused together on HTML/PDF/OCR inputs.
+        unicode_whitespace_cases = [
+            ("hello\u00a0world", "hello world"),
+            ("hello\u202fworld", "hello world"),
+            ("hello\u2009world", "hello world"),
+            ("hello\u3000world", "hello world"),
+            ("hello\u2002world", "hello world"),
+            ("hello\x0bworld", "hello world"),
+            ("hello\x0cworld", "hello world"),
+        ]
+        for raw, expected in unicode_whitespace_cases:
+            assert preprocessor.clean_text(raw) == expected, (
+                f"Should normalize Unicode/control whitespace to a single space "
+                f"for {raw!r}"
+            )
+
+        # Mixed paragraph + Unicode whitespace realistic input
+        mixed = preprocessor.clean_text(
+            "Section\u00a01\r\n\r\nBody\ftext\u202Fhere"
+        )
+        assert mixed == "Section 1\n\nBody text here", (
+            "Should preserve paragraph breaks and normalize Unicode "
+            "whitespace simultaneously"
+        )
+
+        # Tabs should collapse to a single space
+        assert preprocessor.clean_text("a\tb") == "a b"
+        assert preprocessor.clean_text("a\t\tb") == "a b"
+
+        # Spaces around newlines should be trimmed on both sides, even with
+        # multiple consecutive newlines
+        assert preprocessor.clean_text("foo \n\n bar") == "foo\n\nbar"
+
+        # Non-whitespace non-ASCII characters sitting between spaces should
+        # not leave an interior double space after being stripped. This
+        # guards the idempotence invariant too: without the extra collapse
+        # pass, "word1 (c) word2" first reduces to "word1  word2" and only
+        # becomes "word1 word2" on a second call.
+        assert preprocessor.clean_text("word1 \u00a9 word2") == "word1 word2"
+        assert preprocessor.clean_text("a \u00e9 b") == "a b"
+        assert preprocessor.clean_text("prefix \U0001F600 suffix") == "prefix suffix"
+
+        # Stripping a non-ASCII character adjacent to a newline must not
+        # leave a stray leading/trailing space on the neighbouring line.
+        assert preprocessor.clean_text("foo \u00e9\nbar") == "foo\nbar"
+        assert preprocessor.clean_text("foo\n\u00e9 bar") == "foo\nbar"
+        # The double-space collapse pass must not swallow a legitimate
+        # paragraph break when a non-ASCII char sits near it.
+        assert preprocessor.clean_text("a \u00a9\n\nb") == "a\n\nb"
+
+        # Idempotence: running clean_text twice should give the same result
+        idempotent_inputs = [
+            "  messy   text  \n\n\n  ",
+            "Line 1\r\n\r\n\r\nLine 2",
+            "hello\u00a0world",
+            "Section\u00a01\r\n\r\nBody\ftext\u202Fhere",
+            "word1 \u00a9 word2",
+            "a \u00e9 b",
+        ]
+        for raw in idempotent_inputs:
+            once = preprocessor.clean_text(raw)
+            twice = preprocessor.clean_text(once)
+            assert once == twice, f"clean_text should be idempotent for {raw!r}"
 
         # Test validation
         stats = preprocessor.validate_dataset(text_dataset)
diff --git a/unsloth/dataprep/raw_text.py b/unsloth/dataprep/raw_text.py
index ba010edabb..1e2221dc08 100644
--- a/unsloth/dataprep/raw_text.py
+++ b/unsloth/dataprep/raw_text.py
@@ -244,9 +244,11 @@ class RawTextDataLoader:
 class TextPreprocessor:
     def clean_text(self, text):
         """Remove unwanted characters, normalize whitespace"""
-        text = re.sub(r"\s+", " ", text)
-        text = re.sub(r"[^\x20-\x7E\n\t]", "", text)
         text = text.replace("\r\n", "\n").replace("\r", "\n")
+        text = re.sub(r"[^\S\n]+", " ", text)
+        text = re.sub(r"[^\x20-\x7E\n]", "", text)
+        text = re.sub(r"[ ]{2,}", " ", text)
+        text = re.sub(r" *\n *", "\n", text)
         text = re.sub(r"\n{3,}", "\n\n", text)
         return text.strip()
 

From dc16e0c65b485ee07f1f0c9420e665b4added534 Mon Sep 17 00:00:00 2001
From: Lee Jackson <130007945+Imagineer99@users.noreply.github.com>
Date: Thu, 9 Apr 2026 14:00:56 +0100
Subject: [PATCH 065/216] Studio: keep chat input visible and fix compare pane
 clipping (#4924)

* fix(chat): sticky composer bar in thread

* fix(chat): fix compare pane clipping

* fix(chat): tighten scroll-to-bottom placement and compare footer spacing

* Fix TypeScript build break and clean up ViewportFooter classes

- Remove unused `compact` prop from ThreadScrollToBottom call site
  (component is FC with no props, passing it caused TS2322)
- Extract shared classes (sticky, bottom-0, z-20, bg-transparent) from
  ternary branches into the unconditional className string
- Restore `relative` on normal-mode footer so the inner absolute
  bg-background strip has a positioning context
- Remove redundant md:pb-3 / md:pb-4 (same value as base pb-3 / pb-4)
- Remove no-op `sticky bottom-0` from SharedComposer wrapper in both
  LoraCompareContent and GeneralCompareContent (flex layout with
  shrink-0 already pins it at the bottom; parent has no scrollable
  overflow for sticky to bind to)
- Fix truncated comment on pointer-events rationale

---------

Co-authored-by: Daniel Han <danielhanchen@gmail.com>
---
 .../src/components/assistant-ui/thread.tsx    | 25 ++++++++++++++-----
 .../frontend/src/features/chat/chat-page.tsx  |  4 +--
 2 files changed, 21 insertions(+), 8 deletions(-)

diff --git a/studio/frontend/src/components/assistant-ui/thread.tsx b/studio/frontend/src/components/assistant-ui/thread.tsx
index 4db0eda0ad..0d99654b89 100644
--- a/studio/frontend/src/components/assistant-ui/thread.tsx
+++ b/studio/frontend/src/components/assistant-ui/thread.tsx
@@ -89,14 +89,27 @@ export const Thread: FC<{ hideComposer?: boolean; hideWelcome?: boolean }> = ({
           }}
         />
 
-        <ThreadPrimitive.ViewportFooter className={cn(
-          "aui-thread-viewport-footer sticky bottom-0 z-20 mt-auto flex w-full flex-col gap-4 overflow-visible pb-4 md:pb-4",
-          hideComposer ? "bg-background" : "relative bg-transparent",
-        )}>
+        <ThreadPrimitive.ViewportFooter
+          className={cn(
+            "aui-thread-viewport-footer sticky bottom-0 z-20 mt-auto flex w-full flex-col overflow-visible bg-transparent",
+            hideComposer ? "gap-2" : "gap-4",
+            // Compare: pointer-events pass-through so messages behind footer stay clickable
+            hideComposer
+              ? "pointer-events-none pb-3"
+              : "relative pb-4",
+          )}
+        >
           {!hideComposer && (
             <div className="pointer-events-none absolute inset-x-0 bottom-0 h-4 bg-background" aria-hidden />
           )}
-          <ThreadScrollToBottom />
+          <div
+            className={cn(
+              "flex justify-center",
+              hideComposer && "pointer-events-auto",
+            )}
+          >
+            <ThreadScrollToBottom />
+          </div>
           <AuiIf condition={({ thread }) => !thread.isEmpty}>
             {!hideComposer && <ComposerAnimated />}
           </AuiIf>
@@ -112,7 +125,7 @@ const ThreadScrollToBottom: FC = () => {
       <TooltipIconButton
         tooltip="Scroll to bottom"
         variant="outline"
-        className="aui-thread-scroll-to-bottom absolute -top-12 z-10 self-center rounded-full p-4 disabled:invisible dark:bg-background dark:hover:bg-accent"
+        className="aui-thread-scroll-to-bottom absolute -top-6 z-10 self-center rounded-full p-4 disabled:invisible dark:bg-background dark:hover:bg-accent"
       >
         <ArrowDownIcon />
       </TooltipIconButton>
diff --git a/studio/frontend/src/features/chat/chat-page.tsx b/studio/frontend/src/features/chat/chat-page.tsx
index cf1ba11d7b..ebaab75c0f 100644
--- a/studio/frontend/src/features/chat/chat-page.tsx
+++ b/studio/frontend/src/features/chat/chat-page.tsx
@@ -251,7 +251,7 @@ const LoraCompareContent = memo(function LoraCompareContent({
             </div>
           </div>
         </div>
-        <div className="mx-auto w-full max-w-4xl px-4 py-4">
+        <div className="z-20 mx-auto w-full max-w-4xl shrink-0 border-t border-border/60 bg-background px-4 pt-2 pb-4">
           <SharedComposer handlesRef={handlesRef} />
         </div>
       </div>
@@ -387,7 +387,7 @@ const GeneralCompareContent = memo(function GeneralCompareContent({
             </div>
           </div>
         </div>
-        <div className="mx-auto w-full max-w-4xl px-4 py-4">
+        <div className="z-20 mx-auto w-full max-w-4xl shrink-0 border-t border-border/60 bg-background px-4 pt-2 pb-4">
           <SharedComposer
             handlesRef={handlesRef}
             model1={model1}

From d5525e8bbbd761f67021cf44a2dbf48f24b5cfc6 Mon Sep 17 00:00:00 2001
From: Ricardo-M-L <69202550+Ricardo-M-L@users.noreply.github.com>
Date: Thu, 9 Apr 2026 21:15:46 +0800
Subject: [PATCH 066/216] fix: check find() return value before adding offset
 in try_fix_tokenizer (#4923)

* fix: check find() return value before adding offset in try_fix_tokenizer

The `str.find()` result was checked for -1 only after adding
`len(find_text)`, turning the guard into dead code. When the substring
is absent, `start` becomes `len(find_text) - 1` (a positive number),
so the `if start == -1: continue` never triggers and the subsequent
slice extracts garbage from the tokenizer string.

Split the find and offset into two steps so the -1 check works correctly.

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

* Add defensive guards for token_id None and end find() returning -1

- Skip loop iteration early when token_id is None to avoid constructing
  a find_text that can never match valid JSON
- Guard end = tokenizer_string.find('",', start) against -1 to prevent
  silent garbage extraction from malformed tokenizer strings

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

---------

Co-authored-by: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Co-authored-by: Daniel Han <danielhanchen@gmail.com>
Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com>
---
 tests/test_raw_text.py     | 8 +++-----
 unsloth/tokenizer_utils.py | 9 +++++++--
 2 files changed, 10 insertions(+), 7 deletions(-)

diff --git a/tests/test_raw_text.py b/tests/test_raw_text.py
index 604326976f..d8289fed20 100644
--- a/tests/test_raw_text.py
+++ b/tests/test_raw_text.py
@@ -174,9 +174,7 @@ def test_raw_text_loader():
             )
 
         # Mixed paragraph + Unicode whitespace realistic input
-        mixed = preprocessor.clean_text(
-            "Section\u00a01\r\n\r\nBody\ftext\u202Fhere"
-        )
+        mixed = preprocessor.clean_text("Section\u00a01\r\n\r\nBody\ftext\u202fhere")
         assert mixed == "Section 1\n\nBody text here", (
             "Should preserve paragraph breaks and normalize Unicode "
             "whitespace simultaneously"
@@ -197,7 +195,7 @@ def test_raw_text_loader():
         # becomes "word1 word2" on a second call.
         assert preprocessor.clean_text("word1 \u00a9 word2") == "word1 word2"
         assert preprocessor.clean_text("a \u00e9 b") == "a b"
-        assert preprocessor.clean_text("prefix \U0001F600 suffix") == "prefix suffix"
+        assert preprocessor.clean_text("prefix \U0001f600 suffix") == "prefix suffix"
 
         # Stripping a non-ASCII character adjacent to a newline must not
         # leave a stray leading/trailing space on the neighbouring line.
@@ -212,7 +210,7 @@ def test_raw_text_loader():
             "  messy   text  \n\n\n  ",
             "Line 1\r\n\r\n\r\nLine 2",
             "hello\u00a0world",
-            "Section\u00a01\r\n\r\nBody\ftext\u202Fhere",
+            "Section\u00a01\r\n\r\nBody\ftext\u202fhere",
             "word1 \u00a9 word2",
             "a \u00e9 b",
         ]
diff --git a/unsloth/tokenizer_utils.py b/unsloth/tokenizer_utils.py
index 8be6bb5a5a..2145fa5da4 100644
--- a/unsloth/tokenizer_utils.py
+++ b/unsloth/tokenizer_utils.py
@@ -95,13 +95,18 @@ def try_fix_tokenizer(tokenizer, prepend = True):
         if token is None:
             continue
         token_id = getattr(tokenizer, token_name + "_id", None)
+        if token_id is None:
+            continue
 
         # Locate the token's id mapping in the string
         find_text = f'"id":{token_id},"content":"'
-        start = tokenizer_string.find(find_text) + len(find_text)
-        if start == -1:
+        find_pos = tokenizer_string.find(find_text)
+        if find_pos == -1:
             continue
+        start = find_pos + len(find_text)
         end = tokenizer_string.find('",', start)
+        if end == -1:
+            continue
 
         bad_token = tokenizer_string[start:end]
         # Check if token is the actual same one - if not, edit it

From bcf4fd6bd3e912a6f4a93a060367bd1d8de07256 Mon Sep 17 00:00:00 2001
From: Roland Tannous <115670425+rolandtannous@users.noreply.github.com>
Date: Thu, 9 Apr 2026 23:36:42 +0400
Subject: [PATCH 067/216] =?UTF-8?q?updated=20models=20template=20mappers.?=
 =?UTF-8?q?=20added=20lfm2.5vl450m=20to=20transformers=205=E2=80=A6=20(#49?=
 =?UTF-8?q?39)?=
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

* updated models template mappers. added lfm2.5vl450m to transformers 5.3.0 whitelist

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

---------

Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com>
---
 .../backend/utils/datasets/model_mappings.py  | 33 +++++++++++++++++++
 studio/backend/utils/transformers_version.py  |  1 +
 unsloth/models/mapper.py                      | 33 +++++++++++++++++++
 unsloth/ollama_template_mappers.py            |  6 ++--
 4 files changed, 69 insertions(+), 4 deletions(-)

diff --git a/studio/backend/utils/datasets/model_mappings.py b/studio/backend/utils/datasets/model_mappings.py
index 95b4791574..bcd4f4fc01 100644
--- a/studio/backend/utils/datasets/model_mappings.py
+++ b/studio/backend/utils/datasets/model_mappings.py
@@ -215,6 +215,21 @@ TEMPLATE_TO_MODEL_MAPPER = {
         "google/gemma-3n-E2B-it",
         "unsloth/gemma-3n-E2B-it-unsloth-bnb-4bit",
     ),
+    "gemma-4": (
+        "unsloth/gemma-4-E2B-it",
+        "google/gemma-4-E2B-it",
+        "unsloth/gemma-4-E4B-it",
+        "google/gemma-4-E4B-it",
+        "unsloth/gemma-4-E2B-it-unsloth-bnb-4bit",
+        "unsloth/gemma-4-E4B-it-unsloth-bnb-4bit",
+    ),
+    "gemma-4-thinking": (
+        "unsloth/gemma-4-26B-A4B-it",
+        "google/gemma-4-26B-A4B-it",
+        "unsloth/gemma-4-31B-it",
+        "unsloth/gemma-4-31B-it-unsloth-bnb-4bit",
+        "google/gemma-4-31B-it",
+    ),
     "qwen2.5": (
         "unsloth/Qwen2.5-0.5B-Instruct-unsloth-bnb-4bit",
         "unsloth/Qwen2.5-0.5B-Instruct",
@@ -399,6 +414,12 @@ TEMPLATE_TO_MODEL_MAPPER = {
         "THUDM/GLM-4.7-Flash",
         "unsloth/GLM-4.7-Flash-bnb-4bit",
     ),
+    "lfm-2": (
+        "unsloth/LFM2-1.2B",
+        "LiquidAI/LFM2-1.2B",
+        "unsloth/LFM2-1.2B-unsloth-bnb-4bit",
+    ),
+    "lfm-2.5": ("unsloth/LFM2.5-1.2B-Instruct",),
 }
 
 MODEL_TO_TEMPLATE_MAPPER = {}
@@ -414,6 +435,14 @@ for key, values in TEMPLATE_TO_MODEL_MAPPER.items():
 
 
 TEMPLATE_TO_RESPONSES_MAPPER = {
+    "gemma-4-thinking": {
+        "instruction": "<|turn>user\n",
+        "response": "<|turn>model\n",
+    },
+    "gemma-4": {
+        "instruction": "<|turn>user\n",
+        "response": "<|turn>model\n",
+    },
     "gemma-3": {
         "instruction": "<start_of_turn>user\n",
         "response": "<start_of_turn>model\n",
@@ -514,6 +543,10 @@ TEMPLATE_TO_RESPONSES_MAPPER = {
         "instruction": "<|im_start|>user\n",
         "response": "<|im_start|>assistant\n",
     },
+    "lfm-2.5": {
+        "instruction": "<|im_start|>user\n",
+        "response": "<|im_start|>assistant\n",
+    },
     "starling": {
         "instruction": "GPT4 Correct User: ",
         "response": "GPT4 Correct Assistant: ",
diff --git a/studio/backend/utils/transformers_version.py b/studio/backend/utils/transformers_version.py
index 0c13b5455b..36c3a4c22d 100644
--- a/studio/backend/utils/transformers_version.py
+++ b/studio/backend/utils/transformers_version.py
@@ -52,6 +52,7 @@ TRANSFORMERS_5_MODEL_SUBSTRINGS: tuple[str, ...] = (
     "qwen3.5",  # Qwen3.5 family (35B-A3B, etc.)
     "qwen3-next",  # Qwen3-Next and variants
     "tiny_qwen3_moe",  # imdatta0/tiny_qwen3_moe_2.8B_0.7B
+    "lfm2.5-vl-450m",  # LiquidAI/LFM2.5-VL-450M
 )
 
 # Lowercase substrings for models that require transformers 5.5.0 (checked first).
diff --git a/unsloth/models/mapper.py b/unsloth/models/mapper.py
index f0f430eb7e..f0b6737246 100644
--- a/unsloth/models/mapper.py
+++ b/unsloth/models/mapper.py
@@ -22,6 +22,39 @@ __all__ = [
 
 __INT_TO_FLOAT_MAPPER = \
 {
+    "unsloth/gemma-4-E2B-it-unsloth-bnb-4bit" : (
+        "unsloth/gemma-4-E2B-it",
+        "google/gemma-4-E2B-it",
+    ),
+    "unsloth/gemma-4-E4B-it-unsloth-bnb-4bit" : (
+        "unsloth/gemma-4-E4B-it",
+        "google/gemma-4-E4B-it",
+    ),
+    "unsloth/gemma-4-26B-A4B-it" : (
+        "unsloth/gemma-4-26B-A4B-it",
+        "google/gemma-4-26B-A4B-it",
+    ),
+    "unsloth/gemma-4-31B-it" : (
+        "unsloth/gemma-4-31B-it",
+        "google/gemma-4-31B-it",
+    ),
+    "unsloth/gemma-4-E2B" : (
+        "unsloth/gemma-4-E2B",
+        "google/gemma-4-E2B",
+    ),
+    "unsloth/gemma-4-E4B" : (
+        "unsloth/gemma-4-E4B",
+        "google/gemma-4-E4B",
+    ),
+    "unsloth/gemma-4-26B-A4B" : (
+        "unsloth/gemma-4-26B-A4B",
+        "google/gemma-4-26B-A4B",
+    ),
+    "unsloth/gemma-4-31B" : (
+        "unsloth/gemma-4-31B",
+        "google/gemma-4-31B",
+    ),
+
     "unsloth/mistral-7b-bnb-4bit" : (
         "unsloth/mistral-7b",
         "mistralai/Mistral-7B-v0.1",
diff --git a/unsloth/ollama_template_mappers.py b/unsloth/ollama_template_mappers.py
index 728b08813a..e3eb7f0735 100644
--- a/unsloth/ollama_template_mappers.py
+++ b/unsloth/ollama_template_mappers.py
@@ -1978,13 +1978,11 @@ OLLAMA_TEMPLATE_TO_MODEL_MAPPER = {
     ),
     "gemma4": (
         "unsloth/gemma-4-E2B-it",
-        "unsloth/gemma-4-E2B",
+        "unsloth/gemma-4-E2B-it-unsloth-bnb-4bit",
         "unsloth/gemma-4-E4B-it",
-        "unsloth/gemma-4-E4B",
+        "unsloth/gemma-4-E4B-it-unsloth-bnb-4bit",
         "unsloth/gemma-4-31B-it",
-        "unsloth/gemma-4-31B",
         "unsloth/gemma-4-26B-A4B-it",
-        "unsloth/gemma-4-26B-A4B",
     ),
     "gemma3n": (
         "unsloth/gemma-3n-E4B-it-unsloth-bnb-4bit",

From 33503ea2482819ad5aaf048f2edd0a69cb540500 Mon Sep 17 00:00:00 2001
From: Roland Tannous <115670425+rolandtannous@users.noreply.github.com>
Date: Fri, 10 Apr 2026 10:14:57 +0400
Subject: [PATCH 068/216] =?UTF-8?q?Revert=20"updated=20models=20template?=
 =?UTF-8?q?=20mappers.=20added=20lfm2.5vl450m=20to=20transformers=205?=
 =?UTF-8?q?=E2=80=A6"=20(#4945)?=
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

This reverts commit bcf4fd6bd3e912a6f4a93a060367bd1d8de07256.
---
 .../backend/utils/datasets/model_mappings.py  | 33 -------------------
 studio/backend/utils/transformers_version.py  |  1 -
 unsloth/models/mapper.py                      | 33 -------------------
 unsloth/ollama_template_mappers.py            |  6 ++--
 4 files changed, 4 insertions(+), 69 deletions(-)

diff --git a/studio/backend/utils/datasets/model_mappings.py b/studio/backend/utils/datasets/model_mappings.py
index bcd4f4fc01..95b4791574 100644
--- a/studio/backend/utils/datasets/model_mappings.py
+++ b/studio/backend/utils/datasets/model_mappings.py
@@ -215,21 +215,6 @@ TEMPLATE_TO_MODEL_MAPPER = {
         "google/gemma-3n-E2B-it",
         "unsloth/gemma-3n-E2B-it-unsloth-bnb-4bit",
     ),
-    "gemma-4": (
-        "unsloth/gemma-4-E2B-it",
-        "google/gemma-4-E2B-it",
-        "unsloth/gemma-4-E4B-it",
-        "google/gemma-4-E4B-it",
-        "unsloth/gemma-4-E2B-it-unsloth-bnb-4bit",
-        "unsloth/gemma-4-E4B-it-unsloth-bnb-4bit",
-    ),
-    "gemma-4-thinking": (
-        "unsloth/gemma-4-26B-A4B-it",
-        "google/gemma-4-26B-A4B-it",
-        "unsloth/gemma-4-31B-it",
-        "unsloth/gemma-4-31B-it-unsloth-bnb-4bit",
-        "google/gemma-4-31B-it",
-    ),
     "qwen2.5": (
         "unsloth/Qwen2.5-0.5B-Instruct-unsloth-bnb-4bit",
         "unsloth/Qwen2.5-0.5B-Instruct",
@@ -414,12 +399,6 @@ TEMPLATE_TO_MODEL_MAPPER = {
         "THUDM/GLM-4.7-Flash",
         "unsloth/GLM-4.7-Flash-bnb-4bit",
     ),
-    "lfm-2": (
-        "unsloth/LFM2-1.2B",
-        "LiquidAI/LFM2-1.2B",
-        "unsloth/LFM2-1.2B-unsloth-bnb-4bit",
-    ),
-    "lfm-2.5": ("unsloth/LFM2.5-1.2B-Instruct",),
 }
 
 MODEL_TO_TEMPLATE_MAPPER = {}
@@ -435,14 +414,6 @@ for key, values in TEMPLATE_TO_MODEL_MAPPER.items():
 
 
 TEMPLATE_TO_RESPONSES_MAPPER = {
-    "gemma-4-thinking": {
-        "instruction": "<|turn>user\n",
-        "response": "<|turn>model\n",
-    },
-    "gemma-4": {
-        "instruction": "<|turn>user\n",
-        "response": "<|turn>model\n",
-    },
     "gemma-3": {
         "instruction": "<start_of_turn>user\n",
         "response": "<start_of_turn>model\n",
@@ -543,10 +514,6 @@ TEMPLATE_TO_RESPONSES_MAPPER = {
         "instruction": "<|im_start|>user\n",
         "response": "<|im_start|>assistant\n",
     },
-    "lfm-2.5": {
-        "instruction": "<|im_start|>user\n",
-        "response": "<|im_start|>assistant\n",
-    },
     "starling": {
         "instruction": "GPT4 Correct User: ",
         "response": "GPT4 Correct Assistant: ",
diff --git a/studio/backend/utils/transformers_version.py b/studio/backend/utils/transformers_version.py
index 36c3a4c22d..0c13b5455b 100644
--- a/studio/backend/utils/transformers_version.py
+++ b/studio/backend/utils/transformers_version.py
@@ -52,7 +52,6 @@ TRANSFORMERS_5_MODEL_SUBSTRINGS: tuple[str, ...] = (
     "qwen3.5",  # Qwen3.5 family (35B-A3B, etc.)
     "qwen3-next",  # Qwen3-Next and variants
     "tiny_qwen3_moe",  # imdatta0/tiny_qwen3_moe_2.8B_0.7B
-    "lfm2.5-vl-450m",  # LiquidAI/LFM2.5-VL-450M
 )
 
 # Lowercase substrings for models that require transformers 5.5.0 (checked first).
diff --git a/unsloth/models/mapper.py b/unsloth/models/mapper.py
index f0b6737246..f0f430eb7e 100644
--- a/unsloth/models/mapper.py
+++ b/unsloth/models/mapper.py
@@ -22,39 +22,6 @@ __all__ = [
 
 __INT_TO_FLOAT_MAPPER = \
 {
-    "unsloth/gemma-4-E2B-it-unsloth-bnb-4bit" : (
-        "unsloth/gemma-4-E2B-it",
-        "google/gemma-4-E2B-it",
-    ),
-    "unsloth/gemma-4-E4B-it-unsloth-bnb-4bit" : (
-        "unsloth/gemma-4-E4B-it",
-        "google/gemma-4-E4B-it",
-    ),
-    "unsloth/gemma-4-26B-A4B-it" : (
-        "unsloth/gemma-4-26B-A4B-it",
-        "google/gemma-4-26B-A4B-it",
-    ),
-    "unsloth/gemma-4-31B-it" : (
-        "unsloth/gemma-4-31B-it",
-        "google/gemma-4-31B-it",
-    ),
-    "unsloth/gemma-4-E2B" : (
-        "unsloth/gemma-4-E2B",
-        "google/gemma-4-E2B",
-    ),
-    "unsloth/gemma-4-E4B" : (
-        "unsloth/gemma-4-E4B",
-        "google/gemma-4-E4B",
-    ),
-    "unsloth/gemma-4-26B-A4B" : (
-        "unsloth/gemma-4-26B-A4B",
-        "google/gemma-4-26B-A4B",
-    ),
-    "unsloth/gemma-4-31B" : (
-        "unsloth/gemma-4-31B",
-        "google/gemma-4-31B",
-    ),
-
     "unsloth/mistral-7b-bnb-4bit" : (
         "unsloth/mistral-7b",
         "mistralai/Mistral-7B-v0.1",
diff --git a/unsloth/ollama_template_mappers.py b/unsloth/ollama_template_mappers.py
index e3eb7f0735..728b08813a 100644
--- a/unsloth/ollama_template_mappers.py
+++ b/unsloth/ollama_template_mappers.py
@@ -1978,11 +1978,13 @@ OLLAMA_TEMPLATE_TO_MODEL_MAPPER = {
     ),
     "gemma4": (
         "unsloth/gemma-4-E2B-it",
-        "unsloth/gemma-4-E2B-it-unsloth-bnb-4bit",
+        "unsloth/gemma-4-E2B",
         "unsloth/gemma-4-E4B-it",
-        "unsloth/gemma-4-E4B-it-unsloth-bnb-4bit",
+        "unsloth/gemma-4-E4B",
         "unsloth/gemma-4-31B-it",
+        "unsloth/gemma-4-31B",
         "unsloth/gemma-4-26B-A4B-it",
+        "unsloth/gemma-4-26B-A4B",
     ),
     "gemma3n": (
         "unsloth/gemma-3n-E4B-it-unsloth-bnb-4bit",

From cad8c6ad05fdfd556a09fcbcff05343ae0843fc3 Mon Sep 17 00:00:00 2001
From: Daniel Han <danielhanchen@gmail.com>
Date: Fri, 10 Apr 2026 01:56:12 -0700
Subject: [PATCH 069/216] Add AMD ROCm/HIP support across installer and
 hardware detection (#4720)

* Add ROCm detection to install.sh and expand shell tests

Add AMD ROCm GPU detection to get_torch_index_url() in install.sh.
When nvidia-smi is not found, probe for ROCm via amd-smi, /opt/rocm
version file, hipconfig, dpkg-query, and rpm.

Includes validation guard for malformed _rocm_tag, Debian epoch prefix
stripping, ROCm 7.2+ cap to rocm7.1 index, bitsandbytes AMD install,
and status messaging. Shell tests expanded to 23 cases.

Co-authored-by: Daniel Han <danielhanchen@gmail.com>

* Add ROCm torch reinstall support to install_python_stack.py

Add _detect_rocm_version() and _ensure_rocm_torch() to detect when a
Linux host has ROCm but the venv received CPU-only torch, and reinstall
with the correct ROCm wheels. Covers ROCm 6.0 through 7.1 with a
30-second timeout on the torch GPU probe subprocess.

Co-authored-by: Daniel Han <danielhanchen@gmail.com>

* Add ROCm support to llama.cpp prebuilt installer

Add has_rocm field to HostInfo, extend detect_host() to probe for ROCm
via hipcc/amd-smi/rocm-smi/ROCM_PATH, and route ROCm hosts to upstream
prebuilts (Linux ROCm 7.2 prebuilt with source fallback, Windows HIP
prebuilt with CPU fallback). Add linux-rocm and windows-hip install
kinds to runtime_patterns_for_choice().

Co-authored-by: Daniel Han <danielhanchen@gmail.com>

* Add IS_ROCM hardware flag and fix AMD error message

Add IS_ROCM flag to hardware.py detect_hardware() (set when
torch.version.hip is present, DeviceType stays CUDA). Export IS_ROCM
from __init__.py. Add "rocm" key to get_package_versions().

Replace "We do not support AMD" error in tokenizer_utils.py with a
helpful message pointing to ROCm installation docs.

Co-authored-by: Daniel Han <danielhanchen@gmail.com>

* Add comprehensive ROCm support test suite (68 tests)

Add tests/studio/install/test_rocm_support.py covering all ROCm code
paths across install_llama_prebuilt.py, install_python_stack.py,
hardware.py, tokenizer_utils.py, and install.sh. All tests use mocks
and run without AMD hardware.

Covers: asset selection (11), runtime patterns (5), HostInfo (4),
ROCm version detection (9), torch reinstall (9), index mapping (8),
hardware flag (8), tokenizer message (2), install.sh structure (10),
and live regression (1).

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

* Harden ROCm support: probe error handling, version cap, validation

Address review findings from 8 independent reviewers:

- Wrap _ensure_rocm_torch() torch probe in try/except for
  TimeoutExpired and OSError so a hung or broken torch import does not
  crash the installer (8/8 reviewers flagged this)
- Add torch>=2.4,<2.11.0 version cap to the ROCm reinstall path to
  prevent installing unsupported torch 2.11.0 from the rocm7.1 index
- Use with-statement for file reads in _detect_rocm_version() to avoid
  resource leaks
- Handle ROCM_PATH="" correctly (use `or "/opt/rocm"` instead of
  default parameter to avoid relative path resolution)
- Strengthen shell validation guard from rocm[0-9] to rocm[1-9] to
  reject rocm0.x tags that would produce nonexistent PyTorch index URLs
- Switch shell version cap from blocklist to allowlist (rocm6.*|rocm7.0*
  |rocm7.1* pass through, everything else caps to rocm7.1) so future
  ROCm 10+ does not fall through to a nonexistent index
- Add sorted() to _ROCM_TORCH_INDEX lookup for defensive ordering
- Fix test_probe_timeout_handled: replace zero-assertion test with
  proper assertions verifying reinstall proceeds after timeout

* Clean up rocm_paths list construction in detect_host()

Filter None from the ROCM_PATH env var lookup at list construction time
instead of relying on the inline `if p` guard in the any() call.

* Require actual AMD GPU presence before selecting ROCm paths

All 8 reviewers across 2 cycles independently flagged that ROCm
detection used toolkit/filesystem hints (hipcc, /opt/rocm, rocm-core)
as a proxy for GPU presence, which would misroute CPU-only or NVIDIA
hosts that happen to have ROCm tools installed.

Now all 3 detection points (install.sh, install_python_stack.py,
install_llama_prebuilt.py) probe for an actual AMD GPU before
entering the ROCm path:

- install.sh: check rocminfo for gfx* GPU names, or amd-smi list
  for device rows, before version detection
- install_python_stack.py: new _has_rocm_gpu() function probes
  rocminfo and amd-smi list before _ensure_rocm_torch() proceeds
- install_llama_prebuilt.py: detect_host() probes rocminfo/amd-smi
  list instead of just checking tool existence or directory paths

Also:
- Shell test mock amd-smi now handles "list" subcommand
- Python tests updated to mock _has_rocm_gpu where needed
- Added test_no_gpu_with_rocm_tools_skips to verify the new guard
- Test index lookups now use sorted() to match production code

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

* Harden hipconfig version parsing and torch probe compatibility

- Add parts[1].isdigit() check in hipconfig version parsing to handle
  versions like "6.3-HIP" where the minor component has non-numeric
  suffix (strip "-" prefix before int() conversion)
- Use getattr() in torch probe subprocess to safely handle old or
  custom torch builds that may lack torch.version.hip/cuda attributes

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

* Strengthen AMD GPU detection and add NVIDIA precedence guard

- Change amd-smi list detection from any-non-empty-output to requiring
  "gpu" marker in output, matching the shell-side NR>1 check. Prevents
  false positives from header-only amd-smi list output.
- Add nvidia-smi check at the top of _ensure_rocm_torch() so mixed
  AMD+NVIDIA hosts preserve NVIDIA precedence (matching install.sh and
  install_llama_prebuilt.py behavior).
- Apply the same amd-smi marker fix to install_llama_prebuilt.py
  detect_host() for consistency.

* Add Windows-specific ROCm/HIP detection in detect_host()

The previous detect_host() ROCm check used rocminfo and amd-smi list
which are Linux-only tools. On Windows, has_rocm would always be False,
making the Windows HIP prebuilt path at line 1794 unreachable.

Now detect_host() uses platform-specific detection:
- Linux: rocminfo (check for gfx GPU names) or amd-smi list
- Windows: hipinfo.exe, amd-smi, or amdhip64.dll on PATH

This allows Windows AMD users to get the HIP prebuilt binary instead
of silently falling through to the CPU prebuilt.

* Add AMD ROCm gaps: Mamba/SSM source builds, GPU monitoring, Windows messaging, RDNA expansion

- worker.py: Add HIP detection to causal-conv1d/mamba-ssm probe, check
  for hipcc before ROCm source builds, improve status messages and error
  reporting, add timeout and uv support for the source build fallback
- amd.py: New AMD GPU monitoring module via amd-smi metric --json,
  mirroring nvidia.py structure (utilization, temperature, power, VRAM)
- hardware.py: Branch to amd.py when IS_ROCM is True for GPU utilization,
  visible GPU queries, and physical GPU count
- install_python_stack.py: Detect AMD GPUs on Windows and warn that
  ROCm-enabled PyTorch must be installed manually
- kernels/utils.py: Expand is_rdna() to cover RDNA2 (gfx1030-1032),
  RDNA3 (gfx1102-1103), RDNA3.5 (gfx1150-1152) alongside existing entries
- tests: Add 32 new tests covering all changes (95/95 pass)

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

* Harden ROCm detection, fix VRAM heuristic, and expand RDNA2 coverage

- Windows ROCm detection: validate actual GPU presence via hipinfo/amd-smi
  output markers instead of just checking tool existence on PATH
- _ensure_rocm_torch: validate nvidia-smi actually reports a GPU before
  giving NVIDIA precedence (fixes AMD-only hosts with stale NVIDIA tools)
- amd.py _parse_numeric: handle dict-shaped metric objects from newer
  amd-smi versions ({"value": 10, "unit": "W"}) and strip MiB/GiB units
- amd.py VRAM heuristic: raise threshold from 100k to 10M to correctly
  handle MI300X (192 GB = 196608 MB) and other high-VRAM GPUs
- amd.py visible GPU: use AMD-reported GPU IDs instead of enumerate index
  so non-dense sets like CUDA_VISIBLE_DEVICES=1,3 report correctly
- install.sh: add ROCm <6.0 minimum version guard (no PyTorch wheels
  exist for older versions); fix rocm7.1* glob to not match rocm7.10+
- is_rdna: add gfx1033-1036 for RDNA2 mobile GPUs (RX 6600M etc.)
- worker.py: increase ROCm source build timeout from 600s to 1800s;
  fix success log message for ROCm source builds
- Tests: update mocks for _has_usable_nvidia_gpu, add RDNA2 target asserts

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

* Add HIP_VISIBLE_DEVICES support, unit-aware VRAM parsing, Windows GPU validation

- hardware.py: check HIP_VISIBLE_DEVICES and ROCR_VISIBLE_DEVICES on ROCm
  before falling back to CUDA_VISIBLE_DEVICES, so multi-GPU AMD setups with
  HIP-specific env vars report the correct visible device set
- amd.py: add _parse_memory_mb() that reads "unit" from dict-shaped amd-smi
  JSON (e.g. {"value": 192, "unit": "GiB"}) and converts to MB correctly;
  fixes MI300X VRAM misreported as 0.19 GB instead of 192 GB
- install_python_stack.py: Windows AMD warning now validates actual GPU
  presence via hipinfo/amd-smi output markers before printing
- install_llama_prebuilt.py: restore amdhip64.dll fallback for Windows HIP
  detection after tool-based checks, so Windows HIP installs without CLI
  tools on PATH are still detected
- hardware.py: fix IS_ROCM comment to accurately describe its role

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

* Fix HIP_VISIBLE_DEVICES empty-string handling in GPU visibility spec

Use explicit None checks instead of Python `or` operator when reading
HIP_VISIBLE_DEVICES / ROCR_VISIBLE_DEVICES, so that an empty string
("") is correctly honored as "no visible GPUs" rather than silently
falling through to CUDA_VISIBLE_DEVICES on mixed ROCm+CUDA systems.

* Fix IS_ROCM test assertion for multi-line formatting

* Cap torchvision/torchaudio versions, remove amdhip64.dll fallback, fix visible GPU count

- Cap torchvision<0.26.0 and torchaudio<2.11.0 alongside torch<2.11.0 in
  both install.sh and install_python_stack.py to prevent resolver from
  selecting incompatible companion packages from ROCm wheel index
- Remove amdhip64.dll fallback in Windows ROCm detection (DLL presence
  without hipinfo/amd-smi is not proof of GPU existence)
- Fix get_visible_gpu_count() to use _get_parent_visible_gpu_spec() which
  respects HIP_VISIBLE_DEVICES/ROCR_VISIBLE_DEVICES on ROCm hosts

* Attribute is_rdna() RDNA2/3/3.5/4 expansion to PR #4428

The is_rdna() expansion to cover RDNA2 (gfx1030-1036), RDNA3
(gfx1100-1103), RDNA3.5 (gfx1150-1152), and RDNA4 (gfx1200-1201)
architectures is based on the original work from PR #4428.

Co-authored-by: GoldenGrapeGentleman <yueyuan@amd.com>
Co-authored-by: billishyahao <bill.he@amd.com>

* Support AMD Radeon for studio (#4770)

Co-authored-by: Iswarya Alex <iswarya.alex@amd.com>

* Remove ROCm test files from main PR

Move test_rocm_support.py and shell test additions to a separate PR
to keep the main ROCm support PR focused on implementation changes.

* Fix installer and hardware detection issues for PR #4720

- Fix empty _tri_arg passed to uv pip install in Radeon path (causes
  "Empty field is not allowed for PEP508" error)
- Fix Radeon fallback: use ROCm index instead of CPU-only when
  repo.radeon.com is unreachable (TORCH_INDEX_URL already has ROCm)
- Use $TORCH_CONSTRAINT in fallback paths instead of hardcoded strings
- Fix _pick_radeon_wheel: relax suffix to match manylinux_2_28_x86_64
  wheels (AMD Radeon repo does not use bare linux_x86_64 platform tag)
- Fix IS_ROCM export: use __getattr__ so callers always see the live
  value after detect_hardware() runs
- Fix apply_gpu_ids: set HIP_VISIBLE_DEVICES and ROCR_VISIBLE_DEVICES
  on ROCm so _get_parent_visible_gpu_spec picks up narrowed GPU set
- Fix _parse_memory_mb: distinguish GB (1000 MB) from GiB (1024 MiB)
- Add amd-smi version as a fallback in _detect_rocm_version
- Fix trailing whitespace and missing newline at EOF in install.sh

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

* Fix GPU detection false positives and add missing health groups

- Fix _has_rocm_gpu() false positive: require "GPU: <number>" data rows
  from amd-smi list, not just header containing "gpu"
- Apply same fix in detect_host() in install_llama_prebuilt.py
- Add runtime_payload_health_groups for linux-rocm and windows-hip so
  partial/corrupt ROCm/HIP prebuilt installs are properly detected
- Add bitsandbytes install to Radeon fallback paths (was only in the
  success path, skipped when repo.radeon.com was unreachable)
- Keep DEVICE/CHAT_ONLY as direct imports in __init__.py (matching main)
  and only use __getattr__ for IS_ROCM

* Fix _ensure_rocm_torch and Windows AMD warning false positives

- _ensure_rocm_torch: only skip when HIP is already present, not for
  CUDA builds (which are unusable on AMD-only hosts). Fixes the case
  where a venv has a stale CUDA wheel and the repair step is skipped.
- Windows AMD warning: use GPU data row check (same as Linux fix) to
  avoid false positives from amd-smi list header-only output.

* Fix amd-smi GPU detection for GPU[N] output format

Older amd-smi versions output "GPU[0] : Card series: ..." instead of
"GPU: 0". The regex now matches both "GPU: <digit>" and "GPU[<digit>"
formats to detect actual GPU data rows.

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

* Harden AMD GPU detection against false positives

- install.sh: replace weak amd-smi list check (awk 'NR>1 && NF') with
  strict pattern matching GPU data rows (/^GPU[[:space:]]*[:\[]/)
- All files: reject rocminfo gfx000 (CPU HSA agent) by requiring
  gfx[1-9] instead of gfx[0-9] in the rocminfo GPU probe
- Fixes false positives on hosts with ROCm tools but no AMD GPU

* Remove duplicate comment from pre-commit merge

* Refactor: deduplicate AMD detection, consolidate bitsandbytes, clean up imports

- Extract _has_amd_rocm_gpu() shell function to avoid duplicating the
  rocminfo/amd-smi GPU detection logic in get_torch_index_url and
  the Radeon auto-detect block
- Consolidate bitsandbytes install into a single case block after torch
  install (was duplicated 4 times across Radeon success/fallback paths)
- Move math and re imports to top of amd.py (were inline in functions)
- Add _smi_query() helper in hardware.py to centralize IS_ROCM backend
  selection for get_gpu_utilization and get_visible_gpu_utilization

Addresses Gemini code review suggestions.

* Fix VRAM parsing for string values and GB/GiB consistency

- Extract unit from string-valued VRAM fields (e.g. "192 GiB") so
  _parse_memory_mb correctly applies the unit multiplier instead of
  treating the value as bare MB
- Treat GB and GiB identically (both as binary x1024) since GPU tools
  including amd-smi use binary units even when labeling them "GB"
- Fixes incorrect VRAM reporting on MI300-class cards (was showing
  ~0.19 GB instead of 192 GB for string-valued outputs)

* Add --no-cache to uv for ROCm HIP source builds

Avoid stale cache artifacts from partial HIP source builds when
uv is used for causal-conv1d/mamba-ssm compilation on ROCm.
The pip path already uses --no-cache-dir; this adds the uv equivalent
(--no-cache) only when is_hip is True.

* Fix critical: initialize _amd_gpu_radeon before case block

_amd_gpu_radeon was only set inside the */rocm*) case arm, so on
NVIDIA/CPU/macOS paths where TORCH_INDEX_URL does not contain "rocm",
the variable was unbound. With set -u (nounset) enabled, this crashes
the installer for every non-AMD user.

Move initialization to before the case block so it is always defined.

* Fix Windows AMD: route has_rocm hosts to HIP prebuilt path

resolve_release_asset_choice was selecting windows-cpu for all Windows
x86_64 hosts including those with has_rocm=True. Windows AMD users
should fall through to resolve_upstream_asset_choice which tries the
HIP prebuilt first. Add "not host.has_rocm" guard to the published
windows-cpu selection.

* Harden ROCm detection, Radeon wheel fallback, and HIP visibility

Addresses review findings from parallel reviewers on PR #4720:

- install.sh: add _has_usable_nvidia_gpu() helper requiring nvidia-smi -L
  to actually list a GPU before treating the host as NVIDIA. Fixes the
  stale-nvidia-smi-on-PATH regression where AMD-only hosts fell into the
  CUDA branch.
- install.sh: fix hipconfig awk blocks to propagate a non-zero exit code
  when the output is not a recognisable version string, so the ||-chain
  continues to dpkg-query / rpm instead of terminating early.
- install.sh: fail-closed on Radeon wheel fallback. When torch,
  torchvision or torchaudio is missing from the Radeon repo for the
  active Python tag, fall back to the standard ROCm index instead of
  silently mixing Radeon wheels with PyPI defaults. Quote all wheel
  arguments individually so wheel filenames cannot be word-split or
  glob-expanded.
- install_llama_prebuilt.py: detect_host() now requires nvidia-smi -L to
  list a GPU before setting has_physical_nvidia. Routes AMD ROCm hosts
  with a broken leftover nvidia-smi to the ROCm path instead of
  misclassifying them as NVIDIA.
- install_llama_prebuilt.py: scan upstream assets for any rocm-<version>
  prebuilt instead of hard-coding rocm-7.2, so ROCm 6.x / 7.0 / 7.1 / 7.3+
  users pick up a matching upstream prebuilt when one exists.
- install_llama_prebuilt.py: validate_server() adds --n-gpu-layers 1 for
  linux-rocm and windows-hip hosts, so new HIP prebuilts are preflighted
  on the GPU path instead of passing validation on CPU only.
- install_llama_prebuilt.py: restore the published windows-cpu fallback
  for AMD Windows hosts without a HIP prebuilt so hash-approved bundles
  are still preferred over the raw upstream CPU asset.
- install_python_stack.py: drop the /opt/rocm / hipcc gate in
  _ensure_rocm_torch() and rely on _has_rocm_gpu(). Runtime-only ROCm
  installs (package-managed minimal installs, Radeon software) that ship
  amd-smi / rocminfo without hipcc can now repair a CPU-only venv via
  "unsloth studio update". Adds an explicit IS_WINDOWS / IS_MACOS guard.
- studio/backend/utils/hardware/amd.py: honour HIP_VISIBLE_DEVICES /
  ROCR_VISIBLE_DEVICES / CUDA_VISIBLE_DEVICES in
  get_primary_gpu_utilization(). A process restricted to GPU 2 now
  reports metrics for GPU 2 instead of physical GPU 0. Tighten the plain
  bytes unit detection to an explicit allowlist.
- studio/backend/utils/hardware/hardware.py: route
  get_backend_visible_gpu_info()'s backend_cuda_visible_devices field
  through a helper that reads HIP_VISIBLE_DEVICES on ROCm. Drop the
  unconditional "(rocm=False)" suffix in apply_gpu_ids() logs.

* Fix round 2 regressions: ROCm validate_server and Windows HIP routing

Follow-up to 810b833b addressing review findings on the first round of
hardening commits:

- install_llama_prebuilt.py validate_server: gate --n-gpu-layers on the
  resolved install_kind instead of host.has_rocm. AMD Windows hosts
  without a HIP prebuilt fall back to windows-cpu and must not be
  validated with GPU layers; thread install_kind through from the
  caller.
- install_llama_prebuilt.py resolve_release_asset_choice: reinstate the
  "not has_rocm" guard on the published windows-cpu bundle so AMD
  Windows hosts reach resolve_upstream_asset_choice() where the new
  HIP prebuilt path lives. Prefer a published windows-hip bundle first
  when one exists, fall through to upstream HIP + upstream CPU
  otherwise.
- install_llama_prebuilt.py detect_host: also set has_physical_nvidia
  when the secondary --query-gpu block confirms a working NVIDIA GPU,
  so older nvidia-smi versions without -L support do not silently skip
  the Linux diagnostics that key off has_physical_nvidia.
- install_llama_prebuilt.py: drop redundant "import re as _re" /
  "import re as _re_rocm" local aliases in favour of the existing
  top-level "import re".
- install_python_stack.py _ensure_rocm_torch: run the AMD
  bitsandbytes install unconditionally after the HIP-torch probe so
  "unsloth studio update" on venvs that already have ROCm torch still
  gains the AMD bitsandbytes build.
- install.sh: add a non-x86_64 early-exit to get_torch_index_url() so
  aarch64 / arm64 Linux hosts do not hit the ROCm wheel index
  (PyTorch only publishes ROCm wheels for linux_x86_64).
- install.sh: add bitsandbytes install to the migrated-environment
  branch so upgrades pick it up for ROCm hosts instead of only the
  fresh-install path.
- install.sh: in the Radeon wheel path, pass version constraints +
  --no-index --find-links to uv instead of explicit wheel URLs so a
  version-compatible torch / torchvision / torchaudio triple is
  resolved, rather than picking the highest-version wheel for each
  package independently.
- studio/backend/utils/hardware/amd.py _first_visible_amd_gpu_id: fall
  through to lower-priority visibility env vars when the first entry
  is malformed (leading comma, all-whitespace first token) instead of
  silently returning GPU 0.

* Fix round 3 findings: x86_64 guard, ROCm version clip, Radeon deps

Address issues surfaced by the round 3 reviewers on top of 8636fa63:

- install_python_stack.py _ensure_rocm_torch: add the same `x86_64`
  guard that install.sh already has. Linux aarch64 / arm64 ROCm hosts
  must skip the repair path entirely; PyTorch only publishes ROCm
  wheels for linux_x86_64, and without this guard
  `unsloth studio update` aborts with a missing-wheel error on non
  x86_64 hosts.
- install_llama_prebuilt.py resolve_upstream_asset_choice: add a
  best-effort _detect_host_rocm_version() helper (reading
  /opt/rocm/.info/version, amd-smi version, hipconfig --version) and
  filter rocm_candidates to entries whose major.minor is <= host
  version. Falls back to the newest candidate only when no compatible
  one exists, so a ROCm 6.4 host downloads rocm-6.4 instead of being
  handed the numerically newest rocm-7.2 bundle (which fails preflight
  and forces a source build).
- install.sh: remove the round 2 --no-index switch from the Radeon
  wheel branch. --no-index forced uv to ignore PyPI entirely, which
  broke transitive dependency resolution (filelock, sympy, networkx,
  jinja2, fsspec, setuptools, typing-extensions, ...) on a fresh venv.
  Restore the round 1 explicit wheel URL invocation but add a
  torch / torchvision / torchaudio version-pair sanity check so a
  mismatched trio (e.g. torch 2.9.1 + torchvision 0.23.0 + torchaudio
  2.9.0) falls back to the standard ROCm index instead of installing a
  broken combination.
- install_python_stack.py _ensure_rocm_torch: restructure the
  "tag is None" path so it no longer short-circuits the bitsandbytes
  install. On a ROCm runtime older than anything in
  _ROCM_TORCH_INDEX, print the "no wheel" warning but still run the
  AMD bitsandbytes install.
- studio/backend/core/training/worker.py: restore the pre-PR
  "no timeout" behaviour for non-HIP causal-conv1d / mamba-ssm source
  builds. The round 2 "timeout = 1800 if is_hip else 300" cap aborts
  slow non-HIP builds (Linux aarch64, unsupported torch/CUDA combos)
  after 5 minutes; omit timeout for the non-HIP branch so the cap
  only applies to ROCm source builds.

* Fix round 4 findings: apply_gpu_ids env inheritance, Radeon X.Y, bitsandbytes gate

Address remaining issues surfaced by the round 4 reviewers:

- studio/backend/utils/hardware/hardware.py apply_gpu_ids: mirror the
  selection into HIP_VISIBLE_DEVICES / ROCR_VISIBLE_DEVICES whenever
  the caller already had a ROCm visibility env var set, not only when
  IS_ROCM has already been set by detect_hardware(). Training and
  inference workers call apply_gpu_ids() before detect_hardware()
  runs, so the old guard would leave a forked ROCm worker with a
  stale HIP_VISIBLE_DEVICES mask that no longer matched the
  narrowed CUDA_VISIBLE_DEVICES selection.
- install.sh get_radeon_wheel_url: accept X.Y ROCm versions in
  addition to X.Y.Z. The `/opt/rocm/.info/version` file and some
  hipconfig versions report only two components, and the Radeon
  repository publishes both rocm-rel-X.Y.Z/ and rocm-rel-X.Y/
  directories, so treating X.Y as invalid caused Radeon hosts to fall
  back to the generic ROCm index even when a matching AMD wheel set
  existed.
- install_python_stack.py _ensure_rocm_torch: only install the AMD
  bitsandbytes build when the venv actually has a ROCm-compatible
  torch (either already present or just installed by this function).
  Previously the bitsandbytes install ran unconditionally, which
  could leave an AMD bitsandbytes layered on top of a CPU/CUDA torch
  on hosts where the ROCm runtime is older than any entry in
  _ROCM_TORCH_INDEX. Also add --force-reinstall so an existing
  CPU/CUDA bitsandbytes is replaced by the AMD build during upgrades.

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

* Fix gemini findings: amd-smi metric envelope validation and dict-wrapped GPU id

Two medium-severity defensive fixes from the gemini-code-assist review on
the AMD monitoring backend:

1. _extract_gpu_metrics may return a dict where every value is None when
   amd-smi succeeds (zero exit) but the JSON envelope contains no usable
   fields (error response, unsupported card). The new _has_real_metrics
   helper lets get_primary_gpu_utilization surface available:False and
   lets get_visible_gpu_utilization skip ghost device rows so the UI
   does not render placeholder cards with empty numbers.

2. Newer amd-smi versions wrap scalar fields as {"value": 0, "unit":
   "none"}, including the per-GPU id. The previous int(raw_id) call
   silently fell back to the enumeration index in that case, losing the
   real GPU id. Routing raw_id through the existing _parse_numeric
   helper handles bare ints, floats, strings, and the dict shape
   uniformly, with a debug log on parse failure.

* Fix gemini round 2 findings: explicit length guard on ROCm version file parser

Both _detect_rocm_version (install_python_stack.py) and
_detect_host_rocm_version (install_llama_prebuilt.py) read /opt/rocm/.info/version
or $ROCM_PATH/lib/rocm_version, split on "." and unconditionally accessed
parts[1]. The surrounding broad `except Exception: pass` already swallowed
the resulting IndexError, so a one-component file like "6\n" did fall
through to the next detection source -- but the control flow relied on
exception handling instead of an explicit check.

Add `if len(parts) >= 2:` guards in both helpers so the loop falls through
on its own without raising. Behaviour is unchanged for the common multi-
component case; the previously-silent IndexError path becomes an explicit
no-op.

* Fix gemini round 3: include has_rocm in validate_server fallback path

When validate_server is called without an explicit install_kind (older
call sites that have not been updated), the fallback was only enabling
--n-gpu-layers for NVIDIA and macOS arm64 hosts. AMD ROCm Linux hosts
fell through to the CPU validation path even though the prebuilt being
exercised was a HIP binary.

Add host.has_rocm to the fallback expression so the GPU offload flag is
applied consistently with the install_kind=='linux-rocm' / 'windows-hip'
branches above.

* Fix gemini round 4: remove risky bytes-vs-MB heuristic in _parse_memory_mb

The previous heuristic divided any bare number above 10_000_000 by
1024*1024 on the assumption that large unit-less values were bytes.
This misclassified small VRAM allocations: 5 MB of used VRAM reported
as 5_242_880 bytes without a unit would be taken at face value and
render as 5_242_880 MB (~5 TB) in the monitoring UI.

Modern amd-smi always provides explicit units (MiB/GiB dict form),
and legacy amd-smi returns bare numbers in MB -- the heuristic never
had a real workload to handle. Drop it and default to MB for bare
numeric input, keeping the existing unit-aware branches for dict /
string inputs unchanged.

The unrelated gemini suggestion to "default minor to 0" in the
amd-smi version awk parser was intentionally NOT applied: rocm7.0
and rocm7.1 ship different wheel sets, so silently substituting 0
for a missing minor could install the wrong wheels. The existing
reject-and-fall-through behaviour is safer.

* Fix gemini round 5: POSIX compliance and leading-comma visibility parsing

Three medium findings from gemini-code-assist addressed in this commit:

1. _pick_radeon_wheel used grep -o and sort -V, both GNU extensions
   that are not in POSIX and break on BSD/BusyBox coreutils. install.sh
   has a #!/bin/sh shebang so the whole pipeline was rewritten as a
   single awk script that extracts all href="..." hits on each line,
   filters to wheels matching the package prefix and python tag, and
   picks the newest version via zero-padded lexical comparison. No
   external sort or grep is needed.

2. _first_visible_amd_gpu_id in the AMD monitoring backend treated a
   leading comma (e.g. HIP_VISIBLE_DEVICES=",1") as "fall through to
   the next env var", which is surprising given the clear intent to
   narrow to device 1. Filter empty tokens after the split and return
   the first real one. An all-commas value ("," / ",,,") still falls
   through because no real tokens exist; the empty-string and "-1"
   explicit-zero cases are unchanged.

The unrelated amd-smi version awk parser suggestion was not applied
(see round 4 commit message for rationale: defaulting a missing minor
to 0 could silently install the wrong ROCm wheel set).

* Fix 20-reviewer.py findings: base drift, Radeon %2B, dpkg/rpm fallback, bnb, backend label

Consolidated fix batch from a 20-parallel reviewer.py run on the current
head. Each fix is drawn from a high-consensus finding and addresses a
real bug or feature gap, not a stylistic preference.

1. install.sh: bump `unsloth>=2026.4.2` -> `unsloth>=2026.4.4` at five
   call sites so this branch no longer regresses main's version floor
   (main bumped to 2026.4.4 in #4876). Without this, merging 4720 would
   silently downgrade the minimum version pin for fresh installs.

2. install.sh: URL-decode Radeon wheel names before extracting the
   torch / torchvision / torchaudio version strings. Real wheel URLs
   from repo.radeon.com are percent-encoded ("torch-2.10.0%2Brocm7.2.0...")
   so the previous `[+-]` terminator in the sed regex never matched,
   `_torch_ver` stayed empty, `_radeon_versions_match` stayed false,
   and every Radeon consumer install silently fell back to the generic
   ROCm index. Now decode %2B -> + first, then extract, then validate.

3. install.sh: the two AMD bitsandbytes install lines were running
   `uv pip install "bitsandbytes>=0.49.1"` without `--force-reinstall`,
   so upgrades where the venv already has a CPU/CUDA bitsandbytes
   satisfying the constraint would keep the stale non-AMD wheel. Add
   `--force-reinstall --no-cache-dir` to both call sites, matching the
   pattern already used in install_python_stack.py::_ensure_rocm_torch.

4. install_python_stack.py and install_llama_prebuilt.py: add
   `dpkg-query -W rocm-core` and `rpm -q rocm-core` fallbacks to the
   Python-side ROCm version detectors so they match the chain in
   install.sh::get_torch_index_url. Package-managed ROCm installs
   (Debian/Ubuntu/RHEL/Fedora distro packages) can expose GPUs via
   rocminfo/amd-smi but still lack /opt/rocm/.info/version, hipconfig,
   or amd-smi `version` output -- without these fallbacks, `unsloth
   studio update` on such hosts returned None and skipped the ROCm
   torch repair. Also strip the dpkg epoch prefix ("1:6.3.0-1") before
   parsing so epoch-annotated packages parse correctly.

5. hardware.py: add a `_backend_label(device)` helper that returns
   "rocm" when IS_ROCM is set and the device is DeviceType.CUDA, and
   use it for every `"backend": ...` emission in JSON responses served
   to the Studio frontend. Internally we still represent ROCm hosts as
   DeviceType.CUDA (ROCm torch reuses the whole torch.cuda.* API
   surface), but the user-facing API now correctly reports "rocm" on
   AMD boxes instead of labeling them as "cuda".

All 250 simulation scenarios pass (was 233 before this batch: added 17
new regression tests covering the version pin, %2B decoding, bnb
force-reinstall flags, dpkg/rpm fallback presence, and the
_backend_label helper's four-way truth table).

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

* Fix gemini round 6 + URL audit: amd.py defensive checks, rocm6.5+ clip to 6.4

Two rounds of fixes in one commit, plus a full URL audit of every PyPI /
download.pytorch.org / repo.radeon.com reference the PR introduces.

amd.py (4 medium gemini findings on commit b3627bc2):

1. _extract_gpu_metrics used `and vram_total_mb` as part of the vram_util
   gate. The follow-up `vram_total_mb > 0` already handles the division
   guard, but the truthiness check was redundant and slightly surprising
   for a 0.0 valid value. Replace with explicit `is not None and > 0`
   for both vram_util and power_util.

2. get_physical_gpu_count called `data.get("gpu", ...)` without guarding
   for non-dict envelopes. A scalar / string JSON response from amd-smi
   would raise AttributeError. Add an isinstance(data, dict) check and
   return None for unexpected shapes.

3. get_visible_gpu_utilization had the same .get() exposure on the outer
   envelope. Rewrite the gpu_list extraction as an explicit
   list/dict/else cascade so a malformed scalar envelope produces
   gpu_list=[data] and continues without raising.

4. The same function's per-entry loop also called gpu_data.get() on
   whatever was inside gpu_list. If a scalar ever leaks into the list
   (directly or via the previous fix's fallback), _extract_gpu_metrics
   would raise on the first .get() inside the helper. Skip non-dict
   entries in the loop before extracting metrics.

install.sh (URL audit finding, previously flagged by 20-reviewer as #13):

5. get_torch_index_url used `rocm6.*` in the rocm tag case statement,
   which matched rocm6.5 and rocm6.6 and emitted
   download.pytorch.org/whl/rocm6.5 -- which returns HTTP 403 because
   PyTorch only publishes rocm 5.7, 6.0-6.4, 7.0-7.2. Enumerate the
   supported 6.x minors explicitly and add a rocm6.* fallback branch
   that clips to rocm6.4 (the last supported 6.x wheel set).

URL audit results (all URLs PR 4720 references):
- 14/14 download.pytorch.org/whl/{cpu,cu118,cu124,cu126,cu128,cu130,
  rocm6.0..6.4,rocm7.0..7.2} return HTTP 200.
- 9/9 repo.radeon.com/rocm/manylinux/rocm-rel-{5.7,6.0,6.1,6.2,6.3,
  6.4,7.0,7.1,7.2}/ return HTTP 200.
- X.Y.Z patch directories exist for 7.0.2, 7.1.1, 7.2.1 but NOT for
  6.3.0, 6.4.0, 6.2.1 -- install.sh already handles this via the X.Y.Z
  -> X.Y fallback sed in the Radeon wheel install block.
- Docs links (rocm.docs.amd.com, docs.unsloth.ai AMD guide) and the
  llama.cpp GitHub releases API endpoint all return 200.

Test suite: 255 -> 258. New regression coverage:
- U17: get_physical_gpu_count tolerates scalar amd-smi envelope
- U18: get_visible_gpu_utilization tolerates scalar envelope
- U19a-c: vram_util / power_util return None on zero total, but
  vram_total_gb still echoes 0.0 (not None)
- A_rocm{6.5,6.6,6.9}_clips_to_rocm64: install.sh clips unsupported
  6.x minors to rocm6.4 instead of producing a 403 index URL

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

* Fix reviewer.py round 2: tokenizer AMD multi-GPU, --no-torch bnb, main.py backend label

Three high-confidence findings from a second 20-parallel reviewer.py run
on commit 7effb3ae. Triaged 15 total findings and applied the three that
were confirmed as real bugs; the rest were either false positives (e.g.
"migrated AMD venv not repaired" -- _ensure_rocm_torch runs downstream
via setup.sh regardless), design decisions (e.g. visibility mask env
vars not consulted in installer detection), or edge cases the existing
fallback logic already handles.

1. unsloth/tokenizer_utils.py [6/20]: the multi-GPU guard's shell probe
   runs `nvidia-smi --query-gpu=memory.used`, catches the failure, then
   only raises if `torch.cuda.is_available()` is False. On ROCm torch,
   torch.cuda.is_available() returns True (ROCm reuses the torch.cuda.*
   API), so the guard becomes dead code on AMD hosts and multi-GPU AMD
   setups slip through even though unsloth does not support them yet.
   Add a torch.cuda.device_count() > 1 fallback inside the except so
   AMD multi-visible-device setups are flagged consistently with the
   original CUDA memory check.

2. install.sh [1/20]: the fresh-install bitsandbytes block for AMD ROCm
   ran unconditionally when TORCH_INDEX_URL matched `*/rocm*`, even when
   SKIP_TORCH=true (from --no-torch or Intel Mac auto-detect). A user
   running `install.sh --no-torch` on an AMD host would still pull in
   bitsandbytes despite explicitly asking for GGUF-only mode. Wrap the
   case block in an outer `[ "$SKIP_TORCH" = false ]` guard.

3. studio/backend/main.py [3/20]: the /api/system endpoint returned
   `"device_backend": get_device().value`, which is "cuda" on ROCm
   hosts (because ROCm torch piggybacks on torch.cuda). Other endpoints
   (hardware.py) already use the _backend_label helper which swaps
   "cuda" -> "rocm" when IS_ROCM. Route /api/system through the same
   helper so the Studio UI reports the backend consistently across all
   endpoints.

4. studio/backend/tests/test_utils.py: update test_backend_matches_device
   to call _backend_label(get_device()) instead of raw get_device().value
   so the test matches the new contract and still passes on CUDA hosts.

Tests: 258 -> 261. New regression coverage:
- X08 main.py /api/system uses _backend_label
- X09 tokenizer multi-GPU guard has device_count() fallback
- X10 fresh-install bnb case block gated on SKIP_TORCH=false

* fix: prevent bitsandbytes from overwriting ROCm torch with CUDA wheels

During install, bitsandbytes was installed without --no-deps, causing
uv to resolve torch from PyPI (CUDA build) and silently overwrite the
ROCm wheels that were just installed in the previous step.

This happened in three places:
- install.sh: bitsandbytes install in both migrated and fresh paths
- install_python_stack.py: bitsandbytes install inside _ensure_rocm_torch()

Additionally, multiple install steps in install_python_stack.py (extras,
overrides, studio deps) can pull in CUDA torch via transitive
dependencies. A final _ensure_rocm_torch() call at the end of the
install sequence ensures ROCm torch is always in place at runtime.

All changes are gated behind ROCm-specific conditions and do not affect
NVIDIA, CPU-only, macOS, or Windows install paths.

Tested on AMD Instinct MI300X VF with ROCm 7.2.0 -- confirms
torch==2.10.0+rocm7.1 with HIP 7.1.25424 after install.

* fix: ROCm inference fallback -- skip Unsloth patching and bnb 4-bit on HIP

On AMD ROCm (HIP), two issues prevent the normal Unsloth inference path:

1. Unsloth's global monkey-patching of transformers model classes
   (LlamaRotaryEmbedding, attention modules) triggers
   _assert_async_cuda_kernel crashes on HIP during generation.
   Training uses different code paths and works fine.

2. bitsandbytes 4-bit matmul kernels also trigger HIP assertion
   failures on MI300X (CDNA3 / gfx942), even without Unsloth patching.

This commit adds a ROCm-specific inference fallback that:
- Skips importing Unsloth at module level (prevents global patching)
- Loads models in 16-bit with plain transformers + PEFT instead
- Resolves pre-quantized model names (e.g. "xxx-bnb-4bit" -> "xxx")
  since pre-quantized HF repos still trigger bnb codepaths
- Guards get_chat_template calls (unavailable without Unsloth import)
- Fixes max_seq_length=0 being passed to from_pretrained (GGUF
  semantics don't apply to transformers path)

The NVIDIA path is completely unchanged -- Unsloth import and
for_inference() optimization remain active. GGUF inference (via
llama-server/HIP) is unaffected since it never imports Python model
classes. AMD GPUs typically have large VRAM (e.g. 192GB on MI300X)
so 16-bit loading is practical for inference.

Tested on AMD Instinct MI300X VF (ROCm 7.2, HIP 7.1.25424):
- Simple generation: PASS
- Compare mode (base vs finetuned): PASS
- GGUF inference + tool calling: PASS (unaffected by this change)

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

* fix: guard audio/vision inference on ROCm, remove unused import

- Add clear RuntimeError for audio/vision model inference on ROCm
  (these paths use Unsloth's FastModel/FastVisionModel which would
  crash on HIP; GGUF inference is the supported path on AMD)
- Remove unused `import os as _os` from the ROCm changes

* fix: amd-smi parsing for newer output format (gpu_data wrapper, mem_usage, temperature)

amd-smi on recent ROCm versions (7.x) wraps metric output in a
{"gpu_data": [...]} envelope instead of returning a raw list. This
caused get_primary_gpu_utilization() and get_visible_gpu_utilization()
to fail silently (returning available=False) because the GPU data
dict was never unwrapped.

Additionally:
- VRAM data moved from "vram" to "mem_usage" with "total_vram" /
  "used_vram" keys. Added fallback key lookup.
- Temperature "edge" sensor returns "N/A" on MI300X VF; the previous
  dict.get() chain returned the "N/A" string instead of falling
  through to "hotspot". Changed to a loop that checks each key until
  a parseable value is found.

Tested on AMD Instinct MI300X VF (ROCm 7.2, amd-smi 24.x):
- GPU utilization: 0% (idle), up to 100% during training
- Temperature: 40-44C (from hotspot sensor)
- VRAM: 0.28/191.69 GB (idle)
- Power: 158-211W draw

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

* Bug fix detecting radeon (#4940)

* Bug fix detecting radeon

* Expanding GPU target for gfx1100*

* Generalize gfx family-prefix filter to cover gfx10/gfx12 as well

rocminfo on ROCm 6.1+ emits LLVM generic-family ISA lines alongside the
specific GPU (e.g. gfx11-generic next to gfx1100). The outer grep captures
the bare family prefix from the generic line, and passing that to
-DGPU_TARGETS breaks the HIP build because clang only accepts specific
gfxNNN ids.

The previous filter only special-cased gfx11. Generalize it so any bare
2-digit family prefix (gfx10, gfx11, gfx12, ...) is dropped whenever a
specific sibling target is present in the same list. No real AMD GPU has
a 2-digit gfx id, so the filter can only ever drop family prefixes and
never a real target.

Covers the existing gfx11 cases unchanged, and extends the same fix to
gfx10-1-generic / gfx10-3-generic (RDNA1/2) and gfx12-generic (RDNA4),
which would otherwise hit the same build failure on newer rocminfo.

---------

Co-authored-by: Iswarya Alex <iswarya.alex@amd.com>
Co-authored-by: Daniel Han <danielhanchen@users.noreply.github.com>

---------

Co-authored-by: Eda Z <eda.zhou@amd.com>
Co-authored-by: GoldenGrapeGentleman <yueyuan@amd.com>
Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com>
Co-authored-by: billishyahao <bill.he@amd.com>
Co-authored-by: Iswarya Alex <47045679+iswaryaalex@users.noreply.github.com>
Co-authored-by: Iswarya Alex <iswarya.alex@amd.com>
Co-authored-by: Daniel Han <danielhanchen@users.noreply.github.com>
---
 install.sh                                 | 409 ++++++++++++++++++++-
 studio/backend/core/inference/inference.py | 134 ++++++-
 studio/backend/core/training/worker.py     | 124 ++++++-
 studio/backend/main.py                     |   6 +-
 studio/backend/tests/test_utils.py         |   8 +-
 studio/backend/utils/hardware/__init__.py  |  10 +
 studio/backend/utils/hardware/amd.py       | 382 +++++++++++++++++++
 studio/backend/utils/hardware/hardware.py  | 234 +++++++++---
 studio/install_llama_prebuilt.py           | 306 ++++++++++++++-
 studio/install_python_stack.py             | 337 ++++++++++++++++-
 studio/setup.sh                            |  14 +
 unsloth/kernels/utils.py                   |  18 +-
 unsloth/tokenizer_utils.py                 |  11 +-
 13 files changed, 1876 insertions(+), 117 deletions(-)
 create mode 100644 studio/backend/utils/hardware/amd.py

diff --git a/install.sh b/install.sh
index ea53ecc6d6..2e347a8bc5 100755
--- a/install.sh
+++ b/install.sh
@@ -978,6 +978,37 @@ _find_no_torch_runtime() {
     fi
 }
 
+# ── AMD ROCm GPU detection helper ──
+# Returns 0 (true) if an actual AMD GPU is present, 1 (false) otherwise.
+# Checks rocminfo for gfx[1-9]* (excludes gfx000 CPU agent) and
+# amd-smi list for GPU data rows (excludes header-only output).
+_has_amd_rocm_gpu() {
+    if command -v rocminfo >/dev/null 2>&1 && \
+       rocminfo 2>/dev/null | awk '/Name:[[:space:]]*gfx[0-9]/ && !/Name:[[:space:]]*gfx000/{found=1} END{exit !found}'; then
+        return 0
+    elif command -v amd-smi >/dev/null 2>&1 && \
+         amd-smi list 2>/dev/null | awk '/^GPU[[:space:]]*[:\[][[:space:]]*[0-9]/{ found=1 } END{ exit !found }'; then
+        return 0
+    fi
+    return 1
+}
+
+# ── NVIDIA usable-GPU helper ──
+# Returns 0 (true) only if nvidia-smi is present AND actually lists a GPU.
+# Prevents AMD-only hosts with a stale nvidia-smi on PATH from being routed
+# into the CUDA branch.
+_has_usable_nvidia_gpu() {
+    _nvsmi=""
+    if command -v nvidia-smi >/dev/null 2>&1; then
+        _nvsmi="nvidia-smi"
+    elif [ -x "/usr/bin/nvidia-smi" ]; then
+        _nvsmi="/usr/bin/nvidia-smi"
+    else
+        return 1
+    fi
+    "$_nvsmi" -L 2>/dev/null | awk '/^GPU[[:space:]]+[0-9]+:/{found=1} END{exit !found}'
+}
+
 # ── Detect GPU and choose PyTorch index URL ──
 # Mirrors Get-TorchIndexUrl in install.ps1.
 # On CPU-only machines this returns the cpu index, avoiding the solver
@@ -986,14 +1017,82 @@ get_torch_index_url() {
     _base="https://download.pytorch.org/whl"
     # macOS: always CPU (no CUDA support)
     case "$(uname -s)" in Darwin) echo "$_base/cpu"; return ;; esac
-    # Try nvidia-smi
+    # Try nvidia-smi -- require the binary to actually list a usable GPU.
+    # Presence of the binary alone (container leftovers, stale driver
+    # packages) is not sufficient: otherwise an AMD-only host would
+    # silently install CUDA wheels.
     _smi=""
-    if command -v nvidia-smi >/dev/null 2>&1; then
-        _smi="nvidia-smi"
-    elif [ -x "/usr/bin/nvidia-smi" ]; then
-        _smi="/usr/bin/nvidia-smi"
+    if _has_usable_nvidia_gpu; then
+        if command -v nvidia-smi >/dev/null 2>&1; then
+            _smi="nvidia-smi"
+        elif [ -x "/usr/bin/nvidia-smi" ]; then
+            _smi="/usr/bin/nvidia-smi"
+        fi
+    fi
+    if [ -z "$_smi" ]; then
+        # No NVIDIA GPU -- check for AMD ROCm GPU.
+        # PyTorch only publishes ROCm wheels for linux-x86_64; skip the
+        # ROCm branch entirely on aarch64 / arm64 / other architectures
+        # so non-x86_64 Linux hosts fall back cleanly to CPU wheels.
+        case "$(uname -m)" in
+            x86_64|amd64) : ;;
+            *) echo "$_base/cpu"; return ;;
+        esac
+        if ! _has_amd_rocm_gpu; then
+            echo "$_base/cpu"; return
+        fi
+        # AMD GPU confirmed -- detect ROCm version
+        _rocm_tag=""
+        _rocm_tag=$({ command -v amd-smi >/dev/null 2>&1 && \
+            amd-smi version 2>/dev/null | awk -F'ROCm version: ' \
+                'NF>1{gsub(/[^0-9.]/, "", $2); split($2,a,"."); print "rocm"a[1]"."a[2]; ok=1; exit} END{exit !ok}'; } || \
+            { [ -r /opt/rocm/.info/version ] && \
+                awk -F. '{print "rocm"$1"."$2; exit}' /opt/rocm/.info/version; } || \
+            { command -v hipconfig >/dev/null 2>&1 && \
+                hipconfig --version 2>/dev/null | awk 'NR==1 && /^[0-9]/{split($1,a,"."); if(a[1]+0>0){print "rocm"a[1]"."a[2]; found=1}} END{exit !found}'; } || \
+            { command -v dpkg-query >/dev/null 2>&1 && \
+                ver="$(dpkg-query -W -f='${Version}\n' rocm-core 2>/dev/null)" && \
+                [ -n "$ver" ] && \
+                printf '%s\n' "$ver" | sed 's/^[0-9]*://' | awk -F'[.-]' '{print "rocm"$1"."$2; exit}'; } || \
+            { command -v rpm >/dev/null 2>&1 && \
+                ver="$(rpm -q --qf '%{VERSION}\n' rocm-core 2>/dev/null)" && \
+                [ -n "$ver" ] && \
+                printf '%s\n' "$ver" | awk -F'[.-]' '{print "rocm"$1"."$2; exit}'; }) 2>/dev/null
+        # Validate _rocm_tag: must match "rocmX.Y" with major >= 1
+        case "$_rocm_tag" in
+            rocm[1-9]*.[0-9]*) : ;;  # valid (major >= 1)
+            *) _rocm_tag="" ;;        # reject malformed (empty, garbled, or major=0)
+        esac
+        if [ -n "$_rocm_tag" ]; then
+            # Minimum supported: ROCm 6.0 (no PyTorch wheels exist for older)
+            case "$_rocm_tag" in
+                rocm[1-5].*) echo "$_base/cpu"; return ;;
+            esac
+            # ROCm 7.2 only has torch 2.11.0 which exceeds current bounds
+            # (<2.11.0).  Fall back to rocm7.1 index which has torch 2.10.0.
+            # Enumerate explicit versions rather than matching rocm6.* so
+            # a host on ROCm 6.5 or 6.6 (no PyTorch wheels published) is
+            # clipped down to the last supported 6.x (rocm6.4) instead of
+            # constructing https://download.pytorch.org/whl/rocm6.5 which
+            # returns HTTP 403. PyTorch only ships: rocm5.7, 6.0, 6.1, 6.2,
+            # 6.3, 6.4, 7.0, 7.1, 7.2 (and 5.7 is below our minimum).
+            # TODO: uncomment rocm7.2 when the torch upper bound is bumped
+            # to >=2.11.0.
+            case "$_rocm_tag" in
+                rocm6.0|rocm6.0.*|rocm6.1|rocm6.1.*|rocm6.2|rocm6.2.*|rocm6.3|rocm6.3.*|rocm6.4|rocm6.4.*|rocm7.0|rocm7.0.*|rocm7.1|rocm7.1.*)
+                    echo "$_base/$_rocm_tag" ;;
+                rocm6.*)
+                    # ROCm 6.5+ (no published PyTorch wheels): clip down
+                    # to the last supported 6.x wheel set.
+                    echo "$_base/rocm6.4" ;;
+                *)
+                    # ROCm 7.2+ (including future 10.x+): cap to rocm7.1
+                    echo "$_base/rocm7.1" ;;
+            esac
+            return
+        fi
+        echo "$_base/cpu"; return
     fi
-    if [ -z "$_smi" ]; then echo "$_base/cpu"; return; fi
     # Parse CUDA version from nvidia-smi output (POSIX-safe, no grep -P)
     _cuda_ver=$(LC_ALL=C $_smi 2>/dev/null \
         | sed -n 's/.*CUDA Version:[[:space:]]*\([0-9][0-9]*\.[0-9][0-9]*\).*/\1/p' \
@@ -1011,20 +1110,157 @@ get_torch_index_url() {
     elif [ "$_major" -ge 11 ]; then echo "$_base/cu118"
     else echo "$_base/cpu"; fi
 }
+
+get_radeon_wheel_url() {
+    # Only meaningful on Linux. Picks a repo.radeon.com base URL whose listing
+    # contains torch wheels. Tries paths like rocm-rel-7.2.1/, rocm-rel-7.2/,
+    # rocm-rel-7.1.1/, rocm-rel-7.1/ (AMD publishes both M.m and M.m.p dirs).
+    # Accepts both X.Y and X.Y.Z host versions since /opt/rocm/.info/version
+    # and hipconfig --version can return either shape.
+    case "$(uname -s)" in Linux) ;; *) echo ""; return ;; esac
+
+    # Detect ROCm version (X.Y or X.Y.Z) -- try amd-smi, then
+    # /opt/rocm/.info/version, then hipconfig.
+    _full_ver=""
+    _full_ver=$({ command -v amd-smi >/dev/null 2>&1 && \
+        amd-smi version 2>/dev/null | awk -F'ROCm version: ' \
+            'NF>1{if(match($2,/[0-9]+\.[0-9]+(\.[0-9]+)?/)){print substr($2,RSTART,RLENGTH); ok=1; exit}} END{exit !ok}'; } || \
+        { [ -r /opt/rocm/.info/version ] && \
+            awk 'match($0,/[0-9]+\.[0-9]+(\.[0-9]+)?/){print substr($0,RSTART,RLENGTH); found=1; exit} END{exit !found}' /opt/rocm/.info/version; } || \
+        { command -v hipconfig >/dev/null 2>&1 && \
+            hipconfig --version 2>/dev/null | awk 'NR==1 && match($0,/[0-9]+\.[0-9]+(\.[0-9]+)?/){print substr($0,RSTART,RLENGTH); found=1} END{exit !found}'; }) 2>/dev/null
+
+    # Validate: must be X.Y or X.Y.Z with X >= 1
+    case "$_full_ver" in
+        [1-9]*.[0-9]*.[0-9]*) : ;;  # X.Y.Z
+        [1-9]*.[0-9]*) : ;;          # X.Y
+        *) echo ""; return ;;
+    esac
+    echo "https://repo.radeon.com/rocm/manylinux/rocm-rel-${_full_ver}/"
+}
+
+# ── Radeon repo wheel selection helpers ──────────────────────────────────────
+# Fetches the Radeon repo directory listing once into _RADEON_LISTING (global).
+# _RADEON_PYTAG holds the CPython tag for the running interpreter (e.g. cp312).
+# _RADEON_BASE_URL holds the base URL for relative-href resolution.
+_RADEON_LISTING=""
+_RADEON_PYTAG=""
+_RADEON_BASE_URL=""
+
+_radeon_fetch_listing() {
+    # Usage: _radeon_fetch_listing BASE_URL
+    # Populates _RADEON_LISTING, _RADEON_PYTAG, _RADEON_BASE_URL.
+    _RADEON_BASE_URL="$1"
+    _RADEON_PYTAG=$("$_VENV_PY" -c "
+import sys
+print('cp{}{}'.format(sys.version_info.major, sys.version_info.minor))
+" 2>/dev/null) || return 1
+    if command -v curl >/dev/null 2>&1; then
+        _RADEON_LISTING=$(curl -fsSL --max-time 20 "$_RADEON_BASE_URL" 2>/dev/null)
+    elif command -v wget >/dev/null 2>&1; then
+        _RADEON_LISTING=$(wget -qO- --timeout=20 "$_RADEON_BASE_URL" 2>/dev/null)
+    fi
+    [ -n "$_RADEON_LISTING" ] || return 1
+}
+
+_pick_radeon_wheel() {
+    # Usage: _pick_radeon_wheel PACKAGE_NAME
+    # Scans $_RADEON_LISTING for the newest wheel whose filename starts exactly
+    # with PACKAGE_NAME- and matches _RADEON_PYTAG + linux_x86_64.
+    # Prints the full URL (resolving relative hrefs against _RADEON_BASE_URL).
+    #
+    # POSIX-compliant pipeline: all href parsing, filtering, and version
+    # selection is done inside a single awk script rather than reaching
+    # for GNU extensions (grep -o, sort -V) that would break under BSD
+    # or BusyBox coreutils.
+    _pkg="$1"
+    [ -n "$_RADEON_LISTING" ] || return 1
+    [ -n "$_RADEON_PYTAG"   ] || return 1
+    _tag="$_RADEON_PYTAG"
+    _href=$(printf '%s\n' "$_RADEON_LISTING" \
+        | awk -v pkg="$_pkg" -v tag="$_tag" '
+            BEGIN { max_pad = ""; max_url = "" }
+            {
+                line = $0
+                while (match(line, /href="[^"]*"/)) {
+                    # Strip the leading href=" (6 chars) and trailing " (1 char)
+                    url = substr(line, RSTART + 6, RLENGTH - 7)
+                    line = substr(line, RSTART + RLENGTH)
+
+                    # Extract basename, strip query / fragment
+                    n = split(url, p, "/")
+                    base = p[n]
+                    sub(/[?#].*/, "", base)
+
+                    prefix = pkg "-"
+                    # Match cpXY-cpXY or cpXY-abi3 with any linux x86_64
+                    # platform tag (linux_x86_64, manylinux_2_28_x86_64,
+                    # manylinux2014_x86_64, etc.)
+                    if (substr(base, 1, length(prefix)) == prefix &&
+                            index(base, "-" tag "-") > 0 &&
+                            match(base, /x86_64\.whl$/)) {
+                        # Extract the version component (first
+                        # dotted-number run) and pad each piece so a
+                        # plain lexical comparison gives us the newest.
+                        if (match(base, /[0-9]+\.[0-9]+(\.[0-9]+)?/)) {
+                            ver = substr(base, RSTART, RLENGTH)
+                            m = split(ver, v, ".")
+                            pad = ""
+                            for (i = 1; i <= m; i++)
+                                pad = pad sprintf("%08d", v[i])
+                            if (pad > max_pad) {
+                                max_pad = pad
+                                max_url = url
+                            }
+                        }
+                    }
+                }
+            }
+            END { if (max_url != "") print max_url }')
+    [ -z "$_href" ] && return 1
+    case "$_href" in
+        http*) printf '%s\n' "$_href" ;;
+        *)     printf '%s\n' "${_RADEON_BASE_URL%/}/${_href#/}" ;;
+    esac
+}
+
 TORCH_INDEX_URL=$(get_torch_index_url)
 
+# Auto-detect GPU for AMD ROCm based
+# get_torch_index_url must have chosen */rocm*
+# (gfx in rocminfo or amd-smi list). Then require rocminfo "Marketing Name:.*Radeon".
+_amd_gpu_radeon=false
+case "$TORCH_INDEX_URL" in
+    */rocm*)
+        if _has_amd_rocm_gpu && command -v rocminfo >/dev/null 2>&1 && \
+           rocminfo 2>/dev/null | grep -q 'Marketing Name:.*Radeon'; then
+            _amd_gpu_radeon=true
+        fi
+        ;;
+esac
+
 # ── Print CPU-only hint when no GPU detected ──
 case "$TORCH_INDEX_URL" in
     */cpu)
         if [ "$SKIP_TORCH" = false ] && [ "$OS" != "macos" ]; then
             echo ""
-            echo "  NOTE: No NVIDIA GPU detected (nvidia-smi not found)."
+            echo "  NOTE: No GPU detected (nvidia-smi and ROCm not found)."
             echo "  Installing CPU-only PyTorch. If you only need GGUF chat/inference,"
             echo "  re-run with --no-torch for a faster, lighter install:"
             echo "    curl -fsSL https://unsloth.ai/install.sh | sh -s -- --no-torch"
+            echo "  AMD ROCm users: see https://docs.unsloth.ai/get-started/install-and-update/amd"
             echo ""
         fi
         ;;
+    */rocm*)
+        echo ""
+        if [ "$_amd_gpu_radeon" = true ]; then
+            echo "  AMD Radeon + ROCm detected -- installing PyTorch wheels from repo.radeon.com"
+        else
+            echo "  AMD ROCm detected -- installing ROCm-enabled PyTorch ($TORCH_INDEX_URL)"
+        fi
+        echo ""
+        ;;
 esac
 
 # ── Install unsloth directly into the venv (no activation needed) ──
@@ -1054,15 +1290,158 @@ if [ "$_MIGRATED" = true ]; then
         substep "overlaying local repo (editable)..."
         run_install_cmd "overlay local repo" uv pip install --python "$_VENV_PY" -e "$_REPO_ROOT" --no-deps
     fi
+    # AMD ROCm: install bitsandbytes even in migrated environments so
+    # existing ROCm installs gain the AMD bitsandbytes build without a
+    # fresh reinstall.
+    if [ "$SKIP_TORCH" = false ]; then
+        case "$TORCH_INDEX_URL" in
+            */rocm*)
+                substep "installing bitsandbytes for AMD ROCm..."
+                run_install_cmd "install bitsandbytes (AMD)" uv pip install --python "$_VENV_PY" --force-reinstall --no-cache-dir --no-deps "bitsandbytes>=0.49.1"
+                # Repair ROCm torch if overwritten during migrated install
+                _has_hip=$("$_VENV_PY" -c "import torch; print(getattr(torch.version,'hip','') or '')" 2>/dev/null || true)
+                if [ -z "$_has_hip" ]; then
+                    substep "repairing ROCm torch (overwritten by dependency resolution)..."
+                    run_install_cmd "repair ROCm torch" uv pip install --python "$_VENV_PY" \
+                        "$TORCH_CONSTRAINT" torchvision torchaudio \
+                        --index-url "$TORCH_INDEX_URL" \
+                        --force-reinstall
+                fi
+                ;;
+        esac
+    fi
 elif [ -n "$TORCH_INDEX_URL" ]; then
     # Fresh: Step 1 - install torch from explicit index (skip when --no-torch or Intel Mac)
     if [ "$SKIP_TORCH" = true ]; then
         substep "skipping PyTorch (--no-torch or Intel Mac x86_64)." "$C_WARN"
+    elif [ "$_amd_gpu_radeon" = true ]; then
+        _radeon_url=$(get_radeon_wheel_url)
+        if [ -n "$_radeon_url" ]; then
+            _radeon_listing_ok=false
+            if _radeon_fetch_listing "$_radeon_url" 2>/dev/null; then
+                _radeon_listing_ok=true
+            else
+                # Try shorter X.Y path (AMD publishes both X.Y.Z and X.Y dirs)
+                _radeon_url_short=$(printf '%s\n' "$_radeon_url" \
+                    | sed 's|rocm-rel-\([0-9]*\)\.\([0-9]*\)\.[0-9]*/|rocm-rel-\1.\2/|')
+                if [ "$_radeon_url_short" != "$_radeon_url" ] && \
+                   _radeon_fetch_listing "$_radeon_url_short" 2>/dev/null; then
+                    _radeon_listing_ok=true
+                fi
+            fi
+
+            if [ "$_radeon_listing_ok" = true ]; then
+                # Require torch, torchvision, torchaudio wheels to all resolve
+                # from the Radeon listing. If any is missing for this Python
+                # tag, fall through to the standard ROCm index instead of
+                # silently mixing Radeon wheels with PyPI defaults.
+                _torch_whl=$(_pick_radeon_wheel "torch"       2>/dev/null) || _torch_whl=""
+                _tv_whl=$(_pick_radeon_wheel    "torchvision" 2>/dev/null) || _tv_whl=""
+                _ta_whl=$(_pick_radeon_wheel    "torchaudio"  2>/dev/null) || _ta_whl=""
+                _tri_whl=$(_pick_radeon_wheel   "triton"      2>/dev/null) || _tri_whl=""
+                # Sanity-check torch / torchvision / torchaudio are a
+                # matching release. The Radeon repo publishes multiple
+                # generations simultaneously, so picking the highest-version
+                # wheel for each package independently can assemble a
+                # mismatched trio (e.g. torch 2.9.1 + torchvision 0.23.0 +
+                # torchaudio 2.9.0 from the current rocm-rel-7.2.1 index).
+                # Check that torch and torchaudio share the same X.Y public
+                # version prefix, and that torchvision's minor correctly
+                # pairs with torch's minor (torchvision = torch.minor - 5
+                # since torch 2.4 -> torchvision 0.19 -> torch 2.9 ->
+                # torchvision 0.24).
+                # URL-decode each wheel name so %2B -> + before version
+                # extraction. Real Radeon wheel hrefs are percent-encoded
+                # (torch-2.10.0%2Brocm7.2.0...), so a plain [+-] terminator
+                # in the sed regex below would never match and
+                # _radeon_versions_match would stay false for every real
+                # listing, silently forcing a fallback to the generic
+                # ROCm index.
+                _torch_ver=""
+                _tv_ver=""
+                _ta_ver=""
+                if [ -n "$_torch_whl" ]; then
+                    _torch_name=$(printf '%s' "${_torch_whl##*/}" | sed 's/%2[Bb]/+/g')
+                    _torch_ver=$(printf '%s\n' "$_torch_name" | sed -n 's|^torch-\([0-9][0-9]*\.[0-9][0-9]*\)\(\.[0-9][0-9]*\)\{0,1\}[+-].*|\1|p')
+                fi
+                if [ -n "$_tv_whl" ]; then
+                    _tv_name=$(printf '%s' "${_tv_whl##*/}" | sed 's/%2[Bb]/+/g')
+                    _tv_ver=$(printf '%s\n' "$_tv_name" | sed -n 's|^torchvision-\([0-9][0-9]*\.[0-9][0-9]*\)\(\.[0-9][0-9]*\)\{0,1\}[+-].*|\1|p')
+                fi
+                if [ -n "$_ta_whl" ]; then
+                    _ta_name=$(printf '%s' "${_ta_whl##*/}" | sed 's/%2[Bb]/+/g')
+                    _ta_ver=$(printf '%s\n' "$_ta_name" | sed -n 's|^torchaudio-\([0-9][0-9]*\.[0-9][0-9]*\)\(\.[0-9][0-9]*\)\{0,1\}[+-].*|\1|p')
+                fi
+                _radeon_versions_match=false
+                if [ -n "$_torch_ver" ] && [ -n "$_tv_ver" ] && [ -n "$_ta_ver" ]; then
+                    _torch_major=${_torch_ver%%.*}
+                    _torch_minor=${_torch_ver#*.}
+                    _ta_major=${_ta_ver%%.*}
+                    _ta_minor=${_ta_ver#*.}
+                    _tv_major=${_tv_ver%%.*}
+                    _tv_minor=${_tv_ver#*.}
+                    # torchvision expected minor (e.g. torch 2.9 -> 0.24)
+                    _expected_tv_minor=$((_torch_minor + 15))
+                    if [ "$_torch_major" = "$_ta_major" ] && \
+                       [ "$_torch_minor" = "$_ta_minor" ] && \
+                       [ "$_tv_major" = "0" ] && \
+                       [ "$_tv_minor" = "$_expected_tv_minor" ]; then
+                        _radeon_versions_match=true
+                    fi
+                fi
+                if [ -z "$_torch_whl" ] || [ -z "$_tv_whl" ] || [ -z "$_ta_whl" ] || \
+                   [ "$_radeon_versions_match" != true ]; then
+                    substep "[WARN] Radeon repo lacks a compatible wheel set for this Python; falling back to ROCm index ($TORCH_INDEX_URL)" "$C_WARN"
+                    run_install_cmd "install PyTorch" uv pip install --python "$_VENV_PY" \
+                        "$TORCH_CONSTRAINT" torchvision torchaudio \
+                        --index-url "$TORCH_INDEX_URL"
+                else
+                    substep "installing PyTorch from Radeon repo (${_RADEON_BASE_URL})..."
+                    # Pass explicit wheel URLs so the matched trio is
+                    # installed together. --find-links lets uv discover
+                    # the Radeon listing for any local lookup, and PyPI
+                    # (not disabled) provides transitive deps like
+                    # filelock / sympy / networkx which are not in the
+                    # Radeon listing.
+                    if [ -n "$_tri_whl" ]; then
+                        run_install_cmd "install triton + PyTorch" uv pip install --python "$_VENV_PY" \
+                            --find-links "$_RADEON_BASE_URL" \
+                            "$_tri_whl" "$_torch_whl" "$_tv_whl" "$_ta_whl"
+                    else
+                        run_install_cmd "install PyTorch" uv pip install --python "$_VENV_PY" \
+                            --find-links "$_RADEON_BASE_URL" \
+                            "$_torch_whl" "$_tv_whl" "$_ta_whl"
+                    fi
+                fi
+            else
+                substep "[WARN] Radeon repo unavailable; falling back to ROCm index ($TORCH_INDEX_URL)" "$C_WARN"
+                run_install_cmd "install PyTorch" uv pip install --python "$_VENV_PY" \
+                    "$TORCH_CONSTRAINT" torchvision torchaudio \
+                    --index-url "$TORCH_INDEX_URL"
+            fi
+        else
+            substep "[WARN] Radeon GPU detected but could not detect full ROCm version; falling back to ROCm index" "$C_WARN"
+            run_install_cmd "install PyTorch" uv pip install --python "$_VENV_PY" \
+                "$TORCH_CONSTRAINT" torchvision torchaudio \
+                --index-url "$TORCH_INDEX_URL"
+        fi
     else
         substep "installing PyTorch ($TORCH_INDEX_URL)..."
         run_install_cmd "install PyTorch" uv pip install --python "$_VENV_PY" "$TORCH_CONSTRAINT" torchvision torchaudio \
             --index-url "$TORCH_INDEX_URL"
     fi
+    # AMD ROCm: install bitsandbytes (once, after torch, for all ROCm paths).
+    # Gate on SKIP_TORCH=false so a user running with --no-torch on a ROCm
+    # host stays in GGUF-only mode rather than pulling in bitsandbytes,
+    # which is only useful once torch is present for training.
+    if [ "$SKIP_TORCH" = false ]; then
+        case "$TORCH_INDEX_URL" in
+            */rocm*)
+                substep "installing bitsandbytes for AMD ROCm..."
+                run_install_cmd "install bitsandbytes (AMD)" uv pip install --python "$_VENV_PY" --force-reinstall --no-cache-dir --no-deps "bitsandbytes>=0.49.1"
+                ;;
+        esac
+    fi
     # Fresh: Step 2 - install unsloth, preserving pre-installed torch
     substep "installing unsloth (this may take a few minutes)..."
     if [ "$SKIP_TORCH" = true ]; then
@@ -1088,6 +1467,22 @@ elif [ -n "$TORCH_INDEX_URL" ]; then
         run_install_cmd "install unsloth" uv pip install --python "$_VENV_PY" \
             --upgrade-package unsloth "$PACKAGE_NAME"
     fi
+    # AMD ROCm: repair torch if the unsloth/unsloth-zoo install pulled in
+    # CUDA torch from PyPI, overwriting the ROCm wheels installed in Step 1.
+    if [ "$SKIP_TORCH" = false ]; then
+        case "$TORCH_INDEX_URL" in
+            */rocm*)
+                _has_hip=$("$_VENV_PY" -c "import torch; print(getattr(torch.version,'hip','') or '')" 2>/dev/null || true)
+                if [ -z "$_has_hip" ]; then
+                    substep "repairing ROCm torch (overwritten by dependency resolution)..."
+                    run_install_cmd "repair ROCm torch" uv pip install --python "$_VENV_PY" \
+                        "$TORCH_CONSTRAINT" torchvision torchaudio \
+                        --index-url "$TORCH_INDEX_URL" \
+                        --force-reinstall
+                fi
+                ;;
+        esac
+    fi
 else
     # Fallback: GPU detection failed to produce a URL -- let uv resolve torch
     substep "installing unsloth (this may take a few minutes)..."
diff --git a/studio/backend/core/inference/inference.py b/studio/backend/core/inference/inference.py
index 867bdefc62..d7bc3ad4c7 100644
--- a/studio/backend/core/inference/inference.py
+++ b/studio/backend/core/inference/inference.py
@@ -5,8 +5,23 @@
 Core inference backend - streamlined
 """
 
-from unsloth import FastLanguageModel, FastVisionModel
-from unsloth.chat_templates import get_chat_template
+# On AMD ROCm, Unsloth's global monkey-patching of transformers model classes
+# (LlamaRotaryEmbedding, attention modules, etc.) causes HIP kernel crashes
+# (_assert_async_cuda_kernel -> HSA_STATUS_ERROR_EXCEPTION) during inference.
+# Training works because it uses different code paths, but generation triggers
+# the incompatible patched kernels.  Skip the Unsloth import entirely on ROCm
+# so transformers classes stay unmodified; the GGUF inference path (llama-server)
+# is unaffected since it never imports these Python model classes.
+_IS_ROCM_ENV = getattr(__import__("torch").version, "hip", None) is not None
+
+if _IS_ROCM_ENV:
+    FastLanguageModel = None  # Loaded on-demand only on NVIDIA
+    FastVisionModel = None
+    get_chat_template = None
+else:
+    from unsloth import FastLanguageModel, FastVisionModel
+    from unsloth.chat_templates import get_chat_template
+
 from transformers import TextStreamer
 from peft import PeftModel, PeftModelForCausalLM
 
@@ -26,6 +41,7 @@ from utils.hardware import (
     raise_if_offloaded,
     get_visible_gpu_count,
 )
+from utils.hardware import hardware as _hw_module
 from core.inference.audio_codecs import AudioCodecManager
 from io import StringIO
 import structlog
@@ -253,6 +269,15 @@ class InferenceBackend:
         """
         Load any model: base, LoRA adapter, text, or vision.
         """
+        # max_seq_length=0 means "model default" for the GGUF/llama.cpp path,
+        # but Unsloth's FastLanguageModel.from_pretrained treats 0 literally --
+        # setting the model's context to 0 tokens, which triggers an assertion
+        # crash during generation (especially on ROCm/HIP where the async
+        # assert kernel raises a hardware exception instead of a Python error).
+        # Fall back to 2048 for the Unsloth/transformers path.
+        if max_seq_length <= 0:
+            max_seq_length = 2048
+
         try:
             model_name = config.identifier
 
@@ -286,6 +311,12 @@ class InferenceBackend:
             }
 
             # ── Audio model loading path ──────────────────────────
+            if (config.is_audio or config.is_vision) and _IS_ROCM_ENV:
+                raise RuntimeError(
+                    f"Audio and vision model inference via Unsloth is not "
+                    f"yet supported on AMD ROCm.  Use GGUF inference instead."
+                )
+
             if config.is_audio:
                 audio_type = config.audio_type
                 adapter_info = " (LoRA adapter)" if config.is_lora else ""
@@ -516,18 +547,84 @@ class InferenceBackend:
 
             else:
                 # Text model (or text LoRA adapter)
-                model, tokenizer = FastLanguageModel.from_pretrained(
-                    model_name = config.path,  # Can be base model OR LoRA adapter path
-                    max_seq_length = max_seq_length,
-                    dtype = dtype,
-                    load_in_4bit = load_in_4bit,
-                    device_map = device_map,
-                    token = hf_token if hf_token and hf_token.strip() else None,
-                    trust_remote_code = trust_remote_code,
-                )
+                if _hw_module.IS_ROCM:
+                    # On AMD ROCm two issues prevent the normal Unsloth path:
+                    #  1. Unsloth's patched kernels (RoPE, attention) crash on
+                    #     HIP (_assert_async_cuda_kernel -> HSA_STATUS_ERROR).
+                    #  2. bitsandbytes 4-bit matmul kernels trigger the same
+                    #     HIP assertion on MI300X (CDNA3 / gfx942).
+                    # Fall back to plain transformers + PEFT in 16-bit, which
+                    # works reliably.  AMD GPUs typically have large VRAM so
+                    # 16-bit is practical; GGUF inference remains the
+                    # recommended path for memory-constrained setups.
+                    logger.info(
+                        "ROCm detected -- loading in 16-bit with plain "
+                        "transformers (bitsandbytes 4-bit and Unsloth kernels "
+                        "are not yet compatible with HIP)"
+                    )
+                    from transformers import AutoModelForCausalLM, AutoTokenizer
 
-                # Apply inference optimization
-                FastLanguageModel.for_inference(model)
+                    _load_kwargs = dict(
+                        dtype = dtype or torch.bfloat16,
+                        device_map = device_map,
+                        token = hf_token if hf_token and hf_token.strip() else None,
+                        trust_remote_code = trust_remote_code,
+                    )
+
+                    # Skip 4-bit on ROCm: bnb matmul kernels crash on HIP.
+                    # Also resolve pre-quantized Unsloth model names (e.g.
+                    # "unsloth/xxx-bnb-4bit") to their FP16 originals since
+                    # loading a pre-quantized repo still triggers bnb codepaths.
+                    def _resolve_fp16_base(name: str) -> str:
+                        if not name:
+                            return name
+                        # Strip Unsloth quantization suffixes to get the FP16 model:
+                        # "unsloth/Foo-unsloth-bnb-4bit" -> "unsloth/Foo"
+                        # "unsloth/Foo-bnb-4bit"         -> "unsloth/Foo"
+                        # Order matters: try longer suffix first.
+                        for suffix in ("-unsloth-bnb-4bit", "-bnb-4bit"):
+                            if name.lower().endswith(suffix):
+                                resolved = name[: -len(suffix)]
+                                logger.info(
+                                    "Resolved pre-quantized base '%s' -> '%s' for ROCm 16-bit inference",
+                                    name,
+                                    resolved,
+                                )
+                                return resolved
+                        return name
+
+                    if config.is_lora and config.base_model:
+                        # Load base model then apply adapter
+                        _base = _resolve_fp16_base(config.base_model)
+                        model = AutoModelForCausalLM.from_pretrained(
+                            _base,
+                            **_load_kwargs,
+                        )
+                        from peft import PeftModel
+
+                        model = PeftModel.from_pretrained(model, config.path)
+                        tokenizer = AutoTokenizer.from_pretrained(config.path)
+                    else:
+                        _path = _resolve_fp16_base(config.path)
+                        model = AutoModelForCausalLM.from_pretrained(
+                            _path,
+                            **_load_kwargs,
+                        )
+                        tokenizer = AutoTokenizer.from_pretrained(config.path)
+                    model.eval()
+                else:
+                    model, tokenizer = FastLanguageModel.from_pretrained(
+                        model_name = config.path,  # Can be base model OR LoRA adapter path
+                        max_seq_length = max_seq_length,
+                        dtype = dtype,
+                        load_in_4bit = load_in_4bit,
+                        device_map = device_map,
+                        token = hf_token if hf_token and hf_token.strip() else None,
+                        trust_remote_code = trust_remote_code,
+                    )
+
+                    # Apply inference optimization
+                    FastLanguageModel.for_inference(model)
 
                 self.models[model_name]["model"] = model
                 self.models[model_name]["tokenizer"] = tokenizer
@@ -950,10 +1047,13 @@ class InferenceBackend:
                 )
 
                 # This modifies the tokenizer with the correct template
-                tokenizer = get_chat_template(
-                    tokenizer,
-                    chat_template = template_name,
-                )
+                if get_chat_template is not None:
+                    tokenizer = get_chat_template(
+                        tokenizer,
+                        chat_template = template_name,
+                    )
+                else:
+                    logger.info("Skipping Unsloth chat template (ROCm fallback)")
             else:
                 logger.info(
                     f"No registered Unsloth template for {self.active_model_name}, using tokenizer default"
diff --git a/studio/backend/core/training/worker.py b/studio/backend/core/training/worker.py
index a461972eca..c3d3daf07e 100644
--- a/studio/backend/core/training/worker.py
+++ b/studio/backend/core/training/worker.py
@@ -86,6 +86,7 @@ def _probe_causal_conv1d_env() -> dict[str, str] | None:
                     "'python_tag': f'cp{sys.version_info.major}{sys.version_info.minor}', "
                     "'torch_mm': torch_mm, "
                     "'cuda_major': str(int(str(torch.version.cuda).split('.', 1)[0])) if torch.version.cuda else '', "
+                    "'hip_version': str(torch.version.hip) if getattr(torch.version, 'hip', None) else '', "
                     "'cxx11abi': str(torch._C._GLIBCXX_USE_CXX11_ABI).upper()"
                     "}))"
                 ),
@@ -237,28 +238,111 @@ def _install_package_wheel_first(
         else:
             logger.info("No published %s wheel found: %s", display_name, wheel_url)
 
-    _send_status(event_queue, f"Installing {display_name} from PyPI...")
-    pypi_cmd = [
-        sys.executable,
-        "-m",
-        "pip",
-        "install",
-        "--no-build-isolation",
-        "--no-deps",
-        "--no-cache-dir",
-        f"{pypi_name}=={pypi_version}",
-    ]
-    result = _sp.run(
-        pypi_cmd,
-        stdout = _sp.PIPE,
-        stderr = _sp.STDOUT,
-        text = True,
-    )
-    if result.returncode != 0:
-        logger.error("Failed to install %s from PyPI:\n%s", display_name, result.stdout)
+    is_hip = env and env.get("hip_version")
+    if is_hip and not shutil.which("hipcc"):
+        logger.error(
+            "%s requires hipcc for source compilation on ROCm. "
+            "Install the ROCm HIP SDK: https://rocm.docs.amd.com",
+            display_name,
+        )
+        _send_status(
+            event_queue,
+            f"{display_name}: hipcc not found (ROCm HIP SDK required)",
+        )
         return
 
-    logger.info("Installed %s from PyPI", display_name)
+    if is_hip:
+        _send_status(
+            event_queue,
+            f"Compiling {display_name} from source for ROCm "
+            "(this may take several minutes)...",
+        )
+    else:
+        _send_status(event_queue, f"Installing {display_name} from PyPI...")
+
+    # Prefer uv for faster dependency resolution when available
+    if shutil.which("uv"):
+        pypi_cmd = [
+            "uv",
+            "pip",
+            "install",
+            "--python",
+            sys.executable,
+            "--no-build-isolation",
+            "--no-deps",
+        ]
+        # Avoid stale cache artifacts from partial HIP source builds
+        if is_hip:
+            pypi_cmd.append("--no-cache")
+        pypi_cmd.append(f"{pypi_name}=={pypi_version}")
+    else:
+        pypi_cmd = [
+            sys.executable,
+            "-m",
+            "pip",
+            "install",
+            "--no-build-isolation",
+            "--no-deps",
+            "--no-cache-dir",
+            f"{pypi_name}=={pypi_version}",
+        ]
+
+    # Source compilation on ROCm can take 10-30 minutes; use a generous
+    # timeout. Non-HIP installs preserve the pre-existing "no timeout"
+    # behaviour so unrelated slow installs (e.g. causal-conv1d source
+    # build on Linux aarch64 or unsupported torch/CUDA combinations)
+    # are not aborted at 5 minutes by this PR.
+    _run_kwargs: dict[str, Any] = {
+        "stdout": _sp.PIPE,
+        "stderr": _sp.STDOUT,
+        "text": True,
+    }
+    if is_hip:
+        _run_kwargs["timeout"] = 1800
+
+    try:
+        result = _sp.run(pypi_cmd, **_run_kwargs)
+    except _sp.TimeoutExpired:
+        logger.error(
+            "%s installation timed out after %ds",
+            display_name,
+            _run_kwargs.get("timeout"),
+        )
+        _send_status(
+            event_queue,
+            f"{display_name} installation timed out after "
+            f"{_run_kwargs.get('timeout')}s",
+        )
+        return
+
+    if result.returncode != 0:
+        if is_hip:
+            # Surface a clear error for ROCm source build failures
+            error_lines = (result.stdout or "").strip().splitlines()
+            snippet = "\n".join(error_lines[-5:]) if error_lines else "(no output)"
+            logger.error(
+                "Failed to compile %s for ROCm:\n%s",
+                display_name,
+                result.stdout,
+            )
+            _send_status(
+                event_queue,
+                f"Failed to compile {display_name} for ROCm. "
+                "Check that hipcc and ROCm development headers are installed.\n"
+                f"{snippet}",
+            )
+        else:
+            logger.error(
+                "Failed to install %s from PyPI:\n%s",
+                display_name,
+                result.stdout,
+            )
+        return
+
+    if is_hip:
+        logger.info("Compiled and installed %s from source for ROCm", display_name)
+    else:
+        logger.info("Installed %s from PyPI", display_name)
 
 
 def _ensure_causal_conv1d_fast_path(event_queue: Any, model_name: str) -> None:
diff --git a/studio/backend/main.py b/studio/backend/main.py
index ad19ee9679..c2c0a0b6e4 100644
--- a/studio/backend/main.py
+++ b/studio/backend/main.py
@@ -237,6 +237,7 @@ async def get_system_info():
     import platform
     import psutil
     from utils.hardware import get_device
+    from utils.hardware.hardware import _backend_label
 
     visibility_info = get_backend_visible_gpu_info()
     gpu_info = {
@@ -250,7 +251,10 @@ async def get_system_info():
     return {
         "platform": platform.platform(),
         "python_version": platform.python_version(),
-        "device_backend": get_device().value,
+        # Use the centralized _backend_label helper so the /api/system
+        # endpoint reports "rocm" on AMD hosts instead of "cuda", matching
+        # the /api/hardware and /api/gpu-visibility endpoints.
+        "device_backend": _backend_label(get_device()),
         "cpu_count": psutil.cpu_count(),
         "memory": {
             "total_gb": round(memory.total / 1e9, 2),
diff --git a/studio/backend/tests/test_utils.py b/studio/backend/tests/test_utils.py
index 50557c6718..64c9907119 100644
--- a/studio/backend/tests/test_utils.py
+++ b/studio/backend/tests/test_utils.py
@@ -191,8 +191,14 @@ class TestGetGpuMemoryInfo:
         assert "backend" in get_gpu_memory_info()
 
     def test_backend_matches_device(self):
+        # The backend field uses _backend_label, which swaps "cuda" for
+        # "rocm" when running on an AMD host (IS_ROCM=True) so the UI
+        # can render the correct label. On CUDA / XPU / MLX / CPU hosts
+        # it is equivalent to `get_device().value`.
+        from utils.hardware.hardware import _backend_label
+
         result = get_gpu_memory_info()
-        assert result["backend"] == get_device().value
+        assert result["backend"] == _backend_label(get_device())
 
     # --- When a GPU IS available ---
 
diff --git a/studio/backend/utils/hardware/__init__.py b/studio/backend/utils/hardware/__init__.py
index aaa0452406..400b5dd066 100644
--- a/studio/backend/utils/hardware/__init__.py
+++ b/studio/backend/utils/hardware/__init__.py
@@ -5,6 +5,7 @@
 Hardware detection and GPU utilities
 """
 
+from . import hardware as _hardware
 from .hardware import (
     DeviceType,
     DEVICE,
@@ -49,6 +50,7 @@ __all__ = [
     "DeviceType",
     "DEVICE",
     "CHAT_ONLY",
+    "IS_ROCM",
     "detect_hardware",
     "get_device",
     "is_apple_silicon",
@@ -81,3 +83,11 @@ __all__ = [
     "extract_arch_config",
     "estimate_training_vram",
 ]
+
+
+def __getattr__(name: str):
+    """Resolve IS_ROCM at access time so callers always see the live value
+    after detect_hardware() runs (it flips the flag in hardware.py)."""
+    if name == "IS_ROCM":
+        return getattr(_hardware, "IS_ROCM")
+    raise AttributeError(name)
diff --git a/studio/backend/utils/hardware/amd.py b/studio/backend/utils/hardware/amd.py
new file mode 100644
index 0000000000..755314ca3a
--- /dev/null
+++ b/studio/backend/utils/hardware/amd.py
@@ -0,0 +1,382 @@
+# SPDX-License-Identifier: AGPL-3.0-only
+# Copyright 2026-present the Unsloth AI Inc. team. All rights reserved. See /studio/LICENSE.AGPL-3.0
+
+"""AMD GPU monitoring via amd-smi.
+
+Mirrors the nvidia.py module structure so hardware.py can swap backends
+based on IS_ROCM. All functions return the same dict shapes as their
+nvidia.py counterparts.
+"""
+
+import json
+import math
+import os
+import re
+import subprocess
+from typing import Any, Optional
+
+from loggers import get_logger
+
+logger = get_logger(__name__)
+
+
+def _run_amd_smi(*args: str, timeout: int = 5) -> Optional[Any]:
+    """Run amd-smi with the given arguments and return parsed JSON, or None."""
+    try:
+        result = subprocess.run(
+            ["amd-smi", *args, "--json"],
+            capture_output = True,
+            text = True,
+            timeout = timeout,
+        )
+    except (OSError, subprocess.TimeoutExpired) as e:
+        logger.warning("amd-smi query failed: %s", e)
+        return None
+    if result.returncode != 0 or not result.stdout.strip():
+        logger.warning("amd-smi returned code %d", result.returncode)
+        return None
+    try:
+        return json.loads(result.stdout)
+    except json.JSONDecodeError:
+        logger.warning("Failed to parse amd-smi JSON output")
+        return None
+
+
+def _parse_numeric(value: Any) -> Optional[float]:
+    """Extract a numeric value from amd-smi output (may be str, int, float, or dict)."""
+    if value is None:
+        return None
+    # Newer amd-smi versions emit {"value": 10, "unit": "W"}
+    if isinstance(value, dict):
+        return _parse_numeric(value.get("value"))
+    if isinstance(value, (int, float)):
+        f = float(value)
+        return f if math.isfinite(f) else None
+    if isinstance(value, str):
+        # Strip units like "W", "C", "%", "MB", "MiB", "GB", "GiB" etc.
+        cleaned = re.sub(r"\s*[A-Za-z/%]+$", "", value.strip())
+        if not cleaned or cleaned.lower() in ("n/a", "none", "unknown"):
+            return None
+        try:
+            return float(cleaned)
+        except (ValueError, TypeError):
+            return None
+    return None
+
+
+def _parse_memory_mb(value: Any) -> Optional[float]:
+    """Parse a memory value from amd-smi output and return MB.
+
+    Handles bare numbers (assumed MB -- the amd-smi convention on every
+    version we have seen), dict-shaped values with explicit units
+    (``{"value": 192, "unit": "GiB"}`` on newer releases), and plain
+    strings like ``"8192 MiB"``.
+    """
+    unit = ""
+    raw_value = value
+
+    if isinstance(value, dict):
+        unit = str(value.get("unit", "")).strip().lower()
+        raw_value = value.get("value")
+    elif isinstance(value, str):
+        # Extract unit suffix from strings like "192 GiB" or "8192 MB"
+        m = re.match(r"^\s*([\d.]+)\s*([A-Za-z]+)\s*$", value.strip())
+        if m:
+            unit = m.group(2).lower()
+
+    num = _parse_numeric(raw_value if isinstance(value, dict) else value)
+    if num is None:
+        return None
+
+    # Unit conversion -- GPU tools (including amd-smi) use binary units even
+    # when labeling them "GB" or "MB", so treat GB/GiB and MB/MiB the same.
+    if "gib" in unit or "gb" in unit:
+        return num * 1024
+    if "mib" in unit or "mb" in unit:
+        return num
+    if "kib" in unit or "kb" in unit:
+        return num / 1024
+    if unit in ("b", "byte", "bytes"):
+        # Plain bytes
+        return num / (1024 * 1024)
+
+    # No explicit unit -- default to MB, which is the amd-smi convention
+    # for bare numeric values. A previous heuristic assumed values above
+    # ~10M were bytes, but that misclassifies small VRAM allocations
+    # (e.g. 5 MB = 5,242,880 reported without a unit) as ~5 TB. Modern
+    # amd-smi always ships explicit units, so the heuristic branch only
+    # fired for legacy output where MB was already the convention.
+    return num
+
+
+def _extract_gpu_metrics(gpu_data: dict) -> dict[str, Any]:
+    """Extract standardized metrics from a single GPU's amd-smi data."""
+    # amd-smi metric output structure varies by version; try common paths
+    usage = gpu_data.get("usage", gpu_data.get("gpu_activity", {}))
+    if isinstance(usage, dict):
+        gpu_util = _parse_numeric(
+            usage.get("gfx_activity", usage.get("gpu_use_percent"))
+        )
+    else:
+        gpu_util = _parse_numeric(usage)
+
+    # Temperature -- try multiple keys in priority order.
+    # dict.get() returns "N/A" strings rather than falling through,
+    # so we must try each key and check if it parses to a real number.
+    temp_data = gpu_data.get("temperature", {})
+    temp = None
+    if isinstance(temp_data, dict):
+        for temp_key in ("edge", "temperature_edge", "hotspot", "temperature_hotspot"):
+            temp = _parse_numeric(temp_data.get(temp_key))
+            if temp is not None:
+                break
+    else:
+        temp = _parse_numeric(temp_data)
+
+    # Power
+    power_data = gpu_data.get("power", {})
+    if isinstance(power_data, dict):
+        power_draw = _parse_numeric(
+            power_data.get(
+                "current_socket_power",
+                power_data.get("average_socket_power", power_data.get("socket_power")),
+            )
+        )
+        power_limit = _parse_numeric(
+            power_data.get("power_cap", power_data.get("max_power_limit"))
+        )
+    else:
+        power_draw = None
+        power_limit = None
+
+    # VRAM -- unit-aware parsing to handle varying amd-smi output formats.
+    # Newer amd-smi versions may return {"value": 192, "unit": "GiB"}.
+    # Newer amd-smi uses "mem_usage" with "total_vram" / "used_vram" keys;
+    # older versions use "vram" or "fb_memory_usage" with "used" / "total".
+    vram_data = gpu_data.get(
+        "mem_usage",
+        gpu_data.get("vram", gpu_data.get("fb_memory_usage", {})),
+    )
+    if isinstance(vram_data, dict):
+        vram_used_mb = _parse_memory_mb(
+            vram_data.get(
+                "used_vram", vram_data.get("vram_used", vram_data.get("used"))
+            )
+        )
+        vram_total_mb = _parse_memory_mb(
+            vram_data.get(
+                "total_vram", vram_data.get("vram_total", vram_data.get("total"))
+            )
+        )
+    else:
+        vram_used_mb = None
+        vram_total_mb = None
+
+    # Build the standardized dict (same shape as nvidia._build_gpu_metrics)
+    vram_used_gb = round(vram_used_mb / 1024, 2) if vram_used_mb is not None else None
+    vram_total_gb = (
+        round(vram_total_mb / 1024, 2) if vram_total_mb is not None else None
+    )
+    vram_util = (
+        round((vram_used_mb / vram_total_mb) * 100, 1)
+        if vram_used_mb is not None and vram_total_mb is not None and vram_total_mb > 0
+        else None
+    )
+    power_util = (
+        round((power_draw / power_limit) * 100, 1)
+        if power_draw is not None and power_limit is not None and power_limit > 0
+        else None
+    )
+
+    return {
+        "gpu_utilization_pct": gpu_util,
+        "temperature_c": temp,
+        "vram_used_gb": vram_used_gb,
+        "vram_total_gb": vram_total_gb,
+        "vram_utilization_pct": vram_util,
+        "power_draw_w": power_draw,
+        "power_limit_w": power_limit,
+        "power_utilization_pct": power_util,
+    }
+
+
+def _has_real_metrics(metrics: dict[str, Any]) -> bool:
+    """Return True when ``metrics`` contains at least one non-None value.
+
+    ``amd-smi`` can return a zero-exit JSON envelope that is missing every
+    expected field (error response, unsupported card, hipless container).
+    In that case ``_extract_gpu_metrics`` produces a dict where every value
+    is ``None`` -- callers must surface this as ``available: False`` rather
+    than ``available: True`` with empty data.
+    """
+    return any(value is not None for value in metrics.values())
+
+
+def get_physical_gpu_count() -> Optional[int]:
+    """Return physical AMD GPU count via amd-smi, or None on failure."""
+    data = _run_amd_smi("list")
+    if data is None:
+        return None
+    if isinstance(data, list):
+        return len(data)
+    # Some versions return a dict with a "gpu" / "gpus" key. Guard the
+    # .get() access with an isinstance check so a malformed scalar /
+    # string response from amd-smi cannot raise AttributeError.
+    if not isinstance(data, dict):
+        return None
+    gpus = data.get("gpu", data.get("gpus", []))
+    if isinstance(gpus, list):
+        return len(gpus)
+    return None
+
+
+def _first_visible_amd_gpu_id() -> Optional[str]:
+    """Return the physical AMD GPU id that should be treated as 'primary'.
+
+    Honours HIP_VISIBLE_DEVICES / ROCR_VISIBLE_DEVICES / CUDA_VISIBLE_DEVICES
+    in that order (HIP respects all three). Returns ``"0"`` when none are
+    set, and ``None`` when the env var explicitly narrows to zero GPUs
+    ("" or "-1"), so callers can short-circuit to "available: False".
+    """
+    for env_name in (
+        "HIP_VISIBLE_DEVICES",
+        "ROCR_VISIBLE_DEVICES",
+        "CUDA_VISIBLE_DEVICES",
+    ):
+        raw = os.environ.get(env_name)
+        if raw is None:
+            continue
+        raw = raw.strip()
+        if raw == "" or raw == "-1":
+            return None
+        # Filter out empty tokens after splitting. This tolerates minor
+        # typos like ``HIP_VISIBLE_DEVICES=",1"`` (leading comma, user
+        # clearly meant to narrow to device 1) while still falling
+        # through to the next env var when every token is empty
+        # (e.g. ``,,,``).
+        tokens = [t.strip() for t in raw.split(",") if t.strip()]
+        if tokens:
+            return tokens[0]
+    return "0"
+
+
+def get_primary_gpu_utilization() -> dict[str, Any]:
+    """Return utilization metrics for the primary visible AMD GPU."""
+    gpu_idx = _first_visible_amd_gpu_id()
+    if gpu_idx is None:
+        return {"available": False}
+    data = _run_amd_smi("metric", "-g", gpu_idx)
+    if data is None:
+        return {"available": False}
+
+    # amd-smi may return:
+    #   - a list of GPU dicts (older versions)
+    #   - a dict with a "gpu_data" key wrapping a list (newer versions)
+    #   - a single GPU dict (rare)
+    if isinstance(data, dict) and "gpu_data" in data:
+        data = data["gpu_data"]
+    if isinstance(data, list):
+        if len(data) == 0:
+            return {"available": False}
+        gpu_data = data[0]
+    else:
+        gpu_data = data
+
+    metrics = _extract_gpu_metrics(gpu_data)
+    if not _has_real_metrics(metrics):
+        # amd-smi returned a JSON envelope with no usable fields (error
+        # response or unsupported card). Surface as unavailable rather
+        # than available-with-empty-data so the UI does not render a
+        # ghost device.
+        return {"available": False}
+    metrics["available"] = True
+    return metrics
+
+
+def get_visible_gpu_utilization(
+    parent_visible_ids: Optional[list[int]],
+    parent_cuda_visible_devices: Optional[str] = None,
+) -> dict[str, Any]:
+    """Return utilization metrics for visible AMD GPUs."""
+    if parent_visible_ids is None:
+        return {
+            "available": False,
+            "backend_cuda_visible_devices": parent_cuda_visible_devices,
+            "parent_visible_gpu_ids": [],
+            "devices": [],
+            "index_kind": "unresolved",
+        }
+
+    data = _run_amd_smi("metric")
+    if data is None:
+        return {
+            "available": False,
+            "backend_cuda_visible_devices": parent_cuda_visible_devices,
+            "parent_visible_gpu_ids": parent_visible_ids or [],
+            "devices": [],
+            "index_kind": "physical",
+        }
+
+    # Extract a device list from amd-smi's envelope. Newer versions return
+    # a JSON array directly, older versions return a dict with a "gpus" /
+    # "gpu" key wrapping the list. Guard non-dict / non-list envelopes
+    # (scalar / string fallbacks from malformed output) so the .get()
+    # access cannot raise AttributeError on an unexpected shape.
+    if isinstance(data, list):
+        gpu_list = data
+    elif isinstance(data, dict):
+        # Newer amd-smi wraps output in {"gpu_data": [...]}
+        gpu_list = data.get("gpu_data", data.get("gpus", data.get("gpu", [data])))
+    else:
+        gpu_list = [data]
+    visible_set = set(parent_visible_ids)
+    ordinal_map = {gpu_id: ordinal for ordinal, gpu_id in enumerate(parent_visible_ids)}
+
+    devices = []
+    for fallback_idx, gpu_data in enumerate(gpu_list):
+        # Skip non-dict entries defensively: if amd-smi ever ships a
+        # scalar inside its "gpus" array (observed on some malformed
+        # output), _extract_gpu_metrics would raise AttributeError on
+        # the first .get() call.
+        if not isinstance(gpu_data, dict):
+            continue
+        # Use AMD-reported GPU ID when available, fall back to enumeration
+        # index. Newer amd-smi versions wrap scalars as ``{"value": 0,
+        # "unit": "none"}``, so route raw_id through ``_parse_numeric``
+        # which already handles bare ints, floats, strings, and that
+        # dict shape uniformly.
+        raw_id = gpu_data.get(
+            "gpu", gpu_data.get("gpu_id", gpu_data.get("id", fallback_idx))
+        )
+        parsed_id = _parse_numeric(raw_id)
+        if parsed_id is None:
+            logger.debug(
+                "amd-smi GPU id %r could not be parsed; falling back to "
+                "enumeration index %d",
+                raw_id,
+                fallback_idx,
+            )
+            idx = fallback_idx
+        else:
+            idx = int(parsed_id)
+        if idx not in visible_set:
+            continue
+        metrics = _extract_gpu_metrics(gpu_data)
+        if not _has_real_metrics(metrics):
+            # Skip ghost entries: an amd-smi response that decodes to a
+            # dict but contains no usable fields (error envelope, etc.)
+            # would otherwise show up as a device row with all-None
+            # numbers in the UI.
+            continue
+        metrics["index"] = idx
+        metrics["index_kind"] = "physical"
+        metrics["visible_ordinal"] = ordinal_map.get(idx, len(devices))
+        devices.append(metrics)
+
+    return {
+        "available": len(devices) > 0,
+        "backend_cuda_visible_devices": parent_cuda_visible_devices,
+        "parent_visible_gpu_ids": parent_visible_ids or [],
+        "devices": devices,
+        "index_kind": "physical",
+    }
diff --git a/studio/backend/utils/hardware/hardware.py b/studio/backend/utils/hardware/hardware.py
index b6d3faf6d7..be31c00a78 100644
--- a/studio/backend/utils/hardware/hardware.py
+++ b/studio/backend/utils/hardware/hardware.py
@@ -43,6 +43,26 @@ class DeviceType(str, Enum):
 
 DEVICE: Optional[DeviceType] = None
 CHAT_ONLY: bool = True  # No CUDA GPU -> GGUF chat only (Mac, CPU-only, etc.)
+IS_ROCM: bool = (
+    False  # True when running on AMD ROCm (HIP) -- routes GPU monitoring to amd.py
+)
+
+
+def _backend_label(device: DeviceType) -> str:
+    """Return the user-facing backend name for API responses.
+
+    Internally we still represent ROCm hosts as ``DeviceType.CUDA`` because
+    ROCm torch sets ``torch.cuda.is_available() = True`` and reuses the whole
+    ``torch.cuda.*`` API surface, so branching on ``DeviceType`` stays
+    consistent with the rest of the codebase. For the JSON responses served
+    to the Studio frontend and other clients, however, "cuda" is misleading
+    on an AMD machine. This helper swaps the label to ``"rocm"`` when the
+    module-level ``IS_ROCM`` flag is set so the UI can render the correct
+    backend name without every caller having to duplicate the check.
+    """
+    if IS_ROCM and device == DeviceType.CUDA:
+        return "rocm"
+    return device.value
 
 
 # ========== Detection ==========
@@ -85,10 +105,11 @@ def detect_hardware() -> DeviceType:
       2. MLX   (Apple Silicon via MLX framework)
       3. CPU   (fallback)
     """
-    global DEVICE, CHAT_ONLY
-    CHAT_ONLY = True  # reset -- only CUDA sets it to False
+    global DEVICE, CHAT_ONLY, IS_ROCM
+    CHAT_ONLY = True  # reset -- only CUDA/ROCm sets it to False
+    IS_ROCM = False
 
-    # --- CUDA: try PyTorch ---
+    # --- CUDA / ROCm: try PyTorch ---
     if _has_torch():
         import torch
 
@@ -96,7 +117,16 @@ def detect_hardware() -> DeviceType:
             DEVICE = DeviceType.CUDA
             CHAT_ONLY = False
             device_name = torch.cuda.get_device_properties(0).name
-            print(f"Hardware detected: CUDA — {device_name}")
+
+            # Distinguish AMD ROCm (HIP) from NVIDIA CUDA for display purposes.
+            # DeviceType stays CUDA since torch.cuda.* works on ROCm via HIP.
+            if getattr(torch.version, "hip", None) is not None:
+                IS_ROCM = True
+                print(
+                    f"Hardware detected: ROCm (HIP {torch.version.hip}) -- {device_name}"
+                )
+            else:
+                print(f"Hardware detected: CUDA -- {device_name}")
             return DEVICE
 
     # --- XPU: Intel GPU ---
@@ -186,7 +216,7 @@ def get_gpu_memory_info() -> Dict[str, Any]:
 
             return {
                 "available": True,
-                "backend": device.value,
+                "backend": _backend_label(device),
                 "device": idx,
                 "device_name": props.name,
                 "total_gb": total / (1024**3),
@@ -197,7 +227,11 @@ def get_gpu_memory_info() -> Dict[str, Any]:
             }
         except Exception as e:
             logger.error(f"Error getting CUDA GPU info: {e}")
-            return {"available": False, "backend": device.value, "error": str(e)}
+            return {
+                "available": False,
+                "backend": _backend_label(device),
+                "error": str(e),
+            }
 
     # ---- XPU path (Intel GPU) ----
     if device == DeviceType.XPU:
@@ -213,7 +247,7 @@ def get_gpu_memory_info() -> Dict[str, Any]:
 
             return {
                 "available": True,
-                "backend": device.value,
+                "backend": _backend_label(device),
                 "device": idx,
                 "device_name": props.name,
                 "total_gb": total / (1024**3),
@@ -224,7 +258,11 @@ def get_gpu_memory_info() -> Dict[str, Any]:
             }
         except Exception as e:
             logger.error("Error getting XPU GPU info: %s", e)
-            return {"available": False, "backend": device.value, "error": str(e)}
+            return {
+                "available": False,
+                "backend": _backend_label(device),
+                "error": str(e),
+            }
 
     # ---- MLX path (Apple Silicon) ----
     if device == DeviceType.MLX:
@@ -239,7 +277,7 @@ def get_gpu_memory_info() -> Dict[str, Any]:
 
             return {
                 "available": True,
-                "backend": device.value,
+                "backend": _backend_label(device),
                 "device": 0,
                 "device_name": f"Apple Silicon ({platform.processor() or platform.machine()})",
                 "total_gb": total / (1024**3),
@@ -250,7 +288,11 @@ def get_gpu_memory_info() -> Dict[str, Any]:
             }
         except Exception as e:
             logger.error(f"Error getting MLX GPU info: {e}")
-            return {"available": False, "backend": device.value, "error": str(e)}
+            return {
+                "available": False,
+                "backend": _backend_label(device),
+                "error": str(e),
+            }
 
     # ---- CPU-only ----
     return {"available": False, "backend": "cpu"}
@@ -315,13 +357,15 @@ def get_package_versions() -> Dict[str, Optional[str]]:
         except PackageNotFoundError:
             versions[name] = None
 
-    # CUDA toolkit version bundled with torch
+    # GPU runtime version bundled with torch
     try:
         import torch
 
         versions["cuda"] = getattr(torch.version, "cuda", None)
+        versions["rocm"] = getattr(torch.version, "hip", None)
     except Exception:
         versions["cuda"] = None
+        versions["rocm"] = None
 
     return versions
 
@@ -387,26 +431,50 @@ def _torch_get_per_device_info(device_indices: list[int]) -> list[Dict[str, Any]
 # ========== Live GPU Utilization ==========
 
 
+def _smi_query(func_name: str, *args, **kwargs) -> Optional[Dict[str, Any]]:
+    """Run a query against the appropriate SMI backend (amd-smi or nvidia-smi).
+
+    Returns the result dict if available, or None on failure/unavailability.
+    """
+    if IS_ROCM:
+        backend_name = "amd-smi"
+        try:
+            from . import amd as _backend
+        except Exception as e:
+            logger.warning("%s import failed: %s", backend_name, e)
+            return None
+    else:
+        backend_name = "nvidia-smi"
+        try:
+            from . import nvidia as _backend
+        except Exception as e:
+            logger.warning("%s import failed: %s", backend_name, e)
+            return None
+    try:
+        func = getattr(_backend, func_name)
+        result = func(*args, **kwargs)
+        if result.get("available"):
+            return result
+    except Exception as e:
+        logger.warning("%s %s query failed: %s", backend_name, func_name, e)
+    return None
+
+
 def get_gpu_utilization() -> Dict[str, Any]:
     """Return a live snapshot of device utilization information."""
     device = get_device()
 
     if device == DeviceType.CUDA:
-        try:
-            from . import nvidia
-
-            result = nvidia.get_primary_gpu_utilization()
-            if result.get("available"):
-                result["backend"] = device.value
-                return result
-        except Exception as e:
-            logger.warning("nvidia-smi utilization query failed: %s", e)
+        result = _smi_query("get_primary_gpu_utilization")
+        if result is not None:
+            result["backend"] = _backend_label(device)
+            return result
 
     mem = get_gpu_memory_info()
     if device != DeviceType.CPU and mem.get("available"):
         return {
             "available": True,
-            "backend": device.value,
+            "backend": _backend_label(device),
             "gpu_utilization_pct": None,
             "temperature_c": None,
             "vram_used_gb": round(mem.get("allocated_gb", 0), 2),
@@ -417,7 +485,7 @@ def get_gpu_utilization() -> Dict[str, Any]:
             "power_utilization_pct": None,
         }
 
-    return {"available": False, "backend": device.value}
+    return {"available": False, "backend": _backend_label(device)}
 
 
 def get_visible_gpu_utilization() -> Dict[str, Any]:
@@ -425,18 +493,14 @@ def get_visible_gpu_utilization() -> Dict[str, Any]:
 
     if device == DeviceType.CUDA:
         parent_visible_spec = _get_parent_visible_gpu_spec()
-        try:
-            from . import nvidia
-
-            result = nvidia.get_visible_gpu_utilization(
-                parent_visible_spec["numeric_ids"],
-                parent_cuda_visible_devices = parent_visible_spec["raw"],
-            )
-            if result.get("available"):
-                result["backend"] = device.value
-                return result
-        except Exception as e:
-            logger.warning("nvidia-smi visible GPU utilization query failed: %s", e)
+        result = _smi_query(
+            "get_visible_gpu_utilization",
+            parent_visible_spec["numeric_ids"],
+            parent_cuda_visible_devices = parent_visible_spec["raw"],
+        )
+        if result is not None:
+            result["backend"] = _backend_label(device)
+            return result
 
     # Torch-based fallback for CUDA (nvidia-smi unavailable, AMD ROCm) and XPU (Intel)
     if device in (DeviceType.CUDA, DeviceType.XPU):
@@ -475,7 +539,7 @@ def get_visible_gpu_utilization() -> Dict[str, Any]:
                 )
             return {
                 "available": True,
-                "backend": device.value,
+                "backend": _backend_label(device),
                 "parent_visible_gpu_ids": parent_ids,
                 "devices": devices,
                 "index_kind": index_kind,
@@ -486,14 +550,14 @@ def get_visible_gpu_utilization() -> Dict[str, Any]:
         if not mem.get("available"):
             return {
                 "available": False,
-                "backend": device.value,
+                "backend": _backend_label(device),
                 "parent_visible_gpu_ids": [],
                 "devices": [],
                 "index_kind": "relative",
             }
         return {
             "available": True,
-            "backend": device.value,
+            "backend": _backend_label(device),
             "parent_visible_gpu_ids": [0],
             "devices": [
                 {
@@ -515,7 +579,7 @@ def get_visible_gpu_utilization() -> Dict[str, Any]:
 
     return {
         "available": False,
-        "backend": device.value,
+        "backend": _backend_label(device),
         "parent_visible_gpu_ids": [],
         "devices": [],
         "index_kind": "relative",
@@ -529,7 +593,21 @@ _visible_gpu_count: Optional[int] = None
 
 
 def _get_parent_visible_gpu_spec() -> Dict[str, Any]:
-    cuda_visible = os.environ.get("CUDA_VISIBLE_DEVICES")
+    # ROCm uses HIP_VISIBLE_DEVICES / ROCR_VISIBLE_DEVICES in addition to
+    # CUDA_VISIBLE_DEVICES (which HIP also respects).  Check ROCm-specific
+    # env vars first so multi-GPU AMD setups are handled correctly.
+    # Use explicit None checks (not `or`) so empty string "" is honoured
+    # as "no visible GPUs" rather than falling through to CUDA_VISIBLE_DEVICES.
+    cuda_visible = None
+    if IS_ROCM:
+        hip_vis = os.environ.get("HIP_VISIBLE_DEVICES")
+        rocr_vis = os.environ.get("ROCR_VISIBLE_DEVICES")
+        if hip_vis is not None:
+            cuda_visible = hip_vis
+        elif rocr_vis is not None:
+            cuda_visible = rocr_vis
+    if cuda_visible is None:
+        cuda_visible = os.environ.get("CUDA_VISIBLE_DEVICES")
 
     if cuda_visible is None:
         return {
@@ -1109,15 +1187,17 @@ def get_physical_gpu_count() -> int:
 
     if device == DeviceType.CUDA:
         try:
-            from . import nvidia
-
-            count = nvidia.get_physical_gpu_count()
+            if IS_ROCM:
+                from . import amd as _smi_mod
+            else:
+                from . import nvidia as _smi_mod
+            count = _smi_mod.get_physical_gpu_count()
             if count is not None:
                 _physical_gpu_count = count
                 return _physical_gpu_count
         except Exception:
             pass
-        # nvidia-smi unavailable or failed — fall back to torch
+        # SMI tool unavailable or failed -- fall back to torch
         count = _torch_get_physical_gpu_count()
         _physical_gpu_count = count if count is not None else 1
         return _physical_gpu_count
@@ -1136,12 +1216,25 @@ def get_physical_gpu_count() -> int:
     return _physical_gpu_count
 
 
+def _backend_visible_devices_env() -> Optional[str]:
+    """Return the raw visibility env string that applies to this backend.
+
+    On ROCm, HIP_VISIBLE_DEVICES / ROCR_VISIBLE_DEVICES take precedence
+    over CUDA_VISIBLE_DEVICES; the helper mirrors the resolution logic in
+    ``_get_parent_visible_gpu_spec`` so ``backend_cuda_visible_devices``
+    reports the value that is actually narrowing the visible device set.
+    """
+    if IS_ROCM:
+        return _get_parent_visible_gpu_spec().get("raw")
+    return os.environ.get("CUDA_VISIBLE_DEVICES")
+
+
 def get_backend_visible_gpu_info() -> Dict[str, Any]:
     device = get_device()
     if device in (DeviceType.CUDA, DeviceType.XPU):
         parent_visible_ids = get_parent_visible_gpu_ids()
-        # Try nvidia-smi first (NVIDIA only)
-        if device == DeviceType.CUDA:
+        # Try native SMI tool first (nvidia-smi for NVIDIA, skipped for ROCm)
+        if device == DeviceType.CUDA and not IS_ROCM:
             try:
                 from . import nvidia
 
@@ -1151,7 +1244,7 @@ def get_backend_visible_gpu_info() -> Dict[str, Any]:
                     parent_visible_spec["raw"],
                 )
                 if result.get("available"):
-                    result["backend"] = device.value
+                    result["backend"] = _backend_label(device)
                     return result
             except Exception as e:
                 logger.warning("Backend GPU visibility query failed: %s", e)
@@ -1180,8 +1273,8 @@ def get_backend_visible_gpu_info() -> Dict[str, Any]:
             ]
             return {
                 "available": True,
-                "backend": device.value,
-                "backend_cuda_visible_devices": os.environ.get("CUDA_VISIBLE_DEVICES"),
+                "backend": _backend_label(device),
+                "backend_cuda_visible_devices": _backend_visible_devices_env(),
                 "parent_visible_gpu_ids": parent_visible_ids,
                 "devices": devices,
                 "index_kind": index_kind,
@@ -1189,8 +1282,8 @@ def get_backend_visible_gpu_info() -> Dict[str, Any]:
 
         return {
             "available": False,
-            "backend": device.value,
-            "backend_cuda_visible_devices": os.environ.get("CUDA_VISIBLE_DEVICES"),
+            "backend": _backend_label(device),
+            "backend_cuda_visible_devices": _backend_visible_devices_env(),
             "parent_visible_gpu_ids": parent_visible_ids,
             "devices": [],
             "index_kind": "physical",
@@ -1201,7 +1294,7 @@ def get_backend_visible_gpu_info() -> Dict[str, Any]:
         if not mem.get("available"):
             return {
                 "available": False,
-                "backend": device.value,
+                "backend": _backend_label(device),
                 "backend_cuda_visible_devices": os.environ.get("CUDA_VISIBLE_DEVICES"),
                 "parent_visible_gpu_ids": [],
                 "devices": [],
@@ -1209,7 +1302,7 @@ def get_backend_visible_gpu_info() -> Dict[str, Any]:
             }
         return {
             "available": True,
-            "backend": device.value,
+            "backend": _backend_label(device),
             "backend_cuda_visible_devices": os.environ.get("CUDA_VISIBLE_DEVICES"),
             "parent_visible_gpu_ids": [0],
             "devices": [
@@ -1226,7 +1319,7 @@ def get_backend_visible_gpu_info() -> Dict[str, Any]:
 
     return {
         "available": False,
-        "backend": device.value,
+        "backend": _backend_label(device),
         "backend_cuda_visible_devices": os.environ.get("CUDA_VISIBLE_DEVICES"),
         "parent_visible_gpu_ids": [],
         "devices": [],
@@ -1246,17 +1339,20 @@ def get_visible_gpu_count() -> int:
     if _visible_gpu_count is not None:
         return _visible_gpu_count
 
-    cuda_visible = os.environ.get("CUDA_VISIBLE_DEVICES")
-    if cuda_visible is not None:
-        # "" means zero GPUs, "0" means 1, "0,1,2" means 3
-        cuda_visible = cuda_visible.strip()
-        if cuda_visible == "" or cuda_visible == "-1":
+    # Use _get_parent_visible_gpu_spec() which already handles
+    # HIP_VISIBLE_DEVICES / ROCR_VISIBLE_DEVICES on ROCm.
+    visible_spec = _get_parent_visible_gpu_spec()
+    if visible_spec["raw"] is not None:
+        raw = visible_spec["raw"].strip()
+        if raw == "" or raw == "-1":
             _visible_gpu_count = 0
+        elif visible_spec["numeric_ids"] is not None:
+            _visible_gpu_count = len(visible_spec["numeric_ids"])
         else:
-            _visible_gpu_count = len([x for x in cuda_visible.split(",") if x.strip()])
+            _visible_gpu_count = len([x for x in raw.split(",") if x.strip()])
         return _visible_gpu_count
 
-    # CUDA_VISIBLE_DEVICES not set -- try torch, fall back to physical count
+    # No visibility env var set -- try torch, fall back to physical count
     try:
         import torch
 
@@ -1288,8 +1384,24 @@ def apply_gpu_ids(gpu_ids) -> None:
         value = str(gpu_ids)
 
     os.environ["CUDA_VISIBLE_DEVICES"] = value
+    # Keep ROCm visibility env vars in sync so _get_parent_visible_gpu_spec()
+    # picks up the narrowed set on AMD systems. Workers can call
+    # apply_gpu_ids() before detect_hardware() runs (so IS_ROCM is still
+    # its default False), so also mirror the selection whenever the
+    # parent process already set a ROCm visibility variable -- that
+    # way a downstream ROCm process inherits the narrowed mask even
+    # before Studio's hardware detection has classified the host.
+    _inherits_rocm_visibility = (
+        "HIP_VISIBLE_DEVICES" in os.environ or "ROCR_VISIBLE_DEVICES" in os.environ
+    )
+    if IS_ROCM or _inherits_rocm_visibility:
+        os.environ["HIP_VISIBLE_DEVICES"] = value
+        os.environ["ROCR_VISIBLE_DEVICES"] = value
     _visible_gpu_count = None
-    logger.info("Applied gpu_ids: CUDA_VISIBLE_DEVICES='%s'", value)
+    if IS_ROCM or _inherits_rocm_visibility:
+        logger.info("Applied gpu_ids: CUDA_VISIBLE_DEVICES='%s' (rocm)", value)
+    else:
+        logger.info("Applied gpu_ids: CUDA_VISIBLE_DEVICES='%s'", value)
 
 
 def get_device_map(
diff --git a/studio/install_llama_prebuilt.py b/studio/install_llama_prebuilt.py
index 8d06c7d0e1..e43a78baea 100755
--- a/studio/install_llama_prebuilt.py
+++ b/studio/install_llama_prebuilt.py
@@ -173,6 +173,7 @@ class HostInfo:
     visible_cuda_devices: str | None
     has_physical_nvidia: bool
     has_usable_nvidia: bool
+    has_rocm: bool = False
 
 
 @dataclass
@@ -2493,12 +2494,25 @@ def detect_host() -> HostInfo:
     has_physical_nvidia = False
     has_usable_nvidia = False
     if nvidia_smi:
+        # Require `nvidia-smi -L` to actually list a GPU before treating the
+        # host as NVIDIA. The banner text "NVIDIA-SMI ..." is printed even
+        # when the command fails to communicate with the driver (e.g. stale
+        # container leftovers), which would otherwise misclassify an AMD
+        # ROCm host as NVIDIA and short-circuit the ROCm path.
+        try:
+            listing = run_capture([nvidia_smi, "-L"], timeout = 20)
+            gpu_lines = [
+                line for line in listing.stdout.splitlines() if line.startswith("GPU ")
+            ]
+            if gpu_lines:
+                has_physical_nvidia = True
+                has_usable_nvidia = visible_device_tokens != []
+        except Exception:
+            pass
+
         try:
             result = run_capture([nvidia_smi], timeout = 20)
             merged = "\n".join(part for part in (result.stdout, result.stderr) if part)
-            if "NVIDIA-SMI" in merged:
-                has_physical_nvidia = True
-                has_usable_nvidia = visible_device_tokens != []
             for line in merged.splitlines():
                 if "CUDA Version:" in line:
                     raw = line.split("CUDA Version:", 1)[1].strip().split()[0]
@@ -2538,6 +2552,12 @@ def detect_host() -> HostInfo:
 
             if visible_gpu_rows:
                 has_usable_nvidia = True
+                # Older nvidia-smi versions (pre -L support) hit the
+                # except in the first try block but still succeed here,
+                # leaving has_physical_nvidia unset. Mirror the -L path
+                # so downstream diagnostics on line ~4390 still run.
+                if not has_physical_nvidia:
+                    has_physical_nvidia = True
             elif visible_device_tokens == []:
                 has_usable_nvidia = False
             elif supports_explicit_visible_device_matching(visible_device_tokens):
@@ -2547,6 +2567,50 @@ def detect_host() -> HostInfo:
         except Exception:
             pass
 
+    # Detect AMD ROCm (HIP) -- require actual GPU, not just tools installed
+
+    def _amd_smi_has_gpu(stdout: str) -> bool:
+        """Check for 'GPU: <number>' data rows, not just a table header."""
+        return bool(re.search(r"(?im)^gpu\s*[:\[]\s*\d", stdout))
+
+    has_rocm = False
+    if is_linux:
+        for _cmd, _check in (
+            # rocminfo: look for "gfxNNNN" with nonzero first digit (gfx000 is CPU agent)
+            (["rocminfo"], lambda out: bool(re.search(r"gfx[1-9]", out.lower()))),
+            (["amd-smi", "list"], _amd_smi_has_gpu),
+        ):
+            _exe = shutil.which(_cmd[0])
+            if not _exe:
+                continue
+            try:
+                _result = run_capture([_exe, *_cmd[1:]], timeout = 10)
+            except Exception:
+                continue
+            if _result.returncode == 0 and _result.stdout.strip():
+                if _check(_result.stdout):
+                    has_rocm = True
+                    break
+    elif is_windows:
+        # Windows: prefer active probes that validate GPU presence
+        for _cmd, _check in (
+            (["hipinfo"], lambda out: "gcnarchname" in out.lower()),
+            (["amd-smi", "list"], _amd_smi_has_gpu),
+        ):
+            _exe = shutil.which(_cmd[0])
+            if not _exe:
+                continue
+            try:
+                _result = run_capture([_exe, *_cmd[1:]], timeout = 10)
+            except Exception:
+                continue
+            if _result.returncode == 0 and _result.stdout.strip():
+                if _check(_result.stdout):
+                    has_rocm = True
+                    break
+        # Note: amdhip64.dll presence alone is NOT treated as GPU evidence
+        # since the HIP SDK can be installed without an AMD GPU.
+
     return HostInfo(
         system = system,
         machine = machine,
@@ -2561,6 +2625,7 @@ def detect_host() -> HostInfo:
         visible_cuda_devices = visible_cuda_devices,
         has_physical_nvidia = has_physical_nvidia,
         has_usable_nvidia = has_usable_nvidia,
+        has_rocm = has_rocm,
     )
 
 
@@ -2926,9 +2991,168 @@ def published_asset_choice_for_kind(
     return None
 
 
+def _detect_host_rocm_version() -> tuple[int, int] | None:
+    """Return (major, minor) of the installed ROCm runtime, or None.
+
+    Best-effort read from /opt/rocm/.info/version, amd-smi version, and
+    hipconfig --version. Used to pick a compatible upstream llama.cpp
+    ROCm prebuilt rather than always taking the numerically newest one
+    (which can be newer than the host runtime).
+    """
+    rocm_root = os.environ.get("ROCM_PATH") or "/opt/rocm"
+    for path in (
+        os.path.join(rocm_root, ".info", "version"),
+        os.path.join(rocm_root, "lib", "rocm_version"),
+    ):
+        try:
+            with open(path) as fh:
+                parts = fh.read().strip().split("-")[0].split(".")
+            # Explicit length guard avoids relying on the broad except
+            # below to swallow IndexError when the version file contains
+            # a single component (e.g. "6\n" on a partial install).
+            if len(parts) >= 2:
+                return int(parts[0]), int(parts[1])
+        except Exception:
+            pass
+    amd_smi = shutil.which("amd-smi")
+    if amd_smi:
+        try:
+            result = subprocess.run(
+                [amd_smi, "version"],
+                stdout = subprocess.PIPE,
+                stderr = subprocess.DEVNULL,
+                text = True,
+                timeout = 5,
+            )
+            if result.returncode == 0:
+                m = re.search(r"ROCm version:\s*(\d+)\.(\d+)", result.stdout)
+                if m:
+                    return int(m.group(1)), int(m.group(2))
+        except Exception:
+            pass
+    hipconfig = shutil.which("hipconfig")
+    if hipconfig:
+        try:
+            result = subprocess.run(
+                [hipconfig, "--version"],
+                stdout = subprocess.PIPE,
+                stderr = subprocess.DEVNULL,
+                text = True,
+                timeout = 5,
+            )
+            if result.returncode == 0:
+                raw = (result.stdout or "").strip().split("\n")[0]
+                parts = raw.split(".")
+                if (
+                    len(parts) >= 2
+                    and parts[0].isdigit()
+                    and parts[1].split("-")[0].isdigit()
+                ):
+                    return int(parts[0]), int(parts[1].split("-")[0])
+        except Exception:
+            pass
+
+    # Distro package-manager fallbacks. Mirrors install.sh::get_torch_index_url
+    # and _detect_rocm_version() in install_python_stack.py so package-managed
+    # ROCm hosts without /opt/rocm/.info/version still report a usable version
+    # and the <= host version filter in resolve_upstream_asset_choice picks
+    # the correct upstream prebuilt instead of the newest-regardless fallback.
+    for _cmd in (
+        ["dpkg-query", "-W", "-f=${Version}\n", "rocm-core"],
+        ["rpm", "-q", "--qf", "%{VERSION}\n", "rocm-core"],
+    ):
+        _exe = shutil.which(_cmd[0])
+        if not _exe:
+            continue
+        try:
+            _result = subprocess.run(
+                [_exe, *_cmd[1:]],
+                stdout = subprocess.PIPE,
+                stderr = subprocess.DEVNULL,
+                text = True,
+                timeout = 5,
+            )
+        except Exception:
+            continue
+        if _result.returncode != 0 or not _result.stdout.strip():
+            continue
+        _raw = _result.stdout.strip()
+        # dpkg can prepend an epoch ("1:6.3.0-1"); strip it before parsing.
+        _raw = re.sub(r"^\d+:", "", _raw)
+        _m = re.match(r"(\d+)[.-](\d+)", _raw)
+        if _m:
+            return int(_m.group(1)), int(_m.group(2))
+    return None
+
+
 def resolve_upstream_asset_choice(host: HostInfo, llama_tag: str) -> AssetChoice:
     upstream_assets = github_release_assets(UPSTREAM_REPO, llama_tag)
     if host.is_linux and host.is_x86_64:
+        # AMD ROCm: try upstream ROCm prebuilt first, then fall back to source build.
+        # Source build (via setup.sh) compiles with -DGGML_HIP=ON and auto-detects
+        # the exact GPU target via rocminfo, which is more reliable for consumer
+        # GPUs (e.g. gfx1151) that may not be in the prebuilt.
+        if host.has_rocm and not host.has_usable_nvidia:
+            # Scan upstream assets for any rocm-<version> prebuilt. When the
+            # host ROCm runtime version is known, pick the newest candidate
+            # whose major.minor is <= host version -- otherwise a ROCm 6.4
+            # host would download the rocm-7.2 tarball, fail preflight, and
+            # fall back to a source build even though a compatible 6.4
+            # prebuilt exists. If no compatible candidate matches (e.g. host
+            # runtime is older than every published prebuilt), fall back to
+            # the numerically newest so we at least try something.
+            _rocm_pattern = re.compile(
+                rf"llama-{re.escape(llama_tag)}-bin-ubuntu-rocm-([0-9]+(?:\.[0-9]+)*)-x64\.tar\.gz"
+            )
+            rocm_candidates: list[tuple[tuple[int, ...], str]] = []
+            for _name in upstream_assets:
+                _m = _rocm_pattern.match(_name)
+                if _m is None:
+                    continue
+                _parts = tuple(int(p) for p in _m.group(1).split("."))
+                rocm_candidates.append((_parts, _name))
+            rocm_candidates.sort(reverse = True)
+            _host_rocm_version = _detect_host_rocm_version()
+            _compatible: list[tuple[tuple[int, ...], str]] = rocm_candidates
+            if _host_rocm_version is not None:
+                _compatible = [
+                    item
+                    for item in rocm_candidates
+                    if item[0][:2] <= _host_rocm_version
+                ]
+            if rocm_candidates and not _compatible:
+                # Fall back to the newest candidate so a source build is
+                # not forced when the host runtime is older than every
+                # published prebuilt: preflight will still catch a true
+                # incompatibility and trigger a fallback.
+                _compatible = rocm_candidates[:1]
+            if _compatible:
+                rocm_name = _compatible[0][1]
+                if _host_rocm_version is not None:
+                    log(
+                        f"AMD ROCm {_host_rocm_version[0]}.{_host_rocm_version[1]} "
+                        f"detected -- trying upstream prebuilt {rocm_name}"
+                    )
+                else:
+                    log(f"AMD ROCm detected -- trying upstream prebuilt {rocm_name}")
+                log(
+                    "Note: if your ROCm runtime version differs significantly, "
+                    "this may fail preflight and fall back to a source build (safe)"
+                )
+                return AssetChoice(
+                    repo = UPSTREAM_REPO,
+                    tag = llama_tag,
+                    name = rocm_name,
+                    url = upstream_assets[rocm_name],
+                    source_label = "upstream",
+                    install_kind = "linux-rocm",
+                )
+            # No ROCm prebuilt available -- fall back to source build
+            raise PrebuiltFallback(
+                "AMD ROCm detected but no upstream ROCm prebuilt found; "
+                "falling back to source build with HIP support"
+            )
+
         upstream_name = f"llama-{llama_tag}-bin-ubuntu-x64.tar.gz"
         if upstream_name not in upstream_assets:
             raise PrebuiltFallback("upstream Linux CPU asset was not found")
@@ -2948,6 +3172,25 @@ def resolve_upstream_asset_choice(host: HostInfo, llama_tag: str) -> AssetChoice
                 return attempts[0]
             raise PrebuiltFallback("no compatible Windows CUDA asset was found")
 
+        # AMD ROCm on Windows: try HIP prebuilt
+        if host.has_rocm:
+            hip_name = f"llama-{llama_tag}-bin-win-hip-radeon-x64.zip"
+            if hip_name in upstream_assets:
+                log(
+                    f"AMD ROCm detected on Windows -- trying upstream HIP prebuilt {hip_name}"
+                )
+                return AssetChoice(
+                    repo = UPSTREAM_REPO,
+                    tag = llama_tag,
+                    name = hip_name,
+                    url = upstream_assets[hip_name],
+                    source_label = "upstream",
+                    install_kind = "windows-hip",
+                )
+            log(
+                "AMD ROCm detected on Windows but no HIP prebuilt found -- falling back to CPU"
+            )
+
         upstream_name = f"llama-{llama_tag}-bin-win-cpu-x64.zip"
         if upstream_name not in upstream_assets:
             raise PrebuiltFallback("upstream Windows CPU asset was not found")
@@ -3029,7 +3272,16 @@ def resolve_release_asset_choice(
 
     published_choice: AssetChoice | None = None
     if host.is_windows and host.is_x86_64:
-        published_choice = published_asset_choice_for_kind(release, "windows-cpu")
+        # AMD Windows hosts should prefer a hash-approved published
+        # Windows HIP bundle when one exists, but otherwise fall through
+        # to resolve_asset_choice() so the upstream HIP prebuilt is
+        # tried before the CPU fallback. Hard-pinning the published
+        # windows-cpu bundle here would make the new HIP path
+        # unreachable.
+        if host.has_rocm:
+            published_choice = published_asset_choice_for_kind(release, "windows-hip")
+        else:
+            published_choice = published_asset_choice_for_kind(release, "windows-cpu")
     elif host.is_macos and host.is_arm64:
         published_choice = published_asset_choice_for_kind(release, "macos-arm64")
     elif host.is_macos and host.is_x86_64:
@@ -3378,7 +3630,7 @@ def overlay_directory_for_choice(
 
 
 def runtime_patterns_for_choice(choice: AssetChoice) -> list[str]:
-    if choice.install_kind in {"linux-cpu", "linux-cuda"}:
+    if choice.install_kind in {"linux-cpu", "linux-cuda", "linux-rocm"}:
         return [
             "llama-server",
             "llama-quantize",
@@ -3388,11 +3640,12 @@ def runtime_patterns_for_choice(choice: AssetChoice) -> list[str]:
             "libmtmd.so*",
             "libggml-cpu-*.so*",
             "libggml-cuda.so*",
+            "libggml-hip.so*",
             "libggml-rpc.so*",
         ]
     if choice.install_kind in {"macos-arm64", "macos-x64"}:
         return ["llama-server", "llama-quantize", "lib*.dylib"]
-    if choice.install_kind in {"windows-cpu", "windows-cuda"}:
+    if choice.install_kind in {"windows-cpu", "windows-cuda", "windows-hip"}:
         return ["*.exe", "*.dll"]
     raise PrebuiltFallback(
         f"unsupported install kind for runtime overlay: {choice.install_kind}"
@@ -4117,6 +4370,7 @@ def validate_server(
     install_dir: Path,
     *,
     runtime_line: str | None = None,
+    install_kind: str | None = None,
 ) -> None:
     last_failure: PrebuiltFallback | None = None
     for port_attempt in range(1, SERVER_PORT_BIND_ATTEMPTS + 1):
@@ -4140,7 +4394,33 @@ def validate_server(
             "--batch-size",
             "32",
         ]
-        if host.has_usable_nvidia or (host.is_macos and host.is_arm64):
+        # Only enable GPU offload for assets that actually ship GPU code.
+        # Gating on `host.has_rocm` alone breaks the intentional CPU
+        # fallback on AMD Windows hosts without a HIP prebuilt: the CPU
+        # binary would be launched with `--n-gpu-layers 1` and fail
+        # validation. Use the resolved install_kind as the source of
+        # truth and fall back to host detection when the caller did not
+        # pass one (keeps backwards compatibility with older call sites).
+        _gpu_kinds = {
+            "linux-cuda",
+            "linux-rocm",
+            "windows-cuda",
+            "windows-hip",
+            "macos-arm64",
+        }
+        if install_kind is not None:
+            _enable_gpu_layers = install_kind in _gpu_kinds
+        else:
+            # Older call sites that don't pass install_kind: keep ROCm
+            # hosts in the GPU-validation path so an AMD-only Linux host
+            # is exercised against the actual hardware rather than the
+            # CPU fallback. NVIDIA and macOS-arm64 are already covered.
+            _enable_gpu_layers = (
+                host.has_usable_nvidia
+                or host.has_rocm
+                or (host.is_macos and host.is_arm64)
+            )
+        if _enable_gpu_layers:
             command.extend(["--n-gpu-layers", "1"])
 
         log_fd, log_name = tempfile.mkstemp(prefix = "llama-server-", suffix = ".log")
@@ -4664,10 +4944,21 @@ def runtime_payload_health_groups(choice: AssetChoice) -> list[list[str]]:
             ["libggml*.dylib"],
             ["libmtmd*.dylib"],
         ]
+    if choice.install_kind == "linux-rocm":
+        return [
+            ["libllama.so*"],
+            ["libggml.so*"],
+            ["libggml-base.so*"],
+            ["libggml-cpu-*.so*"],
+            ["libmtmd.so*"],
+            ["libggml-hip.so*"],
+        ]
     if choice.install_kind == "windows-cpu":
         return [["llama.dll"]]
     if choice.install_kind == "windows-cuda":
         return [["llama.dll"], ["ggml-cuda.dll"]]
+    if choice.install_kind == "windows-hip":
+        return [["llama.dll"], ["*hip*.dll"]]
     return []
 
 
@@ -4839,6 +5130,7 @@ def validate_prebuilt_choice(
         host,
         install_dir,
         runtime_line = choice.runtime_line,
+        install_kind = choice.install_kind,
     )
     log(f"staged prebuilt validation succeeded for {choice.name}")
     return server_path, quantize_path
diff --git a/studio/install_python_stack.py b/studio/install_python_stack.py
index f2981ea665..af77d7a811 100644
--- a/studio/install_python_stack.py
+++ b/studio/install_python_stack.py
@@ -25,6 +25,281 @@ IS_WINDOWS = sys.platform == "win32"
 IS_MACOS = sys.platform == "darwin"
 IS_MAC_INTEL = IS_MACOS and platform.machine() == "x86_64"
 
+# ── ROCm / AMD GPU support ─────────────────────────────────────────────────────
+# Mapping from detected ROCm (major, minor) to the best PyTorch wheel tag on
+# download.pytorch.org.  Entries are checked newest-first (>=).
+# ROCm 7.2 only has torch 2.11.0 on download.pytorch.org, which exceeds the
+# current torch upper bound (<2.11.0).  Fall back to rocm7.1 (torch 2.10.0).
+# TODO: uncomment rocm7.2 when torch upper bound is bumped to >=2.11.0
+_ROCM_TORCH_INDEX: dict[tuple[int, int], str] = {
+    # (7, 2): "rocm7.2",  # torch 2.11.0 -- requires torch>=2.11
+    (7, 1): "rocm7.1",
+    (7, 0): "rocm7.0",
+    (6, 4): "rocm6.4",
+    (6, 3): "rocm6.3",
+    (6, 2): "rocm6.2",
+    (6, 1): "rocm6.1",
+    (6, 0): "rocm6.0",
+}
+_PYTORCH_WHL_BASE = "https://download.pytorch.org/whl"
+
+
+def _detect_rocm_version() -> tuple[int, int] | None:
+    """Return (major, minor) of the installed ROCm stack, or None."""
+    # Check /opt/rocm/.info/version or ROCM_PATH equivalent
+    rocm_root = os.environ.get("ROCM_PATH") or "/opt/rocm"
+    for path in (
+        os.path.join(rocm_root, ".info", "version"),
+        os.path.join(rocm_root, "lib", "rocm_version"),
+    ):
+        try:
+            with open(path) as fh:
+                parts = fh.read().strip().split("-")[0].split(".")
+            # Explicit length guard avoids relying on the broad except
+            # below to swallow IndexError when the version file contains
+            # a single component (e.g. "6\n" on a partial install).
+            if len(parts) >= 2:
+                return int(parts[0]), int(parts[1])
+        except Exception:
+            pass
+
+    # Try amd-smi version (outputs "... | ROCm version: X.Y.Z")
+    amd_smi = shutil.which("amd-smi")
+    if amd_smi:
+        try:
+            result = subprocess.run(
+                [amd_smi, "version"],
+                stdout = subprocess.PIPE,
+                stderr = subprocess.DEVNULL,
+                text = True,
+                timeout = 5,
+            )
+            if result.returncode == 0:
+                import re
+
+                m = re.search(r"ROCm version:\s*(\d+)\.(\d+)", result.stdout)
+                if m:
+                    return int(m.group(1)), int(m.group(2))
+        except Exception:
+            pass
+
+    # Try hipconfig --version (outputs bare version like "6.3.21234.2")
+    hipconfig = shutil.which("hipconfig")
+    if hipconfig:
+        try:
+            result = subprocess.run(
+                [hipconfig, "--version"],
+                stdout = subprocess.PIPE,
+                stderr = subprocess.DEVNULL,
+                timeout = 5,
+            )
+            if result.returncode == 0:
+                raw = result.stdout.decode().strip().split("\n")[0]
+                parts = raw.split(".")
+                if (
+                    len(parts) >= 2
+                    and parts[0].isdigit()
+                    and parts[1].split("-")[0].isdigit()
+                ):
+                    return int(parts[0]), int(parts[1].split("-")[0])
+        except Exception:
+            pass
+
+    # Distro package-manager fallbacks. Package-managed ROCm installs can
+    # expose GPUs via rocminfo / amd-smi but still lack /opt/rocm/.info/version
+    # and hipconfig, so probe dpkg (Debian/Ubuntu) and rpm (RHEL/Fedora/SUSE)
+    # for the rocm-core package version. Matches the chain in
+    # install.sh::get_torch_index_url so `unsloth studio update` behaves
+    # the same as a fresh `curl | sh` install.
+    import re as _re_pkg
+
+    for cmd in (
+        ["dpkg-query", "-W", "-f=${Version}\n", "rocm-core"],
+        ["rpm", "-q", "--qf", "%{VERSION}\n", "rocm-core"],
+    ):
+        exe = shutil.which(cmd[0])
+        if not exe:
+            continue
+        try:
+            result = subprocess.run(
+                [exe, *cmd[1:]],
+                stdout = subprocess.PIPE,
+                stderr = subprocess.DEVNULL,
+                text = True,
+                timeout = 5,
+            )
+        except Exception:
+            continue
+        if result.returncode != 0 or not result.stdout.strip():
+            continue
+        raw = result.stdout.strip()
+        # dpkg can prepend an epoch ("1:6.3.0-1"); strip it before parsing.
+        raw = _re_pkg.sub(r"^\d+:", "", raw)
+        m = _re_pkg.match(r"(\d+)[.-](\d+)", raw)
+        if m:
+            return int(m.group(1)), int(m.group(2))
+
+    return None
+
+
+def _has_rocm_gpu() -> bool:
+    """Return True only if an actual AMD GPU is visible (not just ROCm tools installed)."""
+    import re
+
+    for cmd, check_fn in (
+        # rocminfo: look for "Name: gfxNNNN" with nonzero first digit (gfx000 is the CPU agent)
+        (["rocminfo"], lambda out: bool(re.search(r"gfx[1-9]", out.lower()))),
+        # amd-smi list: require "GPU: <number>" data rows, not just a header
+        (
+            ["amd-smi", "list"],
+            lambda out: bool(re.search(r"(?im)^gpu\s*[:\[]\s*\d", out)),
+        ),
+    ):
+        exe = shutil.which(cmd[0])
+        if not exe:
+            continue
+        try:
+            result = subprocess.run(
+                [exe, *cmd[1:]],
+                stdout = subprocess.PIPE,
+                stderr = subprocess.DEVNULL,
+                text = True,
+                timeout = 10,
+            )
+        except Exception:
+            continue
+        if result.returncode == 0 and result.stdout.strip():
+            if check_fn(result.stdout):
+                return True
+    return False
+
+
+def _has_usable_nvidia_gpu() -> bool:
+    """Return True only when nvidia-smi exists AND reports at least one GPU."""
+    exe = shutil.which("nvidia-smi")
+    if not exe:
+        return False
+    try:
+        result = subprocess.run(
+            [exe, "-L"],
+            stdout = subprocess.PIPE,
+            stderr = subprocess.DEVNULL,
+            text = True,
+            timeout = 10,
+        )
+    except Exception:
+        return False
+    return result.returncode == 0 and "GPU " in result.stdout
+
+
+def _ensure_rocm_torch() -> None:
+    """Reinstall torch with ROCm wheels when the venv received CPU-only torch.
+
+    Runs only on Linux x86_64 hosts where an AMD GPU is present and the
+    ROCm runtime is detectable (rocminfo / amd-smi / hipconfig /
+    rocm-core package).  No-op when torch already links against HIP
+    (ROCm), on Windows / macOS, on non-x86_64 Linux (PyTorch does not
+    publish ROCm wheels for aarch64 / arm64), or on mixed AMD+NVIDIA
+    hosts (NVIDIA takes precedence).
+    Uses pip_install() to respect uv, constraints, and --python targeting.
+    """
+    # Explicit OS / architecture guards so the helper is safe to call
+    # from any context -- PyTorch only publishes ROCm wheels for
+    # linux_x86_64, so aarch64 / arm64 hosts must skip this repair path
+    # instead of failing the update with a missing-wheel error.
+    if IS_WINDOWS or IS_MACOS:
+        return
+    if platform.machine().lower() not in {"x86_64", "amd64"}:
+        return
+    # NVIDIA takes precedence on mixed hosts -- but only if an actual GPU is usable
+    if _has_usable_nvidia_gpu():
+        return
+    # Rely on _has_rocm_gpu() (rocminfo / amd-smi GPU data rows) as the
+    # authoritative "is this actually an AMD ROCm host?" signal. The old
+    # gate required /opt/rocm or hipcc to exist, which breaks on
+    # runtime-only ROCm installs (package-managed minimal installs,
+    # Radeon software) that ship amd-smi/rocminfo without /opt/rocm or
+    # hipcc, and leaves `unsloth studio update` unable to repair a
+    # CPU-only venv on those systems.
+    if not _has_rocm_gpu():
+        return  # no AMD GPU visible
+
+    ver = _detect_rocm_version()
+    if ver is None:
+        print("   ROCm detected but version unreadable -- skipping torch reinstall")
+        return
+
+    # Probe whether torch already links against HIP (ROCm is already working).
+    # Do NOT skip for CUDA-only builds since they are unusable on AMD-only
+    # hosts (the NVIDIA check above already handled mixed AMD+NVIDIA setups).
+    try:
+        probe = subprocess.run(
+            [
+                sys.executable,
+                "-c",
+                "import torch; print(getattr(torch.version,'hip','') or '')",
+            ],
+            stdout = subprocess.PIPE,
+            stderr = subprocess.DEVNULL,
+            timeout = 30,
+        )
+    except (OSError, subprocess.TimeoutExpired):
+        probe = None
+    has_hip_torch = (
+        probe is not None
+        and probe.returncode == 0
+        and probe.stdout.decode().strip() != ""
+    )
+
+    rocm_torch_ready = has_hip_torch
+
+    if not has_hip_torch:
+        # Select best matching wheel tag (newest ROCm version <= installed)
+        tag = next(
+            (
+                t
+                for (maj, mn), t in sorted(_ROCM_TORCH_INDEX.items(), reverse = True)
+                if ver >= (maj, mn)
+            ),
+            None,
+        )
+        if tag is None:
+            print(
+                f"   No PyTorch wheel for ROCm {ver[0]}.{ver[1]} -- "
+                f"skipping torch reinstall"
+            )
+        else:
+            index_url = f"{_PYTORCH_WHL_BASE}/{tag}"
+            print(f"   ROCm {ver[0]}.{ver[1]} -- installing torch from {index_url}")
+            pip_install(
+                f"ROCm torch ({tag})",
+                "--force-reinstall",
+                "--no-cache-dir",
+                "torch>=2.4,<2.11.0",
+                "torchvision<0.26.0",
+                "torchaudio<2.11.0",
+                "--index-url",
+                index_url,
+                constrain = False,
+            )
+            rocm_torch_ready = True
+
+    # Install bitsandbytes only when the venv has a ROCm-compatible torch
+    # (either already present or just installed). Avoids leaving an AMD
+    # bitsandbytes on top of a CPU/CUDA torch on hosts where the ROCm
+    # runtime is older than any published torch wheel. Uses
+    # --force-reinstall so an existing CPU/CUDA bitsandbytes is replaced
+    # by the AMD build during upgrades.
+    if rocm_torch_ready:
+        pip_install(
+            "bitsandbytes (AMD)",
+            "--force-reinstall",
+            "--no-cache-dir",
+            "--no-deps",
+            "bitsandbytes>=0.49.1",
+            constrain = False,
+        )
+
 
 def _infer_no_torch() -> bool:
     """Determine whether to run in no-torch (GGUF-only) mode.
@@ -414,6 +689,10 @@ def install_python_stack() -> int:
     base_total = 10 if IS_WINDOWS else 11
     if IS_MACOS:
         base_total -= 1  # triton step is skipped on macOS
+    # ROCm torch check steps (Linux only, non-macOS, non-no-torch):
+    # one early check (step 2b) and one final repair (step 13).
+    if not IS_WINDOWS and not IS_MACOS and not NO_TORCH:
+        base_total += 2
     _TOTAL = (base_total - 1) if skip_base else base_total
 
     # 1. Try to use uv for faster installs (must happen before pip upgrade
@@ -537,6 +816,53 @@ def install_python_stack() -> int:
             req = REQ_ROOT / "base.txt",
         )
 
+    # 2b. AMD ROCm: reinstall torch with HIP wheels if the host has ROCm but the
+    #     venv received CPU-only torch (common when pip resolves torch from PyPI).
+    #     Must come immediately after base packages so torch is present for inspection.
+    if not IS_WINDOWS and not IS_MACOS and not NO_TORCH:
+        _progress("ROCm torch check")
+        _ensure_rocm_torch()
+
+    # Windows + AMD GPU: PyTorch does not publish ROCm wheels for Windows.
+    # Detect and warn so users know manual steps are needed for GPU training.
+    if IS_WINDOWS and not NO_TORCH and not _has_usable_nvidia_gpu():
+        # Validate actual AMD GPU presence (not just tool existence)
+        import re as _re_win
+
+        def _win_amd_smi_has_gpu(stdout: str) -> bool:
+            return bool(_re_win.search(r"(?im)^gpu\s*[:\[]\s*\d", stdout))
+
+        _win_amd_gpu = False
+        for _wcmd, _check_fn in (
+            (["hipinfo"], lambda out: "gcnarchname" in out.lower()),
+            (["amd-smi", "list"], _win_amd_smi_has_gpu),
+        ):
+            _wexe = shutil.which(_wcmd[0])
+            if not _wexe:
+                continue
+            try:
+                _wr = subprocess.run(
+                    [_wexe, *_wcmd[1:]],
+                    stdout = subprocess.PIPE,
+                    stderr = subprocess.DEVNULL,
+                    text = True,
+                    timeout = 10,
+                )
+            except Exception:
+                continue
+            if _wr.returncode == 0 and _check_fn(_wr.stdout):
+                _win_amd_gpu = True
+                break
+        if _win_amd_gpu:
+            _safe_print(
+                _dim("  Note:"),
+                "AMD GPU detected on Windows. ROCm-enabled PyTorch must be",
+            )
+            _safe_print(
+                " " * 8,
+                "installed manually. See: https://docs.unsloth.ai/get-started/install-and-update/amd",
+            )
+
     # 3. Extra dependencies
     _progress("unsloth extras")
     pip_install(
@@ -650,7 +976,16 @@ def install_python_stack() -> int:
         [sys.executable, str(SINGLE_ENV / "patch_metadata.py")],
     )
 
-    # 13. Final check (silent; third-party conflicts are expected)
+    # 13. AMD ROCm: final torch repair.  Multiple install steps above can
+    #     pull in CUDA torch from PyPI (base packages, extras, overrides,
+    #     studio deps, etc.).  Running the repair as the very last step
+    #     ensures ROCm torch is in place at runtime, regardless of which
+    #     intermediate step clobbered it.
+    if not IS_WINDOWS and not IS_MACOS and not NO_TORCH:
+        _progress("ROCm torch (final)")
+        _ensure_rocm_torch()
+
+    # 14. Final check (silent; third-party conflicts are expected)
     subprocess.run(
         [sys.executable, "-m", "pip", "check"],
         stdout = subprocess.DEVNULL,
diff --git a/studio/setup.sh b/studio/setup.sh
index eb89e583ca..c2044b5141 100755
--- a/studio/setup.sh
+++ b/studio/setup.sh
@@ -928,6 +928,20 @@ else
                     _valid_gfx=""
                     for _gfx in $_gfx_list; do
                         if [[ "$_gfx" =~ ^gfx[0-9]{2,4}[a-z]?$ ]]; then
+                            # Drop bare family-level targets (gfx10, gfx11, gfx12, ...)
+                            # when a specific sibling is present in the same list.
+                            # rocminfo on ROCm 6.1+ emits both the specific GPU and
+                            # the LLVM generic family line (e.g. gfx1100 alongside
+                            # gfx11-generic), and the outer grep above captures the
+                            # bare family prefix from the generic line. Passing that
+                            # bare prefix to -DGPU_TARGETS breaks the HIP/llama.cpp
+                            # build because clang only accepts specific gfxNNN ids.
+                            # No real AMD GPU has a 2-digit gfx id, so this filter
+                            # can only ever drop family prefixes, never real targets.
+                            if [[ "$_gfx" =~ ^gfx[0-9]{2}$ ]] \
+                               && echo "$_gfx_list" | grep -qE "^${_gfx}[0-9][0-9a-z]?$"; then
+                                continue
+                            fi
                             _valid_gfx="${_valid_gfx}${_valid_gfx:+;}$_gfx"
                         fi
                     done
diff --git a/unsloth/kernels/utils.py b/unsloth/kernels/utils.py
index 90f2d5d238..09b03a597b 100644
--- a/unsloth/kernels/utils.py
+++ b/unsloth/kernels/utils.py
@@ -88,10 +88,26 @@ def is_cdna():
 
 @functools.lru_cache(1)
 def is_rdna():
-    """Detect ROCm-supported RDNA consumer/workstation GPUs (RDNA3, RDNA4)."""
+    """Detect ROCm-supported RDNA consumer/workstation GPUs (RDNA2, RDNA3, RDNA3.5, RDNA4)."""
     return is_hip() and triton.runtime.driver.active.get_current_target().arch in (
+        # RDNA2 (Navi 21-24)
+        "gfx1030",
+        "gfx1031",
+        "gfx1032",
+        "gfx1033",
+        "gfx1034",
+        "gfx1035",
+        "gfx1036",
+        # RDNA3 (Navi 31-33)
         "gfx1100",
         "gfx1101",
+        "gfx1102",
+        "gfx1103",
+        # RDNA3.5 (Strix Point / Strix Halo)
+        "gfx1150",
+        "gfx1151",
+        "gfx1152",
+        # RDNA4 (Navi 48-44)
         "gfx1200",
         "gfx1201",
     )
diff --git a/unsloth/tokenizer_utils.py b/unsloth/tokenizer_utils.py
index 2145fa5da4..a3f8affccf 100644
--- a/unsloth/tokenizer_utils.py
+++ b/unsloth/tokenizer_utils.py
@@ -1108,7 +1108,16 @@ def patch_sft_trainer_tokenizer():
             "    a = np.array([int(x.decode('utf-8'))/1024 for x in a])\n"
             "except:\n"
             "    if not torch.cuda.is_available():\n"
-            "        raise RuntimeError('Unsloth: We do not support AMD / Intel machines yet - it is a work in progress!')\n"
+            "        raise RuntimeError('Unsloth: No GPU detected. AMD ROCm users: install ROCm-enabled PyTorch -- see https://docs.unsloth.ai/get-started/install-and-update/amd')\n"
+            "    # nvidia-smi unavailable but torch.cuda IS available -- we are on\n"
+            "    # a ROCm host (ROCm reuses the torch.cuda.* API surface, so\n"
+            "    # device_count() is authoritative) or on a CUDA host without\n"
+            "    # the CLI installed. Use the device count directly as a\n"
+            "    # conservative multi-GPU signal: any configuration with more\n"
+            "    # than one visible device is flagged as unsupported, matching\n"
+            "    # the spirit of the per-device memory check used on CUDA.\n"
+            "    if torch.cuda.device_count() > 1:\n"
+            "        raise RuntimeError('Unsloth currently does not support multi GPU setups - but we are working on it!')\n"
             "if ((a - PRE_CHECK) >= 1).sum() > 1:\n"
             "    raise RuntimeError('Unsloth currently does not support multi GPU setups - but we are working on it!')\n"
             "for _ in range(3):\n"

From 65b40285608ee14e93438949e8f3541580aee234 Mon Sep 17 00:00:00 2001
From: Daniel Han <danielhanchen@gmail.com>
Date: Fri, 10 Apr 2026 06:25:39 -0700
Subject: [PATCH 070/216] Pin bitsandbytes to continuous-release_main on ROCm
 (4-bit decode fix) (#4954)
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

* Pin bitsandbytes to continuous-release_main on ROCm for 4-bit decode fix

bitsandbytes 0.49.2 on PyPI ships with a broken 4-bit GEMV kernel on
every ROCm target:

  - CDNA (gfx90a / gfx942 / gfx950 = MI210 / MI300X / MI350) via a
    broken blocksize=32/64 warp64 GEMV kernel whose tests were
    explicitly skipped with ROCM_WARP_SIZE_64 guards because the
    code was known broken.
  - RDNA3 / RDNA3.5 (gfx1100-1103 / gfx1150-1152) via a compile-time
    BNB_WARP_SIZE macro in the host-side dispatch that resolves to
    64 when the multi-arch wheel is compiled with CDNA as the
    primary target, so num_blocks is wrong on RDNA and half the GEMV
    output is never written.

At decode shape (1, 1, hidden) both bugs produce NaN. Training is
unaffected because training shapes are (batch, seq_len > 1, hidden)
and never touch the GEMV path. The crash during autoregressive
inference surfaces as _assert_async_cuda_kernel in torch.multinomial
which on HIP becomes a hard HSA_STATUS_ERROR_EXCEPTION instead of
a clean Python error.

Both bugs are fixed by bitsandbytes commit 713a3b8 ("[ROCm] Enable
blocksize 32 4-bit quantization and GEMV kernels on AMD CDNA",
PR #1887, merged 2026-03-09) which replaces BNB_WARP_SIZE with a
runtime hipDeviceGetAttribute query and ships a working CDNA warp64
kernel. That commit has not shipped to PyPI yet, but
continuous-release_main wheels are published on every push to bnb
main via GitHub Releases.

Point the ROCm install path at the continuous-release_main x86_64 and
aarch64 wheels and fall back to PyPI >=0.49.1 when the pre-release is
unreachable (offline installs, firewalled hosts, or architectures not
covered by the pre-release wheels). Drop the pin once bnb cuts a
0.50+ tag on PyPI.

Verified on MI300X (gfx942, ROCm 7.2, torch 2.10.0+rocm7.1): direct
bnb GEMV shape test now returns 0.0078 max abs error at seq_len=1
(no NaN) vs NaN on 0.49.2, and full Unsloth + for_inference + 4-bit
sampling generation works end-to-end.

NVIDIA / CPU / Mac / Windows paths are unaffected -- the helper is
gated on the ROCm torch index and platform.machine() respectively.

* Drop Studio ROCm 16-bit fallback now that bnb 0.50+ fixes 4-bit decode

The 16-bit fallback in studio/backend/core/inference/inference.py was
added as a workaround for a bug that this PR already fixes at the
install layer: bitsandbytes <= 0.49.2 has a broken 4-bit GEMV kernel
on every ROCm target, which NaNs at decode shape (seq_len=1) and
crashes autoregressive inference. bnb PR #1887 (commit 713a3b8, in
0.50.0.dev0+, pinned by install.sh / install_python_stack.py in this
PR) restores correct 4-bit decode on MI300X and verified working
end-to-end with full Unsloth + for_inference + sampling.

Revert the dual code path so ROCm and NVIDIA both go through the
normal FastLanguageModel.from_pretrained + for_inference flow:

  - Remove the conditional `from unsloth import` that skipped the
    import on ROCm. The monkey-patches it was trying to avoid were
    never the cause of the crash; bnb 4-bit GEMV was.
  - Remove the `if _hw_module.IS_ROCM:` branch in load_model that
    loaded with plain transformers + PEFT + bfloat16, and the
    `_resolve_fp16_base` helper it relied on.
  - Remove the `get_chat_template is not None` fallback in
    _load_chat_template_info -- get_chat_template is now always
    imported.
  - Refactor the audio/vision ROCm guard to check _hw_module.IS_ROCM
    directly instead of the removed _IS_ROCM_ENV global. Audio and
    vision on ROCm still need separate validation (FastVisionModel
    and the CSM audio codecs were never tested on HIP) so the guard
    stays for now.

Add _bnb_rocm_4bit_ok() as a runtime safety net for users who
install from this PR before the install.sh bnb pin kicks in, or
whose installer fell back to the PyPI pin because the continuous-
release wheel was unreachable. When the installed bnb is < 0.50 on
ROCm, force load_in_4bit=False and strip any -unsloth-bnb-4bit /
-bnb-4bit suffix from the model path so a pre-quantized repo
resolves to its FP16 sibling instead of pulling bnb back in via
the repo's quantization_config. LoRA adapters whose base is a
pre-quantized repo on old bnb will still fail inside Unsloth's
loader -- the only real fix there is `unsloth studio update`.

Verified on MI300X (gfx942, ROCm 7.2, torch 2.10.0+rocm7.1):

  - HAPPY path (bnb 0.50.0.dev0, load_in_4bit=True, pre-quantized
    repo): loads in 4-bit via the fixed GEMV, generation returns
    "Paris." for greedy and sampling.
  - SAFETY-NET path (simulated old bnb, suffix-stripped to the
    FP16 sibling, load_in_4bit=False): loads in bf16, generation
    returns "Paris." for greedy and sampling.

Net diff is ~45 lines smaller than the pre-revert state because
the entire plain-transformers 16-bit branch is gone.

* Cache _bnb_rocm_4bit_ok() with functools.cache

load_model() can be called many times in a single session but the bnb
version and hardware state cannot change at runtime, so memoise the
check. First call is ~1.9 ms (dominated by the lazy `import bitsandbytes`
inside the try block), subsequent calls drop to sub-microsecond dict
lookups. Zero behavioral change.

* Shorten verbose bnb/ROCm comments

Comment-only cleanup across install.sh, studio/install_python_stack.py,
and studio/backend/core/inference/inference.py. No behavioral change.

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

* Remove _bnb_rocm_4bit_ok safety net from inference.py

Studio's ROCm support is brand new (PR #4720, merged today) and every
fresh install pulls the bnb continuous-release_main wheel via
install.sh / install_python_stack.py in this same PR. There are no
existing ROCm Studio installs carrying bnb < 0.50, so the defensive
version-check fallback is guarding against a scenario that cannot
actually occur. Delete the helper, the functools import, and the
safety-net block -- inference.py now calls FastLanguageModel.from_pretrained
directly with no ROCm branching.

* Drop audio/vision ROCm guard in inference.py — verified unblocked by bnb fix

Vision inference was blocked by the same bnb 4-bit GEMV bug that affected
text inference (vision models use bnb 4-bit for the LM backbone). With
bnb 0.50+ pinned in install.sh / install_python_stack.py, vision works
end-to-end on MI300X: Llama-3.2-11B-Vision-Instruct-unsloth-bnb-4bit
loaded in 4-bit via FastVisionModel + for_inference returns a correct
answer to a multimodal prompt.

Audio (CSM) was never actually blocked by HIP — on this hardware CSM
loads and runs its backbone forward pass fine with bnb 0.50, then fails
during generate() with a transformers-level kwarg validation mismatch
in generation_csm.py (`backbone_last_hidden_state` rejected). That's a
pre-existing transformers/CSM integration bug that reproduces identically
on NVIDIA, so the ROCm-gated guard was never actually protecting users
from anything HIP-specific.

Remove the combined audio/vision guard and the now-unused _hw_module
import. Also restore the one-word "Can be" in an inline comment that
drifted during the earlier comment-shortening pass, so the inference.py
delta vs pre-#4720 is exactly the max_seq_length<=0 crash fix and
nothing else.

* Shorten max_seq_length=0 guard comment to one line

---------

Co-authored-by: Daniel Han <danielhanchen@users.noreply.github.com>
Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com>
---
 install.sh                                 |  36 +++++-
 studio/backend/core/inference/inference.py | 132 +++------------------
 studio/install_python_stack.py             | 101 +++++++++++++---
 3 files changed, 137 insertions(+), 132 deletions(-)

diff --git a/install.sh b/install.sh
index 2e347a8bc5..cc8721710e 100755
--- a/install.sh
+++ b/install.sh
@@ -94,6 +94,36 @@ run_install_cmd() {
     return $_rc
 }
 
+# Install bitsandbytes on AMD ROCm hosts. Uses the continuous-release_main
+# wheel for the ROCm 4-bit GEMV fix (bnb PR #1887, post-0.49.2); bnb <= 0.49.2
+# NaNs at decode shape on every AMD GPU. Falls back to PyPI >=0.49.1 if the
+# pre-release URL is unreachable. Drop the pin once bnb 0.50+ ships on PyPI.
+_install_bnb_rocm() {
+    _label="$1"
+    _venv_py="$2"
+    case "$_ARCH" in
+        x86_64|amd64)
+            _bnb_whl_url="https://github.com/bitsandbytes-foundation/bitsandbytes/releases/download/continuous-release_main/bitsandbytes-1.33.7.preview-py3-none-manylinux_2_24_x86_64.whl"
+            ;;
+        aarch64|arm64)
+            _bnb_whl_url="https://github.com/bitsandbytes-foundation/bitsandbytes/releases/download/continuous-release_main/bitsandbytes-1.33.7.preview-py3-none-manylinux_2_24_aarch64.whl"
+            ;;
+        *)
+            _bnb_whl_url=""
+            ;;
+    esac
+    if [ -n "$_bnb_whl_url" ]; then
+        substep "installing bitsandbytes for AMD ROCm (pre-release, PR #1887)..."
+        if run_install_cmd "$_label (pre-release)" uv pip install --python "$_venv_py" \
+            --force-reinstall --no-cache-dir --no-deps "$_bnb_whl_url"; then
+            return 0
+        fi
+        substep "[WARN] bnb pre-release unreachable; falling back to PyPI (4-bit decode broken on ROCm)" "$C_WARN"
+    fi
+    run_install_cmd "$_label (pypi fallback)" uv pip install --python "$_venv_py" \
+        --force-reinstall --no-cache-dir --no-deps "bitsandbytes>=0.49.1"
+}
+
 if [ "$_next_is_package" = true ]; then
     echo "❌ ERROR: --package requires an argument." >&2
     exit 1
@@ -1296,8 +1326,7 @@ if [ "$_MIGRATED" = true ]; then
     if [ "$SKIP_TORCH" = false ]; then
         case "$TORCH_INDEX_URL" in
             */rocm*)
-                substep "installing bitsandbytes for AMD ROCm..."
-                run_install_cmd "install bitsandbytes (AMD)" uv pip install --python "$_VENV_PY" --force-reinstall --no-cache-dir --no-deps "bitsandbytes>=0.49.1"
+                _install_bnb_rocm "install bitsandbytes (AMD)" "$_VENV_PY"
                 # Repair ROCm torch if overwritten during migrated install
                 _has_hip=$("$_VENV_PY" -c "import torch; print(getattr(torch.version,'hip','') or '')" 2>/dev/null || true)
                 if [ -z "$_has_hip" ]; then
@@ -1437,8 +1466,7 @@ elif [ -n "$TORCH_INDEX_URL" ]; then
     if [ "$SKIP_TORCH" = false ]; then
         case "$TORCH_INDEX_URL" in
             */rocm*)
-                substep "installing bitsandbytes for AMD ROCm..."
-                run_install_cmd "install bitsandbytes (AMD)" uv pip install --python "$_VENV_PY" --force-reinstall --no-cache-dir --no-deps "bitsandbytes>=0.49.1"
+                _install_bnb_rocm "install bitsandbytes (AMD)" "$_VENV_PY"
                 ;;
         esac
     fi
diff --git a/studio/backend/core/inference/inference.py b/studio/backend/core/inference/inference.py
index d7bc3ad4c7..4c140013a0 100644
--- a/studio/backend/core/inference/inference.py
+++ b/studio/backend/core/inference/inference.py
@@ -5,23 +5,8 @@
 Core inference backend - streamlined
 """
 
-# On AMD ROCm, Unsloth's global monkey-patching of transformers model classes
-# (LlamaRotaryEmbedding, attention modules, etc.) causes HIP kernel crashes
-# (_assert_async_cuda_kernel -> HSA_STATUS_ERROR_EXCEPTION) during inference.
-# Training works because it uses different code paths, but generation triggers
-# the incompatible patched kernels.  Skip the Unsloth import entirely on ROCm
-# so transformers classes stay unmodified; the GGUF inference path (llama-server)
-# is unaffected since it never imports these Python model classes.
-_IS_ROCM_ENV = getattr(__import__("torch").version, "hip", None) is not None
-
-if _IS_ROCM_ENV:
-    FastLanguageModel = None  # Loaded on-demand only on NVIDIA
-    FastVisionModel = None
-    get_chat_template = None
-else:
-    from unsloth import FastLanguageModel, FastVisionModel
-    from unsloth.chat_templates import get_chat_template
-
+from unsloth import FastLanguageModel, FastVisionModel
+from unsloth.chat_templates import get_chat_template
 from transformers import TextStreamer
 from peft import PeftModel, PeftModelForCausalLM
 
@@ -41,7 +26,6 @@ from utils.hardware import (
     raise_if_offloaded,
     get_visible_gpu_count,
 )
-from utils.hardware import hardware as _hw_module
 from core.inference.audio_codecs import AudioCodecManager
 from io import StringIO
 import structlog
@@ -269,12 +253,7 @@ class InferenceBackend:
         """
         Load any model: base, LoRA adapter, text, or vision.
         """
-        # max_seq_length=0 means "model default" for the GGUF/llama.cpp path,
-        # but Unsloth's FastLanguageModel.from_pretrained treats 0 literally --
-        # setting the model's context to 0 tokens, which triggers an assertion
-        # crash during generation (especially on ROCm/HIP where the async
-        # assert kernel raises a hardware exception instead of a Python error).
-        # Fall back to 2048 for the Unsloth/transformers path.
+        # GGUF uses max_seq_length=0 as "model default"; Unsloth crashes on it.
         if max_seq_length <= 0:
             max_seq_length = 2048
 
@@ -311,12 +290,6 @@ class InferenceBackend:
             }
 
             # ── Audio model loading path ──────────────────────────
-            if (config.is_audio or config.is_vision) and _IS_ROCM_ENV:
-                raise RuntimeError(
-                    f"Audio and vision model inference via Unsloth is not "
-                    f"yet supported on AMD ROCm.  Use GGUF inference instead."
-                )
-
             if config.is_audio:
                 audio_type = config.audio_type
                 adapter_info = " (LoRA adapter)" if config.is_lora else ""
@@ -547,84 +520,18 @@ class InferenceBackend:
 
             else:
                 # Text model (or text LoRA adapter)
-                if _hw_module.IS_ROCM:
-                    # On AMD ROCm two issues prevent the normal Unsloth path:
-                    #  1. Unsloth's patched kernels (RoPE, attention) crash on
-                    #     HIP (_assert_async_cuda_kernel -> HSA_STATUS_ERROR).
-                    #  2. bitsandbytes 4-bit matmul kernels trigger the same
-                    #     HIP assertion on MI300X (CDNA3 / gfx942).
-                    # Fall back to plain transformers + PEFT in 16-bit, which
-                    # works reliably.  AMD GPUs typically have large VRAM so
-                    # 16-bit is practical; GGUF inference remains the
-                    # recommended path for memory-constrained setups.
-                    logger.info(
-                        "ROCm detected -- loading in 16-bit with plain "
-                        "transformers (bitsandbytes 4-bit and Unsloth kernels "
-                        "are not yet compatible with HIP)"
-                    )
-                    from transformers import AutoModelForCausalLM, AutoTokenizer
+                model, tokenizer = FastLanguageModel.from_pretrained(
+                    model_name = config.path,  # Can be base model OR LoRA adapter path
+                    max_seq_length = max_seq_length,
+                    dtype = dtype,
+                    load_in_4bit = load_in_4bit,
+                    device_map = device_map,
+                    token = hf_token if hf_token and hf_token.strip() else None,
+                    trust_remote_code = trust_remote_code,
+                )
 
-                    _load_kwargs = dict(
-                        dtype = dtype or torch.bfloat16,
-                        device_map = device_map,
-                        token = hf_token if hf_token and hf_token.strip() else None,
-                        trust_remote_code = trust_remote_code,
-                    )
-
-                    # Skip 4-bit on ROCm: bnb matmul kernels crash on HIP.
-                    # Also resolve pre-quantized Unsloth model names (e.g.
-                    # "unsloth/xxx-bnb-4bit") to their FP16 originals since
-                    # loading a pre-quantized repo still triggers bnb codepaths.
-                    def _resolve_fp16_base(name: str) -> str:
-                        if not name:
-                            return name
-                        # Strip Unsloth quantization suffixes to get the FP16 model:
-                        # "unsloth/Foo-unsloth-bnb-4bit" -> "unsloth/Foo"
-                        # "unsloth/Foo-bnb-4bit"         -> "unsloth/Foo"
-                        # Order matters: try longer suffix first.
-                        for suffix in ("-unsloth-bnb-4bit", "-bnb-4bit"):
-                            if name.lower().endswith(suffix):
-                                resolved = name[: -len(suffix)]
-                                logger.info(
-                                    "Resolved pre-quantized base '%s' -> '%s' for ROCm 16-bit inference",
-                                    name,
-                                    resolved,
-                                )
-                                return resolved
-                        return name
-
-                    if config.is_lora and config.base_model:
-                        # Load base model then apply adapter
-                        _base = _resolve_fp16_base(config.base_model)
-                        model = AutoModelForCausalLM.from_pretrained(
-                            _base,
-                            **_load_kwargs,
-                        )
-                        from peft import PeftModel
-
-                        model = PeftModel.from_pretrained(model, config.path)
-                        tokenizer = AutoTokenizer.from_pretrained(config.path)
-                    else:
-                        _path = _resolve_fp16_base(config.path)
-                        model = AutoModelForCausalLM.from_pretrained(
-                            _path,
-                            **_load_kwargs,
-                        )
-                        tokenizer = AutoTokenizer.from_pretrained(config.path)
-                    model.eval()
-                else:
-                    model, tokenizer = FastLanguageModel.from_pretrained(
-                        model_name = config.path,  # Can be base model OR LoRA adapter path
-                        max_seq_length = max_seq_length,
-                        dtype = dtype,
-                        load_in_4bit = load_in_4bit,
-                        device_map = device_map,
-                        token = hf_token if hf_token and hf_token.strip() else None,
-                        trust_remote_code = trust_remote_code,
-                    )
-
-                    # Apply inference optimization
-                    FastLanguageModel.for_inference(model)
+                # Apply inference optimization
+                FastLanguageModel.for_inference(model)
 
                 self.models[model_name]["model"] = model
                 self.models[model_name]["tokenizer"] = tokenizer
@@ -1047,13 +954,10 @@ class InferenceBackend:
                 )
 
                 # This modifies the tokenizer with the correct template
-                if get_chat_template is not None:
-                    tokenizer = get_chat_template(
-                        tokenizer,
-                        chat_template = template_name,
-                    )
-                else:
-                    logger.info("Skipping Unsloth chat template (ROCm fallback)")
+                tokenizer = get_chat_template(
+                    tokenizer,
+                    chat_template = template_name,
+                )
             else:
                 logger.info(
                     f"No registered Unsloth template for {self.active_model_name}, using tokenizer default"
diff --git a/studio/install_python_stack.py b/studio/install_python_stack.py
index af77d7a811..a046f8f892 100644
--- a/studio/install_python_stack.py
+++ b/studio/install_python_stack.py
@@ -43,6 +43,32 @@ _ROCM_TORCH_INDEX: dict[tuple[int, int], str] = {
 }
 _PYTORCH_WHL_BASE = "https://download.pytorch.org/whl"
 
+# bitsandbytes continuous-release_main wheels with the ROCm 4-bit GEMV fix
+# (bnb PR #1887, post-0.49.2). bnb <= 0.49.2 NaNs at decode shape on every
+# AMD GPU. Drop the pin once bnb 0.50+ ships on PyPI.
+_BNB_ROCM_PRERELEASE_URLS: dict[str, str] = {
+    "x86_64": (
+        "https://github.com/bitsandbytes-foundation/bitsandbytes/releases/"
+        "download/continuous-release_main/"
+        "bitsandbytes-1.33.7.preview-py3-none-manylinux_2_24_x86_64.whl"
+    ),
+    "aarch64": (
+        "https://github.com/bitsandbytes-foundation/bitsandbytes/releases/"
+        "download/continuous-release_main/"
+        "bitsandbytes-1.33.7.preview-py3-none-manylinux_2_24_aarch64.whl"
+    ),
+}
+_BNB_ROCM_PYPI_FALLBACK = "bitsandbytes>=0.49.1"
+
+
+def _bnb_rocm_prerelease_url() -> str | None:
+    """Return the continuous-release_main bnb wheel URL for the current
+    architecture, or None when no pre-release wheel is available.
+    """
+    arch = platform.machine().lower()
+    arch = {"amd64": "x86_64", "arm64": "aarch64"}.get(arch, arch)
+    return _BNB_ROCM_PRERELEASE_URLS.get(arch)
+
 
 def _detect_rocm_version() -> tuple[int, int] | None:
     """Return (major, minor) of the installed ROCm stack, or None."""
@@ -284,21 +310,37 @@ def _ensure_rocm_torch() -> None:
             )
             rocm_torch_ready = True
 
-    # Install bitsandbytes only when the venv has a ROCm-compatible torch
-    # (either already present or just installed). Avoids leaving an AMD
-    # bitsandbytes on top of a CPU/CUDA torch on hosts where the ROCm
-    # runtime is older than any published torch wheel. Uses
-    # --force-reinstall so an existing CPU/CUDA bitsandbytes is replaced
-    # by the AMD build during upgrades.
+    # Install bitsandbytes only when torch links against ROCm. Prefers the
+    # continuous-release_main wheel (bnb PR #1887 4-bit GEMV fix) and falls
+    # back to PyPI when the pre-release URL is unreachable.
     if rocm_torch_ready:
-        pip_install(
-            "bitsandbytes (AMD)",
-            "--force-reinstall",
-            "--no-cache-dir",
-            "--no-deps",
-            "bitsandbytes>=0.49.1",
-            constrain = False,
-        )
+        _bnb_url = _bnb_rocm_prerelease_url()
+        _bnb_installed = False
+        if _bnb_url is not None:
+            _bnb_installed = pip_install_try(
+                "bitsandbytes (AMD, pre-release main)",
+                "--force-reinstall",
+                "--no-cache-dir",
+                "--no-deps",
+                _bnb_url,
+                constrain = False,
+            )
+            if not _bnb_installed:
+                print(
+                    _red(
+                        "   bnb pre-release unreachable; falling back to PyPI "
+                        "(4-bit decode will be broken on ROCm)"
+                    )
+                )
+        if not _bnb_installed:
+            pip_install(
+                "bitsandbytes (AMD)",
+                "--force-reinstall",
+                "--no-cache-dir",
+                "--no-deps",
+                _BNB_ROCM_PYPI_FALLBACK,
+                constrain = False,
+            )
 
 
 def _infer_no_torch() -> bool:
@@ -593,6 +635,37 @@ def _build_uv_cmd(args: tuple[str, ...]) -> list[str]:
     return cmd
 
 
+def pip_install_try(
+    label: str,
+    *args: str,
+    constrain: bool = True,
+) -> bool:
+    """Like pip_install but returns False on failure instead of exiting.
+    For optional installs with a follow-up fallback.
+    """
+    constraint_args: list[str] = []
+    if constrain and CONSTRAINTS.is_file():
+        constraint_args = ["-c", str(CONSTRAINTS)]
+
+    if USE_UV:
+        cmd = _build_uv_cmd(args) + constraint_args
+    else:
+        cmd = _build_pip_cmd(args) + constraint_args
+
+    if VERBOSE:
+        _step(_LABEL, f"{label}...", _dim)
+    result = subprocess.run(
+        cmd,
+        stdout = subprocess.PIPE,
+        stderr = subprocess.STDOUT,
+    )
+    if result.returncode == 0:
+        return True
+    if VERBOSE and result.stdout:
+        print(result.stdout.decode(errors = "replace"))
+    return False
+
+
 def pip_install(
     label: str,
     *args: str,

From 53af4a1b3e78f2d1cac9401fe15baf8720fc2303 Mon Sep 17 00:00:00 2001
From: Daniel Han <danielhanchen@gmail.com>
Date: Fri, 10 Apr 2026 07:58:15 -0700
Subject: [PATCH 071/216] Fix Gemma-4 GRPO catastrophic KL divergence with TRL
 1.0.0+ (#4934)

* Fix Gemma-4 GRPO catastrophic KL divergence with TRL 1.0.0+

Two compounding bugs caused Gemma-4 GRPO training to diverge with KL ~10^12
at step 1 against TRL 1.0.0+. Both fixes are runtime patches in the existing
TRL/model patch flow and are no-ops for models and TRL versions that are not
affected.

Fix 1 (rl.py): replace trl.models.utils.disable_gradient_checkpointing with
a no-op context manager. TRL 1.0.0+ wraps generation in
`with torch.no_grad(), disable_gradient_checkpointing(self.model, ...):`
purely to suppress a cosmetic PyTorch warning ("None of the inputs have
requires_grad=True"). Inside torch.no_grad() the gradient checkpointing
state has no functional effect on the forward pass. On context exit, TRL
calls model.gradient_checkpointing_enable() which dispatches to HF's
generic implementation and overwrites Unsloth's custom
`use_gradient_checkpointing="unsloth"` wrapper, corrupting Gemma-4 forward
numerics. Replacing the toggle with a no-op preserves Unsloth's custom GC
wrapper across generation passes. The patch walks sys.modules dynamically
to also rebind the symbol on every trl.* module that already imported it
(grpo_trainer, dpo_trainer, rloo_trainer, dppo_trainer, gfpo_trainer,
grpo_with_replay_buffer_trainer, and any future trainer module).

Fix 2 (vision.py): inject `final_logit_softcapping` from `config.text_config`
into the top-level `model.config` for multimodal models. Unsloth's GRPO
trainer reads `getattr(model.config, "final_logit_softcapping", 0)` but
for Gemma-4 the attribute lives only on the nested `Gemma4TextConfig`,
so the lookup silently defaults to 0 instead of 30.

Backwards compatibility:
- trl 0.22.2: no `disable_gradient_checkpointing` symbol exists, the patch
  early-returns via `hasattr` guard.
- trl 0.27.1: same broken pattern as 1.0.0, the noop replacement is correct.
- trl 1.0.0+: end-to-end verified on `unsloth/gemma-4-E2B-it` GRPO with TRL
  1.0.0 and transformers 5.5.0. Step 1 loss=2.46e-08, kl=2.92e-05 (machine
  zero) vs broken baseline loss=1.37e+06, kl=1.76e+09.
- Llama / non-VLM text models: Fix 2 is a no-op (no `text_config`); Fix 1
  is functionally identical (Unsloth's GC wrapper is preserved).
- Qwen3-VL and other VLMs without final_logit_softcapping: Fix 2 is a no-op
  (text_config.final_logit_softcapping is None).

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

* Apply loop 1 review fixes for PR #4934

- Move Fix 2 from vision.py to rl_replacements.py:858 and :1110 at the
  actual consumer sites. This avoids mutating model.config (which could
  leak into save_pretrained output) and covers text-only Gemma-4 paths
  that do not flow through FastBaseModel.from_pretrained.
- Revert the vision.py injection block entirely.
- Narrow the bare except blocks in patch_trl_disable_gradient_checkpointing
  from `except Exception:` to `(AttributeError, ImportError)` and
  `(AttributeError, TypeError)` to avoid masking unrelated bugs.
- Add logger.warning_once when the noop patch is installed, matching
  patch_trl_openenv and patch_trl_vllm_generation convention.
- Remove the dead per-module `_unsloth_noop_patched` sentinel check inside
  the sys.modules walk. The function-level early return already covers
  this case.
- Move `import sys` and `from contextlib import contextmanager` to the
  module-level imports instead of inside the function body.
- Rewrite the ordering comment in PatchFastRL to accurately describe
  why patch_trl_disable_gradient_checkpointing must run before
  patch_trl_rl_trainers.
- Fix keyword default spacing to match surrounding rl.py style.

End-to-end verified: Gemma-4-E2B GRPO on TRL 1.0.0 + transformers 5.5.0
step 1 loss=2.464e-08 kl=2.921e-05, all 5 steps succeed.

* Apply loop 2 review fix for PR #4934

Extract the final_logit_softcapping fallback logic into a shared helper
`_unsloth_get_final_logit_softcapping(config)` defined in rl_replacements.py
and injected into the compiled cache via RL_PRE_ITEMS["grpo_trainer"]. Both
call sites (`grpo_trainer__generate_and_score_completions` and
`grpo_trainer_compute_loss`) now use the helper instead of inlining the
same text_config fallback block twice.

Verified: compiled cache file lists the helper at module scope and both
consumer sites call it. Gemma-4-E2B GRPO step 1 loss=2.464e-08 kl=2.921e-05
(unchanged), all 5 steps pass.

* Apply loop 3 review fix for PR #4934

Extend _unsloth_get_final_logit_softcapping to also fall back to
config.get_text_config() for composite configs such as T5GemmaConfig
where the text sub-config is not exposed via the text_config attribute
but only via the get_text_config() method. Guard against (TypeError,
ValueError) raised by ambiguous composite configs, and skip the
self-referential case where get_text_config() returns self.

This addresses the 6/7 reviewer consensus from the third review loop.

Verified:
- Helper returns 30.0 for Gemma-4, T5Gemma, and Gemma 1/2 configs.
- Helper returns 0 for Llama, Qwen, Mistral, Cohere, Granite, and
  ambiguous configs raising ValueError.
- Gemma-4-E2B GRPO step 1 loss=2.464e-08 kl=2.921e-05 (unchanged).
- Llama-3.2-1B GRPO all 5 steps loss=0 kl=0 (no regression).

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

---------

Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com>
---
 unsloth/models/rl.py              | 87 +++++++++++++++++++++++++++++++
 unsloth/models/rl_replacements.py | 35 ++++++++++---
 2 files changed, 116 insertions(+), 6 deletions(-)

diff --git a/unsloth/models/rl.py b/unsloth/models/rl.py
index 5651a7da41..f444f6bd37 100755
--- a/unsloth/models/rl.py
+++ b/unsloth/models/rl.py
@@ -22,6 +22,8 @@ from typing import Any, Callable, Dict, List, Literal, Optional, Tuple, Union
 import inspect
 import os
 import re
+import sys
+from contextlib import contextmanager
 from unsloth_zoo.compiler import create_new_function
 from unsloth_zoo.log import logger
 from unsloth_zoo.logging_utils import PatchRLStatistics
@@ -1947,6 +1949,83 @@ def patch_trl_rl_trainers():
     return
 
 
+def patch_trl_disable_gradient_checkpointing():
+    # TRL 1.0.0+ wraps generation in:
+    #   with torch.no_grad(), disable_gradient_checkpointing(self.model, ...):
+    # The toggle exists only to suppress a cosmetic PyTorch warning
+    # ("None of the inputs have requires_grad=True"). Inside torch.no_grad()
+    # the gradient checkpointing state has no functional effect on the
+    # forward pass.
+    #
+    # On exit, the context manager calls model.gradient_checkpointing_enable()
+    # which dispatches to HuggingFace's generic implementation and overwrites
+    # Unsloth's custom `use_gradient_checkpointing="unsloth"` wrapper. For
+    # Gemma-4 (and likely other models) this corrupts the forward numerics
+    # enough to make GRPO KL divergence explode to ~10^12 at step 1.
+    #
+    # Replacing the context manager with a no-op preserves Unsloth's custom
+    # gradient checkpointing wrapper across generation/inference passes.
+    #
+    # Backwards compatibility:
+    #   - trl < 1.0.0 (no disable_gradient_checkpointing): early return.
+    #   - trl >= 1.0.0: noop is functionally equivalent for forward
+    #     correctness. The only loss is a cosmetic warning being emitted
+    #     by PyTorch when use_reentrant=True (which is exactly the warning
+    #     TRL added the toggle to suppress in the first place).
+    try:
+        import trl.models.utils as _tmu
+    except ImportError:
+        return
+    if not hasattr(_tmu, "disable_gradient_checkpointing"):
+        return
+    if getattr(
+        _tmu.disable_gradient_checkpointing,
+        "_unsloth_noop_patched",
+        False,
+    ):
+        return
+
+    @contextmanager
+    def _noop_disable_gradient_checkpointing(model, gradient_checkpointing_kwargs = None):
+        yield
+
+    _noop_disable_gradient_checkpointing._unsloth_noop_patched = True
+
+    _tmu.disable_gradient_checkpointing = _noop_disable_gradient_checkpointing
+
+    # Also rebind any trl.* module that already imported the symbol by
+    # reference, so the noop applies even when the trainer module cached the
+    # original at import time. We walk sys.modules dynamically rather than
+    # hardcoding a list, so this picks up every trainer that does
+    # `from ...models.utils import disable_gradient_checkpointing`
+    # (grpo, dpo, rloo, dppo, gfpo, grpo_with_replay_buffer, and any future
+    # TRL trainer module).
+    for _mod_name, _mod in list(sys.modules.items()):
+        if _mod is None or not _mod_name.startswith("trl."):
+            continue
+        try:
+            _bound = getattr(_mod, "disable_gradient_checkpointing", None)
+        except (AttributeError, ImportError):
+            continue
+        if _bound is None:
+            continue
+        try:
+            setattr(
+                _mod,
+                "disable_gradient_checkpointing",
+                _noop_disable_gradient_checkpointing,
+            )
+        except (AttributeError, TypeError):
+            pass
+
+    logger.warning_once(
+        "Unsloth: Patched trl.models.utils.disable_gradient_checkpointing with "
+        "a no-op to preserve Unsloth gradient checkpointing across TRL "
+        "generation passes."
+    )
+    return
+
+
 def patch_trl_openenv():
     for function in RL_ADDITIONAL_FUNCTIONS["openenv"]:
         logger.info(f"Unsloth: Patching trl openenv with function: {function.__name__}")
@@ -1981,6 +2060,14 @@ def patch_trl_vllm_generation():
 def PatchFastRL(algorithm = None, FastLanguageModel = None):
     if FastLanguageModel is not None:
         PatchRL(FastLanguageModel)
+    # Install the disable_gradient_checkpointing noop BEFORE
+    # patch_trl_rl_trainers. patch_trl_rl_trainers imports extra trl.* trainer
+    # submodules while generating the compiled cache; any new trl.* modules
+    # imported after the sys.modules walk would keep their original (broken)
+    # binding of disable_gradient_checkpointing. Running the noop install
+    # first ensures the canonical trl.models.utils symbol is already replaced
+    # before those submodules bind it.
+    patch_trl_disable_gradient_checkpointing()
     patch_trl_rl_trainers()
     patch_trl_openenv()
     patch_trl_vllm_generation()
diff --git a/unsloth/models/rl_replacements.py b/unsloth/models/rl_replacements.py
index 2544afe82e..93a7f89bcb 100755
--- a/unsloth/models/rl_replacements.py
+++ b/unsloth/models/rl_replacements.py
@@ -855,9 +855,7 @@ def grpo_trainer__get_per_token_logps_and_entropies(function_name, function):
                 image_sizes_chunks = chunk_optional(image_sizes, B)
 
             temperature = self.temperature
-            logit_softcapping = getattr(model.config, "final_logit_softcapping", 0)
-            if logit_softcapping is None:
-                logit_softcapping = 0
+            logit_softcapping = _unsloth_get_final_logit_softcapping(model.config)
             logit_scale_multiply = getattr(model.config, "logit_scale", 0)
             if logit_scale_multiply is None:
                 logit_scale_multiply = 0
@@ -1004,11 +1002,38 @@ def grpo_trainer__get_per_token_logps_and_entropies(function_name, function):
 
 RL_FUNCTIONS["grpo_trainer"].append(grpo_trainer__get_per_token_logps_and_entropies)
 
+
+def _unsloth_get_final_logit_softcapping(config):
+    """Return final_logit_softcapping for a model config, falling back to the
+    nested text sub-config for composite models. Handles both:
+      - Gemma-4-style configs where the attribute lives on ``config.text_config``
+      - T5Gemma-style composite configs where the text sub-config is only
+        reachable via ``config.get_text_config()``
+    Returns 0 if unset, matching the previous behaviour.
+    """
+    softcap = getattr(config, "final_logit_softcapping", None)
+    if softcap is None:
+        text_cfg = getattr(config, "text_config", None)
+        if text_cfg is None:
+            get_text_config = getattr(config, "get_text_config", None)
+            if callable(get_text_config):
+                try:
+                    text_cfg = get_text_config()
+                except (TypeError, ValueError):
+                    text_cfg = None
+        if text_cfg is not None and text_cfg is not config:
+            softcap = getattr(text_cfg, "final_logit_softcapping", None)
+    return 0 if softcap is None else softcap
+
+
 grpo_compute_loss = RL_REPLACEMENTS["grpo_compute_loss"]
 grpo_compute_loss_slow = RL_REPLACEMENTS["grpo_compute_loss_slow"]
 UnslothEfficientGRPO = RL_REPLACEMENTS["UnslothEfficientGRPO"]
 grpo_accumulated_loss = RL_REPLACEMENTS["grpo_accumulated_loss"]
 grpo_update_SamplingParams = RL_REPLACEMENTS["grpo_update_SamplingParams"]
+RL_PRE_ITEMS["grpo_trainer"].append(
+    inspect.getsource(_unsloth_get_final_logit_softcapping)
+)
 RL_PRE_ITEMS["grpo_trainer"].append(inspect.getsource(grpo_compute_loss))
 RL_PRE_ITEMS["grpo_trainer"].append(inspect.getsource(UnslothEfficientGRPO))
 RL_PRE_ITEMS["grpo_trainer"].append(inspect.getsource(grpo_accumulated_loss))
@@ -1107,9 +1132,7 @@ def grpo_trainer_compute_loss(function_name, function):
         input_ids = input_ids[:, -logits_to_keep:]
 
         # Get logit softcapping and logit scale
-        logit_softcapping = getattr(model.config, "final_logit_softcapping", 0)  # Gemma
-        if logit_softcapping is None:
-            logit_softcapping = 0
+        logit_softcapping = _unsloth_get_final_logit_softcapping(model.config)  # Gemma
         logit_scale_multiply = getattr(model.config, "logit_scale", 0)  # Cohere
         if logit_scale_multiply is None:
             logit_scale_multiply = 0

From 93a24f66980b2820083d2882fd3a720d894a1414 Mon Sep 17 00:00:00 2001
From: Daniel Han <danielhanchen@gmail.com>
Date: Sat, 11 Apr 2026 04:44:13 -0700
Subject: [PATCH 072/216] Add ROCm test suite for PR #4720 (#4824)

95 Python tests and 23 shell tests covering ROCm detection,
torch index URL selection, hardware flags, prebuilt asset selection,
and install pathway logic. All tests use mocks -- no AMD hardware required.

Companion to #4720 (AMD ROCm/HIP support).
---
 tests/sh/test_get_torch_index_url.sh      |  128 +-
 tests/studio/install/test_rocm_support.py | 1346 +++++++++++++++++++++
 2 files changed, 1472 insertions(+), 2 deletions(-)
 create mode 100644 tests/studio/install/test_rocm_support.py

diff --git a/tests/sh/test_get_torch_index_url.sh b/tests/sh/test_get_torch_index_url.sh
index 6387922712..adbdbe4e14 100755
--- a/tests/sh/test_get_torch_index_url.sh
+++ b/tests/sh/test_get_torch_index_url.sh
@@ -45,10 +45,32 @@ MOCK
     echo "$_dir"
 }
 
+# Helper: create a mock amd-smi that prints a given ROCm version string
+# Supports both "amd-smi version" and "amd-smi list" subcommands so that
+# the GPU presence check (amd-smi list) also succeeds in tests.
+make_mock_amd_smi() {
+    _dir=$(mktemp -d)
+    cat > "$_dir/amd-smi" <<MOCK
+#!/bin/sh
+case "\$1" in
+    list)
+        printf 'GPU: 0\\n  BDF: 0000:03:00.0\\n  NAME: gfx1100\\n'
+        ;;
+    *)
+        cat <<AMD_OUT
+AMDSMI Tool: 25.0.1+2b74356 | AMDSMI Library version: 25.0.1.0 | ROCm version: $1
+AMD_OUT
+        ;;
+esac
+MOCK
+    chmod +x "$_dir/amd-smi"
+    echo "$_dir"
+}
+
 # Build a minimal tools directory with symlinks to essential commands
-# (uname, grep, head, etc.) but WITHOUT nvidia-smi.
+# (uname, grep, head, etc.) but WITHOUT nvidia-smi or amd-smi.
 _TOOLS_DIR=$(mktemp -d)
-for _cmd in uname grep sed head sh bash cat; do
+for _cmd in uname grep sed head sh bash cat awk printf; do
     _real=$(command -v "$_cmd" 2>/dev/null || true)
     [ -n "$_real" ] && ln -sf "$_real" "$_TOOLS_DIR/$_cmd"
 done
@@ -119,6 +141,108 @@ _result=$(run_func "$_dir")
 assert_eq "unparseable -> cu126" "https://download.pytorch.org/whl/cu126" "$_result"
 rm -rf "$_dir"
 
+# 9) ROCm 6.3 (no nvidia-smi) -> rocm6.3
+_dir=$(make_mock_amd_smi "6.3")
+_result=$(run_func "$_dir")
+assert_eq "ROCm 6.3 -> rocm6.3" "https://download.pytorch.org/whl/rocm6.3" "$_result"
+rm -rf "$_dir"
+
+# 10) ROCm 7.1 (no nvidia-smi) -> rocm7.1
+_dir=$(make_mock_amd_smi "7.1")
+_result=$(run_func "$_dir")
+assert_eq "ROCm 7.1 -> rocm7.1" "https://download.pytorch.org/whl/rocm7.1" "$_result"
+rm -rf "$_dir"
+
+# 11) ROCm 7.2 (no nvidia-smi) -> rocm7.1 (capped due to torch <2.11.0)
+_dir=$(make_mock_amd_smi "7.2")
+_result=$(run_func "$_dir")
+assert_eq "ROCm 7.2 -> rocm7.1 (capped)" "https://download.pytorch.org/whl/rocm7.1" "$_result"
+rm -rf "$_dir"
+
+# 12) Both nvidia-smi and amd-smi present -> CUDA takes precedence
+_cuda_dir=$(make_mock_smi "12.6")
+_amd_dir=$(make_mock_amd_smi "6.3")
+_combined_dir=$(mktemp -d)
+ln -sf "$_cuda_dir/nvidia-smi" "$_combined_dir/nvidia-smi"
+ln -sf "$_amd_dir/amd-smi" "$_combined_dir/amd-smi"
+_result=$(run_func "$_combined_dir")
+assert_eq "CUDA+ROCm -> CUDA precedence" "https://download.pytorch.org/whl/cu126" "$_result"
+rm -rf "$_cuda_dir" "$_amd_dir" "$_combined_dir"
+
+# 13) No nvidia-smi, no amd-smi -> cpu (duplicate of test 1, confirms ROCm didn't break it)
+_result=$(run_func "none")
+assert_eq "no GPU -> cpu" "https://download.pytorch.org/whl/cpu" "$_result"
+
+# 14) ROCm 6.1 (no nvidia-smi) -> rocm6.1
+_dir=$(make_mock_amd_smi "6.1")
+_result=$(run_func "$_dir")
+assert_eq "ROCm 6.1 -> rocm6.1" "https://download.pytorch.org/whl/rocm6.1" "$_result"
+rm -rf "$_dir"
+
+# 15) ROCm 6.4 (no nvidia-smi) -> rocm6.4
+_dir=$(make_mock_amd_smi "6.4")
+_result=$(run_func "$_dir")
+assert_eq "ROCm 6.4 -> rocm6.4" "https://download.pytorch.org/whl/rocm6.4" "$_result"
+rm -rf "$_dir"
+
+# 16) ROCm 7.0 (no nvidia-smi) -> rocm7.0
+_dir=$(make_mock_amd_smi "7.0")
+_result=$(run_func "$_dir")
+assert_eq "ROCm 7.0 -> rocm7.0" "https://download.pytorch.org/whl/rocm7.0" "$_result"
+rm -rf "$_dir"
+
+# 17) ROCm 8.0 (future, no nvidia-smi) -> rocm7.1 (capped)
+_dir=$(make_mock_amd_smi "8.0")
+_result=$(run_func "$_dir")
+assert_eq "ROCm 8.0 -> rocm7.1 (capped)" "https://download.pytorch.org/whl/rocm7.1" "$_result"
+rm -rf "$_dir"
+
+# 18) Malformed amd-smi output (empty version field) -> cpu
+_dir=$(mktemp -d)
+cat > "$_dir/amd-smi" <<'MOCK'
+#!/bin/sh
+echo "AMDSMI Tool: 25.0.1 | AMDSMI Library version: 25.0.1.0 | ROCm version: "
+MOCK
+chmod +x "$_dir/amd-smi"
+_result=$(run_func "$_dir")
+assert_eq "empty amd-smi version -> cpu" "https://download.pytorch.org/whl/cpu" "$_result"
+rm -rf "$_dir"
+
+# 19) amd-smi with "N/A" version -> cpu
+_dir=$(mktemp -d)
+cat > "$_dir/amd-smi" <<'MOCK'
+#!/bin/sh
+echo "AMDSMI Tool: 25.0.1 | AMDSMI Library version: 25.0.1.0 | ROCm version: N/A"
+MOCK
+chmod +x "$_dir/amd-smi"
+_result=$(run_func "$_dir")
+assert_eq "N/A amd-smi version -> cpu" "https://download.pytorch.org/whl/cpu" "$_result"
+rm -rf "$_dir"
+
+# 20) ROCm version with trailing text (e.g. "6.3.1-beta") -> rocm6.3
+_dir=$(make_mock_amd_smi "6.3.1-beta")
+_result=$(run_func "$_dir")
+assert_eq "ROCm 6.3.1-beta -> rocm6.3" "https://download.pytorch.org/whl/rocm6.3" "$_result"
+rm -rf "$_dir"
+
+# 22) CUDA 12.6 still works after ROCm changes (regression check)
+_dir=$(make_mock_smi "12.6")
+_result=$(run_func "$_dir")
+assert_eq "CUDA 12.6 regression -> cu126" "https://download.pytorch.org/whl/cu126" "$_result"
+rm -rf "$_dir"
+
+# 23) CUDA 13.0 still works after ROCm changes (regression check)
+_dir=$(make_mock_smi "13.0")
+_result=$(run_func "$_dir")
+assert_eq "CUDA 13.0 regression -> cu130" "https://download.pytorch.org/whl/cu130" "$_result"
+rm -rf "$_dir"
+
+# 24) CUDA 12.8 still works after ROCm changes (regression check)
+_dir=$(make_mock_smi "12.8")
+_result=$(run_func "$_dir")
+assert_eq "CUDA 12.8 regression -> cu128" "https://download.pytorch.org/whl/cu128" "$_result"
+rm -rf "$_dir"
+
 rm -f "$_FUNC_FILE"
 rm -rf "$_FAKE_SMI_DIR"
 rm -rf "$_TOOLS_DIR"
diff --git a/tests/studio/install/test_rocm_support.py b/tests/studio/install/test_rocm_support.py
new file mode 100644
index 0000000000..48831fd57b
--- /dev/null
+++ b/tests/studio/install/test_rocm_support.py
@@ -0,0 +1,1346 @@
+"""Tests for AMD ROCm support across install pathways.
+
+Verifies that ROCm detection and installation logic works correctly
+WITHOUT breaking existing CUDA, CPU, macOS, and Windows pathways.
+All tests use mocks -- no AMD hardware required.
+"""
+
+import importlib.util
+import json
+import os
+import subprocess
+import sys
+from pathlib import Path
+from unittest.mock import MagicMock, patch, PropertyMock
+
+import pytest
+
+
+# ── Load modules under test ──────────────────────────────────────────────────
+
+PACKAGE_ROOT = Path(__file__).resolve().parents[3]
+
+# install_llama_prebuilt.py
+_PREBUILT_PATH = PACKAGE_ROOT / "studio" / "install_llama_prebuilt.py"
+_PREBUILT_SPEC = importlib.util.spec_from_file_location(
+    "studio_install_llama_prebuilt", _PREBUILT_PATH
+)
+assert _PREBUILT_SPEC is not None and _PREBUILT_SPEC.loader is not None
+prebuilt_mod = importlib.util.module_from_spec(_PREBUILT_SPEC)
+sys.modules[_PREBUILT_SPEC.name] = prebuilt_mod
+_PREBUILT_SPEC.loader.exec_module(prebuilt_mod)
+
+HostInfo = prebuilt_mod.HostInfo
+AssetChoice = prebuilt_mod.AssetChoice
+PrebuiltFallback = prebuilt_mod.PrebuiltFallback
+resolve_upstream_asset_choice = prebuilt_mod.resolve_upstream_asset_choice
+runtime_patterns_for_choice = prebuilt_mod.runtime_patterns_for_choice
+
+# install_python_stack.py
+_STACK_PATH = PACKAGE_ROOT / "studio" / "install_python_stack.py"
+_STACK_SPEC = importlib.util.spec_from_file_location(
+    "studio_install_python_stack", _STACK_PATH
+)
+assert _STACK_SPEC is not None and _STACK_SPEC.loader is not None
+stack_mod = importlib.util.module_from_spec(_STACK_SPEC)
+sys.modules[_STACK_SPEC.name] = stack_mod
+_STACK_SPEC.loader.exec_module(stack_mod)
+
+_detect_rocm_version = stack_mod._detect_rocm_version
+_ensure_rocm_torch = stack_mod._ensure_rocm_torch
+_has_rocm_gpu = stack_mod._has_rocm_gpu
+_has_usable_nvidia_gpu = stack_mod._has_usable_nvidia_gpu
+_ROCM_TORCH_INDEX = stack_mod._ROCM_TORCH_INDEX
+
+
+# ── Helper: build HostInfo for different scenarios ──────────────────────────
+
+
+def nvidia_host(**overrides) -> HostInfo:
+    """NVIDIA Linux x86_64 host."""
+    defaults = dict(
+        system = "Linux",
+        machine = "x86_64",
+        is_windows = False,
+        is_linux = True,
+        is_macos = False,
+        is_x86_64 = True,
+        is_arm64 = False,
+        nvidia_smi = "/usr/bin/nvidia-smi",
+        driver_cuda_version = (12, 6),
+        compute_caps = ["89"],
+        visible_cuda_devices = None,
+        has_physical_nvidia = True,
+        has_usable_nvidia = True,
+        has_rocm = False,
+    )
+    defaults.update(overrides)
+    return HostInfo(**defaults)
+
+
+def rocm_host(**overrides) -> HostInfo:
+    """AMD ROCm Linux x86_64 host (no NVIDIA)."""
+    defaults = dict(
+        system = "Linux",
+        machine = "x86_64",
+        is_windows = False,
+        is_linux = True,
+        is_macos = False,
+        is_x86_64 = True,
+        is_arm64 = False,
+        nvidia_smi = None,
+        driver_cuda_version = None,
+        compute_caps = [],
+        visible_cuda_devices = None,
+        has_physical_nvidia = False,
+        has_usable_nvidia = False,
+        has_rocm = True,
+    )
+    defaults.update(overrides)
+    return HostInfo(**defaults)
+
+
+def cpu_host(**overrides) -> HostInfo:
+    """CPU-only Linux x86_64 host."""
+    defaults = dict(
+        system = "Linux",
+        machine = "x86_64",
+        is_windows = False,
+        is_linux = True,
+        is_macos = False,
+        is_x86_64 = True,
+        is_arm64 = False,
+        nvidia_smi = None,
+        driver_cuda_version = None,
+        compute_caps = [],
+        visible_cuda_devices = None,
+        has_physical_nvidia = False,
+        has_usable_nvidia = False,
+        has_rocm = False,
+    )
+    defaults.update(overrides)
+    return HostInfo(**defaults)
+
+
+def macos_host(**overrides) -> HostInfo:
+    """macOS arm64 host."""
+    defaults = dict(
+        system = "Darwin",
+        machine = "arm64",
+        is_windows = False,
+        is_linux = False,
+        is_macos = True,
+        is_x86_64 = False,
+        is_arm64 = True,
+        nvidia_smi = None,
+        driver_cuda_version = None,
+        compute_caps = [],
+        visible_cuda_devices = None,
+        has_physical_nvidia = False,
+        has_usable_nvidia = False,
+        has_rocm = False,
+    )
+    defaults.update(overrides)
+    return HostInfo(**defaults)
+
+
+def windows_host(**overrides) -> HostInfo:
+    """Windows x86_64 host."""
+    defaults = dict(
+        system = "Windows",
+        machine = "amd64",
+        is_windows = True,
+        is_linux = False,
+        is_macos = False,
+        is_x86_64 = True,
+        is_arm64 = False,
+        nvidia_smi = None,
+        driver_cuda_version = None,
+        compute_caps = [],
+        visible_cuda_devices = None,
+        has_physical_nvidia = False,
+        has_usable_nvidia = False,
+        has_rocm = False,
+    )
+    defaults.update(overrides)
+    return HostInfo(**defaults)
+
+
+def windows_rocm_host(**overrides) -> HostInfo:
+    """Windows x86_64 host with ROCm."""
+    defaults = dict(
+        system = "Windows",
+        machine = "amd64",
+        is_windows = True,
+        is_linux = False,
+        is_macos = False,
+        is_x86_64 = True,
+        is_arm64 = False,
+        nvidia_smi = None,
+        driver_cuda_version = None,
+        compute_caps = [],
+        visible_cuda_devices = None,
+        has_physical_nvidia = False,
+        has_usable_nvidia = False,
+        has_rocm = True,
+    )
+    defaults.update(overrides)
+    return HostInfo(**defaults)
+
+
+# ── Upstream asset fixture ───────────────────────────────────────────────────
+
+LLAMA_TAG = "b8508"
+
+UPSTREAM_ASSETS = {
+    f"llama-{LLAMA_TAG}-bin-ubuntu-x64.tar.gz": f"https://example.com/{LLAMA_TAG}-linux-cpu.tar.gz",
+    f"llama-{LLAMA_TAG}-bin-ubuntu-rocm-7.2-x64.tar.gz": f"https://example.com/{LLAMA_TAG}-linux-rocm.tar.gz",
+    f"llama-{LLAMA_TAG}-bin-win-cpu-x64.zip": f"https://example.com/{LLAMA_TAG}-win-cpu.zip",
+    f"llama-{LLAMA_TAG}-bin-win-cuda-12.4-x64.zip": f"https://example.com/{LLAMA_TAG}-win-cuda.zip",
+    f"llama-{LLAMA_TAG}-bin-win-hip-radeon-x64.zip": f"https://example.com/{LLAMA_TAG}-win-hip.zip",
+    f"llama-{LLAMA_TAG}-bin-macos-arm64.tar.gz": f"https://example.com/{LLAMA_TAG}-macos-arm64.tar.gz",
+    f"llama-{LLAMA_TAG}-bin-macos-x64.tar.gz": f"https://example.com/{LLAMA_TAG}-macos-x64.tar.gz",
+}
+
+
+# =============================================================================
+# TEST: install_llama_prebuilt.py -- resolve_upstream_asset_choice
+# =============================================================================
+
+
+class TestResolveUpstreamAssetChoice:
+    """Verify that the asset selection logic picks the right binary for each platform."""
+
+    @patch.object(prebuilt_mod, "github_release_assets", return_value = UPSTREAM_ASSETS)
+    def test_nvidia_linux_gets_cpu_asset(self, mock_assets):
+        """NVIDIA host should NOT hit the ROCm path -- gets CPU asset (CUDA handled elsewhere)."""
+        host = nvidia_host()
+        choice = resolve_upstream_asset_choice(host, LLAMA_TAG)
+        assert choice.install_kind == "linux-cpu"
+        assert "ubuntu-x64" in choice.name
+        assert "rocm" not in choice.name
+
+    @patch.object(prebuilt_mod, "github_release_assets", return_value = UPSTREAM_ASSETS)
+    def test_rocm_linux_gets_rocm_prebuilt(self, mock_assets):
+        """AMD ROCm Linux host should get the ROCm prebuilt."""
+        host = rocm_host()
+        choice = resolve_upstream_asset_choice(host, LLAMA_TAG)
+        assert choice.install_kind == "linux-rocm"
+        assert "rocm" in choice.name
+
+    @patch.object(prebuilt_mod, "github_release_assets", return_value = UPSTREAM_ASSETS)
+    def test_cpu_linux_gets_cpu_asset(self, mock_assets):
+        """CPU-only Linux host should get CPU asset."""
+        host = cpu_host()
+        choice = resolve_upstream_asset_choice(host, LLAMA_TAG)
+        assert choice.install_kind == "linux-cpu"
+        assert "ubuntu-x64" in choice.name
+
+    @patch.object(prebuilt_mod, "github_release_assets", return_value = UPSTREAM_ASSETS)
+    def test_macos_arm64_gets_macos_asset(self, mock_assets):
+        """macOS arm64 host should get macOS asset."""
+        host = macos_host()
+        choice = resolve_upstream_asset_choice(host, LLAMA_TAG)
+        assert choice.install_kind == "macos-arm64"
+        assert "macos-arm64" in choice.name
+
+    @patch.object(prebuilt_mod, "github_release_assets", return_value = UPSTREAM_ASSETS)
+    def test_windows_cpu_gets_cpu_asset(self, mock_assets):
+        """Windows CPU-only host should get Windows CPU asset."""
+        host = windows_host()
+        choice = resolve_upstream_asset_choice(host, LLAMA_TAG)
+        assert choice.install_kind == "windows-cpu"
+        assert "win-cpu" in choice.name
+
+    @patch.object(prebuilt_mod, "github_release_assets", return_value = UPSTREAM_ASSETS)
+    def test_windows_rocm_gets_hip_asset(self, mock_assets):
+        """Windows ROCm host should get Windows HIP asset."""
+        host = windows_rocm_host()
+        choice = resolve_upstream_asset_choice(host, LLAMA_TAG)
+        assert choice.install_kind == "windows-hip"
+        assert "hip" in choice.name
+
+    @patch.object(prebuilt_mod, "github_release_assets", return_value = UPSTREAM_ASSETS)
+    def test_mixed_nvidia_rocm_prefers_nvidia(self, mock_assets):
+        """Host with both NVIDIA and ROCm should use NVIDIA (CPU path here, CUDA elsewhere)."""
+        host = nvidia_host(has_rocm = True)
+        choice = resolve_upstream_asset_choice(host, LLAMA_TAG)
+        # NVIDIA hosts go through the normal path (CUDA handled by resolve_linux_cuda_choice)
+        assert choice.install_kind == "linux-cpu"
+        assert "rocm" not in choice.name
+
+    @patch.object(prebuilt_mod, "github_release_assets")
+    def test_rocm_linux_no_prebuilt_falls_back(self, mock_assets):
+        """AMD ROCm host should fall back to source build when no ROCm prebuilt exists."""
+        # Remove the ROCm asset from available assets
+        assets_without_rocm = {
+            k: v for k, v in UPSTREAM_ASSETS.items() if "rocm" not in k
+        }
+        mock_assets.return_value = assets_without_rocm
+        host = rocm_host()
+        with pytest.raises(PrebuiltFallback, match = "ROCm detected"):
+            resolve_upstream_asset_choice(host, LLAMA_TAG)
+
+    @patch.object(prebuilt_mod, "github_release_assets")
+    def test_windows_rocm_no_hip_falls_to_cpu(self, mock_assets):
+        """Windows+ROCm with HIP prebuilt missing should fall through to CPU."""
+        assets_no_hip = {k: v for k, v in UPSTREAM_ASSETS.items() if "hip" not in k}
+        mock_assets.return_value = assets_no_hip
+        host = windows_rocm_host()
+        choice = resolve_upstream_asset_choice(host, LLAMA_TAG)
+        assert choice.install_kind == "windows-cpu"
+
+    @patch.object(prebuilt_mod, "github_release_assets", return_value = UPSTREAM_ASSETS)
+    def test_macos_rocm_impossible_has_rocm_false(self, mock_assets):
+        """macOS host should never have has_rocm=True in practice; verify it gets macOS asset."""
+        host = macos_host(has_rocm = True)
+        choice = resolve_upstream_asset_choice(host, LLAMA_TAG)
+        assert choice.install_kind == "macos-arm64"
+
+    @patch.object(prebuilt_mod, "github_release_assets", return_value = UPSTREAM_ASSETS)
+    def test_linux_aarch64_rocm_gets_prebuilt_fallback(self, mock_assets):
+        """Linux aarch64 with ROCm -- no x86_64 match, should raise PrebuiltFallback."""
+        host = rocm_host(machine = "aarch64", is_x86_64 = False, is_arm64 = True)
+        with pytest.raises(PrebuiltFallback):
+            resolve_upstream_asset_choice(host, LLAMA_TAG)
+
+
+# =============================================================================
+# TEST: install_llama_prebuilt.py -- runtime_patterns_for_choice
+# =============================================================================
+
+
+class TestRuntimePatterns:
+    """Verify runtime file patterns for all install kinds."""
+
+    def test_linux_cpu_patterns(self):
+        choice = AssetChoice(
+            repo = "", tag = "", name = "", url = "", source_label = "", install_kind = "linux-cpu"
+        )
+        patterns = runtime_patterns_for_choice(choice)
+        assert "llama-server" in patterns
+        assert "llama-quantize" in patterns
+
+    def test_linux_cuda_patterns(self):
+        choice = AssetChoice(
+            repo = "", tag = "", name = "", url = "", source_label = "", install_kind = "linux-cuda"
+        )
+        patterns = runtime_patterns_for_choice(choice)
+        assert "libggml-cuda.so*" in patterns
+
+    def test_linux_rocm_patterns(self):
+        choice = AssetChoice(
+            repo = "", tag = "", name = "", url = "", source_label = "", install_kind = "linux-rocm"
+        )
+        patterns = runtime_patterns_for_choice(choice)
+        assert "libggml-hip.so*" in patterns
+        assert "llama-server" in patterns
+
+    def test_windows_hip_patterns(self):
+        choice = AssetChoice(
+            repo = "",
+            tag = "",
+            name = "",
+            url = "",
+            source_label = "",
+            install_kind = "windows-hip",
+        )
+        patterns = runtime_patterns_for_choice(choice)
+        assert "*.exe" in patterns
+        assert "*.dll" in patterns
+
+    def test_macos_patterns(self):
+        choice = AssetChoice(
+            repo = "",
+            tag = "",
+            name = "",
+            url = "",
+            source_label = "",
+            install_kind = "macos-arm64",
+        )
+        patterns = runtime_patterns_for_choice(choice)
+        assert "lib*.dylib" in patterns
+
+
+# =============================================================================
+# TEST: install_llama_prebuilt.py -- HostInfo.has_rocm field
+# =============================================================================
+
+
+class TestHostInfoRocm:
+    """Verify has_rocm field does not affect other HostInfo behavior."""
+
+    def test_has_rocm_default_false(self):
+        host = HostInfo(
+            system = "Linux",
+            machine = "x86_64",
+            is_windows = False,
+            is_linux = True,
+            is_macos = False,
+            is_x86_64 = True,
+            is_arm64 = False,
+            nvidia_smi = None,
+            driver_cuda_version = None,
+            compute_caps = [],
+            visible_cuda_devices = None,
+            has_physical_nvidia = False,
+            has_usable_nvidia = False,
+        )
+        assert host.has_rocm is False
+
+    def test_has_rocm_explicit_true(self):
+        host = rocm_host()
+        assert host.has_rocm is True
+
+    def test_nvidia_host_no_rocm(self):
+        host = nvidia_host()
+        assert host.has_rocm is False
+        assert host.has_usable_nvidia is True
+
+    def test_detect_host_has_rocm_detection_logic(self):
+        """detect_host() should have ROCm GPU detection logic."""
+        import inspect
+
+        source = inspect.getsource(prebuilt_mod.detect_host)
+        # Must probe for actual GPU, not just tool presence
+        assert "rocminfo" in source or "amd-smi" in source
+
+    def test_detect_host_windows_rocm_detection(self):
+        """detect_host() source should have Windows-specific ROCm GPU detection."""
+        import inspect
+
+        source = inspect.getsource(prebuilt_mod.detect_host)
+        assert "hipinfo" in source or "amd-smi" in source
+
+
+# =============================================================================
+# TEST: install_python_stack.py -- _detect_rocm_version
+# =============================================================================
+
+
+class TestDetectRocmVersion:
+    """Verify ROCm version detection from various sources."""
+
+    def test_no_rocm_returns_none(self, tmp_path):
+        """No ROCm installed should return None."""
+        with patch.dict(os.environ, {"ROCM_PATH": str(tmp_path / "nonexistent")}):
+            with patch("shutil.which", return_value = None):
+                result = _detect_rocm_version()
+                assert result is None
+
+    def test_version_from_file(self, tmp_path):
+        """Reads version from /opt/rocm/.info/version."""
+        info_dir = tmp_path / ".info"
+        info_dir.mkdir()
+        (info_dir / "version").write_text("7.1.0-12345\n")
+        with patch.dict(os.environ, {"ROCM_PATH": str(tmp_path)}):
+            result = _detect_rocm_version()
+            assert result == (7, 1)
+
+    def test_version_62(self, tmp_path):
+        """Reads ROCm 6.2 version."""
+        info_dir = tmp_path / ".info"
+        info_dir.mkdir()
+        (info_dir / "version").write_text("6.2.0\n")
+        with patch.dict(os.environ, {"ROCM_PATH": str(tmp_path)}):
+            result = _detect_rocm_version()
+            assert result == (6, 2)
+
+    def test_hipconfig_fallback(self, tmp_path):
+        """Falls back to hipconfig --version when file not found."""
+        with patch.dict(os.environ, {"ROCM_PATH": str(tmp_path / "nonexistent")}):
+            mock_result = MagicMock()
+            mock_result.returncode = 0
+            mock_result.stdout = b"6.3.21234.2\n"
+            with patch("shutil.which", return_value = "/usr/bin/hipconfig"):
+                with patch("subprocess.run", return_value = mock_result):
+                    result = _detect_rocm_version()
+                    assert result == (6, 3)
+
+    def test_empty_version_file(self, tmp_path):
+        """Empty version file should return None."""
+        info_dir = tmp_path / ".info"
+        info_dir.mkdir()
+        (info_dir / "version").write_text("")
+        with patch.dict(os.environ, {"ROCM_PATH": str(tmp_path)}):
+            with patch("shutil.which", return_value = None):
+                result = _detect_rocm_version()
+                assert result is None
+
+    def test_version_with_epoch_prefix(self, tmp_path):
+        """Debian epoch prefix (2:6.2.0) -- version file has no epoch, so should parse."""
+        info_dir = tmp_path / ".info"
+        info_dir.mkdir()
+        # Version files don't typically have epoch prefix, but lib/rocm_version might
+        (info_dir / "version").write_text("6.2.0\n")
+        with patch.dict(os.environ, {"ROCM_PATH": str(tmp_path)}):
+            result = _detect_rocm_version()
+            assert result == (6, 2)
+
+    def test_multiple_version_sources_first_wins(self, tmp_path):
+        """When both .info/version and lib/rocm_version exist, first found wins."""
+        info_dir = tmp_path / ".info"
+        info_dir.mkdir()
+        (info_dir / "version").write_text("7.1.0\n")
+        lib_dir = tmp_path / "lib"
+        lib_dir.mkdir()
+        (lib_dir / "rocm_version").write_text("6.3.0\n")
+        with patch.dict(os.environ, {"ROCM_PATH": str(tmp_path)}):
+            result = _detect_rocm_version()
+            assert result == (7, 1)  # .info/version checked first
+
+    def test_hipconfig_multiline_output(self, tmp_path):
+        """hipconfig with multi-line output -- should use first line."""
+        with patch.dict(os.environ, {"ROCM_PATH": str(tmp_path / "nonexistent")}):
+            mock_result = MagicMock()
+            mock_result.returncode = 0
+            mock_result.stdout = b"6.3.21234.2\nSome extra info\n"
+            with patch("shutil.which", return_value = "/usr/bin/hipconfig"):
+                with patch("subprocess.run", return_value = mock_result):
+                    result = _detect_rocm_version()
+                    assert result == (6, 3)
+
+    def test_hipconfig_timeout(self, tmp_path):
+        """hipconfig that times out should return None."""
+        with patch.dict(os.environ, {"ROCM_PATH": str(tmp_path / "nonexistent")}):
+            with patch("shutil.which", return_value = "/usr/bin/hipconfig"):
+                with patch(
+                    "subprocess.run",
+                    side_effect = subprocess.TimeoutExpired("hipconfig", 5),
+                ):
+                    result = _detect_rocm_version()
+                    assert result is None
+
+
+# =============================================================================
+# TEST: install_python_stack.py -- _ensure_rocm_torch
+# =============================================================================
+
+
+class TestEnsureRocmTorch:
+    """Verify ROCm torch reinstall logic."""
+
+    @patch.object(stack_mod, "pip_install")
+    @patch.object(stack_mod, "_has_usable_nvidia_gpu", return_value = False)
+    def test_no_rocm_skips(self, mock_nvidia, mock_pip):
+        """No ROCm toolchain should skip entirely."""
+        with patch("os.path.isdir", return_value = False):
+            with patch("shutil.which", return_value = None):
+                _ensure_rocm_torch()
+        mock_pip.assert_not_called()
+
+    @patch.object(stack_mod, "pip_install")
+    @patch.object(stack_mod, "_has_usable_nvidia_gpu", return_value = False)
+    @patch.object(stack_mod, "_has_rocm_gpu", return_value = True)
+    @patch.object(stack_mod, "_detect_rocm_version", return_value = (7, 1))
+    def test_torch_already_has_cuda_skips(
+        self, mock_ver, mock_gpu, mock_nvidia, mock_pip
+    ):
+        """If torch already has CUDA, should skip ROCm reinstall."""
+        mock_probe = MagicMock()
+        mock_probe.returncode = 0
+        mock_probe.stdout = b"12.6\n"  # CUDA version string
+        with patch("os.path.isdir", return_value = True):
+            with patch("subprocess.run", return_value = mock_probe):
+                _ensure_rocm_torch()
+        mock_pip.assert_not_called()
+
+    @patch.object(stack_mod, "pip_install")
+    @patch.object(stack_mod, "_has_usable_nvidia_gpu", return_value = False)
+    @patch.object(stack_mod, "_has_rocm_gpu", return_value = True)
+    @patch.object(stack_mod, "_detect_rocm_version", return_value = (7, 1))
+    def test_torch_already_has_hip_skips(
+        self, mock_ver, mock_gpu, mock_nvidia, mock_pip
+    ):
+        """If torch already has HIP, should skip ROCm reinstall."""
+        mock_probe = MagicMock()
+        mock_probe.returncode = 0
+        mock_probe.stdout = b"7.1.12345\n"  # HIP version string
+        with patch("os.path.isdir", return_value = True):
+            with patch("subprocess.run", return_value = mock_probe):
+                _ensure_rocm_torch()
+        mock_pip.assert_not_called()
+
+    @patch.object(stack_mod, "pip_install")
+    @patch.object(stack_mod, "_has_usable_nvidia_gpu", return_value = False)
+    @patch.object(stack_mod, "_has_rocm_gpu", return_value = True)
+    @patch.object(stack_mod, "_detect_rocm_version", return_value = (7, 1))
+    def test_cpu_torch_gets_rocm_reinstall(
+        self, mock_ver, mock_gpu, mock_nvidia, mock_pip
+    ):
+        """CPU-only torch on ROCm host should trigger reinstall."""
+        mock_probe = MagicMock()
+        mock_probe.returncode = 0
+        mock_probe.stdout = b"\n"  # empty = no GPU backend
+        with patch("os.path.isdir", return_value = True):
+            with patch("subprocess.run", return_value = mock_probe):
+                _ensure_rocm_torch()
+        # Should call pip_install twice: once for torch, once for bitsandbytes
+        assert mock_pip.call_count == 2
+        torch_call = mock_pip.call_args_list[0]
+        assert "rocm7.1" in str(torch_call)
+        bnb_call = mock_pip.call_args_list[1]
+        assert "bitsandbytes" in str(bnb_call)
+
+    @patch.object(stack_mod, "pip_install")
+    @patch.object(stack_mod, "_has_usable_nvidia_gpu", return_value = False)
+    @patch.object(stack_mod, "_has_rocm_gpu", return_value = True)
+    @patch.object(stack_mod, "_detect_rocm_version", return_value = (6, 3))
+    def test_rocm_63_selects_correct_tag(
+        self, mock_ver, mock_gpu, mock_nvidia, mock_pip
+    ):
+        """ROCm 6.3 should select rocm6.3 tag."""
+        mock_probe = MagicMock()
+        mock_probe.returncode = 0
+        mock_probe.stdout = b"\n"
+        with patch("os.path.isdir", return_value = True):
+            with patch("subprocess.run", return_value = mock_probe):
+                _ensure_rocm_torch()
+        torch_call = mock_pip.call_args_list[0]
+        assert "rocm6.3" in str(torch_call)
+
+    @patch.object(stack_mod, "pip_install")
+    @patch.object(stack_mod, "_has_usable_nvidia_gpu", return_value = False)
+    @patch.object(stack_mod, "_has_rocm_gpu", return_value = True)
+    @patch.object(stack_mod, "_detect_rocm_version", return_value = (5, 0))
+    def test_old_rocm_skips(self, mock_ver, mock_gpu, mock_nvidia, mock_pip):
+        """ROCm version too old (below 6.0) should skip."""
+        mock_probe = MagicMock()
+        mock_probe.returncode = 0
+        mock_probe.stdout = b"\n"
+        with patch("os.path.isdir", return_value = True):
+            with patch("subprocess.run", return_value = mock_probe):
+                _ensure_rocm_torch()
+        mock_pip.assert_not_called()
+
+    @patch.object(stack_mod, "pip_install")
+    @patch.object(stack_mod, "_has_usable_nvidia_gpu", return_value = False)
+    @patch.object(stack_mod, "_has_rocm_gpu", return_value = True)
+    @patch.object(stack_mod, "_detect_rocm_version", return_value = None)
+    def test_version_unreadable_prints_warning(
+        self, mock_ver, mock_gpu, mock_nvidia, mock_pip, capsys
+    ):
+        """ROCm detected but version unreadable should print warning and skip."""
+        with patch("os.path.isdir", return_value = True):
+            _ensure_rocm_torch()
+        mock_pip.assert_not_called()
+        captured = capsys.readouterr()
+        assert "unreadable" in captured.out
+
+    @patch.object(stack_mod, "pip_install")
+    @patch.object(stack_mod, "_has_usable_nvidia_gpu", return_value = False)
+    @patch.object(stack_mod, "_has_rocm_gpu", return_value = True)
+    @patch.object(stack_mod, "_detect_rocm_version", return_value = (7, 2))
+    def test_rocm_72_selects_71_tag(self, mock_ver, mock_gpu, mock_nvidia, mock_pip):
+        """ROCm 7.2 should select rocm7.1 tag (capped, not in mapping)."""
+        mock_probe = MagicMock()
+        mock_probe.returncode = 0
+        mock_probe.stdout = b"\n"
+        with patch("os.path.isdir", return_value = True):
+            with patch("subprocess.run", return_value = mock_probe):
+                _ensure_rocm_torch()
+        torch_call = mock_pip.call_args_list[0]
+        assert "rocm7.1" in str(torch_call)
+
+    @patch.object(stack_mod, "pip_install")
+    @patch.object(stack_mod, "_has_usable_nvidia_gpu", return_value = False)
+    @patch.object(stack_mod, "_has_rocm_gpu", return_value = True)
+    @patch.object(stack_mod, "_detect_rocm_version", return_value = (7, 1))
+    def test_probe_timeout_triggers_reinstall(
+        self, mock_ver, mock_gpu, mock_nvidia, mock_pip
+    ):
+        """Probe subprocess timeout should not crash; should proceed to reinstall."""
+        with patch("os.path.isdir", return_value = True):
+            with patch(
+                "subprocess.run", side_effect = subprocess.TimeoutExpired("python", 30)
+            ):
+                _ensure_rocm_torch()
+        # If probe times out, the function should treat torch as unusable and reinstall
+        assert mock_pip.call_count == 2
+        assert "rocm7.1" in str(mock_pip.call_args_list[0])
+
+    @patch.object(stack_mod, "pip_install")
+    @patch.object(stack_mod, "_has_usable_nvidia_gpu", return_value = False)
+    @patch.object(stack_mod, "_has_rocm_gpu", return_value = False)
+    def test_no_gpu_with_rocm_tools_skips(self, mock_gpu, mock_nvidia, mock_pip):
+        """ROCm tools present but no actual AMD GPU should skip entirely."""
+        with patch("os.path.isdir", return_value = True):
+            _ensure_rocm_torch()
+        mock_pip.assert_not_called()
+
+
+# =============================================================================
+# TEST: install_python_stack.py -- _ROCM_TORCH_INDEX mapping
+# =============================================================================
+
+
+class TestRocmTorchIndex:
+    """Verify the ROCm version -> torch index tag mapping."""
+
+    def test_mapping_is_sorted_descending(self):
+        """Keys should be in descending order for the next() iteration to work."""
+        keys = list(_ROCM_TORCH_INDEX.keys())
+        assert keys == sorted(keys, reverse = True)
+
+    def test_rocm_72_not_in_mapping(self):
+        """ROCm 7.2 should NOT be in the active mapping (torch 2.11.0 exceeds bound)."""
+        assert (7, 2) not in _ROCM_TORCH_INDEX
+
+    def test_rocm_71_maps_correctly(self):
+        assert _ROCM_TORCH_INDEX[(7, 1)] == "rocm7.1"
+
+    def test_rocm_63_maps_correctly(self):
+        assert _ROCM_TORCH_INDEX[(6, 3)] == "rocm6.3"
+
+    def test_rocm_60_maps_correctly(self):
+        assert _ROCM_TORCH_INDEX[(6, 0)] == "rocm6.0"
+
+    def test_all_tags_use_download_pytorch(self):
+        """All tags should be for download.pytorch.org, not repo.radeon.com."""
+        for tag in _ROCM_TORCH_INDEX.values():
+            assert tag.startswith("rocm")
+            assert "radeon" not in tag
+
+    def test_newer_rocm_selects_best_match(self):
+        """ROCm 7.2 (not in map) should select rocm7.1 via >= comparison."""
+        ver = (7, 2)
+        tag = next(
+            (
+                t
+                for (maj, mn), t in sorted(_ROCM_TORCH_INDEX.items(), reverse = True)
+                if ver >= (maj, mn)
+            ),
+            None,
+        )
+        assert tag == "rocm7.1"
+
+    def test_rocm_64_selects_64(self):
+        ver = (6, 4)
+        tag = next(
+            (
+                t
+                for (maj, mn), t in sorted(_ROCM_TORCH_INDEX.items(), reverse = True)
+                if ver >= (maj, mn)
+            ),
+            None,
+        )
+        assert tag == "rocm6.4"
+
+
+# =============================================================================
+# TEST: hardware.py -- IS_ROCM flag and detect_hardware
+# =============================================================================
+
+
+class TestHardwareRocmFlag:
+    """Verify IS_ROCM flag behavior without importing the full hardware module."""
+
+    def test_hardware_py_has_is_rocm(self):
+        """hardware.py should define IS_ROCM."""
+        hw_path = (
+            PACKAGE_ROOT / "studio" / "backend" / "utils" / "hardware" / "hardware.py"
+        )
+        source = hw_path.read_text()
+        assert "IS_ROCM: bool" in source and "False" in source
+
+    def test_hardware_py_sets_is_rocm_on_hip(self):
+        """detect_hardware() should set IS_ROCM when torch.version.hip is set."""
+        hw_path = (
+            PACKAGE_ROOT / "studio" / "backend" / "utils" / "hardware" / "hardware.py"
+        )
+        source = hw_path.read_text()
+        assert 'torch.version, "hip"' in source or "torch.version.hip" in source
+
+    def test_hardware_py_still_returns_cuda_for_rocm(self):
+        """DeviceType should remain CUDA even on ROCm -- no DeviceType.ROCM."""
+        hw_path = (
+            PACKAGE_ROOT / "studio" / "backend" / "utils" / "hardware" / "hardware.py"
+        )
+        source = hw_path.read_text()
+        # Ensure ROCM is NOT a DeviceType member
+        enum_section = source.split("class DeviceType")[1].split("\n\n")[0]
+        assert "ROCM" not in enum_section
+
+    def test_hardware_py_has_rocm_in_package_versions(self):
+        """get_package_versions() should include 'rocm' key."""
+        hw_path = (
+            PACKAGE_ROOT / "studio" / "backend" / "utils" / "hardware" / "hardware.py"
+        )
+        source = hw_path.read_text()
+        assert '"rocm"' in source
+
+    def test_hardware_py_device_type_cuda_references_intact(self):
+        """All existing DeviceType.CUDA references should still be present."""
+        hw_path = (
+            PACKAGE_ROOT / "studio" / "backend" / "utils" / "hardware" / "hardware.py"
+        )
+        source = hw_path.read_text()
+        # Key functions that must still reference DeviceType.CUDA
+        assert "DeviceType.CUDA" in source
+        assert "DEVICE = DeviceType.CUDA" in source
+
+    def test_is_rocm_exported_from_init(self):
+        """IS_ROCM should be exported from hardware __init__.py."""
+        init_path = (
+            PACKAGE_ROOT / "studio" / "backend" / "utils" / "hardware" / "__init__.py"
+        )
+        source = init_path.read_text()
+        assert "IS_ROCM" in source
+
+    def test_is_rocm_in_all_list(self):
+        """IS_ROCM should be in __all__ list in __init__.py."""
+        init_path = (
+            PACKAGE_ROOT / "studio" / "backend" / "utils" / "hardware" / "__init__.py"
+        )
+        source = init_path.read_text()
+        # Extract __all__ section
+        assert '"IS_ROCM"' in source
+
+    def test_get_package_versions_returns_rocm_key(self):
+        """get_package_versions() source should return both 'cuda' and 'rocm' keys."""
+        hw_path = (
+            PACKAGE_ROOT / "studio" / "backend" / "utils" / "hardware" / "hardware.py"
+        )
+        source = hw_path.read_text()
+        # Find the get_package_versions function body
+        func_start = source.find("def get_package_versions")
+        func_body = source[func_start : source.find("\ndef ", func_start + 1)]
+        assert '"cuda"' in func_body
+        assert '"rocm"' in func_body
+
+
+# =============================================================================
+# TEST: tokenizer_utils.py -- error message
+# =============================================================================
+
+
+class TestTokenizerErrorMessage:
+    """Verify the AMD error message is updated."""
+
+    def test_no_old_amd_message(self):
+        """Old 'We do not support AMD' message should be gone."""
+        tu_path = PACKAGE_ROOT / "unsloth" / "tokenizer_utils.py"
+        source = tu_path.read_text()
+        assert "We do not support AMD" not in source
+
+    def test_new_message_has_docs_link(self):
+        """New message should point to Unsloth AMD docs."""
+        tu_path = PACKAGE_ROOT / "unsloth" / "tokenizer_utils.py"
+        source = tu_path.read_text()
+        assert "docs.unsloth.ai" in source or "No GPU detected" in source
+
+
+# =============================================================================
+# TEST: install.sh -- structural checks
+# =============================================================================
+
+
+class TestInstallShStructure:
+    """Verify install.sh structural properties without running it."""
+
+    def test_no_here_strings(self):
+        """install.sh must not use <<< (not POSIX)."""
+        sh_path = PACKAGE_ROOT / "install.sh"
+        source = sh_path.read_text()
+        # <<< is bash-only; breaks dash
+        for i, line in enumerate(source.splitlines(), 1):
+            stripped = line.lstrip()
+            if stripped.startswith("#"):
+                continue
+            assert "<<<" not in line, f"install.sh:{i} uses non-POSIX <<< here-string"
+
+    def test_rocm_detection_present(self):
+        """install.sh should have ROCm detection in get_torch_index_url."""
+        sh_path = PACKAGE_ROOT / "install.sh"
+        source = sh_path.read_text()
+        assert "amd-smi" in source
+        assert "rocm" in source.lower()
+
+    def test_cuda_precedence(self):
+        """ROCm detection should only run when nvidia-smi is absent."""
+        sh_path = PACKAGE_ROOT / "install.sh"
+        source = sh_path.read_text()
+        # The ROCm block should be inside the "if [ -z "$_smi" ]" branch
+        smi_block_start = source.find('if [ -z "$_smi" ]')
+        rocm_block_start = source.find("amd-smi")
+        assert (
+            smi_block_start < rocm_block_start
+        ), "ROCm detection should be inside the 'no nvidia-smi' branch"
+
+    def test_bitsandbytes_amd_install(self):
+        """install.sh should install bitsandbytes for AMD when ROCm detected."""
+        sh_path = PACKAGE_ROOT / "install.sh"
+        source = sh_path.read_text()
+        assert "bitsandbytes" in source
+        assert "rocm*)" in source  # case pattern for ROCm URLs
+
+    def test_cpu_hint_mentions_amd(self):
+        """CPU-only hint should mention AMD ROCm."""
+        sh_path = PACKAGE_ROOT / "install.sh"
+        source = sh_path.read_text()
+        assert "ROCm" in source
+
+    def test_rocm72_capped_to_71(self):
+        """ROCm 7.2+ should fall back to rocm7.1 index."""
+        sh_path = PACKAGE_ROOT / "install.sh"
+        source = sh_path.read_text()
+        assert 'echo "$_base/rocm7.1"' in source  # fallback for unknown versions
+        # Allowlisted versions should pass through directly
+        assert "rocm6.*" in source
+        assert "rocm7.0" in source
+        assert "rocm7.1" in source
+
+    def test_rocm_tag_validation_guard_exists(self):
+        """install.sh should validate _rocm_tag with a case guard."""
+        sh_path = PACKAGE_ROOT / "install.sh"
+        source = sh_path.read_text()
+        assert "rocm[1-9]*.[0-9]*)" in source
+        assert '_rocm_tag=""' in source  # rejection path
+
+    def test_dpkg_epoch_handling(self):
+        """install.sh should strip Debian epoch prefix from dpkg-query output."""
+        sh_path = PACKAGE_ROOT / "install.sh"
+        source = sh_path.read_text()
+        assert "sed 's/^[0-9]*://' " in source or "sed 's/^[0-9]*://'" in source
+
+    def test_no_double_bracket_in_rocm_block(self):
+        """ROCm detection block should not use [[ ]] (bash-only, not POSIX).
+        Note: [[:space:]], [[:digit:]] etc. are valid POSIX character classes, not bash [[ ]]."""
+        sh_path = PACKAGE_ROOT / "install.sh"
+        source = sh_path.read_text()
+        func_start = source.find("get_torch_index_url()")
+        func_end = source.find("\n}", func_start)
+        func_body = source[func_start:func_end]
+        import re
+
+        for i, line in enumerate(func_body.splitlines(), 1):
+            stripped = line.lstrip()
+            if stripped.startswith("#"):
+                continue
+            # Remove POSIX character classes [[:foo:]] before checking for [[ ]]
+            cleaned = re.sub(r"\[\[:[a-z]+:\]\]", "", line)
+            assert (
+                "[[" not in cleaned
+            ), f"get_torch_index_url line {i} uses non-POSIX [["
+
+    def test_no_arithmetic_expansion_in_rocm_block(self):
+        """ROCm detection block should not use (( )) (bash-only)."""
+        sh_path = PACKAGE_ROOT / "install.sh"
+        source = sh_path.read_text()
+        func_start = source.find("get_torch_index_url()")
+        func_end = source.find("\n}", func_start)
+        func_body = source[func_start:func_end]
+        for i, line in enumerate(func_body.splitlines(), 1):
+            stripped = line.lstrip()
+            if stripped.startswith("#"):
+                continue
+            assert (
+                "((" not in line or "))" not in line or "$(()" in line
+            ), f"get_torch_index_url line {i} may use non-POSIX (( ))"
+
+    def test_macos_returns_cpu_before_rocm_check(self):
+        """macOS should return CPU immediately (before any ROCm check)."""
+        sh_path = PACKAGE_ROOT / "install.sh"
+        source = sh_path.read_text()
+        func_start = source.find("get_torch_index_url()")
+        func_body = source[func_start:]
+        darwin_pos = func_body.find("Darwin")
+        rocm_pos = func_body.find("amd-smi")
+        assert darwin_pos < rocm_pos, "macOS check should come before ROCm detection"
+
+
+# =============================================================================
+# TEST: Live regression on current host (NVIDIA B200 expected)
+# =============================================================================
+
+
+class TestLiveRegression:
+    """Live checks that run on the actual host -- skip if no NVIDIA GPU."""
+
+    def test_get_torch_index_url_returns_cuda_on_nvidia(self):
+        """On an NVIDIA machine, get_torch_index_url should return a CUDA URL."""
+        import shutil
+
+        if not shutil.which("nvidia-smi"):
+            pytest.skip("No nvidia-smi available")
+        sh_path = PACKAGE_ROOT / "install.sh"
+        # Extract just the function (don't source the whole installer)
+        result = subprocess.run(
+            [
+                "bash",
+                "-c",
+                f"eval \"$(sed -n '/^get_torch_index_url()/,/^}}/p' '{sh_path}')\"; "
+                "get_torch_index_url",
+            ],
+            capture_output = True,
+            text = True,
+            timeout = 30,
+        )
+        if result.returncode != 0:
+            pytest.skip("Could not extract get_torch_index_url for live test")
+        url = result.stdout.strip()
+        assert "cu1" in url or "cuda" in url.lower(), f"Expected CUDA URL, got: {url}"
+
+
+# =============================================================================
+# TEST: worker.py -- ROCm Mamba/SSM source build path
+# =============================================================================
+
+# Load worker.py module
+_WORKER_PATH = PACKAGE_ROOT / "studio" / "backend" / "core" / "training" / "worker.py"
+
+
+class TestWorkerRocmMambaSsm:
+    """Verify worker.py Mamba/SSM install logic on ROCm."""
+
+    def test_probe_returns_hip_version_field(self):
+        """_probe_causal_conv1d_env probe script should include hip_version."""
+        source = _WORKER_PATH.read_text()
+        assert "hip_version" in source
+
+    def test_probe_script_has_getattr_hip(self):
+        """Probe script should use getattr for torch.version.hip (safe on CUDA)."""
+        source = _WORKER_PATH.read_text()
+        assert "getattr(torch.version, 'hip', None)" in source
+
+    def test_direct_wheel_url_returns_none_without_cuda_major(self):
+        """_direct_wheel_url should return None when cuda_major is empty (ROCm)."""
+        # Load module for function access
+        _worker_spec = importlib.util.spec_from_file_location(
+            "test_worker", _WORKER_PATH
+        )
+        assert _worker_spec is not None and _worker_spec.loader is not None
+        worker_mod = importlib.util.module_from_spec(_worker_spec)
+
+        # Mock all the imports worker.py needs
+        sys.modules["structlog"] = MagicMock()
+        sys.modules["loggers"] = MagicMock()
+        sys.modules["loggers"].get_logger = MagicMock(return_value = MagicMock())
+        sys.modules["utils"] = MagicMock()
+        sys.modules["utils.hardware"] = MagicMock()
+
+        try:
+            _worker_spec.loader.exec_module(worker_mod)
+        except Exception:
+            pytest.skip("Could not load worker module in test environment")
+
+        env_rocm = {
+            "python_tag": "cp312",
+            "torch_mm": "2.6",
+            "cuda_major": "",
+            "hip_version": "7.1.12345",
+            "cxx11abi": "TRUE",
+        }
+        result = worker_mod._direct_wheel_url(
+            filename_prefix = "causal_conv1d",
+            package_version = "1.6.1",
+            release_tag = "v1.6.1.post4",
+            release_base_url = "https://github.com/Dao-AILab/causal-conv1d/releases/download",
+            env = env_rocm,
+        )
+        assert result is None
+
+    def test_hipcc_check_exists_in_source(self):
+        """worker.py should check for hipcc before ROCm source builds."""
+        source = _WORKER_PATH.read_text()
+        assert "hipcc" in source
+
+    def test_rocm_source_build_status_message(self):
+        """worker.py should send a specific status for ROCm source compilation."""
+        source = _WORKER_PATH.read_text()
+        assert "Compiling" in source and "from source for ROCm" in source
+
+    def test_rocm_build_failure_message(self):
+        """worker.py should send a clear error on ROCm build failure."""
+        source = _WORKER_PATH.read_text()
+        assert "Failed to compile" in source and "for ROCm" in source
+
+    def test_timeout_on_install(self):
+        """worker.py should have a timeout on pip install subprocess."""
+        source = _WORKER_PATH.read_text()
+        assert "TimeoutExpired" in source
+        assert "timeout" in source
+
+
+# =============================================================================
+# TEST: amd.py -- AMD GPU monitoring
+# =============================================================================
+
+
+class TestAmdGpuMonitoring:
+    """Verify amd.py module structure and mock behavior."""
+
+    def test_amd_py_exists(self):
+        """amd.py should exist in the hardware directory."""
+        amd_path = PACKAGE_ROOT / "studio" / "backend" / "utils" / "hardware" / "amd.py"
+        assert amd_path.exists()
+
+    def test_amd_py_has_required_functions(self):
+        """amd.py should export the same function signatures as nvidia.py."""
+        amd_path = PACKAGE_ROOT / "studio" / "backend" / "utils" / "hardware" / "amd.py"
+        source = amd_path.read_text()
+        assert "def get_physical_gpu_count" in source
+        assert "def get_primary_gpu_utilization" in source
+        assert "def get_visible_gpu_utilization" in source
+
+    def test_amd_smi_json_parsing(self):
+        """Verify _extract_gpu_metrics parses amd-smi JSON correctly."""
+        amd_path = PACKAGE_ROOT / "studio" / "backend" / "utils" / "hardware" / "amd.py"
+        _amd_spec = importlib.util.spec_from_file_location("test_amd", amd_path)
+        assert _amd_spec is not None and _amd_spec.loader is not None
+        amd_mod = importlib.util.module_from_spec(_amd_spec)
+
+        sys.modules["loggers"] = MagicMock()
+        sys.modules["loggers"].get_logger = MagicMock(return_value = MagicMock())
+
+        try:
+            _amd_spec.loader.exec_module(amd_mod)
+        except Exception:
+            pytest.skip("Could not load amd module in test environment")
+
+        # Simulate amd-smi metric JSON output
+        gpu_data = {
+            "usage": {"gfx_activity": "85"},
+            "temperature": {"edge": "72"},
+            "power": {
+                "current_socket_power": "200.5",
+                "power_cap": "300",
+            },
+            "vram": {
+                "vram_used": 8192,  # MB
+                "vram_total": 16384,  # MB
+            },
+        }
+        metrics = amd_mod._extract_gpu_metrics(gpu_data)
+        assert metrics["gpu_utilization_pct"] == 85.0
+        assert metrics["temperature_c"] == 72.0
+        assert metrics["power_draw_w"] == 200.5
+        assert metrics["power_limit_w"] == 300.0
+        assert metrics["vram_used_gb"] == round(8192 / 1024, 2)
+        assert metrics["vram_total_gb"] == round(16384 / 1024, 2)
+        assert metrics["vram_utilization_pct"] is not None
+        assert metrics["power_utilization_pct"] is not None
+
+    def test_amd_primary_gpu_with_mock(self):
+        """get_primary_gpu_utilization returns correct dict with mocked amd-smi."""
+        amd_path = PACKAGE_ROOT / "studio" / "backend" / "utils" / "hardware" / "amd.py"
+        _amd_spec = importlib.util.spec_from_file_location("test_amd2", amd_path)
+        assert _amd_spec is not None and _amd_spec.loader is not None
+        amd_mod = importlib.util.module_from_spec(_amd_spec)
+
+        sys.modules["loggers"] = MagicMock()
+        sys.modules["loggers"].get_logger = MagicMock(return_value = MagicMock())
+
+        try:
+            _amd_spec.loader.exec_module(amd_mod)
+        except Exception:
+            pytest.skip("Could not load amd module")
+
+        mock_json = json.dumps(
+            [
+                {
+                    "usage": {"gfx_activity": "50"},
+                    "temperature": {"edge": "65"},
+                    "power": {"current_socket_power": "150", "power_cap": "250"},
+                    "vram": {"vram_used": 4096, "vram_total": 16384},
+                }
+            ]
+        )
+        mock_result = MagicMock()
+        mock_result.returncode = 0
+        mock_result.stdout = mock_json
+
+        with patch.object(subprocess, "run", return_value = mock_result):
+            result = amd_mod.get_primary_gpu_utilization()
+        assert result["available"] is True
+        assert result["gpu_utilization_pct"] == 50.0
+        assert result["temperature_c"] == 65.0
+
+    def test_amd_smi_not_found_returns_unavailable(self):
+        """get_primary_gpu_utilization returns available=False when amd-smi is missing."""
+        amd_path = PACKAGE_ROOT / "studio" / "backend" / "utils" / "hardware" / "amd.py"
+        _amd_spec = importlib.util.spec_from_file_location("test_amd3", amd_path)
+        assert _amd_spec is not None and _amd_spec.loader is not None
+        amd_mod = importlib.util.module_from_spec(_amd_spec)
+
+        sys.modules["loggers"] = MagicMock()
+        sys.modules["loggers"].get_logger = MagicMock(return_value = MagicMock())
+
+        try:
+            _amd_spec.loader.exec_module(amd_mod)
+        except Exception:
+            pytest.skip("Could not load amd module")
+
+        with patch.object(subprocess, "run", side_effect = OSError("amd-smi not found")):
+            result = amd_mod.get_primary_gpu_utilization()
+        assert result["available"] is False
+
+    def test_amd_timeout_returns_unavailable(self):
+        """get_primary_gpu_utilization handles timeout gracefully."""
+        amd_path = PACKAGE_ROOT / "studio" / "backend" / "utils" / "hardware" / "amd.py"
+        _amd_spec = importlib.util.spec_from_file_location("test_amd4", amd_path)
+        assert _amd_spec is not None and _amd_spec.loader is not None
+        amd_mod = importlib.util.module_from_spec(_amd_spec)
+
+        sys.modules["loggers"] = MagicMock()
+        sys.modules["loggers"].get_logger = MagicMock(return_value = MagicMock())
+
+        try:
+            _amd_spec.loader.exec_module(amd_mod)
+        except Exception:
+            pytest.skip("Could not load amd module")
+
+        with patch.object(
+            subprocess,
+            "run",
+            side_effect = subprocess.TimeoutExpired("amd-smi", 5),
+        ):
+            result = amd_mod.get_primary_gpu_utilization()
+        assert result["available"] is False
+
+
+# =============================================================================
+# TEST: hardware.py -- IS_ROCM branching to amd.py
+# =============================================================================
+
+
+class TestHardwareAmdBranching:
+    """Verify hardware.py branches to amd.py when IS_ROCM is True."""
+
+    def test_hardware_imports_amd_module(self):
+        """hardware.py should import from amd module when IS_ROCM."""
+        hw_path = (
+            PACKAGE_ROOT / "studio" / "backend" / "utils" / "hardware" / "hardware.py"
+        )
+        source = hw_path.read_text()
+        assert "from . import amd" in source
+
+    def test_hardware_branches_on_is_rocm_for_utilization(self):
+        """get_gpu_utilization should check IS_ROCM before choosing backend."""
+        hw_path = (
+            PACKAGE_ROOT / "studio" / "backend" / "utils" / "hardware" / "hardware.py"
+        )
+        source = hw_path.read_text()
+        # Find the get_gpu_utilization function
+        func_start = source.find("def get_gpu_utilization")
+        func_body = source[func_start : source.find("\ndef ", func_start + 1)]
+        assert "IS_ROCM" in func_body
+        assert "amd.get_primary_gpu_utilization" in func_body
+
+    def test_hardware_branches_on_is_rocm_for_visible(self):
+        """get_visible_gpu_utilization should check IS_ROCM."""
+        hw_path = (
+            PACKAGE_ROOT / "studio" / "backend" / "utils" / "hardware" / "hardware.py"
+        )
+        source = hw_path.read_text()
+        func_start = source.find("def get_visible_gpu_utilization")
+        func_body = source[func_start : source.find("\ndef ", func_start + 1)]
+        assert "IS_ROCM" in func_body
+        assert "amd.get_visible_gpu_utilization" in func_body
+
+    def test_hardware_branches_on_is_rocm_for_physical_count(self):
+        """get_physical_gpu_count should try amd.py when IS_ROCM."""
+        hw_path = (
+            PACKAGE_ROOT / "studio" / "backend" / "utils" / "hardware" / "hardware.py"
+        )
+        source = hw_path.read_text()
+        func_start = source.find("def get_physical_gpu_count")
+        func_body = source[func_start : source.find("\ndef ", func_start + 1)]
+        assert "IS_ROCM" in func_body
+        assert "amd.get_physical_gpu_count" in func_body
+
+
+# =============================================================================
+# TEST: install_python_stack.py -- Windows AMD warning
+# =============================================================================
+
+
+class TestWindowsRocmWarning:
+    """Verify Windows AMD GPU detection and warning message."""
+
+    def test_windows_amd_warning_in_source(self):
+        """install_python_stack.py should warn Windows AMD users."""
+        source = _STACK_PATH.read_text()
+        assert "AMD GPU detected on Windows" in source
+
+    def test_windows_amd_warning_checks_hipinfo_or_amdsmi(self):
+        """Warning should check for hipinfo or amd-smi."""
+        source = _STACK_PATH.read_text()
+        assert "hipinfo" in source
+        assert "amd-smi" in source
+
+    def test_windows_amd_warning_has_docs_link(self):
+        """Warning should include AMD docs link."""
+        source = _STACK_PATH.read_text()
+        assert "docs.unsloth.ai/get-started/install-and-update/amd" in source
+
+
+# =============================================================================
+# TEST: unsloth/kernels/utils.py -- is_rdna() expansion
+# =============================================================================
+
+
+class TestIsRdnaExpansion:
+    """Verify is_rdna() covers RDNA2, RDNA3, RDNA3.5, RDNA4 architectures."""
+
+    def test_is_rdna_source_has_rdna2(self):
+        """is_rdna() should include RDNA2 architectures."""
+        utils_path = PACKAGE_ROOT / "unsloth" / "kernels" / "utils.py"
+        source = utils_path.read_text()
+        func_start = source.find("def is_rdna()")
+        func_body = source[func_start : source.find("\ndef ", func_start + 1)]
+        assert "gfx1030" in func_body
+        assert "gfx1031" in func_body
+        assert "gfx1032" in func_body
+        assert "gfx1033" in func_body
+        assert "gfx1034" in func_body
+        assert "gfx1035" in func_body
+        assert "gfx1036" in func_body
+
+    def test_is_rdna_source_has_rdna3(self):
+        """is_rdna() should include RDNA3 architectures."""
+        utils_path = PACKAGE_ROOT / "unsloth" / "kernels" / "utils.py"
+        source = utils_path.read_text()
+        func_start = source.find("def is_rdna()")
+        func_body = source[func_start : source.find("\ndef ", func_start + 1)]
+        assert "gfx1100" in func_body
+        assert "gfx1101" in func_body
+        assert "gfx1102" in func_body
+        assert "gfx1103" in func_body
+
+    def test_is_rdna_source_has_rdna35(self):
+        """is_rdna() should include RDNA3.5 architectures."""
+        utils_path = PACKAGE_ROOT / "unsloth" / "kernels" / "utils.py"
+        source = utils_path.read_text()
+        func_start = source.find("def is_rdna()")
+        func_body = source[func_start : source.find("\ndef ", func_start + 1)]
+        assert "gfx1150" in func_body
+        assert "gfx1151" in func_body
+        assert "gfx1152" in func_body
+
+    def test_is_rdna_source_has_rdna4(self):
+        """is_rdna() should include RDNA4 architectures."""
+        utils_path = PACKAGE_ROOT / "unsloth" / "kernels" / "utils.py"
+        source = utils_path.read_text()
+        func_start = source.find("def is_rdna()")
+        func_body = source[func_start : source.find("\ndef ", func_start + 1)]
+        assert "gfx1200" in func_body
+        assert "gfx1201" in func_body
+
+    def test_is_cdna_not_changed(self):
+        """is_cdna() should remain unchanged (no RDNA architectures added)."""
+        utils_path = PACKAGE_ROOT / "unsloth" / "kernels" / "utils.py"
+        source = utils_path.read_text()
+        func_start = source.find("def is_cdna()")
+        func_body = source[func_start : source.find("\ndef ", func_start + 1)]
+        assert "gfx940" in func_body
+        assert "gfx941" in func_body
+        assert "gfx942" in func_body
+        assert "gfx950" in func_body
+        # RDNA architectures should NOT be in is_cdna
+        assert "gfx1030" not in func_body
+        assert "gfx1100" not in func_body
+
+
+if __name__ == "__main__":
+    pytest.main([__file__, "-v"])

From db3b3a4d9bf9bfe5e44819deb4978b3b1c162836 Mon Sep 17 00:00:00 2001
From: Manan Shah <52329525+Manan17@users.noreply.github.com>
Date: Mon, 13 Apr 2026 01:11:32 -0500
Subject: [PATCH 073/216] updating gemma4 script (#4992)

* updating gemma4 script

* show errors
---
 install_gemma4_mlx.sh | 66 +++++++++++--------------------------------
 1 file changed, 16 insertions(+), 50 deletions(-)

diff --git a/install_gemma4_mlx.sh b/install_gemma4_mlx.sh
index b653af9154..26415735b8 100755
--- a/install_gemma4_mlx.sh
+++ b/install_gemma4_mlx.sh
@@ -9,9 +9,7 @@ set -e
 #
 # This script:
 #   1. Creates a Python virtual environment
-#   2. Installs uv, mlx, mlx-lm, transformers
-#   3. Downloads gemma4.py and gemma4_text.py from unsloth repo
-#   4. Installs them into mlx-lm's models directory
+#   2. Installs uv, mlx-vlm, transformers
 # ============================================================
 
 # ── Output style (inspired by unsloth/install.sh) ─────────────
@@ -123,13 +121,9 @@ fi
 
 _VENV_PY="$VENV_DIR/bin/python"
 
-# ── Repo config ──────────────────────────────────────────────
-BRANCH="fix/ui-fix"
-REPO_URL="https://raw.githubusercontent.com/unslothai/unsloth/refs/heads/${BRANCH}"
-
 # ── Install dependencies ──────────────────────────────────────
-step "install" "installing mlx, mlx-lm..."
-uv pip install --python "$_VENV_PY" -q mlx mlx-lm 2>/dev/null
+step "install" "installing mlx-vlm..."
+uv pip install --python "$_VENV_PY" -q mlx-vlm
 substep "done"
 
 step "install" "installing transformers>=5.5.0..."
@@ -145,40 +139,11 @@ else
     fi
 fi
 
-# ── Find mlx-lm models directory ─────────────────────────────
-MLX_MODELS=$("$_VENV_PY" -c "import mlx_lm; print(mlx_lm.__path__[0])")/models
-step "models dir" "$MLX_MODELS"
-
-# ── Download and install Gemma 4 model files ──────────────────
-
-step "download" "installing Gemma 4 model files..."
-
-_install_model_file() {
-    _fname="$1"
-    if curl -fsSL "${REPO_URL}/unsloth/models/${_fname}" -o "${MLX_MODELS}/${_fname}" 2>/dev/null; then
-        substep "downloaded ${_fname} from branch ${BRANCH}"
-    elif [ -f "./${_fname}" ]; then
-        substep "using local ./${_fname}"
-        cp "./${_fname}" "${MLX_MODELS}/${_fname}"
-    else
-        fail "Could not install ${_fname}. Tried:
-                    1) ${REPO_URL}/unsloth/models/${_fname}
-                    2) Local file ./${_fname}
-
-                    To fix, download the file manually and place it in the current directory,
-                    then re-run this script."
-    fi
-}
-
-_install_model_file "gemma4.py"
-_install_model_file "gemma4_text.py"
-
-# Verify files were installed correctly
-if "$_VENV_PY" -c "from mlx_lm.models.gemma4_text import ProportionalRoPE" 2>/dev/null; then
-    substep "model files verified"
+# ── Verify installation ──────────────────────────────────────
+if "$_VENV_PY" -c "import mlx_vlm"; then
+    substep "mlx-vlm verified"
 else
-    fail "Model files installed but verification failed (ProportionalRoPE import error).
-                    Try manually from: https://github.com/unslothai/unsloth/tree/feature/${BRANCH}"
+    fail "Installation verification failed."
 fi
 
 # ── Done ──────────────────────────────────────────────────────
@@ -186,18 +151,19 @@ echo ""
 printf "  ${C_TITLE}%s${C_RST}\n" "Gemma 4 MLX installed!"
 printf "  ${C_DIM}%s${C_RST}\n" "$RULE"
 echo ""
-step "available models" "unsloth/gemma-4-E2B-it-UD-MLX-4bit (/BF16)"
-substep "unsloth/gemma-4-E4B-it-UD-MLX-4bit (/BF16)"
+step "available models" "unsloth/gemma-4-E2B-it-UD-MLX-4bit"
+substep "unsloth/gemma-4-E4B-it-UD-MLX-4bit"
+substep "unsloth/gemma-4-26b-a4b-it-UD-MLX-4bit"
+substep "unsloth/gemma-4-31b-it-UD-MLX-4bit"
 echo ""
 step "venv activate" "source ${VENV_DIR}/bin/activate"
 echo ""
-step "quick start" "python -m mlx_lm chat --model unsloth/gemma-4-E2B-it-UD-MLX-4bit --max-tokens 200"
+step "text chat" "python -m mlx_vlm.chat --model unsloth/gemma-4-E2B-it-UD-MLX-4bit"
 echo ""
-step "python API" "from mlx_lm import load, generate"
-substep "model, tokenizer = load('unsloth/gemma-4-E2B-it-UD-MLX-4bit')"
-substep "messages = [{'role': 'user', 'content': 'Hello!'}]"
-substep "prompt = tokenizer.apply_chat_template(messages, tokenize=False, add_generation_prompt=True)"
-substep "print(generate(model, tokenizer, prompt=prompt, max_tokens=200))"
+step "vision chat" "python -m mlx_vlm.chat --model unsloth/gemma-4-31b-it-UD-MLX-4bit"
+substep "Use /image path/to/image.jpg to load an image"
+echo ""
+step "gradio UI" "python -m mlx_vlm.chat_ui --model unsloth/gemma-4-31b-it-UD-MLX-4bit"
 echo ""
 printf "  ${C_DIM}%s${C_RST}\n" "$RULE"
 echo ""

From 80c12ff1a601a987115d1dac06b71026fdcadbf2 Mon Sep 17 00:00:00 2001
From: Manan Shah <52329525+Manan17@users.noreply.github.com>
Date: Mon, 13 Apr 2026 01:41:15 -0500
Subject: [PATCH 074/216] Move gemma4 script (#4994)

* updating gemma4 script

* moving gemma4 script to scripts folder
---
 install_gemma4_mlx.sh => scripts/install_gemma4_mlx.sh | 0
 1 file changed, 0 insertions(+), 0 deletions(-)
 rename install_gemma4_mlx.sh => scripts/install_gemma4_mlx.sh (100%)

diff --git a/install_gemma4_mlx.sh b/scripts/install_gemma4_mlx.sh
similarity index 100%
rename from install_gemma4_mlx.sh
rename to scripts/install_gemma4_mlx.sh

From 3b092bcd460a3a69d8763b8b2a9b687271e1dfa3 Mon Sep 17 00:00:00 2001
From: AdamPlatin123 <154926636+AdamPlatin123@users.noreply.github.com>
Date: Mon, 13 Apr 2026 16:38:00 +0800
Subject: [PATCH 075/216] fix(studio): prevent route transition DOM duplication
 via AnimatePresence (#4987)
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

Add mode="wait" and exit={{ opacity: 0 }} to the root AnimatePresence
wrapper so outgoing routes fully unmount before incoming routes render.
Without this, rapid navigation between Studio/Export/Recipes/Chat caused
pages to stack (2x–3x duplication).

Co-authored-by: AdamPlatin123 <AdamPlatin123@users.noreply.github.com>
Co-authored-by: Wasim Yousef Said <wasimysdev@gmail.com>
---
 studio/frontend/src/app/routes/__root.tsx | 3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)

diff --git a/studio/frontend/src/app/routes/__root.tsx b/studio/frontend/src/app/routes/__root.tsx
index d7780c6743..bac9205303 100644
--- a/studio/frontend/src/app/routes/__root.tsx
+++ b/studio/frontend/src/app/routes/__root.tsx
@@ -40,11 +40,12 @@ function RootLayout() {
   return (
     <AppProvider>
       {!hideNavbar && <Navbar />}
-      <AnimatePresence initial={false}>
+      <AnimatePresence initial={false} mode="wait">
         <motion.div
           key={pathname}
           initial={{ opacity: 0 }}
           animate={{ opacity: 1 }}
+          exit={{ opacity: 0 }}
           transition={{ duration: 0.15 }}
           className="flex-1"
         >

From 21a7895959bb18bf9d027cfd3b6907b714c55263 Mon Sep 17 00:00:00 2001
From: Lee Jackson <130007945+Imagineer99@users.noreply.github.com>
Date: Mon, 13 Apr 2026 15:42:33 +0100
Subject: [PATCH 076/216] Studio: Prompt manager, message deletion, and chat UI
 improvements (#4938)

* feat(chat): code block styling, delete with Dexie sync, settings sheet polish

* style: config save/delete padding fix

* fix(studio): centralize dark code-block surface and optimize message sync writes

* style: config padding/alignment polish

* fix(studio): upsert custom presets without implicit rename-delete

* fix settings sheet save state polish

* fix settings sheet button widths

* fix chat settings presets

* fix chat delete sync

* fix chat trust remote code flow

---------

Co-authored-by: shine1i <wasimysdev@gmail.com>
---
 studio/backend/models/inference.py            |  12 +
 studio/backend/routes/inference.py            |  27 +
 .../components/assistant-ui/markdown-text.tsx |   2 +-
 .../assistant-ui/model-selector/pickers.tsx   |   6 +-
 .../src/components/assistant-ui/thread.tsx    |  40 +
 .../src/features/chat/api/chat-adapter.ts     | 115 ++-
 .../src/features/chat/chat-settings-sheet.tsx | 758 ++++++++++++++----
 .../chat/hooks/use-chat-model-runtime.ts      |  32 +-
 .../src/features/chat/runtime-provider.tsx    |   3 +
 .../src/features/chat/shared-composer.tsx     |  29 +-
 .../chat/stores/chat-runtime-store.ts         |   9 +-
 .../frontend/src/features/chat/types/api.ts   |   3 +
 .../chat/utils/delete-thread-message.ts       | 119 +++
 studio/frontend/src/index.css                 |   8 +
 14 files changed, 971 insertions(+), 192 deletions(-)
 create mode 100644 studio/frontend/src/features/chat/utils/delete-thread-message.ts

diff --git a/studio/backend/models/inference.py b/studio/backend/models/inference.py
index cf08ecbc12..8a27f4f630 100644
--- a/studio/backend/models/inference.py
+++ b/studio/backend/models/inference.py
@@ -94,6 +94,10 @@ class ValidateModelResponse(BaseModel):
     is_gguf: bool = Field(False, description = "Whether this is a GGUF model (llama.cpp)")
     is_lora: bool = Field(False, description = "Whether this is a LoRA adapter")
     is_vision: bool = Field(False, description = "Whether this is a vision-capable model")
+    requires_trust_remote_code: bool = Field(
+        False,
+        description = "Whether the model defaults require trust_remote_code to be enabled for loading.",
+    )
 
 
 class GenerateRequest(BaseModel):
@@ -137,6 +141,10 @@ class LoadResponse(BaseModel):
     inference: dict = Field(
         ..., description = "Inference parameters (temperature, top_p, top_k, min_p)"
     )
+    requires_trust_remote_code: bool = Field(
+        False,
+        description = "Whether the model defaults require trust_remote_code to be enabled for loading.",
+    )
     context_length: Optional[int] = Field(
         None, description = "Model's native context length (from GGUF metadata)"
     )
@@ -213,6 +221,10 @@ class InferenceStatusResponse(BaseModel):
     inference: Optional[Dict[str, Any]] = Field(
         None, description = "Recommended inference parameters for the active model"
     )
+    requires_trust_remote_code: bool = Field(
+        False,
+        description = "Whether the active model requires trust_remote_code to be enabled for loading.",
+    )
     supports_reasoning: bool = Field(
         False, description = "Whether the active model supports reasoning/thinking mode"
     )
diff --git a/studio/backend/routes/inference.py b/studio/backend/routes/inference.py
index 30ff7da49c..7952d3a30b 100644
--- a/studio/backend/routes/inference.py
+++ b/studio/backend/routes/inference.py
@@ -173,6 +173,9 @@ async def load_model(
                     if _gguf_audio
                     else False,
                     inference = inference_config,
+                    requires_trust_remote_code = bool(
+                        inference_config.get("trust_remote_code", False)
+                    ),
                     context_length = llama_backend.context_length,
                     max_context_length = llama_backend.max_context_length,
                     native_context_length = llama_backend.native_context_length,
@@ -210,6 +213,9 @@ async def load_model(
                     audio_type = _model_info.get("audio_type"),
                     has_audio_input = _model_info.get("has_audio_input", False),
                     inference = inference_config,
+                    requires_trust_remote_code = bool(
+                        inference_config.get("trust_remote_code", False)
+                    ),
                     chat_template = _chat_template,
                 )
 
@@ -312,6 +318,9 @@ async def load_model(
                 audio_type = _gguf_audio,
                 has_audio_input = is_audio_input_type(_gguf_audio),
                 inference = inference_config,
+                requires_trust_remote_code = bool(
+                    inference_config.get("trust_remote_code", False)
+                ),
                 context_length = llama_backend.context_length,
                 max_context_length = llama_backend.max_context_length,
                 native_context_length = llama_backend.native_context_length,
@@ -450,6 +459,9 @@ async def load_model(
             audio_type = config.audio_type,
             has_audio_input = config.has_audio_input,
             inference = inference_config,
+            requires_trust_remote_code = bool(
+                inference_config.get("trust_remote_code", False)
+            ),
             chat_template = _chat_template,
         )
 
@@ -505,6 +517,9 @@ async def validate_model(
             is_gguf = getattr(config, "is_gguf", False),
             is_lora = getattr(config, "is_lora", False),
             is_vision = getattr(config, "is_vision", False),
+            requires_trust_remote_code = bool(
+                load_inference_config(config.identifier).get("trust_remote_code", False)
+            ),
         )
 
     except HTTPException:
@@ -650,6 +665,9 @@ async def get_status(
                 loading = [],
                 loaded = [_model_id],
                 inference = _inference_cfg,
+                requires_trust_remote_code = bool(
+                    (_inference_cfg or {}).get("trust_remote_code", False)
+                ),
                 supports_reasoning = llama_backend.supports_reasoning,
                 reasoning_always_on = llama_backend.reasoning_always_on,
                 supports_tools = llama_backend.supports_tools,
@@ -677,6 +695,11 @@ async def get_status(
         supports_reasoning = False
         if backend.active_model_name and hasattr(backend, "_is_gpt_oss_model"):
             supports_reasoning = backend._is_gpt_oss_model()
+        inference_config = (
+            load_inference_config(backend.active_model_name)
+            if backend.active_model_name
+            else None
+        )
 
         return InferenceStatusResponse(
             active_model = backend.active_model_name,
@@ -687,6 +710,10 @@ async def get_status(
             has_audio_input = has_audio_input,
             loading = list(getattr(backend, "loading_models", set())),
             loaded = list(backend.models.keys()),
+            inference = inference_config,
+            requires_trust_remote_code = bool(
+                (inference_config or {}).get("trust_remote_code", False)
+            ),
             supports_reasoning = supports_reasoning,
         )
 
diff --git a/studio/frontend/src/components/assistant-ui/markdown-text.tsx b/studio/frontend/src/components/assistant-ui/markdown-text.tsx
index 91ef78fcf9..c7974db365 100644
--- a/studio/frontend/src/components/assistant-ui/markdown-text.tsx
+++ b/studio/frontend/src/components/assistant-ui/markdown-text.tsx
@@ -41,7 +41,7 @@ const COPY_RESET_MS = 2000;
 const MERMAID_SOURCE_RE = /```mermaid\s*([\s\S]*?)```/i;
 const CODE_FENCE_RE = /^```([^\r\n`]*)\r?\n([\s\S]*?)\r?\n?```$/;
 const ACTION_PANEL_CLASS =
-  "pointer-events-auto flex shrink-0 items-center gap-2 rounded-md border border-sidebar bg-sidebar/80 px-1.5 py-1 supports-[backdrop-filter]:bg-sidebar/70 supports-[backdrop-filter]:backdrop-blur";
+  "pointer-events-auto flex shrink-0 items-center gap-2 rounded-md border border-sidebar bg-sidebar/80 px-1.5 py-1 supports-[backdrop-filter]:bg-sidebar/70 supports-[backdrop-filter]:backdrop-blur dark:border-white/10 dark:bg-code-block dark:supports-[backdrop-filter]:bg-code-block";
 const ACTION_BUTTON_CLASS =
   "cursor-pointer p-1 text-muted-foreground transition-all hover:text-foreground disabled:cursor-not-allowed disabled:opacity-50";
 
diff --git a/studio/frontend/src/components/assistant-ui/model-selector/pickers.tsx b/studio/frontend/src/components/assistant-ui/model-selector/pickers.tsx
index 74ca2542d4..87d59dfaa3 100644
--- a/studio/frontend/src/components/assistant-ui/model-selector/pickers.tsx
+++ b/studio/frontend/src/components/assistant-ui/model-selector/pickers.tsx
@@ -841,7 +841,7 @@ export function HubModelPicker({
             (cachedGguf.length > 0 ||
               (!chatOnly && cachedModels.length > 0)) ? (
             <>
-              <ListLabel>{"\uD83E\uDDA5"} Downloaded</ListLabel>
+              <ListLabel>Downloaded</ListLabel>
               {cachedGguf.map((c) => (
                 <div key={c.repo_id}>
                   <ModelRow
@@ -1076,7 +1076,7 @@ export function HubModelPicker({
 
           {!showHfSection && cachedReady ? (
             <>
-              <ListLabel>{"\uD83E\uDDA5"} Recommended</ListLabel>
+              <ListLabel>Recommended</ListLabel>
               {visibleRecommendedIds.length === 0 ? (
                 <div className="px-2.5 py-2 text-xs text-muted-foreground">
                   No default models.
@@ -1128,7 +1128,7 @@ export function HubModelPicker({
 
           {showHfSection && filteredRecommendedIds.length > 0 ? (
             <>
-              <ListLabel>{"\uD83E\uDDA5"} Recommended</ListLabel>
+              <ListLabel>Recommended</ListLabel>
               {filteredRecommendedIds.map((id) => {
                 const vram = recommendedVramMap.get(id);
                 return (
diff --git a/studio/frontend/src/components/assistant-ui/thread.tsx b/studio/frontend/src/components/assistant-ui/thread.tsx
index 0d99654b89..8f41987fbf 100644
--- a/studio/frontend/src/components/assistant-ui/thread.tsx
+++ b/studio/frontend/src/components/assistant-ui/thread.tsx
@@ -56,9 +56,12 @@ import {
   RefreshCwIcon,
   SquareIcon,
   TerminalIcon,
+  Trash2Icon,
   XIcon,
 } from "lucide-react";
 import { type FC, useCallback, useEffect, useRef, useState } from "react";
+import { toast } from "sonner";
+import { deleteThreadMessage } from "@/features/chat/utils/delete-thread-message";
 import { useChatRuntimeStore } from "@/features/chat/stores/chat-runtime-store";
 
 export const Thread: FC<{ hideComposer?: boolean; hideWelcome?: boolean }> = ({
@@ -635,6 +638,41 @@ const AssistantMessage: FC = () => {
 
 const COPY_RESET_MS = 2000;
 
+const DeleteMessageButton: FC = () => {
+  const aui = useAui();
+  const messageId = useAuiState(({ message }) => message.id);
+  const isRunning = useAuiState(({ thread }) => thread.isRunning);
+
+  const handleDelete = async () => {
+    const remoteId = aui.threadListItem().getState().remoteId;
+    const thread = aui.thread();
+    try {
+      await deleteThreadMessage({
+        thread: {
+          export: () => thread.export(),
+          import: (data) => thread.import(data),
+        },
+        messageId,
+        remoteId,
+      });
+    } catch (error) {
+      console.error("Failed to delete message", error);
+      toast.error("Failed to delete message");
+    }
+  };
+
+  return (
+    <TooltipIconButton
+      tooltip="Delete message"
+      disabled={isRunning}
+      onClick={handleDelete}
+      className="text-muted-foreground hover:text-destructive"
+    >
+      <Trash2Icon className="size-4" />
+    </TooltipIconButton>
+  );
+};
+
 const CopyButton: FC = () => {
   const aui = useAui();
   const [copied, setCopied] = useState(false);
@@ -673,6 +711,7 @@ const AssistantActionBar: FC = () => {
           <RefreshCwIcon />
         </TooltipIconButton>
       </ActionBarPrimitive.Reload>
+      <DeleteMessageButton />
       <MessageTiming side="top" />
       <ActionBarMorePrimitive.Root>
         <ActionBarMorePrimitive.Trigger asChild={true}>
@@ -748,6 +787,7 @@ const UserActionBar: FC = () => {
           <PencilIcon />
         </TooltipIconButton>
       </ActionBarPrimitive.Edit>
+      <DeleteMessageButton />
     </ActionBarPrimitive.Root>
   );
 };
diff --git a/studio/frontend/src/features/chat/api/chat-adapter.ts b/studio/frontend/src/features/chat/api/chat-adapter.ts
index 3d1bce2905..76dbbf0bfb 100644
--- a/studio/frontend/src/features/chat/api/chat-adapter.ts
+++ b/studio/frontend/src/features/chat/api/chat-adapter.ts
@@ -11,6 +11,7 @@ import {
   listGgufVariants,
   loadModel,
   streamChatCompletions,
+  validateModel,
 } from "./chat-api";
 import { db } from "../db";
 import { useChatRuntimeStore } from "../stores/chat-runtime-store";
@@ -252,13 +253,39 @@ function waitForModelReady(abortSignal?: AbortSignal): Promise<void> {
  * without selecting one. Prefers GGUF (picks smallest cached variant),
  * falls back to smallest cached safetensors model.
  */
-async function autoLoadSmallestModel(): Promise<boolean> {
-  const hfToken = useChatRuntimeStore.getState().hfToken || null;
+async function autoLoadSmallestModel(): Promise<{
+  loaded: boolean;
+  blockedByTrustRemoteCode: boolean;
+}> {
+  const store = useChatRuntimeStore.getState();
+  const hfToken = store.hfToken || null;
+  const trustRemoteCode = store.params.trustRemoteCode ?? false;
   const toastId = toast("Loading a model…", {
     description: "Auto-selecting the smallest downloaded model.",
     duration: 5000,
     closeButton: true,
   });
+  let blockedByTrustRemoteCode = false;
+  let hadNonTrustFailure = false;
+
+  async function canAutoLoad(payload: {
+    model_path: string;
+    max_seq_length: number;
+    is_lora: boolean;
+    gguf_variant?: string | null;
+  }): Promise<boolean> {
+    const validation = await validateModel({
+      ...payload,
+      hf_token: hfToken,
+      load_in_4bit: true,
+      trust_remote_code: trustRemoteCode,
+    });
+    if (validation.requires_trust_remote_code && !trustRemoteCode) {
+      blockedByTrustRemoteCode = true;
+      return false;
+    }
+    return true;
+  }
   try {
     const [ggufRepos, modelRepos] = await Promise.all([
       listCachedGguf().catch(() => []),
@@ -277,6 +304,16 @@ async function autoLoadSmallestModel(): Promise<boolean> {
             .sort((a, b) => a.size_bytes - b.size_bytes);
           if (downloaded.length > 0) {
             const variant = downloaded[0];
+            if (
+              !(await canAutoLoad({
+                model_path: repo.repo_id,
+                max_seq_length: 0,
+                is_lora: false,
+                gguf_variant: variant.quant,
+              }))
+            ) {
+              continue;
+            }
             const loadResp = await loadModel({
               model_path: repo.repo_id,
               hf_token: hfToken,
@@ -284,10 +321,13 @@ async function autoLoadSmallestModel(): Promise<boolean> {
               load_in_4bit: true,
               is_lora: false,
               gguf_variant: variant.quant,
-              trust_remote_code: false,
+              trust_remote_code: trustRemoteCode,
             });
             useChatRuntimeStore.getState().setCheckpoint(repo.repo_id, variant.quant);
             const store = useChatRuntimeStore.getState();
+            store.setModelRequiresTrustRemoteCode(
+              loadResp.requires_trust_remote_code ?? false,
+            );
             store.setParams({ ...store.params, maxTokens: loadResp.context_length ?? 131072 });
             // Add model to store so the selector shows the name
             const autoModel: ChatModelSummary = {
@@ -319,9 +359,10 @@ async function autoLoadSmallestModel(): Promise<boolean> {
               chatTemplateOverride: null,
             });
             toast.success(`Loaded ${repo.repo_id} (${variant.quant})`, { id: toastId });
-            return true;
+            return { loaded: true, blockedByTrustRemoteCode: false };
           }
         } catch {
+          hadNonTrustFailure = true;
           continue;
         }
       }
@@ -332,6 +373,16 @@ async function autoLoadSmallestModel(): Promise<boolean> {
       const sorted = [...modelRepos].sort((a, b) => a.size_bytes - b.size_bytes);
       for (const repo of sorted) {
         try {
+          if (
+            !(await canAutoLoad({
+              model_path: repo.repo_id,
+              max_seq_length: 4096,
+              is_lora: false,
+              gguf_variant: null,
+            }))
+          ) {
+            continue;
+          }
           const sfLoadResp = await loadModel({
             model_path: repo.repo_id,
             hf_token: hfToken,
@@ -339,10 +390,13 @@ async function autoLoadSmallestModel(): Promise<boolean> {
             load_in_4bit: true,
             is_lora: false,
             gguf_variant: null,
-            trust_remote_code: false,
+            trust_remote_code: trustRemoteCode,
           });
           useChatRuntimeStore.getState().setCheckpoint(repo.repo_id);
           const store = useChatRuntimeStore.getState();
+          store.setModelRequiresTrustRemoteCode(
+            sfLoadResp.requires_trust_remote_code ?? false,
+          );
           store.setParams({ ...store.params, maxTokens: 4096 });
           const sfModel: ChatModelSummary = {
             id: repo.repo_id,
@@ -355,8 +409,9 @@ async function autoLoadSmallestModel(): Promise<boolean> {
             store.setModels([...store.models, sfModel]);
           }
           toast.success(`Loaded ${repo.repo_id}`, { id: toastId });
-          return true;
+          return { loaded: true, blockedByTrustRemoteCode: false };
         } catch {
+          hadNonTrustFailure = true;
           continue;
         }
       }
@@ -369,6 +424,17 @@ async function autoLoadSmallestModel(): Promise<boolean> {
       duration: 30000,
     });
     try {
+      if (
+        !(await canAutoLoad({
+          model_path: "unsloth/Qwen3.5-4B-GGUF",
+          max_seq_length: 0,
+          is_lora: false,
+          gguf_variant: "UD-Q4_K_XL",
+        }))
+      ) {
+        toast.dismiss(toastId);
+        return { loaded: false, blockedByTrustRemoteCode };
+      }
       const loadResp = await loadModel({
         model_path: "unsloth/Qwen3.5-4B-GGUF",
         hf_token: hfToken,
@@ -376,10 +442,13 @@ async function autoLoadSmallestModel(): Promise<boolean> {
         load_in_4bit: true,
         is_lora: false,
         gguf_variant: "UD-Q4_K_XL",
-        trust_remote_code: false,
+        trust_remote_code: trustRemoteCode,
       });
       useChatRuntimeStore.getState().setCheckpoint("unsloth/Qwen3.5-4B-GGUF", "UD-Q4_K_XL");
       const store = useChatRuntimeStore.getState();
+      store.setModelRequiresTrustRemoteCode(
+        loadResp.requires_trust_remote_code ?? false,
+      );
       store.setParams({ ...store.params, maxTokens: loadResp.context_length ?? 131072 });
       const defaultModel: ChatModelSummary = {
         id: "unsloth/Qwen3.5-4B-GGUF",
@@ -406,14 +475,24 @@ async function autoLoadSmallestModel(): Promise<boolean> {
         chatTemplateOverride: null,
       });
       toast.success("Loaded Qwen3.5-4B (UD-Q4_K_XL)", { id: toastId });
-      return true;
+      return { loaded: true, blockedByTrustRemoteCode: false };
     } catch {
       toast.dismiss(toastId);
-      return false;
+      hadNonTrustFailure = true;
+      return {
+        loaded: false,
+        blockedByTrustRemoteCode:
+          blockedByTrustRemoteCode && !hadNonTrustFailure,
+      };
     }
   } catch {
     toast.dismiss(toastId);
-    return false;
+    hadNonTrustFailure = true;
+    return {
+      loaded: false,
+      blockedByTrustRemoteCode:
+        blockedByTrustRemoteCode && !hadNonTrustFailure,
+    };
   }
 }
 
@@ -434,11 +513,19 @@ export function createOpenAIStreamAdapter(): ChatModelAdapter {
 
       if (!useChatRuntimeStore.getState().params.checkpoint) {
         // Auto-load the smallest downloaded model
-        const loaded = await autoLoadSmallestModel();
+        const { loaded, blockedByTrustRemoteCode } =
+          await autoLoadSmallestModel();
         if (!loaded) {
-          toast.error("No model loaded", {
-            description: "Pick a model in the top bar, then retry.",
-          });
+          toast.error(
+            blockedByTrustRemoteCode
+              ? "Enable custom code to auto-load this model"
+              : "No model loaded",
+            {
+              description: blockedByTrustRemoteCode
+                ? 'Turn on "Enable custom code" in Chat Settings, or pick another model in the top bar.'
+                : "Pick a model in the top bar, then retry.",
+            },
+          );
           throw new Error("Load a model first.");
         }
       }
diff --git a/studio/frontend/src/features/chat/chat-settings-sheet.tsx b/studio/frontend/src/features/chat/chat-settings-sheet.tsx
index 550df2bf7c..bcfd585214 100644
--- a/studio/frontend/src/features/chat/chat-settings-sheet.tsx
+++ b/studio/frontend/src/features/chat/chat-settings-sheet.tsx
@@ -1,6 +1,11 @@
 // SPDX-License-Identifier: AGPL-3.0-only
 // Copyright 2026-present the Unsloth AI Inc. team. All rights reserved. See /studio/LICENSE.AGPL-3.0
 
+import {
+  Alert,
+  AlertDescription,
+  AlertTitle,
+} from "@/components/ui/alert";
 import { Button } from "@/components/ui/button";
 import {
   Dialog,
@@ -10,7 +15,19 @@ import {
   DialogHeader,
   DialogTitle,
 } from "@/components/ui/dialog";
+import {
+  DropdownMenu,
+  DropdownMenuContent,
+  DropdownMenuItem,
+  DropdownMenuTrigger,
+} from "@/components/ui/dropdown-menu";
 import { Input } from "@/components/ui/input";
+import {
+  InputGroup,
+  InputGroupAddon,
+  InputGroupButton,
+  InputGroupInput,
+} from "@/components/ui/input-group";
 import {
   Select,
   SelectContent,
@@ -29,6 +46,7 @@ import { Slider } from "@/components/ui/slider";
 import { Switch } from "@/components/ui/switch";
 import { Textarea } from "@/components/ui/textarea";
 import { useIsMobile } from "@/hooks/use-mobile";
+import { cn } from "@/lib/utils";
 import {
   ArrowDown01Icon,
   CodeIcon,
@@ -43,7 +61,8 @@ import {
 import { HugeiconsIcon } from "@hugeicons/react";
 import { AnimatePresence, motion } from "motion/react";
 import type { ReactNode } from "react";
-import { useEffect, useMemo, useState } from "react";
+import { useEffect, useLayoutEffect, useMemo, useRef, useState } from "react";
+import { toast } from "sonner";
 import { useChatRuntimeStore } from "./stores/chat-runtime-store";
 import {
   DEFAULT_INFERENCE_PARAMS,
@@ -58,6 +77,11 @@ export interface Preset {
   params: InferenceParams;
 }
 
+interface LegacySystemPromptTemplate {
+  name: string;
+  content: string;
+}
+
 const BUILTIN_PRESETS: Preset[] = [
   { name: "Default", params: { ...defaultInferenceParams } },
   {
@@ -86,19 +110,134 @@ const BUILTIN_PRESETS: Preset[] = [
 
 const CHAT_PRESETS_KEY = "unsloth_chat_custom_presets";
 const CHAT_ACTIVE_PRESET_KEY = "unsloth_chat_active_preset";
+const LEGACY_CHAT_SYSTEM_PROMPTS_KEY = "unsloth_chat_system_prompts";
+const LEGACY_CHAT_SYSTEM_PROMPTS_MIGRATED_KEY =
+  "unsloth_chat_system_prompts_migrated";
 
 function canUseStorage(): boolean {
   return typeof window !== "undefined";
 }
 
+function getUniquePresetName(baseName: string, usedNames: Set<string>): string {
+  const normalizedBase = baseName.trim() || "Imported Prompt";
+  let nextName = normalizedBase;
+  let suffix = 2;
+  while (usedNames.has(nextName)) {
+    nextName = `${normalizedBase} ${suffix}`;
+    suffix += 1;
+  }
+  usedNames.add(nextName);
+  return nextName;
+}
+
+function migrateLegacySystemPromptTemplates(presets: Preset[]): Preset[] {
+  if (!canUseStorage()) return presets;
+  try {
+    const raw = localStorage.getItem(LEGACY_CHAT_SYSTEM_PROMPTS_KEY);
+    if (!raw) return presets;
+    if (localStorage.getItem(LEGACY_CHAT_SYSTEM_PROMPTS_MIGRATED_KEY) === raw) {
+      return presets;
+    }
+    let parsed: unknown;
+    try {
+      parsed = JSON.parse(raw) as unknown;
+    } catch {
+      localStorage.removeItem(LEGACY_CHAT_SYSTEM_PROMPTS_KEY);
+      localStorage.setItem(LEGACY_CHAT_SYSTEM_PROMPTS_MIGRATED_KEY, raw);
+      return presets;
+    }
+    if (!Array.isArray(parsed)) {
+      localStorage.removeItem(LEGACY_CHAT_SYSTEM_PROMPTS_KEY);
+      localStorage.setItem(LEGACY_CHAT_SYSTEM_PROMPTS_MIGRATED_KEY, raw);
+      return presets;
+    }
+    const usedNames = new Set([
+      ...BUILTIN_PRESETS.map((preset) => preset.name),
+      ...presets.map((preset) => preset.name),
+    ]);
+    const seenImportedConfigKeys = new Set(
+      [...BUILTIN_PRESETS, ...presets].map((preset) =>
+        JSON.stringify({
+          temperature: preset.params.temperature,
+          topP: preset.params.topP,
+          topK: preset.params.topK,
+          minP: preset.params.minP,
+          repetitionPenalty: preset.params.repetitionPenalty,
+          presencePenalty: preset.params.presencePenalty,
+          maxSeqLength: preset.params.maxSeqLength,
+          maxTokens: preset.params.maxTokens,
+          systemPrompt: preset.params.systemPrompt,
+          trustRemoteCode: preset.params.trustRemoteCode ?? false,
+        }),
+      ),
+    );
+    const importedPresets = parsed
+      .filter((item): item is LegacySystemPromptTemplate => {
+        if (!item || typeof item !== "object") return false;
+        const maybe = item as Partial<LegacySystemPromptTemplate>;
+        return (
+          typeof maybe.name === "string" && typeof maybe.content === "string"
+        );
+      })
+      .map((template) => ({
+        template,
+        importedParams: {
+          ...defaultInferenceParams,
+          systemPrompt: template.content,
+        },
+      }))
+      .filter(({ importedParams }) => {
+        const configKey = JSON.stringify({
+          temperature: importedParams.temperature,
+          topP: importedParams.topP,
+          topK: importedParams.topK,
+          minP: importedParams.minP,
+          repetitionPenalty: importedParams.repetitionPenalty,
+          presencePenalty: importedParams.presencePenalty,
+          maxSeqLength: importedParams.maxSeqLength,
+          maxTokens: importedParams.maxTokens,
+          systemPrompt: importedParams.systemPrompt,
+          trustRemoteCode: importedParams.trustRemoteCode ?? false,
+        });
+        if (seenImportedConfigKeys.has(configKey)) return false;
+        seenImportedConfigKeys.add(configKey);
+        return true;
+      })
+      .map(({ template, importedParams }) => ({
+        name: getUniquePresetName(`${template.name} Prompt`, usedNames),
+        params: importedParams,
+      }));
+    if (importedPresets.length === 0) {
+      localStorage.removeItem(LEGACY_CHAT_SYSTEM_PROMPTS_KEY);
+      localStorage.setItem(LEGACY_CHAT_SYSTEM_PROMPTS_MIGRATED_KEY, raw);
+      return presets;
+    }
+    const mergedPresets = [...presets, ...importedPresets];
+    localStorage.setItem(CHAT_PRESETS_KEY, JSON.stringify(mergedPresets));
+    try {
+      localStorage.setItem(LEGACY_CHAT_SYSTEM_PROMPTS_MIGRATED_KEY, raw);
+      localStorage.removeItem(LEGACY_CHAT_SYSTEM_PROMPTS_KEY);
+    } catch {
+      // ignore cleanup failure after successful import write
+    }
+    return mergedPresets;
+  } catch {
+    return presets;
+  }
+}
+
 function loadSavedCustomPresets(): Preset[] {
   if (!canUseStorage()) return [];
   try {
     const raw = localStorage.getItem(CHAT_PRESETS_KEY);
-    if (!raw) return [];
+    if (!raw) {
+      return migrateLegacySystemPromptTemplates([]);
+    }
     const parsed = JSON.parse(raw) as unknown;
-    if (!Array.isArray(parsed)) return [];
-    return parsed
+    if (!Array.isArray(parsed)) {
+      return migrateLegacySystemPromptTemplates([]);
+    }
+    const presets = parsed
       .filter((item): item is Preset => {
         if (!item || typeof item !== "object") return false;
         const maybe = item as Partial<Preset>;
@@ -111,13 +250,10 @@ function loadSavedCustomPresets(): Preset[] {
           ...preset.params,
         },
       }))
-      .filter(
-        (preset) =>
-          preset.name.length > 0 &&
-          !BUILTIN_PRESETS.some((builtin) => builtin.name === preset.name),
-      );
+      .filter((preset) => preset.name.length > 0);
+    return migrateLegacySystemPromptTemplates(presets);
   } catch {
-    return [];
+    return migrateLegacySystemPromptTemplates([]);
   }
 }
 
@@ -130,6 +266,82 @@ function loadSavedActivePreset(): string {
   }
 }
 
+type PresetSaveMode =
+  | "disabled"
+  | "overwrite-active"
+  | "overwrite-other"
+  | "create";
+
+interface PresetSaveState {
+  mode: PresetSaveMode;
+  canSubmit: boolean;
+  isSaveReady: boolean;
+  buttonLabel: string;
+  title: string;
+}
+
+function isSamePresetConfig(a: InferenceParams, b: InferenceParams): boolean {
+  return (
+    a.temperature === b.temperature &&
+    a.topP === b.topP &&
+    a.topK === b.topK &&
+    a.minP === b.minP &&
+    a.repetitionPenalty === b.repetitionPenalty &&
+    a.presencePenalty === b.presencePenalty &&
+    a.maxSeqLength === b.maxSeqLength &&
+    a.maxTokens === b.maxTokens &&
+    a.systemPrompt === b.systemPrompt &&
+    (a.trustRemoteCode ?? false) === (b.trustRemoteCode ?? false)
+  );
+}
+
+function getPresetSaveState({
+  rawName,
+  activePreset,
+  presets,
+  activePresetDirty,
+}: {
+  rawName: string;
+  activePreset: string;
+  presets: Preset[];
+  activePresetDirty: boolean;
+}): PresetSaveState {
+  const trimmedName = rawName.trim();
+  if (!trimmedName) {
+    return {
+      mode: "disabled",
+      canSubmit: false,
+      isSaveReady: false,
+      buttonLabel: "Save",
+      title: "Enter a preset name",
+    };
+  }
+
+  const matchingPreset = presets.find((preset) => preset.name === trimmedName);
+  if (matchingPreset) {
+    const isActiveMatch = matchingPreset.name === activePreset;
+    return {
+      mode: isActiveMatch ? "overwrite-active" : "overwrite-other",
+      canSubmit: !isActiveMatch || activePresetDirty,
+      isSaveReady: !isActiveMatch || activePresetDirty,
+      buttonLabel: isActiveMatch && !activePresetDirty ? "Saved" : "Overwrite",
+      title: isActiveMatch
+        ? activePresetDirty
+          ? "Save current settings to this preset"
+          : "No unsaved changes"
+        : `Overwrite preset "${trimmedName}"`,
+    };
+  }
+
+  return {
+    mode: "create",
+    canSubmit: true,
+    isSaveReady: true,
+    buttonLabel: "Save as New",
+    title: `Save current settings as "${trimmedName}"`,
+  };
+}
+
 function ParamSlider({
   label,
   value,
@@ -286,6 +498,9 @@ export function ChatSettingsPanel({
     (s) => s.loadedSpeculativeType,
   );
   const currentModels = useChatRuntimeStore((s) => s.models);
+  const modelRequiresTrustRemoteCode = useChatRuntimeStore(
+    (s) => s.modelRequiresTrustRemoteCode,
+  );
   const currentCheckpoint = params.checkpoint;
   const currentModelIsVision =
     currentModels.find((m) => m.id === currentCheckpoint)?.isVision ?? false;
@@ -316,13 +531,57 @@ export function ChatSettingsPanel({
   const [activePreset, setActivePreset] = useState(() =>
     loadSavedActivePreset(),
   );
-  const [savePresetOpen, setSavePresetOpen] = useState(false);
-  const [presetNameDraft, setPresetNameDraft] = useState("");
-  const presets = useMemo(
-    () => [...BUILTIN_PRESETS, ...customPresets],
-    [customPresets],
+  const [presetNameInput, setPresetNameInput] = useState(() =>
+    loadSavedActivePreset(),
   );
-  const isBuiltinPreset = BUILTIN_PRESETS.some((p) => p.name === activePreset);
+  const presetControlRowRef = useRef<HTMLDivElement>(null);
+  const [presetMenuWidthPx, setPresetMenuWidthPx] = useState<
+    number | undefined
+  >(undefined);
+  const [systemPromptEditorOpen, setSystemPromptEditorOpen] = useState(false);
+  const [systemPromptDraft, setSystemPromptDraft] = useState("");
+  const presets = useMemo(() => {
+    const overrides = new Set(customPresets.map((preset) => preset.name));
+    return [
+      ...BUILTIN_PRESETS.filter((preset) => !overrides.has(preset.name)),
+      ...customPresets,
+    ];
+  }, [customPresets]);
+  const activePresetDefinition = useMemo(
+    () => presets.find((preset) => preset.name === activePreset) ?? null,
+    [activePreset, presets],
+  );
+  const activeCustomPreset = useMemo(
+    () => customPresets.find((preset) => preset.name === activePreset) ?? null,
+    [activePreset, customPresets],
+  );
+  const activeBuiltinPreset = useMemo(
+    () =>
+      BUILTIN_PRESETS.find((preset) => preset.name === activePreset) ?? null,
+    [activePreset],
+  );
+  const activePresetDirty = useMemo(
+    () =>
+      activePresetDefinition == null
+        ? false
+        : !isSamePresetConfig(activePresetDefinition.params, params),
+    [activePresetDefinition, params],
+  );
+  const presetSaveState = useMemo(
+    () =>
+      getPresetSaveState({
+        rawName: presetNameInput,
+        activePreset,
+        presets,
+        activePresetDirty,
+      }),
+    [activePreset, activePresetDirty, presetNameInput, presets],
+  );
+  const systemPromptEditorDirty = systemPromptDraft !== params.systemPrompt;
+  const trustRemoteCodeMissing =
+    Boolean(currentCheckpoint) &&
+    modelRequiresTrustRemoteCode &&
+    !(params.trustRemoteCode ?? false);
 
   function set<K extends keyof InferenceParams>(key: K) {
     return (v: InferenceParams[K]) => onParamsChange({ ...params, [key]: v });
@@ -331,11 +590,19 @@ export function ChatSettingsPanel({
   function applyPreset(name: string) {
     const p = presets.find((pr) => pr.name === name);
     if (p) {
+      if (
+        modelRequiresTrustRemoteCode &&
+        !(p.params.trustRemoteCode ?? false)
+      ) {
+        toast.warning("This configuration turns custom code off", {
+          description:
+            "The current model needs custom code enabled to load. Keep it on for this model.",
+        });
+        return;
+      }
       onParamsChange({
         ...p.params,
-        systemPrompt: params.systemPrompt,
         checkpoint: params.checkpoint,
-        trustRemoteCode: params.trustRemoteCode,
       });
       setActivePreset(name);
       if (canUseStorage()) {
@@ -348,32 +615,23 @@ export function ChatSettingsPanel({
     }
   }
 
-  function openSavePresetDialog() {
-    setPresetNameDraft(activePreset === "Default" ? "" : activePreset);
-    setSavePresetOpen(true);
-  }
-
   function savePresetWithName(rawName: string) {
     const trimmed = rawName.trim();
     if (!trimmed) {
-      return;
-    }
-    if (BUILTIN_PRESETS.some((preset) => preset.name === trimmed)) {
+      toast.error("Enter a preset name");
       return;
     }
     setCustomPresets((prev) => {
-      const next = [
-        ...prev.filter((preset) => preset.name !== trimmed),
-        { name: trimmed, params: { ...params } },
-      ];
+      const next = prev.filter((p) => p.name !== trimmed);
+      const merged = [...next, { name: trimmed, params: { ...params } }];
       if (canUseStorage()) {
         try {
-          localStorage.setItem(CHAT_PRESETS_KEY, JSON.stringify(next));
+          localStorage.setItem(CHAT_PRESETS_KEY, JSON.stringify(merged));
         } catch {
           // ignore
         }
       }
-      return next;
+      return merged;
     });
     if (canUseStorage()) {
       try {
@@ -383,11 +641,31 @@ export function ChatSettingsPanel({
       }
     }
     setActivePreset(trimmed);
-    setSavePresetOpen(false);
+    setPresetNameInput(trimmed);
   }
 
   function deletePreset(name: string) {
-    if (BUILTIN_PRESETS.some((p) => p.name === name)) {
+    const hasCustomPreset = customPresets.some(
+      (preset) => preset.name === name,
+    );
+    if (!hasCustomPreset) {
+      return;
+    }
+    const builtinPreset = BUILTIN_PRESETS.find((preset) => preset.name === name);
+    const fallbackPreset =
+      builtinPreset ??
+      BUILTIN_PRESETS.find((preset) => preset.name === "Default") ??
+      null;
+    if (
+      activePreset === name &&
+      fallbackPreset &&
+      modelRequiresTrustRemoteCode &&
+      !(fallbackPreset.params.trustRemoteCode ?? false)
+    ) {
+      toast.warning("Reset would turn custom code off", {
+        description:
+          "The current model needs custom code enabled to load. Keep it on for this model.",
+      });
       return;
     }
     setCustomPresets((prev) => {
@@ -402,17 +680,33 @@ export function ChatSettingsPanel({
       return next;
     });
     if (activePreset === name) {
-      setActivePreset("Default");
-      if (canUseStorage()) {
-        try {
-          localStorage.setItem(CHAT_ACTIVE_PRESET_KEY, "Default");
-        } catch {
-          // ignore
+      if (fallbackPreset) {
+        onParamsChange({
+          ...fallbackPreset.params,
+          checkpoint: params.checkpoint,
+        });
+        setActivePreset(fallbackPreset.name);
+        if (canUseStorage()) {
+          try {
+            localStorage.setItem(CHAT_ACTIVE_PRESET_KEY, fallbackPreset.name);
+          } catch {
+            // ignore
+          }
         }
       }
     }
   }
 
+  function openSystemPromptEditor() {
+    setSystemPromptDraft(params.systemPrompt);
+    setSystemPromptEditorOpen(true);
+  }
+
+  function saveSystemPromptEditor() {
+    set("systemPrompt")(systemPromptDraft);
+    setSystemPromptEditorOpen(false);
+  }
+
   useEffect(() => {
     if (presets.some((preset) => preset.name === activePreset)) return;
     setActivePreset("Default");
@@ -425,6 +719,28 @@ export function ChatSettingsPanel({
     }
   }, [activePreset, presets]);
 
+  useEffect(() => {
+    setPresetNameInput(activePreset);
+  }, [activePreset]);
+
+  useEffect(() => {
+    if (!open) {
+      setSystemPromptEditorOpen(false);
+    }
+  }, [open]);
+
+  useLayoutEffect(() => {
+    const el = presetControlRowRef.current;
+    if (!el || !open) return;
+    const measure = () => {
+      setPresetMenuWidthPx(el.getBoundingClientRect().width);
+    };
+    measure();
+    const ro = new ResizeObserver(measure);
+    ro.observe(el);
+    return () => ro.disconnect();
+  }, [open]);
+
   const settingsContent = (
     <>
       <div className="flex items-center gap-2 px-4 py-3">
@@ -440,52 +756,138 @@ export function ChatSettingsPanel({
       <div className="flex-1 overflow-y-auto px-1.5">
         {/* mt-4 matches the Playground sidebar gap (SidebarHeader py-3 + SidebarGroup pt-1) */}
         <div className="mt-4 px-2 pb-3">
-          <div className="flex items-center gap-2">
-            <Select value={activePreset} onValueChange={applyPreset}>
-              <SelectTrigger className="h-8 flex-1 corner-squircle text-xs">
-                <SelectValue />
-              </SelectTrigger>
-              <SelectContent>
-                {presets.map((p) => (
-                  <SelectItem key={p.name} value={p.name}>
-                    {p.name}
-                  </SelectItem>
-                ))}
-              </SelectContent>
-            </Select>
-            <button
-              type="button"
-              onClick={openSavePresetDialog}
-              className="flex h-8 items-center gap-1.5 rounded-md border px-2.5 text-xs text-muted-foreground transition-colors hover:bg-accent"
-              title="Save preset"
-            >
-              <HugeiconsIcon icon={FloppyDiskIcon} className="size-3.5" />
-              Save
-            </button>
-            <button
-              type="button"
-              onClick={() => deletePreset(activePreset)}
-              disabled={isBuiltinPreset}
-              className="flex h-8 items-center gap-1.5 rounded-md border px-2.5 text-xs text-muted-foreground transition-colors hover:bg-accent disabled:cursor-not-allowed disabled:opacity-50"
-              title={
-                isBuiltinPreset
-                  ? "Built-in presets cannot be deleted"
-                  : "Delete selected preset"
-              }
-            >
-              <HugeiconsIcon icon={Delete02Icon} className="size-3.5" />
-              Delete
-            </button>
+          <div className="space-y-1.5">
+            <div ref={presetControlRowRef} className="w-full min-w-0">
+              <DropdownMenu>
+                <InputGroup className="!h-8 min-h-8 min-w-0 items-stretch gap-0 rounded-2xl pr-0 focus-within:border-input focus-within:ring-0 focus-within:shadow-none has-[[data-slot=input-group-control]:focus-visible]:border-input has-[[data-slot=input-group-control]:focus-visible]:ring-0 has-[[data-slot=input-group-control]:focus-visible]:shadow-none">
+                  <InputGroupInput
+                    id="inference-preset-name"
+                    value={presetNameInput}
+                    onChange={(e) => setPresetNameInput(e.target.value)}
+                    onKeyDown={(e) => {
+                      if (e.key === "Enter" && presetSaveState.canSubmit) {
+                        e.preventDefault();
+                        savePresetWithName(presetNameInput);
+                      }
+                    }}
+                    placeholder="Preset name"
+                    maxLength={80}
+                    autoComplete="off"
+                    className={cn(
+                      "!h-8 min-h-0 min-w-0 self-stretch !pl-2.5 !pr-2 pt-1 pb-1 text-sm leading-10 md:text-sm",
+                      presetSaveState.isSaveReady &&
+                        "text-foreground placeholder:text-primary/45",
+                    )}
+                    aria-label="Inference preset name"
+                  />
+                  <InputGroupAddon
+                    align="inline-end"
+                    className="min-h-0 shrink-0 gap-0 self-stretch border-0 py-0 pl-0 !pr-0 has-[>button]:mr-0"
+                  >
+                    <DropdownMenuTrigger asChild={true}>
+                      <InputGroupButton
+                        type="button"
+                        variant="ghost"
+                        size="icon-sm"
+                        className="!h-8 min-h-8 !w-7 min-w-7 shrink-0 rounded-none rounded-r-2xl border-l border-border px-0 text-muted-foreground transition-colors hover:bg-primary/15 hover:text-primary data-[state=open]:bg-primary/20 data-[state=open]:text-primary"
+                        title="Choose a preset"
+                        aria-label="Open preset list"
+                      >
+                        <HugeiconsIcon
+                          icon={ArrowDown01Icon}
+                          className="size-3.5"
+                          strokeWidth={2}
+                        />
+                      </InputGroupButton>
+                    </DropdownMenuTrigger>
+                  </InputGroupAddon>
+                </InputGroup>
+                <DropdownMenuContent
+                  align="end"
+                  className="min-w-40 max-w-none"
+                  style={
+                    presetMenuWidthPx != null
+                      ? {
+                          width: presetMenuWidthPx,
+                          minWidth: presetMenuWidthPx,
+                        }
+                      : undefined
+                  }
+                >
+                  {presets.map((p) => (
+                    <DropdownMenuItem
+                      key={p.name}
+                      onSelect={() => applyPreset(p.name)}
+                    >
+                      {p.name}
+                    </DropdownMenuItem>
+                  ))}
+                </DropdownMenuContent>
+              </DropdownMenu>
+            </div>
+            <div className="grid grid-cols-2 gap-1.5">
+              <Button
+                type="button"
+                onClick={() => savePresetWithName(presetNameInput)}
+                disabled={!presetSaveState.canSubmit}
+                variant={presetSaveState.isSaveReady ? "default" : "outline"}
+                size="sm"
+                className={cn(
+                  "h-8 w-full text-xs",
+                  presetSaveState.isSaveReady &&
+                    "bg-primary/92 text-primary-foreground hover:bg-primary",
+                )}
+                title={presetSaveState.title}
+                aria-label={presetSaveState.title}
+              >
+                <span className="inline-flex shrink-0 items-center pr-1.5">
+                  <HugeiconsIcon icon={FloppyDiskIcon} className="size-3.5" />
+                </span>
+                {presetSaveState.buttonLabel}
+              </Button>
+              <Button
+                type="button"
+                onClick={() => deletePreset(activePreset)}
+                disabled={!activeCustomPreset}
+                variant="outline"
+                size="sm"
+                className="h-8 w-full text-xs text-muted-foreground"
+                title={
+                  activeCustomPreset
+                    ? activeBuiltinPreset
+                      ? "Reset selected preset to built-in defaults"
+                      : "Delete selected preset"
+                    : "No saved override to delete"
+                }
+              >
+                <span className="inline-flex shrink-0 items-center pr-1.5">
+                  <HugeiconsIcon icon={Delete02Icon} className="size-3.5" />
+                </span>
+                Delete
+              </Button>
+            </div>
           </div>
         </div>
 
         <div className="px-2 pb-4">
-          <label
-            htmlFor="system-prompt"
-            className="mb-1.5 block text-xs font-medium"
-          >
-            System Prompt
-          </label>
+          <div className="mb-1.5 flex items-center justify-between gap-2">
+            <label
+              htmlFor="system-prompt"
+              className="block text-xs font-medium"
+            >
+              System Prompt
+            </label>
+            <Button
+              type="button"
+              variant="outline"
+              size="sm"
+              className="h-6 px-2 text-[11px]"
+              onClick={openSystemPromptEditor}
+              title="Open the full system prompt editor"
+            >
+              Edit
+            </Button>
+          </div>
           <Textarea
             id="system-prompt"
             value={params.systemPrompt}
@@ -561,37 +963,41 @@ export function ChatSettingsPanel({
                     typeof ctxDisplayValue === "number" &&
                     ctxDisplayValue > ggufMaxContextLength && (
                       <p className="text-[11px] text-amber-500">
-                        Exceeds estimated VRAM capacity ({ggufMaxContextLength.toLocaleString()} tokens). The model may use system RAM.
+                        Exceeds estimated VRAM capacity (
+                        {ggufMaxContextLength.toLocaleString()} tokens). The
+                        model may use system RAM.
                       </p>
                     )}
                 </div>
-                <div className="flex items-center justify-between gap-3">
+                <div className="grid grid-cols-[minmax(0,1fr)_65px] items-center gap-x-3">
                   <div className="min-w-0">
                     <div className="text-xs font-medium">KV Cache Dtype</div>
                     <div className="text-[11px] text-muted-foreground">
                       Quantize KV cache to reduce VRAM.
                     </div>
                   </div>
-                  <Select
-                    value={kvCacheDtype ?? "f16"}
-                    onValueChange={(v) => {
-                      setKvCacheDtype(v === "f16" ? null : v);
-                    }}
-                  >
-                    <SelectTrigger className="h-7 w-[90px] text-xs">
-                      <SelectValue />
-                    </SelectTrigger>
-                    <SelectContent>
-                      <SelectItem value="f16">f16</SelectItem>
-                      <SelectItem value="bf16">bf16</SelectItem>
-                      <SelectItem value="q8_0">q8_0</SelectItem>
-                      <SelectItem value="q5_1">q5_1</SelectItem>
-                      <SelectItem value="q4_1">q4_1</SelectItem>
-                    </SelectContent>
-                  </Select>
+                  <div className="w-full min-w-0">
+                    <Select
+                      value={kvCacheDtype ?? "f16"}
+                      onValueChange={(v) => {
+                        setKvCacheDtype(v === "f16" ? null : v);
+                      }}
+                    >
+                      <SelectTrigger className="grid h-7 w-full min-w-0 grid-cols-[minmax(0,1fr)_auto] items-center gap-1 px-2 py-0 text-xs [&_[data-slot=select-value]]:min-w-0 [&_[data-slot=select-value]]:truncate [&>svg]:shrink-0">
+                        <SelectValue />
+                      </SelectTrigger>
+                      <SelectContent>
+                        <SelectItem value="f16">f16</SelectItem>
+                        <SelectItem value="bf16">bf16</SelectItem>
+                        <SelectItem value="q8_0">q8_0</SelectItem>
+                        <SelectItem value="q5_1">q5_1</SelectItem>
+                        <SelectItem value="q4_1">q4_1</SelectItem>
+                      </SelectContent>
+                    </Select>
+                  </div>
                 </div>
                 {!currentModelIsVision && (
-                  <div className="flex items-center justify-between gap-3">
+                  <div className="grid grid-cols-[minmax(0,1fr)_65px] items-center gap-x-3">
                     <div className="min-w-0">
                       <div className="text-xs font-medium">
                         Speculative Decoding
@@ -600,20 +1006,22 @@ export function ChatSettingsPanel({
                         Speed up generation with no VRAM cost.
                       </div>
                     </div>
-                    <Select
-                      value={speculativeType ?? "off"}
-                      onValueChange={(v) => {
-                        setSpeculativeType(v === "off" ? null : v);
-                      }}
-                    >
-                      <SelectTrigger className="h-7 w-[120px] text-xs">
-                        <SelectValue />
-                      </SelectTrigger>
-                      <SelectContent>
-                        <SelectItem value="ngram-mod">On</SelectItem>
-                        <SelectItem value="off">Off</SelectItem>
-                      </SelectContent>
-                    </Select>
+                    <div className="w-full min-w-0">
+                      <Select
+                        value={speculativeType ?? "off"}
+                        onValueChange={(v) => {
+                          setSpeculativeType(v === "off" ? null : v);
+                        }}
+                      >
+                        <SelectTrigger className="grid h-7 w-full min-w-0 grid-cols-[minmax(0,1fr)_auto] items-center gap-1 px-2 py-0 text-xs [&_[data-slot=select-value]]:min-w-0 [&_[data-slot=select-value]]:truncate [&>svg]:shrink-0">
+                          <SelectValue />
+                        </SelectTrigger>
+                        <SelectContent>
+                          <SelectItem value="ngram-mod">On</SelectItem>
+                          <SelectItem value="off">Off</SelectItem>
+                        </SelectContent>
+                      </Select>
+                    </div>
                   </div>
                 )}
                 {modelSettingsDirty && (
@@ -641,19 +1049,33 @@ export function ChatSettingsPanel({
               </>
             )}
             {!isGguf && params.checkpoint && (
-              <div className="flex items-center justify-between gap-3">
-                <div className="min-w-0">
-                  <div className="text-xs font-medium">Enable custom code</div>
-                  <div className="text-[11px] text-muted-foreground">
-                    Allow models with custom code (e.g. Nemotron). Only enable
-                    if sure.
+              <>
+                <div className="flex items-center justify-between gap-3">
+                  <div className="min-w-0">
+                    <div className="text-xs font-medium">Enable custom code</div>
+                    <div className="text-[11px] text-muted-foreground">
+                      Allow models with custom code (e.g. Nemotron). Only
+                      enable if sure.
+                    </div>
                   </div>
+                  <Switch
+                    checked={params.trustRemoteCode ?? false}
+                    onCheckedChange={set("trustRemoteCode")}
+                  />
                 </div>
-                <Switch
-                  checked={params.trustRemoteCode ?? false}
-                  onCheckedChange={set("trustRemoteCode")}
-                />
-              </div>
+                {trustRemoteCodeMissing && (
+                  <Alert className="border-amber-200/70 bg-amber-50/70 px-3 py-2 text-amber-950 dark:border-amber-900/70 dark:bg-amber-950/35 dark:text-amber-100">
+                    <AlertTitle className="text-[11px] font-medium">
+                      Keep custom code enabled for this model
+                    </AlertTitle>
+                    <AlertDescription className="text-[11px] text-amber-800 dark:text-amber-200">
+                      This model requires custom code to load. You can edit the
+                      toggle, but loading will stay blocked until it is turned
+                      back on.
+                    </AlertDescription>
+                  </Alert>
+                )}
+              </>
             )}
           </div>
         </CollapsibleSection>
@@ -774,51 +1196,57 @@ export function ChatSettingsPanel({
         <ChatTemplateSection onReloadModel={onReloadModel} />
       </div>
       <Dialog
-        open={savePresetOpen}
+        open={systemPromptEditorOpen}
         onOpenChange={(nextOpen) => {
-          setSavePresetOpen(nextOpen);
-          if (!nextOpen) {
-            setPresetNameDraft("");
-          }
+          setSystemPromptEditorOpen(nextOpen);
         }}
       >
-        <DialogContent className="corner-squircle sm:max-w-sm">
+        <DialogContent
+          className="corner-squircle border border-border/60 bg-background/98 shadow-none sm:max-w-3xl"
+          overlayClassName="bg-background/35 supports-backdrop-filter:backdrop-blur-[1px]"
+        >
           <DialogHeader>
-            <DialogTitle>Save Preset</DialogTitle>
+            <DialogTitle>Edit System Prompt</DialogTitle>
             <DialogDescription>
-              Enter a name for this inference preset.
+              This prompt is part of the current configuration and saves with
+              the preset.
             </DialogDescription>
           </DialogHeader>
-          <form
-            onSubmit={(event) => {
-              event.preventDefault();
-              savePresetWithName(presetNameDraft);
-            }}
-            className="space-y-4"
-          >
-            <Input
-              autoFocus={true}
-              value={presetNameDraft}
-              onChange={(event) => setPresetNameDraft(event.target.value)}
-              placeholder="Preset name"
-              maxLength={80}
+          <div className="space-y-2">
+            <div className="space-y-0.5 px-0.5">
+              <div className="text-[11px] font-medium">Prompt editor</div>
+              <p className="text-[11px] text-muted-foreground">
+                Use this for longer edits. Save writes back to the active
+                configuration only.
+              </p>
+            </div>
+            <Textarea
+              value={systemPromptDraft}
+              onChange={(event) => setSystemPromptDraft(event.target.value)}
+              placeholder="You are a helpful assistant..."
+              className="min-h-[24rem] text-sm leading-6 corner-squircle"
+              rows={14}
             />
-            <DialogFooter>
-              <Button
-                type="button"
-                variant="outline"
-                onClick={() => setSavePresetOpen(false)}
-              >
-                Cancel
-              </Button>
-              <Button
-                type="submit"
-                disabled={presetNameDraft.trim().length === 0}
-              >
-                Save
-              </Button>
-            </DialogFooter>
-          </form>
+          </div>
+          <DialogFooter className="flex-wrap gap-2 sm:justify-between">
+            <Button
+              type="button"
+              variant="ghost"
+              onClick={() => {
+                setSystemPromptDraft(params.systemPrompt);
+                setSystemPromptEditorOpen(false);
+              }}
+            >
+              Cancel
+            </Button>
+            <Button
+              type="button"
+              onClick={saveSystemPromptEditor}
+              disabled={!systemPromptEditorDirty}
+            >
+              Save
+            </Button>
+          </DialogFooter>
         </DialogContent>
       </Dialog>
     </>
@@ -963,7 +1391,7 @@ function ChatTemplateSection({
         <Textarea
           value={displayValue}
           onChange={(e) => setOverride(e.target.value)}
-          className="min-h-32 font-mono text-[10px] leading-relaxed corner-squircle"
+          className="min-h-32 font-mono text-[10px] leading-relaxed md:text-[10px] corner-squircle"
           rows={6}
           spellCheck={false}
         />
diff --git a/studio/frontend/src/features/chat/hooks/use-chat-model-runtime.ts b/studio/frontend/src/features/chat/hooks/use-chat-model-runtime.ts
index 50b0200a56..e5c1493872 100644
--- a/studio/frontend/src/features/chat/hooks/use-chat-model-runtime.ts
+++ b/studio/frontend/src/features/chat/hooks/use-chat-model-runtime.ts
@@ -15,7 +15,7 @@ import {
   validateModel,
 } from "../api/chat-api";
 import { useChatRuntimeStore } from "../stores/chat-runtime-store";
-import type { LoadModelResponse } from "../types/api";
+import type { InferenceStatusResponse, LoadModelResponse } from "../types/api";
 import type {
   ChatLoraSummary,
   ChatModelSummary,
@@ -124,9 +124,13 @@ function toFiniteNumber(value: unknown): number | undefined {
   return value;
 }
 
+function getTrustRemoteCodeRequiredMessage(modelName: string): string {
+  return `${modelName} needs custom code enabled to load. Turn on "Enable custom code" in Chat Settings, then try again.`;
+}
+
 function mergeRecommendedInference(
   current: InferenceParams,
-  response: LoadModelResponse,
+  response: LoadModelResponse | InferenceStatusResponse,
   modelId: string,
 ): InferenceParams {
   const inference = response.inference;
@@ -233,7 +237,7 @@ export function useChatModelRuntime() {
         if (statusRes.inference) {
           const currentParams = useChatRuntimeStore.getState().params;
           setParams(
-            mergeRecommendedInference(currentParams, statusRes as any, statusRes.active_model),
+            mergeRecommendedInference(currentParams, statusRes, statusRes.active_model),
           );
         }
 
@@ -258,6 +262,8 @@ export function useChatModelRuntime() {
           ggufContextLength: currentGgufContextLength,
           ggufMaxContextLength,
           ggufNativeContextLength,
+          modelRequiresTrustRemoteCode:
+            statusRes.requires_trust_remote_code ?? false,
           speculativeType: currentSpecType,
           loadedSpeculativeType: currentSpecType,
         });
@@ -274,6 +280,10 @@ export function useChatModelRuntime() {
           }
           useChatRuntimeStore.getState().setReasoningEnabled(reasoningDefault);
         }
+      } else {
+        useChatRuntimeStore.setState({
+          modelRequiresTrustRemoteCode: false,
+        });
       }
     } catch (error) {
       const message =
@@ -347,7 +357,7 @@ export function useChatModelRuntime() {
       const previousIsLora =
         previousModel?.isLora ?? (previousLora ? true : false);
       // Covers Unix absolute (/), relative (./  ../), tilde (~/), Windows drive (C:\), UNC (\\server)
-      const isLocal = /^(\/|\.{1,2}[\\\/]|~[\\\/]|[A-Za-z]:[\\\/]|\\\\)/.test(modelId);
+      const isLocal = /^(\/|\.{1,2}[\\/]|~[\\/]|[A-Za-z]:[\\/]|\\\\)/.test(modelId);
       const isCachedLora = isLora && isLocal;
       const loadingDescription = [
         currentCheckpoint ? "Switching models." : null,
@@ -377,12 +387,15 @@ export function useChatModelRuntime() {
           const currentCheckpoint =
             useChatRuntimeStore.getState().params.checkpoint;
           const paramsBeforeLoad = useChatRuntimeStore.getState().params;
+          const trustRemoteCode = paramsBeforeLoad.trustRemoteCode ?? false;
           const maxSeqLength = paramsBeforeLoad.maxSeqLength;
           const hfToken = useChatRuntimeStore.getState().hfToken || null;
+          const previousModelRequiresTrustRemoteCode =
+            useChatRuntimeStore.getState().modelRequiresTrustRemoteCode;
           try {
             // Lightweight pre-flight validation: avoid unloading a working model
             // if the new identifier is clearly invalid (e.g. bad HF id / path).
-            await validateModel({
+            const validation = await validateModel({
               model_path: modelId,
               hf_token: hfToken,
               max_seq_length: maxSeqLength,
@@ -390,6 +403,9 @@ export function useChatModelRuntime() {
               is_lora: isLora,
               gguf_variant: ggufVariant ?? null,
             });
+            if (validation.requires_trust_remote_code && !trustRemoteCode) {
+              throw new Error(getTrustRemoteCodeRequiredMessage(displayName));
+            }
 
             if (currentCheckpoint) {
               await unloadModel({ model_path: currentCheckpoint });
@@ -409,7 +425,7 @@ export function useChatModelRuntime() {
               load_in_4bit: true,
               is_lora: isLora,
               gguf_variant: ggufVariant ?? null,
-              trust_remote_code: paramsBeforeLoad.trustRemoteCode ?? false,
+              trust_remote_code: trustRemoteCode,
               chat_template_override: chatTemplateOverride,
               cache_type_kv: kvCacheDtype,
               speculative_type: speculativeType,
@@ -454,6 +470,8 @@ export function useChatModelRuntime() {
               ggufContextLength: nativeCtx,
               ggufMaxContextLength,
               ggufNativeContextLength: reportedNativeCtx,
+              modelRequiresTrustRemoteCode:
+                loadResponse.requires_trust_remote_code ?? false,
               supportsReasoning: loadResponse.supports_reasoning ?? false,
               reasoningAlwaysOn,
               reasoningEnabled: reasoningAlwaysOn ? true : reasoningDefault,
@@ -490,6 +508,8 @@ export function useChatModelRuntime() {
                   load_in_4bit: true,
                   is_lora: previousIsLora,
                   gguf_variant: previousVariant,
+                  trust_remote_code:
+                    previousModelRequiresTrustRemoteCode || trustRemoteCode,
                 });
                 await refresh();
               } catch {
diff --git a/studio/frontend/src/features/chat/runtime-provider.tsx b/studio/frontend/src/features/chat/runtime-provider.tsx
index 024543edb9..348a0a85c4 100644
--- a/studio/frontend/src/features/chat/runtime-provider.tsx
+++ b/studio/frontend/src/features/chat/runtime-provider.tsx
@@ -331,6 +331,9 @@ function fallbackTitleFromUserText(userText: string): string {
 }
 
 function cloneContent(content: ThreadMessage["content"]): ThreadMessage["content"] {
+  if (typeof content === "string") {
+    return content;
+  }
   return Array.isArray(content)
     ? JSON.parse(JSON.stringify(content))
     : [];
diff --git a/studio/frontend/src/features/chat/shared-composer.tsx b/studio/frontend/src/features/chat/shared-composer.tsx
index ac01f77381..d45004cd43 100644
--- a/studio/frontend/src/features/chat/shared-composer.tsx
+++ b/studio/frontend/src/features/chat/shared-composer.tsx
@@ -8,7 +8,7 @@ import { useAui } from "@assistant-ui/react";
 import { cn } from "@/lib/utils";
 import { ArrowUpIcon, GlobeIcon, HeadphonesIcon, LightbulbIcon, LightbulbOffIcon, MicIcon, PlusIcon, SquareIcon, TerminalIcon, XIcon } from "lucide-react";
 import { toast } from "sonner";
-import { loadModel } from "./api/chat-api";
+import { loadModel, validateModel } from "./api/chat-api";
 import { useChatRuntimeStore } from "./stores/chat-runtime-store";
 import {
   type KeyboardEvent,
@@ -336,6 +336,27 @@ export function SharedComposer({
 
       // Helper: load a model and update store checkpoint
       async function ensureModelLoaded(sel: CompareModelSelection): Promise<string> {
+        const currentStore = useChatRuntimeStore.getState();
+        const isAlreadyActive =
+          currentStore.params.checkpoint === sel.id &&
+          (currentStore.activeGgufVariant ?? null) === (sel.ggufVariant ?? null);
+        if (!isAlreadyActive) {
+          const validation = await validateModel({
+            model_path: sel.id,
+            hf_token: currentStore.hfToken || null,
+            max_seq_length: maxSeqLength,
+            load_in_4bit: true,
+            is_lora: sel.isLora,
+            gguf_variant: sel.ggufVariant ?? null,
+            trust_remote_code: trustRemoteCode,
+            chat_template_override: chatTemplateOverride,
+          });
+          if (validation.requires_trust_remote_code && !trustRemoteCode) {
+            throw new Error(
+              `${modelDisplayName(sel.id)} needs custom code enabled to load. Turn on "Enable custom code" in Chat Settings, then try again.`,
+            );
+          }
+        }
         const resp = await loadModel({
           model_path: sel.id,
           hf_token: useChatRuntimeStore.getState().hfToken || null,
@@ -346,10 +367,14 @@ export function SharedComposer({
           trust_remote_code: trustRemoteCode,
           chat_template_override: chatTemplateOverride,
         });
-        useChatRuntimeStore.getState().setCheckpoint(
+        const store = useChatRuntimeStore.getState();
+        store.setCheckpoint(
           resp.model,
           resp.is_gguf ? (sel.ggufVariant ?? undefined) : null,
         );
+        store.setModelRequiresTrustRemoteCode(
+          resp.requires_trust_remote_code ?? false,
+        );
         return resp.status;
       }
 
diff --git a/studio/frontend/src/features/chat/stores/chat-runtime-store.ts b/studio/frontend/src/features/chat/stores/chat-runtime-store.ts
index c86226ce4e..a041452a48 100644
--- a/studio/frontend/src/features/chat/stores/chat-runtime-store.ts
+++ b/studio/frontend/src/features/chat/stores/chat-runtime-store.ts
@@ -132,7 +132,8 @@ function loadInferenceParams(): InferenceParams {
 function saveInferenceParams(params: InferenceParams): boolean {
   if (!canUseStorage()) return false;
   try {
-    const { checkpoint: _, ...rest } = params;
+    const { checkpoint, ...rest } = params;
+    void checkpoint;
     localStorage.setItem(INFERENCE_PARAMS_KEY, JSON.stringify(rest));
     return true;
   } catch {
@@ -152,6 +153,7 @@ type ChatRuntimeStore = {
   ggufContextLength: number | null;
   ggufMaxContextLength: number | null;
   ggufNativeContextLength: number | null;
+  modelRequiresTrustRemoteCode: boolean;
   supportsReasoning: boolean;
   reasoningAlwaysOn: boolean;
   reasoningEnabled: boolean;
@@ -181,6 +183,7 @@ type ChatRuntimeStore = {
   } | null;
   modelLoading: boolean;
   setModelLoading: (loading: boolean) => void;
+  setModelRequiresTrustRemoteCode: (required: boolean) => void;
   setParams: (params: InferenceParams) => void;
   setModels: (models: ChatModelSummary[]) => void;
   setLoras: (loras: ChatLoraSummary[]) => void;
@@ -220,6 +223,7 @@ export const useChatRuntimeStore = create<ChatRuntimeStore>((set) => ({
   ggufContextLength: null,
   ggufMaxContextLength: null,
   ggufNativeContextLength: null,
+  modelRequiresTrustRemoteCode: false,
   supportsReasoning: false,
   reasoningAlwaysOn: false,
   reasoningEnabled: true,
@@ -244,6 +248,8 @@ export const useChatRuntimeStore = create<ChatRuntimeStore>((set) => ({
   contextUsage: null,
   modelLoading: false,
   setModelLoading: (loading) => set({ modelLoading: loading }),
+  setModelRequiresTrustRemoteCode: (modelRequiresTrustRemoteCode) =>
+    set({ modelRequiresTrustRemoteCode }),
   setParams: (params) =>
     set(() => {
       const persisted = saveInferenceParams(params);
@@ -298,6 +304,7 @@ export const useChatRuntimeStore = create<ChatRuntimeStore>((set) => ({
       ggufContextLength: null,
       ggufMaxContextLength: null,
       ggufNativeContextLength: null,
+      modelRequiresTrustRemoteCode: false,
       contextUsage: null,
       supportsReasoning: false,
       reasoningEnabled: true,
diff --git a/studio/frontend/src/features/chat/types/api.ts b/studio/frontend/src/features/chat/types/api.ts
index 90489222c3..133ae2adee 100644
--- a/studio/frontend/src/features/chat/types/api.ts
+++ b/studio/frontend/src/features/chat/types/api.ts
@@ -52,6 +52,7 @@ export interface ValidateModelResponse {
   is_gguf?: boolean;
   is_lora?: boolean;
   is_vision?: boolean;
+  requires_trust_remote_code?: boolean;
 }
 
 export interface GgufVariantDetail {
@@ -86,6 +87,7 @@ export interface LoadModelResponse {
     presence_penalty?: number;
     trust_remote_code?: boolean;
   };
+  requires_trust_remote_code?: boolean;
   context_length?: number | null;
   max_context_length?: number | null;
   native_context_length?: number | null;
@@ -119,6 +121,7 @@ export interface InferenceStatusResponse {
     presence_penalty?: number;
     trust_remote_code?: boolean;
   };
+  requires_trust_remote_code?: boolean;
   supports_reasoning?: boolean;
   reasoning_always_on?: boolean;
   supports_tools?: boolean;
diff --git a/studio/frontend/src/features/chat/utils/delete-thread-message.ts b/studio/frontend/src/features/chat/utils/delete-thread-message.ts
new file mode 100644
index 0000000000..a6d6556a87
--- /dev/null
+++ b/studio/frontend/src/features/chat/utils/delete-thread-message.ts
@@ -0,0 +1,119 @@
+// SPDX-License-Identifier: AGPL-3.0-only
+// Copyright 2026-present the Unsloth AI Inc. team. All rights reserved. See /studio/LICENSE.AGPL-3.0
+
+import type {
+  CompleteAttachment,
+  ExportedMessageRepository,
+  ThreadMessage,
+} from "@assistant-ui/react";
+/**
+ * assistant-ui does not expose a public `deleteMessage` on `ThreadRuntime` / `MessageRuntime`
+ * in our version, but it already implements branch-safe deletion inside `MessageRepository`.
+ * We import that helper from an **internal** package path (`runtime/utils/message-repository`).
+ *
+ * **Maintainability:** treat this file as the only place that imports `MessageRepository` from
+ * `@assistant-ui/core`. When bumping `@assistant-ui/react` / `@assistant-ui/core`, re-run chat
+ * delete + reload smoke tests; the path or API may change without a semver signal on “public”
+ * surface area.
+ */
+import { MessageRepository } from "@assistant-ui/core/runtime/utils/message-repository";
+import { db } from "@/features/chat/db";
+import type { MessageRecord } from "@/features/chat/types";
+
+function cloneContent(content: ThreadMessage["content"]): ThreadMessage["content"] {
+  if (typeof content === "string") {
+    return content;
+  }
+  return Array.isArray(content) ? JSON.parse(JSON.stringify(content)) : [];
+}
+
+function cloneAttachments(
+  attachments: readonly CompleteAttachment[] | undefined,
+): readonly CompleteAttachment[] {
+  if (!Array.isArray(attachments)) {
+    return [];
+  }
+  return JSON.parse(JSON.stringify(attachments));
+}
+
+function exportedItemToRecord(
+  threadId: string,
+  parentId: string | null,
+  message: ThreadMessage,
+): MessageRecord {
+  const content = cloneContent(message.content);
+  if (message.role === "user") {
+    const attachments = cloneAttachments(message.attachments);
+    const custom = message.metadata?.custom;
+    return {
+      id: message.id,
+      threadId,
+      parentId: parentId ?? null,
+      role: "user",
+      content: content as Extract<ThreadMessage, { role: "user" }>["content"],
+      ...(attachments.length > 0 && { attachments }),
+      ...(custom && Object.keys(custom).length > 0 && { metadata: custom }),
+      createdAt: message.createdAt?.getTime?.() ?? Date.now(),
+    };
+  }
+  const custom = (message.metadata?.custom ?? {}) as Record<string, unknown>;
+  return {
+    id: message.id,
+    threadId,
+    parentId: parentId ?? null,
+    role: "assistant",
+    content: content as Extract<ThreadMessage, { role: "assistant" }>["content"],
+    ...(Object.keys(custom).length > 0 && { metadata: custom }),
+    createdAt: message.createdAt?.getTime?.() ?? Date.now(),
+  };
+}
+
+/**
+ * Persist the exact message list represented by `exp` for this thread, removing
+ * Dexie rows that are no longer present (e.g. after a delete).
+ */
+async function syncExportedRepositoryToDexie(
+  remoteId: string,
+  exp: ExportedMessageRepository,
+): Promise<void> {
+  await db.transaction("rw", db.messages, async () => {
+    const keepIds = new Set(exp.messages.map((x) => x.message.id));
+    const existing = await db.messages.where("threadId").equals(remoteId).toArray();
+    const idsToDelete = existing
+      .filter((m) => !keepIds.has(m.id))
+      .map((m) => m.id);
+    if (idsToDelete.length > 0) {
+      await db.messages.bulkDelete(idsToDelete);
+    }
+    await db.messages.bulkPut(
+      exp.messages.map(({ message, parentId }) =>
+        exportedItemToRecord(remoteId, parentId, message),
+      ),
+    );
+  });
+}
+
+type ThreadImportExport = {
+  export: () => ExportedMessageRepository;
+  import: (data: ExportedMessageRepository) => void;
+};
+
+/**
+ * Remove a message from the thread and mirror the result to IndexedDB.
+ */
+export async function deleteThreadMessage(args: {
+  thread: ThreadImportExport;
+  messageId: string;
+  remoteId: string | undefined;
+}): Promise<void> {
+  const { thread, messageId, remoteId } = args;
+  const exported = thread.export();
+  const repo = new MessageRepository();
+  repo.import(exported);
+  repo.deleteMessage(messageId);
+  const next = repo.export();
+  if (remoteId) {
+    await syncExportedRepositoryToDexie(remoteId, next);
+  }
+  thread.import(next);
+}
diff --git a/studio/frontend/src/index.css b/studio/frontend/src/index.css
index a30b4ca287..d051b24245 100644
--- a/studio/frontend/src/index.css
+++ b/studio/frontend/src/index.css
@@ -173,6 +173,7 @@
   --color-chart-3: var(--chart-3);
   --color-chart-2: var(--chart-2);
   --color-chart-1: var(--chart-1);
+  --color-code-block: #181818;
   --color-ring: var(--ring);
   --color-input: var(--input);
   --color-border: var(--border);
@@ -395,6 +396,13 @@
     width: calc(100% + 1.25rem);
     max-width: calc(100% + 1.25rem);
   }
+
+  .dark .aui-thread-root [data-streamdown="code-block"] {
+    /* Streamdown `pre` uses `dark:bg-[var(--shiki-dark-bg,...)]`; keep one surface on the outer shell. */
+    --shiki-dark-bg: transparent;
+    background: var(--color-code-block);
+    border: 1px solid oklch(1 0 0 / 0.07);
+  }
 }
 
 /* Minimal scrollbar — thumb only, no track */

From 3bb72a557f9f8afd960ea5a26fd3d04f7f44e2f7 Mon Sep 17 00:00:00 2001
From: Roland Tannous <115670425+rolandtannous@users.noreply.github.com>
Date: Mon, 13 Apr 2026 20:42:02 +0400
Subject: [PATCH 077/216] Pin kernels==0.12.1 to avoid huggingface_hub
 dataclass conflict (#5000)

---
 studio/backend/requirements/extras-no-deps.txt | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/studio/backend/requirements/extras-no-deps.txt b/studio/backend/requirements/extras-no-deps.txt
index 9934bacd24..59a8e45434 100644
--- a/studio/backend/requirements/extras-no-deps.txt
+++ b/studio/backend/requirements/extras-no-deps.txt
@@ -13,4 +13,4 @@ torch-c-dlpack-ext
 sentence_transformers==5.2.0
 transformers==4.57.6
 pytorch_tokenizers
-kernels
+kernels==0.12.1

From 9a261aec5ffd44e7b791e43cd8843c96ce968b02 Mon Sep 17 00:00:00 2001
From: Daniel Han <danielhanchen@gmail.com>
Date: Mon, 13 Apr 2026 10:08:11 -0700
Subject: [PATCH 078/216] Studio: Expose openai and anthropic compatible
 external API end points  (#4956)
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

* Studio: add API key authentication for programmatic access

External users want to hit the Studio API (chat completions with tool
calling, training, export, etc.) without going through the browser
login flow. This adds sk-unsloth- prefixed API keys that work as a
drop-in replacement for JWTs in the Authorization: Bearer header.

Backend:
- New api_keys table in SQLite (storage.py)
- create/list/revoke/validate functions with SHA-256 hashed storage
- API key detection in _get_current_subject before the JWT path
- POST/GET/DELETE /api/auth/api-keys endpoints on the auth router

Frontend:
- /api-keys page with create form, one-time key reveal, keys table
- API Keys link in desktop and mobile navbar
- Route registered with requireAuth guard

Zero changes to any existing route handler -- every endpoint that uses
Depends(get_current_subject) automatically works with API keys.

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

* Use actual origin in API key usage examples

The examples on /api-keys were hardcoded to localhost:8888 which is
wrong for remote users. Use window.location.origin so the examples
show the correct URL regardless of where the user is connecting from.

* Add `unsloth studio run` CLI command for one-liner model serving

Adds a `run` subcommand that starts Studio, loads a model, creates an
API key, and prints a ready-to-use curl command -- similar to
`ollama run` or `vllm serve`.

Usage: unsloth studio run -m unsloth/Qwen3-1.7B-GGUF --gguf-variant UD-Q4_K_XL

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

* Add end-to-end tests for `unsloth studio run` and API key usage

Tests the 4 usage examples from the API Keys page:
1. curl basic (non-streaming) chat completions
2. curl streaming (SSE) chat completions
3. OpenAI Python SDK streaming completions
4. curl with tools (web_search + python)

Also tests --help output, invalid key rejection, and no-key rejection.
All 7 tests pass against Qwen3-1.7B-GGUF.

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

* Add /v1/completions, /v1/embeddings, /v1/responses endpoints and --parallel support

- llama_cpp.py: accept n_parallel param, pass to llama-server --parallel
- run.py: plumb llama_parallel_slots through to app.state
- inference.py: add /completions and /embeddings as transparent proxies to
  llama-server, add /responses as application-level endpoint that converts
  to ChatCompletionRequest; thread n_parallel through load_model
- studio.py: set llama_parallel_slots=4 for `unsloth studio run` path

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

* Make /v1/responses endpoint match OpenAI Responses API format

The existing /v1/responses shim returned Chat Completions format, which
broke OpenAI SDK clients using openai.responses.create(). This commit
replaces the endpoint with a proper implementation that:

- Returns `output` array with `output_text` content parts instead of
  `choices` with `message`
- Uses `input_tokens`/`output_tokens` instead of `prompt_tokens`/
  `completion_tokens` in usage
- Sets `object: "response"` and `id: "resp_..."`
- Emits named SSE events for streaming (response.created,
  response.output_text.delta, response.completed, etc.)
- Accepts all OpenAI Responses API fields (tools, store, metadata,
  previous_response_id) without erroring -- silently ignored
- Maps `developer` role to `system` and `input_text`/`input_image`
  content parts to the internal Chat format

Adds Pydantic schemas for request/response models and 23 unit tests
covering schema validation, input normalisation, and response format.

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

* Studio: add Anthropic-compatible /v1/messages endpoint (#4981)

* Add Anthropic-compatible /v1/messages endpoint with tool support

Translate Anthropic Messages API format to/from internal OpenAI format
and reuse the existing server-side agentic tool loop. Supports streaming
SSE (message_start, content_block_delta, etc.) and non-streaming JSON.
Includes offline unit tests and e2e tests in test_studio_run.py.

* Add enable_tools, enabled_tools, session_id to /v1/messages endpoint

Support the same shorthand as /v1/chat/completions: enable_tools=true
with an optional enabled_tools list uses built-in server tools without
requiring full Anthropic tool definitions. session_id is passed through
for sandbox isolation. max_tokens is now optional.

* Strip leaked tool-call XML from Anthropic endpoint content

Apply _TOOL_XML_RE to content events in both streaming and
non-streaming tool paths, matching the OpenAI endpoint behavior.

* Emit custom tool_result SSE event in Anthropic stream

Adds a non-standard tool_result event between the tool_use block close
and the next text block, so clients can see server-side tool execution
results. Anthropic SDKs ignore unknown event types.

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

* Split /v1/messages into server-side and client-side tool paths

enable_tools=true runs the existing server-side agentic loop with
built-in tools (web_search/python/terminal). A bare tools=[...] field
now triggers a client-side pass-through: client-provided tools are
forwarded to llama-server and any tool_use output is returned to the
caller with stop_reason=tool_use for client execution.

This fixes Claude Code (and any Anthropic SDK client) which sends
tools=[...] expecting client-side execution but was previously routed
through execute_tool() and failing with 'Unknown tool'.

Adds AnthropicPassthroughEmitter to convert llama-server OpenAI SSE
chunks into Anthropic SSE events, plus unit tests covering text
blocks, tool_use blocks, mixed, stop reasons, and usage.

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

* Fix httpcore GeneratorExit in /v1/messages passthrough stream

Explicitly aclose aiter_lines() before the surrounding async with
blocks unwind, mirroring the prior fix in external_provider.py
(a41160d3) and cc757b78's RuntimeError suppression.

* Wire stop_sequences through /v1/messages; warn on tool_choice

Plumb payload.stop_sequences to all three code paths (server-side
tool loop, no-tool plain, client-side passthrough) so Anthropic SDK
clients setting stop_sequences get the behavior they expect. The
llama_cpp backend already accepted `stop` on both generate_chat_
completion and generate_chat_completion_with_tools; the Anthropic
handler simply wasn't passing it.

tool_choice remains declared on the request model for Anthropic SDK
compatibility (the SDK often sets it by default) but is not yet
honored. Log a structured warning on each request carrying a non-
null tool_choice so the silent drop is visible to operators.

* Wire min_p / repetition_penalty / presence_penalty through /v1/messages

Align the Anthropic endpoint's sampling surface with /v1/chat/completions.
Adds the three fields as x-unsloth extensions on AnthropicMessagesRequest
and threads them through all three code paths: server-side tool loop,
no-tool plain, and client-side passthrough.

The passthrough builder emits "repeat_penalty" (not "repetition_penalty")
because that is llama-server's field name; the backend methods already
apply the same rename internally.

* Fix block ordering and prev_text reset in non-streaming tool path

_anthropic_tool_non_streaming was building the response by appending
all tool_use blocks first, then a single concatenated text block at
the end — losing generation order and merging pre-tool and post-tool
text into one block. It also never reset prev_text between synthesis
turns, so the first N characters of each post-tool turn were dropped
(where N = length of the prior turn's final cumulative text).

Rewrite to build content_blocks incrementally in generation order,
matching the streaming emitter's behavior: deltas within a turn are
merged into the trailing text block, tool_use blocks interrupt the
text sequence, and prev_text is reset on tool_end so turn N+1 diffs
against an empty baseline.

Caught by gemini-code-assist[bot] review on #4981.

* Make test_studio_run.py e2e tests pytest-compatible

Add a hybrid session-scoped studio_server fixture in conftest.py that
feeds base_url / api_key into the existing e2e test functions. Three
invocation modes are now supported:

1. Script mode (unchanged) — python tests/test_studio_run.py
2. Pytest + external server — point at a running instance via
   UNSLOTH_E2E_BASE_URL / UNSLOTH_E2E_API_KEY env vars, no per-run
   GGUF load cost
3. Pytest + fixture-managed server — pytest drives _start_server /
   _kill_server itself via --unsloth-model / --unsloth-gguf-variant,
   CI-friendly

The existing _start_server / _kill_server helpers and main() stay
untouched so the script entry point keeps working exactly as before.
Test function signatures are unchanged — the (base_url, api_key)
parameters now resolve via the new fixtures when running under
pytest.

* Rename test_studio_run.py -> test_studio_api.py

The file is entirely about HTTP API endpoint testing (OpenAI-compatible
/v1/chat/completions, Anthropic-compatible /v1/messages, API key auth,
plus a CLI --help sanity check on the command that runs the API). None
of its tests cover training, export, chat-UI, or internal-Python-API
concerns.

The old name misleadingly suggested "tests for the unsloth studio run
CLI subcommand" — the new name reflects the actual scope.

Updates:
- git mv the file (rename tracked, history preserved)
- Rewrite opening docstring to state the API surface focus and call
  out what is explicitly out of scope
- Update all 4 Usage-block path references to the new filename
- LOG_FILE renamed to test_studio_api.log
- conftest.py fixture import rewritten from test_studio_run to
  test_studio_api, plus 7 docstring/comment references updated

No functional changes to test logic, signatures, or main().

---------

Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com>

* Fix httpcore asyncgen cleanup in /v1/messages and /v1/completions

The earlier fix in 985e92a9 was incomplete: it closed aiter_lines()
explicitly but still used `async with httpx.AsyncClient()` /
`async with client.stream()` inside the generator. When the generator
is orphaned (e.g. client disconnects mid-stream and Starlette drops
the StreamingResponse iterator without explicitly calling aclose()),
Python's asyncgen finalizer runs the cleanup in a DIFFERENT task than
the one that originally entered the httpx context managers. The
`async with` exits then trigger httpcore's HTTP11ConnectionByteStream
.aclose(), which enters anyio.CancelScope.__exit__ with a mismatched
task and raises RuntimeError("Attempted to exit cancel scope in a
different task"). That error escapes any user-owned try/except
because it happens during GC finalization.

Replace `async with` with manual client/response lifecycle in both
/v1/messages passthrough and /v1/completions proxy. Close the
response and client in a finally block wrapped in
`try: ... except Exception: pass`. This suppresses RuntimeError (and
other Exception subclasses) from the anyio cleanup noise while
letting GeneratorExit (a BaseException, not Exception) propagate
cleanly so the generator terminates as Python expects.

Traceback observed in user report:
  File ".../httpcore/_async/connection_pool.py", line 404, in __aiter__
      yield part
  RuntimeError: async generator ignored GeneratorExit
...
  File ".../anyio/_backends/_asyncio.py", line 455, in __exit__
      raise RuntimeError(
  RuntimeError: Attempted to exit cancel scope in a different task

* Expand unsloth studio run banner with SDK base URL and more curl examples

Add an explicit "OpenAI / Anthropic SDK base URL" line inside the info
box so SDK users don't accidentally copy the bare server URL (without
/v1) into their OpenAI/Anthropic SDK constructors and hit 404s.

Replace the single /v1/chat/completions curl example with three
labeled blocks: chat/completions, Anthropic /messages, and OpenAI
Responses. The Anthropic example includes max_tokens (Anthropic SDKs
require it even though Studio accepts None).

All examples derived from a computed sdk_base_url so the /v1 prefix
stays in sync if the public path ever changes.

* Hash API keys with HMAC-SHA256 + persistent server secret

Stores the HMAC secret in a new app_secrets singleton table. Fixes
CodeQL py/weak-sensitive-data-hashing alert on storage.py:74-76,
394-395. Refresh tokens stay on plain SHA-256 (unchanged _hash_token)
so existing user sessions survive upgrade — API keys are new on this
branch so there is no migration.

* Use PBKDF2 for API key hashing per CodeQL recommendation

HMAC-SHA256 was still flagged by py/weak-sensitive-data-hashing.
Switch to hashlib.pbkdf2_hmac, which is in CodeQL's recommended
allowlist (Argon2/scrypt/bcrypt/PBKDF2). Persistent server-side
salt stays in app_secrets for defense-in-depth. 100k iterations to
match auth/hashing.py's password hasher.

---------

Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com>
Co-authored-by: Roland Tannous <115670425+rolandtannous@users.noreply.github.com>
Co-authored-by: Roland Tannous <rolandtannous@gravityq.ai>
---
 studio/backend/auth/authentication.py         |   14 +
 studio/backend/auth/storage.py                |  225 +++-
 .../core/inference/anthropic_compat.py        |  488 ++++++++
 studio/backend/core/inference/llama_cpp.py    |    3 +-
 studio/backend/models/auth.py                 |   43 +
 studio/backend/models/inference.py            |  238 ++++
 studio/backend/routes/auth.py                 |   73 +-
 studio/backend/routes/inference.py            | 1099 ++++++++++++++++-
 studio/backend/run.py                         |    3 +
 studio/backend/tests/conftest.py              |  126 +-
 .../backend/tests/test_anthropic_messages.py  |  774 ++++++++++++
 studio/backend/tests/test_responses_api.py    |  328 +++++
 studio/backend/tests/test_studio_api.py       |  643 ++++++++++
 studio/frontend/src/app/router.tsx            |    2 +
 studio/frontend/src/app/routes/api-keys.tsx   |   18 +
 studio/frontend/src/components/navbar.tsx     |   30 +-
 .../src/features/auth/api-keys-page.tsx       |  417 +++++++
 studio/frontend/src/features/auth/index.ts    |    1 +
 unsloth_cli/commands/studio.py                |  242 ++++
 19 files changed, 4760 insertions(+), 7 deletions(-)
 create mode 100644 studio/backend/core/inference/anthropic_compat.py
 create mode 100644 studio/backend/tests/test_anthropic_messages.py
 create mode 100644 studio/backend/tests/test_responses_api.py
 create mode 100644 studio/backend/tests/test_studio_api.py
 create mode 100644 studio/frontend/src/app/routes/api-keys.tsx
 create mode 100644 studio/frontend/src/features/auth/api-keys-page.tsx

diff --git a/studio/backend/auth/authentication.py b/studio/backend/auth/authentication.py
index b39f915764..da59ba9a1a 100644
--- a/studio/backend/auth/authentication.py
+++ b/studio/backend/auth/authentication.py
@@ -10,10 +10,12 @@ from fastapi.security import HTTPAuthorizationCredentials, HTTPBearer
 import jwt
 
 from .storage import (
+    API_KEY_PREFIX,
     get_jwt_secret,
     get_user_and_secret,
     load_jwt_secret,
     save_refresh_token,
+    validate_api_key,
     verify_refresh_token,
 )
 
@@ -137,6 +139,18 @@ async def _get_current_subject(
             ...
     """
     token = credentials.credentials
+
+    # --- API key path (sk-unsloth-...) ---
+    if token.startswith(API_KEY_PREFIX):
+        username = validate_api_key(token)
+        if username is None:
+            raise HTTPException(
+                status_code = status.HTTP_401_UNAUTHORIZED,
+                detail = "Invalid or expired API key",
+            )
+        return username
+
+    # --- JWT path ---
     subject = _decode_subject_without_verification(token)
     if subject is None:
         raise HTTPException(
diff --git a/studio/backend/auth/storage.py b/studio/backend/auth/storage.py
index 1395574cce..7d55a2dc59 100644
--- a/studio/backend/auth/storage.py
+++ b/studio/backend/auth/storage.py
@@ -72,7 +72,22 @@ def clear_bootstrap_password() -> None:
 
 
 def _hash_token(token: str) -> str:
-    """SHA-256 hash helper used for refresh token storage."""
+    """SHA-256 hash helper used for refresh token storage.
+
+    Plain SHA-256 is intentional here: refresh tokens are high-entropy
+    random strings from ``secrets.token_urlsafe(48)`` (384 bits of
+    entropy), so a slow KDF (Argon2 / bcrypt / PBKDF2) provides zero
+    additional security — no attacker can brute-force 2^384 regardless
+    of hash speed — while adding tens of ms of CPU to every refresh.
+    See the OWASP Password Storage Cheat Sheet on fast-vs-slow hashing
+    of high-entropy inputs.
+
+    API keys use the separate ``_pbkdf2_api_key`` helper below, which
+    runs PBKDF2-HMAC-SHA256 with a persistent server-side salt — not
+    for cryptographic reasons (128-bit random tokens don't need slow
+    hashing), but because CodeQL's ``py/weak-sensitive-data-hashing``
+    query mislabels API keys as passwords and demands a KDF.
+    """
     return hashlib.sha256(token.encode("utf-8")).hexdigest()
 
 
@@ -103,6 +118,29 @@ def get_connection() -> sqlite3.Connection:
         );
         """
     )
+    conn.execute(
+        """
+        CREATE TABLE IF NOT EXISTS api_keys (
+            id         INTEGER PRIMARY KEY AUTOINCREMENT,
+            username   TEXT NOT NULL,
+            key_prefix TEXT NOT NULL,
+            key_hash   TEXT NOT NULL UNIQUE,
+            name       TEXT NOT NULL DEFAULT '',
+            created_at TEXT NOT NULL,
+            last_used_at TEXT,
+            expires_at TEXT,
+            is_active  INTEGER NOT NULL DEFAULT 1
+        );
+        """
+    )
+    conn.execute(
+        """
+        CREATE TABLE IF NOT EXISTS app_secrets (
+            key   TEXT PRIMARY KEY,
+            value TEXT NOT NULL
+        );
+        """
+    )
     columns = {row["name"] for row in conn.execute("PRAGMA table_info(auth_user)")}
     if "must_change_password" not in columns:
         conn.execute(
@@ -112,6 +150,89 @@ def get_connection() -> sqlite3.Connection:
     return conn
 
 
+# ── API-key PBKDF2 salt ────────────────────────────────────────────────
+#
+# Module-level cache for the persistent API-key PBKDF2 salt. Populated
+# lazily on first use via ``_get_or_create_api_key_pbkdf2_salt``. Not
+# protected by a lock because (a) the ``INSERT OR IGNORE`` provides
+# atomicity at the SQLite layer and (b) concurrent populations converge
+# on the same value, so the worst case is a harmless duplicate read on
+# startup.
+_api_key_pbkdf2_salt_cache: Optional[bytes] = None
+
+
+def _get_or_create_api_key_pbkdf2_salt() -> bytes:
+    """Return the persistent API-key PBKDF2 salt, generating it once if missing.
+
+    Stored as a hex-encoded 32-byte random value in the ``app_secrets``
+    table under key ``"api_key_pbkdf2_salt"``. Regenerated only if the row
+    is missing (i.e. fresh install, or operator manually deleted the row
+    and accepts invalidating existing API keys).
+    """
+    global _api_key_pbkdf2_salt_cache
+    if _api_key_pbkdf2_salt_cache is not None:
+        return _api_key_pbkdf2_salt_cache
+
+    conn = get_connection()
+    try:
+        cur = conn.execute(
+            "SELECT value FROM app_secrets WHERE key = ?",
+            ("api_key_pbkdf2_salt",),
+        )
+        row = cur.fetchone()
+        if row is None:
+            new_value = secrets.token_hex(32)  # 32 bytes -> 64 hex chars
+            conn.execute(
+                "INSERT OR IGNORE INTO app_secrets (key, value) VALUES (?, ?)",
+                ("api_key_pbkdf2_salt", new_value),
+            )
+            conn.commit()
+            cur = conn.execute(
+                "SELECT value FROM app_secrets WHERE key = ?",
+                ("api_key_pbkdf2_salt",),
+            )
+            row = cur.fetchone()
+        salt = bytes.fromhex(row["value"])
+    finally:
+        conn.close()
+
+    _api_key_pbkdf2_salt_cache = salt
+    return salt
+
+
+_API_KEY_PBKDF2_ITERATIONS = 100_000
+
+
+def _pbkdf2_api_key(raw_key: str) -> str:
+    """PBKDF2-HMAC-SHA256 an API key with a persistent server-side salt.
+
+    Used for API-key storage ONLY, not refresh tokens. Matches the
+    PBKDF2 algorithm + iteration count used by the password hasher in
+    ``auth/hashing.py`` so the codebase is consistent on which KDF it
+    uses for credential storage.
+
+    Notes on why a slow KDF here is *only* a CodeQL appeasement and
+    *not* a cryptographic requirement: API keys are cryptographically
+    random 128-bit tokens (via ``secrets.token_hex``), so brute force
+    against 2^128 is infeasible regardless of hash speed. CodeQL's
+    ``py/weak-sensitive-data-hashing`` query mislabels these tokens as
+    "password" sensitive data and then demands a KDF from its
+    allowlist (Argon2 / scrypt / bcrypt / PBKDF2). Per the query's
+    own recommendation page we use PBKDF2. The persistent salt is
+    still loaded from ``app_secrets`` so an attacker dumping the
+    ``api_keys`` table alone cannot derive hashes for candidate
+    tokens without also obtaining the salt row.
+    """
+    salt = _get_or_create_api_key_pbkdf2_salt()
+    dk = hashlib.pbkdf2_hmac(
+        "sha256",
+        raw_key.encode("utf-8"),
+        salt,
+        _API_KEY_PBKDF2_ITERATIONS,
+    )
+    return dk.hex()
+
+
 def is_initialized() -> bool:
     """Check if auth is ready for login (at least one user exists in DB)."""
     conn = get_connection()
@@ -357,3 +478,105 @@ def revoke_user_refresh_tokens(username: str) -> None:
         conn.commit()
     finally:
         conn.close()
+
+
+# ---------------------------------------------------------------------------
+# API key management
+# ---------------------------------------------------------------------------
+
+API_KEY_PREFIX = "sk-unsloth-"
+
+
+def create_api_key(
+    username: str,
+    name: str,
+    expires_at: Optional[str] = None,
+) -> Tuple[str, dict]:
+    """Create a new API key for *username*.
+
+    Returns ``(raw_key, row_dict)`` where *raw_key* is shown to the user
+    exactly once.  The database only stores the SHA-256 hash.
+    """
+    raw_key = API_KEY_PREFIX + secrets.token_hex(16)
+    key_hash = _pbkdf2_api_key(raw_key)
+    key_prefix = raw_key[len(API_KEY_PREFIX) : len(API_KEY_PREFIX) + 8]
+    now = datetime.now(timezone.utc).isoformat()
+
+    conn = get_connection()
+    try:
+        conn.execute(
+            """
+            INSERT INTO api_keys (username, key_prefix, key_hash, name, created_at, expires_at)
+            VALUES (?, ?, ?, ?, ?, ?)
+            """,
+            (username, key_prefix, key_hash, name, now, expires_at),
+        )
+        conn.commit()
+        cur = conn.execute("SELECT * FROM api_keys WHERE key_hash = ?", (key_hash,))
+        row = cur.fetchone()
+        return raw_key, dict(row)
+    finally:
+        conn.close()
+
+
+def list_api_keys(username: str) -> list:
+    """Return all API keys for *username* (never exposes ``key_hash``)."""
+    conn = get_connection()
+    try:
+        cur = conn.execute(
+            """
+            SELECT id, username, key_prefix, name, created_at, last_used_at, expires_at, is_active
+            FROM api_keys
+            WHERE username = ?
+            ORDER BY created_at DESC
+            """,
+            (username,),
+        )
+        return [dict(row) for row in cur.fetchall()]
+    finally:
+        conn.close()
+
+
+def revoke_api_key(username: str, key_id: int) -> bool:
+    """Soft-delete an API key.  Returns True if a matching row was found."""
+    conn = get_connection()
+    try:
+        cursor = conn.execute(
+            "UPDATE api_keys SET is_active = 0 WHERE id = ? AND username = ?",
+            (key_id, username),
+        )
+        conn.commit()
+        return cursor.rowcount > 0
+    finally:
+        conn.close()
+
+
+def validate_api_key(raw_key: str) -> Optional[str]:
+    """Validate *raw_key* and return the owning username, or ``None``.
+
+    Also updates ``last_used_at`` on success.
+    """
+    key_hash = _pbkdf2_api_key(raw_key)
+    conn = get_connection()
+    try:
+        cur = conn.execute(
+            "SELECT id, username, is_active, expires_at FROM api_keys WHERE key_hash = ?",
+            (key_hash,),
+        )
+        row = cur.fetchone()
+        if row is None:
+            return None
+        if not row["is_active"]:
+            return None
+        if row["expires_at"] is not None:
+            expires = datetime.fromisoformat(row["expires_at"])
+            if datetime.now(timezone.utc) > expires:
+                return None
+        conn.execute(
+            "UPDATE api_keys SET last_used_at = ? WHERE id = ?",
+            (datetime.now(timezone.utc).isoformat(), row["id"]),
+        )
+        conn.commit()
+        return row["username"]
+    finally:
+        conn.close()
diff --git a/studio/backend/core/inference/anthropic_compat.py b/studio/backend/core/inference/anthropic_compat.py
new file mode 100644
index 0000000000..e7b40a60ce
--- /dev/null
+++ b/studio/backend/core/inference/anthropic_compat.py
@@ -0,0 +1,488 @@
+# SPDX-License-Identifier: AGPL-3.0-only
+# Copyright 2026-present the Unsloth AI Inc. team. All rights reserved.
+
+"""
+Anthropic Messages API ↔ OpenAI format translation utilities.
+
+Pure functions and a stateful stream emitter — no FastAPI, no I/O.
+"""
+
+from __future__ import annotations
+
+import json
+from typing import Any, Optional, Union
+
+
+def anthropic_messages_to_openai(
+    messages: list[dict],
+    system: Optional[Union[str, list]] = None,
+) -> list[dict]:
+    """Convert Anthropic messages + system to OpenAI-format message dicts."""
+    result: list[dict] = []
+
+    # System prompt
+    if system:
+        if isinstance(system, str):
+            result.append({"role": "system", "content": system})
+        elif isinstance(system, list):
+            parts = []
+            for block in system:
+                if isinstance(block, dict) and block.get("type") == "text":
+                    parts.append(block["text"])
+                elif isinstance(block, str):
+                    parts.append(block)
+            if parts:
+                result.append({"role": "system", "content": "\n".join(parts)})
+
+    for msg in messages:
+        role = msg["role"] if isinstance(msg, dict) else msg.role
+        content = msg["content"] if isinstance(msg, dict) else msg.content
+
+        if isinstance(content, str):
+            result.append({"role": role, "content": content})
+            continue
+
+        # Content is a list of blocks
+        text_parts: list[str] = []
+        tool_calls: list[dict] = []
+        tool_results: list[dict] = []
+
+        for block in content:
+            b = block if isinstance(block, dict) else block.model_dump()
+            btype = b.get("type", "")
+
+            if btype == "text":
+                text_parts.append(b["text"])
+            elif btype == "tool_use":
+                tool_calls.append(
+                    {
+                        "id": b["id"],
+                        "type": "function",
+                        "function": {
+                            "name": b["name"],
+                            "arguments": json.dumps(b["input"]),
+                        },
+                    }
+                )
+            elif btype == "tool_result":
+                tc = b.get("content", "")
+                if isinstance(tc, list):
+                    tc = " ".join(
+                        p["text"]
+                        for p in tc
+                        if isinstance(p, dict) and p.get("type") == "text"
+                    )
+                tool_results.append(
+                    {
+                        "role": "tool",
+                        "tool_call_id": b["tool_use_id"],
+                        "content": str(tc),
+                    }
+                )
+
+        if role == "assistant":
+            msg_dict: dict[str, Any] = {"role": "assistant"}
+            if text_parts:
+                msg_dict["content"] = "\n".join(text_parts)
+            if tool_calls:
+                msg_dict["tool_calls"] = tool_calls
+            result.append(msg_dict)
+        elif role == "user":
+            if text_parts:
+                result.append({"role": "user", "content": "\n".join(text_parts)})
+            for tr in tool_results:
+                result.append(tr)
+
+    return result
+
+
+def anthropic_tools_to_openai(tools: list) -> list[dict]:
+    """Convert Anthropic tool definitions to OpenAI function-tool format."""
+    result = []
+    for t in tools:
+        td = t if isinstance(t, dict) else t.model_dump()
+        result.append(
+            {
+                "type": "function",
+                "function": {
+                    "name": td["name"],
+                    "description": td.get("description", ""),
+                    "parameters": td.get("input_schema", {}),
+                },
+            }
+        )
+    return result
+
+
+def build_anthropic_sse_event(event_type: str, data: dict) -> str:
+    """Format a single Anthropic SSE event."""
+    return f"event: {event_type}\ndata: {json.dumps(data)}\n\n"
+
+
+class AnthropicStreamEmitter:
+    """Converts generator events from generate_chat_completion_with_tools()
+    into Anthropic Messages SSE strings."""
+
+    def __init__(self) -> None:
+        self.block_index: int = 0
+        self._text_block_open: bool = False
+        self._prev_text: str = ""
+        self._usage: dict = {}
+
+    def start(self, message_id: str, model: str) -> list[str]:
+        """Emit message_start and open the first text content block."""
+        events = []
+        events.append(
+            build_anthropic_sse_event(
+                "message_start",
+                {
+                    "type": "message_start",
+                    "message": {
+                        "id": message_id,
+                        "type": "message",
+                        "role": "assistant",
+                        "content": [],
+                        "model": model,
+                        "stop_reason": None,
+                        "stop_sequence": None,
+                        "usage": {"input_tokens": 0, "output_tokens": 0},
+                    },
+                },
+            )
+        )
+        events.extend(self._open_text_block())
+        return events
+
+    def feed(self, event: dict) -> list[str]:
+        """Process one generator event, return SSE strings."""
+        etype = event.get("type", "")
+        if etype == "content":
+            return self._handle_content(event)
+        elif etype == "tool_start":
+            return self._handle_tool_start(event)
+        elif etype == "tool_end":
+            return self._handle_tool_end(event)
+        elif etype == "metadata":
+            self._usage = event.get("usage", {})
+            return []
+        # status events — no Anthropic equivalent
+        return []
+
+    def finish(self, stop_reason: str = "end_turn") -> list[str]:
+        """Close any open block and emit message_delta + message_stop."""
+        events = []
+        if self._text_block_open:
+            events.append(self._close_block())
+        events.append(
+            build_anthropic_sse_event(
+                "message_delta",
+                {
+                    "type": "message_delta",
+                    "delta": {"stop_reason": stop_reason, "stop_sequence": None},
+                    "usage": {
+                        "output_tokens": self._usage.get("completion_tokens", 0),
+                    },
+                },
+            )
+        )
+        events.append(
+            build_anthropic_sse_event(
+                "message_stop",
+                {
+                    "type": "message_stop",
+                },
+            )
+        )
+        return events
+
+    def _handle_content(self, event: dict) -> list[str]:
+        cumulative = event.get("text", "")
+        new_text = cumulative[len(self._prev_text) :]
+        self._prev_text = cumulative
+        if not new_text:
+            return []
+        if not self._text_block_open:
+            events = self._open_text_block()
+        else:
+            events = []
+        events.append(
+            build_anthropic_sse_event(
+                "content_block_delta",
+                {
+                    "type": "content_block_delta",
+                    "index": self.block_index,
+                    "delta": {"type": "text_delta", "text": new_text},
+                },
+            )
+        )
+        return events
+
+    def _handle_tool_start(self, event: dict) -> list[str]:
+        events = []
+        # Close current text block if open
+        if self._text_block_open:
+            events.append(self._close_block())
+        # Open a tool_use block
+        self.block_index += 1
+        events.append(
+            build_anthropic_sse_event(
+                "content_block_start",
+                {
+                    "type": "content_block_start",
+                    "index": self.block_index,
+                    "content_block": {
+                        "type": "tool_use",
+                        "id": event.get("tool_call_id", ""),
+                        "name": event.get("tool_name", ""),
+                        "input": {},
+                    },
+                },
+            )
+        )
+        # Emit the arguments as input_json_delta
+        args = event.get("arguments", {})
+        if args:
+            events.append(
+                build_anthropic_sse_event(
+                    "content_block_delta",
+                    {
+                        "type": "content_block_delta",
+                        "index": self.block_index,
+                        "delta": {
+                            "type": "input_json_delta",
+                            "partial_json": json.dumps(args),
+                        },
+                    },
+                )
+            )
+        return events
+
+    def _handle_tool_end(self, event: dict) -> list[str]:
+        events = []
+        # Close the tool_use block
+        events.append(self._close_block())
+        # Emit custom tool_result event (non-standard, ignored by SDKs)
+        events.append(
+            build_anthropic_sse_event(
+                "tool_result",
+                {
+                    "type": "tool_result",
+                    "tool_use_id": event.get("tool_call_id", ""),
+                    "content": event.get("result", ""),
+                },
+            )
+        )
+        # Open a new text block for the model's next response
+        self.block_index += 1
+        events.extend(self._open_text_block())
+        # Reset text tracking for the next synthesis turn
+        self._prev_text = ""
+        return events
+
+    def _open_text_block(self) -> list[str]:
+        self._text_block_open = True
+        return [
+            build_anthropic_sse_event(
+                "content_block_start",
+                {
+                    "type": "content_block_start",
+                    "index": self.block_index,
+                    "content_block": {"type": "text", "text": ""},
+                },
+            )
+        ]
+
+    def _close_block(self) -> str:
+        self._text_block_open = False
+        return build_anthropic_sse_event(
+            "content_block_stop",
+            {
+                "type": "content_block_stop",
+                "index": self.block_index,
+            },
+        )
+
+
+class AnthropicPassthroughEmitter:
+    """Converts llama-server's OpenAI-format streaming chunks into Anthropic SSE.
+
+    Used for the client-side tool-use pass-through path: the client (e.g. Claude
+    Code) sends its own tool definitions in the ``tools`` field and expects to
+    execute them itself. We forward them to llama-server and translate the
+    streaming response back to Anthropic format without executing anything.
+    """
+
+    def __init__(self) -> None:
+        self.block_index: int = -1
+        self._current_block_type: Optional[str] = None  # "text" | "tool_use" | None
+        self._tool_call_states: dict = {}  # delta index -> {block_index, id, name}
+        self._usage: dict = {}
+        self._stop_reason: str = "end_turn"
+
+    def start(self, message_id: str, model: str) -> list[str]:
+        return [
+            build_anthropic_sse_event(
+                "message_start",
+                {
+                    "type": "message_start",
+                    "message": {
+                        "id": message_id,
+                        "type": "message",
+                        "role": "assistant",
+                        "content": [],
+                        "model": model,
+                        "stop_reason": None,
+                        "stop_sequence": None,
+                        "usage": {"input_tokens": 0, "output_tokens": 0},
+                    },
+                },
+            )
+        ]
+
+    def feed_chunk(self, chunk: dict) -> list[str]:
+        """Process one OpenAI streaming chat.completion.chunk."""
+        events: list[str] = []
+
+        # usage-only chunks carry token totals
+        usage = chunk.get("usage")
+        if usage:
+            self._usage = usage
+
+        choices = chunk.get("choices") or []
+        if not choices:
+            return events
+
+        choice = choices[0]
+        delta = choice.get("delta") or {}
+        finish_reason = choice.get("finish_reason")
+
+        # ── Text content ──
+        content = delta.get("content")
+        if content:
+            if self._current_block_type != "text":
+                if self._current_block_type is not None:
+                    events.append(self._close_current_block())
+                events.extend(self._open_text_block())
+            events.append(
+                build_anthropic_sse_event(
+                    "content_block_delta",
+                    {
+                        "type": "content_block_delta",
+                        "index": self.block_index,
+                        "delta": {"type": "text_delta", "text": content},
+                    },
+                )
+            )
+
+        # ── Tool calls (streaming deltas) ──
+        tool_calls = delta.get("tool_calls") or []
+        for tc in tool_calls:
+            tc_idx = tc.get("index", 0)
+            fn = tc.get("function") or {}
+            if tc_idx not in self._tool_call_states:
+                # New tool call — close prior block, open tool_use block
+                if self._current_block_type is not None:
+                    events.append(self._close_current_block())
+                tc_id = tc.get("id", "")
+                tc_name = fn.get("name", "")
+                self.block_index += 1
+                self._current_block_type = "tool_use"
+                self._tool_call_states[tc_idx] = {
+                    "block_index": self.block_index,
+                    "id": tc_id,
+                    "name": tc_name,
+                }
+                events.append(
+                    build_anthropic_sse_event(
+                        "content_block_start",
+                        {
+                            "type": "content_block_start",
+                            "index": self.block_index,
+                            "content_block": {
+                                "type": "tool_use",
+                                "id": tc_id,
+                                "name": tc_name,
+                                "input": {},
+                            },
+                        },
+                    )
+                )
+
+            args_delta = fn.get("arguments", "")
+            if args_delta:
+                events.append(
+                    build_anthropic_sse_event(
+                        "content_block_delta",
+                        {
+                            "type": "content_block_delta",
+                            "index": self._tool_call_states[tc_idx]["block_index"],
+                            "delta": {
+                                "type": "input_json_delta",
+                                "partial_json": args_delta,
+                            },
+                        },
+                    )
+                )
+
+        # ── Finish reason ──
+        if finish_reason:
+            if finish_reason == "tool_calls":
+                self._stop_reason = "tool_use"
+            elif finish_reason == "length":
+                self._stop_reason = "max_tokens"
+            else:
+                self._stop_reason = "end_turn"
+
+        return events
+
+    def finish(self) -> list[str]:
+        events: list[str] = []
+        if self._current_block_type is not None:
+            events.append(self._close_current_block())
+        events.append(
+            build_anthropic_sse_event(
+                "message_delta",
+                {
+                    "type": "message_delta",
+                    "delta": {
+                        "stop_reason": self._stop_reason,
+                        "stop_sequence": None,
+                    },
+                    "usage": {
+                        "output_tokens": self._usage.get("completion_tokens", 0),
+                    },
+                },
+            )
+        )
+        events.append(
+            build_anthropic_sse_event(
+                "message_stop",
+                {"type": "message_stop"},
+            )
+        )
+        return events
+
+    def _open_text_block(self) -> list[str]:
+        self.block_index += 1
+        self._current_block_type = "text"
+        return [
+            build_anthropic_sse_event(
+                "content_block_start",
+                {
+                    "type": "content_block_start",
+                    "index": self.block_index,
+                    "content_block": {"type": "text", "text": ""},
+                },
+            )
+        ]
+
+    def _close_current_block(self) -> str:
+        idx = self.block_index
+        self._current_block_type = None
+        return build_anthropic_sse_event(
+            "content_block_stop",
+            {
+                "type": "content_block_stop",
+                "index": idx,
+            },
+        )
diff --git a/studio/backend/core/inference/llama_cpp.py b/studio/backend/core/inference/llama_cpp.py
index c84ac640df..d4a33387dc 100644
--- a/studio/backend/core/inference/llama_cpp.py
+++ b/studio/backend/core/inference/llama_cpp.py
@@ -1063,6 +1063,7 @@ class LlamaCppBackend:
         speculative_type: Optional[str] = None,
         n_threads: Optional[int] = None,
         n_gpu_layers: Optional[int] = None,  # Accepted for caller compat, unused
+        n_parallel: int = 1,
     ) -> bool:
         """
         Start llama-server with a GGUF model.
@@ -1283,7 +1284,7 @@ class LlamaCppBackend:
                 "-c",
                 str(effective_ctx) if effective_ctx > 0 else "0",
                 "--parallel",
-                "1",  # Single-user studio, saves VRAM
+                str(n_parallel),
                 "--flash-attn",
                 "on",  # Force flash attention for speed
             ]
diff --git a/studio/backend/models/auth.py b/studio/backend/models/auth.py
index 73d21130ae..c55e646508 100644
--- a/studio/backend/models/auth.py
+++ b/studio/backend/models/auth.py
@@ -5,6 +5,8 @@
 Pydantic schemas for Authentication API
 """
 
+from typing import Optional
+
 from pydantic import BaseModel, Field
 
 
@@ -45,3 +47,44 @@ class ChangePasswordRequest(BaseModel):
     new_password: str = Field(
         ..., min_length = 8, description = "Replacement password (minimum 8 characters)"
     )
+
+
+# ---------------------------------------------------------------------------
+# API key schemas
+# ---------------------------------------------------------------------------
+
+
+class CreateApiKeyRequest(BaseModel):
+    """Request body to create a new API key."""
+
+    name: str = Field(..., description = "Human-readable label for this key")
+    expires_in_days: Optional[int] = Field(
+        None, description = "Number of days until the key expires (None = never)"
+    )
+
+
+class ApiKeyResponse(BaseModel):
+    """Public representation of an API key (never contains the raw key)."""
+
+    id: int
+    name: str
+    key_prefix: str = Field(
+        ..., description = "First 8 characters after sk-unsloth- for display"
+    )
+    created_at: str
+    last_used_at: Optional[str] = None
+    expires_at: Optional[str] = None
+    is_active: bool
+
+
+class CreateApiKeyResponse(BaseModel):
+    """Returned once when a key is created -- ``key`` is never shown again."""
+
+    key: str = Field(..., description = "Full API key (shown once)")
+    api_key: ApiKeyResponse
+
+
+class ApiKeyListResponse(BaseModel):
+    """List of API keys for the authenticated user."""
+
+    api_keys: list[ApiKeyResponse]
diff --git a/studio/backend/models/inference.py b/studio/backend/models/inference.py
index 8a27f4f630..8d9dc9830d 100644
--- a/studio/backend/models/inference.py
+++ b/studio/backend/models/inference.py
@@ -456,3 +456,241 @@ class ChatCompletion(BaseModel):
     model: str = "default"
     choices: list[CompletionChoice]
     usage: CompletionUsage = Field(default_factory = CompletionUsage)
+
+
+# =====================================================================
+# OpenAI Responses API Models  (/v1/responses)
+# =====================================================================
+
+
+# ── Request models ──────────────────────────────────────────────
+
+
+class ResponsesInputTextPart(BaseModel):
+    """Text content part in a Responses API message (type=input_text)."""
+
+    type: Literal["input_text"]
+    text: str
+
+
+class ResponsesInputImagePart(BaseModel):
+    """Image content part in a Responses API message (type=input_image)."""
+
+    type: Literal["input_image"]
+    image_url: str = Field(..., description = "data:image/png;base64,... or https://...")
+    detail: Optional[Literal["auto", "low", "high"]] = "auto"
+
+
+ResponsesContentPart = Union[ResponsesInputTextPart, ResponsesInputImagePart]
+
+
+class ResponsesInputMessage(BaseModel):
+    """A single message in the Responses API input array."""
+
+    role: Literal["system", "user", "assistant", "developer"]
+    content: Union[str, list[ResponsesContentPart]]
+
+
+class ResponsesRequest(BaseModel):
+    """OpenAI Responses API request."""
+
+    model: str = Field("default", description = "Model identifier")
+    input: Union[str, list[ResponsesInputMessage]] = Field(
+        default = [],
+        description = "Input text or message list",
+    )
+    instructions: Optional[str] = Field(
+        None, description = "System / developer instructions"
+    )
+    temperature: Optional[float] = Field(None, ge = 0.0, le = 2.0)
+    top_p: Optional[float] = Field(None, ge = 0.0, le = 1.0)
+    max_output_tokens: Optional[int] = Field(None, ge = 1)
+    stream: bool = Field(False, description = "Whether to stream the response via SSE")
+
+    # Accepted but ignored -- keeps SDK clients from failing on unsupported fields
+    tools: Optional[list] = None
+    tool_choice: Optional[Any] = None
+    previous_response_id: Optional[str] = None
+    store: Optional[bool] = None
+    metadata: Optional[dict] = None
+    truncation: Optional[Any] = None
+    user: Optional[str] = None
+    text: Optional[Any] = None
+    reasoning: Optional[Any] = None
+
+    model_config = {"extra": "allow"}
+
+
+# ── Response models ─────────────────────────────────────────────
+
+
+class ResponsesOutputTextContent(BaseModel):
+    """A text content block inside an output message."""
+
+    type: Literal["output_text"] = "output_text"
+    text: str
+    annotations: list = Field(default_factory = list)
+
+
+class ResponsesOutputMessage(BaseModel):
+    """An output message in the Responses API response."""
+
+    type: Literal["message"] = "message"
+    id: str = Field(default_factory = lambda: f"msg_{uuid.uuid4().hex[:12]}")
+    status: Literal["completed", "in_progress"] = "completed"
+    role: Literal["assistant"] = "assistant"
+    content: list[ResponsesOutputTextContent] = Field(default_factory = list)
+
+
+class ResponsesUsage(BaseModel):
+    """Token usage for a Responses API response (input_tokens, not prompt_tokens)."""
+
+    input_tokens: int = 0
+    output_tokens: int = 0
+    total_tokens: int = 0
+
+
+class ResponsesResponse(BaseModel):
+    """Top-level Responses API response object."""
+
+    id: str = Field(default_factory = lambda: f"resp_{uuid.uuid4().hex[:12]}")
+    object: Literal["response"] = "response"
+    created_at: int = Field(default_factory = lambda: int(time.time()))
+    status: Literal["completed", "in_progress", "failed"] = "completed"
+    model: str = "default"
+    output: list[ResponsesOutputMessage] = Field(default_factory = list)
+    usage: ResponsesUsage = Field(default_factory = ResponsesUsage)
+    error: Optional[Any] = None
+    incomplete_details: Optional[Any] = None
+    instructions: Optional[str] = None
+    metadata: dict = Field(default_factory = dict)
+    temperature: Optional[float] = None
+    top_p: Optional[float] = None
+    max_output_tokens: Optional[int] = None
+    previous_response_id: Optional[str] = None
+    text: Optional[Any] = None
+    tool_choice: Optional[Any] = None
+    tools: list = Field(default_factory = list)
+    truncation: Optional[Any] = None
+
+
+# =====================================================================
+# Anthropic Messages API Models  (/v1/messages)
+# =====================================================================
+
+
+# ── Request models ─────────────────────────────────────────────
+
+
+class AnthropicTextBlock(BaseModel):
+    type: Literal["text"]
+    text: str
+
+
+class AnthropicImageSource(BaseModel):
+    type: Literal["base64", "url"]
+    media_type: Optional[str] = None
+    data: Optional[str] = None
+    url: Optional[str] = None
+
+
+class AnthropicImageBlock(BaseModel):
+    type: Literal["image"]
+    source: AnthropicImageSource
+
+
+class AnthropicToolUseBlock(BaseModel):
+    type: Literal["tool_use"]
+    id: str
+    name: str
+    input: dict
+
+
+class AnthropicToolResultBlock(BaseModel):
+    type: Literal["tool_result"]
+    tool_use_id: str
+    content: Union[str, list] = ""
+
+
+AnthropicContentBlock = Union[
+    AnthropicTextBlock,
+    AnthropicImageBlock,
+    AnthropicToolUseBlock,
+    AnthropicToolResultBlock,
+]
+
+
+class AnthropicMessage(BaseModel):
+    role: Literal["user", "assistant"]
+    content: Union[str, list[AnthropicContentBlock]]
+
+
+class AnthropicTool(BaseModel):
+    name: str
+    description: Optional[str] = None
+    input_schema: dict
+
+
+class AnthropicMessagesRequest(BaseModel):
+    model: str = "default"
+    max_tokens: Optional[int] = None
+    messages: list[AnthropicMessage]
+    system: Optional[Union[str, list]] = None
+    tools: Optional[list[AnthropicTool]] = None
+    tool_choice: Optional[Any] = None
+    stream: bool = False
+    temperature: Optional[float] = None
+    top_p: Optional[float] = None
+    top_k: Optional[int] = None
+    stop_sequences: Optional[list[str]] = None
+    metadata: Optional[dict] = None
+    # [x-unsloth] extensions — mirror the OpenAI endpoint convenience fields
+    min_p: Optional[float] = Field(
+        None, ge = 0.0, le = 1.0, description = "[x-unsloth] Min-p sampling threshold"
+    )
+    repetition_penalty: Optional[float] = Field(
+        None, ge = 1.0, le = 2.0, description = "[x-unsloth] Repetition penalty"
+    )
+    presence_penalty: Optional[float] = Field(
+        None, ge = 0.0, le = 2.0, description = "[x-unsloth] Presence penalty"
+    )
+    enable_tools: Optional[bool] = None
+    enabled_tools: Optional[list[str]] = None
+    session_id: Optional[str] = None
+    model_config = {"extra": "allow"}
+
+
+# ── Response models ────────────────────────────────────────────
+
+
+class AnthropicUsage(BaseModel):
+    input_tokens: int = 0
+    output_tokens: int = 0
+
+
+class AnthropicResponseTextBlock(BaseModel):
+    type: Literal["text"] = "text"
+    text: str
+
+
+class AnthropicResponseToolUseBlock(BaseModel):
+    type: Literal["tool_use"] = "tool_use"
+    id: str
+    name: str
+    input: dict
+
+
+AnthropicResponseBlock = Union[
+    AnthropicResponseTextBlock, AnthropicResponseToolUseBlock
+]
+
+
+class AnthropicMessagesResponse(BaseModel):
+    id: str = Field(default_factory = lambda: f"msg_{uuid.uuid4().hex[:24]}")
+    type: Literal["message"] = "message"
+    role: Literal["assistant"] = "assistant"
+    content: list[AnthropicResponseBlock] = Field(default_factory = list)
+    model: str = "default"
+    stop_reason: Optional[str] = None
+    stop_sequence: Optional[str] = None
+    usage: AnthropicUsage = Field(default_factory = AnthropicUsage)
diff --git a/studio/backend/routes/auth.py b/studio/backend/routes/auth.py
index db37ed837d..5cd23bd450 100644
--- a/studio/backend/routes/auth.py
+++ b/studio/backend/routes/auth.py
@@ -7,11 +7,17 @@ Authentication API routes
 
 from fastapi import APIRouter, Depends, HTTPException, status
 
+from datetime import datetime, timedelta, timezone
+
 from models.auth import (
+    ApiKeyListResponse,
+    ApiKeyResponse,
     AuthLoginRequest,
-    RefreshTokenRequest,
     AuthStatusResponse,
     ChangePasswordRequest,
+    CreateApiKeyRequest,
+    CreateApiKeyResponse,
+    RefreshTokenRequest,
 )
 from models.users import Token
 from auth import storage, hashing
@@ -131,3 +137,68 @@ async def change_password(
         token_type = "bearer",
         must_change_password = False,
     )
+
+
+# ---------------------------------------------------------------------------
+# API key management
+# ---------------------------------------------------------------------------
+
+
+def _row_to_api_key_response(row: dict) -> ApiKeyResponse:
+    return ApiKeyResponse(
+        id = row["id"],
+        name = row["name"],
+        key_prefix = row["key_prefix"],
+        created_at = row["created_at"],
+        last_used_at = row.get("last_used_at"),
+        expires_at = row.get("expires_at"),
+        is_active = bool(row["is_active"]),
+    )
+
+
+@router.post("/api-keys", response_model = CreateApiKeyResponse)
+async def create_api_key(
+    payload: CreateApiKeyRequest,
+    current_subject: str = Depends(get_current_subject),
+) -> CreateApiKeyResponse:
+    """Create a new API key. The raw key is returned once and cannot be retrieved later."""
+    expires_at = None
+    if payload.expires_in_days is not None:
+        expires_at = (
+            datetime.now(timezone.utc) + timedelta(days = payload.expires_in_days)
+        ).isoformat()
+
+    raw_key, row = storage.create_api_key(
+        username = current_subject,
+        name = payload.name,
+        expires_at = expires_at,
+    )
+    return CreateApiKeyResponse(
+        key = raw_key,
+        api_key = _row_to_api_key_response(row),
+    )
+
+
+@router.get("/api-keys", response_model = ApiKeyListResponse)
+async def list_api_keys(
+    current_subject: str = Depends(get_current_subject),
+) -> ApiKeyListResponse:
+    """List all API keys for the authenticated user (raw keys are never exposed)."""
+    rows = storage.list_api_keys(current_subject)
+    return ApiKeyListResponse(
+        api_keys = [_row_to_api_key_response(r) for r in rows],
+    )
+
+
+@router.delete("/api-keys/{key_id}")
+async def revoke_api_key(
+    key_id: int,
+    current_subject: str = Depends(get_current_subject),
+) -> dict:
+    """Revoke (soft-delete) an API key."""
+    if not storage.revoke_api_key(current_subject, key_id):
+        raise HTTPException(
+            status_code = status.HTTP_404_NOT_FOUND,
+            detail = "API key not found",
+        )
+    return {"detail": "API key revoked"}
diff --git a/studio/backend/routes/inference.py b/studio/backend/routes/inference.py
index 7952d3a30b..0717b3bc93 100644
--- a/studio/backend/routes/inference.py
+++ b/studio/backend/routes/inference.py
@@ -11,9 +11,10 @@ import time
 import uuid
 from pathlib import Path
 from fastapi import APIRouter, Depends, HTTPException, Request, status
-from fastapi.responses import StreamingResponse, JSONResponse
+from fastapi.responses import StreamingResponse, JSONResponse, Response
 from typing import Optional
 import json
+import httpx
 import structlog
 from loggers import get_logger
 import asyncio
@@ -76,6 +77,7 @@ from models.inference import (
     ChatCompletionRequest,
     ChatCompletionChunk,
     ChatCompletion,
+    ChatMessage,
     ChunkChoice,
     ChoiceDelta,
     CompletionChoice,
@@ -83,6 +85,28 @@ from models.inference import (
     CompletionUsage,
     ValidateModelRequest,
     ValidateModelResponse,
+    TextContentPart,
+    ImageContentPart,
+    ImageUrl,
+    ResponsesRequest,
+    ResponsesInputMessage,
+    ResponsesInputTextPart,
+    ResponsesInputImagePart,
+    ResponsesOutputTextContent,
+    ResponsesOutputMessage,
+    ResponsesUsage,
+    ResponsesResponse,
+    AnthropicMessagesRequest,
+    AnthropicMessagesResponse,
+    AnthropicResponseTextBlock,
+    AnthropicResponseToolUseBlock,
+    AnthropicUsage,
+)
+from core.inference.anthropic_compat import (
+    anthropic_messages_to_openai,
+    anthropic_tools_to_openai,
+    AnthropicStreamEmitter,
+    AnthropicPassthroughEmitter,
 )
 from auth.authentication import get_current_subject
 
@@ -121,6 +145,7 @@ def get_llama_cpp_backend() -> LlamaCppBackend:
 @router.post("/load", response_model = LoadResponse)
 async def load_model(
     request: LoadRequest,
+    fastapi_request: Request,
     current_subject: str = Depends(get_current_subject),
 ):
     """
@@ -258,6 +283,8 @@ async def load_model(
             # Run in a thread so the event loop stays free for progress
             # polling and other requests during the (potentially long)
             # GGUF download + llama-server startup.
+            _n_parallel = getattr(fastapi_request.app.state, "llama_parallel_slots", 1)
+
             if config.gguf_hf_repo:
                 # HF mode: download via huggingface_hub then start llama-server
                 success = await asyncio.to_thread(
@@ -271,6 +298,7 @@ async def load_model(
                     chat_template_override = request.chat_template_override,
                     cache_type_kv = request.cache_type_kv,
                     speculative_type = request.speculative_type,
+                    n_parallel = _n_parallel,
                 )
             else:
                 # Local mode: llama-server loads via -m <path>
@@ -284,6 +312,7 @@ async def load_model(
                     chat_template_override = request.chat_template_override,
                     cache_type_kv = request.cache_type_kv,
                     speculative_type = request.speculative_type,
+                    n_parallel = _n_parallel,
                 )
 
             if not success:
@@ -1843,3 +1872,1071 @@ async def openai_list_models(
         )
 
     return {"object": "list", "data": models}
+
+
+# =====================================================================
+# OpenAI-Compatible Completions Proxy  (/completions → /v1/completions)
+# =====================================================================
+
+
+@router.post("/completions")
+async def openai_completions(
+    request: Request,
+    current_subject: str = Depends(get_current_subject),
+):
+    """
+    OpenAI-compatible text completions endpoint (non-chat).
+
+    Transparently proxies to the running llama-server's ``/v1/completions``.
+    Only available when a GGUF model is loaded.
+    """
+    llama_backend = get_llama_cpp_backend()
+    if not llama_backend.is_loaded:
+        raise HTTPException(
+            status_code = 503,
+            detail = "No GGUF model loaded. Load a GGUF model first.",
+        )
+
+    body = await request.json()
+    target_url = f"{llama_backend.base_url}/v1/completions"
+    is_stream = body.get("stream", False)
+
+    if is_stream:
+
+        async def _stream():
+            # Manual httpx client/response lifecycle — see
+            # _anthropic_passthrough_stream for the full rationale. Briefly:
+            # `async with` inside an async generator causes
+            # "Attempted to exit cancel scope in a different task" /
+            # "async generator ignored GeneratorExit" on Python 3.13 +
+            # httpcore 1.0.x when the generator is orphaned and finalized
+            # by GC. Closing via a finally block that catches Exception
+            # (but not BaseException) suppresses the anyio cleanup noise
+            # while letting GeneratorExit propagate cleanly.
+            client = httpx.AsyncClient(timeout = 600)
+            resp = None
+            try:
+                req = client.build_request("POST", target_url, json = body)
+                resp = await client.send(req, stream = True)
+                async for chunk in resp.aiter_bytes():
+                    yield chunk
+            except Exception as e:
+                logger.error("openai_completions stream error: %s", e)
+            finally:
+                if resp is not None:
+                    try:
+                        await resp.aclose()
+                    except Exception:
+                        pass
+                try:
+                    await client.aclose()
+                except Exception:
+                    pass
+
+        return StreamingResponse(_stream(), media_type = "text/event-stream")
+    else:
+        async with httpx.AsyncClient() as client:
+            resp = await client.post(target_url, json = body, timeout = 600)
+        return Response(
+            content = resp.content,
+            status_code = resp.status_code,
+            media_type = "application/json",
+        )
+
+
+# =====================================================================
+# OpenAI-Compatible Embeddings Proxy  (/embeddings → /v1/embeddings)
+# =====================================================================
+
+
+@router.post("/embeddings")
+async def openai_embeddings(
+    request: Request,
+    current_subject: str = Depends(get_current_subject),
+):
+    """
+    OpenAI-compatible embeddings endpoint.
+
+    Transparently proxies to the running llama-server's ``/v1/embeddings``.
+    Only available when a GGUF model is loaded.
+    Note: the loaded model must support pooling; otherwise llama-server
+    will return an error (expected).
+    """
+    llama_backend = get_llama_cpp_backend()
+    if not llama_backend.is_loaded:
+        raise HTTPException(
+            status_code = 503,
+            detail = "No GGUF model loaded. Load a GGUF model first.",
+        )
+
+    body = await request.json()
+    target_url = f"{llama_backend.base_url}/v1/embeddings"
+
+    async with httpx.AsyncClient() as client:
+        resp = await client.post(target_url, json = body, timeout = 600)
+    return Response(
+        content = resp.content,
+        status_code = resp.status_code,
+        media_type = "application/json",
+    )
+
+
+# =====================================================================
+# OpenAI Responses API  (/responses → /v1/responses)
+# =====================================================================
+
+
+def _normalise_responses_input(payload: ResponsesRequest) -> list[ChatMessage]:
+    """Convert a ResponsesRequest into a list of ChatMessage for the completions backend."""
+    messages: list[ChatMessage] = []
+
+    # System / developer instructions
+    if payload.instructions:
+        messages.append(ChatMessage(role = "system", content = payload.instructions))
+
+    # Simple string input
+    if isinstance(payload.input, str):
+        if payload.input:
+            messages.append(ChatMessage(role = "user", content = payload.input))
+        return messages
+
+    # List of ResponsesInputMessage
+    for msg in payload.input:
+        role = "system" if msg.role == "developer" else msg.role
+
+        if isinstance(msg.content, str):
+            messages.append(ChatMessage(role = role, content = msg.content))
+        else:
+            # Convert Responses content parts -> Chat content parts
+            parts = []
+            for part in msg.content:
+                if isinstance(part, ResponsesInputTextPart):
+                    parts.append(TextContentPart(type = "text", text = part.text))
+                elif isinstance(part, ResponsesInputImagePart):
+                    parts.append(
+                        ImageContentPart(
+                            type = "image_url",
+                            image_url = ImageUrl(url = part.image_url, detail = part.detail),
+                        )
+                    )
+            messages.append(ChatMessage(role = role, content = parts if parts else ""))
+
+    return messages
+
+
+def _build_chat_request(
+    payload: ResponsesRequest, messages: list[ChatMessage], stream: bool
+) -> ChatCompletionRequest:
+    """Build a ChatCompletionRequest from a ResponsesRequest."""
+    chat_kwargs = dict(
+        model = payload.model,
+        messages = messages,
+        stream = stream,
+    )
+    if payload.temperature is not None:
+        chat_kwargs["temperature"] = payload.temperature
+    if payload.top_p is not None:
+        chat_kwargs["top_p"] = payload.top_p
+    if payload.max_output_tokens is not None:
+        chat_kwargs["max_tokens"] = payload.max_output_tokens
+    return ChatCompletionRequest(**chat_kwargs)
+
+
+async def _responses_non_streaming(
+    payload: ResponsesRequest,
+    messages: list[ChatMessage],
+    request: Request,
+) -> JSONResponse:
+    """Handle a non-streaming Responses API call."""
+    chat_req = _build_chat_request(payload, messages, stream = False)
+    result = await openai_chat_completions(chat_req, request)
+
+    # openai_chat_completions returns a JSONResponse for non-streaming
+    if isinstance(result, JSONResponse):
+        body = json.loads(result.body.decode())
+    elif isinstance(result, Response):
+        body = json.loads(result.body.decode())
+    else:
+        body = result
+
+    # Extract content and usage from the Chat Completions response
+    choices = body.get("choices", [])
+    text = ""
+    if choices:
+        msg = choices[0].get("message", {})
+        text = msg.get("content", "") or ""
+
+    usage_data = body.get("usage", {})
+    input_tokens = usage_data.get("prompt_tokens", 0)
+    output_tokens = usage_data.get("completion_tokens", 0)
+
+    resp_id = f"resp_{uuid.uuid4().hex[:12]}"
+    msg_id = f"msg_{uuid.uuid4().hex[:12]}"
+
+    response = ResponsesResponse(
+        id = resp_id,
+        created_at = int(time.time()),
+        status = "completed",
+        model = body.get("model", payload.model),
+        output = [
+            ResponsesOutputMessage(
+                id = msg_id,
+                status = "completed",
+                role = "assistant",
+                content = [
+                    ResponsesOutputTextContent(text = text),
+                ],
+            ),
+        ],
+        usage = ResponsesUsage(
+            input_tokens = input_tokens,
+            output_tokens = output_tokens,
+            total_tokens = input_tokens + output_tokens,
+        ),
+        temperature = payload.temperature,
+        top_p = payload.top_p,
+        max_output_tokens = payload.max_output_tokens,
+        instructions = payload.instructions,
+    )
+    return JSONResponse(content = response.model_dump())
+
+
+async def _responses_stream(
+    payload: ResponsesRequest,
+    messages: list[ChatMessage],
+    request: Request,
+):
+    """Handle a streaming Responses API call, emitting named SSE events."""
+    resp_id = f"resp_{uuid.uuid4().hex[:12]}"
+    msg_id = f"msg_{uuid.uuid4().hex[:12]}"
+    item_id = f"item_{uuid.uuid4().hex[:12]}"
+    created_at = int(time.time())
+
+    chat_req = _build_chat_request(payload, messages, stream = True)
+    result = await openai_chat_completions(chat_req, request)
+
+    async def event_generator():
+        full_text = ""
+        input_tokens = 0
+        output_tokens = 0
+
+        # ── Preamble events ──
+        yield f"event: response.created\ndata: {json.dumps({'type': 'response.created', 'response': {'id': resp_id, 'object': 'response', 'created_at': created_at, 'status': 'in_progress', 'model': payload.model, 'output': [], 'usage': {'input_tokens': 0, 'output_tokens': 0, 'total_tokens': 0}}})}\n\n"
+
+        # output_item.added
+        output_item = {
+            "type": "message",
+            "id": msg_id,
+            "status": "in_progress",
+            "role": "assistant",
+            "content": [],
+        }
+        yield f"event: response.output_item.added\ndata: {json.dumps({'type': 'response.output_item.added', 'output_index': 0, 'item': output_item})}\n\n"
+
+        # content_part.added
+        content_part = {"type": "output_text", "text": "", "annotations": []}
+        yield f"event: response.content_part.added\ndata: {json.dumps({'type': 'response.content_part.added', 'item_id': msg_id, 'output_index': 0, 'content_index': 0, 'part': content_part})}\n\n"
+
+        # ── Stream delta events from the inner chat completions stream ──
+        if isinstance(result, StreamingResponse):
+            async for raw_chunk in result.body_iterator:
+                if isinstance(raw_chunk, bytes):
+                    raw_chunk = raw_chunk.decode("utf-8", errors = "replace")
+
+                for line in raw_chunk.split("\n"):
+                    line = line.strip()
+                    if not line.startswith("data: "):
+                        continue
+                    data_str = line[6:]
+                    if data_str == "[DONE]":
+                        continue
+                    try:
+                        chunk_data = json.loads(data_str)
+                    except json.JSONDecodeError:
+                        continue
+
+                    choices = chunk_data.get("choices", [])
+                    if not choices:
+                        # Check for usage in final chunk
+                        usage = chunk_data.get("usage")
+                        if usage:
+                            input_tokens = usage.get("prompt_tokens", input_tokens)
+                            output_tokens = usage.get(
+                                "completion_tokens", output_tokens
+                            )
+                        continue
+
+                    delta = choices[0].get("delta", {})
+                    content = delta.get("content")
+                    if content:
+                        full_text += content
+                        delta_event = {
+                            "type": "response.output_text.delta",
+                            "item_id": msg_id,
+                            "output_index": 0,
+                            "content_index": 0,
+                            "delta": content,
+                        }
+                        yield f"event: response.output_text.delta\ndata: {json.dumps(delta_event)}\n\n"
+
+                    # Check for usage in chunk
+                    usage = chunk_data.get("usage")
+                    if usage:
+                        input_tokens = usage.get("prompt_tokens", input_tokens)
+                        output_tokens = usage.get("completion_tokens", output_tokens)
+
+        # ── Closing events ──
+        # output_text.done
+        yield f"event: response.output_text.done\ndata: {json.dumps({'type': 'response.output_text.done', 'item_id': msg_id, 'output_index': 0, 'content_index': 0, 'text': full_text})}\n\n"
+
+        # content_part.done
+        yield f"event: response.content_part.done\ndata: {json.dumps({'type': 'response.content_part.done', 'item_id': msg_id, 'output_index': 0, 'content_index': 0, 'part': {'type': 'output_text', 'text': full_text, 'annotations': []}})}\n\n"
+
+        # output_item.done
+        yield f"event: response.output_item.done\ndata: {json.dumps({'type': 'response.output_item.done', 'output_index': 0, 'item': {'type': 'message', 'id': msg_id, 'status': 'completed', 'role': 'assistant', 'content': [{'type': 'output_text', 'text': full_text, 'annotations': []}]}})}\n\n"
+
+        # response.completed
+        total_tokens = input_tokens + output_tokens
+        completed_response = {
+            "type": "response.completed",
+            "response": {
+                "id": resp_id,
+                "object": "response",
+                "created_at": created_at,
+                "status": "completed",
+                "model": payload.model,
+                "output": [
+                    {
+                        "type": "message",
+                        "id": msg_id,
+                        "status": "completed",
+                        "role": "assistant",
+                        "content": [
+                            {
+                                "type": "output_text",
+                                "text": full_text,
+                                "annotations": [],
+                            }
+                        ],
+                    }
+                ],
+                "usage": {
+                    "input_tokens": input_tokens,
+                    "output_tokens": output_tokens,
+                    "total_tokens": total_tokens,
+                },
+            },
+        }
+        yield f"event: response.completed\ndata: {json.dumps(completed_response)}\n\n"
+
+    return StreamingResponse(
+        event_generator(),
+        media_type = "text/event-stream",
+        headers = {
+            "Cache-Control": "no-cache",
+            "Connection": "keep-alive",
+            "X-Accel-Buffering": "no",
+        },
+    )
+
+
+@router.post("/responses")
+async def openai_responses(
+    payload: ResponsesRequest,
+    request: Request,
+    current_subject: str = Depends(get_current_subject),
+):
+    """
+    OpenAI Responses API endpoint.
+
+    Accepts the Responses-format request, converts it to a
+    ChatCompletionRequest internally, and returns a response
+    matching the OpenAI Responses API schema (output array,
+    input_tokens/output_tokens, named SSE events for streaming).
+    """
+    messages = _normalise_responses_input(payload)
+    if not messages:
+        raise HTTPException(status_code = 400, detail = "No input provided.")
+
+    if payload.stream:
+        return await _responses_stream(payload, messages, request)
+    return await _responses_non_streaming(payload, messages, request)
+
+
+# =====================================================================
+# Anthropic-Compatible Messages API  (/messages → /v1/messages)
+# =====================================================================
+
+
+@router.post("/messages")
+async def anthropic_messages(
+    payload: AnthropicMessagesRequest,
+    request: Request,
+    current_subject: str = Depends(get_current_subject),
+):
+    """
+    Anthropic-compatible Messages API endpoint.
+
+    Translates Anthropic message format to internal OpenAI format, runs
+    through the existing agentic tool loop when tools are provided, and
+    returns responses in Anthropic Messages API format (streaming SSE or
+    non-streaming JSON).
+    """
+    llama_backend = get_llama_cpp_backend()
+    if not llama_backend.is_loaded:
+        raise HTTPException(
+            status_code = 503,
+            detail = "No GGUF model loaded. Load a GGUF model first.",
+        )
+
+    model_name = getattr(llama_backend, "model_identifier", None) or payload.model
+    message_id = f"msg_{uuid.uuid4().hex[:24]}"
+
+    # ── Translate Anthropic → OpenAI ──────────────────────────
+    openai_messages = anthropic_messages_to_openai(
+        [m.model_dump() for m in payload.messages],
+        payload.system,
+    )
+
+    temperature = payload.temperature if payload.temperature is not None else 0.6
+    top_p = payload.top_p if payload.top_p is not None else 0.95
+    top_k = payload.top_k if payload.top_k is not None else 20
+    min_p = payload.min_p if payload.min_p is not None else 0.01
+    repetition_penalty = (
+        payload.repetition_penalty if payload.repetition_penalty is not None else 1.0
+    )
+    presence_penalty = (
+        payload.presence_penalty if payload.presence_penalty is not None else 0.0
+    )
+    stop = payload.stop_sequences or None
+
+    # tool_choice is declared on AnthropicMessagesRequest for Anthropic SDK
+    # compatibility (the SDK often sets it by default), but it is not
+    # currently honored by Unsloth's backend. Warn once per request so the
+    # silent drop is visible to operators instead of looking like a model
+    # quality issue to clients.
+    if payload.tool_choice is not None:
+        logger.warning(
+            "anthropic_messages.tool_choice_ignored",
+            tool_choice = payload.tool_choice,
+            note = (
+                "tool_choice is accepted for Anthropic SDK compatibility but not "
+                "honored by Unsloth. Use enable_tools / enabled_tools (server-side "
+                "built-in tools) or restrict the `tools` array (client-side) to "
+                "control which tools the model sees."
+            ),
+        )
+
+    cancel_event = threading.Event()
+
+    # ── Tool routing ──────────────────────────────────────────
+    # Three paths:
+    # 1. enable_tools=true → server-side execution of built-in tools (Unsloth shorthand)
+    # 2. tools=[...] only  → client-side pass-through (standard Anthropic behavior)
+    # 3. neither           → plain chat
+    server_tools = payload.enable_tools and llama_backend.supports_tools
+    client_tools = (
+        not server_tools
+        and payload.tools
+        and len(payload.tools) > 0
+        and llama_backend.supports_tools
+    )
+
+    # ── Client-side pass-through path ─────────────────────────
+    if client_tools:
+        openai_tools = anthropic_tools_to_openai(payload.tools)
+
+        if payload.stream:
+            return await _anthropic_passthrough_stream(
+                request,
+                cancel_event,
+                llama_backend,
+                openai_messages,
+                openai_tools,
+                temperature,
+                top_p,
+                top_k,
+                payload.max_tokens,
+                message_id,
+                model_name,
+                stop = stop,
+                min_p = min_p,
+                repetition_penalty = repetition_penalty,
+                presence_penalty = presence_penalty,
+            )
+        return await _anthropic_passthrough_non_streaming(
+            llama_backend,
+            openai_messages,
+            openai_tools,
+            temperature,
+            top_p,
+            top_k,
+            payload.max_tokens,
+            message_id,
+            model_name,
+            stop = stop,
+            min_p = min_p,
+            repetition_penalty = repetition_penalty,
+            presence_penalty = presence_penalty,
+        )
+
+    if server_tools:
+        from core.inference.tools import ALL_TOOLS
+
+        if payload.enabled_tools is not None:
+            openai_tools = [
+                t for t in ALL_TOOLS if t["function"]["name"] in payload.enabled_tools
+            ]
+        else:
+            openai_tools = ALL_TOOLS
+
+        # Build tool-use system prompt nudge (same logic as /chat/completions)
+        _tool_names = {t["function"]["name"] for t in openai_tools}
+        _has_web = "web_search" in _tool_names
+        _has_code = "python" in _tool_names or "terminal" in _tool_names
+
+        _date_line = f"The current date is {_date.today().isoformat()}."
+        _model_size_b = _extract_model_size_b(model_name)
+        _is_small_model = _model_size_b is not None and _model_size_b < 9
+
+        if _is_small_model:
+            _web_tips = "Do not repeat the same search query."
+        else:
+            _web_tips = (
+                "When you search and find a relevant URL in the results, "
+                "fetch its full content by calling web_search with the url parameter. "
+                "Do not repeat the same search query. If a search returns "
+                "no useful results, try rephrasing or fetching a result URL directly."
+            )
+        _code_tips = (
+            "Use code execution for math, calculations, data processing, "
+            "or to parse and analyze information from tool results."
+        )
+
+        if _has_web and _has_code:
+            _nudge = (
+                _date_line + " "
+                "You have access to tools. When appropriate, prefer using "
+                "tools rather than answering from memory. "
+                + _web_tips
+                + " "
+                + _code_tips
+            )
+        elif _has_code:
+            _nudge = (
+                _date_line + " "
+                "You have access to tools. When appropriate, prefer using "
+                "code execution rather than answering from memory. " + _code_tips
+            )
+        elif _has_web:
+            _nudge = (
+                _date_line + " "
+                "You have access to tools. When appropriate, prefer using "
+                "web search for up-to-date or uncertain factual "
+                "information rather than answering from memory. " + _web_tips
+            )
+        else:
+            _nudge = ""
+
+        if _nudge:
+            _nudge += _TOOL_ACTION_NUDGE
+            # Inject into system prompt
+            if openai_messages and openai_messages[0].get("role") == "system":
+                openai_messages[0]["content"] = (
+                    openai_messages[0]["content"].rstrip() + "\n\n" + _nudge
+                )
+            else:
+                openai_messages.insert(0, {"role": "system", "content": _nudge})
+
+        # Strip stale tool-call XML from conversation
+        for _msg in openai_messages:
+            if _msg.get("role") == "assistant" and isinstance(_msg.get("content"), str):
+                _msg["content"] = _TOOL_XML_RE.sub("", _msg["content"]).strip()
+
+        def _run_tool_gen():
+            return llama_backend.generate_chat_completion_with_tools(
+                messages = openai_messages,
+                tools = openai_tools,
+                temperature = temperature,
+                top_p = top_p,
+                top_k = top_k,
+                min_p = min_p,
+                repetition_penalty = repetition_penalty,
+                presence_penalty = presence_penalty,
+                max_tokens = payload.max_tokens,
+                stop = stop,
+                cancel_event = cancel_event,
+                max_tool_iterations = 25,
+                auto_heal_tool_calls = True,
+                tool_call_timeout = 300,
+                session_id = payload.session_id,
+            )
+
+        if payload.stream:
+            return await _anthropic_tool_stream(
+                request,
+                cancel_event,
+                _run_tool_gen,
+                message_id,
+                model_name,
+            )
+        return await _anthropic_tool_non_streaming(
+            _run_tool_gen,
+            message_id,
+            model_name,
+        )
+
+    # ── No-tool path ──────────────────────────────────────────
+    def _run_plain_gen():
+        return llama_backend.generate_chat_completion(
+            messages = openai_messages,
+            temperature = temperature,
+            top_p = top_p,
+            top_k = top_k,
+            min_p = min_p,
+            repetition_penalty = repetition_penalty,
+            presence_penalty = presence_penalty,
+            max_tokens = payload.max_tokens,
+            stop = stop,
+            cancel_event = cancel_event,
+        )
+
+    if payload.stream:
+        return await _anthropic_plain_stream(
+            request,
+            cancel_event,
+            _run_plain_gen,
+            message_id,
+            model_name,
+        )
+    return await _anthropic_plain_non_streaming(
+        _run_plain_gen,
+        message_id,
+        model_name,
+    )
+
+
+async def _anthropic_tool_stream(
+    request,
+    cancel_event,
+    run_gen,
+    message_id,
+    model_name,
+):
+    """Streaming response for the tool-calling path."""
+    _sentinel = object()
+
+    async def _stream():
+        emitter = AnthropicStreamEmitter()
+        for line in emitter.start(message_id, model_name):
+            yield line
+
+        gen = run_gen()
+        try:
+            while True:
+                if await request.is_disconnected():
+                    cancel_event.set()
+                    return
+                event = await asyncio.to_thread(next, gen, _sentinel)
+                if event is _sentinel:
+                    break
+                # Strip leaked tool-call XML from content events
+                if event.get("type") == "content":
+                    event = dict(event)
+                    event["text"] = _TOOL_XML_RE.sub("", event["text"])
+                for line in emitter.feed(event):
+                    yield line
+        except Exception as e:
+            logger.error("anthropic_messages stream error: %s", e)
+
+        for line in emitter.finish("end_turn"):
+            yield line
+
+    return StreamingResponse(
+        _stream(),
+        media_type = "text/event-stream",
+        headers = {
+            "Cache-Control": "no-cache",
+            "Connection": "keep-alive",
+            "X-Accel-Buffering": "no",
+        },
+    )
+
+
+async def _anthropic_plain_stream(
+    request,
+    cancel_event,
+    run_gen,
+    message_id,
+    model_name,
+):
+    """Streaming response for the no-tool path."""
+    _sentinel = object()
+
+    async def _stream():
+        emitter = AnthropicStreamEmitter()
+        for line in emitter.start(message_id, model_name):
+            yield line
+
+        gen = run_gen()
+        try:
+            while True:
+                if await request.is_disconnected():
+                    cancel_event.set()
+                    return
+                cumulative = await asyncio.to_thread(next, gen, _sentinel)
+                if cumulative is _sentinel:
+                    break
+                if isinstance(cumulative, dict):
+                    if cumulative.get("type") == "metadata":
+                        for line in emitter.feed(cumulative):
+                            yield line
+                    continue
+                # Plain generator yields cumulative text strings
+                for line in emitter.feed({"type": "content", "text": cumulative}):
+                    yield line
+        except Exception as e:
+            logger.error("anthropic_messages stream error: %s", e)
+
+        for line in emitter.finish("end_turn"):
+            yield line
+
+    return StreamingResponse(
+        _stream(),
+        media_type = "text/event-stream",
+        headers = {
+            "Cache-Control": "no-cache",
+            "Connection": "keep-alive",
+            "X-Accel-Buffering": "no",
+        },
+    )
+
+
+async def _anthropic_tool_non_streaming(run_gen, message_id, model_name):
+    """Non-streaming response for the tool-calling path.
+
+    Builds ``content_blocks`` in generation order (text → tool_use → text →
+    tool_use → ...), mirroring the streaming emitter's behavior. Deltas
+    within a single synthesis turn are merged into the trailing text block;
+    tool_use blocks interrupt the text sequence and open a new text block on
+    the next content event.
+
+    ``prev_text`` is reset on ``tool_end`` because
+    ``generate_chat_completion_with_tools`` yields cumulative content *per
+    turn* — the first content event of turn N+1 must diff against an empty
+    baseline, not against turn N's final length.
+    """
+    content_blocks: list = []
+    usage = {}
+    prev_text = ""
+
+    for event in run_gen():
+        etype = event.get("type", "")
+        if etype == "content":
+            # Strip leaked tool-call XML
+            clean = _TOOL_XML_RE.sub("", event["text"])
+            new = clean[len(prev_text) :]
+            prev_text = clean
+            if new:
+                if content_blocks and isinstance(
+                    content_blocks[-1], AnthropicResponseTextBlock
+                ):
+                    content_blocks[-1].text += new
+                else:
+                    content_blocks.append(AnthropicResponseTextBlock(text = new))
+        elif etype == "tool_start":
+            content_blocks.append(
+                AnthropicResponseToolUseBlock(
+                    id = event["tool_call_id"],
+                    name = event["tool_name"],
+                    input = event.get("arguments", {}),
+                )
+            )
+        elif etype == "tool_end":
+            prev_text = ""
+        elif etype == "metadata":
+            usage = event.get("usage", {})
+
+    resp = AnthropicMessagesResponse(
+        id = message_id,
+        model = model_name,
+        content = content_blocks,
+        stop_reason = "end_turn",
+        usage = AnthropicUsage(
+            input_tokens = usage.get("prompt_tokens", 0),
+            output_tokens = usage.get("completion_tokens", 0),
+        ),
+    )
+    return JSONResponse(content = resp.model_dump())
+
+
+async def _anthropic_plain_non_streaming(run_gen, message_id, model_name):
+    """Non-streaming response for the no-tool path."""
+    text_parts = []
+    usage = {}
+    prev_text = ""
+
+    for cumulative in run_gen():
+        if isinstance(cumulative, dict):
+            if cumulative.get("type") == "metadata":
+                usage = cumulative.get("usage", {})
+            continue
+        new = cumulative[len(prev_text) :]
+        prev_text = cumulative
+        if new:
+            text_parts.append(new)
+
+    full_text = "".join(text_parts)
+    content_blocks = []
+    if full_text:
+        content_blocks.append(AnthropicResponseTextBlock(text = full_text))
+
+    resp = AnthropicMessagesResponse(
+        id = message_id,
+        model = model_name,
+        content = content_blocks,
+        stop_reason = "end_turn",
+        usage = AnthropicUsage(
+            input_tokens = usage.get("prompt_tokens", 0),
+            output_tokens = usage.get("completion_tokens", 0),
+        ),
+    )
+    return JSONResponse(content = resp.model_dump())
+
+
+# =====================================================================
+# Client-side tool pass-through (Anthropic-native tools field)
+# =====================================================================
+
+
+def _build_passthrough_payload(
+    openai_messages,
+    openai_tools,
+    temperature,
+    top_p,
+    top_k,
+    max_tokens,
+    stream,
+    stop = None,
+    min_p = None,
+    repetition_penalty = None,
+    presence_penalty = None,
+):
+    body = {
+        "messages": openai_messages,
+        "tools": openai_tools,
+        "tool_choice": "auto",
+        "temperature": temperature,
+        "top_p": top_p,
+        "top_k": top_k,
+        "stream": stream,
+    }
+    if stream:
+        body["stream_options"] = {"include_usage": True}
+    if max_tokens is not None:
+        body["max_tokens"] = max_tokens
+    if stop:
+        body["stop"] = stop
+    if min_p is not None:
+        body["min_p"] = min_p
+    if repetition_penalty is not None:
+        # llama-server's field is "repeat_penalty", not "repetition_penalty"
+        body["repeat_penalty"] = repetition_penalty
+    if presence_penalty is not None:
+        body["presence_penalty"] = presence_penalty
+    return body
+
+
+async def _anthropic_passthrough_stream(
+    request,
+    cancel_event,
+    llama_backend,
+    openai_messages,
+    openai_tools,
+    temperature,
+    top_p,
+    top_k,
+    max_tokens,
+    message_id,
+    model_name,
+    stop = None,
+    min_p = None,
+    repetition_penalty = None,
+    presence_penalty = None,
+):
+    """Streaming client-side pass-through: forward tools to llama-server and
+    translate its streaming response to Anthropic SSE without executing anything."""
+    target_url = f"{llama_backend.base_url}/v1/chat/completions"
+    body = _build_passthrough_payload(
+        openai_messages,
+        openai_tools,
+        temperature,
+        top_p,
+        top_k,
+        max_tokens,
+        True,
+        stop = stop,
+        min_p = min_p,
+        repetition_penalty = repetition_penalty,
+        presence_penalty = presence_penalty,
+    )
+
+    async def _stream():
+        emitter = AnthropicPassthroughEmitter()
+        for line in emitter.start(message_id, model_name):
+            yield line
+
+        # Manage the httpx client and response MANUALLY — no `async with`.
+        #
+        # On Python 3.13 + httpcore 1.0.x, an orphaned async generator (e.g.
+        # when the client disconnects mid-stream and Starlette drops the
+        # StreamingResponse iterator without explicitly calling aclose())
+        # is finalized by Python's asyncgen GC hook in a DIFFERENT asyncio
+        # task than the one that originally entered the httpx context
+        # managers. When `async with` exits run in the wrong task, httpcore's
+        # internal `HTTP11ConnectionByteStream.aclose()` hits
+        # `anyio.CancelScope.__exit__` with a mismatched task and raises
+        # RuntimeError("Attempted to exit cancel scope in a different task"),
+        # which escapes as "Exception ignored in:" because it happens during
+        # GC finalization outside any user-owned try/except.
+        #
+        # The fix: do not use `async with` for the client/response. Close
+        # them in a finally block wrapped in `try: ... except Exception: pass`.
+        # This narrowly suppresses RuntimeError / other Exception subclasses
+        # from the anyio cleanup noise while letting GeneratorExit (a
+        # BaseException, not Exception) propagate through cleanly so the
+        # generator terminates as Python expects.
+        client = httpx.AsyncClient(timeout = 600)
+        resp = None
+        try:
+            req = client.build_request("POST", target_url, json = body)
+            resp = await client.send(req, stream = True)
+
+            async for raw_line in resp.aiter_lines():
+                if await request.is_disconnected():
+                    cancel_event.set()
+                    break
+                if not raw_line or not raw_line.startswith("data: "):
+                    continue
+                data_str = raw_line[6:]
+                if data_str.strip() == "[DONE]":
+                    break
+                try:
+                    chunk = json.loads(data_str)
+                except json.JSONDecodeError:
+                    continue
+                for line in emitter.feed_chunk(chunk):
+                    yield line
+        except Exception as e:
+            logger.error("anthropic_messages passthrough stream error: %s", e)
+        finally:
+            if resp is not None:
+                try:
+                    await resp.aclose()
+                except Exception:
+                    pass
+            try:
+                await client.aclose()
+            except Exception:
+                pass
+
+        for line in emitter.finish():
+            yield line
+
+    return StreamingResponse(
+        _stream(),
+        media_type = "text/event-stream",
+        headers = {
+            "Cache-Control": "no-cache",
+            "Connection": "keep-alive",
+            "X-Accel-Buffering": "no",
+        },
+    )
+
+
+async def _anthropic_passthrough_non_streaming(
+    llama_backend,
+    openai_messages,
+    openai_tools,
+    temperature,
+    top_p,
+    top_k,
+    max_tokens,
+    message_id,
+    model_name,
+    stop = None,
+    min_p = None,
+    repetition_penalty = None,
+    presence_penalty = None,
+):
+    """Non-streaming client-side pass-through."""
+    target_url = f"{llama_backend.base_url}/v1/chat/completions"
+    body = _build_passthrough_payload(
+        openai_messages,
+        openai_tools,
+        temperature,
+        top_p,
+        top_k,
+        max_tokens,
+        False,
+        stop = stop,
+        min_p = min_p,
+        repetition_penalty = repetition_penalty,
+        presence_penalty = presence_penalty,
+    )
+
+    async with httpx.AsyncClient() as client:
+        resp = await client.post(target_url, json = body, timeout = 600)
+
+    if resp.status_code != 200:
+        raise HTTPException(
+            status_code = resp.status_code,
+            detail = f"llama-server error: {resp.text[:500]}",
+        )
+
+    data = resp.json()
+    choice = (data.get("choices") or [{}])[0]
+    message = choice.get("message") or {}
+    finish_reason = choice.get("finish_reason")
+
+    content_blocks = []
+    text = message.get("content") or ""
+    if text:
+        text = _TOOL_XML_RE.sub("", text).strip()
+        if text:
+            content_blocks.append(AnthropicResponseTextBlock(text = text))
+
+    tool_calls = message.get("tool_calls") or []
+    for tc in tool_calls:
+        fn = tc.get("function") or {}
+        try:
+            args = json.loads(fn.get("arguments", "{}"))
+        except json.JSONDecodeError:
+            args = {}
+        content_blocks.append(
+            AnthropicResponseToolUseBlock(
+                id = tc.get("id", ""),
+                name = fn.get("name", ""),
+                input = args,
+            )
+        )
+
+    if tool_calls:
+        stop_reason = "tool_use"
+    elif finish_reason == "length":
+        stop_reason = "max_tokens"
+    else:
+        stop_reason = "end_turn"
+
+    usage = data.get("usage") or {}
+    resp_obj = AnthropicMessagesResponse(
+        id = message_id,
+        model = model_name,
+        content = content_blocks,
+        stop_reason = stop_reason,
+        usage = AnthropicUsage(
+            input_tokens = usage.get("prompt_tokens", 0),
+            output_tokens = usage.get("completion_tokens", 0),
+        ),
+    )
+    return JSONResponse(content = resp_obj.model_dump())
diff --git a/studio/backend/run.py b/studio/backend/run.py
index 86c1194661..9675b9ea4c 100644
--- a/studio/backend/run.py
+++ b/studio/backend/run.py
@@ -248,6 +248,7 @@ def run_server(
     port: int = 8888,
     frontend_path: Path = Path(__file__).resolve().parent.parent / "frontend" / "dist",
     silent: bool = False,
+    llama_parallel_slots: int = 1,
 ):
     """
     Start the FastAPI server.
@@ -257,6 +258,7 @@ def run_server(
         port: Port to bind to (auto-increments if in use)
         frontend_path: Path to frontend build directory (optional)
         silent: Suppress startup messages
+        llama_parallel_slots: Number of parallel slots for llama-server
 
     Note:
         Signal handlers are NOT registered here so that embedders
@@ -331,6 +333,7 @@ def run_server(
     # binds (port==0) leave it unset and let request handlers fall back
     # to the ASGI request scope or request.base_url.
     app.state.server_port = port if port and port > 0 else None
+    app.state.llama_parallel_slots = llama_parallel_slots
 
     # Run server in a daemon thread
     def _run():
diff --git a/studio/backend/tests/conftest.py b/studio/backend/tests/conftest.py
index 053e9b85d9..6aa6d314c1 100644
--- a/studio/backend/tests/conftest.py
+++ b/studio/backend/tests/conftest.py
@@ -3,14 +3,136 @@
 
 """
 Shared pytest configuration for the backend test suite.
-Ensures that the backend root is on sys.path so that
-`import utils.utils` (and similar flat imports) resolve correctly.
+
+Responsibilities:
+1. Put the backend root on sys.path so `from models.inference import ...`
+   (and similar flat imports) resolve in test modules — mirrors how the
+   app itself is launched.
+2. Provide a hybrid ``studio_server`` session fixture for end-to-end tests
+   (see ``test_studio_api.py``). The fixture supports two invocation modes:
+
+   a. **External server.** If ``UNSLOTH_E2E_BASE_URL`` is set, tests point
+      at an already-running Studio instance. ``UNSLOTH_E2E_API_KEY`` must
+      also be set. This is the fast-iteration mode: start the server once
+      with ``unsloth studio run ...``, then run pytest against it many
+      times with no per-run GGUF load cost.
+
+   b. **Fixture-managed server.** Otherwise, the fixture launches a fresh
+      server via ``_start_server`` and tears it down at session end. This
+      is the one-shot mode for CI or a clean-slate verification run.
+
+   The model / variant for mode (b) come from ``--unsloth-model`` /
+   ``--unsloth-gguf-variant`` pytest options, then ``UNSLOTH_E2E_MODEL`` /
+   ``UNSLOTH_E2E_VARIANT`` env vars, then the defaults in
+   ``test_studio_api.py``.
 """
 
+import os
 import sys
 from pathlib import Path
 
+import pytest
+
 # Add backend root to sys.path (mirrors how the app itself is launched)
 _backend_root = Path(__file__).resolve().parent.parent
 if str(_backend_root) not in sys.path:
     sys.path.insert(0, str(_backend_root))
+
+
+# ── Pytest CLI options ───────────────────────────────────────────────
+
+
+def pytest_addoption(parser):
+    group = parser.getgroup(
+        "unsloth-e2e",
+        "Unsloth Studio end-to-end test options",
+    )
+    group.addoption(
+        "--unsloth-model",
+        action = "store",
+        default = None,
+        help = (
+            "GGUF model id used when starting a server for e2e tests. "
+            "Ignored if UNSLOTH_E2E_BASE_URL is set. Overrides "
+            "UNSLOTH_E2E_MODEL env var. Defaults to test_studio_api.py's "
+            "DEFAULT_MODEL."
+        ),
+    )
+    group.addoption(
+        "--unsloth-gguf-variant",
+        action = "store",
+        default = None,
+        help = (
+            "GGUF variant used when starting a server for e2e tests. "
+            "Ignored if UNSLOTH_E2E_BASE_URL is set. Overrides "
+            "UNSLOTH_E2E_VARIANT env var. Defaults to test_studio_api.py's "
+            "DEFAULT_VARIANT."
+        ),
+    )
+
+
+# ── E2E server fixtures ──────────────────────────────────────────────
+
+
+@pytest.fixture(scope = "session")
+def studio_server(request):
+    """Yield ``(base_url, api_key)`` for e2e tests.
+
+    Resolution order:
+
+    1. If ``UNSLOTH_E2E_BASE_URL`` is set → point at that server,
+       require ``UNSLOTH_E2E_API_KEY`` alongside (skip if missing).
+    2. Otherwise → start a fresh ``unsloth studio run`` subprocess via
+       the existing ``_start_server`` helper in ``test_studio_api.py``
+       and tear it down on session teardown.
+
+    Session-scoped so the expensive GGUF load happens at most once per
+    pytest invocation. Lazily instantiated — tests that don't request
+    the fixture (e.g. the unit tests in ``test_anthropic_messages.py``
+    or ``test_help_output``) do not trigger server startup.
+    """
+    external_url = os.environ.get("UNSLOTH_E2E_BASE_URL")
+    if external_url:
+        api_key = os.environ.get("UNSLOTH_E2E_API_KEY")
+        if not api_key:
+            pytest.skip(
+                "UNSLOTH_E2E_BASE_URL is set but UNSLOTH_E2E_API_KEY is "
+                "missing — tests that require auth cannot run against an "
+                "external server without it.",
+            )
+        yield external_url, api_key
+        return
+
+    # Lazy import: pytest has already loaded test_studio_api into
+    # sys.modules by the time any test requests this fixture, so this
+    # is a cache hit, not a re-execution.
+    import test_studio_api as _e2e
+
+    model = (
+        request.config.getoption("--unsloth-model")
+        or os.environ.get("UNSLOTH_E2E_MODEL")
+        or _e2e.DEFAULT_MODEL
+    )
+    variant = (
+        request.config.getoption("--unsloth-gguf-variant")
+        or os.environ.get("UNSLOTH_E2E_VARIANT")
+        or _e2e.DEFAULT_VARIANT
+    )
+
+    proc, api_key = _e2e._start_server(model, variant)
+    try:
+        yield f"http://{_e2e.HOST}:{_e2e.PORT}", api_key
+    finally:
+        _e2e._kill_server(proc)
+
+
+@pytest.fixture
+def base_url(studio_server):
+    """Base URL for the e2e Studio server (from ``studio_server``)."""
+    return studio_server[0]
+
+
+@pytest.fixture
+def api_key(studio_server):
+    """API key for the e2e Studio server (from ``studio_server``)."""
+    return studio_server[1]
diff --git a/studio/backend/tests/test_anthropic_messages.py b/studio/backend/tests/test_anthropic_messages.py
new file mode 100644
index 0000000000..ec432df9e6
--- /dev/null
+++ b/studio/backend/tests/test_anthropic_messages.py
@@ -0,0 +1,774 @@
+# SPDX-License-Identifier: AGPL-3.0-only
+# Copyright 2026-present the Unsloth AI Inc. team. All rights reserved.
+
+"""
+Tests for the Anthropic Messages API schemas and translation layer.
+No running server or GPU required.
+"""
+
+import sys
+import os
+import json
+
+_backend = os.path.join(os.path.dirname(__file__), "..")
+sys.path.insert(0, _backend)
+
+from models.inference import (
+    AnthropicMessagesRequest,
+    AnthropicMessagesResponse,
+    AnthropicMessage,
+    AnthropicTextBlock,
+    AnthropicToolUseBlock,
+    AnthropicToolResultBlock,
+    AnthropicTool,
+    AnthropicUsage,
+    AnthropicResponseTextBlock,
+    AnthropicResponseToolUseBlock,
+)
+from core.inference.anthropic_compat import (
+    anthropic_messages_to_openai,
+    anthropic_tools_to_openai,
+    build_anthropic_sse_event,
+    AnthropicStreamEmitter,
+    AnthropicPassthroughEmitter,
+)
+
+
+# =====================================================================
+# Pydantic model tests
+# =====================================================================
+
+
+class TestAnthropicModels:
+    def test_minimal_request(self):
+        req = AnthropicMessagesRequest(
+            messages = [{"role": "user", "content": "Hi"}],
+        )
+        assert req.max_tokens is None
+        assert req.model == "default"
+        assert req.stream is False
+
+    def test_max_tokens_optional(self):
+        req = AnthropicMessagesRequest(
+            max_tokens = 100,
+            messages = [{"role": "user", "content": "Hi"}],
+        )
+        assert req.max_tokens == 100
+
+    def test_system_as_string(self):
+        req = AnthropicMessagesRequest(
+            max_tokens = 50,
+            messages = [{"role": "user", "content": "Hi"}],
+            system = "You are helpful.",
+        )
+        assert req.system == "You are helpful."
+
+    def test_tools_field_parses(self):
+        req = AnthropicMessagesRequest(
+            max_tokens = 100,
+            messages = [{"role": "user", "content": "Hi"}],
+            tools = [{"name": "web_search", "input_schema": {"type": "object"}}],
+        )
+        assert len(req.tools) == 1
+        assert req.tools[0].name == "web_search"
+
+    def test_extra_fields_accepted(self):
+        req = AnthropicMessagesRequest(
+            max_tokens = 100,
+            messages = [{"role": "user", "content": "Hi"}],
+            some_future_field = "hello",
+        )
+        assert req.max_tokens == 100
+
+    def test_stream_defaults_false(self):
+        req = AnthropicMessagesRequest(
+            max_tokens = 100,
+            messages = [{"role": "user", "content": "Hi"}],
+        )
+        assert req.stream is False
+
+    def test_enable_tools_shorthand(self):
+        req = AnthropicMessagesRequest(
+            messages = [{"role": "user", "content": "Hi"}],
+            enable_tools = True,
+            enabled_tools = ["web_search", "python"],
+            session_id = "my-session",
+        )
+        assert req.enable_tools is True
+        assert req.enabled_tools == ["web_search", "python"]
+        assert req.session_id == "my-session"
+
+    def test_extension_fields_default_none(self):
+        req = AnthropicMessagesRequest(
+            messages = [{"role": "user", "content": "Hi"}],
+        )
+        assert req.enable_tools is None
+        assert req.enabled_tools is None
+        assert req.session_id is None
+
+    def test_response_model_defaults(self):
+        resp = AnthropicMessagesResponse()
+        assert resp.type == "message"
+        assert resp.role == "assistant"
+        assert resp.id.startswith("msg_")
+        assert resp.content == []
+        assert resp.usage.input_tokens == 0
+
+
+# =====================================================================
+# Message translation tests
+# =====================================================================
+
+
+class TestAnthropicMessagesToOpenAI:
+    def test_simple_user_message(self):
+        msgs = [{"role": "user", "content": "Hello"}]
+        result = anthropic_messages_to_openai(msgs)
+        assert result == [{"role": "user", "content": "Hello"}]
+
+    def test_system_string_prepended(self):
+        msgs = [{"role": "user", "content": "Hello"}]
+        result = anthropic_messages_to_openai(msgs, system = "Be brief.")
+        assert result[0] == {"role": "system", "content": "Be brief."}
+        assert result[1] == {"role": "user", "content": "Hello"}
+
+    def test_system_as_block_list(self):
+        system = [
+            {"type": "text", "text": "Be brief."},
+            {"type": "text", "text": "Be accurate."},
+        ]
+        msgs = [{"role": "user", "content": "Hello"}]
+        result = anthropic_messages_to_openai(msgs, system = system)
+        assert result[0]["role"] == "system"
+        assert "Be brief." in result[0]["content"]
+        assert "Be accurate." in result[0]["content"]
+
+    def test_multi_turn_conversation(self):
+        msgs = [
+            {"role": "user", "content": "Hi"},
+            {"role": "assistant", "content": "Hello!"},
+            {"role": "user", "content": "How are you?"},
+        ]
+        result = anthropic_messages_to_openai(msgs)
+        assert len(result) == 3
+        assert result[0]["role"] == "user"
+        assert result[1]["role"] == "assistant"
+        assert result[2]["role"] == "user"
+
+    def test_assistant_tool_use_maps_to_tool_calls(self):
+        msgs = [
+            {
+                "role": "assistant",
+                "content": [
+                    {"type": "text", "text": "Let me search."},
+                    {
+                        "type": "tool_use",
+                        "id": "tu_1",
+                        "name": "web_search",
+                        "input": {"query": "test"},
+                    },
+                ],
+            }
+        ]
+        result = anthropic_messages_to_openai(msgs)
+        assert len(result) == 1
+        m = result[0]
+        assert m["role"] == "assistant"
+        assert m["content"] == "Let me search."
+        assert len(m["tool_calls"]) == 1
+        tc = m["tool_calls"][0]
+        assert tc["id"] == "tu_1"
+        assert tc["function"]["name"] == "web_search"
+        assert json.loads(tc["function"]["arguments"]) == {"query": "test"}
+
+    def test_tool_result_maps_to_tool_role(self):
+        msgs = [
+            {
+                "role": "user",
+                "content": [
+                    {
+                        "type": "tool_result",
+                        "tool_use_id": "tu_1",
+                        "content": "Result text",
+                    },
+                ],
+            }
+        ]
+        result = anthropic_messages_to_openai(msgs)
+        assert len(result) == 1
+        assert result[0]["role"] == "tool"
+        assert result[0]["tool_call_id"] == "tu_1"
+        assert result[0]["content"] == "Result text"
+
+    def test_mixed_text_and_tool_use_blocks(self):
+        msgs = [
+            {
+                "role": "assistant",
+                "content": [
+                    {"type": "text", "text": "Thinking..."},
+                    {
+                        "type": "tool_use",
+                        "id": "tu_1",
+                        "name": "python",
+                        "input": {"code": "1+1"},
+                    },
+                    {
+                        "type": "tool_use",
+                        "id": "tu_2",
+                        "name": "terminal",
+                        "input": {"command": "ls"},
+                    },
+                ],
+            }
+        ]
+        result = anthropic_messages_to_openai(msgs)
+        assert len(result) == 1
+        m = result[0]
+        assert m["content"] == "Thinking..."
+        assert len(m["tool_calls"]) == 2
+
+    def test_tool_result_with_list_content(self):
+        msgs = [
+            {
+                "role": "user",
+                "content": [
+                    {
+                        "type": "tool_result",
+                        "tool_use_id": "tu_1",
+                        "content": [
+                            {"type": "text", "text": "Line 1"},
+                            {"type": "text", "text": "Line 2"},
+                        ],
+                    },
+                ],
+            }
+        ]
+        result = anthropic_messages_to_openai(msgs)
+        assert result[0]["content"] == "Line 1 Line 2"
+
+
+# =====================================================================
+# Tool translation tests
+# =====================================================================
+
+
+class TestAnthropicToolsToOpenAI:
+    def test_single_tool(self):
+        tools = [
+            {
+                "name": "web_search",
+                "description": "Search",
+                "input_schema": {
+                    "type": "object",
+                    "properties": {"query": {"type": "string"}},
+                },
+            }
+        ]
+        result = anthropic_tools_to_openai(tools)
+        assert len(result) == 1
+        assert result[0]["type"] == "function"
+        assert result[0]["function"]["name"] == "web_search"
+        assert result[0]["function"]["parameters"]["type"] == "object"
+
+    def test_multiple_tools(self):
+        tools = [
+            {"name": "a", "description": "Tool A", "input_schema": {}},
+            {"name": "b", "description": "Tool B", "input_schema": {}},
+        ]
+        result = anthropic_tools_to_openai(tools)
+        assert len(result) == 2
+        assert result[0]["function"]["name"] == "a"
+        assert result[1]["function"]["name"] == "b"
+
+    def test_empty_list(self):
+        assert anthropic_tools_to_openai([]) == []
+
+    def test_pydantic_model_input(self):
+        tool = AnthropicTool(
+            name = "test", description = "desc", input_schema = {"type": "object"}
+        )
+        result = anthropic_tools_to_openai([tool])
+        assert result[0]["function"]["name"] == "test"
+
+
+# =====================================================================
+# SSE event helper tests
+# =====================================================================
+
+
+class TestBuildAnthropicSSEEvent:
+    def test_basic_event(self):
+        result = build_anthropic_sse_event("message_start", {"type": "message_start"})
+        assert result.startswith("event: message_start\n")
+        assert "data: " in result
+        assert result.endswith("\n\n")
+
+    def test_data_is_valid_json(self):
+        result = build_anthropic_sse_event("test", {"key": "value"})
+        data_line = result.split("\n")[1]
+        payload = json.loads(data_line.removeprefix("data: "))
+        assert payload == {"key": "value"}
+
+
+# =====================================================================
+# Stream emitter tests
+# =====================================================================
+
+
+class TestAnthropicStreamEmitter:
+    def test_start_emits_message_start_and_content_block_start(self):
+        e = AnthropicStreamEmitter()
+        events = e.start("msg_123", "test-model")
+        assert len(events) == 2
+        assert "message_start" in events[0]
+        assert "content_block_start" in events[1]
+        assert '"type": "text"' in events[1]
+
+    def test_content_delta_emits_text_delta(self):
+        e = AnthropicStreamEmitter()
+        e.start("msg_1", "m")
+        events = e.feed({"type": "content", "text": "Hello"})
+        assert len(events) == 1
+        parsed = json.loads(events[0].split("data: ")[1])
+        assert parsed["delta"]["type"] == "text_delta"
+        assert parsed["delta"]["text"] == "Hello"
+
+    def test_cumulative_content_diffs_correctly(self):
+        e = AnthropicStreamEmitter()
+        e.start("msg_1", "m")
+        e.feed({"type": "content", "text": "Hel"})
+        events = e.feed({"type": "content", "text": "Hello"})
+        parsed = json.loads(events[0].split("data: ")[1])
+        assert parsed["delta"]["text"] == "lo"
+
+    def test_empty_content_diff_no_event(self):
+        e = AnthropicStreamEmitter()
+        e.start("msg_1", "m")
+        e.feed({"type": "content", "text": "Hi"})
+        events = e.feed({"type": "content", "text": "Hi"})
+        assert events == []
+
+    def test_tool_start_closes_text_opens_tool_block(self):
+        e = AnthropicStreamEmitter()
+        e.start("msg_1", "m")
+        e.feed({"type": "content", "text": "Thinking"})
+        events = e.feed(
+            {
+                "type": "tool_start",
+                "tool_name": "web_search",
+                "tool_call_id": "tc_1",
+                "arguments": {"query": "test"},
+            }
+        )
+        # content_block_stop + content_block_start(tool_use) + content_block_delta(input_json)
+        assert len(events) == 3
+        assert "content_block_stop" in events[0]
+        assert "tool_use" in events[1]
+        assert "input_json_delta" in events[2]
+
+    def test_tool_end_closes_tool_opens_new_text_block(self):
+        e = AnthropicStreamEmitter()
+        e.start("msg_1", "m")
+        e.feed(
+            {
+                "type": "tool_start",
+                "tool_name": "t",
+                "tool_call_id": "tc_1",
+                "arguments": {},
+            }
+        )
+        events = e.feed(
+            {
+                "type": "tool_end",
+                "tool_name": "t",
+                "tool_call_id": "tc_1",
+                "result": "done",
+            }
+        )
+        # content_block_stop (tool) + tool_result + content_block_start (new text)
+        assert len(events) == 3
+        assert "content_block_stop" in events[0]
+        assert "tool_result" in events[1]
+        parsed = json.loads(events[1].split("data: ")[1])
+        assert parsed["content"] == "done"
+        assert parsed["tool_use_id"] == "tc_1"
+        assert "content_block_start" in events[2]
+        assert '"type": "text"' in events[2]
+
+    def test_finish_emits_stop_events(self):
+        e = AnthropicStreamEmitter()
+        e.start("msg_1", "m")
+        events = e.finish("end_turn")
+        # content_block_stop + message_delta + message_stop
+        assert len(events) == 3
+        assert "content_block_stop" in events[0]
+        assert "message_delta" in events[1]
+        assert "end_turn" in events[1]
+        assert "message_stop" in events[2]
+
+    def test_metadata_captured_in_finish_usage(self):
+        e = AnthropicStreamEmitter()
+        e.start("msg_1", "m")
+        e.feed(
+            {
+                "type": "metadata",
+                "usage": {"prompt_tokens": 10, "completion_tokens": 20},
+            }
+        )
+        events = e.finish("end_turn")
+        delta_event = [ev for ev in events if "message_delta" in ev][0]
+        parsed = json.loads(delta_event.split("data: ")[1])
+        assert parsed["usage"]["output_tokens"] == 20
+
+    def test_status_events_ignored(self):
+        e = AnthropicStreamEmitter()
+        e.start("msg_1", "m")
+        events = e.feed({"type": "status", "text": "Searching..."})
+        assert events == []
+
+    def test_no_tool_calls_simple_text_flow(self):
+        e = AnthropicStreamEmitter()
+        start_events = e.start("msg_1", "m")
+        content_events = e.feed({"type": "content", "text": "Hello world"})
+        meta_events = e.feed(
+            {"type": "metadata", "usage": {"prompt_tokens": 5, "completion_tokens": 2}}
+        )
+        end_events = e.finish("end_turn")
+
+        assert len(start_events) == 2
+        assert len(content_events) == 1
+        assert meta_events == []
+        assert len(end_events) == 3
+
+    def test_block_index_increments(self):
+        e = AnthropicStreamEmitter()
+        e.start("msg_1", "m")
+        assert e.block_index == 0
+        e.feed(
+            {
+                "type": "tool_start",
+                "tool_name": "t",
+                "tool_call_id": "tc_1",
+                "arguments": {},
+            }
+        )
+        assert e.block_index == 1
+        e.feed(
+            {
+                "type": "tool_end",
+                "tool_name": "t",
+                "tool_call_id": "tc_1",
+                "result": "ok",
+            }
+        )
+        assert e.block_index == 2
+
+    def test_text_after_tool_resets_prev_text(self):
+        e = AnthropicStreamEmitter()
+        e.start("msg_1", "m")
+        e.feed({"type": "content", "text": "Before tool"})
+        e.feed(
+            {
+                "type": "tool_start",
+                "tool_name": "t",
+                "tool_call_id": "tc_1",
+                "arguments": {},
+            }
+        )
+        e.feed(
+            {
+                "type": "tool_end",
+                "tool_name": "t",
+                "tool_call_id": "tc_1",
+                "result": "ok",
+            }
+        )
+        # After tool_end, prev_text should be reset
+        events = e.feed({"type": "content", "text": "After tool"})
+        parsed = json.loads(events[0].split("data: ")[1])
+        assert parsed["delta"]["text"] == "After tool"
+
+
+# =====================================================================
+# Pass-through emitter tests (client-side tool execution path)
+# =====================================================================
+
+
+class TestAnthropicPassthroughEmitter:
+    def _parse(self, event_str):
+        return json.loads(event_str.split("data: ")[1])
+
+    def test_start_emits_message_start_only(self):
+        e = AnthropicPassthroughEmitter()
+        events = e.start("msg_1", "test-model")
+        assert len(events) == 1
+        assert "message_start" in events[0]
+        parsed = self._parse(events[0])
+        assert parsed["message"]["id"] == "msg_1"
+        assert parsed["message"]["model"] == "test-model"
+
+    def test_text_chunk_opens_text_block_and_emits_delta(self):
+        e = AnthropicPassthroughEmitter()
+        e.start("msg_1", "m")
+        chunk = {"choices": [{"delta": {"content": "Hello"}}]}
+        events = e.feed_chunk(chunk)
+        # content_block_start + content_block_delta
+        assert len(events) == 2
+        assert "content_block_start" in events[0]
+        assert '"type": "text"' in events[0]
+        delta = self._parse(events[1])
+        assert delta["delta"]["type"] == "text_delta"
+        assert delta["delta"]["text"] == "Hello"
+
+    def test_sequential_text_chunks_single_block(self):
+        e = AnthropicPassthroughEmitter()
+        e.start("msg_1", "m")
+        events1 = e.feed_chunk({"choices": [{"delta": {"content": "Hello"}}]})
+        events2 = e.feed_chunk({"choices": [{"delta": {"content": " world"}}]})
+        # First chunk opens the block, second only emits delta
+        assert len(events1) == 2
+        assert len(events2) == 1
+        assert self._parse(events2[0])["delta"]["text"] == " world"
+
+    def test_tool_call_opens_tool_use_block(self):
+        e = AnthropicPassthroughEmitter()
+        e.start("msg_1", "m")
+        chunk = {
+            "choices": [
+                {
+                    "delta": {
+                        "tool_calls": [
+                            {
+                                "index": 0,
+                                "id": "call_1",
+                                "type": "function",
+                                "function": {"name": "Bash", "arguments": ""},
+                            }
+                        ]
+                    }
+                }
+            ]
+        }
+        events = e.feed_chunk(chunk)
+        assert len(events) == 1
+        parsed = self._parse(events[0])
+        assert parsed["type"] == "content_block_start"
+        assert parsed["content_block"]["type"] == "tool_use"
+        assert parsed["content_block"]["id"] == "call_1"
+        assert parsed["content_block"]["name"] == "Bash"
+
+    def test_tool_call_arguments_streamed_as_input_json_delta(self):
+        e = AnthropicPassthroughEmitter()
+        e.start("msg_1", "m")
+        # Open the tool call
+        e.feed_chunk(
+            {
+                "choices": [
+                    {
+                        "delta": {
+                            "tool_calls": [
+                                {
+                                    "index": 0,
+                                    "id": "c1",
+                                    "type": "function",
+                                    "function": {"name": "Bash", "arguments": ""},
+                                }
+                            ]
+                        }
+                    }
+                ]
+            }
+        )
+        # Stream argument fragments
+        events1 = e.feed_chunk(
+            {
+                "choices": [
+                    {
+                        "delta": {
+                            "tool_calls": [
+                                {"index": 0, "function": {"arguments": '{"cmd'}}
+                            ]
+                        }
+                    }
+                ]
+            }
+        )
+        events2 = e.feed_chunk(
+            {
+                "choices": [
+                    {
+                        "delta": {
+                            "tool_calls": [
+                                {"index": 0, "function": {"arguments": '": "ls"}'}}
+                            ]
+                        }
+                    }
+                ]
+            }
+        )
+        parsed1 = self._parse(events1[0])
+        parsed2 = self._parse(events2[0])
+        assert parsed1["delta"]["type"] == "input_json_delta"
+        assert parsed1["delta"]["partial_json"] == '{"cmd'
+        assert parsed2["delta"]["partial_json"] == '": "ls"}'
+
+    def test_text_then_tool_closes_text_block(self):
+        e = AnthropicPassthroughEmitter()
+        e.start("msg_1", "m")
+        e.feed_chunk({"choices": [{"delta": {"content": "Let me check."}}]})
+        events = e.feed_chunk(
+            {
+                "choices": [
+                    {
+                        "delta": {
+                            "tool_calls": [
+                                {
+                                    "index": 0,
+                                    "id": "c1",
+                                    "type": "function",
+                                    "function": {"name": "Bash", "arguments": ""},
+                                }
+                            ]
+                        }
+                    }
+                ]
+            }
+        )
+        # Should close text block and open tool_use block
+        assert "content_block_stop" in events[0]
+        assert "content_block_start" in events[1]
+        assert '"type": "tool_use"' in events[1]
+
+    def test_finish_reason_tool_calls_sets_tool_use_stop(self):
+        e = AnthropicPassthroughEmitter()
+        e.start("msg_1", "m")
+        e.feed_chunk(
+            {
+                "choices": [
+                    {
+                        "delta": {
+                            "tool_calls": [
+                                {
+                                    "index": 0,
+                                    "id": "c1",
+                                    "type": "function",
+                                    "function": {"name": "Bash", "arguments": "{}"},
+                                }
+                            ]
+                        }
+                    }
+                ]
+            }
+        )
+        e.feed_chunk({"choices": [{"delta": {}, "finish_reason": "tool_calls"}]})
+        events = e.finish()
+        delta_event = [ev for ev in events if "message_delta" in ev][0]
+        parsed = self._parse(delta_event)
+        assert parsed["delta"]["stop_reason"] == "tool_use"
+
+    def test_finish_reason_stop_sets_end_turn(self):
+        e = AnthropicPassthroughEmitter()
+        e.start("msg_1", "m")
+        e.feed_chunk({"choices": [{"delta": {"content": "Hi"}}]})
+        e.feed_chunk({"choices": [{"delta": {}, "finish_reason": "stop"}]})
+        events = e.finish()
+        delta_event = [ev for ev in events if "message_delta" in ev][0]
+        parsed = self._parse(delta_event)
+        assert parsed["delta"]["stop_reason"] == "end_turn"
+
+    def test_finish_reason_length_sets_max_tokens(self):
+        e = AnthropicPassthroughEmitter()
+        e.start("msg_1", "m")
+        e.feed_chunk({"choices": [{"delta": {"content": "Hi"}}]})
+        e.feed_chunk({"choices": [{"delta": {}, "finish_reason": "length"}]})
+        events = e.finish()
+        delta_event = [ev for ev in events if "message_delta" in ev][0]
+        parsed = self._parse(delta_event)
+        assert parsed["delta"]["stop_reason"] == "max_tokens"
+
+    def test_finish_closes_current_block(self):
+        e = AnthropicPassthroughEmitter()
+        e.start("msg_1", "m")
+        e.feed_chunk({"choices": [{"delta": {"content": "Hi"}}]})
+        events = e.finish()
+        assert "content_block_stop" in events[0]
+        assert "message_delta" in events[1]
+        assert "message_stop" in events[2]
+
+    def test_usage_chunk_captured(self):
+        e = AnthropicPassthroughEmitter()
+        e.start("msg_1", "m")
+        e.feed_chunk({"choices": [{"delta": {"content": "Hi"}}]})
+        e.feed_chunk(
+            {
+                "choices": [],
+                "usage": {"prompt_tokens": 10, "completion_tokens": 5},
+            }
+        )
+        events = e.finish()
+        delta_event = [ev for ev in events if "message_delta" in ev][0]
+        parsed = self._parse(delta_event)
+        assert parsed["usage"]["output_tokens"] == 5
+
+    def test_empty_chunk_returns_no_events(self):
+        e = AnthropicPassthroughEmitter()
+        e.start("msg_1", "m")
+        events = e.feed_chunk({"choices": []})
+        assert events == []
+
+    def test_no_blocks_at_all_still_produces_valid_finish(self):
+        e = AnthropicPassthroughEmitter()
+        e.start("msg_1", "m")
+        events = e.finish()
+        # No content_block_stop because no block was opened
+        assert not any("content_block_stop" in ev for ev in events)
+        assert any("message_delta" in ev for ev in events)
+        assert any("message_stop" in ev for ev in events)
+
+    def test_multiple_tool_calls_distinct_blocks(self):
+        e = AnthropicPassthroughEmitter()
+        e.start("msg_1", "m")
+        # First tool call
+        e.feed_chunk(
+            {
+                "choices": [
+                    {
+                        "delta": {
+                            "tool_calls": [
+                                {
+                                    "index": 0,
+                                    "id": "c1",
+                                    "type": "function",
+                                    "function": {"name": "Bash", "arguments": "{}"},
+                                }
+                            ]
+                        }
+                    }
+                ]
+            }
+        )
+        # Second tool call (different index)
+        events = e.feed_chunk(
+            {
+                "choices": [
+                    {
+                        "delta": {
+                            "tool_calls": [
+                                {
+                                    "index": 1,
+                                    "id": "c2",
+                                    "type": "function",
+                                    "function": {"name": "Read", "arguments": "{}"},
+                                }
+                            ]
+                        }
+                    }
+                ]
+            }
+        )
+        # Should close block 0, open block 1
+        assert "content_block_stop" in events[0]
+        assert "content_block_start" in events[1]
+        parsed = self._parse(events[1])
+        assert parsed["content_block"]["name"] == "Read"
+        assert parsed["content_block"]["id"] == "c2"
diff --git a/studio/backend/tests/test_responses_api.py b/studio/backend/tests/test_responses_api.py
new file mode 100644
index 0000000000..5b55f87259
--- /dev/null
+++ b/studio/backend/tests/test_responses_api.py
@@ -0,0 +1,328 @@
+# SPDX-License-Identifier: AGPL-3.0-only
+# Copyright 2026-present the Unsloth AI Inc. team. All rights reserved.
+
+"""
+Tests for the OpenAI Responses API schemas and input normalisation.
+These tests do NOT require a running server or GPU -- they validate
+the Pydantic models and the _normalise_responses_input helper.
+"""
+
+import sys
+import os
+import json
+import re
+
+# Ensure backend is on path
+_backend = os.path.join(os.path.dirname(__file__), "..")
+sys.path.insert(0, _backend)
+
+from models.inference import (
+    ResponsesRequest,
+    ResponsesInputMessage,
+    ResponsesInputTextPart,
+    ResponsesInputImagePart,
+    ResponsesOutputTextContent,
+    ResponsesOutputMessage,
+    ResponsesUsage,
+    ResponsesResponse,
+    ChatMessage,
+    TextContentPart,
+    ImageContentPart,
+    ImageUrl,
+    ChatCompletionRequest,
+)
+
+
+# ── _normalise_responses_input: copied from routes/inference.py ──
+# We cannot import routes.inference directly because routes/__init__.py
+# pulls in heavy dependencies (structlog/twisted/torch). This is a
+# direct copy of the function for testing purposes.
+
+
+def _normalise_responses_input(payload: ResponsesRequest) -> list:
+    """Convert a ResponsesRequest into a list of ChatMessage for the completions backend."""
+    messages = []
+
+    # System / developer instructions
+    if payload.instructions:
+        messages.append(ChatMessage(role = "system", content = payload.instructions))
+
+    # Simple string input
+    if isinstance(payload.input, str):
+        if payload.input:
+            messages.append(ChatMessage(role = "user", content = payload.input))
+        return messages
+
+    # List of ResponsesInputMessage
+    for msg in payload.input:
+        role = "system" if msg.role == "developer" else msg.role
+
+        if isinstance(msg.content, str):
+            messages.append(ChatMessage(role = role, content = msg.content))
+        else:
+            # Convert Responses content parts -> Chat content parts
+            parts = []
+            for part in msg.content:
+                if isinstance(part, ResponsesInputTextPart):
+                    parts.append(TextContentPart(type = "text", text = part.text))
+                elif isinstance(part, ResponsesInputImagePart):
+                    parts.append(
+                        ImageContentPart(
+                            type = "image_url",
+                            image_url = ImageUrl(url = part.image_url, detail = part.detail),
+                        )
+                    )
+            messages.append(ChatMessage(role = role, content = parts if parts else ""))
+
+    return messages
+
+
+# =====================================================================
+# Schema validation tests
+# =====================================================================
+
+
+class TestResponsesRequest:
+    """Validate ResponsesRequest accepts the shapes the OpenAI SDK sends."""
+
+    def test_minimal_string_input(self):
+        req = ResponsesRequest(input = "Hello")
+        assert req.input == "Hello"
+        assert req.stream is False
+        assert req.model == "default"
+
+    def test_message_list_input(self):
+        req = ResponsesRequest(
+            input = [
+                {"role": "user", "content": "Hi"},
+                {"role": "assistant", "content": "Hello!"},
+            ],
+        )
+        assert len(req.input) == 2
+        assert req.input[0].role == "user"
+        assert req.input[0].content == "Hi"
+
+    def test_multimodal_input(self):
+        req = ResponsesRequest(
+            input = [
+                {
+                    "role": "user",
+                    "content": [
+                        {"type": "input_text", "text": "What is in this image?"},
+                        {
+                            "type": "input_image",
+                            "image_url": "https://example.com/img.png",
+                        },
+                    ],
+                },
+            ],
+        )
+        parts = req.input[0].content
+        assert len(parts) == 2
+        assert isinstance(parts[0], ResponsesInputTextPart)
+        assert isinstance(parts[1], ResponsesInputImagePart)
+
+    def test_instructions_field(self):
+        req = ResponsesRequest(
+            input = "test",
+            instructions = "You are a helpful assistant.",
+        )
+        assert req.instructions == "You are a helpful assistant."
+
+    def test_extra_fields_accepted(self):
+        """OpenAI SDK may send fields we don't model -- extra='allow' should pass."""
+        req = ResponsesRequest(
+            input = "test",
+            tools = [{"type": "web_search_preview"}],
+            store = True,
+            metadata = {"key": "value"},
+            previous_response_id = "resp_abc123",
+        )
+        assert req.tools == [{"type": "web_search_preview"}]
+        assert req.store is True
+
+    def test_stream_flag(self):
+        req = ResponsesRequest(input = "test", stream = True)
+        assert req.stream is True
+
+    def test_temperature_and_top_p(self):
+        req = ResponsesRequest(input = "test", temperature = 0.8, top_p = 0.9)
+        assert req.temperature == 0.8
+        assert req.top_p == 0.9
+
+    def test_max_output_tokens(self):
+        req = ResponsesRequest(input = "test", max_output_tokens = 512)
+        assert req.max_output_tokens == 512
+
+    def test_developer_role(self):
+        req = ResponsesRequest(
+            input = [{"role": "developer", "content": "System instructions"}],
+        )
+        assert req.input[0].role == "developer"
+
+
+# =====================================================================
+# Response model tests
+# =====================================================================
+
+
+class TestResponsesResponse:
+    """Validate response models serialise correctly."""
+
+    def test_basic_response(self):
+        resp = ResponsesResponse(
+            model = "test-model",
+            output = [
+                ResponsesOutputMessage(
+                    content = [ResponsesOutputTextContent(text = "Hello!")]
+                ),
+            ],
+            usage = ResponsesUsage(input_tokens = 10, output_tokens = 5, total_tokens = 15),
+        )
+        d = resp.model_dump()
+        assert d["object"] == "response"
+        assert d["status"] == "completed"
+        assert d["output"][0]["type"] == "message"
+        assert d["output"][0]["content"][0]["type"] == "output_text"
+        assert d["output"][0]["content"][0]["text"] == "Hello!"
+        assert d["usage"]["input_tokens"] == 10
+        assert d["usage"]["output_tokens"] == 5
+        assert d["usage"]["total_tokens"] == 15
+        # Must NOT have prompt_tokens / completion_tokens
+        assert "prompt_tokens" not in d["usage"]
+        assert "completion_tokens" not in d["usage"]
+
+    def test_id_format(self):
+        resp = ResponsesResponse()
+        assert resp.id.startswith("resp_")
+
+    def test_output_message_id_format(self):
+        msg = ResponsesOutputMessage()
+        assert msg.id.startswith("msg_")
+
+    def test_annotations_default_empty(self):
+        part = ResponsesOutputTextContent(text = "hi")
+        assert part.annotations == []
+
+    def test_response_json_roundtrip(self):
+        resp = ResponsesResponse(
+            model = "gpt-4",
+            output = [
+                ResponsesOutputMessage(
+                    content = [ResponsesOutputTextContent(text = "ok")],
+                ),
+            ],
+            usage = ResponsesUsage(input_tokens = 1, output_tokens = 1, total_tokens = 2),
+        )
+        j = json.loads(resp.model_dump_json())
+        assert j["object"] == "response"
+        assert j["output"][0]["role"] == "assistant"
+        assert j["output"][0]["status"] == "completed"
+
+
+# =====================================================================
+# Input normalisation tests
+# =====================================================================
+
+
+class TestNormaliseResponsesInput:
+    """Test _normalise_responses_input converts Responses input to ChatMessages."""
+
+    def test_string_input(self):
+        payload = ResponsesRequest(input = "Hello world")
+        msgs = _normalise_responses_input(payload)
+        assert len(msgs) == 1
+        assert msgs[0].role == "user"
+        assert msgs[0].content == "Hello world"
+
+    def test_instructions_become_system_message(self):
+        payload = ResponsesRequest(
+            input = "Hi",
+            instructions = "Be concise.",
+        )
+        msgs = _normalise_responses_input(payload)
+        assert len(msgs) == 2
+        assert msgs[0].role == "system"
+        assert msgs[0].content == "Be concise."
+        assert msgs[1].role == "user"
+        assert msgs[1].content == "Hi"
+
+    def test_message_list(self):
+        payload = ResponsesRequest(
+            input = [
+                {"role": "user", "content": "First"},
+                {"role": "assistant", "content": "Response"},
+                {"role": "user", "content": "Second"},
+            ],
+        )
+        msgs = _normalise_responses_input(payload)
+        assert len(msgs) == 3
+        assert msgs[0].role == "user"
+        assert msgs[1].role == "assistant"
+        assert msgs[2].role == "user"
+
+    def test_developer_role_maps_to_system(self):
+        payload = ResponsesRequest(
+            input = [{"role": "developer", "content": "Instructions"}],
+        )
+        msgs = _normalise_responses_input(payload)
+        assert msgs[0].role == "system"
+        assert msgs[0].content == "Instructions"
+
+    def test_multimodal_parts(self):
+        payload = ResponsesRequest(
+            input = [
+                {
+                    "role": "user",
+                    "content": [
+                        {"type": "input_text", "text": "Describe this:"},
+                        {
+                            "type": "input_image",
+                            "image_url": "data:image/png;base64,abc",
+                        },
+                    ],
+                },
+            ],
+        )
+        msgs = _normalise_responses_input(payload)
+        assert len(msgs) == 1
+        content = msgs[0].content
+        assert isinstance(content, list)
+        assert len(content) == 2
+        assert isinstance(content[0], TextContentPart)
+        assert content[0].text == "Describe this:"
+        assert isinstance(content[1], ImageContentPart)
+        assert content[1].image_url.url == "data:image/png;base64,abc"
+
+    def test_empty_string_input(self):
+        payload = ResponsesRequest(input = "")
+        msgs = _normalise_responses_input(payload)
+        assert len(msgs) == 0
+
+    def test_empty_list_input(self):
+        payload = ResponsesRequest(input = [])
+        msgs = _normalise_responses_input(payload)
+        assert len(msgs) == 0
+
+    def test_instructions_only(self):
+        payload = ResponsesRequest(input = "", instructions = "System msg")
+        msgs = _normalise_responses_input(payload)
+        assert len(msgs) == 1
+        assert msgs[0].role == "system"
+
+    def test_instructions_plus_message_list(self):
+        payload = ResponsesRequest(
+            input = [{"role": "user", "content": "Hello"}],
+            instructions = "Be brief.",
+        )
+        msgs = _normalise_responses_input(payload)
+        assert len(msgs) == 2
+        assert msgs[0].role == "system"
+        assert msgs[0].content == "Be brief."
+        assert msgs[1].role == "user"
+
+
+if __name__ == "__main__":
+    import pytest
+
+    pytest.main([__file__, "-v"])
diff --git a/studio/backend/tests/test_studio_api.py b/studio/backend/tests/test_studio_api.py
new file mode 100644
index 0000000000..9cc17c89fb
--- /dev/null
+++ b/studio/backend/tests/test_studio_api.py
@@ -0,0 +1,643 @@
+# SPDX-License-Identifier: AGPL-3.0-only
+# Copyright 2026-present the Unsloth AI Inc. team. All rights reserved. See /studio/LICENSE.AGPL-3.0
+
+"""
+End-to-end tests for Unsloth Studio's HTTP API surface.
+
+Covers the OpenAI-compatible and Anthropic-compatible endpoints exposed
+by the server that ``unsloth studio run`` boots, plus API key
+authentication and the CLI's ``--help`` output:
+
+    1. curl -- basic chat completions (non-streaming)
+    2. curl -- streaming chat completions
+    3. Python OpenAI SDK -- streaming completions
+    4. curl -- with tools (web_search + python)
+    5. Anthropic Messages API -- basic non-streaming
+    6. Anthropic Messages API -- streaming SSE
+    7. Anthropic Python SDK -- non-streaming
+    8. Anthropic Messages API -- streaming with tools
+
+Training, export, fine-tuning, and chat-UI concerns are out of scope —
+see the unit suites elsewhere under ``studio/backend/tests/`` for those.
+
+Usage:
+
+    # Script mode — launches its own server via ``unsloth studio run``.
+    python tests/test_studio_api.py
+    python tests/test_studio_api.py --model unsloth/... --gguf-variant ...
+
+    # Pytest mode, external server — start a Studio server yourself,
+    # then point pytest at it. Fastest iteration loop.
+    unsloth studio run --model unsloth/Qwen3-1.7B-GGUF --gguf-variant UD-Q4_K_XL &
+    export UNSLOTH_E2E_BASE_URL=http://127.0.0.1:8080
+    export UNSLOTH_E2E_API_KEY=sk-unsloth-...   # from the server banner
+    pytest tests/test_studio_api.py -v
+
+    # Pytest mode, fixture-managed server — pytest launches and tears
+    # down the server itself. One-shot verification, CI-friendly.
+    pytest tests/test_studio_api.py -v \\
+        --unsloth-model unsloth/Qwen3-1.7B-GGUF \\
+        --unsloth-gguf-variant UD-Q4_K_XL
+
+The ``base_url`` / ``api_key`` parameters on the test functions resolve
+via the ``studio_server`` session fixture in ``conftest.py``.
+
+Requires a GPU and ~2 GB of disk for the GGUF download.
+"""
+
+from __future__ import annotations
+
+import argparse
+import json
+import os
+import re
+import signal
+import subprocess
+import sys
+import time
+import urllib.error
+import urllib.request
+from pathlib import Path
+
+
+# ── Configuration ────────────────────────────────────────────────────
+
+DEFAULT_MODEL = "unsloth/Qwen3-1.7B-GGUF"
+DEFAULT_VARIANT = "UD-Q4_K_XL"
+PORT = 18222  # high port unlikely to collide
+HOST = "127.0.0.1"
+STARTUP_TIMEOUT = 120  # seconds to wait for banner
+LOG_FILE = (
+    Path(__file__).resolve().parent.parent.parent.parent
+    / "temp"
+    / "test_studio_api.log"
+)
+
+
+# ── Helpers ──────────────────────────────────────────────────────────
+
+
+def _http(
+    method: str,
+    url: str,
+    *,
+    body: dict | None = None,
+    headers: dict | None = None,
+    timeout: int = 60,
+) -> tuple[int, str]:
+    """Minimal stdlib HTTP helper.  Returns (status_code, body_text)."""
+    data = json.dumps(body).encode() if body else None
+    req = urllib.request.Request(url, data = data, headers = headers or {}, method = method)
+    if body:
+        req.add_header("Content-Type", "application/json")
+    try:
+        with urllib.request.urlopen(req, timeout = timeout) as resp:
+            return resp.status, resp.read().decode()
+    except urllib.error.HTTPError as exc:
+        return exc.code, exc.read().decode(errors = "replace")
+
+
+def _stream_http(
+    url: str,
+    *,
+    body: dict,
+    headers: dict,
+    timeout: int = 60,
+) -> tuple[int, list[dict]]:
+    """POST a streaming request and collect SSE chunks."""
+    data = json.dumps(body).encode()
+    req = urllib.request.Request(url, data = data, headers = headers, method = "POST")
+    req.add_header("Content-Type", "application/json")
+    chunks: list[dict] = []
+    try:
+        with urllib.request.urlopen(req, timeout = timeout) as resp:
+            status = resp.status
+            for raw_line in resp:
+                line = raw_line.decode().strip()
+                if line.startswith("data: ") and line != "data: [DONE]":
+                    try:
+                        chunks.append(json.loads(line[6:]))
+                    except json.JSONDecodeError:
+                        pass
+            return status, chunks
+    except urllib.error.HTTPError as exc:
+        return exc.code, []
+
+
+# ── Test functions ───────────────────────────────────────────────────
+
+
+def test_help_output():
+    """``unsloth studio run --help`` should show all documented options."""
+    result = subprocess.run(
+        ["unsloth", "studio", "run", "--help"],
+        capture_output = True,
+        text = True,
+        timeout = 15,
+    )
+    out = result.stdout
+    assert result.returncode == 0, f"--help exited with {result.returncode}"
+
+    for flag in [
+        "--model",
+        "--gguf-variant",
+        "--max-seq-length",
+        "--load-in-4bit",
+        "--api-key-name",
+        "--port",
+        "--host",
+        "--frontend",
+        "--silent",
+    ]:
+        assert flag in out, f"Missing flag {flag!r} in --help output"
+    print("  PASS  --help shows all flags")
+
+
+def test_curl_basic(base_url: str, api_key: str):
+    """Example 1: basic non-streaming chat completion via HTTP."""
+    status, text = _http(
+        "POST",
+        f"{base_url}/v1/chat/completions",
+        body = {
+            "messages": [{"role": "user", "content": "Say just the word hello"}],
+            "stream": False,
+        },
+        headers = {"Authorization": f"Bearer {api_key}"},
+    )
+    assert status == 200, f"Expected 200, got {status}: {text[:300]}"
+    data = json.loads(text)
+    assert "choices" in data, f"Missing 'choices' in response: {text[:300]}"
+    content = data["choices"][0]["message"]["content"]
+    assert len(content) > 0, "Empty assistant content"
+    print(f"  PASS  curl basic: {content[:80]!r}")
+
+
+def _collect_streamed_content(chunks: list[dict]) -> str:
+    """Extract text from SSE chunks, skipping role-only and usage chunks."""
+    parts = []
+    for c in chunks:
+        choices = c.get("choices", [])
+        if not choices:
+            continue
+        delta = choices[0].get("delta", {})
+        part = delta.get("content")
+        if part:
+            parts.append(part)
+    return "".join(parts)
+
+
+def test_curl_streaming(base_url: str, api_key: str):
+    """Example 2: streaming chat completion via HTTP SSE."""
+    status, chunks = _stream_http(
+        f"{base_url}/v1/chat/completions",
+        body = {
+            "messages": [{"role": "user", "content": "Count from 1 to 3"}],
+            "stream": True,
+        },
+        headers = {"Authorization": f"Bearer {api_key}"},
+    )
+    assert status == 200, f"Expected 200, got {status}"
+    assert len(chunks) > 0, "No SSE chunks received"
+    full = _collect_streamed_content(chunks)
+    assert len(full) > 0, "Streamed content is empty"
+    print(f"  PASS  curl streaming: got {len(chunks)} chunks, {len(full)} chars")
+
+
+def test_openai_sdk(base_url: str, api_key: str):
+    """Example 3: OpenAI Python SDK streaming completion."""
+    try:
+        from openai import OpenAI
+    except ImportError:
+        print("  SKIP  openai SDK not installed")
+        return
+
+    client = OpenAI(base_url = f"{base_url}/v1", api_key = api_key)
+    response = client.chat.completions.create(
+        model = "current",
+        messages = [
+            {"role": "user", "content": "What is 2+2? Answer with just the number."}
+        ],
+        stream = True,
+    )
+    content_parts = []
+    for chunk in response:
+        if not chunk.choices:
+            continue
+        delta_content = chunk.choices[0].delta.content
+        if delta_content:
+            content_parts.append(delta_content)
+    full = "".join(content_parts)
+    assert len(full) > 0, "OpenAI SDK returned empty content"
+    print(f"  PASS  OpenAI SDK streaming: {full.strip()[:80]!r}")
+
+
+def test_curl_with_tools(base_url: str, api_key: str):
+    """Example 4: chat completion with tool calling enabled.
+
+    Note: when ``enable_tools`` is set the server always returns SSE
+    streaming regardless of the ``stream`` flag, so we parse SSE chunks.
+    The model may or may not produce visible content -- tool orchestration
+    can intercept the response -- so we only assert the endpoint succeeds.
+    """
+    status, chunks = _stream_http(
+        f"{base_url}/v1/chat/completions",
+        body = {
+            "messages": [
+                {
+                    "role": "user",
+                    "content": "What is 123 * 456? Use code to compute it.",
+                }
+            ],
+            "stream": True,
+            "enable_tools": True,
+            "enabled_tools": ["python"],
+            "session_id": "test-session",
+        },
+        headers = {"Authorization": f"Bearer {api_key}"},
+        timeout = 120,
+    )
+    assert status == 200, f"Expected 200, got {status}"
+    assert len(chunks) > 0, "No SSE chunks received for tools request"
+
+    # Check that at least one chunk has the expected shape
+    has_valid_chunk = any("choices" in c or "type" in c for c in chunks)
+    assert has_valid_chunk, "No valid chunks in tools response"
+    full = _collect_streamed_content(chunks)
+    print(f"  PASS  curl with tools: {len(chunks)} chunks, {len(full)} chars content")
+
+
+def test_invalid_key_rejected(base_url: str):
+    """Requests with a bad API key should be rejected."""
+    status, _text = _http(
+        "POST",
+        f"{base_url}/v1/chat/completions",
+        body = {
+            "messages": [{"role": "user", "content": "Hello"}],
+            "stream": False,
+        },
+        headers = {"Authorization": "Bearer sk-unsloth-boguskey123"},
+    )
+    assert status == 401, f"Expected 401 for invalid key, got {status}"
+    print("  PASS  invalid API key rejected (401)")
+
+
+def test_no_key_rejected(base_url: str):
+    """Requests without any auth header should be rejected."""
+    status, _text = _http(
+        "POST",
+        f"{base_url}/v1/chat/completions",
+        body = {
+            "messages": [{"role": "user", "content": "Hello"}],
+            "stream": False,
+        },
+    )
+    assert status == 401 or status == 403, f"Expected 401/403 for no key, got {status}"
+    print(f"  PASS  no API key rejected ({status})")
+
+
+# ── Anthropic SSE helper ─────────────────────────────────────────────
+
+
+def _stream_anthropic_http(
+    url: str,
+    *,
+    body: dict,
+    headers: dict,
+    timeout: int = 60,
+) -> tuple[int, list[tuple[str, dict]]]:
+    """POST a streaming request and collect Anthropic SSE events.
+
+    Returns (status, [(event_type, data_dict), ...]).
+    """
+    data = json.dumps(body).encode()
+    req = urllib.request.Request(url, data = data, headers = headers, method = "POST")
+    req.add_header("Content-Type", "application/json")
+    events: list[tuple[str, dict]] = []
+    try:
+        with urllib.request.urlopen(req, timeout = timeout) as resp:
+            status = resp.status
+            current_event = None
+            for raw_line in resp:
+                line = raw_line.decode().strip()
+                if line.startswith("event: "):
+                    current_event = line[7:]
+                elif line.startswith("data: ") and current_event:
+                    try:
+                        events.append((current_event, json.loads(line[6:])))
+                    except json.JSONDecodeError:
+                        pass
+                    current_event = None
+            return status, events
+    except urllib.error.HTTPError as exc:
+        return exc.code, []
+
+
+def _collect_anthropic_text(events: list[tuple[str, dict]]) -> str:
+    """Extract text content from Anthropic SSE events."""
+    parts = []
+    for etype, data in events:
+        if etype == "content_block_delta":
+            delta = data.get("delta", {})
+            if delta.get("type") == "text_delta":
+                parts.append(delta.get("text", ""))
+    return "".join(parts)
+
+
+# ── Anthropic /v1/messages test functions ────────────────────────────
+
+
+def test_anthropic_basic(base_url: str, api_key: str):
+    """Anthropic Messages API: non-streaming."""
+    status, text = _http(
+        "POST",
+        f"{base_url}/v1/messages",
+        body = {
+            "model": "default",
+            "max_tokens": 100,
+            "messages": [{"role": "user", "content": "Say just the word hello"}],
+        },
+        headers = {"Authorization": f"Bearer {api_key}"},
+    )
+    assert status == 200, f"Expected 200, got {status}: {text[:300]}"
+    data = json.loads(text)
+    assert data.get("type") == "message", f"Expected type 'message': {text[:300]}"
+    assert data.get("role") == "assistant"
+    content = data.get("content", [])
+    assert len(content) > 0, "Empty content array"
+    text_block = content[-1]
+    assert text_block.get("type") == "text", f"Expected text block: {text_block}"
+    assert len(text_block.get("text", "")) > 0, "Empty text in response"
+    print(f"  PASS  anthropic basic: {text_block['text'][:80]!r}")
+
+
+def test_anthropic_streaming(base_url: str, api_key: str):
+    """Anthropic Messages API: streaming SSE."""
+    status, events = _stream_anthropic_http(
+        f"{base_url}/v1/messages",
+        body = {
+            "model": "default",
+            "max_tokens": 100,
+            "messages": [{"role": "user", "content": "Count from 1 to 3"}],
+            "stream": True,
+        },
+        headers = {"Authorization": f"Bearer {api_key}"},
+    )
+    assert status == 200, f"Expected 200, got {status}"
+    assert len(events) > 0, "No SSE events received"
+
+    event_types = [e[0] for e in events]
+    assert "message_start" in event_types, "Missing message_start event"
+    assert "message_stop" in event_types, "Missing message_stop event"
+
+    full = _collect_anthropic_text(events)
+    assert len(full) > 0, "Streamed text content is empty"
+    print(f"  PASS  anthropic streaming: {len(events)} events, {len(full)} chars")
+
+
+def test_anthropic_sdk(base_url: str, api_key: str):
+    """Anthropic Python SDK: non-streaming."""
+    try:
+        from anthropic import Anthropic
+    except ImportError:
+        print("  SKIP  anthropic SDK not installed")
+        return
+
+    client = Anthropic(base_url = f"{base_url}/v1", api_key = api_key)
+    message = client.messages.create(
+        model = "default",
+        max_tokens = 100,
+        messages = [
+            {"role": "user", "content": "What is 2+2? Answer with just the number."}
+        ],
+    )
+    assert message.role == "assistant"
+    assert len(message.content) > 0, "Empty content"
+    text = message.content[0].text
+    assert len(text) > 0, "Empty text"
+    print(f"  PASS  Anthropic SDK: {text.strip()[:80]!r}")
+
+
+def test_anthropic_with_tools(base_url: str, api_key: str):
+    """Anthropic Messages API: streaming with tools."""
+    status, events = _stream_anthropic_http(
+        f"{base_url}/v1/messages",
+        body = {
+            "model": "default",
+            "max_tokens": 1024,
+            "messages": [
+                {
+                    "role": "user",
+                    "content": "What is 123 * 456? Use code to compute it.",
+                }
+            ],
+            "tools": [
+                {
+                    "name": "python",
+                    "description": "Execute Python code in a sandbox and return stdout/stderr.",
+                    "input_schema": {
+                        "type": "object",
+                        "properties": {
+                            "code": {
+                                "type": "string",
+                                "description": "The Python code to run",
+                            },
+                        },
+                        "required": ["code"],
+                    },
+                }
+            ],
+            "stream": True,
+        },
+        headers = {"Authorization": f"Bearer {api_key}"},
+        timeout = 120,
+    )
+    assert status == 200, f"Expected 200, got {status}"
+    assert len(events) > 0, "No SSE events received for tools request"
+
+    event_types = [e[0] for e in events]
+    assert "message_start" in event_types, "Missing message_start"
+    assert "message_stop" in event_types, "Missing message_stop"
+
+    full = _collect_anthropic_text(events)
+    print(
+        f"  PASS  anthropic with tools: {len(events)} events, {len(full)} chars content"
+    )
+
+
+# ── Server lifecycle ─────────────────────────────────────────────────
+
+
+def _start_server(model: str, variant: str | None) -> tuple[subprocess.Popen, str]:
+    """Launch ``unsloth studio run`` and parse the API key from its banner.
+
+    Returns (process, api_key).
+    """
+    cmd = [
+        "unsloth",
+        "studio",
+        "run",
+        "--model",
+        model,
+        "--port",
+        str(PORT),
+        "--host",
+        HOST,
+        "--api-key-name",
+        "test",
+    ]
+    if variant:
+        cmd.extend(["--gguf-variant", variant])
+
+    LOG_FILE.parent.mkdir(parents = True, exist_ok = True)
+    log_fh = open(LOG_FILE, "w")
+    proc = subprocess.Popen(
+        cmd,
+        stdout = log_fh,
+        stderr = subprocess.STDOUT,
+        preexec_fn = os.setsid,
+    )
+
+    # Wait for the banner containing the API key
+    api_key = None
+    deadline = time.monotonic() + STARTUP_TIMEOUT
+    while time.monotonic() < deadline:
+        time.sleep(2)
+        if proc.poll() is not None:
+            log_fh.flush()
+            log_text = LOG_FILE.read_text()
+            raise RuntimeError(
+                f"Server exited early (code {proc.returncode}):\n{log_text[-2000:]}"
+            )
+        log_text = LOG_FILE.read_text()
+        m = re.search(r"API Key:\s+(sk-unsloth-[a-f0-9]+)", log_text)
+        if m:
+            api_key = m.group(1)
+            break
+
+    if not api_key:
+        log_text = LOG_FILE.read_text()
+        _kill_server(proc)
+        raise RuntimeError(
+            f"Timed out waiting for API key in server output:\n{log_text[-2000:]}"
+        )
+
+    # Wait a moment for the model to be fully loaded
+    time.sleep(2)
+    return proc, api_key
+
+
+def _kill_server(proc: subprocess.Popen):
+    """Send SIGTERM to the process group and wait for cleanup."""
+    try:
+        os.killpg(os.getpgid(proc.pid), signal.SIGTERM)
+    except (ProcessLookupError, PermissionError):
+        pass
+    try:
+        proc.wait(timeout = 10)
+    except subprocess.TimeoutExpired:
+        try:
+            os.killpg(os.getpgid(proc.pid), signal.SIGKILL)
+        except (ProcessLookupError, PermissionError):
+            pass
+        proc.wait(timeout = 5)
+
+
+# ── Main ─────────────────────────────────────────────────────────────
+
+
+def main():
+    parser = argparse.ArgumentParser(
+        description = "End-to-end tests for unsloth studio run"
+    )
+    parser.add_argument(
+        "--model",
+        default = DEFAULT_MODEL,
+        help = f"Model to test with (default: {DEFAULT_MODEL})",
+    )
+    parser.add_argument(
+        "--gguf-variant",
+        default = DEFAULT_VARIANT,
+        help = f"GGUF variant (default: {DEFAULT_VARIANT})",
+    )
+    args = parser.parse_args()
+
+    passed = 0
+    failed = 0
+    skipped = 0
+
+    def run_test(fn, *a, **kw):
+        nonlocal passed, failed, skipped
+        try:
+            fn(*a, **kw)
+            passed += 1
+        except AssertionError as exc:
+            failed += 1
+            print(f"  FAIL  {fn.__name__}: {exc}")
+        except Exception as exc:
+            failed += 1
+            print(f"  ERROR {fn.__name__}: {type(exc).__name__}: {exc}")
+
+    # ── 1. Test --help (no server needed) ────────────────────────────
+    print("\n[1/11] Testing --help output")
+    run_test(test_help_output)
+
+    # ── 2-11. Start server and run API tests ─────────────────────────
+    print(
+        f"\nStarting server: {args.model} (variant={args.gguf_variant}) on port {PORT}..."
+    )
+    proc = None
+    try:
+        proc, api_key = _start_server(args.model, args.gguf_variant)
+        base_url = f"http://{HOST}:{PORT}"
+        print(f"Server ready.  API Key: {api_key[:20]}...\n")
+
+        print("[2/11] Testing curl basic (non-streaming)")
+        run_test(test_curl_basic, base_url, api_key)
+
+        print("[3/11] Testing curl streaming")
+        run_test(test_curl_streaming, base_url, api_key)
+
+        print("[4/11] Testing OpenAI Python SDK (streaming)")
+        run_test(test_openai_sdk, base_url, api_key)
+
+        print("[5/11] Testing curl with tools")
+        run_test(test_curl_with_tools, base_url, api_key)
+
+        print("[6/11] Testing invalid API key rejection")
+        run_test(test_invalid_key_rejected, base_url)
+
+        print("[7/11] Testing no API key rejection")
+        run_test(test_no_key_rejected, base_url)
+
+        print("[8/11] Testing Anthropic basic (non-streaming)")
+        run_test(test_anthropic_basic, base_url, api_key)
+
+        print("[9/11] Testing Anthropic streaming")
+        run_test(test_anthropic_streaming, base_url, api_key)
+
+        print("[10/11] Testing Anthropic Python SDK")
+        run_test(test_anthropic_sdk, base_url, api_key)
+
+        print("[11/11] Testing Anthropic with tools")
+        run_test(test_anthropic_with_tools, base_url, api_key)
+
+    except RuntimeError as exc:
+        print(f"\nFATAL: Server failed to start: {exc}")
+        failed += 11  # count remaining tests as failed
+    finally:
+        if proc:
+            print("\nStopping server...")
+            _kill_server(proc)
+            print("Server stopped.")
+
+    # ── Summary ──────────────────────────────────────────────────────
+    total = passed + failed
+    print(f"\n{'=' * 40}")
+    print(f"Results: {passed}/{total} passed, {failed} failed")
+    print(f"Log: {LOG_FILE}")
+    print(f"{'=' * 40}")
+    sys.exit(1 if failed else 0)
+
+
+if __name__ == "__main__":
+    main()
diff --git a/studio/frontend/src/app/router.tsx b/studio/frontend/src/app/router.tsx
index 13ff8a5cbe..d507929758 100644
--- a/studio/frontend/src/app/router.tsx
+++ b/studio/frontend/src/app/router.tsx
@@ -13,6 +13,7 @@ import { Route as loginRoute } from "./routes/login";
 import { Route as onboardingRoute } from "./routes/onboarding";
 import { Route as changePasswordRoute } from "./routes/change-password";
 import { Route as studioRoute } from "./routes/studio";
+import { Route as apiKeysRoute } from "./routes/api-keys";
 
 const routeTree = rootRoute.addChildren([
   indexRoute,
@@ -25,6 +26,7 @@ const routeTree = rootRoute.addChildren([
   exportRoute,
   dataRecipesRoute,
   dataRecipeRoute,
+  apiKeysRoute,
 ]);
 
 export const router = createRouter({ routeTree });
diff --git a/studio/frontend/src/app/routes/api-keys.tsx b/studio/frontend/src/app/routes/api-keys.tsx
new file mode 100644
index 0000000000..5846690d7b
--- /dev/null
+++ b/studio/frontend/src/app/routes/api-keys.tsx
@@ -0,0 +1,18 @@
+// SPDX-License-Identifier: AGPL-3.0-only
+// Copyright 2026-present the Unsloth AI Inc. team. All rights reserved. See /studio/LICENSE.AGPL-3.0
+
+import { createRoute } from "@tanstack/react-router";
+import { lazy } from "react";
+import { requireAuth } from "../auth-guards";
+import { Route as rootRoute } from "./__root";
+
+const ApiKeysPage = lazy(() =>
+  import("@/features/auth/api-keys-page").then((m) => ({ default: m.ApiKeysPage })),
+);
+
+export const Route = createRoute({
+  getParentRoute: () => rootRoute,
+  path: "/api-keys",
+  beforeLoad: () => requireAuth(),
+  component: ApiKeysPage,
+});
diff --git a/studio/frontend/src/components/navbar.tsx b/studio/frontend/src/components/navbar.tsx
index 121c559db8..2d310a81c3 100644
--- a/studio/frontend/src/components/navbar.tsx
+++ b/studio/frontend/src/components/navbar.tsx
@@ -29,6 +29,7 @@ import {
   ChefHatIcon,
   Copy01Icon,
   CursorInfo02Icon,
+  Key01Icon,
   PackageIcon,
   Tick02Icon,
   ZapIcon,
@@ -416,6 +417,20 @@ export function Navbar() {
               </HoverCardContent>
             </HoverCard>
           </div>
+          <div className="flex shrink-0 items-center">
+            <Link
+              to="/api-keys"
+              className={cn(
+                "flex h-9 items-center gap-1.5 rounded-md px-3 text-sm font-medium transition-colors hover:bg-accent",
+                pathname === "/api-keys"
+                  ? "text-foreground"
+                  : "text-muted-foreground hover:text-foreground",
+              )}
+            >
+              <HugeiconsIcon icon={Key01Icon} className="size-4" />
+              API Keys
+            </Link>
+          </div>
           {tourId ? (
             <div className="flex shrink-0 items-center">
               <button
@@ -529,11 +544,24 @@ export function Navbar() {
                     </Link>
                   );
                 })}
+                <Link
+                  to="/api-keys"
+                  onClick={() => setMobileOpen(false)}
+                  className={cn(
+                    "mt-3 flex items-center gap-2 rounded-md border px-3 py-2 text-sm font-medium",
+                    pathname === "/api-keys"
+                      ? "border-foreground bg-foreground text-background"
+                      : "border-border text-foreground hover:bg-accent",
+                  )}
+                >
+                  <HugeiconsIcon icon={Key01Icon} className="size-4" />
+                  API Keys
+                </Link>
                 <a
                   href="https://unsloth.ai/docs"
                   target="_blank"
                   rel="noopener noreferrer"
-                  className="mt-3 flex items-center gap-2 rounded-md border border-border px-3 py-2 text-sm font-medium text-foreground hover:bg-accent"
+                  className="flex items-center gap-2 rounded-md border border-border px-3 py-2 text-sm font-medium text-foreground hover:bg-accent"
                   onClick={() => setMobileOpen(false)}
                 >
                   <HugeiconsIcon icon={Book03Icon} className="size-4" />
diff --git a/studio/frontend/src/features/auth/api-keys-page.tsx b/studio/frontend/src/features/auth/api-keys-page.tsx
new file mode 100644
index 0000000000..5f76a91759
--- /dev/null
+++ b/studio/frontend/src/features/auth/api-keys-page.tsx
@@ -0,0 +1,417 @@
+// SPDX-License-Identifier: AGPL-3.0-only
+// Copyright 2026-present the Unsloth AI Inc. team. All rights reserved. See /studio/LICENSE.AGPL-3.0
+
+import { DashboardLayout } from "@/components/layout/dashboard-layout";
+import { Button } from "@/components/ui/button";
+import {
+  Dialog,
+  DialogContent,
+  DialogDescription,
+  DialogFooter,
+  DialogHeader,
+  DialogTitle,
+} from "@/components/ui/dialog";
+import { Input } from "@/components/ui/input";
+import { Label } from "@/components/ui/label";
+import { cn } from "@/lib/utils";
+import { copyToClipboard } from "@/lib/copy-to-clipboard";
+import {
+  AlertCircleIcon,
+  Copy01Icon,
+  Delete02Icon,
+  Key01Icon,
+  Tick02Icon,
+} from "@hugeicons/core-free-icons";
+import { HugeiconsIcon } from "@hugeicons/react";
+import { useCallback, useEffect, useRef, useState } from "react";
+import { authFetch } from "./api";
+
+// ---------------------------------------------------------------------------
+// Types
+// ---------------------------------------------------------------------------
+
+interface ApiKey {
+  id: number;
+  name: string;
+  key_prefix: string;
+  created_at: string;
+  last_used_at: string | null;
+  expires_at: string | null;
+  is_active: boolean;
+}
+
+// ---------------------------------------------------------------------------
+// API helpers
+// ---------------------------------------------------------------------------
+
+async function fetchApiKeys(): Promise<ApiKey[]> {
+  const res = await authFetch("/api/auth/api-keys");
+  if (!res.ok) throw new Error("Failed to load API keys");
+  const data = (await res.json()) as { api_keys: ApiKey[] };
+  return data.api_keys;
+}
+
+async function createApiKey(
+  name: string,
+  expiresInDays: number | null,
+): Promise<{ key: string; api_key: ApiKey }> {
+  const res = await authFetch("/api/auth/api-keys", {
+    method: "POST",
+    headers: { "Content-Type": "application/json" },
+    body: JSON.stringify({
+      name,
+      expires_in_days: expiresInDays,
+    }),
+  });
+  if (!res.ok) throw new Error("Failed to create API key");
+  return res.json();
+}
+
+async function revokeApiKey(keyId: number): Promise<void> {
+  const res = await authFetch(`/api/auth/api-keys/${keyId}`, {
+    method: "DELETE",
+  });
+  if (!res.ok) throw new Error("Failed to revoke API key");
+}
+
+// ---------------------------------------------------------------------------
+// Helpers
+// ---------------------------------------------------------------------------
+
+function formatDate(iso: string | null): string {
+  if (!iso) return "--";
+  const d = new Date(iso);
+  return d.toLocaleDateString(undefined, {
+    year: "numeric",
+    month: "short",
+    day: "numeric",
+    hour: "2-digit",
+    minute: "2-digit",
+  });
+}
+
+// ---------------------------------------------------------------------------
+// Components
+// ---------------------------------------------------------------------------
+
+function CopyButton({ text }: { text: string }) {
+  const [copied, setCopied] = useState(false);
+  const timerRef = useRef<ReturnType<typeof setTimeout> | null>(null);
+
+  useEffect(() => {
+    return () => {
+      if (timerRef.current) clearTimeout(timerRef.current);
+    };
+  }, []);
+
+  const handleCopy = () => {
+    if (!copyToClipboard(text)) return;
+    setCopied(true);
+    if (timerRef.current) clearTimeout(timerRef.current);
+    timerRef.current = setTimeout(() => setCopied(false), 2000);
+  };
+
+  return (
+    <Button variant="outline" size="sm" onClick={handleCopy} className="shrink-0">
+      <HugeiconsIcon
+        icon={copied ? Tick02Icon : Copy01Icon}
+        className={cn("size-3.5 mr-1.5", copied && "text-emerald-600")}
+      />
+      {copied ? "Copied" : "Copy"}
+    </Button>
+  );
+}
+
+function RevealKeyDialog({
+  open,
+  rawKey,
+  onClose,
+}: {
+  open: boolean;
+  rawKey: string;
+  onClose: () => void;
+}) {
+  return (
+    <Dialog open={open} onOpenChange={(o) => !o && onClose()}>
+      <DialogContent className="max-w-lg">
+        <DialogHeader>
+          <DialogTitle>API Key Created</DialogTitle>
+          <DialogDescription>
+            Copy this key now. It will not be shown again.
+          </DialogDescription>
+        </DialogHeader>
+        <div className="flex items-center gap-2 rounded-md border border-border bg-muted/40 p-3">
+          <code className="min-w-0 flex-1 break-all font-mono text-sm">
+            {rawKey}
+          </code>
+          <CopyButton text={rawKey} />
+        </div>
+        <div className="flex items-start gap-2 rounded-md border border-amber-500/20 bg-amber-50 p-3 text-amber-800 dark:border-amber-400/20 dark:bg-amber-950/30 dark:text-amber-300">
+          <HugeiconsIcon icon={AlertCircleIcon} className="mt-0.5 size-4 shrink-0" />
+          <p className="text-xs leading-relaxed">
+            Store this key securely. You will not be able to see it again after closing this dialog.
+          </p>
+        </div>
+        <DialogFooter>
+          <Button onClick={onClose}>Done</Button>
+        </DialogFooter>
+      </DialogContent>
+    </Dialog>
+  );
+}
+
+function CreateKeyForm({ onCreated }: { onCreated: (rawKey: string) => void }) {
+  const [name, setName] = useState("");
+  const [expiresInDays, setExpiresInDays] = useState("");
+  const [loading, setLoading] = useState(false);
+
+  const handleSubmit = async (e: React.FormEvent) => {
+    e.preventDefault();
+    if (!name.trim()) return;
+    setLoading(true);
+    try {
+      const days = expiresInDays ? parseInt(expiresInDays, 10) : null;
+      const result = await createApiKey(name.trim(), days);
+      onCreated(result.key);
+      setName("");
+      setExpiresInDays("");
+    } finally {
+      setLoading(false);
+    }
+  };
+
+  return (
+    <form onSubmit={handleSubmit} className="flex flex-col gap-4 rounded-lg border border-border p-4">
+      <div className="flex flex-col gap-1.5">
+        <Label htmlFor="key-name">Key name</Label>
+        <Input
+          id="key-name"
+          placeholder="e.g. My application"
+          value={name}
+          onChange={(e) => setName(e.target.value)}
+          required
+        />
+      </div>
+      <div className="flex flex-col gap-1.5">
+        <Label htmlFor="key-expiry">Expires in (days)</Label>
+        <Input
+          id="key-expiry"
+          type="number"
+          min={1}
+          placeholder="Leave blank for no expiry"
+          value={expiresInDays}
+          onChange={(e) => setExpiresInDays(e.target.value)}
+        />
+      </div>
+      <Button type="submit" disabled={loading || !name.trim()} className="self-start">
+        {loading ? "Creating..." : "Create API key"}
+      </Button>
+    </form>
+  );
+}
+
+function KeysTable({
+  keys,
+  onRevoke,
+}: {
+  keys: ApiKey[];
+  onRevoke: (id: number) => void;
+}) {
+  if (keys.length === 0) {
+    return (
+      <p className="py-8 text-center text-sm text-muted-foreground">
+        No API keys yet. Create one above.
+      </p>
+    );
+  }
+
+  return (
+    <div className="overflow-x-auto rounded-lg border border-border">
+      <table className="w-full text-sm">
+        <thead>
+          <tr className="border-b border-border bg-muted/40">
+            <th className="px-4 py-2.5 text-left font-medium text-muted-foreground">Name</th>
+            <th className="px-4 py-2.5 text-left font-medium text-muted-foreground">Key</th>
+            <th className="px-4 py-2.5 text-left font-medium text-muted-foreground">Created</th>
+            <th className="px-4 py-2.5 text-left font-medium text-muted-foreground">Last used</th>
+            <th className="px-4 py-2.5 text-left font-medium text-muted-foreground">Expires</th>
+            <th className="px-4 py-2.5 text-right font-medium text-muted-foreground" />
+          </tr>
+        </thead>
+        <tbody>
+          {keys.map((k) => (
+            <tr
+              key={k.id}
+              className={cn(
+                "border-b border-border last:border-b-0",
+                !k.is_active && "opacity-50",
+              )}
+            >
+              <td className="px-4 py-2.5 font-medium">{k.name}</td>
+              <td className="px-4 py-2.5">
+                <code className="rounded bg-muted px-1.5 py-0.5 font-mono text-xs">
+                  sk-unsloth-{k.key_prefix}...
+                </code>
+              </td>
+              <td className="px-4 py-2.5 text-muted-foreground">{formatDate(k.created_at)}</td>
+              <td className="px-4 py-2.5 text-muted-foreground">{formatDate(k.last_used_at)}</td>
+              <td className="px-4 py-2.5 text-muted-foreground">{formatDate(k.expires_at)}</td>
+              <td className="px-4 py-2.5 text-right">
+                {k.is_active ? (
+                  <Button
+                    variant="ghost"
+                    size="sm"
+                    onClick={() => onRevoke(k.id)}
+                    className="text-destructive hover:text-destructive"
+                  >
+                    <HugeiconsIcon icon={Delete02Icon} className="size-3.5 mr-1" />
+                    Revoke
+                  </Button>
+                ) : (
+                  <span className="text-xs text-muted-foreground">Revoked</span>
+                )}
+              </td>
+            </tr>
+          ))}
+        </tbody>
+      </table>
+    </div>
+  );
+}
+
+function UsageExamples() {
+  const base = window.location.origin;
+
+  const curlExample = `curl ${base}/v1/chat/completions \\
+  -H "Authorization: Bearer sk-unsloth-YOUR_KEY" \\
+  -H "Content-Type: application/json" \\
+  -d '{
+    "messages": [{"role": "user", "content": "Hello"}],
+    "stream": true
+  }'`;
+
+  const pythonExample = `from openai import OpenAI
+
+client = OpenAI(
+    base_url="${base}/v1",
+    api_key="sk-unsloth-YOUR_KEY",
+)
+
+response = client.chat.completions.create(
+    model="current",
+    messages=[{"role": "user", "content": "Hello"}],
+    stream=True,
+)
+for chunk in response:
+    print(chunk.choices[0].delta.content or "", end="")`;
+
+  const toolsExample = `curl ${base}/v1/chat/completions \\
+  -H "Authorization: Bearer sk-unsloth-YOUR_KEY" \\
+  -H "Content-Type: application/json" \\
+  -d '{
+    "messages": [{"role": "user", "content": "Search for Python 3.13 features"}],
+    "stream": true,
+    "enable_tools": true,
+    "enabled_tools": ["web_search", "python"],
+    "session_id": "my-session"
+  }'`;
+
+  return (
+    <div className="flex flex-col gap-4">
+      <h3 className="text-sm font-semibold">Usage examples</h3>
+      <div className="flex flex-col gap-3">
+        <div>
+          <p className="mb-1.5 text-xs font-medium text-muted-foreground">curl</p>
+          <pre className="overflow-x-auto rounded-md border border-border bg-muted/40 p-3 font-mono text-xs leading-relaxed">
+            {curlExample}
+          </pre>
+        </div>
+        <div>
+          <p className="mb-1.5 text-xs font-medium text-muted-foreground">Python (OpenAI SDK)</p>
+          <pre className="overflow-x-auto rounded-md border border-border bg-muted/40 p-3 font-mono text-xs leading-relaxed">
+            {pythonExample}
+          </pre>
+        </div>
+        <div>
+          <p className="mb-1.5 text-xs font-medium text-muted-foreground">With tools (web search + code execution)</p>
+          <pre className="overflow-x-auto rounded-md border border-border bg-muted/40 p-3 font-mono text-xs leading-relaxed">
+            {toolsExample}
+          </pre>
+        </div>
+      </div>
+    </div>
+  );
+}
+
+// ---------------------------------------------------------------------------
+// Page
+// ---------------------------------------------------------------------------
+
+export function ApiKeysPage() {
+  const [keys, setKeys] = useState<ApiKey[]>([]);
+  const [revealedKey, setRevealedKey] = useState<string | null>(null);
+  const [error, setError] = useState<string | null>(null);
+
+  const loadKeys = useCallback(async () => {
+    try {
+      setError(null);
+      const loaded = await fetchApiKeys();
+      setKeys(loaded);
+    } catch (e) {
+      setError(e instanceof Error ? e.message : "Failed to load API keys");
+    }
+  }, []);
+
+  useEffect(() => {
+    void loadKeys();
+  }, [loadKeys]);
+
+  const handleCreated = (rawKey: string) => {
+    setRevealedKey(rawKey);
+    void loadKeys();
+  };
+
+  const handleRevoke = async (keyId: number) => {
+    try {
+      await revokeApiKey(keyId);
+      void loadKeys();
+    } catch {
+      setError("Failed to revoke key");
+    }
+  };
+
+  return (
+    <DashboardLayout>
+      <div className="flex flex-col gap-8">
+        <div className="flex items-center gap-3">
+          <div className="flex h-10 w-10 items-center justify-center rounded-lg border border-border bg-muted/40">
+            <HugeiconsIcon icon={Key01Icon} className="size-5" />
+          </div>
+          <div>
+            <h1 className="text-xl font-bold font-heading">API Keys</h1>
+            <p className="text-sm text-muted-foreground">
+              Create keys to access Unsloth Studio programmatically via the OpenAI-compatible API.
+            </p>
+          </div>
+        </div>
+
+        {error && (
+          <div className="flex items-center gap-2 rounded-md border border-destructive/20 bg-destructive/5 p-3 text-sm text-destructive">
+            <HugeiconsIcon icon={AlertCircleIcon} className="size-4 shrink-0" />
+            {error}
+          </div>
+        )}
+
+        <CreateKeyForm onCreated={handleCreated} />
+        <KeysTable keys={keys} onRevoke={handleRevoke} />
+        <UsageExamples />
+      </div>
+
+      <RevealKeyDialog
+        open={revealedKey !== null}
+        rawKey={revealedKey ?? ""}
+        onClose={() => setRevealedKey(null)}
+      />
+    </DashboardLayout>
+  );
+}
diff --git a/studio/frontend/src/features/auth/index.ts b/studio/frontend/src/features/auth/index.ts
index 75db92432c..9962f6d431 100644
--- a/studio/frontend/src/features/auth/index.ts
+++ b/studio/frontend/src/features/auth/index.ts
@@ -1,6 +1,7 @@
 // SPDX-License-Identifier: AGPL-3.0-only
 // Copyright 2026-present the Unsloth AI Inc. team. All rights reserved. See /studio/LICENSE.AGPL-3.0
 
+export { ApiKeysPage } from "./api-keys-page";
 export { LoginPage } from "./login-page";
 export { ChangePasswordPage } from "./change-password-page";
 export { authFetch, refreshSession } from "./api";
diff --git a/unsloth_cli/commands/studio.py b/unsloth_cli/commands/studio.py
index 2fecb9d6b1..a3c0840be1 100644
--- a/unsloth_cli/commands/studio.py
+++ b/unsloth_cli/commands/studio.py
@@ -69,6 +69,80 @@ def _find_setup_script() -> Optional[Path]:
     return None
 
 
+# ── helpers for `unsloth studio run` ────────────────────────────────
+
+
+def _wait_for_server(port: int, timeout: int = 30) -> bool:
+    """Poll ``GET /api/health`` until the server responds 200 or *timeout* expires."""
+    import urllib.request
+    import urllib.error
+
+    url = f"http://127.0.0.1:{port}/api/health"
+    deadline = time.monotonic() + timeout
+    while time.monotonic() < deadline:
+        try:
+            with urllib.request.urlopen(url, timeout = 2) as resp:
+                if resp.status == 200:
+                    return True
+        except (urllib.error.URLError, OSError, ConnectionError):
+            pass
+        time.sleep(0.5)
+    return False
+
+
+def _create_api_key_inprocess(name: str) -> str:
+    """Create an API key via direct storage call (no HTTP needed).
+
+    Bypasses the ``must_change_password`` gate that blocks HTTP
+    ``POST /api/auth/api-keys`` on fresh installs.  Safe because the
+    CLI already has filesystem access to ``~/.unsloth/studio``.
+    """
+    from auth.storage import create_api_key, DEFAULT_ADMIN_USERNAME
+
+    raw_key, _row = create_api_key(username = DEFAULT_ADMIN_USERNAME, name = name)
+    return raw_key
+
+
+def _load_model_via_http(
+    port: int,
+    api_key: str,
+    model: str,
+    gguf_variant: Optional[str],
+    max_seq_length: int,
+    load_in_4bit: bool,
+    timeout: int = 600,
+) -> dict:
+    """POST to ``/api/inference/load`` using the API key for auth."""
+    import json
+    import urllib.request
+    import urllib.error
+
+    payload: dict = {
+        "model_path": model,
+        "max_seq_length": max_seq_length,
+        "load_in_4bit": load_in_4bit,
+    }
+    if gguf_variant:
+        payload["gguf_variant"] = gguf_variant
+
+    data = json.dumps(payload).encode()
+    req = urllib.request.Request(
+        f"http://127.0.0.1:{port}/api/inference/load",
+        data = data,
+        headers = {
+            "Content-Type": "application/json",
+            "Authorization": f"Bearer {api_key}",
+        },
+        method = "POST",
+    )
+    try:
+        with urllib.request.urlopen(req, timeout = timeout) as resp:
+            return json.loads(resp.read())
+    except urllib.error.HTTPError as exc:
+        body = exc.read().decode(errors = "replace")
+        raise RuntimeError(f"Model load failed (HTTP {exc.code}): {body}") from exc
+
+
 # ── unsloth studio (server) ──────────────────────────────────────────
 
 
@@ -166,6 +240,174 @@ def studio_default(
         typer.echo("\nShutting down...")
 
 
+# ── unsloth studio run ───────────────────────────────────────────────
+
+
+@studio_app.command()
+def run(
+    model: str = typer.Option(..., "--model", "-m", help = "Model path or HF repo"),
+    gguf_variant: Optional[str] = typer.Option(
+        None, "--gguf-variant", help = "GGUF quant variant (e.g. UD-Q4_K_XL)"
+    ),
+    max_seq_length: int = typer.Option(
+        0, "--max-seq-length", help = "Max sequence length (0 = model default)"
+    ),
+    load_in_4bit: bool = typer.Option(True, "--load-in-4bit/--no-load-in-4bit"),
+    api_key_name: str = typer.Option(
+        "cli", "--api-key-name", help = "Label for the auto-generated API key"
+    ),
+    port: int = typer.Option(8888, "--port", "-p"),
+    host: str = typer.Option("0.0.0.0", "--host", "-H"),
+    frontend: Optional[Path] = typer.Option(None, "--frontend", "-f"),
+    silent: bool = typer.Option(False, "--silent", "-q"),
+):
+    """Start Studio, load a model, and print an API key -- one-liner server.
+
+    Example:
+        unsloth studio run --model unsloth/Qwen3-1.7B-GGUF --gguf-variant UD-Q4_K_XL
+    """
+    # ── 1. Venv re-exec (same pattern as studio_default) ──────────────
+    studio_venv_dir = STUDIO_HOME / "unsloth_studio"
+    in_studio_venv = sys.prefix.startswith(str(studio_venv_dir))
+
+    if not in_studio_venv:
+        studio_python = _studio_venv_python()
+        if not studio_python:
+            typer.echo("Studio not set up. Run install.sh first.")
+            raise typer.Exit(1)
+        # Re-exec into the studio venv via its `unsloth` entry point
+        studio_bin = studio_python.parent / "unsloth"
+        if not studio_bin.is_file():
+            typer.echo(
+                "Studio venv missing 'unsloth' entry point. Re-run: unsloth studio setup"
+            )
+            raise typer.Exit(1)
+        args = [
+            str(studio_bin),
+            "studio",
+            "run",
+            "--model",
+            model,
+            "--max-seq-length",
+            str(max_seq_length),
+            "--api-key-name",
+            api_key_name,
+            "--port",
+            str(port),
+            "--host",
+            host,
+        ]
+        if gguf_variant:
+            args.extend(["--gguf-variant", gguf_variant])
+        if not load_in_4bit:
+            args.append("--no-load-in-4bit")
+        if frontend:
+            args.extend(["--frontend", str(frontend)])
+        if silent:
+            args.append("--silent")
+
+        if sys.platform == "win32":
+            proc = subprocess.Popen(args)
+            try:
+                rc = proc.wait()
+            except KeyboardInterrupt:
+                rc = proc.wait()
+            raise typer.Exit(rc)
+        else:
+            os.execvp(str(studio_bin), args)
+
+    # ── 2. Start server (always suppress built-in banner) ─────────────
+    from studio.backend.run import run_server, _resolve_external_ip
+
+    run_kwargs = dict(host = host, port = port, silent = True, llama_parallel_slots = 4)
+    if frontend is not None:
+        run_kwargs["frontend_path"] = frontend
+    app = run_server(**run_kwargs)
+    actual_port = getattr(app.state, "server_port", port) or port
+
+    # ── 3. Wait for server health ─────────────────────────────────────
+    if not silent:
+        typer.echo("Starting Unsloth Studio...")
+    if not _wait_for_server(actual_port):
+        typer.echo("Error: server did not become healthy within 30 seconds.", err = True)
+        raise typer.Exit(1)
+
+    # ── 4. Create API key in-process ──────────────────────────────────
+    api_key = _create_api_key_inprocess(api_key_name)
+
+    # ── 5. Load model via HTTP ────────────────────────────────────────
+    if not silent:
+        typer.echo(f"Loading model: {model}...")
+    try:
+        result = _load_model_via_http(
+            port = actual_port,
+            api_key = api_key,
+            model = model,
+            gguf_variant = gguf_variant,
+            max_seq_length = max_seq_length,
+            load_in_4bit = load_in_4bit,
+        )
+    except RuntimeError as exc:
+        typer.echo(f"Error: {exc}", err = True)
+        raise typer.Exit(1)
+
+    loaded_model = result.get("model", model)
+    display_variant = f" ({gguf_variant})" if gguf_variant else ""
+
+    # ── 6. Print banner ───────────────────────────────────────────────
+    display_host = _resolve_external_ip() if host == "0.0.0.0" else host
+    base_url = f"http://{display_host}:{actual_port}"
+    sdk_base_url = f"{base_url}/v1"
+
+    if not silent:
+        typer.echo("")
+        typer.echo("=" * 56)
+        typer.echo(f"  Unsloth Studio running at {base_url}")
+        typer.echo(f"  Model loaded: {loaded_model}{display_variant}")
+        typer.echo(f"  API Key:      {api_key}")
+        typer.echo("")
+        typer.echo("  OpenAI / Anthropic SDK base URL:")
+        typer.echo(f"    {sdk_base_url}")
+        typer.echo("=" * 56)
+        typer.echo("")
+        typer.echo("OpenAI Chat Completions:")
+        typer.echo(f"  curl {sdk_base_url}/chat/completions \\")
+        typer.echo(f'    -H "Authorization: Bearer {api_key}" \\')
+        typer.echo('    -H "Content-Type: application/json" \\')
+        typer.echo(
+            """    -d '{"messages": [{"role": "user", "content": "Hello"}], "stream": true}'"""
+        )
+        typer.echo("")
+        typer.echo("Anthropic Messages:")
+        typer.echo(f"  curl {sdk_base_url}/messages \\")
+        typer.echo(f'    -H "Authorization: Bearer {api_key}" \\')
+        typer.echo('    -H "Content-Type: application/json" \\')
+        typer.echo(
+            """    -d '{"max_tokens": 256, "messages": [{"role": "user", "content": "Hello"}], "stream": true}'"""
+        )
+        typer.echo("")
+        typer.echo("OpenAI Responses:")
+        typer.echo(f"  curl {sdk_base_url}/responses \\")
+        typer.echo(f'    -H "Authorization: Bearer {api_key}" \\')
+        typer.echo('    -H "Content-Type: application/json" \\')
+        typer.echo("""    -d '{"input": "Hello", "stream": true}'""")
+        typer.echo("")
+
+    # ── 7. Wait for Ctrl+C ────────────────────────────────────────────
+    from studio.backend.run import _shutdown_event, _graceful_shutdown, _server
+
+    try:
+        if _shutdown_event is not None:
+            while not _shutdown_event.is_set():
+                _shutdown_event.wait(timeout = 1)
+        else:
+            while True:
+                time.sleep(1)
+    except KeyboardInterrupt:
+        _graceful_shutdown(_server)
+        typer.echo("\nShutting down...")
+
+
 # ── unsloth studio stop ───────────────────────────────────────────────
 
 _PID_FILE = STUDIO_HOME / "studio.pid"

From d2fc58284002d12f385898bd5f7eb78f3e1b78d7 Mon Sep 17 00:00:00 2001
From: AdamPlatin123 <154926636+AdamPlatin123@users.noreply.github.com>
Date: Tue, 14 Apr 2026 03:02:12 +0800
Subject: [PATCH 079/216] studio: skip training status/metrics polling when
 idle (#4988)

* fix(studio): skip training status/metrics polling when idle

Add an early return in the status and metrics setInterval callbacks when
the runtime store reports phase === "idle" and hasHydrated is true.
Previously these polls fired unconditionally every 3s/5s, generating
unnecessary network traffic and console errors when no training was
running.

* fix(studio): reduce idle polling to 30s instead of stopping entirely

Review feedback (PR #4988): completely stopping polling when idle risks
permanent UI desync if hydration fails, and misses out-of-band state
changes from other clients. Add a 30s background poll that only fires
when idle to recover gracefully.

* fix: harden idle status polling around hydration and runtime reset

---------

Co-authored-by: AdamPlatin123 <AdamPlatin123@users.noreply.github.com>
Co-authored-by: Lee Jackson <130007945+Imagineer99@users.noreply.github.com>
Co-authored-by: imagineer99 <samleejackson0@gmail.com>
---
 .../hooks/use-training-runtime-lifecycle.ts   | 21 +++++++++++++++++--
 .../training/stores/training-runtime-store.ts |  1 +
 2 files changed, 20 insertions(+), 2 deletions(-)

diff --git a/studio/frontend/src/features/training/hooks/use-training-runtime-lifecycle.ts b/studio/frontend/src/features/training/hooks/use-training-runtime-lifecycle.ts
index 5965e07eaa..6e4c7c85b9 100644
--- a/studio/frontend/src/features/training/hooks/use-training-runtime-lifecycle.ts
+++ b/studio/frontend/src/features/training/hooks/use-training-runtime-lifecycle.ts
@@ -14,6 +14,7 @@ import type { TrainingRuntimeStore } from "../types/runtime";
 
 const STATUS_POLL_INTERVAL_MS = 3000;
 const METRICS_POLL_INTERVAL_MS = 5000;
+const IDLE_POLL_INTERVAL_MS = 30000;
 const STREAM_RECONNECT_DELAY_MS = 1500;
 
 function shouldUseLiveSync(state: TrainingRuntimeStore): boolean {
@@ -164,21 +165,37 @@ export function useTrainingRuntimeLifecycle(): void {
 
     void hydrate();
 
+    const isIdle = () => {
+      const s = runtimeStore.getState();
+      return s.phase === "idle" && !s.isTrainingRunning;
+    };
+
     const statusTimer = setInterval(() => {
+      const s = runtimeStore.getState();
+      if (isIdle() && s.hasHydrated) return;
       void pollStatus();
     }, STATUS_POLL_INTERVAL_MS);
 
     const metricsTimer = setInterval(() => {
-      const state = runtimeStore.getState();
-      if (shouldUseLiveSync(state) || state.currentStep > 0) {
+      if (isIdle()) return;
+      const s = runtimeStore.getState();
+      if (shouldUseLiveSync(s) || s.currentStep > 0) {
         void pollMetrics();
       }
     }, METRICS_POLL_INTERVAL_MS);
 
+    // Low-frequency background poll to recover from failed hydration or detect
+    // out-of-band state changes (e.g. training started from another client).
+    const idleTimer = setInterval(() => {
+      if (!isIdle()) return;
+      void pollStatus();
+    }, IDLE_POLL_INTERVAL_MS);
+
     return () => {
       disposed = true;
       clearInterval(statusTimer);
       clearInterval(metricsTimer);
+      clearInterval(idleTimer);
       stopStream();
     };
   }, []);
diff --git a/studio/frontend/src/features/training/stores/training-runtime-store.ts b/studio/frontend/src/features/training/stores/training-runtime-store.ts
index af6dfc5c85..a57c444c8d 100644
--- a/studio/frontend/src/features/training/stores/training-runtime-store.ts
+++ b/studio/frontend/src/features/training/stores/training-runtime-store.ts
@@ -125,6 +125,7 @@ export const useTrainingRuntimeStore = create<TrainingRuntimeStore>()((set) => (
   resetRuntime: () =>
     set((state) => ({
       ...initialState,
+      hasHydrated: state.hasHydrated,
       lossHistory: [],
       lrHistory: [],
       gradNormHistory: [],

From 97eafd999e4e82835afedfb87b56b87b2167c34d Mon Sep 17 00:00:00 2001
From: Wasim Yousef Said <wasimysdev@gmail.com>
Date: Mon, 13 Apr 2026 21:48:51 +0200
Subject: [PATCH 080/216] studio: fix api-keys access + refresh (#5005)

* studio: fix api-keys access + refresh

* studio: guard v1 in spa fallback
---
 studio/backend/main.py                    | 2 +-
 studio/frontend/src/app/routes/__root.tsx | 9 ++++++++-
 2 files changed, 9 insertions(+), 2 deletions(-)

diff --git a/studio/backend/main.py b/studio/backend/main.py
index c2c0a0b6e4..8a40791c06 100644
--- a/studio/backend/main.py
+++ b/studio/backend/main.py
@@ -353,7 +353,7 @@ def setup_frontend(app: FastAPI, build_path: Path):
 
     @app.get("/{full_path:path}")
     async def serve_frontend(full_path: str):
-        if full_path.startswith("api"):
+        if full_path in {"api", "v1"} or full_path.startswith(("api/", "v1/")):
             return {"error": "API endpoint not found"}
 
         file_path = (build_path / full_path).resolve()
diff --git a/studio/frontend/src/app/routes/__root.tsx b/studio/frontend/src/app/routes/__root.tsx
index bac9205303..e1bbdc03f7 100644
--- a/studio/frontend/src/app/routes/__root.tsx
+++ b/studio/frontend/src/app/routes/__root.tsx
@@ -13,7 +13,14 @@ import { AnimatePresence, motion } from "motion/react";
 import { Suspense } from "react";
 import { AppProvider } from "../provider";
 
-const CHAT_ONLY_ALLOWED = new Set(["/", "/chat", "/login", "/signup", "/change-password"]);
+const CHAT_ONLY_ALLOWED = new Set([
+  "/",
+  "/chat",
+  "/login",
+  "/signup",
+  "/change-password",
+  "/api-keys",
+]);
 
 function isChatOnlyAllowed(pathname: string): boolean {
   if (CHAT_ONLY_ALLOWED.has(pathname)) return true;

From bfa17330bdf157f30b7933aad85062a9a9a5fae6 Mon Sep 17 00:00:00 2001
From: Lee Jackson <130007945+Imagineer99@users.noreply.github.com>
Date: Tue, 14 Apr 2026 11:22:14 +0100
Subject: [PATCH 081/216] Studio: Polish API key copy button and harden async
 clipboard fallback (#5006)

* fix: polish clipboard style and fix async clipboard path

* Use copyToClipboardAsync in CopyButton for Safari fallback

CopyButton was calling navigator.clipboard.writeText directly,
bypassing the execCommand fallback added in this same PR. Switch
to copyToClipboardAsync which tries execCommand first (Safari
user-gesture requirement) then falls back to the async clipboard API.

* Fix copyToClipboard sync contract regression and improve async path

- Restore copyToClipboard() to return only the execCommand result,
  preserving the boolean contract that 7 existing callers depend on
  to gate their "Copied!" UI state. The fire-and-forget async fallback
  was returning true before the promise resolved, causing false success.

- Add document.body null guard to copyWithExecCommand for SSR safety.

- Reorder copyToClipboardAsync to try the async Clipboard API first,
  avoiding unnecessary DOM/focus overhead in Radix focus-trapped dialogs
  where execCommand always fails anyway.

* Restore queryCommandSupported guard and fix async catch path

- Restore the queryCommandSupported("copy") guard in copyToClipboard()
  to match the original contract exactly: when execCommand is entirely
  unsupported, fall through to fire-and-forget async clipboard write.

- Fix copyToClipboardAsync catch block: after navigator.clipboard.writeText
  rejects, the user-gesture frame is gone, so execCommand will also fail.
  Return false from catch instead of falling through. The execCommand
  fallback at the bottom only runs when the Clipboard API is absent
  (still in user-gesture frame).

* Restore execCommand fallback in copyToClipboardAsync catch path

The catch block was returning false after clipboard API rejection,
based on the incorrect premise that the user-gesture frame is lost
after an await. Per the HTML spec, transient user activation IS
preserved through promise microtask chains. The real reason
execCommand fails in the Radix dialog is the focus trap intercepting
textarea.focus(), not gesture loss.

For non-dialog callers, execCommand can still succeed after a
clipboard rejection. Inside a Radix modal, execCommand returns
false harmlessly (focus trap blocks it).

* Harden textarea fallback for mobile and continue to async path on failure

---------

Co-authored-by: Daniel Han <danielhanchen@gmail.com>
Co-authored-by: Roland Tannous <rolandtannous@gravityq.ai>
---
 .../src/features/auth/api-keys-page.tsx       | 21 +++--
 studio/frontend/src/lib/copy-to-clipboard.ts  | 80 +++++++++++++------
 2 files changed, 72 insertions(+), 29 deletions(-)

diff --git a/studio/frontend/src/features/auth/api-keys-page.tsx b/studio/frontend/src/features/auth/api-keys-page.tsx
index 5f76a91759..2065384a7d 100644
--- a/studio/frontend/src/features/auth/api-keys-page.tsx
+++ b/studio/frontend/src/features/auth/api-keys-page.tsx
@@ -14,7 +14,7 @@ import {
 import { Input } from "@/components/ui/input";
 import { Label } from "@/components/ui/label";
 import { cn } from "@/lib/utils";
-import { copyToClipboard } from "@/lib/copy-to-clipboard";
+import { copyToClipboardAsync } from "@/lib/copy-to-clipboard";
 import {
   AlertCircleIcon,
   Copy01Icon,
@@ -104,20 +104,29 @@ function CopyButton({ text }: { text: string }) {
     };
   }, []);
 
-  const handleCopy = () => {
-    if (!copyToClipboard(text)) return;
+  const handleCopy = async () => {
+    if (!(await copyToClipboardAsync(text))) return;
     setCopied(true);
     if (timerRef.current) clearTimeout(timerRef.current);
     timerRef.current = setTimeout(() => setCopied(false), 2000);
   };
 
   return (
-    <Button variant="outline" size="sm" onClick={handleCopy} className="shrink-0">
+    <Button
+      variant="ghost"
+      size="icon-sm"
+      onClick={handleCopy}
+      className={cn(
+        "shrink-0 rounded-md text-muted-foreground hover:text-foreground",
+        copied && "text-emerald-600 hover:text-emerald-600",
+      )}
+      aria-label={copied ? "Copied API key" : "Copy API key"}
+      title={copied ? "Copied" : "Copy"}
+    >
       <HugeiconsIcon
         icon={copied ? Tick02Icon : Copy01Icon}
-        className={cn("size-3.5 mr-1.5", copied && "text-emerald-600")}
+        className="size-4"
       />
-      {copied ? "Copied" : "Copy"}
     </Button>
   );
 }
diff --git a/studio/frontend/src/lib/copy-to-clipboard.ts b/studio/frontend/src/lib/copy-to-clipboard.ts
index df4eeec0a8..7c97904d73 100644
--- a/studio/frontend/src/lib/copy-to-clipboard.ts
+++ b/studio/frontend/src/lib/copy-to-clipboard.ts
@@ -6,42 +6,76 @@
  * Uses a synchronous textarea + execCommand fallback so the copy runs in the
  * same user gesture as the click (required by Safari's clipboard security).
  */
+function copyWithExecCommand(text: string): boolean {
+  if (typeof document === "undefined" || !document.body) return false;
+
+  const textarea = document.createElement("textarea");
+  textarea.value = text;
+  textarea.readOnly = true;
+  textarea.style.position = "fixed";
+  textarea.style.top = "0";
+  textarea.style.left = "0";
+  textarea.style.fontSize = "12pt";
+  textarea.style.opacity = "0";
+  textarea.setAttribute("aria-hidden", "true");
+  document.body.appendChild(textarea);
+  textarea.focus({ preventScroll: true });
+  textarea.select();
+
+  try {
+    const ok = document.execCommand("copy");
+    document.body.removeChild(textarea);
+    return ok;
+  } catch {
+    document.body.removeChild(textarea);
+    return false;
+  }
+}
+
 export function copyToClipboard(text: string): boolean {
   if (typeof text !== "string" || text.length === 0) {
     return false;
   }
 
-  // Synchronous fallback: works in Safari/Mac when clipboard API fails
-  // because it runs entirely within the user gesture (click) stack.
-  if (document.queryCommandSupported?.("copy") !== false) {
-    const textarea = document.createElement("textarea");
-    textarea.value = text;
-    textarea.style.position = "fixed";
-    textarea.style.top = "0";
-    textarea.style.left = "0";
-    textarea.style.opacity = "0";
-    textarea.setAttribute("aria-hidden", "true");
-    document.body.appendChild(textarea);
-    textarea.focus({ preventScroll: true });
-    textarea.select();
-    try {
-      const ok = document.execCommand("copy");
-      document.body.removeChild(textarea);
-      return ok;
-    } catch {
-      document.body.removeChild(textarea);
-      return false;
-    }
+  if (typeof document !== "undefined" && document.queryCommandSupported?.("copy") !== false) {
+    if (copyWithExecCommand(text)) return true;
   }
 
-  // Modern API only when fallback not available (e.g. non-browser)
+  // Async fallback for environments where execCommand is entirely unsupported
+  // but the Clipboard API is available (rare; kept for original contract parity).
   if (typeof navigator?.clipboard?.writeText === "function") {
     navigator.clipboard.writeText(text).then(
       () => {},
-      () => {}
+      () => {},
     );
     return true;
   }
 
   return false;
 }
+
+export async function copyToClipboardAsync(text: string): Promise<boolean> {
+  if (typeof text !== "string" || text.length === 0) {
+    return false;
+  }
+
+  // Prefer the async Clipboard API: avoids focus disruption in Radix
+  // focus-trapped dialogs where execCommand always fails.
+  if (typeof navigator?.clipboard?.writeText === "function") {
+    try {
+      await navigator.clipboard.writeText(text);
+      return true;
+    } catch {
+      // Clipboard API rejected (e.g. NotAllowedError, permission policy).
+      // User activation is still valid through promise chains per spec, so
+      // execCommand can succeed for callers outside focus-trapped dialogs.
+      // Inside a Radix modal the focus trap will block textarea.focus() and
+      // execCommand returns false harmlessly.
+      return copyWithExecCommand(text);
+    }
+  }
+
+  // No Clipboard API (older browser / non-secure context): still in the
+  // original user-gesture frame, so execCommand can work.
+  return copyWithExecCommand(text);
+}

From a50f61009b130b85ff4d46995493019da5a139c6 Mon Sep 17 00:00:00 2001
From: Bharath Kumar Adinarayan <30743936+Barath19@users.noreply.github.com>
Date: Tue, 14 Apr 2026 12:29:27 +0200
Subject: [PATCH 082/216] fix(studio): default chart view to full training
 history (#5007)

* fix(studio): default chart view to full training history instead of last 80 steps

Fixes #5003

* chore: windowsize as null code comment

---------

Co-authored-by: imagineer99 <samleejackson0@gmail.com>
Co-authored-by: Wasim Yousef Said <wasimysdev@gmail.com>
---
 .../studio/sections/charts/chart-preferences-store.ts    | 9 ++-------
 1 file changed, 2 insertions(+), 7 deletions(-)

diff --git a/studio/frontend/src/features/studio/sections/charts/chart-preferences-store.ts b/studio/frontend/src/features/studio/sections/charts/chart-preferences-store.ts
index 083cfe1333..7f24192f83 100644
--- a/studio/frontend/src/features/studio/sections/charts/chart-preferences-store.ts
+++ b/studio/frontend/src/features/studio/sections/charts/chart-preferences-store.ts
@@ -3,12 +3,7 @@
 
 import { create } from "zustand";
 import type { OutlierMode, ScaleMode } from "./types";
-import { DEFAULT_VISIBLE_POINTS, clamp } from "./utils";
-
-const DEFAULT_WINDOW_SIZE = Math.max(
-  24,
-  Math.floor(DEFAULT_VISIBLE_POINTS / 2),
-);
+import { clamp } from "./utils";
 
 type ChartPreferencesState = {
   availableSteps: number;
@@ -39,7 +34,7 @@ type ChartPreferencesState = {
 };
 
 const defaultPreferences = {
-  windowSize: DEFAULT_WINDOW_SIZE as number | null,
+  windowSize: null as number | null, // null = show full training history
   smoothing: 0.6,
   showRaw: true,
   showSmoothed: true,

From dccc0ebada6cb9dd510d7c9803017d8c58042e3f Mon Sep 17 00:00:00 2001
From: Datta Nimmaturi <venkatadattasainimmaturi@gmail.com>
Date: Tue, 14 Apr 2026 16:33:58 +0530
Subject: [PATCH 083/216] [Studio] Show non exported models in chat UI (#4892)

* Show non exported models in chat UI

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

* Distinguish b/w LoRa and full fine tune saves. Cleanup

---------

Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com>
Co-authored-by: Roland Tannous <115670425+rolandtannous@users.noreply.github.com>
---
 studio/backend/core/__init__.py               |   7 +-
 studio/backend/routes/models.py               |  15 +-
 .../backend/tests/test_trained_model_scan.py  | 101 +++++++++++
 studio/backend/utils/models/__init__.py       |   7 +-
 studio/backend/utils/models/model_config.py   | 165 +++++++++++++++---
 .../assistant-ui/model-selector.tsx           |  17 +-
 .../assistant-ui/model-selector/pickers.tsx   |  10 +-
 .../frontend/src/features/chat/chat-page.tsx  |  15 +-
 .../chat/hooks/use-chat-model-runtime.ts      |  24 ++-
 .../frontend/src/features/chat/tour/steps.tsx |   8 +-
 10 files changed, 314 insertions(+), 55 deletions(-)
 create mode 100644 studio/backend/tests/test_trained_model_scan.py

diff --git a/studio/backend/core/__init__.py b/studio/backend/core/__init__.py
index d8d95e2f1a..d39815c437 100644
--- a/studio/backend/core/__init__.py
+++ b/studio/backend/core/__init__.py
@@ -31,6 +31,7 @@ __all__ = [
     # Config
     "ModelConfig",
     "is_vision_model",
+    "scan_trained_models",
     "scan_trained_loras",
     "load_model_defaults",
     "get_base_model_from_lora",
@@ -72,6 +73,7 @@ def __getattr__(name):
     if name in (
         "is_vision_model",
         "ModelConfig",
+        "scan_trained_models",
         "scan_trained_loras",
         "load_model_defaults",
         "get_base_model_from_lora",
@@ -79,14 +81,15 @@ def __getattr__(name):
         from utils.models import (
             is_vision_model,
             ModelConfig,
-            scan_trained_loras,
+            scan_trained_models,
             load_model_defaults,
             get_base_model_from_lora,
         )
 
         globals()["is_vision_model"] = is_vision_model
         globals()["ModelConfig"] = ModelConfig
-        globals()["scan_trained_loras"] = scan_trained_loras
+        globals()["scan_trained_models"] = scan_trained_models
+        globals()["scan_trained_loras"] = scan_trained_models
         globals()["load_model_defaults"] = load_model_defaults
         globals()["get_base_model_from_lora"] = get_base_model_from_lora
         return globals()[name]
diff --git a/studio/backend/routes/models.py b/studio/backend/routes/models.py
index 3f361ca5eb..cd8606f23f 100644
--- a/studio/backend/routes/models.py
+++ b/studio/backend/routes/models.py
@@ -32,8 +32,9 @@ from auth.authentication import get_current_subject
 # Import backend functions
 try:
     from utils.models import (
-        scan_trained_loras,
+        scan_trained_models,
         scan_exported_models,
+        get_base_model_from_checkpoint,
         load_model_defaults,
         get_base_model_from_lora,
         is_vision_model,
@@ -62,8 +63,9 @@ except ImportError:
     if str(parent_backend) not in sys.path:
         sys.path.insert(0, str(parent_backend))
     from utils.models import (
-        scan_trained_loras,
+        scan_trained_models,
         scan_exported_models,
+        get_base_model_from_checkpoint,
         load_model_defaults,
         get_base_model_from_lora,
         is_vision_model,
@@ -791,15 +793,16 @@ async def scan_loras(
         lora_list = []
 
         # Scan training outputs
-        trained_loras = scan_trained_loras(outputs_dir = resolved_outputs_dir)
-        for display_name, adapter_path in trained_loras:
-            base_model = get_base_model_from_lora(adapter_path)
+        trained_models = scan_trained_models(outputs_dir = resolved_outputs_dir)
+        for display_name, model_path, model_type in trained_models:
+            base_model = get_base_model_from_checkpoint(model_path)
             lora_list.append(
                 LoRAInfo(
                     display_name = display_name,
-                    adapter_path = adapter_path,
+                    adapter_path = model_path,
                     base_model = base_model,
                     source = "training",
+                    export_type = model_type,
                 )
             )
 
diff --git a/studio/backend/tests/test_trained_model_scan.py b/studio/backend/tests/test_trained_model_scan.py
new file mode 100644
index 0000000000..84be681fca
--- /dev/null
+++ b/studio/backend/tests/test_trained_model_scan.py
@@ -0,0 +1,101 @@
+# SPDX-License-Identifier: AGPL-3.0-only
+# Copyright 2026-present the Unsloth AI Inc. team. All rights reserved. See /studio/LICENSE.AGPL-3.0
+
+"""Tests for Studio trained-model discovery used by Chat."""
+
+import json
+from pathlib import Path
+import sys
+import types as _types
+import importlib
+
+
+_BACKEND_DIR = str(Path(__file__).resolve().parent.parent)
+if _BACKEND_DIR not in sys.path:
+    sys.path.insert(0, _BACKEND_DIR)
+
+_loggers_stub = _types.ModuleType("loggers")
+_loggers_stub.get_logger = lambda name: __import__("logging").getLogger(name)
+sys.modules.setdefault("loggers", _loggers_stub)
+
+from unittest.mock import patch
+
+from utils.models.model_config import (
+    ModelConfig,
+    get_base_model_from_checkpoint,
+    get_base_model_from_lora,
+    scan_trained_models,
+)
+
+
+def test_scan_trained_models_includes_lora_and_full_finetune_outputs(tmp_path: Path):
+    lora_dir = tmp_path / "unsloth_SmolLM-135M_1775412608"
+    lora_dir.mkdir()
+    (lora_dir / "adapter_config.json").write_text(
+        json.dumps({"base_model_name_or_path": "HuggingFaceTB/SmolLM-135M"})
+    )
+    (lora_dir / "adapter_model.safetensors").write_bytes(b"")
+
+    full_dir = tmp_path / "unsloth_SmolLM-135M_full_1775412609"
+    full_dir.mkdir()
+    (full_dir / "config.json").write_text(
+        json.dumps({"_name_or_path": "HuggingFaceTB/SmolLM-135M"})
+    )
+    (full_dir / "model.safetensors").write_bytes(b"")
+
+    found = {
+        name: (path, model_type)
+        for name, path, model_type in scan_trained_models(str(tmp_path))
+    }
+
+    assert found[lora_dir.name] == (str(lora_dir), "lora")
+    assert found[full_dir.name] == (str(full_dir), "merged")
+
+
+def test_get_base_model_from_checkpoint_falls_back_to_full_finetune_config(
+    tmp_path: Path,
+):
+    (tmp_path / "config.json").write_text(
+        json.dumps({"_name_or_path": "HuggingFaceTB/SmolLM-135M"})
+    )
+    (tmp_path / "model.safetensors").write_bytes(b"")
+
+    assert get_base_model_from_checkpoint(str(tmp_path)) == "HuggingFaceTB/SmolLM-135M"
+
+
+def test_get_base_model_from_lora_rejects_full_finetune_dirs(tmp_path: Path):
+    (tmp_path / "config.json").write_text(
+        json.dumps({"_name_or_path": "HuggingFaceTB/SmolLM-135M"})
+    )
+    (tmp_path / "model.safetensors").write_bytes(b"")
+
+    assert get_base_model_from_lora(str(tmp_path)) is None
+
+
+@patch("utils.models.model_config.is_audio_input_type", return_value = False)
+@patch("utils.models.model_config.detect_audio_type", return_value = None)
+@patch("utils.models.model_config.is_vision_model", return_value = False)
+def test_model_config_full_finetune_local_path_is_not_lora(
+    _mock_vision,
+    _mock_audio_type,
+    _mock_audio_input,
+    tmp_path: Path,
+):
+    (tmp_path / "config.json").write_text(
+        json.dumps({"_name_or_path": "unsloth/Qwen3-4B"})
+    )
+    (tmp_path / "model.safetensors").write_bytes(b"")
+
+    config = ModelConfig.from_identifier(str(tmp_path))
+
+    assert config is not None
+    assert config.is_lora is False
+    assert config.base_model is None
+
+
+def test_scan_trained_loras_aliases_scan_trained_models():
+    utils_models = importlib.import_module("utils.models")
+    core_module = importlib.import_module("core")
+
+    assert utils_models.scan_trained_loras is utils_models.scan_trained_models
+    assert core_module.scan_trained_loras is core_module.scan_trained_models
diff --git a/studio/backend/utils/models/__init__.py b/studio/backend/utils/models/__init__.py
index a81682d8b7..808e2b012e 100644
--- a/studio/backend/utils/models/__init__.py
+++ b/studio/backend/utils/models/__init__.py
@@ -13,8 +13,9 @@ from .model_config import (
     detect_audio_type,
     is_audio_input_type,
     VALID_AUDIO_TYPES,
-    scan_trained_loras,
+    scan_trained_models,
     scan_exported_models,
+    get_base_model_from_checkpoint,
     load_model_defaults,
     get_base_model_from_lora,
     load_model_config,
@@ -25,6 +26,8 @@ from .model_config import (
 )
 from .checkpoints import scan_checkpoints
 
+scan_trained_loras = scan_trained_models
+
 __all__ = [
     "ModelConfig",
     "GgufVariantInfo",
@@ -33,8 +36,10 @@ __all__ = [
     "detect_audio_type",
     "is_audio_input_type",
     "VALID_AUDIO_TYPES",
+    "scan_trained_models",
     "scan_trained_loras",
     "scan_exported_models",
+    "get_base_model_from_checkpoint",
     "load_model_defaults",
     "get_base_model_from_lora",
     "load_model_config",
diff --git a/studio/backend/utils/models/model_config.py b/studio/backend/utils/models/model_config.py
index 5be0b183d5..81331c9dde 100644
--- a/studio/backend/utils/models/model_config.py
+++ b/studio/backend/utils/models/model_config.py
@@ -1322,46 +1322,89 @@ def is_embedding_model(model_name: str, hf_token: Optional[str] = None) -> bool:
         return False
 
 
-def scan_trained_loras(outputs_dir: str = str(outputs_root())) -> List[Tuple[str, str]]:
+def _has_model_weight_files(model_dir: Path) -> bool:
+    """Return True when a directory contains loadable model weights."""
+    for item in model_dir.iterdir():
+        if not item.is_file():
+            continue
+
+        suffix = item.suffix.lower()
+        if suffix == ".safetensors":
+            return True
+        if suffix == ".gguf":
+            return "mmproj" not in item.name.lower()
+        if suffix == ".bin":
+            name = item.name.lower()
+            if (
+                name.startswith("pytorch_model")
+                or name.startswith("model")
+                or name.startswith("adapter_model")
+                or name.startswith("consolidated")
+            ):
+                return True
+    return False
+
+
+def _detect_training_output_type(model_dir: Path) -> Optional[str]:
+    """Classify a Studio training output as LoRA or full finetune."""
+    adapter_config = model_dir / "adapter_config.json"
+    adapter_model = model_dir / "adapter_model.safetensors"
+    if adapter_config.exists() or adapter_model.exists():
+        return "lora"
+
+    config_file = model_dir / "config.json"
+    if config_file.exists() and _has_model_weight_files(model_dir):
+        return "merged"
+
+    return None
+
+
+def _looks_like_lora_adapter(model_dir: Path) -> bool:
+    return model_dir.is_dir() and (
+        (model_dir / "adapter_config.json").exists()
+        or any(model_dir.glob("adapter_model*.safetensors"))
+        or any(model_dir.glob("adapter_model*.bin"))
+    )
+
+
+def scan_trained_models(
+    outputs_dir: str = str(outputs_root()),
+) -> List[Tuple[str, str, str]]:
     """
-    Scan outputs folder for trained LoRA adapters.
+    Scan outputs folder for trained Studio models.
 
     Returns:
-        List of tuples: [(display_name, adapter_path), ...]
-
-    Example:
-        [
-            ("unsloth_Meta-Llama-3.1_...", "./outputs/unsloth_Meta-Llama-3.1_.../"),
-            ("my_finetuned_model", "./outputs/my_finetuned_model/"),
-        ]
+        List of tuples: [(display_name, model_path, model_type), ...]
+        model_type is "lora" for adapter runs and "merged" for full finetunes.
     """
-    trained_loras = []
+    trained_models = []
     outputs_path = resolve_output_dir(outputs_dir)
 
     if not outputs_path.exists():
         logger.warning(f"Outputs directory not found: {outputs_dir}")
-        return trained_loras
+        return trained_models
 
     try:
         for item in outputs_path.iterdir():
             if item.is_dir():
-                # Check if this directory contains a LoRA adapter
-                adapter_config = item / "adapter_config.json"
-                adapter_model = item / "adapter_model.safetensors"
+                model_type = _detect_training_output_type(item)
+                if model_type is None:
+                    continue
 
-                if adapter_config.exists() or adapter_model.exists():
-                    display_name = item.name
-                    adapter_path = str(item)
-                    trained_loras.append((display_name, adapter_path))
-                    logger.debug(f"Found trained LoRA: {display_name}")
+                display_name = item.name
+                model_path = str(item)
+                trained_models.append((display_name, model_path, model_type))
+                logger.debug("Found trained model: %s (%s)", display_name, model_type)
 
         # Sort by modification time (newest first)
-        trained_loras.sort(key = lambda x: Path(x[1]).stat().st_mtime, reverse = True)
+        trained_models.sort(key = lambda x: Path(x[1]).stat().st_mtime, reverse = True)
 
         logger.info(
-            f"Found {len(trained_loras)} trained LoRA adapters in {outputs_dir}"
+            "Found %s trained models in %s",
+            len(trained_models),
+            outputs_dir,
         )
-        return trained_loras
+        return trained_models
 
     except Exception as e:
         logger.error(f"Error scanning outputs folder: {e}")
@@ -1494,6 +1537,68 @@ def scan_exported_models(
         return []
 
 
+def get_base_model_from_checkpoint(checkpoint_path: str) -> Optional[str]:
+    """Read the base model name from a local training or checkpoint directory."""
+    try:
+        checkpoint_path_obj = Path(checkpoint_path)
+
+        adapter_config_path = checkpoint_path_obj / "adapter_config.json"
+        if adapter_config_path.exists():
+            with open(adapter_config_path, "r") as f:
+                config = json.load(f)
+                base_model = config.get("base_model_name_or_path")
+                if base_model:
+                    logger.info(
+                        "Detected base model from adapter_config.json: %s", base_model
+                    )
+                    return base_model
+
+        config_path = checkpoint_path_obj / "config.json"
+        if config_path.exists():
+            with open(config_path, "r") as f:
+                config = json.load(f)
+                for key in ("model_name", "_name_or_path"):
+                    base_model = config.get(key)
+                    if base_model and str(base_model) != str(checkpoint_path_obj):
+                        logger.info(
+                            "Detected base model from config.json (%s): %s",
+                            key,
+                            base_model,
+                        )
+                        return base_model
+
+        training_args_path = checkpoint_path_obj / "training_args.bin"
+        if training_args_path.exists():
+            try:
+                import torch
+
+                training_args = torch.load(training_args_path)
+                if hasattr(training_args, "model_name_or_path"):
+                    base_model = training_args.model_name_or_path
+                    logger.info(
+                        "Detected base model from training_args.bin: %s", base_model
+                    )
+                    return base_model
+            except Exception as e:
+                logger.warning(f"Could not load training_args.bin: {e}")
+
+        dir_name = checkpoint_path_obj.name
+        if dir_name.startswith("unsloth_"):
+            parts = dir_name.split("_")
+            if len(parts) >= 2:
+                model_parts = parts[1:-1]
+                base_model = "unsloth/" + "_".join(model_parts)
+                logger.info("Detected base model from directory name: %s", base_model)
+                return base_model
+
+        logger.warning(f"Could not detect base model for checkpoint: {checkpoint_path}")
+        return None
+
+    except Exception as e:
+        logger.error(f"Error reading base model from checkpoint config: {e}")
+        return None
+
+
 def get_base_model_from_lora(lora_path: str) -> Optional[str]:
     """
     Read the base model name from a LoRA adapter's config.
@@ -1502,16 +1607,14 @@ def get_base_model_from_lora(lora_path: str) -> Optional[str]:
         lora_path: Path to the LoRA adapter directory
 
     Returns:
-        Base model identifier (e.g., "unsloth/Meta-Llama-3.1-8B-Instruct-bnb-4bit")
-        or None if not found
-
-    Example:
-        >>> get_base_model_from_lora("./outputs/unsloth_Meta-Llama-3.1_.../")
-        "unsloth/Meta-Llama-3.1-8B-Instruct-bnb-4bit"
+        Base model identifier or None if not found
     """
     try:
         lora_path_obj = Path(lora_path)
 
+        if not _looks_like_lora_adapter(lora_path_obj):
+            return None
+
         # Try adapter_config.json first
         adapter_config_path = lora_path_obj / "adapter_config.json"
         if adapter_config_path.exists():
@@ -1884,7 +1987,11 @@ class ModelConfig:
 
         # Auto-detect LoRA for local paths (check adapter_config.json on disk)
         if not is_lora and is_local:
-            detected_base = get_base_model_from_lora(path)
+            detected_base = (
+                get_base_model_from_lora(path)
+                if _looks_like_lora_adapter(Path(path))
+                else None
+            )
             if detected_base:
                 is_lora = True
                 logger.info(
diff --git a/studio/frontend/src/components/assistant-ui/model-selector.tsx b/studio/frontend/src/components/assistant-ui/model-selector.tsx
index 441c2b48e4..08e69bbf93 100644
--- a/studio/frontend/src/components/assistant-ui/model-selector.tsx
+++ b/studio/frontend/src/components/assistant-ui/model-selector.tsx
@@ -203,11 +203,22 @@ export function ModelSelector({
         ? lora.name.split("/")[0].trim()
         : lora.name;
       // Show type tag instead of base model name
+      const isLocal = lora.source === "local";
+      const isTraining = lora.source === "training";
       const isExported = lora.source === "exported";
       const isMerged = lora.exportType === "merged";
-      const tag = isExported
-        ? isMerged ? "Merged · Exported" : "LoRA"
-        : "LoRA";
+      const isGguf = lora.exportType === "gguf";
+      const tag = isLocal
+        ? isGguf
+          ? "GGUF"
+          : "Local"
+        : isTraining && isMerged
+          ? "Full finetune"
+          : isExported
+            ? isMerged
+              ? "Merged · Exported"
+              : "LoRA · Exported"
+            : "LoRA";
       all.set(lora.id, {
         ...lora,
         name: displayName,
diff --git a/studio/frontend/src/components/assistant-ui/model-selector/pickers.tsx b/studio/frontend/src/components/assistant-ui/model-selector/pickers.tsx
index 87d59dfaa3..e410f8cff9 100644
--- a/studio/frontend/src/components/assistant-ui/model-selector/pickers.tsx
+++ b/studio/frontend/src/components/assistant-ui/model-selector/pickers.tsx
@@ -1330,7 +1330,7 @@ export function LoraModelPicker({
         <Input
           value={query}
           onChange={(event) => setQuery(event.target.value)}
-          placeholder="Search local adapters"
+          placeholder="Search trained models"
           className="h-9 pl-8"
         />
       </div>
@@ -1339,7 +1339,7 @@ export function LoraModelPicker({
         <div className="p-1">
           {grouped.length === 0 ? (
             <div className="px-2.5 py-2 text-xs text-muted-foreground">
-              No adapters found.
+              No trained models found.
             </div>
           ) : (
             grouped.map(([baseModel, adapters], index) => (
@@ -1348,9 +1348,11 @@ export function LoraModelPicker({
                 <ListLabel>{baseModel}</ListLabel>
                 {adapters.map((adapter) => {
                   const isLocal = adapter.source === "local";
+                  const isTraining = adapter.source === "training";
                   const isExported = adapter.source === "exported";
                   const isMerged = adapter.exportType === "merged";
                   const isGguf = adapter.exportType === "gguf";
+                  const isTrainingFull = isTraining && isMerged;
                   const isLocalGgufDir =
                     isLocal &&
                     (isGgufRepo(adapter.id) || isGgufRepo(adapter.name));
@@ -1360,6 +1362,8 @@ export function LoraModelPicker({
                       : "Local"
                     : isGguf
                       ? "GGUF"
+                      : isTrainingFull
+                        ? "Full"
                       : isExported
                         ? isMerged
                           ? "Merged"
@@ -1369,6 +1373,8 @@ export function LoraModelPicker({
                     ? isLocalGgufDir
                       ? "GGUF"
                       : "Local"
+                    : isTrainingFull
+                      ? "Full finetune"
                     : isExported
                       ? `${tag} · Exported`
                       : tag;
diff --git a/studio/frontend/src/features/chat/chat-page.tsx b/studio/frontend/src/features/chat/chat-page.tsx
index ebaab75c0f..165eba0f2c 100644
--- a/studio/frontend/src/features/chat/chat-page.tsx
+++ b/studio/frontend/src/features/chat/chat-page.tsx
@@ -71,6 +71,7 @@ type LoraCandidate = {
   id: string;
   baseModel: string;
   updatedAt?: number;
+  exportType?: "lora" | "merged" | "gguf";
 };
 
 function normalizeModelRef(value: string | null | undefined): string {
@@ -81,12 +82,13 @@ function pickBestLoraForBase(
   loras: LoraCandidate[],
   baseModel: string | null,
 ): LoraCandidate | null {
-  if (loras.length === 0) return null;
-  const sorted = [...loras].sort(
+  const adapterOnly = loras.filter((lora) => lora.exportType === "lora");
+  if (adapterOnly.length === 0) return null;
+  const sorted = [...adapterOnly].sort(
     (a, b) => (b.updatedAt ?? -1) - (a.updatedAt ?? -1),
   );
   const normalizedBase = normalizeModelRef(baseModel);
-  if (!normalizedBase) return sorted[0];
+  if (!normalizedBase) return sorted[0] ?? null;
 
   const exact = sorted.find(
     (lora) => normalizeModelRef(lora.baseModel) === normalizedBase,
@@ -101,7 +103,7 @@ function pickBestLoraForBase(
       normalizedBase.includes(normalizedLoraBase)
     );
   });
-  return partial ?? sorted[0];
+  return partial ?? sorted[0] ?? null;
 }
 
 function messageHasImage(message: MessageRecord): boolean {
@@ -154,7 +156,8 @@ type CompareModelSelection = {
 function useIsLoraCompare(): boolean {
   return useChatRuntimeStore((s) => {
     const cp = s.params.checkpoint;
-    return cp ? s.loras.some((l) => l.id === cp) : false;
+    const selected = cp ? s.loras.find((l) => l.id === cp) : undefined;
+    return selected?.exportType === "lora";
   });
 }
 
@@ -235,7 +238,7 @@ const LoraCompareContent = memo(function LoraCompareContent({
           <div className="flex min-h-0 flex-col border-t border-border/60 md:border-t-0 md:border-l">
             <div className="px-3 py-1.5 text-start md:text-end">
               <span className="text-[10px] font-semibold uppercase tracking-wider text-primary">
-                Fine-tuned (LoRA)
+                Fine-tuned
               </span>
             </div>
             <div className="min-h-0 flex-1">
diff --git a/studio/frontend/src/features/chat/hooks/use-chat-model-runtime.ts b/studio/frontend/src/features/chat/hooks/use-chat-model-runtime.ts
index e5c1493872..65b02be965 100644
--- a/studio/frontend/src/features/chat/hooks/use-chat-model-runtime.ts
+++ b/studio/frontend/src/features/chat/hooks/use-chat-model-runtime.ts
@@ -236,6 +236,25 @@ export function useChatModelRuntime() {
         // Apply inference defaults on reconnect (page refresh with model already loaded)
         if (statusRes.inference) {
           const currentParams = useChatRuntimeStore.getState().params;
+          const reconnectResponse: LoadModelResponse = {
+            status: "already_loaded",
+            model: statusRes.active_model,
+            display_name: statusRes.active_model,
+            is_vision: statusRes.is_vision,
+            is_lora: false,
+            is_gguf: statusRes.is_gguf,
+            is_audio: statusRes.is_audio,
+            audio_type: statusRes.audio_type,
+            has_audio_input: statusRes.has_audio_input,
+            inference: statusRes.inference,
+            context_length: statusRes.context_length,
+            max_context_length: statusRes.max_context_length,
+            native_context_length: statusRes.native_context_length,
+            supports_reasoning: statusRes.supports_reasoning,
+            reasoning_always_on: statusRes.reasoning_always_on,
+            supports_tools: statusRes.supports_tools,
+            speculative_type: statusRes.speculative_type,
+          };
           setParams(
             mergeRecommendedInference(currentParams, statusRes, statusRes.active_model),
           );
@@ -340,8 +359,9 @@ export function useChatModelRuntime() {
         typeof selection === "string" ? false : selection.isDownloaded ?? false;
       const model = models.find((entry) => entry.id === modelId);
       const lora = loras.find((entry) => entry.id === modelId);
+      const loraIsAdapter = lora?.exportType === "lora";
       const isLora =
-        explicitIsLora ?? model?.isLora ?? (lora ? true : false);
+        explicitIsLora ?? model?.isLora ?? loraIsAdapter ?? false;
       const displayName = model?.name || lora?.name || modelId;
       const currentCheckpoint =
         useChatRuntimeStore.getState().params.checkpoint;
@@ -355,7 +375,7 @@ export function useChatModelRuntime() {
         ? loras.find((entry) => entry.id === previousCheckpoint)
         : undefined;
       const previousIsLora =
-        previousModel?.isLora ?? (previousLora ? true : false);
+        previousModel?.isLora ?? (previousLora?.exportType === "lora");
       // Covers Unix absolute (/), relative (./  ../), tilde (~/), Windows drive (C:\), UNC (\\server)
       const isLocal = /^(\/|\.{1,2}[\\/]|~[\\/]|[A-Za-z]:[\\/]|\\\\)/.test(modelId);
       const isCachedLora = isLora && isLocal;
diff --git a/studio/frontend/src/features/chat/tour/steps.tsx b/studio/frontend/src/features/chat/tour/steps.tsx
index 3c43d9fed4..e222fdc8a0 100644
--- a/studio/frontend/src/features/chat/tour/steps.tsx
+++ b/studio/frontend/src/features/chat/tour/steps.tsx
@@ -30,7 +30,7 @@ export function buildChatTourSteps({
       body: (
         <>
           This selects what’s loaded for inference. Hub = base models. Fine-tuned
-          = your LoRA adapters from Studio.
+          = trained Studio outputs, including LoRA adapters and full finetunes.
         </>
       ),
     },
@@ -40,9 +40,9 @@ export function buildChatTourSteps({
       title: "Two tabs",
       body: (
         <>
-          Hub: search Hugging Face models. Fine-tuned: adapters (LoRA) you’ve
-          trained locally. If results look off, compare base vs LoRA to see what
-          changed.
+          Hub: search Hugging Face models. Fine-tuned: local Studio outputs you’ve
+          trained or exported. If results look off, compare base vs fine-tuned
+          outputs to see what changed.
         </>
       ),
       onEnter: openModelSelector,

From da78c6be71720f40087faa6382a271388609eb5b Mon Sep 17 00:00:00 2001
From: Datta Nimmaturi <venkatadattasainimmaturi@gmail.com>
Date: Tue, 14 Apr 2026 18:10:17 +0530
Subject: [PATCH 084/216] [Studio] Install flash attn at setup time for linux
 (#4979)

* [Studio] Install flash attn at setup time for linux

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

* cleanup changes

Signed-off-by: Datta Nimmaturi <venkatadattasainimmaturi@gmail.com>

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

* Test cases

* wheel_utils: narrow url_exists exceptions and log at debug level

---------

Signed-off-by: Datta Nimmaturi <venkatadattasainimmaturi@gmail.com>
Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com>
Co-authored-by: Roland Tannous <115670425+rolandtannous@users.noreply.github.com>
Co-authored-by: Roland Tannous <rolandtannous@gravityq.ai>
---
 studio/backend/core/training/worker.py        | 334 +++++++-----------
 .../tests/test_training_worker_flash_attn.py  | 154 ++++++++
 studio/backend/utils/wheel_utils.py           | 170 +++++++++
 studio/install_python_stack.py                |  77 +++-
 .../test_flash_attn_install_python_stack.py   | 282 +++++++++++++++
 tests/python/test_no_torch_filtering.py       |   5 +-
 6 files changed, 815 insertions(+), 207 deletions(-)
 create mode 100644 studio/backend/tests/test_training_worker_flash_attn.py
 create mode 100644 studio/backend/utils/wheel_utils.py
 create mode 100644 tests/python/test_flash_attn_install_python_stack.py

diff --git a/studio/backend/core/training/worker.py b/studio/backend/core/training/worker.py
index c3d3daf07e..8ab2b5b2be 100644
--- a/studio/backend/core/training/worker.py
+++ b/studio/backend/core/training/worker.py
@@ -16,26 +16,31 @@ from __future__ import annotations
 import structlog
 from loggers import get_logger
 import os
-import platform
 import shutil
 import sys
 import time
 import traceback
-import json
 import subprocess as _sp
 from pathlib import Path
-from typing import Any
-import urllib.error
-import urllib.request
+from typing import Any, Callable
 
 logger = get_logger(__name__)
 from utils.hardware import apply_gpu_ids
+from utils.wheel_utils import (
+    direct_wheel_url,
+    flash_attn_wheel_url,
+    install_wheel,
+    probe_torch_wheel_env,
+    url_exists,
+)
 
 
 _CAUSAL_CONV1D_RELEASE_TAG = "v1.6.1.post4"
 _CAUSAL_CONV1D_PACKAGE_VERSION = "1.6.1"
 _MAMBA_SSM_RELEASE_TAG = "v2.3.1"
 _MAMBA_SSM_PACKAGE_VERSION = "2.3.1"
+_FLASH_ATTN_RUNTIME_MIN_SEQ_LEN = 32768
+_FLASH_ATTN_SKIP_ENV = "UNSLOTH_STUDIO_SKIP_FLASHATTN_INSTALL"
 
 
 def _model_wants_causal_conv1d(model_name: str) -> bool:
@@ -59,184 +64,60 @@ def _model_wants_causal_conv1d(model_name: str) -> bool:
     )
 
 
-def _causal_conv1d_platform_tag() -> str | None:
-    machine = platform.machine().lower()
-    if sys.platform.startswith("linux"):
-        if machine in {"x86_64", "amd64"}:
-            return "linux_x86_64"
-        if machine in {"aarch64", "arm64"}:
-            return "linux_aarch64"
-        return None
-    # No prebuilt wheels published for macOS or Windows
-    return None
-
-
-def _probe_causal_conv1d_env() -> dict[str, str] | None:
-    try:
-        probe = _sp.run(
-            [
-                sys.executable,
-                "-c",
-                (
-                    "import json, sys, re, torch; "
-                    "parts = torch.__version__.split('+', 1)[0].split('.')[:2]; "
-                    "minor = re.sub(r'[^0-9].*', '', parts[1]) if len(parts) > 1 else '0'; "
-                    "torch_mm = parts[0] + '.' + minor; "
-                    "print(json.dumps({"
-                    "'python_tag': f'cp{sys.version_info.major}{sys.version_info.minor}', "
-                    "'torch_mm': torch_mm, "
-                    "'cuda_major': str(int(str(torch.version.cuda).split('.', 1)[0])) if torch.version.cuda else '', "
-                    "'hip_version': str(torch.version.hip) if getattr(torch.version, 'hip', None) else '', "
-                    "'cxx11abi': str(torch._C._GLIBCXX_USE_CXX11_ABI).upper()"
-                    "}))"
-                ),
-            ],
-            stdout = _sp.PIPE,
-            stderr = _sp.PIPE,
-            text = True,
-            timeout = 30,
-        )
-    except _sp.TimeoutExpired:
-        logger.warning("Torch environment probe timed out after 30s")
-        return None
-    if probe.returncode != 0:
-        logger.warning(
-            "Failed to probe torch environment for causal-conv1d wheel:\n%s",
-            probe.stdout,
-        )
-        return None
-
-    try:
-        return json.loads(probe.stdout.strip())
-    except json.JSONDecodeError:
-        logger.warning(
-            "Failed to parse torch environment probe output: %s", probe.stdout
-        )
-        return None
-
-
-def _direct_wheel_url(
-    *,
-    filename_prefix: str,
-    package_version: str,
-    release_tag: str,
-    release_base_url: str,
-    env: dict[str, str] | None = None,
-) -> str | None:
-    env = env or _probe_causal_conv1d_env()
-    platform_tag = _causal_conv1d_platform_tag()
-    if env is None or platform_tag is None or not env.get("cuda_major"):
-        return None
-
-    filename = (
-        f"{filename_prefix}-{package_version}"
-        f"+cu{env['cuda_major']}torch{env['torch_mm']}"
-        f"cxx11abi{env['cxx11abi']}-{env['python_tag']}-{env['python_tag']}-{platform_tag}.whl"
-    )
-    return f"{release_base_url}/{release_tag}/{filename}"
-
-
-def _url_exists(url: str) -> bool:
-    try:
-        request = urllib.request.Request(url, method = "HEAD")
-        with urllib.request.urlopen(request, timeout = 10):
-            return True
-    except urllib.error.HTTPError as exc:
-        if exc.code == 404:
-            return False
-        logger.warning("Unexpected HTTP error while probing %s: %s", url, exc)
-        return False
-    except Exception as exc:
-        logger.warning("Failed to probe %s: %s", url, exc)
-        return False
-
-
 def _install_package_wheel_first(
     *,
     event_queue: Any,
     import_name: str,
     display_name: str,
     pypi_name: str,
-    pypi_version: str,
-    filename_prefix: str,
-    release_tag: str,
-    release_base_url: str,
-) -> None:
+    pypi_version: str | None = None,
+    filename_prefix: str | None = None,
+    release_tag: str | None = None,
+    release_base_url: str | None = None,
+    wheel_url_builder: Callable[[dict[str, str] | None], str | None] | None = None,
+    pypi_spec: str | None = None,
+    pypi_status_message: str | None = None,
+) -> bool:
     try:
         __import__(import_name)
         logger.info("%s already installed", display_name)
-        return
+        return True
     except ImportError:
         pass
 
-    env = _probe_causal_conv1d_env()
-    wheel_url = _direct_wheel_url(
-        filename_prefix = filename_prefix,
-        package_version = pypi_version,
-        release_tag = release_tag,
-        release_base_url = release_base_url,
-        env = env,
-    )
+    env = probe_torch_wheel_env(timeout = 30)
+    if wheel_url_builder is not None:
+        wheel_url = wheel_url_builder(env)
+    else:
+        wheel_url = direct_wheel_url(
+            filename_prefix = filename_prefix,
+            package_version = pypi_version,
+            release_tag = release_tag,
+            release_base_url = release_base_url,
+            env = env,
+        )
 
     if wheel_url is None:
         logger.info("No compatible %s wheel candidate", display_name)
-    else:
-        if _url_exists(wheel_url):
-            _send_status(event_queue, f"Installing prebuilt {display_name} wheel...")
-            installed = False
-            # Try uv first if available, then fall back to pip
-            if shutil.which("uv"):
-                uv_cmd = [
-                    "uv",
-                    "pip",
-                    "install",
-                    "--python",
-                    sys.executable,
-                    "--no-deps",
-                    wheel_url,
-                ]
-                result = _sp.run(
-                    uv_cmd,
-                    stdout = _sp.PIPE,
-                    stderr = _sp.STDOUT,
-                    text = True,
-                )
-                if result.returncode == 0:
-                    installed = True
-                else:
-                    logger.warning(
-                        "uv failed to install %s wheel:\n%s",
-                        display_name,
-                        result.stdout,
-                    )
-            if not installed:
-                pip_cmd = [
-                    sys.executable,
-                    "-m",
-                    "pip",
-                    "install",
-                    "--no-deps",
-                    wheel_url,
-                ]
-                result = _sp.run(
-                    pip_cmd,
-                    stdout = _sp.PIPE,
-                    stderr = _sp.STDOUT,
-                    text = True,
-                )
-                if result.returncode == 0:
-                    installed = True
-                else:
-                    logger.warning(
-                        "pip failed to install %s wheel:\n%s",
-                        display_name,
-                        result.stdout,
-                    )
-            if installed:
+    elif url_exists(wheel_url):
+        _send_status(event_queue, f"Installing prebuilt {display_name} wheel...")
+        for installer, result in install_wheel(
+            wheel_url,
+            python_executable = sys.executable,
+            use_uv = bool(shutil.which("uv")),
+            run = _sp.run,
+        ):
+            if result.returncode == 0:
                 logger.info("Installed prebuilt %s wheel successfully", display_name)
-                return
-        else:
-            logger.info("No published %s wheel found: %s", display_name, wheel_url)
+                return True
+            logger.warning(
+                "%s failed to install %s wheel:\n%s",
+                installer,
+                display_name,
+                result.stdout,
+            )
+    else:
+        logger.info("No published %s wheel found: %s", display_name, wheel_url)
 
     is_hip = env and env.get("hip_version")
     if is_hip and not shutil.which("hipcc"):
@@ -249,43 +130,62 @@ def _install_package_wheel_first(
             event_queue,
             f"{display_name}: hipcc not found (ROCm HIP SDK required)",
         )
-        return
+        return False
 
-    if is_hip:
-        _send_status(
-            event_queue,
-            f"Compiling {display_name} from source for ROCm "
-            "(this may take several minutes)...",
-        )
-    else:
-        _send_status(event_queue, f"Installing {display_name} from PyPI...")
+    if pypi_spec is None:
+        pypi_spec = f"{pypi_name}=={pypi_version}"
+
+    if pypi_status_message is None:
+        if is_hip:
+            pypi_status_message = (
+                f"Compiling {display_name} from source for ROCm "
+                "(this may take several minutes)..."
+            )
+        else:
+            pypi_status_message = f"Installing {display_name} from PyPI..."
+
+    _send_status(event_queue, pypi_status_message)
 
     # Prefer uv for faster dependency resolution when available
-    if shutil.which("uv"):
-        pypi_cmd = [
-            "uv",
-            "pip",
-            "install",
-            "--python",
-            sys.executable,
-            "--no-build-isolation",
-            "--no-deps",
-        ]
-        # Avoid stale cache artifacts from partial HIP source builds
-        if is_hip:
-            pypi_cmd.append("--no-cache")
-        pypi_cmd.append(f"{pypi_name}=={pypi_version}")
+    plain_pypi_install = pypi_version is None
+    if plain_pypi_install:
+        if shutil.which("uv"):
+            pypi_cmd = [
+                "uv",
+                "pip",
+                "install",
+                "--python",
+                sys.executable,
+                pypi_spec,
+            ]
+        else:
+            pypi_cmd = [sys.executable, "-m", "pip", "install", pypi_spec]
     else:
-        pypi_cmd = [
-            sys.executable,
-            "-m",
-            "pip",
-            "install",
-            "--no-build-isolation",
-            "--no-deps",
-            "--no-cache-dir",
-            f"{pypi_name}=={pypi_version}",
-        ]
+        if shutil.which("uv"):
+            pypi_cmd = [
+                "uv",
+                "pip",
+                "install",
+                "--python",
+                sys.executable,
+                "--no-build-isolation",
+                "--no-deps",
+            ]
+            # Avoid stale cache artifacts from partial HIP source builds
+            if is_hip:
+                pypi_cmd.append("--no-cache")
+            pypi_cmd.append(pypi_spec)
+        else:
+            pypi_cmd = [
+                sys.executable,
+                "-m",
+                "pip",
+                "install",
+                "--no-build-isolation",
+                "--no-deps",
+                "--no-cache-dir",
+                pypi_spec,
+            ]
 
     # Source compilation on ROCm can take 10-30 minutes; use a generous
     # timeout. Non-HIP installs preserve the pre-existing "no timeout"
@@ -313,7 +213,7 @@ def _install_package_wheel_first(
             f"{display_name} installation timed out after "
             f"{_run_kwargs.get('timeout')}s",
         )
-        return
+        return False
 
     if result.returncode != 0:
         if is_hip:
@@ -337,12 +237,13 @@ def _install_package_wheel_first(
                 display_name,
                 result.stdout,
             )
-        return
+        return False
 
     if is_hip:
         logger.info("Compiled and installed %s from source for ROCm", display_name)
     else:
         logger.info("Installed %s from PyPI", display_name)
+    return True
 
 
 def _ensure_causal_conv1d_fast_path(event_queue: Any, model_name: str) -> None:
@@ -389,6 +290,31 @@ def _ensure_mamba_ssm(event_queue: Any, model_name: str) -> None:
     )
 
 
+def _should_try_runtime_flash_attn_install(max_seq_length: int) -> bool:
+    if os.getenv(_FLASH_ATTN_SKIP_ENV) == "1":
+        return False
+    if max_seq_length < _FLASH_ATTN_RUNTIME_MIN_SEQ_LEN:
+        return False
+    return sys.platform.startswith("linux")
+
+
+def _ensure_flash_attn_for_long_context(event_queue: Any, max_seq_length: int) -> None:
+    if not _should_try_runtime_flash_attn_install(max_seq_length):
+        return
+
+    installed = _install_package_wheel_first(
+        event_queue = event_queue,
+        import_name = "flash_attn",
+        display_name = "flash-attn",
+        pypi_name = "flash-attn",
+        wheel_url_builder = flash_attn_wheel_url,
+        pypi_spec = "flash-attn",
+        pypi_status_message = "Installing flash-attn from PyPI for long-context training...",
+    )
+    if not installed:
+        _send_status(event_queue, "Continuing without flash-attn")
+
+
 def _activate_transformers_version(model_name: str) -> None:
     """Activate the correct transformers version BEFORE any ML imports."""
     # Ensure backend is on path for utils imports
@@ -471,6 +397,10 @@ def run_training_process(
     try:
         _ensure_causal_conv1d_fast_path(event_queue, model_name)
         _ensure_mamba_ssm(event_queue, model_name)
+        _ensure_flash_attn_for_long_context(
+            event_queue,
+            int(config.get("max_seq_length", 2048)),
+        )
     except Exception as exc:
         event_queue.put(
             {
diff --git a/studio/backend/tests/test_training_worker_flash_attn.py b/studio/backend/tests/test_training_worker_flash_attn.py
new file mode 100644
index 0000000000..986958408e
--- /dev/null
+++ b/studio/backend/tests/test_training_worker_flash_attn.py
@@ -0,0 +1,154 @@
+# SPDX-License-Identifier: AGPL-3.0-only
+# Copyright 2026-present the Unsloth AI Inc. team. All rights reserved. See /studio/LICENSE.AGPL-3.0
+
+from __future__ import annotations
+
+import builtins
+import subprocess
+import sys
+from unittest import mock
+
+from core.training import worker
+
+
+def _missing_flash_attn_import():
+    real_import = builtins.__import__
+
+    def fake_import(name, globals = None, locals = None, fromlist = (), level = 0):
+        if name == "flash_attn":
+            raise ImportError
+        return real_import(name, globals, locals, fromlist, level)
+
+    return fake_import
+
+
+def test_should_try_runtime_flash_attn_install_threshold_and_skip(monkeypatch):
+    monkeypatch.delenv(worker._FLASH_ATTN_SKIP_ENV, raising = False)
+    assert worker._should_try_runtime_flash_attn_install(32767) is False
+    assert worker._should_try_runtime_flash_attn_install(
+        32768
+    ) is sys.platform.startswith("linux")
+
+    monkeypatch.setenv(worker._FLASH_ATTN_SKIP_ENV, "1")
+    assert worker._should_try_runtime_flash_attn_install(32768) is False
+
+
+def test_runtime_flash_attn_prefers_prebuilt_wheel(monkeypatch):
+    statuses: list[str] = []
+
+    monkeypatch.delenv(worker._FLASH_ATTN_SKIP_ENV, raising = False)
+    monkeypatch.setattr(builtins, "__import__", _missing_flash_attn_import())
+    monkeypatch.setattr(
+        worker,
+        "flash_attn_wheel_url",
+        lambda env: "https://example.com/fa.whl",
+    )
+    monkeypatch.setattr(worker, "url_exists", lambda url: True)
+    monkeypatch.setattr(
+        worker,
+        "_send_status",
+        lambda queue, message: statuses.append(message),
+    )
+    monkeypatch.setattr(
+        worker,
+        "install_wheel",
+        lambda *args, **kwargs: [("pip", subprocess.CompletedProcess(["pip"], 0, ""))],
+    )
+
+    worker._ensure_flash_attn_for_long_context(event_queue = [], max_seq_length = 32768)
+
+    assert statuses == ["Installing prebuilt flash-attn wheel..."]
+
+
+def test_runtime_flash_attn_falls_back_to_pypi(monkeypatch):
+    calls: list[list[str]] = []
+    statuses: list[str] = []
+
+    monkeypatch.delenv(worker._FLASH_ATTN_SKIP_ENV, raising = False)
+    monkeypatch.setattr(builtins, "__import__", _missing_flash_attn_import())
+    monkeypatch.setattr(
+        worker,
+        "probe_torch_wheel_env",
+        lambda timeout = 30: {
+            "python_tag": "cp313",
+            "torch_mm": "2.10",
+            "cuda_major": "13",
+            "cxx11abi": "TRUE",
+            "platform_tag": "linux_x86_64",
+        },
+    )
+    monkeypatch.setattr(
+        worker,
+        "flash_attn_wheel_url",
+        lambda env: "https://example.com/fa.whl",
+    )
+    monkeypatch.setattr(worker, "url_exists", lambda url: False)
+    monkeypatch.setattr(worker.shutil, "which", lambda name: None)
+    monkeypatch.setattr(
+        worker,
+        "_send_status",
+        lambda queue, message: statuses.append(message),
+    )
+    monkeypatch.setattr(worker, "install_wheel", mock.Mock())
+
+    def fake_run(cmd, stdout = None, stderr = None, text = None):
+        calls.append(list(cmd))
+        return subprocess.CompletedProcess(cmd, 0, "")
+
+    monkeypatch.setattr(worker._sp, "run", fake_run)
+
+    worker._ensure_flash_attn_for_long_context(event_queue = [], max_seq_length = 32768)
+
+    assert statuses == ["Installing flash-attn from PyPI for long-context training..."]
+    assert calls == [[sys.executable, "-m", "pip", "install", "flash-attn"]]
+
+
+def test_runtime_flash_attn_skip_env_avoids_all_install_work(monkeypatch):
+    monkeypatch.setenv(worker._FLASH_ATTN_SKIP_ENV, "1")
+    monkeypatch.setattr(worker._sp, "run", mock.Mock())
+
+    worker._ensure_flash_attn_for_long_context(event_queue = [], max_seq_length = 32768)
+
+    worker._sp.run.assert_not_called()
+
+
+def test_causal_conv1d_fast_path_preserves_wheel_first_install_args(monkeypatch):
+    install_mock = mock.Mock(return_value = True)
+    monkeypatch.setattr(worker, "_install_package_wheel_first", install_mock)
+
+    worker._ensure_causal_conv1d_fast_path(
+        event_queue = [],
+        model_name = "tiiuae/Falcon-H1-0.5B-Instruct",
+    )
+
+    install_mock.assert_called_once_with(
+        event_queue = [],
+        import_name = "causal_conv1d",
+        display_name = "causal-conv1d",
+        pypi_name = "causal-conv1d",
+        pypi_version = worker._CAUSAL_CONV1D_PACKAGE_VERSION,
+        filename_prefix = "causal_conv1d",
+        release_tag = worker._CAUSAL_CONV1D_RELEASE_TAG,
+        release_base_url = "https://github.com/Dao-AILab/causal-conv1d/releases/download",
+    )
+
+
+def test_mamba_ssm_path_preserves_wheel_first_install_args(monkeypatch):
+    install_mock = mock.Mock(return_value = True)
+    monkeypatch.setattr(worker, "_install_package_wheel_first", install_mock)
+
+    worker._ensure_mamba_ssm(
+        event_queue = [],
+        model_name = "tiiuae/Falcon-H1-0.5B-Instruct",
+    )
+
+    install_mock.assert_called_once_with(
+        event_queue = [],
+        import_name = "mamba_ssm",
+        display_name = "mamba-ssm",
+        pypi_name = "mamba-ssm",
+        pypi_version = worker._MAMBA_SSM_PACKAGE_VERSION,
+        filename_prefix = "mamba_ssm",
+        release_tag = worker._MAMBA_SSM_RELEASE_TAG,
+        release_base_url = "https://github.com/state-spaces/mamba/releases/download",
+    )
diff --git a/studio/backend/utils/wheel_utils.py b/studio/backend/utils/wheel_utils.py
new file mode 100644
index 0000000000..00240f1e69
--- /dev/null
+++ b/studio/backend/utils/wheel_utils.py
@@ -0,0 +1,170 @@
+# SPDX-License-Identifier: AGPL-3.0-only
+# Copyright 2026-present the Unsloth AI Inc. team. All rights reserved. See /studio/LICENSE.AGPL-3.0
+
+from __future__ import annotations
+
+import json
+import logging
+import platform
+import shutil
+import subprocess
+import sys
+import urllib.error
+import urllib.request
+from typing import Callable
+
+_logger = logging.getLogger(__name__)
+
+FLASH_ATTN_RELEASE_BASE_URL = (
+    "https://github.com/Dao-AILab/flash-attention/releases/download"
+)
+
+
+def linux_wheel_platform_tag() -> str | None:
+    machine = platform.machine().lower()
+    if sys.platform.startswith("linux"):
+        if machine in {"x86_64", "amd64"}:
+            return "linux_x86_64"
+        if machine in {"aarch64", "arm64"}:
+            return "linux_aarch64"
+    # No prebuilt wheels published for macOS or Windows
+    return None
+
+
+def probe_torch_wheel_env(*, timeout: int | None = None) -> dict[str, str] | None:
+    platform_tag = linux_wheel_platform_tag()
+    if platform_tag is None:
+        return None
+
+    try:
+        probe = subprocess.run(
+            [
+                sys.executable,
+                "-c",
+                (
+                    "import json, sys, re, torch; "
+                    "parts = torch.__version__.split('+', 1)[0].split('.')[:2]; "
+                    "minor = re.sub(r'[^0-9].*', '', parts[1]) if len(parts) > 1 else '0'; "
+                    "torch_mm = parts[0] + '.' + minor; "
+                    "print(json.dumps({"
+                    "'python_tag': f'cp{sys.version_info.major}{sys.version_info.minor}', "
+                    "'torch_mm': torch_mm, "
+                    "'cuda_major': str(int(str(torch.version.cuda).split('.', 1)[0])) if torch.version.cuda else '', "
+                    "'hip_version': str(torch.version.hip) if getattr(torch.version, 'hip', None) else '', "
+                    "'cxx11abi': str(torch._C._GLIBCXX_USE_CXX11_ABI).upper()"
+                    "}))"
+                ),
+            ],
+            stdout = subprocess.PIPE,
+            stderr = subprocess.PIPE,
+            text = True,
+            timeout = timeout,
+        )
+    except subprocess.TimeoutExpired:
+        return None
+
+    if probe.returncode != 0:
+        return None
+
+    try:
+        env = json.loads(probe.stdout.strip())
+    except json.JSONDecodeError:
+        return None
+    env["platform_tag"] = platform_tag
+    return env
+
+
+def direct_wheel_url(
+    *,
+    filename_prefix: str,
+    package_version: str,
+    release_tag: str,
+    release_base_url: str,
+    env: dict[str, str] | None,
+) -> str | None:
+    if env is None or not env.get("cuda_major"):
+        return None
+
+    filename = (
+        f"{filename_prefix}-{package_version}"
+        f"+cu{env['cuda_major']}torch{env['torch_mm']}"
+        f"cxx11abi{env['cxx11abi']}-{env['python_tag']}-{env['python_tag']}"
+        f"-{env['platform_tag']}.whl"
+    )
+    return f"{release_base_url}/{release_tag}/{filename}"
+
+
+def flash_attn_package_version(torch_mm: str) -> str | None:
+    if torch_mm == "2.10":
+        return "2.8.1"
+    try:
+        major, minor = (int(part) for part in torch_mm.split(".", 1))
+    except ValueError:
+        return None
+    if major == 2 and 4 <= minor <= 9:
+        return "2.8.3"
+    return None
+
+
+def flash_attn_wheel_url(env: dict[str, str] | None) -> str | None:
+    if env is None:
+        return None
+    package_version = flash_attn_package_version(env["torch_mm"])
+    if package_version is None:
+        return None
+    return direct_wheel_url(
+        filename_prefix = "flash_attn",
+        package_version = package_version,
+        release_tag = f"v{package_version}",
+        release_base_url = FLASH_ATTN_RELEASE_BASE_URL,
+        env = env,
+    )
+
+
+def install_wheel(
+    wheel_url: str,
+    *,
+    python_executable: str,
+    use_uv: bool,
+    uv_needs_system: bool = False,
+    run: Callable[..., subprocess.CompletedProcess[str]] = subprocess.run,
+) -> list[tuple[str, subprocess.CompletedProcess[str]]]:
+    attempts: list[tuple[str, subprocess.CompletedProcess[str]]] = []
+
+    # Try uv first if available, then fall back to pip
+    if use_uv and shutil.which("uv"):
+        uv_cmd = ["uv", "pip", "install"]
+        if uv_needs_system:
+            uv_cmd.append("--system")
+        uv_cmd.extend(["--python", python_executable, "--no-deps", wheel_url])
+        result = run(
+            uv_cmd,
+            stdout = subprocess.PIPE,
+            stderr = subprocess.STDOUT,
+            text = True,
+        )
+        attempts.append(("uv", result))
+        if result.returncode == 0:
+            return attempts
+
+    pip_cmd = [python_executable, "-m", "pip", "install", "--no-deps", wheel_url]
+    result = run(
+        pip_cmd,
+        stdout = subprocess.PIPE,
+        stderr = subprocess.STDOUT,
+        text = True,
+    )
+    attempts.append(("pip", result))
+    return attempts
+
+
+def url_exists(url: str) -> bool:
+    try:
+        request = urllib.request.Request(url, method = "HEAD")
+        with urllib.request.urlopen(request, timeout = 10):
+            return True
+    except urllib.error.HTTPError as exc:
+        _logger.debug("url_exists(%s): HTTP %s", url, exc.code)
+    except (urllib.error.URLError, TimeoutError) as exc:
+        _logger.debug("url_exists(%s): %s", url, exc)
+    return False
diff --git a/studio/install_python_stack.py b/studio/install_python_stack.py
index a046f8f892..71a2a5bf83 100644
--- a/studio/install_python_stack.py
+++ b/studio/install_python_stack.py
@@ -21,6 +21,14 @@ import tempfile
 import urllib.request
 from pathlib import Path
 
+from backend.utils.wheel_utils import (
+    flash_attn_package_version,
+    flash_attn_wheel_url,
+    install_wheel,
+    probe_torch_wheel_env,
+    url_exists,
+)
+
 IS_WINDOWS = sys.platform == "win32"
 IS_MACOS = sys.platform == "darwin"
 IS_MAC_INTEL = IS_MACOS and platform.machine() == "x86_64"
@@ -368,7 +376,6 @@ NO_TORCH = _infer_no_torch()
 VERBOSE: bool = os.environ.get("UNSLOTH_VERBOSE", "0") == "1"
 
 # Progress bar state -- updated by _progress() as each install step runs.
-# _TOTAL counts: pip-upgrade + 7 shared steps + triton (non-Windows) + local-plugin + finalize
 # Update _TOTAL here if you add or remove install steps in install_python_stack().
 _STEP: int = 0
 _TOTAL: int = 0  # set at runtime in install_python_stack() based on platform
@@ -535,6 +542,66 @@ NO_TORCH_SKIP_PACKAGES = {
     "transformers-cfg",
 }
 
+
+def _select_flash_attn_version(torch_mm: str) -> str | None:
+    return flash_attn_package_version(torch_mm)
+
+
+def _build_flash_attn_wheel_url(env: dict[str, str]) -> str | None:
+    return flash_attn_wheel_url(env)
+
+
+def _print_optional_install_failure(
+    label: str, result: subprocess.CompletedProcess[str]
+) -> None:
+    _step("warning", f"{label} failed (exit code {result.returncode})", _cyan)
+    if result.stdout:
+        print(result.stdout.strip())
+
+
+def _flash_attn_install_disabled() -> bool:
+    return os.getenv("UNSLOTH_STUDIO_SKIP_FLASHATTN_INSTALL") == "1"
+
+
+def _ensure_flash_attn() -> None:
+    if NO_TORCH or IS_WINDOWS or IS_MACOS:
+        return
+    if _flash_attn_install_disabled():
+        return
+    if (
+        subprocess.run(
+            [sys.executable, "-c", "import flash_attn"],
+            stdout = subprocess.DEVNULL,
+            stderr = subprocess.DEVNULL,
+        ).returncode
+        == 0
+    ):
+        return
+
+    env = probe_torch_wheel_env()
+    wheel_url = _build_flash_attn_wheel_url(env) if env else None
+    if wheel_url and url_exists(wheel_url):
+        for installer, wheel_result in install_wheel(
+            wheel_url,
+            python_executable = sys.executable,
+            use_uv = USE_UV,
+            uv_needs_system = UV_NEEDS_SYSTEM,
+        ):
+            if wheel_result.returncode == 0:
+                return
+            _print_optional_install_failure(
+                f"Installing flash-attn prebuilt wheel with {installer}",
+                wheel_result,
+            )
+        _step("warning", "Continuing without flash-attn", _cyan)
+        return
+
+    if wheel_url is None:
+        _step("warning", "No compatible flash-attn prebuilt wheel found", _cyan)
+    else:
+        _step("warning", "No published flash-attn prebuilt wheel found", _cyan)
+
+
 # -- uv bootstrap ------------------------------------------------------
 
 USE_UV = False  # Set by _bootstrap_uv() at the start of install_python_stack()
@@ -762,10 +829,8 @@ def install_python_stack() -> int:
     base_total = 10 if IS_WINDOWS else 11
     if IS_MACOS:
         base_total -= 1  # triton step is skipped on macOS
-    # ROCm torch check steps (Linux only, non-macOS, non-no-torch):
-    # one early check (step 2b) and one final repair (step 13).
     if not IS_WINDOWS and not IS_MACOS and not NO_TORCH:
-        base_total += 2
+        base_total += 3
     _TOTAL = (base_total - 1) if skip_base else base_total
 
     # 1. Try to use uv for faster installs (must happen before pip upgrade
@@ -979,6 +1044,10 @@ def install_python_stack() -> int:
             constrain = False,
         )
 
+    if not IS_WINDOWS and not IS_MACOS and not NO_TORCH:
+        _progress("flash-attn")
+        _ensure_flash_attn()
+
     # # 6. Patch: override llama_cpp.py with fix from unsloth-zoo  feature/llama-cpp-windows-support branch
     # patch_package_file(
     #     "unsloth-zoo",
diff --git a/tests/python/test_flash_attn_install_python_stack.py b/tests/python/test_flash_attn_install_python_stack.py
new file mode 100644
index 0000000000..9881f2258e
--- /dev/null
+++ b/tests/python/test_flash_attn_install_python_stack.py
@@ -0,0 +1,282 @@
+"""Tests for the optional FlashAttention installer."""
+
+from __future__ import annotations
+
+import os
+import subprocess
+import sys
+from pathlib import Path
+from unittest import mock
+
+STUDIO_DIR = Path(__file__).resolve().parents[2] / "studio"
+sys.path.insert(0, str(STUDIO_DIR))
+
+import install_python_stack as ips
+
+
+class TestFlashAttnWheelSelection:
+    def test_torch_210_maps_to_v281(self):
+        assert ips._select_flash_attn_version("2.10") == "2.8.1"
+
+    def test_torch_29_maps_to_v283(self):
+        assert ips._select_flash_attn_version("2.9") == "2.8.3"
+
+    def test_unsupported_torch_has_no_wheel_mapping(self):
+        assert ips._select_flash_attn_version("2.11") is None
+
+    def test_exact_wheel_url_uses_full_env_tuple(self):
+        url = ips._build_flash_attn_wheel_url(
+            {
+                "python_tag": "cp313",
+                "torch_mm": "2.10",
+                "cuda_major": "12",
+                "cxx11abi": "TRUE",
+                "platform_tag": "linux_x86_64",
+            }
+        )
+        assert url is not None
+        assert "v2.8.1" in url
+        assert (
+            "flash_attn-2.8.1+cu12torch2.10cxx11abiTRUE-cp313-cp313-linux_x86_64.whl"
+            in url
+        )
+
+    def test_missing_cuda_major_disables_wheel_lookup(self):
+        assert (
+            ips._build_flash_attn_wheel_url(
+                {
+                    "python_tag": "cp313",
+                    "torch_mm": "2.10",
+                    "cuda_major": "",
+                    "cxx11abi": "TRUE",
+                    "platform_tag": "linux_x86_64",
+                }
+            )
+            is None
+        )
+
+
+class TestEnsureFlashAttn:
+    def _import_check(self, code: int = 1):
+        return subprocess.CompletedProcess(["python", "-c", "import flash_attn"], code)
+
+    def test_prefers_exact_match_wheel(self):
+        install_calls = []
+
+        def fake_install_wheel(*args, **kwargs):
+            install_calls.append((args, kwargs))
+            return [("uv", subprocess.CompletedProcess(["uv"], 0, ""))]
+
+        with (
+            mock.patch.object(ips, "NO_TORCH", False),
+            mock.patch.object(ips, "IS_WINDOWS", False),
+            mock.patch.object(ips, "IS_MACOS", False),
+            mock.patch.object(ips, "USE_UV", True),
+            mock.patch.object(ips, "UV_NEEDS_SYSTEM", False),
+            mock.patch.object(
+                ips,
+                "probe_torch_wheel_env",
+                return_value = {
+                    "python_tag": "cp313",
+                    "torch_mm": "2.10",
+                    "cuda_major": "12",
+                    "cxx11abi": "TRUE",
+                    "platform_tag": "linux_x86_64",
+                },
+            ),
+            mock.patch.object(ips, "url_exists", return_value = True),
+            mock.patch.object(ips, "install_wheel", side_effect = fake_install_wheel),
+            mock.patch("subprocess.run", return_value = self._import_check()),
+        ):
+            ips._ensure_flash_attn()
+
+        assert len(install_calls) == 1
+        args, kwargs = install_calls[0]
+        assert args == (
+            "https://github.com/Dao-AILab/flash-attention/releases/download/v2.8.1/flash_attn-2.8.1+cu12torch2.10cxx11abiTRUE-cp313-cp313-linux_x86_64.whl",
+        )
+        assert kwargs["python_executable"] == sys.executable
+        assert kwargs["use_uv"] is True
+        assert kwargs["uv_needs_system"] is False
+
+    def test_uv_install_respects_system_flag(self):
+        install_calls = []
+
+        def fake_install_wheel(*args, **kwargs):
+            install_calls.append((args, kwargs))
+            return [("uv", subprocess.CompletedProcess(["uv"], 0, ""))]
+
+        with (
+            mock.patch.object(ips, "NO_TORCH", False),
+            mock.patch.object(ips, "IS_WINDOWS", False),
+            mock.patch.object(ips, "IS_MACOS", False),
+            mock.patch.object(ips, "USE_UV", True),
+            mock.patch.object(ips, "UV_NEEDS_SYSTEM", True),
+            mock.patch.object(
+                ips,
+                "probe_torch_wheel_env",
+                return_value = {
+                    "python_tag": "cp313",
+                    "torch_mm": "2.10",
+                    "cuda_major": "12",
+                    "cxx11abi": "TRUE",
+                    "platform_tag": "linux_x86_64",
+                },
+            ),
+            mock.patch.object(ips, "url_exists", return_value = True),
+            mock.patch.object(ips, "install_wheel", side_effect = fake_install_wheel),
+            mock.patch("subprocess.run", return_value = self._import_check()),
+        ):
+            ips._ensure_flash_attn()
+
+        assert len(install_calls) == 1
+        _, kwargs = install_calls[0]
+        assert kwargs["uv_needs_system"] is True
+
+    def test_wheel_failure_warns_and_continues(self):
+        step_messages: list[tuple[str, str]] = []
+        printed_failures: list[str] = []
+
+        def fake_step(label: str, value: str, color_fn = None):
+            step_messages.append((label, value))
+
+        with (
+            mock.patch.object(ips, "NO_TORCH", False),
+            mock.patch.object(ips, "IS_WINDOWS", False),
+            mock.patch.object(ips, "IS_MACOS", False),
+            mock.patch.object(ips, "USE_UV", True),
+            mock.patch.object(ips, "UV_NEEDS_SYSTEM", False),
+            mock.patch.object(
+                ips,
+                "probe_torch_wheel_env",
+                return_value = {
+                    "python_tag": "cp313",
+                    "torch_mm": "2.10",
+                    "cuda_major": "12",
+                    "cxx11abi": "TRUE",
+                    "platform_tag": "linux_x86_64",
+                },
+            ),
+            mock.patch.object(ips, "url_exists", return_value = True),
+            mock.patch.object(
+                ips,
+                "install_wheel",
+                return_value = [
+                    ("uv", subprocess.CompletedProcess(["uv"], 1, "uv wheel failed")),
+                    (
+                        "pip",
+                        subprocess.CompletedProcess(["pip"], 1, "pip wheel failed"),
+                    ),
+                ],
+            ),
+            mock.patch.object(
+                ips,
+                "_print_optional_install_failure",
+                side_effect = lambda label, result: printed_failures.append(label),
+            ),
+            mock.patch.object(ips, "_step", side_effect = fake_step),
+            mock.patch("subprocess.run", return_value = self._import_check()),
+        ):
+            ips._ensure_flash_attn()
+
+        assert printed_failures == [
+            "Installing flash-attn prebuilt wheel with uv",
+            "Installing flash-attn prebuilt wheel with pip",
+        ]
+        assert ("warning", "Continuing without flash-attn") in step_messages
+
+    def test_wheel_missing_skips_install_at_setup_time(self):
+        step_messages: list[tuple[str, str]] = []
+
+        def fake_step(label: str, value: str, color_fn = None):
+            step_messages.append((label, value))
+
+        with (
+            mock.patch.object(ips, "NO_TORCH", False),
+            mock.patch.object(ips, "IS_WINDOWS", False),
+            mock.patch.object(ips, "IS_MACOS", False),
+            mock.patch.object(
+                ips,
+                "probe_torch_wheel_env",
+                return_value = {
+                    "python_tag": "cp313",
+                    "torch_mm": "2.10",
+                    "cuda_major": "13",
+                    "cxx11abi": "TRUE",
+                    "platform_tag": "linux_x86_64",
+                },
+            ),
+            mock.patch.object(ips, "url_exists", return_value = False),
+            mock.patch.object(ips, "install_wheel") as mock_install_wheel,
+            mock.patch.object(ips, "_step", side_effect = fake_step),
+            mock.patch("subprocess.run", return_value = self._import_check()),
+        ):
+            ips._ensure_flash_attn()
+
+        mock_install_wheel.assert_not_called()
+        assert (
+            "warning",
+            "No published flash-attn prebuilt wheel found",
+        ) in step_messages
+
+    def test_skip_env_disables_setup_install(self):
+        with (
+            mock.patch.object(ips, "NO_TORCH", False),
+            mock.patch.object(ips, "IS_WINDOWS", False),
+            mock.patch.object(ips, "IS_MACOS", False),
+            mock.patch.dict(os.environ, {"UNSLOTH_STUDIO_SKIP_FLASHATTN_INSTALL": "1"}),
+            mock.patch.object(ips, "probe_torch_wheel_env") as mock_probe,
+            mock.patch.object(ips, "install_wheel") as mock_install_wheel,
+            mock.patch("subprocess.run", return_value = self._import_check()),
+        ):
+            ips._ensure_flash_attn()
+
+        mock_probe.assert_not_called()
+        mock_install_wheel.assert_not_called()
+
+
+class TestInstallPythonStackFlashAttnIntegration:
+    def _run_install(self, *, no_torch: bool, is_macos: bool, is_windows: bool) -> int:
+        flash_attn_calls = 0
+
+        def fake_run(cmd, **kw):
+            return subprocess.CompletedProcess(cmd, 0, b"", b"")
+
+        def count_flash_attn():
+            nonlocal flash_attn_calls
+            flash_attn_calls += 1
+
+        with (
+            mock.patch.object(ips, "NO_TORCH", no_torch),
+            mock.patch.object(ips, "IS_MACOS", is_macos),
+            mock.patch.object(ips, "IS_WINDOWS", is_windows),
+            mock.patch.object(ips, "USE_UV", True),
+            mock.patch.object(ips, "UV_NEEDS_SYSTEM", False),
+            mock.patch.object(ips, "VERBOSE", False),
+            mock.patch.object(ips, "_bootstrap_uv", return_value = True),
+            mock.patch.object(ips, "_ensure_flash_attn", side_effect = count_flash_attn),
+            mock.patch("subprocess.run", side_effect = fake_run),
+            mock.patch.object(ips, "_has_usable_nvidia_gpu", return_value = False),
+            mock.patch.object(ips, "_has_rocm_gpu", return_value = False),
+            mock.patch.object(
+                ips, "LOCAL_DD_UNSTRUCTURED_PLUGIN", Path("/fake/plugin")
+            ),
+            mock.patch("pathlib.Path.is_dir", return_value = True),
+            mock.patch("pathlib.Path.is_file", return_value = True),
+            mock.patch.dict(os.environ, {"SKIP_STUDIO_BASE": "1"}, clear = False),
+        ):
+            ips.install_python_stack()
+
+        return flash_attn_calls
+
+    def test_linux_torch_install_calls_flash_attn_step(self):
+        assert self._run_install(no_torch = False, is_macos = False, is_windows = False) == 1
+
+    def test_no_torch_install_skips_flash_attn_step(self):
+        assert self._run_install(no_torch = True, is_macos = False, is_windows = False) == 0
+
+    def test_macos_install_skips_flash_attn_step(self):
+        assert self._run_install(no_torch = False, is_macos = True, is_windows = False) == 0
+
+    def test_windows_install_skips_flash_attn_step(self):
+        assert self._run_install(no_torch = False, is_macos = False, is_windows = True) == 0
diff --git a/tests/python/test_no_torch_filtering.py b/tests/python/test_no_torch_filtering.py
index 5c2926a1f1..29cadf87ae 100644
--- a/tests/python/test_no_torch_filtering.py
+++ b/tests/python/test_no_torch_filtering.py
@@ -419,6 +419,9 @@ class TestInstallPythonStackSubprocessMock:
             mock.patch.object(ips, "USE_UV", True),
             mock.patch.object(ips, "UV_NEEDS_SYSTEM", False),
             mock.patch.object(ips, "VERBOSE", False),
+            mock.patch.object(ips, "_ensure_flash_attn", return_value = None),
+            mock.patch.object(ips, "_has_usable_nvidia_gpu", return_value = False),
+            mock.patch.object(ips, "_has_rocm_gpu", return_value = False),
             mock.patch("subprocess.run", side_effect = mock_run),
             mock.patch.object(ips, "_bootstrap_uv", return_value = True),
             mock.patch.object(
@@ -637,7 +640,7 @@ class TestInstallShNoTorchFlag:
     def test_cpu_hint_message_exists(self):
         """CPU hint message must exist in install.sh."""
         assert (
-            "No NVIDIA GPU detected" in self.source
+            "No GPU detected" in self.source
         ), "CPU hint message not found in install.sh"
         assert (
             "--no-torch" in self.source

From a31c82a640117906d44f5e8906b93d2ec29872a2 Mon Sep 17 00:00:00 2001
From: TF-MTGE <m.gegenbauer@test-fuchs.com>
Date: Tue, 14 Apr 2026 17:53:14 +0200
Subject: [PATCH 085/216] fix(studio): remove 300s cap on load_checkpoint
 (inherits 3600s default) (#4922)

* fix: increase wait response timeout to 900 sec instead of 300 sec. #4845

* Apply suggestion from @gemini-code-assist[bot]

good catch

Co-authored-by: gemini-code-assist[bot] <176961590+gemini-code-assist[bot]@users.noreply.github.com>

---------

Co-authored-by: Roland Tannous <115670425+rolandtannous@users.noreply.github.com>
Co-authored-by: gemini-code-assist[bot] <176961590+gemini-code-assist[bot]@users.noreply.github.com>
---
 studio/backend/core/export/orchestrator.py | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/studio/backend/core/export/orchestrator.py b/studio/backend/core/export/orchestrator.py
index 500bc9e706..e52aa8c3cd 100644
--- a/studio/backend/core/export/orchestrator.py
+++ b/studio/backend/core/export/orchestrator.py
@@ -241,7 +241,7 @@ class ExportOrchestrator:
         self._spawn_subprocess(sub_config)
 
         try:
-            resp = self._wait_response("loaded", timeout = 300)
+            resp = self._wait_response("loaded")
         except RuntimeError as exc:
             self._shutdown_subprocess(timeout = 5)
             self.current_checkpoint = None

From 767fa8cade240f1e7a85c5728398bc5d0bd32989 Mon Sep 17 00:00:00 2001
From: Daniel Han <danielhanchen@gmail.com>
Date: Tue, 14 Apr 2026 08:53:25 -0700
Subject: [PATCH 086/216] Studio: honor explicit GGUF ctx and default to 4096
 when weights exceed VRAM (#5011)

* Studio: honor explicit GGUF ctx and default to 4096 when weights exceed VRAM

The load-time auto-fit in LlamaCppBackend.load_model had two issues for
models whose weights do not fit on any GPU subset (the common case for
large MoE GGUFs such as MiniMax-M2.7, Qwen3.5-397B-A17B, etc.):

1. Auto mode (max_seq_length=0) left effective_ctx at the model's native
   context when no subset passed the 90% fit check. The UI slider then
   landed on e.g. 196608 for MiniMax-M2.7, far above anything usable.
   Default the auto-pick to 4096 so the UI starts at a sane value; the
   slider ceiling stays at the native context so the user can still
   opt in to longer contexts and receive the "might be slower" warning.

2. Explicit ctx was silently shrunk when weights fit but the requested
   KV overflowed the 90% budget. The shrink loop emitted -c <capped>
   -ngl -1 without informing the caller, so a user who had opted into
   a longer context via the UI never actually got it. Drop the shrink
   loop on the explicit path and emit -c <user_ctx> --fit on instead,
   letting llama-server flex -ngl (CPU layer offload).

Adds tests/test_llama_cpp_context_fit.py covering both paths, the
file-size-only fallback when KV metadata is missing, non-regression on
fittable auto-pick, and platform-agnostic input shape.

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

---------

Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com>
---
 studio/backend/core/inference/llama_cpp.py    |  45 +-
 .../tests/test_llama_cpp_context_fit.py       | 389 ++++++++++++++++++
 2 files changed, 410 insertions(+), 24 deletions(-)
 create mode 100644 studio/backend/tests/test_llama_cpp_context_fit.py

diff --git a/studio/backend/core/inference/llama_cpp.py b/studio/backend/core/inference/llama_cpp.py
index d4a33387dc..9f8c14e51f 100644
--- a/studio/backend/core/inference/llama_cpp.py
+++ b/studio/backend/core/inference/llama_cpp.py
@@ -1195,34 +1195,17 @@ class LlamaCppBackend:
                             max_available_ctx = best_cap
 
                     if explicit_ctx:
-                        # Try to honor the user's requested context exactly.
+                        # Honor the user's requested context verbatim. If it
+                        # fits, pin GPUs and skip --fit; if it doesn't, ship
+                        # -c <user_ctx> --fit on and let llama-server flex
+                        # -ngl (CPU layer offload). The UI is expected to
+                        # have surfaced the "might be slower" warning before
+                        # the user submitted a ctx above the fit ceiling.
                         requested_total = model_size + self._estimate_kv_cache_bytes(
                             effective_ctx, cache_type_kv
                         )
                         gpu_indices, use_fit = self._select_gpus(requested_total, gpus)
-
-                        # Full context doesn't fit anywhere -- cap it on the
-                        # best GPU subset we can find (fewest GPUs first).
-                        if use_fit:
-                            ranked = sorted(gpus, key = lambda g: g[1], reverse = True)
-                            for n_gpus in range(1, len(ranked) + 1):
-                                subset = ranked[:n_gpus]
-                                pool_mib = sum(free for _, free in subset)
-                                capped = self._fit_context_to_vram(
-                                    effective_ctx,
-                                    pool_mib,
-                                    model_size,
-                                    cache_type_kv,
-                                )
-                                kv = self._estimate_kv_cache_bytes(
-                                    capped, cache_type_kv
-                                )
-                                total_mib = (model_size + kv) / (1024 * 1024)
-                                if total_mib <= pool_mib * 0.90:
-                                    effective_ctx = capped
-                                    gpu_indices = sorted(idx for idx, _ in subset)
-                                    use_fit = False
-                                    break
+                        # No silent shrink: effective_ctx stays == n_ctx.
                     else:
                         # Auto context: prefer fewer GPUs, cap context to fit.
                         ranked = sorted(gpus, key = lambda g: g[1], reverse = True)
@@ -1242,6 +1225,13 @@ class LlamaCppBackend:
                                 gpu_indices = sorted(idx for idx, _ in subset)
                                 use_fit = False
                                 break
+                        else:
+                            # No subset can host the weights (weights alone
+                            # exceed 90% of every pool). Per spec, default
+                            # the UI-visible context to 4096 and let
+                            # --fit on flex -ngl so llama-server offloads
+                            # layers to CPU RAM.
+                            effective_ctx = min(4096, effective_ctx)
 
                 elif gpus:
                     # Can't estimate KV -- fall back to file-size-only check.
@@ -1252,6 +1242,13 @@ class LlamaCppBackend:
                         model_size_gb = round(model_size / (1024**3), 2),
                     )
                     gpu_indices, use_fit = self._select_gpus(model_size, gpus)
+                    if use_fit and not explicit_ctx:
+                        # Weights don't fit on any subset. Default the UI to
+                        # 4096 so the slider doesn't land on an unusable native
+                        # context. --fit on will flex -ngl at runtime.
+                        effective_ctx = (
+                            min(4096, effective_ctx) if effective_ctx > 0 else 4096
+                        )
 
                 if effective_ctx < original_ctx:
                     kv_est = self._estimate_kv_cache_bytes(effective_ctx, cache_type_kv)
diff --git a/studio/backend/tests/test_llama_cpp_context_fit.py b/studio/backend/tests/test_llama_cpp_context_fit.py
new file mode 100644
index 0000000000..f498655347
--- /dev/null
+++ b/studio/backend/tests/test_llama_cpp_context_fit.py
@@ -0,0 +1,389 @@
+# SPDX-License-Identifier: AGPL-3.0-only
+# Copyright 2026-present the Unsloth AI Inc. team. All rights reserved. See /studio/LICENSE.AGPL-3.0
+
+"""Tests for the GGUF load-time context auto-fit decision.
+
+Guards two regressions in ``LlamaCppBackend.load_model``:
+
+1. **Auto mode on weights-exceed-VRAM** (``n_ctx == 0``): when the model
+   weights alone exceed 90% of every GPU subset's free memory, the
+   auto-pick loop used to exit without matching, leaving
+   ``effective_ctx`` at the model's native context (e.g. 196608 for
+   MiniMax-M2.7). The intended default per Studio's UI spec is 4096 so
+   the slider lands on a usable value; the user can still drag higher
+   and trigger ``--fit on`` with a warning.
+
+2. **Explicit ctx silently shrunk when KV overflows**: with fittable
+   weights but a requested ctx whose KV cache pushes total memory over
+   90% of VRAM, the old code binary-searched a smaller ctx and emitted
+   ``-c <capped> -ngl -1`` without informing the caller. The UI had
+   already surfaced its "might be slower" warning and expects the user's
+   explicit ctx to be honored with ``--fit on`` flexing ``-ngl`` instead.
+
+Tests avoid GPU probing, subprocess spawning, and GGUF I/O by driving the
+post-metadata decision block directly against a stubbed instance.
+
+Requires no GPU, network, or external libraries beyond pytest.
+Cross-platform: Linux, macOS, Windows, WSL.
+"""
+
+from __future__ import annotations
+
+import sys
+import types as _types
+from pathlib import Path
+
+import pytest
+
+# ---------------------------------------------------------------------------
+# Stub heavy / unavailable external dependencies before importing the
+# module under test.  Same pattern as test_kv_cache_estimation.py.
+# ---------------------------------------------------------------------------
+
+_BACKEND_DIR = str(Path(__file__).resolve().parent.parent)
+if _BACKEND_DIR not in sys.path:
+    sys.path.insert(0, _BACKEND_DIR)
+
+# loggers
+_loggers_stub = _types.ModuleType("loggers")
+_loggers_stub.get_logger = lambda name: __import__("logging").getLogger(name)
+sys.modules.setdefault("loggers", _loggers_stub)
+
+# structlog
+_structlog_stub = _types.ModuleType("structlog")
+sys.modules.setdefault("structlog", _structlog_stub)
+
+# httpx
+_httpx_stub = _types.ModuleType("httpx")
+for _exc_name in (
+    "ConnectError",
+    "TimeoutException",
+    "ReadTimeout",
+    "ReadError",
+    "RemoteProtocolError",
+    "CloseError",
+):
+    setattr(_httpx_stub, _exc_name, type(_exc_name, (Exception,), {}))
+
+
+class _FakeTimeout:
+    def __init__(self, *a, **kw):
+        pass
+
+
+_httpx_stub.Timeout = _FakeTimeout
+_httpx_stub.Client = type(
+    "Client",
+    (),
+    {
+        "__init__": lambda self, **kw: None,
+        "__enter__": lambda self: self,
+        "__exit__": lambda self, *a: None,
+    },
+)
+sys.modules.setdefault("httpx", _httpx_stub)
+
+from core.inference.llama_cpp import LlamaCppBackend
+
+
+# ---------------------------------------------------------------------------
+# Helpers
+# ---------------------------------------------------------------------------
+
+GIB = 1024**3
+FALLBACK_CTX = 4096
+
+
+def _make_backend(
+    native_ctx = 131072,
+    n_layers = 80,
+    n_kv_heads = 8,
+    n_heads = 64,
+    kv_key_length = 128,
+    kv_value_length = 128,
+):
+    """Create a LlamaCppBackend instance with GGUF metadata fields set and
+    the helpers used by the decision block stubbed out."""
+    inst = LlamaCppBackend.__new__(LlamaCppBackend)
+    inst._context_length = native_ctx
+    inst._n_layers = n_layers
+    inst._n_kv_heads = n_kv_heads
+    inst._n_heads = n_heads
+    inst._embedding_length = 8192
+    inst._kv_key_length = kv_key_length
+    inst._kv_value_length = kv_value_length
+    inst._kv_lora_rank = None
+    inst._sliding_window = None
+    inst._ssm_inner_size = None
+    inst._full_attention_interval = None
+    inst._key_length_mla = None
+    return inst
+
+
+def _drive(
+    n_ctx,
+    model_gib,
+    gpus,
+    native_ctx = 131072,
+    kv_per_token_bytes = 325_000,
+    can_estimate_kv = True,
+):
+    """Drive the post-metadata portion of load_model with stubbed inputs.
+
+    Mirrors the decision block at llama_cpp.py:1137-1296 so we can assert
+    the command that would be built, without subprocesses or GPU probes.
+    """
+    inst = _make_backend(native_ctx = native_ctx)
+    model_size = int(model_gib * GIB)
+    cache_type_kv = None
+
+    def fake_estimate(n_ctx_, _type = None):
+        return 0 if n_ctx_ <= 0 else n_ctx_ * kv_per_token_bytes
+
+    inst._estimate_kv_cache_bytes = fake_estimate
+    inst._can_estimate_kv = lambda: can_estimate_kv
+
+    context_length = inst._context_length
+
+    effective_ctx = n_ctx if n_ctx > 0 else (context_length or 0)
+    max_available_ctx = context_length or effective_ctx
+    if n_ctx > 0:
+        effective_ctx = n_ctx
+    elif context_length is not None:
+        effective_ctx = context_length
+    else:
+        effective_ctx = 0
+    original_ctx = effective_ctx
+    max_available_ctx = context_length or effective_ctx
+
+    gpu_indices, use_fit = None, True
+    explicit_ctx = n_ctx > 0
+
+    if gpus and inst._can_estimate_kv() and effective_ctx > 0:
+        native_ctx_for_cap = context_length or effective_ctx
+        if native_ctx_for_cap > 0:
+            ranked_for_cap = sorted(gpus, key = lambda g: g[1], reverse = True)
+            best_cap = 0
+            for n_gpus in range(1, len(ranked_for_cap) + 1):
+                subset = ranked_for_cap[:n_gpus]
+                pool_mib = sum(free for _, free in subset)
+                capped = inst._fit_context_to_vram(
+                    native_ctx_for_cap,
+                    pool_mib,
+                    model_size,
+                    cache_type_kv,
+                )
+                kv = inst._estimate_kv_cache_bytes(capped, cache_type_kv)
+                total_mib = (model_size + kv) / (1024 * 1024)
+                if total_mib <= pool_mib * 0.90:
+                    best_cap = max(best_cap, capped)
+            if best_cap > 0:
+                max_available_ctx = best_cap
+
+        if explicit_ctx:
+            requested_total = model_size + inst._estimate_kv_cache_bytes(
+                effective_ctx, cache_type_kv
+            )
+            gpu_indices, use_fit = inst._select_gpus(requested_total, gpus)
+        else:
+            ranked = sorted(gpus, key = lambda g: g[1], reverse = True)
+            matched = False
+            for n_gpus in range(1, len(ranked) + 1):
+                subset = ranked[:n_gpus]
+                pool_mib = sum(free for _, free in subset)
+                capped = inst._fit_context_to_vram(
+                    effective_ctx,
+                    pool_mib,
+                    model_size,
+                    cache_type_kv,
+                )
+                kv = inst._estimate_kv_cache_bytes(capped, cache_type_kv)
+                total_mib = (model_size + kv) / (1024 * 1024)
+                if total_mib <= pool_mib * 0.90:
+                    effective_ctx = capped
+                    gpu_indices = sorted(idx for idx, _ in subset)
+                    use_fit = False
+                    matched = True
+                    break
+            if not matched:
+                effective_ctx = min(FALLBACK_CTX, effective_ctx)
+    elif gpus:
+        gpu_indices, use_fit = inst._select_gpus(model_size, gpus)
+        if use_fit and not explicit_ctx:
+            effective_ctx = (
+                min(FALLBACK_CTX, effective_ctx) if effective_ctx > 0 else FALLBACK_CTX
+            )
+
+    return {
+        "c_arg": effective_ctx if effective_ctx > 0 else 0,
+        "use_fit": use_fit,
+        "gpu_indices": gpu_indices,
+        "max_available_ctx": max_available_ctx,
+        "original_ctx": original_ctx,
+    }
+
+
+# ---------------------------------------------------------------------------
+# Auto mode, model weights exceed VRAM  (Bug A guard)
+# ---------------------------------------------------------------------------
+
+
+class TestAutoModeWeightsExceedVRAM:
+    """``n_ctx == 0`` on a model whose weights don't fit anywhere."""
+
+    def test_minimax_like_single_gpu(self):
+        plan = _drive(
+            n_ctx = 0,
+            model_gib = 131,
+            gpus = [(0, 97_000)],
+            native_ctx = 196608,
+        )
+        assert plan["c_arg"] == FALLBACK_CTX
+        assert plan["use_fit"] is True
+        assert plan["gpu_indices"] is None
+        # UI slider ceiling stays at native: user can still drag higher
+        # and get the "might be slower" path.
+        assert plan["max_available_ctx"] == 196608
+
+    def test_multi_gpu_all_subsets_fail(self):
+        plan = _drive(
+            n_ctx = 0,
+            model_gib = 400,
+            gpus = [(0, 80_000), (1, 80_000), (2, 80_000), (3, 80_000)],
+            native_ctx = 131072,
+        )
+        assert plan["c_arg"] == FALLBACK_CTX
+        assert plan["use_fit"] is True
+        assert plan["gpu_indices"] is None
+
+    def test_no_kv_metadata_auto(self):
+        """File-size-only fallback path also defaults to 4096."""
+        plan = _drive(
+            n_ctx = 0,
+            model_gib = 131,
+            gpus = [(0, 97_000)],
+            native_ctx = 196608,
+            can_estimate_kv = False,
+        )
+        assert plan["c_arg"] == FALLBACK_CTX
+        assert plan["use_fit"] is True
+
+
+# ---------------------------------------------------------------------------
+# Explicit ctx, KV overflows fittable weights  (Bug B guard)
+# ---------------------------------------------------------------------------
+
+
+class TestExplicitCtxRespectsUser:
+    """``n_ctx > 0`` must never be silently shrunk."""
+
+    def test_fittable_weights_oversized_kv(self):
+        # 8 GB weights + 131k ctx KV on 24 GB VRAM.
+        # Budget = 21.6 GB, KV at 131k >> 13.6 GB remaining, so
+        # _select_gpus flips use_fit=True.
+        plan = _drive(
+            n_ctx = 131072,
+            model_gib = 8,
+            gpus = [(0, 24_000)],
+            native_ctx = 131072,
+        )
+        assert plan["c_arg"] == 131072
+        assert plan["use_fit"] is True
+        assert plan["gpu_indices"] is None
+
+    def test_explicit_that_fits_uses_ngl(self):
+        plan = _drive(
+            n_ctx = 8192,
+            model_gib = 8,
+            gpus = [(0, 24_000)],
+            native_ctx = 131072,
+        )
+        assert plan["c_arg"] == 8192
+        assert plan["use_fit"] is False
+        assert plan["gpu_indices"] == [0]
+
+    def test_explicit_on_weights_exceed_vram(self):
+        # User drags the slider to 32k on a too-big model: honored.
+        plan = _drive(
+            n_ctx = 32768,
+            model_gib = 131,
+            gpus = [(0, 97_000)],
+            native_ctx = 196608,
+        )
+        assert plan["c_arg"] == 32768
+        assert plan["use_fit"] is True
+
+    def test_explicit_at_fallback_on_too_big(self):
+        plan = _drive(
+            n_ctx = FALLBACK_CTX,
+            model_gib = 131,
+            gpus = [(0, 97_000)],
+            native_ctx = 196608,
+        )
+        assert plan["c_arg"] == FALLBACK_CTX
+        assert plan["use_fit"] is True
+
+    def test_explicit_below_floor_honored(self):
+        # 2048 is below --fit-ctx default; still honored since user set it.
+        plan = _drive(
+            n_ctx = 2048,
+            model_gib = 8,
+            gpus = [(0, 24_000)],
+        )
+        assert plan["c_arg"] == 2048
+
+
+# ---------------------------------------------------------------------------
+# Non-regression: fittable + auto still auto-picks largest fitting ctx
+# ---------------------------------------------------------------------------
+
+
+class TestFittableAutoPickRegressions:
+    def test_small_model_one_gpu(self):
+        plan = _drive(
+            n_ctx = 0,
+            model_gib = 8,
+            gpus = [(0, 24_000)],
+            native_ctx = 131072,
+            kv_per_token_bytes = 8192,
+        )
+        assert plan["use_fit"] is False
+        assert plan["gpu_indices"] == [0]
+        assert plan["c_arg"] > FALLBACK_CTX
+
+    def test_medium_model_needs_multi_gpu(self):
+        plan = _drive(
+            n_ctx = 0,
+            model_gib = 60,
+            gpus = [(0, 40_000), (1, 40_000)],
+            native_ctx = 131072,
+            kv_per_token_bytes = 8192,
+        )
+        assert plan["use_fit"] is False
+        assert plan["gpu_indices"] == [0, 1]
+
+    def test_no_kv_metadata_fittable_auto(self):
+        plan = _drive(
+            n_ctx = 0,
+            model_gib = 8,
+            gpus = [(0, 24_000)],
+            native_ctx = 131072,
+            can_estimate_kv = False,
+        )
+        assert plan["use_fit"] is False
+        assert plan["gpu_indices"] == [0]
+
+
+# ---------------------------------------------------------------------------
+# Platform-agnostic input shape
+# ---------------------------------------------------------------------------
+
+
+@pytest.mark.parametrize("platform_tag", ["linux", "windows", "mac", "rocm"])
+def test_identical_decision_across_platforms(platform_tag):
+    """The decision function takes ``[(gpu_idx, free_mib), ...]`` regardless
+    of how upstream (nvidia-smi / nvidia-smi.exe / Metal / rocm-smi) produced
+    it. Identical inputs must yield identical plans."""
+    plan_a = _drive(n_ctx = 0, model_gib = 8, gpus = [(0, 24_000)])
+    plan_b = _drive(n_ctx = 0, model_gib = 8, gpus = [(0, 24_000)])
+    assert plan_a == plan_b, platform_tag

From b2f80f210ee91bc5bdcd359eda678e57ed320b34 Mon Sep 17 00:00:00 2001
From: Daniel Han <danielhanchen@gmail.com>
Date: Tue, 14 Apr 2026 08:53:37 -0700
Subject: [PATCH 087/216] Studio: make GGUF disk-space preflight cache-aware
 (#5012)

* Studio: make GGUF disk-space preflight cache-aware

The pre-download disk check in LlamaCppBackend.load_model compared the
repo's total GGUF size against free disk without crediting bytes
already present in the Hugging Face cache. Re-loading a large cached
model (e.g. MiniMax-M2.7-GGUF at 131 GB) then failed cold with
"Not enough disk space to download any variant" whenever free disk
was below the full weight footprint, even though nothing actually
needed to be downloaded.

Subtract bytes already on disk via try_to_load_from_cache before
comparing against free space. A partial blob (interrupted download) is
not credited, so a second attempt still allocates room to finish the
download. The log line now also surfaces how much is already cached.

Adds tests/test_llama_cpp_cache_aware_disk_check.py covering the
fully-cached, partial-cache-insufficient-disk, partial-cache-enough-disk,
cold-cache, incomplete-blob, and zero-size-path-info cases. Sparse
tempfiles keep the GB-scale scenarios cheap to simulate.

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

---------

Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com>
---
 studio/backend/core/inference/llama_cpp.py    |  32 ++-
 .../test_llama_cpp_cache_aware_disk_check.py  | 243 ++++++++++++++++++
 2 files changed, 272 insertions(+), 3 deletions(-)
 create mode 100644 studio/backend/tests/test_llama_cpp_cache_aware_disk_check.py

diff --git a/studio/backend/core/inference/llama_cpp.py b/studio/backend/core/inference/llama_cpp.py
index 9f8c14e51f..b0d48313d8 100644
--- a/studio/backend/core/inference/llama_cpp.py
+++ b/studio/backend/core/inference/llama_cpp.py
@@ -904,10 +904,34 @@ class LlamaCppBackend:
         try:
             import os
 
-            from huggingface_hub import get_paths_info
+            from huggingface_hub import get_paths_info, try_to_load_from_cache
 
             path_infos = list(get_paths_info(hf_repo, all_gguf_files, token = hf_token))
-            total_download_bytes = sum((p.size or 0) for p in path_infos)
+            total_bytes = sum((p.size or 0) for p in path_infos)
+
+            # Subtract bytes already present in the HF cache so we only
+            # preflight against what we actually have to download. Without
+            # this, re-loading a cached large model (e.g. MiniMax-M2.7-GGUF
+            # at 131 GB) fails cold whenever free disk is below the full
+            # weight footprint, even though nothing needs downloading.
+            already_cached_bytes = 0
+            for p in path_infos:
+                if not p.size:
+                    continue
+                try:
+                    cached_path = try_to_load_from_cache(hf_repo, p.path)
+                except Exception:
+                    cached_path = None
+                if isinstance(cached_path, str) and os.path.exists(cached_path):
+                    try:
+                        on_disk = os.path.getsize(cached_path)
+                    except OSError:
+                        on_disk = 0
+                    # Count as satisfied only when the full blob is present.
+                    if on_disk >= p.size:
+                        already_cached_bytes += p.size
+
+            total_download_bytes = max(0, total_bytes - already_cached_bytes)
 
             if total_download_bytes > 0:
                 cache_dir = os.environ.get(
@@ -919,9 +943,11 @@ class LlamaCppBackend:
 
                 total_gb = total_download_bytes / (1024**3)
                 free_gb = free_bytes / (1024**3)
+                cached_gb = already_cached_bytes / (1024**3)
 
                 logger.info(
-                    f"GGUF download: {total_gb:.1f} GB needed, "
+                    f"GGUF download: {total_gb:.1f} GB needed "
+                    f"({cached_gb:.1f} GB already cached), "
                     f"{free_gb:.1f} GB free on disk"
                 )
 
diff --git a/studio/backend/tests/test_llama_cpp_cache_aware_disk_check.py b/studio/backend/tests/test_llama_cpp_cache_aware_disk_check.py
new file mode 100644
index 0000000000..255c04a956
--- /dev/null
+++ b/studio/backend/tests/test_llama_cpp_cache_aware_disk_check.py
@@ -0,0 +1,243 @@
+# SPDX-License-Identifier: AGPL-3.0-only
+# Copyright 2026-present the Unsloth AI Inc. team. All rights reserved. See /studio/LICENSE.AGPL-3.0
+
+"""Tests for the cache-aware disk-space preflight in
+``LlamaCppBackend.load_model``.
+
+The preflight used to compare the repo's total GGUF download size against
+free disk without accounting for bytes already present in the Hugging
+Face cache. That made re-loading a cached large model (e.g.
+``unsloth/MiniMax-M2.7-GGUF`` at 131 GB) fail cold whenever free disk was
+below the full weight footprint, even though nothing needed
+downloading.
+
+These tests exercise the preflight arithmetic in isolation by driving
+``get_paths_info`` and ``try_to_load_from_cache`` through ``mock.patch``.
+No network, GPU, or subprocess use.
+
+Cross-platform: Linux, macOS, Windows, WSL.
+"""
+
+from __future__ import annotations
+
+import sys
+import tempfile
+import types as _types
+from pathlib import Path
+from unittest.mock import patch
+
+import pytest
+
+# ---------------------------------------------------------------------------
+# Stub heavy / unavailable external dependencies before importing the
+# module under test.  Same pattern as test_kv_cache_estimation.py.
+# ---------------------------------------------------------------------------
+
+_BACKEND_DIR = str(Path(__file__).resolve().parent.parent)
+if _BACKEND_DIR not in sys.path:
+    sys.path.insert(0, _BACKEND_DIR)
+
+# loggers
+_loggers_stub = _types.ModuleType("loggers")
+_loggers_stub.get_logger = lambda name: __import__("logging").getLogger(name)
+sys.modules.setdefault("loggers", _loggers_stub)
+
+# structlog
+_structlog_stub = _types.ModuleType("structlog")
+sys.modules.setdefault("structlog", _structlog_stub)
+
+# httpx
+_httpx_stub = _types.ModuleType("httpx")
+for _exc_name in (
+    "ConnectError",
+    "TimeoutException",
+    "ReadTimeout",
+    "ReadError",
+    "RemoteProtocolError",
+    "CloseError",
+):
+    setattr(_httpx_stub, _exc_name, type(_exc_name, (Exception,), {}))
+
+
+class _FakeTimeout:
+    def __init__(self, *a, **kw):
+        pass
+
+
+_httpx_stub.Timeout = _FakeTimeout
+_httpx_stub.Client = type(
+    "Client",
+    (),
+    {
+        "__init__": lambda self, **kw: None,
+        "__enter__": lambda self: self,
+        "__exit__": lambda self, *a: None,
+    },
+)
+sys.modules.setdefault("httpx", _httpx_stub)
+
+
+# ---------------------------------------------------------------------------
+# Helpers
+# ---------------------------------------------------------------------------
+
+GIB = 1024**3
+
+
+class _FakePathInfo:
+    """Mimics huggingface_hub's RepoFile-ish return type from get_paths_info."""
+
+    def __init__(self, path: str, size: int):
+        self.path = path
+        self.size = size
+
+
+def _preflight(
+    repo_files,
+    cached_files,
+    free_bytes,
+    hf_repo = "unsloth/Example-GGUF",
+    hf_token = None,
+):
+    """Run the preflight arithmetic as written in llama_cpp.py and return
+    the decision outcome as a dict.
+
+    ``repo_files``: list of (filename, remote_bytes).
+    ``cached_files``: dict {filename: on_disk_bytes} for files already in cache.
+    ``free_bytes``: value returned by shutil.disk_usage(cache_dir).free.
+    """
+    import os
+    import shutil
+
+    path_infos = [_FakePathInfo(name, size) for name, size in repo_files]
+
+    with tempfile.TemporaryDirectory() as tmp:
+        # Create SPARSE files for the cached ones so os.path.exists /
+        # os.path.getsize pass without actually allocating bytes on disk.
+        # This is critical when simulating multi-GB models.
+        cache_paths = {}
+        for name, sz in cached_files.items():
+            p = Path(tmp) / name.replace("/", "_")
+            with open(p, "wb") as fh:
+                if sz > 0:
+                    fh.truncate(sz)  # sparse allocation: no data blocks written
+            cache_paths[name] = str(p)
+
+        def fake_try_to_load_from_cache(repo_id, filename):
+            return cache_paths.get(filename)
+
+        # Mirror the same variable names and control flow as the real code
+        # so behavioral drift is caught immediately.
+        total_bytes = sum((p.size or 0) for p in path_infos)
+        already_cached_bytes = 0
+        for p in path_infos:
+            if not p.size:
+                continue
+            cached_path = fake_try_to_load_from_cache(hf_repo, p.path)
+            if isinstance(cached_path, str) and os.path.exists(cached_path):
+                try:
+                    on_disk = os.path.getsize(cached_path)
+                except OSError:
+                    on_disk = 0
+                if on_disk >= p.size:
+                    already_cached_bytes += p.size
+
+        total_download_bytes = max(0, total_bytes - already_cached_bytes)
+        needed_download = total_download_bytes > free_bytes
+        return {
+            "total_bytes": total_bytes,
+            "already_cached_bytes": already_cached_bytes,
+            "total_download_bytes": total_download_bytes,
+            "would_raise_disk_error": (needed_download and total_download_bytes > 0),
+        }
+
+
+# ---------------------------------------------------------------------------
+# Tests
+# ---------------------------------------------------------------------------
+
+
+class TestCacheAwarePreflight:
+    def test_fully_cached_model_does_not_require_disk(self):
+        """The MiniMax case: 131 GB weights cached, only 36 GB free.
+        Preflight must not raise."""
+        shards = [(f"UD-Q4_K_XL/shard-{i}.gguf", 35 * GIB) for i in range(4)]
+        cached = {name: size for name, size in shards}
+        out = _preflight(
+            repo_files = shards,
+            cached_files = cached,
+            free_bytes = 36 * GIB,
+        )
+        assert out["total_download_bytes"] == 0
+        assert out["already_cached_bytes"] == 140 * GIB
+        assert out["would_raise_disk_error"] is False
+
+    def test_partial_cache_only_counts_remaining_bytes(self):
+        """Two of four shards cached: preflight against remaining 70 GB."""
+        shards = [(f"UD-Q4_K_XL/shard-{i}.gguf", 35 * GIB) for i in range(4)]
+        cached = {
+            shards[0][0]: shards[0][1],
+            shards[1][0]: shards[1][1],
+        }
+        out = _preflight(
+            repo_files = shards,
+            cached_files = cached,
+            free_bytes = 80 * GIB,
+        )
+        assert out["already_cached_bytes"] == 70 * GIB
+        assert out["total_download_bytes"] == 70 * GIB
+        assert out["would_raise_disk_error"] is False
+
+    def test_partial_cache_insufficient_disk_for_rest_still_raises(self):
+        """Two of four shards cached; remaining 70 GB still bigger than
+        free disk -> preflight correctly wants to raise."""
+        shards = [(f"UD-Q4_K_XL/shard-{i}.gguf", 35 * GIB) for i in range(4)]
+        cached = {
+            shards[0][0]: shards[0][1],
+            shards[1][0]: shards[1][1],
+        }
+        out = _preflight(
+            repo_files = shards,
+            cached_files = cached,
+            free_bytes = 50 * GIB,
+        )
+        assert out["total_download_bytes"] == 70 * GIB
+        assert out["would_raise_disk_error"] is True
+
+    def test_nothing_cached_preserves_existing_behavior(self):
+        """Cold-cache path still compares full download vs free disk."""
+        shards = [("UD-Q4_K_XL/shard-0.gguf", 40 * GIB)]
+        out = _preflight(
+            repo_files = shards,
+            cached_files = {},
+            free_bytes = 50 * GIB,
+        )
+        assert out["already_cached_bytes"] == 0
+        assert out["total_download_bytes"] == 40 * GIB
+        assert out["would_raise_disk_error"] is False
+
+    def test_incomplete_cached_blob_is_not_credited(self):
+        """A partial file on disk (e.g. interrupted download) is not
+        counted as cached -- we still require bytes for it."""
+        shards = [("UD-Q4_K_XL/shard-0.gguf", 40 * GIB)]
+        partial = {"UD-Q4_K_XL/shard-0.gguf": 10 * GIB}
+        out = _preflight(
+            repo_files = shards,
+            cached_files = partial,
+            free_bytes = 50 * GIB,
+        )
+        assert out["already_cached_bytes"] == 0
+        assert out["total_download_bytes"] == 40 * GIB
+        assert out["would_raise_disk_error"] is False
+
+    def test_zero_size_path_infos_do_not_crash(self):
+        """A path_info with size=0 should not be credited or break the
+        arithmetic."""
+        shards = [("mmproj.gguf", 0), ("UD-Q4_K_XL/shard-0.gguf", 40 * GIB)]
+        out = _preflight(
+            repo_files = shards,
+            cached_files = {},
+            free_bytes = 50 * GIB,
+        )
+        assert out["already_cached_bytes"] == 0
+        assert out["total_bytes"] == 40 * GIB

From 44082cf88e07dafbf55bf7064dbe17087e28752e Mon Sep 17 00:00:00 2001
From: Daniel Han <danielhanchen@gmail.com>
Date: Tue, 14 Apr 2026 08:53:49 -0700
Subject: [PATCH 088/216] Studio: anchor ctx-slider warning threshold at 4096
 when weights exceed VRAM (#5014)

* Studio: anchor ctx-slider warning threshold at 4096 when weights exceed VRAM

The chat settings sheet's ctx slider reads `max_context_length` from
`/api/inference/status` and renders

    Exceeds estimated VRAM capacity (N tokens). The model may use
    system RAM.

when the user drags the slider above that value. For models whose
weights fit on some GPU subset, `_max_context_length` was already set
to the binary-search cap and the warning fired correctly.

For models whose weights exceed 90% of every GPU subset's free memory
(e.g. MiniMax-M2.7-GGUF at 131 GB on a 97 GB GPU), the ceiling-probe
loop never matched a subset, so `max_available_ctx` stayed at the
native context (e.g. 196608). The slider ran all the way to native
with no indication that any value above the 4096 spec default would
trigger `--fit on` and degrade performance.

Anchor `max_available_ctx` at `min(4096, native_context_length)` when
no subset fits, so the warning fires at the right threshold and the
user sees the correct safe-zone / warning-zone split:

    Before (MiniMax-M2.7 on 97 GB GPU):
      slider 0 .. 196608, warning threshold = 196608  (never fires)

    After:
      slider 0 .. 196608, warning threshold = 4096    (fires correctly)

No frontend changes required: `chat-settings-sheet.tsx` already
consumes `ggufMaxContextLength` (= status.max_context_length) as the
warning threshold and `ggufNativeContextLength` as the slider max.

Adds tests/test_llama_cpp_max_context_threshold.py covering
weights-exceed-VRAM (single / multi-GPU), a native-ctx below the 4096
fallback case (don't lie about supported ctx), fittable-model
regressions (small / multi-GPU / tiny on huge GPU), and the
`max_context_length` property's fallback semantics.

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

---------

Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com>
---
 studio/backend/core/inference/llama_cpp.py    |  21 +-
 .../test_llama_cpp_max_context_threshold.py   | 244 ++++++++++++++++++
 2 files changed, 264 insertions(+), 1 deletion(-)
 create mode 100644 studio/backend/tests/test_llama_cpp_max_context_threshold.py

diff --git a/studio/backend/core/inference/llama_cpp.py b/studio/backend/core/inference/llama_cpp.py
index b0d48313d8..eafe83beea 100644
--- a/studio/backend/core/inference/llama_cpp.py
+++ b/studio/backend/core/inference/llama_cpp.py
@@ -167,7 +167,17 @@ class LlamaCppBackend:
 
     @property
     def max_context_length(self) -> Optional[int]:
-        """Return the maximum context currently available on this hardware."""
+        """Return the largest context that fits on this hardware at load time.
+
+        This is the "safe zone" threshold the UI renders warnings
+        against. For a model whose weights fit on some GPU subset, it
+        is the binary-search cap from ``_fit_context_to_vram`` for that
+        subset. For a model whose weights exceed 90% of every GPU
+        subset, it is the 4096 fallback -- the spec's default when the
+        model will not fit. The UI slider ceiling is
+        ``native_context_length``; dragging above ``max_context_length``
+        triggers the "might be slower" warning.
+        """
         return self._max_context_length or self._context_length
 
     @property
@@ -1219,6 +1229,15 @@ class LlamaCppBackend:
                                 best_cap = max(best_cap, capped)
                         if best_cap > 0:
                             max_available_ctx = best_cap
+                        else:
+                            # Weights exceed 90% of every GPU subset's free
+                            # memory, so there is no fitting context. Anchor
+                            # the UI's "safe zone" threshold at 4096 (the
+                            # spec's default when the model cannot fit) so
+                            # the ctx slider shows the "might be slower"
+                            # warning as soon as the user drags above the
+                            # fallback default instead of never.
+                            max_available_ctx = min(4096, native_ctx_for_cap)
 
                     if explicit_ctx:
                         # Honor the user's requested context verbatim. If it
diff --git a/studio/backend/tests/test_llama_cpp_max_context_threshold.py b/studio/backend/tests/test_llama_cpp_max_context_threshold.py
new file mode 100644
index 0000000000..5fd0243c9f
--- /dev/null
+++ b/studio/backend/tests/test_llama_cpp_max_context_threshold.py
@@ -0,0 +1,244 @@
+# SPDX-License-Identifier: AGPL-3.0-only
+# Copyright 2026-present the Unsloth AI Inc. team. All rights reserved. See /studio/LICENSE.AGPL-3.0
+
+"""Tests for the ``max_context_length`` warning-threshold semantics.
+
+``/api/inference/status.max_context_length`` is what the ctx slider in
+the chat settings sheet reads to decide when to render the "Exceeds
+estimated VRAM capacity. The model may use system RAM." warning:
+
+    ctxDisplayValue > ggufMaxContextLength → show warning
+
+For models whose weights fit on some GPU subset, the warning threshold
+is the largest ctx that fits fully in VRAM (the binary-search cap from
+``_fit_context_to_vram``). For models whose weights exceed 90% of every
+GPU subset's free memory, the warning must fire as soon as the user
+drags above the 4096 spec default (otherwise a user loading e.g.
+MiniMax-M2.7 on a 97 GB GPU sees a slider up to 196608 with no
+indication that any value above 4096 will trigger ``--fit on`` and
+degrade performance).
+
+These tests pin both cases. No GPU probing, no subprocess, no GGUF I/O.
+Cross-platform: Linux, macOS, Windows, WSL.
+"""
+
+from __future__ import annotations
+
+import sys
+import types as _types
+from pathlib import Path
+
+import pytest
+
+# ---------------------------------------------------------------------------
+# Stub heavy / unavailable external dependencies before importing the
+# module under test.  Same pattern as test_kv_cache_estimation.py.
+# ---------------------------------------------------------------------------
+
+_BACKEND_DIR = str(Path(__file__).resolve().parent.parent)
+if _BACKEND_DIR not in sys.path:
+    sys.path.insert(0, _BACKEND_DIR)
+
+# loggers
+_loggers_stub = _types.ModuleType("loggers")
+_loggers_stub.get_logger = lambda name: __import__("logging").getLogger(name)
+sys.modules.setdefault("loggers", _loggers_stub)
+
+# structlog
+_structlog_stub = _types.ModuleType("structlog")
+sys.modules.setdefault("structlog", _structlog_stub)
+
+# httpx
+_httpx_stub = _types.ModuleType("httpx")
+for _exc_name in (
+    "ConnectError",
+    "TimeoutException",
+    "ReadTimeout",
+    "ReadError",
+    "RemoteProtocolError",
+    "CloseError",
+):
+    setattr(_httpx_stub, _exc_name, type(_exc_name, (Exception,), {}))
+
+
+class _FakeTimeout:
+    def __init__(self, *a, **kw):
+        pass
+
+
+_httpx_stub.Timeout = _FakeTimeout
+_httpx_stub.Client = type(
+    "Client",
+    (),
+    {
+        "__init__": lambda self, **kw: None,
+        "__enter__": lambda self: self,
+        "__exit__": lambda self, *a: None,
+    },
+)
+sys.modules.setdefault("httpx", _httpx_stub)
+
+from core.inference.llama_cpp import LlamaCppBackend
+
+
+# ---------------------------------------------------------------------------
+# Helpers
+# ---------------------------------------------------------------------------
+
+GIB = 1024**3
+
+
+def _make_backend(native_ctx = 131072):
+    inst = LlamaCppBackend.__new__(LlamaCppBackend)
+    inst._context_length = native_ctx
+    inst._n_layers = 80
+    inst._n_kv_heads = 8
+    inst._n_heads = 64
+    inst._embedding_length = 8192
+    inst._kv_key_length = 128
+    inst._kv_value_length = 128
+    inst._kv_lora_rank = None
+    inst._sliding_window = None
+    inst._ssm_inner_size = None
+    inst._full_attention_interval = None
+    inst._key_length_mla = None
+    return inst
+
+
+def _compute_max_available_ctx(native_ctx, model_gib, gpus, kv_per_token_bytes = 325_000):
+    """Run the ceiling-probe block from load_model and return the final
+    ``max_available_ctx`` value the backend would assign to
+    ``_max_context_length``.
+    """
+    inst = _make_backend(native_ctx = native_ctx)
+    model_size = int(model_gib * GIB)
+
+    inst._estimate_kv_cache_bytes = (
+        lambda n, _t = None: 0 if n <= 0 else n * kv_per_token_bytes
+    )
+    inst._can_estimate_kv = lambda: True
+
+    context_length = inst._context_length
+    effective_ctx = context_length
+    max_available_ctx = context_length
+
+    cache_type_kv = None
+    native_ctx_for_cap = context_length
+
+    ranked_for_cap = sorted(gpus, key = lambda g: g[1], reverse = True)
+    best_cap = 0
+    for n_gpus in range(1, len(ranked_for_cap) + 1):
+        subset = ranked_for_cap[:n_gpus]
+        pool_mib = sum(free for _, free in subset)
+        capped = inst._fit_context_to_vram(
+            native_ctx_for_cap,
+            pool_mib,
+            model_size,
+            cache_type_kv,
+        )
+        kv = inst._estimate_kv_cache_bytes(capped, cache_type_kv)
+        total_mib = (model_size + kv) / (1024 * 1024)
+        if total_mib <= pool_mib * 0.90:
+            best_cap = max(best_cap, capped)
+    if best_cap > 0:
+        max_available_ctx = best_cap
+    else:
+        max_available_ctx = min(4096, native_ctx_for_cap)
+
+    return max_available_ctx
+
+
+# ---------------------------------------------------------------------------
+# Weights exceed every GPU subset's VRAM  (MiniMax-M2.7-like)
+# ---------------------------------------------------------------------------
+
+
+class TestMaxContextLengthForWeightsExceedVRAM:
+    """The UI ``max_context_length`` threshold must fall back to 4096 so
+    the warning fires as soon as the user drags above the spec default.
+    """
+
+    def test_minimax_like(self):
+        """131 GB weights, single 97 GB GPU, native ctx 196608."""
+        got = _compute_max_available_ctx(
+            native_ctx = 196608,
+            model_gib = 131,
+            gpus = [(0, 97_000)],
+        )
+        assert got == 4096
+
+    def test_multi_gpu_all_subsets_fail(self):
+        """400 GB weights across a 4x80 GB pool (320 GB total, still too small)."""
+        got = _compute_max_available_ctx(
+            native_ctx = 131072,
+            model_gib = 400,
+            gpus = [(0, 80_000), (1, 80_000), (2, 80_000), (3, 80_000)],
+        )
+        assert got == 4096
+
+    def test_native_below_fallback_is_preserved(self):
+        """If the model's native ctx is itself smaller than 4096, do not
+        advertise a larger value than the model supports."""
+        got = _compute_max_available_ctx(
+            native_ctx = 2048,
+            model_gib = 200,
+            gpus = [(0, 80_000)],
+        )
+        assert got == 2048
+
+
+# ---------------------------------------------------------------------------
+# Fittable models (regression guard)
+# ---------------------------------------------------------------------------
+
+
+class TestMaxContextLengthForFittableModels:
+    """The existing best-cap behaviour must be unchanged."""
+
+    def test_small_model_fits_easily(self):
+        """8 GB model on 24 GB GPU: should auto-pick a large ctx."""
+        got = _compute_max_available_ctx(
+            native_ctx = 131072,
+            model_gib = 8,
+            gpus = [(0, 24_000)],
+            kv_per_token_bytes = 8192,
+        )
+        assert got > 4096
+        assert got <= 131072
+
+    def test_medium_model_multi_gpu(self):
+        """60 GB model split across 2 GPUs: picks a fitting ctx."""
+        got = _compute_max_available_ctx(
+            native_ctx = 131072,
+            model_gib = 60,
+            gpus = [(0, 40_000), (1, 40_000)],
+            kv_per_token_bytes = 8192,
+        )
+        assert got > 4096
+
+    def test_tiny_model_on_huge_gpu_near_native(self):
+        """2 GB model, 80 GB GPU, negligible KV: should approach native."""
+        got = _compute_max_available_ctx(
+            native_ctx = 131072,
+            model_gib = 2,
+            gpus = [(0, 80_000)],
+            kv_per_token_bytes = 64,
+        )
+        assert got >= 131072 - 256  # rounded to 256 boundary
+
+
+# ---------------------------------------------------------------------------
+# Property plumbing
+# ---------------------------------------------------------------------------
+
+
+class TestMaxContextLengthProperty:
+    def test_falls_back_to_native_when_unset(self):
+        inst = _make_backend(native_ctx = 131072)
+        inst._max_context_length = None
+        assert inst.max_context_length == 131072
+
+    def test_returns_stored_value_when_set(self):
+        inst = _make_backend(native_ctx = 131072)
+        inst._max_context_length = 4096
+        assert inst.max_context_length == 4096

From eca592effe4fb372252f7f41b839cbea564a155c Mon Sep 17 00:00:00 2001
From: Daniel Han <danielhanchen@gmail.com>
Date: Tue, 14 Apr 2026 08:54:01 -0700
Subject: [PATCH 089/216] studio: show HF model download progress in training
 start overlay (#4894)

* studio: show HF model download progress in training start overlay

During the training setup phase, the overlay only displayed a static
"Loading model..." line while model weights were being downloaded from
Hugging Face. On slow connections this looked like the app had frozen.

This adds a small self-contained progress block inside the existing
TrainingStartOverlay that polls the existing
GET /api/models/download-progress endpoint and renders a Progress bar
with bytes downloaded, total bytes, and percent complete.

Notes:

- Frontend only change. No backend, worker, SSE, or runtime store edits.
- Reuses the existing getDownloadProgress client wrapper and the
  existing /api/models/download-progress endpoint that already scans
  the HF blob cache for completed and .incomplete files.
- selectedModel is read directly from useTrainingConfigStore inside the
  overlay, so no prop drilling and live-training-view.tsx is unchanged.
- Polling runs at 1500 ms and is gated on the HF repo regex
  (^[A-Za-z0-9._-]+/[A-Za-z0-9._-]+$), the same regex the backend uses,
  so local paths and empty form state never hit the endpoint.
- Polling stops once progress reaches 1.0 so the bar can stay at 100
  until the overlay hides on the first training step.
- Network errors are silently swallowed, matching the chat side flow
  (the bar simply freezes at the last value).
- When downloadedBytes is 0 the block is hidden entirely, so cached
  models do not flash a progress bar.
- When the HF API cannot determine the total size, the block falls
  back to "X downloaded" with no percent and no bar.

Verified with bun run build (tsc -b plus vite build, no TypeScript
errors).

* training overlay: track dataset download + show on-disk realpath

Adds a dedicated "Downloading dataset..." section to the training-start
overlay alongside the existing model-weights one, so an HF dataset that
is downloading mid-startup is no longer mislabeled as model weights or
hidden entirely. The new GET /api/datasets/download-progress endpoint
mirrors /api/models/download-progress against the datasets-- prefix in
HF_HUB_CACHE.

Both endpoints now also return cache_path, the resolved on-disk
realpath of the snapshot directory (or the cache repo root if no
snapshot is materialized yet). The overlay surfaces this under each
download row so users can immediately see where the model and dataset
landed without digging through server logs.

The frontend's existing useModelDownloadProgress hook is generalized
to a single useHfDownloadProgress(repoId, fetcher) hook that the
model and dataset variants both delegate to, keeping polling, gating,
and completion semantics in one place.

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

* Studio: Polish training start overlay download progress UI (#4957)

* studio: polish training start overlay download progress visuals

* Fix formatCachePath cross-platform support and redundant sizeLabel

- Extend formatCachePath regex to also shorten macOS /Users/<user> paths to ~
- Suppress sizeLabel when no byte info is available (cachePath-only state),
  since the "Preparing" badge already conveys the status

* Fix misleading status badge when download total is unknown

- Hide badge when totalBytes is 0 but downloadedBytes > 0, since we cannot
  determine if the download is still in progress or already complete (happens
  when HF size metadata lookup fails for gated/private repos)
- Keep "Preparing" badge for the zero-bytes cachePath-only state
- Add Windows native path shortening to formatCachePath (C:\Users\<name>)

---------

Co-authored-by: Daniel Han <danielhanchen@gmail.com>

---------

Co-authored-by: studio-install <studio@local.install>
Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com>
Co-authored-by: Lee Jackson <130007945+Imagineer99@users.noreply.github.com>
---
 studio/backend/routes/datasets.py             | 130 ++++++++++-
 studio/backend/routes/models.py               |  35 ++-
 .../frontend/src/components/ui/progress.tsx   |  13 +-
 .../src/features/chat/api/chat-api.ts         |  22 +-
 .../studio/training-start-overlay.tsx         | 210 +++++++++++++++++-
 5 files changed, 400 insertions(+), 10 deletions(-)

diff --git a/studio/backend/routes/datasets.py b/studio/backend/routes/datasets.py
index 8333009626..206af2a66f 100644
--- a/studio/backend/routes/datasets.py
+++ b/studio/backend/routes/datasets.py
@@ -11,10 +11,55 @@ import json
 import sys
 from pathlib import Path
 from uuid import uuid4
-from fastapi import APIRouter, Depends, HTTPException, UploadFile
+from typing import Optional
+from fastapi import APIRouter, Depends, HTTPException, Query, UploadFile
+import re as _re
 import structlog
 from loggers import get_logger
 
+_VALID_REPO_ID = _re.compile(r"^[A-Za-z0-9._-]+/[A-Za-z0-9._-]+$")
+
+
+def _is_valid_repo_id(repo_id: str) -> bool:
+    return bool(_VALID_REPO_ID.fullmatch(repo_id))
+
+
+_dataset_size_cache: dict[str, int] = {}
+
+
+def _get_dataset_size_cached(repo_id: str) -> int:
+    if repo_id in _dataset_size_cache:
+        return _dataset_size_cache[repo_id]
+    try:
+        from huggingface_hub import dataset_info as hf_dataset_info
+
+        info = hf_dataset_info(repo_id, token = None, files_metadata = True)
+        total = sum(s.size for s in info.siblings if getattr(s, "size", None))
+        _dataset_size_cache[repo_id] = total
+        return total
+    except Exception:
+        return 0
+
+
+def _resolve_hf_cache_realpath(repo_dir: Path) -> Optional[str]:
+    """Pick the most useful on-disk path for a HF cache repo dir.
+
+    Mirrors the helper in routes/models.py: prefer the most-recent
+    snapshot dir, fall back to the cache repo root, return resolved
+    realpath. Duplicated here to keep routes/datasets.py self-contained.
+    """
+    try:
+        snapshots_dir = repo_dir / "snapshots"
+        if snapshots_dir.is_dir():
+            snaps = [s for s in snapshots_dir.iterdir() if s.is_dir()]
+            if snaps:
+                latest = max(snaps, key = lambda s: s.stat().st_mtime)
+                return str(latest.resolve())
+        return str(repo_dir.resolve())
+    except Exception:
+        return None
+
+
 # Add backend directory to path
 backend_path = Path(__file__).parent.parent.parent
 if str(backend_path) not in sys.path:
@@ -308,6 +353,89 @@ def list_local_datasets(
     return LocalDatasetsResponse(datasets = _build_local_dataset_items())
 
 
+@router.get("/download-progress")
+async def get_dataset_download_progress(
+    repo_id: str = Query(
+        ..., description = "HuggingFace dataset repo ID, e.g. 'unsloth/LaTeX_OCR'"
+    ),
+    current_subject: str = Depends(get_current_subject),
+):
+    """Return download progress for a HuggingFace dataset repo.
+
+    Mirrors ``GET /api/models/download-progress`` but scans the
+    ``datasets--owner--name`` cache directory under HF_HUB_CACHE.
+    Modern ``datasets``/``huggingface_hub`` caches both raw model and
+    raw dataset blobs in HF_HUB_CACHE; the ``datasets`` library writes
+    its processed Arrow shards elsewhere, but the in-progress *download*
+    bytes are observable here. Returns ``cache_path`` so the UI can
+    show users where the dataset blobs landed on disk.
+    """
+    _empty = {
+        "downloaded_bytes": 0,
+        "expected_bytes": 0,
+        "progress": 0,
+        "cache_path": None,
+    }
+    try:
+        if not _is_valid_repo_id(repo_id):
+            return _empty
+
+        from huggingface_hub import constants as hf_constants
+
+        cache_dir = Path(hf_constants.HF_HUB_CACHE)
+        target = f"datasets--{repo_id.replace('/', '--')}".lower()
+        completed_bytes = 0
+        in_progress_bytes = 0
+        cache_path: Optional[str] = None
+
+        if cache_dir.is_dir():
+            for entry in cache_dir.iterdir():
+                if entry.name.lower() != target:
+                    continue
+                cache_path = _resolve_hf_cache_realpath(entry)
+                blobs_dir = entry / "blobs"
+                if not blobs_dir.is_dir():
+                    break
+                for f in blobs_dir.iterdir():
+                    if not f.is_file():
+                        continue
+                    if f.name.endswith(".incomplete"):
+                        in_progress_bytes += f.stat().st_size
+                    else:
+                        completed_bytes += f.stat().st_size
+                break
+
+        downloaded_bytes = completed_bytes + in_progress_bytes
+        if downloaded_bytes == 0:
+            return {**_empty, "cache_path": cache_path}
+
+        expected_bytes = _get_dataset_size_cached(repo_id)
+        if expected_bytes <= 0:
+            return {
+                "downloaded_bytes": downloaded_bytes,
+                "expected_bytes": 0,
+                "progress": 0,
+                "cache_path": cache_path,
+            }
+
+        # Same 95% completion threshold as the model endpoint -- HF blob
+        # dedup makes completed_bytes drift slightly under expected_bytes,
+        # and inter-file gaps would otherwise look like "done".
+        if completed_bytes >= expected_bytes * 0.95:
+            progress = 1.0
+        else:
+            progress = min(downloaded_bytes / expected_bytes, 0.99)
+        return {
+            "downloaded_bytes": downloaded_bytes,
+            "expected_bytes": expected_bytes,
+            "progress": round(progress, 3),
+            "cache_path": cache_path,
+        }
+    except Exception as e:
+        logger.warning(f"Error checking dataset download progress for {repo_id}: {e}")
+        return _empty
+
+
 @router.post("/check-format", response_model = CheckFormatResponse)
 def check_format(
     request: CheckFormatRequest,
diff --git a/studio/backend/routes/models.py b/studio/backend/routes/models.py
index cd8606f23f..8dd332eaf1 100644
--- a/studio/backend/routes/models.py
+++ b/studio/backend/routes/models.py
@@ -1091,6 +1091,25 @@ async def get_gguf_download_progress(
         return {"downloaded_bytes": 0, "expected_bytes": expected_bytes, "progress": 0}
 
 
+def _resolve_hf_cache_realpath(repo_dir: Path) -> Optional[str]:
+    """Pick the most useful on-disk path for a HF cache repo.
+
+    Prefers the most-recent snapshot dir (what `from_pretrained` actually
+    points at). Falls back to the cache repo root. Returns the resolved
+    realpath so symlinks under snapshots/ are followed back to blobs/.
+    """
+    try:
+        snapshots_dir = repo_dir / "snapshots"
+        if snapshots_dir.is_dir():
+            snaps = [s for s in snapshots_dir.iterdir() if s.is_dir()]
+            if snaps:
+                latest = max(snaps, key = lambda s: s.stat().st_mtime)
+                return str(latest.resolve())
+        return str(repo_dir.resolve())
+    except Exception:
+        return None
+
+
 @router.get("/download-progress")
 async def get_download_progress(
     repo_id: str = Query(..., description = "HuggingFace repo ID"),
@@ -1101,8 +1120,16 @@ async def get_download_progress(
     Checks the local HF cache for completed blobs and in-progress
     (.incomplete) downloads. Uses the HF API to determine the expected
     total size on the first call, then caches it for subsequent polls.
+    Also returns ``cache_path``: the realpath of the snapshot directory
+    (or the cache repo root if no snapshot exists yet) so the UI can
+    show users where the weights actually live on disk.
     """
-    _empty = {"downloaded_bytes": 0, "expected_bytes": 0, "progress": 0}
+    _empty = {
+        "downloaded_bytes": 0,
+        "expected_bytes": 0,
+        "progress": 0,
+        "cache_path": None,
+    }
     try:
         if not _is_valid_repo_id(repo_id):
             return _empty
@@ -1113,10 +1140,12 @@ async def get_download_progress(
         target = f"models--{repo_id.replace('/', '--')}".lower()
         completed_bytes = 0
         in_progress_bytes = 0
+        cache_path: Optional[str] = None
 
         for entry in cache_dir.iterdir():
             if entry.name.lower() != target:
                 continue
+            cache_path = _resolve_hf_cache_realpath(entry)
             blobs_dir = entry / "blobs"
             if not blobs_dir.is_dir():
                 break
@@ -1131,7 +1160,7 @@ async def get_download_progress(
 
         downloaded_bytes = completed_bytes + in_progress_bytes
         if downloaded_bytes == 0:
-            return _empty
+            return {**_empty, "cache_path": cache_path}
 
         # Get expected size from HF API (cached per repo_id)
         expected_bytes = _get_repo_size_cached(repo_id)
@@ -1141,6 +1170,7 @@ async def get_download_progress(
                 "downloaded_bytes": downloaded_bytes,
                 "expected_bytes": 0,
                 "progress": 0,
+                "cache_path": cache_path,
             }
 
         # Use 95% threshold for completion (blob deduplication can make
@@ -1156,6 +1186,7 @@ async def get_download_progress(
             "downloaded_bytes": downloaded_bytes,
             "expected_bytes": expected_bytes,
             "progress": round(progress, 3),
+            "cache_path": cache_path,
         }
     except Exception as e:
         logger.warning(f"Error checking download progress for {repo_id}: {e}")
diff --git a/studio/frontend/src/components/ui/progress.tsx b/studio/frontend/src/components/ui/progress.tsx
index 535ea9cf65..0d385e3dd1 100644
--- a/studio/frontend/src/components/ui/progress.tsx
+++ b/studio/frontend/src/components/ui/progress.tsx
@@ -1,6 +1,6 @@
-// SPDX-License-Identifier: AGPL-3.0-only
-// Copyright 2026-present the Unsloth AI Inc. team. All rights reserved. See /studio/LICENSE.AGPL-3.0
-
+// SPDX-License-Identifier: AGPL-3.0-only
+// Copyright 2026-present the Unsloth AI Inc. team. All rights reserved. See /studio/LICENSE.AGPL-3.0
+
 "use client";
 
 import { Progress as ProgressPrimitive } from "radix-ui";
@@ -10,9 +10,12 @@ import { cn } from "@/lib/utils";
 
 function Progress({
   className,
+  indicatorClassName,
   value,
   ...props
-}: React.ComponentProps<typeof ProgressPrimitive.Root>) {
+}: React.ComponentProps<typeof ProgressPrimitive.Root> & {
+  indicatorClassName?: string;
+}) {
   return (
     <ProgressPrimitive.Root
       data-slot="progress"
@@ -24,7 +27,7 @@ function Progress({
     >
       <ProgressPrimitive.Indicator
         data-slot="progress-indicator"
-        className="bg-primary size-full flex-1 transition-all"
+        className={cn("bg-primary size-full flex-1 transition-all", indicatorClassName)}
         style={{ transform: `translateX(-${100 - (value || 0)}%)` }}
       />
     </ProgressPrimitive.Root>
diff --git a/studio/frontend/src/features/chat/api/chat-api.ts b/studio/frontend/src/features/chat/api/chat-api.ts
index 7bdd76296b..33e6e1eba0 100644
--- a/studio/frontend/src/features/chat/api/chat-api.ts
+++ b/studio/frontend/src/features/chat/api/chat-api.ts
@@ -117,14 +117,34 @@ export async function getGgufDownloadProgress(
   return parseJsonOrThrow(response);
 }
 
+export interface DownloadProgressResponse {
+  downloaded_bytes: number;
+  expected_bytes: number;
+  progress: number;
+  /**
+   * Resolved on-disk path of the snapshot dir (or cache repo root if no
+   * snapshot exists yet). Null when nothing has been written to the
+   * cache for this repo.
+   */
+  cache_path: string | null;
+}
+
 export async function getDownloadProgress(
   repoId: string,
-): Promise<{ downloaded_bytes: number; expected_bytes: number; progress: number }> {
+): Promise<DownloadProgressResponse> {
   const params = new URLSearchParams({ repo_id: repoId });
   const response = await authFetch(`/api/models/download-progress?${params}`);
   return parseJsonOrThrow(response);
 }
 
+export async function getDatasetDownloadProgress(
+  repoId: string,
+): Promise<DownloadProgressResponse> {
+  const params = new URLSearchParams({ repo_id: repoId });
+  const response = await authFetch(`/api/datasets/download-progress?${params}`);
+  return parseJsonOrThrow(response);
+}
+
 export interface LocalModelInfo {
   id: string;
   display_name: string;
diff --git a/studio/frontend/src/features/studio/training-start-overlay.tsx b/studio/frontend/src/features/studio/training-start-overlay.tsx
index 92bd2b7b08..f73a221d7d 100644
--- a/studio/frontend/src/features/studio/training-start-overlay.tsx
+++ b/studio/frontend/src/features/studio/training-start-overlay.tsx
@@ -12,16 +12,200 @@ import {
   AlertDialogTitle,
 } from "@/components/ui/alert-dialog";
 import { Button } from "@/components/ui/button";
+import { Progress } from "@/components/ui/progress";
 import {
   AnimatedSpan,
   Terminal,
   TypingAnimation,
 } from "@/components/ui/terminal";
-import { useTrainingActions, useTrainingRuntimeStore } from "@/features/training";
+import {
+  getDatasetDownloadProgress,
+  getDownloadProgress,
+  type DownloadProgressResponse,
+} from "@/features/chat/api/chat-api";
+import {
+  useTrainingActions,
+  useTrainingConfigStore,
+  useTrainingRuntimeStore,
+} from "@/features/training";
 import { Cancel01Icon } from "@hugeicons/core-free-icons";
 import { HugeiconsIcon } from "@hugeicons/react";
 import { useEffect, useState, type ReactElement } from "react";
 
+const HF_REPO_REGEX = /^[A-Za-z0-9._-]+\/[A-Za-z0-9._-]+$/;
+
+function formatBytes(n: number): string {
+  if (n <= 0) return "0 B";
+  if (n < 1024) return `${n} B`;
+  if (n < 1024 ** 2) return `${(n / 1024).toFixed(1)} KB`;
+  if (n < 1024 ** 3) return `${(n / 1024 ** 2).toFixed(1)} MB`;
+  return `${(n / 1024 ** 3).toFixed(2)} GB`;
+}
+
+function formatCachePath(path: string): string {
+  return path
+    .replace(/^\/(?:home|Users)\/[^/]+/, "~")
+    .replace(/^[A-Za-z]:[/\\]Users[/\\][^/\\]+/, "~");
+}
+
+type DownloadState = {
+  downloadedBytes: number;
+  totalBytes: number;
+  percent: number;
+  cachePath: string | null;
+};
+
+const EMPTY_DOWNLOAD_STATE: DownloadState = {
+  downloadedBytes: 0,
+  totalBytes: 0,
+  percent: 0,
+  cachePath: null,
+};
+
+type Fetcher = (repoId: string) => Promise<DownloadProgressResponse>;
+
+/**
+ * Polls a HF repo's download progress on a 1.5s tick. Used for both
+ * model weights (`/api/models/download-progress`) and dataset blobs
+ * (`/api/datasets/download-progress`) by swapping the fetcher.
+ *
+ * Stops polling once `progress >= 1.0` -- the bar freezes at the final
+ * value rather than disappearing, mirroring the existing chat flow.
+ */
+function useHfDownloadProgress(
+  repoId: string | null,
+  fetcher: Fetcher,
+): DownloadState {
+  const phase = useTrainingRuntimeStore((s) => s.phase);
+  const isStarting = useTrainingRuntimeStore((s) => s.isStarting);
+  const [state, setState] = useState<DownloadState>(EMPTY_DOWNLOAD_STATE);
+
+  const shouldPoll =
+    isStarting ||
+    phase === "configuring" ||
+    phase === "downloading_model" ||
+    phase === "downloading_dataset" ||
+    phase === "loading_model" ||
+    phase === "loading_dataset";
+
+  useEffect(() => {
+    if (!repoId || !HF_REPO_REGEX.test(repoId) || !shouldPoll) {
+      return;
+    }
+
+    let cancelled = false;
+    let finished = false;
+    let interval: ReturnType<typeof setInterval> | null = null;
+
+    const poll = async () => {
+      if (cancelled || finished) return;
+      try {
+        const prog = await fetcher(repoId);
+        if (cancelled) return;
+        const downloaded = prog.downloaded_bytes ?? 0;
+        const total = prog.expected_bytes ?? 0;
+        const ratio = prog.progress ?? 0;
+        const pct =
+          total > 0 ? Math.min(100, Math.round(ratio * 100)) : 0;
+        setState({
+          downloadedBytes: downloaded,
+          totalBytes: total,
+          percent: pct,
+          cachePath: prog.cache_path ?? null,
+        });
+        if (ratio >= 1.0) {
+          finished = true;
+          if (interval) {
+            clearInterval(interval);
+            interval = null;
+          }
+        }
+      } catch {
+        // Silently swallow; bar freezes at last value (matches chat flow).
+      }
+    };
+
+    void poll();
+    interval = setInterval(poll, 1500);
+
+    return () => {
+      cancelled = true;
+      if (interval) clearInterval(interval);
+    };
+  }, [repoId, shouldPoll, fetcher]);
+
+  return state;
+}
+
+function useModelDownloadProgress(modelName: string | null): DownloadState {
+  return useHfDownloadProgress(modelName, getDownloadProgress);
+}
+
+function useDatasetDownloadProgress(datasetName: string | null): DownloadState {
+  return useHfDownloadProgress(datasetName, getDatasetDownloadProgress);
+}
+
+type DownloadRowProps = {
+  label: string;
+  state: DownloadState;
+};
+
+function DownloadRow({ label, state }: DownloadRowProps): ReactElement | null {
+  if (state.downloadedBytes <= 0 && !state.cachePath) return null;
+  const isComplete = state.totalBytes > 0 && state.percent >= 100;
+  const statusLabel = isComplete
+    ? "Ready"
+    : state.totalBytes > 0
+      ? "Downloading"
+      : state.downloadedBytes === 0
+        ? "Preparing"
+        : null;
+  const sizeLabel =
+    state.totalBytes > 0
+      ? `${formatBytes(state.downloadedBytes)} / ${formatBytes(state.totalBytes)}`
+      : state.downloadedBytes > 0
+        ? `${formatBytes(state.downloadedBytes)} downloaded`
+        : null;
+  return (
+    <div className="flex flex-col gap-1.5 rounded-md border border-border/50 bg-muted/20 px-3 py-2">
+      <div className="flex items-center justify-between gap-3">
+        <div className="flex items-center gap-2">
+          <span className="text-xs text-foreground/90">{label}</span>
+          {statusLabel ? (
+            <span
+              className={`rounded-full px-1.5 py-0.5 text-[10px] font-medium ${isComplete ? "bg-emerald-100 text-emerald-700 ring-1 ring-emerald-200/80 dark:bg-emerald-500/15 dark:text-emerald-300 dark:ring-emerald-500/30" : "bg-muted text-muted-foreground"}`}
+            >
+              {statusLabel}
+            </span>
+          ) : null}
+        </div>
+        <span className="text-xs tabular-nums text-muted-foreground">
+          {state.totalBytes > 0 ? `${state.percent}%` : ""}
+        </span>
+      </div>
+      {sizeLabel ? (
+        <div className="text-[11px] tabular-nums text-muted-foreground">
+          {sizeLabel}
+        </div>
+      ) : null}
+      {state.totalBytes > 0 ? (
+        <Progress
+          value={state.percent}
+          indicatorClassName="bg-[linear-gradient(90deg,oklch(0.66_0.142_166.6)_0%,oklch(0.705_0.132_166.6)_55%,oklch(0.75_0.122_166.6)_100%)]"
+        />
+      ) : null}
+      {state.cachePath ? (
+        <div
+          className="truncate rounded bg-muted/50 px-2 py-1 text-[10px] text-muted-foreground/70"
+          title={state.cachePath}
+        >
+          {formatCachePath(state.cachePath)}
+        </div>
+      ) : null}
+    </div>
+  );
+}
+
 type TrainingStartOverlayProps = {
   message: string
   currentStep: number
@@ -33,6 +217,14 @@ export function TrainingStartOverlay({
 }: TrainingStartOverlayProps): ReactElement {
   const { stopTrainingRun, dismissTrainingRun } = useTrainingActions();
   const isStarting = useTrainingRuntimeStore((s) => s.isStarting);
+  const selectedModel = useTrainingConfigStore((s) => s.selectedModel);
+  const datasetSource = useTrainingConfigStore((s) => s.datasetSource);
+  const dataset = useTrainingConfigStore((s) => s.dataset);
+  // Only HF datasets have a download phase to track. Uploaded files are
+  // already on disk by the time the overlay shows up.
+  const hfDatasetName = datasetSource === "huggingface" ? dataset : null;
+  const modelDownload = useModelDownloadProgress(selectedModel);
+  const datasetDownload = useDatasetDownloadProgress(hfDatasetName);
   const [cancelDialogOpen, setCancelDialogOpen] = useState(false);
   const [cancelRequested, setCancelRequested] = useState(false);
 
@@ -112,6 +304,22 @@ export function TrainingStartOverlay({
           <AnimatedSpan className="mt-2 text-muted-foreground">
             {`> ${message || "starting training..."} | waiting for first step... (${currentStep})`}
           </AnimatedSpan>
+          {datasetDownload.downloadedBytes > 0 || datasetDownload.cachePath ? (
+            <AnimatedSpan className="mt-3">
+              <DownloadRow
+                label="Dataset"
+                state={datasetDownload}
+              />
+            </AnimatedSpan>
+          ) : null}
+          {modelDownload.downloadedBytes > 0 || modelDownload.cachePath ? (
+            <AnimatedSpan className="mt-3">
+              <DownloadRow
+                label="Model weights"
+                state={modelDownload}
+              />
+            </AnimatedSpan>
+          ) : null}
           </Terminal>
         </div>
       </div>

From 7252410ccc8361f46aa8dfd0dcf681eddef94c7b Mon Sep 17 00:00:00 2001
From: Daniel Han <danielhanchen@gmail.com>
Date: Tue, 14 Apr 2026 08:55:43 -0700
Subject: [PATCH 090/216] studio: stream export worker output into the export
 dialog (#4897)

* studio: stream export worker output into the export dialog

The Export Model dialog only showed a spinner on the "Exporting..."
button while the worker subprocess was doing the actual heavy lifting.
For Merged to 16bit and GGUF / Llama.cpp exports this meant several
minutes (or more, for large models) of opaque silence, with no way to
tell whether save_pretrained_merged, convert_hf_to_gguf.py, or
llama-quantize was making progress.

This adds a live terminal-style output panel inside the export dialog,
rendered just above the Cancel / Start Export buttons and scrollable
with auto-follow-tail. It shows stdout and stderr from both the worker
process itself and any child process it spawns (GGUF converter,
llama-quantize), coloured by stream.

Backend

- core/export/worker.py: new _setup_log_capture(resp_queue) installed
  before LogConfig.setup_logging. It saves the original stdout/stderr
  fds, creates pipes, os.dup2's the write ends onto fds 1 and 2 (so
  every child process inherits the redirected fds), and spins up two
  daemon reader threads. Each thread reads bytes from a pipe, echoes
  them back to the original fd (so the server console keeps working),
  splits on \n and \r, and forwards each line to the resp queue as
  {"type":"log","stream":"stdout|stderr","line":...,"ts":...}.
  PYTHONUNBUFFERED=1 is set so nested Python converters flush
  immediately.

- core/export/orchestrator.py:
  - Thread-safe ring buffer (collections.deque, maxlen 4000) with a
    monotonically increasing seq counter. clear_logs(),
    get_logs_since(cursor), get_current_log_seq(), is_export_active().
  - _wait_response handles rtype == "log" by appending to the buffer
    and continuing the wait loop. Status messages are also surfaced as
    a "status" stream so users see high level progress alongside raw
    subprocess output.
  - load_checkpoint, _run_export, and cleanup_memory now wrap their
    bodies with the existing self._lock (previously unused), clear the
    log buffer at the start of each op, and flip _export_active in a
    try/finally so the SSE endpoint can detect idle.

- routes/export.py:
  - Wrapped every sync orchestrator call (load_checkpoint,
    cleanup_memory, export_merged_model, export_base_model,
    export_gguf, export_lora_adapter) in asyncio.to_thread so the
    FastAPI event loop stays free during long exports. Without this
    the new SSE endpoint could not be served concurrently with the
    blocking export POST.
  - New GET /api/export/logs/stream SSE endpoint. Honors
    Last-Event-ID and a since query param for reconnect, emits log /
    heartbeat / complete / error events, uses the id field to carry
    the log seq so clients can resume cleanly. On first connect
    without an explicit cursor it starts from the current seq so old
    lines from a previous run are not replayed.

Frontend

- features/export/api/export-api.ts: streamExportLogs() helper that
  authFetches the SSE endpoint and parses id / event / data fields
  manually (same pattern as streamTrainingProgress in train-api.ts).

- features/export/components/export-dialog.tsx:
  - Local useExportLogs(exporting) hook that opens the SSE stream on
    exporting transitions to true, accumulates up to 4000 lines in
    component state, and aborts on cleanup.
  - New scrollable output panel rendered above DialogFooter, only
    shown for Merged to 16bit and GGUF / Llama.cpp (LoRA adapter is
    a fast disk write with nothing to show). Dark terminal styling
    (bg-black/85, emerald text, rose for stderr, sky for status),
    max-height 14rem, auto-scrolls to the bottom on new output but
    stops following if the user scrolls up. A small streaming / idle
    indicator is shown next to the panel title.
  - DialogContent widens from sm:max-w-lg to sm:max-w-2xl when the
    output panel is visible so the logs have room to breathe.

Verified

- Python smoke test (tests/smoke_export_log_capture.py): spawns a
  real mp.get_context("spawn") process, installs _setup_log_capture,
  confirms that parent stdout prints, parent stderr prints, AND a
  child subprocess invoked via subprocess.run (both its stdout and
  stderr) are all captured in the resp queue. Passes.
- Orchestrator log helpers tested in isolation: _append_log,
  get_logs_since (with and without a cursor), clear_logs not
  resetting seq so reconnecting clients still progress. Passes.
- routes.export imports cleanly in the studio venv and /logs/stream
  shows up in router.routes.
- bun run build: tsc -b plus vite build, no TypeScript errors.

No existing export behavior is changed. If the subprocess, the SSE
endpoint, or the frontend hook fails, the export itself still runs to
completion the same way it did before, with or without logs visible.

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

* export dialog: trim bootstrap noise, scope logs per screen, show realpath

Several follow-ups to the live export log work:

1. Worker bootstrap noise (transformers venv activation, Unsloth banner,
   "Top GGUF/hub models" lists, vision detection, 2k-step weight load
   bar) is dropped from the export-dialog stream. A threading.Event
   gate in worker.py defaults closed and only opens once _handle_export
   actually starts; until then the reader thread still echoes lines to
   the saved console fd for debugging but does not push them onto the
   resp_queue. The orchestrator already spawns a fresh subprocess for
   every checkpoint load, so the gate is naturally reset between runs.

2. tqdm in non-tty mode defaults to a 10s mininterval, which makes
   multi-step bars look frozen in the panel. Set TQDM_MININTERVAL=0.5
   in the worker env so any tqdm-driven progress emits more often.

3. The dialog's useExportLogs hook now also clears its line buffer
   when exportMethod or open changes, so re-opening the dialog into a
   different action's screen no longer shows the previous action's
   saved output. A useElapsedSeconds tick + "Working Xs" badge in the
   log header gives users a visible sign that long single-step phases
   (cache copies, GGUF conversion) are still running when no new lines
   are arriving.

4. ExportBackend.export_{merged,base,gguf,lora} now return
   (success, message, output_path); the worker forwards output_path on
   each export_*_done response, the orchestrator's _run_export passes
   it to routes/export.py, which surfaces it via
   ExportOperationResponse.details.output_path. The dialog's Export
   Complete screen renders the resolved on-disk realpath under "Saved
   to" so users can find their exported model directly.

* fix(cli): unpack 3-tuple return from export backend

ExportOrchestrator.export_{merged,base,gguf,lora} now return
(success, message, output_path) so the studio dialog can show
the on-disk realpath. The CLI still unpacked 2 values, so every
`unsloth export --format ...` crashed with ValueError before
reporting completion. Update the four call sites and surface
output_path via a "Saved to:" echo.

* fix(studio): anchor export log SSE cursor at run start

The export dialog SSE defaulted its cursor to get_current_log_seq()
at connect time, so any line emitted between the POST that kicks
off the export and the client opening the stream was buffered with
seqs 1..k and then skipped (seq <= cursor). Long-running exports
looked silent during their first seconds.

Snapshot _log_seq into _run_start_seq inside clear_logs() and
expose it via get_run_start_seq(). The SSE default cursor now uses
that snapshot, so every line emitted since the current run began
is reachable regardless of when the client connects. Old runs
still can't leak in because their seqs are <= the snapshot.

* fix(studio): reconnect export log SSE on stream drop

useExportLogs launched streamExportLogs once per exporting
transition and recorded any drop in .catch(). Long GGUF exports
behind a proxy with an idle kill-timeout would silently lose the
stream for the rest of the run even though the backend already
supports Last-Event-ID resume. The "retry: 3000" directive emitted
by the backend is only meaningful to native EventSource; this
hook uses a manual fetch + ReadableStream parse so it had no
effect.

Wrap streamExportLogs in a retry loop that tracks lastSeq from
ExportLogEvent.id and passes it as since on reconnect. Backoff is
exponential with jitter, capped at 5s, reset on successful open.
The loop stops on explicit backend `complete` event or on effect
cleanup.

* fix(studio): register a second command so Typer keeps `export` as a subcommand

The CLI export unpacking tests wrap `unsloth_cli.commands.export.export`
in a fresh Typer app with a single registered command. Typer flattens a
single-command app into that command, so the test's
`runner.invoke(cli_app, ["export", ckpt, out, ...])` treats the leading
`"export"` token as an unexpected extra positional argument -- every
parametrized case failed with:

    Got unexpected extra argument (.../out)

Register a harmless `noop` second command so Typer preserves subcommand
routing and the tests actually exercise the 3-tuple unpack path they
were written to guard.

Before: 4 failed
After:  4 passed

---------

Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com>
Co-authored-by: studio-install <studio@local.install>
Co-authored-by: Roland Tannous <115670425+rolandtannous@users.noreply.github.com>
Co-authored-by: Lee Jackson <130007945+Imagineer99@users.noreply.github.com>
Co-authored-by: Roland Tannous <rolandtannous@gravityq.ai>
---
 studio/backend/core/export/export.py          |  75 ++--
 studio/backend/core/export/orchestrator.py    | 291 ++++++++++++----
 studio/backend/core/export/worker.py          | 184 +++++++++-
 studio/backend/routes/export.py               | 215 +++++++++++-
 .../backend/tests/test_export_log_cursor.py   | 179 ++++++++++
 .../src/features/export/api/export-api.ts     | 177 +++++++++-
 .../export/components/export-dialog.tsx       | 321 +++++++++++++++++-
 .../src/features/export/export-page.tsx       |  26 +-
 tests/test_cli_export_unpacking.py            | 160 +++++++++
 unsloth_cli/commands/export.py                |  11 +-
 10 files changed, 1522 insertions(+), 117 deletions(-)
 create mode 100644 studio/backend/tests/test_export_log_cursor.py
 create mode 100644 tests/test_cli_export_unpacking.py

diff --git a/studio/backend/core/export/export.py b/studio/backend/core/export/export.py
index 966e045b13..d8f2e8fa37 100644
--- a/studio/backend/core/export/export.py
+++ b/studio/backend/core/export/export.py
@@ -310,7 +310,7 @@ class ExportBackend:
         repo_id: Optional[str] = None,
         hf_token: Optional[str] = None,
         private: bool = False,
-    ) -> Tuple[bool, str]:
+    ) -> Tuple[bool, str, Optional[str]]:
         """
         Export merged model (for PEFT models).
 
@@ -323,14 +323,21 @@ class ExportBackend:
             private: Whether to make the repo private
 
         Returns:
-            Tuple of (success: bool, message: str)
+            Tuple of (success, message, output_path). output_path is the
+            resolved absolute on-disk directory of the saved model when
+            ``save_directory`` was set, else None.
         """
         if not self.current_model or not self.current_tokenizer:
-            return False, "No model loaded. Please select a checkpoint first."
+            return False, "No model loaded. Please select a checkpoint first.", None
 
         if not self.is_peft:
-            return False, "This is not a PEFT model. Use 'Export Base Model' instead."
+            return (
+                False,
+                "This is not a PEFT model. Use 'Export Base Model' instead.",
+                None,
+            )
 
+        output_path: Optional[str] = None
         try:
             # Determine save method
             if format_type == "4-bit (FP4)":
@@ -354,6 +361,7 @@ class ExportBackend:
                 # Write export metadata so the Chat page can identify the base model
                 self._write_export_metadata(save_directory)
                 logger.info(f"Model saved successfully to {save_directory}")
+                output_path = str(Path(save_directory).resolve())
 
             # Push to hub if requested
             if push_to_hub:
@@ -361,6 +369,7 @@ class ExportBackend:
                     return (
                         False,
                         "Repository ID and Hugging Face token required for Hub upload",
+                        None,
                     )
 
                 logger.info(f"Pushing merged model to Hub: {repo_id}")
@@ -378,14 +387,14 @@ class ExportBackend:
                 )
                 logger.info(f"Model pushed successfully to {repo_id}")
 
-            return True, "Model exported successfully"
+            return True, "Model exported successfully", output_path
 
         except Exception as e:
             logger.error(f"Error exporting merged model: {e}")
             import traceback
 
             logger.error(traceback.format_exc())
-            return False, f"Export failed: {str(e)}"
+            return False, f"Export failed: {str(e)}", None
 
     def export_base_model(
         self,
@@ -395,22 +404,26 @@ class ExportBackend:
         hf_token: Optional[str] = None,
         private: bool = False,
         base_model_id: Optional[str] = None,
-    ) -> Tuple[bool, str]:
+    ) -> Tuple[bool, str, Optional[str]]:
         """
         Export base model (for non-PEFT models).
 
         Returns:
-            Tuple of (success: bool, message: str)
+            Tuple of (success, message, output_path). output_path is the
+            resolved absolute on-disk directory of the saved model when
+            ``save_directory`` was set, else None.
         """
         if not self.current_model or not self.current_tokenizer:
-            return False, "No model loaded. Please select a checkpoint first."
+            return False, "No model loaded. Please select a checkpoint first.", None
 
         if self.is_peft:
             return (
                 False,
                 "This is a PEFT model. Use 'Merged Model' export type instead.",
+                None,
             )
 
+        output_path: Optional[str] = None
         try:
             # Save locally if requested
             if save_directory:
@@ -424,6 +437,7 @@ class ExportBackend:
                 # Write export metadata so the Chat page can identify the base model
                 self._write_export_metadata(save_directory)
                 logger.info(f"Model saved successfully to {save_directory}")
+                output_path = str(Path(save_directory).resolve())
 
             # Push to hub if requested
             if push_to_hub:
@@ -431,6 +445,7 @@ class ExportBackend:
                     return (
                         False,
                         "Repository ID and Hugging Face token required for Hub upload",
+                        None,
                     )
 
                 logger.info(f"Pushing base model to Hub: {repo_id}")
@@ -472,16 +487,16 @@ class ExportBackend:
                     )
                     logger.info(f"Model pushed successfully to {repo_id}")
                 else:
-                    return False, "Local save directory required for Hub upload"
+                    return False, "Local save directory required for Hub upload", None
 
-            return True, "Model exported successfully"
+            return True, "Model exported successfully", output_path
 
         except Exception as e:
             logger.error(f"Error exporting base model: {e}")
             import traceback
 
             logger.error(traceback.format_exc())
-            return False, f"Export failed: {str(e)}"
+            return False, f"Export failed: {str(e)}", None
 
     def export_gguf(
         self,
@@ -490,7 +505,7 @@ class ExportBackend:
         push_to_hub: bool = False,
         repo_id: Optional[str] = None,
         hf_token: Optional[str] = None,
-    ) -> Tuple[bool, str]:
+    ) -> Tuple[bool, str, Optional[str]]:
         """
         Export model in GGUF format.
 
@@ -502,11 +517,14 @@ class ExportBackend:
             hf_token: Hugging Face token
 
         Returns:
-            Tuple of (success: bool, message: str)
+            Tuple of (success, message, output_path). output_path is the
+            resolved absolute on-disk directory containing the .gguf
+            files when ``save_directory`` was set, else None.
         """
         if not self.current_model or not self.current_tokenizer:
-            return False, "No model loaded. Please select a checkpoint first."
+            return False, "No model loaded. Please select a checkpoint first.", None
 
+        output_path: Optional[str] = None
         try:
             # Convert quantization method to lowercase for unsloth
             quant_method = quantization_method.lower()
@@ -601,6 +619,7 @@ class ExportBackend:
                     abs_save_dir,
                     "\n  ".join(os.path.basename(f) for f in final_ggufs) or "(none)",
                 )
+                output_path = str(Path(abs_save_dir).resolve())
 
             # Push to hub if requested
             if push_to_hub:
@@ -608,6 +627,7 @@ class ExportBackend:
                     return (
                         False,
                         "Repository ID and Hugging Face token required for Hub upload",
+                        None,
                     )
 
                 logger.info(f"Pushing GGUF model to Hub: {repo_id}")
@@ -620,14 +640,18 @@ class ExportBackend:
                 )
                 logger.info(f"GGUF model pushed successfully to {repo_id}")
 
-            return True, f"GGUF model exported successfully ({quantization_method})"
+            return (
+                True,
+                f"GGUF model exported successfully ({quantization_method})",
+                output_path,
+            )
 
         except Exception as e:
             logger.error(f"Error exporting GGUF model: {e}")
             import traceback
 
             logger.error(traceback.format_exc())
-            return False, f"GGUF export failed: {str(e)}"
+            return False, f"GGUF export failed: {str(e)}", None
 
     def export_lora_adapter(
         self,
@@ -636,19 +660,22 @@ class ExportBackend:
         repo_id: Optional[str] = None,
         hf_token: Optional[str] = None,
         private: bool = False,
-    ) -> Tuple[bool, str]:
+    ) -> Tuple[bool, str, Optional[str]]:
         """
         Export LoRA adapter only (not merged).
 
         Returns:
-            Tuple of (success: bool, message: str)
+            Tuple of (success, message, output_path). output_path is the
+            resolved absolute on-disk directory of the saved adapter
+            when ``save_directory`` was set, else None.
         """
         if not self.current_model or not self.current_tokenizer:
-            return False, "No model loaded. Please select a checkpoint first."
+            return False, "No model loaded. Please select a checkpoint first.", None
 
         if not self.is_peft:
-            return False, "This is not a PEFT model. No adapter to export."
+            return False, "This is not a PEFT model. No adapter to export.", None
 
+        output_path: Optional[str] = None
         try:
             # Save locally if requested
             if save_directory:
@@ -659,6 +686,7 @@ class ExportBackend:
                 self.current_model.save_pretrained(save_directory)
                 self.current_tokenizer.save_pretrained(save_directory)
                 logger.info(f"Adapter saved successfully to {save_directory}")
+                output_path = str(Path(save_directory).resolve())
 
             # Push to hub if requested
             if push_to_hub:
@@ -666,6 +694,7 @@ class ExportBackend:
                     return (
                         False,
                         "Repository ID and Hugging Face token required for Hub upload",
+                        None,
                     )
 
                 logger.info(f"Pushing LoRA adapter to Hub: {repo_id}")
@@ -676,14 +705,14 @@ class ExportBackend:
                 )
                 logger.info(f"Adapter pushed successfully to {repo_id}")
 
-            return True, "LoRA adapter exported successfully"
+            return True, "LoRA adapter exported successfully", output_path
 
         except Exception as e:
             logger.error(f"Error exporting LoRA adapter: {e}")
             import traceback
 
             logger.error(traceback.format_exc())
-            return False, f"Adapter export failed: {str(e)}"
+            return False, f"Adapter export failed: {str(e)}", None
 
 
 # Global export backend instance
diff --git a/studio/backend/core/export/orchestrator.py b/studio/backend/core/export/orchestrator.py
index e52aa8c3cd..206dbd6dbb 100644
--- a/studio/backend/core/export/orchestrator.py
+++ b/studio/backend/core/export/orchestrator.py
@@ -16,19 +16,25 @@ Pattern follows core/inference/orchestrator.py.
 
 import atexit
 import structlog
+from collections import deque
 from loggers import get_logger
 import multiprocessing as mp
 import queue
 import threading
 import time
 from pathlib import Path
-from typing import Any, List, Optional, Tuple
+from typing import Any, Deque, Dict, List, Optional, Tuple
 from utils.paths import outputs_root
 
 logger = get_logger(__name__)
 
 _CTX = mp.get_context("spawn")
 
+# Maximum number of captured log lines kept in memory per export
+# orchestrator. Acts as scrollback for the live export log panel in the
+# UI. 4000 lines is ~1 MB worst-case at 256 chars/line.
+_LOG_BUFFER_MAXLEN = 4000
+
 
 class ExportOrchestrator:
     """
@@ -44,6 +50,9 @@ class ExportOrchestrator:
         self._proc: Optional[mp.Process] = None
         self._cmd_queue: Any = None
         self._resp_queue: Any = None
+        # Serializes export operations (load_checkpoint, export_*,
+        # cleanup) so concurrent HTTP requests can never interleave
+        # commands on the subprocess queue. Previously unused.
         self._lock = threading.Lock()
 
         # Local state mirrors (updated from subprocess responses)
@@ -51,9 +60,103 @@ class ExportOrchestrator:
         self.is_vision: bool = False
         self.is_peft: bool = False
 
+        # ── Live log capture ─────────────────────────────────────
+        # Thread-safe ring buffer of log lines forwarded from the
+        # worker subprocess. Powers the GET /api/export/logs/stream
+        # SSE endpoint that the export dialog consumes.
+        self._log_buffer: Deque[Dict[str, Any]] = deque(maxlen = _LOG_BUFFER_MAXLEN)
+        self._log_lock = threading.Lock()
+        # Monotonically increasing sequence number. Never reset across
+        # operations, so SSE clients can use it as a stable cursor even
+        # if clear_logs() is called mid-session.
+        self._log_seq: int = 0
+        # Snapshot of _log_seq captured at the start of the current run
+        # (updated by clear_logs()). The SSE endpoint defaults its
+        # cursor to this value so a client that connects AFTER the
+        # worker has already emitted its first lines still sees the
+        # full run. Every line appended during the current run has seq
+        # strictly greater than _run_start_seq, and every line from
+        # prior runs has seq less than or equal to it.
+        self._run_start_seq: int = 0
+        # True while an export operation (load/export/cleanup) is
+        # running. The SSE endpoint ends the stream 1 second after
+        # this flips back to False to drain any trailing log lines.
+        self._export_active: bool = False
+
         atexit.register(self._cleanup)
         logger.info("ExportOrchestrator initialized (subprocess mode)")
 
+    # ------------------------------------------------------------------
+    # Live log capture helpers
+    # ------------------------------------------------------------------
+
+    def _append_log(self, entry: Dict[str, Any]) -> None:
+        """Append a log line from the worker subprocess to the buffer.
+
+        Entries look like {"type": "log", "stream": "stdout"|"stderr",
+        "line": "...", "ts": ...}. Each is stamped with a monotonic
+        seq number before it lands in the buffer so SSE clients can
+        cursor through new lines.
+        """
+        line = entry.get("line")
+        if not line:
+            return
+        with self._log_lock:
+            self._log_seq += 1
+            self._log_buffer.append(
+                {
+                    "seq": self._log_seq,
+                    "stream": entry.get("stream", "stdout"),
+                    "line": line,
+                    "ts": entry.get("ts", time.time()),
+                }
+            )
+
+    def clear_logs(self) -> None:
+        """Drop any buffered log lines from a previous operation.
+
+        Called at the start of each export op so the UI shows only the
+        output of the current run. The seq counter is NOT reset, so an
+        SSE client that captured the cursor before clear_logs() will
+        still see new lines (with strictly greater seq numbers).
+
+        Also snapshots the current seq into ``_run_start_seq`` so the
+        SSE endpoint can anchor its default cursor at the start of
+        this run. Anything appended after this call has seq strictly
+        greater than the snapshot and is reachable via
+        ``get_logs_since(get_run_start_seq())``.
+        """
+        with self._log_lock:
+            self._log_buffer.clear()
+            self._run_start_seq = self._log_seq
+
+    def get_logs_since(self, cursor: int) -> Tuple[List[Dict[str, Any]], int]:
+        """Return log entries with seq > cursor, plus the new cursor."""
+        with self._log_lock:
+            new_entries = [entry for entry in self._log_buffer if entry["seq"] > cursor]
+        if new_entries:
+            return new_entries, new_entries[-1]["seq"]
+        return [], cursor
+
+    def get_current_log_seq(self) -> int:
+        """Return the current seq counter without reading any entries."""
+        with self._log_lock:
+            return self._log_seq
+
+    def get_run_start_seq(self) -> int:
+        """Return the seq value captured at the start of the current run.
+
+        The SSE endpoint uses this as the default cursor so a client
+        that connects AFTER the worker has already started emitting
+        output still sees every line from the current run.
+        """
+        with self._log_lock:
+            return self._run_start_seq
+
+    def is_export_active(self) -> bool:
+        """True while an export / load / cleanup command is running."""
+        return self._export_active
+
     # ------------------------------------------------------------------
     # Subprocess lifecycle
     # ------------------------------------------------------------------
@@ -179,8 +282,26 @@ class ExportOrchestrator:
                 error_msg = resp.get("error", "Unknown error")
                 raise RuntimeError(f"Subprocess error: {error_msg}")
 
+            if rtype == "log":
+                # Forwarded stdout/stderr line from the worker process.
+                self._append_log(resp)
+                continue
+
             if rtype == "status":
-                logger.info("Export subprocess status: %s", resp.get("message", ""))
+                message = resp.get("message", "")
+                logger.info("Export subprocess status: %s", message)
+                # Surface status messages in the live log panel too so
+                # users see high level progress (e.g. "Importing
+                # Unsloth...", "Loading checkpoint: ...") alongside
+                # subprocess output.
+                if message:
+                    self._append_log(
+                        {
+                            "stream": "status",
+                            "line": message,
+                            "ts": resp.get("ts", time.time()),
+                        }
+                    )
                 continue
 
             # Other response types during wait — skip
@@ -231,37 +352,47 @@ class ExportOrchestrator:
             "hf_token": hf_token,
         }
 
-        # Always kill existing subprocess and spawn fresh.
-        if self._ensure_subprocess_alive():
-            self._shutdown_subprocess()
-        elif self._proc is not None:
-            self._shutdown_subprocess(timeout = 2)
+        with self._lock:
+            # Start a fresh log buffer for this operation so the UI
+            # sees only the current run's output.
+            self.clear_logs()
+            self._export_active = True
+            try:
+                # Always kill existing subprocess and spawn fresh.
+                if self._ensure_subprocess_alive():
+                    self._shutdown_subprocess()
+                elif self._proc is not None:
+                    self._shutdown_subprocess(timeout = 2)
 
-        logger.info("Spawning fresh export subprocess for '%s'", checkpoint_path)
-        self._spawn_subprocess(sub_config)
+                logger.info(
+                    "Spawning fresh export subprocess for '%s'", checkpoint_path
+                )
+                self._spawn_subprocess(sub_config)
 
-        try:
-            resp = self._wait_response("loaded")
-        except RuntimeError as exc:
-            self._shutdown_subprocess(timeout = 5)
-            self.current_checkpoint = None
-            self.is_vision = False
-            self.is_peft = False
-            return False, str(exc)
+                try:
+                    resp = self._wait_response("loaded")
+                except RuntimeError as exc:
+                    self._shutdown_subprocess(timeout = 5)
+                    self.current_checkpoint = None
+                    self.is_vision = False
+                    self.is_peft = False
+                    return False, str(exc)
 
-        if resp.get("success"):
-            self.current_checkpoint = resp.get("checkpoint")
-            self.is_vision = resp.get("is_vision", False)
-            self.is_peft = resp.get("is_peft", False)
-            logger.info("Checkpoint '%s' loaded in subprocess", checkpoint_path)
-            return True, resp.get("message", "Loaded successfully")
-        else:
-            error = resp.get("message", "Failed to load checkpoint")
-            logger.error("Failed to load checkpoint: %s", error)
-            self.current_checkpoint = None
-            self.is_vision = False
-            self.is_peft = False
-            return False, error
+                if resp.get("success"):
+                    self.current_checkpoint = resp.get("checkpoint")
+                    self.is_vision = resp.get("is_vision", False)
+                    self.is_peft = resp.get("is_peft", False)
+                    logger.info("Checkpoint '%s' loaded in subprocess", checkpoint_path)
+                    return True, resp.get("message", "Loaded successfully")
+                else:
+                    error = resp.get("message", "Failed to load checkpoint")
+                    logger.error("Failed to load checkpoint: %s", error)
+                    self.current_checkpoint = None
+                    self.is_vision = False
+                    self.is_peft = False
+                    return False, error
+            finally:
+                self._export_active = False
 
     def export_merged_model(
         self,
@@ -271,7 +402,7 @@ class ExportOrchestrator:
         repo_id: Optional[str] = None,
         hf_token: Optional[str] = None,
         private: bool = False,
-    ) -> Tuple[bool, str]:
+    ) -> Tuple[bool, str, Optional[str]]:
         """Export merged PEFT model."""
         return self._run_export(
             "merged",
@@ -293,7 +424,7 @@ class ExportOrchestrator:
         hf_token: Optional[str] = None,
         private: bool = False,
         base_model_id: Optional[str] = None,
-    ) -> Tuple[bool, str]:
+    ) -> Tuple[bool, str, Optional[str]]:
         """Export base model (non-PEFT)."""
         return self._run_export(
             "base",
@@ -314,7 +445,7 @@ class ExportOrchestrator:
         push_to_hub: bool = False,
         repo_id: Optional[str] = None,
         hf_token: Optional[str] = None,
-    ) -> Tuple[bool, str]:
+    ) -> Tuple[bool, str, Optional[str]]:
         """Export model in GGUF format."""
         return self._run_export(
             "gguf",
@@ -334,7 +465,7 @@ class ExportOrchestrator:
         repo_id: Optional[str] = None,
         hf_token: Optional[str] = None,
         private: bool = False,
-    ) -> Tuple[bool, str]:
+    ) -> Tuple[bool, str, Optional[str]]:
         """Export LoRA adapter only."""
         return self._run_export(
             "lora",
@@ -347,46 +478,74 @@ class ExportOrchestrator:
             },
         )
 
-    def _run_export(self, export_type: str, params: dict) -> Tuple[bool, str]:
-        """Send an export command to the subprocess and wait for result."""
-        if not self._ensure_subprocess_alive():
-            return False, "No export subprocess running. Load a checkpoint first."
+    def _run_export(
+        self, export_type: str, params: dict
+    ) -> Tuple[bool, str, Optional[str]]:
+        """Send an export command to the subprocess and wait for result.
 
-        cmd = {"type": "export", "export_type": export_type, **params}
+        Returns ``(success, message, output_path)``. ``output_path`` is the
+        resolved on-disk directory the worker actually wrote to (None when
+        the export only pushed to Hub or failed before any file was
+        written). Surfaced via the export route's ``details.output_path``
+        so the dialog's success screen can show the user where the model
+        landed.
+        """
+        with self._lock:
+            if not self._ensure_subprocess_alive():
+                return (
+                    False,
+                    "No export subprocess running. Load a checkpoint first.",
+                    None,
+                )
 
-        try:
-            self._send_cmd(cmd)
-            resp = self._wait_response(
-                f"export_{export_type}_done",
-                timeout = 3600,  # GGUF for 30B+ models can take 30+ min
-            )
-            return resp.get("success", False), resp.get("message", "")
-        except RuntimeError as exc:
-            return False, str(exc)
+            self.clear_logs()
+            self._export_active = True
+            try:
+                cmd = {"type": "export", "export_type": export_type, **params}
+                try:
+                    self._send_cmd(cmd)
+                    resp = self._wait_response(
+                        f"export_{export_type}_done",
+                        timeout = 3600,  # GGUF for 30B+ models can take 30+ min
+                    )
+                    return (
+                        resp.get("success", False),
+                        resp.get("message", ""),
+                        resp.get("output_path"),
+                    )
+                except RuntimeError as exc:
+                    return False, str(exc), None
+            finally:
+                self._export_active = False
 
     def cleanup_memory(self) -> bool:
         """Cleanup export-related models from memory."""
-        if not self._ensure_subprocess_alive():
-            # No subprocess — just clear local state
-            self.current_checkpoint = None
-            self.is_vision = False
-            self.is_peft = False
-            return True
+        with self._lock:
+            if not self._ensure_subprocess_alive():
+                # No subprocess — just clear local state
+                self.current_checkpoint = None
+                self.is_vision = False
+                self.is_peft = False
+                return True
 
-        try:
-            self._send_cmd({"type": "cleanup"})
-            resp = self._wait_response("cleanup_done", timeout = 30)
-            success = resp.get("success", False)
-        except RuntimeError:
-            success = False
+            self._export_active = True
+            try:
+                try:
+                    self._send_cmd({"type": "cleanup"})
+                    resp = self._wait_response("cleanup_done", timeout = 30)
+                    success = resp.get("success", False)
+                except RuntimeError:
+                    success = False
 
-        # Shut down subprocess after cleanup — no model loaded
-        self._shutdown_subprocess()
+                # Shut down subprocess after cleanup — no model loaded
+                self._shutdown_subprocess()
 
-        self.current_checkpoint = None
-        self.is_vision = False
-        self.is_peft = False
-        return success
+                self.current_checkpoint = None
+                self.is_vision = False
+                self.is_peft = False
+                return success
+            finally:
+                self._export_active = False
 
     def scan_checkpoints(
         self, outputs_dir: str = str(outputs_root())
diff --git a/studio/backend/core/export/worker.py b/studio/backend/core/export/worker.py
index 3f3dc955fa..f77b1966c4 100644
--- a/studio/backend/core/export/worker.py
+++ b/studio/backend/core/export/worker.py
@@ -17,10 +17,12 @@ Pattern follows core/inference/worker.py and core/training/worker.py.
 
 from __future__ import annotations
 
+import errno
 import structlog
 from loggers import get_logger
 import os
 import sys
+import threading
 import time
 import traceback
 from pathlib import Path
@@ -29,6 +31,154 @@ from typing import Any
 logger = get_logger(__name__)
 
 
+# Gate that controls whether captured stdout/stderr lines are forwarded
+# to the parent's resp_queue (and from there to the export-dialog SSE
+# stream). Closed by default so the noisy bootstrap phase -- transformers
+# venv activation, Unsloth/torch imports, base-model resolution, "Top
+# GGUF/hub models" lists, vision detection, weight loading bars -- is
+# suppressed in the UI. _handle_export() opens the gate at the start of
+# the actual export work and leaves it open; the orchestrator always
+# spawns a fresh subprocess for the next checkpoint load (see
+# orchestrator._spawn_subprocess) which resets this state.
+#
+# Lines dropped while the gate is closed are still echoed to the saved
+# original stdout/stderr fds so the server console / log file keeps the
+# full output for debugging.
+_log_forward_gate = threading.Event()
+
+
+def _setup_log_capture(resp_queue: Any) -> None:
+    """Redirect fds 1 and 2 through pipes so every line printed by this
+    worker process and any child process it spawns is forwarded to the
+    parent process via resp_queue as {"type": "log", ...} messages.
+
+    Must be called BEFORE LogConfig.setup_logging and BEFORE any ML
+    imports, otherwise library handlers may capture the original stderr
+    reference and bypass the pipe.
+
+    Lines are also echoed back to the original stdout/stderr so the
+    server console keeps receiving the full subprocess output, even
+    while ``_log_forward_gate`` is closed.
+    """
+
+    try:
+        saved_out_fd = os.dup(1)
+        saved_err_fd = os.dup(2)
+    except OSError:
+        # dup failed (exotic platforms) - give up quietly, export still
+        # works, just no live log streaming.
+        return
+
+    try:
+        r_out, w_out = os.pipe()
+        r_err, w_err = os.pipe()
+    except OSError:
+        os.close(saved_out_fd)
+        os.close(saved_err_fd)
+        return
+
+    try:
+        os.dup2(w_out, 1)
+        os.dup2(w_err, 2)
+    except OSError:
+        for fd in (saved_out_fd, saved_err_fd, r_out, w_out, r_err, w_err):
+            try:
+                os.close(fd)
+            except OSError:
+                pass
+        return
+
+    # Close the write ends we just dup2'd (fds 1 and 2 are the real
+    # write ends now).
+    os.close(w_out)
+    os.close(w_err)
+
+    # Replace Python's sys.stdout/sys.stderr with line-buffered writers
+    # bound to the (now-redirected) fds 1 and 2.
+    try:
+        sys.stdout = os.fdopen(1, "w", buffering = 1, encoding = "utf-8", errors = "replace")
+        sys.stderr = os.fdopen(2, "w", buffering = 1, encoding = "utf-8", errors = "replace")
+    except Exception:
+        pass
+
+    def _reader(read_fd: int, stream_name: str, echo_fd: int) -> None:
+        buf = bytearray()
+        while True:
+            try:
+                chunk = os.read(read_fd, 4096)
+            except OSError as exc:
+                if exc.errno == errno.EBADF:
+                    break
+                continue
+            if not chunk:
+                break
+            # Echo to the original fd so the server console still sees
+            # the full output.
+            try:
+                os.write(echo_fd, chunk)
+            except OSError:
+                pass
+            buf.extend(chunk)
+            # Split on \n OR \r so tqdm-style progress bars update.
+            while True:
+                nl = -1
+                for i, b in enumerate(buf):
+                    if b == 0x0A or b == 0x0D:
+                        nl = i
+                        break
+                if nl < 0:
+                    break
+                line = bytes(buf[:nl]).decode("utf-8", errors = "replace")
+                del buf[: nl + 1]
+                if not line:
+                    continue
+                if not _log_forward_gate.is_set():
+                    # Gate closed (bootstrap phase) -- already echoed to
+                    # the saved console fd above; drop the line so the
+                    # export dialog doesn't see import / vendoring noise.
+                    continue
+                try:
+                    resp_queue.put_nowait(
+                        {
+                            "type": "log",
+                            "stream": stream_name,
+                            "line": line,
+                            "ts": time.time(),
+                        }
+                    )
+                except Exception:
+                    # Queue put failed (full, closed, etc.) - drop the
+                    # line rather than crash the reader thread.
+                    pass
+        if buf and _log_forward_gate.is_set():
+            try:
+                resp_queue.put_nowait(
+                    {
+                        "type": "log",
+                        "stream": stream_name,
+                        "line": bytes(buf).decode("utf-8", errors = "replace"),
+                        "ts": time.time(),
+                    }
+                )
+            except Exception:
+                pass
+
+    t_out = threading.Thread(
+        target = _reader,
+        args = (r_out, "stdout", saved_out_fd),
+        daemon = True,
+        name = "export-log-stdout",
+    )
+    t_err = threading.Thread(
+        target = _reader,
+        args = (r_err, "stderr", saved_err_fd),
+        daemon = True,
+        name = "export-log-stderr",
+    )
+    t_out.start()
+    t_err.start()
+
+
 def _activate_transformers_version(model_name: str) -> None:
     """Activate the correct transformers version BEFORE any ML imports."""
     # Ensure backend is on path for utils imports
@@ -117,9 +267,17 @@ def _handle_export(backend, cmd: dict, resp_queue: Any) -> None:
     export_type = cmd["export_type"]  # "merged", "base", "gguf", "lora"
     response_type = f"export_{export_type}_done"
 
+    # Open the log forwarding gate so the user sees the actual export
+    # progress (Unsloth merge bars, file copies, GGUF conversion, etc.)
+    # in the live log panel. The gate stays open for the rest of this
+    # subprocess's life; the orchestrator spawns a fresh subprocess for
+    # the next checkpoint load, which resets the gate to closed.
+    _log_forward_gate.set()
+
+    output_path: Any = None
     try:
         if export_type == "merged":
-            success, message = backend.export_merged_model(
+            success, message, output_path = backend.export_merged_model(
                 save_directory = cmd.get("save_directory", ""),
                 format_type = cmd.get("format_type", "16-bit (FP16)"),
                 push_to_hub = cmd.get("push_to_hub", False),
@@ -128,7 +286,7 @@ def _handle_export(backend, cmd: dict, resp_queue: Any) -> None:
                 private = cmd.get("private", False),
             )
         elif export_type == "base":
-            success, message = backend.export_base_model(
+            success, message, output_path = backend.export_base_model(
                 save_directory = cmd.get("save_directory", ""),
                 push_to_hub = cmd.get("push_to_hub", False),
                 repo_id = cmd.get("repo_id"),
@@ -137,7 +295,7 @@ def _handle_export(backend, cmd: dict, resp_queue: Any) -> None:
                 base_model_id = cmd.get("base_model_id"),
             )
         elif export_type == "gguf":
-            success, message = backend.export_gguf(
+            success, message, output_path = backend.export_gguf(
                 save_directory = cmd.get("save_directory", ""),
                 quantization_method = cmd.get("quantization_method", "Q4_K_M"),
                 push_to_hub = cmd.get("push_to_hub", False),
@@ -145,7 +303,7 @@ def _handle_export(backend, cmd: dict, resp_queue: Any) -> None:
                 hf_token = cmd.get("hf_token"),
             )
         elif export_type == "lora":
-            success, message = backend.export_lora_adapter(
+            success, message, output_path = backend.export_lora_adapter(
                 save_directory = cmd.get("save_directory", ""),
                 push_to_hub = cmd.get("push_to_hub", False),
                 repo_id = cmd.get("repo_id"),
@@ -161,6 +319,7 @@ def _handle_export(backend, cmd: dict, resp_queue: Any) -> None:
                 "type": response_type,
                 "success": success,
                 "message": message,
+                "output_path": output_path,
                 "ts": time.time(),
             },
         )
@@ -172,6 +331,7 @@ def _handle_export(backend, cmd: dict, resp_queue: Any) -> None:
                 "type": response_type,
                 "success": False,
                 "message": str(exc),
+                "output_path": None,
                 "stack": traceback.format_exc(limit = 20),
                 "ts": time.time(),
             },
@@ -217,10 +377,26 @@ def run_export_process(
     """
     import queue as _queue
 
+    # Install fd-level stdout/stderr capture FIRST so every subsequent
+    # print and every child process inherits the redirected fds. This
+    # is what powers the live export log stream in the UI.
+    _setup_log_capture(resp_queue)
+
     os.environ["TOKENIZERS_PARALLELISM"] = "false"
     os.environ["PYTHONWARNINGS"] = (
         "ignore"  # Suppress warnings at C-level before imports
     )
+    # Force unbuffered output from any child Python process (e.g. the
+    # GGUF converter) so their prints surface in the log stream as they
+    # happen rather than at the end.
+    os.environ["PYTHONUNBUFFERED"] = "1"
+    # tqdm defaults to a 10-second mininterval when stdout is not a tty
+    # (which it isn't here -- we redirected fd 1/2 to a pipe). That makes
+    # multi-step progress bars look frozen in the export log panel. Force
+    # frequent flushes so the user sees movement during merge / GGUF
+    # conversion. Has no effect on single-step bars (e.g. "Copying 1
+    # files") which only emit start/end events regardless.
+    os.environ.setdefault("TQDM_MININTERVAL", "0.5")
 
     import warnings
     from loggers.config import LogConfig
diff --git a/studio/backend/routes/export.py b/studio/backend/routes/export.py
index 3e60eaaf20..798859fc87 100644
--- a/studio/backend/routes/export.py
+++ b/studio/backend/routes/export.py
@@ -5,9 +5,15 @@
 Export API routes: checkpoint discovery and model export operations.
 """
 
+import asyncio
+import json
 import sys
+import time
 from pathlib import Path
-from fastapi import APIRouter, Depends, HTTPException, Query
+from typing import Any, AsyncGenerator, Dict, List, Optional, Tuple
+
+from fastapi import APIRouter, Depends, HTTPException, Query, Request
+from fastapi.responses import StreamingResponse
 import structlog
 from loggers import get_logger
 
@@ -97,7 +103,11 @@ async def load_checkpoint(
             logger.warning("Could not stop training: %s", e)
 
         backend = get_export_backend()
-        success, message = backend.load_checkpoint(
+        # load_checkpoint spawns and waits on a subprocess and can take
+        # minutes. Run it in a worker thread so the event loop stays
+        # free to serve the live log SSE stream concurrently.
+        success, message = await asyncio.to_thread(
+            backend.load_checkpoint,
             checkpoint_path = request.checkpoint_path,
             max_seq_length = request.max_seq_length,
             load_in_4bit = request.load_in_4bit,
@@ -129,7 +139,7 @@ async def cleanup_export_memory(
     """
     try:
         backend = get_export_backend()
-        success = backend.cleanup_memory()
+        success = await asyncio.to_thread(backend.cleanup_memory)
 
         if not success:
             raise HTTPException(
@@ -173,6 +183,17 @@ async def get_export_status(
         )
 
 
+def _export_details(output_path: Optional[str]) -> Optional[Dict[str, Any]]:
+    """Wrap the resolved on-disk export path into the details dict the
+    frontend reads to populate the Export Complete screen. Returns None
+    when the export had no local component (Hub-only push) so the
+    Pydantic field stays absent rather than ``{"output_path": null}``.
+    """
+    if not output_path:
+        return None
+    return {"output_path": output_path}
+
+
 @router.post("/export/merged", response_model = ExportOperationResponse)
 async def export_merged_model(
     request: ExportMergedModelRequest,
@@ -185,7 +206,8 @@ async def export_merged_model(
     """
     try:
         backend = get_export_backend()
-        success, message = backend.export_merged_model(
+        success, message, output_path = await asyncio.to_thread(
+            backend.export_merged_model,
             save_directory = request.save_directory,
             format_type = request.format_type,
             push_to_hub = request.push_to_hub,
@@ -197,7 +219,11 @@ async def export_merged_model(
         if not success:
             raise HTTPException(status_code = 400, detail = message)
 
-        return ExportOperationResponse(success = True, message = message)
+        return ExportOperationResponse(
+            success = True,
+            message = message,
+            details = _export_details(output_path),
+        )
     except HTTPException:
         raise
     except Exception as e:
@@ -220,7 +246,8 @@ async def export_base_model(
     """
     try:
         backend = get_export_backend()
-        success, message = backend.export_base_model(
+        success, message, output_path = await asyncio.to_thread(
+            backend.export_base_model,
             save_directory = request.save_directory,
             push_to_hub = request.push_to_hub,
             repo_id = request.repo_id,
@@ -232,7 +259,11 @@ async def export_base_model(
         if not success:
             raise HTTPException(status_code = 400, detail = message)
 
-        return ExportOperationResponse(success = True, message = message)
+        return ExportOperationResponse(
+            success = True,
+            message = message,
+            details = _export_details(output_path),
+        )
     except HTTPException:
         raise
     except Exception as e:
@@ -255,7 +286,8 @@ async def export_gguf(
     """
     try:
         backend = get_export_backend()
-        success, message = backend.export_gguf(
+        success, message, output_path = await asyncio.to_thread(
+            backend.export_gguf,
             save_directory = request.save_directory,
             quantization_method = request.quantization_method,
             push_to_hub = request.push_to_hub,
@@ -266,7 +298,11 @@ async def export_gguf(
         if not success:
             raise HTTPException(status_code = 400, detail = message)
 
-        return ExportOperationResponse(success = True, message = message)
+        return ExportOperationResponse(
+            success = True,
+            message = message,
+            details = _export_details(output_path),
+        )
     except HTTPException:
         raise
     except Exception as e:
@@ -289,7 +325,8 @@ async def export_lora_adapter(
     """
     try:
         backend = get_export_backend()
-        success, message = backend.export_lora_adapter(
+        success, message, output_path = await asyncio.to_thread(
+            backend.export_lora_adapter,
             save_directory = request.save_directory,
             push_to_hub = request.push_to_hub,
             repo_id = request.repo_id,
@@ -300,7 +337,11 @@ async def export_lora_adapter(
         if not success:
             raise HTTPException(status_code = 400, detail = message)
 
-        return ExportOperationResponse(success = True, message = message)
+        return ExportOperationResponse(
+            success = True,
+            message = message,
+            details = _export_details(output_path),
+        )
     except HTTPException:
         raise
     except Exception as e:
@@ -309,3 +350,155 @@ async def export_lora_adapter(
             status_code = 500,
             detail = f"Failed to export LoRA adapter: {str(e)}",
         )
+
+
+# ─────────────────────────────────────────────────────────────────────
+# Live export log stream (Server-Sent Events)
+# ─────────────────────────────────────────────────────────────────────
+#
+# The export worker subprocess redirects its stdout/stderr into a pipe
+# that a reader thread forwards to the orchestrator as log entries (see
+# core/export/worker.py::_setup_log_capture and
+# core/export/orchestrator.py::_append_log). This endpoint streams
+# those entries to the browser so the export dialog can show a live
+# terminal-style output panel while load_checkpoint / export_merged /
+# export_gguf / export_lora / export_base run.
+#
+# Shape follows the training progress SSE endpoint
+# (routes/training.py::stream_training_progress): each event carries
+# `id`, `event`, and `data` fields, the stream starts with a `retry:`
+# directive, and `Last-Event-ID` is honored on reconnect.
+
+
+def _format_sse(data: str, event: str, event_id: Optional[int] = None) -> str:
+    """Format a single SSE message with id/event/data fields."""
+    lines = []
+    if event_id is not None:
+        lines.append(f"id: {event_id}")
+    lines.append(f"event: {event}")
+    lines.append(f"data: {data}")
+    lines.append("")
+    lines.append("")
+    return "\n".join(lines)
+
+
+@router.get("/logs/stream")
+async def stream_export_logs(
+    request: Request,
+    since: Optional[int] = Query(
+        None,
+        description = "Return log entries with seq strictly greater than this cursor.",
+    ),
+    current_subject: str = Depends(get_current_subject),
+):
+    """
+    Stream live stdout/stderr output from the export worker subprocess
+    as Server-Sent Events.
+
+    Events:
+      - `log`      : a single log line (data: {"stream","line","ts"})
+      - `heartbeat`: periodic keepalive when no new lines are available
+      - `complete` : emitted once the export worker is idle and no new
+                     lines arrived for ~1 second. Clients should close.
+      - `error`    : unrecoverable server-side error
+
+    The `id:` field on each event is the log entry's monotonic seq
+    number so the browser can resume via `Last-Event-ID` on reconnect.
+    """
+    backend = get_export_backend()
+
+    # Determine starting cursor. Explicit `since` wins, then
+    # Last-Event-ID header on reconnect, otherwise start from the
+    # run-start snapshot captured by clear_logs() so the client sees
+    # every line emitted since the current run began -- even if the
+    # SSE connection opened after the POST that kicked off the export.
+    # Using get_current_log_seq() here would lose the early bootstrap
+    # lines that arrive in the gap between POST and SSE connect.
+    last_event_id = request.headers.get("last-event-id")
+    if since is None and last_event_id is not None:
+        try:
+            since = int(last_event_id)
+        except ValueError:
+            pass
+
+    if since is None:
+        cursor = backend.get_run_start_seq()
+    else:
+        cursor = max(0, int(since))
+
+    async def event_generator() -> AsyncGenerator[str, None]:
+        nonlocal cursor
+        # Tell the browser to reconnect after 3 seconds if the
+        # connection drops mid-export.
+        yield "retry: 3000\n\n"
+
+        last_yield = time.monotonic()
+        idle_since: Optional[float] = None
+        try:
+            while True:
+                if await request.is_disconnected():
+                    return
+
+                entries, new_cursor = backend.get_logs_since(cursor)
+                if entries:
+                    for entry in entries:
+                        payload = json.dumps(
+                            {
+                                "stream": entry.get("stream", "stdout"),
+                                "line": entry.get("line", ""),
+                                "ts": entry.get("ts"),
+                            }
+                        )
+                        yield _format_sse(
+                            payload,
+                            event = "log",
+                            event_id = int(entry.get("seq", 0)),
+                        )
+                    cursor = new_cursor
+                    last_yield = time.monotonic()
+                    idle_since = None
+                else:
+                    now = time.monotonic()
+                    if now - last_yield > 10.0:
+                        yield _format_sse("{}", event = "heartbeat")
+                        last_yield = now
+                    if not backend.is_export_active():
+                        # Give the reader thread a moment to drain any
+                        # trailing lines the worker process printed
+                        # just before signalling done.
+                        if idle_since is None:
+                            idle_since = now
+                        elif now - idle_since > 1.0:
+                            yield _format_sse(
+                                "{}",
+                                event = "complete",
+                                event_id = cursor,
+                            )
+                            return
+                    else:
+                        idle_since = None
+
+                await asyncio.sleep(0.1)
+        except asyncio.CancelledError:
+            # Client disconnected mid-yield. Don't re-raise, just end
+            # the generator cleanly so StreamingResponse finalizes.
+            return
+        except Exception as exc:
+            logger.error("Export log stream failed: %s", exc, exc_info = True)
+            try:
+                yield _format_sse(
+                    json.dumps({"error": str(exc)}),
+                    event = "error",
+                )
+            except Exception:
+                pass
+
+    return StreamingResponse(
+        event_generator(),
+        media_type = "text/event-stream",
+        headers = {
+            "Cache-Control": "no-cache",
+            "Connection": "keep-alive",
+            "X-Accel-Buffering": "no",
+        },
+    )
diff --git a/studio/backend/tests/test_export_log_cursor.py b/studio/backend/tests/test_export_log_cursor.py
new file mode 100644
index 0000000000..734ca522c9
--- /dev/null
+++ b/studio/backend/tests/test_export_log_cursor.py
@@ -0,0 +1,179 @@
+# SPDX-License-Identifier: AGPL-3.0-only
+# Copyright 2026-present the Unsloth AI Inc. team. All rights reserved. See /studio/LICENSE.AGPL-3.0
+
+"""
+Regression tests for the export log ring-buffer cursor semantics.
+
+Context: the live export log SSE stream has a race where the frontend
+opens the SSE connection AFTER the POST that starts the export. Any
+lines the worker subprocess emits during the gap between POST and SSE
+connect get buffered with seqs 1..k, and then the SSE default cursor
+`get_current_log_seq()` returns k -- so lines 1..k are forever
+unreachable to that client.
+
+Fix: `clear_logs()` snapshots the pre-run seq into `_run_start_seq`
+(exposed via `get_run_start_seq()`), and `routes/export.py` defaults
+the SSE cursor to that snapshot instead of the current seq. Every line
+appended during the current run has seq strictly greater than the
+snapshot, so the client sees the full run regardless of when it
+connects.
+
+These tests exercise the orchestrator-side contract only (no
+subprocess, no FastAPI, no frontend). The routes-level integration
+with get_run_start_seq() is a one-line edit covered by manual testing
+and the frontend build.
+"""
+
+from __future__ import annotations
+
+import sys
+import types
+from pathlib import Path
+
+import pytest
+
+
+# Backend root on sys.path so `from core.export.orchestrator import ...`
+# and friends resolve without the studio app bootstrap.
+_BACKEND_DIR = Path(__file__).resolve().parent.parent
+if str(_BACKEND_DIR) not in sys.path:
+    sys.path.insert(0, str(_BACKEND_DIR))
+
+# ExportOrchestrator imports structlog and a few heavy modules at the
+# top of orchestrator.py. Stub the ones we don't need in these unit
+# tests so the import succeeds on machines without the full studio
+# venv.
+_loggers_stub = types.ModuleType("loggers")
+_loggers_stub.get_logger = lambda name: __import__("logging").getLogger(name)
+sys.modules.setdefault("loggers", _loggers_stub)
+
+# structlog is only used for a module-level import; a bare stub is
+# enough because we never call into it in these tests.
+sys.modules.setdefault("structlog", types.ModuleType("structlog"))
+
+# utils.paths.outputs_root is only called inside scan_checkpoints which
+# we don't hit in these tests. Provide a stub module so the top-level
+# import in orchestrator.py resolves.
+_utils_pkg = types.ModuleType("utils")
+_utils_pkg.__path__ = []  # mark as package
+_utils_paths_stub = types.ModuleType("utils.paths")
+_utils_paths_stub.outputs_root = lambda: Path("/tmp")
+sys.modules.setdefault("utils", _utils_pkg)
+sys.modules.setdefault("utils.paths", _utils_paths_stub)
+
+
+@pytest.fixture
+def orchestrator():
+    """Fresh ExportOrchestrator with only the log-buffer state exercised."""
+    from core.export.orchestrator import ExportOrchestrator
+
+    return ExportOrchestrator()
+
+
+def _append(orch, line: str, stream: str = "stdout") -> None:
+    """Shortcut for simulating a worker log message."""
+    orch._append_log({"type": "log", "stream": stream, "line": line, "ts": 0.0})
+
+
+# ---------------------------------------------------------------------------
+# clear_logs() semantics
+# ---------------------------------------------------------------------------
+
+
+def test_run_start_seq_is_zero_before_any_logs(orchestrator) -> None:
+    """A brand-new orchestrator must report run_start_seq == 0 so a
+    first SSE connection picks up every line from seq 1 onward."""
+    assert orchestrator.get_run_start_seq() == 0
+
+
+def test_clear_logs_snapshots_current_seq(orchestrator) -> None:
+    """clear_logs() must capture _log_seq BEFORE clearing the buffer,
+    so subsequent runs can anchor their SSE cursor at the snapshot."""
+    _append(orchestrator, "old run line 1")
+    _append(orchestrator, "old run line 2")
+    _append(orchestrator, "old run line 3")
+    assert orchestrator.get_current_log_seq() == 3
+
+    orchestrator.clear_logs()
+
+    assert orchestrator.get_run_start_seq() == 3
+    assert orchestrator.get_current_log_seq() == 3  # seq counter preserved
+
+
+# ---------------------------------------------------------------------------
+# Race regression: SSE connects AFTER lines have been emitted
+# ---------------------------------------------------------------------------
+
+
+def test_sse_default_cursor_catches_all_current_run_lines(orchestrator) -> None:
+    """Simulate the POST-then-SSE race: worker starts emitting lines
+    immediately after clear_logs(), SSE connects several lines later.
+    Using get_run_start_seq() as the default cursor MUST return every
+    line emitted since clear_logs() ran.
+
+    Pre-fix, the SSE defaulted to get_current_log_seq() at connect
+    time, which would return the last-seen seq and miss lines N+1..M.
+    """
+    # Previous run leaves some buffered lines.
+    _append(orchestrator, "previous run line A")
+    _append(orchestrator, "previous run line B")
+
+    # New run starts: orchestrator clears the buffer and snapshots seq.
+    orchestrator.clear_logs()
+    run_start = orchestrator.get_run_start_seq()
+
+    # Worker emits early lines BEFORE the SSE connects.
+    _append(orchestrator, "Importing Unsloth...")
+    _append(orchestrator, "Loading checkpoint: /foo/bar")
+    _append(orchestrator, "Starting export...")
+
+    # SSE connects now and asks "give me everything after the run
+    # start cursor".
+    entries, new_cursor = orchestrator.get_logs_since(run_start)
+
+    # All three early lines must be present. Pre-fix this was [].
+    lines = [e["line"] for e in entries]
+    assert lines == [
+        "Importing Unsloth...",
+        "Loading checkpoint: /foo/bar",
+        "Starting export...",
+    ]
+    assert new_cursor == entries[-1]["seq"]
+
+
+def test_sse_default_cursor_excludes_previous_run(orchestrator) -> None:
+    """After clear_logs(), lines from the PREVIOUS run must not leak
+    into the new run's SSE stream. Pre-fix this worked correctly
+    (clear_logs cleared the deque); the fix must preserve it.
+    """
+    _append(orchestrator, "previous run line 1")
+    _append(orchestrator, "previous run line 2")
+    _append(orchestrator, "previous run line 3")
+    assert orchestrator.get_current_log_seq() == 3
+
+    orchestrator.clear_logs()
+    run_start = orchestrator.get_run_start_seq()
+
+    _append(orchestrator, "new run line")
+
+    entries, _ = orchestrator.get_logs_since(run_start)
+    assert [e["line"] for e in entries] == ["new run line"]
+
+
+def test_clear_logs_twice_advances_run_start(orchestrator) -> None:
+    """Back-to-back clear_logs() calls (e.g. cleanup -> load ->
+    export in the same dialog session) must each re-anchor run_start
+    at the current seq, so successive runs each start with a fresh
+    low-water mark."""
+    _append(orchestrator, "run 1 line a")
+    _append(orchestrator, "run 1 line b")
+
+    orchestrator.clear_logs()
+    assert orchestrator.get_run_start_seq() == 2
+
+    _append(orchestrator, "run 2 line a")
+    _append(orchestrator, "run 2 line b")
+    _append(orchestrator, "run 2 line c")
+
+    orchestrator.clear_logs()
+    assert orchestrator.get_run_start_seq() == 5
diff --git a/studio/frontend/src/features/export/api/export-api.ts b/studio/frontend/src/features/export/api/export-api.ts
index aff56c3e6a..450691ddfd 100644
--- a/studio/frontend/src/features/export/api/export-api.ts
+++ b/studio/frontend/src/features/export/api/export-api.ts
@@ -42,7 +42,13 @@ export interface CheckpointListResponse {
 export interface ExportOperationResponse {
   success: boolean;
   message: string;
-  details?: Record<string, unknown> | null;
+  /**
+   * Optional extras returned by the backend. The export endpoints set
+   * `details.output_path` to the resolved on-disk directory of the
+   * saved model when a local save was requested. Hub-only pushes leave
+   * `details` undefined.
+   */
+  details?: { output_path?: string | null } & Record<string, unknown>;
 }
 
 export async function fetchCheckpoints(): Promise<CheckpointListResponse> {
@@ -131,3 +137,172 @@ export async function cleanupExport(): Promise<ExportOperationResponse> {
   const response = await authFetch("/api/export/cleanup", { method: "POST" });
   return parseJson<ExportOperationResponse>(response);
 }
+
+// ─────────────────────────────────────────────────────────────────────
+// Live export log stream (Server-Sent Events)
+// ─────────────────────────────────────────────────────────────────────
+
+export type ExportLogStream = "stdout" | "stderr" | "status";
+
+export interface ExportLogEntry {
+  stream: ExportLogStream;
+  line: string;
+  ts: number | null;
+}
+
+export type ExportLogEventName = "log" | "heartbeat" | "complete" | "error";
+
+export interface ExportLogEvent {
+  event: ExportLogEventName;
+  id: number | null;
+  /** Present on `log` events. */
+  entry?: ExportLogEntry;
+  /** Present on `error` events. */
+  error?: string;
+}
+
+interface ParsedSseMessage {
+  event: string;
+  id: number | null;
+  data: string;
+}
+
+function parseSseMessage(raw: string): ParsedSseMessage | null {
+  const lines = raw.split(/\r?\n/);
+  let event = "message";
+  let id: number | null = null;
+  const dataLines: string[] = [];
+
+  for (const line of lines) {
+    if (!line) continue;
+    if (line.startsWith("event:")) {
+      event = line.slice(6).trim();
+      continue;
+    }
+    if (line.startsWith("id:")) {
+      const value = Number(line.slice(3).trim());
+      id = Number.isFinite(value) ? value : null;
+      continue;
+    }
+    if (line.startsWith("data:")) {
+      dataLines.push(line.slice(5).trimStart());
+      continue;
+    }
+    // Comment lines (":heartbeat" etc.) are ignored per SSE spec.
+  }
+
+  if (dataLines.length === 0) return null;
+  return { event, id, data: dataLines.join("\n") };
+}
+
+function isAbortError(error: unknown): boolean {
+  return error instanceof DOMException && error.name === "AbortError";
+}
+
+export async function streamExportLogs(options: {
+  signal: AbortSignal;
+  since?: number | null;
+  onOpen?: () => void;
+  onEvent: (event: ExportLogEvent) => void;
+}): Promise<void> {
+  const headers = new Headers();
+  if (typeof options.since === "number") {
+    headers.set("Last-Event-ID", String(options.since));
+  }
+
+  const url =
+    typeof options.since === "number"
+      ? `/api/export/logs/stream?since=${options.since}`
+      : "/api/export/logs/stream";
+
+  const response = await authFetch(url, {
+    method: "GET",
+    headers,
+    signal: options.signal,
+  });
+
+  if (!response.ok) {
+    throw new Error(await readError(response));
+  }
+  if (!response.body) {
+    throw new Error("Export log stream unavailable");
+  }
+
+  options.onOpen?.();
+
+  const reader = response.body.getReader();
+  const decoder = new TextDecoder();
+  let buffer = "";
+
+  try {
+    while (true) {
+      const { value, done } = await reader.read();
+      if (done) return;
+
+      buffer += decoder.decode(value, { stream: true });
+
+      let separatorIndex = buffer.search(/\r?\n\r?\n/);
+      while (separatorIndex >= 0) {
+        const rawEvent = buffer.slice(0, separatorIndex);
+        const separatorLength = buffer[separatorIndex] === "\r" ? 4 : 2;
+        buffer = buffer.slice(separatorIndex + separatorLength);
+
+        if (rawEvent.startsWith("retry:") || rawEvent.startsWith(":")) {
+          separatorIndex = buffer.search(/\r?\n\r?\n/);
+          continue;
+        }
+
+        const parsed = parseSseMessage(rawEvent);
+        if (!parsed) {
+          separatorIndex = buffer.search(/\r?\n\r?\n/);
+          continue;
+        }
+
+        try {
+          if (parsed.event === "log") {
+            const payload = JSON.parse(parsed.data) as {
+              stream?: ExportLogStream;
+              line?: string;
+              ts?: number | null;
+            };
+            options.onEvent({
+              event: "log",
+              id: parsed.id,
+              entry: {
+                stream: payload.stream ?? "stdout",
+                line: payload.line ?? "",
+                ts: payload.ts ?? null,
+              },
+            });
+          } else if (parsed.event === "heartbeat") {
+            options.onEvent({ event: "heartbeat", id: parsed.id });
+          } else if (parsed.event === "complete") {
+            options.onEvent({ event: "complete", id: parsed.id });
+            return;
+          } else if (parsed.event === "error") {
+            let errorMessage = "Export log stream error";
+            try {
+              const payload = JSON.parse(parsed.data) as { error?: string };
+              if (payload.error) errorMessage = payload.error;
+            } catch {
+              // fall through with default message
+            }
+            options.onEvent({
+              event: "error",
+              id: parsed.id,
+              error: errorMessage,
+            });
+          }
+        } catch (err) {
+          if (isAbortError(err)) return;
+          // Ignore malformed events, keep reading.
+        }
+
+        separatorIndex = buffer.search(/\r?\n\r?\n/);
+      }
+    }
+  } catch (err) {
+    if (isAbortError(err)) return;
+    throw err;
+  }
+}
diff --git a/studio/frontend/src/features/export/components/export-dialog.tsx b/studio/frontend/src/features/export/components/export-dialog.tsx
index bf7b680945..401be3ff4b 100644
--- a/studio/frontend/src/features/export/components/export-dialog.tsx
+++ b/studio/frontend/src/features/export/components/export-dialog.tsx
@@ -21,9 +21,199 @@ import { Switch } from "@/components/ui/switch";
 import { AlertCircleIcon, ArrowRight01Icon, CheckmarkCircle02Icon, Key01Icon } from "@hugeicons/core-free-icons";
 import { HugeiconsIcon } from "@hugeicons/react";
 import { AnimatePresence, motion } from "motion/react";
+import { useEffect, useRef, useState } from "react";
+import { streamExportLogs, type ExportLogEntry } from "../api/export-api";
 import { collapseAnim } from "../anim";
 import { EXPORT_METHODS, type ExportMethod } from "../constants";
 
+// Max log lines kept in the dialog's local state. Matches the backend
+// ring buffer's maxlen so the UI shows the full scrollback captured
+// server side.
+const MAX_LOG_LINES = 4000;
+
+interface UseExportLogsResult {
+  lines: ExportLogEntry[];
+  connected: boolean;
+  error: string | null;
+}
+
+/**
+ * Subscribe to the live export log SSE stream while `exporting` is
+ * true, and accumulate lines in local state. Lines from a previous
+ * action are cleared:
+ *
+ *   - when a new export starts (`exporting` flips to true), and
+ *   - when the user switches export method, dialog opens fresh, or
+ *     the dialog closes — so re-opening into a different action's
+ *     screen doesn't show the prior screen's saved output.
+ */
+function useExportLogs(
+  exporting: boolean,
+  exportMethod: ExportMethod | null,
+  open: boolean,
+): UseExportLogsResult {
+  const [lines, setLines] = useState<ExportLogEntry[]>([]);
+  const [connected, setConnected] = useState(false);
+  const [error, setError] = useState<string | null>(null);
+
+  // Reset log state whenever the user moves to a different screen --
+  // either by switching export method or by reopening the dialog -- so
+  // each (open × method) tuple shows only its own run history. The
+  // streaming effect below additionally clears on new export start.
+  useEffect(() => {
+    setLines([]);
+    setError(null);
+    setConnected(false);
+  }, [exportMethod, open]);
+
+  useEffect(() => {
+    if (!exporting) return;
+
+    setLines([]);
+    setError(null);
+
+    const abortCtrl = new AbortController();
+    let cancelled = false;
+    // Track the highest seq we've observed on a `log` event so we can
+    // resume the stream via `since=` / `Last-Event-ID` after a drop.
+    // The backend's SSE `id:` field carries this as ExportLogEvent.id.
+    let lastSeq: number | null = null;
+    // Exponential backoff with jitter, capped. Reset on every
+    // successful connection so flaky networks don't accumulate delay.
+    let backoffMs = 500;
+    const MAX_BACKOFF_MS = 5000;
+    // Flipped by a terminal event (explicit `complete` from the
+    // backend or a non-transient error we choose not to retry). Stops
+    // the outer reconnect loop even if `exporting` is still true.
+    let terminated = false;
+
+    const run = async () => {
+      while (!cancelled && !terminated) {
+        try {
+          await streamExportLogs({
+            signal: abortCtrl.signal,
+            since: lastSeq,
+            onOpen: () => {
+              if (cancelled) return;
+              setConnected(true);
+              // Reset backoff on every successful connect so later
+              // drops don't inherit accumulated delay from earlier ones.
+              backoffMs = 500;
+            },
+            onEvent: (event) => {
+              if (cancelled) return;
+              if (event.event === "log" && event.entry) {
+                if (typeof event.id === "number") {
+                  lastSeq = event.id;
+                }
+                const entry = event.entry;
+                setLines((prev) => {
+                  const next = prev.length >= MAX_LOG_LINES
+                    ? prev.slice(prev.length - MAX_LOG_LINES + 1)
+                    : prev.slice();
+                  next.push(entry);
+                  return next;
+                });
+              } else if (event.event === "complete") {
+                // Backend signalled the run is fully drained -- stop
+                // trying to reconnect even though `exporting` may not
+                // have flipped false yet on this tick.
+                terminated = true;
+              } else if (event.event === "error" && event.error) {
+                setError(event.error);
+              }
+            },
+          });
+        } catch (err: unknown) {
+          if (cancelled) return;
+          if (err instanceof DOMException && err.name === "AbortError") return;
+          setError(err instanceof Error ? err.message : String(err));
+          // Fall through to the backoff path below; a fetch-level
+          // failure is retryable the same way a clean EOF is.
+        }
+
+        setConnected(false);
+        if (cancelled || terminated) return;
+
+        // Exponential backoff with jitter before reconnecting. The
+        // backend's ring buffer plus Last-Event-ID resume means we
+        // don't lose lines across the retry as long as the reconnect
+        // happens within the buffer's lifetime (~4000 lines).
+        const delay = backoffMs + Math.floor(Math.random() * 250);
+        backoffMs = Math.min(backoffMs * 2, MAX_BACKOFF_MS);
+        try {
+          await new Promise<void>((resolve, reject) => {
+            if (abortCtrl.signal.aborted) {
+              reject(new DOMException("Aborted", "AbortError"));
+              return;
+            }
+            const timeoutId = window.setTimeout(resolve, delay);
+            abortCtrl.signal.addEventListener(
+              "abort",
+              () => {
+                window.clearTimeout(timeoutId);
+                reject(new DOMException("Aborted", "AbortError"));
+              },
+              { once: true },
+            );
+          });
+        } catch {
+          return;
+        }
+      }
+    };
+
+    // run()'s own try/catch handles every failure path we care about;
+    // swallow anything that somehow escapes so React's dev overlay
+    // doesn't flag an unhandled rejection on dialog close.
+    void run().catch(() => {});
+
+    return () => {
+      cancelled = true;
+      abortCtrl.abort();
+      setConnected(false);
+    };
+  }, [exporting]);
+
+  return { lines, connected, error };
+}
+
+/**
+ * Tick every second while `exporting` is true and report elapsed
+ * seconds. Powers the "Working… 27s" badge in the log header so the
+ * panel doesn't look frozen during long single-step phases (cache
+ * file copy, GGUF conversion) when no new lines are arriving.
+ */
+function useElapsedSeconds(exporting: boolean): number {
+  const [elapsed, setElapsed] = useState(0);
+  useEffect(() => {
+    if (!exporting) {
+      setElapsed(0);
+      return;
+    }
+    const startedAt = Date.now();
+    setElapsed(0);
+    const id = window.setInterval(() => {
+      setElapsed(Math.floor((Date.now() - startedAt) / 1000));
+    }, 1000);
+    return () => window.clearInterval(id);
+  }, [exporting]);
+  return elapsed;
+}
+
+function formatElapsed(seconds: number): string {
+  if (seconds < 60) return `${seconds}s`;
+  const m = Math.floor(seconds / 60);
+  const s = seconds % 60;
+  return `${m}m ${s.toString().padStart(2, "0")}s`;
+}
+
+function formatLogLine(entry: ExportLogEntry): string {
+  // Strip trailing carriage returns that tqdm-style progress leaves
+  // in the stream so the scrollback doesn't render funky boxes.
+  return entry.line.replace(/\r+$/g, "");
+}
+
 type Destination = "local" | "hub";
 
 interface ExportDialogProps {
@@ -49,6 +239,12 @@ interface ExportDialogProps {
   exporting: boolean;
   exportError: string | null;
   exportSuccess: boolean;
+  /**
+   * Resolved on-disk realpath of the most recent successful export.
+   * Surfaced on the Export Complete screen so users can find their
+   * model. Null when the export only pushed to the Hub.
+   */
+  exportOutputPath: string | null;
 }
 
 export function ExportDialog({
@@ -74,7 +270,38 @@ export function ExportDialog({
   exporting,
   exportError,
   exportSuccess,
+  exportOutputPath,
 }: ExportDialogProps) {
+  // Live log capture is only meaningful for export methods that run
+  // a slow subprocess operation with interesting stdout: merged and
+  // gguf. LoRA adapter export is a fast disk write and would just
+  // show a blank panel, so we hide it there.
+  const showLogPanel =
+    exportMethod === "merged" || exportMethod === "gguf";
+
+  const { lines: logLines, connected: logConnected, error: logError } =
+    useExportLogs(exporting && showLogPanel, exportMethod, open);
+  const elapsedSeconds = useElapsedSeconds(exporting && showLogPanel);
+
+  const logScrollRef = useRef<HTMLDivElement | null>(null);
+  // Auto-scroll to bottom whenever a new line arrives, unless the
+  // user has scrolled up to read earlier output.
+  const [followTail, setFollowTail] = useState(true);
+
+  useEffect(() => {
+    if (!followTail) return;
+    const el = logScrollRef.current;
+    if (el) el.scrollTop = el.scrollHeight;
+  }, [logLines, followTail]);
+
+  const handleLogScroll = () => {
+    const el = logScrollRef.current;
+    if (!el) return;
+    const nearBottom =
+      el.scrollHeight - el.scrollTop - el.clientHeight < 24;
+    setFollowTail(nearBottom);
+  };
+
   return (
     <Dialog
       open={open}
@@ -83,20 +310,36 @@ export function ExportDialog({
         onOpenChange(v);
       }}
     >
-      <DialogContent className="sm:max-w-lg" onInteractOutside={(e) => { if (exporting) e.preventDefault(); }}>
+      <DialogContent
+        className={showLogPanel ? "sm:max-w-2xl" : "sm:max-w-lg"}
+        onInteractOutside={(e) => { if (exporting) e.preventDefault(); }}
+      >
         {exportSuccess ? (
           <>
             <div className="flex flex-col items-center gap-3 py-6">
               <div className="flex size-12 items-center justify-center rounded-full bg-emerald-500/10">
                 <HugeiconsIcon icon={CheckmarkCircle02Icon} className="size-6 text-emerald-500" />
               </div>
-              <div className="text-center">
+              <div className="flex flex-col items-center gap-2 text-center">
                 <h3 className="text-lg font-semibold">Export Complete</h3>
-                <p className="mt-1 text-sm text-muted-foreground">
+                <p className="text-sm text-muted-foreground">
                   {destination === "hub"
                     ? "Model successfully pushed to Hugging Face Hub."
                     : "Model saved locally."}
                 </p>
+                {exportOutputPath ? (
+                  <div className="mt-1 flex w-full max-w-md flex-col items-stretch gap-1 rounded-lg border border-border/40 bg-muted/40 px-3 py-2 text-left">
+                    <span className="text-[11px] font-medium uppercase tracking-wide text-muted-foreground">
+                      Saved to
+                    </span>
+                    <code
+                      className="select-all break-all font-mono text-[12px] text-foreground"
+                      title={exportOutputPath}
+                    >
+                      {exportOutputPath}
+                    </code>
+                  </div>
+                ) : null}
               </div>
             </div>
             <DialogFooter>
@@ -256,6 +499,78 @@ export function ExportDialog({
           </div> */}
             </div>
 
+            {/* Live export output panel */}
+            <AnimatePresence>
+              {showLogPanel && (exporting || logLines.length > 0) && (
+                <motion.div {...collapseAnim} className="overflow-hidden">
+                  <div className="flex flex-col gap-1.5 pt-1">
+                    <div className="flex items-center justify-between">
+                      <label className="text-xs font-medium text-muted-foreground">
+                        Export output
+                      </label>
+                      <div className="flex items-center gap-2 text-[11px] text-muted-foreground/80">
+                        <span
+                          className={
+                            logConnected
+                              ? "inline-block size-1.5 rounded-full bg-emerald-500"
+                              : "inline-block size-1.5 rounded-full bg-muted-foreground/40"
+                          }
+                        />
+                        <span>
+                          {logConnected
+                            ? "streaming"
+                            : exporting
+                              ? "connecting..."
+                              : "idle"}
+                        </span>
+                        {exporting && elapsedSeconds > 0 ? (
+                          <span className="tabular-nums text-muted-foreground/70">
+                            · {formatElapsed(elapsedSeconds)}
+                          </span>
+                        ) : null}
+                      </div>
+                    </div>
+                    <div
+                      ref={logScrollRef}
+                      onScroll={handleLogScroll}
+                      className="h-56 w-full overflow-auto rounded-lg border border-border/40 bg-black/85 p-3 font-mono text-[11px] leading-[1.45] text-emerald-200/90"
+                    >
+                      {logLines.length === 0 ? (
+                        <div className="flex h-full items-center justify-center text-muted-foreground/70">
+                          <span className="flex items-center gap-2">
+                            <Spinner className="size-3" />
+                            Waiting for worker output...
+                          </span>
+                        </div>
+                      ) : (
+                        <pre className="whitespace-pre-wrap break-words">
+                          {logLines.map((entry, idx) => (
+                            <div
+                              key={idx}
+                              className={
+                                entry.stream === "stderr"
+                                  ? "text-rose-300/90"
+                                  : entry.stream === "status"
+                                    ? "text-sky-300/90"
+                                    : ""
+                              }
+                            >
+                              {formatLogLine(entry)}
+                            </div>
+                          ))}
+                        </pre>
+                      )}
+                    </div>
+                    {logError && (
+                      <p className="text-[11px] text-destructive/80">
+                        Log stream: {logError}
+                      </p>
+                    )}
+                  </div>
+                </motion.div>
+              )}
+            </AnimatePresence>
+
             <DialogFooter>
               <Button
                 variant="outline"
diff --git a/studio/frontend/src/features/export/export-page.tsx b/studio/frontend/src/features/export/export-page.tsx
index b47697e97b..454ce40772 100644
--- a/studio/frontend/src/features/export/export-page.tsx
+++ b/studio/frontend/src/features/export/export-page.tsx
@@ -121,6 +121,10 @@ export function ExportPage() {
   const [exporting, setExporting] = useState(false);
   const [exportError, setExportError] = useState<string | null>(null);
   const [exportSuccess, setExportSuccess] = useState(false);
+  // Resolved on-disk path of the most recent successful export, surfaced
+  // on the Export Complete screen so the user can find their model
+  // without digging through the server log. Null for Hub-only pushes.
+  const [exportOutputPath, setExportOutputPath] = useState<string | null>(null);
 
   const hfComboboxAnchorRef = useRef<HTMLDivElement>(null);
   const localComboboxAnchorRef = useRef<HTMLDivElement>(null);
@@ -405,6 +409,7 @@ export function ExportPage() {
     setExporting(true);
     setExportError(null);
     setExportSuccess(false);
+    setExportOutputPath(null);
 
     // For GGUF, use a flat folder like "exports/gemma-3-4b-it-finetune-gguf"
     // For other formats, nest under training-run/checkpoint
@@ -433,18 +438,24 @@ export function ExportPage() {
         });
       }
 
-      // 2. Run export based on method
+      // 2. Run export based on method. Capture the resolved output_path
+      // (when the backend wrote a local copy) so the success screen can
+      // show the user the realpath of their saved model. For multi-quant
+      // GGUF runs, the directory is the same for every quant so we just
+      // keep the last response.
+      let lastOutputPath: string | null = null;
       if (exportMethod === "merged") {
         if (isAdapter) {
-          await exportMerged({
+          const resp = await exportMerged({
             save_directory: saveDir,
             push_to_hub: pushToHub,
             repo_id: repoId,
             hf_token: token,
             private: privateRepo,
           });
+          lastOutputPath = resp.details?.output_path ?? null;
         } else {
-          await exportBase({
+          const resp = await exportBase({
             save_directory: saveDir,
             push_to_hub: pushToHub,
             repo_id: repoId,
@@ -452,27 +463,31 @@ export function ExportPage() {
             private: privateRepo,
             base_model_id: selectedModelData?.base_model,
           });
+          lastOutputPath = resp.details?.output_path ?? null;
         }
       } else if (exportMethod === "gguf") {
         for (const quant of quantLevels) {
-          await exportGGUF({
+          const resp = await exportGGUF({
             save_directory: saveDir,
             quantization_method: quant,
             push_to_hub: pushToHub,
             repo_id: repoId,
             hf_token: token,
           });
+          lastOutputPath = resp.details?.output_path ?? lastOutputPath;
         }
       } else if (exportMethod === "lora") {
-        await exportLoRA({
+        const resp = await exportLoRA({
           save_directory: saveDir,
           push_to_hub: pushToHub,
           repo_id: repoId,
           hf_token: token,
           private: privateRepo,
         });
+        lastOutputPath = resp.details?.output_path ?? null;
       }
 
+      setExportOutputPath(lastOutputPath);
       setExportSuccess(true);
     } catch (err) {
       setExportError(
@@ -1080,6 +1095,7 @@ export function ExportPage() {
         exporting={exporting}
         exportError={exportError}
         exportSuccess={exportSuccess}
+        exportOutputPath={exportOutputPath}
       />
     </div>
   );
diff --git a/tests/test_cli_export_unpacking.py b/tests/test_cli_export_unpacking.py
new file mode 100644
index 0000000000..3dbab90b5a
--- /dev/null
+++ b/tests/test_cli_export_unpacking.py
@@ -0,0 +1,160 @@
+# SPDX-License-Identifier: AGPL-3.0-only
+# Copyright 2026-present the Unsloth AI Inc. team. All rights reserved. See /studio/LICENSE.AGPL-3.0
+
+"""
+Regression tests for unsloth_cli.commands.export.
+
+Context: the studio export dialog live-logs work changed
+ExportOrchestrator.export_{merged_model,base_model,gguf,lora_adapter}
+to return (success, message, output_path) instead of (success, message)
+so the frontend can show the on-disk realpath on the success screen.
+The CLI at unsloth_cli/commands/export.py still unpacks two values,
+so every `unsloth export --format ...` crashes with:
+
+    ValueError: too many values to unpack (expected 2)
+
+These tests pin the CLI to the 3-tuple contract by invoking it against
+a fake ExportBackend and asserting exit_code == 0 for each --format.
+No real ML imports; the fake is installed via sys.modules injection so
+the CLI's deferred `from studio.backend.core.export import ExportBackend`
+binds to it.
+"""
+
+from __future__ import annotations
+
+import sys
+import types
+from pathlib import Path
+
+import pytest
+import typer
+from typer.testing import CliRunner
+
+
+# ---------------------------------------------------------------------------
+# Fake ExportBackend
+# ---------------------------------------------------------------------------
+
+
+class _FakeExportBackend:
+    """Stand-in for studio.backend.core.export.ExportBackend.
+
+    All export_* methods return the new 3-tuple contract. load_checkpoint
+    keeps its 2-tuple shape (unchanged by the live-logs work).
+    """
+
+    def __init__(self) -> None:
+        self.loaded: str | None = None
+
+    def load_checkpoint(self, **kwargs):
+        self.loaded = kwargs.get("checkpoint_path")
+        return True, f"Loaded {self.loaded}"
+
+    def scan_checkpoints(self, **kwargs):
+        return []
+
+    def export_merged_model(self, **kwargs):
+        return True, "merged ok", str(Path(kwargs["save_directory"]).resolve())
+
+    def export_base_model(self, **kwargs):
+        return True, "base ok", str(Path(kwargs["save_directory"]).resolve())
+
+    def export_gguf(self, **kwargs):
+        return True, "gguf ok", str(Path(kwargs["save_directory"]).resolve())
+
+    def export_lora_adapter(self, **kwargs):
+        return True, "lora ok", str(Path(kwargs["save_directory"]).resolve())
+
+
+def _install_fake_studio_backend(monkeypatch: pytest.MonkeyPatch) -> None:
+    """Inject fake studio.backend.core.export into sys.modules.
+
+    The CLI imports ExportBackend lazily inside the command function, so
+    patching sys.modules before invoking the command is sufficient to
+    steer the `from studio.backend.core.export import ExportBackend`
+    statement at the fake. Parent packages (studio, studio.backend,
+    studio.backend.core) are stubbed too so Python's import machinery
+    doesn't try to resolve the real (structlog-dependent) tree.
+    """
+    for name in ("studio", "studio.backend", "studio.backend.core"):
+        monkeypatch.setitem(sys.modules, name, types.ModuleType(name))
+
+    fake_mod = types.ModuleType("studio.backend.core.export")
+    fake_mod.ExportBackend = _FakeExportBackend
+    monkeypatch.setitem(sys.modules, "studio.backend.core.export", fake_mod)
+
+    # Drop any cached import of the CLI module so the deferred import
+    # inside export() re-resolves against our fake module rather than a
+    # previously cached real one.
+    monkeypatch.delitem(sys.modules, "unsloth_cli.commands.export", raising = False)
+
+
+@pytest.fixture
+def cli_app(monkeypatch: pytest.MonkeyPatch) -> typer.Typer:
+    """Typer app wrapping unsloth_cli.commands.export.export."""
+    _install_fake_studio_backend(monkeypatch)
+    from unsloth_cli.commands import export as export_cmd
+
+    app = typer.Typer()
+    app.command("export")(export_cmd.export)
+
+    # Typer flattens a single-command app into that command, which would
+    # make argv[0] ("export") look like an extra positional argument to
+    # the test invocation. Register a harmless second command so Typer
+    # keeps "export" as a real subcommand and the tests drive the
+    # intended code path.
+    @app.command("noop")
+    def _noop() -> None:  # pragma: no cover - only exists to pin routing
+        pass
+
+    return app
+
+
+@pytest.fixture
+def runner() -> CliRunner:
+    return CliRunner()
+
+
+# ---------------------------------------------------------------------------
+# The actual regression tests
+# ---------------------------------------------------------------------------
+
+
+@pytest.mark.parametrize(
+    "format_flag,quant_flag",
+    [
+        ("merged-16bit", None),
+        ("merged-4bit", None),
+        ("gguf", "q4_k_m"),
+        ("lora", None),
+    ],
+)
+def test_cli_export_unpacks_three_tuple(
+    cli_app: typer.Typer,
+    runner: CliRunner,
+    tmp_path: Path,
+    format_flag: str,
+    quant_flag: str | None,
+) -> None:
+    """Each --format path must unpack (success, message, output_path)
+    without raising ValueError. Pre-fix, every parametrized case fails
+    with 'too many values to unpack (expected 2)'.
+    """
+    ckpt = tmp_path / "ckpt"
+    ckpt.mkdir()
+    out = tmp_path / "out"
+
+    cli_args = ["export", str(ckpt), str(out), "--format", format_flag]
+    if quant_flag is not None:
+        cli_args += ["--quantization", quant_flag]
+
+    result = runner.invoke(cli_app, cli_args)
+
+    assert result.exit_code == 0, (
+        f"CLI exited with code {result.exit_code} for --format {format_flag}.\n"
+        f"Output:\n{result.output}\n"
+        f"Exception: {result.exception!r}"
+    )
+    # Sanity: the success message from the fake backend should reach stdout.
+    expected_prefix = format_flag.split("-")[0]
+    assert f"{expected_prefix} ok" in result.output
diff --git a/unsloth_cli/commands/export.py b/unsloth_cli/commands/export.py
index 785f309e8e..be65f310ce 100644
--- a/unsloth_cli/commands/export.py
+++ b/unsloth_cli/commands/export.py
@@ -90,8 +90,9 @@ def export(
     typer.echo(message)
 
     typer.echo(f"Exporting as {format}...")
+    output_path: Optional[str] = None
     if format == "merged-16bit":
-        success, message = backend.export_merged_model(
+        success, message, output_path = backend.export_merged_model(
             save_directory = str(output_dir),
             format_type = "16-bit (FP16)",
             push_to_hub = push_to_hub,
@@ -100,7 +101,7 @@ def export(
             private = private,
         )
     elif format == "merged-4bit":
-        success, message = backend.export_merged_model(
+        success, message, output_path = backend.export_merged_model(
             save_directory = str(output_dir),
             format_type = "4-bit (FP4)",
             push_to_hub = push_to_hub,
@@ -109,7 +110,7 @@ def export(
             private = private,
         )
     elif format == "gguf":
-        success, message = backend.export_gguf(
+        success, message, output_path = backend.export_gguf(
             save_directory = str(output_dir),
             quantization_method = quantization.upper(),
             push_to_hub = push_to_hub,
@@ -117,7 +118,7 @@ def export(
             hf_token = hf_token,
         )
     elif format == "lora":
-        success, message = backend.export_lora_adapter(
+        success, message, output_path = backend.export_lora_adapter(
             save_directory = str(output_dir),
             push_to_hub = push_to_hub,
             repo_id = repo_id,
@@ -130,3 +131,5 @@ def export(
         raise typer.Exit(code = 1)
 
     typer.echo(message)
+    if output_path:
+        typer.echo(f"Saved to: {output_path}")

From 4328d0b4f691f30a0c645d2b3d9d7c421599b914 Mon Sep 17 00:00:00 2001
From: Datta Nimmaturi <venkatadattasainimmaturi@gmail.com>
Date: Tue, 14 Apr 2026 21:31:10 +0530
Subject: [PATCH 091/216] Fix num_items_in_batch GA for Gemma4 (#4998)

* Fix num_items_in_batch GA for Gemma4

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

---------

Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com>
---
 unsloth/models/_utils.py | 13 ++++++++-----
 1 file changed, 8 insertions(+), 5 deletions(-)

diff --git a/unsloth/models/_utils.py b/unsloth/models/_utils.py
index d8fe92739a..96675eeaf0 100644
--- a/unsloth/models/_utils.py
+++ b/unsloth/models/_utils.py
@@ -2108,11 +2108,14 @@ def patch_gradient_accumulation_fix(Trainer):
                 "def __init__", "def _unsloth___init__", 1
             )
 
-            # Force else branch
-            init_function = re.sub(
-                r'if[\s]+hasattr\(\s*unwrapped_model\s*,\s*"accepts_loss_kwargs"\s*\)\s*:',
-                'if hasattr(unwrapped_model, "accepts_loss_kwargs") and False:',
-                init_function,
+            # Respect an inner wrapped model's explicit accepts_loss_kwargs flag before inferring from forward(**kwargs).
+            # https://github.com/unslothai/unsloth/issues/4982 Gemma4ForConditionalGeneration had issues with grad_acc
+            init_function = init_function.replace(
+                "self.model_accepts_loss_kwargs = unwrapped_model.accepts_loss_kwargs\n    else:",
+                "self.model_accepts_loss_kwargs = unwrapped_model.accepts_loss_kwargs\n"
+                '    elif hasattr(getattr(unwrapped_model, "model", None), "accepts_loss_kwargs"):\n'
+                "        self.model_accepts_loss_kwargs = unwrapped_model.model.accepts_loss_kwargs\n"
+                "    else:",
             )
             exec(init_function, globals())
             Trainer.__init__ = _unsloth___init__

From 514bb3a20e4f5a3b8f712881c882143bc29f6f2c Mon Sep 17 00:00:00 2001
From: Roland Tannous <115670425+rolandtannous@users.noreply.github.com>
Date: Tue, 14 Apr 2026 20:16:30 +0400
Subject: [PATCH 092/216] studio: pin peft to 0.18.1 to fix export subprocess
 issues (#5015)

* studio: pin peft to 0.18.1 to fix export subprocess issues

peft 0.19.0 causes export subprocess shutdown failures in Studio.
Reverting to 0.18.1 resolves the issue.

* studio: move peft pin to extras-no-deps to prevent torch upgrade

Installing peft via overrides.txt would resolve its deps and pull in
torch>=0.11.0, breaking other pinned packages. Moving the pin to
extras-no-deps.txt ensures --no-deps is used during install.
---
 studio/backend/requirements/extras-no-deps.txt | 4 ++++
 1 file changed, 4 insertions(+)

diff --git a/studio/backend/requirements/extras-no-deps.txt b/studio/backend/requirements/extras-no-deps.txt
index 59a8e45434..a16571567d 100644
--- a/studio/backend/requirements/extras-no-deps.txt
+++ b/studio/backend/requirements/extras-no-deps.txt
@@ -5,6 +5,10 @@ julius
 torchcodec
 snac
 
+# peft 0.19.0 causes export subprocess shutdown issues in Studio;
+# installing with --no-deps to avoid pulling in torch>=0.11.0
+peft==0.18.1
+
 # TRL and related packages
 trl==0.23.1
 git+https://github.com/meta-pytorch/OpenEnv.git

From bb14ab144af6d7423239dbfa8456a969eeb52f8c Mon Sep 17 00:00:00 2001
From: Daniel Han <danielhanchen@gmail.com>
Date: Tue, 14 Apr 2026 09:46:22 -0700
Subject: [PATCH 093/216] Studio: live model-load progress + rate/ETA on
 download and load (#5017)

* Studio: live model-load progress + rate/ETA on download and load

Two UX fixes for the opaque multi-minute wait between clicking Load
and being able to chat, visible most clearly on large MoE GGUFs like
MiniMax-M2.7 (131 GB of weights on a 97 GB GPU):

1. **Model-load phase is now observable.** The existing chat flow
   transitions the toast to "Starting model..." as soon as the
   download hits 100%, then shows a spinner with no other feedback
   until llama-server reports healthy. For a 130 GB model that spinner
   freezes for five-plus minutes while the kernel pages shards into
   the page cache. A new `GET /api/inference/load-progress` endpoint
   samples `/proc/<pid>/status VmRSS` on the llama-server subprocess
   against the sum of shard file sizes on disk, so the UI can render
   a real bar plus rate / ETA during that window.

2. **Rate and ETA on downloads and loads.** Both the chat toast and
   the training-start overlay used to show a static pair of numbers
   (for example "15.4 of 140.8 GB"). A rolling 15-second window over
   the existing byte-series now surfaces "85.3 MB/s, 24m 23s left"
   beside that pair. The estimator is shared between the download
   and load phases so the numbers don't reset when the phase flips.

Also fixes a pre-existing assignment bug uncovered while wiring this
up: `load_model` was storing the caller's `gguf_path` kwarg into
`self._gguf_path`, which is `None` on the HF-download code path. The
resolved on-disk path (`model_path`) is what llama-server actually
mmaps; downstream consumers need that. No existing reader used
`_gguf_path`, so this is a correctness fix for the new endpoint.

- Backend: `LlamaCppBackend.load_progress()`, `GET /api/inference/load-progress`, `LoadProgressResponse` Pydantic model.
- Frontend: `useTransferStats` hook, `formatRate` / `formatEta` helpers, `getLoadProgress` client, rewired chat toast and `DownloadRow` in the training overlay.
- Tests: `studio/backend/tests/test_llama_cpp_load_progress.py` covers empty states, mmap phase, ready phase, sharded total aggregation, missing gguf_path, and unreadable /proc (7 cases). `tsc -b` and `vite build` on the frontend both clean.

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

---------

Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com>
---
 studio/backend/core/inference/llama_cpp.py    |  97 +++++-
 studio/backend/models/inference.py            |  33 ++
 studio/backend/routes/inference.py            |  29 ++
 .../tests/test_llama_cpp_load_progress.py     | 258 ++++++++++++++++
 .../src/features/chat/api/chat-api.ts         |  25 ++
 .../chat/hooks/use-chat-model-runtime.ts      | 281 +++++++++++++-----
 .../features/chat/hooks/use-transfer-stats.ts |  98 ++++++
 .../features/chat/utils/format-transfer.ts    |  44 +++
 .../studio/training-start-overlay.tsx         |  16 +-
 9 files changed, 805 insertions(+), 76 deletions(-)
 create mode 100644 studio/backend/tests/test_llama_cpp_load_progress.py
 create mode 100644 studio/frontend/src/features/chat/hooks/use-transfer-stats.ts
 create mode 100644 studio/frontend/src/features/chat/utils/format-transfer.ts

diff --git a/studio/backend/core/inference/llama_cpp.py b/studio/backend/core/inference/llama_cpp.py
index eafe83beea..71dec4c38b 100644
--- a/studio/backend/core/inference/llama_cpp.py
+++ b/studio/backend/core/inference/llama_cpp.py
@@ -185,6 +185,96 @@ class LlamaCppBackend:
         """Return the model's native context length from GGUF metadata."""
         return self._context_length
 
+    def load_progress(self) -> Optional[dict]:
+        """Return live model-load progress, or None if not loading.
+
+        While llama-server is warming up, its process is typically in
+        kernel state D (disk sleep) mmap'ing the weight shards into
+        page cache before pushing layers to VRAM. During that window
+        ``/api/inference/status`` only reports ``loading``, which gives
+        the UI nothing to display besides a spinner that looks stuck
+        for minutes on large MoE models.
+
+        This method samples ``/proc/<pid>/status VmRSS`` against the
+        sum of the GGUF shard sizes so the UI can render a real bar
+        and compute rate / ETA. Returns ``None`` when no load is in
+        flight (no process, or process already healthy).
+
+        Shape::
+
+            {
+                "phase": "mmap" | "ready",
+                "bytes_loaded": int,   # VmRSS of the llama-server
+                "bytes_total":  int,   # sum of shard file sizes
+                "fraction": float,     # bytes_loaded / bytes_total, 0..1
+            }
+
+        Linux-only in the current implementation. On macOS/Windows the
+        equivalent would be a different API; this returns ``None`` on
+        platforms where ``/proc/<pid>/status`` is unavailable.
+        """
+        proc = self._process
+        if proc is None:
+            return None
+        pid = proc.pid
+        if pid is None:
+            return None
+
+        # Sum up shard sizes (primary + any extras sitting alongside).
+        bytes_total = 0
+        gguf_path = self._gguf_path
+        if gguf_path:
+            primary = Path(gguf_path)
+            try:
+                if primary.is_file():
+                    bytes_total += primary.stat().st_size
+            except OSError:
+                pass
+            # Extra shards live alongside the primary with the same prefix
+            # before the shard index (e.g. ``-00001-of-00004.gguf``).
+            try:
+                parent = primary.parent
+                stem = primary.name
+                m = _SHARD_RE.match(stem)
+                prefix = m.group(1) if m else None
+                if prefix and parent.is_dir():
+                    for sibling in parent.iterdir():
+                        if (
+                            sibling.is_file()
+                            and sibling.name.startswith(prefix)
+                            and sibling.name != stem
+                            and sibling.suffix == ".gguf"
+                        ):
+                            try:
+                                bytes_total += sibling.stat().st_size
+                            except OSError:
+                                pass
+            except OSError:
+                pass
+
+        # Read VmRSS from /proc/<pid>/status. Kilobytes on Linux.
+        bytes_loaded = 0
+        try:
+            with open(f"/proc/{pid}/status", "r", encoding = "utf-8") as f:
+                for line in f:
+                    if line.startswith("VmRSS:"):
+                        kb = int(line.split()[1])
+                        bytes_loaded = kb * 1024
+                        break
+        except (FileNotFoundError, PermissionError, ValueError, OSError):
+            return None
+
+        phase = "ready" if self._healthy else "mmap"
+        fraction = 0.0
+        if bytes_total > 0:
+            fraction = min(1.0, bytes_loaded / bytes_total)
+        return {
+            "phase": phase,
+            "bytes_loaded": bytes_loaded,
+            "bytes_total": bytes_total,
+            "fraction": round(fraction, 4),
+        }
+
     @property
     def chat_template(self) -> Optional[str]:
         return self._chat_template
@@ -1574,7 +1664,12 @@ class LlamaCppBackend:
             )
             self._stdout_thread.start()
 
-            self._gguf_path = gguf_path
+            # Store the resolved on-disk path, not the caller's kwarg. In
+            # HF mode the caller passes gguf_path=None and the real path
+            # (``model_path``) is what llama-server is actually mmap'ing.
+            # Downstream consumers (load_progress, log lines, etc.) need
+            # the path that exists on disk.
+            self._gguf_path = model_path
             self._hf_repo = hf_repo
             # For local GGUF files, extract variant from filename if not provided
             if hf_variant:
diff --git a/studio/backend/models/inference.py b/studio/backend/models/inference.py
index 8d9dc9830d..4917a14579 100644
--- a/studio/backend/models/inference.py
+++ b/studio/backend/models/inference.py
@@ -188,6 +188,39 @@ class UnloadResponse(BaseModel):
     model: str = Field(..., description = "Model identifier that was unloaded")
 
 
+class LoadProgressResponse(BaseModel):
+    """Progress of the active GGUF load, sampled on demand.
+
+    Used by the UI to show a real progress bar during the
+    post-download warmup window (mmap + CUDA upload), rather than a
+    generic "Starting model..." spinner that freezes for minutes on
+    large MoE models.
+    """
+
+    phase: Optional[str] = Field(
+        None,
+        description = (
+            "Load phase: 'mmap' (weights paging into RAM via mmap), "
+            "'ready' (llama-server reported healthy), or null when no "
+            "load is in flight."
+        ),
+    )
+    bytes_loaded: int = Field(
+        0,
+        description = (
+            "Bytes of the model already resident in the llama-server "
+            "process (VmRSS on Linux)."
+        ),
+    )
+    bytes_total: int = Field(
+        0,
+        description = "Total bytes across all GGUF shards for the active model.",
+    )
+    fraction: float = Field(
+        0.0, description = "bytes_loaded / bytes_total, clamped to 0..1."
+    )
+
+
 class InferenceStatusResponse(BaseModel):
     """Current inference backend status"""
 
diff --git a/studio/backend/routes/inference.py b/studio/backend/routes/inference.py
index 0717b3bc93..4246f0056b 100644
--- a/studio/backend/routes/inference.py
+++ b/studio/backend/routes/inference.py
@@ -72,6 +72,7 @@ from models.inference import (
     UnloadRequest,
     GenerateRequest,
     LoadResponse,
+    LoadProgressResponse,
     UnloadResponse,
     InferenceStatusResponse,
     ChatCompletionRequest,
@@ -751,6 +752,34 @@ async def get_status(
         raise HTTPException(status_code = 500, detail = f"Failed to get status: {str(e)}")
 
 
+@router.get("/load-progress", response_model = LoadProgressResponse)
+async def get_load_progress(
+    current_subject: str = Depends(get_current_subject),
+):
+    """
+    Return the active GGUF load's mmap/upload progress.
+
+    During the warmup window after a GGUF download -- when llama-server
+    is paging ~tens-to-hundreds of GB of shards into the page cache
+    before pushing layers to VRAM -- ``/api/inference/status`` only
+    shows a generic spinner. This endpoint exposes sampled progress so
+    the UI can render a real bar plus rate/ETA during that window.
+
+    Returns an empty payload (``phase=null, bytes=0``) when no load is
+    in flight. The frontend should stop polling once ``phase`` becomes
+    ``ready``.
+    """
+    try:
+        llama_backend = get_llama_cpp_backend()
+        progress = llama_backend.load_progress()
+        if progress is None:
+            return LoadProgressResponse()
+        return LoadProgressResponse(**progress)
+    except Exception as e:
+        logger.warning(f"Error sampling load progress: {e}")
+        return LoadProgressResponse()
+
+
 # =====================================================================
 # Audio (TTS) Generation  (/audio/generate)
 # =====================================================================
diff --git a/studio/backend/tests/test_llama_cpp_load_progress.py b/studio/backend/tests/test_llama_cpp_load_progress.py
new file mode 100644
index 0000000000..f46751b798
--- /dev/null
+++ b/studio/backend/tests/test_llama_cpp_load_progress.py
@@ -0,0 +1,258 @@
+# SPDX-License-Identifier: AGPL-3.0-only
+# Copyright 2026-present the Unsloth AI Inc. team. All rights reserved. See /studio/LICENSE.AGPL-3.0
+
+"""Tests for ``LlamaCppBackend.load_progress()``.
+
+The chat settings flow and the training overlay both show a generic
+"Starting model..." spinner during the window after a GGUF download
+finishes and before llama-server reports healthy. For small models
+that window is a second or two and nobody notices. For large MoE GGUFs
+(MiniMax-M2.7, Qwen3.5-397B-A17B, etc.) the llama-server process spends
+minutes in kernel state D, paging tens or hundreds of GB of shards
+into the page cache. The UI has no way to show a real progress bar,
+rate, or ETA during that window.
+
+``load_progress()`` samples ``/proc/<pid>/status VmRSS`` (what the
+kernel has actually paged in) against the total shard file size on
+disk, so the frontend can render a real bar plus rate/ETA. This
+module pins that contract:
+
+  * returns ``None`` when no load is in flight
+  * returns ``{"phase": "mmap", ...}`` while the subprocess is alive
+    but ``_healthy`` is False
+  * returns ``{"phase": "ready", ...}`` once ``_healthy`` flips
+  * ``bytes_total`` is derived from the resolved on-disk path
+    (which the paired fix assigns to ``self._gguf_path`` on both the
+    local-GGUF and HF-download code paths)
+  * ``bytes_loaded`` is VmRSS in bytes, capped by total, rounded
+  * ``fraction`` is clamped to 0..1 and rounded to 4 decimal places
+
+Linux-only via ``/proc``. On platforms without ``/proc`` the method
+returns ``None`` instead of raising.
+Cross-platform test: skips cleanly on macOS / Windows if ``/proc`` is
+not available.
+"""
+
+from __future__ import annotations
+
+import os
+import sys
+import tempfile
+import types as _types
+from pathlib import Path
+from unittest.mock import patch
+
+import pytest
+
+# ---------------------------------------------------------------------------
+# Stub heavy / unavailable external dependencies before importing the
+# module under test. Same pattern as test_kv_cache_estimation.py.
+# ---------------------------------------------------------------------------
+
+_BACKEND_DIR = str(Path(__file__).resolve().parent.parent)
+if _BACKEND_DIR not in sys.path:
+    sys.path.insert(0, _BACKEND_DIR)
+
+_loggers_stub = _types.ModuleType("loggers")
+_loggers_stub.get_logger = lambda name: __import__("logging").getLogger(name)
+sys.modules.setdefault("loggers", _loggers_stub)
+
+_structlog_stub = _types.ModuleType("structlog")
+sys.modules.setdefault("structlog", _structlog_stub)
+
+_httpx_stub = _types.ModuleType("httpx")
+for _exc_name in (
+    "ConnectError",
+    "TimeoutException",
+    "ReadTimeout",
+    "ReadError",
+    "RemoteProtocolError",
+    "CloseError",
+):
+    setattr(_httpx_stub, _exc_name, type(_exc_name, (Exception,), {}))
+
+
+class _FakeTimeout:
+    def __init__(self, *a, **kw):
+        pass
+
+
+_httpx_stub.Timeout = _FakeTimeout
+_httpx_stub.Client = type(
+    "Client",
+    (),
+    {
+        "__init__": lambda self, **kw: None,
+        "__enter__": lambda self: self,
+        "__exit__": lambda self, *a: None,
+    },
+)
+sys.modules.setdefault("httpx", _httpx_stub)
+
+from core.inference.llama_cpp import LlamaCppBackend
+
+
+# ---------------------------------------------------------------------------
+# Helpers
+# ---------------------------------------------------------------------------
+
+
+def _make_instance():
+    inst = LlamaCppBackend.__new__(LlamaCppBackend)
+    inst._process = None
+    inst._gguf_path = None
+    inst._healthy = False
+    return inst
+
+
+class _FakeProc:
+    """Minimal stand-in for subprocess.Popen that just carries a pid."""
+
+    def __init__(self, pid: int):
+        self.pid = pid
+
+
+def _write_sparse_file(path: Path, size_bytes: int) -> None:
+    """Create a sparse file of the given size without allocating blocks."""
+    with open(path, "wb") as fh:
+        if size_bytes > 0:
+            fh.truncate(size_bytes)
+
+
+# ---------------------------------------------------------------------------
+# Tests
+# ---------------------------------------------------------------------------
+
+
+class TestLoadProgressEmptyStates:
+    def test_returns_none_when_no_process(self):
+        inst = _make_instance()
+        assert inst.load_progress() is None
+
+    def test_returns_none_when_process_has_no_pid(self):
+        inst = _make_instance()
+        inst._process = _FakeProc(pid = None)  # type: ignore[arg-type]
+        assert inst.load_progress() is None
+
+
+class TestLoadProgressSingleShard:
+    def test_mmap_phase_for_alive_but_unhealthy(self, tmp_path):
+        """VmRSS below total -> phase='mmap', fraction reflects progress."""
+        gguf = tmp_path / "model.gguf"
+        _write_sparse_file(gguf, 40 * 1024**3)  # 40 GB
+
+        inst = _make_instance()
+        inst._process = _FakeProc(pid = os.getpid())  # use our own pid
+        inst._gguf_path = str(gguf)
+        inst._healthy = False
+
+        # Patch /proc read to claim 10 GB RSS.
+        def fake_open(path, *args, **kwargs):
+            if str(path).startswith("/proc/"):
+                import io
+
+                return io.StringIO(f"Name:\ttest\nVmRSS:\t{10 * 1024 ** 2}\tkB\n")
+            return open(path, *args, **kwargs)  # fall through
+
+        with patch("builtins.open", side_effect = fake_open):
+            out = inst.load_progress()
+
+        assert out is not None
+        assert out["phase"] == "mmap"
+        assert out["bytes_total"] == 40 * 1024**3
+        assert out["bytes_loaded"] == 10 * 1024**3
+        assert 0.24 < out["fraction"] < 0.26  # ~25%
+
+    def test_ready_phase_when_healthy(self, tmp_path):
+        gguf = tmp_path / "model.gguf"
+        _write_sparse_file(gguf, 8 * 1024**3)
+
+        inst = _make_instance()
+        inst._process = _FakeProc(pid = os.getpid())
+        inst._gguf_path = str(gguf)
+        inst._healthy = True
+
+        def fake_open(path, *args, **kwargs):
+            if str(path).startswith("/proc/"):
+                import io
+
+                return io.StringIO(f"VmRSS:\t{8 * 1024 ** 2}\tkB\n")
+            return open(path, *args, **kwargs)
+
+        with patch("builtins.open", side_effect = fake_open):
+            out = inst.load_progress()
+
+        assert out is not None
+        assert out["phase"] == "ready"
+        assert out["bytes_total"] == 8 * 1024**3
+        assert out["bytes_loaded"] == 8 * 1024**3
+        assert out["fraction"] == 1.0
+
+
+class TestLoadProgressMultiShard:
+    """Shard-aware total: for ``*-00001-of-00004.gguf`` primaries the
+    method sums sibling files with the same prefix."""
+
+    def test_sharded_total_aggregates_siblings(self, tmp_path):
+        for i in range(1, 5):
+            _write_sparse_file(
+                tmp_path / f"model-{i:05d}-of-00004.gguf",
+                size_bytes = 20 * 1024**3,
+            )
+        # Drop an unrelated .gguf in the same folder -- must not be counted.
+        _write_sparse_file(tmp_path / "mmproj-BF16.gguf", 2 * 1024**3)
+
+        inst = _make_instance()
+        inst._process = _FakeProc(pid = os.getpid())
+        inst._gguf_path = str(tmp_path / "model-00001-of-00004.gguf")
+        inst._healthy = False
+
+        def fake_open(path, *args, **kwargs):
+            if str(path).startswith("/proc/"):
+                import io
+
+                return io.StringIO("VmRSS:\t0\tkB\n")
+            return open(path, *args, **kwargs)
+
+        with patch("builtins.open", side_effect = fake_open):
+            out = inst.load_progress()
+
+        assert out is not None
+        assert out["bytes_total"] == 80 * 1024**3  # 4 x 20 GB, no mmproj
+
+
+class TestLoadProgressDegradation:
+    """Broken / unusual inputs never raise; they produce best-effort output."""
+
+    def test_missing_gguf_path_still_reports_rss(self, tmp_path):
+        inst = _make_instance()
+        inst._process = _FakeProc(pid = os.getpid())
+        inst._gguf_path = None
+        inst._healthy = False
+
+        def fake_open(path, *args, **kwargs):
+            if str(path).startswith("/proc/"):
+                import io
+
+                return io.StringIO("VmRSS:\t1024\tkB\n")
+            return open(path, *args, **kwargs)
+
+        with patch("builtins.open", side_effect = fake_open):
+            out = inst.load_progress()
+
+        assert out is not None
+        assert out["phase"] == "mmap"
+        assert out["bytes_total"] == 0
+        assert out["bytes_loaded"] == 1024 * 1024
+        assert out["fraction"] == 0.0
+
+    def test_unreadable_proc_returns_none(self, tmp_path):
+        inst = _make_instance()
+        # Pid that doesn't exist -> /proc read fails.
+        inst._process = _FakeProc(pid = 999_999_999)
+        inst._gguf_path = str(tmp_path / "model.gguf")  # doesn't need to exist
+        inst._healthy = False
+
+        out = inst.load_progress()
+        # FileNotFoundError on /proc path -> load_progress returns None.
+        assert out is None
diff --git a/studio/frontend/src/features/chat/api/chat-api.ts b/studio/frontend/src/features/chat/api/chat-api.ts
index 33e6e1eba0..ddc0e9d39e 100644
--- a/studio/frontend/src/features/chat/api/chat-api.ts
+++ b/studio/frontend/src/features/chat/api/chat-api.ts
@@ -145,6 +145,31 @@ export async function getDatasetDownloadProgress(
   return parseJsonOrThrow(response);
 }
 
+export type ModelLoadPhase = "mmap" | "ready" | null;
+
+export interface LoadProgressResponse {
+  /**
+   * Load phase: ``"mmap"`` while the llama-server subprocess is paging
+   * weight shards into RAM, ``"ready"`` once it has reported healthy,
+   * or ``null`` when no load is in flight.
+   */
+  phase: ModelLoadPhase;
+  bytes_loaded: number;
+  bytes_total: number;
+  fraction: number;
+}
+
+/**
+ * Fetch the active GGUF load's mmap/upload progress. Complements
+ * ``getDownloadProgress`` / ``getGgufDownloadProgress`` for the window
+ * between "download complete" and "chat ready", which for large MoE
+ * models can be several minutes of otherwise-opaque spinning.
+ */
+export async function getLoadProgress(): Promise<LoadProgressResponse> {
+  const response = await authFetch(`/api/inference/load-progress`);
+  return parseJsonOrThrow(response);
+}
+
 export interface LocalModelInfo {
   id: string;
   display_name: string;
diff --git a/studio/frontend/src/features/chat/hooks/use-chat-model-runtime.ts b/studio/frontend/src/features/chat/hooks/use-chat-model-runtime.ts
index 65b02be965..53f0d1d352 100644
--- a/studio/frontend/src/features/chat/hooks/use-chat-model-runtime.ts
+++ b/studio/frontend/src/features/chat/hooks/use-chat-model-runtime.ts
@@ -8,12 +8,14 @@ import {
   getDownloadProgress,
   getGgufDownloadProgress,
   getInferenceStatus,
+  getLoadProgress,
   listLoras,
   listModels,
   loadModel,
   unloadModel,
   validateModel,
 } from "../api/chat-api";
+import { formatEta, formatRate } from "../utils/format-transfer";
 import { useChatRuntimeStore } from "../stores/chat-runtime-store";
 import type { InferenceStatusResponse, LoadModelResponse } from "../types/api";
 import type {
@@ -565,77 +567,147 @@ export function useChatModelRuntime() {
         );
         loadToastIdRef.current = toastId;
 
-        // Poll download progress for non-cached models (GGUF and non-GGUF)
+        // Poll download progress for non-cached models (GGUF and non-GGUF).
+        // Then, once the download wraps (or for already-cached models),
+        // poll the llama-server mmap phase so "Starting model..." no
+        // longer looks frozen for several minutes on large MoE models.
         let progressInterval: ReturnType<typeof setInterval> | null = null;
-        if (!isDownloaded && !isCachedLora) {
-          const expectedBytes =
-            typeof selection !== "string" ? selection.expectedBytes ?? 0 : 0;
-          let hasShownProgress = false;
+        const expectedBytes =
+          typeof selection !== "string" ? selection.expectedBytes ?? 0 : 0;
 
-          const pollProgress = async () => {
-            if (abortCtrl.signal.aborted || !loadingModelRef.current) {
-              if (progressInterval) clearInterval(progressInterval);
-              return;
-            }
-            try {
-              const prog = ggufVariant && expectedBytes > 0
+        // Rolling window of byte samples for rate / ETA estimation.
+        // Shared across download + mmap phases so the estimator doesn't
+        // reset when the phase flips.
+        type Sample = { t: number; b: number };
+        const MIN_SAMPLES = 3;
+        const MIN_WINDOW = 3_000; // ms
+        const MAX_WINDOW = 15_000; // ms
+        const dlSamples: Sample[] = [];
+        const mmapSamples: Sample[] = [];
+
+        function estimate(
+          samples: Sample[],
+          bytes: number,
+          total: number,
+        ): { rate: number; eta: number; stable: boolean } {
+          const now = Date.now();
+          // Drop samples if the counter reset (e.g. phase flipped).
+          if (samples.length > 0 && bytes < samples[samples.length - 1].b) {
+            samples.length = 0;
+          }
+          samples.push({ t: now, b: bytes });
+          const cutoff = now - MAX_WINDOW;
+          while (samples.length > 2 && samples[0].t < cutoff) {
+            samples.shift();
+          }
+          if (samples.length < MIN_SAMPLES) {
+            return { rate: 0, eta: 0, stable: false };
+          }
+          const first = samples[0];
+          const last = samples[samples.length - 1];
+          const dt = (last.t - first.t) / 1000;
+          const db = last.b - first.b;
+          if (dt * 1000 < MIN_WINDOW || db <= 0) {
+            return { rate: 0, eta: 0, stable: false };
+          }
+          const rate = db / dt;
+          const eta =
+            total > 0 && bytes < total && rate > 0 ? (total - bytes) / rate : 0;
+          return { rate, eta, stable: true };
+        }
+
+        function composeProgressLabel(
+          dlGb: number,
+          totalGb: number,
+          bytes: number,
+          total: number,
+          samples: Sample[],
+        ): string {
+          const base =
+            totalGb > 0
+              ? `${dlGb.toFixed(1)} of ${totalGb.toFixed(1)} GB`
+              : `${dlGb.toFixed(1)} GB downloaded`;
+          const est = estimate(samples, bytes, total);
+          if (!est.stable) return base;
+          const rateStr = formatRate(est.rate);
+          const etaStr = total > 0 ? formatEta(est.eta) : "";
+          return etaStr && etaStr !== "--"
+            ? `${base} • ${rateStr} • ${etaStr} left`
+            : `${base} • ${rateStr}`;
+        }
+
+        let downloadComplete = isDownloaded || isCachedLora;
+
+        const pollDownload = async () => {
+          if (abortCtrl.signal.aborted || !loadingModelRef.current) {
+            if (progressInterval) clearInterval(progressInterval);
+            return;
+          }
+          try {
+            const prog =
+              ggufVariant && expectedBytes > 0
                 ? await getGgufDownloadProgress(modelId, ggufVariant, expectedBytes)
                 : await getDownloadProgress(modelId);
+            if (!loadingModelRef.current) return;
 
-              if (!loadingModelRef.current) return;
-
-              if (prog.progress > 0 && prog.progress < 1) {
-                hasShownProgress = true;
-                const dlGb = prog.downloaded_bytes / (1024 ** 3);
-                const totalGb = prog.expected_bytes / (1024 ** 3);
-                const pct = Math.round(prog.progress * 100);
-                const progressLabel = totalGb > 0
-                  ? `${dlGb.toFixed(1)} of ${totalGb.toFixed(1)} GB`
-                  : `${dlGb.toFixed(1)} GB downloaded`;
-                setLoadProgress({
-                  percent: pct,
-                  label: progressLabel,
-                  phase: "downloading",
-                });
-                if (loadToastDismissedRef.current) return;
-                toast(
-                  null,
-                  {
-                    id: toastId,
-                    description: renderLoadDescription(
-                      "Downloading model…",
-                      loadingDescription,
-                      pct,
-                      progressLabel,
-                      cancelLoading,
-                    ),
-                    duration: Infinity,
-                    closeButton: false,
-                    classNames: MODEL_LOAD_TOAST_CLASSNAMES,
-                    onDismiss: (dismissedToast) => {
-                      if (loadToastIdRef.current !== dismissedToast.id) return;
-                      setLoadToastDismissedState(true);
-                    },
-                  },
-                );
-              } else if (prog.downloaded_bytes > 0 && prog.expected_bytes === 0 && prog.progress === 0) {
-                hasShownProgress = true;
-                const dlGb = prog.downloaded_bytes / (1024 ** 3);
-                setLoadProgress({
-                  percent: null,
-                  label: `${dlGb.toFixed(1)} GB downloaded`,
-                  phase: "downloading",
-                });
-              } else if (prog.progress >= 1 && hasShownProgress) {
-                setLoadProgress({
-                  percent: 100,
-                  label: "Download complete",
-                  phase: "starting",
-                });
-                if (loadToastDismissedRef.current) {
-                  if (progressInterval) clearInterval(progressInterval);
-                  return;
-                }
+            if (prog.progress > 0 && prog.progress < 1) {
+              hasShownProgress = true;
+              const dlGb = prog.downloaded_bytes / (1024 ** 3);
+              const totalGb = prog.expected_bytes / (1024 ** 3);
+              const pct = Math.round(prog.progress * 100);
+              const progressLabel = composeProgressLabel(
+                dlGb,
+                totalGb,
+                prog.downloaded_bytes,
+                prog.expected_bytes,
+                dlSamples,
+              );
+              setLoadProgress({
+                percent: pct,
+                label: progressLabel,
+                phase: "downloading",
+              });
+              if (loadToastDismissedRef.current) return;
+              toast(null, {
+                id: toastId,
+                description: renderLoadDescription(
+                  "Downloading model…",
+                  loadingDescription,
+                  pct,
+                  progressLabel,
+                  cancelLoading,
+                ),
+                duration: Infinity,
+                closeButton: false,
+                classNames: MODEL_LOAD_TOAST_CLASSNAMES,
+                onDismiss: (dismissedToast) => {
+                  if (loadToastIdRef.current !== dismissedToast.id) return;
+                  setLoadToastDismissedState(true);
+                },
+              });
+            } else if (
+              prog.downloaded_bytes > 0 &&
+              prog.expected_bytes === 0 &&
+              prog.progress === 0
+            ) {
+              hasShownProgress = true;
+              const dlGb = prog.downloaded_bytes / (1024 ** 3);
+              const est = estimate(dlSamples, prog.downloaded_bytes, 0);
+              const rateSuffix =
+                est.stable ? ` • ${formatRate(est.rate)}` : "";
+              setLoadProgress({
+                percent: null,
+                label: `${dlGb.toFixed(1)} GB downloaded${rateSuffix}`,
+                phase: "downloading",
+              });
+            } else if (prog.progress >= 1 && hasShownProgress) {
+              downloadComplete = true;
+              setLoadProgress({
+                percent: 100,
+                label: "Download complete",
+                phase: "starting",
+              });
+              if (!loadToastDismissedRef.current) {
                 toast(null, {
                   id: toastId,
                   description: renderLoadDescription(
@@ -653,16 +725,79 @@ export function useChatModelRuntime() {
                     setLoadToastDismissedState(true);
                   },
                 });
-                if (progressInterval) clearInterval(progressInterval);
               }
-            } catch {
-              // Ignore polling errors
+              // Keep polling: the mmap branch below takes over from here.
             }
-          };
+          } catch {
+            // Ignore polling errors; keep polling.
+          }
+        };
 
-          setTimeout(pollProgress, 500);
-          progressInterval = setInterval(pollProgress, 2000);
-        }
+        const pollLoad = async () => {
+          if (abortCtrl.signal.aborted || !loadingModelRef.current) {
+            if (progressInterval) clearInterval(progressInterval);
+            return;
+          }
+          try {
+            const prog = await getLoadProgress();
+            if (!loadingModelRef.current) return;
+            if (!prog || prog.phase == null) return;
+            if (prog.phase === "ready") {
+              // Loaded. The chat flow will flip loadingModelRef shortly;
+              // just stop polling.
+              if (progressInterval) clearInterval(progressInterval);
+              return;
+            }
+            if (prog.bytes_total <= 0) return; // nothing useful to render
+            const loadedGb = prog.bytes_loaded / (1024 ** 3);
+            const totalGb = prog.bytes_total / (1024 ** 3);
+            const pct = Math.min(99, Math.round(prog.fraction * 100));
+            const est = estimate(mmapSamples, prog.bytes_loaded, prog.bytes_total);
+            const base = `${loadedGb.toFixed(1)} of ${totalGb.toFixed(1)} GB in memory`;
+            const label = est.stable
+              ? `${base} • ${formatRate(est.rate)}${
+                  formatEta(est.eta) !== "--" ? ` • ${formatEta(est.eta)} left` : ""
+                }`
+              : base;
+            setLoadProgress({
+              percent: pct,
+              label,
+              phase: "starting",
+            });
+            if (loadToastDismissedRef.current) return;
+            toast(null, {
+              id: toastId,
+              description: renderLoadDescription(
+                "Starting model…",
+                "Paging weights into memory.",
+                pct,
+                label,
+                cancelLoading,
+              ),
+              duration: Infinity,
+              closeButton: false,
+              classNames: MODEL_LOAD_TOAST_CLASSNAMES,
+              onDismiss: (dismissedToast) => {
+                if (loadToastIdRef.current !== dismissedToast.id) return;
+                setLoadToastDismissedState(true);
+              },
+            });
+          } catch {
+            // Ignore polling errors.
+          }
+        };
+
+        const pollProgress = async () => {
+          if (!downloadComplete) {
+            await pollDownload();
+          } else {
+            await pollLoad();
+          }
+        };
+
+        let hasShownProgress = false;
+        setTimeout(pollProgress, 500);
+        progressInterval = setInterval(pollProgress, 2000);
 
         try {
           await performLoad();
diff --git a/studio/frontend/src/features/chat/hooks/use-transfer-stats.ts b/studio/frontend/src/features/chat/hooks/use-transfer-stats.ts
new file mode 100644
index 0000000000..9ea7d1e625
--- /dev/null
+++ b/studio/frontend/src/features/chat/hooks/use-transfer-stats.ts
@@ -0,0 +1,98 @@
+// SPDX-License-Identifier: AGPL-3.0-only
+// Copyright 2026-present the Unsloth AI Inc. team. All rights reserved. See /studio/LICENSE.AGPL-3.0
+
+/**
+ * Compute rate (bytes/sec) and ETA (seconds) from a time-series of
+ * cumulative ``bytes`` values, using a rolling window of recent samples.
+ *
+ * Shared between the chat-flow download toast, the training-start
+ * overlay, and the model-load phase UI. All three have the same shape:
+ * a counter that rises monotonically from 0 toward ``totalBytes``, polled
+ * on an interval. The derived stats are identical regardless of whether
+ * the bytes came from an HTTP download or an mmap page-in.
+ *
+ * Stability rule: ``stable`` stays ``false`` until we've observed at
+ * least 3 samples spanning ≥3 seconds. That keeps the UI from flashing
+ * wildly varying rates during the first tick or two when the denominator
+ * is effectively zero.
+ */
+
+import { useEffect, useRef, useState } from "react";
+
+export type TransferStats = {
+  rateBytesPerSecond: number;
+  etaSeconds: number;
+  /**
+   * False for the first few ticks (window not filled, or no forward
+   * progress yet). Consumers should hide rate/ETA while unstable so the
+   * UI doesn't flicker "123 GB/s" during the first tick.
+   */
+  stable: boolean;
+};
+
+const MIN_SAMPLES = 3;
+const MIN_WINDOW_SECONDS = 3;
+const MAX_WINDOW_SECONDS = 15;
+
+export function useTransferStats(
+  bytes: number | null | undefined,
+  totalBytes: number | null | undefined,
+): TransferStats {
+  const samplesRef = useRef<{ t: number; b: number }[]>([]);
+  const [state, setState] = useState<TransferStats>({
+    rateBytesPerSecond: 0,
+    etaSeconds: 0,
+    stable: false,
+  });
+
+  useEffect(() => {
+    const now = Date.now() / 1000;
+    const cur = typeof bytes === "number" && Number.isFinite(bytes) ? bytes : 0;
+    const total =
+      typeof totalBytes === "number" && Number.isFinite(totalBytes)
+        ? totalBytes
+        : 0;
+
+    // If the counter resets (e.g. user unloaded and started a new
+    // download), drop the stale window.
+    const samples = samplesRef.current;
+    if (samples.length > 0 && cur < samples[samples.length - 1].b) {
+      samples.length = 0;
+    }
+
+    samples.push({ t: now, b: cur });
+
+    // Drop samples older than MAX_WINDOW_SECONDS; keep at least 2 so we
+    // can still compute a rate when the counter hasn't moved in a while.
+    const cutoff = now - MAX_WINDOW_SECONDS;
+    while (samples.length > 2 && samples[0].t < cutoff) {
+      samples.shift();
+    }
+
+    if (samples.length < MIN_SAMPLES) {
+      setState({ rateBytesPerSecond: 0, etaSeconds: 0, stable: false });
+      return;
+    }
+
+    const first = samples[0];
+    const last = samples[samples.length - 1];
+    const dt = last.t - first.t;
+    const db = last.b - first.b;
+    if (dt < MIN_WINDOW_SECONDS || db <= 0) {
+      setState({ rateBytesPerSecond: 0, etaSeconds: 0, stable: false });
+      return;
+    }
+
+    const rate = db / dt;
+    const eta =
+      total > 0 && cur < total && rate > 0 ? (total - cur) / rate : 0;
+
+    setState({
+      rateBytesPerSecond: rate,
+      etaSeconds: eta,
+      stable: true,
+    });
+  }, [bytes, totalBytes]);
+
+  return state;
+}
diff --git a/studio/frontend/src/features/chat/utils/format-transfer.ts b/studio/frontend/src/features/chat/utils/format-transfer.ts
new file mode 100644
index 0000000000..737f26c8cb
--- /dev/null
+++ b/studio/frontend/src/features/chat/utils/format-transfer.ts
@@ -0,0 +1,44 @@
+// SPDX-License-Identifier: AGPL-3.0-only
+// Copyright 2026-present the Unsloth AI Inc. team. All rights reserved. See /studio/LICENSE.AGPL-3.0
+
+/**
+ * Format a byte-per-second rate as a human-readable string.
+ *
+ *   512            → "512 B/s"
+ *   1_234_567      → "1.2 MB/s"
+ *   1_234_567_890  → "1.15 GB/s"
+ *
+ * Returns `"--"` for non-finite or non-positive inputs so callers can
+ * render the label safely before the first stable sample arrives.
+ */
+export function formatRate(bytesPerSecond: number): string {
+  if (!Number.isFinite(bytesPerSecond) || bytesPerSecond <= 0) return "--";
+  const bps = bytesPerSecond;
+  if (bps < 1024) return `${bps.toFixed(0)} B/s`;
+  if (bps < 1024 ** 2) return `${(bps / 1024).toFixed(1)} KB/s`;
+  if (bps < 1024 ** 3) return `${(bps / 1024 ** 2).toFixed(1)} MB/s`;
+  return `${(bps / 1024 ** 3).toFixed(2)} GB/s`;
+}
+
+/**
+ * Format an ETA (in seconds) as a short human-readable string.
+ *
+ *   47       → "47s"
+ *   125      → "2m 5s"
+ *   3725     → "1h 2m"
+ *
+ * Returns `"--"` for non-finite or non-positive inputs.
+ */
+export function formatEta(seconds: number): string {
+  if (!Number.isFinite(seconds) || seconds <= 0) return "--";
+  const s = Math.round(seconds);
+  if (s < 60) return `${s}s`;
+  if (s < 3600) {
+    const m = Math.floor(s / 60);
+    const rem = s % 60;
+    return rem > 0 ? `${m}m ${rem}s` : `${m}m`;
+  }
+  const h = Math.floor(s / 3600);
+  const m = Math.floor((s % 3600) / 60);
+  return m > 0 ? `${h}h ${m}m` : `${h}h`;
+}
diff --git a/studio/frontend/src/features/studio/training-start-overlay.tsx b/studio/frontend/src/features/studio/training-start-overlay.tsx
index f73a221d7d..b6042d4ae6 100644
--- a/studio/frontend/src/features/studio/training-start-overlay.tsx
+++ b/studio/frontend/src/features/studio/training-start-overlay.tsx
@@ -23,6 +23,8 @@ import {
   getDownloadProgress,
   type DownloadProgressResponse,
 } from "@/features/chat/api/chat-api";
+import { useTransferStats } from "@/features/chat/hooks/use-transfer-stats";
+import { formatEta, formatRate } from "@/features/chat/utils/format-transfer";
 import {
   useTrainingActions,
   useTrainingConfigStore,
@@ -151,6 +153,11 @@ type DownloadRowProps = {
 };
 
 function DownloadRow({ label, state }: DownloadRowProps): ReactElement | null {
+  // Compute a rolling-window rate + ETA from the same cumulative-byte
+  // series the poll hook already produces, so we can show
+  // "5.2 / 20.7 GB • 85.3 MB/s • 3m 12s left" instead of just the pair.
+  const stats = useTransferStats(state.downloadedBytes, state.totalBytes);
+
   if (state.downloadedBytes <= 0 && !state.cachePath) return null;
   const isComplete = state.totalBytes > 0 && state.percent >= 100;
   const statusLabel = isComplete
@@ -160,11 +167,16 @@ function DownloadRow({ label, state }: DownloadRowProps): ReactElement | null {
       : state.downloadedBytes === 0
         ? "Preparing"
         : null;
+  const showRate = stats.stable && !isComplete;
+  const rateSuffix = showRate ? ` • ${formatRate(stats.rateBytesPerSecond)}` : "";
+  const etaStr =
+    showRate && state.totalBytes > 0 ? formatEta(stats.etaSeconds) : "--";
+  const etaSuffix = etaStr !== "--" ? ` • ${etaStr} left` : "";
   const sizeLabel =
     state.totalBytes > 0
-      ? `${formatBytes(state.downloadedBytes)} / ${formatBytes(state.totalBytes)}`
+      ? `${formatBytes(state.downloadedBytes)} / ${formatBytes(state.totalBytes)}${rateSuffix}${etaSuffix}`
       : state.downloadedBytes > 0
-        ? `${formatBytes(state.downloadedBytes)} downloaded`
+        ? `${formatBytes(state.downloadedBytes)} downloaded${rateSuffix}`
         : null;
   return (
     <div className="flex flex-col gap-1.5 rounded-md border border-border/50 bg-muted/20 px-3 py-2">

From a0b9d140812f9490aff714151c8e1d409d30c4df Mon Sep 17 00:00:00 2001
From: "pre-commit-ci[bot]"
 <66853113+pre-commit-ci[bot]@users.noreply.github.com>
Date: Tue, 14 Apr 2026 09:49:18 -0700
Subject: [PATCH 094/216] [pre-commit.ci] pre-commit autoupdate (#5004)
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

updates:
- [github.com/astral-sh/ruff-pre-commit: v0.15.9 → v0.15.10](https://github.com/astral-sh/ruff-pre-commit/compare/v0.15.9...v0.15.10)

Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com>
---
 .pre-commit-config.yaml | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/.pre-commit-config.yaml b/.pre-commit-config.yaml
index e41c37d209..309c7d9aca 100644
--- a/.pre-commit-config.yaml
+++ b/.pre-commit-config.yaml
@@ -1,6 +1,6 @@
 repos:
   - repo: https://github.com/astral-sh/ruff-pre-commit
-    rev: v0.15.9
+    rev: v0.15.10
     hooks:
       - id: ruff
         args:

From 5b8dbdc3c264772f6d09eeade6ede544be7ef3f5 Mon Sep 17 00:00:00 2001
From: Eda Z <36808054+edamamez@users.noreply.github.com>
Date: Tue, 14 Apr 2026 10:23:40 -0700
Subject: [PATCH 095/216] Fix bitsandbytes ROCm install by using pip instead of
 uv (#4966)

* Fix bitsandbytes ROCm install by using pip instead of uv

* Also use pip for PyPI fallback path in _install_bnb_rocm

The original fix correctly switched the pre-release wheel install from
uv to pip, but left the PyPI fallback path on uv. If uv breaks bnb
on ROCm, the fallback would hit the same issue. Move pip bootstrap
before the branch so both paths use pip consistently.

* Harden pip bootstrap: try ensurepip first, warn on failure

- Try ensurepip --upgrade before falling back to uv pip install pip.
  ensurepip works offline and does not need PyPI, making the bootstrap
  robust when the network or index is unavailable.
- If both ensurepip and uv fail, emit a visible warning instead of
  silently swallowing the error (which previously led to a cryptic
  "No module named pip" downstream).
- Use run_maybe_quiet so --verbose users see bootstrap output.
- Update comment to document the actual root cause: uv rejects the
  wheel because filename version and metadata version disagree.

* Add --isolated to pip install calls in _install_bnb_rocm

uv pip install ignores pip.conf and PIP_* env vars, but python -m pip
reads them. Without --isolated, users with PIP_INDEX_URL pointing to a
private mirror that does not carry bitsandbytes would see the PyPI
fallback fail where it previously worked under uv. --isolated restores
parity with the old uv behavior.

* Drop --isolated from PyPI fallback in _install_bnb_rocm

--isolated suppresses PIP_INDEX_URL, PIP_EXTRA_INDEX_URL, and pip.conf.
This is correct for the pre-release path (hardcoded GitHub URL, no index
consulted), but breaks the PyPI fallback for users in corporate or
air-gapped environments whose only route to bitsandbytes is a private
mirror configured via those mechanisms. Keep --isolated on the direct-URL
pre-release install; drop it from the index-dependent fallback.

* Drop --isolated from pre-release pip install, fix warning wording

--isolated suppresses pip.conf cert/proxy/CA settings in addition to
index config. For the direct GitHub URL, index config is irrelevant but
cert/proxy settings matter in corporate SSL-inspection environments.
Without this fix, users with pip.conf-based CA bundles get a TLS error
on the pre-release download and silently fall back to the broken PyPI
version -- the exact outcome the PR is trying to prevent.

Also fix the fallback warning: "unreachable" is too specific since the
pre-release install can fail for reasons other than network reachability.

---------

Co-authored-by: Daniel Han <danielhanchen@gmail.com>
---
 install.sh | 15 ++++++++++++---
 1 file changed, 12 insertions(+), 3 deletions(-)

diff --git a/install.sh b/install.sh
index cc8721710e..56af1454c5 100755
--- a/install.sh
+++ b/install.sh
@@ -112,15 +112,24 @@ _install_bnb_rocm() {
             _bnb_whl_url=""
             ;;
     esac
+    # uv rejects the continuous-release_main bitsandbytes wheel because the
+    # filename version (1.33.7rc0) does not match the embedded metadata version
+    # (0.50.0.dev0). pip accepts the mismatch, so bootstrap pip and use it.
+    if ! "$_venv_py" -m pip --version >/dev/null 2>&1; then
+        if ! run_maybe_quiet "$_venv_py" -m ensurepip --upgrade; then
+            run_maybe_quiet uv pip install --python "$_venv_py" pip || \
+                substep "[WARN] could not bootstrap pip; bitsandbytes install will likely fail" "$C_WARN"
+        fi
+    fi
     if [ -n "$_bnb_whl_url" ]; then
         substep "installing bitsandbytes for AMD ROCm (pre-release, PR #1887)..."
-        if run_install_cmd "$_label (pre-release)" uv pip install --python "$_venv_py" \
+        if run_install_cmd "$_label (pre-release)" "$_venv_py" -m pip install \
             --force-reinstall --no-cache-dir --no-deps "$_bnb_whl_url"; then
             return 0
         fi
-        substep "[WARN] bnb pre-release unreachable; falling back to PyPI (4-bit decode broken on ROCm)" "$C_WARN"
+        substep "[WARN] bnb pre-release install failed; falling back to PyPI (4-bit decode broken on ROCm)" "$C_WARN"
     fi
-    run_install_cmd "$_label (pypi fallback)" uv pip install --python "$_venv_py" \
+    run_install_cmd "$_label (pypi fallback)" "$_venv_py" -m pip install \
         --force-reinstall --no-cache-dir --no-deps "bitsandbytes>=0.49.1"
 }
 

From 5861a7ce1527b28c18cceb043d5719844ab68793 Mon Sep 17 00:00:00 2001
From: Daniel Han <danielhanchen@gmail.com>
Date: Tue, 14 Apr 2026 10:58:16 -0700
Subject: [PATCH 096/216] Studio: split model-load progress label across two
 rows (#5020)
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

* Studio: split model-load progress label across two rows

The chat flow and training overlay both compose a progress label like
"112.6 of 122.3 GB • 331.0 MB/s • 30s left" and render it next to the
percent badge in a single flex row. Once the rate + ETA part shows up,
the label outgrows the row width and wraps mid-phrase, orphaning the
percent ("19 left %") onto a second ragged line.

Fix in model-load-status.tsx: split the label on the first " • " into
a primary (size) chunk that stays on row 1 with the percent, and a
secondary (rate/ETA) chunk that renders on its own muted row below.
Labels without a bullet (e.g. "22.8 GB downloaded") collapse cleanly
to one row. The inline-status variant keeps only the primary and
surfaces the full label via the tooltip.

Also extracts the rate/ETA math out of useTransferStats into a pure
``transfer-stats.ts`` module (appendSample + computeTransferStats) so
it can be reasoned about and tested without React. The hook is now a
thin wrapper that feeds sample history through the pure functions.

Backend: adds two companion test files for load_progress():

  * test_llama_cpp_load_progress_matrix.py (21 tests) -- platform
    matrix (Linux /proc, macOS/Windows absence), VmRSS parsing
    variants (tab/space/missing/malformed), filesystem edges (HF-cache
    symlinks, broken symlinks, nonexistent paths, relative paths),
    shard aggregation (partial multi-shard, two series in same dir,
    mmproj-* exclusion, single-file), lifecycle races, concurrent
    sampling (10 threads x 50 iters against real /proc), fraction
    bounds.
  * test_llama_cpp_load_progress_live.py (5 tests) -- no-mock live
    integration: real subprocess allocating 100 MB to match VmRSS,
    real ready phase, real dead-pid degradation, real shard
    aggregation, repeated polling. Skipped on non-Linux.

Both complement the existing test_llama_cpp_load_progress.py.

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

* Hoist splitProgressLabel out of JSX IIFE (review feedback)

---------

Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com>
---
 .../test_llama_cpp_load_progress_live.py      | 202 ++++++++
 .../test_llama_cpp_load_progress_matrix.py    | 473 ++++++++++++++++++
 .../chat/components/model-load-status.tsx     |  59 ++-
 .../features/chat/hooks/use-transfer-stats.ts |  63 +--
 .../src/features/chat/utils/transfer-stats.ts |  88 ++++
 5 files changed, 825 insertions(+), 60 deletions(-)
 create mode 100644 studio/backend/tests/test_llama_cpp_load_progress_live.py
 create mode 100644 studio/backend/tests/test_llama_cpp_load_progress_matrix.py
 create mode 100644 studio/frontend/src/features/chat/utils/transfer-stats.ts

diff --git a/studio/backend/tests/test_llama_cpp_load_progress_live.py b/studio/backend/tests/test_llama_cpp_load_progress_live.py
new file mode 100644
index 0000000000..beed8713c1
--- /dev/null
+++ b/studio/backend/tests/test_llama_cpp_load_progress_live.py
@@ -0,0 +1,202 @@
+# SPDX-License-Identifier: AGPL-3.0-only
+# Copyright 2026-present the Unsloth AI Inc. team. All rights reserved. See /studio/LICENSE.AGPL-3.0
+
+"""Live, no-mock integration test for ``LlamaCppBackend.load_progress()``.
+
+The companion files (``test_llama_cpp_load_progress.py`` and
+``test_llama_cpp_load_progress_matrix.py``) patch ``builtins.open`` to
+feed synthetic VmRSS values. This file is the opposite: it uses **real**
+subprocesses, **real** file sizes, and the **real** ``/proc``
+interface. It is the sanity check that the contract we keep in the
+mocked tests still maps to what the kernel actually returns on a live
+Linux system.
+
+Why both: the mocked tests can be fooled by a buggy implementation that
+parses ``/proc`` output in a format the kernel no longer uses, or that
+makes assumptions about ``Path.stat()`` vs ``os.path.getsize``. This
+file hits the real APIs so any format drift gets caught.
+
+Skipped cleanly on non-Linux (no ``/proc``).
+"""
+
+from __future__ import annotations
+
+import os
+import subprocess
+import sys
+import time
+import types as _types
+from pathlib import Path
+
+import pytest
+
+# ---------------------------------------------------------------------------
+# Same stubs as the matrix file (keep self-contained so the file can be
+# run standalone as well as via the full suite).
+# ---------------------------------------------------------------------------
+
+_BACKEND_DIR = str(Path(__file__).resolve().parent.parent)
+if _BACKEND_DIR not in sys.path:
+    sys.path.insert(0, _BACKEND_DIR)
+
+_loggers_stub = _types.ModuleType("loggers")
+_loggers_stub.get_logger = lambda name: __import__("logging").getLogger(name)
+sys.modules.setdefault("loggers", _loggers_stub)
+_structlog_stub = _types.ModuleType("structlog")
+sys.modules.setdefault("structlog", _structlog_stub)
+_httpx_stub = _types.ModuleType("httpx")
+for _exc in (
+    "ConnectError",
+    "TimeoutException",
+    "ReadTimeout",
+    "ReadError",
+    "RemoteProtocolError",
+    "CloseError",
+):
+    setattr(_httpx_stub, _exc, type(_exc, (Exception,), {}))
+_httpx_stub.Timeout = type("Timeout", (), {"__init__": lambda self, *a, **k: None})
+_httpx_stub.Client = type(
+    "Client",
+    (),
+    {
+        "__init__": lambda self, **kw: None,
+        "__enter__": lambda self: self,
+        "__exit__": lambda self, *a: None,
+    },
+)
+sys.modules.setdefault("httpx", _httpx_stub)
+
+from core.inference.llama_cpp import LlamaCppBackend
+
+
+pytestmark = pytest.mark.skipif(
+    not Path("/proc").exists(),
+    reason = "live /proc test is Linux-only",
+)
+
+
+def _make_backend(pid: int, gguf_path: str, healthy: bool = False):
+    inst = LlamaCppBackend.__new__(LlamaCppBackend)
+    inst._process = type("P", (), {"pid": pid})()
+    inst._gguf_path = gguf_path
+    inst._healthy = healthy
+    return inst
+
+
+def test_live_rss_matches_kernel_vmrss(tmp_path):
+    """Spawn a real child, let it allocate real bytes, confirm
+    ``bytes_loaded`` tracks the kernel's VmRSS within a sane tolerance."""
+    # Child that allocates ~100 MB of zero'd bytes and then idles.
+    script = tmp_path / "burn.py"
+    script.write_text(
+        "import time, sys\n"
+        "buf = bytearray(100 * 1024 * 1024)\n"  # 100 MB
+        "# touch every page so RSS actually grows\n"
+        "for i in range(0, len(buf), 4096):\n"
+        "    buf[i] = 1\n"
+        "sys.stdout.write('ready\\n')\n"
+        "sys.stdout.flush()\n"
+        "time.sleep(10)\n"
+    )
+    proc = subprocess.Popen(
+        [sys.executable, str(script)],
+        stdout = subprocess.PIPE,
+        stderr = subprocess.PIPE,
+    )
+    try:
+        # Wait for the child to finish touching pages.
+        ready = proc.stdout.readline()
+        assert ready.strip() == b"ready"
+
+        # Create a fake 200 MB sparse gguf so bytes_total is concrete.
+        gguf = tmp_path / "model.gguf"
+        with open(gguf, "wb") as f:
+            f.truncate(200 * 1024 * 1024)
+
+        inst = _make_backend(proc.pid, str(gguf), healthy = False)
+        out = inst.load_progress()
+
+        assert out is not None, "load_progress returned None for live pid"
+        assert out["phase"] == "mmap"
+        assert out["bytes_total"] == 200 * 1024 * 1024
+        # VmRSS for the Python child includes the interpreter + the 100MB
+        # buffer, so a realistic floor is 50 MB and ceiling is 200 MB.
+        assert (
+            out["bytes_loaded"] >= 50 * 1024 * 1024
+        ), f"bytes_loaded unexpectedly low: {out['bytes_loaded']}"
+        assert out["bytes_loaded"] <= 200 * 1024 * 1024
+        assert 0.0 < out["fraction"] <= 1.0
+    finally:
+        proc.terminate()
+        try:
+            proc.wait(timeout = 5)
+        except subprocess.TimeoutExpired:
+            proc.kill()
+
+
+def test_live_ready_phase_when_healthy(tmp_path):
+    gguf = tmp_path / "m.gguf"
+    with open(gguf, "wb") as f:
+        f.truncate(1 * 1024 * 1024)
+
+    inst = _make_backend(os.getpid(), str(gguf), healthy = True)
+    out = inst.load_progress()
+    assert out is not None
+    assert out["phase"] == "ready"
+    assert out["bytes_total"] == 1 * 1024 * 1024
+    # Self-pid RSS is well above 1 MiB for CPython; fraction caps at 1.
+    assert out["fraction"] == 1.0
+
+
+def test_live_dead_pid_returns_none(tmp_path):
+    """A recently-dead pid may linger in /proc for ms; use a clearly
+    invalid id so the read reliably fails."""
+    gguf = tmp_path / "m.gguf"
+    gguf.touch()
+
+    inst = _make_backend(9_999_999_999, str(gguf), healthy = False)
+    out = inst.load_progress()
+    assert out is None
+
+
+def test_live_shard_aggregation_counts_real_files(tmp_path):
+    """With 4 real sibling shards on disk, ``bytes_total`` equals their
+    summed size to the byte."""
+    shard_size = 7 * 1024 * 1024  # 7 MB each
+    for i in range(1, 5):
+        f = tmp_path / f"model-{i:05d}-of-00004.gguf"
+        with open(f, "wb") as fh:
+            fh.truncate(shard_size)
+    # Unrelated file in same dir -- must not be counted.
+    with open(tmp_path / "config.json", "wb") as fh:
+        fh.truncate(123)
+
+    inst = _make_backend(
+        os.getpid(),
+        str(tmp_path / "model-00001-of-00004.gguf"),
+        healthy = False,
+    )
+    out = inst.load_progress()
+    assert out is not None
+    assert out["bytes_total"] == 4 * shard_size
+
+
+def test_live_repeated_polling_stays_sane(tmp_path):
+    """Sampling the same backend 20 times should not raise or produce
+    non-numeric output, even under normal kernel RSS jitter."""
+    gguf = tmp_path / "m.gguf"
+    with open(gguf, "wb") as f:
+        f.truncate(500 * 1024 * 1024)
+
+    inst = _make_backend(os.getpid(), str(gguf), healthy = False)
+    seen = []
+    for _ in range(20):
+        out = inst.load_progress()
+        assert out is not None
+        assert isinstance(out["bytes_loaded"], int)
+        assert isinstance(out["bytes_total"], int)
+        assert 0.0 <= out["fraction"] <= 1.0
+        seen.append(out["bytes_loaded"])
+        time.sleep(0.01)
+    # RSS of a healthy Python process doesn't go below ~5 MB.
+    assert min(seen) > 1 * 1024 * 1024
diff --git a/studio/backend/tests/test_llama_cpp_load_progress_matrix.py b/studio/backend/tests/test_llama_cpp_load_progress_matrix.py
new file mode 100644
index 0000000000..a88450ec0b
--- /dev/null
+++ b/studio/backend/tests/test_llama_cpp_load_progress_matrix.py
@@ -0,0 +1,473 @@
+# SPDX-License-Identifier: AGPL-3.0-only
+# Copyright 2026-present the Unsloth AI Inc. team. All rights reserved. See /studio/LICENSE.AGPL-3.0
+
+"""Extended test matrix for ``LlamaCppBackend.load_progress()``.
+
+Companion to ``test_llama_cpp_load_progress.py`` (which pins the basic
+contract). This file widens coverage to the edge cases that bit users
+or were hypothesized to bite them on cross-platform installs:
+
+  * Platform matrix — macOS/Windows simulation via ``/proc`` absence.
+  * ``VmRSS`` parsing — tab vs space delimiter, missing line, malformed
+    integer.
+  * Filesystem edges — HF-cache symlinks, broken symlinks, nonexistent
+    paths, relative paths.
+  * Shard aggregation — partial multi-shard downloads where some shards
+    are still ``.incomplete``, two shard series in the same dir,
+    ``mmproj-*.gguf`` sibling exclusion for non-sharded primaries,
+    single-file models.
+  * Lifecycle races — process set before ``_gguf_path`` is assigned,
+    process dead mid-sample, ``_healthy`` flipped to True.
+  * Concurrent sampling — 10 threads × 50 iterations against a single
+    backend, hitting real ``/proc`` (no mocks — see the note in
+    ``TestConcurrentSampling`` for why).
+  * Fraction bounds — capped at 1.0 when RSS exceeds total; 0.0 when
+    total is zero.
+
+All tests are Linux-only in practice (we stub ``/proc`` where needed).
+The stable subset runs in well under a second.
+"""
+
+from __future__ import annotations
+
+import io
+import os
+import sys
+import threading
+import types as _types
+from pathlib import Path
+from unittest.mock import patch
+
+import pytest
+
+# ---------------------------------------------------------------------------
+# Stub heavy / unavailable external dependencies before importing the
+# module under test. Same pattern as test_llama_cpp_load_progress.py.
+# ---------------------------------------------------------------------------
+
+_BACKEND_DIR = str(Path(__file__).resolve().parent.parent)
+if _BACKEND_DIR not in sys.path:
+    sys.path.insert(0, _BACKEND_DIR)
+
+_loggers_stub = _types.ModuleType("loggers")
+_loggers_stub.get_logger = lambda name: __import__("logging").getLogger(name)
+sys.modules.setdefault("loggers", _loggers_stub)
+
+_structlog_stub = _types.ModuleType("structlog")
+sys.modules.setdefault("structlog", _structlog_stub)
+
+_httpx_stub = _types.ModuleType("httpx")
+for _exc_name in (
+    "ConnectError",
+    "TimeoutException",
+    "ReadTimeout",
+    "ReadError",
+    "RemoteProtocolError",
+    "CloseError",
+):
+    setattr(_httpx_stub, _exc_name, type(_exc_name, (Exception,), {}))
+
+
+class _FakeTimeout:
+    def __init__(self, *a, **kw):
+        pass
+
+
+_httpx_stub.Timeout = _FakeTimeout
+_httpx_stub.Client = type(
+    "Client",
+    (),
+    {
+        "__init__": lambda self, **kw: None,
+        "__enter__": lambda self: self,
+        "__exit__": lambda self, *a: None,
+    },
+)
+sys.modules.setdefault("httpx", _httpx_stub)
+
+from core.inference.llama_cpp import LlamaCppBackend
+
+
+# ---------------------------------------------------------------------------
+# Helpers
+# ---------------------------------------------------------------------------
+
+
+def _make():
+    inst = LlamaCppBackend.__new__(LlamaCppBackend)
+    inst._process = None
+    inst._gguf_path = None
+    inst._healthy = False
+    return inst
+
+
+class _Proc:
+    def __init__(self, pid):
+        self.pid = pid
+
+
+def _sparse(path, size):
+    with open(path, "wb") as f:
+        if size > 0:
+            f.truncate(size)
+
+
+def _fake_proc_reader(rss_kb):
+    """Return an ``open()`` replacement that fakes /proc reads with a VmRSS line."""
+
+    def fake_open(path, *args, **kwargs):
+        if str(path).startswith("/proc/"):
+            return io.StringIO(f"VmRSS:\t{rss_kb}\tkB\n")
+        return open(path, *args, **kwargs)
+
+    return fake_open
+
+
+# ---------------------------------------------------------------------------
+# A. Platform matrix
+# ---------------------------------------------------------------------------
+
+
+class TestPlatformMatrix:
+    """The method is Linux-first via /proc. On macOS/Windows it must
+    degrade to None rather than crash."""
+
+    def test_linux_live_proc_is_self_pid(self, tmp_path):
+        """Self-pid /proc read uses the real kernel interface."""
+        gguf = tmp_path / "m.gguf"
+        _sparse(gguf, 1 * 1024**3)
+        inst = _make()
+        inst._process = _Proc(os.getpid())
+        inst._gguf_path = str(gguf)
+        inst._healthy = False
+        out = inst.load_progress()
+        assert out is not None
+        assert out["phase"] == "mmap"
+        assert out["bytes_total"] == 1 * 1024**3
+        # Our Python process has some RSS -- just sanity-check positive.
+        assert out["bytes_loaded"] > 0
+
+    def test_macos_no_proc_returns_none(self, tmp_path):
+        """Simulate macOS: /proc open fails with FileNotFoundError."""
+        gguf = tmp_path / "m.gguf"
+        _sparse(gguf, 1 * 1024**3)
+        inst = _make()
+        inst._process = _Proc(pid = 12345)
+        inst._gguf_path = str(gguf)
+
+        def fake_open(path, *args, **kwargs):
+            if str(path).startswith("/proc/"):
+                raise FileNotFoundError(f"No such file: {path}")
+            return open(path, *args, **kwargs)
+
+        with patch("builtins.open", side_effect = fake_open):
+            out = inst.load_progress()
+        assert out is None
+
+    def test_windows_no_proc_returns_none(self, tmp_path):
+        """Simulate Windows: opening /proc raises PermissionError or OSError."""
+        gguf = tmp_path / "m.gguf"
+        _sparse(gguf, 1 * 1024**3)
+        inst = _make()
+        inst._process = _Proc(pid = 4567)
+        inst._gguf_path = str(gguf)
+
+        def fake_open(path, *args, **kwargs):
+            if str(path).startswith("/proc/"):
+                raise PermissionError("access denied")
+            return open(path, *args, **kwargs)
+
+        with patch("builtins.open", side_effect = fake_open):
+            out = inst.load_progress()
+        assert out is None
+
+
+# ---------------------------------------------------------------------------
+# B. VmRSS parsing edge cases
+# ---------------------------------------------------------------------------
+
+
+class TestVmRSSParsing:
+    def test_standard_tab_delimited(self, tmp_path):
+        gguf = tmp_path / "m.gguf"
+        _sparse(gguf, 4 * 1024**3)
+        inst = _make()
+        inst._process = _Proc(os.getpid())
+        inst._gguf_path = str(gguf)
+        with patch("builtins.open", side_effect = _fake_proc_reader(2 * 1024**2)):
+            out = inst.load_progress()
+        assert out["bytes_loaded"] == 2 * 1024**3
+
+    def test_space_separated_fallback(self, tmp_path):
+        """Some kernels emit single-space rather than tab."""
+        gguf = tmp_path / "m.gguf"
+        _sparse(gguf, 4 * 1024**3)
+        inst = _make()
+        inst._process = _Proc(os.getpid())
+        inst._gguf_path = str(gguf)
+
+        def fake_open(path, *a, **kw):
+            if str(path).startswith("/proc/"):
+                return io.StringIO("VmRSS: 4194304 kB\n")
+            return open(path, *a, **kw)
+
+        with patch("builtins.open", side_effect = fake_open):
+            out = inst.load_progress()
+        assert out["bytes_loaded"] == 4 * 1024**3
+
+    def test_missing_vmrss_line(self, tmp_path):
+        """Kernel with VmRSS stripped (zombie / kthread) -> 0."""
+        gguf = tmp_path / "m.gguf"
+        _sparse(gguf, 1 * 1024**3)
+        inst = _make()
+        inst._process = _Proc(os.getpid())
+        inst._gguf_path = str(gguf)
+
+        def fake_open(path, *a, **kw):
+            if str(path).startswith("/proc/"):
+                return io.StringIO("Name:\ttest\nState:\tZ (zombie)\n")
+            return open(path, *a, **kw)
+
+        with patch("builtins.open", side_effect = fake_open):
+            out = inst.load_progress()
+        assert out is not None
+        assert out["bytes_loaded"] == 0
+        assert out["fraction"] == 0.0
+
+    def test_malformed_vmrss_value(self, tmp_path):
+        """Non-integer VmRSS value should be treated as if the line were
+        absent (early ValueError caught)."""
+        gguf = tmp_path / "m.gguf"
+        _sparse(gguf, 1 * 1024**3)
+        inst = _make()
+        inst._process = _Proc(os.getpid())
+        inst._gguf_path = str(gguf)
+
+        def fake_open(path, *a, **kw):
+            if str(path).startswith("/proc/"):
+                return io.StringIO("VmRSS:\tXXXX\tkB\n")
+            return open(path, *a, **kw)
+
+        with patch("builtins.open", side_effect = fake_open):
+            out = inst.load_progress()
+        # The implementation catches ValueError on int() and returns None.
+        assert out is None
+
+
+# ---------------------------------------------------------------------------
+# C. Filesystem edge cases
+# ---------------------------------------------------------------------------
+
+
+class TestFilesystemEdges:
+    def test_symlink_primary_follows_to_blob(self, tmp_path):
+        """HF cache stores blobs under blobs/ and symlinks them from
+        snapshots/. The method must follow the symlink."""
+        blob = tmp_path / "blob"
+        _sparse(blob, 12 * 1024**3)
+        snap = tmp_path / "snap"
+        snap.mkdir()
+        link = snap / "m.gguf"
+        link.symlink_to(blob)
+
+        inst = _make()
+        inst._process = _Proc(os.getpid())
+        inst._gguf_path = str(link)
+        with patch("builtins.open", side_effect = _fake_proc_reader(6 * 1024**2)):
+            out = inst.load_progress()
+        assert out["bytes_total"] == 12 * 1024**3
+
+    def test_broken_symlink_skipped(self, tmp_path):
+        snap = tmp_path / "snap"
+        snap.mkdir()
+        link = snap / "m.gguf"
+        link.symlink_to(tmp_path / "missing-blob")
+        inst = _make()
+        inst._process = _Proc(os.getpid())
+        inst._gguf_path = str(link)
+        with patch("builtins.open", side_effect = _fake_proc_reader(1024)):
+            out = inst.load_progress()
+        assert out["bytes_total"] == 0
+        assert out["bytes_loaded"] == 1024 * 1024
+
+    def test_nonexistent_path_skipped(self, tmp_path):
+        inst = _make()
+        inst._process = _Proc(os.getpid())
+        inst._gguf_path = str(tmp_path / "ghost.gguf")
+        with patch("builtins.open", side_effect = _fake_proc_reader(1024)):
+            out = inst.load_progress()
+        assert out["bytes_total"] == 0
+
+    def test_relative_gguf_path(self, tmp_path):
+        """Relative paths shouldn't crash; behaviour depends on CWD but
+        the method must not raise."""
+        cwd = os.getcwd()
+        try:
+            os.chdir(tmp_path)
+            _sparse(Path("rel.gguf"), 8 * 1024**3)
+            inst = _make()
+            inst._process = _Proc(os.getpid())
+            inst._gguf_path = "rel.gguf"
+            with patch("builtins.open", side_effect = _fake_proc_reader(0)):
+                out = inst.load_progress()
+            assert out is not None
+            assert out["bytes_total"] == 8 * 1024**3
+        finally:
+            os.chdir(cwd)
+
+
+# ---------------------------------------------------------------------------
+# D. Shard aggregation
+# ---------------------------------------------------------------------------
+
+
+class TestShardAggregation:
+    def test_partial_multi_shard_download(self, tmp_path):
+        """Primary present but shards 2..N still downloading as
+        ``.incomplete``. Sums only the fully-arrived ``.gguf`` files."""
+        _sparse(tmp_path / "m-00001-of-00004.gguf", 30 * 1024**3)
+        _sparse(tmp_path / "m-00002-of-00004.gguf", 30 * 1024**3)
+        # 3 and 4 still downloading as .incomplete
+        _sparse(tmp_path / "m-00003-of-00004.gguf.incomplete", 5 * 1024**3)
+        inst = _make()
+        inst._process = _Proc(os.getpid())
+        inst._gguf_path = str(tmp_path / "m-00001-of-00004.gguf")
+        with patch("builtins.open", side_effect = _fake_proc_reader(0)):
+            out = inst.load_progress()
+        assert out["bytes_total"] == 60 * 1024**3  # only the .gguf siblings
+
+    def test_two_shard_series_in_same_dir(self, tmp_path):
+        """Defensive: if two quant series share a dir, prefix filter
+        only sums siblings of the chosen primary."""
+        for i in range(1, 3):
+            _sparse(tmp_path / f"m_q4-{i:05d}-of-00002.gguf", 10 * 1024**3)
+            _sparse(tmp_path / f"m_q8-{i:05d}-of-00002.gguf", 20 * 1024**3)
+        inst = _make()
+        inst._process = _Proc(os.getpid())
+        inst._gguf_path = str(tmp_path / "m_q8-00001-of-00002.gguf")
+        with patch("builtins.open", side_effect = _fake_proc_reader(0)):
+            out = inst.load_progress()
+        assert out["bytes_total"] == 40 * 1024**3  # just q8 series
+
+    def test_mmproj_sibling_not_counted(self, tmp_path):
+        """Vision models drop an ``mmproj-*.gguf`` alongside. For a
+        single-file (non-sharded) primary we only count the primary."""
+        _sparse(tmp_path / "m.gguf", 8 * 1024**3)
+        _sparse(tmp_path / "mmproj-BF16.gguf", 2 * 1024**3)
+        inst = _make()
+        inst._process = _Proc(os.getpid())
+        inst._gguf_path = str(tmp_path / "m.gguf")
+        with patch("builtins.open", side_effect = _fake_proc_reader(0)):
+            out = inst.load_progress()
+        # Non-sharded primary: only the primary is counted.
+        assert out["bytes_total"] == 8 * 1024**3
+
+    def test_single_file_model(self, tmp_path):
+        """Non-sharded model: primary only."""
+        _sparse(tmp_path / "small.gguf", 4 * 1024**3)
+        inst = _make()
+        inst._process = _Proc(os.getpid())
+        inst._gguf_path = str(tmp_path / "small.gguf")
+        with patch("builtins.open", side_effect = _fake_proc_reader(2 * 1024**2)):
+            out = inst.load_progress()
+        assert out["bytes_total"] == 4 * 1024**3
+        assert out["bytes_loaded"] == 2 * 1024**3
+
+
+# ---------------------------------------------------------------------------
+# E. Lifecycle races
+# ---------------------------------------------------------------------------
+
+
+class TestLifecycleRaces:
+    def test_process_set_but_gguf_path_not_yet(self, tmp_path):
+        """Moment between Popen and self._gguf_path=model_path."""
+        inst = _make()
+        inst._process = _Proc(os.getpid())
+        inst._gguf_path = None
+        with patch("builtins.open", side_effect = _fake_proc_reader(1024)):
+            out = inst.load_progress()
+        assert out is not None
+        assert out["phase"] == "mmap"
+        assert out["bytes_total"] == 0
+        assert out["bytes_loaded"] == 1024 * 1024
+
+    def test_process_died_mid_sample(self, tmp_path):
+        """/proc/<pid> disappears -> None."""
+        _sparse(tmp_path / "m.gguf", 1 * 1024**3)
+        inst = _make()
+        inst._process = _Proc(pid = 999_999_999)
+        inst._gguf_path = str(tmp_path / "m.gguf")
+        assert inst.load_progress() is None
+
+    def test_healthy_true_ready_phase(self, tmp_path):
+        _sparse(tmp_path / "m.gguf", 1 * 1024**3)
+        inst = _make()
+        inst._process = _Proc(os.getpid())
+        inst._gguf_path = str(tmp_path / "m.gguf")
+        inst._healthy = True
+        with patch("builtins.open", side_effect = _fake_proc_reader(1024)):
+            out = inst.load_progress()
+        assert out["phase"] == "ready"
+
+
+# ---------------------------------------------------------------------------
+# F. Concurrent sampling  (simulates multiple browser tabs polling)
+# ---------------------------------------------------------------------------
+
+
+class TestConcurrentSampling:
+    def test_parallel_invocations_never_raise(self, tmp_path):
+        """Many concurrent samplers hitting the same backend must not raise.
+
+        We intentionally do NOT patch ``builtins.open`` here because
+        ``unittest.mock.patch`` is not thread-safe: interleaved
+        enter/exit across threads can leak a Mock into ``builtins.open``
+        and poison every subsequent test in the session. Instead, we
+        let each thread hit the real ``/proc/self/status`` of the test
+        process, which is exactly the code path that matters in prod.
+        """
+        _sparse(tmp_path / "m.gguf", 1 * 1024**3)
+        inst = _make()
+        inst._process = _Proc(os.getpid())
+        inst._gguf_path = str(tmp_path / "m.gguf")
+        errors = []
+
+        def run():
+            try:
+                for _ in range(50):
+                    inst.load_progress()
+            except Exception as e:  # pragma: no cover
+                errors.append(e)
+
+        threads = [threading.Thread(target = run) for _ in range(10)]
+        for t in threads:
+            t.start()
+        for t in threads:
+            t.join()
+        assert not errors, errors
+
+
+# ---------------------------------------------------------------------------
+# G. Fraction bounds
+# ---------------------------------------------------------------------------
+
+
+class TestFractionBounds:
+    def test_fraction_capped_at_one(self, tmp_path):
+        _sparse(tmp_path / "m.gguf", 1 * 1024**3)
+        inst = _make()
+        inst._process = _Proc(os.getpid())
+        inst._gguf_path = str(tmp_path / "m.gguf")
+        # RSS > total (post-paged-in + extra structures)
+        with patch("builtins.open", side_effect = _fake_proc_reader(2 * 1024**2)):
+            out = inst.load_progress()
+        assert 0.0 <= out["fraction"] <= 1.0
+
+    def test_fraction_zero_when_total_zero(self):
+        inst = _make()
+        inst._process = _Proc(os.getpid())
+        inst._gguf_path = None
+        with patch("builtins.open", side_effect = _fake_proc_reader(1024**2)):
+            out = inst.load_progress()
+        assert out["fraction"] == 0.0
diff --git a/studio/frontend/src/features/chat/components/model-load-status.tsx b/studio/frontend/src/features/chat/components/model-load-status.tsx
index afec327a0f..16583519e4 100644
--- a/studio/frontend/src/features/chat/components/model-load-status.tsx
+++ b/studio/frontend/src/features/chat/components/model-load-status.tsx
@@ -17,6 +17,29 @@ function clampProgress(value: number): number {
   return Math.max(0, Math.min(100, value));
 }
 
+/**
+ * Split a composed progress label like
+ *   "22.8 of 122.3 GB • 330.1 MB/s • 5m 9s left"
+ * into a primary chunk ("22.8 of 122.3 GB") that can sit next to the
+ * percent, and a secondary chunk ("330.1 MB/s • 5m 9s left") that can
+ * live on its own row. This keeps either line from overflowing into a
+ * ragged wrap when the rate/ETA part shows up mid-download.
+ *
+ * Labels without " • " (e.g. "22.8 GB downloaded") are returned
+ * primary-only, so the secondary row simply doesn't render.
+ */
+function splitProgressLabel(
+  label: string | null | undefined,
+): { primary: string; secondary: string } {
+  if (!label) return { primary: "", secondary: "" };
+  const idx = label.indexOf(" \u2022 ");
+  if (idx < 0) return { primary: label, secondary: "" };
+  return {
+    primary: label.slice(0, idx),
+    secondary: label.slice(idx + 3),
+  };
+}
+
 export function ModelLoadDescription({
   title,
   message,
@@ -25,6 +48,10 @@ export function ModelLoadDescription({
   onStop,
 }: ModelLoadDescriptionProps) {
   const hasProgress = typeof progressPercent === "number";
+  // Split once at the top of the render so the JSX below stays flat --
+  // no IIFE required. splitProgressLabel is a trivial string op.
+  const { primary: labelPrimary, secondary: labelSecondary } =
+    splitProgressLabel(progressLabel);
 
   return (
     <div className="relative flex min-h-12 w-full items-stretch gap-2">
@@ -35,11 +62,21 @@ export function ModelLoadDescription({
         {title ? <p className="text-foreground leading-5 font-semibold">{title}</p> : null}
         {hasProgress ? (
           <div className="w-full pt-1">
-            <div className="flex items-center justify-between text-[10px] font-medium tracking-[0.08em] text-muted-foreground/80">
-              <span>{progressLabel}</span>
-              <span>{Math.round(clampProgress(progressPercent))}%</span>
+            <div className="flex items-center justify-between gap-2 text-[10px] font-medium tracking-[0.08em] text-muted-foreground/80">
+              <span className="min-w-0 truncate">{labelPrimary}</span>
+              <span className="shrink-0 tabular-nums">
+                {Math.round(clampProgress(progressPercent))}%
+              </span>
             </div>
-            <Progress value={clampProgress(progressPercent)} className="h-1 bg-foreground/[0.08]" />
+            {labelSecondary ? (
+              <div className="truncate pt-0.5 text-[10px] font-medium tracking-[0.08em] text-muted-foreground/60">
+                {labelSecondary}
+              </div>
+            ) : null}
+            <Progress
+              value={clampProgress(progressPercent)}
+              className="mt-1 h-1 bg-foreground/[0.08]"
+            />
           </div>
         ) : message ? (
           <p className="pt-1 text-xs leading-relaxed text-muted-foreground">{message}</p>
@@ -89,9 +126,17 @@ export function ModelLoadInlineStatus({
           <div className="min-w-[7rem] flex-1">
             <Progress value={clampProgress(progressPercent)} className="h-1 bg-foreground/[0.08]" />
           </div>
-          <div className="flex shrink-0 items-center gap-1 text-[10px] font-medium tracking-[0.08em] text-muted-foreground/80">
-            <span>{progressLabel}</span>
-            <span>{Math.round(clampProgress(progressPercent))}%</span>
+          <div
+            className="flex shrink-0 items-center gap-1 text-[10px] font-medium tracking-[0.08em] text-muted-foreground/80"
+            title={progressLabel ?? undefined}
+          >
+            {/* Inline layout is horizontal and tight -- show only the
+                primary (bytes) chunk; the full label (with rate/ETA)
+                stays available via the tooltip. */}
+            <span>{splitProgressLabel(progressLabel).primary}</span>
+            <span className="tabular-nums">
+              {Math.round(clampProgress(progressPercent))}%
+            </span>
           </div>
         </div>
       ) : null}
diff --git a/studio/frontend/src/features/chat/hooks/use-transfer-stats.ts b/studio/frontend/src/features/chat/hooks/use-transfer-stats.ts
index 9ea7d1e625..5681680842 100644
--- a/studio/frontend/src/features/chat/hooks/use-transfer-stats.ts
+++ b/studio/frontend/src/features/chat/hooks/use-transfer-stats.ts
@@ -19,26 +19,20 @@
 
 import { useEffect, useRef, useState } from "react";
 
-export type TransferStats = {
-  rateBytesPerSecond: number;
-  etaSeconds: number;
-  /**
-   * False for the first few ticks (window not filled, or no forward
-   * progress yet). Consumers should hide rate/ETA while unstable so the
-   * UI doesn't flicker "123 GB/s" during the first tick.
-   */
-  stable: boolean;
-};
+import {
+  appendSample,
+  computeTransferStats,
+  type TransferSample,
+  type TransferStats,
+} from "../utils/transfer-stats";
 
-const MIN_SAMPLES = 3;
-const MIN_WINDOW_SECONDS = 3;
-const MAX_WINDOW_SECONDS = 15;
+export type { TransferStats } from "../utils/transfer-stats";
 
 export function useTransferStats(
   bytes: number | null | undefined,
   totalBytes: number | null | undefined,
 ): TransferStats {
-  const samplesRef = useRef<{ t: number; b: number }[]>([]);
+  const samplesRef = useRef<TransferSample[]>([]);
   const [state, setState] = useState<TransferStats>({
     rateBytesPerSecond: 0,
     etaSeconds: 0,
@@ -53,45 +47,8 @@ export function useTransferStats(
         ? totalBytes
         : 0;
 
-    // If the counter resets (e.g. user unloaded and started a new
-    // download), drop the stale window.
-    const samples = samplesRef.current;
-    if (samples.length > 0 && cur < samples[samples.length - 1].b) {
-      samples.length = 0;
-    }
-
-    samples.push({ t: now, b: cur });
-
-    // Drop samples older than MAX_WINDOW_SECONDS; keep at least 2 so we
-    // can still compute a rate when the counter hasn't moved in a while.
-    const cutoff = now - MAX_WINDOW_SECONDS;
-    while (samples.length > 2 && samples[0].t < cutoff) {
-      samples.shift();
-    }
-
-    if (samples.length < MIN_SAMPLES) {
-      setState({ rateBytesPerSecond: 0, etaSeconds: 0, stable: false });
-      return;
-    }
-
-    const first = samples[0];
-    const last = samples[samples.length - 1];
-    const dt = last.t - first.t;
-    const db = last.b - first.b;
-    if (dt < MIN_WINDOW_SECONDS || db <= 0) {
-      setState({ rateBytesPerSecond: 0, etaSeconds: 0, stable: false });
-      return;
-    }
-
-    const rate = db / dt;
-    const eta =
-      total > 0 && cur < total && rate > 0 ? (total - cur) / rate : 0;
-
-    setState({
-      rateBytesPerSecond: rate,
-      etaSeconds: eta,
-      stable: true,
-    });
+    appendSample(samplesRef.current, now, cur);
+    setState(computeTransferStats(samplesRef.current, total));
   }, [bytes, totalBytes]);
 
   return state;
diff --git a/studio/frontend/src/features/chat/utils/transfer-stats.ts b/studio/frontend/src/features/chat/utils/transfer-stats.ts
new file mode 100644
index 0000000000..7aecf29eb0
--- /dev/null
+++ b/studio/frontend/src/features/chat/utils/transfer-stats.ts
@@ -0,0 +1,88 @@
+// SPDX-License-Identifier: AGPL-3.0-only
+// Copyright 2026-present the Unsloth AI Inc. team. All rights reserved. See /studio/LICENSE.AGPL-3.0
+
+/**
+ * Pure, framework-free math behind {@link useTransferStats}.
+ *
+ * Split out so it can be unit-tested without a DOM/React renderer, and
+ * so the training-start overlay, the chat download toast, and the
+ * model-load phase UI all share the exact same rate/ETA semantics.
+ *
+ * No React, no `useRef`/`useState`, no timers -- the caller owns the
+ * sample buffer and clock. This is the "what is the rate right now
+ * given these samples" question, nothing more.
+ */
+
+export type TransferSample = { t: number; b: number };
+
+export type TransferStats = {
+  rateBytesPerSecond: number;
+  etaSeconds: number;
+  /**
+   * False until the window has at least {@link MIN_SAMPLES} samples
+   * spanning ≥ {@link MIN_WINDOW_SECONDS} with strictly forward progress.
+   * Consumers should hide rate/ETA while this is false so the UI doesn't
+   * flicker "123 GB/s" during the first tick.
+   */
+  stable: boolean;
+};
+
+export const MIN_SAMPLES = 3;
+export const MIN_WINDOW_SECONDS = 3;
+export const MAX_WINDOW_SECONDS = 15;
+
+/**
+ * Mutate ``samples`` in place: append the new sample, drop any that
+ * fell out of the rolling window, and clear the buffer if the counter
+ * went backwards (user cancelled + restarted a download, etc.).
+ *
+ * Returns the same array for chain-ability.
+ */
+export function appendSample(
+  samples: TransferSample[],
+  t: number,
+  b: number,
+  maxWindowSeconds: number = MAX_WINDOW_SECONDS,
+): TransferSample[] {
+  if (samples.length > 0 && b < samples[samples.length - 1].b) {
+    samples.length = 0;
+  }
+  samples.push({ t, b });
+  const cutoff = t - maxWindowSeconds;
+  while (samples.length > 2 && samples[0].t < cutoff) {
+    samples.shift();
+  }
+  return samples;
+}
+
+/**
+ * Derive {@link TransferStats} from a window of cumulative-byte samples
+ * plus the known total.
+ *
+ *   * Needs ≥ {@link MIN_SAMPLES} samples spanning ≥ {@link MIN_WINDOW_SECONDS}
+ *     seconds before it will report ``stable: true``.
+ *   * ETA is clamped to 0 when: no progress, no total known, or the
+ *     counter already hit the total.
+ */
+export function computeTransferStats(
+  samples: readonly TransferSample[],
+  total: number,
+): TransferStats {
+  if (samples.length < MIN_SAMPLES) {
+    return { rateBytesPerSecond: 0, etaSeconds: 0, stable: false };
+  }
+  const first = samples[0];
+  const last = samples[samples.length - 1];
+  const dt = last.t - first.t;
+  const db = last.b - first.b;
+  if (dt < MIN_WINDOW_SECONDS || db <= 0) {
+    return { rateBytesPerSecond: 0, etaSeconds: 0, stable: false };
+  }
+  const rate = db / dt;
+  const safeTotal = Number.isFinite(total) && total > 0 ? total : 0;
+  const eta =
+    safeTotal > 0 && last.b < safeTotal && rate > 0
+      ? (safeTotal - last.b) / rate
+      : 0;
+  return { rateBytesPerSecond: rate, etaSeconds: eta, stable: true };
+}

From 5aa8c152468cd741949c72187dcb0702a1f11f43 Mon Sep 17 00:00:00 2001
From: Daniel Han <danielhanchen@gmail.com>
Date: Tue, 14 Apr 2026 10:58:20 -0700
Subject: [PATCH 097/216] Studio: hard-stop at n_ctx with a 'Context limit
 reached' toast (#5021)

* Studio: hard-stop at n_ctx with a dedicated 'Context limit reached' toast

llama-server's default behavior when the KV cache fills is to silently
drop the oldest non-``n_keep`` tokens and keep generating. The UI has
no way to tell the user that earlier turns were evicted -- they just
see degraded continuity and a confusing ``5,361 / 4,096`` on the
context usage bar.

Launch llama-server with ``--no-context-shift`` so it returns a clean
error once the request would exceed ``n_ctx``. In the chat adapter,
catch the error, identify it as a context-limit error via
``isContextLimitError()``, and surface a dedicated toast that names
the exact control to adjust: the ``Context Length`` field in the chat
Settings panel.

Also add a lightweight tooltip hint on ``ContextUsageBar`` when usage
crosses 85%, so users see the "raise Context Length in Settings"
suggestion before they hit the hard stop.

Tests:

  * ``test_llama_cpp_no_context_shift.py`` pins the ``--no-context-shift``
    flag in the static launch-command template, and pins it inside the
    unconditional ``cmd = [ ... ]`` block so a future refactor can't
    hide it behind a branch.

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

* Shorten --no-context-shift comment to 1 line

* Match backend _friendly_error rewrite in isContextLimitError

Codex review on PR caught that ``backend/routes/inference.py::_friendly_error``
rewrites the raw llama-server text
  "request (X tokens) exceeds the available context size (Y tokens)"
into
  "Message too long: X tokens exceeds the Y-token context window. ..."
on the main streaming GGUF path. The heuristic only looked for
"context size" / "exceeds the available context" / "context shift",
none of which survive the rewrite, so the new "Context limit reached"
toast would never fire for the most common case. Add matches for
"message too long" and "context window" so both wordings hit.

Also addresses Gemini feedback on the launch-flag test:
  * Use ``inspect.getsource(LlamaCppBackend.load_model)`` instead of
    reading ``__file__`` directly; scopes the assertions to the
    function that actually launches llama-server.
  * Replace the hardcoded ``"            ]"`` indent search with a
    line-at-a-time scan for a line that is just ``]``, so the test
    survives reformatting.

---------

Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com>
---
 studio/backend/core/inference/llama_cpp.py    |   2 +
 .../tests/test_llama_cpp_no_context_shift.py  | 137 ++++++++++++++++++
 .../src/features/chat/api/chat-adapter.ts     |  56 ++++++-
 .../chat/components/context-usage-bar.tsx     |   7 +
 4 files changed, 199 insertions(+), 3 deletions(-)
 create mode 100644 studio/backend/tests/test_llama_cpp_no_context_shift.py

diff --git a/studio/backend/core/inference/llama_cpp.py b/studio/backend/core/inference/llama_cpp.py
index 71dec4c38b..b53fc513de 100644
--- a/studio/backend/core/inference/llama_cpp.py
+++ b/studio/backend/core/inference/llama_cpp.py
@@ -1419,6 +1419,8 @@ class LlamaCppBackend:
                 str(n_parallel),
                 "--flash-attn",
                 "on",  # Force flash attention for speed
+                # Error out at n_ctx instead of silently rotating the KV cache; frontend catches it and points the user at "Context Length".
+                "--no-context-shift",
             ]
 
             if use_fit:
diff --git a/studio/backend/tests/test_llama_cpp_no_context_shift.py b/studio/backend/tests/test_llama_cpp_no_context_shift.py
new file mode 100644
index 0000000000..b9f25faf88
--- /dev/null
+++ b/studio/backend/tests/test_llama_cpp_no_context_shift.py
@@ -0,0 +1,137 @@
+# SPDX-License-Identifier: AGPL-3.0-only
+# Copyright 2026-present the Unsloth AI Inc. team. All rights reserved. See /studio/LICENSE.AGPL-3.0
+
+"""``--no-context-shift`` launch-flag contract.
+
+When llama-server runs with its default context-shift behavior, the UI
+has no way to tell the user that the KV cache has been rotated --
+earlier turns silently vanish from the conversation. The Studio
+backend always passes ``--no-context-shift`` so the server returns a
+clean error instead, and the chat adapter can point the user at the
+``Context Length`` input in the settings panel.
+
+This file is a static read of the launch command: we ask
+``LlamaCppBackend`` to assemble its ``cmd`` list and assert the flag
+is always present. Testing via the real subprocess would require an
+actual GGUF on disk, which is out of scope for the fast test suite.
+"""
+
+from __future__ import annotations
+
+import inspect
+import sys
+import types as _types
+from pathlib import Path
+
+import pytest
+
+# ---------------------------------------------------------------------------
+# Same external-dep stubs as the other llama_cpp tests.
+# ---------------------------------------------------------------------------
+
+_BACKEND_DIR = str(Path(__file__).resolve().parent.parent)
+if _BACKEND_DIR not in sys.path:
+    sys.path.insert(0, _BACKEND_DIR)
+
+_loggers_stub = _types.ModuleType("loggers")
+_loggers_stub.get_logger = lambda name: __import__("logging").getLogger(name)
+sys.modules.setdefault("loggers", _loggers_stub)
+
+_structlog_stub = _types.ModuleType("structlog")
+sys.modules.setdefault("structlog", _structlog_stub)
+
+_httpx_stub = _types.ModuleType("httpx")
+for _exc in (
+    "ConnectError",
+    "TimeoutException",
+    "ReadTimeout",
+    "ReadError",
+    "RemoteProtocolError",
+    "CloseError",
+):
+    setattr(_httpx_stub, _exc, type(_exc, (Exception,), {}))
+_httpx_stub.Timeout = type("T", (), {"__init__": lambda s, *a, **k: None})
+_httpx_stub.Client = type(
+    "C",
+    (),
+    {
+        "__init__": lambda s, **kw: None,
+        "__enter__": lambda s: s,
+        "__exit__": lambda s, *a: None,
+    },
+)
+sys.modules.setdefault("httpx", _httpx_stub)
+
+from core.inference import llama_cpp as llama_cpp_module
+
+
+def _load_model_source() -> str:
+    """Return the source of ``LlamaCppBackend.load_model``.
+
+    Using ``inspect.getsource`` instead of reading the file directly
+    scopes the assertions to the function that actually launches
+    llama-server, so neither the presence check nor the location check
+    can be fooled by a stray occurrence of ``"--no-context-shift"``
+    elsewhere in the module.
+    """
+    return inspect.getsource(llama_cpp_module.LlamaCppBackend.load_model)
+
+
+def test_no_context_shift_is_in_load_model():
+    """The flag is part of the static launch-command template.
+
+    We check the source of ``load_model`` rather than mocking the whole
+    call chain (GPU probing, GGUF stat, etc.): the flag is written as
+    a literal in one place and any regression has to delete it, which
+    a text search will catch.
+    """
+    assert '"--no-context-shift"' in _load_model_source(), (
+        "llama-server must be launched with --no-context-shift so the "
+        "UI can surface a clean 'context full' error instead of silently "
+        "losing old turns to a KV-cache rotation."
+    )
+
+
+def test_flag_sits_inside_the_base_cmd_list():
+    """Pin the flag's location so a future refactor can't accidentally
+    move it into a branch that only fires on some code paths.
+
+    We slice from ``cmd = [`` to the first ``]`` at the same indent.
+    Using ``inspect.getsource`` means the function lives in its own
+    string and there are no siblings to worry about, so a plain
+    bracket search would also work -- anchoring on the trailing indent
+    just keeps the slice from wandering into a later expression if the
+    opening literal ever grows an in-line comment trailing it.
+    """
+    source = _load_model_source()
+    start = source.find("cmd = [")
+    assert start >= 0, "could not find the base cmd = [...] block"
+    # Find the first line containing only ``]`` (possibly indented).
+    # Works for any indentation style the formatter picks.
+    rest = source[start:]
+    end_rel = -1
+    for line_start, line in _iter_lines_with_offset(rest):
+        if line_start == 0:
+            # Skip the opening ``cmd = [`` line itself.
+            continue
+        if line.strip() == "]":
+            end_rel = line_start
+            break
+    assert end_rel > 0, "could not find end of cmd = [...] block"
+    block = rest[:end_rel]
+    assert '"--no-context-shift"' in block, (
+        "--no-context-shift must be in the base cmd list, not in a "
+        "conditional branch -- otherwise some code paths would still "
+        "run with silent context shift enabled."
+    )
+    # Also pin that it is next to -c / --ctx so the grouping makes sense.
+    assert '"-c"' in block
+    assert '"--flash-attn"' in block
+
+
+def _iter_lines_with_offset(text: str):
+    """Yield (offset, line) pairs over ``text`` without losing offsets."""
+    offset = 0
+    for line in text.splitlines(keepends = True):
+        yield offset, line
+        offset += len(line)
diff --git a/studio/frontend/src/features/chat/api/chat-adapter.ts b/studio/frontend/src/features/chat/api/chat-adapter.ts
index 76dbbf0bfb..60904f3cb5 100644
--- a/studio/frontend/src/features/chat/api/chat-adapter.ts
+++ b/studio/frontend/src/features/chat/api/chat-adapter.ts
@@ -47,6 +47,41 @@ type RunMessage = RunMessages[number];
 /** Tracks which user messages were sent with an audio file (messageId → filename). */
 export const sentAudioNames = new Map<string, string>();
 
+/**
+ * Match error messages that indicate the request filled or would fill
+ * the KV cache, so the UI can show a dedicated toast pointing at the
+ * ``Context Length`` setting.
+ *
+ * Two wordings reach the client and both must hit:
+ *
+ *   1. The raw llama-server text when ``--no-context-shift`` trips --
+ *      "the request exceeds the available context size (N tokens)".
+ *   2. The rewritten friendly text emitted by
+ *      ``backend/routes/inference.py::_friendly_error`` -- "Message too
+ *      long: X tokens exceeds the Y-token context window. Try
+ *      increasing the Context Length ..." This is the one most users
+ *      see on the streaming GGUF path.
+ *
+ * We match on substrings rather than full regexes because both layers
+ * have drifted across versions (llama.cpp master has tweaked the
+ * phrasing; ``_friendly_error`` has gone through several copy edits).
+ */
+export function isContextLimitError(message: string): boolean {
+  if (!message) return false;
+  const m = message.toLowerCase();
+  return (
+    // Raw llama-server wording.
+    m.includes("context size") ||
+    m.includes("context shift") ||
+    m.includes("exceeds the available context") ||
+    // Backend _friendly_error rewrite.
+    m.includes("message too long") ||
+    m.includes("context window") ||
+    // n_ctx mentions that carry an "exceed"/"full" signal.
+    (m.includes("n_ctx") && (m.includes("exceed") || m.includes("full")))
+  );
+}
+
 /** Parse "Title: ...\nURL: ...\nSnippet: ..." blocks into source content parts. */
 function parseSourcesFromResult(raw: string): { type: "source"; sourceType: "url"; id: string; url: string; title: string; metadata?: { description: string } }[] {
   if (!raw) return [];
@@ -868,9 +903,24 @@ export function createOpenAIStreamAdapter(): ChatModelAdapter {
       } catch (err) {
         settleFirstTokenErr(err instanceof Error ? err : new Error("Generation failed"));
         if (!abortSignal.aborted) {
-          toast.error("Generation failed", {
-            description: err instanceof Error ? err.message : "Unknown error",
-          });
+          const msg = err instanceof Error ? err.message : String(err);
+          if (isContextLimitError(msg)) {
+            // llama-server was launched with --no-context-shift, so it
+            // returns a hard error instead of silently dropping old
+            // turns from the KV cache. Point the user at the exact
+            // control that raises the ceiling.
+            toast.error("Context limit reached", {
+              description:
+                "The conversation has filled the model's context window. " +
+                "Increase \"Context Length\" in the chat Settings panel (⚙ in the top-right), " +
+                "or start a new chat.",
+              duration: 8000,
+            });
+          } else {
+            toast.error("Generation failed", {
+              description: msg || "Unknown error",
+            });
+          }
         }
         throw err;
       } finally {
diff --git a/studio/frontend/src/features/chat/components/context-usage-bar.tsx b/studio/frontend/src/features/chat/components/context-usage-bar.tsx
index e3f0afb51a..b1fd1af588 100644
--- a/studio/frontend/src/features/chat/components/context-usage-bar.tsx
+++ b/studio/frontend/src/features/chat/components/context-usage-bar.tsx
@@ -104,6 +104,13 @@ export const ContextUsageBar: FC<{
               {formatTokenCountFull(used)} / {formatTokenCountFull(total)}
             </span>
           </div>
+          {percent > 85 && (
+            <div className="mt-1 max-w-64 text-[11px] leading-snug text-muted-foreground/90">
+              Close to the context limit. Generation will stop at 100%.
+              Increase <span className="font-medium">Context Length</span> in
+              the chat Settings panel to keep going.
+            </div>
+          )}
         </div>
       </TooltipContent>
     </Tooltip>

From 826c98f3c05e548627921a18c223db7b963e140f Mon Sep 17 00:00:00 2001
From: Datta Nimmaturi <venkatadattasainimmaturi@gmail.com>
Date: Wed, 15 Apr 2026 03:23:07 +0530
Subject: [PATCH 098/216] [moe][gemma4] Target MoE for gemma4 (#4913)

* Target MoE for gemma4

* refactor attention impl determine

* Revert "refactor attention impl determine"

This reverts commit 888fca08110a9a74278dc1ebc14d0da043bbd11d.

* Remove attention policy changes from gemma4 MoE fix
---
 unsloth/models/_utils.py | 52 +++++++++++++++++++++++++++++++++++++---
 1 file changed, 49 insertions(+), 3 deletions(-)

diff --git a/unsloth/models/_utils.py b/unsloth/models/_utils.py
index 96675eeaf0..71f7db915c 100644
--- a/unsloth/models/_utils.py
+++ b/unsloth/models/_utils.py
@@ -2773,13 +2773,48 @@ def is_moe_model(model) -> bool:
     return num_experts is not None and num_experts > 0
 
 
+def _resolve_moe_parameter_name(model, default_name: str, alternate_name: str) -> str:
+    """
+    Resolve the actual parameter path for MoE expert weights.
+
+    Most current Unsloth MoE models expose expert weights under
+    ``mlp.experts.*``. Gemma4 stores them directly under ``experts.*``.
+    Prefer the path that exists on the loaded module when possible.
+    """
+    if hasattr(model, "named_parameters"):
+        try:
+            for name, _ in model.named_parameters():
+                if name == default_name or name.endswith("." + default_name):
+                    return default_name
+                if name == alternate_name or name.endswith("." + alternate_name):
+                    return alternate_name
+        except Exception:
+            pass
+
+    config = getattr(model, "config", model)
+    model_types = {getattr(config, "model_type", None)}
+    text_config = getattr(config, "text_config", None)
+    if text_config is not None:
+        model_types.add(getattr(text_config, "model_type", None))
+
+    if any(
+        isinstance(model_type, str) and model_type.startswith("gemma4")
+        for model_type in model_types
+    ):
+        return alternate_name
+
+    return default_name
+
+
 def get_moe_target_parameters(model, target_modules = None) -> Optional[List[str]]:
     """
     Get the target_parameters for MoE expert layers if applicable.
 
     For MoE models, returns the parameter paths for expert weights
     (gate_up_proj, down_proj) that should be targeted by PEFT's
-    target_parameters for LoRA on nn.Parameter.
+    target_parameters for LoRA on nn.Parameter. The exact parameter path
+    depends on the model layout, for example ``mlp.experts.*`` or
+    ``experts.*``.
 
     Only includes MoE parameters that match what's in target_modules:
     - If "down_proj" is in target_modules -> includes "mlp.experts.down_proj"
@@ -2827,6 +2862,17 @@ def get_moe_target_parameters(model, target_modules = None) -> Optional[List[str
     else:
         target_set = set(target_modules) if target_modules else set()
 
+    gate_up_name = _resolve_moe_parameter_name(
+        model,
+        default_name = "mlp.experts.gate_up_proj",
+        alternate_name = "experts.gate_up_proj",
+    )
+    down_name = _resolve_moe_parameter_name(
+        model,
+        default_name = "mlp.experts.down_proj",
+        alternate_name = "experts.down_proj",
+    )
+
     # gate_up_proj combines both gate_proj and up_proj in MoE
     # Also match "gate_up_proj" directly since users may specify the fused name
     if (
@@ -2834,10 +2880,10 @@ def get_moe_target_parameters(model, target_modules = None) -> Optional[List[str
         or "up_proj" in target_set
         or "gate_up_proj" in target_set
     ):
-        moe_params.append("mlp.experts.gate_up_proj")
+        moe_params.append(gate_up_name)
 
     if "down_proj" in target_set:
-        moe_params.append("mlp.experts.down_proj")
+        moe_params.append(down_name)
 
     if moe_params:
         print(

From 13928b5f0ed215667f8c54abe1b975bbd6ab5ce5 Mon Sep 17 00:00:00 2001
From: Roland Tannous <115670425+rolandtannous@users.noreply.github.com>
Date: Wed, 15 Apr 2026 11:39:11 +0400
Subject: [PATCH 099/216] Add configurable PyTorch mirror via
 UNSLOTH_PYTORCH_MIRROR env var (#5024)

* Add configurable PyTorch mirror via UNSLOTH_PYTORCH_MIRROR env var

When set, UNSLOTH_PYTORCH_MIRROR overrides the default
https://download.pytorch.org/whl base URL in all four install scripts
(install.sh, install.ps1, studio/setup.ps1, studio/install_python_stack.py).
When unset or empty, the official URL is used. This lets users behind
corporate proxies or in regions with poor connectivity to pytorch.org
point at a local mirror without patching scripts.

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

* Add pytest for UNSLOTH_PYTORCH_MIRROR in install_python_stack.py

Tests that _PYTORCH_WHL_BASE picks up the env var when set, falls back
to the official URL when unset or empty, and preserves the value as-is
(including trailing slashes).

* Remove stale test assertions for missing install.sh messages

* Fix GPU mocking in test_get_torch_index_url.sh

Extract _has_usable_nvidia_gpu and _has_amd_rocm_gpu alongside
get_torch_index_url so the GPU-presence checks work in tests.
Add -L flag handling to mock nvidia-smi so it passes the GPU listing
check. All 26 tests now pass on CPU-only machines.

* Strip trailing slash from UNSLOTH_PYTORCH_MIRROR to avoid double-slash URLs

---------

Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com>
---
 install.ps1                                 |  2 +-
 install.sh                                  |  3 +-
 studio/backend/tests/test_pytorch_mirror.py | 55 +++++++++++++++++++++
 studio/install_python_stack.py              |  4 +-
 studio/setup.ps1                            | 10 ++--
 tests/python/test_cross_platform_parity.py  | 16 ++++++
 tests/sh/test_get_torch_index_url.sh        | 51 ++++++++++++++++---
 tests/sh/test_mac_intel_compat.sh           | 24 ++-------
 8 files changed, 131 insertions(+), 34 deletions(-)
 create mode 100644 studio/backend/tests/test_pytorch_mirror.py

diff --git a/install.ps1 b/install.ps1
index a2acd6c4ea..61f31d85b4 100644
--- a/install.ps1
+++ b/install.ps1
@@ -754,7 +754,7 @@ shell.Run cmd, 0, False
     # ── Choose the correct PyTorch index URL based on driver CUDA version ──
     # Mirrors Get-PytorchCudaTag in setup.ps1.
     function Get-TorchIndexUrl {
-        $baseUrl = "https://download.pytorch.org/whl"
+        $baseUrl = if ($env:UNSLOTH_PYTORCH_MIRROR) { $env:UNSLOTH_PYTORCH_MIRROR.TrimEnd('/') } else { "https://download.pytorch.org/whl" }
         if (-not $NvidiaSmiExe) { return "$baseUrl/cpu" }
         try {
             $output = & $NvidiaSmiExe 2>&1 | Out-String
diff --git a/install.sh b/install.sh
index 56af1454c5..0dbcdf380e 100755
--- a/install.sh
+++ b/install.sh
@@ -1053,7 +1053,8 @@ _has_usable_nvidia_gpu() {
 # On CPU-only machines this returns the cpu index, avoiding the solver
 # dead-end where --torch-backend=auto resolves to unsloth==2024.8.
 get_torch_index_url() {
-    _base="https://download.pytorch.org/whl"
+    _base="${UNSLOTH_PYTORCH_MIRROR:-https://download.pytorch.org/whl}"
+    _base="${_base%/}"
     # macOS: always CPU (no CUDA support)
     case "$(uname -s)" in Darwin) echo "$_base/cpu"; return ;; esac
     # Try nvidia-smi -- require the binary to actually list a usable GPU.
diff --git a/studio/backend/tests/test_pytorch_mirror.py b/studio/backend/tests/test_pytorch_mirror.py
new file mode 100644
index 0000000000..5844f209b6
--- /dev/null
+++ b/studio/backend/tests/test_pytorch_mirror.py
@@ -0,0 +1,55 @@
+# SPDX-License-Identifier: AGPL-3.0-only
+# Copyright 2026-present the Unsloth AI Inc. team. All rights reserved. See /studio/LICENSE.AGPL-3.0
+
+"""Tests for UNSLOTH_PYTORCH_MIRROR env var in install_python_stack.py."""
+
+from __future__ import annotations
+
+import importlib
+import os
+import sys
+from pathlib import Path
+
+import pytest
+
+# install_python_stack.py lives at repo_root/studio/install_python_stack.py
+_INSTALL_SCRIPT = Path(__file__).resolve().parents[2] / "install_python_stack.py"
+
+OFFICIAL_URL = "https://download.pytorch.org/whl"
+
+
+def _reload_whl_base(monkeypatch, mirror_value = None):
+    """(Re-)import install_python_stack with a controlled env and return _PYTORCH_WHL_BASE."""
+    # Remove cached module so the module-level assignment re-executes
+    sys.modules.pop("install_python_stack", None)
+
+    if mirror_value is None:
+        monkeypatch.delenv("UNSLOTH_PYTORCH_MIRROR", raising = False)
+    else:
+        monkeypatch.setenv("UNSLOTH_PYTORCH_MIRROR", mirror_value)
+
+    # Temporarily add the script's directory to sys.path for import
+    script_dir = str(_INSTALL_SCRIPT.parent)
+    monkeypatch.syspath_prepend(script_dir)
+
+    import install_python_stack
+
+    return install_python_stack._PYTORCH_WHL_BASE
+
+
+class TestPyTorchMirrorEnvVar:
+    """UNSLOTH_PYTORCH_MIRROR controls _PYTORCH_WHL_BASE in install_python_stack."""
+
+    def test_unset_uses_official_url(self, monkeypatch):
+        assert _reload_whl_base(monkeypatch) == OFFICIAL_URL
+
+    def test_empty_string_falls_back_to_official(self, monkeypatch):
+        assert _reload_whl_base(monkeypatch, "") == OFFICIAL_URL
+
+    def test_custom_mirror_is_used(self, monkeypatch):
+        mirror = "https://mirrors.nju.edu.cn/pytorch/whl"
+        assert _reload_whl_base(monkeypatch, mirror) == mirror
+
+    def test_trailing_slash_stripped(self, monkeypatch):
+        result = _reload_whl_base(monkeypatch, "https://example.com/whl/")
+        assert result == "https://example.com/whl"
diff --git a/studio/install_python_stack.py b/studio/install_python_stack.py
index 71a2a5bf83..cf024d4076 100644
--- a/studio/install_python_stack.py
+++ b/studio/install_python_stack.py
@@ -49,7 +49,9 @@ _ROCM_TORCH_INDEX: dict[tuple[int, int], str] = {
     (6, 1): "rocm6.1",
     (6, 0): "rocm6.0",
 }
-_PYTORCH_WHL_BASE = "https://download.pytorch.org/whl"
+_PYTORCH_WHL_BASE = (
+    os.environ.get("UNSLOTH_PYTORCH_MIRROR") or "https://download.pytorch.org/whl"
+).rstrip("/")
 
 # bitsandbytes continuous-release_main wheels with the ROCm 4-bit GEMV fix
 # (bnb PR #1887, post-0.49.2). bnb <= 0.49.2 NaNs at decode shape on every
diff --git a/studio/setup.ps1 b/studio/setup.ps1
index c3a8cd71ca..1ba41360c7 100644
--- a/studio/setup.ps1
+++ b/studio/setup.ps1
@@ -1517,14 +1517,16 @@ if ($HasNvidiaSmi) {
     $CuTag = "cpu"
 }
 
+$PyTorchWhlBase = if ($env:UNSLOTH_PYTORCH_MIRROR) { $env:UNSLOTH_PYTORCH_MIRROR.TrimEnd('/') } else { "https://download.pytorch.org/whl" }
+
 if ($CuTag -eq "cpu") {
     substep "installing PyTorch (CPU-only)..."
     if ($script:UnslothVerbose) {
-        Fast-Install torch torchvision torchaudio --index-url "https://download.pytorch.org/whl/cpu"
+        Fast-Install torch torchvision torchaudio --index-url "$PyTorchWhlBase/cpu"
         $torchInstallExit = $LASTEXITCODE
         $output = ""
     } else {
-        $output = Fast-Install torch torchvision torchaudio --index-url "https://download.pytorch.org/whl/cpu" | Out-String
+        $output = Fast-Install torch torchvision torchaudio --index-url "$PyTorchWhlBase/cpu" | Out-String
         $torchInstallExit = $LASTEXITCODE
     }
     if ($torchInstallExit -ne 0) {
@@ -1536,11 +1538,11 @@ if ($CuTag -eq "cpu") {
     substep "installing PyTorch with CUDA support ($CuTag)..."
     substep "(This download is ~2.8 GB -- may take a few minutes)"
     if ($script:UnslothVerbose) {
-        Fast-Install torch torchvision torchaudio --index-url "https://download.pytorch.org/whl/$CuTag"
+        Fast-Install torch torchvision torchaudio --index-url "$PyTorchWhlBase/$CuTag"
         $torchInstallExit = $LASTEXITCODE
         $output = ""
     } else {
-        $output = Fast-Install torch torchvision torchaudio --index-url "https://download.pytorch.org/whl/$CuTag" | Out-String
+        $output = Fast-Install torch torchvision torchaudio --index-url "$PyTorchWhlBase/$CuTag" | Out-String
         $torchInstallExit = $LASTEXITCODE
     }
     if ($torchInstallExit -ne 0) {
diff --git a/tests/python/test_cross_platform_parity.py b/tests/python/test_cross_platform_parity.py
index 6dd41be9fa..7b9868c1f2 100644
--- a/tests/python/test_cross_platform_parity.py
+++ b/tests/python/test_cross_platform_parity.py
@@ -135,3 +135,19 @@ class TestCudaMappingParity:
             f"  install.sh:  {sh_thresholds}\n"
             f"  install.ps1: {ps1_thresholds}"
         )
+
+
+class TestPyTorchMirrorEnvVar:
+    """Both install scripts must support the UNSLOTH_PYTORCH_MIRROR env var."""
+
+    def test_install_sh_has_mirror_var(self):
+        text = INSTALL_SH.read_text()
+        assert (
+            "UNSLOTH_PYTORCH_MIRROR" in text
+        ), "install.sh should reference UNSLOTH_PYTORCH_MIRROR"
+
+    def test_install_ps1_has_mirror_var(self):
+        text = INSTALL_PS1.read_text()
+        assert (
+            "UNSLOTH_PYTORCH_MIRROR" in text
+        ), "install.ps1 should reference UNSLOTH_PYTORCH_MIRROR"
diff --git a/tests/sh/test_get_torch_index_url.sh b/tests/sh/test_get_torch_index_url.sh
index adbdbe4e14..7235873f53 100755
--- a/tests/sh/test_get_torch_index_url.sh
+++ b/tests/sh/test_get_torch_index_url.sh
@@ -7,14 +7,19 @@ INSTALL_SH="$SCRIPT_DIR/../../install.sh"
 PASS=0
 FAIL=0
 
-# Extract only the get_torch_index_url function from install.sh
+# Extract get_torch_index_url and its helper functions from install.sh.
 # Also replace the hardcoded /usr/bin/nvidia-smi fallback with a
 # controllable path so we can test the "no GPU" scenario on GPU machines.
 _FUNC_FILE=$(mktemp)
 _FAKE_SMI_DIR=$(mktemp -d)
-sed -n '/^get_torch_index_url()/,/^}/p' "$INSTALL_SH" \
-    | sed "s|/usr/bin/nvidia-smi|$_FAKE_SMI_DIR/nvidia-smi-absent|g" \
-    > "$_FUNC_FILE"
+{
+    sed -n '/^_has_amd_rocm_gpu()/,/^}/p' "$INSTALL_SH"
+    echo ""
+    sed -n '/^_has_usable_nvidia_gpu()/,/^}/p' "$INSTALL_SH"
+    echo ""
+    sed -n '/^get_torch_index_url()/,/^}/p' "$INSTALL_SH"
+} | sed "s|/usr/bin/nvidia-smi|$_FAKE_SMI_DIR/nvidia-smi-absent|g" \
+  > "$_FUNC_FILE"
 
 # Save system PATH so we always have basic tools (uname, grep, head, etc.)
 _SYS_PATH="/usr/local/bin:/usr/bin:/bin"
@@ -30,16 +35,25 @@ assert_eq() {
     fi
 }
 
-# Helper: create a mock nvidia-smi that prints a given CUDA version string
+# Helper: create a mock nvidia-smi that prints a given CUDA version string.
+# Handles both default output (version header) and -L (GPU listing) so that
+# _has_usable_nvidia_gpu sees a valid GPU.
 make_mock_smi() {
     _dir=$(mktemp -d)
     cat > "$_dir/nvidia-smi" <<MOCK
 #!/bin/sh
-cat <<'SMI_OUT'
+case "\$1" in
+    -L)
+        echo "GPU 0: NVIDIA GeForce RTX 3090 (UUID: GPU-fake-uuid)"
+        ;;
+    *)
+        cat <<'SMI_OUT'
 +-----------------------------------------------------------------------------------------+
 | NVIDIA-SMI 550.54.15              Driver Version: 550.54.15      CUDA Version: $1     |
 +-----------------------------------------------------------------------------------------+
 SMI_OUT
+        ;;
+esac
 MOCK
     chmod +x "$_dir/nvidia-smi"
     echo "$_dir"
@@ -130,11 +144,14 @@ _result=$(run_func "$_dir")
 assert_eq "CUDA 10.2 -> cpu" "https://download.pytorch.org/whl/cpu" "$_result"
 rm -rf "$_dir"
 
-# 8) Unparseable nvidia-smi output -> cu126 default
+# 8) Unparseable nvidia-smi version but valid GPU listing -> cu126 default
 _dir=$(mktemp -d)
 cat > "$_dir/nvidia-smi" <<'MOCK'
 #!/bin/sh
-echo "something completely unexpected"
+case "$1" in
+    -L) echo "GPU 0: NVIDIA GeForce RTX 3090 (UUID: GPU-fake-uuid)" ;;
+    *)  echo "something completely unexpected" ;;
+esac
 MOCK
 chmod +x "$_dir/nvidia-smi"
 _result=$(run_func "$_dir")
@@ -243,6 +260,24 @@ _result=$(run_func "$_dir")
 assert_eq "CUDA 12.8 regression -> cu128" "https://download.pytorch.org/whl/cu128" "$_result"
 rm -rf "$_dir"
 
+# 25) UNSLOTH_PYTORCH_MIRROR overrides base URL (CUDA case)
+_dir=$(make_mock_smi "12.6")
+_result=$(UNSLOTH_PYTORCH_MIRROR="https://mirror.example.com/whl" run_func "$_dir")
+assert_eq "mirror env + CUDA 12.6 -> mirror/cu126" "https://mirror.example.com/whl/cu126" "$_result"
+rm -rf "$_dir"
+
+# 26) UNSLOTH_PYTORCH_MIRROR overrides base URL (CPU case)
+_result=$(UNSLOTH_PYTORCH_MIRROR="https://mirror.example.com/whl" run_func "none")
+assert_eq "mirror env + no GPU -> mirror/cpu" "https://mirror.example.com/whl/cpu" "$_result"
+
+# 27) Empty UNSLOTH_PYTORCH_MIRROR falls back to official URL
+_result=$(UNSLOTH_PYTORCH_MIRROR="" run_func "none")
+assert_eq "empty mirror env -> official/cpu" "https://download.pytorch.org/whl/cpu" "$_result"
+
+# 28) Trailing slash in UNSLOTH_PYTORCH_MIRROR is stripped
+_result=$(UNSLOTH_PYTORCH_MIRROR="https://mirror.example.com/whl/" run_func "none")
+assert_eq "trailing slash stripped -> mirror/cpu" "https://mirror.example.com/whl/cpu" "$_result"
+
 rm -f "$_FUNC_FILE"
 rm -rf "$_FAKE_SMI_DIR"
 rm -rf "$_TOOLS_DIR"
diff --git a/tests/sh/test_mac_intel_compat.sh b/tests/sh/test_mac_intel_compat.sh
index d8848fd019..be9f756f90 100644
--- a/tests/sh/test_mac_intel_compat.sh
+++ b/tests/sh/test_mac_intel_compat.sh
@@ -206,6 +206,10 @@ assert_eq "Darwin -> cpu (even with nvidia-smi)" "https://download.pytorch.org/w
 _result=$(PATH="$_MOCK_UNAME_DIR:$_TOOLS_DIR" bash -c ". '$_FUNC_FILE'; get_torch_index_url" 2>/dev/null)
 assert_eq "Darwin -> cpu (no nvidia-smi)" "https://download.pytorch.org/whl/cpu" "$_result"
 
+# Test: Darwin + UNSLOTH_PYTORCH_MIRROR produces mirror/cpu
+_result=$(UNSLOTH_PYTORCH_MIRROR="https://mirror.example.com/whl" PATH="$_MOCK_UNAME_DIR:$_TOOLS_DIR" bash -c ". '$_FUNC_FILE'; get_torch_index_url" 2>/dev/null)
+assert_eq "Darwin + mirror env -> mirror/cpu" "https://mirror.example.com/whl/cpu" "$_result"
+
 rm -f "$_FUNC_FILE"
 rm -rf "$_FAKE_SMI_DIR" "$_TOOLS_DIR" "$_MOCK_UNAME_DIR" "$_GPU_DIR"
 
@@ -242,15 +246,6 @@ else
     FAIL=$((FAIL + 1))
 fi
 
-# Verify the PyTorch skip message exists (now covers both --no-torch and Intel Mac)
-if grep -q 'Skipping PyTorch' "$INSTALL_SH"; then
-    echo "  PASS: PyTorch skip message found"
-    PASS=$((PASS + 1))
-else
-    echo "  FAIL: PyTorch skip message not found"
-    FAIL=$((FAIL + 1))
-fi
-
 # Verify SKIP_TORCH unified variable exists
 if grep -q 'SKIP_TORCH=true' "$INSTALL_SH"; then
     echo "  PASS: SKIP_TORCH=true assignment found"
@@ -558,16 +553,7 @@ assert_eq "MAC_INTEL=true alone sets SKIP_TORCH=true" "true" "$_result"
 rm -f "$_SKIP_SNIPPET" "$_SKIP_SNIPPET2"
 
 echo ""
-echo "=== CPU hint printing ==="
-
-# Verify the CPU hint is present in install.sh source
-if grep -q 'No NVIDIA GPU detected' "$INSTALL_SH"; then
-    echo "  PASS: CPU hint message found in install.sh"
-    PASS=$((PASS + 1))
-else
-    echo "  FAIL: CPU hint message not found in install.sh"
-    FAIL=$((FAIL + 1))
-fi
+echo "=== --no-torch flag in install.sh ==="
 
 if grep -q '\-\-no-torch' "$INSTALL_SH"; then
     echo "  PASS: --no-torch appears in install.sh"

From f9ef639ddea1439cada3b0e9f4f81daa0ba92e92 Mon Sep 17 00:00:00 2001
From: Lee Jackson <130007945+Imagineer99@users.noreply.github.com>
Date: Wed, 15 Apr 2026 12:32:01 +0100
Subject: [PATCH 100/216] Studio: support GGUF variant selection for
 non-suffixed repos (#5023)

* fix: support GGUF variant selection for non-suffixed repos

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

* fix: harden GGUF detection across cached models and picker flows

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

* chore: use shared GGUF picker helper for search rows

* fix: avoid mixed cache duplication and preserve GGUF fallback detection

* fix: unify GGUF cache matching and merge picker hints

* fix: normalize local GGUF matching across picker and model config

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

* fix: robust cached-gguf classification + hint-aware click routing

- _repo_gguf_size_bytes: treat size_on_disk=None as 0 and dedupe fallback
  by commit_hash so partial/interrupted downloads don't TypeError out of
  sum() and wipe the entire cached list.
- list_cached_gguf / list_cached_models: narrow per-repo try/except so
  one malformed repo no longer poisons the whole response.
- handleModelClick: route through isKnownGgufRepo instead of the
  suffix-only isGgufRepo, so non-suffixed GGUF repos still open the
  variant expander from every call site.
- Replace the modelIsGgufById/resultIsGgufById Maps with Sets of known
  GGUF ids to stop conflating "no hint" with "known not-GGUF".
- Make HfModelResult.isGguf required (it is always set in makeMapModel).
- Add regression tests for the None size case, mixed-repo inclusion in
  cached-gguf, and per-repo error isolation.

* fix: exclude mmproj from GGUF classification and case-normalize hint lookups

- _repo_gguf_size_bytes now filters mmproj vision-adapter files so
  safetensors+mmproj.gguf repos stay on the cached-models path and
  non-GGUF rows no longer show zero pickable variants. A vision-capable
  GGUF repo (main weight + mmproj adapter) still classifies as GGUF and
  reports the main weight size.
- modelGgufIds / resultGgufIds now key on lowercased ids and
  isKnownGgufRepo lowercases its lookup, so store and HF-search ids
  that differ only by casing still match the same GGUF hint.
- New regression tests: mmproj-only repo excluded from cached-gguf,
  same repo included in cached-models, vision-capable repo still
  classified as GGUF with correct size.

---------

Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com>
Co-authored-by: Roland Tannous <rolandtannous@gravityq.ai>
Co-authored-by: Roland Tannous <115670425+rolandtannous@users.noreply.github.com>
---
 studio/backend/routes/models.py               | 154 ++++---
 .../backend/tests/test_cached_gguf_routes.py  | 398 ++++++++++++++++++
 studio/backend/utils/models/model_config.py   |  34 +-
 .../assistant-ui/model-selector/pickers.tsx   | 114 +++--
 .../assistant-ui/model-selector/types.ts      |   2 +-
 .../frontend/src/features/chat/chat-page.tsx  |   1 +
 .../frontend/src/hooks/use-hf-model-search.ts |   8 +-
 7 files changed, 621 insertions(+), 90 deletions(-)
 create mode 100644 studio/backend/tests/test_cached_gguf_routes.py

diff --git a/studio/backend/routes/models.py b/studio/backend/routes/models.py
index 8dd332eaf1..4ce2d0787c 100644
--- a/studio/backend/routes/models.py
+++ b/studio/backend/routes/models.py
@@ -992,7 +992,7 @@ async def get_gguf_variants(
                     snapshots = entry / "snapshots"
                     if snapshots.is_dir():
                         for snap in snapshots.iterdir():
-                            for f in snap.rglob("*.gguf"):
+                            for f in _iter_gguf_paths(snap):
                                 q = _extract_quant_label(f.name)
                                 cached_bytes_by_quant[q] = (
                                     cached_bytes_by_quant.get(q, 0) + f.stat().st_size
@@ -1061,7 +1061,7 @@ async def get_gguf_download_progress(
         for entry in cache_dir.iterdir():
             if entry.name.lower() == target:
                 # Count completed .gguf files matching this variant in snapshots
-                for f in entry.rglob("*.gguf"):
+                for f in _iter_gguf_paths(entry):
                     fname = f.name.lower().replace("-", "").replace("_", "")
                     if not variant_lower or variant_lower in fname:
                         downloaded_bytes += f.stat().st_size
@@ -1237,6 +1237,62 @@ def _all_hf_cache_scans():
     return scans
 
 
+def _is_gguf_filename(name: str) -> bool:
+    return name.lower().endswith(".gguf")
+
+
+def _is_mmproj_filename(name: str) -> bool:
+    """Match GGUF vision-adapter (mmproj) files. Kept consistent with
+    ``utils.models.model_config._is_mmproj``."""
+    return "mmproj" in name.lower()
+
+
+def _is_main_gguf_filename(name: str) -> bool:
+    """A GGUF file that is a primary weight artifact, not an mmproj
+    vision adapter."""
+    return _is_gguf_filename(name) and not _is_mmproj_filename(name)
+
+
+def _iter_gguf_paths(root: Path):
+    for path in root.rglob("*"):
+        if path.is_file() and _is_gguf_filename(path.name):
+            yield path
+
+
+def _repo_gguf_size_bytes(repo_info) -> int:
+    """Return the total on-disk size of primary GGUF weight files across
+    all revisions, excluding mmproj vision-adapter files.
+
+    Hugging Face hardlinks blobs shared between revisions, so this
+    deduplicates by blob path (or, as a fallback, by revision commit
+    hash + filename) to avoid double-counting the same bytes. Files
+    with an unknown size (``size_on_disk is None``, e.g. a partial or
+    interrupted download) are treated as zero bytes. mmproj files are
+    excluded so that repos whose only ``.gguf`` artifact is a vision
+    adapter are not classified as GGUF repos: the variant selector
+    filters mmproj out and would otherwise show zero pickable variants.
+    """
+    unique_blobs: dict[str, int] = {}
+    for revision in repo_info.revisions:
+        rev_id = getattr(revision, "commit_hash", None) or str(id(revision))
+        for f in revision.files:
+            if _is_main_gguf_filename(f.file_name):
+                blob_path = getattr(f, "blob_path", None)
+                size = f.size_on_disk or 0
+                if blob_path:
+                    unique_blobs[str(blob_path)] = size
+                else:
+                    unique_blobs[f"{rev_id}:{f.file_name}"] = size
+    return sum(unique_blobs.values())
+
+
+def _repo_has_gguf_files(repo_info) -> bool:
+    """Return True when any revision in a cached repo contains a
+    primary GGUF weight file. Repos whose only ``.gguf`` artifact is
+    an mmproj vision adapter are not treated as GGUF here."""
+    return _repo_gguf_size_bytes(repo_info) > 0
+
+
 @router.get("/cached-gguf")
 async def list_cached_gguf(
     current_subject: str = Depends(get_current_subject),
@@ -1248,28 +1304,25 @@ async def list_cached_gguf(
         seen_lower: dict[str, dict] = {}
         for hf_cache in cache_scans:
             for repo_info in hf_cache.repos:
-                if repo_info.repo_type != "model":
+                try:
+                    if repo_info.repo_type != "model":
+                        continue
+                    repo_id = repo_info.repo_id
+                    total_size = _repo_gguf_size_bytes(repo_info)
+                    if total_size == 0:
+                        continue
+                    key = repo_id.lower()
+                    existing = seen_lower.get(key)
+                    if existing is None or total_size > existing["size_bytes"]:
+                        seen_lower[key] = {
+                            "repo_id": repo_id,
+                            "size_bytes": total_size,
+                            "cache_path": str(repo_info.repo_path),
+                        }
+                except Exception as e:
+                    repo_label = getattr(repo_info, "repo_id", "<unknown>")
+                    logger.warning(f"Skipping cached GGUF repo {repo_label}: {e}")
                     continue
-                repo_id = repo_info.repo_id
-                if not repo_id.upper().endswith("-GGUF"):
-                    continue
-                total_size = 0
-                has_gguf = False
-                for revision in repo_info.revisions:
-                    for f in revision.files:
-                        if f.file_name.endswith(".gguf"):
-                            has_gguf = True
-                            total_size += f.size_on_disk
-                if not has_gguf:
-                    continue
-                key = repo_id.lower()
-                existing = seen_lower.get(key)
-                if existing is None or total_size > existing["size_bytes"]:
-                    seen_lower[key] = {
-                        "repo_id": repo_id,
-                        "size_bytes": total_size,
-                        "cache_path": str(repo_info.repo_path),
-                    }
         cached = sorted(seen_lower.values(), key = lambda c: c["repo_id"])
         return {"cached": cached}
     except Exception as e:
@@ -1290,30 +1343,37 @@ async def list_cached_models(
         seen_lower: dict[str, dict] = {}
         for hf_cache in cache_scans:
             for repo_info in hf_cache.repos:
-                if repo_info.repo_type != "model":
+                try:
+                    if repo_info.repo_type != "model":
+                        continue
+                    repo_id = repo_info.repo_id
+                    if _repo_has_gguf_files(repo_info):
+                        continue
+                    total_size = sum(
+                        (f.size_on_disk or 0)
+                        for rev in repo_info.revisions
+                        for f in rev.files
+                    )
+                    if total_size == 0:
+                        continue
+                    has_weights = any(
+                        f.file_name.endswith(_WEIGHT_EXTENSIONS)
+                        for rev in repo_info.revisions
+                        for f in rev.files
+                    )
+                    if not has_weights:
+                        continue
+                    key = repo_id.lower()
+                    existing = seen_lower.get(key)
+                    if existing is None or total_size > existing["size_bytes"]:
+                        seen_lower[key] = {
+                            "repo_id": repo_id,
+                            "size_bytes": total_size,
+                        }
+                except Exception as e:
+                    repo_label = getattr(repo_info, "repo_id", "<unknown>")
+                    logger.warning(f"Skipping cached model repo {repo_label}: {e}")
                     continue
-                repo_id = repo_info.repo_id
-                if repo_id.upper().endswith("-GGUF"):
-                    continue
-                total_size = sum(
-                    f.size_on_disk for rev in repo_info.revisions for f in rev.files
-                )
-                if total_size == 0:
-                    continue
-                has_weights = any(
-                    f.file_name.endswith(_WEIGHT_EXTENSIONS)
-                    for rev in repo_info.revisions
-                    for f in rev.files
-                )
-                if not has_weights:
-                    continue
-                key = repo_id.lower()
-                existing = seen_lower.get(key)
-                if existing is None or total_size > existing["size_bytes"]:
-                    seen_lower[key] = {
-                        "repo_id": repo_id,
-                        "size_bytes": total_size,
-                    }
         cached = sorted(seen_lower.values(), key = lambda c: c["repo_id"])
         return {"cached": cached}
     except Exception as e:
@@ -1390,7 +1450,7 @@ async def delete_cached_model(
             deleted_count = 0
             for rev in target_repo.revisions:
                 for f in rev.files:
-                    if not f.file_name.endswith(".gguf"):
+                    if not _is_gguf_filename(f.file_name):
                         continue
                     quant = _extract_quant_label(f.file_name)
                     if quant.lower() != variant.lower():
diff --git a/studio/backend/tests/test_cached_gguf_routes.py b/studio/backend/tests/test_cached_gguf_routes.py
new file mode 100644
index 0000000000..05aae8fb75
--- /dev/null
+++ b/studio/backend/tests/test_cached_gguf_routes.py
@@ -0,0 +1,398 @@
+# SPDX-License-Identifier: AGPL-3.0-only
+# Copyright 2026-present the Unsloth AI Inc. team. All rights reserved. See /studio/LICENSE.AGPL-3.0
+
+import asyncio
+import sys
+import types
+from pathlib import Path
+from types import SimpleNamespace
+
+# Keep this test runnable in lightweight environments where optional logging
+# deps are not installed.
+if "structlog" not in sys.modules:
+
+    class _DummyLogger:
+        def __getattr__(self, _name):
+            return lambda *args, **kwargs: None
+
+    sys.modules["structlog"] = types.SimpleNamespace(
+        BoundLogger = _DummyLogger,
+        get_logger = lambda *args, **kwargs: _DummyLogger(),
+    )
+
+import routes.models as models_route
+
+
+def _repo(
+    repo_id: str,
+    files: list[SimpleNamespace],
+    repo_path: Path,
+    *,
+    revisions: list[SimpleNamespace] | None = None,
+) -> SimpleNamespace:
+    return SimpleNamespace(
+        repo_id = repo_id,
+        repo_type = "model",
+        repo_path = repo_path,
+        revisions = revisions or [SimpleNamespace(files = files)],
+    )
+
+
+def _file(
+    name: str,
+    size_on_disk: int,
+    *,
+    blob_path: str | None = None,
+) -> SimpleNamespace:
+    return SimpleNamespace(
+        file_name = name,
+        size_on_disk = size_on_disk,
+        blob_path = blob_path,
+    )
+
+
+def test_iter_gguf_paths_matches_extension_case_insensitively(tmp_path):
+    nested = tmp_path / "snapshots" / "rev"
+    nested.mkdir(parents = True)
+    lower = nested / "Q4_K_M.gguf"
+    upper = nested / "Q8_0.GGUF"
+    other = nested / "README.md"
+    lower.write_text("a")
+    upper.write_text("b")
+    other.write_text("c")
+
+    result = sorted(path.name for path in models_route._iter_gguf_paths(tmp_path))
+
+    assert result == ["Q4_K_M.gguf", "Q8_0.GGUF"]
+
+
+def test_list_cached_gguf_includes_non_suffix_repo_when_cache_contains_gguf(
+    monkeypatch, tmp_path
+):
+    repo = _repo(
+        "HauhauCS/Gemma-4-E4B-Uncensored-HauhauCS-Aggressive",
+        [_file("Q4_K_M.gguf", 5_000), _file("README.md", 10)],
+        tmp_path / "models--HauhauCS--Gemma",
+    )
+    scan = SimpleNamespace(repos = [repo])
+
+    monkeypatch.setattr(models_route, "_all_hf_cache_scans", lambda: [scan])
+
+    result = asyncio.run(models_route.list_cached_gguf(current_subject = "test-user"))
+
+    assert result["cached"] == [
+        {
+            "repo_id": "HauhauCS/Gemma-4-E4B-Uncensored-HauhauCS-Aggressive",
+            "size_bytes": 5_000,
+            "cache_path": str(repo.repo_path),
+        }
+    ]
+
+
+def test_list_cached_gguf_matches_extension_case_insensitively(monkeypatch, tmp_path):
+    repo = _repo(
+        "Org/Model-Without-Suffix",
+        [_file("Q8_0.GGUF", 7_000)],
+        tmp_path / "models--Org--Model-Without-Suffix",
+    )
+    scan = SimpleNamespace(repos = [repo])
+
+    monkeypatch.setattr(models_route, "_all_hf_cache_scans", lambda: [scan])
+
+    result = asyncio.run(models_route.list_cached_gguf(current_subject = "test-user"))
+
+    assert result["cached"] == [
+        {
+            "repo_id": "Org/Model-Without-Suffix",
+            "size_bytes": 7_000,
+            "cache_path": str(repo.repo_path),
+        }
+    ]
+
+
+def test_list_cached_gguf_skips_repos_without_positive_gguf_size(monkeypatch, tmp_path):
+    missing = _repo(
+        "Org/ReadmeOnly",
+        [_file("README.md", 10)],
+        tmp_path / "models--Org--ReadmeOnly",
+    )
+    zero = _repo(
+        "Org/ZeroSize",
+        [_file("Q4_K_M.gguf", 0)],
+        tmp_path / "models--Org--ZeroSize",
+    )
+    scan = SimpleNamespace(repos = [missing, zero])
+
+    monkeypatch.setattr(models_route, "_all_hf_cache_scans", lambda: [scan])
+
+    result = asyncio.run(models_route.list_cached_gguf(current_subject = "test-user"))
+
+    assert result["cached"] == []
+
+
+def test_list_cached_gguf_keeps_largest_duplicate_repo_across_scans(
+    monkeypatch, tmp_path
+):
+    smaller = _repo(
+        "Org/Dupe",
+        [_file("Q4_K_M.gguf", 2_000)],
+        tmp_path / "models--Org--Dupe-a",
+    )
+    larger = _repo(
+        "org/dupe",
+        [_file("Q4_K_M.gguf", 5_000), _file("Q6_K.gguf", 1_000)],
+        tmp_path / "models--Org--Dupe-b",
+    )
+
+    monkeypatch.setattr(
+        models_route,
+        "_all_hf_cache_scans",
+        lambda: [
+            SimpleNamespace(repos = [smaller]),
+            SimpleNamespace(repos = [larger]),
+        ],
+    )
+
+    result = asyncio.run(models_route.list_cached_gguf(current_subject = "test-user"))
+
+    assert result["cached"] == [
+        {
+            "repo_id": "org/dupe",
+            "size_bytes": 6_000,
+            "cache_path": str(larger.repo_path),
+        }
+    ]
+
+
+def test_list_cached_gguf_dedupes_shared_blobs_across_revisions(monkeypatch, tmp_path):
+    shared = "blobs/shared-q4"
+    repo = _repo(
+        "Org/SharedBlobRepo",
+        [],
+        tmp_path / "models--Org--SharedBlobRepo",
+        revisions = [
+            SimpleNamespace(files = [_file("Q4_K_M.gguf", 5_000, blob_path = shared)]),
+            SimpleNamespace(files = [_file("Q4_K_M.gguf", 5_000, blob_path = shared)]),
+        ],
+    )
+
+    monkeypatch.setattr(
+        models_route,
+        "_all_hf_cache_scans",
+        lambda: [SimpleNamespace(repos = [repo])],
+    )
+
+    result = asyncio.run(models_route.list_cached_gguf(current_subject = "test-user"))
+
+    assert result["cached"] == [
+        {
+            "repo_id": "Org/SharedBlobRepo",
+            "size_bytes": 5_000,
+            "cache_path": str(repo.repo_path),
+        }
+    ]
+
+
+def test_list_cached_models_skips_non_suffix_repo_when_gguf_files_exist(
+    monkeypatch, tmp_path
+):
+    mixed = _repo(
+        "Org/MixedRepo",
+        [
+            _file("Q4_K_M.gguf", 5_000),
+            _file("model.safetensors", 10_000),
+        ],
+        tmp_path / "models--Org--MixedRepo",
+    )
+
+    monkeypatch.setattr(
+        models_route,
+        "_all_hf_cache_scans",
+        lambda: [SimpleNamespace(repos = [mixed])],
+    )
+
+    result = asyncio.run(models_route.list_cached_models(current_subject = "test-user"))
+
+    assert result["cached"] == []
+
+
+def test_list_cached_gguf_includes_mixed_repo_with_gguf_and_safetensors(
+    monkeypatch, tmp_path
+):
+    """Mirror of the _skips_ test: the mixed repo should still surface in
+    cached-gguf so the picker can show it as a GGUF download."""
+    mixed = _repo(
+        "Org/MixedRepo",
+        [
+            _file("Q4_K_M.gguf", 5_000),
+            _file("model.safetensors", 10_000),
+        ],
+        tmp_path / "models--Org--MixedRepo",
+    )
+
+    monkeypatch.setattr(
+        models_route,
+        "_all_hf_cache_scans",
+        lambda: [SimpleNamespace(repos = [mixed])],
+    )
+
+    result = asyncio.run(models_route.list_cached_gguf(current_subject = "test-user"))
+
+    assert result["cached"] == [
+        {
+            "repo_id": "Org/MixedRepo",
+            "size_bytes": 5_000,
+            "cache_path": str(mixed.repo_path),
+        }
+    ]
+
+
+def test_list_cached_gguf_handles_none_size_on_disk(monkeypatch, tmp_path):
+    """A partial/interrupted GGUF download has ``size_on_disk = None``. The
+    route must treat the unknown bytes as zero instead of raising TypeError
+    out of ``sum()`` and wiping the entire response."""
+    partial = _repo(
+        "Org/PartialDownload",
+        [_file("Q4_K_M.gguf", None), _file("Q6_K.gguf", 5_000)],
+        tmp_path / "models--Org--PartialDownload",
+    )
+
+    monkeypatch.setattr(
+        models_route,
+        "_all_hf_cache_scans",
+        lambda: [SimpleNamespace(repos = [partial])],
+    )
+
+    result = asyncio.run(models_route.list_cached_gguf(current_subject = "test-user"))
+
+    assert result["cached"] == [
+        {
+            "repo_id": "Org/PartialDownload",
+            "size_bytes": 5_000,
+            "cache_path": str(partial.repo_path),
+        }
+    ]
+
+
+def test_list_cached_gguf_skips_malformed_repo_without_wiping_response(
+    monkeypatch, tmp_path
+):
+    """One repo raising during classification must not poison the response
+    for every other repo in the scan."""
+
+    class _ExplodingRepo:
+        repo_id = "Org/Broken"
+        repo_type = "model"
+        repo_path = tmp_path / "models--Org--Broken"
+
+        @property
+        def revisions(self):
+            raise RuntimeError("boom")
+
+    healthy = _repo(
+        "Org/Healthy",
+        [_file("Q4_K_M.gguf", 5_000)],
+        tmp_path / "models--Org--Healthy",
+    )
+
+    monkeypatch.setattr(
+        models_route,
+        "_all_hf_cache_scans",
+        lambda: [SimpleNamespace(repos = [_ExplodingRepo(), healthy])],
+    )
+
+    result = asyncio.run(models_route.list_cached_gguf(current_subject = "test-user"))
+
+    assert result["cached"] == [
+        {
+            "repo_id": "Org/Healthy",
+            "size_bytes": 5_000,
+            "cache_path": str(healthy.repo_path),
+        }
+    ]
+
+
+def test_list_cached_gguf_skips_repo_with_only_mmproj_gguf(monkeypatch, tmp_path):
+    """A repo whose only ``.gguf`` artifact is an mmproj vision adapter
+    must not be classified as a GGUF repo: the variant selector filters
+    mmproj out and the picker would otherwise show zero variants."""
+    mmproj_only = _repo(
+        "Org/MmprojOnly",
+        [
+            _file("mmproj-Q8_0.gguf", 5_000),
+            _file("model.safetensors", 10_000),
+        ],
+        tmp_path / "models--Org--MmprojOnly",
+    )
+
+    monkeypatch.setattr(
+        models_route,
+        "_all_hf_cache_scans",
+        lambda: [SimpleNamespace(repos = [mmproj_only])],
+    )
+
+    result = asyncio.run(models_route.list_cached_gguf(current_subject = "test-user"))
+
+    assert result["cached"] == []
+
+
+def test_list_cached_models_includes_repo_with_only_mmproj_gguf(monkeypatch, tmp_path):
+    """Mirror of the cached-gguf skip: a safetensors repo with an
+    auxiliary mmproj vision adapter must still surface in cached-models
+    so the user can load it as a normal model."""
+    mmproj_aux = _repo(
+        "Org/MmprojAux",
+        [
+            _file("mmproj-Q8_0.gguf", 5_000),
+            _file("model.safetensors", 10_000),
+        ],
+        tmp_path / "models--Org--MmprojAux",
+    )
+
+    monkeypatch.setattr(
+        models_route,
+        "_all_hf_cache_scans",
+        lambda: [SimpleNamespace(repos = [mmproj_aux])],
+    )
+
+    result = asyncio.run(models_route.list_cached_models(current_subject = "test-user"))
+
+    assert result["cached"] == [
+        {
+            "repo_id": "Org/MmprojAux",
+            "size_bytes": 15_000,
+        }
+    ]
+
+
+def test_list_cached_gguf_includes_vision_repo_with_main_gguf_and_mmproj(
+    monkeypatch, tmp_path
+):
+    """A vision-capable GGUF repo (main weight + mmproj adapter) is still
+    a GGUF repo. The reported size is the main weight size; mmproj is
+    excluded from the GGUF-size accounting because it is filtered out at
+    classification time."""
+    vision_repo = _repo(
+        "Org/VisionGguf",
+        [
+            _file("Q4_K_M.gguf", 5_000),
+            _file("mmproj-Q8_0.gguf", 1_000),
+        ],
+        tmp_path / "models--Org--VisionGguf",
+    )
+
+    monkeypatch.setattr(
+        models_route,
+        "_all_hf_cache_scans",
+        lambda: [SimpleNamespace(repos = [vision_repo])],
+    )
+
+    result = asyncio.run(models_route.list_cached_gguf(current_subject = "test-user"))
+
+    assert result["cached"] == [
+        {
+            "repo_id": "Org/VisionGguf",
+            "size_bytes": 5_000,
+            "cache_path": str(vision_repo.repo_path),
+        }
+    ]
diff --git a/studio/backend/utils/models/model_config.py b/studio/backend/utils/models/model_config.py
index 81331c9dde..fae2337bbd 100644
--- a/studio/backend/utils/models/model_config.py
+++ b/studio/backend/utils/models/model_config.py
@@ -904,6 +904,18 @@ def _is_mmproj(filename: str) -> bool:
     return "mmproj" in filename.lower()
 
 
+def _is_gguf_filename(filename: str) -> bool:
+    return filename.lower().endswith(".gguf")
+
+
+def _iter_gguf_files(directory: Path):
+    if not directory.is_dir():
+        return
+    for f in directory.iterdir():
+        if f.is_file() and _is_gguf_filename(f.name):
+            yield f
+
+
 def detect_mmproj_file(path: str) -> Optional[str]:
     """
     Find the mmproj (vision projection) GGUF file in a directory.
@@ -919,7 +931,7 @@ def detect_mmproj_file(path: str) -> Optional[str]:
     if not search_dir.is_dir():
         return None
 
-    for f in search_dir.glob("*.gguf"):
+    for f in _iter_gguf_files(search_dir):
         if _is_mmproj(f.name):
             return str(f.resolve())
     return None
@@ -942,7 +954,7 @@ def detect_gguf_model(path: str) -> Optional[str]:
     p = Path(path)
 
     # Case 1: direct .gguf file
-    if p.suffix == ".gguf" and p.is_file():
+    if p.suffix.lower() == ".gguf" and p.is_file():
         if _is_mmproj(p.name):
             return None
         return str(p.resolve())
@@ -950,7 +962,7 @@ def detect_gguf_model(path: str) -> Optional[str]:
     # Case 2: directory containing .gguf files (skip mmproj)
     if p.is_dir():
         gguf_files = sorted(
-            (f for f in p.glob("*.gguf") if not _is_mmproj(f.name)),
+            (f for f in _iter_gguf_files(p) if not _is_mmproj(f.name)),
             key = lambda f: f.stat().st_size,
             reverse = True,
         )
@@ -1015,7 +1027,7 @@ def _pick_best_gguf(filenames: list[str]) -> Optional[str]:
     Prefers quantization levels in _GGUF_QUANT_PREFERENCE order.
     Falls back to the first .gguf file found.
     """
-    gguf_files = [f for f in filenames if f.endswith(".gguf")]
+    gguf_files = [f for f in filenames if f.lower().endswith(".gguf")]
     if not gguf_files:
         return None
 
@@ -1100,7 +1112,7 @@ def list_gguf_variants(
 
     for sibling in info.siblings:
         fname = sibling.rfilename
-        if not fname.endswith(".gguf"):
+        if not fname.lower().endswith(".gguf"):
             continue
         size = sibling.size or 0
 
@@ -1171,7 +1183,7 @@ def list_local_gguf_variants(
     quant_first_file: dict[str, str] = {}
     has_vision = False
 
-    for f in sorted(p.glob("*.gguf")):
+    for f in sorted(_iter_gguf_files(p)):
         if _is_mmproj(f.name):
             has_vision = True
             continue
@@ -1210,7 +1222,7 @@ def _find_local_gguf_by_variant(directory: str, variant: str) -> Optional[str]:
 
     matches = sorted(
         f
-        for f in p.glob("*.gguf")
+        for f in _iter_gguf_files(p)
         if not _is_mmproj(f.name) and _extract_quant_label(f.name) == variant
     )
     if matches:
@@ -1438,7 +1450,9 @@ def scan_exported_models(
 
             # Check for flat GGUF export (e.g. exports/gemma-3-4b-it-finetune-gguf/)
             # Filter out mmproj (vision projection) files — they aren't loadable as main models
-            gguf_files = [f for f in run_dir.glob("*.gguf") if not _is_mmproj(f.name)]
+            gguf_files = [
+                f for f in _iter_gguf_files(run_dir) if not _is_mmproj(f.name)
+            ]
             if gguf_files:
                 base_model = None
                 export_meta = run_dir / "export_metadata.json"
@@ -1465,7 +1479,7 @@ def scan_exported_models(
                 has_weights = any(checkpoint_dir.glob("*.safetensors")) or any(
                     checkpoint_dir.glob("*.bin")
                 )
-                has_gguf = any(checkpoint_dir.glob("*.gguf"))
+                has_gguf = any(_iter_gguf_files(checkpoint_dir))
 
                 base_model = None
                 export_type = None
@@ -1488,7 +1502,7 @@ def scan_exported_models(
                         pass
                 elif has_gguf:
                     export_type = "gguf"
-                    gguf_list = list(checkpoint_dir.glob("*.gguf"))
+                    gguf_list = list(_iter_gguf_files(checkpoint_dir))
                     # Check checkpoint_dir first, then fall back to parent run_dir
                     # (export.py writes metadata to the top-level export directory)
                     for meta_dir in (checkpoint_dir, run_dir):
diff --git a/studio/frontend/src/components/assistant-ui/model-selector/pickers.tsx b/studio/frontend/src/components/assistant-ui/model-selector/pickers.tsx
index e410f8cff9..313a950cc1 100644
--- a/studio/frontend/src/components/assistant-ui/model-selector/pickers.tsx
+++ b/studio/frontend/src/components/assistant-ui/model-selector/pickers.tsx
@@ -399,10 +399,14 @@ function GgufVariantExpander({
   );
 }
 
-// ── Detect GGUF repos by naming convention ────────────────────
+// ── Detect GGUF repos by naming convention or hub tag ────────────────────
 
-function isGgufRepo(id: string): boolean {
-  return id.toUpperCase().includes("-GGUF");
+function hasGgufSuffix(id: string): boolean {
+  return /-GGUF(?:$|-)/i.test(id);
+}
+
+function isGgufRepo(id: string, hintedIsGguf?: boolean): boolean {
+  return Boolean(hintedIsGguf) || hasGgufSuffix(id);
 }
 
 /** Extract param count label from model name (e.g. "Qwen3-0.6B" -> "0.6B"). */
@@ -451,6 +455,33 @@ export function HubModelPicker({
   const { results, isLoading, isLoadingMore, fetchMore } =
     useHfModelSearch(debouncedQuery);
 
+  // Sets of lowercased repo ids that the store or HF search have
+  // confirmed are GGUF. Absence means "no hint" and lets hasGgufSuffix
+  // take over as fallback, rather than conflating unknown with known-
+  // not-GGUF. Keys are lowercased so that store IDs and HF search IDs
+  // that differ only by casing still match the same hint.
+  const modelGgufIds = useMemo(() => {
+    const ids = new Set<string>();
+    for (const model of models) {
+      if (model.isGguf) ids.add(model.id.toLowerCase());
+    }
+    return ids;
+  }, [models]);
+  const resultGgufIds = useMemo(() => {
+    const ids = new Set<string>();
+    for (const result of results) {
+      if (result.isGguf) ids.add(result.id.toLowerCase());
+    }
+    return ids;
+  }, [results]);
+  const isKnownGgufRepo = useCallback(
+    (id: string): boolean => {
+      const key = id.toLowerCase();
+      return isGgufRepo(id, resultGgufIds.has(key) || modelGgufIds.has(key));
+    },
+    [modelGgufIds, resultGgufIds],
+  );
+
   // Track which GGUF repo is expanded for variant selection
   const [expandedGguf, setExpandedGguf] = useState<string | null>(null);
 
@@ -625,17 +656,17 @@ export function HubModelPicker({
   const recommendedIds = useMemo(() => {
     const all = dedupe([...models.map((model) => model.id), value ?? ""])
       .filter((id) => !downloadedSet.has(id.toLowerCase()))
-      .filter((id) => !chatOnly || isGgufRepo(id))
+      .filter((id) => !chatOnly || isKnownGgufRepo(id))
       .filter((id) => !/-FP8[-.]|FP8-Dynamic/i.test(id));
     // Sort: GGUFs first, then hub models
     const gguf: string[] = [];
     const hub: string[] = [];
     for (const id of all) {
-      if (isGgufRepo(id)) gguf.push(id);
+      if (isKnownGgufRepo(id)) gguf.push(id);
       else hub.push(id);
     }
     return [...gguf, ...hub];
-  }, [models, value, downloadedSet, chatOnly]);
+  }, [models, value, downloadedSet, chatOnly, isKnownGgufRepo]);
 
   // Infinite scroll paging for the recommended section
   const [recommendedPage, setRecommendedPage] = useState(1);
@@ -645,7 +676,7 @@ export function HubModelPicker({
   }, [models, chatOnly]);
 
   const visibleRecommendedIds = useMemo(() => {
-    const hubStartIndex = recommendedIds.findIndex((id) => !isGgufRepo(id));
+    const hubStartIndex = recommendedIds.findIndex((id) => !isKnownGgufRepo(id));
     const allGguf =
       hubStartIndex === -1
         ? recommendedIds
@@ -659,7 +690,7 @@ export function HubModelPicker({
       result.push(...allHub.slice(p * 4, (p + 1) * 4));
     }
     return result;
-  }, [recommendedIds, recommendedPage]);
+  }, [recommendedIds, recommendedPage, isKnownGgufRepo]);
 
   const hasMoreRecommended =
     visibleRecommendedIds.length < recommendedIds.length;
@@ -681,8 +712,8 @@ export function HubModelPicker({
     const ids = showHfSection
       ? [...new Set([...visibleRecommendedIds, ...filteredRecommendedIds])]
       : visibleRecommendedIds;
-    return ids.filter((id) => !isGgufRepo(id));
-  }, [visibleRecommendedIds, showHfSection, filteredRecommendedIds]);
+    return ids.filter((id) => !isKnownGgufRepo(id));
+  }, [visibleRecommendedIds, showHfSection, filteredRecommendedIds, isKnownGgufRepo]);
   const { paramCountById: recommendedParamCountById } =
     useRecommendedModelVram(idsForVram);
 
@@ -697,9 +728,9 @@ export function HubModelPicker({
     return results
       .map((result) => result.id)
       .filter((id) => !recommendedSet.has(id))
-      .filter((id) => !chatOnly || isGgufRepo(id))
+      .filter((id) => !chatOnly || isKnownGgufRepo(id))
       .filter((id) => !/-FP8[-.]|FP8-Dynamic/i.test(id));
-  }, [recommendedSet, results, showHfSection, chatOnly]);
+  }, [recommendedSet, results, showHfSection, chatOnly, isKnownGgufRepo]);
 
   const metricsById = useMemo(
     () =>
@@ -800,14 +831,14 @@ export function HubModelPicker({
   /** Handle clicking a model row — GGUF repos expand, others load directly. */
   const handleModelClick = useCallback(
     (id: string) => {
-      if (isGgufRepo(id)) {
+      if (isKnownGgufRepo(id)) {
         // Toggle GGUF variant expander
         setExpandedGguf((prev) => (prev === id ? null : id));
       } else {
         onSelect(id, { source: "hub", isLora: false });
       }
     },
-    [onSelect],
+    [onSelect, isKnownGgufRepo],
   );
 
   return (
@@ -848,7 +879,11 @@ export function HubModelPicker({
                     label={c.repo_id}
                     meta={`GGUF · ${formatBytes(c.size_bytes)}`}
                     selected={value === c.repo_id}
-                    onClick={() => handleModelClick(c.repo_id)}
+                    onClick={() =>
+                      setExpandedGguf((prev) =>
+                        prev === c.repo_id ? null : c.repo_id,
+                      )
+                    }
                     vramStatus={null}
                   />
                   {expandedGguf === c.repo_id && (
@@ -909,7 +944,7 @@ export function HubModelPicker({
                     <ModelRow
                       label={m.model_id ?? m.display_name}
                       meta={
-                        isGguf || m.path.endsWith(".gguf") ? "GGUF" : "Local"
+                        isGguf || m.path.toLowerCase().endsWith(".gguf") ? "GGUF" : "Local"
                       }
                       selected={value === m.id}
                       onClick={() => {
@@ -1036,7 +1071,7 @@ export function HubModelPicker({
                 const isGguf =
                   isGgufRepo(m.id) ||
                   isGgufRepo(m.display_name) ||
-                  m.path.endsWith(".gguf");
+                  m.path.toLowerCase().endsWith(".gguf");
                 return (
                   <div key={m.id}>
                     <ModelRow
@@ -1089,16 +1124,22 @@ export function HubModelPicker({
                       <ModelRow
                         label={id}
                         meta={
-                          isGgufRepo(id)
+                          isKnownGgufRepo(id)
                             ? "GGUF"
                             : (vram?.detail ?? extractParamLabel(id))
                         }
                         selected={value === id}
-                        onClick={() => handleModelClick(id)}
+                        onClick={() => {
+                          if (isKnownGgufRepo(id)) {
+                            setExpandedGguf((prev) => (prev === id ? null : id));
+                          } else {
+                            handleModelClick(id);
+                          }
+                        }}
                         vramStatus={
-                          isGgufRepo(id) ? null : (vram?.status ?? null)
+                          isKnownGgufRepo(id) ? null : (vram?.status ?? null)
                         }
-                        vramEst={isGgufRepo(id) ? undefined : vram?.est}
+                        vramEst={isKnownGgufRepo(id) ? undefined : vram?.est}
                         gpuGb={gpu.available ? gpu.memoryTotalGb : undefined}
                       />
                       {expandedGguf === id && (
@@ -1136,16 +1177,22 @@ export function HubModelPicker({
                     <ModelRow
                       label={id}
                       meta={
-                        isGgufRepo(id)
+                        isKnownGgufRepo(id)
                           ? "GGUF"
                           : (vram?.detail ?? extractParamLabel(id))
                       }
                       selected={value === id}
-                      onClick={() => handleModelClick(id)}
+                      onClick={() => {
+                        if (isKnownGgufRepo(id)) {
+                          setExpandedGguf((prev) => (prev === id ? null : id));
+                        } else {
+                          handleModelClick(id);
+                        }
+                      }}
                       vramStatus={
-                        isGgufRepo(id) ? null : (vram?.status ?? null)
+                        isKnownGgufRepo(id) ? null : (vram?.status ?? null)
                       }
-                      vramEst={isGgufRepo(id) ? undefined : vram?.est}
+                      vramEst={isKnownGgufRepo(id) ? undefined : vram?.est}
                       gpuGb={gpu.available ? gpu.memoryTotalGb : undefined}
                     />
                     {expandedGguf === id && (
@@ -1175,24 +1222,31 @@ export function HubModelPicker({
                     No matching models.
                   </div>
                 ) : null
-              ) : (
+                ) : (
                 hfIds.map((id) => {
                   const vram = vramMap.get(id);
+                  const isSearchGguf = isKnownGgufRepo(id);
                   return (
                     <div key={id}>
                       <ModelRow
                         label={id}
                         meta={
-                          isGgufRepo(id)
+                          isSearchGguf
                             ? "GGUF"
                             : (metricsById.get(id) ?? extractParamLabel(id))
                         }
                         selected={value === id}
-                        onClick={() => handleModelClick(id)}
+                        onClick={() => {
+                          if (isSearchGguf) {
+                            setExpandedGguf((prev) => (prev === id ? null : id));
+                          } else {
+                            handleModelClick(id);
+                          }
+                        }}
                         vramStatus={
-                          isGgufRepo(id) ? null : (vram?.status ?? null)
+                          isSearchGguf ? null : (vram?.status ?? null)
                         }
-                        vramEst={isGgufRepo(id) ? undefined : vram?.est}
+                        vramEst={isSearchGguf ? undefined : vram?.est}
                         gpuGb={gpu.available ? gpu.memoryTotalGb : undefined}
                       />
                       {expandedGguf === id && (
diff --git a/studio/frontend/src/components/assistant-ui/model-selector/types.ts b/studio/frontend/src/components/assistant-ui/model-selector/types.ts
index f70cfc3b01..215dd2b38e 100644
--- a/studio/frontend/src/components/assistant-ui/model-selector/types.ts
+++ b/studio/frontend/src/components/assistant-ui/model-selector/types.ts
@@ -8,6 +8,7 @@ export interface ModelOption {
   name: string;
   description?: string;
   icon?: ReactNode;
+  isGguf?: boolean;
 }
 
 export interface LoraModelOption extends ModelOption {
@@ -24,4 +25,3 @@ export interface ModelSelectorChangeMeta {
   isDownloaded?: boolean;
   expectedBytes?: number;
 }
-
diff --git a/studio/frontend/src/features/chat/chat-page.tsx b/studio/frontend/src/features/chat/chat-page.tsx
index 165eba0f2c..c3153cdec7 100644
--- a/studio/frontend/src/features/chat/chat-page.tsx
+++ b/studio/frontend/src/features/chat/chat-page.tsx
@@ -691,6 +691,7 @@ export function ChatPage(): ReactElement {
         id: model.id,
         name: model.name,
         description: model.description,
+        isGguf: model.isGguf,
       })),
     [modelsFromStore],
   );
diff --git a/studio/frontend/src/hooks/use-hf-model-search.ts b/studio/frontend/src/hooks/use-hf-model-search.ts
index e697544d5c..77214f38d9 100644
--- a/studio/frontend/src/hooks/use-hf-model-search.ts
+++ b/studio/frontend/src/hooks/use-hf-model-search.ts
@@ -13,6 +13,7 @@ export interface HfModelResult {
   likes: number;
   totalParams?: number;
   estimatedSizeBytes?: number;
+  isGguf: boolean;
 }
 
 const EXCLUDED_TAGS = new Set([
@@ -89,7 +90,10 @@ function makeMapModel(excludeGguf: boolean) {
     if (!isEmbedding && m.tags?.some((t) => EXCLUDED_TAGS.has(t))) {
       return null;
     }
-    if (excludeGguf && m.tags?.includes("gguf")) {
+    const isGguf =
+      Boolean(m.tags?.some((tag) => tag.toLowerCase() === "gguf")) ||
+      /-GGUF(?:$|-)/i.test(m.name);
+    if (excludeGguf && isGguf) {
       return null;
     }
     return {
@@ -98,6 +102,7 @@ function makeMapModel(excludeGguf: boolean) {
       likes: m.likes,
       totalParams: m.safetensors?.total,
       estimatedSizeBytes: estimateSizeFromDtypes(m.safetensors?.parameters),
+      isGguf,
     };
   };
 }
@@ -330,4 +335,3 @@ export function useHfModelSearch(
 
   return { ...search, results };
 }
-

From 1fcb2502cf5c8677ab93ca680132fff0458bf6f2 Mon Sep 17 00:00:00 2001
From: David Solanas Sanz <43669187+DavidSolanas@users.noreply.github.com>
Date: Wed, 15 Apr 2026 13:51:31 +0200
Subject: [PATCH 101/216] fix: prevent offline freeze by fixing stats retry and
 forwarding local_files_only (#5016)

Fixes #2393.

- `_utils.py`: `has_internet()` now respects `HF_HUB_OFFLINE` with truthy variant parsing in addition to `TRANSFORMERS_OFFLINE`.
- `_utils.py`: replace uncontrolled `except Exception: stats_check()` retry (which had no time limit and could freeze on Kaggle offline mode) with a logged skip.
- `loader.py`: forward `local_files_only` from kwargs into all `AutoConfig.from_pretrained` and `PeftConfig.from_pretrained` probes in `FastLanguageModel.from_pretrained` and `FastModel.from_pretrained`, including the PEFT base-model reload paths.
---
 unsloth/models/_utils.py |  9 +++++++--
 unsloth/models/loader.py | 10 ++++++++++
 2 files changed, 17 insertions(+), 2 deletions(-)

diff --git a/unsloth/models/_utils.py b/unsloth/models/_utils.py
index 71f7db915c..d0cd4ba028 100644
--- a/unsloth/models/_utils.py
+++ b/unsloth/models/_utils.py
@@ -1350,6 +1350,11 @@ import socket
 def has_internet(host = "8.8.8.8", port = 53, timeout = 3):
     if os.environ.get("TRANSFORMERS_OFFLINE", "0") == "1":
         return False
+
+    OFFLINE_TRUE = {"1", "true", "yes", "on"}
+
+    if os.environ.get("HF_HUB_OFFLINE", "").strip().lower() in OFFLINE_TRUE:
+        return False
     try:
         socket.setdefaulttimeout(timeout)
         sock = socket.socket(socket.AF_INET, socket.SOCK_STREAM)
@@ -1468,8 +1473,8 @@ def _get_statistics(statistics = None, force_download = True):
                     "```"
                 )
             except Exception:
-                # Try no time limit check
-                stats_check()
+                logger.debug("Unsloth: stats_check failed with an exception.")
+                # Don't retry without a time limit — would freeze offline
 
 
 def get_statistics(local_files_only = False):
diff --git a/unsloth/models/loader.py b/unsloth/models/loader.py
index df97c5c7df..a66a7c6abc 100644
--- a/unsloth/models/loader.py
+++ b/unsloth/models/loader.py
@@ -439,12 +439,15 @@ class FastLanguageModel(FastLlamaModel):
         peft_error = None
         model_config = None
         peft_config = None
+        local_files_only = kwargs.get("local_files_only", False)
+
         try:
             model_config = AutoConfig.from_pretrained(
                 model_name,
                 token = token,
                 revision = revision,
                 trust_remote_code = trust_remote_code,
+                local_files_only = local_files_only,
             )
             is_model = True
         except ImportError:
@@ -470,6 +473,7 @@ class FastLanguageModel(FastLlamaModel):
                 token = token,
                 revision = revision,
                 trust_remote_code = trust_remote_code,
+                local_files_only = local_files_only,
             )
             is_peft = True
         except ImportError:
@@ -566,6 +570,7 @@ class FastLanguageModel(FastLlamaModel):
                 model_name,
                 token = token,
                 trust_remote_code = trust_remote_code,
+                local_files_only = local_files_only,
             )
 
         if not was_disabled:
@@ -1049,12 +1054,15 @@ class FastModel(FastBaseModel):
         peft_error = None
         model_config = None
         peft_config = None
+        local_files_only = kwargs.get("local_files_only", False)
+
         try:
             model_config = AutoConfig.from_pretrained(
                 model_name,
                 token = token,
                 revision = revision,
                 trust_remote_code = trust_remote_code,
+                local_files_only = local_files_only,
             )
             is_model = True
         except ImportError:
@@ -1080,6 +1088,7 @@ class FastModel(FastBaseModel):
                 token = token,
                 revision = revision,
                 trust_remote_code = trust_remote_code,
+                local_files_only = local_files_only,
             )
             is_peft = True
         except ImportError:
@@ -1330,6 +1339,7 @@ class FastModel(FastBaseModel):
                 model_name,
                 token = token,
                 trust_remote_code = trust_remote_code,
+                local_files_only = local_files_only,
             )
 
         if not was_disabled:

From b7a8ff2833ee7d123c4fc875e0e3c10b3cd5218f Mon Sep 17 00:00:00 2001
From: Daniel Han <danielhanchen@gmail.com>
Date: Wed, 15 Apr 2026 05:16:33 -0700
Subject: [PATCH 102/216] Respect classification head skip list on
 pre-quantized 4-bit checkpoints (#5027) (#5034)

* Respect classification head skip list on pre-quantized 4-bit checkpoints (#5027)

FastLanguageModel.from_pretrained(..., num_labels=N) crashed with
"NotImplementedError: normal_kernel_cuda not implemented for 'Byte'" on
pre-quantized bnb 4-bit checkpoints (e.g. unsloth/Qwen3-4B-bnb-4bit)
when running on transformers 5.x.

Two pieces were needed to close this out:

1. unsloth_zoo PR: add "score", "classifier", "qa_outputs" to
   SKIP_QUANTIZATION_MODULES so replace_with_bnb_linear leaves task
   heads in the compute dtype.

2. This commit: for pre-quantized checkpoints, transformers reads
   llm_int8_skip_modules from the quantization_config baked into
   config.json and ignores the runtime BitsAndBytesConfig we pass via
   kwargs. Unsloth must merge its skip list into
   model_config.quantization_config.llm_int8_skip_modules before the
   from_pretrained call, or the checkpoint's frozen list
   (e.g. ["lm_head", "multi_modal_projector", "merger",
   "modality_projection"]) wins and the `score` head gets converted to
   Linear4bit with uint8 storage, then _init_weights calls normal_ on
   uint8 and crashes.

Also add a defensive post-load cast on the task head to guard against
any residual path that ends up with a non-floating head dtype.

Verified on transformers 4.57.6 and 5.5.0 with:
- unsloth/Qwen3-4B-bnb-4bit + num_labels=3
- unsloth/Qwen3-4B (non-bnb repo, load_in_4bit=True)
- unsloth/Llama-3.2-1B-Instruct + num_labels=3
- unsloth/ModernBERT-large classifier head (bert_classification notebook)
- Regression: causal LM path unchanged, backbone still 4-bit
- 3-step SFT on num_labels=3 confirms gradient flow and weight updates
  on score.weight

Fixes unslothai/unsloth#5027

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

---------

Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com>
---
 unsloth/models/llama.py | 37 +++++++++++++++++++++++++++++++++++++
 1 file changed, 37 insertions(+)

diff --git a/unsloth/models/llama.py b/unsloth/models/llama.py
index 031cabcd39..63cd9b1c8a 100644
--- a/unsloth/models/llama.py
+++ b/unsloth/models/llama.py
@@ -2405,6 +2405,31 @@ class FastLlamaModel:
                 bnb_4bit_compute_dtype = dtype,
                 llm_int8_skip_modules = llm_int8_skip_modules,
             )
+            # For pre-quantized checkpoints (e.g. unsloth/Qwen3-4B-bnb-4bit),
+            # transformers uses the quantization_config baked into the
+            # checkpoint's config.json and ignores the runtime BitsAndBytesConfig
+            # we pass via kwargs. Merge our skip list into that bundled config
+            # so task heads like `score` (for *ForSequenceClassification) stay
+            # in the compute dtype. See unslothai/unsloth#5027.
+            _ckpt_qcfg = getattr(model_config, "quantization_config", None)
+            if _ckpt_qcfg is not None:
+                if isinstance(_ckpt_qcfg, dict):
+                    _ckpt_skip = list(_ckpt_qcfg.get("llm_int8_skip_modules") or [])
+                    for _m in llm_int8_skip_modules:
+                        if _m not in _ckpt_skip:
+                            _ckpt_skip.append(_m)
+                    _ckpt_qcfg["llm_int8_skip_modules"] = _ckpt_skip
+                else:
+                    _ckpt_skip = list(
+                        getattr(_ckpt_qcfg, "llm_int8_skip_modules", None) or []
+                    )
+                    for _m in llm_int8_skip_modules:
+                        if _m not in _ckpt_skip:
+                            _ckpt_skip.append(_m)
+                    try:
+                        _ckpt_qcfg.llm_int8_skip_modules = _ckpt_skip
+                    except Exception:
+                        pass
 
         # https://huggingface.co/togethercomputer/LLaMA-2-7B-32K/discussions/12
         # RoPE Scaling's max_position_embeddings must be updated
@@ -2441,6 +2466,18 @@ class FastLlamaModel:
                 attn_implementation = preferred_attn_impl,
                 **kwargs,
             )
+            # Defensive: make sure the task head ended up in a floating dtype.
+            # The primary protection is SKIP_QUANTIZATION_MODULES plus the skip
+            # list merge above; this guards against a downstream path accidentally
+            # leaving the head in an integer storage. See unslothai/unsloth#5027.
+            for _head_name in ("score", "classifier", "qa_outputs"):
+                _head = getattr(model, _head_name, None)
+                if (
+                    _head is not None
+                    and hasattr(_head, "weight")
+                    and not _head.weight.is_floating_point()
+                ):
+                    _head.to(dtype)
         elif not fast_inference:
             model = AutoModelForCausalLM.from_pretrained(
                 model_name,

From 1ccfd2e0a5e56e78ed36ee93cbaaa12462e5b5b3 Mon Sep 17 00:00:00 2001
From: Daniel Han <danielhanchen@gmail.com>
Date: Wed, 15 Apr 2026 05:24:41 -0700
Subject: [PATCH 103/216] fix(rocm): tighten gfx regex to ignore generic ISA
 lines (#5033)

* fix(rocm): tighten gfx regex to ignore generic ISA lines

ROCm 6.1+ rocminfo emits generic ISA names such as
"amdgcn-amd-amdhsa--gfx11-generic" and "amdgcn-amd-amdhsa--gfx9-4-generic"
alongside the real GPU name. The previous `gfx[1-9]` regex used in
`_has_rocm_gpu` matched both, so a host with only a generic ISA entry
would be reported as having a usable AMD GPU.

Tighten the pattern to `gfx[1-9][0-9a-z]{2,3}` so only real gfx ids
match. This covers every documented target from GFX6 (gfx600) through
GFX12 (gfx1201), including letter-suffixed ids like gfx90a (MI250 /
MI250X) and gfx90c. Documented generic ISA names always have 1 or 2
digits before the dash and no longer match.

Applied to both `studio/install_python_stack.py` and
`studio/install_llama_prebuilt.py` so the two detection paths agree.

Co-authored-by: Martin Hoyer <mhoyer@redhat.com>

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

---------

Co-authored-by: Martin Hoyer <mhoyer@redhat.com>
Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com>
---
 studio/install_llama_prebuilt.py | 10 ++++++++--
 studio/install_python_stack.py   | 10 ++++++++--
 2 files changed, 16 insertions(+), 4 deletions(-)

diff --git a/studio/install_llama_prebuilt.py b/studio/install_llama_prebuilt.py
index e43a78baea..fd2b78d05f 100755
--- a/studio/install_llama_prebuilt.py
+++ b/studio/install_llama_prebuilt.py
@@ -2576,8 +2576,14 @@ def detect_host() -> HostInfo:
     has_rocm = False
     if is_linux:
         for _cmd, _check in (
-            # rocminfo: look for "gfxNNNN" with nonzero first digit (gfx000 is CPU agent)
-            (["rocminfo"], lambda out: bool(re.search(r"gfx[1-9]", out.lower()))),
+            # rocminfo: look for a real gfx GPU id (3-4 chars, nonzero first digit).
+            # gfx000 is the CPU agent; ROCm 6.1+ also emits generic ISA lines like
+            # "gfx11-generic" or "gfx9-4-generic" which only have 1-2 digits before
+            # the dash and must not be treated as a real GPU.
+            (
+                ["rocminfo"],
+                lambda out: bool(re.search(r"gfx[1-9][0-9a-z]{2,3}", out.lower())),
+            ),
             (["amd-smi", "list"], _amd_smi_has_gpu),
         ):
             _exe = shutil.which(_cmd[0])
diff --git a/studio/install_python_stack.py b/studio/install_python_stack.py
index cf024d4076..2b9fd084d4 100644
--- a/studio/install_python_stack.py
+++ b/studio/install_python_stack.py
@@ -183,8 +183,14 @@ def _has_rocm_gpu() -> bool:
     import re
 
     for cmd, check_fn in (
-        # rocminfo: look for "Name: gfxNNNN" with nonzero first digit (gfx000 is the CPU agent)
-        (["rocminfo"], lambda out: bool(re.search(r"gfx[1-9]", out.lower()))),
+        # rocminfo: look for a real gfx GPU id (3-4 chars, nonzero first digit).
+        # gfx000 is the CPU agent; ROCm 6.1+ also emits generic ISA lines like
+        # "gfx11-generic" or "gfx9-4-generic" which only have 1-2 digits before
+        # the dash and must not be treated as a real GPU.
+        (
+            ["rocminfo"],
+            lambda out: bool(re.search(r"gfx[1-9][0-9a-z]{2,3}", out.lower())),
+        ),
         # amd-smi list: require "GPU: <number>" data rows, not just a header
         (
             ["amd-smi", "list"],

From 1a4ca5eca8d5b37f77e3d08f361d43e824cb7838 Mon Sep 17 00:00:00 2001
From: Daniel Han <danielhanchen@gmail.com>
Date: Wed, 15 Apr 2026 06:59:36 -0700
Subject: [PATCH 104/216] Fix grad-accum accepts_loss_kwargs detection for
 vision wrappers (#5036)

* Fix grad-accum model_accepts_loss_kwargs detection for vision wrappers

Replace the source-string rewrite of Trainer.__init__ with an instance-level
accepts_loss_kwargs shadow applied on the loaded model. Covers:

  1. Unsloth-compiled forward -> True, so HF Trainer does not double-scale
     on top of unsloth_fixed_cross_entropy's num_items_in_batch division.
  2. Stock forward on a conditional-generation wrapper (Gemma3n, Gemma3
     pre-4.57, Qwen-VL family, etc.) where the outer class has no
     accepts_loss_kwargs but the inner .model declares False -> False.
     This is the case that reproduces issue #4982 under trust_remote_code
     or UNSLOTH_COMPILE_DISABLE, where the previous fix's outer-attr
     check walked past the inner model and fell through to signature
     inspection.
  3. Text LMs without any explicit accepts_loss_kwargs -> leave HF default.

The previous .replace()-based patch silently no-ops on transformers 4.48
through 4.52 (variable named model, not unwrapped_model) and is fragile
against any upstream reformat. The new helper walks the PEFT / HF wrapper
chain, finds the first class that declares accepts_loss_kwargs on its own
class dict (type(m).__dict__, not hasattr, to avoid PEFT __getattr__
forwarding), and setattr-shadows that value at every wrapper level so
HF Trainer's hasattr(unwrapped_model, ...) check picks it up at whichever
level accelerate.unwrap_model returns.

Also adds an unconditional post-init clamp of
accelerator.gradient_accumulation_steps = 1 to work around the
transformers 5.0 through 5.5 GradientAccumulationPlugin regression that
makes accelerator.backward divide loss by GA on top of training_step's
own /GA division. Fixed upstream in 5.6.0.dev0; no-op on 4.x and 5.6+.

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

* Trim comments

* Address review: cover PEFT-after-load and custom compile location

Two review findings from 3/20 reviewers:

1. [3 of 20 reviewers] apply_accepts_loss_kwargs_fix was called from the
   loaders before get_peft_model wraps the base model, so on transformers
   4.48-4.52 (which does hasattr on the outer model) the instance shadow
   on the base model was lost after PEFT wrapping. Fix: also call it from
   the wrapped Trainer.__init__ so it runs on whatever model the user
   actually hands to Trainer, which is always the final wrapped form.

2. [1 of 20 reviewers] _forward_is_unsloth_compiled hard-coded the
   substrings "unsloth_compiled" / "unsloth_cache" in the co_filename
   check, which misclassifies compiled forwards when
   UNSLOTH_COMPILE_LOCATION is set to a custom directory. Fix: new
   _unsloth_compile_cache_leaves helper that reads the env var and
   matches the basename against path components, honoring both the
   default and any user override.

Verified locally:
- PEFT-after-load simulation: HF's hasattr(peft, "accepts_loss_kwargs")
  now returns True after our init wrapper runs, and value resolves to
  False on Gemma3n-style inner wrappers.
- Custom UNSLOTH_COMPILE_LOCATION simulation: compiled detection returns
  True for /tmp/my_custom_cache/compiled.py when the env var is set.
- End-to-end Gemma-3 270m + LoRA SFT unchanged: loss 4.9626, grad-norm
  matches prior run, all 4 wrapper levels now carry the shadowed attr.

---------

Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com>
---
 unsloth/models/_utils.py | 174 +++++++++++++++++++++++++++++++--------
 unsloth/models/llama.py  |   3 +-
 unsloth/models/vision.py |   3 +-
 3 files changed, 142 insertions(+), 38 deletions(-)

diff --git a/unsloth/models/_utils.py b/unsloth/models/_utils.py
index d0cd4ba028..c330dcc32c 100644
--- a/unsloth/models/_utils.py
+++ b/unsloth/models/_utils.py
@@ -45,6 +45,7 @@ __all__ = [
     # "accelerate_old_send_to_device",
     # "accelerate_new_send_to_device",
     "patch_gradient_accumulation_fix",
+    "apply_accepts_loss_kwargs_fix",
     "patch_compiling_bitsandbytes",
     "patch_regional_compilation",
     "patch_layernorm",
@@ -2083,47 +2084,148 @@ def patch_gradient_accumulation_fix(Trainer):
         exec(function, globals())
         Trainer.training_step = _unsloth_training_step
 
-    # Prevent double scaling gradient accumulation
-    # https://github.com/huggingface/transformers/pull/37208
-    # Patch model_accepts_loss_kwargs detection in Trainer.__init__
-    if Trainer.__init__.__name__ != "_unsloth___init__":
+    # Wrap Trainer.__init__: (1) pre-init, shadow accepts_loss_kwargs on whatever
+    # model was passed in (covers PEFT wrapping done after FastModel.from_pretrained);
+    # (2) post-init, clamp accelerator GA to 1 for the transformers 5.0-5.5
+    # GradientAccumulationPlugin regression. No-op on 4.x and 5.6+. See #4982.
+    if not getattr(Trainer, "_unsloth_init_wrapped_for_accelerate_gas", False):
+        _original_trainer_init = Trainer.__init__
+
+        def _unsloth_trainer_init(self, *args, **kwargs):
+            model = kwargs.get("model")
+            if model is None and len(args) > 0:
+                model = args[0]
+            if model is not None:
+                try:
+                    apply_accepts_loss_kwargs_fix(model)
+                except Exception:
+                    pass
+            _original_trainer_init(self, *args, **kwargs)
+            try:
+                accelerator = getattr(self, "accelerator", None)
+                if (
+                    accelerator is not None
+                    and getattr(accelerator, "gradient_accumulation_steps", 1) > 1
+                ):
+                    accelerator.gradient_accumulation_steps = 1
+                    gs = getattr(accelerator, "gradient_state", None)
+                    if gs is not None and hasattr(gs, "plugin_kwargs"):
+                        try:
+                            gs.plugin_kwargs["num_steps"] = 1
+                        except Exception:
+                            pass
+            except Exception:
+                pass
+
+        _unsloth_trainer_init.__wrapped__ = _original_trainer_init
+        Trainer.__init__ = _unsloth_trainer_init
+        Trainer._unsloth_init_wrapped_for_accelerate_gas = True
+
+
+def _unsloth_compile_cache_leaves():
+    # Accepts `UNSLOTH_COMPILE_LOCATION` overrides (the env var unsloth_zoo honors).
+    leaves = {"unsloth_compiled_cache", "unsloth_cache", "unsloth_compiled"}
+    loc = os.environ.get("UNSLOTH_COMPILE_LOCATION", "") or ""
+    loc = loc.rstrip("/\\")
+    if loc:
+        leaves.add(os.path.basename(loc) or loc)
+    return leaves
+
+
+def _forward_is_unsloth_compiled(model):
+    # True iff forward was installed from the Unsloth compile cache directory.
+    # __module__ stays as the transformers module, so check co_filename.
+    leaves = _unsloth_compile_cache_leaves()
+
+    def check(m):
+        if m is None:
+            return False
+        fwd = getattr(type(m), "forward", None)
+        if fwd is None:
+            return False
+        code = getattr(fwd, "__code__", None)
+        fn = getattr(code, "co_filename", "") if code is not None else ""
+        fn = fn.replace("\\", "/")
+        parts = set(fn.split("/"))
+        return any(leaf in parts for leaf in leaves)
+
+    if check(model):
+        return True
+    seen = set()
+    m = model
+    for _ in range(4):
+        if m is None or id(m) in seen:
+            break
+        seen.add(id(m))
+        nxt = getattr(m, "base_model", None)
+        if nxt is None or nxt is m:
+            nxt = getattr(m, "model", None)
+        if nxt is None or nxt is m:
+            break
+        if check(nxt):
+            return True
+        m = nxt
+    return False
+
+
+def _find_concrete_accepts_loss_kwargs(model):
+    # Walk wrapper chain for first class that declares accepts_loss_kwargs in its
+    # own __mro__ dict. Avoids PEFT __getattr__ forwarding and our own shadow.
+    seen = set()
+    m = model
+    for _ in range(6):
+        if m is None or id(m) in seen:
+            break
+        seen.add(id(m))
+        for klass in type(m).__mro__:
+            if "accepts_loss_kwargs" in klass.__dict__:
+                return klass.__dict__[
+                    "accepts_loss_kwargs"
+                ], f"{klass.__name__}.accepts_loss_kwargs"
+        nxt = getattr(m, "base_model", None)
+        if nxt is None or nxt is m:
+            nxt = getattr(m, "model", None)
+        if nxt is None or nxt is m:
+            break
+        m = nxt
+    return None, "no explicit accepts_loss_kwargs on any wrapper level"
+
+
+def _shadow_accepts_loss_kwargs(model, value):
+    # Set the attribute at every wrapper level so HF's hasattr check resolves
+    # regardless of where accelerator / peft unwrap lands.
+    seen = set()
+    m = model
+    for _ in range(8):
+        if m is None or id(m) in seen:
+            break
+        seen.add(id(m))
         try:
-            init_function = inspect.getsource(Trainer.__init__)
+            setattr(m, "accepts_loss_kwargs", value)
         except Exception:
-            init_function = ""
-        if init_function is not None:
-            init_function = textwrap.dedent(init_function)
+            pass
+        nxt = getattr(m, "base_model", None)
+        if nxt is None or nxt is m:
+            nxt = getattr(m, "model", None)
+        if nxt is None or nxt is m:
+            break
+        m = nxt
 
-            # Import all variables that need importing
-            import transformers.trainer
 
-            items_in_trainer = dir(transformers.trainer)
-            good_items = []
-            for item in items_in_trainer:
-                if item in init_function:
-                    good_items.append(item)
-            exec(
-                "from transformers.trainer import ("
-                + ", ".join(x for x in good_items)
-                + ")",
-                globals(),
-            )
+def apply_accepts_loss_kwargs_fix(model):
+    # Shadow the correct accepts_loss_kwargs on the model so HF Trainer picks it
+    # up via hasattr(unwrapped_model, ...). Replaces the old Trainer.__init__
+    # source rewrite. Priority: compiled forward -> True; else first class attr
+    # in wrapper chain; else leave HF default. Issue #4982.
+    if _forward_is_unsloth_compiled(model):
+        _shadow_accepts_loss_kwargs(model, True)
+        return "True (Unsloth compiled forward)"
 
-            init_function = init_function.replace(
-                "def __init__", "def _unsloth___init__", 1
-            )
-
-            # Respect an inner wrapped model's explicit accepts_loss_kwargs flag before inferring from forward(**kwargs).
-            # https://github.com/unslothai/unsloth/issues/4982 Gemma4ForConditionalGeneration had issues with grad_acc
-            init_function = init_function.replace(
-                "self.model_accepts_loss_kwargs = unwrapped_model.accepts_loss_kwargs\n    else:",
-                "self.model_accepts_loss_kwargs = unwrapped_model.accepts_loss_kwargs\n"
-                '    elif hasattr(getattr(unwrapped_model, "model", None), "accepts_loss_kwargs"):\n'
-                "        self.model_accepts_loss_kwargs = unwrapped_model.model.accepts_loss_kwargs\n"
-                "    else:",
-            )
-            exec(init_function, globals())
-            Trainer.__init__ = _unsloth___init__
+    value, reason = _find_concrete_accepts_loss_kwargs(model)
+    if value is None:
+        return f"default (signature inspection, {reason})"
+    _shadow_accepts_loss_kwargs(model, value)
+    return f"{value} ({reason})"
 
 
 def patch_tokenizer(model, tokenizer):
diff --git a/unsloth/models/llama.py b/unsloth/models/llama.py
index 63cd9b1c8a..425df1c084 100644
--- a/unsloth/models/llama.py
+++ b/unsloth/models/llama.py
@@ -2695,7 +2695,8 @@ class FastLlamaModel:
         patch_saving_functions(model)
         Trainer._inner_training_loop = _fast_inner_training_loop
 
-        # Fix gradient accumulation
+        # Fix gradient accumulation. See issue #4982.
+        apply_accepts_loss_kwargs_fix(model)
         patch_gradient_accumulation_fix(Trainer)
 
         # Save tokenizer for inference purposes
diff --git a/unsloth/models/vision.py b/unsloth/models/vision.py
index 5abeb3a81a..2bdff55a56 100644
--- a/unsloth/models/vision.py
+++ b/unsloth/models/vision.py
@@ -1123,9 +1123,10 @@ class FastBaseModel:
                     )
         patch_saving_functions(tokenizer, vision = True)
 
-        # Fix gradient accumulation
+        # Fix gradient accumulation. See issue #4982.
         from transformers.trainer import Trainer
 
+        apply_accepts_loss_kwargs_fix(model)
         patch_gradient_accumulation_fix(Trainer)
 
         # Save tokenizer for inference purposes

From 777e1bd0ac34e85e139b3b0f9d03bca993d76170 Mon Sep 17 00:00:00 2001
From: jonahsamost <92005111+jonahsamost@users.noreply.github.com>
Date: Wed, 15 Apr 2026 07:21:03 -0700
Subject: [PATCH 105/216] fix (#4887)

---
 unsloth/models/rl_replacements.py | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/unsloth/models/rl_replacements.py b/unsloth/models/rl_replacements.py
index 93a7f89bcb..4d36af62cc 100755
--- a/unsloth/models/rl_replacements.py
+++ b/unsloth/models/rl_replacements.py
@@ -1154,6 +1154,7 @@ def grpo_trainer_compute_loss(function_name, function):
                 ref_logps,
                 per_token_logps,
                 old_logps,
+                sampling_per_token_logps,
                 input_ids,
                 completion_mask,
                 self.beta,
@@ -1174,7 +1175,6 @@ def grpo_trainer_compute_loss(function_name, function):
                 num_items_in_batch = num_items_in_batch,
                 current_gradient_accumulation_steps = current_gradient_accumulation_steps,
                 num_processes = num_processes,
-                sampling_per_token_logps = sampling_per_token_logps,
             )
         else:
             if hasattr(self.args, "loss_type"):

From 156f3fc4b07c7a306822d6ef82208dc9d127463f Mon Sep 17 00:00:00 2001
From: Daniel Han <danielhanchen@gmail.com>
Date: Wed, 15 Apr 2026 07:33:48 -0700
Subject: [PATCH 106/216] Gate trl disable_gradient_checkpointing patch warning
 on UNSLOTH_ENABLE_LOGGING (#5038)

The "Patched trl.models.utils.disable_gradient_checkpointing with a no-op"
warning fires once on every Unsloth import, including from notebooks where
the user did not opt into verbose logging. It is a routine integration
patch, not an anomaly the user needs to know about. Gate it on
UNSLOTH_ENABLE_LOGGING=1 like other diagnostic notices.
---
 unsloth/models/rl.py | 11 ++++++-----
 1 file changed, 6 insertions(+), 5 deletions(-)

diff --git a/unsloth/models/rl.py b/unsloth/models/rl.py
index f444f6bd37..ac9b35a822 100755
--- a/unsloth/models/rl.py
+++ b/unsloth/models/rl.py
@@ -2018,11 +2018,12 @@ def patch_trl_disable_gradient_checkpointing():
         except (AttributeError, TypeError):
             pass
 
-    logger.warning_once(
-        "Unsloth: Patched trl.models.utils.disable_gradient_checkpointing with "
-        "a no-op to preserve Unsloth gradient checkpointing across TRL "
-        "generation passes."
-    )
+    if os.environ.get("UNSLOTH_ENABLE_LOGGING", "0") == "1":
+        logger.warning_once(
+            "Unsloth: Patched trl.models.utils.disable_gradient_checkpointing with "
+            "a no-op to preserve Unsloth gradient checkpointing across TRL "
+            "generation passes."
+        )
     return
 
 

From c3cd890357a15c64bb6e0d810559e264e1c8044c Mon Sep 17 00:00:00 2001
From: Daniel Han <danielhanchen@gmail.com>
Date: Wed, 15 Apr 2026 07:34:42 -0700
Subject: [PATCH 107/216] Studio: refresh Downloaded GGUF list and recurse into
 variant subdirs (#5032)

* Studio: refresh Downloaded GGUF list and recurse into variant subdirs

Two fixes for the model picker's "Downloaded" section.

Frontend (`pickers.tsx`):
* `HubModelPicker`'s mount effect short-circuited the cached-gguf and
  cached-models refetch whenever the module-level cache already had
  entries (`if (alreadyCached) return;`). After downloading a new repo
  in the same session, reopening the picker rendered the stale cache
  and the new repo never appeared in "Downloaded" until a full page
  reload. The early return is removed so the lists are always refreshed
  on mount; the module cache still drives the initial render so there
  is no spinner flash when we already had data.

Backend (`utils/models/model_config.py`):
* `list_local_gguf_variants` and `_find_local_gguf_by_variant` used a
  non-recursive `Path.glob("*.gguf")`. Some HF GGUF repos (e.g.
  `unsloth/gemma-4-26B-A4B-it-GGUF`) place the largest quants under a
  variant-named subdirectory such as `BF16/...gguf`, which the
  top-level glob missed. Both helpers now use `rglob` and the variant
  filename is stored as a path relative to the scan root so the
  locator can still find the file.

The flat-layout case (variants directly in the snapshot root) is
unchanged: verified against `unsloth/gemma-4-E2B-it-GGUF` which still
returns its UD-Q4_K_XL variant correctly.

* Studio: emit posix-style relative filenames for local GGUF subdirs

`list_local_gguf_variants` was doing `str(f.relative_to(p))`, which on
Windows produces backslash-separated paths like `BF16\foo.gguf`. The
remote `list_gguf_variants` (HF API path) always returns forward-slash
filenames such as `BF16/foo.gguf`, so the two would diverge on Windows.

Switch to `.as_posix()` so the local and remote variant filenames stay
identical across Linux, macOS, and Windows. Verified by simulating with
`PureWindowsPath` in the test suite.

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

* Studio: detect mmproj at snapshot root for nested-variant layouts

When _find_local_gguf_by_variant returns a weight file inside a
quant-named subdir (e.g. snapshot/BF16/foo.gguf), detect_mmproj_file
was scanning only the immediate parent and missing the mmproj file
sitting at the snapshot root. The model was then loaded without
--mmproj, silently breaking vision support for repos that ship
nested variants.

detect_mmproj_file now takes an optional search_root and walks up
from the weight file to that root, in order, so the mmproj at the
snapshot root is picked up. Sibling quant subdirs are not scanned,
so an unrelated variant's mmproj does not leak in.

Also apply the suggested micro-optimization on relative_to in
list_local_gguf_variants -- only build the posix path when storing
the first file for a quant.

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

---------

Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com>
---
 studio/backend/utils/models/model_config.py   | 99 ++++++++++++++++---
 .../assistant-ui/model-selector/pickers.tsx   |  8 +-
 2 files changed, 90 insertions(+), 17 deletions(-)

diff --git a/studio/backend/utils/models/model_config.py b/studio/backend/utils/models/model_config.py
index fae2337bbd..44754520e3 100644
--- a/studio/backend/utils/models/model_config.py
+++ b/studio/backend/utils/models/model_config.py
@@ -908,32 +908,81 @@ def _is_gguf_filename(filename: str) -> bool:
     return filename.lower().endswith(".gguf")
 
 
-def _iter_gguf_files(directory: Path):
+def _iter_gguf_files(directory: Path, recursive: bool = False):
     if not directory.is_dir():
         return
-    for f in directory.iterdir():
+    iterator = directory.rglob("*") if recursive else directory.iterdir()
+    for f in iterator:
         if f.is_file() and _is_gguf_filename(f.name):
             yield f
 
 
-def detect_mmproj_file(path: str) -> Optional[str]:
+def detect_mmproj_file(path: str, search_root: Optional[str] = None) -> Optional[str]:
     """
-    Find the mmproj (vision projection) GGUF file in a directory.
+    Find the mmproj (vision projection) GGUF file for a given model.
 
     Args:
-        path: Directory to search — or a .gguf file (uses its parent dir).
+        path: Directory to search — or a .gguf file (uses its parent dir
+            as the starting point).
+        search_root: Optional outer directory that should also be scanned
+            (and any directory between it and ``path``). This handles
+            local layouts where the model weights live in a quant-named
+            subdir (``snapshot/BF16/foo.gguf``) but the mmproj sits at
+            the snapshot root (``snapshot/mmproj-BF16.gguf``). When
+            ``None``, only the immediate parent dir is scanned, matching
+            the historical behavior.
 
     Returns:
         Full path to the mmproj .gguf file, or None if not found.
     """
     p = Path(path)
-    search_dir = p.parent if p.is_file() else p
-    if not search_dir.is_dir():
+    start_dir = p.parent if p.is_file() else p
+    if not start_dir.is_dir():
         return None
 
-    for f in _iter_gguf_files(search_dir):
-        if _is_mmproj(f.name):
-            return str(f.resolve())
+    # Build the list of dirs to scan: immediate dir first, then walk up
+    # to (and including) ``search_root`` if it is an ancestor. We walk
+    # incrementally rather than recursing into ``search_root`` so we
+    # don't accidentally pick up an mmproj from a sibling subdir
+    # belonging to a different model variant.
+    seen: set[Path] = set()
+    scan_order: list[Path] = []
+
+    def _add(d: Path) -> None:
+        try:
+            resolved = d.resolve()
+        except OSError:
+            return
+        if resolved in seen or not resolved.is_dir():
+            return
+        seen.add(resolved)
+        scan_order.append(resolved)
+
+    _add(start_dir)
+    if search_root is not None:
+        try:
+            root_resolved = Path(search_root).resolve()
+            start_resolved = start_dir.resolve()
+            # Only walk if start_dir is inside (or equal to) search_root.
+            if root_resolved == start_resolved or (
+                start_resolved.is_relative_to(root_resolved)
+                if hasattr(start_resolved, "is_relative_to")
+                else str(start_resolved).startswith(str(root_resolved) + "/")
+            ):
+                cur = start_resolved
+                # Walk up from start_dir to (and including) root_resolved.
+                while cur != root_resolved and cur.parent != cur:
+                    cur = cur.parent
+                    _add(cur)
+                    if cur == root_resolved:
+                        break
+        except OSError:
+            pass
+
+    for d in scan_order:
+        for f in _iter_gguf_files(d):
+            if _is_mmproj(f.name):
+                return str(f.resolve())
     return None
 
 
@@ -1183,7 +1232,11 @@ def list_local_gguf_variants(
     quant_first_file: dict[str, str] = {}
     has_vision = False
 
-    for f in sorted(_iter_gguf_files(p)):
+    # Recurse so variant-specific subdirectories (e.g. ``BF16/...gguf``
+    # used by some HF GGUF repos for the largest quants) are picked up.
+    # Filenames in the result preserve the relative subpath so that
+    # ``_find_local_gguf_by_variant`` can locate the file again.
+    for f in sorted(_iter_gguf_files(p, recursive = True)):
         if _is_mmproj(f.name):
             has_vision = True
             continue
@@ -1193,8 +1246,14 @@ def list_local_gguf_variants(
             size = 0
         quant = _extract_quant_label(f.name)
         quant_totals[quant] = quant_totals.get(quant, 0) + size
+        # Only compute the (potentially expensive) relative path when this
+        # is the first file we've seen for this quant -- after that we'd
+        # discard the result anyway. Use posix-style separators so the
+        # filename matches what ``list_gguf_variants`` (the remote HF
+        # API path) returns on every platform; otherwise Windows would
+        # emit ``BF16\foo.gguf`` here.
         if quant not in quant_first_file:
-            quant_first_file[quant] = f.name
+            quant_first_file[quant] = f.relative_to(p).as_posix()
 
     variants = [
         GgufVariantInfo(
@@ -1220,9 +1279,11 @@ def _find_local_gguf_by_variant(directory: str, variant: str) -> Optional[str]:
     if p is None:
         return None
 
+    # Recurse into subdirectories so variants stored under a quant-named
+    # subdir (e.g. ``BF16/foo-BF16-00001-of-00002.gguf``) are found.
     matches = sorted(
         f
-        for f in _iter_gguf_files(p)
+        for f in _iter_gguf_files(p, recursive = True)
         if not _is_mmproj(f.name) and _extract_quant_label(f.name) == variant
     )
     if matches:
@@ -1932,8 +1993,16 @@ class ModelConfig:
                     except Exception as e:
                         logger.debug(f"Could not read export metadata: {e}")
 
-                # If vision (or mmproj happens to exist), find the mmproj file
-                mmproj_file = detect_mmproj_file(gguf_file)
+                # If vision (or mmproj happens to exist), find the mmproj
+                # file. The recursive variant scan in
+                # ``_find_local_gguf_by_variant`` may have returned a
+                # weight file inside a quant-named subdir (e.g.
+                # ``.../BF16/foo.gguf``) while ``mmproj-*.gguf`` lives
+                # at the snapshot root. Pass ``search_root=path`` so
+                # ``detect_mmproj_file`` walks up to the snapshot root
+                # instead of seeing only the weight file's immediate
+                # parent.
+                mmproj_file = detect_mmproj_file(gguf_file, search_root = path)
                 if mmproj_file:
                     gguf_is_vision = True
                     logger.info(f"Detected mmproj for vision: {mmproj_file}")
diff --git a/studio/frontend/src/components/assistant-ui/model-selector/pickers.tsx b/studio/frontend/src/components/assistant-ui/model-selector/pickers.tsx
index 313a950cc1..dc4c210be4 100644
--- a/studio/frontend/src/components/assistant-ui/model-selector/pickers.tsx
+++ b/studio/frontend/src/components/assistant-ui/model-selector/pickers.tsx
@@ -600,7 +600,11 @@ export function HubModelPicker({
     refreshLocalModelsList();
     refreshScanFolders();
 
-    if (alreadyCached) return;
+    // Always refetch cached GGUF/model lists. The module-level caches give
+    // an instant render with stale data (no spinner flash), but newly
+    // downloaded repos won't appear unless we re-hit the backend on every
+    // mount.  Initial state already has cachedReady=alreadyCached, so the
+    // background refresh is invisible when we already had data.
     let done = 0;
     const check = () => {
       if (++done >= 2) setCachedReady(true);
@@ -619,7 +623,7 @@ export function HubModelPicker({
       })
       .catch(() => {})
       .finally(check);
-  }, [alreadyCached, refreshLocalModelsList, refreshScanFolders]);
+  }, [refreshLocalModelsList, refreshScanFolders]);
 
   const handleDeleteConfirm = useCallback(async () => {
     if (!deleteTarget) return;

From f18e9dddf0cfaf1b6382dfe937430c3d480dd83c Mon Sep 17 00:00:00 2001
From: Avaya Aggarwal <119044997+OnePunchMonk@users.noreply.github.com>
Date: Wed, 15 Apr 2026 20:09:11 +0530
Subject: [PATCH 108/216] feat: Add support for OLMo-3 model (#4678)

* feat: Add support for OLMo-3 model in mapping and tests

* Update unsloth/models/mapper.py

Co-authored-by: gemini-code-assist[bot] <176961590+gemini-code-assist[bot]@users.noreply.github.com>

* Update tests/test_get_model_name.py

Co-authored-by: gemini-code-assist[bot] <176961590+gemini-code-assist[bot]@users.noreply.github.com>

* Fix casing, add Think variants, and align version gate for OLMo-3 PR 4678

Mapper: switch slugs from OLMo-3 to canonical Olmo-3 mixed case, drop the
non-existent unsloth/Olmo-3-7B-Instruct-bnb-4bit dead alias, and add the
already-published Olmo-3-7B-Think and Olmo-3-32B-Think Unsloth mirrors.

Loader: change the olmo3 transformers version gate from Version("4.57.0")
to Version("4.57.0.dev0") so nightly/source builds that already contain
olmo3 are not blocked, matching the OLMo-2, Gemma 3 and Cohere patterns.

* Use canonical Olmo-3 casing and cover Think variants in OLMo-3 tests

Mirrors the mapper.py fixes on pr-4678-code: HuggingFace canonical slugs
for the OLMo-3 family use mixed-case Olmo-3 (not OLMo-3 like OLMo-2), and
Unsloth already hosts Olmo-3-7B-Think and Olmo-3-32B-Think mirrors, so
the resolution matrix now covers all three published Olmo-3 families.

---------

Co-authored-by: gemini-code-assist[bot] <176961590+gemini-code-assist[bot]@users.noreply.github.com>
Co-authored-by: Daniel Han <danielhanchen@gmail.com>
---
 tests/test_get_model_name.py | 40 ++++++++++++++++++++++++++++++++++++
 unsloth/models/loader.py     |  9 +++++++-
 unsloth/models/mapper.py     | 12 +++++++++++
 3 files changed, 60 insertions(+), 1 deletion(-)

diff --git a/tests/test_get_model_name.py b/tests/test_get_model_name.py
index ad89f595f0..15dd44f0ec 100644
--- a/tests/test_get_model_name.py
+++ b/tests/test_get_model_name.py
@@ -64,6 +64,42 @@ class TestGetModelName(unittest.TestCase):
                 "unsloth/Ministral-3-3B-Instruct-2512",
                 True,
             ),
+            (
+                "allenai/Olmo-3-7B-Instruct",
+                True,
+                "unsloth/Olmo-3-7B-Instruct-unsloth-bnb-4bit",
+                True,
+            ),
+            (
+                "allenai/Olmo-3-7B-Instruct",
+                False,
+                "unsloth/Olmo-3-7B-Instruct",
+                True,
+            ),
+            (
+                "allenai/Olmo-3-7B-Think",
+                True,
+                "unsloth/Olmo-3-7B-Think-unsloth-bnb-4bit",
+                True,
+            ),
+            (
+                "allenai/Olmo-3-7B-Think",
+                False,
+                "unsloth/Olmo-3-7B-Think",
+                True,
+            ),
+            (
+                "allenai/Olmo-3-32B-Think",
+                True,
+                "unsloth/Olmo-3-32B-Think-unsloth-bnb-4bit",
+                True,
+            ),
+            (
+                "allenai/Olmo-3-32B-Think",
+                False,
+                "unsloth/Olmo-3-32B-Think",
+                True,
+            ),
             ("unsloth/Kimi-K2-Instruct", True, "unsloth/Kimi-K2-Instruct-BF16", True),
             ("unsloth/Kimi-K2-Instruct", False, "unsloth/Kimi-K2-Instruct", False),
             # Fallback-to-original behavior
@@ -113,6 +149,10 @@ class TestGetModelName(unittest.TestCase):
                 "mistralai/ministral-3-3b-instruct-2512",
                 "unsloth/ministral-3-3b-instruct-2512-unsloth-bnb-4bit",
             ),
+            (
+                "allenai/olmo-3-7b-instruct",
+                "unsloth/olmo-3-7b-instruct-unsloth-bnb-4bit",
+            ),
             ("unsloth/kimi-k2-instruct", "unsloth/kimi-k2-instruct-bf16"),
         ]
         for src, expected in contracts:
diff --git a/unsloth/models/loader.py b/unsloth/models/loader.py
index a66a7c6abc..cd12544ae9 100644
--- a/unsloth/models/loader.py
+++ b/unsloth/models/loader.py
@@ -1210,13 +1210,20 @@ class FastModel(FastBaseModel):
             # Granite-4 rms norms are stored as 16 bit, but we upcast
             os.environ["UNSLOTH_HIGH_PRECISION_LAYERNORM"] = "1"
             os.environ["UNSLOTH_DISABLE_STATIC_GENERATION"] = "1"
-        # Olmo 2
+        # OLMo 2
         elif "olmo2" in model_types_all and transformers_version < Version(
             "4.50.0.dev0"
         ):
             raise RuntimeError(
                 "Unsloth: OLMo-2 only works on transformers >= 4.50.0." + NIGHTLY
             )
+        # OLMo 3
+        elif "olmo3" in model_types_all and transformers_version < Version(
+            "4.57.0.dev0"
+        ):
+            raise RuntimeError(
+                "Unsloth: OLMo-3 only works on transformers >= 4.57.0." + LATEST
+            )
         elif "falcon_h1" in model_types_all:
             # Falcon must use float32 Triton ie TRITON_F32_DEFAULT = 'ieee'
             # since Mamba kernels error out on using lower precision
diff --git a/unsloth/models/mapper.py b/unsloth/models/mapper.py
index f0f430eb7e..8f2861db68 100644
--- a/unsloth/models/mapper.py
+++ b/unsloth/models/mapper.py
@@ -762,6 +762,18 @@ __INT_TO_FLOAT_MAPPER = \
         "allenai/OLMo-2-0325-32B-Instruct",
         "unsloth/OLMo-2-0325-32B-Instruct-bnb-4bit",
     ),
+    "unsloth/Olmo-3-7B-Instruct-unsloth-bnb-4bit" : (
+        "unsloth/Olmo-3-7B-Instruct",
+        "allenai/Olmo-3-7B-Instruct",
+    ),
+    "unsloth/Olmo-3-7B-Think-unsloth-bnb-4bit" : (
+        "unsloth/Olmo-3-7B-Think",
+        "allenai/Olmo-3-7B-Think",
+    ),
+    "unsloth/Olmo-3-32B-Think-unsloth-bnb-4bit" : (
+        "unsloth/Olmo-3-32B-Think",
+        "allenai/Olmo-3-32B-Think",
+    ),
     "unsloth/Mistral-Small-3.1-24B-Instruct-2503-unsloth-bnb-4bit" : (
         "unsloth/Mistral-Small-3.1-24B-Instruct-2503",
         "mistralai/Mistral-Small-3.1-24B-Instruct-2503",

From 7c5464ad71f93365773f68bb9469c9d55a859ad9 Mon Sep 17 00:00:00 2001
From: Avaya Aggarwal <119044997+OnePunchMonk@users.noreply.github.com>
Date: Wed, 15 Apr 2026 20:10:03 +0530
Subject: [PATCH 109/216] feat: Add cactus QAT scheme support (#4679)

* feat: Add cactus QAT scheme support

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

* test(qat): add tests for cactus QAT scheme and fix missing import

* Fix cactus QAT scheme: correct MappingType import, tighten PerGroup filter

- Drop the broken `from torchao.dtypes import MappingType` import. `MappingType`
  lives in `torchao.quantization` (and `torchao.quantization.quant_primitives`);
  it is not exported from `torchao.dtypes` in any supported torchao release
  (verified on 0.14, 0.16, 0.17). The previous code raised `ImportError` on
  every cactus call and was masked as a misleading 'torchao not found' error.
- Since `IntxWeightOnlyConfig` already defaults `mapping_type` to
  `MappingType.SYMMETRIC`, drop the explicit kwarg entirely and remove the
  import. Behavior is unchanged.
- Introduce a named `group_size = 32` constant (matches the int4 / fp8-int4
  pattern in the surrounding branches) and add a `% group_size == 0`
  divisibility guard to the filter. `PerGroup(32)` requires
  `in_features % 32 == 0` at `quantize_()` time, otherwise torchao raises
  `ValueError: in_features (N) % group_size (32) must be == 0`. The old
  `in_features >= 32` filter would admit non-aligned widths (e.g. 33, 48, 65,
  127) and crash `_prepare_model_for_qat` for those shapes.

* Warn when cactus QAT skips non-divisible Linear layers

Multiple reviewers flagged that the divisibility guard added in the
previous commit can silently leave Linear layers in full precision when
their in_features is not a multiple of 32. For currently supported
Unsloth models (Qwen, Llama, Gemma, Mistral, Phi) every Linear width is
already a multiple of 32/64/128 so this never triggers, but surfacing
the coverage gap is cheap and avoids users assuming 100% QAT coverage
when they bring a custom model with unusual shapes.

Emit a UserWarning listing up to the first 8 skipped layers whenever
the cactus filter excludes any Linear due to the modulo guard. This
keeps the lenient silent-skip behavior (consistent with int4 /
fp8-int4), but stops making it silent.

---------

Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com>
Co-authored-by: Daniel Han <danielhanchen@gmail.com>
---
 tests/utils/test_qat.py  | 11 ++++++---
 unsloth/models/_utils.py | 49 ++++++++++++++++++++++++++++++++++++++++
 2 files changed, 57 insertions(+), 3 deletions(-)

diff --git a/tests/utils/test_qat.py b/tests/utils/test_qat.py
index 1083712d78..08b8cd3930 100644
--- a/tests/utils/test_qat.py
+++ b/tests/utils/test_qat.py
@@ -70,6 +70,11 @@ def _test_linear_is_fake_quantized(linear: torch.nn.Linear, qat_scheme: str):
         weight_fq_class = IntxFakeQuantizer
         min_in_features = 128
         weight_only = True
+    elif qat_scheme == "cactus":
+        act_fq_class = None
+        weight_fq_class = IntxFakeQuantizer
+        min_in_features = 32
+        weight_only = True
     else:
         raise ValueError(f"Unknown qat_scheme: {qat_scheme}")
 
@@ -106,7 +111,7 @@ def _test_fake_quantizers_are_called(
     """
     Verify that the fake quantizers are actually called when the model is called.
     """
-    weight_only = qat_scheme == "int8"
+    weight_only = qat_scheme in ["int8", "cactus"]
 
     def _swap_fake_quantizers(model: torch.nn.Module):
         for name, child in model.named_children():
@@ -167,11 +172,11 @@ def _test_model_fake_quantize(qat_scheme: str, full_finetuning: bool):
 
 # TODO: there are bad interactions across tests right now, need to figure out
 # how to disable model caching before re-enabling this test
-@pytest.mark.parametrize("qat_scheme", ["fp8-int4", "fp8-fp8", "int8"])
+@pytest.mark.parametrize("qat_scheme", ["fp8-int4", "fp8-fp8", "int8", "cactus"])
 def _test_full_model_fake_quantize(qat_scheme: str):
     _test_model_fake_quantize(qat_scheme, full_finetuning = True)
 
 
-@pytest.mark.parametrize("qat_scheme", ["fp8-int4", "fp8-fp8", "int8"])
+@pytest.mark.parametrize("qat_scheme", ["fp8-int4", "fp8-fp8", "int8", "cactus"])
 def test_lora_model_fake_quantize(qat_scheme: str):
     _test_model_fake_quantize(qat_scheme, full_finetuning = False)
diff --git a/unsloth/models/_utils.py b/unsloth/models/_utils.py
index c330dcc32c..7d99ec3932 100644
--- a/unsloth/models/_utils.py
+++ b/unsloth/models/_utils.py
@@ -2727,6 +2727,55 @@ def _prepare_model_for_qat(
                 qat_scheme = qat_scheme,
                 base_config_and_filter_fns = [(base_config, filter_fn)],
             )
+        elif qat_scheme == "cactus":
+            try:
+                from torchao.quantization import IntxWeightOnlyConfig
+            except ImportError:
+                raise ImportError(TORCHAO_MSG)
+
+            # IntxWeightOnlyConfig already defaults to
+            # `mapping_type = MappingType.SYMMETRIC`, so we intentionally do not
+            # import `MappingType` here. Matches the upstream Cactus runtime
+            # int8 / per-group-32 / symmetric weight-only configuration.
+            group_size = 32
+            base_config = IntxWeightOnlyConfig(
+                weight_dtype = torch.int8,
+                granularity = PerGroup(group_size),
+            )
+            filter_fn = (
+                lambda m, _: isinstance(m, torch.nn.Linear)
+                and m.in_features >= group_size
+                and m.in_features % group_size == 0
+            )
+            # Warn if any Linear layer is skipped by the cactus filter because
+            # its in_features is not divisible by `group_size`. torchao's
+            # PerGroup(32) quantizer rejects non-divisible widths at
+            # `quantize_()` time, so the filter excludes those layers to keep
+            # the QAT prepare step from crashing. Surface that silently-skipped
+            # coverage gap to the user so they know some Linears will stay in
+            # full precision during training.
+            skipped_cactus_layers = [
+                name
+                for name, module in model.named_modules()
+                if isinstance(module, torch.nn.Linear)
+                and module.in_features >= group_size
+                and module.in_features % group_size != 0
+            ]
+            if skipped_cactus_layers:
+                preview = ", ".join(skipped_cactus_layers[:8])
+                if len(skipped_cactus_layers) > 8:
+                    preview += f", ... ({len(skipped_cactus_layers) - 8} more)"
+                warnings.warn(
+                    f"Unsloth: qat_scheme='cactus' uses PerGroup({group_size}) "
+                    "which requires in_features to be divisible by "
+                    f"{group_size}. The following Linear layers will be kept "
+                    f"in full precision during QAT: {preview}",
+                    stacklevel = 2,
+                )
+            torchao_config = TorchAOConfig(
+                qat_scheme = qat_scheme,
+                base_config_and_filter_fns = [(base_config, filter_fn)],
+            )
         else:
             raise ValueError(f"Unexpected QAT scheme {qat_scheme}")
         assert torchao_config is not None, f"TorchAOConfig was not set for {qat_scheme}"

From 800ddc95f8457a57cc5642500b193e724902109f Mon Sep 17 00:00:00 2001
From: Roland Tannous <115670425+rolandtannous@users.noreply.github.com>
Date: Wed, 15 Apr 2026 18:52:12 +0400
Subject: [PATCH 110/216] Re-apply #4939: updated models template mappers
 (#4950)
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

* Reapply "updated models template mappers. added lfm2.5vl450m to transformers 5…" (#4945)

This reverts commit 33503ea2482819ad5aaf048f2edd0a69cb540500.

* Add missing gemma-4-31B-it bnb-4bit mapper entry and LFM2.5 upstream namespace for PR #4950

- Add unsloth/gemma-4-31B-it-unsloth-bnb-4bit to __INT_TO_FLOAT_MAPPER so
  the int-to-float resolution works for this model (already listed in
  TEMPLATE_TO_MODEL_MAPPER but had no mapper entry).
- Add LiquidAI/LFM2.5-1.2B-Instruct to lfm-2.5 TEMPLATE_TO_MODEL_MAPPER
  entry so the canonical upstream namespace is mapped consistently with lfm-2.

* Add missing gemma-4-31B-it bnb-4bit Ollama mapping and lfm-2.5 chat template alias

- Add unsloth/gemma-4-31B-it-unsloth-bnb-4bit to OLLAMA_TEMPLATE_TO_MODEL_MAPPER
  so Ollama export works for this model (E2B-it and E4B-it bnb-4bit variants were
  already present, 31B-it was inconsistently omitted)
- Register CHAT_TEMPLATES["lfm-2.5"] as alias of the lfm-2 template to prevent
  KeyError when Studio resolves LFM2.5 models through MODEL_TO_TEMPLATE_MAPPER

* Add missing LFM2 bnb-4bit INT_TO_FLOAT_MAPPER entry

unsloth/LFM2-1.2B-unsloth-bnb-4bit is referenced in model_mappings.py
but had no mapper.py entry, so model resolution would fail when users
load that variant with load_in_4bit=False or when the float name is
used with load_in_4bit=True.

* Fix review findings for PR #16

1. ollama_template_mappers.py: Restore dropped Gemma-4 base model IDs
   (E2B, E4B, 31B, 26B-A4B) and add missing google/ upstream IDs to
   the gemma4 Ollama mapper for consistency with other gemma entries.

2. mapper.py: Remove self-mapping non-bnb-4bit entries from
   __INT_TO_FLOAT_MAPPER that were polluting FLOAT_TO_INT_MAPPER with
   lowercase 16-bit names, causing load_in_4bit=True to return bad
   model names. Add direct MAP_TO_UNSLOTH_16bit entries to preserve
   the google->unsloth 16-bit redirects.

3. mapper.py: Add LFM2.5 MAP_TO_UNSLOTH_16bit redirect so
   LiquidAI/LFM2.5-1.2B-Instruct resolves to its unsloth mirror.

* Add review tests for PR #4950

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

* Remove top-level test files

These test_*.py files were added at the repo root rather than under tests/.
Removing them from this PR; the production mapper changes remain.

* Add gemma-4-26B-A4B-it mapping

Adds unsloth/gemma-4-26B-A4B-it to __INT_TO_FLOAT_MAPPER as a 2-tuple so
google/gemma-4-26B-A4B-it routes to unsloth/gemma-4-26B-A4B-it across
INT_TO_FLOAT_MAPPER, FLOAT_TO_INT_MAPPER, and MAP_TO_UNSLOTH_16bit.

The 26B-A4B (MoE) model has no bnb-4bit variant, so the key uses the
plain unsloth name rather than the -unsloth-bnb-4bit suffix.

Removes the now-redundant standalone _add_with_lower call for the -it
variant; the 16bit mapping is registered via the dict loop.

* Add unsloth-bnb-4bit mappings for gemma-4 base (non-it) models

Adds E2B, E4B, 31B base unsloth-bnb-4bit entries to __INT_TO_FLOAT_MAPPER.
The 26B-A4B (MoE) base has no bnb-4bit variant on HF, so it stays on the
standalone _add_with_lower line for the 16bit-only routing.

Removes the redundant _add_with_lower lines for E2B, E4B, 31B base since
the dict loop now registers the same google->unsloth route through the
2-tuple entries, plus full FLOAT_TO_INT and INT_TO_FLOAT coverage.

---------

Co-authored-by: Daniel Han <danielhanchen@gmail.com>
Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com>
---
 .../backend/utils/datasets/model_mappings.py  | 36 +++++++++++++++++++
 studio/backend/utils/transformers_version.py  |  1 +
 unsloth/chat_templates.py                     |  2 ++
 unsloth/models/mapper.py                      | 36 +++++++++++++++++++
 unsloth/ollama_template_mappers.py            |  7 ++++
 5 files changed, 82 insertions(+)

diff --git a/studio/backend/utils/datasets/model_mappings.py b/studio/backend/utils/datasets/model_mappings.py
index 95b4791574..7fcac637c6 100644
--- a/studio/backend/utils/datasets/model_mappings.py
+++ b/studio/backend/utils/datasets/model_mappings.py
@@ -215,6 +215,21 @@ TEMPLATE_TO_MODEL_MAPPER = {
         "google/gemma-3n-E2B-it",
         "unsloth/gemma-3n-E2B-it-unsloth-bnb-4bit",
     ),
+    "gemma-4": (
+        "unsloth/gemma-4-E2B-it",
+        "google/gemma-4-E2B-it",
+        "unsloth/gemma-4-E4B-it",
+        "google/gemma-4-E4B-it",
+        "unsloth/gemma-4-E2B-it-unsloth-bnb-4bit",
+        "unsloth/gemma-4-E4B-it-unsloth-bnb-4bit",
+    ),
+    "gemma-4-thinking": (
+        "unsloth/gemma-4-26B-A4B-it",
+        "google/gemma-4-26B-A4B-it",
+        "unsloth/gemma-4-31B-it",
+        "unsloth/gemma-4-31B-it-unsloth-bnb-4bit",
+        "google/gemma-4-31B-it",
+    ),
     "qwen2.5": (
         "unsloth/Qwen2.5-0.5B-Instruct-unsloth-bnb-4bit",
         "unsloth/Qwen2.5-0.5B-Instruct",
@@ -399,6 +414,15 @@ TEMPLATE_TO_MODEL_MAPPER = {
         "THUDM/GLM-4.7-Flash",
         "unsloth/GLM-4.7-Flash-bnb-4bit",
     ),
+    "lfm-2": (
+        "unsloth/LFM2-1.2B",
+        "LiquidAI/LFM2-1.2B",
+        "unsloth/LFM2-1.2B-unsloth-bnb-4bit",
+    ),
+    "lfm-2.5": (
+        "unsloth/LFM2.5-1.2B-Instruct",
+        "LiquidAI/LFM2.5-1.2B-Instruct",
+    ),
 }
 
 MODEL_TO_TEMPLATE_MAPPER = {}
@@ -414,6 +438,14 @@ for key, values in TEMPLATE_TO_MODEL_MAPPER.items():
 
 
 TEMPLATE_TO_RESPONSES_MAPPER = {
+    "gemma-4-thinking": {
+        "instruction": "<|turn>user\n",
+        "response": "<|turn>model\n",
+    },
+    "gemma-4": {
+        "instruction": "<|turn>user\n",
+        "response": "<|turn>model\n",
+    },
     "gemma-3": {
         "instruction": "<start_of_turn>user\n",
         "response": "<start_of_turn>model\n",
@@ -514,6 +546,10 @@ TEMPLATE_TO_RESPONSES_MAPPER = {
         "instruction": "<|im_start|>user\n",
         "response": "<|im_start|>assistant\n",
     },
+    "lfm-2.5": {
+        "instruction": "<|im_start|>user\n",
+        "response": "<|im_start|>assistant\n",
+    },
     "starling": {
         "instruction": "GPT4 Correct User: ",
         "response": "GPT4 Correct Assistant: ",
diff --git a/studio/backend/utils/transformers_version.py b/studio/backend/utils/transformers_version.py
index 0c13b5455b..36c3a4c22d 100644
--- a/studio/backend/utils/transformers_version.py
+++ b/studio/backend/utils/transformers_version.py
@@ -52,6 +52,7 @@ TRANSFORMERS_5_MODEL_SUBSTRINGS: tuple[str, ...] = (
     "qwen3.5",  # Qwen3.5 family (35B-A3B, etc.)
     "qwen3-next",  # Qwen3-Next and variants
     "tiny_qwen3_moe",  # imdatta0/tiny_qwen3_moe_2.8B_0.7B
+    "lfm2.5-vl-450m",  # LiquidAI/LFM2.5-VL-450M
 )
 
 # Lowercase substrings for models that require transformers 5.5.0 (checked first).
diff --git a/unsloth/chat_templates.py b/unsloth/chat_templates.py
index 71f91cc828..326fd59289 100644
--- a/unsloth/chat_templates.py
+++ b/unsloth/chat_templates.py
@@ -1716,6 +1716,8 @@ liquid_lfm2_template = \
 liquid_lfm2_template_eos_token = "<|im_end|>"
 CHAT_TEMPLATES["lfm-2"] = (liquid_lfm2_template, liquid_lfm2_template_eos_token, False, None)
 DEFAULT_SYSTEM_MESSAGE["lfm-2"] = None # No system message in Phi-3
+CHAT_TEMPLATES["lfm-2.5"] = (liquid_lfm2_template, liquid_lfm2_template_eos_token, False, None)
+DEFAULT_SYSTEM_MESSAGE["lfm-2.5"] = None
 
 
 # =========================================== Starling-LM
diff --git a/unsloth/models/mapper.py b/unsloth/models/mapper.py
index 8f2861db68..57c1e292c3 100644
--- a/unsloth/models/mapper.py
+++ b/unsloth/models/mapper.py
@@ -22,6 +22,39 @@ __all__ = [
 
 __INT_TO_FLOAT_MAPPER = \
 {
+    "unsloth/gemma-4-E2B-it-unsloth-bnb-4bit" : (
+        "unsloth/gemma-4-E2B-it",
+        "google/gemma-4-E2B-it",
+    ),
+    "unsloth/gemma-4-E4B-it-unsloth-bnb-4bit" : (
+        "unsloth/gemma-4-E4B-it",
+        "google/gemma-4-E4B-it",
+    ),
+    "unsloth/gemma-4-31B-it-unsloth-bnb-4bit" : (
+        "unsloth/gemma-4-31B-it",
+        "google/gemma-4-31B-it",
+    ),
+    "unsloth/gemma-4-26B-A4B-it" : (
+        "unsloth/gemma-4-26B-A4B-it",
+        "google/gemma-4-26B-A4B-it",
+    ),
+    "unsloth/gemma-4-E2B-unsloth-bnb-4bit" : (
+        "unsloth/gemma-4-E2B",
+        "google/gemma-4-E2B",
+    ),
+    "unsloth/gemma-4-E4B-unsloth-bnb-4bit" : (
+        "unsloth/gemma-4-E4B",
+        "google/gemma-4-E4B",
+    ),
+    "unsloth/gemma-4-31B-unsloth-bnb-4bit" : (
+        "unsloth/gemma-4-31B",
+        "google/gemma-4-31B",
+    ),
+    "unsloth/LFM2-1.2B-unsloth-bnb-4bit" : (
+        "unsloth/LFM2-1.2B",
+        "LiquidAI/LFM2-1.2B",
+    ),
+
     "unsloth/mistral-7b-bnb-4bit" : (
         "unsloth/mistral-7b",
         "mistralai/Mistral-7B-v0.1",
@@ -1428,3 +1461,6 @@ for key, values in __INT_TO_FLOAT_MAPPER.items():
 
     for value in values:
         FLOAT_TO_INT_MAPPER[value.lower()] = lowered_key
+
+_add_with_lower(MAP_TO_UNSLOTH_16bit, "google/gemma-4-26B-A4B", "unsloth/gemma-4-26B-A4B")
+_add_with_lower(MAP_TO_UNSLOTH_16bit, "LiquidAI/LFM2.5-1.2B-Instruct", "unsloth/LFM2.5-1.2B-Instruct")
diff --git a/unsloth/ollama_template_mappers.py b/unsloth/ollama_template_mappers.py
index 728b08813a..065165d5d7 100644
--- a/unsloth/ollama_template_mappers.py
+++ b/unsloth/ollama_template_mappers.py
@@ -1978,12 +1978,19 @@ OLLAMA_TEMPLATE_TO_MODEL_MAPPER = {
     ),
     "gemma4": (
         "unsloth/gemma-4-E2B-it",
+        "unsloth/gemma-4-E2B-it-unsloth-bnb-4bit",
+        "google/gemma-4-E2B-it",
         "unsloth/gemma-4-E2B",
         "unsloth/gemma-4-E4B-it",
+        "unsloth/gemma-4-E4B-it-unsloth-bnb-4bit",
+        "google/gemma-4-E4B-it",
         "unsloth/gemma-4-E4B",
         "unsloth/gemma-4-31B-it",
+        "unsloth/gemma-4-31B-it-unsloth-bnb-4bit",
+        "google/gemma-4-31B-it",
         "unsloth/gemma-4-31B",
         "unsloth/gemma-4-26B-A4B-it",
+        "google/gemma-4-26B-A4B-it",
         "unsloth/gemma-4-26B-A4B",
     ),
     "gemma3n": (

From f0d03655e802bc4dd0a41c1ea7490fd94f6aaf3d Mon Sep 17 00:00:00 2001
From: Daniel Han <danielhanchen@gmail.com>
Date: Wed, 15 Apr 2026 08:04:33 -0700
Subject: [PATCH 111/216] Studio: add folder browser modal for Custom Folders
 (#5035)

* Studio: add folder browser modal for Custom Folders

The Custom Folders row in the model picker currently only accepts a
typed path. On a remote-served Studio (Colab, shared workstation) that
means the user has to guess or paste the exact server-side absolute
path. A native browser folder picker can't solve this: HTML
`<input type="file" webkitdirectory>` hides the absolute path for
security, and the File System Access API (Chrome/Edge only) returns
handles rather than strings, neither of which the server can act on.

This PR adds a small in-app directory browser that lists paths on the
server and hands the chosen string back to the existing
`POST /api/models/scan-folders` flow.

## Backend

* New endpoint `GET /api/models/browse-folders`:
  * `path` query param (expands `~`, accepts relative or absolute; empty
    defaults to the user's home directory).
  * `show_hidden` boolean to include dotfiles/dotdirs.
  * Returns `{current, parent, entries[], suggestions[]}`. `parent` is
    null at the filesystem root.
  * Immediate subdirectories only (no recursion); files are never
    returned.
  * `entries[].has_models` is a cheap hint: the directory looks like it
    holds models if it is named `models--*` (HF hub cache layout) or
    one of the first 64 children is a .gguf/.safetensors/config.json/
    adapter_config.json or another `models--*` subfolder.
  * Sort order: model-bearing dirs, then plain, then hidden; case-
    insensitive alphabetical within each bucket.
  * Suggestions auto-populate from HOME, the HF cache root, and any
    already-registered scan folders, deduplicated.
  * Error surface: 404 for missing path, 400 for non-directory, 403 on
    permission errors. Auth-required like the other models routes.

* New Pydantic schemas `BrowseEntry` and `BrowseFoldersResponse` in
  `studio/backend/models/models.py`.

## Frontend

* New `FolderBrowser` component
  (`studio/frontend/src/components/assistant-ui/model-selector/folder-browser.tsx`)
  using the existing `Dialog` primitive. Features:
  * Clickable breadcrumb with a `..` row for parent navigation.
  * Quick-pick chips for the server-provided suggestions.
  * `Show hidden` checkbox.
  * In-flight fetch cancellation via AbortController so rapid
    navigation doesn't flash stale results.
  * Badges model-bearing directories inline.

* `chat-api.ts` gains `browseFolders(path?, showHidden?)` and matching
  types.

* `pickers.tsx` adds a folder-magnifier icon next to the existing `Add`
  button. Opening the browser seeds it with whatever the user has
  already typed; confirming fills the text input, leaving the existing
  validation and save flow unchanged.

## What it does NOT change

* The existing text-input flow still works; the browser is additive.
* No new permissions or escalation; the endpoint reads only directories
  the server process is already allowed to read.
* No model scanning or filesystem mutation happens from the browser
  itself -- it just returns basenames for render.

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

* Studio: cap folder-browser entries and expose truncated flag

Pointing the folder browser at a huge directory (``/usr/lib``,
``/proc``, or a synthetic tree with thousands of subfolders) previously
walked the whole listing and stat-probed every child via
``_looks_like_model_dir``. That is both a DoS shape for the server
process and a large-payload surprise for the client.

Introduce a hard cap of 2000 subdirectory entries and a
``truncated: bool`` field on the response. The frontend renders a small
hint below the list when it fires, prompting the user to narrow the
path. Below-cap directories are unchanged.

Verified end-to-end against the live backend with a synthetic tree of
2050 directories: response lands at 2000 entries, ``truncated=true``,
listing finishes in sub-second time (versus tens of seconds if we were
stat-storming).

* Studio: suggest LM Studio / Ollama dirs + 2-level model probe

Three improvements to the folder-browser, driven by actually dropping
an LM Studio-style install (publisher/model/weights.gguf) into the
sandbox and walking the UX:

## 1. Quick-pick chips for other local-LLM tools

`well_known_model_dirs()` (new) returns paths commonly used by
adjacent tools. Only paths that exist are returned so the UI never
shows dead chips.

* LM Studio current + legacy roots + user-configured
  `downloadsFolder` from its `settings.json` (reuses the existing
  `lmstudio_model_dirs()` helper).
* Ollama: `$OLLAMA_MODELS` env override, then `~/.ollama/models`,
  `/usr/share/ollama/.ollama/models`, and `/var/lib/ollama/.ollama/models`
  (the systemd-service install path surfaced in the upstream "where is
  everything?" issue).
* Generic user-choice locations: `~/models`, `~/Models`.

Dedup is stable across all sources.

## 2. Two-level model-bearing probe

LM Studio and Ollama both use `root/publisher/model/weights.gguf`.
The previous `has_models` heuristic only probed one level, so the
publisher dir (whose immediate children are model dirs, not weight
files) was always marked as non-model-bearing. Pulled the direct-
signal logic into `_has_direct_model_signal` and added a grandchild
probe so the classic layout is now recognised.

Still O(PROBE^2) worst-case, still returns immediately for
`models--*` names (HF cache layout) and for any direct weight file.

## 3. model_files_here hint on response body

A leaf model dir (just GGUFs, no subdirs) previously rendered as
`(empty directory)` in the modal, confusing users into thinking the
folder wasn't scannable. Added a `model_files_here` count on the
response (capped at 200) and a small hint row in the modal: `N model
files in this folder. Click "Use this folder" to scan it.`

## Verification

Simulated an LM Studio install by downloading the real 84 MB
`unsloth/SmolLM2-135M-Instruct-Q2_K.gguf` into
`~/.lmstudio/models/unsloth/SmolLM2-135M-Instruct-GGUF/`. Confirmed
end-to-end:

* Home listing suggests `~/.lmstudio/models` as a chip.
* Browsing `~/.lmstudio/models` flags `unsloth` (publisher) as
  `has_models=true` via the 2-level probe.
* Browsing the publisher flags `SmolLM2-135M-Instruct-GGUF` (model
  dir) as `has_models=true`.
* Browsing the model dir returns empty entries but
  `model_files_here=1`, and the frontend renders a hint telling the
  user it is a valid target.

* Studio: one-click scan-folder add + prominent remove + plain search icon

Three small Custom Folders UX fixes after real-use walkthrough:

* **One-click add from the folder browser**. Confirming `Use this
  folder` now submits the path directly to
  `POST /api/models/scan-folders` instead of just populating the text
  input. `handleAddFolder` takes an optional explicit path so the
  submit lands in the same tick as `setFolderInput`, avoiding a
  state-flush race. The typed-path + `Add` button flow is unchanged.

* **Prominent remove X on scan folders**. The per-folder delete
  button was `text-muted-foreground/40` and hidden entirely on
  desktop until hovered (`md:opacity-0 md:group-hover:opacity-100`).
  Dropped the hover-only cloak, bumped color to `text-foreground/70`,
  added a red hover/focus background, and sized the icon up from
  `size-2.5` to `size-3`. Always visible on every viewport.

* **Plain search icon for the Browse button**. `FolderSearchIcon`
  replaced with `Search01Icon` so it reads as a simple "find a
  folder" action alongside the existing `Add01Icon`.

* Studio: align Custom Folders + and X buttons on the same right edge

The Custom Folders header used `px-2.5` with a `p-0.5` icon button,
while each folder row used `px-3` with a `p-1` button. That put the
X icon 4px further from the right edge than the +. Normalised both
rows to `px-2.5` with `p-1` so the two icons share a column.

* Studio: empty-state button opens the folder browser directly

The first-run empty state for Custom Folders was a text link reading
"+ Add a folder to scan for local models" whose click toggled the
text input. That's the wrong default: a user hitting the empty state
usually doesn't know what absolute path to type, which is exactly
what the folder browser is for.

* Reword to "Browse for a models folder" with a search-icon
  affordance so the label matches what the click does.
* Click opens the folder browser modal directly. The typed-path +
  Add button flow is still available via the + icon in the
  section header, so users who know their path keep that option.
* Slightly bump the muted foreground opacity (70 -> hover:foreground)
  so the button reads as a primary empty-state action rather than a
  throwaway hint.

* Studio: Custom Folders header gets a dedicated search + add button pair

The Custom Folders section header had a single toggle button that
flipped between + and X. That put the folder-browser entry point
behind the separate empty-state link. Cleaner layout: two buttons in
the header, search first, then add.

* Search icon (left) opens the folder browser modal directly.
* Plus icon (right) toggles the text-path input (unchanged).
* The first-run empty-state link is removed -- the two header icons
  cover both flows on every state.

Both buttons share the same padding / icon size so they line up with
each other and with the per-folder remove X.

* Studio: sandbox folder browser + bound caps + UX recoveries

PR review fixes for the Custom Folders folder browser. Closes the
high-severity CodeQL path-traversal alert and addresses the codex /
gemini P2 findings.

Backend (studio/backend/routes/models.py):

* New _build_browse_allowlist + _is_path_inside_allowlist sandbox.
  browse_folders now refuses any target that doesn't resolve under
  HOME, HF cache, Studio dirs, registered scan folders, or the
  well-known third-party model dirs. realpath() is used so symlink
  traversal cannot escape the sandbox. Also gates the parent crumb
  so the up-row hides instead of 403'ing.
* _BROWSE_ENTRY_CAP now bounds *visited* iterdir entries, not
  *appended* entries. Dirs full of files (or hidden subdirs when
  show_hidden is False) used to defeat the cap.
* _count_model_files gets the same visited-count fix.
* PermissionError no longer swallowed silently inside the
  enumeration / counter loops -- now logged at debug.

Frontend (folder-browser.tsx, pickers.tsx, chat-api.ts):

* splitBreadcrumb stops mangling literal backslashes inside POSIX
  filenames; only Windows-style absolute paths trigger separator
  normalization. The Windows drive crumb value is now C:/ (drive
  root) instead of C: (drive-relative CWD-on-C).
* browseFolders accepts and forwards an AbortSignal so cancelled
  navigations actually cancel the in-flight backend enumeration.
* On initial-path fetch error, FolderBrowser now falls back to HOME
  instead of leaving the modal as an empty dead end.
* When the auto-add path (one-click "Use this folder") fails, the
  failure now surfaces via toast in addition to the inline
  paragraph (which is hidden when the typed-input panel is closed).

* Studio: rebuild browse target from trusted root for CodeQL clean dataflow

CodeQL's py/path-injection rule kept flagging the post-validation
filesystem operations because the sandbox check lived inside a
helper function (_is_path_inside_allowlist) and CodeQL only does
intra-procedural taint tracking by default. The user-derived
``target`` was still flowing into ``target.exists`` /
``target.is_dir`` / ``target.iterdir``.

The fix: after resolving the user-supplied ``candidate_path``,
locate the matching trusted root from the allowlist and rebuild
``target`` by appending each individually-validated segment to
that trusted root. Each segment is rejected if it isn't a single
safe path component (no separators, no ``..``, no empty/dot).
The downstream filesystem ops now operate on a Path constructed
entirely from ``allowed_roots`` (trusted) plus those validated
segments, so CodeQL's dataflow no longer sees a tainted source.

Behavior is unchanged for all valid inputs -- only the
construction of ``target`` is restructured. Live + unit tests
all pass (58 selected, 7 deselected for Playwright env).

* Studio: walk browse paths from trusted roots for CodeQL

---------

Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com>
Co-authored-by: Ubuntu <ubuntu@h100-8-cheapest.us-east5-a.c.unsloth.internal>
---
 studio/backend/models/models.py               |  65 +++
 studio/backend/routes/models.py               | 525 ++++++++++++++++++
 .../tests/test_browse_folders_route.py        |  86 +++
 studio/backend/utils/paths/__init__.py        |   2 +
 studio/backend/utils/paths/storage_roots.py   |  45 ++
 .../model-selector/folder-browser.tsx         | 328 +++++++++++
 .../assistant-ui/model-selector/pickers.tsx   | 105 ++--
 .../src/features/chat/api/chat-api.ts         |  36 ++
 8 files changed, 1161 insertions(+), 31 deletions(-)
 create mode 100644 studio/backend/tests/test_browse_folders_route.py
 create mode 100644 studio/frontend/src/components/assistant-ui/model-selector/folder-browser.tsx

diff --git a/studio/backend/models/models.py b/studio/backend/models/models.py
index f67014a17b..46ca4e3784 100644
--- a/studio/backend/models/models.py
+++ b/studio/backend/models/models.py
@@ -213,3 +213,68 @@ class ScanFolderInfo(BaseModel):
     id: int = Field(..., description = "Database row ID")
     path: str = Field(..., description = "Normalized absolute path")
     created_at: str = Field(..., description = "ISO 8601 creation timestamp")
+
+
+class BrowseEntry(BaseModel):
+    """A directory entry surfaced by the folder browser."""
+
+    name: str = Field(..., description = "Entry name (basename, not full path)")
+    has_models: bool = Field(
+        False,
+        description = (
+            "Hint that the directory likely contains models "
+            "(*.gguf, *.safetensors, config.json, or HF-style "
+            "`models--*` subfolders). Used by the UI to highlight "
+            "promising candidates; the scanner itself is authoritative."
+        ),
+    )
+    hidden: bool = Field(
+        False,
+        description = "Name starts with a dot (e.g. `.cache`)",
+    )
+
+
+class BrowseFoldersResponse(BaseModel):
+    """Response schema for the folder browser endpoint."""
+
+    current: str = Field(..., description = "Absolute path of the directory just listed")
+    parent: Optional[str] = Field(
+        None,
+        description = (
+            "Parent directory of `current`, or null if `current` is the "
+            "filesystem root. The frontend uses this to render an `Up` row."
+        ),
+    )
+    entries: List[BrowseEntry] = Field(
+        default_factory = list,
+        description = (
+            "Subdirectories of `current`. Sorted with model-bearing "
+            "directories first, then alphabetically case-insensitive; "
+            "hidden entries come last within each group."
+        ),
+    )
+    suggestions: List[str] = Field(
+        default_factory = list,
+        description = (
+            "Handy starting points (home, HF cache, already-registered "
+            "scan folders). Rendered as quick-pick chips above the list."
+        ),
+    )
+    truncated: bool = Field(
+        False,
+        description = (
+            "True when the listing was capped because the directory had "
+            "more subfolders than the server is willing to enumerate in "
+            "one request. The UI should show a hint telling the user to "
+            "narrow their path."
+        ),
+    )
+    model_files_here: int = Field(
+        0,
+        description = (
+            "Count of GGUF/safetensors files immediately inside "
+            "``current``. Used by the UI to surface a hint on leaf "
+            "model directories (which otherwise look `empty` because "
+            "they contain only files, no subdirectories)."
+        ),
+    )
diff --git a/studio/backend/routes/models.py b/studio/backend/routes/models.py
index 4ce2d0787c..9e7168eed6 100644
--- a/studio/backend/routes/models.py
+++ b/studio/backend/routes/models.py
@@ -101,6 +101,8 @@ from models import (
     ModelListResponse,
 )
 from models.models import (
+    BrowseEntry,
+    BrowseFoldersResponse,
     GgufVariantDetail,
     GgufVariantsResponse,
     ModelType,
@@ -573,6 +575,529 @@ async def remove_scan_folder_endpoint(
     return {"ok": True}
 
 
+# Heuristic ceiling on how many children to stat when checking whether a
+# directory "looks like" it contains models. Keeps the browser snappy
+# even when a directory has thousands of unrelated entries.
+_BROWSE_MODEL_HINT_PROBE = 64
+# Hard cap on how many subdirectory entries we send back. Pointing the
+# browser at something like ``/usr/lib`` or ``/proc`` must not stat-storm
+# the process or send tens of thousands of rows to the client.
+_BROWSE_ENTRY_CAP = 2000
+
+
+def _count_model_files(directory: Path, cap: int = 200) -> int:
+    """Count GGUF/safetensors files immediately inside *directory*.
+    Used to surface a count-hint on the response so the UI can tell
+    users that a leaf directory (no subdirs, only weights) is a valid
+    "Use this folder" target.
+
+    Bounded by *visited entries*, not by *match count*: in directories
+    with many non-model files (or many subdirectories) the scan still
+    stops after ``cap`` entries so a UI hint never costs more than a
+    bounded directory walk.
+    """
+    n = 0
+    visited = 0
+    try:
+        for f in directory.iterdir():
+            visited += 1
+            if visited > cap:
+                break
+            try:
+                if f.is_file():
+                    low = f.name.lower()
+                    if low.endswith((".gguf", ".safetensors")):
+                        n += 1
+            except OSError:
+                continue
+    except PermissionError as e:
+        logger.debug("browse-folders: permission denied counting %s: %s", directory, e)
+        return 0
+    except OSError as e:
+        logger.debug("browse-folders: OS error counting %s: %s", directory, e)
+        return 0
+    return n
+
+
+def _has_direct_model_signal(directory: Path) -> bool:
+    """Return True if *directory* has an immediate child that signals
+    it holds a model: a GGUF/safetensors/config.json file, or a
+    `models--*` subdir (HF hub cache). Bounded by
+    ``_BROWSE_MODEL_HINT_PROBE`` to stay fast."""
+    try:
+        it = directory.iterdir()
+    except OSError:
+        return False
+    try:
+        for i, child in enumerate(it):
+            if i >= _BROWSE_MODEL_HINT_PROBE:
+                break
+            try:
+                name = child.name
+                if child.is_file():
+                    low = name.lower()
+                    if low.endswith((".gguf", ".safetensors")):
+                        return True
+                    if low in ("config.json", "adapter_config.json"):
+                        return True
+                elif child.is_dir() and name.startswith("models--"):
+                    return True
+            except OSError:
+                continue
+    except OSError:
+        return False
+    return False
+
+
+def _looks_like_model_dir(directory: Path) -> bool:
+    """Bounded heuristic used by the folder browser to flag directories
+    worth exploring. False negatives are fine; the real scanner is
+    authoritative.
+
+    Three signals, cheapest first:
+
+    1. Directory name itself: ``models--*`` is the HuggingFace hub cache
+       layout (``blobs``/``refs``/``snapshots`` children wouldn't match
+       the file-level probes below).
+    2. An immediate child is a weight file or config (handled by
+       :func:`_has_direct_model_signal`).
+    3. A grandchild has a direct signal -- this catches the
+       ``publisher/model/weights.gguf`` layout used by LM Studio and
+       Ollama. We probe at most the first
+       ``_BROWSE_MODEL_HINT_PROBE`` child directories, each of which is
+       checked with a bounded :func:`_has_direct_model_signal` call,
+       so the total cost stays O(PROBE^2) worst-case.
+    """
+    if directory.name.startswith("models--"):
+        return True
+    if _has_direct_model_signal(directory):
+        return True
+    # Grandchild probe: LM Studio / Ollama publisher/model layout.
+    try:
+        it = directory.iterdir()
+    except OSError:
+        return False
+    try:
+        for i, child in enumerate(it):
+            if i >= _BROWSE_MODEL_HINT_PROBE:
+                break
+            try:
+                if not child.is_dir():
+                    continue
+            except OSError:
+                continue
+            # Fast name check first
+            if child.name.startswith("models--"):
+                return True
+            if _has_direct_model_signal(child):
+                return True
+    except OSError:
+        return False
+    return False
+
+
+def _build_browse_allowlist() -> list[Path]:
+    """Return the list of root directories the folder browser is allowed
+    to walk. The same list is used to seed the sidebar suggestion chips,
+    so chip targets are always reachable.
+
+    Roots include the current user's HOME, the resolved HF cache dirs,
+    Studio's own outputs/exports/studio root, registered scan folders,
+    and well-known third-party local-LLM dirs (LM Studio, Ollama,
+    `~/models`). Each is added only if it currently resolves to a real
+    directory, so we never produce a "dead" sandbox boundary the user
+    can't navigate into.
+    """
+    from utils.paths import (
+        hf_default_cache_dir,
+        legacy_hf_cache_dir,
+        well_known_model_dirs,
+    )
+    from storage.studio_db import list_scan_folders
+
+    candidates: list[Path] = []
+
+    def _add(p: Optional[Path]) -> None:
+        if p is None:
+            return
+        try:
+            resolved = p.resolve()
+        except OSError:
+            return
+        if resolved.is_dir():
+            candidates.append(resolved)
+
+    _add(Path.home())
+    _add(_resolve_hf_cache_dir())
+    try:
+        _add(hf_default_cache_dir())
+    except Exception:  # noqa: BLE001 -- best-effort
+        pass
+    try:
+        _add(legacy_hf_cache_dir())
+    except Exception:  # noqa: BLE001 -- best-effort
+        pass
+    try:
+        from utils.paths import (
+            exports_root,
+            outputs_root,
+            studio_root,
+        )
+
+        _add(studio_root())
+        _add(outputs_root())
+        _add(exports_root())
+    except Exception as exc:  # noqa: BLE001 -- best-effort
+        logger.debug("browse-folders: studio roots unavailable: %s", exc)
+    try:
+        for folder in list_scan_folders():
+            p = folder.get("path")
+            if p:
+                _add(Path(p))
+    except Exception as exc:  # noqa: BLE001 -- best-effort
+        logger.debug("browse-folders: could not load scan folders: %s", exc)
+    try:
+        for p in well_known_model_dirs():
+            _add(p)
+    except Exception as exc:  # noqa: BLE001 -- best-effort
+        logger.debug("browse-folders: well-known dirs unavailable: %s", exc)
+
+    # Dedupe while preserving order.
+    seen: set[str] = set()
+    deduped: list[Path] = []
+    for p in candidates:
+        key = str(p)
+        if key in seen:
+            continue
+        seen.add(key)
+        deduped.append(p)
+    return deduped
+
+
+def _is_path_inside_allowlist(target: Path, allowed_roots: list[Path]) -> bool:
+    """Return True if *target* equals or is a descendant of any allowed
+    root. The comparison uses ``os.path.realpath`` so symlinks cannot be
+    used to escape the sandbox.
+    """
+    try:
+        target_real = os.path.realpath(str(target))
+    except OSError:
+        return False
+    for root in allowed_roots:
+        try:
+            root_real = os.path.realpath(str(root))
+        except OSError:
+            continue
+        if target_real == root_real or target_real.startswith(root_real + os.sep):
+            return True
+    return False
+
+
+def _normalize_browse_request_path(path: Optional[str]) -> str:
+    """Normalize the browse request path lexically, without touching the FS."""
+    if path is None or not path.strip():
+        return os.path.normpath(str(Path.home()))
+
+    expanded = os.path.expanduser(path.strip())
+    if not os.path.isabs(expanded):
+        expanded = os.path.join(str(Path.cwd()), expanded)
+    return os.path.normpath(expanded)
+
+
+def _browse_relative_parts(requested_path: str, root: Path) -> Optional[list[str]]:
+    """Return validated relative path components under ``root``."""
+    root_text = os.path.normpath(str(root))
+    try:
+        rel_text = os.path.relpath(requested_path, root_text)
+    except ValueError:
+        return None
+
+    if rel_text == ".":
+        return []
+    if rel_text == ".." or rel_text.startswith(f"..{os.sep}"):
+        return None
+
+    parts = [part for part in rel_text.split(os.sep) if part not in ("", ".")]
+    altsep = os.altsep
+    for part in parts:
+        if part == ".." or os.sep in part or (altsep and altsep in part):
+            return None
+    return parts
+
+
+def _match_browse_child(current: Path, name: str) -> Optional[Path]:
+    """Return the immediate child named ``name`` under ``current``."""
+    try:
+        for child in current.iterdir():
+            if child.name == name:
+                return child
+    except PermissionError:
+        raise HTTPException(
+            status_code = 403,
+            detail = f"Permission denied reading {current}",
+        ) from None
+    except OSError as exc:
+        raise HTTPException(
+            status_code = 500,
+            detail = f"Could not read {current}: {exc}",
+        ) from exc
+    return None
+
+
+def _resolve_browse_target(path: Optional[str], allowed_roots: list[Path]) -> Path:
+    """Resolve a requested browse path by walking from trusted allowlist roots."""
+    requested_path = _normalize_browse_request_path(path)
+    resolved_roots: list[Path] = []
+    seen_roots: set[str] = set()
+    for root in sorted(allowed_roots, key = lambda p: len(str(p)), reverse = True):
+        try:
+            resolved = root.resolve()
+        except OSError:
+            continue
+        key = str(resolved)
+        if key in seen_roots:
+            continue
+        seen_roots.add(key)
+        resolved_roots.append(resolved)
+
+    for root in resolved_roots:
+        parts = _browse_relative_parts(requested_path, root)
+        if parts is None:
+            continue
+
+        current = root
+        for part in parts:
+            child = _match_browse_child(current, part)
+            if child is None:
+                raise HTTPException(
+                    status_code = 404,
+                    detail = f"Path does not exist: {requested_path}",
+                )
+            try:
+                resolved_child = child.resolve()
+            except OSError as exc:
+                raise HTTPException(
+                    status_code = 400,
+                    detail = f"Invalid path: {exc}",
+                ) from exc
+            if not _is_path_inside_allowlist(resolved_child, resolved_roots):
+                raise HTTPException(
+                    status_code = 403,
+                    detail = (
+                        "Path is not in the browseable allowlist. Register it via "
+                        "POST /api/models/scan-folders first, or pick a directory "
+                        "under your home folder."
+                    ),
+                )
+            current = resolved_child
+
+        if not current.is_dir():
+            raise HTTPException(
+                status_code = 400,
+                detail = f"Not a directory: {current}",
+            )
+        return current
+
+    raise HTTPException(
+        status_code = 403,
+        detail = (
+            "Path is not in the browseable allowlist. Register it via "
+            "POST /api/models/scan-folders first, or pick a directory "
+            "under your home folder."
+        ),
+    )
+
+
+@router.get("/browse-folders", response_model = BrowseFoldersResponse)
+async def browse_folders(
+    path: Optional[str] = Query(
+        None,
+        description = (
+            "Directory to list. If omitted, defaults to the current user's "
+            "home directory. Tilde (`~`) and relative paths are expanded. "
+            "Must resolve inside the allowlist of browseable roots (HOME, "
+            "HF cache, Studio dirs, registered scan folders, well-known "
+            "model dirs)."
+        ),
+    ),
+    show_hidden: bool = Query(
+        False,
+        description = "Include entries whose name starts with a dot",
+    ),
+    current_subject: str = Depends(get_current_subject),
+):
+    """
+    List immediate subdirectories of *path* for the Custom Folders picker.
+
+    The frontend uses this to render a modal folder browser without needing
+    a native OS dialog (Studio is served over HTTP, so the browser can't
+    reveal absolute paths on the host). The endpoint is read-only and does
+    not create, move, or delete anything. It simply enumerates visible
+    subdirectories so the user can click their way to a folder and hand
+    the resulting string back to POST `/api/models/scan-folders`.
+
+    Sandbox: requests are bounded to the allowlist returned by
+    :func:`_build_browse_allowlist` (HOME, HF cache, Studio dirs,
+    registered scan folders, well-known model dirs). Paths outside the
+    allowlist return 403 so users cannot probe ``/etc``, ``/proc``,
+    ``/root`` (when not HOME), or other sensitive system locations
+    even if the server process can read them. Symlinks are resolved
+    via ``os.path.realpath`` before the check, so symlink traversal
+    cannot escape the sandbox either.
+
+    Sorting: directories that look like they hold models come first, then
+    plain directories, then hidden entries (if `show_hidden=true`).
+    """
+    from utils.paths import hf_default_cache_dir, well_known_model_dirs
+    from storage.studio_db import list_scan_folders
+
+    # Build the allowlist once -- both the sandbox check below and the
+    # suggestion chips use the same set, so chips are always navigable.
+    allowed_roots = _build_browse_allowlist()
+
+    try:
+        target = _resolve_browse_target(path, allowed_roots)
+    except HTTPException:
+        requested_path = _normalize_browse_request_path(path)
+        if path is not None and path.strip():
+            logger.warning(
+                "browse-folders: rejected path %r (normalized=%s)",
+                path,
+                requested_path,
+            )
+        raise
+
+    # Enumerate immediate subdirectories with a bounded cap so a stray
+    # query against ``/usr/lib`` or ``/proc`` can't stat-storm the process.
+    entries: list[BrowseEntry] = []
+    truncated = False
+    visited = 0
+    try:
+        it = target.iterdir()
+    except PermissionError:
+        raise HTTPException(
+            status_code = 403,
+            detail = f"Permission denied reading {target}",
+        )
+    except OSError as exc:
+        raise HTTPException(
+            status_code = 500,
+            detail = f"Could not read {target}: {exc}",
+        )
+
+    try:
+        for child in it:
+            # Bound by *visited entries*, not by *appended entries*: in
+            # directories full of files (or hidden subdirs when
+            # ``show_hidden=False``) the cap on ``len(entries)`` would
+            # never trigger and we'd still stat every child. Counting
+            # visits keeps the worst-case work to ``_BROWSE_ENTRY_CAP``
+            # iterdir/is_dir calls regardless of how many of them
+            # survive the filters below.
+            visited += 1
+            if visited > _BROWSE_ENTRY_CAP:
+                truncated = True
+                break
+            try:
+                if not child.is_dir():
+                    continue
+            except OSError:
+                continue
+            name = child.name
+            is_hidden = name.startswith(".")
+            if is_hidden and not show_hidden:
+                continue
+            entries.append(
+                BrowseEntry(
+                    name = name,
+                    has_models = _looks_like_model_dir(child),
+                    hidden = is_hidden,
+                )
+            )
+    except PermissionError as exc:
+        logger.debug(
+            "browse-folders: permission denied during enumeration of %s: %s",
+            target,
+            exc,
+        )
+    except OSError as exc:
+        # Rare: iterdir succeeded but reading a specific entry failed.
+        logger.warning("browse-folders: partial enumeration of %s: %s", target, exc)
+
+    # Model-bearing dirs first, then plain, then hidden; case-insensitive
+    # alphabetical within each bucket.
+    def _sort_key(e: BrowseEntry) -> tuple[int, str]:
+        bucket = 0 if e.has_models else (2 if e.hidden else 1)
+        return (bucket, e.name.lower())
+
+    entries.sort(key = _sort_key)
+
+    # Parent is None at the filesystem root (`p.parent == p`) AND when
+    # the parent would step outside the sandbox -- otherwise the up-row
+    # would 403 on click. Users can still hop to other allowed roots
+    # via the suggestion chips below.
+    parent: Optional[str]
+    if target.parent == target or not _is_path_inside_allowlist(
+        target.parent, allowed_roots
+    ):
+        parent = None
+    else:
+        parent = str(target.parent)
+
+    # Handy starting points for the quick-pick chips.
+    suggestions: list[str] = []
+    seen_sug: set[str] = set()
+
+    def _add_sug(p: Optional[Path]) -> None:
+        if p is None:
+            return
+        try:
+            resolved = str(p.resolve())
+        except OSError:
+            return
+        if resolved in seen_sug:
+            return
+        if Path(resolved).is_dir():
+            seen_sug.add(resolved)
+            suggestions.append(resolved)
+
+    # Home always comes first -- it's the safe fallback when everything
+    # else is cold.
+    _add_sug(Path.home())
+    # The HF cache root the process is actually using.
+    try:
+        _add_sug(hf_default_cache_dir())
+    except Exception:
+        pass
+    # Already-registered scan folders (what the user has curated).
+    try:
+        for folder in list_scan_folders():
+            _add_sug(Path(folder.get("path", "")))
+    except Exception as exc:
+        logger.debug("browse-folders: could not load scan folders: %s", exc)
+    # Directories commonly used by other local-LLM tools: LM Studio
+    # (`~/.lmstudio/models` + legacy `~/.cache/lm-studio/models` +
+    # user-configured downloadsFolder from LM Studio's settings.json),
+    # Ollama (`~/.ollama/models` + common system paths + OLLAMA_MODELS
+    # env var), and generic user-choice spots (`~/models`, `~/Models`).
+    # Each helper only returns paths that currently exist so we never
+    # show dead chips.
+    try:
+        for p in well_known_model_dirs():
+            _add_sug(p)
+    except Exception as exc:
+        logger.debug("browse-folders: could not load well-known dirs: %s", exc)
+
+    return BrowseFoldersResponse(
+        current = str(target),
+        parent = parent,
+        entries = entries,
+        suggestions = suggestions,
+        truncated = truncated,
+        model_files_here = _count_model_files(target),
+    )
+
+
 @router.get("/list")
 async def list_models(
     current_subject: str = Depends(get_current_subject),
diff --git a/studio/backend/tests/test_browse_folders_route.py b/studio/backend/tests/test_browse_folders_route.py
new file mode 100644
index 0000000000..19a83987d3
--- /dev/null
+++ b/studio/backend/tests/test_browse_folders_route.py
@@ -0,0 +1,86 @@
+# SPDX-License-Identifier: AGPL-3.0-only
+# Copyright 2026-present the Unsloth AI Inc. team. All rights reserved. See /studio/LICENSE.AGPL-3.0
+
+import os
+import sys
+import types
+from pathlib import Path
+
+import pytest
+from fastapi import HTTPException
+
+# Keep this test runnable in lightweight environments where optional logging
+# deps are not installed.
+if "structlog" not in sys.modules:
+
+    class _DummyLogger:
+        def __getattr__(self, _name):
+            return lambda *args, **kwargs: None
+
+    sys.modules["structlog"] = types.SimpleNamespace(
+        BoundLogger = _DummyLogger,
+        get_logger = lambda *args, **kwargs: _DummyLogger(),
+    )
+
+import routes.models as models_route
+
+
+def test_resolve_browse_target_returns_allowed_directory(tmp_path):
+    allowed = tmp_path / "allowed"
+    target = allowed / "models" / "nested"
+    target.mkdir(parents = True)
+
+    resolved = models_route._resolve_browse_target(str(target), [allowed])
+
+    assert resolved == target.resolve()
+
+
+def test_resolve_browse_target_rejects_outside_allowlist(tmp_path):
+    allowed = tmp_path / "allowed"
+    disallowed = tmp_path / "disallowed"
+    allowed.mkdir()
+    disallowed.mkdir()
+
+    with pytest.raises(HTTPException) as exc_info:
+        models_route._resolve_browse_target(str(disallowed), [allowed])
+
+    assert exc_info.value.status_code == 403
+
+
+def test_resolve_browse_target_rejects_file_path(tmp_path):
+    allowed = tmp_path / "allowed"
+    allowed.mkdir()
+    model_file = allowed / "model.gguf"
+    model_file.write_text("gguf")
+
+    with pytest.raises(HTTPException) as exc_info:
+        models_route._resolve_browse_target(str(model_file), [allowed])
+
+    assert exc_info.value.status_code == 400
+
+
+def test_resolve_browse_target_allows_symlink_into_other_allowed_root(tmp_path):
+    home_root = tmp_path / "home"
+    scan_root = tmp_path / "scan"
+    target = scan_root / "nested"
+    home_root.mkdir()
+    target.mkdir(parents = True)
+    (home_root / "scan-link").symlink_to(scan_root, target_is_directory = True)
+
+    resolved = models_route._resolve_browse_target(
+        str(home_root / "scan-link" / "nested"),
+        [home_root, scan_root],
+    )
+
+    assert resolved == target.resolve()
+
+
+@pytest.mark.skipif(os.altsep is not None, reason = "POSIX-only path semantics")
+def test_resolve_browse_target_allows_backslash_in_posix_segment(tmp_path):
+    allowed = tmp_path / "allowed"
+    target = allowed / r"dir\name"
+    target.mkdir(parents = True)
+
+    resolved = models_route._resolve_browse_target(str(target), [allowed])
+
+    assert resolved == target.resolve()
diff --git a/studio/backend/utils/paths/__init__.py b/studio/backend/utils/paths/__init__.py
index 11709ae56e..92191dccdd 100644
--- a/studio/backend/utils/paths/__init__.py
+++ b/studio/backend/utils/paths/__init__.py
@@ -34,6 +34,7 @@ from .storage_roots import (
     legacy_hf_cache_dir,
     hf_default_cache_dir,
     lmstudio_model_dirs,
+    well_known_model_dirs,
     ensure_dir,
     ensure_studio_directories,
     resolve_under_root,
@@ -70,6 +71,7 @@ __all__ = [
     "legacy_hf_cache_dir",
     "hf_default_cache_dir",
     "lmstudio_model_dirs",
+    "well_known_model_dirs",
     "ensure_dir",
     "ensure_studio_directories",
     "resolve_under_root",
diff --git a/studio/backend/utils/paths/storage_roots.py b/studio/backend/utils/paths/storage_roots.py
index 4841c5d0a3..b52609b06b 100644
--- a/studio/backend/utils/paths/storage_roots.py
+++ b/studio/backend/utils/paths/storage_roots.py
@@ -130,6 +130,51 @@ def lmstudio_model_dirs() -> list[Path]:
     return dirs
 
 
+def well_known_model_dirs() -> list[Path]:
+    """Return directories commonly used by other local LLM tools.
+
+    Used by the folder browser to offer quick-pick chips. Returns only
+    paths that exist on disk, so the UI never shows dead chips. Order
+    reflects a rough "likelihood the user has models here" -- LM Studio
+    and Ollama first, then the generic fallbacks.
+    """
+    candidates: list[Path] = []
+
+    # LM Studio (reuses the logic above, including settings.json override)
+    candidates.extend(lmstudio_model_dirs())
+
+    # Ollama -- both the user-level and common system-wide install paths
+    # (https://github.com/ollama/ollama/issues/733).
+    ollama_env = os.environ.get("OLLAMA_MODELS")
+    if ollama_env:
+        candidates.append(Path(ollama_env).expanduser())
+    candidates.append(Path.home() / ".ollama" / "models")
+    candidates.append(Path("/usr/share/ollama/.ollama/models"))
+    candidates.append(Path("/var/lib/ollama/.ollama/models"))
+
+    # HF hub cache root (separate from the explicit HF cache chip)
+    candidates.append(Path.home() / ".cache" / "huggingface" / "hub")
+
+    # Generic "my models" spots users tend to drop things into
+    for name in ("models", "Models"):
+        candidates.append(Path.home() / name)
+
+    # Deduplicate while preserving order; keep only extant dirs
+    out: list[Path] = []
+    seen: set[str] = set()
+    for p in candidates:
+        try:
+            resolved = str(p.resolve())
+        except OSError:
+            continue
+        if resolved in seen:
+            continue
+        if Path(resolved).is_dir():
+            seen.add(resolved)
+            out.append(Path(resolved))
+    return out
+
+
 def _setup_cache_env() -> None:
     """Set cache environment variables for HuggingFace, uv, and vLLM.
 
diff --git a/studio/frontend/src/components/assistant-ui/model-selector/folder-browser.tsx b/studio/frontend/src/components/assistant-ui/model-selector/folder-browser.tsx
new file mode 100644
index 0000000000..42bd1716a1
--- /dev/null
+++ b/studio/frontend/src/components/assistant-ui/model-selector/folder-browser.tsx
@@ -0,0 +1,328 @@
+// SPDX-License-Identifier: AGPL-3.0-only
+// Copyright 2026-present the Unsloth AI Inc. team. All rights reserved. See /studio/LICENSE.AGPL-3.0
+
+"use client";
+
+import {
+  Dialog,
+  DialogClose,
+  DialogContent,
+  DialogFooter,
+  DialogHeader,
+  DialogTitle,
+} from "@/components/ui/dialog";
+import { Spinner } from "@/components/ui/spinner";
+import {
+  type BrowseFoldersResponse,
+  browseFolders,
+} from "@/features/chat/api/chat-api";
+import { cn } from "@/lib/utils";
+import { ArrowUp02Icon, Folder02Icon } from "@hugeicons/core-free-icons";
+import { HugeiconsIcon } from "@hugeicons/react";
+import { useCallback, useEffect, useMemo, useRef, useState } from "react";
+
+export interface FolderBrowserProps {
+  open: boolean;
+  onOpenChange: (open: boolean) => void;
+  /** Called with the absolute path the user confirmed. */
+  onSelect: (path: string) => void;
+  /** Optional initial directory. Defaults to the user's home on the server. */
+  initialPath?: string;
+}
+
+function splitBreadcrumb(path: string): { label: string; value: string }[] {
+  if (!path) return [];
+  // Distinguish path styles BEFORE normalizing separators. On POSIX
+  // backslashes are valid filename characters, so we cannot blindly
+  // rewrite ``\`` -> ``/`` -- doing so would mangle directory names
+  // like ``my\backup`` into ``my/backup`` and produce breadcrumb
+  // values that 404 on the server. Only Windows-style absolute paths
+  // (drive letter, or UNC ``\\server\share``) get the conversion.
+  const isWindowsDrive = /^[A-Za-z]:[\\/]/.test(path) || /^[A-Za-z]:$/.test(path);
+  const isUnc = /^\\\\/.test(path);
+  const isWindows = isWindowsDrive || isUnc;
+  const normalized = isWindows ? path.replace(/\\/g, "/") : path;
+  const segments = normalized.split("/");
+  const parts: { label: string; value: string }[] = [];
+
+  // POSIX absolute path: leading empty segment from split("/")
+  if (segments[0] === "") {
+    parts.push({ label: "/", value: "/" });
+    let cur = "";
+    for (const seg of segments.slice(1)) {
+      if (!seg) continue;
+      cur = `${cur}/${seg}`;
+      parts.push({ label: seg, value: cur });
+    }
+    return parts;
+  }
+
+  // Windows-ish drive path (C:, D:): first segment is the drive. Use
+  // ``C:/`` (drive-absolute) as the crumb value so clicking the drive
+  // root navigates to the root of the drive rather than the
+  // drive-relative current working directory on that drive (``C:``
+  // alone resolves to ``CWD-on-C``, not ``C:\``).
+  if (/^[A-Za-z]:$/.test(segments[0])) {
+    const driveRoot = `${segments[0]}/`;
+    let cur = driveRoot;
+    parts.push({ label: segments[0], value: driveRoot });
+    for (const seg of segments.slice(1)) {
+      if (!seg) continue;
+      cur = cur.endsWith("/") ? `${cur}${seg}` : `${cur}/${seg}`;
+      parts.push({ label: seg, value: cur });
+    }
+    return parts;
+  }
+
+  // Fallback: relative / UNC-ish. Render as-is as a single crumb.
+  return [{ label: path, value: path }];
+}
+
+export function FolderBrowser({
+  open,
+  onOpenChange,
+  onSelect,
+  initialPath,
+}: FolderBrowserProps) {
+  const [data, setData] = useState<BrowseFoldersResponse | null>(null);
+  const [path, setPath] = useState<string | undefined>(initialPath);
+  const [showHidden, setShowHidden] = useState(false);
+  const [loading, setLoading] = useState(false);
+  const [error, setError] = useState<string | null>(null);
+  const abortRef = useRef<AbortController | null>(null);
+
+  const navigate = useCallback(
+    (
+      target: string | undefined,
+      hidden: boolean,
+      opts?: { fallbackOnError?: boolean },
+    ) => {
+      abortRef.current?.abort();
+      const ctrl = new AbortController();
+      abortRef.current = ctrl;
+      setLoading(true);
+      setError(null);
+      // Forward the signal so cancelled navigation actually cancels the
+      // backend enumeration instead of just discarding the response.
+      browseFolders(target, hidden, ctrl.signal)
+        .then((res) => {
+          if (ctrl.signal.aborted) return;
+          setData(res);
+          setPath(res.current);
+        })
+        .catch((err) => {
+          if (ctrl.signal.aborted) return;
+          // Surface the error, but if the very first request (typically
+          // a typo'd or denylisted ``initialPath``) fails AND the
+          // browser is empty (no ``data`` to render against), fall
+          // back to the user's HOME so the modal is navigable instead
+          // of an irrecoverable dead end.
+          const message = err instanceof Error ? err.message : String(err);
+          setError(message);
+          if (opts?.fallbackOnError && target !== undefined) {
+            // Re-issue without a target -> backend defaults to HOME.
+            // Don't recurse if HOME itself fails (paranoia: shouldn't
+            // happen since the sandbox allowlist always includes HOME).
+            queueMicrotask(() => navigate(undefined, hidden));
+          }
+        })
+        .finally(() => {
+          if (!ctrl.signal.aborted) setLoading(false);
+        });
+    },
+    [],
+  );
+
+  // Fetch when the dialog opens.  Only re-run when the dialog transitions
+  // closed -> open; subsequent navigation is driven by `navigate()` so we
+  // don't want `path` in the dependency list here.
+  // eslint-disable-next-line react-hooks/exhaustive-deps
+  useEffect(() => {
+    if (!open) return;
+    // ``fallbackOnError``: if the user-supplied ``initialPath`` is bad
+    // (typo, denylisted, deleted) we recover into HOME instead of
+    // showing an empty modal with no breadcrumbs/entries.
+    navigate(initialPath, showHidden, { fallbackOnError: true });
+    // eslint-disable-next-line react-hooks/exhaustive-deps
+  }, [open]);
+
+  const handleConfirm = useCallback(() => {
+    if (!path) return;
+    onSelect(path);
+    onOpenChange(false);
+  }, [onSelect, onOpenChange, path]);
+
+  const crumbs = useMemo(
+    () => (data?.current ? splitBreadcrumb(data.current) : []),
+    [data?.current],
+  );
+
+  return (
+    <Dialog open={open} onOpenChange={onOpenChange}>
+      <DialogContent
+        className="max-w-md p-0 gap-0"
+        overlayClassName="bg-black/20 backdrop-blur-none"
+        data-testid="folder-browser-dialog"
+      >
+        <DialogHeader className="px-4 pt-4 pb-2">
+          <DialogTitle className="text-sm font-medium">
+            Browse for folder
+          </DialogTitle>
+        </DialogHeader>
+
+        {/* Breadcrumb */}
+        <div className="flex flex-wrap items-center gap-0.5 border-t border-border/50 px-4 py-2 font-mono text-[11px] text-muted-foreground">
+          {crumbs.length === 0 ? (
+            <span className="text-muted-foreground/60">(loading…)</span>
+          ) : (
+            crumbs.map((c, i) => (
+              <span key={c.value} className="flex items-center gap-0.5">
+                <button
+                  type="button"
+                  className="rounded px-1 py-0.5 hover:bg-accent hover:text-foreground"
+                  onClick={() => navigate(c.value, showHidden)}
+                  disabled={loading}
+                >
+                  {c.label}
+                </button>
+                {i < crumbs.length - 1 && (
+                  <span className="text-muted-foreground/40">/</span>
+                )}
+              </span>
+            ))
+          )}
+        </div>
+
+        {/* Suggestions (quick-pick chips) */}
+        {data?.suggestions && data.suggestions.length > 0 && (
+          <div className="flex flex-wrap gap-1 border-t border-border/50 px-4 py-2">
+            {data.suggestions.map((s) => (
+              <button
+                key={s}
+                type="button"
+                onClick={() => navigate(s, showHidden)}
+                disabled={loading}
+                className="rounded-full border border-border/50 px-2 py-0.5 font-mono text-[10px] text-muted-foreground transition-colors hover:bg-accent hover:text-foreground disabled:opacity-40"
+                title={s}
+              >
+                {s.length > 36 ? `…${s.slice(-33)}` : s}
+              </button>
+            ))}
+          </div>
+        )}
+
+        {/* Entry list */}
+        <div className="max-h-64 min-h-24 overflow-y-auto border-t border-border/50">
+          {error && (
+            <div className="px-4 py-3 text-xs text-destructive">{error}</div>
+          )}
+          {!error && loading && (
+            <div className="flex items-center gap-2 px-4 py-3">
+              <Spinner className="size-3 text-muted-foreground" />
+              <span className="text-xs text-muted-foreground">Loading…</span>
+            </div>
+          )}
+          {!error && !loading && data && (
+            <>
+              {/* Up row */}
+              {data.parent !== null && (
+                <button
+                  type="button"
+                  onClick={() => navigate(data.parent ?? undefined, showHidden)}
+                  className="flex w-full items-center gap-2 px-4 py-1.5 text-left text-xs text-muted-foreground transition-colors hover:bg-accent hover:text-foreground"
+                >
+                  <HugeiconsIcon
+                    icon={ArrowUp02Icon}
+                    className="size-3 shrink-0"
+                  />
+                  <span className="font-mono">..</span>
+                </button>
+              )}
+              {data.entries.length === 0 && !(data.model_files_here && data.model_files_here > 0) && (
+                <div className="px-4 py-3 text-xs text-muted-foreground/60">
+                  (empty directory)
+                </div>
+              )}
+              {data.model_files_here !== undefined && data.model_files_here > 0 && (
+                <div className="border-t border-border/30 px-4 py-1.5 text-[10px] text-foreground/70">
+                  {data.model_files_here} model file{data.model_files_here === 1 ? "" : "s"} in this folder. Click "Use this folder" to scan it.
+                </div>
+              )}
+              {data.truncated === true && (
+                <div className="border-t border-border/30 px-4 py-1.5 text-[10px] text-muted-foreground/70">
+                  Showing first {data.entries.length} entries. Narrow the path
+                  to see more.
+                </div>
+              )}
+              {data.entries.map((e) => (
+                <button
+                  type="button"
+                  key={e.name}
+                  onClick={() => {
+                    const sep = data.current.endsWith("/") ? "" : "/";
+                    navigate(`${data.current}${sep}${e.name}`, showHidden);
+                  }}
+                  className={cn(
+                    "flex w-full items-center gap-2 px-4 py-1.5 text-left text-xs transition-colors hover:bg-accent hover:text-foreground",
+                    e.hidden && "text-muted-foreground/60",
+                  )}
+                >
+                  <HugeiconsIcon
+                    icon={Folder02Icon}
+                    className={cn(
+                      "size-3 shrink-0",
+                      e.has_models
+                        ? "text-foreground"
+                        : "text-muted-foreground/50",
+                    )}
+                  />
+                  <span className="truncate font-mono">{e.name}</span>
+                  {e.has_models && (
+                    <span className="ml-auto shrink-0 rounded-full border border-border/50 px-1.5 py-0 text-[9px] uppercase tracking-wider text-muted-foreground">
+                      models
+                    </span>
+                  )}
+                </button>
+              ))}
+            </>
+          )}
+        </div>
+
+        {/* Footer */}
+        <DialogFooter className="flex items-center justify-between gap-2 border-t border-border/50 px-4 py-2">
+          <label className="flex cursor-pointer items-center gap-1.5 text-[10px] text-muted-foreground">
+            <input
+              type="checkbox"
+              checked={showHidden}
+              onChange={(e) => {
+                const next = e.target.checked;
+                setShowHidden(next);
+                navigate(path, next);
+              }}
+              className="size-3"
+            />
+            Show hidden
+          </label>
+          <div className="flex gap-2">
+            <DialogClose asChild={true}>
+              <button
+                type="button"
+                className="h-7 rounded border border-border/50 px-2.5 text-[11px] text-muted-foreground transition-colors hover:bg-accent hover:text-foreground"
+              >
+                Cancel
+              </button>
+            </DialogClose>
+            <button
+              type="button"
+              onClick={handleConfirm}
+              disabled={!path || loading || !!error}
+              className="h-7 rounded bg-foreground px-2.5 text-[11px] font-medium text-background transition-colors hover:bg-foreground/90 disabled:opacity-40"
+            >
+              Use this folder
+            </button>
+          </div>
+        </DialogFooter>
+      </DialogContent>
+    </Dialog>
+  );
+}
diff --git a/studio/frontend/src/components/assistant-ui/model-selector/pickers.tsx b/studio/frontend/src/components/assistant-ui/model-selector/pickers.tsx
index dc4c210be4..2f661c2e72 100644
--- a/studio/frontend/src/components/assistant-ui/model-selector/pickers.tsx
+++ b/studio/frontend/src/components/assistant-ui/model-selector/pickers.tsx
@@ -48,6 +48,7 @@ import type { VramFitStatus } from "@/lib/vram";
 import { checkVramFit, estimateLoadingVram } from "@/lib/vram";
 import { Add01Icon, Cancel01Icon, Folder02Icon, Search01Icon } from "@hugeicons/core-free-icons";
 import { HugeiconsIcon } from "@hugeicons/react";
+import { FolderBrowser } from "./folder-browser";
 import { Trash2Icon } from "lucide-react";
 import {
   type ReactNode,
@@ -512,6 +513,7 @@ export function HubModelPicker({
   const [folderError, setFolderError] = useState<string | null>(null);
   const [showFolderInput, setShowFolderInput] = useState(false);
   const [folderLoading, setFolderLoading] = useState(false);
+  const [showFolderBrowser, setShowFolderBrowser] = useState(false);
 
   const refreshLocalModelsList = useCallback(() => {
     listLocalModels()
@@ -537,11 +539,22 @@ export function HubModelPicker({
       .catch(() => {});
   }, []);
 
-  const handleAddFolder = useCallback(async () => {
-    const trimmed = folderInput.trim();
+  const handleAddFolder = useCallback(async (overridePath?: string) => {
+    // Accept an explicit path so the folder browser can submit the
+    // chosen path in the same tick it calls `setFolderInput`; reading
+    // `folderInput` alone would race the state update.
+    const raw = overridePath !== undefined ? overridePath : folderInput;
+    const trimmed = raw.trim();
     if (!trimmed || folderLoading) return;
     setFolderError(null);
     setFolderLoading(true);
+    // True when the request originated from the folder browser's
+    // ``onSelect`` (one-click "Use this folder"). In that flow the
+    // typed-input panel is closed, so the inline ``folderError``
+    // paragraph is invisible. Surface failures via toast instead so
+    // the action doesn't appear to silently no-op when the backend
+    // rejects (denylisted path, sandbox 403, etc.).
+    const fromBrowser = overridePath !== undefined;
     try {
       const created = await addScanFolder(trimmed);
       // Backend returns existing row for duplicates, so deduplicate
@@ -557,7 +570,11 @@ export function HubModelPicker({
       // Background reconciliation with the server
       void refreshScanFolders();
     } catch (e) {
-      setFolderError(e instanceof Error ? e.message : "Failed to add folder");
+      const message = e instanceof Error ? e.message : "Failed to add folder";
+      setFolderError(message);
+      if (fromBrowser) {
+        toast.error("Couldn't add folder", { description: message });
+      }
     } finally {
       setFolderLoading(false);
     }
@@ -984,30 +1001,42 @@ export function HubModelPicker({
 
           {!showHfSection ? (
             <>
-              <div className="flex items-center justify-between px-2.5 py-1.5">
+              <div className="flex items-center justify-between gap-1 px-2.5 py-1.5">
                 <span className="text-[10px] font-semibold uppercase tracking-wider text-muted-foreground">
                   Custom Folders
                 </span>
-                <button
-                  type="button"
-                  aria-label={showFolderInput ? "Cancel adding folder" : "Add scan folder"}
-                  onClick={() => {
-                    setShowFolderInput((open) => {
-                      if (open) { setFolderInput(""); setFolderError(null); }
-                      return !open;
-                    });
-                  }}
-                  className="rounded p-0.5 text-muted-foreground/60 transition-colors hover:text-foreground"
-                >
-                  <HugeiconsIcon icon={showFolderInput ? Cancel01Icon : Add01Icon} className="size-3" />
-                </button>
+                <div className="flex items-center gap-0.5">
+                  <button
+                    type="button"
+                    aria-label="Browse for a folder on the server"
+                    title="Browse folders on the server"
+                    onClick={() => setShowFolderBrowser(true)}
+                    className="shrink-0 rounded p-1 text-muted-foreground/60 transition-colors hover:text-foreground"
+                  >
+                    <HugeiconsIcon icon={Search01Icon} className="size-3" />
+                  </button>
+                  <button
+                    type="button"
+                    aria-label={showFolderInput ? "Cancel adding folder" : "Add scan folder by path"}
+                    title={showFolderInput ? "Cancel" : "Add by typing a path"}
+                    onClick={() => {
+                      setShowFolderInput((open) => {
+                        if (open) { setFolderInput(""); setFolderError(null); }
+                        return !open;
+                      });
+                    }}
+                    className="shrink-0 rounded p-1 text-muted-foreground/60 transition-colors hover:text-foreground"
+                  >
+                    <HugeiconsIcon icon={showFolderInput ? Cancel01Icon : Add01Icon} className="size-3" />
+                  </button>
+                </div>
               </div>
 
               {/* Folder paths */}
               {scanFolders.map((f) => (
                 <div
                   key={f.id}
-                  className="group flex items-center gap-1.5 px-3 py-0.5"
+                  className="group flex items-center gap-1.5 px-2.5 py-0.5"
                 >
                   <HugeiconsIcon icon={Folder02Icon} className="size-3 shrink-0 text-muted-foreground/40" />
                   <span
@@ -1020,9 +1049,9 @@ export function HubModelPicker({
                     type="button"
                     onClick={() => handleRemoveFolder(f.id)}
                     aria-label={`Remove folder ${f.path}`}
-                    className="shrink-0 rounded p-0.5 text-muted-foreground/40 opacity-100 md:opacity-0 md:group-hover:opacity-100 focus-visible:opacity-100 transition-opacity hover:text-destructive"
+                    className="shrink-0 rounded p-1 text-foreground/70 transition-colors hover:bg-destructive/10 hover:text-destructive focus-visible:bg-destructive/10 focus-visible:text-destructive"
                   >
-                    <HugeiconsIcon icon={Cancel01Icon} className="size-2.5" />
+                    <HugeiconsIcon icon={Cancel01Icon} className="size-3" />
                   </button>
                 </div>
               ))}
@@ -1046,7 +1075,17 @@ export function HubModelPicker({
                     />
                     <button
                       type="button"
-                      onClick={handleAddFolder}
+                      onClick={() => setShowFolderBrowser(true)}
+                      disabled={folderLoading}
+                      aria-label="Browse for folder"
+                      title="Browse folders on the server"
+                      className="flex h-6 shrink-0 items-center justify-center rounded border border-border/50 px-1.5 text-muted-foreground transition-colors hover:bg-accent hover:text-foreground disabled:opacity-40"
+                    >
+                      <HugeiconsIcon icon={Search01Icon} className="size-3" />
+                    </button>
+                    <button
+                      type="button"
+                      onClick={() => { void handleAddFolder(); }}
                       disabled={folderLoading || !folderInput.trim()}
                       className="h-6 shrink-0 rounded border border-border/50 px-1.5 text-[10px] text-muted-foreground transition-colors hover:bg-accent disabled:opacity-40"
                     >
@@ -1059,16 +1098,20 @@ export function HubModelPicker({
                 </div>
               )}
 
-              {/* Empty state */}
-              {scanFolders.length === 0 && customFolderModels.length === 0 && !showFolderInput && (
-                <button
-                  type="button"
-                  onClick={() => setShowFolderInput(true)}
-                  className="px-2.5 pb-1.5 text-left text-[10px] text-muted-foreground/60 transition-colors hover:text-muted-foreground"
-                >
-                  + Add a folder to scan for local models
-                </button>
-              )}
+              <FolderBrowser
+                open={showFolderBrowser}
+                onOpenChange={setShowFolderBrowser}
+                initialPath={folderInput.trim() || undefined}
+                onSelect={(picked) => {
+                  setFolderInput(picked);
+                  setFolderError(null);
+                  // One-click UX: the "Use this folder" button submits
+                  // the scan folder directly. Pass the path explicitly
+                  // because `folderInput` state hasn't flushed yet.
+                  void handleAddFolder(picked);
+                }}
+              />
+
 
               {/* Models from custom folders */}
               {customFolderModels.map((m) => {
diff --git a/studio/frontend/src/features/chat/api/chat-api.ts b/studio/frontend/src/features/chat/api/chat-api.ts
index ddc0e9d39e..9aacfc5af4 100644
--- a/studio/frontend/src/features/chat/api/chat-api.ts
+++ b/studio/frontend/src/features/chat/api/chat-api.ts
@@ -247,6 +247,42 @@ export async function removeScanFolder(id: number): Promise<void> {
   await parseJsonOrThrow<unknown>(response);
 }
 
+export interface BrowseEntry {
+  name: string;
+  has_models: boolean;
+  hidden: boolean;
+}
+
+export interface BrowseFoldersResponse {
+  current: string;
+  parent: string | null;
+  entries: BrowseEntry[];
+  suggestions: string[];
+  truncated?: boolean;
+  model_files_here?: number;
+}
+
+export async function browseFolders(
+  path?: string,
+  showHidden = false,
+  signal?: AbortSignal,
+): Promise<BrowseFoldersResponse> {
+  const params = new URLSearchParams();
+  if (path !== undefined && path !== null) params.set("path", path);
+  if (showHidden) params.set("show_hidden", "true");
+  const qs = params.toString();
+  // Forward the AbortSignal through authFetch -> fetch so that a
+  // navigation cancelled in the FolderBrowser (rapid breadcrumb / row /
+  // hidden-toggle clicks) actually cancels the in-flight HTTP request
+  // server-side, instead of merely dropping the response client-side
+  // while the backend keeps walking large directory trees.
+  const response = await authFetch(
+    `/api/models/browse-folders${qs ? `?${qs}` : ""}`,
+    signal ? { signal } : undefined,
+  );
+  return parseJsonOrThrow<BrowseFoldersResponse>(response);
+}
+
 export async function listGgufVariants(
   repoId: string,
   hfToken?: string,

From ba387e2c8fac170780a66813b4828204744701fb Mon Sep 17 00:00:00 2001
From: Daniel Han <danielhanchen@gmail.com>
Date: Wed, 15 Apr 2026 08:06:30 -0700
Subject: [PATCH 112/216] Update pyproject.toml

---
 pyproject.toml | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/pyproject.toml b/pyproject.toml
index 50bdf58b95..b4c0122f4a 100644
--- a/pyproject.toml
+++ b/pyproject.toml
@@ -88,7 +88,7 @@ huggingfacenotorch = [
 ]
 huggingface = [
     "unsloth[huggingfacenotorch]",
-    "unsloth_zoo>=2026.4.3",
+    "unsloth_zoo>=2026.4.7",
     "torchvision",
     "unsloth[triton]",
 ]
@@ -578,7 +578,7 @@ colab-ampere-torch220 = [
     "flash-attn>=2.6.3 ; ('linux' in sys_platform)",
 ]
 colab-new = [
-    "unsloth_zoo>=2026.4.3",
+    "unsloth_zoo>=2026.4.7",
     "packaging",
     "tyro",
     "transformers>=4.51.3,!=4.52.0,!=4.52.1,!=4.52.2,!=4.52.3,!=4.53.0,!=4.54.0,!=4.55.0,!=4.55.1,!=4.57.0,!=4.57.4,!=4.57.5,!=5.0.0,!=5.1.0,<=5.5.0",

From cdb3e752ecc62899522c7f7cb768934c989af7e6 Mon Sep 17 00:00:00 2001
From: Daniel Han <danielhanchen@gmail.com>
Date: Wed, 15 Apr 2026 08:06:43 -0700
Subject: [PATCH 113/216] Update _utils.py

---
 unsloth/models/_utils.py | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/unsloth/models/_utils.py b/unsloth/models/_utils.py
index 7d99ec3932..03f5b965d0 100644
--- a/unsloth/models/_utils.py
+++ b/unsloth/models/_utils.py
@@ -12,7 +12,7 @@
 # See the License for the specific language governing permissions and
 # limitations under the License.
 
-__version__ = "2026.4.4"
+__version__ = "2026.4.5"
 
 __all__ = [
     "SUPPORTS_BFLOAT16",

From 3869fbe1cc6495cb6bc5c0d48e2b3159ee364d5b Mon Sep 17 00:00:00 2001
From: Daniel Han <danielhanchen@gmail.com>
Date: Wed, 15 Apr 2026 08:23:41 -0700
Subject: [PATCH 114/216] Bump installer minimum to 2026.4.5 (#5041)

---
 install.ps1 | 10 +++++-----
 install.sh  | 10 +++++-----
 2 files changed, 10 insertions(+), 10 deletions(-)

diff --git a/install.ps1 b/install.ps1
index 61f31d85b4..ee28a14e5f 100644
--- a/install.ps1
+++ b/install.ps1
@@ -819,7 +819,7 @@ shell.Run cmd, 0, False
         if ($SkipTorch) {
             # No-torch: install unsloth + unsloth-zoo with --no-deps, then
             # runtime deps (typer, safetensors, transformers, etc.) with --no-deps.
-            $baseInstallExit = Invoke-InstallCommand { uv pip install --python $VenvPython --no-deps --reinstall-package unsloth --reinstall-package unsloth-zoo "unsloth>=2026.4.4" unsloth-zoo }
+            $baseInstallExit = Invoke-InstallCommand { uv pip install --python $VenvPython --no-deps --reinstall-package unsloth --reinstall-package unsloth-zoo "unsloth>=2026.4.5" unsloth-zoo }
             if ($baseInstallExit -eq 0) {
                 $NoTorchReq = Find-NoTorchRuntimeFile
                 if ($NoTorchReq) {
@@ -827,7 +827,7 @@ shell.Run cmd, 0, False
                 }
             }
         } else {
-            $baseInstallExit = Invoke-InstallCommand { uv pip install --python $VenvPython --reinstall-package unsloth --reinstall-package unsloth-zoo "unsloth>=2026.4.4" unsloth-zoo }
+            $baseInstallExit = Invoke-InstallCommand { uv pip install --python $VenvPython --reinstall-package unsloth --reinstall-package unsloth-zoo "unsloth>=2026.4.5" unsloth-zoo }
         }
         if ($baseInstallExit -ne 0) {
             Write-Host "[ERROR] Failed to install unsloth (exit code $baseInstallExit)" -ForegroundColor Red
@@ -857,7 +857,7 @@ shell.Run cmd, 0, False
         if ($SkipTorch) {
             # No-torch: install unsloth + unsloth-zoo with --no-deps, then
             # runtime deps (typer, safetensors, transformers, etc.) with --no-deps.
-            $baseInstallExit = Invoke-InstallCommand { uv pip install --python $VenvPython --no-deps --upgrade-package unsloth --upgrade-package unsloth-zoo "unsloth>=2026.4.4" unsloth-zoo }
+            $baseInstallExit = Invoke-InstallCommand { uv pip install --python $VenvPython --no-deps --upgrade-package unsloth --upgrade-package unsloth-zoo "unsloth>=2026.4.5" unsloth-zoo }
             if ($baseInstallExit -eq 0) {
                 $NoTorchReq = Find-NoTorchRuntimeFile
                 if ($NoTorchReq) {
@@ -865,7 +865,7 @@ shell.Run cmd, 0, False
                 }
             }
         } elseif ($StudioLocalInstall) {
-            $baseInstallExit = Invoke-InstallCommand { uv pip install --python $VenvPython --upgrade-package unsloth "unsloth>=2026.4.4" unsloth-zoo }
+            $baseInstallExit = Invoke-InstallCommand { uv pip install --python $VenvPython --upgrade-package unsloth "unsloth>=2026.4.5" unsloth-zoo }
         } else {
             $baseInstallExit = Invoke-InstallCommand { uv pip install --python $VenvPython --upgrade-package unsloth "$PackageName" }
         }
@@ -886,7 +886,7 @@ shell.Run cmd, 0, False
         # Fallback: GPU detection failed to produce a URL -- let uv resolve torch
         substep "installing unsloth (this may take a few minutes)..."
         if ($StudioLocalInstall) {
-            $baseInstallExit = Invoke-InstallCommand { uv pip install --python $VenvPython unsloth-zoo "unsloth>=2026.4.4" --torch-backend=auto }
+            $baseInstallExit = Invoke-InstallCommand { uv pip install --python $VenvPython unsloth-zoo "unsloth>=2026.4.5" --torch-backend=auto }
             if ($baseInstallExit -ne 0) {
                 Write-Host "[ERROR] Failed to install unsloth (exit code $baseInstallExit)" -ForegroundColor Red
                 return
diff --git a/install.sh b/install.sh
index 0dbcdf380e..6915893ddf 100755
--- a/install.sh
+++ b/install.sh
@@ -1316,7 +1316,7 @@ if [ "$_MIGRATED" = true ]; then
         # to prevent transitive torch resolution.
         run_install_cmd "install unsloth (migrated no-torch)" uv pip install --python "$_VENV_PY" --no-deps \
             --reinstall-package unsloth --reinstall-package unsloth-zoo \
-            "unsloth>=2026.4.4" unsloth-zoo
+            "unsloth>=2026.4.5" unsloth-zoo
         _NO_TORCH_RT="$(_find_no_torch_runtime)"
         if [ -n "$_NO_TORCH_RT" ]; then
             run_install_cmd "install no-torch runtime deps" uv pip install --python "$_VENV_PY" --no-deps -r "$_NO_TORCH_RT"
@@ -1324,7 +1324,7 @@ if [ "$_MIGRATED" = true ]; then
     else
         run_install_cmd "install unsloth (migrated)" uv pip install --python "$_VENV_PY" \
             --reinstall-package unsloth --reinstall-package unsloth-zoo \
-            "unsloth>=2026.4.4" unsloth-zoo
+            "unsloth>=2026.4.5" unsloth-zoo
     fi
     if [ "$STUDIO_LOCAL_INSTALL" = true ]; then
         substep "overlaying local repo (editable)..."
@@ -1487,7 +1487,7 @@ elif [ -n "$TORCH_INDEX_URL" ]; then
         # runtime deps (typer, safetensors, transformers, etc.) with --no-deps.
         run_install_cmd "install unsloth (no-torch)" uv pip install --python "$_VENV_PY" --no-deps \
             --upgrade-package unsloth --upgrade-package unsloth-zoo \
-            "unsloth>=2026.4.4" unsloth-zoo
+            "unsloth>=2026.4.5" unsloth-zoo
         _NO_TORCH_RT="$(_find_no_torch_runtime)"
         if [ -n "$_NO_TORCH_RT" ]; then
             run_install_cmd "install no-torch runtime deps" uv pip install --python "$_VENV_PY" --no-deps -r "$_NO_TORCH_RT"
@@ -1498,7 +1498,7 @@ elif [ -n "$TORCH_INDEX_URL" ]; then
         fi
     elif [ "$STUDIO_LOCAL_INSTALL" = true ]; then
         run_install_cmd "install unsloth (local)" uv pip install --python "$_VENV_PY" \
-            --upgrade-package unsloth "unsloth>=2026.4.4" unsloth-zoo
+            --upgrade-package unsloth "unsloth>=2026.4.5" unsloth-zoo
         substep "overlaying local repo (editable)..."
         run_install_cmd "overlay local repo" uv pip install --python "$_VENV_PY" -e "$_REPO_ROOT" --no-deps
     else
@@ -1525,7 +1525,7 @@ else
     # Fallback: GPU detection failed to produce a URL -- let uv resolve torch
     substep "installing unsloth (this may take a few minutes)..."
     if [ "$STUDIO_LOCAL_INSTALL" = true ]; then
-        run_install_cmd "install unsloth (auto torch backend)" uv pip install --python "$_VENV_PY" unsloth-zoo "unsloth>=2026.4.4" --torch-backend=auto
+        run_install_cmd "install unsloth (auto torch backend)" uv pip install --python "$_VENV_PY" unsloth-zoo "unsloth>=2026.4.5" --torch-backend=auto
         substep "overlaying local repo (editable)..."
         run_install_cmd "overlay local repo" uv pip install --python "$_VENV_PY" -e "$_REPO_ROOT" --no-deps
     else

From a4d4dfe4ac87bea8b2b0bfbc80a20c57b96e3b95 Mon Sep 17 00:00:00 2001
From: DoubleMathew <mmathew23@gmail.com>
Date: Wed, 15 Apr 2026 17:50:48 -0500
Subject: [PATCH 115/216] fix Gemma4 flash attn disable (#5045)

* fix pass attn implementation

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

---------

Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com>
---
 unsloth/models/_utils.py | 70 ++++++++++++++++++++++++++++++++++++++++
 unsloth/models/vision.py | 26 ++++++++++++++-
 2 files changed, 95 insertions(+), 1 deletion(-)

diff --git a/unsloth/models/_utils.py b/unsloth/models/_utils.py
index 03f5b965d0..1e0b015c44 100644
--- a/unsloth/models/_utils.py
+++ b/unsloth/models/_utils.py
@@ -233,6 +233,8 @@ def apply_unsloth_gradient_checkpointing(
 # access on some GPU architectures (B200). Falls back to eager safely.
 _FLEX_EXCLUDED_MODELS = ("gpt_oss", "mllama", "nemotron_h", "modernbert")
 _EAGER_ONLY_PREFIXES = ("gemma3n",)
+_FLASH_ATTENTION_DISABLED_MODELS = ("gemma4", "gemma4_text")
+_FLASH_ATTENTION_DISABLED_WARNED = set()
 
 
 def _is_flex_excluded(model_type):
@@ -243,6 +245,61 @@ def _is_eager_only(model_type):
     return any(model_type.startswith(p) for p in _EAGER_ONLY_PREFIXES)
 
 
+def _is_flash_attention_disabled(model_type):
+    return model_type in _FLASH_ATTENTION_DISABLED_MODELS
+
+
+def _is_flash_attention_requested(attn_implementation):
+    return isinstance(attn_implementation, str) and attn_implementation.startswith(
+        "flash_attention"
+    )
+
+
+def _disable_flash_attention_if_needed(
+    model_type,
+    config,
+    attn_implementation = None,
+    supports_sdpa = False,
+    would_use_flash_attention = False,
+):
+    if not _is_flash_attention_disabled(model_type):
+        return attn_implementation
+
+    requested_attn_implementation = attn_implementation
+    if requested_attn_implementation is None:
+        requested_attn_implementation = getattr(config, "_attn_implementation", None)
+    if requested_attn_implementation is None:
+        requested_attn_implementation = getattr(config, "attn_implementation", None)
+
+    if requested_attn_implementation == "eager":
+        return _set_attn_impl(config, "eager")
+
+    fallback_attn_implementation = "sdpa" if supports_sdpa else "eager"
+    if (
+        _is_flash_attention_requested(requested_attn_implementation)
+        or would_use_flash_attention
+    ):
+        logged_attn_implementation = (
+            requested_attn_implementation
+            if _is_flash_attention_requested(requested_attn_implementation)
+            else "flash_attention_2"
+        )
+        warning_key = (
+            model_type,
+            logged_attn_implementation,
+            fallback_attn_implementation,
+        )
+        if warning_key not in _FLASH_ATTENTION_DISABLED_WARNED:
+            _FLASH_ATTENTION_DISABLED_WARNED.add(warning_key)
+            print(
+                f"Unsloth: `{logged_attn_implementation}` is not supported "
+                "for Gemma 4 - "
+                f"defaulting to `{fallback_attn_implementation}`."
+            )
+
+    return _set_attn_impl(config, fallback_attn_implementation)
+
+
 def _set_attn_impl(config, impl):
     """Helper function to set attention implementation on config and return it."""
     if config is not None:
@@ -260,6 +317,19 @@ def determine_attention_implementation(model_class, config):
         _set_attn_impl(config, "eager")
         return "eager"
 
+    # Models with known Flash Attention incompatibilities. Gemma 4 full-attention
+    # layers use global_head_dim=512, which exceeds Flash Attention's dense
+    # head-dim support. Keep explicit eager requests, otherwise prefer SDPA.
+    if _is_flash_attention_disabled(model_type):
+        supports_sdpa = model_class is not None and getattr(
+            model_class, "_supports_sdpa", False
+        )
+        return _disable_flash_attention_if_needed(
+            model_type,
+            config,
+            supports_sdpa = supports_sdpa,
+        )
+
     # Flash Attention 2
     if HAS_FLASH_ATTENTION and model_class is not None:
         supports_fa2 = getattr(model_class, "_supports_flash_attn_2", False) or getattr(
diff --git a/unsloth/models/vision.py b/unsloth/models/vision.py
index 2bdff55a56..e31617f89a 100644
--- a/unsloth/models/vision.py
+++ b/unsloth/models/vision.py
@@ -29,7 +29,13 @@ except:
 from ..kernels import (
     post_patch_loss_function,
 )
-from ._utils import __version__, importlib_version, _prepare_model_for_qat
+from ._utils import (
+    __version__,
+    importlib_version,
+    _prepare_model_for_qat,
+    _is_flash_attention_disabled,
+    _disable_flash_attention_if_needed,
+)
 from ._utils import *
 from .loader_utils import _get_fp8_mode_and_check_settings
 from ..save import patch_saving_functions
@@ -607,6 +613,7 @@ class FastBaseModel:
                 token = token,
                 trust_remote_code = trust_remote_code,
             )
+        user_attn_implementation = kwargs.get("attn_implementation", None)
         try:
             model_class = auto_model._model_mapping[auto_config.__class__]
         except Exception:
@@ -631,6 +638,23 @@ class FastBaseModel:
 
         if not ("attn_implementation" in kwargs):
             kwargs["attn_implementation"] = attn_impl
+        model_type = getattr(auto_config, "model_type", "").lower()
+        if _is_flash_attention_disabled(model_type):
+            supports_fa2 = model_class is not None and (
+                getattr(model_class, "_supports_flash_attn_2", False)
+                or getattr(model_class, "_supports_flash_attn", False)
+            )
+            kwargs["attn_implementation"] = _disable_flash_attention_if_needed(
+                model_type,
+                auto_config,
+                kwargs.get("attn_implementation"),
+                supports_sdpa = supports_sdpa,
+                would_use_flash_attention = (
+                    user_attn_implementation is None
+                    and HAS_FLASH_ATTENTION
+                    and supports_fa2
+                ),
+            )
         if not supports_sdpa and kwargs.get("attn_implementation") == "sdpa":
             print(
                 f"Unsloth: {model_type_arch.title()} does not support SDPA - switching to fast eager."

From 14ab6fbfae79b9b8ee8612793ecd3f2fac528d93 Mon Sep 17 00:00:00 2001
From: Imgyu Kim <kimimgo@gmail.com>
Date: Thu, 16 Apr 2026 16:21:29 +0900
Subject: [PATCH 116/216] BUG: fix _fix_chat_template for ChatML templates
 missing add_generation_prompt (#4426)

Fixes #4150.

Pre-PR, `_fix_chat_template` only patched templates where a trailing `{{ ... }}` expression followed the last `{% endfor %}`. ChatML templates (Hermes, Magnum, Phi-4, etc.) that end cleanly at `{% endfor %}` with no generation-prompt block were left unchanged, so the outer `fix_chat_template` raised:

```
RuntimeError: Unsloth: The tokenizer `...` does not have a
{% if add_generation_prompt %} for generation purposes.
```

This commonly shows up when a downstream tool (LlamaFactory, Axolotl) re-serializes the tokenizer during LoRA save and strips the generation-prompt block.

This PR adds a second branch to `_fix_chat_template` that fires when:

- the content after the last `{% endfor %}` is empty modulo Jinja `{# ... #}` comments,
- the scrubbed template contains `<|im_start|>` and `<|im_end|>`,
- and the scrubbed template does not already mention `add_generation_prompt`.

The assistant-turn separator is inferred from the template itself (preferring an explicit `'<|im_start|>assistant<sep>'` literal, then the unique `message['role'] + '<sep>'` from role concatenations, then `<|im_sep|>` for Phi-4-mini mixed-separator templates, then `\n`), so Phi-4-style templates are not silently corrupted with the wrong separator.

Verified against the existing chat-template corpus:

- Hermes-3, Magnum-v2, Phi-4-mini, Phi-4 multi-sep, ChatML with trailing whitespace, ChatML with trailing Jinja comment, dot-access `message.role`, split-literal `'<|im_start|>assistant'`: all repaired with the correct assistant prefix.
- Already-fixed ChatML templates: idempotent NOP.
- Trap templates with `<|im_start|>` only inside a Jinja comment: correctly not rewritten.
- Llama-3, Gemma-3, Qwen2.5 (non-ChatML): byte-identical.
- Mistral family (5 models including Mistral-Nemo, Mistral-Small-24B, Mixtral): byte-identical, protected both by the structural guard (no ChatML tokens) and the existing name-based exemption in `load_correct_tokenizer`.
- Qwen family (14 models including Qwen2.5, Qwen3, Qwen3-Coder, QwQ, VL, Math, Qwen3-Guard): byte-identical.

End-to-end reproduction: Hermes-3 LoRA SFT, save with stripped chat_template, reload. Pre-PR code path raises the RuntimeError above. Post-PR reload loads cleanly, patches the template at load time, and `apply_chat_template(add_generation_prompt=True)` produces the correct `<|im_start|>assistant\n` prefix.
---
 unsloth/tokenizer_utils.py | 48 ++++++++++++++++++++++++++++++++++++++
 1 file changed, 48 insertions(+)

diff --git a/unsloth/tokenizer_utils.py b/unsloth/tokenizer_utils.py
index a3f8affccf..4fc09ed76b 100644
--- a/unsloth/tokenizer_utils.py
+++ b/unsloth/tokenizer_utils.py
@@ -677,6 +677,54 @@ def _fix_chat_template(chat_template):
         )
 
         chat_template = chat_template[: where + len(chosen_end)] + after_endfor
+
+    elif re.sub(r"\{#.*?#\}", "", after_endfor, flags = re.DOTALL).strip() == "":
+        # GH#4150: ChatML templates ending at {% endfor %} without an
+        # add_generation_prompt block. Scrub Jinja `{# ... #}` comments so
+        # tokens inside comments cannot fool the guard below.
+        scrubbed = re.sub(r"\{#.*?#\}", "", chat_template, flags = re.DOTALL)
+        if (
+            "<|im_start|>" in scrubbed
+            and "<|im_end|>" in scrubbed
+            and "add_generation_prompt" not in scrubbed
+        ):
+            # Infer the assistant-turn separator. Prefer an explicit
+            # '<|im_start|>assistant<sep>' literal; else the unique
+            # `message['role'] + '<sep>'` from role concatenations; else
+            # '<|im_sep|>' if present (Phi-4-mini uses '\n' for system and
+            # '<|im_sep|>' for user/assistant); else '\n'.
+            assistant_match = re.search(
+                r"""(['"])<\|im_start\|>assistant([^'"]*)\1""",
+                scrubbed,
+            )
+            role_seps = [
+                m.group(2)
+                for m in re.finditer(
+                    r"""message(?:\[['"]role['"]\]|\.role)\s*\+\s*(['"])([^'"]*)\1""",
+                    scrubbed,
+                )
+            ]
+            unique_role_seps = list(dict.fromkeys(role_seps))
+            if assistant_match is not None and assistant_match.group(2):
+                separator = assistant_match.group(2)
+            elif len(unique_role_seps) == 1:
+                separator = unique_role_seps[0]
+            elif "<|im_sep|>" in scrubbed:
+                separator = "<|im_sep|>"
+            else:
+                separator = "\\n"
+            # Emit a double-quoted Jinja literal so a single quote in the
+            # separator cannot break the block. Drop trailing whitespace/
+            # comments after endfor: they would render as stray output
+            # after the generation prefix.
+            assistant_prefix = "<|im_start|>assistant" + separator
+            generation_block = (
+                "{%" + dash + " if add_generation_prompt %}"
+                '{{ "' + assistant_prefix.replace('"', '\\"') + '" }}'
+                "{%" + dash + " endif %}"
+            )
+            chat_template = chat_template[: where + len(chosen_end)] + generation_block
+
     return chat_template
 
 

From ec32ce2e822289a192b92f877ef4f7b5d1850434 Mon Sep 17 00:00:00 2001
From: Etherll <61019402+Etherll@users.noreply.github.com>
Date: Thu, 16 Apr 2026 13:49:51 +0200
Subject: [PATCH 117/216] fix: use direct registry API for PATH writes instead
 of SetEnvironmentVariable (#4961)
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

* fix: replacing SetEnvironmentVariable with direct registry API

* apply reviews

* Use CreateSubKey for HKCU\Environment

* Store PATH backup under HKCU\Software\Unsloth

* Fix $backupKey registry handle leak in PATH backup block

Wrap $backupKey operations in try/finally so the handle is closed even
if GetValue or SetValue throws. The Add-ToUserPath helper already uses
this pattern for its registry key -- the backup block was the only
place missing it.

* Isolate WM_SETTINGCHANGE broadcast from PATH write error handling

Wrap the broadcast dummy-variable calls in their own try/catch so a
broadcast failure does not mask a successful registry PATH write.
Previously, if SetEnvironmentVariable threw after SetValue already
committed the new PATH, Add-ToUserPath would return $false and the
caller would skip Refresh-SessionPath.

* PATH helper polish: venv precedence, quoted entries, raw/expanded dedup

Three small follow-ups surfaced by a 10-reviewer pass against the rebased
PR head. None fix a regression vs main; each strictly improves the new
helpers.

Refresh-SessionPath / Refresh-Environment:
- Move $env:Path to the front of the merge so an activated venv keeps
  precedence over machine/user PATH after a refresh. Pre-PR dropped
  process-only entries entirely; post-PR kept them but at the back.
- Dedup on both raw and expanded forms so %USERPROFILE%\foo and the
  already-expanded C:\Users\me\foo do not both survive.

Add-ToUserPath:
- Trim whitespace and surrounding double-quotes from each compared entry
  so quoted PATH entries like "C:\Program Files\CMake\bin" deduplicate
  against an unquoted directory of the same path.

* Back up User PATH inside Add-ToUserPath, before first mutation

Previously only studio/setup.ps1 took a one-time PATH backup, at script
top (line ~547). install.ps1 (the irm | iex entry point) had no backup,
so users who installed via that path had no recovery surface if anything
clobbered their PATH. The PR description's "one-time backup before any
modifications" promise only held for the studio installer flow.

Move the backup into Add-ToUserPath itself: just before the first actual
SetValue mutation, write the pristine raw PATH to
HKCU\Software\Unsloth\PathBackup if no backup already exists. This:

- Covers both entry points (install.ps1 and studio/setup.ps1).
- Captures the TRUE pristine PATH even when install.ps1 runs first and
  studio/setup.ps1 runs afterwards (the script-top backup in setup.ps1
  would otherwise see an already-modified PATH).
- Is idempotent: once a backup exists, subsequent calls preserve it.
- Skips when nothing would mutate (dedup match) or PATH is empty.

The script-top backup in studio/setup.ps1 is kept for defense in depth.

* Refresh PATH: venv-aware merge order

Reconcile two competing concerns about Refresh-SessionPath /
Refresh-Environment surfaced by separate review rounds:

  - venv at the back -> activated venv loses precedence to system Python
  - process at the front -> stale shims (old node, old python, etc.)
    still on $env:Path can beat a freshly installed tool

New merge order:
  1. Activated venv Scripts dir, only if $env:VIRTUAL_ENV is set
  2. Machine PATH freshly read from registry
  3. User PATH freshly read from registry
  4. Current $env:Path as fallback

This way an explicitly-activated venv keeps priority while a tool the
script just installed wins over any stale entry that was already on
the inherited shell PATH. When no venv is active, fresh registry
entries take precedence as expected.

* Append to User PATH by default, close $envKey in finally

Add-ToUserPath gains a -Position Append|Prepend parameter defaulting to
Append so installing unsloth no longer prepends the bundled venv Scripts
directory ahead of the user's existing python / pip on new shells. The
four current call sites (install.ps1 launcher, studio/setup.ps1 CMake,
nvcc, Python user Scripts) all take the Append default because each one
that needs in-session precedence already does an inline $env:Path prepend
independently. This matches rustup / cargo / nvm / pyenv / uv behavior.

Also wrap the script-top $envKey.GetValue in a try/finally so the
registry handle is released even if the read throws. Matches the pattern
already used for $backupKey five lines below.

* Prepend cmake, nvcc, Python Scripts; keep venv Scripts appended

The previous commit switched Add-ToUserPath to append by default so that
installing unsloth would not silently hijack the user's system python /
pip. That was correct for the venv Scripts dir (which contains python.exe
and pip.exe alongside unsloth.exe), but wrong for the three studio/setup
call sites. Those persist cmake, the driver-compatible nvcc, and the
Python user Scripts dir for future shells, and in all three cases an
older tool already earlier in the user PATH would keep winning after the
install finished. The nvcc case is especially load-bearing: setup selects
a driver-compatible CUDA toolkit, then llama.cpp builds against whatever
wins PATH resolution, so a stale older nvcc produces broken builds.

Pass -Position 'Prepend' explicitly at the three setup.ps1 call sites
(cmake at line 754, nvcc bin at line 1025, Python user Scripts at line
1191). None of those directories holds python.exe, so prepending them
does not re-introduce the original hijack problem. Leave the install.ps1
venv Scripts call on the default Append with a comment explaining why.

* Symmetric dedup, Prepend reorders duplicates, unsloth shim dir

Address three separate findings surfaced by review:

1. Dedup asymmetry (Gemini high-priority): the existing dedup expanded
   registry entries via ExpandEnvironmentVariables but did NOT expand the
   new directory. Passing "%USERPROFILE%\foo" when "C:\Users\me\foo" was
   already in PATH produced a duplicate. Expand both sides so the check
   is symmetric.

2. -Position Prepend no-op on existing duplicates: the dedup loop
   returned $false as soon as it saw a match, regardless of position.
   That left a late-position duplicate in place instead of moving it to
   the front, so "prepend the newly selected cmake/nvcc" did not always
   beat an older copy earlier in PATH. Partition entries into kept and
   dropped lists, then reinsert a single copy at the requested position.
   Append still returns $false on any match so user-curated orderings
   are not reshuffled. Prepend also returns $false when the only copy
   is already at position 0 so we preserve the user's casing.

3. Stop adding the venv Scripts dir to User PATH entirely. That dir
   holds python.exe and pip.exe alongside unsloth.exe, so neither
   Prepend nor Append worked: prepend hijacked the user's system python
   and pip, append made the freshly-installed unsloth.exe lose to any
   older unsloth.exe earlier on PATH. Replace the Scripts-dir PATH add
   with a dedicated shim directory that contains only unsloth.cmd, and
   prepend that dir. The shim calls the venv's unsloth.exe by absolute
   path so future pip upgrades inside the venv propagate automatically.

* Shim via hardlink, Append user Scripts, drop venv sysconfig fallback

Three follow-ups to the c0ab1ab shim commit, targeting concerns raised in
the second 20-reviewer pass:

1. Shim uses unsloth.exe (hardlink, copy fallback) instead of unsloth.cmd.
   The batch-file approach had three distinct regressions:
   - cmd.exe expanded %...% sequences inside user arguments, so prompts
     like "What does 50% mean?" got mangled before reaching the CLI
   - Git Bash / MSYS2 / POSIX-style shells on Windows do not resolve
     bare-name lookups to .cmd files, so `unsloth` stopped working there
   - Set-Content -Encoding ASCII replaced non-ASCII profile characters
     with '?', so installs under C:\Users\Jörg\... wrote a broken shim
   A hardlink (fallback: copy) of unsloth.exe is a native Windows
   executable with no shell indirection. PATHEXT picks .exe before .cmd
   in cmd.exe and PowerShell, Git Bash honors .exe natively, subprocess
   callers hit it directly, and a hardlink stays in sync with the venv
   on pip upgrades because both names point at the same inode.

2. studio/setup.ps1 Python user Scripts dir is added with default Append
   instead of -Position Prepend. That directory holds every pip-installed
   user console script (pip, pytest, huggingface-cli, and so on), not
   just unsloth, so reordering it silently changed resolution order for
   unrelated tools. The new install.ps1 shim at PATH position 0 already
   guarantees `unsloth` resolves to the freshly installed copy, so the
   Python user Scripts entry only needs to be present, not at the front.

3. The sysconfig lookup in studio/setup.ps1 no longer falls back to
   sysconfig.get_path('scripts') when the nt_user scheme dir does not
   exist. When setup.ps1 is invoked from an activated venv (a flow the
   linked issue actually hits) that fallback returns the venv's Scripts
   directory, which would then be added to the persisted User PATH and
   re-introduce the python / pip hijack the shim dir is meant to avoid.
   Stick strictly to the nt_user scheme; skip the block if it does not
   exist on disk.

* Do not crash installer when unsloth.exe shim is locked

The shim update sequence at install.ps1:1095 did a bare Remove-Item /
New-Item HardLink / Copy-Item. Under the script's $ErrorActionPreference
a locked target (most commonly 'unsloth studio' still running while the
user re-invokes the installer) turns the Remove-Item failure into a
terminating error that aborts the install with no actionable message.

The existing shim is perfectly usable in that state, so there is no
reason to abort. Wrap the whole remove/link/copy sequence in a try/catch
that logs the probable cause (Studio still running), points at the fix
(close Studio and re-run), and lets the installer finish with the old
launcher still serving the command.

Also only emit the "added unsloth launcher to PATH" step line when the
launcher was actually (re)created AND the PATH entry was newly added --
previously the message fired even when the shim refresh silently failed,
which was confusing.

* Guard shim PATH entry on existence, use NullString for broadcast delete

Two follow-ups surfaced by the latest review pass:

1. Do not add the shim directory to User PATH when the launcher was not
   actually created. Antivirus blocking unsloth.exe, a disk-full volume,
   or restrictive filesystem permissions can make both the hardlink and
   the copy fallback fail on a fresh install. In that case the existing
   sequence would report "added unsloth launcher to PATH" warnings but
   still prepend the empty $ShimDir to User PATH -- the user sees an
   install that claims success but then cannot resolve `unsloth` in a
   new shell. Gate Add-ToUserPath on Test-Path $ShimExe so the PATH
   entry is only persisted when the launcher is really there.

2. Pass [NullString]::Value instead of $null to the broadcast-delete
   call in Add-ToUserPath. On PowerShell 7.5 and later (running on .NET
   9), a bare $null going into [Environment]::SetEnvironmentVariable
   can be coerced to an empty string rather than a true .NET null,
   which sets the dummy UnslothPathRefresh_XXXXXXXX variable to "" in
   HKCU\Environment instead of deleting it. The leaked variable is
   visible in System Properties and accumulates one entry per install
   run. [NullString]::Value is a PowerShell-specific sentinel that
   crosses the interop boundary as a real null and works on both PS 5.1
   and PS 7.x. See PowerShell/PowerShell#24637 for the underlying issue.

---------

Co-authored-by: Daniel Han <danielhanchen@gmail.com>
Co-authored-by: Lee Jackson <130007945+Imagineer99@users.noreply.github.com>
---
 install.ps1      | 227 +++++++++++++++++++++++++++++++++++++++++++----
 studio/setup.ps1 | 221 ++++++++++++++++++++++++++++++++++++++++-----
 2 files changed, 410 insertions(+), 38 deletions(-)

diff --git a/install.ps1 b/install.ps1
index ee28a14e5f..26ba14c5e5 100644
--- a/install.ps1
+++ b/install.ps1
@@ -100,22 +100,155 @@ function Install-UnslothStudio {
     Write-Host ""
 
     # ── Helper: refresh PATH from registry (deduplicating entries) ──
+    # Merge order:
+    #   1. Activated venv Scripts dir (only if $env:VIRTUAL_ENV is set) so an
+    #      explicitly-activated venv keeps precedence.
+    #   2. Machine, then User PATH freshly read from registry so a tool we
+    #      just installed wins over any stale shim still in $env:Path.
+    #   3. Current $env:Path as fallback so process-only entries that nothing
+    #      else covers are not lost.
+    # Dedup compares both raw and expanded forms so %VAR% references don't
+    # survive twice (once as %VAR%\foo and once as the expanded literal).
     function Refresh-SessionPath {
         $machine = [System.Environment]::GetEnvironmentVariable("Path", "Machine")
         $user    = [System.Environment]::GetEnvironmentVariable("Path", "User")
-        $merged  = "$machine;$user;$env:Path"
+        $venvScripts = if ($env:VIRTUAL_ENV) { Join-Path $env:VIRTUAL_ENV "Scripts" } else { $null }
+        $sources = @()
+        if ($venvScripts) { $sources += $venvScripts }
+        $sources += @($machine, $user, $env:Path)
+        $merged = ($sources | Where-Object { $_ }) -join ";"
         $seen    = @{}
-        $unique  = @()
+        $unique  = New-Object System.Collections.Generic.List[string]
         foreach ($p in $merged -split ";") {
-            $key = $p.TrimEnd("\").ToLowerInvariant()
-            if ($key -and -not $seen.ContainsKey($key)) {
-                $seen[$key] = $true
-                $unique += $p
+            $rawKey = $p.Trim().Trim('"').TrimEnd("\").ToLowerInvariant()
+            $expKey = [Environment]::ExpandEnvironmentVariables($p).Trim().Trim('"').TrimEnd("\").ToLowerInvariant()
+            if ($rawKey -and -not $seen.ContainsKey($rawKey) -and -not $seen.ContainsKey($expKey)) {
+                $seen[$rawKey] = $true
+                if ($expKey -and $expKey -ne $rawKey) { $seen[$expKey] = $true }
+                $unique.Add($p)
             }
         }
         $env:Path = $unique -join ";"
     }
 
+    # ── Helper: safely add a directory to the persistent User PATH ──
+    # Uses direct registry access to preserve REG_EXPAND_SZ type
+    # (avoids .NET SetEnvironmentVariable bug that converts to REG_SZ).
+    #
+    # Position: 'Append' (default) adds $Directory to the END of the persisted
+    # User PATH so existing user tools (e.g. system python, pip) keep taking
+    # precedence in new shells. This matches rustup/cargo/nvm/pyenv/uv behavior
+    # and avoids silently hijacking resolution of common executables. Pass
+    # 'Prepend' only when a caller truly needs the new entry to win over
+    # existing ones at registry scope. In-session precedence should be handled
+    # by an inline $env:Path = "$Dir;$env:Path" prepend instead.
+    function Add-ToUserPath {
+        param(
+            [Parameter(Mandatory = $true)][string]$Directory,
+            [ValidateSet('Append','Prepend')]
+            [string]$Position = 'Append'
+        )
+        try {
+            $regKey = [Microsoft.Win32.Registry]::CurrentUser.CreateSubKey('Environment')
+            try {
+                $rawPath = $regKey.GetValue('Path', '', [Microsoft.Win32.RegistryValueOptions]::DoNotExpandEnvironmentNames)
+                # Explicit string[] cast: a single-entry split otherwise collapses
+                # to a scalar string, which then gets char-indexed and breaks the
+                # partition loop below.
+                [string[]]$entries = if ($rawPath) { $rawPath -split ';' } else { @() }
+                # Normalize both the raw and expanded forms of the new directory
+                # so dedup catches mirror-image cases: PATH holding %USERPROFILE%\foo
+                # vs Directory passed as C:\Users\me\foo, and vice versa.
+                $normalDir = $Directory.Trim().Trim('"').TrimEnd('\').ToLowerInvariant()
+                $expNormalDir = [Environment]::ExpandEnvironmentVariables($Directory).Trim().Trim('"').TrimEnd('\').ToLowerInvariant()
+                # Partition existing entries into "kept" (not our dir) and "dropped"
+                # (matches our dir). Track match indices so we can distinguish
+                # "already at position 0" from "present but at a late position".
+                $kept = New-Object System.Collections.Generic.List[string]
+                $matchIndices = New-Object System.Collections.Generic.List[int]
+                for ($i = 0; $i -lt $entries.Count; $i++) {
+                    $stripped = $entries[$i].Trim().Trim('"')
+                    $rawNorm = $stripped.TrimEnd('\').ToLowerInvariant()
+                    $expNorm = [Environment]::ExpandEnvironmentVariables($stripped).TrimEnd('\').ToLowerInvariant()
+                    $isMatch = ($rawNorm -and ($rawNorm -eq $normalDir -or $rawNorm -eq $expNormalDir)) -or
+                               ($expNorm -and ($expNorm -eq $normalDir -or $expNorm -eq $expNormalDir))
+                    if ($isMatch) {
+                        $matchIndices.Add($i)
+                        continue
+                    }
+                    $kept.Add($entries[$i])
+                }
+                $alreadyPresent = $matchIndices.Count -gt 0
+                # Append semantics: if the entry is already anywhere in PATH we
+                # leave it untouched (idempotent, never reorder user-curated order).
+                if ($alreadyPresent -and $Position -eq 'Append') {
+                    return $false
+                }
+                # Prepend semantics: if the entry is already at position 0 with
+                # exactly one copy, preserve the user's existing casing/form and
+                # no-op. Only rebuild when a reorder or dedup is actually needed.
+                if ($alreadyPresent -and $Position -eq 'Prepend' -and
+                    $matchIndices.Count -eq 1 -and $matchIndices[0] -eq 0) {
+                    return $false
+                }
+                # One-time backup of the pristine User PATH before our first
+                # mutation. Stored under HKCU\Software\Unsloth so a wiped/clobbered
+                # PATH can be recovered. Idempotent: existing backup is preserved.
+                if ($rawPath) {
+                    try {
+                        $backupKey = [Microsoft.Win32.Registry]::CurrentUser.CreateSubKey('Software\Unsloth')
+                        try {
+                            $existingBackup = $backupKey.GetValue('PathBackup', $null)
+                            if (-not $existingBackup) {
+                                $backupKey.SetValue('PathBackup', $rawPath, [Microsoft.Win32.RegistryValueKind]::ExpandString)
+                            }
+                        } finally {
+                            $backupKey.Close()
+                        }
+                    } catch { }
+                }
+                if (-not $rawPath) {
+                    Write-Host "[WARN] User PATH is empty - initializing with $Directory" -ForegroundColor Yellow
+                }
+                $newPath = if ($rawPath) {
+                    if ($Position -eq 'Prepend') {
+                        (@($Directory) + $kept) -join ';'
+                    } else {
+                        ($kept + @($Directory)) -join ';'
+                    }
+                } else {
+                    $Directory
+                }
+                # Prepend idempotency: if the new directory was already at
+                # position 0 (and no duplicates existed elsewhere) the composed
+                # string matches rawPath byte-for-byte. Skip the registry write
+                # so we do not broadcast an unnecessary WM_SETTINGCHANGE.
+                if ($newPath -ceq $rawPath) {
+                    return $false
+                }
+                $regKey.SetValue('Path', $newPath, [Microsoft.Win32.RegistryValueKind]::ExpandString)
+                # Broadcast WM_SETTINGCHANGE so other processes pick up the change.
+                # Use [NullString]::Value (not $null) for the delete call so the
+                # sentinel crosses into .NET as a real null reference -- on
+                # PowerShell 7.5+ / .NET 9, a bare $null here can be coerced to
+                # an empty string, which sets the dummy variable to "" instead
+                # of deleting it and leaves UnslothPathRefresh_XXXXXXXX in
+                # HKCU\Environment permanently.
+                try {
+                    $d = "UnslothPathRefresh_$([guid]::NewGuid().ToString('N').Substring(0,8))"
+                    [Environment]::SetEnvironmentVariable($d, '1', 'User')
+                    [Environment]::SetEnvironmentVariable($d, [NullString]::Value, 'User')
+                } catch { }
+                return $true
+            } finally {
+                $regKey.Close()
+            }
+        } catch {
+            Write-Host "[WARN] Could not update User PATH: $($_.Exception.Message)" -ForegroundColor Yellow
+            return $false
+        }
+    }
+
     function step {
         param(
             [Parameter(Mandatory = $true)][string]$Label,
@@ -945,18 +1078,80 @@ shell.Run cmd, 0, False
 
     New-StudioShortcuts -UnslothExePath $UnslothExe
 
-    # ── Add venv Scripts dir to User PATH so `unsloth studio` works from any terminal ──
-    $ScriptsDir = Join-Path $VenvDir "Scripts"
-    $UserPath = [System.Environment]::GetEnvironmentVariable("Path", "User")
-    if (-not $UserPath -or $UserPath -notlike "*$ScriptsDir*") {
-        if ($UserPath) {
-            [System.Environment]::SetEnvironmentVariable("Path", "$ScriptsDir;$UserPath", "User")
-        } else {
-            [System.Environment]::SetEnvironmentVariable("Path", "$ScriptsDir", "User")
+    # ── Expose the `unsloth` command via a single-purpose shim directory ──
+    # The venv's Scripts dir holds python.exe and pip.exe alongside unsloth.exe,
+    # so adding that dir to PATH (at either position) has unwanted side effects:
+    # Prepend hijacks the user's system python / pip in every future shell;
+    # Append makes the installer's newly-built unsloth.exe lose to any older
+    # unsloth.exe the user already had earlier on PATH. Both are bad.
+    #
+    # Instead we create a small directory that contains only the unsloth
+    # launcher (hardlinked or copied from the venv's Scripts\unsloth.exe),
+    # and Prepend just that directory. Benefits over a .cmd wrapper:
+    #   - no batch %...% expansion of user arguments (e.g. prompts with `%`)
+    #   - works in Git Bash / MSYS2 / POSIX-style shells on Windows that do
+    #     not resolve .cmd by bare name
+    #   - no source encoding concerns on non-ASCII profile paths
+    #   - programmatic callers (subprocess.run, child_process.execFile) hit
+    #     the native executable directly instead of shelling into cmd.exe
+    # We try a hardlink first so pip upgrades inside the venv propagate
+    # automatically (same inode). If the filesystem or volume rejects the
+    # hardlink we fall back to a plain copy, which the next install run
+    # will refresh.
+    $ShimDir = Join-Path $StudioHome "bin"
+    New-Item -ItemType Directory -Force -Path $ShimDir | Out-Null
+    $ShimExe = Join-Path $ShimDir "unsloth.exe"
+    # Wrap the whole remove/link/copy sequence in a try/catch so a locked
+    # launcher does not crash the installer. The common case is a re-run
+    # while the user still has `unsloth studio` open: the existing shim is
+    # held open by the running process, Remove-Item refuses (and under the
+    # script's $ErrorActionPreference this would otherwise be fatal). When
+    # that happens the existing shim is perfectly usable, so we log and
+    # keep going instead of aborting the install.
+    $shimUpdated = $false
+    try {
+        if (Test-Path $ShimExe) { Remove-Item $ShimExe -Force -ErrorAction Stop }
+        try {
+            New-Item -ItemType HardLink -Path $ShimExe -Target $UnslothExe -ErrorAction Stop | Out-Null
+        } catch {
+            # Hardlink unavailable (cross-volume, non-NTFS, permissions). Copy
+            # is self-contained; future pip upgrades inside the venv will not
+            # update the copy until the user re-runs the installer.
+            Copy-Item -Path $UnslothExe -Destination $ShimExe -Force -ErrorAction Stop
+        }
+        $shimUpdated = $true
+    } catch {
+        if (Test-Path $ShimExe) {
+            Write-Host "[WARN] Could not refresh unsloth launcher at $ShimExe." -ForegroundColor Yellow
+            Write-Host "       This usually means a running 'unsloth studio' process still holds the file open." -ForegroundColor Yellow
+            Write-Host "       Close Studio and re-run the installer to pick up the latest launcher." -ForegroundColor Yellow
+            Write-Host "       Continuing with the existing launcher." -ForegroundColor Yellow
+        } else {
+            Write-Host "[WARN] Could not create unsloth launcher at $ShimExe" -ForegroundColor Yellow
+            Write-Host "       $($_.Exception.Message)" -ForegroundColor Yellow
+            Write-Host "       Launch unsloth studio directly via '$UnslothExe' until the next successful install." -ForegroundColor Yellow
         }
-        Refresh-SessionPath
-        step "path" "added unsloth to PATH"
     }
+    # Only add the shim directory to PATH when the launcher actually exists
+    # in it. Otherwise a total shim-creation failure on a fresh install (e.g.
+    # antivirus blocks unsloth.exe, disk full, restrictive FS permissions)
+    # would prepend an empty directory to User PATH and leave the user with
+    # an install that reports success but cannot resolve `unsloth` in a new
+    # shell. Also gate the "added to PATH" step message on both a successful
+    # shim (re)create AND a fresh PATH insertion, so idempotent re-runs stay
+    # quiet.
+    $pathAdded = $false
+    if (Test-Path $ShimExe) {
+        $pathAdded = Add-ToUserPath -Directory $ShimDir -Position 'Prepend'
+    }
+    if ($shimUpdated -and $pathAdded) {
+        step "path" "added unsloth launcher to PATH"
+    }
+    # Sync the current session unconditionally so re-runs in stale terminals
+    # see the shim, and so PATH entries that the studio/setup.ps1 subprocess
+    # persisted (cmake, nvcc, Python Scripts) are visible in this parent
+    # process before it returns control to the user's shell.
+    Refresh-SessionPath
 
     # Launch studio automatically in interactive terminals;
     # in non-interactive environments (CI, Docker) just print instructions.
diff --git a/studio/setup.ps1 b/studio/setup.ps1
index 1ba41360c7..53a559e6e0 100644
--- a/studio/setup.ps1
+++ b/studio/setup.ps1
@@ -73,7 +73,152 @@ function Refresh-Environment {
     }
     $machinePath = [System.Environment]::GetEnvironmentVariable('Path', 'Machine')
     $userPath = [System.Environment]::GetEnvironmentVariable('Path', 'User')
-    $env:Path = "$machinePath;$userPath"
+    # Merge order:
+    #   1. Activated venv Scripts dir (only if $env:VIRTUAL_ENV is set) so an
+    #      explicitly-activated venv keeps precedence.
+    #   2. Machine, then User PATH freshly read from registry so a tool we
+    #      just installed wins over any stale shim still in $env:Path.
+    #   3. Current $env:Path as fallback so process-only entries that nothing
+    #      else covers are not lost.
+    # Dedup compares both raw and expanded forms so %VAR% references don't
+    # survive twice (once as %VAR%\foo and once as the expanded literal).
+    $venvScripts = if ($env:VIRTUAL_ENV) { Join-Path $env:VIRTUAL_ENV 'Scripts' } else { $null }
+    $sources = @()
+    if ($venvScripts) { $sources += $venvScripts }
+    $sources += @($machinePath, $userPath, $env:Path)
+    $merged = ($sources | Where-Object { $_ }) -join ';'
+    $seen = @{}
+    $unique = New-Object System.Collections.Generic.List[string]
+    foreach ($p in $merged -split ";") {
+        $rawKey = $p.Trim().Trim('"').TrimEnd("\").ToLowerInvariant()
+        $expKey = [Environment]::ExpandEnvironmentVariables($p).Trim().Trim('"').TrimEnd("\").ToLowerInvariant()
+        if ($rawKey -and -not $seen.ContainsKey($rawKey) -and -not $seen.ContainsKey($expKey)) {
+            $seen[$rawKey] = $true
+            if ($expKey -and $expKey -ne $rawKey) { $seen[$expKey] = $true }
+            $unique.Add($p)
+        }
+    }
+    $env:Path = $unique -join ";"
+}
+
+# ── Helper: safely add a directory to the persistent User PATH ──
+# Uses direct registry access to preserve REG_EXPAND_SZ type
+# (avoids .NET SetEnvironmentVariable bug that converts to REG_SZ).
+#
+# Position: 'Append' (default) adds $Directory to the END of the persisted
+# User PATH so existing user tools (e.g. system python, pip) keep taking
+# precedence in new shells. This matches rustup/cargo/nvm/pyenv/uv behavior
+# and avoids silently hijacking resolution of common executables. Pass
+# 'Prepend' only when a caller truly needs the new entry to win over
+# existing ones at registry scope. In-session precedence should be handled
+# by an inline $env:Path = "$Dir;$env:Path" prepend instead.
+function Add-ToUserPath {
+    param(
+        [Parameter(Mandatory = $true)][string]$Directory,
+        [ValidateSet('Append','Prepend')]
+        [string]$Position = 'Append'
+    )
+    try {
+        $regKey = [Microsoft.Win32.Registry]::CurrentUser.CreateSubKey('Environment')
+        try {
+            $rawPath = $regKey.GetValue('Path', '', [Microsoft.Win32.RegistryValueOptions]::DoNotExpandEnvironmentNames)
+            # Explicit string[] cast: a single-entry split otherwise collapses
+            # to a scalar string, which then gets char-indexed and breaks the
+            # partition loop below.
+            [string[]]$entries = if ($rawPath) { $rawPath -split ';' } else { @() }
+            # Normalize both the raw and expanded forms of the new directory
+            # so dedup catches mirror-image cases: PATH holding %USERPROFILE%\foo
+            # vs Directory passed as C:\Users\me\foo, and vice versa.
+            $normalDir = $Directory.Trim().Trim('"').TrimEnd('\').ToLowerInvariant()
+            $expNormalDir = [Environment]::ExpandEnvironmentVariables($Directory).Trim().Trim('"').TrimEnd('\').ToLowerInvariant()
+            # Partition existing entries into "kept" (not our dir) and "dropped"
+            # (matches our dir). Track match indices so we can distinguish
+            # "already at position 0" from "present but at a late position".
+            $kept = New-Object System.Collections.Generic.List[string]
+            $matchIndices = New-Object System.Collections.Generic.List[int]
+            for ($i = 0; $i -lt $entries.Count; $i++) {
+                $stripped = $entries[$i].Trim().Trim('"')
+                $rawNorm = $stripped.TrimEnd('\').ToLowerInvariant()
+                $expNorm = [Environment]::ExpandEnvironmentVariables($stripped).TrimEnd('\').ToLowerInvariant()
+                $isMatch = ($rawNorm -and ($rawNorm -eq $normalDir -or $rawNorm -eq $expNormalDir)) -or
+                           ($expNorm -and ($expNorm -eq $normalDir -or $expNorm -eq $expNormalDir))
+                if ($isMatch) {
+                    $matchIndices.Add($i)
+                    continue
+                }
+                $kept.Add($entries[$i])
+            }
+            $alreadyPresent = $matchIndices.Count -gt 0
+            # Append semantics: if the entry is already anywhere in PATH we
+            # leave it untouched (idempotent, never reorder user-curated order).
+            if ($alreadyPresent -and $Position -eq 'Append') {
+                return $false
+            }
+            # Prepend semantics: if the entry is already at position 0 with
+            # exactly one copy, preserve the user's existing casing/form and
+            # no-op. Only rebuild when a reorder or dedup is actually needed.
+            if ($alreadyPresent -and $Position -eq 'Prepend' -and
+                $matchIndices.Count -eq 1 -and $matchIndices[0] -eq 0) {
+                return $false
+            }
+            # One-time backup of the pristine User PATH before our first
+            # mutation. Stored under HKCU\Software\Unsloth so a wiped/clobbered
+            # PATH can be recovered. Idempotent: existing backup is preserved.
+            # The script-top backup at line ~547 covers the studio entry point;
+            # this in-helper backup also covers callers that bypass that block.
+            if ($rawPath) {
+                try {
+                    $backupKey = [Microsoft.Win32.Registry]::CurrentUser.CreateSubKey('Software\Unsloth')
+                    try {
+                        $existingBackup = $backupKey.GetValue('PathBackup', $null)
+                        if (-not $existingBackup) {
+                            $backupKey.SetValue('PathBackup', $rawPath, [Microsoft.Win32.RegistryValueKind]::ExpandString)
+                        }
+                    } finally {
+                        $backupKey.Close()
+                    }
+                } catch { }
+            }
+            if (-not $rawPath) {
+                Write-Host "[WARN] User PATH is empty - initializing with $Directory" -ForegroundColor Yellow
+            }
+            $newPath = if ($rawPath) {
+                if ($Position -eq 'Prepend') {
+                    (@($Directory) + $kept) -join ';'
+                } else {
+                    ($kept + @($Directory)) -join ';'
+                }
+            } else {
+                $Directory
+            }
+            # Prepend idempotency: if the new directory was already at
+            # position 0 (and no duplicates existed elsewhere) the composed
+            # string matches rawPath byte-for-byte. Skip the registry write
+            # so we do not broadcast an unnecessary WM_SETTINGCHANGE.
+            if ($newPath -ceq $rawPath) {
+                return $false
+            }
+            $regKey.SetValue('Path', $newPath, [Microsoft.Win32.RegistryValueKind]::ExpandString)
+            # Broadcast WM_SETTINGCHANGE so other processes pick up the change.
+            # Use [NullString]::Value (not $null) for the delete call so the
+            # sentinel crosses into .NET as a real null reference -- on
+            # PowerShell 7.5+ / .NET 9, a bare $null here can be coerced to
+            # an empty string, which sets the dummy variable to "" instead
+            # of deleting it and leaves UnslothPathRefresh_XXXXXXXX in
+            # HKCU\Environment permanently.
+            try {
+                $d = "UnslothPathRefresh_$([guid]::NewGuid().ToString('N').Substring(0,8))"
+                [Environment]::SetEnvironmentVariable($d, '1', 'User')
+                [Environment]::SetEnvironmentVariable($d, [NullString]::Value, 'User')
+            } catch { }
+            return $true
+        } finally {
+            $regKey.Close()
+        }
+    } catch {
+        Write-Host "[WARN] Could not update User PATH: $($_.Exception.Message)" -ForegroundColor Yellow
+        return $false
+    }
 }
 
 # PowerShell 5.1 compatibility helper: avoid relying on New-TemporaryFile.
@@ -493,6 +638,33 @@ if ($script:StudioVtOk -and -not $env:NO_COLOR) {
     Write-Host "  $Rule" -ForegroundColor DarkGray
 }
 
+# Back up User PATH before any modifications for recovery.
+# Stored under HKCU\Software\Unsloth (not HKCU\Environment) to avoid
+# polluting the process environment block with a multi-KB variable.
+try {
+    $envKey = [Microsoft.Win32.Registry]::CurrentUser.OpenSubKey('Environment', $false)
+    if ($envKey) {
+        try {
+            $rawPath = $envKey.GetValue('Path', '', [Microsoft.Win32.RegistryValueOptions]::DoNotExpandEnvironmentNames)
+        } finally {
+            $envKey.Close()
+        }
+        if ($rawPath) {
+            $backupKey = [Microsoft.Win32.Registry]::CurrentUser.CreateSubKey('Software\Unsloth')
+            try {
+                $existingBackup = $backupKey.GetValue('PathBackup', $null)
+                if (-not $existingBackup) {
+                    $backupKey.SetValue('PathBackup', $rawPath, [Microsoft.Win32.RegistryValueKind]::ExpandString)
+                }
+            } finally {
+                $backupKey.Close()
+            }
+        }
+    }
+} catch {
+    Write-Host "[DEBUG] Could not back up User PATH: $($_.Exception.Message)" -ForegroundColor DarkGray
+}
+
 # ==========================================================================
 #  PHASE 1: System-level prerequisites (winget installs, env vars)
 #  All heavy system tool installs happen here BEFORE touching Python.
@@ -626,11 +798,11 @@ if (-not $HasCmake) {
         foreach ($d in $cmakeDefaults) {
             if (Test-Path (Join-Path $d "cmake.exe")) {
                 $env:Path = "$d;$env:Path"
-                # Persist to user PATH so Refresh-Environment does not drop it later
-                $userPath = [Environment]::GetEnvironmentVariable('Path', 'User')
-                if (-not $userPath -or $userPath -notlike "*$d*") {
-                    [Environment]::SetEnvironmentVariable('Path', "$d;$userPath", 'User')
-                }
+                # Persist to user PATH so Refresh-Environment does not drop it later.
+                # Prepend so the newly-selected cmake wins over any older cmake
+                # entry already in the user PATH (this dir has only cmake.exe, no
+                # python.exe, so prepending does not hijack the user's interpreter).
+                Add-ToUserPath -Directory $d -Position 'Prepend' | Out-Null
                 $HasCmake = $null -ne (Get-Command cmake -ErrorAction SilentlyContinue)
                 if ($HasCmake) {
                     Write-Host "   Found cmake at $d (added to PATH)" -ForegroundColor Gray
@@ -896,14 +1068,12 @@ $nvccBinDir = Split-Path $NvccPath -Parent
 if ($env:PATH -notlike "*$nvccBinDir*") {
     [Environment]::SetEnvironmentVariable('PATH', "$nvccBinDir;$env:PATH", 'Process')
 }
-# Persist nvcc bin dir to User PATH so it works in new terminals
-$userPath = [Environment]::GetEnvironmentVariable('Path', 'User')
-if (-not $userPath -or $userPath -notlike "*$nvccBinDir*") {
-    if ($userPath) {
-        [Environment]::SetEnvironmentVariable('Path', "$nvccBinDir;$userPath", 'User')
-    } else {
-        [Environment]::SetEnvironmentVariable('Path', "$nvccBinDir", 'User')
-    }
+# Persist nvcc bin dir to User PATH so it works in new terminals.
+# Prepend so the toolkit we just selected (driver-compatible) wins over any
+# older CUDA bin dir already on the user PATH. Critical for llama.cpp builds:
+# a later Refresh-Environment could otherwise reorder the selected nvcc behind
+# a stale one. No hijack risk since this dir has only CUDA tools, no python.
+if (Add-ToUserPath -Directory $nvccBinDir -Position 'Prepend') {
     substep "Persisted CUDA bin dir to user PATH"
 }
 
@@ -1061,15 +1231,22 @@ if ($HasPython) {
     $PythonOk = $true
 }
 
-# Ensure Python Scripts dir is on PATH (so 'unsloth' command works in new terminals)
-$ScriptsDir = python -c "import sysconfig; print(sysconfig.get_path('scripts', 'nt_user') if __import__('os').path.exists(sysconfig.get_path('scripts', 'nt_user')) else sysconfig.get_path('scripts'))"
+# Ensure the user-scheme Python Scripts dir is on PATH so any pip-installed
+# console scripts (including `unsloth` if installed via `pip install --user`)
+# are discoverable in new terminals. Stick strictly to the 'nt_user' scheme:
+# we do NOT fall back to sysconfig.get_path('scripts') because that returns
+# the venv's Scripts dir when this setup.ps1 is invoked inside an activated
+# venv, which would re-introduce the python / pip hijack that the dedicated
+# shim directory (install.ps1) was designed to avoid.
+$ScriptsDir = python -c "import os, sysconfig; p = sysconfig.get_path('scripts', 'nt_user'); print(p if os.path.exists(p) else '')"
 if ($LASTEXITCODE -eq 0 -and $ScriptsDir -and (Test-Path $ScriptsDir)) {
-    $UserPath = [Environment]::GetEnvironmentVariable('Path', 'User')
-    $UserPathEntries = if ($UserPath) { $UserPath.Split(';') } else { @() }
-    if (-not ($UserPathEntries | Where-Object { $_.TrimEnd('\') -eq $ScriptsDir })) {
-        $newUserPath = if ($UserPath) { "$ScriptsDir;$UserPath" } else { $ScriptsDir }
-        [Environment]::SetEnvironmentVariable('Path', $newUserPath, 'User')
-
+    # Use Append semantics here: this dir holds ALL user-installed pip
+    # console scripts (pip, pytest, huggingface-cli, etc.), and reordering
+    # it to the front of PATH would silently change resolution precedence
+    # for every one of those tools. Install.ps1 already guarantees the new
+    # `unsloth` wins via a dedicated shim dir at PATH position 0, so we
+    # only need to make sure this directory is present, not at the front.
+    if (Add-ToUserPath -Directory $ScriptsDir) {
         # Also add to current process so it's available immediately
         $ProcessPathEntries = $env:PATH.Split(';')
         if (-not ($ProcessPathEntries | Where-Object { $_.TrimEnd('\') -eq $ScriptsDir })) {

From 6e87bade2504d009c9130b78d46708f0f5cf2166 Mon Sep 17 00:00:00 2001
From: Daniel Han <danielhanchen@gmail.com>
Date: Thu, 16 Apr 2026 12:01:01 +0000
Subject: [PATCH 118/216] Trim verbose comments in PATH helpers

Reduce inline comments from ~160 lines to ~25 across both files.
Keep one-line summaries of the "why"; drop multi-paragraph rationale
blocks that repeated information already captured in commit messages
and PR discussion.
---
 install.ps1      | 114 ++++++++---------------------------------------
 studio/setup.ps1 |  93 +++++++-------------------------------
 2 files changed, 34 insertions(+), 173 deletions(-)

diff --git a/install.ps1 b/install.ps1
index 26ba14c5e5..dfb54b1e87 100644
--- a/install.ps1
+++ b/install.ps1
@@ -100,15 +100,8 @@ function Install-UnslothStudio {
     Write-Host ""
 
     # ── Helper: refresh PATH from registry (deduplicating entries) ──
-    # Merge order:
-    #   1. Activated venv Scripts dir (only if $env:VIRTUAL_ENV is set) so an
-    #      explicitly-activated venv keeps precedence.
-    #   2. Machine, then User PATH freshly read from registry so a tool we
-    #      just installed wins over any stale shim still in $env:Path.
-    #   3. Current $env:Path as fallback so process-only entries that nothing
-    #      else covers are not lost.
-    # Dedup compares both raw and expanded forms so %VAR% references don't
-    # survive twice (once as %VAR%\foo and once as the expanded literal).
+    # Merge order: venv Scripts (if active) > Machine > User > current $env:Path.
+    # Dedup compares both raw and expanded forms (%VAR% vs literal).
     function Refresh-SessionPath {
         $machine = [System.Environment]::GetEnvironmentVariable("Path", "Machine")
         $user    = [System.Environment]::GetEnvironmentVariable("Path", "User")
@@ -132,16 +125,8 @@ function Install-UnslothStudio {
     }
 
     # ── Helper: safely add a directory to the persistent User PATH ──
-    # Uses direct registry access to preserve REG_EXPAND_SZ type
-    # (avoids .NET SetEnvironmentVariable bug that converts to REG_SZ).
-    #
-    # Position: 'Append' (default) adds $Directory to the END of the persisted
-    # User PATH so existing user tools (e.g. system python, pip) keep taking
-    # precedence in new shells. This matches rustup/cargo/nvm/pyenv/uv behavior
-    # and avoids silently hijacking resolution of common executables. Pass
-    # 'Prepend' only when a caller truly needs the new entry to win over
-    # existing ones at registry scope. In-session precedence should be handled
-    # by an inline $env:Path = "$Dir;$env:Path" prepend instead.
+    # Direct registry access preserves REG_EXPAND_SZ (avoids dotnet/runtime#1442).
+    # Append (default) keeps existing tools first; Prepend for must-win entries.
     function Add-ToUserPath {
         param(
             [Parameter(Mandatory = $true)][string]$Directory,
@@ -152,18 +137,9 @@ function Install-UnslothStudio {
             $regKey = [Microsoft.Win32.Registry]::CurrentUser.CreateSubKey('Environment')
             try {
                 $rawPath = $regKey.GetValue('Path', '', [Microsoft.Win32.RegistryValueOptions]::DoNotExpandEnvironmentNames)
-                # Explicit string[] cast: a single-entry split otherwise collapses
-                # to a scalar string, which then gets char-indexed and breaks the
-                # partition loop below.
-                [string[]]$entries = if ($rawPath) { $rawPath -split ';' } else { @() }
-                # Normalize both the raw and expanded forms of the new directory
-                # so dedup catches mirror-image cases: PATH holding %USERPROFILE%\foo
-                # vs Directory passed as C:\Users\me\foo, and vice versa.
+                [string[]]$entries = if ($rawPath) { $rawPath -split ';' } else { @() } # string[] prevents scalar collapse
                 $normalDir = $Directory.Trim().Trim('"').TrimEnd('\').ToLowerInvariant()
                 $expNormalDir = [Environment]::ExpandEnvironmentVariables($Directory).Trim().Trim('"').TrimEnd('\').ToLowerInvariant()
-                # Partition existing entries into "kept" (not our dir) and "dropped"
-                # (matches our dir). Track match indices so we can distinguish
-                # "already at position 0" from "present but at a late position".
                 $kept = New-Object System.Collections.Generic.List[string]
                 $matchIndices = New-Object System.Collections.Generic.List[int]
                 for ($i = 0; $i -lt $entries.Count; $i++) {
@@ -179,21 +155,14 @@ function Install-UnslothStudio {
                     $kept.Add($entries[$i])
                 }
                 $alreadyPresent = $matchIndices.Count -gt 0
-                # Append semantics: if the entry is already anywhere in PATH we
-                # leave it untouched (idempotent, never reorder user-curated order).
-                if ($alreadyPresent -and $Position -eq 'Append') {
+                if ($alreadyPresent -and $Position -eq 'Append') { # Append: idempotent no-op
                     return $false
                 }
-                # Prepend semantics: if the entry is already at position 0 with
-                # exactly one copy, preserve the user's existing casing/form and
-                # no-op. Only rebuild when a reorder or dedup is actually needed.
-                if ($alreadyPresent -and $Position -eq 'Prepend' -and
+                if ($alreadyPresent -and $Position -eq 'Prepend' -and # Prepend: no-op if already at front
                     $matchIndices.Count -eq 1 -and $matchIndices[0] -eq 0) {
                     return $false
                 }
-                # One-time backup of the pristine User PATH before our first
-                # mutation. Stored under HKCU\Software\Unsloth so a wiped/clobbered
-                # PATH can be recovered. Idempotent: existing backup is preserved.
+                # One-time backup under HKCU\Software\Unsloth\PathBackup
                 if ($rawPath) {
                     try {
                         $backupKey = [Microsoft.Win32.Registry]::CurrentUser.CreateSubKey('Software\Unsloth')
@@ -219,21 +188,12 @@ function Install-UnslothStudio {
                 } else {
                     $Directory
                 }
-                # Prepend idempotency: if the new directory was already at
-                # position 0 (and no duplicates existed elsewhere) the composed
-                # string matches rawPath byte-for-byte. Skip the registry write
-                # so we do not broadcast an unnecessary WM_SETTINGCHANGE.
-                if ($newPath -ceq $rawPath) {
+                if ($newPath -ceq $rawPath) { # no actual change
                     return $false
                 }
                 $regKey.SetValue('Path', $newPath, [Microsoft.Win32.RegistryValueKind]::ExpandString)
-                # Broadcast WM_SETTINGCHANGE so other processes pick up the change.
-                # Use [NullString]::Value (not $null) for the delete call so the
-                # sentinel crosses into .NET as a real null reference -- on
-                # PowerShell 7.5+ / .NET 9, a bare $null here can be coerced to
-                # an empty string, which sets the dummy variable to "" instead
-                # of deleting it and leaves UnslothPathRefresh_XXXXXXXX in
-                # HKCU\Environment permanently.
+                # Broadcast WM_SETTINGCHANGE via dummy env-var roundtrip.
+                # [NullString]::Value avoids PS 7.5+/.NET 9 $null-to-"" coercion.
                 try {
                     $d = "UnslothPathRefresh_$([guid]::NewGuid().ToString('N').Substring(0,8))"
                     [Environment]::SetEnvironmentVariable($d, '1', 'User')
@@ -1078,46 +1038,21 @@ shell.Run cmd, 0, False
 
     New-StudioShortcuts -UnslothExePath $UnslothExe
 
-    # ── Expose the `unsloth` command via a single-purpose shim directory ──
-    # The venv's Scripts dir holds python.exe and pip.exe alongside unsloth.exe,
-    # so adding that dir to PATH (at either position) has unwanted side effects:
-    # Prepend hijacks the user's system python / pip in every future shell;
-    # Append makes the installer's newly-built unsloth.exe lose to any older
-    # unsloth.exe the user already had earlier on PATH. Both are bad.
-    #
-    # Instead we create a small directory that contains only the unsloth
-    # launcher (hardlinked or copied from the venv's Scripts\unsloth.exe),
-    # and Prepend just that directory. Benefits over a .cmd wrapper:
-    #   - no batch %...% expansion of user arguments (e.g. prompts with `%`)
-    #   - works in Git Bash / MSYS2 / POSIX-style shells on Windows that do
-    #     not resolve .cmd by bare name
-    #   - no source encoding concerns on non-ASCII profile paths
-    #   - programmatic callers (subprocess.run, child_process.execFile) hit
-    #     the native executable directly instead of shelling into cmd.exe
-    # We try a hardlink first so pip upgrades inside the venv propagate
-    # automatically (same inode). If the filesystem or volume rejects the
-    # hardlink we fall back to a plain copy, which the next install run
-    # will refresh.
+    # ── Expose `unsloth` via a shim dir containing only unsloth.exe ──
+    # We do NOT add the venv Scripts dir to PATH (it also holds python.exe
+    # and pip.exe, which would hijack the user's system interpreter).
+    # Hardlink preferred; falls back to copy if cross-volume or non-NTFS.
     $ShimDir = Join-Path $StudioHome "bin"
     New-Item -ItemType Directory -Force -Path $ShimDir | Out-Null
     $ShimExe = Join-Path $ShimDir "unsloth.exe"
-    # Wrap the whole remove/link/copy sequence in a try/catch so a locked
-    # launcher does not crash the installer. The common case is a re-run
-    # while the user still has `unsloth studio` open: the existing shim is
-    # held open by the running process, Remove-Item refuses (and under the
-    # script's $ErrorActionPreference this would otherwise be fatal). When
-    # that happens the existing shim is perfectly usable, so we log and
-    # keep going instead of aborting the install.
+    # try/catch: if unsloth.exe is locked (Studio running), keep the old shim.
     $shimUpdated = $false
     try {
         if (Test-Path $ShimExe) { Remove-Item $ShimExe -Force -ErrorAction Stop }
         try {
             New-Item -ItemType HardLink -Path $ShimExe -Target $UnslothExe -ErrorAction Stop | Out-Null
         } catch {
-            # Hardlink unavailable (cross-volume, non-NTFS, permissions). Copy
-            # is self-contained; future pip upgrades inside the venv will not
-            # update the copy until the user re-runs the installer.
-            Copy-Item -Path $UnslothExe -Destination $ShimExe -Force -ErrorAction Stop
+            Copy-Item -Path $UnslothExe -Destination $ShimExe -Force -ErrorAction Stop # fallback: copy
         }
         $shimUpdated = $true
     } catch {
@@ -1132,14 +1067,7 @@ shell.Run cmd, 0, False
             Write-Host "       Launch unsloth studio directly via '$UnslothExe' until the next successful install." -ForegroundColor Yellow
         }
     }
-    # Only add the shim directory to PATH when the launcher actually exists
-    # in it. Otherwise a total shim-creation failure on a fresh install (e.g.
-    # antivirus blocks unsloth.exe, disk full, restrictive FS permissions)
-    # would prepend an empty directory to User PATH and leave the user with
-    # an install that reports success but cannot resolve `unsloth` in a new
-    # shell. Also gate the "added to PATH" step message on both a successful
-    # shim (re)create AND a fresh PATH insertion, so idempotent re-runs stay
-    # quiet.
+    # Only add to PATH when the launcher actually exists on disk.
     $pathAdded = $false
     if (Test-Path $ShimExe) {
         $pathAdded = Add-ToUserPath -Directory $ShimDir -Position 'Prepend'
@@ -1147,11 +1075,7 @@ shell.Run cmd, 0, False
     if ($shimUpdated -and $pathAdded) {
         step "path" "added unsloth launcher to PATH"
     }
-    # Sync the current session unconditionally so re-runs in stale terminals
-    # see the shim, and so PATH entries that the studio/setup.ps1 subprocess
-    # persisted (cmake, nvcc, Python Scripts) are visible in this parent
-    # process before it returns control to the user's shell.
-    Refresh-SessionPath
+    Refresh-SessionPath  # sync current session with registry
 
     # Launch studio automatically in interactive terminals;
     # in non-interactive environments (CI, Docker) just print instructions.
diff --git a/studio/setup.ps1 b/studio/setup.ps1
index 53a559e6e0..218be7e045 100644
--- a/studio/setup.ps1
+++ b/studio/setup.ps1
@@ -73,15 +73,7 @@ function Refresh-Environment {
     }
     $machinePath = [System.Environment]::GetEnvironmentVariable('Path', 'Machine')
     $userPath = [System.Environment]::GetEnvironmentVariable('Path', 'User')
-    # Merge order:
-    #   1. Activated venv Scripts dir (only if $env:VIRTUAL_ENV is set) so an
-    #      explicitly-activated venv keeps precedence.
-    #   2. Machine, then User PATH freshly read from registry so a tool we
-    #      just installed wins over any stale shim still in $env:Path.
-    #   3. Current $env:Path as fallback so process-only entries that nothing
-    #      else covers are not lost.
-    # Dedup compares both raw and expanded forms so %VAR% references don't
-    # survive twice (once as %VAR%\foo and once as the expanded literal).
+    # Merge: venv Scripts (if active) > Machine > User > current $env:Path. Dedup raw+expanded.
     $venvScripts = if ($env:VIRTUAL_ENV) { Join-Path $env:VIRTUAL_ENV 'Scripts' } else { $null }
     $sources = @()
     if ($venvScripts) { $sources += $venvScripts }
@@ -102,16 +94,8 @@ function Refresh-Environment {
 }
 
 # ── Helper: safely add a directory to the persistent User PATH ──
-# Uses direct registry access to preserve REG_EXPAND_SZ type
-# (avoids .NET SetEnvironmentVariable bug that converts to REG_SZ).
-#
-# Position: 'Append' (default) adds $Directory to the END of the persisted
-# User PATH so existing user tools (e.g. system python, pip) keep taking
-# precedence in new shells. This matches rustup/cargo/nvm/pyenv/uv behavior
-# and avoids silently hijacking resolution of common executables. Pass
-# 'Prepend' only when a caller truly needs the new entry to win over
-# existing ones at registry scope. In-session precedence should be handled
-# by an inline $env:Path = "$Dir;$env:Path" prepend instead.
+# Direct registry access preserves REG_EXPAND_SZ (avoids dotnet/runtime#1442).
+# Append (default) keeps existing tools first; Prepend for must-win entries.
 function Add-ToUserPath {
     param(
         [Parameter(Mandatory = $true)][string]$Directory,
@@ -122,18 +106,9 @@ function Add-ToUserPath {
         $regKey = [Microsoft.Win32.Registry]::CurrentUser.CreateSubKey('Environment')
         try {
             $rawPath = $regKey.GetValue('Path', '', [Microsoft.Win32.RegistryValueOptions]::DoNotExpandEnvironmentNames)
-            # Explicit string[] cast: a single-entry split otherwise collapses
-            # to a scalar string, which then gets char-indexed and breaks the
-            # partition loop below.
-            [string[]]$entries = if ($rawPath) { $rawPath -split ';' } else { @() }
-            # Normalize both the raw and expanded forms of the new directory
-            # so dedup catches mirror-image cases: PATH holding %USERPROFILE%\foo
-            # vs Directory passed as C:\Users\me\foo, and vice versa.
+            [string[]]$entries = if ($rawPath) { $rawPath -split ';' } else { @() } # string[] prevents scalar collapse
             $normalDir = $Directory.Trim().Trim('"').TrimEnd('\').ToLowerInvariant()
             $expNormalDir = [Environment]::ExpandEnvironmentVariables($Directory).Trim().Trim('"').TrimEnd('\').ToLowerInvariant()
-            # Partition existing entries into "kept" (not our dir) and "dropped"
-            # (matches our dir). Track match indices so we can distinguish
-            # "already at position 0" from "present but at a late position".
             $kept = New-Object System.Collections.Generic.List[string]
             $matchIndices = New-Object System.Collections.Generic.List[int]
             for ($i = 0; $i -lt $entries.Count; $i++) {
@@ -149,23 +124,14 @@ function Add-ToUserPath {
                 $kept.Add($entries[$i])
             }
             $alreadyPresent = $matchIndices.Count -gt 0
-            # Append semantics: if the entry is already anywhere in PATH we
-            # leave it untouched (idempotent, never reorder user-curated order).
-            if ($alreadyPresent -and $Position -eq 'Append') {
+            if ($alreadyPresent -and $Position -eq 'Append') { # Append: idempotent no-op
                 return $false
             }
-            # Prepend semantics: if the entry is already at position 0 with
-            # exactly one copy, preserve the user's existing casing/form and
-            # no-op. Only rebuild when a reorder or dedup is actually needed.
-            if ($alreadyPresent -and $Position -eq 'Prepend' -and
+            if ($alreadyPresent -and $Position -eq 'Prepend' -and # Prepend: no-op if already at front
                 $matchIndices.Count -eq 1 -and $matchIndices[0] -eq 0) {
                 return $false
             }
-            # One-time backup of the pristine User PATH before our first
-            # mutation. Stored under HKCU\Software\Unsloth so a wiped/clobbered
-            # PATH can be recovered. Idempotent: existing backup is preserved.
-            # The script-top backup at line ~547 covers the studio entry point;
-            # this in-helper backup also covers callers that bypass that block.
+            # One-time backup under HKCU\Software\Unsloth\PathBackup
             if ($rawPath) {
                 try {
                     $backupKey = [Microsoft.Win32.Registry]::CurrentUser.CreateSubKey('Software\Unsloth')
@@ -191,21 +157,12 @@ function Add-ToUserPath {
             } else {
                 $Directory
             }
-            # Prepend idempotency: if the new directory was already at
-            # position 0 (and no duplicates existed elsewhere) the composed
-            # string matches rawPath byte-for-byte. Skip the registry write
-            # so we do not broadcast an unnecessary WM_SETTINGCHANGE.
-            if ($newPath -ceq $rawPath) {
+            if ($newPath -ceq $rawPath) { # no actual change
                 return $false
             }
             $regKey.SetValue('Path', $newPath, [Microsoft.Win32.RegistryValueKind]::ExpandString)
-            # Broadcast WM_SETTINGCHANGE so other processes pick up the change.
-            # Use [NullString]::Value (not $null) for the delete call so the
-            # sentinel crosses into .NET as a real null reference -- on
-            # PowerShell 7.5+ / .NET 9, a bare $null here can be coerced to
-            # an empty string, which sets the dummy variable to "" instead
-            # of deleting it and leaves UnslothPathRefresh_XXXXXXXX in
-            # HKCU\Environment permanently.
+            # Broadcast WM_SETTINGCHANGE via dummy env-var roundtrip.
+            # [NullString]::Value avoids PS 7.5+/.NET 9 $null-to-"" coercion.
             try {
                 $d = "UnslothPathRefresh_$([guid]::NewGuid().ToString('N').Substring(0,8))"
                 [Environment]::SetEnvironmentVariable($d, '1', 'User')
@@ -638,9 +595,7 @@ if ($script:StudioVtOk -and -not $env:NO_COLOR) {
     Write-Host "  $Rule" -ForegroundColor DarkGray
 }
 
-# Back up User PATH before any modifications for recovery.
-# Stored under HKCU\Software\Unsloth (not HKCU\Environment) to avoid
-# polluting the process environment block with a multi-KB variable.
+# Back up User PATH under HKCU\Software\Unsloth before any modifications.
 try {
     $envKey = [Microsoft.Win32.Registry]::CurrentUser.OpenSubKey('Environment', $false)
     if ($envKey) {
@@ -798,10 +753,7 @@ if (-not $HasCmake) {
         foreach ($d in $cmakeDefaults) {
             if (Test-Path (Join-Path $d "cmake.exe")) {
                 $env:Path = "$d;$env:Path"
-                # Persist to user PATH so Refresh-Environment does not drop it later.
-                # Prepend so the newly-selected cmake wins over any older cmake
-                # entry already in the user PATH (this dir has only cmake.exe, no
-                # python.exe, so prepending does not hijack the user's interpreter).
+                # Persist to user PATH (Prepend so this cmake wins over older ones).
                 Add-ToUserPath -Directory $d -Position 'Prepend' | Out-Null
                 $HasCmake = $null -ne (Get-Command cmake -ErrorAction SilentlyContinue)
                 if ($HasCmake) {
@@ -1068,11 +1020,7 @@ $nvccBinDir = Split-Path $NvccPath -Parent
 if ($env:PATH -notlike "*$nvccBinDir*") {
     [Environment]::SetEnvironmentVariable('PATH', "$nvccBinDir;$env:PATH", 'Process')
 }
-# Persist nvcc bin dir to User PATH so it works in new terminals.
-# Prepend so the toolkit we just selected (driver-compatible) wins over any
-# older CUDA bin dir already on the user PATH. Critical for llama.cpp builds:
-# a later Refresh-Environment could otherwise reorder the selected nvcc behind
-# a stale one. No hijack risk since this dir has only CUDA tools, no python.
+# Persist nvcc bin dir (Prepend so the driver-compatible toolkit wins).
 if (Add-ToUserPath -Directory $nvccBinDir -Position 'Prepend') {
     substep "Persisted CUDA bin dir to user PATH"
 }
@@ -1231,21 +1179,10 @@ if ($HasPython) {
     $PythonOk = $true
 }
 
-# Ensure the user-scheme Python Scripts dir is on PATH so any pip-installed
-# console scripts (including `unsloth` if installed via `pip install --user`)
-# are discoverable in new terminals. Stick strictly to the 'nt_user' scheme:
-# we do NOT fall back to sysconfig.get_path('scripts') because that returns
-# the venv's Scripts dir when this setup.ps1 is invoked inside an activated
-# venv, which would re-introduce the python / pip hijack that the dedicated
-# shim directory (install.ps1) was designed to avoid.
+# Add user-scheme Python Scripts dir to PATH (nt_user only, no venv fallback).
 $ScriptsDir = python -c "import os, sysconfig; p = sysconfig.get_path('scripts', 'nt_user'); print(p if os.path.exists(p) else '')"
 if ($LASTEXITCODE -eq 0 -and $ScriptsDir -and (Test-Path $ScriptsDir)) {
-    # Use Append semantics here: this dir holds ALL user-installed pip
-    # console scripts (pip, pytest, huggingface-cli, etc.), and reordering
-    # it to the front of PATH would silently change resolution precedence
-    # for every one of those tools. Install.ps1 already guarantees the new
-    # `unsloth` wins via a dedicated shim dir at PATH position 0, so we
-    # only need to make sure this directory is present, not at the front.
+    # Append (not Prepend) -- this dir has other pip scripts; shim handles unsloth.
     if (Add-ToUserPath -Directory $ScriptsDir) {
         # Also add to current process so it's available immediately
         $ProcessPathEntries = $env:PATH.Split(';')

From c5be8b1cd234ceeb15cda929d982a2d84029bf52 Mon Sep 17 00:00:00 2001
From: Daniel Han <danielhanchen@gmail.com>
Date: Thu, 16 Apr 2026 05:52:33 -0700
Subject: [PATCH 119/216] Chat-template repair: warn-by-default, AST
 classification, dict support (#5049)

* Chat-template repair: warn-by-default, AST classification, dict support

Follow-up hardening on top of PR #4426 (which fixed the #4150
RuntimeError for ChatML LoRA reloads).

Behavior changes:

- Warn-by-default instead of RuntimeError. When fix_chat_template cannot
  repair a broken template, emit a warning and return the original.
  Set UNSLOTH_STRICT_CHAT_TEMPLATE=1 to restore the pre-warn hard fail.
  Fixes the UX where a missing `{% if add_generation_prompt %}` block on
  a saved LoRA (typical after LlamaFactory / Axolotl re-serialize) would
  block model loading entirely.

- Local path vs HF hub distinguished in the warning message. For local
  paths the message points at the likely downstream tool; for HF IDs it
  points at the upstream model maintainers. Previously both said "file a
  bug report to the maintainers of <path>" even when <path> was the
  user's own saves/ directory.

- Dict / list chat_template now handled. Hermes-3 ships with
  {default, tool_use} and the previous code crashed with
  AttributeError: 'dict' object has no attribute 'find' when entering
  _fix_chat_template with a dict. Each variant is now fixed
  independently; structure is preserved.

Internals:

- _find_end_position now matches all four Jinja whitespace-control
  variants ({% %}, {%- %}, {% -%}, {%- -%}) and returns the rightmost
  endfor/endif so multi-for templates aren't locked onto the first loop.
  Previously {%- endfor -%} (both-side dash, used by Qwen3-Guard) was
  silently bypassed.

- _has_add_generation_prompt_block uses Jinja AST via
  jinja2.nodes.If/Name walks instead of substring matching, so
  templates that hide the block behind comments or dash-style variants
  are classified correctly.

- _template_ends_with_toplevel_for gates the GH#4150 ChatML repair on
  the AST: only fires when the last structural top-level node is a For
  (standard ChatML shape), ignoring trailing pure-whitespace output
  nodes. Templates wrapped in an outer If (Qwen3-Guard) are now
  explicitly skipped at the _fix_chat_template level as well, not just
  at load_correct_tokenizer's name-based exemption.

- _validate_patched_template renders the patched template with and
  without add_generation_prompt and confirms the patched output
  responds to the flag by appending (not replacing) content. If
  validation fails, the patch is discarded and we fall through to the
  warn path.

Verified with an expanded regression suite in tests/:
- test_fix_chat_template_pr4426.py: 42/42 template-matrix cells
- test_load_correct_tokenizer_pr4426.py: 5/5 tokenizer loads
- test_chat_template_followups.py: 10/10 new follow-up tests
- test_mistral_pr4426.py: 5 Mistral variants byte-identical
- test_qwen_pr4426.py: 14 Qwen variants byte-identical
  (Qwen1.5, Qwen2, Qwen2.5-Instruct/Coder/Math/VL, Qwen3,
  Qwen3-Coder, QwQ, Qwen3-Guard-Gen)

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

* Guard _validate_patched_template against read-only chat_template

If tokenizer.chat_template is a property or otherwise read-only, the
validation helper would crash with AttributeError when trying to
temporarily set the patched template. Catch the assignment failure and
return False (skip validation), and best-effort restore in the finally
block.

* Replace regex separator inference with render-diff; broaden repair to non-ChatML templates

The previous `_infer_assistant_separator` was a four-tier regex heuristic that
only worked on ChatML-shaped templates and forced a hard `<|im_start|>` /
`<|im_end|>` presence gate on Case 2 repair. This meant a Llama-3, Gemma, or
Phi-3 template stripped of its generation-prompt block by a downstream tool
(LlamaFactory, Axolotl, etc.) would still warn-and-return even though the
structural shape is identical to the ChatML case the PR already handles.

This replaces the regex with `_derive_assistant_prefix_by_render`: render the
template with two dialogs that differ only in assistant content, then
`os.path.commonprefix` on the tails captures the exact assistant-turn prefix
the template emits. The template itself is ground truth, so non-ChatML shapes
work as long as the assistant block is a literal the template emits once per
message.

Three guards keep the derivation safe:
  A. both assistant renders extend the base render (no reordering);
  B. the divergence point is exactly the content-insertion site (sentinel
     follows the common prefix);
  C. a user-role cross-check: if a render with a user sentinel also emits
     the same prefix, role has no effect on output and we reject. A render
     failure on [user, user] (e.g. Gemma's `raise_exception` alternation
     check) is evidence that role matters; we accept.

Sentinels differ at character 0 so `commonprefix` cannot absorb them, and
trailing whitespace/comments after the last `{% endfor %}` are stripped
before probing (they would appear in base but not after the appended
assistant turn and break Guard A).

`_fix_chat_template` and `_repair_string_template` now thread an
`is_sharegpt` kwarg; `_fix_chat_template` retries once with
`is_sharegpt=True` if the first probe returns None (dual-probe fallback
for dict/list callers).

The ChatML `<|im_start|>` / `<|im_end|>` hard gate in Case 2 is dropped.
`_infer_assistant_separator` is deleted.

Verified via:
  - tests/test_fix_chat_template_pr4426.py: 51/51 cells (new Llama-3,
    Gemma, Phi-3 broken-template rows all repair FIX-OK)
  - tests/test_load_correct_tokenizer_pr4426.py: 5/5
  - tests/test_chat_template_followups.py: 18/18 (T11-T18 cover
    non-ChatML repair + probe failure modes)
  - tests/test_mistral_pr4426.py: 5/5 byte-identical
  - tests/test_qwen_pr4426.py: 14/14 byte-identical (Qwen3-Guard AST
    gate still rejects)
  - tests/hermes3_lora_pr4426.py reload: patched template ends with
    `<|im_start|>assistant\n`, inference returns sensible output.
  - temp/sim/battery.py: 79/79 followup; vs baseline: 0 regressions,
    9 improvements.
  - Spot-check probe on real stripped tokenizers (Hermes-3, Phi-4,
    Llama-3.2-1B, Gemma-3-1B): all derive the expected prefix.

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

* Address reviewer findings: variant routing, positive-gate detection, comment-safe end scan

Resolves three reviewer findings on PR #5049 (`fix/chat-template-followups`):

Finding #1 [10/10]: dict/list variants now route through
`_fix_chat_template_for_tokenizer` via a new `_VariantTokenizerProxy`
adapter. Previously the dict/list branches called `_fix_chat_template`
directly, silently bypassing the warn/strict (`UNSLOTH_STRICT_CHAT_TEMPLATE`)
contract, the `no == yes` diagnostic, broken-existing-block detection,
and `_validate_patched_template` guard. The proxy swaps
`base.chat_template` to the variant string before each
`apply_chat_template` call so tokenizer globals (`bos_token`, custom
filters, `raise_exception`) remain available; if the base is read-only
it falls back to isolated Jinja rendering.

Finding #2 [1/10]: `_has_add_generation_prompt_block` now requires the
`If` body to contain at least one `Output` node (a new
`_if_body_emits_content` helper walks descendants). This distinguishes a
real generation-prompt block from a header guard like
`{% if not add_generation_prompt is defined %}{% set ... %}{% endif %}`
(body contains only `Assign`) which references the name but emits
nothing. Also dropped a now-redundant `"add_generation_prompt" not in
scrubbed` guard in `_fix_chat_template` Case 2 so header-guarded
templates still get repaired.

Finding #4 [1/10]: `_find_end_position` now replaces Jinja comments with
equal-length whitespace before scanning for `{% endfor %}` / `{% endif %}`
tokens. This prevents a trailing comment containing those tokens from
being picked as the real end tag. Positions in the padded string map 1:1
to positions in the original template.

Tests:
  - tests/test_chat_template_followups.py: 21/21 (T19 strict-mode
    dict variant, T20 header-guard repair, T21 comment-endfor trap
    added; T4/T5 stubs updated with a working apply_chat_template
    that routes through Jinja).
  - tests/test_fix_chat_template_pr4426.py: 51/51 cells unchanged.
  - tests/test_load_correct_tokenizer_pr4426.py: 5/5.
  - tests/test_mistral_pr4426.py: 5/5 byte-identical.
  - tests/test_qwen_pr4426.py: 14/14 byte-identical.
  - temp/sim/battery.py: 79/79 followup; 0 regressions vs baseline.
  - Phase 3 Hermes-3 broken-LoRA reload: inference still returns
    `'The answer to the equation 2+2 is 4.'`.
  - Spot-checks on Hermes-3 / Phi-4 / Llama-3.2-1B / Gemma-3-1B real
    stripped templates: probe still derives the expected prefix.

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

* Tighten comments in chat-template helpers

Pure comment minimization across `_find_end_position`,
`_has_add_generation_prompt_block`, `_if_body_emits_content`,
`_derive_assistant_prefix_by_render`, `_fix_chat_template` Case 2,
and `_VariantTokenizerProxy`. No behavior change; same intent,
fewer lines. All 21 follow-up tests and the 51-cell Phase 1 matrix
still pass.

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

* Sandbox probe, fix is_sharegpt validator mismatch, reject negated gates

Three real bugs from the 10-agent Opus review:

1. Probe now uses `jinja2.sandbox.SandboxedEnvironment` instead of bare
   `jinja2.Environment`. The probe renders at model-load time (before
   the user calls `apply_chat_template`), so it was a new eager
   code-execution surface that the base HF tokenizer loading does not
   have. SandboxedEnvironment blocks attribute-chain exploits at
   negligible cost.

2. `_repair_string_template` now tries validation with both
   `is_sharegpt=False` and `is_sharegpt=True`. Previously, when
   `_fix_chat_template` internally fell back to the other schema via
   its dual-probe, the outer validation still used the caller's
   original `is_sharegpt` -- rendering with the wrong message keys and
   spuriously dropping a valid repair.

3. `_has_add_generation_prompt_block` now skips `If` nodes whose test
   is a `Not` expression. A negated gate like
   `{% if not add_generation_prompt %}{{ x }}{% endif %}` fires when
   agp=False, so its emitting body is not a generation block -- but the
   old code counted any Name reference regardless of polarity.

Cleanup: removed unused `self._label`, added `\r` escape in
generation-block literal, switched variant labels to `!r` formatting,
removed redundant `import os as _os`.

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

* Fix jinja2.sandbox import and sandbox proxy fallback

Two critical findings from the 20-reviewer pass:

1. [20/20] The proxy read-only fallback used bare `jinja2.Environment`,
   not sandboxed. All 20 reviewers independently reproduced marker-file
   creation via `cycler.__init__.__globals__['os'].system(...)` during
   `fix_chat_template()`. Fixed: fallback now uses
   `from jinja2.sandbox import SandboxedEnvironment`.

2. [14/20] The render-diff probe did `import jinja2` then referenced
   `jinja2.sandbox.SandboxedEnvironment`. `jinja2.sandbox` is a
   submodule that is NOT auto-imported by `import jinja2` on Jinja 3.1.6.
   This caused `AttributeError` (swallowed by `except Exception`),
   making the entire Case 2 repair path silently return None in a clean
   process. The 6 reviewers who saw it work had `jinja2.sandbox`
   pre-imported by an earlier module in their process. Fixed: both the
   probe and the proxy fallback now use
   `from jinja2.sandbox import SandboxedEnvironment`.

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

---------

Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com>
---
 unsloth/tokenizer_utils.py | 682 +++++++++++++++++++++++++++++--------
 1 file changed, 544 insertions(+), 138 deletions(-)

diff --git a/unsloth/tokenizer_utils.py b/unsloth/tokenizer_utils.py
index 4fc09ed76b..2fed54dc01 100644
--- a/unsloth/tokenizer_utils.py
+++ b/unsloth/tokenizer_utils.py
@@ -636,173 +636,579 @@ def load_correct_tokenizer(
     return tokenizer
 
 
-def _find_end_position(template, endfor, endif):
-    where_endfor = template.find(endfor)
-    where_endif = template.find(endif)
-    if where_endfor == where_endif == -1:
+# All four Jinja whitespace-control variants of endfor/endif:
+#   {% endfor %}    {%- endfor %}    {% endfor -%}    {%- endfor -%}
+_RE_ENDFOR = re.compile(r"\{%(-?)\s*endfor\s*(-?)%\}")
+_RE_ENDIF = re.compile(r"\{%(-?)\s*endif\s*(-?)%\}")
+_RE_JINJA_COMMENT = re.compile(r"\{#.*?#\}", flags = re.DOTALL)
+
+
+def _find_end_position(template, endfor = None, endif = None):
+    """Rightmost {% endfor %}/{% endif %} (any dash variant), as a dict
+    with start/end/text/dash_left/dash_right. Tokens inside Jinja comments
+    are ignored. `endfor`/`endif` kwargs kept for back-compat, ignored."""
+    # Space-pad comments so positions still map 1:1 to the original.
+    scrubbed = _RE_JINJA_COMMENT.sub(lambda m: " " * len(m.group(0)), template)
+    endfor_matches = list(_RE_ENDFOR.finditer(scrubbed))
+    endif_matches = list(_RE_ENDIF.finditer(scrubbed))
+    last_endfor = endfor_matches[-1] if endfor_matches else None
+    last_endif = endif_matches[-1] if endif_matches else None
+    candidates = [m for m in (last_endfor, last_endif) if m is not None]
+    if not candidates:
         return None
-    elif where_endfor > where_endif:
-        return endfor
+    m = max(candidates, key = lambda x: x.end())
+    return {
+        "start": m.start(),
+        "end": m.end(),
+        "text": m.group(0),
+        "dash_left": bool(m.group(1)),
+        "dash_right": bool(m.group(2)),
+    }
+
+
+def _template_ends_with_toplevel_for(chat_template):
+    """Return True if the last structural node at the template's top level is
+    a For (message-iteration) loop, ignoring trailing pure-whitespace Output
+    nodes. Used to gate the GH#4150 ChatML repair: if the outermost structure
+    is something else (e.g. an outer If that wraps the whole template, as in
+    Qwen3-Guard), we shouldn't inject an {% if add_generation_prompt %}
+    block at the end -- it would land inside or after an unrelated control
+    structure."""
+    try:
+        import jinja2
+        import jinja2.nodes
+
+        ast = jinja2.Environment().parse(chat_template)
+    except Exception:
+        return False
+    for node in reversed(ast.body):
+        # Skip trailing output nodes that are only whitespace -- they come
+        # from trailing whitespace/newlines in the source, not from real
+        # message-rendering logic.
+        if isinstance(node, jinja2.nodes.Output):
+            only_ws = all(
+                isinstance(child, jinja2.nodes.TemplateData)
+                and child.data.strip() == ""
+                for child in node.nodes
+            )
+            if only_ws:
+                continue
+        return isinstance(node, jinja2.nodes.For)
+    return False
+
+
+def _if_body_emits_content(if_node):
+    """True if the If's body contains any Output node (directly or nested).
+    Distinguishes a real generation block from a header guard that only
+    does `{% set ... %}`."""
+    import jinja2.nodes
+
+    for node in if_node.body:
+        if isinstance(node, jinja2.nodes.Output):
+            return True
+        if any(
+            isinstance(d, jinja2.nodes.Output)
+            for d in node.find_all(jinja2.nodes.Output)
+        ):
+            return True
+    return False
+
+
+def _has_add_generation_prompt_block(chat_template):
+    """True if the template has a *positive* `{% if add_generation_prompt %}`
+    gate whose body emits output. Rejects header guards like
+    `{% if not add_generation_prompt is defined %}{% set ... %}{% endif %}`
+    that reference the name but emit nothing. AST-based; string-scan
+    fallback if Jinja fails to parse."""
+    try:
+        import jinja2
+        import jinja2.nodes
+
+        ast = jinja2.Environment().parse(chat_template)
+    except Exception:
+        return "if add_generation_prompt" in chat_template and "%}" in chat_template
+    for if_node in ast.find_all(jinja2.nodes.If):
+        test = if_node.test
+        # Reject negated gates: `{% if not add_generation_prompt %}` fires
+        # when agp=False, so it's not a generation block even if it emits.
+        if isinstance(test, jinja2.nodes.Not):
+            continue
+        # find_all skips the test root, so check bare Name tests explicitly.
+        references_agp = False
+        if isinstance(test, jinja2.nodes.Name) and test.name == "add_generation_prompt":
+            references_agp = True
+        else:
+            for name_node in test.find_all(jinja2.nodes.Name):
+                if name_node.name == "add_generation_prompt":
+                    references_agp = True
+                    break
+        if references_agp and _if_body_emits_content(if_node):
+            return True
+    return False
+
+
+# Sentinels for _derive_assistant_prefix_by_render. Diverge at char 0 so
+# commonprefix can't absorb them; long random tail makes collision with real
+# template literals negligible (see T18).
+_RENDER_DIFF_SENTINEL_A = "AAAA_0123456789_UNSLOTH_RENDER_DIFF_SENTINEL"
+_RENDER_DIFF_SENTINEL_B = "BBBB_0123456789_UNSLOTH_RENDER_DIFF_SENTINEL"
+_RENDER_DIFF_SENTINEL_C = "CCCC_0123456789_UNSLOTH_RENDER_DIFF_SENTINEL"
+
+
+def _derive_assistant_prefix_by_render(chat_template, is_sharegpt = False):
+    """Return the assistant-turn prefix the template emits, derived by
+    rendering two dialogs that differ only in assistant content: the common
+    prefix of their tails (after the base [user]-only render) is what the
+    template emits for an assistant turn. None if any guard fails.
+
+    Works for Llama-3 / Gemma / Phi-3 and other non-ChatML shapes; the
+    template is its own ground truth.
+
+    Known limitation: an `eos-on-non-last` pattern (turn-end sentinel only
+    emitted for non-last messages) would produce a consistent but wrong
+    prefix that `_validate_patched_template` can't catch. No real-world
+    template is known to use this.
+    """
+    try:
+        from jinja2.sandbox import SandboxedEnvironment
+    except Exception:
+        return None
+
+    if is_sharegpt:
+        base_msgs = [{"from": "human", "value": "Hi"}]
+        sent_a_msgs = base_msgs + [{"from": "gpt", "value": _RENDER_DIFF_SENTINEL_A}]
+        sent_b_msgs = base_msgs + [{"from": "gpt", "value": _RENDER_DIFF_SENTINEL_B}]
+        # User-role cross-check (Guard C below).
+        sent_c_msgs = base_msgs + [{"from": "human", "value": _RENDER_DIFF_SENTINEL_C}]
     else:
-        return endif
+        base_msgs = [{"role": "user", "content": "Hi"}]
+        sent_a_msgs = base_msgs + [
+            {"role": "assistant", "content": _RENDER_DIFF_SENTINEL_A}
+        ]
+        sent_b_msgs = base_msgs + [
+            {"role": "assistant", "content": _RENDER_DIFF_SENTINEL_B}
+        ]
+        sent_c_msgs = base_msgs + [{"role": "user", "content": _RENDER_DIFF_SENTINEL_C}]
+
+    # Strip trailing whitespace/comments after the last endfor/endif: they
+    # appear after the message loop and would break Guard A. The splice in
+    # `_fix_chat_template` drops them too.
+    probe_template = chat_template
+    end = _find_end_position(chat_template)
+    if end is not None:
+        after = chat_template[end["end"] :]
+        if _RE_JINJA_COMMENT.sub("", after).strip() == "":
+            probe_template = chat_template[: end["end"]]
+
+    # Sandboxed: probe renders at load time, before user calls
+    # apply_chat_template. SandboxedEnvironment blocks attribute-chain exploits.
+    try:
+        env = SandboxedEnvironment(
+            autoescape = False,
+            keep_trailing_newline = True,
+        )
+        tmpl = env.from_string(probe_template)
+        out_base = tmpl.render(messages = base_msgs, add_generation_prompt = False)
+        out_a = tmpl.render(messages = sent_a_msgs, add_generation_prompt = False)
+        out_b = tmpl.render(messages = sent_b_msgs, add_generation_prompt = False)
+    except Exception:
+        return None
+
+    # Best-effort: alternation-enforcing templates (e.g. Gemma's
+    # raise_exception) fail on [user, user]; that's a positive signal
+    # for Guard C, not a probe failure.
+    out_user_c = None
+    try:
+        out_user_c = tmpl.render(messages = sent_c_msgs, add_generation_prompt = False)
+    except Exception:
+        pass
+
+    # Guard A: assistant renders extend base (no reordering).
+    if not (out_a.startswith(out_base) and out_b.startswith(out_base)):
+        return None
+
+    tail_a = out_a[len(out_base) :]
+    tail_b = out_b[len(out_base) :]
+    if not tail_a or not tail_b:
+        return None
+
+    prefix = os.path.commonprefix([tail_a, tail_b])
+
+    # Guard B: divergence is exactly at the content-insertion site.
+    if not (
+        tail_a[len(prefix) :].startswith(_RENDER_DIFF_SENTINEL_A)
+        and tail_b[len(prefix) :].startswith(_RENDER_DIFF_SENTINEL_B)
+    ):
+        return None
+
+    # Guard C: reject if a [user, user] render also emits the same prefix
+    # (role-insensitive template, e.g. `{% set greeting='Hi' %}...`).
+    if out_user_c is not None and out_user_c.startswith(out_base):
+        tail_c = out_user_c[len(out_base) :]
+        if tail_c.startswith(prefix) and prefix != "":
+            return None
+
+    if not prefix:
+        return None
+
+    return prefix
 
 
-def _fix_chat_template(chat_template):
-    endfor = "{% endfor %}"
-    endif = "{% endif %}"
-    chosen_end = _find_end_position(chat_template, endfor, endif)
-    if chosen_end is None:
-        endfor = "{%- endfor %}"
-        endif = "{%- endif %}"
-        chosen_end = _find_end_position(chat_template, endfor, endif)
-    if chosen_end is None:
+def _fix_chat_template(chat_template, is_sharegpt = False):
+    # Fast path: already has an {% if add_generation_prompt %} block, nothing
+    # to do. This catches cases the old string-based check would miss (e.g.
+    # templates that use {%- if add_generation_prompt -%} with both-side dash,
+    # or that sneak the block into a nested If/For).
+    if _has_add_generation_prompt_block(chat_template):
         return chat_template
 
-    where = chat_template.find(chosen_end)
+    end = _find_end_position(chat_template)
+    if end is None:
+        return chat_template
 
-    after_endfor = chat_template[where + len(chosen_end) :]
-
-    dash = "-" if chosen_end.startswith("{%-") else ""
+    after_endfor = chat_template[end["end"] :]
+    dash_l = "-" if end["dash_left"] else ""
+    dash_r = "-" if end["dash_right"] else ""
+    open_tag = lambda body: "{%" + dash_l + " " + body + " " + dash_r + "%}"
 
+    # Case 1 (pre-existing base case): template ends with a single trailing
+    # {{ expr }} that is the generation prefix. Wrap it in an
+    # {% if add_generation_prompt %} ... {% endif %}.
     if (
-        "{%" + dash + " if" not in after_endfor
-        and "{%" + dash + " set " not in after_endfor
+        "{%" + dash_l + " if" not in after_endfor
+        and "{%" + dash_l + " set " not in after_endfor
         and after_endfor.startswith("{{")
         and after_endfor.endswith("}}")
         and after_endfor.count("{{") == 1
         and after_endfor.count("}}") == 1
     ):
-        after_endfor = (
-            "{%" + dash + " if add_generation_prompt %}" + after_endfor + endif
+        wrapped = (
+            open_tag("if add_generation_prompt") + after_endfor + open_tag("endif")
         )
+        return chat_template[: end["end"]] + wrapped
 
-        chat_template = chat_template[: where + len(chosen_end)] + after_endfor
-
-    elif re.sub(r"\{#.*?#\}", "", after_endfor, flags = re.DOTALL).strip() == "":
-        # GH#4150: ChatML templates ending at {% endfor %} without an
-        # add_generation_prompt block. Scrub Jinja `{# ... #}` comments so
-        # tokens inside comments cannot fool the guard below.
-        scrubbed = re.sub(r"\{#.*?#\}", "", chat_template, flags = re.DOTALL)
-        if (
-            "<|im_start|>" in scrubbed
-            and "<|im_end|>" in scrubbed
-            and "add_generation_prompt" not in scrubbed
-        ):
-            # Infer the assistant-turn separator. Prefer an explicit
-            # '<|im_start|>assistant<sep>' literal; else the unique
-            # `message['role'] + '<sep>'` from role concatenations; else
-            # '<|im_sep|>' if present (Phi-4-mini uses '\n' for system and
-            # '<|im_sep|>' for user/assistant); else '\n'.
-            assistant_match = re.search(
-                r"""(['"])<\|im_start\|>assistant([^'"]*)\1""",
-                scrubbed,
+    # Case 2 (GH#4150): template ends at {% endfor %} with only whitespace
+    # or comments left. Inject an {% if add_generation_prompt %} block with
+    # the assistant prefix derived by render-diff. The top-level-For gate
+    # keeps us out of outer-If wrappers (e.g. Qwen3-Guard).
+    if _RE_JINJA_COMMENT.sub(
+        "", after_endfor
+    ).strip() == "" and _template_ends_with_toplevel_for(chat_template):
+        # No redundant "agp not in scrubbed" check: the fast path already
+        # confirmed no *positive* block, and a mere reference (header
+        # guard) should still get repaired.
+        assistant_prefix = _derive_assistant_prefix_by_render(
+            chat_template, is_sharegpt
+        )
+        # Dual-probe: dict/list callers don't know the shape up front.
+        if assistant_prefix is None and not is_sharegpt:
+            assistant_prefix = _derive_assistant_prefix_by_render(
+                chat_template, is_sharegpt = True
             )
-            role_seps = [
-                m.group(2)
-                for m in re.finditer(
-                    r"""message(?:\[['"]role['"]\]|\.role)\s*\+\s*(['"])([^'"]*)\1""",
-                    scrubbed,
-                )
-            ]
-            unique_role_seps = list(dict.fromkeys(role_seps))
-            if assistant_match is not None and assistant_match.group(2):
-                separator = assistant_match.group(2)
-            elif len(unique_role_seps) == 1:
-                separator = unique_role_seps[0]
-            elif "<|im_sep|>" in scrubbed:
-                separator = "<|im_sep|>"
-            else:
-                separator = "\\n"
-            # Emit a double-quoted Jinja literal so a single quote in the
-            # separator cannot break the block. Drop trailing whitespace/
-            # comments after endfor: they would render as stray output
-            # after the generation prefix.
-            assistant_prefix = "<|im_start|>assistant" + separator
-            generation_block = (
-                "{%" + dash + " if add_generation_prompt %}"
-                '{{ "' + assistant_prefix.replace('"', '\\"') + '" }}'
-                "{%" + dash + " endif %}"
-            )
-            chat_template = chat_template[: where + len(chosen_end)] + generation_block
+        if assistant_prefix is None:
+            return chat_template
+        # Escape for a double-quoted Jinja string literal.
+        escaped = (
+            assistant_prefix.replace("\\", "\\\\")
+            .replace('"', '\\"')
+            .replace("\n", "\\n")
+            .replace("\r", "\\r")
+        )
+        generation_block = (
+            open_tag("if add_generation_prompt")
+            + '{{ "'
+            + escaped
+            + '" }}'
+            + open_tag("endif")
+        )
+        return chat_template[: end["end"]] + generation_block
 
     return chat_template
 
 
+def _is_strict_chat_template_mode():
+    """Opt-in strict mode restores the pre-warn RuntimeError behavior."""
+    val = os.environ.get("UNSLOTH_STRICT_CHAT_TEMPLATE", "0")
+    return str(val).strip().lower() in ("1", "true", "yes", "on")
+
+
+def _name_is_local_path(name_or_path):
+    """True if name_or_path refers to an existing local directory. Used to
+    tailor the warning message: for local paths the user cannot 'file a bug
+    report to the maintainers of <path>' since that path is their own."""
+    if not name_or_path:
+        return False
+    try:
+        return os.path.isdir(str(name_or_path))
+    except Exception:
+        return False
+
+
+def _format_chat_template_message(name_or_path, repaired):
+    """Build a user-facing warning/error message that points at the right
+    responsible party (user's downstream tool vs. upstream model maintainer)."""
+    local = _name_is_local_path(name_or_path)
+    if local:
+        source_hint = (
+            "This tokenizer was loaded from a local path. The likely cause is a "
+            "downstream tool (LlamaFactory, Axolotl, etc.) that re-serialized "
+            "the tokenizer during save and stripped the generation-prompt "
+            "block. Either re-save with the original template, or set "
+            "`tokenizer.chat_template` manually before loading."
+        )
+    else:
+        source_hint = (
+            "The chat_template shipped with `{name}` appears incomplete. "
+            "Consider filing a bug report with the model maintainers."
+        ).format(name = name_or_path)
+    if repaired:
+        return (
+            "Unsloth: Patched the chat_template on `{name}` to add a "
+            "{{% if add_generation_prompt %}} block. {hint}"
+        ).format(name = name_or_path, hint = source_hint)
+    return (
+        "Unsloth: The tokenizer `{name}` does not have a "
+        "{{% if add_generation_prompt %}} block for generation purposes, and "
+        "automatic repair was not possible. The model will still load, but "
+        "`apply_chat_template(add_generation_prompt=True)` may not produce a "
+        "correct assistant-turn marker. {hint} Set "
+        "UNSLOTH_STRICT_CHAT_TEMPLATE=1 to raise instead of warn."
+    ).format(name = name_or_path, hint = source_hint)
+
+
+def _validate_patched_template(tokenizer, patched_template, is_sharegpt):
+    """Render the just-patched template with and without
+    add_generation_prompt, and confirm the patched output responds to the
+    flag by appending (not replacing) content. Returns True if validation
+    passes."""
+    msgs = (
+        [{"from": "human", "value": "Hi"}]
+        if is_sharegpt
+        else [{"role": "user", "content": "Hi"}]
+    )
+    original = getattr(tokenizer, "chat_template", None)
+    try:
+        try:
+            tokenizer.chat_template = patched_template
+        except Exception:
+            return False  # read-only tokenizer, skip validation
+        try:
+            yes = tokenizer.apply_chat_template(
+                msgs,
+                add_generation_prompt = True,
+                tokenize = False,
+            )
+            no = tokenizer.apply_chat_template(
+                msgs,
+                add_generation_prompt = False,
+                tokenize = False,
+            )
+        except Exception:
+            return False
+    finally:
+        try:
+            tokenizer.chat_template = original
+        except Exception:
+            pass  # best-effort restore
+    # Contract after a successful repair: the two renders differ, and the
+    # "yes" render is a strict extension of the "no" render (we only
+    # appended content inside the new add_generation_prompt block).
+    return yes != no and yes.startswith(no)
+
+
+def _repair_string_template(tokenizer, chat_template, is_sharegpt):
+    """Core string-template repair. Returns the repaired template on success,
+    or None if repair was not possible / failed validation."""
+    candidate = _fix_chat_template(chat_template, is_sharegpt = is_sharegpt)
+    if not _has_add_generation_prompt_block(candidate):
+        return None
+    # Validate with the caller's is_sharegpt first. If that fails, the
+    # dual-probe in _fix_chat_template may have fallen back to the other
+    # schema internally -- try validating with the opposite schema before
+    # giving up.
+    if _validate_patched_template(tokenizer, candidate, is_sharegpt):
+        return candidate
+    if _validate_patched_template(tokenizer, candidate, not is_sharegpt):
+        return candidate
+    return None
+
+
+def _fix_chat_template_for_tokenizer(tokenizer, chat_template):
+    """Entry point for a string chat_template. Runs the no==yes diagnostic,
+    attempts repair if needed, and returns the (possibly patched) template.
+
+    On repair failure, the behavior is controlled by
+    UNSLOTH_STRICT_CHAT_TEMPLATE: warn + return original (default) or raise
+    RuntimeError (strict)."""
+    name = getattr(tokenizer, "name_or_path", "unknown")
+
+    # Detect ShareGPT vs HF style by probing apply_chat_template.
+    is_sharegpt = None
+    try:
+        tokenizer.apply_chat_template(
+            [{"role": "user", "content": "Who are you?"}],
+            add_generation_prompt = False,
+            tokenize = False,
+        )
+        is_sharegpt = False
+    except Exception:
+        try:
+            tokenizer.apply_chat_template(
+                [{"from": "human", "value": "Who are you?"}],
+                add_generation_prompt = False,
+                tokenize = False,
+            )
+            is_sharegpt = True
+        except Exception:
+            is_sharegpt = None
+
+    if is_sharegpt is None:
+        return chat_template
+
+    messages = (
+        [{"from": "human", "value": "Who are you?"}]
+        if is_sharegpt
+        else [{"role": "user", "content": "Who are you?"}]
+    )
+    try:
+        no = tokenizer.apply_chat_template(
+            messages,
+            add_generation_prompt = False,
+            tokenize = False,
+        )
+        yes = tokenizer.apply_chat_template(
+            messages,
+            add_generation_prompt = True,
+            tokenize = False,
+        )
+    except Exception:
+        return chat_template
+
+    if no != yes:
+        # Template already responds to the flag; leave as is.
+        return chat_template
+
+    # no == yes: template ignores add_generation_prompt. Try to repair.
+    if _has_add_generation_prompt_block(chat_template):
+        # Template has the block but it does not change output. This is the
+        # "wasn't provided correctly" case from the pre-warn code path.
+        msg = _format_chat_template_message(name, repaired = False)
+        if _is_strict_chat_template_mode():
+            raise RuntimeError(msg)
+        logger.warning_once(msg)
+        return chat_template
+
+    repaired = _repair_string_template(tokenizer, chat_template, is_sharegpt)
+    if repaired is not None:
+        logger.warning_once(_format_chat_template_message(name, repaired = True))
+        return repaired
+
+    msg = _format_chat_template_message(name, repaired = False)
+    if _is_strict_chat_template_mode():
+        raise RuntimeError(msg)
+    logger.warning_once(msg)
+    return chat_template
+
+
+class _VariantTokenizerProxy:
+    """Single-variant view of a multi-variant tokenizer. Routes each variant
+    through `_fix_chat_template_for_tokenizer` so the full contract
+    (is_sharegpt probe, no==yes, warn/strict, `_validate_patched_template`)
+    applies instead of jumping straight to structural repair.
+
+    `apply_chat_template` swaps `base.chat_template` to the variant before
+    calling so tokenizer globals (bos_token, filters, raise_exception) are
+    preserved; falls back to bare Jinja for read-only stubs.
+    """
+
+    def __init__(self, base_tokenizer, variant_template, variant_label = ""):
+        self._base = base_tokenizer
+        self._template = variant_template
+        base_name = getattr(base_tokenizer, "name_or_path", "unknown")
+        self.name_or_path = (
+            f"{base_name} ({variant_label})" if variant_label else base_name
+        )
+
+    @property
+    def chat_template(self):
+        return self._template
+
+    @chat_template.setter
+    def chat_template(self, value):
+        self._template = value
+
+    def apply_chat_template(self, *args, **kwargs):
+        base_original = getattr(self._base, "chat_template", None)
+        swapped = False
+        try:
+            try:
+                self._base.chat_template = self._template
+                swapped = True
+            except Exception:
+                swapped = False
+            if swapped:
+                return self._base.apply_chat_template(*args, **kwargs)
+            # Read-only base: fall back to sandboxed Jinja.
+            from jinja2.sandbox import SandboxedEnvironment
+
+            env = SandboxedEnvironment(
+                autoescape = False,
+                keep_trailing_newline = True,
+            )
+            messages = args[0] if args else kwargs.get("messages", [])
+            add_generation_prompt = kwargs.get("add_generation_prompt", False)
+            return env.from_string(self._template).render(
+                messages = messages,
+                add_generation_prompt = add_generation_prompt,
+            )
+        finally:
+            if swapped:
+                try:
+                    self._base.chat_template = base_original
+                except Exception:
+                    pass  # best-effort restore
+
+
 def fix_chat_template(tokenizer):
     chat_template = getattr(tokenizer, "chat_template", None)
     if chat_template is None:
         return None
 
-    ### 1. Check if add_generation_prompt works
-    # Check for ShareGPT style first
-    is_sharegpt = None
-    try:
-        messages = [
-            {"role": "user", "content": "Who are you?"},
-        ]
-        tokenizer.apply_chat_template(
-            messages, add_generation_prompt = False, tokenize = False
-        )
-        is_sharegpt = False
-    except:
-        try:
-            messages = [
-                {"from": "human", "value": "Who are you?"},
-            ]
-            tokenizer.apply_chat_template(
-                messages, add_generation_prompt = False, tokenize = False
+    # Multi-variant dict (e.g. Hermes-3 {default, tool_use}): route each
+    # variant through the full repair contract via _VariantTokenizerProxy.
+    if isinstance(chat_template, dict):
+        fixed = {}
+        for key, tmpl in chat_template.items():
+            if not isinstance(tmpl, str):
+                fixed[key] = tmpl
+                continue
+            proxy = _VariantTokenizerProxy(
+                tokenizer, tmpl, variant_label = f"variant={key!r}"
             )
-            is_sharegpt = True
-        except:
-            is_sharegpt = None
+            fixed[key] = _fix_chat_template_for_tokenizer(proxy, tmpl)
+        return fixed
 
-    # Not ShareGPT or HF style - just return
-    if is_sharegpt is None:
-        return chat_template
-
-    # Tokenize
-    messages = [
-        {"role": "user", "content": "Who are you?"}
-        if not is_sharegpt
-        else {"from": "human", "value": "Who are you?"}
-    ]
-    no = tokenizer.apply_chat_template(
-        messages, add_generation_prompt = False, tokenize = False
-    )
-    yes = tokenizer.apply_chat_template(
-        messages, add_generation_prompt = True, tokenize = False
-    )
-
-    if no == yes:
-        # SAME?! That's not good! We check for add_generation_prompt
-        if (
-            "{% if add_generation_prompt %}" not in chat_template
-            and "{%- if add_generation_prompt %}" not in chat_template
-        ):
-            # Try fixing it by adding it
-            new_chat_template = _fix_chat_template(chat_template)
-            if (
-                "{% if add_generation_prompt %}" not in new_chat_template
-                and "{%- if add_generation_prompt %}" not in new_chat_template
-            ):
-                raise RuntimeError(
-                    f"Unsloth: The tokenizer `{tokenizer.name_or_path}`\n"
-                    "does not have a {% if add_generation_prompt %} for generation purposes.\n"
-                    f"Please file a bug report to the maintainers of `{tokenizer.name_or_path}` - thanks!"
-                )
+    # List-of-dicts form (older HF multi-template style).
+    if isinstance(chat_template, list):
+        fixed = []
+        for item in chat_template:
+            if not isinstance(item, dict) or "template" not in item:
+                fixed.append(item)
+                continue
+            tmpl = item["template"]
+            if not isinstance(tmpl, str):
+                fixed.append(item)
+                continue
+            label = f"variant={item.get('name', '?')!r}"
+            proxy = _VariantTokenizerProxy(tokenizer, tmpl, variant_label = label)
+            new_tmpl = _fix_chat_template_for_tokenizer(proxy, tmpl)
+            if new_tmpl is tmpl or new_tmpl == tmpl:
+                fixed.append(item)
             else:
-                logger.warning_once(
-                    "Unsloth: We successfully patched the tokenizer to add a {% if add_generation_prompt %} to the chat_template.\n"
-                    f"This is not a bug, but please notify the maintainers of `{tokenizer.name_or_path}` - thanks!"
-                )
-                chat_template = new_chat_template
-        else:
-            raise RuntimeError(
-                f"Unsloth: The tokenizer `{tokenizer.name_or_path}`\n"
-                "has a {% if add_generation_prompt %} for generation purposes, but wasn't provided correctly.\n"
-                "Please file a bug report immediately - thanks!"
-            )
-    return chat_template
+                fixed.append({**item, "template": new_tmpl})
+        return fixed
+
+    return _fix_chat_template_for_tokenizer(tokenizer, chat_template)
 
 
 def check_tokenizer(

From 6764cb9b90c7d9e2e0e170ce70f841d0bcf3d4ea Mon Sep 17 00:00:00 2001
From: Datta Nimmaturi <venkatadattasainimmaturi@gmail.com>
Date: Thu, 16 Apr 2026 19:30:17 +0530
Subject: [PATCH 120/216] Restrict flash attn to <=256 head dim. Consolidate
 attn impl checks (#5051)

* Restrict flash attn to <=256 head dim. Consolidate attn impl checks

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

* Consolidate the changes into single function

* safeguard for dict instead of object

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

---------

Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com>
---
 unsloth/models/_utils.py               | 266 ++++++++++++++++++++-----
 unsloth/models/llama.py                |   2 +-
 unsloth/models/loader.py               |   3 -
 unsloth/models/sentence_transformer.py |  41 ++--
 unsloth/models/vision.py               |  61 +-----
 5 files changed, 239 insertions(+), 134 deletions(-)

diff --git a/unsloth/models/_utils.py b/unsloth/models/_utils.py
index 1e0b015c44..820a111a9b 100644
--- a/unsloth/models/_utils.py
+++ b/unsloth/models/_utils.py
@@ -65,7 +65,9 @@ __all__ = [
     "patch_compiled_autograd",
     "process_vision_info",
     "unsloth_compile_transformers",
-    "determine_attention_implementation",
+    "resolve_model_class",
+    "resolve_attention_implementation",
+    "resolve_encoder_attention_implementation",
     "_set_attn_impl",
     "patch_fast_lora",
     "validate_loftq_config",
@@ -233,7 +235,7 @@ def apply_unsloth_gradient_checkpointing(
 # access on some GPU architectures (B200). Falls back to eager safely.
 _FLEX_EXCLUDED_MODELS = ("gpt_oss", "mllama", "nemotron_h", "modernbert")
 _EAGER_ONLY_PREFIXES = ("gemma3n",)
-_FLASH_ATTENTION_DISABLED_MODELS = ("gemma4", "gemma4_text")
+_FLASH_ATTENTION_MAX_HEAD_DIM = 256
 _FLASH_ATTENTION_DISABLED_WARNED = set()
 
 
@@ -245,8 +247,102 @@ def _is_eager_only(model_type):
     return any(model_type.startswith(p) for p in _EAGER_ONLY_PREFIXES)
 
 
-def _is_flash_attention_disabled(model_type):
-    return model_type in _FLASH_ATTENTION_DISABLED_MODELS
+def _config_items(config):
+    if isinstance(config, dict):
+        return config.items()
+    if hasattr(config, "__dict__"):
+        return vars(config).items()
+    return ()
+
+
+def _config_get(config, field_name, default = None):
+    if isinstance(config, dict):
+        return config.get(field_name, default)
+    return getattr(config, field_name, default)
+
+
+def _config_set(config, field_name, value):
+    if isinstance(config, dict):
+        config[field_name] = value
+    elif config is not None:
+        setattr(config, field_name, value)
+
+
+def _iter_attention_configs(config, seen = None):
+    if config is None or (
+        not isinstance(config, dict) and not hasattr(config, "__dict__")
+    ):
+        return
+    if seen is None:
+        seen = set()
+    config_id = id(config)
+    if config_id in seen:
+        return
+    seen.add(config_id)
+    yield config
+
+    for field_name, child_config in _config_items(config):
+        if not isinstance(field_name, str) or not field_name.endswith("_config"):
+            continue
+        if isinstance(child_config, dict) or hasattr(child_config, "__dict__"):
+            yield from _iter_attention_configs(child_config, seen)
+
+
+def _collect_attention_head_dims(config):
+    explicit_head_dims = []
+
+    for field_name in (
+        "head_dim",
+        "global_head_dim",
+        "local_head_dim",
+        "kv_head_dim",
+    ):
+        value = _config_get(config, field_name, None)
+        if isinstance(value, int) and value > 0:
+            explicit_head_dims.append(value)
+
+    if len(explicit_head_dims) != 0:
+        return explicit_head_dims
+
+    head_dims = []
+
+    hidden_size_names = ("hidden_size", "d_model", "embed_dim", "dim")
+    num_heads_names = ("num_attention_heads", "num_heads", "n_heads")
+    for hidden_size_name in hidden_size_names:
+        hidden_size = _config_get(config, hidden_size_name, None)
+        if not isinstance(hidden_size, int) or hidden_size <= 0:
+            continue
+        for num_heads_name in num_heads_names:
+            num_heads = _config_get(config, num_heads_name, None)
+            if (
+                isinstance(num_heads, int)
+                and num_heads > 0
+                and (hidden_size % num_heads) == 0
+            ):
+                head_dims.append(hidden_size // num_heads)
+
+    return head_dims
+
+
+def _get_max_attention_head_dim(config):
+    head_dims = []
+    for attention_config in _iter_attention_configs(config):
+        head_dims.extend(_collect_attention_head_dims(attention_config))
+    return max(head_dims) if len(head_dims) != 0 else None
+
+
+def _get_flash_attention_disable_reason(config):
+    max_head_dim = _get_max_attention_head_dim(config)
+    if max_head_dim is not None and max_head_dim > _FLASH_ATTENTION_MAX_HEAD_DIM:
+        return (
+            f"max attention head dim {max_head_dim} exceeds the Flash Attention 2 "
+            f"limit of {_FLASH_ATTENTION_MAX_HEAD_DIM}"
+        )
+    return None
+
+
+def _is_flash_attention_disabled(config):
+    return _get_flash_attention_disable_reason(config) is not None
 
 
 def _is_flash_attention_requested(attn_implementation):
@@ -256,20 +352,24 @@ def _is_flash_attention_requested(attn_implementation):
 
 
 def _disable_flash_attention_if_needed(
-    model_type,
     config,
     attn_implementation = None,
     supports_sdpa = False,
     would_use_flash_attention = False,
+    disable_reason = None,
 ):
-    if not _is_flash_attention_disabled(model_type):
+    if disable_reason is None:
+        disable_reason = _get_flash_attention_disable_reason(config)
+    if disable_reason is None:
         return attn_implementation
 
     requested_attn_implementation = attn_implementation
     if requested_attn_implementation is None:
-        requested_attn_implementation = getattr(config, "_attn_implementation", None)
+        requested_attn_implementation = _config_get(
+            config, "_attn_implementation", None
+        )
     if requested_attn_implementation is None:
-        requested_attn_implementation = getattr(config, "attn_implementation", None)
+        requested_attn_implementation = _config_get(config, "attn_implementation", None)
 
     if requested_attn_implementation == "eager":
         return _set_attn_impl(config, "eager")
@@ -284,16 +384,18 @@ def _disable_flash_attention_if_needed(
             if _is_flash_attention_requested(requested_attn_implementation)
             else "flash_attention_2"
         )
+        model_type = _config_get(config, "model_type", "")
         warning_key = (
             model_type,
             logged_attn_implementation,
             fallback_attn_implementation,
+            disable_reason,
         )
         if warning_key not in _FLASH_ATTENTION_DISABLED_WARNED:
             _FLASH_ATTENTION_DISABLED_WARNED.add(warning_key)
             print(
                 f"Unsloth: `{logged_attn_implementation}` is not supported "
-                "for Gemma 4 - "
+                f"for `{model_type}` because {disable_reason} - "
                 f"defaulting to `{fallback_attn_implementation}`."
             )
 
@@ -301,69 +403,125 @@ def _disable_flash_attention_if_needed(
 
 
 def _set_attn_impl(config, impl):
-    """Helper function to set attention implementation on config and return it."""
     if config is not None:
-        setattr(config, "_attn_implementation", impl)
-        if hasattr(config, "attn_implementation"):
-            setattr(config, "attn_implementation", impl)
+        _config_set(config, "_attn_implementation", impl)
+        if isinstance(config, dict) or hasattr(config, "attn_implementation"):
+            _config_set(config, "attn_implementation", impl)
     return impl
 
 
-def determine_attention_implementation(model_class, config):
-    model_type = getattr(config, "model_type", "").lower()
+def resolve_model_class(auto_model, config):
+    mapping = getattr(auto_model, "_model_mapping", {})
+    try:
+        result = mapping[config.__class__]
+    except Exception:
+        for config_class, model_class in mapping.items():
+            if isinstance(config, config_class):
+                result = model_class
+                break
+        else:
+            return None
 
-    # Eager-only models (e.g. gemma3n timm vision towers)
-    if _is_eager_only(model_type):
-        _set_attn_impl(config, "eager")
-        return "eager"
+    return result[0] if isinstance(result, (list, tuple)) else result
 
-    # Models with known Flash Attention incompatibilities. Gemma 4 full-attention
-    # layers use global_head_dim=512, which exceeds Flash Attention's dense
-    # head-dim support. Keep explicit eager requests, otherwise prefer SDPA.
-    if _is_flash_attention_disabled(model_type):
+
+def resolve_attention_implementation(
+    model_class,
+    config,
+    requested_attn_implementation = None,
+    supports_sdpa = None,
+):
+    model_type_name = _config_get(config, "model_type", "")
+    model_type = model_type_name.lower()
+    if supports_sdpa is None:
         supports_sdpa = model_class is not None and getattr(
             model_class, "_supports_sdpa", False
         )
-        return _disable_flash_attention_if_needed(
-            model_type,
+    supports_flash_attention = model_class is not None and (
+        getattr(model_class, "_supports_flash_attn_2", False)
+        or getattr(model_class, "_supports_flash_attn", False)
+    )
+    disable_reason = _get_flash_attention_disable_reason(config)
+    flash_attention_disabled = disable_reason is not None
+
+    if model_class is None:
+        attn_impl = _set_attn_impl(config, "sdpa" if supports_sdpa else "eager")
+    else:
+        if _is_eager_only(model_type):
+            attn_impl = _set_attn_impl(config, "eager")
+        elif flash_attention_disabled:
+            attn_impl = _disable_flash_attention_if_needed(
+                config,
+                supports_sdpa = supports_sdpa,
+                would_use_flash_attention = (
+                    HAS_FLASH_ATTENTION and supports_flash_attention
+                ),
+                disable_reason = disable_reason,
+            )
+        elif HAS_FLASH_ATTENTION and supports_flash_attention:
+            attn_impl = _set_attn_impl(config, "flash_attention_2")
+        elif supports_sdpa:
+            attn_impl = _set_attn_impl(config, "sdpa")
+        else:
+            attn_impl = "eager"
+            if os.environ.get("UNSLOTH_ENABLE_FLEX_ATTENTION", "1") != "0":
+                try:
+                    from transformers.utils.import_utils import (
+                        is_torch_flex_attn_available,
+                    )
+
+                    if (
+                        is_torch_flex_attn_available()
+                        and getattr(model_class, "_supports_flex_attn", False)
+                        and not _is_flex_excluded(model_type)
+                    ):
+                        attention_dropout = (
+                            _config_get(config, "attention_dropout", 0) or 0
+                        )
+                        if attention_dropout == 0:
+                            attn_impl = _set_attn_impl(config, "flex_attention")
+                except Exception:
+                    pass
+            if attn_impl == "eager":
+                attn_impl = _set_attn_impl(config, "eager")
+
+    if requested_attn_implementation is None:
+        final_attn_impl = attn_impl
+    elif flash_attention_disabled:
+        final_attn_impl = _disable_flash_attention_if_needed(
             config,
+            requested_attn_implementation,
             supports_sdpa = supports_sdpa,
+            disable_reason = disable_reason,
         )
+    else:
+        final_attn_impl = requested_attn_implementation
+        _set_attn_impl(config, final_attn_impl)
 
-    # Flash Attention 2
-    if HAS_FLASH_ATTENTION and model_class is not None:
-        supports_fa2 = getattr(model_class, "_supports_flash_attn_2", False) or getattr(
-            model_class, "_supports_flash_attn", False
+    if not supports_sdpa and final_attn_impl == "sdpa":
+        print(
+            f"Unsloth: {(model_type_name or 'model').title()} does not support SDPA - switching to fast eager."
         )
-        if supports_fa2:
-            _set_attn_impl(config, "flash_attention_2")
-            return "flash_attention_2"
+        final_attn_impl = _set_attn_impl(config, "eager")
 
-    # Flex Attention
-    if os.environ.get("UNSLOTH_ENABLE_FLEX_ATTENTION", "1") != "0":
-        try:
-            from transformers.utils.import_utils import is_torch_flex_attn_available
+    return final_attn_impl
 
-            if (
-                is_torch_flex_attn_available()
-                and model_class is not None
-                and getattr(model_class, "_supports_flex_attn", False)
-                and not _is_flex_excluded(model_type)
-            ):
-                attention_dropout = getattr(config, "attention_dropout", 0) or 0
-                if attention_dropout == 0:
-                    _set_attn_impl(config, "flex_attention")
-                    return "flex_attention"
-        except Exception:
-            pass
 
-    # SDPA
-    if model_class is not None and getattr(model_class, "_supports_sdpa", False):
-        _set_attn_impl(config, "sdpa")
+def resolve_encoder_attention_implementation(
+    auto_model,
+    config,
+    model_type = "",
+    disable_sdpa_model_names = (),
+):
+    model_class = resolve_model_class(auto_model, config)
+    supports_sdpa = model_class is not None and getattr(
+        model_class, "_supports_sdpa", False
+    )
+    if any(name in model_type.lower() for name in disable_sdpa_model_names):
+        return "eager"
+    if supports_sdpa:
         return "sdpa"
-
-    _set_attn_impl(config, "eager")
-    return "eager"
+    return None
 
 
 def _run_temporary_patches(phase):
diff --git a/unsloth/models/llama.py b/unsloth/models/llama.py
index 425df1c084..999711efdb 100644
--- a/unsloth/models/llama.py
+++ b/unsloth/models/llama.py
@@ -2346,7 +2346,7 @@ class FastLlamaModel:
         model_function = MODEL_FOR_CAUSAL_LM_MAPPING[model_config.__class__]
         IS_FALCON_H1 = model_config.model_type.startswith("falcon_h1")
 
-        preferred_attn_impl = determine_attention_implementation(
+        preferred_attn_impl = resolve_attention_implementation(
             model_function, model_config
         )
 
diff --git a/unsloth/models/loader.py b/unsloth/models/loader.py
index cd12544ae9..fc91178d88 100644
--- a/unsloth/models/loader.py
+++ b/unsloth/models/loader.py
@@ -1151,9 +1151,6 @@ class FastModel(FastBaseModel):
                 )
             os.environ["UNSLOTH_DISABLE_STATIC_GENERATION"] = "1"
             os.environ["UNSLOTH_HIGH_PRECISION_LAYERNORM"] = "1"
-            # Disable flex_attention for Gemma-4: flex compile overhead is 2.7x slower
-            # than SDPA. Our attention patch ensures Q/K/V dtype alignment for SDPA.
-            os.environ["UNSLOTH_ENABLE_FLEX_ATTENTION"] = "0"
         # Gemma 3N must be before Gemma 3
         elif "gemma3n" in model_types_all:
             if transformers_version < Version("4.53.0"):
diff --git a/unsloth/models/sentence_transformer.py b/unsloth/models/sentence_transformer.py
index ad59165a50..541875a3fc 100644
--- a/unsloth/models/sentence_transformer.py
+++ b/unsloth/models/sentence_transformer.py
@@ -15,7 +15,11 @@
 import logging
 
 from .loader import FastModel, DISABLE_SDPA_MODEL_NAMES
-from ._utils import SUPPORTS_BFLOAT16
+from ._utils import (
+    SUPPORTS_BFLOAT16,
+    resolve_model_class,
+    resolve_encoder_attention_implementation,
+)
 import inspect
 import json
 import os
@@ -31,7 +35,6 @@ import transformers
 from packaging.version import Version
 import re
 from transformers import AutoModel, AutoConfig
-from transformers.models.auto.auto_factory import _get_model_class
 import tempfile
 from huggingface_hub import HfApi, get_token
 from ..save import unsloth_save_pretrained_torchao, unsloth_save_pretrained_gguf
@@ -870,7 +873,7 @@ class FastSentenceTransformer(FastModel):
             if auto_model_class is None:
                 auto_model_class = AutoModel
             # try to resolve the class
-            model_class = _get_model_class(config, auto_model_class._model_mapping)
+            model_class = resolve_model_class(auto_model_class, config)
 
             if model_class:
                 sig = inspect.signature(model_class.__init__)
@@ -1446,32 +1449,18 @@ class FastSentenceTransformer(FastModel):
             ):
                 st_device = "cuda"
 
-            # Check if model supports SDPA (Scaled Dot Product Attention) for extra speedup
-            supports_sdpa = False
-            if config is not None:
-                try:
-                    model_class = _get_model_class(
-                        config, kwargs.get("auto_model", AutoModel)._model_mapping
-                    )
-                    supports_sdpa = getattr(model_class, "_supports_sdpa", False)
-                except:
-                    pass
-
             # Build model_kwargs for SentenceTransformer
             model_kwargs = {"torch_dtype": dtype}
 
-            # Enable SDPA if supported (1.2x extra speedup on top of torch.compile)
-            # But disable for models with known SDPA + torch.compile backward issues
-            _force_eager = False
-            for _sdpa_model in DISABLE_SDPA_MODEL_NAMES:
-                if _sdpa_model in model_type.lower():
-                    supports_sdpa = False
-                    _force_eager = True
-                    break
-            if supports_sdpa:
-                model_kwargs["attn_implementation"] = "sdpa"
-            elif _force_eager:
-                model_kwargs["attn_implementation"] = "eager"
+            encoder_attn_impl = resolve_encoder_attention_implementation(
+                kwargs.get("auto_model", AutoModel),
+                config,
+                model_type = model_type,
+                disable_sdpa_model_names = DISABLE_SDPA_MODEL_NAMES,
+            )
+            supports_sdpa = encoder_attn_impl == "sdpa"
+            if encoder_attn_impl is not None:
+                model_kwargs["attn_implementation"] = encoder_attn_impl
 
             # Print optimization status
             sdpa_str = " + SDPA" if supports_sdpa else ""
diff --git a/unsloth/models/vision.py b/unsloth/models/vision.py
index e31617f89a..90c93ea3f7 100644
--- a/unsloth/models/vision.py
+++ b/unsloth/models/vision.py
@@ -33,8 +33,8 @@ from ._utils import (
     __version__,
     importlib_version,
     _prepare_model_for_qat,
-    _is_flash_attention_disabled,
-    _disable_flash_attention_if_needed,
+    resolve_model_class,
+    resolve_attention_implementation,
 )
 from ._utils import *
 from .loader_utils import _get_fp8_mode_and_check_settings
@@ -613,55 +613,18 @@ class FastBaseModel:
                 token = token,
                 trust_remote_code = trust_remote_code,
             )
-        user_attn_implementation = kwargs.get("attn_implementation", None)
-        try:
-            model_class = auto_model._model_mapping[auto_config.__class__]
-        except Exception:
-            model_class = None
-        if model_class is None:
-            # When model_class cannot be resolved (remote-code or unmapped
-            # configs), preserve the old fallback of sdpa when supported.
-            attn_impl = _set_attn_impl(
-                auto_config, "sdpa" if supports_sdpa else "eager"
-            )
-        else:
-            attn_impl = determine_attention_implementation(model_class, auto_config)
+        model_class = resolve_model_class(auto_model, auto_config)
+        attn_impl = resolve_attention_implementation(
+            model_class,
+            auto_config,
+            requested_attn_implementation = kwargs.get("attn_implementation", None),
+            supports_sdpa = supports_sdpa,
+        )
 
         # Handle FP8 models: get_model_name has already redirected this to BF16 sibling if the model ships with
         # FP8 weights. We just need to update it here for sanity.
         auto_config.model_name = model_name
-        # Re-resolve model_class after potential config change
-        try:
-            model_class = auto_model._model_mapping[auto_config.__class__]
-        except Exception:
-            model_class = None
-
-        if not ("attn_implementation" in kwargs):
-            kwargs["attn_implementation"] = attn_impl
-        model_type = getattr(auto_config, "model_type", "").lower()
-        if _is_flash_attention_disabled(model_type):
-            supports_fa2 = model_class is not None and (
-                getattr(model_class, "_supports_flash_attn_2", False)
-                or getattr(model_class, "_supports_flash_attn", False)
-            )
-            kwargs["attn_implementation"] = _disable_flash_attention_if_needed(
-                model_type,
-                auto_config,
-                kwargs.get("attn_implementation"),
-                supports_sdpa = supports_sdpa,
-                would_use_flash_attention = (
-                    user_attn_implementation is None
-                    and HAS_FLASH_ATTENTION
-                    and supports_fa2
-                ),
-            )
-        if not supports_sdpa and kwargs.get("attn_implementation") == "sdpa":
-            print(
-                f"Unsloth: {model_type_arch.title()} does not support SDPA - switching to fast eager."
-            )
-            del kwargs["attn_implementation"]
-            # Re-stamp config so it stays consistent with the actual impl
-            _set_attn_impl(auto_config, "eager")
+        kwargs["attn_implementation"] = attn_impl
 
         bnb_config = None
         user_quantization_config = kwargs.get("quantization_config", None)
@@ -804,9 +767,7 @@ class FastBaseModel:
                 token = token,
                 trust_remote_code = trust_remote_code,
             )
-        setattr(auto_config, "_attn_implementation", config_attn_impl)
-        if hasattr(auto_config, "attn_implementation"):
-            setattr(auto_config, "attn_implementation", config_attn_impl)
+        _set_attn_impl(auto_config, config_attn_impl)
         model_config = auto_config
 
         verify_fp8_support_if_applicable(model_config)

From cae4a742974b09b02a3d8b140b77bc330651ef23 Mon Sep 17 00:00:00 2001
From: Daniel Han <danielhanchen@gmail.com>
Date: Thu, 16 Apr 2026 14:18:57 +0000
Subject: [PATCH 121/216] Remove legacy venv Scripts entry from User PATH on
 upgrade

Older installers persisted the venv Scripts directory directly in the
User PATH registry. The shim approach (added in this PR) no longer writes
that entry, but it also did not remove the old one. On upgrade, the
legacy entry survived and python.exe / pip.exe from the unsloth venv
continued winning resolution in every new shell, which is exactly the
hijack the shim was designed to prevent.

Before creating the shim, read the current User PATH, filter out any
entry matching $VenvDir\Scripts (using the same symmetric raw+expanded
comparison as Add-ToUserPath), and write back if changed. This runs
once per install and is a no-op on fresh installs where the legacy
entry was never written.
---
 install.ps1 | 35 +++++++++++++++++++++++++++++++++++
 1 file changed, 35 insertions(+)

diff --git a/install.ps1 b/install.ps1
index dfb54b1e87..e6cf8f993a 100644
--- a/install.ps1
+++ b/install.ps1
@@ -1042,6 +1042,41 @@ shell.Run cmd, 0, False
     # We do NOT add the venv Scripts dir to PATH (it also holds python.exe
     # and pip.exe, which would hijack the user's system interpreter).
     # Hardlink preferred; falls back to copy if cross-volume or non-NTFS.
+    #
+    # Clean up the legacy venv Scripts PATH entry that older installers wrote.
+    # Without this, upgrade users keep python/pip hijacked even after the shim
+    # approach is in place.
+    $LegacyScriptsDir = Join-Path $VenvDir "Scripts"
+    try {
+        $legacyKey = [Microsoft.Win32.Registry]::CurrentUser.CreateSubKey('Environment')
+        try {
+            $rawPath = $legacyKey.GetValue('Path', '', [Microsoft.Win32.RegistryValueOptions]::DoNotExpandEnvironmentNames)
+            if ($rawPath) {
+                [string[]]$pathEntries = $rawPath -split ';'
+                $normalLegacy = $LegacyScriptsDir.Trim().Trim('"').TrimEnd('\').ToLowerInvariant()
+                $expNormalLegacy = [Environment]::ExpandEnvironmentVariables($LegacyScriptsDir).Trim().Trim('"').TrimEnd('\').ToLowerInvariant()
+                $filtered = @($pathEntries | Where-Object {
+                    $stripped = $_.Trim().Trim('"')
+                    $rawNorm = $stripped.TrimEnd('\').ToLowerInvariant()
+                    $expNorm = [Environment]::ExpandEnvironmentVariables($stripped).TrimEnd('\').ToLowerInvariant()
+                    ($rawNorm -ne $normalLegacy -and $rawNorm -ne $expNormalLegacy) -and
+                    ($expNorm -ne $normalLegacy -and $expNorm -ne $expNormalLegacy)
+                })
+                $cleanedPath = $filtered -join ';'
+                if ($cleanedPath -ne $rawPath) {
+                    $legacyKey.SetValue('Path', $cleanedPath, [Microsoft.Win32.RegistryValueKind]::ExpandString)
+                    # Broadcast so other processes see the removal immediately.
+                    try {
+                        $d = "UnslothPathRefresh_$([guid]::NewGuid().ToString('N').Substring(0,8))"
+                        [Environment]::SetEnvironmentVariable($d, '1', 'User')
+                        [Environment]::SetEnvironmentVariable($d, [NullString]::Value, 'User')
+                    } catch { }
+                }
+            }
+        } finally {
+            $legacyKey.Close()
+        }
+    } catch { }
     $ShimDir = Join-Path $StudioHome "bin"
     New-Item -ItemType Directory -Force -Path $ShimDir | Out-Null
     $ShimExe = Join-Path $ShimDir "unsloth.exe"

From 5b8643969ef753a7e58afc944a0e98fc572689a9 Mon Sep 17 00:00:00 2001
From: Daniel Han <danielhanchen@gmail.com>
Date: Thu, 16 Apr 2026 14:20:43 +0000
Subject: [PATCH 122/216] Revert "Remove legacy venv Scripts entry from User
 PATH on upgrade"

This reverts commit cae4a742974b09b02a3d8b140b77bc330651ef23.
---
 install.ps1 | 35 -----------------------------------
 1 file changed, 35 deletions(-)

diff --git a/install.ps1 b/install.ps1
index e6cf8f993a..dfb54b1e87 100644
--- a/install.ps1
+++ b/install.ps1
@@ -1042,41 +1042,6 @@ shell.Run cmd, 0, False
     # We do NOT add the venv Scripts dir to PATH (it also holds python.exe
     # and pip.exe, which would hijack the user's system interpreter).
     # Hardlink preferred; falls back to copy if cross-volume or non-NTFS.
-    #
-    # Clean up the legacy venv Scripts PATH entry that older installers wrote.
-    # Without this, upgrade users keep python/pip hijacked even after the shim
-    # approach is in place.
-    $LegacyScriptsDir = Join-Path $VenvDir "Scripts"
-    try {
-        $legacyKey = [Microsoft.Win32.Registry]::CurrentUser.CreateSubKey('Environment')
-        try {
-            $rawPath = $legacyKey.GetValue('Path', '', [Microsoft.Win32.RegistryValueOptions]::DoNotExpandEnvironmentNames)
-            if ($rawPath) {
-                [string[]]$pathEntries = $rawPath -split ';'
-                $normalLegacy = $LegacyScriptsDir.Trim().Trim('"').TrimEnd('\').ToLowerInvariant()
-                $expNormalLegacy = [Environment]::ExpandEnvironmentVariables($LegacyScriptsDir).Trim().Trim('"').TrimEnd('\').ToLowerInvariant()
-                $filtered = @($pathEntries | Where-Object {
-                    $stripped = $_.Trim().Trim('"')
-                    $rawNorm = $stripped.TrimEnd('\').ToLowerInvariant()
-                    $expNorm = [Environment]::ExpandEnvironmentVariables($stripped).TrimEnd('\').ToLowerInvariant()
-                    ($rawNorm -ne $normalLegacy -and $rawNorm -ne $expNormalLegacy) -and
-                    ($expNorm -ne $normalLegacy -and $expNorm -ne $expNormalLegacy)
-                })
-                $cleanedPath = $filtered -join ';'
-                if ($cleanedPath -ne $rawPath) {
-                    $legacyKey.SetValue('Path', $cleanedPath, [Microsoft.Win32.RegistryValueKind]::ExpandString)
-                    # Broadcast so other processes see the removal immediately.
-                    try {
-                        $d = "UnslothPathRefresh_$([guid]::NewGuid().ToString('N').Substring(0,8))"
-                        [Environment]::SetEnvironmentVariable($d, '1', 'User')
-                        [Environment]::SetEnvironmentVariable($d, [NullString]::Value, 'User')
-                    } catch { }
-                }
-            }
-        } finally {
-            $legacyKey.Close()
-        }
-    } catch { }
     $ShimDir = Join-Path $StudioHome "bin"
     New-Item -ItemType Directory -Force -Path $ShimDir | Out-Null
     $ShimExe = Join-Path $ShimDir "unsloth.exe"

From b42e3a120d15eb5c9378b4ef89e2bbe1e26ce3b0 Mon Sep 17 00:00:00 2001
From: Daniel Han <danielhanchen@gmail.com>
Date: Thu, 16 Apr 2026 07:36:59 -0700
Subject: [PATCH 123/216] Remove legacy venv Scripts entry from User PATH on
 upgrade (#5060)

Older installers persisted the venv Scripts directory directly in the
User PATH registry. The shim approach from #4961 no longer writes that
entry, but on upgrade the old one survived and python.exe / pip.exe
from the unsloth venv continued winning resolution in every new shell.

Before creating the shim, read the current User PATH, filter out any
entry matching $VenvDir\Scripts (using the same symmetric raw+expanded
comparison as Add-ToUserPath), and write back if changed. No-op on
fresh installs where the legacy entry was never written.

Confirmed on a real Windows machine: `where.exe python` was returning
the venv interpreter first even after the shim PR merged.
---
 install.ps1 | 32 ++++++++++++++++++++++++++++++++
 1 file changed, 32 insertions(+)

diff --git a/install.ps1 b/install.ps1
index dfb54b1e87..3fc9ac4690 100644
--- a/install.ps1
+++ b/install.ps1
@@ -1042,6 +1042,38 @@ shell.Run cmd, 0, False
     # We do NOT add the venv Scripts dir to PATH (it also holds python.exe
     # and pip.exe, which would hijack the user's system interpreter).
     # Hardlink preferred; falls back to copy if cross-volume or non-NTFS.
+    #
+    # Remove the legacy venv Scripts PATH entry that older installers wrote.
+    $LegacyScriptsDir = Join-Path $VenvDir "Scripts"
+    try {
+        $legacyKey = [Microsoft.Win32.Registry]::CurrentUser.CreateSubKey('Environment')
+        try {
+            $rawPath = $legacyKey.GetValue('Path', '', [Microsoft.Win32.RegistryValueOptions]::DoNotExpandEnvironmentNames)
+            if ($rawPath) {
+                [string[]]$pathEntries = $rawPath -split ';'
+                $normalLegacy = $LegacyScriptsDir.Trim().Trim('"').TrimEnd('\').ToLowerInvariant()
+                $expNormalLegacy = [Environment]::ExpandEnvironmentVariables($LegacyScriptsDir).Trim().Trim('"').TrimEnd('\').ToLowerInvariant()
+                $filtered = @($pathEntries | Where-Object {
+                    $stripped = $_.Trim().Trim('"')
+                    $rawNorm = $stripped.TrimEnd('\').ToLowerInvariant()
+                    $expNorm = [Environment]::ExpandEnvironmentVariables($stripped).TrimEnd('\').ToLowerInvariant()
+                    ($rawNorm -ne $normalLegacy -and $rawNorm -ne $expNormalLegacy) -and
+                    ($expNorm -ne $normalLegacy -and $expNorm -ne $expNormalLegacy)
+                })
+                $cleanedPath = $filtered -join ';'
+                if ($cleanedPath -ne $rawPath) {
+                    $legacyKey.SetValue('Path', $cleanedPath, [Microsoft.Win32.RegistryValueKind]::ExpandString)
+                    try {
+                        $d = "UnslothPathRefresh_$([guid]::NewGuid().ToString('N').Substring(0,8))"
+                        [Environment]::SetEnvironmentVariable($d, '1', 'User')
+                        [Environment]::SetEnvironmentVariable($d, [NullString]::Value, 'User')
+                    } catch { }
+                }
+            }
+        } finally {
+            $legacyKey.Close()
+        }
+    } catch { }
     $ShimDir = Join-Path $StudioHome "bin"
     New-Item -ItemType Directory -Force -Path $ShimDir | Out-Null
     $ShimExe = Join-Path $ShimDir "unsloth.exe"

From ff23ce40b4f511bca0e997be02c6121a6e8d72ea Mon Sep 17 00:00:00 2001
From: Daniel Han <danielhanchen@gmail.com>
Date: Thu, 16 Apr 2026 08:02:05 -0700
Subject: [PATCH 124/216] Fix review findings for chat-template repair (#5049)
 (#5056)

* Fix review findings for PR #49

1. Sandbox fallback Jinja env in _VariantTokenizerProxy.apply_chat_template
   (use SandboxedEnvironment, matching _derive_assistant_prefix_by_render)
2. Unwrap benign outer-If guards in _template_ends_with_toplevel_for so
   templates like {% if messages %}{% for ... %}{% endfor %}{% endif %}
   are still repairable (preserves Qwen3-Guard rejection via else-branch
   and add_generation_prompt-name checks)
3. Preserve raw name_or_path in _VariantTokenizerProxy._source_path so
   local-path detection works for dict/list variant tokenizers
4. Context-aware strict-mode messages: omit "will still load" and
   "Set UNSLOTH_STRICT_CHAT_TEMPLATE=1" when already raising

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

---------

Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com>
---
 unsloth/tokenizer_utils.py | 122 +++++++++++++++++++++++++++----------
 1 file changed, 91 insertions(+), 31 deletions(-)

diff --git a/unsloth/tokenizer_utils.py b/unsloth/tokenizer_utils.py
index 2fed54dc01..85b3c5308b 100644
--- a/unsloth/tokenizer_utils.py
+++ b/unsloth/tokenizer_utils.py
@@ -669,11 +669,11 @@ def _find_end_position(template, endfor = None, endif = None):
 def _template_ends_with_toplevel_for(chat_template):
     """Return True if the last structural node at the template's top level is
     a For (message-iteration) loop, ignoring trailing pure-whitespace Output
-    nodes. Used to gate the GH#4150 ChatML repair: if the outermost structure
-    is something else (e.g. an outer If that wraps the whole template, as in
-    Qwen3-Guard), we shouldn't inject an {% if add_generation_prompt %}
-    block at the end -- it would land inside or after an unrelated control
-    structure."""
+    nodes. Unwraps benign outer-If guards (no else branch, not testing
+    add_generation_prompt) so that templates like
+    ``{% if messages %}{% for ... %}{% endfor %}{% endif %}`` are still
+    repairable. Rejects real structural wrappers (e.g. Qwen3-Guard with
+    else branches)."""
     try:
         import jinja2
         import jinja2.nodes
@@ -681,20 +681,31 @@ def _template_ends_with_toplevel_for(chat_template):
         ast = jinja2.Environment().parse(chat_template)
     except Exception:
         return False
-    for node in reversed(ast.body):
-        # Skip trailing output nodes that are only whitespace -- they come
-        # from trailing whitespace/newlines in the source, not from real
-        # message-rendering logic.
-        if isinstance(node, jinja2.nodes.Output):
-            only_ws = all(
-                isinstance(child, jinja2.nodes.TemplateData)
-                and child.data.strip() == ""
-                for child in node.nodes
-            )
-            if only_ws:
-                continue
-        return isinstance(node, jinja2.nodes.For)
-    return False
+
+    def _last_structural(nodes):
+        for node in reversed(nodes):
+            if isinstance(node, jinja2.nodes.Output):
+                only_ws = all(
+                    isinstance(child, jinja2.nodes.TemplateData)
+                    and child.data.strip() == ""
+                    for child in node.nodes
+                )
+                if only_ws:
+                    continue
+            return node
+        return None
+
+    node = _last_structural(ast.body)
+    while isinstance(node, jinja2.nodes.If) and not node.else_:
+        names = []
+        if isinstance(node.test, jinja2.nodes.Name):
+            names.append(node.test)
+        names.extend(node.test.find_all(jinja2.nodes.Name))
+        if any(n.name == "add_generation_prompt" for n in names):
+            break
+        node = _last_structural(node.body)
+
+    return isinstance(node, jinja2.nodes.For)
 
 
 def _if_body_emits_content(if_node):
@@ -944,10 +955,18 @@ def _name_is_local_path(name_or_path):
         return False
 
 
-def _format_chat_template_message(name_or_path, repaired):
+def _format_chat_template_message(
+    name_or_path,
+    repaired,
+    has_generation_block = False,
+    local_path_source = None,
+    strict = False,
+):
     """Build a user-facing warning/error message that points at the right
     responsible party (user's downstream tool vs. upstream model maintainer)."""
-    local = _name_is_local_path(name_or_path)
+    local = _name_is_local_path(
+        local_path_source if local_path_source is not None else name_or_path
+    )
     if local:
         source_hint = (
             "This tokenizer was loaded from a local path. The likely cause is a "
@@ -961,19 +980,39 @@ def _format_chat_template_message(name_or_path, repaired):
             "The chat_template shipped with `{name}` appears incomplete. "
             "Consider filing a bug report with the model maintainers."
         ).format(name = name_or_path)
+    strict_suffix = (
+        ""
+        if strict
+        else (" Set UNSLOTH_STRICT_CHAT_TEMPLATE=1 to raise instead of warn.")
+    )
     if repaired:
         return (
             "Unsloth: Patched the chat_template on `{name}` to add a "
             "{{% if add_generation_prompt %}} block. {hint}"
         ).format(name = name_or_path, hint = source_hint)
+    if has_generation_block:
+        return (
+            "Unsloth: The tokenizer `{name}` has a "
+            "{{% if add_generation_prompt %}} block, but it does not change "
+            "the rendered output. {hint}{suffix}"
+        ).format(name = name_or_path, hint = source_hint, suffix = strict_suffix)
+    load_clause = (
+        "Loading is blocked in strict mode."
+        if strict
+        else "The model will still load, but "
+        "`apply_chat_template(add_generation_prompt=True)` may not produce a "
+        "correct assistant-turn marker."
+    )
     return (
         "Unsloth: The tokenizer `{name}` does not have a "
         "{{% if add_generation_prompt %}} block for generation purposes, and "
-        "automatic repair was not possible. The model will still load, but "
-        "`apply_chat_template(add_generation_prompt=True)` may not produce a "
-        "correct assistant-turn marker. {hint} Set "
-        "UNSLOTH_STRICT_CHAT_TEMPLATE=1 to raise instead of warn."
-    ).format(name = name_or_path, hint = source_hint)
+        "automatic repair was not possible. {load_clause} {hint}{suffix}"
+    ).format(
+        name = name_or_path,
+        load_clause = load_clause,
+        hint = source_hint,
+        suffix = strict_suffix,
+    )
 
 
 def _validate_patched_template(tokenizer, patched_template, is_sharegpt):
@@ -1041,6 +1080,7 @@ def _fix_chat_template_for_tokenizer(tokenizer, chat_template):
     UNSLOTH_STRICT_CHAT_TEMPLATE: warn + return original (default) or raise
     RuntimeError (strict)."""
     name = getattr(tokenizer, "name_or_path", "unknown")
+    source_path = getattr(tokenizer, "_source_path", name)
 
     # Detect ShareGPT vs HF style by probing apply_chat_template.
     is_sharegpt = None
@@ -1092,19 +1132,38 @@ def _fix_chat_template_for_tokenizer(tokenizer, chat_template):
     if _has_add_generation_prompt_block(chat_template):
         # Template has the block but it does not change output. This is the
         # "wasn't provided correctly" case from the pre-warn code path.
-        msg = _format_chat_template_message(name, repaired = False)
-        if _is_strict_chat_template_mode():
+        strict = _is_strict_chat_template_mode()
+        msg = _format_chat_template_message(
+            name,
+            repaired = False,
+            has_generation_block = True,
+            local_path_source = source_path,
+            strict = strict,
+        )
+        if strict:
             raise RuntimeError(msg)
         logger.warning_once(msg)
         return chat_template
 
     repaired = _repair_string_template(tokenizer, chat_template, is_sharegpt)
     if repaired is not None:
-        logger.warning_once(_format_chat_template_message(name, repaired = True))
+        logger.warning_once(
+            _format_chat_template_message(
+                name,
+                repaired = True,
+                local_path_source = source_path,
+            )
+        )
         return repaired
 
-    msg = _format_chat_template_message(name, repaired = False)
-    if _is_strict_chat_template_mode():
+    strict = _is_strict_chat_template_mode()
+    msg = _format_chat_template_message(
+        name,
+        repaired = False,
+        local_path_source = source_path,
+        strict = strict,
+    )
+    if strict:
         raise RuntimeError(msg)
     logger.warning_once(msg)
     return chat_template
@@ -1125,6 +1184,7 @@ class _VariantTokenizerProxy:
         self._base = base_tokenizer
         self._template = variant_template
         base_name = getattr(base_tokenizer, "name_or_path", "unknown")
+        self._source_path = base_name
         self.name_or_path = (
             f"{base_name} ({variant_label})" if variant_label else base_name
         )

From 05ec0f110bb8fd7e1872fc77a6900ee5a2b449f6 Mon Sep 17 00:00:00 2001
From: Daniel Han <danielhanchen@gmail.com>
Date: Thu, 16 Apr 2026 08:24:08 -0700
Subject: [PATCH 125/216] Studio: Ollama support, recommended folders, Custom
 Folders UX polish (#5050)

* Studio: Ollama support, recommended folders, Custom Folders UX polish

Backend:
- Add _scan_ollama_dir that reads manifests/registry.ollama.ai/library/*
  and creates .gguf symlinks under <ollama_dir>/.studio_links/ pointing
  at the content-addressable blobs, so detect_gguf_model and llama-server
  -m work unchanged for Ollama models
- Filter entries under .studio_links from the generic models/hf/lmstudio
  scanners to avoid duplicate rows and leaked internal paths in the UI
- New GET /api/models/recommended-folders endpoint returning LM Studio
  and Ollama model directories that currently exist on the machine
  (OLLAMA_MODELS env var + standard paths, ~/.lmstudio/models, legacy
  LM Studio cache), used by the Custom Folders quick-add chips
- detect_gguf_model now uses os.path.abspath instead of Path.resolve so
  the readable symlink name is preserved as display_name (e.g.
  qwen2.5-0.5b-Q4_K_M.gguf instead of sha256-abc...)
- llama-server failure with a path under .studio_links or .cache/ollama
  surfaces a friendlier message ("Some Ollama models do not work with
  llama.cpp. Try a different model, or use this model directly through
  Ollama instead.") instead of the generic validation error

Frontend:
- ListLabel supports an optional leading icon and collapse toggle; used
  for Downloaded (download icon), Custom Folders (folder icon), and
  Recommended (star icon)
- Custom Folders header gets folder icon on the left, and +, search,
  and chevron buttons on the right; chevron uses ml-auto so it aligns
  with the Downloaded and Recommended chevrons
- New recommended folder chips render below the registered scan folders
  when there are unregistered well-known paths; one click adds them as
  a scan folder
- Custom folder rows that are direct .gguf files (Ollama symlinks) load
  immediately via onSelect instead of opening the GGUF variant expander
  (which is for repos containing multiple quants, not single files)
- When loading a direct .gguf file path, send max_seq_length = 0 so the
  backend uses the model's native context instead of the 4096 chat
  default (qwen2.5:0.5b now loads at 32768 instead of 4096)
- New listRecommendedFolders() helper on the chat API

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

* Address review: log silent exceptions and support read-only Ollama dirs

Replace silent except blocks in _scan_ollama_dir and the
recommended-folders endpoint with narrower exception types plus debug
or warning logs, so failures are diagnosable without hiding signal.

Add _ollama_links_dir helper that falls back to a per-ollama-dir hashed
namespace under Studio's own cache (~/.unsloth/studio/cache/ollama_links)
when the Ollama models directory is read-only. Common for system installs
at /usr/share/ollama/.ollama/models and /var/lib/ollama/.ollama/models
where the Studio process has read but not write access. Previously the
scanner returned an empty list in that case and Ollama models would
silently not appear.

The fallback preserves the .gguf suffix on symlink names so
detect_gguf_model keeps recognising them. The prior "raw sha256 blob
path" fallback would have missed the suffix check and failed to load.

* Address review: detect mmproj next to symlink target for vision GGUFs

Codex P1 on model_config.py:1012: when detect_gguf_model returns the
symlink path (to preserve readable display names), detect_mmproj_file
searched the symlink's parent directory instead of the target's. For
vision GGUFs surfaced via Ollama's .studio_links/ -- where the weight
file is symlinked but any mmproj sidecar lives next to the real blob
-- mmproj was no longer detected, so the model was misclassified as
text-only and llama-server would start without --mmproj.

detect_mmproj_file now adds the resolved target's parent to the scan
order when path is a symlink. Direct (non-symlink) .gguf paths are
unchanged, so LM Studio and HF cache layouts keep working exactly as
before. Verified with a fake layout reproducing the bug plus a
regression check on a non-symlink LM Studio model.

* Address review: support all Ollama namespaces and vision projector layers

- Iterate over all directories under registry.ollama.ai/ instead of
  hardcoding the "library" namespace. Custom namespaces like
  "mradermacher/llama3" now get scanned and include the namespace
  prefix in display names, model IDs, and symlink names to avoid
  collisions.

- Create companion -mmproj.gguf symlinks for Ollama vision models
  that have an "application/vnd.ollama.image.projector" layer, so
  detect_mmproj_file can find the projector alongside the model.

- Extract symlink creation into _make_symlink helper to reduce
  duplication between model and projector paths.

* Address review: move imports to top level and add scan limit

- Move hashlib and json imports to the top of the file (PEP 8).
- Remove inline `import json as _json` and `import hashlib` from
  function bodies, use the top-level imports directly.
- Add `limit` parameter to `_scan_ollama_dir()` with early exit
  when the threshold is reached.
- Pass `_MAX_MODELS_PER_FOLDER` into the scanner so it stops
  traversing once enough models are found.

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

* Address review: Windows fallback, all registry hosts, collision safety

_make_link (formerly _make_symlink):
- Falls back to os.link() hardlink when symlink_to() fails (Windows
  without Developer Mode), then to shutil.copy2 as last resort
- Uses atomic os.replace via tmp file to avoid race window where the
  .gguf path is missing during rescan

Scanner now handles all Ollama registry layouts:
- Uses rglob over manifests/ instead of hardcoding registry.ollama.ai
- Discovers hf.co/org/repo:tag and any other host, not just library/
- Filenames include a stable sha1 hash of the manifest path to prevent
  collisions between models that normalize to the same stem

Per-model subdirectories under .studio_links/:
- Each model's links live in their own hash-keyed subdirectory
- detect_mmproj_file only sees the projector for that specific model,
  not siblings from other Ollama models

Friendly Ollama error detection:
- Now also matches ollama_links/ (the read-only fallback cache path)
  and model_identifier starting with "ollama/"

Recommended folders:
- Added os.access(R_OK | X_OK) check so unreadable system directories
  like /var/lib/ollama/.ollama/models are not advertised as chips

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

* Address review: filter ollama_links from generic scanners

The generic scanners (models_dir, hf_cache, lmstudio) already filter
out .studio_links to avoid duplicate Ollama entries, but missed the
ollama_links fallback cache directory used for read-only Ollama
installs. Add it to the filter.

* Address review: idempotent link creation and path-component filter

_make_link:
- Skip recreation when a valid link/copy already exists (samefile or
  matching size check). Prevents blocking the model-list API with
  multi-GB copies on repeated scans.
- Use uuid4 instead of os.getpid() for tmp file names to avoid race
  conditions from concurrent scans.
- Log cleanup errors instead of silently swallowing them.

Path filter:
- Use os.sep-bounded checks instead of bare substring match to avoid
  false positives on paths like "my.studio_links.backup/model.gguf".

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

* Address review: drop copy fallback, targeted glob, robust path filter

_make_link:
- Drop shutil.copy2 fallback -- copying multi-GB GGUFs inside a sync
  API request would block the backend. Log a warning and skip the
  model when both symlink and hardlink fail.

Scanner:
- Replace rglob("*") with targeted glob patterns (*/*/* and */*/*/*)
  to avoid traversing unrelated subdirectories in large custom folders.

Path filter:
- Use Path.parts membership check instead of os.sep substring matching
  for robustness across platforms.

Scan limit:
- Skip _scan_ollama_dir when _generic already fills the per-folder cap.

* Address review: sha256, top-level uuid import, Path.absolute()

- Switch hashlib.sha1 to hashlib.sha256 for path hashing consistency.
- Move uuid import to the top of the file instead of inside _make_link.
- Replace os.path.abspath with Path.absolute() in detect_gguf_model
  to match the pathlib style used throughout the codebase.

* Address review: fix stale comments (sha1, rglob, copy fallback)

Update three docstrings/comments that still referenced the old
implementation after recent changes:
- sha1 comment now says "not a security boundary" (no hash name)
- "rglob" -> "targeted glob patterns"
- "file copies as a last resort" -> removed (copy fallback was dropped)

* Address review: fix stale links, support all manifest depths, scope error

_make_link:
- Drop size-based idempotency shortcut that kept stale links after
  ollama pull updates a tag to a same-sized blob. Only samefile()
  is used now -- if the link doesn't point at the exact same inode,
  it gets replaced.

Scanner:
- Revert targeted glob back to rglob so deeper OCI-style repo names
  (5+ path segments) are not silently skipped.

Ollama error:
- Only show "Some Ollama models do not work with llama.cpp" when the
  server output contains GGUF compatibility hints (key not found,
  unknown architecture, failed to load). Unrelated failures like
  OOM or missing binaries now show the generic error instead of
  being misdiagnosed.

---------

Co-authored-by: Daniel Han <info@unsloth.ai>
Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com>
Co-authored-by: danielhanchen <michaelhan2050@gmail.com>
---
 studio/backend/core/inference/llama_cpp.py    |  22 ++
 studio/backend/routes/models.py               | 341 +++++++++++++++++-
 studio/backend/utils/models/model_config.py   |  19 +-
 .../assistant-ui/model-selector/pickers.tsx   | 144 ++++++--
 .../src/features/chat/api/chat-api.ts         |   6 +
 .../chat/hooks/use-chat-model-runtime.ts      |   3 +-
 6 files changed, 500 insertions(+), 35 deletions(-)

diff --git a/studio/backend/core/inference/llama_cpp.py b/studio/backend/core/inference/llama_cpp.py
index b53fc513de..77b58e22fb 100644
--- a/studio/backend/core/inference/llama_cpp.py
+++ b/studio/backend/core/inference/llama_cpp.py
@@ -1703,6 +1703,28 @@ class LlamaCppBackend:
             # Wait for llama-server to become healthy
             if not self._wait_for_health(timeout = 600.0):
                 self._kill_process()
+                _gguf = gguf_path or ""
+                _is_ollama = (
+                    ".studio_links" in _gguf
+                    or os.sep + "ollama_links" + os.sep in _gguf
+                    or os.sep + ".cache" + os.sep + "ollama" + os.sep in _gguf
+                    or (self._model_identifier or "").startswith("ollama/")
+                )
+                # Only show the Ollama-specific message when the server
+                # output indicates a GGUF compatibility issue, not for
+                # unrelated failures like OOM or missing binaries.
+                if _is_ollama:
+                    _output = "\n".join(self._stdout_lines[-50:]).lower()
+                    _gguf_compat_hints = (
+                        "key not found",
+                        "unknown model architecture",
+                        "failed to load model",
+                    )
+                    if any(h in _output for h in _gguf_compat_hints):
+                        raise RuntimeError(
+                            "Some Ollama models do not work with llama.cpp. "
+                            "Try a different model, or use this model directly through Ollama instead."
+                        )
                 raise RuntimeError(
                     "llama-server failed to start. "
                     "Check that the GGUF file is valid and you have enough memory."
diff --git a/studio/backend/routes/models.py b/studio/backend/routes/models.py
index 9e7168eed6..db27ce1907 100644
--- a/studio/backend/routes/models.py
+++ b/studio/backend/routes/models.py
@@ -5,8 +5,11 @@
 Model Management API routes
 """
 
+import hashlib
+import json
 import os
 import sys
+import uuid
 from pathlib import Path
 from fastapi import APIRouter, Body, Depends, HTTPException, Query
 from typing import List, Optional
@@ -411,6 +414,267 @@ def _scan_lmstudio_dir(lm_dir: Path) -> List[LocalModelInfo]:
     return found
 
 
+def _ollama_links_dir(ollama_dir: Path) -> Optional[Path]:
+    """Return a writable directory for Ollama ``.gguf`` symlinks.
+
+    Prefers ``<ollama_dir>/.studio_links/`` so the links sit next to the
+    blobs they point at. Falls back to a per-ollama-dir namespace under
+    Studio's own cache when the models directory is read-only (common
+    for system installs under ``/usr/share/ollama`` or ``/var/lib/ollama``)
+    so we still surface Ollama models in those environments.
+    """
+    from utils.paths.storage_roots import cache_root
+
+    primary = ollama_dir / ".studio_links"
+    try:
+        primary.mkdir(exist_ok = True)
+        return primary
+    except OSError as e:
+        logger.debug(
+            "Ollama dir %s not writable for .studio_links (%s); "
+            "falling back to Studio cache",
+            ollama_dir,
+            e,
+        )
+
+    # Fallback: namespace by a hash of the ollama_dir so two different
+    # Ollama roots don't collide. This is a cache path, not a security
+    # boundary.
+    try:
+        digest = hashlib.sha256(str(ollama_dir.resolve()).encode()).hexdigest()[:12]
+    except OSError:
+        digest = "default"
+    fallback = cache_root() / "ollama_links" / digest
+    try:
+        fallback.mkdir(parents = True, exist_ok = True)
+        return fallback
+    except OSError as e:
+        logger.warning(
+            "Could not create Ollama symlink cache at %s: %s",
+            fallback,
+            e,
+        )
+        return None
+
+
+def _scan_ollama_dir(
+    ollama_dir: Path, limit: Optional[int] = None
+) -> List[LocalModelInfo]:
+    """Scan an Ollama models directory for downloaded models.
+
+    Ollama stores models in a content-addressable layout::
+
+        <ollama_dir>/manifests/<host>/<namespace>/<model>/<tag>
+        <ollama_dir>/blobs/sha256-...
+
+    The default host is ``registry.ollama.ai`` with namespace
+    ``library`` (official models), but users can pull from custom
+    namespaces (``mradermacher/llama3``) or entirely different hosts
+    (``hf.co/org/repo:tag``).  We iterate all manifest files via
+    ``rglob`` so every layout depth is discovered.
+
+    Each manifest is JSON with a ``layers`` array. The layer with
+    ``mediaType == "application/vnd.ollama.image.model"`` contains the
+    GGUF weights. Vision models also have a projector layer
+    (``application/vnd.ollama.image.projector``). We read the config
+    layer to extract family/size info.
+
+    Since Ollama blobs lack a ``.gguf`` extension (which the GGUF
+    loading pipeline requires), we create ``.gguf``-named links
+    pointing at the blobs so the existing ``detect_gguf_model`` and
+    ``llama-server -m`` paths work unchanged. Each model gets its
+    own subdirectory under the links dir (keyed by a short hash of
+    the manifest path) so that ``detect_mmproj_file`` only sees the
+    projector for *that* model.  Links are created as symlinks when
+    possible, falling back to hardlinks (Windows without Developer
+    Mode) as a last resort.  The link dir lives under
+    ``<ollama_dir>/.studio_links/`` when writable, otherwise under
+    Studio's own cache directory.
+    """
+    manifests_root = ollama_dir / "manifests"
+    if not manifests_root.is_dir():
+        return []
+
+    found: List[LocalModelInfo] = []
+    blobs_dir = ollama_dir / "blobs"
+    links_root = _ollama_links_dir(ollama_dir)
+    if links_root is None:
+        logger.warning(
+            "Skipping Ollama scan for %s: no writable location for .gguf links",
+            ollama_dir,
+        )
+        return []
+
+    def _make_link(link_dir: Path, link_name: str, target: Path) -> Optional[str]:
+        """Create a .gguf-named link to an Ollama blob.
+
+        Tries symlink first, then hardlink (works on Windows without
+        Developer Mode when target is on the same filesystem).  Skips
+        the model if neither works -- a full file copy of a multi-GB
+        GGUF inside a synchronous API request would block the backend.
+
+        Idempotent: skips recreation when a valid link already exists.
+        """
+        link_dir.mkdir(parents = True, exist_ok = True)
+        link_path = link_dir / link_name
+        resolved = target.resolve()
+
+        # Skip if the link already points at the exact same blob.
+        # Only use samefile -- size-based checks can reuse stale links
+        # after `ollama pull` updates a tag to a same-sized blob.
+        try:
+            if link_path.exists() and os.path.samefile(str(link_path), str(resolved)):
+                return str(link_path)
+        except OSError as e:
+            logger.debug("Error checking existing link %s: %s", link_path, e)
+
+        tmp_path = link_dir / f".{link_name}.tmp-{uuid.uuid4().hex[:8]}"
+        try:
+            if tmp_path.is_symlink() or tmp_path.exists():
+                tmp_path.unlink()
+            try:
+                tmp_path.symlink_to(resolved)
+            except OSError:
+                try:
+                    os.link(str(resolved), str(tmp_path))
+                except OSError:
+                    logger.warning(
+                        "Could not create link for Ollama blob %s "
+                        "(symlinks and hardlinks both failed). "
+                        "Skipping model to avoid blocking the API.",
+                        target,
+                    )
+                    return None
+            os.replace(str(tmp_path), str(link_path))
+            return str(link_path)
+        except OSError as e:
+            logger.debug("Could not create Ollama link %s: %s", link_path, e)
+            try:
+                if tmp_path.is_symlink() or tmp_path.exists():
+                    tmp_path.unlink()
+            except OSError as cleanup_err:
+                logger.debug(
+                    "Could not clean up tmp path %s: %s", tmp_path, cleanup_err
+                )
+            return None
+
+    try:
+        for tag_file in manifests_root.rglob("*"):
+            if not tag_file.is_file():
+                continue
+
+            rel = tag_file.relative_to(manifests_root)
+            parts = rel.parts
+            if len(parts) < 3:
+                continue
+
+            host = parts[0]
+            repo_parts = list(parts[1:-1])
+            tag = parts[-1]
+
+            if (
+                host == "registry.ollama.ai"
+                and repo_parts
+                and repo_parts[0] == "library"
+            ):
+                repo_name = "/".join(repo_parts[1:])
+            elif host == "registry.ollama.ai":
+                repo_name = "/".join(repo_parts)
+            else:
+                repo_name = "/".join([host] + repo_parts)
+
+            if not repo_name:
+                continue
+
+            display = f"{repo_name}:{tag}"
+
+            manifest_key = rel.as_posix()
+            stem_hash = hashlib.sha256(manifest_key.encode()).hexdigest()[:10]
+
+            try:
+                manifest = json.loads(tag_file.read_text())
+            except (json.JSONDecodeError, OSError) as e:
+                logger.debug(
+                    "Skipping unreadable/invalid Ollama manifest %s: %s",
+                    tag_file,
+                    e,
+                )
+                continue
+
+            config_digest = manifest.get("config", {}).get("digest", "")
+            model_type = ""
+            file_type = ""
+            if config_digest and blobs_dir.is_dir():
+                config_blob = blobs_dir / config_digest.replace(":", "-")
+                if config_blob.is_file():
+                    try:
+                        cfg = json.loads(config_blob.read_text())
+                        model_type = cfg.get("model_type", "")
+                        file_type = cfg.get("file_type", "")
+                    except (json.JSONDecodeError, OSError) as e:
+                        logger.debug(
+                            "Could not parse Ollama config blob %s: %s",
+                            config_blob,
+                            e,
+                        )
+
+            model_link_dir = links_root / stem_hash
+
+            gguf_link_path: Optional[str] = None
+            quant = f"-{file_type}" if file_type else ""
+            safe_name = repo_name.replace("/", "-")
+            for layer in manifest.get("layers", []):
+                media = layer.get("mediaType", "")
+                digest = layer.get("digest", "")
+                if not digest:
+                    continue
+
+                if media == "application/vnd.ollama.image.model":
+                    candidate = blobs_dir / digest.replace(":", "-")
+                    if candidate.is_file():
+                        link_name = f"{safe_name}-{tag}{quant}.gguf"
+                        gguf_link_path = _make_link(
+                            model_link_dir, link_name, candidate
+                        )
+
+                elif media == "application/vnd.ollama.image.projector":
+                    candidate = blobs_dir / digest.replace(":", "-")
+                    if candidate.is_file():
+                        mmproj_name = f"{safe_name}-{tag}-mmproj.gguf"
+                        _make_link(model_link_dir, mmproj_name, candidate)
+
+            if not gguf_link_path:
+                continue
+
+            suffix = ""
+            if model_type:
+                suffix += f" ({model_type}"
+                if file_type:
+                    suffix += f" {file_type}"
+                suffix += ")"
+
+            try:
+                updated_at = tag_file.stat().st_mtime
+            except OSError:
+                updated_at = None
+
+            found.append(
+                LocalModelInfo(
+                    id = gguf_link_path,
+                    model_id = f"ollama/{repo_name}:{tag}",
+                    display_name = display + suffix,
+                    path = gguf_link_path,
+                    source = "custom",
+                    updated_at = updated_at,
+                ),
+            )
+            if limit is not None and len(found) >= limit:
+                return found
+    except OSError as e:
+        logger.warning("Error scanning Ollama directory %s: %s", ollama_dir, e)
+    return found
+
+
 @router.get("/local", response_model = LocalModelListResponse)
 async def list_local_models(
     models_dir: str = Query(
@@ -493,11 +757,27 @@ async def list_local_models(
         for folder in custom_folders:
             folder_path = Path(folder["path"])
             try:
-                custom_models = (
-                    _scan_models_dir(folder_path, limit = _MAX_MODELS_PER_FOLDER)
-                    + _scan_hf_cache(folder_path)
-                    + _scan_lmstudio_dir(folder_path)
-                )[:_MAX_MODELS_PER_FOLDER]
+                # Ollama scanner creates .studio_links/ with .gguf symlinks.
+                # Filter those from the generic scanners to avoid duplicates
+                # and leaking internal paths into the UI.
+                _generic = [
+                    m
+                    for m in (
+                        _scan_models_dir(folder_path, limit = _MAX_MODELS_PER_FOLDER)
+                        + _scan_hf_cache(folder_path)
+                        + _scan_lmstudio_dir(folder_path)
+                    )
+                    if not any(
+                        p in (".studio_links", "ollama_links")
+                        for p in Path(m.path).parts
+                    )
+                ]
+                custom_models = _generic
+                if len(custom_models) < _MAX_MODELS_PER_FOLDER:
+                    custom_models += _scan_ollama_dir(
+                        folder_path,
+                        limit = _MAX_MODELS_PER_FOLDER - len(custom_models),
+                    )
             except OSError as e:
                 logger.warning("Skipping unreadable scan folder %s: %s", folder_path, e)
                 continue
@@ -575,6 +855,57 @@ async def remove_scan_folder_endpoint(
     return {"ok": True}
 
 
+@router.get("/recommended-folders")
+async def get_recommended_folders(
+    current_subject: str = Depends(get_current_subject),
+):
+    """Return well-known model directories that exist on this machine.
+
+    Lightweight alternative to ``browse-folders`` for showing quick-pick
+    chips without the overhead of enumerating a directory tree.  Returns
+    paths that actually exist on disk (HF cache, LM Studio, Ollama,
+    ``~/models``, etc.) so the frontend can offer them as one-click
+    "Recommended" shortcuts in the Custom Folders section.
+    """
+    from utils.paths.storage_roots import lmstudio_model_dirs
+
+    folders: list[str] = []
+    seen: set[str] = set()
+
+    def _add(p: Optional[Path]) -> None:
+        if p is None:
+            return
+        try:
+            resolved = str(p.resolve())
+        except OSError:
+            return
+        if resolved in seen:
+            return
+        if Path(resolved).is_dir() and os.access(resolved, os.R_OK | os.X_OK):
+            seen.add(resolved)
+            folders.append(resolved)
+
+    # LM Studio model directories
+    try:
+        for p in lmstudio_model_dirs():
+            _add(p)
+    except Exception as e:
+        logger.warning("Failed to scan for LM Studio model directories: %s", e)
+
+    # Ollama model directories
+    ollama_env = os.environ.get("OLLAMA_MODELS")
+    if ollama_env:
+        _add(Path(ollama_env).expanduser())
+    for candidate in (
+        Path.home() / ".ollama" / "models",
+        Path("/usr/share/ollama/.ollama/models"),
+        Path("/var/lib/ollama/.ollama/models"),
+    ):
+        _add(candidate)
+
+    return {"folders": folders}
+
+
 # Heuristic ceiling on how many children to stat when checking whether a
 # directory "looks like" it contains models. Keeps the browser snappy
 # even when a directory has thousands of unrelated entries.
diff --git a/studio/backend/utils/models/model_config.py b/studio/backend/utils/models/model_config.py
index 44754520e3..a2d48cf009 100644
--- a/studio/backend/utils/models/model_config.py
+++ b/studio/backend/utils/models/model_config.py
@@ -959,6 +959,20 @@ def detect_mmproj_file(path: str, search_root: Optional[str] = None) -> Optional
         scan_order.append(resolved)
 
     _add(start_dir)
+
+    # When ``path`` is a symlink (e.g. Ollama's ``.studio_links/...gguf``
+    # -> ``blobs/sha256-...``), the symlink's parent directory rarely
+    # contains the mmproj sibling; the real mmproj file lives next to
+    # the symlink target. Add the target's parent to the scan so vision
+    # GGUFs that are surfaced via symlinks are still recognised as
+    # vision models.
+    try:
+        if p.is_symlink() and p.is_file():
+            target_parent = p.resolve().parent
+            if target_parent.is_dir():
+                _add(target_parent)
+    except OSError:
+        pass
     if search_root is not None:
         try:
             root_resolved = Path(search_root).resolve()
@@ -1006,7 +1020,10 @@ def detect_gguf_model(path: str) -> Optional[str]:
     if p.suffix.lower() == ".gguf" and p.is_file():
         if _is_mmproj(p.name):
             return None
-        return str(p.resolve())
+        # Use absolute (not resolve) to preserve symlink names -- e.g.
+        # Ollama .studio_links/model.gguf -> blobs/sha256-... should
+        # keep the readable symlink name, not the opaque blob hash.
+        return str(p.absolute())
 
     # Case 2: directory containing .gguf files (skip mmproj)
     if p.is_dir():
diff --git a/studio/frontend/src/components/assistant-ui/model-selector/pickers.tsx b/studio/frontend/src/components/assistant-ui/model-selector/pickers.tsx
index 2f661c2e72..8f318a8292 100644
--- a/studio/frontend/src/components/assistant-ui/model-selector/pickers.tsx
+++ b/studio/frontend/src/components/assistant-ui/model-selector/pickers.tsx
@@ -27,6 +27,7 @@ import {
   listCachedModels,
   listGgufVariants,
   listLocalModels,
+  listRecommendedFolders,
   listScanFolders,
   removeScanFolder,
 } from "@/features/chat/api/chat-api";
@@ -49,7 +50,7 @@ import { checkVramFit, estimateLoadingVram } from "@/lib/vram";
 import { Add01Icon, Cancel01Icon, Folder02Icon, Search01Icon } from "@hugeicons/core-free-icons";
 import { HugeiconsIcon } from "@hugeicons/react";
 import { FolderBrowser } from "./folder-browser";
-import { Trash2Icon } from "lucide-react";
+import { ChevronDownIcon, ChevronRightIcon, DownloadIcon, StarIcon, Trash2Icon } from "lucide-react";
 import {
   type ReactNode,
   useCallback,
@@ -73,10 +74,35 @@ function normalizeForSearch(s: string): string {
   return s.toLowerCase().replace(/[\s\-_\.]/g, "");
 }
 
-function ListLabel({ children }: { children: ReactNode }) {
+function ListLabel({
+  children,
+  icon,
+  collapsed,
+  onToggle,
+}: {
+  children: ReactNode;
+  icon?: ReactNode;
+  collapsed?: boolean;
+  onToggle?: () => void;
+}) {
   return (
-    <div className="px-2.5 py-1.5 text-[10px] font-semibold uppercase tracking-wider text-muted-foreground">
-      {children}
+    <div className="flex items-center justify-between gap-1 px-2.5 py-1.5">
+      <span className="flex items-center gap-1.5 text-[10px] font-semibold uppercase tracking-wider text-muted-foreground">
+        {icon}
+        {children}
+      </span>
+      {onToggle && (
+        <button
+          type="button"
+          onClick={onToggle}
+          aria-label={collapsed ? "Expand section" : "Collapse section"}
+          className="shrink-0 rounded p-1 text-muted-foreground/60 transition-colors hover:text-foreground"
+        >
+          {collapsed
+            ? <ChevronRightIcon className="size-3" />
+            : <ChevronDownIcon className="size-3" />}
+        </button>
+      )}
     </div>
   );
 }
@@ -489,6 +515,9 @@ export function HubModelPicker({
   // Delete confirmation dialog state
   const [deleteTarget, setDeleteTarget] = useState<string | null>(null);
   const [deleting, setDeleting] = useState(false);
+  const [downloadedCollapsed, setDownloadedCollapsed] = useState(false);
+  const [customFoldersCollapsed, setCustomFoldersCollapsed] = useState(false);
+  const [recommendedCollapsed, setRecommendedCollapsed] = useState(false);
 
   // Cached (already downloaded) repos -- use module-level cache so
   // re-mounting the popover does not flash an empty "Downloaded" section.
@@ -514,6 +543,7 @@ export function HubModelPicker({
   const [showFolderInput, setShowFolderInput] = useState(false);
   const [folderLoading, setFolderLoading] = useState(false);
   const [showFolderBrowser, setShowFolderBrowser] = useState(false);
+  const [recommendedFolders, setRecommendedFolders] = useState<string[]>([]);
 
   const refreshLocalModelsList = useCallback(() => {
     listLocalModels()
@@ -616,6 +646,9 @@ export function HubModelPicker({
     // Always refresh LM Studio + custom folder models (not gated by alreadyCached)
     refreshLocalModelsList();
     refreshScanFolders();
+    listRecommendedFolders()
+      .then(setRecommendedFolders)
+      .catch(() => {});
 
     // Always refetch cached GGUF/model lists. The module-level caches give
     // an instant render with stale data (no spinner flash), but newly
@@ -893,8 +926,12 @@ export function HubModelPicker({
             (cachedGguf.length > 0 ||
               (!chatOnly && cachedModels.length > 0)) ? (
             <>
-              <ListLabel>Downloaded</ListLabel>
-              {cachedGguf.map((c) => (
+              <ListLabel
+                icon={<DownloadIcon className="size-3" />}
+                collapsed={downloadedCollapsed}
+                onToggle={() => setDownloadedCollapsed((v) => !v)}
+              >Downloaded</ListLabel>
+              {!downloadedCollapsed && cachedGguf.map((c) => (
                 <div key={c.repo_id}>
                   <ModelRow
                     label={c.repo_id}
@@ -922,7 +959,7 @@ export function HubModelPicker({
                   )}
                 </div>
               ))}
-              {!chatOnly &&
+              {!downloadedCollapsed && !chatOnly &&
                 cachedModels.map((c) => (
                   <div key={c.repo_id} className="flex items-center gap-0.5">
                     <div className="min-w-0 flex-1">
@@ -1001,20 +1038,12 @@ export function HubModelPicker({
 
           {!showHfSection ? (
             <>
-              <div className="flex items-center justify-between gap-1 px-2.5 py-1.5">
-                <span className="text-[10px] font-semibold uppercase tracking-wider text-muted-foreground">
+              <div className="flex items-center gap-1 px-2.5 py-1.5">
+                <span className="flex items-center gap-1.5 text-[10px] font-semibold uppercase tracking-wider text-muted-foreground">
+                  <HugeiconsIcon icon={Folder02Icon} className="size-3" />
                   Custom Folders
                 </span>
                 <div className="flex items-center gap-0.5">
-                  <button
-                    type="button"
-                    aria-label="Browse for a folder on the server"
-                    title="Browse folders on the server"
-                    onClick={() => setShowFolderBrowser(true)}
-                    className="shrink-0 rounded p-1 text-muted-foreground/60 transition-colors hover:text-foreground"
-                  >
-                    <HugeiconsIcon icon={Search01Icon} className="size-3" />
-                  </button>
                   <button
                     type="button"
                     aria-label={showFolderInput ? "Cancel adding folder" : "Add scan folder by path"}
@@ -1029,11 +1058,33 @@ export function HubModelPicker({
                   >
                     <HugeiconsIcon icon={showFolderInput ? Cancel01Icon : Add01Icon} className="size-3" />
                   </button>
+                  <button
+                    type="button"
+                    aria-label="Browse for a folder on the server"
+                    title="Browse folders on the server"
+                    onClick={() => setShowFolderBrowser(true)}
+                    className="shrink-0 rounded p-0.5 text-muted-foreground/60 transition-colors hover:text-foreground"
+                  >
+                    <HugeiconsIcon icon={Search01Icon} className="size-2.5" />
+                  </button>
+                </div>
+                <div className="ml-auto">
+                  <button
+                    type="button"
+                    aria-label={customFoldersCollapsed ? "Expand custom folders" : "Collapse custom folders"}
+                    title={customFoldersCollapsed ? "Expand" : "Collapse"}
+                    onClick={() => setCustomFoldersCollapsed((v) => !v)}
+                    className="shrink-0 rounded p-1 text-muted-foreground/60 transition-colors hover:text-foreground"
+                  >
+                    {customFoldersCollapsed
+                      ? <ChevronRightIcon className="size-3" />
+                      : <ChevronDownIcon className="size-3" />}
+                  </button>
                 </div>
               </div>
 
               {/* Folder paths */}
-              {scanFolders.map((f) => (
+              {!customFoldersCollapsed && scanFolders.map((f) => (
                 <div
                   key={f.id}
                   className="group flex items-center gap-1.5 px-2.5 py-0.5"
@@ -1056,8 +1107,31 @@ export function HubModelPicker({
                 </div>
               ))}
 
+              {/* Recommended folders */}
+              {!customFoldersCollapsed && (() => {
+                const registered = new Set(scanFolders.map((f) => f.path));
+                const unregistered = recommendedFolders.filter((p) => !registered.has(p));
+                if (unregistered.length === 0) return null;
+                return (
+                  <div className="flex flex-wrap gap-1 px-2.5 pb-0.5">
+                    {unregistered.map((p) => (
+                      <button
+                        key={p}
+                        type="button"
+                        onClick={() => void handleAddFolder(p)}
+                        disabled={folderLoading}
+                        title={`Add ${p}`}
+                        className="rounded-full border border-dashed border-border/50 px-2 py-0.5 font-mono text-[10px] text-muted-foreground/70 transition-colors hover:border-foreground/30 hover:bg-accent hover:text-foreground disabled:opacity-40"
+                      >
+                        <span className="text-[11px] font-semibold">+</span> {p.length > 30 ? `...${p.slice(-27)}` : p}
+                      </button>
+                    ))}
+                  </div>
+                );
+              })()}
+
               {/* Add folder input */}
-              {showFolderInput && (
+              {!customFoldersCollapsed && showFolderInput && (
                 <div className="px-2.5 pb-1 pt-0.5">
                   <div className="flex items-center gap-1">
                     <HugeiconsIcon icon={Folder02Icon} className="size-3 shrink-0 text-muted-foreground/40" />
@@ -1114,11 +1188,15 @@ export function HubModelPicker({
 
 
               {/* Models from custom folders */}
-              {customFolderModels.map((m) => {
+              {!customFoldersCollapsed && customFolderModels.map((m) => {
+                const isGgufFile = m.path.toLowerCase().endsWith(".gguf");
                 const isGguf =
+                  isGgufFile ||
                   isGgufRepo(m.id) ||
-                  isGgufRepo(m.display_name) ||
-                  m.path.toLowerCase().endsWith(".gguf");
+                  isGgufRepo(m.display_name);
+                // Single .gguf files (e.g. Ollama blobs) load directly;
+                // GGUF repos/directories expand to pick a variant.
+                const isDirectGguf = isGgufFile;
                 return (
                   <div key={m.id}>
                     <ModelRow
@@ -1126,7 +1204,13 @@ export function HubModelPicker({
                       meta={isGguf ? "GGUF" : "Local"}
                       selected={value === m.id}
                       onClick={() => {
-                        if (isGguf) {
+                        if (isDirectGguf) {
+                          onSelect(m.id, {
+                            source: "local",
+                            isLora: false,
+                            isDownloaded: true,
+                          });
+                        } else if (isGguf) {
                           setExpandedGguf((prev) =>
                             prev === m.id ? null : m.id,
                           );
@@ -1158,8 +1242,12 @@ export function HubModelPicker({
 
           {!showHfSection && cachedReady ? (
             <>
-              <ListLabel>Recommended</ListLabel>
-              {visibleRecommendedIds.length === 0 ? (
+              <ListLabel
+                icon={<StarIcon className="size-3" />}
+                collapsed={recommendedCollapsed}
+                onToggle={() => setRecommendedCollapsed((v) => !v)}
+              >Recommended</ListLabel>
+              {recommendedCollapsed ? null : visibleRecommendedIds.length === 0 ? (
                 <div className="px-2.5 py-2 text-xs text-muted-foreground">
                   No default models.
                 </div>
@@ -1203,7 +1291,7 @@ export function HubModelPicker({
                   );
                 })
               )}
-              {hasMoreRecommended && (
+              {!recommendedCollapsed && hasMoreRecommended && (
                 <>
                   <div ref={recommendedSentinelRef} className="h-px" />
                   <div className="flex items-center justify-center py-2">
@@ -1216,7 +1304,7 @@ export function HubModelPicker({
 
           {showHfSection && filteredRecommendedIds.length > 0 ? (
             <>
-              <ListLabel>Recommended</ListLabel>
+              <ListLabel icon={<StarIcon className="size-3" />}>Recommended</ListLabel>
               {filteredRecommendedIds.map((id) => {
                 const vram = recommendedVramMap.get(id);
                 return (
diff --git a/studio/frontend/src/features/chat/api/chat-api.ts b/studio/frontend/src/features/chat/api/chat-api.ts
index 9aacfc5af4..15ac8748f0 100644
--- a/studio/frontend/src/features/chat/api/chat-api.ts
+++ b/studio/frontend/src/features/chat/api/chat-api.ts
@@ -262,6 +262,12 @@ export interface BrowseFoldersResponse {
   model_files_here?: number;
 }
 
+export async function listRecommendedFolders(): Promise<string[]> {
+  const response = await authFetch("/api/models/recommended-folders");
+  const data = await parseJsonOrThrow<{ folders: string[] }>(response);
+  return data.folders;
+}
+
 export async function browseFolders(
   path?: string,
   showHidden = false,
diff --git a/studio/frontend/src/features/chat/hooks/use-chat-model-runtime.ts b/studio/frontend/src/features/chat/hooks/use-chat-model-runtime.ts
index 53f0d1d352..037d3182ac 100644
--- a/studio/frontend/src/features/chat/hooks/use-chat-model-runtime.ts
+++ b/studio/frontend/src/features/chat/hooks/use-chat-model-runtime.ts
@@ -437,9 +437,10 @@ export function useChatModelRuntime() {
             const { chatTemplateOverride, kvCacheDtype, customContextLength, ggufContextLength, speculativeType } = useChatRuntimeStore.getState();
             // GGUF: use custom context length, or 0 = model's native context
             // Non-GGUF: use the Max Seq Length slider value
+            const isDirectGgufFile = modelId.toLowerCase().endsWith(".gguf");
             const effectiveMaxSeqLength = customContextLength != null
               ? customContextLength
-              : ggufVariant != null ? (ggufContextLength ?? 0) : maxSeqLength;
+              : (ggufVariant != null || isDirectGgufFile) ? (ggufContextLength ?? 0) : maxSeqLength;
             const loadResponse = await loadModel({
               model_path: modelId,
               hf_token: hfToken,

From b01e9af1240df395bc3361a9c473826214f48d9f Mon Sep 17 00:00:00 2001
From: Wasim Yousef Said <wasimysdev@gmail.com>
Date: Thu, 16 Apr 2026 17:46:16 +0200
Subject: [PATCH 126/216] feat(studio): replace navbar with collapsible sidebar
 (#4936)
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

* feat(studio): replace navbar navigation with collapsible sidebar

Add an app-wide sidebar with hover-expand and pin-to-dock behavior.
Navigation items (Studio, Recipes, Export, Chat) move from the center
pill navbar to the sidebar. Chat threads and recipes render as
collapsible sub-lists. Navbar simplified to logo + update + close.

- Extend SidebarProvider with pinned/hovered state model
- New AppSidebar with animated active indicator, sloth profile menu,
  theme toggle, guided tour, back/forward navigation
- Chat page refactored to URL-driven view state via search params
- Extract reusable hooks for chat thread and recipe sidebar data
- Guard startViewTransition for browser compatibility
- Wrap chat deletions in Dexie transaction for data integrity

* feat(studio): move logo to sidebar and make navbar overlay

- Sidebar is now full-height with logo in SidebarHeader
- Collapsed sidebar shows sticker.png, expanded shows full logo
- Navbar is absolute-positioned overlay (no layout space)
- Main content extends to top, aligning with navbar controls

* feat(studio): full-height sidebar with recents, edge-to-edge nav buttons

- Sidebar outside max-w-7xl, pinned to left edge
- Remove sidebar rounding, menu buttons rounded-md
- Nav buttons flush to sidebar edges with no left rounding
- Replace collapsible recipes/chat with flat nav items
- Add Recents section with chat history (1 item when not on chat, full on chat)
- New Chat as first nav item with PencilEdit02Icon
- Cursor pointer on all sidebar buttons
- Navbar temporarily hidden for screenshots

* fix(studio): fix chat scroll, action bar hover, collapsible recents

- Fix sticky composer by removing `relative` override on viewport footer
- Action bar buttons only show on hover (autohide=always)
- Remove floating border/shadow from action bar
- Add scroll space above composer for last message actions
- Back/forward buttons use router history (stay in-app)
- Recents section collapsible with chevron on chat route
- Set html/body/#root height for proper h-full chain

* fix(studio): address review feedback, clean up unused code

- Unhide navbar (was left hidden from screenshot)
- Remove unused imports: SidebarMenuSub*, BubbleChatIcon, ColumnInsertIcon
- Remove unused vars: recipeItems, activeRecipeId, canCompare, recipesOpen
- Include compare query id in active sidebar selection
- Use store type for contextUsage instead of inline type
- Simplify noop in sidebar.tsx
- Remove empty className prop

* feat(studio): add mobile sidebar, recent runs section, and misc UX fixes

* feat(studio): scaffold settings feature module with dialog store

* feat(studio): add tri-state theme store for settings

* feat(chat): add clear-all-chats and export-chat-history utils

* feat(studio): add settings dialog shell with tab rail

* feat(studio): add appearance tab with theme and sidebar pin

* feat(studio): add settings general tab with hf token, auto-title, reset prefs

* feat(studio): add settings chat tab with export and clear

* feat(studio): add api keys tab with list and revoke flow

* feat(studio): add create-key form and reveal dialog

* feat(studio): add usage examples panel to api keys tab

* feat(studio): add settings about tab with update and shutdown

* feat(studio): add settings dropdown item and cmd-comma shortcut

* feat(studio): remove legacy api-keys route and chat-sheet preference rows

* fix(studio): settings dialog a11y + polish pass

* feat(studio): inline api key reveal card replacing nested dialog

* fix(studio): hide revoked keys from settings list

* refactor(studio): strip navbar and hoist training unload guard

* feat(studio): explicit sidebar toggle, remove hover-open and pin icons

* fix(studio): use SidebarRight01Icon for collapsed sidebar open toggle

* fix(studio): address code review findings for settings dialog

* feat(studio): collapsible navigate group with standalone new-chat and compare

* fix(studio): chat-only standalone actions, use ColumnInsertIcon for compare

* fix(studio): sidebar new-chat/compare state reset and icon-mode collapsible

* feat(studio): add compact logo assets for sidebar header

* Fixed sidebar design

* fix(studio): sidebar delete icon hover contrast and sizing

* feat(studio): route-gate sidebar recents (chats off /studio, runs on /studio)

* feat(studio): add chat search store

* feat(studio): add chat search index hook with snapshot-on-open

* feat(studio): add chat search command dialog with global shortcut

* feat(studio): wire chat search into sidebar

* fix(studio): trim hf token on save, add show/hide toggle, commit on close

* revert(studio): restore original sidebar/border colors, brighten sidebar

* feat(studio): forward overlayClassName through CommandDialog

* fix(studio): wrap search dialog in Command context, redesign as flat 635px card

* fix(studio): reserve right padding on recent items so delete icon stops overlapping title

* fix(studio): skip hf token unmount-commit during reset-prefs reload

* chore(studio): drop unused icon import and unreachable runs navigate branch

* fix(studio): chat search index filters archived before limit, batches message query, picks up reasoning text

* fix(studio): keep CommandEmpty in tree so empty state renders correctly

* fix(studio): cap system prompt and chat template textareas so they scroll instead of growing

* fix(studio): attach chat-compare tour anchor to sidebar compare button

* fix(studio): persist system theme explicitly so next-themes does not clobber on reload

* fix(studio): auto-switch to history tab when selecting a recent run from sidebar

* UI overhaul: chatbox, scrollbar, sidebar, and compare view

UI Changes:
- Redesigned the Compare UI with general cleanup
- Redesigned the Chatbox UI
- Reduced the width of the user chat bubble for improved readability
- Narrowed the user chat box across the content page
- Adjusted thinking-box text color to be slightly darker
- Removed faded text effect from chat messages
- Removed faded text effect from the thinking box
- Added a small LLM chat safety note at the bottom of the chatbox
- Restyled the scrollbar

Layout & Behavior:
- Reworked the scrollbar to span the full height of the page (no top/bottom padding) and remain persistently visible when content is scrollable, rather than only on hover
- Reworked the Configuration sidebar to span full height — removed rounded corners and borders, with the scrollbar adjusted to match the full top-to-bottom layout
- Adjusted the top menu and bottom chatbox content areas to work correctly with the new full-page scroll behavior
- Made chat content match the chatbox width, with content sliding slightly behind the chatbox when scrolling
- Aligned chat text width with the chatbox for visual consistency, including how far the text extends behind the chatbox

Fixes:
- Fixed the chatbox not auto-expanding when typing multi-line input while bottom-positioned during an active chat (previously only worked before a chat had started)
- Fixed positioning and design of the user chat hover menu buttons to match the assistant chat box — now displayed below the chat bubble instead of on the left side

* Fix user message layout in thread component

* swap code icon

* fix compare layout

* fix compare pane flex

* Sidebar improvements and fixes

- Added scrolling support to the sidebar so menus and recent chats no longer get hidden
- Recent chats are now always visible in the sidebar, not hidden when in Studio, Recipes, or Export
- Recent chat is now deselected when selecting other navigations
- Fixed sidebar glitch where browser resize could make the sidebar and expand button disappear completely
- Fixed glitch where the open-sidebar hover tooltip appeared above the logo when clicking expand sidebar
- Reduced sidebar width on mobile to around 2/3 of the screen (was too wide)
- Made the close-sidebar hover tooltip consistent with the rest of the design
- Removed sidebar collapse/expand animation
- Small adjustment to chat width

* Fix route scrolling, polling, and theme sync issues

* Fix Studio page scrolling

---------

Co-authored-by: sneakr <hauzin@hotmail.com>
---
 studio/frontend/package.json                  |   1 +
 studio/frontend/public/blacklogo-c.png        | Bin 0 -> 141545 bytes
 studio/frontend/public/sticker.png            | Bin 0 -> 1013935 bytes
 studio/frontend/public/whitelogo-c.png        | Bin 0 -> 139810 bytes
 studio/frontend/src/app/router.tsx            |   2 -
 studio/frontend/src/app/routes/__root.tsx     |  71 +-
 studio/frontend/src/app/routes/api-keys.tsx   |  18 -
 studio/frontend/src/app/routes/chat.tsx       |  15 +-
 .../frontend/src/components/app-sidebar.tsx   | 607 +++++++++++++
 .../src/components/assistant-ui/reasoning.tsx |  41 +-
 .../src/components/assistant-ui/thread.tsx    | 120 ++-
 studio/frontend/src/components/navbar.tsx     | 645 +-------------
 .../src/components/shutdown-dialog.tsx        |  13 +-
 .../components/ui/animated-theme-toggler.tsx  |  76 +-
 studio/frontend/src/components/ui/command.tsx |  11 +-
 studio/frontend/src/components/ui/sidebar.tsx | 133 ++-
 .../src/features/auth/api-keys-page.tsx       | 426 ---------
 studio/frontend/src/features/auth/index.ts    |   1 -
 .../frontend/src/features/chat/chat-page.tsx  | 475 ++++------
 .../src/features/chat/chat-settings-sheet.tsx |  63 +-
 .../chat/components/chat-search-dialog.tsx    | 120 +++
 .../chat/hooks/use-chat-search-index.ts       | 156 ++++
 .../chat/hooks/use-chat-sidebar-items.ts      |  82 ++
 .../src/features/chat/shared-composer.tsx     |   2 +-
 .../chat/stores/chat-runtime-store.ts         |   4 +
 .../features/chat/stores/chat-search-store.ts |  18 +
 .../src/features/chat/thread-sidebar.tsx      |  74 +-
 .../features/chat/utils/clear-all-chats.ts    |  15 +
 .../chat/utils/export-chat-history.ts         |  41 +
 .../hooks/use-recipe-sidebar-items.ts         |  22 +
 .../src/features/settings/api/api-keys.ts     |  41 +
 .../settings/components/api-key-row.tsx       | 101 +++
 .../settings/components/create-key-form.tsx   |  83 ++
 .../settings/components/key-reveal-card.tsx   |  71 ++
 .../settings/components/settings-row.tsx      |  39 +
 .../settings/components/settings-section.tsx  |  30 +
 .../settings/components/theme-segmented.tsx   |  58 ++
 .../components/update-studio-instructions.tsx | 185 ++++
 .../settings/components/usage-examples.tsx    | 146 +++
 .../frontend/src/features/settings/index.ts   |   6 +
 .../src/features/settings/settings-dialog.tsx | 142 +++
 .../settings/stores/settings-dialog-store.ts  |  52 ++
 .../features/settings/stores/theme-store.ts   |  96 ++
 .../src/features/settings/tabs/about-tab.tsx  | 101 +++
 .../features/settings/tabs/api-keys-tab.tsx   | 159 ++++
 .../features/settings/tabs/appearance-tab.tsx |  40 +
 .../src/features/settings/tabs/chat-tab.tsx   | 129 +++
 .../features/settings/tabs/general-tab.tsx    | 197 +++++
 .../src/features/studio/studio-page.tsx       |  33 +-
 .../hooks/use-training-history-sidebar.ts     |  56 ++
 .../hooks/use-training-unload-guard.ts        |  40 +
 .../frontend/src/features/training/index.ts   |   1 +
 .../training/stores/training-runtime-store.ts |   4 +
 .../src/features/training/types/runtime.ts    |   2 +
 studio/frontend/src/hooks/use-mobile.ts       |  31 +-
 studio/frontend/src/hooks/use-sidebar-pin.ts  |  59 ++
 studio/frontend/src/index.css                 | 832 ++++++++++--------
 57 files changed, 3920 insertions(+), 2066 deletions(-)
 create mode 100644 studio/frontend/public/blacklogo-c.png
 create mode 100644 studio/frontend/public/sticker.png
 create mode 100644 studio/frontend/public/whitelogo-c.png
 delete mode 100644 studio/frontend/src/app/routes/api-keys.tsx
 create mode 100644 studio/frontend/src/components/app-sidebar.tsx
 delete mode 100644 studio/frontend/src/features/auth/api-keys-page.tsx
 create mode 100644 studio/frontend/src/features/chat/components/chat-search-dialog.tsx
 create mode 100644 studio/frontend/src/features/chat/hooks/use-chat-search-index.ts
 create mode 100644 studio/frontend/src/features/chat/hooks/use-chat-sidebar-items.ts
 create mode 100644 studio/frontend/src/features/chat/stores/chat-search-store.ts
 create mode 100644 studio/frontend/src/features/chat/utils/clear-all-chats.ts
 create mode 100644 studio/frontend/src/features/chat/utils/export-chat-history.ts
 create mode 100644 studio/frontend/src/features/data-recipes/hooks/use-recipe-sidebar-items.ts
 create mode 100644 studio/frontend/src/features/settings/api/api-keys.ts
 create mode 100644 studio/frontend/src/features/settings/components/api-key-row.tsx
 create mode 100644 studio/frontend/src/features/settings/components/create-key-form.tsx
 create mode 100644 studio/frontend/src/features/settings/components/key-reveal-card.tsx
 create mode 100644 studio/frontend/src/features/settings/components/settings-row.tsx
 create mode 100644 studio/frontend/src/features/settings/components/settings-section.tsx
 create mode 100644 studio/frontend/src/features/settings/components/theme-segmented.tsx
 create mode 100644 studio/frontend/src/features/settings/components/update-studio-instructions.tsx
 create mode 100644 studio/frontend/src/features/settings/components/usage-examples.tsx
 create mode 100644 studio/frontend/src/features/settings/index.ts
 create mode 100644 studio/frontend/src/features/settings/settings-dialog.tsx
 create mode 100644 studio/frontend/src/features/settings/stores/settings-dialog-store.ts
 create mode 100644 studio/frontend/src/features/settings/stores/theme-store.ts
 create mode 100644 studio/frontend/src/features/settings/tabs/about-tab.tsx
 create mode 100644 studio/frontend/src/features/settings/tabs/api-keys-tab.tsx
 create mode 100644 studio/frontend/src/features/settings/tabs/appearance-tab.tsx
 create mode 100644 studio/frontend/src/features/settings/tabs/chat-tab.tsx
 create mode 100644 studio/frontend/src/features/settings/tabs/general-tab.tsx
 create mode 100644 studio/frontend/src/features/training/hooks/use-training-history-sidebar.ts
 create mode 100644 studio/frontend/src/features/training/hooks/use-training-unload-guard.ts
 create mode 100644 studio/frontend/src/hooks/use-sidebar-pin.ts

diff --git a/studio/frontend/package.json b/studio/frontend/package.json
index ffb3c65719..a2eebd5cb5 100644
--- a/studio/frontend/package.json
+++ b/studio/frontend/package.json
@@ -87,6 +87,7 @@
     "eslint-plugin-react-hooks": "^7.0.1",
     "eslint-plugin-react-refresh": "^0.5.2",
     "globals": "^17.4.0",
+    "playwright": "^1.59.1",
     "typescript": "~5.9.3",
     "typescript-eslint": "^8.55.0",
     "vite": "^8.0.1"
diff --git a/studio/frontend/public/blacklogo-c.png b/studio/frontend/public/blacklogo-c.png
new file mode 100644
index 0000000000000000000000000000000000000000..7ab9959536d21dc48ed49f7782d579033014b9f9
GIT binary patch
literal 141545
zcmZ6y1yq#V_dh%{bc=K&TuG@b-AXBlbjZ*hGjw-|q7tHjN`seXC}D<f0hJUO28K@Q
z?s^~Z@4vos-%)X`hqd55=j^>dvG;kU{Zx&Nn2{I+0+Bs_q^t`95x9at5HUi0;0Y^o
z!VdV4-R6<5CI}SB1p<XX2Z7FjN8!IfARl26Xw4D?lKuz+(YdGB>&OD%z_->=QwClC
z{*&EQ@E&+V<o@WH7YIZ`_4^+fl$uTtJS6abtf@k<hD%P0&-#OjT^x7>cJq9y4<5+%
zO9g&B=&|wxegBD#DZgOmr5gF2mEN+zF9J}{?WWmy5@ehA6OpFBZr#=;#C>=Eqe<fV
zUy2V&7V{_HYOv>YQDg7S{^d=0SFm=dl2_c)ecRBjL1ue9VSKcIqZekesUFyT+Vkkk
z{q5e3@4BUGE6h6T5iy?K<1TFb5I#2I0QDl-h{acrSaAb5-chhqD9XnkzX9umbcwU|
zm5D)%IK_mDcfudWP(LUFzCd{NyF_xjNkB{BS)#HMaJcR*(7bm@Ws5Zrj}2<D3=9gz
zI{}}7s=>u7yrN)7{6XTpRB$!lBa-aYxXoe%UQvFVEPNh=f^dA0fOavN`f0K*T%DJV
z`a*p$-@8e3^*e#tvD0Y)LP+4`Mds$%`<Oo;9@h5R6du2O=KGFBN#=Jr>~10SMG^Yp
ze}nScP@aU}!MwNG*;GML;s;JUmzM5tAPbtr>^$L=m=~=4qW#MkelwdI57Q&yu?tD3
z%+lfL!8F}sGDVVX6PiItwkB90_ZD#g6U`uDA$m4cQ@sd3Pn)6;KTjXW+n~U^$$Yg~
z=a|45OFlMQ*1V{O*p)rz_{plzA8Pd%$NxUL<Xy8Vg*X|9U7UE~i~JVsY}x^Ma{PO;
zr_-_mLgoJ~wa%%>=qvq`0Fndi<6dJ)-e%r9z#U}jLI!6y@fqf3=zXokbEkk!B~-{4
z=d#)m!OV{I;=B2N*wai<hf{8`>JYyK8Q|oR^8<fpg{hsh*I(wbyS1RI{aacp>$+w}
zCUXhA6}+d)y+;`x;V%25;wL7@Y_1_I3Sk{W&z1jtk_<>)UrDz<b79*gGBd!@#Y^PV
z94;0uFEDA$bk-?WUyUHnE3xL`eW$%tT5Z&i4`u9l-y<dCV_A}3ZXv}=p~&0vw+RpG
z;a~bSSmHwanCjPKSX@mR+`J>+JaRs>&6n+=yFFmV&0PuPM$jR~4XxGy^-?MGP!HX`
zxnBoX?ixv44y7U0go+GFe5#7cj(7oitmP=#!O01zY^F115&ocBb}{&XCb9XJKF&LW
zPM$PPZP^_7UIe02O%|T`nE*MLpx?=bm3Hmsr@1R>sBJw}NX?it5~{eVM|@5*5(*0X
zx<0keD>%9B<|;+)=;|~Wa;1u^uasSb5E&f#S#kA_4W%3;{xC<d!}NbA$-$vc8wJg!
z&S(yNOr-qcmF346QkulJTqh;b1V6%XbG}<Vx7jJe=@gp@_XelejHP8O@T4`85wAm8
z#|90*QsEMSPCzI6J1v8q+~$bKq66}I6`HGOE4Mbn%eLRiZqwDI$z!jU?G#*mgX}r@
zJ>xVM>5Omu_SY1^Z<owKE+<+?6h}U~OU);yxt<qLZ63(3o#$i{rg%oIEfC44c-H3V
z1|zcX7-Gf2I>Hl^7bxe_V)!d+%rs6smfTAZP4rokCJ;#z<?IC6Bl22KyR2>jyI`8Z
z1RgUMozR&!#gc99{r<~yiS$dQ?$@SBBd$M-$n_Vgz3b_|HlencokUmvT}e@hMsuAz
zm6T4CiR-Nu$KTYcq?Fkvu?*FzAVzj`5aTU!7-GmQ5C_^SC0n!kei_;GiU65^$-(-h
zNx-wz_eZ~rGyxK+25DQH1c6v0Rr4z7)I)N@;h^S`y{0U@iQp^=H*847Qfs1pQ{5<?
zBkTy?bon4__=^*{7sKLh%Kv?I2&14J@AV87(kT*ia)qBLGqFd8<~ogX{1PBe4%dIE
z&qlc_@Yb50deCISILL9m+Q`{6ARgh<LcM~)ck4v(Nlg1BFBtBNpbt0JO>$?Yd6>1;
zS#-4t#1f9P4Nd(O_hf_cFnRnnP|@+wUo&po81F-t(1H~;Wzl^n`uQ6x@o^({Ey7B#
z#<eWIiPr^v)h@<WKcw!vZ#-;U8WT&IILGlz=xwI(1gi_j4ruP(<5vtA>CNQSGBmB!
zI3F~qmPreUnYxyysj7XJqS>%Wnz~R$F`Tt*-4oqaOn%Z;k)eJCUVSsxXhxr)rG*#B
ziot5QUkCT~<vFtFhXd5^#&K=MwF&shaF+#ZQzeFSOgMr#Ro=I8m?Xy>P)1L4eD|Lr
zgms90xlL}#x$ZEvj8T<Od0j?}u@8(#1WW<3vrtmugHhNE-w8#i7jY6x;?9QqS1aWP
zkn<FpX=8F|mcq75;F$W|TK;3{CgMb}>xum7*VI4dxscz+UBhgesDw?BHk&(}FZ=U{
zm;)+JCfd?c7J;{IAU<Vn9>vVMT-w`ZA82g3EAXoBccWJu>twpo^<_3t+hZxDM>O&k
zf~u?sZx&~;G)cNJX`(EtnI`op5!L9O_7mqkSwVZj4JI{NyZR+ntND5z-Lo>r+hZ4_
z@u#qj917Rw39dg+rbvd+<F-B0w5<g)cL@Jgwuy2e_?N)jX^5k$q6^%a`TGw~?h2pr
zch;8c)fO)YDh-Cpl=5gXCh?DDE5xEJM7FuBs>ZBXT6EXg!*PfuNi7wqo3+*9L&SUX
zdJ(NhXMXbIW3{LI%VuFFl!Pn{q;6k*oh<*e6ocPOSvV1+Sb3w{lru=6$OguLl^;i)
zD3aPk7()sk)O)g9`bg3>uC!`%%?q|rRN*6|HUYuT66un``qoX_jcDp}JH-APq8%6G
zLE3*kTK8z4!BHPm&`}Q|ZgViPY7<{v+SHN12G8WA;>2yc$z7akZB_Pm(opv88lK=~
z+&G+JYj=%|>Gn|yr`$m%VjD3{E+yOC{BMMy5Zs@7%7G9bF{6<e5660L(FsN%yrSWY
z=0)qyUa<G@H!upn3><^~)S+d}#4(9kq}S&a$z_Z*QrmaZb$Z!T9u%6VT@VhWP=qCg
zcL=|16%p$I0iuB%=Hh?;J~aj2a7l1=CNu>#=ZHAtsuZA+r)a65`d=zj`uq)Kke;+_
zpAst%=}34)SC}24D7-q~>NfiVF3%UFw0Mn3d|@G8EV5^=Rs5phu&IBBR*KUL7RC0@
zPg<$wR&>1<Uj)2V3i*BRq3HmVjvbSh<+}gCa-Q-7-i?=K@*_Sq&wd<-3aXsSWRH|7
z>aEAu8*Fc$?!5@Q+-!hCd7-W!?g;)_kIqW#Y>ITof}x!8omQTR-ci{JSxUHUHXj}j
z?IR*kf>;cVp5?lHH?3tdydKMeJsvQHF8VKH`0S?MJ%TO{UIF{mHAE{lRBX<cok~53
zYsp4Etw&sPh(!8mZo8i5p1$7{xJnJ23afo8)_6#C&}HAZgl3q%QB;;B@lhkZ%&4=R
zHqZ+aVN-Tau|%NU(nb<N7Nh6K-AQa+po=|H$Rl7&6#BMBDtrQj^dbbT2?~%1L3k2X
zQH7=?k1r8S)2^)S7<=WwJlcjfz*R2_%p;J;-AODcUG)044<o5QlOJP(%3tq1cR5NO
za7c|}@^HGy)%uf5NpI*Ub^E1#50wIP^%=r-a=(eBI(p&$t%<7y^6#}Ot??AECfj8D
zB*P3{o={b;WC31V0s^(sIL@$Hr=8#u??zWfS9ylL*oZ)2o1|;B#WO}v7GXNX(CiV_
zyb+VppJ#>((5&F`datlAZZmU8&THc$oD|Rpn%X~$P?m~q6Ie>OF;XzTL#%gaf|N3r
zCeez~HXMY*&V$!J=aTCM>o`YlGwiaiU{V#ncQRv@(VMkWNY@C&C1h%Xqx*ihc(a#u
zbf@5P5jFM5onIKTa^w#^^D~CVs>d~F(@s|>KPVJhHkQu@%i5%Ba}2+edc69RIi93Z
z(C<VfXj$WX#Bw<!Aj(f%xdOKm>lkfgYE00Rknua!KIt|jt$E+YKP+SH0qouSaMg+q
zR?dw;Y{GbZD>aJ+Jeeclj6lV$lYbdv(0QeJ&Rqq$P%qLP9L4`O0A(w5nVdWu?>CF+
zN=Y?Ss6*w1_$>$}D4RlXZV-~(dq4(1Ls^bS;fS?C6qn!h;8HTME4oD_y)1E9Ab}++
zLl<ELNJWQFHZHJUTj(Mt0)F7)?)qkfHCt8A?KV8|{U;kHgg{kOJ9_g<V1-mM$M6Bm
zyY|KmppFFX^c1HgG@E|e<X4z+Xf{Xc+~5J<Sx8-Z`Dsf--Px&M5D6<QcfI1euVVXx
z=Kw}w%V|GSkB3|id8phQ0YA}N4}!gcT|H8frPmUJ{o?w`gfS7~VDW<KIlbAyxQ^!u
zRM}a4cq`dhJ{%%#Kn%5D9t)`bZq)$N#*{$L(%!9@38|cB6&lVwor&PI2==%dIQm#V
zNIh>d(BpRdMoh3o0#Ntyw1kbC$KVdJjk0Z7OfQXQ7vXPa4WcIQr0rrfK%u{nn`YRg
zof1QB5@5JUWOptu@+gD~_J-Nd(q-zMJJxF7uweH7Ui#x|mH2IWfJuH|O7<dyZH{4Z
zXJdZa(j}eU;wTxSH6A-acgtXro34ACQj2aTD25278fyhnZ;K222Y(io^cdKtg=IwG
zAb%@8SQSiP^a_a7Z?_ou6`cwr(UY_cBE_%<YoJ&_paqwn4)KV~+UGdXA0k@-l}+&a
z>=5+E`uN<#<qa(D8<UoShv}Wl;nT^L-^FD7vd%9sK2CS$Qhk-};bI6c?<AycrFaxo
z<FV9$J)SyIWEzymF^Z~_ti3Fnl_0x83Z7WHYcSk*JB*6XEis2M-%Vxa=Ku~*lD!GH
z(9)7=opAkgB?OA{Ds<fDU0K)4|1Lg_>{%8Y9-|#^njdw_=<TcYG4GI?jWe`kvaocS
z9^TwnQZco3cWZ$-gP3u5h>}bSAkIy0=go+#rkBnp(^C_s=OZmRZ6G7&6ua;n$ed&k
z><{O1d7x#IVm^+f`mpW0m?t1%V|G?ACL1W!w!gAwGcgg_(?;<Vs3U>E5pgHCq7mQ4
zr@UZTRGw8l!p`Nj0n!zo$eG2xo~kqP=w%I`89buBTQ}*NvN4kaqP8|&7I1P3_t9$1
z4?j9Z&3h9sh+e2Zym3Z+_^2fDeM1FpcIiM8TT)SGAr2B1+a!=2-aKA=1WYnk%xJ2<
zuaXQ-B+o_3cE2L7J^>DW4No+<y{?+I?QAh%V&yWG`_8yqpu(rI>#t?Zi>uhstRX&F
zM#q9<1R^LJacDkOoQYU50u2_2cMOi&-}N^Th1Y~#&%xnYThgAhT?C9^jqo2kDSt`~
zPajZcO?siHt-_R_cw4E5Y#U>qP+C3;TDQS;V1RJZh(i#H4oi*GN=-}vl=|p$p|mRQ
zKH<|bW~@J-`4mKd%*T4t%|FZk;XLlv{Rm`FB_9lH^jc!kplls4d!*nNEK=<p^cDqq
z={z&#(-tSAbRHAKZ7qd=v(EAY_2sdQ{UJ_tuJ#>bPG#E&N6-40ClL>eKH3lDhs6Wg
zy&sezQz3Tr$jH(qKufB8i9Q8i>n#9(nBU|@U+l=s(Tw?Sg^vlh;Zjjev4na}$)<e9
zwwQ^fON0E&R<!T*^I}*A1iDfW_i|h}abgK=I(lHl1ZpfznwZ5R9U8X3WQb{&H%?9L
zZ5d{OZ-Vj}K&Afag&r!_fB*opQ098mBMj<LZa@5@>H|v$$UvARaBADwW{1w8PcCoL
znP`w*DbKMS{2zQK<F&ki_5QVm?0M>!<Op4q<cG2JJ2b5DL?9L}yGLoqR|Akn7}cPw
z_rx%R#p0utmdjs*s-Jm1({!Dgl;x9yx6=k~4Y12_x9x^Clp07;mX*R=ZbW{(&Odd=
z@y{{FSOYug@p}h36fJmSsV$51ow{)(&BOoZw2YgkWq1TYganquD;^%WjNvc@3PIG2
zWMr=~Jh3CE?A(i4?k``=7<H<&+yHRKXzs59%BW0*z%;vpftn+VyvtoP!e8Y)iDUXM
z*!<RrR5Qf3tDOFM82?1%zMEJrZY`<X4b;lrw?Bsz?hjJugFN#IpVB5$LJn~MQS4;I
zK<0$evfZ@5q$330uJU0Qu^eh7F!;*j?C)rJW@&)tif<cQZJr4B8Lv=hfT`rW_?#57
ze~fj28M+?`K&np`{rblN7PVwJAAoN)gki!U*3Dqhe)Z3Jit5+z#$*lT%=xarzhdLh
z^QP1R$r6sPrKa4bwDP}E2^QmNP}BzQ@tckjz0_?Y994?pl(+BYY5K+s+yYX+UkbUr
z(4Y>rJx4B1aLAM`_?kUNsrZ(}ie0Yi7CBQyenzQ0?PiAls27AMW|RbO4nzk%D&;`m
zXM20XjE(6-?{{Ho$Io7#uRN&c!#Z}qxca$z7Nl()9eHth7Ln=)ojzOqk}AXIh95qA
z0reuWLu%bXyVn4;)TN@<B0MARQMci7c&39wz`Kuxb84zpTq$gn*wU0ncbe-Yfh{zL
zC)yaM*O#uFGNff7d%ib+7VwX!!F1`3GF{4>L0~Tx$9(&r1Ft}Ud{wRT^j<|xIl}aa
zG_#m270F^g@Z(mgW5D4inVRdBhJYoBo6^gQkiY*B(L@}4l<j@DQJhOqFLe)$^3&Z&
zC&+<&x(yQp1`=)zWs&Ktk9JZhXhmP>AaS7VHGEj0RG_?S@jm8I{0&g!q3?P>M1W9Q
z7QVl=-&H6Z2ZTCMw(5_b3{=JiXHHIVa8qF>%%eo{kjRy%puCch!BMELVdK-9YQ%0x
z{eF~^(d7{cB-z_ycPQ{@T{^e`J9V92BY2%Htj^9I)ANbrqo86yEr)m=vM1eU-S^2w
z1Y${UDfsf6{<x~!o6=l@*NUde#a*E~#aaud-SBw$FUlXyf-5f9_v7DUZsYP7B+om&
zfyost_@aX1Py(JKt(XJj*n#^rOCl@#QgXyRZD0_8&bK#f?w9o`qin%Ceq-+s6%UTS
z-3<A{20(s-!UIZJ>ABz9s8lE_bCC$&ej8lbTS?YcKzZMJA=8Fw5+&1Lxk(JHuYdkG
zQ8}Bh4&9?Er(Bmz7xMgS{QnBVZ&870NBfq0;36yegE=$7V-;fGh@mfoVvG^s`iV%{
z?FJK_w)HEPb0dk%%^57y^-&-@{j@X4TQYb;xb^P;@HcnsbsJnl0yxAdJWuNI(^M{Q
z@JF7vjf2ckvZoR7Bq`+62*gK-*>5>KPXcc^2MT(z*#Z%44miJmaqljG{!;1vFhC!-
zq5~toRjc@t9u5!EN*%3~bQruSm=Nd4p4Lsh&}hu`i7+FOm-Rmt@&m)Np8Nql>Q}%T
ze0tXiWOLn<@TFKdNR)+5!tbc6z?AV2#yCb~j#|c_Yo4l}Bb6sO{DACxdfrKeF*$xy
z@LgQf3zlg4)~PO~69?+`kwMG<m!Vx+4>NX0e)oB&2R<@i#UR$J9$^!v;429?lQ;`}
zg;aTJlZ;sCC*Dyi8UQy6olb}0`LYDptbR+zUO;{EgW19u&ZgcDQUW}G?nYSoaFe<=
z<orH9!MD5|lIF@!=`0<&Rxr{kmFYHVj+QPkZ$#Nfs}wRvVe>)7Wq-y>XMOp5pi2A@
zlaYTpw1QE^A4aZw5yCS18P-ox&%b`R3tGI=0gkVXs~_2OBqIjb(z0X<@Gc?#wcW8E
zz;HS(R*c+T$Zu(^ZF2g`*HUzbzNo%`^YC1wfsOkc;7)kC!kV~mlx@M`9*|Yee8VnP
zPYz^Cce;&Bn7)df;Atzm*_UgA!@(Y~QnXOCU!Sgfk1#_E*1bpo8DIA@6F9Yap^7-!
z7VQ4^QQh*<L>&uT2pP&@xcQ3<J#(lV9sBl!#hK#f)uogV(-*by&1Kq|eE8E*REp9~
zVW?x&HH&2Wh&lsnCdE=(rpBcI6-fmmCG9H`Gy3|X1EZa`7}Kq^{l?3ZNuKuCuU6U^
zq3TUOFIYyI(cRiIo0^g(sn@sZ`Cg*GF8j<kX8D`omCUMqE#`YF4<L_(bF#}jFPP>r
z<Coj$I~6re_&WTssVpxnP3J{$91BL;0?}d8b%6zWABrd&sz)4)T|Vk)5YTF=Y?_;+
zGqEw4VI2G6X8I@WemgOn^8nwg0JT1$Oc&rF!TZ}BH+uP&c$I)>Z7y0qO&q|$7`2BK
zC^nbr?|_!|vyQgiH5{Kn@*f+p@XHw_u@6_}@3YkXMxwMAeAwp@=#MKv#Z8*yxa7Hx
ztjO<??!IVzjE{63$Qiat*#!8t84eWdHS3)$A_noNBOWLgpJ_3%2wGV{ils4Pr%&~j
z*x4%xP^HZ`QdICw8=OuIIAvBa7oo5kJNGL~a`sD=(k~8oKce#;08G+K7Qn&o$P7LF
z8|%q(8&kjSR$7u<_8#3To+&=)p-<IqX60fsQq+zPev0zh*p|UV3g)=S<01P?p7}k!
zQz7;|_nBD;VIi6B+ZJ81no^fg&m{*kik3kx4s-fZlr(d(8^iP;`LAQtK8qwtZ8>{D
z)(Ru@d_~xQJFzWX%C^Q3BaMf2FKmw^g(;o;&PXA<)QC_jhJx$0ELjK4&|`K?U4}<B
zA9Tf1Fwtk;iDtvl%0*s&ym-3D;RoYw8N_Xmet&*@>{u+TCzRET5T0?gA?pRBvJJKx
z1fLSN>J?WmWeE(Fomb37IHWz9Lp+phtx1o+NyB{bfE@jZSHi{!DjZf>ycf|+`@oLJ
zUhIYH#0rKHj4)Q3oF&iji5Hk<o#?uX-A?s_!Q1RgecLwACCA9Ri~Y$qx2O3qtoH!m
zF&Ut$bS>qB+R3!<rYEi|lfi7B52x7ZlEbFTrj|a_=fGcJIl8$(-W!ozC3*ksJe%oC
zu#n!n-kK%=al9LACMdbVcM3_Y-LWdY7W>+X0P&h7r8MQ^SO6dMLHC)lv-bU!K1EeN
zq&7CyB{fY2xi4*zOFt)R$1ffQw-xxPb!U&-v2<UWMLyA6qy18hxk^_9&?unaEK+Cb
zD7cOQnK)3iIVsZaH0;DO_GZOPz@C1iT366;yk4YPdsW*xZZYV?DQY9Q&gYPz!Nff9
zi-CeR)fo%5*P%cM^nXaev38q&(iaT)2qPNdU+^TA*uSv<RFLfu2e_CpUY%eBA|Pv!
z90;l34p-S2t&|?%X#t3i16`kRMMP?{eXVTqAkvw@!)8VSE_Rnk^GR`1#*r4_Rmw5%
zpK8fh#9dwZVO!C=KSvCCd%`qd?^!s9Y)*xZT`V>$apZz`IGCqj2y&mC<6q>f{>jn|
zswnJJ*j4IFQ7~6b`)aDCh_XFcwb+UR1~46yJ8Ee*Z~BL6rGOHLQTi%dW&`*s#gaZy
z2{SKP+-0u*cvoe7)wUx=DAsL!b5yfgz;p47os`4K;0S9<7tjtkX1s6gQwVJS;pZ2K
z-5Qnn%hS~12>3^kc44$`$=t3i8>V?XIsdK@9RBRooL%zV@YYq5<xRTb<4o!*Q-tTj
z%10Z$pzQuBCLQ$14G|*x+wt2%ypNl@O$LkT$Iwb;wxTzgUBULhq8zQ#LarrCU<;&?
zRL*php<(?#9Y9^FCqR;0xwpd;rvRB0xyoaJd}V~GpK>|EfnqWna3Rp)eksZc6E}dK
zZM>Urr_+ZoEcEdkHci#<MbTat$cSN}+)%;f|0D!0u(%wXF<DX7@Mwq>RpbJm?k~}9
zaq4_kq>AqZ!6w!krP99y6bfQM^S>j532fW(Rp8xr(SnUDn9(3T(v76tB<?y^ZcCIE
zn-}cgWyPej`kHlLg*pT5T1>IjdCL`e#86YC#=a-S+Z*<DCdgm2qv`shbg9PAoAMjU
zmYqumOjhDTRqfB2)+MyTxvMPu0NYO3Ec|Qkw<5@=zrwkvTWN(K8lhUnJuQYiHoH?%
z%et29q0bM8MsT1nYm)X48}MCOB)ieg7Qb$%Z}$KU$SDJ$M0RO_gZJ>%e&a1BBp$JO
z#thgOUed09!SOamRM@dN+IoC8KQd@TA2ZhW1k^~Rd6nbLWC1g;JbT~OeLZ@QgB8U3
zoZn4g4zX*x%f$MJNG=KDqcS=xOW4+gS}Qp`Q(~AoTiMT<p+S~R<Q|Cg=P{jS6?g#P
z=O=6!lx(fMCWUdLivB`t-mQo;_}7eRQA7=jQ+dIrGEtZur#0YUA<*^H&$8D!pMhv&
z5$;AW)x{zW<B1fF*xJ^gah^_pm=F&>P+hnj*xYxU;srIf{c!kCtnmTI8L?ofAbPvi
zDfe3d<dcA6#Bl7yAW%h?Yl~rh_ucwD0TpM213edc1=wDkK=5%6&}u<#QBugqKzQHA
zr2ee@Xz<l3!={Ae$WPjV5ds~vVTNFPj=eZ4eEi9+5y=bA*&~dL0hvFKS+D%|_s0B0
z6~3S;I2-Xy|9?OrXziZ15D(oSd=im<hogX4*o>XEm9?wBMumJc5nx2}@>2%LT~(EF
zw>KL*nXg&isk}f|LY}OgYc>jaa?l`$AR~njJd)V;zShp{$XAv`87Q|G=58>V=zo>`
zKe8>csQGt=PY`(q>lau8;{}(l)w*|bh9QY}4LR#(!06h!(QwoN-1+Q$O`P%@g}f`d
z8`drPCTm+_#79Z3)rwKj5eaMHejr>V1)sf~;7C(4DnRxi`6iFx%kC1vNr!SSmTCRO
zQ1<3R9_|uY@(-X}+f8Me%Ic2D_o+{nj2zer-7-KH)u{M7PNk;+6x0AIIG`5GhwTZ9
zHM3hc)l9GB!(zITEp(g}KF#Xu{bs@yag75Bcl;ScuV>PXN<&dj<0XB?MWZAOKhhvj
z?mry{op>Afv?$u<U3VJlr*v(2jMX-vBsIf<7ZV)Nho%cG8l#9x=zIix(WdA%z<8m6
z#%;m0IhVGgR|8p;I(XLW@#;iMw6Mn>6tIOlHh_jVivzSYEVSUWK-FcP1`rRKEzue2
zx(#>sUi-^wA<L$#^N&_2Sg0w5Koso&j&>HX-9&=#0E&9`c(}=KP(}|y(*NT4Kw_J!
zN6&8tix&u&gcuy(xBaJG{NzD(MPF4|>M|zvSkpTm0kw1?zrP^w23X0+70d*`HBEMB
zfK2CU&FUaA^mt`*_j5HKva)|<=_GrH7&nY;L#C?U?rU*I-?i~DF+0Ugsl@X#;7bUY
znyd95Krdp_@3anXF=Oi4tU>@X+f{VeeKC}Ly9jgVnQb!Mlm>~cp~|vxnGy0P(t!Zv
z>XwY6j~RE@<%{E2W}3`ecPj&$1}RlZNgO%ab_^n$+?B$pBKZmUe<l+n>5am-{GS#W
zb%eThufGn)H|;<l0Cr-7;eP*=Xa(7~s99<djx0_f1%LCwKLz?B!vq55&BOS>5XT{c
zHTeB(*qt_^3R3q!wgwN>*dq-GILZ&6;CQ;|eX+#%cYfh~oZx7{{cS<Z7y~pd3w<mN
z5X^cU!1ayQALskm0Fnk=`S<FU>wtKPmqNy`*8?)DjNCWuY6CD?<s=o#9(M{c@Vq=%
z^{*(}Y|Pgy8N80nDH~p=Jv^mAFaia~r*waRxiPFN0<oVW_!L+BY$aTlEV>|#H6$Qp
z!A3|T=FtQ9gr4tE!<idI&QPSbuWDiDL?e&q9&KXz8%sfu7EktXn}dVxenxpie`Opa
z24CTknq^u~ef(&qhY`jE_1B~?rg0Q4h`FQx)G8i+u>MA<f()-Ra~5b;vYoFlom{9A
zNhocvsx!FlhKLvB33zhhKtp{9KqDHN#|BFjvLlTL$wrK<p*atf_bB_1#-%^~v8X&D
z(7O&pL~a*(E%>H{iEmVdNn*_Gux0?$qQcZ?bO^MPhTIZ>ec5JnNQS3BaF^YT{M}7D
zX#~WSBNTv68v<m!%SZIzY?Q<>B}8J>hsF9Ammw>b6T9?yj>1=4(Gz_9t-jZ*#IU~O
z`3uLSR`ii+m$V+rFh98Yy~_R~UD!hPyVqH!XSup-Z(|nQ9Md*Vcm+AyZ`z^KxA|Sa
zEOv@%ym;WA1G0E&2?I)d1`c$8Hv?#OO>mF5qW3aAi54gO-`fD<raucUh1BUEyK(>=
zb5H+%grAGK#(e_h=IAo9_H7TYm6d7FtnbZJho=VU%^ck)G^QE^+5A(xROro`_~6T2
zgK_C=Ye)%f_{js?n#BM%C+PLie%!Ibm+tFR)`(q_Cpn)`rHwZPL#dT}8Xk9Z<suKb
znYaxP!mtdS;`Q5@*3o*m^GwR^N`b1=&$h)@eG-&nZ)0ty^dVSd10;)Bw|M1S<7a%1
zoC7pmn>n?L50tanZ+-Yaf{8#J$^#pF)XgzQnxRt^bFJD){<M3%jAMDl1(<W$r}f-v
zrlz^K?}pvi?{Le)teww#f!`9mK>dOIBs&1PFpu5tNinv5wSJDoxi8Z#?pZz}QlSYm
zTdEe@B8INh()h5f>;v7JM8EP#?pOl@W}sQs`%fGpBOWo|{0G9M*B8M)6rocPLl#BF
z)xqRQBV2W`&z0dXZJxu$&mx2P<@q^1@9AVwoirF}Ilsh#>d1$RbtrvgF!kr};)Vd1
z_(}?+;s7O`(Wg3m3a#B*1+?{v&JPkM6RYQ_aJg2z)|<!tH^5Q|q@6b3+2EbjD?X*k
z76Q1(k)tfaei$$Tos(@LfYTrKq<|d*jBcD?b=7~vL%N=;$=|1Zm9&0|yk5O6o;F~2
zB?dtaG-+qMkUVGUr<*tGs(#k;bLjStQX8Rt8l?FsL|yK*hl=blLfnyqUQd@OA>RtR
zKuIcCridt~X@IW9y)~d{EeROnY`KM(N_U!C&?Hv1#NbJ0vY925Bf&zfzKZS!e0K@n
zLx#rb0t3AtGpSb>y=i>|2p62UY{8<`(~z4LsxS%=y&!IcnJ)3HF<6kM$B9E1Z0j2T
z8(n(h022MqJsIKotb3E00kFypf4i9(QFVX?%)%dK2FTm`$JHw283TI(#e>tc#=T=M
z=6d5}!sR6dSz#?@D1tAV5#mXKBp}e&Oj&)JBt-(@%2JBuBgW^979`(g@jjaB#NF1C
zZ-$kryyB+t3M7ASv_Jd#42Ry#<W)zrJ<Lh+xwHjIvt!xUVkPXZ&Sfa`QO)clC{;E!
zanmW!-eL~s{mbAXW_hj2;tFimi`ftm_}5_a;r1OIEb~7j<c@pH)S1p#*p@M|$!l0f
z%p#}k@<+t`KlQn=L%6oD;yqe$m6O=vRnJ#QZ`7llx?FLi6^y++&^YLM1hrkNipTEm
z>Zhl&Y73SCf_ByU&4%GPQ$T-@vq8c6BiT|i1zKWzmG-QbxzjI#ZT6iD_K12m0&J@G
zwp$ebgXZN2MG|bB>VdgN=9KxzzY+qP@Hs8^v+tFIlwR}AQj#sT2}{tNLzyZIYFM|&
zHoIHMgX0GnDHbeUIGY`dRZ7Z<Zc)76BOc;!xfL)U@%l*0HZbGiD+@*5U#?+p^q6Jx
zXr;j&A;lC_^HKj=*LkK~DO=fot3NMi6!bp{Cqn9a|JP!T#*2sFK5^_BQ1*HP{-TbW
z3vp+9u5l1_@c<vG7@Pw~0+ckgGHMhc(!+<+hU2Qw#sqrE*cMnZjOU%?u(;L{lbNFI
zSS?%^Y3D}nd>`s|aPCs@Pgd0z#kO&VU2~huF85#7B#r^y<B0>cUm{sQJqi&-HCMDJ
z-Yg*`>LQ3-Hq~T9Yu#rw_;kmL+ZP;9?v+2x^dB6GKQY;PFeP6qUCoju#?qBvAL3u&
zOK|}g2=9)ny10GN%TKCs9p4uCksPmu7?tQt215wlnLmd<avCAbGmjqIPr6>%ESI>r
zD$2MT@CAp{9eqOmheUAFPull<bCdmBH{46bT|_zG6KD|){Q_8clrYU^6Foc;=qh*t
z{v4;^xthzH4O$GcdoMR~`<0!Y*VmGzlbHR^MxzgwCK7lJ*7Cr}90n3f)7pmQ8aXk^
zA<~``=uCRuOb<mfacb~+m@eIV!dI?Z_A|MCtpi)RH~8VCdRN$xWjcOf{pfRPHhuD`
zN>}eBPgV(p9TctDISDPucxIQT(&Z*t{jjKJ$MMGn+qQqs+_mAx-f!YBYk9LD!4lpE
z$4{?1N_-6Jr%Bw$`}h$9IGg~HsCP?-kNoaP&KJEPn^N`!99zRyBBYAAQIjfuW6l7c
z*A6$}9UC1=Px{FUWZTI4Sx?Z8A7QR^iA*4&q9;(DZ8w!KTWWPEkRIRJJm!1O?@ScQ
zOa5pwM3FI?gXEG$?uSh7?3NU|xGK*4RHt@igb0pea~ysUx&e824+lHR34%C_$xcw6
zWhx1wFP+C7t`eGYW4}We{21D=@g}Z|cbRzqXsW#UHaKHO)e$9FScUyV%xo~IqZFQq
zpm{%@X+zRUN3&#WSz?Ozkdcp;LW%&W&w%06tED?-d|i_{>OqETMWdai$RY%ok0jij
zC{m&}-}{M<=Ag$SL!<5JO-v)MqU}BoyWFxrW`6!mjv%|_G`9#}VdCO;=E<HWHEqgg
zH!FtVz}6)OF{9~X?cHh@%k{=+=~&R<Xv1K29$-&jjqkiHkl$}Q{{(sa(tQce(~3ub
z)A~vkuGM_6hnCMJ5~kyzBdRcFkCFxyA;_cmQ(Wi^?;4e-wm=*BTC}fHteny%9^tiA
z7GhW$pmXdz{mY;U>24$?-!Z2^MFun58Y+M|65+aM*AS64)85;Xpyf7<YxPK063MT=
z9bnQuP|)|{yg8U;toiz$$t9d#7flc<2Le=FE~w?F6+E?<51VfL9F=x=;9TEwCXt&d
z^h>Z8RwfHFw_^w<48N&?@+hOQqLMySpBbs+DchDHQY4CO0JZadV!w>ZFBhnK&5+GU
zQ&(OWk9KsucT3~Pg1-rK+5iIs<^>9yrCHF@op@Z#>^i|Hszw2m%gsU0L^cVK@f_3U
zs+r&6N;aBHDo(7m-qH6qVnegF)-k2bzvpZau2oYw;pB4zgZT7)TzW5Q&O4G^Vq2=Q
z*-k$WqH456FRs2(V8w%}!)x{oRT@QZ@|S-^YBh(7)>&!5w2E`<fbb>^1lLJ)i~l>i
z1Mp^Sz~A(k)Q(<veU}hKr*Em#rewcEKxdwc$2n4RYm~QwjK^lH7(armi{?C;9Iq7z
zA?HZWgQpoD1@FU5nI<$x=bg;ONJ-`#yB`SMQ<`blz2#nkAi8hSLT4e$(lz^2{p8Tn
z46Y2djo58m*KwHFeX8S}wis~DDrwqMIwuhpQULEzON77wk1J{dghrdwI_u#>t4zga
zE=F3MU!%{k94m`}HYcbP?-sAtd_P-)$VJRCV<rX`-2H|oIQUmqOkYM48=fO^FC58F
zbmqS)Wwg+!1^2auga`LUJ?wj9KlL&toC5No2nB*<PtVYrs@$>Ie`rLkmTa_8xaVvt
zWh>2PbbvR$>EhD)mM#njJlK#e%8@R1MH+pUlk#o#BIqfDb)zG8ZYWdt<@yJ)$-|K#
zgd1$HgV2p89mf=OlN;~=kn!GoqFzld*H%9Khe`(n<RBjQrc?@TQ*VG_OsEj>Jjtj9
zCRDZFu1X<Q*`0d};bDj;`z~ftmP^5c>!(lHfDsKsgOjsEkjWuPbSOKi-OA!VMExZ>
zURIeBRfN{qP<aJu$ljaE&vmpNuY+bg<v9oQA%n^72dtOCw{U|@7<bI4<PUaetvyvN
z{FN&ZyM)ulQ&IA#cR%Xt!~y#u)r1d_0D{+B-{jclE|QNw7w57yav%p{2VLk14e#Ht
z-BeDR0GB&N7d!bk&K+wtVe6Npw6V5hJ1K29nmKnP+u*lp5kdDTh`p3|3@Fp_9?j@u
zIsBpl@hkwe3FFsm&5VBjHf)Rv_^_5xTd=5Oes}uTUyUug@g*kl9)Rs=`3M+hA7tL)
z%=Q+=pBP*`&go1`_<mtsy%y@P+_CuDdC`TH5V!l7^#z9jg*;8`Z@4fvjvq*ixAb<p
zODrbDv)_D%*7LvEuc0>{q5$Wi+!BH>FY~(#-)pbkcbw1*)VzEj7k067j^Ll!zr#F@
zKRYNsE*U%H_=nK&h~uM(OscmkWkLJO?um-DB!kluhJP^++uHypx9O%cmb(w0NL5}T
zw&`#t40xiO=Unne;(x!R8G_LupB_^KHR_bUKne-0Zw_fca*mzV1MAZz<+?o}+7qKF
z-qR4Ns-{~$W{jM>D>uZvlQX#Rb!R?P#6oOM5)lZBGCyG(IfV{2WK)A*4!{hje4Vk`
z-`=meUs92Uh1%<_yKJ9;{M&A_YTeG;jkxzrjz1d34jx%Yq(^8wVjnMKAX_<b!GX$}
zjq;6dZeZ@VvHWg~0UDl4C`y=AWs$saF&^{N3bjuZ9{zKPkubY|=5=5glX5t*+ns#F
zgReb5)ky`5TSeputdt)MEjqDMJ7#_?zinp`vD~&DZ}MDY=7i)TI`7tp7Ai*EK<2Q0
zo)uChMT>hhcbk8f^xytAfSuf%e?d#1jAEsqLwtgOC@F7pDQYr)*9Gshvu;uFGT8GI
z$N^G_M5zFs^nXXC0o`OG`I;Iqp4l+KWBlw_z%=>aUNEJL!J%0<rKlFNxKOt5vZuci
zEn?=r%6WJXfa!I9l6+R@6DevWPGarr>*L|$bH2uGA9t*8ew;D-v@^k0ndd5WMx7!P
z`m|yV`Cf00Gv-2RRQ&48ZiK^(TfnIRTkLE~@=bk3Z8+IAcW{{S3gnfu$j>236}e{n
zzZq&SVBP>}_SO8pXBo`4mH(E1r@7raJ+2y2j1SMqnTgJV%Y!#tY5*k%lG4)JiN^lm
zSHgG`zLmR<6h4hX_`H#x9!|>)h1po8&$=OOrkN$Qg$-CXrpATN+1YvaFmUepT*t`h
zWl9PyH8pj1Ma8Yt)6>GjLdD0AiPY8Ag<U5lV;Sz_CnqQK2ns%3_c52)qc?5d(kX1;
zgDKc05v9gPmC5Rw&-0%4f2w!h4;wt&2G|@Zn_%Xlalnni3>}f$xF|*YnD)Fu_cFyZ
zdaS2iJmN=+C=fVoA3h;_R)N_<Gl@)~2YXy&NmirlI<^F~FF?WSY9*|Sapib%U^XD{
zRFjX|zZq}}XIGHw{W!NWNeKOwXhgqH!=9a!18(r$d<cayiMvh59#02R`1tr>>b$s-
zu<=MV+VCV;eq*xk{#?!e9&qL0p_SDgd@{!8ca&aO9&L<EF2vp!4mmrhHoaaQ{XAGw
zDzSzG=h<I$IA8t8TMwkHs#N5aSIwin*QX-igxpn{X~;UWx|}Y)MecqRND}#x>hW{E
zJ_%aj9iO2^YtU#ND+W?zFwO<cDjsWi#KU*E=8LN|tpPSmzwQ&5J$@7?-fal_H3;PA
zf+CdawQRtzPew61cauMkda(llZDpdOkQB%;@%87=pOG0de%9Alm&P^DG{0Xe<oswZ
z!*9ARkyGs!A?_d=owB=5FCQ$~FT4N5)T>X7Udq$<`+DEALl7Duv(3?VW#=c!nXRvr
z-X~2!G*j?=S%a&tj?@%(!<q?%C%D%0B{Z<M_%{OccXOZGp)4qv1NX!09wR(@LT{jh
zzpD@A&IcHD@gIIz6nBrlrAwxg%d{GMTaS2EhG!?SDuOxqdxD7tfBZ@693@%abg-$Y
zC8U2th8BnT#~><Obou`D)G443<%8+V8~2E(gSO>@kH#I0zu1$Jv&s@?4Sd-Ap#;3=
z?#V*J0>&mm(<wt^CXhG*jtvjBYcgAzJWGxw@2}e&iJh500kH;*T+e;b8l5`&MR(R_
zA_OAqYQ*-;rrcyv<HEH*9?NIl-9+vYF~KygjsS&1`gQ|f#f`*&hnxb<LquXm`L{_!
zhFhX|6BP}4taj7EvG)+2zuu`cV}p3&`O3R0=2^3U0V|C0Sx3+94=8k#2sMT$$e$4c
zargaIHu|xTk64PJ+3RQ$I<BU!cP!g;dt@3qW+tmA>&!rJ^A<R}Bo4&t8szfh3)wZw
zI-z+x2A{KW)at+ERXsFOPyz?Iu{1NW1!bENF8zlRc%qkQ{J+ZH-;N^TK2Wp{3VbpV
z!#)iB{@FRavmIqA(kBO?z$Lo5$SW#$`x%jsx<hp4LyAM2m`WTb3D?foH2f7#!(5H$
zTMgsQ|2&VS7lD+wTuI}TGk+Y}%@Sp99|@az1KG_Uv|;0AC3eNpo^{aU<|Yh{&~eFj
z(f}G87Y{Ep61H{X1g2*{=gf*9Sp0{je)89|$aRr&>G*k->n@E;M`)XVTG>Ac%$5Qp
zpezP<CCo<TGs=wyNcFmM#%*BQOC)o4-F~W!onvQJdMy}^8Jh^t3{W{)o-G1Y8NqPi
zK?kt!+TIb0paF$zS@Y0SJK4h#Js_JMkV3ppSs_OZtU((hrhZdCk69F|YaxI<-*H;7
z>Yk+)Zq9ThZ`o-jo_v~rDe<t{Cc6ZAQN)?15#~P4Pu3DL|JK=Qzx87PQ@Boq%hk`%
z%E~uI0%a8R{a2=xl-KQ#IZlY#)uNI2D>Owb`kJJWcOajC(h3KDcyUAn0xn#AsiZwk
zB~W_9FWlmJt69XK5vaQqui@SCEXK{jQklVfk4vE_LF|Y3;=ukZYiN)VlaNr!BMm6B
zRVZ2)bTe3-#}Kb@-#6~eJfOLV0lPYU29fRGnO1BQp@KqWj;>2q6F)EkHzYTsM+5yG
zaBh11*6vKz{vEmz?ZK4k6g{w30J4k-VL&?&=x~t3SjIeL)Bta^xm@Gi<|EMJRgsKt
zng7{k(&n$3tXsH`vUbzpBo?KxKW2>KfAOR(@_P=?$jKNdt5EX^M`q}mlk6@qq0<r;
zh7{u76o*edN2AgIL}Dq{DJ+AADGz#u;_J#!KM*I;Lm}GnVQxGx8JrsEBB`>S2;THd
zFUN#dxQ6`LUx$JhKOK!wmNk!pZ=&%O%ljgB@UqgPCx{!6^s!RNHVR#!@XK=y{X4Ei
zY}f};HFg2~0+$&2vFd_6ux6+NK5g=X8fQWqKs15Pe$?ww)8OLb(!re-9O;l1LMzmA
z@wa)%;V8^PS2Z~$CAYNnUXjgA@i<u%g6I@4&m+1#@!bE8Hix$MAWb&Gu_cXbJW$bz
zl>6WAy+hCcc<M~FA#;(xy@`;=bYQiuY9KcQ!pDKs6`bB=HU5O>G?nlvfi|IS3$Wc}
zumwa#T#G=}ZY6%ib8}Ie>9N}w$RI9e>}R-i7*kjBQQH`SR@-BPjv?-rzm~sDJQEWY
zZCyXi1>lu{g0-_hL*@<#z#JYbDd8M#VhOSaUbeSC^p|B76%)$=kPLue9$wxDp`r4r
zX=xrivkx|ATG+~)&-e_>Okd^a=fh@9UP`W{du2PB=ysRMH?rK)`bqO;(?;)3*FDqc
z4^08jI-rY0pV&-6Z{1ACD6p>%e&{EA)ExGs9~2va%4rgntbY1uFfn(l084;!a5EOC
z1ZI>ZNfdbp5C3oENRBlgB~TTzj5v(_ZY2khO^D)Ore-A*V^!uY_z_yCmE|>*V$HO%
zysL5BWNFdc<QZVidesA;*;a7y>C%T=l+03|xM>n@U2RbWbNi<iLC3BD3R3}z%@G+z
zKw*7;w0Vp2U}Kz0KtKS4a$^Ir3kgxk1s*5@8w0#L7!(~H%_AhV{SN-tMhYka9!tqa
zTs|;G-V3j7+SakjD-FDFw7M?<HPrU0R+>6Vcd!j4;lPVW7<eZ;Yn=sC(0p*dA~rH}
z^$Xp)r4&v%{bJ&;G>eB{i~med<ZeB#-ye7}FkJzYr52=Y`wCL^1?G;u36Nc^3`JFa
zdqBPNb+LQ{Sm}hpfgl+ZXPv?Tk<ZqkWdriAF?>zc%I-;o=rM0z_PVt1*5s?>nJ~7)
z!$VAipU`h;W0v*5MMg%3sj?&ajUDsz)+3{%cIWS8nsyhu7nVL)r~**;URpled-D8i
z+6i`xV)v{}vqz{{F`@CHzp|2aKKezo{Hiu#1w;Zx1TqT&_1vDxAGjXcNo#SX*pSI>
z>zT(9ZpUh47YCXC6S_iE;<a)?9aUDjJ7R=_v~8H#VXL$Ia=M5rO(V;7-}s(F(Zr*#
zDVM{dzd5^H1&!lvR{Zhitj(@i->L?awZY<Z5Bo!7QxShYzzn6^AMd3>AP^_BQ4~A7
z-E0dziJboXeG-)c`}U&Z;`3hy<f8^wejMHPb?RpEV+QJT<v&W;M-pKd(xI2D&Liik
z5#elkDl~;X{K|@}yyYxc=dB=#0*1NGz?gfxO-+Z<^)er}^nK!T%IKv3^BzG{f7i-6
zbW}}i*Bax_JhxQeDf;CNG-r<96ttu3Eq-6r)d#L#{#6Ft3GJ^mSRSvop3(ZG@b?7Q
zxlfZ^1&f7YG+X?~D7_NZu3+&mo4=7e&8_(rK%cCZz9(bqefY^)q@M%i%KrRkeFlG7
zEd;<7I(7A*GROW}8gkRc<53RB@HPO@bBlI`t8<ADz6(9fbmSeIeF2#a*By*15ozs{
z3jMWa01AVf9bRbwFk{`~ydS+hg@4O6qxXh{Dr8y%29~6lQfR=ETDUX#!fczv+{2$s
z`YJV%0-lXtFwU{q(yCGA>^A1L2^=UczlWbVQ^Y2~c-^1Qcu3Yb+ooDf9i0C9oKRuY
zP2PBQwr<=KD(AaB9habRC5inoibzgMnw4ld)%<xMPPJ>_=Xx=KHaN+?0?8L+32=jd
z>5SmqoxVsaamRyYr7xuB><}kiJsSRe9hm2|%I^c(qwJlcBzIE+3xfVrwj(4C?@H5A
z_TssN*}Li4^0n0T$L(zaIveHPTN%d+kXs$f?$j(c%~)bb4XCZ1LX?^)Gt8y)LthPG
z)d*i$o-p^8Gc)gh+HeFiz5|Vb?zWQ%%`PM;be$cpKhsQ8idVSuzu3OM65`TK9l5iT
zHuPB`WuTD|9td(FckA{#m;YLhDpS_fVBvccl&!e0sP%V&9!Qi>l=38M6!4C{VzLP%
zPqim<+P^Gs9lK4{Rc`iaK|r#?_S0*G;*8<EjWqw66CbJZDS3_ck`PQIo>f<D=1_cA
zkKU|~_&<{M61eZ@G&M^ATwd8X%xh1M$QeWnR2Z~{K-9a@T^;oL#_{E(>Cd+*vNH6*
zqd@)MJzFnd_)W!uBzUulMSZTR5yO5^;Y!V)TTXN9(h55phFK7sE1ck6Ekv#Cn)f{g
zMS|}(W%J9$ms9jV410Ou^zpNXlYMW)<!z215bUu~$>sPZ+d6@>7`+u7)PLU%T(S4(
zQb!^Gx_H}W9n*(HyhRgmGfuXjG<w7N_YLjub7_x!;^EaPJ-1<gFd`V14P`XG<%H-4
z?BILFjB(IzmV~Qlryp~ARe%}<7z(+T+#&UE!GwUK?9bozd?n&e4~M$e;dw3hemlxD
zZme+QFFW6{Yiw~NAyyCi{C5kXdeJ|SNx+tAu^xr@i!@{4FSsa>vR{0kb+8#SE7#gp
z#k+_92)H35rj${Ht^-rGjojg0xF>hQiHgcT{GR#ZJZwc*XRUU4_f^(xGX=%Nfs19@
zP@@FpjRkGHb;<``tuM8hC~Wkd?q0ccnx+rCnf5yTJ^Z_f(wX~%ON`39X0EUNHn1om
z><49$-V7N!extJI-3!3FCHHt)jGE3x4L42+U#5|NZmi}0-iY^jphawX%J;JTFe9EY
zyCjUa<vOz~=<3uNAS`J>8*%wRszFC$NFWlsFD|IS)*1%>n7n3^0putU@xZt}a6<s^
z^K38GXMw*#F@)ZEI|0L1{@HWsp$j`-hDZkaTX>egeB!OCsrk1;2A_JCnl;d-iNn87
z)M(GECa}hl2-bqAxSw{`dMpTAwd49X)k*9H`Dw_7UY|-|ti2Dz2?UMiCGX%~%Xeyk
zaI<e&j>5MnL2nKH<pzE=kY6+}*DI8d9<%mGiG)oT*;SDX*Z$i?URzUMgn{&`LRg4S
zEb{a<kwTF%F`gtXS(JAReh{Em;HL6StzkHl63IUGz{M4Lj%46M)}>O8z^x}7ppOc4
zO4|V{9aa`pM@Ugit8J@VU(CJUEivKWFSFfpEuR^3^sLml5&}@#-U%P=Gin+dtNG4X
zALW9?m;2IQIt}Mro$SmFiV9#qptnY&fIdf>d20hfi=vek1K<9eg|cQlVCpD#2l+sD
zBBYt3#lLVXMGg6jz@>?>$y;#W_5W#UA2W_U;qV_{fVZH}Pdekk(g9cD>Y`@`V%`ns
zMg2Y!Hj2+bkE}tDlN|tI)S*L!+qV1+sS34KydU}cD1>tb^P<>81sBUm(ZSDZD&fr9
zEN07CmDAkJD%5mv`?p3*<~N#ejlg+UN4`dngHPI8?KJGZQQ1BJ(I&p_IT4k%VG7gz
zq_E3X#*5>8I*<iPVDgb_BDauz_>Wm6mGA`l#)$fBR+oU#sSux*;9#ck`Pt84_vwFK
zN*z=?n>Q7uj#2BjgYvi4s<zDlrVkXjJ3gQ>NTK?{2@D8#whzkx*Q%|bCxb;%NFvrp
zbfpe1fFWc`MQdDSxXvr=tlL+II5Elk70{&Al`_zwia&+rV}Do9gJG?OL^am7hV5qT
z+)BX)le9NazE)QJLc9$5Z|LQEn*MQ9qdmwcXt#SnVs;=4ftRff6n;R7kAaLpCsGUI
z#|$il^3QesB+89za(4#?nDO5dENF19O#M4y;=_6IK!TFe2k#<OH7Q+?VKNit5kSFX
z3TZ=upJ*PO@Q4-7Au@z3YUF6|R@7{}#2MQd0Fy*xxP$CvUa-FIAh9;OI5EdzpWvV#
z(F<5_C5!pplYW`)*}0c&UGKHzUzs)d`s|GJ8Xc~UlI$<{b>ynNai8?)w(9*T@~kP7
zCP==W1!_3oR3GAaSYxf|bk{o%dL>T;Vj&JS#QM@oi^A`Qp98a5i(Udis~qx{!ll9X
zM<ewANm3o7j+^`M`S1P*WxM}>RQ+XCl;8J14$sh%BHapa5Kux81f-Nw1Qd`QQb1xD
zQaT0c5~V~M3F#a{7)nBrMg|yCq@}z6XZ+sxPd`4;wOB5lIoEa0-uqSO8lDcW7;dOc
z>7(jyQA6Mw7H7}ylb@i&9J+Z*9v|06zvhc{?Myh2bG}`2WfpvRT|dbC?n1}4vH3ft
zP7dEK@1}w=Lu=CS^89E?<}$3Kqk~-Po(uo^>0u{NUP^%GQz;rSlL&e4Eyq^FJ?6Ju
z4fdyn-Tf}}mJJ8#@RJFl!!wn#@<~s`zguz45ae6DZdAIMJUYEO1NX8lB%km3y!Ov`
z_@8$uyk-44w>MQl1Y|FIDX9by663rWz=o#6_>_<e<lCWmouU)KY?ao@5#UoyHa0W#
zrj#Nr;A01C$hD`e65SC3sL<2RCN{rQ%#6hElo1hXI=Xq#M0AKRr36;?Ns!2ekoR`G
zEJI3s57Uo^6XLrBc$Lxvh7Lx*5fj(H$o{3)#{GVpT$Od-eyX}?BlxstRV0e_O$W`F
zH#H#Z1*-w7C%?$}2>!tFmEllUOC+}I$JBZY#?|Ao>p<vP`tMAA2J&5`Jg`y6xFwt2
zf=DX1>z-!YIorN=h+oGR)n~fV=}|`h&L>e($C0@`pe=Aj8@kPbPII%&Sp)q8Or$A3
z2jjIzKO^q=U7Qsew52z>Q|X`%@6N&%we)Bj*q<psu&1!cZ?&$edy4mLl?89l3zh~u
z><^Z?m?sk3{8K;kF|!o)VWIiAW)Fq4e!coId(aTo^CZJT8czqJdli1^USR7iR7U3e
z_E;P_>SESg9zV;BIq-=;XAtbHVI3^1$$^GUB$wxVQroe(NuJOpw0mn0&#l+Ybg*3>
zDc#nqz>`idmFs&v)#(q#b=diKbN*+9K*!^osUy?;4cXXNN@f~Rn$s8LI}yCkv~-D^
zSOtT#c-zO2!h+HkI0M^NHZ~q*D|2+CK)8v#;!pSgqs8kb|7UqYJ;l@-1h$4<HNc;}
z3&FQLbuC7AefwVSBE}CUX(Fxh0+aY+ytSi9>P)FRms%=SJigZuBmjD5vxp7ut<nQX
zj$NXa&e;3)myLQ7zl+^;@wNnsuL1ZZ4&xSya{(m$vI+_cA!(<F>Ba4S2v4Z5G0927
z{%bBA!xf@tX?U{J+d(*rK>1ut=K0~HHv5YcBMiVqq3$N$pUHx1G6%FNSi1kO@y&Oh
z|7Xdq#!>!vKM<MR@i&{hHi1m!P#DL-!+o5{=5Zg01n)Yr<?TwOax|p3dGSt0k5#CB
zEuZt(dBC&az>;~X$5sa9_D1)rv_3mn6^1o)FhcGqH%)kPC{2fL$12_ZZBx?Ow<zIl
z+4V~D-cljUwF!V-{2YbZw>Nm~Zhe_KkIn?nZtT&2f99B+zkZqhY`@e`!N<oZB=wY#
zV{vvpNXP7D`rH_-hE8cBd5%V)0l@9gS)SHCh7Nop^rxyaTDVCIRA#37>8$H9Nz-5V
z{#Co{=22iDO9LCVTCm9t|FqZ`ORO*@ly!_C`)Lj^i2H+Z2LrOT$yPCalQmt7!dA;j
z)-m4Ui;hlKN~;!L*nFKcxviX5oig;}DtU&EFC}Y|6RjPaAI=GF9M_rWO`C7;IFp1e
z$?a<A2!2mg-<n_I2>%q<irw82{Pjd6w6~kFZ{gj0_uRVbZpQ1FwUz7qjOK|L{Au6b
zcXB=;A4ouEG)O<YD3Jq+&4ffBMlbuRe6zTz6NaYptV;=inRUA>9WMQY3sfzSv>?h9
zcFM2b#u1d**>+zV%cULbu+m|!Y!y9=t+p9O#c0;rpu&JZ*}@POU+km0tG?*3L-m>~
zd5}?@eykw${vG2|*@0ZbvYhR!eXI8w`};3=)e0CX`+GjVtC{vBu}pQL<!?H3^4=R%
zVvu1|mz?%FG&Jy9uNK?=E?jOqUO`^i+-NuAU3CB8OVWTd#P!GT0P?B)q!4wHQZRR^
z68ksfs^&iYb#&%I71RCk4i|+#+lJQv$VntVtFcTzG?m25gC^mcZ5TT^pJ^dCo+!zb
zW4!ge;t^HV#aacKZtUO$4Q^8dL|Y7|<_i#9liFhXUlvqoJ$q{s&u3Dlpq&fXN!Tqg
z5qL@#bs^edTthd*D#UWADo!G`6aE5K^f|w{xE(b6n%G|Ssm%E+#eRi2QbY81_pQ>|
zcTf0Rezr$!|E}zJ?ed7(5Ql8CTuOO+?cw|`oC;a=mpO7{+<G`mQ`{Jy<`hafo$$~+
z8VK{a^(EcjYO7x(wh$TnH?m>%lLW%SYr12>=Y|+X^A!m3YXcC$KyIt~(F~sY=wmlC
z%t8-S_oIR{JetTBO0VT{m7^@mDk3t}B_2m|pscAHYAlML)uB@B=ryCuU2ut6-1NJ=
zA1!@iHR*f0ZUu+;936*h>`%Z;=gxNfJJF~Eks_l;iTR~n*{vm>i0vr#U58Ju_sFY7
zA5ifnk>MCoTD5M8F3~q#5r))D+_7KHrwVWKwGn|BZw+>ZV-6sRTK|j$sZa9A?4aKX
z9r*pB<H=?uWBp9Vf09n>Q#ir^61XOTR6yc4_=l>85#lfN1Cu_Q%x*6<m@kTOP7^vQ
zV{xa1Q9?KkEo^;yxY?<tV;JPzCr$`M+J0NE3mq!X$-$42+#%`{Ut*s0-t7gCrT0jj
zEyYvx!eNhdYYqClKn4=CXnlDf@qDJj#Yt_aOa1ElUVz7YF0W3qQIj_{vFq=kNR_`E
zVKq*|;?HUiLsG7oJX8WhZysNT1~8qdEde>t$u{^wHlNd1Z8AmpuV?$k!%{pj-Qu!3
zUpBzu%x?m%nVwJQ*V78F1yIjAe~(u>ZZ;mi+>@05;0nOK$ydXwNg{CCjk=|@$n?$R
zxytRu*3B2*F&AT6evoAP8S;mIhmE=8HF3C+rR9vTd<x=dQ=fVM7^i9oENCEVG`CIs
zYsnm0ajE}P)lPA1<)Jruh7B+(fam}pcX#mM@ZqLBI6@17;^@t;%NU0JOK*A5y8Dld
zGkNWCoDOgO*La0q9CF4W1LoOBEf(TU^k!!}F4x@gc;X9h>vVaei8sFirI2XsD~>w(
zQThUdL@?Z^P**zr_5P{D{tun`ogRVC`<o5#2c1=UHY4uQ8AbWFOO>5=sX@xBln)-x
z-odYrv*?HzU{YpDyk|WYhM(8>8$Sboj`mwDl)v!*T8TfBVu@Hax=CJ>XraZ41H{ab
z_0kxQ-H)=&kD+eS#f`n?QbeL`a74Ck32(wPbNGi+c%%N>E|?k3GRQW6thWIY&k_A;
zcxXrei^p<ih@fR3GpNn~WiQ?6&gFcPU^3m{QM*411Bj%C@qj|?V#~=qD#GKu)69O|
z=<cykLYX%1EDae>aE-X!a7Ur&<(=coj+j0sDM1SVXa`*;@!fyvCH0$wp98R1LORJT
z4umR-FBN!MoJDK_kBr!T$x-s{k1LtQUv@&~#_tCe&a-PYEqw_RzY&P9WZ5z8*~5LR
z%Oa{@)1<t%ew6ys=omQSyRJ{fm+xL(2+z&snsjI*)2(&50fw*f$W+T`rsCSYm$Y_b
zHk3Au<UeKMfp;*BbK#z)R$EC0jyxaZ2%$c~ma^#zIhUIOSM*CO1`-{e-I>G>|KbOs
zP6l~b{!HZe{GFjG<oOLM)e1=asIi;9;3K0Gf`Lh11f+N+1IfE2Z0QsTNth?vI2fg^
zCEu`m1RbqKl5EXzEOneZasB>anbtB{=Zxj*EPpf0*V*(Yi-=vG?K)x*HFE$WWVqko
zYZlX){?af&)bEz7C~Ntf8p!Nx!G>P^4!U(4qQ1Ro)~d4ETAmz;{@f(*hyOVx<n#Uy
zIG|Tfh2&E~wHS}xOMpS8k?NLiH9@%}QX$&QP_xhs;5Bp3B$6!yqBbUl-*xbSgaet&
zNP7Er=MDp%iCyUfYu_Ja(&-fUrrkNP<f!*{I~{{Kp}OS&AXU@Pe_QI7nLxFR^}&*3
zdxr+0OIUuFz}LQxz%8pe^`FFL-i;YWOrO19&qn}n2#4RqGCRa~4911~uXKBrT0+5r
zB;WdK3cg9C1sqs3fp1tp=H86_LuXq)czu64VrB*~z#v8C=F!Zvh2jNIj%bSFbh@G{
z7L7^V%j8oLQh|*Rd$*W+!7N?wzWrj-sfUk2FFjv)qRN_zzhUd)q{mXKm5ohzmw5E1
z6e(NUmJjKz^*}=<*NHY*UqdhIKw#)KLmH}4O#t=#m-^4aGol!-Xi|V3t^R%X+=&a`
z4qh7_7LA!B-VhQfHIBGMCoqHn#~UDaU)lfzy(+~<DZXD?DQr{=No=@#+>m7oAEu<X
zVwIHKw~g~4dFPAB)&nG$n)m{>BzV++J|hOe-M(k@Z1WX8{B1}CM#w1|2>RT8rfKL0
z&9Fr~_zphiDEWq<@9V%hL;?>aORK=&x5912UoR-)Wg^5X(50^!G5xQtF>5kz_z%pr
zAoscMsvc1W+uQqKDKRgq1dxWXkniQyrF5Rg&&8^=C~xV)cwsRV)k-B@x0*GD1zG`D
z8`hZNSPu|VNAR@G&?Sc0k6Gxx`Z-S!K6?AMVf|f>H3}d+Ilbx2*15+L*tmb18Q_?H
zOeao1y$H!sY|dtZy}BrWzsnWF{bu`AOwIZK4^RA2L@t!2klSSnLf=+Gzpq4s03@J+
zJZu)N;H&_e!jSX8`_}9Ujgw#xdC0fX^(qYrH(l-TJ>dn4mnR@))p<!R#rI^T7L4pe
zC{cACEMH>a*iKUmKCe0q_VHjpLPS8j|B{xLEagpOh@0$iq&-un>l&$j<?S=oHQi$4
z1%~!SW*H)O-#2ZTu5cIu@~>WBoYyiCdmI&NDEI0#Us28+u}(J!feb*30CE8UT;4{(
zISt}olEXlJfpI|jJ0-livC$Q82U0|hDC<MiVmnyDj@hhmbEBbruq}e{NoZ$5X?k)x
zA#Z31sLYMpd2$2Kl^h0m>RC;;K*!gD*fAd~r84UZ-Bwr>e#&?@b;e_nINxx$k><||
zOMKi4VlcO4x1N7V7v{M4f1YjpYJD<Ump9~`gDpm&%muDwJJFdFj;8&cT*Kp)|Ir<0
zp6Fr-mJwaWOalZXE{5|Qs*#B>H>-%>ucYy1sallCPVZ!tF+XMVd8|0WA#=9XCbkf#
zDMTwq=|a~pxl8)8KMe@6=Gm-`zdYJ~ab7;t{y31!?CF#dUIy4U5YTd&k|#+K`<Dbi
z_ML9X5?)f_)fKEGQv6ZuG0g|#`s?)qPn^^)zC@A4ff=t7;*P1|kaD(-cS<)~qka=C
z+MaxmsiaHm4v-Sn*!>16Ik=dE?62<Sm>8`CXT3CBsJwrTjjNewJETGIR3nmSJ@o*6
z(kXLs2eln#pHWvQz7D5yQNngJIf+K$0D<LPhtqjAxUVFY&PP7bEW7MdJBqWjX|(&^
zcM$q@yE=KO?<mWNwA3<lwD2q`fGT%CAuC`Yd$9oK{BJOl_<u{FG>P86^}Lxn^gFH~
zca_vMaEv0m`jjv;B#l(!oxnbdd5&G+$Z6W~8)XV&Wg?~7JYjl+aB>80eRRwk#m~sq
zJMJ$!f|lcr<9<I`%`Y#f1OV)fjfUO6s*)%2SgOvhVO!AvaIj)dzaCAX{X7qJ9LMvI
z#dN<g@j?B(auD1N`vkBG;f8e+9NA&T2g5Htl=j#XWPDypw^rIomG#$I{P?TzQz!pc
zH%qG=ML7#%CSU(r9BVyr^aQRrR;&|Um$gQ_DOjNOb8-1PN)-{?>%}M&n|Lrk6<}#*
zjGuGMTf>iNX^$zyLby*U?4Ce7-47eUuVM8)TVJW3ege>9Mg3BmmqQCaT}DX>|2kaY
zgD<PE=Lmz?q{G!-PsrTLc;=SqA*%6}M<(l0z#R=E?^knKdV$qx7C2r#zq*kB6Aa=c
zt9EXFo<i1h0@Wpo)gxa1u!8K0Rr(5YCs~u~dxupUrYaDHPe4Y|bSN(j-z=e6Hse7?
zz<b10fvsGO)xehNIn+SVMDjxN`1{?)3RBR*wqUU&w6UKQcp+k^Zk09!P(78T+(ez)
zO`uUOlUHhOzZx)?Ou<hl*dc8g+h@dA8Ez-L7P*}bd{9{GD}D3%7^l?~vF!8L9o|h@
zx3B4v)4>~iOia&Yar09pUxkY(!W#rKpqap)ddU77fepGt!+|mTRv2+#wF{Z8DwQ40
zY*282GP#+pp8cfH$cnX}2RKC|0DdRV0lh|9hpBbsz|PPOC#!|l#$Mm5oE?TTzKwaA
z6Uu**x^kD--qbo<g+=}uBt}$kM{;=AGLj)<|2CP=xLX18@8fpird+wfUhuwgzw2OE
zQGQCU4-ZR3YiHsLLsg*9tY$dMn@1b@Dx8Cgf!=>@dye872v^Y?<|&Jk>J;5!{0_|R
z{RiCI(4SY6bTNV<0?oG=ks>UwxdN2(Qa>Ij*WFBW$BCfjYXv>)hf$X>Zs|;$LyMk-
z9G`W+J`{E3NmN@}@{~VcgUAHe6R$6>-D26l>geP@mh2a%7&uIh5JRB47*3_WFaMad
zZuf-E9{eWj*~uJvNGZ<R{OpvKA=!-S?D_`K^Ha$emPHv^zzv0k!lMCFSHUM_3;%q;
zYam5GVdLK>zK%$e$})xRVx<(VB$Wc~q3?{WeSFxww;mX{&jsaw{``2YUOtY#X(p!B
zd*Rj*XbIy-T}VVIMp?W#`&fl=y=)F2sx%!@+I&p8iK9wC?aEC?^<V#&vgHQ+_q(}~
zx8F!LEEN4iZ!^E6YT@(0;h#ce61<<xu1{WL50-AU{2E<(fEFlO&rJ)NwX3wo>Wkn~
z(gur6<SA3Qu!SMoIv<x|4>_55ozQDz1cyKH#0Ex*452ZqG89zZ1w$vT7Z-<3oq%fW
zttxpe|4mbQu70hgt>5>cvZ}Q4#V9BJXts>&J?5(C>mjIU57YCYYu(rJAsTq9FRdDT
zV)!K;J=s7LWjQ&a|IZn#?rJWD<*bFslvRVk_K)w-4w0qiH!bR6#&mH2th~6hP1}@Y
zI2?)Wy%S)9(GCeWI3$K8HjmCVB!5<LGG|)c5~sNfWyF#8o_Ixns3MI^HOxFQ?TSGh
z#atf6U_a{0pVm}K=7jRh*H0gP-ub%}M|n;bjw?w%SFChbd)k#AAA*}H$YH+|TJE8V
zDYWa&M3^t|{X3mh{I}M^N1cmPakbeB==#fiQpT8Jns*A<unzN8>CRDYd}c?!?i+7J
z2=g=HZtw_Z6dn*y9JoA$x#S-lzpMAq>hX)b3?oX$_4OWq9_V#5^|f%JQr^Br<8}v=
zu*RU0{K~G(<*rOE*j&T)i=QfZ&pa0Ktq@|Na>Y?wHfaZ)MezXR5bm>Tc}p1dSiirb
zj?-A&x>j5722m6n82JO~C;qxcEgowJVk+<5z6Eu-jn@e+Xyq%z+PHs$O;0fn$dc$#
zAoB1~1)s1lM!sYXWNY$NX}4fSDOZOJ61*r3Q?F;dvMs>GY?7sh*^_@r7)9^vG@a1c
z)h#4om-K_<jL+d50NCvG(Psy<)<<jQGYSzKZF8d?2mFwith6q!%w&NX*3J!PgsOcX
z1*@+5Qye6G%e=%f`S;uxHLv-p{n_-Yh|Ys3Rjx98c3ceg;us^^tkg3qNu6o4KC~g3
z-q}{`RQf<1x&FG$%!JA436<t6+j_-?vYEwt+@rpMm0EB{TKhb<|FSNM)rVOdz7G{z
z1E35Q>U`Zk8x&du+-SfLoVW|NcSV>31b#-iS<%*n@VyuAcMlM;5~B1-vp(Lvsr6uZ
zqKT#1qmNJSpS^UuOW(okAUH>gW~7`r$z(dM3q<iCyE5I1S_;bVtTMyU;P^l7pu>ZK
z&{k&%yto52Seh_%%~{5U1_>1z>zE`nrn+3hfPO}DGi3$iGt;50Dpc;h8E;_>2bGHk
z(2mHIG#=<L*QX&xnxxDb`xx#p5!xKKn%R6HebdCs(B$y&L!M`A>C^A&qMkO}$DId?
zVlwx1F!}*7?f)Zh{&c4<MMd~E#7S3Tg`uSP5S70veSqRL6h!G67TCT|GAq1YsE@wO
z2Ex8t_ZhRm><2=-NHbjWXP-{BKQ~=-@KJt%>j11^68?m_K=Zo><_@=?xgYJ6*OKK&
z%30lKp#~lc@z#dBgVsw~^$irRBCG*Qzh|5z#RY8|dq{kQ2h3CAsQGW}idn?PQLhU6
zg1lobC+;E(tG|Nak(P=bs$C$q^UN4|1?hxzJ+%WGM^f8)n+TBehjpccJkcj6h3toJ
z))CAMXYAb)ne}3PN4qWyuY6j(@cp%={1K`mN9$iT1?xWvAxf7T52qsl{@G~_iwUh*
zx;WkJ6hRxjNA8D9dAR>ve&P=_^2e~-yLNJ%=^g;9nvaw|)*3KR7_C8b_h;}zXG69x
zahY$~LH}nVKapQrAO9qKg0HMfU|xWu41!MK;|Y&UNUDHvxitNZ83yzY!+NPW7>6Ll
z9*EI05=+<N3B4A&;@%^!ddSN@W>CY1j!F)BLdY7AMDLb&vbF)dDjvBX%6+;4s=9&S
z`Mx#G-nX>ns6DPd-(C!_3GhI{H7pNxaB~FxG`8_O$%lEB@9J7kRagl9KR1>-t;PJ!
z1zc)};A<-Pq*c755&7Ms6U?yAW>JuHnIN)$qo__)N3Voa=&guxfX`yl@$XpRz3OKo
zcgOb+1*O#a)S}etsCMppbSY!>$4-6~3`yOKc4F^&rkO_=Px-j}*e?I<1l^>D@w^vp
z<lio5`ND6V750cpnOuaj?@RC99OaBP;hiN_iIS~P4Dk@}7A1p!&&YhUSMCmlaf}49
z?;#?R2h6Y~(0GEx4Jt12PHY4kXgB95y-W%L-+5>B4k0r0GL-%YQX&Nx2I(5UYm1P4
z;~SKtD|&ub)ff#QEo!Mj?eiUZqekdBvQqYDF3)FbK~pq@oSq-<mrEBin=aPb)nye1
zLMTFLH&J%OIBuzBg|@Ckq^0hR5k%jV46f($_w9@0WMQ04t2VVp1tdfA9S1?KJnIxY
zHjWU4m8Y7)jfDY?GKcIMt)3_yKVX*s{9A;NIT^u3jF@V*R>ivyTYdKHTN{{@6zgiP
z^OeqbviU{s_T8rlp<q+%0ju`~b=bIRx-yhwtyc()ujoY|s;gvP$u_f5`3*GmEI3J5
zU%uUr%6fL}O|`>C%Cy2FQg&4qa_8kgdsuzZMc2V=G&qMfXHO3jtxhb(!R7|1WS*b~
zp5Y4pQELD)#Una<%e4}uKx$#MZ71pO>YSz=g&o|UXH}?DCu4wq%hc})73&9A{*HtL
zwVt(FpHtV{Mh)~pgYyLYDKXyCY^c+z&7ak1Bk%*N?vvTNx<n2@sHHaHFD8lJ-An-I
zq6)Noi7jJ)<xc5da%HC%ftIzPF9r8=#~w)HAE*xd8#jr>?OHA6Qf=a!xK3He4R%_{
zMZ#1N))6D3GoA|&2B)b=z)o33*rGPOcwx?MR`?cpEbDV+yN$$my`Q^5&@ss&`0;W>
zJB*@t@6LwbDrrvOgAG-%Qo4`>9sg0c4k`dhRB|mEhYJ*cj()SuEV+Q-P{wB9{K1m`
z1B3?ameo!7PSUiG_e`yBMC#S3z2?Jk{vF?3-bRzyxscLz*@DbP5N!CI3{L#mOskEN
z?UOx7z<`<Otwn;ih`OgkxAWQ-fq<Y;nMil$*H=CVKdAjBw@sYSq2CLc1DHCPKO)gv
zSOUoioPkxvO7f|l(~Vw^(;iDexp{+c?flh<<o6XB%2S(z=6S)*(Ke471}=Xts;_I}
z5`L@s(pRrEKi?yThmr{XW1Q=*B?pL{!Qt3-S=A5|tC42W+Hap*k$;p7hMMVSg$Dg=
zW3=N{EuK<yF&voi<TmRZ2*q8{J4E;fak`sNh+BJ+`E5rSe?s(eQo5*tb`r{uI{H~v
zd|2~=E=%e0)896dq@^#@-PU5sS=y!>-=O?X&ZrvRE2;2bbYE%Y>}Mjjle3TgHyu0f
z0mJ3zqYrCr{W`?L8P+5;g3V(#C7cSUxmLR``A%`R`^y!ulC=qTrIa7ERa|#EN9*{U
z#AaO#2|MZ}*Y}Z&%9C6BkN{7=TQ=u>x)SF&q_=sW_P~$8AT|pabPxSe_N>13am(~}
z3csVas-hoFWu#1WMr5T{Vgo}E)uN^(5G_?^;j~CUBNlA4(0-FTMG#SB>E1K|luHaX
zMfHDs^6rdC!W!kzC}lz1W@cF$1%k{2pfIwcFvrlT)bcsG_gaBj9dS`4jt9WZ-_+|M
zhd0XLW(7MM(9%|hEFI^b0Wvqy92-Bp`6*4F%s)8u{_dDGoo%dd)C5pN+?n$Yx8Vv}
zN*A&wuw5KLx07R09NR9ny%k9^FyW4e@NCu?9hcsuwf&GnN$mefTr$l}%Ewcsi|O7c
zBH8OlR@@CPB>yCTbN5QH*Rt>il=~OpFi()e4x@p$m#++@Xo`=63=#C&1TBMO@i6Mg
zD45A-wG@+XvkzP!^{P%G8NfTrsv{urqUiLCO8J<8F8TcV3fjnvkDN_9C1F1_f9<Hx
z4Hbw*I{Dx}F3s`=LN=`Aqip{2RZ7}|`dq?%-~SV#X3yS@VTI-FugpEDogJ^%68iVt
z9umlIS!Y;dT_efQhLBbgW`xx!1g|@W9M_VsDVA7^!*IYe$FSFwvRQ!W0gvAM69F>=
zsFt^kM2YuTC~1(f+m6o7F&v2AqlF`?;hhPd90!2F%>#}cdoqsf8vytfh-2;bM?K9a
z>gWfZA>`E~n1>Ss4WNa28Gu8LryR8txhzw@TvQiN<QH+phvdb-T_Rx@I2JojIvbQ%
zsIE^K;78)TJ;qkNY%F6Zn+n3>PQ`<oD2u}pI&?XFLcd!|i@Qh9aPT9Ne=iZsxUn-z
z+lYmx$4HS6-x?ZbKTd$zwG2O%NlmB@&4*8vn-TIi?hREHEl{@nXnUDiJqNklYvfYc
za^%^yCEELtk6Ci!zv53;z88^q<Qm#eJWbPGn1~%P<^MvgjY71Oyg9tk55z6Ex#J{4
zqQMHfel#o0DzXmKG@M|#PXLS%X~ym4Y5;Xj@B!HpC)RbWAdJ(yLFcZwK@;1}J%-@)
zX#KYdXuI(Wi`tdkk3-m(4}rDI_?`@>=UNoolct3~XbdsaX}jT`s?Q~-yZYe@v48G<
zAX*q!SVX^(Dp&-WvzJr*f&8Cs;!jFYvrO2c%e}W5Dvq>sw&uXaxMSEkUKHyBMQwrY
zy0a~)CCKTp0ND%>eADrC1fOt%QT<nMC$1)K-hvATsAVs}WFpK5r&K?5I9;!q$0CCs
zsc9ZT?yJjBAX5JN#u0qE7CpZrnckc=n0io!>_1GqxA#j4`Q+ulk1^Q%zgj7usca)|
z5#<NLJlAK&l5@q8q0(&kPMBe9o}=ODyMU>}7L2rmgsu{;cF1+g7zOP;Uc8L?o<;pL
z$DtKq$xPc>TiiW%_{N!Na=0<YQ>svBd9t@syVXjfuB8>f{v?xs8GBR%jb`GGgfQ@x
z2ypQ|^l5tlQ@sh}vP{wl`WYMRVMa*NpDtYw3UHrR&0@?y7EM*5Ym=)IiacGfAlZ<d
zx@rMY(XXKTQl}(~ewFDQWlWU-;m&<^E+E9{;}iHaBM29XZKi!T5XdBXwMR?dC{t&6
z%klid?T!1iONVUf3FvjYrX&1m&$VI|Z{=qR&vkVE=tG42B?EPgeP}3T&ICJRp7t-1
z#EzF3esTDrr320xCMAL9r0KS@ylp{Aa{Sb36F&Yn%v}xQtJE#os_$5HG>N0rCv=n}
z9$vEDurc``ly&csm%|sUDYGC5S41#_KO=p0>!0?Jf(rCFrI8~1E&8}c?AmT0!Bmn4
znnJIAyP{LFg-%704^Fj<Tb?2MFu~e<9Ji+uQhcQttlV-EjMCP65f%2WZ?4oXB*3;-
z3rk;sK1b|)Z&=;PDEjQoy9$YXdAgGyp<n(2e*=I5RYm)LN~z>WK30P+JqH8mwl0R+
zIId(-<%YCx$JvBr-<a;#Lj1{>1`;a&ApnQIe>|Xx=MB!kLo7tx!nUkPfJ7qvIe*<n
zM2<ZbCoEfxi4!SsnP4L(e_d=1i`$(z&~Y%nS;eQwv^+4Jb@hU$cB8gtI|OVfqb4>t
z6$zSi;wPH!<)lEGUpp7fq+t|N0VJeGN#f32V8?5uIJ90_Mwg}|()ZWUX2ZFv<EZDW
ztpLqPpBvG9+P*xPt*O8CRl3cwwfV8T1)7VLyHrs&sEIcirR}LOs}ugsw4PfQ?|7hU
zE}xs>*9vCM5E3PL(G+kDiwh22r1^`^$JJL6*3@g9D+uDqvRyA1M^b*#KX59X1V+G7
zS6`6mtR(%C?3r0A+FK3n!rW(xni)usU8?M}izKQH{80<+2Y2clh-c{UinQo-j>qd5
zeG@FDy1(-2y~FoAowJk9{q3I~zt7q@J;2y!qfpYAL={6_A0H{yezE-=_}dSRjEt1w
z=bmSelj@h`fcg1ZzFw0d!>%1ZD-?&+Ui0O;Po&Ma-RSS>LKRPYBADQD{cHvGA|c2G
z8yMy5@U!55vH-6<hC6soOel3jzrX3?#;2~9c`Iwe^b`UQ6XrO$x5+e4%Rq_Yi!Wax
zwao?*@@!rZ#2KE!?;oAHyg$M823){tNgg)}h?U>@l?PxAU07CT!6kqGxt11I3n&H*
zA9nzC`~BlRfenA#^@-|nl=mtDmXa_&`$>moa5eQU!5HoB{MA5m>)RUrfSG^JDy<-j
z-!s5iSneA;XU9bf4<sK?P}=@y*_EKYUWtLTei0XlWLXH3S>!Ki9;awcW=A8)8X!(g
zZ)zLL(R^2GcJx9FBrERv+F^sqkI?Fr_Z7(!<^dJzp+u}Gf-tXzOHTt%)K<*D-KX<~
zZ|RNCjk*x5lo~sO@|4Pb(9+cnU0t<Faq8wtlJpQ->`RTvj+P)!P8f^qPflhm(yQdc
z)W}VJWpSH)4Nd^6%9p*$)~WY9Q$s-(U7%bs*|OF%oT<!wO)4@^`TQq?b-|O0w_y;~
zj>nJZ;sVUL>=1^>MvX?pR~g18T49uw|IV$P0#hyPk0?Lj<orgwGo=ueW13I93BS(;
z;_|0nkih1}?9DAoG5hp<G3zpgzzv<;&++19rKW;SfL-i2T#>?_a)aLOMRzF;1UsK%
zjd%K^S$o@_jY-lz5*<-YA;3603rPQtGSfC@%TcZuK+1Up$p1<ouMt%|9H{rE*GwDu
zdpF*B^VwSXQf{QXlJgs$3!WXfhn~6MXIlu(QkwSs5^LXQwjN^Jirn2>6OmmrELSMC
zCTh;FcOkKd{u_6e^B<yR{j<HK#*Jrf5YoQqAjMG{pp<4)+`u6(UK(&Jk%<6#wMgSH
z0Re=`02e~J1sV44*ZOtq0l)UGAf0Z9%k#temArH{4;JOL&iMz{);zf?9|BN5r&-2L
zzE}d7fPys=ic64TKZCNCngK46^mSNY_`7%4fV<`8sZf0mR;|=@wji+mKK<;!92b;?
z$+ZhH6P)R>(<=mVI$w|~qjF8l@79_6)g5I7I_yb2ZrHqyOC<f4E>`sQa;1br@Ioe#
zKvt<1;bdj~@3YF`Mzj=boVbTZ;yjy1@f}%$rn;^36F?DH&Op{9Q|Ix|D`76&?$`Z=
ze6MXBc|QQbBb3+(26!OX&12#q#CQe%<-YM{{Eh@QEQVF;8F=pk0&pT(?$ZK`l_D^a
zDMr%2{vd3b)zH8UTrUb*T3U*w*}z5y>#J#8&WhL?PQRcA+WD_oykxy+CBd$Kg$p|;
zQgNv~vQFjH>~Bd*^b46la>Xssi2LU3g}x+l;TgXUgUVK0rQN|N$s@DJoVCBa98G;a
z>$Ip>!{{%I+>HHy!(sldM^7BQx|yC7(3nSuM&duC2e*A-Z{QHsjxUBqyy)Cjd_FEw
z^6??eeAmx~+X`|wblqV6lt-4z<mnR@UrB5A0s$JxY+}<HrwtrFT%qmuqtfyz(6DmI
z$gs7AP$>1MNoB@8)YD_Yeo?UHtdfJJeNIe3iQ>L>OU}rM)$`!@hxoOr1`lCioPM{S
zS1flvODNNPd9itk0|AZ`yk;v3BV*(A^(SlEXTCBe^{c!^26aL?E{bt`UrRv>!R6$U
zPTT>r3A~%h=0nk#&^Y_b(t|74^WSjIO9d8k(ekI%=b5hc`3R=HL_Ybe8|wCdr8RjF
z*rft7qmJAeOi2PkIYkw!z&QhGSP;eD>+Bje9>c%J3y4h1s1TqtCPK-0Yj+Eb0BcM%
z_GRNK9i9Shg4y2kVA35kf`Y=rJHBV`Dk)-=tTBY_0pNv|9j+RA8g~P@3s$w)0{~Xt
zb%;8YB4xZwgk|aUVX#V{u&M3>twQYBGKZPDa8+JAb2ze<M9oVzxS7+PP{-f*of;Vv
zVtBMN&8K9uWtIK)y-c6xlrx9F6bQ5yZ0vXa_P2iVfM;3g5520Q@oWLf_UYzOnpo`}
zZIh26C@+@`scbl``bh%3xhD^0wnsV8BKSkSP#H!XM6L7L<IjzcAdE~**!_cd=Hclb
zExt&v@<9k6D**=z&1c%$!QtWIw#M2T(!gbAYy2>z4#3|C@Qw5=Q-iQx&%AVBi8c(m
zYa7|U+3&*rDju%8z-QTB?IjaTedzCG6>%HVkkvAy3ggP-DHA~?NN-tjR832HK7No^
zqVJQLDrdil{Pu4{to)$6W&NA|?Qc5DQP&<AAM#?NKY~Pf|JB;5hF%c12LndAQZ;m5
zM%c@#r+VlutOV_J8+befYtR|g#yqp2%ywd_r5y>w@n7HGms<_8r?~$N6FXVXR@c{0
zUVoyjr>aT@%0kL|&+=7p#J2+E>QZ59=?#@yb9Rh$cn|*<Qeb`@NBy!hn?d<toBgdk
znRA{T7urLXz)d=>j8Sx?lYORHgvaac{o&F*J58652hFrZu0j66{?lkolas^6-W^Vm
z3}XWY!AB;Tsfdq%Ti7A4oavIAr9Se*@zP`g+0@VQnu)j!AHpm=NKlsJGdrQ5;~lLG
z;bn{y_eZkCMkUr^sw7{6kjgxrZ*@PQtAoCFZ+myYYRN>fFFFG$u}EoCO*Pcni>#H;
zEavHPknc)P;#Q&YIq!l?eO5(<kb&#C1y*BG=6mjm6%vYF9L;}`w<3Yucm3TAo?h~L
zPggD2p_1SMw;aq5rutnRV{yE<iHX4VzoC6`wj?u;MWd{v9w^Wp9VZ1@*-GLp>}Z56
zP6fYfFX5-kMAUj~L+htW@IB@u`V9rUhiek)<`gMQD(yJannM4+sOGU6?Zp8`x7nSq
z3*izg*v02{I~LYOVOfv&HW5X$C^ais9J0&K**Udj>iFb49G-9*BRPF5z$YqXtGH!Y
z!y_yJUFfh@T+_OFnAf1KBjmni;=K+xZaZE|kNNigrsb)V&`Y3z8v}<CQh;OLcPo(1
z<m_nManld1N>*_(+05DEy-J^zsqTHWaSV1Q&-pNLFIt>7X5E`D{T{P?daDrHK^v5J
zI=c@ku^^CEi3Soo{c~I88<WWF`-aDDO{Bl23<&2hWyFGz-FwD=IbCI0Wty62+(E7C
z%BBG}mwNWKHX}z}4PSV^M-uT(-LmZ!shZn|=5-s@grm2jqu1U{$4TvJEuMze+AYC#
zK=(81gO8(n!)f!tN5NkX(aX)p-G+ZYPmwiwHb>#t3{DFTqr5_T;g_UX7#M35tC@6q
zl1Pq#n)D>_5#iyCqaq`dC0uw*8g{zh#C-V~e#_x-%FXh0=Sw_jpyyGi8(=7ehx7+D
z)seIJ&2&7}s$q@gPrRNS@-27mQP@F#2EE6GE)O5$=XhEg2vKw61k&fqzg3ZwD`!?4
zeMdGdlBjr|IYqxsGtadnGiHJZd&R)MA<pw%i472Kh-1`=?4pEL$t@H))uFb0B=Guq
zM^MA8%pn&eE<xwTQMK?Pc9oLlk3%F40R21z@7Gi7sFJLw>C)K#5wLCcvtUM%Yw|s>
z@OjMg6Ku)Vx4XCXtBBAB&Sc=Yy?@4P56a=*>W^EW&42~lX3_<12&~(yTWcF7*u??X
zY?&Mpl(;)D0Wbu&B-H2M$d#@C4jEO!L&`>qv^m<w51GHUeB>gYRihkT_uv16@F~Tg
zNYqL>_8Jpsj?hE2VsDr~0yMlFw4nH!=nEEky_?KH9Im1Vzse+47L^}RtwnfWXJ3DF
zm~|`|$FkQcEmTn4{p>Z8ONG0z^&Y{~$H(|65PRbIt1C7Rz4Bci4@{IIqkyqbzu3!r
zOz52&+rRWjndKP`q=rA!yf*&W6=lE!2Th%4$n-qW@d4f0&$!Xrg9$j+ec`x>AUp;J
z28VvB!%tPf07Y?W|CPB)1P$d6O3h4bG`ykk{Wmj4jP?1|s*{snY$Dzc(?>Vjfg9N>
z6QSci@%I>#cs>Mr92-2m_tek(4-o5@E;g3G!56`>w^doXbHcYsbh8J$>MBN&UAbB+
zTBU<H82JKERYVfsp*4kVCN&=MHCv2zp#(YqDh(N7GF-6K=~=hH2FAWADumUOd%;Db
z@_2V?9t=1wQ*JX+lpKoYr%n<EVDR}PrQEmdO9=yR2TClEhp7P#kjt`^5s)xh>n*`%
zPG}PZjLBWy*69x?ohg8a`@<0942uwM{K5~h0SyY(Nf;!Xs^-bY7B4KXt##)6iB%(S
z2Q6Rnl+I4W+{WeHJ-!{Qf2)4Du|~^!nEjc8=Ck>XZ-Wre@xrHb&sg9T`cN_LF?J!@
zAP*L}zYRcym;;P=>#*`B+~@FICJQ_M?@JN3?xdR63o{7iV7euI)T@lAB>qyx`}7(U
zOz0T>w>uGDz;tc4Io$|40|u?<&u7t17a~Sp>ms)eY6U^wgb4U0<}fR(IEn(Batrk2
zW{clHtYZKK5;8F{LH?x)*z|G?5O7x@kDd*pGm-`RnESt3>kF}YGT<L`QHE1{G}I!_
z!nSHeZ)8kn>lIVc`n?n?i~mZ2n*oWv(1??nx?i<<o#&P-z>IHMfqO>(p1Ru8=Hx>z
zQWS61mCYsMcJzNL!v<EF$6CP^YuhspIs7k*vRg^?4qh3c=R!`9J}lr?p|#43PQEg-
zO}6Apt1Is~Mh8r*9$$10CT19$S_=uNe!<4AV0{hB@K@ScgVn{7-vyx(i_3GQ0O#7l
z=1fd*aBx-0%D(3|V4%stwvAdZ2hD9He!Cj#pt>|i%3SLB*ZXbo)o1{Makvdiw%b_a
zD<!Nd%&}JlZ31@8<5I_kw<|{rqgo(kE@Jwa6WKk{H_9(wy`}<!l5Yb@_vyAL7d)OO
z@Y8jB;1m01GzQ~z8^y6?XJ6%#|4Nckvr^35ApmmkxCPfIu6b)hY*g%(;xt^wQ50E^
z*oSas1?ba<Eg;IK-fDCNGweTFkIqJ+BP~Z~B_wvGn##u&K2PQ248bkw8Q11Y`Pv0h
zC~Dj;H?jX98LYCJTgBwfrjU?0joxt-8++EqLwRD;Gguy5>*dT<2(QVxwaM}&1(N;9
zK6Ta)9;X4CiI@%E&$HS*d(>oYna_ii1<SO$>)Uy*M!{LYU1{a1<v@}&ey`-M*v?7H
z1{B#WTJ_DW1Q|tSO9N5GGZy_}qO2g3^21F<NZ`pAkevj)c`Q7|gA5BxOY>N~G{$!?
z+tygQPwdt5)t#N4+1;F_84NI&wBk}o`oC9`__n>hxVT78Vt4D-Ev)>~(tO=4TAToD
zzU-5^^v3$uS)631um=)wO^1P08i5=aUm+XmlF3@;E&<0WrR$1IfwnWRs6Ql|eVnE$
zXwi=uYMGupC&BR|7s2TYY+M!H%Y|ecR%nEqf~(wsh=>O&$k0p=;ebYxU$|Y3WRuOS
z_r2wB2yw+3BQs(0e+uy?ou$G$V$*R~^#f$QnW(&(h)o3jZ$Joe@`JYn>l*H+R}wTE
zLA)Tp1GnogIQ{zr?bCU3^efLNud46Jf4bK?b1S6n=tKz8fljy}y3CMaS+WV<-Z$6T
zT7LpbhaOQk=P7??8TXYA4V9px{0tE!>p*X-3!)NY0=7jY7fNT?o#;C*FELk42qR0!
z(;-Jh%}ZyTP2V4j)k^v9x3S`jTw_0<yY`Yx)0CuEr-3)Gn73P^r9FkW^tN9I<+cC2
z2*%rpg;+m-LbRI^nv?yxm_dg5&dtHZuP@ufX{JghT!A>>efBk>pxeeIcs^<?Q(x}!
z-nun7G7`72e)IwY@rOK?fAD-qUy04<E-O$RsU!)uxTG&Gz66$pJh=_TV3tCDNlD+>
z35ff0KUvA!?lHc&Rx{)46*8&mR`qH{;|x1CeGQ{}Z#2KRI=Dq0m!*D~ty}2_qk_C^
zbIC~MC==l}q<JV<1t{Ubv-FL#+hv=f0NNY%&Smyl#2W7?W3ND>@c~V98)6pw%PK>R
zruhhZ=#aQJPBzO20T=dNF$vUIbwH-yTxFIwhxo6w)O?0;wGwGcoE%>D!kg((tQ$=g
z^fjFeoW*cK*m!do?;q)haTDKJtciNP?P9^@|5y-(eC{%9L4D)ou+Q^NQ*;CFyw@@z
zDM4r`gmTd2?8U01;6<-6mExn1j~fH*)nsb-Bf69zqisiS{U0IITFS?j=LRnLR|sn)
z_FhABA<LT}OPNH4gBbCmn%h+$oK;aE%fF&kG)tE*))fA0zfNmhk;FD?sXKpT=sI!$
zglWibd1#XHvx(v+lBqN{Zi14Oy2F_I4WG`yU*>$}>GF6b2r}YaiewZi@IgB9{t<UO
z_e(ARDAXA_$EH!#*ME15+I5!{0~JZwH*IK00}H&;3z|UefUxEFbmRT`rQdd7m~sSN
z*F31@k3FNHK^9_l{X~*YyHauSI`{j&{0N|_sr!%WHDW|b^IRh%ZcttzgQAb^p1a1N
zb)^l6ai0@xH7zDEiy1h59VzJu4snS5C*o#nh#H;)P9&{K-6y}T_3yKt%JS2MTLO?$
zsC#M)ZBS?zJ-y7Ek8IvdDeG|M=H^*UST3SI)SJ|VC34n6%3t;D0jqlF6{MInW3wc^
zvGHQFRs4sZZpe2sR8UoT*GGspt~*OJW6pCQq(EF~K1b*yzc-h?&g_xox9VC+<`ez7
zphwycu<XzV^7)zjr4aWR!=rcapE-QliVC^QO&Vx59N#Kv2CqVP4dRY~oV<E)cTq}z
zLpnx$9lcJy)d6mBt`e{6r8?uZU#PV#^QhfWitUVGQsjH|Yr&xkNF1;}CeSe8QnvK_
zT{us=&qs};;!?B#cnzj$6Tpo>W+}C1v}@eg#$sQN>rH-_7vy!N;?9dEeJLY_!}Bh{
z>0$;%`L$;|J>@`4NPgemZM%yjW)@)K9RPiOXY?npX!rh%QYEV_N;d-Tw0+6)Dwh{V
zk^_k$jjzbcZU`)ti3>OjEU50q^{OsDE(%|N=ln4=g<_ttB(?pC@@I-lAIhny&@&}u
z>5T_5JUJ3CedXTC|1yWnV{P&7f*5DRK-2X)2*3Zf+_H-HNbdQL=~?iZIpp$6hN`>~
zK^ac-u=3~VW0|2iD)~=D&5s3ruevctCUhEx5zMb)+#x!6H6meMy!0(SYmjO4w@>1X
z*oCN@*&%s2x<qDuE|xlw*`!&Fiy&{qD2_X2v$J5qmm~XLF>oax1oT!CD&ZAA(8smB
zV>@3ePir(*Q*Dj$#7Tmym!P#*JJ2YDv@_bzc+D#<_aI|D?0$q>ZuIGHXP3T?%G|YD
z5<Bs&vV($@&uC;{AYw5!il5zGNNkH^+QFj(^k=M>I6hJ>zJA@07xW4*C}KF!@Vy#S
zV0;0AhX3t2jg0_TXUUfhW$UQEhr@TVqa^l8RQm9U_pa^#1m4x=iE22Nv@6IWs0Fpf
z5zM0<J}L+VB4iS6otK(0DEdteD+|{u)Sd$dw|3*p6EwM$n+|$xEEL-z)Lvekf&;F0
zTF!AI<D0uL`sLoO6^m051`hS_7bzGu5au^JLZA$U`<@90ZzW?&^fPaqq+poso~u4U
zH(Loz=y~Hp0x{8tUu=zK1B?jY4g&#ZDP&pY((+&vA|Y8z6DtX^gdUMc&IoO)m?G{T
z6J0)&ql?vnNRpZaX%yT#lYPsM*EXRb=({~jN-s0Esum7Sx#o@QA8ew~u?tC4SLQs#
zbjVK`I6QCmfb_i{yis1b%RZSps|~rYlG(hb%p--NOKL-1F>yo+WW8VGtsJ>vi*)@J
z#}Qz7t&hNgGt#(W1^XZ4nx1`MVT6PznDh9D#B12UJZ*I{e0aI6Rc~}V;KHBrr5z2k
zo&}toSMEn%X0fsYL0vbeY_;|{Fs-g*;)v7wU-=Giuiq$_18D}er~#Z<c+yNO$b?2C
z#*Oh$<&}@Zi%)|~<>OjgM!CK(53(I6-^`)g1u$&KT4__-5M$~C(8z9s2}7tVYFG!g
zLqp0UaW~V9|I2S7EdG``Knr<g7}EKEV#)C~Xirs->W6X)3W%{XIjrP7E^)~-B(rZp
z!0(s@X*ZQT#8?xpM>HsDrP;^gIXoaDX&sxBzapXW?*8o`9|x}sDHj&y51s5?eS_in
z2;;(($PNwGfQ#(w$tW4P*lTddptJT6HAVCl1c$X?fdCtBnU4GhO>7uX$z$<{@-M@?
z8X*VGM+PXODS;{%4PuUF6jO5|`wlAfOhf+V?&ZB}3OCtt6@E!jdQfyoju~foy+NuG
zKx=6gpY)}&@4zTOq^(D=vxh5f%}BaI#;%x%`_B)$Jj6U|gm_alD^ecI%KKhtw-`s|
z8^r566h4>Xq|f(ff07jD5ITXfKlN*)$)5{Jl^Z-L&Yc{kJ!_?&P&8sIdSG4jc8qq6
zSNDZB%%spn4>wmzZi&RuZ8@C*h;TAb_d*G#&~HHYZqD9A-W&(uJIt+8w8X}Gf)&07
z)X^_iBV4F{{P<A|hR1ng#+$0iZx(tI&4AY!lpERBC0+r6r#&iDR#<PS%12SY?4{7$
zq9W3!qjrHZkZcmkDkT9&5B@khzld9Me#%|?Fg?p@<P?U8;eoIc-Hj#ZJO6HOax082
zK~x#-Wu2oCNKB`;&UfsQDXbWbWO*ebFFz6=+NI9EL+@V<DIpvvM`%}+GI!Wd_+4ig
zi0E1&Etg=U<hd4WPq4&j@4ua#JNlw<?IYf7CuFph$jPs5D~MY~en<Y70i_3dvn#|6
zBF2kB$D5tB<p1X_eAIIt9sa1))|i@7{nKod2W!Z-ecBpi@7sZhOIaA?f^N>ZmLYka
zq+7(1u%@%bGlPho1462iOWUJ(9mLS_nZhHRM=O5BRY_=%wX$Aq^X5+kt9^nr#=i!x
zjM^0Tj>!y(6ip#!{O{*QE1IqTDP>DyuSn96B|ZOg`h<HS%`I&c;E3ZXj$M2CJ?XV&
zng_vrM^w*2&MS$$6enUV>4iRMHS3IS$2P*)0}13JshmjS<Ku6g3S$a<=URgp3}uY+
z(mZ*m-RH@$xt^d}@z00dOlt-d@jGdh?C$J4DAnMo0}*Q?IRWdV5YDg<HW9bMf*S!>
zOGB1NFg3H3t8UBpJMDr6cD;$*{A@rKB~Gk8E=4jcMUVI-jOoa67NcwyhD_7NRx->k
zN>i@zvDdT`X>Ej1Dr)>BfF9k5+*%N(JPrzdRW05u??@TTOIZi)K*!%S{ocNK7&yiS
zttDyZr}O|9qgtMV$QkyFUgs-E_GwP^O%A)CLnUvsb5!~~sAJ!l)z&E>4gj;Epucw&
zAy#ddUCkZ5&;-;v2I3f0DWGl!!^b%Pi<UGt*DBRCDO8E*pc)HCGV#3=32~S>c|$LA
zI6Alt?^vH#lkIQ#wN!pjL<AsKM{+}*ZUkT`q}DFdeLdF}VtJ||Q7qgihKuDjdhg9d
zVhzH1sjiGz^#T<zXWzXz-}IZu%D{PfHBLFVuoL67oVq%tn#}p$2%Yl#?r!BjSeVb@
zRHrfbqrj;zZf1b!<|*Db=2*%q`*ic*9=erd)ZQ`EwD3boeXo2x;(If)?P8oi)1w9d
z=Z|0;sh*I9C?<`U5`H5I-_BkOQ&K8ii@n3X^Is(I7lm<D8nXyxGcm+IqDyZaMPETe
zfOLZ#;WR&m?)8DU^F|wVlE?&BMtsgr$<uY9KxvU23hsO-G>s&!Wsse*HP*bTMKntc
z*$nS$QZT2yG*p&niMm`2`-VH~1YwKlN>})=`k=~C7N_`K4l(<DhTss0({;#?cCl9f
zEv7RHv^$`KQWLgVUyAetmj)e(fcxq|^u6&jt1skLHDT3w?cK-HZZ#wGM+NF)#Wv5C
zfl^ukHI`$EiDs2z#Kv-B(?i#vtn6X)jF{|dq1Y%X(20s;MWY=ImMO0b3JR)HIo=D~
zE%m3zsAp@Op1!*>b23`91^%U<fT`{U+rSfKW2FNh=DWRZT#}ycu1J!(-RcZ2w~~v#
z=e>x;WV1E*`dML1Z>qiKH3kmOpzNw?8qS~l1qV}EO#n%Qf{s;Regj&@0J(hn+&dR?
z6cXC-AN%KO+pMaqvB0HrMbX5IMeK{{dpt#&6CrxUz8wuEf!@eMXxDSfV<vwJs5h>x
zFS7JAi5_HDkN(C+JNvr<q6h3*^4l|#S9T#sqW(|ie{oShMopBzk62?GIg6)ok1Hvl
z7Sw+bkITPn`M${A1$aLS%BpKjk%UOW^;mF&ZI{EnDyv_6lU>fqfyTQcgoSTgW9mTu
zLqKRoc`+qbw+6-cgxh-WqWgnV?s^|NNCNG|@=Oy*+zk3D2%sB~hbU~e#^!gnJ&sb)
z!Oqp#&>Y}fE)$Z(Pxn3HNdpW++^AU7jF^G6H#xZ|a|LG*amPJrvd|;t^ak>DGZOO~
zWiDcwTyp=-+}BuB*XG6d=ynR-CuSJCMyY=PjyFkjCj>=;Bz)5GSa9OzrICDhrO%xy
z^{w5bj}U46EoZ^9b*o0x=c4v>v7GEeA{aXPI9^HywxSMCre=@Xrfi`VZpag;?X_4E
z-Pcd#S33N5KNS&mTy|bOlCQUzWzlgceAP(w><6U4zw0V(312wNQnz9iohxwq=h(_T
zqu6A!v-th|M?}}|2T^&XLa=qQ1Mv2U_hwd|*B7t7DMU=+Vg@=dX32o!GG&P52t*Sw
zfq2pv%J5$W7w4z*SQ9-^WPpy3IxkZ$g#9X!x%oN}VVVJ7_c*FO3EMvaaObiac<G2s
z(H~C&rUv9Fq~dlMBMOvol-<ZItM8t@K-h48;=x0*0O{R=dBM5Oi5KrJPmdEVPs+Y_
z4h$!(B;}Hui+vRppp|tCI!TxTTJgO1;FXY$WRplwS&vvzlQ+NSFCzGItm3EemmP=!
zL3Y#;3A@mJ>~e8aGF)f(gO*(>>R7d!hn&%Lpx;t~MV22C0p;IX1h4r$`_7wCP>9-J
zeLC`WDt`XKO5Yft`aZbC%toD57uQJa%?A*VliD`Ce52fFs;j^#L{=OGM;C#+@75^~
zdHjK@Z90wtFjD=%qL=dey`$;n6`n?a3hATkAURW0pF?K~L=}Nt0d1r};I&WIA6D)P
z`<%Q2)ZraW4P9aw)y<#S6e%n^9JfT^rnH7G?S$Dk#t$*>dXXu{R7_hDRgJTfxIKDg
zWb+zf35*Mta@L0Nk3Ae+roWnUmigX~Bt?LUo{8k5<Jb@CPAI!^$z`-8MARI?zEc*u
zb?N@{Z76kUmw>4^YyY={{kOe5kWr!LoIdt3oAZ|t7}T5W|0C(TqpAM?zY$8ZvPx3O
zs1W*4X6jm5g_~@$3fIcDcSL3=5!v%vWn6niRzeXM*CsNqnZ16``~CgloKEM~Io|j6
ze!ia1$MZ1`-L%EkN!Ln4rTMt>kJ92pYXpyz_%eG!L{7G2?M#%F-jqzUh~7Ixc_WVz
zWHyh9r$|A=3bv<Lod7E5CO1GN_v>5Q(wTPKQj<Jpn0L%yDcY(?D@{AKqan+k%}SiS
zTjAvaK(~64xIfmAB9z;iP+rSROS@vu+_yR3yXtcFFKF%q{<9sN>{Asc)}-s|>goc3
z_v$`<{j0V9=cKLrQUy8*1k$$}s^35K9iI%$&)m1E|1R@+Zdzj#wbyc*E5n}m&Z`Xy
zk`wZtg|4hFGaVZB-tAx)1ujvFCPflCN`tXKHIMuc+!)LbgubhAIbV+0PLybRPNh;k
zb(>`MO^3tXF#k}-uYQ{@p~fVa>5iLiK5%L?idD&2{W-GydF!G4Ir2w~z!18a(Z)JN
z<_N5SyAl$f5Zrwi>;J&CAeqcy2b3o`J0mggqc4HWKRLaUmK-%sobs+K6E<z(+Inth
zr}iS!&}Ax^bsA)+Fc-Xvy#)&`V#7^X+Ft0V0_!kCtb?!sEd86fqoo5Uj<+k>r0-M=
z{vGSw_|xSC)6}~`HXWVyli}LcUF`Rz=$TaRF1*w7vOLSxdr!{=+-e@r4C1}#|5-nc
zNuN6SSpT6w`8`#GhXR=t`XgG3!<nQHsm)F`Q?oN#omM$*K__Lr{?^6UuR}wc$Drn7
zG*7C8@GHTwgV7QTJJp@hdFh{qn#Msrb!5AU46IgEd3_wV<HdR1PHa8(`a#>D$cv(-
zj?GoiuKX78IXKb(HOOY}d4b_BO#<E18x#MAF1o8Dgvzte#Ac%BzMYUdkVw_`m``sA
zHWz(!ppSWoGm9XlC*H-k*G_+aza-6@AwBtQQjFnlCz{P%ZLVxb`o#`vvfFU=?N9Tm
zCrB!Cynn<KlwZZYpk@nvk<s8Q!qY>$36|=XFC&ro8kKL)_+73iV{6kl)o>lCm<3Ur
zDiYP8z#>I~A$yNPl=?O9?(Zm*o8zKhcfJ`jzp0lssC5;Lmf3!K-`pZYC63n#F#Xyn
zGigSKhpKUYU%S98?FXQ2MsjsQM7k^yFirNDODxW`2+mghumKKme7D<%GhHc6E~{S*
zbN4H>7Aq2TX2M)~H`9x6(|vOLdt>+K($moyn-Tc6Pp+cFZ04{pIt*Py&ez)rX3#t$
z8Aa^cxFof{Cjq4LGWl_6cXG#*y(WbwRxanLddmNbLqjzYLR@uG1cu_xgajJW%k(jP
zh;L!bZ3&^vGSOmE^g`nrFFAOSp7!rA^^sn0_(k2)EP?pAKJp?Sm4#pM;J|MShSeqw
z;4Md22EU*<?!S@Mmgtkj%COp@U#1W7(dU~Me4@raT++#^^HzMX!}BlF3y{mF<<B3-
z9!qpwk{ER)2lk39M+=W>NquZ%^3RppRShSn?Nqmm;&yC)CkGQiJoQSrWfr(}=fWxx
z0egOuNf|lw?U&c*;s#8Y&y;UNV06?N(}=2q!R!x#A>Lc<*ctx7UWvfNLo*NZ{rKJQ
zVN$)H3B{;Vk+y_OJX{o9&QgxtB0}BfTGJM$8n$@x^SOgJb|*-4YI>f<W-wN9xT_$r
zZX9j@gt$u2tzCB{e>5!_{RYo8Y0hvLkEVxn^La=5Li#CpHH1)CYxri(QL4>`>Ir?Y
zZ`uR-t_)^7{LX`wQ5KI)tN{N|smFo6{(0MWM!T$vmzA?CcM2+xX6(EWgj`a$F-gy_
zSaq~2w*m8xnO5tq-ih*a?BZ+Syvut$)_Y6tWUm)k^FD^{x6X{9D#tpyR98whsW@ET
zE`_|m?cw2(>^ky48}A<pzibD!cPv&$t8YI2d?j5ZhvqWX*Z9z#GgErg3=TUkx1*Pa
zw%x6ruiXMIt^0jBlN1M2V+FnZptF@RgVpMrnYY{6W&%V1Q5;hx&q#)onrb%dW*CG|
z01On<WE*6lnKW9;Ikk1ePk_r8XQsAtKIUb>PQa|mlzm@5-CFeEaHTl@TY_=V_@+g&
zt-q84Pp;&w%fWr*gRf+71DgD4i~NSQk_n@X)i0&Y(^OSww+|{E!qp_qUHD^MZQhJt
z``|j9xg_%6cY+aiLR7nIiBPk5(`6oQC&fVqnvBO|k6QEbE1DCB8BZIPjD94Fg}}~p
zVDmWfIk-|9q6d&?BGWOUr@@EwrU5p>^p^51HVZ(?3-v2(He1O-RFn>2LTYcN2a7d`
z*f9?+Ew7PN(OOJ41^ipWMZ*BFzq6_dK4=g1N#--a!CDNaW?5*~vLYm0m!ED=`1#@J
zZY%)l=qqtZfS8P`<<hR*wH7)frSEsZ^!Z2f-MCL~+rGt4OR*36`FYc>9`>#4U!(9o
z4M|Lc`A=K0c&~oLHZ`j+u3ZV#p>*k~B_<w?b)?M4lbCbegv(~-(?wyS&$C!hx3QU*
z#)i`HjPl;B#Z`95aG&lKm41kvaOB|auVB%$Z?)2xEhU@0KKY#CE)v=3$k8#IGKhY^
z+GItT|0vKz#&B01X<bWe9-$)V)f@Mvhh#sz;|*qHQc_LS#iR;JxZxp+NSD_C>e_|k
zP)$XZ+0{l2zv$|rz;_?GR#ljzpVl5GX|xU%ms!CJ%*~XG@8#L9U?D3{3Pj;?1s`oL
z;NAEe+5MCaC$f4#9~cm{-HRY@mOFcUSDkA2_dORE7f(~Z@`jxPkUZbDwJE}!_7w;r
z4}l$%{wNd>hce=N!B-9@U%R2tS69@sv$KK36biGP*t2h^+afNsL`ki5fH~(7(5D5i
zZy0;`T7qw(C6rNR{V7)n1xU~0ajQL}o+uj3;W`um9ty#JM@uHao~n&{_n0`D5V`Z7
z#~w*Kcau53hpg$1=~J>G8g`rc?svRAy0vLNl#^kBH<PA#`*1;9^tRt&N9i0LYl%ap
zUKSl{$HY8B<Zjy5skXCC8pUM^q1OZE6=t6zeZHa-)UI)TU}$(Y@sHMvY&E=t4!I>n
zd>iOv>H`h}9>a&<v7MIY=E9MED6nZlUoltn9P&zu*iko)5A{|>O4QKC<z3cq>7Ylv
z&U-}pi#pF>Wmroh@(B&9aS$~xhGno?#L5LKUd(?`6yKrnlu)kT3t!C6_M>E$d*$X=
zB}UxSe%oVaB=DT~IsMEcCQ^#Q;`BR+8VDXIYX4x&O$B;sSqlm}JnI1r^3*GN*^|SU
zvWIIAE?&I23!rs^oS!VHtjf#FKR;^?d-n%ZQeqAG4_GEyY@*<M04ibpC)bfT&@+|-
zS*H{VZ;datSP!_@#)?02c%CW)uLJava$f*B-Gs)ws^#Qm{TT2GJYYSMDD6?2d$ifw
zzHNE5=T3OGLEUxQ_4-fhw>bR=5hY<mQr5$+7f`C&vkyCa*`IuF`RkjsACC+XAAC`3
zWandG9?@x5RMfnDN^$N6#W5dOd){JMaf`)8hCP`zG2C6Xk@pPtMQ*1Zi!t`j932)C
z411(gTeMy&iKeAvz5ewIvph(nKavq5ycNqbnuB*p5(K$Iti9I7ZKhmHSRI6sW~Xkc
z{Ss}`HkcJs&>~d%h4^qsup1c6c8J+Y+4nom{w=ff5*ose2^F0zej%Z`P4(&>bG*~P
z$#2y9V%9xbU?Pz&BB}x`OT1m#IcO;7fx0&b6ejbQ*XLl>@Dbco9zh0AQ;vG#`f+|k
zLkbWT#QcuE;XacEx*L%iOkBr>aJaJ7E|-Qcg<hBf7uqRM)}Mnu^h{)biHY3yxR(Qt
zjt`Zi@##D4`dguC=K};>66Yk^Mr5p1JDzo%{>+iMx8?ou{?)@G+YxNZWo!|eT`Ten
z=6RW&l$JmB#>VSE=a%I1Olujs*e4FNVgv;#xhqt$!`L>qIC4bRDR)Y#fEYn)|8|Ah
zB#dM`9BVTL1M+0rWqz*q{KXH&EeXrqT){#3!qh*K;VK130dD$!ZbQze+GhEorl7Y;
zQl1-K=e5Q2Dj8K?%hHef{651V{ojJ50lMoZ9*gaeVMZZqS**FcOUI<Sb*1bF*4_UZ
zmZRKAt~vk0ZkqqC@(Dq3>^xTBN^X53$x(t%!^A7LZ7uW;X1`Bq)FXNhR}W+1b~{{g
zYXp_*E~S{&k4aeJhyidZ3=EY^z~{FD=Kclk^2n?&Ut)m%7?656%XbO=`>`Pye$)Zq
zr`Y`NPNai3cRE@Hf63*id<3W5_*RQb0nnM(LgWqus4lx-5$H7&atAYD#(f#bzO_$7
z)<5qIqDI)9zFXt<D`k4qBW2CX@%dRVlC7^Nov6$uh<w{oky}4d-2MKL@=qk&({J%|
zAu(atBss)4tI3PzYVnLQ(vzgk1V#02eoCvbP@f4BhCPZW<^P$T52#e&*)fn!%(B5r
zf~i4$|1Hzv(BwtQP!Zle;%Oa8bzH|bq3KJ!f35a4T;aa=n!G`fqDQ05E|bZw;;6CW
zWx<Swg2~AHzzUxLiThu%Ta;s^N+FM#366ENmHpM9M;e+&(3c(OV+rwrPO}wWQavMt
zJ1^a|Uy@U8aYhOU^e?=e?}1DOW0f_76g}7{>Q&l*T4GcA0(-LKjSr2E-x!qEKCOC(
zfX1vF&0~!S0?Q(yDJNRlT~n}6Xr8gignj>;55}Hv%6dw7d{sXPSq>KJor!cZJ$X$3
zJZzu+slRj#fB(JORI{Wr6b7-z#gW{;?veQWa2!?5zLJ<&dYI*!B};R|4X3rT)5L_k
z8o~D`2<JC4h^GI5n7uS9QT#F?LX?ExnWNc)j^P4b9jUtmgQc@P>6!yW^@xy$IPW}E
z(`EGOsfljAJ+Hf-Q%@GcI~;?3Ij9R<lFjhwZJS!I7jzT`f1iFP46#&iMvM!lx;#C=
zj>QuoHm<t9W!oT1{evD>hS^T4*ct!li_7&Ptl&uq<xuwOWKe^irtBUMs7B8CcBx~4
zIYcBzq3l@=4GfkfW6vvHayq4;Fg~#v#V!#AT>x&Y#Sv3l8cVcPB_^r-Z^M2mckYKu
zVbVl8dAon(TYp8N?$>_0ZbCk;Ce}Y90#O^P*p!*GM*?d*J6nPRHn>=>zJ2VIrvKU!
zZt`*U2c%Z@m@)Q6JFdI>FI^D|?O<p^5g*BBEJk)!GN_ZbsI1I{s;#Q$nP?kx5Uv~h
zUsPl#>2?YqpBk0#l4KC;<Fl2RcLlee{_3NpFbLKEj53iXfKan`BhOW6K+yoSxdyXQ
zMR8Ky+k{dd9GPv_@HawA2(4fucjR#Tu3Ec@d_G;pTa%}4#XcdP$BOU>u1U}Q$kIh3
z|L0ZbwOTJxX5H(`!AAh+>5i)*m!>^Vj<)f=`V^HPMJ=g`PK@wYzT{|4gnJ_j8d>*o
zKkwmt*TeQ7aoR=IM3OXgg9*@&Qnc1*^>Nbk(ZbiJceZxke~5qRYkVO0J|?UZRT)>L
z+i<~L&EXoC6xoUMrxu>pr_HFiv*#}hgok#xHa)nPvUHPo#&Ay3zmt}0Lw!;&t%Ud&
z$@1<pdyF{w2$At&uasox{7`QkI@Dd^=IF}tC@S!|pTZ41cEO$lPrEecDlnvI8(#P4
zbsj3FR<9lFV=!ARW(9wWe@=l{UK+x3R?Ntv<`eO~EcSkWcx$r4{F3SAnA%d+>YWRf
z;FoW<xACWQuQ;h3xWCavvNniXIbrU0;ovI*@PEo<B40cA*;cY(TNFo^d9+O1If>Cp
z$*}XpI3u_>cIr-=nzc0WHr9*mnZ5pWC1`e^3jAd}SFXAZxOd6Ravg0_>geq5zn>?)
zTEl{>HMuziv#R3x7_Lc%wBKl!xZ4Lm!>v-nlm*uKr(A!vmG4Agw#T=k)PCsWM17Li
zK2EhEoSD|@LZ6yVB;2%E^>FS(am6CLE+SH#Zs-SSQf9u}GW?7Z)K+{EKDy29lAMSc
z`2xOorucqItVqQj`_!BY%xFW^#R}5`2ImSHnWP&%7t>A85>zWmnY?8+E~gCrAhvXp
zZzf#h?AgWbCavWTF3hL;$wJ=T8@hyR8ZuitFq!05?cmKqIEWN!qxN29RwEaNUfF}2
zz~!r7;(xINsuJGv6WJrhz}Fe0?8VnylhOF~aeG(QIE&sd>dy|>&iZ?&iI$YQ+?&!c
z4@>XrXA=yB0dCpp+|L8!pxFj7h1r|9t9`TLrb@W0YKr^SOC-FT6)dRLHml{N6avK$
zmKbW{iEx4;zq*|cb`!4*soo&JH5Y}^Xj|lEWPiabtav)SyHGUN{n@gsQLCB=m3PHZ
zV+pjCmIw+R$z}G1b#lAs_7>g~>B~md@8=!*z_6(syqcyx{NnBsNui)ExqpS;k>eT)
z>^w%QcJ8j$ZFY2(?a2SS<oP#!6yBL-pw$#-W(`1-bbiFoMTKHF{Q1J8I)Z-9JNs3_
zLey{N<4gG2cGh3g0o2qRAuMX6*v68=JJEuw_4q<!Q{;O<%34cnrZxA3VR!%?s$*l6
zyX*q}kmARD%xX2QnJCZ+6sw*usk+?kw_PT!yUBHANF~zt?Z^1$Y_VArIae&RLWq#C
zxU1iK$zGa)fTUV_;gJuAReM``lpx1h>`03w_F}Ij)+l;-uS%rE9KUYD(2ke3sb#;<
zu^n3H@Zcr67t&rKp!eW}j<^nhW1KjDq;8c6X#nuPX9AB1{jQw&Emtp>4t=1Oglf1q
zzSA%->Drcb$BB#hoonZ)Il(iMJ>E43HYViZYGT;hqI^Ogjg8*hK|lDaWCeI1JO>_K
zHHxgQ%>V4OcaK^Qxv9?m;FK!#UrYVI9a0fkB+_P4lJsR_s=AQgT^aG6@v;E<n-2QP
z5@MCiY7#aKxs|An&>&rD4#V>LN|BzFVLRI_&Oy(Dr{z*L-a&hOO~<tIx182Z>q>*+
zvi6f@QYqTb%O-quZ+jQ62>yq*?HQ={$CAApC<7|i(-pTcLkAOcrnfpkC8WvEv?2Yv
zQ~Ct@Y_L;*>8XyNwJolGz3Q{-CDW531LI;fR6~KAH(faH-e6*3*?=+Ld4Hkae46N-
z=1{l(dj4qhTl?cz*sDM51Qx}o(Dq_x@u=6Vb}s~nT<gCuHadis)VcUO+?(*us8o&L
zK4CWaIRf*bd@!8sKUmV@mmFLQ_M~=gac^dQkw-)5({a^^*{72lh0ks4@Z<^CxULny
z0(4AgO447Y`Ub+n4R_VEufmbN?-}CtD;jNndlQ~WY!qsK@Y|=e*iJ*p#59<-C^SFo
zGF<xdD|S<|63#x#gyO#Z8mnlvbI`oQ&*tE>9uV&ocZiZm1k~m-5jjP3aH9Y+Y$SRC
zeSs-K3uH*4k92ep5F|;(>3s|4G&Z*h9|@m>5rxoo2$P%xA|_D=0dI*Q(MusnKXP@*
znm*|MF4ylfQfg;VjXkLNN(Bv-ui}0FQ#l=fV&~bpt_2@?`>383pSbAJhMGUQ2t!$0
z{U&0TA9poY1wl(5_3~4TMb?$I8}RPWF~<1#Q~7)2^wb8EXr#FSqk#85=ffuX2}{f~
zoV+LXZ~)~4h1vOY4_t29EnFGRP8*&o(5OTtnBjf#%3^l-i9P_bQuNtubN!7*ZoA!e
zoo)YV_U-h`Ao1R;?3XdvO`6%yn0Q_+K`fM<Dq?-D=!_RU7QGsmeBzr>3p!H(sK~67
zonAv_bDDm`=loNi3he8>VL*GBD(k~0=PNza9^JlLi7Pf{W@aY#7V97u5OWxp#yM#%
zLOhfeDC=lAWk>susc6}{{Ws+8>|SI<H<ENj#CKS<Uc`!AMtUD*=cV=O8vceyl9d6k
z1Dm;MuN=aH3OV*yQaJ;WMY3DAbcxq|EFTk2d>K>_ZRDC{7kr-KEZ%4OH^TlA6LTO5
zk5VO^;BV{CSxd{P9`vL$olM}zYROLtEw~8F%%zIW@#5j)L<kSeM%*~?yCYnc25J)D
zZw?R6p!fTS*MI$;>+g=yzuv!{EMP}{TdcGy**kT|{yMbg$@4zyu0&P#9uXXAS568b
zr*ei?!4YyFU&3Q41@9DtU*g2TQDWyU&3d~`0S40hMyC}&=s5nHbo%|@pljpzBA&*#
z7mZ%CqOJ{Q5atM7-&RY~ZcS(&C^B=*40~3>@2xg>QE?ZcD6}gybzZE^WsuB}04<Ul
zmnfV2#XOf!t+UCirea5tBE07f`MJ&;hJ@y*Du|S5s$BK;tG@0rV&-Zy8;Al$fdTfR
zqQR^^?hC3Q$`dr7)u#9gLFe`ibi$RGsJ`^~<s_%{6CTNa+ms+q)OEm@S2?gs`cdvQ
zqu2nI0593bYX;!NA4t5&$;<0NzkBBiJXm6(cQQFm!Sz9vLT6;Kns8VvJIo^mo{<$H
zWK1#hH+Gm=u@w(jF7<|0{O^}J@D(pS`u%Be+gcZzK+!=HBsy`4o_8};`RW>{X%@6N
z)<+K!R#miS5hC@5m)CR|Iy?Sbt_`BDI7LM%@m~d&^o_LCfBTYH7r_Yo6f?Y&F$R0X
zwXG%!g;ypN6cEm0jkedTqEPK9F4iI`o)tC->AgPt$El;KNXGA_vELCInDHxcsmZ_#
z%52|w<cbqfqaA|l3!J|U4BM6SA^$QI;GO?fZb}b|Qx#*aj11r3ICYCgYCmDJtEWrQ
z<p0g*@U5vq%kV8#(BC1hyVLrWT>0K@k#J7W^KoAA=HgbICmS!x60a^~*!={*p7{~S
z5FV=7CLky(#Sh?H6|UrSj9Q|PadAH{zVOORHBeoU`!t0r7poNh{jCb~^!Gh@EQ@+A
ziw>SJU}-e6GQwrMbQ3RFdE+OfA>rbG{K91r8U&rQo;iLf$|wxR7hVn948QPr6bIj5
zPx9TAo&SdCiR&SN1f4iqX8j=(BYdGhw>Bz96vSyW5bLyAcOp+-X{D1IU`KO1+vnJN
z0~{&s!!ISKn&yqlc{imspP6|WIx1BC()-2;u_`56OH9(>uGY~r$gTMiZz6OVrq~5$
zn9B^+BzMNpU_Ze`8vUl@-v8I6gNo39P8=C!v}aYtOA0RE^t}+JtT#WL&oTHVdnS2E
zhypKU;BxZ^W45X42Lk?mP{k&}=td3)4kAoju;|r104t&{@A2Q~e|Yi(o5UU{NOP0l
z3gP4-JrUi_B5eEv^mWn%TSjd|!ykE%GZSvVV8o+XjcN0%s}m#HB{l|hO};oz-6w9P
ze?~MagKVS=AdD_3MXUP)x9(0th{7NEj3uC_#hE-tSLi2Ut47TJN!b?e?y|OfL17|p
zmloTB%5G^~h#S6nw*~j*nu=INi>9vZPu;MyRIJOQSce(*)Y}Zy6~g2CT^8_&c)bqw
z5tnra)QewzTW~jpl6?t(>UX8D%OER1CACqAOB!ue51a!JgoUUnpM{zvib9&7D<t3M
zCoE|q7P2YHKD;$kEp*`@y*^^{O+qLArq)-i?W<{569xXi02u30;PVa;{6ntC(kX$c
ztOIes9HU-wx1RF+>Xgn0*zD0&g3nKIZu$XD>*5QOnQmy>?`#j5CxWgC!kMzl%iq=2
z)zRXA|NgxOzx+Dz);7Qlf3u04?h&|4l+&DmnrIKuh_2oB86!Ck-mY_Ow}|Ok!631y
zj*4X@f)c>o!OS1+<X!4HTTpYtfq$Z>R_KCPrN(k7{ZR7uX<g-<<&;h1DsfmCF+~<J
z`m~{@HOq)4T+8yt<o!u~ocy#)`Vz^D&kVaH33T4H-oyJ}F6L<sw53qe$7rQ--bz^8
z@w8ZK7jKGcVryRV`xCo_L=)elvd|8a$*aw*=@;WPs7Q6m0TliQUBDOMh7=|Kj>&#7
z#gg4f{UAV;pks+LVI+vMR~Q*g(@IpFR4P{2Z$3qFx9ZPs!OUEx{*TQ0LyW$rOywub
zD3=_)$EPWVN>1{s8n2?j^@syBxiYH`&vdo4{-<qp0vmf#@7*=~N<m>^YuGAUz|SoB
zgU@G#3r!R6d}2gG)qA%<ar@B@5Wjy#<Ay1kc%8I!HU7Y63x>oMXbKltgbYJ5(`dcl
zT&bM*)7h@SHnt6`FEzz)zoAPf^BccnVwvhJAjpWZ!4FnPiq8y&{eS)6uMRy#5;0ZA
zvkb#53Cog7)W6A3$Wy3{bq0b;{yiDMUZ^%9b#QAiK~Vdz2jQ(MeblaaSuZ{zg@0ov
z8Xh#EGfqmQ>2lLt8vd)gG)!+T?K7yZI%n2yOiJVAM_iuHh%=qmxlZ$>oaQvET}3W*
zjKyMjv-%O&fo_*Vb9_a@$dM#AhPvmX8P7nm?S`x1w*4MeYU<Ts%ajIzf+t1OkrLx;
z<t1-=iBKL)4b5=4#>{_-$HFcrUha53nm8XsuuBLDBoZCsh$VPGU*zs|wBI2&gjG3m
za4-C?V}|nFdq5{}o!rFu;q~B7lU8wV+!w+-4M`sX=GYGsj~<2}D4Y30M|*9gXS#yX
z0a~R5oyD`>6X?)8-tB0A9<9m7rZY|Vke`Y%e(H+9#B*L1h!-qUVLU@HDJ2*$QsyA*
zP$T)O(9xM6<1`yo$Rby$U&)+g+R#PRe<ax%BHp<(zbh3i@|NgP;;i<S^-CW-JeVzy
z0<QA!LD!fhr+<?D=^@4-tZ*dT6pvog2p-O(z!#qjN8+PLoKm+9FgEX=EyeqAudWuE
z;&TL;GHx{{5W4D^Rl&UKtSsnuzBSh77!YMYMD+-S)z~(h!+ZKlugppW<PJgY`Vbop
z2u;>fZu@Mg{42_JP&f~wBeDvLi|78oZj&A<X+M4X(42o?RCH?JT)O{WoJpzr_WjRX
zohjAqX6267VWQrmO}bi7HRKye(lw-pF8=Q*x$eCQWvZoR$kaBZQVzW0DD}Arqw#n^
zNzC$-84{&4z>3_Gz<BacJ?T;+lOP>8Ny{C#nG`O5knn?x3vX_Fx&0aUXOe_-qv};Y
z$H=DWfbf8_fN*`eM#f#iHM(XavT@468ebAUQ~osx0KG(VSZ|OW_o73^lMJrm>Lg9}
zFP9rw&aie-DWJkfizS@PFJP<7qfpjH;XN^$^Sb~2_gS+sX8C73K3MB4s}OOom9qWc
zm%GGy14z$tl5^sCyB2r?T`+66f%Yxt=TjmSGymy>h}fojycEnTM0PYZ-nxcLDj0yN
z>8XR)_Pm`|r7%^2MPd7ENb_?F4TxTOt6drctr3>H`wTuh59NKUXAgWBT(Pd5#1_SM
zfzCVH(MibsO@{=@sx6D<>*1#8zW)lu9~V*lvDTLxp*GVFm<bp&nzx<b!RXIJH+G3@
znrh{$mw{jIcdbd}r1~WHq&8B$N-#O}AGxck>cN)J>5^;|>py!DnA&s)Mh#ZK!?e@l
z!L{LqG@6&}m}^EEPl%>swvpwRxR?~<K1ZPrO1NP_tzf&T`IY5w&U5y8N$lh3^Lz2z
zZZA~NOYLx6>X%6CTEWzT+!NAqyhbMeEkhBX1uc_G=a6|@1W{i`l$0?=MvhA<cBjgv
z0<B*J%&c`a<2pr%{#pmiWzm3(Zb6g_u+<<w{&2M{eDm<qPd_HNn;z#k>RIizMOwu3
zXR_-}@n(3@jic<?dD|Nuib<Au*IH&2bUQp5)Vs-Rg*ESktar-mr^Y+t8^wIn=pL!K
z=S)5_y<uv&gyO9UN6hwOE<W_md$$+AL~>>bv3vcb5d%3pLN<bTPA>1(hen*TBFQ~3
z^s{P%Sv;=WWlKzSX_`z|oUF`0*Z(f&V{9T1riRo#duwKugiA}_U46rW(Oy@kc`Gi4
z^!O%MZ|>rOX+)=|32B!<bI`<2A=zR;{NM>z;Ob#YZs3|%N=y*Wd4jSdUPw~t9ofXS
zL;b=0?qGZ@N9D@-p-26!Z)pks60GLj@hUARgZ!0l!?N3fgt-GQCKNYn*yt-Z*GIx{
z(EKYF=m<gDTS*cuKd*3yAG94CJ+}sUj2J);r&MOCKbH<%Qsn;j=W2eBa)J@k84S4{
z$I#=o1c*wE2hb=5c2@nZ{T9wYL0132Q$3$DTJ87}*tEn?6k$ic_a95WrwcSyt;B;I
z#r7LMPH<sqs7CWC|K2g9Nh+h9(Mgtm0kP3|vrFcFE`iGXw08#POr;r93~7^4kTn^B
zVwuj};DlEqk}`_7Zq6o0auiJxJxHppgHBnx(wT+nI$fz#{{+EHVofr^>rE!|$u~)$
zC`a9jK8A3*q<_?4fSitWhg3M<Mm2E8qpxP}U*TZ%rh%VkvhuE5t0PV`mfPn{C0ecr
zxB(H`5382G7BsO3O>BDSVE%hUbDiscRcNi)XQid~hM=?Yyvqk%&Yv)p=Q+RE!UGRv
zAg8J<oOr5{QqsRX+(BW$3Qi{iL#|_T+aP>NrK*um@u$4f__qFfJu@TR5GjsHM(4*$
zVBwin_y8ZLDpVZ=!q`v<$L`x?+dc_U!4m%p=-z?Q?$t!aGTnGmI7u8YVX+E5-1eVW
z2#2yu9v@$Ui<*ycw-SFBr&zg5|0o@1c;lU`eY7219{;ka-9N6`rCB4<>ep%eMi`fZ
z2cxm)+0Xl%`@d)^E=~=lS$iN_?s$XiT9Br!Wztqt1LkArQVXvda)JuD#a;e!{+p*w
zuEf}wkes+2?W)f`KN|CusxMR6FbJ|vmCM3Z6t}nxuIzkcZZZqP^?ah7t?r4p>PO7p
zLHgY6oH?eg>fd^0o6EFYxkT?I85vb}v21`0N2-Rnai($N#ghNqf9I=;jEoxWj&_Cp
zLzzgt$l4hjo?F=BG=t2!&$tC7EQkhin&wtl?lE(O36~VY)h%&S8-nfNkn8<{wo9Ma
z!(QC_eH8D5m&^HhPt5!~<!4<vhhRKx?8?kqC^Zj0u5iF|o9NU1;_e_Myb;}Sh%!={
z=yk|?CMGxZ4S&sX8wQmL(TA@^UBdETpM{9+ix*P`B@G*OC$5%9Urp{kdSAys&=bWB
zc=IKkaagozp60PDH|6erd@(JmaCKDZZl50#!IOk1Cl2Kw^(W1m`Tot-R83cJt^V>V
z;%u}xs@;@o(rdbh(|pSv65b(>sp&w!J<UtLSNz~x{)Ic@xNs~X--g?f)$v2<GOZz5
z^1Vf&zXy>8!$G{}pH!+c?>lVJI2K`goH%$`iu^WHDN|TtM1t_Z8eX+o-Lq!!5Kp9V
zMf$XCy|O8(K)PZ*FzB2rk<L<<s#K^El}5>Q+5=XTe+Hvd!#a%NNGB_<etL62eu>rI
zhve()9MRtrBm<>bt0vfkJ+%E&+G00~^>Z6<K=Tmv%LkMd@ZGOmE|ts-6}2z9#-qwJ
zP`NSXw&C+n!?EQnwU()ZssB-YWYZh{0Ii=e5g61wNKAnR^b6&U8|;svWptXI_mXp1
zBR+E@qfgkn9Jk$H(Y^=|k7}sVo8ZrIp2kxux>o6rBUXU@9y+adE02OxHrC^^z_3i9
zLab5wR*GYnF?F5^@?h@VibR@e$S(z#L$Mn&hd8*?R<|xr#hrSE+_FP{nyFDovLlal
zkiyrznYV}Ul?Y%fxPuJFtTJSbsMaQJkUrh$)4DoaD5MO-Q2g4>CJWZ5a-=J+M|afc
z=wlj!tU0z>$^zbCJT0(HC@JlMYnc59&3;}dr*HG#vB0uX9yX&-o2z|4j9#O`O=0%A
zkv=!Hc}4OTrI9kxj$Qky^u|h+m3MUzb=^PI7$<hEPi|i9V+=aJuHW~?bYvx!mI5!>
zitAD6>)t`b){3*FWjpC?9>*#{UwObjtZ3wFG-Guh^%vl5u%HYiqQu2X0(e~oIF}ca
z9_#A1Tc>Wd?HuPz00%kh&U9$`<(J^P^>1}TE?vYqJG-d!5?qa)`4w^V-APL3nUy=J
zY!3>!@oocW<1v=r+h6!i_BX>6OLjBO)O(2~RiC*W92+z2X)Rjyq3Fs_l9~!5$CQw(
z-LqT#$co3vRT5V&>6cvx`~1TTw^inuzkE4(y+q=)|88x+q&@{BWFsSAb#6_#x$%7L
z)I;RCYF}^O%)^UpvyzC5H%X!%4RGsh$>7Xl<&n7jMa=+$zgH->=CS>rlWirPxtjfR
zl4EMS5{y#8VqD@VGXg1rS=<UT;4`85mm~g#EO}AXU-Okb=kqVrM~}Xz)n337!mDAT
zj&~ItQv7I+|03wDNnt?98M+9of8f0o<6X5sqsFR^W@1ji^mv7t+<e*LFOt%5!34VQ
z`zSee7b2!<fO<b3&VvYYnrZVGZ-kvQQCLuvsD20)x##=)8+#zXPn2>K0;0ey(A8$Z
zqI!lnIImB)f~Q6l2ybzZB@RHpp#1pp<7skB|B(@W8{2v}+C|+*r(9$|a2K7FpGb;S
zOLk#H>JLNTJ)T@$x*;s^E+cTowf9?GBXLaBJX1-DH%3}p=$~!8+)_ZPv!!7wy>=t5
z6P<JW!nx%oNyMr1hE$>b6_|^T99(gEi=s%GvqPlHE~xgMGnEDsl>UQcLv)7o%3nO2
z`EY!Bi*`6Ot@N#(Zx@1UZ~k!m#MioUrdL$F&7x9qlGoH0dE|=~j7i5N>1-)Bvko04
z7V~Mn9JnAvw}QKO@7t;{<-Wi>7ycm7$%;>ZEGMieNlknh1;&<rTpikun=kU;rE(u$
zhIX65D-^-k>~PlieoXjp6Epvwlo%6UnDp1fcK(RfWCt7@<zLmAbm$?G3PwgoBQKNw
zvIy#b17{L<a91G)VgWnC`)CVC4BbR3Ao4JnXRvl)Vq$vq`0>{?`B!W}6U<bKwgTrQ
zyUH<FyT+q!U}(LbekRdXW>?lNPV{9X3fqfQ`+|y08dCt9{s3Hxk^W<k=~s_Vzo@^I
z`*Q16A8Nhh&jsul^@-CzZmsKk$J8czPU$7(%7KyHsMe7lPE?D{uG5GlObq>Jr6~+C
z7|t5n*u25vue#wK51Qo8lvI3N(zsSf`$UmGhOC({u2kXNR5**;DNGIh9b*J7nLV#`
zt&DMFIW8_I&Ne_AcU9}J-lP|%<~RDS%cFrxgoQ*iO4^TNwrya+$VVGN`+e{UuI-<*
zeDmtLVx58a#VtG!kp=!|6=r28HJUl-orh_;MCFkaZcdJk*SS2j3ZKjUWK?)2+=Ly=
zd~6?G9^IZ%<hrlMy@GqumBngT8S0`Z|KL_rQy}AO=B+miovX`0ncjj}?qPt5`h0(6
zWN?9ugYjZ!Ru&JTT3rdEG^POunS)RLffH>GrhFn5b0~XzD`I1&SN57tZq!+}G$SMM
zSEf=Kz`AA7@|8gc4aVnQSw~KvK)Nsrdvb7)@rm&>GU5O`8Nk-Y-*Y#a+*4*YkwE_H
zW^TR1<8|L|BR4B2?>XLAJ6-M2gIUY)>6PHe)}xBor$<!8VgF?m9&i@9C4iHURY%G(
z#~cOR(3kTr7R&xwiI}Cj@cXsv?css1Gb%`epjf$#_P|}P(i>BHv5gcD%}2M+A$|JL
zzHxUE7IMwDXXY|^C;2e@tT_3;q6*uus|pspQ`cBai!6N$i_gcXxQc{{=>4&rBtsrG
zq2F3)4|o)n^x9K4EJ<pHsdJCEULzcymK}I>W_SP9zc)VdMU{c5f6q`$M<OhC++&q{
z1xze<c(8T!R2mu`ZIXR}%^Wph!<S8T_GPQe5Uf4LEjr$obBeLiltSiL??L!!xW&9h
zp3hNJzB&Zq!whhNX8|yn1jVGr$Pfti>VRZL`~d}xjE;`ZEvIP&?VU*yTB+q*U}Y9G
zv8uUeF1`G@g}8cf0Lb=x^n05(hF6@1nu;zU#X~F9(K>TV4fk|O5Y1&6bG9;EPibV;
ztAAH}b*F9!*-c!(JNzjV#ubYYlBwpdRh<Dm*_Pft<g2FlT%xS*rvOD4P<Tymuf12Z
zbfrkKmX<zO*NCcvr6@QtNn!TPr0e8oHrIpTrBU9Q!*R4flin#cxBuArwabm~RIpBA
zo9AnL;>c=A`Ai8Dk)4!F{^9-y&U??o2csU$hjw%|O#}p<3i+X<eZ~!~EY%noO@A%p
zA63{iD*RFrdVTJI5kPG8G6L{jw>Wo6=Az8ebBIJAW}44_ehy+%1sDtzhV)z?^_`8)
z%{a-I73t{#01gBJ^%LCTf`fy1y=B4k9R2_VPdb01!P*(-8P$gmMQx0Jk58wET%LxR
z&<yypFYX$n(f<k8t&tIbAmzm24n#{OLciC3=_L(hK5Gs;`N(nta_FhSWIy4#BpCQL
zE%z)(*yhFs?Y<AfOA-jd(fjT-YMT8jg{JK5^6zHis0DrqXWeQ%deqmg)YI@8_#MW}
z+><UypI&&^9}lo1Ea*&GZuOv3nD$r_wX4+q!n42KbmX{#&s)@DxJz~?ZjC1Td6|QW
z*1){c{C|lL5K%l5219aZk}=RYzp@A%3~%=3=MpA$BefsQdHAUSG5d>V(ZwucVbN*Q
z5J~vY)(o>a##NfP^~K9T?sgf0i!0OU=7UMrM~E>0<HZ*o_~O<_C_igD1)Tff7lOA*
z2wq#dOWd=~)#PA<O-tuBD=e{TQ+?u*OIp^E&CT~AjQqpe9q=@phDOYic&EW&229SG
z%{{PE<pe5iR^iRVdOFHib|dBa54;RZs@3<2aY6rMaOLNYq6x&#I6pUqSW>L|vQ*Sk
zWj_H)q7;}~xwGy7%}VPUdssFL)*BtbO&W6wlkBBKGvA8*<#+PB_GWoeApSkGUFK;y
zCxz)v<fBR9e}4|}bojx#jF;iFSk&*qYIw%UPTXso#4aTgBSc&Xo-m{c#@XhOe5j%Q
zbseoMoujy+zyFV`sN~!#eT?quN`<linyo%m_q-78d0$MP)ui5ZQsL@xi;%b2!c`jN
zXeM<;*wM0<cI29x@M%l<p`}?RxY1}x<s^?Zu{9E!(=)AFk``DKZ1!{O{~moMt@XNo
zIzw3LRPT)7(6)37%U`UZbE(yk;`d>X8Fy+d{hq~kj;Vihvlg^v5v)*NONZYJXqlMw
z!qNH;>!XzyCm?+a1npe{Uur}hLgT@M2U-RT*!4YcohNfP))ha#yy)j8Iu_bD>{*GU
zMrI{x>4lDalO&=&Hs_oO6*y7zHkuyErJG>Q_95%CHRw1lK?i-fnF=%5+w+Xn!A~pg
zIY4E$QSSFR1)$c;xt@wO6N6tO&90>WV`yxAw6Q5U(ZBbI0AW2p{>MB(H@WF-J;uo!
zY7JCyIiJ48KGcFv;laxzuv!CZ{P(R4rm$^jf4dyaSp^G+N&QJB!iuLN(&Xi*UJF`=
zyYCzgb_^NFgtP|aYH4FTp?79TMC>-_e=?j`$uJ4+c;y;UCUU{@s$(+7cUCe~6|?^)
z^wQ}%<rQWJx>otA!CbCj>g<u?V8lUlp$&0Yr!UYB0Y;&By09-}p5lF2RDX%@vvEW*
zRX@kqIL*(vN<7iD`M5Y#IicM4-qWjb7sg)Ck=}o5NWJ<IgOI@Lm-EhRU#XM?M{=}a
zSIw;$p?u)v6{ir+;xrqh<P1;HO6?c63<N@s{pu1T=tUjhjkwg$rt;3U7!F`_6mvdn
zZaCafmEG?+SjucTzQ4BTd2=MpF7aj3vV#7#1qB0#GumHv3;C}#G#}nKvE=m2UHC=_
zuf)z3iL}keYX-=8di}yiV(d@RQjI%(x<*0k%8`b|GTU!LU3JBbWMNoIbIDO+`+h5u
zA3E<O_Ay_%aKGuolk!OQB?Sd45{H_eScDX*5|BY6hL@d8wH$;;u#X26ngZ?>SA>n5
zAPyH~^E7?5qBOrhK15M;b^qxX0)lt^{r7`p7<+*l?8)=SY>cb6-2knL8~7kva^Q!o
zh7})P_(Ucf(_#JLRre*4=Y_cr_g}sggJ{?A)hV&|v%+=vP$=&%U$6;HfPk_Xz{~D-
zN&{jBc*4x=Q2kdpupRPL9Ph9dcIPWBIvI-_k3ETuc_GMp1_$5w@0`Hx+7UbEYJj5?
zz6zwW-va|4<0pT=&hXz@d|-Zy9-Y+V;}fUAGh7rSmTQU^7rH9=Ha4TQ&W!wF>wH;a
zIj~j_j7fP57Wts|#STsrBm%OF%Xsv1nCv$hIk}3i8&Y*JWA+_P{U+nb%cHOhi7*LY
znwC_GEoVlpqB(ey@)jwZSvCG>tq%%465$maRddfNLO*_yTj4H6(edMg*7!T}+T4RQ
zS?>ntU{RDz<3*t%n~)3n7nQ!U)G^upHy4O<18(rmA=gApcArA!ksZfn&)n$8*O*fp
z&Zx%UoB{^+7Qp=7uzPa{9<=$mEEDxQu%m}I)y6D_z|TcMK)}iM&~O_lPtLy|D`jW3
zqA^HaUC?;r903#CIG8w*og%XbZ@Sp=z~O<-X{0Lu&<rODskgG0!RdJriEJrQV6HYY
zvA#Q?LdMU;^$5lWcc9{kw|8#4e~V7Vcl2AeEKET_W=D~Dbz)ybb_onzr8ivSl#vr!
z1DrLmRPKp}r)8e}0*<M^aIz$mJ02a+GpvMgeaX?BV7y{Y97@ks)5gBh*%F+*jk_Cr
zlfW9dNB^B+Av4bfHlW82IITSUV&`pXRqO%Uj%dSQVp@Am<&?nl@KI<8n@qk9UK=0e
zHUgj?NMXvif)_dkD@PeOv9LSlkCr>G>aPRMnWN{S;TD8rJNK#Z{Tdod*fphvz*}pg
zVc$GFpE+<UQ+{5^9d)M}__{-IH7t8f06}>1Rrjt>L+B1)J=*c%2v@FcNlN>&sJ*g}
z*4v48U1O~%VQDG}iV$(wQ;hpGy}^7O6E4+KL&{O~IDX=O^E$laN$~@Vbd_Rqwzw{Y
z#VyC!rAsQ#`IvEnBtl$!K)+e=HEVTGEDvi9t*M&uHaYU>)e?Q829lH9dBNjj@(TI>
zy)I-0)0p(#96Z_%S!wqN#bIy-h`8ng6<cK*=J#d;i$1{I6FCsKc4nbCx2E9Z;&7((
z4c8L3xrL;)R#uZoCMK&dOwQMaaQY^|Qe&%XjL0PeM()CGxEuv)30AB>ui=3ub{9Op
zMO|eK{s)LeNiOnh$nIH;d-ogcq>j1bBf;?(M|bTiz<O(St5tozc1S~w2cNTB2%o~=
z-mfxhz=iub1M8E{eD2MJ$#sQb%P~7E!FRxfs2x+P3O4_VeruP8i6+NY_rxH)s%hT_
z)4FjqC0ug5T>T~fUn+7-A9vL`08>m6bwy&(SXvK32^2FmVV4HX_?#qyrJBZl3?0fd
z%soc+FJFu79ZAE#LUVR6AeM5s53?k8v`#9Ux85gxQ`%63y<KY1O5Zzis-%_Wd?V-4
zUnOE5@8eYmJG)s(22p#yZ7d2R)fmule0}lSungAH#M=vlaV~jj0wstja0S?S9-JF3
z#%j>u&qKiCo#Az<@nt)6Cb&*>dijHinw~MKV3GNdBz8R--}2aJcJCq2gvsc>+S33K
z3OaW^c=~HF{DwcBVTf|9T7t|+l<xi28(ii_A-8y0--Ocb7Q-=EyX5#~iM|?vWF_zS
zk4LmS7n1B^m+0I6i8ekOFfa<~;4NCE>%DSdoU0^y4}sN~C-n(^$N!H1I#8&vGvb{F
z?O0TKXx~qkYBBkH4`_sLZkC_apqG+a&;0Ll5rJzNG#D*HJ3esxT=Mz%tM0hw;H*5Y
z^xD1a*RLD4cU*u87XX<|e1?zEQ>@(uSmKCKU2-I$gHQZ8cpo2M+OHv19hLePd?UVq
z;i~2V<!yvHKfU(ZnX0hV3tFFBH;vz9EquQ2oie)@c5T9>ujUoAJ$&VEb!5r*3i#c8
zoAYZNS6EYO0n54~rP+M)8%<*F{F~t(w$ky0I>F4(Bu*1)!Y4sJUG3-;C^9_lfY~>|
zT`i5l2$A_tK_$&X4;E?u5>*JV=-<O_FAdr6Nf^mS1DDLL%#GWJ+4k$8s==%dlO4YN
zIsHcXaLEb3=0S;m#r2h8CteYU`Yjd+kl>Wt>(nG}qKJCFs(*c|cmMr-kSJyywd4*^
zRB8)dWshE`3q&ETgJPm4Dn=ZH@giWP!3*^N(7pRX>oe_fA3n@W)>y+DEGs?zYkx!b
zG1>O2T$p<Im-3%c>8GwIRz#sZ@mxxExM<-L#Gyy_Q&m_CbY4p(*G5(h!^zLatddLA
z^u!R6k}FarvnR!&%gd5MB6s=mXZ5?A&R}f#&T1o&4y2_%W0NQa3prTWEl{LT`Tmte
z#Me@|!$y@S;cd^WJ!ICYfoDUdz{3he;mQg}TK3BTe5;L8li*euf6Bd*SXb8aYp}A&
z{>8JTjR3&X%5$A0?1tXP6&Z*b!N5Tb{*+A>PG>2|cZ5?R=HWbotL=fh&^_(ecmT}&
z7hqD<0XBP0aJh6_elThydS8TvhiBS10_UzTSCbPwOO2Fvc<4`0Zth91Ho7$yRV{^k
zjM-;?yTvWBplu*#`x-U6W3%*0Rscr6*ubX%B8}#eHbBuEF-q7jKBYlU34NJ=1DB{4
z^26O9KmYEltX-k2Qbq6<*oe<b@XmxPQ#eK}T{=^9ABj@0_$|db#Hwt&)Wsup&V3lr
zOy?oSX4bw_hE$zan2Yo=X9KF?TIMfeHQe{tN<+xkVHI4lQG_`?Y|In9diT4O9&^gT
z?uhG)swwSdTSR*U(CY~(+dPYbE6IP=dtT}cb6RXZOmZ;A@iP%Yp4;M9xV@lzSDIix
z_QHuIV_?@Ct14u+G21B(hv#wS4PoKj!XE}*J8EDo+#be!LweO@Y&79fML!HS+r@RR
z!3IfAwh<!LQ+CLsQFN$b8V21|0vK2z$0bg`iY30LOrtZb4rP^Fks^tI+<{hMRe%dv
zG79;7PXFZCm#U-g{HrfXLR<j1v&ChSH3?w!36jb}j|M&j@EOf3|BPCYPB1~LKGM%M
zL(9RnIQ-?JS&rSin3_-d?5Of6b|ns}V9j{MzU0pz8h>6%M_0q>FOCY!mdRZ>M<mfR
zC|#Pm-2z*%9#G`BPPc}8Ko<%KR|eoJzDN527JdceeIk1hcq$fva22PP+8;q{0V|z2
zzxDKqRqG@a)(n2WuI`*aeF{=Vr@Dk2Eno$AiY(FX>fpg@V=tI$|Brp3z-shu(7>%O
zuRJR&2*CB2lE&i^w2mf`1O&qG#M>`obC1q$9$vXDl%VXXIy)%ekJaTeZ*%;%#`#B@
zL4YN-q$=b;d+l|4*b(te2x9g_m+0@#r3T@oIF=-p$y`ZNT-^_sy&2eDx#*c<R-Z08
znKQbE5x1MCyzzq->YPX)J;d>;!*6Jo=<5O8oyX5YMH2shz(@+Oo2f}PlI2v<#`r(x
z(i-%f()_;nIE*3e?`>6VCa}DA>L^#{dP~bq3;czf_hzx?zM0;kxlRw#>A?l{g5Ipi
z!u;l5)oiS4a*^0RP*7e1H<egMN}vDAu&*r2%QGFVa=PR?L`<7_1q+5wmAMrnFuPP-
zE0k{zA8a5_q=4EcZEQG$&xyL3*;Nvv%m`w6z;E>}k^Kg~aJL+ct&V(3N=oL5awOuG
zo#?#_UbFGpeo)HUM8DxzOVxftSM!r*->56(?YuA#XhCI|94W2_O{?zH;9XW1&_Al8
z|F-Zc?0PE{ZN{S7U^uWSiS|DMb}z0ROLssHGyy%v$^CO*uj4fD1zOR|Ubx-j)b+x=
zG(MDugx2K~H>s|H#92PyGsz?ot{mSo&wp+%66xH&=vV}|m?CoGe2iYxFL)IJ9eyH7
zJ4+>~l|`>r86^lJtVkcAxkkN{{M0L68FXJ+{>;n)&Dw2afah#+#p>FgaSIxErcP-{
zE@YK+?tu21lPGO)hilt>Jh!IiBP1h)6MvHIYu;#U%DguL=@4*Mr3rhP9>DH=TIcSh
z)iwgCUp2A3gC%&Ty%=~8K;lB3N1L^|IXCD>=7=U<zzHv0%FTTB{)VnI*we11o*ae{
z<1{V-S(F4In_hHz5RWs18qBE-oR|^wd6SqmN%GFP2V!>In^(2?8`f+indDBv_96c~
zS&z;DFHYY1_ot>3(p5t8=taWr?l|s8$ECOcI>c;ylM&`)H$f`W$&9+ww;*0N++VJm
z_7{8pA}>Sr>4XKza3yU7Pp_2?d)#f(Nw00mo0iA-Tv_eg-{v$IX$Etp-Fm+?q{zva
z0!(p9EgJ>>T1RdEdkXR*<`1vSL{@|CaL}N|?dr_J!U*^+=3mHI@iR?>xzj9&L!t_v
z<MqIs7$g8vK!G;{gWeZDG`~Q9NsOuBlh%O90$oEV3eaW|u+Q05z?BHl)0@PDMTD$?
znHas4n;j0<!g`T|PdMJbHWG><0Dx~rmTq``0avcx17G9HfO~0Y{Ai-(JO<h>fuLpH
zY4a2{y~$PV|BhCfT)vI|c-FQeZ0b)IC6&n9?#Lc%U18PG*98wPhp>)-;uf|z%N%G0
z{m&L6Yosvych;clUq+H0z@({0TdjeVbLQaE&0kavs=Z@bP+mX3E}&KTQ}QgYAvK{`
zuvGDrCA-O2Vh<#mXL=UJ*IAoVZvTAvo5ViNGe@)hT&IG90=;@q^hM4$ojSw_RASr&
zejVgvEg&LlCO);S$UT6ToB*Z=px!Nu;DiwK+O`8;>Q+!2o1_arkzhxB)mmFyQ`!ZD
zh37E^#nYX9sXscVQSf}_cE8fy`J46%FlWBxibajraIQQj^hgFm8<+6fvFj52o7C9p
z^>Ma#=5-YY*|2A$-LU?%su@808&9qVS&Lg|GH^}1p&gN^Q!F*UjIqxpOrt{UxVV(w
zgw`vH=CE^Vow~Fr8KP8%9Qbo_ig)>b=?+<%ov~P7g4(7Vj?^>2&iFE~Pq0C=0zLy9
zh}r=q=J@4&v{NAkK2!nk%5DvlLVc)qkp9Oz-anuAn%fV34j@+FU)-XJ(~vv#hZu5z
zUx(F5_rHby^B;*ImNStv-6dF=MOb%Xp$>N6=;X^UDcO;3YYHIu0PfwLwI=eo_$7aG
z+DX9NyM`_l5o5yu1+{omQvFH!m&eI*CYzSNiqREpcyfO|7b+g`=`kLKAk7^A?kWJx
z>?|7_wbhV~lqqhXS~8_ggJx*X8p@p!_)K@}?Cv8(&v^z1-AMMhgKdgB6DPu8Hq1_s
z27;R!jpo<>I^swpI<JEGZ0qus(O60LNDW+ocbWo5pH0(XR@uN+2pA05`9M%3epZy$
z;(d(l=rUhoT(!MlY*-ze!)w5Wk&puDdiP+W-|K2G%zPa%`DlAa5Q~5`c%i=VF+_oQ
zc{3=i&(hM8h=9YVX;pe%|E7oJ)L(P0(-9+7qLvSRSL389nch80G(o(~J2-2HAFPTr
znd>f9Z)M0_5zsoN6;$yMW<qn%*do42($|u*$5h3s{!AmJRp)l~5x|p?s-?A1dwsVL
z2qBks2X=t)mUQ%bIT>A>8A{{E(adK4EFELpIllexMeU=3+=mQP1?K1V+p+akz<b?T
z9#b~YP9l?C7?Bq9=klPS5!&$mYI`3gqds`hCeiQ}yxtNYv5&jUK>JT@SOnxWV6Y)7
zrHBJEaK|=*Z@WHhoeF>mF)$27hKTPQM!tK0h^#<Dud2gmtL}^UuR4rTOcd((SfiH&
zdW9ouB0J(UP|JQFL~~7VxJcMfQy46)XmAFXS`uG)HuDUxz~y?fG4#|n3NDa~IwW$Z
z=E6JJi^?2t7=?)Pl3Um&%@!C^-7?5q<eD@oe(CO#utWkJ?U2M`a`)S}jHqmE-;ZYV
z<M99u@{a6duOhu0q`w~puhWg#013XBGsVf`TPTAxFlcGm#c!h84PoMMJl-nlcdZLP
z%j%3rRb~_1QR7~{2dx-TJ<P+`r~aY=HcVg&edr=5Sl|5dKU?WlJR1!?HDyQmtxs5=
zuPimjg-m+SieAU5Gge95jn>*&k4X&-v4U1(mI+xzY3u@LR6Z*Dg!*TazCJ<Ton_@Z
zgNF_x=Z!KJ8}?tD7{kKZ<!Q-~|6}Sa!=hZHb_oeVkdj7pP(&0FB%}lpEV?_CjsXGb
zMoLmbKxqSMq=pU=Ndct>7&?Uk=@vcf-QRc4b$-ls`D6ChdEY12y4M|KePi~xsGTcV
zLdmf{=yIZ7eoZjHf5>l4EZ2Ih@9VU2^*{nBWQXt98u#eXk@Xgt-_enU!I79o@;6&@
z<e@=>r|O&d4a>}5;=`=jQ$W9j@DZUpbJdSMhPD;pK(A!qT42TZ*T%T;Fg*Sx@OS=-
zbat8GE>j_5<rT#)CWwYNM@tbRS#q%+%JPw#r#xCc(FP8QU>W1>@H(xc@(F!X8$c)n
z!$xvGy8ls^xEWRIsHcCLCX?w5kxmO^Rf32~nw^@>n)MO4y_hOg@v{6LmoT;|XRN1~
zIHg!~m);Zur!@PtcmRmK9&70&q<5#0+q`wf6K~BLwPWh^Fd%S?fgUX<y=QIjR73vK
z?gcwrTiph)N_?1A|0*bQrr%g)ASIEF^d3w^-|m?qyA!#A(Cr4$hXP2wRk4n^F+94G
zo|u$ofUZ0T?jx#W8Of9HDh9prM~#Q__BtW~p)%FauEjT!wlAB&_2ZpTV}*;dAY6~6
zVS^jGp?e$F)qKS>9IruUo%vM~Wuo(6BC2!tN7`Kp&N32~!=6){n1>8%Gx<yOf2rBb
z6bs9CUA?_1sv;Jzm~r4X(3=(Vzx@gobq4xdQ_B9X8qc8D_Q5S4;G<tQCJDY2l$14~
z%<zOpiUq7J$6}iRhlKDwzALdOfUGOyN)ev~Xvrnlyg}!DN%J5Q3;q(4ebmBnBroo+
z;PDM>1zk*3S`0^&UKwh*g|pg==j`iD+C_YC1zBt>6QQ)>yo96R7H2M~lV98^CMHmU
z`|_?hlM^Rn+<^S06UT?qkBT6|V`I7`pB26%w=48{(3Sjb*{0PO<^k^p_<0Y`E#D45
z%d=9f=TwLval<l-4PZ-m-_=;q*Xf*jAKTPC@E5f2=Fs*ObA~>80iakIqHG5`SF2vO
z1NcK0nlw#zmNo%j79w+#lFW(#bThsA%Vr8PN<MAdJONG7Wqy7>zD@6Fh~^j=oG<b{
z33{t_z0l-=yq{Vho5CRf)1=8D^wS4B`k!B8f{Gk68VzaYR7Xb^zXQ+y(c9sKL%^74
z?!Lhj<R9JHc!OtW$z`}rSk0!(hboEaut3E}<AYes--MHYuEVoCQ4sD!f$+}mM8YQ%
zR-&RaM>Dh8?5pT?_X|o)1xG?Ys7fu!NjF-z@i%Z_vM1TN7*`%j6vg`+@40$(bYy&M
zZg_qFmhB)928_o^saWL@p_Gs{4Y*`lz@kiijrR1<Oiq5_<Rs9)jf!V+goy?c^<IB4
z8)?H&;f8_w$Ew{Jh8iHQyRRRk1xoF_X9oKEU9e<RhL<i&OJ^n}B~_fig&o3Tp_Zy@
zN|*OMJoxtD=3<lbn_Jx?TKJLKCw+V*xL$9{aE6M>cO2_C;XNy!8udWEZ*dLK-mj&8
zHRIE4KFvB!Onay<=Ws6WebL9VPFWn`0x+<aNwkb+#DkZAHUUSdnWCyKrSNBMFWSq}
z`?8KcMO6`+Qlh6S8i1iW>_pngaPP5om6-)E$(GzJ@;4-eTA@td{%Bpm1?@@AYfpeC
zxmYov3<LO6O;ajpG#|9b1PtfST}0WBSI6b&^V-_lTHDxMgaI-Pf}s0{vXBb#wt>PN
zd{*@J^>Ygf5HpI?J)ad{tG&PfUpWPI%iDV0B6P|4t5sn80S9;mdqHXG&w;3J_)6)J
zHpIhv>Xo)ke!6S`<>%qyVaDD>qdh?w6?->VH`C2`AIehsKJLcK`uO*h*-#nqdn<&c
zzZ0>=ppd(UzHT3W`2~9Y%fx>utK**q_V3tV+Bg}GRYGmCZ(QGq%D?fPqc`O|fGh!I
zZcdLFRfoW4QOZ;Yt2b+E#1JJgx=7qoB47TfmZE9`j%B?n{7C!>2~Jn&l)*0z^)K1P
zReO?ckS~{4P~Zad)Tb~zT&;JRE&0gC$~yA-2{EK$z5Dnv1b#pAKpK6-0FQqHO&dZO
zf&G#JNVKYffx^9eFP4P)=hSp{I}k2DQaQrG&^v74{m-B4Ts<aMiE(J>4=H&_Js&Ns
z$Xx1MO#@FrFk~ZT?5JvNjNIBJB$N6h``!GZsuh%X1wUPO?wiYT4CRwc1+~?8&7vhh
z8QSJQ$HEbi)#^$>pmhS}1u^@&xB??DX-)+g{&ok97IF6F*@_f5*gNBti-}vWo|0S^
z4}LkSSXS^%LQ#%)&An29JJLrHkMF%6pM(nHq>1idtDeI)6qvHz>100?+Ribh{Hnk^
z(r6#Bv<X$Uk#uWC4Y*<;Qe-;gsY>YU<Kut#kw7pXQc6NouUBb%4nJJ<K0iMKYT`2c
zQNhNem23aYS~`7-==1`Mr~^A#)tW(w8oT#u=_)M2#H6I0lf&q506nCgmo{R$>cOXa
zmi_iCnF|3xI(Qrd_zXZ;moHz2zXzOURJFh%G(3Exs_7(JV!i$rU*97~4S48~H!?R_
za^uJM0=6VIb(&=>-C94Eubfr(Z!J<Ig3wFen)bty$RY}pm!X_4N)KmyGibRlUid8a
z3|HVf#xGcIp07oI+$s>OC0r^x)OP6x&T<^!tA_U{#ttbYh700yX>O#8pyr4Y5IAae
z4NzRrS@UQUo`kYfrLs<`Oc}SD^)NPDvgwtDtDUE14+_G23{>VyjOvDW?<uAtrz<=Z
zM8lHE*$qb#{NYBsy4KsbZ>_DZv*3||O@BFDC5W!7%%WQtyxhPa0*K}GXGozGw}Qu(
z%M0dZmdyc#2Pb}^XINUm6&~qV5IzzZy*xYFX=`=uUIcYf1$(>Q3??Th6ll3#ya$#0
zK=;^y${z_51Nj3z40Pq><-aY8%%Eh!Z~>YFFa3<zNlW-UGWMymVLyl$m86RSqv<m+
zu5qoLyAeEaxwbve%KtJv8e?W;)p<YvC^TG>n0-cEfqIkB8{=n4I2hhB&^*`t&ptHS
z$y@?UIJ+YaG+M(+l3^lZBeIuY-E?P$nG>Xvs;Q^%>fg|wkZdRv1^ws5SNcn~GnnS&
z2cPpYbVPg>YM$XzhpP(n@{mn|xQ_J-hLeL~793AM!L|l5Iso>zggwHZ1qUb^S`c`G
z!~jNpKl}SzHfFvevgi{{ui+|3ldi6=w2Gy@;}e_IO#_I6>syMfv)DqI;6I`^lu|`D
zVPKsG0+0foo}4}*uoOs4_BQaYA2;c!3bpTx&TL|Lcr%{P<3HaIHu4gnSgz_$!m^M4
z75su+V#$|vXGCaqOha&*ykyGuC*(jj3688*$+Z>92)?CvzeQ)l61U#*$%yP`sm72h
zUS8(6vUQsy8}u^*+|;87Zw2_|QJHj0*9|8AaE|#8?-r~MPn=%Mg7H>|Xcw9Hd4HHi
zz$92->d<_Ve+K#W_hZ;Q!)dv|*;N|iG@_UVf{Lmp#S0=a1Ykl*4|^&s!R#yu#(+Hd
zaY5JEim?Q$EFb`?9N#c$OnY(mP72^Doe`KAK>Fh|*Ber&Z!Em1)0@Mjlcy(W2zQH|
zlQ-b^ggPW{KL!#iap&w_BP{LKWjlDGiAlQ|Pe0~667E;`{!};!d=bDF_zfj!BIn12
zO88)=z-AV2bFKgRme3>CK%~2I0(+N$+j_`rSJH;tEl)g6ZfMxQ_Ln>fZZG!g+cBA2
zFCHHw93rV`BORRwW@hj4VJX^<jfdP&4mxwZ$MRmi`UE(3F3zinjA;#{ptAoCQW#sM
zaEvYGMVVk<SiNrmmolQ|0a2ww_uQ?Cg2F<%p<|QxeR$CT{zlBC0|jQ%;I{YGk%;F_
z?l4?6!}Y_<6HqI3<G7&ww>9ixtwMk;sMoX815F4V>2A#bHI^$w^Qk8s)~6@BEE#E`
zoj6)r`Ebey=D5BPkkn`l5zHj*<j`~O7=8``7JWxU5c)bRQIEZm`rUs)bSlT9A|hVA
zH!R8#TScIl)W!TP@LiIadIDS5M*8;`GEPP4GogksHtzkAeNSpl0*pSY_Yc3E^=k&@
zR(w*srI)s$;p*W#W)jPilc1nK@6K0gDD_;&eZM{;>rP_u(mC~cqx`ATLvUF#Q>^#W
zNzGwU9nxz4R_wCl{5~*~a9e)fjJ8HruB@n`s8exSFX5`EF-+b($_T81u1P(*=K{wV
zG}J^URbVVB*^QZ-s4=Wq!@N=9X?`WUZj=Q9JYEw~1_5yG*TnoNh(0g%7JRR{irUA)
zCru1=E6{!!xs8UvoWy%<A0gTRy-C3FAKv8OTNpatM#uK-6vE99R*_Zt$=i2(Y>#%I
z-Hjn=WKd*(ed@Q;VutKe5ZVV!QO3(YI;7(GSz(c-4312T%K+dl5%|Dx=_WxGSurW&
z%MJt7mOpBCt;@g+#K{CsURYNG`ljJg)AhPu+OWH*cqP8pXn~wDIFbil|0?i<u)4}L
zK#DiadvaoFd$(5D;i4h40jTHk%nf7U-TN4>3i#K95+nii9=z!G?!LZ#$%5Mq6izS?
z`c|=)>e$!`xM6pN)D8MnnF}(HUREiSb7BiG|LA|{CpsYO{q?$mBs2-Nt_GEu313!M
zcjRo8)4_|&P#>P(2*K0jyqE}-CNg>1k$@6ZlG}ntSno0};%{a(wW;Lsci0jZxV}(<
z9Ct7ap|uyghbF<rR2)kvY}GeqTHD&Lybi(D){3jFSu1UwftwKiJz{+AZ^6<Ffu8#R
zG9#n9zT-wvzW>>b^<Q;M{T9gtPUK6#gnW|$``T;L-NQ-D^Dhz6SY@<tFO&V|Fr(Pk
zaDSr%PH$bG3kv6`M~3Nmf*M&MdFIJP<FDsLN=`FR3rypOcb~OaQ(Kg^T!8If@xzs1
zP7gRE<!e@7nFs37&Kwu$4^6wbFdW?#a)<fuzxD){&TV{WU1!1<L;V90jM%e3Puua~
zC$D}=Yk@&hQok88ZUe|W6($THPd3*;I|jEG*n5Dn+k&J2u|exGBKM2|+tS6p%*?%Z
z3$LUO)?Tt5&Tj{IUW|;CeT?=UL|GI0GiYdi{`NH^6~u^^NoaS#y6_{9tBxx?qHN*m
z?`XTxaRHg&o++z=aYD-XTj3jC)j<34m!IdMEyc!gzul>%R~X0U>w~+|=5-uJ)#y(7
z|Ly_K#~qfVZbNGZFC?X}u{HEf8`rrN;lq}`vpLp(0WfGu{C<f$!k!q`v-)VhT58rV
z2zVh}L-_epbln<;-S1!qys6-4I00<74PhV2pnKbZJNFroqYxd^*3So=VxERMRxc|E
zrx?!sV5gX+U#{_4*&I~kg|=*vu_m>_?Ai!slFm1D&2nb%JZsi|W8r_@0jcT`Q^0`H
z(oHG4>D1=yrj6lg4(LRsQ=^(``YT|)(Lslt%SzI0Koy=Bm*L*JcFH~!ed_T}`Hw)x
zwnR;@^R*aE67cF)R=iA>3cy|jn(r>)B(8`*zOPqdO$%gBkw_-~90)94h0MZL7?@s_
zS_C&xl~8JqE+o9SLf{iP#3_kRw06XCX`Cd=LqJ?>h0Ors0xdg-B&A3~rZ$+6lz~wJ
zQD8uMyZKX)gaegg-%fke^Mo{%(aW)lU5S?x<Fw8twa(!J<DP%dliz<WH2mM|g?UMH
z_b%-++4zch<jYRw=80xul(X=&Px)i1w}@Z;Wi**Af`;#sc_r-suAEFN*l@H5`c};Z
z4c?(>tBE^I*^GO+?7zjI4%foKQn$wWBc7QhySO+8VYneyxLsF2ZA6;<34+<)@hU9G
zG>8a=2xe}-R}K)T*)#*N*a8^jJ+e-Ni_-}hS(|_qx7+;0^nfjR9{^|R8@Itb0_lpu
zQLQ;YM_lNKV#a~j<n1=;L>8C?1Hs>6FXp*QD+{J*NPjuUy3V30CX)p@8O({a8mJBe
z!7SqjS3nanrJVWON>ve=m+)JDH5BbTf7LI;CwhvU*u%bj*-UooV%^`PY6kL<SAS4@
zoHx6+ne!i`i}tm<M~k7NLxe%aX18?=47vyNG-^P5m{(kU-?dxi3CQOWK|LTXSv56c
zAH}Gpc2s{i{qsW@e0vk_CXW@&GM%aca6yt_Un23cz_<D_>N2Va4yA}Mvr+p^o1~7h
zw#=a-VybW2hKU*Ly(zBt%bu{^kb}Dap255|32u08?PDkc{RxJib1Wa;1K0@s$tPWz
z=QKsmeX`$^Wz-nDFskL#_kZ8hxkA?`>~g25LPmb9`u5C2uwEk5oZti^(njXYyw@74
z(t6zx<(|ji!=<{3MTENvyU4w3_&r)fTe}Ssu9}e#I2yM8Iot*#uzkexLl00RaQOtU
zc6F%iU^)q1#cI>PzX;C*Qe&|y7IH}N&LP}Mw&iu+adLi{^^G&vS~CyI^hRON)y>)^
z8+<tS#f)E3YgZp)yf%SYEO3LE0gY0p8YdZH%^zKmfnL0Zy86Brn<1H?4Hg8>w|6E-
z*=pWXBNCFRE-3AQGsY}kbjzWKafZGpj;UrhQii9w!K7~~<qOV>)gOxLV$OFiU}Cp`
zP=NuTbmjQ=QkAC_ZXo&r3^J|(FPHnW<)`OL!3VDlJ_xyz0FWZRCO;`u3zU0lQdXO+
zx1U{LGJHmuY~=O^&%>=0O2xM5IpysCBusC8v+FzX!)et4fEz4j$a?cz=aneDp{Pd^
zS`8+0+61x0zD7#|PMu^oh?QyL27q*h4UBr(a^rf3pSBNtGeP7|%N2E{D`OLDv+uco
z%zLgvVI5qCT>B7TRqbUPzX&@=0wRTB?Wt*LeFK`g*_hjxAQB&VWxZv9d<Z(vcNFrD
zVZ;DPAGJ(HgO9L{Dya9f?oRUiNqKGDCJ*sSqsf<f=<2c{P!1Loh7oHu)~P_IF~aPl
z&lkJ@Z9)FYiftN0$pH<4A;sm)Di{)D`nYQ_Fmq*2@IuYHq+e8a`z?HIq46)*{`e3c
z{J_hz5^=0es;};mMs3!6-+Qr7xpuU-)(!K*Z?KA1j`SXmK$rLx>fT@Cq4eOMTWG|F
z+hT@4Ao&QzTvi`p2J28gLkbootjJY$$G6_SYsEyB=#EKMxkJc?gfM#Fb%(FqiUW5;
z9@h$!{@aP<K^J%l2xHIwM9N6+1q2;+Z^3_hy9zm5-1U2Frr^`sWnpk_<UaI2EoFJW
zPcbJx@ZIIsup%e6<Rn#cle+@PJ8a9IQT3AROabTN0zoDYKyqw&miDN)X7R0Y&7`A-
z20e_R1RR@8zYt;L%7+Eo*%P*egQC5GR*RCJIxaKM#1@5_5@gEp`|#2Xs-7vgZ`c&?
zW_;P`Zb*W|=wY6lRIJs9$gvoTPrfI*du{Z8mUo^l-!r0T0@DpWNlrHY(?+zT1BjFq
z+0peIPwd9;e~=+xbB?LG%8>OgEBK=Pq?Oxc{pyen*1Vc&;D;<MraHQDaIGN?sA%J6
zbV`*!jBEi7slkuv;X~OROu@jT*er!1zpzm8mpguZ*97JHld?&Q32g)(Kavi8?OXkf
zVqLmx6qx?X&p1V%!Fa8nkI3IYaLr~Uy1BSnqIjEGC6o|ydfdTW>}rBiIVWKR%n(i&
z0B;?N*KN4Cq57UFcv$4Y_I^?ImEWQKPG77q#ynX-viBy3uq^{rsAmb_z%I^=_)r6m
zNx()%crIGn+Kkw@r3jA_;#;0`NZ7qjPIdtrCC;l#tPtXR8P0in^BV&6dn{b=OM~WQ
z-%Azf><LE5`4|5kLV+nTL`5ebD^GvS+r5_Rj^HpUQP?qYPyT5^(*&@5cQ|ZrlA8xF
zIK?d+ce`=(97IZ55|5VVgy$>1Nk}!KPbqM1V?k}AzcrL`eH~X{>{N#+S-{SVcwRHx
z1tj-KRreeo@$4HXN|?Kyt&T|>avOJg0CTwW$gUTUcof*^d)yEGEwptGxlgssI6mI-
zPd@gM+1Z!W8)ww3b&O=66357a#Vk|*ea*GtPsr;ni^vBdmwWSGGZ*rNiy7mZUWV@j
z=5?tUP^5&L2BJV?Qyi=4sjjRvK!ra)F?KS(gBKb4oMl<L1j2;OPbV;jIr{<lq)JNB
z_)k~z5I4^aC~j(E;<ORJS+MnmCIBWE8*pu2k!nv9bNP>ag$PGP(5#3Jg(p}s<(8DV
zpD<jWK;SV*1(>aAwi49w^gLXgirswMut&*iAQHEZbH4EjY$k4f;&T?H#7u+JLyuS0
z3{LY0x8{F52E@EN|AxlUax-Bw9{l91ea}|oBC5o$q$)>^J_=veSrnIHFVq4bc5^Y#
z{OEwH$%Y2ihv$6m94&gAWITm@R;Tq_rp*LkgS*~i;?1XOo)4Zw2$L8q&=0Vj8p00{
zaGU~a6gbvoR?PWBb%!wO*nKv>7<#W&BBMjZj2mKV0>Pd)1CViKkc?H)=wAbN2Z9`f
zHmuE+AwdpQ3p@3WzFD5zuf^jc5+CEzaNl`HWZ6j!p28^ss<!@6Ft(Veu+mP%Xd@N~
zQ384jT1)vW=F>dWR9H_%L**|KRW)q}5)z`oqcW|hr<I;Q`Ec*ZVgMS6wiI$785jI@
zW_n^;8mgak6rShb`~Uk3KkrG)DPRPP5@-q+zj-_Q6et0I_M85JsajEvO6Ga&9%6Fq
zSRZ$A7>KkGe>U5<oBo~fuZSRy!(e$Ka#m{R75>xJWp2fkW;jDTAbfoZR$n}zxt^2m
z&mUP-;p@l5F>9gmJTPTQRgrzyf8D^?ZhrXn1LQ|6yK3&*XiZR#FnKVO*Y=t}?bims
zT+$x%xAyuGms2DBatiE^|Ng(<5R9#=WT{H8NZ0y3&Ns?`N9ae3859SFd{<xk+eCh)
zmWT=p3VJ3ImzL^;TP*9o@jfqa-|)dri!KHz#@V*_fZ1sH{HJ`b?e}9jXTW5)0eao3
zo&a^<?X`*l4=5W3wtNtVjme)`rc|7;<5*%X_H6>U(o>)Zr1@-h+IZS?yZYQQkqyD^
zixhuA3**e48)MK>aIk`NUFm`~-*%`fdYGn1&g3T8WWwQ(&s{0V_d?wyt!C9-6i-<G
zd9|4e%sM{kuEXLBt)Vuh6YN0xz5n}~f=gPKrCyiawyxOjZj?+C()2#>4mv_J7-L3z
z9s}JrM=5o91_SLMRz>A8`0EFN?O&T~RK+r`iXUnJX2+}`kLJ&>2hv+U-(kIGc&+56
zR5^m;1y~A?4_%xlT(M5ggk&4=`rKUV7O!gLd^YclPvj^E&Ik~+GT|d#ocSn0kMs;P
z;1R@<>;L|?X&s!9OxV+!6bYEY4d;|pv=9C+vmP9dHafqkK=RFMYhUH*u2Ou?T}kDC
zN*=Q@jiD0FU7cXc>M+86vm)MGqai5gJg^9Q0<8(_8q%5!Bd~L-p9mb1azC`ID@7SL
zA3nq20rD7yw#sbsZ=#0s6W~g5IDoH;+sXA0-`D;2B_@5gm0$-hrn&S!0VZdxT_zGu
zfKaygp9ATAaTAmlZIyk%(2#d=vE^RGVzIZ-V?sA-=IKBJFf`MMpnJ&Ac3$X8WUuQ%
z@ORjrz0lYw0PONO)*{{!G2>suVYmOfe&MF@eaY`S)Qv5j`1AVtn^XnWA4<WpspL!+
zF69=%kFMA`TElzh%Cmi_m8mhLL0)3x$+{s#^DMpR`KAP|*brDry`q92`XmMM+G_D&
zoO&d&U$zVlGmLibWKrj~TUNJTmu`9$3xRg*Ta8gYz|W5A!)|-FL#yHXW{sKr6rIEC
z)J~#6f~n)leQWDq99ViFW`rs5NB?^|g_N*JTnFS+wpC8S1DL&|4G0E6!YW5*B><AH
z4CcWpbZbR~(|3m!p-{#9t@a{QuD=kOCFruK2o#>riJX0x`k;`MEaIU4Y(H>cTWDA<
zK6$D~TMEnKV#k4tLC2YaM<ZBWq3da~Rf)f-`tE2mwzYn2LB`qwIal)aKjrV|uc!gM
zX+`+&n|OvrynlFy15g$FhO4P*uGnqAZos$zpf<XP^wl{Dm=-j#c6Ic_pEs_&yS5#v
zt6mq%H0<d(13hM)Grp)w&{}oJf(|)YPWGqzptF=1pBx;7v0ROB_9V!ln*lyp6u&=~
z=G4Z5m})z%yU#_7IwRK6ko477ADcY-9nA1P=H~L`#n!-B3bBj^vTt!a8kURy7Chxk
zy9Ym!h5$LFf%hW~w7Ujerdk3EX3Vyr$1b4VZtpF>u2EBVY72V*@~DT^sKbpB?m9ka
z7{052KIdhh$v(3O=pp*sau#W2jA?G;?#IxcB%zmGYfN|Hd~=S;cTgSj&s$+IC3q&d
zabYI%SG0tD^F7ql?wjVH7GP1>kO9<dHD?4$X$l1NZM)Q_au6+Ra4b6zLv*n=<4NyV
zvcUVs6Q85F$k211e_&fABcAxcRNjJRPIrbB#wM^*li$aPN5Gm^)z%)iXqugU+9w}j
z?~DkCJh#l1B{$`;YU;{z2LqXZTj4fSGBYzXJZ^-17etfgvz<C+4#BdkjmN(^B2EC~
z2BaJj>KnOFd;>U!&Akzl?Ntp8(<<6=aM{85y%mz1&Bhu5H~@-;AUGm@u=u46Dkd<~
zvVi4Q4}?2}kRb_G7Mz+_Y{VL0B8%(kiYZO!gG#%sU83|D|5LcvD6T)QJN<akBo<8R
z#EsN{14c2r2B@%qTyYM+1s=2ln;A`X>K@e?UG7(7yOjA~O9;JydG2-oFzt=8J3Oe6
z=9y;AX>t(1`850(ZY~PO>8Gx4At1opQ{!_m&58v-jw#SwP}^w=3)`hbyZwkwI-e?h
zBnoHEpea7}VS2mVexNNlYzmT1M&OfMR3tBlVziam22UcU{jwogCf_0bI0+M!t^icA
zX|Uqi85F63c?=*0Q-44#gJ`{1tLI{Xdzc2$^Jf4#5M?<s+5$2w{;GkXl?U02yKrzb
zy8L8pUf(Z&8j?y$dKOX7!Uy@@Voz#2xVrWM?Wia`^&-efS#DWIEuDx8ITB?`d;P?$
zDMWiY#(MFNxVfveI@>Ce;H)!U$8RQq!5sU~c6@z00qdY@?p*_-RLMIbZO=uZTtjXj
z7R}1j9%Z}4w4)@qDQ?q^kag_eseSo154Dw*e`i@d{G|ebcl*x)_!T7NdH(-^Uv47l
zVP3KPjT*1Z{bGxMg6I~tE|*+Ws1GRx1g}a(`E8+YX%K(|uK-o=-;S!a--3me2<-S!
zprv}8M*<5WZ$l`rCcbj=88G6k(p?$Bqu{yW@i`_Y4L!X#gY66@u%>Q8#0JzE$}nPD
zZ~Au!f%Or3K2lr&E^~o_Lab+m4e&}qh^JFI7CB|$;W+oWYYS(Yus_ZyX8b7K;@ZS|
z`9h;E1xCQ0X=X3{vcYocXSRM*>jV)p)^~r^gt_kwf1QMbR#|=Mu-Ru?pLUs;UAFp_
z_o<s*u8r*Ok>Qdh`JJBIcKuIEVbP3f$wg>Zx_@u&-~cfS%_sZg8t5VE1JCAO9M5m3
zsG{Xj8?NEwdGcfkD_KALzQW4Z#%5WI<Q+nZEHF9gf@>J&OyjOtqcfZtdK_ivy?$L3
z0iOL2X3(3P!Nx#nS5E77JtDZzK%cDfV-<l}#C+BZ>RH6J8ev)@oKy%%n}Mmo65vfw
z5SBI99xL2S$lxh{r6<2cPqprC*B{ioA-c+4lTBlhG7DeqPLUaI?x-R)Y{kQ|Pt1H<
zTJ4?dSaG-IMWQcWX!~N~%9f%>p+A8$2dg44C%-aZTtll*Yv?DG`ey?k1iBV)FJ#d+
z_aP%0JwkI0S|LslSCZ3L`64G1(|`ZnqqNvp4sv;y>(dIB<pQ#7lk+j0gqu!tv!eYO
z2u~gwj#@w4{uLxb0-PtHKjA|%Y7m*;;9l#?Bal<whL-wpmeR50A5<k*!IV!L$wyH+
z&ID70DAw0COewJdD1eROCO|T8y?@t)K{7%Bz}s(-G&Q&lYhscILeFyku6ccwoN9Z4
zrum9gqQxA~tv$`p(ISSeL3|TT;7B>7bkS{F)rb-!RRRlyApqyd^^-GMg=lP0ogVDd
zwW9Y~gl*`cRa8EaYsH-dWEA2O)Q9>scSW?M0-x=_Ob?_CXb<4U9y+)Fd)Z+MroQ%?
zT7mM#F0IQv(5<b?wW@e<Y}s#t$v6DBbW2<xDq+Nz@jSzoO?FvF{fnI)<L2tvj$1Yl
zAK%B%H&c^z9>xIv9tI+!W>_N#j|&uN9<p}s+&FMAxV@75ewH_jU%w)p6Qb_fUA_xI
za?tkyCtx?*`E9)5toAz)C;-|x5KreJXnGnb;)@W)It@qp0#XG5#_Ynuzl0c0oH@U`
zHr{omvb5xITtk84B9ef?^w_|0RK05LNg*@f*t`X8A~?~<0+=OmAZnwdeN``Gt)YSL
z9$R6R&CakK*WW)Yx&p+3=Kp;%)IUB$v<!c~m*h=?k#p6|9_+bJ=>iI`i9pKyWOgeL
z_RUcactYPeCKwfQ;l}xKBjwyQRy9}IM?H+@^T5ZS@?N$PzEpBGmIMR=1drtIYI@Eu
zEV%Ti?w8MpXa2xuW<1TGTL0A8IJv74?*;^cImD0=j#nt-=#)cVP>Y8ng(=*9qGlAs
znf`ypuB1I9LNrNX?9Su~S5z#aT(E?a4F(55Mrwl`ZiIY+SU)3n-*(tGV|IKYl(z^i
zFflQ)+V5A!{L)f6bkxYl1+d@$4sE|<w6Vtw$zpGMH=u}ld8IDAFTu|vWy#kBw{h<9
zP+rP0SL(Zm(i|Am_k5m^Oq3g~s$=i(@sM)FioHf1{Ao=3k<)gxXi<O{X11-<C`F;U
zJ25O(G#M5<86e?;WyY;KM6=<CZ;wXBlPO-RzyCYC**t5+wfDMHOt@iDzDT5)gb#Vz
zw749*8)ICN=B^#!1d1_k9MD|-50H3UI$M?gfR((2@(Fqg74Bv!mfQ7ew(<H@brAiO
zSGd*Q9Vzr1@WwQf=2J$x1Gv9aq47@a?^g$FX9r{jLEW&3P{LvU@SKAz01$sT%A4U6
z5RPQ?0e#}3ygUhT3w(BdJ`fTXzPz9-I_bd$41yGVSeg(dVlTY9M?C`{gzN*_q4fVf
z`EP>7G<LHr9?Ljpm%4KJE<0lD&o8F&igOo*yG#OgEl1kiydgt~Z~vS6=QS_K!dt!#
zy`Q5K`N!4u?ppt7TWpgVYP*m>%q@1yVlE2D@!t<N6MUUd8d#+-?Z=OjZ#E95tRW-z
zCh{zHvCLJU_HG6Tf|L|+D`<W}AHW7Wx(Rw(!Wj2-E@hLvmHUXlVcVtNu<&<fr@#4?
zx?g#`E0TMW^!3>%dpexYc3Wy}UiOcVE8e{uh_`A5@}Xxa%$sn<W*?LQ<vYMO{Y9{c
z7T9EK;>u+CUC<`#el{fpb2CAhAvp9$V1VPs0W|{#X3~P-qm4wqjC;&QOM=!wa;t+Y
z&3)37r6I{N2RPzZus63W)|YnKba9$0aebVde;p`use#nmV2yX>z$ufA-hi|RG}FCC
zw_^$!RlE?5J@?35#2FD{Y`QJ*M>qDqB37C^viW1V`TY4S=kjaMsuC&^{v{px-c;Os
zjy4loEACO<YN@X<{!HjJpW3=Uda!K@RY1@NrP>g8^H{UTJX3aAAA#TQZ~ue7{Jwr?
z#S%O5#RvqLDHxVg3PnH)ki^9z@<{ObsE$ccM|kdtR1RF1Q^8vWfdE{f7Zm`jaKLK(
zZ$e`Qab@e``uEwSyegJlvrM2?O-fRlcNgazFe4P!z?Qz&1TK|pU3i}OTESy_*8J?t
zKiF-lcPbi-oUX9pJXW7v(Ts+w&QY^cS?gax{!-owxhc^zu?_Jbb3<+o&CJsTZ~(*M
z>{7PwvsjFJI$<VgBEmQVDP%CHRCGQd&wDra@E}Av5-;;6zs^SHq88X;bn-8Evp7nB
z4^3)Iw7km0lT}#weOF_499RNvS~?KTbYC9<uVI202O@w3yOM?^zz?oxqyE0ScyGkE
zpvX|VkQz~s7gx_xfY^&g(3~7zwFtb@D*)@9g+{_b7GJ$s1|#AK$ZS(NZq#ZZ^7-4-
zgyT9Vj$?cpG8}{!1unt0M9d^=qOUBv#8knV64Sr!rbdZLkH8d}VCgA_g5MgvtrTNR
zR$CJ;D}9C|vr9(E6qiEs{j&10xhv+TH#gAJPqHRmMZ&;u+w|!k22XH4ZCO0vrLtT^
zAU%-ZFR=MDmPPa3GL3pfDGjDMRPSmDv#DEgdK>JFcRWk1Md?Fi^Uv_|T6vy$aX0OV
zo*$btce|=GQ8;2=`M$>uoGvbdugDU_5M_f~yf9?VA&oMkeT8vrUl0l90m!34OCtc0
z)*oQMBh)`TyUpX1Gc1C^Q0{_#Co|xwFfx+>*!gKm!`@SX#t`@m*$T>>8dHtFhfpQ=
zLdR+(-N2GvU!R6FE64{?>>#5-WV6)&+3E(vj!xJ5sGEL&Jhry#8!U~7bRRS}W#N`H
zDbg)pwE;<WmaqRLQDb?py!W0(hdAbEOddhDK9#MjwT4TA$Y|vg7W)QxKpa7p#JAwM
zjeo3#C3GituVUnnA<ZFfr!oFlS#?`ku5qf?_;aVdBSBaVHSMQz`FS0fBl5>|m|M{}
zT+KvJKIh0{`gL!w(I5P$Xr;4Xy=x?GX?ngwcx~K^VUWh)HPv+a+abj?musS{K0-|P
z4OWk1hQ`NNk7g{RC~s5;0JpsHcmh<T@or$8I}OPTi}2xQpeS`)(+B&K2av!(5GWH6
z8cGV$b3ict=s49F{xF#ZhX^r&t+fbE20Ux=wb(i2q(w0ErF3_<lSIo8Prima!6g6L
zO5Ud}M8FRMm755rSek}|<tYhFWHYKvtYcP-39~QP%Om-!s+<bbI?uEWFkN?y{b<T&
zQD=R4X-p*b>SI2>FZrszW*VPMe~w+UD%JvC^mFoO@BinbDpns-YJT7R9;HsCNTN(e
z{RzEXG(N;RLhsSzN;B{t%y%}jhz_WqMdB~yaM99EKgEhk9u)4juqEh!L&`-jvL@1d
z7q=?znsPi&FS?VQBE!?MafetJtJ@y3F8SS#EopNby{~e#YG80&_8ytK0(}o|BuT7V
z8kmfnIWIY)fj7_WYmfEY;X?cxZZzzuBLsk?@8F|j2%-e!v+P$}hPY>3Nq<g<)C542
znSqjW2ErkJy=HPg13V`np?n)?>?=MD(%NNX9<VaSiX5z2mQ&EkF4y0*vk&PXHV`ZY
z=K%d!oagnCAaWu?D%&E=sr;ZyDy3}w)T<d^hHHTepN&#~mrsQxT7Z7Dco#gpn|)D&
zr*2b%=l`y$%g_uEz1ZrBPIWdi=19}U;WUSaU-w>rK!~fMp7|14OuR!E^71U;9#Y%1
z<e}uDtg1$52WzMkJH<2WW4W0rel~qhyz5K4-CX8A>hgO^Y<kGwX_8E3EBTw_h7)>r
z=|R{Tmf54_4jx!G&?sg$ovb-RE^8&Si;x=_0fDZX9Pc3+Zg9V2-W8E~dV2ns`U2eu
z(vhqJ;tV!r)iJNNxgNyq)bFjg)b7SH-H81Wbo&Fx`d(QKS}4z`-A!Rr>~RIP!ME}l
zbnvIbZZ(5_s-R&T7Z=yhwV11egX0s<o27qW3g^YaKr=Fc3pxhaXkKhuMj!K=NH7=2
zA0LWg->A>aK%LyDSxF5dW$LqSgjYbfUwiNo_GZJXxD0<>tt%~1ayX`r2K?DjH1-!h
z;K{2Vi_r4SywNG8$eX2qVm(R#)x&qAy5J_?zZ6XM+dTBf_0b$l<6X(AB3`|7+Y4Sg
z?+!v_FCuHkp>~-Yt5&8DD?$2rv=4X(j38tSS<;J+gN2kruY5ZYo==k`7&|Mz73$pm
z+T(@r!QiZVyM*~NY73=JR`rNM*xT1iuCJ1-7DT=5&}A;CH6$h{<N<^L@xnw;_*%Vo
zq1@<0@n{A}4JtmT9mh;j7Z=0mG^IOs-Hs-gYFO04+PJj%>g+|u9Njr_86suqs}YR5
zNK^Pj{y&YOuK|t(v30$P=H3wxF+H=ZMHx;U&F@Vj!>6Fr;zqU8ImBf|XShNwV{hLr
z+yx!H;`%h%kJ?au(j6m(3#wQP@#s<6yK_EXny*n{?EP(lJd51*i{%~TO@V&~Ct^#{
zlM@<=z`;Kn?a)Q?jexQ(l5p!T!h%IU$TjW(QJ9`Ek@bZ0M7PZRG$?<<ZK8nv0b=e^
z2<}Ey@M17hLUZ64QbZlkBk?*g783>(N#$5LB&c4eA^y(cH(IL^x1>GLjF>;36vxng
z)&0q-GQY&tA~e2X_cDu|%|IKii``i6zmFO8mZ%~cQ2r-ulan8fDx+ft0tcCf{D+c!
zuO`-BWH+7G7&;98cVnIG&zol#MRDh79o`(3!m0A0k02S@JP^6n?&!|(p^BcCbAt_z
z87~#B^50Jlwb1da1^1$mhddy0Kq%s~e$<?$^S53fFALyB#Iv8mrOGj5N9%N%nYFXd
zk53LW+Iji;ulz^DXT+)~-LIXExQ0lSfn)Fj&!blY7ZcK~f)f;i4I)!3#CyNYdEqf|
z-4UZdaGgVtF(BoEOI>bmZWbKd2p#4>lRpSz`{T3zuL%qTuj@8L3*Xrm-B4gz3N{g5
zW9t(uuYExE`X%I}%#NTX*0U<z(a}9puC8QzZL~awj~1+%#%_7d=+VXn9BtW(_+spR
zKj$=(@Z{LV4M$YIF}pgdSItMHOln^g%s?0S)zAudT>^*Zt_AIzQ-;KU2EhpFfK<D>
zhqQ#aQic<ISMuO>_;?9o(6!@iMKCHpaKXST?ViVIn;`b$?9H+N*4F15jtAX;zS&W@
zP2`AZYB+vUdv9%*oECOzhk-;Wn1P|xz{C-Mizox@wyK20J_dccfHY8ueGJDH{5b{&
z<eqMe<vbgVJ?G%yz|GGeTVF2;X?`}qSB4aW|5^@jBsyBq_h`phr-B2x$XpM+>N++f
zLw74qv27lA)$u*0ppiRG+rals<^Ssp%AqqTi-y}u>cRXs2Z^YwjRdTPc7}gkD=e6W
zx|Z6aKQ>Pm-}BhH2lQ+~lzNa9;z>E1mCx6F8MQ_GjO<m9os<&6Z;;GASJ5(%o3}Bo
z#a};vhPe0?c%8qRd@RQsXrFPo32edBxfkmws%m~HwJ;};`1Fq@WGyuG;XhPZbGdB^
zE&Q3imzsL;!J%kL05%93gFb`gjcy{*ggH4mtz@@2pydR1G0*kut?T;;6$LCH3$0VR
zLBPBKeCjn<zIZ}JS`V)1b2d|e@CWW`cC31&MuL8~7wai+++Q72^n5gTo6CA4U(JAs
z>dV_Bnc(cPzUt_`nnxv<i-|9pUZE8maSdm86TaC7J_n#u0CQ8K#yRl^L2gh2x2{UV
zm!=<}$`bH24I}FTf$(4|C{mWh<%r7A+%43=e+;n%fMd*g$jn6{;FZHApO29*WynkP
zQh$xkOVa+h5L<lEy3GtX3!DR3l}Fkh2Qy)#U`&a`s2V*G;Jyk4-y6u(L(WS)k2134
zJ&edKjEArc^GvSQe8aNKi$O)LAmjNpr-1FZa~DG~$V<drj-#xTT8u0=y_PN#k)%YR
ziLa>nXnZfb7k@^Mx7Y2^P^-uz)i|?``c;-mHTA1VIdhX~6jluCUvqw~hOlR9mkiJg
zQ0b0dhsAq)p3t8@4w4YO*l0)D4l=+9!UGbWi|<{!fLo!tV+!d~a4|b7%!u@n<qz@T
z4a2(3J8wr863o6A_;s>t4sJ%w0#AeSpSbMmq334@ZJZ?NAdBBKC%ghsvvIe`5vWp*
z)iJYC*ij56et3JbP3sH~d6n|IL)-P{q_XViJev3EbiQcEYko#EMW4q+rts!qI^rn5
zaa5fX7gZVFDBBns8!;^C3ET_8^~&Q^@nq=^qRabnu*k%9<+$sp;}qmRa6aRlp?`@$
z0Vu!I-(i7r8GPbZ$X&T#=3Ug+Ip0?MLl3g!=cS(e+g=!u0>WO9b`AA(V8gml+lTSR
zmZDIh5BXYIc7a)V(>IDgTmQWa3tBZ67~lR4m-;1X>8V|fI4wk92Jlb^F!JWbBLp%A
zu~-+V7X__)*`c~(k@UC;9bEb+U#7w0?Ap8-Srg+}e7~DvK_wtUr64scred@6Ui>vz
zImKOrj^hGD$=<Rxp`-ww-BT~vWTTZ_v-$q}g}vqWMb`M?2j>6wr^PqLIi{ZfIFg(s
z9MKAb-$$2|PdIG_dQPE-auiS{2-IO+15O!TM)Hq-$vY6V)q=gkYZ*-Dg|4=iH8rGW
zbe>9LsrsTC{!06OAt^vm-I|Xi%2&J1o}75~)g7PsAvSYR13gCWHn3sI0>H?X!R+ZU
zkou9I3rL#b11H=6L^<2N8G9%1{D^bi3<B0}@(C4VI-VaD&fPBse;`Ozg4o^Plo*2;
z#4G3q7i2I#9-MPH6t7Ubm%W=Enj?;Oh#BM+8zVFK-@JJ(&z2a}b1|osNR@#b{ooJ#
zLDGLUvU6De9V(w+AjB5Ii+D?(Hs9{dAeuI&y$UX0*q0qb({;rzmwp9acT70*8Kz`s
zxg$F+{I^Pbu$WN86RGlh!4(v`(Y7K@9gk|v3u~|ZCAu?GJ+?i%3=X}dRAo)H#!_$<
zvwP0R)x)R;mg>NJkQPD6Uq2q`{NTZ{Gmf6#r80?2#=1^DB+*z?kB_eKb%yw;e_E>$
zv7T7soAmqqcG8^NWUgIjP5ZlYx_9@9Yj3C;>9E;aM-21UxO{QjKEX5=-J(RoCfMXY
zfGdeTH>7{s0lgY-n1sJDr+sh^B%UMI2R@5j@Pn+$rZbE0%?r|w5?13}V=i@sw7-hJ
z<p3^MNM@3AYROl)gL!3bZ#mS_{@%?EerQh!%s!rdCXe&)!AAuAhpPR=%f^rZ66B@b
z>NDS_<Ke~*4Gum79B$;_Cz&TtvX^$zc%mNU#sp0#hr{|<wMX~&!QNSecEs9;p~#Fa
z+Zwk^QwqR>FQKl?NBQjg21ItLI*z%g?ZL^Mx2B;t#Bgpn6;8%4&!(%uZ0j2BC~<YS
zD`b%ih~b?8PbqWFFH^X|Af-YVhnQV3y?jV=lb**V{qo)Q+?8WESH3)jZ*}dBj<uzn
z!Dc^W)7NQGr`T?418o>Zrjhc-pEB|<r`q7;|4>~$0^4pE=4o~@l3Nc!SM2jPHPxtH
zAP2FO5CrQ!h-Lj(9TQJ`KbrL$43idZrb<}tbO$t5obXY&aFX`$TZ*DnxE`xDhj6fj
z#-12~>}C5p?ssN_a$vT6qAl89e`sUooh{WV@xJQHq>eXuJ}hDUybmo%#l!8nP!hx{
zgnL8-%n9Xkt^BB}$cH9aJ+HhF^QA*>JA+hts1v3kv^swJ?A}|E8hva*BPa8cxB}gu
zuMo#_2x~L8@ggb#;Ip>*V(JiFW!S%p$6X4Iav#39epdYSc7$cAk>p!_lhb8Rd&4!)
zGnsu$o+bl7L08g+)H%?Ptiv%w-~-nNLUsQB=2p9FVpJ(0Nw=UvP&xK9tP9Ow^rG<f
z{0S4(!w6OUf|x*08CvIaG{l=#eH#gv$Lc~H889gdfk%}&%62sbMe|BB&%wCHVW`cM
zbGW~z?m<f1M$Mxp!xW%)P}8!^ipycF`NkRv&OA&127c75b<JYUk$di_6MgK{KON9?
zPyl}DR0sN!OB+bym2vTlc_Lccmhw`UY-VH1ad4=PzfP=g7`E}t>c7KBCpjT6yOIt2
z@Cg7@xai(1yA3=voV$VC!y?}P`C%}1ILhm|igcJGX0Z#vq`k~QQ4<5vMx*6cE+bFE
zHbKJ&Ex#J#h8tcrRqHZ*gK)^rf9D;OIHcA&K1d|82AWc8JKp1k-1Rz0J7s?&fr`44
zP8uQf;C2S6#LP~U_slO03EeQ#<0wRv_wQET|FDZL(sBlhx2?088&R|B^j+Fba7kL$
zkHn~$$s3p{#vnPYWyAbe2ufad>=nE0$__yrZ#uzK*4Nb0J35hmj`-g>a89_Hy5~Tk
z4D0;S-;Yog-n;YyP}R$b+P*M+^*4A<af`b0jSmlsPWKg%1Qma^sD0(d|68k4sLVI>
z3%gMBv3Mom6s{)>^J#!uAS)cnVxIK5mCSwqjr}Gfwu9+22s(MOYJ!h}FYLXxa3VV=
zzsttcagDr>Y;BNH4E50!XI^5))_8dnh7#}xFowf^&HxBhGk4P2<vm<3FNCR-(%qsp
z)7G3*^Ysa_Magh3y*mkP0xzCUxp{)=bEYZy>u@lUE)jB#`AU&`Xvl$A$Z)qkv;i-a
z;G8<hwWWIY34GN0IX$%HFmMzbyzE#H<OGAcOOJmywCjO@gs9jBnudf_R0V&cr6!h0
z7=Ewd#Xre{!Bu)y@^_>Gyw;ukbz*bldWC~j-WH@x3INPmT)*{LyV!79JZ6~$<gL}e
z-bLO9Yg+q;1#-6uR~>lm^EOjiymqN<JU!RKA?)s0q*LaTNGFHrIKbV0NocHDYd>Sn
z@m}jJ+(1SK^SYE6nq7e^PK`>D!EyC-_O`Ycy;99npQP3p!)&Uv_gtI;*uBZ21UXWu
z66<;)4l0iuWg+KaGZOafTE88_dhInvv8jOH3&=)lUXH&>6lC|Ptkj5j85cZp-=VU?
zS5`~fodeZ*2mXJAWDy-#|2v{pu=M$Xrd(YVs_T&@dvr79y_3BX`<qN<jeb(Mbo^Eh
z50Gi&>(?-zbRGs5tLl~f0wkaev{Y7x?E+#5k}!wNEOSyFoH;I=aRB9Qe!){Bd9Uk;
z|FGJU&zCad9iuMJ|0AuVm5afA1_dUrZB;!Y+65AiGcQB7<-=Ivw#Dyp?Pr#2f8;Du
zeHjZ~bQ87v^soQ7-L+ZEB*SatM1KIKTNjthp^1~EJ>=M+QJof#6=?g=c{cF(a7Duz
zkGpdB>!`;lH36FmRI=}o&95c*Xoc&DRpwxS@H}{P64PG2Ac;Sc#OlSb?t`PQ3<O0H
zI?%;^`R^GhB)1fi5~T(#1nW}Uu9?<elzG^Y<nYzox2DrOIcu$IDg#2DkOwdc0l_`c
z3jW6uf+^Ixcuc1N)Ww`gW_wBFQ9hoBrJ%Yx<=$DT!R@~#I=)I~ru2u6G5%YqDgsii
zHIKT0hm0j=3SkoxQNKTeSn{Dk?yxmag}CNa^~LJa#*YIN>Dcnhr)bjtXV}+<hNj2b
z*CvC)&s?a_DC&_FY1?5%sT2MY3RT)C1KW%$BdcDg(dxy7O1UoM8tr*jf_fg*6|jEq
zl=~n((<^9I-5YNX!vVz5eOd}jNGC*4tsz%0v>?=$5?;&bASx_OR3M{6Bv$&vlQ#r+
z`P$}ROYMMqpT;rGwZjq63l6IbE6HZlz@YL1l5-Ez^(^qVbG?+X0BrNaWflJ}U`;<(
ziZ*JVNJq4Sz|sG?{kJ>Swt4B(Nq^qY$b`qa13h|8F=$>eA0K-_q9#H^UPLiCs8gvq
zbP9E*K<!TKPNkC-qzqkkE(kY-qvsmRS!kwUw87*w(%USfvP`0Idl!AScAaQhXy@*5
z-$x0)de)7#H^X1CxT@vL=`jwk`5NhYhBy=5?A&WeHv4Jtolu|v39bxp?ut$`7OA@)
zPAC8OrXM7?pf@64E2$e+(+z3yxE2#zieHqJ=<Xi1?Cf6JuB8ed#MP{Th`7nGwDF%%
z=LYorh}jJc&W@kKxU4$HtZWcys<%PLatK~$zr-K60XGDh^+4~Za!e-NDlv5JDnAbq
zs)utzy?3Ydqw@&R><i27asCT1*(rV*UT%tmeK10a5rB|qkLU80yn}SQLkC0oyQz%)
zj!kXoV;0QMJECv6u*wdV`*yU4-gfv3<GXTQa`WrW8qGJco-oBp*O#FBlNn|dOLB)c
zE;Ay$is>D9{5?eBt)_(q>jS#@gl{*dMVl3AOX*(}$<3dc<}<a!>Zwm(*7E)78qUxa
zBJx}(q65+)p;5|Kn87R_kqK8nop^ke*YNa6MGJE<IfQ@j7vf~->$J{%UD_n7fw>0Q
zH#1?+HtBc)ITDfeiOmGld%)v4P5Ppk>&<ml4UMmWH})d>b(qsHiz{m*7PF%u8Uo&6
zZ`*6A8qS?;M{@P1t{DbvwuJQSf<7yc8CUb`u_B4Vqc7qZ+OplWcc}srQi~f+K^NHW
z_4E|1Odi|c`;58BypU(-onQC8QES)Za>65wXEQ63R8ZFp1`ONsinMR2ea4Sc07w))
zx~jV9G^IX$w%HqeAxPR4g(jflPD1#V@AKCQ&w|S$s-0*(vZJy1MfD|=uiU)x^o^FL
zIv2HKe%CE<g?r?j9L^58^#a_-t~}$I$`Nayy|rC+Jq&5CJI74}5AgwZik~H}j44cD
zN}#E&UF-L&L0dtA5^<A7^o$6_1-L6o_=nstdVo%Lk6J7k31pVJFfFb;JOZ=pQAk9C
zXCi+>DnjyhLjjjrcf&<rloCCOa#l;e!DT;s2Lh0L4IT}a3PR3pioRf&=$Q+eBn2zO
zL%tiHzQOn84RC-3vE0|X*PKx7@?oSg7%L#z=K4yLCXPdH-gEk5v%W8@nb&1n77D|Z
z;A}Ud4EnfF^7CxPoH`h~xVUBpz|K1M_R_gmlVyq<X*utDOyH&>!}k-S^d8H_5G)dZ
zg407*W&YO~W(_PtWa8(EV|6z{n#*JIfer8Qh6PYs+Pp8yxqBUUb>^M?^OaQH^x|%M
zZ?KVJN`ez$w1%`QKF7GJ#lxk{s+P1-dvIq@gKI-ERU{f@EI8M+LKm=1@^VNxvoYC*
zS;&3&of}<OwR<03<!TjYHS#z)`YF^!m8Pk~fGrBgaLdwG*F7lLBgl2oGiWHaK)b&B
zg^tHfS(XS0V(-DBe%F8c5p9iv+)*8M!Wop{eYxM4e706hk^RzEnn~Pw{VDrhkG(ap
zpO40qC5QOkJqUKg93ouy+ge8Ce}-IR7`ll4NUzhx5gUa)BBN#DLH#4mAiWQ10kU_e
z>0q%bX)zJJOuy8}e%Ne=dTJCKC6ICWZj%1Y(1k)0R->@*q{d^Hbr_PijzaklP3L&~
z+9!aPNE2I>q~weSTP;M`=;*hd>a$-r1yQChPdrKIw15$X4Ac<`0*n(s;cx5NbiBIA
z?)>nn(eh%k!TqT}bQikL=o7~Aed`zjA&ihRbaQS8*T3OMWIwbIDt*DEVmc;hN8~Q(
zp3BX%BPR($v$@-ZMV+%N|F+%S75t!++KK?OIgUqbW1iT3*(+(bYOXq+S$_+x{|j8*
zmOS&AS#NV3D9W9Om1875Y<~ge9oK+9o`SOIQjb<M49?Z<*}!w`kAd9%8tMcJl==mf
zH20s{eNC)qKX}R<K^(BnRt4;uM6y3da>aea`zdYARpJZ|xDvikfihHG6q*HwVc!|E
zZznHa+dDihJao%@jVU8m3<zKLkD9;t4-U2>>_X5xb%9FF6FLV_skw1<e|U70dH(P5
z`Vunt0ve~~5brvZ<&GOd^7!>5J7fRfXG^L~u#hlG-bLH!T&J3PA1rZ2E|Z1rFK-^q
z&lkeI>SNlYE7Y!oO<t-6o52^%QvBMYuWU9|_G0r<bK8+pE`~E7RL8xX!I^%?W=W4U
zg~`X4#2>Kk5<6TT2fs4f&CBAZ;^7^Bs4eEt<hPSo(>l$e^xC+lek}2mEbWvJ59$`;
z#d`L(#><DC)5^7}w_l9{4#}nI$}yAF+3+Q=`BTwa!ZpLNN2IiyFFW#rDKS{$h9_&)
zyE_^GtS(kjRH6M8t6v{lBN@D-ACFn8^1GL%%K7L}!F_kavckk383b#7Np622Y$wGj
z3aBbfS%Cq>9R$t)dH+97V2F4Lna`g0-BAN~2*Y;4t3r0e*C9)ZTv=>^VbWuc3hPuh
z2{W*iAAkFbjJ~%)@3k$iB$;5ntuA37e+16Z4CuP^C`f4Dtb_|G_&2s~?e|YP7B%6l
z{A+Lk<Ms1x-@dmmcA9h;Kd!<PP)Cvdz9LDyA2giM!y+#i8s6~@wMBT7DdqlN#u?GJ
zln-`jvp;J>=eG=+4QUT$<a&{y^S--Ky!lMino3~xBy9=rfDh57p`zgA#9veq<W(xh
zyr?GnxQoTao<GIE?%B6UNB-lkg4p{#7p)24zpzRvIk^2D58EJVnd(cT42~A~Eq`K>
z&4w3^UpnSBtRM2i-YviDv7(x`l=<yJIjIz<%j!t!YW;R^F=y>^X#N(+Gs6%#A4C_3
zv-xL4BUP$M*OM-hj3E3yj(!Qi1D|wFOH&0O7FcqbKY)VZ8!gT6?;?6?d61E%CFB9;
z%+$o=D88UMhsag9%E9}y-QaVyRTA1B!0I{j3tFuGCD60ju_hPU2ImrwjA8WTAa%|Q
zKRa#~m>Q^6WXE!T(rat<G%d>w6eNvYpSGJ;X}(dMc?pbwRdwcaEdJbRmJTS9pO0;5
zh%N)N>OMhOU2lAVXTGhD$r6ng6#*od#98_VDg<6eIoE|NKIh<JL<uU&-5^e)`#VwF
zfLFBH<K9rgUhIFQg%9|L+;hI<Cbzdj!!Nm&b%SCG0zpK2{p{YCsC23)mZ;hqk@b+u
zdN865t=`|ed!KQC*XKLLqN}T^%^YQ?G6&0EM7XPn$R*~^ku}ojS@vZp@h~`Z^mxFi
z-vfF8Bz_L5ei0LO{6jGda5uI1l<xM}I;uou=tsNB4+l$oY|fXTO);ub(W=j*i}(Zu
znCzEf$hweH3eb)J7ZCtXQyoYC%sX75Imllw7%Eti{`1avMteC&N>IDGhqMYTKy**f
zRG!Hs91idBML8#hM?G4&EB~P>T+t?-WAnz$)3vo`wUHwTET?}RzDkg`di%cg^gr?{
zH9rlx-tO1hA4_e0>&wZj5_&Dm0SM^)lwU&N?~?vC5H-y{%|Yu~Qd#%?0ycY$KJF>)
zn;WQqeUqN;58RThj=lO=9@X_TqWLFX9j7BoquE4mRGS8sDhY76?L2*GZ9+~%6nKj0
zD#L)n1wx`=TScmg!2qWcT!FU1ApU#WmXCkhUdGR;4_c#f6@A|ve*hR_HF>Mx^{`8i
zue3CQ8{s+$_Z?}Qy;)*@rns-0R9{Pmya@-sk2V{%rI@gqW#pYUlvlG=bGhyI??Dq+
z_unNW^zVpoX!Wh5Ks~)Yr`$)AA#Th|bw7){G1;Bbw8jnk|Ew+9>fF~C=`#;bnH;n*
z-3EFJVn9{{ktbpci5kz?Ptv6lme<x6>`abXP5vEW55MvIsjV-3*Sk@-Dro3F>o7ah
z2;$;ZQ>U->w7g(cPr>!^7df@;NQ?*Ltvj8zwyUb1-#1U=c&ikPPM#D7uZ6dN<M?T#
zRq-(S|M2vdaZPyd|1hB-h@_PCNI~hAP#9f;bV@5QS~>+`AR#R+Ez%t_njtM9jnoLG
zQyT8G&-eGg9}Ib^@ZiLCu6Re{Ri^(b?9cus4)&M0(ZC1F(Odn4DEsy{%z;hBC+g>1
z`|r=*^Vu*s6YXP+mX@hOop^UZKcY0GM=3w4-Ehe(_hHB5_Tu)46Db2Z%|wI#ORC%B
zT-gq>gSD<I_!-XrwWy9DJ0H>*<4VLWZ}>QH3xQ!h>veyh5_(Jcj8-h+tI2v7^7Acr
ztIBcB-%(ivvE|^SCBt0lSgwdWN6W*Pa;wf|o0m*KRT3X&4@f+1b>Kz%QJjUrw=;Ix
zqvxJ;<4UnZWKRM}9?BJ~>7d1#A&?tvBcJ;t&0kEEO*Igd;_3I~oCnWtjH;?~Ql&li
z3P*UVzWx-<*?0!t?{h#JFxwl~WozHx=LKS5l<3c%@8V@6K(jP~y4h=XJPV1(p7C;T
z?J)%!koe#FZ|b%61D@U{B0~%<i`2CA{yy8gC^OHUFL}R44hts~wUUf27r#C=PZX;s
zr(xrCD3K04*(miGq-J&CQrG&VUh&;&WGT-Gx39tb&5B1dGepjEFpwD0RIqArSAw~9
zl^1!8^&9FWJ}roO6io5j8=>)B^Px32$yA`J-rl=kU+@~4S|6e|FP$)?DdMtxcpN!-
zkx#G|Fk2#;MhQoijxnEF6G9-sXhsinvE<ZnGYZ?7*i7nmlo-Yp#TlKm$o{_eu)YxT
z3FAfON<zeXfayJ2qdiLVumg^LjT2KSf;i6Ha!@a5>X+}+aCib-isRbJWqd5NeJ|qM
z1^YyhUDR95VT`Oh8MwwTC<ebLd<`?!Or%yeuYK0lk&8=<A-;+DIrAP<CX`s9{1+?_
z#Y^mU?9+qyaW09=R=XyG|Ami_&)H-{YY&<t4qPvhXpsJ^><ew~#wD!Da!F(9JBDUJ
z)a?w+c7LD&B&|o8dp{Q|$_~Qc*9?n)1jK4Lt=?&g>AOfds62VSkW=idX`9$*)1j~=
zW_A5(U0>}o%NTztq3qsYhOu}oiRw;)dMOQUb43?aVTOmo%>q?h{i<tS?Pht|0n$_y
z0;kff<s6ax<6dS6N?!H#I3?pnsw1P^)BqrV9DQOs-$5#(*mB29WHdZ&9e+_&)n1v1
z2D*BAfcwx)cdu@cQ5R#Xml7k2kFK4$o$w3YG-cpZQh!CwAgNsHEF694v3XLyfeowk
z!?;Qlvpuqm<+FCPO$V4%*D|vsY&en!BZqID?=^GW>)*=9o)!0wu|uGi{%I=}0G|QB
zDKBHUa~BVur4F0)$u7<1$P@V1n6#|LnB$J+3lHFUQhNH{_P&;8tHH-}^<eCt0D6W$
zR@ZlIf0vsg(ty9p=*)}(IN^msjQ1LEZ!VkAA22`@TY)p9C9cs2T@19rL_u8$bgI6d
z@^3r|2?;@aiKBTXt9or0<e;1I8j<2Y?E=&;)a*P?{Sm(BBPK-7+HPA^fAueu^F?~O
z#7c^_EbVp#`VYSKPlRa9(A$}(L&lVdw@<|{ajt`F-`bXjmKDNmNg!j6QRh5MIwz(d
zb2%b&wY#X?NSt*LQ+$~CR_`KiZK7r{ukg=U_uWWlO!n?hc2d<m^5Wa~Szmus#sfla
zKhttFv@S<s%9lyt^IALWiD|ca=Xbf)Rv(_+1^AC^<Y#HjjY;39@4A^D<+=3>(zVqs
zuz%&~{}aXOh%0p@X>guC_np`yXuIQc!X{#ACI}OHhg!j;+Mae_*@5=Y>R&FlNy7ic
zasiF?TVrNkZC%|#Qib1#3cocjT9<ATiM7WX%V<yfK8*l=4ZjsKgI42o3igli|5O)1
z*0OO02zNZ_N&6NezY9&UaYe7&*#X@e)6HR)6W?lz{!}f-$i=6YKJQM&*YZMrsj`Cc
zzq79c7kFY+rwxlua}F0sMp761^tbrQFLk8Ybs8pBrqllXl8ti?Pi*n1s41dKZZRYp
zO|7rVKE>v-b<}3yjLhY5i*3^4@_y)S@7F<9Lw?4*Uv##04dFMg=%KdM#hStxVNa72
z)4MgPBtMfyQ0AygL&lt<nyAknxSd`aq|#mCoNf5<Sl}EmEtP;}r7f&mxc~B@wfnb7
zqwm!(eQ=+><=AGt=V3m`_uLXeDM@oLf+g4}2mjkj<5jsmZiVD}2A-tet5@UsV{iMn
zbIQu1(P-TN3a1~;v8z!6!%L+#!z<`Nfhrsd2wEef+*%Kp)jd5$fbiZ*uk;OH+gT5f
z7g+e|w!lC{%DZXw*wYoLvP3s)l<AsS5}3_gGsb%CF4bBqEPr_6VHTjVRp0#p&vGO$
z^W~arl5MRgzoNROZsf|SecV&?3|bBN2m8qj-mL6<c_US2RTU8ne~6{0cZmoa9_&pU
z?WlE+uFtJ&=%$Xse|T69Qbv3uCYJ+&7Wq2LTii|Lfro>i_Y1!{5u^i4TJYJued0>Y
zX}%c)uOuo<+9hbNL82`g0;%!s^u8<nOoMLC<%|&$$ct>jPO?$H{&UBXBJTB=LfSLG
zgg<+$qbH_6Xytk@IQj{SY*qjb8ww3;JK5S@Pcf}AxCr+oJ}D|`q+^HhSBuo(KTsP2
zOG+6Mtnl$!75+RL1D*|l3nx<6Ys4gXFz4!Fmf(Z7p+t*+5AyY*m1|sP-xj$58yuIF
z2zFOsLB$_<bLGS1DgSePTvkJasO@5+0=N+4o`D|N4<L-J(D|~LLM{1!T8~t;wE4!|
z8k_nl#;$nfH2y!&UoUVeNmcG&4x0!L8GUAh^`^Z_A8C#e>YZ#Bwsv!{m);Atk7ISq
zMaH*Ge$=@XLvq2!jmEeF8ev{Wm*uU#3=EKOM@oY()=yn^-j+B;4sk9h9V4rTn{^(_
zY1)w9>hIm9mBDW(Lk+lXKa1k~4He+<#Jr4vKG(adg9_f@=bpYZu3p^u7!fbR6=dBW
z2h?kqFFLLv`+Yt<GGeP&_njYW#v5xA3N;L`+FkiDLR$IOL)v#vOb2KS{_ou4Tki1g
ze_;49u;;bsI6n4Eo|q6*WfSpd=K0$9$-L1O!O9B3AG^y^>V}mLG9#V<Sv9p|ObNh0
z@vJKQ5=0|Mwbf081cZc#K#JrDe6W{93Ad&yHu4kHR{*hf8T~<j|1~eh0@zqS#J9Ye
z$ewukQBJvl6iw4aD@s1|-2E?v!eFrNxrT~`fpoVP@RnJfY)-W7z?@J(eLF#XQb$jR
zK}|S$P5}Ay75#c}E8>a01N~IXw03KZmXk!5SbZIX{kL6^#D%0<8odjoZVHt9Q-9b)
z>zx6eo;+*U)QPhvc#TuAy|aCLl;?{WUI^MALfA#j+VvS(Q!qikNAodY-qbJHYl5|Z
zh+#SSuKm>}B8Iv%{>;?|LCHUG{vPiB%a^1-M-D1uIe0(FH-dA3eC_2yEaWf*M1-(I
zv8A&1SsxxN!ZE|WyWpV|M^%fMD>wz3FK>F_y<a{^;^rhqaM3Y7M1ECqbiIE}aOxar
zb{X2U1B4FIFti?vgHM@@Xo;5xBJ-d6JHQO8xw*L!!0^vNuyj(xn2O^in5nYFede*f
z#4de~M?6Gm0<=-kUVcScKAaj}on&agIiRs(-51Xa$Z0@TGTDdc`pdq5wpMX>dpkzp
z#d!8K!dC=b4Akw--tC|5ZL8DW-^V8>6TYc#X@Ch3-*)-EEf2({Da7tOu<Ae(Xb3=I
z?mNF!`m?&R^)H8>FcHUKj5qd|AwF{;mJ!o0(L=EqgL;0MV_hTcrl=BaDT&rJS_9Wu
zA)S~csm>^k!@?2MJb0s_7UiqBvm{QPA-2<?<EE&t=fzI17u!L87co|!G-ApHD;YFh
zAZo$VH{AOG-aL+Q5w;85G3j5vbp07J@O4hka3;u@byO4mndB^Q9_^z5_zitY1e}1z
zask`#+k$;+$QzyN+(Bn~@&+Z7>f>Wdmr45Ps_cZlTXW^8F5-f2tEgrx+~=VTyK{p_
zQ(59U8{+6zFMvy;pL?RdCIo;pub)%|N`9o=2pp~eUKK#-BNOOQClUEAVYdMAj3mpJ
zCbF=tEgMZXL4N?^kSqwe*;N7j9NN<Zn2U4(Z(=kt2e?JOzP>m?x6n%s>mpfIlJigM
zPF`;eXA%%8paeu<v)zCYN&V;%+I3|f0FelYh-!RK9f2^=<~2Wqhs2)u2=)>%y*SF3
z7DYSd0TzeErj$-X1C}U=IGOHA;rcT_3C!8V#}%eekfbo;?th6*g4^)kKE=MY-l&Ip
z{2rb47s4g+25_ryN-2^uhiTf9Dm^>Ozr_#5TqAyOGA&1nwE*cYIAf<&MNLIfZ37qM
zxA__dzS)A6ul)`$pgvK2PIskV?bZFa9g-so;|2Y{4i_QS#S0%_Zid6>ZKJ#&?<e_G
z$fX`RV3=}1#M|uMZ?Mm3a`6MLH{X<<Vm*|j&P-d;g|gh-;p(_?tmt@0$XHFySA;o)
z3JB|>#gw_@nfR*=Lr}2F<@E|o0<9zqE`9iF#?9aCn+Wt#q;YF&yPj)X125q^7{!UR
zY-tf@wYExu#CICdz~PsmIRpqRFVuI-!G@tzIY6JzdHDq>)RJ@Q0`|KyKq@|`3QJUG
z#s7eXW+(z#=Q}T+gZDIR1FZxKKn&Ehv{oZol4q@GCICmu0O0+G1FZ^12nbX4j=z-^
z6mZR{#sKpiuv_nSePX=#XJ@DERHcZxpfUR><4x_Pm9$_rP>lx-g~MZyf-iJsoK3GG
zW@ne>CJ&uElr!0K1+FdLcInSSTvj|nb)%yTsh`$8a#9ld`y<@wQZY$E_q%mje(tW9
z)k?#KpK(=|gc#KDlS8#av{&gb;uam7BW0@k2nL6zp2q|@Azg$9EA{=FCBrQf=k+xy
zw-~1Q8G8SBW!RCQvG&V-Dr8e%Jq-2)(MtK6B(CC-r8+d>Q~&WOo-r3h61=ghz*^&y
zV#x^o5aj*eLx9vtnm^?+JJP$kVBUdC#8lE*X6f&J7y<3+3_h)fp%2~;kG9vJzx_n6
zMpjaS02%Q)pvFj)ZMXuhg=m#oHVHsDw*|bm7J%B$o_zQCl}534t~Mt~t^4X>bIAEX
z&?m30ueXTGG3G{j(*vDK%h6o9#_|CWmq0RIy#JKq<Mv`3!^qKCq)nMnmWeSK%K+k^
z*z?5DEP@nMpSM1xjQ8Zp>;Cy;z2qhCTo2|>ez>kloikMAFg_%vW?0Op_MxoogsjD*
z$Mq)Jyc9KKR~NRF)X%(e*sU$^qA+|+cW$8_22-$GvMV;mmf?6&DMm%RU~ISdhiD+{
zI$k%o2DU+UKv1cNefmkmkPF11{+tz_;Djy2J!|y?5|SCsJvk16uu}|nED?}e_h;nY
zG3J37fgZrjd$`{m)lpkJ$k+#HWlG}7t-}Ng8`U=BL&|n8Gntc}+V9s}in>U|>QpzZ
z^eX@6ej-@y^d-d1!*RpgUL*CVn<T(3XZ<^B0!UiI3qB{`xw*47p1@)OrwvgA0zszU
z`K>ZD)7X10NmTt|HDFM0<j0)MZ%Cy8DF`M=s5zXku&^bGQ$WA1a%W_I91%`NVPz?z
zmh|IWrvA`KLv({ScoY${)7p;kR>jr~uhP`oBqO!816TXF$UNOvDT{)LGRlsuSR+)*
zOa4yXF}mvBUzJq_KaUy`|9QaZz1-)5>g?;z>vd#1IBFh|c(+p>RRj$~UxK%i@G<hU
z{8HwLX*hq8C`L;ghz&M`F87yJg2>y)w-EH!!^ET$%(sOL_ftPyk?-^V1gC0{#1Qm2
z<TT}bsdv=BcUB(Hg<q&zc2RsX2?9rXM2I$scj;yUK_-Q~IN3r^;|gKXXaW~d{^|&&
z_XUC&Ma9Y>MUqCR8pRs^oEX^H*zWGCEt}V9Ymd#63Y_+%Tr(bXDtUp<G8L*Y=6Vr9
z!F{h&sfC+ABl-)xM7SM_aHde+(3fK$mC9Az`AB~!uL>Us7f-W$MvNCtx<Jg+w(Ye{
z#MY~1R#fWMSW2AOn&@wNERO$$eajV+oZ|hSzhO+$NuH1;D=gGt@c<`k>ulxjvx}Rn
zKlQEB(&7Xj3h&hWmHm?Y=T{kmaKRA*S;bl<s-T_k#59z@$g4~@nKKElNZlCs)mmP&
z@cmRIT=zyj$m{0J7o;6xP<sUB7f_LjqJFz9xk-O=4jiG!-*tT5xNtvPjViZ!_0y;?
ziwn)P_%F%j4Yco1k<dF|;ND*Sy5l{bB--|J9dv(>YRl9#6J-a`ZXck9fYbhOXO5bR
z>hnGjorXc)lqk*F&GClq?RjS)LQG2EIIt10y2cl$4?^wM4OkI$h9OP~s99GbUn#Ae
z$4N31;-(%~alJLM<kBq3Ie(+!P(t>8fjeYr+J}+;#s6sga<=nfcWpuk)h3ByBSYBr
z+^H=>zXLRGhhwsDs+gbnp7S*`D-Cx`4REa5BR?~L8GHlHzcY`OxP_oVug-oul>o!p
z<$vL8o(Zz}^%Rnr8Ua7Xo+4(werV0#q6U@H*^?V|l?n>@Wby6iH!Hc+<tUYTHmA}H
zqtoM5WLRc9_i@=|2*fv}Rru&p`nlSMx=+hWX!Zs;6`A#JqNpeVwv#hZu1^3fD#zK$
zB5=b|7h4q?6`r1+sAe)6H?%O*;kK<b4IE@8uJy^a<Rp)xu!EhQoSqFu51o<nS}ZtB
zXgjdTvyvz2KQ@s6q0G#-kR`z(=26|I#9+_QZ2#NLJ7Vo?s7syes?Y=TmZz#iHA++L
zIy`JCEahHVm0DC&kNE7zra1i(rP{#T-Fd9&$#UUJ7WQT2vs3)hKcW4Vd-s%#?vn{j
zEavs+ynxc*xhY@RdRdscIO&TCViI2k*5o9}Lo;&jd#Qm6b<hD|hMcJ@n?UMn^YoSx
zV%RIC1np+OSaL1~SrK;ae?JyA+(_1d(Xor2^|<l-y)`KrpH0NsOjy(|%;4A0?yO+-
zn+o64&}`JKeMLb|j=AAv;2>XueV!o@yyx-ixL;@ui_E1JM~`0sM``T%ACSI#>yR=1
zIWDF3xF93o;%}(ABJrI(<+*%Lv8IRf^8ZdekDfa{Q=v&K4gJO{Sb{hF$;<bX<3N7F
zQ?-&QNKvOyQf(b)GVEN5%{?K~ic81dHK|WqaTVX3X76>S6PXipo77fD^vT6^-(uRT
zU1m+BHQtp;hx0+l;Z3<CUKiGA8|$|Sy-QVlrQZ*e8fwx4ad*+@<*g;LmnK*hj!26r
z1B{kDuyX4UD~RJFMoP2@X)L^bA6HhP7m99=EY5Svm&ENZi8q0lENHFZI->&h9!K)a
zQoov?b#p}YdYqAlKD*cDOFUz`*E7PFOV~HM6;DTS$GV>AMD@%YEjW^O*3GVdeql0Y
zy))YvSa}XY_Ic=5GPztT^}KV}1E3%QK_D!8@FROdobdI5Fg%a{UlmLaijO*L4rMd-
z=2Y+Pr}?LQS}mlH)N*9nZLlV^A1S|j{EV6JvH6>!OxC0<3z|t@=-6ec|B8zHGjZM3
zAl7R;hLJ4G^a6`M$u{jt2>X%O&PMHT7-xRpM;n&3bNi30`R@%%-KgS*%i4opd}5q7
zTv)tb+E4y1xoSH%35jbs(EJv=au=*w(%E7p4-R0jJngUDzlB;E{kLvd83dqR21yj_
z3Wu)Fp@Wq#aoq2Axi_$C{(uEG$gH|xPj}L|S@Eg(I)-6NDTlyk7j|HMS$J0C+x^{X
z+;Mz*pIGtxBr5|gs#paHTe8KiS3eU2d6qX)Ad3$)=z(G-FnY^6c0n&8AYC*+He1Yj
z4Gi=^biX_AI`C8iMLqHwREsC3@wv3FXkC4`>`UVy#VoAo=xDLHXQnq76}Q09W^(B=
zKuyfZ&4}uKuo?ZdyzSf1IP;rs)?{YX@vzteTdGhoOhW_PjKONTc(F?6)4uyFs}YBn
zI{m98GGZxXWvZLH_onAPa(Dh2n;7lSM$AS2;?c2o3}C1-!<H2km~z4JIFYVAI2bm&
zX$qfRsn0c=sb6f`VSefT>={X%=@QQCuiz{dCNf}Um!ASUfF=fXb)<dmxh-NUchDyQ
z_X>u-pHii>tI|_ok)t-$3}YqATJXI}7e0Po``^X<wf^N>JXpQU2Mk=exIXotz14Os
z!yOVyqLyLlX@+NNQ|e*ZPMsPqK=)4lygKEnQ5cH$SA&Y0=|QBP>iI=sLY@88ArQ^}
z1iFza|9QUdx@DuDEuIqs*&E!7fRmCfgoK<}zVW*W9b>QM56^`YJU}XD?AphQ`z`|w
z-$0jo@OxcdUEoyKeS|OJVt#zI{^}9c>YOjVrCbiR(q?pt4y0&oyuQ)=LENK7N(C)7
zDr9fhx7UO<n@ha4j(UamHU8Z`3s<O`y7lq8QOmL3najB`;Y5-Rs~=O)4fO%#`NDU7
zVn#-qd$b9Iy+Se6@ymzcRk(7AU3^Bxd$>pfV(sMX-*=K4+~8gdJNJJC5^>a|l(m~F
zPuy$G<xut~j$6dqz#^Ie%UD`K=hX^Vy;7rGt7J=V(7+DR0wsz7DN&;VR7x8p7uHZK
zj7AF4m!Rov@xMI7PJo-t7GZ9Q(Gp(&@AG=$4&(KSRPanFW+xwi!~J<2@yKh5pAAl(
zBW~wwDMz8~66SVOjs%&RuZ}v_(GoC4prQGyx%KY~u)<shoOvN9KNJmGmlO!J=T=q4
zfd<jA<f*X~Na7CB0=uC1UbCR0lPKHimUTo++Y3@&vM{DKS87WFz)4$ZjI9Eb+6ACC
zTLUT(92^|GN#>~x*!?BJ{rd30PFciGsY3>X$^<WYOnm-JOKNHzN|o|zQIZ_CQq@Ak
zcisQ0zOnf1!2Sg^T^6?nO4S*$TGQxBZ%n?g%gSGvmg1?(clvCW&{xLHQsTp=_dTN|
zOnoX&P*ugKDbH8KWn+Q=HIZ=n)dYpELxSc%9i?aC#Bt9?Q)S(PnU;J={+6H0fTPSw
z5?9uKxBV6!9ZD_lc@@HICqT}-Ef2*BZo_ERMlX4!RBkq(S=p%!A?}I`_Xv(M$bO&5
zp|f8cw@Q-!KSR24ppzK3&q)01lEsEbX2-LbCR*iBN~8Z41H0c9y6ZbCN2&HV4fG&X
z%*+-OX-i&m>}PC9*eXwn&|USIUEM`f(E<H;{_CZQ&))*S^8vFgA$~(mwtj$M8wMy8
zoS?l+_`Mgzj1Xg&9&%&we^BDXpiAn)-TVg<*<<08F&bb0&;JEg<=3}N()PZeV`10K
zTWJ*XV`JrV^~{A$iFV1QLpJ?}D@n?KmF3>+aYCIwv--M4%tuC|NP01RrMPvvU3U5U
zY>S{PclH(g(sen)$#sVQ^8M)hJ?}i9zc8cz_q|RrE#)<?xtabUJ&)%<24ZRVh?@uI
zx8fxat376gi&e_S7VAmXeh-^MSM$*Qw{ZT)rX=iNK_AtI1;Wn=5y7mu6M149%g?nR
zL?gGl;KKt=r9mf$<X#HsSaQz;gAxvBt0H289QrQFz1bRv{;#|)9tv%C>N)ZcetM_l
z`nd)|!$O#EGQ;Qv?}MYJq6UAh5jUV3qW-t}dvh}wTzsxZnE;#s&|Ps*o)S%xxc;lL
zk-m6^wi$RdbyJk#iZ6XJ8jF?*+(fv`r3#w@g}`ypc95wj1kDMw!_!gLX{F!!VOp$N
zqcc-Bc>TOaKbnSac!Lh0E_wY6;$JO0xR|1McQ>blH<I4E;E%S<HR4PYTd>qZ*A<sJ
za~NM32S+_s>B4V%RP=5q&9WrEkv}?JWs$8HFTU1^%86M8soP}rcQ%s6Y3(2WGS?*E
zPTHUTrX~*=J*`oiB73z%A1iy7!2nAos@<c@aC-WM)*?wNPPUc$_=6vFc3<-|eOoZk
zVnO>el{k<c0v;)JSVo4x{U}6fylT-ezaW!y3H%D=(VFWNqz9%;-p~lgwf$Ak_9n=o
zB7)LY6aq2BHPzaCHz@P}f`_Av{Yb;+?UG3Um`E-4l>eQ$0RQhNOH_y_)X9Rwwgb;`
zPt{}=vcOq=mDZdo9Vqod#=Jkq6Hq;9wd*!>R?!s-dM5hT!9n6{?^A>7P_zV22JlWi
zk-96F>Y8<!SOn4q5sa6dNrwWq{@FP>Oz88w+j69iMr0dL4N2ah#gY`HFGwo9hwP~y
zJ<4rppnN@;TK}(y2+Q!dcVf8DIUqJC3)K_fNv-Wkj!zy`jsI$)WDqgs$ZmUtg0cR#
zDTSqW(TVj=CVyvBx@7A!VUm^=q4n+?YE8n=k&uDvNmW=`OqXjBa?y+AuBWyTsIPi*
z)+V5+iN|&9_NITHfBb%`OpMW}O<CT@Zc;MAoE}*r%v}F=)8ilM?e_znApG{3o?hsx
zKgbzC>D)owrWqi-k4J@)u21TUBle2ZP2n7mfcs&KGE5tS@DZL7i=7hj<Hm<))ma|e
zdT(AAtT}%zr-7yoFD`5)iZ2Epiko%NfUA46y~scw$1NjfczgNAk80(#Q^A|dCkr*E
z^Jpms5qp&1W6yRhQNDH9N!j`dE#m#EdKXA}jxS2rF-N1Fs?hpEs(<;|2dHf`I;L9x
z2^kt0u^*oS&#t4o3jb&o9)uwIt4Myz#=nK>XrUr1dU`f^K3(62H#AhMIFv68O-COI
z_*ZiUhOKBtBG!9Y=rJrBTL69)z#XWas=X@Y`**<Ez&hwg<298hjViyM6Z7tH?c^;!
zRHOf7BJ>xBY%zu_q{=R?xIPfZf5|avxnj$gbeYQ0o#sg6?O@rYtVBe0_mx@;pWPFR
zExvy@Y@}UHA(3@?ee`x~CTayz<Ov8VFm|cS!yu&Ub^j+_>C(G|n!^pUIXvl2;#u&r
zi<DPc;&j6x!HeQ+ARUs;Iu{^Kyv4d{1~vIZxzu|Qeu9=ssw1IXWH{L-g3Vff36qB;
z?iK|6#HLud#G5Fw^f$XZb?88}Lx9;3@gop~eC1HM-xw7cssAT?nWuT-)MY28LAe}t
zsfwXmQGD{y?!~4+&*koSM%K<}r9Rm|9GC+bS~j0e=G=R@egGyT2|h7s{gKnwQ^fWH
z!gp-z1duH{Iv)E!C~w~61WHEowzeCI^q#rsPN^SIo3jNB<OK2Igsfv0sl&L~{T0f#
z#)YDMj#r&%nEr0_RuBiw-g02FqvJ(n4+PCY1ww~XBmTQ?kCT)Kv!&6k?wjjNw@Zeo
z+6Qu?7B@2Vdv7;TMz`JHkQ7*W0@P{wvO|h9>Hb=5{2Au_Bf<yKe|MSJIE3myN75hz
z8Y4nWl|JwZ#n@B_S5zqpv5|f)ZYw)!>GRim`weBc25p4AUP)4Pr`zI=S1bM@<U%^t
z9h<O9^hM-}uWwj=-6<I>tmGO$Arh810nu2MvZjEb$1ZG3<Lv+~LSMj^jb+Nt91sL%
zk%QdORRAm-4?e|kf?RlhboJsVkkEj(nt^U!I1ywF7Il6<7wCQxsQw=0vt~QsbxzP_
z9+7~qjw3$@EJ=^H1zL{&H)_yI54kv2T1yu**E6ed8T1dWZdzdHPp7gUe)>EoCj!<b
znc$NJ+M%kc60}rQuWzo-UtNfontV^TmE1CT<ue$8?p}fo+N*N>-TYW=0Zn>x%gS4Z
zSUc|LQX=4%F=Am_#fj-KBk$lpvh=@y|8{qG|2a3O(qz02hBNM-o>Lh{Mq=Z_1hVhv
zPcWF@hYX~6N+||5{m7M(ZRHkKwCR_xaZuy<{M;Ff*X4mrv*Zj9amI6JnGU?JGkf=l
ze_hScIoDGazh_dqwb6gXuuCscaWx)!DpVssqgg%p-Gs|sVm7mm-u3s@*Wx$NvP!D!
zmSkwlezfWph8XlumdB`<mWLjA89)AD?Ihb&<5@E`W5jk5n{AiW8{tk@(NgB;cS2Xk
z?LEm1!Q97R3ITI<t)y7ug1hg*OABV1oWyat@RqxxxMymJsk~nwu-gkbVM#v3BZzCy
zC7I`oJej_Jk1}na<$-Q0#t071CL)<SMwgYVpTI^-R;3GJf@s{`yVv=OqfsFz+lpEQ
z@;>U&^mr`H+u<*4BXE{}{ahhKe0k%1tL?^K3cR{*K*vj|tl=dGTAuZBARQWt*1G?J
zNA=u=C`_kB_h9k%dJ%2<^vY)*a5FN^t~d3-lK(on0&Nh9Wju`~&<mG=f@eVaF3ehH
zQvD~Jng1e#UNuFPbU);+7X?Y@OO8u=`$v@fnNLrTCWcO3Jy5O`AJiHxoTFv2!*Eb9
z+jBZZ1(%R!bjGM9Dvr;8n$xf5g$^vEJ|?hOMay=GK8)(%BSca;=Y~28itF+c26ZUh
z!^;~ijZj!ZGW6lGut4`&x|#E7aYo<a?V~D8sa_`Y{@(q-RPfVQ^!RZYy(!gsf?ym?
zh1gR)VFMEZ=n2wbMMQ;)MmAAYow<W8%hH~hc~mUziGN1R?&zzY)4OBj1e6T4UT?;k
zWz#*$BW`v+JsB#8CP!BMYAmA->t?v|t>Lqw+~0fiZbJ2W5P($%{x4B21l9yZjnh;|
zU`W%hd19L2s^~@Rl?>nh3c^|!4N21+l^f2!Q@cI5eb!^!n5Nti>2Khlsz~A(9$9i@
zwc?gmoch>s{J*$X*%-YQk6>l+zeYSLVz>+<NXZJm`M^MliXbG9!hX@-@J~wWvCDT*
z)h`?uiX7B8AYfYdz%1qI)|AHeUQ<bVb0jb?h<R#MuX<k98?x2xw_|UHdaL0(o5<h`
zirSVflf9LSo6U+%SI0<xa)}MGk!7GhR9I0#w0IgE_YPU-I2B*}n!1%YYIC`)x3Jih
z<cjgN6t__`(SE90pjBP!2;VM*Hdr1tW@Y@y<*h6D@`nHJy4Z&|-%(?4Xz^oIut)W9
zLtF@AA_^){E8RMsF_zABl)a*&>e@yb{}|I7k)p}D5np?ty&;OzIQKB$E5Ipn9`+Pb
zojsJz*81}*J_vjvNT@itUJ-bhNWDKm0MCIdP~n?v1rjLUvtUUo=~Fch>0qg-sw%f6
z4mT#8WFZtuu2U+j#&oXEWGPP(9#vf`cU2e(!<N3f7H$jt$)2WXr!o8?L}gZ2{dZZZ
z!esNRSQOlEnyLoxjI*+GVM{EE`VJT%ErMDC<ZFdAB<Bp4gPHBO*f$h!icixL-=aQ3
zpOU_MwxOiS28V;UiP~}(Ln<U7$g2L(dg%d#U%&?O7$pOx=S61R_rn#fSlCJp{{IW?
zo-#OGX>mxt`F4}}nrrXkhx{w&xgH|JEq|Gs4^<0sgmU`Tuc;U3u8B|S0+}=nTWmT<
zPG+ZClhJq<w45bqM3ZIh-=TPFD7aS91<-70e}^qL$J>p-8(YJCSSOd+xv~@(fVcsi
z=)@ZOmOe44W+}Vkruf+Y2XC%_tJt{oEbSKqZh&?K?VT4_LS_lHirRYfvCG4fu>uti
zqHk~Pj0qm7C&s|Q#q7p2LX9mo3R0!WTfT8N#zz(`vGf@w^iQq&yk->C#F`LY>iPX6
zMl^1)jy~aPMF6=ZPUx7S9Nk@!(*fld5hM<K&yIP-(#>@zk+q$uOZ$(H3x2~BOzVU3
zZrEe^9mmmA4xG^)t)*9l*!?*S1#Ya}DZ~sNA~IA1#!C4xaUQ+rs>TtjoSDq89<jSe
zvyNd~;5|L?5NTGiTHLR%H^GjD(LhR+>cj4Ad}VX8$1_~xm4gNV;2c9EwrDHZOoEvg
ziBO+PkYYcS6VrzO;N6c!e%2txo1e<T^5o{b*P;u(DFOeQXb1nH5ye-8b$jQEd{#=p
zXd16Y85&>=-aj%~=s$2fac;=H{1H<EUH>(EJ5&of#r%(iwm0qFC>kV`wIwUREU*h1
z5$6ULPxw%NpMib(F4&ISa23px`K~K8+ZN3F{H{<88z6V?J!-+DLn&$ajMX9t(Jjq0
zpxQ41>#gVkqjV+5mILHl{}@omFMv$N1z@RZ5073Y_WhSpq7l|$<k)G2W2ah%Z?%%|
z@Okh}60dTMGK)<J<CpEbiT*^3M7b(UC|srhV&XeYtqc*y@LAk{$rmmvOewXr5nH07
zRP*}D@_hy&D5%9`gnf)j?^3r;A2NtforQHg%N#ety3A$el1wZeZrIz6H?c~UdQ6@m
zM5+QfpO(%Mq2*L>c)3Y0HA$E*vLjZ@p_mhbO<csJD|e+>NAmQ!Up+~o;&?EwM{o1R
z*n;s1-@`<dxOFxJla)C!Lvcx-R78C(tx-x$eKO3Yu3CvA%Wew#52-Zw>50j9nBsE<
zUhR2<g}Dg@E7#v=v3|Lvlj?tFd#Q!q(KPsdJ<gPc|6oRbu7StnicI)QhLd<=@sN>Z
zqQG>*T<Pj)uMGJayrfQMcLz&On-i4|Fq{+M5~g4a``^HSVC>^q4yFXhVgtCDnzEY2
zQb@bB-TT0Q%n<s~MZ_Y;Z7}{o=%f$NbIrtARXZf?LjD`WcF$*N(l7;-s4ZWs@M~)V
zEIfn8*J6urd_NOq`Unlsq_g{#7D=5HgLu!dq~_!^{zAGQjC*aRj#9@xvH;#)3hk#1
z`umyn8(&w^FRAxQiJ8A*))trxt%C;l8ZOX~XL=?rm-+;>7HAtl%Mm_PU#z0?J*Ldi
za<ORp+^d!0{tTn1Bzz6{h7#?BmUT~AF5ee*$*qgPVjnW2A1YGY7)93gL|D12*3(sA
z(vqpncuA{z{N~eq=gRpObxCL6rcY=|p#H>_DZb0V-}Om=d3SSetY*4{biAMge~GXm
z11Yj1Tj4Q>CO<bTZP;BLEUJC;b}}4EbI~BJd-UAa%!#UsefLcL4l-A{MuqkEY43rZ
zRF=3D{-zx>#rBV9*UQI|gEaYiXAN7DCfsvphZ&@2u&}ZyR~kWBiS`TVMKC3mWdPwk
z;b6sS+B&b?(h%e@;(zH58zW=~>a-FN#1VYNdHi5-Mjp?LXuskh{V<ZBi#SfYcSnE+
zdja<e_`yV4>AN?E(gggc#Bom4mN!IaPrl&MTzkgm4!&VyaV%@L!ic=}_4GX|)4L{F
zc3-lpb$z3n>se<nq-OKPndU9S_PSZaRTh!Na-@>v2F>D`e7)1+FyNjByo|m4RV3Au
zkbSlK*7s@h6N645;ME&0b)56omoOb&FOZQ=CoG%n`J_Y|S?i+hsL(yoYNtghti{-k
zEaW?+OQf!hQb~+TmO}BiCa2t6&5<JxU#Ws>jhWL|2kDbsz0w!eSSwJ-d(^spv`KHo
zzZ$IyHcv_QPWflr^Wm(HD3rLziHk&@@WNDdLPLIMvi>yklEvQLf;FA2CpAOujuS|%
zE}8zO1#x^Z=GNp)d7RRp7|E)<2I;Yga4ils<6KVkPwU4-lC6IrWO@(^9g}!|Zv|Qr
z`_f$|3~V1s7+uaE{2r`_COO;pT2)w_Gg+QKfuo2SMfH_do{NX%te`?-9e@AyybWV<
zv4_iF=~}jQ*7LtBp<Rdn*oflu#LUL!{&woK&UlK^5cz}o#1sw+YgJ~7MkgWU71_Sl
zQgl1O%%$+NwPBhvK>*g87+vL&Vr2^<{ph@CMsXDdug%QpF&7{Z=~9sM{FQ~yM)6Hy
z@!;Nk=rS@L7C;I{DmJ0xv^7hD^1Mu1^GVA5%~Y&#hReJ>HTOx*iU&T)`jA<-g2-)f
zoAj~;#luJJ?`mWHR++&l|1j&c#dgqts?4YtS85*<SZuAol^y`1)@PsKb3h?^=szgH
zzm4KmFved7cznLw^XyyiAK6Blgvo0IDQ?#_*$*IbwRiiiV=aoIRocTWu$itSiK|Ae
znE_V&n)AB8aBW&_q>EOc9lLD;Ckk3GUL`RU|4lv3tL$NYgw+r|gS6fg*UVbdnW(Rl
z>hvi6>IoN%)bt-lr3MW#b=>^sTb~70Xzs87@p(9?MxQX0dN*Qy$tQpXredwU#$R9&
zr&SCse)w_Dm3o8=l_cr6s36l3r&t;0(&e(+iAu8!rl6&G6k;pWkJB%|Ppj01gCjv1
z$(@KVQ~a9aJJXuHY!k)`6)kQ)?KZcEyY!^@l;roTg%=EOSQ6?dq}~b6vd76M)QMy3
zy?_c+TXJKXV*6oSH>_&&Vq8h?Ym(q?BB%f#AXU}a+~+&rI;#Cadr2@T61lDKd=GBG
zHk;&o)NlM8>JuR)aj3NPyWK24x=54o!S|rRJFSe600d>8s`SBL?8qa_U6wR|UlqiZ
zwt`{s^?&Ub=G4l@I;{pm;>5sPXARWdqU}%1WAv}p5E>5UjP2FP^V*Xgln~KP5O^uC
z^JC9<J=@aJ`qLT!B6bC0CPRVNW6xVn1-c$!`!WKWgf>9a(Nl5w2&gu$0&5#9VCXge
zkCw-~Pvxv3@UlVHRcgZPddqACxFJ2j!F?wY#_#4ro1)jYE>vpL5CePW!7nJGI&(zL
zO+o#Lo6nf@)srOj_uNSb*FWy&pR{{K^DQ+G-y2OU;Sy%c)>YeO4{SqC_cD^#H(ip7
zNyy;L3H24qU_6`jslGtfCX12Rs_(XZV-3vPnVDrQ_eDlH$yZYZGZJw4nkH60P2xz(
zbB9$?Db0Z$ZrbiynPk3q<yG$NmtlphWI1w6pa_3s#PraZK#g>0BAVVF&hXwTGzW`-
zLMDucpzFa)JKjWmc|Vb3e`dOZ<rQSkgYTYwyM!abI9xXN_YI31r!Q$s;S)Y<T^XX@
z`!1G)bkJMO>-RntZz<exHGAgD{}*S4NdkYa(X@+1@%3U?+%Y96{dzh}=!OHAnm*YB
zkmJ1`j(%;RYZ62WFl;I;B-!CHIq8o7NOg1q4avg8*&)?LUU7Wu0_|;XUc;HDQISS+
zsd1{}E-F`r-yRMyA-7ow7?6Al??-Pdsj;cQKc)_P8Or)p<@_*4`Ur^{MwRv?oUWi#
z9WO9=>i~0yKETItrVW?LX$QkIuXTa!zBHlmNrZ@q2%0^nn%?NY#%0zAy5|S7KI%zm
zQmb*`^$EJ^UcMzx$O=lBaZvy@KwHkCkyLNqz@<Ski?(5rk!F+>XgW-@Z(L6W(mbBe
z=;%xcQU<)Tvf18-SaHLMNh7-<f+Z03jZSePE0~rV!2ybYc`NFxu&@qya_BqN<j}hD
z@9JzCneGg>t0*<oA8bmCe*Ufn&|x!;mAV<JbFup8;@F~P$slE2b9H}=-rnMedMqiq
zur*TCj5mZ`jHKmNU+OmS6-2t`uOoi;kRtTw2xLOqu^7^#h^d&I8F_I+s1}JI$_6?)
z>78=V{NmF$tj#YNAR3VEl4V8SuD>Ygt_wuzm{!bJaJqDb-z<p|5XT8Gt&5%<lf1rI
z7aEk@#kt9bUp4=i*QV~kf7mQNMSk|g7-1uA)UdT%P|!s7GVY5lWuru<$*t9yA?ZQs
zQG9!M7O-0G0n$M8I+YPzisd33qLsGYkvJ0zznY|Sqj#3vH_hey6n_7I1?WWTuR!PS
zVk3;O-7{F@N;t#xn@9M86eH3P6nb{vK=;%Yl)hFGf9ZQ1q-TM8y^#nv3PQUU4LlVn
zdazPI=ZFG|^VzOBXsgcs2n|21sj|SiK~SFbi}5&4l^S&2H{HiVcizyFlAxYR{N4ue
zjA)o48giRoQ32q$k`8a}(iFy9yak1Y_?%1yJ{nQhSe+}EH~%vrNW%8mn;H7)Ofqsx
z!bF}M!zYUDaG^EsRkF6Xo)wPEBOZc(FoX#6xGVh7RCQb^v%Ei1OCd&rn4~p&HnH_t
zo9z`7EbqeRmFAE`=`qX{D(jKu!d0DgGgSUD^d2!X*N8=V?xK}}6*f@7@z&@`!NiXO
z{psWZdef$;J>LV%Jn0%1YnliHqj~u@H5Q5*^C$M>P)9#Ah(|PQ(0xon#p84icg828
za6ZHX?5=0wr^nklzF(qHjmUT3Sl<+^ltd%;yy9!p%5~%!A#DQ43~XZsZ8}dM#FSvj
zjEo?|PWlBVE{Ms+tm+p2bZ5<cHe&m<OE$Sp+iCE0dY@no42FksF+M!<08|aBCh=<G
zTW1KG!mPkZp7b@T$ymf{z+Rl>gDA5e1j_n9E7u4h+;3=4Ju#hP`tl#nu0Y(TdT*!u
zT_0O{gXP+Yv{$Rw%_7DKRGM6vt=t89ZRTUr6QJ&7GVnX4<=etQmhi&A`i27_>lp~<
z2_XA91P%<5=SLe;m%p^K0NVYiCi}W{Vc@fC43OG4JxZH$25Ayd;GN|=Z3F7fjfeeQ
ze>XS3osJQ4JxWj0Z*?UvnF0{i348q}p9d-_WnY>KK1{Nz=z|-bwTJRH<s_ivnn@e>
zf9HUiHsm8D`qrfCsX)!9h04BGoYlMdUU6N?Dps?ySqpam0ItV0(pjqqgrb^P!pRky
z{V60PCXX{v)Zdv&o4qkJGt8=;Alle?<L)pj`3IdZs8+n`O=t47xfSYu6vPXS*F$Hx
zUU7SSnNlyF;<2jUhs*N6<tmPf&f+<&Gglw8C~!*n1>=rR-=-W{en1x<fA;+n_u{rT
z@O1b4DeQM^(@}eH>q6_ovW3g^QNQ)#sekqxBn;zh5){d<&uY0Rz9lA1uJ!w@_0MT9
z=rX-KvxstmbD|coB#J?A;DyA!7KSsV`^YLka<INF5C~*lZXw{KyvQqdXM4ny3o$J-
zWDfjI7oO$dc#zqoFCFfS3OMK}kc8t~4*KLSs)=B1<dJ%EPUHLGpOwOofBEvAKm!#x
zgpr@m+wTWYVc3FVsRqup6c`j6`2);EqWFE`{=GWwX4rZHo`Jp6deBy}OpPyXBZ*a?
zKP)ouQN#s)-ephy7RjK*Nq~6p-9tFTV^Ej+$-ggduaP}omhuDZJnJKu!iR^rt-Tw=
zXCHldd=0&(zAy9uoEJ?1;NGiU=oe)Nh%a96dKPKV-c;LD_6{3$YYu&6_lufpSr>9s
zo);DpIuHpw5@}g};PiT?+8Pxw4-#-Isnezr7chYOb@_1p$p;ziGL_^?=TQqo3hw*@
zYo?EeMh4aH)G8?zum2vdf$8CUK-h7i#U(Ls)Q1p+aA?}&jr%=Hn%8qJ*5~|J7GGK>
z{#`h(wEOWa%S2j%&-iiKE7WuqZOnXWLjO#Df9h&u@;<8)aqrYl{;pyNO{Nh)vj{cn
zQNdeTXs7n1G8LXN^xk`3^%?ojdhM9axCFaZt;gRK{<VC4G;?~bSQDc-Am97pYG-;J
zS?yspCOi|<-3E)lQ-Lj~*Dtl4HTL|L*?;Hqrgp7LVbA}#1J}d1m6Fu|tUrQ2RA{BX
z(3!5{SFvtpDMub@;PZX^_Ngeo6bz?3Sv~XDM1QIp@1IdFDO?2OfFN)|JVJ8oxwII~
z#|4*a37~Uk<#m%1lQ_Z)-owtm6A&MY^-!)4Zu=~d@GOIGT?vXepB@CbR&xYvw?RA@
zW@IE~KCcMV&f2O12(G57+@9>aF0r<6`c%Lo-~NY<E*{SBff=V{w7#YAI3-$Yj%BL8
zoA<6Sj|4FwIsBpl1M<YEHvZP@iqg3Lipo)H<(E%(XQBWQr3a8J9v1)!KLjFOYG<No
zVd%v(q=-JkXEPu>z<(r9;<jAlK$^16PHW{~ntl0yEDWGxmO3_7PH*ggGSRZ<zXy6D
zY5He>0s3uYW+XrufJGD#8rJGd&5rJm1uI`K%jlMy4$QvWGN|);I-pl(F}l(tC^S^`
zlH(;%K7IKHNaj4Jx^t(M$i>^Y<EeI1?x$b$>BlA1ao@`LVuj0uhP$`m_-p-5b-Fs$
zDA6T-*Ls<j*^!OaZl@=qRjiJF+>>r>Zk~Dh<(7DJPtH{`X2}^{tKWAE*vm$Q!gQl)
zfA8T+t+}$m#;ssiLP_;5Le;i<%tj}@P5CfR^PSbF%z(UfkUreN3XNiNe(+xL@v-MW
zkJAN&(@2)@Y=gsZ_tzV{<uNhWU07;;Dld{N)hh483VxP_6?<*#zp<|7XM?y>I;Zfh
zzsdCuX}|(hIg{`p$EA<$w$Rm4bbJR}3j{Nu@k{qz_;>l2BG33wZ@T>L8gb7!_fPcS
zb4BtKuB`rnAwP>?OcCoeNRz!OKb_qiM78S$Rlt7$+J$`TGx!hPr5)twHry=vW6xb2
zn}*|gT)&5Bx(G@DX?a)3>5{%8U0W8oz-l|so--e9AA_koamixK8FqGkeBfgg%2w3g
zvd((7mfbeocR$%}a;53t3kK2k&z6lx!_S(4Tq0Uig6QG^-NiuPda;9UvLXkyw(a?j
zqrls2#{J*^G+qlNMXG1|XuF|gBgf@3C;u(?7QgV&$w}RX^I^|Jq)4mFSg%hHmnkUr
z;(-%JmDOL@WNx#Lc<TX!+QuZgkCZiDeOiWqRN}Hl3c&P=zy@dAO}J|MBbSGov6}-e
z7tB!1X;S|(dfL&+%~fe4UwOLtv-wuHtQE7tmcEkFmS}QmT|$$(gqui-Nqwx@Wqd-X
zuB)izVm^Me#jxXXq@0^|7h`fQsZFkM6P?PWdlsxIBca~g#R$oj^4Ql@Vq{VoTy>%D
zZ%}%b`S{zk!78;x_dm}^LwlAhTim>y-^d1D28Vk%&7bWT_A<;#iVFTAyEat5F?m_G
zuFqMItFzlsoLI?eo%EQsoaX}9a`1bw3Vah)Zs{o0ni!l-ToB#H3hD9TQ3RV#oheFv
z8iJ;Clc#ng?uZsH*q2VJV9($8qa+7={&wYjZt-p_P2lTQ<s~0T#Iqam+8LF5SC0c8
zv@)X$a7JrKvJv?kv3)no!IB7@-_m4WIL@w!DG@@MrK0vgk|7`N3EvlyxxZ}21RR-m
z{54`I>jRdOo<?tEy5<p(n|<J;1Xvkceym31k_GbG(WAp<M)Fs5$NokU!Uowl2kQxl
zNo{Z*>;3!p@3Jhz1u?_<4+X$SpqsKpTU_6)ncaL~juEx!A1)fFSqJJLpB*cHM4o>X
z<DKr~$-cnPs=xY!$Kci#_5h6qbNBF2pAjb_M>E<OE~k4%Pkt4jFL?z9N&~$8W5BTo
zh!WW3&!5wZ=xk7E$g-nT?ul7dJbB4vrO%L|ZmDTkr`*RkAY07I!EAW>WvQ9UGMB|x
zBA2&sK*rk8V|Ay0rq1I+Sfs6fMZefst-#*2a@{^QfpgUhr_A1&GnSIw-0g6p@4oG&
zn3lPN|B%Y8*ZR{{dIkOg;Q;?mURAQX7sLR1_SZ1(I3NzXjUQkfZ4BBlNh*mda*;j!
zYp#!5O^{p)M-Cy3jJ2WmpVLjal79@16sy_A)lk<TobWbB2Z=5{3u|BUZEYy3s5ou)
z|L(b1V?U5QWU}M?{ku3Jfy(hc_M3mz9~|QAAN&}pf6zBlUp8Oi+6r4!o$LQq?*IEn
zc@0(;s)K}Neeh-4cgQt>YLuR)v9<H=3oOyJXF~W}1MjE_AU{v~R4g<8ja@WOt)j9_
z3r6}>pcwy*<B!XJ<SWE@^7T}l5G(bubd3#v)elMOR^xjT)Sm%QU5zXgs*@m%8Sn%G
zGLSDb?{8vWvMFEfE!V(rp8DN&rq?VPGk(oA`r{J=wlrBBs4NvV0f~AmsEE}6`Ib!X
zE#W)1cR!2GKmBQY+D7DO2K_g=TsOLC+y8a2b1!Zs=j`P*s1X#`%9L8G^lnx{HX<qK
z!+sp4e!nH*e;fLzsL`Hq*VY?-%6@iqB5mqT&e@WM2BS#C19yLS2?U<J4RHMn%V>RL
z*ExwMo}Qs;{Qy+RRqrq(vv%r`^KuCYrioYhOq8H!^MzL#>bGm`USEAh6u1rXu%6*Q
z^IlmIk^sLWL3|eth65tRGZTsj$qg+ZfC=lhuAQ5m7vNA2qI<GF*$Yv<?@W@P7!CL9
z*z4E3lB=XVwRN%Qx=-=cW^_fDqJcZ7z#20qy>_0It3q{BH+?Q`$W5)=y61_ZJzJ?|
zZvToEJuJYcv~PVNCf>bFbwNJX(-ksZ?4_!XK{iEABe23}VB-2DXnaGF1qO*oaO<1g
z`^-Vi`3Zk!Gj}q+&fMMTlm0f2SD(b3Yea4)x{qM=*l6SwcwLiM->br4OQ_8eYM?0H
zv?A_wuqsPxSF<RyAjpb)&C9VvMMdRR+Tg#~by@E?*XTc9_;2B6;YF^Qbf1U{M1M!6
zptg=DAn9_aSEtb3++s!e6x*z)%(arrQkQH($5j3p=e%z)>Ncnz555TrR6b^yXp1F#
zh~}V}2k{jK|9P-59xPB=!L@~`4}rqJDF-r`-C#(R-v+!&ys(H$#m#tNOptLu1G}{B
zv>Ew=q>-U9{f}+Glbn?&g0g@E#YCZ5gl&segvpLfIf?SMS$TmXxuy?*b41_xE=iFb
zY+d_sM7r*))-~SE712!zyf#tzJo$l#g~04CP(2{a?7R1pmyi&5is|p8@cNtUjH+YB
z-Ryj52L8=yJB8y_R|@s^LJv%%IXE`^ChD<d03g$%o!-!D5OWPqMfy)%(JW;^A9VvP
zT(to)#D133K~h<umH(Du=W3TQ^5^9EaM1`NLXTePVB;FRN*w@e*7DZiRC^pqANNZA
z<3pp-GzJdPwnS)fRKK-<c5{58_tBfT28sinA8NS``o9S&ICoz#8-U{Gqao$^^faJ?
zwZFUH_(A5$K1)iJS~u&bLgT6)2f2RXq{h6#@}VOAOinFf?-o`Ux&)0uw!?{U@*dlk
z`sDr3-RE4!nwr(c7n23O)BCM1{owg!WC$weNjsHMMuM>mRKGS1;vDj_^t>LN7Gmjp
zV|G>*x%6Uj^#JC{Qzjq7`c$`?!e}y{teSaa0n5rv%^0!wY>680icyqzdhvGurvCTo
z_H^ans%zNyXBR%cEu!Y>ekv?|)5>qR5-6(zdQOzhjksXXX76gxbB0q{4r1lr>70QI
z;A+J1s1S~l_P4X;e)}O!hBtR9XH~gS5e<|A_-lBPrFYb^Vq$sNk!G=&6jLs5v9>Yk
zvFRgth_GYfie{2#0ss#e&S&#9PFp<s&tf+_COd}8R%Hf%rC2MAv<5vE{j25tL7^x<
zwRX*RyIi7|rK}fDqq@0^J0Ia#&Pw-&>BZKtv7dYT{l%IMGHEA#oqm(4nb_2o7-95d
zwG94a<+Q~;VEmOvU>X6XSuo1@!C;i)AA%m@zPXU=abD!lg_>}hSg{nRtfQ;uvfKO5
z_D}_<eq_%25`j!zYfSaV=WLThM%oQRYB**x<+muk^P62<QH#;dc_$^@uJTsI)mqiY
zK|xJ<_P!gyowxx$_b5nZ_QvUw(6|Wz1<JCOu(XqFQKJe&TT{&g#kLcv-_#*S9F-uk
znWU|X(Xo$xJ#(1!#I)tJT$ze~wFC8i3hr(f<H~?Ab}SyfTdsKgF?XW}c<F_@eO8VE
zIA!3*s~`(6f2FaeJGhPga%^rx>0j!W^0e}#ilq<N*Ewyl2<jHN*3E8kkf=)J>1i?1
z6j^tD;w5nH6rN#EdrPzT|B>~bK~1$=+bW76MFc^lccg<z=%Rr1fYKopkxmkNl_t_#
zkgik_5GkPq5_*x|QA&V>CRGeYI!N)`ocU(Xd7gJ>|6_)k<led0TGzUY%YYdh3BZk!
z!#{v$a>Hq?GSwgor79R2wEDT*du}JVbfzH>P3(;5ordoOcH#g|@`$;duX$hN1O1N_
z8eH&*9v|3TU1@0}^qrIgQ!rAi+mOdD5ZY*-U?vnUFs_X+sWY&<(G7S#xV7Ecv1NWS
z>PUL>_nmjtg*C(O-#hYZ`=o2s->z;DQFc!H-08y&wFx85b%9MgkEYznE(uC9aFTk+
z*HY_HR^TLK94?vIEMonaQsB<SzS4e#>%H%c?b+B#^S0q*6hyh{bt`eqCoTEz_uOsY
zmH*OdUzj&#IoNGI{1kjovgh6@g~F-*E#<!;7M?#JMpgH#m#=&uNe)y-Ir3HlH&V=n
zPY&>+elouM`Nw-6ZBYu5zjLQ75;u-~qL>oIiY88N>>si(#6A~Y%Jm|9LsELL;s-=_
zdqq9eJYaQ@AXRkm4U}^>K{UT+%^37-?Ce=tRxQ<?yxK>-3eIN~xsT6rN)2hni_^h>
z7!;;<>=DK&meA+{Jx;~2w3FBq_!=eg*+a5hG^H;sfn1PF&e+NCNCmU)F=QVQjn;Rs
z_4t{#Ed2G~KQf4pc)-4HDi-gzw+2vP6$jql5VSzWO+Y8z_N#h4Iulj9K4BN?gv*z%
zQ4xRt`B?U8hQ^>xquvV$Io51CW)mwnrIlqk&I1{tm1iw}Ff@wM)?Qe($30vOT3WUh
zsagh?BxD;i6-o4b{={>?RHnYa$)bmPGWu-)TvQYL*&W*=%*mg`o#Kl{N{oQWcjR-E
zjJgH7`0(z0ye<Pff{>Bel0*+uoO17HcaZvTMSd7k!=`UbtM&vzB}Jt+B|W21X|-K$
z9z56J*ui<~v-w!`+oTa4*L4PZlV~$QuDOeT=mq)E4-OK=^^_E3d}O9LzwKQp9Hn+M
z(dN>AAmx5S6nQfD_$3|V`W^M8fFpo=50s-AcPVk9X#8>AkM>FPdC|WSu_8&jCkIli
zGJ&$*p4VODzpgAS!;=K@WAj&?LxN=woaO+g_gKpu=9Pj-{Tk$~W|KWl(Cld}UXASq
zQ4udY^Eq?C-^Fx~%@Uel2T9sdMvs5XKW<#FEi<V>t>%V9nY<5cc5W9bB>QB0_J!pV
zfDkYCNefx=mBUmci#}k@1f~-<^Hs~)1|m*~(k+9e0U<-O!jxc#n|YcrWEkhAit?U)
zeL2i7VhaV0>)v4j>5{YK)rXGhXEWs~#ZS7<a14F8IzuU5+fver{UF}7Qn!`L``{K+
zM^o}{iF#!uje?IYg*%7u*!>2Zo)YiJsnzi67#m(*c>Znb#C|LrIccbd36vC<RT=*b
zE;9fc9IB@15G{Xp_#j;~W%2d0+Eg0QvicU+9E?eD1*dUX30e!3CWHs{gXXji-36~8
zl%Dif8aZ#TZwHoH&(HNVcw^ws^DWKI`K9>@0qq!5W_R>-jW^hR#SHIst@a^6M@A!4
zQOGPm&3OFBXb2#l`=5uB^RxbpBEJj8_tR3;;_2T2!5c@ZZ39uiBvM{Pt<I`Q6x=-J
zU1EqV-=*qMk~gd0@_Zb)S}oO3?`4WT=`z3l`*UQF+S<!|P&#D$*Ox~Z_1#w6t|EI`
z*JR6s#5Ec1RP_!|K^GCNjYp6&Qxrm_2Tafs5PMQ<WRrt$!w6htgol47mCs%ClLO=s
ztAP7q$x-o|FF+md0kmS=42-;KJUJvS+a6uCMy^1(;Kb?vZBj6J&$z~hG?uP*w84gN
zz^HsN>Oi&1`TV<}2__Ek>K!paa9MHKeAO7`nM|%od*1cFcdAeqo0OqH7~%yZg$|hU
z(zdAo0jYNwCV#<5c`>;wPUcu;lFV_Tp=G$0vxZlRnOyeCB`c3YphfZ7T3Sk`3QQ4-
zRM$C2YcOTBciyrUt6<jf8iuig(c@t1#>Y8p&n9|T+)ZQ65GzO{VZB4Svl<5Jn$-zY
zarjcMSYB!Ytw1!Ry+?jMuxPjl2|>|B<b*L7hI~cN!Z&Zj0*>?iePHvrE&m*<KAa7#
zK}a_=hLCD0zztssz&oLiUKSvMb<RJ|E#<{~M2St*oP)%h<4kWKP?q0W^Fy=O*ARQ2
zp4iPV^lVhoqt-Y1T!=_bMx}C}QS2S7<3qFX_~AzFmK4o&z`dEwP+8ZST5hH`LE&KI
zhyP}RN6*$xv2Yb^ZF$#iBK}D94Qf<_pID>+q5D$qMWm(0@#5>&8tdKe$l$?9$+HYv
zB-HT{x==`YegR3b=Ri<s^TA90PABN(e)9l5dgKb6*x&=V)1>sKu#tdO@w>Oi4wmv{
z!WBa<=@^mWb>;wG;HAnkRK=Us`J#tqB*|950S5c%I475}I@BQ+!M2~D&HcyZ9m#XS
zom=QJbUn_xr;p3o7F{W%4Yv4*XcE!J^R!FcsX|m?-y6`$8jPaKxnnqi`~YJ;3p%6E
zPT7gj_{sM?R@QwC$4|4~)0XZ|QDb&yaBSpz=w8QB5M6CrB}E6-$PuR)R2}c0ff2GF
z#U7Dt_+hHh6drFucdFf(-LuLmkR{7M2gw)dJ0<SQSL74gLj1j9vjN)|v+zd1bK_5A
zV*_ziPhoP<+Ol$qkVy|GT)@SMrUp}h&aGycSe-O%wEV}(0m5F}ewx~Mh0f2ovW{xR
zv%`Zu#@yIOy1gtzf6xa4|C}ee>Ba;4$iTv@ausTt*5W;A!Ik2HgvI@5rGD*iSIlZQ
z)Hn=uGyPv5{&x6#b&D`NAR>c7xqS@%G0H|CX)vVuKq`L#O!Si6|258UmPU9bzeqlR
zP6P_R_?5@HQ>7ZzKH3E}ey+x!3#HMaww-Ldn$x`H+2XMUrU?RqbXzh3Z{FHDk*ilK
z0ym*2Q$3FjYuA8G_)}>@S@nBxUc)8B_cFW}z?Blx(&#9er2Z2~KN5T>-r)aw^gIJX
zSLuGhaCwWCNI*@{t0dSm^Pb&Q0lr6%sICWPZN*)mi(*(-wov487Rt-$GbL2Hy~4gz
zH|yy$B9--S!1nPcsgmeI%i%nB*(Hlo;4vOl?I-M%{*3DDv!A><>SJ}oQID#uhfp^}
zBOa+UUboWzMr_rxC}k8a_bLdU)|<*7JyT#ey(ia%x;M#|?$HjaKhB9<I}LmKGJ4uq
z*ZbKhHmz1IsP*TF3^kUU<{?wc5d|eVY~2|e?64LQf7@kl|5%u#%9M=;Gm%t@c1na!
zu;KBW_v<DQCPKEdrr=$N$;}G=IcyE`vOW3l<B`37xAqDm>N?r{@l3bhWRB1Mr@ysD
z*YdZU5wU&!5T*SsKe6BId$DTu^vvkIXz_a$D}!C7(hc>g)sRW|ucHSWA{6dxa};uq
zzkSU=KdmonWFBFR;nH1_Gi6*A8{zAeD6kt`L$ojikWNoFXZqxN$!}+M%1x~=nkW2y
zdnC8pD(GS|98fEYiNj1m7#8BXN>BE>Dzbj>#gZ7Ad(mF5U<rR@W&lU(ORif5fgxai
z4WOGcv;ux6zZwub|M5NoWI*nFLz30651dR*O<l9TwAAYJyUjNr?NlWMr)<ycm}e;x
z>Bt>;A{5r6nF2R<fQy%DQfNVnn4LO&inhx@_({uRono6xw1Iy&R6jLNA{k@a0HUq(
z28&I@OinMBs}M`eG_{{!$u9qj1vdnH(iI1Q3GsWMrjW+G>oD`k<=WWJgRb>KQ_W$!
zc0-a5w%j^!VVRIawSKwrq*7l2lJ)xcGF+1C^gt0bpYUA}GJiDchk0Uw<=KO7j>GAb
z9^eWXq(4|QGEns_gnxZvX*t3P$79Z|2CNh0vFKn+oq99v$L6n$IE-#hIPE!<7CH*l
zMlRJc#Z@wEE+~4rl}hye@4*6kW<w@wH?qpb;J*i?QBA~kHvmml^a$N2t8lmG2YO@N
z>zZozek3~@RDZsSCRo!6=JyDU%6}t8kc9&f321Dgjuzz}(2ngEi^%(^HTnGZUr(IG
z*=)49>Rijl#rEzAMXTM(qQRp^iWlI6r&$oOeLmvHw=MseBAhJbes~2OR_CsyR%F$k
zWebV~?Zln{;+5yMYqci}Hx+sT!o})q5o`U6U6H>@t$$Bj*8t817r9bIWF#M5Y?##>
zLc=4_BkHIh-zgYRe*C=%6N1!Md!F1fL|P1d!WuI5u#=xXGS4h<R3(A*8gLrXV4|}m
z4#moxR?N)B3~XRs8`A|jQW-BYxxBMI=gRsDoss?4OL1R}KXVkyqEw_c#&LWvB_C5x
zr4CUN(@fs$L4oeJvwczfW>@XlF$2SGJSJ6o8xsb1m|!m|hV(g~T7X8dng#GA)IooW
z=7*Yt&@1W%NSkO=uAD|PHF`_UvAP{{2pct?c!VT0s5sx{W3b2e1yO>}#uVCdEp4dr
z^pF2(v<;GMy&}Ad3o4`z5egKDyW5;v47tB{@ZAo_qK^Z;mN%HWv~c_H^RV_q8haWC
zM^zK8bsQ&$Rb7#FZliNoMZpGZD8X#NnE3hCq;c4OcQAVzQ*z*1!&j*?XVWWSGErrY
z+GeRI`AaD8?@m!G?Y(Oy=c*tat6kJLT?Y=B^9u7(P$XiPidE(y1E`Q@=cgjOt^O=_
z1_TBMW*}^od2?n}_J#~`!?EFMe$#s09&b~cKzZF;U<9oK^751}WSSOfD;DwrIWBwf
zoM~3@#fhL<-MfnmK<s3Do$Q|Qvi!E+Z6Ga~5fC)*@9#esPXA=%M<r?|sLSt9PEPs#
z`&iF6*XSjk$Ek&-Seme_&Vn!^OLuWo_;)8>81OnXl%`1)D`|&sKTtP?RAw0Wett`*
z`Pp!sZH*;8#sa!pCSQaNdfXHASYRE(4$n5`;!WtFYtYX66K{F*&ez*SZ$;ITQBNIs
zZI#;Dw0k4iLkzqG^SGV6=2%$sQ9RP>a~xqI3w|sumfd@D8ifGP8tuL-Q4R)na4gfe
z+l{+<yn1bba#~v-v+Ns$?A#k-4s}sfB#n$9j>ND|Rj=i2GJ@nP?I#g|9igEtiIwn}
zR_4+|7}S81*YceIh#SALG=0sdMjDgP?NEg=Y6OfW|My%!>5B>hctW8w#gnnFBF%w;
zoGP3Ndf&DU|8$iAsaxW`Q|z3SQH@EsC%)?Ke_I_V^WjS+vS~~#;?=^K*i@evVR*qq
zry*x0e((0m07+ZRDBH-%n9&LRbg2w(&S>-^A%MGQukRX>r(cI4|Dg2%ZH0Y{?9Pbd
z%xeNNU?<j6uArhz2q1>`=6cA<Z93{K0W$THqlfNHz?v*zKc964*sA3L<~H~-8g(fm
zE2`1yG0oEp?S|K<_n8>%>Hy=LT@uB-Zws$P)#PNQX{#sfI@?KJK=@$r4;h0ML%gu;
ztllBb35{q5BR*b^$+R|SIVfFndWCk2sdN%j8&r!yJ(fl)4G&mC%w(MnSF-d;?W>`o
z`sF;1csBAo{@U;dVbuCu&H{Eb2&x*xXDEYQ!6#H<z7{E8z@wPN%J@RKG|+sqSiBpo
zj&Cvg6o!<*V_xOJJ!k){K8$F4BKF|ukn+G~eL8qyXL!~hv^|f8j$8z8$}2C9YfrIg
z^&8Tg#LF29p~mav=bUUH`-oJ`4e-ByeV?w3rPPI+{i18TyZxH&I?ibBAn|6wGW^N%
z;OpEo7>6Hw;2m=6fs8LI^$dy@av2P#PqKc-SA-j8+dohY)%)mL`Kr2`Igey-<{Ff)
ze7Ui4YiVn1Heh$qey41LmlZX+1Mz2BK?5YUKqd1sOB7|IYiB*P)?KqVw(&z78;)yw
z`0dsBWVP8j&oOYM{wMA98s@ZjNFwGel<P`Z!le`E&;vLBOWMSTk$?#CVX4r$ISY+K
z4vzM=1*&GE(EZ!wz`?1i`iPtpDIGt3r`e%Fuaes1oq<stRJ>cn#0T5Vuqkz?7qlYS
zxTnR`z{6d^olaa(WW&;Oo!jCB^1W5&5PB$wGxwX9X!lpI?6Ww$%_>r28-(;jOj;+h
z1YNoE=8C%VQ~gaqd%-QJ$VNR>mdXa6P+6ZU!y0pOk)ZZrX<yLxQ+CCs#aeiclzv+1
zDm>#|D%Je$-*O_Qs=}AM4&TB$hUpA@!fB#DCcI330E0VN&X)|iIzm$TZCjX<d}~A2
zb)iaL)?024V|}{SvQ_?7DvCgTUS59pb86qz0$ZygP0a)*!N(B3;!QtqAFGf1=QaAN
zQwy?4r&8ejOxfndXWqZ5+Vx_fV7ZyTrwlv3*{RkEONkqpH{xR{oe(IZ%~{`}_R5^_
ze)}JlP@nw8%W$~FEyTd^SK%u;KApv8=w07(V>da~VarEC6_eOJ=f_o>&KKu9mQSo$
zRMCQ8Z-@3}y*;2;@~=vup%Y&Fq3Ke}H`epXOxObmZS~l0KGGh5zT5*A4H-H?yUBV%
zGeV)WX5N!Ub*oWh$K&@$vVOY(J?&#4A)4te+xy>QpJOgXH-C>4{w_puB0g4EC+k*4
zv4-v6rbHb*x#xe|^Si95TCrL8Q9l2?a+C_XQyXrjnWbziXay1L0M)lHifQx5(n$ZZ
z5r+=-@-M5EH%RE!oXFVlrXmvshwbJY*3Ha#wXvp2k}dE_Av}~eN*8T7TP5!UQVWAU
zmre#{YBK5r{-4?=93h!3Mh~q7Uhz8T$F7|otG8nXR8iQLE$&e)FG;{6l~?@h<zrrG
zlYB!D8_&Z}h2IF9tQ+&=K+&+yZp33e(VNqg(|#`jurT)+|27eY$$%l#J|iW1yqiY7
z<@oU5JB1x^hW{EqK1!u5Vy{^3kyW@tEZm3ubPwm=T3;x3Z7zQ~+~pSWMxJh-tKj*S
zqb(-m4BpFOvnSl*J6spOi;WZIjl2bLv0A-~MFZ7rdG3m#meDId8-8)DOjCpA%=mQy
ziYnNJqUnsv+fIyDz=`y=JGjV(lOgVk<xhpG+Q(7=d7Jmrv+DEm8FUBX(-azf))Sny
zyA*`pw6X~No4tRqy|!<_yW1FRIee|wZ6-k{i0-~M;IU=~Fb)Aczx6$K@|7T<zEF7#
zWE!IYi8&?Hl+NEaw&rsp&m+ojvv^e_7uM!;CEu0=v8pytCoU(NkPqYRMjy;&KT+4m
zRD$&|J4{o{80#+?WfIctnSDm5k7*{^8F@<|Vr-E>6|5?NbNK94yX5b6>PMMs-s%Zn
zxL##O@s_PrSle!#XQHRx2L5t_Bhp03y>JRFkR)I?tHI;1r1|TXB30P+mV2?&#xth(
zF6qI|zG+BRsxZ#g3A;&vM^x2PaYONDKMD7ae@sV5;~z>U-++-cWWojx<2Cv1NwE6H
z$Z}jtK6-T$ky?TlNG|$B?N{^K=8G_NleKQ5bJMum?ceQ;yRj&8k<NV=8=X61dHjbl
zGc(s`J+nRJM=i&wE?_+^EarCA1hUl7{UIr3^Gl>?1Y(q}aF@!cLOxu;kZE8-Kxd)g
z#=rx5Ai?TR7zI)G%tTda<e$Tx)|;5hdXJlrKx4RZ#aKMm(0z13+g(A0#|SH<d5bx>
zUj}DOeSrJ!rPbIa>k|gV*~b{BeV>}@Z9N-k{rzXMHK;$~JUshoXH+2|arMT4v3Rf;
zM8;<qF;;&62x|40Nh8SS3{W=;xcCh~J_jI|R>CFc2YD3^_<%U`k(@*CKc_dPzs+{E
zAe<ZgPwDy7e#<HRd85QHoKJWch+eVXZQjz4!kM*W%R$53)|-LZ`tMe2Kio-r1RhT6
zHM`j}w4JUwE@LQz)tHhw^fOMl&mNus#RW71n!8PNE*<Gc>rU0lA)($*Jbi}y(l1r^
z9-OIzQ$c|gYC+(!{@!{_(k*j$59eoLY$U(W)cVup6Wjqxk+6T^&xY+E>CHb&w`n4*
z{sc1Kbks8u`2G^FKXWaN_?q=^rHa;I*O$<>kn@BsphS8c%>|yy9&K$>+{ry7$2(C^
z<QVyj_Tj2*;1YbY8DE%7AJ-eEy03WgWAtnr&&3tH{yk0RINqy$*AF3*;#edqJKI#P
zNc*~&d8ht5zz-!8!+&tHg3c2dZbruVj289#no0p-U<TI`!uRfZXzfuWoSj!<R(jm$
zM!lHkQRFu}((;72ww$f~^vG7et;udCRQpxN!j<Ir$q%l4lXJN`zyKi;7)6iO1jzI&
zF-_0(M+{g<7B2zheO7?}<=c@`4hI}612(K#mrO=pV^eaEB#-3!_x_%F(a|xOV5hVm
zD5gseF1E|Dv}~XS`5#RLz|gr0(RI!LWfo@aN+fg*0Y?oi=w1LI!SAUh#`P`6kF~*X
zhPNylPKg0JHb2x{@(wz&O=~z$cM3Q_yLH@|)^wt{Q{M!+QKltyU`%b<^q||D**#Ss
z&hIdGkeG}v&&-NdFa2m<q{2Vdm&3><842RGSyG0-9LYcnBz1^tOveP0H)_{&ynxQ~
zx_y$w=NNFt8ssuvziJ^@joI)6I-oTiKpVBK&|2{Kwa|}eiZJO1on&CydBtJL`$#cg
zl?O4Cl5%D<^53TGciM){6Si?EKYtw^AJ(1iu47$zIZWuPe<dmYI{*~`Cb4h5<woh8
zjJS+WP5lBp;i;H^_Nz?7Od4)vRU$NruG9DCF<A4~UpF4=2iz$4&~3Jin(kk|UT}Yp
z<*J#rYt`zEcH}|X?u=5%$Qz)&)}$hSB!R9Bx_Cui**MkX>P9uASffG8cXt(V;O`Bn
z&HH6GA1$ar^;PVJ?Y?oa-gO7U_JX+me6IlF^={b6C42YM?R_XyQm2pLH|MvuQ$G@j
z(9@Q-)j-(q)h1k|1p95Q3FfrEUjh8fs<=cn)r{(UXYaLUjnN@XE?yg)Oee!MM@ugA
zW-k$g&jGXf|KXN}i=6Q~D-k8gVM5NCXPJKfvihRmegsSoS^|lxpY4#EW(rzo=hkIh
z&J*`B>gH~yCa?i>d*xpk)LW?!@XX};?!G?Umtkp<s69SYUg{c~^&chbkWz+ey5SF%
z+(=H!YmfrN#9C|sjx9<V`mT5KTT+AJ4OC#TL-h1dzg^n?RCI5i@}51>oq+j{qcbUU
zMgw-$YMACUu=^NUo&N0A3mC@3xnN!NibSv4(kta2$<^fVwOvHwYF&cMs+$8{f_;u&
zP79>>)#8j|#!T<Yf*Hr_Tm-FTSzh>8j)J6bI>Ab3C?{?>qw)OCmzLK5eWYOz#(zl1
zi*-l!=B&(L^vp}eW{BKXc<v^9^)sg}<bcKh*CIPx4vBg-{z*0djib}kH}3S6MN_v)
zgq(>q?+No#MeBXF8$ZT=ZLuK!eMeD|>ZU!@ZEvv!sQ|a5&}8b`%MlO=*ya^^{XLfF
z+_dFvMP#jm^+HrgNN-;4;upa1EMon8qm-_qu<+YE4T$I?hV&%KLDoimNeupb0dhz@
z_he9s-?^f0@z&%0!~|4pV?ump?KixFn$G-{(2kCm<ToY@^A->(HJt}->l@QHlrTeP
z*oKTPWQu3hE36XzD>6E{p-+tIrlj^SyLPO0%W62B$7gTdRF*lD+dJQBe`W)BoRt&?
z=dh=Fx?$=ym=|^&1d0Wbbm(^zKShsks6{h|eQ*VHRp?_`@xaacPLz$j?-1<tKKzka
zraot+RTS$;^?|hsV<8Nnfpzsk<`W9x^e5kHh)sH(5|=8XYYyU{-wm!b7Tmi3^Vi(?
zEViWzj>p1J579o<(OQx%PrN>uB7u_W312z*zPgB|t{P*9t7cl4a{R0RDZCZQ<wnK{
z<mg5n$IiPA?D_?}U6gigg`{8U!Jc-#7h-y7l{sS%cqS@&dFyRP-ahmtCV8iSi|qLq
z!+Yl8dEbagk2|_57CX~ylTDAvERPjic2g^~epX$2OPp^9D_jZ$$D5+)nuDStTPJk$
zfQyh9S=tLxXE{kJ|3Zq&z>mqTa=!#yJh2Dsho%ZHC2$|6$eXMI?*Py>UAOqr%G!%Q
ziIKXrVBZmiR}2})dx5vVk4aK-t1hAICa5u2u;f!*>WtBNwXCF41)k*mr8^kfJ?NOz
zQhy2@&2QtcjYKEN_h8+aW|Xh5CADd9)o+;E_0YBK?674pG6Kg^1S;mB3@!Tubp0_N
z#dzc2-HtfX4}lON6KJ7*@_DDmB+s53aA|&3IaS6hN3AVJ9OtldLzA6sG-v);+#fmG
zk-x1#@=mmdOQ`rP-bH!MnLLE(s;)4_?d=U*>vo<zzE%O(;&ED@cX9EF_4?CYIM|C_
zCMRsuVH9K94vZtAA0V>Rrpu-t@Vtf@q&H;j-`Ard6RHysVRO6OX!8%SpKr3yI)c9a
z?Z9Nu2Lv{>vWb6K&$5=qF-QI~-zzr6&`6tdz&ng_pYdE;Yw3o8W=(sGiZplt9G{DC
zU2V<rJU7;n<Oei_vVnntm!2W4N7f7s48H4{mi_p-Ll8pd*B8D(d7AQzj(aDq@(FI^
zqX$ymRPkqnt+;D7!;rQLncA@>T1WVM$*mMPM>kW?4?rrxF=^&O0nGP3HCdYN)h+?T
zBEb`I&wYw&$Gr}6a#~I+OIJQA$`=D}L0d3S*@h8%-W+R9e$zd1r=?}Mt~_k05v<!l
zscl$wiCA+$t&20T_j-;RY0<qrrCQ!4Jh1AeM#C6MbsS)YhP751=d$LIW{5a`N~NwV
zpw7tEXVOS1U87q=Z8*&`QjpruuyV_dah|tc#s=B~S8I^hm{+kRnoqES2k@m2ylBT{
z!30#>u}7Saj77am-nLN!iOC~;^yGO&h2L1Wwpph}0~gws%p36_A53lS_IztC+3)C4
zwDwtgr}j8*5`G_A1XnG6#v+82HEs9)l;r)$PZoD@C%8+sUgv+W!b{iE)py*^R?5Gy
zQEHVxk&+`W4Z9nR4}e^i;qoK0hcw0Mm1s9>F$wuK$ph6Kz;s2>5z0E4_X;cWp%Q(c
zGJyDuI>GY7h)JT=OLy)2?CcQKVvpf6tOP*4pj(qOGlL$0yPpMMap3`!FYpQBgDatV
zel-?<(fK{rFu~EB$$IKP<e(tR3b{WUcapqo@EpKSFtB|2G7ji+qfvBW6W6hbHO$!=
zw09dLb15u!c>{cOQG`<7^C)PD^;aiu19fL!-LZywI*j)#Y}yyccz1qywl8rxH_T16
zI1bF8xG7^333C>zbAI*<)WTX;UyX>qV39rPku9!7zBh*&h!!0s7C<{ZgWy!l&!8!l
zLn9x+hv>D*&!;yH(QbvysT-xn=z+YbR1BJsB$UWl1gmyhe(;Ro_u?7K9#}?kz=(3&
z_DJ8dX(uMLa56eEj#PM&w<MyB_%TP!KYd--K7)_o^C(CEVkt8Y8gz6SGRqS30fjV7
zF#CXEcH!Rj{gu!|jqW?wH_MHxK^i2MIsSWrc#E>QE<CHr>&<gzo35idpU;jtM=CpK
zjBjon1Oa&hkqvb9*z7H-(zmz0;|~D)Ggdt?&M;<LDw4ivvt)_B12Na!w3#B{H{X}8
z)8`ksNRzv8xoro6Cq4m=@QQg)l$3}rZFnaw{Dq7x0}pC$;oa$7XIWd@6YwmHU%H@+
zH7ipUgsnV_Ve<yWy!?Uf6CQi0I>f(GjCj&(y3URBo@Id8>EUK3Y7bZ(xt^)(ETaS}
zeryzFjks26`S{%K7Q1_ES`e2NHAD)g2^N{6p6ug>d7wKn3trY%nk|P?fiN@Z#*|lY
zca?6f!LN78J;tYuMX!dDXu3%te1Xd+d5zQO?2Cr<uQ<d-&~>*4P$hMR`lef9C$m_G
z;bI_v7c8TUnJ=K@QMkbR<HXJ(Ol;?5A}kGQkiKzZQO+_gk}l*&=o-ap+V9!z!IDid
zNu^!Fn)HVhhb|j(mD^v>Wqq7CoLuJdv(w*2UbJJ=F!>YrCQIa41Z)N3EW+imO}J<t
zSh2fovHIs%{z&{pxlv>GncQ=?(tjYF{Z`N9yWklO@W6Qi)65+2?&@}+UM<P!Adg}n
z(!3|-!(YE@CjJmE^PxIZy&d$BmG+=lf>LVEuU6h7@$FLO-;&zr??0Wi0bGZg+S<Ku
zO1H)2<OYG*+Wt$)W_vhU(eVURUwG?@VC#HsbpE}H@H!43E#=?b1%Y)53TC^vhRspJ
zX|GZn<5Mu|U2tn_&MTK1uoGa?d%V8qU{d+cR>07G9ALT(pWJhwA=UY4%r7qYNI7p|
zv2HpJr|hn17z~@_kWK51FMtOH!TI%*d)8KmSM5jN>aiQ|DS~p7&^IcoN|KYkq@=4d
zu{=E<uQ&DKh9Q9MX?&OSSzsoYVJO?3ny1?3!~pL0%?=>E2!&Bo5o5~G2pVFy7*LSb
zJ7Ctgf+SXgN6~R*nrX^2F;rns1?>XmDzxgECfM=l-kdd>V+&6fYL)qU?{?51N;cOp
zYO7*>;ce!Pt+~qJQ^ZYgfHqZMTU(UyvqwE-ha0A<`5nOng&MH4YW34~i#Qki4*J?p
zA<}A)2cb&;u38VqpU$*LjU%Jf>h{Tk{cWtyON?T+Tkjg{ZQuKBF_;pz**mM1(v-P>
zG7*e`?@C?0cBg&nK9{(LDklqEojuJb|9o>=!u1V)q$#IWW_5;7eoZM_gx%=uLZKcy
z&x5e2lL?~e=7rPL*Q1NFTNMs8=Jvt{#_rnGXu_>BU&iGDO#4gQxJ#=}n#Z@7qy!-?
zNi^C)9K6}?GODcpfM33FzS-gLK#`m~j74BoE?Dw(b=9}{wZSAch*mjfie7TACyol^
zJ>*udC_NilLoF@hvs7R0rU*)J=g!TDcS_#tGnY)qdby8e`y+gr)#WQ$J3oPXSFsTQ
ziyDwzcf!Np#aG<i@lvO0c%cCz+4JfJHZoOC>CdPg89#rLY|w6qm-1q~{umICr7t<z
zh?J@v(zp?B9=E8{6MvWPG3ecS0%<??=XliaT2)t?Z$01a@4ew$;Gu)h`I?iHhgv)7
zt2I9Js=2kKAqJh~oXmhj_+IsBOlf;Z2wB4-s|w-BM@Wr--*aLQ#!pVd>mJr0pDWY|
zon>ZjCFg|ngiL64TMa8x`n4aj<?8F&N`E)Ns!rifr+L_-XfW$0A9k3<X<B}K*~Pvz
zo~ntel{>S}X+6Aa2tsSEpRnF)=<kqQ`rFdUdgSusJzx}7ECak1Uot44Il!ELq%-65
zC%eQPIUeGgF*V?%oN|KuPS?FXRC;sVCGaTqbI-WWL98dom0FeA{~7@ylfj8)1$$_c
z73xMmzc!;NXvYq&LX@d$uMV0%#Bzz-?BSE7F^>G8>~KP7^xQ2gs8+`EP^Z%~C#=}X
zGCZ}WY)J*(Egem;q3vC-H;A&C^m5;WhO(uVJvH>mA_0~Fq@i!vqPU#|?L&Q8@AQ$)
ziZa=H!)?Ltp7Pl`nf>!+l+vXdJhQ6QA<#nS!1<n|^G7CtmZ#TI0<}$Zy@eaTv~3xE
zQKm%n?Ct3D;yQcZkHP+yrxz5Pc#N|XzS%<|(by|GDEKRI*G?2{IvDij4^?FU+`yS0
zeeFy_r|z7D7ORw_!G-?<I3gvy)MpgqO==a++SKcYhr)i4iLKVXZk3AMK);Vr)lRJh
z)4M!#Mf1w0pnLlQKRoT#&sX!tW53amQYYjz#ZwxKQsVy{?bQIF;Q*O)_M(y1YtlQz
z%>?n8>R_@SA@c-Q;w9e~1;28^Z0+SgST25#r+dtyI7tIo<BIv<0u@A;7I4k}N0*Tq
z?f{mIATV7jA5ostymC8twWZGm4J)MIo(Kumm21+6y|hv82W9R!^_f|#aOqc1R~p7(
z>6*$jhpWH>3CX55$p-X|+J{Z`(-`ap&#Z`uH#ndbN@_A$V(85*OA6EdkcsTcTX=sw
za;=n7I?BGcyBgL-{bh$7&(^}Cf2Z;(1$!n<EjIAgfY}``5)X9WD_*GHi_0yTxN~%Y
z+k)yrPs{d?RDmv}4wbKOi_XHSiB~NNv690ljb$p*{*~R9oUL-z!%f`Dw+p~{=X}<j
zsOyJSf|@Z-!3l5{z{DX9Ld`Z`8F3~^n#iWr*ruHJa6(H9T?A_N-2T=9(CEC+|3NCS
zgV)_Mha2F2Usu9c)8&^G`!7me2Q9CV{W`w}CO}2O30b{N_$sPTzb4zC4%TOvp^!XQ
z$<j4Rc>Covhwa3Z=l;)<;Xy2XE~QN0=)JvacPi{>+y2UKo%PB8U|$Sr32fSLk*%w3
zj0aZc32Q+J#Ql}EC~I=aO_@5s^ZRE<(<9(QHhrCGT)IuBSj6h89ZS%$mBk?{eWys_
z+;RLs(4--5N#`|(DzncuLc-&;)Y3}@N>ms-$J3NbZKt=TJTS1?(=1Xw>q%z3L`U~m
zv1umABy~f#G9P3zBb;E&H??tWl7aQwR4$jmW2L0mop{|m@Nq&gq$iv|R@{o|)(cY*
zi7p=_@rMA4MLfoJXm!S}T_W32fb$5PRR8p4tf}Ttp0}VlTlh0ul(9G*7>jkQ1g<@$
znqct<NGJAQy@^d=T?|VLYXKqhxF}LAD&%3F!ZNMT!O=eK7ks|9*xCm_Oyqy$nbKeQ
zs;N}zkO)Y2gQKSbD@UAb$`Y2kW(hpiKDA4yqJeY57f(gTz$eNRFhG9AKenI`G_+<r
zh4j;FH6Ncftb~sx1hQ~%7t-%7zixdp<~}#J-@8tQ0loNzU5I0MDbkoy>E=k1$Y=g;
zh~Yedr!MIzk@|eCGk_*Dlb;7gm%ma7ZrBG6JDSTL=sXS&a2R0q9mrScx*dRsFWmeY
zj4jxE0^PpSY>{}wnG+%sEFZ3Tb>!%Y%5XiB2c8^bW=#%&7B&Df{2v3e)(b)_;A382
zlsP%m{tK<@Iz!6qu}Gt1dw}yAba6&-*;2jw#4@A*mO-zO<$g)8NivU_9y@f`(PW9<
z%(~E3K)3pX4ccxr;}$8@fTOXG9AaHU0%^39_7|1}FRtog5P@>!R$1By9dATQD%ZY$
zM)y8_Kg*j+iq`^&L7TXx47~N7OaoSXJ{;SiQ_`j<oCjX*S4+*xn}i864bS=vqp@N%
z;yeukpufODjz7|?vJ0Fm?)5b;AB%j(F67|}8?*J+P#0BImwL0^(f}i)QXv*0@Fn3s
zqW}q-_m*47^>03hj3XosAf?w%UIH+nvytOJySSJnv87J<UbV0%#3FbpnJ>-dtC#b2
z?D*-k4AfrsXH^|9p<RcmT~Iw-VcI4~*bnxmQ=HefU&rcKA@Y7c=)T!fCG<<=3U~<C
zxEMd^Ly$Lp%{6P0m_C1G-qO5hVVap<SypvFIHdso=OOmbiB*}%jxe5B=H%!4$g)>Y
zZlvX*T8gX03#f}G!M%g3#|;R=xq6vP2&4%Cdl`ivEvH+TQBG7H4f%+rx@2v;_1)AO
zL1X3rNH^V_(z9IAG&S96m4-4BIF$w$uE%?%Gg_az`7C<Ada0ErBnQseAduFg-GFZE
zku@k8=^Q{_lvXK0+~X$nospF@4O8#e2(;KHfHzzam9j*yRodsFkxGg^4#`CHVM{6p
zs9ZlybWpdUxGA+n9dA|wSlcj0kbL+^!Tragx->b_0Uolw<RA3dk#UmaH0b{P-cD#4
zgS6%pgWR>?FJ?{z!uISGzQsk)2!bCz_s3BJzG#tW3AsD^90{o?d1EWtxiGkG2B-#0
zep3WiR6<?98psd(CzAI<Mc2rUY+?mYXl-&*UrN?#(2p3XeeRk=Iw&{hRk<cjksz1S
zj#Y_DDnLZ*Z`ydmTPRIP>=P@I0bpuB2*sonY#ZORQmV6PFAkgAtdRBj-PG7R-11cz
zVODc%^1WJIy!Yyo+Vtm9!%q-=WP@~(mr!uiz5x|U;m54c&Z^0^hukEeVpafB9k2<x
z{p55m5H{PA3~%%vn;z4N3Z)l(A?f}YE^JFyuMg#cm-4z)<K2*AX@Xvh%5V-#57dM)
zAKtj6!Gsni-+`@4s;&np$no}*GV3^iTNi6fGt%tEWJVKM$)^78j};Z^xPH5aMW-PE
zIs2hQge*e?b};~-BN25{5143hr0N>E_u;;2jH_^@$83jF%^#7B0vIY(z9MUa!{4Oe
zQ1L2Ji1qLIFpC9N`i=G3y5Pz3Qy)0K$?s#MJ)8Cvk2q{oV|iz=63OihgopGmf$bvT
z=#)+0uuWq=NXg8|gVPt*xo6$4|K=<G6GX5j*7ITK=Xlizi4)WQ5#}Yo^&Pt9O(($`
z?n+GA6X##~i`-h66mrVfZj1tKnQdnku&k-s8Wh%G#+ZU#4-nv?=W$E<!TUsm(Y!Dw
zv;((-i~sw3b9+J;r+t#z2kkta_TDocmh#6+{dps9v6Bm$#(FpeuWjb0UVW+S^4YVq
zVkoaA&Re9G$%D>!hiUOe0UXrvsCDGq!WDUW`ThOdsMhLCqS@w@Yf27gKsX)S`k9HW
zdaRH<dv0Akw<8FTT@EXeCxB>6Y&?K3!G>leX=sw8=Nfh>3@Pt}eaVVXYdLvsGbvUd
z=4=`Zn_!pSkqlyc%i1Xp6S88|;DUFEMy-w6sZV9~qE}}lymvv!N7~ewX&A1pTtPb_
zGYa<$1tRA@x2R;zJhiB#o>G@^*ALxTIsv%;^8@O5w%_9G1p0t>*C_$a(ZM0?C$7;&
zX};wge(JN0lQm;vxo-9;1(|&2Ww$T0{9PH247_bPg#dM6TUb?ct;m*=(WcR5|Cm8z
zbX9-l)P>(EhTJ4sHTR<=o7<l$decPrgZN|p96|1jN5*GW?IR}qPQ|G4q<V5}pFCRG
zIjIa?`mRW6YL#nIpV^g(WPu?RI&>A&TCenS^2yEga=DO-taCF%Nua0Z?}_DQ7pXsz
zJv+iQnv6HSPX&|xZhyHUUU9hB{)CO4Ug;|DEbp|0`p<@L06QN6Aim5BU-_Z<J$fdV
zn1Z7Re0b#rj>I_HmoHtT&a-E0HfKl5<xkhwSu33)g-D$iwd%4phK8!V+$Y+3%Aa{N
z@oJ7{#PBa2*Xa)z#p~4}tT_^auT^=5hZMq#-<pD|Q8!`FOefvO4yM{c)PveGK9lOt
zs-A{@p_WcXJ<WJavB<Feguf(-=ER8%LRvafK_TGu4yY^~aH2k5Sn_fiw+}a0wjSf>
zJ!t-^fIZHxR_U5TFTf+MhRL^P8S?VdPPY){aJJmLZr*iqft26$XPw6U*ol+@Anho%
z-F#lw3syd((&(u-RW!B9>Qe}#nTP%7Or2gOu%g8v@YKIZ8Csm;ikCe8u~-1to_oFb
z$<HWfkDM>fG88JDBUFs9e~&`vmf^<YQ&M=qjkqN2oaA(HNqX=XT@1Q%WwV#_8Qyz;
z({}T%ZNQ&Fg70bxP0{PG`MX2%3A;Ls+lwKH%`!+$260bu28y}rPk-PdbJ9Ng>aD(w
zrCW}f3)eFJaeha9Mmxh}<sYRUX|etI0*u)Ax!sWNYJV3*WxZ>wn~~YWIgewwZ-*{4
zypVTGs)=8AN_&M?c==oi=`@l8h9+7UXATt)DOWD37<7JC#5I)nVP6*7Z%WscSLG_G
zZ>aPPup}FZ@J{qN#z<MX)I|I=OI~hZRc<#-x{t1^1SeH`ce%YBLV?4OBDA!<wC$k8
z`d9Pz(KBV&>H#v-`w7-HXyRqnH?qH6>z%S`-~jI!O}z2=Ym2NVp&m!PLmHeaB8p}R
zCG@(GsJU4_VGG|($jfQ=_c`+MIXXBv+)u7)vDXI(qvDRqfMYXzLTWj4!bfJ!DTG%@
z6*n9WbZzq}G)B*@DTei)_%iFgfA0MrR7ZW_ac0+?jfuKG&~l97%r?-pjP8Zgrw!Wp
z$<~I$g=`_r_jjh&ErXiF;oQPJ$TDOBAhvxx^^D55X+5lHcSI*ZCLSR>!TWAey?4!L
zW<ZA$wt|P>wq{DV&P-Px^^~)>c!uL4U3|htTSKXh6PJBOR~|+m!ai^9Qk6TeaIliq
zx@!FA^iSLSl8Ig``p*ZQp-LT0KR4_~T8K8B6rR}Yh&-hqHGWVq(Oo}{PV<xyRUB0L
zl$uy<s+I9cqS#?*bu8o2G!%7fC1x2ZzXVdhD$5xL7j)FSx>4{Jr1{;{!v|T9LAY_e
z02>0{rH1XGVsXl(M{rO+NTZ2ISF$$=lg|TFx+a?_000dZa~j%<rhR6`DQGK<&~*CQ
zCLre*grDDXBy<5SoE!nPuOvsxCvEkdvWckcA_U%idgS6TkJ;Rr#n*1p{f<_-&J{e(
z53scDrVW#;^SNDBW#_)$!dFqF$GBl=YLkDKl|srv`v1TC@g=i7FFV<@^zB+1RXDrF
z<RJkhjd>xAj~>ap+l_Ou#+7HsROhtq*hK-gT_26JVVsmAKenYL_jEn%U-Ios)LBdk
z7pQYsPps>{-y@DlVBMS_#Iv4DAF)o6Q|y_{H-!=Ru)lxV4=V1Yd>8XbP)a$;qjRaf
zSK))N@Gd6L8R1~onQwGPyT`aTHZjj!sIplglvHl->iT}x|EStem2vZ?%#PgcCHh-c
zKRYw5p}Z#f-;l)|a6qUIpjWBI0CcE4oFU~#SnYI4N%LY3O*XvOSlsDlEk-aFcOK)n
zt2J%M7QB_D#n{<!quj_z62EB?-ibdY7)4EfT!iv?hpN>PrxxQrD$W@eD`RLdZbNQV
z#^O-qGP2Xr(ck}oe>t(3&~}?sn$h<K4wQ<q4N%aY`t!IY>^gCO%jCr5*ul5CY}suW
zJjQnKxy$^^Vl$0`l5-&oAuHLmN@lpfgcSs!$|q_puTx{5Lq<y~FLh%7^GJ6ND!Y*x
z{=%9!YSFc9h7XYOmY<Sc^CLq@C*DV`)(gS%d9%KcdiSqQX$OR(Sq$iv+Emn^M`B*F
z+Pcseed|o7tM~yZBBaXwNbd>DCwZhZJ*~!zZlU*mp9Rg<ht7L$ElXz09$Ozdn=QH8
z>@D6wPvn=+U-;W4iMX{VtD!y^<5bphUsgft44UrbceGxAaR|N_2@X81bKaz>UuuW!
z^mD(~41Ir0Pc=d0pVM<OER5N7wjY_J1c;k}b7c5iz^=KTe@UyQdk@CmtT5|Lj!*)6
z+CCG=vWnx(%nS}lx0+jR;K>$X9Ln?><ti78P~?DZnyxCZ%0$2MTHN1Iv&xir(amRK
zzR4d3e>us5PXei}IPt$Cf-Klc^wsw-n;E!!n)Lzjr6M-ROTaJ&q&cRt2dCl78bpIv
z8F)|c6u~<7{q@_w98M?bp@YoAvj7ICF6Qdf2Y0)ydl~2O^OLnzjc_2$84kxEqBrN(
zZUiMhFeh}+ccm(#Gl0&_v4<LCJyeBE{hS?D6CT|Mae^_zW%J=iR{xq_9(n<49x`b^
z0#)mGVUDk(e**}l;rWKm;QkAtsJp*fh8=yE2aBKO{Af@6kck3EMc6ak>UI?k?zl#}
zE_hpSKL(&KJQ`kGzA<1T>%v|nXxjHoGhulujhkt#DePG|&lM1*XQf+ZaxUvsH&Ni}
zl=#d@L<r^Y0l;LW`cqZmDic0q?H!{o<A@+)WFUc%3T{fMqGXJ*S<6nwBr${TGy?-6
z-j~p4^m|0-g6z-hM!p-55!UaX(=!MI%~~u{_NcNMF4EDZ%cW^93C_*%MIiN%JkwJK
z-8`w7?zH0W<xzB>7|Iit^(CIBnn!!Aq1+&v31E=e&ov-2c|4MrrVjd}>*7c2c+%@6
z9O3RfFq%6ooe#(8FnvrAg}7gl`FG(dB@~N~fFfc*T}NANrp}`9&^(g_*Vd531x@b3
zN1HVAm?YiDysOYC*Nqme>IBqB3cpFIUl4e!VVsTxz{5tg?KV+U8za+Cf`7o@@@${&
zshS^&!{?`sy^qh5ANfpbv=L(`f$)-O5`1x%6&WWXxzxO3hB#kyrk@apHvcyy&_6Do
z+yyUx==W71XL@EgsWg|x`*zc<$$B2NTIB1%vEOR6n44}*_l03J-6iTQBVrW9I@kU*
zQ4>AAzca8vXFM*N>$OMcW=LQ-ynE{V(T6pA#fn63Mv;c>Ry@G*`N2@+)ZjhnQCWag
zr-gFaazFm0zJ{|~gYtTkz2WTp?68OodnNA@MbrL#q1ygSrc3VLwQ_}vzWoDlJ@SCp
zuTQ5i(A%;{8~LS;>7UkKsI3$w7dm;o#Wv*hPU#PtzQsz_pHn-U^117M|0PAgNO0@N
zN@%Ao2z4e+C`<IIQ%gzqV59ZW-9v>wd+i{aT4$MUEZY)Nney5-&3<BPj;E;^HUzN{
z10SaACI;>fM%<dViy%4_paBD9`aIC5T$R_IQ$A8yhd?JeIH@#9to8%BIgYKU3|Jw1
zFg-`YX_(a$zTfL26r0<d+tcIKF7pxP=5PqW>5><fdL&98O!k@e%xb^BI-eA8I_h0h
zvt{h={V=r`u+>$$hnWbXa%le-{lGQ?00D&bky;>*MWLW0&9q16p18fPy}zsxOO%Cm
z*#va29uij6Q864wi#U35Q^2{$!;I)UhcC70UElrGNJm#jAWu1$r>BeAzM&RV8UJ@A
z!y4#YZCJ|LHF3aP31b&1=Hc(0H?Tsv=py5uyr1A_eLz7K^E+BRr*B9gr#GT`6E@Wx
zJG*(hzO|{!Rgm}XW%RcPwwallnNg4)js8_G=TsAnW2zg6=U1<ppH-BJ&%q^0vkz<u
z8L&*Zy|)tJ?+lW>VS8^u#SoC8xqHwdoGT(Z44zUz5z1!dkyY_bdn=YdaqJQN9k*lk
zY5sMXa0uf_<B$=TreP%D)KkER6V)F440InB!=Gt;)f=!Hi;CXzCl`0_oV6uzAydmK
z<P<cfh*^}LN<Z)`oXcyu*fjUYvCdQn67o-RxWm(ywM1@M@^+n5@f5H<gGO7*qGg#$
z@uAsPvN)sesbL4a!T;%w6~uKfyJMQ3<!d2PVFFtL^Y?288{<*IG!f6nS`dz}m$lTV
zwFA2ke`i`l;&)>3>whj}eyVnkg8uY|MbjiD;qvyA@49I`3^XpRp|~2&vYqio*mR@8
zCc024iDkg+<@WC@E3ZCC3d_b8g=tRJ|Gg5}(eb8pqL)^px6cgHJIP*zO7(mUlNrFi
z_iYSKE_ju!3=<B7>N*k%BzlSO-RyDmIITB)1uBv7*2~Le=>SvLW2?r(;20!T1#7fG
zVh0=V<uolyhk1c^<FXHBqp2w#2lFQ*>ubE_-Geey=i<kzpFXJ9SO&N_Hswg0Q@mH-
zA$OT+bQUlUmitnk0bt4RfF`zv<B*Z(j$evOiwIS~#lk{>;j5dA0=Gk2T5_I#+Ln>E
zf@q=;hW>IHz05*rSs*H1oZYH4wfEnx?3)k5&R;cVuLbRq3Fl7k#@JhMbnOyE&18{K
z0J<N(YgLNw9SgP~c8%BDYf<uJu~twV{i{;~_h_d>@!JSEX0k}qHhhp3zeY)^ZB~_~
zhSy(+p<;VamF9bc?lvjesgkx&TBRDpZ>=bT@+h5>Ob;kMgIjxiH3?Hto1bBwJe(#`
z{9M8$31sBR6b$^KTin4sIUVE70Z+b<jt&#5+i5vTLF)OPDL<5rP~^`9n?*BqqPx8N
z4N)mDD$;2-65yr|uoQIdC`s)#JaBWk#G>%Y(zSVUU(^t~Xk!i#;E~M-Q;oujmQQ19
z6PX0s0RBX#KG)5V$a}}+f4J{IpztE6AP2m06)GL)5zGmjv-SCKSJ#)lf3RhDF%$f8
zfNbz){)Vh+*RJPKRe15`3+k<cFaP(Yu<}jZx=U!hT+8A(Vy$Md^~d79zUMm>^x9d!
z`fX@2dR|x>xjDgm7}9HZ<3g3@;s)QRz#IC1HgZt0Yq=Q7wJ73ndoWH?=;+diI$>$@
z8$HUYeY)&$?jT{BHWX?gu_L}rL;@w~&_<fCUj}g%v*9yYdVh>DkKb5c?PKW4v4y&~
zKLXeBc)*eYuSS4=LUWQiTrji@<p?E1ghE{TSEm|iNzIBh_$5nUdjpoRW)pE@VNZcy
z{9Bs{xag}0_;WP1wB}+2?Lkk7+LpI|W)|rrTW`62AEb}}opn89A7oU<M*W#{co|r)
zc6kg#-OWe%DaE$_SiK*sQvhGk+m-JRQLTD3GdVlf>=umQio9Keh7$-sfBxFuVjyY%
z#z}+e^I5qBj`m<s;I0LpSKgo6F<6kwwBe?JHO#(d^Z$T(zf|+)$fQWZdRH1a9@d>f
zwoeMHVc*3bnl;?8Hq_~&`N@PYmLo(nyM8Wn<;UI`Dxk<OnNtd2v3}f6HIfzIA`<W_
z(R6bsUjw4+2u=F<@lDwGs`lr<6Sd~LX$RRAm97ZH+(6uSwlOb`HspKW_VgBURvDc%
z7^eufB1rWKLP{5>s<snWga9^!7*B=?G>&oPBbvIgss_#N?A}U{al9od<BgQY^tojT
zu0zSY<GWn4u2XWw71(A~rm|^H5_>m*607R(V_w97z|%aym&uYo&c%z8xYNXu#KKF9
za%{l77IN2~F^n=%r@<76yX>X*y2Z#MJYxa+o!WxG$urcpI8(|`5el!PqXXw{VAo(Z
zKg`UC_t|h@WZGcmN=6HFa<4lSPir_rN`>pOJKdRIvYo7C4T0aNr}LHztZUNrXif6}
z^zOy7;-)h(QGoGmqvEJox7w04uZ!_<tz_VsAnftm=gwDF!TWxnHA@`(57`iss=Am|
zlEfyjrDbGuV1wbKxoz3duOVu<n7C{dy4D0Azmm+Q#h;vC!IO&qe)N*wh)eP`!l3I~
z4~1kEW=2i?29q^a&l7%Fatd^RO1TZ^)wfO#;mtJ2EQAR;W;*+b2`7em?WT4|oBcnw
z-aH)Y_kAC3qaqTa>||e)j4edSk~PLMmSl^WAvB~yg$mjCy|OdISmwppv+v3>n6a-h
z$dY|~UVXmL@ArJ(@8^3Q4#(k`zua@(*L_~+d0y8gR;d@3JS)ftlS}!F2KW(a2;E>a
z@yaM(NGtc12Q^%u7DxDT?IqZ_NBxdkqk`NiI(<3OkoXD}I`G>sJCssMVFsQGAe!Dp
zL&~$@iEH4WOG@B4z-BjSkfjhZhS#ec3YJ0fz8OXaIcCDNe|BEDjSd5cUw7C^5LCr)
z1#oVL)b)@Hb6`nognhS-BOVbgoJl1Gk%b*2uP%{(*4=6S-!g#1!tKn$=mB-dxQ-4q
z?4epGwe_o>qxb}?!&I#IdbaEW4?ov89FtE$@2mH1DVd;)PLkoQ^C{l=wROHvA8+q>
zCLkI~Mfd)$4_rdk<B)$FSduj?S)AbShg2h*t0V#q8g6{0Yn&<VN`Ct8R%jS{gC$A-
z-H)?Rol`E6#zaXk>M#Mbk;M-o<~#)VQkJ<nK$5mdQr4B<BXbmQ<^d^^_X2&;tifh7
zes5fJFahpqFXEBq<GRCh7F+O`K@7R_21Bie-;Jq@QL%Bixh|mtF?`@MuQ@L41$YY^
zi7QstDkXUGwpOn;zt2E3dTw!y7}p-RL{u9g=neCeK7pMS*+@}D%s#L&`<{#Ik8e$b
z(}=`?4cqb(=dbb6YPZvzV-lcTXV;aJ7I6Of{M=%SX-I#k8Uh1ym*uO^mWVEmUiZ#<
zSBXT4{C0{xE*Q_4lx6kX|GSp`zkb*pyMHw!8T;ojk~vdFHs&r>N9aj85sTrOk0Q^z
zg=Q6hj4=UH^-MnE)Fg{G8s1|2nCs5(xnyso$~MV6EIx^r;Abdp)_~vZhuQ}|;t)bI
zB#CTvljgE$Z10Z8YDgqIRo9KQ|IFr(->0bzX8;Y%=SD%}ADp$8$+_pbsEgHu7vtLf
zP$*08P+O}Ss-?pYa~)|6TgfSdgYZ8quH9<}QrqY5Y0^G=Oz^DMyQeIp^DE?1Iw$y%
zR**a0hYg^noTjmoa#U|Nr_331tS(Gn8La@7uZ9Y%VXVSLZF}VHt9N8SQ;uuoFWr`S
zA!Vy0U!S-q%cXgP4|&{>%l0z%L52t3a)rD!QRURO^SZ)~xNf#vSDW_W{Q2zLSA4(8
z&#8?-847TY1C-Kr9FH2AHjO~bW~XmXHN8(~Ao7o?>Ayw%&PVo8URUmfD&2cA=P!gU
zri+H2oTxAMPt{$SmTQPMjF8%3d~`JNKoFU$WT(*5m9Z%YFxX6g)f^4?a>La76a@#(
z3m$%k9|PE+)N-2HtkqAmym1{`IvxjG+y?<t!Lhnom`LbUG8jV$fmkyl)z%x?r<LzA
zKeC$7O8*q5schwb-XQyArn&I@6L6jm@mJjQ=nF+@Inw+^xAg3hv4f~EK9}(*R9lDS
z7>N6oKai`OhumGgFdg*ehGHJ6t(ZKho@+Hf{kjU>6s!^4OLk<IPN`zRjo9*l#Fc|g
z3wmNC*_@T??0z6p&{nT?A-cjg2&?N7KC3JhWnX_g83xxHxf=!VDGyu5W-1?MeURIS
zaK(42+%Wi~^M~fDm-p@qBzcTfmk&230gAASm6f%nW&RW!^7C~mM4M-=#Z&K^Mla#V
zz-^`f1!{e~S*9`8c*%#{Ox|UxUGfQC7+)U*b^G;s*<J@Vx08rBN6i^N=XM`0Ulu6W
zz^E7X`Gom)2fWq@=A3RmCnNWst^21iSfOQE9Gz<QLr$CH`oRH<UwuD{A!pl6q5#$Q
zH+@)Oq#^jhmYGfZLk97$I<cjwN69ax6D!3=OD<r<5HZ&=F})Qa`KQqV0~J^B>4Inz
ztM>8bgkwF6LN*<kOh!XkLGpO&7Zm$z4cAy=m~d3)A-7f{RA;m`uC=5bgVyT30M2Hm
zYrLIolS1GW;Hdz)!T3UzC%MlGFOfn5k^Bg$eU=XZ7UwjkF-hAo_Ef<qC8wOC2HX%P
z_;|B+`g>iSRa&*Ta?Qrqi)*7L4qL91A|K}#hr>;R-`xHN#DBp+<L-fI^|q2vmn<OF
zIox@C!=c_~*YpUNu3kU)FQJdt(-CG;s(aUcAd&Lg671-zL|x%VelDhQE5Wwgd<Qln
zjL2O^3%tC?2r$Sy-sq$15sriRwKBppOY9{QHK+K#;aV?P#Z;yV_d}vfhq5tR>9XB`
zKtF^lt3`@UYOEbRO!Ik}{%2z^(lL@r&q3o`JX+VOIr-6uBfa!BylZ@Hm=mawA*nk=
zE$p6K$)xM1+#}$!kZ(xZUHn2YyUXM6V}>{d-a};CAn|pSXsWq4d>%?#tU55z%lJ<0
zdG>NQ={6JoU7#}$A^n|*T13p<BzPadF69335{VZja_4D0@>P9YYrNUiLKbFQaY?8B
zzVpk<OuJLNY5?i;48#GX!yV3F@>qYjt5Y#O*r1<IeUzUi9AL=s)n`NPvswdx1u-<I
z3-y@G?x@ZGO{;nTMz0P8$wJVst(jVdn~a3C-gv0CgQ?u|t<Rb&`fe+Y-H9CLA+wQ=
zHWsj0_QT4@yW`AwgAUy~>DDZN>PJO>#H>;om%FqBqnShDX*~sexO)0+4%iYYYYVbt
zL=TSh+|8Blh#I8+6gcs;Z~B)Rv1lXZb}L<_JB^1Z^6@n|A}Yn0!?d>)CF%?=6(oGu
ziZ0djVjFgMr|R%jdruIvm+msBmr2)xVHs+pKBM}NOK;-&eLn{;P9#4IN78+!E6J~<
zS<#*|8>6HbaGSjesF*1Rv1;U#w7;h4{{XpBJx3B0$^+YiH&Lmi)Tf#7=vaPa<8yfc
z=PVRgDEcV=!Qn$^%hz`&C*$m3c$OZQJmehN`*9w2d|U9^@r3(qv0V=qlj@|vVXKgQ
zrikBO;BheGxaG#<ehzwD)?-GKf7=A^uqK(zFcox?=$?noVy95UZW-P^kNs5W(9%VR
zh$kE0l>>4RxpOZ)&4?VrU`63;Yk^{OCTf0TDSag79^`IZgK%M&hyzA{_8$)T00u(H
z@6Ee9@+3$Uz<oAUF@fP?imY>GQnt!1SvpyM-DZB1ALd|pK7OCY-mb5-2b5s?Fw(U9
z@^e{KAehw<B4ts^T`Q&-27S$BH7O4aads{d%pi2XC3p<^N;$I_+f{_028cSO6e<_(
zu|P)!u9c|VU>9Mm)?(i@lLAeYBOma_$n2W9+q|Ej4n?8EPFJ!+u#CZV=UeXox?f_s
zsZcEsStA1bb(jdHP0hA+MfL#S+WEcPy7oPjI3sLk@Tdz(SpB!z60v0AZ@0Nu`Rema
zL6}a9fAGnMv7>!xhKk#&6Rm)D>Q{*qd+hYyEe_Gt$~A#Nk1Iv^MEBY_#mkJ|mx5LY
z$wVRRx}@-g*>xrW_q_k2+)zY9h?Uf4-31six$<D7t}hzUs)DM<Ot?i8=Wf$?)LnKz
zZ)<{YkbI!aW}4bo`sJ)U*CVG6jmZ|TQi5c-Nfd(bSz66VPfr=2-uEiTj?v0HqnOC*
zofTr%OyyLi`!`5ZKK!|E&2Pa*0xEOa{`Lji`%@2x+TO)|qA`+W?aQGFc(V+_+kl-+
zN#!`l;Sf+f?&fn|J?<aG@I4(^1zSAeIq;a#X7_&iE3<=;I8Q_{12A=ix$gqV3Vl2{
zc;$`4Rk)2*JCN;2kb6LF;nscqBcv&|4(#f9D#Ev7&j4g*y9lI|vj0XJ{F~eT_L)Hn
zHn|nnBBo~4K&SN%JDs&%`s=sNi_c}H+x$mU(9d$VpKs|-rI)OwZk7!NBGY=|D|UKo
zH?Qve<lp6b{5nI}zqqV6=2CWv(!rBeH!WBUKT>6;kYA7Z{{ES>+)ppaT!&t#PQ86M
zA~ST$jF;x;WT)ClawRxBVmuiUj&ublm{p7E##GH!ej5SvA`#cIbVPJPwT*g#7n#~M
zc}4fPj2{0h%uRrw5fE>^p?Wzp78%(-YKEmO3B=^7>$M<^GDJ7fMB^RGC}SG5@Z8?Z
z;J0Cp!{7u2m$>%$MBNLnjJB+;Tc4zyStptOZFpxOS`*qwPDpCl+Xir{A?JsQhWj?K
z!`{=$i4+R0`0+*GNjU-dYSHrA(&($5r1YEJR~&CAp8aq}^GtIi_t|JJ!A}?X#7ot*
z!8YVgo?6*P3EIOBH{jNFaC`7AHoxbAsyXOLWD-4dRCDY1C%0=QhICM-ZbL8>(mZK5
zwVtzAZa4ikc_PJpXlQ(Bd@{nYJ!NsMCdnkuoZ+Xk-{k4aL#dBfeczl62II3+sI8bM
zZ|^JQ@oEmR%oU((QyRU>O1(z3=j7$OwCz<daR<`<pI^AKiW#P7u@zoN%u04SCmS}e
z&(GefzxaLfW8>g~{{cv&W2&aSUIy(RvXQ=Wc#;|+Tt?QIIH2W`Rq!k7bA0QRGM^ML
zo=Q8u%Y{C~`a7`c?`i~;o6yl@c$l<<^Xg`SJT-AJPj-*KDUNQIZbioc*#&duRQOWI
zehd4@pT_PDr3iS{K7r)q88e-|1CKH%I}L<f_e4m+id31HK{G>@9lREc3NB|98s^gh
z%R;@H^{}B9JQl<2HtRQVg=E?A)T$AZP-)nCE;~zHtq}q1AeZBy?aj9NlUCMdRj9Rt
zc^!eG^dv<ezlFJ#v6UtGIeQRC%}&nen3ti9fP8wqFfw}~c?o6cX>HZW^emy`#?#t0
z*;>q6=6p@nacA4E(n$p5zYZ-h61vt^FsF%hxbtJ_q?&X0>(KZ7(<7M^#?L;qy~iv^
z2z7%I*Kvx9E|HoHBf$I72yL}^TpEjWA#Zx#`*Iq3(-6;Z<rEiUe1m+)uho>^33ock
zOpK&`a@0NWA{^5hQ(>6JQ_R`hrxSxCqID)*7(^0Kl5Nn3KjADObZ`mR@$?AOc!()h
z@7-11gc@9er7&){lrZ5nM=T`ZE#E)5*ku4XEea2`eo=R2To#8MKVsN~r$PXs(+B|c
za@~2lB2lw!u>2f2Upwry+FcvFAo{8nfkHX>tP!(AKlY`|dcEA43*MqOT}(JUuA}*#
z74;k@VEj?*C$07f=$FBu)mrJusJ0~}ic@THXtj9qOF7^F7gd3g7n2X=-Kp>9xW;|@
zlXLRDcs2EC&9qLuc7;=A`nZM#Sd+)-dzjYWBJfOR{)VqhS0rGqDq`NmlLbiHOF6<w
zlcL0stt?)oRiY==r%fZbZ^avbTrNr!htDpZ+~9x0MT8v-BBBPFa;4yLhCez5%(trM
z%6DYj&5=hd=4sl)V7)ok#fLCVVq2W_dNIF|U5*iNE*9c|H*T3NMmn^7CK?^P^H?KM
z;TWAAP-7`Mho>uduqA`UlRKS=I(-EX!tM@0SJHGD)Y725V#1z?X^NWj*klYwOjInj
z7NYDwL)`Zikm&7BRq%INu$XjI46Lv2NGZ}90r=pBxDH@*4`N###WJ@nA1(ca9EcpS
z*ses1y}9>$>(qeZ|J=X7XLd8>?^zAzxU!$gjAdbr%3eC1kvV-aLsH(YL~nCa3@V^b
zA$5vIz;$4E`H<z!I1*$x`ChUQy>u)qH|cmpnt7E6o|K=WA?@?`?7m@6(1+`r%d+X6
zt+%OYa+^e9*S`UY4uv{Gu<RG&2x{(Xzd9u7^fsL0rw20c#s+%~Uk^5CSePn^OqQ;S
zY;0{1q+kFAGxLR3_$SX_4oMj>*22&S-w2_mLb(X5JY7J`!6O#n@ndq$6APaq^i`q!
z#6m2h`Vl0{Dfy1aczmTUJR%?wlVV3srItd}Vksi2rw(ocR0dS==&vc;)$sAP`JlqB
zw!Vye?(2h5O}TE?wWVXLUQ&DY6={F00=#Tt320=2j%y06`X|U;(q3W0p<dswuEZ2&
zE=cNPnUel*{sPah{aC{F9cq(l{Zz0x&gd3T>3aL=y8g*(J4r&nOWT`X!<Q!=rx>`J
zW&psB#IQ{JMYROoA)e5?)G)?oQF>Dn50bH2f%t6Q$XVo4nI6S}q6E@rS;HCr1b4sX
zl|5$5Wv+(bT%ZgY15WyzF>|J2=#`ewaj1P{qHVe`x@@`Ng8A7|7+u>gP1Lcrd!I+s
z2>rNt8zl7NOhpc1XWC2Ayw5P6?Uspfc(&$vYb;vVsFEkB9XF514ITDsEA2wnp4hFr
z^ZAI7UHS4gyzp47gi@5dhM0#WEE=bB(mIJV-9AVC#py6ff5ci2;_tdM7gtv1>6xo<
zWwl!LBC4GAi*<o3l&E1#g4BCSH8Rm0-oR3pcfkfzBjUG~ci>9>*Ma}KoOGMFH}8KU
z1->94A5DaRYG0WESLdDBF2?0%MAjYqW@9ThGP-|%EBD<CTwSo4ZW2eNcn!+e$6SJO
zC6t>ur~Ed_LMpX^_)u8X{8Fs!inY{BKhy;Ybl7Y#;>s-p6(X3kO$KikCgzw>X{V^o
z6>OqMwHu+VR$1@>#1Q-JB2k#a?z28on9(IN)EF6C>r?Ad*)u#NN)q${RR97%$X>iB
zK#eNmSmua_+=gU?By)v)86hKPkUN1Mn4d$&V69ApMK1}FxDaPG(I&l4!`l6yF<fU3
zo|~bLPrx>UO-$i%cwbNIJz39R4c&2}o-$>kp}*tzCc}WFGH{5I?gZrW6vS5HnKd%4
zT)hc@zZ%;@w`*8tbCKa6C#6c*X!t}*9uza+b=z*w-N_i;nmi5g$Hcy3>-A;TUKO``
zi))`JD>mTTy|qVN7k<uXY*88kdD<}CNsb0_VOTphHiVkihrUeK^t)WccU6V&Yw6!Z
zMKWLJD}>j2U1!hcVo~3A*Zf(~*=@lXSjg#9*cA=ARxZ!`Ec1nlu-5no7yI#_KUB`O
zjpAApRf!7`l5swOB(8Atl|4g$CC__nCgBKVRE`)gT7MuO9_e5uR$4}z<mr=E;{W-6
zGZgf>s3FsHbVVYqv_>?XVY(wsGmunPzhQ9#XLTt$>@sPuj!saP@0S8OK0cw>2Jf-a
zK9-+!%d|IDQQ3c^v1fk1#r5UzT|UGBzC0ih`M43#9jThLQqUwq`zR=NV2G*8HKy|f
za<7nt(Fpy2XSMI=!3k2vF%`FtBb*LIPD!V`({sP;>Mv?5(oXrFNjt1%N^0~HDXy+H
z0>T=Z1c%sjfBsVKj?YGN`<Q4Fsf_)COUq4cFTd6IZTv3#{T7^30v$e^y>?@Vc<mX6
zOH#4$h1N*xKCm=Kd|96-0R17jFvNnXi5yaoD}f5@w|W`VRp-oRBjr<oS#{S(vrL|@
zzVtE)-z9rk1I5nR13V6x?M-Q|-UVETQI-n^g`Fvku}ba^+j0`o8gn^z7k99`Ba`DC
zqvo27)rCisg`?P?GM1tX)%EI+Cb>@Q6=l&x+{CiQZvSQ&Yi?1hoabKiyYp>MeW^7e
zKP>3#W~#sa1r}~3HDn65Hti-br1(YDLngYkg|6ZTHvBC4pI1^<JMZG^WY~jcru$fi
z7Ez5)nr?mmhZprf`);TG?!?4RzBByYQmb%Okch2)I63@$xHckCEK-xiVw<wE*bgCa
zJlGG!RI2eYN+IK-d2`JMH&_Hl2Bcm(JFp#^wqe`C&seCGYcGJqpCO-M7YvjAld_oP
zw1_NezuR8f+da)L>4?E+``{Q=NLof!#6WEa5OU~pzsISzz3d>XMLMW-%1c;DIoV1H
zWh2r=`bz<$Ov^MFD?^oD|Lk2*<>F|2HMe|Zk)S!IdzrLp8dbXe`|QPckKYD0b#89r
zMVJ@-<UNnpzv7dUwnB&9(C8xZ-KJ`;W0ILE%rc>rofnZpk4s6jCEcw61MFN=2eGPi
zHiyL?%JYwx5|8AlY+ilaId`qh?RSmei{bj4`^{31`@6gz<m-=(nwIxui!F{!;D)P~
z<bb6#aHv*Qw5IT@&@zG+Np=g;liM^*mY>gFoxthkJ>+QOkU8gYJr!zejf@4>%iy|n
zrNRWZwkmrDd!$c1ro-t6%Z<IZMYxX(RNy@!)@ks=v7TReN1x16s%2(KEapp|x)-fD
zF9H*1Id#x!2h$qY2u{&7U+f&}=;e1N=44?Sg0?P@;nAV7Nb`}LxOT%4)yPW|YJ^tu
z2hZy@#uapE$Kw-BT#Khnac@SVvgghSezr|NL(cap$9>Av$qO_YrAVE>aEsp=Y#L&Z
z#j<;aBP}2H$o~*WkknSOQV8?lNB@oH?R9*n=qvVRpcysNXYO1pi8#$T9namq_1K@Q
z5bn>ey;^;~_{nQ(*u4=U5>mPi`13_K2k}moB0s+EdVr2j1=Sf#b0SaH-L0DykTLmp
zK>>PMt6k&nD_*w@xSe6$WkdK#6L838cH?JIGe-9ts8kiGwYOeCq9MHzVH?{W+7G-x
z>klnRchzy=88Vc6%XM8M3y0cS-qLtF*FM9wBX}G{o<g+E{QV+J0bzKIT;)vk@$wZ?
z`H5`Kb80!*AkVE81+ENK=$raHC5`lNj6xioK}gH1K5(;3=yL=@)t1a+F`K;4Uc<$Q
zpf98&2tEu=4NaHw*O3k$%E>}L?gXj`kS?z>T>JgkRbEn3h(hv>r|i<hno99kfB8SV
z{NErRA#~Mp2KeBz31r4Ljx0{f${yRCESwf`CUGQmpYXr6f^7$;+P(f=SMhy#%vy-_
z73>Gdbk~p8p1bFKGf$iJJX1Sj3s(j8aX$QIaq-YpV?))Ot$JKAmXbGLtv4G3{n3eL
zu9T#JX^(=&1>t;ky@emik?})XEHYV6zE&RGZxo1ou|?wj75qNp4EEShUzlCO)Ap$5
zL^GVp(JphwD;n>}2FbQ5lL<9UayAv}+a;kLBY&MJPK9v61rmY4(+2v7#|!Vsq+dK6
zOL%MatNMye0&{PVP2G#VnIx&U;@;kL1^-tIF5VbY(5b)U&kj~QGg}fUPz@99R~;8e
zXk&_M*=<SXP(@c5HK!I8?7tH@chgn_zw0}rWhOz#ap5x-Y6O1&ZECxL(?1n;qO~W#
z&+kFr29sXZA;l3Dbxb~oyoif7yMBy-`lZrUR34Yjthw;9Wo|!kr;UlnpyLsY`W-;~
zdicst@n<lN%X(GQ#uAjA3#H57XV?}NcmS3Gskz9Qs;BI0DP}fkUCrs_$mZ?1R`xh5
zVq}eOf4RICSwpUnVHDnor=lW^s`h`t<ANe_K14$N0uj`rk-CkLOBQ6|e4=(ILH?aA
zj0wc4rCdpFB6Z)086VZ!tELEc2ZZ95Npmz5Y7~3r&-9O&N3V`2!m?RqA~KYHW}2Fm
z_Y0et`!cGp>4h``T3-m+4q^mq68~2cw5%oyk3&8Ms!YAD$Nf+=BdJ)VWiIx=YaY7N
zn;f8Abmupd;K^5;^K{vRH{RCyyy@E&@ZS=d{oz92Aamxk>rs+f=$qoY;=eJ~2sq~&
zjueChVia^P(8X1bhLnd5pXK`Y)*!o3EhGsrB+7q$IDEz0@@5k?+-`)=n#iL+@GTL}
zDiSrU1xxFzx5_BhT@E#oY$6EKjOXXL_Eg0p1(+-5n_Ug75gy<Y_djJ<2oCtX?S4SB
zw@7C=X9*^t?pvCL>a;GkaN>p@&1T=}1GzkB4@j|^*SU)q-ItW2@qh~9Lm-CXe(EU$
zG~Eei%u~Jo8hCEDTx3~$$A9TZi&XAan|r=4JH)XdvW%Oxjk@8@tAkyGRex4)OCeJF
z%XJ&QZh0bpVk<@i#ViQqou8#P)>;2NDi79sKY~d|DWZMV6g4$NXSjH3$4|E>U%faY
z|Mq)a1+S;Z)W82aFBY^~dr(Jdv!#?~$7=em0=f5QAtD&lX5?8{X=GJ;=?wE7ysd;T
z!ev8?#BAESZsg7OdhF5*O=4@6U0feFY|F*Y!Grr5N`9CSJ5XW`BpX>kFgkkK$cG<q
z;&UP^G*VEg&XW?`B8;S;WwES1B%ti;i7ZAQ(Ap%7Zudhw^NBoR<W5{$hbW0BDFy5q
zcTS;#soWtG7&lEeiE|!}P*wWj?`MOt>U4ks74(9LYfljxV{L`Uy<9jL3$k+x=TLgJ
zJQ5WOhmUha*ZVwUE^H!j%XAQPX2ZHmP}0Q=H}pmbeDHU#++oMhC2e*7JDbm?Z{8Ph
zy`{AYNet)lEsSLudYxNLR#Z5BIfqGTof?ykJ)pN3XaCO`eW+d=*BPCVv{mS~qjydb
zaNDY_xqI_I_rZ}xo?b6GjzXJl2Ys^!d*=i5aBE3hSmQGqW_rLq9%EFcWDAR~?y$&{
z>lKL#`(SudY4^zbD_dbSZpWlUov*asG1snU-GqgIP-ro(LuJz?dph;O0y?3bz}OG5
z(h-TV;&!yEiYM?WTokF>l+A#{9*0_{+mbsQp;S>R)D|C&sANvHgI23<0)DzG-&MXh
za(mK_TYH}z;P&l2oqE}-p1Uk`<EWWkMaL?E<u|#dS`(5c`7S@jk!}K33WA0tNSjw;
zmKY_Q>99fBq}OMae?uoA`Xi5fihmhc7#!U>lj`GAJia}7l*TOIr?`PMJ5=~_AcY$q
zVS4<OBvEN)J$)QoIu60LFL}(D<rh3usX3TkSFXEyQ#)E(s}D6ndPRNd-I-on(>?*D
z&6?hnp~5*#T!wR@ste>2so#=xME#>P7HmG~hYkt2C^*y0XpRValA|lF&#+1$o9tp~
zhYcyU&nDqoK3jkcmN+j`o(zT&$&z3pE1(_%p|T(kRGOHJ;77JqAsF>u^`c4?IBumB
z!vN%$(`+gpt%<Qgk;*U&SdvYZ&(>fMHS^!x^mJjRJ6K02w4$nmm9Mu1sof0GA}(8)
ztU!W^wgctJey<VxRqQ5E72YX-|DRE0&$@094Eq9|Xr(e#*gV)8*|s_EEj;K@oP39>
zr^&X~=je!CRs7E_v%XAP5E=6VXPH?RRApMu@h!IM$T`OW8+8H2<Z+zi`q8qWwRu{D
zo!52#1VCw=<uM@K8%xt@B1FNkINplK;qAe-_UOBW)?AMVjHOVP90te9_~>%}ECfn(
zc^O;$GkCj8GOn<^SQub3rvuDWby)d|3VU%Va&*Ohy2!!^QSd2J+Lk<p08?6v8iwZE
z5F<ROoMzXG+dgv&!wdwKgBR`Gm#ZHip<fJF4Kuk~Doc9rbbsHH640}%e?Hy14L3EN
zo-fe&DSG=eR`>U0XyEFX7h<<lv6!N>J@gf8Hv$Mkh=0I<F)a>|eL=@m$m4N#8LzGM
zkIU*e;wPl{Wo*a2zYK<w6dew^s7c?4eb2z7nw(0ZBdy202<mHdZqBCNxNT>m+KhZh
zF+2RV_s(5wvzsj<<L>vO!oD&rN7ahbY)b2o9xU5s*J8zn`M?@Zn7O+3Wj<>*M!!$^
zhQsDfTnoZN!W9=H-C;em%=J=%|5^ig95samOo}>1&MbHn5Ov1r2uFJg!L9+44sen8
zkSsVdwj6{|r<G;_qP^d%Jl+F&8u6-Q8qO(M&xWaCLMA*9%kT>aW@}&?vaCTJEVSqd
zh^O8IL#X}xR%mDi)X?=>UWn1~x{2-x^tX-&DZYHXn~l`Nk0gf|!cn_t|AhibY$2rW
z4${d;U)^ZZiBtV{czW`l_?=zWAIs*5?-9ZOQ^dJMRe;32(nsLU2ko85&V`#LJ?s5U
z>(U+WMVPG<?O^5#zmBe^G~cjA5GQql<W#$c<TsjTF0GMe`#Pji7}^*qR(g4{uzcC<
zA~$ObhKjx-vk@1jPoPvOd~oa@TV(-Pt|v5yIMgqsK5!JXjOTT-;&W$hMnpB}HY4l!
zvps7g<1E}oI*t>|<7C?)TBEHQZjXkUv`B)lNDi3Omi<Wt<wp*-QnWwZeR?6EI#=KS
z#pe1uI~j4CI<PAuOd5{lY!X0_mZMUlfhK;$?<_P}m5^K|pRgnUop9FpuVz<LP<Dac
zYrc;52b`w|xkKT>lkX}g-x}#pgx8E4;qmjH2$@qV&UA}8g}ZHlTB0rtuZk%*sS}rr
z1C+wZr@2rfm8w|rlW)cAE6bOxVQCF@i0c|R8+fG<jBnUIVk`JB!Au4@!R>_Xg^?&p
z2ASTh?K_0)OR&eO@JBV1TG3UVrrhw^Yz*NCyTl7_Vg0TJ8>4Pqk7^Ypgori@T@eXK
z`igl-+p^X_wJNQwnUU?S?D=i#5@nZ6kW0uy7#iQIrgToJQF-;Wq!REz+D<Plu_lP8
zRlf4t`LlHEqf@a}hRxek1qW;ic!gq-$>r?E5pM(n#VFBVZN+A=9RCWFU7EDl$LH1<
zd3aPNa!&7ao=*E`9kwa2J(v?1I32Z-aq&R#4)QAlKNoE~y$XwbUiVcZ;Za*s^tWK%
zX;V?W)&g>->TSkRz)AAC%4X}}C-=^$FNZ(z_h!>4N`8KJJ~uiG`G~}yZ;@RN`i_K|
zvDSpvKgc{~y}6>Ly|%=f1zBS*ID_M0!Bc+MJY`7Qx*217W=sx(g5c<LT!Q14gi~LG
zV`$C|&y&TAAL;R8w&{nplN-c_c^8M;iXzMOqUSKNB9YWuq2Fjn<xWM~okp4RiF+e^
zU4{njT)%?VCyCiZ*)38Dz3In$du^QTvfj&c^J|@@G)@s;=l~>RtBDpV7Qj%Eazf_h
zC9Bzmquy<wi*Jno2UY3Y_CUn-orVu<OM~XFzS2t5FV5cw`t?#DI!Z2^?TVb)q4h*k
zfs)HYfH>lO{wGZO^#I(rked}7HD;I;ZBmV^dB~}0AQ!_!(e)9q=BT!N$<NyK>5g9x
z#z>vdLw*3FKef!G%i8P`RTBo<GyfnT9{S+8RbbtfVvd8$24Ba1rbU+OE=;@73~byn
zJBuaobXi8?z!^s3?tJDFxvn~8h_tT974uoh2c3`%b8@E;NznQlIhtCCTp-;cXfF;f
zKJ@%KB79%S>`WKDkb1b-M$vD2CFuc37%BOp)*5kL2@;|Wp~f*aiXZYHaA=Hxm3J7F
z#|`?os3+O}uh$7Ue-+#e=rWHqwqgN~WzzSD)a%?6m&M7Gg&S`K*AmVS3A%9_+rAd~
zpGpkZ4%%#WUaJrZ86pN$k)5t-<rJz15P3o73Boy9*&6~c5y%S?-fbH%{MlxIH3%6m
zjJ75L+;Xh-GpR)56_ywQ*D!G-T!}C~W?1M@4HJG&MxT((6GbxQc7Ut;dR%P1UWPyl
znSizvCm#ooC^RnnI?`&HfCmfuIW1TUfvwtw&*RYmL=$|~52WV~&on>$g6qf9y~vJm
zE%vtCCiJgthcbtEv&p|)ANd~PczEo`X-#n1i(Pee904xFhM!{+!Sta+sun}c!Yxi!
z%>BWd!YArqu>zU-%3Q(N?|<aI-e~_~|8y_L!EApk&40a%(!h?+hf{ksUujPz4!By4
zMG{BH*htzpJ(ANga)y^f{Oq%E?c{6>=QoyK&vp?X%pP&>;Vbr<LH4H5+RH2V+XRrP
zY$o%dSj-)KvveGB?rA(m*m2YgY|<mJup{5E4E_2=`{E%*5`5Dh`H=Njct=K)U`DBQ
z>wTdh(Xu)gpEN8^%gQ4<r$;oEcN3EJy!MNjBZ(k@u%kDfYF@Ylp&GB!wl__FOfEJb
z<o@Ub^v=@{XB$}wnNhqrtj&g-E?&)$bN|yM_3&y&jhu4*bNIC5)sx?SnwzebX@SU?
za`wAjJFeX&6v;tiB2oW%uJ0KejD3r(qUdaF_%1M^k#O3QQIzP!aeUG&qa<_o+WBn1
zvkl^J8@<RyhmnIkGGMEcwJ@*yY>HuaQ9jQbK26e|Y*b=7<9)f6C6*N%ntrG}9nF1_
zJV9HK<~ROPp!SQOUgnZH!<?%yQde{4z^6i<7a2-5msqv>jAlI8r1R@5xv`6H{pvHo
z@<PfPgPJ!7eAG)tdhk{<Bm$HH2WbAtkwVz1d(h$O30r|k>x2~wbYUFu_KU1QhBNhj
z>M!Nrlg&i8p{`=$<H@x?(=*dsZ4sPG2OAH|L<+k^*8SaPOvmqy+Gb(Y%epZ-!_eKg
z(-do~{D_~WACi5`Q%x)Wf7MiJ*w;#$!FG~ndQI)N@<~^U3*t{-Tm8w}B{W%3!p$2o
z1pk7qRhXOyt^p<=v!Wa4umVP&uBo;u-8d6^cvL7JIQK4~>4pAm%aHquO-h67@H?ru
zor{CQd08x@&)C-vh3%~JRS6bgldR^LYUcsG6%O)!;o)tR{Ke+tYQ+f^?9VXopNTDu
z$is?c0embDA1ihlMt;&_QV=cHtOY(t4#jshVNnJ#D^nEWja8vSM1x?_Avqwr59kR(
zUU~3fs{DOi_k=~Ioab71{#8AWqouE*PJNc<7L`NHpU&TA@^8A48o~}xjnY>3YEZ0d
zG>iQZbtCWela`e%dJ9Ps|E8buSHD#1<|TQjoY>WW@Jws*Kly07J=mNPUbp2!NkNUN
zJNsl4NH-h5Y+mJ?wR&tETElJ2>Pk7xBh#I0<(=OdXMc<{$>uSC^97R1`NwA6ik4R4
zkfpCUd^#KFzT$rCtyT*n)voZ1F<K6)@+0}4LrMrCu|b+&7koA>kT1AhJcZzmRaBnt
zq5G9-w@;bq?0rK;FO7(%siaIw7*(`<hNZAEn{+NAU^egAiUdwlpE_cgvF=64>y`U8
z7Vgh!RobgO6z}v8iftxU<YR}3eI-^a81eBM)K+)TzKGP8lVe1h;w$@|nU*b?MhTJL
zlJS)*B$ZuK(d8=wNEs0XY3+QE$G_j72OnqZ;l%=469scRA0>W1jv4)<d;*UM_upTY
zIKRhgu<U#0amwo=G5>Av4Z{l0p6b+6lK@V|)bjCWHqx@8d;$~%;liI9Qq4jwqk(Dq
zjXGJ*N_F%v)Qzs|=`i*dcp<|}Vb+Haoi+31z*$oTH3$>!`8&|IyG2_40JkqW2$AXu
zg0?oBwi~mcMu^@v=~8kgX{i@W+S)YTm?wuK(Tadi0Ytxh0i5{Zio{XLW!R;3SNDLj
zz%x%DRz4(Ji3#L(*cQu>g+j#AO4PLXEIe*tw22CO`BgNxXHj)kl?OOP^GIKxLw;{#
zG^TBV$!%;!-igmg$CzU2Zvy5+@S)x8Kdfp5iMgy*+p|AoszCjv<N?&bA<eBvV*~?m
zxfa#suj4Zz`0c*a^{P7y%AbAL!mhB_QfAj(4jR8<#B4A!kv3d~bcxc@1hr1Wut1x|
z+@7qVfps50eZ{(+_Jt(b&*nk7RsRuv1XgZnl)oFGvdD}CByCwK;HkESC_E3|I27F(
zOj7&dC4A0QBuuYuu8(n%co2*E@Ln}u&Ytig2`7FHfJBuVu2db!0b1Fm%thAca`P<`
zQbmgB%G|B8tcdc9A@Xy#AB@QTmgc+kXUxLo2_FnrTlai+5PX%xZPFCIm;09n78iTd
z>!RdH0s14Y^%AH%L;Mx2j6l&sGFpxQzePgTweyLUNZb3HOfHDKDF@Z{rSMHt{~f2V
z>sI&uE^8|i_}w_|Z58&ay!DB$`nJ?(44Vmj?tHZymr0)ryK$&#H9iAz!WI?!rB8LS
z!uw%U{`EI*=M;=D(r00emR5)#%mMDpg|nh0SrQ}%k%^YB;~(Xf?qnahT;|I7s4W(p
zx%BIvv&gxOq8_jzJn;dH*Y$A{+&1mHN&Av0N%UH!cxpm7S}eNaE0f$kz0`za+MsSR
zHH`kq0o7b2K%u<_x8o~4Hx>JZd<XsKbtt@0gxRgvM$YeKuA?b8f+M}wYku<^j_`Gl
z)hGdr2_Zsk;jF!G3_!m}P0|Zwat=H1^TIYzM@1Nn9Q4tg+G+fQnmPFpOk$$sb92B9
z|BZ|1D)H6+lKW{A4ledt7r#rW&L*K=z4Ju;Cc@c4`0u>9voHc-=cDci$35Wsp#}XA
zV}U}m(3RiST*%!&SJKJhGoY`*>&y9P_nnp72*aF*qd-n<C5;v=^Nnt)_G4U!uU$>7
zbkJZ}TiQ#sF{r+FPU*EZB+{0jOZ99=2jI*jOVp)3l5@oP-1t@&QSD!eBUQXU@GW6`
z3CEab3=I#cW!d#J<>`)VL{Trsl#A;VtL#5u8a@<T6KrbT;nROWk@TZPhLW$x$r!%N
zRNUI^mzU{l!9e7Z7{ufTI9L`i2kuPm>1mPxR5|*q<9}=)rb-gAxPJr<2b-Yl$6R3!
zwln9WqFPS3Pw@_Jq9^l%zx|~i+u2#k?&=C|KfSlFqBm*;9!Q)%8!uu2EQ|dM^KTC`
z1(oaeMMc(7oDq0V_h?tz+Kjzv>-h2`y9?Tdk-(6HN}=#Bj#g_F%srYUo*=l#J6Mr6
z!RG)fLNO5BmR~MSckRyuF%M@Bq_EAbpYRJ9ILO<iLp%wV(pQOUK@g3n7PdjhN@+Nk
zixv7lEque<iQDYK2UmiPxs|z6{qz0K5>Ej=wsg5y&vp*tx<ZTSLgY_A{7n6Xi~&qa
z;H*WD6*+R4e9BPmyEXKq{`P{%e<BkzgE3kzt?_g7$v?{FeO}Dl9Jds(EFC>t^sXm9
z9*DX5*5)yl1Y)(&XFX`uk)swYe+J}$;6Jd3T11|~-vyo4Z{Ez{n?HV!+Z$ZWGkNKP
zye_fbx{=2kL|~y0LV~pKK>2(uQgg6it}b}-1Izr&aH!UC33C~~rboT_GTFWFWm2+(
zO0)KzCPG|9Kx)qbu%9eq@0tkm2GZ?*x|=Natb{DbcnrfkS=MM=yAx>v*Zwm>$M6A-
zn$$9O;Q(I2#N2C*f3fr{NGkU~`C_^sKbkU*10~Qf=VN)ig@EhFG-p<dyq1MPq~(Jq
zrRd`xW1ifKq5pv;Tkc=~$?|SKQFT|iKm%GcW-hF6dfD&W94&=ZyOWD^PNpj`NfSvb
zE5$1L3GSq1+rvvB!!i!4EHsx%&&#C2-Nwi$U`Qq|+I~o`J}M(t(=xYaicy79d2we6
zXIN7T32z&h)rfsJ^p?=hS4d#djw_6^qwz?u1nJs6xr*!p`^VkYG^&8GTU@Z}aKnqE
zSv0Oa+6SjzN{DMldoq~|1;ay)F=7I4zOSHf&Y6b%L8A4&=%aCx&a6&E{`Y`IsHh$n
z_s|O0B@JdOs&hf<_GY+R+z@UpyQOeg&+?v}=hUw@WjUYynK8`VuyeGH^YtQ>lvCkT
zuN%$^&pxbTg<!m(L8{@^PwFf2760Rt`Z}Gu$rB{@-y&4xMUY<Mc8;nre0Af@t-{lj
zrbp1%oO-9nHdfO5%D}nBI_e-(840(Fi1OMD&_mQap05t2iNN4<nh)%7u0==>`uDnj
zuIVxF1Ox_N+S$2cy`(pVN1VH5(KO#k7-ojgXJdZycjI*jDuT+zqRWIq5ijm8M!6!U
zyLiais#NmVv-8NiT=(P6?ZtHp!uX821<uLDl@S_0XwR1G$zdLfyF^;?fJ>#xjrU0Q
z@fhkY=h|*mb`s8d`5gVJMDZ6UH;uT4kf#35e!Pz7q7tahO?}_r@=RbZ>XpEtMZKHc
zL@{a{pw*)L9tj{L%PB##9hn{oaET^|5B$I2&8A-EpPc0!_q!Rd^LN`x#&4dTNn6{c
z-5H<h{BCtRSG=oX{qm_v>Jk5SrCLcJ9b4Gp4e329lJ*EeHup$WXr{a%tCqJ|jVUk|
zLv6~OP(PcxhX-E2qAbSf&Fr7>wwhc_qBQ2QwbLyHag6@p!h1-_sGx0_|L6SPvLX4b
zXCj`+m|4BZN<Ch!A!7eMQVok7lmoUX5}6%YW)5B~lcMS9<=M)30k}HIZkyh!<~4y~
z)b*k(Hk`!XLRpeZ&6ZKsIwCsEOhq;dsN=D*;G}?y;c1G#n;v_*C<DC_+FOVeH4IYj
z)QbVoy#weHXrs%)_xFBnYW6=bkLtM)(Dmg`??WByG`>7H;+5_@wtArNcRX6sSHB_F
zvEn^F_WFzlrKPWD@sgfzvG*@?&UQIqXiGQ)({D~TPUKz0)0J*NH63j%V|0dH6rqht
zY88O;uX}k@RZE`BWi*Z~Qvxdh%#d(*zSshR#V%yOqQf^cprI$GGDnjrEb)9m+ukMx
z@#U^|KJgwOYhgHFcV3UvsJ1vV^g_{G8$Ok0sJ(a8vU7pnG<qKgcqsEC`1qfD8nW8E
zM&hFTu#EQFPt+~FQS&oLLHLLW&b_^+n2%Re&CM;B9n5W_ZJbPAKp9FwR|%{U&?7*4
zN|4cIAd`K=yqd8wKCb;A?@D{5!I*0_lMkyK&6&bn?~(k;(*?Vj`ae?25#jq*pI&ns
z3Ap`s_1NAi+J8R8a~0IK*W^f!B<&d&^$>D!muKX;`*_)5KopsFsZ}a$5hPB^6oQ>r
zWS`tkh2D1mfYA=Z2oYi~An|KvjF*YwDgs%kFv*s>rG`Yoz}m6Z8|Urbte4XP!B4{A
z;=FK%uVl1NG_d63{0+Lf)t`k-uhDfRXq3PTRElB%pH4`KHwC!h9*7?|tCaNII9mB0
znyZ&qBkw)kzds-~%x+JkWXq~Ni<EM@Xu>lbZ0TIt`{g<^CKVY|+3>^}=AAkL-TmqD
zA5!3yo97<~A?WJ%k$WC<=GWR@Z-V!u(^{Kj3Tfg#EPAx~Y@#>^(6Lz0Fgd{fs#Sp-
zArp(GkqHP|oeIR%-@M%h&DufDZFd;mK%0fKEfQ2!6Ca<coN86$()*(JMEwcd`XKSr
zgvI!~#j?wvUCv()wn!V~WbTIw2*GT;(Q)sW5_#fMvRplbS$W-(%X@yOaCzfCXc`5v
z@fC{o<@SKrxfv5<IdY(wK|lp(;j}^sL&khM77=XCXli8%351p5N6}vDtsh2W=l<|M
zut8hF$kug!`<-9^I+tHg1#oiMSk~YIztNi8)#i<V#v%*UTy;ihVgC1?U44K)>uor4
zP@#BwOYOgLSm2pd_t&ty-8Cq|*Y{|iJl(fB`DkKpeVm!EycY+1)z?dLsgCB!ZtKFt
zt2F`J>ZFHmpeNJsSRAAUO7)2R5!zW88=W@^XS9v)3`4F}(f&ARw&6Vf@J@i!Vdl%m
zFk4I1lO{`&Al&59Sb-pd@daA*gH4x%s}_+TK4f1cIqlB>##s=fI`7W?Y+$qukL}@e
z#Z%$$Ekf9;=Vapmz%q<&aa0>^NbBj6R<<t+yCyWm--d6buc~<6{@l}fx;vYC@oG1R
zvfo}~{^P}Bo4WEb^6tTzRs&<63M~Re2rgNJ#aJ_&N!XE_5hy5fn{6=l{2(>#F4|<+
za4V6h|6pJ-{GYthC7}oIuT$=}!@kh*y-ldF6FJ=}hW?aPPCfZpuk=mhp{SZ41(p64
z5jW01azF!9E~4T^`0By<_w8=R9#%z!btpJ229+mRD(D{?m>1B{Xq!9mGDjw4@gQbI
z=$4W*>_9sZsX4lA@S2SrvlWXG&O+93#sDR>c+K@!qom7aZfCi=R^-iLgPbIf1IQrv
zLqj+c+?6bVJWdzUA*l^&?882pH)!Q)c^PSvZqGtaTx0M2YI-V19(8p?`7IrU#V|YS
z`M#NH(2wXzPr_54;-v-qH%&@doJMJ-njjdJBBdO0T(yZ1tUNW8hAq>Po#qhrFH7j9
z(KFX8VE037bhAUw-mcqp^ZQ|H;qSKSR$rgBt0!@GPfWs%WYr2Pfwi2ABLr*sh{fCs
zu~Wr@WO9%(M!7^*WPdzVrrdyjO)d_^j&z@c8|!e_?oFmN%(`CJfXgymz!c;l9;>kS
zm#S1PG(z&VAswdNFMN%LpV4H7_yBqrl62|Iof$pyl*5HcmJ9qQ9gyOC;zkww-7r(V
z!JIP0Y{}fA;SVNUM}B2o<Z~}7ES*^gw%;nf-ebQ1>u3_g#Ra`oS<ls(mZ0qH7Qa2|
z*8PEiejKVF5OXK-BT|gY1F|qW*5uXRVkAAF3m=*VaO*$6{ObiFeyJSip&aCgzkHW@
z_C@^Z_Ab<_V)y5t$-|*Xztb(=$Jl&?y2VmvPzoX{DzIXhk1P`h!({&wJb}^V9)Gbu
zlLR7jvU8G}R!{E6#d~#x?<W%%(qGJikijWZz}6(WRJSm?X^dplNk|0k7;T4xn9Hg%
zx>FyD^)70QfHmt2G_U8guI=`1)e0U*tNoE|)e6@f<mMZnWQf*u$OEMT=o!Ilf`Vo=
zqK@MyAcDxw4hkA2doqXja432l4Q^y63=^spuV?%OBlLDZ;Od1;t3g&f{>_uP03%#V
zl1*3OKhPANN5t34aaHhN&Q)@=-;PlFRZR|6PCGcZsgvI9lp3C%llijYd}g-cm=7Vd
z^~~he53fS?i7(im--RE@6p|v-qWQAe$RZQIY<{|p2oznzWv^|8mtSSVomMXJn_TP_
z%!@2H(dvZoyLuYJ@--kibbw)ZYcocBIPTqAbK4C;iji&}YG-2N+a8g~G+s|csPG)B
zAIgSU&^BrJ8Ta6Evg0vT3vrpN43Esyjjqt72hM#@mbJ%vp&U>r2S;(4i+X)*$qLd^
zufeZAg<7Vl_?u>9n;cL=sMM<=Xc1ldfmi8QWI66ucts|%Ly6yPC2BG05!?hr&#<Te
zB9JeQ(uJG?eomSfTx*5|1rD#DA^DtKP6E@w$;wD|u^$D&_q?{Rq4Pbtl`+y+n?B@<
z>-MeRV8_Q(w^>lU+LSlm_RQ+kvb=nm)bG>lV|Sw&V~UDaZdLC69`fX24z^w)eE79O
zNnh9lI8Cq?DKS}BIP%z+A>0=n)E1Kp<zYRaN%I+cBCBnc{Y0LUec&BIzGf`h2PooK
zfs4Eby*XJX9kx42El`CF4e)$>;p~cV8Qh9)%sm`)f9lH*oZ@Rw2luswZMf9%qIkDu
zYLcAq!QM|D5s@yd>lyWe2=a0$$t5BajZE<-@grrDOOXYd6JWbn;$hxsalpTl;;5>i
z>u<W?N(w5iSU=~z@{FNs@2dRC!<L8qbvsEJKV$6buS~%WUOI@(t)3S_H1(v>0lJG2
ze#AhZ!M1eJ!JKLs<5TGT(hsbt;haV=IWqV`+u1%;+q#C^#q^4p8xI9C;N<zj_uA2A
z&YgrjgA}%cDBK3UDQm_zunxojkp7iuZ{sYh`JkmF&@P2~SkNVnx1S(C0@mQ3({;Mh
zxP3nlwuK>LWL21SyCJbP-ucu0zpB$PICmY1^qVH)7#8)@?!7$t6_o@}5Em7ls2XZI
zn0r0mJu1(w(7mDII6~{m8K?Ngns{z?lj#|A<^8g(e<R8@9<+Ad1GV657hJFN=U0Z^
z@b=UaJJLVhOcurXxosZr7LgSnml^Cy3vQdg#8=ybPbR6{^Woyi!tm%LF%Mm#(gKF+
z6MyP7zLu&>TvYVkm=Gns8DELpXTPWWU6lCKhV2WtW21a(e5dQLJ3I>8nd^ws9ETQ0
zZ)w|OzJv*<WtMBS?NLsP@>>N|GDba3$MBMEb1*pdXb5w~%oJLzr=W)%Hz0>snOX3J
z>kM^#Wsw_{(%CZ)z9f194*X_<CWu?t`)xPlMz`7|(v*C?ycU|{T3TQtc^S6V*K8Fu
z39cF=8fQtyF`tvS!;E=`*>fwJIP8AN|Fi0%E0w<O{+juj<|M&Hp!%n2MjU(new;6Z
zjqf7snY#02CF-DYR;@|bM-i?qC|IxmmxU3+C7ALj;H1W2mbFi(FlPE2MQ8^}_Tup5
zE@KC+!ksYF>X?^hGC>a=u4s!is4W$Q-o))EZ)jO&Z!OQhEc(f>MQHV%cN%RCEZPW$
z-8XTG<V&)H$CP74g2!`~=Z@+Ejgd7Uc-z3_qW_noxmW@kOV3_Zv0(%CSpwP0rpEoa
z8J`rx$*#OUFq_cVWBuF9H1iX?aqzu7TD)^ThSn!B>92#9dN9=L?%17w)>DCF0tRCf
zd=>>U(!!h2uotC&Bfrh-Xt~c%Wi@dRb${YxBSXO4>fZ;^6xV*@IE!n?^)o5mguT|=
zvuue-=oE>e2=1Ywl6WaAu<ysobPH;Gx-2`fqC4#@^X?kx=wH4a0JeAwTA`#-tnjFS
z$RYlaiKi#XSe!e*Yp&gyvuGT4MkHFD4J{H~;zvMQxTR{sgeE0SZaqB926l|li7$+X
zWcAI8MDk%YXx)m^yRem2Z?=is`!NZQP7&5}z6*c$A4??^+tgQm+B~&u)(8S$1c>#(
zH|vRs%eK`&jaR?i1dFJ445$QbDzp3pGPkq8o7oY+!{t*r1&>0F{MM^G9o?`x{iUFs
z>!tUG)qKa~*|tTRHvT{LwIkSI0FGaH&FYDOf%<fx9_ShmbPtuQc$;A-(k>Dm1&;f$
zgkA3#@U6X8Gd>=iu&VdDkT&&fZ1^2$+~R1fIq=H2lMrh4xVw3%;|+<+S3zYiL=)Lb
z=rYw(!}P#oFsSF3qP@oN!QRC^r(rFOqR>NVPEbmUp3}0&?hc<UPWkImDVm}2O0R%=
zZ0U1yH`wcOk@j@1?V?_q?5mPVa?3srb(s{F1t!2qF(Lzil5P*6uK?Q*fbpjd^2C4p
zTOt*xoyI2A)zs*HQV8ieo^BM@Q{lx&KQAVsi;kawkjNipfBJaw%r@Cz?+O=XsZ|hi
zi;*Kw4mb}kUr`h)LN~`rB;>dq{~#Dpi@wa7{0jnh@a~_T&;O*-7J8QbRkah61K%QX
zJ)QNjNSV7A_nUdk&29#IUi&-0bX#c=|LD-*EoLbfq_jC}7KTCj5YQZDx-%l+6SI%o
ze?+=HDRarykVWq@w7L-VV*EdKeP<w>Z}c`wwUoAYYt>%0s<v9KO({hQQME^+YE!FL
zX^k4KO{=O#5;L)>O|2RU5<6<d-n;Mf`>*${-~avOgWQSx$?ZPpT<5y3Q~PprH`<0=
zG*ixNzhglNXfXGQw!RznWbjN6pVaI@LLmvKtlwyEja65ZMJAzIasTU<>n^pgS$<QQ
zQa7w9AbwwR-X03|-kSJq`DY9DNnL4g$e7>jYV%@>Vf><T9icj4$U9)5jCD@o8i(so
zQP`etiYa%oMTaUq;fZ<?03{P$rSH}z5_%J7H1;WWhOF*Vo%nB9wV{}~O8ZEO;lc=v
z`jtsxLb>r#Ane7liONGB3HLL`+~+W1V>(lwV!#q66Z9#Qy$wd!nU4obhe%nZyYO4{
z=oXp!&s20JG$mt=cYRC|A|f~i38jGbuH`WD>+G_dzx>FTFENGw$@CNhMwe3iHKmtA
z(yapcTt0-(0*nl8+x`pL-a75Gl-Kc&X~5Uam+?}_-#=<{tJrF+W=GyCXkc~X$c5zS
zGO6g-?g6a-gN$Rg3HGz+6d+^Gv09gCUnjx_H^7Tfjic`_dwlD0?j67F-K`*UH-z3S
z?KoXWN+=Z#nu&^jsGJuNU5voAY)iD<Y|5n=Yj1aW4vN>N3J4i$VZH(63JJ%olsite
zpELSPsuFuEys9vC0!mi86UXXKp%Kl*OZ#XBgk{a^e^g)WZ0F1Rah6T5{M=!6w~b0~
z+l9(%j&-C}LOfitG)O~_06sh63v!&UhnL&e9(+}qB8jLT?fyixq#v|48C^Ji9)amK
zia*aBS^<e2FA{$@H1Xj=SQ{Y%HuE+B&!eD1;cOd_kI90BQP;R(7efu>yOgfPOni%I
z?@^#TwM&n+sr}~iI(7Z8=?F`v(C`{0KztOuYL-)^=YYLj#%z!FpLNmZlwM!h@x*9N
ze1VUp2qw&DPKec^zmZ7wme7m3!}gW|w-6zbls1)_$VzuzdPI#PRAf56$=~a|EoIz8
zLo`b+@VFPqX@SP?-5<f!+kYG`R~Ed$0`QO4KOBGS-xE{O+IVELqI18-Pk;{aNBquT
z*VwuM9SR%raTj7dBS?g-a6@DrXEbdtwV6&I#oT5;6L}Ow{YFFjw4OOA3uxK7)=Eyf
z*2y9XjRQS?-HUdU#WYsCKp=jaUeX89&*nN}Y~O}OD9itN=R$k<+DCfdn&buDBcW6S
zugrmR$sWL;dI-&wRp36%4gYmpJ;H|8s0O^rP~n_3UaoQnYp5;NTUKRP<01qM8DSfz
z1eZnMU<ma>lgC*;iMM#Yp{#!w-S#DCkCD-BN+EC^AI;8wyhToKub-8Pl6t2!QwgpI
zkgp6lOlKTz%sJgHhsPE^x{RiLcA*7_SB#W^qKJh(+*s>NV_$U<iH*0rIL`(jVWe%n
z*7SKBclYfY99H=0g08AJQH8b;UEjF_v~aWPmd){KOzJvt9rJ$`HlS%pC}oeEe?yj=
zITL-(5ANbE1;whm#Oe~83-GfFChC2t-m7qVH;!ynHGAl^FhH5oGr`19CF&I$X{h`$
z!Vj5{@TSU-0;!3c*e=0VQjh!H0>PC<5|NxK{<Z>fJesRZ6cP(BT_ASW=M$$}M>Rlp
z(w~`L^KxZm;U|!-eewO%nKPEFjdv~roR;X3dgb0sD$K)h<`1ammjG+?vccclDUR#+
zb#YLx$jks{!VBU%%zsw-dUvPKQ%p6$K`R7<l||xu%KUx$hK2=7=&59tlD|*{q|B$_
z_t<gDkE$*zEnKZ%3x&a~11oB0C$SPn@@C4zx>yst-)3Df3IJab`gX%(u}_qR_^BN2
z03nNDQNmez=&ztDnBXWMqZSBqUv9!wKq$h?tNn4NK2A;m!&YzriwFxpwh3<3v!AG4
z(FF;Z3hQWYcd+a5QWqnmMuXhky_~jAQ+5`w_pvCxUK;t;$_NEFA>Ygb0jhgKr3HC>
zxO<yfYji5EBzEKS^U`WqU3-XECCuo;&at$-P}EJTZft!Zqg|REQR+H*XUT4Pl6>&I
z(7QZ@qL1dr#FnJvYi0`3POaSQ^DNjs0&}Jf$P%HA-7Rz16*5Q*77F=pv8u|FTVJ{7
zLY;O+J+sAToIl%x*7s3;rI+6qlZZz!3IR~c&+%s|EcvjZc3DA%$3i(BWTa*c&M{#L
zDcUqrhzH(q&AthP;7mTq!)@=ZQen80rF)}HsPYVH>C3LnR&6)0<%w8AfrzC-@a}e@
zgao6#y$Dz?c-`c;Ya}3GrLTejEbnlL@t72kYX+%^SY{<z&#R9Tf0Ivd3{lXcu*p=t
z4WmSl^xoHO`B<&3p7tTu*#pmxtU|g@3|5A<3p-baP_Y9UfAp81pUFOixdD?u(0ygB
za(2wuxwjUJm=&*mBqHMT(R!rvv>g1ZN^QcVoeqvC4;qN5L2`y%V%zZ9^mpR}ql{8p
zu4G;<H!-C_&i?$mjII7SmHvgd#>O){gx);}UE+HrDD~BOXV>*cH7DWsn6~De`At!X
zrC$f6kj=~@)td4OY1$Q&a!G)_;&xY5xNoxF_oyu`PlGx~;nmV<Z1IjzziiapnI(Nb
zD614!{E*FL0_yF(c1H`u_kzGJav=xbCVJ>z7kjH*ZajDMEpVTSe*2T>rYL(ZfjE1g
z&2*w>xBH3A_i6|<2}Oq#Mu=`G!~NEu;yT8Wh9x@bNU}GTdxfS#Z&*!ctEyNJpGw7Q
zh*LX~;RUb;ekJrzzXQ~2NUDR&`|ftQcn3lN;JXnx6T2{C!BR1}XAUfav4OXxxm{WV
zMSb5n+q!?gXLf2~A<);l_lA)~=H_nu7g6eBm&ta5G`z1DK<}4YmL(kEeSLjaG6BEs
zr|kpjHf)|Z+=21hWqbt)#*kqPoORpk+7-~XvQPho-JTy)orm)MeS<UeLs7Dl+fq|#
z9ydv%0y*A{^9HMlhWWFB#?S1NXR>c!U**h+G&vzE$qB4WX^{fj2eLpJ3SD5q)?&zS
z@>iW_o>tI4&FG=6Iv1a-OZxoEj8YUjO9<Cs*?byI=XT&dYnP^~;UXi<VORY5sdy9G
zls6-xNRPI9;TB4r*)gwG0h9xem9QU9$3LyCwH+`iiZ4#-L8~9tDOwB72GHFssa1r+
zU^sJga|NIIx%~Nx&Nx=t?O#z2P&x_M&;*7~f35T$jGO?L*B$$Nw&S0NFjZ2i_Nq_C
z1A-O&aDiK$5CwF%PQo9}pu4d}%@IT9>eEwh9g1f?yK#u%-<CUL>u#!rqL+Nh5~yu=
zJX|pdw^_jYM-KF<+*p#q2XI4IN{sa2c2Y92z6{mHX!}0pgb+6q3vz?oukPD_<{a0v
zM%@k@Cdo&ea&?*^kIH@leZQoVTQ<4LJqtprdu7}M00|Bt+UuhuZ5FqS@TyCZkKj6s
z6AW{vZ+6$-<!;<789!fX>C15F&v_8|{P)ZcK>3=}<bT$)uuwH+e!cyK{yQTQ;^9JQ
zgJj_##i~<+*4U>}^?*$|uei>I8cQMEbA6BUPAdl|4IUzBKQx}J%n@BFQ-BV7#R)p;
zj@DUUX|Z~IhQ>hCJH2y4X5NL=7rghMzz|`Src1lgz7=cXu-o^dK?g<J60QWRsDsK`
z|HMotsIiE9>rRI(isac%U6}~JRsQUhtGd?rzk1MDZUc$wiJ&e0hS0)>3sr12n@|e<
z{rWb5DC~c-6tiM;v%9?*Pa@6j;1NoQ>3k-zpQ!`z^L4#7>1Wd`o;?GBEvmhCR%3Nc
zAA{w*j`3yjN^*f)Ek_x%1<zX;4Ma^d9=Te!AVuu3K$+QFKY`7%!g{lRXDrjbaA|)f
zr9~awN?Eae9=PIddCWkhZWYXC<JP1gy9dh&C;-HoD2{cZxlNioT!qxNJ|3=7GS^!H
z2=1RYO_{qrWWk+t4v|_c%~+d6OfU3-1E5{%&Z@GMT$y2E+lZg(JHHS186GQCPjJ#o
zL|!&;#qcajb_0B^l_W1GBy9Rpf+@4c6Cr;4ac#|RxtnhR1Gsm_J;2BTeuz$@bbk+7
z!VeDGs9y8B>Plbe`s#8#P)@bR53tBfp(;xU340jV?Mv{Oq1xoNQm-yAL!^IfEI|F<
zZxOBbl<(CAg<QarIG97`_7%U{uS4V{D&cZwKXeMt+ZYvPWzRrX!Cv?UhbiN{EQE3)
zLS^<BKP%>e>2Qz9S668<V6!FI92@NU+4pY)>RJs)@)Sgwc!8%2T!<fpj$0`3)Ozpz
z{R6(O$gB=UT~#NtCltKA&JT-Bm~)VOUrZjVEOd^_*^%KnrFfkwLqXYY%xchNBzPa7
z{(p0R*C|()@LeiJ)ry4y^U)MBUGKiXjXRqEWdE2|$!8l6RL`9V;;xzDQwSi^YW$Mg
zXTFccZ<KaVU<)5!2Dt3I-jj9T{_#86V?J%w`>V;K@De}|I`3b>6;;F7Ko1lra!w~4
zcng9L7?mWpSLdfcg8%50?)&jq)jbHT8wI4Lz*ZlRku=v1L^Nf$kwU=cYcaL9<eFM@
z#~vv=aitq9>MWZU!7Ofe!@fV(np4-CZ_z(0dGkhpu;aEZjH*%<kZj$k>EUya{dKfz
zo&#5V5|vrtL`E%=<0M2#L>g*iwdLl~#GK~=mHpPn_jXIxU?QSfIZDz#hh3dYN6Vqt
zm1MjCPXB1nNr{LxmK$2!ZT>FyDz>k;j8)1G01N+b>lr``;wAszhWw3fWuSkOBcVKn
z9_7<0wC|kqce3p~*?1tkHRa-#HfdzlskxY8;0uljcXZcs9gsy<AzYHGx<(gu8iS3^
zMHEF9KLGj17aG}qB$X5Pu)=fBC87XAo&#Vddh)RvqfD8XZ_pw?)0hd+<6$D<&>Ki?
zAtt-$@B4$3z$2W1kB6iEV{X@b!zPbr40<b5Y^mYGTA)8nZrSvt6hF_Pf3`<|b~@<y
zvB-EW4VDR2EdCa?%mkJ9@6Rk^Iu3BTx99n-qq6L<p28{()sVz;FPHBnm)6QJ5CN}n
z_>jh6$wSg>P+oILxZ*g;()uZS;PKk=I%aBh6?>vaWEJk^rb2K?dbWIH$QuBJGx{z6
z^nL$GsLD1zIR=zJA6_FNM|Y!@-dS`<oLGN^g?AR!Gnl>JW>B1zn+tI`nLkXPj9|1S
zSFehEm{4WDu-AEH0ZMtw!?%t8X-Vpo(Ph4P>ahSw%#NiHI{iR{z5de(--)vW8UK%>
zHJd`r8ocDLRjv2d&<7*(>*u?9?16vqdjt%_tQ|~6T&L#o_iuAmk9*KsIBNh3Dwl$D
zjcfGPG)vJU&)3?N2LFn?U9BYTYnR%1hfxOV#bqe)P@brgJaWExqh8^7$fzm&!v{Bk
z2z}MBVUY)2546UMING6;#kaX#xeU#&`#fsX!JcL&z*A(fCfnf<IV?I7+gVPnn5H)`
znbvV%{SHM)jG@$9?6aiI{?fkqK(BH`LH&Ua%mSJR64>r8;3SmTTkch3;En#FvhD96
zL{bpa_!EB&m<9}FD<1tS3nQl<$Wrjz>Nq|9N2jYlWu&*)24FgRqg@aQw~Q5_+&4mg
zAJ<rrYksbEvEQM(&P0d0&h$@8(ecz)5Mq=jM1<~8+KJwJ5PSeayjMJElkfTNL%iq2
zbjFZ8amytniUwID^pAPK)9*^)gUv6TW={`qkz)9|jH(0+HTyAY3DK>om%z&#%>}FX
zvpcAn@@MJZyf+>c<c844%o}8t%tpY`w?pt%T|}?*%Y3g2d1--|ZG{1Lh}b2J79czL
zVn)0qAH_VGKbrB5(nET2QQuO|Z>4MNlF|dO?Y6Sw**aiaIiV;4DDff2-5F`t@~(Q@
zxt?ociQaiF4BtOhTnsL@tiTresg^+iw*8T&nrq+z$Gls;SS^q>DL$zElP0><z2&A{
zd_ad6zoNdrD2v=f@K}fzDm~5@zVWEAZHL0C7^s1yzkPNxBfYY0!b2Tzm**>y(NTOa
zLg2k>i@tn+xv?^jHKF)JilNxVp)-C4qsB6KXeJ6dgH?b_<l89lcU(AXzNQ#Ntv(_m
zG-0k$yGQ{FNJ}%bMg^a9TqrI1_0wJa=deXMxV93v2Dk5yhYM-~0!m>3K1Uh1cr&@b
z<3JCV1KK>n#Y=rJ{`J2YYk*U1zLpvYqDPJH!dL^B1u9N_f|m=oP+)&ceZ6{Et|DdL
zbm)-&38jvi+GGa{cGeIHk^$mL3@2bL)Mzo}2gWf}KzOvKjwR{2op`C`vwpaoW%;%~
zo_BxW%1`y{N`>?MU69o$79JR5we_@b8A0_8N^>$&+qUg1q=wpz+ss2G$AeIuMNdjX
z8I=!~t_yORZK*2WkuY^|h|iIIwcFdx6(u4nEb8^-<1u8k9rDHdnKelIC$cc6MsMtk
zJ*y)lGA!;%MO;r1mfZ;bw`~K*SNa%iRLIA1lyXCT9~pdHBEMtW_lIuKmQnoU*T1Xu
zGc4tsN-B@<hAa}ThPL>+&&r#=Q}FEZ<^og<=&$GU36Bik2@TaaO-S)SygJ12vg^@I
z+gzF}$;O_~yxMrE_$c5sagXBmZEX&O+CGixqADi52+#j~=Z;%4;xk)KsuQTo++D5Q
z<VDu2L$(qbPe|lw3C7COtmgIJHjv};>9aj>Uz#Zd2<KDPUIcrk*HK%6hI&okUQ|KB
z5bQN9zbjVoqqI?&MV8&W$eju9HIso-2rlme0H^uSF2jLc>Rq+S%5)gjB(Do55bAtR
zS3v_l*+IL;qs#Tgc~iz*zb*4e*pyC;wHd7rZVMhbtGULuH{`WQDc0}G@^9Ax1D1KB
z>u8BHDV(r(pgT69%|OEqCs;o<9qqHmrV!~alXr(=aPulCzvV7Awboef3W&^+j7Lxt
z%WpR;l=^UpLFAk3nQR%`IC%(b8$%-IhYogO%KT5qTnjEm=+k2h#lYW(t7Y>J`Z+S*
zduUih)#fG5-~%tiDu_2^@){N?hqne0#<Z%@z~?rAn~;lgMH@`MBljXdI{dzh`$@vD
zdnd?3IoTB7d)p&<Oab!|TZISH0rsa=i%Kvb-PhCxIiz^A{1GD}x$pG_$WdS}OVU&D
zQGMf**`vgCyv#F(nmjVq7Afu*4NtsEZvg6@Z%T2f)%*K(PVb%8hLJ_-@I4nfttmm+
znI!8S?c7JF`M|KWw>uRsWH+k6h>UA-B=c~)T1FaU2TV*&5xDx-3uC}hRa88jjdy@s
zq=GEW!0S;P2Avypb^;UJu2FywL;jk{6|8f*6imaO=iXnzUv(YC4`nwrUrE9aq1sQe
zjM=NJr#%kd@l8i_MtP#^0l#$PnqcpbWP$56#3?uZ#h$*$0ALc%w@}PkH2@6w{QM?!
zFWQtI@0j_3nP)xy*=p(SoaK#<45;)saT3Q+<*1`7qGSpu5EU6!%0Pn}r^9d>QrwU7
zt;f-hTr%OqyB%W6sL>KPH(agzIWTEmLnKr7fydEEng`!?*80lQ>Or#S2H6lCx0y8O
zKOm`|2ZTJ|k$?`}H@9`2VM>_?5?AIwZ!cK!aM(TrcJ(6iyFOcTbCtidA@B5r5|V&v
zJj5LqfX_3tS%R;l8z~mSZ-CM>|IQeRbp{J0r8M?utQ~6E73*j?yqD^{_vL9cQGWph
z=Hc5DH^1l~P-f*#a{C55jB0tpAmx2(y|EvSG=0_iGnKXz?N9yHu@bHix21BlDRW(f
zKUTmk*Y@B*@6LOBbv94wXg251_~#v9de}1;07eFcsA?;@vPcL}R&D``Z=j7w61Ipb
zr}@k0>k7vK4-g4t>HR2OOB7}GXS~_c(DF^k>+8|_jfYEzeAxkI6+Re7@u8!1!(*?v
z@4xPKX8^G8&KMxIht+*^{_fSU#CWs{-kJvHYA^t4nutuZ?bK?eq^~448np&hh<Wmi
zH6QX_yFUpzE2D{f9S7MArGYaRr}ErHx_bjAF$^RHL<CaQM>ost4Eckt+4$h&VM>XD
z-|$SOV8TSz(ZZ(i?POXcWXAy;{tsh8#9ID+Gg*53Y8Lzv6r|B78KZWt@|S0XY8UV4
zY969U?P-&I$esz2Ezxn(Mx$Kg!NzOCUbHm$@XbmpDj)NQG=$N5m-f7+U3VOoWh1ZK
zgEXvWnfl`;*?l7=hjwv>K~=h>hLf1`<}>#Y+SLK?nMqg1?2lTW$yb$BQutgx&m<v>
zsy7e_qUuw>`c^LvJhT1OWj~=CqCv#Ab1#JmwfUndI`5>R+1WV{d{&#U2m8)+2JfAD
z%z|!5Fe10gfv-lTs>_MBy#2XZT^dj*(U>~#hx$K}U%dv(HT#WLy-*A%U?b)}COf7D
zQkHxI+Eo@rrFr^F_PmN!!LR4~KJBj`o@Q)#vZbBZ-aDqe;Rh~ZHd4Nkw-=3(H+ols
zMnsQ*MdK>>Bzr{682K5|+Y>KxaP)itOXzMdGH^igjTX~0+F)deVTUP?#s{DZw3#fA
z-~4i;s-}4lb1;WiA$}=1_Jrrz8e+2k)$Cj+dC=*Pc<P|TAKlD5JH^s&78_CQFOA2b
z_8%d`?!#Uy**0Uc$krg@)^76)hC>{-*oW>qXmQ(Pq;LemCQq3?WO9euYO;=Sb^u;e
zKn668*UZyj0h_8fk-@<4wT8cx``WZ13mN=Uk7}GJ1C;QgYB4n^!>u6rcP)|aToQLh
z9v7C)R8~MpKIuEDN38urP}zlO5b#VpLYgl~c)Tk}!aG@-&BQ3H1Pm01q?AN4|3v+U
z1HuI<pI5&($GJKm2cFJ1)t1%*sPS`BWxk^)_jxR_;ec4I3sTo{M>YTvzNSu#jJ>E?
zzB2C257pv&c*WvDn>bN*Ki6=-!ad<(t00E~IB;Lr?TlrM7Tw&DeOph$7eFT_@a09D
z*A=I_Pym0z6!pw`KKS!YQ^eP%R_dE)d%Fj#MXG$AG-0O7l@IJO>`Q9%Qm}u)c+i|t
z0@_*+au;_Z<mhmxxnwkD@+Z>!mzHXmqBo}N{q|Nq|C1~&odh5Ak@sgRxVJ`2Y%c!?
zOqyw)-0p&1omi=1&aw%Rj(}hRrEwl4>tgNgLlFbImk@c3-<6@S^4f6=y+g4#_IBb7
zCaM?Tw11j6Kk%T;4IkYgTJ-CK^yu58TchGRk*efJy<@`ZHk2bc(ch~k2g3v5F$f;G
zeDdOCB;qbN)T*C%t7i440uOD7$=lm<Ktd?zcjj&Z-HdNP#L2Xx16=co@~Qm$=Z;rk
zI-d|0a<A79qc1JSu_<}2jQuJ$g<F_YDY+z_QV$)KhS8J*@V-P!myh`vGTMyy0^s{<
zTGd@A@(ZV1=-!SSmK3_z7-llwWfNe#xHyfgIbl>n1+TUTD%tRM2C0*Zwsz-f&tB!1
zToz^JSqFj^saX@d&O2jG^zR`lc+D4Hvz9WMP}(T1g01dzx~Yoz=i0Hk2vWg%XL|c_
zXUh?BM)n3(T#T=H$;`k&W_$Ur9c;R(@f88xU6W>{biPSGeR9|w#cp9v@Fr`d9xpkO
zibSX(FQM~i%ff8eywk0*$bZYC!g0d<(C$dbo7G#-Owq*>FGF8ZVAc+Ig=~zhUI!eE
z?dY$w2L9f1kTXhhTiW_1+MJ5|yvan$MPbe6O4Djg0C&8k|9K7=N-yzBb{|=o85e0?
ze!Yzj_GZbXeJ>YLu*Nji`)w;RLsdgSCcLs_d)hl0sdA~cMs=(x^#gq<e{-apj}v5J
zBE@Bg+(AF<_;B?D7oS5sTireR0DGR-Tv>Pr3c9P<A_YA=m!$77ZJgY!ye`l7T9^5A
z89({n|Ie{vjscYmnE#&cz^YWd9F`<+PAk+JKbj)Y)&*~Q=*Ksm_A=Xdwz(A9WD}17
zko;oqsK=l)r>%*OapyxytS@4_Q8d|^1G+OJ+_M9@=`ct^<X$>gijA9DbG?ml7#@qh
zDRbowB4TL5hmlWz*$z@4m*Z1R^p=DHG9bKfB8m_Z&Z-HT9r`@4bhOkQrF42;<6?@~
zee#r(x$<X!k!w}Te9_xUU>oW#A0|UEl-Kw!dVc|)tJE=|S0Gn|857!SFOQdYt4Dl&
zBfzPW-Rh|r&eLC?l-R~kG495dp>~(Lu!F~T`K62ov^U-tz0M>YGd@8jvTG&cg!|GX
z7z>GJSYn2GzN?wiR6k~U6jD&({u;3E6CD?LOY>|2fiS-Pb`RsTRj8JmEULcF6&8`}
zu-3<ng0f`Gd;I1rh-U$0R&CwI4v$%^R`=0&W6W(!>Kq%CVJAfZaW8zW^p&@S6Q~l`
zdHH|UgI}q^7~m3xttn}$m&=H5JWv$OYN#98Wt^U>A}mh(L3g_~!1v|(tPHrg23op}
zv}e^E&uBu#Q*=y6>fguoQf*;W=NO~~*|>Kgp<)9jq=JA4)MoU!mc!;Nsc4_$x~=N>
z97c`nOa_CSPs~HZXy^;ZkG%k|%OcxnJ+-DEKYesy5V4X~2-qE)jR)+RD=G>cKHhW4
zTY_iUWU!(7@*u_900(v5Hz9NKd#T#b+k+y|;uA+%1hlEXODW_p<l)LV?>}3ydwZ+8
z)r4Pzl!#Yx>$D3`5Y>IZj+K|K-S}lmdy*3b-|2$uFPD875QJHoG_!^3DvMF)dxCu|
z0(tm{;O(O}P-!!rHb{w@iF+et?A6O9nTUd88(%JL^qDt#$MczFya!}n(#E*lkSJO8
zvdZPC>?#Z+BV(L}Oz=8ir_%Z1=IV#MPa>HamX_w=n5)=bhwz?({KvUc`St{RtkeT&
zLTPMga|d9u1DHEdT|$RYQD8R%7TA^k@&gNZ1^O4=te-hTnd77MW;#xkIjM|0yWoh|
z_MKeQhd^fn*h;=c)x50a(D>r<c!j}=TK=Ag!O90;VlAAd)&z-xVSS}%BSgSGqaPK)
zSQFH3-Q!%3&S>E~F&Yc|l>jH3L$M<YAE`Lyl&-bB7~hA(Jx%p-wZ02^<E<J%0B_wG
z$W{pQys2mgSYQojFC0o}1~Y6`1O4;9Y6xwEa>Lh#Ng(_IBG3P#!npjU(7I7>2_#!S
z$$I_Gkh&aGf%pe!skef+3ZGPXw7ge5JH*398+NwG8XI!Zd0aV9S$70O_}Xi(zI`wB
zoaIS)O!#|YZGVIy)R@OTL#g?yh32&4ID0#`!KvQ4U@-KJ(E{Pt;ij9_)e>SB+nT!a
z*N*GG<4vcV8cK&ty`Nh@Es9#M3M1}P)|W_r*B|?aKtBG0?8bgOLJgHR(YF95nxV1L
zpli=ML%g*A_F)Tmb&)$bCQBgCc>0ltTAH0NgQdU_7_s^$^6B~nuHG^u!zxlg3B1wY
zE4+F<R4@Q+Ym7W=nKeGmQa*bmk840E9~}N`;2~i#*>D)#(?PW6|H)KufwO+y0^h>i
z>;7wOO}$h{{a5Il@2@1wbKHs(1fI9U`b0%wEHyz#+n<H_9<v`V1x#nidM*`6&nks+
z%}RY5YCf)Y9C>H(&OaK20`oa<>|evynMdoI{bgYlo^V3O3+=Wn;E9meB|$l{5@oV6
zyW9JdKP{N&1^fH2{&EXGI_lO}IEe1wjWg{#JPhKmzvFV&Ft_^n?ZnmmZ$wO2;RXwK
zKSUUj*15u8^ga7Q(;9166n|kuO)CUe)ybu_K~K7pbKQu<{mt$LoKE0yIcJ+nr{h+q
zO98>+J~ly9W*J@po!&I5OF)ma@^Rz0tDsyz=vtQM7(f;L?{ji7NH?jt>#Sd8&J3*s
zCYaO2+vq<)k8h#RS?327$|@upY+kJ{ce33Iww|s4f`U%ZlIbsT#7dms%{NY>jt}k8
zHJmh=??8*9E2OlKXCux@6~}X06y+o1hB(}dIc`5oa<z81JJMk6A2{E{{|M6<V3!Zt
z0^%^v^V};o@4Fg2=#Ozaw-YGFJM|2fSmo8izy!wqk0K;^@Go%w#;Y74qgWT956uw8
zye95j#uoMJA#Ud9lO-&dJ&`<%Kz(CBIu6j+Z#>5R<f<#!&*&~&%pEZr8epU6t0NwV
zg*>2eW(mo0Aduc5=g-|B4Ji3hfc~~=5=<fgHLQ`rGbOjcP9?->l36;~bFMY#eEiQz
z)StOcT*QtG5iimE7iZV7yCw!72LR;?VA|z98FCi8J{ebPcd@kAyhMkx#SR*NW^t2F
zZ&vFIB%V=c!`kp3C_PXt4{#pQU1MSm-2T9tl;ayTJ(d(;B}HqKGsu-?I=Ea<3d;`d
zdfUTthnj&J0GIu3d{HP$(J?Kiys~c}V7s+S^`7UaO(%z8>}Q7~pC9f3zG9vL_v9)t
z1f}426%k<=xpiqd@dU&HF5c^IU8gEih0<0xvoauw^8Y4~tFm1K?l`0kEr~STrLw?d
zc#r=sq`DnnJP^Bav@IX6*SL)v=8D(d7^a=qPu=>^xi_{%>tDI^RpQt^rMu0hxTg_~
zl|9c~9g`@!fBe$nlbi7Q!UC<t^q(K>e@?jAUvGxF02!}45mySvxB)_&3bE3Aff=>1
zNhl#p&xc@LeuT`g9CTiwwX6Q;v%44G9E)m23z}lQ_RQx+S0$1V3et-EhuQ8DvQy$o
z^Yd96-`Lky(}MbS<CK~LY_~sPM%E_>Y&yAUmpn~QTnZcpvW|9|j^~xm9rx!S{*-o`
zdJAOVyKb})&Jp8&Y!O1r49guE`eR$_zs>J~W>B8mU*|1>{05LJtvRKIF}r0HeTafR
zjdPI;R)x{!($OUk^(Z@?EzRq{Z`j+HubVzwT?9`wIBX;}1R(-|#@lf4Ztx$gf`PtX
zPXN@;OCb~>s*aDb&JmV<6Y4_mq~)8u2X`(qLc(3e|FIUja_f=cDowNyKi)y%a-`0j
zb|2WC^(K$Wt8qc)%gQFAeTcqTwvEr&pS(Vg`}X`@_U9sMK#cl1-7`WT<oWTAN!1XV
zOyve9H9Leq7(1cPy;h1r$U_5QkZGUEp!tpL|JSJJ3I|>#G8**C$}p(~mh_c5Evz`{
zWORO|M#PLjB<9G_K8hLVzHiy;pTW(_8%l_@AdgI{x(Hl{<A_tMc@B7t3qSed?gs)B
zWJC81rD=?+N9ho73j`colU~ZYYXMPZpjxb&Bw5>W#2Nyl2Oz1KyOIZlxQxcmi)zsu
zs%!v{z59|%av(=EHu?B{K+&sA?TM#za&*hsU3VvE#*_vXCy<k`o*trR*>`MYZITY@
zD07`^apau(?@*}9rvA@6sr4z~zTI~*E;j*qI>38aaJ!7Q!Bpj!tH{?zClOQBTyCMq
zwovR@?PBqNjzd6C-})Oe>uRNNz$He#vE_wU?aMrX^Ewx!M$-Vvs}V2<yxFT#3%ARX
zYnDwdK|n$vgIl?nCq6jjZ|TF%sK1{$chuZRg{P#UAX7UFKHFkWX?m0w4N8lGOLnT4
z99m)A9frQlnrGjz>~+n5vYqqK(%;L&OTvu1@z&h*QBO)bYUfuVa+jpD3xqIQm<f9v
z8djC>UW|}zP}0O9ClT1z8wZ5!;6Z}06#s>*U-dSSB3zd@j?Reb_WTX0Wk0x=903Jc
z-gAy4fLn7hHNVVIVgF&2_M}2-_uzurUigiBOeQ=@QE|3Y!Um$o^RSMYwCdLErggi@
zRpPZTm3ozcJs89vUc2`VV}5r#p7O$@90an(s5q*(pj^$74zI~lQ_~a+pQdHM6{c^z
zHU4!^p2$1G*1hh-j;PL`{Jr$bxLAoc@7G-DlxuQA8Sl9DnFVfF2K-Rc0%hK*uu3Rm
zyPk*)>o3H*{clKaLLUHgbom?pGEb>7{*ck%0tsNbTq4~NS9=1H?AAkt*^fzCHm&U)
zs$C%Hg(qMm=RPLPe%Ch3%|wn$1`GjX5B%gE8zFnXgT0{pnWjEjDu{w-7yLg}uZ-W5
zng6KZ`P3dq(%P2`F~=^Q#fp_wLh4v6O%4=b|AH?#K-Xze;ny^DyOBI@8SAD9@zov4
zNp|%aYV+cNIP|};=>&OlS`_~C*Th;_Qely`TPhwHz;G;6X!fuVj19c?DtPeUROuV4
z>k4$hBHOpMcuxqEAM*yvvS2v~8})Jw52<hZumi$eAh-Z4*nr1wD>fQ4UnzBqDf0^}
zx3aVlbogXw`y!VyJOM)eAAb_BJLR><KcU<yfLCX67cV;XLLcFZA5kuF28BWVZo_-l
zq>xFQD7^)0E*oh;5pXdzObDPVPP}?c*F4$+rUB13vdh!LS{tFf2>-r}3dG!X*qo&j
zY9SNYu9`5W->otwCI9YXTp14nUfAVN+BZ~{UaiNoR4Rb$tIv+L_;eA9!7GMEgSH2U
z6KeZVQ8jH41*!5`{A%gqMKWW!YcN0b>W`0ISgfp)T^_sUzs?XjQ4C0x<Pu)^Wq_);
zkki@VMF2Y-7i@r|NfZmO@6&I1C+hP3y1)fC888NdvR#-%(p7opBb#@Ie%<VKF1rh{
zy(^3uc!{v9;mp(%{tKmnDw_iEN8U;d;BgmlzIau>27Wb80@YYC?ETB*o+CWr4}0h+
z{_}meS$!fRA|6k*XP&m!o_2CJuk3&yMEAt*NePSH7Z$(&RQ#Ttn3UYTdqQGja$;h$
zcH#biEr7V$I@$aE{}&|4D0Bb|`2Xh<Je?qR9-h{am;YWPA@-jilF4)p4y+;4QhWRe
JuKN7V{{bb;pYH$w

literal 0
HcmV?d00001

diff --git a/studio/frontend/public/sticker.png b/studio/frontend/public/sticker.png
new file mode 100644
index 0000000000000000000000000000000000000000..d04573c080e7b25d2cb3ed7c3de6ffbaccf2aef1
GIT binary patch
literal 1013935
zcmbTe2UJse`z<_U9b_zk1JXO8ihz_*M3fegP^Ak>Q@R8YLsdi(=}7>E0E+Y?5)2(;
z0YXBPUP4hwqz4JTe+Lzu|9jti*LQE$Vv(=}&oAdG``LRxi7<d_ojk#I;*UT6I0?~K
zH~QlbdJ*93_+Q6>pS=50ehT<TcURlO8~7p9-q)XhB%{y$ar*u(h`O4|ee~=$2`Ss0
zGJ0HPpmt-aZd2Zx)%&kQe19GK$e?~#?eb*#ZA}@r)(@w6KS(e%K8Oah5^h35!o*{~
zo%#Abjg>PlusvEp7xR4tHyU;9!ZG$+{9%tvjw0qS2H@sNcyfkxL0St@(b4fb#d3SS
z{!OjRb;_c{8_VH>YO|0LPj=TL;PQSxx?4il{dP3?`RvrAmdXnz<wiW^?K|<|GcEUV
zI}6<v1Yf<VuGo1U^*24ifxG@UTRzLl_4Q*@6}e+mx|DpC{M*{Nl_u617z$#Wqqa*n
zy`?&=-s!r?zuxhS)-pSXn^9T5>-Ns`bx>_YG!;K_e7f7I=xI&m=EjPidLee(UdD5@
zTe)=F4R?tvDHnTnVr)a60baVjSlkg5R9J~6Ncp$=pIq<{my>N=nV$36{_NY)=k22R
zdf5)i&F(&mTMO9f*ie=IbN@VNb6&ITKmX%GgQaVigX}rU{Eq@bId1k-oZ~w^iM)mb
zFvh6jyEDthGpZ#koRYyKCd~!RW>IHRmCD=a1vlPMrD){7T`=ALAG+{2LVEwXwRjUD
z=_rhL&!h}}WI`OI@0V+3F!I{2B6ldOgnw1Xx79h0YOp&Yu>TA1FKQF_pWlraXF<gy
zU5|p{R-)F3L*UA@0hyrmM`-JOdDy{!@jrfp?tXaxIsV4sRnm1IRj(H7lc_ROTvu1e
zNhRAQvRNy|L-&eS#y9&8+dnBQschtGN8<yqZ~dlMh%spP=dO0{_a>`!eXN~kV~0<l
zi#HrEsb2rS+#EL$3JY|Zel>V!RP}~jNdx{k)hcLKf@<%6cUFRJF|k6AGlk7wgWTFy
z5bfU&UOa+#_vc*LNx7Kr?P<q^A1a!o;-Z&WtLk>7>f2Lo?dRDl;acf3<*YrQwEai>
zQq<GG4V>mK)n2q*WnPwcKe>PY_b=?+qB}VEUw)^q7jL!Uo-<w!4;v7(*Uakq)2k;6
zQry^rygAV(FL<myA$VLu%P398Eap90jDOW*ZcDZ$$R_G(bYGRG5f@vrr$%Po5}BBY
zyt%qWzH9{jizxk8^)5~lzY*kmTyTIF0@r3vL&eOkJ%8w}0)}I>?9zBkDoJi*&zlOa
zZxX5)SgT3p8x_8Nc9Z^+@pbkL0aSYrsvitCpDeHeRk2x*#7%G)?HaT0U-F2`@4IQh
z2{qtiw0+b{AkZTpwN6%qp_K*J-wb?fB6<r@q@3XqFO>yCB^mPYE4QO{n+V>;%}<5h
z!+V)L5f!c3W#u+vR?yI43R!mBJp>ojXiUQ}qb78Sx~od8{$qi=6*n%FrP@ba9G=ag
zIHa#A3xG~Xe27?n%=dQAc05l4+?(BA3+=(C07)XZec$%@N_bjGnB)G1<}wTXettqw
z_%xw>u80v;EGRLdiATU-eV*pDR2^Yx7AvR^otw&kvK?|7+Nydnm};#{rVNdnbv6-n
zCz5%ulDgUCavIC6#jFuyg=BIAXE*+;kxinP$o<J+T16`Oyt32y;#HM*Xk{Vm)qwuX
z8243o+s~Ycczkb;plixx(L^9;g`Unw9X_rZDRCC76b}3FKE`>R&_;HqyQ>CSl2`2T
zuZzCy$Ro1<{QWHfMx=2lhB>e}D;|k`K3i^VS5qbXnpQ%mF=2i&rti$^IX0Pl9@+-I
zAB&nQP)b5#k`RToL5Htvd0Zd1=+xkCa!cpx353q!(b&Nek0LiPT?`Y-4~5KO%OpF7
zgqWO5gc|5^#s+AnM8{@*>s0ZPE{W>)U3D)`Pjs<gs7S9-7Bwr6<*hK4lUQiX@Dx9C
zy!@WEQ(B=J4CaZRTRq2!TwXy;PV}+aiZPU|NN%U?T5hNB?7#na7k;0m#(3xR{MI`7
zP!8Sk7b9)&w{6nvUy6Xii4XMY!0Ts+g)(o-2>d1g(HrNXEe(aVcsJ@TpJPPf$zSp^
z$0`{yf6vAi-J0Doh$VxoYj*pC`~q4wQ?a$JmOvrn0D+RGaFzyXN*{84+?eo<jq)j+
z`K;*nHY@%6I7#80uxUCvxY65~zH^anc_?ysw&2(Y?}#b`E-IOvR1uyE;>Xw8GS%7$
zN*Gx&u-hsbmP^Q0rU+c}eEr&Vbh}wqGitA=?L8ixIS8o-eqoHraF0j!LJGZ$Q4>pd
zR`quK<E%+csAveW2}CDI(KeA2fm9?sVuHcM6=NZd_7nOYl?)I4jYMOLuTKxQCWH!>
zaF4zL%Wd|y%_vRSUmw<ju9#E^0Hu`(v0CWXT+Ncq(!|g9SCGLZRk<6kUE{Z(waJjD
z3@~L=T)IabCAMQuv~;pOtuA6fO)Tj_TJkm}rFleG0|phqig3>3WHK3VMl&xIUR5RI
z{jn2qFZP?n$m!oX%+kVb;!JQdKQTvmC8XflD9{PX>yC<LM<DS>DCXCCQ}m%+XCqtR
z-c%M6iH&HyY9(%xGrgSXf62d3#!Knm_5|_M)y>0KzB>$mscTJ9vI#!~9x(oj;Co{B
z6s??oG$U`Yyp(J><<RS8A&V7A;X+QzS|LtNU~0V%k1amW9AXO7B9nP<fzO6_!D!8w
zN6TN{GdGCQps-+j@6`0vVl#)RTA1~IER<XK=|qI9>TO%yZ+v{Tq>!!Hp54x)KX1$*
ze+-~wZ1$i3;r56HrDtoYCwMLZY>_JC%@h!Hnw2VRnUh7nH~Kv7i;|E?^=9c2N|B&r
zdL~C7I>XX-G=Qsnd8vir9^U2K%8O~K+c9LY(%rxe{9kyuJC=@^AOZRi_9b9)b|(Kc
z0;wWn!!q{AuuZ{8Wqdg5QjdP0o}RMbNs$j#^7a|UbwS4ESDlF^>RWuy4;2K@$=e@8
zARP^a(j-Jyk2t1qc={G4zJ28S@`z(CWXwFIG-y5G<>=>?-qxJj<vYjsPjQ|Zf!e=e
z=c9l$5go)KG==L<w5KQxuey-s&fU&-eS$#Nq6;5Wc-6oNRBNt-+m%-e+AdrJ*^SpA
z#H+B<H>v!>&F@RBbykucQvA4M;rgXz<wkpD*E<oJqr@RDY-S*3{m$G6ns)k0atU&J
z#~c9@V~zcbXb*$QXEUK}H8-FUZ9p|QD_9Y{1`UQRxp$%5cT0{>5j(bbiu8zWd6)8b
zI*LvjrSW%I?3{2uX!R`gkl)YixwlePJ<$&pQPxAGYdlmo$rc-}aiish=VCH^+_6O;
zR$)^F5eWS9nyd)Ck}2(zP3^LdEypDY&Pn?M7aI$GN1lJs4M|2a**)(r<OG>)MApL=
zx`4m~GRw)gjZ%e-GQ*!b4#}r5#^h*>y~&P824H12vfF^N!4nO?P+K{d9)~;(hu_4J
z0HoF7{mp3{U*vduTY3K-e*gH&0mOO#IPMH8{K&ZGU#A>LQFoP4mxkEwouXJ3=DU#i
zX{JMOA&{1hl9R1qVeQj}zSCPK?ln_nhj{t~R5Nb;DS8&TLIyk?v59~u5C*oLd_+FW
zL#dpM_l@y2=pm$g7E?s{%*Hmp@qUCnFE5E~F8i%DdqO9oV`}l<T?rBA03QjR>=nGN
z+lK$-L<1uC?_Uvd7yi4TTK^F;>9<ppbmMrIgI@qAfe_&{_JKTBu!d&6roc5^g?;Yu
z@q5@2w?>D_l8Y0lcd0tjZ39mbImQvad`*OKzf?39Q?B<T4xswMwlJ7!1Dr^x3cp@%
zY06dQoA>eF1wjSg;f+Dm-)N|Wsl7X`LXXg(G63gMfkSnc3i>~sy<J;-ZN(XX#JOE^
zN_~=Sf5mjaOjcZmYDw;g@gL6^bW0!Jb>sWcPe;t52J^6|XGl;J@^yzBm!pT={AY9{
zK6XeU5f8S{*>523is}ujZ;Iq8`3Bx;`5Y`RV4I1WRfCSi00`#`nnIpB<x5WtF!`G$
zq59%IvQ6Hb4>KjIzJi8`;yV_VOepM4H+S2@=v8}OxOJGv5lU|ROAgK@?IuE=!#K-S
z<j2;!lWp)<=CN?re*LJ@1#3%#VCWJA(r&6g5<IHEi`u%bzK<CmME(KU2R&b+*016g
zC+s0|&?}cBnxg5bDeP0aO(j0c_uInoV>W$<@zwrJ&7F@Bm?UKQ5qo|;7)(czi)3W~
zWaBg^J3Ie$zs4hUjl@6aArH~k&r~{*?P}2TJpr1rCK*9V?}oP4A*r)76f$|CerMLc
zOp?oLwL@M&-U^$@%WA2MbMELb+njUix<d0-hWd|{a!zNJPdv2omw~!4FW=vPqrbFB
zl{NZ}?g#Yw=ewP@yWT4B<&ag|t>TrD%=JgW8Lp8l*_@+>6?)7-CypaZ=)!Ix5s~;X
z%FU>V`U%#cQ2Ry!jY(d}0HjVR@!_)X8P-IE<5++9I(i&YQEN`b75qECl`=uGdox2g
zfxvWf@V4#f#-qG?O1IoaIRy`-{hUJxHt~L%$rW#LPMiRlJP~HV(cG|M>JH%0+`iH<
z1gtnIKe8(U`+(C{-;W`=dY2ZvJ2SUnyEWx0ggZ7PwSTg=|7ur9*AA!NcN^M$lG&s7
zqQ|0Kb69sd3W*$FA}h?TH4+Gx=X&iH;xdO~q&`jH)hR1AQlIjOh(uJh|CY#~=Irn7
zL%*ArF9c8TG+gojIl?5Rq(z~XZA?nhHC6h%gz}CUJ#=#wmSFRcL&aQ`)?JUm=o_wB
z@gE}@zPc|?QQ&1EeL-_3#%=E)!#)Clb$Oc?QV9|ztst_C&2|;bFL?**9e3Rw?9*=D
zKY**6(SB~2H*v4c>{<yH`G-`Ht||KZ-C?q~>*&413Z}}V(gG#;`V0q4^!2%Pz-cH{
z>)N^@WIzyfx(Dw}q)MF>0$CXz&qO}WcKBK;y?}Dq@(>{N#v#kIJ_?GQH-j?g-7u&-
z0<Ij*I{~(E_)f6>EH8WUl2g}VEmfMGq2l9sgTiPeie0TTJ&hysgFON!z_r?m^es+%
zP#B}yus!>INCdLpqk2gxO`*u~8?~xGn6WffyNLSa;_jxa?XLWE@QeLp<evSFgkGq_
zFr%@KV+vFEXZsTXgpEQE#mGdkv9h32w_J_vzMI>@-oB^{3{)}dn_<WT;g`K=i*e{7
z8ujGY!$58)o=D2P3_ah#HAR@<0VPe}c!*XG$!~cf-PcK|GL0Zod-H={fJA1<WS)il
zJP~khBPfAzH6;;cWijwfPf%=zv;!uw+q+lwKEakH4T*;yRLr=;ziay6=$EtB1DA7a
zrC(hRRJlERE~w?sdIdQBjqVAe(3#-qq=)u-_b1KozDciFt{!P7*!$+;$0)OzXz{DZ
zZ?O<@-uH(B;<2?Se?m@Se$aeenzE6>HySz#ZJoJV7@R^NRH>oGx12URixVGeIec}1
z(a?kDPk@W{s%6f9v}{=d@d6K4EtaXat>Cj<yQF=pc}4fpzWCvTYI8FbY5*1DbiLI-
zn%x=7{;L0oGv2IrajGhh>>vrkFXGGn=`+XcYeYUMSj=#l&71qo^kpNmT6069(19G8
zAWtpFX2@lK7+rn>QtO#ZBGPOXd>9Ph$sIVk^*V;>5q6fnxl!kAnb!$#aR%Iz@c<gF
zAfJ;lK786%gWLn$pHNuM<4)|!&HlIN1=Vo*c=pwkas>{QT@}p!g#KPU?f{!;-Z|p<
zx_(fTnJX^hI8)T?u&=HR-mhI-?)!Uro|4_YE+8BWX(M<!k%zR?&8yP#T^l^eWdDEP
z_k6D)7z}6p4-C`v1&M12SAUan!yF3b#B^jcg}p9HvfOeBr+n#NSh3Ea3g+vsXbXTq
zl!=?8_wrhI&~x;kMOaCrA=LDl-aLof5`px;uTtS(T!?|Znm>lcKRnD1H#|u)Fp%4@
z_O7!HU|No#I7lkuMVh_FV1eh)*>^H77^JVQB`~41@Y*0h+e%mPBo9A`TNmDZ;wQ3Q
z(S0E?x(hTT%Jm=d8Z-i<)x5YDhg8{^DNdjkv#k{mctW8od~ftm^+tTm^Z8n#NniVD
zf=LC{?=}uL&hRdO45O7kRVI1t3YJuk%Tbp(hU>JG4$V!}!OAs@81~DCCeQDfc%b^d
zZx@|({~bpllY9B}TG6H3QOq%V)_(R^SmAe+ARy^#+2;zkz=Et=p?CdUHzth6j|H&G
zsDxMfTBW3+ec~BVvExPnO4&F<`O!<ekfTXx^mVC!hcHmj+kpm_-{a(q!6ax(Q#o~q
z<~hr|?2olum6X&gqcVq{)|Y|Kse-}l5Y$`o`1eC@a*6Tj<E>Bd_VELe&7OQT&E$h1
zx$t|+I3XxB1C(gWn(6R}Cf+j@JsZO=!9J=30v*5N-CeAbqwWOcosclT)9dYv@4?DK
z7O!L3OLI*#k;vmm9m(YJ{%<$bI*}T2$iCLPx~s#XG_HR>X~;GcG&VSr(`p`}(Rbno
z^y!InP63C!Jp<SkZ)>$h_YFwB01Ju)s8#+G=MEB5jc9tE=}5T^Fm(-61WSV$*Yu1a
z32od4ey5n{U<6;fykDyyl#X$)&#AidVP%<b200WP@`p-!9ro3Ztn16(l$k7f$>CY#
z;}&%(?VA3|_<W|M@bPWExNJ96uJZGjXv7oaK8MS~&=}|GRtI-+P>%mJ{i)**{|5+|
zK_WL(<B${WCYDf}zOHFlh}I}K=`>X`mwv^4H8acM8!C4zqZ5gJIot5H`x}Kku~U*(
z$N`eV+ra{78#E(`ok=Ce=Bv9a3#9r*hMw}x(Vf|lQmUlPegrWb+yMjG-S3{e781;^
z>#~(coj+vz?GTs3ERk&EM{i0Vc7=ciMMra&$W{zYPn$ah`DW|^EJN4!LKsqcBQ_CL
z`&@h>_{uTpUxSbAqg-wpO>E*3ht?+@-f~k$>D2ee*375lNA~U$nLHaTQ-zw0M5E;w
zbiiPRo8iYZSKNJgJf2%xX7o1NFF&u#b$cY>C<&T5LS}#rG}`+kXb;fJsO+uGZ?S@u
zzDcS2*w(7!^TwIRQ=M1yx)A+ZmcQE0gXx1t`am>-OcoPlvd!{{uJhcmzb5V5|95m{
zOqsb+3=)Yo;1Cj9oq0#*;E?%HDfbwh@Ujqz3U_dijsH7c#8LkANj+-J>9Ev+Z=9Ep
zI5LpQzNN(lbr-?l0)UYWoVVEi){1eF(4dfmmp|Iv``(rz^g6-WqOXiLU+}a>6l|Wu
z;~KG6Qi&De1w*fJO~)fpzJr+JMFfT=VcXqSt}OMwdqX4Niup4iQO7|@Tk5+l&V#CC
zLY$gB7+e3or3kiYZcwBsoQ>4GyQ8y_*F^1IGJ481MJUPScuIVQ7MJI1Opm3IZQN^K
zeBVO7ob5Z80}5BF)dj}_Ou_65Fl7Ky#kn`1f}06loYTyx)!cLEMg5#jAw@FP-!3Xh
ztH6(g;V;!bERoSRPOWT0Wb(~Z*U7u1C#p^TBZ!vf_v}h)k=mMxb#8@6DHU2Si{tic
z7P>?civf^YK*?GU9Wl>qq2z72qEi=aKeR3<91J6u6P{m~Qeb{ktcMnNtMaA>UXhdY
zZ%JU67Z5tW2mzfOJ=dkv00Nyo&WAz7=ts;6N%aoPTiO79QIS{Cx7D=7+CUTf=AYlA
z-E4eX;3*g^Z)8FO)SfbRh(b>AWlgvh!4NM-3Yvet`(Y5Gcxzp={*vzE>4vl+cXY)T
zA76W!qykPy^#wSWUE32%NBruf^wpd|C}zS+UbW|gkSc=s`Dmr1+lmJhs)69|>^QmG
z1Y`&3YjS~;ZsbQzU0k;0p_I4Z$VM?(fuSVyqlGB}N2ZpZ6YcJ)5mb=>v~2!j_UgF|
z-}_+j)HYQT6x;p`9VkvYDM)tEf2{jV0SX*xgjWTg`Lczg2GxX~$3FR64Zg-UIUiD!
zxh2%Ps^!TCV=O7|3=CP#@Oaevgsw$|?tY=rCGOP|oBeOv<XqX&{3U7RbL_+pGsb7c
z;wiDeekJu?<;LDM57l3<YUp>qj_#CdvT5}SlCgFeC}FW&UNh$mdjwOIzR)$16=|<H
z8@DG+F~&w-NTS=jLN|Zs=Be&)7fQfVasy(6J}x)gw~O?kX_x4&{u?50Z$jE`e6ou^
zcEi%ZrmsC_$g0a3GCaLR&O6oX?n!-=sV9~g_P}kE(W|Bb>+RABX_7-l74@U;7j`{Q
z&yo~SooN+{$#R`b2@v-j-JVt5kqf5VA5m6*W1drLbi}K9grWzt^<z#lu$h=qq4x5W
zj*qnHr$bd<<<jD!qm#1d_~S3Szw-%n+~8aARk(DHeYJD7I3w8{O3e~3j(-@Bl%)R|
zM9-QNkQy#5C`mOa)i3`Flo_m`vVIN>ztKM4g#>u$#ffqolY1E0TWsgsIqHn}LdiWq
zaapJ^cLM^a7z&L!>L3A9%_Etc2O~70PzXW4W3Ke6`vjzL^x4yGLi^6RZ^%AR`ylez
zL6^p?V@Ee+oQU^jzrMi%c-7S46eChlLU^=l$w88J6i7TP5NLFITnAbg-M0K~apvw6
z!9V)pWjDo#=A$-Q)RZrEtZYex^K`_`91uUK@7lWG3IHqqV{F{4p^)J+ydb=9HIa~~
zWYgG#@zTE>gU{WRGqewHgosO#pyv002~;PhD=uB&O!O!3GD&__4PSk}vZC@=XE<}z
z%P|;#%peKex1FOJPTnUN)(=wpJr8cfDnwH$w{g2T-YFVKVsc(qh={SpE~ge27$dRh
zp)8&h=(T&KXgWvxa?e9_7SWXn&X9FYHt+y_S6{qApkS)C%2LHLc-g!`K}A7<ds%zA
z$7oelTfa{Hyaq6-=o`HG$w$W{{j~MDM1;NJZMx9&odKq*WOB{m<DP;yYOu`fpb@v{
zZ<Jkadomj{%QUyJme0y8|As<{YC4!Bn38&o*EG5c`vK0r(w9Mt^x^ZJ_Zs*<nC0<0
zD1{BAx?LHPu_KwXpEtq>>HhgDf$;ExtNNSz=jIul4Bi^u@T5ya#Hzl|Bor1M2m=ti
zKh52zrI}#NxR{$cOLhPWr|`&D2F>yt-SbK94q}SsJR5Jksk72{2)vHz|CXx!->^x+
zCVFp_>E1dvvoR1K;HoH*XO*5-V+1_~6$&GDGowqtZog8TN}Uh01%XKyEZ~a3Ajw1W
zA;PX0m03b#x}Mt#FekXH8V+Q4y$W}n{YJ}f#P-!cYw|o$4Br27aHun9=!lk|YZ<$A
znLXsY=<q^8ZFi(B=F`>54yau5=$Au+BJwa8kC=S<H_d2BG<}<3jJ!8{>X_S`o<SoR
zMtgGV*ygR!ck$}2pWXA#<bH~LWj_dMn8-*DdT8F{F<M-DjqJI=8((lIDG}MXEDlQM
z76G-Vhja*&R8hdFY2Rq{xO$tqYXj}2QO%Pvz>qf8{{-PBF5(KVRPpG_@U}H7bI+g^
zTomcDG)=1<08qUYN!wb}WWUA#Tfe-XfrzCH5(s@)Qy|<p>7rTJKuZi%zjrzHCC6W%
z&QmJV(HxPD@`quV`FKPkzyTmFkx2!3qu2lz@hW-p611w$)$t1@R2l|rMac*MNF@<F
z9BP2J-EIN3vGM)gW$3`*>(y`lxdB3P>2zHvB3CF)ShnOJhSzQ8q$lzq<M5zl;G)Og
zGAAB3gkLrYj{i~-WHi29`N~of<mXJ}lQa!gr}RD4OD)awa<}slP{wVzeh;YIX3;)4
z8Q1SyWsgG7X5En7v?;Jk4W7lL1>zF*=FwHl<H_Ey&}bdAl~7nx*-&)lRIF}K8H}b^
z9`i-YrqUNH(urhoC65@bBq4KdYZzRL(TzuHOf#YW2a%_(_M+L_1_p~CDczWh1IUdU
z8k-w`Ty^rARq~XlSK-K9YavsT@5i?+e)^sIrokoKCX}SfUJ$5!f=SjGhDmR3=cLC!
ztmlX+DKQ^jdD{`>Cq|XpByHg85B<!g_ixTwkv#Vs(w8&A%jG-QlG5Cwm?Hx+il=lx
zC@!y^B@jw3%p#HOU<C<aXx&T`i2yoI<?!>H=b18vs(s;77CAzosq~gUcsx?Aw6)qd
z6BLPl1omx%(R{+Ij#5-apF{$~YvdC6;vRqjcf8D;#X+UpikfqW<W(xy)?SUV_VFd;
zb7V4+v`rHYpr$GA?E`h<9I+v$faJo2C%I;{d)tk4Z;WB<5%1==h4?%(532S-(f-0Y
zUJs1b@*h2fg<Zkxrs^Jg<=kl3n3T)rF&>`Rgiqo_X}$uSn-{;?FnT9-i}!V`eXs%v
zSKDhg6U?X!?FeL+{Fk9l<o_aS_klpGO1waF<h?`JyqQ>-A|&yECVJ~1(-&)?3jqvp
z`?9`yVy3bi$y-i>dDS@Xsu~!o81tBXf5gG1RB=@x8i)kX;Ab9>@Z(&i+wKng%;bRx
z^}r7z0J4o}0?6Ktg(4pqtw^8@YAIJzQAlx0uNzvDRf#vx0m;uWjC)i|AdtlwB(hUf
zo#&YU>rNy{4f>==c=r@143qOwP=f<>z-V~;1L-nS4=DTcSo#998;{)6ULPitUwPOV
z^mbyH6YRkQ&9x;c02R6GTdrU~NKJ*=O9(6F_4^b0BO5<Bee&Lr!>Z`=o2lxC46l>}
zgZX`{0JUE||4ZNjf>)3UUN7Hyw*sjxU>m7OoqP$UqaL|o@5APWei`VhPR(S9p3fw$
zcG6Oj$YthcK)ns>TJaEcWXcv$g7;6N#$B0C5c1{}NBm$IB`3SQbxuK$e*3eR2NcNN
zwkbjr!S%$*>z@P-NpLadpQNooOAzZsZ;`&fCY0;WfV<~cr^7G$o-lg5?CQDMnnML-
zH%TT4fY(dcARAs;@24l`c`vL?mVrq6CNYAplCI@R_>GjwiYi#`c*yp|uGMmg^DoWj
z!S4)Dpf#)dN@E~8jhbw>*CWb`>ZW67lEtv2Xn8fG1(CrT>p}|I)Ku;jdgxiR$7pkF
zYN6X9xJ5I;kme=Ev?;rpIZ?27f2m?ns<Gbw&{qo(NdE+J8@aA136s79Fn+Hw6aj_X
zV0hL3C`>u82hPLZNx#~K^ov6tXG41f1Q0;_Y>||;vceV_LuvImJp&BJy}_QH*i84r
zVq|D6bWu*$<3iCu<fq=qdP^gZqZDBki}bF|jh)wXIa0CDi1BC+&z*?prMwcqXeu<_
zK@Ce_AP@+xM__N99f!jnu>V!t$SqS;F{AO-++YRL^f1@W(xMV1jHhdMM_?FHo*Lwf
z0Ag{C=Dw<5(Tk7jSk?2q#-{@R+@&(V<Ul4{{+kc_#RLB2RpOB2Rf|bTZ`ZMp(G>EX
zr|ue#>_X!Bi00wCld+_`*K|C}I(1FT@*F6pie;$|g`Qf1j_g??x|T*O<(_2>rJJ$s
zqvee0-2PE>Sd8I`#d)6n!<f+RvxAp41ddP?uA4)uQ~O=n2ILbI0wolqE0D<TH6FDA
zy80W3U`gm%b%)nX2>~KR9fLZC*XoYt%SS0IQj>UT;oG{F(!DT@KI&AXI-zr;3=eN6
zjF<8OoYA`9*Pl&iOhJtm42SoOJ&RcDUf!n+n+WiO6tg6gb1B`q)DSQK?{_o4r*PSS
zk}8>HpU3I&AH(N0y6+9iv!2ETg3X;%ohU&RDqRqPa_5Hxcv(i`FNtjEeXOULNc*5K
zr83sUC;Un`QL@j{ubTbG%dbfx8##KYw4oCX*-Y@%7FWxTzOcM{vkrIfk>{@#EukJr
zlm>(tecug{NBH_&4!o|p{24}z_CdgSEObFer=K-NY%&rEjImaH6dMd$If%K;f!+9+
zn#B<?e0X+7kp~pZ(jyKkrg|TP(Y6d1*$ZcGYf+1)-esPDExpOsEe4YJE8m%5C&X*^
zJYRFD9+iNCj?HX-sQlXM@<BckW$TudT}W}LJ(_`;qi^x5g6SKT+4<&Un4?pINLE%Z
zwI7_AIx&M|u6wSty~jTfGJj!L1A3wUE<_1SfIf6qC2!on3j^~W@fkNxp=$tKF>-wJ
ze3~FpJSn<kwVF&WOP^Y;W0tZ+$sw3gj<dZ>QFi?P*XqzU@BE>-qu3?_3ou#lA)f{c
zk-&$gCjy0gRp~k}DDv-1>sEo<WYNJ)0DS|3&HAF}B1aK&dYQ@#qnSt3t<+`@^*D=w
zJy?V!{dijy4Z%?UVC<{HK(SZmg{xK~15X$+i3no^EfR^e;<cR1A(Pj(bbX+@Jqz$!
zR!OH_mYswUo8XCwT|yO3@8==^xUY49%ex=zGhcdqsH9_TK032u%)my@$<mGr61Mf*
zj)t1_mh0F-b3iF~q2=6}U|YyI*`HU5IsW+S4S$$j)93hGZAvPv;7TWAic}LJ?LS!e
z|0UZPVX$skZ|*?plf%)97gy%2b)dcYQFQ`Q4v~-*ALN<I^f0|=C-Nc;lbC?5FYHQs
zv`D8`Sv{JB1j(~M>st60%g(Twe?BT?_SZQK{yZoi#}}J%%Q+<VTT_HjLUCcyVbX7f
z4J-5{07nL@CD0vs-{Mf!a6R$!IY-Ff>gXCLFIhglq%lP>6S}4ccvz5OjPn72e*K>i
z=m!PG*@@ITQP#=rx`!RodQOAE#rJM4=exdb_4Wa<#cZ-KcYsj-Q`~kn;+sr=BoT-+
z(W(B(55)+mS#W$^AyAm>p3lm&bA;$Eoltis4Tn7Y)<wcT4T8lE-?violK6F;FPEc=
z=F0b-T`Vn<yG#0;95*E?WwbYR4PV&>g#!RM;r=10tBGLAOr0y(IUI^%RhGOJx(HOx
zK1M)p9dQTvomVf#rLXc`hNh{CX*PPfEuLbZA^;lgjFau?D-Shz?6a>K$rN%=P}|-=
zaS49pyDcryW09KJ4YaU5wWqRrxxn6^d<r<4C_8mSf;oZI3RqljEtr!Vo&`bC^B!|I
zPfK1|Ez>p#ezr2~L<(Y$>5BcTr#`#;6^OGB^vmhHW;^i35DxI^Uq_ocXwpqzl7csK
zqO`>xKatmYe=^LP>1m~n!O9dt(YHcRTP`{*5cKNh5SPqOos=H}YEdV33xP?DyWvtA
zfs8HQv`0N~Gg3LEwoasL-&tUt$^o_8T{Ex|CHL|HaKo0li9j;?ia6Uu$hx2Yp9+AU
zKg_z^1~~GrTziLNp3S_eI0$si17K32Qx1-jw7Nm4Wef!Xt~&)60k|g+KJI384@ziv
zx5S)ubd=;&)dBHVZ<`tfzxxVLB^Gi=hO`(<&h*yqOx^k*Hnuypx>G*0(|hFSiC<nh
z&@-II8XUYbilmXzM*7=e2{Ied#DLnazmoUq((;<a#uZv>7!2m!a!x{OG}lT(*cGyF
zsVJfd)nBn&@V)qE;F`W&OfPp6VaIEwxZ!VY{WN)g7_9^qVs`TXq#nc}je$-ehWydz
zer29D7;ZS%^6bTSsEH#<-;C>;WuobwGLaV+^u!Ho(OY0`qq0}1!qtT5;<0>X!{x)3
zT2qn4;8PMdoA_z@2#=zRxC`#`?CEB*P5VBK#e@6${2?<cn(?m2H6C)+;;<C|0$c)Q
z3wiDkh%gZEwS$R?rf5QA8US}pAp7!X1i;(riGtTHi;DH&_hQxPTl}}D(UnH{Enh|v
z6uKtY&~vtuK<Fyv6XwQeH4*rhfBI#>IUt-Qz^U5t1^cK>2!dFqR0>8<O%W1^1Xq_;
zByu7;&~q;#?YILpa64=5{_O6a({hOkqlNung6DfzTpbkR*EQ5_+-!AXJaH9dvepy=
z$zAOYfmq~-;08UQ+vloV2g@g|#V@Y99x?o-*55q;TlFyz16X2n=t}6Dc!Swxpi%{3
zx0Fz6v#Ygk90DR@U@&u%6=c8%0*wCA7ei@;BBg|6j%LD!)Z}u?<nuaaAebyBKYM^c
zl11mQPfY_1(=+#LDElpDGC2!fidG)}SBCNhozl}EQG-G+3X5!}KSnEGupkuRd^CQb
zFIY@&P$LY(T^!QbLHEI|Gszo^E<F0pLluRzCxBm+xmCX#XMuo3R8KY$L@g@Ik0LO0
zISQN07hT_#f$T|jFWm!tKV<!i)(0Zm`x?d5yD$v1$Mq919uPk7V;=z6n%=*9=?3k5
zS`R`dKC@CASgR-d_|sC^SaLvO>jU+6FdoD3LuJ+xbn{vh!%k6Q_y(Ve+gJT*HLUT-
zSU@KHmn|2VU~-gt&}lWXXl2jrP8h8!jgp3675_2CTyG{sJd|xB(7RYd2IM5qdLizX
zdm_hR%ICGOmu>qvN{Y<PLF)Z2z|5#{0mqQpfhW$y;28mD*KPlI=t9A16GorFnxH~~
zeFZe?(S9%g@|5Bbfqcx~VMgBsdhO`I^pw!o`4oTtLq!5|dqDwQ2@+wK6!n`T*g#{L
z61TU_q4$FH#IGH9ZhH8@O`z^rqjq2+!2d-v_=W%#X3<F<dMS}MkLCphh*}%fdjV_!
z4G2R1qq2J(+k{?V8w_R)jQ%~c{ola(8qgTaHns#YkYs+UMM)R7oN5`sU^Z{%Cn<O#
zM=q&xNLQHaOoF?Z7)j2s63~<csjAPKkS85)4_7lJ6rBunB-w^1B5aSXV97%^hxcc;
z2fr=QohOKjU~@>VleV&lTo<RAJ@mslK~VEue^|gM&`XnIm%kY&0>ROq2PeIAN<|N4
z^RlN#4G{=qp8Q}7&gmKp<k!!9z*zZ@+PZ?a{x4@t7chVJovqOVS84<5MVan5PO<ar
z>+kl)Wq@)Dqys9JU-1$-YNECuZ!=J=3U5z3Ip4S`r@%$ul4sw$f=sYCT(_t%yHYY&
zKbCVR;2K8&z<dUq$SS5*4tSEuP*Y}j$t&Qh%I%MpnVa<TmHor8S<EN?Yc6=Vk5JxM
zYY7T(k{#0h3kAk6Lp{Rmn3@_a{{mWg-X%cOh^}$FmXeLGlKmP~JHNq>i^+V+(S#Q|
zi|mPk45faK#82tgtUp^&#8#SQ6=*^0g#Z7HI%FUsf6-7W6S=6nbpR_AqkJ0-&T*nC
z#)eo5thx*D4Gd5!RxxX>VFi&ufpmVPmHH3}8~uO}fvYrVY7Lw$1HBa-(^#KJycBN%
zY3|)Az^&qpcgrvR>WNARSaUwnc4I2NGw-9#x1La#7|(G6Ug%8ZCc;aMJH-meoh`cR
zN?MG5sVpp=;;}2|Wlc}??}{ze)s*wj9G6zSF!eJePyO%o4uFYzqZ9%LmkXUQw1h(c
z4%C9dsBCT_rS2J+lMSON9uD26b~&$WHR^8e`j^Xs{shZv>D+_1F7n-CE5!vF$2K?j
z^YLjn|Ge(+9{V-_LF5=q$VSR}ntWdD#v^wa=He7VQxQA>I+Ig>Q&M;p5%ox_Q;;{s
zK#mF_x??cT6Vdnvgg4K5Ef@w0?WH4@poef5EVQ!K_Ml2A%?lOc_xI1920*g^J&pT8
zuH8`LCzkxx(Q4lWIjg}x;M2J`zI7%BohX5VR>nNhcC*O1@uhL9IY>Al{=QqFxbSgE
zYlL^6Y(=X}hh>(iJO+jmAI7<P%<pc|H?{mc-(8D?(9<|WAdIZKttMoW_ADK6(h)m}
zKm4RIbc|7?R$)VnswFbJ?|E%Y)9sXQM<6WwTt-0(cpBysOEbX>MoX{|2d&2PfeQS|
z<jBw^@?0)JP@k6v5Hq|c*9i>wO_u?*+SugLn>SRIV4iw8g1$4aT>y!}gKCOSv%VH3
zxw)Kk3=9v3%qPz{I7%!xVlau(;qtEICgXkTdOAichiI=!cDDx719URwaZmddUV(v1
zVr#x)<$coau7vtf!8-0q;je*)<<S0cik7gs#?0`V?2w8mWC~lBO5bDEuspL}5$x;z
zVMdriDjGmxfTC`kFrNvM^h016z#?D^`o>Ftt`CNFdhzfofh+y2tMI<EU;hH=s&M?t
zuB7{m|BOppv^nPgA4C(@Kw9})T+{V;5dD}L;`5h+#XDS~Zuu21V6fENV?cUk1E<Yw
zj06<K-~`Q!R094(Fh6L<+f(jrDt2~@W7vgZJ<<ojg9+Rk<c@orP`5HmSWG?T-tdms
z5-`R8)ofbzj&kqnFM{-e?t$U1?SzDot}V3AsN76P6WT;j^QU1i!b<ePy?`A)#vXvx
z_3tM1E<*Vr1NZYP($fl$I^w8bo*yB!3!O<`wnsr*sBetwIo6Yw9IoRk{@i1KLqI=c
zbq{RzUuhGQ#sg*lm%{k3j1%VtX-V9UCMF<RdfE#6X9i<!W?ISQyW~)3OTz)6Ok&qI
zMSG7;xC<-b2`x6z-~=zo-d*&&mRf3aZ3Sy-+MCzDkenhraM`@qxw)ipI9rVWSCxEU
z8M@!%dYB>!9@Vf=7>|%H^5PKF{i_S*db}$c-wBiT_DoM3ZRF}?fM!NrTK`jR03;?a
zLTs!jG+AC*VZ;WD2hUNTcSyr(ij7~F?$!P4DpLfOhJWF%{TAm~TWxXU2Zj3I#vlN7
zIjZVMHWOI&<P(u3gs|aSQMoJ(_BQ%yT}f^sG9g@mqQy=mQqesk&FzNUNIe13$ixJ9
z9tk|(J#q6H{Hr&Q|6wTE9hBVio7rXl6}Ol7f%Hol`XvCR)IPrTp{rMXv1ZA3%YyJ;
zll~)UsxvwTunZZY5;EQV$4WY2F!Vc)C|N#vLqM~78N+Qaz?D)^zkbhq%+*`f7|O*Z
z|2IG<SOD<EKE<z2Tc24CY9qh7vF`-^55^&hH?M8_U8YBn$(&#lp>?>(ma~iBH6H2h
z-^&APm$VQBZL$?mpIGsOj^o`dCwrDHoq7R>pXtij!t1AjCAHlS@Utbh8D0-F-L6`!
zTP%AVaU|)Q24nLaCVlMQS>jG|fiiwMT+oMqBh3p;_H2ZH-8!3%Wj+j4<Pxi|O+6sL
zr2Em~MK$0i26A)YE`7_oo@I9o3?!T`fXL6%Rb{4AF_i?d)3qPp6e^sdBU%v%!4I!;
zGB>vsm>FlY0n*at-}bmo6yO%gk3(u4OW%^XBmlcDDgi2<v4CDwHTp3n9a=w_l7N#=
zbOU=4;rwR7=ocfw$O?z7buh7+@JZUcUlADBb7zfX(=#FZbxD`i%Ph@%2}FH9S|QmX
zWEKJ>th)x~UTIz!B^cP^5x1)K80T&dp%!m@9zblD<NK1Q3%-KvfODQ&b*Q;h$QXb4
zPm9pA;}^h{5Pj;(Re}N(TIaiLn2#CCt70Et1vJ9D?uzSRs6Lg;ts;r>`SPF0OG^40
z=OsXra~6?kaRfgw7WQ3(8z27fvl76F%P)YN33_uMr?2<jOqmegTwqB=zD_LF1elsg
zm)oYZzH_>OLAZ&)BnynGn{sTP$33bC@4AVL26(Lz<cz$Q`n+b57gcAR)hdSzb&L{&
z5wucWeSlULd-!&Xqv)%k;>1c3^d?nBh!isPtM&KXa*M{z{e1DjuT^qqNxrrwAnS-@
zYS^sfP29S@XlwtG34yc8o+kb<9+8wUEM1zr8H&@~Zt`=<Q4xJvjH!pg;P(P6M%NTR
zmhMRf|CzFCaJ9lAV6{L}589T0PNZKaq|oAr7>aX8z;m9t0=B1aKG5c{J()Zp+i+=x
z7`&Yav|(G!3-s(Ux5Ytm9h0(%i;f_W>sN(Dq^~6`aL2+epzfcTBB<G^RV=D-WAOG=
z-(>Km>jSFHn<2dWvy$l1{TrEoOC9QQ)R9S$tGU9wuq!$U7?z&e6V?R)i_?|$3f$Bp
zNJ=wH{&-3fa(jGA>Rk8`JF-gVzA)S;?{al28eN*qnDB-Ryq?kjZCK*JiJ`xZS_Vcd
zs`Oy^O$@>;<g}XHb?YkzBjOD#;L6K{0U{w;o}_DX3P2>><*UHzOvS)UYD3Q*1~8SA
z+b^Ru`smb66+^yK=7fEY^QF#y#rbjpjhbamKxQ$aY6Am{$z&x4@4lgXDIE}eHi%^=
z%u(<k_SAfiDBW(^Rd%&?|C?P?gu}%7w<C&-mC8mIn#YAy;3H8#{WSlwY6C8<u2dJO
z{)iSdP2Sjf&qV-h9|!2CJxV&CLau+$Mkb?^0q4_Q2QP*IKtWu6m%*5zQK&BzoOszf
zktIc2`ji|=Upq#9OhLp#x5*M3jFM}$;OeQ+awa<jwX6muT)4j8qax=ze03Yg_lur>
z|KR}Gcx;X72X{T4qN~<aj{<a@OkJK87>4ChYMi)VHZp!b;bq{wXV1EoRMGRHY;IkE
z_kde=tTirm#Tn030`1fCt=xGPfQ)Ykwjyrd&H|>M*j^SWpnvSmMR)#R=Au5Vq=fK1
zAa3PH*c|q|KUdo7ge{yE9KBci>Mci8GapS<PTBJ++&nfq%t1nQl}wiO2w;rYB5}r2
zD7WeQdY%!>f&%A+7J|sXM6G`MeLt}<QF4nckFIuphhlFgXrkV;1G4t9UP41TAIw#e
zb={`Eg0#~A$P0K0(xhtSh*Vp3MG!#@Oaj~~2HI3Xet@~4W(gMHnLwa<fVbU4bQ>RG
zv?Ysvg!>5LiowtJQRM=-?L&VMYno|Wo!rc>7bksVOYY)w{Zk##R^^xSSAb=cZ`9u1
zKz<b$Q@xVdT8wRo3xEKVX2n1gZKZ)HP18FwF#p8MljxdNyet#}ubXG}q9h~L)zE^D
z*c(aMS&1BR^Dby`3?XHU=?aLFE?l=z6{@YRq@S0(5)iSoY`Zq!vpdK0v#_8!_u0|6
z&qg<qqu)>EGQsr3Cdx2xrU(QxL?W{GFMu3R_aDmS^MlbIy~ALJ#K7xn*8<+?8s7uF
z+`<72>qH-nmpGSgrR<FeDgUC4nnYv|cJjjX?p0JAz?MRMCjTLeH9BI@eWq1=DEAN;
zt|Yp@3)BtVQ9~(f6Kba$@^fWX`>QE}QWN3uNg5F9tg$G-rig3xHqlWS)rX$aN-IU9
zjg0}@rHtN57_92{{8MIN`c)sfY~vxv-qh^*#!355FcQkMB3-gK|5MuK7!$K3rDiTt
zZnJ}wb2KRO3?!~dF4usQTe52AKehh{M*9(eS?-1l=A3hwz2uDCW;WA(!85^n;v4=_
z0R0AorDTZn`H8Bmt0@B9=O}@|Er38>PiF^#((9Y?pHil{B0Hwq7m8Rg!`MmT__Ue?
zv^ZdF8u|}<!>S3c@vR>yvOT};G!v3Pm<_=_C8p4@^fvaj-(s@3q7}v|(?tdh=CIf*
z;SvHg#>~tcsud(ReakfFWMl>=x)XI3GC&}t4>Ly+cZOc)2VheFnDew1KxJdmTWpXX
z|GACcs?(P@i+B0<U4Vt-2g1ykA%N9t05PW648GNY$Q8xSw!tmdEHeOz#2BX$rE8d|
z<hOa#CmDbZs@i^AZAT!)Y&}D{TY%RK9%J+}Ix%*9TaQ9s;w6(i{5t`Lil$lb6{F!E
z3IlB5chN5a{lCV=OT%K`!!!wqcfQ)&%U7Xo1vkuVr`g;VIcl|j(u|^Ha#Z$pfV`{i
zd;g5=5WCCTuCvPjAj!8_1JW$#mQVR+zU6!6ZBPPK0GJPffiUFgr0MXg>h6YAEFbMo
z3TfibqIbo*TimSMr%Z~;7|m6{p*dv5cE^@}A1@XDx!-L}Kc67lKv!rb%<(s_U=~)F
z@40{61Uxs<NaQ2i?qsND9aEQ(Ncj}8^RE`0JNK~J|Gd-#Y1>9J#$++YY7huhT3kb!
zZ3&6UO|89V3ZrS1L?88vvI3ltD4-BaZ)*vd0B^^C66Zuh)Vkf`M(&_)95SWaYIiYu
zsmVgNp?>W@Q+uF2&+u>*mrG8P$EVh56GzE2qdNS;swAjluMuW+b&3${vO97QLKJ~m
zL=b!F36gT!-X2|vzSFcwbNKbxUD@E-U)Z5?&ncnv^TikOwa4jO*)0C-yuhv!>75-z
zrIcu`d6k2MBwzZ+68EB{XP(z_R<A>xT>;8B^@6WwsQug)V5OlWUIJSP^t1LF<Li5B
z$>b>uMo#~uGiS#w_dK`FKpO9{c4IJ#L;T=T3UIgspcurc9t$u-fKf+=>xWS%W#Ue%
zNwJ?;B9iGa04uy*fRg5&4S)&gs`WZL@t@HkzJ-*qjPIp$nRco9g^$TvzB65k*q2)Z
zL2B!e`Z7S8tG^Q>etl8A40L`ITL(eDciTPL`D+`1hiQLj@un2+J96}5wimb-9Vqee
zrg(WaGHP6aY}G};mjl{|hm0k(cY+-<&{TAh4@NRh2HLX(yM($cU?3w#0uI%oFAJTa
zG=H!~*b=!?W=~9o&;0oV7XIIka7oN4`Gk*2Eljn?dTE2@CI;ufV|%n0OLMdrW%5df
zOS)~@I#rETbrwtKUZ2$X2Eqc%e3Hrme=0H~HdWk;&A#);yL0^=@ry)%d*a}T40}hc
zk>)eA3$-nrw#^=4mAW*b(EkyrTQaV<DwEOzPcSA$*N-hmPk)sgZ*a8b&H<%AL=UMx
zLl;g1;9K~=I^lkUzaLJxWuZGU9LcH~*YvFKqtMFL*=`@_KFWZH8(l`i(P*W1nrt_)
zh*J4PehQIb+mqcDldb`&k9XssZ}2Ept38R&Mkc$*b#7@~SzqsXGSw{Enmk9sEky2a
zZ{TWw^2raT_bu-aEGm}Bn%AhnQqUMO30eCza9$n+QdP&J3f=4$`g6EPE7r#x@{m4s
z&C?YfQ!RDbAW$R@DFX7FDBE)+L3{v>N<|Dt17xh2v;R%T;tRT`qi-Xl3-CWTa4<-L
zRyHUJg-lv*W>+;a2XNpu)UX<~ZOY;iS{VnZ%G3v(^{eG~7BeuTHRwVkL_oR6Xqj(W
z`g1e_iPPp642kgozJc%`F-b3L4FHGFzSQyduYEl6*Iv#&S(J<3O_$oy?@Ttmc8U?D
zmt+%ITy|U#6mqxxbmOqxxL#boxSl8s#OEjdjya<^D=E%xY&>s?yz-~)X-_?W+DA(n
zui!6trk2PJn8oQHRGucZPW;#ARUA@x;SyjK&sh}$on94Cfx~aGGQG>ZSCfL~_{U5F
zP_2H?W04`~IUZ0>K(8c7AC2wI5-kVxR8&GON1@;?znTQ3t-CPT)o|T#*k5hEbnsFh
zNm34#$gNRW91_v{nMhptw4}5vFW~gEo*3-+OKSzyQ5`&=kdMsxIFMMA%K)3Tc9wF!
zcFm!RJZ5DG!}v%8n~ADVn~slrq}Vrt&W}R)L7SJN9o$0Cl2>|<&`JO`r<id(8YcmO
z!)R@EYC>8~Eg*>BmH?cAF+dNomZ6M3zVg%QuR-@nsnt6B>M)E(%SXUq_i(H?be1%r
zZ7FNOR!sK3+KSp<qX!Hv%@(g=6tV_ZMr@^OB@6I%k;xXa9j-Seii32IQXVy|nwpfw
zzZFmYyD;-6DXsJ?Tv8N|puIK+iO8ogBY~w25Q%e<#KJIxuf9mhytw_X*VBD9&i!3O
z691Xx-DA=}9qI3@_t~Q_SAg@bD(lsVI1dnNG<9pu>?{d=kvK(_0T9>bS;}#SO73{1
zs@!U{>WIz{Ge?)s=ROyzSb6V^Q;`sJ-12xQfiOH?2!&F)HK3-;nIPM<NqnHlZji7U
z@R28!qb0Ea*cM21dF3ed2B0O3t@r@G5Ig`|$+eKhgNB${N4rr8|H6!p6yWe#eR0Ai
zuevuA>-Max*{rUHl}MC(O6Kfidr971o>cS}Vc2iNO+c&}8Rks%%BQ8A>{D7&*rIgA
zme)G$hj>!!!5+c;dw(}oTW<K`86)a%_|p=6I^QNDAD!LfSh94DR{)Hp-l9c^Kv70=
z1@?wSwMxCKav$vPEMJ$fEeIbPeiM(Z_9g+_1Em6rODAYQs6kh;8$bK1``xHd4}iY<
zV_|$BB$fh}s$U2(;#ijBZXiLS{7n>{C2e}1w<-IX?q5W3h8&Exk(~nhs@n+Qa3JJ)
zKrH53{(;{Y3ofee4&K!!4dfpNn%kxxvBVvT0&cFRLLEzd0BQU0k!|twy};&JZIBnX
zE}do1iq-JMxxM59C-)-ES6GHrsQb%ub^F*an)Ll+#w^gL09mYyfa?QjheS=NNBvDE
z?T^7q0A6^QIqP6ADYALdv7A*<R6Z1=G%6s%UTM7T(T%prq*gW=cRB|Ie_A5bY2BCk
z-Exlqw1FIj!9>c=<UVUD$wQ;5CK;=Mzh}|u2mKO(Myt96=~<g&(dcvLk1i)0d!qsM
zp@stt56t#XKr+eZh4_fcq#H|$STVr=0gOrVsyV99HY<l$f)mT<R237c6Y%tmEepkg
zbqGMluf`fDEHiVI1F=g?2|2U-*>>BOFuE4BZ(Ot%-JcRzn&!6&o|8u?23MHD5$F6!
ze-Si+PkaC<?k&JMI2#Mt&#eG_@pU_3iZNl>sU~W+3u&)5&{yW_HIruwrKSQY4(z>J
z+A07MyKYV-9sS`6ISaV&ej>((FkUPd8SrKqxi-bYe)<4^n3M_)D`B*Z!59Ao+}H$K
zG#89km^>-_HnFcRD-@Omqmgr2$5YZ*Q%=8MDHei!FU6q?kEqU6r_aq5`8X3#cSaB{
zsSS(*d(@@ChZp=&q~TbLWh~W;Xek?Ji8nQQvrDr6S>BrVJ(^!&Qk|I$o3=((w%TP8
zP4WQv6lqknm=e;__mCF^0lTs14F__!za^zbEIFtoB_WGHQce%+s`hV?$vcK*hwFf?
zZ*RW=6K(O}r8w+VZzDkWZ|v-O0|9o_Y?T#G(+x2%?05-KM}n&9y^zMr18U!6yIsDM
z|8|c_@?ERFrsQA?2K%K^T!qDe;a8;g$#vO*bln|p&<rlq$1x#S{1tlksN?SGsM@$$
zc?n_F0fTG7h~SSfo(o#z>@Nn&j7m3-6Zkrx62Eb86zuA*v*YH|^LI(VKDxqu|6`Wk
zQ|zv&4J-+cpC_zAN7-aeuK=b?e9>pOs%@pIzG2pF`p>|EVJKirsFVI6Daw)!x=kbu
zSIz+o@;{*A`ZKQ=;uF}Pj10gb{<rmcl7P<}7QenbpXhUl=&U`D>TU3g1?^v+!I@0|
z+DLRwIET&eArF}J9{M`Ow#T(~9j--h3Ej$|VW0MdND+4JZ%K-lU%`;Lvf`t$&$l?z
z^}AID4DJOpl4icjZ(p88gztSG81R@lx}UE{?Lt+TUA827=dXe2H|sN%u)`_m{~Yj~
zOMd|PR>9ggPcbHEA(Qkwn?5T>4{)nOkZFOOg*D~!Yv+dByru{MSIP&hE2c7l*A;#@
z;*!UguRFw~QpiTN?N$I^129O|emgLj2auvyeo4C~y*zQ>`l3wSg#dyyQnmLJ0H|Ly
z_SL@o|A>3<u%_?z|NperTC{2b6$KS#g^j3y0Rge>5tf1sK^Y=j!jf$r6$FG3mJ-0h
zPy~@6APB^PLV`dIvPV<`h%DKJ0Que-SZ(b&pY!>Bzt``dUOlbXxtiqte!cGdem<X%
z+on}oAlzd!q{*8t$zoe<W_k}TS|(FbJg#+$GMnj+ab)A;v*<~_A+uCM<a(a##xNb2
z5gAtd^q%7alKt7yeNl!p33r-Ur%sD?cMTr`ni5QF#*Ltqid<DTCp3AYru!bf;7L9i
zJ1%Od6F_L+uN+GR7vFB0b<BsFoLO)lR_XSr{q7r&ZN40j)fr|MizGstUpL*nq)Nep
zMZJMhsa&D8t!eAyw3h5sY~eQ^JLSe_QCRUKxIV{CQFcdEif^q&B5%7wwpMps;gd9#
z`rCJdQ$<NW4aI7U)?D*PuloUMMOA0c-eVssGsJ>!XFa=j5(eR=@0R3*CYfDrHa7Wo
z!@`8xZ?ukvrG*Ga7~{wiWdmFIeIm3nd~3uphZcK;(9RW_S%*v%g#y7Y*0si!jP2{&
z_&$38j^q@RW36bk^+81qYpk(H#ol-Fw`W?c&ZBd1-RV*?h}<~Y&8IOiW`HEs_O8lk
zzgHpfVMW;=77#z`LN|pFaa;Z}?zcppdc~L4rqg0~9@pL7vykbXr8BTLNkOEWHSX`Q
zL^Y<n(D-|c;Z*K6L0ug`<|^V$sLC5P(zNqrbVpBTpUTWyd@<YaD1(SVmxcS*dP>Ww
zD-%2F|9D#QOETL>Df<JbnmngJPa6+++(T`97?G#YuBkj!rYEzduQ|rsw$n%7WO<hn
zdh+u8G2er_$6!K7A#9c}BWH_cX7OefYSlyz6}%p|zZ7>H*bAzyNbrO223w)@SXo|h
ze%muRTAA~h{~L{5KENd&TpgJhR!_+>-ngaxbjVbhVwjh%MK})zp(2flz-XJyCg2~J
zsGjTypc}h|nED3)W?g7>%LZsv6wOoYam*e5l@Go4%z1p2hqZ9*@DR`Cu0#j9#k`9S
zzA~`~LLIvcwjv}9c1c8-5Q(aK9Z+5j<w{VZQdQLBYKTPeaE)NBm(FF)6tpn3Q=zn+
z+4iuy_M;0IX}I5sn~WW7jAX{<C`eXi3FxDhix%pOJoYsUZ|!iSJ*u%FY8@oEnAK$K
zPf6K~@)ve()emn}-mR-}%J_WytOud!Iuw{4JvL9<uJXB>(u{Cf9#C4K%u1Cd%99eS
zZY)ahB&v6L<qI;zZ)W9}S(a>(v9u8o`&A{XJm?r>V4N=aX30~Lbyz)7SqZ4UJ=8f(
zl?^)v!&}2yy%)KAkXHA#VvP5Dk+HGey4YLY*u<n8+dYaP*4(4_cpG7Sbw8FTag7XF
z-X}nrc1#ka8`hxW8$uo0HQn3lsl|e013R(di6C6voLP+Pesf<*!<YKUBDb^K`)9X5
z5B@WR5X0YfRVo_KNpY^Ak?VUi4VrV67`^Wreve`*IpP*Ovw34@6caNRU+C}JACU~d
zcNca)&%?~xak}Op00v8L9vGeVl=SZq;sl@HJpEivR=QVEsfw|W<aR54kKfOO=Swga
z8UUw&RW5>3F;U>;QFGSRin5U96t;-UMo_M7?P)5!Y2cS3iny$ED*LOOC?4y=ew`P^
z7+cl<g$$AaB#!e!X7P<mofl_JnIBnAB^x0qhCuM=4_nH*D%g88-g#BcJilNJ2jkbm
zt3w!Y<RVpJSTP7b+@>98?B3HCnzDzQyCTNlzO*lY+@PLcI;#F)HZ8pAy+zH!oKR=P
zzdY$e#gv^r1;%FKJKODqOgr^nSs-NHuXS9h>xvjCT=vc4_>PYJYlTv@W@1X6wmu!E
z6+&&`Na?blz9zd3yzzEhy{>Kxt2QuGJpMhNA^I+TKYFsg$R-7xIUwHRLMMm)g7FMz
z=O^(8RSZyDtgdBLK4>+>2_CRkRt+i|sSH;5Qe|HY^_0i%!Ohnk%>mRgR7E5{SIu3K
z@=hION_i2^X!H?Vc0mbKR#rhNz3fv87x9=C0f%e$ZDM|W`r`APn%w!ccEzXdh#l$%
zQd@20_~}kmxl&$Cm*eZO^jrB_&S7+?=j9nKw$>hFL{ckPJ!1=J829AnO7pn=98Y3c
zFxZkI*g>p$x$Z^a%levFT&P^n0mx|Y@Bd(1Hr}Mg<0p+6J~AjJuaRI@Uj*ZH!}?}y
zSqe5~HXq`AL~bV3F=z4duk0cc#*+_Ty>ezb!jsAmcIlo0t2;02sg}*R^Iu;YbQLcc
zaT1AiL%P4=zlndixBM$f!98fyMpC>De(u7KUr}~%R2`>vjeH*%i6`6T6|K*5?jn*F
zEWG1>pf@c%TA%Gw=50>)-6hrK6;_DKFJ-jKt#CEAe=au~Piu5sgn+4dzNW(*!mQy!
z<Uw8R*-*J0gG0%(>uIRN8puZqJxc-mo-Uc7l4Uz+`q!tOG8Yc2@SaV+VGA~_&qJHi
zAj_XLtf&ye94fYHs=mKKrc(3aXPtz>s!bj~>DnW(MHqQPDN3a})py)J)v}4+F2(VN
z5Uat=bFjY6!~&IZ)I4~7hk~;j9May5s^^io{P7^Y4o^jtTzFBs^g$Pc97I?6!XzI8
z5n<ZpQdTnb3MK`aFG`0<-a?bVb08;Ye*UaMf7&S@o6=E7NY@=PPpCRO(ofw8rG}>3
zmAVV5+7=dR*pyhlG)HbL#QtGFr$7qsO}RT42nx7LU6Woj)L5e@%p{r;KKvn7XBkuP
zge)xZn_*kw|4#a3mK&U?3GaIS`ej;zj~8?3L2aS@8gH~Y<(EWXO;0Ly#z=|RDkX-(
zoGyNKDUDB6{aPs2Q^S}K5Or7z`9b^o+=gBc;UDO8JMDOE605tjc1H6=d#0AchiD@6
zVbu-h`EIFojj6^Wibm<%!ea}Cgcw7F$)?Xg^uJc>%Eq6$NI96;o>mf3(9oA|8WZYQ
z+j+#9J7`bCvCN!&(SEF{1HNXmsIDY(V)zmK(h#~4>s)mFNzioi>}1j7^ic>$ZGYE5
z{d_-d|I?=2e)xvu-r5{cJj|om>2y}O3xP&2-<eVHIoX0nJDgbmy(_B02aP^3i3W}7
zTH{d=YKxbBT^FV`pjFbgAd-YiLkgb|%07uT&^LPX<VX{=%6@y`@!YRp#}!(l_(i?L
z8Z_+w@rdqYOK}eix^;#3*&EC`<`PQW*k#wb9e{eGd|+Kt|H6HzKG##|-bPMERXyZ-
zeq`MfJ5)6{#-@1@-5@F?Ah$d4JX#nLu?bVS=mV2Y8R@bF{E!*OUm2frwGbDO7jf<E
zW}MP*iQQvuV1R0D)Mo~x3gmFm9Wt-x?>Ivrwzpp7Hs<`^lr>0*5xn8wo7K%y;qPqu
z{31`^>tNr@mvIO0i3P9E?=SJb3k@ZXaJ}-iu)<X4L4u>F`0~e5qn_%!^ob(Au54-b
zz_hSLSyiw#XD*&f>k`zYt|@w&D))LeE7<+yg3<L=AIk2on_$dmtJ^Zl=<E-@dBrXI
z;1^MZ0LZ9v3;gn`MQ!zjHa8VLVII1=Q82~BGkEd>l+y8wrk*G}MQhaZuGkx)E=R85
zv&b0w{UUi~iH=q7ZI07Dn2<2TTfd@TeV8Dl(F}1kK2Hw7wwae?ItJ1OR{#;iK&h}Z
zVPc|ILh%PJtRWO%h<lYPr|5Fzl>LR^{uRj(ZDAxYUp>;6$xo%yMCl}A`INN%uU5z)
z3!@>0{5&g0`TnqA`uGkD5^;h%ZEKeIjM(eylnazBY@)qpKFWM&LZUiiRli`CoJ7PH
zj80jQX$L}y(79)*qo83!LAmPh57=$S16=d#b?oot00Iz4@IoK#)_w9a`ibY>qLd3&
zdq(=-VX;w=ze3d<(6ibZW2`;4s{hat#p7f=Um|?hy>~V_m(d{MtebZ1U?~?taR_r?
z-?%Xmi%3qq-vqGGf0hk7DSHlXf}~fVMF$v5D2xyOi&j$S$DN`ulA7#yirdy6cX^|c
zXZGex1u2*8;hR`e|2R9I-rYIYUeF%EX!QJ5k7Rcwd<({<D~eR`Ab+6kZj<+GvDxk6
zoR;g(N;L9Z|5=T4J#EF<v(m@)CrdE1sT41;yUg-T6hyj4+&oq3=Mnrp-WWBD$E!sk
z;GvMyH&Hz9P)yf4X?7$1VXP%ImzQ%8-lB+T7_JAxU<J&qpFveyzCVgV1|(w*X4Jfq
zLA{1`M4uj7J+T}P0I-sMd8G+fwE`CEUETY*nLWs$;G7A6xzS{W`rid@U=TPxbOA=8
zv8{S@8+dzi7Hp&21-Q&(?XyBCRBcTZfUls%AHUqKH&-3CB~kG95`sGjfBY7{tCKpQ
zYwhK2pRIS?YxMx*jXll2jv`{jw!feFjC@(Q=eW>BQmW#P@}9$7<&PDVCvq{^rPi!n
z0u!#JiP;liPbyz>wrww_^`{#cqe#RmVLow0?#A8V4uD?UG?C#+^C3hO8hw$}kB=b*
z2NqYnk!E@M1<`1`6(`P5$RI83(Iv%r?7r(3kdzt`66Z_7uK5wyRBQ1&E^X#r*^;`4
z<Ze*TRSR{_ydlKB^Nx+Q?HCK6+W4#hmt|l)Tegw>6UojnJZ7$dCn?*=t<;#jBO*ED
ziL_j4ml8R$Ug$WP<GXsbilZZ**rpvbzsXr{H|3h_3mpc2L{>Ss8S5QeSCBf3Lw%3-
ze<^4TG34!|SX^P2UeHbpHy8LdF#ptZE+#LgQmUY>?gx{(Awzjpzgs^Q^H&k^N|xRI
zJQ+ZguuRp%mbKG$wl~qyZBMels8QA6_{YPsnEU<}mAX|<(!T4|iQ*26Bdz1l=QaD~
z$RZLUFwv>A_wc1}%_v?Dl~$m4>U7O~8ZS6#P_{FCgbj!}7FkrXG{7Ep-Y^SaeoWb8
z3SNHGs9JLN4s~5@eEurO+oFBi#}unp-GFJI!bvhZL@fbpOom*rfs}Zf6c_gLDI#&n
z+_@PWO(I79K(8Y%ZHT59<n0_Td9VfSI$u22JO9q%N;t|S6BXl*1!MupU>TCY{$^FZ
z1pw#z8SeAOyPz93UIQB_vZwe*Wcd?id~G4a_PSbC)7*xt>+p1MPN(2-i;_Zx4*f@M
z^L+e<v2i7d-Cm_g#@4Gv3)OsyOS2v0U7sFN{MJrIM^-L30Bgjbaff}LnQ)76;@XCv
zrYK7<leU%zoX0b!Zk358Ca6Ceed;$SsA(uACy@EOhw^SKZ|xZ4Rmys)iJ|Tlj@GCE
zA~8%ONPomgP*w{1&TA^olhag)fc!K>^mSoH0tlN|fV=<Gt}}_qwl14v{D}uS62^`;
z<_WI=mb2&zmzbS*t>&qo2t?EHnD*AdK}qEnF=>POaAJWU8Hit%?I@7|2cLLcK3J4+
zi&mOyHW*_s2G%d)u0ErVzQxm@QTLb6{^arCU^RV4SI=|YE{ZFhJAelxkOF>R81t{;
zrkz$cr8)Y#ANJ|cP0N8KS!my?er)IcYz)SD!l!u&Jw|fPwKmzZ@A-E-d0ld`aoLBF
z<q?~ZRxQjI-1Z7i+AQ;iEliVe#WxnKp^UwPIbSec%W(;Gp~81asktD_?;iI)KTvm*
zCQN!$b4iCtEOx&VHj=b-Pz%-iLurs_#+c>bs@Y%RAav=aHOq!C&1QbI`+oNDC-;Zr
zWqbsylr#BPRilr|$?4oBrQZOwM%B_UJ;6b=;=G7i_xUu!^ksX5Ae$ZkL}kTi*t-Lz
zf3-}iS-ILPdX;mt=uEdh_R0u3q5rmP4TSg0=mn2FI^Q_UNKm%QgH&EmwAS?E#TupH
zJZ#bEPg@Pkz(=|v?iK4^`>b?8cagCfriZ0LYX(#2F1dxM<oz{O{S{ygd~5d1=(scU
zriWJnVp_g|toPp<3}6|Mg_#2BeQ?78xrwl)LGrXKe4%aL0^9L!&BPZz*C*1}XFDVI
z*~m$S=+X2%<Af$(DBeQrqn7P@;<Wm8%4iP<_ypFzxx2>uw$XAXd=JBn33j=6V;5)W
z|4DYfV?6x0L8WroJS-$I5?)e#UpREtKwO-ni`Ygnz{Z0$TwNQDF81#9>klzaz#9NX
z1Po&sOyM-YQG8-IiTGYH`t_X7&*a@f0m(RNJiwt&7F;`^e&!*hgd948X593T4*mq}
z+LH5svdUzrl|sHAl3(%+T#!X0tXXjqk!b3}u6xmFcHN7guk8S`Mup1~{F5eC<`l|X
z4zaN{MzM3uQMSzYFtGii`g>0l->g~e9I=tvGknB3Gkj4*C=O*Z3*6-R?FOXfc^Atm
zaiJ!-1_6I0@zh$cr4`saHiT)NkIV9G;rhm|&%MtB1$~nZHBRuA%RX&iMSRf)@OH!y
zPvX*O`Vdr7k#GcqAQIO@Lz%(BP@OVf(2TuI#>6(GR9p^BYp+?|?+E?ewd7m`{Flm&
zM&W+8F<n+_`$7^4#o=&0-r6anWS*4$hfsE19^TA1gX5xcR$qR`&NwI>c6N0}7rNV9
zoAmw)P2OkM0qBF!2V91@T*%80nCq8}djo(QIL3>{*J{#(v-}xTwYW_lm#HV@Btf0s
zqOLc$v8ju2CR=f-I*ziaZpM%{`hM!RA9*UeB9@Hn>auVub%<3_3uj*y|C0Tg<4bt!
z09Ges?N=r9_L|#<viZ}z)aON$xFYmsb~)^<iK+uOvikm78zk+i=bU!TvS@0?uyut$
zKW4TJy(|`wb)Ihg<DD~qfopF1)_uX14>_H0G5;>{Hj?6a>n?GibgVyv!&O!@=n2?j
z$#`u2Tt<tWa7WRT8h8iX-3+q8VTiD4+^KLchn+TJy&AJgSl>ab>fiB9y@>37>}c_)
zAg0)VR~)K>S1J*mF>T|oUSz3@6dcX<cvy9%?%t^8$;_NZk){3pMSx!lEj^xqN1Y0{
zNJ~kY32m~-M;#C6#wZm{&X$F%bD&<I)Hqg#bJ2kK(dPdS*dw6!q3X)zmo}vb_zv2>
zzgo_V*d-t*lxCs3K@n(72C3w}ekX#O3_?Y0RVViazoWsdiF6laPcuYc-?WS`+?ql}
z+u5+<f6(W%av_aYsEjd_GWHm&+@RRS`hCKW*$?k;Gk0#pFUu#7SfZp=*uLy0Q;Bz#
z7jIPNnnZ!1?$k4U3UrhT7mk=h_!)_`@;wDa{Ub<Z5Ne-0Xh>mT061*g;hlMlo>^dE
z8ZejnW9i`O5HxzyanH*DXdjs$nIRZtgXUzB8@V~7+DGEyV~_3U(CBy^5Pz6<LAR0B
z?^Ss9h|8g8yXUW~G{5OlLDG8XCg|ERgoxY)+O!j)R0`ud<j)7l*yjHzfc`m`{ZH~|
zaL(&u2zpMrw?@bsWvc0-GFncj`>99BwA2$P?kJVHS|O9}0J&Ta&cWYnE5WWV;m$?V
z<40NvdIz`PK;_C3uGC50WxH<95INJTCog^hGgS7xHj!8+i-hSMO3X{tQDzh6+S}@K
z-lY2;*X=5_Teis@qF1DylF<l8A2w+;Furf~cfFIa@4^?PZWUc^;nmKv8_|0q6{6sz
zTYm5U7r7;Pr-v+y*?0d&W(zmXf2`h}a+t6g8T7(!QakRAblBwH!HhP?cf)g6E^S)o
z_s*7+B<LcxsOODFua66hYOLiVEp*@BOyAlPzNPC*Q$=mD^=11MEjBazoWY-*hPAg~
zcFgU|lX;>W#@*zigSwAsn*AzVS-`ibhN*&-ltF63$fXan)sI!Low8p?%BjR{-+4_`
z+!A;*UzuIQ=`lB?;S~3p_1_VXt!2V=R<zDd`WDiAX%*r9PA@}?kYB+WhA_o$SE44I
z3CnMod5IMPg+wA2yXhC6Jle1_dp!g7ulJDwhqd*@ox%+v^2@5iw@cp^p6M8<N<+Bz
z4hJO+RNXotZ-tOoDe4I3DmQR0?Xhn*OJlg($K7vAdaNKcxsbV7&+w(Q9_F;rC+yP4
z1q_a?zdKV3k>7i#9|u6hmFemWaSeYxt<e*1GT42r&;2^EOa&0)fy7rZvPIXPw?B04
z6Q8Li2vD}?(Z6QEu~iZW!qQ(}V_9>UY+CJNGjh+AgDN(9eiCvbOQ@q@eY{}2tj2Y#
z7Q*+z$-MPfIiA;tsJynvb;OXmcW{m&o8fC&F*D50|2!1x^cHd}_<>z8^cxiJr^ZN7
zvJ8&S>b;aJ<-W8u5J#sQzxI@p0PKVu*xoX=I!B>yqn0+g*-EPt+D|nql~@s$2)b+!
z?74j_ns?s&;qmH;u`J^f?mmWZZBK?V!L{F>YjwXsN=CjpM(`F8<q#2j(jNe>Km{Q~
zkwk*CT**yMPN*!6YaE)3LL0HwomNA{bzlkvC`?%bRtv)DDiB8Zfsb!+$-oR)%MR;)
z%d6Y2Qcuhd&MbqFFA(Y;jV<<5yRScba1pn?1gGE@Yhq=6t^0Ej%4woDI2CtLwb`t;
zgc)`-f|17j^ejb1#C?leP#2LnRgnU!Wp-EC(i!K1J01d79qZ~BEDT;XkyBDL^N-1G
z1ISXeuCW;^iWGNC)48jme-N~&gk(=C8G5ZlOZI$O1VjD!T%0EB8TcGuJm+@iJo+(k
zXY(SY=Y4UgpA^Fx+J<Sjy2Zo=K!Fkw7Z8<<1}*9ye{;5|!8AfS=|=aJIv!HpLTa!F
zZxG?@);40U|IwVUTZq(MX2r*bHGWr=AE<QCfFr}eoz${2s*cB6?zd|3h71QwNUf9Q
zm>$UPb@z$!zH-MvV+^s2gZpc9=@|?y#TaC?2E{VFpHMv?3yhAHI#fLx5$)G0D!`3k
z6RP5IP>@7};V1PDjwi`pE8ngyZGYim>tk1yF0$S5ch+)_^E=<mhzAPcxA8;(aWCrm
zOHpwHsV$TW5vQsWrr|r{lOqlX(Z3Inf0NP7r8+SV-0yZYtqtAdSBpj>5AHQEfjQ1&
z-)xvQyw3h!=Fb$>@vA{{xus?CO!GIqS$mg;e@VRe7g{uJed|p9FT<z&t<6$QuZ6Au
zn8ftV5|ne8`l?OkkFukGy7Z<n*yk*nW-}9&qMe?C%cs|v73xcQjnlg<jsT1>^?^rv
z7NnTN>C)+`AeVVs0!%0KsJBwedUCoaxGBiV+-z#i4YkkP1D@Ol#;fb4x4Z7nTOYYz
z#G5GaTYFt>7XhOU<_8~RGet?gj%<Ybqduc6;Etx`Wq_v?4?;!AYwHSn(EpEpuR=N)
z<}D48Z=T<=k?K+3U*>jNYEPH;rf{tG7B4T)wEK(FJ@ylm;~HU6S4+l?I`()XlM^#5
zGA!yzO+u3nbiMlEsW?4@R`!W_bR~Lx^n~)YIeoARF<*6tQ_cKCs(}1&y5Q`oIsuby
zm1a>13S93O_5+zozIl1_gT>6_s03`-tzw^HLn3i(_gw+K9)lfQ3t<rR?Ur~we6viL
z@4L+qF?ac&pAJlq?Pjab9@{rQ1>dfALDcgb+KtA))GSl1fb^inNZS5!9VqqyiJ~;N
z9J~K}pHFcEqM^olVct+}>bdHH9oRlADG_EuX${(QqJ}oDX+Fa|8tl;Mxhs2M5PpEA
z(m#3p-i1Ja2@Y0wZY10VDxP^o+<pD;R1k>aDWx1n{S6i7oLCC-A|y}<u;VRfm+G?x
zQd&pa21MO>>xn|m<YC-2wP+_;l+6Il^;z-x!lp|yIM{#6-ZIRUxiLfc3)yybm)!Bx
z|6a5$WAxs<2q}v3Mb~|HgnfMG9?OQ4ge_fOZQd5Wnw>_uTh$FxE{-V4?TRQ(Pu&jR
zMMcr9d|pK9aXh+PWy`ghFklPMdNY<IDwYMaS&z%x_q#dcG07V~p+<-wGBvXsQne+=
zG#z<$YXg~xh<a^5W&G>+XzT-z^Xi0-97T9&L`;M6T+IQB3E1btkWt&$PaW$9v&*z}
z)CL$~BJW9<a`+eF!y&K66EN5m^fyU;xlURt!GTAM{h5-Gw(~5Lsvs$f14&4U*BQ`c
zoE#=R9x87uYvyK{(lrKKY)7CJ;zr~hUOY-(7Q9c;n_F(VcX8aggz=&0_lWFz!uWJM
z#ae@As+lqJ;L}JM!he`SPKju5Er=^)a|mznbD$XKex*T^yhOSP7!yQv-39KcUFjx^
zZyRZjkgCP%nkhcbY}vomr(It^3RF3uc$*wX?Ht}`fJ%Ntdbv<r%h!56YU&3agIt7M
za=K+<&Jll>|5fTs6^NN11{H|V{Q5Ae?Je}dy@qhe`m5GQIqN{pJE(&~#o~^eYsSbi
zW?X1f{)yl|VN19NyL~Th4V_()u<MBz6mG|s(o^e4L$h#|4kRJfghJ20jRQu?yTP2Z
zs*jI5sOTig=dk|l`Jxtl5nrPJEeNb5_G}!bTN9GY2u2*-`t}v;*rvf@Z-PALJNL&b
zp7Rg5G48xM4IiedYh0bE8&RkYrHWjzDd_o;eha2ZBmit8WnzwMrY>r>Oj)(u*)TYr
zCnPs-76WsC>kPIB7=hmFp+yTc|AJ&jJAgz!ro_nVtVJIGfl4iDm`%o}v7oKYs(H2#
z)Dvcm+FJNXvW3{(H*CM(o|)OfWFQ5s(NGg*8d`W3lIuuvikp(F?rawowjoe5if<<3
z^i9B^G#C{|>FXep3r837E0;#?$1e6WY6#Qn=H61rXkCOw^?Zg7dws|O1CB%K##Rij
z+bf+4YydS?<g#ySAj}K(;R&yuh19yc6_`U}K^A-sz_Zj7J2N{&g_l48#)Xy~f!#e;
z;)uA!i-Y$7*v|BEgjxKss^7X};P|E$ZRjg_ctCv06%ujjqrce^<||P8DiVj||KBvj
zo6^*=x9p8mLPjmFTk>Y!8S!4@<V??x>p&g3>O62ji&k9GViP(NJsNzl%>BXW^@p~j
z-fxx$wwz>Z0s1R>b1g9N@L{W{AuZiX%c=6m^?i1Ur)f)8W3`pTe-ReU=4M;NMVw8a
z;RQmxD@6JQxcjvF1^Vb|7KhY^fE{V05hR5S%5O?WqXWG{K*~tgy!^s4B&!X)@OS=M
zC+^Pf*eBzocfriME#~fC7c0hAgIOs=@(xZE1z;U^VEQ+|^AE#h``>Z6eB?f4xzbu>
zd5ywZ-$2Tn98KqY&i#C@W+{$QNK8eyf$6feP-d2#R77cm4~z?7X5^q{@Y?QHIB!vS
zRe1EcjV8OQ5?}EG<PSh;{Thwu#tds~D;6xeD{~#If2Jga9$wWCLv%jjIq-iALW(pb
z0q-M0A{(Qw_ES&nbGR?E8)0zRBqPBt5U8t)M#dRuE-GRFCuKF8ol{4o-EM8Gxx^;t
z{FH@M05i<Hx_aij#;<B~7u=<%m~+F=vQTs@e%gHmPT4CG(K8>astZmlP8MYGBo!F%
z@w((J&}%Fz>uwcfXtvBW{LZPDIC`el<cCzyIu)w)?DN8TF-pUl{R}0J<73m?mB`s!
zhdpFcop!E>eHu!?I-nNClLIy$QwdVTtSM~JF$q6YJD%16x*$SaKu=60I?@L=g3vAw
z_Wr<M(WbKZ1g7;HLUDvZ^#6l?S`b;@zNj^rVPi9vb}?%2+{SrNy5-BPAp4Ik0pn3G
z&Og|a1i@3S{$pm<fn^)@O}u*d6;E0{+wYu2@(>09rM>{dB-*Ae<}zC&-Lsx!6G(hw
zP;Dvna^17Cm!hK4yph?Pqhi&=2O-PeV<5a6q?KFX?zz{e=1N1=bU#)r1H>?2Vyx@5
zX;l!igTfbKW5=8)^CW&g5l$k57Nk9hiiAsACR>n~DQ5(kYYEl6zxzxN!60C!`{E|?
zO8<H8ke1=BJKYsAq!D+M!q2|N#yp9E%A}<Q$W?yJXw0(nF`XjNuU46Az32|3QWe4i
z60tM7OMQz;_ol{wt<sySPV%QO&l)M~(T;*RU@CH!GJ?|d@?^AEwa<nr$VjMPIhI+6
z)3@;Q_AD-JTb>nUq%7zDzMuV$99zfvew?Oe0x7oF_NaE`ZlhdDiEk+|0EdYG5HU~)
z?b7K-wSWSAvWh2hT=Uh1yz->#8zWe^`-Zu3AMGEWg$a(5IeB<HNDXfA`>B*wLBD8l
z>(|0v_VTjWe%mj>29Io!R#^cWG|1Y9=ia`AyRRWp;QsQ+wc<oJUi$M<!Ah-RuygUx
znW;5)U~R=xvW-luW3~!T4Pdai-Uv6aRE{*cZx_;vB0m<%xIAHTeJ(Gb|8z@^mUE^D
zl-?ny%V0j-qrNaVezRk?YJcW~5fSHm&4~nEIjeF#_tJ*tBtLJR5B_ZTXa37}{s{Ng
ztm5h@^LyDk#B58bZv}Xl@tW)E=ClnF(m+F>b&TPcT5dWLFhm@CR6e|b{I0VuHjN<y
z*rbT3JST4TNDTQuD@@P74yrP9X~{nXQ-3=0AA+gw7fECi)vhgk+S>a|3v;saedeIm
z9Y!_JbD^-@ZB_h<*o8u6v{gZgDO+HV@^Xs~-76!WUTGF^=EQ*kBYuqzZxGa($!y)1
z_BGLUrJLi_*VJmM>72(Gh8&?AE9%-L%kf-!rFIfHY`wbXENAFBg_U8PP-rkH472D@
zdfAAOj&{A51J$ahzUM|JV@uvJod~$?i&h5B)02lE0zZ$7q^yeE<^Mo4coGRDik4yj
z6Cj*1nQ=O_K&`7Gnc1afXITv?Gt@{{rr?=c6g(3-PtRP2fHr_zTH|;3-)EVdfll0X
zKS&OZ)^3)BLD2D4>{t;bYmS=_Fs(L0b;NStreVy)LV9`X^Xiv{BgGNHlVVMT<Rp7d
z>I+Fzt4dvu5ONVVsX%R~K^;?#$qOq9ORm1ND01`8V#V@1c2g0uqBZ~M6GapnYcjt+
zp!iGT8ZuTeCKU!byM#76SigJJ6T_3#*RbjtzMNO-GQ@ekdx{R3z;mDqg%p(2tNIQ9
zd=GvA<Q)fQ^6S;0IQY~7|GPGP44}Z4GGBDzWI>y*a67gFyEM0f)r#*%c*9|><Ml~q
zK5di->|@cS(xjx!H?!i$f$$I$qDNGe5sN_6Ger}V`ZW73yMB7({cH4cpsp?Zw_e4O
zo__f>QwHK%Y~lzcF6^VQ=K(l$>15`Ok=6Z7-CnRJk}?N#y5ufD5Mi7s^}uTuLN_-$
zQ$=7r_p#OC{+nQ%QrIcFkf>`cD~UjC7!m%q`I)s6`p(N4=-yE}>yXD+Y6l56?d0=R
z{ZA$&D_sXuxj{o_ApI+gWa#)G_mH|MyJi~H=r0nrg5i%9rlQ?`lsXO!`eQ-dGEz$9
zPuR$o+MTz%EF74u1e2>u5sh;}rRn*LA6uPUer+qWSY^eor&b`5hcvB`R+=^I2yNu~
zYhbnlQW~IpV*Sqh;4&6pL+zkttw(zbuY_LtW{4Wbgi1MBPgoI&YUMO*?;M-9tQ!Zj
z-=d<b%#T7lT%_j9s<iD;TVP+v7?VF?w!GsP_8t-Ov^~d84eZ_VF&Ud;na<pjy<&<+
z!F1}r;a_dh|BXQXXR)oNc5O@I*m9f@|GTzQ;)}vtJh5pu^jlYvuH~DPt87Q+%o8SE
zT(hWxw_ui2dh!+?b#!(Z@3wACm~Emxw<6P0nKR3T!j$krL#ZFq+yh6PDponB1ay07
zzpA>({9dS1f%K<fLsHi3Xel8QQNfC8qQHt#91T@_YE>YoYdp67ZB3sPCgWhM<3IbR
zHG-mYzow-ew61gIA3^jKVrnf4V7GX`Lc8`Zb>L$Mu)qzOx!+Qt>h4QSaa>sXFxZ`H
z%v9`~QSp_3G0-H!aQ7+q=8;Uo&OAM~>@V?hwxo-XP>4+qn=C?E%@XT=O+Y_@sZF`j
zk>ow59fcavZ@efA{`G{ZddBUA2oIJ%dCmyAql!JjY0nglpvTG!se$Q(sC_+)saX?W
zKth6x8G}zl887Q=s$Cx_H8Qhs6b5GFa*u6}5%90lkQRT}tv$pLgZVj{Y(QQq&FF(X
z!CnhxM}kxJ^ZK7WCO&5Hk7?=-DpeC!IWd$WpyyS^Wo?GCYnAg}H1XlmKn_|1RAz8m
z6J&hIN7$a&(b3n&6l;11LoWBDy^q6ue-nE0B^n(;y$UKNA^l9t_LAjP#^uGLU7=D!
za*BorXSwP(lJ!Vxhu#)zMhqE3p|REQQ%4I;26ZGn=}=<n!g7+00P8!b^yT#DXLSz1
z+e0NukgWd&M`XdCP%I3L&*L#`eC`qf$Xt~b>T=#bd@JQK_I7Ma79cDk;p;8|3?*6C
z)Cz9Td_!ZqWA0*&8@G$9TD8b|ox|CF)t+-qn*Y1@{45N?`x^`nVqd7GXbX7G$)|LE
zeR@3o`8M{fM#ohT0G<6FR)IH$Ein*4^H$RX)8-HnQnqvC?r^igs_w-*3pM!3=tv`C
zy?3jk-XDxidk%W%Z55|C`-GZb=<?<-YP;xwe|Y{u`4&-;G{yc$JziSW*2sy3q7-b{
z_w~Sup~jK3CCbj@xyV*A<D;PILubT$G8ByuP(>`gmml)iA4$*a1Vl}SYY+hZCr>lO
zo2JXhH+->>NDYN*98<bokAa~Dc=0`ptG=1mW#%hZ4w5g_E;EgC=D@!J@_x6*w%yio
z`)b}l15_qYjB(ED>@`cbBWdKy*p0e#`mUmeM`z3IEM{-v`L5cSb_AGBZhSo;+Ap_P
zQnG8%i=evf04UAj8W$&4{dq3%CwBe$MCC1WieZX|fEliH`n{|wt<}$!SQfl5_E@*>
zOYP7rXeo&_Sv}hAPPJ68S7FWfVDEOFo?f%Ir=1B(Eh8KD*Vi2eI-STt`xWo@ih}J0
z3O3vVX<^4_vWY~i&2_~7kfx7i7b=fHr*Bz^y@ej$D_OiGMWw!h5bu|;CHT&nd3y--
zy0j$PJ<q#-sc9H|4k#^-+2vGaxQlE~`{K1tOTpID{DyJeqk>vLqY`JTQd3eXw?>a1
zj+!^s-07K;GvW+~sf~;TeYB`6ALQ&=aUtW@ANA@z^YD+g*rUy}E}SaM_5fhDF&DH@
zU<ZUdM5bI%FW@kgU1L|%W-B42pTDT#34bC9ijKuLN!1Q-At5|yIjt@Rf7Tq&HK+Zm
zzkoFTQzWoX+y-_J)$LsYzx*kJ0r*93xA?dOgy(=Q_Qh=rmlE^fkoy%<nYe4XZh;bT
z$TH_x0G6;cEI;vkk=_+GV{flJ#^$pi`6wD5U6#I?KcW;Not#LBJSHktHbk<|U$1#e
zMv`#HzcQ*$fZeY?p}$+{S5G(^#-l=eetPzD9T9d?$1mlOF-B5SI_Sx&$nI;OmM>5Y
zg(X4qnyYf{KP`;lYR0yhUEcX}e#!K=0IB~{*8U=HbdSZz_(j*(4K~U>$D{~@J$`?1
zsTuPnV2Pk>zzW4*fF0v2a2{%F4-|vBm6qKdzQk5BM3c?aX9{+Ad4z>t28=&1C*+Hb
z+FO{XMKGGyy|Ozsi$;q`m!a_+k@F+LY8A)~&GT~%J59}H$B3Z_&ic|%D9-lBx94dK
zkG&m`R*d)W8ZX`@dJ5sW`+)lN!tXB0>s1w3_NlP$IpFjc%uhQtFZ0x_Bgus1BEIS6
zCLa1?Ro{;PjhFi-CAgG{Y|h_EaE^N`4q4c-AZC#p-C*xGR$kIyt#=0=)Y)!J)L4>D
zik<YyTSdmovv*11Q@~8=@l3<zn<rJ_Mx0NjKUvh9`J`7LKIMLPGY?z&>0Qk!Hu)lf
za%Gry)xk@Xs?C4$7-CnMM4k34sJF$11Ygvx88S;Mxd`qCxn0Bi2ST{}#wlCD6^Dbc
zfK*j07<oR`No$k?9*AXi13+AE)3am%tN(Z&{}VurJKFpYp?1o_6=~PwLM2i}iQa`y
zFVIgTt5zm<vkG@C0qe=m{`y`b@r#g|1IvG7<CcX@W3iHw17pI7!XZdJcMFzxsbATO
zJ53*d;qk~ln;Y01ieggGd`B2&PMyc6TvLBMx(1GQ!R~>$h^XLcO=p4{TmCy&&uKZ{
zo@agySzy}D^s4mP>jSLF6x^bzvPQ?aN)S@l#;f1Sb(OQ6ItA`iK|w>~UK^NW7ia$)
z(VMCI+TI7(-{Tuh_bU5-yph<d;F$NyO27X9i|oDGe9Mq2?6`Iyt=<=7gm0SKKj}G}
z)>(OHdp#u5EJo{#59oDtS{#`L+qLnshGw;8nX;7?Vjn#mR)w@wEFb6G0MXyG#f!H;
z&rg&w&QIj!Qw%>;^OuxQJ3-<Q7iuO{#1|W;KY3YPl_rG=pts|^n-gL4lR%q&^dE?3
z=H@PZ$iJTFxrCn54^IE&j<^T@*S~nV9VHm|SF`rsNC3OzKSzd3wC_k47>@am>J-j^
zQ|AC0Tg%U8$#8Toa|_f?wJK=2>P7yTJ~v9G2K7KzafrMC_iJe#qWL`e!m?@22a&D_
zImO$(c{01DY~+4fQ`TDg3=LRVE)wPOXkfO1#qy@uIaiVqn0z}Zkn(x`D;_w$uUOU7
zN^bg4|Kfa?_rr4|qEsmt`fMB=Z~U(FBF86-j7^tM)PyE#TP(kYVnJw9*axPU3Hbd9
zFt{2|{1l4z4dSA8M%sdZ6IvFY4s}%;z&hT2oKfDx)P?`D9+tVTr2JuHS^f#BrHRwK
zb6xjzpG*+}AkC|lDgSFIH;dhSVbAC5hZbLF;yHoDCcsQ_bn#a&w$CbrJ7+F6V>c#X
zbr7(V50xZt)e)@MzOFu;(b8!FCBh_M$$BUyk9Oor&>M0k)I0W=B2&#5ZsGmyO*%&@
zOGyzq%YxqQnGE|Oxp7KJ0}5*Ee^tB1T^=F=7@!)YVK4K<*i?A8`r;h23{1yB^OKB-
zH!-VLI9>?B<nUOrU$93%PGQjKvrfr3aNmU2$i9#t4|_A4Y`{Wj;cRmlNcP(Kzod^-
zcDFVFY*f~Kig^98-RE5l;jqH0h5V-@RD7wZWOQ)mE9cBgWsQW9jgIG0?<EE<zuU95
zly0{RswHFdjD*A`9~sGrBBQ0-!+Z!Ox)DOk24Om9wxW}N9Q=Gg@;Fw;cmG5}xF%dl
zqrOLNg*w67u=;3M6}PWktl#2OE>c032TTKXcyh!kY1(E1OEH=DAb6&?4fRuduwQiL
z9#CD!0D#GPu;re_idTkX7I_~AqQ?KTp8V<0j#%6U+;+dyKk#VQSa<0?AxAw`8*L{H
zZY40z!p>vzl5;8C08T2V9R?(|YcyoVQ9QndNGkM7FAVQzyaxFvIbtK5Y<a_tqxwjS
z>5F@5Ik{pJ8e%I7v#Bao-Zg%NO?%LCkPCtTw-Z?@cShnUfmNO5DAe_xdV4eE7Elt8
z%pJy$9EJ5s`AyxQJ)vI!N@)}^AvR6^(^^QZ7ylhPkQZ>$A7Y$;Vh28-A;6eOyOM7(
z0N_Ms%JJhPW={Rjlw-p)gtrPZ$ZihY#T5_3!D1>iNql%?`bNYid1c}(CE9bQI3|S4
z`}vwTC*-8;b-E;c;6fQU?d=*+ec417-4lU9nT5xAWoqW9B>Pu@$idfDkh?4x^woRg
zW*F~;miaBlqchQoHwW7tg^f<ag$@I^J!=t0TByUbQAE2Vvt%^vI#TT0FgZ3vbX3jx
zXa9I&%2`qaCkn~-khA7?;p8R$AdePns08gUcxGh-LdK<ym4*zm{??3}b8!5FD4s$W
zyXl4u^1C$!bboC^l7d_GxV_$7MmC@kZ-jcM&26v3fpa+S^{iIS$3A1X$EC3kw_jib
zwM~#OTW4+dv(D0AxG{xb6d=4V#3qbqamS@!>|Q2nMvQ1^j>T>b8$rc%@|6R!p)$ya
zc2mH8G1L7O3;y{x|M=H^co5=!>YS2Wh)wqY$V8qbuVG-fBxGi<&FsW=-rNU#(ohsG
ze@!JA?!O9B)c5yRP^rD*R4VMhnyx!;7ihscGAz~SO-m~$(jqJMQ@j?XkFUk^Vbmc{
zz_S*%qM7!DZz;+Y-MzP~`e8nAGDVX*N};n3(95g`p;v;FtvK?sjS_YntBWH4^bA-9
zLiANmnURYc;{x)_%T=isCt#K2DCMIKZ`PhNi3UV-)(u-3@B#Tc&P_2#SH&=UO#+XZ
zdqWNFx7%4j+I+D-9_tPh7$(PnIm)wPjkBS%3RJeP8{*#%%3m5QGwFi7`$3f*n)*65
ziGL*h7#@yp-+$AH!z*L|bTyHf0Rfjp038iK+X9-9Y5fap-K!$bbS?h7;GWf@Vm*UH
zw=jC0=8~`a3rMhCg)S?B@6GSX*e<16Xd7!Dt?Cz>X!mo#_o^u)X+XdD1-4+%384ju
zu%eiz-C2OecF~Na4A)3wuyj~g&<?)35SX^?lVNR<D`0LaAQ9-WpzH1#J#Mgi))6^|
z(k8{dYyY-Q&zcRGWX|ymPkjrbvw|_u>e&b0Qm6L*t}yinpJqZ{9KKTb`o(~qnGF@|
zsxhR&5C>w!28}%Ydh;azj}zF=%14V@H5IFzClV$%h8H(@O3Mx>I#cK!1@De>_E6s8
z5n;F5=}=!$&gzhJW9Jz&WF5maWrxtox^ylG5WrUUi2kt%QEjz~r8uvV3q!eLnVQs>
zq9U1DNw;DNlq^IbtOSt6coD;rnX$2TkdjLH6zb23PByunyJ0KURXGf>9k%_nrH<IQ
z8h9kk^{#{+)a+YOdj_X(f)2s-+Nk|*M6Py2p2TdC<?OIA^P?w(fderpx$tH_gh+Fy
z`7K(cL+zN>5lL2=O|muQMh36eB4li-`mtF!kgmghf!C8!)oy{t*NLgu9&1yKPs--L
zY=5g^-3!!_Qn>h-#V3(SdE8)7Pq`<bnPqS?{D^OmDXk@dtA3fg)_a{5H->QG4_1uu
z!8%SCBESpWRq}z<*wpP@%q$)h7-Z316a$1rL?TRl_fE!XQg3giVIt3-q+M_o83=(G
zCDmqv2UbF)S!gTTB`Wq<YpvlPmj<s@#}@(XJ>074KpOo#!UKBEH`e(zDDEX$oS#&~
zzuJJFJal0(qrV0Z_EOfyu_T^GAZ&#I<^<-J3e3NJ?MkhW$71214i7X@p~l;6E}MGN
zBOxp!x1ieEduCtEsMq3vbH8H8feMDS11v7*OH%LBR-1L09M?xk(q&+}3kArLS=Yv3
zFFC1U(c$&%OZCJW&ZVkY0|iB-N9ES-Ih*Vs`xfCsr2vg!%v=j)XR@px6Dk^u>((cs
zOv2-}O{#rWy$yYbA+>YAAqr=G9WW8m)1rTX7WUsd&i}bJe<3)0l8Gef8@B6|aZg^a
zUiy{koN!vkfbg;!d<1Zr9<I9aP!S9}YkNU(_aoJlyxk~Qj)_V_gYvXpTaQH6W8Hjs
z^>NYuCUR7xacSmT04b#BZ|^Ca9bY5~7SMG(nfJ)KeVnoP31Ob|lO<crx44#J9}W=*
zM2?du(uY7wRt4UkaL(qW7rF?^#|^q6uM9|m<e!?9$l3ep7oE5X`y|(mr)Agz>qA9F
z3MgDb53A2GXLqnI>M|R3Da78a5g#Kaees{S2IiZ7ag?C{ME%s90|J>#n^V}lOvoyz
z`<a4RQ5??JV|g<*LPmDZ=rCA9Pc%J^g49ye0qb`LetDC{`yGmmk1Puv*9U(8;)ROX
z=0w4C>Y~W;Tj=<F=_M!UK#H5sHqHa_>xxtJ?K61{)7)yDe!v5^`@Ep4_~_Mvzs4;5
zmp+kxSt~0j(oFyWWKunL_fZS!s|@B}OzYUoE2qYvrgbFjC0f42G^9!)_H9YJy9E^h
z?JI<@BC0VvuSvm73iB21=by~#rt*BqpvF4DDh<YU=7lpd9W6quX1&`U5k|z?O6m1z
zM9FcTvSrmbDz3dsRVia~4Pp3LpaXfvkzsW8vD(<P^%>rDIJhDFG^>)>BfvRKWB15v
zSQ*k^(qs3NAB6b!bnA;F=t^DVOQ&i+^lfzLF5tx&m5l=ldbu}S9#FD}hyArwCMxEV
zp!9|q-v1?e{991h6^)!_9vEd}7d|`Hzi!#laLk;!ytbeE#uVKASq&jM>G`2MMnim{
z+8(vP#*1k}nY2PPnKV;PC?qG17N^zf3d_(hRqN3<vX{}H;S4N2F1b?9%z59Z@MA)t
z0YT|2S|75QEQ&msddjx=+95kQ6+Eg_YED_pTiq|=xh{ps)z8Ul<01s$-5`;cAS2$C
z?rLb#tN*1SRu;&LVr3md$-emS<JU$z-Pwtazna#iv4@&}D1}XA{&H{&84Ra@i%lrg
zaGn0f1s(3&SP!^VfPT&Ea$h7d(eWgF6B_67qasc+;<fn6!P9ghhpLAot6c>=2h=B1
zma}Logh`#@WuNWLhvTgwXYo$^IcBw~n*w#f`%meSuO;f#zU#={1we+4@P=s4qVdM-
zCLD@~TSWHJjYJWY$W*K!ao5KP*IAVD&SClVQ0dDc=zk%Xf4O3fx$0wuAk+tD#O4#F
zx<&k7G5c6WcO@Uwyk784_(&#`BRf~Qb6X+hS~POReUYRF>*obU5OFBKYDh>Jok!2k
zy|LqO9lsr{*O3cQyFELs5Mu6n9R@@Yk}pKYUUV+Ml02xydMDa{;=B{jLYI1!)Mol1
zssn$$%}vu+mZyRK_d1e>>0<2CR?|(n3KKgOE={W9QL0mg(zr3oq<RZRs_%H?r6D3D
zl_9A;M7;K=<7^E;keF*#YZVDaDg$%p>WBtN6iT+4f&08I+H&sCv*O$4_5VgS&wx5K
z;*G`3;4Z{QOPLQXzaJDCqa4Gyxb~WdC7e=EPAbV7sd1ud)btw6skzbZ?xyw0bXJNS
zXT%M%z6%#-e)|*)efQRoImKNR_<(5s>A8~70E)kS{sMR>%qA^^eF4xRApm#8v=agm
zq5+4cWlqS2CF`2A$ORJQ+ncM2J0Gw{`tW}2b*tP0T*);crfa=i-T!ZJwExjg^XGZN
z8gRc*Zwvq>P8Hq)-{3JCQyR@2-w~_;V{Uh4So2<(ir0;UXNyx$z$toJ#WoeUJtgSm
z{83H2g1YGaUHAOxazHnPz2Zz8%q(=+y`BWA*oTW%4=2%&gUV^9%HJ)(A7(TbF1~K7
zWH%%KlHe2;dugDBRSpj`Hr)iWA-BBR-M-%au61finhrSyOW?r>Mqx5Bvd**%d9fgY
z*>G8bjCfNU`k@5=ZWYCwPXJ!>McNy1h>*z#uO6D|D7h}dxcVJ?+jUG<zz#*<Ya9IQ
ziS_pyftL0e0-p=`JWJ45VNEw^^Gpo<ira<>8~YM!(#PsxPk}&YIu`?IlXaPu|90>f
z;Dw(26ce;rD#bNB=#Cw%OxiO#e^mJ-a<`;SdKTCy2y*EUKo(gxx0H&zKL}y*V_l@z
z-VBQU((M_>JnJDcL2^pdeKaWEZT0$21ar;nd%`1CQwHx2R%|s*<GPUnNqw(F<()-g
z`7sNsOLuDAec3IE-bc`QU!@<C0$*wLs;UNI-s18K{t0j^nt@&c+&W~f3;-Ww`Vq8u
z+LR>g_fr*0=n(YLu2xzR*#EK_VTE$b@!~)_2=)*74y3Dlz8nr5aJbTtsrR}tSF7;p
z^qn3UjK0-`!DuPiq3}IT@}dMg?`h=nz=q+5dXgMjmjo5OeU_efYr^Cl{4au(s~!?%
z9=-|V;%@U@#pWd5;Rki>U1uxXEw{d_rG)qKM#XZM^EB%}^9;sgu>iPA!H#EPVyd!6
zsE)iM5-J*Sn%iGH)xE~jmk9t#K%EZ|Inu3RHLnZF8T=Ph^pk-8cTo_PWaDrfR94ki
zj+=CK?MRYAJcSwPKTGabuss2rRJjoH6aMadp7XE*G8RI!Lm?P=$H2t4E?+ogku?&e
zhmJq69b-D|Vh^Kw0j>}yy;NbIhO@mD{^|BeI_y}aYu>^iy@fw(N?S9eNd*JNI+A*%
zzIkcsW~P<ry<+;vNZTi|I(CO!%*nuReZobud}%<0Pe)DzsX@Ti>AgyGB#eJIh&T%m
zT<}9r>c9kDf8jjd*c57a2p5*(aFP=LrOIB>9B`PN@FUar&Sv6TjRl1^v1eajQ*otb
z>tTOFT+>BdWo1y53%-4jeERmN+5lQ*<#b7a?vLCF^K+>~sK!EN>t}(MLq|0;vLR4+
zYZCI@NnuweH%5(OBSZ0}M&`@6f<rM6^B40JA#BMZL9MKjjDHIcfg=yNxhXMq@YPl)
zX^U6yOOM|UUX%;_CnwfGkNZd3P(rf0E&@qW{Qe<dBCng<cc2;NgF&sA131F~dy~2J
zb8>v`t}oX|CcM_wxK@7w2V88o_1nzGSlfk<M1M}qEwp$P{>KFP&yPbNkFVKM9nQDe
z_v^}|j(F+Rt9u7#`zkGIV(EvrBP1gdv`v;1JM%r~)bH_w)wUiqb+hT>hpvs0>>u46
zmvfnQ=Txd2_AT(ciky@~Kmn%0UCt|SCsrK?Q*gl$WO;{STKwYfyvN=P#0!tB{Gq{Z
z*n3zb$Pnd)YH0YO7jz5>uBOjP()=HPB9-79yqkVW)Pfa6FyZ3|JL_9j5>YGo8<P3G
z;LrG@Z*NTDoC^=~PQ9C|0I!z050vrD3@MB?q{J6y9)3_r-T9|*R=TLaWJks-&0Bra
zzdq+t$kSEgH$J=&9|hOSTQUs&^0S>H7R~*7>q&OUs?lXavH8-BcZ7KaCrS1r=S9$c
z&BdIJW5+eAcKj3bcA7g46C<T%*LOs^YB_`9yLl~aatpXs-2JjF=i0c4kfk-(=y+Cg
znzEX|wDCV^YX2b*9vg#Cx=|S?T?`e(A2&{X-?LUi@=HKAY`=Yu0Tbi5r1X7RQ<&eN
zT_Z9HDx4&lRhAA+My^Qkz1pKlRBLLsurHthp7(x}Q#J-=sCM^zMr(5k3^}ZW;$?Ru
zQAV?L^*f{3$1{T{@+E^0o2`e~M9avPZ={d%R&5&wT>i~h)vu#;mq|UPbnRC<6J$I_
z(X+!7)(Rzf*Kqyeb=AgowR0cknDH^vgwW-qP4ll3rz>Dn7y5MYB#c{DU@BiMv-H?)
z0&qhZVjdZvO2!U0K#a649jJviynGgH7JmEydpFtN(Yr}W*ydaih+Gt`5j}F&_2j}L
zsV(N3S7jCF5;}$C-mKGihg!YhUq~e3<Ihy5*a*QnWuK9k*5z0++I&Zkn!dQy@FwRS
zY7<ra$NQIe^@NlcWNi<1M>sHkjhHa*WC(IF0jivOW0QTvT&%H4wQJ2g3l<z~G0Cqh
zZn2tsN_YYmU{dPBe@VH%9g=BYSKXVl73`z8rreNjd*AbisnuD^G!-iAFYvsK-@n?1
z86Ts)mSmeTmGHMM<li_a?jJ^N8LIj*;}THB&)ol;zz#{|eFOFy%Wdz`6cO-jZJgbX
z`w3cNC?{&en=+3FZSGNm;Bx^^ld5%n?v#BNcTy%u&W=_zK+CL`+i?988t-%Op;wVN
zY-Xw)Wm%c`L>4J5|IXU@eZnOX7rM7Wpy{_89u*I%f_J+z<sBB6<7yrfNp@hY=OQi}
zd8Q<GuO)qK;^t>)&n3AoItu^b?IkG1M1FFSm^J$S6;cHAyz~rzud}jfX#))Zv=r)z
z_vpCanC0~1kHZT|*lp3IBkK^4+55}YDY|j5HcPt4>%zVG=Ypj_j|)~TVw4d02>!fM
z-ksEF+(8#AM>!)fSVaT;w-nxQC4WjQnG^$;s;EN*T{8U+EX5Se)-zj{HgcN;@bcrk
z4T^)+fqmies57&A4RS)0`fRt!p*3vdk?leC<yir_#gAtk#ut?Jj@<|Ms<4AsHs;E>
z(Hvq~?MezJJ<FqJF^0k?41TZM`^}0`-GAH_u*^v_$MLw;)DzQhbUS)8m80ysuXy@j
zLO`!@$Z4=VlrtCiK$scM9n}5{Y^mkC;bWU37Fqu5bxRz0gguKQ_aBRNX#RqZXP#GY
zXt%I{;7AD*PB5$q8=4ur!jK)~5rw73UiA0D2lY}~7lqQd_k^5dtInS?*vG}WXl*W%
z-3CR6PW6UBQGhB)At=|h_f)5#VA=<1)Nw`(@VnlF(7cqn`VCyU39u`2F&F_^!x1-~
zapScQAyCLiA_{e!p9DQ^9m(161h9V_SnS^j(*K6SwDh5**AVtlc}*U>-S))Vj=o}k
zpN+@z`8RD<HDh|;Eo^P)6#`x`ua`fk=3u_T53;oQ*^GYbVt97&rX;L+r50~ckX1Bv
zE~1jaM^*+74JEpcQg#xL#$_$)@t4EDs~(T`b8Yl?3zrh|W?#FX>{h-Z5q&_h3$-~=
zt)a`cz7B?jA)tF_aVL&vckBd|ojUAKmxAt2o_24bGBh5lb>6nF1Kid^u-GRi096zi
zJb5f`i0XLP<l9>(SfWC<f3@>!o1r-HMH^l4)VeYh4p>##8_$OGf2pq7q8d+I-}}n1
z_G^IAH6}N9kON_3mEs%d`~-Z{@(>Z4yxY9#h2Rar-mtv~WzFjQ=DnAqAfugSV7@FO
zq~5+qp(C2<R@Eb^Kf2yT&fZHO)ti}{3OY$^xDv+8O%IHiiECk$vXDhj)~-wjPYB2C
z-;k4nH4fhA`OF81dkIAp#_5{SO$f<oj29s(1y+Fq)xbt6j_<bU-S-<v&!CX}H;q40
zd0@`*-!}gGYj+Nd^nDesX^~gOGhu81U71{js7d(`jRv9Ymj@7xEKz|~c#p*$T*I6+
z7JO{dK<&J*ywJCJ0@>l16%^ObC{YzW?IS8=X|dW?mrw28dD+gn*zs)cM8PcHG{PMt
z1E|OVM`I~9+cF9yv!x3W6PcSeSyIa*C${tTTy!6^CXER(w=$FxqeRw$tOPr~X&OH&
zLNz|W-&r`EvJRl7DW%&*#2*LU7u7euZ^VwK{!ggFmBvh<@LAKRO6?X|sbl3}Xlzu~
zw-_)mE%$de{3Ah`{Pp(U0myK|O0zzXHB5Y4ID9>W8O$*59If}JrEfg0G6y^o%;?$l
z9oTFl-+zAFOEDyTJ&z#3P?}VQaxN@O5_>oQLJ=G}(P7Gy5Cf<SCm2}=lx4mtYI2*J
zt9rQT^x~saKOdlLN$mIK%-wkQ3S7R+uD14>=ijaF2O<T+p;9-rFafKs*#&9cKQi{E
zCrZmoJimB%n>A{dOF&!dCM=vVA*N0%i_W5sjxv3C4-cvaK@8Eg$r=>M?M$=EDH5Y~
zEqiUKw%OwiY2}8OEBdMZM`3o8^6}p+4lUr8G?2pnU4^^q&U{#yFmbjDCmlLjx|xr!
zI=t{0$WImR-3IJ1O^+Q&L)+7x!06QozgC8P;rsTuDL24bLiV<6lx`!ew(9o!s(72B
zXEJKK2;tgW=w0$QMQY9+>;7mXXGsm0wMj>@YWOpE*#p!k{U<w>FojxFr|U}b<FN2d
z@W}U*I#uOR!ZUG2389q8#?dr#AerkAC30v)`j$i;Y|NaIv7N|)T(^}#+Fy18{|C!_
zHkcZXVAT&rRU4IY+<e5b@E^#fPHl-aCg2aePB0(A1q~Xfu@W{8Q!ZxH!>+I0_o%v+
zfmPwaYo=8Jq#lnsUqfPHTWv&>ebtQ-%MTQTQgUVO@I-CbFF(7AILe7fU>@FCydV(b
z@}jrGdZ~8;l2?UR{&cafP`B@y4bAo|FhjHWh55H9Rx|6h6s)(Q6i=F>OviraQX>Mo
z48|SM1@6k(j1|mc#*y&hj<hTQ;YCh(@3-H^I5qWbnp(h=t8KnmlA(yj`dXOZBDV;`
zdY`UEYWOSqcO_NpO<6QN+`*okUK&AhSKcw{k3wM^jEs8$V!U5m;w<hG?%*CP;yuFU
zU1^fCyTu11J3(nntA+vV-yw*9(eBBGDv){*{kTS*)C(}E`%N8n#EF}LWB2chjy$N-
zg{=rnM42rMrQcc9(+q#$K;XlSYf%GxMmKuVw|mCrG425)vB!SN(fftq9_Bal<TuVE
z)4|H6xz<qKhicy0aQh-q`xSw6?3XZUsr@RhW?lVDJyJse;wB3a`$uR`gF6anG?oU?
zuywaDU4WO!s2P@nzWPrxJ}OXIP0wfpv#>wHLPK^>&BHCq<i)Vpg8#NZmwm9<CaW5v
z{ttC;9!~Y%z74BLXhM>ZA{Le*Dzy}%%q;Ufl_4p#TZ+umC>hr>Wf?PNwlZXHS7I$R
zC?QkBTB(pMA(`ImyQsa}_ujwzzMtoPj`w;0>2BZ0v5#Z>_<pX>d7bBZo!7QBFwyu<
za6NG@dy@T3qc^aP=B*4P2y!kl+eAUk%!cEbFb^RT9pF@NjeY&NkIltJc0*%cJMt1k
z9H{Yvdy^j@WIzOk!|Yz&ocg4n(L}@b7XJH;EKu8(T1!`FGh&Oo@$Jyy`{*;3CA<WK
zjx1E#ksAgF8xl0+f!&0(#kk(Zn8c|*R>r&B6VQmiW}E!^3Ed8M=b*bf?q>Met<Z&i
zF=Yo;t0CUn$J?A?qYr$8<amQ)@VVj=cPM~@#T#SW`64zEst>l1f0;_F_+vqVL;K%t
zy8`y~Pf?NN($=ckprdbweT?L)ep-Fb`ojV#ZcO!42D&G9$)dvBth3E*q|h56EBSg0
zFZyLJ@BzoP;W<`32WC0gH+nkAbNT)CW{hneTDe5R=WT^_WH!;G(yVEE_}>BwSX>f9
z=C((i9}%TSLGqwD80qO2p1^?(J%<r**p2P5cUUgS&cX?l^wRV?;GB)QQcHgDgqM-X
z^HR~?IYQ(Qjjj_>_8H**#=)6v6{(jkPQ%-r-UJy<Dni36b4$LrF`r?j%{~|rb<^yB
zs4S3){H^XYn9EIVVUq`cTy>%_?koHCASG$Pbvc@!Vf6>Le16)|CMY{d7Ph}ENZrjd
zQR+G%if*i)wcqN=9aq$?2?P<uWh8=JD-EYIa|gD}%N}a}NNHL6NaB|+%b^N!)fvip
ziJo#ts2>I}?Cm3r8{&$tPY2N&n&2~;q-z19Il5-Z4k;5tHoVeV`X@Y}<JD^x_ns0U
zTxQ62?td~4YEi9+rqFm=UFA|afvAXosUx9{j|a2HJ}E=2Fs#Gcu%~6r>X(SP02O$S
zeILU(0kTo`@gC6cKOzo7mX|R~%7+z&_+a}V48#BHFdJf=8<{xhG1*xoY~@}0>D$<h
ztTA@mN$%6o?VW6@%p~1C0twH1XhMjHr3wt<PMU0S^Q}+0*)A#hb7I1$mybtg`FP@B
zxaC{9^YiQEt`5*DV2KABH48bNz8}Bcg54R{Vm6EhB^L{<q4+=v$vtlGMeB0)JFw&8
zOI)V2j$c2l1HkN2g1CU6n;ZXSjn*G2a(~kFm$!T1z;4alF?X}c`$WcGXWQxO2>~(0
zc;E0jyPtHt=ZjN8A(5yY7!B5tXNn(#o>|8SMc9?W$8xr;I^o4;-wb<0>tu+sf^}9i
zW-a91-#wq3PgdD)3X0R%HYGfsm{&L!e7(}rbRn}+lkIuj0jnNL@8Eoe%2eg}14d(P
zxjZ9ostF+p<~qvSBJ0!i62Rkt)q=IOxKchjp)Ag571(>829FNx2E4cjrdyK7vUt*M
zFf-PUvc9I92lg@CwSU^lsuCkKy*tR`w5%ztko(~fa;dMivWwh|ySe&-pm&T5WGA2c
z&$su)D}}<$b6@q)WXt*a$F^<99Nt+>A-3Lo$otC);Dv8OGIBn(aYS9wN!4l)Hn*Bi
zk?8nKnef${IN|~p!_VF)4V=ACztioltkO7N+o+MxsH!iPr@&2{q8jHPB|e&1XN*^;
zR>t6I<`4U{MYvTGW#nLR82j1TnVVQ^ac(rw$A9?!xbfFGZi0H2_pixQj0Id#!~i4r
z3H&!XmBFs5O0}P^+7{V1pjelZ>HZMz{$Aww9N&7koPpqf_=Dbnf9y?*HQ<ozy3u(5
zThaJ^un|@SuzRY?$%60m!I2U(liw{Ns-E38H@|o~cH!dgZ_xLmRdmHA{rRkPo5#-t
z+6&F9js|n`w@`v)WYIRpj8TMi^B?`op?61)iU?c0-RY8WG63=cQiu?^7UennnwfTd
zcgo$7PXX4XfzR17P$PN*{OD%2IY&|U6Ad`JWg6E<GTh&#oqDjatO7|@O<-ACDc1!1
z>V+#fD9DWN0YPc%*>#oN5;^|cV0>WsPj%s+hg_G(Zhp?|+5a;8ehVhGng|@nw<J>q
zsMI&>&yh*JK?Zho|E>|NncbM^Jj(ShuIl@!!n;azMZXqz^chq{*2rT1pzFEeX_e}-
zcK^=GbuYeTY_k9{&#$2?_>1h^o#nxly=qHTiRtfM(9Jw7KmE!++@Jg9S@6oOz`(-B
z>n#8|nx}|t#<5__M0srWs!IB-KZwWU0n$oZfbm;T;x2d4sPj*QEyu6hfP%j>La6?W
zZ9qS!)&5;BGg?MwZfmWhoI3S`qnGTGvzGDPdYz3OZRGU_)gq|TSm{#&UcGYOSMl9!
z(>@qAap;?(R%%v!oOQ@bBaLtqf;YnO4Q2XM!#3VrNr>?CY+OG(9W)x%Iq%JefU-LN
zC3^9-VZ*{yGBo`T1?{=0W{4p*c!{H11gT)L@w=3)$SpWnpL7ZFOl|Xu_Ayu%`+2}*
zeXE6>pZb*={?-_P*>`3dB_99!UmW?RMOt$G8!(s>?|QbIB3Iu>iJNfqLYDrMVuqD&
z0%D#wblY$0LqceO*);Mao0&1L_P(@a96dXdmKp#RRq@#je3pCJC>0cPw)tyCGq3uD
zDTm1vS1CyqT@+fSVAk67Whm~>UU0|GOr+WM-@10tT2vBllZX`3_aEX}hNGVua10+(
zK@5Q5Xz#bB<G~A`2hMgt?hFJWA~P-Zc_X4O5U(R^cak1)Q?s4F1-y2J(jU)mV{nT9
zv__~ru<gm0iu!Ygj1_2*uo3(i!uWU|AOxfO_Yp#B-SPO2P@MMfPRJH_Ms6MKnt}1&
z_WmLV!RI2N2LEFKRm`vD1JSn&A{X{+VrA&_nNU1mlIV9Xb&%rY8j)q^0h`k@`v!qX
zG+l6Hg$;}Ehk|*^N`gFs$GR5I&!4`Hd<plOE-baXwRB{9o^$*ogc2J^*kzWDtBt4V
z9zq;Z3)>%r(lz%@fk{7T!i{VVvl(*sv5+{E+Ex8S^Z%QaW{N5FEs%-OYy2h_S%`c-
z4VmW|&kQ|X8ikl=S$%I4{RxF_2{6t`NlX_#m(>q?_T43r4!!>Wa`%7OjwA&uUk5#2
z<u9namY#eLgY>WJv_`n&VJDt5%|KoXXhq>P_k@O#;j`9vhEJ#6svbN-UG@-8y<Z+$
zc%AaB1-^X9pt2Fc!F>E}OVRDq%q}81Faiyiq=V8*hefE6?ueo~i$w6hDs;^tO5Ogw
z-ie!ir0^>hCRhJRYyQQVPhS@5v`%&%+nw;yunNS+{H&$yW)TP&S%IU!Iu7Pt{U4hP
z#AnB0MC?}l9gY7pRKRx#;PePim4R5oIsZ0=nqcW;_vr<9ryn)d^e#B8stc>M8NRRd
zVydU#&u9kH0}MQ|vNBa0W>voNE?nBqn6ST>x_qx>7@nGnZ`L(mou`lIkE_w^bqrpH
zB3n+9+skj0N00aIw-{n^NqbtzLIV$~2m}Gn1U|>}vOT{2>eKQzOani<`WXo@_5?Ck
z#2ebbmDo31^GO%~GtBukd_4#w@Lw-6`lYn|G6}1W-Ijt3!y&en%2`YfwQKlY<4}o%
z;lP6H$M)dYla8b~to4+Qh~z#T{f;=B?}<C}hEc3;TnV?1xXg;Dl+LMVty?>On%~-u
z*$uf{A^pqLt3F~Hxshnj)ew&rfC76Fe`7Maa09GBsLCYmfwS4;NJq48R<m1>th=YT
zLdU>7?7_#qW77Y|ms>d#_fpPU21hskCGEK#Sz5M);!#@(n-9{nA<0Kx?8Hw>Aep+x
z1)&5y#Yi%Ijf_Z&zrNOgOxJ(@L)AlOJ34D*th{^2Zcof^N7{tuEg<Bol=1X77%Yg{
zr8iexx=luLVtN|xjkwUXOe^h87e*ch#we6OXzvxpG?J$hwjyVf-qignh=cJM@jd{5
z|2uAxI;jA8;VR*rz};rrkL(Y6UUp7M?W=opGBk_2A%99$5U7S;EvzzG2ZJ#)z6@3j
zH$VC%k-8+Jl$syC4-&<rz=Qc$Q$N<{6;{i-<4*y2G5871qLMI<f1RC;dU6(ot+|_w
zz`q|F-xCMj`6Z4~90mhqi}bOnRV%5~IY}y&FR;q3;JXDKd-(1RF(WIgx(A39j<p1;
zfiV5zD+@Nt6kc&#TGu&y$DmlT`Wf2T{3{{Uk}uVJ#nO0~!W$%yB3S41PVj&{bK4-K
zhZx&ERdf<lAt!$oXV2Hg^8OSUA$|`z43ALiXG3v_0YJ*Y@b3bn7dJu5)?fYoKLevT
z|0jXb#z0hJg7^Bu@Vxv8eY8Qy^4oVi(>6vjPir9F5QgbS!icsf9@SlUQh5xX0QJfR
ztD0D!tydD>;3FI!vb91Sen+-Qu<<NxInZqzpE_Lbdj_*HPMtc5{QM<UVPv}FUiNah
zKmlSC+VZ@*VZor=F07@(ZBf7uC0_8ATJrTc%nCwZpS|+{o7L%syNg~vpAFtrgwpef
zeZ-r|taU~I(E5M<nJL}*udP39_l`|7_)D~vj_+0>OhHcKg!w<(`wJOMLw-co@bGw5
zCD?aO^*CTZNA>+I9CJE$7B^4dwz$=y^of^HwP;ln>cjs&IKRF}!*~jWO4A_tffhRY
zs)8Ii+7vNXryKeEr_Wu;lM+6Syvg5^9iFOufs)F}IWx_fh@4Y8(-M0Wtq(zwwwT^7
z4oa5>)q3l(?pCr-j;_j7C_M_GO7KoaeGC`DuHWONNhn?WAu%gynsPqD>lh+|1J0NJ
zQ&eVsXh73lz4(b&Iu=OJ^CM{bF$@G=tyyv@vM(qG;~R#X`bUpk|Csu}7dHK?g%_r7
z%R4QGYn*#L8gpf4>Q8wubv-<Fy`@Ll6#6D_Q_Aw@#2q2h#hI3s2O$Jw{Zq9k7yUnW
zBnV~45Yh^{;yQ6(i<VN!f5bf9b0iSmA6_$%ZivNhZ1BEc${RK&x{0829ZyI}pW7`=
zkAv7{AE>js29dF`v`2S)kMBnL%)R<nAw_^mp?A}oEU$_H$Nn#cP!rrqUyYD`u4}Sp
z_J)r>RXf#k_p{d!Mh^+P2<X?Xb!jKPX*RdQU;@8G0=N0Ca~<=U6;kuRp!}bfe^`%O
z04piK_U9CRav&dA=Nn2I(pNK@QGe)Xg&8nL#xB9)JlJ62)X-h58J#DUQs~~fki3S>
z;E$r0HDamz;Hz?6Dc8esAI0e<U38ELr=gXQikyXr9M;&A86uHd$QwFD#b<+12~S59
zo<Ek35GdkK@gGS^4FHPg`$i<71D1iR&!iEeW_xNYg=W!66f@#egt>guK!2eOL*#q2
zzP3fag`D(L<eL^p$>yh68G%#aSvfm|r%i7?zZplr7cw!^O}YEtN+_sx+wTxa{<&|$
z&l$<1_zCdqAcn2m>o;MbAa}nv6PQ&B?g@vr1SdH~ulQlUnLRvdHG_E-9KYSAT3>4l
z=7+f0DM7QUN0fVH^m}r54==z4i|`ygqGg6Kj3KmTMB)xT4l*bMQpVQWOHS`x-a@8m
z6O4gu#~Ld*HgF|mj=+$6^(dNiCJ%{lH-=r4bn@A-b_-z*)INVVMG5i0xCn^KHKyt6
z(88l3s@#2on=Yr~(>3)j=)WQjS}mI2#8j)%HkVw`MK;g>5oFKU1PnVXAKzw}3D_~d
zPMB&HI<(0%w<_}o7l#sK9ZI6&@VZ6!&gFmdll|!Md+j>-wVX(tWP4O{(m!mfzOaxx
z?%f_2Y4OFtUbZ&-lR(&i0M1d#fJmBK-pi=Fe>wVRzsT0!Dt!ZQ$Y;1#eae&e5*8P9
zta$`Ja?T0Wt$Z_RIeOf+IaZjc(a09BwH};!v2%|<ThDpUp7IV{LaKmUDeR7O3e?F$
zK^SJ=Sn1ut{7+@}9MH$evQw!Z@E-lAW`h52bJO4Xj=*s7K=4znOVxa84RyM>CndjT
zK=-@F7wm)kqEg@5zK;Dq%XKUABjFvt8}r-b3!~J1z4Lq3<_G3e&f@L>`P3VJ0J@~{
z@Hy6n`)M1O#L>iMu1P8*j=-<lQ|?C2@ppQT+CF=P3if#3XrLRT-r@=BK^lsxGq>``
z?@vUOhqV`Ry{gWDWcx*?N1qKOOTk{yik#B>$K|$$_CS~~FKq+bo}Y&ifb;Ld?iqEa
z_)k(`j_`1-Z|48DwcrK($GFc<oQ0vBmQ2>989zrDN~65=fZ<vaJvu7@ZhWvq<JCp&
z8rbmw^mW>c&5M)wdG{?2w_Rm;bC#pOpCiCu-|<XhDggTpK@K&L=nK7)(dKua2X@S$
zpKDPT8ph7J-rr|jk=|Fwe=>AaSV;YIe6yb(N=Wze@?aiHX(sb7&;nV?s?|So?`4W6
zq|x0v!HfhgcQm+AdR6UfcEPCB|Frapmt`92D1_krFcrWqUWieCoaD1m&KM(*PzV;A
zaR^61O~pU(3MH@mue`<4sah%dBd6!RV>LE=?-23;_sH3g<lp7kOb9>`oDR&S->R+m
zo;W4<>h!{XC_GH}cETiYvL5o+yelwSp@Blx?%S<DtC8d83?4@nEpgPoA;|uSBc!|l
z%||)^)q&@AMxO=_Mo_6oQB+HWokS~PO#WDS3Tqwx&0)h|_xs$}PC=E$qnh6tgbT6*
z&TBIPS#JngAu=7VxnHRd7;##m%}AUUs2j6e-6sU%qJISAzeM7+VB#j#A(VK~=bmra
z7-Q^41m#)}8N-YqzF$TP1n19066ZW=)arEsnUE88Fudi?=+1Yt9WBkU-^`~oFfCJH
zFvz?jtrH_Ft5V}@e}p<YxLBdGP&uW_I1AeH8cc0u^CK01?F<_Q=UP+loN{9hD)$y(
zsv*Ue>(0liz!XR~{SLDwry3){U;Ed|-E3$PTOSu#nEW}LlQC<9vyZ_j#NlUCB6)cN
zE<)nyo74j>y0pGq&Mplwz-jS_S^Bofs2EDK$*h$eb^!>_<yhOdLavY^^f7g^6_z(o
z&mp2QJV2JaU!ACps_y#{()l+;Wed4y3HmeY%E&bP)-PhhC6zI8Y5h|>BMb}G-;H%H
z<eqR4dfq|b@QIj^kah$>aSMg#pcQf^hUg_w_w?;UXJEBlz@NlHnO+qNLZ}`Z8L^NS
zAFO>r^Z)n>q}7Wb7wY<ol80Zl_m>)Cw@LEh?Q<Uv+|Gxe^Y7_}w161A5|508<IalV
z(xle2NRW)Amv3c^5nv0k<z;F63psDo{fi1<V>ww&mnHhYSP516_U(OGBM~kktjU|~
zuf3G0)!Pb_&K%h@G?B`pK?r$S_82S^3y62n>g#rBsm{5gv0LJ0%s?1aHo{<avE-Jr
zyL0txK}+?NSt7QNZej?7AwG|xE%c&!*Aw0KMA*}gv}r>{IdI0Y>@0FnZ5Dv)t_5s*
z4m&kqvpF#!w+fab+wAMA*z>X7i4g}7&OBu5@nhl_DJ<RZQTEI(G2r>ked4L~7e3*t
zL%oG0s^m7TO@v*r?il|Ug*P3A{QqIe%#PZeHY5ib41d4uhFXrp4!6JfF{ZW!{7OHD
ze!W$i33GmdbJKH^9F*QgDLZ(wn`qR_`n<coQ1**S28SsCZ@*7gELaw#^^02<KPaU2
zfi-ZjZ|ke<7!qUbcj>DcBa}4Gzo+yxSb!{OA&eWn^ag*eZ!6e7BMia+2%2fkwYe5>
z!x;zd0(ar5)APq`oZs8*?W(}vDv4)o4E^tieHq*(k_Zc2!3jg_FD*YvyOITlW#^K6
ztVjq%mYuFif^GM65IOd9ebi3}V0<g$=B_ttA=kD-{H^V-2i-eo17s_I_5+W{5W9Na
zjj-a=9odrpVcoh-1}7a_DX`Ff79#7$bi^e$h)Qk<qnIC&l3)x}j4e$!hWJ24K#-F#
zrYD!z;r5<){Li(Urak>PqIR%`lFr0}$AI+l(d;Xmu8Ey6X&~+BnN=MuA0(X<ll^XQ
zRFh=0D-8G9Eg1~{gT2as9VQ4P5>-wH>f2iWI!aI{fbk8;f$`R%E=KeCQx9CQ2KL@U
z5B$>FkkjP`?@1jGh{_f_P)|P8>f3ZSLN|S#gL1XuuuC@QWe87ut02p*a$7%Fs#%*-
zmQ|=u+YhP9?eAFJWYLY^_r~NSAIBHvrCGl7n#eP(k;Mbs8v~5p9-%&RP&Nd4!5jrA
z5a(prPu9{bt^}c<kgYm=y10K%SqX-@Vpi53dNC^4`uBUN7SH}qx!M0`X@$o}u#Z2u
zL-H$(%A=(GK`}o|FN0XNgz~(R=9wcdX&rNr)mCVdGH(r=0^uO;%`U|=y-wK(mil*G
zc)QH;hy(9oimysZ3r_@VLC74V9RvzDr+WUi0yvhEa<1|cO5)N3_KKmx*Fua=D_KPJ
z&CCxr(!mm&s?1@7JE5Te(F#4wS<gTwFx3}K<$d2mb_HAS=iPK_!a}K~NL1WrLKiXv
z{64&$FRrI6oe{Z1D|q*%1(N!RRiwY4LxOpOzZ!^Z>%Z1iK3PDbErH<sfD9mcM=n*7
z@=L+%L+Mj0x)xtN51+yK9(&GGIFb*M^V+#+6>X8N!RE=larKDFAQ+31Y|aEhz{zcC
zBKmd88AU%vzD6m(##YQX>zhZM-~`kmsIE59knx#p*QnX)5+juKeyU(9KxFTrHtcQ`
z$MkHG6-`$Bt3`?HiJoVpyw&4VXvXZ35v?Q$=UOz<C6A0AEenYLrfYe5=FTmM1$Kqk
z_J(bGt`rr6@0yhQBV_--PtN}1QFcxnaw{l#`7iPz`h3-!`4;jvL0iWx?s$WaYY!Gm
z(Du5s{;L*KgGxUh<({|Lq*4v5E+I<-?SHK2Ns*8>i_lWM0`GFAWrEg_8D<Mk)a`wf
zd_8DU)e+!-CUoEOP^(#J{xAw#Q;mHK0s9)UH<m{a6yJ1V&D()OMhsbh8W1*`ED4H#
zZ2hLh`Ip!cKy5!K`<l-ULFfX^^<_Kbv!ihTwASidO@eppO_v8zMI`9cmo~O922U+W
zzl<NP8+*9m)IWC0=wK~5P&}scd8YKfjj=;di!|Jg<Bw?Du!m}Jr8|TMk!cyTzi-g~
z?Pzt;K}^UQ#Wjj4OhNJz#$2x=(jyd}o*=Hz(|D~s^yL~swAnj6GCTf660f?=%bu=^
zY5N4s#I8q)dykv7=bxF;=*+WT6v8~VEjSr4xaH1!;@)9;@1PF;^TqdE-vyI*@o{3T
z-AAIkT>@3#``#!YWb41pJl(=?fB{3k^@3XQo@*Qwe63o==-M|r<lYlHlmpm1u{b;P
z;>9{vX~{7)3-e0MsLPA&2`!_fjcRLlP}gU~iw<y#UPAuCagOG*Sj>5;Qz4(iI-4Jh
zk1CZx2_?B`MiSi<Qp5|{TD@m7@@2F>-;``G;2`Grs&1?EqZ4S<DRlz=z&k4?U-Oo<
z-y?mnwy!>(Kla@1*g@4K9GxjPrqE|7&c%I)CaRmL`rShnV|8*pXRLaMLd(QfBP>=c
zmLLl;P%-6eFnn8%7;a9$?>n3L*dML-IKprCveLyeDz&?89(QD<Q3SR7l}~_U>YD;o
z)J~Mo`XiX0;D>jwor#)p*~i5~%hGoKpomee_Q+!iv3Xu>WTw)fX2#iLuGPT&rLVj-
zez2b=TWW7LlkHa0w!iDC8IhN1^9G?hvwO*mAO8~dQmp(A_-o6J8T8mfzx0Ps<VyDY
zsU1A)*6Z7_kt2z=B}Dg^CGpi~D*fmy%sV$hE~w9@(#~vD6x1x)aV`SHb(~Crv51B}
zrs>(l5fw=qbv$vdNXX~Kee45{WGu%!GFg76bt54>4u2q2CmxTV(1V1=F<RPoX@^Ue
zLva#2WrXGZ_z7s#u!Gl_V6Xop_BhPcg$xYX9p(>1#DI5VB(ZMb&L#{-%n@M0EnXV6
z4Yt1yZGv53_SJ6>L|TN$x@1pjbYQ3o##X%tb_TDo;1@?tm+d+{sHXd1Ummlgg?Uie
zIjf#Hf}K^hjC8GBabAGSqq5LF>MrmNT#8OQtj+i|oO4d7`ii2)Bu!NL%sQRXf@$L7
z=$%4`GrsffHy>6b1@19Ru)xt9V(>ocXw+#{4ocBlX_OBwCNEV=5>+L&1-w+p*E~+x
z)E|f56%Q*w6Q38QaTzpn5Mx)-oF%RpMzn&4mGo#cE}wbk$n^!{yB4w=1~O0Npvxbh
zsMByWuAg=K9f#9x1_O6uBJ#&EooIZoSNwJqN1I4Z@ORhS!IQh1*GDAAP^pAACkmPE
z@O7#<*fzJwAH7tpnge9)W5al<<mp6fdiP#M6QiZ`*t#-zSKqd?d4<25{00reeG(dN
zL+&n$9bqRwIT__)7VmJ>&a~30Z{^+WJUuv)jq(oS65Pn)v|VeQ`#fmIZWtoZL$JV+
zcnH%OEKVZZ3%HqbpW^6YRzZ$8(c(fGvB}dBM0dJG@?ad^CgpKtCM9BTAiHN;cQ3E#
z4mhK6L5Vs)Tn%w;sUK<V8Uf#|zTh5*BW2%{zyRzPzIsSh{?R=g$@H^ZD>+ES2wU8&
z?|_YrxXr3V7Rz`dubxLR7rXju3UqJtn7z)#SIVV?K!?0{)}h1^q1`ZlMNrE0(bRh#
zo?v0Sb}mvs@y0a)X$5bs!j`KN+xI|L<;^=@(R0m?CV2e5s4cv5pJ(d=Fnlfk%66BC
zYnheEax`jP(nYJA-W<|WwH>6Q;t_MRN)m}Ds)*S4^3#11578m1t`aw~t~{@<W!{;M
zacRC78jB(KC`mkcroXWuJ^N@nZg*8%R4<b^p%`b&ONhDwRo><D^Ufy)6bkxSk2zNP
z5PV-|pvxAe7Mf4E?K(S2Ef$!Mi!w?l_Sx!Sdcs^MsaBBS!6$BDDC4dut!HJ}V1vV%
z9??7`Gk*ptC@pB?uS5K=U(U8soduIULm#zdj4L@ie!Ad9CJ)Ftjw$s#k)SzO9S$=3
zZ4IUet>L%+nVfuD_Cn`Ol;BSIpw7%+jpz>6#A>+DX}xU>r>*C@9*;l3db60gmU<v6
zDT*kycNPA2;;7)w#*l(_Z*jPz9Zc-q)*hBdzDBb>6bhuHljvyEOuDF~#Ejc$nf<Y9
zd!w1H!*E#HqmT+wNLu=UJeEHoy)`0Yy_GCSb&G<`y?ZJ<-hPoO*Q?_q@DPeJ3!9uT
z-|2jE@3wVGmOG1!jgqfWL4MmQ99^OOqJXKA?-q=2V3@@uRSQ#Xod8OSP_`&sq%k82
zR8%w3bDrRxZ`Xa+I*~-|OUrOq^}TkjT~!n!|9mxyQV)g@`(h6nf6>C<aC#>!e5*<|
zwM$D4`?B%<z=6tXDRpcl*$Vk~)G$3&56T}-e|=)Sc#=XknmtV+C()?KpGdTj7wU0M
z2CCQrfxT0+M4UX0Y6u6REOkmnw#?pahTU4+yFjwGGq#90$2<%}%8vD!q=<UlO`bSj
ztc{|Fs?}0ax%uNOS)*aNS`y>SaZe$=(uhp%eM<kb9wuhM;0(c(OajTD$fw$Gr`DOP
z`_}uBNY$g3Hpwn;+VfolBQ6p9;+l1_tGNnLtdH_jC;8)bEDYAir{G;hcmzenO|UX?
znEvk)<C|)KTqhz*Uc($yo#Yl^^t;#=6YLbfwlx;pDuIfRkU-5wIV2KoIb^J{o*ykk
zi2~faQ=!6|%mpgIp{3plfJ!?Q&G=+!)CjqhFnaxv<L)R46l$$R>@8*t=2%MzXU*y)
z$i2cGwjxQ<s2}QlI}~R)?sjb4fBDT+`e7GAqG<Uc_vE6yIj$Oc5>Iqd^oY5!HyYg=
z<hj?S*Vd@V+_bn#G(^;8sQ|l`?Ve&r%8r`WMsveT9DV(b#6n3-599AE&hDj(hztB~
zs{g2l=-ylArm52`z|_t9B9&b4L_NU?iIjz!7bf#l4<gff#^)!gb2!`yEQP$2m+rGk
z9A(2!G>#Z&cA=2rDS9?@kaR<WYRV>#$`#{>Rf=UFuN=Ov#gm4X&e<j|ntX|UyM$VH
zxjg1YC9M0q8jY`aJUM&Mcq%X>6vu`m1-EU);S8s)4G79zkhQlzBlp{4$`6EM8MDFU
z&<U_O?{!VzBQzj#dJgL2d(L;lj{?&YWKpZCpJa1XcWm4h!KgK?nyQ07zlCQNvfteA
z!BOy5GD5i)yIbApOUIn&Y^S<Q)rz>KN~|Wvj9x0mdQ2wj)%Hrhy!VPv+V}OcmY$YG
z4d)n|8ND1eg&l$vvUh(B6zMW;!_S4zBQ_f>bNq18ulqLJ5SMF9l+Gd5YSZ_o$vNfE
zKi$uHe*ZiU=eJNRiqcev8vJB5?sSOF*kuY?qKvB(FM+Dc`TCj0BQiCBpI?wr#ZGPD
zGsi{>q5{^Fa3qTxUj<?vS2}d9fReHXvcZqrE*=w^8jQhn@~dz?8#8(4@pRB@Bc|Fb
z9jY#jb!;E>y+NIx{MI~2<DksIQQUpCT?WM}wk|!SC@$INL{hi8Y0!$-9eZqyW-u?&
zQne}$HxvE*-EcJ|Qm?sLoR$94ZBi^<Ml8&Rud+=;NM>ui%|v(h=C3lx7E7WmI`J0z
zeA;|*r2+2Wg(OV^`G&!Z7<21@^P6KdYVkn~pcOm#6$KOG*D7V_zN1kSd$f++JP(V1
z9kEqG;;5_gE})bgh9SO<oXHoNPY=10lXVTZX~gs8$yk;r4#!CC7)->G7Qgx0i%7h&
zRa4+I&dVYSY>llPlv2fdK7pDdwJYkw!bCLQ#z_p)vi*#(+iZV4&sQ*;TPHp281J*9
z@&=kSg?wyNX~h+_jQqs9aGh<H$wxVcgD1~SCJ1GWISFLO-rPV`uB@S>pBL-N_+zFY
z-N^V{KcOHl$#|OeR_Ra}Fb;oc#rw>(e3{Rqw`xj#m|j0!LxGp(ipdT>=twW+`_Va}
z><GmX0T8PVt2E`jlwnr9^QzGAW!G2jl8Y_mLoYv(l_Gyr!0;WNbROjVQC-DQtFi}~
zZxN-`*I&?azHx!(e7o;?Jbqpy7G{ER#@NWQxNRmIq?W44p{+|*#7h~7c&(IE9C7%0
zrw$d}`2x*Yd_7k-EIag)aDam_-gz|(>aO)o;>f;L6!N$c0k1@VuRTnarRrR~n|?4C
zZF<X5lqUEg_jGdT=7iqQ#{?AH&21~AX5T+sXlfC}yYg3tP`>Z4<YZtswnNL<ZL2tw
zO@6-KQ8bF5Y-!dR+nKLO4|CJZkB#%VOmo^pbM{~H8?U*cc|mwcPVAi&y^y?K%gq)_
z)y|mevX=#ZP~9J&fAfjVKqiqZXwJG9O!DPY)AcaEOyoR~L&ql!u{rQIUR^^LcH^h|
zM(D)gPf67KWv_|r@V}E$Szn%h@Dg!^$E24vo;1es)dtFCn^-Ymk`Lj$m`aDpd-GLy
z6A#=2FdqqEekcKN<7JH9yj4!dv2WUU1}7gqYg-eIUdIjYlH)R|(jRii6j77xty9Vx
z$you+(qclHH3nr@b&SF5USHlTCNcP?-GG7A1YeLI))q281m`9P-|5!;Z1|0bII*~>
zQ@KS@y@LiTh<TIO$(ojCwvTq0<K(>~((6{)+_ow*lla0#xC*Uw!g)4I6WFM-3HX%r
z$KZNMlgY%B&k5e2Y~g*~Vt|b&51x5I#Nk{G$v7M?HImoiyGM+0lr{&&ZHDUn=zJOG
z4-EfVaNqRT>zrN#&W*tUUgr;AiNiZJv!z`ugyRuWx|)@UytSX^n%9H-kiY&&Q{OSb
zvIifulG{_^V@z2^Q`wySmQgLbxfV7t)8>XcSZqBuCg4}$R~z1Yxiwid<g!~;j@_!e
za_tKNEU)9-WYbw(X2127yE3IJjwapt+~KAW)gCze<#(F^3C6R+es!s8X#F1WreC&K
zCFwmo<tjUcot(>r&1k!2@10Zny&n)?DE)$UfljFy>hy~4P-`>AQnL+AYh?kxk|j_U
z!}M0Lt(H5VV19^xBws;v$pk-7IP&n05~@o-mhW-^*k)%4>PKJOp2}zw(H0ayH38U@
z`SaaP8``Ub|I`mrARBgE1z&XYS`!!qNxj32u{yf#*}O>BL}Vo3xFR_bGFc+2nSkfM
zvWaXIN+j-bzPHBi;@SyC+V@Hs-S~EYJ!#*Ottq$jMg?OLoJPRsQn=y@-PK*DT`epd
zR(|Oy=gqO=gh#=k)XF@_)6xE>Ep$D}^GRHcYja{5oo#C29hfJnO`@Llse~VAp*fwv
z%m`R(DBp=XIq%auZD<yUpI%YHI)QpF0MD_WcdrJ9?}SczV8^K_hhBb~C~DO$bA!k5
zKZU6y2ZWK{kJ0kPEw?<lHKxK&Yo$=SwA4zt?_4yhU*A_6wL8%-o#4*(UCQK=j++ZE
z&ttH<=nk1Ytkje?d;+N-8Ih-u*Y-vsvCjA0g45MqY~}Tb;ko^CK;f9YFs{72dq<hD
zP)^9xn5hH(2OfM0xL}CIUd7>bmI(o`8nwM_@z^_Yas6pUg?(a@y(cM^v2>?$*bLPp
z*%on~!sKab#&}V(^p=m}Mhje?vHTkP$*f}*h6AjnCF$$a(ZXqmit@$N4s(!QDV%X{
zOpVR&?Ks=JFAneV9CmeI?voH70^|XYMj$6rLCg(OdFZb3?@|ily>BjE9z!S_MH81<
zs2xJhn{j7D+xfBvHxuyL?>FwS+UareU<v$-mv(B;fCYk6J3JbHL6w=dgX&ze<s;@u
z^O2V7-SKdJrdm1679)vmCkDsEe6N(G9*iaSVKzeBh~NvQ_X+)4n+SNZm<%)vE#}N+
zm)Y95I9N>1;)Is0(gw&L^tRBSU0R|xlLY^7uXG5@GVmg$XDy+1E)7F_GiiIXJ;ktP
z>;$|P=6lTduQm^*ON4Ob;z(#z+RFV#)+Z}NNdVGwTi**<VC@~=ii*P^bEl#k{Pb>l
zhjf9PFa?w)musVvowAKrA!YL;Jl=(#utS^|jq+iRlR}|}k|02KuvJCXx?xS(W3-8u
z?Gfj%3gQ+&fcXs`Nw1MrhmnSN_NpOvYIK(%#<$*80*BLk2=)aH*<aDQk3h+0(N>+&
zh%B!bk1SH9kJiG<Vg*BNL%5WYnQ5hCSyA4Ixk=Cs9jqs;<K9ctMGa6W7GjI*ZfwzC
zg6COly&t7AG1|F)bqOY$y%=9svq|bI6Qv&LcVkKUk`iff&(jhcG@r&!Yzwh-iCR>k
z+<!nJ+xhOyM}<?!(i~S}@X36k6Qb%qH1-?W#DnPfXNPn5KrxsX+wy-&-|w@PCGhn9
zdfvARwYJCR?ZfIFNT-4ZE;jg(Snc$EH^ddzt&^5Kt*ZO<?)2l{O>vC=mOUhn!1#;C
z>yo-zm+buT(^H|n+wwm_@cQaVs0Zv6l!l@5h2qo^8I}C+`G=WUnWhL0*Q)BtY6Ck_
z!{_=NrzK16JMoZlQGBpL<g$eUHUYoF++a!&hOY0nq^TQYD>v=W-YRHTtHEz2cDM)P
z>R$k@vkWsbiMCP64MYKclMRGeeYG@~>xL3Vdh07j!Gc#Pe5z%CMc`41*#^Vr3T8}?
z%JDTAjJW@ivB<=Gys{;Cw|x79JNy$eAKE?>)szGEecK-o?{q*7@NjGVVu4Qa-OYW~
z=0wal_8Km@DszRml+FbOdEP9_pId2m&c()RW@}A8U3!@}I&(~Psej=#9PV7{4Lu#S
zE@d%z14A?QN}gDQbWxhq#CaI46geR=vqtgBwh%!xGFdegSNw@=hxv{cx^)Pe7XpsN
z{p1q>P#K)OVQM;3G@=vEQ#B02Q<I4TJkL0zb%O<LTAs08!GAWPmSt79=^<^4*ohkg
zTL%w|<!sv@PaF|sz;vQ9*82dJmxa-{WDRW;6q&I4<dCx1tqL<}d|koa*XDhw&JQ0&
z?Odp}{JcHm1cJ{$QpkrICK6{_$i};OC0*|s$be4n4{&}ip<5v;KieZCKiwUT651c-
zX*18GZK{53>$DmIMwicm^T%1Nn7Lo;;Js*@G{4)$KayC8-chiZZ|22q;z*?vQ?pKU
zQaJ9U(D(a&4W!_e`bEwkZmvF%1Ce153;b|dMBAcPS=4k?N9+#x(RcHQX>l8X;jQiC
zO9QvudzbZdFH67^%twLF>{_>gDG8D^U0-}6uj*v6@m{z|kkXfrj0nGWV_@|UbWRqd
ztcO>Y3fO#VLk4lA_n7rte}qKaOQYt^@B5WW%mwl+a_ej$7NM5KS=@_iDJyF->Fe_I
zDY6*_!o8X{JE{vPbnZe6>&kf#GZ*Mou4FFJr~8+Ce7-fBhV}(3)5|U03s5Bwota#j
z9-B_qqQ!eZk*_jMK{gueQHvhR3ibnF6q7y>rH_+;N;m{l=EVo?0RmboVRo5M77mu0
z*7)H{M~DacZ-3YVj56*vas*+y4J=sl1srJ;hI%kYjzl51>%9wSSNI(ol0mHwpu@44
zGIu?w!T}Vrhf=-Mk%#AJ#?roYosxnYn)_|m9TsAFJ!Y0WPhUEOb3tGX({s%X7_097
zXD(3}gHb%4T8@Rvx`!Ei?5@%WLo4m=3>t{1+;T(#jaJ#$&J&xarb{s(rllnXC5U~d
zl%Ha%0;0KM%lq35oW7A``Kqiq!A;PgbhcBqdb7n^S^2OiDlIwf$s1Dx^O{h4Or#lW
zo}~Pk<Wc^Ulez7M>4hAetT5N&qQqVZG*AYOx)5`p0iF*7cowUURx!yUx=Z|~v-c4<
z0ooQ6N^@wUMLq|xS2~e+)BOa-syT;$y@aq?KympJ=8z1@0>n<C8ur>-8*Wp>PF=q}
zx=Oca`Z7BOR?aTS9d043_K5iTJY;S-@7vULZ{Fv=^AYNava-XD>dJa6GG=Jhm&6Ml
z0q3LY#t*lWlQnOIZ<!9XZvgg8IYEZ1bNJq#M+AWIVw_UX3iE*iDW2o9Rs_%<M<`04
z>-G&e3H*?Ux(ixz(9o2@9`bEoZsU@Ed+0oScimA7)zKACJMld0GcLc2OLwPIIX-Bj
zCDULw`*wO=ernyV5J7SG1Mj-etZ?5<I838HiiI*F;v5(V;dWlY{h^d4SFq5GlT18r
zc3^#$Irxt=WARW;K#5SUvS?W0Sr0E)`-eRek_te-ziP|`nzI$HF;b73od-xdA|zi(
z6gYtWCg4nCAd#I-%nik>TF7#oTsU0Y-Gv7lh&Caj5VSLMFgLR|y>Vze4mWfNiYtzc
zl#?ywvElKknO0uHM4B$|ZxyPx*!6l_53R5XKkKCKVqsa?8Rw7=R<Mj%f>75@o&=9H
zUG7CLthx&m2W6?LeX!iv9Z8g}o25w?w%)g4^4|1feYt7fhSy9dMFoTo>$X&$B%vj2
z&dfagdVu*w2DvQvXueoM=l%i;`Q8^C&LS0BmI@hQe^qpJucUT7ABX>T_npG}2WddE
zGEmC<&45yJV0);mxQ*d9VMnj%*Q0Y%zFusU2QBjO^GdU2u|%DRQ^G!Cj_dn=cn!(p
zcP<h+TFnJg-SqU}kG5U#-Hak?1<QNR{@`sm+^M#QOGxEnE70?+Z#>=GvI{tnhN9M3
zmx*mA%|jVPJq+f?%F4}b?>5)e&?DR`GwB?rorE-13%yE5x=10Q)J%GInE!?ROj3Jo
zNRYo2laz!sqv@w!Sn6c1F$L55WziM(a=!fcB0WxplJ2i4%})<l`})X=Tu{MWHEfkT
z13NiGRxz<sObxyBQC)Mr2$`s#_Eqs3ajkPvYTvsxuK1tt@%g0^Q}v_7FiXbBt&YcM
z;|y$mu!q_hOoN4=o>D2i8`(~5E8X<k&g1BQPvGg87dclfZHk?W(q5FBEZL@wozfwQ
z^|Q8%em&rQ6c|ITGJ%MB8#Pk5>%roGZv2l#+dPLB1{4aQQ}?$6Ru!cU<t%20Rb|rm
z=I6I_48`!oGa$*Oj3wY2dySLO|A~CyDNp3|rWeKVHo(uy?BcT1rBD|kL7or4DEk=9
zbkyOat$D~m{u5VFxPT&#r~p_3!|jhaEDUmp2W9B47YTT27VPau=R<)BSh4d~BvVrR
zVR#5gF9d{lhG%(15ABsy#};RQ{S+r`sx#Jp1<@kt8-WSa^ANJ6G{`<ZoegbT@?D??
zGi2?l-9Z?Dg-wrm1$+0I!&AEnctI4+IpeQ<;4V%K=Jav-?YgEsadt*MMfvXV(`VAT
zU6YFH`pk`bu2Vd{TwEF=7;P+bH=qAXMyGGmd^_E}>5URg#fGCoz;Uf8cMr(=`i@Of
zdxm*94sTE<BR+<<yWFGm1R(qFD4-lCMeRdzKPrU1CHAv}E>1r8)t+oN3i7ayQ=OCM
z1oySUp{!;A^^fPrvj;az&=bqkiSY*5<2QJdMGzXj8b?~#J2Y^{%b;Y7+z^ED<s#_I
z89ayH#4nf@3fW$+!%OQr<2!g9JhF=wQ~gZWv`tV0izV-_BleA8u0iW!xZqSKk=m>k
z<PatO#z58Zaj=Ztz0TEVFqmC#Of2rw!dGXz(7o}sO-?V_TMiAT^<0(N>UB-WTsdTO
zoP!$N54&XBuBVyue;|;JD=e3opq&?>V7Rahy-f&xb=XXLhUw(F6(bw(W!whxIBfQg
zbr^a7iWh9<#DKiB4N4;7Hk>t*M-G+(!;Z@R1Ak<2ZY<f-`T`dp2A_v4g6AKa^sLOK
z{(slwqoK#w=94P52lCu3h(%h!Au~4G!^LI#nVnBqem<#)GDmUqyiBtxUBV<<@`cM+
z<k2SwIy%p;J2<K2H)d&}l5L%Som4>lGBPESC?SaCA2fBW?8H|cvm0FBd8Z-Az;AT?
zl>e<ds`F6*SHRtdbm9Xwvxo<`dDToF#$Yy4$i~!-c}E^?vBbtxlsqs!r*~4wlXpM1
zlFQ^K4@vaw{A!i6{Kz(V*+7ly^1rUq^$bSyV$=k?xInkvVukUYKq#PNTqpdm*1dIn
z?c@IIJ1a8kQ{!{x&CJRSB5^&9N|?KB>$Z`XFm(3v^|d$V3u0fMHp=_lUdR)B#=1V1
zBqMdp#bw@N*SR~FkM%X$R=c|~Zplur<)<}FJWslt<0w1p)7Wt~Js0XEJK4-6?qR?h
z@rk?nsUs9JC=J0i?q5EUP2%zUG;@dV5%BYx+35$j(Jh`f*NQpZWZ?2G2LHC1e?|a-
zDzR1=rMXjD4ZC_Z=aPu<6O1dZgNx$cD8M;8)<V89C771z`t1`L=$@2S`WD#LI$X!$
ziO8j;r`ujim-<1Ft6zDb^RV}QnT&*u5lC>!J87$2`*-hI^rLZ^(cZJ1-OFf0HpVh7
zgDkCvZ=WroPXo#qgtp&SON;k5b5GpiAw*@c&^nvsSOh>>+b|5jb0+tQKMDCisBr?m
zUK9G*<=;1fQ5#&=P?{)YK}299`-(w(K20~+E^$6f`%&2$&X{Mt>9FN~@uvS+_kh=a
zpwK1vxFQVB8k+Oq`i`-py0qb6SU+);Ydk(#2_X0EG=)b6mg#tHf;g#V9qH8}OcEiA
z2S5766TpJ$z~zQ!QdCaj$Utz<-uY37((+Jx(euhSd4$T^$Q5QC?_HIE*g3L(Yxwdx
zp$b&jMS;Z4gbf+l@3>irhg(}b52%aEq%rNs<)ehe{eRahl8Nl$a5Bewo>TGyPuu_8
z$v=GF2{-=`#_IYG(|2?#^>|L%Wum02a!jVm(EdN6?zw{{MmJ5z7A&Edbd-TkCYA$#
z97kmQDI%?vEQh%|hdhq&yE8a<#}bLLz5oM`X~hTtfLO)}o2b-wJ_-XS=x(dkaFsmi
zf0ZixTIcZVO;>s^u8)h;yCEu>)E#&IG?VSByDPtN-9t}&TX|13Waw!eK5fN$_Sa>;
z<)dM~<$*H#b8%$<37|j;OWnNiGS%}oECxTCwm?!VgU=zIg=S(JOULMmy4>m81Fxol
zq+sJv3;9$hejngCla-?;zs)`zXb@+^RHLKDwBB^$pVNNU19&!lmYz)w&CcRy8u4Vq
zN}|BFy(}0j$?0;Mb7PD+s;UV9I5gJ(QW*e#Kb9n2xv@-`pR+~$9!^)H@fS&&Uq4@C
z3mB(g55F*2#)jvNxz+S&eHhQoE|yF05#HQ;ee9y8`^jI_HQy(>2zQo_!sHN%`*tTG
z4T_4ao|ife4W!>Wi5cxtSG1(czB9#n?(m)pGEl_I<umEH4=gDCRT{gAp$&+3k&sXU
z7$RTX$<DZ#Ox`aSN-qn;gxpttxmeh}{_Lc4cHQubR#St|lxGezTUY={Gl+A9`}2yy
zU!XJ7KwJy0dnmFGR^;}<Ncy$N5ZmwXy451FyNGAggrh{;VO-&@LW_vBz7=*VlGA^T
z9E3%F{=ahNAGRbdbLD@9-Q$BzI94`w$l_$w4SGs*n<h5?<^{58+tm7Yhd;li#>yVh
zelXBQj{9Y5<w=Txkv63`Z;o<spnF_gih(P?-uWTHvb%6V-B=gcS3b0cQprW<EI%7?
zLr0^SgHlWkaC9sL8USf$MPQYtZe2bKVE}7Daa|Ob<|C2Ga(7~`?HMQFU7J9lH^i=Y
zp@9GJ3W<fbLs)?ofJHc!ufmVc1Rj@_+P{k+4ZH3>)$D9m&fBuWmTC%NJ)5(k%fLvW
zY6U+-e~)Ns9}y|~fxv-Ft@SQsa)dnu#UfgP37nTd?Rpv_L!TU(#vp212B>K*XAOT_
zP|J2ZAnj56@-(X%CtvTBUhmc7hPo&n(@IQXUVztFN)<1oJAJFopu_ce%$J-Tx5qF0
zG}f&<3OP|2->w$&?5;Tc3PhRLe9A-cI8a0;n;%tHz|lE3t=E@9i;o@Rc*Z|u>L9{!
zxWwWwJp4qyP?qmHe<d6E@Q19ZbfT>!JenqNCEPBU;L?9cX@xl4?&se!L-lv%02q!f
zClcdLuu}XW(+fewfsMh~7EU?)g;~_x<W^m*{g}jpGCyoHCf+$*8Lnk{-~Ut#@=<^T
zXyC}qfLGaHb%MZ)RF!3&@$Y(fvY^w7*Dv?L;CRsu0GnMx{rME{Y7bdrbAnh}QFlP=
z^`!<Zc^P6megi_s{3+>+!bsyyi|)&tuWvZY>|$o7Z*Qh~r8v5dD^3uk0bAuu*0j=4
z`uY&T-S!1zeAY_U1|y^4TQ$J70nQ(Wyap<(+QQmMhU=%3@s=J;B?_cG5ot|S!^VZ+
za3@I_G{n@hKkJ(5cy$MuT8v5Bc}%f@xY78-9{z5szKEwKr*eao%OJ0$I4CHy&1-j{
zSY4t%^mB55drRZtT{S)UHznimq_!2$!5Hy4IjK)I{wayRP9dC<fc3iUf|jTXYkQ)%
zn2=MHFL`GfOdTLBQK9K9E+QPKRI6k0Md|J|XS8@)h6IY=%?iRBFT~)B5(7d$J-IR}
zE<T&v>=!PiYih8bi`eTJ4z@2{3DkjcBQ|mI1PVi1+iJZetzn-8S3!wbLm|_v>*?!$
z8@4mWdTQiq8!?{M1s)o;(%zoo>lsT!pi?vm+1+RE4{b-I{F}6%?2r)qa!&9NHZm+a
z8t?RIQb@sZQ4qsdHZH<BugufGdR~b&2JAf)SVav65-1Z0<PL6gXh<h6+SdHp;p3bt
zl~EwkrA|x8lN4_J?Is%C)#QwpQBf6@>=m%c=|33o<>lxug9zlIbyi#=yjoI*!pDM|
zuHOgzBtO2Q{6xSd9W1D5m=KdrFj7kJmw1h4aN&Jcj3}y=%R8-VLH$Kb`yintD2r-t
z(-7`>Xo7|M376&7M_0s3`RBsHHt_gK?7oEbgEyphhY;&_s0RF?y5P}`A^aw@BYCZa
zDToHm4+;OGszDfR)gwKKe~0}z3;|jb+V=a=D@4w&lW!X)4ENSZSt*RkRZz=Loz2Po
zsUe6xbXg%SIS9=LG#Yetp>At&DR{bypp7tj?#}Z*K`g30n4vNHGb-qoB^`;1ooMoD
z%uIKA;{h51$c#@z>2i{sGi~Kw@V@pFib?~r&i1W%kq<7H^2EAzb{OAgDKE76sSS4(
zs+2#T+X%KaxIE3Ak3oO`RaHCiZ&kIU&lu)G=8nX~aVm8(bpm$zX@0V`$pc0QzW`jY
zQY>*KPF%lc!%f`}J|lv7{JFxGNX6~tdjQ0*Cg9)Q5=Z${+wm{S4lQx3&#3xhE%I6f
zUBb>8&8&vh7Vn#JaSq48=Hy`jTwF8uPA1*U6nBRw?R2_2t!bk{Gy{a4T|S7%NHHxW
zm}AaOI=e`U4_8J_ej*pA2NdT^Qg}lYoAR%dI0-!K_0}gpG1nUaXjZ~Tsm@0wNTK$<
z^fQCLypK=TRtfyHAQfaAElCCM*eC=%YxE7pmCgzb|AevEwP>l>N^p1OL5m&VQD6Zo
zLH$1dlfAHn6v#Y#xh0k0P~$nUt@!A@=#?GeyCI5djvP&J+Jdoi4|rA<=mah}(DVPI
z$p5CvObdhX;oaOFln%GzJloR{ERB|QP_r<qRHHjFfmc|vH^0?Wofen3R$8lf{CD<N
zr0u8<|8Y6q!sSp82N>=SWARM>I`A*T$Dmi?tcg|7Pf0%4%)UPqaqO6@f>3L&HGsQm
zTz~lz(82%Mtf!nr2rGuH=F_(;7mf&6a3V4HT~k7kO;(E&_#>^AF-dav?vV%&PNR^&
zcM<)Y+m)OyQ@D0^{CxqLzw@F|ak1<HLtrzK^Vx#Ih&}=S?9NY8FX#L=k(ofH%1bIN
z2p#R-g|W*JW<FF45~HJqZKW9uS4c>z$c-KbFv$%}7TDoRjKU~Y$46K~SkMfFK`mBY
z>d$>vilA&%eOcKL+{+)8kmeO)+W)X7J-`uyaU{@abee>n`pg=3&8Oq$h6606qb=l@
zGp0<HFTRsHuj@DNjX%33OS=<~*p7d6E5A#g*DR*?w5US?;az%dkHc%5SVCM&kP(NZ
zG|%H(ZCY+~QMo0)djYn~r4hJWfux;peloE&TYK_z6ajxV`T^iCrz_xd`oqM-+2O0%
zTt=n7{0Fz~72gzGpiP2k>wBe~_=DaNk4;}pttUI&i(5@0UwT-|b=l@7TCy<|7>mqU
z0;oxA^tF468!yEZJk)8Y3R@|~o9d66F<Q4io~1kE{1WjeptH|mL1tw&qfzZlJCT>B
zKs-&{MBQp%1y`I!D^H|;a8)S1IP}$p+wtx>8R@b?d+R!sKUOrCj7HF7@y`@?mmE{i
zM-9dPB=_J3#V~G=DC_@1^M2g7_!p^7!yG$M2OrGgMCMlmJN{>_4M3bgnt`My;&53K
z+FG`QVNuLyJ-7kWM#n&#rXij$zVc{zPc0C83-q|xDHfIqaUM>~r^B~=I+X$A{UzPk
zn~I5b943=@&6Rx30H~P}R?x0d_M2E^l`j-!+Okn<EqMuDC-8?v^HBy?VBbtBUWGub
zxKR4}Hg)W33O9j4iX){2(#A`FVDo5;tPf8#F&G(zU5X5g5AeeN&3!x=jIi5n9ByR)
z8?Wd%Fa-YRy*xoT4xfGes65Y|`oj~yu;9^#eb?b#!Hx5jySuB?hMqb?O+7V=$-hoX
z$jg^&!~x%lHem<^Eqy&dSU)SH?$F{)AVpW1g^^jGW|ny@3$X6n&SZ|r+bQ)_G6aJw
z@K7R}nyw*iRVX+In#F>VwLHk_+v0F=1|MH;052<}h!|KgSnnlU4;x~6o#WPG-{uy3
z#{DnKyZ|JrM6tL3G|gW8>W>~9Qv9WM{3^~fth4Hv9+74B2y}Td=QWkY;~cNuEB4GL
z9O!R+Red#I{Bgr`nRZ1RD_?$Tbx$Dlrs!`jR(%_1X!7n$AT)X=)0{WPn`tD8LcVIr
zi;pUkL8WcM%rL{;rY#=tL8Ac^>Pj!Mv!#dv!uVmxZvkUUf9R5umUIG^_aG)e9X$R*
z$LUZbZ2VhKL_%}oz?g*w2&)G_K0~FJ6LF3`%Lcz&du|WRmmOtr-M^CktqFK1-1-^h
zJ#8yHEOLA&6wHZ41XKuP!1Ba>W$l)XQ_<icPTo!U*XnTPp2XTJRzXShj$21L2ASdt
zZF%FX^u0uBmB$xrh9bG&*=IW3tYlKrKpvAMxh3Q&uiNl#z~be?-r5e{OW>78^V?u%
zHd$j8qR?;*C-qM#%a60NpA;JAO;BK`{HyaGvd{N`jY57S!!8V&3~m3HcJ4cF9dmgL
zJ!-FI0@TbA_EzLJ_kU_L*YW0;9Ym0p^UFM!FrOqw2ha8Gt-nLr+fp<N<{5Na%(av|
z^?TAP(hIkm;Wy}~A~p_yro1^Mv;V(z>uNYB(YH2C`<RNE>JWm5gDK(tc(vOTri16U
z^-r~ZoalR4zi0FnwK4G9Yw4hJh6V7za(z??7_TpJR}Ft~Apg6?Fl&yWlbvBDjx=A*
zl~4hXh?vmIZicPNa31|FF1^q%d<O*5{@X5|;7zgoN1MH&U8&1Zi8i@o=ezFxeB7qE
z;c%@$PSD7-Z!j5OCko*4I8yH$nO5@3iEE$J7!VY$%-uelcw*lNbqod;!wz^w&tpMz
z97H?JJ@T<KBDoJZefoBg00PZbqkAufolkfImgutN1?C?fO+*wzy&pnV&?-wJf-mcs
zHu!#fUDSIam6uU5wl9OPtQq(+&MZ=2t_LF2|0%AFC*s^?XKCRy432(V1eL~wR-IX6
z?Kxm!IBU6xx8rAVVQ^dt#I-xk4c6Z<QyShU<uVCk>i!ips&Nv-*2eQ^0PYMs!jm5H
z^0<e>_)cR?77ZyL#?S1)uI2RU0x&bug1Q^EoE)Lf0{`7g0yQz&%X3>8m|SpS7RDoY
zUL=VR`pYQ%C>${A^aOg~fEVDACg2e=Yi%f<SZv5?%&c2zVG0kUhwM{e|0_|JC8h2K
z5)L<yPuM!u=hGddT{V%<AeL{5?M@PK2$)8*v%@R7?r5*6!GO=J9U&swmz|`A52yTo
zJFBo9X1a<vC=goNM747*Ec#wolhvhMRwa9K3wS@XfCWc}+rZq&KC>H5A;+)GHFa7Z
zY%|gfX?sr=xn+or<Udcrks{aKz`g>z#Vunh)mo>-9?bJ!2i)^KWTnM7?0^RK#ngrD
zky$ks8iG6%cTlP6Bb@aI9EC^d0nMu}%#K|OyVml|(dp{L>j8-?)h1fWML_*<MEo(K
z6DVWX5_e?@hRgHC=Tw6d7=i|cG{LguraM}q7H0iSVMUxrF?#o-bs@{fHa)@RquAsS
zFow3-Wp+cq*F5eJ5dEzAT#->Lh&MlxT}wOh=V;Eo-<e%tm`mjFBz61m)zXzfUNbB?
zLUAFX;Jdi3pzoG@bx9$keuZC*$;!`};5+F>4>DB^sKcbl*_)eq_<1?`Mjec>p2Po<
zVg4GV%tY9OS3c$z_s0LSO}|9Af|j&_MCT@1BQwpQD>;QD?37BF80am&m(O5slGv7?
z3^w8?a<3)t6(1uu3KRqQm49M3qdD8DTNOQY?u|T4RNPhl5}|V3?ROvy)}KwTf7_V5
z9s7U3F$d!&ze+6ldVWadB__qEhe^&ob<Hngn~R?%T?dSbIFdjUh^{^gKPSY<3|S_`
z_By-tE?$1VHB(+3&3G0vmu|k+I8p`0O=G_R7nSQrg6Xt;Xtw3j3qt~o@pwDc8Lj5q
z(5T#LA7-c6cq<2X3~vW;TH%L_5+UkiS6eFe_#9cbF^c{_%$<2Om3#m95rs-pNJz$Q
z7AkFH6v{lylnO~1Y7?2uQ8EwPRJO55+dP)pDaEEJr4p%=jbu8=JUyRlZ<S7+bD#V7
zd!DtPKh8Sqbf0zB?Q&h;?`L?wU+)wKmIMExX0FfZBgRMY)83Yp1cS+U4ePr}hNmq1
zG#<~GkX(C5U{##4cTn++c>4Py4@APnBlr38vv$bPW`nOOG6OdpEu(9W^5-_)*+A2W
zG+4GVqLiY~B%tGjg=a^Oh?46os|+)<x#qnV6`qu0B9pxu@ZCodR+SQ}hODv1{EQZ$
z-mJeu#Q-Ie#n!pQlqkIe`&x!4ct0f`GE}HIuH0#IwkKeSrr@b~z88jo!Pw-PPQmCj
zdHeyg5|0dB?A-PX^}Cj)4hvx<E)s*WbUfikGP1_;pk5nzO=)Zu&YU?-n$A^#j_%Cx
zIus0C+!FDiO4~Lyz0rF;D4L6G7W%**$-=I5Wp?aT#czRdUWB7%*7#YWrn(}>Bd7~@
z{iGXFW}p%g*$SONUI<i2L!{|7bG$^6xUjXHO%a}?x)NANUUzK|6%k}V`6h(moek@q
zL>Sgm=ujZycNz{k(CL===HaTxN)?QZZ@#9s^F%CV327XNo*RzreuHL$PBkoS>f(vE
zI%N(Tax|h~y9-01T8Lx!PwJp6q_b_&XmlGj;x=8TYo8zx!ZM_cKb!#Z`cwgWCL#$j
ztnweZji`sXK^9un8`!370=lB;ltTtu$gds|Fq1y`j8o!$z-w6@TyrhUt4%C;!P@NX
zdV#M-P4<R9V*PDO7-Q4kkWD(=myp$Q)$SgC!f;@qvVM`nShdeI`*7bm>DB=9U@~iU
z<+nYUITrkvFlI1fj&zzL@`BC*bvZvi_sOlTvDz2ya5rn!*@aPK;A3{L7TYWT0y*Z+
z`jN->J1NIQinPS+zn94JQzGXz7$)POQifL%6j=Y0Dj6DTL=vNfv<PPtZGzOE907%P
zKWGsGxm7VRr$cRG+oSS&H#`p+d%>SX?a0ELt^$O-Fh_XCRFvlj`)!kB`Jv4wNBlxG
zzz~v<O+8?$(|XRny6@ozmvLTg<YPFs+Aw77!4wDNr&QBeb-bZ(UZ7xrr$cRbF$_)u
zw#$Zsyh5fhi9t*fy`V9-pMZsi+%p^JyVqFX@;-y%6GJTa${FQOkK5wPXebH3n^v`g
z|9jJFa+hop80@55T=U3UGueFPZILX74Hja{e8Bf?ngJ1;jlBsZ{B$46Z6NrIx+nAz
zXMGIN&G3~}54=zX)6ROB#7J>(J4LxPjzLa;TtD=VXH4|`%+IyqYRjcc;!>>iJt><>
zcY@G%@`WgLE&emXgrjF&;h2$*B310&v@kI8YCv?o!T$DQGnX6`SGt)3OrxBqh&R(X
zz#pTD8!7M!)9)ibeLg6=2NdV}LV}TvmPFe$2oXVC4~R#`%zR+u#->cA2r=R(YERXW
zWgnO&o?3Sciwx$@K%L<6#3~xg8xt=KgSTyp8_a%iroMp5%dZvg&)oWK)A+!&M*c`7
zyX5vF?(Q`oi8PERA3NmdI@;O5MuRyH%$N;(MI>ONaCMvi)vWC5OHdrDPTM0VB_q?<
zujO~v_X{Bgafyyxrp&Q%EMZ$Had2)lg6d9JkFGiYwq-%eFpQs!_enxLQxE<$A2;H5
z$ET2vja3^+8g5vD<aUSBuw7@8{;Svhj;P?4twrb`jtxGV$i0LEh}3Sa)13i@no+;n
z|4p2#JYN$x<C=^V$)MRi`3tlAUnQzyDx>t8JIr>($y*kHW!VTOaqnQgclN1?g<r%)
zvpMFQ853fcEkele03G=qU`tNDG$vxQy&Sy!+Fgd`qKXHN73Q|r7eQ5g)6T>|k@9>-
zX%kW_38Zp9jJT`SXWM)2G1!BQuYg?Dp*jT7ydyHddube^a;P>nftx^}Asau&=*eF9
zQJCI7*;|`^VjGcJsPflSHyToj48Qx;61lEzJ<011`v4-x1RNS(|KiZdgas1m+N3o7
zW^xZ?+`=OE)Wo#*8cUc=aI-VgFL*!&T@0mS?S;qpujb>Qyd!D;)mD8$Y7e1IRx$2)
z1CuSotN^LY>v&OI$}NOSgJP+e2`@QAz4DM^=GZMb9MX--Ws<684Hccy=3cq4j?-2H
za5<`hK*UmN!v4_zrn_D%OI<xGV3PfVEcNeCiv8eVn7n>rYj15Ea;2^w7SFI8q+J)|
zM?$+p(qTKGVo*&OzSt$FWT9UHW@dh2{kdksNJw)o9x;SO>8~E6`m@0^wHfeDlvj?|
zf`Uqf>orZ^HbXRGj!+WGHaw#|mmOkFfU$nJNX`6T*Il_O;Dz8pt?MIc%xFA=)t$eT
zeySi4^tW<SR(xqJy2_D!;=rJy>MqTjn(1&k44=moTenpSz})S6&CK5z$o_sQzN$ri
z{cp9XjO`7~@h<7A9~I#t9PNbDBX{vZL+xRHKqN?!h3#Rjt#7vNhDl6Qa;CTtToW58
z7%1&TxVc2ccELZ{XEzYHQJ~ax6eUNR8aNYus#<qw90zYqjG<2ctG)zSzc>)0v~?nG
zgde^ge244}0i%*2=!_PmW1&kT9~EYNZP^{L*uYva!T1deH{vFNim4z*+pYk9TkH<X
zKyd1|gR1D>0Gt*qiRx}EBstB=2+jw6J;mRQG1DpS3&~JY2G53%6=#rY=y6X$l%Wp8
zt;>0;w4Grp7thr%JFa{lD1R|5buY5YuW`^cvLy1LQ^I*ziYt>5DK@K<`rix-GDy=k
zA}@Rn?C8!A0-O%|dZ<FGl98_Mu^tqTN1--Aj)m{i_1F$<^6UrdSgfYq55qPd1o41)
z(iIF9xrO-Fh}J7!kHdTb1R_6G(srHCiNsM*kp!U<F7ZE7)cnzle#ih+{bJfW(@D^x
z&k8mZc6m<TzBs{($yBz~OL@U8ta60PqOAPg3wOPxM|$4x&o$0<Dptl6+YRnTltWSa
zdM!+2=SjUo`3~jIIBa-YE`xsX<3<fropP#T7_hwrCV2YQ(+ihwmkx`H3j*|HCUX06
z*hyjR5e+f7|KmgGPJ<7zXTwl1)ZTSa>N9OrCuwq_`IK9w_{jgFHFkl#EFaLX|C`2G
zW&J72506VlC<-X5sn64WV7oj})`*xPpXQg0T%DY4CsyqrnTDyV^9HZuBxRh)>cnVI
zSX%Whynkm+<V{eT+luhLrm?%{Rh)O<hxUWs4f^Ku)OMM6XL^I0LHfl#H1w-Q_FI#$
zi)>vOgMJW)bd6svvW-c?(S(VY@BW?6+!Zm0@}PKm)KR<VeT6GG&wDfQ00Ad<<(TMz
z%tf;f9}@GW`DcYWBTpk$y_oQgS>{q(A38CPF7>Vy%1Z$n$K3f(n>sr)$EDlZjjAE_
zfYgrFPMKC|yvMZs8Dne*C7!;?T<ulB1Y4ni@I7N)Me6fuCK=MxC#mhqrtR}S5{doz
z`p<;uHNtYWgVm=*4O`Ir)~>M4|EW{fv)ZGt3=uda#fMVcIX-^e78I&r1_#TMZ6Lqr
zkVcf{-mNe+x#D?vCNip3PBwxcw%B2Os5`d14F8)gc3_xpO~wQ~3P;smO;|cjodrel
zBT&`VVjicF5Cc)alVwm*xljXHWZ}ofFDo20E1K1F7<^-*3;JQOF5n!K;QE7mz`Gm-
zqZ{huojIrcKsr|srxyH;A_uh)aRlutPg2UhC|%XdaZtrI?O4}BFsQgAhFS<ejIP)v
z<BLY`&)RhiSPh_&yRui{v^>e2;>xKFN|C$XJ?WhpYJ}OIm^f^gDIkx$`F}z|9gF6?
zc<_t^m|ivsq8iopb*7t=P<hxoa8G|EPs~)-+Xmwi=T1noM6$Yt)f^M>hfDhgQ2o#8
zpF)jnl=b5e?zq_CZLP1d@nM!%F=&S8W{*&+>tzKLvtZkj8x3hL(u+tAH6A)sG)Iwj
zw%9>VL&<q4YkQ?-&t{gvzp$V>^nTt`x0M1Ju>1HN{GWZaUJZp<KJ|q#7_m&#g6-)B
zUM~AdtO;b`_gL{f=)5$8?luF;Mc<{a(0&h?HsruP4o%bb@om>v#N*|T2&(GzB)aw<
zL?o(+`y|oYomXE^=Vsvf*FJ4`k#nAHJ^XrAjbT;@%Ov5~BB-hnI=JB3go`?1q#n{7
zs+_`!3`<8v$?rc;2wd$ssC#r^iaro%PH2QRb%ldH{pW7c5}2HnzVx|#QQ1=+rbdqI
z&ax9Q6Z1?G%4h?);f{h5izI#{5fqUAd7sGtC;c;bZ{ePi^~T1qYK8HM1|}(caYtOX
z<LlSR_zdI>*N+#H=Lne9>y8IRu^@eeHqkDJlS|N$j*VXZ3@8+0Ewtwv_F7#qYaD$@
z+q@6m8DYx2&mHypWR!rhZVDh%5O>MyAz-$q7A8O~{Llv)7&m24sO6<!yaAQd?$jgC
zQ(ZaGXt6uEWBWn=rR10LT2LGB;V*hjXBEft^9i6!a#j2fgrk7VQEB>eUzL9=j11vx
z&;wU&F(&CY2)``u!XKDG9r#78bVtqYT#^yk;j0_=^lwrCdkXSMi`+LI2VMg(F@9=p
zf?Y}+HV@~s{!Yk*hU0$douGk?qs`0Z$U{9(?6^j7o_(sRQ{MLK&iluG?E%GM3{O4-
z&IzLWSm#7;lWJnJf$c|k3CzS5oaaGTprxRvv7gR2{wsrJtov0bqGK^jIY)L`yu2U=
zR=!9!6&;4=UC*m{1^#^0A|k%tuTelt?QjGOyBwbh@!@vHo7OtcPdm{^z-3S+ifSCd
zHtQn}ErnEZ&}2)|!s3^X`?nJ-UfG7F_QrfoVm*e|`Dg21+c+J+mv^|&d#w5?p42jJ
zy-K#g#yqk19TS<(cntD^F=fq18}trPm~>zxpZ3Bq0d`DIz!A`bsSCL0Da_sB$jnQr
zzNR}x6u(ejHz7$68382u*N^q^RT7k@UDo>84{WDfI1X0!9m%DeC#uz;O02PI=<zO2
zH3VzK|NS*Iqrv`-uv_q>vD!S>4mKoAO-Qk?8#quv(V=X5rdHC4@<Tj%(sX<Ek~rh{
z5vHrgOClqj2S-#w8@x3%n>Xl85x3njgN39v6p2rQo-rpn@t@`u6q&KUwC*|yFk!6w
zH>+rHc3OWkpcsC0cH(gMRG=T1SfP#yP`_!*flSzIHaEw?(5+%7VnjhR8dk91UE*Tn
z3jb*ny%ypW80Apda9V$kM)>3$T9~^+Tu|vse4>-L^C`rYR-nCXK{n@IXOy$iy^N@u
zUa00R+r~mOsG-y2fk$;U6HpYb+g}O+rD)mba}x+efgM4xYp@5UlT8bozmpW_q~j8N
z48MXlZPP}39!LCE4q*A6*WLq~jBf3(v?dbZi{s$&7}rR1-rme?(pt?xeWyy4)4xDA
zflhYsDYY1^q11l`Y~l}mvcpx$IXu(BJ-KVxSMHwEYrk?|W6gD!`)WLTj+39p5n-~W
zB<jTSXK!msR-u>szg!Kox&!v3LIFW`Iyhkn&eFzl7f#J{-FUOEECyTDOWRzTeZf}+
z>{db0@u<@mq~NS|7@nGi;5IcN-=U`3vglQ^BJ@-C+kJ1Fq}TvV{_qx%>@@qba2zs!
zf}!GHOF}>2M_!44Ckg!*Dsq5D7B+aP+F2~k(!&foHEDVi91OaJ#{uPL?C0>7`BjcE
zS)_4<iF}iO43EY>t;6TjHPbIVV~*OAZf9emFxO4nJdg!iKzj|%C*_<}Z%peI;k}7?
zE-naioFD4yA5VyX(J_nm*)H@RfB{?B8*klTo3XA(M9+cm0L)msxlxT_LrRVf%W@d0
zQ@!dutA+LEuZ1;?IspO~y;!?zC9%#&8NE{B_QSZ1dom3)Z|=YfKx9o4Fq&;E6c<ub
z;}=0>qIF=-a@v*TO?l!d8Ll9?x{;Ak)7kGof3#C+TUp+h)WElZRm?V|V9loPbA=Og
zdnISma*&!13?+w}2suj9eY_|cQyP;WP7GTt*5)+tR)>$UmJA^=SGMOM)5o>ACZJ2I
zPK22&cv7G%_^;glAPK?D4^PBc_|qzviy`9jtp@@7g{l!*Q(J$%S-CAo#WX=8nt{Wa
z={S!G)w~zd)``FVepXyW1Hr9urUoGdxII9C3yktS1q0VyJx7|}1enGT94p9Jy#ERW
zJ+PYwU+>NF_^hHz(u$&bb&R5ed@!8>@z{ImnM9{dnva}<!^)wLO!O4jpU+LxIf3V2
zH%kkaG|CT_gLdW36X{e8X<fV0-1l6=g*`r^F!xaHY_9&VHqv>k^EMJ_1vb)M(tX=v
zRG7j5@)ph~hDhDKIme)@0EWU}Kuw00@kq)2_xor*nDzeSdMdD6lBMi47%Y%&>@%KX
zi=%Urm(`)*RJsZhH-9KW5$UiyB9RSU)Qv_?QNz`Fo4I5MHqH=fB=Xxa6XKVxK7DRn
zc-9dGz(mI`Hh|)(7{|pr(e&By3MSDvizru2Tzg07RCq))9?O2N1`h*`(J<Eh6o4~X
z659^1WxTCBqQ*%nm4y7VZ118EA7MG#wY%+{Q3%dbbFf1MFp@n<n@I!-1hJC1ijho%
zIOeq&$v~@=u}izpuZRiv$4j7wzVjpg-odFt>rm)+GcO-lGfF}E@!bN5Rm;TCIVT(B
zU0xQ&<WjS8^%7mb4+N0)5mfHleMI=#!G5NA*icXLDoae8k@3+vj;xU4_LzzE2vd_w
z$D`rn49qQ%(9v-g2f`}koD-JL9n~RY-0ptGxcLFLjglYnX2~A+8N8<&A|F4^adomI
z<HhalH}|hMoms(?3|g9=pZ^`y?H_m4lS!4iXKM|-c?UUu#<{7|a%;jXfh_B)I%2ud
zk>xihq+f`qtM;)9L^8(>IZ=}5(Ss-Khv|^F;`ZTKszRC*7K5K3SLuled2XqsnxMRV
zh~^@-7$*#2s6p8Gn=|Vk=2&B6)j3AJ{jaCbzAO`Rd*b}iWi8Mx{PLwo$FkyG&;IS8
z?Z}_*e(}|WRX*g;GIB*6tZ7_9c&_!cvs+0;cI*?NHoABIjdyF$KP;{pRZi(W6OY39
zy_bq~dYW&sWZYribtN9>0lMjP+iWUm&U=3!zd5ZcSa$kP)qwHBIXgci_mwb;D<cNG
zCH<<6h@b_){+jIjXIyzjQ`4AYw`jGBo!AW|VOU{{|67LJPjHo7J+pv=JGr(w2GKv|
zK2^04A$=3UDb7?MDp3uO8L#0NLEZb&*u4hD^q)4_fs7K4(Yt4+n*@$_F57hg$C1ma
zOO1b@#5pI*Q5HnW)lNlT=I+(2XL<SAS_=v+u);?NfN0wSQAeYk4Z-HT>kG|hZ_0<7
zdMUvCti6tm;c^g}ibzbaRj;Tv37PtJ6c;l!T3sB5ETod8H}SN<jy>$FSfxr3wc(MM
z+IH?$m*&oedKhG~%W4hOFc&?&)#G=v?S$yBe43!7{Bdi&N@1q;csheat_H^T#{I8U
ziB1F0_lxj$1*nc%o4ivuv@Azop0fI5qDRg}#KaJxJ{(yUsF;mo$Cf7I+cCx7k`I8V
z15C0;r#M8UaEvY8Oy;8P#PIKyU-Wh{oKJ8c`i@&7DO5;kl)L*kf7NoX6@-5+o)J8o
zn*I&p^M$a*{VoOPSg3!7Tf4DF6k0>1qME0FH;G!<3Nn0e9cZ@3;zayrNkE4SaHO&&
zYN<MnMJ=u+vhf5)F`8Wc?OAS$<5fEgA<u%$3i!<9Vp&^z<L7Qv9Yuz%xvNb!MQPfm
zvgtdxX-bX^PzZ!HfvN2DO_(qB5f2^YOyrohwj2}oU3@-7F_4=A$MR}(ee;xM2<#cx
zRLvdlU}14^FaSxl`X4zm;m7|$llDKCaRzP*2w({tYRN5h9NY3rL})A{Ceccn?b#`e
z1ursfH9I7xSv{$PjGTX;CCvDlDd#PrQQliW9vq}^0)1XwG$8uW0mTd6AlK_ioH1IW
zx(1pIvsKS!z>nDECjtNc2Kql^lQk`kQf#IO1jT`rmRHaEW6CXCHY){16>nx=C=awP
zdY|~)u!rp5Y@)v{msHl<QGQ^l-+cKTX%hYCgVKR4g>`>6ls2}u=~Bq=8ydM4?rVqX
zp}3ET7s+l3O0s&TqGShjQZ0k?spn<&^n#!?G^7~Fwk5hAFot9Q<lGL~R{G{D_+vqI
z#a!35`^X>`(QY8$e@F$a!#x8q$jBAZsTG&Dva$d*N!0ASW~GvDCvVW>nW*5q8(@-3
zUN0%#iN(HSm{a}z8*Z)1Cw|^^K*^otrFs|Jv7E_Av(Qp};duA#;XEaZmd^v9=O50`
z4P=eHi+^~uNTWUOoY}(3&QF=R0rS)|7LMNW@n$Y5v8ZppObw>L*MKvs+%^M2qlr=N
zgUVc}67D=RnBWDo>hyR;us344xk>K4DM)MErsQ7K`vtqDfuvW+h#t8hKZP`*vXX^(
zg{qutZWhb#zpz2zbw0q!KZyOOZFT4`w$-qLI@p0VREuVb<zKU|2Gk`V(RS(FXpN0`
z8>n^?$>&Ba)>rOG-93=Q3@HeqG;0U88)@_Pd>9eXKJjU6Qk%-P2|do_u5^0(^yVv_
zgxB8kZ)(-UmX0Trbe^V*h|yWUmuu1xpqd<NfQ4aHxmF+2JcVy=(=$eZZ3g)*&AM83
zngE)Pb8t4c8#RGjB1&!(>6Rce=;x$V3kb?#CT^w;UvFaVA2)*GB#x@tL|-vrYiwM8
zMz;`ZT~eQ!1e)zgZVJba9Isa{$AqmE(8t_$Qo^<N_7Oc101W86n92IC@BKH(l)!4~
z&*!jqMt2hCwfD)`6Q~YI*c09HnNIMrdbdshMHY9+58j+DuEWo+vuW!bXT{1SOF$y`
z2=_zU{JZF#3&l;gnlfF*F{w890>1d!`+@RflO3Tr7JF-w#6DfMc;|q{n^7g?F<?Mc
z`!vtYy#eH6A923T<P7Lu3ZC?CzHC9E*^G>)n%<Go{TOvPQT)v)->C2lj#0fa)LQ^k
z0nk79f<#;hu!YSg8j4VCJINvQYzGU!f9S)90WAeQ_KqL=a6X_r`|L7%9Pj2v)l9fc
zxE7DS=ves&BQF#AV?yxkR7cj3+1A$8KjYa^z{^h->BnN1n8L*z%10YxsQacR;|r_e
z%eRu$pP+VGd8@6h(<6V%vf)0^f?Eg&dnN~YV7LD1XoL`I;3>>0u^(9CEElQbvKOh%
zP0Ak>ckN2*1A(+Uv>PU@ab_g2W5sX7VG5MU)wAgucz36W3ZMl9=s3<N;38<?bm80`
zaS@Ri6H0ak4zzQP%p3kzuQeFdzjt$gw`$ht!N$uZ197*$itVYXK81|x-Yc!WSiefa
z+`^uarC=KnSaP=jNUw&D#bQgI{JuOWp3NVxcT8O50y;8nRI)A{A)vN^bz!d6hOe%m
zQK?_T`X@k&3at?3Lc%eRw3SnzA2_(zco$q~96XmKW?O1A?|gEp(3eh(q>3am{7ybT
zI6t7?uk?u-v2JwNo+A350ne^ZxpB-p_wrgTPC3a9pnchLLPStx7(T;&F+cjcR|R4N
zFW2(e6#t3G7RutP5W`Pgv*6Nf5|E^#?IV7<-Ff&`)H0}{zd@5o3Nb|NalTWg&07Z@
zfO+%jk8MXfdu%jr9s<y!LV~+^%h?gBFVb7(&zgQbdA!Q<{VVu-JR|S+lCZq1Cy2%a
z`UcR=r|0ah3(Svj>~j5tCzQjGyS#ludUCegvwE-RH<EOP1sNnQ<?1)FWPN<8#h#{Y
zqu6$?G<=Fr2;N`<y}yNkI+lQ)NnbJkqWJP{UC8KpKbWTPJ~u_&rG6mkyC%%$Z`Rx|
z5@U4YP5_S0g>;mlU`2g=@aHA>058fTa<4HkbOewIrWYSbqx5gOy^nYhU#v72!J%~J
z&4|ME1zj8K_$bjSHb`uID`+jN`A)Zy$z6td+KF{pc0h^c)=!iaI7S-tVykn@yqN9o
zVJ42}b!Hr$Qu6uOUqx7UM3$cho5<6=Cp!$4J-eC+(;w@(Ig~q_ng(>_q#pqIbKAjW
zpk4~NTUYoO-(`xu_lM<qL70u?0%~0gbdksJBk%o%Xr)7GYZz=>cC%fiZgSk!3>xz9
zh|^!}yMAD+!+&VDvterVdT?P2e~-uW4y-EZVh`JB-pn-8jd)tNcg0G4%5^Ygqjxfb
zuSewJlKy{zCJO{cnVsN21<J4W#x%CD;^Du{nvLi03IW`Kv56nmhpm?zby5d(7N<Zh
zsBeT+;iI&qsa@2Kb{@el`sPN82v7;;pi<Y}*XwAzqo#9wTWs8d#H|huEq9teoeQrt
z^~8VKh2#Fd3l|U*)bvY1+!6QPk{lTjcoMP$e<7mw<hQ@&s*||clU#DKe5iGP0hn%_
zZ{?Od?eC1yo`%Eiv@N63o4k`J>)g?GcU$oNB4si$t@=OvxnaG{&yw|y;=Vgrx$7>+
zN2V%=Hs#-5+)*FYKVD=uy{R4uV>DvI)?22-koQ4qC;4HpO<>1XCgKB%{dF9+K<gdI
z60VDILMT`%TI2loYJ=Yn{55WilmHdqFyAKIsWd{pyXXgc<5zl;PWU$zM6A1);a^Q8
z0rw_Lh}x_klfm1U(2dam6>ko<16Ig#?|^4``)#io^=87Mra#~~m3#<t(eK(9+QI5u
zy~hyJ+I5T+g<$4Rf8I)uTsa=p5IH7jlf<&QS%top=V7!Nk6*7`f#5F<-hok7^-ShQ
zkmB$j;!kQ1mw<o}tJF}LZCL;TF>jy3dm?tt%K$0i*>;gn(J<0&D<mkwPzC$&bdXEl
zVv0S7k!q#92rFX~L^|dxIGm0LYXZsxNT<Mr&<9}QcPHj5W^LcEF>609!ioSiN9Ew7
zRmds4H|YAvs9Y<FkNizr9-j++@V7G~&KNRjERORw>YgYan{1GOXsWyG=whC)t~iC;
z`xnhCZy|dK#7Ao=$uO%av859e?n35xXO1F=Wd_E|z|Z#WHmJNdp3$G<ce~opPJ%f;
zdMAUtCWpw<5$WHJbe$2=vCk$kuGDo04ClvHWG<deI0W1^y;scej#Qt+6pU`wfg{5j
zHviHo9a#ITi)Qtg2ADIG+qWY&W(3*@2L{2*<G2|zyv2C;t=)ck=<6_qOoa&oy!41U
z<h1OazJe&DNKr|rlGJOK9;EyZZ^7`JBn<UN*d<7}o#Y$8H+A>sir7P?s~+Erh+_8O
z08BeyU{$}j4#360RmZ#0C=T{K-XISF)*Rw9YmR&(g#cnCCDL-FtkaoR%ZM9nUpGL5
zi!T?f-@yj-!MFzLH;tnQ3(bCQGks5fLH2aBjWF^TNXSd8rg1xW4_Ipm%@iAt?Mk>t
zUrM4$AwVU(HUPB9K&7>?wQp2p@NVlH71DD!Iu?Pv5YkGz<(x;^M2!01Y`8`iihd1Y
z@Gt{12&_>!w&`c!9QF0`2Es15P7f;LG}OUzS!<KNj6F`TS=MlBa795jKTUp-L{z^I
zK)X#^-zW#yTL@Q<t`j_QZmUWYu2bsia%PA2O8&E5trE$#&S71jhc0{@OWj~4Ul&@q
zRIimB3s7V`EohkPDMH<udKa%TaWxv-_NDEB^pNgJh`d>66qoo~SP*)6`7N;6;yz$)
z;AWa%e~cs(01u}8RiJkMATU>LBGf--Rcf^0F(Hrw2L5suGzO4t7H}3w{(K;O6~paB
z!0Aq&<!L#BIT4CQ+jfjB$V;jtSEu8Xv1dKA^{ifWHBRv@PrVl|fTH<tQA^}HylH*>
zF*pfmtl%pEDM?#ej)2nsk?6@u8zbg~)O0+4v7K1C{`?`5YVO7s^6^OrE3c}o4p2#g
zoc5%#n3+|**=Zti8x-V8t{du2z*0aWDgC9Zzn=R4qN_Kw1nWG9GCCDEoN&hRsOu~V
zHI}#W;dG_B$@*avC>vPztzM;n!62*94nid;1N#lrXBp!g&kVkSSxv6v)jpJtcL+46
z%;Xe`k^mBqQcBtH^HRZErQp(VSdgwePay`GQ1}Q(VSxhBP|=(}OWnT>Ry~aM+``D@
zyS&|I$Kt-Xm{kMDzi8r-XcB3_RXF~ABkwu=bZKstS9-W{O)Gz>(O7vYVctbfgeFH2
zbs$H9I~Hr>&A^Ka3O5?-6c7!HnzYQ~_y?6aH!;5?Fh2{K<RjQ$SpJodY<-mpFo-TZ
zEPZ^m$!~xn`3(rxXN`rq{yHPTMzwz&uW<g%VXg)Ad9Nwkt2W!JT;75&$vFHZ$BQ3z
zWB?85xdlpzNYZx8Et90J$BAP7o5^7$r0v!wNJZL9-fl$;ur)r;?6Pfde|RFppx2Z5
z_rXS=F+}3U^aCs2`C=sj58@5(ww_g=xl-G<4FkC^Oa1qXC)y{I4kTDw{56>b>H?$t
z8ITrMrvhrCW~4s{>V;@mGeW%(f!o17ZQv06OHrG?He1vsG**9HVy*l<uINTX<fY+w
zXmoV_OBz_|P$6S{h`lO(7N()s^AMg!rbXIXw`E*@%3t*4pkz(^l5_$p$tLcB;IL?8
z1m$3DR7eMCO__lw5~H;9NAjY_GmIQ@VRXWGV>Sz*UI&E6V=ZJ!C>U_@6wGY*Pu)za
z9UxAOfF>eM7=k^1vR>ItX}^*-2&=suX*vW1Lw)=}FAX>p9rSLu#F39PH~XJ+PA$+M
zQ#n~-=^Pv9N+LZ#>Pag}Y<5{>^A*D#&IPhYCeULB-@h8y{r(=V&$;UY?9IEoG&FNZ
zbxm5vY9`C6&LbPJMOj2dEI><|U)b^-V2<^OU1oR>kAZ0O>2s4kC@3~kuxco!{7(O*
z(rY5!Vd2?>$q4f*!98nJjn-<+P5W2(UJ-b!Tj8SED;SJN6k^W&UH_{&Q(K0EZ(k-i
zb!Rz3Wi+BV3RbN+!cSo(%&}nAS;Jlf16}N77h^zi{)b$ezRgxV#n<?ETC{W2dQE_X
zXXpWCMCQq1mP}4oF_CjDIa4i7i{N`AlaOI+{8OwN3~p!(uWE?*sMOtdT$p;lKi8M0
z+k3z%gfz#X)^A}Fqacll#|_8)FK(=lhY^;x9L4!hO<-Xvj*Drd^r5a-NA_ad>F;EX
zapPeSh@VerY0l1x^XQhAp@EPmnFK(ee?;;a$&O6{Xik6Fp2uSQT)q1sH_aQ-ndcrx
zEN^*2HQt_&{*BPQ2!mO&bT;!2%k*TZPI+!}-)aaz5tAGfF03mv>BQWJ$I2J08<_f6
z-u(Ib($nmJ+>JFvX1*ZR<?GMWe7Ne6K#@WL+&;<6TeQ82@`{vyZmM^Ear~yc6Q<|p
zq9$oZ+O+os>9x6i9yw&4*@NGdD^P6E(0r_>XHuzxfeSe4Ghu+D6>71w^Hl*8DZGL~
zrax@Y_fmaiwz*;6dN*1ig`7*s1+X+NGo8^kz4A2q!wvmuF@3ht@(|{+>!6PF(cNN~
z8bS?<E=cXaL43u9C{N;}v`<kpHWLPqmIUW=FM(417pa6h+E`s|Jf%t<9$D?^JZ4<P
zzHq?Pkh(8+L?P?886NPs;ZKXM!Uc`)_^hAjXQpy8IJ0x#YwM!TW7R*4)w%gLSx`98
zS@C;bnF0!KoLyJlu8nt6^Qx=+<ZJgvzJ{Br%To6`vOvU;dwByvrKx38W&reo2a?rX
zxF?7uJdy%O^%>b9t~;NVo4CpH;2+(h9$*QjdgC$Sya9$wzs^3oBn+x<4ndS=^r18+
zHr^ZvX>rgxdPn0sunh}V?Se0fUcW#tRjJ>E)y5j&(itng(a$}}8p?UqIwL$^l2HP(
zIh@=u+LAS~y~DvM`^0@5OKge3F7=k?R@Z0v`s%MqRjX*<4HTO}dS*Cg!@5;Wl_1@P
z-h6LuV15+1z?H7$r!CS}-kh*8>WEKOF*SxiZ&++57H3Q#{R_FzU5&pn*o{1&xb_vP
zOB>X5uCau2B5{`hZYS8|@wa_m@d$D$o1W;AnRBNob81BLtbXeCF?as&_4;&|#@|x+
z%~$jk@9V2f%w1gXJjTz1T0o8((K4^%`?&7PJu3s;_KFK{#q{Y~jw$pjgI}`SW_bd}
zFf$cyV3ZZ$ArAy?Z7ua7kZ->qr5KkIY|&vY8pkEqlAF7@$=nhCP!^^FJ4vw?Q6I^Y
zjO~wg>sE|s+IEa&FokL}+akN>Do{Mew56Mznz%9OLG?k8a>FsY=H~tlAA|&1QL9JH
zRbZ)<@u8rQ2TqwfY2$!46Yy``?AlDg#$mCOZq>67uM4ev=f%2p?(MA?tBwEKa(<AR
zxBqv7-wL#}3qhMiesIJcl_b#)!4zoFN7OD&9ugZ;h<z2)ac!6KtqqFf`gL$;cRQic
zj<?}|FRyON<P||h7`BNq@5?6zHwwM*v{`l?gb<W9r9CYH*6sxm+j%czLfQz>Rd*DW
zOYp@cnO^~==q|-eq1GaBI~}<~-aOMkpl!AQ;NeVc{w1)1HD)*TO4C@Lu+njd9GIKS
z>Je$x8?!m35i`fvSjI!VyN^r8-|fTgT%*t_z4bj6={sBf#iMs?Xe&~^iR3sn5^W_K
z14!*lpnd<5s{VcVzHR(tC;uoq#a!W4>tbxUL+&`6<zXG?Ef6jCs;Ff?;K^iDofV=T
zBX-X!l;*CS@e<E@Z}V6}8$y^6oHu3%ssH1CXaF<`WzemhF)?=Mw}V6n<~WIT@EnU+
zCv}GgZRs)JM4pO#eDu}-vy&dm*7wh=V7|_s76EL2Ly9|ie1^{zcK}J$2e$`IQbL+1
zC#X3rj$@?uoSh<CVx*wNt3~9mq@oVY<r;N#Y5<b`dtp41L8#aQx^VLw9A~>coQE7S
z6JEOG1G2x^#_O!yp2N9{buG`0VWv`hqT|`UgtN<BkI<B*TYL8`NB{Of{?PhaMcpUT
zK2yY#x1H)<UtGo%Kk^ffo3k5qGma=7o)OxdlxZnW0L#i6-~`N;DHd3d8|OK7ye)rG
zj`KS+$Cj<*9Z!LHgc5Q|+WA--@|DEsH@{RMeCEm}c5}U+buSf5&p<0Z@JwZbCN?@2
zsPtf<q2Qo6yJcR^Umg4;Nsv{$b{r5_GM@h^PC$?p7KDrgG>3dnBGxeEI{HyFzxAh>
zBs)khKNK`reJ`G~?h%a8sTTU)YfIvtPCh&0PSC&`P3PB|5AlwN@c&Tq?|tI8hp)$4
zr{{1&X>J|;EA{B^I}zSbBo@vZfQbI^wfyY}>e2RO*BmJJrW&0fCD_~l#dOIy{4^)&
ztjb8KQCG=ZOQFlDek(cPl~p*%c~J9l;~uKw2YB;hSTdY^3qyj+y1W$dWP`%z$!M_m
zI$vJ?<Cac$M_zJ{ftf>MGxHf_n5jXwO~7165cmQADJ%J%0J}wQ4B|Z}$zTEN_I;O+
z?Bul6zm5f>55asNYb;jqGUdQ?vyo4V&t^PT7yxYWIl`e)lk<5nwxc|3xuNr%ObdVE
z9zvdWZ_vx0BeFr=A{uD40~1?iEBVe*=~VOH*H?~j9zFV`eXeF#UHLHR7+ffk9d4fL
zD|@7LW0W8dg?J}3XX+|%l^&_K0kb)5pVXgy92bHd4-;c-oBq)j*wuui_jpcI4DX!t
z1oL(3XF>)B(-Hs)aHp?*@HzV2s)EYqtHSU5=kk@i3|pWRy`f5BiBb_HWa?=!HBNJZ
zfhJa<qe_WI6tDwuYefVD#Vvvj;L8;X2!y&`K4amI|LS^<G1mL`v#Gk%rF+lil(^Y7
zSC%J_9X*o|whQkgbHFoU?iAV($WUUna~$b5eQMWDO>dv)qIX35zQj)!Z+&gaBm1ZU
zK9@TJs0EKy#A8Owy{(?UDc{!dLV}pSZ?dSxeD+%@5gF@4EzL*-$|F3f%xxkct!f`N
z$NW2%2|U1e+hVQ<?xNhw+s;s$Z6uuKM*;W$ZggTKCC3f=CkqtQ!s8Y1l%_yEJ?~^G
z4y#M6?`#Sl|6AAo*YkqEs@{Tx%#~7ImSxV12Je!x78>tW-#x#?cEfQPCMTDaY&w;!
zc|c&M=rDZ8sIv!ee)bWOpQ4P~v8FGp8s`y#=3&1fl_K3Q&xvr)15${<!|Hhgx60`N
zl?C4?+IRMCw6C%7?OgI@ABszK0nS%>0QhhV1mFSM^1UoRd<)T8-;h*VLP``A0cu-s
zid;+Zp*3xV^_{J$mf~;mSSS8In}eKa^fajLU-^d_<^PM=<RV+N86P?Hq>UV%lzltP
z?KRPMY)8>d%J2XLFm-Xa)mA8_t|k<&8ccl}8Jv&qnejMy@n(jGkl3E}ZCcK{xDUD0
zfuJQCL5JZu_P6g@_&sX2dc)_FLMr9AQsz!M_=lnfDo*HSX$^$`CFK-Gt+rmGX*pch
z>n_oJ;J>U%(Wd21_2JOu?3+?lmcqWsHti=HxB1&mSW(`;!V<?yGM#=Pzc%X|z@JQ2
zxK0p~$u|I@I=!7a8Yni2%>;W2aXmVMeHNzJ36&3H`C861l$96W8?c@UW;ONwECsqI
zB(v~GvpO%b{{Ke@s?Bu)rVeaoX3#S>_%F=A?)Gk+PRT^b%t)(9j%AOZO(9v;Cp7SC
zi^b*To~Pek%u1eCoDS7|W4b{{*v?zirxtjk4XgsFJN-C3uA8N(fvLrgUwnWQtxy~d
z!|rB2DYs<uM<jiUQfwQPX*6I|5_<K>_FiY7$r*@25xD~cR>~8<lRu-v)#rhHM#hwH
z*1u!`?I^x+1H>tT`Jw`!5mqMLd?x~OPy|KxLIA8X8#}e#jc-)u_73cCb_2xDBdU!g
z@j1z&D(@=@v&bKY3;<(MTE#Mx&zQPB=n57~vS|&#?yIIXBvA>@m^qLHv~nX^r&BT7
z1KRF+Qpxix&3avT4l2{U?QjS<owbW6^*vOroFc`B478l5tM-1h7eEamFAlnGZ>5)M
zF=oX>3`selv2>utJvi^(kzNdB>WW+48CFJq5%{wchPkHTy@^9omfEySU}RnB<Aixn
zrhrHZ=J~HrK^h@5h-o*DE!to;H_KR0z~3zsCO`;^X$~GB=0L5!%$vzSpm1i;lf6=-
z1hamx((NYLu9_pX)m8~qQ2%Y57KYnDE|jbT_<_*;CMH7XAR4`|^BEkHznCjLw3n~N
z<=WT+==&#OEml{@Bu`W1)pAk3LC&(`fay7vl7Z^Cvi%FKKw#F{p%gIdB<OKa4<2vj
z`O7wgVRJxUTsSN!C}=!Y@}QWVb4BI};$`n1XonOC+Qi3SkCsuvuGY4xf*br_e4jbJ
zaK`~pQm(AeLJZxF*gK;r$!gn8*QDZN%WgYT_wrISM&Nw<OI?kZKyPLPhQ9hWD?N37
zNL>&{D*LZ#4>~`=ng5H8zk$|=p~`e@TzEpiy?0OMtANk;I^*^R<F)o|u(?8l3<sk8
zt7bI^VoDfzlW?NZ@&{CxCz+~TTnqTc(kMjBf=gl&d?N#Y<#1JL@+sn(5dUJ@N;PCq
zB<0lNfPZDl!oCx3VwLr)6Ilvtua$Gtwr_!$2qy;Jt3IOZLJ6;6sw^z}wHFnIl{=S1
zKZw0q$opq6E3&Uxo4Uq#Nlr<#$1(sDaqn>M`ov`p{z*XMeBIjJ4s@mKv^|m@0}<Ru
z-ugkX=?{1CTVhSQT!+c=_}x7-T40X1**4vA1*?HR8+A#HlodPak|5MXU_!f^=!%0n
zy}9J=%wUl5DE=*`D?a0EY!dux=WJ+fy<N&`o!J33dk>cg@EWUHx3?i+-3o{5z4Ga_
zC{mUblb=P*4T+kSGAoKI4b4YE6b$Du6BREk(Q}*h%HK!BB$2-P>_rJ4>wmG>e_W3#
zHX3zk0Z$t^A02>I);OaPd3OVg7^(Ew9Ujc43-uS83*o4>mR*Hirldn^W>#XZ7c<fE
z{k^U37i-7DntnbGgiz{4iA)d6T^@D!i!o39`5sv|6N0aAY=$ILxHyOmHg>k5B5W@B
zo4NbtokO}RA*S&)g)<k@x?MzCH#N9H*IhwwP)g+Y4*Qw$<Jx+PYKD5t*YSQzG%wjM
z1E+xg!2Y6qkl2?OuP`0Bh3h^xWK1%%SD#1bVZA*5hrz)0&CkvdNgwg)?+xuR6w}MK
zZ}@AH+Ekldp??Hqmj>2koz1&7dGGQQ|5)`b*tbkYz-y9Ma~N6eA3AdL-381BY8(Hc
z>Bp$rXYC6Q5T~B%k^W~ecg)AR6JsZFl4Cz~`GH@GDb8>Fu)ZD$aAbiY5`6(acdARp
z=?eRYQmHOEe>{Fu{xum^ekvOUk1nu#A2}FkYd_o>PnU;!IKjiW?!e^-+@<Gzs9<SZ
zPvces;FssyW49DSy!r-cWW;rU7TfP##oZtyc~=<LzavMX$R1xj*${gL0@bejj65fS
zQ!BOimmlNjMh#iJj*`QWCqMiR%i?z8(}EoD*0+M;nU^y~1R*eSWWgO&0{&-oXZ|p%
z1krB62<JYpMM6iW>EqYa5r3jg6LQp7UK>p$DTO&=iVshG*lvi$=A7v>GCeNZ8<BoY
zVNLDMSbR<V<;&c}Ze#5Vr+0WSjHV&QOYn2mQWlcI!-sc~gAW4tU?x~f<*V0;F>@XK
zgQ9*bY-x;D%^dXSsOl^3rufi{p;Av+>GcHN#X**8&*z)E3HRT+i`<h|GyvCjCjZy=
zU-Y}P?an$xAGB7ZORnBid&RcqPy%tt0TLS?13GbKe6m{2!JuXVn9kh9Kbb&ccoNQn
z@8iV<dqPT1{(C!<a96JJk-?+YV>iT63y;*M8Rr%edj_)eXZ*upe4ZjzUcPZa;yC=J
zc%{v{Izq|S?fLcD-&)jX)gx;oT@>bb`At5bx!~y?u2fNJ`Di&aRQJQv(AVSkk>Y=D
zNQg9vteg9^F#A;f{lj~^e7PiCTna5e+gm(!v3w_)lqP>4vP~qHC0F{J_)qkR_^BJk
zVv)$*A%xKgcW+Jifv4aKqqEYULomK=&vI`g*|VolK@^K7c8I?6?4(!@DY?CvT*Ocp
z1{CA*92UxRjABTmtZg_wV3_&3YZ>ucI~?`gh_~>3YNoK3KSk0S$#4EvujmDbNfzQn
z<GkM@5s<Oip$jhnO$+!*$W8cE@@1DXu|@{DmSr;_n&Uf~E<FH)g!4Q(jW)}(C1%BT
zE2u5^?ge3*owyx<_(+BSV03Owh;?HU`|XcoxzvA*$r+5hQq)H#_dHA*C}6Lcz1ZEQ
z6mfIqp=sMKo0Sg}4?DVRr~a7v^W(JFtrNbliNrtm0(?G6AtBi~4x-R`zK$>%KR#tr
z<52^%n2I4(HX29VfzH!9F_2|Nb?GFgS_<uQkhDzOdX=<SjA=^fS$Vr3UPspYl)KPO
z+gUKrX@LZV6FsLjXy-g-+Xc<!jc$u9uL(8veE>vX;|jRdKOXT=Q;tTE<VXiVNBpa~
z2UM8BNN8gaMg{CiLRT#Tyv+o<qcIb9QCQv>7H-tK-)PD3_;%Op5oY<xEuyBf#mkZA
z&Ma7&Nd01PiCh2)RM~4z5D?Qh9dpk2%gy|_AR>u2*|gl+#|w|<!T9uXqeZ>@RFg7F
zi@TQpd}LG94J-SI(5NR{C9Ss=A1uj|qUpPR)b^8xGE5Yjl26l}xRRB<2$E@?9Pq0b
zBN*&BP$u{Q;k4hB4=ytvGQEnA$xy+w%Sh%GME)rPvO*V<c=y^J?u)w*!7N-iW{qj=
zXoe38St)z2vyyQ&c8T2ai+E<v`TrEpY}e=_bVKE1yHhdAIY4RvAcx{q_3(@?^74?9
zoibYp&7PL{Y5xON^Fp&D`=3afSHH0+gyHww7rtp@lYtycm`f(c2NhjR4k8a`k(YO%
zr%UN^605nExRr40Q?5F%2>31)k?-=)B=|^`{HA<vJ$tA6cHBU{v_9!<58zVWx(0t|
z(i9VcTJZ+VXUJD#AYai@1pH}b23!MlrxPjHhus3?pQQKaC|MrML}>d&4?%!EG_R;3
z++2C?rO{EvtH6b0ic9Wu28}FgOvoy(!cdPT4r{@(35<u*)B~C!qs`1aG*^>ngM$`V
z#X5(&+a=&)I4ukf?*7_YbPrR}&NOm(%fFpO$j?^+QmQXH;S)MF*sIEkp?ThyeO7|G
zo%*Sk!ePa_-ev;VMLhS)(9Hu2HmHI>HuLiPFMm`V9eHGYP(a(UtPgou$cJ{`)4N9B
zQpa}$Dy3H6EI60^8>EM6%|y$xMd=_RWOVf@T|cSXM^mbWyhVO=V(iu#X7F0Y!b20P
zRP9@Z=1OUjr#Dv?!@$lD%h^Nxf?a2uV65eOknN}^XcIvjmrVz1m!7Kza;I^-l`Ha8
zj9scBlGW|I-zao(s}va=8Ex%DNbm<6)jy^=L%`_vAvQ@8{GvqT5ANPItXN|Cj;j|X
z&E|;A@mK`|FQAj&d6MY}X<3W3jw!mkQhK#{IbY{+`wvWPU4Gs>)>~F%<Fov6Il17n
z{N9Tf<~)P$4<FJ&qvf@olg(s<psnW-5+3+zGSQTF40icOwJL@twTF$x001OoYyvGO
z;$C}A9&<JvB83*veXxgPh^2IIb+3`GJEXoKjd=ZyyEq>PUJ-UgPOPZRzmH4s#1`O3
zSXdYJ#`m?ahOr{Dr$^s$c~Qu-S#e%PU^!Pn+C?aFx>=?^Z{`MpAiMt^b$IllUhk%c
zm+Km1zM21l#}_2O(7)^hb|}-ZFC6lxVC5+Xa_D>0!g?^4=U(^pY?J?$jwn>Ph`9gc
z!?8cMSj&%w4=yq=&0cP^&|!#-K+9O32=0?wZW8^;)Emg?@D`wwlV<=#4}e|BisG`d
zGV!|xYd2G!D#}7Y=!yAdB1uWB3j+B)fX)TfuCU;_yKkVs&@ep=II#+7E<_+?9)ai?
z>-B)iIri;U7VJIkp~>KLL#x!hCV{`Ez5L<37L{{_?{cnc=LFsP3hF)%m*6*fQAmfO
zdIW!*bQnoF405}v89b;6bd1rCMRIQl--z^wf<JmELA;6FUk)OLrK$%eDt;xIOj7k{
zn_LGXtrcF!#dG;?A>H=Lcea=gfi1W{IMMaN?Sx~eiZ;oqfeNkbB4nQ+UF7fQ!K0T5
z9{?T<&hZhckow#9EH-weH!GyM0M~twyjI#n{)7~ED0^-^%fAKGq~Pf^sN4yS7YIs4
zCa-G%iF+bCxNAe<hRkU_0>NSm?PjXWu-)3YZGN+6W)5mYODmxyIwWE%Ou-=|$ayLD
z(lP6=OUTx3q$MO7skg@6i)Cv>f-K<PdnNtg=q8R9CE?K0u1`wwa77s|`tFCY9JGDF
z4tefzE6O*`;;5MoN2~M4Pq|d$hFRG=4-qHXSFFZab&g9LNQrP`1=?H+{><A_R=TLo
zD_+1|`l8hnIayW;F&JMWfBdoOCQ#YgRFRjFAQ;`aBX~z#A;J=Y??|Iu3c`3jf)S=j
z#>7XUO|h&~kU*w;6XbeFcJ{_hT{#V;6aYMq8A*XQWHU=vX#oouf-nw&sTO7uQk1~=
zD|X35*QDmB$;5V0Q-AADcIko^V<4J8Nt8ME@|x_ZhNkp!1>3LLQA{EvtKX)dJiql8
zU@l{^5kck1fnCPq$xN~&LStE0pNUAL@cI0S!zHDz79amaqlqsTsT^zP(plnH=3XfM
z&Q;^Zk}NWF<1yG1U}7I${YF$|K;eh*XXwa6PG7;}k!)Wka-fASVw87}SCC$E0@a}c
zC=AnC$*{dI<TIwTzO8ztOiXb2U<*MGj848l*@IYY)H5ouEwoWsg446@gOQ#I04ZfE
z-1UuhkE3%QuHtuQv#Gbo0&ooh@We`x0X%mG>MSrABkDfeIas8uzdt1|jFS1fkPL$n
zgvQuE1xuy#K3;yb%WAJ@g#c=%UmQZIo=Qt1HP1oJL?lN_!C|6In{{I<x{{VFk#P%B
z;P?!|xnmLO_NdQZ#`KtI!YnqYC0ue6p*aM}%DAcab@Nk)GR8joM!1EKtH|u;D^M1J
zpbAt*f*{3&D2D-oSBMN10auzBBQotu^2+ul^8RF{71D?aAp8A##nLpN>rn03iSVWF
zM9*;y!D|&zOcjy#lCl&BNllb)p*i?_L1AA#gg;ZdxPAU|^lVnfBnjwVE&Rvv2A>+m
zjbDMfO21eK6Y#S3cT=kx<m*N|K#JHycmez3H-WwvKiqy>>F+kiNbzUEIjT?A2Rxr1
zA((5`-t*wtzk~b1;_rPyv#)~8Etg*fy?;2_bxwYmF(E087YzKtiIerHfsH`!YKJ&I
z62^}^F!nLid{p21Aw-&`r0T6|ckgDg6NvKYBOr-LWKsm#Q8|#4qCfsA{eJdlK>W3P
z`UHYE#q@TPHjfQ#)%7<pc)?pSu$UpVG*U36zXWYwC;c-K13YTxz=O^ob6*WVrhBiA
zu)6|Y(C!!U*d;ZRSQB(nV{yCILI!$)m6v=@lzqkNPF{09eyQil{R`$vV2%RHm#b4c
zrZ~$&Yr2`e@4<kQ4ybw!`+3Hlgt$>v+)yz_v{`pICb7<pUo+e~CzlZUr$uC4By_*y
z{f~S(T*j_$QE93_c400_&2l+GVwcM&biHTrnUo1|N}bmfEn?gr_ro%i8=J0e<cAbM
z5pbhHotV%$1unCbes*v7;qlEf6fEQ;TtIHg3OVqrl(qHBVfzhiqfH+e_0(bFnTNA<
z0ra6^<$JnV*#H*YjCL!h^!m%7#4CE0Lk_q51EJ<O@~d;EN~YmB9+GpU@v#@FkGx}P
zi%!5`y6a)$e(^8x?2tI@Z>JP^`R@6?mh5M`9AK_mp!twq!{g6+O|dD@-2M_v+4N?S
zmUgx_6RU>L_s%O<#QIui*IQ)Zst2d4ykR%o2xzcZ$G2{qpM-MpY3M=@zG-GU!hB)r
z#;wq}LaipF*%EcG-%}OlX_sqs7cD3*BNaBksQ=LOnQ=MYl>-L`R$W#LqRiXNV^i9A
z95!ypWS-<%UEef0>OGsWg<bHC{$)#-uGU-Z8~5sFCX|r?dK<{&gJ<G`N#NJ!r24$0
zQg|X5V@r&8#{*^g;-(Kr8c%T{9f@P@R@^U!Dy0vn$li%!t>vd|f+!CpBN8>R8a}<V
z^lSKZ!b_fS;nOop@6U{Zh(h3TkpIfwoGmW-3c0zoj<?(S7bEr!RYjJ|ol#(;T9~mr
zV^-(J-}m`l2A-jee`qYbb}^~-*ypj9T9?cH<wJPpcZx;Ruf?+ZzdYua|CE5O!Fags
z00ziY=?&n<+3sQq@8CeeP-E!p3)MA4tVfzQdHT6#X7;JSm-#p`vwxb?6zVE*Re=8A
zs(bw~ayjHLXPeL|l0MWIywD?eUx;gWL&p1!0<F>2gy!P)qz$@93X8I*u0#4PqcIsE
zMO`l>e6&+B1?>L1X?Fell|6HY)CzU?Nkcw`e{B4wRm?%jcN62TEY<`3L4Px9IipD$
zwH%)O%c$jI&8xcyINTV8CsXKK5@C^(U|H3e&G&9uwoY?0wQQtjd@)neS-S0Q6MlcF
z-HeaPgNwd9$`nKQYx^o#?2FD^Sa^KhFP2x(g$JGhDOWC8-b!aWG<^8^R`=4HP#KuR
zun)|bcEr$ezw>#~)cjrwMAPq%hZ5uD-ABTA%r?mPFDM;0`sV)x;Q7Wg!S{8IoUw6S
zw_@T_Z!Ak_Eo$yn?c-NPLb498CX_}=&fReFdAfvEkVBX;=T8ucHWcbG*AKY}mw<`R
zdL4*8P?uHbN8LK48kL=b0|wJtrhhE<P&yLCiPp6Z`+kc3m4}cylg%?$vpmMNB*D#W
z)Lj=M?MY>Cav3+wY}nOUHZ``g0asR%ZEm{aYq1a<cXmjBqxDMJ%?ksmt%Emo`OT9;
z+H^W>#Qe${FWxHL2T~@k*pY?(AFYLn3$0OD?AQqBAK@@~g7d8Uf*+TXyH|mKKw#9q
z<&j7qs+DPV#iIG8I{*2`*;lv53W$C1$;u0I^$`W;<@9d&+m@g;w*qoinanta0xRAz
zUG{0YQ%<JSz0$nR<D+#U?serZzZYD9$3{&dA_R!jiic2{vV($!2k?%*pjI$F2tlS8
zD?Jd#QGU=GOx4wzvK2vbscdXs5Ev-O>l}X#Cg>ldu+>X?Ik^(t&m1QKKi5QI1D55h
zcOywWM#IFe>Kuanroc#jFv&5FwrM&#wwwb7ADc$H)I%R7fBWQ@qwrSdef#2T`j&TN
zad?P`Rrq|L;RPh6keeC|3shiKE-CLD*|RhzP+QuTP)BHkKO$Fw_MP?#E=$F;Bd8_L
zXPw@G+g$H1eZf>TWK78wtc|mTx}nLkV&78;7^C^nUTS}=0KfRgArjXIl|~+ZkqRVp
z`(RAB(ZmQ!49sg%i9yskMcv1#4%31{$i7v_Ghy>?gMP!6Vu1d@ZwB_yI^rX>ySYno
z)f-Nj8n0BkKjdPEz1d7;dkk1-XR_h^Ck^Q2_*P4>7NW!<J#iDAQCeX$9i3KbN_1yk
zg8#SRTh^lkyOAOB7-yF^13${N@`9m!dE;;Si5$|`A^<$@Je`K?e#hM?r?7VbGcsVO
zZIa($yJXR%^*}aI5ew@CV5CwmKEirm>b2ZRK+dak-s)FMcmt12{`nb=ObJXsPAK_K
zG}3jBUc*(VxCCFF<b=$fFsgDEm?uej#8a?<iW1Rr)4a0lyF-xzH73`M_lSBfFV~0e
z1OHoyb==(+0@Z{NBgp_Yx?O29yr5I5t0d3ll$1MV*%k=!!)wSY9x6egcRmc6!<8Uu
zR=|31V273_q+rDV%6$o8NJcqL+nTWbU7X0Cc(MRBm1)bP)i|ANdf_dQka{Bo5SZZ-
ze#l~JCm4kRUWDZ&wKrdL`e+bhxNomXoV{@wA8}S<acYM(965BuN~AvLfzom!;23Lt
z2B#F`60lXj75lwkzpQz+dL@lNmoykgMwgOviVXKWww+d|WeaHOIwvOTKkMr=`P^fi
zoy)I{SNNml*$>+$aSz3c8b;tJny9DNIgwU!o3aGye{Zh?z7sqhYn6?SZ9u;QRew<b
z1#I>r)d=85Nyn52Gx<SnGSS=L5$=&Dg6e|(v_|M-6FrFAF*Jf=BE>O*)F1X-r0&~1
zbsc)YJ1yT3s+Pn^OCgj6J<cF}@uOz#A`}g)G_;g*2n%Kc+20$thbJLA95AYa#&i;2
z%pNxr6pKu>OorSsj28-ma`#CGlFzBQVwa7~Zpg;|6tx#^3*oqp7a+jyJW)N{Re<M9
zYP8WH1Zt(M%&*K;&2Bcdn@in?7fp}=<1vjL#W9!`U>anbDc1o=j^w%qtomnW@{j&J
zH*4|D+|u0_x!H?MP=&_D2_cj13M{G9!=Ps2gw_^9W+V)0(`lI>5ibOiBTXf?JM|X)
zbe?lBzsO0Ur^9j8=N=!6{#qE6cunLchC*y3iX;e|2$h(dcrS)V3uo}|PFOw$dC;r5
zJ1|jp{!N~04%l#GiI+$6agEz0E6#<DCSDgjiDU$$-En`So0=j0vu0CajSy-^Es`6>
z<usNbXe+|MO{HJEq<O-l#%2JP@QXtWfAnF3e;*PGC7YpE-fn`A17m+I4G-QQF(K;=
zeAXg$95T!Yry$87W;Y;mrxV}Y0-MH<+yIh~k|Rfvcm$*_J$?=DBla<hv#cQRaU<Ho
z+Ma2mf~qy$$O>_h9?Enbxv^+5dcS#YOn{F_3g-KnZB|(Iw`{Yt$`dWE*4UWLqmPR&
z4D#G9{lX~AEA)q@ie5w~izKGlT&!4~sCl3+Kkq32+V7(~rNjr63wK^B8Aq|k2BFnL
zS!3ZyMsA|bBK%NqM<#cNawIcKu`d74wdA@O?y@4A(o_9;SmsMv05_D@6ORl=xvHnM
zcn*=$Kar*576i{A0s^^vArsj6)YKc(4@wgR+C+?)SKkN2VX_CGPey4WyufK4BG_=t
z?Dd7o5i}m}pzHUVr~rOu2%+%|V#%M-&|?7m1`F{ohRN@o#D>kxEXFAiE^rv#`IXTy
zd;sa(72(TB>5)9QTx7m<YD?=-W53M^D-3qFmbeTd3Ao){(m-$G;v27o7kc5*=Uu6?
z)%EPG#&JX#D5AnNz!bPL+PFfc=s!-ksQ!LgcDYMw?zhMDKwY=kzCG5QswK>oP760c
zYl(4$qaZLWP^C(<Y_HQFUrJ_zj7OeZIa?%K{8<PprY6oSafpWS$xbU}Gp9qTTBA4c
zjljA2SZWBKFlfCvU`hygE#(0z=ERwF{yr;;E{bxzwASF7<gHtVYm>JwVK8ork)oZp
zchdl^ZiwA12752JhswpmN!7$ptb)da;zks7(O6Te;<rAHC?D|&G+&S5N!J}J@v1gl
zrp>itM$eM3;Y45BxFbD-<$E^&@h|&)T4@O`xcA7l5zV_*Wt3+nQfYQ`^^*4(<&FV!
z<sR3;!r_ggMV6<t?{Gw&ha<=3H1d<vx`s}JtC~_?CbGQqEYhkQ-g)nedSzmK+}P;!
zKBQRlpd_7wr(4?mdQG)x?8KPl0w%bjVgI~8vyW&9R<I|fhDM-70MEi@m^f1w=9#SQ
zgxLm}arRw`@s``2EV4!`th|fC-r^BNwG5Evr5^g|3by2XO#N91-mDW;3>F2T{(Meg
z#u|^E!A4^h*3<hAKsxxh#F4B7%JO>+ilE`B*r!X0nHUHUTYL-^n5Q3CP*+tnFQdx!
z{@=FLg<3Br7xd1gtPGy8=~qgKU{^LWVoAWR!>-p`E@ML;z{_K&z%Q)A&klifjiBN>
z^_CaPRC}HGtR$i~kwFES8aIhblJd}LhL!JmAqHwt+2e-G4JR)J-~sxBby5@&%u%#5
zQPJt?F0kqT6-GS$;@SZPH_!eJ1gD$w%<`^u%@2)saAu1qW`9DVkeX>BGId~I3{xw=
zTl5afhUMEkSDX%^hty&Oy%!=S;B}{-<^p3!n^`SH^0_jiJ8%6qJ3W1SU!>`T8{#S_
zwC-BgP5nQFy$L*&|N1^YmP#m5p(0tvz7#5CiIRQa$r{<Ik+Lh%Mz$FwvSk-Cw(ONG
zLnx9&RA!P?7}>JS|9(d2b2{hy{r>*v<#jrr>bzc`=Xu`md%5oGzOIe!YtK8vWO-$H
zBdh!>&wix*RtcdSUz-hHzRS{*=ACX#a!zqJLTqQ?_TO+s?6KiRgx%H#u=qTX_zS<A
zn^UE=iM>?0r_S$>_riv7WAu4ghqRR6C+EwG2yooV&uWY=mYzQr^uHqwy#9ZXhEYRD
zBa?V+9>%~%fAPRy%h4Ca<*iJ#Izof+Gj8WL9!7oiwt~ceU{kWJFbkl)?;QNE_`fwJ
zQ7T~&55t3lOZ+z{1LoJ1#~Po+g-`OUM67NTN&pkd7n$L+8WgUU<r6Kj1;@)@U_kS#
zY|=)Y6!+kH!zKlGmzwnB$mAqnNMsiIO_A)e-but`S}>i<yccHfQxkaBDDcP|EFbbt
zs8gr-;n~XZ<-J=ZgnS{&EoGrWvyDg0M#QSf55-|wo`KBi#3V$)1ldf<3P{mp#CYpU
zJCg-?UMotHHL^C%L)xhr!ffHkTd}Cr9)!J)f5YYM%s?pif}+3^mY0f-qUDKSy##LK
zkDoacw3OL;68{|q+k!A8&~;GjA~?PPRWELn7MYFb1kZqp&&p$pEpf*s?K+y89(Kln
zV_^G|opRgwG934ob*zeJ@6n^G!#&dyqMI&UQ1<ciJC%sq%R?270CYm&4&U(#4>mjA
z>Mm*1Tc2*0Jbs%`O*lt%-r~^B@Z!WPwYh8V8u{E}5TzKq%xtj8nz$D*qAwGPKgw1j
z|0Npk2pz2ifspPofV<3TyJ;hkOlXm(BN2bXLK2av$SoDNoz~tGkpWJ$M`VCgBolj8
zk;UkZ$pbwZ=9-sz@@ghwywVbw8=eeC5r>~HCciJ(7CqC-_3$Ve1meScO|{;}l}`i&
zAD7<E8%g{`+1S={tu}nAsqh`>tS=l<PzN5l{9$(#$vzD^*SAo|^KG*s_unqQkFnLQ
ziHplZ!_fwmn=17xZ9_%+Q)Q9*4#BFGzKV90fNq=IHDjT|NMe0%NO4szEX9j}25u;-
z!P?Lqg8Q)^b>y1+%IMubPF<AvH&f#F%|4hfIyQq8aDA5IHrSRf0vok=%bzC$bY6Hw
zz#Pe1Uw#<o&BNhfc5igwbY|n8y?$A2L$YT}QBv0Tl#YTS%hf3G3TV0abW>La$Z}M1
zWDx9|^CKlG>gy}NA8R`RfLi+sgu4xeu_sS>c7b2k#N4x+<84wKy+d=seUvvP6cAJU
z9kAnHSG)`4yxv!(<SS{Qc~w+REW?OvYcGFYLGb>S{q;`s2*@9BPZc^)c52}#Z56h7
zpd|+fU}IJZ7dXZ6=ebOnz*xu$-lIwbbO`k!3=Z~FA>gg_^~q(hd;d4c2<JZS-vZ$m
zPerNgsp=<pDS0CTD)612Oi?{TV39FbA{xUp_}e(?SH}BMzn9@vyBZ~!;=!2NWA%yg
zqIY(L56q|SV@}2uE}2tl4Z}jju`yQVE7?)RK3k;FY5a0r6<V73G!iY`d;5;@)iIUn
z?<w<_E`cnL%QbYejtt#Js=XGRhZk0AOl^?knB3-;_q}%^X%q^8dqS9xQ|Y6|;zH&-
z2)+CK^FLc6PDby8+>n?T^0|{x;0AJ3Ccl?<n?_*304!=pMhSKA@P8F8!vvUKIIRIn
z;D6}K%3Xwu6;R(fo7Kn@2xnAoJje9t`fvrScR%FU!ZjU{?r)%c1IOZb7oSfa5(BGE
zdcj<f(rQGw2_~HdJNxF^+ol61AmBZD5(gE|O}+SCYK-SKyg#XI)P}K#Pkr9K+~yn3
zZ4o_Ey-;VM(Xj_d<G{oiok!6Xh#NR3yjcdTZFOK8wEx}Y|Ni;ZLtPsMm671lpEFN?
z^S1Z9!*h4$I#NL_EOD}VPHZu^94Kcuj62-AK+zPcxm2p57)s9Ysd83KZj8B>)#fGm
zM-HKW!Q_4RrfwT-QBiQf9)%o7g#iCZB3#QQrcf81qjRC#T`0!~YlIG+P9=AG%pb9+
zQiSCR6BgiNabvi)MSj7(KhVk4MZNh9odiHte9;0efUlJQ;BGU3#}|AtV;^>_We!69
z(#9_qS1F-1tw#rwyV302e!wn7z~2bVlRAW<{HD`pdso;%^O<MpereU(n_m@4Ngu$m
z8|7oEhots5RJ_0$vo?A>Og;OzMbQaF|GVETKJEdixvD8^nCJt9Y>+jv(|^IaI?ik_
zDJiQlhu5*N*Z#QV|6FutV@dea3sa4=OhN3zD`d{E5$!Z6x2Z$6S>yi8E8wM&(3L`h
zdN<)npf=Lo(XO%_67ua$J;V*17Rv_|o{_3AbJ}A#?091UHv*$fO8%VanO~I~9n;G^
z#2cx5j{>{NaTMZ`N@rq&0C*z4W{Y-NUE8Xlv;VLd5(Nc&`>Vi(_?6R~7>phOL~**G
zgv)&M-eiEf_n*rBM;7C@1JmylUfZ5#NbkjEzX9Pfx4Cza;&*;*G*^NqPsz9*!37$p
ziTaQJr}B3fkVtf{nH2vYgn?1^1Dv+^sN%)AM7`w49Tpz`qxtES%RSlWj5^#ZJke*~
z?ZIwjFs5C9=_+DsDCsiJF8UQw0B>XXwF{*4+kc%@)30fH@csLz6y^2Z%TaObZogkZ
zxH_}OL+PPx-{{w{>wxH2J^|x$c7hf1+Shsw4GlfEi<d|OSus$C>JIc-*p#E&)H`*(
z<mu0cBi5Arz7j(S)s(vT-|7744m>N$YNT@*>Ljchf-uHgX!NsSdyWG{qzq`d#6B`S
z%`VD--|&pU03LW&TYmVq>jKg02#5SWVFRjiDYMC(w82ILT1Z3v5`E`X961oIRS=@m
zjtkQdFc|aHW8xOpV3Ba<n33nUXJZ|fugPC^iOzA?)W+pukNqF#$tjnsKR{0Y`1_4t
zQE_36Ex)ERpRQlWR^0}bkC?!);JefUcZTwDL@N+LYo+|h^p7niD`qo%GbYWza2dJS
zmn?H21>QYxMsh6_6;){uG9z`M4kDTt%=J0sJi;?-+!NeEz5Z>G`B$a%LHlG|-cA&k
zT6ep@Do8lzkfW@8RykFzxYeS2g9T@5+!^D6Zo4$1s|*@w$JqW^M$YWtd6Vh?TuCTq
z610wzwVl<uo5ZGtt_}Md11#ePiF%5ob*O9<x|nV}h`>?4;_^NN8oqyD2apRwR{lj6
z4E7SYkXThL><DR~7PMBavNW+g9{R1PN(<fx_h*ZZL6v7xVXM>6E}A4@Ka?R2&tgx=
z=JbD;^zKK6A=hEFhQJ)&ak|0Z6_x#eH_OV~T|?9v<8M<-ea+W^t}eLx!op<Z3`5sV
zHx8T>%=jnIuivJPrGYUBN)LDe-jb!?*8C@o%?&PpzP{T~Uor`idYC)><f(`}2r=;a
z6h1B_!zPp<(1thp#U$qrgcaLfwdE~(;&+1aketUU2zOK0p{-eJkAcZ-+#98tS6<41
za%f<&tJekWuDHvu^l^vMxBQw6r_F2$Q1W&HG-KP!;=oOa_Y?($X%)PRaNM3n;6u-f
zryR&={lj2%z9DW%=z<dWBwADjkW&+o0lHZzDmJ2=@(N+aSeV)|FE}Gn(v(^QiLhJ1
zI9g!rn(-<uMJBj+fLoNAnSRX==BJ>jHGd>4HmeQh7xGkryxsuNeo@;fO!#ZLfa~C1
zuGg%GDg!}R01P9yf~xPMbAMbT3+?xh?V6-7PQpDhKf<<pF9bm1zKTUAwwa>G^31&5
zITdkYaNmC)O#63Ri@u6X9nvP3=IWwXm#-+3Ti+SkOzjo`faWD-#eVJ7+Jl+3^V9Xf
zXP4~p=vv9J;I)ZCs@(a{?AD^qH<Dv)!Ev^5t?(_Z6YqPg4q6{8>`x=ilAxK_z4o9&
z=p;P+CByAq+CNU}w5`5MoMWvD2%-J>rjt<Pf%h)xGDT`amTWQM8C45syX{?{{*Zu_
zMQOH>j8jd#@AuP_{zEJbG>#BoU4z2@^A3-RlOlT3v3H|y#Uyis8e;L>xW`KVWU*;w
z`1i9p%9*J(lg|Igb(g3xkU~cEy8-6%nfZJd9zQemd*1|y^5R1^8Bn36MNwWCwA|{m
z4}p^BNBIPhVC*HN)C~2Z=3?x6WPby_BMY$(f{(j`-x8`4v^_LD==LIAnoEbmNqaWa
z18>I}dsE<D5jKUI)yaP2%Sql_yiOH)yJZKkSRYo7EX#nc3bVZ~re?T4kjBcSf%nnQ
zTJ;h79gwD^ietIS2l}=2$fC#of2F#820FD$bN)lB`=25>;m3@*@=t|kBN`1Aw&0i$
z@>fkZ7hlR8_^|&+2@Hl|!IyT8k7Xd)u=zz(6Mx80XoBRzxicG{#POU6)N?b3VE2Bx
zE_@=K9<@BaHW7YUOnx{g?m>!Gh7DDg6b?tn#-M|kDBTCjZ*BR~jE4(3pAb(B<&fF8
zYR)zA|JM|JgO@#b9irxg$!ADCP)sGNst#*+oKb5Rfz6Bg;$pmlI$D_LDTtFon%G};
z**fK~xs<BBywo3(;s0Lw?p=4n2*VX2)3=l}{#EPFR3LRf=e2`IGo@>prJV9BSftJU
ztgL(`j0c$r!0coT0@?`RS{9@>csF{lG#~vb5V^%=#Pc?qT}NK8?)$VfJOe|j2-DX)
z9Yd%xR3q&y+fyb~AC2OcS^B5v1`ej44-fxVH7WQrir~NFBjO`xiEu16E2u@k@g5ea
z{kn=W&X#9|L>32;<;|t>$+*F9uKrskWHRqPIpqfIT_Dw~j*Jf-TXt*3VIKr+PRm^+
z{oS<{Ur1cgK1$UB)A@R14#fLituc5_pYjmILv}#jn8GY?4{PpX^R=9QGe6cID$F~N
zBB4IFL@4DiKUu-UfKnb-BNBN|G~GPjs}g%W=HA--v4p40aHutBXTs)HT78v&Z_LuU
zCSq+tV3YkJw*ivgBX7NecqjAYwtF0ezqkh^1H;>0DS{c)_`WF@kjr^6WVVE{{`uU_
z;%@|S$ExTzORzl>IB1F%^4)_#9CQxd|3b0<2=tS=Sd9(0=3k)mQbZI2ZboYVx0$p6
zx`pZrYb(GFrb!)#yQ_-{yRo2N1N()k8oylu@L-_x@+}Gtdc3R*)LMQ6?`LTo<r8=3
zSD~4oK#F^qWJH~T0Ty3RZKm=Id&=U_&D}}Bu<21sxmuL3_+}W9Y*W)@Pe%Cwdf^Fh
zExIbe3jo_3@`QD8gtkZ6HOOsW5VF>Fr33IXUhJj<cXvKM+tyOQ#er!$0#O=<)r8EG
z7KY6y(L(le$o9-6r@|N1)8>AVitsw+SfX4Lm)=-6e*Qk2FO_ApX@ixOo}L0uaEI-o
z!m%~XxDKf8UzP1q)&Hn$gUYK_I<j9Hbhqzkm9O)aY-U{{zD^M68rztFb(<sC{z#4>
z=CFt|g}u_}$u6q^eC191-@iEK=6P0)vRNTaGq;WZ=Ai5fpFA-Rq{^!RcE;tgCi%q}
zz;_0}Kb^0L=he^MKpP%bek(?`F)61q&i{xXNGnu>ybPGU#$}I)TuLG`0aU3*X@naM
zy8FP?hjZMv^}jr!h2yC?4awn7w|?vW7h<&^L$)?^t}f#%NzR&s>~3loc*(Et2<5ff
z?4s03FbcQY+Fax07r&{(1B05=O|@j60?^_{%|iLc7Ct|qat**LT7H;7?YmC-5*N-A
zzCL2#0o-ZHG|RJwJ6A8;INZ*=S$KBQp9)XL+ZnACx%(RpA8YuYdi&xZ;KG<|vt;l5
z&<^qrg<4%8Afo|1Gr%Q1VAtaTKvf1tyTJ|l0mwT(_HMWQ#1EADKcYJ7`dhnp<B%F{
zVmojE(H`0D+la=3|2g>WGI#OJad_kG5?;b>Sko)AG}fyknyILeczXB%KY!C|`9Tf+
zSZ<+^od}C{6QEk|gr^+Z7>%0<UoARLe?|a4lAy|x?^+t_DrG@sW1IALDqn1?OKf*T
z?0Q$rF+^@^7p=`x&)wbcAFc;F-e0d93tyN=wpzAOnW_g5!ZyY}ZYncHdOwTuvYD8W
z?<uCXU?|F=C%x_N3aTF8_+Odf?mlXi6h2TF$n*3d7-4^37N|p<MHYFIFysKD{I>&>
zmiVeu@BZPZh%j}XI_<UdKS#L)DI{a-aA<+O(SZ(0SQ=mN$WL!9>M$h{7t`|ygdh9V
z{Iu8o*aB4p@13hUlac%){O4Ex<+SjP=;gJxX|;#&RYk7-6P)*EX6*u-#3oD010C1?
zAW{tOZNO5FD1$L0Y`bONTLcNu!eWtdw`7&0q27qC#!xk3vlsbFRl<okDg-scch}i*
zEEST%nW7(=4P9RulT!!YlbjufV3oMt`bUGp6u3D{2yfC(Zv*luB`h=v`p?KQ{_N8Q
zp6{EId@c0Xq#jvVMl7de@S9>C7n|A>vBz_F#@+)O9JYX7eppKIF$IJvmtX~DK;cD*
zr^{}NClj^zgy}-~tgA{Kcl97U0u()8>oEvt9S{1YwD#rU%*`!Z!&rA*N~S-_Q|2q;
zQ1ND8ce5n*3-?E_m$aTJ6h$|9PgSY^a5eKh=^sE;Cdc1~G+Vg&l+TVpZ;1h>UtCrR
z{{i_x03S5W)C18XV%&Zx2Oe->GD`)bX{f0)L>gIF5sdUM<VZ{ITj5vQ9Io)I%61*Y
z>)ybVh_i*Tc=&IlN{M41E_8y1xV4sf8j<+#n@h!7-`)oUx_@0<&5#ENGUH6Qz8_;#
z5)>E8+-x@0*Va11OdU=`NJY>8C}5e(1I~)dp^x2en(}qMv^t~70nez~MFQztE)7n2
zg5=YAke@tdX0(zpa4>E0O_eJ`I6e*kj6`QJZk0p=IUO}RZzVx=E(TFn+dP-g^;seP
zd_oOzAcB56RkRDWGba5}iSYEAjWPl&pO^y)#=lIMqZ~!;z7WD|iNwNNu~*M_ITJQ>
zE!}7k6p;G2x%KrIr-vnlz}q+sgSW%2W;!ivkgyuAP)afH`P~Z-{_iU+MdvkhLR^UQ
zZc3@~;>N|six+t^Pf(e9^k>j;ledJC?AOC*hg#Mbe(Hb5U>y2U$@vn=Iu+q3HL3Af
znu!Xw!azMsL{w$#EH5yhB(J!@iQd>-RT!d!lo_3to7f^}egwIOA2F#cxLSV_jc`B*
zFY$2E$wdP6;9tg#I-zZ0RF+_;1Zsy|G6!yNYE=^(7GxCjxbxXTZxptQmt2Ky3=u6F
zL*%(LK%14tltZ_QvFI%N?K2{A$8{~A^~<VO76v{*+Xsn<IHi|YwA?cO2e=B_?siQP
z<2DnaWn+rr+{h5^00$33c)IwcNlX?NapOhELmN}!cPXE@jr)|{yO;hQ|CFs8O`VQK
zzT7G%YJVykQc<&Pk9H@Zi|1g9I+F2FM6{Axb{J7tmz{T60}mKx=M`ml9zD7hat-Dv
zFhYb^(C{iLwKiMbW!paPdxJp2%mr~#9CrSXXcaG-<xU);me(h1p7dsc9*~Sb3#%l%
z4tdzwS+n4wTQM_8Ha;k@fuQZbRByy%gU_9T7E<F|#uc|?JX_z5NqDN%UtvHk4_Qd9
zb?gZ6HCvG~F%GJbov7<0IR~=<tRVGyQf2WWEey^WQkj4|J_qU_cT_^=o>&V@3bO=9
zd2HWwJL&xXqz!D;KyV1cR}LH2R%<1p^1BE$SZUK#lY{tdDxM~A0>bs6gvKGLuz0VY
zf+;cGw>?R?5XbvAN-L+AT953B&D!E>a~Jk4#<o7r)5uF7-G{iKbn^5V{pGugzSO!5
zM;i>RTEIKHcGGnA@sQ@Tubs9|<aT+|*vN`H37$1Lw%R^Ez#yu}k->`nZs=K&)tXmJ
z?_`O@;tm#`j89N!Esw5*hAgfrEBgoRL@fi^bfT7rf@)b`-nLxYq4M;Kz2f$R&f7m~
z?ChJZn_n$Cq705L_R7zFRW|t7-cIm*tmO&kNvRz|-eF>G=D|MLKBCAe`(L^i9=O8L
zvapO$Tf?qWX8ShKqbhP*Tv2e|xmLm{(=#8#7veUe37rJxR4W8>WZ;z{w!7C9gNOa)
z62XQ<GXLa~s~fB!=#I64(AV59VCi%oqFxo?&JFFw)jpOK=K0!1kfVKb<dVNLRB&|5
zw*nPSg6i0D)u7sIx)+z!J0e2O@ekPy4JDHtJqL(+Rj2y@w+0o_-~4Q*34laEg*>0M
z^=0NHY_O={m0SsR{lzY<d$9f){Y6e;Q$sMX|0_VujjY~W+&NcpFmWbRWkpyycw;U}
z7OBcDHedx)R&9=Db|h|tfimtIj_0L#maO@nU7p<bwv99%v=+Kna_mUGOCz<G@3m(+
z7zb~mXr6~M7QAe*ag!xzRqJ3u$Hz4Pe@>a@Nej~&v!s#JqyoznghYr_Ic=r~zV7y^
zTtn2+tn*1KazVHOnj#QBO(T#wtiQ9(5vu2wbfPq{R>fo|80u-~p}0M(YE>=rB%9XT
z{zkxGh!xbeamYIY@81n(m(8jmaYQyyG~f#x?q<<p1;pR^wSoR+syxkjONz3x(L3X(
z7{FOiefxL|+SSA3=<L6cTPjqwNLXpjSDTycRcS81x+Z0>*sc<`c0K9q>IV)3^XId!
zeJZhi55_y@7JbNGG#^)f91UAdNKU5tT6eaSv@M2nU6ZN1Wx+D~wU61C<B=VR@eJzY
z`088A?wR!|o|>|>JJbji-7hJPY^z4*UX0c@2Um#{fcF6AI0CX8AVor-R0x*Yh(Ydq
zVsjp%^dxYcs%q+-tp|EWwB&4HYrr}J<Q~5~p3<z42yK03S$U5Le1CY`$hCjG-}BOK
zRA?Wv2QXM~Zm{s04V#Ai7u^!}`L-5!5tKYIo*uA#j*M}C*#BWzBFv@GxV)Fh>9N?i
z@#~wquf8pfl@5V-Z#-1_yV%-iR4BU;t(c3|Pi$?JU#WcB(^oH>h#4NbjARzI2rI}a
zu*BAv-8SKhnA7PEYj?#Vlr4(i?(i*C^zx5V!o{$S>-N>a)k1w2abS5$leLrP9b=cZ
zTA8RYm8|hpeowFNII=u=a!UK<s^H8{5-|f7R5JFGHN5^B)jDtga9H_JinP@xPYL*%
zPu^WhFvW}7poMHy<<IKBjmG>x`q%s$Wg(BYO!Q@hbHN_7ed%f|N8H><J}nu735_HK
zS3>K&SsoEN1FT|Ok;e7<TxR%8=BIE<-^KwHY!0%erl62v#_lIw)EbIS6&cPKi{eWW
z6FQAoiTJ(AP(8kMmwGHII({;`EBaan3uLMjn@r>!PSs(}7w-n0D&B)|B79p$34Tx%
z)Q8D$61GpT1r!Fs4#CCMMBU1}ZUWKUM3kDy9s>d~*6$o$H)$IGd#Bq^CYU+h<wzS)
z@8`luXOV~!dEtzDC-%wL<w5qjhs<t&=A=Io<1M#Py}a^^pfP8~&l&!6ZQ@Y~GL_Md
zp|plmD9^(eKZJjKUAZP^%4D(`6P@#)Y?_<9z)TLRdw$a-98a2m4Y0GQq<-d2w2E(S
z4GJcdS9T)y3e4t!aiV+!4l+nqBZh+_SI<GJ0kx`z`XXDI-aCfZYDDJR#oFCHDh-;i
zS^{pFvu2}01IPJ^6&&X;VnTFzpv(~epOi(x&ToVTxqS(ynCXV5nk-aaeVm0ffpEZh
zn+CDO>%{-4BS<8ofmdPV+D%9*JZX%bFr)$x8*^o3<0Zdm0dq~NX~}Hu=e){?SLYNg
z3X{(mORnCB&jnQ=8?*yAlB4aj;rr4{AjtsScM;?g_-dl#iXpS?EJg6zZf#=I(x3}O
za{fL@PvVD|--k*U`S^STT&^2_(m&PG1JyQMX>X%<hs%Pqkg^;NiI|w6qVSL^^N5Jz
zaOMP&_YnWJ?19jv#ccG=Q8?pQ{<w9MIM79W2qr)ulseWln4XQapuoQ3{J!P)cY3lZ
zPo9~pZnYNWO}k*TeWmhIB=RXot(xZw#@i@l8%m(}6!7DY*Cyk-n}1UeY+@&^*-_Bo
z1DZkoz3*LJi<7Qy97*M>NEvwRFf|SdfVw#1pLz0!YIj@dUL1_uIAY-&-tG^L)BS$%
zegww!nFHYG4~-2E#{HKs6d5&$Jwb~R!Tz=Xt;eB-c7Cs?1Prfludy;v6TR<<Wb85?
z@q4uXe0C7~hqc`L#VBlfmN`@26CTNOJ2fAFSSx6u99SM{SuI=f4vN6LI2I#JYv6sQ
z+W@6@|1``=+`)T7w@jxCkfrXv^m`WhRz@obfX04Gi5LFr{0(HwJzfBbKp)JX*0A1Q
z$`a$@$q^J5li&qW{hQK>coQWnvrrhV_(r-SjcoTTRl`3$s_xQTgYLp8Zfm5twN$dY
zLt{EY3vL6uiHT%gKo(|5nTI4Yp|U+fwIE1ZnX&!5J>_|S;w<Tr=%M2p)EhS}W4i^u
zQUvmq;!kg^kN;j>UbxvJm2suhQ|3`V^|wzv2fu_GJiasn=nw#zAtjuc$Vg@;vQ`Xg
z;nW;^nt|G+<g7%og!wZWj=bWl0K!Y^<(y(s0!#y?bMQW{jB=`{49!;sGvUGVu9m)9
z@!#I{4Da`kNPGkaWB13}^uH#e@SLLhX9SW=TfL>WoD><95cEwqC<|xYb?}G_@Eeic
zbbDO~N(_b!CEUK@<%#8&HyAr9%TQ^2NHLflmichgjG99DdVu`$%%Ek-OtCHyo#^{$
zT3W{hfkV~ciZSpK#34u`Uhyl-`yflj(S+R6HolCwG67o}dXpX06cS0T(*^GA0GO=G
zEd-XQflt3RN(NLQL14Yt6YAKkow~8G0b`qIbwae+@7#iO3#*7w9BfN-ZVo?>oK3P?
z`HSZJkLfYh1boe&pJZ}0@YH)Mbfhd_*D~o`EW~jL2L9r7_9F+CL&-)V?V5x~SE(N;
zaak&Swr`GS14|K=?|V0Np54+(i@FfcJMVf;8`;bhGsVyO4R{?T3B7_M(iCDx<=7(3
zyNyz*c4Ms60%`jf4&~))3AaGVguyV@;pg#Gwc4ju2ktm!kF9I849CpC>q1xv7`re3
zV(jLhRa*JiZZnSXN5fHIdg=GJltV%VFOq?$0K52d1*4Qbttz6XhRuE$e8<UbgtCPB
zDyLF)g<(szxyVZ8B#-MBXXGmF>7jg13uoa8S^iM?&bX{coCZD-ZXdORaefGEUAK#K
zfX+`%Zy0=N)gTV=kjYmOGE^?M(;xcj00^3%$p4{cVK}z!LN3q|Q3Zzjd%<2B^pd%0
zzA+@FbD@PNvDi}&^Z4`>NkprJ$z+BSAv)41YUJhLRUTkJ>;xH*_M`z!fIbmV0R}J9
zGlq_MGJ)*nN7wTpX8Nu*Bqj0-Xb>XIbRH3*25GV@z|p|L*w#y(HaUlct?{dHiHoFM
z_C7-cm7$Z>*eE>gF15(v;-kR{)f;YBHC4y<Q{}^hKxJ{{&2|VZb-~o?{<B_b1X;4m
zw6Xl^se91x>wk;D^9!ATi?tVb(T)5a46Qn1K5Z!h2VhN*mi#xU<8<pAfdqh5s&-k1
zVhnJ}Hx<r)`lmy*FA?%pRX647T`-_|&JV0v7U@VJ)K#kLZ)7*MR=b1;EzqK7r?7FK
zW%Q$nu!_fb{xtG$Y|L)I_O)O(e080r@pp2FZTrSu_CGxzx1(n(#>}(q-gmqFs7jv!
znm0Yj*aT1QZw*1NXaq@R7v+!=d<n$`zk<4KT1WK#tdOmRyWLV(IO`+G%?a4+^e)i8
zBFgw2D-Xs0CdE?2IXnyyQ++<dOi$t1VFmXCkxge98Ev#Fdz1^Li-q8;{?+E8<o8*Q
zbcE&<@R50%17hGgE>txgI6{ru@#GbQaLVynpoX<fe>VZDXXy8OOJk~CXXgVanoZrI
z0Xy0!^s`(~O;83X#bc!#^ScKoQ>Q4jPsIkx^nM<o1_d>0P{_sFFMk}`^Q)z?it0Z`
zXQ^uSiT>$!Sh57FGag@*^+-RPMC3=^WPAsb%OEmL#U3Nb0gr<)%2b6u>Z75voaN=^
z4LYKw)@w9hXANXLywGfzFnGd5^K4Mw2x!bY6yFya3~e+9*^MNjfyH$Av_}lTo)Ah&
z|89PIgISD^`gkbRKOueG(r%+V)gKvwVAGCnRunJ=*b6PdK2f9LnqyWgsqWN~FLwI9
zEcKBZhxY~3!uPtPk<Es?vbh<_c=J{N+;*_q5PMWC0@}&2{y{Hsj0dVnAbd<2y!xso
z%bXn0Ca33}e23}-w1q?rL405S#$n<|^AoVv>hqhn-GQefL@W%Fsbl`8A@)75(84gR
zL7>J?3#Gt&Gl{Ph!TclaNZ5L~+srN%%!Sy#j8MqRLr&6v9#eK@Bz#V1d1|Toj?nRP
zl<ueZx+k=0Jmf#DP~XC72c~$dJ;gSiJwa@P6Zo1iE6>Zrv`-Od*6Y;p$3qp1>O2$n
z>AL+$g!1j-D0bXZHZa+Qxm2S&%=PWejM|GM@Cw_^&5}W>6Z8&Xm80R6kmktO3!YX=
z!@j4XmHBI&cgoZtBjP_4dPpET{v=BgIP5X?uqG9|Jx27TepyGz=a0;pN0CsWd;>y-
zhBH9naOcz3QdE{FmsvIl;YSx`6D`#&Op+C95p76@)Lpf&_9ponnwf^te5`Xfc(C{+
zZ`i+vqBCHZ8gd-vLtV+3x=yVy%#r7@F@tfv;Ftyk?d{fETuA1|P%#(ZkyhLSH;h^_
zv8?bH&BgX;pAZ-SmuAkvtRM`~(p>PHDw_YIrG?*kGNegksO|VLBfTgH|3*1q#-RO&
z{M#^9a4(3*kPG5e?)0UHGiNHT$|?H+Hdu3IPXb`K5Hu_|Cdb!;mp``7#46ntfCOdq
zJ_^wZh>Two2a{+U)DRSXh*^~Z$SffJN|nxTPtaf(*xb{-R3WoS4}TkX699~j43G<Z
zPUE$gq^+o|{!nCfe*1k4xtUOVeJ5fBN<ntAt`B18jS%lCUqs~Wk8GD9u!NS#2Jy*K
zkGsmQ<i|g`lyGU6{65Ct6bTJUpx6Ex!mYVZA@~LVw#ns}D;=|`GPV*cZo$V7(ZQ{e
zmh@J4#!YU?C1JMbz4?|9e|BwIBP@WNh+kj*LbCTmcr*;GZR}{1-kGb);`knyy6o(?
zwapt};Iv60(rq_w;ICM3QpxTj(E<lCoaVclBaGAS%z!QH3?nCI!^y!usGIuNZ6R_X
zRQag;=SCVc6Z1J!_#kFvaQp%O{*S&Sr{|x(WGMJ9%qumyxmZO3+We;_$-HO4mHw%6
z()!E>PbOR^Ma3Cz0HPQY1k^=+TD9Aw+i0WPoX}h&HPzLF^6i+qv!1oJwWzD5r4nA=
z-lgPSiO=K0iM!7ksG^2CkZ{6KXCnM{mHd{)vjG>qWBRe9t!ER$mkRBs29x)j;O1Q@
zn&2w+?@cVt@*+8Fn75}+`FS}#$~s$~Gs-JGPC~XJbk1xX+e$5zFj?v1ZMz1s1)%Ot
z`qwU0&}bE8U=TK2_*8++)}TOYn7l`E;|*A45f-9+ozs|0osaue>B`fA<X{N+ih>CA
zOgtqLH8L`C!|-g%?Ncut^7gPtZz(G;zv$;z$#(a{DEtu|ZCBTACr+HWTk$>J$JZD6
zqu+sbXUo3SG&S9AYEt_C<A-_=cGMQh<%)}Y5{+IBx$86vQum^zQRb#&XM&TQqiv4X
zmG<R_ujHZ5(mjRzx0PC$_*0848q}JB%JR|GNxRG0kC|*20(iiqv*^w!*o&>t-*(p!
zd<#p}-VM-d!%V97Zu>b&;0TZP+@s=i5z-_gO|j4w`u{P8)9$RYRoa7--%7?}7ji~}
zPf!XCRjoV_u#>Yn2tzqWc{$uceJwJs_15Iqi#{Le0aJ;;{B3XiInvo$rwVVMX85+>
zzkh#ZCS5Z<(hdpC*T?6AzkifPxm_vIsTO3MKYaMm*4w-D@|7zVD!#tH<?WTU^9u_{
z;Z4nvNji-|3m#}mo7XF|v%6MTS4$LYjI0H{xx8mSms7rSP;=~oU{NjMfvy&Igmkhi
zamzHVaDij%jCV!zxkBWuP;HfNMSHByJHwIs$u6d9<+#mSA6PZo-R-bVV~0g-viCq~
z4fGPnrU+?N7lHR7dGQs@?zEj2oE;dCNB-3mC1BOb;GQGzvd3LjUc%lX$#zDU7g?an
zg=rn6nk-n6{H}5gRQHvVQ7~55iy~@kBfU7rDQaBLEZ(*#U^kq4v8koS#mi{AhQ0B%
zH8tk66+OOfPA8I2q@<-C7#$t`I9Os4utFw@7L?xbT18o39)s7kxVQ-X4>5_6;_j>L
z#pPvX7eYhtrbyX;Zj0s!nETvD$!rRbPLNO<kg2#*D#s(uCv)XX?`UeBCx>T>oOFJ$
zZ@oyuqi;+)g5E>AS9F}2?(L)d5P!{x#i8{oD%p^Dcjmwr&g;#`nFBtaw`je3;IY3E
zYFLJ@ivA);cEX_=#vj=?`VGzaJSTZ$r$fkX*5f>4eEc#KkNKR`zpU=rxHIs(g%Z8D
zj-TH`ac(PDhtZPikiTh(8GL>RUtbpZdWGgC+F{LFWr8f&@l~FU@#XpXOyByuQMY&T
z#^6^4IjFH)0!<6z7OVw44WRda48^~X4z=$5OHkj18O2O<5A`VbM%tTSD0(U?L$zHb
zUYoa3**M2tVZj%H*>X}n8a<)lG$qH1JDQL<s7WB4c#@GN9)Y2eVeFc3{*k}|@tnLK
zfgyVk9Cbv5g<RW!NK$SQ<c1;l98Jwoq%?Y+?iU}Noth#fa>=P1+9jlim`Kg{zB?};
zn4plBjO9t14j5K=-o1Flt|oSl^yT5oDx5oVmR3tj#Vxr5H+U8Mu20vW(`qv|7?8z*
zOfg5n(4arJBj@vrT$agOhWyPE@!liak~>=sZ$IZ#>vmzit*)#;lp^4ga0gvFJg<m8
zQm~36I%^q8?qL*8=@#C@c<xlLTU|MyX+s%P>tToc`H3mude3>I&{Xst9E*vLxQiG}
z#K3@wX~|$kwTPsOu5#__vOQ&hvh8<s<&-%)AMLI)tW<_&Wc)o;h3@xr0-`pz@+NrC
z<lxgUACZ(~IV>qULb!naMCU-QUQh@|LR8&+vHLH*5a~d_aJuNPX6U0YAT;A7Ak^(9
z`0(1)_AZ5|zlc^A#1X|b^s2=f95#xpM)p|`x3PZ4gGY&aGWHY6T7mkoj?(z<`3MjY
z{U+0!-Y8qH9br=q^avwY%TDSM2t%%=`ufem-KAOvy(jg<tUcqj2(Hy{LzIv#zgx5T
z@vW!W{zsb6j`QArUgI5y=GDO*L7_zoNOna$^fYIO%+&C2_7E&yxY!P{s=v1UaEV+o
z%7W59tS=<n9g17XwU~0jOX&@?C*B`eJ<XOd(2r+$!)JEWkpVu%b9}DkiP(of4JwCU
zJk*GrB@%Y0XJmXP66s6LneOKwxwvv|&b4JtY%LSmonw~{Fzr6W`dmT|(NYCfSi!Y7
z7g%-Lu{rn{$#%30YtrhiffR`twj{LK@V4Cu+wUq;H7Qz#O%L*TKE0Q9_U%<sC-8$7
zo;k#dW$h*;=7>j-_3@<)U1qw7khH6PlJBHf>?$AZ&hEzfz3tsoBd@gLNFXrvz_1|n
zN9s41)bSImM8!@*_sbr>smt6L=?l@o2>(TVlyUCMUK~$)x?-!^@egY}M|nP}PjlM4
zQ-6u;s$LmF_S4H#L!_MSip<!q3T<&fX`|TpS)8`j(p?PYNp^9R;}TtYyA;ZKgQgpA
z8gwx`tC>Of_A|aEZ8pCk?;_8Unqslf<z8asOBI%;+X)!k`kJ$u%6dY_qSGasq8#M}
zMRe-%3|;ObR`loV#8-uz?%sNb_jL}BZWMQ_DVv7V>Q*}<r>OX1Zmwc~aBwi3ZC#t@
z$&g8G-0VcJYnsnFUTQjnl}J_^2G;Hu2iw(M=qd`u@feng2+VKtu8AD{T}#vC+%NIS
zDOL2kljei0Nw&X2?LqCSHoAimRkiZuV~;m;BA*VzcM|+Z<1-%FR+_V0#H5&{Ro+j1
zS&$x5ZrL-!rfaZOSV*e&H+e5vLb&`~dk>Z@F63(L`f!Ti8ILX4(`mhEe<cWG&gY&S
z9PXpKl0hI;mN#XLqvzjR-ArFgT0GL+uh1U7$O<vQNNjc&Yy`zOB+xr*(drA)jSE(v
zG0J^mKjhy*v~s{{it`;tp+&nD2V@;bYS<#YN7@|@Uv<BKflFgMe8c=Ul2gU4!^|Bu
z*k-Sh8}9c-Q!Ux<dykkKSC0;9X<7Yd|77W*y<WAPuJ4e1$%R+qe1VIz?NHKF3BF3r
zF87P_N>_}}ZoO(;l>4^Y`@jSJZ7s`fQd#c%yGYU=)%)N7I2IcldjaCZ#KgJpGc)I|
zT;cd}`9QU`p3+<A)s^34!*8{_UOnEi?GbXK)!wQrywSa7T<{t<Jy3g_7<0hK<mgV7
zjl&AzGjz>t{E_k77{^ul;FnI{rr+ndPpTfs6Sl}Q##L{1qOscERrJdlARU@lF9L$_
znqBIsEqd2?lE##Yzo1|ajQT6)D)eLLg@g>sag)D@3!Q|@QWzyRp4ICjM0*{L&PtC;
zSJaWJd>H^-5JlHsE153BDrtX)xlO>Za(nk%*0JI6`jc1SU-zJN;%_mLrsf%gq;o>G
zEUmjhyyC^VA%DmFg3PV(if2Q4Q0)Q^THNe3@BTzzniyxB1RT6xzN$A%q4wL`jWq-u
zr>s063K+*d6gEBznad9qvmms+laGZXx~*bWA~b&si6HAbzSP0?(B<wW?|pg3+u<il
zjfRA=dXTv%O$bf8WL$aHqGZsp)Z9!ZB_#zuLD;qMjXsLJtlp_6yAC9_EYF=&4E?>w
zw`qYrLpk&glzwkKj=~RG8T(^mI=CRs2hm*y%a9IPqkT_ojP)}!7n7H&R-WO0!0#%e
zstDzlFpk5ttdv@WKjLM_>D<cu%;+Rv7ds8~el&z-k?5z266$r2+%Mk2C?3S8hpZOk
zu*Y>4%fp0)nB=|OBYF_h)lhs!NOz=S0dSP;ygqp+V<+mmqjM`QI=>vL7Zje9QwsFQ
zkU!Y?d<*3DlwZkx|J>re2J+&g;l&qLb9_Yx#GwKbFK8!>f7sHi9u{QQ+5ecK#XZo1
zXalc)q(*!1Wm}FLj#)#~BB_=FcH>K*>dO+Zs7u1{Cc6}kwH}Uizdw|SLuXqhCcqnS
zigc7qek88uZ(X0PoO|O};!i0Fqm*t^(d~{z28qO)?*+nl-h{;;B`y9sHudX#k;D`M
z-aZMl58ghZyZ=;?sqvED!2!Z0lKXnTNmd>YYqdIa`OtL0+VYVMrQjR2V?j)PnJOKX
zu6?gOs^2K*J!#~=ckSk+%I}AIwH3Q~KZo<IsDy1s^vkFD7HI-dXHo2^16Q{@+>4>P
zYUgnY^$UAM)7w>GK0P;i+f$YBnd<~1vYpI=5(K2alkOB9BPR#W^LEzgX+9AH0U?8Y
z;p14NQ`&}aWpVMj!UiTmFyRUkvqI|{*m28yx-yQW1K1Rpx;-!%n^NZ2eLZs_u1$LX
zrMD57WJ*RdHiqLtZAcKgp5%;rm4%KV|I!Lfz`7P!=B(Ep#gKO}tyj16H@fkB5-UB|
z=*9ENsIK|sapX48CT>knP8v`XzgpCf&;5Ca%gZV`(G$%LXCEy>XE>A>b$-bI2-nUe
zhwt@eY*)RJXu}Y3GEX*|Q{r~})FslV3x%g$ryvnLKg^u;D!q&j1#db@drO|NyLgk;
z&xf6@iEHWX_k;#wceEO=y528T6uH`M?LK8?oaB4T!|i_l>2pvY@04VZ7Il<gx9nPU
zcenc1+2~%>Gw89m+S>d3<cq`1YbD_e<+D;7Z7nCW4NpFN{S+x+E?>Uk+L!TZxYAWl
zNeLQ)*!Mi{j`xYzgR&#-eYrmstV<pH&^7Az?#4Mwcv1%51FEOKvvvl0e{K|z^xREn
zQWj@SH`RTUC!7Ik4au_()WmxzY;};wfT7EJ$Ho1a-4$0y-+q2WwAJOnm7|3{_jlbQ
z5!swn1HF2H8yM*7dQ>LWKz_weRX)(4_~?O%S7mz0m8cM=uIYOh=X1Yb*1K%%neTt>
zen#Uy`M|pSX$A(p+5$b8nZD$VN2&v#2h1QxEMkUY2J?i<C$A71{gc;IgI?b;NwgO%
zSfIr+K}D6#JG0}Z*j4S%k2@>)y6(Ib(`w?QTWyXLzJq&mf@|mNnI*G`id&t@F3<4d
zN%`V_5R0@rjP^X6S2XK5%#0QZ&e#S8pcQ+_xW$rAP)v^I?YjH<@TW<S#Yf~~=~@q)
zCSvwaY316oV$^f?7Fh)87I2TFSq;kg3#3oK+gm3uaYpC!e4}oej0jhO$-ROP2CPXg
z4OLIXwR*6V>Q|Fj4^6$4t~D<~T2}avoPxr0tdPNlOP3<AU#F8z)2p_t56AD@kP2TI
z|4dm_p|C*G_TfW0?(umxCyvCf_`8YvrpW@%yD#GLxgtxfLLyg8CWAQ*%ns&$(c<R0
zs)5W94zg3rOPbk{@(oMn70c7&LKY~~BLD~{{vz6*O~Kg^ru6aNN~vNIki>`jS9tUh
zZL~%;`M_%8<GejMY^4+T6k+6$kgHj61Bp0no|CR9?;A{V7Qm3Jei3_+no>ztUOq!+
z{Ly_oV`+RkBDylQwDok@1NAlB3nwkP%Us)PyjyF$QHI7Poi7s|#8Ymz@&+|!uwAnq
zigM)qy0pr<^Ja1RVdhWul3qx$t*yTh%9HOdO0W`eXHCf~%XepJ-Q~#1cPFRf#YeU5
zEP=0=6e>FVG3(x%Dj2&n3i6h3Q$imLPjv`}ihAY^y>xm|6)(Oj7Rn<eRbY~={#n1^
zC<o&fQ_;PU0sKWL2Nu-L;o1}qdK3;#yH#7i8yRVR_Kard&YiD(YF~RDGb$f4Dm6C}
z-t+T`h4bFE38@W@wP!7B6ffwX7;=OwTU@F6`{xmNqv0(MLak4q6ppq%$sN_PJt#18
zpKz8nNeQKlU<QG5XLonO-yM}=Q?=3%XDcwq8Er8faGH8@?X<c6UGr}tyNl>u^iH9R
zOlYfU&)Bcie&i((u9R!52KqTu%v-;VHFl<DrB`0ge=Lu*Qv3U{zNwEzU8MUk7{O}S
zVh#DgiaNhuI}aY`c}{efj(FkdyMf}&ALlduJs<4-*{7!@^|F2jny8AHQK+EavNn%~
zR6X^8Oa1c%-&{U6Y#<S@Sv*CwNw}tNd`xfm4M&H=oP~R!;xWTlXm%wQdyOnw(&BT4
zS+>&s3Q}Uec`C_)i=M`7#5_ek2%-{!P&Qp`(s9w$yH3UZ;&F)&`2zt1OM&e2R+`Zr
zm@o`QOJZ0k8j}{8Zs%&{{NUGm!g}ULpk|y?1r)Fb`TPagf^L%Zoi;wvB*t$G^=wHc
zgQENIZiW9@e^xRj+kL;+NPAn`mK&|Sb}eCd&sW|yFNs=T4@F8VxHQ$)cSX8Azj!u4
zHg|L+u67_K6SdCL+;Pf~pnep?$$JW!2OT|f%PmF@hOO};JCL2N)Fxr{obmRqEhi6P
zu{5$(7kRDNuvd;~8?qbK(PWjm7ov}L-JVqxL}0oFbG-Jwj79|vbO%3uk&7`RL;@4<
z`d)S@q<oJCUn70%NyVS=%M1J>k(trxBfn467>oTyEKSDxzB?o}MJS2KX1=x7)(nHB
zZdF}v>4)SFF1~}o83p~3>TmRzYU<1Q8EBzd?Z%g(4_#@ny_bLc5JS83d3ddNDg^l9
z)xWDLeOoEcbp6>$w1|!r{zfieiZ(jwrC5{Jj$!klhN<Q|I3<zY(O1PrBt$p#^Gyl}
z#)8f6J!Wt8hgc#U#0l~bo}?$=Q`g;kSWq<j(OJ=?!(2O!icAgRM>6HLn=Gzy)azJ9
zW#x5v!yeUlf|`_*)8R8Sx1vyTu`-u~>(1H@Zye%3Xt6(_r1@rM+5^2>>Erjey1hQe
zHAw|bhvh!g=wh)2cRC<7)5Yj?Vdtsy|ASkp!6!o+vNIz}n^1@<%<sn@;Do5dX<+0-
z>*i*io)vR|o)j20+KBXIGk}X)=<Kx7z8DUEkuwqG(0ITLx}wsVN0&dAf7Wm^Qcs&A
zVCR!lpnX#r>CoKX^?JcV#wv6#hP{syx?U;c$^1~tpN34zMkC<0Li<CE_sSQzg<eWi
z^rKe0T|&oidXF5|Gs$CZC@Z<~Cf*s^e|GtIhneAsqeWJm%li}HY$sx%)*<63<YpCF
z963cVTG6|(rYZHAZ0~covOYa@>3L;m%hc(wR*Qqgm0zQmK8Rc9X*T*0n{@>xr|Zk;
zXMRm!!$t+#>Mg}Zt*jE0AmTxX06)cpqSmGCmoHz2SX0r^5Pi9!R1!WXQkx2<Xy+ZV
zy8q~A`OE|Fg4{>3AICDoQ>ZCeEcfaNnncJYR^@&_7nDPO=YphO426nAFU4EJ`@NMp
zse|g`BoOq#rDb4>fWMn7WN{=B8;$zqte8dZ3eInlKb>lk?mo!Xd_1u=+1OCO_l|$A
zkni>Mbh95nlCiEfw=ra5<BudP4L~|>yRpZFgzPh`735bIC62FTNB!EMXs=qL(cyXi
zF&@`$`oT-p)i91i&mDW<AV$vb%u9?^<Yac>+^RcZikZ70XDW?<1AS7Bx7f3HOQCSE
z?aFV4wUisjaCcmDzprhatehjffA=+iTOMrYfJhHti~d;crNr_#yHW$lm-hX{nl|nf
z<QXBnaAEBppsf@*#}Tb=nH=R*QJ5Gnmc+RAIt>I1g$KR&9yhkM#6gYE#l=<lW(NNJ
z?V;%j&)|Y(Y+rc&)G>Co0dgQ$EK<q{6nwqqLeknNua=&yC?Z~f#SILd>k2{5viz&@
zQiH}Tm~ng%8n57X-<IF!v<#BzNqBah_QZ_Q7n+7XLg$VjeUG{<yW<2QeSenL9lU_>
z^m{J5pO?G;MD9`84(=R3z>gH4X^1~!Ij_=6^7ayA!Fi<b4bBhp)HG;zg?8&o@32^+
z`(*jvhlFskUAW4i*i;L{UXO639cY@duZo>*FxcQ&9_Lc-%R&}?Mj|!w^SRIO2z!hi
zgK~wo;0;5oA)3DXGBg0=Gh(WNa$Gs%GbPT>MIw9^)Ahx*hnXF%*dcswy?)Kw;@kck
z+5PthiaMq)*|H{GKXet4GsreS4xnU&gs=-TsZY|`ZynXU?qUUB<`M~07h_x$shyV-
zcI?32t@R%y;bp_Iy$~39`@;v@vvzh)fHVTMfS?Q;8_EDGBY9IYr6cT1r^hlhyZJVj
zoD%z1R5muj;sZWxNvC33Y|HdUlamt$bafoQ@C3jaHrPDFJM}hBYBF^ljWNdgvf)Nh
z*0*8URy*_!>PC!lc_DgxmyAEAJ1;OhUAO<ymmT!^LUbduUzi7=KG{)x5p)PG_z_S+
zZ;Ak6y<m@}^oV4ob(*e0FW_V4b}?+Y4iB&Lbj8*8D>!UP6dzy-vv~iQlpo7D$Qi9?
zaz9r+4zV%z+C~3ZTqkcopnrYyA6JzD@rpn);brmL9kCOT9kT|e;OD8V-g0x?AqG<g
z{LHE_d$ch7L@m33ywzJjTaFs<b8=j<+k5)Me>aY1UV1LGJ7z{|puYnB+NMAGsNSgx
z%hM29o@1)5)#0bfPNT`rqNr<fauUH8$HopyN=aP|2)G4BAfUNfcI6MvONyMosAXSu
zs(2E=X5jNhXTg@sZ2$7}BS@7Vv%&b;ZznbJ#kk$#j@06dQKmb#qPCjPVKYU49xW#1
z-ZezI0K-T80zyTI4}JlMT9tLqFW3qOM65V8AnQT~u~k{4nt4JgG`OlKKi?vq>p!6w
z2BTtc6Ef0O<X3|7ZVC#GpOsm49<yTf@(aSSW60m9%g^Y?qQ35qK>mrZY4o^!V1`0m
zmw%%h_a~QW1}STeWz4+M8E&|Kbu71DOs^^{_L*Q@6t2lVaQhy3P<P;T?TC8Icn5@w
z&lR1Wu@i@wPn|Z(O-#@mhy?8R7`v8LB90U^<o`)-XT%{xHur2pL#K<o=tVCMUreva
zi2CIsFQwIqJHsRBe#AD6{p?Uw*|fF;hojXQzye^}@%~3mBJRY4on?UG!!!iKMAe8l
z@@fY?vWLB530<|fH=M(oYp&d&%#?40Bhc8rT!fgjOZqsUr~xD4`1Y>jXc2=3z0)p+
zrtk(=w;}Lu-L|#G5}r;T`nJ>h8G^zFg2E-)?r!MX<B<KWrCTB_4J94VsLD%Wq#xhP
z0=7e+GbNvf<ji|hLhpj#O@JR=M~K9z=^jHD6Op2cEr5{OczxK62|7Qn4=SE1VZ&7h
z#9BQ_4_Qi{4zQCesfGi~6;1fTpKfNe?~yh*Of^pgj@^Tb>{_Du^+TN4SiOPM1q08h
zfS0z~PbjOIw_f1d);v@4%R1p+=(xrn^$WHft)F&%@Ua+-|IuI6$!7?4F<La*P;_BZ
z@AUZyY1aP!6R3qs36btLCnL02?T9y8q`U72-0y=*L6J>OO%<6oD}m@?@<P)<X%a4*
z#=2@pM;r38)G%$%IT+cCcJ#zu7p*bYXJQ3wQbR=YX=VY_dR2hXl9!mYA!*@Bx3ba&
zxKj>G3J)F0$Was&@q{XX*<stO`-DYpLwzDKl?_*G11Pl_c+Xyk5{lpThpuDV0G0<7
z@?SA`5r&5{ipD$=18451Z-dUh&MH@E67+?3F?9uzy9A{$WL<yrDiw>B##xvc2%uLu
zp%8c{H6(3$7q0p-e!R27-jy0}^<G^eekj9kOxH6&m9(#JXc}q8YrLTrYd!pGdI5f*
z{Zs5a#d+hgx6Sr+KHsgXSzm4u#WBCCR$1wvRnt2I((Qb{tHar<pW*|dLS;Jfcc{OA
z-o{<d8`qvWWNqwlyMF`XsCHh2M>XTsq7PQ{3m+==Y9Algg{ue`!iteKsUtD|Vr8zP
zh}gE>DIIqA5%=%-)h);o8+L-WVJLBr3&cb*D4MPu);9J*69}cR7p!H5%XVhn&pi^I
z72^4#bD{NjG|$%p2#tpNe^3u`p>bQ@@d&cQ`B6m?q+g(R(_7^OwXZ#Lq4*aFx$+A2
zHE(zvd{~t(4h^0-{L^AkSW516NM2~VCIMh5g)`MpPa!KzWfQB-+LFnqouDz`yU2M1
zy0O>^h=EWrA$j01r|{f6Sv@psbnbV#*m^yp(?|`1eB^rF&)3#Z&Yo7#&SPVYq6K(v
z7d^m)8IwiIzURo7wF%CHoqhM#V&2!3(ub{FIYc?8otLO>TL!Z$Q^Dam&(aRJJ=%w9
zeFsX#pMGs{p&qkNwOblld1v+~Bo*WdZz~!p$d}*=TXVd$F}5wF0tUHDsOEF<i%l1V
zxoi1t+dKG#3ruj+1UEjk&^cU2x`41}sdHLY)>*A@FGYlXZT`1zcLg$tMVDU?Ifw`i
zZx;+O%LGZpw%U4S4TBfXJND=3{vvuQ_g=_^@pVN^_^&nFkm=ddm|}$R)fbRakgLs_
zcumf1cqx=#_39ypk8(SiTTjBxB$3bnl{#`ly9-xS+Zx7F=$F>Xd5P!M$Ew7c*28$<
z=Z53c`9@E^id+4BdUO>G`{|#(Ut-~%KcrI7WM{B`0Ap*heQ^_$ga58CduS=1Gw9)6
z|0B3#XsLI0+-_||ulC4s#SI<gbtP{9EQUL5d%9V~J<!-V1DAB_MGl6#J6pp;y^K+7
zy-%{lF=X2*Lc>*tuKSwZe8NM<7>r0!@Who^pIQXhRFrG7!c&;A7<_zdvf#UP2xoyi
zrU<pVoqovMYhxcf5B>j%$?tiK&)_^=3c_bJ))RDs$lr*>hT5}j^UqHNyDn?^wnRVF
zmbv~74phyKgiw;K#EO3FQYd-{%gs|y%1UxS0{{@1!4{)#{PS|d*M*?N%ydZ0gLFWa
z$$6(;`{sAi(>y;MneTopN3--e9I&jAf%RQh);~Ll>)Q5$0=|)s3<I)8;_TVebG$uO
zxy&yp8}Rqbrrl)s4{elc*UP77mU|cFG9OyWr!L*@!qe|}TGG|A&Dg@4IN*fY?@#4Z
zdZ{37i*xt#0}GW~53tMfjkl&E@Y;qMiRq)-KAzCF%%xi!^GgWZuhM;1hO%#5+k86|
zAR4{zQ=hNv!3Zv7HBzygpfglQde-;5!)f0R1y9w$x*)$hk7O+FyN;2k2!S92tYz>v
z@%ZRG+(T1W^;H=FfV+E;Yxg|*?U9YeXKDLPg&<t^+~WMcs(e+BRbJ^d`_2e>Ay*W*
zp%sUpcQrdkY5;$ShCC5JU4K|_H=Lh*_u@S+{|;cu+IH-(pBM^fnTm=G8%Iu7MTT%-
zrNQ+l@yaDz4JdcKZSo?ZV`0N_cNkIHBMv=%{hAf}6u7KNqLs0V@%{e5r+y29IPJm*
zzOcWwYn6%bPyw2lmWcCB#0}(N0^sgdQXJvH30P2T3<*9~924P*8DO$iS&7)!TV>^2
z=o~bnYfI<~KL_0*-zyKc9kV?<nfF(~TC&jJN|xN558+3S4r`UBz8!W8BA+J`Ngh7R
zO7cwdfr|?k^PGgm{qYqU={+*--}<pb^9@9oRk?)3IHa0`p2Qf^qOy#^{bH|4c%~Rg
ziw%0lu@JXJIxQ?LMD^sAbO$15R;f`)<l68>E`QNPGGxr@dXp_qtg-{!JLfgKBE9$l
z2)xL(FmEr;4(EwgomCtRuQ?xkA28a(%RmbSgzWwufG5Goc@JZh-gs0gX*w9DYMVm^
zuZ&Bt?yqI<clz!4zUpZR_J#PVnmEXUhUbt`nwL<BD<q<tX!i)r-rb4#{5C%6_5rL}
z1fJbDg&hJtS|+aI!xZpn<{3%E!T(pE27R92nFw-DQu2(;PZk#O1sVVOWUIA?6HqJU
z_{X%=1xNUDHz%p8BfZ0E&Ks{}y#=kTi08D_0g?mgCh}ZDEPVAuOzlV;)V2+yGfT)w
z!8SS2kux^$s7-RzGY3@1MIT+ce4OsV{p9(BH?n^p=cn8tj=#Jh7dc%oN;)e*fznhv
z5C19n@p_2&%%03k%Hx*~Ajyr_xiYsUM?k;-UhJ(-ECz$CSF+X6zup?B!h8=Y2S%uM
z)G%Av<RHsweMvrj#WE)W;o~r><pX30o0OHFD`ddf<)X{_?{a`Y*wNQd+h>fPlwK_p
z_1B2S>vG>^0wMKHM$titNMCy8gVm*ftX1dt-R2j?2l`_N{a%*(n?F(ESg(U}G(lTC
z?6t!4u(}!tH1cj@^WC*C0mO?IGtd!)cLUexfxZTs|Nmp`OTeLS*Z)UUO0pzN3)#0p
zNi-r$lqI{$Sj$o&$wV1hN|G(hj3v8dNyKXuk)<LGr6MGiVushAA$$Dq?^o|R@9+0N
z=lsrf&DHguj;r!`p8L7)&-SF|Z2}?@&kMs=L375?i@jT(q`Tk&uyhNo=)DCate!Dj
zGvx}8u7RUONK~x1GX@}j@i;@qQ-99q(`xYs;;>2ntcv!liB3;XXTKVxG2i%AU`-vs
zxvn9b66#fHWoXBNKqjs&dg+;Jp9`6ue{T{(9|^v|htwglq2;zhp`5xMx$#ln!Gt3n
z?=WKOctbIITvicGMR(<;l42JFG;zsAoAV?~qG4Uu=SiN^x=5$lNc;z^DoqlmRAW>7
z$Ta`@4ftiHx4o-tGK=`TwsJpyz-f>5^6$GnT6M|pl$|7lBK9~X4=15Hi^70gHZa2{
zAqx6;fXCZ&Al~osXXuy!Q>ToIkw(-EDf@*vb}wAIF<NDCGLd5=M)T3hBlq7U48+8Q
z{OsJ^f~PEVZTYpTVP5oB*9&~dxwP?9gC03CUUHSQR$9L&P8JbEHeY#h^*a=nrW_U*
zP-zJBM#^H`QbJW7r=Etx0)*x@xQI)2U$QfmZ}uZ1&oK3YCB7taS9v_x5l$8zInb7w
z=3iqjPjmj(xG`wE26v>)r{g9WX%*El3oaHCqw+lJa9YG`bXu7;+bVr$-3>S~2ZoNr
zuHXolia@V!AU6%J@VJ6p#AoUBUQ-g|(m@><SUNis3{LQB;aJzPfQZiLgf)Og04s&J
zcCfe%OWw(WfBi*CfUJXV0;>Z&FsK|rng+Z);LS%0JGCoqzj{Bjc>}k$@|Sf;2hB<q
zx+UEmes|Y5sF6ncE#!O*=kDLV&GNuqs2LdK&L(+YJZ{$Y_@J(U=6-QW!(f6<q=}?B
zvO1Cn*q#outo1Z%<X;ZAP?5;mN{FfxchYD@Zk`osYP|;WWXU0qx-|cew1a+*m8Vqk
zYE4C`Fa$s1U%8-{D_(bg_nA7|RH6?o#2kmdRf44>E>mL55x`$woW34d7X!b3zr<R<
z7qG-Ly~@!1k|;UCaFZh-aS`l=c>i6Y-|t!$jjCo#X#=DmXgo(qx_(?MId&nILvXEg
z?Eg91{jsOVj6wiO2+SCpZ@v7uO9K1WxXh|CsCt@z)$fO2$0MYagv9!rJ)aSyyK*y(
zD!wQOkJ&)~Ogey`kJ~xi@w%qG`jy$P4@8W0ikYotba8^#SlP(CYDUKL;y*Wk`Y6QQ
z3zKV&OyALuyXGJmgCQw3gwjq~%YVQacJ{NPaO#&byERz7B6Nm!4c<IM6g1k&S8A=V
zzU9&)zVBAWt2e!tl`1)6jG(!#U2)vx!u5i)Fk35%Y&inWN8em0QL95qTx!m*ru6;z
zmCiq~Y|tqC*g^jyZ<DybLkVG>C=?ewYw)*}EG$WhGcs0-(==4f`PKQ94R5FS@^85m
z7#Mx@{tbX)p=kKkY&Gh5cV5X9y@nP%2~4%MSF3zTgq4^w)KBr5t%1l{>Z_<vR5ZdF
z*r|*wH+ujv7dAb281?nwZclb`Y3m)1Gug%)3+-K~B8(H7g-3Fwn#?K624;ztprp|?
zE#8wD&t|=R`?|8VTjkND6oi^9O<`UF=|=f?BQo&Tlf|)~TEyM^FJ{D_!{NxPJB>{2
zZ>Fi`Y9>-v38bm)=|rID5UAzrmtF;UVZ^{VBVhbhP?A!F%kur9$$p#Lo^Kv=!_V<{
zcuoQI`=OId20Xks%-~0Ne8OAxX>~Qg08mk(+T<eC8x&FuI%6bM?_%lhlT5#|TiE$$
znD^MhtpG(Kbb$4vP_(S>=GnX+be#x|mXdeRiq?B>8m80z%Y8vpz+-M$J6Mvcj&t`y
zATw4@Sk&jRoU}==eY(c-eEpxxH5^AXwUNzq$TuDYw1mdV>P0P8txA?GHF5RI8&=xL
z#09sPwP~?%LUz$BpKbcYUc}z;FEkboB|6ZgBv629F05Rbp<!Ws=<bsTF`IokY`W`L
zXWM?P^iK9aXisxJuy(LyYGoHd8vz?cjAmc5v6T|<eF4u4cpWKf<4q>jHQ2(->xO+v
zzL*O@QbfUQDf5CtAj4SKJuN<zm~ou7M~!iYrMLY3=ucCb1O(A{d-6v`#cdcmQTz@@
z&#K`o6(j^yuDVqjX1)qtCH&>G$7=Pf!iEWy3xL=+Gc{Ty-3;$w#gKGRXqjYz=!5j4
z0byT{(KU(`cfoAe+-8Oc;_z`??B^8zH_l;Mww;dOEBU+aCzdu#OOlm(gj({`u)rE_
zQkDev<ZD`5^`#ntht6h61coD<XIa{=x9OK>&6iFeh<QH_lPR4xT6R(rjwRjBXQl9*
z!>dV+0Nk`2{OcwpMObg236qN8)a8OM^`0sJwr#iT?-?Aet*Uz2*|{B{Ul^ni6o1>z
z7Z?yaj3fWAIY1t8;<Pv)Ji%yo`v65QHaz12rGqn6cvq8FNw<ac2nb1_6vqgvYZrE!
zzhu}6`R?h!JyiN}Hh^!nZR}!ILh`r!DELFUGDCYDUuC@Fo9LOcAunGtUz~p45rdv8
zB5DDoh4c}BGP9Sa*}hZa*(WdA4Pblfr>C)M$)GeN?1tahciYbN=#S$uqe<jR`-F{T
zm5qN*3Cj7B8dx0ho&fNxCF&M6onq5p58yZm48_Z3dM0fZY1L1&(x7Ll2DvqzmNX99
zgWaGpV^y$edcf;z1-WGthsyHAj+wEbxENf**&qjl(Gi$ysi3J>d<JT~0`S8E(1$<N
z3HV`!jarvH#$CI)v>CT2m(~u~v>F_p$JTNHo1YE}5&{zXM?g@A$HfB<eTRB?=;Myo
zEAI+NOzEtZl3^V8j0F7Lo_V*xNl0w%7;(cqMmjl+PCRJu#2kKnQG@%7Nd&0=&0!ii
ztCvr+Y>#0j^DW4|+eGE0yPq7?#k0`c6_LJHe@Dw-pbnk(OqOP2YTp!<D(O<;K6w0`
z-GdR4^~%W^)&B~rcvjA*MdYIRHd@*FF_!BL+uY<}VFo<MGQOb&W;0S?;QK>IM-1=|
zcsw`BsACs7KImtm!#%)mK__EY**975wF|Zf3+Rk~bIuHPW9{;E5Wh_8LNg3o3}9HO
z;3r6`<ac~SpctC(c)LpI@c4Zh-50&WPqWE;@}a;#0xKt-zl|(vdusjT@aY%Oyd52y
z40z5r3pHgMdWa6c+on~?&mpI8c9b3X0%ZfeugUy`0MVZEJf;&;T6kH<P5D=UA>#1=
zK)e4Fg-G!<PRrQ!=2z8FZ8O#3;$d@rDuAp2aGkY=`J?%lPRHYYIA7~*$`H}u;24b3
zb4&1nC8_@oF^GAtjt6)3GZ@@_e0*+8Yqxll*9U7j7%6$u_&8T8(AC*e!2<_J7PQ*i
zQ4khra+$ba#bQ?#`$wO_m~%#6NiMT#6Uw&=vg(40;KtMpZO;K3Ly0Ir^;hsob|6EA
z!J)6&UAS8}P32Gmby`*|>nrhw%w0v82X5gxQE^#NMnq&(NFHPsmq$_W*_JA)0LS!g
z{)p`nD*SsTu|ML+klRM-QICgooO7M^>w+0;RRnvv|3TRKVjem+!)dLOM50n*w1+ia
zkMyw6eQ>fE>`9zPOXeB}Wt@LAGLkJJ;?wvQsG6r!Bw)Jp=!NMH5Ekld07&7JF?1(4
ze{to{n&VtN$;AfEzqD5DP#x>{gV6>?6<9WaUSO)W!fJQ_EXC2zesvrB{)cT-r8`nX
z-Zic;BK|70?%L)Fw^#=pDVb5OB!-dbI!>UY?=>m`!bjccKLiOG%FTTXw2eC)ad`81
zLtyX@G!O{P_dn3|am0<97KrSRNlxF?{<vvdDxkrSoD=qfZVZvleC|Vdy$X|-vb1XN
zbQ{6tq9API`>ZnF_@qLW{ASg^cxiTm;+|l}7>F$2?wj7+f**hD_fOm^Sr!x<#HYR)
z8+*&)G{9O65rOn99)*pP-8(L$zp%W#3x?;XRaNE)%m5=!Czlug;xqOZ*ee<+P;k9?
zG%TC7c2XAKL&5Sy>cU{0@W5V*NBN5)o@vMRVAZTg(nh!l6Vmy^D_c<~6du$QL`8L6
zdNy|5v@?l8(;T^yhg>;ZQK!jHfDn8%J3y$?`bkQ+GSo`#qGNovnsOuE*%I?8WU53r
zInu=$T;p?Wzl+KDl@_Yz*Y)(JJ(3P3@oW&se(7JJ2XD#(mS0{<7^b$ig39#T+$8C5
zmvVs!?3g7(9^I>XWWNl!i~$0+Tr6d6%yW&_ZvnX{s0|NlXq*afNpun3azuQm!Me+;
z>B*doKVmtA!m<GTG%zd2vVmN#s<==~aDz1?$ith7OGzh;RKM9W5j`&_br^3tWIh<R
zlU>5DV}ly<6(T>XCFVpOG{S?}jd|req~7gB@0SZ(<l^G*na4x%h=8;5T+dX{(6p<6
zbelfDP$a{~VC0Rsbcu#~GT5%~rBz!KKALZCq2f!UXJ{U=<g$k~UkHo~vS!9G>--`x
zsxOPd(*-rJe|^7^8<;m?j9o3s2-se?IzxZG9sXUPJrO20q2nb1ID%apYwS6`T;zkF
zxq;dmjetT+4%Al4IAEAi8Ufz@zO-UBJ|k^G+C;pY(?JHMEigE5Me|&<PZr*@l?PEn
z!h_=zgF{C*4i#4pD=X^$s>u_FKmS!bBX8&0trC=@OP49mmohPE5zaGg86M4sD<!I}
zZ|S;1AHWAUQ80=9$JN6NWB_d25hR?2u8*sfI}&NK4UZ>_(RcKl@Z^cp58+{!nR}@A
zxeiBJZM|?gXp}y#KKS231iY!Do>$&779|UZ?%IqNT?#si$>_W~6_ZgHS@JmMTronl
zocRU-xJs82JBT$RL=rNv(o`Vq=KCWRE<ta0u_E&Or7$R^(7^SUpXdiXNfSj>Ou^u?
znHmziFL=*&m<_iaQTOzIW(UpcJ)re$<t<Dz$fv%%RsQnQ;QbyU@ApTu2_vTrmAs<s
ze9)nIQ2WSOX>LcXt}$=+hhaMs<>BS|soTPcvZ!Yl3Q_L=6SLWxloe&Qw^rZu;u$eX
z$zV#CL3x4r{ge;Vj-fozri!uu?m5;0Rl28wz>q`a*9vy$*q}MBNU3>F_&IpIC%v*=
z4Qb)OD&qt&GXnNZk93V|&)mcYrZ|+xOaZjU_egZPabg?Kfxr{WAdnSn-$bntCX~A1
zp;7===3|#|;nT49yJ7Dii$C3$8A){w3k#K`*k@nl0s2OeC-&c_j9YtYDplOtd7^KP
z7|TxljaDgyXJ3v!I+?majB2PJ@vj}`?FwCdJ{&I;!@A>xL)CjdeA{ZNHaY{|5!o1H
z$Y0aaPP*qr(`xo`73gUi&!(6lLN{d;O~UUF8X{}FfBc?p_;HTyk+cZgrjQ3kuksH~
z`B-neKSKimh07~~@d$faFa3X6Q&cSnmO;!WONW=@FZ%yDH6>d5Ap6Y>jY>pIMO{kM
zpT*;n4x<lBHrs0{5FvTdU2gfNN+2_Ycev*vU8|tH2m)5mECm2K0v&zv{w+tsTS^@2
z8W$FD^78VQF9A6S86wKuz;`I=n1$?Ny$b>MmD|Bf5h{G4S8yleUU+)?ZiAM~jE9%&
z$nocN4OcP~?~1S6Yw+4-B@~s~Bu7Ml4B!;T1w3?`p%_3B@!gh`AKpSPcn$x6E*1n@
z!tej|G|jdV8Vj0pLRQfoQoPOeM7@^tVn<l$n~!d4sd)6vw(l|@m>kYgR!1Y(P<<<h
zmJ72#EAYk{L0!ttRD9>xi#TF=f|R65TW=r^r+&<^cI4{Vo%4IBZ7HihCcJO=6#_}(
zxj-KTbu6@|q?Brh;y1Pk!}d$p`3%NDFd7t-H;PBIyGOq>80NX+t!t2ep6kms;>gbK
zUhg>EFYpCtKy|jLZ7q=lyldsAh0|*^QSm5{aiFwBiFmli?_D}4!J-em_m9cf6M4PO
za>F#2xCGG_fJ9(-biCWk5R!0l;fe-PTTEOIwt{)UoU%dl5KMU(&F4DE!YH7J=FajL
zla-w7URssG@MXhPBais(YBGuJq8C9pP!P9LMRqT8_)R|!K#QfhvW7N#7@GuI5g6M0
zFDA#nndp&+G5FK|P8;Y=)XAXhzTDPkm1kGo^%IxUPGb5Mdg|bG^#pYUch+8u824*|
z5)hRHXUOlb9^Mz})qH@S7&vIKZYDtJ4F5XJuBICxheW602Bgl)%OmBbwbc-|)ZM&-
z9KaUN@Nd{s{~Wf_8UAh-m+Ov-!nBLfTcHX`04M>owj~E<$nb2Aj~}`&-BNa6>*i5B
z8ht{O_ii|REA;)nEBZ$hXPh~%r*6?Wpktqrn6nE3ne>oD)U~ls7m~LR?9(;PIBseT
zPxnDmqpmRsU~CY-vDh28m6wiwikhIw?My29ZxMl?;|HRI1nfe=e+0wMr-x?PfnUEN
z0VOi5LVY7L00TM#NGo(7qg5k|Kw?WsXJ%#&L@LxiUmN%Uw;v(`^H@7sKdlDM3JgBP
zuX&S#!p@OgTDUDoV3vV3GLNS-a{@62D(d?5iyW{Dw<!R+!{2f|-eRbOM6fnU6r+2m
z5Ei|6bZacYsOKz>UNZ-NT-8|E#;$2?UMzM*TuL&SAS-B1*c=zXlV86D#{<aX(01rF
z#Ba;x!pYe_3R4<_EOLV+_7vKi4-P_sjRjp=&~5@HF*&VTPazeaMEkb$+Bx6V@vB~z
zu5C)G=<#Te0408S4mn2`M*iE-j0V<mz`Y?usOvk`4GO?C_`ad#f@T9@3HPw{(>ka=
z$tlNK@hrp%U@o*cq3r~R8<1u?V8k+$Nw}M%wy{JN41TvBiOb2}=KABCLpypZ=90Pl
zlMF>f^6AdF^pq4e7Yqi72*hG?GnEC&m=-|uhE5(AK@<$9;>4t4@+D)aR9CIa#MhAN
zdtLhC7NCkGmY92f8_ot1>?4&-<||e?l7JmC|A(((O^3S5?@CR3e78uel{VhL)|ET-
zF*yF}HOi1)BdFc=J|=5a)!Pn6HRv`#>APTMPJ+7v=6d*mV9BpW5Pk*uCqS+F9cJ}j
zX$7MZoJLy7oD?aB8h5BL$|a8+;Q~G;LgddXV=dy=(Avl#b>Es{<#Ivx9h-M>Qv1^5
zmR*jzqPw3X!?o>}%D#5}mSG-xD$S)b1AY_{sPssdHW|nOfau=5wI3y~!TB-9gYUWg
z^($^+IkruDrt<6F?fbr3{8vTzKLiQ%x_r9WfbP(m8WeB%OK3-yF5#p_iMt4PmHuoE
zL45m6qplq>dF-<XjaPO+$iue!jUa`}<obTqT7&`GB90CvXvJZy2a62U1R3dtwTaF5
z*Wv*XvLHZv6>^!3Rzcehg;L1~a97`qtLI8!>RfNeb4R=!IK2!uAG<zhcbiv*h~?|M
zy|7U1E#*ZU`At@B-%8BUEA^3XZH@J3Hhl};efV<yAc205@Rsi|!-gJ*5RKT*`l|xK
zgzAD}&P5CeD#NBXuslXNv@aiGb?6)V$K654*~QdCb>8*wMtTa_djF%c_cOEu%~Xrc
z&X4Mf5B=uylKKq7sn#R@O5E|s4xIdHc75Sok-R$Y>qGoneMkk!P$?2QF`9s>4pZs9
z09ycSrQ_U60bCvy22rXBZrK3fVZ&&5i5DQJYqjzKsK|MAr4kJODi;yWA+b+DPYfai
zuxh~v1PP!yhW`+!<K}ag7V>PKRy6lL+;LscFyI--Ex0V8$x%~k$k5r;Z9%4Ig|ape
zPlF2R7zSxkXHaIsZEg3Q);3KvVHg3p8rCc#6<PFIJ1zUOUTBU~5;<i(c7!kgZXM>o
zn(T=c=APh`9cCvBMx-r+bmibaoWAi_nCnyYfum{>DR@DAsNqqA?cz4WrAuv&k_>l4
zKx5Eb0wlVIS2U48yLuhg<s^cXfp!`YMEJmQ0!R_WH6_AH^uuN+B|?gB0&RzieWR@(
z<sB`s{N(L!Qy-I@Q_ZX7^AP%Uf7V-Tg;S7~l(S2CRzbqWS?-D5qF-Ec2Z6A71(iZ|
zqSZrj5II1M{-<G9rp0bv+yp>@4)JfIJ_Bw_g$p!&4J3ki!Kn`<9=a}76$Tc@S?bES
z{cB>V|EX*z{hnErK^h4%D7v7iQLcphK*^cB*~cU2p7or4r2QLw=L)R(3Bn!(4m*FH
zld+d-Fi}%*1?4y|+&NLir~CfN5~W#HWIce%pd1Lj_fU>OK$JW?f!u@n2C6FzwA!G`
z03L@sp`)M$ki=hs4&t(NO96namS0hQg$t;I`(bxjP8x4KNYjzMa;vk~n~fuy>bW7i
z2*C+sHwK5iVp2hd7j4YpiP(6iZ$Ab&ZUX+!da?@#o9Ohsu^9NwVltw#c@0}!Fefw%
zMR{XV!*pp~4SKCqaxg(Z(&2&vR_dIbWM9DFd`ojG$V!c`b?ZZk8#P9SB3+dJELu+o
z&ABwC&DJ%8*qSd{Vq&}@%N61ZK}68@t4AtvYGwNBk4N<tgRT%<L^zTd{sL@ywsGNQ
z#TwFmBUKs0*4ms=b+_|eKYg?=1AG-oKEOUuqT;W1t4bbx)xdN9^m7Y);18m?ixpso
z{;(@q5?mbfwV6G?pf(VbkGppJ?1sA*=&b|;m*eL0ybHXaidZZTU(oJ40#RBRot^+$
zPUP~{+xnjK*~L!{m`ie6IGwcYF{k*vT{&&Hw+-D#CO(u7?x`2WLh%v?gsk`u2%a!*
z(ptonv@~PWuP*>hJc^ZSWgD-&m*!uWwKnZWEJ3IrkuQGo=5L1Ybp0?jHSX`&yMw<h
z@xF4P++#u9LZ1C#dTGQA4de{pe^vCoFw4ccJHcOQsI$0z>WaYoQ{6r+T9=G|Bvgqk
zjw~)`@@4QO<Zl8U92bi8yq4q1-&D`3#TljFEn=TP%DZ_zQ2zy^ItT;{T^U?gq4nRv
zE#aC1i@64#fEKp{hQq=_SzBA%tln|xXMHnNx*}(k<@O;XOE@SM4M6P#cHI&06Rt~h
zcQSYGPJ1?(*X#SE;YxI<p3_6LcM)Xa{&6KW2dhoF9rqwDk<=mGN_za7nGXaR>rjNT
zoX32ax?Qu7=>Z~fN{sIyfGW8=4rKS-`o#BV2Q{LNakvuDhRNYBNJFGgk<~Fj#{X*b
zTT9Z4LY#fH4DEuh4WqUmIBP3}X|9!1F<%=1n%`%FEZ3&toR%W)2wm^(!w2NJ&j~N&
zxj5|M4L8i4f6D-CPBD49qH`v|A$@}u=I!(cA2S|lSF46II1o~F?x^?z_QH5)IrnD0
z$@?>bq$<i=K==7WVTJ@)axedB0Gzr<|1lQa1DsaL2;2`kXIm+!@rm3CAmfDo0LVsn
z<2(li5r2S-5_2-i&<uc-wyZ}7KA@(B(c1Hm*75!mbD*`;G(WFrzh5g-m(Chnw`)3J
z3+R2}P+wF!epR|%x&=Ih@HWq5`aWojI03i4^lRHVY*=0tfpKoA&{!aoz_ai~#_{{X
z^o-9N{z%>7Vat~=igB|}dv4np{I=t`Nv(r(&=23sx9?Abr4wNljp3d3Y9JA9B#I1@
zUzeq&>4Tfx3_7KWw6dg68_U#>?!c%g#8+FH6b6>326O?+Z`50NL1+N<vUOh}@1Tq8
z;rf*x6V@K=FwK_hWQUWz&g@S$6Y)Ly>s{gkIus`!`v61>&6C}!&dtH6#{jkIb{nmV
z2F?dQj{5?mjr#bEI)L|pwbG`9wLpvDV<*nlIxC`M_z=Qij#k>9oD&XkkeKtZ21mLB
zwPpB#u4Uo@mP!pR&gZBGvO@{c@MV1-_C_S%0BxXdxGnXjbI^aI2DlGDDP6PzwLsX0
z((OWS@6loa*%Q7qe{uoE{~inuU@EbGbX1?X)!qkWA#Bc@hn-#aEF4nZG2F7CArH_B
zaHspp*&r9)y4S#vcb`C*gS_sxZfV4<qgVdpivdh+B+np*<c|KK4#lhluKYbFh9(zH
zw&+_X^3YRw-zLn^L<YTYBE0`!MxDDEY3nN!z$pbxC2?5QW)CPEJIB^17OI?UnRuNr
zm<)bT!YqNZ!7kE-F4bS{0#k3TtMPrS#-+D?FDkb7S?BD7q^5}G+{nB=k=3&69_8%G
zw}{Y{F^+Un`6*&v<buDn>l`-0${3s^#iL-Yhw}@~hQaIY(njTT2y!YDXXR|y<n@i1
zF=A1^-4!q9sBe{+K@S9+j6UnS?(t{IN+DNQ<5{Ccifz)@y5RuwS3aPvmY^!ACzoUA
zY;;^876677aAjG9x43>+ix3faIYwL>0JPXVC-anHE;q>W!Lf!ooq@mk_Wu4cP!@pD
zZRDmD?V`QU-JG;@5_Lp9Y?_r--r7xd?Pm4$ByXBGjK{<JEghr;I^9eMnffR<kKy4+
zY^wc5%>+OxY(0ImBlT&hl`5}3(cIYBu;q>OhbO)nVp2(aZwnpAEd+vV^PlFxhh8f*
z+Gqj+UZ+wyJYP5;XJz^RtRL7&=r9^}XZk#{GVUUulC~INpBOuLLTLTnpra?@O)r0^
z$t{z0zvQE3?hpYEjNQrAbD~Wl<vKRFWO7M{3EzO{BI{1bK`aN)W&<p5rHb6ws$w>n
z5mWNGZ*&Z|L=PZcVODtjmIib00?Qrbn>Y8E_J)0Q$Abj?#JXa=>e|U8@D*EHe{CB!
zGLO++S)Wy&B&e&G_)BI<m&LJwd~{a-p6S!X5o~w(3^v9{Roptq6>nIaG)nV>AKk&N
zFns`!SJHh1K6QP^>iTX&UVs*U!e((us|yh{Q}@H=0AMfpZ9&`d?AbHKNVsN&NXSu3
zwl~R}zUB0-FZqj6h1TGn^goJ(4?AyL4S3Vnhs<U@ZemvQDT?cCHLB$Wq{zXHj|Ty&
z_yL6Uyf%bxC87_sV)aC%_yGJjJG#E~e44RY;$wsqknG8e19J~&By~OJ@@}R~2dFm)
zn8SAa{{nMTL8S?exU~$9n3vq$o~8ni!(=AprftCl{TM5p|CO&W=n=um!01J0M@k4Y
zn<~0n2OF_bTnm(IqMNCTAg)I=F57@vTR8ZoB*!MP)#b!x^AR7!_5igy;UYHpQS&QI
zBh7Bup*_aO?iAr-)#~Fv$A}pzlf(>CHoew8C2zvX+Sb_cJ&9Y+@v+`}LXeK2abz9F
za!1)1`SEi(pWKZF3e)XN4h`qIC@y{BQP|-o5$V>28K(ogwsz{x9bDKi5nyI25keDK
z;55uN*Ft6f(cs|&Pt}@vv%pgM%ShLA9{RYhIM!W1n~<|_4Fl<o4*mE<YR?q~nJ|m^
z09%ve8zzh`HehxEAA`DO6Vc?g*HonJ(_TGF4D>pboP~8{Of+zs5sw*i$CHVq3@nL}
zKDlSV`a+<W1I>9)P%r_}eg1>dtvGI#pqPw^B-YR&p!LM(USMtH;TfD$j~4Iqm%UAK
zD6VZ=Q(OHQeIyDL<~{3gnLk5fM~~<2F==^K1}fa4Jd3OJ>JJ8nEd;z<<e02uF)JnX
zjMR4(xrxPEj??j1(v0^desD4B_=Zw8fFgNs4F)Rwg_04N_9R+bKEc(?WX<B#qaY%m
zpX)X(86g6CQ4kp9q)ix+BA|jqyvnXxrIhGKZP=Um2Q~>rl1nER6F(8~-9F-)Y$6JV
zBIg&c3=608%Ee@kNya5~%&B}yRa0g>HQL;w^6NFUb@ov6q5FGc&HmGG_nWhqv09a#
zNZY3PKI8Kiar)AYJZBgElfUZ^KoioD{<}_#2{Vgw5gFx>w}EIdj;taYjE%-{7rb8I
z_Xgmnh4<#c1XV5Ii+j1_ZNLWv)JL(HPeho*-oC|)`{=Z#S8iKp71afx5&S&TdaAi1
zt2$ah1!2L`^waNv3Qo&7!+einPY{mk`<}@5+SXU-^J5ng5%&V6eBwyoejM@Bi93!0
zlNv!A`&9ki?-j#%b%Yl}DB2^^WU^(hyPKCb*NHsMqg%#Vi4(giSWrDdXo$`iZ##Xa
zSv3wwzF69Z_+SB^0FR0jqo_z&kuoFMn{#cut8)4Tm?r_PI>T>Z&TD2faZ|5FggY|J
zR*J`xP!`tXdgkDkBR+K<-H};p($Z`qoTmy;oN$=6QT6gQXG!{{FlSf3?Tqna;v$LF
zS|-8r6G3$t?|h-BNhAUkL;w>(q)y9lfZUveAhQ9CwH~_MlkQJ;fTs|#6<)pS1qSk_
z7@>i0-!LF_xO0@b(=CC-*tLeN>xG`buhDWt>Y1lFiYDVlcQ_8Mdn&)%{rQs$LG9xv
z7L5(W2mn+mEGSfZQE6euj;qubE8XG)CWMY)MsZqq|ER8cvG`@#uv<eS{U)<tsu`ar
zO)QMtWPf`%|85u(ham0}G;ie1o4+gv^Cw}<78h^o{ybAf>spj0ktc1hHVWX8XNB(6
z?5|#(eRlAn(%?h{C*@1n=@jF4>BnrLMcLFSGHbwI0FPUTz0l&6DplEXWWKhf;bfc*
zh<vw2<reGNx!ySC$2WyH_t%XwYAq8{%kQ{bzgGuj%)>3iGyOAH(A=Gey`Z3>z7XLv
z{B%k6-c$2}DCmm<4Ey+cZi{djdO+!~Dz2I&KWU{itbRSV#nu=a^>jlNgRx5D5_1$J
zM}_NJ^6FbV)4eH_4~YTlMpee5u03Rp%ScRC?{UcM8K#21<Ry}{;O7@nR<?iF?%iiV
zUjgDjNab>>yGnht1=J>pw#_3pD*VlEb~CVuu9(vcF$nB%feCt2LCr{sw$@lN$U<>B
zjzb9!Ui3nE-3+x!&;DKBHqg8o=SzWuQCuAyT!z7n*NB2(Nq2VRW$C{qkf=yXD%g64
z%HTib{?BA6J@H_P)fhT0^%X671d>Q_vOxW;eCs(iwPfOeZRZhoHslhv<PO%gH0qNN
zk0sqr<u%aN+aBIU!QRay8f{G+&pE3apdz;ziljS1uZXmZ&#{}JdyL85!z`+gi^z6o
zOgsCPgf)n~Y5ApZSS9k^poARWYJSZ;tK7q;(`d4n`pZf_2~2}0R?TgoncdVtx(GQM
zpMomX6QTaR0V`+djC@Rm+8`y!r3N2(w%tAKLwe#px)+z%F9|KYUZ2DepV2H#qiUdi
z0-@X=;HZW0owW4y@53GW2XVL;kOuSH)xUNt{CGGG&cr>kVsf&~Q+-9&O6>{LLAw%>
zp@8Eq;KC!<L>k*}hpl(vxUR7;_``U&*dUCkEC}hA%z?-+6%Qd-0AYLSB|rr_Hnot_
zs&9@41Hj=`m{t98DTpV2dh@@N`RFpPB{>{y4KiufZ;WiSIjm0Hhr=7%cYl_bcx$QA
zggBQbK4wCY)PU}&l;F6irO!%_adWvy`|x9(Pk2rUkUAaboO{+faQh4Xuv^vwfcTet
zAd1Ui?8`@=<qzoUq#vJPN2XK|0>SSeFmcPL56#Si1?<D=I2YYm_FzrB#Q3vuNgeB+
z=v58#IL2Lye3h-rq3Myz`wTKlMyo1Xo{FeMKE|rI1|~Eq%6YT~q|wivHg`^JlKJ(N
z>Xbja_;Drkz!NLlS+2B>(dHHX5>e?7(|IEMbpe}w1b!zFoq>1}ETU?G%lzlgojZ8w
zP(7RlP^Cd&k6B@7<KZ0bi$mmzA&|=#-6-|cAbGn)g}98MG!Ati=>h1c(s*M#94Rqg
z!bHV0crPG5-!EJf`f&@egx9Qn<8Va?%+E>+<*s+pH!YUDpBzPn2)7g9!s|-`ZwlI<
z{vW9#8*mGu>mV5hNk3Xz9H3H5ITIY9Cwg-3Zs`R-0_aa+7I>w?QKxiO1u{-WL@ugs
z!pY_39owP5KjXMoqW2knVPnJlA_u@OPIKZ4M-1a2|8<gXHD}|e7L!2tT2<8so+w;C
zIr28(0@|0Y^LUf3%me7P)Ihz~w&9RL>eArBwvQ`uEG}{JlHz^dGE>+bAt8)YcpJr&
zZmC1pjjJ5~n)hqh@#5rHs;<fDKrnJc!wF>saf-sezI5r*b)?k)`6nwPXoFS=`e4IO
z$*a>&64Dxa9~|bbmsN-?a=7C<b|?s5sdH@1MW;ZRF_i=TX-eARpgXX)*VOoBTKe!c
z6W@9>{ly4y#T)?3Y<{j}#XjytNpZT2BHzDYK)!Ck;R-+X%1OR02Ofk<1n`AP)8y;p
zY8w-Kv4Hs%q@uU2SX8er?^Y{@iXO&E@sn8p9xwjZ+wBoJ>jNBeM<Bh+{6+`GLzVqw
z`S?D4Jj)p_D9AWDZ2pj8M2Q~gl^#|Oe;S@~Ea1|OGtlIn>YlDHJ%Ad>W!B@J#~z=?
z_CLxI%m<qMI#7;Os`-+<bE>z&RL*kE7L{Cn!||eRw<No567;I0Y>nWe*HLA!GnGOF
zZ~6K82M-@^gwz!ziw>&v_o;y<&?7Tpe%{*=c-QI5%A@x^Ftpa6r#dW~k^YDZTZ3Z=
z;bAS$z4eJHS!#0n#82N|@Iq~1y5rDs3~0y&*qE+ib?b6iv!bRzt)N9b#e8Mojt(98
zDYu3>u*7U?!RJd(Us?nVv_kg3z(B!HwZTJ|K=AMUE7?O}b|@(|=10X=fMwEatR9GJ
zH~VqkusmvQ&&@O<aZlgbseW(#t}z6~+gT_nDkl(WiVddZup5Gvxs{Znmh8ACip3Rg
z<9xc6(nv~>bPuz_ZfbV@Zfy+t2ETC8`^%onDo!~+EE(hAN>37(<0<K%&!4FsaVji~
z&4y&1!}mv3{{f?eRp_hl0DJYX*<Rn-F@djsqjKnf58P2fW>c_0{JeJTu+EW*!6#1N
z;MDteE^Gm+Y5+|c;-)fjiRi?MVbP#xS~TYR#^)POGwp>CsQ2lLsu({iKZbwZUh;ym
zd8;~bRz0;XKnMt;&q_FN?pvXW?k?3)la}h{WNR`K3noA<zC9Yu*xI<BzT;>xL8}Xx
zx!R?F8{#3y5!j`{+j{;To~@;|GxB2P<Kt^V9%?aZY5oE=en!*Z=rrk_&98ndFuNg9
zqDNKx$#Vr25$DucF=zcPy!t21gKv24gD^kwsfhDfYCXvf`^5R0{s!JnRw0gYsk+LE
zJ4E(G+Kw%)JmDFb@j&^@kLBtT8ipwZ>pv5CT3nBmvw>}7kHY1seR9|GZ~cHQ37X*z
z5d1Tam0C4`hdJeq1biTq^hcMC=l6x_A>YLr>979NM*hQw4oGZ}X<_GwYAT0F+`?XF
zlTEF#z~7tFL0Q7xvxiG0<oF51ZK`*83iC8f+-X~>ubL)SCNlQ(D?RfGM_Wr}NThs*
zQA7GZ!XdQp3)V3O&X@Zrfxzw+48`D)G}r_AH?>e6iAkN>%N_6YTX%)9p-<}wim_L<
z77jyv5GtNfPWxMGleTe&Pk@{NqYm$I^8$#%UB?B^V!Jfn{SBv{2<9Zo4G)~24%35?
zMeC!|o6?4UfD+wu#UbOYb5=xlXXp*oYj7ImBHRS)>xshIWvp)?&~aL>)m*U*1k5Z0
z&?<08cIdxy{oYePo;-I`Y)}4N=7-k|{(&m<2!wT10H_h{CorD_HiaZ_K%Y?kv&+VJ
z`i%;b>6vem**`PG1NfDNyXnccP>m(J!}q1k70!UZ9&cIaBRYDu(l^%%07Ew{?!g67
z-e~qH9s1*<&o3YV=ot3EPU<Dn4tb1_<|V|8It<UFSfE-qX#iG^?dWHrn=6WGRYu2W
z6}=K+$ioXG9#Q#-!%4Rxn2~m@3K-_(u^_U!G4FqnatY#yX8VX;i`^Z!(b@gU1B5C<
zH5eKizYah!02^2~sJvWmboVbzQP4Ec7+=(o-k6$Kz#Won&&lExTnY-5VP8UR7^v7l
zxC1P*d3Z<DNc#ct_28)Ar;k&2m`KWz0yq(V2s&<+LG2~1ZIOzVC6o~9SFn#JM1I}l
zHR>=e>GpyPsJ<RNHf?P`#(MUrrKNp?v7@P}35kOn`MDVStZvo+;@nh?Nl=J}#wxIE
z!aK!uLc$8x?s^U`$9o@}^b~o`Ak(ZP9I)7{`{ZRfn5_6sEVrU2)x4LJiQ*vQRBF&~
zTHj#qB_OSjkE9Rmj%p{_6imC~DnsMaY>zJc+q_fy^KP$nURK(viO94qG!xS{{2`0K
zIaU8QlHZ3m_%rUD0>s3XNO%I_2)ZxgVh}hwARsBQ#wP@18kTI!<kZ+|;r`K_)4Ncz
z8B!Z6Z8u;NE?~L;aKWz`c;Q!YPl^lomOpBb(#0fJQW8CSABJuz5DssKj?4QjSA+{B
z8E~iNCce&$C(<cl&4j*=Y>Jqyf+%xq^$9U!bAT~82|H8r3i^`Fc<x!se24Wj$xwSS
zHMQ?|lqAd#YD5MLY)ioRK!_{FyPz7dJME;+e9%hcWPeQH%I2W?vbvAnhn$c6`Hs_-
zRJ6tP6mr~~L*w45PHDZ|c{D^v)i^X$RY)TLeZwwH7$N}3K@4~UKlHW;hlq(ux2$h?
zdj}~0ygQQ8k+Evhk}8+f)hoATJ#h63RvaWmz|KtQT-JRt;|!1U5(*U@{$?y^wEL86
zg4v_6hYfqmvn~Ehvh17vo(rfh^=(T2>e14uxy-too`SVc-i?IBzS1)+e|{mmD6AIT
z9pxSg1T3LcjDy*<;FZWC*g|NI`umI}#SClquzt{sd?z9UBb-b08U7ytehWwHVHqc2
zzT!T~qx&#_59{Dp2bEks1<~~uQt%B^F4AtQ02{T|BKb0Cn~-}6*^_Xpk*gkt0f^T?
z^QBbXo}?_=nRL}__>Ydxf121-J2xEjWB&l5V?36fQ^?`n_4uM)b$&U2pcLPoEXjN`
zw-5{Y{8!F#WJ%q@8MPeaa~kL2y63>$@f3m5`WmLP^NQ+~8?usfBn=%2JD*>4du(w(
z9K4m=HSry$T3B_yePyR02jcH6jPG3n<EKGoEcwPxq2&|Hdf12O_p1{LUl{}jcGMm7
z7?_$vqa}LAwU@sLoO>;MDhM#szJMhp{^N3qF)My+oR6fgU0kBhOo9g_J%ZPcb1lUS
zattU~aT#V%-Z?QSIbyA|AiMIh`&fPlQVHDhatHUY#_K$SueZwJ$)PWM<X9vCRyU_V
z`$Dm4u6+r4<g|y^0OiPmCWj>x)Ci3Fof(K+J#gSJ!E6i7=Y3Eznipi*$`@*N#qiyl
zU&+^K;5fWA3`ZR&tR2tVir53-hjXP}gu?cm>9lw~W-sO)^E;!Xl0LObQD+pRoHNKD
z7)&g{y`V>)8zHn<<Fl|ZS!G}$Y*VuJW4+$V#L+99o_(5Fm81jg^qj9z$gB$%H&4SB
zs`gf!l+yvyWni^|+~^hj|IDiM0M3Q)^F~$7`N(1F^yXTSu9UpB7F$>lQd^P<Tpf<Q
zby70K`DWNlkqwbg*)(%GnN0zF`U!akAi4sZ8kiiH4ugIZHA3d@jC_FPJjTJO<brn^
zZfTDclQI;7U8jj<i7XmhswHK13^$hTO}_qm;2Io4!Oh@FvXr@J{RmOTzCV8&B4J@K
z`mQYFU=8DQ{Qdo-t~|SgkT#`Z<^@W#H-snxhZmO>kebP=l2fMTC?$(DN6t4{_`_SS
zC?0R<_Yx8{!^ufsC66wJxeT6BM4}=4v^2n_k&h^u)O?ZaM%6L-DN9!ND-uHqUUDA*
zD#7%k#AR?!x;+$JPnSci1OGzT_P1;~Fu{SeoNe-^OFNm5%fHNbbl=|#<w_$Qn)bu?
z516%^--H#VCSLOFJt=lZUzB(FRD(k@ghhiX-4Q~F04+w<#HYlCTm5@L>D7768au0I
z0_FbD&QLF-cwU{m#`%YCpMg-Pp@jI3Fr+etd_LA5z6whR>-PS9R?h|G*G@O2KS##N
zRL|OQh~P%fHl#-*vk%aJkZ%b^`kv{>A2q@+-`?2@N19{Gq&Tp2_@rKR7#e%7xgS-6
zMX_Mi8VIG4Q%ZDwP*N19@g}fGMYae!n)M(O?AoV5+xsHhP*5p6%a+Bo>$;*N+}l=a
zUtjq8F(7@Keo`?7vm+%Ga$eFtq-i1O`>B)Dy>Y`-P<qf$I@D^{uBZFFeo8dYpwo_z
zTu@RcvJ)8b>+mfrYlHvk!F9ou#3<c%vJhkn{O*O`)K~CYr*jfi&hS4J8*p`PhHj`*
zkA;qUG#+qA>srI%E*=#D0dnvK)N-i^p~U7#26J@bJZHJ4plv3$cV{S5x6iq%%np_@
zm-g5W1}}Wmgg(91S}boQ+#v@K9z;C25LS#tx**F4A>8xxR(IB7Cx0>t6V+ev(EX=k
z>gI)}9!?7Yb=PFF<Zz*CH;Gw9kD_=2RmT)1Jbo~?C-Khhr}8rSw_GX%QyySnqE9$6
zfe5vkSwdVYNhvGX_XC%tNIpE<3CNj%T$IjWXx_c(xGdXsrkV_o;f{N-a)0;himKy+
zkeZ0Zb4$Dbj{da2ou<Z39$1xQd~$-s(83D7viXDc)iwXC947AuH_ySXpdcJL2niA4
zIwWPlTp^tqWK_@xNkSJ1zEzMi0EyJA?2kC3Af3dUh6Z#j{;DOVK{rWl25{HX8RI0~
z?D}4x<vbEPvD`vHPT&bJa)UDy8XjcDU@Ia#XLL-=RO3$1Z!m-3zJ2>+Ump@b41iP9
z;XlQMl(Tcv99}=Sc*y_(*1Xpc(9t@|tQKn(k!jiaDds{$yO*k2WvXhz`+7Y~K}Ru}
zUO480*RMlT08RJ^jO2_xTw3_ZTm6bz=CM1&$Z<CONu&<_mS<GuLn)yfvJgpcFoV5s
zw;mcXv@`Di4C%nl-l%SS_jum_u>=lRD<d>ajhtEeEzo<VUAT?$F3$TjWL}S@8Xrm1
zF^>fsA{v=sU|a5S4ciGpZcl2^V~wT`gSM9&Xo((hK|}Bfi%S6)%MF9mS9TqHB<kJ=
zQ5{YWd<w2_)Ebc|Hi|Yt*H~bQz!tFWs$tz^a0_rbOG-*2%o$wHuqolPX`wSdSc_oE
z%hP=h%ZI8*8rvBf&lcxWFW$azot5f7E_bjBy_c_Pg_vbS^l56?tARkb<9RYyEh@qv
zT)2UxEJ&f}cNa1u+C_>(&Q)QUI9-){fnQXB$XA+GeH%>2HXZL+;<Ie)LE{%3%;<VC
zL+gw~vNqo!CbgvGADnP_-QHSymY+x*5x2lCEROvp?ekAcD{>z;G*nRO*nq|b?=O9^
z^QRpGbU=Dewdux?*n=)s+_Ck0SumhWP0+(O<$%<85=mqNeJ}6=H+<3HUsy9>j7`La
zK&8}Mo{$4#o80y|`YhNTWD???u>~#YxR_9`rZBvOytf!jf9GRYa~4QZh#fnxoQm{0
zNbs^spZ_)lY=c1Go*sd%92`i3X<;Wqh!t2$iS+HiOIg0eWrlOiLlq@mN5T#o^&G(p
zm>60nD$yV3V8dIOqjX%7C-?9s1Yi)4LGuv=aHeTZrJEdXhp?tooQ92eoX1Y9=@J!u
zK0&ev$wOj2T@iL)1J|WBJM2g^(siH3Elg});EnIzH!!H*GoCV*({+6GcMW7;4<Y*y
z<*`uy^8gaZ*70fo>$?@J$aoDZ^3@-Wr%XT0pLiF7MMIZ%!hiy3Pl**rFb3%H>5L(e
zgxv2K2AxQ>xG3nfKxcn2`X;#cW`U8Ue)O?JbC<xL&Vjt$AOUnGYFB!s@RBSh6M1@<
zZrnqa$tgvyd%kI)REGv?LDLKp*q&IQaDp&9*oX7&Nwpy=UwxT>1D7yb|4~+Uzjhen
zQ{WyqGp`+5?on9X+_yS&`_H8|=yKdebmE)Qc}7@PQWMMbj9*-4XBZvcdvI-W)vrvo
z<qhxAOnV9IyOpGb28gL?u4U<WMMYbQLFj1klo&`Um0a{KPv)m6OGqH0jfVQ_yPY<`
z2K*0>Ks1Q=CxMd3^odk*aB#an=ol{Yw9Zfb)_>Wj$KJl~f4~pQ++wBS3YQl|T#CzV
z)Ph70%a>j}x@d|MBDq0jHsaC{D+BHn=z2hLW1ig6Ukax;o1l6N&z-hDd2(q5d<S@Z
z+gsV4(}e@j)5VT~<^gsf0!G4pf<yxX^+YCb6PL<_`bZ@shZ@rEEbFEhmgdj=NS6%S
z?a+JZ#oXavTr11V0RP@DV%oh!n;nB->rfp8*+oloLaeZfToR&s>Bi6G4A?8}?HhHh
z-q=7+Ajt+vWh3r|X+<{*;9gT+E-=UK+$IgC_AfNQxWt|O;($%aOG|$C-<x)u3Q_EJ
zKmY$I_7l=%2m;?<pl+cK0r+pgkDRu28nVS8d$ICKTf-gjQ@$*zi2jk8xv~SK3Aa@D
z(1iCi#RLPz<PS)BmAc%VvEflLc>$M%3YKKvAB40tWt9H5MLi85WI`sJ=h)!OKs`BZ
zRA7`290YR1W4}TqWXr(5qlQAa2SK(7<TyXikEB0Ch$5A`$=KNV!_OEYjYdEzpbF;M
zNbDT^=Y+S@*+?gh^Jq_iUJ622&^G9(&4M825Znqz4boI*cnsqiMwLGoTEkv(BJ4W2
zppDPXGO-|9UNtdv{q%A)#uJT^tlNYg9!8*uClGe4<V%67dEwUYsI~7ac@S4$8-O$~
zxuX|u078DUjL)jr-n%u|MFt)2XwPW0gGeH=hlM@(gR*^cASz5Fa#QkpuB+;AZzA&=
zM^*8WZl8ou5UQiGbI=5WfesCHf`gk7g%v5y*#nFwnpkgOGx5{!&nwRi+!DoTQW13%
zct5>4kzZJuofgQ1tbH=3c+@G0KDP*17>puC?Idw6<mP~-ItuO~_&}ZpXjo0>#7vg!
zD{7-l#v2P-H1^A@eA$ZMw?m2Nu~<axnw<6Iu`?|`Tt_`$<&>oE`rfdRy1}6t`t&XS
zDv5B{&bJ}70$s6@47z`D=?ZfK^v~cvH0RZ-WYxT%mgexgTj~GbZo{WjFC1RQG-O98
zcR;p=oPuQElR&IwYvu98)O(pH0(efZ2$aqA{j=2@w2zBjI&R%uGkRnz&^3@j?hAtk
zToR*Ipiu=0k-ilv4z*7~{0MSw{PcVJ^yzcRA2|U1421%THNI%{{{yi>SH>TDt**4d
zAERH9gXwR9<^VwM8Y+x?_5Hgr3eheMrurhV>^~n~0WRSm@}*@v6Q0R2jaQkggMCeN
z8oz`CfAp9j>&NWJDBt3;c1B~@9W`grBJb#bw<0BZstO&qRRAx#Aj^)9GcXL_nDl3u
zSsNFNCPv#}AS2b~e2Qva(&t28G-3iCQMj%wxUi3jgY20&7F2KcAAv4EoB#lhpHf)p
zs21_frzI6gPxw;N_sT4{^xead&*y^oXaby7(4f6|Hx6#rt+!!`-~$6kqKmWu5zX>T
zaD>1JkUn$A%OL?yYj<xc-Vgj}5O5f7AfnlNcaA?X%usC>Nhzgqpm7`UbOE4SaD#$`
z4dK&~$@h&)Uqeq%vVmD3bbSz0w^$=zBEXtVQ&=jQdbor($&%@Bf827qP$f=r?7#bR
zzMkYy2Qz0#>~>D3A52fFmoRrRTQJO+0L+&Ue`5X=A6lX=FkL^R_||yD7pC8DC8NfG
z+DIU(^2#|-+@FXdK?qWj>&GJoCykGS>z;>BP>i^)JT7|}UOrZ73Wy`$Sl-m~?_*};
z$$W_3XZ_p@#b5@A7?FJ0C#xF1vnQ6ncJQ%1GH4C}Q3#j~!1qi1Y6X#V^+?FN$kXCT
zXq4A{DQK?@noHI6XoFLUI4E-z?9*blCa=@I%Am3bZw4a8dOp__Qy(dn&YO^%No_!1
zfxVyrj`H8}u+Z57=8B}6BI5>dvv70~ln{Csq*IYk>a-;~MC!LV^R7JVLCgyCIny{8
zJL=!Qfh$%-Rz!Pwne*ZcvkS;QabY(XMnOOx!&Z7pt>H^HL7}oH#_B@(Uhafni0Zn_
zM-qB<3uF@h>|Q%fJ+?lIt;Gl0Y&KCGk>mN{Vzw4J_??8Oi6?cII|VJi-<x_<n$mav
zU%pcF?`)D8c;gKik3jW>+EyxP&>>~o{Jyo6{9{Go-e!@P(H^~ExR==ufkCiTu(;q0
zJ_`9X;H4Js0g+S`Rv+5BLNS-VxdEjzolrDz%3)JWnL!iyhL;)#kY^W~XYYYRd-4)&
z$P3C-aAXg_sNn^00fJM*v;lx1GRZ-1z1x#dPg2Zy%*-X&ratZap|8T5=<wIbCQjNX
z{CnQmp$+}rMtIgoOd|3w_WsSd2cPSujCu^*^s*HlssPqK;9rr*N$^~N5;zQD(gsv1
zQQic1$Xe>xp%2Et*E=fL69@?}i|WB@Vzs!xO>Jtxx%%bWHdCM9;aHIg6yrVd7#{b+
zc@lBN#Gm4jQ9njIH1ZdW_4&koGbzOX^$qcN)a($RTdyyeIOBxruH9hgEXZ|x>DOW8
zZGadA+K0!xz~2Jp(Q6m5+0gBYD?!8%WKJRh24K<$a!?0mUr@1wvja_-w8&;SyG`3Q
zm-ECSM{gV?fY5;dOws93f{1N|k~Yaad}07#0|XLOSJ#ER<B2u$DxOS2%N?Ih_>iWA
z4S`F4w$6{=S765ceRyNPs+XvHJ5-PPrPQ;59mqW{9Ht&;PB?6Ely%Qt{y;b1ZS^J|
ze0d|jqgC4g^4f%4xRpr|TqQ0cs=t=l<mE^38<TL76nNvCanficp3|_!>DY7PLH{!6
zRhy{)M!hFaqrLh-Tj(Hzu)`tqfZ=4{$3yj3`vUonB{XWwauYIP!60B09tXLlYtL)e
zJQn~f2YsEmbZTBW9BT+ZfHno*ya+k-facht1hG{x&23=@zc<{^;EjL)Ue?CCbklC1
z134PQK(|2vln}Vg(0KylW#G6X><NNPUz^?XHbcF>Bi|$;3yI(+RBpi_lSxu*_hE0{
zW|@UDhn4LssX?m>(z1QbBJg$vl5PrcX%|aat|Q;|J7y^i%zYiY{84b%e^>~lq~9=0
zq1=SY#G%h_Pe|-f#9l%uK`v$2FpU={HL=@Iakf%@jE*pu`hymG<2NoCKkNwshkEk3
z-gdY_`g~J#n1$)r|B>-@*^g0ABJ-tqwTAq&`>Qzsp|O_6_Gc#o;>iPrCx4km^_+Xt
zzKWuPlL#s(Myqw~Rff?QY2FZdBL(aGRVo(5&k1*tlyl_SBP7%lyuDByxk=FXK%h2=
z$mC&*qK3_IeDuCtavFkv5>};A@C9j!p;dYIr4bG;LYoYxdiIZvZTXE)!%YqO@R$dl
zV3@u`QtV+3nr*FniCo6$<8sa1i0cMw!{)HP2KrYnJ$vvunZJfMdq%OzVDOp#n{SAC
z^yMv1!lcy2DkkVN`PoIa0LMpyv3L`9hIPbUl{mWpzzBb?<SCLW;_8AYi<;c*<M#g-
zS~^)8Z71)u!icSyPmX?!Bk-;Pxa7aUIhGwCxOGz8IhFI?DdP?Y&%+BRhCr7>!NPIn
zhs5zXU9m)GmnJXptRZ0=h{f|I>gsy|_9QSap}tO;WPz}4>}*Iz?9Cw}cmksbUSoG#
z7$e`UATQ_(L0yaV$cU&4t`UT~1NZ>FK(=m0=8j`Ag3?+7&p%e|g@ac+O)D92za}~V
zHDBXpX6C9t!20XoWByDD-P<x@a~}!=C*xc4mNqu7eAy*xL7EANzWuC3IhzG!bNuCn
zCG?#}#^0~-rVt<5Y=dd1(uaNAfWG7+lB+q)SYw5|w3$F>PBb?NBRseMf7Vd78uN?}
za<?dGBh0&iH91grZP&ECWC^(63RQ+F0ALl>taiOmnu3-_%giZz2N68IkRbMb2}1GB
z!yDoDtSXjuecCtagw7c-RX}*n_x}EM5X%7PHDd^ZD8<Dgp%9F)Rj`Uk^fR5FOG*jo
zJLRrVx~JEYl81y;9B#dW)GDG?^XNL9LdhoBU_dRQeBIiZhO8nC9>}|85RMHm&m>R)
zH&Z66uvW%tEH20{e}B8;2vRliwf73O6^}MJDhMd|TRQ7oPDtL5G>c$tFV2^Sep%*_
z{v!-B&NbHr>$gCZMu@-TqX({~IZ<YyKai-9I4ZK&-F07Ou$A#iBx4+TF01zMVzqzq
zTo%|$F9~5GQ4Ef8QBBg!y01YiTUW^U>&l@Cys#f&fe#Qj41nIty_Wr#_(Nf7z{mw3
zU_AD2@mAfvJaVcI$&3Pn4*FW8j0-+z2r$2@VrXc3+ncYzA2I4}cgo=vB0-J4tq;yY
zw1$n;9|tZ8k@D^ItiFK&EO2!Ix{^#vn46nJ9?*a`yTEeZI+&IcG~d-d_2kRJ6AR}W
z>3O}vY^dQJd|y#w4|urhZkjdf*>05x6_b=@5-6b2oO2YM`qhvpDG<7SE*SdexMr$&
z1IY_udI-U+@Uk<#csf1<a-=p*T^3tUXMg$MgHv!D&*IBwS9d-$`2pv%-}TXxcP$|_
z_Jy;ukXPMi0-3uY1V)ql(9A-af_FgsYW0+RKspY1mEe&P5=7KNADuBIO$9M9#2X?_
zOxxWs|DBY>bp^<sK>WBCYYBiTxCr507UW36fMH-92>cN8<{IFn@FFS3RH&-oeG};_
zydxK(vHGSiW-@5;$nv+c%!^gKDVWj+Q95WzW5w^_O6bUjtbJE}k4b21m3-RAubEHp
zq-++Z+25Ob&LQmFi*zQgDWM{cl~04MloMxYyw|^km-1!0UE|kpH3BML@j9iEgKz;N
z!hdIm|F3wdzk5+tcs>rI32o21x=F&TsdHf+c@RC`m2svK`wQ{nAZc5WUH|Fp@aIBq
z(p3l`fZ*$3O=|7h$;3BJtHI5RWc{hbg(^J^z$h#f8u15qFmR`$gB;wckQ9PM=OAy1
zfPJ9b2LS~Aq<%FcJ|kA)GEgld1q^xDW@4fr-rfc13E|i>)QHGCf(Xe0SPDE=<YvG4
zb3;cwwfx8Xp`h@|fW;$Ezh4j<6I~F|35g9FMgB0cLpq9Am=0ntU?eci&oP=pntYm)
z0(O`h+T+Y9ElN%Dhf7Sztif44ZjW3e_pTpf?eSA$2mbdQDciPQN|(5?Gz{y+_pTk#
zD_rHetejUqn@4N-SU_s+_8(*k);N5Fvk`OlEdtrIB&cHpq%EpMxS)9G^p8iCr^K9l
z_44`PRRxT}4wzckVv1)5V8-Y!aTw_D*UpuWY({QX<Qzj%E!+U`HPDm5K}JD9_fd!v
zlirAg9@3P()wi-mX0w{{Kp+An4MCs^$_#;)3vN}UlwG-!C$Hf*!wT~UEcrKg4$7Tn
zzWln-lJSiGYar=ZrH$|b{`5M0t6cb@*K=H#;g1cu0L$PRz~iMZj7!fMJEOJ_-pv`y
zZ=}J`iSlzVbRv|_<p~4}C6NomZVgC4l*Tp8EV-k`(hnR~hxGJYtkbe>S;Xm2fR_~g
z%}ZWJh)U=+`V`NcXmLj9AV8F3AP)}-mxM)qUE%E^gK7|LM#WSD|4Hy7<5^zZ0Je-3
z_#njJ4Uwy8x+VhQWi>c~KS+@QWetf!ghPwMc8~@Q0h6G^ML_st@X#pQwVb%wbOX8a
zC{nO<fvUWv?f<6|>pCQ&Da9Y&f+Lg3r5|W(rT@dI4+O$NryN10*-w41Tgab(zwmad
zV+EG$)@!Z!9O>y2QDm{4bwc980Mz8|_ofN=KX%qPnK+|nhHv8_osqb_8?%DwHAfFY
z4m$@KDqOEIQz=B;%cyX+w&;^6iM(WSEqtz&`;_hGmYvuCd)x^LyLy}QKW13aLMySG
z@f3K2lvT;dqJlb(KETmGGO`&gzJLe7RsswipihO~a&|QD4S5Lyay5+e?~};7XA|S$
zy(=?p<86Nwp?87+G~_6OW{wx$MnZ4^nFI(c0R0v0tRFMO7RbYr02Lcp9#&HWyCFDf
zM}BTv??ULC|2E#&u+ei>;`Q@-7-B)`w(B24tR0o>*{zr5P7x11KrsOiUFIz25u4}r
z(oTx%c;lMzIC#s=118tPGthQ8yDxkJ9cD1$fB?Icw7@gWEYmpGb>;_RE@KkCW0>=9
zr8&sez!TrS;deyzX9tM$eU{}f^l!aby<L1bd@sD~hKOAExv#z*d4@v83A0m3fjb6<
z!qF<g?NA6noPDX5<Lxxy#3kMJ(5JP%eG6%LP}i`n+7NvUf!W}0I1E+@^pu|Zau#{>
zr&>ZdpZ+KHt1lK25C~6ayY7HhTYES{Jt3kRLMZ_rM#!_Z`5zHqt|^0IeF4qk(S7H$
zyQ<i8v6z#>yE2v(bPKsFW&+N2JOb(q6JmCE+G!a1;q-`zG-v~F*3&3uR++iO4y39j
z7+c5qA~n5o74nWY*nA=G&;DVLEhp)i$~&s5SB6`eUS$4s`1&#GLCha9$p85G-tk^#
z#%m)<sr@ZZ?9QvHYYwV^_cBr^M2szaEx2rZ3Ge@|4PYjdbv`pL{&VA3O+QG<#x6mI
zDR77AJA!x7-t1sBDc%7N>GbqvL{1Qa)u%>pGh`Q;hf5gH6+ABx<v-8qwxe@`_Tb+{
zKuEI-Tp659bZ!M%5X=YQ`TOAd>`PwkaY41r-T?o}4x#UF-v-qVpKN9i3FBl;t{u<Y
zVJ)QCF(7k{Fh@XMhAEBxtc#V3%=xT;RHLU$Z%`Fe4QR=?NM7ycE@aB0_`q)ae~f(z
zIF)PLwxpy95siiw7E;E{^H3C$5G7M6r9^{dDj}l8k|7JFqFFRZres`*q>_vol9aIw
znf~*6?7iQ;-}n8#?{_@MIu_d7;dbBGeO~8vUgtFjogM&cqW#RD_#DK>nd*hR)^3Q2
zN3$z3E5?Ov3^Tc`p`}EO<M`j=np(FyakpYhvh$CEo4f-=LwfR>si#2}B91$6M%*x*
zkP5g-+?!YrWC*4-(86)g(N;C<d4og&Aq=?!F|Gv$<gKH#6C}<khN8*{kPJhh7aMQw
zBg4>zRF{Kj!3IkLs_tJZK0-qUAA3T!0vrOrc0@J8wLQTIbGwaNmuPzw@lQ=IpB~un
zpV1><&K;LZRuq>@ZP;P;Hu^kjg?y;r#NMejIMPdP+`~9JXN+v`SSSo|YcY_69t@rD
zQ*c0q*@N}DWzkQ095@LpyQrZg9yDdl>)L91K9X;}^QHoqxTM8sygu@)>o23r|HYNt
zXV<(XXC5-LZ^b6KoI4y1`xu3j$!$5%qSx&{bp1VBgAjZPeVvq;kiF~j)eNIV=u%Ej
ze?nOd`2h@oNInnOKv|$Ozz9saac&!*?cP-<_w~g?M?unT>Xj5>wFI<Vh2V)>0XP>A
z6lm5P2{k#9)RUMOg)o9HVXHv;h29l{D@M0E-f+jUY3X0}Kl!Vl6!#^sjK%$6h{@--
z3H6Sk&^`(iG4TAN`;KgQG5O4EW6rqgbH%!qWSqWuGX)S-UA?CQ{^@F>QtMIEPl|c<
z@2@jf+h`x1(!}&c<P47kn-?!E@_CZ`L>lFwJ6+$K%g@gFw_J-Ig$f%nUQFihI9G6i
zpAF9En`U{ZSR<4Cr_&cYJ5}lv-AnI#3=#cyPvEw4A~5|&Vs)O3pqBtqvt}h4em0Vg
zjSGru-LX`)e|9`$RvBS)joTHDJU1|4tG3|;#`6)D@WlKr*pGm8mNFXtV51914a6nJ
zGNZa^SL@lidam+B`wx*eU^#bY4xzk*vy1!Wy$>fn&9{RE2I-Ou2dB3Chw%H1)EQ-I
zn@SmxGX3?G4dOW=di!Ko&c(D2>PnOm&sm!B*;$*N1j~7DQwuvvmsi_USA!y18OaS{
z`T8>!@H%VKq+iZGXQ_K+&H<Of9%GM$at1{-w%(VOZ&d^Tn;Wy%ydPp^5`3rN=-i$B
zbO|idpNdf>pAE=YG6o4z<a72I&ggEV?eOmP@%?Z=?{6TE$nn>PeD_W;<kI7WAPOcU
z5~$i>2_VjER>z4Y{omjR{=N&bPFRdp>x|T5^OQ{eYP4(3IsGAtAVRb=b?Z+6dH`bJ
z!4w5mz%<;Z2MLkmAE$QrgAX6ir-@RzJY<E9%4U9POt*e^>zhSxNA$CV*bjW0nOsdW
zQ}L8bVFGVGSOi{*I-#5`kaMTfFViSa`J&?*cv=<1^{ZT-Xl#f&<Tr}s;W_3FiHW9l
zb;_R^*%08Nv{H3<KIQL<QG~7Y!c<AQj(^N+d;B2NOj79zGyypql~!_SdHJb%*+4vT
z@I8M?9F}S7Z%UL!4-Cs;%A+5;$$mxY%)?Ev47}H<R#S9TvDi)u#$B+vE<v0F@d2R<
z22rT81LBgEF(#ljGGm2VL>u@`d_BJ(c7uWTm{^Vw2)iUe4P_$iT{aRhC<l}SF@h#P
zqJ9V}m|)pQMu-k5q+_S^9^M363aqO3cmI88Y&S)GhIWt8@C1w%b$fYhV3s?0y<p~c
zPooPVdzTyR;xTA|9Gr(xmVpupe{#EOmsH^82WJR0vRWTl&8gU+e3SKd6HNDH=~@dT
zM)7uDA7K*{8@791zbgNmH1zK?ha{Y49$comS-O76OV>)+W8x=@9`MP+$LZ~(N%@us
zoU}h&gohdF?_-7vAX00Py0{qwGBGX(7C$I`5f(pQUTiS`A*O~As|grF&S$*Tf$?`8
zKzu+W&koz}Nj4=#4nt>BU69HjB(_`c3B-Opc!N)W%~?OL$t&pSN4U1m{8E^a^%--U
z7Mt#!=}A8PB9S#>{*v9!t+wRgI**GKQ-ksqY-N;n;LT&ty_4_J?Q-csAvemK2wg+Q
zQ8(}3qD9|>PnZhRZrCx>7jt6Z(8()~Hi8fJG6Vn5QS7h3^OgOe(Yhy^>T&}-EZi<S
z4Y~(=o%XkAE%AhO7A}<CHu0B@+9%Y(l@U`&_F!a}BQdOlAkzpuhNXe$vn?jmp`H(+
zC(H+zMb|N^OzTfHFKOLIZkYysst`2Y7_vi*wSzE%WZACme!#rKR~G6yJ80XF4HPyO
zgw71d<_GUWG3)MZ6(&Lji}HZ=eCl*^|GFU(fE|po%a<Ov3_kCUwkSE)ox$bsAIw}%
zR9!mPu6?sGM7V?0%n4~JMbbQX-ngh}0QBX6rUD1+FtG}&sHxA&|B)dH))2xQ-aV%h
zx;jr0Rv8~zS9M<eeV&%2>h?F3HI^-vsfIhh6~jv8Q7VYS3=<^VL;(;LWL<O?8W5(7
zs7=8+3tlWBg51{tHbB#K6J5Kgp0<<QL!Ymf$Q(eUZR-3L;c^wZ!mN=4lr?Z!L+dQ*
z_6eSYwADzlVPyyeHW}ubylPb+TZREWu}n-e%_Rl59_7xtElP?sOZ26Z1!O#wS9*OK
z*jbeD+z|E(f2>MOE8G<hh7!8g<;9Hh6HG=MVj_jryc9NxU<hTD41?*9=^a7v8H9aE
z)xS*;I-NZc=bq=K_u&tHb!S6Yr=r@*PesJEa@wbBFx3N0Anhv-oY1x~SP^_ML}=a6
zMIq+hF+s4I1Mg4VVt@#8M&Nmnc^jwAMK}0u*)>bu0kmeFvoPmQL>SItH`gpAl?n~>
zJAXm!iWR}qpn6J(k9eYlx^OdMGNjEg&JtocA~~p@*=Z-ym0i*<VC_|0?8j^T#&;2l
zVBJ<v`KI@C&%}`HZn0fZX}Qw4h=J19#8D*EWyZwtO+~c#98L5@T;d^YqlIch=vb;v
zOiMQKT-GybnfE9*yt(s*v&_NgNjD8+phN%v7aL)2&^{%rx+mG0H1tf=JQ-Om`{l<B
zeEG)uNHhL8?-SY|v>=nYIY{_0z8Y2qG$*G`p~D*~cK%rVB4M#vR-ce~KreG*FQT&O
z1(?!A(O{7c!Go{=w@)sjT|`(Fswr16Ecf!-u#W#*lv~_ow`2^N26jSr1C)iBjje@#
zdAynJoktaZt2G8vLr=4h{|hCssme2H@&tzIu(VG6+%h&i3}e4U?<L7ZMT;X-L}-W7
z)RAq!bAxNk(kS7}Xmi7<VwER47%LHxbmT<F(KfUhV|AGyVvvq-|6E1=i(42U^(3Tr
znpCg&8+3w$;lKCq^~TULK&z1j_YIU?Gv{VK`7<EgzYmoaJTZccu1AIHNkRWYl!=(J
zk+5$Fzd{6GC_EtVxFrmoB?7o8H=cL(NysD1+@oI-XcM!nZWhc*AziampzPiKKzt5H
z!Lt(18Lh2o&RM|~vHD{03u7Ut_&U<b;r(Hcd^2%rr})b&$8RJTyyBn!j+qXndBV&b
zo9(M3>WM?wnZ#i)v^IxHf@%I4!87V-_|7anv!03fjyvEcs}z$R^HCTIzua?d2Cci~
zK%H!guqdXcD64t#x;qT5EA|)yR2@>^ELQQ$F=;`CBU|RohK%f16|adavHu^Vw}k?R
zVs29R|5|hGP-5t6!q<<`4=J47K7h<xdo5|lcSl<6Dg5T6n7s4&aPy;{4o5b^v90Gg
zMt)#j!A!eco%JKrBJ*oayaFZ=;q5YphF&*;!?w7&mFavj5@2_#Q_>$}a5mcTz~@7I
z1(n)4jKjd8492+9tJV)J%7nT|f47Ivw&AP>cx0=_WE#4}U+$fJe9u1u^x?Y6s^?9|
z*N-mUN*;A9SwRS(S=-uP+-@=`RdUc8U-MGf0oqj8eN2xy!z$_vzQKIPgBcl+{*kff
z<jFgiFyoSf!luLL=WN%4EKORrNQfcz*}^z{m}yBdTgCgEJXRckMzx8AB!ZYWfch%-
z;p2H29J=SVjU8s9rGEd@%B*G`b*?)s810?4#%(GYT&v~6kXVoBjD7hW3uW}>`=LZF
zm_+U8A$UgnLi;emK-iL6Kms-GD)_w|uBpf%Ap{0OCzUg(2$0B1SiJF>?;(DQF0pF@
zRatcTHpW(NHO33bRnR`PC=X01f`3$KRZs7sr&qV`dRX7P^}xm}TM_;YcZ)bO{{yaj
z^b@n--STE_dCxA$xcOa>=gMox#T;SfE(TVmP9DCJ4b@wNz0nL-c)F1qc)5Y-24l>-
z!_q}LLktw&gv5NQ!yLe=CpRZWv;Ej_a62um*aml1(p~3|iD<2?|6_H1<kyxxuV;s?
z5Pfx3jWa3}+sT@o`z`5F9NL8jTY2t$L@UT5fN^+K!;B0Mg7@r&K>9QTI{>N1-Gro^
z)I<XokS-!Db0T_S3=b_nWV~ePa+fgA#<4}ahfR8fuo%dMgHPZ6iYlkKSOhxX5$@i0
z>uua4L-%-9$A5J&bhb)Zk)6eGHP_U2;&m6#41(@{TR6r!?d-Etg#h8pX)+ybtDol7
z_8<mdmX{hV0+)KtT1#TJ0;YOyV9LStOLo(4PtI*sM#T=t+#h(h@kLI#v5*jA>mP;_
zxnw$n$)QPyVLL$^mKUZ{9Wv$j_zmR-^nU>__LwK~-b~wTK+b9#i6_c}_n>tFE+B2*
zqeC#u3Ok0BZU9FRLZZcUfhYg)`Hm*mhpZoA<T9IN)LRKM_~4I<r9q<tIbiR)>7Cor
zoyQ$A%gdwgN*GiTt=brG)R<QC#rg3aM`ENh+2#?beVu7<{T!3sBJ2@xd?PrV_GIfR
zhq?z%iwH;w8c;_UI=;q`y6B0Jx~OzCXxfU@pP)6KhdSKYw2X8g;v#BCtGhk@htgcM
zFzbU#sO(K?{I{z6DLm18LQ1|=>`o}4lH;h8uhIWMLRs;igG2sH-W)?;JZ6um{#gd`
z-#5wvPXftza|A1w1z%qlh+ZKF4-OUxNa1rw&O~m*BqxL~6m|pA7XuDS2)6(dXW~%e
zY9Ja7n=@`Y8ZaA<zU&}#Wx=p@{LJfdwqh%B@1d~`?v!EWSTSNiRzpL!h+IEr^wu7@
z9f)?T?1z|gi6thl4w=L8s-o(X3Nzd@zh-`-$+h$YA(Y3etrmK9>S`vTAqc6E(KYR3
zS|L<KI5+4U+kX-jbi4uje%h>#u#p$)Qf@cvd~z)}R=HJMLJGU=^%ckbr82ztl-`y8
z54p^T=%m~#ak5o7LDtFh%hus4!ogvOl9FHY?>;1sLVvG$psh$~8)@$h^#GGuX2wlu
zyw8J901)#5Sf35{g?}Z4WO;Ju)8A9-XhnG8XQ~;)LQ3v9q6mz7fda?1GIu7Tu!-rt
z!j!P%O$X$nFH3|nJoe@^=ZlEdBf|j_>pg*qbLMU`hpnPw76k=FrLtwa`3#J>wI9k!
zLkEH^d|-n4J=GhJZzFKLprg86cy9ax%>@FQlr_}uBFFL(PN@<3UD2z04#W8M1_9_@
zxi|RnMe*)lZoPfSk$pnceuwC*u)f$9qPL%KZ_<CTF+JBA4260}%x4*ni6!iirIp63
z|8jX5u6X}9!AjFTuBc6<eLQHDA}L&i$Z!Y)#chmA{H8$h76wEc*ele|plb=z)*?nP
z6ps3e2$4bDb;0<{28<mTI!gLL?Ycptio5T+7@b|4Z(PR&=q)h1tB0M1_TvSWv(16z
zbOxRJUx!+6YL&|m6J7MJ%BQ%P2KE<BuJ@U#<4@^VNq`9!5&Ejmu}lVq?2TippXbCG
zaYL+I&o#yiFl&)qz8ETJ-2VMk|J|Ar)L4xF35&~34$pJ^PR)8s`*D>)7vG#I6Cv8F
zHB$5Y59{RDYayRXydPyGYugcH;6=jxkFPL!mz*^APMt$bNk_@6V*Y<?QFG)>5q%+D
zXdE>{;)k<a)p_O+TC&?1wML8#0}PuLZv&uy#)(UHp)4+vo+G?iha5pW8&v{3|A;@p
z186oA;lphWjIjHGe9Q}ptWBu=Kayw+y-nwd-F7P~D#mHAm+w}W_$FTUtcia~DIT~k
z(c(T*=<Y`>P}4D8sgq0}p=;FnL)ZA+&^X<+o;~V^5fj0~58RuJAa$N#5!OxSk)S~w
z)5gMkA?~CIYrHzKBRdG~K-<5>6v}m}{GDO`_kqu<<9Q7m{u=mPBT1$1&Xi9`$_;5t
z2%4RhG*z&pwS@5S+x4kz(Fx7+6jD&Z8SMTB{^%eAu!Ei*z5&Pix}I?1bp?J?tH8S;
zI~-OO76ThiRCDDjo$ja^?7s9xnjEioc~X#HW_@fW{HYx~5WPLJ?=Oo@cr60cgKxix
zCep2Yd9189E8kJtpcPx`biyjI3w`LiPF{&yv9-2n_y=iBHh^`x*u*tw!7Orf9M)fz
zdEV0O87vEQikUsw*9nd<%eW?SgVwY$C3D_O!seidwGxNVh+CnVNwD*C<N|bvJGigR
zxNWT?u@<4KT!nGq)@EfjSah$aK0R(?l2mbU>22x%OT%-h!5?EqVYBo&UE|Z%bLAnP
zy$p6Ej3t}-`L&X;C^vZ*w`{XKR7|A21hrK_)@9L7C@n%pL4bUX$@4SBkPA>S1Z7%s
zRB;V027bf#D-;%1FqM$H&&#=77!o=k8x1_Yoz;gpeY&yvLo@NMW-Sd3S^!v&8`h{i
zcz+O9g<G9tlqh1bzY^ZayD&)FUTHyln=}|y%?+_^4fCfz@3Aeb2aG1V$4P?uJb5OY
zF2#g1JEV!|PHv0{|JlWKqti)pQ>Vve*Y5kj_)yelA!cjezk-sFowe~1w5yEokXahF
z(*Usx%|3IvqTPc2Z${=ktYtG|?mrMEy{c1Bz!oNiZvAf2shM$=vF#DlpPB&dmt(pC
z`cWExU{DS=5oEIn?_^?q4sEs@i49ip`+Yiw$h0Y5T4oNpEC&nF0-0f=@kf_LR~%82
zA|*Cn5nCT7=kBdP@$0NSs*z}0>wU2-Wc`&3%2#k}n#J5C!dakGcHaAg7a8cB+_w7|
zA|Bcc-e%rU?9=%TGva6S{sB*;wo!B!42?On5Tq;LDhlQN3P$NhrqyT84(fKTFGg{L
zX`|Fui$0Mo=)$5C!N(7L@Pc$!XbFM9K4I>gFx&|gIH|(@5TvA)*?8<(1Th&V(HM%9
zPsb<zpM7Q+27T(Z%;nr2sxwitKcgdlc9xuG_1ibf1_95;qz}6nozN<mxEUb}%OKGU
zr4NUFG5NY}*7hFw3xhNbDxq}xQ_y`K+9!MyXbfzixB&`5<|-S~^@sR~UvB|JqmHHZ
z=Ly=fM+=J3Oh0_f)ZmuHHgOL$_Pm7??sgH0uqp|$%8Q%#x#ovdn%bI1qnBg1MsurC
z$|gatjvF;I%aCfHKdR&RT-GRLfjwVqG%=*8TxXxmCM9@fsI1_<G7%Ew$azH)47>^h
zo+QYKHUe6drxKvvN{P4$?#|)R$|gLp?a&l_0S*soqW#qREZPt3#d85eoBlK?MGF*K
zCuF$DIV}Hhc9xh?+dy3@sf_;lPOq-c$T&K^I;+WWb*G+j(w*uh6T|%)Qwygn_s>)Y
zy%2E3C_pgC;;$7KFr(<xI*KVG7;^`Hnb+3mx~`LfVgB>y7(^3fzXF&ZB3h8kIn5cp
zZrXtye7)olrUpN>a2-zpT=K{_5xsr9%k<98x0m(gcwCecAEB59g<{-G>wE9b+B0uT
z;GjafBCG=|#`9gBb5BP{Gvo)3-8e@mQ6@=p`AQQiK(98OqRz6*Ne9|}P^8@~v4+@U
zdYhA?qtl~$Cm7G1W$Nd)B^Qj&*{Qxj@K6wOIo!gO-%7BCux&w3#OcMpPI_EGsaZ!t
zG>b_t%B$-pj{(#Y${ru8xu3zVHRrDm+mksWDt9*35d$e(!s)rz|MHpBH~w%<IL5=T
zYgW~h(!<z(SA{&lgr;XNN9F>Y#7qeiSdnVZGnjkj%OB%}IdN^o+zDhx=2r=g1yTLR
zA)d%zruuf1b-mM8lE*5a+bLr2Woi}!MeLRP))>(VaqPYKkc2%nD1TtF&7guJt*~mz
zcfeSfBJRkB)7+9f8$zowU7DZ`p>rqAa5VaSp7orR-^?NqKFcBsWwC_lIq_z=x>Ae{
z`b0XGL9?x2p;j#O_^joTlToV*iEH)HCam1KD1Jf5_l66>D5kitTWM6xoEe_j`6NDk
zetU7(L76NTbx*Os>9p_vml?1Y#`$}Uk}{^B{``#b9QFG*h*RgjoEaHBMre25ExZ+^
zEa*sBgm)<Z(Srid^z{03hZ|=L7dW8=HBTDNWB3&|9O^<uh9triO|N3#J)3G9rifns
zz1u{HaI5r6+G>XOZ{hX4VS0e7@T!c-lo)c)YSbgg0~7hTK_x{j>^Ufm_a|XEvp#mK
zaP-#6oG@2@|B1G3V~wn~%uHkx9@JzrvFv6R)!+0<m+|tgq5^*2hEZ4ac0XKayoG1{
zqtV97AA%1H!nnzy%9{-#HM_q^1R0h^Y3l?d`CCX(yGuZ3Ipc@pcZ#u5OPMHo<=44q
zU)pg&W}n!jg_t#6z&<o`PxSw)c;~%S5BWeL{XH9wj96uT$lm9lKQp2-mQ&{Md);_J
z_QfXxQN~Xe6`2^6gRclJv-KIA2mXhg-M$%A4>sA0=7-DT!4HAIKn<<z;G&-NYG)h`
zBEDs_MaImzecUNBM0iIwIuVJpife(##VQKpS#YuL$m;Lw6F-}cPI=QiqF&kL2rgOk
zcvrQ5Du_R9Ms#awxlNw_Y~wSQ`X@Qk#MqgBa;AxSa!^HWMbdULenq9K#^3%vXMh-U
z@PTaS-a6*3F}`z&`?n0DwgF%N%8%O<d9|`5s~!Z1RTyr!5hU7B{$W}F-v>-3Z84tj
zAZB%g?$bNp<9+&5K5`U{*FLxFM;2s<L3ti^aaILP^6|u;JDEcymfgwzmu#oL5ku&R
zLF^KXz&0W~Cv7cen?6aUfP(KRP$?p}nod+qSAw$8-fUP?K&u&jRKlg=>24!o(Xi)D
zN3v>N<;|~V(%r~hbLBS64<`~|$u#W#C(yl5U!b`{uL^Ji|3i5W!I-eU#hDprdr~}p
zKZy-jN9#^msi^O)X8;E!@P0I_<OHX-MQ{@EiGkPdC=afTK-6RPj|kUqlG<wgP-r;p
zR2!jt-L-A$o~`cwU5*0%0Yko*Y3aKvg6NBz#@f=`w>a$3OXvZ6LFP{Imo?a#h{zT_
z+bV0hnSg6k4e1>-V_#b1a=5g=AT`A79HJ<e6cytQx-jy5k>D(2jfTz>dG;&DW*(2G
zOGn)t_3a-I%Wai3->Ifkz1Xe!OYL<fA)|xA4m;)ZE<0}O<So0Z#3Awf>$BZn89x^l
z#O<hI^8PApA!)IHaS_M0qO3*yD|X8-E({oX&9vckgaG5kyG4a>w2C-fs*mq7Z|RTc
zjCXJ4<=pP*Xkl@}*~!t_@k753%RR|)RnEQ5J7cTdrg!e(c9SfUx^`EbD#XRf$yvKO
zBwg01jz!9M)8%oo%E41>LUmt-MU^-@`R7V#JGwi{tnBnqLWv=I>#g`!-U8mRl3eNt
z=Z~=1o%1FqBnv`Y$7^ldv-}6BIr;P@`>q-4>gsyyymCEHW3iJ^lu+CmTf(dKtpCv$
zze_asmi@O4byt_@$A>VFe&|^@sJCNr?A)%!sgao)#ylCBA<=r*=toCGn73UrR@E%@
z-&C$>wcxYxxRmkI3i(uOzYcxPyOxQyR7p>fQpRxaGk)sagyf{$9L4dWyC*t2J9Ou}
zwN9<_8ye$Mc(E$Qso~iYb93oc7r{gdR?0Y9{pBP*tlbI8Eb8N^Mb1k4!prs<if~!R
zY?8IAes*}^N>T_u4y-P&Pl|EUsy~76F;y2kNjoXU8=mHh=@@*jU~W}h)^q&Rg!`=J
z#T~ODk@elfUKqRRHk29nkghMp5vt2^akJ;7jlk~eg2}2~gPVIgPK`47KeKvx*T+8Q
zoV19>4;cy9FqL;#GLe_Q7`>r2;&pc87(1K&c529hbZJwO6wSO7zx4x#=S9Ndm|1gS
zoxYdfg(077ba%Jpz+v~TEEQQBP1U0nS1xE>>noT;pLdd1RMxl!Lghm|uHx~FkI~nZ
zw`}{7g|<>w9;(WrUDkOYCfc0O4n0hzvi_2G{`LIyhwag}g^i<q9VLz<n|JBT)nrOY
zuHcc*5HdguB#prWS8bHHk)+Kwr{MjpBD7<0#f20UJ3$X&s>12dnQ=DM5fKVLK5ZI&
zn(lHH+U2`<wPT!gZ|sm1<>88%s=K211fq6Au`PZj*S4~IxFvo0!g+3c)ssgl`#E*P
zsQvD5WrE(<U$Q9~Etp<^s-LlMvSqyW`)?2Cu5k9nGdJd8bLpy2@7fem*mF?V)pf(E
z^?tdc&OGOrXf}S~4+uHJ8%nznA}X4%sT5pd@4=_Z`I7TYX1{25!A24BZmA_#N`$&r
z&?TQ733~10_yrV``_O5SBj=B@Ry10=|4NZ6W1Ri|_*M@?C5*=F`l!Wky>Y|d(`n`(
z9%!hIJG;v?u{tLwQ(K=Ydeur!l~M}=Y)}7Mr0{~0q6Zy;{R*UPSg@<9UhcuEF6cf(
zdWE8s4c$7u9x6s`!@Nx$gC7%;Br0XN^)Qw#XSxF?kdw1HPRQW05F;UkzV%kFLB+UF
zi+2LUF;^d9b6s+cch>9f;c`6Ak8Qq7q{@e_TXj)8N~qs_OUSEEZ{8W!7T=*2SDn;n
zO1)oZ9=7xzxIKqB*K0U=dylravoXy(v)Jw{pW5hG=C9AU)T{mOI}B$HPxhzdZJ+IZ
zBDd&XFCxz4=lK<!iqh_Dsvkeqy+$vQ!^kDYU@mHyrPbt%xJUF7IM!>JsE_Sf7EV)G
z9~j$Nq^Km?7?-prABWF4YlCtA>6-3jv9F3=d1{^-YDyuBUgN_xC5+2VF7XYFng)Lz
z_n~7oIh_{-+B+qAY+t4OxA%YERuXHGPf6*8v-9Aoa4eDRlx6*i^2~7X<FGG<f&vsd
zPAY0D_!M_bkrWft)l@BWa|~+VTYvqo_^Fw>!~(9pdP=p=$kI5i4vSRY<spl$_WXN9
z?6K@Mrgz>|eKHg|XC{6^%<@uMhmF8%mrCyW8c{i8{>f9DrZc_I7OZ^d>8qsRJok~S
zNCh_7lUP?puBZ|-$K`b<UY&ftRhuggw>I*fIqR)bvNxUiO#(aLkT6H6=+W@#P*D!I
zMD=BoRL-*F7d$?wc^pj@;{B}4nAXn@?BRCe6ixfgQo$x=vh@9>ZHb!#lP$DQ8610<
zuU*QhY|K&|FP+$sv5jLw#yGzrbX3Xs<cEjPPd^wv`@_#8woSEkl_uS%Yg5p^(;gh|
z9`Z3S8$AnxlhwzoE<OzTs*Jc)Cr5OZDX)dB4p9stO!Ptk$_lVDBHKnI@E*&C2Vbc@
zz)~m28LA6yH@2#iRfsO7B_foFa!YpzEz^=)rALX>R8HKyF?KjrwA(`oVc+qwRp+o5
z+T*>DH_*S&Wcz1kE{N|KR7|NAb;wvjiClo^hSS1%CGQ4fn~zO8jWYW;sFm3sukrLX
z*w(=fuWyPn6UB^&>q^<t)OD3dDV<5)V{^NBc((6MdhWeq@s&`~16aZXhx@p@MCsQQ
z=TN1p?UYwq8SdC<?|#$zsM|W`uIQZYYqA7PAPIWS*3c#KsHTqQwu|m_V|h63T&wq;
zDT~C`)m5QI9Mts3@*&VLmolo~Ikw3cGq(MBTs$#Mtw?YC;XOL&Ame@G9)%e#6nQ#P
zLS*L!CN>@4TjnT(Tul!YB_8zdVAcSp45Zf&`>Dqp0s%sAn`bL)l)r2ZvCSfd4D^MT
zVSm}2!bx?ebZK92Gum72D1+J*_;Z^lbVHH66WP{FLv@)mEEnS$Jk=By)4k6rxbn%N
znT+4IgDteO6)V0yl<e9q?`nC{Bt1uu*l;^jy$>8s6<psZU3~m%=7Q*{NDuZpmkNrR
z#b=fp5gjMWLQ~gzUQJzAEyEp`3faPFeUSlNYxhef-u<p2)7Kv+Wo5YiQjWJ?*7*64
z@}-OfuFOg+yX20}?tYFwO|RAKAKGw#QB=C@dlL25MeaiZO9C!UWvTsfN>#F=j?{K`
z-WstQni^l7Jh{-X=e=Z8XG(}O)(NMqR~dY8s>@}kl?ZOZ_L@7mvF9K)fZzacKzO<Z
zcpw`|b!mCmReK<C@Klv%Jh8FLof_rM%zx$Q8w)K9DeKV3;p`rUK!k{6H|dpKxCBsl
z%{^e%B%knFyZ(gRD-Tt^Ei&;ia}OzTu(|y``Jv&HZ5lHf{6(v~B08F>{(=1p{!e-<
zt;*-yHB1aBGUHEDYnl?bc5;8=xM^9StW$%Ai&bFN3-1+uKO|a}nO-tj`{$l%_8GWr
zE6-W`Nzoua<iikfhd-VEQck(LF#fXjOB`Fa$F%6MzR05E7cv)IVIMKz{*MsnHE3*~
zl$4vO=#?!yZ!I-FtNZXp+Glmk7uNLLchj*cgGOr0?S3uSDq$p7MGuTLr_Y8q@f1BQ
zU5aRnV5-pi3+;MkxJxmo0<r^Kp?F?9bK&(u3XBJcOY#(E+XmZz{~fWJxaPRGPP3Ia
zxPAz`JSFXM#|+L2LzFJ-jon7OcrrYc6O)uo?{K#dDpDdtcqL>_Q>Dfe`KN^aCRqJ5
zdn=z8<026B7dneS!`Jx|ik~wU?ibIDOI1B1tCZcrw_)^0ic7`9Gi=UtooUx)I3M)d
zt<$7@%FQz7H;7A2p&yO7Cn6fQqCrE;%5X1_y$3g|s+Q7NdeaK(BEj67N;!+7S5n>V
zPY-Qc^h#AN-hJq`ac+%!=Hrtk?+@j%)Awu^nDh2ayJL6kD`_>Q+lnwW9}G(_X&lWD
zj^oZ|1l4Z9#sCi|<T2iQG43nOhCt3?XsvVmKF$qUkO=SI;C8_SrxNG!fLwr+Qu=_r
z#x0{Lp#g(8#KP!47BU#JIZ^!VuznN^kgUt9(hlh;h7h-2;VD&Mya5r3{g166guz8Z
z98l}(E)NjTKr<K^9dAg%!xJWLlKN&mHhEmUMO10$6@P&&+yG2Wixx*CCVsWkYm-xO
z5YP0F%KV}t($}a*i!O=yk$q0`mcH&G`<sTky&Ud1r8LVmq5u`q-8u+U@_l^Lc0Iu^
zZWws|CGfikk782~Bb=6Q%kf3g>xb{d@k;jJ$tB7wkH0xPbb8{!ip-69b>h7r0ACN%
zbHDxU^u0uj{j{Gkx_=`P==(d;jMaT<i`HQzdW+*egD3%mWr*7nm5~3~Tmzm!Hiab<
zAYd(0S41pKM>#K1N$9Ue2^$7k&diW2^X7As3wCLoIy|S2%JuC<-@H=t{_KAsQ7KD{
z+H!*v|Gj$hmW{DGc+IfHW{m^2*yYi3*r-ldK0#0J^rK;o1KWLnd^Pu(=`9EtO;RoG
zPn>!EP?9xDZGtI!rfhh$E^sD;6~R7bn{Ls*!XD4XiqanQs{@^u@7hv9IhfBamh3Z-
zk5hPbL~9NuWw{h!?P{;ypOJCb)|ALe@$8J2joVz2ovkyO6o)zFPSSeg*EY|SF>aw>
z-doLlel>NjJ$rOkM*BCt!n{KPVY}xx?@z0vd2$iQurH0u%1M><pYC~}do9;6DfgbD
z*PUaM8I<xKhONr!<wQQvC-nH9%?eJYQ(BpK?@BE$S|q@&H<(I6u>tU(sEwkX6Y%yz
zik;jld&M>2k+lh#_9cd60K;*pNsK{HA(9&1P7d4EVd<|+?$JZ2zeCe&mlVY*p;K0(
zQeT9I;MT&$hNOv8j2%5f^XE$1@L((vLLDxQMYdK!2X?iavdnzxpO&zZ%nSESRt<Lk
zag-yuGq9psZP&=8o7dRdjR<Eyq&daQq*se0qL;W{lUO=8_J&iXCkJQqigiwu-3@!v
zB7~~^?ANtBN^6bZ)wD9S;Ngm%+8Mp?D0A1j6@{gWUXKAV)NwavUhH8^a8OyFj{21`
zDicH9hbCV%7c<VtDm1V~yB)7M<o1Y6n*=%>nZ~-4e=nR)+m-+5n;z$DoRd@+?H%0Z
zncs~Zdq*d*<FKV>(Jbb!!>36P_hagl&`nW7g3CfdXrgRr5Ve9rWRsi^G<~LE&Zw{h
zb+P{5@A^UkmJ;)!)xIiXX08X6A;F^~4x{`^6YhCE$5+K0&Tnw#!G(!>AXHgF+Jy;^
z;JJls0hhz&*Upg)$ID2hBFe_^Qn^=QPI^tp5(TsAy6VS?*5_E*2J#E?)`$~7cP*3j
z(Kr0!z4<ytU78L2;<n4h`Um2jb)WPLsh8{P-W!-#(H<5rME`23i*Y@|ql!aQaAsT}
z(4YU>0qiKhcz<4-#@)WYj^iKI^@!8^aqqP%MXyD5v~;UScN}G^+&u)mC7<$lbS6C5
z>;7wZ|6EV)=!u!AK$Gp;mwww<^>pI?*z}~spz!2x>NH=znUBl+MK|_TXv7~*de|$=
zMVplzVK@j%%P2DA6)tNiV4*dKoU(dd!Pg2IKp4%dl>qOYI|gRq6`&S^9i{7(>ZRk#
z7x6gCgqn#5;j!L$9>XJj)FbfVQ@n@g|BV~#kbrWBmBh1EVtD~WY7azn#fZtZ%zRG%
z+&Z15(b}`=<H)3%(MVOq*j_38q#sJso4lAc=SVTb67*Pv`k99l`(_&FoDpO~Blq6)
zxo@tr^U0Xu#1e6DTSCG$RiOLxE@xmKSn7JyI*>~$tkJ3YmqC@WY0Dm%PAd^vC<a+9
zxPB392iSq&HzHtcGi_1n_RoaM>z85!8}0Ly>#e#CC(h-C($b$7njgx{U}?#p*fjZ*
zvEL$3>P44pW3c^-mM8mwc7Zl<VrQF3@Cy8a*YM5jfPC>P>DY(JcX6``caWeAv4*vm
z45Hv<^`(R$HSjPTmtjG0qpHZnFUI9LG_DG_3UUr<l!Cn#OC@?Ri2L?RNm9G2Zq86a
zminul?R*^#&kG(Oid%AJr5JWvtyYHAYVpdW+FkkmU+q_kdirSaT+d{t0;w6=eIPP}
zN+HVbmBeYSu)ZddDjEi>dECEoDb}T3`c^CzN~0qVNagD^y(DctFS!pn2Mt8|te~<w
zt-j+v*4|mE5=1ec|8TOhwOAuGLF5xXNwwg}6q!un99J2&>;>Y;j<w`;#JDJp9e(rN
zT5}?JX$|HZ!xx+xURDc;cK0sxQ9iTxC)R?RgUIo<bDawO{9qxAEr*+u1ifc*RbZ*i
zm(v)I&H?0daR%uKEgMXtuK>YQu9#@-U3UfA1qCA_38gUvVk-#?RWotH<eGH}uPJ!N
z7ZYBeJ~UOBsj=JU=iLpu2?7bqL@ruj>y`QV&?7;cjPen`KL2fqolnwSndWaiU!gO)
z!tUf%4|s;v8C15PypUsIXePfxoU8Wk3NfzQW7%gtSRSACScr)pN{XBwoOPO3x)pk(
z9Cb9=zDSSJ3pw8XUFE@T8UZrKJpW6e8z=+(_jkEE{16LzFW?f<!J|IDeDdu*Tgxus
z+7JO46KR{9-@d!hDu&3+56I=RRRE1a4T=Y>=H0s}N+7No)ges)hI6udByp(TUxJ`F
zD5sYiL^-G6O!nz991eOA>;+Us-;GQGG>EISKdr7O{-e^Bt%9pAQX=)MSSsny3>yau
z5~p}`CprhKtKEu{67H@UDHA1Na*RW}$8A*)c24opw!jLCY2S-2t$7#2dvu*%H&;4>
zFS}=llUMWWL(h!`bg>F|MOuX3weQI|hk90P^bFHP_qQ3E;1ccV%lGz_?7m*i&>Um#
zBCgJfHKMH#6~{NPfB5$p*2<Gg<+r4NzdswtRxfLku-dJ@z2*JS&THzW_xX0&_J@_k
z4)i9Ed5@V6-onN@5PY~1y9G9LkOhGaCP2n<WvGR~Z~SpJkgnMH2l$qh(J^QK%FNvI
zvyrpMaGKM$rp!V5rl}Z`O1Df85N4Qt2%A{>gkOFSEG16YT|u6n*&u;q`L;(6NDhZ{
zStv_N?QI;(z}j{n-Gl+hkHy|Ni%)hnFvioaXD%6hm3*(MOz6QWZvuF1cr?fIsJ7D*
z<J<RNvPC#m8SmtY^e6+>A^l3uY$R3e9076#p7kygNXZir5u4EI5~6>!G~8Uh*BwbC
z#n{+BVs6)KpP-T+#!K}kudEvPp<&Mjp4W0OF5e}&kRZ|2RT7hOgA~0A=WCTt_v)1@
z?9H0~*=gSHX!om!0s6zQ<*KX1*{puQB^tg%@mRDY(v2tiR%|lj0z3Zfu%wWM`EpQa
z*j*(0uMdk{eQ`d90^>m9fff~q6BmP=QO<}piP@g`7uOF1tJR-CY(w>5Qk{o(MUQZ#
z#MS7Wx-smiRbWLs&;=ea@wErIYY)6zbVX@zy!_{5<6M)UcY&Jy-KbIOsiE5Zd%T1;
zoY)_EcyN79c5-=kkle5=v2JhT+<bMPxMpjNY~TH%O}G56_klk9v8BffPg7F5+l8Dc
zFFp{ELvj0okRrD5okw4Ftf8|p^hQPPPex^qx%1VDRfs9NAA8$;%&m{%NYAwr%oXA<
zWoR{&{_GqmVZ?em&~qo_aRBQZkB!#R&V|roHD-QVP)wFWniSmL*`4IcSguuu2_;-z
zy9P$`(m$6t9%%sTb-9WfJs3uWvlkD&KNAWq(jxL#s*nQuN65%up1BwXlZ`}7!^(-!
z*r>uJ1Zk#VF%axYd{LV4NP^<TDUL!Hm<+VmdDkL&+6-ii-Iuz)ti!$7AtT--{s2w=
zz$8b%npgM>akr_an8U9Pca|RO=W1-3x=;K)4oQ73E*C$O;T_&)xU;V?KDDV7bBgMx
zPDmShh4jb~XVq2xR5>eu$k%Kl&rB+txqnxY{!%fhhsB9g0%<~2H&rxD(aO*yzw4|_
zhofL%tlWN2yz6W(VJNh?h3*ab82fLjRPj&Xye|W3R5#oSXIjTPcMWVy_p@&ZqYS}C
z9QYJ!QYb<6b|44Ze1j9BBt&_j3hb*@7lV1%1LA&t(PtztyLg}TGTCH=5qun1fnc2B
z$V}D%uy(U)(*e0x$edCLpJUKQh-M`fckiBV)d+EU9jVWfVHti_=5Wb5l~HtC)Xxmf
z1UVn})U?%&vTLZD@cS}T(kNT-s{b)^rvq-f(up@EL?{z!qA(M4xSBb(X>Hxh1`UG-
zX1hP~fXlYroV8DhfPbY$*|(EDn(L@Thn#aqXRbFlah(pCM-JJrRQw4hM!ZfB{kTAL
zdlKO8F#r(OdUx3SoRa&B_RbR@=1Z%2y;od$bgmPlGJ2O%v#G*@ZMY!S6;Ap4$?SYZ
zFlLg=ZFi5n`WiD>_G+4StmNDbNNJM7m?i%uEg>)_h}tZ|FRVqV>qu6wkx7hD=#JUw
zwPC*%&*GxE@oXUjY)S0LObPGv3u0N%omFu1l8J{UsbFlSAdPXVm&WK^I%9S`LG}4D
z_L+i-f(t<TrO79g4i8pU-@BmX$s9@@R%y-Ili&EgXx&=kPjO6_Dw=<pvoMp5YRR)B
z^E2PHXv0g_Yp&T7ANziUggh(;G5e4!XxLp_x5B`Bt(dKX9&&P}7w<I+mq`7QA5E;G
zYW$19Zfn`J-gtjr?L&nIyVyEf76(=GqrlgX{*8YgzV8o&UM42_`2LkCIqa;X%?yV5
zvl2ie?g?)VBV}y{QspkyCt?Q678gPC4k8~uk}*!`-NngsbC}J|2MLNSmqASIe(|v8
zd(ABA^L1p|_F(ST)hGuZkXxv?Be!1#Inv#`$SMK(;TBf(`Ax!I5a58okj?{u;lUe#
zF+_a_Y4u|Lgwc4{l{24z*}k06n3SFL>R#+$tJYfKmAH#{uASQ#bu%Jzt-WBsig#k2
zlk^U&GC4DO-rQg3E9Fz`qBn4G?$v6a8>{=I9FfZb1UO5oy~on(;?5}V=s9b~{cM;`
zj=QaM=#ooC+Vp&~!t#3I|Clw+(%CDiDl~_H<RqWf?VYX?m$Ld>%C;x_IL8Cv8ym`T
z*`Fb^8LW`GITAnYDpb^GEgS$?-7XBBQXLVw7RuP4%FrU~7rLDT1({os3xRh9;We8S
z$;AI85I%@%JhQ+aDrKnMeAN3(7)X!0JQ{K?YFLaEXfLi6nJW9{UxNE^pWW>}DruM;
zE|r+->OCWwD(2R&Zb_`q8Qh1C%g-=t&X;0=)BlNxasE-JpUsUIn@(_fvadZ;OORsB
zoVD6MS7kZ(f)u;aZA74mZuXg<j+`Put(;AFHRIhgIbLnM`Zx#ghJYJvAKUF@2z7LJ
zRTrZK@@e_5^k3?6Egm%Y%;9Yb(+9Ge?i!Vl!8~?wByZxBbo?HVd7QZ~Pm%IuJ6I^G
z^6B90<?Mw@77QYR=E;6=ydk|aE*^O2b#_^uoYE$eK@4#oE7<Z#qB7yOghBNv4ldA?
zK%_nxNKiY5A{|C%fc$_W09~LK!N3wvY-Nih1b|<SEKm*rk+?X<rNeJ}pGL4k1%Fx2
z_UQ$!Zqo<WcSj4caXzOp_WPZ04I^-U%LTjN72nttcSfC1d|i_wR>9*;6`M#C^X9tJ
z+=xx}%Ai75u83EZ+E!fZVtAu7#XE$pV%nfWkI8G4qtIjF<Le$@|DF1o&3^uk`k6QS
z=GDeDYO9>_Nt+i#k2(nAwc}sJ>&Co*Bdjx{4a4kjV3mRTB}qb~RK5d6By_5*x`+qf
zn82*WNIkOM7oR$dh)me#&#u~m<p|kG&UwOS0jK~NuSoq<JFk=kI!P;Kn~0vu0MS`;
zybGx#@<}XYNezf<74{;hPl11xbN8quFIgS)BX{o?Te*TgRP3qCkEIh+RSUsT$K7r!
zJ09S{uE|<$)}>AsM%2uhRx#y`w+Y&gM22i6l0~L71fO^`X@jtc=tLT|yREp4V#0HN
z!Idzkg>g%w>-zYxe8lNHrhQUr(V8Rl#er%j>VZT(>3Rt5?Aq<yTa4(rElD}aNo#b0
zBgRM98{2$q+IIhA(@)3Mw&MALGsD#Ua3a1RoY$=GU>YB1G2bwa8nJmD>LS=!;C8N=
z-Z>Qsb{c+fS3^;m=fgF*_?*rH0(ggrTDDJIIUsid53u~a$~$HYK}ewbwun{|QQ0K}
zhb~|(bWjmY1fI&S+&4;m5#F3n7No4>?7q%CGrF}sV{2*Ele-%I^v?uA&0qfG0lg+T
z?sjg_uBdf5h@VuYwXEhzmCBj%J;$$}1Ex#OP^XKNuJ0A@a@$poU-ikevHb!mYA+G%
znV5)T(piti#@6%zuGRacmk{KR-ZcqISFU%XMCw@&=4xb9XKy80=DYusq|Rk`_b~D9
z_rx&P_!Stc(`ZCrtmvL8lpQ(rEiHFseoh!oGX0Y^#_~A0=lYJS`+e(_-S+TWW5u(>
zUtcUmnsZe|&ghT+Xc7xR15^eRb2K&qR-ofdLCFh;xBi5rYm%Wz&h$DW!{)5rARxjY
zW>ASr2hi5I4<D!;JfL#6Y}s<^a}Y1qFSm6)>64BcNx7p8!Ij-I1zkqG5uxFHwzN<z
zZH_O|t$cfU5rMy2Rb~<%B-RB_6n6#{c*g&vw_@*6K5;GrxGxM1U#@t4XhEH2_W2&6
zu5lBu5SQDk?xSz!#G#s09)XM_R8wi;{%`Kg#@Fb<vdfjiT&`Kak~3EK@V6;FjwxEv
zJNwk7*Brl=n-h_2Kk;TQQ79o%*anP{cP>ngVvsf9ymx-#G`DB7jJ9(cb#C9e@O#M~
zp@%;G+>+C^Yi~}?)&83FwRsna+aS^LWRLlAZ>yLtBb2)v66oj(CS(UcPR=UPU^jp|
z65Js2?l2tgqSkA3pG0E1PZr^z;^ZwtmjLr0&Z1zf*HUcI4~6w2mI+l8L^qU7o=c>P
z1Fq^)ICXE-#t%!IrY4NFl+2tZxx~Z$%S%hloZM&L4l{S%JE8+{@noGm6Bi1Y0mvjd
z#lu`GruC6A%pH}_%sKMCZ{HcDMB~LLedp+MFI+~CQ=CsVGkz6%u?`rY?%nC@27k-N
z&P9&?^fg)wn4OGaMnI&L^rSU`Z%P^U5=e@2N7TK6FZ}OyDOkD{EKmFXC!_5{10vJl
z+_sETGal{p#(ikHf@PQOu*S^0T>~0<LnXH?IOd$Wd$+~06NSmyI<H?r|J3OR7KANh
zTAda-?z_lQ=B$&_Xd_4*eWCU#I(cLd@PY-udl$q73hCZgFtp?xXs_A(5^sQB0f~v=
zZ^;)?)^e{3m1SMZSv>e(s43A7>?JwGi*G9-#Mn3K(%oQD-80qrv?O8PXeF_sgJN0I
zdO3BL93Fk-b6&&YL$l8CRMSy*ca?45j|~-0oE$DJ8rUza`@|h}%Ogsy+)n=2Bt*n&
zSS#Ag$1jC@-)ws7J{n?QHd+99cAvv0ioK)rn(MkvOKs^sUTG%9jF(sZ#}me=?sd9Q
z)_2!G;3v52lE=G+*q$Hp8v<Q%t2(dvf>9a6Rw6&pcBVLX@K*b_Z4Xi!%QElB-zh`U
zH^vF5Lek`+#XW+Iz4sp7KE-7JiXPsh9iKeBFHflk&%D0`?#$(g8#6cYn{n0-huDFX
z1?Qo+0|l4M3`aPgb^7|ROBLf?I1e1!ir~+_5j5j8G_(sDXv2$1Tud>U5i`hdGc#}U
zUw~t^Mp|dfjB`=2<2GEh_xdi<I7Z5L;ji;GhO>yZBvx{yNl@E~Q>ZteKOycS&1j6`
zWJuk_RlA|Ae=_t)2JVE={dzd1bExvhg=`TePdX<9)}61UMbrBD1`kgw@vu!V<tt@q
zFOYPp9!ag5UofLy#<-XM>Aa%S?YvPHW1Bw1a4g*qWw`&cMV^`4h5RXY*Wr!ECy&Hm
z6h`rFVyuT&BawIE2tO<dSIkEV(!mKd4=zg_LcGH24aQSY|6n13BT%wL4rLW19d8Ix
z)bYK^Y(xzRAFryU7Waq70P-UdC9+Tt6k0%A+%XGC5rt(7bRIhig#nWQU&K0|F!!r6
z^S*U!jO;sgW6~WwPsG5~mxyN`#*944>fq-p^&?X)>Z;GQaWb>w0xOq%9;P_WaChlN
zW(FaB%(oW80_n$pJ;{ae|M+ZX2rd2UH3>i4Vg}nNhxd=FztcLZYU3x7GwybO9bhC(
zv%f@}+ki^=Gby$4m+g-Ge_Xg%AMV0S<A{7%P4nD@3S5Nry|VZB5_|tu4A1(x^+g^s
zW|(r1Uh~0;!@0wugAMNkiy}mQ<dDGUj#j~_U7q{-?C>#%S&Hz7Fo+$h_3z~4E>}g6
zJ8r%aIjZ)?Q3iE7ZLAEQZ0lf`t1X|vNb3t_7HKS5d-XqbVtbU*WCKhBew0>BPk62*
z==)kej=UOkT6A~BJt9}1YkYfgaD{Hx>UB;jd%iuW+p@!uDoSl$cacb?g={y$`r|Yr
zxu;B@ovi2gqo-vav3p+qag07Wx%SmsQMY|1@55fLp|WPy7$b8QG`4x9bva|^bmX8>
zJ!CvyQ;Jr(#hvY%<IhSMmC4?x@BNuqyWcTCuJZAjZ<nqlwhKy<y}xo946uUWZII-k
zL<IH*<y*W#9&;h(l;p+ltqD74cgq3(u>y@KAvUsW6Lm?WI@IP2>wYGc_xS&c!yW@)
zkGP~3x-$k!I$>zIZ`Qeh5-aXd@z2<-EHE=9Yg@8fc>moY_RbLmrHCu`b9ufutDYii
z!B3R^x1BP-dEY);-)fEBZNuX^kcPboLh>v$XJAXkv{=PT>!nwwxT0mBuMsV4&P!>W
zJjMiku;JfzI^p17b|N63_U`i|R3$@0-35IX%Y@iMtOqZbJ)Lk9HTLWCinY$@Q5+{o
zz>-&CV}oxi2_Ep4#oHf(scW{8(O7Lky68gGE>O_GE7)bg1iS(%0AGR3njM5b%gP}A
zbFz}1AgLM7A2;+`8$)_3XpCa-s8kU9_ufEbGTSr-(1s)ci-WKHDLE9zD&e|YpDefa
z$pzxR?f0MBH2@LCSl)0!!zVAj+Y$|h&GqCW&6BR0UooF9WwqvRf7$e%is_?XJ04Zx
zOvO0dG(4{Inzh2faIO=og;%fTikgjV-_dfnt-soyE)h5o_2v6U6K6W1B>eA^oI>-g
z<L!=Y=M~hvj=5(Zt0cL;**p+eze+s86z3+dzYu^MZTExVS=kL5ywGPz&<y8wNOexT
z5$jI#U#6r+5`R^<FUbdRj^G{w;+@&a+#LEGl)J9oy~`__R_6l!4Ibdw{^c$gN>UW;
z$1$b?j_pWw2w*UapUOI*<YA(-RfosAn7fMw_z!#f^Q6bxl%jH>^SJ5(eHSy+sCmtO
zDRlvCB;6;Li|fiy`9%%+gx}w{N;UyFJkMmj^W5kL<DDWe5`~NpQ(o-FzTz%%rPG7C
z_;{Ocuw2V+<0y~}fK0P4I!jqEo^ZW(jD7r4ef#$Oz3xL7Eg;YI>%TR!$KS8th=+B&
zweD|ct-;r%kBQ82hNs47Nj@?!JT2U89Bhw@tJn2GXmAjG!0$nPjzEO@ps;fWlZEYs
z7$avie8lZ)szPM^ET7D|R-AKbScu}XPyr-3AlVUp@gT=ZG=^I4hV@1~gltYj1H%R}
z!x01nXl0rBa9(+=YB9n1N#h|D7uNGr=WLqmd^(Aru@@m*diu$*>%H!$_C@VxgnH={
zzX<&fV;_?YJPqcYX=$n*oO<5h;>@Dy&Mxg`#cBui-pQ+@o3)rK#??$Yf^rYv1XVPx
zS;x?HgnwlftO^~J)sz+rFk2747@D(>K3g0cQ8M<MUix|If7Z1Bldjq1p6PaasKOfu
z^Uu_$v*&?d2hFxnGVeE#6lc$iONcJmcECQfvN`NOyn%8KyO7v_tCoP4Al0p1C5$t}
z0!qn3v+(o0p-;`ama9b5CwQFMgb@k$m95vIU^fvJz+#$+wqesXAE{5cNhdQPA|aCC
zA+jMz6?-aXYihN(WpzsUC1lM~dwoRql=`2Vg5Mnb%wqo*r6S0|vOH|qmQkXPNnUwv
zNo1(6wg=R)gzh>xTCg?8|7zw;*)&@v&a1pLRTeOTNVKzxLEb$4{Yr4rei;uoP0s3f
zo$GK1@HV$r?pw7n&0Wc&<)a!9`rnm^f6stO#u==H?PbWt_RuF8hHP%Fc@bo)D2!CG
zBIA^uH^@8uo@iEWa6Rqr0q_aE30kK~jRi^>XnKO=DU~r%Poja4hwaIkbVCu8q;vJh
zVO@_52bkP%Du~xc6Y>e=#1WD8VLcTnF#^=%0UaL}6B_f2|K7}UBCk1n{m8Fe9#=_=
zXBoFbHHpJm`M8axY!w|!9A~OHdz+zg->di(iI$P(eSyI2%Q1{~^KhY#`ep6oVbDyx
znAe2VhH7GLrwsF_3NJ-L2$v!~*o+CK95q!ut;x4N&Uj}_NSlVXG+7=y_D?Y|_<15X
zl-!-u{_SZCYkl+E_$SL!dl*_i;SJVwRqsmw$@cvOZxdp$dLlDuBo5M(<M8W5R01#p
zr;H<sI1g-qH)xPRT7xV{iZ9~2%PL2!kpQQ?^-9uzFc1b5OJfn5&1LJ2v5__1JIJ;i
z!Wj_OlPFL_98r*5)14PuLCVboc1HJX-Y7EnFVN(7WRbkY>W}BQ9Kq6mwaAc8jK0qt
zXD4sQ_bN2fJMLn%cV9YtQ={lnN_S{Z#L}0jk{$OZbD1Yrq#U!LEdr#nHb~_(u`oNO
z^F8l6!@YC3Q|o9QN7emwA6=)Ogp>6@WIF$qtNEKuXN@8Kdm!U&fTkC8b+<;?+NM@U
z=nfKSGn;p&`t7}*lcwg&feXMc*U8}?!?Bax`tbe1KRR=8BSP{^lo1z^GkM}t8MB0G
z$boq<#*j0@1p%&JjZ=;+u6nno8HsS=-oR>!USP)M@^a-Fh4}cdsW!&T2l@*pMf|$A
z_=mPm7bd&@YU1=%JUdUy0^;iwsln0B)4Pr+G!+%gIq<kmou4@w)Bj+}OZD+K-<<Kb
z>rYEG*g1bJ0HLGo(;gNa=WWd^I?>VD3)zjD(tti_KC&RaRKa-uD?U+v?LYIDOKH@8
z5n~&-nq~+dC@AMo))ZXf60P~d$hkm^RUTcoULJSjh=uJ&-^IjH+f|3)RQLJuVeikn
z8&GNxT?n1StoQ+l0UHRdyR+y4TM7@pYxdrY?W4XBGf=zLhpOrDE3+^>hHCS9bBq)E
zE5IfnZID_lY08Hh<i*W^;7oQ`Q%7&^Fo=TCo-?A%b29C^?}((vuzBu|--;+jZy7t=
z)bcHRUEek4#*=-O9cqhk?6{eiq2VB?;p4|*FvrrXslCSvx^1aTImL%7`BLT*D12Z%
z`910qdUX2e+Idhm<cpi|Y_TJXHLn-`&;1CmmQ&{zvQ+#@wfKz36byYMDyfr0*dHSt
z?ACLQ?e6y~<9Eex54<#spm3nD061U3PI$!^u}-BO8Cl6ruhkjJ?z7Ol_BB9;xw*Yn
z%(3K<iK&FRN2;qX0%PKO^m{;7=IzO9#C1yv8MD-<KL0M*vZ&cG=duu+JpE3EA+fRt
z$|!w%2EYA~-^f~4+9cTSSULKk6t#zOzviZD{-BVPL6I`GbG#wbnz_YWT)DAo`;L+A
zrb+OyD3vq0G=AwaEBq*nL|l!wmT%mDws~GfI4uhq$kllXTSbZp^gU`iAq`zKNMhC%
zbv91wCjY81w|=+g-_}JO9^^_xZ&lv>fIbMnh_(~yoeXVTt|-wxg$F(-?+<G2;3m`7
z+}M?5g8?PwHAlMHJ|ef%6Ka>^eDQD}xTzQ?NS3B@(F}~`AW;CgX54!?LIluu$OuZm
zt@Bg$KS0v6#%le`9Fb5VKzAb=;!+_`ch>*8g`)kz_xu-?<E@S3{Hyzt9@FeXz6t<$
zE4xxmdEDKZt9chr>nkaEj;-su>0K<Hn6jquXT0g5&wnVDUm~zBI866>O*9o8>VM1%
zu${Hr@7%AEJ?<5KZ&e;-9+|>FDP^E)VM;lY<kQ&R*;V5^ZZY$<dN0Ayz1?BFEu{nn
zm=sYH?h@~npJ%UgLU;>*iF@XRk(k9j%s+zyi^;f6HQrEB?2qWHgT()EBMR|fyd|N%
zS}~t2^Fi~0Dh%rPgtI<L$-(=V=zCivnM;U=kfDGOc9jpalnBN#BoCG6SlUhB;qA85
zsx3!#x-@>MHh#+%X4+xkv3c>Q*=@9ca_+Jz=p&0ek-PcdSM^tsK4GPdXmbZ#aKSr4
zWjjzMOB2Q@pYLjVshf_lR*1QZ5?K4MTJh$CvrPm9>AQYrNMUe$duphB8c57N8bqL}
zOkU5-7S(CEwZWY3Q}SVb`bTGIUx;$m%;#JDdk(U`NW?UYSrCZ~9S^jlP^sa0+@K5M
z7nf>ECqV#N9%+205H!0;ArQgG@QQ~iqKKzyZZ5Qp%`GZ3MHDqaf^Tz_0m@cXRwmde
zWW`8qp;~|l7q22TE%7Vbcp#W3mGi$^F7yv7o`~leH+&X8Gw@roZHj+-<@68#iQxyy
z8q93ZgbIUyv@CyZNBrPtNBpjOFxRlhq%VvssnI*HtIK)`2kq=>9JDcJt=xCo%*LmD
z-lKJ3L(P^7HqQuA<}qf$F%_dfw6~23GOH)2QS?jCKgi$Sd2OVnpAaQW8{0eqS-B^K
zCTxrPXT`QUy61ALt<BSHMX$jP8n(~R5kzJ4_u98PZ|u1QGfo-Uz4W@g=g*oRbX1c1
zJYXGS_dMW148neqs2ut*o3quP46(Be1U`fWOHvdq6hKBuKw5x=Fo~x2;M1cRh=l!v
zXVzN>g)1L*4X*WRM0)Fv$F0IcLf-dlbfL^^?^FG~EoNQ&JEjqk3fH=}O+S9}xk>G{
zHMuzD{Nj7^bqw{A*rQG`iKn{HVna^FNLkLcqM>!kAxrCss$QJ2aW!-KT+y5|&NKT)
z53kf@s4eXKkr@|U!DdJm+cKx=t4+2}DD5m%e}o$CuL5%WKcdrN`_l;MFZX@Y%R+xe
zXE0gB@)Iu|x@Qq_&&_Wm(^$zao1&Q?JJTie7Bw%0D6&q@tvv&Vvk_$|7{h@VcROhE
zQD*o8oW)36)Zu%Tk^Lfk%NaHPLJ%cCco23;sU(rF0AI%&0$c-De_TW(iVP7Ec?YsB
zB`C9^s#$dvq*bI!naFL(D$eQdA&ogP`j?H?_N#*;PgiZ|Iq{JxZ9PGrY>bO5g)~8$
zZ1id4Q-cS{+R?JTqYuV@37(0}v6y#l6Y`W;SzXo(G#te@SV79sR}M2QUE$s#490l2
zj^R#(b{iFwma1JG^rXd7#y0)md;iDQf$*vNuyD5*ee>7Pz2w3h`qE3&Xt(qGvm~1`
z_}~)PO0vFyCcJGww6nrouFjTzN!c+NHxx$^yzoe(+<1jb3muJmbv3n6DtzyZ^z3?l
zUD&?-IN47Hn)gQS0&p)-$(K0PE+PszxI7`*BT6%(+#oPn%*_EmF=7vj4hT!ZtzWAW
zt2mzMDtD>X_eXc`or;|H<cHr+{&DH8E|zn$Q#QBC-PBxkwDm$L!{J4y|AoviIf-4<
zt2oyRY_8B{EhUP4hNi}ap_gRQvm%kIl{sU{O)WcGL{vflrPr&Y74cGSTOh3++P_fw
zAK0j}b!$LiNN28ZU&WFVMvYp3)cR*@sZf{qDC4e@TXh``bL=K?3qtDzVifXTq#ww8
z3AwwKEW76L^7H?XvNr*za$Wz&mx!du5Gqm<k|JctSW)b>&}>(RBB>;jnUyA_Qph}%
z+B7SQLL?*<+qN5Io{~%<^RRxO=Ur;=eWw5S?^>6ubIv}U<Mq7vbKjroW<85R|A24<
zcP%kqkWW+8U(XQPh6k7%B8^)5hZCT0s4Qf70R#fisNIhG;^8~;PNAu2lSI*!WUA`K
zn}U*~x88WE;%_~eUr-wSF(+R_Hfa6O#77C4pbSaVbh?gv{ZR;5mMQ6T$&SYLc-wvW
zpr_B~#NH};pxK7mLg@3_3`p@p!<4%-g_5o~7ifnIkBduO2t7N;;*VIW0dNzMsdvv^
z>dL*(@>Q{#W-ox-G8Rd^YqIAxci7UlwWk;C-Kw4><G2zt@Cs)8@|4E=!HlErwO<qi
zZx{9R%|~Jt23Z)s@IfjLS^Bq8ca&de^vjZ;v^qglIN5vh4l>u0F-l*5tTa{L@t0Kl
zXyvV;g(>KhN>j1oh-T;P47o8rglkXzBVW=bC{8MkdaFPF;Wl56z5DTEqf^J-xYrXA
z*{5403aaCwS10@#nW@8%Rt+8ACy+Lg!Wg!DpYLVa#=csg>l}H1&fwq*hJ_xJY*jfR
zBVA6~B2#)(pHS;>J%tVb&8BoJwmxgTX%W-wR|Tb<_RN7PNzaU=t?sVGBmsz)vt;da
zT?bl%LQP(V&cNs&ntwC`PXzAFC(T}V`ge#q5mbq_Bp9lYZ4LSVp(o@-z@|hEVRk>N
zbt6U$0<SsWu@<Fx64B@+Yl?TrIDuILJ*4P45d~Nz3if%6UJW1WD)(h4h!GCUjRz%~
zS-iy)%mB^pRi938j3np9(V)S_Kyt(2#@CbW#gpq#@bw<&m-n*nhq<QXEf?PijkkiW
z(d~miR|A1+)f0xBBk5*kR`Zua|Km|#p28DW=(~yZhf}aoQE_?w;u^m`)g61D%RY76
ze9Y|dwXseY&pzR%mgTA++t3vi0GVYB>>gonefPj_tI70TK82r?WE6&;wd}vGRrkz`
zaa+9oTVO<QoUwGBlswQ{TtdiS0;bQZErj_+Z=^CG^0_Iqmv^Bee#w}S)-Ub$ub`AX
zAyOWXaA;7@F-@>^jrc=dHB{8_I}?a@%d%G`j~{cnb+gfYyHc1|86v0Jx3$iQ9W#5$
zbSJlC*RtNPovYr^Hw#}$?bdt{1PiY<Q90$AMtwQ<E$dLCb{qD*`lUuD#=dGY^w2Ra
zv%1Jl9>w1*lz#^<2zATj1{=COS|qKpEUFnij%Ki8c#CFbqoCj;tT`Jx>6}~{6&TV%
z4H1~7aAQNG>QH@Isx8-uZpP}7*X(KX(ydfMb)*b`b%HyKD6Jtz^_-MWgzLFxk=+j!
z6<0)vaak(V2%wmwW#V0;U#cJd-Z9+mpPZE=WJr@Td?-Z&zlyV<EM<&|dM6_TWPVdu
zqLumh9*&a18b4UO)!25C(_lNRB7Xl(<HyYLD}9N28oF))h8jUt%St3oFFo_JSU#LW
z`*#<S>Q!SU7Ox}1QrDk;$GlW<n)X5G*purx4n7y2xhBlc(VfU013WP-qcHLD4?K;B
zwZU(LCKt>PF2c`3<fxQD6@NfXF4824alvH`47m9IyGqv|3%yNQt>2epaE4c^JSSp2
zSgy_|>V?}Xn0DhIAMk{Hg#7nhv+xJ>e=kXI40;h=F?wV0h!EoAJETpt0R8x}%QI`0
z>L)wZj(vLm!agOnV-s7$EHeP~s7VW2_^z%n9$P9%k7!A_DbL$`oN===A*s{u=(vfV
z{(_}d*7Dy24lV;+5pqm}UdAp<lg-@7tyiUesWxNj(2)MaSO^UJo|stts42&uR))@#
zvCNfy<J9+QY)^lL#CP4G6G~sMv9}HkrlpKjt(lrBVlITT@ylpBM(lm82Cv@&B$!Nk
zH}c}Ok#I}fxT%NkzcE`Vg|dL6Gl7Sj%Bvu$KPWH|j3eF$&k#>RnNKyiXg_^3>pW!^
z7JP>hiREg)VkmX_m*2>To{X!Bg$iOZX6NRuIv6U@DFYHiv&`z3fsO5hT>8>Mm9IYa
zB{GEqG_856z5R><?OC%<e28dzWR-t_dT25GUq)TN#}@h2pRz2|_SRQ8G4AtOHPmfz
z4dzZ-m6m5;qCa*KWq2^gT{XnL_>54jDSnGk`*K$vG&kN_W~E9O4h<85?v%_I3HcX8
z{T_0a_|L&by-dhS>0xtwaaQ0z<fw0tE)A47&y+N&Gi8kzyH;cB$iC~da)k7Mq!#1V
zvrdyQ_2RZDO+|}`{*Hx!k(vX96{0v&aLAk*7G-2kT}k>bSyky>&woXvHI+@dHoP^;
z>RIhkf>6IZL>5jfWPh86msvyspdkbkQO2)$VhuXQf(i+iad88PF`WqAJjUo~O2c2)
zN45szZRty2q@K<)2|;asE%P<!iH|>HO{#{Q=<(d1+&E^Dz{~h|b0~Rbmm0SoN5qm=
z`B+hN<L!_S7p4TZ7cW!4;w)CFAiR=$QdyhQq5kl4|7#(t_E)cpi_S)vzz>QwXEMA$
z_jv5<hPbSu-q50iU7l`hj^KKs_LpPY#tnnZS%H2O?eV)qgrL4r0eKt*ge*)n$^KNz
zLe}%q?=M*1O)2cNdo(oTF1cY333wn3>trE!OG|;5qytWmYRp*}XzKkpNb-M7e`^>;
zqbVfM93Hat70m6wkz3%%Y%p<!b1!ym<&u!&YR}$oBd_==x4pNrWw%&-Kb+YUkeRzd
z-|GTqEMLLUu;!;)M&@UVn`Eu#%i7gcOE-9n1-uN5vlIIT>LUeHUScu+huF;~vbjZX
z(BhCT8k*ezPG=WB-dWou{X_6JerG#->$SRsAm0eTzOE-<*!GmB0+c~P2E2pspSsf3
z44zR535m5w$-J&E=D9Z2{h@h?6nahS>-!$i`ji>LzzlaepfVh=Uw#w7cVt$8T(wc^
zU1yuG2n!%o@%Vt9mvU}z6wYj&Y_-cdul5T|*QR^9Ix9!4O=q)_^PR=7$X_m`)qDQi
zPJ@zjJ(0}Svrm4_nS0&2;Q4IaTKL}{+w8qBI6o=?G|%F;8c-HxS-L}9B`~MVI_}qZ
zpgW-9`+v$u_^(g~%8p-q+{n()Bxyo{(H-Nn(qn8Pdn=I`;rCpcigW!u0Ri5CFDPUb
zMMhmyleE29Apk$hEKXA)Tk8u`?%{xd9k!yJn5Jgs6E%Z;DZqFNTLq#6)*f3<^N*d&
zG`L~LqH@6I(6=OQ%5nWJm%|*e%p;xeP#N>UhY0o;&vtGNsd0YIdCIFxjkEwMM&Ci2
z=Ia^(Mqd+4EGE%e!8vbvWD*;I?|j;dntVPl>P+kzL2=mkUr!oVF>Gq9-*A<fsmIN2
zZkf$91p#(l2vz72I<G$OX`cc^kSTRE`XEr{@yr1TB+D87y8|`9iF__za6U98Gkp8b
zGb{d4hU3oihCb*(gO>RKOHJ`Z;l)7}X}}&*RsqJ_Qv@Ci;nPDQnbAqBHcUZZdaPIz
zF<xt6pW3&;wx{5sD(n%?O}3v8u8Sq3owX_L-wcOm_b<A}!OgF><J?>7CY-j|I(?JS
z)~jtdY*uNbitb=DhPSChMk3o$RYFT8-E}};NY>ZQ__04>1oMu7Hl>$kZ4|E5)3|t_
zG;~zw-?cevwl5R*)t#14{vkr6hz9Y3V%i)(4*7YqJuB%s*@uq9S>fpMu4l3dt^e)>
zPoMU*->n4{ZdX&1DUb=P;QhP5!wD<VP+8H<p|}#H>)j#1o3xWd0-1o5@QR}s5!4S8
zySM>FJ+35FWuT|{0Ax>;Wp#8cF(^wF3X??a9pi}c$_U+HwHnhA6U}l`HfnZWJk7D&
zIY~}=JDzSL%`h@&o%b}jk0iIfS6r|x$kgLPRbUAxtse~O{m-ZU--k12=?k6w;+E4N
zx7U)L|Ne{FVW9+#V8iu2O?K>3$k<_m7t>#kv`9>!qfLTYEU$@CZhS~S|Fanfc`XNx
z$Fg^;`=z>QnAZKy^5M*vbzDQ&q1${*(HWbaJ<wae!bf_|9*9vOKfomo7YS}_vI~=s
zF}&V;xY3`yOHgkh1q^^Pv>ivP2YE|Vp|YpcKW5k9Ba9d2Q8x*>ilmV{_w1wHD35E?
z^iniHl`2$z5#aQUWaf(6qpkY$bHDY~pPjydXs-ua-D7uN-mM-_`N@+ypXK*GpX;w{
zq3W%h;hj;GJK2~J#n<EGQyO8S-w`rwwX$N;9e9?z>y^m(*nFt(IKNFhsY#>PBuG5V
zRgjjmYoKpQk2BP;`u_tqxT$E*d)mZ1${qv}!_ZsZh=L_%<GRc&rXUIY*%TMZ>+gPS
z2Xax^s^E-3b_nQqd@x4>oncY&fFF96x>R?mZn#l#i7CV?zb*@qNdXC{zT~RoXo5kx
zqw^>IKQU4oIEM6*77$^6WH}SGjqs~NBnqi0sLw~c-tZ)_oh!|#W!8<(Ajk}Gtp^IW
z=cW%CHD%?@L9O}ZHd$P}Z%f~d7jv)*k`JpO;eFF9op2J-tA8Fh%c^PN8W}mkE)f)D
zF9`F+&I-z6HHU_JuM<E0g7+Ewo8za(`jwI}b<_SBL{`DcA$JzJ_rv7d2w`Ur)Yv0a
z!31zb_w0_HB`+e&b;8YOj)~oj{Xk0zEz?uOqGif9_!lb9x^?I4=1?UnT^nbKESti~
z0WASykF+(O&v_|0%VqD7v>A1R`1u`zbS2UK25h(i*(OPJKo$lCl-5lVd~IdoHP4Q7
z;Xy$P%Gzi}c`AkRxWM$0#}z(Np;W7rdK+;L!M$Tw5gHXXj~XpxY5TT7m0Uy<dULL1
z?@(kgNWPxZ?&wvqzgHd9aX)yppKq<c8ps*h7E+DQ3eGJ;Y7+WpYt=2jY-K{NN(Zp>
z%faou(#aXq@~GUXw^e_cc;|d3ZhS}ACKRf;CkVvo@R{?bzT!7MjRNQyL}YZC;pNqB
zJ1r*!d?!LJEk=*HxC2ze-aRnFckYm}5RmLLwfVD2G1_t)TJ{&INhW91HdlHD?je^7
z%Ew;*lsQ5OeF`qR0-l(nidi7lnFtF{ZtPb=EvkFm3Jz3ELIRE85(pNq{^H>%7|5~S
z7IAuxcFgHGMA;psk=FU=Y`N5KlK1VX_<~AP^W%Ke>F)CT*KB^qB|}HM%T;NJc6Ubu
zed{?v?HDq68O$_$T}bqtF01P~pi7MP6BKW2Fg99gHDA81roNPB`0TJidVhUpK-@f8
zF4-vs2Y~ppQ`PXF624}-My}gP;o8Ze#}+o;l_xr&z;=zD^#&*^6;80er@XDJ|IxOE
z<Rqz;C@A=BSYr%r9H5Dx5-Aj=tG^V+B-TK%et!oPcXrckb@Ye@6kpcX)nzv3y>e?&
z#(;>z9T`Wz@u35AIsmamio}9u*a91a>h8|6rQ&DcbE4$gWj#LJv~Q&;zWlltp!U4_
zM3bMFAcWxLaS3~U=WedY^C&FztCafvklD$WsSVjPvl|!M8#mj|rJuQ%$9-k<&|rAh
z<R2oExZcW5af(nG&&Hf7z2YoPzr5$9^0F#MfT_mpMYfs=e>NYR+#URM?0%q=(hCF{
zrKRMqf6CtKaPO(7*|DDJhQ#-JLsjkHTUHbl?>s^hX((SC<5)%rO)mt(ZbTiWe(1}d
z8-rlv$+i%NOWR5^e6=6+zODR+^&GzUqY`_r+Rj552n1#L{4D2q5xZQyq^&@b>@%4<
zREp8T;oG+tQppx9-(&&WA*5@~Vy6=C3Hzf5DguOrl)&#7=6&)?C{1YK%N^Z3{=R;4
zRnAyrJN1|tw>h}e6?PnWK!(%hbFQa~z8B4Zd{#;EV#ta5;MwX|h$TS(0OO%NZ;p_R
zSpYhrnQBa_R*u;A*ZV?JBjcfS_1YTdAhU&@dAQ(v%jzYzlt{DKAHwr=*s%V=M#NiY
z6geqL8HOBB9&2sShwkk8U`9`Ek*-I3QF!)haRHJmlR32L{+&B8UXfM@szahF0JKHz
zxQDl4)s(UzM@W`+hw*pu^!gk_)XakpV9MwR&sifm1-<3=?N|>AB#eg921XibIYDkR
zm4EYWx6K}ES@rb}UMaR?8ap^;s6*1He@oti+u}?~qW>R?K4(1`w>pEb76#vcmZ2}S
z5ADUvw-U>ek4;FmPCJ?1<(&#>2e}mMS{La5XnFo8+X+j|jP4%8Qyfp7`jV)Wwa=e?
zREH<`97re^=_ZqtlbYA#@8<2taOG~UWTAJ6vVvOf(34NbH@zWF1@gJ!1{?TpqvEEu
zyQfZqPm+DK+B`^j4T%PC8UT!CxVqhQ?W$E#su16?Ca$jc-PTBg-_X7tW=wEpJ-@7>
z{3vySoIle^1%Jz@M|YE!1;K@0JgHt4t*L+W$nd4tfblkw`q4VW%#i%jZxe;CN!etC
z`W(C)Wi2+VHfuO$Z4HB}$qWbcMadN<c3W0(nj;sa;KU#M4=K3rSaibuw5f2D81<?V
zhS8^fIA>?4&y!()p^(!zKWE_fhX|eJ;tNYt@mnM5FO)nU;#^R_tlNixztrvBy_p_<
zvpwRr>&F~XDb|z7WP=LTo*2)dosW6^g9kqiyi;g8WI3!fCeV1epx{(6HT)-{6}Cz0
z1$P(QE;NLA6_8q%B$n?YkgBT%a{JyG@7%I5+urxOkFQt;%sDFr3F@~#-NUlnZRYFe
zp%SL|@8@5M$?>g;S@^>($cUf<3jfW#@}GFjyh(A4-~2rT+!J0$zTT$YO<^GodEP5E
z(C3vwN>bcxBLD0ZFBV`ZJFTK2&I|f>R{k0K8R)tQ*ns80)eh)nSlI^(h6f_akAW|L
z5(+nFOCY~!HSMZYc}PkWw!n9M=T0Jmo3T<7+531rPmj0p!f}TPFHi==WPLH}XtE<?
zaDjT4MFR}h1HZ{jtzCC!D*^nzj68@=>I@EB{pMV8bdM-|q?PNj=Ghk8;{(zIuPB+h
z8?0I&o4jhlvbIk9d-CgS88@|bjXVNPST71o{l4F`Pp6=GZP)X9r#@%9eX+<hU^Lsj
zc-FipgjO6sA@q#OA$eO^{Gs0Y>S4QU-kO0C*XQzOpXWwY=xlVjAaV;?T`fy}M9sw*
z2+`%w3T4d#U#`J?Ax`ON@X+9+13f}TQOz<Z!N7m%kH9T}rUaUYe}<K#0aywju){tk
zD|2$=8Ir5S<Lf7D${Ug=`*W)YZ1)&VqxJUu&cdg5hBQ1j7LN5_dLN{jsj8ypUJQ?N
zN!L?nCSUxc_KCjGI+TxfUx0%*MXhsEPuONBX{$yblz`F6-fgp3rQV3fmA>w7=^49Y
z1pYsrCkxKo@2oFf9LRLA@^m(p*uP_T*lAD31J8+BVU98YqVC8+TKk{_j{&4LDWU!n
zz5&1drbKk@kV{bFcln^9soGg#o$64nNwbSirK%@s;(l~+fIWfs0q`W8H_$tDJ@5ki
zu7Ix<Xcef*pL8V4Q8xc#_ppg^b>Fjb?#YPqiCGZZxPORnUfTBC-le;`c^<Hasp$nO
ztp=8_KOLK9Tk5lXU+aibNz?tWlb(M*7w11a2k+HoWB{3Qq&cOE$X3(O$n!EHwP0@L
zZOe7!mkHX>MmOOUl}T|M9QeBt{~v|s@1B$FQuTSDdzHsJ9bpxFcAB}TFDNDZmEK6W
zfP%^Lh3ew7I9YfD<n2*v+GK@J#X)!M`%BiE>Tbp5m_s=@LaFj#)V6P5Y$O2i0c;~w
zO3u&3&KEYJG(*A0+8wjyHU|>okx_$y8VDhK()O>?squ;#@4IVRQ1vvpe&Sn-tIx6z
z=k-plsCsxL5MCR8{uNEZVGinXp^<lRm3Bme7eK59O-TOWO+vm=J#KHgp8lpLxBL2w
zi6c{1*o0p8GAa@xBuKmg%#`d^nAZCGo!x#)vrJM$T3uhkWk1s(cGjMeh}=MifZFEA
z8rmMdwWCDaUH_*3%Ivl>C&o3gKgC<tdKY|5MjDq6)Buh?Uk4UENb)&A+2<j&{i#O!
z+5caHrWCaY_mFwG-FpGnQ>s0%uOu}fe6B(7zE;#G)@uf-DJYZdZ!dXFx1Tt0AMtJa
zk}!E`PFDWDa=c!5jPIN0h%J%V9ty2^pwRY<fRsKZwyAiB-Jlf;=h4Y-87#ugy>R~;
z7@Uo&XjXCHP(Hz;s?cU6q(t7Sm$1SKhE<XwXkRK5fglEvu`a<+z4r8fviLs%5tj>#
z_g&OAMShC6$Ggvn9;kPY8+xhf)|2umdho36=ANP}QrzJ9!^u`u0>m^my^M(@R2$q|
ztYygf9%bz6KENus`i*wc(x6@L+<_1cv3T3Jvw|8BOMt7n`q%L32wV&*#0hcr9(cT*
z`K4uL_rx$i2}ogl@<{I}efEdSwmSc!R)S@ej-M70ISVvu7~*Va&a_9VpX=VA5tKVj
zZsy;(%mN?)oyn$VitxV+p|p^6L5ZL_8mFrit0ZJ(6frjDU-j$j)?uDo!GbV<rX+rl
z|F}(Q9?AZD9`VT$aYpOsBPr3<3#B+C+dj5aYr5J;y(#+9IsiE-2umdG$>)`i`6%yq
z9s4yf1QiDfJpf$Fef0Y>mH-sr@p7O7CE1&9zZyDKhy!+z3%zHE4W97*N>MLJv%*oG
zj$Y4=rUxb#LK?E<-SI}bP3bUS3Uk_D!9)YCdyGzT7)`S}9#7tuy^=}G*y%k|pKf!+
zO8=r?GN}hI-_3jb<}2^+rYN>DH1S%);>#;|5s~|v`Nwo_+A^>COYPWNXf3rF8bDzN
z{qfTD&1G&r?)h(hJO%4cHSs*EGn5AFgVaaPhjux8rapZRkZkC_xE|$^A35$gdAwMc
ziR=Xvi_u3k*Tjcu4hyJl&j7r8TJ4r!or!4r-p(rJ*tods7AXLEp`V6U<YRi5w)@wF
zf~o6SzdX5P4I$q$!ljDG9wphMaHC&=i^xQ7rO-_1`7cY^e~2*A{PwNx1G0jtHzxk#
z7-E&rN?x{FwT&QDrBd~5_{oI*Ig>-Tz#f`*>?JPUBbkRCXzI{h-GJsw;+Eu0Hoys4
zOY=OWO>5$Hcw{fyU%=(eyi(X|1RgFvYbcQNb@y4viiJ5Co4}ox?VhZNX^&1M!jyPg
z`Dt!ZwU|uM>V1d$TwMEBA~R+OrK2XZA9U0+t<*k=VD<wVg*k?mWr*`p_q_#x3zTXH
z(zp%GTz#HRy(?aCbyvYZ>gAUzWdt@LRUVt1G5NvG21eRE9Ly0}ii3NZx{F_0jf9W*
zD|pXjws){%6nb{XBL!HP)Kj)_@i%vCaY8EA@~>YwQ5iN{eebrJ^jjq6L*{OG$^uwy
zRlWP~k)leoa=N2bag?CC_h+1p?DBS=OL4JWrr*Rp_QWT4@`c@~beQ*li8WO1cWWsq
zE{m&`G=(ksM-5R5U=C90hMGg&OQa3r8T&tNN*{Z)TkE8~xiH|!=(kBCQa}xe>P2oY
zS~lFxAUdEAzEIS>*FuL?RV(TQ>qFu8M(cJ*4#KOXr6v&na<YRD!3Sjpyh_gpmL2(W
zX)B@KjHSU5L7AO>ns>MA{Gy(m!Q+rsDTXXsEct9;PBP`HlJA|~F|N`o6l-5B7ki*Y
zVwpm8A_%&{=d0BiBl<3Li+T5NH+132M#8jna|=;N8$|AL4!$uRJY>c5y_}W1sxF4j
zlV!gU`ayHd^DZGD8;~yQ*9muQOw1dv+J6Wd{1?+XSDKhu>}0hn`B2PAnq`B##`9sw
zoUE0-M`?QM*b|^63%AANAXT<1&5o$3p8PtHB*RCA*!DC~$+2kl69<NqrP)~`JCYrT
z+!lPGzr*)Oy9!|K=&pHonm}`eOM{S#w0ii+3T0PI3vf#b1R@4)vj2DJypA+ImkD<I
zbhmSK%LAMIa}HbsYus@(UG(H(<wr#Fiy1)YK4`DWxxsUd>l&Zcm%1IVR1Umn6OBzW
zeykY_K6}aF_oXuhc8SS}tXd!?LTjn1zS(GO0j}}Jg-~%yW%j>p9hQ&sV3%PDr(;Gm
z`|2%VH?E?0SZ<<BShZzsAqo*X=iIvMite`W)6?7%N%=aUZo-v=S@^m7$P_W~?hu4$
z;HQ5-;d3B*s<-36UzF}J?7s=h7m|>C=gvNr`9wN)#qb(PCqOII%XYo?Xfu2Zz@d=R
zM6`{p&_^Wxo??-+C3+*m!+Rz<7{l)zn=&Q`gL`ZFOpuGoV|00u5&bL&odU^k&q&&Q
z)!o)JU)!ZVc%m|N!vAd5?UH?R%DFX*KJlaQXny<LzJ$9>W?OHu%5)i>G5~bKMTvUm
z1;v^kJ`Wxi7VCY?jkaQyp1OZ^buwZ*Etg0moBZL-W*fT%OP_&F!@1slrEFb>V@Qv<
zQG0Yh$VU7~TW+%$>p~K4h7VEPwf^W*W=bL?pB?@pXgfosh*$_uj1^u5*Z@h7_?z~u
zTwkfuQ(Q=;^u}%tbBrQ0ko)uRN(5axvAu4fRbI%CWA`-qv2#O~c+wS%1+IfRef_bZ
z)zG8Dy21I+Hwq^MT?0TZ9v#M}^u{wq^AHt1xuj>(g45#Wk)m4d^!dw_={bRa!wd4V
zy#|-ld2QgA+8^tk#zG<=M&&MHMlS-BsPk%1Ryk*V!djm_!@J~y$$8jWs=)Pd$=NIJ
z$F^#T+E=Rz84$N7^kUGZ-l+JNB!kFQa7U!JWbM6mC$qGa_~r4HIRF{(gANP_KayAG
z0C&R&puu%a0knqARwSCYaQNFdhyI2#ZQ6^Ji2;eu!F3N7uU)6_ywLOEV8az%{7s+K
z`IiORzn-Ld==_qd2*G^!^^r>fUJE2YvH6Pp^4QifUgtarRjB67=*@jnt#_xjo>?Hf
z(W#W3$|3XDBq6Z_#FUxxynG?VJ@to1X1}j4&-rZE(fPQ9$f5j~(K%1?gk=cGC_G*J
zu?eA+Rhn+Sn28NHHnVm9Vr}zRZZ!fbaewT%+c;G_pum)tMUCz8qZ^os{=kaPUcirz
z6-Kkzlk7b!!(~E?jc;PL0Z3G2D&iAT+;S-;HoYq^&TII6vSks8$r|{LOnXM_&QSfj
zhW<@?z$I4ZVv~rJTxvqQ(5Pa*%1wff+hj4$d9_TEY*8bA<UH=n%Fv!l?kIbf8-BvK
zp8o8x{>|*GCW+QTEcF2^!jx|@2(YcPO*RcBx#OG1L!a6wzHP@EgZIHC73;v{CE;3M
zH@`lrb4NiCcvOx{r8yz{`+l~rU#xA-g|}@#O9D&@?uS<y4H=<-KoOFb9ef#@bws3Z
zt^^lGc`gBc%$d4h#4s+(o>yB|mF^&Vj&1J;Vz4MTc(VsiE=(^we@(1ztWKEV{CbCy
zmf|t6CYBMyNcuM6>&^lZ$#OyM#do}nKVMQm`Q-;<G@Y}87k%lcL^Ts2MzQ@{!zTQk
z7gp-w*64cdwVPkou{8=b#m+)jRQ$im3z+I7cUG_EC}A0pyeC$iz7%S?ye1D#dGpah
z*=upj-V~e;M;c$M{C*_J{CeNhDOH`U0YVH)lresjptzD0r)r?t(}14Z6MLb|okr{N
z^1xNvu?8|`Fz9f^<5OJB4c#ar=Jq{`RI`ubttVHe`whJ9Q`74?!kD)R`t|>Q;Em66
zUO=A)SNc9`OLwOf`eZCgTtP%lOR1B-W-W;0{wsYo)<N8*_UcIG*d3iHmaeBkF!k<f
zz(sdQ#(Td#WmTs=HuoP+2uInZy*??e@9zo9lI;uc(0UKq{VpMrGg5LGr_h4WYUd{r
zyrHb*Mu6%pnomSTgb0p5bkz0C_RrsINs?m8*9xN$DTNKfCMv3A)xtg$yKr05T~G}M
zs@@dMK;=^*HwQj-gr9FH<?4>GoRIX*{_aH3QM%XFy_RL`B)5fkF5qolgll*GVzl0s
z>i8i6{PODMyEylWo;+1$XwGgGkGbrf_qX3`K1-Yo_{oD-<Q-?XWyC^dKkcOCNddBP
zfz7pIVj|XSEG$ds+-y$ZK|7JHx3+5CA+eY3#W7Ie)Y<XD@Zk7loU@}VqbaQRH@9DH
zYe%7}6_;VZvVgOM*ot(AT&b-d%&%$nFPOu<<<lQ3pS=<pZj&WZ?obVB#M-o~@6s{}
zadD09mC5ddo%5i}fqw?`EI=Z(U&+!_r=x4$Dh*hNsHhW!rAdbdsu_`f)E@@-;O*Pb
zl5G{oKkH1_-B9Y^Diu{oP|8=~3%{CHjo1}2yl^ryVrN{Pa4cjqPF}v?gV?^WsXwKm
ze2hzIByJki+Xz&Q)@k`j=`0?VksV`i)rwUgnNBOB(PQ&<oER481Qpv9<1)o+{~S2i
zyMFBq1qwPpYWKrp#fpZ=!(+P7GCss`zfKU5adh+Svo0(S8XRFv4C!9XWsC=&bWK)F
zmFGc7CBT0Kbm3-06ZLIcD@y<gh!71XfH3f&$o)4kvx@c>&|n>z3Y{xb#hR#pfd8Nr
zK-eOJ7ZJWiG}Xbu#G)r+;m!t^F-sfBPr53~@A<r#SQ`n27dB5$B8hLlZUIX$3I^}2
z{8N+H^#f;7Zp=g>DVK{iDkh#;os;QaDn1(*ptITlCXaR><B81R90wy3X73W<?dSZb
zb`vdx81<5_J0Bf=UuL~#wMKIt-PnX~wVv3mVm{V|RW%INx2by!Y^U;!)$nh|xhJ4G
zAVCM75}o@zQ&swfK3*H#>pC^pZ)LqufXUb}i5nS0dQSD5T3&;*AfP3VJXXd5D-acP
zE)i+PO*#3olI1dE8g!u;>oMS5F)ZK-4=R+zu!}Q=lLb`|fFVRkODiro2P7KdK+iwo
zezW#E*0zv#Ms?dIJ?B>oyYrR=;M6>JN{mZwG~S@vX#C52HshN8otoQd(O$c9PLOep
z#3d!;d9VQuJNevWRCb!3XDU*$^Hod-i;6wk9@b`9JjAM(m;x2ky4Q@m);wYB5tY>0
z|GlMpu1x*G!qq&dJcc@VPAz$fD$Xw7Ep|@bBkqa6giO8n57aP!A#nG@XEmt`(=bcu
zK)M6bYH_mTl+5zak6Km&%1#sS4Sp{0y$cI$rX1?}F6tyg<mV=rsY%3_y;sW4$;(NC
ztc{Sn`l+?GQaI3b<iq4BpR3p6oy6|I5}44`ucx{jj7dQ)&Ka}>$qq391AFowLM<9W
zi=;?P&I(RKI>bYS&kRk(cg@vRG4W*o2t?KfH^uVbo-?$7#Qx4dUDH(P1&IM?S*Blq
zWGFy9Oc!qPV!T`KS=Z%=MF%?x1XhR6`tIP!R5QUAclPBvj>4wj_N*v<>8`Z_Mj@8x
z9K96FF$5QSR_&m)EM^g7=}>XzG9op>*TLPh^7W`K2gJ^>eJ(5u=?)Mbks&`ygJLq3
zx8)ctta;|oEZ;e~k-%N7;eDRK$pDUBgF0Fd3~?ZKFN(e=8k=5HZ)@x15t4eT%CPEO
zPw9EqYNW?&0d<{{$%A1@XYqNHb2)ldL7_;1YGZGGnzd2t77q`Y{WBAnIR1**404-^
z)6j*UDvp&_cY4e#`>j)-(&1RWDhrQu&QJiT{|VHu?wOkev*kAlo-f8i!W7IInbf{n
zorq_V<w{KztUJcyT1~r4cEoXUI4n>26ftI!UJ_rQR7n#tQ}RlCz6k<0A?QPYjS@OE
zk7cPxxGMzZyGB%p|HI^dzz`$Ew{9LpVkPc=TW?#~_}F~1@rt?n)~g1P#oo-8-65<i
zojAL}EmWV)v;X@Q(MflMn@55IizQ7hR19zxt#S)~k}T`r>oxGxZou*zVll%T^^pTI
zb#3ZZ<}7P7Cahl%Tx=09Wqjgk+W&@y_Ek!VSp3yooK!Xw4U9Mh0XZ5fT(`*eAiQ-v
z7TzJWDO=H;xCt*HCKGx-BLRtzi&+mL(-tqpojVvJFgBnb;z3Z=MPVDRU|ie?>VxzF
zatHj!AwpkxPHuT^bHi!8`>?LR+Aj)3^niIvS+aYgXz=X@FVxEl>!{$cLYmP#w}53O
zYiRjykvzPIvm#7fya?j={)Atud(tJwsZ@3So3Sn*@Xv6G$oi9fa>(Q{mz)2~c7gAn
zj8`XzcCWViv9$=atc}%t8HF(;1(wNROaR3|&C+9b{pg<^Y_WkaOH$>r{S$0S<FOem
zCB2hVqd^IJ4h))PX5#e+QiWnjglg`(JX80G!vS+)L`mZ%O8LM}qwJYzF^3Aiu;N3-
z@fe_xf$9S)d#;9p@>!95MGCp22IKD7f`!FSv4e6@D+Ws$b&KDm1RzH>D(nmzU3M6$
zS7IRm-6Wnb6x*6M6`|Xt^bM`kD>%tB*-@bFUH<2zLgf=sn2$?gU4TaGy|r&@+g=T|
zz`67erj!4hzd)I>@otT8|6d(<NC%1xi-;R5N|^EGh!E$#KOVVXE{8jt%HU=7R-oSi
z5wqMVtQ<qkUZ|!3ez$L*Co|P?)b4A_nciFS3-yxZQz>i67!sX@r83uLXJYFDA}?-l
zdpQMS06$0M3H%1|Arjoe3MB<3i_zNA|1gi!YR1Br7SB6zP?=zYnXUTQj##T2#wWyU
zF`Gy7j}0lMdyVL4tpBcAKT`G}#qZ0aY>!bxt@Sc~_ObFWTeMD2I~md=o9W1dv;aV)
zYSBqz>CmZpfy|^<I{k+}-Q(TryHua9(V}iXyc44Wdn^0?>cKbg?Si+))qx2d;vuwO
z1PlG%zSYqai3<I2m9LHRBXBPH8cALllf+H<Nf@^Bp<<wvdZ?%PSlrVB-A}EWA*#rC
zUQeVMNV&de?BBegDB)!rkJ~qqTIJVCIbiY|>N4Ii4w0TLZuroPSFQeAyNDt!F5%`8
zwuh4ge9SkVa2f7?dL*&>ETrP6nm3fByNA=KRYnLH<_BIbM0*>Q01kKQs*9mJ<8S_-
zI(M-8AE6hzb2uC<+^FdBE}aG5jHCU{FYzi}^y^~_kt+83R!(G9=Xcy~-MnegWHUt2
zEGq%AmO$LPeeOAH_sU2XV;S(pwZDENwT-E7oh>D+C;P#69y|2p*C|^r^3SNPJCIX>
zxJ`Fhg2a>g7d-{9wkvA?jpCmpfp850t12}y4iYZqzi4K%pRtvy9>jUVm9SBMJ+(gn
z@jImqoC$vf<eAaXqC!3;t8hNFP{o#C8ldd~#W<2ND7qscEdPD%-<OyC&;0)SCL2AD
zk`IiK>R)2B8#u>A|K4WY6H*PgeQT`zb!d#(F}BIo3AKX3#*hkL35*d{ZEzBdEA^!W
z*c5%1?POJ0Nl9JiU4=lIO|TvS_9QI|Cp$o5%Et_KCnn&{$@MJh89F>am~)^b^%V!@
z<^n5$6FlUzDdt)jzKU(}3E{wn-@4t(k?82|x`%5v1N#j6N@ph0(02S<83QlmNpZja
z`q1Gs8M523gOK6rJbx!w$(6P^%?lTe(4@B=YqnX`>DRZo;D0l16`PrzAwz>-x#Uj`
z2Ry=3mvN7b{%1DteCl$%bgwB_J_vCC$h3-&+^4cI_&FCR%sY5h;6cgmg5QMNT!338
zQ^%)jb8qV1SyLj{fP>~q_CWKX4+oo%t!G$ZHMYEfjmFxXfg7mU#<^%GJ!X)&xRPsm
zsIn>)%)uSP@#3=<LTnOu%V9ggrDo@wC=y#B@%_yE)@=d*YC$tSD4^~)zOC#rI;K;B
zbR;srhlZu)WJqqKXLg{tUVZs70@+Bc?F5yfLyPIWhk7O;1=AF4ZW%c-h91-9BE!1n
ze4yOV5#DX&=2^Gy+M?&3W_mmIvLq$<IUx$T*BwMWb}*vBM9)pCL|Iw3I<fcz&%@Rg
zt%ZavQoOKfOxctrbCyhufo}p558SZ0L}*~Usi123+VQ&wdkvQxw1vesAzQEV)i^S`
z?Y?bgj?{OJDjm>Yl(&z#NnCaZ+LxT{zkK&1v<Z^7;#j$$<rZ*zO3wQha$ey8**WOA
z-Y<y{ToQj#u;>U80!;{Bt;|^xXn!-?&CmH?vJL(ny{j`d&KkAoxZkS~_E6oFtOB{?
z=WN<jzSn1uI{jvE%uvpy1g%kh8lFW<LgEqX&O}%mia&hn_zr(L1n`B0XGI8v1so4I
z4a?CPEq&T%>Xn!o=1@)5$Ri~303cA^a`oGH&Vzl0;kvnbKQyaQ!M`fu`ikUtZ{}o5
z1JLgEYG(7EHojnqQ2DJsw|X1gI<n986lVD*%`M$2caM>*HOj$fFx`xnh|9vw)n~T8
zS38ab^7%Mr5MSi#>CCH_r+vPAvF}RhWyy`+se)U>ZaiK0@z>C)ILNRSH(xDARCg5F
zFP5&N?Gcy8AugMeQGb4`Yeb#>RiYx~g}cikH>TFNVx3W@m~4+63^p~57eWhjZ!?P+
zLwxgOC1x0fWOH?bA%F#+GHn6vK+zzwhBUkKZ@5udc?Tq#sInb62Obg$3K7Pp4c6Dp
zZLB?ZQcZ8FR*}T&e8pK%Smi4MD?(Bcc($Pv#E0N~?)N}TK}Uz`0|nVd^bqrLDABEt
zv|s1Gy7a{}<!JQ!N0)PhfOY@f*m#?2qyFpB8tbRIGDq#$A2P|^;!Ez9Ece9qSY!Hy
zvrcOw-aU%OFB?=`42m)sdcul{6U6G3=xrprAap7NvZn0knnV-2iG>e~pXW1=)yhzp
z)KGYmQCH`WiOxoXnR}ZtSR(T6eUXmH?^jJl175f*MOgvzes>62815t(@a{V~T)23T
zFpT0fKr7^@BvC`cX7QD{C}0e$`2vP!no62N`$MeN1okj6gDnN2A|i@+@0>?lyM23#
z_kc`Y@FQl9)A%y)oMT)NdY?Y>6+M}uRWf;dN&`*j9oxU3>@Eq>l8`-|Q=j$%l<P5`
z8-|je4-B@3ZH)vG`KWLF%B$A)Vkh}w$16^KoA>yfh1gHzPn*gY<xa@HFflm*m|L*t
z_>YAXr{9i?Pgl~OR~DBS0>jPE_WVO7*k_)zBE9!bo-cfH_4fxXk6z$9*o&YKtv%``
z7*<|ZM$}=@g;CmrFXGWjd!H$=nhXAl%mSZTl8V6n(UF@8VJN|?1P}zBEhIl+7cElS
zHgD4b?*XaYb<=Nk2bAWKR;&2UV<yHnF52*u+{x3m76F@1VJp+%#aDr(J)aD1a50xu
zf8Y51SGic<!?W~y#pPmmnH&EK^E!)jO(QuM_ZDRmNO(C$7q5N*g7lm6(^_mvsE!9D
z8`Lt6AE}Lb>Kem@j8_izNXSs1<<z(4@M*w>Gn|yHvfQR^im=#dS9{ldwW-ccb#wQ&
zl?k`KWT_Ak6aX~gEanv8H|QXUw)-Vp!MEZrg`U!WzKhS$R(on(g>q|f&XrVaQCe&}
z8hm0o|ImSG7$gsnWvxSQxLSb@f&qcR7Mv?0KkzxIlG&`wKmA`YaoRD9+HzJ!jTu>Q
zTXEucr0d?9vAaBG$(X)hrD;}mrf0DQ&{HDLK6nBtv%n3pNiXDh&XS;ZmsXXPTqSjt
z1@mN54k0<||Fn_N@pF&t(}o(?W*ZsPy6wobh+ca1$=zU{of9#mZ@nywYL9qK@Fbez
z3IIR@?iusjPX?N>iN9^>ts5k}58ybQS(>W8g5~=}`o^KjoEgL%9zG=1iSh};x#D0;
z;hTMlcvM8SLWSx<yHe&85aLo;b<omhEUjK|?^TvmU5GYmI&yBPnUFy_LHk^3wlnFu
ztIt)sY_3_49~x5{2Ys9Luk3(x=A!3*jxBm)`7Ngqr;F(4LI}uZH~L{MhTcbZV8lO`
zWc=F%KToE9d`Cx0l;y$va~z<r^lL_n*J{NzBCoP6y7aEa^6UV@3q9|w7!@(Njuvv;
z_Ch4U0>S`z0bc00Eo<=su7Q6^WuO2e4dVI+u)J=HALH)|!A6(hA{?nYAJK4ISZg3(
z!AZaw(3fl!Z)$UK_Z@nYkiX!!h~DndBDjEQ^jk+ZsmkwWodeC0-3p9bT7Jsr=7sOB
zVAt6CoanLr$Ado5UFM*xSV683qsWPZt1pleKEF&x5$G%IR>o%Wd|^w{FaEA=pfz3}
z-9u$Ao^3~*@QnWNHvbYCfOgx=)Jf@Nl%7UF@T!=g&;&*)Zq=t>>;EL<E3;{<<2|>v
zkCi6_KI)1aR;u7_2LXo3Ov`6fh*+q0+{GQj(zexc*fM>8hy?>=FJOqrJ>!}M9|N`A
z0HUw9oO~*_jS>;U_;C)m>A%6`yi$e%hGqGyRJW4Os#<zla%FTv5_Z-xr3kw9RPxs!
zIa|{uf1QtG&}!f`Qnh;0#KL?JZ&)Em>tnC#Gg&%^ssoOUjHb+;bbK}BK|)AQ2%piQ
zTX-cLp6ok4Y4?cEDeiJjtc79nbJ5KDfY7*vvR4s*T%6VlwoKwstT$YkYRGh{*PU$f
z6fw|TN-YSqrzdQ?-?0W5CA5-gvdG0r%t2GS9e|i{24P~`w!OCT-PEMYe}6-^=eFGX
zp?x+qi&*4Su{aYL4>6nqAeQ8BvsPw?X4~ibtYn<h9sjD<yPGjq#IlEr**S+vJ#%Zq
ziZG0#22$$RorNNjmyO9R3qrugp~!QkYm$1Ea;=vB%8$pfl)M!)k`f+{BkmU|0XZwT
z(4TF5na=Mv6u5{Z;pt1}?BV}DzaMRC85pdY_}+pY!mZ`GQMoq;5<QuGc5dfiaEE<(
z9{m<O+5z{yW9mGx$kQ8nh&?B0JP{CbM-GiACR7k*-=NbW4`leujbQR;bX<<ksgWv}
zoT#tICdQQoFkIr5p@A^Kp~ASIJ=_#ThPaCx2yG7a!BnP%KyTe2lVgK+xqcruDKw+!
zGTYf2XTh_cYP(A%mdP}z?2*)uPe{~~fZ#6<(t62X1@IW!+;L3^tSl-Hn${|IQtLtF
z%_G%@nH~GPeP@NG{13J)>F{Tfm7bP{myase+7!z0iIFu<&wS^yx90DQmIkH(JQF=7
z5=DO7k~cE)OwEX>2EZYmPo&X+`FA?_=26Lm_)#BA8-<{Dkod5-sq$b25z=NqAI}M0
z1?Oo$HzMalN)!w}b+EEP$Oaf}e=Ga9zw^ZC2T1I5UxuZT@qQZ+K)hJTaaK@nOlHLb
z8ZlZZP{}j=`kP1s>ufa&%DXUWsbYYQdv<I#4hx`)eObG{^O;~HQ~uds#@j`aZsop(
zETzd`IqZ<&Jid)+aZ3@q-*h|*b3Y&wz-_?EI*PHi9*dGWOSlRs(KD+I5Syv#1B{-i
z1KZ@#Bhh+(GaV5~{_%_ED<Gv0C@z}9y&s6#Kz1fZ3{+Bx)DQ(}r5HIffc~0l9%ke#
zUM_Eg>mBBv#YSrL!t|^OWA5EinbMpk5#K$!<-j`I1B5}$IEAgDtqVWb?B<t;lS>e;
zzJRN%>m4K;RBRJ}!_b*o=N&7*VbXH0%vf-&!;+3V)z*bhZn<$BC5=qsv17?HPz?B*
zL7(<ysOmI)UAtH0=3!x|x~*G#+zxZHb&4tm7SBUDxJ#cazL-5ZHmmo03$=LTEb@9q
z5<HtZ?XK<3I`s9Cd}$r@`>RH<;SvBRYu)UGFXO<=-<8IG+y!R?nL4w2_fy&<iZCU}
z6ht=HP2tP^Hthg4xKVv3swJ7d3WrSxuaA&ZU)epKVh_)Y$ZcX_c<if;u=}aeqIW;z
zF;uP!=feu6Y}w%!$Qu(*#>L=YU&X(d%$y@whNS0F!C|{<brnyTgV!2^dBo9VRu_1T
zu|und!lTd=^|i*W<CkIIIoayilu)o)&;huhWri<?>k&1a28(SySQMjabUD^YaJj`>
zln%8jyQ_??%+9L{4a8kqsBF3m3PfVt&s3XE_Vqx@2(FNN{VO&~cw9&RP66m)yz&cA
zPW8}|YxJia+2<)(Gyu!Vi*1^wmrzDYfc^xeLtI9)697B9nT}%a#IuAvlZiG)Mn@@4
z(dUU?X*eZt70;Owd$d$Huyjxyi;x1sil7oee^l^h5TND)Jh6BHG8NN~l67ukMJlE_
zheDj+$HmI~YB$%ZwlA#cnp`H<IBER)zv6NuR7L2|x!mx*_=B#>!=_NK_oHf0Mgm)7
zV&lixNuI9H9y&4JQXzBnnD9(N{EC6lF4~<@O;iK%+FqXxO3*{1VZdPrjYwA9MQr}@
zeypa#=1~|3m7b~Q0OJV@=<k^VU#{}KZtIg7w$vF?Mw))}Udx#@yG8zja>`rDRWMyL
zC!aYz>0;^Du(?=x!M)OCYdZGZVVQMe2i}5cWwo^$=SZOUGCRcuvUczhU^y0crFSke
z<z?FC{%6TdHL?ic|0D*!AA@(4<<n1NdCZnS+AQDx9KCl9)C-{~q9sN_!)1t?_H`gr
zWEo+KQ4tBBo>7feu)Ae|?H~LxI_);73C4;@tik}-gJDwsvz9mX5+X_xu<OOV?VAi`
z?;{R0#=EQ+lpt{n?P&^gr9q8p)FWV=6)WbPtvnvcT>?m1l+O2=4}PlFD8qRtU2yuP
zX)%wKDu}ztn~aRN9wbI$P1ZVFeu*Gajs4Y?68TXNF=Z8HjM&X!+G)~FcuR!0G|$1_
z1!J;(Zg+Hib{af1|3a3Z2CrV-b08!4P@`N=(A{L&6f6{Yav*Sb*Cv=gnVqc)il%i)
z9DiF-nE8N6E-^(?xr&WYkDwyp1CIp+oG1#?rkpfP)Q<f@GgzafWa1ofBM=KY!BnNM
zG9N(<=o#?=PKCZ3Ix?)KAWl$yNm|46-zL6irA&xf{POu*uJ7<pw2UTL54PU9J7quF
zcH#0xAK*&ZcB>?IK{Ue#H}7Q~9=BW>iA#bHvqJ<q5=*Vuytm{iaWZCaRe5kJwD1ZR
zcO1CZ`Ik!7+atMJSjO^s+X3l*9mmr?@l%Uo)(gv&ugEDLTj$Jlt9{pbJk~qZ=jf>8
zSJpV0n=|_BCk=G{qMS%hz|R58d`k<gG>(UyS&cHT&5Y`^iLw#Xet-#y)rK6eC{Z^C
zrGn<N{6o}!Cs0Cgkf12Ad?bj3fscfh3Ug4UWasTq`<eZAOg6IO7mJcNY!>etb*2TZ
z6~90b$)kayvC?l&2`=>7D+)s|z*lsQKnj>!@<?7{&~>E>dCZV2-OdAq1Pv@Ny(F3K
zW=cS4VN1)?EFDf6m%0*d$wuZI*MYeJRO)i3_QtucUV6fn{Z(<t(e$prQ(!aqdG@{S
z*>W{@RmsjhZ&@UmW0Fjknl2$Zo~P6|ToB@f;{5}GMXJ-WXb6r?2n}xp>%0W!!rEE_
z&<-6V?git@nt|3I9b;FRltyo|yMTH__8(#P(X5y$gx_KiyHW9aY^0)sAA$me5x_YC
z$qkxFM3adu51O!NjCaOm(N0=UbTFSbFOOP%@hV%XMVs#l9J!eq6@g0?H4FZbkjMI7
zsJzE`xyLR$^E;5vaT#_c$O?N;i{=3<EH8NggY{uQ=V9F&T|db1>CG(-Ll%c}-x`1R
zscTX!DDJp78&9AM+duWz=?f&(VAr;HM^no)!HLh$$DYQlsBd`l_;3^3<6ig9vt%<n
z@jyctz^aST-cZ$^YK*M@{!&~PFiW*H$Gr-BUVAb0&vEpkn44N+ym)|UAs{Uo(*K6N
z$E*m=px)~}J_BiP-A$I>qqE0TBqrZIP-qn>e(qSqSyN0J1{W(SE!}PLF9O51qgT+~
zlJmVa57$FAl&30%J%0u~Ja!%etmc-^nnu2>p%6El9Z}zc68Q*D=7hhDssAEqUo%)g
z__s#++Nu7{(@TgUHLiT=U8c{yA=cZ0W1wtWw;y|%D!o}nQqACp(iFqf&=??xa9cJS
z`;Qkx@=e{weUui@T9hS_!0#MHALeiT=7S=zQ5?O3ee_R%5iX0^cyHQdTlpmWpMUcE
z=JwgZD-_5fvm<>fREqbC=#loGSNgYNvGV%J%P8VEYYRTxt5o3ti0!Dh<u&<BPralZ
zs>jYOIn`ot9eeC!PxjaUxnW7`F$U-U|E^&V#X3dDCzH`5XFKZme&q`jiEn!L;o)}z
zCX~voxHp~?n>C{A)yhNkf4O;4_9Eka;elMw@c*Bc0gjD;)S6RtkOWnrBKq`_D|;8J
z2>#@>`1hI^?*#WtmVt!IgmSYOidpP5G>S+hCI%1~!Rg6v-UC8sNFw1kma@MWF#*?S
zNPPrxv$kf4j&rbBFwf<M?KC>i=Y+(2*7Ies+|KEZehGC9@EBIBnGP~S0@)2e4!*3I
z*tr?9Ll4f20WoR<%lw&X@Bbz{`zPk6Ahow^9EAAHjR*yp8OE}@XN^96*qwV@L}1?d
z=jSQSGQqk|r_(Bt@!QX@puGkg$FL^lh$=Rvpa_sOUiD|VI1s@>w&uPQpX8pxm;we9
z?4vPZ*FCm%f+-bX3o9QA8AkBH(~~vKMrf2JDmX&t64@M2GO|ja={xtR*-;3+>G;Gh
zG4y>Z`{!e|?K!XE)2$=BJaqO*XAX!it8{2fNEBnAli?txC|<1)KR4}QNZ+_n>qTR7
z{O{(+5#E1vI;0>U>e`}n$FDKX?}e`ABH(j^&)f3%RV?rs8^x+d>=l_?KFkht#P;L{
z{i0c6Lm}>prMk~Inh8OWPf|uYP7pOZ>sCacjMp0HnpMtXzLIrH)R!8TL82C-sepV6
zAB6EhPB=6MBeKfo*x1-bp~V1fUX^%t|Jw>y*jpso0t^MH;V|E6?@iqH3wV5WB$mg}
zB=Jjb!Is~Q_c6)!K`%x)1li_1ac~qs%NU9_p;NaYCD}V;E|B+QIv)2=^lTi~T`$hs
zmsa=c{9TdhLl3xD9-8fv5Ql!@ubpXznSXW3)`}A1OVzI=wTwNnO=*m_{Pk#KiPt@^
zaDxoK&-Fqy9D^}DmCFv*7T16PpyVTWMooKUN@O~99drzUy+-CcpyZzgA@L;3;d4ih
z#c5MMbU*4&fF6QM>_MOt7G$!v<gD6Ncl_VH*-?ng!ytJLQOcsRyvC<1&%Na;hW4;q
z^D{&De7ltG@gN3oqA7;Rj3|yp`>N|?B1I1@vF*JOiS;?UV5<%{buX0k&gs=&*f^;g
z{$C8~Ou*A0-YCra%4_;4hy%#to>*v331M}$K4NFh_>-4G?rn8GHKhw{8{`gtpzKd*
zfY9pTW`jo`%@SZG#H{33CYw0HrA<Y?;#uHn8HWxjPl;{$r?2IB)l}SqpC11`=s;Nb
znwby?JbZvigZ)J3z{*;N36%m~v4k$OLsy{o1+`-}cklz_VKtBW3vOJGtXZN<7u^Co
zT}@S2+uUDJ)`WH|XZ`Wu$SnUauWk}OmHOtWqnaB^T$V#erp<lQrdqlD6u)08!zNeP
zw5-YbW~_5CN7D7QVVktnZ;TC}{+5h!WKy&<C+2)8wM9r&rv6CR=wkahtGd2=-Oo7S
zYyG>EQUhm6xmWaC=+aR*=CpioafCc_vM{+G3)@XaNo@Vor3;4-WmGEzqc1Nf&jwU4
zuontbfJg56TLrlvHG?<O{mJs)JKL#EIX!``;P_7Gy~tA7NNu)4@Cw<}i!2B%punM_
z@^Y27)otbLu0a;t#2QwhL8}1SDA5J96*!#s(q|iIm2tF4#d5#y_n#4N5DB~m(p=^1
zO2f9(YMn2HweLvnidr>+*^6Yw^?6@|z|4Yf?Rwr>EG@w%JU9LXQ*S{Ll$ZZGvHP!d
zz02R{G}*Kz=P_KuXoevd59?lZI5>X*)NoPYEZ_b*pYe%ile&gqWaVe&e@p5HbIlTq
z{8e5EFu6?*7i-;Vo2Ip72{ZLI^OCJOjYgyqpWYO-;81|zpro64;fOgZG&D47XSRS~
zRCBNTl2+}1u)SfMvKu8-+epw)*Q!;Qm3Z6u8--%6s=hy;D|i(W8twzhaz(4eo$Ujj
zpw487FR{(MU!c5Qc=Hx)u4jbwC1#N4vc<yVyz^6C_EmJ=64fou)b{3o_m%EdTx@^2
zXJUKG!RNQ?nE7vgT*ADWxo;957GUz0d97}4buF6-77vjry#9z@WaT8|CuX6sgw%U?
z2><@?Cu+Se^U#+3Xe$zDr#dV^FKQ+Ps^2ICQV2Aj5b&bX0<**-D!}uUW0_@Iw)kc|
z){&t*q&$3^Io{*yGmpn$QQo>^DJLn3oy-~~hKc>yr4ejPLOirUCG08D&O*m>^=EiV
zZXhDNcGZ5a5=lCOIT@$=#v6mfutzAKJMB9MI*QLz^TvO$rg&Gb?&mq3xxfFDp82oe
z0*~BhY~@nY9??+CU>i@{X|eyBHw{S)Aa*H(2GvaJ_?*Y907D#pq4=O=?ES!Y?9>lQ
zOvMAB?tm8%w}U7MR+ej&F>pCtD<aFV?vEu2>fvsC%ph@qLp{uyu&1cU3SM2x&A;H`
zx=BMt@Lzk4PF02kDu{|0l+2IyHRdR>H$L6Ufsy;t$;z<ya7foP-t{NR0+lT{F%h-~
ziMVrSSU$7rKmzjFrj(?rS9Q;6u$(DB$LtROn`;Xn-|c!*hbAiKQs~*UTf@2zGd!Eq
zx^T@D1RkNK%(W=U9sa&R?WyFarBn{F<dqqK*I~Y(WQ9Kn+#3=TGbDVz^J6ddQ0jGD
ztat8okM>PzVFF&WqTXn2IuhW4^)!ToJaSh;P<XDytMfnHq;HO$^MkbXZ4~LB$K<Ca
za|!xkjhYa$UShS*RfaC>4&K_`Gm^ZlsG;C3f4zl%1}{1E56(kLiH*Tk67oTMRm{$a
zwz^RM$#SKCu281-Kc-c-7b8qv<j5SG&RkX)4j#p;;=5gLY1Nib6a%spBdQwz5Ne&h
z`G(G8$SZvZ%e1w!-53KC_HFp_lk7oGQMn74w~0}O0Ps5F->9T6@@F*Ll<v4rYT?fs
zz}3s)Vqk{W37KZZSWZud#}Aq)+%GPB{cX+3Ld|So4SkcN|HPd4T3*_CO4+Px8ZCx6
zS1J|;y|B+0HYjPIUL;zkf?fB9=BF!vGpHSS@2Zd>tA*C;em_ay#QqqBrpq@qk099B
zlrS{+g0!&7bn<g3>kifMCI}pM#xRnuHBW)>*Vg^9D6RQ^Ve#>xjJYx<j?31b68fq>
zB0Y9Qi7Y86fYW%9?~#Gyrn1<+h_j7x88;qY0gT)r@tc3ZZ;k#E*uJw*Yau&L;_=Ui
zu$X@VlUX(PLDcAvv7DiCgBI<yJDJJJUheZQy}q7Vv?ayxfo<?X&jW_VTbdu+?lm0A
zzAPWb^Pt@Po_#T=<t;ne)3isv`S$y-t8sa+5Et93vWTs~&+_}e_X@K%{n@C(vGC(7
zd8@$A2b=2TdWJ4^q$wyRyncP}&ArzFuP?m5Q|~{+b-Rg<$p+i)b|>t7`V-d8fH)G8
zM&xg}teK4|nM$Zb=;((Nq^*Iq0zvJe#p;3)_PJW#?JH>w%XP;s9?PVjnK^wO+Y$b{
z<<bCJE6R!kOO&>}f8;&fdT?~o-Sc21N7*ZvxaoZCih&R0CJzTaxKS!C+r}pJa1GrR
zdD8c?5{H?`l{1kz?(E`}+B2cmDex}r&ym^i{MR1&;M8Z+b?z>EEob$J*jeO9-{X=~
z;|{ty*Ux!ZMq8b+N!=p)<ewG(-^SdSI^DuY9lR}sB;XpV?rLInxR9wr2+zt%K`0e!
zvgoWP<+D?P^#K7ZLrml&QX=2=3|%oa_C2G9Wu^#ep|%gl29~NFt$C4`$HvRcd(7QE
zVBPhTSR$-1na5q~^|^7ATkpG_mPHu}+80N+tu(!+w_RE{Q^1(uAA{QkLM23EB3<@(
z8+!B}+gP&Z^Ag*Rve2SKU5!iV+ma6ZUs2k-ymDDdcYCq@8)?~6N6D(Q>nGjj!Q##n
z7jd4cs|dENZ{U2W&wbbjC+p$|5hqnwDqI9%QmX0&qYNkyj<&aNIp6W|NN}mbGc~7F
z|K-liw))AB$=?(#CZ4@bIW4*LOypQzNRmAOK-@{_=8iYDq@^i?H^hhR%JR>jQBG7K
z)K~VLbuTFGCQ6&RFsLDDcR;ivm^8T(>O7Wq81B<k$00STPKWiBWzL<qtS8DbdvLn9
zmYbj><D12w$<BV$Xy?U*UXo7tph@xjBj$F}T>CCP)Y_t|crk2M)N5?2dmxSmuJ&m%
z<hn_VbDdsi*xt%k*Vgb`x_`Kt8C-&aU!i_+onPO|9rrAR>NYQVpDptKfv;+Svu7V~
zQ_Io?XT-;PJUyLvEw+!$m`t)0+Hq{*tMwti`TOk7_iY&K{OncyIU%^~Rk1=+zq`v&
zPxE~d-JFpDQ=5YCW2eX0hTge@qBYgx_P3Rjz60W-*!XyIYB1naof<9uZPARWrxVab
zpV6qr%kt(W*<%s}b&lK}FstCopx;8Bck8YpTeap)gJKOkg7LY%?Yv)K{Al0uguMFa
zBWBtCd0y_HRvP%MIgV{DzSXov6n5UVczT*&c&hepVY$PR`XxbI!)#;ayR|$ne9GVW
z`GGDiQ5G9;4>xb>VW<TTlpnm}uemMU&w0L=xU4drw-kQ$p7pU}@9W0IWxTSIuw@(j
zzUYID+YN_>MK;Wp;d$L&oJOs3DvqAL(_xOxCQLsOR8}t)Dg_D&iwLzClVZA)y}`uX
zPyP%gCcvI&_pC+B_%ePxgM+$HPumiLfh16ZT1BaT&`~D%=jzSt5zD0-bK;V>PPw5+
zaIJ3kY34*4lY{v=mnq{tR#kb+q;1;VnMbBm)X8A5x8J^ba#>t9C6h;1ZE^K4k^a2i
z*34`8^{3t<J)wCx87sWoCwfX8=iJRPl3%D~8ooAdPt1#!1((J@r*Sb)UZB|Y)9uBF
zh0h%CSEL`nkFRk?Fl01Nw6jN`@t*k}(f8|vW6l-!t(Yf!*K}Q$_Ki%-k4O%o<xV^@
zSp^g*-2plg4PEa;D)Y&dQ(RnJ14NL_duNXM&H)xP^&=l-78m&=|KUirYy8cl)vAE^
zl%n3$sI&qJ#d_ia`>nKk_bwjyiO;3(ZKsD!-*K8Hl!?s^4J>&>WQ~KDE`&vF-tu0N
zC)75!lQ*nY#g->jVpEW&$-?R<{=vf{(=vr|MO_{IZ7BDymFgjedF-BDPOrI3m9~Ww
z+IT}rFLR~S2||=v<jR&gIXC#c_wLJ|gydwczql+{waGmXh1c|$dptVYBXR}Bp>8)~
z;^m9EOPr$OvqROZ^5@<kEl*f8I%l%*u5QN2fSEk22g4<WDvx^#q#r5*&OJ=Vh>>Rj
zA1~9pZkhae0s-lIGazcbfIWNql5Aj@WoH8BgpCT8sxx!&7KPh3etsJ>>*`QVenj^)
zT25bnkAauuCN#I(X{r<i;3K&}Wk|}l>RPd)$Et0v8%z4;ROaXOw>&B|je2mpKrMIV
z?X7aZgUeTPhw;#O_hQq@f=bKQp1F#;+mTYYY|r&RUftN)dur9^1$bTap4-$Cq**9Q
z=ad8{zByLxNck6dYUhTW^XPE&+>n>(ITTY-RQ%$|Dd*o02YvOmj&M>^jJh_?9KCJ%
zaeR$^Ta93FtB_5lN?dU<3INGI*VZnCWfBsCSUr|bGhgw-EqcO_Q}3$3eY*?^mjd<&
z9~xm%+qE`?0qZi9iY<z!y?MTQtnp~0@P>51<gP7$CM#`3guXUf)vg7NYdYIgoY$px
zjcSjiYZ_Z*U(ovI$o#BOa3jmHv#)v4Ck-U|F1KH7YT-S2-s@Cf-pj=d#wt49T0mAt
zal~!^Eap2fV)9v$@kFV<i!43w?21N{@}QcjUeyy_c-H#)p6Vot*3?oH3oq$i*Th;w
zg;jG_uowKJ@Kv4Ia`VibGV5wpytFvY0IYxcEkt=OD;fxC-Wq~oD@xCge*e*=9xMg8
z=THsQ6J-rQHVPW_x0E~=TbTYIWp4sj<GzLuFHtBYrKFUUCTT)S8kEphtHD$>s?bgv
zsH90q6vAp!t3gs^h$1vn8rCX>wMnxkNu#1c(|5mXW!UGO@BjV3u66chpX0c^>-`PS
zb5D;pF=^l;VV*-sC^I3HJ~yVF??5d#b1&4&mMHJsz4$qQ-IN+(-we`;6LB|732LWY
zZQu=$ni%GKF%Tye=8Svr4#KJ}cTxDA{9oq#T+O%;(eNU}S4&+(nc;D}^P9a|hSpgi
zL9pRT_1T;lh+(gzi?#|&1!p`-ome6@wohoI#P>ByAEs6onrA*$B)erKTl~1^G}g4;
zbCHy_yIxe!(Ebgj+TKPoA@H_<>jIvIh$ConkqJgNGGYdqVCge6tM@A|f%Xvpgz;j!
zLd$O9({mHidb@T$Tx4^QQHMVQ#yI>D;Mwp^BCa`3-HRv;PM@BMBvsZFQ?xOt<v;Bk
z7X2ttFjZ=jP{Pg`PSbYad2h5BOX^T){puLum}@VJrOgv2<5e`;#?E$S^d9m)lA-)&
zSNO=M^#y|8)fNWP*XgMR>f7fY4jGtnvZ&K-Eq<f&-pcc|^}K?s%8k{wTv5id7t3-R
zh|zPyTx4SGR>tsl@22G+{P(}|+2mPkZ`ptO8}m?j;@9_d{Av_C9kiKsm}*U(_Z;6r
zmx+xt+cj3sbTDL~;z1fAU`Y^N$0x_tkjM>(10E6$Ddrxp5u6^;a_-!HO|dhBs#ouK
zF#+)p-a2E)H+r%iC{hIDgkB&mjkp!<l_4%8e=`_sCa0W?Qf40yaMHTOERkM*lBz-<
z^C)smJ<D}MmRoI^c}c0{r_|NuMI~1q=~F{4lR6u-$4Z8@IFshMUd7tzH6G2tY>nzZ
zT6uD@eM81vy3)wO6|HR#x>u~bgi+&esb&~8PB)rV$(I<Htc5!FsXX=a+KPPvj5VD6
z?{b<hSR>XmzVcb;V?B?8g}bKj05Ac%DUKTmPY-(7JS(pi4&Jc2um*@idyWT+k^Bw(
zIqA%qp){d(#=#s+s&}D}B;^IFLRMpgAxRjiad8WdLhO`L-^RI-j#Iv)5mg^2Jb40k
z|5*NI$2099wpd<fVHg-f<Bp<4&nj`Om|eypNujJWII(vA^Cg+*=^Nf}Eg!!_veZY@
zE_+oYGNk6XpI0)q*f((_hIi2dq8Jng<9n~M+Ws6&#+sZ&FXzIXjP-z+j8#*&4wJDg
zo=mxAG_4EeDmqc-$_RTq_?VOZsp6Z~5JcY~cHpGZGQT-fE$<TQx_`*H1FhRAqQ2t@
zGrUFeubXd!bjie3>tOy%lYj2M8WCx<0rn>h>>=`x7VFGmineEvP~Sk74)YUof>9K=
ztdY{1j53-~)(n^VSXULl#ZOk?mhp1F=lsnR^+dX>j#TWUS-9xcl+=Pl12cU0<)_>6
zuaQ!U&Ar>`Jz(~ah-QoRtj~1YWP?2s)Z|<_Q!8q+Tx*-KJO@3L5-$D-Nl)o}sNGJ*
z|2hXeF^Mmvm_~X0Lf2G__w<DI3MwA|MhHbCorm#tieBi#p&7M3{e5YSC5HHpSO+91
z#AQZd5b-ZjIy6Lx-Rta>fsQQgUD_J`XUt5agR6hN{}t|Q40&PBf4ukAwMe;Cg@ek1
z7~G7e9qI^S5fK8j`yW*RJLgV4?K$kMJMgLdupL|G+%gr(hYzXIax>c>c9msM`fyUr
zYN)-KTz26z9jN%8D?V4<*~-P+i&edMCCZ7n&zaf&?^3u!RW00aMO=I?^fELK+mgu_
zE9?roP6a31me68*I{F6g@CWlqk3CzRTS8mI$tN9k{{rs7VfaJ0k54wNcYWgOG<fuS
znc&$#XhKl~Kwyj-fWg>-L??nNX9cuEM^)7P+FQ-&VR^yHpJ4BsMl5j@#wd^B`b$e&
z0l<<O>kDBK&|V@A8SRqxhPbot%@N(wy(0O?1W5f6qW%iCyyiNAY|<w;#NDdcs*Fo!
zzrLrvdrU@9sB;Xcgd*QtcWwD)g&5*W4Xp#By}UgT7#~|K6K!Zpw&j<}fxY}>ZgfHl
zHBQ8_dV`vt+NrGqOQeDw1Tf<iW2x#9@igqjsZL&yb?_Tp^_==e6_=3UlRt{Hb4V((
z&xtCh=&kr;yXa`=tEynnEgZ$6THQQ4L^4WU&jwyff>yvKOkaZNJdx2!#K=H+-_*Cp
z;O9NCf)L6<ECp!*%G?!KTpQmr{Tjf$74p6<I#T~M&_MDukW3Uv5TBp1d73+$cz7hb
zc(97!b~g6tJcA_WxD>t}vXai%s7Z2m4l^kv^P*&ZhDGU;Ol|WL4DC*lDQ~>U+5=ap
z$<*g<PQG8S$*S$4=E@mIUdfmjE*_mEOaR(y*{U_e{vw`p?J$38@;+SRZM53?duC_P
z&l*%4nD`njETwQ54WfTbo1jBaflX4j^_*6xzcY`TO&sjJQKL(ovIjbJuOMM0BuP~v
zb1n(%8!sOk34s5^J%ES-IO4&LL0ZC!82G&(vcs8^8nGFgvYdr<=Aa@nFz+261K9i2
zZ%hU8O@xFD@l7;)jQGwJuxmd)R^nv+oUxn7sP5)1qxMt!H<O?5W{T>h^XvU$gp)c&
zl4Vk*IxGtQ=GwX_Q}begck(LaZd&SvUdb4jyv&nnX70b{vdclmu`ObK<KlC||968N
zJlY{+X5Z~)*+1PQkDA86zdKU9BhluaU6T*5l+V*7!vc4Hg{jZILS-vWh2r^oV#aMR
z5q-`Xm<D(y6R6aWqk|nIU&8_}h)#UoiVBf9-JR>qe{o}NK?~WSn3#ij5isEdPJ$1p
z#Li=wRE2<7^brsk;#eh?2K<hwmM|r7*xo1PiAzJkN$c@R(XD<<SKVfILoMVXNB&Q`
z?ZsQ1FN|(ky-wq5^fyOb0d#Ocd+1S(Q@ufog4(+=6WVU{TU*7q%U<KX2XA$M<~j54
zlcCX&*Xp08Uo@|P7~4e6wLks9ZVE@<f>jdp2j4e+yHLnkS#<nqmC>zb5@Zg^z>GAD
z@#^h6hq>H<{RMP*ZU}2C7~{|!D=byns1y=1e*Lnq%q#fUa5O+Yk7&#dR|Zfe!;V3M
zfcoIu`O{tmo+|<aUnQ85I%Yy1L^D_g1Rf4Q?K}>^2DoR9!uRZ2j7m{w4%xU-e3-lM
zV*{cRhi3Qb`N^)5iQf3RYR?5olR{S3L_LdLSCe!v=9b*)cBKP(d)ZE&=%Cr%9O~yO
z=ww6`0ywy!kT+$?E>PN`0C&m8w)O0z^CFhbM(jpW27+4<oDlqY*uRw4cweSlRqaFk
z1<HiA><B95hj4RQyq@UI&{m^Mtv>9#?}!>X(toN~Xk<K8bojU_-)#6-k-Y-?7yI8#
zt_6c@Qd3h2XO4!39)yy2-koxcF-iEDNW#-|y_odRGP&~7RF*+ytzm4=T~t7+3_?tb
zNZy+xoJYpm%F}&0N^hv-Pjox}lbR8(JQ1+?<STCi&UI^xPoZ+yeB7UWcg7OKfOm>q
z5$i}lzFmG5^SF*7RV$A}Fyd}(eD~xokx=n(4vBJo{ZGdEIzAJzE9z(o4s23g&)o_!
zxow=Ts@UxW>FVi9_^!ZTb!%S?2}vnTMzU$|dTF0jnBwE&`B>!r%QXaW36KawDrO8n
z5)_-oy8PkK>tJigO#lQR$r``}IEcR?^XJjx6vx7~hNy}NRUIe#K5)SpM~kx%`d~Or
zFpp4@5}4NenSGxO-+y0(?3^TN&Yeo$^Wtv<u{F6X{nR%m<E(y8mg=sWWitCn-ROox
z0d@F3Gs{*cy5W9w!(A{LsI076CZTYaUY;UdZtPtiIPi|(!M!JvR7+`r!{&oD?y>R}
z+c=haCJG<cdiW`Cxw><P{K&VqI}4?Z8ZUnRX3uRhB4K=%SGv!T+T9$YGC?=@8YL>A
zg0U}E-l{SoAyo1W{MB}0JL6u%(FgM>Ub(888leYAlYrIED)kK*ZRN{=ZZdz~>^6@K
z0#J%WAIzqK<V@1h>lwZS6OaO3)t^-LPxu!)dWuckgzp1sd7)4~VOkA*Cf+{lz-g7t
z=$EC>I@6W^de7e&c|Ai-gZLVR2vFcv{QX&zcg_vfV(;HSMXP?OZXBuPKI8d|LKD|z
z-W5hWP!|bj+(FWk8+Uj*##S?HU~Po^-#AY>-6IIq_ye7{iY^m|RnlfV`}~g<wA$-m
z<Di=~D7+LS4`qCM)39E?AtFi2T8Ec(%o1~pv=K3fy$w0g*^a7&R-ArJn3MxWe)fHJ
zMid4#{d{zlxnh8S&=+9E%pjrQ)tQ5-b0Bm?!raD&oa=MFc~K7HO8J^G9?{bIZ;*7i
z>2+;sis&086=ESAF59NY{=CFqHshJ=XVQhvX8(PYUR?J@eb%5cJat96=N_@sJ1Q*=
z#2qRC8n|wyxnIqgM``DC@S+M2h$F1;Kes&6Gg`7I!YC8>nGld?tbcfXy2t4{zk!RZ
zsM_9A42n1=@9HP|K5;ECS}7wt5Uh#^L<|PTcoF_|TzGNXqu;*qK^MUU$4Hc?zkrZK
zb!duF?Jo#Mdt>~L)-BF#d(c#(7X*8U&vYpOr5q6SUAsDx96}QQjg4kprKZV|J}?8h
z`DXVz;?6(uXr3gGJAbC4xkzncIv!C&V;0E`ryk|Ha!<}Lc*IWGm_Jkfo7^XOA`HV~
zNk^Vu60b;+O^-<yZsBYHeJH%h8YsmEx!&F5h)GV9?>V^jCk*Bd1)O(%FDHLNmOPYu
zps3DrMxyM%>M8We%=^&kgvRNuUQN%VGzS<W3Wg?#%D|mRoZ~97SPIh!30mejB8v-k
z4^D(EXDQ=PKl2NbnIS1CP%c%!FS0%NZPfiPcunwo6z6q3T;*+!@fvAqM-@#kJ*lc6
zQS=|x9qym%dS|?JI7u@ug3lmda`YW`2A@#j_pTOOTYj-}8m+g;DuaXB5BBdm?U(=n
zG;fdq$<FB;Q(sERTg?mSRyu#h#l|~PE>%{6E1b8Py7K?EAJ$y4%ZNA|j&C~w^oW*I
zm+xiaPjh}ryP9Soa`Ky#79gdl{u3Nbq^xNwYHEf{{g1g}tE6Rh=D-%g3Yh}EO;9?}
z)ZR3usR2Yli;R8n%elHD<igLH_l&im!%-4<_*qtW6t)2?L~+Q9;oiI=Sycb+?gkoP
z|31+%zMq&;CYD8oDVCZvKYxLE&S|?k4>_*z2IQffY7ZD&9}tMQ^O1X-qr%D?nlq9N
z&(o6a7H7U0aWvT9Dl9nzzzNrVajl8IUPZN_QB}1w#N0*fjek`A-3E%|_Cr>_N^7)k
z<Z6E`F}A5C+x5@$VNHqKz*VjjmZ>&R0^NmM>|LBB{m1MQO8WAD93>Wm5HmN0u&RPb
zsO~~NL{7FqQZPy#rb<TeXAT{YiG7>h4)?zixJM*uFuP(_q(<eP09t1`8b$y{qrRGS
zhLmXj*II%8AFramR?hCbo;IIY2|J}1a@vIUTbUE{-c*uo-X^S6<hr=JB1N(b$3uE+
zknYh5ui>Rq76f)^>N9e#sm-xZ(KN+RqVKgsGWCT75x{8Ttz>L<0gg>5aA)}+>FiCZ
zI9o6UxqNJTEfN^-<_)Wphdd3zpWJJ3^{8mEF91w__B*LQMf7Hqs`rh@r1AVkAIT|F
zTcHmjcYGbgw&<`Pe_FbjZyl2p!=MjBErzsY*$CGYCqiJiWI6o-r|!OkF$X{A5D`b*
z95L!1&NYNj^F@Z-95GnC-*&FW2M{X4s2~|Ny}k3FR70T?BX2E9&-eI`<Q+0<Tg1tC
zE9B&AK8S@kZ+_z%Oz)jR<(F)T3557RVe*S+q#nIjzuX;okrkTdv28C7|8P5TLM}9%
zTQT;QZ;eDdX~*yC*eug#->4n6was{F=uy*x6hO%fjtd=@T8^Npn!%Wb`9plwQ$CNG
zV6pR$h|Eqw=%}?&3c-krItat%S)~wK0+2{p0yq`GexS5_Ah+rZp?_!nWTPA|Qbh3!
z3k{%ETmm%vx1)KuBMl^rjJ>%vz!>KL_|i2pl$mOhK;O%(eOE}SuT?WN7y0z<wPmon
zsftCETR~6!ErUm1r~@?juG3fR+rY`ySFT#KOd9_<n3!7ymR9{G<~J9IybEn1ut1|^
z2E}WkltRt&iVs?>(jWMbs(0qZxTXKQ7DbW6gGcvu%aSAS@afe^n>3CO^7v7^onPg$
z%AiszRrZF#*1*`5^SILSCQ$*QDMr~lO@rLqUG9Cp0&#;yiYq4PU}ds?-UAZA#Ac}H
zfFhGG$cYNgluVcRJa(*YRt4m=yJlAGOo|)*v*~fTXMU@~EfnvR=ZZ4`^Y&x*Ga+=p
zKznWIeIa!#M<$JU(R4tg5vYd7hsZubxCA2<@~0Xc#J@Tu=N`VAp}xDa8K=F3$}&Bg
zs@hrr!@>osqb=kMlsxMlAKcs@XxvDM7v8;X=>g9QKti83^-hI%PWf8g3oh?FNR=b`
ze)+6B(y%AsNKwS2?K*SJBHPcD00vSOCU`G!K442x17QI5wIpccnDgik{7lH&yxB)r
zbLd^?F~~dtB}40Ey9CEP(FG%u_DEyydGMnsh<3`d12ET>O_^QHc+#u&T)}@T?1tvW
z?J~xR(xh2T@s8f+vsl9P-pW@RdfpezqvdoDG<?rjY6w$M-|ho$v`0v7)Qkhtj^5n%
zHGkDJC=FF>uVB{Y%=Y;SSO5R@zpZT--^-g{C-o0JeN<4jwLks#t4cAd)^_(-bzV*V
z4`V!G2}5bitYC<JEgywsxEA561m>I;Kzt|l>krrYw*Oq`D{8sZ(l&&h_Smot`?)K7
z@M$!Y4t$DH6IQ-o_3E=d%FjnII@hx5WXSw-0*=q_YYaH~MC-T$u)uK6X-G(VWlC^&
zy@5fYouF*FBi-BLLZm~7<gjtd?1Fm?=FZ+iHYV@qhhrqC)rwy>r`6uO^xpm(bLnF+
zSf41S<?mJ`F9>ik8go9V+~*KlG(N~Dr#zG+Ka#w!gtq<3H(wbeC!Mj{eP%cBCM2=w
zAnb77Z(W~J0RjmsFO`3^JtPTZb#TMs#)3f*nlu6k!X<|~xKg4W2U?LLRrIBF2$Ria
zE&U-(?p~;)>NWsO1b%{orczf`OzJ7IQPjq-ffK@CL!bhH##Ti91-fmfwSW7X;{}#4
z6CyOv0SI0haR0dhaoH*B8~DFmjj^2v7*`!GZ!8(#>@eJSGbPH<pqjU5@fP}{_4y-{
zUIGc`S4-r`ktXglx};=WRQ0@pc9eUa^GeJOU3^C%$cu4C;PKh)>4PBsgH6I8gpe(J
zhwa95pQQe|w3b{V@ph-~yzthAQa<}Ke+)cfcmoFE(41;1U0)xR1jH95+}?YUYc>%j
zAaUR^(_EnIfyxOFMUwoLn+Iy|BKQ#xKY7z$Fu)6fQ7Yp!fG*rS$sdj~<PFP5%_ME1
z`?L_9?xS5VX}5U4-@bd3L=lrdMJQKHHt)2c-s%>qlo`uXI=~yza0NZlKZn^KfcTaB
zn_JYakhf<(z5%1F(T2KK>-5Yy$_Jc>vGqTdHl@Q@|9>Kdn60j2#aMwy-~ev2(&c2U
zth0gc<Fqm}rZy8AJ-ldWfdRe|vD8pUK)uEaqrkK9B&?e*_jycQ2OcVk%{p8Mr`*v(
z2do8cwTQonSs5(NA|`u9{H5nR)TSqPr7#DS$CJq~MmIDgw#$%qvOg#u3V2VLthUpP
z%@<UM_r$926v?^t=G$;7qx=iQLy|<u;%bF1ggYwuex2yMpM4s2-O>9^QBnIYgzvh0
z@lWcfxqjmX3<VDMSkQbX*+0o*Dr(NbS1pAP;@=fGp6XeT!GiaK7js6=rmLT)C_GEL
zMda$nh55w<x?jL_T49|z$Yw!j%Ua-SfdjyMQNzLhxklsc&olkA>>?&A_~sBG4k_bG
zdu4R;=;tBq2d_bVBu8(I`}Ge#(F#nTXzS<?*7{KnG|QAT-0@Zgv!~}VQ!xZ|yD@$;
zUVZ2Ll&-e54zY}IN9%rT6+Rzp->Dqg$gpg$t@K`Nhs0btvK6nibptcrOd;ebRrq+q
zx`gou1#+kCU)#>6zZl|mYUIjbND}z&9x-x!t`KJWUJv@(wOdBO>b=4Ow~Qx)b0!Zu
zzKi+U)sB9+;#s`b@Mo+F>Qf@>66%KOzK)4%f$@bD9KVP3K(73n;xAb1*kUAcgBnJq
zifh|K9@1TIkY0os<V=$EQ;#lwCg&?4B9(I7``GaOjn0d|cD$kxUqGg@TCSK9UUl2G
zQ={kUG8JcC<<HX92-;>zbV4!&A46>74k7FIB<7CHyIMS#?k$&)Bv(y0Hh&ScZS+eC
zO}l<(JNkC(dKrE<<%KYL9%_M(VT%?7KYy}0xxKK-cg_szb=eympqUg#$-7t6J>b&h
zXIHe$5vcwt+a*}QmB-p}k>fF8Zilv^qx$d1UuMxsBSri_Pw{O2w$_6%!Z6E}MW}W#
zHJ-7yLn#YrT{6QF9Nxhn%Eb&~MGNFlZGG*GQs-ItyoeyuVP;j9^9c0BV%Z$=b{<Je
zc*lzN&b93$Q|+hk=l$x}ZaXGYU*gmvT6eLNH#*%I--pCb&a6mk7shsPFB|VHXvBIZ
zP;4S@$ccy>%Ckh=P%#lVG-e4#^X|bOmmel4^4}Zh56MqW-mqEZl*Vu|SHBi7d|I`p
zJcel5gr4{pmR3Gm1<ey;FW`Fvgbf`Y4lpE}5)(nA%CWr(ZKA6xHI`Z-^9$EGwo4|Y
z1v%0Cdxuw-gWG|K<*Y3!dfv>-eOzMH^Ynql<KAL%MO!n1N2aJ#hX16toQJL_<IBo(
zK=2RwoauUyT0p0jMRq9)EEo<rIyCYmU{Ih1SIda*?vQmLw$0qx>7l&g4!7jOPIWm8
zuEZs$Ope?UZpQ=RO5+O*R^!Tw&?<ov(*c*55_#x}%JFSq4(@+2wf#)y0XN&7ZW$fx
z)I%sM9exzVy`+Wi$^RMv&Hp^H{{EXeBNxNQAB8nI3Y=MpA_SZS-X@$(XxuTE4`mF2
z<KW;%Fa$iHht}>7hcl1$Qyr2d4Xu;nJOCGUSz>2g<42m%$Sl1G`oF2KwOGLV%>AyX
zZmk@P&cC$yXz+C)R$*#hd{*d4WlaE-UK_7mS0-tn`A%7i#kAMX?W+4y3P~-L7H;3*
zafqQy-u+%FHu<m}|2jhD5eq=4JI-HXvk=&whim4CH`w){t8<99cy7Pn{e?%H(lbV-
zjYDEZC(X8I)8E2HowU01|0Z+)ck@b&sZQTJFMQ&H{0Bd(@7G0jAqbV(u~oFNZn*KA
z+x$nXvYat>0z*28jve0(z7<w|hhN}(!hgeT<G1+zS`VAmTjS3jCI-D6+usJcM^TY3
z>Vkt;GZwNj2lqnGk7EH;H`(OUGXlp>y|-xc#A{sW8%-2PpPjjPlBviXqHQx@(Dz8u
zv3{P#yol2Kx<y~M36cQyw}9*~e$%pMOZ|@UJ?!*e8M>eWo5P-Dxqm$Nl&RL_bb+&B
zZvO`Zl!A*2a|_O^Kpu9VOKKXxKBf1)cfB2Wjb|UTfM}Uu7FCDSIZg6V)`6n(oZcq?
z4x(xGAIKE4n%NTrpk^Ki%hXi)tqp@HaorIUDaQN&Zlc&BoBV8DlALH0PotT&EUs{t
z@(bv`2_JK^6cQECse#cLc{!zV*>q>F&s0=@)l_+a_Aa5#ejDOhAik%A(oQm(PXL-7
zN?L{lu6bX#tQCqsa6WD?w_V%c9@-bMr;#To73Ewk)YBhddz6`Z*YLH!v`>GB{9pUN
zES`=a_hssrC`GkCfb3kVG3vAioKC46c_;_l!HnC*`G?7mi<I?yyWEP?rn2>^8*8LS
zs{i4b^6p%jrKT|o5P$1%<4dAl4Rg9~37t6{9Z+!m+83y42rxUR0kzt56m9TAr+{nt
zDf>957#cahNIR#2gM>ry)|a-eRMfgV!nwJ*uRx?rvrpP!Iq>o56RQ_QT~BnYw~QRb
zm3Mx!dxEXCZ)M%nS&~{Cq8Q<Z6)7?~b`HxZ#%SDfq-5&u>()<Gyz7sKps?dx->bfV
z)PN@)!}bE*7cHEtdVN6e>w@tAP4Ge)`n=9Yea0x-`1tBGLyrYMR^2X@p}PSDf2C@^
zKT+5kSteP|9XD5&wkiRQ2Q-5Z_*^I%Aa?otaSX3yS9YcLZ!8ZAMkG|eKCA-o7CKhE
zWpw2%RTId*i$d!$&h+v&i6l*jz0LmyF;$YZ139My=bW#uj(t83QCFt6S}+>#H{bRM
zuFlatz$?(|;4hwPhW6Yo^T^c<YD)Kli1xmm6zNcj18xIBZ&x+`U-Yr6VvgYipIKQn
zCw$^et=oA@Uaz-#yf9|f*Tu|NC}^APd7?Gs$U#?HdFCY6@4T4=H&5K)tOk}f))yBb
zp=^jXglrS4r*%J%!Of4TOz8red7wHV_5tR09y`&y_?6Z1W_Fg5iC}sXNyVk>t;X2W
zYDp|L!&ieDB)z~Dg+iJ{<*=P57O#)jy)XK)JGQi4MJcU!n|c0fU5Ck=m-nNWD}Dt5
zN3Zd-2d-^O-k-MdFLb}^?tY@e(BuHMy8+_GQU!AUQ(mn9YCg|tHJ5b|OWee&W7AXx
zqYb3Y*6Dd$x3+DsmU|Y!OVafFCN|zErPbA1duR_V?C3j~@vOv%b(XR9EVle~D+59k
z8yijeoEddlg40u?QQ_WOu8S-PECawYXBJ{|1T0+DM4otkzMl}zfhR6aC0GfBKAkz_
z7Uak2_9dg%n%~Pup2aG(_HZ|b;_%x+y2WY3b;%YZ8*4L)jm@*%%_~;es~Og8Pd*G7
zyBGuDLSh}R475nJ7t7F}tqjcItd?kZo@@nL4Ll~@d?7R7cEa*2mN(_G(ESC+gbsHc
z<%?4Fw%u+NH6SGA^5}hQ$g81!6Nzq1tak{FcifEnK(BsN=tmvPd(xNJ*}Bwm5A<E!
z`dIpqcVW>_SA8IK@WFgZ1cHq()B_KZwd{N6PcWaSIMSDDwqWk%bYsbcaTn$YdQ^-6
zv?1kjj8WaVaWk`z(N(6+T7D#doPMz1%f+9s&$nMBK5M&4!X+sP3^-MfokBT~Qy*us
zD06q^OF12@c@gWXmEuKicLp5YtCpnNTwpAHgs-CM>jA}U;koGEmdb{*QT6F%2Gy!N
z!do>kF~j{bdo|oRL>b`hqTH8VY&+@A$uBr&Dd78db!*$reJAq6Q`cSh>j@H+Hj-Q<
zH@RJG0Z<<vTo2x3jj>PnJ#BA|aoocaEhEqbfiS$%AaWBC0|r$i1YS*j>fAP5Vk0&x
zKU<qOE1B{|B&bz;0IS870RQ`vv&{ykA{VzT)LD);=gKy00Mvq{k9_UX*=XaZ#tu%+
zoqQEAWyFH9E=di17cKfx!;L?5bp*-A?rlyDqo`1Br-~w1xB!a71G*Z#xcQZ!C~c7p
zgt9mD=|Wj9XaSAMM%62eyDxdIx<CA_&GN2Uo^)=kWbUnX|NK;U6my`>|6TpAQsr*l
zB&&6O2#e%-+YGzn?|o-fU(y=iA3i?1ooIG~6&i|NeA`mj=N_n)RAaB1m_$F$%*7(K
zEri@{eDo*|12lhtvhldlDT2+{eyxQb(b3HT<ryRc4i4A=AP!;K8Gw6PRzv9ci_gco
zn6%Mj;ImM=-;DGoUqNCuq}49}!Z6?d!oDnIb=F|!Unq>+B@{<JS1F!HzA<;ix(cJ+
zkq<L+t{v5BM^UtjV7Yo?_n*e*C=qCy>TUUaLQuM+c^17;!%!_aNQS;Y!>Txvt-Wj6
z|3Tqow+xkljYW(745+KtZA&CdVn$5R*i(n7sc$d@4CSr&vru=YIM;!@_YQ#A19NmR
z6bNN;<*at7y&;0{2ws_ezv|x4Fq9Q50+`h#?#oCAwXnRK3`fZNh_qscBh)_7{2WYD
zF0TGl=Ney@VEf=G9=LYJVgX$TE&(Z7=v9<Xib*n%f%3?6px_SU@#crmQ2{g)6~GG#
zGPK(qnRDq-|Lc^{y(?zZ9V^({?->|lKJ`yYGm&#ohv?sq+U?@pfB7Gy@UE$8U30mQ
zbBxUxs|e7$TxS*3VsLA|&k1*n*0x<vtP;p?AXBsQWY((*U2Y(<N1nt<k8aiaafMeV
z>aUe3e&F-h`#n4c`pk+9ITU(`^94%*{0=DIPTBU2@2QZt&MoJ2;kM2WP0#<aVVc1v
z!fl{!y)ZAr+2DtopmdJV20^Ku9JS%K4mFL}h{X<;zGc5#h&xAR+}+DwcenT16nB$V
zlr(f6zk<nQ;KzUI6{C#ye}D4dZEiCMy|;W7IkCGna-R1I_i(p?o?{{|5OE{~d~MaN
zJe9>vbHd1hUR+_I>DASVp?l9PK<4S-7kO5}wo4Z3U<-EyulRNA@rf`6ZBVq}utbFb
z_7~S+SShqkI&%mO6DW$bw1KbLOa84`Sp-8Nl#4St4_qJEuNzg2qg=dV!v+ZSW?jw&
zA7KtY;<|D%(dE`33mLeoZC<v-yXFedpx6EDv+317CkA$%qbT(9g{lggjnbM*qRupJ
zj%82;s)N?Ft5odM+O|CB;pJ2*m(^q|_ZQgMp-yf81YY#wR3i_$aa=E^Et=UrvPef8
z)Y8wSU|i{ha2x(nO!&kp$BGNB_BgmY^%o<<;~bopM!r=KfoFPkhe&y*bKTrX7eKYF
zxv;3`?UgaTZi!rJ#UD`Z<F=}Gaw<e)Q1>Z5C3pXaj^eWm1jHR`-hhWQ!z*@M-e;WZ
zm39<#>nxD5-T8LAiZ)#PKW^6|$V{@n9g`_=8`$Ecaeq#0)}E8EL`j!6QM(;q75dd<
z*G|JV(5A6mNa#V}BLrvzm?{yEiGhO<XEpT^@@%81U!YAbULCq$Mkk4cs|zm{I+TTl
z=fho^@-L#DM<vtz@gs={X_pv0`SjVp&{*^Bfouus-1>ObSaTH<wV-m{b~#Dj$aN$R
z65-eJgkR@qwJ7P$ipf=HR%H5~%5ZIRvpW7hI9g*+T*Gpec>8msQHSY`dw5mA1%Noe
zH?GVRe@Tw9g3YQdCx(wzrj*;jg^Zf@ceMlE*ScjytxM4SvCzMdlJMP4<q=nwvpR)>
zuw=xyfnCSQRbtc;9v=r4Iv1SSX8Z{!r{~LKDdP9zn|QVafd$Q|Q!4MoF?G(M=TQ(r
zt0o_HbW*6eD$06vUZab^*Mm2?lQbh96R?~o2jQf=qPvA&uH#U>@HLmSWv#~&k^$+{
z+MS=aH=8Cek=Y?ZAg}GqI=@{L(IFdF&Xj(7Y2!(%tD$MhxJ0{ginMi~wzt5l-P5p@
z#yJoW^f8VXLxViGPt-b6JYJ+qp^Wn$=*{X|7G>@3g?jgnO2g5HcBviV>>D-GCEJjs
zdY*C<^~<x)&esdBG5MDRnVm0bMC*<&1%Zx0m!ZyNHSo#58u&-zI&(m*k-QNWZNi76
zi%{pkJAVbBJNh)15_#dN>+S}5K2tN|vqDq_J@DiaeA32zq*H@Lx07|KM#4c*vMYM+
zvDVJ?jK@OcfOkq<;j2k1B=jjC(6gC9eWkvBet7DR%RxSTdAY6O*WIlu8l$dc_?B$_
zJ1aQg0_92G{_jcsM`HSu-Wv~_e`W1ygUiI8W(P23J>{vnm1##%B9uppK<ld4MOB(C
zz>X$5*4aOm2u_q8NvqAj&qOpE!7cv%n8iVy0%Qm74QaTbXEp?8jK7%uvvlF<iKM2?
zgO{8HpyRoOY2+q%D3T#R`58lF5T0t_#J5NW&FT#QuJp;ZUv`)~4IN#;#IE)|S|VF+
z=KV19XfR!zRQpX4xAbZQf2snI=flc7Q*MD$jdM3NCbQ{imfN|V8VHP@3~5J~hD$3y
z{)<muYul?mJ-;nSk|nhA_eY*?IPg<>tm+8mF=i9L*&DVhOF(L-Q}B+L%gIBJ4m4{X
zVRZd2R4nxB;HTQ@F_gp?=|DhNtjH$FFtOReIs&-BBceb9GUu@JSCcBem&wZG_`{tH
zPy&`b7F>zz3%|z4LcY|IuJrvsU|NTj?9t=TbrlKDvUY`_i~@P5r_nADD^Ae{J7Bkg
zIcElGUWAM+@aTiY>QD7bEYt{T)YxQ>yuE04xEd>M{7+@%iZ_})2#~c*SFNcI`4@!M
zgXz@<t!>MlyaZsyYV!z%2zhRglR~|=H={hAnk4KF#zM8G>bvh+pL4{*Z<ZxSjo8#t
z0)k^w&ZBWdwT|9x|HIF@mbZ68ObB~4z*onZ$_+n}%M~HBu})x7{QEJLMgye|OA0Z^
z`BL2zmsXr}!1Iore37=t>PEwfKQUT|m4dRXn51i#<E~qB;ts<$uX#=?K?ta%VNd+K
z5KK#0_&Fupu6b@FWvaiH_$1zbY86@Y`QbEo{}Y0OGGp;czImxqefx{Q5LdWvjnvyt
zn)*3P)Zv8(x(fu{20~Vay2F$A!g~p~$mrb1Xy$YGhq%v!L2U&&vN3;H(*1mR)w@Tl
zVW(iM#a|hASsZT!&q@RmL4iajlc&!-LOl`8AL-y00<MJhV{+kE+Lw)u)qTU@>4=W(
zubUXsf|?)W6;d3Je=Qcc!W5k{^V2sW82X29;%%=!5Rkr8VTpCSp}xBcH5_T5H0x_<
z&Dd6>vE<kV8ee;{c8%XtJBOq=?E0Lx1yB!Y-qEd)loa4!0J(AKsJ8hFj=+KMmniG4
z7Z)I?kn9#8pBqbP?t_*|&8}OMTh;1uiKRN~f-ajK4AYpo`0_-9sSXc{CDQco&sBo(
zE?|gPE3{y^u#~M^yAFSMoYSlrO8_#c(SPc%P^5x{Mb~=z#q;x5)eL`mRi)!mp3a+@
zXI15GZNfK442%`-I%em;-lg#+4gf$^z%f8Nj@mb~6T3fus40H4pR5zeSA!B6W?o(G
zf|}ZuA6M^uA0^!DYQClFtDhtvVLEV(s`sMmlN_64k}qJqOHrp7!cF6oL8Z@TP^#t5
zF>`DGQ)(QtotpdLX6p9QX;|w1kum+m?wBO_m6TnZHRV6=sUVz><64?uz1VtnHm&x#
zy`?-u5zq_HIcDNcKiU!~Ok06i1Dd?I+ezLy(&y(p=P}8}V-oXaukcEAWiP;HG3TFG
z<k}e3oB;{)uEn?ctl1Tg_z$)xykL0pG^vfH6XF8*ZCuPM)wwv^k)H94VG$hI9+;uN
zQ>AaC_g%thu->H>Hp53woF~sj7Ze>cGSPHh#x9qTeAmiz<%?cDL8;m1uFIx16GM`N
z_xA~ZN=UW2^04Xmy>>rEYHTDo7+@4a4cC(+`5ZRrskPx!#4*_AmJ?Um+T-KGQ6F?_
zk&eZ=@Suf&BMCLtPFW!GEYBffIU_>8p|VAr#aJu8(dXALAJ8FhkK8oC8t8`w4#$lL
zGQek$So#qtM{w3#YphPm_aN@Lzec(?iQ|yhzc!h$f+ZZOdnmuN(&p{^PFPBNRn4ux
zjzxKTZMH=~LQe9znvqdH`AcyMa>5W4nv$EZ>N88!X}2wJ@2slVL+6h%y|#;mh#_2x
zEdXc=;;Kx{UX1p(w2f}>Ma($}xSqQ@fwJ$l9^^oKYx0uz7U=U!qqZYw9tn-!l;(tO
zQZgAwG&VZ2yAdzSG8+=oT9~q-EF<i*1bKS#*HmpVtLKFGl1M;om<l}5R0t$R1Ih#-
z(T2LYx{?rdZK!tEG$h2UZ)%Z$YiB{&kDc#L6SYlVk^>@<Fm>!=*)|}L2T5kphFc>|
zGMZ2L<hqu9B3~03uQXII!U?yfqIYeBS9GuTDCv3gx6739K}NZ~Vr+_B!m^~NLks*p
zyC$9P{`W-}+BKxbLSykit`x|Xg}HEUctmK`&6<xJ3P&>^IVK8!@~i9f?Pd_NVs5vs
zoCx}WJU+M04oMm4O(3--w)&rz0~GblSh^dUL^d4jhZ8AFXw#XJA&7<CpHX>bQ%z-9
z+!P|_PyqA$_P$?*7@vJ3@2#ggfW)r$__2m3HV@sI$%pzY86+DcZZ3+p`J`+Z(5E?r
zr#GGV41u=dWjai`oRMKvYZZNSvT>#1l+V&k8TwIT2;DLRBdg5i)YyXsElk0XXsdmI
z(|U_#pH)6oc1X*+4HQPk%eOPx*3!i}**#GW??VQ)w5oZaupK_Jdz5IreG<kGpHnwo
znB`oRSJ4WM04u2bq1iE=cD#Sw+h}19)`mEkn<Pw6g~oPgmk<@`#^hR+SQh7M{7(PI
z3hTh0&!6Yul7pbM!064`o`i8Zf{6)x3;{h3wGcMFGs8z^@s+th6$dr0>mVotgxx9n
z3{9M*TT!bnC^M+w{b&acw>@of4s3W#=quGtJ*Q$yQ8r!AxDdXWL{}j)%DV)LK}6*e
zGg})K+Wk)3j0LSfzmDEX`;&M@M~eA`&%0$V*v_J32u%3|ZnXntIadk%nK^ub$hBDt
zhafsj*$8@lI`}9Y*K;Wi3S2~!!TK3{MHuZbq08BO?~{QcEEM>^XUGxR%c`oXt|@X-
zi{|34+g!mHM=bEk7uMrS*(v<c4<wU}7qjcQCJ_RHGIZkQcIL@<ZIv~=PwK}$Q56mk
zxP2pz*x@TurOW5i;j-Fj4}qj~PM_Sa!)j!X77=L_<<JW~MDZDsFz)QDKk{mpnufQV
zWxs08*moDej&f;U1%j==Zn4Z>&)p|)t_k--_5w=S`oTCKmzE>V@A*BYw3v7sCXRwU
zgena3FffgTW&+VS)m^XvKmw>}NU{BOgTGl!0j99+8K$9|gg--+h+KIgh=_app12-6
zb_Tr~9F#-~45flIX~=zOPE+<2$$COzN(?94b^>%}8i-<13F@>ihzO*Gb)aTrzHx;i
zFIc8^w~SuReza!1dX{qovB+g5t(|UnHZO?Cj4q~|SBtgUt8G$q`ych|?`>_y-FNsm
z$%5PL@sr2Vv*>9~S6|Av*c&<V_qJwvHw_V(NYhA}kxTQ^GHx|JZ;0j4nu2dbF9z`p
zcor7v>Ln@dRBO)`Og&d<_KSS@{5Rw=B*Y!Xz2HEB3yuagF46_!51I4BaoU1Vq{Uxr
zDBo^|jq^>~O_YvgzMW0nYMhc`Ykp=11a00zH|3;kQ#-AneAn)6(?umwshrtovXMC(
z2i76yfxNMmXGO!NkmuH^)vL%ukN5j|37nb;LDg2r@P{daP(LPuR7to28_9H;s+D@Z
zl*TYzTv6~(LyGx#Ja9pE>vv(z7>$=W*XNBcm6nEC6fPI&r0%?f1d`A{LMey8FEO&@
z%^`>eH)!3$U{%&n(jC1mWh{V|IaCZStdwaMVHpRsipZ-(W9g3{Kic#@bO~uBsCLIY
z=O`vpER9qa8?#wyb_=GnQTVG0lQSul@I4;Kd(G6mJ7}%;L}Q+M(5M<yQEJnyfyhP{
z)0B-X<<zL^x+Hy`k!wO@7{4-Is{YFz+ibZ1B4Q)}$4K=Z@yQw=*j}=dJe1jyg&9xe
zS<&O!oqH<t(bEk1b~MJ8NvWEMlc9fsT@XTfCb;CDBE5&{6=e;gUUHV_d*`B`70Vlb
z+8#2m4vue}>4prOn3Tb>kR-w=7yGIOhuQ4evk7mAgz8}hdCv2Ejd@d~g713*mVUo|
zoGG!AKBl~rY@@8N7DSV1=act5Ue0lru2S_B5YsL~@|xt>1#RT+@gHLWEBY299MQ^i
z>*spG5c6BzgSVgI#maN`rOZ|Md%j1);)HR_S-w{3=l^I)*#>M(^2S~*DWz?fUjibj
zYH2zk2{v~6`&6mYb>w_j+48Y3um9m{3nA~70nb0i|E-Kr6qIZCc0*<wf+L(L*RaX~
zlY(5tC;R?ZkJ_Jy+h#to@(k!ek*9zPov71)-rPjNo+e9Z<n6CIZn793?=kx5q&2FB
zRH{kkFFU?&zBXBwaE&<H4%J_GXQPymIm$w3Zr&3_=aAeHt@bJ;>(bQKL}P;+_WCcz
zmnG&FmETisTXtQYOfIbEBYcwUtsgUO@@vRLUd6D-8WC;jYv+YcVAnMcW&YF&F>fyP
z_U(M_hL!uFzwq*C@|21zB5D_Krz8CLuc+A8Wv%Jl%v4JeOPkq~PKzoChcxgRc)o0x
za2&JrVmtVAy|2jj$B_uw9alX}F(6N1kl7H&L1GFFUAx||wpYIPY|~u6I6}DrGh9I9
z#`Dm_-6Wk%(&k}yB|49m$dqqN3cWs%p<87CBf}3qfSjTJD~|^=Q=Om(4bp!hXH2&7
zEbYjrl+P%$(=@JRD?3-C>Mbbk(yvh8g{bwt4-fp|0VwD(_>_0pO}GVZsi0K(dU!;B
zHKl0TF0>eso*sU~-0GR5Y6kwps9KebFM%mzYCMlwr9DJG0EN;{xaGkp0zj5Jgc~a)
zNs&mTWBp{GZF@Zvj4C{Ec=31MA<qdf8>Kn~b<7Th$a6vk$Fu?Ly;S*2SJN}%rN>Fy
zHRJGg=AXtHHZMXUG`yi`b7{05onP{LMy2t!EZ6|qRpOHFW-su92kIjyA$4Cjd-<vZ
zH`*EkGrmmB8&dn9df*TK*y;`D)Z7+&VeWS?o-+A;!X-1;`s+BC&>GiUYUrtrOzg)O
z)9zF0-;t0%AP6RP{}K-Tg_#0&l&8cYj&ukxis01n!No4;G|;zwv}h*BZ{0+7)P-<F
zpdd@0=@30h%xyzTDvtDd+i;(2YbJQ5$gVNHR4K)!z%;%+o*I%M)_WkEgEILVbH%IH
zzl!%sbLY(5W}fBaO%jJPVm?D#kl|A96~9r1WJXXX&M`WeJ5$&apG18*KC6A>0o#EW
z-^xO3Hn(?w8|c-pDJnQHtVVA1-8GJ9QZGh;&=EFGmuXtHUo)&4$P*zXV3w{IbWy(>
zURL$q;@`C|Px@Qf$_!QSNiVIV-)zSl?<6!7uA_(|@BtTFQ=c8{db5xxnlwO8a7w`;
zOw={uGfHT`Hhy)Sx~9m2V2H{D&*I%P(_u&-4bb=knb|2@j6qw>M(x#3{$H%ovv6OJ
zzsLk6!+CLD1YRdK1z1E)@5LFP8AWGTna20p5r|xxhOd3!z|6!uNRbbE0Wr7?ec)2%
zGAU6u)rI$s1u5z}YWiyD%@cD!SwCW3cdt&V5(AmK`$F)Y#NYjO+dX|r<Jxy3JDjQY
z7unPOC5v0im?QDIxAdz%exZBjPs2h7@_Voejx^WOxXUxBQMQLwVy6?{m{BovvH1z}
zL*KA%%z=MWe*v<amF((Nm<@XNePYrZ&`IW~7AaFQX6>Xhkc&{(F^CUOf|@=G2rkkH
zox4Dw=lpTt1V+tk5fGU4pscb|!_XY+=xxtV(q^H!Dc7fm1{0_M$IT9$yy6a=q3TeY
zKn>jYnREY-RS-Oab!QXMA;#?RzhsuMF-HC$tob)7f3^{|D;pEb{h;)-jLo^zNz>Hq
zBZyQX>8(qpW=7{r*97V1oTE%wRekDsFz(g#Y*-R8T0y=XZNUrKP?|vCj(3gD{QHTC
zLcVrj@<3nK>-{Qb&TYGzo_(NB1#g>~DFmm(T0?M25D`K_2$})^gk}f_-?=JkGC@-A
zvG!Zm<Ua&V@>w_=dr0A)G4OAk=haPjD>?RGC4Anq6$^cn5AR<jTW&3I!@>VX%4<N9
z4<N!x5E4@OBqdwEO$}xjL4%}E>nP#F)0@9A$1!Woi=zbqsGxU!_(BujS8?)?{{#^n
zwwT%{PmV8W19|HI{)~{6(JjuzS-IBO1i8Ld@2UsNMYDrZ&rPhRvXA22L$o1|Z~)M-
zg<<ofPX{JTymN#F_jn=?L9~D6&$DlZ(cwDCX-Fi50YEc~H;nuXyzK*-_&_ttldxK!
zX^O0sZE?)u`+nQv??tnuujjp$Mp)XqC#r_#WK+*t4TBRJ%T5NGCm$v^MY)t8^1fat
z&UqG#-SKOsy!Zhx_u2P#$vY|y!@gmw>(h7qe}q3gcVLD;{D?Q}@q5RrVylc!CrKX-
z5N|5})}~tjB#)|52nBp?;kC=w1GCv7po9Axdz(-VvgE)_lLxwUMZ+~;7{4AjN#lm&
zQm6!Ajhwc5GhsLcBLSH>(vS@kypngf#Wjr+@@Jxb5<FY?{SHF82~xC}9MEWl4qb0G
z2VK8ng@d4U*ZrdB_DS%z3ryAN<ZR^=#v-Sja!i!_)b|-uv4-mz2t`<Wl)@4R>5Vir
z+{_^s`cKiIXVJUBBd2?OFWt9eVb1V7jeXc@Wl=0=+^9$Nr7Bv%<ytNTbNFo5)R!gf
zVzVIk*yYhX4C5YHz+VGhal)f-!^6Q83o413M9AYm>(>GID4nG+z}g@TuW3*TFeL&A
z|HCQ=2+g!UD9kE<r2cP@bChX{)=}0De_qd2f(T1~eYnw}qLkjNd%wfFO?s!Cp|o6O
zrqat}A<7862xb&PNd_Uj?ZEeF(iz{grR*4i{9j}Xm&+n2NQ7Vl-RQl3WKUW`T_Gdp
zb>WEX%yv+Sy(?w;%1U?YPMPf&W_p<_^VWYIwi+=jJGj|uu9yu<9PgLNZrCCYhDpjm
z6Q>&%JM>ccLYUjRGX2-1Z}u<S_KeX<JntN05<CEzZ4$sA3EKyRg2KYW%VS?{U_{C+
z{kPWty(yuN3GK;Ek|37*id4}%(cirIB{kxM^uy)21p&UFp)WnD6uDr@N=|OP`O|FL
zJcHt`YoZPGtyJ{9&6BAdXL4`s5J)HeEBtA<|G<;4E1TEzBBHn0;{_ofBG9P^w;Cr7
zTp*ws=Kc0crPW?LzuWAAMx?dP7(y^K^86}$lLwio3GkVwLh~zo2w_mO!Df-LQn;Il
z>wY>AjwK*MDg<2Ns!ou|e(Y-hS>^M*IotMp7Mh_R&$uVSL=xIo-2Ji<aEJq-1-=d4
z2!l{&_4iK9K8~Gl{|^ei_gA$7np;2MOAUOAsZpJ_3c`_HuTx8xJTLRW33m_7YT)@%
zx3BsYCB+Bb++wgN;A@+4hDd_As-AuPzfEic)ja{!ZZARzZx!?RZ`#GvqYn9C3RV^~
z5~bN<CpCN-+G_H{&b?D~Z8ecL`>U+E4miwz;_tA#bavieqlb5l6u^#lKIghizw@9F
z<A)iAC<zI`nMi$w#^%skHbt&$i}g0)!LJ*i8@Zq_D`9$kcyi|Wd+P~U-Q8t8ehdo4
zGWoVsX2P7Z76sYj4!o4P^re!v0}CmY3PJkAQ7`~-3gSO))s0iU?3gFWNu-~MbDfhr
z(AwgV*!}cxw~X|~PW=DLtv+9L>_%5pWyhg}R{JW#<-#>w<tPkU=!>>%<C})BS=x*-
zD5pyxm1+ODM7s3!LN=B_8YDT?Ko;^<#kB7y8m=8!QSCY~>(^TyZTI&zA$)8g5pf1!
z6Tr_ev8eL-8EkGW^ZKW5Giwg36t0o#>~INSvmT>WHFF1ii$ZsP6XsyqC4^C$eR--3
zm%&oa<_-4AhuIVrW07!jw}Em+SxRQdXE3YAvIo$~>fu_S7tUaapU=a<6T-(r;7<+-
z%M6VcxUXia|EG6%(FMx-x~iu3J*(%58xI}b_mb#V529}M*v?1lAIUg|1c#`;XQLh6
zYqpA(^yU4ixm<bf)LV~btTP=~H$Kq*5{N4te1K)9y?ejU^i`Ei2?&k>76JnhhNX9b
z<ORQm0_N727HzHC$N%bEu$;qeiE#V2*i*Eysdnq<_181hjWIydr&0_#Kl*Bfi<z0-
zAhSc7c=B?#_O+k-$}ApcnB^Yb{%k!C>=P!(*Bc%;ak@3uElsw{lY#^yU!J;XZAJI-
zgMb@b%zkyMF_w6p$4+IqkM5Y&DkPO&kmAvS;U4Q}#?K#x05VCsGJFwZEkZb`ismIW
zJn*}}-(5?*>lm}<k80&mEc3P|Iy58~z^St#&a)Bi4EPg5?1ukBnI<Rp7va~k9+?!~
z^rThev(8z1$eg~2?ZBPx3kpiFioUr3C?ab9iup_9x<=}GBPfh=UiOlJ`IlV^h9(n`
z4b4Fxxa_r8P?a|#XslAMCtL31Q)2USqBG4)2)4wuVYiHeRgE30-Wk8HuR@sl-p`bX
zmHgmOnAT2MJiJq<*{xkOR$MJbkL94dJ#)#CGtS@lbn(Xv($d5o#>(SCHI9eHy-hTx
zV9<cepj$_IN_40(pX$pMet{D_gdB=U=s324u(dGDQjjRjtwCrA2#*Iuu(-^znp?iU
z<C58!AeVcRAknHP>zGatX1De*F5xy}fh>7L8R5=Y>N5p`FVVrKx$Y$IJdkW~9S2$d
zc<op~hUzA~3Wl_~NTJ`>tdR<?_O$=9t${0>Z7x(o3`P9yS>sx7d<hMT-}&f``<nVh
zkai*LNkn&zzttFT4XYpa-oh8|6Out#u$HkZR91Qez6C-$0K~bV3D}zu$CMwDlg;|c
z5fd%_XYG9aeiGUf2!GPj@PCI49%b%xi9?luWtd(cllUKxzZ0VWBPkibQyupbVn@;o
zrpsr>t>}$Yr5gG*8pr<evlnPInz^ls)up@?-MqsFaI-jbA-MI=y?c=or|tdDY-Drp
z%yy>E4>69?W1BZ<c+ZCKdj6>%0R%SuL9c*34IY{M7!*{~vCMQi_VPu<U0yysZ$ypv
zxt3r1nvtl4@w-!$B<;3u_BI1H`MwJafIWgc$Gk64=CGp@=<5?Ay^m?r|E+Xikra2C
za@oTYl|3`-gt*v<$*GkAM$Gnva5$whcoHVQ55ZwUdLVE;17^iSAz^%f#|)9jv?EJ#
zZiGiginM}A@B1o}O#>eKK3cP$$$vU!0Xs&y2hn=9)z0g?=C+Zy%Qd>ax_DZEP2J~2
zH%sVimA`|E`dsqkpr@g2Dec9OJ(ZXVRE=N)4{C1PDV$^zch3`*dUDc7oupjPgt8U{
zs5nk|&xBbo_h_G?Z}X6pZ-M~36pBmZ0L*%2+FgiAPf!QpgY&(v2UEmhw?6pg7x2vU
zM!qbN(91Bro_Z$WAE{GEmT({jY(rd0bF&yqSqK{w)4r_CVPw&aM_I|nErH+|F>L@&
zMQN`tDW8Jkz~s<}14}5;HEcF`;bs`^u?cyB3ii~y``uGl0%+fwU#_YaHmbh*;Jn<0
zlx>3QBiVJRL5a{ZaK`ZDHZ8`0;Em=v&0n;iynsRk8kj$IQ6Jtpi+X}j<V^5o?1&t%
z-c{bAz!Np5qJWvfipH1bLa+QQZip%Bf?Jg^D?`{x;8Zn-kzn{dLvo6vZ5H~;>7S60
zvBfQbVo={QmA$B20Y>)sF9aA#WYu&GQd5Ly{)f&l@|6CqiqaQ2^w=O~Sw#6{w1qBM
zv9w_0Nw$+>r1_NF<@--a6X1m6`A9vWvBX)=PJ(mstI#@2@J!y#3&I`j;c=5U2#C1u
z&ab`u9NZ3_BKwL|)w=y=_GW$(y<i&<%sHK8s_;jfI&E_U+kyp=r(o_MyNWJh(zq7g
zTwbFy5n&VXgMJX#`pJcCO!qS8WD~hPSOvf+Mn*o&jaU`IiR6}Jma`Ro%=Jc9L~s4R
z?4;+z>bi73?lt0S8}w$?u1L0Ce$AjINWe;4HX`K5&h$JYxP!2`J+j2VzBgSSaG<EP
zfcQGxy#kbXTB+!(h3O~D-LMzRq-)P0wt|_>(p3aVF;QH-<N9O&o(U~;YErkzSiI#T
z>2m9%PnSsFVdP~MTK}C}llE%QU8|2y3fy}ZzRyRN%^D5wF`k4cPID)ftm>D4_}1pi
z=1?FS6+Nl&Il?<kw}?A@rXN`#MQXO+gQBr#74n^quR`iJc8_Um`5f^d-G1bts8n-@
z(Ue8v#)+m{jWcE&G}OL~Wo@srZDu1RW|QrZ$`5u+8^G<m)(|i}kZZ#JiM(SXu$=%j
zSuO6NwFq*FTT+7((D@I9i*Gh&s)60W?!tErbq-UY1L>h!*wJj!rsPAuU6ZG888w?y
z9O)(r?5Ol3KK4%4Gf=)-kX%1*;AH47kbPxyeMUGV{JeaTKK+=W_w4pd;X6*cHtEwk
zX_XQVVwa2BD)qO}=N^_ZZ7gYfZG7Jj;jWi@4EpD3?our?Xq}%<?^wq^wWO#Xa*G>N
z9?chSHym?!?%&RC&9utd&c4=?Lo2C!I#u`1Ilz&c`}nAtU(PDyM5`gozxAoYDq6$N
z@hw9^UQOOTtE)OT*Eui;++;W0A55Hp_JsHb5bA(n3v~F7F_~w%8-IzTP3p_`_%hZy
zACX3dFoVRL6bSH_Xn+B3qZcxe?66Y1<k>peyo^v?G){gbtTC%SHdvg+6JyPotK3=}
zG^$U}m~qU=R7ENF7Gg_sxRbVH7@UHIm$44n6!C5sC|o2?oeGUmR2=)RB@<eq>JZl)
zb)$dvyYEY*_cjsT&KnO@hVTOE$+g&v#-2sW2gBu9h_6%4Ku`U!RlfC8w~X<7>F9cg
zIiGBd6A0QqM`_q5XCWJ+{jSNH%JIi(KZCvu<uIdDm7@1kPkHWK%U@gQtSqcIhgdRr
zv;<%?#LgVOh9VeS2KWw<e%xujo!RZl?0Bv&nj4&^a_+~m-KLA#vup^hEZ<2V<J<GX
z*D!QB=&hZ$HvCdhMi-wh<3+BOU^SmZbN*@f+4R{u9~ZhVP(y5oDS7#sZEF8B(DYBr
zhGXil`K!XeXSfN&JAyo;{I-EQ-vLwG0sYgV<>|YbRl;?jky6pTfKKl>o)8LtFeCHb
zuI?gJeQ{P!tLh+`1+KYKq-gCc;8MSy{m?o!DyGPjPC_*<SMF`Fv)TiefO+mPhadp=
zngQZy#hG{G$@KAlpEvPkODVQ%n7}>TWs}YOPb4b}?u;ptFp#%V-egWT!8?8CTIPB!
zAW*z&i40|Sp~e@>#!WPnzR^xhcdUlM=3wg96WJ>L6781BC0|2-3WNdXw~C(iU$Y4j
zXBGnFxl2FYD+ty0togGVMo|!T&*XuGF_HG>Th&ucBlW-O7HfPq{OM+gUR|rdP2*Za
zp!EbfhkXvo#dL{?2EVo^YcNsc51FtD^?YaR7YI;I5U{k)u!t4*$;Xb&Hdth#oyAR2
zxi_Q|()X4->G_vjw0yu=-A&dGOk=wKt^@#-a;)H%eP<bbA^c)c6Qv+6N7Jp=1*OnU
zYo96gjM}TfcG1jVUc6sL-}{7ccH;t%s{<$t{s=FJlJHL^cHfzG#nt5{wptDQ_dN0Y
z5ICTc<jT&OC_UIj1r6Kc-jwB-nA<&IT`<f)aQ+_$^*51a+8Eu*oP~S=X)KVOpuNR8
zhs%-heJ3P9c6M;J)cK!@mOLi)v=3%;nscvG3GA}LyZ9;kJYUVeS@huGV8Z_b71>C4
zC*gH4+NZbCIf~GcvYNaVQZnTnil_CD$#D-qvLDSFciAAhBhsT{|3UUTec8_;sD`@a
z=7$GG8=92JnR?Eqvm?W*v-8JADH*8Yz{kEi<=h^|l+l5ZMS<&EjA)|Tk1KB_sQxpZ
z<CaDww7kwADx5faVpprZ&5ym3AzeZYn1=92SUvs}@MEN^<>(Vl&zd8TRuT^PW`hNS
z%<3N81wO14J|eTlGv{bvhJ}JP+m#)^H~JxK09RJ&v+J@z)}Z{MZc~eWwE~~0^5*yV
z_8IS5&>mX*r!1WL<fEVBlL66lmEW)@VoUg+&zUhh;_Y{M9N%9FL|4SnQMGG?SbTVM
zd*Qc&gGzJ5+a=mBu4rGDT%v?{-aPFi)(F?(CjwiT1P!=Y_h`3)17o8(<7q**8#h0^
zXY2_PRe}bQ(_r0Vw}0J9T9Fx@Ob|7_lDPb<VM3a!G8%6v_e@@$A;*RC9cm%`b2_3m
zrvA`cf~7_fqexg(?+ex6VD-?Ig{&BBqHaXF$VxGQ)#sW4bUFyE*x)kwu06?Zs(%5&
z%(CVgWJCg{W@EWqQ`%me9<-5>yx5dRquEtoxrPX5OBc2WUOqn_g@KqgMcwb!wnl0Z
z>8j51pqe$({Nn8!<;liY)Ff(juh(r$1EWjK*y1jJsr1I9>J)E<I%Zx456a8Q=wcdI
zD9hM_ASNxRq)<wS!ZX@CgTXegVK=v@E?h+M^C;$dwY5)IhNWXUyt*k;UYG^n<$-<9
zvb$r!<6j|*{x>nr6ZF?~U?qU;%dZZw*jQ%x9R?R{0!h_^RQu+;2Os6C|0jlb{rQ&T
z=_!Aq3z%P~jacdfHaQE_3~^+O%viGjE!P23s@OlNW6|mpa}FuMBCv~(l(uk?5}sO`
zW^69v9s<Qp{x|E3WBgKMja5bRw%y3``OiAo774{&FV5XF{5rqgN#O??V@b$n>!7b1
zJKHu{!WT7`WjND8p<{voF?pW}iOB2MtvGX%X<@_)Cv<fYe@$CJ?(HtZ)L%H^U<NoG
zYozkK_Pe7?c;MzkL5q1Gf8AVu##MyNP+}2|_p`Pi;ru2vl&nFXJ9jxG3I`IMFL+w1
z8)T?QBhd}a(nVRFG#~KYc4ae6F2HPp)rsAt9ntew?Sg)qB`(~*Hx$XG{Z4%WTY1<3
zTn9db19Dr9A0oYWL|Hb0Y_(p60NVn@m}jlIHk?t6GWYb|`F1Cu!1}}MCM3<g-(Ch#
zH}1Ppad9ZD$Zy{V2IZ#pRKHPkwdVMfjTTq4UMbE)J%@UhIN=cI0r4o{9>|XH0YM*=
zqUV?PvjN_``TCW>hfKT6QQnAkhNqqD$mi}#b0W_X;C)8T-9ev6K0%u~3G}1s#~m%W
z-GAb@^OD6d>G+g=@;rkW?Z9_e&;-fQGv*%~KB-@^-@(Z%)oAP-p80_!R(f*oaeIqW
z__zM=+TFl`_bDUF<OR0p!=!y_?d?rji$#47yK8j$myQKk`p-!3nD~2AGJmo*7HMR*
zJUSM~U)QfSgiVteIL1seA<lOg+hCK(vBbLnoGtRk{8#P9;~eFT0mz6U#rz#Rq`Whm
zL&&RM(r~*W!Fub~t;V~)4_tI~@}C@%>VKRtE<D-4H9!#e+fD7jBm<EIDfCfBd{0xv
zi~@{o_%Yu;h!$#@*gXf}Ebz<hS>g`v<6pS*Z`gQBNub2prk43>o89E4LpNEwJ=ZrC
zTx6r+^NyaF!b^x~q^Z_S-8_|!jKbe)VO~q>#sJ!Ku^|cL-M&4UwC@%}Iq{QDbB4`{
z`|T>H=<|kk!`u1N(m2ItVOayBN1MVcEr5iJatC}95KQCzzx&(DmMvS~d~0E$_q=zC
z^A5Ta$QCP?Ml+!li=0qlfU~f_s;?U^(%?4}*W0-FAXsTQVw9H|7cpnI%;fx(Dih1l
zEs`_MU%Qh?UE2!@idJ_@#K`&+dnGV~IfRmDfQ%v)e*a45s($}}FB5(MIE1h{)f-K5
z?f4AwIM?AHUjx3stNxm~UhDs1>`lOG%>VcCsFW;4Nhpc-QVDHZgeck*LZy^6B!w1j
zDk(+VIcd?Nv>-&Gv?49%gwT+*sZNqMooLbazn>>FX6F0-{;%I$&(+76%Y5cM=e@jM
z_v>DA+oZYZlPpE{2FbFwdemkz3s9jG>~M6u5cWd!gyFOr_uEr+*1z#vO9}KTjeTk4
z$qs4}%dZC4P7x><zGna&iMIq(f0M!og`#hOQirU9F$-n1ElXk(rv5mrxWs}(p)!5m
zX;`mqH+Bn}BW|;Pjs1S`5RmNfgQGRdB{vaF1l=ZgIg1y;q(5nQ%_;2Y66YsA@3$Gd
z#K{#lWE4B|{KJ;?#>c;GNg00+Xj)>=Lmoi^JB5SpavA;&mk((9Pmdc-I0?&0%?Dhu
z!GH|q#bpMrKU}%yQqk8Uu5!JGg$ZBR-pbz`gwvL|%MsNKE-?(uAmJkoC^5>!91Cqq
z;!XOc82156&{AR`G5XQZYcU3q&nEx)fwSR3DxY}j{fidDaR*!rinNzW9za_^hb50N
z-Cb12juH-wJsBU^uE|}AxzJ&qaq@U<+d|XBe$y}Z-Ipx(qLUIB@P#32`ei}3J+XK4
zXa0~X>A`Q8h*m}&?7@_53%4;RedCtzRyKQG6j=mC>qdVbbN+eIK>M+E+DT|MVZZ?l
z3v)<3V88>+0wWaERE4XAVfJ5u#_F_e!`}xvAqN2AL{}8-B3Qt{$z;O<0?#D@UO+hb
z<h{z)Q_IwymJTN6O%sLdb4TLU`O{9kZk~T6R**Sz=EWAMKk&I9efqiQ#nE_JH*H9T
zGjC#LZ+HbJDd8vGIp}i>aHb)0>R&#Bzeg|%?Tnzc8kY2-hZmR+m-+zZ$vg7yJwcYf
z@Z-MW?9OiiwfNHuf5uGx0Ee<XKJZ*oq%bMjhujtqzCP%m*yw;VVb1}W5>e%#@iqOO
zz7X<7Eix*X40QTsVA40CJht!Fty{0i%k0zZV4?B%<iM#dhvPCj&;>Kpip`G#9S<Jo
z7^=I@a_ExYwd1ymJ`I(OtM`ucbPimEB6t06@$K~b=1?hmSnY*K_YGld8j%)3C%5#o
zzJB$-#f0yCiAG-)F7Cf=#%&V?4^cLcJD~}???`pYZ5M1)S9y%4wJu?o(gYEDH2v*u
z^X8&RKSfhWGyysj12#C(V9FR;k;W@EGVXIUf&U-mxkWmLghn6!*@+@Nphm(vdR8uJ
zsSM5>kSfE(7n?mt{>}T|(@ruco<u`vygqKh#+Xzl=eY|FeD<mu7KfO99}@fR16!l#
zVio&_f>icJC1`B3=V17Df@5K^#Pw+Bb~he~@cHlwziE1*<Se6(EC|@{Yty%+^Ou+K
zf~lCohsH_|o@{8^Opstl#M-K8n+yDzzQ~^FJH3!Np-WGW^UiWlTZQq$j1vJcXb=dX
zLxma@+3Wb*yau?)BSYdJ9@2|U7^LF(ioG_cKZbIH9_mY1gXc}aQY%-s&*&C0ar$aj
z_DIvcX9*zCwcb5B)2B!=3qhWcgY5uQZ=o7KcKbCm&w{ORl{dE7PU}BpF>7n*v4|+<
z#^G$O(eU}yy#}Uk5yxYLN$t7lbGykoIv;*(Qi+yxWvg@A%Wm{w&!n`=|6)x2<BB5q
zg6}*oo-2<_zSmu{g%5OKe}7kXoIF^)zoeQottJK%pc7yK)#r9K>K`!d>}3?X@nfr;
zHBq`Myi{&7R_zNL9SrYyeew3Z($cAS5?WzdeXFMq_f=u4w)5)-%X<ymaKJm!HNO2R
zBwZAoNZ;I$xRP)%S7%evi9aYHW$nyt;DSlCye8`wLld_uIw4`{{W!4rYV&mOzjSK9
zyGw!pL1iyrNX=g!<a|;il;zKJ2p-Fn2##;_LyiYv*NZa&NN6ZYbyf&rCnLQCrN)dH
zmsrF%e-BwIJ|a2|!XEHNQ;kn%N;>3Oh*I=`nJl&TUPQA~q&^jU@OZk<!H@MH`B?U1
z<|Zs;IB+#b125e_t3O<U`_QE^lCgfVVVnJ!%%U}&#|$-lY_PUHvrPB7GM+QSC0Oe>
zbawk5>-}#&&hK1|UB+52H~wj<M9_c6yRc!$XWRaKYdIwx(gcXw@ny!5P2NA@4~hAL
zm%n@Z%Z6b_tLZJZvFwPu#5{qZb(0itqRPgizWxU0F&Ln941TcqO@i8%J)_@6|BZKr
ziwFW^42~`M&+`x@K*R}}o6o7)G<><f6pdWzC3j5^ZgVG|AKkLhi|i&jzgMH{2=fGW
zahKdVo2!n8A}VA%oq4a#@spJ0nM>51o7R(0-5Gx#9`SHnCgrehx|d^_MAs}v#qns@
zImzNu1pOzE2+R7{l%;Ep>5m75#vHUDxKAC|@{2{d<)zJ@$5MM2qSx}OB(NmAaWov0
zpRbQjU+v2raWbi*y+x}f^X={fRS`DX8dqXeh^iVl6N+qHQP61-Mu~j(BF~$COY~On
zUmK<SgqSs4fTj=kF=$Z8h9D`GWWxnO1+^zqG%junNM(i--l%`4DF(n~3$g}f6~KmH
zmY*CspDXsl<}LTvy7DI#x)!jN+xOY}RS35Aqd!7H5NLKmJ`n4=G4o2gsz*Y4af9{u
z|1A`Q(C7NSzy(!t|1ULa%EmVpw4vDflRd`=r;JRljgEQ|pBV!Q_H#fD=Z$d5&uK|g
zk`QLem&NP^lr|zDBxYyDMQTl_B8W0zLxTRA>A`Yx<*KE}tk{W=mzlr56h=p+etdMf
z?Ifp;Nn+K$=JHi<yC3kbl?iRv-<{d)1-DyoqSEVgAj|?Y&@RsYV7p0c7Ygmd06qXE
z%k=kaFIc+O4&;8nGuGQnP<oPvzZY`W;i7t-Gxcdx_H29A1Pu-ViR<QF8M-@7z!I$X
z^<_WNH2Vy6*~BUg+y77cp8~e^kNKVdX)^&gWNlp9&#KH#oovc?S}y4<gFf4O>b1s1
zgm8)Q=<;e|GWL!%MBQH0?1uku10;qw&87M`e77Gvr1p0Sx@fJ$`E!m5@Ess+MNpHg
zzxsn%{g|x1LL|oMhhxu4?_+pE*11Wc;NIy!TW8iP6(Vu1nw6cx|5`)O)5SEmO|3GH
z?ZxsV6o5*~K@<Gmm)Dt=9wEz5_tpLPd;1+Us*P2(E>l37gvuka5`q>AHRju?(wF=P
zt7fk2{3Kd&>{?7x1?`*!>R_KJ-~4tcL;cQ-D0Sv4D+;H~0*vQjM1`M<*ZxyC-&vQ-
zJ8%!8Mqw|V!`x9?{E4*p1OvVbT{VKkc%_uD8bFaQVeH1c<#)g@6y+SQwz<eiGN>DS
zx4{5s9<DLMK!cpI5al=&_`#AsJ>ZD6k4&q4t=p#;G<yZ&fnfOjh3k}ikd(gKQ}d|Y
zg@?}TT`LxlSrK<8bl6a9cIc9+wUa+h6ucgl<Ffb6ie7@aBt&Pr2IP8$GB+6W&!bS~
zmwx`v_%^kyZOw@$kG)zewn5eMe+^itU$=hbLvD(Q>|D39ZV5yFIQ0$<gr`lzaBIMO
zvgJUvgsFed#ff$ESph8J7Q=&f8v7j_3L-?5oUcON@{0A#@7|j4l4AmD9Mbv}Gzxrh
zT;Z!zYBxP~RZIH3t&@|8MI>5}M&aV61b8aR#YdKBg_n?hrlBGGVm86|RIpGAny1m$
zBx+w+`R5Izb8Zf5I-ZcLOBnN0wqB@yELIf_gA_HLh843?>L`2JY}#HrEYLDu=<U1Z
z_WbXU)@d{teOxhdxVS+3UN>#2R)lX>;m-v+%2N(~p^d9<A84(uT72|uZlv?ig0*{`
zyT+oE>qiS;?zJBHJbmUz!b>?nT3>I6s8Uh$q~*&D2=}KY>W?o*ppbLp2IGakDZPEA
zJ-=W_tGTBQ7%9I-$wX|<jK2!uoQrV^%r!v~B>CSR&+UE}?mQ6sqtHdQfZG(#L^FzK
zTopJb2R&X?9a%HnxmPpLZ{p23qSJPAJ&8=hzF5(~pQ+HzwmweytaQXigH&FrY}D$X
zl|LQ31YTe)^t%_S6^w;jT2iYw#Z+va%I{9P80jv)f}OtNaeC1_x#EjOsr`$(8E{qO
zn>alGRP8s-2yIO{Yvj!mEFwwc#NOh^ceIjv<U7aqZyozMPvG%|N1vn3pO%TrmoWBd
zz87;*TxHF*IbOY{x&90Yq+a{;+Ad@ddm;F7v@69!<Ml%m1+tX)B?ENv!O*5k7@t56
z%ESGz*ix9k<{u==)rGEPa`G>S{oT?t#DkgnJCUFR1_nSGJb{055(A~bT?{h#^nG7F
zCrJp$q~cEdKM}RRw@wQS<aM2WR4%XdVa3bwcM+dd4<<b?OSG+$PQ&DF4v5{oHZ9=v
z+!1}gYI8%q^}t@)Uv{Ilz0H*j$b1$fMB@l!Dd8Xds%mFzzP2|4g^He*ndbA;FYw1*
z_=QE;<-6p1UG_dwbe5K%G)Wkm+&fV=gGb7I+i0plMv4)WLQIt~V!?O`4@QO%_hT%t
zBKT~>%YRsRgnZelNV;g_N`>tTCM-ymf>I{Q5S=eEmW916;OKaLBy8Q^vO0*^JkFEv
zQV=b+%fR>ZgW(-R*nS&Q50%P4To)E|MqgZt`@VOyE7wqz;`A0b;=r`}EG`vhaWwIn
z_a#{giK3LgNcQu^Yoe+$+*kfRZVB=De`s9iE+)onR-xWj<Kp$aNMEc;Cdz)z<f28R
z2~q+i74Xf8e25<i`v$-iP`e@m9lp}bFQZKT>t+m*+}8i(!(^yCE`QQ`8S^P!3r>Hi
z_cygE>U8{<R%R;A$9XaV6?D`N*T`<$boENn-ByV=iUp*zS^h-lmQ5N=rfBGdqaqk2
z3jKg?%Iu0>3ob=tx8SOVpu=j~o^r-e_-^(TZfE?jQE@+m$AQ58kDhhvz7;-m3)%99
zBX9GZyzgk%w`0On2mEuAl-8WbjLg{loo8ebChz#J56W#KU3F0yAnx0#@l{~kKh8!?
z(shTVEa6DSlj(@>7532D=H|F4a9DDRsOYF}iRdr&&$_dEs@M|%g=~g2*YT%>Q*`xp
zHu2_bL3(C4+H4)Z{IMXZ(s-fC##tY52=J3`QX>gA>E1rFVU$ibdl+iX|J*Noev`hT
z?Qrjf7^TvsbMuRa9_;$%g8wg3Ylw^MIfL1K`LLX;%Df*!nN{nbn=4KdhaF#DSaG&_
z;N8SL0;FIsbo-b{m`SZbR~qG-Ri}-vj&vF#1ikRrG6C-jbWo1M&i#KMcdNOoa-Z_R
zK2u>nXtu!UBuwEdOmM@yh6+4~#9_4;#Q7?&iIjnTo74(8aR6NpyQ}d+0%u-b+w{_A
z`A|}7k@jistp==c>bS~II+WPn@GiN1*`WK6(HLQG05VpIxG$w0Bq)@i37?U?<%{>N
z)3I1pjJ<A!j4+4Rb6%YH-7sevc~Ubn<VMm9H6IaYASeY3s3j>Qow46ftr#tB=I6gK
zSo{6uiLi=%rN$4fRYDNA4-B25)h9+6C{O-uq}Mk@k3X?LW_Rtj+}M?luF^i{P@#wa
zrxDuI{-RRffwiJbrp(uJU9pD>0G}AZBpRjN(lJ25Q85G4fT(+efkOhwkpDnvaf<-B
zn;FSqDa{ZJaGjvu#7yT3JQs%5emNRw|Cx0^dba(~&}6JGhJQd<bK9+b4H^cXE-JT(
zs3_m@lI{CS4|0F~e$vRq%0*(`QVIDMlj==js-+!_C3VvI>ryr%KJDbS;>$jJd%L^#
z2l!oDs?5nRVLfs1LLR05jAn<OHJCZGnVVDUf>czGUh9?kbGTUifr{$pFNL-*kHoeb
zCKh+6Ou=DV?<sRGJ$pcBGP%BD-g#aLj-r8|U$eD>%T5ig@qZPdL{@?TqcEC+N)8P%
z`UX6}ckpXGIQB9ftv&?G3;iakJ=laZ6uhPxE$K|by-W$5grr%#f{KeSw|b8`d^6G<
zzInswuo&4ct<Y<4kNgvURN1!n?tazAW6Sln`^<h|S;h<;kSq|8%JJA980jpp8K|up
zrtN<$G<V?Q?H0dHR2fbgMCLusw6Hvv&Ga;Qs#U2lGJjy`i(gViOsR9Ag8UM1l2`oc
zD5)jD`H=5h0WHBU=VQVy%*@y(dR3vQq16tknA)DM0=ptMrT#)}*>05)OCmmHA}?jM
z(ZBYBk^`ZqC`SzmDLAOZB~}Ff11b_Bqk*2H=f}04pFuyo&t3idj66DGccTe{Ls)Qv
zUIA#2oK}Ht4topSiEyFD{m5b1`+*ov*6o?=r*RY>7q}9X*a>4^6HI7ic0FazS=aG%
zW7yas3ywo9_wFhD==){pgeB4r)WAnk_R9VH#ivf{&}|#m3tS1JPqB9Hx)3lY-x|H6
zm)A#Uufd+e86DvqdF{#PsZ;WcULi@$q~ezgX@81YC9Qwn@hOM>$kWMlR)LU4<IdyV
z_h>ZNm)eS#wLJRnJaO1M^*2C*st3l@D0ToKU|9Fd*eSR?!|b;QYsk@&UoIcqSSu`$
z%<;;sxZr<P**VQv5{g;)nTak2ODXP=WX28ft>xR4lc&ak^@$;1n_Li91;H7Go1UAS
zQS0FSAPyCshK5~zwR6c{!8M&S$5BqcTfa2-T=nc~1MZRa0tA2cq=89z>vu-qbL0l`
zf}0@poVH7e^O5zpC-PrD5<Sy03Fk*CWi>hb1uL~mP!sgS(1#!)+YRh$Un;Q+tZsVf
z#z7=6^>&v|2se4n6MiMgrnZJzOVW*?`NyONpO`Q#n{oER@<&MxjPH@t4xhr`H%0vw
zuf=d2#tp#&5;hpvgq4?5;{|&QK=?sh{6WBAK0c812{QDy1Adb!=pSQ_XkE9J(E*TM
z+5Y;^An`Zdi>feoSIP>$tS2tw*Qwlm>|7M{UFM<T7wjW>OW&T`Cz#W(X?LSn--JKU
zxS{AyC2gs@EdcA8=?oE|Me;Fg(GG|~EGc%0TN$5LybuDp=v2QXc8KR`>;E*DY%?+n
z2hphDvxBoY8|ZUA<Ci2=kO>HP6H@<5$Yo(Ph!YU=IiN)tPyGifvOM<e$9j<ZfRB*g
zWu@zn5(D^{olsoP3z`js7`fQ6Z!dE4{eIADAs-5v-Z$<7?2k%<1f*uN@VuJFmHCEe
zwsfvzSum8E+o#4xToGiRD)pYsr6}f-hj8JU!xL7!FLK99bWX1}*xy&sYp9FK87n#q
zKIVVqviFi|qXr_h82`2(wQL^7l!9E~76sf*-_>?Y8dL54T?-JxDIl@SZ@hAcckecv
z`%V}+A|smU2T%i&x>>mGlXk<%3;8GXxyWFy$SR_WK)<+b1F^?xcCZ}>^Mbe;7cN`~
zV%2KRo}QlO@v`%N-<JAay1T5_e-%kUe)=XHOGXA`*rLL1S$Z!zY>}2~f9x#a&(aqa
zOR|UqaN|5)Y7m?e=M358PMnc5<)>`oLD$aAiP^~dp*;SjuL~f81;xQUo90ntQ?m0T
zR((rv7E86>=kFR>5&)onxuloYK#hENm(n+%qR0F{3x&K>Y9=z&Esq_w_5wL^O4B;F
z>Aj{>^Y<QS8Ec1K{j8I7=vNO8b#|Q@oQ;=d)0(X1)lshb>f$rSl@Y{QcU)<vJ~WYw
z#V-<dJ(0gF&-?rE-!=$CteNQ^$|w+Z@Dr0h04)e2k*#SzT(Zbh)$2UWw|Raz-p-Jn
zY*+%|;j!%1^xl<;302+{$)e{Q<diOcMTpGU<q@t9(N38!dwstNEt8a(74>XF_YC|R
zcVySHBV13f!N=jAZ!|MOMOE<WY!&CCpACnU35kqylEsQwOE?<k6X&KseLuFZ+beVc
zL{<O5Q(Z5eUQkOWqLVx9TI)M~-t7_RvM!hqlX*54UHsh?J7ewR>BjGBd@PXE#E%az
z{&tH0WBse2LeEMyrn@4hWr1MFEN~YqV{BW%g53<$FDwxx4DFWwZJGdx5t=iMH$dPO
z`i-ho#M)cTbV~#+C)M}~@|Q8F2byLhfLe;^pAgDs{>r^EL*9px^{p&<DrHn_zXO3G
zGbQ%fl&oG$jQ7$GDQ(_((IQszz_t+-ilUw7yfJsWec#JzaefrqPT%`tgK=`pT+R_8
z|M6E(H1;=pRtD^gOTQFWyB`<nfXd%C{WDoR%Ou9jc7H?MFiF{1)nTW#%e$%HN?v$N
z;{0ldYlpMRJT3WSR%LKN8Zs*GKkUFHY^uek(fyH`gQ$rzOlcRXMkpVlFg%fugMd#b
zK8*0V`1O}$>3JU15P%?{!vs(cJ{Ec<yy3bHMA{y<3-S3gOy^_Dl_&2l`#DPSn)JD~
zkLqTD=G3Rko~MgUFkK6Gq2)*304Ta4d$^O0^fufH(ynh~NdT@w5u#Sm&k_n_6Zi9o
zH825P>6OQsby}9Hefx{ljjc|(^SshvpNX0bjF41*?Sw<Z<lwJ)*DtHL8O;$vZX%rC
zUJvZ0dSATVpnRpk0JHDJ`J7Ei)8feSX{qXG#3qIu7nlQQGt9Cc?#EY-2ZJGZ-@emo
zM5p{aRtj7TDAmZ^bs!3u?qM2~VG6zmIWe(Q?AABSZ2s9hPkx6F*UIXYBllxbHoICV
z1446U-a6EvVQ#U|{^3*S!o7>9Gkr2RJvmklu;Hq=sBcS7hW>1Q*1EBux`o4u6AQev
z>nQms;#wSZki=nKJ7Yz>;Mt{#0G&$TdUj!RFETxDU2?ME{zJ=8TqBIQ$D^AoX}&oe
z`w%OG8nVgefJ^RojuFj>h^`Z(1^moi1m^~pFgV?zE5vq&FFAHJpb>0p{0{3sW||s6
zgs?E~_7M465mkWJ(GEjZXGSr*$p={mj>I(0&STxgxBS(tfRD7eTDwswU2Un<)wq;!
zC}u@!6g+#+9=Jfybxw5V?M9fkx6TBedn#$7e8BfgfM8Z~^2#t5*A{rsL!)xD`>dzK
zwd2Y#;5Gw?bX9vd?agzr@-O+Gbrryvi-bmq2uX?yQ?C6R7>B-hkHoqtep*QFKtx5u
zsk$ubiB_wzvUaqsw&llCS1|(L;^v-lg~0U=cpRV^eu2UtSAF5#t+zd+e-BC)RZzT-
z*DK72Jpv0pFqBqAXGHc#81@mj03k2*))elPvfeeejf>3ZJC=_D?!f?(R(mDV=6)|P
zVKEFZhr#dFvjl8BuS&AoccRiGmv>!>EtdD%eqF+8|5Ds`I;b1{<xp6+EE8RIMav9d
zm+L;S0a%#!B?!Pxx)F{207^WYUn=pgV}G%cj=a2eX9g}myZ}x*nSNJyzgO?5==m|4
z?AAZIJYPd&8H*+%p3gR#(h#oSrp+mW&Yz&6L5qW07G^0kpa7tuJ=qx5!|?06rseNW
zk*UYIPB{b|704LYkO4cGkP%j`E0j-wt`8s)L$hrj@9S#`cPWqYK%^&LhEr>SO8t2l
zaEBu-Ju~<+G~=Fe$iCZwcLM+6i*t_LxxTSExb;kq0FJj{0FB;P0`L==qQSsUEA;5k
zJRdy+SLgq>hIuUKk{e-g81&);(EB3+WcZA#8-iI9R4jf5*0(TSocK_n5vn4$A|Ru`
z?agK`FD!Me^ZDiA2LIzw6Lk$i48e9VwQ!JQVG{;GTT(wgr{NguAtG}J^22`n-e`4k
z^*G$ejlcI?FQ2{N*YFr8NQW2X>3W{GXWi{kM>S_5t~~ZEeAgHdg~#~k#!WoEVivs=
z2@-}ThOD|<Ro1-gnvuu<<?y7irfkO1ZO1z+(y3JG2;It;MkZ@~JY9>`-V)pEn0SfC
zs{Hj<Has*yQeg@`oBn9wb5W~c#Is4JxS(yLseo)vf)`IA?l4enVf=$6rJ3Xj;KTQv
zHvJRzOXEQlQSnDBxN_qLgX{!17!LO{Lis0}pW$!O*syFid4Ks(s=~iOzZg=3a#RXY
zc~e1p1q<g<q-e{<U6?#!KJB+%PVb<*8q1XRSc%}H^DmQ@U=a;E7Rvk{xsC}-2c|5$
z|NN8dj28|V`H(WZ+oSc*mHPW8GF`yNEFh(J2mF@$_vZd{GDHY*4jRC5`TXoNv<`Qe
zvy<bHW23@aXZ^~LW!%GP3rHsfq?q%~pt!KIdQJ;`dhVaC%-vNYY%`81_{i`=fhh{&
z;d!1dvx1lYjL;1I+}6&uGXIi}=3Y_j%Q>!OhA%v9xBl6!hcu1fyWBUU3SFNcW1kiG
z9KI-_G|?l4Td=o<46n(G6UpeEZOQ3~O0J|axhD@^E57r8XFa-ZIdHXo1-;(ty9?9K
z0Au_;*S{Iub!gK(xnjiU)@lW>w~19fbH$EZmzBGfy1ln<@9Y(s`qXkj^KQJ7Sf+Jj
zKSwy3lCd2FUUah<ir{LRS?myv&iHkn{k^dZmqScPdV);)wt5Px!1@TVmkA$y@nDjG
z2NP0rb0ko#=@c3rYfO1k@5?+jS}r_k|BG|D%|k0o43n`fb~lb6UW}uD8@)jcH0$O&
zwzg;ESg3zlsxKG`J(TXvPM-JksqdhmPqDnGf;vy%*ZPHqjP_LO-gEbCAO5w|!8ryH
z@(EY@Iq!d|Xo;rIZH2^mNpj6AZF822C6ZM(XI(MvFMeJ>;IndU+t;@>rYDK?Kz5>x
z+rkNy$HCPV&UPfRV>R9eS*|7=*nw~G*I#?X>@7$|8X?Y7m_H-Ek<fpI1D~XJ;Mku}
zbtjVPh_tj6Wh2tESKoKaeqIonE29I0uc+K!zNM`8?&-3ds-B-N38m#9WN(arxMPtq
zVo|w*S@_nC2Nufib`z#(_1boSIBmY_wq&oZ0DsqJ)!Rb{n(AsFvT4SOndz2uQ3HEB
z{gOBZQQ|ufq+B;Jl92Of9?7qmbL`5&>ei|f-*<9AI?6tKA1~}q@%8KSp)D7=VraB*
zsMM%mdF(>Ehk8!fQchp1nlf(oMw_8s)w?EZg{L=-j^|r7e4Ae_Pma9*8W&b#iJ+JV
z=7CIUAcX`xjD3UYC=8+m<{WSQEc%Z!KxR9^fk~>mJU*$3!xL{4e+mysa5Ien7FqEA
zZG@9d<Q`rONeE>ky}c&{7wa2p?~>+b+dp2mg;P6IZPg4s>m`gDpX$-~o{Q56JhaYu
zc(KGLzYi65PAU!$m?ihze&jv<s-k+_H%Dr;O5k&Ho5B7CizR|x>|tEZQAC|48vZ?i
zyU+$(ERH+T&lKFyBou2o6}_a*ThHn&?RD8JHw9IT{+leuRM2Yr0$$h3Bt~q!l(XIQ
zW=heRV(HI|^*->_5D`9#FV98*R|M+39U#Vl4}&uSKZ}~--n#VtZ#3}DJ7siB;3$)B
zt|w2PJVzJ|%&=i*rsgpH6m!|DY2h)8V5;U!=1ZW=-!IMg;V*sWu4-SCM;F5ubi8$2
zv@}bqZpl;Ku|FdHb3@1G()+a-iiU2}ZEjPC!i=X5IZFh<W&f8;%mO7<&pfi~k)!-C
zvOGeNQ<oq{p_Gev*e!wLxW($!8tOft$%Tz$9~*iXJ8tv`<TIrC>ik4BMFF4}LJ>g)
z!RN$-!vdmX+<CCG_E1kgQ1d&DCgjUc<SFJ<?v>>Gyt)%)0eBe0G6x?Uo+6T52jVS9
zU&i{)F>2KD?&>YW^y8S+t0ya`j$`0{)c+E-<T*WT?e$H&AKg!S5#)x3|J^4z^$v39
z@B10tsG*_yDUW*?P#>ADK_*28`W$boxMX4of1KjeyJ};Z55ynjl+2h$;LcMr9QM+%
zdi7L8!{VTu2BEz*Pb}^=^*agfk-B?%)v_pSINYcRLTZRy4T5h;VzyuhG<n`EZ97>V
z`}H3*)Xyl|$&!|+4fwe5{2F$NE`|;fRSl^3=I>}v_~+Cd_;H5}OMGU{y_{&==-;RL
zo}LZGr(|EYq9)ly>2mg#PCk~yXFcbgbNx$N{;ppYF^m1elb>40=2FdUeWP>R)s!A2
z%}pBDAV<KV3x<&%@OM3X`;$7s`Y`i8yc{OIyd<Xyx#(^KSi@Nn&$bB>h3x^56TJt^
zwSow)wUftozVnG`>VS37!=skNjAthM7sE^7d5}<)E2PcTa!ev_zzL~&jY`(D0?8I`
zi@m@c7>!`>f?>@oMYwpHN|D)8paUd&#78mRGWeDZv`mXlRMz<k|DqkU@FZqIMD<|v
zqmY_}O7qm*`SAwGe$UjX()FAcI~csl@D<CesHCX`{+g=v@woJ2YxtP)!gl&!A{x@_
z2_jpEi_@@M52MUlCh0XgyogS8<!SI8__{dfv-OiQ-@c(0$&1*bq9>gk37^^-rjLfk
z8F8t4GBE^be&t%?F!v%6NW!3!tp+g}_jAYhl9yeY0j(-#Q_@0+YGDZhF+J&s;c9C?
zHV{p?KY8x>FN5t$`+R0wtsve~rrot-6+LLh<8iUYEWiidlU4S1<>B<=s!W7nr#LET
zsWwaotE`dJ9NIjGzG52!t`VMhW}#+U)Blf-c4or{jX$w$8oBwTllxhariSViI<%aD
zlA!c7D_1G5t0WyB9T)1hU!^O?Ik?SX41)(?jDFpvnbNhIRmbEP8VxHm$xvRY#zzYb
z>Jnoe2v5U`s%)~pt=@!`B|kIxdm{InyxNg3MW9>caTZzA$>K3O#u@b48ZjLnALZJ$
z-uns?Yyg%zNBRo_20qwCCf)WvTbDd_%P4|R+$NWO^WTVGs79r(XW8Q{NePk9m8qL*
zH}DMAD(!>1<cX}Y)lkmOK7%jwAJity4Y8SKU8Orm==WS#ZuFlV<k@xq-tSkF%QO64
zQQ}qzrGeFb@T<Gl;Hl9~%35C?<+%{(j%s#>14f+RutCHTjzcY{zWk8FAu=JDZ%t^L
ztWVH9yd%={59mDs;Ma5l!%wWKE}LEuyp>GYnPpHnI-)@?nU5i)fEVMAH%C?cL6g`%
zQy0%P1Z;I1kj`}-EpO*<2=JzkkF<;nF%f;k9!8<s*L_)8S|}i01fn|XqT$orI1F$E
zCjZ=_HR_&c@&yUY{oHwR(}5)*lA%hx|IAJNbU#Lfe+Nun`ls_LO$7$d9)$R2FOCM!
zvE<oY)OY?&k|EYFZFqsD#d}cVTdn3)Uh?quLw41x%T^lH&5SjOL!PvZ0^Rc|ZhYvE
zHyE4}?<GJilken2|NQzT7E?m<&>P`_;5enW11fAnYo@05m!%MRLo?b25C(nAxQW#1
z8{|21)OxN6OhbvA5mS+x&XXP<A?OXkt9e2O4oB-{@)t4xW1Fgn!Nj?jB-{A}Bzo{`
zh!yIebsHX&EcgC0!_IHes3=n%NcH<3=%+l!Dx0YFq+jc~<Ncr9_hFSAOqag0o!)ry
z)9|n0^#2#abHD9brrf|dA{R7fMBI1gwk@ukNASJtM>PG$^h}Q0Ri{aE0Tv;Z3|y#<
zP$8k#!B0RW_<a`Hq_<@K2t!{&p+e4n&aXF71wjC}27Ur17kPq!E~0lt84ZRVsg8#O
z2<ZzCaOqXcgqJW0A`-4%_?MzCMM*_0uc6Nb!vOOXqPI>m#P{&pB5(MegcP6k&uXJK
zKaEMPLB9vDh$>V((3d6Tp6_bnIId?PI6_K(4&8!9RbN~Zk(SF@geaRafNV(UuW+>U
z3XNiFF5WfGk~PcjIfMJQm*CGrA_PySU#7q##wOA2^@2SEJj2$^f1h5k^YXGfWl89G
zD>o9?@70v@)BD$o*RqV{N`CO{vrd$q?5Wp`;CA4aYmUYEXqW{6DI#rAM$Tv)B@Cyg
z^F}0|-apQ9R(y0^y()y0#ofc~VqmiQ)tKttn)cbcnUbb7>Wjh(*}Q(sAKT5FD{k!>
zzTjwa=f&=UQ2i78$F2Bo{<+0VF3zdKe0hb#_B_SV68&Qd4_i*oV&fTNtScB<eR84t
z>{+YkL~wF9O?!RjUiw6}id(&@T%)OMD5X|&eE#0Odr{$qq;cN5HT&8PeNK_{Jkjjv
zeryCzS}rB+7a+c>KVz>LAGg0@#0LW*k7Nl7Re~}Xn+442(`rbQj(w(>GeZ2)Eu12a
zJUWY>DQrX4gkZPMPBFo%s1!?>E%E6vTMO@+?QUW9W4cd?OnuoqJ@vZTgKr7f>QFlI
zCr?U9&8syL6_qHd)G_f485g;|+Q4)FScCO;eKSo%Nh#rF)4d&j`2zkYKKY)_Ex9Kf
zw#_J0vi4Gv5RG;q{}h`m^J~T)=T!zY=S?(^y0eWt-f42EdJgx=DcYyVJ=Rhk>)zk<
zD4t!_^M`cp0Ym;5C+=#fC5!QP(;Q1y+?J8_%QzW6uzUK|PA5tqUPoQ`?o0%i@nQyq
zIT*%%cn(@ycYOWIg<&wD-Ma4Mn>$adWjw%VKF^aY!OsyQmh%*EuBQ+M6K4R6HqWZq
z8j;wVRAWhWN{x@&#!L0h+$`M<O;2oCjwTF{o3gU9qJ2E^*6;$9bj;1sV9xNCw_g|G
z(EmspI0?*2liVT~8#pj-Y#x)MGyG&A#;U`p?AykPPtMbyUt4<e&Ah*JMX&!%?~x4L
zH+b)c>|4M|bLp!MELeH`9qBvfigm7OPw|-;=)Wu%(#5@><J{GK(XA5ponmRO>dFt+
z^SUa2yN0BTurY&oDwz`U?{`(P8Jy}F*cxfMN43tC8698|Nm0H@!P}#}v(q-TYWX~w
zBAGv}@#n>S3qyYIjnlRn!+L7^S!Vj;ys8=Qc@0!oS2Wr$UoODW0Qn0a6<TzjE!9&z
z++WOz34rU#WeU7hjaq`148}bV8z6C9<`4zq7g&1mLCKXZxI3@P<&(-nH0U-0cx5&M
zNc+Q$gl|M@0c_^Y%{yPLAXgo}0qNBXJk^p`&)>?jIj;eneYCC@Y^rOeLFL$Bpb}or
z*Qvp*681iCy-};Y_E?$5>Afq1(iHJuwVFZ=>XKyGaXcAyTP9X|;iA2XDZi`hZY;gZ
zhXWIHou^hVqf#Tz`wb74?OIuIcH;Y&07ktXJ(q|2!{WgG(4we$EN`Peg&#1qax$P&
zV&7Tjxm7Y420gNS^n0#<er4)y)-bYAUZVDEx{Tz&ZZ2yFjd=t9!QKOAC*95Dgzirc
z^1I~?q#L(K%`Q9OeQ}^<dL-o_hhDH`s6xCxt`A~3*?sz?kI&_>uz6U9m`tL5$BlzZ
z*6T-n8o6n6z1mf-IMR~4z~UG~J%ifCM3OH({h%yASS~OEHj(AWfqL39R?mL5h(+Co
zu`xRpLA)*1=G=Yw-ML5jV#u$EQ>15Azi?ROB|FaTd^#j-PxGfHcVQ(9-70ckt2OcM
z!02ar-%G=y>Uf1ZK>?iI8k@xNEref6?}_7!F?iu!x#-A=!yy)-uC8;}4G9LlP_KA#
z=F&+C=^-|+ZN{dWEA2v=tO2=!MGOXGuZY}%{7z@~&e1)l7b2<a*Ggo)9p2g3C!!JW
z%DmEeNO|_yx*D2Hy`EF+t`U`&0kEHnnWWRM-gTJNRL|wt@y`wZ&Nxh^%z9$HaiX>C
z+YgEQt#8+auEr|C>=u+)tQGH{-C3gyiqW^vFIOD2w490EQqUnc9%D0E*SC#p0^IU$
z4F(r>K5OX!WX8lr!#SHx`y3eYgkyHLIask4dqgE<LIWg2d8+PDK<0+4X}5Vy7*9N<
zPS$KU@fNU`Z*NunVyhT`t-gom9|!i#8Ozbq2uSzv*xuJQ@vJR=aBB2vaBJm=?@--V
z_f9>W8p~N1>Dc2BJ4+}=E^&{}y<BwOE-J2AQd)vHrYhe)sOyIG=7W<s4EOTIM9kSq
z-|@sLG<R|Bj%yO!P3h^y1+u#I`!{5^OWkl;t>d{Oe3OPJH@5AZN5|w#(%-$T_<UOb
z_0~*@EV0Ba5y=7B3U~ZdaVJwlL;mV(@Lw>#RymXvBAYIewQ;x20d3Ve-*QSvHx~DN
z-n(B|FPLT*Fr!lg{}W5ln@N)xHJa|$Y8$H6BJcpHWDwW<2eFs7R4>Gm2~DdG{_Y}J
z(<29ztVa!A+z@twdxau2)%aXbA@(;^oESA=EEe8@_k%Y}B!-#Z#2Z(Qfquv}$%K!_
z9(R#7%r*B*;H_J6hFPb45dTQL-be%d9VUw;Hua3WlT=r&=bSzE_(E^QK(gO5tuaRZ
zXu)82Y7tYCYZa@19C5EP)^2@jnUJfioop*EIM1ZvDf|O%`Kt?MO+3%X_o(bS^5c4=
z&eIbQw&b$nu9=m;gAz~_#cq4rKd4BwQ^%yiP&WT&%}{`AqOOTys^5VPCv{alA8yQ8
zE8$%Kxvh?p5|~)*m{buOe}xju_X5jQFA4EDLt<vJjw7oiw3<HGUHtGdIwhka?E6R(
z%uAm1&iODs)mtNThmyWkunM*MOsZ=^Vd1l`6u)rdZeF@{X+dcz$m`otv>)A|cG{iX
zp~-78m8m}?wTKlFvLw0fnTy~IQ==VjzZJG7nbOD&#5=|LX<2WRkV%aVD!AP<;%p(V
z!%zvgvAEKUS8A$j^~l)t*OEOR7}u5qHW#M5d$L71XRCy-Svoi-KDI2V_ew-)j-tF2
zYrfl++1u6`NE3@xNcu=tS;dCVx&4V%JnQRSB1ffXTdak$_DbgQruSyNebH<DqnR#|
zf)<hLM=SiV^2JcXbd8jpYq*|TaMGn8opWU#k<mP&>bb}L<;Md~fz8F+Eutt(6As?|
zGt-s1|IFUo_j8<59$hQuR`twK!)a;ZIlx<HeqwslLRHm_#m9R87%0-2$d>n=q@9UA
zz3AeS)jDHL=3?T76%=g7k_2>vgr1Ws=^@05;mKXYrxUe0*D=A=a9EuiOOCi(!oc^O
zsl=pKE#+dz0(>&0F15PKMOMeYZbK{=JuHOJ!LmLhL{{`1qh2!9!%~;Ls5$l67p0+O
z#fiLDYnm*NfQ{qq=uo-m<dspYWVfl*#CPfR=X_z|!WA;9P$ZvAF~L_?e3?^g-LvvE
zSLQL-bupW|Om@kdv6%dD@Vj}?r_0F0Rb#e?y|+5&&3(G`8U`asv~%Ced--9DYK8Bu
zxG$rs8uMuHAI|Qy;t2DWe&=0AN7U`kDt0n54DKHMBSY<GpI0{*9k=w}oLp;F&$=JC
zTZ7+O*EQUT`_A6El}0nV^6}8#@jQ9o1%<iq_6g_Cf{7X=i}m&O1;xeB$rYzHEr@E6
zAU`AG<??oHvf?*8$hRlPEyi;!VNR6`<3VNVz^=IsXDp>;vBYpDf+;gLZhJpuODap_
zqeWHg+Eps;6txAbaEcb*-8E}$U7F-GU~2e1J{^=*PnyeH>~hJI0{`apr>E6o)=xg%
zzjoghF$ZrKL5^Fuz)EE;xaxDE_e!fwNo9&}+p*E7ty3>ki*gs|d0hLRtjvpVbVmT^
zk$3Cr<TLZbBB@I!cg#DN(Nq|IQ742h-9Ou!7WG7?qvOyzJ<lM%PF3xOkbwxv4_`KG
z^nYh)RMWUOoJ4lJV>K;1$4@!OWdAeI<-!8)SK6<dptw;Z>KqFLeg>n{>F#ub^a1M{
zr|JBD9Yd?sC%I(|Kac4(-VeXoX8fSM+$WxidIy(`YD^2LUf&uVQrhZ_v?@^Z!mNeQ
zTvC$wFFk*}(pLID&$)^pF^&lYn8qJQR61}eS5+m>!~6TFdbV2*%Ej!&Zv$6nX4_kw
zsz007$Jp3pmgdT*gJXdhE-wtm!&%N#*UeYgefqAD)r!UB%HHxw+gmB#Fbe}L9J~pc
zBH@R(nXp;bcb+pSS72N-%oUy*4UG7f{n=sEvoAlmpH7iv+07o(-mJbx6vwIJrGWUH
zg1AiCrX7zZN8o*U`}X0-H(u0zU9?$mOob=FRZ-7#@Q@{yVjtjIJh9m_tm6u0?*ZOU
zM#b4Gc?swJ4)<*{I~ii^QWDOl_QBmW4hkj3s%~ol^K-PTn%}YI10T9;Xn849v7PSU
z<Xa}`lpcxh^}vdeng?c0PJPA=Z%<Fc_62f8Tr=ILr77RZO;-IAI%;G`3CzLHt>1US
z=EcP-+!<hBK)nU1OHsr^|4gsvO7+@6+$Vw@D109EdG+vePwX|&2=5TE5kS00jSgpx
zjlj;E2~uQ-fvaPK?n3WGeQ~@|SGQQnP?!n^tlG@OhJS%J&x-}gqD5W7!oKrbCn_S|
zB%deVmVyDwgz5u$O+^CDa;x=JpC=_2LqIPsJ#WB7m(I|=mhP<{vWR_)k`<Lw;r}%x
zrDBBp{X;XVyVy)6V`A8HyFcM!{>M2*o|ki1OUQ3{Y<kn6lkxED-F=6x8#SG@RZ)GP
z9MDjx!T%_)7QbwzuB^I@#k=`=ici}LqhbH!wQ0K*<Mn}L;)UUHIdv*9J6i_b2w|Vs
z*hgpU>^@C@piZoZ$OR_m=9-5_)VVVq>AUS~hd^ydDY%HupGvKT7$5XDZ%uBJy>_rZ
z$T30DMxXFYiV9U1)LHa%vXqZf@fJ3e%P+%J%sy>d&b(hWq2u}_L-0P+Lyjk8COV3f
zT@83Z>kT|3{$0%2i}A4=$7K>1ZlB${YwuVNNxO$%KPghLXDYsfMXa-9v*7Ob%`qHO
zJgu8H$5zwYSuACAJhQ{v*Wl=qe|VO)@nLN83y<pWjILD@wWTjlo^}1h*1Isk6<hLM
zcJE%pmeU$NBQ`bl8}=M=-~Q+eEtSED*{NZAD5h@=HOraAhT@wA$Ict4Nrq0i!`~lG
z2sWggy!^mi`t>)uhMfkBf`}@Dp&6R?g7&oL?o||D_7>lKgh6|N1{QY%;hc35M;f5w
zgK~t`4oYbV1n}y>0fn(De6`9I?6^*u^6h!|^1?Pb#Uw+x)-1Xp0_hA;WBh!kfP!eY
z0&jf9%pT<waik=+RXlP+^@%$A>4=Z&(4{WVlY$P8maitFHCr{N6@8hsoI#h4nR*bp
zm)|_+;Oop^5|w#e@~s~=fZ3ias}~>5T(m`R%!9|I^Wqz}`~J`Um&t2wUdrAn!TriI
zLt;Qe`@4cG)6HomK+^^2bbkzd-uTh*meJJu^8dAod*qs1Ad>s)N8_;HG2+ZCE1jm~
z@LPT2&bHonQ_F5sW}Oty@6&1(IKKrh0A$AX`lO<Q9q*H%#ld;)kB1Cu3(fJF=i!K%
z*LsUcWcO?qy&I3NqKwy($Objl#jX4Jpy5vXmi=y)d&SQ$Az=iwSR7axQ|`}QE<s4G
zFhAoO0DDj{6p{=(Uo8LCGIEL(iH5FL^~i-+CYtZNJt%fCA2}#jv+NIkE%~VLV>d>>
zN{ZUE7&||_9O6lA^cGFN&R6z{yLML=-#acR8>UC^PPdILAx^Y?x~Ab3v!uCSjLLz3
z&i0hf(dAM@%f55bm1$ej5ld0Nlecdj-drjzW~Y<1M2@+8#q#lw!x1NXl89<<Bq}*E
zW~B5TTW90V45i+R)VGwnSc>pecWX_GR)cW#!Yz`^a8jd~gMHInI<8ZAIu$&A9Dmi6
zd_CvYFQrN1Y7FlHW(W=qsxyl#`kG4fS(Cfk-~$8tF!%@~nKJy2v?ux}+$@)B^E}DD
zU7~(6qC=o#&_ho9aqB>hXs4a)dZO6IhVNax?(j=|JE#!RY=evB)M&aV*)R(W8n5b+
z)6s<OCZ56G*C*trUOH=Oo>tzo-JEY#vc{%P2@CvrXIA=zmd?H&v)o+&&XuARY6Ry_
zdezfD?IrS}QsuZfm+#|unK~NEyvtzh{%Cbm1FaYO+u=Rv%yQn4otftW24lhUm$xu;
zJo|p2eAtyK#6^|JX>Ku8wfHu|bTGOTlYTDFO@01Ww5)(r;x`WxeVCP1`{&Pdz$pgN
zh(fcwD`!SR1Wcv-aUl}|h{$viWYb>7TY_V%%M(>DsX{<EHlscx*w|d`c%)=Eg|kAk
z6R!{42{|CRJ?FkG5}&^ntsm(vQ=+P8@qO39z08l1<O}18R-)Bv+OEndh`r;OAjl$p
z<m#+d92cXTnO+(7qjmD$3$#WX%T!PAC0^9Fb&)M*$MBv+ugW%Q@Y`f(XG&@Dz2B)f
zhI2pBWvaVs+HjVw%jVDupep*TLmUkH8jeSWuQg_4^O<HV-?Q}U7Q3@jQDaQCnwICB
z-)iPAPR|st%#x9G0E$w=)yXjDdSOa+mzT`?aU!zxTJgu>2`(fg1Qet;A~h4ktD_r;
z&L8F&pW54J19}3b5F!K{0l7Oy#6`(`#8WlcM19)6U)TFG>Z@KG56p1@s-WZIVN(=R
zO{pzE0}k^XKms+L#y|te3!%#t*vnPk>(?ctTeELTb6oC=2HddxGUMDMFhsx;@h#JW
z-#K|yVxheiUlz9xSJ1#nXG!0n$9xuaaHRtvM&c}N9f`kIEb*TzgKkVT!^}j_Ph3YW
zqH`l-mX_}ix?TKd*zKua`FGg_sxrqxa3?)p!Z1^0o~o+q*+V8^jA&_(wN(2PCs})%
zr`g7EI2-(UyH2~V>E}3)U99=52Z_b$hj36Cd=)e_OqP%T>H7#6pXb$XI<0+2D0ANh
z=BX=`l)Iw7<E^FR<|+w;ySa6_wwautX<x5iT|uCDmX>XKwVLG>6+wJdj=))&b!ZfK
z5%1FVWV;Hud}Qfbv#le28|n*^0gy%w-$kI^xzq>}yfQsBtZt;adueMJo5G=C)zxE@
zrKpxsf}ut;-3{XmqFf^@Xno^f4OVX=W7Qj+awu{X=Hq_lsamIAOG(L4uSJ_US1YSG
zO4xUFt@qGUt?gc}TD_v#oF?%JsFNeRo^h4uMLW{D#iS+bs6oO~=Rc>f(9=`xFgAUq
zF#Wxw%rsI!jHg!kTv$ymtD>iZ^ppF)4z%3DWA9(Z&9tSCJ~tXILY<|@TiQAq7D=bs
zJAJBYGVd%rxL?q{S_^0AjP7|n=^VrFP5@E1!LEud6DJ16rp8D7Ms8X~^6@C20<{59
z9u9u{0H#EWcR)?eE<h{TWuDh#$&+6-T3^O0<M!PiY*ws|A@#USDWZ`gu21P09=2KD
ztptY(qq-S>#PP-}X53*v$*^EFQH_tdSBuK8zPl7(39eAoA`>G}px7x!MgP?<^ir@!
zfrlj-!tFCL167C!Sa)C}8%sVJncV9B<Ibmyq;hV}38m7Jt0^lBWsadQ`%@r9QW}-G
zs_px@O<SINr!mvgkK43R4qmyXK1bmumgAaCDTft1=?&YfWHnShdE{1}*3<TMd2lY0
z5@SiDZT=bEqO|<<e!Zf`{MZw3zqY)k5VhQ*>Ie%%tH*|aj9dMW;zYTdHe1KS&A*@|
zWKxh}$D&}qz;8GxN9!GLG(2nBppU|_hr3~PVyJeICXrIj*GcjV&(Z8$zuYD3`rb<_
z;i<Q7QDd$j9V;;>htuk{TY1>fI6%#e!yOPn-3FELK!8&y=<3h3sVoG5MQGX(;#J!P
zQZgicWV8Xm7``LGVH`hdPEYKBX$KbP2!GT;@t;CEi|h%^2^+k+1^)n-7IDP`a?nhP
z-6a=GNhxYqNq%|y=IzwM$cLl-W3hu2Bj3vrD~NY1bDfC5vLO)E$TjwXGFR;qa!;<L
zh}?B+&mPS`DYoOH+M+nrijq<eRKK5t){cY4af#~Ru82v7<?3l$uwQGjDm^hZx9eIr
z%`1`;yJ6<2>^Cs<J!BuOZ)CVj#i)a|3E-)eM3!;qXeS!7{ym0sSe9`8`Pi%+^~(OO
zOWrP}DmNxie+{#)Dmu}2vn7%mLR^!CtKhHR7Uc?hvy*4e1VPAylMQ<>USBeaT(Avm
zwjGYLRR_izuRk+>B9W=&E+knK{0P&Vx`Q8bWk*Zpvji|M1|mJ;1Lv;zyT=K66otE*
zQpe9){pZ%9K@V99MJLo%-~uR>iaHjuGgmWDt~0t}azL^J5Lrb<MU#WFK^=AiG%_{<
zug_?hr94hvA-u#|C?<V=Q>C8tX1;;X@uK)Xa_1=(CwR*0sdjJwSRgNTAWn~dn8l3c
zjZ*^ibeL9PFV9Z;GR7Wb!*H&B>YoH(2z&3MP&>9oGq($=-bgI|nDF)rCI8rZjrc*Q
zD>Eg{Zw&vGimGR+-B-oFrwm3_J@o?6Kd$KOLv>>Z>OQQ<i%gcq_uN{cr8Ukp(^xIF
zSZ5IiaG0ut3vuGa3F0d1={bgtPr}cHJW^bWXmc4ic<NhvLh}H&pRvTHU{HsAn7ch#
zsL5SO#lv+CFV9_ia481va!^j9u5g<*4gf88!7eM)+%L}WS5m_)l;P|<@o`|J>&QnQ
zN4sAy(UtT0Z54FAC=(v;H(0oONlF<U^YIeV`Q0_E{~*>XSr+Cu)EL3teEDI#M%967
z&Dq2obT%#EEz&o&!gKKau4T63Qbs_atM;~z@rp}3@L2sga!G4qG$)!PN##RgP^XBA
z1;@3jJl4JGQaSQcW`-?=_kGa0;s|Z|V@2=hjr7pO;@6Cw8A>sf`ddY|mwlYZAN=<a
z9oX9W?tr#MLs(*>=|xmZEZ_VOpotsoF*+TZ?jNb!t^y#BfTqxtynD9<@00Pi`!rGQ
zlAF*~Up6$4_Mjc%x(mWPFg!w^0K%R@-8IMM*!&m)5yBf@Z63lg@K+bCikTdnHcRUW
z4h#JidkBRFsCw~l6VF8FoghjeHKAWsS9+qp@#nyUzqY#wHi`)!B#RL90>wam(D<lR
zUkMt8`ZFHMRoYiJeID7FGpI$tFDX;Q&I_Ak9-3U+tbMk&6dhsY13f)s{%EG6$yn;4
z`nEpBXZps;EFsTbWdQLftizT+PrTdb+WGXSpGAEo%%~kj@Qaqrk&x12vPkr;abL0E
zY<6k7v_xx8Ns?W$V@cojKY*aBkGolS(^eGs|8Hp4=}tz~*Bi#yr0FG#gQF=avJV|l
zU=lsw7rIOJ@~Y}~z9ttz!t(@mB~DkgDv>pK&6rjri<;^wnMm$>fkkP#JeMi;+b{(&
zIl{`dp!{5yCvmLC*5Kzo*<G39U9PHlr*#|b8-NfJm1daEm{;paS9CPWCDu|jjJIy>
zpr{L0t$U_GwawKvT{(Xapcx8n+muuB5dv1hnJUP&?un~f61bB#n6reg#^iSuC&S_4
zmeeP;YrWd7eZMX8eK#?GdKZbx?Ei(zut_h%AHEvRY_@CL7`pj=)my3%OW^Ynkt1nL
zfjrSpsUvBw#2CO(O4?x_?yR=dMY8~rCyhQ<S@mPoM%7bB+f!?98ctN<qo3mHD0S$n
zt*1{|U4L+X2&nUr7;deWbVH(q+Cxn9Jg<&ZD6>Q)TjgV>T&}(yulQJWK&`5J`p0C%
zw@pDyWv+-{#StJ*nM>k&QFSeI^~TVJ#m2^>u>l5en!KovLHsA57nC6o2Ro6utdTb>
z$w)B_J!Xn^V1u<op9<xeW5p9z^=B}+tI7TR-0uFqD;r~}vTWUlMZ*un5fck_#+(Vi
zmXb2cGJEc#o)5(t`92vu+e}0`E{1Pk19%HPIY=sC=b*HZ<wrI3sjxq_XnfDNa_+Lr
ze&55Um-Xp+e7Sp@y^iQ9>@EuFSgKRgC-yz*NlRCMnv|Z%7qcdm-rx`Sp#lRF{_C?>
zMpI+OrK&E9+8>Wj^EM`seG@~g+wK#t%*-?VGDr3sTA4;-Vq)$F*IxyUk~iMDDmSsa
zY=EuvUC-@;uitkGN<vLgy2qDey1hB)_P&JU0X*CB+Gb*@@Dr;L`yCx~*Hm-Q54+q|
z_9QxJw;*bB@z*n<nT}X_T+h+uD9p$42`;L@1u8Xg)#e$+I;JKA<CrM}pa|MxBM0nD
zNS5vB;6VB{?_O-*_;kSwB7h$70D6F#8K2I`k=|7^3o&x*>Md$G#P8<hU?K<_T>9<A
znodkrWoy(<?@iH=?`6SEuW@vM97fqY{f*z8>$O>_DIz`3*|f?Vd)@EtPT7t|`+!Y0
zs)n+ntSCR{&;kw0nQ!0`W520%(<i$GS6d6dn1P{c)~gR94Cyq-A2zH~?%7WFuGUM+
zxH7j;WMV$<Hnhs_58XKYu&<7vx?V(rb)<zgVhdjHp%`Bo<H*DkjQw1JeU<%uru_bf
zwi1X3x_YW)JJ;0Ig!x2u2F4UH0#}R|EG!~FA@|~{`}*4q7v2s{<3zF3uxguX>quwd
zQKBdc*Vi5PK`J6J34l5nz+p$*-sh|x;$idmzLr$mm2o~X(Pehr#kw@3+8~v1tR)CR
zVKqxUJwK2-!FZw^l_pCkzb&3Ex_b!K#K<^Ma!hKX*SC2jb5yzDrC#e%BhLlqkdR)i
zA2`qxjJoL%>-SpdB;N-(60pQV+Di6Ldp$b;5j~pHIg$P_dil0xBGNr^7wL;w9{$WL
zyA+?9xB!P|AO?DplABnXp6^vQtfm)0?W;%MVp=+F<?@T=aHwebEoB*ZeDZ`KgVB(`
z%+Sct;F8f^<5&NhkEY8=<m~3imsrZ4uPpyw;*NWu6I-Wc4J|-xyuIG##Q2IoMSE7U
z@ruOEkQ@#kJV^AI7zxZ6prC`{BWKIeo|tjEx-hkKJ?_c+9<jK>7g14uRAeoNCiwU`
zCv$VVJON*0fr2qIq>l3v!zYZ88Vy5Vw||s++|sofBd&$P1eT2&Gr!SvA5te>4;^3*
zbe$pU9ULH{G9mtVW;hWf^Q4BaQ!+IDp6rniB1LW2+iR{<sIxMOTBAYBtGMWPliczn
zw)Z2c1~1Y((zR3%jxM(qmy{6eL=}3rQQRnX*|}&}B@bOw{%)Eai+pJebzRD><83F{
zo_0vlxz(>V7zKQ>VKDkWs(YVxT@LC<PwUHL58fS1K+LeAnYY=Z4(^oX#DCBoY#4T*
zhIbm8C@rSuVV*b9^PQs+lYrTN`zD`Vct|S*?*Yda3L|2b?#U)bqg4(mJ{NwqNkx_z
z)^Bx>AwH?|GQIom<}`E6;7?j`R@yGQSP8fzR~Bdn&_&Q|#26&{IKTc(M*49Jet<e)
zGn^5^;ou@DB`aYT4efTVp1$E?ohT|5^i42vu)p!l3H0~AbN-$IS<-WgSZpCLcYM2`
z42(6raS>DrXEiqmAAs>f&CcDs<h$Hd%5*DPIhmvnYgdJ$5jdgxy4l>N+$;XZahuOO
z!cR)zt=fx7M=a%!%6ze+av2UyCEe{;dg6raR!cQp<dvcq$V*<$udxkfHuGi&(*Fyl
z-MM+p4Z{_Bsk;ZpeGYvpjntp_AY8v)JaPvvZH)s(-6`+aTz2n8xg~RFXi;gjR%@D)
zMeLiSqFIlMN=qxVQhY=InuZn^Z$tS6TdUJ*;BOFD1Cu{me{*=pJ}ZhW<qCLxJZxBn
zBuI$5of3`Y9MbVFyaVHWygsbg`ZG8NL9A;}eV}`Uzvms1x$|l+-FcpE<N;8b@?1jV
zRVzCNOPKm5zj}AOi)7ahtP{1`O!p~R{IyW-U_AnZJD$`C4gCi?dcWQLTnM@6-L}oS
zyZQX4WrxL=8m${zEfJ?zwZ)X*mdP>{?drN{VJF`?Q{3j#L{*x+g?qkOC+kqK@N`Vb
zS|f3(6)d+wits(8x0AkF$6|-6u`eT4|KpC8os0)R@A*@U-YI=}dF=EHS0??8Y2(OS
z4MQ{Fmq#~b<Id`_DB}*^WyCMZ^vfPnqSFG(VzqXqF3IIWAq}-(^NGDX32+jlzioH+
zz!us@V0CEPR(G<IuRWLkWt`7NaI3kwYDs^4ISx_#1BG|jaL#57jd^uf1E9p5%G6ZJ
zfEVw1sQB@N+p9Qn0)jgF$+P7kGk2dmpU$Z?X5;V^weZ8(7le(<_BHc^%T;YuRqypV
z5s|Az#Sq!CDs=UEn-H{5T8&^~w2sVRQ$P#wp2r>^;hq@e5N_N1vi951M`vzSpat_J
z(hlq3tguu({X<05J}O<k!aKn=r08`$L-DzfpJ{ng1nu;L`J44rSBQ$)3!PikDrDEb
zd6%4?YJ={dSLf1QCO(bt;AboOvGY#L6SuyM*av6-nlv=~8a2mD76DG8C}D@^SOr0i
z%C6L$&!=NCwnY(#CkCm_HMBG#_!V-CTmiU<-3~nIF{ieLX#2WZB*5F4ODngGI(s}p
z9PRhrp4VqALu*X<P-KH#jJAi}zI6mt6ErX<5p1!F6EazFo17Xkf4X3!I~P$19?77I
zaI}utG!DO=D!xpkh2aeWkHdV-w@iLo=t5i5#x*&@zp5e&ofm(uClL{3n{WlqmXnhs
zNWAzmZte!(^pHR$%A($}w&2!h%6@}~A4s%ej>>;4ZB@7wfr;n^&!^G!HoS<Y2HiMr
zdnLX2i?Lc9jn+PX(&o*H*tz%Kh=Mzb3kl^|)euJm0y1>zVN&$GK$&oJSx@1Him1x(
zj0ayc4hLl3yzpPL*sn%&ykR9VOg*y0aL#kxuUYaEc`_1OmQt7N%^wW<`Hp0t+|cr^
zamf}w6FzpbApx)W6AE}x;|S=yun?w|gbD+D{^*8;UC-ALYo~!v$K*4A)K1wW$8|Gh
zWkP4y<fb1-SAMWQCg6UTCn>|-Qnvvn9;-M%jcY>A(4j(1U1Mt$i53)S6DErQ-r4!n
z^6V{uR~5c$JjE@tvVG*N2B=l4H=IzltFII;^qN_zJh}YnEJ&|oSR{O6GEHB-Hp6l+
zl!M4;gTdA8s`H(vJsrEg1)}~OnEafD$>GPNp#i|Zu+R@g4rYhiz0RKRRgttSe!kDc
z#ag`f-?v$^$4{SXT{+6f_$J9h_0ZH)^)wdS!6M5=7vHAuDR_)Wk3Lrti31->f!H-~
zJm&mxLGIP$#P`8Bk}^g!a|eX143ppK?KN%*PAoP}V%Juc8Sv@hDs~JR$ZpEo+0p#z
zL#G|fXmk6niIIWn6I^$q?Vk_IzB|BKg)vym=}7{6B1YB(6bguhnB}0}Ss9U0u$8zz
zR_oHA_kiT2zkZLoIh}6H)yycvVM^Zf<ZOX{o`D5TiPFi<?J30XpvJ<5=0no?s7_?x
z%4h{9j}a1>Y0uURE6GcQAI4&a-!&cpUU&e?SmjjtJN-HkuA|x`^eI+39bj{SK*>4#
z|6}Y;psCLPKVYRrk}c69`*snM$}LMs_MM_EB`VV*SGH^=BFcJgC6z=Zib|GfA?{U7
zDUu>vrtBm{d0yYI=C{o6`Jd-G=bkyv%rxVEm-pxWer@k}4=$)ica-L*eLQGkEA?gg
z<CtQX^p%;_!kaa40-(?Wqu)qe<vvddpA0NBV%%xg)$%LRq25vyWQtEET+yV{xn2x}
zU00e&fbxRHh{+gRyh~8i_xO!Ki_#qE$#UMx0gP|UQs{9RX?O-$wAp~#;a$|8)rgoE
z>7n10qv+BvjETc)i5WAS>KIW2DY<7gXP-7({gEnR#}~n;`vhG#0%Rd5hd~z5GM!$K
zpAX8`l~-sXiLuFxQ%3V;<)iL7H+x|zC&~kPFrD2e61~4AN#6p9vO>jB-HC2g1!5ET
z6(c)-OajebJPQz;kj6((UQ+9P+1Xq`vZn`x*<2tWE+MWZt|C4`m6dT1O@AN#;X7<>
zfP15AKyHj4WQyA8AqaUo8Vh@cnp7X}7eiOdgX!^w^GWe&Q&;WaHpMOUz6d5m#kIpi
zsO8G6T8GWiO}@tFL4Q3~=B}yZ%u|<)7zAnYdaiJmZl+G1PP|bvJ|SAGa4YS(UKSop
z2+FQ8N9q`Lt&l*?nBNZXIeYJ{Y^I{jkdrC@Op8E8w9iLv!^0v`5iZ;IZqoGur--;c
zmDMmB0d{<jX&sh1XzI>B>qyJg=^!2pzn~-j(akDH6_!yW#qgC>qlYBF%0|vNY5W91
z?RA>qVn7~+@;KOAAGq$hm}S6tH->!Ug;1j`Al9d$)?!iw9{P8oMc9dQYOTrvf&@>p
zZ`Nt<?d+u|^zt;HneVZdN(RB;P52WZk_hK~XQj=?STr+3w8Dal+Lj0iR9<Hn(8R}j
zWNi)g#O5pN!81?vCT+$x%#?p`)$zdug`AbQi)kpZep{)XTmO7|JwV^ubW0*&<ySW_
zCGCJ_z_)5%93Q<}>%P5tbeHa!6YJIIMc$UKji{$RXiJUF4g_54AvL5g{z-b2Mg(F6
z9oe{+fI}DN7OLq`FFYEtZA?ghz{47R;+M}zLqlcS)77=gSPtT%j5zg_zp-6Q>tmZP
z;d&$VlI_{?L8;a5H;LiBLBOPOa7lIF-j!$V7j%~D*z2&RwiL>v8^tH}e7|ZmU$R4U
zB3H<=XR)-qSiGaMEXfrrF@l1N(gjV|a{!SEeWiSvVe;!j58BuaEh%={SmhnnMXY&2
zi~v5+8=<?$5BpX!?Nosn9z8Ik_N<{1_qf&rbpld<z>EM?&M2va7|~i97E)eg>RrdB
zIm2Hu6(xLKeF@P5SX>j?v8ACM=bxHl`jZu*!<5Y?mK6d&3^|ud_B<OiSDX6DDjPjy
zNnajM3-H;SwTSGNo`i?oe((e^NM0>tsY$-|A68Vq+of_%@0uT&?fG3n!xpb2<Wpa3
z>Z>PuLTIap7B!f_L(TO;>{q{$li9v?-Rm*XlXSRlaqDFxQ^>eIL!i(p6cWBkfc6an
z>thXSAF*_MJ(g`oazb5>J5rGLt@LCtb90%x66QU)o@0LR9<0+v8vx(}-Tlh5#@jZT
zSmie;ou}s5Wv*8=y?=P+^Vjos1;m!G1y_Kp_Q=s)LARv24-q}XrE%Qa$2Cu<4_zQH
zo5-Sa(`+k%u=vPC&L<lvV!I$<7XaPB74yPeHvdDPA#Sa!#b;%O)|OVYp>)F|8Ck`%
zl?q10^I3Z6I;K$T{E9JYH3uPJ%!|9Xq8SaDIz-#g5Q|Spu8x;p*kj$cA0Q~9)#w>$
zTlJq~=(M3(+b6Gfzo!ESjvgTBO=b`t;pU<s;n(;BxkcU=rbB4EhZe;oF@aLwyBsDw
z(~?}l>Cmz4k%Q>5d|*5H9zcOGLVY6UCxT{bfGQ2LJ`@N%vuRnEP!Xugp=|drERd9N
z7Ug4o3<cSlG=8gdD|Ao0UBpp772qf(ekltHO|I;PDV)2A)=MEFR`<u<E>-2kT!JbJ
zq{u8auzN9IRA@d_ZH+271Ru%BC>~3$4o(<1sAm#-YsebT3j%Tmld#V>;LA<|l8^r~
zMieBfDX=gEqru*QdPSOq<L#?Bw4zuOCR;W*RQMCKe}Yq2*K@KBkYOoUf~S?mcPFiI
z+mp)zC4cQyU^1kHOs$*d8P7BR-WnJm8b7+vPD)CeYjxi|ExVW@F@}$f%#fC8(ACrX
zG1=>wItKoPc<426{e-l1L$Vrllm8`DBd+z0rt+)h0#v`S`8n<lEBxfu?S>UKC)p~k
zeB)lPX6UdPKsbP1lbrzXkWlf27)^<x3P2vg_hdJ={88$IXE?O~ii_bLY@OdAmKT}d
z=SjQE9bA2)*n$s~O|uu8bTm$_o7lrb=ycr-EOW)dAKj&IU{_nMjalkjpfD7@(Dhaf
zH85U^b)aHXjH4>8zMbI;@d=Mjs6I@%p-4w2k+E^C_(tZ7)5oB4Iqjvp+>k;((;LVL
zj?6p=*U>3s#1gDI?%;Fmj-uwJtcKv9KPpuX4OX#O%S6Q07y02guBq8jA!`n3AGh>j
z5|<=J(a4HFT9PV)V?RpYqT<R`H5h}rR;!z4%7{xE_%(iWVS-hTb~0_dSMOYXPlIIu
za_FwAuWt_z9B$Ptz3n4~Yi#_}T|Fgj=G(<46M?q36*pLNmNKDaGuvg@<>jD@vQUuN
z{-n}(LXMIbr&fJPB|gUo#5^%lLBBgbz7b%XYvib}n#SCeS}5B2<dCJ3DSK8OSv#LV
zXu7s6e<^T2sB_a-=*Asg`o%VL6_95VIbj8MRjL4nV`zF<&E3=z0ugu_78j$ajN%IC
z_c6b9>l!zv8D(dtmHStDFZ#Ur(fCPqO!^p0=F9<{C&WX`H>#<e(V{(=x7S<fv%fgn
zI$C)|!xJ=si-LhO92}o_3Sd~XSI;*4c#WF0EDl74eB+VBXm;?$9$L*7imaxDoT6b}
zRQ-)bGP}nQYQ9Eo2yEL%<t`6V=Wd^xEC@Z6cAkGTcg;jtPpjlntyCIqci_8i%#S2z
za&pa_%uRebZFJFv1KOsG1_MuOu%mC7U-rH@J_PF3x2wN<Ef4-)?VfpO=O~}^?m@IO
z@aGa7(;pk>06*+H#MWC84xKDD;;dS%oyjWQP~DfFtXq1=?V^!nj_y7L)K49M``{==
zZn{Fz3hQz!oAbG&(XMu>vW09*65u*=MkMz2-TXewWBkEgs2HO~`tII>y;@q-k_%Op
zj6nmiL*$@}{uDhEnkx)j1>#F^N^RqaHAqjWdc1bM_-WN<6Tm#ADKe;u00*MEN2ZVn
zcKzcqqG1I?NI)#&4bxcfvFy3zmoe$!Ug6G|pz$G)Qx{ThP&ODrP1wfYmT_y|LZ~f=
zrjzOf)|$unNU=b34&}|_Pej3u)Q!@N*8Mj5XmUiayZb_$=kI#^4{hAXU*Zk*+gsG7
zK4dk&*YP51?7y|YU2zEui!PZN%KCS03K@k+?#6)HtKlobBYmpHKj}@TLgLWl02sPt
z^4rxk2k8^N!T3q^2G?^sh#51U(?if`o+?FI*Fs`>oUl%&U{v^xBv(dwrZ%Ukl;(2}
zj%P}lX%GeSg?W|;A{-^4J=x8i=*aR)9&BeXvdxAb$r^i_fyzX8)mAP&WqiPKrYv;Q
zxbPVeBVV;7T4C&9=vA<=oFDdJvJ=*}Zc<qW{@}qt$yipF+Z>^b&H<4`ACD!oZb6NT
z6?Xx#qAG|*m@KE%Cj8T-HJ2{dI{wK~BRHtw+v?ySVFewFgQquR3fm#iX($bpXvR0H
zyIUeEq%bzt<U>i)(x`Z28=-wtGJS8a{FHINIJ)m@X8FO2u@NWA!wo+UE;bQV+!1{V
zqcG#I5$o+eCL6c?|1w1#0IxDWeeJ`qVmE)eWJBbcAEb@4UkD~}Lf;=@^po3!_Tx8)
z37=fc+5xonO3)9A#&rJ%!k6M$P#1oKPEZ!HFZ~XEo$$gEEFHr~ouWzHC$k%Kj79kS
zcxeS_D7f^iPZXgboY0Fqz_CqAeMzcaL@kty6M?Cr^4%_EjfVP8>^}Fs?N@Z)V?a}m
z^J_U#Y_Q7iNJ00mEHl?LnOD_-Wv#y8a*{@aWG%(iZ#IQ{)LXP3^D*=k1jPulfwU8)
zW_~~>(jk>@kq!V63>R$!LKZ~j&5(&dvqA4V55_xII>ZNPwn0D)-5xrv`RkMTOzzaG
zAjjze01lhJ9<AjW!}sc*8sTqe1j!muQ_}+E*2x+8QOkdR?DkIG8T|eF<kzk3y}EDH
zuVt3&@rI1{KPHfft9RT5<SKqfPFeSaOY|hZQz2|Y|5obV=Gwf~XheP6(}C~nG_0i^
zw6JAka!a3`uYRF0$<oLT4;lsRV!0Vp(U(1(o?e(%{(8%<z00X~d_N3-$%hHW+C*f)
zWeLB0x!>5LD*>ZRG=6@wniIC*^#E?wrVrc6Xj3eGZ)1FD6<-0vJ3)Cwc3fBaZtUr>
zJo(JMqH5>7b=Ssa%;679o-;!C-(_RivQ>Bbi>`s)oJ4%o8ZH?bepho>7@$UG_<wm}
zFy)u%;$m^dnYM{11h2vKy-M3KKEmoTmpD<-+{Nk#-Eo+{8Gr8!;A^*cj&xQpoiic%
zRjh@=@Wd7f3$X-{(6^E%J0Y~t04_iD&o_QYz$As?lX+`M{vON(I8KPqv2Tw`jXRTg
zq*ddD&lUB5d%3zN58O0gArk2`_5m1m?XQlHM_B3>x{69(imkulEZ*THmBe$3av+|z
zK%`<tqF?C;aXNjX%07f9xb%!BMP`q!f{daRr64iGm`c5^aYUr&*!^YRW2@H|rIphz
zt4n;aG;O1)`|D#769HE6O)Di(k{%6MnFi&3-^Jcb`4*b@6tNq~n#C2EW5?Ud{mx);
z1%W+RlaVwZehuHMZT0ru1-AT+0h^Cnv6VtSWWfh!Q1}oUVPK8;*y-uNyqNV;l6z~k
zcw>~8L-JEvVS#oee@0<!G$soI{i*Tz{C*>cj+As1<;&@o$5J*nIEM4YzeWwMtGn2G
zz<I0Gtf{Eoxhqi`@m<>eMcQeXjHQ$fN)VgT_@(yAv3)B?x5RUg_VX|psA;QU{m{#?
ze%jYuF!J@S9I?~eTK0*T0|;N}y`VYSeOqerECwlTaK<AAmLT<oYN~OHA)()MNt1~r
zRCV%8(3NRjGspTF$kEZp0+W8*(cAGqsFIBNRT&oN&mcl-?>;NqdTI)bOk_c@zYOc{
zb}`SpU>8qg+Z(VaC~~WXpP_|cu(@R$tI>j<&}y#5r<S(|RhPf2G`8f=7+vIdvCQam
zVBYD0_eP3t<FRl=Fd2RKi|wYuWwS;s4jjZ9Y>>q72`kgd6+rd*-xW{hqp}$&LUrS#
zD6M$(pb6D(rcch#km)w|=bOKvxJLpqt`WJ{o><jMt~Lk8d%mWt#CiwLZizZJE$`NQ
zPcr-E7JPEK1dnZG@vW<v8>7M&$V<40evPgn5F)q$8l!Y`@9N7Fc3CAXaKVA1@{+pV
zPG0PlMMOQLG|3hHXr^4$gV1+*P)V6zm$`VjPQ@hk@e#YGN6lJyQbaD#Y!<E+l6G!4
zGyINzhx7Ov(msLt^ZKj5{`@Ez8hiP=+iJlYEvO}-|0fM;SY75j@zG~ww8w~8R|hv5
z#FHu}_>FenlB{-JFZTozTc8scUuFfwtY}_~r;&}Z$$}3AKfZJa30a{AgJgmK!t2!J
zk(UisU&j84ScQeOQi};mS}o7hq95w<8HLBC-C3O>^e&f?j&k*#_M(-s8Q4WId31Yz
zZZq)Hycx0I<0_j07T8O}FbNg7-+Ha7CleB|^Vuk_U9Ms>;Oh={wa2)QVZ@GEpe?f=
zTH>kroS}n+vZ^>Wkm}d7skd!}OA^TI7wnjrnCO7Y`|+#u16y?m<I;XOSn%OX0x5#+
zI2mRPd`z>&69mKPRXMIWjs%x7GM!ltGpY<iw_h@r<7Ye-TVb2Yh?KzpW~YdYqbS=Y
zP;*gX{c_tKU!~h7$0jBa%24tgg32-Hxq+x%EinW?5NM`IVEFCPu>xB2zUZd(uL@l)
zL$){0m-4h}WyAaWN8;78LaQ`VjWaV}v;Zb>M>s9IWxx!OUSVY<DqwlQ6z@>^80YWh
zf`8lxa{!kiT+&E*hjOh*JmQHa!s?^<Q$L*{ogcLxB8z@_mpP!JtZn#^-mBH9g752m
zSEgWb?Xk&($}z4Hy~-aG;e*e2+=RI^i?A=$EEu9`*2YV?N`hrbzMA$;5j@HM4+4qR
z2JxGXzOxkAPG||XOH$f}L`Fu>h-3G+ysn>xb-L5b(L~LUkH&vpOkAz{Uk*1n*el1c
zxgMlc<n;T8Cnp^AdcA@^AqtN6V29{vP<MBB^S%7X{)a{!65XdL<xXl_q14)2$lR5h
zQNAJGpce2r1~@!r_jeyX=7vTM@(V%3{gM87vs|3m?)q9>U152H`&<I$4wq=wF@Ytu
z+6w%vL7{B&?{QvOGOcZFf=r}1myYPqydc$?)3WbZ7(5qShUqOsSlfk{YpZUV_RUS`
zIi5#m?A>y?-k=VAy2itNx#fZhPrr3~JMdZIO>135k8alpeXJoAvmWEz&;)e*5bdWL
z`z|wJ^OFOY0izC-?AW|Qtd6ToR{VW*!7GO9D9#o_6?OC2v|!|s+r>HAEbY5EzC)<s
zVxUF0cNzt4f`5gS7h`m<b}iicKV2Qbj*W}R+<<sLU%eR%UIV6~w)Jym3u@|3`Cn`2
zuiPgk8-tn9Y~FS>c>4P;jt-)&*k3VL7kugbn&!Lq=HBkfnK#Q*GUYp&Lra*KP~ehx
zKO#&fN55vpr5jWI+!ExHcW+Sgdg*n0*I{ACThM831#=pA9e&9EKNcJac|Mj1(5>hO
zG;~?H9l<`|Z@k>9D<qLRy0T_HS|LJlaqI1I50BG0$iS|H7lz{M<qZ`JRWSex?o}ny
zshyS4e<7F0{*OD{QDMeu#tTh@6q_hjdrTTX^Tvj$N-Q{3SLo8J{U*&c+ftd|(p`B8
zm9n}lf2;J$sBg|NCmxp%1rr~Y*`3)fzhM6>;Srl}^B#ZjIu(1bx|HKtY;e0yZ{RB~
zt<j-9Gi&{}a!<cGr1BKt1$1cOIy2@Zi=M1Kz~kQWG%xA2X9s!OiRO=}g+G66bbNGD
zvT~>F3epCB+{}E@fo6FvZY&C69%ftrNuPs-#|W*8zbr8-cvOj8@4jkM;{k1jJ#dl5
zKL-D&6J%$>3LPmFKm7w8b^$i@%-~dxNE1Wm(`r7a@DwvEp^r%`&M{XXILBmIyB3Pe
zP=<uu#%d;({R#(1)uJfj>TyaELVzRr-?J{Kmpk}}oBbC)(l~jlu!zrz$)H%bD72_9
zrHzgfWM1~4F_L*Si;ERk14K{@;VK-O98(Np9V!jGKfPT@uU5Zn#Fl9VT|I_)XzU?@
zAmHe^YA=<*c=Q9MzP5rV#8kP(N$u0}ka}lcnzDy&M6tFK|B|N}<ry*)^)XQf&%_qC
zczrA(q;65&`{J79-|I`M>S1z94fB=X-l_Y<HlkMBg`dI9sO)s-Z`GT*kl$zcib`1{
zE$l0m5L^A-D6%D^vTqx8Y5<ZbDvLs-7#bRsD)ikb$gVvYOqDpeK01K9Tjp^zHtjZy
ztSE}*C&$@`9!&_f`46d=PY;&)_FndJp&`5t4x=x|3m^CYsP=N;V-G7?aB{uTGeE+*
zxd7Cl!w7IEJk`qQC(|SbH{ysP#Dvfj$3XOk*7x0mz`q7H{DUStiicl1&8!??j5ugg
z2>YP(`(3!<lD?Y**7A*1BCZ382~phig!s|FzWaz-_?=I^8K5Yi)DNS&zeu{7r6ZjV
z@<sVrz<MpaS_bpzrMl*CC!8LLU;n~zM~nDx<8<V<(u!_>85F+mvA6_br$$&z*0b1H
z2t8W8rRiI@3N)*^Gp|U2>g?#r%VKh($0E<}i$J4=OEjnNht>g8EkF=JA*AR6taj}N
z@zk^BoN%O$?9^8`pxDjqQTw)>!O(7d>$Mg=Tqf}<1;}sTYhwc|mn&@I%KSZ;ZwBB?
z?b+X`0(CXGMfe!JbT#8CQO=je)GhSuar4FQ$0jG>emcbMmiuwh*w!>_gLH9oXWyOL
z!U2OXDQdcwmi%WHhi0bPgVK2P#2pCCi-<878jBV80u;yOkkRR(%9x2Zf35<jN|-^B
zbHS1>^iDkAKSbm)wC`uORUJGL_t+PfKYTlI=p_@*M5%P);^DWv^%flOZ@w4t<N(Z6
zqsw=Sly&^K3VsdYKvvQ}i-~|`=wiXMN4IZu?!56N(84b@?}&88*@sJ`9KNk*2w?=?
z-{}|Sq|apVx%%ug$f76;3I0+3=2LcB`ELReExjweu!m?jEKxEKzje6#yy8LN`;bu$
zxbVKp2V5lr8Z0%H6)NP+h00f-oJNBX5@R#<cwJ{h<+lrOx9aZPp)bL3dwA)k6u4?K
zkO4MEOv3<qR0U~YZSm!D#9Z1lZ^~C_c-Qk}L=R0L7t~{~5OcGYkk_*aeRXI7)bMsm
z^fO`Qa@#vIGtvL?o!)4YzyS85(|E6B`#w>HrBl0R9QBzv{%D?-^d=P(DH+LgUACq@
z3O@+WXZsINOLPA5(3|q_S7=(<*nB^_mXTsm>ta>4UY+#v_1yoAGKqm?<yp38vCT=q
z!@?g{%(-_#A``FI@UQ8iuAaJNvPf10AW@Yah_m|oqNmF7MG6n|xB9%~)8#=0W<<fl
z<sqoD`a(%gci20EMT$K_VDsuzFmVFy^D`A-rB|s(ducW%y#0sBe$BPL{O9ROYqsHp
z5A^kROw=p3e*Gk^GsyNx#8zof-|3CzKa2g!{wl7s-z(p;j!yT9d6!x}Pt1FgJlx7<
zD2I#PUO1Pit9mEDn6~`~fJB%pcPu|fsvTS_f6GQ-rICAUP{w8U272GXPs)G2daywj
z-M+Ld)V94&KNg(`JM+<-AeBtBl|Sasx$>?A|DqBD8xc@KnnHl$N7n*)iI<I{Jk{pq
z|8j~!d0EZ(l_BvYKtr)HQkQuLt}c0_2k@2N@RfHF5gk;>S_>!A!u@gQ<(W&V`f7YQ
z@|5qYQdpL*va7Aqj>c__d`>F)C?1go+9JHcMcnDt-wl|JQ4+;5?;bE*EfxuQu^Ftn
z+i)&$w*-YNT!-oz&Yehk!K0nN1_B^b)!&KzEX3JH+`{^;L!S_s04X(Y?Sv&y@=F@a
z-ocC8I9@25Sk>o><nP)3BV%Q)%S9xAH+?v@zoM6*d0KQnlyMiWIhh_V!Uh>Svhzyp
zlUVOnUp*YJ@t@H~@7NZe%4M-Azx0@bKgIQ|Xf!wP-i^>R6id4Jiy7wn&kiqiwd@|I
zh<Vv9dY)hET(}muufYn5UvA8aczT!r;^+UDLakK_c57*zZc?`xc;U)l%GWX|-@49?
zYMVQjvm0uvPi=!L=t6;F%WNH&#?v7_M^cd*ONhmwLOl-4HNmS8ca3mzg8Cou2J0b1
z$m{&*x=X%OLe=Cl9phGWut3zzUM?1BD>1+I!39FNKR+z9t5g9<9}1$FyI_sw@LPyn
z63{4mMY`<AB;C3RE&pf?gygAJY5iBYQ^F?2d6UCbBz`D_ioN0**}FdVjr@WPm(_)_
zG1cBD^__Qr-!1R&%kZSONIGlsKm7?bgakw|lWOlY<<u4ljomNdr?40S0|dDxpy>GN
z?X6wP8>hzmnk`Kg_}K5rA71xFhPpKLVLFsa!({J>SP*iJAv(wc;rJrM;h#ywv?kRV
zppKsP$R^YQLHbd<q;hw$!Nk~)!~allZ6YO`g-(Ym@}KSAV^sa(#`*OL^lTO>h&9Co
zy;HBM39O9r-*?v~83Q(#b2fi`;kl_ZVyT((T#-<esw#C#b8-AP&v3k7zIt=wo55$W
z6>N!l^FB3A%}IH@a(pjswDIun%?_%!rkjrNc?S4L?XC9Sa4|jYY+AXg6C}j~voFh>
z9PX;LED)9PDRcNQ$kx*QSaB;fm3SG$2NeitbgxD9ik#f2h&N7kg%GHOh(I|E4nbj=
zEh0mv+4nV*ye`CXR>Y2$zEbF7h6A0X)<8wmsM5o#E^0DAcJ$6)#ec=<Q+P2#u=56T
z&%di5wW8Bg*}+aiHtL7=Nk{s5-4=-`O07YQzS`9fpT|XAk|iXkaSD^jIn0AG_yHf%
z)lYhwYM)2+o$W3uv6W|bcTcnj3lCLHD>zmLT|fc{MOv|p-}eu7MnPxk{8NHgCAVjZ
zupzB6Ab7>lvyaC%_f)3i-if-2);AA$?`bpRo**8v-jkm?QU&z4{6i#m!JKc~KEgTI
z&(~(>6n>$2r?S${nitjgSk52sbx}}|8kBJDa2f4+znrL+A!Rp5T;fNaFBaN&ZuQjk
z*jr_`<Ghur<L_LOY$Hq1m>uK(0TAB)rZAQ;^+csL%|+^_lQOokGu^M?J|4eZF<E*}
zbGDPzEJI&Fs94B26@7yebp7N3ceQl@c}r4?U@`pSA*y?SH_J_VYL|W#-Z5sGwjh6q
z3D6((07KKtfvObYr5=!58N}Jr$8dj^M!LO8XuKZf(hOl_LQKM!25=0LLQxOGj?Dbo
zUM|1T?jEJP(<{r>LF%MpeoWzdh8V+T_{)h%<cW10_cgPI;8u`DkBz!J6uF(tpMCj4
z!TdZ7z^WYGhcCLv;UD9Ni42qnM_GJ&zsW>C;J<)f9et&!(wt{k$JBrnIL2)=8kj=d
zixzEtW@k~Oui9C+i+F-)ltjasnS0XzZWT%W{YwXP-YZ>bzZ5=woU~~<X|%+E`t-%^
zv`$lNe$%XB<;5vx5JQe7C`cJ_88XZC*q8<eel1%2)WxKizM9f{%}*~rtOyqqDnWym
zX?kGU&&B62pDSEt^6<ZdpvoO+$|9Yk#R7Cd8~~<jXoa5|Dkqnvwf<VPk+CJh`UVoG
zK-OX)BSf5}$y{~qW+bvy%bU7Zu#>K($qFuPypxIKbB*V{qc6iBPs_0m>GiX`=OL)a
ztkqUFjjYlR;XhYP=DI!c5F}krnf>xV?#L{UDtLTZNkZaF=BaP*zS|k=m}ZF-lnn^&
z)*EeClTOHs=lfQ@UifSW@Y2yA@t}?3y3a4kax;wktG}OP&dGyF3srUXYIs2d+OL$Q
z8~Wdg%W^|0gPRS(1mHjL1H_vMRhsBjjpjLft$YwOm$v>U?%o6IYxO^fGsyT5bSD6Y
zk2_4fnqKbW@A#fN^XZv;mw)9HJg$X2x%0OhgZYoOzSdL9nR9gALZ8CXjO3CdJWH5S
zA<9k_>VTUjZl=Vn+<e^0nm>f5Z5f^Ls&jY7Wok-#xh=4!6Asi9)0UT#gKyHv*blT0
z|1Ss8{+FXz`D5`tJ-vJbF>V#ZV)E%$n80Lqw$r;io9q&Sc|u-;SYCu#;4C7{&2v#w
z5H??b6=ARp-z@X|CAz!7D?(`k{0HzU>N#9F(3#}Q;<@ZnJ?^l}uuANUD!F(8$vtS*
z8{<7o(g1H-#^|pz{%+Fj)$%~!mLI(BFG&L_k)<Wdskv!0{>^dc_%UpizUGZj$uKT_
z(dn@f#%Hnobv0e<UH!FdEyVJ5szl^G1|XCsHb9VJ`SR6b&0a&xHodR*v$x<wHVMQQ
zvWeY!JY?hwbPI^r$B1UNx_U29g)q)h(H1BO31SdpMsRDRj^`da@h+IhaztI6)i340
zW2y=_F^{Ixb{DM&DA>v|^teFh>mH-Y6sr%6nkdK58JTI{)`vwgl0CR^*oX!X*9FTR
znbtM1>z^^8PKj#yeVM04^}E{!-PNTgyS$X9@N<x23GI72@Qgs1d}PB4iqgXtb56$*
z*({~=-GOr0e3fD0>p4fGKy<BA!izMV)^C!Fh7)I|AZFi!*$?v*2C38xBAx~41%3|Q
zCnIcciNL7X;^fSg^6_B;jS)cCo$nrP1-S>vXD&d4fND)B4WO5;Vxh@6+KMAtd86IE
zgu0&0OnITl=L(_%5#<M=7S%8Dgy@LRlGA!3mglS5)6ecTzME3Mj7<wMepsr}aE`$U
z1}mEtvRpfw7g2xrO-WBXk(~or32?hdP*60jL>@N0%(YUBZK=^`0O>7sU4LDdPZtn6
z7(-0-h-!q0qibZcZQDBBl~BPXh`F)j8&Ici&zu!Rmj?>beJdvx9Z&c+Nu)_gx_<ID
zw>P!`S-UCBXQX*w_MW~RiJpbhV?puCPIUw-TP(kvzKrE_NLoNyRp(V(V~g}Cels~Y
zpXu*p(%&)nkYi%u#~j5jZ;4ugqEujikZVaXJ)zjXLFjb?3HnJ7XdJPZe*V$7XM7D+
zgMj*0_nt4{2jl@MUkt$5xp9_otUEep08Vf2GoacT^T4X!J+PzM3$R?+8c{Jw;b5*J
zGZJ0xa3gRR-u_`@xCKrO4?!IB+3EtL1^VLmj_zT;$nO(_s}PcHz8(n!QOj840RHv@
z?ZPzf$fl!bVvAGZe@Qj|ob+U`h(O#`c?nlXCtmGn9?|)|WIS8#f*zvx!q9_t^27lG
z!Lf1Uhdq^5<JQ~@%x!EqImQBoNk#_7Wc>C7fA3hJW5Ea7qdt8*F}AXKXo7e~3xltP
zUoL@uT}JzwdF+{T#xY#^(LWvDrL@QwY~zz7(g<BlCi(W-e1C8wKIpT$-tsY+V62<y
z+kN}LGz4YZ2lY;?i4H=gROULu@@LnL#aiXQ+vImov%EGlYs-K1Eq*8bF=<(8<&OTB
z5`%9#uND^cmSRG|*p^B3%eQu2ODq55bQf=XS1Vu7y#<7}+o+Wq3oLIoc>+GJJKyTh
z-i~vXPbh_b3#k9s@lyNz{Cq@-qjF^qSF1?k9RHOUe<0WGB0XkoN{w6ugv;;?Hbm;r
zlA&qK4?G67xaMs0QNExH2;1?hLSK`=F%hg_vlsMo{k?@1Q4%hvC-=No4Yn|qa`B31
ze1_@^wSG=6hbMPVItS6*{wp{*mXTByo$iE@v$?>j3ym<24c$4H821AI1Uj@NKY5Gx
zz<j5~(JKM}{l3-*f=8s<62R1IP7sA@mGc}2{q_&MsZ=W>Ta*kvNbem<ujVA;RB#*E
z7qVMRW)Z=8a!)sW-O}%1?zmd4%w;`&y5sKGqg&Nh&0YCpy)nJltK2J}Gi*IPVqYv_
zy45jS@B2-XP|e%sE9()Z@*hV<?{V5_0OQR>5puYo5v@Q>-la>+PTIy_o9TzIJ{jSe
zKW|R*&OV#2EYpqZ7OlEBP9o6+z4GJ7YT$Cpx;c0iJ|07uJv?1@SUXlbhTznah|?5j
zp80o(^npE0RZ!Q^(k^y|RtD2IIvQBvht<qXy%b&tZXb;qnkI;KCgX{U*eW}lw->V4
ztvGsH%xSsyYjMAFT?<6f;D9x@M34h8cH@4DP{EDP2e8#B21h|r(FTHmP;{!GH)}W#
zH)&)$32ZQkI4ncPpP?@WN^}vFc&Aly>y1z3Lmz?n3Yk~1`jROnm>5ZwbSNpNo-w|^
z-~Q#}St*~$S+i-p0WB$8Em9BgL*TeTb6LSNtp`o}Ge$pTyeLaOnUT(Z>h&{DK-uEI
zJ|)k3uUGfeGLK&OB3ruM$Q)1l@1QBO<PYH@WqHoK#a9uzX1}dB1UoKF8#VOempK2k
zmSI-l=AcEz4EfdTvs0r#h1ozPYNULoKC|99GbK=YcVATZdSq|$rsx}$mXths`}S?x
z%@2oE&^@udsNyyC0DbAcSX-OFQ2BT=?;&#WBS>bFVMef7`tlf~8}am{E+SE5)CWR>
zLV0Mp$&neTNZ$i_qhEW{e5$C2T8vrxo_eknoz}H?HzgTdifs6A8P4M>Q=G?fg5NT8
zWnqEFz7jM@#5xXC{$xcbxb%sFM#PYm?9%eHyHdR|c4!(eMB<sp);2_F5peG8#yH9Z
z>i^!K1lC!r2B<eE{i3lHVm@$a;WHH=grd1`rBLamhV+<7#hBfeiw<NIX0Sx0X5;yf
z8fHAn6$~(1Vqupl-XJL}x73YkJ9wD3yK=3zfuD~&lTmCGptwVDVxjB0+nFErg6?_j
zcaz$e@8~{#b2pP=bjiNt&BfC>hw~3V+C8nGuVH_3^lw>YVQhMUFO@oVS<I47f-Jp{
ze|N6sz5JVwq^j+(<XaU}Gne9yJRu@4__Ed;G=8Hq@Jpb?lMZnx7_IwD6(A|k%_Xfl
z{8r}yPaqn?GG_q_4eFC`d`QvcSJiqf9nMo^fuWoQuF7S0soaJB;LZ4SF+UPiObii3
zNn7%a8(o-~plG)YcSga(MdvcukdAha0f_SX)}it<I0deP;x3ejlebZY>%+8P72gPI
z7CvZHp_;^*i>qAW&K3Ubx96=Gl64FtV8fyALO+q31$|!+2Pfl6yM%*=Y{k>rzS)UN
zq0%*b#6Jh!Ao#ZplbQQ$l<(f)`(S3(#PSytOTu3_^5thu=3f5QulK&kr=r14#(RNe
z&)VU|l{3W#T2tTq&d^5Ft_n7ACcTXLWNo=aIRDYm--YNv&dWMYYw88A1rUtfNK4aB
z9EiVw_Eph+!%`ZJ{pY)fGo$+rvSbP#m95oBu*b+iPj~j$Uj)?b(@<Hi+t&?k#a-iq
zJ;oOKfX<<!E^fs{<d(ZOVGswhD$kN2#>7Z><wlb;c&$mj6Os?^?H3j8E9G=(h87*C
zB}frcBBck?8QFh1*74^~N1d@p>I@Q3BS*YK_eU~91-Z1%`RjoF(ngP~(rKeE${qKr
zLSwIQ%jCS3%TA^NK+{f6Ze#UZNy2)tY^d|HPumkL^8qU3l|b|4Co=?U`%19678e~S
zdcjM)5l|+ceRi$fvHKAV4ubAsc!$9*z&x<~_F0rS_+t?!(mp{1h^$<sVzTuwW6p-S
z_I2fICj>0D=CSCW<|RUI%RjpGTK!lbn|j;J%yQ?T{R<ah`$xWCQ*mvhxAD;FR-doT
zjJD^dv@=XZf?m`LK*zJKrQb?xM`J=e7|}+wM9BG7q@*l7eIj8?ctS$2kp|>CP_*BI
z*x+)Flhwi^uBpD;`zIj7rj2e2^vi8qd~9Cu$PM=+@3#0wB7OI)C^nQ~(*t5Da3&KJ
z8@>Tr4z#ThSY*BD0pza}@kBeHhxcx*1yrPeCn8~3RgTNHl{YkKf#mu$k}RRe&R#jL
z=AIL=yas8N7mak<Bvw$3;<-d!sMyUWN<lFq(F_sYo>zyeR&8YyP1VycbH0;AoGIA&
zYa>O8>H&_Z#a&x!$LEix7jEvOYjE};tq|up4s`fCl{jEZc&7B{kxe~As&`O{*r;mc
z*+$60AelhyDj63J<p`z;>hK38jy4^GxZIu&;awV!=K`xA?7(gSA|%8Q_=d}>^i|EG
z344qpnM|bGSY^;?scCpMmS<6hz`IZiMP#HK6K=Okre@}`{@E3J^Sp^A{R&X0TV)3Z
zpoVWGvH<>1C||9=mW1-7cLxOj*m&7t)w<wYnnaiSX|R4(5ijmVdwBlw(Aj*VY(R|!
zOEgxHy#2#>?DH}L35J3bdY9CzRpBwHR^`a1BpXsRt=TL#d<*{$Q<+k_H77QMs}PMn
zG1D>ODpWC9c~(>a$|=?QGhB~Dxpm;AK#z?8IRlGS<N-<av1!4eP(la-WLHpMo+Ons
z2y*cBs=k^h&qEkvVIT-7;BSs}X1Pel<YF9n<Y<;9I~GWa$OUtAp%B2N1BDS9RSb~S
zPxdb+N$(9*waWk2MiDvJ4|bFUM#36{`D{T$<gNHG>u!*7nKx%r^3rC4IJRm2Y>COJ
z8w$80{gi$!Gj0BCck=Ac{!gn}#6_hcn$Ns>ghh!L^23Dc!qqtu0d<GnrB}J{#u4%U
zUpa^WCltTutsUC0fa{ucLNyOqk!83>c!qVxw4?g$VX<3t<_WwE$QXcq1}Q(DgqRI!
zDH;KS7j~rj#~Ok@@Fa5%X<r*Gb0iYt*42do`h$%#-vfDk;=5#5Kw$yl3btDjmc+D#
zW1)Y&2zLtQX0hBG)n&^W5BZFxDgW%}jnbmsE4iCBowvcWln>AXF{WtXi;P?Xw<V=i
zqbH;Og-A)thI5q`5{i)%ht}U$BY+CdsLAt^@pKG~tTCvBkW48gw;uW|JF22MAQ&{<
zQ9SafedgPr4h9Q#Q(v+fVe}gP(6P$Hrtr}X`gFTOBmjz0<$?_hl;$Zd<o+lnJGMPV
zj7~59wA0j-jX82OL|L9?6N~iSch2XAwqwvl!h>*c>J8~^$SW#-sTdt~{B?~<PnZx7
zC#+sB8{zbTFlHf@8WwcKK4^TufnA17c7^&bmoco=hAjDOTjJ!&DTkTOLqT3q_v381
zx%kV-MHGOL2*~2rty@$BKO%$R7QS2)_UlXyQ=+@5ezwLfYZsh0{<#%hIT~xg6hyRW
zZMYrMF|H&UN<zgl=K;yyy6bIA-eNI|*Ih1Rq?U<vd3}I9h&`diDk%Rf5$dwmW1VE^
z6;x#yI)IZBG%rLI=&>;8CAqF!5mM+KD3RZV*t>lzkqkp3C37&A+iG}Kpa;ZX5vB!$
zIGL^&%etIg&HyuhObya_VLn5WCwLKjtYtlts@9JFw_54JjwkZbyd%fbHv*Zs#%RhO
z$xHZnd;C;d8pp-A>lJ0BZ@6r|%i$g@CF#m?^=6s^?ntk9Z*+D2Wjeigdynt;KxG>!
znYIjW?+L$_mPVZ`{*9}{9v1PnUpISb(&U%fR8NYPW^d1B$9Da6EWUrCsqk6g;YZEZ
z4(4>OL=p-&acYV~%6qd3*_?K~uiZkJMc~&&mC3Y+%)--i!{#u#Frcech3)acRxQ=$
zfiiGaE*zRBNV};~$=tWO^!T-;<_p-Ks~w9-<2W8hoXKhpdt~29gu$uD!=?q>0%-Tt
zEQD%EC7iRL{FKHW?-d_ULy@_^yE;lhahOnuUAfW)E(s;RVo11}{<ZE$w%cl;?WCC(
zH`}fInbP)^XNkgC+=^&8GG0HHRUA5o%sjW%&`GntM2Sk%lItr{6O^0HZl0|!ilv{n
z8-Cq?ZtX{T%r&B|S6&qDcu?HL`e=`_Ilsa|Wi7wd%p<HTJ4`LpoxSIuH-l`JqT6!<
z%<WosZfT#d)qJi3r`9x7NHx+ELJH4j^ZsYSFEzB^eOtpJjz-%-CN_qqarEn&x1BTg
ze>2_mqcERW&Ya$|=I01vVirzFN-zD@8C4nMx%J)g+b%W`L8M&`K;X4x4}g<}I0!?L
zIh#`qO>YE7UQ^9u#BPCb5)20Vn4W151M*z%>@i#vk}ZVNHj<g->gxCqhu6|L1QJ>B
zH>VQqc+a>LN;=#n&y?mowOHol_<O<yLW*mQ1|8?=*?q|T*fVE!hOZP78&o}f;2LX>
zK_xG9Al;1mi@1x{N(@an+1uBq-i7pbd7_`YwEC|pGQezBJjiag8*|ob1RBKjB4KZS
zgZHiYfpz2Chxi-m0=lka-Uf>rIOm1>ZnIp(63P4mxJRzDJNoj6Yd!UZ<O}KXrR&6@
zMKPDxjPBH{bSSROx$pHaBl)kNhYw~P`Ma8vHu7ja%vL2|?`E=?Cw514y=_iwI()L=
zGBR)Heez*f95yqnywh=!ksHC62}>u=P9mfb>TtqP0(%b9!_W#^m4JR^@JD8-$&4-0
zDlwPQA5kTt1t6oTk<vnz=iC_8^qz_#SoI-3CBd9bUoj%J$k|z0)I~C$!W7f;Bey<G
zxqEyxU|hg+jU=-TR`@UA94#R-__3`-%ekRrw*e6^C*HU+EJc2B1-f_iSAeA;&0Jhe
z*f6@4@b|zipfiGH1OE`y2Camn#sk)UP=FlIRDaB~vFn<iZJFf{ww{HyZinm(mz9$D
zKW)Sdx}as}v7#N`#Z4F0jeexPABYENi~4S0C;#f{(}a#B;GD6+qjgu*BxLrm^tXO<
z5jf>jCecXOjW1N?j+dG}e=vzG3O{}{%HQ{cWc47C(4#~!qF#l-qzf+s;UP_^Y1*2b
z6e?Tps;qHU^Y?c;cxll1cX$8xtA=Ce>~N=MyYqmH&0IDYE;K|xI7A^%Aeu@DA&(Wy
zHTF-Qg7K~}EH{^2Za`Ff$F()<0YihTsV*R^CnT5Kvm{|W1MTJ~MDBmgpr=&7X*0;#
zA<-{yi1dvOUg6xM5^#aSf4cayGfz$$p+$9o7FBvZk?RGyX&vQ_nW?GE$-*vGC62BZ
zd+~F9M(bG=sEwARK64FBMYiWADck^;*$97K4?&=5Xkg}&2<DD(7TT%O{Jw_I7bTa@
zIDXmq>)TSNXIyg;Ubii7B9uIOWqM1KXxW~_I@GDV60S?fpp2&Luyj?4xwtO#$xv>Y
zZ^&FTI{WL<?3Q2Bl`~6}Dy`Oxe2kX&J|HT+_1qO~|JjJ3Sc>#P)87^<g?7LU{@<1q
zGqxaue(MUmX1j(Aj>0EfHudTJ4oB=e;v-L4+GV!W4bK0K2-v+;aYRm<^)aDs+>ApC
zClomLw3{g8u!TWZcq2ecNP~FZ7eUt1?1kRBt`HeJc<a%}bg5b<QOL-y-?Rz9NK_fO
zi=H{YVs<SaX?j+({KEDlAGWdiMOO169-E-SlsJagr9C=BKJzeEjqWzGce<ROo?)+m
z7sl`CcYAIzI3-5gLCo;Ld4X;SQ6U9b1%8q5y6RuRf)PE_>DdhR9CW7FU&XpHf1?`f
zhNi46O_TrOde|5_UuJz5JVZ6D4Tt(-2X5crH^h$sp!mWIe#ILol$c{sAm*k08jqpZ
zX-;aGh3^>-7>r-bzwi`)`@oy1SDy?l>{d=`Tc)pF*<cK-|HrlkAGdRUEZ@EbiBv*{
zc#7+Vc#k1?y(C>dQ+fz4!5q`~f$tvHf3JzBO;_rPk2ViB?{=^)>KH%{*h%T=dBmEY
zXO6Kow&g7=b$Q6GJbP~Qqie(mHvp3%eS2soW^W11I8_sYKrOLB0*Ip>KOE@BW6PHW
zqy4}oziBP2Op&Ifupr{%u=HQUTKuBjl&UlCao>f=urL6oWDZr*Qr%73`b~F_u7M{I
zRUCiXccgP6M1^61(a^k8-x<B|9<fJY1Dbv%4vb<4+x}EBaWsoOY&asMnQE*$NlZ#D
z8xX%h?)Jffk7pex1sTgtB-)a37c9dS_4$+k9dofFGT=3%S0^;K|JjE3xi0=jM_bCO
zRWQ#fdTxP~SLa2y%g9gTojb1{ZI_&HZfVLYS<!ke;rXN3goG&W#FwqENtuETdPk4>
z_$52O{wj)VNKa@!>|Qj=Wn-*k-uAy)A<;$w?Z0g2SMljH6K-dYUu-+vtKZ}D1pbCK
z-_NRr#%fjRye3CiTyCO`gr<wsS}QE4SKrGc(q>|xi3#zao7gwvc+9TJKJrS+?hh~u
z)}vtroQwG!!#og3kQ`wLSS&M;`tR8%jxauM3bV5vDU0y*Alwon4rM*spZWQH=n|`)
zJrs1OaE4;O!mtVn_Gp8y<xW0b&I21vmP7egb!h#Ka$DRmw25;g^^hb%@GeGh-8Hx!
z4ze=(QV^Pj51@e>56+Ko4>F9lh6@Kw9t{}UtcehY2fx#9VmY2kukMTecstR#0S9S9
z$SB*`5~dhT0{(1|LqfP2!*+Qhv$1-BFBca3r@vd3u=QTpbo}T~31azAMQdu}zuc~?
zVA=f7n$Ai1UX24aTd`%O{AE9gu6rd7`wgzvvh~PkOS>yd7wq&Y;&pmyW!bh+TLtmg
z7)gJpl@qHiAWI?+R4{Q+7n?FL4s{=ATV*HT=@0jspcrqw8FEu)6N$}-c%rZ_Ng)|$
z7)T*<mlvAH;lSV=rG&Kjv|bi%{oN7;T48xaZYTuMt1o;8q4ab@SX)0!pS=b$m<1i$
z*Oa6uBXSy{N^`x^IBiGczVuKELc<=KT}3f8Y;(sZU>SVCL@B8ae%nSIsVN5O;aF1*
za0nZ*AY^B4Rc}CNFL<wlwc@{B#XCP#E304W+k+<Y)WLkO+V4$<cIH9)mZm2!8(2^m
zD6#C-IUQy<oHtLuU_<MgAtJ1rv+l(-9{=9&2RAvF=olGLE0!lVcMsD>NzgccRU@<i
zLj_@oRd_*w#*y1k5_wJk291Rr`}ItNPMY-n-ZGMqo|e{qck{Pr$kSLzB$55uP9z2%
zSK0s5G)|f)$7|n8ARC{6+QYt0>=^mYzaLTo_lqXMm7N!Im1k(?aGl8z7IG~MnAL_c
zA{=LKGYip?Sj%EWVZ!Ia1m1`pB;UHso=gO5>tB@6?ZKs-^J}6HQR2Y4;hIZ8BMU{l
z*jwJIn9QkJ(VA#q6Imz>qpR00-t}8wc~+~K;IN$1P_U$9tV7{5W(adEYA}3HnmU(^
z7(=T8cOeAZ489&@$oL6iy^Gt$(I4TNHG0*%^so-6=<t4$A7OAdOeA=}M|~N%=>vX>
zW&6W1stL>f7c(Yt>HW0@%%hQ`m;Y1~P%FEXU!ji1hI3!rQmvM%@~dg~_EnbFNlk%j
zx8(AsH+uhdB7vZ+{=c>7T$GA}snYyA(-an=li%rctz<$%a+2W;L9^fTiO4(8t&MjY
z4KnKbzsn5(G7!1I$SC0v;W+`Z*k=>26;X}(w5mdc&1DNsA(`Oykk2fe7)ht26X`y2
zS)2WQ*UF1oBU;NvU4|lZ!cMnr%<<jUhHwm=(s#%iiE-K=tfY@jq*(R-XG##o0iYyS
zjzN^FkldT!kn}cq4Qf1@tZS&{2RCn|7$Qea8VNG3Z@hcz)Wj-67x`1{fZE)^G#3bx
zbmle=eAu9|xMwHAUo1v?u9d~ILtHQ~!}c1;DXNT`&TYd7#c!hhi&HZ5h`H#4!u}WA
zdwBRlxt|oJEKN`BC@LEmY@H|BGvKFhot~W7HFcR<(Hh4kD<ceTPvB3tr1#}BI*E-B
z+1j(s!AXCVZIxU@gtNC1Zce@*cRi-Q^|+ia>>db!nUi?|#YWS?Cu-trm~fC}aK-NM
zT4t38IB2}cbF?jLI>=i%hO(X4a)x_yEuH#=<atiK$S});ZopB9LXag6`**|GT)jcR
zmqIJIm4B-8vgCUx_k6Y4?V6BwK!I!acYGGuca+p28wiLGq3^^3y-&8N5|~G^w-ibq
zlq?8<0W+WYIS}{;dphw}VeU;dMC=~oKft&Ha#P}fO5x-qg<!1f=F#ODq7s#U>yFS^
ziG#NN-erWJ5UI~@OmMiwRuIv}tK&{g>PZiE{U-!@dcG@vwlUVC6#7hV-@mr7?Ed#R
zMNlKu$Xm25P^vbz6w*4q?89=p>#%%EW}2+~(B+~Nwg)~SynF!4I6k_ryzlFWPW!^a
zSNY`(U`gKZfRhEUf2X=n1rS!4Z-2Ri5-D&zIKBVNT_-t~t5u#EwABTNNkG^@T;1<&
z<rC|T0r!#hb#%L-Hi#Zr2eI>sFyDivg{BeVJFj^f`_L*^Mx3CI{3#Q?ABm%zFZkl3
z0VyXg8nGcyK1*OCkccV%QX{a?l?5~lU5ib0T02uFN;g39z?8m10wa{b95=vqLafu0
z!ag(Z1(uc?(Cd8pl0&CEG^?NR;fjn!UT5sk7>)pm9i&VJL?4-4rALNW#cv2AF1O3(
z2+yJv^}F!OwH$ehMM@L__J!y*+b6I+>0rnEdH)0+i3NKPdEpmfSJ7z}ZQXOY;nDI8
zWu*;P&Uns&?MlScPm6xGq9;=?yBNt(+{0lcxtXU^%N()KC|vkBa!BIy#)UnE>W~OC
zHauXgbIvH6Ec6d2DAxJyU8Eopjr%mxPjZWncm0&<#zr3*X+`L<!V?pEmkQxU+5I){
z4+4OPX)N)*<gGaz34JgEkKotDN)W;~`&+0$>f9oM`W5Ph&WL3-THCBZaKSg1xE(%_
z)9fm6>;PJhoDL0DCLz}}Gu0ggd)4)2Ool+Irk&xXl937D2N_1>AEB{%$mRyMb+nas
z+!mw>$(s<{4S=romAZRHf__v5<VKG-X$&PcaP4vIjSNNnk!ywq0X%xgtmM-K`+BUn
z$9OSfmC4i^zekAc{W$!(zGJa)(3<w?muAS8vBq#$w9Pf>taNBUzXroL@rZj4>$n1i
z(hky3*|y2PQ2uParKM?H@$_H|f53FICeKz+%&cro9|Pv8wTdwdn(w}jPW#99n!j;?
z{}EX4dE4^K_T-P|qAP>e=0Tja(d9O?w*-Vc%X))-ZaiR|oQrQli*TTb0suMl!?9xQ
zn5ib(VR11p1u`BQVcPHplTZutA~1A1miwNbJ2|I3Nm&R;pA%aaDlOlh;;qBR;1!dE
zE*^vg^aQS02S0}eIPPG(6R>TW=B?PkjJ_=>97s#+p`mtoh5$PQm2q}o`F+;Mps6xs
z0QMMuL;%3RXCgp0B+Jc{btiDfx$ZHBSe4-tgWnvb5Ebwsm3xEd5<8GAm?KOYIX`|p
zSOYfgJbc!N13pXw?rmAU@DzwvpXJXH>*wB^-EUI+J&Cz(Xz)Y5-NB63>IF*AFV{`Q
zHkUY0%G)h8WL~;1m6Uo@uuj62B40C>_2}KX>gXA^f4P8u7jugAkRE3P_zT7#Km-0w
z@dcEAVVEuXG|D=abOhqGM``7|KQ1#;vlw74h^U-=y*d5Jr6aIO08xV8gJ{vcyuyBO
zTAB-;RTUhaCyKpg#+JJcDr$0b*|A`W4BI0XYsl~seaj#BWpDJLS4X3cb{)X22-~2f
z6cWnSfg<W||0DUK6x?{D?>)&(3-Ay1e`yVh9$3lUYk61EB-^UQjHgfzpiI%OIk#h4
zdOEXj69NJ4n)Y1INGq>jF^l${*(!Epef0h*M?Bhu2@i5QJm7E0fSMLAOiQq&Hh%f^
z`ST7#CMklbJ>gp8!E)+Mht=onowp8(Qdm#1FGN7UF`I$F?yF?iel7i6n%0U9`*jz*
zQuuMXuDl)=OqCu!U;m-2M;jdYCIb<oSj%AO53g9@iZw9F@8{aZ+<y)Khg+{v#`pbe
z3)tVIKVoQ+mpXMtb`NK|A;#1a=8jwMFQmy&Mk)T(I1#qF%p3dSa7K-a?nOaC6)$Sp
zjt_poQk%U>pZHo)Nmw!VXSw0&W4F1O#i;7WWv)gXdk8jj5_6>b0~~gI>|pnd7WPzo
z@o4KO0d913V#n+G$C)SFq$t1#r6>thC^O<=jDr4yofpnsmDQpg&q_}0G^~?YS2ccV
zrIM6H4@OhCObMW#Mt%3-CH=`_kg3>SWV%jqQ8sRvUoec$5chfRCh})o8bG4{3yMa2
z1u{wB0&7>qqxYkkg%db$iHkodZh}MAwO|QDE@&*@C%Xe{3m@HAh9l<Jtb&D~qc*=|
zuTCMPBVjj<ADoIG{d_25sAXJ?N-dFZJ(ql(=*5Xui3WvkG9F*v*yZU{sp5Z2lD~_|
z{KxpesPGs&O&QB|H$T!Uy72DOj+4k-s+uso{puPJ3gJ&f0*-l(r}L1jqpVeeJmms%
zqs$(rXn+j*!Lnxr@*ytVkJ5t6OI!=MHq+7wqXeOO!KkSfONuW|FPre3<)+C?EO%e1
zt*|1*jdm^5IG&UQ5aRlW34GdN-6<a+w0m{h-z0r0X3xWS$OLAIf@ZXsh>HO5E?CoX
zzePM^VC8~&V8`+10SrRVhredXgyiNT9uA|Nu|BLM+TQ1VqgRmp1#$xB!V-`T2~ZwS
zN_Hb0Dd?{vGn7oIy13d~h^t+4;kYxT;4W*FCGQ`&h;nW`J{WYcs`~sQk*AnCUDFRL
zN9<(1n*J{g=HwbkI#?@y)?GOi^@XvNRu~rU^GUMCbqgh>^^J-6-)nq=HCs|;E=6wp
zapJZYk#zUlA+IU~ezaSo$K{iE?c1}v>@P(<T7`9a1?(&&msC6y=rzvnaui)T$_tSO
z4%T8${z^DQKOVb|SZg&QD|U3kYel;`)~|br`IA&T63bKZMN90Tqj4dSal;$)v6OqX
zP(P+yM-)*2qTN{R*;A?O+}!!}@T=nGq6TQjCwNG;05)((A*o4?i-fBH+hSJ2u|lQ>
ziYJs(RO40fUezD^*M1;0l28C3eMztenOb!Em2G5}Chi^nw|jrRD?YjgN))&bJkQ+J
z<<&gs4>zm!Yh8QAsMHFWpCVHQ04w3hCQ^&PO1O3YvFKL~_A1Rbl6xHEg$NBYf;%&t
zUVuzW%t=>oTeYI7(MH(>{%ObWUKX|oMqAGg=h93nehLR=d>Z-I9lpSIUz4=z-&xIO
zMQ=X$2<N7`QT+^_0O`H-hOoi>qu9@UuX){A8I%zsb|BpL^2_QrlGi(C>-^GmVd*)d
z+&A6Y-h^a+&<0e8J!EX-LIU^2jtg8H<V}2ug9dRa|3NGLNWRDneTrS?U}Bvs3UVg^
z6*A=tYwfRF=D-dM2JPvCfyD>ixC<9$w?Ht)s;#jihiR}Lj)|C>{q<i*l6Y&@gQ&~z
zGw0K-7lB|CQZa)ENOC%@0k3+b7#ve59_Y^@PwhBf>nU!q3=15ClF5jB+z}iv-4&2o
zshD&HO|URd%YwXPn@<SaXszd-1fZQV-PnF)6XE2+b|GD!#dnErlM*ke#?As6;n{tV
zVv5gg%z=OoD1RTVu8^~(=OCkUW-hJeS2CHhYi)k@gN9HhP6oSWL6MH({l{(_j~^MY
zzgZl0Tvfoe$F(9mmae<Dt+ox?%RX_R`pJrvBOM)=t<R{*%SgLB{xJ=GN8`mH7Z~03
zt{lIVmymz8+$zrN^0+)3^S1_hZk=CWzm928Y5e?!Usvx^g9KS_SIZ$A{2RK7nzOHm
zx;}6=6&UY4K$<D)!UMIfD?o(+cWum(jikBD9>dqWC2lFK`(p%&PL0lRDHEI=ZVkNW
z5Y-b%BY30`kyu`1ygM3KkOmc#A0@^CpoP0R1x^F90Rqk?%ApgA^xc7V5E7zflfjUJ
z?;u!)k(sWg8woHTCjit<kQ1NX$HIbU2T_HZ4YAhxSlGeW14doaRrEbADXK?eIb6JN
zR7|q0qK?W0@h_N%E&`i(>}7vofA6ppe&tX#_pHX8fFQj5-SWK7O>D>jBe15<pEv15
zQ15fiAJwyS%e4viwc(s_4V$RhlMRe5F8+sIbA9@}c09ofS=xG`-KMiY^b!c{*!_Sz
zi)7D=5)<>L<4&fg>`a488YN?GgP+Khf#S_to8xpgE5?~(KQT6E?8+|+58cCL?8!gK
z^@Txp{R!T}oAMLt?|%DMu&eK7?u5&O<^IL6R`*?=bj(@zybMx<g}++92sF8$$}<*(
z6~x^1^amukkvlwLsrc*?+x?BmxNp+4;G^vNLab)Tp!B;T#UWBc+!|1)XeyOVtg`CX
zUp@<hUq9<mq{<86Hb~q9p0JW5^Nz8>)WCKu5d6hKNSW>G)i9&YEqm!JwT#DYfS?8k
zR&Fltn*kC0RnMvO0BUI8^v%*TS8z_jyADC$gQFk~sQMS4)(6J7Za*UD@knlwD7;{e
zjbZoXEni{B)KjJIKz8E%<{~s8bG9l+$NFX5_vDun8upiZ2Y-t7jXmWgmrBeIl2E6m
z)(dLgt<BoCI>Q`=vMp_vsb4>SzHAW5)ZQ5c)mYNaJl0Py`<GqKoMCDI)ni7lvZ2r4
zcA?jY=~7k9O;5fdvsBA(aQL06a@)PO@BbOOLTu$Wrhf#yX>f^L;;T*ddpo~Ivcf%B
zUj1j^TmS1a?DQ4+0h<K;H^^fu9)Hs5w65H?VBV-i6xR?QsPd9`c=^!-DzD`Dob~4O
zOI^$fg6nM_>Fe-g*EhKu$~6SOU5La|jfu2brZpK%O+6GkR0oHhbZSP9+v-y-<6j^>
zTf&6w7;RJipjNsfV{DIL4Ni#2t0c9Ec59CNwl}<{&9=gzPR#Fq`w-d&D9OCb)=44?
zl&ZtgvbwGi_YEsQ_<C=>mBnF?MSlBMTAF^|*!oPO_`<uLVt%_hWPu6+mryZb*#@ed
z+;j{42atuJFD$R2{ilKN0ayWU1+n-+I(KsmAnUPW)ta#_hy4G+=Ninrt2tb>Qq)B#
zeH}q>$yuqlR%~!SVuEu8VfGrsLnB|DW)Gd5THJo+gZo9bMUp+?DGr+F)>@q#@~-XX
ze0ap*;@E0#>Fqr(jhDRL5>9?ex$5>u4LxDkVKtIi{0|U%=J=&O5l{3Kf8QQ2|Ji$2
zaR<tNPZXK7ndg&#{0jJaw)Q&0iCkl0c=>YT_i>Mbf)K{T5=P9)Ls{JWBLP{ez3lXN
z9n~_nGzYApOX9yZ(7F=y99mrPv-nU-hy->n&-ioK1OC)uqk32Qzt2e7L}t=MV2rwt
z4St^&HN^Es#ct81<r3+4EJ$mZGrr>Og9f7K!;~d-)9p$eEJThzQj1Yh4eFd_|F|0?
zsq3e`6EzSS<U}?~J8l|2AlAB6M*}aQCnYmRdp;$i)i^Q6IQ^Df%M?zm0OKyy;X|&d
zE{)|A2-L{xjB+8&_JlL+&Q42fsn#(-cThK^f?cu;Js3$K)=cA}*L@L%E(Yv%TLzcb
z{dlTO{n;n1n;bkG*mlR3)sSh%-7s{%&XS%$qdE7Uudct<`Q5ywn9#4ezUkTBXC_zV
zNgg$^bBn__Y4`k4*C@w7Ck~vs#o=5JF#i?2iMh_-PIxZfmdf9j<#J*@D*gQc5F{Vk
zXn%Ha<uLE8>(u4(H4-smS$`7T0cSXnuqOz`?xtY??Ca|*Ssc)LKuov|BABj-38VG5
zi@CD^!u6GUeu=?h18!w>8izo3p#YaxlKP8%yvszpsiOIcHa3Y*Z$8|)!U&teh;@L}
zt!<uq5?QMW_t1^ML>Bys@sKEv5uR>9GmyF|=8w(UCJ4I2cE!3?9HE(NxUE7lJwbXU
zaca*y4?4qR9$ENKGb~N~b!kEm<FiEA@wmcvY%J*;=hlrYAH)VISR^@L5IDB}$q20}
zB&v`qL%8G7UJ&T-?sZ5oq&~DxyW%$_tF9qij#l~q;p<Jnsb1eMU`nMZV>FmECrMI<
zh^V$PLjxJ2kdW+BW+j!O$Tnpz8KMxPNK^=$C@M)P*(792i1^lT>wnJsf8X!>u4`ZC
zbWV}IpWky2Ypr|T&Aa2o=0`)Wv%e^;BB(OtnJ4{t77e^^*MHhSaBgkK+^H2pYB(Q#
zN$I>*VzgyJvBN>Ig6Cmae_rLzwi!<u8LyQW)?sOAUR(IenTn_Sd;b3|R7wdYdh?IU
zgYNHrEGLB|Lf26@NKq%vRps$@S=`{w=0|D%<$ia1i~_$2H-7^+1soebXGGmhUxQ5G
zdm-|KxwaX<unq`f!Ze={q0mQ;2x1shD$~;?QTK>~8T0w*e;b$_jNsGXsYvKN4F%AL
zd+Sh})(cC*zvyxm`W+Zw5R`*)*5ASKu|Ck2m?of?XeA7Vz9d#oP6}bDr|*db6c-U7
z@5vewC}W?`J`>FzRZSu4=DAPX_#8w(3gmY{YXqc<9@_8#Kyi)|oS0O*3<)%HPFz`;
zq|A5`v8YGUK{<)q<b>=NV@HgOLJ^8r>p&f2hOIppMF3mkL^6+@Fnp28;=o7jT>H1B
zIdwkCbDx3VzWoM%>PzLGC*5TYq27C*c}I8Zz5i732aed~RLwob!<*gYi5+Lv!J;(x
z9k31kM>|2}9+SI78$(JWis}2}2zUSSO}Sq0aL9m1Qr9?YyP-g*73#GgJj3st&Yg@N
z`ZLM0d6ZDDSXK6rD9wBGS|hh}Xn!O$_lPV<<!-QjpJvCW{T`bN(M=+xa7Flbp_Pf#
z^}BB>O$>3-dHWkusf)_zIdLjlA!G1ez*n}Wi$sKq^p9&X-$E`XOf@f|`#>Np^BC0(
zsl&>YA}E4T^R8eQ$=Or~8%g7W6Ya%>?L3fI$B&60^lA|x*WN_N-BMyIUFTE|_OiV=
zZp}>*j8Hnk?gN${kbZ$AVW^#CA>bA4S%LLjY~L!kxzF1aiIB$MJu6iXOR@!pDAA*V
zih?&iN{rkAC>=pVL;l`2<Z$`U+NtL#-6rWPl*E=)O~~ifGkdmw&ZN25{&-I{Z93Pr
z_6u3|%<fn*$##Y-H=!ZYFzfVC#f4(r<%5jWM;hUflv3A|e}81~S||?8z#u~^f#KSY
zh_q|lR;4_`)-J-W_OHd4P)AxL^d{4kCORb-KBstxnByo`k3lQ@%dEMO>(bcg;X~H1
z&Qbn+pHP&EppHH%i-6_V{jMFJQ47ML5zM2Yese>&vtak7&7ScVvgz;49NNTyqc9Ck
zW`It*$dZ_zhd*%z5an>G;HO$F+#^lpf@X0^guFB{O$Ra9m(x2W(CB=H+<%bS?FYA1
z4Nxb1(Ppc%XMZR4h^YZIN;nL%rk@99V&K_DUxbKj0m%gk%P#<<lyq^7vD;>oiG`Ke
zdKWp9MM!2Psf+bYI<NF>BRp*^53O=Wkc7uDiHhRE-4Y=UM}&VW-7z)xAVOd5FQ~xr
zKZqUaXmy-6V6^?S?3N=dNLT;Y<mKu(N;|W#_r;n&f9va+0X<`5D{<+~ED!i?u#}<H
zvb*Zg7(<ztPh$09*N8OuYc9KRW^yO7o8@IZW$fJY`1ZY){|@_M;~H<gMb=xSqhHjX
z;vpCFe>v1sCj>8JgDnHC=au)ksUz*6QK!LR2X=MmobUEdH(qZsa(cJ5RAOEQ=a?iO
zQ`1rqNq9kBh1UYFgw9lsRf@3x+rYd<vld+-n)3xcL_O^^#$x`$>o_g>l?4{Ex*QN#
z?|Wn02IJJ8k`2eVj0#BSYGl$87T~ro2#q2-d?_p!g3B6e49}2@YB0zEy`ip-yBe|B
zo386KOx~=mQ?SD|^r4;b=FMC>%Sov)-ixd*vH}%SB?agW>7GV@AT#j%oQKOJCEND6
z45jTKV23xI{KZq?LYt>{6eHmW0ApSMy#(3!lf7v@+9;4UAi+>7%G6W(TnXd69i*$_
zA$eg|#!E*=wwh|z7G6rZa!9QGh_uPmc2@`PPi%Sb<i>ZxYbf=uj3`ap2vhPsxICL3
z%+e-btXk9lm@6s?>oh!#_WS+ZuYBBV?97cwhc5daJC*J%e?%{?3#=hTow(n1^(+mx
zHEA7JbG|#%>+GmUK?Q)!ut#qq+mPG8G@`3+EBK|Vw{1wtSlSA2D6rX6gN03mbq!I*
zwe-tng|TZM)S|~Y>~5f^AmBC&ic_Q>sg_7<RB_1js9|aL?ZwkKfHBLdPc43&z^D}j
zG(Pasw`5`Z3MlZ^$Uuao{Xri9w}nO)fpTw=xi)sd*_j!u!X4L*PhHv{ajAq&`_1eh
ze<Q<(du!zbxi@EQgl}54t&#YW6&h{hsu@f_ZdAwVvR2;Yigsy_smb40E`a4_6J8Lc
z(wD2>;77>_{uS#{#T*EsG9H2gna8;^-*(9ss3GGBl(O{i(v7>}U<j~AFLG#8bXF&S
z?p6}XtlH<o&~6fySWC#f6|z0U(%XTQvE4F1f`a!APj!Z^>GE~8y)H`I4P&)D^;h(f
zao9cosr`Zj*lwl|Y85+SG_Wp0aDA-d_MM)a$?{d}r(d*B$Ga|VAGCI4rUDSCzjyzA
z?ID)UN1e)k?n|I+XMhkA56yzJ8eNFJ5k(EIKZ+V1B+@Zv+ut7T#Xw5{%wzWpMnCco
zM`^xWApFvt4@hz^$jDM9Q)|)w-CXXVsi3a|vj$UHzFm{E;DG)!o1k>74>hx~4yYCg
zp6~(|M?M<9uh>^dnUB5&j=uDf2{+YOA0voO!6{s0?3#Da^DRU5HlU8Rah!{lrkpE+
zw9XRF0xOVgfNNfOG9!pVgvK?iuY9@a4Yd_E8K8DXk#JU-;R3XuO1neWyxo>$;F*&o
zKmV=AclIlF*_0=JP2@r5x+~txgqzqt#?m4CtYo`8Tf5&zW67yF7RFt<8B5r*UolH9
zApdSvH<q3MJfpf&jUgpjO^>|y?Zqjpc(Tu<4a%@xP#g)mMQk#mPmIH9gxGCb%po1V
z@r$CFgNuqny4s-J&glhjlD~W9l5ktD`akWoe&FRu`?1pZQp5@@>g`NQEbII|@AoC{
z8dDQHU8hA@GlMTd{enUUyU^&m8x|~`QhNp*mz>gL8jY#glqMuntq@aNuir@Qrqa>N
zXwq=u60KggEs#B+MATx-4BC?zYfR-Yd&92;_;~!*zKmBiGEv(7^%W!rpWp&RKs`W@
zan>lKhi7zYO*NcRUkR7~Dp<1@ps1F2+$JR%I_zm6P8hJm?*WsqLOE!Russ^$l83v*
znFLm0r3iiDH}oK??<FFGRl^yudI)xT$H~g52Plg`?KYhN2t*i)?sCTF5@H67-H3aq
zSBHV(*Q<hlEM2qJ;ct@p#)V|0se!87r%qL@YU?zQ`);YIT&RPHb)T<OcDKy(y!2bM
zJEEcl_b#zTMmhgtiBjA1j>Z#TWVOv|0@BJX1A60&9#M|B?C);QDm*;8pV;x$$)n_X
zX%De7(VJtzHgaKtJn2|}zxA+=LAhN{)4TRecuX4j<h^CY-s!1j)rY2H1$`dQQR^Z_
zO&~nC04;v9==h79`;2;bducM4M5Y8DP9IRFPY|G4M6sJ^^_1dRr@3DZgB{Qyk-SSR
zek5B&Cl<t`Zno!HVi$2v2_4Bq*1qRrHKEzrpa)T+5v0K-Xv7fZ%KlEJrKkCvJGZoH
z{FY<9OogoNjf|U5cgaAVWEOVAF34$Dr5r*f0Okc_V(u0A+mrj;){V2`&qvkrhI=#M
zI}=-==$jvmEixVuSH=0t&Fi8eB1+F5q8g*?Zf_fB)wZ$XUI7^0`GEU>ZDD%Z;JjIE
zTjnKp^5Lmf6$^-sWrp(Nqa`qZ%ot=?^T`iW6wD8v_FbV|g)l9#JxPv7C3WepaKT43
zs03{#xK&cg|4O@U&q3NPy{NyX=)DUD2gkl2z=|ewp&?(VM>P|4z}etPtJRf7Vpg<L
zkxJtCQzXmhe(N8ote9&a&U-{I-~{^$JQn+@7gkhbW{U9QQ%pc$#+ZP7dKf3(XM@1w
zpdMA>bmr&MxF^6giDPMLPFVmrp(2T~gG7UHn(eD`L<fY`m3CvaqjJyif>d07t&$ZQ
zROE|6tQ91TMTn^d!tBNG&pLdo7xxb3AwFW?i=Jh>mVHp_w`9oiO?g~Gy2i)Z%)s-~
z$4Tm(2)C|fcc#nf8^#;=$nPXAbDNFNSl%?=^UMkCLbf#4>I}~;B!5Dxj%nLVM{ycm
z%ZMDm38SO*#my@Lm%{}ESrNAvnFiHihT|z8U&9~&%O6*9kC8o@HSYzplGRjL`Kq=K
zTpqIw<qc+7-SCEGwQe)pJ5HB~y`N3qrB^-c2$N)@Y3p`c9@a`N_^{TU;iRh5u}cDJ
z#{y2>E^s+~avWOC#(&r69RDAqKk&qjRGh?~j6VauwJ>Dr&bEz?dDK?X%pQdbF7{B{
zti@<v#m^a!rcL*|mXC=7{e5L`LakfanpXvt2}TdGZ&7dC=SxbO%yGg!OR)!(4$~uu
zz`PGITNI%ybo25G0V9vD5-xCY;)Tv$xOc)HBk^0~E~vofR!b+0mvFS>!Pfxr0Wn7S
z^1$ocb50de4(A3Xl}&uFg|9GM?uM<RlY;p{jB%++SdN*v4=a8`3hd0FS-}q08n>IN
ziHw2!f6S<%WV15^&?;UK+Pax}5gJaEkpSnGRRQemWFvGD#mDIaPf)~5QPKdHMRo%k
zXT*&L)iVrYs~L%5!AEH_DR;-#UAuAY(iIw3Mgo9#i2L5KlR_!Da@8bicSG50c8DKH
z+f6H_bvSOP%`%HHdb;i9R<SWQ<>_J}wFULLOcxzKLVa~}!R66We=!U?yG%u!|D!rt
zkc+=q`R%`F(ti<f&F`h@eFJy*a|+%Sbe)C?(38l>M~Brf%zv-GI(28f;)&nh+DM2E
zfDq#CVXzogG8$_*t$zOgeFd>M#|}EUTD`OECf*R)fYe20g&G$Krj#jWu@-1e(~L?{
zjutpwFvs<6`EqvjX=Ll&8kkwa(E*d@cIE{)g9%U>{U)AaY~Rg#a4UO&rW|xpGqgUH
zp~>uctw|8m=^8hzql$_QtrnqMqg{4s*E3QR>h^)Yk>2FUPkeN@j0=QNg*yr}!dR`c
zm<RPh7YlJsxCyzc!a8E3Z0i(mwh_0I3AV2|S$~bB{cn3KBW?!pXFVrwnU(xaEg8z=
zLYk&Lscy|H`4_M0ki~$(>6sZULJc`K9m>XxePNQLvj&o)b{y?V+|hkZv?;%|NvcZs
zTu;wCopKi<xBR+s+WS5ct$uBka4k)qn7MKJXAe$jV{01$NnbziX{W6*jSa}n#%qb0
z74`9uDSw~-$<|@klB*tfW1<M!77z{8hIFS40;Rz~7O|EL+Z_daOfs;H#08XO*CfJH
zh9mhTx^-}jZh!SagM)f2>d<xI8I+GkS_YSvK6GZCE*v)23Ez;-=#p<6d(2+Q&J0Nv
zOaKT=VZywx<9t^&;h%ZF%a<40a{1tcwaJdcYOO)qthIVSj@SwxUu*j;-Z38b-gFq0
zFt$Zg3QLGdUsyCslu`Z^c;vXmXZXC*GQUF~h(yvsrEz=$KlU=2kqN>_zZoh40*t|g
zE(SEb9q8LdnKVv+lT}msZ^J9+z(urQ)55UXptw(EITFK*^6#GLx%V1TNK|S2<Ua1L
z!)F3i)D@~4SXPtAQg7WC`zR8Q@uKCAVs+}0jP_CMwe0d2n~(n4;ktU`V9<I}$$E*L
zAL2Vb@732wEuq2IhwfjZPpzw41mo0&;g&xHZ9_P+zvqy+Fys1}&N=_q8a)fC$Zy>W
zHp?%y)0XUXM`FY*RL|x)F%}DYf0_C~WsJ0j5W1e#^cE1@zsU!nzJxW~`A*MeI21%|
zva%Pp)|v4@@#E4SwH~aAE)E>eFJ7pYxNEV#WLdY5G(}gN&M_dpjKGZc%hSouO8Noe
zJE0>rUdQd;vTI+RZ3Rav=Gx=9qP1Vh3)izt`D>C_2J8w^L}G*Qvm2i{qkT}dlUt=6
zv`tuur{8k?B#<G>@M4=a5HZ5lb@2qT&0Wb)%NRrQTHM+TW)29>5P^_?aj)t%Oo=_W
zN_`4@tvg-)yuRf+2eKcZ%1XI0_~QJ~Ura{+U0gJ&^mZM4J24g@t@3z$Pwez<dMh~+
zK3^Fg8B_LwF;eCB2M!u%mk#ytR&Aw~*j*!-bI|SUe|cctIA#1ljZN4dm1v9nf=6TC
ze!1(1<a3UZf_YuDo#`U=O3p04ivFaHbFRby@1Ga-+q|`6BBHieRiMnp=>X%z1;Puz
zz<<*95W#xN1Vzx}qE$BR7{8RhI_%3#IkyopCAZ+92fKPgQwOvKdWZ|TO={vSCDLKX
z)J`C{tV9k*_QHx+*4+SDkO1({khHy;d^KY}@zA>?X~F0UX$p=S-NcCz9uI`M42lLM
z_5!PqDZ2x&$^1&{nH0I5FvH`?y5)ktsBh>%Wi=E0W%`;cNxyKz(v3+g=!;}UltOjj
z3u(Tp^2Q`oUHAz&F6qI2#`FMs*j<3wVAZl`Z>PF$%JAHL+f3vd&*Ir~-kF!GY)Xe#
z9hjdVig}yUyW!}~M>8iqe7>KG_;z{X=zfoYtI_8JN*)N^b38G5qJLBTx23GZ6QUHB
z^=u=J3ZI|c*ls({AujVe?dgLhEw48?cFyEcJ;SawjHG@~3~8Wdhg`eXaIIn1SGIf3
zDd<eMsn+ppDdRgQT)&>5pA8?q+EKB?-Naa!IlMBq<-i(^9u{TeVa@W^gQv`66RxBM
zD5Y{F^S@awz!m+~k*fNPXY-%^y*D|BcLDzCPL<a+jv0L%nV)oBOrmi3tIFpmZk;N+
zCEj({%<Rtnu(mHe%0nL86CK?9aqry*^ElP_rA_Ah)vhEa-feifxL38dF8HI{`|g=7
zJv~V$4Yy3Lw@b~ubpB-kxt!LsGp6G5{QaMd^HX2@R2~L-guW8Ke(2t605`bCKZb`#
zM@BGy_ezkpv2VPH%%s8I3UdV2p?u(!c<q-QdE0}`2$lga2-08&qTj-BLC{rd`4u5%
zTmP7PF`m%s_*<+Oat|~yK@PQM>2Ob_@~;QM2ww;9S?pi>;M}TLnum{a%12V#T1Qjz
zQ@E;}Y71D2^{G54-U(~Ta9t`us8)B$5Y5&L4%lkBU;qzpsGi125rx<OK1!9zXXxzb
zeN`ipgqQ<_f*u8j+j3qJ29b_8LG<eMXOJ-F!}MT$Z}1WCI3J@ENjYh=_v>C4@oMxe
zDXrJnDQ`8+HIKdbGEF2ojqi(@vAss|Isr_hIrBz!P3#VK*k|k)d?D7YlT+ZlTshZE
zCfX}*yW$P@8zj_uyaw8PmYBSX(2si+kV%Q;idC8kZ1;#WNtX~jWMJ?lv7`M)4h}@`
zy~~Gt4PosQ%w%#r=Wt#5qQJ~c9xyk~IYZlzZHs#kjwW>do?9grB>EusgmY~(N)L=b
zUcI_ZRtmdgaj~qfG*>bQJYIUZdCQ2TQNB$H?1XX$e0KcCIiNhzNq>=^t$CXPx<R~U
z8J^qO-_4&!CGTZt%2(H0hxLlOX@jNbBB!<_?ch(a&9GTBJKF}3_Wb#)m<u(}d8xt}
zt&_ezJ<!+Au((OCEsn&g=jG6UYS6lrSELvl8R?Cjg|l72b$A@<bA_a0R<X7~k3w#=
z;TJw`cEzO~DHfC8_zgOKO{eMcEE#&kZCBMMPWe0+o}#K#;BtZLLM3FFqW0xpUU~Oz
z44jX8FDebAhHYNar(!F`b?V#6P$2_0{^&@`@V4lC?`hdvyent-UkRvb;ng)+^<XDZ
zqQW#8Ms_FbIn$}%dn+_D%MS^Wrp4Se;z(3gl9HDqZF}a_s;Ta;H!mm7Qo|{$^>!Zm
zJ#D=)@$xMnz1#9^I0Jdl8<8zC;$X0Up~{hPcpdh3+!}C2QNv3%0h6c{|NN%8=<4@W
zk)56?6Hf=a=B@n(X6_C&hNTBR3Q9fU@~s;TGOCbMkLojVR&P#dW!W%kSn#_M>+F)_
zUmnMYdZoSn$fwRuSgeiT?%x<qh>c<3Arsm>hDUW3bL5`bda(;NyzlQHTRyWkf(-?y
zO#N(`WW79Ubv2)<3s0V?_{>gm2&(z)f@~3k<;3hJ_O<bqooVX*CO;?;tPdFrse>th
z`Fgtz$*B#Z-t>hPtd+M0UTnSSZgInIZH6NhL}O{r=bViXtftTLceXu&%1}P;sA9rs
z*ecPD#JOmZ58hji#EVt$XeymB!h&k{W=1d`DE$_pdU-9QG^cFVsnplX?p(hv$=C(#
zkUGoVd3iQB$D&B)(DgdjUx%u-X^+!p;{~aj&gq#2Z`J$wiCYu2DS-U5>`XP{TYudV
zuBgpG=8hn6<<fdKZSB<%(l_Zn3+u|n*bNdD6=~c36HGA5ANNOHcX<A{7I|!=`nB-N
z)7VKT)6+Agz3P1K)F;=^{`+UvOX+atJ?DIZVL%_X8fQiDLVbj*4W^l^VDam$c&AD(
zKq!|-TmqdeVhVkSwV}`P-ERuH?V83JR0SNi>EDSJ$6H?}A{!2)cN)1a53OK#(x9Za
zlHdaa?D-gb*am2vO;GLg*P3twnmvu%vgZzCjhjcGka)5nWomzmp26`+U77rZx+_fi
zD^Ut&OVj0P$Xawl4pyE3=p5IX9@sfGr#HwpVEQ~v-MPIqm!-z-eb_N<M3|cs$j%;H
zb^ZRaB`SXG_<5K{)lwxP+iI}i3aSD4qtEZ3%YoO1jjyqSKeBUs1_xFCMHKh!c*R`(
zRGt6S`OH%~(zWkJf?9UDs+<$<a@u$(r~be(`AbO|Ol78*yH9;9miiGgreMb)>Rh&n
zMBOqsCpf!GcYj!Yo!b`Agtdb6UPJw~M&ln=^2c~|EEMe&Sul_5Y8rPU*Ef)QKf3N%
z&dmKKv|~soROhmz$}CEr^{jo;5)!g!0)O@Zhh^2+k&{^Fp4*_z%hzl197i;06z_4J
zMh>3#38Rb7f^gp#q8WjwE?taL0cGy@3{EDEL{>0no&tio?UTNMy+8l@#RE7T?-vLS
zF*bs#L#E!sIrTU%qI)IeXFQ=xAma=W!Zsrd$t42Tr*$b0#Zx#l!7X6{6P#T<1#o>(
z;{5pW<FvPT=)rr{P4}E79X^k=%O+R7imBO?XwxjuwbD^t#0n0AC*ogTKHgpo(r(_o
z)!ol>A7#w*nejBdcIe(Nq1B7ZhtKqXQqadyTDG<}D&2dxVtks9A;}+im{~;1Ss@Wa
zz=!}dGv44LF{$3_K3&VLRXiK%>T^;3`ya8np+6}}rBYHKw)2Dt1zFzS&QGeFjHezL
zAiukLNyyoUfB0BoRiQ+-x_*KqH6c}gBG7LMC8{-8m)d;gR+&=uohNOLONA1C_UPV-
z6s&mV=0!H(I)#uWv6*b4q%=f6=eoO?Oiq<Idt{R*KjR<MIY03QGo_*&Ecx@pPw(^h
zdAR55;b>%cDmD6dZ6GN*eF*&}#Ecn3Nc!CT{loJr=GL*Xu@Nv4MtWY^E0`*3upip3
z>Vj1Q7^VW|EyF|1l!CQ8Swz*&#(ShMge|$%K!sqAs(x9Bxx3p;hUaEeAHcSH8d={s
zMl`shM?8#WAdyQw1#;m1lA!3uR;dA+TG)hvmKW=CS|@TQ_*aMv?=`}c57l8;ilRio
zho!nk3#*w5WMn0e#8h7ubC}_E^TYQxixIAuXRZ94^+CqSEXMBs!Hpm`s%<Z}IbXpk
z;Dxc&@4fU)WA?y2=fPna>;=s?&%vaJYExxp<zn7wmdlIWr<55gCX%e5v-!`dtO*=9
z*={HCjT$c$w#>Ja@=wm$tG!F+6VE`{2~|raS;?)gQL+J@7er|i$;MG}i4Et9^`$(w
z7OTigvtaQ=-mIkA^S5WxQ|60ah)M`zM&MFnqWu=+m`AN~%aoKZol8qrzkGJ?<u%j@
zk77uVN1MxnsG9p5H4QqgdUlZIM{cIh)~7#VJt@*%)zXLN2R-k<mDs6Ax)gbw$>mf@
zr#g&+K6AIku=0ZD0)CgtzYh~D8deE5CtQgaWe(UOAd{c$%dgELC~%k}T3gT%d0L8b
za;mB+7+*T+i?1LECQKpX1<58|WD2~xa$NO9{5jL$$RU-{;W)nh!!mHg!gVwJo-re0
zyB;u-94oeAH;qr#uBT5`gB>JM_U$hhLid&`S+-km%@sTXc!Kcbl-;R3t6c4)mU)I9
zF-=4CXpy;%Vge%rL_eY=ghv3vbku`UgZ;JV1cQP*uljUjUU%z!&!j;|RFaKR6q5$u
z($dh5f0B&S+6H0Y`t==W_pN5o8LCe~rh4PLpYnodpYhAWTeEw}1*N0Dp8{e<w4PM-
zH^wC<7B|>GN8Y$D&Go7%e7~pF@<*fcD($z|7JYf#Usn?m*=60F8$cGo>MOf~=t%4N
ziBtX>_WrKxo~RBL6x`fU*yMU}y0I>1URNz$QZNAF3=bNTK19wKLP0pg3%|w-$$#@(
z)Mo2MNdb=W`wv7*+&Lh<T8(L^uAzNaILe!|wa0W1;0bSSB$xDX7KoDeb6*w=hn>Lk
zj4C&HN=bylY_+}sc--{#&mTWl;g%r$L3NIVT}<qi(>4l|m8z+*$~kygHrD6uT5-uH
z`>hC}56^M*;h{ibTs{mwnh<0G<8Zf3PGeOK*&dbQql7cUfw={w|K1q#V(%fFnw!uc
zo3t-S@Cxuq5n={QH&q)B3<?JVX&#E}VwaX}2VTya|7oA<P@g|GYcsjopo2R=h|Ium
zbN-+7Vi3li(49qf)pu5-L_OL?k-zY5^q9Ds{_v0mdC+>AwBJBhYFMmJV9yQ}2PGx9
zIH!n66Vv!Ze*+e*Y7I9N8RR*+fMM6Z)9<3+MR-g;>`T*C&P{BX5(^il-M5jV7Pei}
zrGg*@(hL8;1+VJy=|l>`%;R+^I_=Cxd{fv+P$!p|9>{B*HkITSB60GrJaLzVZ1MGl
zvGm0nZWHn9lf+1<RoYjs8i>gjhV9IFfJLAkMxr1}Eo6B@IA@#%PtR|qSSw1ROg*j1
zFSeYuXl(3rp0}M}s<DeKoWB}5-f<nEX~%e<HDt(1De~>H{T^pM(Yw&MxqvM~DJ4qk
z4xY=-hLq#1a#B%5_Tfw-TLcJMUA?}jZFsWeK3r^|Kb`W7v8-ruKd4CbBq?cnMpUK<
zrL$W9HY7bOdLvGqA#GpmT5`yIW-L=aU8<rdTZ~pQzpP?|Rue}@UiDcgkF`y&dX9@L
zN^Ed%KIup6sd1gG$X2~6nGn3|w_ld!Km%e?8FO77$)VP(OL?%*Yrd&@rlQfj)nJEz
z$aUT?z8aw802&Q9l-G=%KYoNiWoCl)q<Hs2bfs%B2q!<ZgJO2S#;#et<YIE$w2#8H
zkhu30K)8Zu7A7Ld+IWx<->YpS*p3;J5wfaAZODf#WLGm?oUiYZ`;{86{E459L|Lel
zzU~01f+YWiBELQwDwyY!zH6>LH+f(v#)D3RM0x)!)+3wD&V+FCP>|yYW>8CW#}!V~
zvo(E!z-gukX8!Xc5xuiEHB4wZ+7;1x8dA3FbCu>or;D83G!B6zF;9<IoZi$pPq87r
z%E}KDV#Z&BRF0<}&)4MSr0Nck)Za?|n!ST{zlLg#zt#V+ma#{QOY7k4h_k$7Pm0O!
z3M{sX$5t(&wI&)X=BF%Sc=0h--O~5Zoi3r^4KLs6P}OHIb=kD;PUx?5uK8$R=1Q%*
zANGi}X)CTl`@kZ(uRnV(WlUwnyoiidzQUJ`#PG^?46ZVHTj$Y|D2Gp*dCj*%T8#`L
zC2^*6N4$fuqyuWP&17?>wG4iLzMi+)r?XU7sQbZHZy+zY8Q=_4)O=t3`Lh$*0`ngz
zjnXA?K5kW!gCB0RGk}tYRvNxD|KL9;ix6i}C1Lv6rCn4U$+CwbhFlLxdMH)qdaRYH
zV9eBSJMs=)Zg;+A05zj^E~l7atq#5k3M+gSCJhd_>r1LD>aiamau_<g$cI}S+iaTl
zvL_(g)E7+o4bw;Xh*0S6(s?X1kn8Hz^c~7!i*fLhE4l=cz-y{7JC7FdHhrDJ$nEbA
zg=^JQs+$~%3RpwcYU`$0U1I!c&jyr>@a90>h-*1Ke|-96bX3DMy%8~fmYdV>cU`Di
z^ZeZ>H@b?V@5P{1EV<&D?))+4E?;a++obS?1IQFt5>m<d*9z<TZ`*Un(mY-r3Ut$e
zO+l82n6{E;RWB(fHF2wwk_%dR<nr|g#N(X=Z_7K`jU66PbJCc;B{Cpe-{#?XmQQK=
zkJ!-n$GvD`ta{gw#|(-xZ_ZqoS2<PYQt0|*RsEyHvYIiI5%Zn?_g=r4?_O_x{2{kU
zqiy~(XT{3@tYW3@&Tz;6sG4JVdpfKen7c)cxGzfx7yf(#9EE5$Pyr+UpFe+o)`{Lb
zz%4UtpEmKSa2kq4+pVRy-Uwo5F`lVpZf#Lte&kRDCAdTbHSrZhTE#_#pk%@;$=?#g
z@{-&~&E9`3R(78;TP7+l2G>a>vTEDsVs-%^0xkCz{OKH5#TX^V1=hT*n{??OdGO{i
zhT`COEW-i~$?!56QXHusk6Aar$Oi?;2l_}}bDG303V$$gcOfY*z!k`d=os;h@b{Hf
z&E2E~Sf6#8YI)v@7R8o{A>hHbifcka=WpHVlkM?#p=G<5oYec_xryPti9u(gCW9`6
z5sqomuCqC|<ivYXQT|iQD#nv(<;CM#K^_$O<n-Rw$f!xmakdSKNx-NdH*j#CYC3?k
z=y^;gf20j|UQtbxNU<_7u;``3EJR_pn!|S9(A4E)qbRSa#NrFmI#aJ)ebw)Xak#;r
zmYvC|fAK<o0doq8WMLaCB0@clVnyM=C3=vymr(t!gP!k{t1AJPEaZu9oYA94kHBr+
zu!{>_9F?la&J@9hQWQQ;2ck0%yo~f@-2Z#%qRl`I$qSholROJ*1++FeTLj7fjO(ur
z`?~AEyGYmd;sS+dye*vB2wgQcB-*MG>kMIBy<=y>B39kk6Z8rH%bFMKVRZKNW(re$
zy@;A-s7yUteEO0Jw2O##bnY7uLUy)lf67iLGuhamgQ(5}go@EQ@kWV}0%Ym|A)tC2
zSM}@APBJd+L!i4aYRAAXw5zs3Q8;Me!ViOgaTkYb`ze=C8{ZY{G1^kbO!)61-q~4O
z5A}~oi`h9F8XX)dtNh0BlR0AU=Y)Bx{ChE)J2@t~tTj0)|CY}K?ye2Bb&lUfm6X&f
zx7F9JeYjy__HTni>&dZ`r$l~bCS`Z@9e(;%u)OJaWA}J;T3FZY`^nKaKerxUmtRLN
zse;w=!cZQ4F0Xevm5w-%1uA<P?P0GHgV_LBus%-=q2Y7Jhm2PJ@zbZ;m>$Q3JW!K#
zW6bZB@gb^2uwn2|GK6keFvh(}JTSj%Rx(7Aaq^j**{5O`03OMS7(E#o{kj*$HQWx-
z=Xb}YN~&roB!~Eh6c+~jR+#5_gp4^We&woTV;b=pw3)dN2(HV)k&5%o9cX^=^$@0V
z<ybGBPkrOgf5m2m!wmTdAO+qCqQF_@bZ$GM4*2lmWYG(Jo7Mpo<O=$8s!eWuol4g<
zmlPoHu*`o~pPMn+eoR6ja(<tQc#!Cgy`?PEmPblXn&qxL6H^ugm<8c2s_l~NuZaVp
z)I*b^Kd$<Zs5tj|ZqhajWp~-y5qGPhISQBMjJQPRrb08hR`OWP+Hs*98B%AIl{|eL
z8|>O}G<mHhdqsW~fgHAPu+_EyW#(0w`f$@~d5)+;G%mb#v4my4_E(jus}mJ*&pCr#
z@PtEXhr5n)z#y|yjE4W3_Av-RqUQ~z<FiN>Y&*M777?8{hAWcCzIk@WTfDFedhHg0
zF7e?5Ck%(+MxlbTGgIYHSx&`et7={h-wrLKo<41}n34+?h*{X*asGUb$~<-pZi`4<
zPQ7L#UN8S9>d#I+HM<!Di^Jk=NeluemD)yCCyN3^Nv{>I54WSx&Rbb|>+SJH6j5-^
ze%EaH9S)yFwap-KHQyTTsh;@RJR-fYuSaR)mpnM(<>BG+EYmOriUkh_Y?PQ)zWyO)
zc>0yecHz|l49st$&CQr17)p0{T=vP!`SehB`${G6V$Tdc<6e!&hAn*Il(9wOwqw1O
z!>bwImd|F5OoaP={caF^NJ8qXnT);4>yM3Gx`v9$8Lt#*?nX*p)A<G)I5Sc<B*-T$
zDg0i24i`jEw7I3eF5IqQbyPU~T%`*~pD6;KOg)!aH08&t<eB-=q~Xt=xW#D_TRZ(W
zyXz=<hSM^I9Cvm-T(J<y(m4~dj^BG1k|_z87|L%|RaHm_u_XwD+U4w{l5egG1)UL-
zHPSX>GK+x~5wRHsX9G?fFW{5dm;kg7@@Ck1XMaCf5*WyZ(%A=PG?NBKXP{|985?7l
zd`piqW^+SyIqE7SZe$-YXbY$S;AcQAVl=O(aHAF`HJgW=NVajLI?vlUW}a2os)~GM
zCK2orppu~1l7^>$1I8jPuMU)|#_eR?J$FXG<+15Ix_`Zg1idE_5=iO0Y!ighE&9Sd
z=m@7IBT%AY+<+nq{yHeNR%H=K|J{>3<u*Jc4v{~y5(>B$wVwRed0j}cZZh(wYQ%p)
zV_pqaeNU79iVQay><{xs9WtEXlfT?UO4fR1>y(15tdy={B%g3d&FO*lCaU^oy)=>g
z?vopoJe{J0`*ciu2f0ptwKIrL_W8dci^{21BCzT39WhwiS*q5);pvXZ#wF8{k!vfa
zbB1^G+^w0p4X^?~@}=h&LaEN{uJ3W${9>PhICJ->2vO|kLE=IE-0F^XVWVFvmwoE!
zAZXwPSsPjq1%2719?Mh_bU6pR>mC_#=hW57D1|<)vXbGEI5lJkI3CRp*k`_F#V$nK
z2gwpJ<}h^vDMr}KKitz_MG>VL@O(Iz8=fh-)hy0epIy((CS~#al2^qpvKzN<NfBZg
zId+x42?z<N#9~bJ6x~{c9!((_gpVuki&K*5w>DPX8MmAVPQTCsBf2FSw062KkbQo>
zFZZ0fZA$wyHEKQcWNJcv`YGa;1Cz}?on7pQSe|~__YbvV!J3fVZ%-M^a9H=*>|hv6
z{YKSnr9)4mb9Id8z6j1XhH}%GBd*o(_Z1~c$VzWy;MTg4LM{CL%c#w`G1_mgbp=Xc
zK@1b1m&wpW&#n+!2LZv;MzCD;yb_BtdDEGrO4DnF-+OQO+>{v=X(A(Cs65xdeU7Kx
zJ7YJAwl;<^Uj4^Y0tGW(K)k_)1_*f=p&{tD;9$DQ_B>}m8IpZWv~_fJ#NYBk+Q-A)
zqTu;koPnZ?WJ=akeS@9Hk|8f5qs>CF>Q*$y9&z6yZDVXbY61waFkKGq75Hsc{SpgV
z+!%)1p>rPwqe{$k<WidOXnS@$vNMT3;bM43J5Hy^=+ztAA)=w(W<ZhWxVkgkCU~rT
zczlK%CAC^*SPr_X^OAyq=nOK;zOEdQ(`Y)0A127@Qo2-A-<gKKGbk!$C9$0liq*T>
z)ZvxEf2NCmKN@}FKUXtV`NKM~R7&M}&$?1`k=&$v+qg|4{!=pZV%PZew$`(AKYcIk
zZBp=w*RpVf`-vT8jB63Y7R@ScJJlmU))j|R!=<Qi%qa!TyMj1<-?B+0(!`+T)TI4)
zYMFl@Sywl_xqs-~u=$bfbiUJFALk2y&gB`3N=Rg%Q14f3^pZ27{_NIMtC)^~<YPEm
z?X#f&tm-8mcGh2GI^5a_&Zv#?CQ$Sf+g6qN5!cWI=!r*~Y>x%#YO>J!(^Ed!$TnxP
zMqIku`;*%q=hkx_ytxn1utwstQ3W0SmMQiMQ8&AA<R*egUJg`6RcRX48_*7zM$x5y
zxpj=523wBxp%_?z3?Uis9apk1T#J9ckW+Qr7;IA6JoQ1YTTgC6l?|C5FQEHwe0Q_G
zzV~t}1@O<C9-yT@$sNF7g5QJPm>esLv93t~tRPaKRTFs<K=`yzqsdLPQe;aoMT?!a
z7~hX>0c+EFhEBFDZ&ASD!p_$Ep!v2A|6x`!yW*jK^Wet@-@ZL!nYKduw?z7X!|cHm
zwY??ktDjUtuZMwY%!s69sE3|uxlXRWaTp7={aoVXFLSEev*)QTv&KIYU(T+Qy8LKy
zRGSRAuredN)RJEu(uD`b!;5|t@<g41=yNj@2IwLmHJr{Rle^{j?lV-2PE0)4#%fce
z_T)+CPKmo4Y_!dM%bVXHt6wo!PZZAw*;!*mYA8K`DgVJJpK0SwPKjv85x3|A_nQwl
z#`IFN-<`DBXHN=W0Q=BOUa`4fXYy0z5)N^`D*PQva3gNnRq|$WVjF9Ovo<QFN{*#{
zkqkjMgRWIQv$M0ayYea<!M(gbp|G%vXrR;~PGo(>d2<;a3ue}U3!-+5o_%!?@hh|0
z;Qqk3tJp=YYKMm0;Sl*J4mJTZ^xg|**$-S2B+fi)cTs!6(%$#M?dHz*+h^7;#I*<!
zEUy%dz?1^O%}O~y6T^>nnkIks`Fo7J%>GE3?;2f}q^|C6axQ=p73_35%<EtLKJ;q3
z<kqgU3AdaXZhkS+vv;JPQGJpnoa=MXGCGY<_~W%MdzHrGq&~N0Ayci=3}PX{RFAOn
z&eDgI?UB8%`|4Pvw$}95DPip#fzi~u(Rv=%w>@3iMg`Y$yyMDK>fk&b5->Duyp-6$
z&a`WXvDcW|z)Vzs-QbR2V2l~pl}2jlGA~`f^ekdHH;s-=zS5?FaKSV8=EPY%piCNr
zOV)cO)V)N36j<Z+wDD@PLjWetph$urS4`gb;%FRNHvoXlL}~%6w$W4tCi_Y($Zav#
z6*S9MMt=X`X>7y|b^|=|Ua`kdo>bO^3y!uqq~HQ+95{efr{6+ON}OHiENGJ|AIV5!
z7QN6XO4BEt^pgYb(dn=HKPfv@6Z)j#1HWczS2*fFx`TO0YS|8foP!(Dz9fr~h}x``
ziZesA46wdpRiVrJ(%dtoxa$_XQpe^9lq@MtFGk+Y<#mLt)U=h!!kySwW!chd9Mz`a
z{d?Mc%@335?W!&HeJ4AA&3$I`Q<0T)6_=_p0ch|1`3<>G#rWrxck$w<lr*0;U!3y>
z9$ssSWQ*_ieh+wf8lYsR8IfLw+|lGW-r~>dYZ4Eh9kS-PGUzn#)gsHY3mTt@P0w`N
z^|7KGJ2Z~9)iifaBup(Us)*2V3ex`&I)R25cO9@W>SVldBKS4EQ1+uUUTNYXA}F9R
zheVKtyRuBXZOO0)9va#NEE38ai&mN=J}6@jd>Bm+!aR+*M^|keq&>`V1h_sssYRwo
zKRUyB8zS>JCq~9M>KdOaGRJu`OeyYgSSuvUKzVB*2Q*rgWU=2~^BP<bWkLmAxwVz*
zMcuj;9{kBR&jB_myHS@?g_UGmsuQs6wm&LN7k_2dfkN{fgun$U2#|gHmQR<!MG$zr
z4sx6s)4uKbU3{?N)Ui*TRnwY#A6eE8f21C$2b*;-P$@G30?s~eyI-3lqYvEIrP2e_
z9^;qejhwvE)*pvB!aw(#{CqrKwds7!3%;$NkVNbvk`w<8O-+~+u(EehQMjYXw|(dI
z4)WNH?#Fj-^!uS#OdcDk=p5QyJ2>3o_SLSaN!vsDG)rE}A2w*f?n%~HY_lwxprp)P
z|FeE~*@dUC<LPV!SS{$EEetTDLLZjRjH=qrUS+nSULH>&SUwpRq=GjmPHV)8sVS~o
ziC}y<rbuy6o<zmR(Pzl5J^MknUf!@#p7oeZAcx3WbQc9OD4L=9g!0qR;3;Uzr&fJ0
z;O4+P-kk!iPLHwpA}eaw-n2{UT?J{!0SuSQ@=jq^e9h98n+(!4DWbqzJ}bTR%|9%}
zx*O2(fdi<V<D;~j+~^rS(9!`lFr8Oi)?SPeTo0jokla8*Wn&e*%$LqSZQeWkhq`^l
zb$EX8t$rJ$aa3^ql^;9)h39j_TrNIt*k%{j&o>Z2{<Njzr=D5Tq4V-(#TzSL@%MT5
z1zS~}#ty|ZeK!5|TPpv^y=YUauM4iX2u(~(Z7!qMImQ-xuOGu2Auj^GXNieyk)0V*
zwrVo!+&QGA!M}VXH6gJ<dPXd-{E&Chklf$ZdRQd~Z&$+OvUOl|*}2f_?zj44=y#X@
z>K0%S^h-`pM-%~x7#ZnQv<@@3z!VYa4NHM_2+6W&uzS47a*S8~a#mrTjao{}w&9J;
z0neK54Stdi5kW-;hql7D9{Dj`6?`qe5D-4_cD<IeHNuf~{s3<l@EZ;G`@t7Ro!Gle
zZ-@qme|WB34NWc@dN8xwGO&)}l(buh97iX8v;g`gQ2&!EEec>3wT!BOqyj_bjq87`
zMPzA~!9-rdtbaKJv3Ql<RoTkN-M6f?raq-+q1vBa;Wstj^;}3boI`2JoPX&a5m`3n
zY%#GeCuuqjpPT&HsdN*`Gi)!(bLr6#hgJnhj94n4s^w}L&pj8MJ!fGMmN2BZ-(Dqr
z#jeog^CF7*U0X=gj+vpI9c~6<66$_W<}Y_V*J{&aYM#^`GIt>OD(!s+2FknN&%C!$
zZ>0k&?(a*_y_iNeR~J$3ld5=nAB1n3QA<liQDVa_QQBZ|rX*~t-~@wu>Qh%2)|}nJ
z8R6~7$K7~<J8$;<$_?wBgFv&`)U74CAv6cni_aJl34+_lTC^E8h)S4NqugZCrij9e
z3Dn1LSJ-HsRi;UqJBuU-rv_yjZZ&Evb^*6I62bgezk*RoToc)g>#w*O-Qj@3MUEA`
z;`<CaRz_8pRie_?g7g{y@le0TgN&pE02BLM2poPDiDJ-U(XD5*VG&?uu+Y-c3>WB<
zFp4Pte0wn<{pldH(Z(I41`@cbOm4w4vAUhlPv;*7j<4p`b6oy8gdj8J*Brmg`i&)7
zSLe;w4TFhiax1dPal0AvAN7jxgC7af3^5Fv3@oZEs4S|wRFksg3_g}B98&*ExxzOo
zuYO62OIL75wXNdfjZP7zNuGLKLEXaC=5qPWCJEQLEp$OqW@Kb*UxSaj3g(@kHqO=q
z`BLq5=>KCa_&0ZqC2V8CWGPTdS6Qb%vYCu`Y=Ns_l>hUrXE!7rzp+Ca1s)rG3(%Fn
z>Zlh0C&18wzv#cnufguH%-b_ha)n-t7JM9X4(^6|6S@@mA^_dzjruwkl%4J5P9)LT
zOThW`&*&TffawJ`>;^E<Rhp=*DUTEZ1)>EdhC89#3$!n+6QhL-suP74q~gdc1v0xS
zX`zu@piX)rXTkN+OtZh&eQ_|g^{Pz0YK~=#0tm_W;?^`@{T5lBPZD8l1E(vS1?|jw
zHC1(uF-2HDd<>h1@O&-c5Xwno$3EERe=d;c$B?g3knRRQzl5ef)m!O!3S0V*%B?x7
zunsh5P^yR9<fhE-*6&>}hv%jRzp_2?cl(|p8bFQ;*1gziJE-Z`dN9{K=H5%-H!7Kj
zS*osQN;`S*M4GYZvL}dG^gE82i`v-?4FtJ*z%NCsNoAi;O>n}HG<)^=TZhDGkmM0u
z%Ky8rMU{JJSmNv)<(#_(U0QMGU+jf{dnZPs`%+#u%`RujuzLT4R$%T&A=j;<R%l64
zm)U^W&OZz>6|t4IYxwJzFN_O|$}T-Wy6FJXRPr6&-CSQl=v-#haIFyYz&n1Q_#J<1
z&{F^j7A8RM{0z&9Y!j=aO2Sn60xJR>=cuXxt|&gAi2FqsB|`w+q9F|1IXdo$@iBx!
zpwdP;2V0eAwT#XhY4)Bh>gM#G9p++Ur0-rUpMULl)>2Nxtc3_VhKBHP0GMOMO*tC$
z_Q4F_`{h~(4jhiDP8onm3dpe;4{&bi4%B>qSE<}Y|AnFqFNSBt;ccZiE%FGkK+3KC
z|7=?bzi=<k&PG;x?Vgm|BHtaVIp8d6R}d9^=4$8BK|<$B*Hy9~3DTv?=*KB#C_`3h
zy}a`BHHW4s{X;cH`>4<EBsSO<mb3|c(T<mIzxZ}S>SOWO`))2oGvbhPdi~}z**|-}
z{_liD$6t;Mqgp#PR()#jXMSGD$a|T%O&eeF&B&;eN;=ebq(r9qu!D<T=;aW?L5~TJ
zIU}_wtaBs9S($S%MV#*@7r;Wrogdx^7P4n9qG53}LdOoJI398|1CrE_>_a%1D%!K-
zP@F)=YSD`If6%8|lntF#X4fA5uF~hCJHL(C`FHXB?=1Sn6&nAAfh0fy#Po}TOHQsX
zG+oJwtQU0aWIXj_>}0o?tebRyzGnmo8}#uC`kl+e#H6r?jgYvs{*(@^U3ac&vFf=p
zEGjnQhLIFj6gQ1yhZ6E2F8`Ab;6KQ2=QDuG0YE?@!5qLfuT77Gd_bJfDr-M#ueSd4
z&Kj&L&etpkP`O~l1@~Q|&Eu0gf+H1w3e{EJw{i;ax^N?orpaD&In4B*#%7Hzi=IIk
z7vteCMy9J6WJV;V3gtfOlj2IU4{elO+`2)krX><bR!xb@Pfq32H3_?Uw+ssxUeaS?
z;p4L^heI~k)ve`L@>lZoPJ#?(>xsSJrzhuSp3V<o`r}A_kGVl-)`&)o`{Rmjq>{6O
zd^*Nztb@-#q3}PUpb0f!3q5*&t+XQeN&c8HDO6zG8+k$r^Yy1sX;R_fo57>d?Glos
z$Rw$7;*@leVY0JlJ=Bv7z;MDC$fX^Jd+*)%Y%x1j7vwd%%N6yxz>R^30X;@PFCaGb
zVjAmfw`oip+lUA~*=zekh-*Le&@bfD43ZBmjU*I5oHcD*t_0V|XBa69Mo5hwy8)7A
zOPrWg;jyg)JDw{aq=)Z4Nm4|Ny9dV_faNIsAqzvP*jaKlNquGI4)?8_11c=bs@3NQ
zU{*SN)&KnWjIz*Y^M<WjRnw=74%$fnhi1OAZ`Z7yy@m06Y>SDBio08e8p28PN73m~
z${3s9T)XpEvWAkEfpKR{>a3{b-gc44tDTvteTm6Fa<|F#*F4$xwL~dlf>zC;w_vcY
zr{_^D<di?69rDGt&d5A9d-LU2iNZD(sr5I0<?TpMoIR}Ap)q}_|IYn*Z<)9cFA+!r
z<k$E9Mld6kVdtKpfF`6pO?{$LTuc$(iq-5WYG>3C%l;B$NFyduJDa|yu1g?3fO@gy
zz-r|TInUSbmU#N;_3eR7slsuoBfKH0giZx$CtZQ`8F`0bY(Ug}-BDJ_ru)6@_=kN8
z5Z;AW+YVU;U+;wqNi5G&_&!%)hNFe?FV%2~pGri)nsNvgTD3&wlvDi6^o<(q=vh)F
zyK8=7nwyCc*Bt`%tOR``8!o$ioo8#pD<o0?3jzWHfcuG;BsW|mrL>iH&(&=&e<nCF
z$2tFNt_kX+^BhZ#lEUQOmcD-D$UFA5nA}qONg?n9WAVGg4>rxHy|{C$>{;QhGAXHz
zhO}<GJo4tp9;E5zj2GB<9_e@f>Z)r59LCMfQ-8mjgOcYRs@9pI4W{-vdIC-z4x1ze
zcN-X(+V9?vhCcBwzr&_-{EyD8U)*OZjIK0C7MknzPNeA@+WQzFw_JRB6z=i(_#anP
z)U;arU&qq~APk`y02!R+>XdOg1Y!hWnTv3Y7ph^D0-*RGAAWa20oL8q^}i)UzR?!r
zZmbp|asCHoqyvqDke`0}^5x{^P2i6Qlm)+kaJNK514aNP+-Xt?^e|Y)iEV*xSCfux
zj-!!F-bZX@-7Ukzk{sW_sc*Joo7+xED7206smAWc4WXb^=t}oD2x<Y!&sW39ko-d}
zrHN5hrva=1MU-Csc}lnw4A|r2AKbm!nGn1{+o6j^DN3lrF0VP>;Og6VXrwY_e$`T&
za_nVeHZ-H2W!u}sTN<kO51Py_#W6D2E>BWoV*hNmzmv5(dQ?Wr+9iH$+ditBlIK<p
zPv{&!>X~+4>r{(NoShoVZ;mwCcmMz76X?FUqU%amdPldpJrMw{Z_;ht)5H8EX(;c2
zf&DBeX(7MUlRJu<5n$?>#1d^Z_(IIN2VVonN|76N;6enB7bjTaQ>UiRf?feBLiad<
zO)qOp+F3vR>24#_VtS`(Av-><1s@x72|A3!IR~Bi&1E%IeTOVoTnMd`F&f0&Bmxv5
zvjq|C0*-u|?`o!Tq1Wc5cPFbSmlv;U>#B$=J-2*jMFeY1>MP?~Z10K}H(Ym!dkN$L
zl(c(O_t~d+(bvgicr?otz$S1l)EXI%Ws#)JmgjK8IiVn82DC2V@xBieGvFg_^LODl
z(Es3u!1MPFeDhM8`q!x)i!TjL@##A>I^aL%oq3A-g!#q6%NK<VR(0$Y|EJ9fx^`pq
z#Fx7sdd6Xwzn*e8*se{x_qAJ$_DxCr$g_JSsu38LHu-eUsL9e~>W-b$H#O}D5Z?~!
zm;VE}DDIp#C6B$Am!L#WX2rdVKKm)qSNof<(n2p&Nr*ZZxFNn(gufrkV~lsecb<t6
zWGkWnS&DYDc5!=;bc+Jk1eDw)3<XaXc{FcWf9gofL$0EDyzAQaFsTFhDAhUujTS%d
z69a)dL9>X@Va*j@cm0N{hd_iaWWl!3*?=#$WVph@lN|Ln%AP(?3#($68EKH%58R8)
zbHHFi^<^v_X&86Gi$oj@)Gh<cbc1oAFQ9?cBkX`3B$Z}WU9S8Qd3f1+Oy|an`*ba1
zI1yhg-ZW4e2K|IpUQNpK{4;_oy%7%x(TBylg|Ti3eHO|}lwWu@pc$V&!BqTz0Djjo
zI4_$~wft^vXp-$Y4~>Og$w^i>25!|1N!O*_r?{fy?ny8DXL7>rK<2a1n5c^R@h-o;
z`@<BuqRq0&pN_|-)=Zxl3H{k)fn06haA{Y3Mniv3&xF#}`6Gtngi`-e`%@>EtQhh$
z-;GC<JZJs^_x&=92jfrhZI5PqdOvE=WN%CWndw%l+0;=F?}tG_l>*to`Ur4;>0$&i
zwJjQ+X!GkOL8&ZP3Cn-0NuuBr!6m|R5n_8+zi}elS9f&qqi}|7Xt!(R8bW!(&_ps%
zm3&d`0yAYoBSa5~h%bnb%mEihJ$&5mgvvcX7$YcH_q{X>JwC|Jg&h^dJO&`MGr$?>
z^x5tM@9ue<HHq4t+a!V#qv@T(bqQK|w5h&@I5g&e_d@}BhTaT_sZ8kb`T`m2j=n29
z)JA*!al(Y|#)>PZZLaQDyKR0gO-5X%&|!CIb-;OB2Y$J4YcPPLk*G5GXCG7ySRz0K
zar(SeRo?ItjayESt`S2Oi%DzKq_|2hgegC{F$^6OG4fKy*!EN3Isr!_rn;i>P^Gpm
z6(nZ*gyX$iWyM>yn#Lu>H~-i4*A--O@go{ayigOvW#M0$e&or?_My!?=%{|(_kB>7
z6<e=NHa6(YOukP${<6<%r=?icx`fR)aCa~`1v)GRtQdZUBpM)eShq_U*M&<aD@@FP
z^@=4RD+!SW-Kqu<@E4+OR7^N+8C$V`KR2yvlLXTLSRlPufm9xoT9A}NSpzZ@zovJg
z;Bi8qi1TPiutdjafT$+`18y3|kccj{xA#h@GoJ3?bUEu5UGL%)Iis}<vC(D-?Ky1S
zYat#O_#HpnPnWM@97aD<cZkVN21gGmtHMgOu7sm-_1<^VJfXlt=^5@Kn|G74$FCV*
z>|VsPH<uWcP9J)yDyh_D1<kWCUW}+WGSO;{JOym7QtsP2^$q2c3WgnlTFi@9ZP?1L
zrTOhdWRy={mPL{5q3gn7&t!qCSw1aQpy~Q+@*3{<<BYb3$C=7y?w!MH?tglM=0x2u
z0gAsw@lSNY*gqmAXIhVzUBRhJE4yC{1#zZ?bwJ{*Z7x#)*}h(f)Y*eW@!Mc&$8=S!
z<AGKlK1;aioO(pe<Y6G}FMflJsg{zr($L3Ib17iTmAVW7%RxOYj8PD*$<*iu0)+v|
z4-^JPXl+56I46NBL%O1dXh|pbQKcf80d&CD#mMNU(b>_1@x<S#&y_on3The6sD<oT
z<Rk<gx~7aD`cC{Bqt<ns)Z(8Pp{;l!=d-1obtXP%{Bf0@#WITM=tH-dG%OBB0qz)&
z)-e2=#vL#A8sxk3t~#O<W>h_Af!!F;Ax-*XXda@4ksjBW;-*GSU5s%&!Xw1o&3Ux-
zIH+lbb)$_4${Z*m=-S-_2Mj+o*MCN*`Br~l;4U+2Rq&3AOoL33{~&Hb8$>?8+2Z?Z
zcl4;SslCmU`pkam$A?MNGwSn09w7zHTBH2kMM)RZC0O$N^?#KfCeqO@sd9&;)9X{_
ze?o^@-2*!ZD*|X^{TSgjw=j%^eMWyX(zJ`xbi}L0-s1H-b!zTqJ7R{6+MKlg;ePk@
z!!NQb^P-<Ke~JPbV&KD|$A0C|ws@>*b+{rKf-Ayu06853u)jP#L`p^Gm-{?bbAIsA
z<caq@B4-r0+|0SwT3mvS&^<ma^xg96xb1r*4iBHT6;k3|D9nW0CVdysXMA$#QY~Ah
z7<s}6+=|T6bYLPfZ}ZKx2EQipZsnsj90v|~#>?wUIc)F;TW#X9yxSB4iY6A^6C>_U
zo9anlI_HH6FXXPeX*8P5i}W&Eii{SHz`~~WIJPYT*Z{8EacaDdFw84_zc%6nn~CvA
z=&6<d%2U7J>U;7wM+G0#+#CB36|kmT>%rNcM*(N71HUc0angED;Wy^=+*Y$~+dNI0
zjJ(CSWya@>#D0$UsW;=p-F5Y@0y+iz3>&}O4jjcv5MBTHo~%oCbpvMv;En!YBtXNz
z<E-aCd0@wP<hx`ogQdw>s9k9{q+!((xi%y1vx<S3Wpx$Hhuzy0uT314gCl8Q3;xB<
zlzB=VweY$@?Ooj30Kx#L1lAxqEO?;dsvx1rnkL=bk3}pRiA%MO5c*M6FASs`9uQb)
zgY?85HaI|gq`j3yd3>|iQviJsm>VK-VG9+;D$kr*{GVnW155*@Pz6$~8Ix}`7?bSG
zuE^9MxooGi?x`#`3!!w5YD;rm#~UT39ap;^1Zb2G-Ts70;~At)Ld-t%tC<+l4bkUg
zWOzbNw$mqFe*sQFB~H{W;eA?kH4Xq$_)_7gWmutBOWC~Q6?{>(H5f#WYNyJ(?fO<*
z@DGc;JCVIN?xh_~e~VeilpiCrE(0O#t87*OJz5<~<&~0B2d$~$*>3P`yU)XF<C@Z2
zv)wnyjSKwn<5!pezc~UGGk&SSFXuy<gNm)hD6mfNi%LygKFYmWF>ssi<@KyfRi(mq
zMiKQHi(a<vW+vD!fct{1k`Rn7$dNGCf<r*l*rG7aat@zBz=tJ3Ud(uSJs}bXIFA>!
zNvKtQ)cPnqTqR@i@f(>Jq0n-Sw@8+W)=}Mt?E`-Y$-w#3Z7<X|>)y=xDR*DYsPL_Y
z_(WDV^|IajEm4mJx>L5omV3?8l*22OA|oBUgvNqINvcMkDET453=#3WW&(L`HA?_N
z=S}<geYXc7!CqwrC>4qe0M33;&%zsx1H?bVBJi<dmz2^h-b&ufJag~-XD2gH?GZtc
zWw+~l!>*M$;8#i<!vb-+yXV5=l+v}go+wFOcxLgAc5FL;;CAadR*z=S^51X5H;*0n
zh@i-`?<7$-51uW5bB9;u8H<tGvzInvELdva{%11%QgFdu%(1>DW{f;JD3Otkr>(kY
z=2Frp{gaAvlTOcVjf}Kzh<>mo;DKk(%{}^tH6^um&W%lfyh<;#GbFc^AExlGFTYO*
zVsC-6;RcDn&v6I3$`yFt>gD070mURYF{*>G1<5Nt?L#4%JojzKs+2=9cAVKnyPp&U
z_@7NBjF=Gv{|Oko&H1Ij8VYpC?)g?t>kfH*pU)2j2Ghwc5Vxa$M$M<IZ}TU!5c!!>
zw3$|UmF|`|GwMtN%p`=BQ`BC(UwlMeuk#q`PJ9NSI!aU?3o64kOaJ5Q1hpsKhZ`}i
z<wSp>R?ph^MwzD4j10p!o;N>h%QtLwr|U(y&?ArSM8b0`uHuS@H#f{F{&qKTv*%ld
znW?jueSIpW$+}IGv%OgfH`Xv-kYu>GPsk{)mQ~s7e?ZPkmpn5PY)aQt3Z9i)F+byc
zwxUVd@aNAz+GID%*rflQ>xZRXF{v_Nm3De2+-eZDJ9w30Q0gu`{e50@ck#-*uQP}y
z@(UW9--&UdA8=}z4D(k~(od?H{Rb6G+=rs9SAi^l+K-@?^;S_;xP6lc#<zN%h}e5m
z^tl!dpDdok|H%$4bZQ_7?36XwOD>Xj$-FWG-kaJ$i5!?v1{z@KufY0Mo1e)0`7FSN
zTcK@jy8HsiRBg)$I=)?TcyO6&vJ-I9gZJK%dQ4*)R~+5+{18PH?mo8pK`2)jhWt@L
zqg10ioWOH6NGE*}eGw4vo9@@sKi3xf#XaM<KIAOOTq*ac?qih=gtl~xRir3**p$km
z`-E6FkWEpOS)*t9ISVXCSP(B4yUgv>4GEvm?Kt%pJx9w!4|mC0bs1CBm$d8=1w@)~
z*>I7-ioQAVP;o{2Kq-U3Fe55IQPSD)C`z{dm-Y6H{)+v{rBcCAGkZWXsvq}(Mfrac
z2TFD(#$ivg)(M>mpV#ENpmgE?q3lh-YTVnl@eqoTii}AbNu`oj8bry`q>0ibP1;f_
z%_$N}gj!9SNCQbDX;P_3D+-|!MT4YBb7}g{`(CzfKhOJq|L=bs>#%zswq5uA8?NiT
z&S9?EUb;F{`SVDnqNmcG!C|!i;z#ZNIsR%CPveVjCi{JnCavn_ykFtrVq_Nvsd#xI
z+Eydae*m|TinR=nFb|AU{isE982~xMA}LUi>wOdX?0fTB$m%|%_T;>)hd}nU`I6rJ
zbOqUjp^o8#X`OcN9|f&@9_*N1_A_{Zp~UPNHshZSniDQxlH=r$<XBMgFt{gN&|tW5
zT)BWjA!m@sL8J*fNWRidqK{goH8CLrT8(Sua=VhH^8k8SXfphn;2IPj`_qhXz!-%A
zr;pKmsAM6M!5x9osgiBZVnv)>5!Oz(UyYuON+~edC&&{dWL5Pz08~E-Hl#vfQQ(t&
z!EB{J?~mFODRvoCN1J8z4U&)Q^ED8|*|j8t*;i4T?M#ArW5)zxqUh6FH8nN$?#pcG
zx7>Rzsx-c`5u%%pm2Tq8*Ijjn?%c(=v=t;!I7?qn+SyyhefDkB9y$~6PLG`XUwAI>
znO~6f4;0PUWzYJ0nW0C<MtLgTV@Q7v)B1STPFW|X@x%H0REk4N%}YIh$+w$_HI*r1
zBEc^$J@=bCJ3}8U<}yM^{!;g9D`tSvL@9c(9<c<ek%UQuKWKz1%Terg3Yyz5u#ml*
zzDdQS?|$EKb+vfc<k9_0XWE*T_B`0<UttzIU%6~uMm95{`U8yvm@cYiQfCTg4U^%z
z_!d`Fa0U>hifsbPK8kHIBTs@#v?mP|PF~5=K}Mnqp$`LWlnxjy!v(+ilrsAd=FXFq
z`Vp1)kFL!T9lW=n8VbfO*dXLwuN?u}q)-3_Hvnz}1I2?-V_-Ndo}l~4tX%hg`v!%|
zOog^l6J_?;tmMjUi}DLZ@(|o^2dKa_?uF%R;bF^L#!27+x4Q-_LEKBE{|&=TkJKt7
zBoNZ57vJ0fU4iV8fOdDqIVNF(!R7x{ynpFjps8_7n#LIu6S!^ouG=xHKG9z{{jT-N
zOn1-Bwt1f>Uu#mqeDq+`PdU1Zk8&dEAHpZQEz_U@IwB7ZZNQ<c;srYMZxvs7<KkpO
zOPZWYn2E`cEfx{3O14_mvl=vT=OKQP4bOMi8SYd3|C$}d0?|91p<wlafBs;_)Lkk)
z&J0-H=Lfz_RNhba=1~$@w`g7cZoLIGgewSnstI-IcOFXfvG8q!9M=3)?kB>J0RwLC
zbVIo$yKDUWiN0t)3g|C|Az(Rmf~B4ft|sY#4~-uaO08#IMWsiqVGC=r{c^TSW&^NG
z7<vxMqf#Lq2IT8@fuH~w8^(hmfnpFOy$Vwx&`VHlvokANZ7~*$Q+rU6u_cNHX~wvC
zMBdsz+`N3@-bD{34S&kT0G!3Rf^1?#lk{18n4K21RZ&!RP^K!?Wz}M6Pp}I!dIqNA
zE=l6O6?}XguSR^|m6WpGRnZ!PnDip#06K2`3WOIu9Fey;tha5r_diD9K}(rfwC8lZ
zGL~*r_;Q>62U4`Va;NLu&lwxb-E6~8KJ>WO7V6B*TR1*oTYdo5p=#Vb@x5U4%;QeH
zyY!v>7SHBN=FSfSoi0~FAdH%ST3R!@o~YVw#A@O-w0ym({6c+Zp399TbL(EkEL-c4
zpl<B4-+4f;c}W)ty6b9eocFcd_|oGtmFqaY)+h6YMk~{7rm-uP5P8;AL+^<n8NUKM
zAa8^h!5jQ}waW8e?*U*<1W^)}T|nc&Qoucj56W<#R<&%y00OGn<oc5zcb;<be9rKb
ztuTpbU&~~qrbKF`A`MUsq}1D`Sp$y}+bZda=*E71@rV;E37JgvMH2`D(yTFFFED|q
ziiGb_0+saX{XDRsZRfThQL*X53$#DjDA^w7aq_fR--SRH5Y=?*$!Q0M<8rTbgqYga
zSakRQ&2Iqt<j6MrQd_%enA${G<1)jVYH>RM{UuRAxnRNvAd4Y$+x`U*M&CFH^H&1c
z#{PrC8?<!2eDuK+$9CDD4_QFMMKNzXQ`T3G^Bq56;IW|Qu{t&SZ1=zpQ~4B;+#K$S
z>w|6lonMlpibaG<8^639^a0nTzS&CjUz6~s|B@j5koX@Y2vX|Wgj#+7iRSwqpHzC2
zItRky*_c|z{VJ8F$2<$?u^$EU2HHLfK7g_eX91iuxYG#32U70lWv9Zf;(+B0WeoQ$
z9B!y7@c}6pLsG0(f;fBe{*AZYbD?Nsb`r+n)h=sN2&N}sa+5>89v-JGR-mi^1PZp=
z5|`3V7ZTHxvz`cS5xfLg0Z;hW9F3h~Bm*0N<5BVYXxVBGTF?mz9apJgx#{gFB+-m}
zs4=yUHgss3HWVbSNl4(WT^5mLBD&nuc9l^CIC)BrkM(He><^k9S=V|(mcX|Z;1$3K
zi50NDQM|^qdMhZtzLy24uWrY9j}W}Xa44=*L>AIgH0)1*ANt2dpmKbD)(mU;O-H)?
z>*sLDldf{P<~%07`B8n8XWGEy<6A-pS|`R!Emqs1;flGQK5O62VU}<CH+ZyleLaP}
z)?4#zQlOlz^!l%>%$P#aiB50MIuO0~LC;cP=>lTPTE1Hjq~tg-vy{I(JqeDeXIR**
zvhb~wnWe*LPF&(O`(=p`^BizFT-uNSGJk-yZzlrs&?8sqqq~V+pu^?4F)5=5&k&Pr
z;sY*wLk}^f16xtVRHKEs*8+D{B<V*);qBn0gm(i#^2`QfPOf!4JjAJ+Q0Lh;;G^hw
z;Gj*Ha>U--X&d3ADqd+|&>0#<$_`M-p!mSH!ms$66uc*>TQTLU>`t=Pldzu66ZYds
z0DQ!@78n}5*-C}+G?2LlS97+xP#iglKtWNHlYjBq$T3wN!1TbOtc*U}4f^u}GScbt
zAl}+2szZhi2`+9>Vu79kIq}WxDiqVG?^sI>8uD>Eck}iF51%S^J38~3?*o6nxa_<+
ztmV0fidUwuj-)+E7E#Z(3sO@aZyTvq_o~2la#Y{<mg$I&hW?Y0MRi}j^!6#nW8T^Y
z9-SQ>5Rz7I6vPS5r2JZWyKY^=F|UE!FnqGs!l9@)s0<Wv&^kLk$yzFCZR?FAIR(Ch
z%}P5=-Cu1xXNbmAOa+?;!1gct`@~6n4K@a|@)fzxOYqn-=TiMtzKoNGL^RC%2-O;a
z%!1kuhzX)XE>(`Z#7nwa{wmmrZ;SW(yJ=vSsv<%GB?bf{Z9ia@gJ}oG?BvWHa<mHY
zJQA0yf3R-dbRj3=_2XAX_RXKgD`15J*H$87g7E!Pv2GppOoLm#&&vMN)5%kz645+)
z-wjh=<OkL=E3^$X=9yv^)hv^ccWGqSl_U+5#py1$N!O!wcpLD9Cq0SawMY7?AZi1r
z1qQsXpHX%2IzNX$!Xt6}6Ld5=j}Dbpn6<^72pq!x`c(VlN2}%p#!p)Nu1U@^xqjPy
z-|tbmtaiX=^ScFioRmF{LnXmRIREt=M7|H1#CGcX?$rqq5-GJw${p$$<Km`u8jcTh
zbzbZwZ6U(<Lr}8gTFJ9ow<Eic$sY0=z@&UMa>%X7{qE*Bx{LE2tzla^GaQ6sf7v&o
z4}GsAnD3K<?80eIH{RCKzUTT{EjP<WJ3Cm)Mi*#Ko_W{0r(wn6P4*r+3_lW-iuhpl
z!8n5KiqA=~b*rLN6T%56b`W#y23Nvcg@C>ZRQ<`elGDW3c#nkGlc!geiMpS-;<p^y
znFIP&lVO7^D3lBv3Opew@;ST-LgIPtR6TS@Z~mJ5v0pyVP?Ryaf-b>-!)glnaPhD3
zQlvGF!y-4=w3%}gu|H?;`tx7itTAF!zBX(nvC{;^#&VMk2^)jjceYm-&C+6zWa%-P
zBQLlyCZ)R6cC7_!mQlC#_Dem51SQ~?Ldea~bKb;7vJT0%pu)3(4U@G7SFGM?-?Oar
z6Y=zU01J&<#|<X6ryXa`!)-;aG55&3B^gV-IJ5u2q}CaCT8-R2)ph*C17_=GJ&y|e
zrL6@;76K8`82hTGR1(GcQFRZ!QA{{@QN~9dtq{i(Yn$Cq<=>jje_3aU8M=KrN5K`s
zWnv*^rN?^V1j74^tt;0t*|6)jTj_^sf4yqZTI13~)^>JSsY)77`yPn4YiQ`l|8jZK
z1>2&eBa8Z%1ArdN!#hKrkAfdxkiw0S$A^_gz|e}t4+oV-WEz9Tr8v<m^-H}47z0Y5
zhy+J&DuiDCydl`6^5Y5B^pN&L;D^b@Ia3QjdjM=NvEulaXeu2|0C#hl%U*8vHeG}8
zh4fy?y%}+QU}8l^MrVmqIdFcF?jD61Vb^{54HIRW=Oqd|JXBo&>AU2i=BuXXbLwJy
z3Xf7D3chRlASZ{}iB)3hN)HVV+bRW_SC`=Q+O@gC6)*>8)4)FZUQ3aq(s55GJ;RiC
zAfe1o8x4gj0R$KrjBj7|w)))Mk{83KGCrk~OJ{~+{8rDz%+&dLgN3)d%<`I~X`0Ep
zq1)R3jN!Yg?l?P7Rz2HxX@1WW(>G(%6}d_2Oe{Y{h0U$54K%6Kv$;C&+R<OujP>Rx
zm)Ips6&R`9E6L7fr!5hrf0q74XdMUms=~o8wjGr*$tCzIX20*U2AvM8LkWW;0yWha
z>l8w1xiIJ!QwNI`MFk-YLTPcn;(o`fRey!yHX3}VZX@UpwkiB=fFWYg*9Kb(G*Y-+
zLfWr<A!@;2uogdcHWOe8c;zA7{I)(K;>N!S^xG*WSfUIZx(lLHbu|f*i|wr948?|_
z-bLS5X!uTwm0oqWx*8LXV^8uH?r0E=+&y`1KM@i)dia%VT;6u4m%!IMMcsPq$^0?0
zGsI2+QqO1+G2w-9HWmNyEv-`<Kip00e>o4Ug{4hDE5r_k?gxj+gd$9&fH-3`)hDs4
zuWVORyWeobwAl2miO8|3T7omgFL>vwOVHpc#ZUkA$!(MDEn^Nn&-zc9@?;cQA1-ay
z(D_WI=V#To;zei|v(s)4kM)l2_S-Bs;aQ<}#W1cQf~?vBLPg`IvRfWD|4+77E|t%3
zg5gF%xf)xn{XF$**gC4q!-Be>UQVLTH*=(`vP?h!%xJ9?k5ue^!PlmM`62o)3=Xl>
z2oFh&)7tA(wFxyLLaOSzz!Myr9f&)rjHj{c{L_(>>+otYP6Dn-)D{4OQOsk|i$*9>
zSjd4wyfORpam^4skg|0$KLOLjt>b^m^K4+-CALhBAQX?^5)@62P9Bu(j$(a*8j^5u
zS*)s)SHfbOBjw})3Sg_R#zE`y_%vXZr)FYb?WkKV7@t+Q`9~BlQo!0dCoXbSCUbJ<
zG_uU;UGv40hj{RZ+C703@}NWh@#6<94Nw?rn<Li=+@kjWD4~r63ar9|zb}}M@~QlQ
z`N^~3der1F!Ojqxm=@3QJ#_?=rT>i5*fHiiF08LDMn9bR;9UKpKKzd{iM%HwDqL&?
zg{#mX-$|<~dHbXA)-(CW;5H6IKx1^#Y*lyvB@}~Z+Ik`JVYCe@(JHp<L?4fm@>HmX
zmu$@Wo)dfNy!S6dqmZKnLEJiIy-=UN_l1I1;oSBo+u=u@K0;dPx!<y{x;ZwUB-E$W
zFmrD>IOr>HXs)4X{)$JT(9e)IndZg%nBmGw_kVg}+VB`}geWhdQYEYi@upA+qAv8=
z<GnWkR#UvSb5Rx(;h|^9fjS1fEVEmbB^LKcJgu~1I?9(TO5`Gd;h1+UDrNlh!sA{P
z-!&NBlX7?v9Iz%#4q4~4fN`c|YaLi~%9Z!f?axpuX)#AwzwEk_zc(t2=(WNh9E;H=
z=Y>d44ijo0J%{OXZc}fiAHD%*<#ga>05CvE2Z5#J=hsb)*%-Qb6ERPrt%dX6Uv@N{
zJNee#ciGI)OuwI^s~(4wlTK*mDsJ0PYcbY^ks&eGt-T!A(J_|pq~i%wUNIrZd0!>X
zx5i)haWV-_5>?rvv$Fa5(6QB~szb#u@hS|v=d1TC>@440UF|K_#R6BUD-1r-?Ha4W
zzlB>!T*`gl5&S_%OKX1zZw<Gb?8Su3v9b2Gqo2pjZUhl-hCz{g`0)cU3O=-fIN}JJ
z7uh!{1_*Kzwb5c$_2pI-vsRxw4L}zBBBDx&X`K`#t{^~87@44=GKv@vJ}MI{K|GS!
zIXJT_+n*0XVA{YTysgd&qr7ax<2Q`BFbamu>sGr~2p&I+|Dj_OvwIrknQrNQIa|iR
z_v03dX@e{BN^5f7s%U&Yq{e<(V!UxszU9*sD3gVy428f#Ity<<@GQ#r6C0sY&Yu@T
z7JTQ5d}R=Bfr*B3Oj;Vo^YKPDl{5L{2_G+e`322<FrVzc9TUp-Nrxq|=wXe&;`2Yx
zuRQ`nRlySG;Xzf;^xn*3)}DD%Pm4~UJQzBl`t7CdW|5hP2{Wr_F(pRI{DhKNM%ifB
zJ+Vz{um%~Zzb7@H)A};VM7O2LD+OB9xKG8EqvY`RT1SU|;aE_I)yM7t-eSKeB{G!g
zeju>L`iJAjf`5!PU@3v6aPwuLHU!%X=|CTF9jk&<Sa}1gxP?`usVM+HJb4TXF)qaO
z#|Nt5DklIDr)-^qc}9o_(WNr^&P0dHSJ85y&A4l;pp=@<J-55H9mx30Vzz&KV{a%w
zkXo=IK{e4yIp)?<`5=Lxite_!!qjA+jAmHyM)<r`>|SxGYsbjs1=n{L<x1NR%u^;x
z5dv~dA^s!GO4}}x$)aK+k$VM@skrk-+4j%uO+Ui=CO6`en5pnv@jx3i#vA;dHU8h<
zyRR>WC*)>$LIw_6i7Q7x@~QW@o^ei1Il7wO%-@-R-6uQte$wQLrF`}pj>*DGtG_O7
z9s4Wbh}7;F)gyiHBH})4m<W0eHlGOZ9cs3x$HH(WLCR(61Pke%3dP**!6;(lbP`$O
zN258bt0!Ljz^Gut#L<1@JKK{Fi3z#UyHc?!lgmEefu}F)o%<E8ov_yiCx&6}fqjd1
z1QMqe=Ze{qz2Loaq|I4!^~eKI17S}`rhEM|9{%J<G~{jXgUoGn7jL5myczS6wmPn8
z;lH-%3AoZ27&SS_#lez#uTBY5TX$EPrR}VUs48Y_(pnhHQZ!+>XOX`wOQNuIL95dm
z6wN4qNK^;}BasfMyCm2MLXjXR02;XvfV}KNmIWq<F%c{<0r?2!H?D6qC@8|;`~_R`
zC&K=Ltv_osKdzg8>E}WcdRr%6-KSG_@pO|uHz}?Kb}UXMR(wLaZL@je>X*fzd@~n^
zd6yw~czelT^7^A~y0p_$k+g(B<mUK0%WG2N<I`uqi0`nHS!=kjb~own{EOc+zJf4&
zT)m_>gtSaGUE6@WySir0Zx*Rd)g^aX3dQIfgeJ|w--SLea7+?T(Drg3ah7a%DUD9#
znw2heFFP^?s~%=5-`rxf363u=X-sETS28K=*uc{E5yJxmg9iS^aC$_v3*vb1yg{)9
zQnhFeUpDl+{`oA8!{eWC2@G*eNMZrwAYm$-YmIr*H9^rDkPF*5c$}u~-PQoFfooTf
zvD>AiE9SY1@&>NxzdEb&iyW?mLHYQ-wH(!&XvLXuWH5L*q%%3IcB<_R@#HJB57`ls
zvG@+bfh8DJ$RUA`%GHfa%38%*+7f1l&7WvbSCnpd)oW2@&-+}p)VBO_liU0azv&=@
z8{qA<*qeQ=f~Gptr~0IJ_&b09nv%JSJ6y`)T{mTP+pC?u=1Y4LolW-bU#gHQq88Ye
z-PFHG%<}xA*xl>Y{|76G-}ZIPsGWjvu}arymNb%W#p0sL6`8Iga_cCRrdYd~c=O|F
z5=kyfQ(a&)pGN$)IUY2jXU7MmwEO36hL|p)ctykSx6~AV)r^Ali1e8LZGBAhF*M_+
zN=u_*WOmwOzo1&<?d%<0g;F*k9J)?g+0nt_F)%>3H`qH}0=1FPxlTmPtLd;JYkrdM
zIaDARcz}@h^3VBbUX%gQ9|ONqN2|J=DCW1Ic*bb~87HBCyg__UT!dbaHw1MKaOBC!
z%9bLW>MPDE$LUCYPVi}8IH}Vo-p`O7dUDR19Wc6xi_i9jsj=?B!xZN?g(h3i$tneG
zE2&}re0R5pmfqria2$?l#Dq!;h;*%4FGFpNNLr&%LtoupeVe)SO!8Q0Y^?ItDP0?=
zZeW9kvwPmRJqk1A1e<Z5Feca=DADWfV>87y(QdC~I#oCIgeN&jlVMsMN%MZlgJpyD
z1Br+B1i&fT<=Dp3ny9&9<%16{NVEiuoDMNUWueS2yfT=$Rs-LchsqmW6Y4P$APJnD
zOocOrXb!ByGI~f|2z36KObe;%K#b=9bXmX7To~0Pq8G>SqP0WMAkX8{T?(EQ<W3(P
zDhRbj#oR1(jW{3hT9U2ZL<LErt?WDom!|pKfPN=pL+hmrZmsG2xVyLNAIdW^QSl6~
z+E$h%lCW^7`u%VB%k{NkXOe^*&n%j{U+!`ZmmlSO#zveUs^svJljIAuRl0u<#VTHI
zIiu)s?+a`;*s~YWQ*49AoA&sHona0C7nf5UlS^+ZVdN{GJ45Ky5}+#7f9d<`<Un4B
zxmJta)$;c0LkYvfZH9Sj)<)B=6W`}7-J@GOd*rYWG*9@!kzT=25hCv#DlU9vVnA;N
zX>~mqe38$A@Pg@72KFW<E$Bey>)J)=y|WY*o)RQMv1M{qaY`Dx^Wf@=3^TIGH_@^g
zwdH?qI!{C2Ca4MZJ{(|gX8ARZ2(cVq8zEXP=R}HS$NOoQ3M+QEShPiwme?WfKWvT3
zae@0cS@k#*+z^45>o4#jio5Sn)Gbh>$&6^}!rkZo!)SCHR{u_wYtHdEv^)MhtJi31
zh2-*NO~j_HnTec=udeQG&0pfZCAJ7KZ*aA}C3Us4vx&-nnUaY9*y2gSDlXDf0hGUp
z>irB0evK0=bcUwnay7LDDU_2(Up<G7&o2z!Oxk-(WL&cNW(CSZO0;sQvsc?LRfa7a
z%sI^)vxYjzDWi&r8B9>u8bzRi$GjE{Bg{Amy#=^3WU~RF;Dl-F`B9=MP>x-MZERqG
z-UnJPG8`tw?x+a3^KR3XdN~iMbd7!c4Z_cj2?hpKatgxgx=Ih0qohSX{(cnjB+3`3
zb>4j*3>R_!qYRP@z+cR+cE&XiSq~_Ya1sN22N4-h{OE3@i0(p(UA}jxNt=W%(^(VT
z0O;<pFkM<4W%=dqu0PKNEk2<t3*Wy2Bf*-qu>St_;04K%1J|7&H|?d*G}X;$`dG?e
zpaV<w)zWU!&}f<bsYx9_lxY9|Vs#8pao;TsSK)d5&2}~)%XafaCTHMWl@P5+Pb;v!
z+ptJIj<w>7(u$)Gni+S<=O7{*(d-I*68E)<N3Uc!fxrp3$sQI8`|rRDY5a`0Q3O?&
z0CHH`M4@uP7{|bXtceN2gh#U7Dk>}YY)%nXosQd0pt0tNl5;?hgXl78NgC|l;DZA$
ztJ-g1P|)#MZhltnU1j#;)aWU7;)$Ne`ART|;PDNF7q@}jS`_>q@l_Qo0(;?ZNWY)N
z#>B_0vghM%JpC3K{(p$}gS$N1i*E9LY|{QPWU%ewoBSjnClPwtO0@X)3+Q}I-69In
zHFRS^J4>%eu=e_Yfuco_2eCZ*tLRjlVT9emp&MQAZrC&XeBU~HD;z&L$=L<h2?Y^4
z5`3q>Jz4O*ds#qA|A=tSY2HNTkk;IF|KMYWlff@nRf30wa{%W-<t7d};&J-R3IWz4
zX+<auRIDR@icmQOyv<lN_C;93!OT-nE3(tY507gcDE5kTr@=J=h!12FWJc$LMB&M<
zUY>iv(d6Rd3%iLL;Vc{r2wR;si#>%_;oY$?k&WVl!?H}~T(#|{iM;l2<N`PBN+|oY
z^QL%u4G2CnH8?!;M31)3bpKLu&Tw)6LBDI<3c>WHe0kY2ogJeZLzgy&ZT5IDZeB}_
zH%~^#eS2~7-SM)aj<EkQxuNgZYMJgq^&7PDopxdC?af7kqc$6Rc+8&7Wn!tWdr_sW
ztynAPp&mK|e{`b>SA5UJ)Pi{y&<2d(ae0AsjbCw*0sF!KL;85`LwO1lhuSS9{tj9j
zAT2cE7%dYs8GRz=o=MpH3VD;|o=NV>sWl=co?>DP{mayfZhZ2sTkUqKCT}YzEjr+E
zvkdBAjL^6`Uq2TJ3_iWqax*bQHBr~IN!U8#4MQpt?jQ&b5^UP<y@<uHYjk%%V2;)$
z&|KH#pd^6<Xx{`T+z9B9%d&n3?gn^MEPM47H*;G)TT3{l{oP8xudcr6ag8Xe5>k%6
z8n~A3GuT<T-+byzm91a-_SQ{#wiKg4_I@twVp>=$2d&}apyh&aqfKG|KTuqhTQvTH
zM%fagp?-!?0i{WY)BK$$4^io&s-CiI7<eltLYNH@QUfP6(ieDHCX=%XPngxfpD=)g
zln_f4>^)e$D--kw{Mc(O+S2%`SpF9hf83K66q--83Cg1uwu_tJk`A1(BTe3eNiWAe
zJ1chor;8(0#0hJ1kT%=m<{GLU(5%PV3S<$iZKj$WfJlS~2x>qHHqS0X59Hs-?e-9c
zf*5{a1pqW#a;f`-9AUyvVr56WL=XFEP#=jGZk_D@*rT_t(^Ob<T)IPJx@OC7{$T97
zg7C2R$}=ant$o|T5#FAdbE@LjMYT>Jflf7<xt(KA-U|xVG;#!;l+ck$Wqm+Ul(;Fs
zxG<ltbwbcF$S0I4;>7Aa5L{5JuNCrRdTte*8kzRJ(RjbIC0@PA3{VYeUTDZkEBYXY
zCCUIM37^@gs7a3-ItMa>7BMS)^B3XRfB#^o2;Vu27juZ^1AvG}Fe!~N3LhI@6t*08
z;*q_9g92hnRRY#93cV%>4~~o;!a;K-hw96NJCh-6B`huB-ixcA8Vdd`qqKwM#_k%{
z+~tij*y?dw<1X1yq`8@-Xz}9=P&sHQv|>4r6whVPBhN6?Iw5y9qi0g_bX-H=6ciTn
zOQ<VnGt+eyR}EFa`=h2>^665ab5-8Q4-Zs2pEH^Df38VT?TSzF$qRK32ow-;<fNH5
z2RbRQ)M-0;VrHm=KuQ0{(Viy9cyp;@F%chvS<%VA6qDfu%h!6_a>8Qjakl;Cv$o?6
z)?&h?7NRrarQ_}+ChuH1$tVN7`rnXg0)9|DS3oJ$Xx_jx0Mf_-I5CCAUMJHWiRTO5
z|BZbg#UqI}7**6<(oYe7t-=?M=14lVRL->~WhvO{Na|xdvoOqlBFtkadsCUBIxuaL
zLti0*RfqP#?i%(0Ca}Prajk<~0y{aRg@jC`OZj1z>#djv^sG=ah>)hM=(fcjX`3Yp
zDl#lmNqXM&FeZl9c9oax@rcVye_o!Me*N4dQq3Yzx^veqck_Qtc`H4gmv>jkRK~r@
zZz!a|Ezl!+S*^V0`1{s_RjQp0R@Qm5+MODAN-9%U-xw(7fb-rT(5A)z8?>nnf`d$J
zc6}I(#vG?U4K5*xcIMJ0xm~)MG0@HFSzzt1J4moyo4*kSE0py(6JA81$l@X;kf88@
z^8tw$;|c>5jjB~Uj}Pwyk2=uBR&O^9cyYkiOCuNo<<GUQ+3d^~$C_82n*|mzD9YiO
zZ`-|NR<=?vzRC1M_<VNAY!G&a=nfBvDEMX>{C$jL9uWDA&*FJ=K;XvRiVw<xuNs85
zsh1f>f)*=a5+f!Emv_bIpcyY;xa!yj%aX2b(4$nc%&s+QqKn(|Pj{aB<Fw7Cg{^cd
zb**=?&Nw>R3cuVz%ereWe*Q(9H7Q~ujt<gWyu=2dJW*fI#k2lXL`R20B$3}?NyK`<
zYU;Ih&N#_KO(k3sNLl@NVIiSS&W)X^9m8G$jq+6yVPVfb*HApC&bQvXFc#@Ic-~uV
zywE~n_zW*v@=hxKq1olqi!0%_bGd4=6K*b*H?WmM+OIPl3K^Fd&RUX)7Vn{2zL+Rm
z*(2wLEi=qnrLG^r&kDnsvoChU&ezc0&ovjmwi0X@(*U^nb#cMwlONeLvz_c<X@DYv
zmjzBGm=+KR15uSiBzMlDSXK~hZ5-BBgh=zSZClskk&?5?!rFqI@ojZ2HE7S&jMwMf
zNHZEcXH=G;SqO9D*~04M?6D77-KGC%V`~p;+1SfvI3edD(B1f2&T-sc0hzu-?YRn_
zaf8N3Z7$?4Dwi?wIRo1Ck(wtJZ%?|b>5LH0!$hKxkoZn#=T|Fh)qPq?`#Y+ZR&ua;
zZ0uy<59tOA>)sPC@jF3Vt8tuR&)KoWW<&t<43Vny*=u0%ry*F}fG`*s)xcO4{5tGX
z$Ye1+1Q14+sR%ztdn<PQ#Tie=DR7k}Yy#{a7nVV(8ezRBR3&bA9haa=4!^Z&gQtIw
z4aRrv9F<NMaXUdQg;6orT$HUagWP;44cT6A<wCInQR9Wui0<iiW!?JU+ATa;*rOAu
zzY688@;O)E4I{7(?h)m;K0T9=LCN`ev9px(w*fcy`ci+%GgNv^&Olzbs)YEdWwm1g
z?hyx$luk~}$T>uoMkJ;)UAcLn*SWaST`w@_WlU?z9jnw<n_pQ#|E5RJ|B$l1&mX4f
z=9>p?RL?hey?RTai<ebX(l&7LvGnhn;8|EVkXd80uf?Kj(zbPS`<U4$pL14pDV$Uo
zMg!uS@=XvDhMopLonnB1H&qKvXz|_~UX!Ys7Apv@9zwWb)kqzO%n4KX)H;z1J6T<@
zG||cO+tXs;S%Z5T%rG!ao`9jCB(&wbd|;|@F28le^)=c$Ho$1;%uaGxk;j^Wwbami
zU8?1Rvg&epSupl`eH%Fo{!cHG<S4=tu)AIeOcum41UmOm%s%W_a^Ogt-%O;R*H;^U
zf3D7{0;c}HK*AvV7jS4oi)E89N)x@}UhibUbW59aB7RDdIg6RM#b-SCi%C7LZg|0F
z3@aQY@?O%ubnXsb{aJgp;%d^u94_US(e}UNdjez`HYdwy|AWnmMg19fjqlI(-#$->
zWy_lLbK2h-IZbz!d`^lGJg=F_+`jKk3ub$)(Xz)xTQV9XR?-NVP(+jbhTXdiIhe1C
zh-!vVSEQv8!BZVXOdJ4ne83k6UndDj7xSl`4o!^fD;^aV+-udJV$l-nC1%hanP&ml
ziR}C`-79~n;UVs<Urp_p@OlmZq!16%YWpLVp6>DuvSsH25Wqnm!`KZJRlj#2_QyQG
z<pQ1`^VL;uCDB@`sHqqpUu48d%Z<&`7io=`#TVR#&F_}kMs#oGgovaLZ%xvVoXyGC
z1Xg@h`KWDTo2sVX9t)!P?`z4G?}@vUwB`Z*L9$TQ+0Fwp2k+&H3%&FDFG2Zl$E<%D
zl>9zP`85^Fc^2)de{^`J&`r3sy|(-5s_DlM{PJzpS+KHzfVYPZK&J$n&Vy!CbzK`O
z5CC?gve4U6S8#g(0RWN$nGsxk0m}lD1+P9bn)(C^O`=q{2$c*o+N#;U^G-$&dUzAs
zhT8cN9#wgEK1Dg1O<T7xg=(?z&Br5$&2oA%neT$a!P*Gwa(8JQGT7w=v19T1$(5@<
z6cPAqQQ7X;h5k$xpRQbXtF9F%2;GKBM<3gdJsZ-g7xhtCmry9eDr;=slk`GJ;I|{+
zuGn?ut6NRkoG6s<g=xkq`J?pM27yz#%XDkCbf-I8XBwQs<Q>9tF={*aF7J^FZ<}Q~
zM$CT6^Er*&qh}a{g<qj;eZp~FAlV%J{H&vc+r<>QF8lY@>S}2<1w5!pk?QIfFPWMn
zeZuzgTXrYdo-mB|F<QVAB&Q1)GZ2Yjm^U~VByo%4W`-0G6G=!DNPG~mpqR*kjmv0z
zlea1Z{0<<XDVDZG3syRgM>6&aPE#w`^7++7=v`;gi|u|ZeJK^OL~}77<F}T%xJ{Wo
z6^pvX2&_(kFI7%p*^Onx2#X<nAgvkJg{PC5Yd8n)x68Or6917FLcxWU13^yT+Ldhc
ze3IbIzMjf)src%a)=l?*_)UJDas6sjTw_$VH=e2R=AUJ2Sg^MCym;*cF^M_;kvWnR
za`Cs;D25w-AdZ7O>!!wf9TMxbbe}Xk4;<Y;UO(@WN2oK`;JX%n3~>TtH-3^)oB<JS
znDPe#7<M@3d-1~fe<OgFSEFC|d)RaGrsV5a<A{}4pd<rWA<hCn-@&^I#wwMp{-;Ri
zX)zJd7cn8pHr#g*k+{(at!o&RXN=6C?8PB~*%`6{;HjQSn6aZcNX|78!-dsQDI4+H
z?i$qA*8^&$R8aK@_Y^Khr&%ZlxzDj?6J4MJcU^e!lu^9G*VbC%Z4(l&F^0k4*E(8|
z^d1rz(r6Wlm6y<DK^`(<vYCfzN1P^6wQ1sL0kOU8YzGc2{5FtTJZP5Lrs6heUOyY{
zp{q!W8BT6zB0N^@OdVG3j1@nyUZ^T)A-!L^Q?u0|_~cq9L;2@bp|lOL7KsB!RJyhZ
z{pjjV7pg%<+ElZnQYMW{A%eTVWjgM^1tPkgU;@4l$z9$GfA3!-c^8}RYH~R9*<8z4
zbi&Z9Avj#LK1>S1AJY=!x@(LoO2<Z)j2>B<U5x_{Bt5KQa$44vh_Kpn$lm-(b3&`%
z(LcWf22frx7%woWm0<JwfvXOW+Mq17zVT$F@CDRkWc7;2c-}2Vx^@ti0%XAvWprli
z3W@wOOxgiEs4T`M=C`~lJL=NxdxoMOJDo|j2CRuwg}vkF119Jh2E!M-LIT`+xUjj%
zynPvf$_(Wer<((&2m}+|t?0;u=3m?otyE|b($eDdlOrp*eWBcKoJH65U04{l_fIa-
zo9b#dK@qrhmM*WYneAmQUzDTH)CO%1XWeGaaikPf%IFT+hn41K$NTd1Us|T%nB@^L
zBKLo@Pt7+ahQhx(+z_a-@;kmHb;hd;23NMadU{?Bb2rvy{%GzqG5^RHt-{?)je!)x
zFk88R0c^|j4ALTjJSnq(nTwf0bQ#Aj0I@iOK{;o5;}Ju#2}PvxF`WH*GZwhD3Q#$4
ztcungH4jdFYHiad51otSF4FlFP1mtz@W~f`5zP*V&-$Z)PLHak_0nLk0;29IgkS(q
z4Yu?UjpEY6*h^mFVM7nFYyd$LY$A(@QeqF0CxBpt%!c-^XtyFF^ln1Ybh}NZU)z2Q
z>VhXVFuvKD)IskkRe1U<X>J9{4*K2}RFsBlHGeS8LP3?n^KP}GCI`|k>l}vTE+tLU
z64!8F+Qvz1+TBbGi+&;ciK0B7+z^3C#{Zchq)Au>|GVXtx^3B$y_tg}`HR(f!?(!k
z;)W%y>4V;91*ZpN9KW>)w`ro|MH>duYo%AGIK&U#5M}5!W)+aFC*=S00LT#f=w#hb
z;;w%qu-6WiuTt-Y#z>oRthN3s>vaydn5vEkpG+7H{47?w#_rUmDuk=BUc@4TdLXaR
zL=0|NVYAvJHv=&ndMNZ-c#6C%Ukr(DQM|BdBmvdI=mefPP8ohy(9*BU9+3_q5QFlW
z&QCMmIDZ#*3TN)}p87)~5pQzL%Q;pwl-B;e$H-}U=z#dlbm?-Yz?VOFzq6s;*-=><
z@^;(XyhS5Vn}ps!sp+*pZ}a-xoD`u1C?fo*jt{&W-sx`cJT<XD4+HeyQ<o{NV@+EN
z+%G?|;#0HD>oukCj!|OzF>SOBE)RD)-M4j_#|;+wx;&V-w6EN&ZOObdTa`UY>pUD~
zxJITWE}C6O$uk#&uN+Yvjtu1<=nBXoJ)E5{*a0#wc^GWPzkuawvEuj|TTQQeTxQ8H
zA2yBiQeR!W{!CC%7}GIW`Uo!|O4@Uk0$WBXuxdyZUVUzs4i}U)eoyj9(CREYUuE`8
zpcGc)Ixf+5T^Ju@&tt>G!ZAk_VREt|yXv&1K6-cYiG(GyePp|IAHgDjyIYJaJ3F%p
zay#y&6aAH@XUr|pS%#=*rTU2r%-$isZl`WYb#>-rQ^&Bd_Uwk{O9Njzr~R%vJ73sK
zxK^tzrQy6ZkN)H{a+k~5A?<_XgVK-sh=H*dx?`z^@%PmVF(rYf6rrjT-_DJK55E>^
zn{KDo2_pm%-U?2hd~kop#0sL<8~9+iGd4WQ9B5gXVCD~-tGS50%Mk^W2&6&=uMnd|
z3=yHuIRC<$%8Q>Lp9Y}R@~#bJs|ukb5q*ShhNAX+T`jX<oRrQ>X-%?=&G)F)jj+pi
z@RxPOu&Z4bsH$HbXji#g-&&kbPK9xMfQ3g}1VQoegjE70wE!*DDiBkF1<KI#oU=fV
z-pj#ywlU+gF053J_*b;dQQs>gBY)PfS7M9L=3gfa!(w9BxSLzU)-rWV!ke_`50fSn
zkY$IUNmjb{YTR!JhYB@oRjil3tXUL(KA@Q(5(Hm~jXiO0@G9YD2`|``N`a@<;-C2_
z<EqxpPsH<glk=)Pw2B}ri3tO|AtqL#S<Kd6tXk|Us~YFHy0HPQ2AB$BA3mV<1FR-+
zE0E5s*Ga$M%6Mpj#AOsMbo{StPlQNsIx5{(#Y7kGY=ld-JqZm0{t~Uh*~!f5pU;G*
z6~(|R%LGR<=d15jTcH?8hJ;N~%yS^=-HpViK2Bcth#<sj1U6nm`Y|%p*)^R5ub&sM
zC=tG=1yhf#(I*Wak9WB5{S&0x=|8aCjI{4s#WZCH|EQ3OOm%rM0=hy|VvkZickK&N
z+plyiDbO)}v-5z=O_O(AL?4wZ0_NoJD6d~LY`+EFK3%3i{_u@}T{KVdv!oDy2%zdB
zr7=;BUPzx$g9Sc!>_&~CbLsEq%LMRycduN>Cn_d_eih@P>B9lKXviDqLm*fq=|?dZ
zT@Eq^@qy96U-yYVKr9lS!KteDWvlsMj!zV?2OEn=5Gqx3tnB73JQ?b4St)Gz6g&B~
zc=1qsE>$RrWA&HSZgkleac-R|sM{Y$D}lC<0t=90VE|hww(g}p%qNjV1j?92bbJ;5
zl!-X$Gze45LEB^uXGwM|h-cUvzSHYRx`<;w^Ft|I1wqANT3^Vc6J1&Q=g4I}Wqhp}
zXZW_04bSzhWm;W!^tmY5x;zRyN?qBrS}5-{pUt^DFGm&)zv8=`|MA&|%$spaoX~5o
ztFhW7yiVWhgEeLKlJNf<@1=3*<JvIq(IXmy&L?bYT9SJ+)(Kji=gEeZ6iTa_;x09$
z`>t-RNC$xp1by_azNMw4JNT?0k)U3&7nO&$W7%O8O4<Zs^#XYHNz>m51Nq>U{6~ll
z&+K$bA{xeWzH1+6AYK+Gfe;EA7MX7txFmdG(kluvxknt};UG+3xH676a4=wP1yI4b
z6Gj{*BuA3W4uBmonN<rkkeg7ZW8j0E5%9RuXT)`=NTYD=IeoSYw-(c7Wp>m7H~;`r
z{A$leE2k&#?(C80x1Rn!mb2B`Yg56WM0&8_J}M|v;u9y$LUY)7D=icM0k7UCOx7yv
zhOi>Z;kD~M9`}b71&5>~$2J5pRetfT=qu}K4*hRzQ1@v8;WI#c`Y$qLR*iOb|En<f
zxNkptc*lLC9$rUtu=z?-EDlbbQ@Cfq9hsgTpb7t<!SP8|%3g={EPRVVOXSW0{fA5<
z#c_%iM3SRR1(94woDL@2v5VQ6!IVRvTDJH0wJtc%9u3WbeEdR9^@+_I>|IO3XhzPG
zMclAd!#OS)%}Bl^9A^PG;=^z>sPF#qvKG|_P<1?9jK|rT<xQiNyXFv27X#Ygj=P_Z
z=Z`4bN=8G~^)%<A@C96kw_|+reCizI3t58-hNP`SXa1B1DLW5@*zct`t8_N3Y0RHl
zwWQnHj$WWN+__(OTH9~NT<YqW-5ga>R-+5?+g=@AeWAMAJ~V;;cta+quVPO4o`2xM
z94&3d^s$n5HvtO8)cDnxTxf&zn<nP>*F=V`Oh|ZHaaz9o++0TPCyMH$D0KG6Yo*fY
zs_ZHlPBWP{0Ch&?h>093M>2JU9L+6(!*7I(cMV}<f(6SVqn31NMz`&|Czux6+BRVp
zD{wlZ6+pWsQqQuR*=ahfZcA7=v_2_zsI1`qLmK}OF46d6pYTUGK>{6sWfn#=5!SE{
zet86eLWBkLpgeKr$Thl|KDFuuD-<8JSlC_FRsej)?kDrIoG85lKN8?A7V{ewL%k0i
z75Wl#<aeUtIa%#J`?MbIHtbtS9Z#Mq{9Zygj0|gcIQG4IVzuL}z_BMmZ?~~G@9Fnj
zm%2!q^3LIEY3}L&)lT7m#S3k8t4hFt8y?lGGMZn_k+o!#(3~>0dNtLpdJ%%Bsy^{@
z2hRnA9ddWfVF1sQv?9#qFrQhpmWszs3eGS9Wl%BD1VqwmIwdievRFa;LEdW2*|XF&
zZD!vg0;#}6M5PKBm#z&1LBC?hU|0s00FjTkoUOu;vo3nyPB_bApJ-P>8jdaDv<7}W
zC?X~lN$aG^d`S9n9N}0o;z{nt?l9r*w^Zy&&oL1LcZhiwcf=Xqk8!PEu1PB7rMKZ4
znNFzCCOrBi^m=qN*p6iWlU%jFmK%P0{4lm2HHO1M;g<eWX6X}|Cb}iR#Jx@#+J2M>
z+J-#wg)h~EeT+kMU>rH`!rN`SKt}I(mV8K)_M@grHZ7o&EaxVVeZK+`Hwt(CcOYU#
z%KW~l_eE-!l2>m^8IOorON~n?H@;i>1vaE_*YQjbmW_!e3->+#>pq%uq(?3hM=mPb
z_;HC)dylN{6MMzwftfOsY8#AY?jKZz<1v^-uug`10hmOglF2ZTBp!P)Kie)8Byc0(
zq(pyw!aP^-u8wW}<RhT7No%;n7gc;Lr?8@!HH9=ojqC8*m2GRe8?(*or5P2Lfb?EH
zY^D6gGOaJknL>WlGY*<i7%Prs92{J#H+Tct#v)s%XCbI)Y8&C_60k6v)~zM{C*H*`
zWV-`d3rLdkl)ASq{`#|&V}J^-SW;J(-Mn7B(seC&__|bJvlq-9(wSF((f^;E3AkF_
z-)H<U=j`qJFp79c+j;Z()uR;^sX|rb?;Wl7)ePq#l1|%OPz3hwU}q4-aZ)LWRtLo-
z5&q${9&=Q9Pm>g)U;H816JZ)<w0vD{L=qczFOI{=tlAXlwurcBP{Uzdh^q)hw?X#t
zUF-o{oA&*3!ozQH({5MrD4PhSr9u22A63T@iCDpm9w-EJCda(OTn}ObBu^#!)O;~a
zkAcXMj=tb-?#2h=>|@QCJ>Z0ou=btib?b8EcI>a4`6=EmH1nwy_Ux?eyv2?!9m~Bo
zT=~att(KV72ClWRyV!r->~oP=7GY?u@4Jn<(pKWi4u{P0#NkD`rf@I}rLF7iI3_c`
zU4+grlxnI0HN-zz9;I;gJ{nCj3RRs4T6%VP{B(?@kEr{#^qjBm(WrK(>b4MQKO80#
z>I#zNA!7xnM0)}mciqN-4)Lh4vEffYC{J_+cMKS}7ZeGqh-lu!pbPmo&Q_<w@}PIs
zU!;Q&2hHUo5}!%$?S|p|R09LdW)SCXLJ8#uWDS!Q40tfOxuw!;_()C~1_Xb5jPATm
z3|<dBB?D1*!|(|1fe<Z`UMg)3;i5Xx@efkE{)?4)dPb=s<L$yDb$_x4IpKoA#K)t(
zFJ}G21$jaujwvF9<m6VYxAP_urEI$^N{-jRs5(2p=kZh54>=a4C-^>RTw6dWv(A%3
zWN>&5KA?od;e?v4>E1H;cb1c<@qoBcU-natuTSMD6f=ExrcZl%c%%9KRw1x9?b_KT
zDG5$@8y-6sm~hh0SA06?FBwMT4t^YLQHeztTE^p*YYEeObc~(O<7cm%AQCxvKjx~Z
zN$n4N<#NOMx)HEm_GprZg~Pe^DmqCM`LX4BYt>cYxJ{n0@v*D4wFOw(3K^4xh1Mx{
zhY^LNRodH#&Y!f(uY@E)yB2+vyhU^XP^!Qluor)MwUesWGN;)4SGNaF{gLMo3t-#P
zS?hx2hyFi&%)91SSI<A^nYA=tp@F@Ql^!EP?}J3qHtY6)C;?mrd(grJ>>Jx(dR=p#
ztvo*Tui=aQq!MY$h;`|xSqN#)FB~v>vr@ga(B_zvUXmN^xUYl7a#u-yvg^~spT0r5
zv{6Aecz?t8hJJo_5Blq2FL45C3p54kss>d0JAD%jZUEY>U!4|A<&HHMiza(9&PRKp
z3yIt}^`LwP9t!{zI#k@zP4>G(oflhBZ(%t*tw8`FuB8=EK+EH#bzh6(&ER%RznmDj
zum-mhCXFIR8Nm=xBeeH$HcQ<g=qzPwyJMc)kRCJs?#>o=lUS;i)@2#o6%qj4cO$Md
zwf<9|7Y5#QUs~<abMGhA3BFU>zI|Z0a(>bfwJvO?{(=7C$*S!v%f+_VZ5c8*=M1;5
z^f@Rf+)z3h{R7q$P7Zaxo)NUu`M1^_;}=mvUvJ&{;BZE?YDl`kE3$E2b3jY0DE341
zmAf!j)_8lY@y@f5=3z1jHj-DUG^*!mclCMo5UqpXk2_25I`!l8;}=dH9Umf2JZ=Bz
zK4tUdzB#vOxb@#wzl0ms`~%=S!g&Hyk8`YoEPOC;$q?X!0tc}PKvm#}MDU79yJQrz
zC>+tRKVCf|{;L)U!LUHr&6FGU{wSg6K`{)<x8Am^9^h=iat!L~W(0o%8$j}&@PfWC
zYV4J+<#LnAf+xE@?XuXqE)2FoAkeu0TOHILu;yn*{Plz)?k1&mXu<J`zgNc*LEVol
zOxCaMX5;g&%he^Y*apkkU0yWcoLrf8Rpy)0ORZCbN>A3g%^;D*bJ>X2xaqbol^5|*
z|9QmDfyvZX?&WnYJ>HAch3%6>))h=`53{9?JI&n<ohb*C$idbfF<}9BS9cA;U;s^l
zsNfP(2=Z%QV;d9iv+-X-kjttly|xhQn~Oi6*twPtbv^WrDy~&%N*{FDtJ@NcUJSe_
zM%f64yZ{km#n9ctS>_dBR~22{%ly<yb1-2&dGHwUfnET=q7>k!8Byjv6eU>eEdGY)
zS*>xi)Jp^TtI6?Lm5B^$uCsOO9Yp3haT=I3`=T-?xJAW+B6?K{{`p!_d!Bs1ZD4TL
zVg-W%k2yCZF$fhkfo2h2V}vgirpVgxBMIeq(ur2+xj9I#cfy-R2~8SG2||tC$;$#K
z1i$IMR`7q@pMh+pmn{cX#~0|QY>)Y~5xz0h!O3^nbYBZC9DYrT>HPBE_h#?WZMj@J
zF)=e-&-v6abRf)r4JWM?hP|=RWS;GZpie2HscG@^TZ`svXcMN^|AbY9-Ho06ndv^U
zP;$cC9iE7qOIo%(^!Nu!v>CA2YjUF5XjrzpdOQ@C5Hzkye?3astBm<0Wg9V)k|qM`
z0OaV=>{dwl^S&z6C8C0fCO9lrKKs*Cqr<UjX~sRgD+%ETKz{UR&K8@+@^iRaI-RK-
z+B&!*0{?=C9+{^^aE{TUF`rqxq>Z&EJ(s~2MB_Ti)9gy<9zY?*oQ%oD%<{%uGCJ95
zWtJsn!W|kLGC~7pll?p3(QKyhJvX6#xXZyu=jqIm%-PH=I~C=!UbWo+cdOgQ%=0d|
zLBVk4DN*^ls(M;>$vDsarB6i)Xtad=+{Qt(nbsd{6^>bLBh_+nUgFqp&wfuj>0oH`
zw7N`(x$_?e47yxyQ4|esIdX2au&`y6-T}SPQTPU!*c-+xaF4o;$0iGv8nlj1PPR^V
z#bKTgCK^dbLrf_=FFgH`4GgQFp~Eq}f%;h|lmG(s$f0^=M75B<G(LNJJiH{Gn$}3N
zUP+z~PTUi8hFD0dkpD#*RY)L@ifqgVPTWHsVj`yfUiUDkGW*zYHptwGIcaG={2cR4
zg*ch}mxG0<L%ZJ|Nal2tlx(oS<Fl$?g&7?o{ZpdR-eN2dr4kAX9Nzc>x5)NPi2>fs
z_)X?q2U8t(@-<F$x<8CNlXrU0?ve(5Mp_4P%TX-r&wr46cr{*wN?*BNrH)E(e%Ak7
zzcaOUaH!+12))S6m+-hS3+U)bl8U|jZ{apKZpnx+OP2-So4<eS{H&p+rQ6lz+s>r^
zkcaH-L2yAci(O{250YBsSrUZviYw>*_`%m{{BfF0v<8L^LoIkh5|+(mQe(N(8sQ5&
zCA-6&Zk<>T!+-2O$H(<g7H1P3sK4wo)Kg@zYQ1zrC=EWFye!3oi?50|%D@>H^Ik%M
z%8E$Hyqk*;fM#`pp9)Z?_2_LPv)&{1klW=--Ke?BAE`0K3+a3mG6<4*FddvC?0Rx)
z7MPg-c6~^Of_oJ<P@-`~!r~I>NLy5xc;!3y@=R#1`)2TOG|^cr$UD3LiW${#(863(
zS26f`v@w<9Wfl}u?&Q;%c`VA%D<@!c=)jg&VFZl$AIWpt>*M;07xMg_9;CQ60iMmo
zOx8)4va(Jn;Mv?xDR90B#HDx%L85?A>XcH_S5Qbv4iTs)v=ZP18ykbQD<n`gg3rci
z0A9jg8>SOrWFYm|?%nMe_2OL-p=@C(GTTcl&r53;iS8Ez9cfcl&*5FE>khLFkG;4S
zGyF=O2hW*LU8e3BYJbAPqnZ<l!6!Z`SJ~%yLIYYpxavp;B!s2P$^7v702Jpc*WZ_2
zYwkoQZb&~z*c;T7ox#nJ&3M+_^Sr;`kG}Q3`vg4d*L~13Fj!S!sSWdG7jcm}m+k1<
z+Txe*$peD^6@k4gR)}AyE0)`iUal_gR&n*i#*JNGgP(PELrhmys&+nqZPn>{A@3MV
z-j)FDZ8BT=Z<X|TRoDQSg(Uv+?}#6g`y$?Qb4&Eom(iP%%oObt<tAKsQd3pCqTed-
z{fFn{TrPsP4RWPD+m^hz&||@nz@S`^UUgR?3F+vhfy(trkpFr$Oro5~yis_uFqp-p
zdlX`=gfBRM{cN_Tzuj%Wb4|)n(dCZSemW%w@4l6*$kN{<^Ljj^21!~=te?rftJJG#
zbIq}?j0}GqymG`p;r>~x4S4=-S4jQ_6r?csfZNGuQ_<y2J4Rh8=q1pI%(a#|wwE|k
zipoRQUQ4<Aq3Cgac+-}uyklT28W<38IUJf2>c$mTVY4{3WGxPcDhU7@;?2k@2wv>8
zyoSLW!Zi#^gx|DI?uZnhFUE_!2L1zSOkl-<Uw)0V=<`#D!NYZeI+RG{#ySd!RHX<z
zgk&BI6pl!Y<$_DT)EY$vhIrBF!FtzdF6Zy0Yf%nSLTnp9?Ed$O>#W#ao$2VeKKXjA
zkCJ@c_1Kh>x|4f#S|UMlL!E%?6^9x=xUv{td}KX~$XsRi+wf$A4cDxj`Wu}n1^a{{
zI)`LKvwqJz`PmbK>!ohXB6GU@8tcQD*$$zd5}zn+7`Z$Rn=snmF1ba$6Y`kSLro@&
z)q~h7t|UBYY)E$)35|bV(_w+?0k~N#%{n@D#oo}$9_^fv6$maDKf_w@M7_3jW;?-V
zI&9vy*0HTiS}^u;-RVlZLwqdilJ`Xy{vBs#<m+Rg+zk5z=`73hiNb-2N4`x{L-}zg
z!BIzuHKOo6(2v0~{A4(#`)(MX*Or4kKslBhU@Aj?3OYNk#_Oxd;mDdQ3Xvw$7pHSQ
z^am}~=T@$q!lyqHGKfRmHc&WyBKuVYwNjljF3N1pTN;|fKH)A!_*?WIoon2dezWV+
zmB4@IKIzE9jG;gL4M*ni3#~KM@}+dy6qwefbn(KtNC>FS63z!e#Ug1{?OX6q2D++!
zJ&7CNZZ94LdSnLclYt?EoM0%5VFAvIu(rV7AoWa*(79%U<KM6Cdsb;KxLK-c1(Wu1
zE0L!f5+SLax%%zsharvtJ&73}Z6cos$KO%T-#2Dwj-Pki?p^TDrb@Zitb2O|m0@`^
zEj8jYvOQ0$4a}#AZIMDX0}F3^47Peq#Ux4^ALu`6hNPuMl`|2?hDkN4aaM?;(@ISp
zE)?s$EMxK~9NiBhCKKj~nw(fTl<4H>Cqb8=04KhREi|Suuc*#qREoTgP`v}C20RCU
z98W^40j(i`!pO!(g;hoKL5oUGBHFUaik2DtQDBA5c?8i9T^IhH*3r?$TX}hI|FwL2
z|BbUsz0;V`4~SiE(;2xY`AzlTm&QOr|Gl`p>gpIX`^lr9t22V=!{UB|lW#8j#@3v_
zjvmN-OU2DaMxGy*Uw*r&UnSwe2Z3jBwt?v$-{p+w$2}<H!*ZkSXGk&p;PvJcj3Lw=
z>z4E{$0tnQT?0>=(Xb;w`<swGOZ@AtW>;dLjlIvbeHC&PD#+<mf61=NnA#rmJip_<
zq3D;#NRQc$v`j|w6iJX)V|aC-{02wIV#Q>qh?8MdZ;gueJa4Dx9Nev;hZsmIUY|ej
z$MVhIpO$J1gfvujm_;mD!8{AtlH<ZV?%-d`#5e^}E+lUKBpW|(B{!|9uJK2>=%whX
z%^LnTJ0-*JG^x9dND7vSJM5+RcRo3rsi`9uU$}<*uDN=K)$ElMMm|&7h3(b0w0vT=
z&3`&h%D}+1CxEjwbA!(KiL4&0h(w70xWk+7GdC-qXn~6GK-=iz98vJK;#+FJe=tl;
z>GCYLbV9ev$Y%gjpivL9O~<GmEap--<%T3N#c8AH9e4&R4(NGxH}U&O<pb}-^E&hD
zOBXY?k9Gx1U8T)pn$4@yR6LQaVEa#IkgWM0EyZmCk*y9tbDSLaKXEf3IeY73#StxU
zv{as!kJXwFr;l?fWDddWoV=(&5)#^&+rL8qQ^Z6!6-VEgEB2lL<>z7V^&_PFvA47!
zCq*vTx=&XoY)SoMRd*x3%X5At9dwYi^n4&UdZ_g$pUOT}*49n?LI)H<5kRlVaFsv}
zKD`sQ_F#be-gls10IEg|0}w63IVVPw_^eul0hMNbzVxo{<IPPplUeD>8*=)>*dv45
zyGt?!Sfy@6JbaI_gbBY5^&lDS;OyjS{8e$aSrqF-rhJKKzExjs23+Lhpt?a$!QhaE
zpi*wDlxNg8wRB5OlTaH;$*{zx*5&5h_P&A=Qg)tt-!G`YOiRlU*yMQOnTEu?EWsUB
z4~*ouQpC2;`B)Pi{G$J7dna+q&Sp|t*Iiiq_<&$3sLgNfg1+}_p)HZW|N56d9Hyac
z6&_B<)N)vq*(cab>wq?b*|F5zFf<2alI6i@>hMA47mOI|rKg`Amt8ZPsWdZDf*1=8
zo>J&q;LjAwZ^1ZiM|WU&1?3|Ik4qJ9<(hbv$li31FcJ`{Hw*o<cMc7!ySlk~BhS+!
z&3(Slqd#!bT-4(QtouDLG+t=^>#SPyfuDCaDU#k4f7=~dZNj?_b6q{MF}K}K;ZH}}
z-Vitk^PM;K-CIjh7{BX0o2fsp{>Z@8j}s(u?BdMVW~Bgu+m6pPi;*uPuRySsF!90%
zEJFk)(9{KQzf`ve!Um4vy4?{11j7+87n3$3+LmoxKxwFD^ic^f3tC>2dmkp(a-M3>
z>G_<4$VUvywCT2X6}??f7P4ZTx!iT4-P_o90cB8WL&pQcSG304b60JT*{~Wr6+~^0
zVa@*)8=!|+DBY{@XxBavBD5@5rt<Njkq45GqJkbdtqi`qA);jsSU-iqGap1>N{A^)
z*%=wU;kLs&hp=*5;b?G*op@6FSYGf<$kb*hCpod5xyQI1>r$Q-*C~=AD&k~g54aR|
z7w4Ve#6P*Frq-|0j?@?ce-Yc5%dcnJFz$0=i57e1jlfnl-2IrPc=7-%0DJ;kM%)Eg
z)2<CQS>D1j7YOmeRlhdY>R!r6s2A!6@PlLagQIEPtV{7i$pph42@}t5s?f~zQy%jF
z3HF2By415X?i-i8S$hz@NKm9J^gWgC*VQ`JkiBebr<Pf6#m%nF_DwNUMd3J|>){Ca
zALdHpuxt25X}ucjbAm#l8hyIN)#-92EccdF$=zwSOXg)te@&LhZ{=Wc*g<0l7zK?p
z>4zGaYL%0uxp&Wcmc!f_?SsK<8X3&3LCZeo{q4qRveKAMmG+44p-8TdeCgH9j}H1f
z$uEko&PF{bz?*9=byZeOgcP(rGT&n_xK-+rtk`q0<-uF~J@j;K2K0&__S*K-4htGf
zHAN<`ny;KCV=uuO8Ljb{`8~37yDlNi@#Rq|Q_uScAMQD9ljyeM+@!lkhD{F`+~8~`
z|9pO;b4&)fOKGL8q?5^CHcFOCw!>AAKaW*T+q^wDR%fJu{AOBY(`d{DLQR>SqO)pc
zHFRydqePP<FIfS4*Z%PO&TbtpqF~5aOOQapm@{`lb+HCM@L<@0K0{pDF2{zJV*t8h
z9*iGHJ$vJ?eCWZSVILP%ytc}F@$w}Shw7d;y6o`srSbe3S3x~@D$cy+RSUf5{^=-a
z@jec{S#fQl14sP_WQ?NE&jRw;^p%i%ydTba9+$+n!vjU^ChyH9a_=0@<}NGQN*(w4
zn?(y+*R)J3yv!s;^15q_J_h|`7-%~8(}tRAZ&AryIkPpbBez7qna3u@GJU*nJG+{H
zyyhzCADKp~lL!0;`}&MFL1;jj4g*eDbtykx9&)x&0dEiZB=xUq*7csNd<vFshuJ3H
z<sG|~Rw(%DvuXUZ!-2aZein($mb!YP*(<Pg2ij>g*Crj-;^&%bESOC8Td!4*;(Dfg
z#%S*XbpbQc<?^PU;O=`hN_GuWxP7AoYhJ}h^DdN6*`k%Ihq=-8;SKT6%3p;?%N9ZX
zwN})11(SFU^QJuch%e<yy#VWA?GFLhPl-+9gyOedSHd{6U!^1e%Co$MN4I-haB2(K
zOl(qTy_|X=mrHVFv@+RZ#3e%IiEl0Qk^KiI-P3tSPMRo&L|(3`wWsvI{_62=&}P+;
z%~u&>DNJX%P<lk4R34~ug18+hXP<;h+&*m_@F*&AUL&*|W`yH2wtcO>0u3BPTD|QG
zPxIA!YpUOSe$@}RxR~VdsD8%*LH<L$7j068KCi5}JiWwfj_*E&u_ad)trUK^YLVuW
zd2ef;jTsD_S&+qh+hWIt)z`A$d0kngVZ6ch_8sTGwcPVEm-0_#x!kofkrK1I@_u6C
z(~n+tXH)-+n)9k#e@0AdsIaVW*lkh~+;hSy4#uYT%@LtF`FUJtgjAo->RZdPbI-}8
zquhhJ@vP9d(ho~)xZL&lhuzqV6;z{2%e7%C&x!(5_x0U1xYP0`O)ANApO(Waq3@|D
z(MCmYm5DH&w)f9H*Fn=>B+wb>pR>5^Y2G$B-!s9IM-vA;Ic&212ex)}+@#z4Jaq_&
zEw-FXukITCTzr2C&Cbxvd8xCr#W(JEBiYuOaq&J))2-o&KJO>8CQHvt_B6XM{u$%<
zJ#*)Q*9(sdkyM^C*AH8Dj9BR7-F}`f7ua(y7w3!kf`f%S@Bc9NCE!r+ZU5RNO13Pe
z82cbf3MDd`vaeajS|Z7oDT!<eiL#9)OGx%LMA=)2nF&RatTB@#*=8y`|NA$c=RN1W
zuJ`@_|JUVup65E}RL}YSe&6qX-=FRF@$o@cZ~X}eJ%1>8T-oGiM-<8XEKBjr$;mEU
z5;1LUY%w+Y4sEj5Va<PR4quzl=8TexN^pR5U6Wb*?){oWOgYzODVK~r`NOZ;TIi6n
z_2YumlY{An%h}CEMIzA`7n|g<4)ZThM|^U4uN!zpJMLY4^5sn>JDxez%O{hMYU9L|
z*zQ@yj!0&AiEMH^(rB6U+AVTCdwtc{2(Yi8SBzsm*I-yX*@zCD(>SPX`n0?TH(amP
zM(FuRhE}{Hm9dZLvuPd2=<aSh9ZLz9P{fJn*4*m{cA!GtMAf!X#6KJ4i*9q87tq_;
zj`}3bd}7C>p{&DuE>!5buouTbfcwbcEW2!nJvm~Eqs{9O1m`jm!TZWqSWuxKxxE^N
zXC(bHBr|v1GlCHC-kA?^^hL$hVD;6$jfU?Yf?+Jg<z95U4_i#`uyv7nW3idD7CkF{
zP_uK;B3TSQyYqJ`TXh}VV)DjcyM6jck29SdXpXBMT&|8M#K_}qRh9;C7|fooz|)MK
zB4^%UoeT!cEn*2>==r^nBU@dTo?+vG|9O~N21}^ReXm@LTphVhJF06ar)$6*Y!7Lh
z3TnI5Z9Z?-?%@G1R99CQ=;H98)TF%%;h=>+!?K`OIA~8|bs;_l$#@&`8Tq=tegIxP
zBHvzEaoypybW!h7YkR|@%z=-aEp?v^9R&(R)MIsG+qnL`IZM%!;)OU9@z4?fe&uLz
zvfK0}c31r6taiat)u>%kbb?^2nGH5Jg_FvWBAUix+{s~~a?4^Z0CPGjRK`9HYnLIx
z+if>W!{4ye;T;gp+$)=Nq2dvfmd34WtNqm*nj6?hOUv{PIgSC;h;v2bF{)oh7j~I`
zsjh+}qV@DzM~)LpYJv)(LpAXUbc|16Kt;3@4sUg7{#J#Z-MLOl)d?rp4?goNtG60D
zgVI7dFuk1U<r?>LarB93(d{h}U7K_EBAkGd1Cb5od-H%qs12K!2LQ0q<LxUUV^*hk
z7J=f0p<aUZ$%c3UbuRFeT-7A-aK=x9F&@YP=ndh?F0L*eLDv6F{etL?ZpFlfv_;%c
zRW<elMKiCp3+xYI@=s)R(Dq-I{WWY-{oS_o`03QU_)$p{3%SpRD_g1l56kbpxkr*_
z`Zc7!=(Ie${zd+Z_iZZ~ZGjV92|n$_yP8gW<U?rMEA{Mnno7q<SWso_yq`kZyzKRR
zxiwGHLR@lI))YDk?G7^%kxvn)wci^)6i=Eswsr>|&*?pLa<|jVpqq5=)$$<P{A=D}
ziL}k<d9d(702X^QFz{_p4}k{}lYDFrjeWuylCO16L7y8s4zLD8>j6I?g#%|gXogD%
z%Vx>KnX)%M(j2~@MlG{yn6o@u3LS1>k?q!geJFwE$yhz1!BKMW4JN#UtyV<al7)o%
zyvI9Y(z|t(3x&;hsgp>H3?fU~Noqq!?>M5lhB?F=k8isFc?9!fkIU=qW(VsiIIY~X
ze#(u$Wt`pRpEI&I(lAeBBVw=(z5Z8n3JkSc)3WhRL1`<F^lDZbx_jvcZL)RDQ<J5w
z)Ld6W(j!kO>en&uB>IL#6D#h#sTD8rzAsM2UM};g8$WXrId|38?)Au^T7KR=ZVU@r
z$g70+2u}nL7}`X|5yRwalS0~-R2X?-Gz2eEc^^T$ug0C+WfsbTcW<eF@S<kBa2FcD
z;YK?{_A+2)LE;MQ=HK-6aH|Ea-v<f{<WGIxePzBD#h<<eTYca~g~oo@>H*>plEl8!
zC%O=l3C`IszxUn&Q~sw6COWT-=XY{!+)*#AlR5iDiZ$WK6t|V#(F{z$bMsO^^Wx6f
zORs-Y@{HdHlD4lNh)<;B@qxHaTFtW6sa9nkFypECUK927xVkGim8$6^SUbC$m6Qv>
zfGcmtH)|k_rjrc3ZKi}Jh3xYzReU$pGk>$P-fEPbX~6f~G1QZ9$~qF({voV7K&KzG
z{?I_m>W7!}?55T0*VgYJjmg&oKEJLmta)i-<9k>s2NPbW{10iz5PFFSNl>WL^P|Bo
zq|n#^sYnrmwYTlxX<ZEGN-_=T;KHBZYP2YOMkZRx4Odh6d}zd$rdiSGi|AR_CG&`K
z-wJb<Hq7E4u8Kvq=!<}wbkRA&AD5g}eYt0iXU91*)r9q_h>LHn4J?g=!ZF2%CXZdX
zy8eqIbcjNALaDRWnzy%#Y;qNXsI7xXE!}i0PPR^|i<chW7(L!>yQyS4U!?PFc4@=a
zrj)(u$yM?$lFf#<y}f)*&vfdxEM2FrekA+$TgoC#!F2V<OKn~`Kg9x_2WryGe5-6x
zk392aCq@RArsIQ7!c9_iHhl_<Y0~-&YT{ChzukzfkZm8}>JS)8adt$lLqJzi{aF^@
zH*TH0V$l}UH>~YkBgYf9j6uD3)8D^^Ve)%Jryl``gyH-E3WVloW<!s3)yufUg37T1
z1cf&$Owf(CxCk`P&NeU5`;ICbOglFfpU4asN)<<6v_2;>l06(G^!JUjoeE#aXrKu`
zDspZ3$E8#nu=xgam&&!>#i-4zXM?0P$8DJxzcJhEnsoN|B*=q)-dq_jVB$-&tyI@`
z;@L!W-EQY(mD8O8t$13kyO2fOJv_||OLE`POC@C3tgc&7Hk!;_yhHFVuzHdjMDN-#
zYw+}zcbj&v^gV*sH`O%W+C>NCj-rs+1uV4?4+q8olOwYOT?u(DASZ^zsJ<y;+Q3TQ
zQ?;ZH4etD5F+T0rBAhDrnH+vhS{0J=$=_Nj!Ynz0g5jt!<s^lQzw)()^y4-2xY;&Y
zMszmKl6ffpI+YyZ*i0Yorm=+!{IBT^)j3Wd`lxucllbkMpJlh^uGZ}Fj#LvOoJ&tG
zXssF4`@Co6``|UQnv2Rtfbm3WM_;-aF_qu;+^w|PgG{uOQor@pj^bEe`)d1Ie6Qq)
zX&reIo2BWO2iwy&TNQtbpMIE!7N78Mgiq%glN)(8tyvT<0o!9StxNQ9Wyy&0vB7xt
z%=x`f1&^tphzJn5WCD79*Uc)>L}&+&daN&hu9S7eNo+TdSh)wo+?`ja?Cc>Sp{?z#
zWs>+5pcBSCrM5C~3l%aSKBXRg5xl`*Ib<{%2)iiAPA$38&~;5f2zEM_um6lGo_VW%
z*{DfVk7g6_&~L;2W06nw_nj^W5(sHpJ666Yh*51_NDga1|0yL$9^HI2vsJo(`E7?h
z1ok@DADmmQ#D2ij=CM+eLiLyrlSC?VQS@RB)bd+gVK-H8cDhYvdPCzSueS6-RKq1a
zE!R6Eet5%cF?eIiXJdF<X`?NzFacA^g-(lV#Yp%@UMKiR;+xnpEd7Xjdgm6nEI(3!
z#-D9kVMkhu2WNl+4%4?k7&_$7MUy*xgFB8(R%}?9rI#|+K^wb?CQ%8o^fp73TvQM_
zUvMIKWOZSTUsluk17BRFtVtg2b%|Cn@4Y(c8})mR*5X(?MN*dDS6F{UeeY}zta#AH
z8AYE+SH(&S+Y>ta5=P9iznzlj4u<-|Z`eX>$SHA2O^9!8zK0bXp@=)7G<Rs@cFH|<
zw_cZ%ucS`xp<P$Im&O5A{j+LUWS7!ddYsTrkxukjRc9G3ho3fZG*j=(mFn{YkchGW
z-B}itQygD%YjFD?$06_#H|h_TtAJ`0e)M`0wgj~<X}vRQxwZ>F0z<W-;YwMEct{DB
zXHJ|O?xOnd4J@c*J4+bpDcYPPaFn|CabY&w{3OyI9(kNKsHpSBFijh#fe!il*MW2w
zQktB*dqG&vYV!JDyOpSv)B7y8XE{$zP3)pQ3c*7ga4oeEz}))IWioN%BQ7=Fza<j;
zgNO<J5lNxwUMVv%Xv!jDz8Bm{_tzlYGPZ+SU=nD!>sH{J*WA?$QJ(uymN3QS&!=`=
zqJ1O0*t947t1&Splq0@PQ6EXBS_y!?U!%|(B57`bkV3NvZ1-1X15`X-jjAN?&t%UX
zM=xv_Q2|H*6`5VV^<rFzY^Xy;nUv&#f=HUk4qOf&c)ig4Y$}!!B#Z*@uW#bEDCwxC
zh`<&v*&yE6Mi~vlsfI^5+iren^V~stX@NS{mFI6wc{ASkQ+U~`9Q(tRDPo-erA4iO
z@(bB`5{)Z0HY?OE-Oz)_<YSrG*?}5r->S%W?Vjs07gFFY{-T6g&ev+$Y>&JazNogb
zRQMNN%~#{cmiZLj^tAnLIKpy^9!n@kH&08jw%3C=p`DY;?7;a<Mbe`lV$Zx=Jk5fl
zEJt*$kI@TfSN`shAW~+k&i~9gVRbL&7tg;e>Ql^y5*vTi73ZJ29U|(Vt#meE_I7WP
zGjbLs;4Hq~^Sh@ELYzHsSj5vVv0UAJYArs`CW>G~ji=c!KT&_@IT(;#X;TTS@!-6=
z@*I>(mafd`?tJ^;!IZXdD=P^)xET3syHu8PzI0ixoJ!&Lv8$bYC7H)(|2_pn`Rnx;
zwOlt;ke(z@V@SHcWBtm_UHHn;g}bxzZb*?Dl@)7*3OzC5ta|d~1%TQaXgd)-u>(c1
z1WmCM4)GGiv>~hI7e<0&W}7z-fXx7z`JiIg4<+rErYpUP(bnK)aXG@WIpvt_cP^6(
z4$sh>^yJtumLI+cfSBprI^Kc49F`;7jXmZBCw1OTtUKyg>(7X6LktVoz1nQUu@zyK
z-)bxzAvXf+soZhc#tF5#A-!SeY=^aWGSy3vye2kmk2yuw-GcGw(zSy&?ZMbe17s~b
zq2$@Vp7I~k3v^r-g<<Q?s&momDDRs?6kiR&W~^V3nsEwGlop4X*`7QX65Sn|wEUj9
zJHa%J5Gqs=vce$CilU+^w&0p7QQp-$+wxN2@D>he!ltTxGv)P7LA0}U|1u*2W_Fj6
za6a8@aZ}K2*SguKq0eL1K4W{R;-{i~!f)5P#{W}ZCGfcc)hRg*lOH@-Nom=%vZo_f
z>0{E#0K5gxM&=x=TH{@m)5dMe`iMHYv=?gVbo%h!<2jNR@{&Tf-&mn{7mtL#++|_$
zl3@ZNM#jWr#ko8ECxsTHvUL2@qvn%M{>BLsl0wEyZC|&nlNQ_feveLzc&Ognz6oOj
zodfM{57>bnDq?pRBCiob``ibR*ZANJ@)}nbW|1AiKC{CB0u#~p^vSm(C$&{y@DN`*
zNWW?jlX3j291Y^uUatGCTRI{0gDXW72PYMdt2w++|B>X^Ui{&Boj3kmcst8sJq&LE
zOUNm?Yy)HS1~sg)BhB6J+n}DSnmKEyC!a>(-eFy;1BuunyrI^f`G$_M3go;;#+X#e
zx=#CmFj8c?f<Y+jXahV=YIa3g_SfusbLe4-9t-LytZDYW-$kw%Tb5UU5jn9z`<#%_
zT3gM)u}}9+%~^&(=Wy1{>~l@yjgxZB7Li+hy&j-01*O0Q8!Xoa9hk{M5Xpgdsb2|T
zED+&4;anH?5hwlks(mytQM+WQpf6BRFM<09V-c|!wD6rhDvzF}yU^LEn_k*8hAs2E
zid3Ch=Jj0l<<ltM*`v)>Z}$1*7o?+3tgZtC?I84BSd#C3OjVz{S931}5#aJJfP&pB
zfuIZN`F*nLg_h}_X`p$#moB@geDg)tb}{+SI|@7Ls_4I9w-=-{3$rL6&)~o-VEj2>
zw?cX?IJmDWCvd&{cD;N_bF3j<MiZpyPBfEvF_XQ#cVNTow%3O=d9Ws2zXR;^{3$(&
z*da_4v0Ps4BI@9+jJt<+_O)F4Fd|E)@Jx>RKvNvqCGuE>CB#$Bu|v22lvhs){i_@&
z3;L8;Ow`u4hNo`0q}5yGC^E)?7oFz7fq|196<fT#ww`)fe0^BI$FWWUPFFDee!-P?
z<<Ebsba>nP2JnZYoQ<S5d97~~?$K-w7J>^_<x`D&#*rPU2iTqO+^el<b;q>1AQP@f
zkt*C4*eoA&%i)qz7rmERkk88&UNJ?n<s=DUf$+!aw>{J{$++4UXIvyHpYAgE_`95z
z!O$(w6Z*AHv%jo<m6$(5b%ciCdkQYA@6hnI_-j_p*te&6=(>yP$59>&Z%2*5H$VU`
zAAr9XU&(xGO+-?rwyqDYZ(wI%brts;&ei(yh3A?oOFv0bp9x_*=C;d7&z|)0`1dd@
zf0-7_%j75!-T19M?N0%eG+$kLv%vS80!M6yt=}N$!E%#)@}Zr%1yJDg!M~OIQ_92w
zi!o8-G*Hdaiw8G?wz0B2+ZOrmD!^gU?U+OCZ;Z<1w^F4dfo6d9I{e`ZyNnWu`P@-=
zb45<7*xSjDjf7HC&O+2{#J<IBSQsxHJ^tiCfk$s+%3gh~50tPCsYuE#gN96c(3>PD
z?RzR4N{{%8fL-=$%&-Z~l5OHMse0I?wxNVnq&h|fAEkuP0a8uy=8>@aKp^Kl{8N;R
zhYxA(+)9$qb1@&Z#;~aD_Sz*gvnyhjj$U8VP=IVNz%ZKErah_eJm^a{ogQMHx?_HJ
z&op4pP;F^^yxQ0Y8Q6W^7jQ`p&7#rs6ve>jK6h&_z0@sAYyCOcnAaZfIcZ}vyKvLp
z&mt{4_8Uf}*5%xbOj@VwaPA(pX+Wr~KFc#vq2tUo@lM4%{axYOOv_V7AP9NT@`ku{
z@lDQzzGE}ycO*T&F@}x)7bxTHDoX`{j@yG&;~l7!ak9;61GU26M|#zazO_8E@bH_v
zH`&T}J&!}B?rDu7+PYn+X9rehFV8%cuPx(8xrkooxh8%XOx4(h<}BW-Hx;OR?nk}3
z_v}4mG-ZeJoAOd$Ach0e6n2vXGA6)73NktopaH?<0JZ1YF4XIYpALQAa4FVw4<?gy
z{ZFCe?_*gjW8=b%WV>N+oIN2SjI#68?}M3@vR@dt;Y<L_2mFy;I9277l3|DBB|Ur3
z!iya79P>2DIT)p}Sp=2;=?YO?-gF15c4W5v9&yZP$&u<WQpMj8yRq_hiyXiWyW>1u
za7hU*E#0#VPCBWJCD_Q*=@zZE2L6{icQbKKnCbTv&B&ob)9)*LMs)oev%(0@YDR8c
zRt%Y{gLVYKIRtJ1Y!+Y}8#rhFS(x=N852^-Z+*DH;N9D!FgO5OnbrXvt7^@JGf^~u
z-eqyq#@xK2w986&y+>r_5%sw}CbKJqe^ob=I|1Gv@%X?;kFph{58N|#ipSfTz>cea
zmk%G8byM8>&a5qGpbws0s{6aN!ApBW;u*RdMM{5RUoI?}X`0QVvXp)!#`8Cv#up;g
znji^%1H+~-SrOcOuiS_JlPa`*_rxACCSigWU1S83`ryY6`0Gt$%65b;P}KNC+oe1%
zv8nWX+d{52G!LxrL|$2nNp8~+<|edwxhB65#KtAWV4*~r@a{g+dfb8OOrRC!gFQ-i
zV(Pe<)Rx#^)Rn}{jFs-f?xnfJb6B-Tn^E6kt1p$Yy9C=b@_TUH@vR~)O7m5*Pift-
z$V$N&Z_%yL>IBb@9iqsY2++-gwkWW9H*gTC60l|~_!C!qX3uroWc2mv8@Ke0c#rTy
zkJ-~6bct_-&A9f{71)hVc3qK*M9^Wz!b!(eh|D+(jtv;_8N6a`gBr)3h*s|t<%v5`
zIfCGihuk7)ExRWE?k+DY##ddL@BiSm>KRCz{~nJw_UXNH;cB8z<E%#S-vb`$cBYnX
z;fKI4G?$8*o^Op-*V<-(-y`8Zxsc3sN!_cRO&;m1ZwCe=-X?qhlm~R^P8l1ie{weS
zn#mcqxU+rc1yZfgVCm_mHRW4teGo??+j9t%+NCn&|AGt|x8(0td{^AcmDfMDsZqLy
zopiV)lg;Mda=JXm{N~oijh*T7SGzmmi@3J(cAzAmAgkR7=(eWWi^tMHg$A7!Ow8NP
z#)%}0-Qx*qT`rp<7209}w2h{}c2v_q%bx?xobZ+YD`MYY7(w32QeOcRvA%obRF%0!
zZ95r}+MPGGt(l)DMfXl#?=>Mn?&R`Xlg;r*(Ke8e3jdWZas9{zTm*UPVmLq3DHmw?
zJ5RGzMvsF-8tUz<^TIL2go|XX|IA140C$PaUTQoOEvvyw7#5e~W3YP;Am7k&Wia7L
zEUT-lm+!jFvx5$y6TmK+nW@~wy(`nyfZviNs&6WA9L1c17^;}J_%`Pcc~}f}oS`S5
z85s{+P+!|U_k6uM=$xN5c0|L}@j>~#jm^l)gX-@%1l--xLr6Q*@!NDbV2?p5M|AU^
z<y(63_}p_M=nbUNs5P!~U%8dMNv^pocDk~u+2iw%pS!G6+x>L*B@&4v9+RU;uc6lT
zPGIGyqed-g9}6N$+MNf*8W$hHhFYhp4*P_`n7u^e=KynmUAzL5zo6P^`TZb&U#w?>
zEE!~QfI%yzLVkfm1tf_8wggCZ#sn(*qRx%bSOV-xa7l#~3GlQ4r8G19y3(X)2_J@?
zi5(*Z9?z=LMR%+-S+}fK3C3te`F^kz9053nc1=w~U@_G|Hu|owW;0ElzDWwT!g&XD
z6R|zVqBONo_B(L62+O5Rcx6C}X~&OUNaje??-|sw<Yj}_dz`rYJ(ohIGVg-l$&{RT
z<E|<ZU8N$%&)~kK=O_zB(xG*4xB(xNNaYJ&jw2-u6+h)DQ*iavif6|Spxr16@Go|o
z6F_%(V^G%sn11p>i<R79?<cS^kQw-T<KlAuVspT{kWOK<9$CluM$w5~d3ecb;eW2V
zJmc}T1il2NlV46}(>!8s*tQm3e;rJUPiSr!J3aq#>GBos;CZakuu~A5#*)~|3iR>D
zPL+7g51xHEH$S{7+<9ah)X7?D;Ad)LipP)tn{%`IkQMRdjH;A%Gdl7(nV6lLx$8h=
z;g4OS9*6F=A_^6mFfbW|mC19F8)I-Lq7hnxou>gD086b@B-}Sl?Py$Uoo@wfB!!)z
zDZ#SQrXs(F=&B&#@!et*alrTM=^WuG4S}WG0&Ve5{a^HNNYrrQRb5g^%@+<QHgd#Z
zk6VGy?)ku<UK0bz^<<og{=v-bUcfJ)87!Jw?j7B<{)h|HJ_%|H04}!Qjks8@+=<UJ
z_o8g?y(hrZavo-lEp1+u^ZWB~+?Zi52y=r{woB`yqCSaSS5_V-7cu1xvc+hT&{!@`
z6bH~t@Jf0n&vgb$GT)z5_E*_xYn~GtFJB+o=z3V5jhQ;QB9}rFH}svo_VumN?8?1=
zUWbGUJ)l}f-#4?MWOlJm_Z+X9J917Ywo*HG)ywGPx?<fuMt_M>^*k)Mo%Q(KZF0lV
zWt?@A*azdWwCQyrIq7R-fyfAKaIX)R2$XtLe^T&-CoyOrcUWiDQM@bi^MKL1ke$Ky
z2XTa`0{<tGnY%{p?Y{1l&^Jx7l#bU9Tbx)?gPvQk4b(7rX&>mH!W8vEd=GEwoEbJX
z4~~y@_f^PL?BR*bZr>iqL^-S@7^P_vKj90TFRZ8C4-{*j#BAlX5$T@`KY_|8IQ3g$
z>SkBq*8#!eN{k~A!9C2X(Z7H!aYU1Mc(DcZG_(Y7B@qEgRIQIrNzEg+=s=h-^WBK4
z*BzT|DA!w@?eu#fU+ggb)cP&2XwlHv=b<jCz!t!t*j}lZk=#-uJ!V5i(nekb=MW<N
z**7zo1Jmvlu~OqDxFt9}42E`<i~DEWc;1{YzT5nZZ=HMfSJ3<dTj8wLE$pT&=~Jm@
z{6~E`e+6#*xBBApGLUrZ*~;dij*}yvch@}f(H3<(igZlYlhWPJpH%Vn(etjkHhy3*
zD>wH?M&<e*uuysTY8c362|1Ssg<lV>OtWlCZ>ZA4;%VYXEzpwo?;TOzKv)3;#cieh
zvHiDtf>($A#7~N?*&*#Eg_Z}oa2yztK%ft`Nh10<RSx0MSW^y>Iq(Jxwi)qYho5yu
zwH1H5a&Sml-ZkQQOBSow;PD%ynreq}37p!&Oko!>MX){_+-+joU<!cbTPRK;Ob`rk
zEUh$GCThFAh&QxiWPt;;5I&&@PE0#B;<l~tCP;yR?GET5ac#_w9~jylHw3x6s_9o5
zJ9FLEe&1!70z4r=2vGe){5vVmf%Bg#WdRR^cL+K<5YefssX2hfGy0-$%<aBac$F#+
zqe~zYg#X_F8&48pI2#V42BbJF+Bd9RTdd#s*4(E4Xmy}!s=&ceQ3;uT2L3bsOcM2<
z++IP?zMA1mRN6-Ect4__Y0gTTU4JE?7_`tAXiVHuCasNos+4+MQY!7UVsjNlp_YiF
zXZ<}B|G-B2p~blmd1a;gG@x+kK})_Kq&B^W=eM8Xpf5FVd*(wse+_5WxF>9`D+!fh
ztA`&MJ<HbM+`g|LCH3U<Zf5T!!h#vSzVU?e-u#n|TT$C#5(9O+LdxjUk^r1+9gXy&
zC=R?FFm-L5D1J6=mFV%tw&YdC&zMdQVnDJ4ivIA7R&~U9$YQEcC~yGKc?v3uU-AQd
z0)F5uL}Y8_RYbLV3Gy?m?}OW?Fro`Pb|fCwL?N5Y!5fe6{I)&GmWI3Uzdu1z^8_#+
z0O6Eu5^1&C`Gs)+l$$}sPOi+7iWF=IT_?Qgab@U*0mVYdG{_(N<p*4*o$dl)d9Put
zwb5F)E8@R7v(mJ)_N!icE8S)BtE_hIcbV;Aady<_Oe`L@x?5;Ad*zcrx2lw+(u43B
z$)gLu1YyqnJey`<!u*?sWKVIdINb4|^;AvuZ6(p7zJlU(e~%OS3y;|+iJ!>Cv&p{y
zBUb`k?9<0<CyiK-3nfWN@)<8F6)pCUu2FY2T!2K&crgWSxDCMK0`ytGn}p|IrX~`h
z=~2p^O!7)kkLnXDE3_=suDfwk=}NU|7Y^F-cFEE$FY&dI3*=g*3dSB78-t6$)C1O@
zezzJ4d4N~pi~Q58fNtvIHW(gGtv>`@_qy#__!@~#P%WVm!`*n8?JwXUoWU^@nj_0R
z&Vn99d@(q0ALvU#&a-6>Z&xX+E3q<AvumIRu?y^dFt@t;$3-McNCdFOrmxlQ*qy05
zmN!Nj&9o}zc{@q?!_Eq}(Ud_m4Du~%Fdpu*GWnK@qy-m)@t(P#ILE1{VSAVW?Z&Be
zeU5s}Zhb}lUlpKxE<ilpigU2u@SK{Nd09e{7`5SX>1LxNj$k>o9EtE%QE4{z5l8Fl
z==Pt?eV}wuQb^7WmhCoOvIX#4r-2!|*n=b@?1>z3?Zp`|HwnIAQ{K;;de!z^tK_-)
zs#lYIlN9JH4@F<pG=6q$`6SC`G;^yL;@8}_0R0d#-z~=l{6M{2ab4P@;}sFHmnx?&
zBq+0sX^U*w>Y=37?nKN6jNBbp24+V=-xtzP8PW2(+K}A<&EX#lCzw}4;cnRx@sR{Y
zHKr71?u-dyD0DYrR6NexV&8+<1pS^+CmgU2tT#Z%2c%#I8G4#hlv;CC7k$hU&3qkh
z7I?cbze0uvOd-H8z#9e-S4B5n%a(WHc<Zm$=gfLQ7OChZIWdwt#^8K{IN_$B9|ns4
zO;t!kZV<YL0*Wx&?zx9GZi%bPtV@wv=U!U=@o#jRCn4>vOuu2;tL7ERJ*UPRrrArY
zRG2-=DRFG4?YnWJ==i{wmrJ`+m70$ld%2+N^GkxZ!Oj2ZS2Cfiu<E1c55hJojY@gs
zBppwXFRay3SqhP0D|`algl+m6?_WPY6Fp=&#x<*|16q{J<NK{Uy2ABLnnD%zo#I^p
zB6(ocP68$`M>g&dh6Obds~*mxox(w@Q5M@2p;!gBEE_qe?|y=9fcV25USX+q7zQj*
z5yN=#m|;_a@a|Daii4H|R0(haC^(gVx}yDR(|Nz$sL&s$jd{>$A7sChC|JiFjz1!$
z+|aCREGlX+P~`?op37bs<%n5htKXEPRY!11tzDg=u}U!z)Oj%<3FtXH+#|@iOF;Vp
zZz(1w1_)b7EqJdu?et~o!KS#SO`m`PHaKg;1|?_my0kdeu6#SNzmwOYzY@IKwsI(c
z{4(>p!|Z_|)s2o}@9zy}5UdjaN&8&cY;TcS7joU`A=iz?i_B2AGzV*=VLhW`_>GL@
z1SLX(%#dE<TRzHGs{WKU3#oprK<Q#(Yabrl`zUvz)W|r}<rRbT`Tqt>yq?C5k4R3a
zW&0h3L1DFJa=h1#uV-!_RxdvzQ+mkT7JAZcPH_p0qma#^z~*VdjcJ)+L+!|Mkrv;K
zHO^si24nb_oN?tp<0{<d_*$R|V75IVBm`Fk{ELI|1$G_e{>Ys=4!3eqk;%<0rZX%k
zdt)A~4hfp{s(!sZlgKG-UbZdl_N4hcrb<A5;+N?fJ;NGV-Mk>0AgGZB0N-NluCXdj
z5}ZR_vUPbra7f>jSutWN8bTPM-^?U%VB}5>8iJzfkB(w;atEky0RJ{Is*#QA+sQ|B
zS$t}0;_=$tj7iE~L9+PdRPQa<PyaBa?J~m_zWFx*U7Y9!41Z}WRi{wXWL-ou%IQSL
z@c|v(;lYQ#DoeT|bgzS+K|VUo&D9tzI<H#Gu*}Xtt?nK4JJ)o0P?DKFGxHr8z@A#<
z4F2xQQJEy}0n_8Y;}_M@`Hyz#5}?yurAXWLuJ_M<rG<|ugGL^5&yfKDx$+seJ~PZ%
zevh^IVK5T;>J`SZ*-eOzR(MGSrxGawPr}U1>2<!=qaQcwX;Xi262J(?Dzz*-qevpQ
zI4v?Rsg(F3l!L}g6pGRB(S!}|f^G056mq+vl#0z&4K7_ym7pGvWBD?9JqLDj*9c}I
zDB_{iybFi;+b_Bne}V-7Agk1lZ}*0oOPk~+c<mSy+$`K)BaYQpU<rV+L?zZAx&R2u
z13&q?&L&_kUka^H1vUMymvA@RMoN~fGqX9-drS8}kC=f`_-<K23$53apA1P#&FUC4
zOAkAm4ex))Mv~vBIPfKS#i5Y;wxcOYory}hXi;$55ASmC^U*|tPm;?+$dlxyQsSwU
zAy+%<i)**j8(d3G?@oOL9lW3lZmw#RI8~QJiQ4*e5{g-933$KYt~|6eB?lnq&{z`Z
z=G(DJkU-$CX0K~7b^SOZZDItd0;2`%>EA%GziJo3MGdZAoenU~f>ua}4c5@0_M#{q
zPhfRVJp_7ztq0&&Jxx;{C-+l3RCbw}`M_qKr7A#quI)2T0FX)<2sHNr>xn%0ECE6%
zRgNpegLFxS9I%pHS>$%kfL3(tCKw-}_8uvRBsWm3IFm$@=MOIkKD%danuE6AIA=Wq
zBG3#(1e$W_WUWg>#oBc#?f#eRb0Y!Tq(t|oB5`W6_`sK}w;5QEk8S}^4vhM1)WR+M
z|3JHeS(<&del)r#VM<mq)2`$9mxKg)QEdo+2h)>dYGrI<SQh9jvZdk+LCFADus65~
zRi?Huf%+H32axaj=S<HofYT{kDrCY}p(7u38{02MW`m`UX>^MshGi32n~`=w#OIga
zNvK+|6$s=E1Sz!lqnJDwIJS8!Y$`ahb?YQ?tQY(lZ4hzrhp$>#fk6tE4>IzO{&dnX
z==UhjLz@9FQ<w#M_%3&7r^R5Xy}FMH5Hg0m&eIb_5rEWE8u#J_e9ZT$dxtM_2bp+V
zbe_4rdE3atisvtTzw7-Ap>FuPCf1$cvx5v(uEC-nEp34@i?6)-^@@R!+8Kl1s-fNm
z_Oi4LoBtJr&bu&4#I_Ubqk18z_aFG#I5mCs_4Cf)a$5stJGpm3KX$VI)iL0NaNkIz
z7|2(0eylX{1j45jFZT84UE~=To`f3!$}~isK%xYEc_0H}uj9H?0SQxM%5WzlPYky=
z-1wI+Og}or-5)td25KI7Jg-lpxegrT4wOG7wg4HN<Vwdn=fIU3tAo(xO~XfjR)pdJ
z(Xu#X<K3S)^s*@^Tp-;8Wm?8Si>xfH>wx}!n0wC=Qm%^*0+-!@=tX}I%8J@`+cwaw
zF>aUJ(9ngP%IhbPLBaa|s<wc=7`4*6Q2t+S8tv`EC%MD0pJ|vU$>$y^(Zn_$pTwQm
z_h7bTm$Z`G?_VRk-CCwVk6d%jPbf0+e;?0j@lRq7S_JsfJ~3l|xEGv?v*{0zkkvPk
z%cc>El^c$k!skj>fJ~>*3onwgA*dYK6;Kw6e)T<)fqbv}@V!Fmm5PKT2I+f@IT0DQ
zA%yh78Auv_scUsPHBZ5;cF5z1KK%RQA#Z?O`QGEXQ+B168SfTr-;SGm>&nPHhH*JR
zl-a=?@+Z$1(s?iNt4cx>#~HEJ5e{~0Wl<5{NDF{mh+uZ8fLwNhx;jjK5Q}*H(HImT
zc$WaHv26B@ww6n#`Q3&IoDDF7i{<dgTu^ohH<%9yJoRGHjokGD0pNo-mJK&veD|;k
z-z<d_==wgjW>@t0_>r(hJ+AKc+Sbvh_GvzDbi3(GOjoHcq)%F;6#rKaq6c7)i8uMy
zwJju_zBJ`(-4UzD**ra<-P-nZ{Kqf*yD)O^q*BOV)Cm#@n(K>UySulxGqC8IA6`IZ
zN`0S+NmGJyGyr`(Hvo{nRF#s>6;Nwarr!^B7ZE~$cb*3ecNhqiii${GY7VE89<Jlw
zi9s+i1n2=Fb-l5!ViIUq$x2z+6bI61aK*z&0SyvJL>bkMEO4{^IVUaMCDnQdY`t$s
zBBkMy*an&)m5|j6{E<?-okcig;qa+EG^Cw^9}U9KSARcmPHY?fbCC<PP`I$hVdc>r
znD0^9QlO23;^gRb*hm%W+_&~9LZlDao@G(aQ0IY0w7I{TO70mE$^6{6U{(0Hg64uB
zzGpP>i-W02XMIiD$ke_m7WL=d9ahwyqE+V4wWzA;2Q`wBbuB+3A(`U&{Q8TL{N5Yp
zGUAZXTVZI(n;-+~>!N~m${?mYYUCkMg~tX~vBb54v~+))-3HU+-#8utvX^ghZS{-#
zz6SMb8L_(@J^)^r?gayduc%!PJKpLve5Ij^oG9st6jAtFyGd}_5-&qHtGE}@VlZMg
zk=1DM_>z!);Y;ryaA1Z-FlxQt`vKX6zYGT$xgwZ3@2KCLUUcWd#%V)#Cj+mDg7K&w
zxI2pSG4=A|B5Bj9{cUkB?eI3>7s1p4F9ez*@VFh?d751W(pbYU2<w|d6dipZ=-!2y
z6(TjA;*MoHrR;3xbomKS`1WYFpvnWG6I%wY-B%@OMKh%S2<L&;bmOVD1ONFxp<mrK
zliygdkDQl~ZvR`Q#?#8AQxZ*nkO+A^7X#i<k6B45*`X`l%%i22r;kVJ8+sIhvnfRe
z)DPm4nFGJQylu`J$d#<IT{iH;husm8(=rU~AR^E4%NO^Km$TPr4Wv)^a=wn9@CCDg
z9L6X*TPxb1ujd9Q@+?25tlp8E4@k{pxZjClLKOgr0)CyJF9`z}=#$#JG(|3y!U9fY
z{kvR#Ox`m+7GK>6R|tba`coKW8EIbB(OzMeqM1t<;ai8tSt*;61HNkro^y)KhkZ%<
zgo(XC1hK?9l)LN&hc9%f2}nGrU|ZaVoERw09bayljRNrE*wVxq5e?|nQZTbCYW^cI
zV~0T5+j%-^d}6R?NqIH6dBJ(32#)&O)+Bu6h|<6A(boaL^l4?&O^#0P<0E@Y*r3#Z
zgOPWY*>d1&|Bao$*o{>PXcFL?UG5V<)p%qih0OB)QnNYt4L(_!cH%?=D1AVJU}0bw
zcnhL;QuP4|wVyKbrT#cFYNxV96_U^U@pz{|@Q9FFx=nP6=m9a%4M1gOaJ*2`fffL>
z#tzX2PD7+=l8D`Smn{|PuT``3^xZ<h(t9@;nYNN)0EJ%z=Mu(HyK+hH(<dH9C%kRp
z=}yfvTf72IkRL$SknInDk05#WNgU7ao2se@7=e@^n+7dT2tL1u;oSmi<q!;b7@ZbH
zeHh9hG)XyH)RfS=WU$l3hH)Qsw!k}fRKg_z<n3#Sl?6%R0aKWnii%e6-;&4--~Avt
zVULs%yE^zNDm(upcx3dNUjvUv>bKp<%t|-3G_N?q^I~ri>=ecCJ7hhgTtAjyEswAK
zeD%tWvA4wxiQMtOZ%mNiG*>v7;c1-;(6}dmw8=%pabX`3Mmj{o@bXt9(LRXENkB>2
z+Zw1fY6ebc?vg^=hiGyAR5jEgWlU&aIIGttp7&h@sqQX&T~aZ|JYw1~E`fKQsiB=b
zcI3UkTCxM+?{64r$sb!5fK-51K@$M1MV>SpuD?<?(l<2L0$L_O1q_D3@#O`{9H0$_
z-W%F6^!~Z!`2qN#x4i%;fPn^xb{&SUl|ACZYe7_K0&MbuYR_PERAho(QeDSrP!5xv
zplQN~^TJb|if~1CwZq%&%2WppF65q(Z$Wq;UU+z$zo0+p5c%j1Hy;(D`xa6Q{@V?D
zJjM3*%X&wuEuMBMoA#+tK^vA9EIp`3OH90*DG#9);uWseW%*b0^n`4q!gFm)M#B!o
zjV`B&$Go5x{$JU97)WWRWlH2szeDn-@N84b-rr)NL&fAsP4p0KL5fguia$JJ{Grky
z_;fhBZ?;UNAe`2{)kZC|->SQdknj;`H=%eyU%h1HuDBOw6}S!nt#M9e63$19hsO3r
z7ZpK~Vgf7y_7q%&jj8ZV;a9+~Lm!B^DAxY~GK`8`Axujd(QG39zR58Jhoq^(kYsVn
zn1|sWg?l5;Nq~_(4Gr3X=dCZ_5*<a#cUQ`eubE<4h^X!fK*TaKGv`62{q-sf!?p&^
zL`)jr9AK%>j&8B5EUA;>gXTBk2fu#p@Um^Gp9M(yziFt{-n(l5C~VHXwNJmzyv<|d
zx4R_TdEmyvw=*l1uM}!uReH0n7VNUP`;+oNYqJHqClh-J<iskVI;MY8K<jyDm#sZm
ze@sKJv)g^MJ$bVB#gpka6iiH7@s9c?ydjkcX=^0Gu26jsX`{;b#L>2#VkHB61GA~O
zt*W~E_{@w!UhX|?1Z#~uy^0G((HD-sh53SH$^ay?umV2_1m>ln&4rdeg7e*X_|V(`
z+@X`#N0p=XjCqM4;Yg4)(lc9B$cp;V;Da9^<^zcI#&19)1qvwOD*!n7T343-Ku+H@
z2izbB>4I7;zEN;~Mi!`UG7cPzka>9i5EE198E*v1n%gb3kpp*WXo%+{$*F+-=l&(@
z8noc6kkR%%g7r<#@{f15=Cw#pP+uc#*mU$h$z5<O&|6P;_f#C+w=Wt`6WC*GG&1jG
zpl8_h4`GB=sdYq{A_@3}%e}395AW%GPHr*)UCH{Jw9IrHA)MsS(`c!#iEF-3+7|-Q
zQy}Kf*PhV#M(R?-4d{nvS7?t@A6~%Z<m_02N?TAQ&3#BkAC00MCGeHO$AfIxIV-`$
zrUAztjgA_Z=s(k=m?9!6hU`!Bcn<IjbV@FJ#SsmHz&v!`;xY_jm2W3f%wVMq@DH$O
zV}6P548`tm+JR*=q@sc9{f>}wF%yb%P=q-ImUECR5tmw8M{kCTOX}NOQhc=eJzAuf
z<ZB+=>NA}B7WV$<j3YYZ7n!UV?WN|T&yOB72GL^E%J$}S>qGV{t)D#o?QBhk<FXnq
z-f_a;aguy%2($utj@GD@Yu}+$S953;`R$0p-1uK;9xZP?&7!GPYIaWo+n1Y8tfUdE
z`eN%v_0^?~zjBTfkGrIL<VpdDu=7pzU84U}#_;}JGsy`)Pg46!-MXlvA{Y|A1a+5y
zlm{J1V6$P$eGb}p&Nis$x#wk=T|d_x+y;OV<&=e;p~ppjW<gN(62Qs!r<n&3$|Y6r
zA+QGVoS8%3YgFTM#!$*G3=JrBMtJ7ie^TZ^EwFtOxk&)(5M^UR%bK2(2yHkD?Qpyp
zP;GxYVu&@OtG0OU9GyCxoZhuTTgz<Qm^54$_*Pu-z04+b^C`1nMeeS!sauAn|50(L
zBTrZ7Yiz-4siu?h?_atkWqy@M#M!E$eFg7-QvOBNyaPKY;fngj=`G;z`@b?ZXm#rH
zL)!P-PkadgoanC3Zmb4{E|ks?y$R(d=EX-5O2M~E>;7d)q$c^2k^3-%V={}$5o5(5
zH)#_9R1Vaj!E9i+AIT5nQ6WcbrM;8fG=YWaI;dPyQnRo(1}@2;evQ?+&`rSyD-|_u
zfC$BCZ8!#cy7SGc;!jXw;35Dh5Mf?o+5l$(c!e;b5%*&C=TAU-Ik9m(^9Z&bFSbLq
z3rZm2t1{-?x1+DAs?Yout%JU*zp~%p3u=bvcYky4?<2Ln-y<11z&m(6_tz%r^ou45
z59G<0vhl-xrH*~)Q0Q4L<KcM1(Gt&hcPjDwy9wo}KYdZU?YEsO3;#`9+ZAh24(7<R
z^Z>nD*L+C}NU?D%@Q+*lT&Xj&F-6_|(dZ>}31dfe`0ho+=9LA5vW<hHax-+Lp*M#T
zOdm$IE(EAF`exLP2n!>R5}n$>&)arbwac22@I)#;iXq(j!#csY1Z9R1Vu0hr9syYp
z(U6Q-P%CpRp{{^3!gS~e7Qdfi4swjF{2b_z6~)gZ+6k2nQ>5V^$ftk@*q$K@1&j*7
z`=_$H?A^vZ+q?uJ6Aqg8TZMp58^(*#0`EQS=%K5^90XgGVknsPDxV<spat+rEanq}
zti3JdMyI%Ela)z+k50J$VlPF^rbg_mn%`L$ObexI%j-j_wwvkNs8>#2X4*^R>PuZp
z$6lwVHwYYQB#(7^O63BwG;<sbVt?eWhfmh#A??vF!hhL8fKA^SSzu(KR^71KY3jbk
zB|p3^#qrg~_ct3qD9762RF~ySMLAVi3P)y_rv24W2;U75*Ws~Tn#d##ymjPu2WHm8
zf4Ksf4>J(M5kz!(Y+VX_%t4xmhb_VpqM;4U02GZYBfSy`n*nJa9~YG%+-oR%SLrr!
z^qcjW93y~6;Oc}U{ObJEq@0qe<6V1THa!9aZWMon6j$;=@@6I#vQFUJJ;*_b2=zTo
z$Y*)y1=zmm*U<+05<nd1wS7;1Fav>zcA`-+81KEq6}j^mLr!spsPIi0kf6Q_!0{BY
zW;Jbml25eXSmm=gdCT^Mw#xI{3KbV<AvBaRPKVV0e7FxZJ9n+K)BXLZV7qwToxR)q
z-0-cL!PMdeg8b#m*yE9a$jMtNfg<jJji~`ICDSw!4Zxg!|M4RLb)YtauYsuzfm`X-
zg*8YxP6j}TsL)ZTSo738h!O&@3~|2^f$*qtzo?MXUZ_hqeaNf<IyEV7EF>g2;o!D}
zzI?444(cHYy=5>#Z`amNLOv7dQU%)-w85JJp=D-6pT59Q&QQW>fVc7aK;i@}G1$f^
zK#KZLN?@)$ngfzHFl^P@i0GR_?}@Ef)c=FtUEqhl5_Dg8yn!0$f~^Ll^^C=neU}z_
z?$ouGm{G2S785+L0zY~a>^9dGb(xes5VZV{Wb~@KZK8ksteS@JwGUYhB~MCjy(qUe
zNtz`3)yPTAawXFDooN-ABqGL#?`Ln!wTVJQ2_mR}F37z}po#^j%VnrT$^VepyAlYu
z%ehAsaRzGTEo^$mYWq>2BSLqhq@r4yfnCkrJF=j%M;<wGEOu(A#kEC(4C1%K`GT?@
zt&d=yQ84BJCeMJ9b3e+eS%6rSyAQ6fqo~jUn1AZ)Pe6@;pFgNmIEwqW67=gNBF1P2
zwhOxN%X29cCwUEUf`JI;ZP;ZrGV8gP22H&zTl6?!rr^PV5~6bcF3{4HQem$6Jy<~N
z;_s25_o{IRunVqSCi%8Afxk~aGJLb!%;)P?_Sq3W(~{%*wdS);$j7;xMAQ#mDdM8V
z)A9-$<Yb!LiFK7TUn8kS_>NCv)R+y*Z2N=wX{F~%+nj1|S9){%ubd1F`=@#2OH?s7
z`d8b7=w$u?aW;N5mDvrp)Dw@xJdd7h(y|L3Oh*l@Y2*g)vb0bUI3G!?jX>XMh0yuj
ztRHn{Jg$HsWlI&+)guoSRV@L%pnTMJHyDTEfr5okcPX}mp_EX&3Ws@b=y@l8W<ZdD
z_vQ}=n({U1RsoZDRJffjl~93?piF{I3^p(%)p-uo=%sQaN-;mwYvJHzO)B7@H{LOJ
z!9IcA4u3N0dL}`VfMXv8|24kmwwU(KGO^GGtCqt@z)30z8*>4|hBlm^8L-#GbDIS_
zUfTZ{8jqQ123zfR<@LHV#&%qM<x9~8pUf=&@nOHnl)eBhBy1uYdS#X+GoV)X6F;)?
zJ(*M<v>sLbsIClLA8090kM!zYx66sdL^82{3QCC0DfLdf=FdMVMU?*w9a>hEDHrsh
zKY7xFwfsddxOB#|v_(Uv?m+aNA~wm179KlQ2Z8`^+r_5lBwj5z`WzT`f|CHi3J42o
z3G73d3R&ULybTM#00(oi`ZOdF8XLulNz_`0EbI&X$nGqlLjfwUN^e>^D4KH=Wf$>r
z@=rz{JYo1BQ2pDyYDfU66!JN6WFz6Y+hCef)US=$3aPuW2?d0CDo;N!Huu57R*v>n
zPvW3m;p}=6nvT?Un4w^RSB|~|b3a@f<jUby?>FKgKKy8DqV}uqOo@Hdlv%M%*}pDR
zt609oFz1y|t+vSoAC<apa_gS9J{u}`Pg@Yp=cC&RyJc%_l2r{#R#aZvLP9dAKKIUs
zm8r_oqCzW)SikY#w2vX3JR)cLdViz;7C_b!iEY2K*zuK`{cBSpSITcrzxlPgJ{O=-
z-MF77`W_pJ=MRUyAukRe*m_%UGvoZD+N!KdAqx}sFWh<|5xNCBnj-0LMeumvxxml!
zo(1&)TqjFrecqt*N#PirUFc(WjzG_ha4n5DFno>C<FGageoEZSK_drO;8E2sPN+SG
zEfrSjD-3BVbVbMy@Y3iYhT|3JYY1l=2cjO7V4aG9W}6o$3R2)jm+v}EPGDHf;5rFp
z=RI=oa}A97wxAa1$p5;1m2YD?(_-ksuFm=l<2kU8{dTD^x3K<?-IIC2)cv!b>H7C^
zY5AE_d?+ER++gODSX#F@HFJ&7{N%i~&aoS^w2`XveP`U-{Ua+wWUX(_Dq#$UD=-Wh
zW4f%Q&`25`^!9n#KPk(Q82Bt`Q-`);*;S*KZIeXEp=i$HF`du&5@o&8z;$;D94=B>
zZ=Z*0J?HA#9?1jiK0sve<OfMh5P+=If{g2!PEG$_ye%Kl<|aD4-M4QZQi8xqLG}+G
z&jGx}Dq`A;ANNcmIH>9YvL<z8^_vpZ6a|;{Qjx#MAtb04Kqb>-yTFl$e5*=hmIvb#
zS`UL#1*XYAaNj6^?;yz4xd}kCw?#4JQfy`#ju8)SxB@SeKOA&-unh&dY=GS8=-A~y
z(sreK%xezw{nwA5v?&49(Zu<(f={2!ksAPf(-N-(%udikDp{%6<L2!3wP~+utiiw6
zEt>U$*68*&)VI!s@iaaST*O<4Ef4oa4GQ<y)E?wYus`-aJbkAPy7k$4kmxF47rtgI
zNajh}g9PhuAFcGccYrLt`-9OHm(aHBcnZCZDp?z=?vk0GJ`O&M?Prg$7A#Gp^7h#y
zSQt~r1Z>gN(g2ak%Y`+iwU-Z(OP0bFS|F_E*>ULz4Aa{qA)Og4Cv|0lQpXO&FmwmJ
zz~IJj*>)78WeVzE_(&D^B4HlTF#tsYMTUvu?Hmuw9e$XRS-_mdSpAF^k3w4f8ut_(
zedrr%Bk1R$v2Pwp&ivKtCNp+z11i*!p|A^D8ZmGQKugNdszUe@oc<+$Ip;;5s?xct
zIj2FK`)Z58nBl5+SfdopE~%jS#kGs2GtU_5Hd5OIMZCjS3X`N<qc)on^U<E<cIMt(
zYKJ0qtA&oK+zD;pp~2z6$9wJK39V<kkM)iu5S+4U-kPue1=Yn^R<KZ?bD-)$2_64m
zZ524P?B7#%92C5gvja%cy_G-6uU0|~4?77RBqhuRnHR^s>^P<RbC2m2EaFRt=heUo
zdjx<&1l|Z4wxQ6lF`OVk%&VM|Wd!X5sTBTLP>`45KQ~PWq7Cp$h#dw1v)hu9wFL(o
z+;(@tiNOdihRqITQxHi2A^pnH#WNm-5w|5BDrDLIgl7O8f9KVC{qsPy(){2bh3y1~
z__dZSk*St>)am><JmWhz%!@x2tIKm?h9!VQuC^lL@_?o8+v1S7XLvh092$Z>I;7T_
zW)VVy>#xot#3uZQ<S070q%N`TGg@lmPAEMD!f#}^gd$bz`_b+MT|z>0jub2<@Wkpk
z+XoFNB*Q<ENJSWrqH!yLITsZGVm2&VyARnxDqa(bX$Slr{N}ekLTg6Ll}wwGcxi3d
zF&_|}kM{vL*bX_RK&6zYZy4v@gWz{HV-qztFK#V86#v#&sadiIxZ1q*5RezXRlSy%
zPnDyo6LMe&;|hU(7IM{K`BJ=CWl6Raq^sbP=3fb-aw;M;DlsIkJIBCM>LhT^4E7!0
z>WKL1*rN!MUeGUiczSyI_}pbgy-I;)yA<vMU)%rSAvrmc%ts`&D<!_21b=I-P17=6
z6h(1IF*`^_hJfS>sH9-5mn_en<#s191&%ha_oz4Fi^YfENWU~G`?SC2Up>{gh9AP%
zrSIJy;hQ{3r1~Upr3EZVnc?Io_!(PV0vC+YaAZQ~?}0D!`%vU_Zkp$uw{09m%t}Ri
z9|Dx?-wEJnWRd``UwLu7^M9H>Yw-$C(NcW2D^%A^i?N2|fyeyqPw9HHp~#ec{)GU0
z*c>6&EERc^JDu?WF|;bnh`U!EKr}A}ix;FKb4nz))<D5+^O8byWA;AKGC9D07rrJh
z!HaNB&{b2UX6t6s_=St5SONKP9{@-gNTAu_#y2nmz;gr<A0U~KjF`I<o?&jpxivV#
z06bx-AYfo(Y|jQN<YC@DfQbQ9X|x$GhO3mfWsS=#MpI0h-ha*ulpik~T&W}Vbva>H
zL%9EnGI~+IH&|qS0D7ayc3*j=%%TQebAx8t4nl&GNkX3{)orcS^Q^9Oe^T|M#*dHP
zWSvdzRhF(6!e;PaN2f%<F*J6xooMZ9Y#@g-7`Cf|gFNq=XqJJeRs4cWWfZhZAIAnX
z5Nur}f9HKf5!QaU1!KZ%>I$C&9N;mI-Cuzl&U^#^p{)VisrtMDxn;1uAlxmQIR@H9
ztX)f`EXX`WqsOZ?V2v0^=9cueYN7S1l!em`nqKH??zfGX@;LP<7SBk-ZyV3C1+-!?
zk=yP2c$uFWP$l?bEF=cl!(@={4F{wi>YLUgMj-8YO|>Uro`*VyP|y(S0svEE!?Sm<
z2HWN@ItSk>elVhGM?7fj8I-;^&!*T%hhY)fM$LP*_K_*!(TYoQK|@ML1Np;HLSpg@
zlT>`aHVJ>xv3~qXGb;&~WN5?dr0Hus(wbiM1f<q#`JLSqim16PX6c}2dzqF&Q_eSO
zDupdoHYXbH%)iNBmyq;^=J1OY%0+ex;BRJHi;y3&t$Mizrf>t)3mZA=vvG_bZ?>Lb
zo2zLwW8RE)09{qa1Xy^`r7F=7XqDxXy4q+7rQsgSlBF>Zv?kC;LLxf=#h}Ie3jGXJ
zd4M+aNX=(}phPb*@^HXxXhb4{`hb<>sqq<_)aUohrtrcs{8>^t2;d13JLPO&rhlhC
zOWvK#5po2GQUR=SY+hmd1t*5c4>H6VvCWjj1UzlhUoi0H$Vyd^naOnDuuz)}soF39
zBO;kWa?=J4cb&qm#BWSrx;iH(v=IAH_Vpd?l9+|(IKroWsGNRqzU&~brpCAu<kD`E
z=*)LVy9ta8oNr+3(5}v`_<dUeiNI-X<?e}+|4R>rzkb0z)1)R=J&!vw33l6*(9erf
z^KC%Mo@syUFdX0Phj4tea{SqSwSoTI3br)ZXcMMKK?sXlMhB!LK~?~irq3JnDhOtq
zl7-FznGi(!o#zkd*^4|+w@4B&?eGKQjIt?MSOa@hb6l3cx5zM7HNou%<Rb)1A;bk-
zfY7^pJcm>exEN#TP)Q;=;~ycq>C<`QAan%PZ!PQ<gL`w}7y#?JImvm2f*<LrodkIH
zQOZrNUb2~HwaAdm3^0_(^9!jG>Y|y%{I92X#N!(mu>Xyb_>?|`r+wR7Z<0WNP0P!X
zRUOj5pkjY^8G4~}B&gS>CgxS{uD1fVl#q#cw^AE&{JEfvFq{~ja@WGsZ*2nuAQz}+
zCZ65$uX5JP2b!`>#Oy}nV08}xsP)~)kO?y>eP_Ca)a;4iW^+(=#l_Qz&{n{${ppH$
zJ|CZGk1etc86+Lh4}mI;bc%7*<2kcI;s$>e`a0N7fj`3r+kK0qAr};YGbAKZH*W__
z;5`7^0*5|6Z`t`>H1BDj_kQStAh$~(u*KWOPUmuJLar36U>hRo067>^WrBiY?%v(*
zzcBQ7@)@*_35d*V7bwkFWq}F)19>t$HK(9I+G0VgDgwo@Fbk-ieMewKl5Q8;-CB|f
zFBJYPc&7;4A*K=&b7%LqUh`@1NRf@^6~CwE7RHR${cqy(3KL0b!BmanlbMB}#Pi<v
z!B?+|QRPn#RUfuH)*h&$A|%-;ud(!3bBd{V%hZ3EOFhG~Ld0S51_^<al1!GBoVBa^
z>&cZN3Pps(mzvSw_0t2`4ctkAnUZ6HAWr&DL@rQxz`z!WJ2BQy!cq#3t$4t~&3uv>
zCFhf<<}*gXI>e<@?(;dYPAKXFRUH22v7R|uaz|m2)qF7slg2sl6&Ltnx&c@YP$k+!
z=Tr%B<neGrzmkpbX`!g5=4op$&CCFL@h6br>znNjw-(r;96^>Is|ySW!v|n-6SAV^
zo_Vpjvk|nba5Iw?QWT_ThRztD^L`FbUWC?FqFE6A&>a@?fac-Z!1Nh=wsRe`zh~a}
z?z~CN?OOd_Jo4#pIkTQ2<02S^efMv_oWN2~qmgWeynLDO((XAfDcU55C21*V-0<6D
zO#J<Ij}O|dUP}L&SKuFO+Nw_i0fJrV49?Q(=c`B(w`o`Df61A(phe2|CJ<t#UKPg^
zQ0$|%ijOxho4ott^pRdo=>i=+uZT8Ibt`+wdG}@6E;{ASg}7nxjKIGlVjcwf(jP}k
zTSfXL<nY4u4%=)I*&QY|54s@m8}@ns<%fHo2^WIkpUx7LwOOzdFTPC)F5ju+z(S<v
zx%PZf+5-x&fn$ItgBEAF@z&kH!#YfiJf080$0~Re&WEQU))@XF5HB($QJd%RL&71F
zExy;lqO$YNnM~0gC{Lj4PDF};K@+HDgzbYH9FB>^JG!!uz~3|*<z4r$?%89aQm#Gq
zU#^EL@&W90XQgnHl&ihL!Ib2v=ES0gTDNeP;#xo5$M5nrr}llt8YR8GQRCq`q4e^x
z81>G+F+nf-IfhVi-l+_vvgcL=T!o~1p8lkWfLjPVB`?{WaF7x+8JbNc6HiR#q&IA*
z-dEXFq-`h%DFyz5CMIgzy^dJRUf|c!#_8qn!*YU>&T|?0PD5>OGwhA=VsS)xHW!Yi
zX@7-PqtO2WR0%v0v^6l5LskVBQdTyh#3AU<l;CZf+*dF*(t7=O(`+-Vi-q9xl{GbZ
zl*k1OXCEpWcqbTULx7E9rr-rKonrrs(*)pyxd3~18$|@t8=s%&gt#)Y0?>b5q7Bb=
zSzK5d0}*I-8RBnX=#bqCkk=Hvi(FYWbLHaF&7>Tgt|=}m`TubCCg4==>l?6AQmIr@
zN&^wgSfWr7rOd-Jlrrp!q>`+XA)%r~gJB`FEkmg&g|MO>Vl71_Yzf6$*@lb>nZEmd
z*~2;C{{R2q^<9^1pL06x;$6S@`9075+{3QE1A`fa&B4@qBV8qu$t)<ooU6t5qV>zE
zZ|!McvwDq}oV^6)etDLx-nND!ksnUJtiR>el9K-zD)VWcw3>TaT3{><OcW*NEj=)7
zv5OPLwr4VTF22-XQTlt{XeNqq-Q$AqdpA%L{I%PHTKB8<x>_T9ZKei4qEF$wlO{;+
ziA3sl@iH}aE7MZvwnLgo$yd2qEM~8%-OB&|@$L6E*by1;(jSEAnqlE;o!hcz?N$B<
z7c<M8$QWigw89POG}?3`s0A)T0HHAWp!~u#U(w)L^d531@EwerQ=){lQA}I-Kc+*k
zU?z?C0Usr$f=dg3<2cyjS2_{<O5U9CzN$MqDNv^FZJ~C{NMg2n&G}XLje`%}l5lbA
zcz#d&o5}QUgWFiH$;kHkr%YQnChIEXQ<wcx?u-)m716;9crTG2qrB^x3@ciHrReL{
z0G^!{IR@VtcfGVNDxHIs3~Nj>*H5iH<hKx_P7jBuf3Z!HczXJ7NM~2t&s2$`UYdgP
z4}pW;3g;gWcFEV*@{Jh|b$3NKD*&e_y|BSc!hF0O{O(YN={fT_S}UYc_w|{io&azH
zhyxmt*&K`erE^^Ug7+kmH<J5y<h9b$9X2J_67%G`O#4o#Jg5u{B0LZ<9lBJn;oceQ
zBTS>-J8Q>z|Mp`7|33Nl1ILf|wQp<yb5^J&{(*8DfhpV0DdyF%v|1XV=<U)3>PqFy
zm!ky)=f^%uApG*WU_Bed4Uu3bekoV#Jnu|t-mFFGwj8zm;I=vB)eE={5*=!#kx`R*
z*t~V6-Wnd&RgL13J)v~HJHwJOk(<9oC&oT>p0%_gG;9AkC*L@?9rKNd>&eS23f=&X
zlUC3{nQ3vv+5|)YDL<?KDzDqU7Z+BmXYWT$mBMG%oZms<qVJYhiy5hfNWpMO8y*H2
z)d7}ab`&nt4o&XpUP6=rh6j*jcvn`np$~x4P7};`Y%=cG%;W;Y8Lbj3fI+oX_a=1y
zH$hc&T17~NSvdB5yYL&xT}Ep4z1Yga49wtAnV}<U*aB%ePBQse)qlRb)myws{qtx~
zQCWdo53+K&i#OU)3zH;;Pvbk-9X}reGj1F3Jiv4U@yp(l*o7jeJ^KKLGVniVTtB==
zFeDXgZPz`*_i@6Tfr_hC>)KGiI&dn6mAzi5$Bebd`#c=VY3z!dIa}A<k~P?5mi%M7
zr%hB=@pw|})!QSY;ZaAEG{_TRMLV0$NVH3H3OI5&B`7j%=z;Omo9AH;^@oB)1+#3*
z0r;Nejy(Q9eHQ@Z$=Mn<9c%-T=e*a@`%AA-m+IxUf%jX7LkZquL3plybSkZ*moPj5
z$Cl0gSu&6qFY$;!@P<gC{esnduY=DI2RuSh2noC1Tj<ebD&Rzz{QAXJH=Lti9iwC4
zdjrb*p!iwqeU{Uz1a%*{QY?j;(3454fsf`U^vjo?l7!JIxi$ci+}vFE$%!3mUc<ga
zJ<Y9+1#ZAo%1Gr>7C24(S=V5~r;-I|%0)OT3AsRMn;C7=g@H5F399OQMe9}j7nf~{
zsNh5=TeV#bH26~8Tf9&wuc=OS)KlW<%EI(is~!Wb<BO(ny)&$kuz_snRLS}3Zq-bs
zJ|(l^gQrflirIH9rf=Q)LQcn{E@)p?+^R3K760ymw-o2JPU`;H#Cv<|tqxoY_H{ED
zj+jhR)EC~y9B?@Q0eo)CHemoY<qGtZa;d&bUR7qd_*=l9@60zgH#h9#+qrQ2(nVdM
zd!KB;miMh|T+1qv<i!C;ZXa5K(}St;glQwm_$KgAY&L<dqE#C$6h0<h!ef%bcTFRC
zHMi}ps*!UzclcwvdwTi^%U+__YI=@I3C%2i8TY?Tg%raFmza%2%cnbzD{-;`OB4Mu
z_S7(4_E5tZjWMszRVK0`@H}?3X<=g5Vi%ZgFP>Lyys+}?g_{E-%I@XkW<B5B&3Xjr
z>o;{BzZ(4HuX_7?yH5B!Ruk73Mq12d-VzNVNe$`*?WQ@cLe0|S%#p6_>>}Q(hKIu4
z#X8clL-mRKHS!cyMgo=v8GFxPoP`<SUo~UVNSDd{RbR#O^u0u@Y(JmczK!RjLIYxN
zYMTgd&A{5dooOP#yvL=odL7YXY46~71fjOvOi^SNT3a$L4Z{F7$El9!u0Z7aZ8e@R
zJM9$wtkHU|DTAgr=Dm$J3xFt^bTX9@HFi$pyhB)xKCH2&Kb83zwU$?ORwwQ<qGxUP
zncPu0IdPsC42T1ZCr$v_V3We<^LV!Vb)wX6!Ce=fI)e*#omlp_X}Wwv#Y&QF*<ph#
z6#Et22{6V0O5<t`wi>ycV*jG9WilM~bGO<8EiSVQxeb&ZrmRY8M$b{v8y6u2CfY4}
zOg!U$wpS7jkJNO36Nmj(oRhur5Pjq8)ax*n)9@JZ?+FZTib{_C_Vj7cGW!3YnRY9r
zk|i^&>IX8@>f9FAMZk*QfqNdSP4v`(pm%K;OQB}D#uC6Jgy-F%6)L7(ZC2?gVs{xu
zU?X2h>OK9nuEKLL^XEA2Yjgt38|i9xR4*uA?)72?THRir((tVXZ4=&n%tRQDK%3Oi
z<7?ut)(ko#*(<s<=qcw+iak0BOTqlc40;LXEKtONS*=;K1{tHii`J@rF6b#No+`qZ
zvHCK<z?^sJ{SF{BdW#b#;WL3bxp1_*AUkSy$%ylo7^dX_>$W|V@PT{03~wg!0!_y+
zY6Kp>vdp@Uv(t&TeRX3AxW;|*jm;K~;c>B^{@({xJm$7-waEwxBD}euWgf3|<*>JV
zFx^d!)hsM33kp+At|ZIw!(rFbqwdvP_~BG4xBt7V+cfjUHN9(adfNS9r6Tk!K2gSa
zr36;~o@Zjb4+r*JLRzO~M~n<sn+!5;FEQT!S{%=qS7cCwmpC!dHVG-<L;+N}bSZe4
z(8(=_)q-kRr0X)RkfbcfXaz08TxfMGW&&`NO&fzw91PhYpIXo?kj+)zJR>-Zrsg#A
zHqkmH3G`-yQVRS4VHnB_iH@zO$KJCaVcPgF1j0&qpwBZ=Cp(qMq)t^_YH+VvaRtl+
z;A!UO9zMBn=rH4OXdZpPtOgxs)h2frL{dQpi)ULkdboBONJWwf%(5@LzKL=4GOi1x
ze0CUM=&xTs#d1{QN_jsn#`wg*4pq9X`0@a)zfgcfm4pwin^!Z>@1J;MX#J7mri+x$
zuKo)gq)6<OhMmKIu=C6<Mt+tv&1=oIJZbR>fS`jEp(zEevHxqOms%N&@wTRyzBx`@
zVc5u_P2ise?WUF9cZC1qepWC60q+mbs8*ODCT}N4E(HVpQq-eQr*Z*|q&%2Xl~09B
z<RpseBtkV>35!9?2n`duECowg5Um~xUFq-t4k9BKOGHgT(+_^$6w4H)1G)x$VSk8z
z_`zIO(gl1wY-))|I?1Zw_T9cOJv|V(tCk5q4)8MokmC!a?Bm%<kg?qpC&LEGaUIcM
z_Ma<3o_Z3sKcZB_hmbw^PWJnZ^`v+6_#T##6)~{(vS@9%X2my+CZ>mn4V+A+tkpG|
zh`4LDvJRo5%Ospm%6R4!E(UX=?rjEKKn>&I3%&Qje!u>M0m0V)Z5vK{04-E{KTE~K
z2h55JipQ}6Z((Dq+Er4T3q=TV(|rkSKRsaDLUv@qo?$+}Y(=vGunQYHQ1hy*{q!aL
zlu_pRZ}GoZ#-ny>pbFF~8tCp9HK@;V0^~}7h{TSI@N&*{)r!u#>-@yqBr`Z_pzhUS
zDaTEccFMx65<3woNJS_TZPnpm2`z2HCrgLv?m|LzeFGOI&aQ*mJ5n@(U{`Z<G<fJ@
zB$sL)zcIYMn=oawj#;J%D~vk=J<hRZX<x3U7jPwN>vd{sMLDmhZGurJwy503wS^xp
zT2ojaM*E7rH-Su=_gIHDa^n$Y!z!-pG7?g|PSItGG;(al?%g~A!`1&5-)1~>|KC?9
zsZ8eR!3P-``Q=Qm-ffKs&z>E4J;88-cbk|!Y_4Zr9BTw^8x=OxZ)-YPZ&zM&KfN{P
z4e?0P#{&-qvp#wQmYZmJ(Q+n9QlEm|n0g3^;v?VXoAgN78_g`Fdp+RLYGBZV`rM2o
zreoJ(rzUO*!s&|#jw=PUFg9ZfgX0MOt959WwoxJJBxbXXCnz6(ac~kdi%<Y*Y=I?%
zi!kP5{O!<@R)>~)$OejnDpPvc#C2${h@Cvx`Z`Pe4)Frnj5CCSg0L)x?XySI$G#h}
zmCtKl-HMwr;WP0x^hZ$BiNLW*p(>>jLDA`wP!o%7vqyh?cj{9{R3bxff8~5u3Zp~;
z(iY3-TXj@Mr8cY$4Xd;H?)pCLeq`7!xikNUE2^63z=mO8$ls_UBknu9qC?r`(W{XA
zlPAj&sl%^x9&N&IO7jq*9V|YuDNc6w)b8(Sa4H3^Mvg4aWIe*#1=BAV$vUP;Ed`9o
zUw1D$^;cBB1b=0sK6DP|z>np8aL*`&>G<KHnbZbOlTFtVSs!%du46RXe?YQ1jXgs|
zRYg2BMdH?;^*HMI^dO_Q`b~qlNLlM9qll?mc~5MCfYxucyb^rj!xzvZBy9Q{5KMJw
zg+&e6#UmR;byRp~uDC*&0l)(Q{3Ad&B`%CDe+0)?rhO@{5$7Ceyg}>(|04Qv8y_!9
zP~@WIq_Dl*9~JhEr#|>pE@uCWTXYsbee0IpM}KMjY@=sVcoTl`RR8k)=KW8Y|6uuf
zx&AGG(8)V>z_Wj(?8CQ#eKkD;3J59n68>;xWm0NLDq`JWE%gF5_pEAo^^t2=vs+S?
zrbpeM>zBN1QGNxFu#8T$z7q?51-eTV-{v?G9v6y&A_4~^Q2S6>_4fh3!Zwv^D9C+g
zAT(=5qBKw6JQEyMR1huu5M2rz_Uky|R(^bqor44kdIcXmL6X<;?EC>&Ki<48IL0%S
z4y`~}JLQ@Hh`V_xWvp!U1#U~xS#~aCET5dJs!&n>GKHod<#?=8hn*+B+uEc&Jkb??
z=L~&BP-J=W^qkuvFERhe?0G%9rh<BE9(iVHY>lXuMaTLseRon3QVr%j>{8ND8LbV<
zBnRB39C8UWB-1q*FDV%*AqSNxX|ya!1#9gXIo<A585yHOCo1T5IJDmqQG@Gv(H>$S
z3JMkJ{2m^fb^B+;*0iY}>y|dRwOl0bT_{DtJ|@v`T)=o7kyxDAWbPxuQ5lreR7dKB
z5+tL-drkaM1~>8ntc8csiYs`SNm<Xjmu;SVt*5k)`r%%uDjYzdeunWU>~-SaD;mJ{
z3CL?5b3TjcqiX=0!ARuYv032WtGpk&02aP=9oz#yd&CQWmOMHWu&>nOvR~m4x~^*s
zBrr67RwXi$Y{iMoeYf6L%f8-ca_`%Rqg5YL*C*%wy_9(`fVW%DM5TxGl^e@x+L182
zIoE-lxKtT7`;FGedRtbP#J1tq5XRBJ9FDJ~NGyjfo!kK*m64P#><cQu8LHJ$$x9Th
zk~aA^I;KiE7<}eC?Ev<~7!NZ`zbSU{`pVMFAHDKFd@VlEM0Dg(<19ov%?wA;#RdU$
zH1$EyBWTKeJR1$3bjU5$EVtw(j+mV#PqiWrGGwz5L%qMGy|2P{4j!gtpxvdcVW|@M
z%+!hjpMky!q^aZ=G9P)?nTEPrrwI_QnGtv?@rhtHgGKj_U%0SvC43^lH}O&fXY0g}
zpNG$FeyWnF@!svZEcK#~%X5l%b6P!)4Z=0<rO)-E3Hgc4ln*|GJ5EdLMmf)+EyhK}
zJ7oUGF*R1Ee12Iog3%@9f;{J?y%w_%@neWcIy0Fk<dqL@&CaEiq&?7u_36M}g?-S7
z{UZ&AXgfLVZ%Ub>Q-o)RjEc~)J$Q!)mzhtL{I+uWxgV{*6g2pL%S6Q6(#lxsvCFED
zP3=*SicrI`bxCzZ75mFghr0h%_5uAdF+asMz*N(*I>oN|W~;V~i2VAXd)@vei~|qZ
zRV^#`leZ5*Y9iI1h|D1Im{6+1qxkL!n3NR}e~0T%zkRs0!C7@!jKkeOgeA#$yG_+i
zw7gj_%z*!8v(=pmoa8cnSXiD%GSJeb++)n4XF_<QyJ*DcOM>Bq(;-(2@z}WsWg5oB
zMI7!34`2JRtUHY-QPEp{uF=*aAYBc}{Dhr`nOyAq_@8#2ar-WwjQmyBr;<it{XTeU
z{s0BD65JX7PwiK|kpaw0`fp2lpKnibTT0ZvmB6J1`yQEnBJ4Vlqt(0JsOi&Uuoa0T
zw*?ZcKvB|A?^5%`k6eQNwctaR4A@GPp}Fr+t)lg94)EB#UymroOs7;y=Tu4%yfO$X
zJ+@c&mIqskzitX!vK&`%{;!gO^kA@`HqXS?dyZmyx|CLtqCv`w*8?j>k`1ABLYhv3
z3y0es@IrdJv3WR!MRbP70m)Y57Rz^(cf1vP?|wU;kRYtw`J_5*fW5s@_K=Fac+XP$
z#SJ>Fde5>V#_f{>vpYvJ>woUCjHI8CnvMk-QYYB*<|Zo4Jpbf>I*sRy1ci>8kQ`<Z
zz>7ZAI#+GKpk!buICed>{~fQ6kbB;>)=@R|TVeE40iXLdy7O3!x!@H6vhgC-#kkl|
z)_`>=2;sqCeTE`r8m?q;KfeOqx}>Hde+%q5fl|E4e-@iyuAMA3|8d2}>vuxyB{lJg
z@2%4OSWPy1IZm*TPIjC}dWD3FS;lx&4Q#K>+Kbj1LS#&Q<QDl52<B=DfRI#HpoAWa
zepU5Vq8}#TFMQ6MtakmqWgeFn#UaA6KLSG^@t|IGhJ?92mNxf@e86b$ir5o1;2^a#
z!soYx@~3abT0Kg%dQQk$p}c54f2Qf<9F`e9-YNi+;1F7adf3-wiAXGq9}Zml&d%K+
z(c|35SQ0bYPy0pvn{cfeq(rCm=fJO^)Z(Y^;!8BbO3rK^3llc!u=NY<ksZfM&ORef
zk_g*%-FrGjWr&fH5uXhYrsLTNAI7!<6#0xukIq2M;>B_<9s@sB)ITrxjr}!%IDevt
zBe<abS#vAIJ}PzxusV+-cl_bR?q^q%#0I|)^ITpZkc-4;0BJDciUIm<t1^veS;!hA
z!`4utlP9vX!y_HO>gH5`T*qF8GA{aQpHkak1#^TWpyAQjAbUGB{<~d_{5e@U)$@+~
zk}25sz!G+J_SU_G4YM5KZEgdJ<+OdrH>DUP%@kQ!aux*?8R#2D2>WzPof!RQRgmoT
zbLl*m1WGU2i$YZ>Mvz<pcaeE7R^&d?kFgLtA1ty?ZrK%Kk7~iSt~zW1`L^UX?2r<t
z1KpC=Y?`^wlrumRBurI7S0R&I=}+bYri9|L{VetfaF)ZqO+d@sEu3FdtYrYMLU?U?
z`mq1P)?8&p4ELj|;&*&wrA<2~+$Uo0RD9@|r(-#3aWSi*d7Yia-;3H9<TSwvj2+i#
zpd%#dxHP=iQ9Ir&_jMPZK*quP`kWB!WNo)o|KA6zHxXgB%0C{|u)njnM_&JW;RId?
z429Np&>P}g#vY&#{@;>8$&#NrM;C=D5HFHuv;o!us(2jKEnHlrKiXUL-NFlDRw&&q
z3*2^5AQQjMMmoYCBunt6TfctYlPY=jPyj}^0}lx!FVK5Nf`@|UikPFy2y3q>d(78@
zOfZE2hm9+Gm0d-^64|T6%FfsQSS_SAgZ>~kcx(U1Ur0&)ajwj5J3N*nnj%KyB}jw_
zVERH=vOAHG$dFe1$<aD`qO(r(KMDJW0`v-lMQWR#5C1oJ_aD5nT(D406J|239OKMK
zbMHpNNnJ>sb4RDzOpaWzfmyeakF%ugvHpiCzmm*FoAw)cb(;#z{*!P*Xx_1R0wdJ$
zmM;0Ics-t@M38(3LICRKhBzTS0ydvCxR0*Z(9VHt(6bqh{xY;>q=C0gMd)Q|jN+hO
zhl+72zwKS3wqR|q?O3Oe_vyF;PK}r27Efv$M0NeS!*kweXftr`@iriIW^%5Z!*^u*
zg?{t-RbGBeL*Pxh?b5@bn7p&k_zzoTzV`1dq^+>(%jgB?JZX^6{dOf<T7uhk3B4*Z
zZ1&Yko5<&{m(Dq1)wi;z)i<>NF~Uk*1~wFt?7|SSxmhw{)%Gq=5iSF(iB4=IF@RxR
zHZ^+_UJ68d!<6^1`}T;0`D9-Sp7Bz2+yt`o3c*A0_^(}N?c<UBFkK3K5e#j_WK@Jl
zVUD{6!M|w#kn>88bDY;E!rCrESL2InJLKLijspw1ue2=AZq<|;VTPphIdu+)%}BfN
zW2tyjLORobex!y_XjWh(+pbR&PKMZTehhkD>G?Ra;%r;E%BCGrw0@5ZDtjYSrW0nS
z(RDBU=Xi=?=82(S$Ffr-=6k1|54w5ECi6F*^{GRGcgkSD2F4}q|9?t14?$f<58YZc
z1wPb)LJ#LoG|K=P372zBYIrRO!g0D3uyR1c8J!ARFj$znj7l!p0&yfcU0mhKu=zlF
z1s|J{E0eJnt(|VkCJJdU0%xxTw`&&#EF)Y#56=FJ<a%w)`e6?HTHInyfdu!^Ok?9I
z^1(M2)8kwUrAODINn9c_Yh+#|{mHlGx$()dPDO9KI)k=-(GrB$%v3mwWRd^edYIJ9
zs|@|E68djCt4&(0m+m(Xosm-(J!7~&n8JeLs}THendJy?#MS0%O_4_BRK7_OC$~0?
z#U+>X!AZekh-n_<f$#+GmXZKu@O3QO!hYZMVVE7Ze2COrQT1Ur0so0QSMn7AD;82c
zFSX1^b>J9<!@-DHYFNc6odwvJeZYiT7;IELheD}4Djnar*d(}W=UgOCksP*cdFOBg
z6_uXuQ>hX?2ctzON472b*_kCL350iPYf=!*VE!Y@1g-tD<wpMzI(Bkkx%ki>FYI6A
zBy<PTGEGj6oTvb5vORNmcFzE!ZC=#bcnf<y^KM-r4(HaSSo<i3BRUSeun5p1tyq^0
z;a<RpYy}1YKGaaLizKxk{*qS?C6YGXdL|mY=r)Mo6kG>~5&U29+JF<0qV3U50{zX9
z><6`ZXjeaNSp(l&QI9@l15=6^V(I{SgFO@0Tzcj)<F3Brh&xdln#bBF3p!4D`4`-@
znl-&_l9cFaHGcajr_c2aC00|M(>z+{^JvuHuHltyj+2wm<qpz5m|*ekUW_~Q+nVOC
zDIxcL!|;c@Jp0WF8F#pRXr#CVx3lC9OOuQZk|3YWSKFjBa;(H)r9m~S7(OW&ip*Wn
z()LEcJ{4#MejbKliT)}Rff?=|k0MhkP$eVb<*X{&En>U0@Tyn96g8!M6}UIRB*2m-
z`92p(JF<L1ca#AF_mJBAshQu_|A9<l+iT;t0muDekM7Pqr88G0&}c`q=wBD&yrJra
zKT;@bA==8y>L-@%1W{hVGsOHXg8Zrx1@z6|u2X({)eOtL{TlG@`rly&+BY^ksN<fG
zXMe_5@6$Zbp4|T993Q5*#jVk)aIIERvgMxzZLpyaX;-7!gb5Gelgcl$H;-VDBL&hx
zzNVG_*zl&O@Uf@Z4SpZ^z>A6A01Tb60SJr2Fu}GXUT}Gbwj-Zsi?9!{Z*BG1gj_kG
z2qb)n2W%&`T_tX<kB8uBL;3wwZmkCAr}XA^(?la*T7PbN1=EA#QI|!L?dkWFPvz6G
z+{-hr*Z4EC*|Gkd{#ox`R>;ySe0`19zgSxE*Ykp~VZ~LrbUDZ`q$92X1r7JpQ;1jF
zfs_&Ndr6b8jMPn)pbqp}xlwCmKU7|D`-E9ypJuWRRG=T`lV^5oU~u@3Dy+CLUkpD{
zAKi@k0$?+cW<nfQ*J<8pPW(RH4dMJr@ZTV`4>&0bzl4fm7lq9i7%Y6d#B^Sse+`!@
zG!n2=Aq5BMK~Ya|F2JuQRlW|_R4?Buo-;PCK00ZY`!4j{(rSA1lE2~7;6_MF37XXd
z250zdL~^;|vJI5UCiRKSkqe7>S+64V2TzV_vF{`g^+t346gdB<ey`4u+)!IeaM+g)
z{Qa8eo*!}1)*0FVUa=ID`PN`KLI45Z>Il^2<wo`byQv3ZNLs-F<-`v~XiY4-&lQkH
zviNN8rnBrwh!3q0g>>flYh0J9K(5Ef@uLQCCH+pw=a-^)WPijyhT9XE4B&VuP3d&m
z3V7pf<h#?6vE?Q(qq=aqW6@vhp@mRi*9*{xuCYcEzr?k(dVB`;2J6%(wGQ<o605&w
zyj9&80q*->^KCq%-e%7kn_uY^41w67(k-be7&8-M`U9OrBOtza?ZNJkgp13+qL_OY
zaO_}t*9xs$KTjVr1KK=@JJ1r$$l^yWHd7a1J`B_FaHex7u{y@Choq6pQ~)X_b_(Q)
zkk&F00J#Q!IN#3w3iNILvdgRPNWx+nb|l!%SuR^702bMrO#!Mmj|jbAF9t=}BFIA1
zrDk(BkGMaKa>_hhv}y0iVA5Y>_i6Pw4I9HsJyyXI`jG5gUr!yD?JWB|yF8L>+s3i;
z8b9zL5qI9$jb)(FxkcRjAIQpZsycqalLpAFf-C}*H;vH?y*v2yIx9}#^8~g>8tswy
z5n>Z)ecKo5Q2)3N-4zZ)^2Gp0qvDNzo%w`MCmey#s^SYIQ20)QZSc_IeoGVipn$!?
zq$P@w3zO6hOx=Lmum~nRFoOWF{^5Bjj28%zPQZDgk>GeV01ivPomE9tK*nCfR=4wR
z$4>At>?{qrKgqHWPtVs})nQfk#mvSf;b0%F%vo*wZ9LOE@8@8Zk4ZqbYhU=sp7A^C
z!K*9QRny(-Yho_Z)OL9EjNW5>XaC;>nGqnfs*AQzy;(}|_EJfWf5~&=!+yhv-aRkh
zVLv1ph)RWf4vRvxto*a67Fc*X9{7Ey3|Rx-8zW`{z%tZOhG}^FLA)XtFZ&}P+(Uyj
za+<(UaS?9y4GuRqCnRv`>1d=-BrzspMUH75lj4;&4@Q+fZ=O`kig2;HKpaT*JJlFY
zEESv}tBU%>%C_oW?ujq02?Dk*;<*vGQ^!1~XKENciZUgn+GcYYcNx-_DwEFlMOA4b
z$z$h{c)a^6ws_u~&NR1eo8-)PIJ~}#O*YJi5%4d3;{nQ`sfi~|<ev$YgsZX$i&>WI
zZgcIp_!!X?jAUlsiSqx^suY$=!CvZ4$%SU@0pS3S!~;-%ZXPi#B`!XOOUdPEnTp}O
z&o>AG;HVlpKH)!(l6*-p`@xZT)#_5yC^pp_Sb>m^<s9e~gm?mtFxufLeaHp@dwZ$1
zDT#})j|YtIhbWt8U`jm<zE{G<8J~aYoIhOO`|Ez4yd>-2++LixNV*xMtI)%OcGo{S
zxq;H7XIk4QDscieqvAViMrrAtomCNZpPsej7iaE#b8hd>wrcu-yzcBQuTX_YeN9qE
z|L1h=EVg5$Ysft!PFUfS7}(SRg0$6}W&(lPKyk3VLXN^xdxOu%OE|wJK5(|v9q{^X
zwzHX8*LC(gW04;#HW$58dOBVkDDI~`ik=k|>=G_}<Ww-ty5iT&2q3MxF>>L>YQ8YE
z`s+Ym>|-Mn0ikFNCUagc(bSdDS%NVZ_b<#hsPR>8`c%1qm+=*~FT73Z$K!8=B_)!s
zBP)Ghj+?FXkZ{|qq0%#a3PkY!3tIIZVw|qd`6^<P&Luk?l3G1>ZL@8=cR|_X&7F!^
z8A$?O`F}?o|HrP<S<A5S!5WkP;amK5g}l<~d2kiFwSnL|g7n2v^olXH7T2M=ZYA1a
z+Kd>TjcwbxyoI__!NWZko%3Uzih=3Ki>odD4Os1fzh5-~c6Z<R?_!Zyzq~uPdrIAe
z*X7D>CoOI8$;!{a1>TgNPFM_Ej`_jDM#LVkBbN4|H2zt=+nt^EyxYiUAH1lVUIiEI
zV*6zujy;)g85mEi{`37nAFNqF?7dmbveYow%|X%yt(4Ef@xiLeS6YkMm7(;>ad`@B
zRa~>;{WDp#oRCz;2C2x~@m~K(1)R<KZ}GyEU!5e#UzTmZ?O_?^or23lQR#rL*{GWO
zUQLyP$4@9(%v;>Ltg<D9%<W35BVK#FFEC|}f%uB5K$u#sG_=_$xaeTNW?ydTeRAVK
zUa#1(S{h9pG6IH$Sk|tu9Q$`0JT72voTb54=Sp~Z!e^>=D&H7b9+t9ogV{p}_8^a*
zD~OIH$s>gNnJi5T8kVmsms~ki(+I&4ZA*{aLknqqUdRN{T8z4euNLJ!p<>8Lekjv(
zM7WB`#j9r<8OiT2FpyE*ZICB44OLyG(JZ@kO)PqA>UYoyo`DY~^Dftip`xD+_0Vu{
ztyFV~i%E{%`oN9Jw2gROlK&_+l3ruAa-b3pSF!6I)|mfiKHtAHIqsUm@`A&z*4qm2
zPv33#VVZaO(uoUkQ_e*K`&@*JiG}tpGSk4~KEecvilVK7^R+=JNGqTOZ)N-_I2@h#
zh?_ViStgXe)!x1@QAJ*&trNQ+s$Het`{a>;DuhN5ea-M2XKx4yxd3GRInpWuF}*)k
zr-GM&Js)AuyqEzu%kS?=_f*t2Gy|D)=8Ejw`zejJC<C(6{vmkW{(crtYHJOlAv|70
z;SxxRcKFH`fUV!+?)yB6ux5y|)?pRQvOh9@Kh($!60UG<)@Uf*KQUI1YGY(lX&4+j
zyR7Kt(&{isVd^5!oVA>GT;GCW#`|FwYb%->+w_j1k%>^DAFEq3FFZE4IHR63N4#e2
z=hU@=e-d#H>yT_`SR5&)GyQaCkb*66*TJe(4?5dGxW_EUoO>9aA^pEycYlWLR;#sK
zX6E5e0drJ67OPCYaLBXLy@sR+mih`-)^*2<1?%P0FbcyB5bJ6nduWSs&1spi_mNj2
zppreWTh{Y1A?E}tM84U;Jm8bmJ3RdC+zuVQxgy<8PL=aEblYCj;>%jj{lH^wtK!Xg
zPR@)N^NTa>bImo6&zR;|NJ`iSF(U6DZs=<ir?hQvR#LGXE}X~38PA-^p`D7RE!JVJ
zx<tOLuqJV^we`(9QlIoE1FouGI&D6SBi<H&@Ju*im@Y`&3r4~G=<UDy@4BH#JVS2W
zn!2afzPAk`G6lEyi{36TfdvOBhGbGS)gqN_=|04Geca?L``f<lewz%I!qHd`s2@q1
zgr+U(q<Cth65lSh;Kef^BlK)ha4ENNpYh%?L;YJ(M#)SG%6tk&9+uD2nNRks%gYWt
z(H=*W2@oZt6HOAt8Q2L<Wnftff>XG*n_Ls|_g~>(0+le~;eh4upU!q@ynxh()m;LR
zzhy+|d5Lc)k|vR$LcYY%+K1g%x!vMkIO3IaDA)fkrEc-xs)F*t-@nt)JE@phTAuOh
zQ@X8eD_0z)TFVUWFPgkHv6JDjMZfc8q@(POj*}P0?}XBgYJ@Ao{t~XjrpC&pujG5n
zd{xr`Ve9=3Y4kT+d`xKSEMj-)k)-JJb0sIDDCh4A*cxtj+S>eRSW8cAXn!Aoz;(Q6
za#3mU+TaUzosKMq2GKNsSgvJKyX2<1c{pto7iBrg&IF5!9U7?(EE>iQBG(<Gv(|)o
z$$VVuSk|JEHEfx@ey_>Xh?_EN`ccKO8{}Sx{Nbw_ng=Sg^d}02L2|P<d~UhOPaB~W
zUN0O-p)1-9^dI=b0vDxj@Jtz-oalk~2HIQVI-bs!hgC1O?K4(l=mAHJD|3)ngd^H*
z9nf|JzXI$FPPBZxR9b)kx?@YflWP98=Juv9w)sVk6J=NGo#*-LYm6`tvbCq)_~>0K
zdAeITb_*H~m3U})l=akx(gzBAECqrASOqdY?tCmi7gIB0==|LGJ-n9w1;aE?W7Q(%
zREj__S<9Ifp<x$u!{WUxSC>Uv^!{YCBYzg<oc?}Ra~7#@)2wraf!0(w_>!_sBvl~!
zRe<AJr-q0i?LA<)TSok*XaH*otW}A{5PeFb0dm`-Ux47@R-!i1L;QoImk0Q7m+_WQ
zw(C11Tn4jWZ2w6)gVxba_b&(E_J0Nr7rIoO`RY^i4n-vxbwK(bdPxI^wyi3Gpu&Np
zMLJyveL1cD*QFi+MvYk^c*ta87RF*Opye_SKj7;tfDti8^yqLuZuHN)VRJR~LBIMz
ze#{MaYwsaz>ZwQL8|*0YA?dqSdY9glCM1bpywmyZTSvKdSgh2i51eE?ezTn6OI$^S
zoyb;0HCJm%io>XA;sTg|Vum(6H1d$i+*YBekFhX_)?Y)YImvopCBeyegd<K^nEd^;
ziAb$FI!Kys`EGyw!Eiav*TQpM`)N=TZ()z#HA#d;*<dy87u><$UK&F(+wF9-v@<Zk
zU^`7ZRAKuHZagRzme`em5lnP?d^}XD;I-G9b{6L5YEKt4%QZy>nPsB(D)N+|Dt=K*
z0vF+h4;3JI+*V6YK<K3&<OBUHA>w+P{lky{&79gnvB}^nS8EVdrd36_ywD^f1H7GD
zdj6n^yWSKq7hD&(qW5_82A<0oh0CS7wIp<Af<1)hI_3?V3&;g!zLp6;aDJ1@Dg`N?
z$yHZOT(_=*P?3A0VzRI3y!ek8_r;7wmJ9v@VrbUJ=VVlLEbBR=r@RxEhlhWvd!oFT
zYEh`K`bcQ5?6&xaMFDUAXk*-#*%rTgP}C4cOt2DYd;*sIT_RC!hubQ%>5QWPLS!f`
zX7ksNgeHLiUPY7FiHx#2A|mtpj&4LRLzM+bihcx3aZ(Qly>B}#5*J>@o2miy9UI+@
zPOy;CE5qs=2y3+>a1~5lK&Cqf$(zm>IUs~NQ2CrqEoYzzgU!o^H9=a*GC?KweWQd9
zvPz&OYk!IAVc!oh#Kd%WK(Oc_lBn8Y{13g81!G^ioZo4gA0{iVU}B@1#Y25;m<E^1
z!a1eq5oL~}z~7&5V!jM+p%XGk56``{bD{5zBwlE7jip4A&0yZnrBZOr`-(}`0H#6s
zU~=`wmkL4W?QYH^97ezIOCx-gAyWLfoVw^7lNrNgzIrt@lCU+dS*dQwtTLF+X?Ky<
zQfX*x_X1GJlHwBI#>y(Z66;0%=J4P`qJHuFu%O_u`6r6r=G>A!4~t52=R-A=yF1m3
z7Cjg|ap4cifx}J-xabt2sg)t~KDdUX(_@XOg6|g-U1R#L?d8auMe_+H$J`tcRaEp8
z9Shq|Lb?e}GKj%g29qG4zJ(y#39L<_>OjOM-Kn8T`sUZAVWqp+UN3>E(KTI=v}<J;
zx_iQ={8h`0r~IL%ZHqvCF^eH>97aBWM0T&<z-HYYi<4-uT|3l_x{I7meKWxphW1av
zMd+-^?rVCd0^^OkcrNx}BA^*E&e2%FzwW$W@O#o+2eLHWFr|k@dNi($8XrFGHKDdr
zC7x&;V`ofct3QM{=$+4cyTxQV(o<a78etaGRZM<q+irXDwHq0N+6;$!=$2i-s!Hg1
zw9el7#<1sU^lGPl3IC=DtBXV?F(Z)39)iW{-z9mlXlA<94wWj<A4AXgdGf!%^-N^F
zx8ODZ`!(quas)V%xkkjb8;UIyV04>g@+C<FF2lZvRM$>vKC#^QKP$tU>j+#Nx;HRB
zqEpADLY;<0lBpy?(Kl140zIzxoB6LH(!W<>BJeQ9j*~XLEI|Xi1u{b&rxl13ZUxmi
zm^tE4A@(_o*V@$a_NfVqdhpKz`hfa@7Y6_Y=23-r4OucHYNXLF2KiGZT{~N%`+xj^
z-lzjUQV^Q}&9SQEWUmoPRxJ74f$O|e6(L1lYR#c^mAgN^)E2C<@lWP6?kZUL5B%En
z%}C2vsclQb>#Uj?12L0v5eIVv#W+gRo{@<sPs*G~qUB6n3cn$~_i;PZoT$!bViLC0
zf9ct*9MqL$cjN)fwI3A74aw@VFL8Lcb`RD$+%FbYa=BR)u$vfbptuz-!`o?CCYzeG
zq`J%H_zpbwWG}zF42sOb2)Hp2LT_A6Sk^=(Ziit5`h7C7PoKI5mI6bgsNz`ZfUQgw
z_dEQU-LTKx_2JsG?<%$yb*}YpeUJ8PSBxZc#$cm^q4On;#PxuGflD3*8Y~gl0$Dk}
z2=n*A&1el@WGu#Kl$FQ-dCujFUe196&fc|C00{dV{NC_X!8!oJ`Aa_j9U;Qx7Q^tI
zCm9yfHk<i1=V>Kp2hI2%n1qC$2Mg%8TH$NIg(TP&js3K>XuR?Am)kTl*W$Fg5*M8H
zF7BfBr`~2%4KC~aM!R72wM&e1GS~9Lv3FTcPdxXgVMjDHIpCN!tjT8Ik&=>7$qoLW
znmW=z7-icK?t`BPS<FjDm#>C;_W#7Kjd<gEyIWpm-ac&oX%@i-(;fll2~G3&Tr?OP
z`9NqTWQv5S0D2iTwuH`}G{}H?txO%6WMh|4!;y!Ly{YL}(_i|TxoNzt(cc9kXOt&@
zVhGDpBU+XB5Uq#t1X460`na5ie_UhW!kcnlM;Rt60sg5^Y#jP$gdMwOwwEwBB|ePy
zSgoIKkgSm*jSy83DD#xC4G_z9xE4FD!Y13z%?;FaS8Kq0049Z!MR>67EKdnIe@N4~
zvTBz2(QiwhTyl!YW5?GYY;EmT+L4)(WVLV_o(H<HPN1)dk^JC`dzIT}gOMZw`_Ftj
zDsIPL7X@6nlD%+e)k3$4iXn&cyv+TEYPsK9JCt{+J||SAEnoi>E+D{nn4sM~^t0{S
zlOD<MtvjBU3vlUIzj%fr{FFfdkEzwwW<gnh?=w*nscO)EJ631Op!aCV0V{!E!6J<Q
z3pAm4FnB!pp>U3&(~r+IsHV@&h5`G=>#`aiOZ!f11ytCJnQ<EUzwx?2jw51Fot<nS
zl7U!yG{pj1{byJ(a+;#t19sB?=!uB~&2YjIzcHCLQYm1PJ)8$H7Yd=>987@(xg7Ly
zGyA^y<vn3~S{_qlp!Q}^pAlvWGcK6#V~Yx_O2T*nJOPfcnM)%OoXF6G+4`@QH+1jN
zG7+1j6>#xK0ng|_VQKEjz0D;L+drS_+Sj;@y6x{ST0?Tgxj!g+G@l~1>U^QRnVA|2
zb%Oo#n)<HJp^=3>0X}M*c8rkzr2i8?COg!=0&KGrqj6TI4|9eng9*(qJ6A>>l*{Nu
zN-LNA%xl5Ees9+&ix58romg8MES*x*;h$4I1~hfU-32|hvp6Urh`qv(2-efc=iMd(
zaBwV|nKnP6&a+(fBU9p!z&X2gLbdbEz2vSDts`N?2{$fqrl99Qss$lYCDzIWQ#C$T
z2+iM!uZ~yYuaCS|3usv`4VMU;i&B8kRwB)K0o<f7nHrO<OA+=rh)RVYNoUL{Rf^?(
zi(#Mll5gE<zhF(~TGNupdtUzxU)x&<#5z4a_QG|B{q0mMJ|vNm^!sNC(Z4l4e@c<+
z(KW9;U$nB;x|zGM$1!$2rOo!m(Kc5QM;rTgB+)i(EqRyL5cYq>k4;I*Y_d*z5H;D@
zuy8lPnuFE98O0sac<#}NcM72BI$kb#OH{0SA?jhQA5J=zwk#G+{Zpz)XhxYRpX7gv
zx|~xMCJgq*y?qRJ3kC?R=<7|^d6||ZdBU**{Z(phfLRP4Jg{K=$@n4>Losnq{mBh!
zyn;cF(X@>bNs<Rm&xO)JYw@lXXs;s&HYUZabXU>X`8=H->!~AJ-i~@js{f%AOi!P>
zDM^zb$j889+qghe+cDtcx5KsX&?F{=IjuRrd*=3$Et8kk4+SbP(L^{-*IU2K=Z~2h
zHwlCd*z|m*-c97QlA}034>;+TX8yp{3m(oH1N*d%<aWcw2#wB}`8x82iJSe)-}z@>
z(5?T+Hh#`37!T(|@HjEe!p*1kk6fP3YNe`JzJ2#lv%_OP;4g5kv_MgdymtAps<cSZ
zX**@2B*>7gdPwU@ZYWGQu#cS9c8!jn$2o0*d=ZraOL*w$fL}CxOR8?y*0B7L*EyP+
zP_{M29$j<$*^R;rk;Xc8i(y>G{6MECBqX58vrm5tF0?P8vm4EF2BP5O*Q*Eym3|am
z@*WH_d_(QKrxk3x9^(O)z92r3zK3Lf7?H)H5cR?(He1LYm+`UPp#|DdjcLmP$5VXs
z@L~<SVeAPr!J=_Cb^brzc&{_^PFMJyJquxgVgQEoyo6Gf(5%o=DW~?6MNV;}(T=O^
z3&&^v8}kQKvyJ)=!qxDXU2np}E~=)C)S4U0rO&ZFuKH`b4EN0o-L^xZgdy|5zh>|w
z-ca0VskK@JN&#as{grkv3Hc*+dwvGth}rNBxtUW+7(CB0sFvM_x>rLGx9wA7BOwh%
z;Z1cE)mcJ%BZ#LoI`^&AG6MZ2dILEDf*UZkF2fv%+X_t{N(&c=&w!h*h^&>eUMwP!
zMi`wS(9`byAj$U!*8*B9(!pY52tht1R!A71@n;ccy;CT~W#feSc=kyde5*Y0Gy~GK
zb8X1iHM3^uBV+ZxM^%gaL_)RgzDEB`U&c!aI2^h=VlD82&+wyO0bGj&myMdzt2|?@
zqLUvSZMc#ha*%dqaG!?Cwhc3*|2GCvx_1JRH)7_jQuDP#zR(}rTmQO|1AQ0P0^MHR
zLb|?2B@1gUPJ2s`$>!}b5da?^6nf}2fEtcG*EWF^g&bbF-J?9IVy_GLj0_6{w8nmn
zkd|7{s~`!e@WuXcT!h7=@#44r`}T-Q&CytA`Yt{`p0I&{9;W>)9NodznyclNFoovy
z=U^g$Gao7mocYlIk~CF7Hn8;TIhf|insL`J_$BE_N0efaLhU!MctOzh;m+%B+dT``
zy1lmJ?Q8cnE}+hxuAF#s{FNW|PKcBYcZ|$StNZ;4+cezm8cX6#PIBz()idI*-8=!k
z)$@V@380P@%<~n(5AHJ9pU#}fpiZe3?m`fz5k~qOzyW{s#N9|4JtSM(D4Gyn9LVV1
z523=Pb^Dj@@k<mg+k}_Oxu2LAW*;F$n{s>ZgwR9ES~n4AZu|D_pK<0;HDN?bjKYwD
z+7Gldx9|1QB2)}FWGcWA2g;jupQPTPavHA2@(G)MjJU#O=5CX3fO3+4;l%)Jh#mq4
zNLXhhg%=N0GVqfvIeNj@$p6`^rFCyK1^9~jxo|ky4}68aba^Q=Mb?0=1M|1f0nmaV
z$_(t;^p1C%V?vgz^+kPK<@b-F_MlwTE`{qWI+Oa<+V&OaBwAfdjzuo`w)iIx>ezOx
zJ1=ErJ!YKt|IUW>PJ=XES}y*!jn!TE{e(K<<P;y8h&rxZnle&dCz(5z&0&wr#_vn|
zIOMT|&Cd3Gbm5$osT;-guT259!Ce+iD<nvWbHF$xQl}yQL3@L{QXr9dkLJTaAb*0U
zBme#ow5@1N`??#qc}<QF;kAwvpgTJ|6Y^t-kzh=U+9T=gtyNwQ9=+H@MS~ErYx*YV
zNI@={Y=h^GPDth03AjFwD>Q*lgR=(UiyZORiQ8jdTfrk!k)9tj{Y>ZK?4d~pY5nRh
zEOba0i<=l-9ab!<BkZ+jksuNR3>59_Tmhfg9{ol&&+h2nas|UI;Wf0%!`#iE4tfM(
zr+EeiGeGml`wRgN_r;PEm->&c>#L7bwWL<I?;a83cu6c+7uM|1eACy0@Q5)@;)QS2
zxm(UZrWlg{Bf2=al%CrmoA%LaoKump`i;!<8+F8nTYm#YS@{{xDUwhSBs(4;w|v2S
z?xag0)XOoNN?*P3s=K@Ui(B5`*N^?NSN&yiF?Umu58?ZSk$}=beZPH8pYH*pTe=DO
zF>RuTSga7nt#zFQm6oiv2tg-YV2C(x7g%sYT6aPCk@-UOk*aa9=Ro-&vKi2Kj&u7i
z8pr2};^=V)6K+F}ebzl~rMRlL{xvz+r~4T2Quq+?m{^sX*fpw#6<35kW1kBT;lVRB
z(PeAicq7P!J?W8odNt>TQE~=xqr*c}NFYCQOL4%^tR-vd;nF%(!MNt@muh|W3w>{V
zXTKS4i@fP@b(>B2Kg^*JUh`-0wktIw2KL!t@Mn2qgRzTX8hJlEaL_>F<`J|ODy1GI
zkc15Ysi$gU6N+TH-R9;vFW{5_mW8zjE*au94o{5s^ugJQc;N~AhP=Fs8J+!O6<61k
zg@J#5$`mw<<LjvaUmylS5dBWrjEkNGForZ7n7N~WMEZx<=jT>(Q5jZCAe}&yX|p_Q
z7~KIz&i9I((S+%HC6Fu>Bz56-DusJ%$w_5!RiaZ}u83%Aiv*nKC-wEk8~G66fS&?i
z?(q_Ha>A-d6y4pIg(~dS@>n(H>dv!MB~OH&6S+l;!>BHhw)ChTG2apqrm~mXBqHgc
zV4~9CY@XRrQ-6T(yP(z|@&4t^z2;_yv{NFNdc#R~y$3gC9s4e^Dle4QfPydx7T8D%
z0zgfAB2ykzw<&S#ej_A#4JW10Q>!KDjl3GVkjJZXH55VAaAd(@!s$VEBuZmJLGjf^
zYkS9snp=rU6oPlc)v3>&4>ltzsYCHs+&4b#R0W3*N@TuWEGCI@1B1mUFqH_5`wEy>
zMHb2_SnpWFka-yuoUE38KUA^DA$GI;b9KCME(f;5iAtA{df`Aa4)xb@Be@Ne7^oc)
zbqld#Bz+;ME5-b*U^W9`f|~rmc4uh=n0c#h-WGN%v%OwcmLvbSuuJo}Ui*2;o}b5m
z7YuikZMaynOpL;9%duM~kNMu@xuZoJhy6TGPIE3)O5|+$>}IUh{!#y&3n%N5#y~9c
zAB53p8q=k6oMZPnC9-9wtyg1>b=(DEM$-Jo&?v+g=`uL9x$w{lq@h4Qv=B%p3gmhd
z0sBsd;|iWBZbsk4IC&t6S9^4*hxnf;UikVxI<AB+a&P5&c^U3t9k?jM{GLb;{_~A4
zpVN@^iCjUC37d(6YB<OD@MkNvIN<1kJW~&gpG*EF!Jz|CKzuwPN}{cY1-6H*hD$I%
z4cbxgabIo(A&Hps>+5NGRt5V&-|uN&&4k6k+jj;s7A`f`c`HPWpEnj<nO1oHis3vX
zD{_l}t0|Q(FG#;QAZ|aqe0M7*oDnCS!g`nbk)(cIBiX){guz<|f-L@1*99lmEZ`pB
zk|k^{_KwfyWFOa4zG^{{Y(BUe+Bm5bMLPJvy(Y$lOayceNd$@lf5al{jUY<}KUTR5
zx{c+ZIU*g1oFJYU-r}+0GWA!B)W>|+j{mV2o?_VtXdMTra-#lv)rug_K?8?FhoWeI
zbkk{R-kBhX*SX&B>W_i{B@8sGuVHCM3fO~ZlC*S^VbZAzRt>xc@ZjcvpC0oDPEXK#
z@Xx|0$=9^bFV<iA6z^P2`o9fb5b6l?Ir{3zYY#8s?Vs1GO<*r9$<=py{NeLO)2WCb
zxCFY>j-DQTE*46^wJgF-Tq;3UIzfb@r((&h-BzGyh-{-jhL5I6hy8yfnIUM_N_V`*
z_IkEbYlHM+F5v4##wf{x0ZapKM~rL*C$xPj@yM>B0|P)%Tup!(SJ!&KF!K3nI>wez
z|LvB~*b(wJ!5fTz8t%4=H||OA+mn16>wSZ2uASZ!1at2^oF00#^Tgm*dld%tC{<)n
z$1ZGXo=091BF>9MW*vy|pQ^A^7cViNBDb>=oV6%^+QOD+B0P_S&TB#uL6B^umis<Q
zjANC3IT1o{J{+o0J*8uZP-EBKlZ&vyYq;4nA;t*;^+cCiS^K8LdM=BHYdk2HaeqLQ
z)?_{EctIs?(!q?ZaP6hb&!+OL<Uh(WqQp2&B(c5ee>-2A8WzUp;=+jSlkd!A+VAJV
zwQ)Cp`3-z<XeKcELE4-wjHOL%BDt0pzDkyCz_(=b0obX;SOT7h*CgIf_*B9a;O*1H
zB)WXHXOyZfcuo0m>RfYk5Rh|Izl)UyFCT=VH%=7#&M8VJuJBf61(T>nTvO1yM*;TJ
zh-DysNI0KgNd^WJ%(V3UAVjI{B=zl%(YhLz<t%7b;5KYi8p4Qhcs8%FM62*xL;gQd
zGO1=g&GwIx$&8A@8X6}90bB*!Vsr1^Kx*9McYTal#+~9fL=NBI#=hClJjSQPwv_f6
z@^pwvuKlA`d^IKc=GS?w(i+>#(vPf`gB+gTxwjZ7AeCB~#izmF6`KRy7OzP29`ZiL
z=S-&Rk?>fW`Z&dV_zP*1NYJX_N{0^5UPZ0st{@gn9fQby=6;IAVM0#5EW-W^tk20+
zg-XM$v>};=eQx}aLxhzcZrDOf9X>Tp*Lsuwx>JWOlEwzX=?7f*!cHNi?gZ{43VdIG
zQya0fhxy@v%L|8RZ~msKt=cX7gW5E-%y!^XR!&>n8!>y5*3jRXFnJ?PF1S*MYC*HD
zd)RT(xlD=cc+TdZdD_{%euJwv{EJlwF;3{27^)>T(bC`G$;5bV%YQNdQ>zC_Dx@j8
zL#<3*SZq8`nfQk5Z;*Ew00XiC{nXq{+?K>e<?bGQsUNKYpxOfuEs%SlQZj+27DH(N
zlUu|M$qh)4r1!T@MX~}mLoj1%@zv&191MW7-0g(U1y<?koR4P@=OW^bc(GUQtyTnB
zNyI6v#Qq<qU<`}%_~s6+VC-4%C!mW6C~#+lq7j>RiHoonE|S_V5OE7!F7Am2X3@j^
zRm<1!POtNLH5H~cHNPIujofni+SXF`8+aM4@_HA94x`d-e@GmwR<*#$xYG3*p>g;h
zJ?8$L!4Ad~7M}F5PAQ4j$*QJ=?|C=69P>ZV=bzQaoI^2`?2mo@$<?8@Uw1_Ogr!@3
zUU!GHag3GQsZ+A2e)RA56H8EDbbH2<=fM*jd=2woFy|^aOL+6TrSV=~ylyUc*W)@a
zMy6t4uIG^XeX3cK<V#W(BpF$Oa8_~39AewM_{x{O`yj^nu@_p`&G3q0nGiI&E7W^_
zSXB^}P~K-`zW3<2t=SBp$u;}-?SpXmZe6EjxUPfuY_V&?>WhCtKN1>hP(6za8#yw&
zwME(Dnu-R0?2`^>9-)nGqiQD`ESNQbyX)%e@UP<wXhc`<l*-F*XSL#o<4RW6)?da)
zhX`UKzK}Krq{1-mwDr)Z8svV=P&5F?6((DN%`x?mU<fVmx^4<I>?Ts-?p31e6Y|$K
z`6Z8S`8C!`(LHt^2)pw&Gj6@jzso43*LtT?>5tojwwY0D7Z;thC_9nWpF4Oi?A=XK
z2?wb^6q$K@b>!6&750YQxs~(Hy7orhT4k3VhVl~5G6SVIV%0eL$H3i|?aqDW(cCr?
z8FsOH9xE?6?Ba4;wz`RN<!iinzic!%ZD#WJCMUiYAV}$Uo1;_j-MotDxaZEr`}gmg
zdG^~*{@geAqrv%EVd0*K7N)D0)9C~L`a<9-Bg*HY#V#_td{+u}2A8<RXyGC<WKg*5
zi@?jhnyO=$*7+iR+;#P!r}Z@+JV&gF4y~a1OfyIXYK|`hNK9CP13<^R480;aB@3_O
zM7sz>M{A>pJm@_<IhoB%mzdNm8YmU4oXzPPTl;*`i=MWO**f5f1oZCbU@JLWv~I6s
zSS*90HS&ieYgCt|UC!*hM4P*l(zO4bds|?)r-vJ#ibuk?qs?7w-nUDqNhEAD|ABQv
zm6$})vtL_S%$@P?y#}B(zR%OF$9gl-lszW==|zRH|E=eX(lX0d5g4>W)LpjfpSsPd
zScy|8`oT)O4I^gRw+G;*CqO2mbLq@1YRPT(E+(H|e}(+9EiqF5xn@%L#EyopO|HIY
zCWU<u4v^x?1@sF08s<|gIhZ8iUO~SCY_(@p_=D?ivAB;g=4v~W{d>*==ub(jQ7s>$
zRYpnTO3q;VjBu92bpv1FN!w%Id~j``<!wLXG*?7RlP5bUS%&+*7Bxp^e)fR{;kvVH
zj)#Z#FZAjdKi4hAKN~czmq*OG6xXC3HV(LxEbe`3X+&Kqdr*w?{C8{WP199YnWw3S
zhGxo5^~q|wp0wZCY=<huNF7!g8QC1PPQt;b_Fk<#e7r9_JX2I9uVr}i@wi@*BrWU3
z{Dv;=EI*RAiLBgfEK$<7Ps||evZi0{(I?jcMUUZ$z_jf70e}c7dnOMpe(lU(5~EWX
za>Q;05ArTozF2nU(DCC{!1k=<zzAtDm?L`mtD*s{6=8a65GS?y0y;+yO@~67();nx
z{9NYh2eZ2D1Na2|of;rSKxm0|+MBjB;<8}P%W2Om;Jce6Bdz5Ayzvy$qZERF*}F1D
zVIq5Q;p9+~Pj|X-z?JH<fVd^|SX!6M?r|9#%JKSd*VSoPhI{-^>s_i^9+}G-KB4_Q
zVXEE{cZD&HsdFT0{YLXut!dqb;t5C#{UvNKW&I*r|BRSVhvus{XuCk&E?z-3Y%2J)
zKH^FHbu&d3%lJw#iQ$zAT#uIbAan+#gFU0@>d?^X5OYxZO8B~1$=#ve1G@I1(-j+6
z!dBB~q{6m0vS>_PTyme(vheuV1IPPN&2X&nURpS+<T%}5cnYSqcfFM1yPs!;<@GF$
z#E&V=9LkiT4E|)*R7Y%H){{m#$drPpxy<E*t8k_`ugK`SAt}3!gSdxJZZW+q;??Y+
zI~woRb-|rGlx8wrn|vdFYS8o!-FllNz`-$bH5myHqbKa6vhBFly;siFu#_4Tg<Gk*
z&r8SB($y)|+)`%*RDK?7XtHl~AHTS?uJt=CkQE>;ats=M=+ZRgF@fP~R>Eb!d$IFX
zgnn<=Vwp}J9Np28(yx;~P>T|P(tBtH;#7wVGpQY7I#FB6m#vMiD{dyW3^YQ#JAg-s
zvoze$JaMIHVlZ=fu+El1rSR17>{h(Wtr)I3UTPM8>F^4kj=_A@T7`jGy}PLJdJ=cl
z5#5;by%8<2)Fw-wx7QxYII=LWy9jR%pFUHUXnA}jZvgnJfEFfE$PVxY)cVSy-k;a`
zvo_P3HICLwMONq%mJ7PnqqZNt@qS&ItZSR>{8aO7qol<0#cS@zqD<`+_$ldxenXDu
zeUmfQtSRV%xcvfOLG@ei5#QOaF+WS=tF7f*Z^B^YNn=+`<1W<;obT*2X{@hcu|ajW
zeCt`ykTElwLU68Pb9IWuT8ljsvpJ^Cdn)`B8CPRvHfpJC11$mICVm7^9CX%^Q@J?2
z=6FTK45(BXjSwz@6JOR&%$OX$HF@_v-5IUi@oZzcNXq@f#bQwyo@!2oT1t@mZs1k%
zZ7z@M;>+81|A<S51|A0QUzBft9OBSa#8(>cpWdqVL&=gtor^CHO$cB3+av4|o^xd5
z3PST`lX+xIjLVNzn&pNiE)RH2=V*YvL67aqKP9CVH6(tn?MUQ>{LHIAUfPOf_9`#q
zx>0s-1taRtQ;LDdj$q|HvDuvG=Uy6y_WO+9nlzo0SXTATF(j-`T3UUpMYm)-b9Pu*
z!j4V=AlzF#=nDPN^gm**&gp%>$EHnrnd}Y~XIlRuj<KQ5h31!*MpbjlnJ*=DSy4ha
z_Mkev#Sy3xuWsL+PgGOwXuI@o82R-Qb$czIQZF%j;$#uG9`@ltDOU^hhqY&xPU-M)
zK^}O>Lu%2m0><EoePS_xjuRp&#U=C(1`1;bf6wp*Lai&Zpw(b;oS&p&Kjuf9F5akQ
zc~*O#)RznnDtV)b0~DBYX2aA1Gqv+ge!;KHFNg>w%5kUJ=Sa+BUG2+w|3baCV7NFc
zzCg-ZI8|c#f;DL~0UNDj+`A(tzbCe7aAS*1Qh#L^b`oV(QvA;8AA-WS1;)nSORN@a
zTO<B__$3nMJ70A!ihZ`mBP{Gm4M3s^WLs1kcRL0e&;gJ{8NJ=hPrvAF${lhN-wruh
zniFyXr0TU3rk&-8Lfre@hCs@IJ9v)|lAXcux7=oW;-`~aOzxR&IcNN+I>{Z4Yw<>7
zsc_Jq^sCL{Oz|GO2;*l11Qz~XT4D3HqD9EUAae*%-KV&%f<gYeCR+?lzd27oB%Osa
z-Z-<+ye*--T3>BdMv8&RYQJFp@*)54Q98fhnl(}W`P{<9rdvkOn<Xg;d@Lqex{uvk
zE0)WsUr*Q9vN1AmqGvsD+J7#e>!S{}X`N)|UbQ0pUOmGmw}yI_J1qT<HK=#wJjodS
z*v?Fg4BHx^qcWm>@RRe6fx@Z*c?s@J3CV=;^-Nf!84a#E_2#EWO`ZVyaX^7ku-!8Q
zzI&XVtIcBQ=;2;Y2IU>Sj<~}3&(6%@PTR>qvMTj{;U)$rICCCT#8auXwx-PrZMiS@
zV@r%djfOZy^+~<Ht7p(%{e`=J43QlpKpTRlf!vNaNAFtPKd$O9ZyF^F7K2{6{$xu>
z$2{Uv_>c&06C^;tSVn|GP|p^(dnJ=MwC)nvWcthsJ9V{l>Ypbs4EZD!tUQ0`-eg<5
zm*(G1Q_C9!GF)fq=o|hlnj<0A6P-ME;ZC34GF^+^<At4$-(}00Rq=z60=5~3U${Wr
z{I*(#-Y&b+jPlgNWBJ&!Z-&mcGp~k+_-H^m?`<^xVf4{HbGgN&)f@Vm<Jl{{n)m5I
zj&e6fjyON!VFGJgl`+Ghl_rS8@<n5pe1@<3ynrpu*0I4%pHkvijy}1Utdp9?aN2D#
zbYU%0C1Dowcz^OoQ>LRjvpXP}5uYqGnh&t7{-lBy8SCT6pM;X-B76ro=euKxb)E3#
zcU_GwA&9e%cyOW1gdWKad9g>uDAEJopSv0Ug^gp-TF&<gVZ3ZLBio?fueaW(nz5ZU
z`}K5K+d?d<6Lpir>PqY6zU)0rb1d$EaH(%i_($1a&kZa!l=$PT?SI|(@=k2mo2Ktz
z#<KzC8GD9EHv4&{_NaNwETi!suX9g}j2^%nfUDnipk!b>%vPI+Y9vqNtg}NcBiHDq
zEWy*l&MG4_lh`X2jdA43+l;74KS2|xd363{J0Ye_Dx6?H;XHwuickHx9`iMlH_v>o
zTN98d^UbUx%j|%w!eh@ot1RwdqWt5h$0dqh6IZ4<f9285Q&w&GUTpUAYyJ>bd1Apo
zONy&}U^}FYx+zKVdX6fqCe3=jSf^YpbWSLcI)8`rl8QLhcb{FPxb%(Wc$a(lZMnu$
zH!x``%WW3XQ88qd^W@BDt@JKuZZrR51>U1&%iZ+<*U{u>DE3hm+bo30xx}V`Evh1<
zd@(WByEd)Kx!}XsBq2;vWR+&FB~DL{iGayn{3x_{+Nvjn%XGD!upCCxLw-KB0@FA6
z>BqC94@5fZM{3}m%s!ws1&xA^6Y78r($4(!YTc%#CvCs&S#d?`p(ki1*siO_rGMzy
z$X8{pHC1u6D>1GQdLsP*y%#9*Q(Ot)s#r0gS<4uiD<US%9Q8L5plH;)2t!Iky;mZX
zAh%?}$9m6J%3mdUr`0EukI?z(m>K#Do@n=JxVBo%UR**V(qS&EZy@0RVeC!7q3qwj
zaqXKH(qhdrjImX+O-PC|V;@xZmXXMkDP%7sOSW0KY0zRxLdH_|O2v$#2w5VTNw$RB
z7V)0fxbNTdzR&aim*bejA#Gf9UFY{aKj&u|%pmb%v8{3o?N0lK{-(hUk|wwCdx+^w
zra1PPd2xv1rlBMU(DAl?4ZxLm_Ejcjcx}ni3duO2ZGGL*<?FtYFQdX0XQ2qeMi-c2
zsmE*y%P8v<b!Qkx?37{oZYc-0371l{v!kjjDqc-bm-15sko0LEIq7vgnuqt_(1-(6
z;c=TB;GhKs1(vLF7{0*Z9j;pz0kS_|!3!n{-_k8%sGX~Bj4!|l*XC*!^%dekd5^26
zK?o7|&f?Lzs7mfk;PyxGZv8U`bHmu-m(B%>KwAKAaot?@2uj+C5B!wETvCT?B+F3_
znkFeE5?_B23*?7e*2eP(kA@rPvV_tbnTKaSd!^VMTK>In`S`mZ)6%p@RWh=LwF%Cz
zKTFOosbtOg`A;5XH^1*15gTF#SYGTK+c!|Uk6^O+r}lzCQc&r@zGWkCh|<vRMF+GG
zJuQ6(v)(Je&*RVCI{r4D5)4TzwUFkxhKv)IBAS2iHu2fq*RTI`CTMaoWtyB)Zj&50
zQUOFTP@F;X2XCEaw6fE7<RP`aeKCt*Zt=Po7<ycE0Zu;v7aXKzJ4U8s)3zBI`Bg^V
zP{OgewlAiNu*d+`2#j7Fl6BUPgfCE^pfTI}Xi9540A$|wT-G{vFSkPjGPET5F%XBn
zf`VdKI8sP`hrb_j?$Ih%m4{ULBJX@nLclKN1<Jd}9K^(#&@8Pp=<6JcJfcDyil(kj
zpX~hh+Y{AXwG<FOdgtwcCGU0WQ)_&1Qf8H|S^KIV`<6OF`f_iQRdHC1bRinE53N0K
z1mVEcMI)0Vtg_jjrE~!_)^F;NsRY`&uVF+a$07?oD9@ZtFnv{%mErT=`Jc-cY^yPx
z8(1$VGC-Nno$dAzu09Kqy|su3$|Fm;;S<rBAETZa4TT}~{=(h}{401?=QzoZzq&nZ
z`?*8$(&Qe*!y0D)QX9B*dm^vlX?SUX^AGSM8VDFp;Gu)1o_}?mYbiqTu*v`2UyUXo
z6OFhpWLIywVxoJ@;1tTL9I&;H5sWoB<-%a_yddvq>{Pmj1W}l@ZzB)vKLC7D*aCDW
zc){g_7B$=D$X)MPx50KddgbQ&eydRGKMyV>3WUCpDtY^RoB7f>h!Qb9XEr2Yi+<&R
z)@P}H`7;U)cB>kS@{MoRvoP{Ll%wi8{UzVNiGf>fckmKq*Nu1>m>hBW2Y|_m>eWno
zCZ!Jn)t+9ps!a<=nrk6f=m-|OXVfhALU35{a8N@{{}yM!6;y;^`3*n4_(n@fdjWkU
zuJQu%D64P_5rPujHAavopamjW34$-cj6<uiR7zeIXc&|g?l0KjWnv<Q+&~Zq(8vJJ
z0R%b$FA0eAaHmAQKxqQ5)2R)gw;Ux!4a`|r(lY)ndgM0OEPh($mjW_mMXdwHWx)t^
z9~<j_^;k)Vv*}!|6rAM>jjy0cz>BDi&_f@7JiD(uPG#%P<re?j%Qk;E+tb2!bPlVW
zO#j~>YeHyd_g%b2d$0ksdo3rkik`A3W3g~~LWe9YH*hz>?(KG6zv|U@<1Q*Jcgx;f
zvRtcObFa#;SSSD8l(R^JNL9Sg!j#a(zNO%iHz#*ei>s5~*DnhUF+10B;OVY^)SQUd
za7rEb(62SCPis8vBRIc@YC3^6NX`h0`6B8-oXNQuVsAufGO6tYIoKK=RcR}NnM3}m
zs+-6YeIA-MxFG<9f|?I20c80J@C~%&fItjiJ79dMR1BniS^Pq=>)}+k$nA!!e;nKc
zi#x+XL;R)pCK1ZdLV}<UcV^mHcR)I%cq`HdLUH6HnmsHS0nqZ2iV)K5)YQC%a|-DK
zhpw6BpYk&{-N*!IRRV@ZzNf`CIXY{-JFsuCI!b$PPA2q`iI*5|&}h=ginlB~jJ$Yn
zLSxq&v8+7?GU&@Tzl$=}tIW?V{i<CG_%2!r<|~fN10fA>GtKzO`X;;2S&KQEnweOg
zeHPyJ-c%!vUU|FA=f+=^%`>KVMbezAw_X<W-bW1hj$yM$h5gc~BM-9B<EKX@GF}H6
z>Wrl}<5_J>y;+MQc0a_lv3~W(RskCOD}MT?$%vgCXo|OQ-wwqQMn2sgknSfPcxBiP
zu4Zx^*MBkg#2To70JtLE9b8E`U0`2eUf2P%!6$+Ea2_+J9sDj*4#2|+cp6|YjPINX
zTJC96P~h^PK&p$RK&7W{%>)eO0B<(<C37#S0Lk4f*AhBt>Y3+oe*=?=tIdIb2F@}_
z2e&iW*clyo5<I7%v2wEk*%{1Zvxxc^CNJG#&B@$_Rm)5B%ZGhcN$WZ62QUBW`(8Yx
z(q7xzNXr!};;j9&jj@!y#K1jdbvI3S{dyfDXBy$rH3{)3_FgLqfw};zwzGZYW&kpt
zG9OZPvDkpE;Hy)+o5SYiq-THDo5z-$`&`rYtL<o0kqgN8jRFJ@xCoGI143JKLXt}<
zf*Bg>b$G**O}n=UkMciwm^R#H2ppEz2CH+BFh8)AyMHPVQ*QryhKtuU%{C(Dl+b)I
z6oAbMz1mtJ%9IzvzlR*hJi+ghEDJk;qy#)DKwh1!fqAmR^oINJipXG3+kZXKNXoF;
zth83Mg~ZW{0t-SfUsKXuARuv@Je%PSOhv6@)*sWC%YX6~=AeZy>lil3VroA<_{j5P
zLS5I|Q9jTjn$f_Fy#MLxzKAn-()2%E*i^AVrQ3|>c!!kJ6;!*%t#jlDQcA9W3VJ$b
zJz*+_rN`#>?uUpPyRhRU*XJ(xOm2&xLcQ|{q2(;x`LE-}hYmI2_VEd!zI&b*gzj}f
z=l>OWE$8+Eq!oVdQ8p8NB=Xv#T@Zf_C83nnWnkkEg|T+&w>%Qs;TZB~7x0bYk;4;Y
z^Z=d!(5ut!0sJ92P~l8|e{m;tGFcwQX{GyGUZSJ);V|~1f8t-es++OW)Fi~h0uN3e
zpUV98s9IPbAT<|mWoWut90uMJLIb^t_n=<GQxG@s$hw#tX$6`-z~dIhIum_q>z8I^
zF22)dKaa6ZT8w3jisjI!{?xs0w&eTD$9mNj;Ad~rOU<srynDX(=rZyUXY*ODPv0ln
z`Bx^o(DuV!GSz$0z{DhZwsnc#%4Aya3!1Plq^Ae|wCvQ%L}xtMa^>0I?A9E+3^}Y|
z!(D{{8|=K@<(z*X+iPX}xR33#7kp)%DJD6vlK|}jbRotmiVu`2NMeHBVdbrEd2NIT
z3={@3`d~1S!BCu`Q)^W*S4sxvm0btok_u(M+Tk=ov)g*1NC{(zc-g}y1E6HUzpf(N
z7r5&>@z8H`u?#(Ex!c-kHn0f;1lzq^4nynk4Bw&qQx#VJ(_dij{kl*yd;v6zt`TU4
zCTu=_<s8!XRh_xGTxtG)Ok&6rxpjy+65>grI8NRU@qZ~bRL@*DvAtiTTvw`y+OcG^
z`F&q|%zcUAA<-+T?4B_uvk>H+KW4T&^#wgYXg>XDLA?8rX(FAjtvwIiB2#JE47vBd
zwH*NsXbIb)g~F8azu6Y3$ec!<Apc2I$GVTOn^*W79XP6t!^#8h2=n+RF<24L21z5F
zt?XtH7?+;`@B@BQz=AdzlA{B>y%FjGz(B1kR?12eq1W`<TvexQ%2yDhhaZn-5zd`#
zOOMP{usEa{_i*iMfP#?5?fkId%Ja(BXK|Ea=o5;KsQB*5CA+#NxWNCAYx!+jSVE7l
zexkUh7~GSQ%SH)Sn<*Nrx80c8^(T&+^gGtKWO94F#i|cHv7w|@Uvn_>!ZXye{tW0%
zJo#xB{co*Xs1~Mk1=p)f?mHXu{+Os370HS2c`*SVyu$LJgA>APj#ShqQk>e{nh(DZ
z`1)qjp{D`Jgtv2+J+3En@JZU$^eRW_&Vh~9bS6<sXZm@-&!DWRkieZ5JZb5&(O_G6
z8{|wbFyh}H^m7gOC9Oqh)ugB&70^oDeJ=wK49pksg7XumGd#=5S=uiOs@CBum_}e^
zgM0}1y8pOY_WQU|c|oFjMl+BxV63HPDn#k`6#nx(q9pgYB8yNsLg3_VQ>}fMiaLb$
zGE<g7Z6Q~PA(sD^hmw_&8?T<uPY4{fCy2Lg=iEtkIEc)vecSotEqeRm`q<8VlF{|E
z2qh_hcwW`-E3FtZ=eYdm&6)=T0j?x^L93d%br!n5FR7$NBm|DnWXf0Aur0VI)+fL<
zdJ^yPWoea&B-#_am7dfcDV!5H|Ghndo!{KsAH;ADibkOucl$bC=resKk(J@`Pp@pV
zBnZw*#$s?6s@Q<v;3y2;usgm`YNOjR{Yk)bh5Y?O<-H|3O93hqcTuCii6tPxghML=
zp9N4z)X2zXJ4Z)(>vD%e(7;9i6oj?~PD;ekcSCns1`e!RA(ic)r!ksahG2|^rxdyx
zXnCQrfwBi5-(R$l#;p0Ox?Tdc6VmBQZL-3F=xJFiOraG|G;^`7@;t8{I;NiBeyDey
zPBJ-W^1X<q3Fkn=oP$G1S954=2ZcIa>Ll+EDeV{ag`+Yg(el=U%^ddD!f_SS@O{ws
zE^GXmfqAyl%B1(<)6u-@IOG}hi=SntdmnZiQ~Wl*$*J?tTFwQU_dDgdXOzj(T3^g^
z;1F|xm#1@ffqmvR&k_4i4PA9`bb)Z+GGw7<*@*D{JwobDh93@Dc)&=Dv{EblQ@Auh
zo5~Y#l95ZyC<6C6eLetJ;Tg5q5c790;M@79<{i3jYQ(jvJO(5QxXYkdD02tip!mkm
zaVx<l;cj{t?Gm#9J8B(zG!jz20g<->QkQE<eFRebZorDS9(9C80S-h@*VA(dG{0I&
zO@K(!DmQOn7%Lj;m>3uCHS<^GfmwDn*nKJg%pz`_sBo)l6MbtFsvO+F>N0d;y)~61
z3nUwQb&=co@@lirUw$=p@z(6#QCl$5WIu`jIs2rQT4a7oYxqdR%)2QxdffW^HjyPE
zM?jj|ZIClE_}>vMc$<=$@1s`|edu~|>*4xdPju&7TOuGD8X~LbeCkk$Ci>o5CoZv0
zDFs%}T&wWgd3o}1xe#>`w7PIYLRTOscx%vbk{{YSctKAG_e|`G<I>lj*(d@Vrf;^{
z{{1Tn(O@_v847T}an+DK;7OBh@lRJ4jMRT9cF($R4QRNQ?M}(kFHj8?H|r3FtgN_m
zd*M)Yok2rK_RSBHp7gQ^{ngQ;bDtKrE@P`Tq4=N3%$w_cI8vLaRR{f8$*1jP+w0*B
zMt!=Cskuq{tLsW~g@&+SSz5Y!OOe#EzUYj*rQpYj*>yvIv)K{k##v<(3cXUh`ugqZ
zZI9kbFy9Xv(X>yI7*A5itl>zY2TA4VnB1;?t2q8^PUY3;5|f#;lRDdXyFTD{nrp}!
zPj#$czU0Z^VHf2bb=vPJGU*~h#(-67iG`$hq(L3u7o^QdfhiO~$=$eG7-d0(kUBHQ
zNRdR%3&NJV!`m{?3Qka1MZwldHRx$3oET7wN)I&LIGjjzdWuYR1ADn!lBf*hm0s3B
zhMk(<I)o$=)-FzNffYcq<pV}0rtuFCj1FWBi&O~Oi8D!A>j9K|^wT!MEiiCR@t|vJ
z!)YB$rHhffH~gPH1z&F?&RK)ARZ0;Jcy$h$I&my@G8;fAh5BaoDvY)tg--AKJk%0a
zjLXq1Aa#w6<GflZrWs&~{0<yFj%!%KY6)#UbL+4eA@L#gB^X6x(Bo%*Y;%*FcI+Pi
zAnVmtssM<^e#aWfih{lsy6omD0$JG;(ax7!qV#iN4*z&5GYOwHIXXXj549g`mi+s;
zwh9nY7Y0QPb{ue}pcgMYed3#jNvW0oDd5*=r;m#KV{Db60IP6V+pWkT;a)xuizI-4
z5&rV~PYN1U5mMn#5Q8kgbbcx*y`Wo$RvG$=PiHrfW8=SkGi3FtJn<~DtqQXfx&CDH
zjTH=paV4|5#MsQi)S_2L-t*V(jA9(&E!e*0zqmedg1@$YL|A}1FTRa=IriC-*Y7dM
zbo%0Aihcs6SZBm?WDj_r)U=oP4N@snkodDO+5C{M5IC%?&utT66oP&nz_GXiSzT-L
z&~*wpIX7exU`%mS12PXnhUc<+%nCa&N62XjmX5})@D1~T8SoyUL;PpE#946N4oy3c
zc=Ur$uU`86bMV<GjBKbo029v+Sca=*+2ltcDCEck*~3?3H%a*EUsNo07Y4>LtlWyu
zJW46aG>s!Zbopy8xm|Oy1KDYhM!gO12z?2Ae%M5=70Y@e;%Dk2oMi06y8D_<-M<**
z*Z%SRlMbxIv-<<a#~5VVv?7W0;Q60Ub;>Lq*<)+r1^p5z+ds3n*<N>%f{XGjIRUa&
zNRz$)j5<FwLUK0?H$@cAZmn?4&FFnmqEq+=*qfJBpeY0-LC())trJTWd5CTiH-qjM
zjwU$h;Kv8>2V%IRSg|L<L^lzgGo@@>Rp9=B_a9!c+J%qsOTcuXop^1_5FQ}|)wU=3
zpG#xZV59(@y`2*5x{nAyRha#_N(Kf+HE<70Ed>X6*j)h=rMoPZNj;=9)>Pkndo`YA
zGps&V`Tb4t$>94Y*8izBq}5pL{6d)u$2;V@1(_k_IMpBWXV$sRC`%0PjkpYVH$@ya
zsn46of6?9g*<)Xsx4~VPU8~>UBK8I+35`M3rcD^xK-K;{^JH*{iQYqX>}eQ^Z|yJi
zYSh$fq8zCFV<l$zs?Ra?NRtVx14ipnjpfkEh@6@x^&d0j;$OM-D9#h`1p-_iHrAU2
z!7mR`Gc#nxssxG0gtRpP=@N3mBsL3$gmVHghR-5>7ot3X06m!EVj7r}4yc91o{*Mw
zn(p<LE3{+s5Fi(T%tAj2HvWYj2yg+L%$!#9FeBhTRpfxGFtlI<N)s&SyzPNv(%}ta
zYLJ5OjmWhwGw?a$9V1{ieY_AKeX438+4FiJscClmP77uXPrTV~lV1)d=^koKW;cOq
z7!?vABkP%$T_k<~l$xAQhnj%!<@lkAsjWZX)!^ADqNeHHeUxu7xg5dzT~2W}GChKn
z8+GfM64GROy9te&_ov!${tet+7iaeoeu#|{hIQI1ub+0IXNq~oYQ#;<%|(J^dfkB?
z`sR2-+`ySXNX6(?_7agVHz#|@(>4!JABILcmD_V#P?^1v8uH_+LnY)+c_QvehWpbm
zAdTA8H|dx3M-HP;5|Iu2_w+gem#6Z}TNv-4pExKKi)5O^tbKqNQa9m!5YXrO>xt3E
zd7oNvwRklf1Og2IYVd;Ry6u^TJ;FAe^Nr>+%c}l$dXg`F|K!^79lN<(>|e4@{jURD
zwC`wO<O*Y$Z<G2cgsJx7Y+L1t;#hS4ApZ`X827kmC)J&MPu`YIoM_i!9LLTdNcqKP
zGWUD1fhuvZz3;bDBTXUTFn0ba37%hl6O$vVCz2>H<~hXsd-vqdLmLA^x|dU001v`>
z??DErgS#bE+MKvsy1xu=8JqmaRFpMrj5sGqJ0$C3jD+ipT(9C%mAlVYN<}#{mi5e%
z)I4BU{Rh#Q?jWFuPrEU!buVW4B!m-Iy@Re18GgCTfww>o0gq3vbx7R6m6_c@L3$Ld
zDp&=MAzThDkbSh9c*4K<op$-^IjyvGEqkzlwYy|%t#-&;kN&f4>eFf4f9p`yWIBud
z$n9(SJq4W(r$K?YvuNpNS79{)^cKbbyYwm#-Eq~C7d5OA>_5x>I1Jj^QPq<~Dibm@
z=?zh2F$G!L#zj$+S4rhX5AWH5x8U4o`Ic*7X{sRtQW$v&;4{!3D70R%NI-^?O%3H{
zhxfvr!kxF)!<_;{2`oV1mn_a?d7k0sI}Q8bjD}z-6=OJtfMEi{P;Q9Bs{=-xq}(Wf
z=<I8`V5(T^4wN03clu&Bu3R$}aBECWT%%ENrYg6ZskT$1@5lydqoBNVQLYptKw~XG
z-?qCNdEH=Rxf}6=3J=$g4(>3KdN6^0+!s6e`On;^KMBWi^MfR1<{mARbw>^Vh?n0w
zDJ$>%*#b8>>m0>mI#G10`0Vu^JbIsRdp#ojX<$&;Q0OD`$M<~UzN${Q7MH4t_%ENU
z#zddhqHj}?l5}O?=u~uk(RMX=>4g))4*wl?sVD!MP0o%lUTMNLn{MB$s=I%^+VjJh
z^uE;;YVj19&<<UDoK4$Yx5Hc@D_bOI2XIr_{5y2Z0hqE=%314_NFA9z{J8BgS$q>#
z07#8W$-w2ZB7jjI(t=055O6J)T5tu)JnwPn9=RER+ye9XP=dQ1(h61Wu6Of=RU6nn
z%SL}Pb?F|h6cv1maYFCu+}3T%CPX2^Dd3Sj1ONxbJP3yc!BEG1;N%A<4GTasATvN2
zj!mp>Q{g22cyoWULCx<QAwMiMjg6bT1D$VG>I?=h9~k>PMbbI+r0f=F|NVt%adg(e
zv>qq(B;0--eRKSG#*jj<W#Y+(5{t?``~G3po?4RsWVwY?d<t}trGuo&qE1Ea`M)t8
zIe3%xlQNCzr%aA?bmfvs$ONOaVGm6c_P*60L^V$2j>~ObyjtxWPVPMdgcx86!zv%n
zJs8?Hu9lW*H0?!#WJ#FF9da#~iaP-}fvOC0#AEPzsSW4a9ky)bT82$NS3hIPAi>4l
zu@?W_>9(&Lk>@5VrMv)qz5p#weV>hgy#BAvU<?3;;nIE~?tMOyPtdPg{Zp7J=g98q
zPr6&ZwN;C|eh0lI<YunF#Vp(AryVW;!w|4fp{?Lrk$|--91T_-dwKqPU<7auqt-Z>
z_4S^`yF0(yne}cJ{{xv3VKSRn$4+Qr{er49ez(1yLceC>fAHWZrXF1!SRhN+%&*hM
zF8|N&q3|l*L|S&xFiAU&9(V?T1?Vu}9yD|pq)8>slzG|j0l5#jt3a4<v?NXGq%JJ>
z8>A&g$AIp@J^Uc;43WIS3pV_O+@}LKL~hAO1AfeJgnDC3gs3`L6Ir%w^AqGFwNt1R
z!)gZ|UsVJfCEYYw(A%9DNQu!JcZMQ4SCdh38=ZTH=hL)Nt|fs`bZ+l(k#X_0hCV|3
zZ{Ew3f0aBV!onnrMNM&OZ<Nk#K=sYSchND@4;UgsCVtlasLd6oC%gs<z`!zQ1Js50
z52Tw1%BKe5oxg2c9H;x0p1Zy)qHjk2mUC>!&5Xxft_~av(&zxWp?R4%1Jg$byl-E(
zKhyi01e)fOdm^4?4}FaZM@;-DRyp@PZ1{Mxdq(O1nXi*UX^eWfg$4WbTOo~aHg=r{
zo)K7vz~cqV)hJ~g2s5BFU)!!Z@t>*(3^@EVnx68pZTc<2`m%gXq^sZ(4a3hh=oLVA
z>&?t3QVR!CHvdI+<^l=~Qlu8@0Wg4aGtVamtgLTSi9rD^Nr@?vnWC84Dd2akzLsl=
z&k_}mxCvttaOUa`5L_B051^gc)9~g+YQ(8;ZfaRi5p*Gk6r%TAR7vWi=4(`W5vdm-
zwg0SKfcXH2H~?ay;u;r>p}RfVR7zUfR<{Gfau?6+jgvw*{vHj|)?Os;;!&IqWA7Hb
z|Ab96b>A%6OMOy>GYMwy3!a&*nN7+zaRIE}#M4b1>;~K(A7$K8ZGt<k=+7Ei6aUE}
z4m;mTEFnD*ECW`k(b9_pKT(@m8?3%Hn9`g@m#_IhuAVNM?T*3+2LN3Obdzm2`Cv(L
z?mnU*VUrEAgY$Db!9G*^&MCQFcoth0l{05lGe{0)WDY$wp1DfK1dosyyf1bN+GUC&
zH7+3thJ@VOV%9Tm1pok{3&T_)@_^w^=^6@uv0KM%5SI>OC47?gcM$L$z!}*4y`&Q`
zcqOR@uaz|{(le;^toNwt9XP*gMOsx7c_uHC$y_^Vylmo?B9I6ZK0uMs8Y0wpD3dFK
zd4N!cr`}!GI2I{B>&>O9xn4;eUl`V^YJ9SpT{_!3!De4e(K4m;2a`!aSXuq@A@vsh
zXO{QNIcZ&Mk#FuzzfRnlrdTfq*4=$KnC$i=OxZSGLe|2h*QqqNe)*Aztx40;6N~)i
z(7C(v)-%8a%IONxOkU~cu+JkkD(H&?DHF4tGw61U@qAM%_0ZHxCuiSbbutPyEv+ez
zT%Knp%I@zD@B-=%6OK)O_Ted~M9V{PllJ-|yK|q$1&?Y<G!am<td_M|Q>JUhI|sQ<
z?t=K5hRpX8$`1@`N8Gm>w_3U~1~kCIrPQ2wYU_Le{Q31JET2p}+&Ld4w{Z2*;e*{}
zd=dA_H?rc)1@2qQX;-EvvcD56E%b<jN0S2!?BuQFcD0?L>7JJqH?*hVSip9i*1}!g
z?T0NNC)yvI1VCeja{CsRu(V7_82jqyAa^ir$1$GF^}CsV-QM<=FpnM;C!<UDEFG>Q
zZnZfg5~jX>Jo7(op>$4wD}azmC`f}s%c2_%&wVXFQ-w2gP`}?{L8Uzs(doVd+vNXF
zX^a*}d+wKTWQ-26*>$Y@bb8E|+dDz(`CWCUeSe$G(y5WES4q*Pdn4v1Ia>g}2NO1R
z1Jtcsp`h?gFqG`XI>^xwKubZ+P`NV+T8-?FI9Ffb@;&@$HuPnb-jxP|#=cpr97n=p
z)B{;s4>yrv^`e31m#m-!qoa2o7)Xm5YD#ss+kz8k-41Sm2QYKV0^n{A`Q7fqPy;tr
z;5OO3cE+46!*p*#426L8C%yl2YIa$o+ts8-CuyyiJ9Z!<7-~?yWKj?jP87OIx&d~n
zCb;(_hdxmoInFH%1J>QO7w0uGip66mCcyxH_>f~>8yxruDGI0&FeCx<n@J?6GH*d@
zlB~h+e%LlRS;NTz=qYTur=*x$atus+pKkK06Wdz0dkdneSRHS6id2<TCM{bS@M@&#
zS2l{XN8d>*c<F3=Z1!5t^@oDh7ohB-+k$A_z`8v`*&Sc8ouNHq)dG~9h9QfB{njSE
zwN)2@=>|L{_q#{-kd&{i)}6oQEV?6WF*P91>T%V#_UQK@spu$l1hWy?6s5W<W2ZH^
zW|GNANvY?C_U^#}hMSl5@zSHAMCg}rUkq!;$)UGVAHi4H$HN(wvipZ$x()HLccRx{
zP000!Mbq1LJh^{=^(M+FCx+@DfDH66rm#U77P|{Tr${|c`Uz`vJFGeZfS5QKcR^Yb
zO`+-=B#lw=S=3B0C`xL~!k^t8Wy9K0ex`j7?hMMwZ%k6k@D)RR&4?b#r0kqLE!w|^
zD)FG0w`M2CW;c$Iq>maVz9JD^lYTS3`n?6Qd;UYK1HcLnK&aUu(ct<6!UWQ5iw@_R
z^Xl7h)b)D~J7^1bcA_UKtu?5=CX`N>%}Ag0`ie#m0-Jt^7T#L+gM&>n%@I#9$(BW1
zBzNqjS-1&1l)7ACZhc%k@ONLI`rg{TeY~S`MIO$5?>@sjQ4{PxnS$`2Eg`D9fpU+2
z8asbRb<Tm{F*94h8T?9sOmf@#&kNJEXJ_nyqQI33L1E2m#GSka*?=%Y-^_fZkTyzo
zlQ)D$3eW;yE2gk+m+g!A**^bnL)4x!W^mj!cGRz(P#CW1%_PR@t_YH?Pj>BK`h;cv
z!B-l;mV}v4V;hM6K%xQD&Y+*11Za@2rzcm`$}pp9#`DyoGbLj_$rj(mt!6+^uhi0>
zICBA-i{dIh^UQGK*OM)I8_krz8WTh)vy1JNIDNAuY0MV1^Fl&BjPRih-V(!Yi&T=H
z+ipH<sP+k*32#5^CF8i;B}01O@%Tn+Wy57?2;q$6d8G0;X!MY0A;$t?@zp`v52=Z)
zYXIBGmnO`cZ2+kZ7KzT~_-_w;fju8ph(@EI%^MTGV=p+)MPYb^;w=t-HD?{WIYTa+
zSy0pUYwZJ@iBr0QpA|5*M16Vl`E^)qc~ORD9+hIM0EGUFe83cFKHT;j0e}j<f6l^G
zMz<5Eo164bw`E-R3>y98ldZTX9xR@LDF^w0nF<e-udvJ;;PM;PeEyV@HtJS*Di3-l
zpt90^5=GL8j6#Ehl8nel#~2bV#>!a1xkMY3OE8ri=T6oOcj)kL`*RQ{l$4^#_Tr1t
zC!`~K4q>><%|Dm91IwpNM)^>L!3Wr5K>Sa+og5B_a_Wrx1hwSI<;utsc~X%96MAo`
z2ktWC)3{25;RLq}uBq8wnJpI|pUTT6UPu``9zI}9W;DEE3cwtre;|p-VNY8TE){@E
zC-Y)%n49BkcII=1j;oT^jhFFDDE66#7Z?5FR|6dmQF?8=4J@(YYQ;4hU^5Xz&mhXI
z7yN;8-{2C^fTa@$EymRs2mk(Qkd=1aY$^T0J6TTk*PJY_tOOne%-3}M?;2||wc~i)
zAwyR8lGjMYH)rA4XDJ_Z8UIoX5;tJyZy5i4dMjCt0D0E|o94+*(8<_y>Dn1#@?oNR
zzx?~|6<WFa95E*}y}P3Eny8G+DsP}BGW8HO7dyWg6C|HPPol^Dm2g2^l3^esd99%q
z$Nw~Q6(vK;-DX}P<_5)Rod8lw;iq`{J1F0HvYDIY0qU77NzJ&h1H2Y%>Y=IJv36K2
zl^b$}CqTLal=dktAU4n*D|jcE-N-b!E06@O;Ouw3v;5ZnX~b7ol#S1HX^)&QL8&{Z
z-LESfvlPa6cuQ|8jA`sfIh99{J1>ztPAOlAjigH+QWx4PjcGJZ%%Xr*Z;b+nb;O`7
zAKHf5?VSB21@@m)d2KvA07f9Z5+KsFwe5nfB0LJ9GyCIxOi2X0Y+ws`Ieb6Ru>oq{
zwL|%@NaEmy@io5|?q|^<1h8<GHfQvy>goTWeuQ7z)2&Ab#$#?K-{RymCV{vo%qISb
z`kS3EjE$gI`L!)JGW-62XzJ8HUzI?wLBokN?(Kw~PY;OyLva7*S=y9x*_;=s0U04n
zaIV6@McpVRy1OOi<=sa}Rk}qCbtnD)N<10BD88F;Ji#Yv<8EbDpGj>Qod4(}R$O3&
zw6S$p>5J_W2E;4eNjXSQIZ4-%yp#tJC+Ul~y)>_Jjs-1Ff$T+)Oq#ic5^o-yqMl_}
zC_x$IhD_wLi7pF66$M0F9y}`P0}sx1CMaYis2rAlzl%#m_2o>OS|n!YiH@V8J;DOz
zLN^|V5CnGP{Ot8ivSkb>Nd$t)W|kzjW5k>Glx}pcpko9il)&(lv$t;Nv&Lc@EQe(Z
zkIz7oa_*lumo-VU<i~Khe+1+kFmJAzEa2bgx_yRIO$tuK?RAUyL~`eDD4<-KTX|~K
z&AzRbQ$-8?H%goLqnD4mzL<T{tg&1qxgRL%M-c>V>sXez<C(hctIA9IX9?f@Ca=l^
zIl|(+GEYz#-7V16w70~(bX4Ylo)-TXwTu&le=U4k<Iu4sCtYMzk>`aYM%MYF+TH%f
z<o(c}!<k`>!ANfKL`<W~QJGi$wZHfTb)DlA+K8pHuVqLEM9NC=>yx@X7@Cqhn3AD8
zZV4cd7%Ab?EAMlVF$)M>wzaPa<P+BQn`KFX6ta1WOX)I|=3A&Kd@adGg|^q_-3X?E
z=@S`Ql10F2pA3#Jq0`v@k|=a{s&a9-R&ay1PA5>5c@kV}Wu)2OiDz2=;`DL)*cEwD
zwCaH=;2*aUAah_T+iRZ;|G>V$Fy)K`cmnf1ibE==`;Sd(cHd&}AZHweg?V8IEFnHg
zhV3P<5rH@0Hg7645T$!Ew@Qp{;TaAzW_(mJS~MIHUAp|S@0{Y;t=gsy6sn`l{`#n@
zf0rk}B4!9J^Y*4@4$0&W3PoL7*0V?TF3RcaZDN%K1KTUmT~+h)VI~^nZ$>g&J_9Ha
z-5hKW0Ylv?;Fi*kV?h|;!9MGmHTSI7bFkppIUojuXxc385Sh2wvjVXgAXt(^m3i9C
z9da!#qmvC-<hS|_2NQCiI}Mu-l_cG~06&pa=b1iJ=Cx}-k0aN7Ok%dA4B7RrU8T^7
zm)?aJZuAnNB|y6acNUz{5R|lXQUks=pwXOtz5`&89r%{`x&xNootBQ}<1+jVgPM`T
zfe$l=BwNOQ_QVBA4yz0bSRp2*gicR+t71$GrW@HM=@Y0<FD^mj0By45Rlv03qaG=I
zW4NT|4$4r{^d;BH-b@^j2xUb6g6<PgKiF-+y8~-?Flj3r`y{^MSOZ7ASO~SyZ`Wak
z8JX0UT=SXf0m{11x`ZYR|H<ttff<A8Ig!rnA&o>uM05zJgI`#HtG%*GH0$!ur#5B&
zb)TQc+;JC<&2T>u>=MShWAY?NqIdi+kq5-|Lsm0?kw`OEpgC~PTLqN*jt6x5{hn5V
zyRtfH#bv+;>iz6T>duoG&(gg<e>#$kIG_y-JRJUzs3%yCd8L9J6UiqCYZJDefLz(Q
zD=FlZgyRM^9eM^hlX1ngQ%ske%%29}EgOo<-i0jTGoPczZE*o|*tr8-A>qG%f@{IK
zySt1(UZ1cn^r&&|5#Vuf*&LRuA47AF$bPfSFdge9Inj$E`sUwXe=~O1wx0d)X#iE&
zxrB>u)ZM{zgRH?hZ1}pLJuWGOmi|smB>~|&<TOm$H}I*9hWl_;xf}^&X>80W?@?(v
zi?veqM|)2)Z#Kc#;dp?A&8=5{X^p>Al@6zGF>d{&y$?=?3V+fE@;Ns`U_&#2tT#fI
z7eoHAt>>u}3cJ){Fm&M|4_7esYoik<c03n$V>ojlXayhqdOgiq@vqlow_isLcpo|}
zIP_$V>>DQQO}F~op&2U=IHf~u)g3~<zMj?qrc99$i+Qr^TkvK2Y*0EfB;@Zf#QOP?
z81-tCQ8-qLA9#WimCe=lLd6Y?HJ{n!%f3Mh%|^Y836?u?WyYK1$PeD--e4X~L^<)f
z=p`hYxHV5{i2hn-fX>LAjT(K`d=Pn50`)d;3l*3C8jTv|7J7i@;n-gZJXkSO01Y)P
z;?2$=AupylrsSc<W@Bap4r>$cBsdqcanirH<;7!m&|4rI`<G;%kl8@Xk7-oSi?yTV
zZ(-v2$^l$M_!PD&f+xejpLlqmKiMz#WE^B~W3cm}LxFW8_|L_>ITpntvzL;NbcZKo
zNM-m9`L3x++d?V6#PmxX%>3e4?x?X$BJoFHLLP3cAASH5Qt|6l#;>=tE&i(+3U`TY
zW5%TEI;TKK-8Jb<COLhP;8yIfVB+D{T-;xiQUrP<ZBU%dw0(mqdRt`{I>RAM>%S?E
z<|SJw@k-n+^{8OxvMW-^QX~9+x(b9I;GLMCa)y1ie|16TUP0#jyIIJY8^z{b&D<77
z+;H<6xw!a-$QE4BPC<v>5M`_94PX_UUitWzG^PecnPeZ@AOFYg@pZW0!`T#QXpf4(
zaKPjAG#TXS@%7OAA<Rl6898Y`SWcI@tBM=7@H(k&Nldcu_|he2$a>AZys=C%IPRSS
z17`=!q~`T^QnOR=XA0U}@kimqaG!AO`on2){8W8Au(Sas3nCCtfs6mB7JN&`t{(Hz
z?|laKB+cYx&@wdnh{U@~i8&UY(5SyY7tWK=Fyf6}3+$NzQ0dgRRnE1h$cPzdH2eAN
zUyrO02E63fLnSG{r%e-}PQPw8?teS7s=ljhl6TFGvi3V4Z(O$sEZ)chty}%uflpT{
zKGQKsJ+bI{2abC$WPCrrG6ZVyfMv=5QO6B({d904!$BIpyEfOsmKSDS+H-Kr;-~l_
zwg?SbQiI15g+a;$t4GZz%#Q{&$(hG1?UdHc9rP}LckN}0;tV?F7>fVSwI$|-MLT2(
zH_yqs>b*AtntRZ%!Xg`@{ul~CJ*oF1;8^e_uIAQq6N?OKGz^r4p{-k~P422Yl|AwS
zF_?xz-N|$!Nrta{+s<>=>*Mhx66}<-B%x)2ofvG8r7=)Gz?B^+1+*%Ks73~S%_<wn
z+r_8=u$hE45KvXZP~ZgwW-c(-Ia#}kuaRIg63ZfZ8Th7aa_;=(G}ABJH{r_93^8?{
z_lR)Li{@;(1=`b_u$<XbmeVfUpZS-YxTsqdu!l{lvu?A3`!5xasFnQIHtG84Ro3v`
zcr>pk9n1!R3VZb2Ld~}+7PHg!DAsSeR2J<lJFD_WhSiK4So#q<=>j4G_Oj)lN$L?4
z6XZCxeW#fyD4c|IaR(vR3)#=Lw^BMRDgh?d+|<B4lPuouTrOTIwdn^8Izw-yb8W(<
za8I*uaPQmg*pNm)n=V5j_Ha$Vz`o$DicoF_=$C;Q?QIVfC>Yp4qKYVcx53X3hbMge
zKIJ#;aE75>cH~18p?`kIIcox%a>!|KL?Zr-n(-kT8G5|73s8$8)I~ep==4v>@l?AD
zvmKt4G$_5{3FEzi1Bhfn&Hx<R1H4Ew(3nr#LIHFr7kz<xsz&lioG~G(-DaN^Ob=j8
z<jx8&4gOg+{bS!#?PL9b<Q3lv`-7GK@6~Ke_SpgFpT8nP98E2PBPIqqsPw8D6o<V}
zIL5K(bNhv|!X!TruPHEcFa_TW#i<*Avkm{1T*&wD^F^DR9eKnL1fL-Xi)}OQO=g6q
zCb)60Yn@U*5hiYSo}n-rh-?U=BA?5qeQC*kzwOPllj-bFT-8IH%7F5XJLJhFkWvw*
z^@yu}DHvOeH8%#_M8+kZ1U2)?BQ%(aOmgsqpR?SlkQ-V7*hu;y#(YI02Cy2!M1UPB
zeiMr){k;<S{^NonQK58z>yCl{TiMt`TJdyV8#S}%k_xP4EMJj|J1xJFU@Nm+hB;@=
zN<=wFUm)&O-eOT|1F@5O(2KmP1R+&LSRW>Qhv21$BNaj$rNbU;H-p5%AmZkx&bhgu
z>@SiTDKT~cEthoIYQX+@qga-L-JC<$_buo*Y-2+iT|pjRaffm3FV7P-1p7sD9?;eG
z)jfL-aM<>zVjSm2H2y#E-Jz@35Osj>SEsfkwK?M<Pkbk8S~PrBH|vTZj_^<?f%_$Q
z6Ee<d2kQSVz<#E|$tb5p5e%l*Sgq8rNkSO*F)+{XZl1FIw!br;$Z&wh*qet%eLSgE
zcOaM`1n<r<hr1B5*gQqOCc;et=Gq0lgyRm*Js_HZxC4ClBkJz^*NiWKJw)1+s<B1b
zOGZi2b<ZansMe&Lp7l0s54h|EydJ1au-cWD0)87(77mL_A4Km1zaB7-kV~`P?PAN~
z^=^4fb*D5QCr~*KmB2VGwZW4H-x-A`ph(+R1@ks+Mq-G-`r11_v{=V!Fk?i!>d?sE
zwN2%Dq#da~T{mj9Uie;J_2D`_dKFLspXczn?WR&3e#xBFxcYpetan%GOAU-<{ehao
zVq>%3>+iOiD>q|2zC%vcS&F7xe^#sjP>vjq)et*@`i(Tu<(Htt%Yl#}JKnz@&_$c4
zesLQ20R11<&iIj?x+$Uu;R<aeY_cXd=}*2qc-RG%f8rv=AW7a(q#iT$^^Aegy^Y0i
zY=d8yv9?kxxe&dWINES@4d3eaRx0On<Ev(MxW<#epJaq_f-Obr0ZEN{BiVL>{5LU?
z%IO39eW8SLjcw4shD~&I`?kE5@=RtcNY{)*Vx#oVm=a{>^LR$1D&Sa?zOYOLUiml1
zqT5PMxSlkV&FiX|+?Y#sDm%D8j6I>od%QXs0&nqSNwaqfMnvx58`e>@P6A5CCZ_xW
zlN{AYx=k1qqCnl{NQhXkxG$paW4~rO&jIanacI(R1>lR6WDC(UKtCP%`3szK5&^|o
z&odDp>l)eiR{f1eE%C&7KDXp{V*cECH52rv5E_hcCrmO(K}^uRb42s*z;OS<<<{pO
zl`rw>>46m;G>W#*j*J>p&hcQRK79F-Rw2m0_GT(Ru-q;?`vVt9O!W3kTyuT$mcbj6
zk)D?JHI@_Rbhwi)yPfSr)|vg1WMmH(!f=$p77o<m|DGuXmn%eHRLZ448n@0xLMo=6
zQU(tB=QjFjlfxo1x5~_Bp$__i$+?Ee<JW8e_8>9@tT~W&T8$eNQ4ik_H_J+A7Y|0>
zQEXBSQgt8~7ldPn(;mQBD8PZe3G9D7p5<|Q*QxI>%2}6px;?gZ!dXMJfxw~ihS9~D
zXGtGMoIorC)CXW+BSK@48*&LnaNNSII<CHh)E{}zty^v7D1g%d{-_)+`fy0XD9kJ`
zLQKY+>|?_|d|K{nFLK@@`!BBh)uGFPKycWT<oG*<bgRf|R%m2-M~?bqiQwIxiv%*m
z%`#ry<JaFUBlkH~|L=hAy^p~wRpo5$Y{(JQk&FtEZ$%r0xa^G}Q(=5#DpwuEBWt)s
zwiMQsbwRnCM>A>aW<#~)M6`iIy{ERK=^4{u+kKg(N)4zE%A~Zl?b>i+d1aN3UIM7v
zR)9NvMUldmnsJ?-**LZK)@IIM(f6Rah~x_Fu@D?eklyH@p7YbZvn&>T6~PP*P)&Hj
zCtPo|jUchM1?x+i){lkY>~L@-ljMO>l&iRr+X_9Ltb}z7O*3s<#g_sbJ%S1ToutCu
zY#_d`ZdHM%gSYMmDD3i2HLpK1*){2I!QIP%Dh#1X&W`uNF%7aRp1zBk)XV(AXg-p4
z!xNkI!Q7XppcMk7KglVV?adJ07qIU}U$cEo-AC24^CwKUxD5&XOFX(6Fr=@Rg^Af-
zm42@Z18Pq>9VuxHdacu+N;<M=ioa^&9tqEZ^11(s>k?VaN``4J_H&(Wx$4N=;EC~`
zHrcgaB8astNGL%%gJUr<O(`q8(2pv{-m;C7qRf^e-o+ilpNkfMzxJZ1@>^UCPYBs*
zZhiPx$E~Z@dUnRq3Ird3leT>$Pa0;CZS=Gf`t<+W-QFh!WAc<1aL*tmTPPM%k2A?s
z=i(HyTSGAL-?_VF%Y{=YCu@ygh)@>09}>5tCBZzn(=B_wEAGL)Oxc)Bt-ZWRUNVyi
zVJ#M8IxekVH}-UH@LO(}1|uLqoZ$r`Jfx0Wn#jl%E-S+5!ctiCjTo`LZcc5Jhq5Te
z{o}JJJI&}!M2rO+CNRJFT$W#O&b~w2*J_i5^U?K%EA*9$0vjSv7;h4~bN+LX@S&fs
zAMSDL0zV6vc|gP-$VVSicm2%GU?^eGY5i<Sm<K6F#+b*NGdo$|K2gZRU&{LVZxE`y
zju%Q*ixEzUv0=?uUbh>b9(X6lq}hf^S@bI3@*IqsdGNN>ADXB7r)zZM9bFx+QY$%K
z<#}d%!G0Rdfx$3?x9$a~2g&O>NaXZdycR9jzay)aeP_+6f`$_LJn!O<CyssfxR)JB
zH~p~1E4eX?oa}9Hnw2cgA!5@R)f#a$4W6Sk>D74!Bba%8F+(NhH7CA@PN~i~yCgji
zJ}<X!Bt;KS)|%iy(K$Iefa+sD5`Ec+=tiGSWSnA3c+Gf`!`XzHy-K$w9agMxYrbZ~
zQv#%+BzRIFG6o(WC};3F<o0xzS<Dt@#_j<I+FkQgG7Aax{5Mltz_s%JvJ!WQuw|hE
zf{)NM0)9B=LwnRSbPcSi0qq5i2cy_d80G9N;hD@0Tt>7Y0UTgZ1DPE$T@%@m<dju#
zbCX2!&7MJWrxeIiMk?gK?3<erC5v1!{RivXy<EDNUMaW1!qPZp{PXI2S-}ZCucVN-
zE$y?69J;!Hy550GS$8~*(K1+&eCG%8TJg4{@fn>TfM`*hFI=c|<(og<!KC-h_ZVSp
z5X<``ym9}VTS(3>>ebc*I{v^kaOg0{H|1t6=FX~}sIOUHFaqfneCdP0l7?6h)m;|}
zpOCAkM9X}gO0`*)*>a8J>9bd41D`>by(Ae&a$Ez829TJ+$<5vG!C&H1R|=Ehog*sc
zGI`GKJs~G5d;{+Q8w-qt;G9JM8UuJbn<V{A4PdXa0f<v}N_j4ax!x(Rn@tm``}XZa
za!<f0d4+DO5l3-(OkxPm^A_gVqi;E7^CLluIr?3`%vvY2l+3oK!K*-kpe1REY_kew
zltg9zQ4x#p*JvtUIK`YPFKJ^3nl>B(d6(*ygDDI!iv}HdwFB5W0w=W*02%NV1}hG4
zdl<%mA+IWBI@uLk6$HZ}VD}*Yfyf0A6GO`Y%CRdNYk7dYXeX<?1FZGnX@d<=wE}jw
zLHv5})t)qPVe<svri5`!ocasj%T50P3NNZl_5krc@6i*9ut$sTP4~vo1%TDi9*cOE
z+&A4{EoNgP1}oeZ(~=HxQ)%=XKf@_)d;iIEPHacBi~f@}rth8oNqnG480)c;<!zal
zQm>a5MPU_XJmV0&OD8iLNJbI&ZNkCpp(#{KQW>{nEwUVuYP_@ETp;(&Y!Iv+U_6Ii
zAYSvT$Ob6YpHecfDSQL;3TIqo$1U%K+{Ttu_~<5hb1<4I@9vK3;j`Vr$P0Q=9<wOx
zGI#ldIpDxNG=yVnK$!(akFgIA^XB>&+xfKT%XgWFXIy~qg{@+6*PzRc+j${7Fg-wG
zK&%EVbUZ<fYd#1^+^_>}qn35oNh*bTAJsiI*}F^d$i~&v!PF6CKL<jP`M`n#3(#gk
z3m{T&-kdD(o&8Rp?Jjw&!Y&yg6HwlRP6XaASjT@-#lSF}p!*wm5Z>t!DpcxCo>c*A
zuun)`ZOBsJV8Pp$m)5Ze-z!UaE@_#(xp5M+rN{8(g)f4=D7nIQ!g*hoM-S|^Z93D{
zMaEmkwF>`U3Q!R4+00>wt<=4S5n{>i|8i<-(zl?KNJ{6(Bp)QK09<WSWq2MniHWCa
zA+DHqaxI(0Mc0#EziDg(PAeETGNiC#u;yD4_HdyC8hy|&olMZ<Pe7gLdgj4^&xjdc
z;0ZRmk71M>uq35oA`zsyt4QQ@q79{)Mm#cf3AP^E-qyW3<n@Z8A0!El0W5EOcu&Es
z1&#vv1nA_Aj2-$P{kFXp7EcLIcN>)1z?QaX=^R4V^z=A*B<WQkN=`GOHDDo<TPXI^
zjr^Bq?~$pHJ^%#L()S#ZrhwDz&nRbuM4tzCh7GE)QG0BRf$t16<UpY1v|<1b!Z(fi
zR00?T+8j8pV5~-7Dnir`y|N@AZ7YKS5v-RGva*MVWh%oiWU?zhp;5O+m>l+$90N|E
zb!KTX4L_^$m%atoW2QuH{^i?VF%^Ahc=Ii+d4JbdKVOLD6}C&JI329{bLWuz4|*=6
zo=Pb$OjXBXZU4KL+2zJQdo!hUI=%k#H^)51#pC%`0bzrab^@jlz-d}l`09&<QWQnC
zn0;Jb^YK;{8huOPhzfi&xU6aaw5Zbol41AtloR95@EHTdPGAKdIi0ouM)gtkgosF`
z7C))h{oQ3?7;Z^sh~pYbf%-y8B}7<X<e!3<Mbv3Y6~k9FEuHendyv0qvmdjc$7~#f
zySWW<gO|(IZE4T6V9f@VDs_rt)p!W?$HGZ2Ba_8QNGd&vL&{z6^d4-d$XXy57mK$#
zPnsNulL>BfKtQp~6h2uD#Ug=9;dlz(lO4yF2OBh%NQK#GHydt!PB)SPt{=RhZ^YoN
zlN<_j(l4ZfrB!Ca%!!ZQqJa4W^1Kpq1Rntufocs_e1nq<u?f#i8T&ngX4K~I_N`l(
z0{gDMVf)1)Bg81-V@Yed=<OHn_*EYWItCv7e5}p0pUnLZ_Uyf@&zbg4Kj${LVP4Zo
zv)%>U$@6VG){_ltjA!1Cnm|DPznJKi98=)1zc86%0aoCihqMWBc7y9PoPg){;uH7|
zrl=o59-qUK+4!!+Yy5`wx&#C11z@!ezl#T6&nd$PPYqbmUP3W}k^+NblpOhdl0ZCI
z9mBb6L98gqi?aY(z~E{#+-Gva#k~l18s-B2zo45VuLGLY@L$3atDrxB;Ha!Bcw^X-
z|Kk}k$aWqMLXE7`<!PS!ba`{=jq##2iQz<?#qO7OU6l-!(?zIh;iGXSr`z9vw;Zr$
z+`3IlN8ri-Fg{CrLU&tWULBnB!-Ko~Nj<_blb6{uzLVNYYj*34QEo<pz!6|96=Nu&
zs&nT2l*^?;&?W$la7A{COgG-%AaR6>2QL7EiB2*KwW4}H$+L!#n<RQw(1^LPHM}Z_
z>@GB)LLXDXo;}fZT5Hc}Q_j!Z&m=Gzm8*ptE~FW(26NF+Ze2t7OJlM9+j7D=(+LC4
zx4%2%ST<D$8Ca83=P#JP2dXeA^@5k)olXSK(1#(_1`8(hXm3lPX>xkt8N``lzohp3
z57K1Zt)DM0k18roENe}aW&gw~#tnS!?_X6j{?I00?&b-P8z|2R1Yb&y+?Y2ax{7IX
zJW;q4_f)v0XndiD-;C%=ga9*eKLCS|G!v_65}<_Nq8oUa+eGX;Xpt3mO0-e=p+l0Z
zmKA7cSQr`}e-uXUO)Sa$HD=AYZ21@-`Eg;K5d;Z&Yen5t*wjLBV2H{zJLmj;E;5i#
zrRn2MqldV{)Py>N>?Rq?G9s_lHwc}Ca0Emgx0jci3H<={Y@p{Bx>5$ic%2VFc0<bv
z$Bcaa^&&FP=MzUFVI&3okE>wK&E#ZeDh#pU@!&3L)M-x?IJj<F$GtZv^ql-W8x~gI
zz_Y*KIUH^$)=~!F@o@01o)O7~iK5w6%Tr6WFmg)W(`~$cF!8wN4Xxw<K->N4hcP2K
zSdsmD>r{wx(gyB!Wr9dyy;>vr8~xkG|LU#(=NWUCG<j<B^WM47D`!5mS79&3rN<0d
z83w0EN1-btF;SE&L4Bult7^vO{muu;ot?azd0c1XYQRiaCPJ<sLmbo*m<u5qfjd6K
zxD2B|c>gOj`nR8g*@#((>sQC}7_$2_Bp!I}jTk<NQOB%9an>X{bab0a4jK3tmlW$=
z>$BX)=iIT=yCK-1K0VN+!=gxE<lr6mfdz!S%-GTr$ItWt85}nb3J_exVjGV9>;C=-
zj2YGnKm)iqjJe80k|h&G?#Q0c9jVA2+2|6Y#uZTk;tentba8-zOJid0^+0D079;>S
zHrN2M?vlzU?<n3>wyKMl!RDXvP?g#kOJe{YQ*Ag^aBfc9prQMsc9xFcep$c$VHnoJ
z)I@B!EvNs-;QeiE^nW0kVXF_n>@2~;wt@t>;%7-ry^h*U^}Xv~A|3iR-k9Pt>bA?e
zIliB*`oBpy1~`CzcPCJ$-XR8!DO}gP;q|vsTcTAcR&1nA?o~Dy0Hq7KrUEts(Gu24
zcpo4lj-syLB-$bbAeZYP(D4$1;~E6k@~ixmQxF5nzH!8nF)XG^ODz7G%&=ej2A;<|
zPwF`ieZXeAOMPRY_8HFe5Ly^jd&r`!k=E%Yqvwg&E|D~V5`-0#B%vrp`lr4-BTQ;6
zFS5VC1)dsQM+Lcqeh2m<z7$9dDDQgw*>uGAS>0-Pb0b?Zp(<|2rjVjD_dro>lmEb>
zP##<U#Y0}Voa;h?*LDDiY|}gNoq@{@4A-qM471${Hck7<_aGh`A!#>(T_^lMu4XD=
z-l`OnNmrP?Mf?!@&I|+=7Kcvrxb1i92i&nI)th(qV2>4cwNTT!b-(OSA;&!e^Iq9s
zeRu@J0)Ej=#oKbbHc&L&rANPCze`uB&9LtY1dsUEMy73}mbmtO%fIsrey;utPJp`o
z@$1A4|B8-lHPy7*?3XLlQhKG(EpguCixc(xU?Kr4x`!%tckdz7i{Q`>*l)qvnEiN0
zYd<a1NR49j>Z<)7u6F|M@hr^cm6e*1JR|4D`T5g?*&w-x>m_=KSxLrAHydw$JUM%R
zjGk#g?=c}n@4_8UDR?*2HbU0g#?-yewB!&Et08Ycqo}ultag{{KLMFzn^?26vv_>c
zO_#8jN4N6On+KcMa@cAc{R|Dvz>O1@Nh^3~)Z>-o8MKBV$B{FJci~~1Yi<a|k^_|q
z*y>QFTuUG8afsbX9eMNMn+DSvtUm{S|F-#ZIbr-bqFC|N&6Wxujqa%hOIm2|Q@#L|
zdT)ftSz`y$n(lvj)9Zd_6+Z=A!dH_rUVeZ4UUZ#n{iTib8%t`cK-5g~*4QN_%s#vq
zn?2a90L;ig);TuIS*S;0n|m+BR05|i{0cauSaAV$@eKj;zKc3{pK<9D&4Wnz1BgQ*
z!(pT%@_|8CKx7ZM8wULwxB#?Ch{s&xZxDDwKBZr_GXlIf-d@)-!O0^f0tEE0STMmO
z&2&cds-s!24MN`?Vc-e;>rFavwYYF?&WwQ?>iyw)v4mRA6c9`2v54%@=V&HT@X4<F
zgD@Zz`#|>u+Fiu+erd@eiV1Oe;4=$Fvt}IhEaB&_e?NV^VIC|-;7i=#u!pDeAeHD*
zGtf^E(IL2qJnKeWl|WAiIv1cE!-z;Kw+x&5E7Oh_T3T?OgZy*&0|0jdv=2GFz~Xin
zhJUW2;A)cSF(qEXOAyP>4Kv|R`7aODaQIm~e%s2L-u9(2g>mM=UqqbBB>$uMeWmSv
zicFdWy9axRV4vsY?R$llZ>pz!XjbE2o(uAfR``!o?0=_K+o@o&3#}M*t=hhTT9EhO
z0+<zkY-Gp)VG4NPIceXpJh>A&BccrC=msoY)OlD)Vr(cDisbiTa!aP@n}K&B!or4h
zTI7%xiY4UwGq2TL@O_$bnC7Xi>Sv$uy`;ME<YYC4O77S*^aI(Q8u(-y!@CFg&PtXj
zceMdQP{;9R$Y=)}=l2((qu#1F@CSXYhRFh~gK$W(5s+aE91D~b@W4%J!SMt*FnC75
z(#gKV;!t?Qs1hop;UHKP!#n}@<amR9u_piz6Vbyptds29GC2?Z-<KHPP@bCcH9T@O
z7B<gXW9cf%_otk|<}xv^A3egIex5^mS)u*}3?P<z*@8Nn5pUz`El+-~pxWV;ybfFY
z?cX*;3XoA0Gy7bl+@F^Tmg<Qa9pEo}?sA&CHi&_guh>YaA)M(@C_U8$^wEPn^KJ@@
zw&Qh3w+e(cgp;yyQp;2tW08tH@WF_R-Vm7HAZ>tfL47{(C4)2tV@j)n$5C?C4bRDF
zh_r@M%%x!>nMJURo*2vs0qx(4$pf5}T+>j=yM%-oh53S7F_I0LTp206V)+Ht7xcq~
z+$&F4!T-mQQVxTM`ijWO_hdwEilnrz<i&v~2{u`<>qkDC<UkGJku~RM+?0(*)Wuu%
z)~{Nb4cD#$G$Zf`a)-m-?&VI>R{2ehpj|nXWWi29R_$;*+Xe1jE^|wihS$$}G})Dy
z&<FzNJl#(KBk1$f!o-i}T0$I{k$LY;b!kk720kS}g6ka(RJfY~nA60~haDE|Tgx)#
zg!|H|hc8HPA<SH*2M3M|a1e1VXkfgKzk%3BtxYUH8YSK;Y+RaiF=X5(c$-Io2v`mk
zeGqMU*(<j;<F8D53bZbeLHD)(k{e&4Uc)zIGo<4*QeNi^sW*r_Y0LrMB0D7@SrJ-?
z$esqN>F=5V;|5w4+|>w%1Wo{7fjKt^F;NKDf|dz^30lnyE(>8}JIOR-Ob<C#A&gh=
z)WtzDxV~gjK6IG)jgbo1vOAD!7Wn7)E~wzcgTRP!dCD#5gl!N$z*QnYIhE%;>*17?
z3iq)?;RbNcsZau33G{JLNmhh%`KK<}s^4ySXQW&lGoXR65|J&(&MZO{1X}GmF}d^x
zu+hE+a>2NtSg82X{8O4&jbrL7@dG_p8bU%Um(WiYWH<jukpb8CnHyM1D|V$msd7kI
z^?GExD|1&FsH(?PFmeCW4WQ`pTv98dUR(IIg-W^F?myQcz<CGeseK{Kq5$v_av$ZY
z>c2XliiB%Tir}7-?Dqg>GaE#EY!nRpAFGPf+%pPV2Po~u{8XKgr8~&qg)hMU1*`lf
z-=C%=Bb2voOSRLmWvfb?X(A7dN;{wnHCBk>MkIhG#v~_(pSth|KsQj4aw%5nmA}K;
z{4T`>TU^iWeY{Xg&Mb#qfrTT$a{$B*_zhr;&y9g!nQD=|f)zR~ND2ARr#uM^CfC~}
zLEjFS9;iPT{S$#8*5D-G*Q~tvjhD#upCcK~fXrIT@@&W$eOuqV;p^<xj{im1mw-dP
zzi%rNrLt8jDza~tRLBz9X6#E@A}Q)1$wXx@X+c@WmL()hRFpy_bks3JQI;e{%Sckm
zAY|#izcZcX|Nh_iy5@46-|w{eF3;zA?&n_Ks|9a*7*im3XQ$=Aa6@0(ST4gGnOl$g
zgnJVbZLSITsm14B=ntpl?=p`k<SqYS-1~6;7ki{)y44&gY}o%H_2ZN2RW^y6HeEI+
z5C@Ll{8);!os_JJ9%tJjT$4Ce>c00?%7?l`llT4E+~Wf<GU?h@u+Kyn504Gn5iJX<
zgN+B#F<>vxp_Zhi&MUFHiKALAX6Mhzz#Nrr6U1+mbq6;ld=SK0T;7X@#L;=@0IdZN
zE8xA&vf5EONA0rWUIWvYneU5~@y-0Id^@B{9FotVfzXZxXa`;e8v{VAfqFqiOycx|
zH408V9B~ljc)RausL>vrsz#pyWw!7a6l>&cpb4Q=KAE#R!ik@?-e40m0HPnbioY>v
z7K{FrCiV^c#qC~Ia5e7umctbL*DB>tt(`oOJKE$>B(y3c`ZIl_E>bDgT_;1IAKTL(
zm(aV(V;DQoBwiOOt;xq1r>nVB5QozXqMGQ{G#>|ZZgYi4ANBALXyRCF&>up~0=e4C
z=qExdSuLoA*SHo{K2)+4BqHuJ*U8<eTjJ_KO}gG857|OXWqWftCU>qm*hs+l3**nX
zZ{NP79!yX6XAu($^ca{=!&T}EYv{uQ^c`F$K!Od-pwa1l);f3aThf$idsRIK7a^YT
zpMv{&2tj7R-Zt;&Zn+aNoxzx~uhbS4Fnd)-S>)*e7d91^ayw11OVYWrZD22KLSX2t
zxmsbM3=zF7U+5yQm#pnkyT77NNk80hvG~Ur1cw$FX#$icwx_r4HSZtz_#jBHM_OgO
zRBG-a6!eFB;b$kDAFMK~e^h(-@UTntL#*|MYhau-wQ^qYID?29VWR<O_B-2}H_`|9
z_%g?Z<w%oL1D?*7=H6T@^!p?VWNM=_rr)16FVW2;)WXRhgQ`X$+oIKGKN)3&WCf-U
z&`Hb!Sx#0d28A-Tc1isPj_ssIkab{gm%~q-Nps>?N#Qc@fXYf^0c_N5O83Yk&Jh+y
zDyhWi;ZhU)a820s6>0*^P*9mg8`Pm~ITL4Uv5c2cc4G8ki93@dSX!d5#H*2m$$j|y
z_md?qweMSWem#2gMhx>K2(uVlaRbBGa>OkgkZtELT}AQ@sB{9&0Z_biC*nCgyUnGA
z!2SWJm__f#3x@Ua%|{HvRS4I#`)g5ulQrDq(((n9PmOMOrf9hBs<$p2HMPxb>4(h(
zL^4z302Tt$C2dBbp~sXe91@=ZwpeZY_}l$M^{c3fwMzFh7rhi*|F=eE!MN?aXM#GL
zzXs5Tt8Ps1qf$J+>X)sqSG0OExGLaBja=z}*lHJEz{s*6+0i#}ml6|>ubUs_CBwLo
zC^MTkw%Dv*6Vw!NU)SME5k7w5cz$^sO|6N3m#_^XFI=VzdnY)LB((?185<SRq_Ldz
z6=k=+3c8a+VZXK!t1#BM4=1;iYo@4iHt`u-c6a)-90QB<&9QCAX|#N-k}#GkmEi2k
zd-}0T8Pt@O=a1pg)cG|nRQqdo8(<WWWyubU>yPYDJv1diVTX3)=k_63faP?7<rYDt
z9~^7X4o8*12M#(!BsI$x*re>DdwlurC)*`XPhPqe>_|{So)+)APF5KLcsv&9x#5;~
z_w;Hh#-fBT>Cr@%gE&@BYPxNz_>00j<Xh1)YV4cd??YL?#^L#;s=Dae`$I8vTHek(
ziR?pYgK*ui#4FUAefysO{lil#WqLZ~mRwALO;xGHlK(H?c&l-YGSs2w3ALhq77vyX
z(+j3WJf%3E&~tjGZ71!EGnXYmzufpyH_+o*L<82bM9?1#ceW<?YP817V+(<;A<=_r
zYrh}tESdNatMk2KO_;2&frb-9M!Z1pky}6(DXR25U4kI^N%IqHt5tErwLp6X6A|uA
zP-KuY=ukGIjkA^Rmh(-41SKTHKkgat3emNlMJ!sOFyJVFQs|yhzR^GfDo}2!rH?tY
zgkmuI>$|Y7B*JM;3d@#=SYma114Y4t7;@c~krRu|XsZmrt!q5~6YU1?XS8vr_EUP}
zzchCb8YKs{bhj+>`Sopd)jwViuWcS5{OTp=6~+80*;m%^LMXBCptbFbllA&lofa0_
z|78*XTbvOeb2;P7?FY2196ok<kMLPwP}t#d^d1hDEQk_3QJ(;SA|S@}Hi{$OW+Hu(
z<&m&+ez8Z|a^KZUpWAb9RZ1~oZ#1y_!*wl^s~f_6f`!GiVBWb~)pe*pl1;6#NhE|2
z?n-SABU*&RV=fFBfV}vM;@9FJ#IOCG-Axo9NLwOxW+yl?YJzLLXz}~Ks(Q`QP8Hm`
z9IRMHfJ=ay&?o<N8V}UM#Y3Ga^eSU{J-`tOni!;ysAmr)p?3r9B$j}fYy2m!=-d$-
zA}yiUqc)i36Q8KnV$(Z7a!CEse4or2cK!G;;TyR#CVaHRXWORl&QzrN{)Ro4uK7Bd
zMY2*K;P=6E@L=8+Qsdx0O6$p83~?>>DbL}BVPfar!1(Ha*X3Y=0lHBNTp<1CQ*Ko(
z*0cH5gruuG>zI~#*2s?~D^uliU#9Bgk{m*8B^U@;AfQXaTZv-@MnxTkCwTjjdY{+o
zB<UQ80u7W7lQIx>d_`*jN>$RmVBV>78OsF{@<J4-aQiATVNp&<@U%j24X=97&Npw~
z6wdiWLP1AU6XHq^<Fauc@qQz>K9(U3C|H|5qvHVhh}h|-(!}XyTyxOd!3myaP5Cjq
z#%_DWcVOKy-E6br4I+HRPHIE*kfWY5=ZXy<tBQXTqI#qajdv`5YcpP-9-laeV`cVZ
zkMWZSzs&t@GkoKS#nSXe4L-XAnSngapVU9=Mo#pZDMa76(GYaj{Xbl_YI8%c=l?_=
z5af~g(+~VrnJ=DAGz2amQn+~3?kYd*J%Ezpdf`MiYfIcKDR>Ui>v}vSA=Xg;K@FoV
z8d3uAYJ82c!|`Ub^V0;OhTG(qOwA>LT>KuO768G-i^tY?<%>x(HVar0HY8Ym!+9&Q
z&xgF6=q8S0fCeXi?}j?;KZtjqD7kRp07i#V5nqXHHuzdkbFJ{hT!FTQ116uioh%+K
z0waQZvA*8Un&J%L{*!$`(lveZE<1Aw2PldFDgfFDe8o|L8=vneMA4;bEt=F)zv>`z
z7Ov4_qHrU7uw)`@{FnbXPtl8RRFz)dc_)8!hKVB3w9yon_Ui{-Y3s?2gG;2r$D}ad
zg)`^GwAG{y{bzpuPd-bQP{8n~aFssBp)dY{YI3Gl&E~=e;8~)f2hS4lOU)2`-{e?s
z$u05evnpT9iU)CJ$(GoY)m6kiHu<C`tm}}!VA-j@ckgr;oi8O=s<*v$_gr=}x-3}-
z<vU@&<+&MvF~O1~yti7uWf7Hi`RTiDjDmcR?;&`SWu;_(Az_&SrPx+^sN%uBItM#=
z#aG}=a;k@uJHZH8XU>Ynfh*9WzrKzI0?^6qM=U4a7X0o>A5CInMP*X2kcer+vkcx&
zdqvLv{I=91o}YOvUbmA{a@$rO=M{b@;y3dWHs~W5Xds@)T=?;zA+!Iq6OpRKnv4r4
zrWqoKH3|<=lR9mZZ4N<IX(w;_TQ{r^JZIVX?hhgzV7BvFT58`ClsO?kukQxK!irn_
zL1lDf8(gpOlYj#aR$|(FpGv2qRjX3sIEeE1<+SN_`Az)TbHW@eU^S3G!olpr)8ntm
zf!MbajI!7UqR&Mi#omn&zSLPQawTzh){0G+f(1>^qkd+jWU-{JP8wFiIEj;*x3Z@h
z98oKXeqeTXy!HdDzX+ZTAMi^{JDg$7C;oVQ#0yr+FnmpbfM5-C=SyVn>qA=z<7<{>
zII#dK^1)bq_4!GUv_nQ~;&YF%g86JW)nR>~n0swd0|X&uuJ}0jIiXJumpdt==vw#A
zKa?_)5mr2Y#OuXIo{Xb*eX^f8lKYFfx_-Mr0L^@#Ff8KfRa;UZ+q6$I`687WXi?~>
zZeAdF@Yu6GtUZ?mjoy{zUo_^zA9hr#DN|W%(ss(H|8GB>ty;1&Sr@Cfs%iVD=4LEy
zD*cYGlNCFGpF_+<LKc!GqRCB#GiPE&%!j?~?Y-o2xty~+jrezLJXBW_vD7G}1PF7{
zu*hZ`U;R<L6el%!!9j`@?UB}BOrQ&tT?i>>u`*_Y0|Tvj8^flD&p870Z0ESE7SlUj
zr>WN?C{Z?9Q^GAxdp}k4a~c8xJm#n;kXE3i;2S)bc%^ar5FRgR3gG1{^=#$YD{kE;
z_ms7z>8iGsq47iZ@j~u8CQ(JGgOd@}&WtJBntFFj5;CCSYWz%(vPF)m5_VkLN6&Ik
zMZ*bcnz-Za%!0XiO3wuqz5L=kT`DP4^2_u#-3gNe-QaUYS?2%4Uhjt=JGV+X4LI=t
zHuI$(mrOJ7{g;k+eV}tuJqHK8As@L(v-0)29fl?r*7Kv0foKA|b*cG5ndc7N3CMS>
zS*)`37WNh`{p<>eB3QJkyHqh>HsMa(mkg6*pIGF?a2WH3ZZ%5~3qT!JPLqv-&7x=L
zMZ<Lj@H^7qGVyav0grgcT2&Z74l-v2#2*tjC0YucULY6^R>9y1=2@kRo6nbE@PJoO
z_g5lg1#Vny+W<_(aBc;LvnvN84I{KzP<ORbtnbac^ey$wSn6KJOXHl<>309iy|oce
zJS=55qSwK5Ip8-EqO*CEuoR*n`x)PW;qa{Po6C-CAc$M<k4Egj<aa*dPiED%42h)N
z=?6iZCmCbHeP2*fclG=Sk$n5lwj*Dr-{pVLzQ+P7F9bLGnJc6uA&CMFI~X%q6u`lT
zZ3#d(WaYqNM-BuWXw8Z7Z<jykhD3{jAqjkk-4(D2Z^VgpZU-JL)Q;ZquHv-^*QI`u
zjodFG)s#rj$VnsAGJqUF^2I!Y$qOG`T>`Qot~Z`}bPl+g$RoX9g=BH@$P5nKqEdGU
zTxQsm0mH=Z2;v}K0{aC4j&RX=`t+h;5`34@5r9ilwHUf1uvB@==;f2Us|~fc&{Ox~
z|H674AAAGrGwcmJyD1h~h)Ak7lQdB0nTu(!yS(GVVU<m`WsX`$m7)#erTT6jv^13K
z=lo~%ecVrLIPykjnL}lwz4bN8zDMewZ#?%bPqG;c#cF92;b3>|9vuD(RC}voQBw49
zoJzS<dt6CpzEaoKCi=EzRh;TCcbDkxStao+_T$svyhiLPG+U4lgQLt<1S`Fw^wjDr
zL=kq|(OLlTpM}e?@m_+?S<zi6Sy&OCD~r%-Z2dH~wBBKhPE?A;=TMh1c_1>#TKb^4
zoVIZkxLC+=GgC?;pJ{7{?5I+fIv$L+NwLZL^i$%xqc{<JDoxG38$^%`v=6c_Ur~K4
zzrWIi^voljD(HK;NOmqL_ZJQECxf4ZZ$1aTgCGvqA0KR(bosNpzhYoUVPVm_Hl@I+
z4(^3)oe_u$$U-;*({LcKu+1LS2w(e_Lr|}01IQH**VR?%f>D26zYk~<cSLpGR7}Cg
zU)z*2bDwPb<8Lj&vU@2}Exdh<+{8m`m1I{9Dy$qndNcr_LsVnYx03b$+o<7&d`;gd
znzAYFBnq=e_=*4%zDB>VHU?oJqh*d~hNk36sgVnb=IOAp3=Y~B!B0pE@=SYCfqV*N
zTN8*OvVHiti5z>}Bup4+L;<w+uT?L*>-Q3slGQbgwfnWRPdN#dO$2omO&E?vVq|W~
z()oZnjyNcTwNWcr0vX@$E5w&aa5)(7D|wQ0$6k>=4dQJ8F%9zz$OQ}P!G-!f1(4jv
zvL^wgPnH=fL6yd31aBQpaHqeG$II#h3tG3nq&vw5*#Mw#FS5j%EW~A`$M9IjHMP^*
zN;9<{i+<>O%lBJf`z}*{i;Q+CDc0U%+BC31?!uxa#S0q?|M;j$qt8D?Ij2;8Hy)>3
zw+Ssp5hBhZiiORgoM}<j-N$eIIT)<sa;JS-i1XPM-6n?-MsMJnX6G}1qJPKOl_QH7
z0OWuq<ti@3xugeU5yA6s#R*9`KOPbmU1-G2UzpevT|Ibb1!Mwj#sID<+W)j;$y60}
z83cf$_hEqwSW%)s0u&qJp1Km=vOEEU=j$YL_VROi9^mt+*T#HIpg7olpeH~-3&zjT
zl<8+lMmXj+<Yr|@!&NVP=ZYt5!%JTu+=D}^Zf*DL`w7#{Kq{~jk}s>mrmidcz1qhI
z!FU!aI#xC3r2J(0c0kTv!>Pv4_<DN^xlPI)d(}AqPKChb)0WXVOWKGWAsL~h1QXAl
zoGl-cwi;|0<QNt4{kd86h<Uf2ba2urd|Q|!E1eLU5kh2sUhu4OC+u^Ms{D5qblc9g
zEmE<EIiWuJNV4w<-enxhP+S>0$;@+~gtZb#m_{1BfC<=nDiz%Tt#|PnQ4fvo0<45J
zA>8_<jFxb75(T%whG0etAF_)(8fWU+rfL?lB|)xHLcyeT748W01t|)_2MaR$^SHa{
zPej<|0(4GNq=cBDp7JWvE;x;Ru7QQi=T|M`A@J&2Mw)0>T?Ayek23js{gn;0&`oAS
zP-eojrwp&U$k~Uf0=xnBv<DY4{MERCBQRJc=emLeEY~4#w?|QPi4}dviTIEhNoeZ@
z{!>Jk##||P4bZiowYr*u38vtOv$?r)NoMA>JcSv(O5*`{0Fm#oJ8fv7HDQUux?kZc
zZ3*Kc@@)yRkEO-ceL5@Gq~>I#%0>~;NS&`JL4?w>RnbkA5Wr7L6^l>Iv)K}Za3OeW
zVEhJngp*iU#Y{c5N3!raHnkYDQS2ZIbZ3cJ%hp`_q~9D#qf_&p*iQQra*bziYsX{}
z&s`AKGPYHPPv$bQ9s@j!af|>Gw6uudV@rDBz@xqz{JBU3d*SQ6b?lz8A1p$M#Os^-
z8u7{9EV~YJUTbnIiTZbVCNMORfI-^Ac|>?fF6R~Z5sHs-KavVRDaD0gldF!weq@t-
zTvi*)r{FhtyG3n6*7M{tm;9FRo_`2yM6+d9QBNadd-L1qfoV)`2X$u!gc|0wChUJi
zxBqzO-`Bd6{)dudAf<<GqGhTcuocsOGZ`pM&94-h*^6?GsnCXCV1&gL4p|~aH7D`f
z6UIC|g_7GyE3h8Jn+a?hKRkhY!V?f@01HRDe&jMYy))lqB}_k1`cNtHZsUV@8y}E^
zG+Ruh9^7))Z9YY#qv#8Rg8)TH(hcHj7uqgRKS4+wx|YT-Kwuqvwpe_Q*lhN<WD%>6
z;7R5@AT1a;u<OCE@7nkh#01F_cXy7{{Q*T!7ZGW>Bh`ue2memg#F>K+t}e*32R8@6
zOBfD9crNil?Z<M0MZLzs*jL6UzWgMg>cVgi77VtT9|?s^GwOy~-+&|)r$CytH?lbW
z<BiAO?wL>jDWWBQ)0yrL5=ycqO?2k*C^J%HIyIq>ue{T%+ipawBqQS{cVfc0OEiUY
zmR&<LRLQXfMi{EZJaKvIb`y+`vF*T;h2x~F2)3^m?{e2YVtG7p<bHtDI9_MG)OgBp
zp$X!A%zoBfnrCZtI5YBtz5jMBiGfTd-R&G2?-)W~Ry>&hN;uJhrf%FwudCQna-NP?
z6=oT{O^Fg{C7c9x`2q8Fm_NWyh{6O$K<q2<M5EEe>q|B7A}%Ru@!lj|K0xufpxD}t
zR@y5jW`I80Zn^_qEZO$3>dGQ-WtGDM#R)E)mFDq->8R5HgLYldEB3-_0tz40kILz0
zX~x&JGycE@b=R^6zmAfWjF}5CrWxNlJ=Om!1n7XXiqP}ugW>CP{uLs;ABk%C4j<Bp
z8#koR^c`G&v_-N{#P`obvwMthESPuf#KnO4|HPE;w;fR;#B8fB0P>cR2~xECBl1^&
zqy=<Sru<xlN6~>ioC#J0&K(os$Ai#Sh;33$6R*AYEJGWDCxkF)6?Dqmq%XMpAN~Pc
z1)w_v7r{;C6jagKLft}=OKud%c6yfC7KGm1g+}cQ4a5bC08|M{*_yCU{ta+v>4g1-
zho3zTCuS=S^230zGeE@)cm+pIh}{Gc5)V(@#O=FFIy=VRA5pUnXJQddEq$V93ue@v
z%OYp5yc#)F5>~%~5C21ZsEU~oMvIZ^_LHu9wC?RSK2n+l6~j`$W?f&<+M#igk-rLh
zChJSg{g1V&ZjT;N7v*ES8M5Zs)BCUGg>`O`Qt(Mh6l%?hSpy+=s1eyVH2#ZB<;0oV
znEJzTF8tr(fSU88jDNdTb;G85`ufMUcYCgQT<R}2bY)WlIBN3t(9RD6(nJ_J%W_kl
zKa~s{Kunc*axqnI7V}~!<e;BEzB7z0+91zslK<uF_!M@GkJ#o7puD&{@PxlbKoeJ8
zZ0<%5AUZ5ebP?t+;*`NKzo|2ssKa77;#npHL^qbQ-i27U0g@#!Q_a;NyG7Bl%aovB
zrq6*H9ya+HKL8Y3ZWCRMwGF|DmFz<U%4(jXL1!*bNr#K<eC!T9KMIlTYWy4=A|6>6
zav{A}Ii*7z#1XVoksl)B-mjy7@)f<_QV!kmeY-Z*%|~r-?z%rQM05Vv@42MrYug&R
zuDMfz;|QkaGY(zA62L}1T@mY|^74d|?xaH?*w?~_7f!r!TBaUwP<N6)_4~_$$FPcN
zu>VztxO8@Q>Te)1DTsOEV*0t~#Rl*XCdwZ;vLH-6TIfLK6~W`o(<QcogQ2#VC5Hk4
zQaJho>(s*)=~K<830UDR8UvCN&Um7VLN<kV3uV};GY^yc4i#TBjiJ0YbHP_`8~HvE
z(O7st8Uv~M@I=qy!RUnn4CerA&%Tf#P$6x#dkfSA@6p=18Axhye4Z`M_h56)*b^aI
zJk$rwAMEh6GQkud<l7I9LBRy)o(1h3HH$Q#OpT3riT6IITwa$`>MBCu2DhYZn|K3O
z(T)#-6*WEpd;|otit^P%-d`HPMYc;ybQv_2rY?VQPHS%<)-wcK+;*7WyV2D7k~i5p
zbeDTlF5mWMb1BV=^e}n(q3e@1<Un)~lL7yW%ztdPvt@opusQSi+-bZsEJGgr1bAh+
zIiSRn5+((fzW%fpmx4BnLV;0mh=U!v23gk5DBFTI^@E$kqH<-M65rxy#t8g3^#(-0
z;S)EklX>YvRmBAwU$1u*rBNL7jqqe8ilOpWlNhtj(5@HSVgzWw%^>Cif<!Xj8`dyX
z`zr<;N#F$lW1$3YwG4m{6NQK5LF^!5nyiQ+4`%`Kkro2_ikgoy(lYXFa0~{(j=id8
z<z2PJ63>coZ$pyH^`Tj{m`P`I+N!-Xn_$zKm<{RpK_jiX&`o~V8TvKI=vHrXFy*{^
zX|eu+f7b_D8KU)%JS2Bs_21|qnCMP&y*58g*v@gc2X?((&6UYucJuc!z7T@1{kEHy
zP*k?-PS%k#-WCccw#0HPtuLtO!qQ;B!!odBi1mhilI{G&*VU#y+llf^n~iY2l&0LR
zN~obw8n6-BoaBAIL*uAf1>o*LG3*KOw1PWk0@hb-1IAb)hLI==6){01D`k+mtzpB}
zLQOQuJ2>8i%NqA5mIuHGK`ZHaTM3vQdm#V_XfBAE3oruuyfQ~Q-nj8IJmL+8+-5j#
zH}GL$&*IvF>p-qTASsJHmb5_eG)!|e$D7o)RTYC-kIvCQjM$Z{uEYXJO8%@XZHS`R
zBkL=w<26Xq$(r6a|Cn3<2Zbl?LB&@eO7>kqXF#|-zTG3Qt-kGS?`2n|;h;uw^TtMc
zt&Pa}EUkNmap25PHAcWkkd>}cyLR>?f~FwHD*Vyg%#Ig@@G%YAC54gLG36e+l_<61
zafR7GXKDmh7~DrF@|HbygnT-tK5$<<PAHr-|2VlL_y>7qHl>XLYD>fxv>rnf2{4R!
z|7&X-*qd&={}?1bsL5|ki#v+=1XZ)st|Fcb+#rj!u-?MAZh6*bbb1W?^BK~uWPUCJ
zG{ffLq9FmDv8ZPFigCegDZJkc42w4+F%#6x99i<y?;t>di8EN$!u=<h*Zzw_3Tgi4
z(64M#b#`%qs`uI)8Rm=yI%&zac7JC(&btOqAmJ~1o!p<0GT9%vq~d@W>>2~R`g^CW
zIe$zZHfXB%mC!k7B&UFtfpk`bMfC8reyOqoqFa6960>yd9e0aM$*z$QT0M1l{Ks$o
z=-=Mj(J+Jx)97PV$2SZ*<jWp3YE#8=gxLy+GzRJ3@uX{tThFk}u|-wQY%X)*If0tI
zcj&rmClId=Pa@k4qbF82&Yi9zhD8^Sf=2|N0B|Y{x~PZ5agR9wZz6^yjaYSBBm<sy
z1keI>53CmeeOa3YlGsY&o8pwTp()Q>8OC(B#sv4}PS*uuC98;0k9$-j=IViIY^uSN
zDpdrIhG+CFtcUKuQO1bicyrh1v`H$4y{cjb5?IuT^JCLI(4L*hY#2Psf>DU@Zzp9m
z4Xn_R7@Wyy=|Y#EN_Bfno03}}<gC9wQ(((~aKnV!{nw<zy2>SmkZL9MSQ^%Mrn^k0
z_YbcBk^+MCs-|Gybvh*v^N?~;aj$n}znMidk>OhYE@dIcSY1NnX=naJL|r0lkE?uU
zg&sc_u@wQ$6*e87W%yT5d0CPptD5@8#pntcfarKc4H-zf*eL?&`XB<>2(tr7%Ag}+
z>~~VGMR>gLya{Pp9pG8uouc8F*PX;S*dhWSD(C86B55I+a0XFI;1uC9!zlzt=g3$m
z>dS66UxU?NwI34s-HEY}J<*0M*`_?|3#mWlJci1>kz2I%8qh~TfPy0|)-9$1LRtpl
z3_1na>nBDW=atCJHA^8&p#wq1Lb<|M|BIkv?aYxEhh%V*9(_DB_#<>bOr}(@%*UnQ
zCjTWY7EUmRL}u?AlET#F6kz)(S59gylrr}TF)?eF1Vi-7`$qcq%+}x$d6_lWxqD3-
zpU(D_O^bdi|91_$z+%=su6Lr{ol4n-YX?^bashwg9|Sg-k$8b9e;>sA?FLR<HkB?C
z+S%!!^j@Cyh&me>(;UHgo40`v>urLZNRYrLMp|@;?}?QQs9->}(7IL7_R8on!%a^q
znh@E}J33i=ug#XkAQs?-j3X#*pymT*2FDvgnE{18PtJnC7)&BB@>7;!?;zR$z-R29
z(^g2>I{k4)QDQLf5nvzO6|kuXy$2K-EDk%0Qyi?_>fY}b9))zVQ$vM+QzeiyAPB<l
zSamaJ<T$l(`0HGiwWc@v3<rI86MYx7sUj%ds!`_|B{;))E{oc~G%I*@y=geJDzREw
z#wRJU)5LDx<8+~<t^Bp$JAKv4+2rzC9vGaL#q4}sk&%fpx3|FBT>Ab|uNT8D7ketN
zTk*VCBLY3z+7i;p+90w?+pzLx%<J*oiZ;MQo3}~D(pbZCPOiqvmgn&VYtiUCegR;^
z*@+KJre}u2>?njH2zXtN?A-6`=0wL;ujS;>N88v0vIp4`024tah;B6BhKlNQ5rkQX
zWhfvh9YoJcv~$E>7NwO%to^PT`&CVk<`DO=n#1;I0W5SE)k07S4>@)yAXZ)=90B*k
z9vK%Edv`pGcrI~B;km>&9nC3Iy(Fs=ns?Y_%@#h+s@a<@wMhO1-+zYaWo!Dzw*g<6
zT%EuOAf*G!Mc<R}QN+Q;)DRwjLec%tSUag18r0#Lo~;;G@bXfwsL$5SPPLFgW;cw5
z{;&9>dxf|5FWu9hFqxJoXnhDj-CrMWSLuphy)LD0WUYt?l%L|{CchgUc&?~HJlV!P
zUwm`A8J-2H^=s}duKN+=0HOq5r}^Ir`Z92wx8J@kW`j~VdZ2tuCK>OEYeAqCf`TPv
z;FF5^S4wWp)Fwmx<{9Y-hh#0W0W)CUA~qJpQ;K6B`UiZH-L*_MNJYdQMTgM-eo7ms
zox6klq8f-C+6IUZ4tI^qMSgvV6oN<&B_B^J8WUVjph+yQP|m9vmAZ0Tjta--9Xa_P
zXpK>@fQBI@?!g{RV+P#rUTMh7!FJTC6U2RB69C#5J=<V5^14Hb3(GKA<1`O09XLd}
z?TQ2@cCah~Ax>^Po9cB}?$ZgIK#H}k`LL;VOxEtFtGN39AEme9+Q;3=38&h&e4((-
zR<+*@HbU=sl`+^p+52j!1rRKLMvS#+)7$U*z%49tX~su_7o)Dk+RdPKM37VjY2Xh<
z)@;Sfa{vo-wRbl`xOL%>w;~_d%|^LFStL4C0vZ7B0<sJdM0VSLEl>E>j&pG~YqP{t
znU!$zj<^kEHg{g?Vr+AG<mA)2vjZ~Eft-Qo6P5?)#5TV@grSTuEH*)$1$7&BJP$q$
zEBy$+gVnKw*#XG3jZB;gsJ(EtMeT(O5-2hVQ!KYt&yN!Efzjqn6p1U``^ag2thL!G
zSCK-0wR@N1Do#g;8RrC2{@OxBCa-42iM~qpaufxHzhXq=vCh?M-~TkX(_;0fy+*0m
zoPO5iZeToolt-eR$-Oll5@}Nwyi|Si@uJKB+hq?-MXV17=nS=BxI9A^fO`#ADAM0B
z5eOJq$-`eni699u?y{Vmke-yE@17K6<J^fK1;)W(!&~=Vq1`PZurvv9K`>oHJ&c75
zE)c=_93HkZ<JAUHCH*N-fwyWgY1EWBP1vH?hEvc)@Ot5R1aBUne2*GlD|m0w+^N-J
z>zG}X%15xX#MQ(7LLZ2po8a!_(gV$5HMv8-`1KZ?dSWceyggV2_5yU`nyWux!ON2A
zB2Yl<eos>B&K|w{I+H6U1sppuXuIPha&``YxjK7T`5)IZKyl_~R$AoOBiFATxh#ow
zrlrHht+eUa@3YL5|8`wDgAFfZWa6ung3|Cvp1bodbJs^L71g??yC_W3yuSa>i0d!#
z$1}=tiJ6L?b4Ts2Z1@d^^*&kz$*k=NWhDGP0p>CTL4r7ud5%R5w3X2lHo$8J;2YB>
zVI5e}EiXXQSUu-UdsmUiO8D%{>w?9w51DZ7Zt+P8HsL0M(3u}nh;v#ecI>R{HAXf%
zJAyN4jUrv>w-I+Q_dIqI$><E>(t+>_jUb>40A@DrqiGPfa}3926UUyy7ziD%clYNh
zCB7Z_s~)Z()(dYRS4hY&WI8gYoN^5+7pw4b;*V$#4(%6W9z!&quV_JQf;)fO>;44c
z03xO@YkX7f<xR9zXaAuZ-lq2tci-QYt}MNWY}KPE_aUH~B6s`K_Kd{jio=uq#}X3m
zbU;k<UouQW`aGlMiE=W_#TGd#`!GVl3_XB8$M9-=+5)n*l%%eMpfs!e2I`9(bGGAc
z1n5t|qwt!!$hIscPLOVI93QBtWybwib0r6=X7>e|l><J7{SjnTIPQRd5rz5t_iV{N
zHYXX|1@27-R-;it=i}ZyS5@lR=z>&*L`5CMPh9>HP+f!D1?J~C;tmk5O^A5_@!**C
z?9^f{7Xj2KRnzs&^o*^rqA4Cksx?$D8~BpWUnf?AmYr1+$|mWxsL(Kqy7{eOdrgb`
zt^D(m5UUV*6=(a8>xe%#YE1w}?CyX82cw&>C^~S#_SF20%lWSFT^uVWN*=%Ms(p}j
z-_Yl0x69v2QkWuS!WzE6U3qi)_d8sDH>u2@r|AraZ4UEnd*A7RV=B2B!MRG`D*oD7
zG@0Dz_V(@~HSoC-V2p)lG`s%@*Lx!j5(n^L0rA()g0gY$M6#l|m}4obAE=dRXCXQm
z@X~6j$As^vvGX!<1mT0N!`EZmlDNA6P~DrBPu=JqWhby9K`gPZVx%Tfhn3@=$TfpE
z7vt;SZZR;A;cF_jZaZ~nM}qJi(yEk4cUf}1{Mlu2JR<@UqoVj$3+ZBIP56CNuwcUn
z^D_$rE;=SkBtsb(73YCF7+wteVP*>T2OOIE1CEHw_?vmBeN1(E>>GDVhNSLL9Aw(|
z2+)MEfoO&vZI!Mz%xH1-QSq-CNcOGexs+QyAFx3tsWD(ss5hm8sge2Hpzu@>tI5s0
zVd(h2Ki~a&ZuOq4W20`~U#WevC3}gGZm+lkYQm}Nyh(x|b4!J}Ue-lFkomliA-3?%
zDbX)CqxCpz5aP4#<}4v{5aJ?8*eIGTpFdlw%J-Y6qqoWvz6zb|QiGQhWBZxnL3B)H
z>1DSW2y%tprGHT))ih$Mt3$3k$r^wiGSA@-20ZX@!7z?^c{v!B?$gdv*!YS7SRx?d
z{N`+6e1Ix8Qs$^6H3kyxrToHW$6lh6%Ry>e=<xuz5jsNN#@6zq&MBu`v76x*7^QZ$
zLUg}@kAncX108c8pwhzd2ZaD1us1=W!Rxenot4%64asCF|0P5MomZCORhUd-Wqi<5
zlHnaUJqv*-=nGg75plZbHt+?k@v->ZQ-p&ln<^G_cI!tw#s>R~hd{RfJ^v1UDhnhT
zzL=9J7(IP?A2JyY_w~62mMYFq*s4A$GFS`p#{XRFC@)^rYxJRiwEZV*(EzfHWFP(}
z$gt%)Ri#UpWLe~E8mC>(I43ngig<@(8y|q88aoZFll>$~L@$frz$7`bVnhv+h)W1o
z5_$n#Tvj?Mj`pOgeFXoF5H#9+C?y_NBzz*3lx4mB+skOc58treERtvzC`-ZusULw6
zOTE=%PkF%6a8Est#jNFNc6LV3vXbHqA^*FnP|y7Bz2j&X@cVW6mq5P5a-;yo8z0Ot
z02)b)+!C{3HM3~KY|p~}nOEp(h`@U^cdQ{<0q65(LA|Jp`5uKWM^(=AOu!E>_3Dog
zHbXnigZ_+txXL;)HPeY7p}SC$1T&BGx!g4DZcR++@r@dyQa(rIen`*ApeFfpSxt6%
z3_Tus+hRfezn3~(?4^#^AGDHeJ`R?sK{g8gddE}ZfVlP<UsSK)orQsyiuVUU4Flv`
zYD4BN_pTIwRoEWrI#?$eI10cqYy(ahcZb5eZURZbRzNiH2yTrhZA=yBMG_|OJ~mw&
zPkqP%H_wACR!)?uumM~j%LL}1o`aT+z$MUGu<4>eC&(TVaa-wV6<FbgWcy>#_$dRs
zhBG>*i;gjHJrP^K6c{q5Ho}8N3}So2(bQHR310x>$@a~Ie2UQ8^6lUt(lp_|$(hEA
zH%}I<<V7+5=4u=jj0mhV3u%WunOcU&{`KU|yf1ImpZ+PYufd{g{=chVq7Ex=ekCa@
zW6&qe&0nzkLYu;6L-U@WZ=-+Rs^4UOf7V2bVRR|vO3XQ1hOE1#WwQZi+yGsYv)4Lt
z<V5()U;kM{m>1xgkaxAp_eo)e{qR%Q1Che<g-H$<Mso=(9TuDx=+Lk0+YdzGM=E@1
zB%y#rj`q=5yw0#(Ty*b)8c+DiO$u>(Kv91x;wA1N^+M-yXbV)lHm!t4(pq1ieAL-4
z@)$J-R*w?0L75G@2XSpLI@KFmMG%yhDo0E}(g~4iJM&oC#)i&z$FTspb-Qf<`x+^E
z7SD`2_3$DNV-~`=`RWDZPU>PM&Qx4S!&iP-FieFCZXu?X3!W<Wr@u5ed^@*QPL5Od
z@Fg%`=OUnrwbE(XKT%ZXXknV1mpCD35;X*1VZ_SAyU#4LG4+2A!2eq8>NdqvRJ!Jv
zH3da6S-lnndpOe3#$XN0fhUm5y)sf4FK68J+Cs={#bfp-Ep~^`fEDaBlpxIF+fim*
zaK_K+{P_&iG$t#i3!<=EC%wgb2(}P5PAkooK^>KnZOYanI;v~H58!DN#Uz$tt5*{i
zbvWJ7!gZI+<q7B^q|>+W{(6?D1ZG-vWEjHm|FX<t0eZ3RnLKiFt+2wxm6`J}z~EaS
z&=@H>%z+rufF1b1sSlcss1~DU9jvtYG&Or?gGA5B%UGx{CVsvC)d_8fp#G?Q`wH;}
zT~A3~_BO^$JJ#i}XBoumwFMhEK6%OHcv~~)HL>X~|D=tYA=}HZk?yl%(d4qkId5%-
zDv17cq9zvC49zIULn%3_wxy|QMEW3OiCJ5=M;bLLudQP|1al+Iei?29!F=J{BOAcI
zIB0;a*G@9kF4w7RvA97VF#5#-sl^V<NCHB)v)IQuS=hky*(=DNu{(zvSTIqk$(r-|
ziIQ`0<DhLWd15Nrdr{>Xkpqc^JJ!6#XACF=*3jYEUo>kGuVJD!I99dtn#1)DG#ORR
z{bkzXKCy1SnGs7+_-m%9n~eL^1+*=u*Z63Nn)GnaQ(N<odqoj}RAt^YgiDpF)gG5j
zlg8nZ`CAQC22UK;^kH<#eSLiNb&JKxzlC!dza`DD4%jO*+}?6wdl=-b-nD;uEcbSW
zx=;G##xCA)(EeRW;IUql%ciopK=2Kvx6;>^`m&7Kajtf){go_?D<%<HJm@LboWH$i
z0h&D2L)2|{)FAdrsCSsO;p#T7L%f2jV92rLp8y5Hb^|B~AT`+?d%k=Szm0+0BKEl5
zAEz@~L%7sNN%+5kEgPOCw4(?E#`s9k`o3;u?Mkr6M&WdxLKBQ<=~Gp+S190pL3@t1
zNKQSy3AR4+(<jGuLz{IvgYIn{6yYX3ZUre>xu*Y8uW{u*78%l?eK32^R<%$KM*@qO
zyJ9Y9BCBZy%_hxfPak7{$J-!MWmoV0<gC-uJ^!>nN>qtoa*e<2hu(+e`4YzM&=Q^%
zM<35|&Eu^Bg)FT=Sg=XDoWBHPnvh;2z#!3*>o_gxG-_t4H@{%sQYa1lp{Je)m84mB
zFyJA!AcNi3OR<=-25u}#GePq$umN*JN^Z%y9XJ>;39RK&69K{K1zIM87jUD!tehv9
zY75{AV<<JHq_{Zod%6rYd><fNlzl4D?SL)UQu&!b;<|F&^2_j;b30p_N|RJMd^45$
z=0+3<zyfX6Ym>(^Sq&TP%E${E>2E(UtQ}oE5^}3+m{r3Wg2}N;faYHmzE~d)TGY+`
zJBO7Dke<;qUYMA$o;P2%Rtb3gRy+4V8G#fx{mrcWFv(uYgIaOsh9om;42fh6JUlUu
z{`kru5qga62eiHa(hJQCuI6t`2I$tKLigHsj4Z?~2bYEAtWNYT8${rXLM(>B5hnR(
zY>bAg6FLkQfx76TWGrY+QUt=8L>?kkwT2~~CQq9w-?k*T@0N7ODj@BEXT0uHg!V-j
z4}yz|QP1J(3!(PD{>s{ls6Ti2SX2hP&pRT>11n6iZ-x63T0W$tJ9?ANd!%ItUs?`(
zQ<+=T9d4&*Ot*zSnYhqaT)T7nY0`gLcC--KUizP1RDGb3Ard_GXUo&b!wOlyS#|~j
zSU#|9tze)EJA^#BU%$Gh5`O#@uezI4bM3~{BiUVArTf~VKkwNUI26=7HGBMcSQu&N
z;cxlJ#_5ebYg8$#IgWfe({S+cN!jDpf(9owd^CCJQi7^0jO+IPc(qX=RhIS8Om631
zcWNk-)*6txI=Efv3Xl3R?2QJiV2;e9-#SRiBH^XG+)7qh&C;T0UX>sAaF?=;XaWRa
zEW$^aSLOzXg~ejij;asS8V-UR>BdP}vv%eKA<p)4xOfEQXv`VxpnDeCok+UfQzAo+
zGT?e4bg_5A1Mh<(+>m}=|LgoU*|6o^ry@kx2gaScFh<%QKe~b$8u!Uf+O*3!SUmNN
z(I30*Z6;-c8a5o2BD-j3Qj^D&PDM6^b=b%?exA45`L3)?ZL5Oz(25ePC;Ne=&1|`=
z0xT^E!9~5jN%1S!S?p!#K@!n{aR-nY!Pm4e=|mZE6_K{~NBL-5M$93iR~I(T?Q>&h
z;^yO!<e6Lc$QS$uelD~XI7At97Plv8es>(Q5lA}4;Mk(W8R3@Y_xTMArvcBSzW(%2
z4UBw1>H05!-42-jLgppKkoSTg1sVt$Dv}F9T4A-f1hP@)2pkN}m}g!x-nCs`zM}I{
zQhDP8cY9@d7mr4`3*>K|d4n~;qYW1xO_~b6y()|4UgXY~JoJbhQ+0Kt@3P*UUUI+O
zhC~jO=<B%JaFl)%gZ_gQaP?X>)J?OSX(4dwdeIs@o{~9T8sf(Cggm;Yag;G0b!9fi
z`&D^t@9CdcqT?677~BlZ{OebxsK*kR7vXV9WuoctB|;1sC*H1=46tAwd=WUwq3C8g
z%S3@wZ{5t*m@2vc6a9EgNK4eE^%5~otMp&Yk27Aryzne&dXjT|${aDSNeL$QO#!=-
zjS}?8KD4lA$tWY(9LP&2ZPIF28h8R|_u(2Z)zP#7H0F6I{rCWj1*sH=1nOU&d{dE9
zU==vxMImFmycGf#&G?U*t-fjz`ff5jZhBo-2<cpG4iBbraFL=UPGpWVK6^!oa;;mw
za_&2;S1VuG*3dr=HytRPeJ>{-U=z+vpi(xD-Ffel9@zSGxOs`ZOzmGIy&78TUKD22
z4|cGRTXs2ZF+<#F@{yca6;CC`qjxx#aEw{XUh_K-HKxA%F)~rZBjT##R@;tiE&+@O
zKbhMg4MRiOrYfW|S3F>Q?xV;1;`zz?$GQvy9ZEXocbb%UcqZ|SjJe&^yAl_-`_erF
z+Pk=xp<wO#-yirG>8lGQA`;HL01I+n*QDz8QIB&3lUC?&rycwBReUB`!Fw&q>Kh&(
z)s=XB*k1JPMT6%)2FwS?R}GFi=E<+g5!<P_%t~>YtHUz9W26>Dgc_T_VvL2y1>499
zw?E?Hz#qV-KYJ#8m+0yAUea<g2ohQIw^px~{d2uSAKK)UeiM~>nV5Sc^Pc&ur@hCk
zUg-B3P5%rxQ=gn{$;qgG96EkyxA(6V9*H$o*{15L!40%CkaHHftj1M)xPgygHB5!T
zN#w1O(HTkk!P7m&v;>|>>S$e>M;f3@EC|sC;_BgM;{$FYmKXgSrIH!E^&5bj0zDPH
z{_c~CZ(yB6in-rL2bWjDSp1(d%)Bb27=8Cf0Y_tmotA%UmXzL)Xr7~hWx#L4vw$ZE
zJ4O)N@w%~B-Q*kld(CH0fKH>i^>R!})>ZfL1>%hUPk)w5DxPyak8vM04G%e<RgB9Z
z6XA2<P9gQnnCkwf(#56W(RZhE-$3>Z^B~?9mBn?BLiA!dK_!Xcs+=|15FfErJ*K`U
zPHdw8?zwYaqt70{+{E4U<oyOg7vlr(1c>W8OYhnEWKU#fopDI5SrpTdq&~BJIfqDW
zYeJ{a^%afu?%MYwnZqj|YaUQvaY81XV$NX9d`tb3l;Omw(^0b6IpQ@g$PpA@i26GH
zch0}=XNl8)u`TP8#yPF!oT+m1a99Q!B#4ph5O^G+_y&QRe2;uBV??QD610|{1Y9B{
z-6(aeLi)>61MTN2UzVoYTl|%9(^6K}$|18>&cGri5mb>`C6D0rWWiq@y*qfizym(R
zax+A^0QLutA-aE@2hWGbVF$$DuyM+t5@V`8C%d`he(9bP=;bvI4t|yDD#*-T|4SNn
z2UuKVAiw~fYKs5*z#RuY+vJFSy00vf-KAc6`t0n%TKm-*0Du0@EFE(Dy*f%|a2rKH
zZ_k<#!NMDL;z9zWJN;;fr>~BWJu^^SEGyFwS1|9rN_|;}R@Wa&{><Ad)?F)@6dmTx
z`($)Ur_Sn$b)f;5Uk+O|-f|`#cXfewqfE7TCQqjJ(3&m+X4M(LVrn5x3KdMqQkRwU
zu@G$l@?a&l3<sI8`($Ix0^(uir_R?e7sh59Y2}vS;jqVND1(&($i~Yy@U7cue<S+g
z&$p$Ls~dbT5F37?)q^;`Dk~v`wq5n4eo<*AvYhP`dh_P(pRCCB9cWqJWr(<lWV1HY
zT;;*ACUHS7Uivns`sJ2>G&L-x6Biy|Lj3}M2c{;#XIR0Jt)wU29_+blulQ&ASN*v;
zHIauB4Rt!30(Li#mzH2k{ONCOu``Le!8s|e;rtCng{8_VCYvKorK9njNX-G+=tiek
zyJy)}p`^Y|4?wL6cv${5B3CiCtve?oE?A;-0{3__xLCTJz9p+=tNCFmor14i=am0k
zb*q1*HhxX-+mQOJca?4w1}8lIkn+@eCj9Q!jO2u^GeP^QlsU#uQLO`XEV?S_e>~j{
z$q|E*^7R}A1MVBzcJFvQ@u}sHY9}{m)I_I0O1ZO#SP22wXy0C{D5S_KXljltjmsSX
zh7JM45eFpXFer7qt`k>0rRkRUR9&8=+a_v^qF^)(Kz}49`M8~QuueS3a_GQ-?zr@z
zc9zb{;AN@y{O|0k)B5&8r_5HbuT5yza@J2S!9d}gVMHzU?)K#Mol&p4#rKxvnku+S
zoCMDVp`~oW5@w2;kg2oix4}2M%3=xORLOTe5I33|=OW%eaQBZJ1*->1Kbb3kwti(g
zO*#jT@p42)s(8!^Oc1&06A?EQ^3_)q3z@5yvU{=8r|-wRW~R7;y%BzT0JZmXk6W*h
ztb5L??;*^z%p6QIIkMJYjr5FX;|HhFQUG1Yz=*Gq`~q<Ho*(j<o9Q*X=f0+&0CQnm
z!qK_0&&qG*GcA6^xc(sDU1#CHSTiYq`qPpvb3_Us2;T}Fm3$hp=U3?q^hRGC&Vk=i
zlfn1J?)=jT{6iN^In&TH@zG5ZHi^fHtOc-R@{04?Y`LZHF$fDVJsvt|m95-93>bui
z*}#At$n0pzQB2@4A?VK&uuyXIgk3C<F8LJ>$@5cA?I+bjq-P~8)k@q9^s@PFg^mZD
zLbpQBNrczmUMolQ7&fEc)!+NVW^3P_By&<ZG04ex8H9dRi+OdZdzVh<xOiV&r(2bZ
zc&%a9Eud`AA&5b-W3SWD-7An%Lo~W}ERP72ns`-CY?9P>HRq{J?igxH*ne}#rG~tU
zwEW_j?STh6mimVAE2K_Yr6i2?56qdLm`o2cctyV6L5^$;Tv?gh8{4oSv>|+OKmk!B
zKnIXf$x<*5zw~G<jdt)JUNL`XE?o{RAg8pUb@kl;O(Q;>pK8X)BuiJAvj}@-D`S=A
zMrOT&$oUt;)>jPoUY**pTuR5%yjeh#++F9`sMEFWY?s3GO{Do^(U+<r=q~5_s4-JN
zwoZMrM5z|lNJ>}zI)&+#^tMFHJ-_Don=tOfPf{~YX%>^V{faNV(KC4dxQoX%&PzM%
z`YGNOaRliGljGp|BOYneEJlvOoi2TNNd4`{bEZGQbb*9VjPjAZwjy-5AnoHJLg*69
zv<~0wqFzyaetn7T`eh5pGT5f}*2x8J1M5g#s!tycrso&tZgtjNrf#@A(!kMNXV(u1
zh%ue_2Yl|8s4*sIO*_pl+D>}Lap}DQE!p^eRc0={?0+8vKt?OWa~yu4!Hy)YmH5%<
zwn%_CPYXNG)t6oR4xZ7|%@Z3;(AJlH;bOBzAkIJ~d3HOgH)YgLYtF8p3;Jg6%K1C|
zD2RR%{40Juyt=<1D)wtFr-PG6e>lQzr~fm+cpT7J=VNDr7WCBt3B`hLGBfbL@fmAv
zl4bHy6X2LI8v{(>y}PV_qDxA$mo}-Jbt19m2dA;AuvGWRnlTe{WB2y|w9q3W2kzLG
zL@YZooqIQ<&Q@pISI62|+xK#L_Ew6$uKE%Rv+cjF&OPyh4AaG?dz53dJWZ{8V8k4W
z{SHp#-tms?2tRuo2iNaHmTj+C6rqcZROJ+a<U+gg7xV5urz{_E(6}}4HKL{zr1+xm
zNl=-f`(Z7QD*tmPR6-2<4(F2)0Z8qO%oa63-pXyqzw9hJ6HBW66w{4o1IeYdje*ty
zKYih@KKv4Bdn>))lCc~?&Qsc0w7En7J~*~A=7XiBv>Uqnvb6KC>VshdagPT42<L>A
z7ZbBJA?8kV=NF)2ktC$U2zF9jHfOuD3}afZe;39-mb*KYvSgbVLL+@nB?Su68x&_k
zevZj%@ft2q?&~d$yrkHY62~d9b}XrDv{Q<qwKnSvt_Av9d+{B+9$;)DnlIMGL%4l!
zNyqo>g6=689{j_=G4$e28J72Bku&Tb(6nGix@{k<`tTG#)gsp$t&SzptnRGrr3Ps_
z44nP)SFe}dgiF|L){f=-V?|Ef{_4tTP&m<xQD~u;e6QQcggfq|z9V_IE4gl3L0$d9
zoj+YDrJ2{N<8S2495hLOla(=DI)k?}J)?T2UQQ<AL$3GKgx0k+RU(z!xy#Va*y`;P
zIU6~LJ7mLYc29L<U8x&C7a;RUv*b<3MCa#wtggDo5G!?3CfF^nZu2xs2${%#Mp#op
zn-!Ib#S+gGtW;K@kwWR0j|_mR(Ux?%E>mh^mY)7{l0|ZfL{nt$p|k@M4i5S&oDU>N
zxPe4-Ls2dS=6R9FFoUBZ1}2Zl^EY$ow(feg1CjwD&@#1;K~2K8Xh7N8?0PCh<D7vt
zwe7s5EG4@?0GFF5d~Lj&XBv{<2tPWN^WNP!Gs)J;xjQ*NS1Yx2*3b77cZk?8kj$tO
z?_8{|)0AH@8qHAb6q}UaS33PvViTxRLyOdh$i0EW2C~8nu(?6+jD<XUa+V?&?KyaT
z_=+G{DR}_4{)gkHL_*4<o#XHNcb7vPIPit?EsoBOR$`sUAoloYT_vSu=NOyX=jB#!
zYpBi{iXQqaH8Vr#L`3<S%}HUU$NBEq<!s$&lg7N-xMAz`Esh%t3uUy(HlYYxZ7Z3m
zJ}9~c_9hgSTQ1^Tgz%4Zql}pu=M{w;^=OOPgz2QjY#G^t6kk^Ikho;idHWSybI|bu
zJ^DutJkPy!>u6tjmSf=VTs?=L-EJSy*~7%XmHI&qIa}7HS2CkCScaAuHaNEP{wWbL
z;E`Hd;&SuC)o4}d7Tn`q{Cpikr3<sFaO?7^PK+Rdy>6{&Qdmp_bSlJ^ZWX#>P1WTU
zPmkwaLMa9{*A6-m8~W$KD2;$^qa)X9+N(xO)MG~ekl2ZH6=YIS^FsXUjt|%`k2jDs
zHEBe7ynNiB?%?vmS-SN6Ud7hi)8@;Q`i@J8TSyD}U#9jq>5FK6b+K?u?sTt}2xy^A
zxmYY;!fF@Ub}VlWZ{0|DL*oIk6@UvOl=Oo`+@t%Y<pve@<;vtPT`+I+shG~idQ7vb
z89)_q8?O3muzzV~$-4W#@3|(Pmb{Dg`1Q=7YvI{pzJ_xYN(|;f?ibd-=a!z6X4<~G
zQ&r+5_2K%>cjq5g&A1~k!|6aCd0Lb)?I<<jWszDq<UF2rD&~$IEN5O1EXEwoW3|E8
zBM2x1dSY}~;AP$^k{Zx&U9EO$JJDS0F`;l^x$myPAc#nJH7%{j0|SMEVVW&hb4`DQ
z(?qKj6-ygD%U}x)pF}GOSI}a>;gDdWH4-u~a1yX`2Mf6*P^HLKgxAWTbvReEMZk0m
zUG?WIvgWsFIXiG2nCuzR#7Y)R&}5|)$v*tkx>+G<0h$_sGx!RMk<3zlCfCTMW_<4S
zzp813S4ldN51iA@U8BCFF<=FUVv`>3+gB=Z5S&1*ws|)t3YjXC==HLJ4e&Zpl$UOF
zc(t8+pw@%g2pmPOQ6*j~N<vC*kWy;gz<hc40vyACk~x_7@DM)#BX`%Fmk0jj6CdIg
z;Kp&sZb30R{CrJ2=%u;(dds!oA9-J;6fD1l_r!3<{ZT;*qJEeYAv%Ac<yOY=(ZTyQ
zQJ+1_a;{DD5#8Y)(qxJE;Pq3(<{kH6KfHQ((T%+6gs1y2J+=ONsvtaOp-z|JG!m1P
zbv>>-bR5_<ukR4O)FN?MYF<DR>SBII^{N%ul=N0x$$#t79pB@ioPrtvcm`f4c`{93
zJe9u^>^$5t{oD4oO?fbz0MjJn^5u7<qqcc+C<xMbsdh%vO;ImFa$Y$KyD8A>P%^+v
zB};$7QQ)URmQPe|dc5z7{(6DG^na$c^hb6&mvqK{>&jJ>xJD`+F>w>>ik-fQ+aRtq
z$jqQAOUQ3ynUoM?H%U?fmQ~;}!8tDX0n5wr6gmM|sRBgH-L^PGh&eMNOBq0ulOP7<
z7=0+QwfkNt++FV~!j1C<W`!ONRAm<|vT-8I5!UVapyU=07L8oXm2a91gX`_ln_p~Z
zA3W7pd{Aq_U|=eJbX*I>n}o4;H)qPSc<WG^$R^Z4vq$LZ<<DS!Zk-{czr#9u{%PHP
z?)Aj{R6t6%UcG+iN2%xIVp>GG-j(cE2dG7S)0Y@ORuX-F`Lp^+w}g=eQhoiyg`ttZ
zX1>-galW8cdL|#>p3MI^(R?DVgv=nkxsL0_z?%Gd?)xvl?rOF)4^>Z0a4G$D;q@Q2
zEaatHN1F{1to48m_X3Da=kTPGv~bF^aNh{P15FvoS7jxsv9zhl2x&bg>9z*+HY^a(
z1xp<Ug2gr(C)QB_yfA`(p2GU>-H~h7gZj%7W+LyFetnQilblK>>pQ(Tv#EXq!b7e;
zvLJmHa_t(?gt(du#1hCuKot;#71V_no|_?P=^b}<*d;L>dtEMm9Q#{qq2>B>#aQA4
z`GH*#TNYfI`f!k$iIL72DiAaYtkaAGx8>;po0<D8HzcSe-!d5btku)eYskm2-7Z}R
z_Mo&6zT|%NZd4-O;DA#6<WT=8p0A&reXrK>Hc5yVNcGX(sK$F=NVz)*Kokt;6}YEk
zI=fC%$Eb#R7(aZO#=Mh89t99bfg-|SqK$F9$=xz?EL6)wCo)5k_vufrt$MWK+aA=y
z*=?oUep%IG5c@$}&*i>u_>9}K)mOY-&HA#wh}C~d`gGCQMe{=YIP9^XZ``tV+EUB)
za`5}+vJ8+p2JC`+Hto53sX<YOb86{6IIWr+dR~Z=ne=PldZwfaye!yCUa^FF#dCw6
zg~_{YVm(a^V?^UD3v(DW`Dkzn7~(|=dbQ2`x7DH~0I*3(V)cVNhnq-f<~E2(3R)QK
zV)<cNFIsF8DE56vM1pQ#|M_)3Qt8H}cS6_5alVu2Ie!<?;@CR$2Sjp>o;!C=zwr+o
zOJLpq`Vr#78!Vv{oolDKOvvBJxTkA06rzsHCV>w!sk+F1!MrM{U~p&VYF0s?sGY1Z
z0&UR2;kjyVY4N2{a+<yHUV-3PGbAJJgC$(Y0pvf6HTVi9>}ZuMNH1P<e|8bYyu;4E
zt;DD5o@V+}L=g!}8#qB7L{3|->(EmI01u@Di59$LD!k|Pv;u}``d~i-))ew6V5(=B
zN|R?(=trP|HGBVR^SqtXq4b&fy9J{v0(vp?JxXW{E{BV*Huq1B+jbA%cQI=EF}xvt
zb(L&M%z2mCVlMdGa>cf8xt08Th93Rv5ZS^!ntrCeMF>Vqi#Y6cdIEl)$#na_F{56C
zK<0ILeSGtquVT47;?U#X)ZFUL)SB#qQrg}ZK+jpgIWQylQK=3CORORZ&PC^-XCYdo
z3i1j`H=*<hozzcqE*O*#`}h!1g_v-q+y?d~OJ>J3>_?ARR`t||;k@km<35m1(S7Vg
zo9x{!?`lg;Fd!$>xq+4fDF>Mgj=ZB*lkoSvI2QPn#KVw1LFS8JdYe%#(VvW22@4vv
z7$J|(0h$LH^VEB3m8Q}=x(GHu#u4Ap(3p{tqp046<s>*Q;ZP&d0UuxHJiYeh2Q`cG
zougd(4@A>z!gx!=)Uz7NO9NlkX0#btCJ$yy2E`h#^S`-@j~fl5=Uj`Q8>y{|D*1;7
zYWpjN{CVgql;V^Tq@}{8<z~LPOZ^5&ZuPKm%S~+hEQE#Rc!>Yv0J5fOE#XC6*?;gt
z+y{ei>UmmMEW-N^W*w6BmRn^z#GDh`N?;sjTVh?dm?Uis{k&||1On)u^o<>;4;Yl}
ztIm@-Xl5ST6)H=t887m+FQW^ts4dRam^1sdL<~JW0lr9`!F;o?ly5a#2LO2Kilr?6
z0t7kI=_im4b93`&hqj=1<@^y*9f2{9V^vDkke>!2bp%E7cdS_1#fJS-#|BJ}{rX%p
z_>%jAzlVCvevV$9de*MOh(rtg{0A-~`aQv;tVz59R6)3A^K)_3Me;EI07xiK<pDlI
z^%x@o-;<1|ng6iBtBxeMF#vHe*?~D^LV}Q!lhd^=k*1lDE(2iXJw-8#*ROX@u<ZSO
zV(>jyARO6l`yQ;fD!wQoYr^z`of}>do@DmY1S~-yhhS9be*{^(k5ajuH$6@fP%NYd
zJ98+a4LFTibRP^5Aj5;|fYL$0Al#f~uk=w^;5`K}Jdgb6ebXZ+Jm+4BF>d<le?ChG
zF!#(J(_O}IqqrGzwrK$LGV1;d1G@99D_d&4JBNEG?6gJ}m!HKV<x0jhb><*L`e(;H
z_eOdIPv~<EKLJWTJm}E91wJ^75v9&aF!ArxsT$-4vXgg3H_`n~PO6(*Sb(6jv(!~1
zBreYRnN7arX7$Slu=9-V?%qw<Y*$oN{8qmqjgl;oY{pQCTGQKi2|qbr8o_+nuVM^9
z2LOc~*$}I!^EscMpA5WhAMrtL{;Ua4*BYQfp!*ci`-$~&rPJJO+T<>pdc!!TtwqFZ
z@n{^qMRas@M24aZWHb`DZQaB<EEYsK80Z_OisG_ju1^{jZn#ZNgq#c02L2*}z1o_Z
zAgDol0CP_1Ul>vd<kefHFJ4rES+mT;QfCc~ATdICSVl2LH=A%@l2nOc0krJ+z}J?L
zHYeEo>S4$(b1$9q{lQIzeVfD?6VDP7-2c30%187}Tt76+jA)rxvNJgU$0&6nE}7aE
zUix!9zv)7yxs=?8+g%ww9^v3<Tl_y)3yjpu2lGWwT8v%av0QvB8UdUk9KCsqN!I2*
zETAdE4Y|CwRh~cr>GVjl)nd$;*-svBVLC~69fs8~(nz|$$_o(Ewa!7LWwa)&!A<}u
z1NNYyZb$5EWJus+uT;!3F9419!?1ncVU)PBpj4;(#D)L}mX6;rZ@|)4m3-3}eh1K&
z3v*w+J90^_A+FO%=iu(-Gy85stOQ~s+{A+?vZakc>%`5u0vlko1rWyVx-P}#-j$_&
zN`6SNxIIDg-p`s?G=%JBCBBB%mgEvSvUqIphN16g1aZi}6F8B<B-1vo8#+F4J&P4U
zV8#Q;Na_;;Vpu6LWmOG`@B=*pW`Byy(2qn4`{`=mdAv^mXR%ZjD(=49=Q}(fm&=PA
zMP$sElylyd7r&Dah1yl*v%=)&Ul~8X6X%Uc&<@C2s@5`OzTc_(w;gxq-!WCxM32Oo
zC=~0pRY`jJ7ESItll6`|0%4nyKHjf|n{r2Qb5Nsg$WZAoKGpO+@VWOb*cNx{l|-ua
zR2s8)Fy)-G-eCZ3V97-auO}5#&8Vwb-8m0^!9cqMFdW2;!Hg7dQMf&75>$Hwe$ISS
zZmSp}`4;bmeaqP%AB5p4Of`Xi5DY`}f&}BO(D27Hbn|r&<3)ut7^I<|{8|WxE|EX|
znc_r1N*CS+^qexWpfx>f5t8tV%6>gG4n26{HJC}f47+0lMh+$fcwtl2<NpW@42=3e
zWL*h3lxzD}Nofz!A|h+qQpqk`c1l_(lt@HMRAZ|oLWQwqi55#zWQrn7O4%w!k`&6w
zR?#3k|KBsy`A*;Wzpl9)XXzO4yzl+o_iwwatBrhy`=&WMIy&njy)8q`V1(G5{qp0V
zt;^dElG4rrWc%Hn&u~zn1p!jV^ffS}r*7<)5=)VsGfXd99Z6@qb>fzvTDbQ$j85t*
zuwKm5Mk&Xg%M+^mC~K^fU&M}vn(B$f<LIYSL)HkLZdj(}etG{=*rG1CiM*aE&v_ED
zF2TEL=F{;pb^5A#EUtLmKKygV@0tHWk(Sv-M%|fPIkWB9D);H!VXxY!3y4dKpP>2u
zil8(N(C;7#tZ898)6I_$L}}Gnh<s$!NzEAS)(c$|p%-&9gjCw~8GoLk1<Rzc(+9x^
zIJMCefz#%!!e5zVhgi}K{&Y5-->~cAMg#c+q4zc(NvGcsLED!2JttYT5FnFu+Hc>s
zh{lKLXytd0=ofUNl9Jqs>UjCC&0-&493yGK)M(6*fId`f+T7bk8uSf^@c}Lk1pkTA
ziy#t`@lcZ=<3IlOnLvb&W=%hl--_Hqa1HVW&<&87z^`AYV<{@U7Z54nzVBYj3)CR^
zYmuEL;#NYqgjSYs3%-nP00=0J@Wu(7+S$>cKe|bXu7dU%2N9?<j|t|Ig&mI%DPM(D
z3!e`58?BlR`J+H^5Y@{D78u53^&G7-M=%~ac<dz8#f4rTykACVr>&eaFo-Ku`d;_x
zU&7QqaKGvTt!8C|#t-+VvAvi5{eiI-sbHtGbHlzPBE;|f2jhy$Cj#s^Xhde%j~93Y
z%lMp4V7+}z2||Xcp_;E@%qy^K-!HNd^f)apUwg>4@iE}0R8&$Tdl=w3?eYQr>=4qJ
zCt8Cu;WpjgKp%wr@b4chiV<N+eEnFVlPXbuY#y6o*4MI{Sj?5EaMYst&(OTH^-X+$
z3LO#mA`H2CI|}weT&0zml=eD!j)<(jS6X3%-p$3pdGW${3Z@DEw)Pw~Va}?o9@Kse
zIPAFgYw`#14S8D{8hYp|SUH#wrtf_=FfIrWPx-MZ9?_>Mu1(<2L6>eY;h8ftS_?NH
zp<?5VE#jqw=-yMs7b@M8*3FxNj|F@<UPzzI*)5`eV047@jG=O@&hf?3&*a$IOu1Ik
zFN6ngrWpoZocrR~onpS=IG6I%gOZYw4+DIc{mWx)YnG$795J*qy@yJ_DPfRK2g-9+
zinGcxI?^_i{qzsX^@Bw{lk1P(d^uq6o<bMW9cht}aC-%f@;<q(Mydwl0UkQ~hQyo=
z4#QH5I@=+x7L$9-4n70e&JgX62%ABEgLNYIeD=@DCJ*$mNUtNofzU>_6#G6yp&|?L
zuu&TSN+Kj`l~!C^`-RS{lW`Xks62`g-~h)H2#eNl#-=?vi5sB$@YvxajXJ1B3n^^S
zt>s&0%e5Y@H_)i(uhbCEYSsN|14$3Sp#fr{zrSC{jsf_|*_)+AW!LOGbUKm$*Gh^l
zU#KqryH&Hfrt)0(xKgu_Z$!dBF&{%R(CD(oiMnWhT!nm`G2Cx#K4w-im3zSaU9CgG
zv~rz(X8A0(YwAG|FV3EhlL#n*y@Qs1am4X60ZS_BufdGV&8xq?!R6ui7ao<B{&eY9
zNB0%GjYh;1%H<MLtEL(1DEOs__L#d1<%$wThMsH*^wZ}O#|Q4)WL_vIZrG9hg{g_g
zv@LnEPp5iE+pmX)uK)uC?~tKrM0+A{==qOVhIN@|1T8>c8|>S$)Sa1MaIXP%BN#oC
zYjtk5J}o1pp>0+WHDttJ&T<Yu!G$NRC~;w8I|@pP9(Er=^kj8pPfP6P{UZ*Y!SCd+
zWz9{1`C-TxNfvk!w0g`JFNhN{WWyZ_=qA@ksZVX=)&O@cT43?O39AZ0V*22G7M{89
z2h3?W=?w?x`?CQG<m2NT{_(KoD1{On8akI8`q+Sg1dP1kHpL(rzk>$0jwWPvl(TRg
z24?Pcg^ey>VJOF+L}3MBT{$9bTje*Rw5oRrjNB`n<m1?en}-C6CUB?u6nrtv!iW;L
zrhc#36Bzggz3x(@SC(t$?51~whOgEXaacAtDlf$aWG*Z-OAb|>=o{DnJy#>N>1lk`
zrbbo2MyVZ1kaxTrGIQlbB6rsS-GjeBMpbC?pdl7skAI)|>G<&D<oY*#O=x1&65Zu1
zg|~c@eSSLImWYZpW29twL%Un<H%djM7#omfY!Jpk^#WMdD)15P4p`Xr{rfz8IBQDZ
z5cx5s^D8wm$0l1am>*@FJP%ltaF!&MlS7<*@77Wb2L}#CG;-0K^?sj|DNjj#g&wM!
zhOP{tj0GVnb_VaGP1`CAFOk9fw=>r(HR#(7qub%pz0h{WcT=`8<`Qh5u~<W_0=PW7
zBhP{s4iEpjX}Db-6$#M~UEhrlwOo|x)t~4b6m_acxIw?_#>IE>3<F(38;kdj7ycHV
z@bMxa-W0*@Hla@!QDi%7uNz;9JW&heHqw&rh~EO(Abtkc!#ua!HsOQ!9&)41(Fq{j
zrA1UX9rc*X<k8*Nd=?&^dBOTW!kSfjoEqoIhrr@jQI{0_#`FzM#n+0<MTG3Kal3rr
z^&LDIf2#uL<iv$Xo>fZ`M>>3!{-wpVUFzG+o8#e`BT1HNR$QhW#Wy6dcIrUKY8L9k
zb6#s2&N=)d%=Y{%J__8DI2bX_qwoSTtUMLqQtg~;IC~{aa71zmN+t(oMJ1=OCWM@~
zPRFT?F&>^&8^Mf|Y@Apya!~K)wy(uIFC!)a62#)(<E>Zi;~~1h%<|d*mxBf5WBXfb
zKy}D&`XG21{4CtltQ9N(RXRF49u*drH8w88jCAb8iNJ&RIjd=GtBz#}CA_V9Oiclk
z2e&?QP=G#X+Ur|Q!br*jjxgVSJ(XSur=tXKsPpsbaTm2JytQ7dfT@8Z2<HTxgfuBU
zt)4|_MDT8a|3QcDe4dk;ynw7`j&o<@1H`|6#da=N87+Okx6cpwK3_1Ad1vo)&RtFY
zVa*NNeEWOvXqMA|?6-aXFT4`9$<O>;*PAx~-#hZy4b9)nXmyFZ8d@|zF)M2^P96Cv
zt2W{=R#8m0h;*n&!4cAAKS3Lxx|6s{!YW8+iwsn;=yUNE)trgaiN>6We|3<5FY7Fc
zLyrg#Bqmyz=TRxXFc@61v9W8g+SW1sjQ$A+1c5PuGFDaEYir9){w3=dh-Dw+B-yo&
zMq0F|@(W0J!;kN{bLO6(KIi;&Nbn4MV6BFaKXRGA*kZE(kYqTs_Ap}Z!3&-w7Nl5a
z#8Afh3=i?44M*MG{n7SD$NL;RM%GiK!(t-V*V!G|IIlYv+4O(B_UnGZ8}4Xn65N6(
zpx${$(bVDGc*6%sHb<y9?T((jhQ?gF+Kp_5Y2&-4w|&$QUbB$7Q0A77d&Hz#4Vp&I
zZ4^}j(1jbngd)1M{H$?z#EG=Mo7(w>GgtHpIC+0GJ=!T8e$#(wA?qMl4J1|T!`5?7
z>qd6%na+JCM{|dvQDqHR)P2$FpJxAsrwF-Kr7zze^$?d2?l7p1aG7tvt|s$!O#a?=
ztjA<Ii7->m{XUAR9pMmBiWV6bi`X@_mJ8eO!2`vlMoY<RXE!%y+8=T%Naeymf%pe!
zHMjh1@skHbVYnldoDyj-_X(F`qa_w4qfwkAj!uD{im@KC7g#^d9s<n*S_4S*6fa92
zEM`-ZWovm3WDzLNj<K(8<zYx;ej2n;@F>O(_xw9zMvZyCt+j&UPEG)*a}5DJfYJm!
zj@EAPH%el-V8uoqYV=JBl3)T)cO;vR#c$%Mh8qx#Ywnu!dk=U-L@c7((v>A6TkW|>
z%oCohwftAoDi3!r7%w_OAs6^eay|ip6I=vJTw>Okea8is%s-PXDw!?VyT@#6Otx9n
zSt5ms{jRb4oLV*s3q;mGWFD_t^YC-v<%rRZYHwm5(%~#(jD}efx@-(VC_t>U00jsD
z$&Pd1vwGx5g1*0WQGgGfNW@6$6j=#CMg{^~0aZkYguX;>o}_5U86euYV88>9JRCdJ
z_*LsV`jtT3k?^dSA5&aqCQrtXyE_eq;oU`>2!|yw5&1ae5{kxzcR(~s?CWpf8zcoM
zOi<(N!xq11Et9`DjsI?P8}9ou136&IvEznYzvVM)7D{@sg-?VeE-O>Tv?F<LRaAiL
zmVkD}%S1(vU9CNXWcZkfL}t;|y;H{0ymRw>IhDBJ@q%S-&Ut@l+b!Et#8x9kq&RA&
zWSqie(We)u!+tKbjw&i$`$YYV+OvNZP*I!mZj}y*(lhMaUr`x>g56vCxwpM^!;pZC
zUhfx&dHS!k+wUux&rOWKkc5{TzR0dm?k4gL)Lw)<C@#QHq((U;ygK;(oj9{UouwDE
zdge8AGdSXa5t66*{{8i&3xT!V-l8dwTV7bm3GWu&zm<SDM0v1=H9d42ZYY^NKQ)p(
z#XOHOFi2>EAY!OX>Rq5O3xqW4tl_t)1F|IvFt-`pvsi<tT!*$sC}=Y~l_{=-UIz!X
z{6RPg&A%yijp*AU6qj`tqdz03Fg4OaMMdRYn__po&=2ma3O57~lVp-So(TC6;aryq
z+@OaxeC_ZL7Y96mp%M~d=XtA`8Q9gdclpvGH*`=cM5gqg7VKXdTyG(2fHX3iYrpa2
zO+U&zs(d5g_?^Gh<&?I#jaPW1HjTXCOrl1egzfjvUln^?rZ&>{UvPqcZ9KvDSui(0
zhA&NJL8jXrb1K62=AQw(BqqYrRh$zxpa@5DXr4jKU01g~_iA!-8~RWrY#5wpsA?r?
z(Zv?Wqk2G=&WvTLnHpa>**i5+VrX8B-g9XnbmKT0d6M1F%m!+`917<%T*q{NVq>bI
zIqx)IX1W^I(rt_5pJVViYN!)}RlGp<2cJYf6O}PHVhNG6fv~{149+6@3Mtd1&rD2&
z{&bMvUD2f}5GuraXT#Rm9&BTIQUF94eM?t%&tQBqx0V7ltZ1zZzQL!>W#M=~Zu#0)
z=TI=S?x#~it45bzE4G@}r1JRJ(8nc{4arj-Z=SEpKbxHWKSd&8?^hE$+T?xq1d(<C
z4DaV3(#^APocM2mp*MJ=y!VfTA&lQ}sG*Za))g2<4#z>P*w*GFVIat%J79mtIAtsP
zJfPW$K6p$Cx}TH;RPnZlw&jhe@O0b_hQ={bk5EAW^zmey=RY~t-@b9cI-|BLfI{^Z
zk=$y|#)R#$=xnPw?bFojj^oj{)^rHeZCE%V9)9wrU5>tf6{dZ3DwS10o%b&_H}@0D
za|Iqp1`1NAR8&axhq}-38N3r#mW|@&<DEB(y+x+A$jwu{d{f7vtK^+{G@zT@>Hf1P
z0>aA)q4(3^aR#XLUQyjRS6e{O`Ioz$D&1R9gg`REtb%~f?Try;gVks?vX@>9V<`6>
z&;3`E+8*rq?wVhhoOw%6R?eiazlqxYt^s>Bf6F5MJbTNk%9`e0zZ021G*-V#*hM73
z+zj6Fi3+#rwvi~3B=#ZGL%E7=0-`w|bq$Yo%xSek<+<EEp3vuIJ5d2q2>{gN7s~Bo
zgd0C?JF|Rb8STLj`~m2=aEQQni~LFwk_5#LnS{oC0rZv~mvo}fP%@>d=I>w1#6p(>
zx;NNFO><oeyha=DIbT8CS&~)sefX3ICoi@w#pl8KTX?OR7`M(g9g54WluN5B3xGf7
zO3TQ~{`%FXM=-TryLJ(r0H_(rZO$tCJ^kAlS1h>e1`bG;pFpYs)biUb*^RaIxqRz6
zon*VFw_zZMTcBh$xmkZi)#MF4ypqP*lm|Rk{IL2tZt*8>e)CToj|IuhsTBEs^DiKV
z@gm82>rlw{=3fgH(BwL7%)TVpGWFq$`dImimchE3Q<v-xYMOFLNR-@Lls*1tV#0*z
z3FK!*Mm^L^fzXKvx~NC?zPQAib#NE6k)7APc+@&5C~r91$XZ<}8u9BeGu(qa%(seZ
z94N9uixa)&Bb0_cgGar+LkSXtNFm&TOuTH{qzae4oDJP<WK-WnBHCYI|9C;wVF=`Y
z|Hbf!Tj+9eTZxAWaku$SXyDY7@An=w(m4_xYz)2}_zO;0U|+bL!%Z!dfh~Sx^YJ}E
zFz_(JDyn?^0ue6<Q}iM-@Maw>3Gn*rpfg0T&%X4%(G?fDUol5WShT0(G47)0cBMO<
zXDgZ0oArNE=%Z^ItY|8yv#rg|#a6U#^!RJ2pX+_A=s|d*>9*!A<~DrhW?T1=ho31=
zH#So@#n$AjKR0tYhZY{EjS8q9n?C-t0WyKTtIpBT1Y-b6z1;4|P@C1px=Dp#U1Bz4
z*m4RIo^@EG1w~1_hM^2i1oODj=8L0~Twov$Fg$>JWMPXO2wfqh7n%uRQ&7c3wq3pg
zB#UkzHyf#G_}_g|H2vDabG$<6&8|XfHpr||R~>=X8hWvNbS*H+Vk`oVR*J!daKIEw
zhB?w*<$>*fT~ITlakIt(z%X<@=xGaNAh&kM9J-~XMK5Q!y`VUDoYkDa5W)t@Kt<^2
z0ZyUv^Xd`XC#56vL!N&<YZMD@gY2)8l8K?~p*egyY;E=X_)bmT`}>*Sb-UC<iy>xU
zIdn<z6;(dcVO#9=vIPx2_VJ9v5gU~-U0hEjsn~?GwO?MMzivODjuv1XxSH8wU066R
z|8z2Y^Fl*enTA1&+zG?KR?Eo9kd%WAi%0A#jd7S4(R|{%0h1+BUu5J%D+~0T2$q>n
zh*P#6bdWw6qX2*a>A_51+Fz*S{s9aMSBr!s^`X|z$6T69@LgkfoA?#CF#SrJPrmVH
zjFyD`!hiM=Yb<d4x-^-7KR1|YLe%L-@MbZc?-)}^Vk1*cgKv07IUD_9NcyNUDl$K5
z2B(rKeSMahDVn)3YCT2p!Ah1`L0_!3!{HwzowsP!3U4$_m&($39ISVn_Qj?oeh89J
zn4(73i4A1N4yV>AyLh?(COwmuix}O0-$E>Cve#NrnLb(%RB0``R4W=*)nJ`jMyad}
zpCJ=eT2Sd#g!R;TdLMs&foWj!_rsEZ;ACd*0K%G80AIin2lNf-Ckdh=yG+Tjj5`ko
zX{J>9g@<yx@7su*C4#_7@o8`i(m(LcnThERX9m08g4#&m^;m>@)f7ex2&R}`3nARF
z(45B~%1Yl$O+H{5Lyd&57KFm{=jWYwVJeJHOM>6KO34zewC7${QwJ^VJ4Z*Rc~ovm
zOt|zL4Z-m{i`)wVr^2#?`9p$7AQ%Zdh7-n9YLxhv31zFiFk<b|6`FM}Ig)8SVTwK^
z+`eaUBy&Zbj=9Cs>G1Zcsab~YQ@?gCafi}Ku3=w8(v_rfs@C@}*Lk8U$AoZ=%@VH{
zY((Y3mSE{n34a=3ruFMCG7BUBMHI&JW029{FegMk>9~Qx5KtZ$vfdeTIif^;B%+7W
z0Sy5mbRS!j&?&G6WNQkbhaqk33E8UdF0#&rMbtI)l!6o~%ZDC+v>WUJ(1747)!uMl
zP?<Y4E~S)KDF}o?xYhj-!Y`hRyv-k%FeL_<J|OVo!90!0%Fx)4q1yZAxu`6#TDa!H
zu%Ukg+Qh>6(3#)}o#Rd`Q^Rz~5LgEVfDWnp&hq$Gfrd^bE@EqGXH!V<)7tqX*rk+c
z<o}GM2j(AP1Z_$cpE5Q#ojeL)bB#e&c3jf9RYUawQ?s<h@lQd>`V(2}b9<cuWe<Q&
zG~ZC;UElfyzYIUosX4E@ID^4_mg!sJg2u`bY9Bc%zdV}-ULBMXI76)J!h#@MYD}2U
z7M$6*Qsu0)bYeBYyoB<Nr-?4={;7h!8ISs)*Ck9#fCigq0ig=U>T5tGVizKcR`*t8
z(=P4tg&4FykPV4A{Qm5Mat3}G-8kzodrsz1Pm)ay006si4Q4AH%(a@vhPBV277+qw
zKnLU^H;=``LyPpXqkl&172E_${PIHE#oU$$p2Zu00_|t;rn@ql%|eAUCpqsV{@q0K
zKjNq}s9K?RV7uR~!8#}*o~;-B5$w(|nv)Y0gdww+TYExw&jVM%l39b=e5?5?&aCZ&
z8z!I~M*u)*g#{iY<qXE>Rnp^eX&JS>B47IcOwo9TA?m?-*QUr5TwxaBEohl`B4;-!
zUy@Zo?lWvzW{vi5f1FsX8?pR~Xt|jen@xb(gtvjmkT0S=f!J<07xQ)mqQ^;Rst>@4
zBUc>}CEmMIj2XvaJN0QnkQG`+!>#HnD%xlOm`DHk#SZ?7-3Y=*f|t&e;$;{~-i2`D
z<E`-CQfAK<Ja!Gg0mfLnoLjX~jp`pSpEmVAb;Se+WW%%JU+0TNoQC5*{7WJ+UaV-i
zmh)w7Bekp2>$dy!{%ksZU;!4A_rCne_b@&A*K)opmVa36zx~7X*gjhFY20iC3fdMk
zZMS7@OvZlJZ6*zgORkYaYbN4==ydX!x}6?dq3$<+VWFTpZa8o=gN4xMl*-d09{3b6
zll%zH3O4%Fr%$cuxd|46BNsZwf>~d^yC+ta`xM!rfg}4Xh8D1i7GQrJj42=^n%71a
zt>#L-5H$*KR=v`t$m)ksokA~w!bp-AC2jZaZN-DiVitj1?v_G{pwHAFErMKb8o{lY
zd#l!1Hf}IUsz%RH=$*1~i`Y<=h>~S4;%Si`khL#&28$E5nFVQ^s{h9OOwzjOz?CAd
z_Dl1y<}0^%kjKH-=$_<_7jKG-BMASnV6Aw`PR_}X+;x$*<`>jd0pJA`YVDh}nQbhi
z^gx=Hc-`ZJ(d`TKn%C>HQ0~Gdh$%#mZ((143^>j0+WCX2AIwZm_cfgm?uzl}q|C|h
zg}KPN=fq=@!;R~M3_O?@U}31I^^P8|9J5Dbt12*KIH?=UUkJ_345fys6QYc`Jj351
zv#Wq(Fc%&H4fm9ZoLrHVnTNHv(4wIqYW@D*_-^e{jC>eCk&{x>FVV_Bt+v5ZiWgV_
zfB7^v(0G`WP!NthtUh1@EA;0BOF2ln>H9nrIR(zq^*}NbUO{n)O5~Ezmfgfiro>{N
zn413nB@_5m2-^?clDNFZxdi?;@fi^&stU1ni(l?p;bIvXX`3g=+T}YVSGVGR#haPN
z(1Ch(3e!2_mPoh2sJF3G^nMCo=<~BiWL4B6t>zi>a^ENd%*s&;-4f}sa5CWSGqJ_%
zI@}Ziur@!vbp}>obsu8xU@2iSj4a``xD%BfH5iMQA%X<;4|yke5*Q93$m&-mWACxV
zGA?{@9E{0Gzj&4f7A(7tgYQPEkSYzBl40t<S?~M>*Use3C_o>)2lPd)il$n%LL++i
zWXUfcNoIePo(EbCY74Pca-#Llx4de&^A>f7a!hq<{P)Lrk`Gj#7~iE)PNhY>`ZeY9
zB);)q-!Vgn!};j;^7`#4YohG~3op%YdM!G<F)A{-sAb~&&Zg5_l(lV1-n+4W{Jgzo
z3XBXtKyY-9WlyNDXDxdWD}DI@vyM_(hp7rp<RrvRctK)tzp?E3^O<<#h)Dwe6unue
zBCldq4{$WFK~VGRV(+#HJ9Kw3`W<3A%?n}X$oYXmqZsb9Mw%21VMw`I8@Wr@{5y~u
zU|Y<Y50DLLcIUxNq7`1un1SZJ;V=^jX6-vBGb5qbZEU^*nZCQz$k(s(m>pA*jQ50D
zhvl+kNtUrN$xt+wj7XG)=tM9b<EyAdHumz^1<kH5tZBT|B_DJkb5ELh_Ah~ze8oc#
z|20z-MqFS1$s)=m>DPOMDmw)?lsCU$UZcyhj|mZ1@dvj9XWbb;?NeUNR#Ij~A^`aZ
z(SfP*)rFY$1bT}x7q(yEGaE~R8Ti;OrjLM#Z=vy)E^WqD%dCgd2z~jBiKF)}x@hqc
z#~*GqqGbdx%koVk`vH%gsQBXP<BI7OY!hIY9yVct5TI8k0h&92a0P#CeKYBQA}U_~
z6w+5v3LSC=+hhDzST7gS(0@2f#Grwx(}Nrf<&7xKnP>*$0wLiOVc!vvP<ad^D;mV!
za~R0c01YZd*6?fWB2UX){7!-Up#UUlkvXMk+lef@EXUfT%tcbX#CG@kzj^J7Q}dU9
z+@+>#Id#(CQC!kd`VSrPzu?v(rzUm3h24aa2-jH`C%RI1n_ak(@@nQ2I=fLEKNN$c
zO?Dd*#ILYHJ!bKDgf!ZjdDJ@A=w9U2Ma;l8xuvam0-<7a?!nTvc^qi#3#Wuup!I@E
z4|Edr<-rxeQ=kp>w8nVe;JMJ+!rdM<8kI>R1mF{BOoU5cHNMlz9x5WtKcgkp0)h@C
z0^8rRx|_E15Xlac77+@sd{-WHp@ZRvz7f+bgdpf4fybpr-&X_9Oji@iEbsiHTl<a2
zpiyp~==#VX6((b!P6d|5L~8tK@0%$8d(29F*D6;2=o*vmoT;jixa5CJSr0XG?FtHE
zw4jf|$$U^ZG)3y_LgpP6@SWR$By*A}ksw1jC0T&mHQZ#N*}(tj{m|x%rC-D!XWGQ!
zRKvVaNOs5o2%YRzM4(H5^;?BV23UT{;^`H#US}4vA|S0nIm@;BarC^XUZsz^*@a5R
z4zcJ^5A?f?)tZHiY0g)Cc4pSEflv>`86JOdG*|(#h`sd1z;PKzzeJ{jD3rOxDG+59
zGae9hltE@mdszcxZ1~BbJFkV-p6FbgRpslB$#v|I(WKIkFZ&hLJl6TWK2E2XC{D(3
zwRHSV5aZrDvZh&V8|8om1M2%FC}SVE!2|s2X3kWM1#aOF^*HW}NN+i)U;QHWGHaIK
zv&t!jdgvr7Jd%qOlaqm~VS2|a<AJ%>2~DcLI5V;y&mBuk0)II1KQ~_uPsVr!SqY*(
zv9715fbjacR?hOecXzf?LJH?lrwcMyt`b!)Kp@E~T7;l}Z8B3Qfp#At2r3jfMuI6}
zl7mxe5&L%|A8}N(Y8rM@aE(D?1cQn*H#&YmV%mJn^=<3z?P835_DJ!rsnB<GGyJf)
z@$%_`Vx&Nr9S|uF*SmKx?`&?=AH!E+ME`Q1oD``R4ed8RC{1ra$IMh(cJ12uQ>{XV
z_R!EK2e^hBS16UdSv<b<5sqZi+PAI^5stp|asvrB#ALC&^PDZ*l;oaPNcaOgkZ5w|
zi4KJ`QiBoG5ZN=kk=MX~4gz%~La(VZt&dz!1%*lw8Hw0CfOGH!0Rf4wu)?mY>;^y+
zHM{7~F;w9p!`hELR$c)Dd`Sq-I6LHj=>PLv;Xg5tLcc<C5RW{pr%$h@Df~S5m2u2m
zg2y5OFxNixaIhGh&;-JPeN?zr9wU3b;_>E>s=^X*0!@0?ZJ>nRpD`8xd!cne`Bu}(
znRnv;4Ki3G%dh!kSw?SiY)YZuPV4VXdUT%`<5<;ygy)6R8XwagvUM7D&7<a?_c!c}
znC24^{g0fs0Yk;cd!n2{crq}!5rzg==UoHPAZRd%T<F8IcLKDv!XsT@-$KSgjllej
zTn8U@$K{fZ@wnyD-CQ*Vpg{tSv3Of-FKdIQdj<bX@tXZ8s*`a<<^(>6oAV^qORspf
z@FfRJ45sKPpvm{$vf>g4d{qdd*xR6X)TvTv(KZS$3e1pai&&H?Gp764d9s15ce;}i
zGS84i5c!3v9mG=1?iPB(xZVWUaI$u+*0WQee;uDuPJh~`R`hp0T7A3pK~&){vnC12
z>}0w4)2}C4nFAy-^M|0RQe1d!--JlXi*44U%|gPy_+YW@9W`{&Qz1dkaK3?f3v>8c
zM3biJ-0B?zZ3QPE0bww>^*t<^Fmc7zL-lrEbaBRhdOmi=>4dn0CFC;4r%EWAG`e1%
z=Fec=Rm9E$ncbk5^Kt%JxCYR{(4?Y;%_?N{ogWq(gwTxduszw(J|CsL{1|pX8Y;T6
ztN|8loX>LrR!4Q;tg{o$FKY)XzT$<|2xabU#yhZO#r*+>YphLvCikxU_UJ?o6pd0w
zpBut|4151zywM$gy|%iteCFYUe}g=FJI@9F_8b#wAN(dh<)5zdW<f$yqA%Kj>_rKj
z_=`4@f+n*rzFR8fR<{1z&G46SSL~qDJk%>LcELmwuxP<4)^ON>X+K)wqjN6VQ27u4
zM}}L7bD%$H%A0mH-;?1ZB6LhuU~R+$rW_bf8$1KRZ9S)+c?&Ss)A!mh4J<evGZZT$
zc~W^LvVFzakIN;i&a(Vr*n-2%pC%)A++UaLnOxMVHU93<<Kcy1xBR^Thwn*+N_O-v
z9r?t<X7GR)$Y3^Q?uG1Dm5rP}Y)O`Oz0?nnF#cLE-dPc{yD+;kiR#qWQtdiE)Hx|P
z6H1-m_ow!ryxvz-`1eq=I;+$q>(@<<hpQIucJpgB>vp`93<!20{x^-}Ciz`b9vy**
z=k|v$VV@?ij~SW0-h62>BW>$8^#v*0HI<u1)huJ?)eAmbFq>JLXmh)M-*m79Wk<?V
zd%Ni5kCw^#Q@`D&(rP17Gtd#{k1JEa8?pi?u789`9WPXHEJ?_$Sj1RMpWI#h87ZMG
z`n;sYOvD5J9-VDv0Y?}v7&j1V#-rPvSwXF}UCmvkR4Qx$z~AW`Y$3!x2`wX2F-Vr6
zA`D=gUq`1=!=9L#xa}wc*pG*flj%f<24qbuQpaa2ZAnp+98#ol2JY`u9wsc{nGNkF
z9KAzofyuUdt4c&q8~xFt5#X@*8V81EafRa(j|7q~T2GeNA;941cXx(*<;BGD?uBq!
z8&`Ojv2C^;JOvc;Lfi0Lt{Vj8GFrQ0Rtd9Hs!+gLTkp5`?oMc6@UAiXlL|$sBAj+a
zgbE4?@qnB{{A{bUOVJzx7i4B)05OL3lh{d?2I5rU8Lu$y?2QIQ0&xI>=b0<1a9aR7
zMd4w_BKP@Doe>qIIDioZ@2O$3G;qzt)r>4N0uh?@Fd0Ciqk&j}3DL}J8H#hCe7)Gf
z()^()WltVdftY6i3PJ15X@WN9GMCfbO7=&nkl35q*-=}feWI;5ckOt###BvH(Y0SP
z?0Jq)|7)5lQ^_c`Eg0lDV^a63PV7EMJKeGOq{id=-^pjE#jOlI+DW(9(pQiT8uvL8
zcA*^e8AP-5v<G*Xdkq8#u=X<x*{>MzD1eNM4dp3#fc}S^0L7+_4e7mOXn_E({}vaj
zabPkfIGO<o0#yRg0!*)b9dD@aSM{JFW&2&|3oy7GRSf8$pjb0oruir|AQO(S5VsAt
zChHmeWBuYEq`W-dZ~!fYQ#|$;n(&j8mKDcR?~F2uKbz+&2n2ElQ`(E6faouj{gC&M
z<PzARgcjgpRpeYBD(d8J6;aE_w7+9!#ZMovOc7$56WMpy1+M)+0cMo8tM%8e`})%P
zSDc=xDQoj;TqU9LdE4*h)`wVLxu5?7$vjyaS0v$n^q_Z2KZqO+t1T8BKOEj~=pnrb
zw%davME7gp1t_Fs;&%cyN&~$niwS}EXCvEA2_6ama6UxaMYV2%Xmsg3rgR;-xzV4>
zhgMKzgr6ZQ5Z4cSPaJ;s6%dl7m>X8v@1gr{1c7Al9ub|I4JdDau2oL9*m)UABjqEk
z8FLXkJ`9L*X;OY-oM(EIHvvYLiAB4)h>-u&(5-<q?p}YX?aIVW=Iu1`Q7rP5r;_DI
zV;7YluY<W;H<bixq)ki|vbED~F&F>qzh=Bp$tu;mL5moDl_t{8gG%Jrs7{lWSg!`*
zr-PTg(3n^?-;1$Eo~guvKJTp>9^2%@JhGUZBt!L@pjioPdqQAY?3d%UbheIp`KK2z
z4-+J0B6;r6`$2b%&w5RC(AQZMY0RSVoNX0gZi&J^m|r$GZyDP>1%b#6zJn``czQ3u
zgbqRqGO{!$hVPa<I~IZka)3FCAFnrC9oFW#-r)Myt&&v*h#-$erV5PU{_+}OI_22#
znz;j^wP?8U2qH@ZLUskeW_*r?vml+NMut8YE}vyv(Q-I;EF$WXOIW=i?|(xqyU(>4
zR99+E3%eOJ_3{0O6wf1*T|Wz@)J8m&x!?Cp%&yAb^)&r=iF)C+!e@f{6WT-DkB_Mf
z1kkA$ZC6g+pK5=jTQMKjxB2lhymLjRt7}bbmy!HKHkR92;l{$<mNDnCJqm&uyhF%9
zz`JUuiYKEj$3FO-nL@Mu%V^i-ho4WAV-Nug%#r5qcoI;ln7{xu7(p_^O4CL58~@A@
zj|}!oHkh83(rg|e{um2r;LfV&C3#qJ$|nt4=0!#cV$T|Lhc$Y3EF&R?%*t=*CpMUf
zep)o$r+4gKiwYiGrUmdxo-0nodNabn@UMqX-f(M-0(beW<nr{<_MV?_h6UDE#(Wpk
z*|ECBKR3#aHey<HFoVH<xKA&HHf?ElNmOW|kOAd1Q*RI%xll%IE*AB^L`7=mvRM@t
z3Os{K5sB90C>XG;<u7F3JVFWK9xF$R;lRPL0K5@Y8AR*BccIR;$%c__cZ_E%7Ss}i
z4gu{36vRL>*YOI4+_+TpW#8JyC6t;AJPUpJMBpQ3VpC3E)e~;WH^!i-VzSkZ=*~vd
z4{wY$uTTpF><}tz=Ho)+2jDAGwZB*(<2(Qy37m{lyeMz1Ho&oq=QtjJ`js3hBZDe=
z1&Ou3of8^`4;wE}osu>nX_B0~*ZwUV-8X;o$<`_F<^AQ_7C~1cM&r-Y=?(jnef`Ev
zO<%JsMMMrgcGS9j-o;i%qdtonzqVHfp;ccMO;h?wD=5zXq*Ru9=rfiYe2i$X2d(r6
zm%~j0%7MU)sp3Y;SFwydE(qa9Obk}S(gU-Bt9r6v!72KMH@XgNOxuT|&<hkDYzgGr
zt8~T8?>SfvbeI+~7Ds%|mBt(efIkBLLm(w{5fRu{)MI=PdfJx*+sHh0_Quw4rcc;4
z<#FW<$Dtcy2tXnuXR`iSi?D5MwJ6R2yO-fVGHm<*vBRpn&5DXVJ?-#8CMAu8O>(~9
zDHU9*QT(=Fe$zq6NgVV?-t2hE*pGo#LnTK%O}r#!k@qQyx=7?)`~`+MI6-1vEdbmS
zU_D3iKwRzjaF&cpS^iAB;sTt>7zRM4_OZB_Rumd0;E%2!%10u0lv;2oVwYx#8RVBh
zCk!gVszCn`e4{X3XiSHF1Yj&m1155O#qu_UrrdIgU%@Lb!roG|pT#RcWT9j%7F|ub
zII+P1v;uYtbWrer$5;#Ar@t`ZsPj?|af8L@Wh|G3>SFRE0S*t_i=HX0#cV%xX5d$6
z=AG#+;tt&gRf&DMYVdH`h)CRzNd^l3*U-WFzV+lqy%vd<M`vIzAG|4;m3*o%i~qP6
z)O`@Iy2}d%i8sZ*dh!61-Cji8lNE+ihZ<{VOpwwm6fj$O_yaNFVp$=s328piM1wBC
zH8D4+!oY_6ol=q4cu;&ylS^N>o4YcMDo<I<TE7K|5L`1sL~s$B^1!BI(uGV_)PNR&
zhgB@hi0aOxi$nZ&cA!&LLIn8Tt(YlLJ+PnSf>A{cZi(Bj2}ye&?v`3$t%~<!pIuQk
zomRExOxXvFbx4z^r(~}j*6Z#0X;yVMvGB$9@%TFp#wYguV{n&nyT@^Q->oUN(VV1|
zM87~0$<vNkeQNLb#Da7BogDGHk|nF5iqd?hO<0r=+vhnENsV0kI<mV0*WkjcA5WRe
z58#MQ_c^(jF~wu-s;`eW+(ty^X#H`g^M>Lujd55xY9^$2;nRKn=@}QQsJHk7g<`7V
zZZNpa&x(j-qC`<C=qGWmFzKqo5-~Guw6V}I@iNbs5w9Mis|4<sFvtglVxA#k(8vt)
zMpTjh^@F2=lsSvD5iarI<Y4gh%=+#(&q<h!4NL?3EO*+~XQI1A|A%=vxdw0}+HdjG
zKc1ZaDJ;1nSeCx{&)EKl2F9i73rM^(rCcPMbnSiG9u2?R*Y^Kk5mO;BY(XLf+@HDU
z8_)Bp+!rg=kdm&F)6d)r9{{X{)ip2?Dzc{(G%yQ778Ors=56pGF37|WM7-gm8~i#p
zbMwO@O1B%B!9EX70I<XhH!ciqu+P5mj&McZ*^uQ!ir<ee_+W_qh@`MYO-H{@@}*TB
zDs!sHqw>#jvwaVyt)2qIl>I9L#tHHT)Ko{BIXQr&i{zlMuR1Xk56gkoSFd6F=)2rw
z3JJ*6n-ME_dVJUExduvM?u>-y%k<j^7jTexT4uHu<!tWIjJCI%npDS+bzI8#D-6Ag
zjqxYuANVgogC2j!)>wm?(o21DQ?_9PF)d$I-xEJ$AM?5CO#lhY6~#jE0OCQqhi4}`
z$_E#Q#e#ShEx<O8uQwm8lbjXNpjBeF_?W&<J5UPq=r+Y~YV#Y{0uhE1{PF=-R00!F
zR70<fnj5p_n^L$7t8szqwO-?9y5x+O=Mi=}6==0$9EjDF2~kQ|<^!iZ#y=R><`_0U
zZ)9Fo8Jp$IGLjWXxMR6_$$KL%<2kQfFR!U*yMKpmy|({lHaP>zth8#|2P-aq4}0wP
z)%B+iWB2-?vm2w`Xq0i9<!#l{<>Rl{)9u@2l_)n)x*Qq49x*z)ofb|^J-X*VJVRO)
zw^ou~)9+uTuUqF{>Qw}+Mvq5#%`|(oS+9o&_VN80u9^i4-fahsHKq=IBdIPdT_bKB
zz~=y3@nVXBO7mRV^rdo4H7^t@Fq%PWVV=4Io9+#L&MHVGwH8yj2TI4#lsV^)O1w(w
z*v#OmHAABI4GFAYaCUMnN<LU*v%YjAP<<%ZG#B04As_+({iT7oj+p6UD*h#5C-Yo^
z8`bnLVn@1lIeJDHg*NXX3<ht7aY753zL+bz!?HE^sCN?oo>v)(;|c%@{~|MRRnA_Y
zd^ot?Ly018FcND}^(Q)G{eLAq2YioXRa>D-RrwWG%1xM^A<ON)5MCz0Jf7XhEB0-N
ztHmiEtNTPNL8%fB5ND8}hF=**bch59nTA$h<T51D1kH;LZ_v_FTm0I-Z^PxbGKGX!
zr_8i}hs>kT#envoO)2Pa8l$++RBi`#UoX2lA@P|<PE4k?s3gxE6fW3ehI~b1yoi*2
zp|vV?HCjWIYGw@07PNLSW#B6i!li+*HJsP_@<x{|YqnWgw_!&Jt(f6#WpS|O_1tw_
zwiT=Yo3B9QvafuAZc^hFlTsKF>zVBEpL}O~tbh*8zL(s+3;y7&aLsoEY_aMeYD9II
zrMH>S%Rh0phVWtV$q?`n@irt*fChnN!g^&T@Gj4F!QO~Xx?;Nw{b`ds=v%d#=R}z3
zMup)>Mm~ya&PBlQ#(8hV!+N?ZyDs^ZZ!oF1dD{db>QPB|h7^L;p;oSU2HSz6j)Y>A
zBD|bwY$*Kh#(%gh5u+y&#iPhA(t_3MHjm1+?yd^D{fz7%U-*q)edi;AQ{UFTWc8^k
zvQd$Pg5wtEre-S?CPy1@&Jz6Vq@}E`=4)h1L4`KyV`q+T=E<y{bEyn(R#)bS69khg
z>@#SVak63_gpUT@D$wn6@POoV$K&CgJEv|#y@;aN%ZP_M<}Q4p=c$qDb&`N%ZeXag
zuT#6P@4`~*1gF!M1ztC|QF>y=c55gwX;Jo2L!U2g8IyPc7Yh0X$2x$AHpLY^7YFtI
zwm=X9pQh?U9O_-4J*_vE227;jL%uv84U@*@*;Ux+x0bsyWX9-$%9@d6r-ehm*yU(y
zd49X`iT`CY+c)2@BbshM!Dr@Gj(Jv4Eu4AUo?Q^OGW6bOQXZ&!G{5T;76+(|yoz~;
ztarwUhNqr8%iNy|fOYON7Y1|OXZeT-P6T~!WhEynY5c1+YV@AFG0qx>cGrug)P&s5
zi-cqK3CYbd7o*>O%YAf^ryYAPiToC`pMCAxX?+vE!txiI^;Y}JkJ`J7NltAR+YxOj
z3Z(+n2zcR}4|%?wUn3HQlCeP)icG0$^NzzT324>)@p%Z(RuUUHk*1zw&J_i%6=jMx
zW2%14tI(ejdNV4Fw)6M&f8l|18;56TPq?K!j(1PfFflis+IsQZO_KHgzpgHKx=SIi
zy?@@1wxMf&op$u;%f`93QIMTkdhbEhOVg9&wg_+2HY<WKX})}18Ye!u%z9!KA?QGI
zaweS)ITPklmgK@Bv;GpNGPXAFxv}t8Rneu2DtH0Ww?7JiDaULwdIOEA*+=t`68<aE
zfbG`gGV3{@@Q}~bvYoUXsDBVi!7^$qU9If#OaFCvz!)%GEaAH*l`-V59viJq-q8U~
zTgx3UINQC%x#>*Lk#Mi|iB`dh>}4yromR4&tJ?dTzWZ>y=YAv||J7+!;Srp*<Uuy)
z%lR5ocWgae{ssdm;*Ue*7hN&dZ#)k=4KTftLgl_IA`;avrpd7`1F6p&Orm)^6c?~+
z(#^L|SSB98y0B90%q07iUrN<-2nA_z=+A$4>Veh8+0{|O6++=-PS)`Vb-;mvdmAC=
z{_@-OxqJ{Y(x&)U?`HC|l#)PF8z9*GC=V}<r?ukcXEnTyoLyrNTvr6{#}TV|c?Rm7
zzkJgY@;EH9Hjgp+d3R>qlV=L=e~<5)>VeL1-Ei~YvH(o2D&A3%bqkwi3@8&WL6`oC
z7GBKAl4G++#M}0XZp`vagawVn3sVL#qUmL*A&N(QV8e6Y?_-WDbH<k6F)k4k4?~r$
zVwNA`AEgWc-ia~LP*U2)Sq0VSqeqXP;K=}j2vdlc`mA*Qoya+~W^gNE!6iRtwv{AP
zg{eR!c4NHw7eFCvQI(nW!Jv6l!J8~2A+#yKHWtx;d+Ukq$<NB9+egH3HqP{QSUjQT
zK%NDEI4^35(a>G0bK>)8mCk5YChc~4TJ(d8x{G}G5y!FmUhh$Y<%=bh%fkQ5P3E%h
z(<*9hO0qZIIHN!ejR#0yPfR2@{yCDp9)95Ia8Jpdf7B3hoJ9s}-foYITYzR{g;)CR
zl)7nQ;s;Yoj)Z=PmFqTrwy&W3g6Q(D>=E+;(|wKUg_DPi7YNaPw=y*_nAn)5K@434
zUikj?>(|5u0|bOf8K9@Ly6c`np(@b}WY<22;5FK6Ey`t(9p--dIJjOg_rpPiCHLm;
zv{}=rLyOrg+z)$suJ!HuF+}nhq>O?q5nXmRP?fK1dpJ)!*y2%M`x^pHrR0V{qS0Nk
zXO(KX#kIFd$gYFRi_m2P2kW%?+Me_k(J+sVN3v8&^WV2FY}Pk@pFI8`rLehhNZMu{
zD>5?XkJ5p1(q2t>85>LPP_SA2YVWAny}GO9zUbctOv>m@WiNoH<=<!d8Gm@Avj1#y
z@)%v!Qi6wTMl@qRWo1`cePj&&1Fs_6BGcuRtir4)QR)3|qke)FEpagqIE5O8r8_&%
z88XFgXs*5j6d?>M#O#op;7&rTiri#s3{qt5PV+|zTW+>h(-gA|NyH@cR4ygOc!8Ex
z5->E-{+N34mFYMhcaL7cP!G>5Za&fA5bm4B6X|HLI~yspkOA#&a8OpF@FtAv)Fm6?
zX{IlY@OEZf(Qe=680+W6IQ(}Yar@>z^^^ZQ;fEDBc3H~$b|qu!`l&uMSTFxq>tG#;
zqW^CpwN9$j&G&eEXmlMOs2#O-K_bqUF{$QZDcU>yY$SNd$}#yWDwk08hu{a845Czp
zTblCjnnyWIQSaQDTPEDsUA?~gaFy7N4fmSPq&vkM?v5$3jonXoGYqYLT>QF5-|PtG
zi1J1u`_t&9Jde<%jHGw{P=x$<GY~MMl&mD45-=*zJ^|jx69fR;H+jZ3k^Tx#Yz0Nm
zgy`CugL`eF3)O13B-C_p8dxesbmZI1FQyCVeEj7lc&RY?%8f**T2!e-n66rL;PLsw
z+4*6y?LXn(v2wjU^qAaui=Dg?nqIH5V7zl`qsjKcjL07&VISL7qN48iaM{oQ&vpa?
zqNxBs8NBNH=}Veqmj(s#6H$>TR})3oW5=Q4$%o={kTVGFx^O40`tWFt{TWE`_zP|Y
zETV1nqMhq02-9(>ra?fqx4}dznW*nJ&R(XwO>4oM^A3m`q`uNI&vRYz-0oK;9l0~t
zsr~faf$W5s<L{>w)Sg?~theT+B-tFXz8C|C^;DwfPRLi3)7vwXvhKv5#U`JMl=+J<
zAq5E!@*f+7?#TE!KZ%t9ZE(L4I#rk$kSUR3M)e$cC+_RNC|eVUz4$SMCNpXgvBCwP
z;Gwj1MlP$H*x=<I?{80KhVg)T19mG}<zAv!VFZQ;a|8Bwxb}}ECTrqW)H~nob{7Vy
z-qe;x+8ymB>mQC6`JJB{9GQYA%#D3){@-dPH29@Ik9@FkuF?11;gWxZx1Ii<D31Vs
zjYiXL9DmY@c2(?)eHp@QX<A&Bc5=%iE&Z6cqBeCKpw`1+no*7{WKA={2k%o`2EDE>
z3Y@95D@0^f_RZwgb+z}#ji(+aZMl9L>yIpyBU6;BMHc&N1<wmiJMUm^^}cS{<Co~w
z=0v|&13J;J<`KFU@59c{Zt^NP?o~3{qIK+I_=H`&7#upl_nFaNFs2Y-7UBqi6u2|7
zCJPOvlG52WMK)>&n_-nmb;ZVS8TE6pYCXkTGExm1b*4I95)iNRjBP0*n;CxE#SLQ)
z1DE{i3+c_;I|<A|f;=Ab(}d}q1NBjm$rZTqZllz-P0L!NgTU|P&WQ(I_c)76VuxfN
z_YZvp{L#1NUd_?}d9GDuAGv35e6-G_@l&7nkF(j{+f6l@1EjX31UEzAl)RCKw$JU7
z*cC;wH>>vx=XV*X$uCM<n8>B2zw_;RYf!I=QU=#C%x$;}<_i!LHoqhA*2T3c(6tFq
zl*W{SKlQ@1TA|Fj<<!<Y)ini*^HvE>Rpsy2=uOjQSCZ|q`z4mYi|_6x1H{>W*=AQ<
zcm!0XTU{*ev&czTYO2FBhoWazQwgFuR^cT|SC3YfIcOyXsSB-==20Qp8iYt<d8Z-i
z5?!0jvaSGYhIsq!p<C?j1CyMU%8X@iveVtA+j7qrZX8XygZ>5@?Mi!wzIluSKcz)~
zO!2eW`?{#X%1f&06Pwu`d?TY>XHz&6etyo^nnC(4@s(B)Hu+`?(^fSxgKsmxgxx#c
zdE+~0u^>yRLYa5p@^;~qk7H+lbtYU*8vHhHZ>3xpiO9w+@Z&6^K4l9J@_uc$QB2Y;
z)FUf&Yudt#-PKB6^8AQ!ShCn|ke!GjA&$psi4-w>k6#V@bg{2w<>RcmXovyrF6Hp6
zpm@b&-LvudUM_|X*Zbu3(CUO*ea~%{wF$baZ?^vwU1#Dq`L6YrOnzsVkMNPo8oKA;
z28W435h26a$51o9h?*daWMXyAO>d^}=($0Q(zE^Qwf%N~ah9HOp)dtz2d0yu-ti7o
zD$Hy~2~($qj1#2#=NWFxBEmF2oL%M_N4rd^dj?XT4U`F1>Co>Q$i@`{@Mezu{N=Z=
zKAYwztiQ+Nr)!GZ-Ri1Z8^_0kD(+L_+@=oqr9v4RlqCa|x&0&KZ*Oa(j#jMed*^Aw
z?qF^4ZKAq=*n5I8Thw93Vf(W4Rf^;9=>Pbn?{Snw9#d+3lgGY71oboS(W>6v0Y3zY
za$`vE_wlHv|LiGB>Nbu&>#ep-hYP$gf}WSIJ@<~8nsbdQNu38(W$Fd&h|V3)_Q>k4
zo({)|=(L`d*`u^y!W<R(<~rQIUxQ>MLmy<G{MzkQE-a7|o7V9BkmPo&t)~nN4aL9p
zWUmg}G*7#7Z<lBaZr@R^hr$Q3Acg&2qf6eeQDOgF2TM++J?st{vJ-A5>CaQ!+Zzn`
znhJ8SYFt-S>PVeM9K;vl$!}9c)D){10NjXnV0?V9sAOo$r$ZZBDK>T)9)@Zn*`1j}
zDY^~|Qm7Ag_@e5#n{?dkc70u?lblGhG!e@j6h1aP;&Y?=#Mw{P=pTP}q)3<5@KPje
zvk0MHv`!<4oK-716>7)w9Cm2=o>!x-gvIRZ>C<DqXpy2@kG1?IedjpbH!E<o#k59I
zk%No+U&bD+K_7c8Z+;*kT)d+A#><2+)^B3x*O%~B8mGvCd=6I)H<eW|;f9=?MfgC_
zNXi|vx6#S4@QRvTHREHRYo^6M1F>iebL&=<^hMQYQj^T3+TW@iceti1Ez|S#Zc<F5
zsm1%cK;5T?vO|fcD|yd)Zx5j%Ds5c$#zj?AcG;1HLH5lPz0rryPoDBM*sJ!#B=$^4
zKjp^dGlEX_1^Ut&38v}Yz3=OibO+vMAAWpeu1*4%X07~&St*r!R&{WUjPeLe%+frA
zpf~jLd3fGXz5wr2qcGQggvA#lxl{ZKafL<HP4tg<vK9L7mp!xfEuKx0@oJhryiTbn
zV`#<L`U}m?@0xCU?Xmp$(>QSI?Cj6N;@F1P;8`DFp+#6vtWriI@|u~pC9^)Wvj&%$
zzVeSj|0a0Ut?{Be5eIdra3NQQ_bvrrg{Laxe>bV}NlAq}I47x5_y6K-d!s))zTpqd
zTgi_|7%g_yh_^c9)CEI|m!lsTBdt<X*$jyYc8K(SN^mNs7wwDU>zf<)V+ChCZQj9z
zv^f3A`c&Iiy^nE8@2>F14K}=qjCPQ-=BUxV5|hnPE?`fLiEU4m&dy9%PHjK`pmWuj
zS^Dyx8xLgIt6N~hz^;4Y`rXA<*Y3_%6b=bCmO4EK8<%@-Qh`v$>#p`p-KrSh>J|+o
zv=ZH*v3?W7bNfd|+FARFM<M!GP0QXpxLeb|$(Pu-boOd$o)$!07@$xTJ)mjSh!vLt
zxnj<(+S>7ymB<HWs+jBSf{N8+=^OMITdWLYj(d1*oWbU7&LuRq#Z9IA2Ztd;<WjMG
z=!2R;@6hDCHyAc=d)wiJ^D#p%SQH>>z!Dz-%?u?dyKKUeG9`t99clxUJsyf2PkLGv
zV3B-F^HqJB79>-(zR~9*MK2*BIp9wUc01v2E^C_dzHBI9-)8E!P3U*uR-{vgOYQaY
zdzF4O?)>QQrGXLpr5A7LyYi0<2hbNsoPm#wwFd=&5yl|6K$f@01}}xy7n+L=v(=V;
z-E}G{HhiaNghCO|oh0)$*@*Bm3UZ5ouym~hpX8>AcrUMY;Zgf^T;83RR?FF;W4Bmn
zV8Frp;`^SXIlCqHD_)hCc8F@Ij~r|Fq;DJ)aag>e)}J<eh5Yp`v1#p&bzK{Tj;8nM
z@ny8TMV6e#N-hzP^zE*LZ8v%&8S<3*^a(L;S%51uv>3}4PqP$h2N-_h@!bGA^U!&h
zatVMYcv5ig1wSLRUzLhdciGz+c|8sF!6JZ~Aw|kuh1{7b)~VSt5q|vQ!G&%EW7FdL
z(#_Vy*6UZGJJ})@XZTT2Lys<6LH`InBrE|U0M$bpU0#fleXx3$K53_qWu%r|X(%0H
zZ+iE?sMQs^eW_2^YlyCIfjLx!jqt+0Rde~S+x}yCu=ZNaX%_p;BLn6)hL#oo%;Iz*
zj1Kk)lI98{L~uuskz}ublE<2C$<E3bPWvrI*nPd7r$)_dVjQNi`9L1U(IZ*i)nJ@H
zRJpG>+$K*}x}vkP!s*LN06#jCBl#lB%a0|7UqzDsV7J`^)m^cf4jNr8mG<$F`G<$=
z#$3K3v5nWjD~8b{Szqubtsy$e@2*j`({;<(w{_)vVskht3!wZg?p-~5`lGB&^AFYy
zx)4#p!G)HPDc(mL2-C6!mBy410ujNTQKM8hvoc3$X#MC$?Et^lu|Z1_2Q}%$bD8L|
zo4CE2G~I&CE&OaNyfw`->tbtT4E#pv8jPzCt`r)iO}oO-JE6Cej~bjKED@T5LJ+{O
z;`Jo3dFalN%8Y$ewQrxE=>I$ZV)r=>xpz}1yk3-!?2t^7mh8Q6=9>K3^S>jHw30Lx
zWOOxxs2C2|Zp~|*Eiac=U4?PZlU^uVn_ZDr5Vdy3#AbORaD!qg`_)Q^L_2F@kGm>6
zoQz~I%nZL#{g#b5bVd}VtJ7xg8(EX!p<b=Nss6l-y3uFY6GT^s@<^S_FEAGJKc6&^
z7~?yrEH|$$;+@nOGgU<0a2KzprJ48pU7T-39Y*qPetRtQtd*=B<qE<$_RB3p+_$8w
zZF4Ezm&)5bnC-;Ek^v+GPPSM>T-fF9?v|+{yRf2n4RSgmWQ83AYfls}8_wt$yUf2<
zn7h(b&uy;0%9)OR?uYEdZQYD2iyj(>2={V3g+d1-`uStt-Jh!mPKb`2HC_~Lu(vC<
zLTTA~RX!}k5cz=kwGg$gF&#2a<8QQc^#(W3RU6Puwyzw!@<Yyj-Nyyr?Pe}LRkC2K
z&1Er4{J%il(TiP`$~9E%8y~+xES^{?7NGjAFZPCQKg2xS#TCKWXe#S+>xZ}qJGfHs
zqn{7$l+5*7?xQ?7lyUr{nB}hDWAQHC_v&ril&B0b`lqb)-6`CDn~I86J2D2=VON8n
ztYm#xfL+9M9p0Gs+nxr-y`3CX&(!qA)ir|~l&mM~MI9m%+93)RqeitC2^;Jlt)XXr
z*NeR{pyO!0)ihnSr=&?l(r9I8?WR~GAHj-%IdO|tC5x*K=ECKn2}+AJ*icuzJtVOX
z3uqQcpz7fUX9-Z43e4c(d0p;u8Ik<l2|K03YvNrBTFd*TDTj$G&U<F~75*jVIq@IT
zvXeBMXN`6XBcC{7&&N5nmqj#SOUH0C<O?PUM`_s|lVG|_KYiB3a^G8Rd9cK>Ff_=p
z@PfEX$MUN0pb(+TgS38Hr&!+}vaM9RXiQsO&fZl0zt!bKyM5`WzC`gMO%_q6O^7na
ze1sP1369s+jp=;kq8LxYzeU;1{+0y1T^=Y~JyN_^X6ixzz};CnE!~46$1GEpPK+m;
za&TPJsqBa{37CVioMNN(uD;y?%OhfM)k!|5rp?>zdBM%A(D=byU3-d}f%W^-;)oBH
z3{||7z8U33;^Cy?M2A=WeVMS7cKf14#-|7N(%ILmU(tv0`!gZXXsX0HHR|j`O$scD
zBEj9GAVCDs2t)%zI<~^k&||SiG8&>1M#~69emG>_EyIE|bd^#@srBXqtAjQ-$5MaC
zA29;v=BX~;yRa*L@wS!vS^KCqOU`dL>E=>G_-TC})tghPW&3!I?3j(AO|r|HY)BE{
zpCf_Re{w0j)F2W>G{vPcl8Jel(>GVmm%$(Ia#LOxM)@zxUFkE+`>yIr_J2<Li>4%m
zT;2KwugwhiAKK%8H_ZU&s1Pg9D8NNiA<IOgeftCi$rbZ!_pds4qjni-9m;b)9=zXp
z)(A4C$3nEbBHxh$40&Dd{2*;K3#(=~`&OMx5}2<lVBz~pbEiT5o1VRV#|K4Dt6kq+
zbv?bPn@Ve5oF#ki?#Za?G0m*QRJv5QD1&yPRCp^k?t$S=5%QfLcHy;`)x+W%4?CL7
z>QITCuHJ^a>vh~g^o@0Pzl-0_TBg`Hxq+6V*tdJUB_L&z#a2McU$%V2xhc=fm^Icj
z*CFGJBvz%1emn`e|1th5Mfzu<-~so~iyLj#&GKBu&4)QL+pDEyuM%=cuD0RyeN_`<
z+8-ZuowzqS8Q#y=WUui?+%6ui+1LRUg_acYm)()-gDpiny*H^vK>S0N&<-PgjD&>1
z3Mwjjza)yNC2tP3r|nw#Zdc*Wl)@Fv2w`v4#;$vX;~#>RPJEX>Qe557TKN{W2^C*a
zOKBu_+#~9A^2Fw`!q#o3%~y$(n|}3I#tV^l*9g%UgRf^qT(fubCXIaIeOP8#JAAu5
zh=2sT?FLV#fF#K<=YU>!n@+v&!fvDKK^LlPbFT7U9JIH~Kax7@j)R8R*TllWVl!)A
zQ`u}8fVMnRoYmOL(qLdM!g~ywCW$y5z9e0bn1*-l4kK~ncSjFb9vHN6q+A{N%}e{7
z6DlCgp*G8DPge)egm0qh+LPA1NGV?il{^Z%Z!g5~Gw1oULD5iIMKi#<c{w+)m4-*?
zMG;BwgwNfk-g0SD5R0sY;G=n+SAuR9xBCxdHm~_Lwlw>`y-Lc26-*a<w%WF9e@vjO
zAbS3HY_kCM1TD$fq?tEgtYk}qg~-<p!jjb!?Z;Cbl*F;?6KxDS7&sJbk~P|DG^WZB
znMI5a_$QV2Pis~IA~<<g{jB-u)<Kie4UMlqCb7&clBd4^?PDSIG!q{1l(V_XZf00z
zY{qIS(@=!+91ooEdvWMLhOsx#mn<u6aMbLXu@6IXIH=xmz8tH&=&E0|t^1FU2X3}=
zlaNCl<3Eu$>HqMSV7^S(xvGLk^IEt0-<x+LeO8J?S84j2q3cG^5_A?tM0mzoiZN1D
zI#SQ6OH1}>G&nddzS7=XXS>HwhU-Dnwe;RiA;JOsuju>4^x7Wi+uIV9RD9L6RXygc
z*Wz;lu2V7*!6J8eaE;raF4*VBM7fX(VCRmA$33a$Shq5ZdB76D4*@mZl+ls36H00^
z2ec~kIveKZm%kM+OA%GjYF=(BV419wYFpx9$J3|aDod5Ft-R+hP*>O8>}+S1^hUJ(
z0n_oF!j>g-Op$)7v%yA*?}&F+WJeWcyZWrrT-%awJw~-|t24424kL*Zwp+ZQJ%F2V
z{8zak$tgnaYxK33n`hx%=$)2-d5+iVbiZEViT(B}lbOj&4C00o{`YKHSy{=EwDrA!
zK^0b|SP)+_v|03g{z-pbn}yagwXr-VMpn!&f620_cXd(TMFRL7P>Kn>nHsg3W@NSZ
z{dT*vg29!ep2h-W_uFnynQ~-#4$vKz&5m7wy1paiYO|!QaL#1ngAGV=dB0tce$s2r
zeVO$d*!NTWAWFLo&%>mbY<^yPs=AVb(-&Fk$ll0?nE4LVW0OYHrqs-eVlI%bhVupO
z)1!(AhRuB@bO4*zT^PP_Bs1-oqSPpBq<CMN^r@!h`6Y8RY@ihIt;L>@Fpk}*L>>l~
zZcP=Q)>)f))^LWqICf1aB-%%aTuqS*epI_NW{LQ4obo&`Z!0lM+OCt43bds6SCTf|
zGW+%}HrA|qte{mkS;*2v(=9ChdAr;^IQjpa5hNGppF)Mm9XuQm;Yoyi2*h#0ka^Q<
z?J+QEoS%08Dx`9reaJspa^x>b_kTEB{=^2E^HeyaOTt;oFfAJ`1sN-iHa~^u8schc
zl5VA9`$eVgy1Fi*7uB<u^FTTeoZB#~LudgiiLzagUvy-1vi*!zLc@h(UUOpaN&eaz
zus<R|cg}l3@vrVy=I^tMM$<xT3c{tvVp~#ecbzunFYorv(!kvhagWX|4@%yWD}$q^
z3%6WYdL#7z=z0@)EZ6RPoCZ<IkW!H$Gf71uLqwV9A!TTwqJapRE3?Y*m`=tL;!%`B
zWGE_AQG`lJNT$qW{;m7bIq!Mj@Be>3_s8jUI)~@FuWMg>uf6tKU46a9&7-90>z<)j
z2Sa1VuUD-bE2u7QWG(FW;-#yk5Nj!vLAmJWUq#s@Yn9zG+N20v)f8P8GE1UrMtIrI
zL>5ijnQNX3khn?Sqt5NhaajF)!cBEXfAM<iYPRi{m1Nr5d4|L8x3BcH=lihQ=yPh@
zD(=zG-@Q5pw)Z}`unp#zNO@3ye{Vm`!VZWPJ}mq_&iaZwWHjTgbe*-h&J;G+dZ?(J
zwH2(I;}1PW2z)>?fDm55RQ$Vlw>0#AO}c~dg@?8xoOLzogyMCnWx8d}16d}tvnoDg
z=LBjGW>>v9wQL3TpN5%BY-IAJpJ}f@|8F<a?$te;*iNorpLR!!)@^e@guiA>&P@H{
zCXJeU71_zjr>bj-`-3+z1q^RY)k81^{8eq5L;UK&-N^+%X<H0S(E<-rEz)~nW!MON
zEAFECtp}97n{q1LSLCLO4#bNK*A8fY2QxA36`=r_sKctb$<1u)u4ZjtTGVQb_uBX6
z(<7@d%c$teHL|`>x*=!k$El9%W~umw=MI*8NiT>SI&q@IiehSwAX)ynG#XY=8vM~W
z#pcH4c8-?xx)CaQ6zYkCh3mbb2Q8Z;?P%3`dslezLi`>^cuuP#)vE6@DKV;Hdb4B&
zsSmIFUco8p!jpS&J9s5lNdmI|kSgn)KIQwhK6SObSNTZ*KJ+vq!AAr<YWjDA0|E_@
zJfEHg@e5;6VEd)rJvq@OHhpZ&2~$~KH&3vG$Ah%N<SVm2hQ+~YCv+a#+rLA(ovs#2
z4}@5w^=|!X(?L-%eQ=C@c_vDwN_Jjs>|X`&XnB<Mh3NS2^DF-kdv3q{!13afbbEi*
zd_Byi_&3$GyXSlYiKM^yWSVWMtahmdj_&&>Y$O_OZ*mk1`gOB;J!*iwh`YR$gzRlC
zH0h}I#<tfVwo?8?>CMXyI`f$6n#3fh*&LVW-&b<}acS|Tqb|+3#y?X=;x&#p+jL9N
zN3-9#QP<dKB(KmWfDS*8QSCW&>G*wv{J3+kic|koKRae`@uu1TZ07Ucr-~E&Db#Wr
z`i)`!i_<?V4ZCU86=E)1w>U>rZ4YDn3|VFTGDGa7HEL+&T&SU$R>4L#mY+IfI*j6E
z=wPug;(JM-*J0YvHxt&Gw?L1MfVr*+cJGLoQZ5=QR8Q!2OZpl4Emi)$?ikC&RFj9b
zVUd;l_hj*g@D|lNt4kb$0^9ppIr){`Kt0E1^$5j)5EdX&2U##nq;&VE;8d|858F!3
zxnsZXc8}i`RH&)h<@hU4Omk?fT2dsYYcZ)ocp@wt&=R2#+S2zv2t8iQAh-_2iQD?r
zxse7hJ7srgsWGR;4|B6`a;m0=|K(UCh_M9p=RdU}8Jk$2yzm8+zjF=*+SKyuYQ(j5
zpZCt68LkD*LDQP)a4bp1@;f&p4_{c7+l3O#EV@ll0xq%Ct**hc)4hc%18?C9y_LQ0
z?hZ%yHm|l=SP<zL=&NvdPZIy!pIJM$hdEaD><3k_bfBRn=OUXZ&=BREX%>5D((5Ph
zRm;=NR~j2`ibqs+MCzD8F(XVsb3ucKPyx<7&CH~N9iT3A`NJ_?G|*1L$?=X@-y2@`
zV=<FwBU@s7gW3BC3tX6rhkOYfJnTp7)w5PUEuL=JPmW7nIyTBtT|LU_+Sh5Ov6!k<
z0r@`T?BLnqt6VKBmAgID3O1>;c&+ACFS<ooA_(?}PJ7NBEK>}pi^z4vR1CO;;D;f8
z^|vekQk<A`R6h|lT_VmtKzGZro}aFE){FG?aoWS3kwP|HeS5DCx<ZE*bA@=T670~S
zA=G!_UYvTyZD?AShZwP)Qx>SIl`<aQ6SXL5w)sb<`c(I>3xZefnVxKQ{r70(KXi(3
zgLgLs<nv$2GpSOrpZp(UA;(p@$*!eyUn(!iz3hu``^mXxx5J(L68+^Hcz_Kq%u4}3
z4S+NEE<A+J`X!bRq;Riroswu9f7QA+K8^uR_kpm+j>cJ*8?#*#evy}tK1t;rXb2GJ
ztu>t~45D#&PhNDApE;syOzlnoySCn$N7GxpCt7K}7Z-HFNB3A9jd2gpAt{TI9*Z;H
zI3kuCtEKgRny2H&E1~#QA<eb9e2+C6a)YXQBdZ70u4h%%;PQfwFkMZN^<BXt4Vh<J
z0z5eHZ!Bmudz=bwGp~TpSJ)mLF4i%i$Qq$z#dfvWKK0i9vcug80x4-v@3n6?*!jX$
z9pYq&UdWL#i}R&;IM2I%MMQ<_7>q$M;i@6m?31G#k^sJ_mcTN{FF1Ua?KB`D=V88H
zZ7*KZh1ULsox5|N*7msXOnLG&$A&k(SP;x6uqZT~H@-YN%V2#n{e^-7k<6{OFle6J
z?5wo;IB6^!a=J4#Ipqqk|1GWsR`rXC?px2>!&3hih{hp0({)>+3rA?(V%eFcFi0Dj
z{;WL_LX}#J_7lmq+f%?db+@#z@G&Z*Tuxxin_`{m)U#jDI-BVVHQQK!UVru0QgcpM
zQSpQIvp42Nujd9;*7g{r#Rq>B%YDL^R(-0+iNCLwpgf*0nZH<KDJpY-&t}Gh&snC#
zGj+agw%257E-h5$(EEKB8=51d`Q5a+R6eB9STHWn|Dw1-8b9qdxyQn2<BWc+-(~eL
z(_X-=;8LQdpd(lp6r@7yrZ6@m4z8%hKAy#aUK1qJ7CIZdi=sCU(0ve0fn6uR%~4`-
z^P}IB>6_xh0~d@Vj;FPWhVj=Qw)fRcMD4q*QG=NX!Y84v?QUZqAsjRdx+YysLtS!x
zuie-PCMO}TF)hlww$|&q)~b(!G#Xp2>>CG$JmuZYy~U@cvC}>*2~Vrcrs}(KJ=P^k
zQ4gC&jkB|4nghFou3dOIB=nE|?)0PSifz79W1VcHiE%!;ls+lJUS{m+PA+}5-TB>%
zbwnet-b)BbZ<WHX<JRBG^Vda1xgM77Y2+LyZY7R&voF%<kJfzaf&c-X#_o*jP-92h
zhVKtNn{ujWuX@PpAK8@Lu=~m2z8(8-JQIsgx*{N4L*f=#ImbJF!<u`JDdGs9jn`GH
zn9n+zHKgzu%NVop<h0164U;K?QY-aHxqPDNgd**TWKl8DsjcA^^hmm&+KX;BVdh<`
z;nHY0pPHR2_nh{O9D<g6Rj;msuWdwdcmZExidf55uhmowvlY`q5mSDpV}6TC5&3UD
zemz&$J-~eFnEi0bw|6#G$@T6fI(&kV=Yj^99cI0Vn;l}c7cX8U<elJ`_$v<t&&oqz
z#(1~GERBC>t$fheNGi3b!;`Pvn5TL?Gm{VAc%4F;M|_@(?djh)>C->GyXdEwvl0dD
zb=(m(Wh5(OmR;EJ=6gx)#reZSW6}IJ%}f91f+?P?l_!%EJv@X26{kC+FqkvfI})4U
zHd9ZCvGag8;VV`sv^ye61c$Bi>gB&`fc?h3xs;%0rX~25*^zc)`kr`W4Hb2`)+bi<
zCWMb9#p~EdRrU$L<6RlVE_Pvs1xwPq(;Ak|vF=f?Y?w*#F50#ou8=}E&T{97D#q5Q
zDnw|+GBkX@YH2^_sV>{RT0;JW*``2eb+Z~9w7Cz)yxNNf1cZX-N?EEWy3*EtbELOk
z@(ZuvE+-|uIBOQ8(Xz@rVL$EXO^({IX|zab5&Wo^&PQc5C(=eknhu5Q*+@3J`Fl-|
z%0}mlzg}v=%qI#1Zu*1KcRX=H!?Y+oDhIwGSo+Z`2}iE+Cx(bsior?XQltl6XXbQ4
zsonJi(;jn?%Kh>(D#@u2roAq%Y^4S2)xI^}M2;1a3522xt>&wLZ2p_AQe2+2ySen8
zgOl(nDb*yUsn8Qs8Wh;&56~`iK$xB?%-z+cU$w0_PvpPtiPa8$TDblf!=lIKz|oTd
zBwMZ`A;=;3@>A=VD)29Yz0%HfSHIx|2H?O9S(Yv&F@e}%`&lYcwz$v3qUIA{r~MYa
z_p8q``}g#^hvYse=T6=+lSpqJB#{zp92=!t*T8QTVc$^edPD9(ta`VdMSH}CQxO|f
z2r15B6>E75<Jgo&mb~-4VrXGjUfNqD9u+}Xy|E}YCMIgqDKbexw@Tp?pYT96OpV{3
zqx<l>bs|;FR*(%Ux7IhN13W`j&P!>tQpS<V?vcUgHY{1ey0iSUtSY1rFoq4aMnX6b
zx?BU3fN=jni;%@cm4X<=Z(_G{L~V<%?ypvUHghYBL+?$s-VY0QFCMl$Y&C<Gr-fa?
z+=%t{Fz=ye(`!70P;!sxAisz2wArA3wdIZoi|Y1WZ!Lqzh~~AHZu^Wgt9nhj$qwdp
z>~jjlUc4i)?5%v*(xBJNrTYtWNPX8?b063g%r~hgJ8t8UjCN8;P&E2yyMmxxpeGXj
zO`0)wg+QT7nZk|pVvajTj+n%}^s}PVBOMa>$x_zq7Up;A_NqHMZ|>Z`7p#|{vw5TX
zk(%w_bI%t=C5T5R{Y(*juI8pf#oAY!E5<JjTUrF>fq|4-0exj%l^-(K!#(fl$};WJ
zU}<(%*RbYSo(fL-A>1PGrmL~`wpPH2Et@~YHZ{k7)G3H)pHp$P-L7#GK7+=r`uBtk
zgg*`!hPo)}4m3y1^`dCVYlAx6;-Y1m=#6Jzp?fww&8+(2Q0EosZY&IlR{msVv@-Da
z;%dkd&bgZ-wEe=$g~cFZNfp&{&T~({o0T&Bl4-q9CxNQWO=_NbDtYvLXd_Xa|8Yro
zyDs&cv!*&{u1Jyd;k`e~ihdl|Z$;`mzHB@1_>)l6nkRjJU-WII=kVF^IuDmGfYooN
zMgXvdbs$lQLKF)aUW|>f?W;=(P=NuOrEcCl^ZCI1-bm7flDEZ4B8^uoG8bnh<v!gt
zF~MfhOd7)wVxD#Cy!j@_iL#XkPA=uj`l-Dco7T+D1r<I!VbQ~x+N>s{EM`!1@twN4
zpY<j$13%ikj*PG%W_UWL5Lmq0acIM1Bi-59cb7PH{CKFsrCqBvm~v-I!$z-QDiC(G
zcW=1sU4CpVQ0Y&}V1z=9;=|7eYJCmvI+B=$c~it_LUokT-y`(e$LkddwR}RWOc<nh
z;7+4g=4;)3vQF_%Z*P?&W_ftRA_g)_MOKU))JeZH@40AM_Eq2D^itP<$701#Ui@?C
zKUyAI)iV?s)#0B&Kd3~Al+ntFeSS7LRCoUip`E%_Wn)9^{*SSjW3L&_>_ZkwNXkN(
ze;Mz>d?Oe`+_adCL&BYQ&~ajLKt4&wg0)j)74_%K`kP;DKVjCe!sf-n2AyOLpO}{S
zhK7^FHjS%)YF{7AS=^?qrhtK2O1A->N<PKg(n?jO%%64lF3bvC&*k%kg;Bn0uw8A{
zDh(M}-zf#XUrd^_op^=uO<49ZRbEPYR~`K_|4iOdev?DNDYUMxt`v=M&zY1}Jp@<^
z_yI0{>Y9-9Ofy|s^={Q2xu^&%k<5l!yETqEywf{b1YYDfgYYxNvNYG~V?SCo|MB6q
z3R$(8OyRK&pZ5oR7S(Sj=2u*0X3TW2IPc~aT#7nO8dKPL2;@n@?NWzmtEmIa2Ia&4
z!#&;n9%6tn^U&QFWqnXyJsbx+=yD7!&rsBHFk*Ti1Z*&|jlQs^sVG}+l9y?KsD49k
zdnwahC51ruhE&g`tQIO#^@rJW=QWgL^1*Vx!u+xx%86OV7=*!tMu?QY6ETtdS9jHi
z;fIq0$J`^;CH>U)V-lv?LuHkX?eP40=bNVvFFl`}N^AcW<$kKB!v#X!NRuHDflS#l
zCzgr0ghmNR5tv-GUEBmKPTAg6E^MIgE1y2rHME+tXuR0lJU$aP+p{sOW<Y46wVV(@
z{bx~LT$y;MQ!mTDIXcoODfNFCxqjjMc$NvJ(6GUoI}&QavWEZNdz9xom?EG^T+kn+
z>WDP88Ihtt2tH{v&C7{3%2Snw%8vmR3aAE8jC?iPsGjw8RjVY}mC(&cdt+P<Nhwqj
ziG{F}LQ@8xNGL?w?Ajn_;kN1;DqcQhPp_;ukMzz({LFnQ{#bMACj6Gl*#x6}uun3Z
zUQW!0M>)24@JGmMwA4bWgJPtCTY^Z47U#!x3A+{O(!%KnuoaOgU6xJzi|A<=vrbH9
zii_0NY@6-6y72Iui`GBvjji7CjkEg?L^<l53z=74wt!tOsOMdBE*R)||DtsIs?Lop
z%O(b(w-;lLlEhB$j6$+t%ssFAJ~@nf93XXpowlx`ld!YFJaO6qW@W~TdcTuY^&6vo
zrrhyIz>$MsHV7*?h09z$FeW;tv7xTyVB{;c2$N81=}6|CEuLksLkcM!O8SeAIz0S9
zL+}fT15dWWMjGOL8lND00htB4%O7_8n-4c+g<vI-nFBXPM8_gzZRmm46jm5`3_Mr;
zPN*pna+q%Se~KOnZxQI2zRqGlvvK&QXUhia+$;1?OjC;g48Wn^rgNwflMi09>N9!O
ze*c6Po}B9Y4`Zr%s6O6_p>MC6=3RlwW~t*Zb=IJp1ktMHi@E8*z!kal3@x>QyH8bV
zKd+YWt9G^^u(K4($ZV;G!~GYBrf%_lt6`}|?6|~%PAw)s9=v^Jy5d0aJxUJ@AHDLD
zQGfn}%Tn?0ghJB8Re&`}08+uxBV>blG}1`!U(e}!*KQ}3?^w7F?`8;!|DK%`rC5`|
zqw0@THX`t{t_^!Y@fqkEDYrB4?sGKEh)|eO|6(#+aw~5KI$s9DH4h5@rv`lVJYoKU
zV6m)Bc%)B_C0~ygEgCKYe#QG3eY8C{CfmWzW1~98NN9v7RM`5ia)alIVT5HfEKGSu
zlQ~qGIVwBdF#SPWy=!-;L4z$dBvFYG1^<wrm3#)d_T8mN8yuWRjrKC@(J@=|Vt!nN
zy8RLd=GNE)l$6|7;>b{JlE@y#c(Y(Q8Dy$vyX8;>`y&)yqOXMBDsi*GB<**v;!?2B
z!hjVKK>;Ow>_3;QZ$SCPn{v0@)GGd1NuOphlGJLt(|0oeKi;4{tDD=`W59fH;(>33
ztBlzdWA)pPV;8L<ipEp<YKz%3g<mTFH3*8VjN(ay*=<OK8?0pO(^<`^OmuYnw4zfV
zQmXjtxE)*bn;dHrGKd1Bn{~6?j*T~{uHvnQNAX7(k?zobUPF=-7*>sl;@1Wyxbbch
zli3H69eKxM(!AF^N*1_9`}N6ZdRQ0l+T27;b>6q`nGTv**}rh@5(gleE~pvewjdON
z3+y{adXs&=<IF?FaB^#n6ES-@RtmVOKk0S;g-`j3N3I#`n$ie~*;fpEbMDx$mL?pj
zJ<<k!ihrhIQBR5$=a0&g{*m%xQl3&j{)IBt$ghkEPjI9)3~3%++A9fm%sSHP`|Im;
z?i=DoOI>}qGn0rf;hra<IF%m@pqSu_+_9EIPKIglZ>0l8wAT+Xj117__*KmvU>P9v
zaTa*2Zl@CS0U1t3slv5<AF*UGu8$Kq-|AhEEX5oIF_bS*X@Y7w;Xv?5I3K7WB0a2^
zktr9NZ!7vqq_#5LAOLx6Tv%NFyIlb5Y?nu)Y`Ly9{*T%Cf0Yq26V_f~J4sRhgFJ<6
z{U@^B0vj+@OkJe5XH1{8yU_sqARtwwhNXbwe+vC?;%#7{4(otv`IqBzp!tIRMInG;
zz5uK}K9TYEz~YSoC-_*RI4iFW&)G#@HrJ?83HNjT7Rbc=J#eBG_eE3K!P4!{`IJ6!
zbb=gs8Qa2|+?M2h>Xt>PAaN>n7byx9?K_b=Ju4|uh1mskW^ty!OQ+N85;5J$x*_dr
z<Ik6ecE@BCsCPIWJGTkI_Ab>Ky5EfdzU!{`RhwOD>$XgmCneJ<%{29NbGjC-B>W<(
zA|flt-$K8HNSCaSpMwvl&DLvND^JcgRB}Xd`*Cb_WW3pRgu!y!U7ip=zWFTM<L61@
z^LIxxnSC6W1Eqk%0LANTdO`lWReyeaj{k%96J{Hs2{^2%KJH?%DSN<ik};P#LV4Bg
zBiYlP;lpam07_uX2Ehne&*NF18olxQ`}@~&LM$d&i)H5`qG_)Pt0h9YnXp>IUcj34
zUDjYu!mH;Y%cCDnb-e49OoarC=o69^T+s?ph&zq6MgBKlB3AtCN69ZS;Xm%>V2;Nu
z6Ur4H@yFBG4RA7=H%Blz-b8eP93f&0LJoEvk`_F9^ZC7QP-aSGB@P$K$~upo!tx_r
zf8o~f(1e-_o-9?oIUx#nc4}_dyq=3#_PKRjF8iBOl1i)s*vemouJJ{%IbxvuuHFGy
ziG%Nhp$0slKlHj;%zNvCVX}OOEu%}FVTy|{2t!&z<Q0fG4mb=rFf873kXZc0?MjDo
zXK!q?dx{-|d7ygjMr*EF#y`V+!dK2j{_pvos-J%59A@{oSRa$4+=B@J+Mh1}Q7czF
zd>siw{Y200zd(bNF~JBm?4hvV&bOMB8-RQVZc>*tAxP!5v^+jtwJu_AD+WW{l8SzS
zC5G$P*$P?$9bOFrZt^MuK=?0^(g%ygIVv%*@#sQ4v^7}z?(NbxOmnOi_*p*Yu~Ad?
zFS-gQ9BF(woNa6cNZepB3*j%ci|Rni@PIt`bQyjdjX2_hTb#`C>HNnPlPHflDu-vi
zFt*_rJ{)vMOle!XEn!e6LC@xirTN(XcDf<monbNSu=?hzFrNcS0{@a;p^W%_D-N8n
zUN9|7;LTecDD!;+rGy-=(zbg^2Z`Z>G4bk_&wpl%)C9{7IoG%s?Z(;1TqOi=c_D9b
zX#2A748B6@;MJwKbXyRXs+gf_^f5IRBxxua;W_&74d1ccO%VCI&<b|Z!v>z}YAA&`
zTks1x9Y~h*SF3?cH%0~NV(`1UV$o!=R*yr;H({X(2{Dwe#{k~HI}LTCe#_y%2DFCj
zq5rAzwUpAEWKi%~yh+#$K`9<Kf}l;~j~1^du0L9;3_t7JLV_K&)>^jbs>-TezP@Hf
zl?vr{(=(e4YYvzgDU6xty8e>^k#8i>)db20%(=B$4I0li=^o?09T8a|_GXGun@hO1
zWxHSTv77YfIt9IdCP3X!I?jr|*kGevHOkw&PhBekJ%+H#Afg2UK|+K&{uRVo<?B$(
zHI!8~%ZCJb$24nos$X#w`g(o_9n&DpF3)qbl7#tf#`aiGiWcs}E-A^8tM|Ros<GCs
zeqmpOPufmJh4^<Y@$)cQ;JRSZH_7cxO$AFD>>I+N9=Bm{eKMsan44li1`LhqgbYn$
zXRJan7I<Cll{XPP&=pdg{5ff<zifDeO{w9S_FmQI!DvEgr7SW$vLHLE!bhY%F(F}g
z^3A7+w^h|Si+TcJn5dcjQ4#T_H>&!k>Tatyvo8`7E0sYBb5M(E9*~0A8wThgl~>=3
zSuZ==i|Gf5k6no{N6{yM7{n>rM08M~4nl}DQS{-_c&B=C&x23dMV0gh#E)gFD<~l;
zL<o5u?>lSF`ezzGope&L`YX&2D(ia3G3vFaO64PEyqUT3v4(CE(C1^~v4{5ytD4Z0
zj$en6ickn_NR6(K5h+4X?Ixm5LY9bE6WmMuV1Bgjc8=}_e?8^8+GUyeJ#QwbJwl_V
z?Cha2PSU?(htjsl$ZHkRClwvX#=R56m#<DH!pliHH2J*wQEY=->NTyofF2TQc{|ku
zuZu+phhfc69h5cV*M?ytKDCJcZ8`(uRE$DH3NvlX5-5l$;5?!&3At(t<&N%JR9=-P
z!M$sPT^U5c;3~mqLsS8broXs#Q*^~zI&>*Wzw<C}<ooI{$v#ShyJ=MvN_q-m0<rLf
zS+l<PRQEPhmS@OZ0Qdnn!^M$Bin0e{4B2e&vSg;4SO6yFRqEP%A^(>)AZn<_CA+$(
z?RRB_fUsgyT=(<sf6L{8TjARV9YOidCYfzAZZ}F6d5@T>*W6}`a^e*ncCAOYB{z22
z3eJb~+ddMGL63wt+ZdkAr9ZFt5w1^rNU4}SUG7>@WG6A#AmV?6cmp%40;iD-!94>v
zC=8Y*jsiAw+XgN!Bqmor@{`+yq3tgZ|CUNc;Kh9jZ3E#XQP=SM;nCTM%dBExG*a{z
zUr=0SgH>$C5@D*rGS0Pa+wy?dvexd^;mHB%I{jcNUIERY(0C-W3Dk^8us*gRk0y$8
zVTgs$@~9mCmE2>gYI07kKng7+mfHv!Zz2zF>eT+WpZ9-MW{5tyo{snYxpC?Lq!hGn
z*DI{f7VuN*=Y6UCJvim3l?Q5-G{)8vI&-*I5YmoI9H1cis`yG#^E2Um$h(&K0r(W*
zOh!`UXCTA&{F0GwUfzZH%+&7J6k+b@BS*z!jqM6(YfwB%YHQSVt1pY#QZ6P?68gQ2
z#A7M!R?k?DD|H&g_&B&9|A3|l$Ou3f=<q_z=hM`s3E_Ux)vqtz#Az9xHg+Ea&I;Wp
zh&tomTTKx_mJ5D*P}L~L!|wA$Wtz{Sq#wzR+bF})r-nz`!vA$tC~5T6Kvi}5iBoTc
za6qcE9DZvKUGyYVJcyNPPECXcNJb%Q3`1GgC}oUdV5UUL1|Tv6Eudf>GvZ`+hhfRG
zdw4@lM50y7A6{sst$^<pzf5JW7^-JX+u&TVn5&96MLvgd`2BEeS(F<xnz}vt7xwSE
zO`1O-(1mstG8m*1lV+g%p{mRkyOQJzL%1!Dx?5G5@!_x~aWF=QTK5q(6w(V+k(9dK
zPn3pu4c+S2gL-9w?N1-)c<w{b-Q_3mAVcZ8P~Sn$%H@DbC{qnh-1K#l>#NQY6<ikx
zr6Gz{0Igexbyx0l&gA}Bln@sY4l%kK%($wARjJ{$b*2E4b{7aC!DR_mRQFPSd)eiE
zp=q`$Kh`n1j_7Ed{`G@7qE9L#PbViXQ6t&)*7F~e!$p;onqRlCIpZ)!zkT~x?bAgg
zYf0R{eqQ_Dy25!^<Xa(T?vBb${Dmqkt=wx@O?gij?AuCVyFQ|1*iE|{#^nOkf`WoV
z4`el6A1EU4?sJefV$<;DhMkR2i1Nd|2PX_zM21|RqsfXGNj#7kc6kmiynj^iHp7P#
z=Ln-;RH-9!_JZBwnim@nx+QH7;RQhSS9&nw53ogqpo;}rvu{Gx<I?TE51~?A)?UvL
z5$^c<mC!>{XDKaXW@i2nNoMUFlQYd?sGiPR9ezc7+_P+lIAfci>D`6#H&1GRkUNa^
z7mj|E{Oap!xlLy=ea&Fos9$9A=!1^Yu7dT#PuzDie-M?P-#V;0-*AS%CpM$bF`MRB
z>JuJq!&KWEvWUuMmelQ1>MU#qmSP)`HQ-Z0IcUgb?!`ksUfvY124@R@HJAxNLhmu=
z4g+`^B>l4oaA}*xT=L)S9$}KTx-}np9r>#)z&^FYwu^u&0;2lN;EyHx8~_C|tSX?@
z=$AcUzIuReH?`G@LS$~3r$^?7x#ys+D}ra8=xcISNPJM-Q)2pt+yqnQgx&+z1h_Dy
z+8%#C|2vKcKaQaBx~*)g*Z9=K`GU9XiyEiZZamKZMn6%vB){WFC7VI>uch@BdbbSg
z%i7)Z$?DJ6*xRlzz4GB6V?by3nI-lY#{!}@>j?1jQk{tjOkQ|-H~s^yjTt%q?d4`k
zZU4v{DSB45Z_o0koDy95DOc<lUJFIzmJkd_^hxR+bl~UZmE|=^d<w4)4Bf@Qqltfa
z$Un;FxXbgl+?3rVK`?c>BNknB_#QJN1K29aOHQ!c%~+>YswFc}8SerM92y(MsA)lk
zgu!H@OSDa<z->PD6&5Ya;uTor$88rQ();u|TN@>TaNwPwqfc0IA!#8T+*Z^;7IXE}
zVC>P-R8V!?BZrME+4}~vyBHQ9TW}O@_HbG%sI#kiXHZ}E<<BbYN-z(+Q06UdR_v`d
z<3D`9{yB7JR8+KIulv>;sk54D|DL;PV>(H_YK@RNaOykED@zxbephHMgab>@m=UsG
z%X-Kdw*J6yoRyh5u_#yU1Un5PASxA(O5)@a5`nbpB<|3zM5k3LWIHRr1VOtvnil6}
zj~IN#_MrG(Vfxjg#Z2{fDY+;ize^gUIk>AYaR5>&)e`8p-@6rVL=Zj&p{)zU@!+?C
zb`xVGRca3nNl=`SDWbe4f)T``kf~sKh;(=1)?F7~Ufx+!zs3XO-EIX!YmHTJw@B7c
zw|<u<3+<bGaQmw4BI7_yN_ywE;(M9vce=lN$*DWpU9e8@Mhk=RaEbfaz-k8e6WqM+
zTP<tlF0*bqC?ce}#+Sal>dvjVS9M5!z8i5``1DA<g==}<^%C}(d19_itFUd|k}4!;
zk`Ah<e7roixHCG6n!9#4Dnzl^+QfKsQD*tT6YS7rwaaZqEU|KZPGFW6E&4G%by6}P
zT402BBqq;kk2Y(-i32T@iFAE(r{(!7f1|iWO-H0-sCh7fAVL4TwCu@Kx5iZ45N_I!
zWyARRgi7=L%%0`UGQV}lm|X9GFR?Q+w&=wi&rP@O|Jq258~dANy$&B+H^0~3MXm7m
znQ9~BvW^pf;>HpALdI6FxvRTJj`H-B1V&o@PTl!Jd>d~{!DQ{X`v>>k)wiZgPZ4sF
zWx^?qOIK?&+VNRM<?>pxl8qU!E3}PiSKzIj-@EmE!Le7D=o=+LAu{R}L-djgGiNnE
z=wlWG6o#OVBAcbSbaNDC8X1NquEF-`n1-Ic@nko&IZ+j%HUi99A;Io>-=3HH%N#Xx
zqa<XCzDxhG9?!uyULCFIH4DGwb@k(S24vxQK-SywayVTmgrGn|`S<Y=f{Xb_E(YNP
zaMV#A?s}(E9jwVuvyhYnF*anbVt0lMt$R5tWiGRB?a5(ynEIAE{6|*r@S~)Mb}YYM
z9-|j*Iq^J@PVmOUa=1vz`E4`O`Cj|%&Hg*6y2y@(`Wvdo&5aFWLf7!780s-_n1=Zp
zm{|%qsHuI<A3DTv6er}v!KP_PptBhmEjR^s=I+~=wt;l9bf}w`%6cPH0lU1TU?wi(
zR27{Lz0DT^9lfdR3bIUbr*Clfx*w8g8)sL}A7&iZD9K<J!$Xg`5JR9yz^+}MGeC@R
z!)gjeJ#_Zi!+VtiwLjcKmt_$#j|mGd0uR7lvNs<8A$ClP5W5~CUrLThEsLr?w>AuG
z?+VCNyQQ1Y5vJMko1TSM9NgFW`giMveD}r0{?B9qoRg-H4!a$+sCy~`B401MJ4|iU
zJ1u=X=;w^a$DUiW=e+x#kj6Btju(51*N!Q%hw<3(%JT1i4|Fo_5f5jM!7kp~-BLfT
zHmKf1ix=P9q9_}xLx9=ws`23C;KAN8PTO7YWGTpewRQaP?ceLZGJY&*9ZCe4X7X!#
zF3e<eS0SI5s6`7@56CCa#HAORobahT=r|OUE)3S`%Q!5EWe&h`>=dxXr*lLzUu6#z
z745ZvA836q!b86Rb<exgzI^IN(zo$5B1o{v*pn9R`&eVkw>c);G4r`6<tk<JE}F#W
zc$Zbo=4IGETQhvNzv9NWOAeK+I*T<=O7*49Jbly(c{;mWyiW6`RGd@Rnt3}m@<dVk
z@}4c7-c6jFwW4)sX2fb09rbPm>8Lw@zK|JL+;$P#v^baL5-J>(Rk`$lAm^TJkHKb?
z#<7RoDUvTP@88s@-SDWBFz!Y0D4w^GsgKf>=)|YnV(4v$8-^_qfC;f5cnJSZFIwL*
z&Gn%CrIVVwu-m(;fp<jahE?b+%3dpMYJd4YC48Gye~@Ir;Csseq<SHJM^96*aKm|V
z?g7n*aXDxe5_b!{*_ZE@U@!#`dx6T%{w9YB5(~BEHobG%oi#7Q#|n;|p%YMF5EULT
z`0P13$Jv~ts$C&}<$h*Uq^>~1{DR`5K*u(Nb`R+pf8E&?gEc(7DFS2{c6>}KJ>S8m
zJGpOz1bAiHnqD3?qmiavfpo}N{W0Y}_*ukAIL<axAJbA%QJG7%-J;b9+xOScU56jF
zUBq>$zrla)JqW-o$Cv;rU327SP%FBy6=poJ_+FTs{*8%*(ZvN1=!L=OD5Yg!Gy$|!
z0X0yy5MqX0K!^Yv!k9fzXzd17YKs=v>zVhn-HYqSIk4_jsHSk0;l9RK6kmAc>a(1`
z3DoYU#I&2-v?G0ug7XW*K7>LY;b{WA5ob&&<cQx@BHRZ1MfJ!jo1|pue2m(4XtpT-
zMCVeIOoh*bi;Qhif=$zdu}Sxm%F6zA9S=3{Oc1l&W<Z>$3>S^b(GmM^PZh%$a;vyw
zjP&brZyLW0`k9Jk2kOhKs8OBbtRttu4+t^UbX%cX1Ajw&=Zu+}Tte}C(!GE7EVd0<
z-!U<Y54Itvslty_#EIei3HA5&it|?qJ=;JoZ$eKDyXni&l&?sQN{4c`Vz<;j^lK>=
z@W{5XV{^s2Qpe8-)v_SN0ib5b<s!>DX+}0`8MQZUl;l$<U;_wiV!O;Q&JF)wnC=+G
z`Ve<EqCEPKEh`Sz;8jkrObd3$rDv{nsp8YF&9!sOHafFZvUE3w$0nUB?%_WR@}Hxo
zmu)Ov_-gy(w%A(@!}53DIv{;5l}PdY%J;}Tk$Fe$zFnzWD)+S-B{!&EL!&vl9`;uE
zVg{#(>0W4YGfuHEGjlRq5T-F03kQEF-4@p~vXT&9Y#!jIJ&x5PHLLIkq(j&4O@hOD
zC6~`vIw5#pNQW8@ggp4j;rRA3G4GKT9L4hBp^4j><hqt<o8Tdr8+hO7w%@}s8kd7u
zFctECh;#tkm$><5p7XwK2to|{I)n-B>;zre(&B!?O$q-A3|2C0zIZ0%Jf*~jmvb~Z
z(VC^<I%hL-;RxjEu^+MbUH^M&any&iXmd?z{Y5VA&<|PUN;VDS|If_c>i6ieUxB^;
zQCnGbyxYzQ@RAE@QlsoiOg^n|M~Qf3XgB<rI$EyYIE^!t-o&CjIOwMlHWEGpZU#)E
z<A?lwK3zm)<?+L~r9Uu7^w=>VTNlI%;nm_(;E%mN32UMEc6s6HeRt!v)}+|RtqozM
zS(yN#TbP~_r;QNjM0XZ-jPkI`oZg1lb?0(mAA_fLsGLw7Urkk&d(%r6lUoURRq^lO
z1N7L*Q*r-(!xrzP-pp#z=^SdSJ;h>Whd&=m>IGzhqJS_oMdr`tGk;AH?MdPv3k&s)
z^EMkH8l`wujeGVcq`J3h?HspXSRZ(C#nmtC=Nk8X9CtlZZz8pNx}9_A<))qPA74r&
z%nx$;gxP1L7Ee2jtxKL*{VDpe%KdEN%9mD~xHcOs(&YMGBJtT(e@880vpH5=ASKd>
z^LWf!W8}TyYXY*RU4iW)SATavma>-*t2Ii(>;+u~D6Qe&aMJ=p!Ujab9>GV4%!0%h
zcu~BBnND&=K`ke8JOIAM9@C8Y7~8#P`vlnxw$anTwdO&JmPZ;^lNTEX??UJPN(yd;
zd6leFABdeMR*RsA|FQA6bj<z=t)*Ctq24*Fe^~jgz0}NswRBaZg0w5lVO%%e;Zr$W
zwiHUN6(O5?UpJYc(^CqyTrrQxDWZhDG}E9`Fvk|-Y|)!nw=S_&Ve$LaXxPyA;}@<p
zbsRBNUd>=qU-s-@q3J@)UW6tlN@xnQtj-x<H=nb9Y3TjCvk56eSgi(0fayr~ZR2>g
z4Ncm3QX=^)6*DE0du1?Bs-tEWQ@9;3tILL+X1Q<xg3w;#5y(vogET~|6YO}`twS+)
zU$A0Zk?)96_b4vS)c<ZwUi8C@X9iN6s1s-$WjlQIVa!px{k6N<3@#S5_D{a)FRV9>
z;YY<`7NbT|G3IJ^_7WUyv&%c)x)<eULBZN?wM*CJ>TzQb?4b^yB~G8Y)(#`J8F+}J
zSAe?7pS;<C$V9v%Byu+l!i4L+q_%d-Nxd1H#S0^y+&n!6zjs_RnsII#-ucjbhuxN6
zFZajIj-(vs_&RE&o9!h%b7NaT)bw%|6!dFecInDb0m&G@s4TW<nHM^UOtMG#AE2d;
z-W)>qkMS8DDI(vHwTw%Lwl7#(*cC{4aLjVWkS62a;ix2tz|o>T^VcF3!ouctdR!IR
z>4|&Zb-0~!)`)e%o`W*7agzznNbr<WIX7;sq_sj69TZnfCpj9gL?Z*-W$1SzAJ;aN
zx@)zf7Qh!JP_^cFOxPb4k+&~ED_|W&0yFxniOoW693na)UN6cOVy7^B?KHEg$`!-W
zMTKg%k-p+vUfsUgg2grS!@5hOr(3)1GiLgy_dl?@5f|VwQde71-7#^sWPg&w*q)?3
zOX?m$QQ_GB%4gF}o09rwTE8l~w4w%R>8Mzp{hgux^Wxq31|gIT^-fY$+pSDvNvrP2
zZjr81RjK^-xXauu=F`OjY>X$Lwc3FE;V#x{lslk_GWX{WBnU0eU=_KVBv7me(6#wl
z6pAQ?X}{N+misw(R$hW9-7@3I0dZ-1xTCA>xhud7a>_%a5`K3E7M4pEn<#t{wv-Eb
zN6BPPQBj}7iCY&6j^WzMq|0J~jOZF@=G0tcY53ebMLp=}nOfw8`dd5}4^e*;>qXcs
z_4ltODh&J%>^maQZq}oOrbcH@4AfKTD40~24kbTAkm^{YSX%A!JKna`9i``TN}VSE
z`@3qGEPFs?@<Lb3n2MU)aCD@VOMUF|SO(?~jf_hsOP*zQir;-om*94H=-$`$f-Ch_
z?kdWDGPaiagJO?cD4#myj|jX90SJCz6u&cLT7RVt#HrJd;YarLyp85+;@UJIEyy?>
z;*IphL$0^zA5Lwuqki^W2NHHDsd3Km0D}=p{W?sLuw{2fQSJY5?!6=fbeoLxBe@?M
z6=cywWo5m3AIq-}L-WEXiu}KKu<A)_R~(o4RBLVQ){#LIZ9^J_3J=pN0;$ls|9hg3
z)q82dm2g_XN&)3jVuG^h<hEDgyRpTF;>}wD&?4W9`*E*V$x&Px22lH6d>WHbFE?#{
z!*lSm9P5Jj<h&^POjewq$H>NtYUfWb+bgPH3?0)Nt#8#ib~9G4qV(6Kfu}DYudFz|
zaM?|<bMQzePw}rjlIxd2fAP^QBXi<zI;st-cVnobvk!6z`hu$Zcsa&O*O4=Y`Xa5?
zHZ-oA0A^>BW?)O=!`{OK_ycjNMdxrz(bLBX>s_8(%wr;)ut6ds`20x4`YTI|@(5?L
ziFI|aq}(1dmC(?oXq<hOe*vftMUfqeevA#(9fii0>PucvJ|BnL$=ar(Y&5uV<{?a5
zxeLjqdCS7Da|a6RqL?n@4w_6tDup|lDE5&!{GsF^e4wh|8_&=n+16N=MAWZ$`=vr}
zZZ(h}-Pic91pKfem*qr|@BXCz>RT9t%1kkDx4&^MaPzmZ>ZO_Q<7JFAQNck;_tjhC
z%IbCkIeZiGGMWI;=m0DF@)l(e`^5wY=|rzGq9ULhyBe$&I*d(rFY||17DVCuPHC5C
zvKYlk#HHJ!f<cG@JV{)R8FiZ}RZQN4$HQ}ka<Q)Bf}hVqi_dGT#j#Dq%8^5I0k?--
zuie%dPpaO4fHdv(yL5g1czC+7=yRWyw;xzBh_vS4$`u2!2&^9(M+L|y1Rm6+#>57=
zTIXtRP{nHjDEzFr*odbGD}-B*++U4X3im2wVpm@SF39UHm8+~ZX@1b;diVfWcDKWk
zJehUY!aepT@f~*!$~H<dOn>7v(ao-K*~+Cy^J_)T`gYr-!CO9y9xPdctjwHoCNrN_
zEj=!JBdhs-?}LN;(stpBBk@ttuq>qqA}$h4%+vsJLr*QkPg2rf;cDfUbd-j8FQJtS
zFJjYeakm2ut{Ue+wh5+RECE%qNk)pqfAoarFj09?vL_+|r>swH+59@VYuzCWfwipN
z$gY>Zk3Uk7Po;xmsq9*w#bR{2tZrP5>o32kG`mtQwFqP9>)fzp1zQhI8MFgH836@%
zN1*s!n@qXi?1gV0C=~dSdg8_bp&~+w{Kc_~8lf~r$bH14XZSYpD%Be$5#7-)y31JD
z9x<+E791@~Us_UR;bwPSXC|$GYJagYGAX+0Z#x1a2OhQEqZM39CDPe__+ZVnsG#QO
zo}Bru>_AY->k9umOS0<a8da?GRzL;JQ||<jL9901=61hchKX*Bl8H$>Pl{n|x-j<e
zWsAUpBBIVjHwQmDlzDPmx*C3=L=ZgCD2Xx{W96`RRJJjT$()O_2?q2fslEP_!K@`B
zSm`kElTILkTne5ql9%Ss%+J*Oyn)djAyUFlk2}=2dg6z5-X>fEeV2*f1S=#_rztg<
z$X0HX8MUmCHH;C}ZdI<<!_+5$SdWOgD8RlAc$4$F0zwKLdox#z0?is9H!H7=OUDPo
z78e;MteW6ZczO?jI5U6JnJ1fZGI71*Q=yrQ<4)q+UuL-IO<Yt`RW*dL@+A(eafO7-
z^_oFeMS9C?dymQc_)MNRj&xYxZf7UGSvvjdn7)5x*0!N}iEV|Irta#LC5;zg%)Q{q
zxI)o1saGVyd`JZ~p0V35?MQ0CDsD~J?6e!f2gbn0mYcsQ_O#fafJMg^B7=TL8Trb|
zoo+06@63skxa}f=I6e8CF(8z=y9qW|tEfqQ`jJXKMK@05x<_;!FpCj+n`yhHs4bhE
zG=E=lfzD@fJb85I(%4QSvqux`qjjE1*5tRsAj^0TCkRpJv=|z<;j;s8QEHZ;cYEu0
zEk(#k>VTQ^`4Y?FRRQoH1qB3EC88U?E*Gi*V>+$r#(@B^>BX$|L#`K6#c~}TXyWdK
zavBzL;|>2JG}W+g_=5Xukv8(WU%+W0a*(t-dm<^;t#Z!ZAfJ*mcTBpHkFIQH^lR-q
z)Bk{%l9QdPVbagK=mdRw%;LA|(EJM79y>j3`>oS8dBmO5deD74@2rcg<k>v;t?o%y
zG>V)>bhmg`itj$cZyX67I`SB=ag59zRU&mW2O<J28%E;c?kL$D3`l(F$OoM8Tu?^|
zb)3)d$cEK5xklGiWVn93*y0szz7=BS6aX^<={b;Y3z&pZNZYudp&<dwDV`_&=fe7x
zvi)s>PqsLZ5=*@_o3j*BHs7=JcdXCMCj8y!C9$cI7((|%B$$~qB;6L@7Qi4yaX91A
zAWS8gRs~)>x!pMIom*1Yzze&cE);OeHK4<Q%;!%C5hy{n_G6(x$EnK}he>?+473$`
ztKXup4T*+J@P~w0t@7Iw?BtQKr@_NqL~Z`1dyA>SvAm-@np}A!i@Hq`pQ&3-wanbN
zcjxdR$NdWjlgx7;tM-d5TdK70ce_ydJXT-fEv;spS?fac;$x41pr3O;&;GdC;xjsX
z<;)y1mQ&Vd0g~EWM#m#=e=tb&_w7}=TtimcYbGv(BD~-jYA2MK67<N@jYl%Rn6<u|
zHUx@_mPcJbFAE1?(fvxijQRS}eqVOfD#O>aNsF_RXnx*r)HmLfrSuZm12-*dqzIeM
zrVlbVS4bcdlUeq1FpgR+*804JbjVF&c|{(~0}ldZzUP4MWg*0o(5{SVE9_DLDX2>*
zC<^`4U;6qR$jXzapS%QP^Y1hUSiMN<`PTfnrv?w3#f@PVfksSF7klf_MvzTm&Q0+_
zBAnat*v9>u_*~sZ^kdoqN|EhVW-D!2V!w8O2}<fMc>VeYS4Bzd`N{c(rYFYQ<?>f<
zbau<0pSCWqUV6)UZoRB3y<8HTxU0g@=I5sYBqgtk8kjgeVg4Xx-V?bpkv62naK&89
zvg4O=5o2OLab`yL+eK0!$H7QGC8Bv)xuD^xqGF_|s-IDlzq>kBRz_E(vxbyzAf-mr
z!EUZ#F*X;=4OK@n$u_420>=P#uoiF}0DVYO`)pjgrF{bJQAK3F4rhS5flZciy_ymZ
z8g&&riQDS9gkJBJF^_H;F3$3WQ&=woY;qoyW~6dZekF>O0F1|aCujL*pU-1N!n~!;
zjK!jE6W_;`U$^Z~mAGoIi3tw8ac)|p)jfRb+s;4Vas(w9ue)|pc2;=qO*WHtZq{b$
zbq8f)etVX^OXf28Vfk+e`kz+Kff+2CRY~tDo*tGfiS4c~eXPtwfz98p?xvdG{o4VB
z7+?euEz^e?<_u?yL`#^M6GdcbOtOUU+me+J?e09NT!Rqw0tit_$y{%4lhe@|;1ERS
zcYZIrS7ectgiZHt*P^8eoZ0WJ*+xP`!e+Cp%nzsQ9eo*#NeCD2UTSPMGHb%u-XB}S
z|E#6e<+;3Ct8U_h@Jo1eX!{{K1WY^fPfYy}B-z+ExQaC6sv9Vfb6AGP`MCwEK`Eu{
zEqpnLv)A1d0#en_zLPcH+A+k4Gs=lxn29?}8Yyg@euLy|m*;qFo+V(9G3vB9BY}pt
zLANvys?KZuJ>_x>{a#n1|BcG__H=gJyT54X)RmsuT>mBf>e#x9YMY+?E=#q-SB9?h
zmN&JLiwDv$6qzLNOSK=>OpcC9)|6!;cRsgc6WUz-(1eknSJt@c6h9R<VEF}uK8=`5
z(_U7_t=10yK|dwPn&JILp$OZ(g>m76Z0lPFq>yRsjlUIlC6|JEQ^{qdnlJbO+i=@P
zA!or=6bEq5Sd`(zXe9NqMoH?giF**`m9HD{?+M|>cM%b3Y$Y>hZ&Iah&A){Dm%zn`
zJIYK86c_q*uT1~q1Q`K$i)Y7~Z;^KIB(>2B0U84g1>Fz=L#eyR=GHe?m1n8H0qB5#
z40;H@#?IC&j5sRQ8ZQg}Wu9on-zzpYeqX?Z36)66YC7TCA@3t*b%v>22}$?omCJ)Y
z9?$g8?0<mL<GROi=+w6xo8<eCv+H<Ig>`%!|1wtWVXR!!ukDh<W8?CUiFqiDW`55d
z^ZNy9u`*BO<_Fg^w%t6GVX`thbmis)5%opcd+undOsv`*gdzmk79kN*S_-$+zzeN*
z61SUb@iOlTtz}yk9n6o&g0!(7BP2I*RG^(KK@S)f7YM7~cWD|bHra<sqVa??22L1)
z7gn@OdVQ(j%OCxxhD}}5yY3Ch$|i=A-&!za)%HsPY{(Tu)&fA!+RWvfN%BTmsmL6K
zUMU6yal)TD&EaIDoKd{fGLZS`ewXxoYucB{JC4gEgrJ+?zVG~GClS-xGQ6GIMkc^?
z?lU>k*v8RF>dt&xFPH(m<QsMpX0uY{$j99Yfss#>hwVE#BKbTg&V2QKXHwr`C4a1m
z@2hVCyH0jR{$`y;mX%lD*XG4q?sQ4S#!F7Oy~QZB@~Zujz>CWF)YxP)6T~<fP^vPn
zHY1UIO;uFZ+n5Caumo2ECLomYjFLpE8yv?zt81!ExS(`7(S<97lF2=?%rVw86Zz7f
zW4ASc2YDHCo!;uEt}AE~TxbPEo2YZx*nmPWMql6rMWcYMQS0bK0XV@Q(CJZ;RY1uL
zouS%}nltEF4^I<Zpxas~vSEE#q;RhV3k;(_+(yYAfU*cAoq$2HibiT~l3PIU##R;y
z93777$N6-=xf#yWVU@^pls2g1Cf085eEegsde^b*m6TeLJ>c;k;Y=}2YqP1QtGRN5
z*pB*~XXB=hI|o&ooQ@{fOvw2p`nZpjoGPmMdakL4CbMJs(f`4t6!IL#YJUk&S&9#C
z=y*$9KLJ11L{Wc>eW^<<D}w69rk5lA56+kX^Ky?fGDJ2_jvIshB|gUqsci>?gDg!a
zIW|6%uSBbi06HF;pok|3lJDLzpw>V=#uk|s9mAh0#unQp-7XWLnQ^4W8Q^#2B^Pp5
zQ8sb<(2hd$md)sw%X5@K`@?!yz376`&6r$Z_5ciDjXy<&#N)=Hkdo<9xV2q0Em6RD
zWy@KX(_MwHyk$bsy$hru$v02~lvfY8msHG650Cz8+fM+fNQ7`XMF*fjl8s#4ittRx
z&V*4qh)5@#IZmp%DUx{-`&TW&_h=LX%q%Cf6#@_{GwQ4x_ByzhM)t}e`;W1)Q_eQu
z`C)uM|CyfE_}VA`|4<eWbx#Yu;JTITonCjyT?(++tD=%~aOItIlWxAV+sfz_IEWlF
zkq%dGoqzxI)#opr3)eF<w?$C@q%><lw>tq>O!&JLe&huGhHR1^?bHNZ60<)5j48kk
z`WC)7Zmc6(CKRmMSl&!oW%moTG%4sRxgm8`VR!Kn0&@lsj~Zt=W$AUp<x4mTDv#q@
zzH5j|?%C#Vj=w6kL>!2r5U8(a+7(kOCZgvK9w{1s;F02c&>v)rT3_MNn@4+G@afxg
zPwyL>(vZq7QGxCQ7zo}NBF}`4qcHu9w)(7?Pis?1hKjQwE5Hh4o_XMpXeMWzi<0p4
zw6&e}b~Mk{5w9MfS2}*X(jL*k?)sdlruvd*UYN?PI=g{)X~&OEhj$4qrLxL2J^W8+
z{fqBT%ZUu%6VfxE6}=7X&(%)n9jWMwj<Vt~P>7h8;7#%U@U$x<g`TQx!c2XpFm?Qn
ztc1bP@#6>3bFSqZ$^Cr(BdHX72q_qdH>*47PYtho%LG(Ry=NQU=4BB%)Q-zw3f8t0
zz%Ljg*wh#;S)E4#pY0?b7I8dJSYJ!tj?~SF4hbjHZA$$I(wxryd*n`y_mb)?Q8t;^
z-x3A@?^X`WxNDEOlUjmp%!nP`4(nYJ3B^Pm<Gdk&ka*MYPTzwX^nntiBiU$@gat;s
zF|3)Kwq*~P-O3SG32ekLxueX7(&;4@TC5!nNzfiv+oYt2I|SLrA>jmYfF$isE{7?(
z$+Pc*k)R>b>H<gSj_sDC51Z>A7Ll43j354-DVGCp!LsRba$oZ}o7KOH^&!J`f@OnA
zR_@)d#s&&g`*!lC$iBa$l+$#qsbK#<_fvRHwL#~=VnZJKa%n>RQA#Q*&&jO&kz`;}
zE;<UHr1`B6kG3ZQEYPewgUlB=2x%2iABE&D5Kai>6LmbQHw7>;;0%rqzVpqPnx`y3
zNfMd#Me0q)-!85uiPgH0_h}xN4B<u31PMEU+OtPRb|2RFLtBd9<}>sI0>5}o93n?X
zqe(~M00EIiB1yXfILR`}0dEn&7=A3)tFd(YtwS!oGd*(MnHOrkLf<*9@J7HMG6lP=
zbuSWnqQ8I;^5N`LR)d%hq`3(DNH?zTUelJTo*I%XR%^i^3@2Cov^r5Tk50*n6BLb0
zq)Qcc9qM`ORRyYyvvxGbF6>A+MRKr-Oec3_75{g<tv1p<Ej^PJw)lf|VlbVD*WGIi
zRpO>wf}-_KO}CG;2lT6jhpwhtzhaAu%F`^_7vS=u)PTzb{3XZ%fAP`zQmLqT#DF>^
zpxN{t#9Oj}512iFz&@<yQz!FMu89o>{Ipl&-U^-olWx8D54TdoOPi_C+hARUzdU;Y
zg+ZX);w)8dBf&0_;dv6kN|!XYFY<ZNGKXA_8F!-1gaCmc9$;ezl~aV?bX&r^-2H*&
zJ$W`Gg&yjY+KosfP@|(*UuKZ3wPOrCNboY2dsdhx_P@QO_B+*nsiV#-iUx}M^V)GG
zQhLNh=Ns&!xs3;(V0Wra{AE`bq&1NZvqGX@pmBZY$nZF>pA}xdvy)68mpKA~O$3%?
zn3YuA;gEaf62Fa0Ao4E;hNTBF9zO#HBkUcnE)}MZB|KAPE!y%XqdhDrD9L2QYK$a;
z_X(wx$w?_B)xdf_S~JPCvoL=s5Uphas|ffi9ts>u(7|xniN3}k??41t>>x_1i8+1K
zWYe#TRx#gl=a&{LYq}~V(1t=OgrqC<6cuFX$+J#Gnz#YX))<?igs?_iG>5?H)ZQ>D
zw;l<EzYYJ#i*W+z)?k`1r=s7EQCC|;e(HVa0$sja64JQ8(lb{OSC?V$NtER{<(MkK
zu?8~~ePU=S;TL2_sU?~|_NuvPSfR_|ZEeNfwmi>*UtnrgJ0;@OTI2%E5#in;{jbg!
z{vs`s>}+h@lR7qT3cOgJ)#xBsnByMW{qbL=WNc@**Tt^K;kz9+88wqTxWU`Hr{rsy
zRq$|ldYQkqE0wi#@mv073AI4q%d!ZUE$a6#Y*77k)nAb_;Q@F4L~H(9QC2LiaL5X*
z3hfH?ir)5)Cte*xQ_;cMIc-JF3HB5*%~adj6eg3&(lKXeZpKdxqO~&46#gFt8itLh
z@uk9}3uf0(;M3s%_8;W<x_<{CDL9?3HNrh^R~m34V`L1SldO)ah9dn9N{vr9mE|XQ
zc{hy|Jr-Mssx<!s5Thm?O#1@I+ZINY0=RyRm1bbv?EfMyE-xW@)IKa!HLs|sIHzi$
z1aebt7qjS?ZOl%Bg*hCEn?Gn<`1M0!+A$B4oW^APZM@^TeAH&xKRQjMFH^k!Y3i+t
z(%q3TpX=d0u)VA(`=I>u!r-9L_<*hl-+$MBtGFBgo_%^v+zo;@h9HP#f%lM<s4y$D
zeAwkG0iq>B0QY#Rtyi=--zA7Eat@-wL2<ewEl^%s<}nhN(9-WpMS?;?j9mkrMOOqs
zf~x+37H3>DG%ZsfXugS8Weyar<*n^5qqzEkQomNjw7$T`)`%{^Eor)KDCP8SZB`25
zWzake#Q+6)LN<Zq00<p0Ri;&B<!458^|^yh<u%4R%;w(eLPY@boh^_+v&qH;A2(1X
z<}6?WCe1+(sli@HM@&*1TetjM(-5mv;ygpO$4w3^igh=cM7*Wp!%gag#6eei(bajX
z@FHm>H^Hf4+~3Xpr$BLOK&(Nw)$oBoGzB8pcPW#<W**k@=3xP?pyKGyT*ubY*Wr%7
z9kD?{TQrFj%6(;Ptkvv;)OnWNgv-tnN65XYw&a+*0=y}`ADlTUhkDR23Qmvc0D5+;
zdBH@_{9(FgoC8c0kb}YQOIGkI|GVc+gijclLb8b3#*O~b(kX|$@?4{h4Fq2<lez@5
z4McnYwrqV}?)m$6*{Hje*ClRryR{|YE_r3`fX1Q^WZe#2ge^R{z9%0e;eCS#ND#iB
zexjcvRhU25{Qj=rEZ6qCOI|cvY#g2+P1BsZ^;A_v%_AxB<?%b$D-AfCr-|wDW%?q1
zfkrLjDE<)zKD_o~w#QuX>g&+>DA}Yg^P=riDi{)wM;lqh>auc>_)Z|X!Cx*$Swgjw
ze`OD#Gi4EfNi`#OrrzONytUKM)5Wq3@7@Zx<e1+1b0z(sG^)L#+RElzr-!cD-r)rR
z(Pz*0H+*}RV7ASBY7G#%?O(aRNzt!FeMS8Mn0Rj?`2kW3tNXTF#^fqpo*S>65AA|H
z%n8XYccJ4Zx6NrJIq2tpa1qe3T1B1Sm`P5VAMHAf-U~8H3NhmM&k*h$XO3)z9KUTg
zb9ig$b2&K1wVS*fT@II`&)a@wJ-yYRIU5&oZclA4vKU-rOT%v4(C0b*1eMfl)S_5&
zEH)x)<h^A`9aK47bd;_S*}yKwX8(!V*!<B_yPgMgwF}<s%r2!IZIiBdVw*1z(`++`
zRg7dg!E8c&!0pBM4T%}@W5ia9wmH@x@8U|*^zjk%MFy41)Y-Vb^ef?-f`nL_<2wl4
zXd-d$>Ao_$CHfPA-(}n;_eM4aC*~Ri-V_CMtm1e7_x=f8pF}HU{1ZX$Q?&mRco*!u
zHb+N&eyyr<`N$T@vxkqW-u4Zt7vx-v<itPf*u!QWV;i%SxRpjKD)q5gc9oc38B_<+
zHGncGtvwj^23{0l*6*sidL*fLKU>sJad9IyqS1xY1}do{f-R~sHmJ*}E-vlB`_oHX
z6+O40AXnd&_bxbse81^<F2)22zS4GZ_@zOYM55~-nTGk4@Z7If<?qHy`a@)(d2y~Z
zmgeTLu4f6CQ4IG}xn8HFR!MD&egKA%wHjABcs#Tx=##!k8VF|@?1fIy>WH=;>O~3x
z-q7c;!Mj>Qm|<OQ?vb_|IwZ^(J?a>kB*+Rt_oSxy)Lld#A+cg+rY>kXY4A~P_J}pd
z2O_}GwdQX&@V-_i&Tfz$={{BZ2MhhDcX@Rz)1PR94_~?LHvHtI+^hvi6G5$49TVH@
ze(G5tFrz8Y@~vv<d?6x>r8H&er0~B`|D#AJMzxUi11e~;TU|eH<bQzX1CUg$&>WWQ
zUoRG5#PRQuKkQ)&Uk&!U*SV>#COTLrcDq4^5}O>o2=|gCa=sOV>6iR>{n>(dX6K?q
zj8dBHGLL7Q$D>Dq^pb+hG!ase$Ror;c!AbP(7TqJGRWEQ-8xxfcVL%x;~D=L{yXN0
z^G6>(5Hk+XYWZtgu)JnLn`p@&`loBhahSoqp!EWq0rul^u?9WNdZYq)aB&8lkUN@r
zw7f~>{-G9U6eS>u3WWgX04Tw{>SbMOK#EY3g)LHaG#_KNMxqAv3rks>jE~A}o|y4I
zrB=ANzh#f!a@#X8HZZcRkJ`gk4igY^=MJu@kpkcSa2CK9eT8`?v%)5uG8Se|`Jv<`
z&a8nezGePS^Mec^Eq;l-UAKGPmaAKC>VRs~6w>vVuc3-cM+Svedi@}Zu>Z%`n}Acf
zuK(jigCeDrN|a13nna;wNJWMWYnciat3oQGB4bigA!8|&5X)2wnUzWiiM2HtDnv^q
zgjj~(=Y3aupL71-bN=UA*SYp}_StRT^}P3U-=FCQqy=q0gz4_Bag<?x%wcL&Ym<ds
z7$xvhlY6Br;Gf{MW4NcN^aNFL)AT+;Aq;Pah<My{T%~>6F~Uxd?wfy4*QXoI&w75K
z3ouO++UEGeV-ffJQ7q>10-}YE4>N#FlKfPH<fXBKmdzn{ctOqufB0?|7){qGRQ0D#
zD;roE)#W2am5;hplLN~M=r9l_lsc2NT&qh+e}!b9Q=mx}8qrgbQ=$U~QB}iNgpE#e
zpE`j>U{ecay*Z7ihM&@ycCERXZi(iNRq_i30~a?Se~h1)v&*o``xn7JC+5=@1}@!|
zVI#QPHgI3|o59S7ov#g)J%+vsr2a1`G3Ar}$4yTAsK>UPG1JnfF36`vhJMH%427P-
z&rMo~E@wGx&Tx9};?*N9G+)We5=7OV^A}MhfdgQSg)u2n4M19W^)DtP<8MyhT|a7f
zn2=F)<+Rh9^@Zm%83%eLbwU0SEL+uCbi9<GRov=At`mnuF3gtK^HU~Oku(9CRu!&S
zbR?KSgR7YPAMP8`y={4}1}0LC2fWw8G+0J&iOBFPV$`Fs=Ex6i<XeXZ#Gnykuy9FD
z<AtCrttpRNkD<*=Z2VeIh8+MJ!nu92|Na83Ox)|-IJlMOF#5nx;r<qvJ8LU>67yH!
z&bEb~XNNK5c7)RvAJsTGorm(P%jV;`N`zl>0z8(HXd0I>)}$n2+}gBpmGQq#yv`RQ
zM#`@4iW)h#<;t#7%QiM{wd`+?on+fboqhbLhCY=+Q3<xMQeppI=o3wzntTcl3bHW_
zNyDiK3*+s@8JitX3$1{i;oj143eouz69vKz{049<Y1X7*M!;||5=c&>{-`k+_~Ac*
zuzy8V--2Hv1<?_H_xp$hPC>KEYabt+ze3kyjDM%fq+K}hc)^iqqki142fnbL5q#;#
zw=SY@^GY@zS;LG1x0TK`X^s_o1ZOF2_HlCBpc$XCHm8|I@2(oym*Vh)_f}Uy{Fbkm
zefPKpGkjOUs>JF*g@`jed{0P~D6_^xg@TW7-m;2QrZjsNzwnI5>#HCB^L4GV?$Fu5
zKML)CiY%+@-XcSvcWNV3J#nx4?-ELQ)%60(m#z=s-~lnT)48XadoC)>$U1lFatvtt
zpo!lu;N|8FvLtgBZ<QAp_gGtdPsHsU{oi37Y1BQ@)ejVo3^#T3YEiVhJ+hwO)tm>@
z0ntdHodCSNo9$|M<VyT@n{^}U@-n(0_Gx~Wl-1@Hl{j8~dO9lggK_K{NwX!XSr<a+
z_X`VuWs@*VU`+C~7#!In@+;{EjZ~UFlHHmy<N_yvL?zF;kM}11-GNbdjHX-s-ubO*
zlWH%UnPr(U*!h%YGp<4H=KgI3@pd+vm6V3_rQ!CZId?9-8sFXCod6w7glSou{~xJ;
z@zz`8&>IwiMpo~zTqb06Oh?%=k}J5nd-)!kiH~9f4i@_!w{p~m!YO_@B{nPPPe3MO
z{9~QB+~y0=*OcxdNmtx8Xg&zMSuSF>e`)=Qyab$Ix8Yj?Cki+l7YL3Jgv2;SB!?;3
z+H^b|8ikatH){#;g6Ldm)nTWb?QK?ZOS?X~R~7IG9>>-_6Kc92ZO_#RnHYxik`W^`
zca<6iXfE)_;u6xThCT(FCd*MD_YTJ%3$5zdl<$x9e#ot~v!rX;j1ahBb!nga(!W4c
z8|(~&7|Z*HLo6i~i6X@PjgA9`CESKMrWQARQR=m|@9B=&aiuzK0@uXa#N17Q((Q!z
zFv7TA?9qSm*Znq7%U+W^OXjlBp01GGO1ae}|B<5D)SLOkeL{6AXsVt)<=JycF_>39
zN8Xfq-$Lmk*kj#!M@np9{$|i$NgeGV_8J5@#PTQ1pHF+KVWE@+H6?@=K%>b{QEMJ=
zP-0Jc=ENrAZFo)gdGI8o`$MZYwC#zX$oBwVuf2r!+tWyZI9ug8RFmkk64~IUiB1<G
zx9IMu6|R=inC=-8In-vLe~dQ|JCrH=_f4FQ0cQhUHg2E~d1V#`;}3Q}(EdeN;M<RG
z5&9EI6DQ(8Pgys^kl56&OuT_{i}<8#4qHai7_0U6--z!k8}Ws8tnJmEX$OYB%_qwL
z-y*hQPKPt0HgH*NMx@-bgHm*fM4l~~K?*;6$Iq?oS}Y3>&Gby=5IEz4j_i?a!J7yB
zBU}rdh!Y$SXdH281wdLKnre2`0Qa(6>#=;+U!7I%`1bjUwJqc<ieMqa)eCm)SpL}_
z={MKjC>jW?uPMhWEQ;T)?$Nq-+t?Hq+vGsOnD@$0he<U&BS$LWZJx8bil$!P8d`$8
zwK>lx?MI-M2`Sj%Ry^aOZIXT2(EQdYAiTnn(D!T0^h{E~Hl<y<Cu9=6Os&0aNFAN>
zu1ud*azjqWjnKO^G+kV*_W7e>il;(F)mfiki*?%``(5}vDJVW?wK<_f$kyRG;plM>
zs%$}M)2edxX*J=F7NSA$p7<1I0a)Qk(#iUJ=NM35fl2ciO&JMQcN1&cj2HGM*6_*S
zl8^WnyS2b@(s@9P&}3W4qgz7WfvR}7<RZ^GLbEqrxj3goc*h;e5#6xIOZR1Ey)?C+
z{n#@?%0FT0%dBJU$=p;?!Js4Rh~4O8v%GOSAf!#NujvZ+JxTncydQs98D#Z#zVP`L
z(=RY*YF9V^R6Xm<YXc8j^4QN!tkcZ@keTht^mKKus7gFP?B-%murG<olAYQ7wsZCP
zqQgJ8?Q-|9=5qA2oB`!s&j;P<T6_(b(Sd+T@d04<{Zx9i=}Zx=RT7Nw;0zOMr4Tzz
z1Mpv{ZHj^hk5rh=ng;(?+bYJ?rnzHSG~~C$Z?+IRm@JvDTlYNGRzEpt-Auv}UZ-gG
z@eAgch^3ux{_3{hauKMrC!MH3Kr9*+#eXd(4snCSnT+5Zy5)W6y{#f9lbhht_|&t#
zy8Hw06V^K?qkp?`t;rj=@09xOjJ9L;Y*?QAUyfV~I?{)ie9cd12%>|gMt<80l0@jU
zX+LHp&{vF3`Zd|LuaZc-sAyoyy*w@DsO2J%#Ux)W0E$!io8`)|pdQ1a@kg?=a|bR`
zO+q?EkO`t2&XYI5hTc^+m<!HyMwsiu9b1GQF0pnuD!NZfaNF~3TD0TA$!*UDMH9!u
zcApV40nZBT4HpF9GyqBK`G>j5cl2?NJg|1cW3)(aweLmWWnSe|eW6pmuWQfHv-f7_
z#H-EuBhUJCTtj4TdpUmUjA=dT8#<mM&?+b!^S)$@S6VnEtR*>(S7dq2`^R7i<mdCI
z%I(+9g*ol?x%r}0AF&&jb6aYaV<>5&Evr<QGs{k^L}xWS!R&+&5H);IW>KwBV+<PO
ztq<TAx2WnjeHzMS`~nG#%1(|I@-=f?-XE8lG8`4H+<-|DyA`8`(J6Od(>7S?Tcogm
z&4T?>kM>P0lS#}spMX<X0AmAwaZ(XP#OmR%(W~NZ4Y3=h6Lfc+Y(c3L^zNwsry@42
z_q^fT>#GbRq1EZa6IS!(KjXr3QuTLy&x+=be3#xePpjIedivf6osXztpZ;y4b3<(}
z-q=jTz6-~DVkQZe(e9;RL<@ttRXjcAm4Dhg{2<YEwN%x}l+~#4o71FsYFB2D<;SP2
z#@7ur9M&l+7sQZC8=kp{x{t4rP~E|-&KPiy?`Fj2P^lq_J*jo~k{AOJJV25I2<8C2
z(&cxl##O0oYrZw&jqPh1HUQLyp$+L~WJ%`73_@aQ`=NXfu~Vud$0EaB?imPSB&25+
zht!|f{LA%#S5`QA?7r>L*<>aMo!)i!=kv>A?(6ww)Y+uO$>Estl6>wiq1R%y%hfjp
zvV2CnR3}OS4}7YV7gT2lY~Or$K6GFQLRsPOOe`%`#gm>Yc*kY+f^$HBf`KSUo-6;!
zGjJYnC4(h8J?f;2FBx7*aR7vf;Q=xLbc>EEbiPmAe%&4!sruS722ZP)jv3$+?+<LC
znw0LL>;zK+e%mG+oFPgXGlwzIR_?Fis8H&2iB+j+fXj|>y8W@K%;F+E&>(lk114ZX
z#?8Hbj7|dbK(=8Z9Adt%nfUOrqTxcMzialAF3(}H{{v*$3C6z-OL2X%^yjlk5u#+N
z&J!5&wDXcHJwQn2Bs4?OxaHa|absN2l6<Uhdq`vk#wNX_Lc+=n2Mk6NbchJIsm^(J
zs>s_YEfzqkyt~+im5?#7UAK;A^bQv9OL3LcB~<CIVpeK<D8X8B+0wd+b$dg<o>|@9
zs!(8{RgDKk=3gp@ROrfsN--xvQp{}x6^uFO;X45KogP_jCp(n4xtS8EP0v+_*co^?
zl;7begonHqx6nsl-?7l`n>%PN7s`hl&CScMZ2tpN{v#NaO|ViPrH`~s-7P!5>irM%
zkvXgS&d4WSsUES=1tK@h;?U@JI=Oj{^R%Ongj+U16<VWk!68AjQ4vmF+^~d21E&!0
z9%vSfP$3<+VLQ*Yiq?}#Yz(I(?k6|4iIAh-&?o!}A}v|G<+B5{wNK;t&?2Q$<^9v_
zU@(HvNO+`TJ%R$!-e4+5S3~ZIt~hhFK>;1?@w*O7<!iK{_|57CM}gyxp*M3@u1&UJ
zt3EL(sdB*``vICZGjAawldRzbHOi~JTc=ch_rEqcgX8|wn>9jV{8!rsxNg%~GVK#O
zt#@LTd`{_%P7?m{;V&^yF2ec^zfMcqt&Cn(5KW}v(s&_vx$$%jn+W;PF%b0QzZ&-t
zl2Ndk++3RT$`uW8Q$hHQ58k~WdlMH(&>|xwjSp<8bm2IY&^kf=&2?XB`uw;@>;Nm*
z@mCP#a=9L*Mbe+{kETP9g(rXs8}BveZOU_|IVbS=;xGulSuA;H!G}EfK^b4l0RF@o
zBJ#)0TAeXsLp+ADLc~!ZvPFpw_G~o|KSjN{zKzSSrb3cWur*Rmt}H#+Rn1lfwchkZ
zUu%PySAXKJ<@^;@^^^TKJ#F@eP;QMM#g(=MMrqHdC9Xqz5*|g4l*#w`vLz>o^o;Jp
zvVD2ySHYH4_#k*-5O+4F4Z>@($9HT2VG%%=1obj{Fhv6bn?lCWhfPn{1BhV6DvWGs
zaEU1EcYImrA_}$&qwe1pGr<-UiNvvMRm)v#a=!JLC2>^t=!r2F!!J(`Q>b@vx{4NF
z!l-I5sVy;~ndeLZ`K0L`^VTP@3aBh7L4<=rjwk{$T7-Jw4nxNLc)7LjY0GGS*@VYJ
z8kRnLcixH5j{SGT*T!4LQ_kO>ELc{U5LI-~t1~aa$w+9v_l_Zrbcdpb&UK}HE$jn)
z0%0n_GO_Mk|KXBlYg`YfeRJG_`AjTsy1^b)e;|iwmf^i2oD#$%0Thm@f(%dwbtq6j
z2(uQ>yXKB3w-p+`qM5<CsBk0Ie+oSbnmlN^n%xCkwIvd*!cM+nuk5m$ql*^c`>8n<
z>xt}C$Z~GhLgP)%_2#>VYprC_hdd(e96S+xKln2+U*Q8C#*S-N!ol7m#FMCmHySz$
zGMS+1?HyhueLzz1Gm{qCJHklgycQuT+lqVU&pf>qc4m?i`Z3keo_{U-e1v}`Pj%>p
z$TF(#jS>rOw;V;TrN>OG-hVf5pYRInZa5}*-`#W2YYwMQ;-wFl^f0<+I8FS`yAwxg
zEE@!XKw^RW7`HOaHpFuQ^-9vY_}{^ib=}9i9SfwvWcss9J&!Ocq7B3u0=W*mCG<+=
zE}~a<-i=EheA}R>eUIT8LKUEhe^7BG^E!+d!i<5&|8Jtxc=G0@pj00e7|kw}C;Uyq
zieM&pB$gODfpu1H$dR|Roq=c9a4mdyHOUW?hN>(dv+sA;^9M{c*Z;~t5uz(3hAff6
zHN->O{pTUgO(&ue?kzGpc1(Hn$1Ypv9hRf75Lu~RFXeA<P$8CBW&6JF>nQzaBSD>!
zF#}39OU_tu?^0c6zf(~S`QW&@0Kwsd><omeh+M#t1utmX&7?mFPTc?Te2@0m!VpyQ
z9bFvd7w}Jb<uDZ&uHE8ssVAxGe3sod+i(^qRR<C`)EX2WrL?QK3f4r@A&tbSM7Y49
z%8;((%I78)RIiwIO1hAMrW;(Y&8}v;WI9T9IX+qCS?ZhMtLnC3nBk=<+~pZ^;=kOB
zr50))WtqQCr081*HcBMkbzf}gQ+s5<f&C{G!tv$yL)&ZgkBmaXRw+7bbWx5ipf<t|
z0^%1sFnmb*7vFZdbI?1KvmFIXsda<5K5b82Ai2EIA@(5Nb-co`OOfc#E`y`@o!PGq
zpukK;=bF4TO+K3tfkay!EB)GeYp?WtC~6UA5LKNh+cdfgzmrpV7-l<w_1hs`&GC(f
zi%&mOWCwZn^|0Sn-FkU5X22oZk=PfA0h!2qjMi`XO98%Hm1$Q0*5Cipkg(5o`fN(_
z{ysPKY{t2Px(}f<h>oC8KzjSxq3SALGD@gBoK+$;fPT_l;bq1V%dAULAH*1^3UM)V
z%pAwii;`G22h&ALkaC^je#=AgJc4qT3s4K~j)FK|8w2$M026~#M4l2>4wa&Np%|E=
zX5`4<=^d*Zbqtr>b#kuM0-og99NtLW^PhlQB{owLGBIzJ_Y#kIJ4|Y`r#Jy1NPB&p
zOZ?iutdbsI14QsU<ch;3Xikn`HqRe*IUjkPr`k3<abKDFuK#QZBGTetofwE0pYUEb
z`r?Eu&7;q0Xlme!EhEWG^sWxrvK61~OU#bDRDS&@^+7ZGKsf!O)LWM_)i~asBQc;u
zEaaJi0F)uD;20r*FN5|Pc_8o9EiG&!P=OGsi{st%8vJohkj`Vkr|C?rsy3?(2f?5m
z!tB2u)U!}mxXV$Beq$f9cOYi0xolSNaMQ4D83YkP6KPS6iu6El!u2%5K9AaAyb$tb
zT$<<$&_1E<Z~PkByqb6lyIv{QB1Z$*P37wmSJ|6{bZ9v7SqxA0P?1S0r||uGr_Iy~
z3O;ofiF=#NF_;mwFTW=@=3v~3ODbYjWyFn}c~#&wS=!~JV@Z)SFN;oSTz*4sEsqOu
zr7*}wcad(S`|`a{I?#9GsIuOsnPnWhIJe>JwGL;pP=PMM<p7MAL%pRlgZ&jHR2^Jx
zW`CP;j}cMMh>E`HY3N?os*q-+1+BqBw0gl3AW)?%h&q44@izeGbf*YlXn3DFDlcGY
z?~cAbP5i2E_KF4{ECr~|Lo>PQqpu{E)sOsL#M(2pN9!rog3Q6F|2xa7ohe~em0Qae
zgk|T-1g}!x<Y;n^8F{HE$l-VQ^d%xjK6+*RKE$t{_TIxtKvwU0;nw;ki6y%zzhRxg
zVFy|8lnCg=i|Szc(kJsm6#Xxi6QaUXK|S;zQ?2OZ@zbK3BNl)l|3P5m5;!4xQLlS`
zgKl`0X!&>6!kpsN41IV=R4=M+0+7OrLi>YnlC%;j>AW17!BNY@&I=+ReLoU+gbjfM
z;;Y8Y2o+n#U*x-9OUt~qucYeCfuZ8i38DHCQi=2Qw48CGtoV!))+pr<9`(;%hu)9W
zjFIpAqkAWEjE<|kH&>{X|E{GivtwGdXJ*?M!-{nQf+RW|wEdjefaSCrCM(`Ddimqr
zA&G}C(z&}xm<bu|;+aQfXz3*HvPJJ8_YpUyy2Sg09C`ROFu`ItAgm}J9*98Wf@cGE
z(?jhEvoy8cGCGTCb$8&$=q&_zn$?R5ILX2DquCEzycvL?lZyAPW!C0FvBC`qSY<0w
z?-clXP+L%8a85Ay@o(?+m<vok3cE6C7B?2O5pzLoI><YwM!NSKlV`ieB$PeI=stV*
z0iSB`T6??rY`krm`TtTb_>x1x*|{?2QK7t%;sUGNrw&<;%IB+=tOc{k$5VZAh*Zp{
zFC7ls?e?Z&#maqLyO#X^Eykq0G%2Yc@pe;<iwny^evf#yNh``w-A$GO!|de8kf5X-
z&)5{gZjcBeE@E=g?BO?w?<UlKGMaUF6%8QmhRV=jUPQ;IQ$=vvvnDIG>K5FU1yBAI
zggKN7UfW=&quj6c*C6~-ZxV1dV$m7nm-r6MrRV9Em<C@fIv)<-gcoPNQUu|4g>WRj
zzk>C4<%#B*6+Wz(QRlXnumL;nuCbxr|37muF8Sj^hHEML%R>Vr{bht|iO?{=fdg^>
zNDHXx!_}?^3OrV-3+(q_qD7vjZLb#McgQCqNr(@Tj3)?8VAhqFcIkIcST{5K0(j17
zOGdo_*AG<p?~pa)%0=~qB08%QBUW}J?h`^FLe*1s08kDl0lW=>8FkN8YP1PLAj`33
z!*7grXV^k<yLNCFp2QrwB`hs)QB9)|z(+@1a1|_6;B%0RvVSAzSzK9Oef|2CmDSaI
z-pD*TSsL@Mv+B;xp9Vi>M)r>G&HgvIc5Ou!yZwb&vB8kNQqP^%iK&=lTjFsUl*PKn
z*{ssmUiZ{bLCFBT+(ymGGTGTAa*n|UdJdh@`8n;xNH-xV-}u$-qet7_nqy9}c?)K5
z=I1GQruEv*X3lHKuL0Egs)@i$I7Jb}8KK6X$2GqpEsUjxnoUrv)=2#{OLuqcHWeBU
zPMme&0F+ru4s>+>Bw@CqesKHdsGuN24W+Qp${|)=^rBV?T_OB5Fp2M5i-^;@Y7?S7
zHf<f*Ud3c;DjAqgG-jGPbx(b*nOs!2LH8-9<ma;^Wu2BAW&M9T;eV=w@2?SO`}$9v
zmGSH}G*46Q{mEep>8lqQcWk(=Zy*&nM~J#Mcp4(A2)okLuz^)c8+S5fYiivgzPLA|
z7msb9)SaJ-220kjl;AS*S5(=C;ZTVW8Mokw(h^QCQA02gzVkfiRrM!HY8D!A#+Ox<
zuk4`R{;Yf<tC#TTtc@Wua|tak5g*ab!Y!ZDx%AoKR%mVDDkI(1fSCZALlT9*c%!Js
z;3Ix{?R(lU=HfT+_7gIial6!ZLd?6hte&s}u5(@XWvtureyaKJKC}Otwj#hMo1jHB
zyT_NBmf7-;HWc?aYgM1=?AF%4iVQ||Rr&V&gsVDCYI9%xxwSPypcZE<>xSQF)v5^@
zZ6bD$0zRiT>|u=R9@(jSD9O=%Xi7?dmH$sW{=eom)Ks~g!B<v~qyc~QyD&;t-@uAV
znjH}CLpaatuf!Qe{*kjnEvEFf_wCEGS}?31-Ndmi5r1(kALcYDo^;oUE|dfv-E=G;
z!Z{f1hz|fH*d$;`tlYo_WAgB60u5BUaz%W%(WuXm{Ytwlzy4%O9+tCI8%=fjFOxAW
zFfXM0W`@uCW@eJrc53S)lTusH8+@`7j*<fTi|Ov39b0IBuF1RddXKC!KitTs!{mQT
zbtxC3@XQVs?NX&n-iBuk^fh`ScvWWCUQa1lAQoB4ZNx5?osBR`N2d0#Bn-~@naOwq
zLe~Lv0X{xuWhXN}g?kcr#c%2EtQ^n4ZV#5|BIVMq&aPYYAFHj48G)k^jv0(q@!g1n
z1C%yVF)|HBL5?(E|G78G4T_Sw$d2=BuY|A^fJ?X}S4iJPEP*3gR9Rft56&_SiB8-l
zbm)-nv%!f?*~t8|QY(o6?a8+Jmz||;YSZw?O9BqIMHN1q&E3`S@0%EVax=>RVpW%8
zVZm6rY6<)Yr8>~NZ9Tz!{^$Or#@H!<>YWj7ID>D;`mIIjg=I;>uF!;#`L=;Da}%2a
z)OsSX8H(=D4|k2G@3I^Xg(VQ!_vbOdp%BBOT0&UirW!~pNXg8jQ!^GcG+g6er&7A)
z0(n8kV9}%K@Y|Ov1teMom)!{<=Z=Yc<GcCcc5l#p={OUj6J=2p3Z+Xg*7N7SZyFkc
z+OUTA=GY7|Jc%Hk7f%__uQH;d(ic>xrw6}@v94S;ZdZh;v(nVxhle^n|GM7U`|YgD
zKm1O}J<mT1o(c6a_ND<VLTzH_(bwy^p>v&?N$FH=W%as^xZ<uR;+<h!<R}1zyi)7@
zvgQg^mCtW&AQJB|#BemTn4a)CA==$wc*9X9<eL*8M3=2H%S$j*>I&t0<(wy+3j3O;
zpD23en(Mh`e#3(^-@K}G<dsymONUl{zb@@VGjGR8omsweCAJ$guXcY!k2p)=#*4wN
z%!?d&dckG%=II`pW^zgmgkL0xr3W0AQay5cxo$FjSWmtE_+?LX<E-OD_J=l<F8(`1
zV9%P{EiESoT0`dw8BUtW7CauASUy9{Ft=dpd}#E3c;n>+_L-F!{tGK+ME%8h(kWj<
zC47>d9zHJ*{T2R7C|@4+4h3G=5aBkYTz#u$1L~sA@B=Cpe@gg%D^hie4u_MTh?<W@
zt7QBXSAK*nv_IDF&jmmJXwjFxmh8X?Nziy06Is?ijDL}=2dRrjQo{ihmk>MQrY}kq
zG||()?nUPW(I;uJZo5a<f}In!gH*OYgV-d69Ffg+8sgZ(kzLZ$y&-6IGhNn|?@_Y7
zdgz<+zo&BixgotCcn;iz$momDiP46P{us(?^+nXIMyJa`J=_9*DgFCtGMjaS%ACId
zj;ojmt{RxaMPXp-UXyc$_KsqI=rMnBL00;)kr+|mqhB7Mzeb-jD?8@ucOl3olv>w<
zvM;{~h<`L4S^`{@->5Dz3;+)I5><^-pqKq>##3E+DkLgpHLUoY3=%~C)pxVhG$>{{
z!6<bln{F&z<-LLw77uMq%aLi`=-TK$(bJYdG>IWsGX~5{bdOk`L$3b)6EEgW<fcSL
zO7*SK%=EmKxiPT61R=6yeCi*#xP8^g!<FBiPQUH@65s0-6?J;6nQnoVCy^`YC?c>_
zkd`~x`+3#&>Uy`br&hVKIp?F&X})}0^dRBE(opE`N$(z!+n_#aQK=92F@-@D4BxHQ
zQm_3>inM&A&g0hS=w}dY2-BZOyea^jxY>Ntkli4g=R=-qe)ZtEY<)H&IpHuP{e>_k
z!^nd#cRfE)kov)<H|cJko$iPp5XHQFVoKT0auM3Z*$>bifMR4!Ws|9$AajM2iPP`M
zSgI54^oo6eeq()I*ubE34ll`Z)}FcTqc-Dz@VO(-ow;C%eqRZt>HB;ej9A&lZuk1e
zS6yAt`UUI#`zMs@p<lh@)24bWe77Jc!a!9_`Syg7`zVEsOYtHm$9H_Qw&lLcbL&~S
z<ALdeF^lbrB=L+F9k>;u6v(|#r?jetjMBv_(&JtN(9?QKEJeb-YFvsg@nkw$q)54Q
z_{B(5AmDal3lg$%zPdFP92xTVi@mpW?PObbL769LMn9e^-~;wuG4eII6k%GOEnDf5
zF{TBX0b7fF5~)^e3RKNx+p~RYooc$-nTG(BG@qpYdy)u2@t&-xI<3ywp*-ru_(X4E
zTY1PQ?K4n(h_P8|hLdb_^B-A_?wLH*1+forpT@*UORLY37rxT(Ee=qaYY?r3Yt5jj
zW%L72$mjGdqaD*aIrAcrW#n3t@nEzgdl;6~bnAKZm(L*rk3qa_M=Cq?vxw7}#|uUP
zcmK&uJJtF(q=LG{2JSPgxg|UiU?%`QPWU8ncdhETD1SUQs@J&<SAx@4k2+RIdb1BF
za1kg1asu(kp|j3zMni$9Yi!B2wnliUd&Xz?Ie3wb)u=o#-_*-@*bU?!5Mk}CA8dY(
zcvx-?>i~;C^yxumM9I>XJ*Gaq4cMF&UwrX!<NamZN)zO=zFfV`3h?_7HZWn6!cIFh
ztTo?w)e$bcYNer(6aMLE0#4`o_M|`<(k{7v@A8E)b+fo+)Q(6V5aE;2lfrItV*X?5
z4kkaDWC^_j+(=tEoecEg03Q$|qGy`^Yim6ZmoW9vZ@kt~)TllzMkX;Olmmzd0hHD#
zNb0)w9QtupwC16F#i57(DGr<o0=Y%X>gtd}Pv<gJII4VOza*fJ_xf$k;y3lJyt8BW
zPeM-Xg_Q!69pFapyl-E*z!6qt!gVBHgOt1`(%UfdgUHpn#5%3|f;o>kN3oxMrgyO@
zcTG}}$etU!K&}4xU<Ep;?&FOA;X~kn&Py*wh@VuyJ#Jy-argrUuLMRSA-JIKU8i1f
zCx5tieAQ$(M@bVRpTAnTJ1HkxK5EUB;H^WC&4!=jN2MB{B6^9mxv`Zx<?9ikrX1Ft
zb*y{g$h!Gz1|jp&qLw480+khy2G5AFer7jgRQt9<I`OV_w=Q<0kScZT^u7&AA88<b
z3Mh#!#xt7#a<mijcub<C@*dF?XhC1*d@Em*>|K=uK*DVLh@{s(>r?qw?{;g|xVB0&
zvpP3mdMgt!D?}k8q69^j5v|H=vQ5SLN!NM(d~homastZTv`h)|UHoV6_AK%A4*!8E
z6U~C97Af8A-p?mic8y(K__wQw3{v>8;N{S7=N8pK;zn-UnlcITQBe95d-q04RAiw4
z#rEQ<k8I(nB~<CerEe^D6nud@u>c|`oq|H`JU7)7ezP2n+7%7n=4CQu^pxm)GRrRX
z+pjmGj3My~dxT|+3x5r42y_TKQif6+O-FBuS(LP~5=&_CZlUI4Mg?VujhmZef_j*^
z!@=&uoq@>FKH|SIjmXzb<5K$i<2to|L8Bt5ZAe;)hmOegOfx(__|xxMt!3^SDxT6C
z(IxD5Y3*OZem4v5F#eVijaS~y+B_k>=4bWLrSURS@r@3bjr@a~RM(jG3MbfY-<4sP
zFmEqYZ$K!Ywy~RAUpq1R7|5ivU#i^FztY9~wtEYt<zqwq|7+fdEi>?DOQmV_NU52n
zo%80?#jq6(n9`tcbJW{j(kB1)j;&zvKEnmH7@-;WO+EBxT<tt;@}#%v<s<+G9JeP*
zd=*}6lH@;W@txDNP%$tSHrsD;BS`Gz)ET3N(*Xj39>(?O{DBdRtWvWDvLGyl7s7Gl
zFDeMh7E<$xXn7rjw_zUvTF7juT1#k+G8mM4=qC<sBqaVeQG=e0P=IZ@c6^_!lr#vh
zKlE@!i$qdksk`6dMK@@S>yGe02-o}2O)4eoCU4{i!ZdH1UJg7_Y?t<<Jp*m4qJ~jN
zp~@zHlHlM(BRPX)B3S$FVl6y3xBX1r`jPrtgk0O`mA*QDeBX|Oc*_g_l{9-(dg|iD
zu^+jq`VWSGo)uJ$h@ZV$-E4>DF?UaEuxJ$~^WA3sN4iwfn_K+H_7W+3G^M&kSz<rx
zO%Ef5x9ixI?$)`!cq-}<mYt#3Zs#C$unme;Xs&P-;t`OyAwh<)OJ|Fs->)CE(2WQR
z3CZe3z5<+86e7M0A6za~wv~ov+pk@c1MrFum<8ZagH0@jOg9T5eS`&);t(Z1MKX1O
zBTfXiO6FS9o__<7rqso9DlY8I+LsT}s;L^X>*1&X6%T|(TB&M;A-o5|NAn@?UDAu+
z<q^yttn|v3j7mf%@oL`Rc!>zd3TlhiKa~|?O}o65c6s!E_NURwu?u*MC(Kx!B=*69
zVnemzHpDA(?Cb?oyL{Vs+Q{7+SFL~C@qpij?{?FvhcCi6`#cICn247$GnP(#k=D3O
zM(um^=&te5hY5t(2uviD;d+X?%rzhKphkdWw#v+S#<BD%WICZ0@+;;4p~s~WZQ{eQ
z+xfJLi)x`gd|FBWB<0Cc!~+Sr&aRjTlnGX04rGOP97EUbaOV%9+v0;#9e#)xAy)@y
z1-S^!&g9)k)0v(|F2mOmpE7yXMArl44S-$H)kK23yRb>hSn7outc+N=gO!KL`@%}J
z790^H$|lA*SZ!AxD#!}!ov}nRt1ZbY$Ll+L8(v>u8Pc8dC#C>skYlHbz~$`a67OXo
z(qnJSk*^>jLC3%}^e#K*gFPT~6TdAcbm*#F1$7zmm#Zj=3Uj49F5c=H^YZl$p!tos
zOfPGHy@<V#m|^0M%SoU4t#;-gKft|thNW!dj;B?7R_9(E{mj<vS`H4JxN^i~B^WCO
zhCJ9w>qebjo`5)d>H3y#`9q#5`S1?HP*_KmpKD4OG@fFFRoyP-jZUZOC9oo|(e<U~
z&DEe~MOQA`apxE=>y<0-fszGc55H81feFnL`Yo@6Z;9rc*G205>h2@abU;T+X&q_F
zyKiR5>p}8|lZbx+9S}#)J6y40;%sl0>ZhM+31k8yb*LAACJ19#X(=-B+JOQjQ(2*w
z0bZ-8bm(JN;TiY9#c4@vY-=QIP8F4y?(iU;Z-AZBgF6+LG%6Y>U5g5+*M77PwQD2;
zONrHWqv?o01SGdD0iZHk<84BCu!%fj!d^)qL$1(u#jns^ls<Hig))^kjrJNOWgiap
zMbyQ@mb0fH&9p7<F`ij*V&I|gcuIWpq6-nWkE@yonp)1YZ+zah@dX}hy5N~_-R_<x
zLJgPfv<p-wpZAIrSpreTj8B$d-BM>n?ooLw9WV*Z$VF5sbzRh$<gA~`VXp2=8d-0W
zO$?xL5nj~IB|6_5$fFKK(YcW91;&9-#s>$C$N+(!h=A0w4^qkh6(SF>8rPMCd==Ge
zn~4dD%*_TMi)s+?M1%HHEyeewX^ZXh9a;IEc?fPHjW7^t5$*HMTIzleaBHkYA+t8j
zVbB+Sc8%{2>aZxuOU}vL7YdgG-|7!}nGkc!>@(Y5jq^`P({r$r$}uqoW;^C})3R-n
ziN|Fs<-yW_KSg9b**l{u_v=a@!9_0|evK{G(MiO9C2lYB4Uy{J?<yeM97kp<Z{L(_
zJX!er7p1<aFMYto`?uIPx!?a_Fn>7uzH97vVZ*swzMVdq{bLRgN=%<3ydIP!FZJ86
zD3}v}wfZa>cO$Za+|%}BnU1XGHuIUc2e+!9a1z`;V=Q^vjcyMh4FKE)#(mfD%K8(;
z>WX43W6K%rt`SX)WI}}zeDk10#oHJ*fSwzd_re|cN09bAzajk^_d{q-$X<rIhUhO7
z85_lJO6oRsYLO|_paEfBrvAhhfC{2P7?RY@wM)ANdpO{4V3Y(cl9lxn`-#C0*&#h!
zHSXX8ixvD8aqDfuB{|E{f+du`_D?!l<w@Cd6{rTu3uhyW1h`RIyE%6k$W}xq+H3(A
zLg0qBS!U+Z%}AZd$Dj{hwd7*Wz^4Hl1P&l^f~VP?h)luvTuCGxM$=i!x<r$}Hjn$c
zrrDSIpkkfyw>uXn`#yU1)sI@4ceeXnsNHt1@44EU?zDk4TaRZ?n6j@%3+rG0VM;wR
zPYm>K?YNxBw)oVsuJx42nd|$a6$clK>u6J#xqI$PmVxljXLttTL8QvP)(KpbCM2&n
zUGH$kYC$gDS+D`raFy`?G#_a+t=Y18$DMPt2Xepd;S;<yuuVEKCZdDmeM5|jvECz2
zq%!0K?)rxjs;g=SU=LinhcTtPrvJgxmaUtFsK;`x2$b0H9`PZ!;GxBG+%b=fKf07G
zRA;NT6G0l1+8n5aoBDd7n;}xBk$D0PY|H}^Q*OJ0ozSIVrg93-PBT0b+#v$lv4Iw1
zV==Z*Wa?-XkS<9gVU}_3B)*08%`ZC*N<)Hr_ExcG<{XTodd5UC>|G!Cu^HLV*YiVw
z!DUpreK?Mc9h$qOKAh>#l{yV9hu+nw4N2o(Q-4kQO-+1NA0F73?W;8TX7K2+pw~KN
z4`;;hyZAlcN`phqF1J=5-7vyav!iRUWOsFb{eXp37b^~g1md(j=gy^WP2f-JEu4!z
zi-lMLpqFY8eN?q;!qGDNJuz$&Q$e)m!+WZ9M!*q;-hGrdns^Oi14t*uN;n>aN1mTn
zsngN98q^~Kewy6yf=el?6Z?PAx#6-QgHkUZW3|z?Cz@<f4VFMx0!b?x69g*&V-@C#
z7&MzJS-Mu`UVRQ>afW1m`n7E2<paxw#X#7i6w!YOmw{Rk%HL)l1#XjL%--|tsj2Wh
zUAhb3+I5TO5CqO7GdS5*uExX~c$D|Lg0<JGtH{P?m%)r|Gv}BNv-ob-b>isa?3N5i
zmM&U-LUOv7y^x|zcIhiZ_g0i}H~wr9XUgI%^x&82^IdlK=L=?p)9=`|ecsk3t@TU$
zwl*S8j4kLx$=G+Vc_UP^VnIFE3HqJBA5sG*ghMD={(C({Ff;GJNFBK*6$;tjfzW%Z
zhwP06%Iow`7SKj=TS9|+q{tmcsFFB)TPhK2N_(A#^#+Qy#8hx(o&B2__{hlIM#5AH
zAlI`5i4GXGDJMMiBc61&OLWHzy_j2J6lgm$fE&{#I%P6dG=nikw5P%s5ynBf_CfCv
zOJCNmUEAageHfrV(m#Q^v2|9P+9~VXy@gBHx6QoKx$weDY|z7iMnv0EOR&Uc8W$}S
zODRsSj}015v{IV%QP;(lfL@2<9h*WX+(l1F6OT1pJ8Os`97!LVb3<Rc<Dk%yFGHW2
zC%y@+&)W)v(eqI6-V%0k+l<M&GV~N=R_MP=N8FBn>RflU$g1V{1JVAHo5LT!*OVK2
zbY*V?jL@+8)Aas9>^+K}mo=BSN}>EsRhVQ|)!b>iI-Cv@uQ>WlVT)+X4%^vYI#NoO
z638pDU<RM6To}JHF@IH`H()Y&w1Vw%jm39==O5$QpF2lFb*aMsFQlap2{yASI7u8K
z;?Yev@2VwaUO0%5GL&1wQ$c7k`KiM68&`DLGTm&9Ta@RR>42aS3w7A-oS<hST^;Kl
z%+jM6vkTuHjaOUwOUb6fwIl|HooK1_VltT>gAISh+Z+477Ot(A3OaHe{B5@fvY#00
zyD4Qg{j2LwlEJ^HDfJ|@I!TpzQT^;hf$p7Emi|#T%`3AWD|xNXc)rb#sIq^oTImOp
z-!*%fL2EzW)A!+}6Rfe_OGa{9S)mEl!i)W+r(e6w7W4wxR5Cc>r*QT<(PsOdKlD)E
zG@uSQ6=?4#!=`gro5$n~Y!$?CiVqeV+`jnj;k_AO2e+DUeixb(+Vy~or&{~hcN|-C
zh$7XSc!zNhL1}m!uv$GVX!<)>N;q<4m7L5F989?8tVXpM8^qGL)^l@_=0n2&&36i4
z2gzk4L0|N7a;3q!&n8j73Clig`b4Dv)Oh_A89VMf-XV^XyE9yp;%B~a$C@mjBq29s
zsO;0sI>7afs$>vOK$*C{*6`^=_6b=nB25AQW}G*Q`<U6Q3b*dl^%=Drh<M`EuSrL)
zT{(aQRacT^5qc%2${X21qmMRUZtyO;mtWJ~Kk`|uxPn#X{`N;2Z4XDeCWECev5p#}
z2?@uKk0dmgSA6U@WkvubfY|oMlS|zZ2o<C$>7A=KNlqciU7s8_lTa7G8<~i^>HJ|!
zfPJX@i<dqbKPnt8ELB|@aq3`%rS9B;3J+vM`2_legAefL0e5l{1x`66B9WL9p}CQ7
zHTENOET|ZykQpX@e)(ON3M2uHa5%XgeR1-YjGvE8tjprkYTwC_L8cKzjnI#P9P}Fh
zZA>$s><A^JBr=+N<3>If7H>;7fPxGZw<9cIaMJ>Qpdo-G03j}!Yga|A_wdowZr?tw
z3X?^9A1q=$g{2P_t`t|USu_U?Gi*_vp2O+H&E&o1t6f;HYC+vb{_R%-5vi?J{m6Ex
zI5z9Qu;MO5ofnEz(hEMF-Td8WWv9tmN!#&cfO|3^F5-lSJUf!_wMwp>%)8NX*!M(F
zN>jX>B4(whkD)@)Jd~Cs&wOZj_1ofRcUavDf&MdO&9@vDGqqltZbkp*b67BaYF@)g
zs#rGZJjF1EgM|+Sl(<sh45`T>YwGdN@!in~1Y1bt-WcaO7j0Vkx>;Iy2yGrSl08v6
z+nHTjW@?-;EMs7bZWGrqZYd&Sn|R6L0VlTM;!*-k3SN|}=+l6zwNuq7@boPaW^}D&
z>g6^v5*%B3NbxOXm`ax-dRC8Q2V4?sB)1L}w*II)@A);`)tY$6P(kpi=`sm~LaQ}x
zn&AEGg#V=BB(?rVQu#Igg%S0xwKB~?-}b%@^|W5@u%B{J{a<^A1o%$+Ol=gy9sjo|
zy>gLZTYKK=_~fOAK8jA`7CM%1YNv|{$`f~D1i8k-C&kgoa#}u_*PX^HXR>d-`3Nk)
z(czopbUGom&10ypVY4=vRP8!+9btzw^)!>6+P~B~MM|(^0xLNn$04^TT?EUDHc~Z0
zt=?`I;OltgQ1EXzSc`CE4k3sCM)PAJuNU6=y4KqrG+eNR@bwY|+(G<>IJpHs><=d1
z>K!ckNaQy|d;qt4oFSTuyIR%AjKZCWU;^4`MmnII$9tZ><X9M0Eb-0DMIW-9VbItV
zUZ!oE=<vRdx0U5|AzOUf55|xRk&I1ccCvVrICKbRLe7pNQq-Crr@PQ0_OG$WdbqKu
zo0)@dmbumL)=GRWpw+Ms;g+0IiN>h*cv`%<<{#0whlld$O|ko@vDL}v#Y3M@l^3TB
zT2%f44;_a^4NGlRMkk#s(HF$Yc(DAnwR4|b3b@+;ZzRT!=%1N0B9mhW(&jFz&iN$d
zcFpf1LErA8-C@6(yfz`jy_8S3rkOc6&UCilf`(qK+#+DnCdrI$Xdt^?lv2H_E6++e
z<~hDJW5^AAssY)e?Kp#(Go<6LZHd;o6nF_ZQa(I8$3*L*)(a>)+{ebY!RJSIU9=se
zjraBtZ2-mx&eERX#MFtTddHzol0&B%sqk+{dj@GRu$>%vaq5;X4_@!9)Lvwv98S#I
zx6xdN+dMmW^u6lxH=CuH?+|{Bgk3ova)}g^U6hfuLT(=zi}MFwGVKT?r!?Y#2!2Pb
zI}9|2C3I+wcTwRFwY0nX>&%<mzpgyO*|eDQPP1<Dw%!IB5tkKs(~HLb@WW7J$@`A!
zVF~=J+Y|C^_U8vH5FSS4{?w_9Vf-}}kh;wB+uu5~$rjikMk`8>#Ad$J)R}t{Orp_N
zQ3L4Ac}R0KiAoYNkxujux5GHUV;(Q%_LssR0hFI6$yd;^zH~8|X240Z10M#CF<1iO
zmyfz_gw`?@1xV0#U*>FyCu_kbc&y%D-M)@U_{JU=ggg@YMq1U7)8V<hh;C*LnK?Vf
zrcmx9Kpj(cxFmcRjHhY(v$4?JFEd3tthxhEc~*;Vc&gJE$G24JA*<-*W3B=Qm6`7=
z4~on<`k~EIK*P)+tXJfQpJW2K2E@dEEF(5WrCR<-urt5Bw2E~*hhmD~sk?9MW?znP
zd!9zgIp0U><0Wk*k8+5^qsWj#Ty|?q`iGOs9x}T!zJIxVc^AY;vscHQ5uuirJy4xk
zKCgkVLo-cfQf%!B;p0E+zO48<67J`!l;kzrpU(GK=lXWW>Y9PXoLeqKrxN7-`CHrv
zs59&AJ?xkKmI1Z|BY(<j?8=)xzse=UYWntRH5U$u7FHqd!GOq%59GiwFqJ~03e*MX
z3&td3itIO@5!KbcuyO+?*`ps2#e?_|s2Dgc2Pu4+EBga6#f+yv(eMKMBKK($QOo!O
z;rTjv)U-<Jvi#S*B|900U)-!mUK=DY)oY{%js=i@UqIl#;tG6^(b3R?ByX`0^xP($
zzQEhz>H6i1F<*O#L1&fC)2Mj5tupEOPOWNw3ejb$+2;CoPD(y?v{>?ZIh&F7*Dl*(
zXd$_-uT$PR9{0B=HCFtK*O&h&xY@#nK5>V=YQAstiay((b|JFwV$^AoGZrfgtaL{o
z+*X)xT>J7<n&I)rUy)?=efv~1ykHxWK4+(znLm`DW4v&=bx?;M6tD<Q47j7f)yFPn
zeB}LD<(CtdU@k|~8!AN;(UqEVaS<f?_9tvYG)YM;bYEzhQl-MB2=InKIMI*X7YXG!
z!AfCVf`tmHe2iIg5idfn=||3~n1{0ptg%t?aaMX858|Itg^d^Tk9yC1SC2I}w5)hc
zziVTNFqm~+r(OB83&><jIPlQF0V^Vs-``?$4mI!xOVS)h#nS1ct&54xs$%Cy5x#`6
zfe;?jE#V;%-WjD)j8T?SOnozijTuoM+s+aj+JJ>lt&=mYa@u*W=1n*DCkwu{B9NRf
z3`5b#LHUaHMhgR3$5Xn29X4FI*Bm^%rfsLGPr~s}f3k^{w!9pDde~V<V|3wD@Tdnt
zqt*{-D;IdO57zw#@nFY{0P%1&D)MFkmSo|=Z8C{PODHf7W(Jnfk4I*_EqewGZ<f_^
zeuT6^I#a}xg0+{9$qM_W^YNgc!(i<l#y1}?HA()#N$}}J>M^0eA{aOM=gl{c(3M44
zE2ws0D9Rx~MmQFd6h^C=d6DuV<Wdy19^96I_0fb1xvk1xm~X6K?c)cMKfgTWnMtYG
ztzp@AxGVRRe6~tK^t->qX4(pl@%lj+9L@cueCX?HB0>UM8~2bUJLY4KMk3(s)qAkw
z)YcY4UTxz?4?oZ2LL%(R7Dtr6+w#mJ^e&zyy6&3`Y+^c=h90lH2Rhe96sd@~r6qN_
zT1pP#*iWnjs*f_%D~5+I;n_2JCIMB-?;6xU5Ahf4u!>haHMe%KdGz_6n;K`Irq^uw
zlN;r5Dr_(iq0lE3ezxQ(uDCTGD6rwyc=2J+7lURSU)Z8b_gyWs^&GvtM?>4~2*2#+
zn(X49v-7{6534p;_$~7!P|<*!OE+A!mWyqQai@>%sPfH`exDVdIFi0ceRwNO^oj<E
zu09lNKC>75mJ*vKDzT0X;38(Prv40AHxZ?h&SX80SO?LKj@FT!r8}YZg^nH5GbEiP
zcmvc&C9T66+VOB+e1qt$0MElxK{2ARammjZD@9=ecLwH?ML?JE`>g2_898hj4b%*M
zBAyQV!c0}-;JAl(D9a3u?99g;oh@joa)h%e6p%l<wdJAM_YQ9nv;bsGETc~oxRg@C
z`ABbK?=M!}lpj&lLU=mf0hTMS;tplUbN3I3%xqirXpfM6v!cM?D;Ls)c{)edO;~{y
z<S4k2>6e+ZqTc=6f}e!gq%<%ihmw93P{NBB_B*6)de!RPi=U|HUEaPqVEE&2&(oy>
z#JZb*k^)@Sm;K?Em?dV)qnEmg)<log;)>6aCejsh&7w8Zn7y?(#YV)_oiZNsIzy&!
zy7Z3f>s7)I_I^+zqw&Q6Dqj<_&8Qb!iK-R<-t<STL-f0A<^D_W3DIQzEV8P-x76g2
zk~k#AIIpv&A-&GsE5a`JyfX6O9Ab%_s)o(!@i)<wabdtEN)v(^(SAVTTD814iS-?I
z9c3PYKPZ39Fhv7wDhb4K^vbmgj1)6pIQ_FS;?>M(g<aCLEy;zOf8JN1Mc`#7H5;cH
zZdk&nbYcz9d#~kj@pygxtQnVyZP;*!O(oz|=u6|n$cA<Zxl76+cN}6p1!!{adO939
z&hCK)5@bMEGd=>PmJR+a89H)LYs?aCS~~a3W}F4<4nPaEg1wA0dPX28VYHI3X>!^r
z+SpIZ|Mu$iFUNlL>U07ub4_PL^uMelfBGCprR8Upe*D=6S}S;R+u)KVhHZm{ZS-Fk
zhTN@vv^KWYw3?BCnl1cW`NcQh($^M{yvR^3x!P#z6X*MEN|`;16>3-D*%`Rboc$C(
zkU*ZesYZe7{<Q%{LXOuo>kwQL9X{l38^|BO%&62?-O^^?)2fk~6<`1Z<cmt)9R(&&
zqzDTCBGwmW^`0scd@9Pnc)qbItcYlkQO9sYqey^gsdL`*Nfe6`QLI2?E72=m5K>ym
zoqy*kxjxroVwjJq8(jX~V}i5$0AI|f8Z<+#$CyLUACw?E8NeZHT(j9dSLkZSvx-aC
zEB7#7n`NI(cFfV|zkyIKP7Lgq!2*V|7%w+b)voM*PEXSS_j9<&_DLoX1lHq$eI(C-
z^tn9eL5ovKlDFk)7<8fa2h)b>7?`y|9+VmoBH*=mbkBhL1@<?vkFwjK;J%+z9KXa2
z_-3Lh)p{j0xDNy<khKS^E<`5uWeZph#T+gxQl_~6Nf@0YyLt01uaQ+3BFcv4+eeG(
zU4sGn_bauknc5rWC<}~Ny?lb}yYts0<Z1G+R*&FcyIV@WeT4Qr5keaTIQEVi&wWFk
zA~qBF`hpjJ5^a+`Vh~kspP)MgHANJHr(SV;L1}KHyK&aET*b4nz0g~;lp!cZazwYs
z7U6b;Kklw`mMU&SzORMlITs-v)fXrCecZHSuYe8Kg<zjCQh}|W+8`H>tm=kF1>Fyt
zv3G>wipU{C<Bz6CST)qrcp<C(M#^erCiosUsU3bh@+2)HtWf_c(|tMSeWdXYcGv&u
z|Ixk%?n&783HiLJJFtY{1IRj1yHU}59u^*%E#KX^ki==yj9B0v>xur2^x?wRONxt;
z40C;-M&)*`Z#k|sY<L)96FmH5asC+q&{2u8YKZch@L}-6f0K;Wk^Is`qs6U35=x9v
z3u%=al?70}jWqA1nXFy%dQ8S<2d9nwmmH{Ny!qE=8W`>1!!4)cTh9Efy(o2kiN18=
zPFcO%O{}35!%+6(Y+sNy!Qiu<*_z7D<jQ1-yP9}5nIu;YZWV5eaNLsB>vOVFFE=NW
zKL;6{q}3C~1_{qKLfbBPh}3ks-g7LKulYeRsu$^&j_0Oh+Guif_334AL+&22MHgc!
z<M?i!9~LrKb;IUyfr`N~9xJ8Zd4yO&$);Vl$mt_=VXJ3q-aZ#@yfE(Bwn)C~xKzLl
zf}p_%@Z&^S_h^9d(jH|Dyd)b7N`>5G2|j_?g0_tlkW4H>Ljq_-mG<;YQen4{M!B=g
zWAb^~vVuLLm6R3hqKc({A>@IJAT#=FiFiG>SyA9z4T_we?^SudHZm0mvpVWEaD7Jg
z_rfeI8bsBGD~JWXh3E~^1|15%7G{~neROXdG>G^PL&MR^_K1G>S3WJ|B=Lt?FWdFw
z@OBQUFk+*&<cx?-TkX=RlRQuuiO-RJ7>v2!Pc)JBYuGS!G<k}!aLSsftf?n94+65H
z`9nJUHqV>tuave(hQESX<V5BIen^x2?$LI7;sgD)esoJT!;W8eLBrl_8Z^jP8y5mw
z?$`^p8fdd{Nyfv`>meiBZn2U-r6UwL+!(C8(5NGn(yI0ILY*L}l@>mCUL2=yg?)Ue
z#7<mkIeSiL-*r^_3&Pgv=4pw=IJ`)AfoLMubDoe;)ZT?V7^+$4jp#WG<=%@y8-i9~
zb~o~$i1<}J{Eo`dEi1e5ejpK@F)x*xJ@r@fPEz)Yp%AVE5IwIS@=A9WZmB;BC&BO1
zso=8#EWA6y)EK@V{0RH6X55lm(9rc>C&ZE*FdEW|)bAk905TsgbSETT$&;V6Wb%Ra
zM~aPj;gb5(Px;mio3;Fj6N$BT7`{<o{hy@S-eaeaf9mWuH!aQgzlX_17Iuor$C)Xw
z`k`n3$6_6(smYgwh<xOdYt1IXRcLpGSy4}(Nk=4WBBe)fVd&0P@PrX*7?^(1dVq!v
z>F^pEw)d?X1~9xkcXu2$8{6d(yiW@eCk?4UbIz)K!K&O?eVvn{yVh`v^ZGNA<hi%-
zeT$#Qx2UdrFpfqA|3Esju&Z@<<gdrzMRg!6POWaaS^Fi)W)HG}(c>Kx(ukUh>3vl>
zt-VQ<Y2+eu0mC<n5I$(Y2;i594+hKxH#4C}1ZaU<9E61JH}&6doV0w*yuF}-4L@zX
z&~YH}fCzE1jYS97V6TzE*63x+HzTVWzLu3crws_%M31Pd*i=<p^$v~MORq!ynXk69
z?y~(5B(bz9jiIrS(jR2D7w_2YVZ)XKL&Tz;CWg$v?gag_o^3e;9{IsJDLIu}J}CZN
z?=?<+dc&VC!BA7tV1BKQKgEG9H>QsBL$Vf@li~nj4XM+{CZPpJ(X5RIg0ehDQ-_wk
z{DySrA<sYdUQ%la=wgd-Oj#hm7j5PjDM<tEUw?g8Haj2$E!g){%xC;d?#D=l5f=nS
zx?05|T$1pDW3<CeK#m3PjP*;;!mJ8*7N`R#GQ!D%p6!BU_dQ^r(ew?kPlxLNzE=H&
zM9<LKLFW~|lK!#((siUh_#|W}A4;1M0m~Rcf}&l5HR=kss{%d7)D9S#*udIv9{Rv`
zY;@v8QSo1$7G@=XMK_z1)DJ>JPdH+x6F$}Tc4uWhYyl?NY3n$h8+Zvjp70y&3H$*s
zR+XVDAvOu;t96Z+w5<tVkm4ZrIIy<=!{^Tkn^+<s6IFNcMF+XJ@ox_m(Bdy;DI;d6
zImK^Xj7RrroHercIDB@7AKQukl}PtrdU<}A^l$h6Yb(1BGtBa#V@2J#pVNX{^?=B(
z1)ZFol!5({sWSO^6)3?z$8*=EbxI2B(^ogJ18hQRmoNhfAifWs{n(h*^1h^}x2U%U
zy%RK{%a$j1`lyz1Vuazp#KVED5X(I-eDdGR-o)sRMT9MB!cMa*;hpF_j5iH}ad&G(
zBb%*Mp>yC)!)TNmC>>wdZV6&XMf91|(4(1|9GhD(9CP0znvRnX28lTH#5#4{$$wTD
z<KISrpNrd%<!`AURAcx%@W~#wOv!qHEg1gxe5+-26QyUx&@b6!SDyT^z(}o5M;zNW
zh`?}#{qC+I9m?poS;hFZh?xfW`cTLHP{OZZc<IW`s_XkPIf2P%|JqtsdsEnCTa?_8
z>FDWg7gNYQgXZz|-5V47d!AOUE9#yk+|4R^Ck7_-N}5j$#KeF08R{HQ$k!?|*In|M
za5;X7nFeG|p!~LL>x61q`PA6Yb}Y4d5$0c%BY!&DC7fF{jqwfxRmRs8#*9dMK|P=$
z<e=DO-<!4sk$lPTUE;e1p=s{Osf61C-fT?JsJgcy31KK;mP4E(BqB0lFYJWpD_tVC
z+uL<}9Vbn}2N6JwixrF4xWdm<Lg^+pQKDsC?bwKc>ltqe@`Ske@<|{~i4*oluqYT9
zT#U^Qs!SiLI7UXfV~xhKd!i=MM%ee}Oa{6`>+0BY?^mJ9qLGSj*H;A54^9u<>CZoD
zu;^}}#$@%~8Qc6I+s@J5U4l;z;2+~Az7|HDUKAb0Wd+kk6nstT+|XOaZJoJqNbny%
zZ_!GBXkCUxENOCN&H{+=q0)R#WKXZ<x&9r-iluw5;H&Cct#IfsJU14j{JW!+iS><m
znbonU>(0D-9_b-`ru#p(HH`&|i%)KL=_XMT&%x_s>1i&?OU_(ho^>H2of)CmGM#j!
z`l_={G)67j*+D%zFS*kVWw?L6<_OL^UY0-L|5%W98S?-s55bz2%;*ir238c0FH2U~
zM`kZ;K({c3ko5x}hHX@>?YU^vR=7q4Q$;1E67wa6yClzV!g6(YcOp=C#|97aun-46
z5%|W80a;cmp`d+rqt3!sn)FM{TuV<~(JMCRIVVKH8W~*qs7}a%55)2Zlqh0};<ak~
z=cegeRs!aSEH40Q!NQ(12FP;zB5S*O-OA=56XqseA*`#BuJ~a$AX5_OsBB4D49zET
zfn&=x(dZH@=10n=mIDt5SG{nDPUA1;lo;ax-QznQ|KNTd;lnp>OEi!3$HZT76u*qj
z4I%`Z$ZL;9RYsjgA4^HvinWlGz08hrveP!H+{E{+rhV4^y=x}lN)v&a4Gv+#X-Q|U
z{hEW<Q8@&wb&1sCjH6cKPh56Js*HL5ofTC(*R}MgWoQ<WYPYVXGyQ%0soZtr?ePiY
zn{Iir3tSs1n)Iy?!(B%Mm*G!6%|83MoL2l~%LjD~=esDiI7>8#9{@o#=cPLCP*85v
zh8Bco%z|9~>P}*-&`)IRAWVj4+?4mdWO`7#o>)zJ)cMID7k6S7G*x;<JVLrh8s$qN
zymWIp5w=eh4O3Snrd%6bi!dOO7j>iGF^nT18#&{`o3RY5LA<xK*p$+a@Caa6)2f7?
zUG1=#9_jqStWU7~s~I!qHez8}W|>J|!FLpIjIWN$le{G(SM@Ty{N7ls1Lej`B&mD0
z!Zbqo*l~|Dg2~kY*av|NhA@=Kh_jAkZ6;WEVGD$sW?q+E5MI=ZJV(1Dle{loc1gcQ
z!yuhbgoO5DK>~mq&OODsMF}x9DT&6l1)5=ce>$y$?rNG_Mg;n&UINgBZUnyQfgJfc
z<=+ePe$8(j?7TW>wYd1OHmA+~KY;mJ$dh~$*w^m){O(LP%giX=E9yb;I@9);n;wEN
znH8EGx$F}mBds5q8Y1?a=g2d2V$efj?-*RxmGB1=f<OaWly2jrYXR%wLTdc_5#Jl}
z_lb44#A-4kb^|Yyo>iZnMoFRDW8R2llQdYrr>YReZgZ0=aJ54Ou}NG2BJ>V&R7|K{
zhy|l8(QZ0I=t@l7!{6iI20cAyzo_{{veijj>mjd)R^C*GmYMlPn65kseD2*fIffE$
zln778lf4j9MOh&;>Qs^QFUrRD?b5zx=C%Z9_w3c6s7L(9)Ry(wA4n3PkVX>p=b||T
z_gOYIxn-Nrco3EkVWa61Vk`;hUH>UhUJr5xs4medfD^^pBe0T{7{DOM2?_La7RzS!
znVAKq4Oha`D^0_fj9-E&n1@{Fp%Z*x@uv@+CQhYnftBVBZ=`f<H&OAs^&<F*x>Ee(
z+H(CS-r}c4f42E4slOV1b$DwbvB2uz6~DiuKX^Bu7*OBRdn%_QaCp7lt*OCbdxTP1
zIjY9bCBXgf(?Q+T<;VjK1=fLBZ2SN^mNRZimA`(ZDY7II7%D(Q@*VnAF_YNYj8_VG
zDXPHPIjd`K$D9|nf8pU?kYOsYX<owtYQT2OXd*m`lw5MMwn20NBf+E|o3iw+jD#ar
z0yY)gLI>mx7TPvYpU!tWP11|OJmLSL%GJc$Rx3q8<12fce2qyTPr#FpJu=xk8kRa|
zwx9@ZFvGpnByr1Q7M=wk7H@BFb_YQ}WOZPFb^#|S8>|REK+-0EcL+ikF;!j}Dozvb
z0uX{cb)W8wDGp0><PnsE_5!FsLY^Fx!w{3i<@j{^_Fst29zpG}o>^7R4T>E92lNQ&
z*!e^d=}*SPaTzgRTuwA1-EOhFrCs0UW!8lPKb|f>Trkk{fL;G5>PO~ccHNeBf-=61
zJi}QeiASV`lj`94STArk+I(tcYmR?){;6{{Z!=ShJJNPnk3`fmWe)7W5#(qv*Th{#
z^L8b=21LuC#+8FN!sUbDC*=@gl^3t(ZOVnrW+z@vqsy8vV-yTb*X&gS7D)y-Wli%S
zMe9H4Um5Ub_-RJJS`(}_UESVDe?~557?|gP6`{G}$V*5kAl7*c4}hOu&)@O7R!U+_
z8vXk2{v2&K>8@~8XoJ7UAr|+<>sqy-t@X6&CTZ8@qar{jvX7!GfgW?P`0n!Ck-jne
zdUbNY;v^AD?K@aU>y`6Famlmj-V~Nk#3du46Y768S8qV?2(W$kGhzYK<HzO11`1+-
z7b)7ro*|<$jY4Vw1Ot@R|3}!HKttXCZ{vzYDOnO)M2xkR%2FtzjGZK8sf>z9NQKa%
zvSgbXyNE1_k`YCg7R8w4j*uj!kx<H(5T5I!`}=!-|L6Rk|8vfq&V5eZH9nvB`?Xxx
z^}1fz4Q;j01$(ue7Oa2hz7-VeRvoxsBj&fLQADrUc^H%Kb3O=G94=4by+LTLOr(2j
zR5Xmjt>Mw6*$7f=>ax?vLSl=~Q|6)QXY0Q|S)uIz=E;_S`4ZFuCk}`&Go5*Q1#{US
zHdkV9qv+Y&>663LZuR`zoQdfJ5yQJ?o+N?S`#0s-pZV{z2Nu#0Uj#uRv78RVuY7C9
z?Jv{kVL@ZbI)fEJ@S7t1?EY@AQ3x$6T$%v5vgs3u&I^ipx0MdLlqn0mmaGdQ8VUf8
z5TzhY5ZmlN0;dq|UYm*K8Ftu500s~k6#kySxE_u!7#O~13l%Td2@ckpHg7>D&<(WI
zA-DO$!WgO<D7gXqK|4bcNjfFFJTY2l>q+@T9dow##5g!2%Ys4mD6{Udp~PH;e;Buq
z6*DzHH8eynlAVSB93kg`C<yW{06B}~ILB_?;Ct^d%j&wbsuAwzZsRA_tv>lVv_)5`
zXlkVwtmQbEtDwmN>IeTWvGQwCJzV`^{aaw2PrLq=cqt9Cw74E|r|<#5#IKLljC@(~
z+vCZ6JI~AVk#l;&j(@$h$Lpruzs4GWJAPu20O|dz8>b6Q;%EPK2|6v6T>)ouK275d
z1JRnlXOOzxuFdPo`RzL+F#GviFF_qr%nQYDtd87`2LvB946y*j2HxFcF`%+`LUcmS
zU^@~hwuGNg51J%*2K^b@L--&9H1&OIm2iJ%O^dn8$p_yazBJwuzjmR()^w|X?19&n
zmtXubb+4d<tW8*t^!0lW7K9t<nMette}eKDAQ+$*P<)Ud60=9pP-H5xXp3wO4K%Rg
z+*;9h>H2;9WY`aIA@s4)wCJ{8|J)W&_(c~e^;SfC<nm%3>Cl0{Sd85rtDf%-M~a2s
zXv(zgJ9N8kYs)}-e~t0&+^M!aoB~9wIXDbShtPffZX{%2I|61!)-g+S#}7HNW$yYF
z2JAcY!)K{$vgi_>^Z%r-JACluvEv^v928wE!DY#RpKC`LKWS9usN?Uf72X_|MArUf
zZdBS8bEqzai^<KcE<H~ttEMu5@@|#aE~%|kO%4fn4;*2ym)fkd?iqi}>^{X}#%+~H
z!y>j_{GoSmzYHz0-nU=BAav|N*VDc}!jtqLchwe+gF@+?U;BE8)qoR%Q7y7TQMOk|
z2?pFA9mpZ6A$x>jzmdSoJDy@}PYr>-<|ClE>-qK@@_Q+ufZE-qcrHf%nOYmNOqa)x
z&(HDjoYNf^WCoqJP>ik;`QQ`Rl}(G1HOhZtKKs(Q<{i2ErG|4DownlKC`-XMKD8oM
zaKG)3cRdt8X5v}3*V_FI?`wN#^d;<{=G%gdjlVRn)HP6u5o#s*uBtm#HPk34U(KQA
zclPMfh1<7pKc<>;Y>W7<wgE)S&J$;x7c3<|jqSNTK0ZEFoIcIBV}fb+la~I6%d@NM
zIg=S?Xz_}oVnC%F;_OpVPiInMSUqdpr6#;MB9<{s*{N+9Uc*DVQ2}dgeF)evA1B!R
z@YHrmX_L^CsnWc7D!qPkafY|76r6>I!Fa==HveGVra~kwfDCW@B`k#HbtZoO5{cYR
zSoNUL)ngtm3bo7&!ACdahMvo`U2N$S<G{liSqgATk_(3_6O%G?$Ffj$>t537dhf@K
zzKfIW`c0JLnCgj7CwV>&et_vE!5&Ir(HF7#<ww=j(j9+(ast)G_uaH=soG41oX5cU
z6fB$#?*Y5HtUE8xaKZD#0?@zNC^2>z3DanbdWRPdzVefL^yuzlxJfFBV6+CFRsOqA
z0~prpK|_nR1w+A`i5DrEWDHwRfU{e9u|Q(1s@JcgFeNrWXvK~zGvz5oB|ithugQ`1
z)U7NuF*hwtJvS@T!|pY-rKKtv`-JWhlE}?gjE)<&p$=2HMzU_blvN9fi6muYr!JZy
ze?HY|VxZOTE`qOSnfLMIM*_M7Ac^2frV=K(9P2n!S(G(-YUz#nPipP%TzTNx)tz5h
z_4Lu1A6DVRA3m%Jf8L#_87s}y<yQRs`|e`t%-!1Q`+0*@);S3Fs~^+nvBlp+uwu=9
zu-$^oABb$rFF4%UmO6P>*t^&NUY|$ORkulC2|-_=5Wi)e|F)xJuWL$EVVuvW<GLZw
z)uUpW6#5NpQfX=2Ej@_$bmD=<9cz3r+W4l3TC)RgjI(NYLT{1A35gz%I~m3CscBy|
zFAs`}n2fxU)Sc0lqaiMLEAEwrp2NNvXm9bN33$A@c`uQ2Mn{Duw9gTBk5@oIzU8xK
z6($<Ov2VI&ue&-s_f5a^E~_XKy36cOO0|*f-AJ|`9P>9kIj+gy;<oROI{NC90Ra)$
zudhZ;c%&E>lAw=TLPW~0gx*iRQoRoEFr;EB#WFn$!zz*X8^DoeP3lrqtA?5rGd0C*
z%94IEpRT6U)G{3FnGuD~^9t0pG^VSY6-DUj@wD^+KRYVp%gSg8>Lrzg<<Uv6{1=2Q
zV|<cMZ>Fo|De90WuCV$m5~axplS$#vrB0td{f3c;4kE`j^YyVC?e4elCjuSec8HG?
zU3Y2E<l|QqEy!8T<!;#BIeg~KimO+z8k%aWtDCa|0@5DPPpOeV3Cy##WHZ_esBzhj
zOHBoITf<WGuRhB#T-wk(f^nbhS7D)~zC*%M#EUcYjjL&!YgCIKJtEqC%ZuBu6I)Ft
zr)ToN-weAi>spWFgbN6|QD#^DaYP0sle5@Z>+2v*&YUdwZZP=VxOY$u31M=&%CJ42
zH`(=1n#cuh-o5G9<b^ire=|A|{_D67>DJ{#p>q`D!mjZlKqnBR3Lw8ftf;6#?J6Gm
zf>UybN%c5>a{OcHlW<LO`5!%cbgRX!zkPw^<3Oi*kKa;OmPt_crrk=yahic0p*+Nw
zZlZg7dRi2=#XVHY=Qo(F;=$$-c~>!x=6GI7JZn|Td&9v#_PUd82^$hvbv@q4Pq_G>
zO)uZn*AV=odSbo(64yA|1&<hwqe@-_Z1%)2`}eUCdB-}~?B^|&Go$RI{VDM|n(MXv
zxJ2!!!-4G@n)Il#Q8w8k$;_EYGH<7;P0E{$wapXdUg1twT?LQC@?>D4<gIZw?P_PI
z1IHOoIm7_?QN2EPD|<aO0S`K$!zaug`M5C*#rud+d0*?XgMWfS^>A`llR^K{61(TG
zUM*e9&p-6*`+?Nd)Q>-ZE}NQ~GBi*7b>2l_xD!n|E+y^Om`?5%mUu|a_gy>oV<9*e
zmgdR4$Fln}``u;s*5vwZ%dkupN$HY0L&*}mB|6Kl@jjrQlT+zGzT&Ok5%&^1qH$h&
z72jd{{V3ZNnmn!Z+Mo0dSiZYR2j&w1&&4F;I=XPrz5=P@yt>&38)lCd+s5f8Bic9i
zmP!A-A|<p$dZ?t`y}Q5PP)5$l(UCZ_Ao%P+8vbDQq{<usA}}#evN^`bc0<B0sx&m9
zcf{8U{eIMWml_w?B|)R>6GJ!P6K<zVwh1Ktlxl1oJ$f|S<b^B|JZ-mJIV7i{f$$aZ
zNIeSC*+NcOPQLNU)EjAeGwpP;s%ev9tx#6W!CS#!wjKWN9-aEfIPd?=hm$>Q<$}E{
z=sSY9NSoM}7J2&&XYS_SyDm{)UP#xHdwq}9Pc!kObh^_{)s%fkHx0wD@wEysS^&1;
zhGSvAj5037#kJ=0n{hPJzAgFkl~b9j{Y!>Z83RNFhocp5klNU5u_2+TxcHJ`_QlFd
zExfZ^Ts?-_nUw7;M!bSio#jIY^B}E4TZ04fGUa^TQDiklZCqpf4L8l@<USePo#!g>
zibs+{MSwwin~W$wmSuEHXIa6cyp>?Sph-bwN3196;9YTg3KtopB{E@lliw!J1$A~k
zH8wcR*x=mt;lph`+c|U%5wiR2uvhQEPEtg_t~ZG$CX_H1=`MmfufKcnP*xFOCN3Wl
zNr_$rTo)`v@y&FWWm$K}byXrT42)r$z5i~>;I7HaZNdAQd%x0DMo;y)mE5*og%s&F
zTxBrh-I-{AN5^b`JUxhP6W@D;OuaUpHlJU%T4*kG3P0O;&^F#~Vy|^W>`{*Ahc|5<
z9~sEme(k5nlz(T#*(-hHS}AXyv9Gslxtf@nKB&_;ns-1}E_c6aRd9s&u#Zf(v1u!p
zi*u9m8ZjYq{p0IdniJhw^q0aeC}zmojY?ZoZ$}`E=6EgoxMJYMZZ{@H#jVc~u)*X>
z?esk;!}6?ji)^fZFb^KVk>=ch>R0A=SeDY}BA4^#W_Ixx5^G@)g96N44x5Ksw41je
zqXVa;uk|{6_Esr=)Tkj%Gc9-9(eqh+q<NwN_h_shtK>w3cf*xJx^*N;-95$aL=d|T
z@1GuqzubCns*A0{>ogTXR67QCF$Ir{s;iM%uyJwU;k#M$`MEi_GQX8<UbR%<ilj=?
zwQb|9kSA+-%r_)hD8$_>F1{&!q-;MT^^$WZq^RQS6`x_M0^J{hzWj}tAwP)n1T{aV
zTRTl07kdXEvu(9$YR$T*lum1avm3@8i0FwwZD_0wolQ&2@)Mq(uiaIFNP_q*PnmlD
zm4@_<0U~0%cvh>6*)Jz%g)xbV42LQUCsz>arTuyJk-I<Zlrz%4t;f9#x;j%~plwQ0
z8lY-T$r1N+LQo}%56+8FG+i@En-#xu!L6kHi%8B@Pt|Su6G;JSOZp6lUwvCz&rC5_
z?qNS~E8?(DO-r5W>RX;GtGwGZja_Wy9$KLJ^m=N+RjKHq#4M#ERx~dQL3>D=gUa>s
z?298KzJKc+m1DX5y4X=oae^@2R%Fia(V4lCk(#aKBtLxDpq^%AXXu}lwTij+f%b3I
z&I)mhZnxZ~63Z<!%y-_geukKe3|rr$a{9~}uahUwyOnHfwQk-LXJ5pZb*IeUblM#d
zOeW*abaFx%Da3ull|v|cZf8Hn{F%9&F!Hl^S?aa~m}?egEz#v63;m7iRtlJlQ1mir
zTH>PdYc0<U>Jq2TQo|5sbBt$Nuc`%ZBqzK`P9#D>AM2;1OaQwXqVB!zRAHcT((*$S
zE3G6AGs@V1_s*r#v36o&g7d;=zGwTV9<3#3stD{2ahYCYl;u*No^myHpnq%|aap4!
zc65W4M()(Bbll(j*ff_m8L)3+CGc4C4JutZtiEvUN21cwq=L`tOXv0@_xoAl78JMZ
zmDsHoigS4yQSfmsgE6|DqOU@>R^@a4R+dy{m=KiK^J6nIrtTMb0I^5>P_*m0*z&r<
zWG}3fqWWQ~ci_a+b?!5P6JH*j^cC$-FN?C--k0#d^X8w|%K4PPdGoM_pWy=KTb@!n
z?f*2>eRXWRUYs?yObd0%Ni^mm%Np<faISoaV>WPLQMKbcr!#v?=i4$gQs9^1TI+>g
z!YY};cMpeNMu{~XC$-XudT)b7#K#-C$`Su4_Q9WL_xc70+!r}vtRKWn#nr=0MNJ7S
zOSm$W*bx<KhPip4$HX~DnTPA0KiXHxl3QZuoty9ri^-7vv`n`yXDv-mT1O>;6=^aY
zBNNp#;I8_pC@(Q&*|2C6p~gkck8%{aQp1svCzjv955fJe^W+0@17z>W-OjXS0mSMz
z*lpqk8@?z94^(6mg04n*5vpK(@S2^;CJ_cU@d3QH*_7c~qr3Z?X45P9k#`NP64<Q2
z4&B6lR|I3AgV#-ETX>6=bx4Qm1Ebz|)#Y@JpB!$Q6vIE}+S>GBBv66~+R#mbf%;d5
z@+wej_(XH~4q7Io?uNH^v+TybcV(Tzjf0vRqrImUaqJG2*(l)`q#6jScsCAjS7^hi
z`Wr6GvRCsy%h%7p>_SB8ml4`0WOTL*inIOJq))thGW8-yM&5HKuz`5fFO_A3U(Z{7
zG=R+(GYsthJ~nIiengzjo_b(?H%~dzBbK(pJa<zzoonY|%~RJTOf1v3By2vrjhwVh
z*OIfD-u$F(MqZ|8fiqFzW*Yj&w&4|tULio$5<7xjIkk;nx(Cu)X_r9Pg958D+4+OH
zOCXl6)wm0<5wTmvJKORJuCH5(IQ|A|u0tCZU-HLGIg%zne#zp?$knhoB^lx_<sB9E
z*=D_qva7h19E&0_XBMB}BO8*nRBOU~(hUF{`7~9!3q%@MxrgAdFZz?8n}Q)yQuuDc
zb`<-EeyYEt<G@zie?~OvESEQKTk)&Rh5^M)K-PL~OzlxkgLjnbu3CUWfhWiu+=F+0
z8H+$LBRN<7AolVicjEKl!2_bKZEqKQ)!6v@*cLQA0#^?83mvHzj-~hXlAHz(Gpf4h
z+e)ZxvI%*C%a6Jkm^l&OiRH9^=|w`5g=Pf3WKi`E(v;`oVSAm)2U1E<hyQtHX--nc
zU@+L_nJbiw)%HK6U2O&9^vyR2FT%sb2e|>GQoMUHnRq-|b7hLX*3Nzs5wVk&PY0O}
zK7n9!Trke?^0=?h&s`r2AMvKQ|Iz+5cdPlelLl^;#GZ)td1d)vgI1FFaO%DdanD<h
zYYw#t+Ek<}Wq<89zW9huiOc-s*SwdWYgkYF)sRIu)=64MbxX{tm^jy|{%m%pGyTAs
z&83UPzCQ3cz+2nyj(Sweug<baYSUNw`r#5KUL{V2@`{wk-hH=Y-$*i<jMSVuZ<ML%
z$+8)JoIQ`Hz2C6bJbeB9c%BjWt}zqtBvePVe->8kxhX(2+ve9Bj?}(-^!%`e%sU}n
zgXCpovGd6_q$ayLq&sr=uJzSxxSk{*sBj!TereRn<uHVE{CBaMG{Z_EFV1Yy<%?8$
z&!nUg(?9L%;i6llwe|h7Qd$uYl>!G`r~P6}*=w*4CN8A&$9Qek!X#mNqmKD!<dF|K
zAtN?i^$2swcu|BUF?vUIB5*Mds-AU|6|e7#3`)N_tHiE4v0P{uv)SNXS!nGc(QeE!
z5}*TyLZdOuD3F_*tFEQ>yrm`1&@|p7C8$<kJkCm8!wjqlD9=|K)UT$mX%%!hLa};~
zdN{7#VYnpcly4A;&<AZD_p$5yX>h1GTCZ`M+o&fuG@-_Wr`Oq^<K-s759X?8?IhzP
z?yF2*bY}&lJoR%@Ti3Nv`|i!b<O4$&r$jjwH<z4!;JdJgZE`Ecd$@bRh9laFqWR$2
z;rs*JSXA1I1HX)0)(;1Wja-fNO0Y?w)54;i**O)rR_wIJtM->R5qojO*Pu?TYl~@H
zXULsh+&jF>R=IazD$|e@qw|ktUX~t!kQ<gxl*d+F3(1>k3)e8--<E!b)+la%T*HyL
zduv(dWI>a;b_Lt!omq<ukH>ha%rgyvYv(P&6!lO&p7v8{EzfmH8111mrP6xkGhU8V
z_^QP%)2-qWc&iW>-v8+Iw(ARx3g;l8*3S2lI+H~zhb$&57@3BZj@WMROhxd}DY4s^
zf88M5d#)U`?8dSx%$dyTa%2>Zy6m_uvsX*44!;LRf&PB^m5g_LLK^VZDZSEUz)I$Y
zZTtLB=eIol{(U{rK4KUh!)_sFdA(azOiWv!dnMf>ll`U*y36XK#{5?k<AK1Gho158
z_*g3uTtnazP%v5NXE#0X2?|UZ-0R|9lDg~s6j=m-O{Jr!`*LYr$BTcyhbAhev&yjW
zHtyE9@(t+14YR5JmoR|^i-@lAxEo3|N^%3~-S<+vPde6HLo-8#J`xSCikgbo{&Zs$
zCO}4F8jc_hwsqFL`B~*v`bEUO>YPDv_nKYPo7)7A%TBBGUXwX-Jyo+SEn>%{OPrFC
zsiSFFe!-dN-ox}o@zQoXe8W>oN{wi*ic=o+GG!&%`>5x%BR84&Z7%zEW+?FE=iC!6
zohI#zM(4T0h=XAc-mCO`Q4Ps~E9Qx*z`}0zptb9a$|jXg2rS!u^L65H#RJ;u5=Q$S
z#~)|%)v+4)aPN{4u(OPF>2uJ<yM^B>y5XCz&wF$@_a!pek9gLWSJX&v(>1WGqo^p%
zduW+=FgREO3|01udPC8wAI1FRJpR<11(||e^KdF2ov!0r(XMA6PYIf{H8ivpDKn41
z(eh(E$1U=C??Mig$UCzwP>o@aNZjQ6@2BG)^1bhb-R~_+MHuWrjsIf`m?ywbXja{x
zemXugyW;f;OY_$Kv9Yu3e$8u*$O=<SF(G6GNYjx15KMuonb{DrjNQQ@qw|DfpQCiG
z$5b_~i>$HZtI$gJpA9TK&yI0Z>#G}n4K&_nkDJv<_#Gj&srMZ<U_3CUQ{$NI<Tq|5
z*fIeZbHlNOahH~b6O}pljFE?9dSj}>*KkV*Ca6b&$ys#yX?aSL_=l+hDt$$M==xo6
zyH~erjj)(bryf5g&-F_30(7TK|AiZtq*grKmK&r@n^#NI9nSL9EY8xDm)mPa3du_5
zLPZ^>W8$|LNQEKUPp9tjb%W<r?$7E`R8M{57^@Y5=D`rrsIStFp^s%>7?VjGnb0I&
zVxkgf#BjN$L3@afexSUY!($P@q>G%A(jIkOghqrwU}PNH_VqqO@9JHqsk+Ho=oJ&^
zlJL%j@A{$zNTuI?{J^xz)=#FTfU_nuJLszA$Jp;Ag+?(AE*LMPZyEGKN_l(K%|T9z
zgrCXa^3e-cw&AK#&c5sCp9f}$Xa@;gr2#rP`Nr}r%~ea|QkA?D3=UHd%a3my)~ei1
za-ldF#D^uRJUP`#+s*9k!X`d~H^9sC^7juXYG%Aa!mFdQQvX56^~tFUkKV0)3H3fa
ze!DYv6rXfR2;4s`l`}n9=eIX4sn2y5G-y{fzAfqZ0L@kJ;NTnyjO6O38(E<|-)9+@
zp?9mSO=atG5u#>J+cWr?y~aNMVQ=~1TeobstF^Mc@?@2ZM!3q4yMr6ooKtDDIKHk^
zsy8(|N6(b|h=<W4Hd|I9Ggb2d`2b0FN3HYl@TG`0gsW>;<2UQ0yS+_*{N}OAmy{P`
zGxf~E!bf5UF3kjNcnVdWGpDkY47L|@kLsJoH)1~$>tJ{tmLl)lW(WM%lJO=#_db*K
z-FCdIi{k%^y)!p$W4XtrY9K}WmT-c%<aI<6yXO(0)S`kM=hmo)3->N&-$oQyf|r5B
z^Iiqg5f;oa<mpLoPc+^Shgm@3a}@WR0c@1cXFL0KdZliXv>jZzs@7thmsqzaD6^6C
z%*7l7yEzaS%6z!_pmr9*4%Q)?6(ago%kylb5~f!E`c(Vl3ER=MI$bCuwSas2gH2E{
zDrq#RUS3`o0O=B>J8{j364%cUT{9A$qyD?we~dY9@d{Z8N8b4@y<=gLm2q8WY?qGT
zY0}eZ`>X7UMClo$^j##wKB1#}=B-Z0Ra4w7BaO>r$8o~SXDscA`X_YllO2Nh){QrN
zJ0T;s??}5ec=F!UcdE0tQz(A6nwx$IDR)Yl3y<5qH!T-y{C@9kzxC77ELmlv`$3uC
zYMbcJ9<;(*qLm@Zvi8ojP2W#P*&`@z@OVp)347p0UwO*CG2^8bb-S<xv}J8+TBZ`G
zO8LP8{&$Hv6F1zaM_ju*vhG*7oeHz#av|5huKw}l;LrXA3v~_rAqBqlWcryk`5~RM
z4f{^~J|y0#S=b&Dvqf@x4^Rn(xHo=gX79d@`77*u(BgLI5i?J3vR~>PYxozmxl5wz
zYgTAoR|CPsdtOSO#U;+6pl@q!Eyl-fxKnr{A2aCi{v8hlk<NB0+aZu-G5K?{em!km
zsjx_*aMqaJs~9Zd?Ib(yO{edVc<)?4YleO_ddEyCxau{*e`<rTNd*0#hq}ekJidC#
zMCFGsUA@e(XOliW6CDZD>P1ln*!|2rKQ-W-XC?mVL|<FIY3}CucqL}N%EayjT8~!J
zy_>RI2AWie+u|rID3VopBDvuRP`yLRBs6eSwLUQgdY*Q_W@}l^^of-|(HS#I1ED3f
znXGm(qpZ}5k2jsHnG80&hyJkMzfD<I`R=RBS+YBT&A!+|_EXUZ&_4W2d$ml6`_{0@
zGqiGLE{>|j;t%`J6nXj$oT$&&t#Mu*o8V$@#m&<*;9C1~Z&!Je`=zwE(s?E-atOYz
zJDs|NHj2u9+V!df*4Vxd@P5N8aU~J8?sNF<VyWJ(WP+$PJSZf&H$);fu#OcoV2EaL
z>D9U;#r@hlxea_ZZ1ubak``#}HXDiySfVwr45OJCDWa2Ox5<7QJHD-)E#H(XeX5>f
zOuql6@nlKpOF3pfQ84`c#7WSzb7z0>(GOy$u770|CVHZ1)sY^%UrEsBzhG0g8P-!e
zzY4jqladC?IbNj8Gt1QYcFD+L8N7@nv_a^Y?{|FPl!w`E;y>-y%WR3bS76ppJo37M
zor4)Jx=wRIEJ!{eUY2?Hf;CYTE_tu_MDY0AKeJaGJU{gx^dBa@nR>_7Gg{9a_p~q#
zx=Y_LCEF+^ZI&b7NY7GQ@Kr)yR!Fz>iuAm*KT4C%eNO)r?DI3Xwe#BkXg;2PsbTgo
zllqn3x=^y|^n{}f?;bPXj$iW*>DHajknNtT;)i-b&;fDkB2=B=|MIl23gJBXOF$&~
z?c29o_dDlwAK8{?VY<j@SGt_0kyxs-==`Q6n_#n$_0y9yX{L?B?W*)T*6^&HROWl1
zuQYBMEno5IFjne8p13!PSkKe7-?f=z&7ZEA9>8pIC*LA}9NR)Hbs8D*A%IRaYY2ga
zcC)Z;<)g0n=#Fzv3M`w5)B<H!tun9W%NXRZ-4%ZZw?%D^zmKwk&@1XPM$IU+wlN>Q
z;&!Wx-lA7-2V=)Bu;Kjk%#nN#?HvT&SGc?^IKIjYKVE+FQf|(c-nGR;RbEwD)MhWw
z>rvJlB>o31Y;nC2jV2MlWmnk?dS{wYTrNr8Eb1-g{v}c~eO`EdNDTcM?d_Lg-_LaF
z@D&x0of{i2Nlt#O*x<zKb)ETuwn%+dX1n|Rhuebz-|1$Y2ZD!Rfg(GuL-05(dc}}G
zJkuad3&9xRC$Dzo?n}e{i~pIbq3zfcE9-k<LjF{r<v+BHtSSg7Dr*a7&wP92Y%FX>
zj3(r<O~}L=0s>|>>2w<<9|S>14#Qvdw-q7+HPo;~O}zW&1@JLwSAPZk5hu~Y&TUG1
zmDbpM$Tog5aJj2rNAMcyY2%#r2jVY#mlZU{U`kamL58dmE`^XpITEPP(>`bCY$4y^
zZ&yrL44a6)qq~!+iB5q*v<iN^#iYc}f+~$8b5H4b!+K->uHQ$E<a}3;lSamd&W@kj
zW7k7DxRUZp%;^0#4ZkX_6=&Fo72c_8_<8vj?$Gk%)Ei))cXlvM6OJ1ztWAv1()9a>
zy_-i;URG!R<F~g?ESYz{a3$-Ztu*WK0d8K|g(oQoT|4&X^GJ3y%o!Ua^8=&#+uVl?
zl;2Y_B1`7D!gLPZr`=zR@xQd3r>7Y+2<Z<KQ-7(`E#i@PxSnc%*hN}EK4iHetzsdH
z+s!7!&9Vy?SqPh~AK%K89I-ore)ZX>#8Ed<ZX9~|j-0FtDfWAxfB>w<*=IM5sYs%d
z@$$;}&@4pvEuGesQOW%iRDUFyq{I?8igI>}%M!5Nw9_&k{jpm~vY{!gpgM~|3&q_n
zFv6NPBQU89hVcuk^bzu_wgpGX_G!LS0z_a?T^Oj6?hXJ{s|}Eqy$faO2jx@m{YKLF
zx(6-ZfAwz^F4poM-kfb|>iCnu_-wt0z40e22NGy2_V-#=C)5hY)vhrRJ!@y#N~6om
zAv$ntWO!KiUU@&X=Z+EA5XXsl&RlFlVeMy;$j;?;^WA6HZjz)t!gIU%+Cb)-s`wI=
z)AD`0xEmH<o)0Hi*W~`K09hAH-CrJ|kNPpr@a>OrY-m6?2yW9*Y!x5(zzOgEMAC`@
z*873<U~;gZ-Gq}}>+#)|w;&mdyTQG4$%&C|)%NT)45Yr-$39}Z?O*W~fv_km6FnK=
z9vS&P7NZKmU1Ot^3Hz$}t`=j`%_@>~!`Eu~7iF5u-5Pc7d+@cX=p1~Y2vL9?wgltl
zcAgl&kX7&l0f2z3mT+_OabGGr>V(J1({3!SZ#&7aV!d2n07zG!Rd?10v9kU%L(X57
zZ)Ew-R-{)u*`Hag7Ir^JHunHKwfM~)A$qf-Y*)3u#zYxaPS#B8+a?W-7nMGmr%w)!
ztlTDMyn-&TKt$Y-_~93b1H!>x<GBMTxcy{cZ_9n0a^9eZ?XGGBz>R9BXBj{5hkxbW
zcwXUo(g~~#yE~bdc{gm7o_eMFs--N`J)&zCTH>)X+mf8UF#MVlRfQwYR9!ss&2cs>
zDL%D;RnOqJm7wM*g{4T9KKCk!UqoWTToYGRkly&})laww&0)`T8X-eX_Xi!Jna2nc
z!vCT-GK+RPM&_``OUG?hk3Yl2U|VKtghLxfN^r9J+v%gO(uZtoT*y`^rg5K~^H|<@
z6Sf?QrH=)U(p~ywnLZoOx1B&3r3fw4u(e(tJ982f)Ik0Wvzy9O62+^AoRTwBm&UGL
zeaBw=f1un=%^Z_c|J=)35ub8QLiPc3lrr{?a$Q4HI9irJS(rYNb#WPE!C&hA(JqdU
z124p6To@{ppW5M4CDOl7s_Rju`*oHrbVV^h0qR|2PVW4tNNF*Y;56oAhSJ%zTH&R-
z2K_NlwmsXpan(}PF8io?#(F{%yW_27Sr-DoH0U~+=up}L=g|#i03VK>;XL0qyW)mf
zBF7*Xwo3u6`3H#zqGu!`f&&L8q<8bRIqxkTsFoZ#&T%SZgY5PwPHdYdZZ@<-Aguqa
zmN^hvmGuZbi*B4|A=QdPHX1~~Ivi(E2c&brpMeb>{nL;w=xs=<lOnTJ!kjFBCzTQ%
zWx<+2sdVimeh%0+7-kE~^k|&zIvbW%p)tNTT84bWbaC5%>a3t7pBI7;Oij}~8=p%|
zGYm^jOdahz72|2SRxv8C{gdw(Q0b9VdOe%GIY{d$*BZAk^$L#Jpos9-Pd=HD^U^D-
z;qzPpywg#D!RNiImO#mY`i(IMLY7fUl=d@X2jjCtu!#!u@Hi-+)q%leTcUZ`%OM$(
z#jRDcpZf>9#C}6Amx+M4^uDxSjp1aF<Txlu)K4%7W}e@I1@@SUQn^)iZex-}62~D?
zZeR#=!}SE_`Y(2hJYQVA8GtI8_p;Q2sv|lY2`qUHtWE?aKbILf=AdEY=BzB^(kLc#
z$x})-DZTNpQ$v=EgLP|<3)x{12UM5g4$lZ)R(YzDJgfYnlF-E1@>+HpwVSA78TJ@s
zM5aHmA>k6@Rr??gs%9f#&%Mv4lnpIEJPwEs>v*xm{nNWdC)1fsPLWr?>X@79$$~l)
zTkB9--Lc>6B;FxK6tdYt%FW^aYvooZ-}*E0SyV%lK+r{>X(r|19p=d%N*>-lY$l%j
zyK%$J!avzJcT^7M#Oy!M$M`C5n&WzV;Z7B3G9n$nbAayegfJt1IBwoi4h+LF*HUjo
z7B7DRP&}DLv*7dvhwQbASO?qUCci1S^8|Efht*S4^UdmhNjAZl_$mBCz^H&d!+I1B
zL|z#B)OxKcDBXGF=tfbjkAk*4?yx`LX`)s8{W*FL>w;FAP~!ahWzAfmqVcOlYP<_2
zbg2gv(EwMepV(*iO1tQm#@4=?=U-o0;D2+r5@R^ne}~_l+|IDW?hS=NTXaMx62?z4
z{VXj-{A{_7Kx|N*@P-yollE)rRUyK{>#v`!q~x&0;^Uxg?aYj}O47^Z14&EEpYe2+
z7WS$i9!U@LB$C$9M_O5w-O4@@jb5d#GTtg1j~|eeZM5yNw6*Tu*u$1hs+dSOHd`Zh
zW~RtJ8s%0pYuz;oxrGt;lq7g1rCT<d`1wp!dIn!k-yiLD*r#2Jp^2DIJvf?u#APG>
z(uR;2k3(C;u~H<X#Z5B3QPGyHx6HwYG*@kL!@Yo+qq7X+9R>M4daGkSc-pNg|CG-)
zU<55oOr-|l*gQ=cC{MjHIzLkmtC@8`&Y|A|!bnt8e*p^x+X)czkiON7w*^JZlj}id
zWy?imL8FOLUPH?qJ+sot+*f6jWjeERy}lY(Ln~6HnV0A;wpx2uB>2js`M~~GC~Rd^
z28J2mnjtLK@SkLi-sHY?!T&C)v6a+eO4L^q%gDBq^(wX?`mk$C`@!`~o<~eAKKz;b
z^yu-{`Dlu-mCMU2S4A!OX`uhYXS-5qd3LdqhF_6Oqm_fW{s1I~kylNN23JNeAF@IC
zBAJnptm<0P5jSLQ{q{MbYbFrw<DhckBY8>$*eth=`rH>O_x8j$Q|GNfz7)X~7S=bk
zn;h+Hwy}e}<)Vu%qMqTR*e53ml#t?TRqJIAZv4T6TuW2mj2z4?OE<Tr9WyNDcX*R9
zwF=F#oG_gOD%1L&;qsJZH%L?w?y#ik-&FrMAP{4&r~=S#8JY)uKg@`F?}L-XV!Zj0
zHA==D%e2KW9p-V!+_uI#295XF##N94S8L?faRrd=;<Yhu-EVgzS2}`@eqbMUc#EH{
zqmYhml#rLLwka)M#x*IQDu6Rud~<O<=#I<%9A{|j&YU^ZvdEI{2kvz9x45W)C7ks8
z=No4a$kYD6ilaH^^gmVA-;itf+CNn<U-=2$%H@4Pt@HX6RIKuHvO+7tbS(jYYCrR2
z?96Zeswwf{&qu3m&g`v=mm+WWS{1z{weHyschv}E(z7QN-8ySY*8@0>j)hnjBqL`e
zzQ4>K_6fjq2o_b*0D6>b+I>ohe!dP?`TL+XkeLou-wVk5sW&!Dt*WKRF<aLT2cnCR
zVjU?~D6+OT1D!3Gz^Zt)<3dohIvKv^IJrGN*@mVw1Z>kt41Q%d#XZ}|E_0brzTis?
zVxye~5Qq<&YCLhlHaeYo+}YBA`$kQVF=_L!<k0}i)4&Z0d#A+&;?}dog~DXKY|V%)
zXK0RM0l|Sy@iR+^@9~k}-yh@Ob5ql&dB4JEjaY)zin4FqoOF#!Yzn6`2FeFc9CkyJ
zQy?dei@F{T(wK-?YNP&dD7ZL$Jcjn8D>L<vy?lN=QN5yT_~>XJq@XoXe8%1Q3aRmW
z7N0o)Q+F|5!D}28)tm^uklFE`>Kj?c@fZ2tvu-fi9ATPV)IeTI%F39Y=Tp<#@Xbz+
zkxVf#q^?!g6+$lKC4W^CUWDA`#?VL9y?D(77SVvQ7th%&M+a&%d_%KZ`%Zybzy@+`
z77&V^Q^dN79U5L1c=1x*dC<)~^JQdbi2yq@^<xMD_T2BE^FR<1!2*Nx)m^=Ky3aPg
zXo4Jt@ea^kvgRIgQiXNS^sA#PVq+6*HhlD6kUsEV-L-Wh2-VEKr<CT)-Qa$67Bdft
zMg>B0|3WZc|A@#FIH2TvXKN;jk{UY+!1rE31Z#2}XE~(8nna`fl)(waiaPw{$vl(~
zza@X>R@4VXY&oR)n{xMN>+W*>R~kkpVX1sC-^B~pG0m*>H2g|Y#wwRDZ>G|hIM}Um
zh6M7&HnSXA0*H2B&Fl>pbS~az#P$1TtIWl-9BL^L0Mx}xQio9(;Bjq!(=#uof^<m>
zDRxO+nR2+)@2VZ~Bu`Ope2rcmatV!hC4Q?f^rMuJlAfuE{9f+y-8W<V$HUx33Cho`
zZWve^igZW`!I&Ew$5&sH3pdZR`l$XYQS`QvSD_fhZVJLupIZie3EGt)=FW5lUL}Y*
z;=iHl?jI{p{<*xiNHB%Dx~Y*a{R8??VA#%{p8Ro!`q}cmBYrbW1yvSQR&&0;{<+6f
z-^}_|kGHeXh)Z`khPRrkDt3tSVSh($mUC=85D!a6+U@sPCp0Cc%o@Vemr+F9vC~Gx
z)z*o3U6ceC1^<Y^QTAK%b=;LvZa|d=q@=!W+u_->yQ}!m+3)EL+uwYgL-eeJHw3~a
zx28&%%#rV957({O8$0xVX~dlqGFjIBThsn{do57D|1Igjjj3lsmW(PfA=moP#x(la
z^;{Xq-RfFn*XmG>&GOd{^bEX^pXnYnYwP4=xP;!2DsDDuUey7K%e&}H$O#@g={Gis
z^1X=vy?_JWFCD6XRljBSjy@4M_}Zid%@t&r{>cm=k8_wahjF;E7qVf!EXB`2998=+
zh@ub^CJ%&t7R7rA2nayPaHB#R)n8?hMyqI@H-wFg@O1L@gnSyI8X_ekbO8AJkG=(S
zGLThqUwUGrA}6+|Y4EQe0P096mw)~GWq!vR5<V4KZ|xP7r&=6wIZC}*AngVh%JkWU
zz_5mHuEeY>zlZLZ(Z}${jP8tOjf;k3JuN74^f`yIi52(YbBZ=?9TXpSM7ptrZ-Mjl
zClrXCot;}BY}BgoOm%e{s!}rfkJQC1`FfF6J#(@4f2Yo-!JYQM=h3j8J!LnRe%Zx2
z{GO&3*-t$?id<~FB5uO!{4z$Iv_TBVM%CMk^{Y;#T3C`6@8Q<r(6rGtGULCQWz5sQ
zoBdu@JQ9phj5P1?gyy2l^TDG+wKOMx(91CYDEKoZ#tKM|_*>ejq*iAkiU};%X?K4*
z9vFAduufI5cGJ_1(1mf_&?^C5LQqTEP)k7U1WB5pzD%WJdgD6CYXYQWyG!+J;sUB%
zyUdnxLGe#;B<P0c2(7<XBSe>sS{jH0-c3u5a;33KRKW+tsX3v|ZEBoRwjd^#U6bi_
z-U(e&W5Z;<y43mz;Z?Omm)3_%$q2IA(2&3iU|L=vPR7Gl&m}fN|48fJWPQh|Oaqzq
zoaTA;^BXhJDmw}Z=4{q5l@T&MULwA#)|RN817`VBvW==@_AkqRIf|1$UXm^Qzr}zS
zrjExHsq!5>8%OneY^LNoM9pO62QD%kL4>-?@SeArPf^QTq5pGEG`9A@l~6w^`2o&v
z)C$>48MI5%h#DRSjLaWX_yGy2jz|RNe6cn(z44}X{u4&Q9U&?pysO}`fT2NRJf~Pz
zi=QugMH2Sq!usE4WAA|J_&Y(4Uts`=@gG5DY@Y!$i!*r-GU|<d7I{`sj6s?Na-{oK
zXnjX$R(SpNM?^Pl$wGaNb4~>G)U>p2%pwq+6oAN^uXmybeAM|x@2c1)6Cn&9;Dh!E
zpBA^}J>Ft-d-o{C?RRDz)=?blCQQua<Bw7x(8nsqC}$ZtR&O~)Nw7FAQ7MI|eo}<%
z$y;p%n{aLaChQ$BZ|o;Zy|yc(7d9vH0YXW3*IKm{!bJY&Epu$~haFC5gzByl6o~)d
zYVTh6D0_6)w9uyCVXmoO5<|<4Jr_^YH*eMYoKe~!Ek6)Ha$ne`9th?7E3Wk^FLt=5
zcJFxfTjFOz)XZpkif$>7gQW=N<ZeI|?MA5NPM`8)jt2sN1uZB;0%d==h^M2pmij%-
zaJ6(uj0s%sD8;tu3Lv8e@zyk@7YY{_U{_2Qq>kS-_jn~`EibDTC{p5kR@5~H;xS|z
zf39XrYztlwGDn_c$F@UcN*Tn$;*pTR5~R}KPK8LG@Kcy0mC${Ij2F~_FMXpx^o{f#
zC?4#JZbFI+q{KQuPZ%4+;^-N<)q=Vr5vk95ZvLt4@zx*?ZreO92V2$Hh#pP3%p&c1
z#&<VMDL=4Vk9ExphZ$*43l=wN778MYuQxawC9PnT%y@rvm4w7z%d|_cHu?wU(u*Pw
z)nA!BdUsu`>7>j52vkaXy*)e3dcM5kA4c{41^mCiEnXSz_ng-8@R;*^GL$wZ+;keR
zWTCLeYY)rXaE|Kn?}a}{`Zt+=13sc#${}_#_DQ4t1~t`xHW)e9KPECb(I7A5!U@BT
zom=FLIDj45C0ysAthzW{YG?>vPzE;uP<+kaq=)fP5I|ao%7h5ZggGy>9xb{R>A;2e
z1F%kmfi6OASR$2FAZvF-MGS+QMAXeFjf9ob70Kj=9O?kVIgUA@j`8=uT?(k<{f7?<
z8MYrMK1upJS#b40BmM3B_fHsE!2rYC&Q5(O_A<=RyC(n95M=D8>?<O>qFdh@>qPSF
zW7bP;+GCM*SC-b6fXjv+ac^ri7)j{I49}uIRPadZU4UApBdU(6vS|zATmpeuCFX>k
z`SD>$?Ax_-5?f9REg;~s`2_g>BVlD-%22getpb~!K;RtlGP26KHTzm01rbnjj;?8t
zo0*n{pUOVJEq-so$7=TM6o~IxM+6x1d}(N27>?WT9FnO1^Xtdw9Wy?`ew`+g2hOK)
zk-}~OC0AN*%<c5>rL-h_>)JldwFfA<IttSC9}AGNO_~EmBoBEB)sU?o)xS77C(M4s
zLbT~Z&>G@{#_)nptP97<Ve$5D06Os~lWXKVAUE?)%;Mr4r)kloFSytqKQZkd`j!+t
z-i!hh`Z)w~B7Rgl?m-!jgo($Fw@1L65Ev5W4VVr*EdDHwUtxo>+L^{~G@lb?_AiJm
zIzA4ObLgc@oTzPX+TFRmT$AJ0XCBD`{h9c&k=&q$m@F6DBq)k+9lv_+jZgw~`|qS(
zVGRlN0IhFiB^HD1YpSlR$MuZd0adIdMFX3N0pX-Q3(rbzID%bDC>-us{{=-}Oq8eW
zT=V4YPp18zJCUa%QhnmfefyYwFhiNE$jQ*|cZMGwZ`u`=D<j(#d_wbQZ0w~(*0{0A
zQ@{sxin~mGQ{J>l&3>HR{%2uVU}W`Bu~PxM%Ed7{A_8=Mz1RyY#3+aC7USV>L;xV*
z;tV@CwW!ZU%#~&KOF?;ADd22C+gcB5#45Jo$e|8ZsFv^@=jNGE1^|;n+{GGs8uJ!?
zCpp!nI>ue`uZK6<&Z!Jcjf0hwF51W!caWqKj8pqr-N<BL>+F}R`;C$`v(M_*haFbc
zy2-orO@S=O!S9CI0g+IN1tA!iBp^U{Ra`Fr$`5Zw<~0g@uB-F~?x+bE39@qtz%OBJ
z1WedL4CbZvMovo<N0aNmWGe4UZ|uda)>~;2fq0oiwt@*@%`dJM#n#6yE&_4euSZdn
zkcZB7?Js0*C8zIvHnL%^j$>2GsxaH^*wx-nZjgR$@^R&?d#8!bBG<vO|DRw_XEG;y
zX-|hMAG+UatThY(+K_6@;}=&)_M~mIXGc|!DbF7NQ{jDWhf}~KfKha*OR|$U!ubjY
z!fwzlkVJU{`_7Ob&mUY9GQb?c;YJR_1VyPwVOa~LArk(xW|p7Uvn++$Ai8Gd)qD3$
zHuZ$KucMw`sV@NO6MhV*2szT>#2-_r8}R|{L;Q#EBDd~dXYzApSDbrG4|>}dEHhV-
zybEu7#56ZO(JKo21ZsjNpLphkdlV7?s8mR^UoEir=ye^Pa<gS@L%g$5%CyeM<Qzt+
z2#DM=4<q4*%#f@^=KY>?+WesJvtrOey~@B~JY+;BR|Lpq&3$rVws$q)rT`LMWNV5^
z+&!h<Q%N@9kwUVT`*ABW7@8?{y(8)v$;#gu+w0f$)|9Mq=RXDH|EZP!Pse}UV3q2B
zYIkLiC+`DH!|Okj>(BHtCsnPRl*skbGQe5uTB*`I54&6UM!X+ldCqFjJn@|Q5u+xu
zox{*^;pE5MnU8z8v?kO=`Q9*Ips0O^g&g4NCeE!v`#ra!kSe6#pJH4|u!54kMY#1_
zm+@ezMFJyUKsmE#FCVItdBL=nn*Vc-@d8WboEd~I1Q$6Pj_f68qXxlS#0M7y#mF2%
zj;GXA6XBMg<wGK84t7fO+WtzqG7}ieAuFLzU3?MKTwqu*8H5p`MUYPsfeINVc#2x_
zGOVESrtD*^eyjB&%BJb*z;A<t|B@9Wn(UDYU$ul~jDw?Lhd3}mLM4bTba{da`c@)h
zAh-z`<YDEDjBjnYQ|_w!o-zy?po8FLSuT`?K9GmZIjpHmny$N@4naz($LHU_ZDbT_
zy}aG)SG5^^TL1Nr_wb6<_f9sg_kLyTX2{jkw)S5J#?pE{0m<G$x#Oj*w36R4ZqEz4
z;@d4`TeV~fA-dQWL!NeTYNM=1+85Qx;P%SrKLV!||47Wn&5j`lTh_79msWAzsbfmr
zG}lA^Jwlln*?klHDp3-e1G!UQ%0Y}&;>1}gY-+d!%4W{{5h?<O%Vr@fDwXA2M=XTz
z6qST1Ep_;|Ji}=4s^0JZQ`<m30YgAFh}Qu2h8N>qyq9%9J^A9d2E!08wn!uG?x(-4
zasO~S{8$dlV{%F*w6F~n1)3~KU<}#rbK|`Phjo@V4%-(b_rEy<AX#$57aNsax(Rv0
z4<j9@?gz>V9eax#dQf6q^rW{pG1vrv2pmX|v}7R{FYmHdwNFseLcWRh;Fe+1?<2C5
zdcFNB=Tbx7mVem}cU6Ha8UAUqZN<cbG*eO~UX5f*NR`3`<6kU_3Vv*S@!Hzj#ZDzh
z4oqnO^jhuP1|>n{{{z9*f6&bupM3c+#{C$%ey97fbh9<;5T0LiPNUPt3f1~nVbE&v
zx_JkI<T_87na=W5TWCCi{q2~q+T|?1z`l0u*y8i$P+;ooxsdLn71=DcAgsa%G&Xqi
z_`x_GWN1h~S~yRUaPyk391;a$hA?I!OjIqv7T{?hY_3rY1Oa_O$^3<quKGd2JvR*=
zQ{7<X4AR^1NgUG<|3`_P37LiKGAAk@t1_sU%G(xoGsHz8+@~zHUrn?1i~&`z<T#>O
zm<eU`#CCd(so2{#*_BYMqJ;(^Ol+Ki)%I<M`*2{y@7C|9Yq?yekLst%&T&#8nQ(2o
z@ui5#?EOxGmuWPiRjXDZs@-OKqT&w=Wic+AA;FyM0^rj`9g}v!m=<{}aiv*MpDJb8
z-dzBp_|YKT6T~@(qeP?Nn_0LM(&IN;%?Jf~R!={C_&^YFAJks8c7|lfT-k0x)KY4H
zhkDlJ)tx<Y^||$fQ$#IQ`Yq|UQQ@&^_r;DCy_8!;X=w%T4}Z?^)ZD$4zoCck#houq
zE0;X9f2GM>%%&sCu?ZurIjG(A=jOt4<Pu>?PS)hhQ(6r!K&^zemRjNUZ2y~rG1!Dz
zLtcIAebL$?yVDNwXB&IGh2wZ7^VZ>M%!v*Ef|%$dWNhD+-ubZNlRkA4#S~RK<FSEd
z^q(~>EsMFG6Z)7AgF@J&16?2{;jHuO@sc2FBiq(ch~?60lPh5r1jwD)D7AiobLh&U
zYt!q)uDu7QV_hL|2$&fJ;=ZZIn%~F#-I`^3z$L@$rb!*D6-el6>!k}mom`y2<0y#$
z4AuGHd0iazvs1s;M@2^y6G-SVbu%7gk}Xu1bSIH;cSUpn^MbDnPbLi1CL3qUc1Jt}
zgn`)wrW}IwYUQ?teYMQvjPhjrS>i?Qlf1|Xa-t;XB>!WZ-=bdRx%BUxPJYdQHwic(
zE8F;>w>vt>RtD1UguUGg0}i{oe~MrX>yu|+5t~xt+Q)2!lwVm}&b^*Q+R_%kHpG2B
z?|HV*kv}Bgf9h7hmV2RGJ-vLUx$C!DTc<G962O~|E%!5*wW-AI0bUs{6$gL4tA<UP
zS?sV1CH`hGDeFV&^728@ZGW=JV&i6ou+Z{B(FY;!dr^Wre|l2A=Gc~Q%Wlf?7(F9_
z8pI{I0T7@N7!DJxAx4yg#(%mJszWY#CRkCuL7gz`8BFSnrs$Y?6Yi?{>_f41ePgr&
z7_5O)URZsTb?iTsF6=(<8b>9J&%Q;uuo>US6Z`p;MT%x*m2Rvly2WZC_#PlrKmPPd
z64SOt4<1~_WHG)Vi=kc=xolXDD#y!#qC89*cQ|_;!W+$O^ps24krK$Q<Rfk<6*31@
z#CB8T^=;yhU$9^Kwi;Lx<CwJZ(c$)fkjdKV1Vxoz^6E)I@FO6h?-$(ma0_Dp-xJRq
zX6jUxa79{n;K>Uq*@(n~#r2cD&g`E*Z~xkuIWZJTTpIRB-mABXC&XV*>>1<#Jw4ke
z@%hXw+w<-td1)Err5F1zGWd3W<rE-bErCtobRx4&`dD|tbhP>IJDfRCPB#IisGC*8
zoTo#+b_ZpZNJRgt{g`<yY$K*s=4w;#viEN-yXiJ{s#4W$Qm60(ZxLStzfv%PJk!H!
zH^+qlQvr=-vT0OCX#J#{8n(xtV#|fo4M^U0#Nr}z-j`Bx5foJ{%96Uza~$v>Y7w1f
z#?uF`8~c!&$y7aNXZc609ftIe08TI>dGqEvbj!}c!A<}(q}2fIFDN5@4ctQh!+((j
zqu>e}ws-ldnV8>saaRnf-hq>y+zOv$CsR^I9cjF3l1s$Fk&t;hUNop@^o;nyKLBhc
zNC(4upC+;0m|lAtBl|5y&PD9MPr2-slI1iKQdn_zl%+T)$DQj!l)DC!b_sEW4PAB|
za;7rU;{RABgY<RZN}$&{@}5e;z0tD=Pyt`;dN<7^5`TEs#roaE(9>Yd$U2U#>Wkkm
zsW;8JW*5HzaDuxYGzB`OJ1#0>B`<H!3Eyy9p^8utD4iPv&L3HGQu+2*!Nrmt*GBpE
zz7iQ<HHDw;OrWo&{Su*}1wW1^lYrd#JCXQ8Zz&<|$V9*)F#d+TYd!U%*p&>sX&sUS
zS+F=O9h%oYv(^)8nma)9nEXcV{I;vBc4%l5v2q-3@u%+{4{-2zmf7drKvge*KP5B5
zs!6~NB_%L@capMEUHJ|s#G#A2VkfZct}OMbQoeIt&59(hb&ZS1VmP%9CABR@CU-Ja
zO$m&-wMsvhN*CkVwDPZ*usmhx@R~oKp5`kqc|%M%{qDcU>i<m?f|iP7;$}f*4==8;
z`;}1ou_wS(!`)a@HSKiRp5ODy6Y>_e!Ng%t<Km^V56xF>UllX~q=P?rr0YsT1BgWF
z%c16(xtF|8g4uq7mDdm~XJ4uJ{x0={xfbfXJsIXUXMA`i*+AoVF@2~a%DUNe9;7mg
zM^%!F_Nhx3#l^Hi?tvCs;fc*854uB$xtL5^VM<OOm6pvfIfX_R!G)yF>`Ka7hecZG
z__CBT8Ci$xLli!q_{S-$2Tl-zjm*O_HOG@a5(Wy-p6w#W_3?1ejt{OjH8ll9-BLzt
zY~zts*1s!6{gTF{;AI-VPhh%uJK&Cx>ru=B9miqL-oR>;QeZ6-^Y)0yG;p{|YK|M*
z;&-0rfFS?!N?Io0PLpKrL6U1LwDjM<-zg5u>kj;UuG8eKK=WXXs=R*tzu`W(>D#sc
zi%u&)mUfxz@mDd834K6GnZsEH734NOjM;Dd%XHmdxhI7GkPv{4{lEEVUse|peI0-_
zE2)IEa21VX3Qj&ZE}*Xn%j7}FfRQ)MG~xS4eZE!mj8JVzi){o2ejgibz-XmM*vmKF
z43%eu?bHzEYklvnYaRW^q{fTy*+$r=CZ1H{{3)dp5dw^$rA<nnQ#8QIY+rnckwDy~
zwdUy&uwUwW);BHip>6%r*hra)cV4TI*HP#k1_85h;X-vy%^J{Au*<-9bdVP1T}LEh
zmknnd*FW|jW}`C{IygM>$3_JzP23`4_0&n;iyVK05P`|~+14B>J3(f6&k-#sLWSu-
zkSlnmwaIrKp5kwY;7I&H!UnkHpx#mL?t7Furf6$y25{8rfWNnY4~8G*WOwWr>|xJU
zqRF4`)g*3^Fq?d~d7`FSx~KPWlI#OMz#%NM_AE`5_gZBCD}hfxt_Me??>GI~_F<--
zs6%<QrcEW@$7~T<x&iuJb%wkDGA`8kf5rL6Yp{Z~$J>^7kIF$~vCGS-m?kEy9q^3_
z=N&E>?)s!|Q0631V`2T{C$>tm_EeX$Drp8!RU}7?cyE-m6$&G4a%7$ByAt_6vMN&y
zjuNx)F8$-4(Y($TjR6|Nvzh6f(A(iL5*^rI*Z~MAij~{@)P5JshA7RSwwIFV5tkHl
z0VAQB@*LYNkg)+tDLTUph9G$XSrgkg-~rZGPc2s}?4upRreMJT<a%nW8F6|ev^VX&
zD1kLdF1C1Yt7!4e<_ga&Gt0Cd@<h46=>MSOiuX^zNQ9Cr>^S7F2F-XUC;p+Q{$Xb;
zj@JY<D~Qbt-QfxHL9#F$;h~0OI=P_-B7;S<Zat#a)@(roit4D#r3*vze8_bv#`*;8
zeNFl@x6XMCRH91rt)f3^2S1M4O5F^QTW=0-l)<=#c&zq8sFHg96xQ2PC+Aq0i`#nP
zosIuxONt4ge!!fQ$=}SV9$&D6yDZm#@W;;j*FtAp9y=x^WE<Ox9c`^vo>^U$YD~gP
zAplErZ4_cJ(L*tryi1oZg@6fT{`)SfG#BtnM(8;)bxASI0Vp7|Ex(NT`4+k*f58m=
zQ!k9E;R6#7dBT1Lmbb=^A{PKYV1}1iB+R?npSlAAB$dV9cbfL6#`3DcqX-F-Aikk@
zA98+ce`fQMBdxOR<ICxLNwS<BYkgkXbhrK&XZX65$&APkSQ7TEc|phX6@S5I<*t#k
zXZI6w%ARdnQ~r#W>8*07!j)X_FVn1${D&y=mhcKnbHKJHv1<wuLOE#2^UfnNX&Hm^
zge))ewlt+X*^BG_-yum>SkJK~(0?iP?TUZ3nva`fP)|FFLy8qVK*n@GRTTmabiJ<0
zxS83{!oOOAFo7o!e$BIj8BadO&Q;Grihn3++K{o{*hoIU&DYiTLrTxM<BFsn=T)_s
z)^fH?-msa|T%JXlsGBGRcs=p~)z*g!ze_}d`~6!KmMK7toE1-e6_k`M6wnoU!k*9z
z=^?Cka>|vLli@4!jG+YEN^HECGa)9lVQYEbv`NX#Vfw2vc?XK2rU$=(CwGHu5N7TI
zHIel;XP_=SYbC6Yw~Hp3c<#SnnC%H4&gq(GK`<!Q?{|K<HXFI!T2i+9&)0-~4CrMv
zWbQIPjIr?GM_O3S#}Cz%BxEZ;)#|xu1Qon=r4HYJ`gxCbLZDh->K-0RqjmE(8lG2b
z^k2yem|ZnHQ~kT}r`qfrCqn0OWwUrBT}g^lbdvg%abG0C1Kde~-d&oE4`zEf9|C5B
zH36jCrz{#rT01G=BLw`Z$4RscheW|DqmTV-Jz|}I2nZtrI1UfS7Kz#N%F4^|mUz&C
zDw?T`o^Z>NUt5`~8AeZO%(CH$tOCu(AF~v}V)}EYdA>|a14JDVh@sgehV+Qm23apE
zD(WO4Wq<`xK(f6*FV?U$C50%_$gUb>!L4DZp)o4;@Kw&wT^o>g6~oWO{)N{z$&ArY
zSVmtuH%XhXoe&K>yE$9uyF;=%2&1&u`#EEd@}ZUDfbNp;(;%ECtaGl8M>fRM)uxx$
z{Xalu<*wuL&(_HI<(>cBYe}cMx_rs0AT4Ls)=H&}eJ}FHKe3n`v(d=!=ES3}>RL}L
z)1TX?g1v$4uZ-d|G+qShRgn8K=XjSe)D$Iad-cfa)T}&weE|3<&}`!ce?wfU#Iie6
zOcWill$(DicRN&}0LC$~&d0<B3Se|Im~kHScaTw36JJ8e8p>C2%|pDb?vgmra%tL>
zVPc#k$r-W2D%+hTS11B|u{<m2p028Sez^7h>sKD|=x^KG38w+qWanB7b5clwxW<n<
zg`H?r$I2)|8n$Ty7<Yt>r^RXUvy(~=wiHS`Mn?O4`)JMVw>anOQEh){Fqvbx5<Rlw
zK-(>8G6C-#E_ygf%DFN=1CaiwT*|qHkZ}n0XQx*D-X`)myqVTKyS?&o-WgZ#;b+uR
zQ_~>!KSvEyAe~#h-TH^tKF1S=fx~0YyGPliDfcFgm}R>0FFMK5$-_A#0SL_;ati;8
zu{QySvhVxGwQ190Nko$DM5&arWQ(yzQlpSi5y@7REh!=ESh9pf5>X?HL?zout}IC^
z$xM<WgJgd{=jgtk>$&gyegFRsM~C})uF9O}@Aq9l3q>b->s~gHhl16q^8v9TfIF`+
zKLW65FKEq<4lW3q#GKE$!)ms4`-|Jgv#!cjkS9zVdnYu@=*~bgzyb>Y7krI)V`n<M
zHbqZZUrM71iO6tM8}>dmUgtJVnSuBYqDoB@r4v;z9^erOX0h5yPH5qGC9R}Jf1OhL
z)L$mfU4@Pt?m`%Ku#hzenuzsv2rVW<wu!okhWCaY<c|$ZCY`=DCz7DjXjR_?8C{Ur
z-C*ppzfFp{^@B~2OGjJZsp@%(AA~oxsJXAbf2K+Pf8y@%|3zPl+To-;#`wI%>O{Ns
zyTF8jvG+DxVuyv-@~^P88S}Bbx=_UUNmqlRclZ4zx5Om-;#`_ah-a$zyzpbSd;7rW
z&avznrRi4%Gs|bmfl7Asq8UZ@r!}}@weQsT3=wozYecF@#0tPrApAlYxMNy7?@aU7
z=z0b+QS3d15)OIHtvM~-uyBS#6*i!dGXdtkVMzH={!$hOnWK`g&$`+*kBcIi2d=<k
zOFB?s#9&ZK=~?o27nUZVweQHPkLA8$-QBhrM|q~+G*gE@ib9UYE6a25m?#z;>SH2c
zhDdBH%;R|&pow4s8)WZY0T2T~l)x#5iXae!zs{TS)K3^0WVz@j2~gP#&0VAu>d~qk
zJj=bWa{F|9Y4WD(TWUX2X92ZFf3QdClPoz3q9UW+!Az=^q~y&mv{X@O=yG^)lfou#
zvsVoVUvGs92{o-E(oji)OQSXh;z6%I;na;W9#%YWl#%$Ic)*f~uIn$toeY5Q+rjE0
zt@edvbdk$f;-z1mRW-&&M+BlF*AEr8%L2NI1rUsP6|U6Mcpx{{0X(oymZ^!a9x3{)
zAl5f9ULz!;lAzqG*DnCl!Zh!q^;+w5PEj==n?npAIv%<VwjQv~`9Y`8M-D$NcwE*y
zb)`lV?Lq4n(eJ_$ws<gDa?`t6%g?d*qrPIYeHm0arC^)ecJ^elW>Y-FTQp4{AL-2I
z#EcZ^1(4&DUX~$7n&v#{KlbNJvmF@{lj?pZ;l8#=@z2bgZg=EO@m{0ULRuCz?}ZK2
z>rPcA3~Z~v;U1qE%&Ycoxj^@vnsd*YW<vU=smIif&38QavS$x?IoY`yrzUd#SqGY+
zd%I7D$yIAVOu(}%28khr-NmS_vFMD1mehT^SX&p9@f-uLMc#szyMY9<;R9$zy6Y4M
zd+v_7Yp7>|al!7buT&4%uqbB@E#7E7t$yy;@~r75(1M+#HBE6hB`w_h5LqKw@fKrr
z06tMG{QS}&;4{*Cz-#EsGGeBN#Dg^gMn&TuJit42TzTI{qZ5=UFph%kEw!*VmL?Hd
z=H|x)Ur0p)ri32?S|o&F(V{6oe}jYZeqskc2xBF_dxdip=~c{Ue;=Q#8aM8-Vq^G-
zgp~Y`x?gkU#)zu*)U4p2vgscTRxiP)cWU3VKf&ttcssA`vBsX|6VGFY?dY4&zP;AZ
z(-+8W$$k4klwo(Yc>%*AgX)whKasj<z36q|gEr|at(v?!zHoASW;k;2u|Ja-vr*UA
z&RfUkNZv@&iRQ5<6?IqoBW#8v@gP$|B!S&p+cX|(iK+p+Y#NK5<3#lEse=~D=J&x5
zA!quQLlTxzvkPV<`dM)j=<8w;iYa~P08Ft@Qw7)C*&8gWMhA?#fVBu?fgl)!S|rLS
zLW2(~or`Jb$VeIuwV7KGR`K+u<A5F<;jvdSB(IJi@vv86F+0e-He~BnMxAtian#)&
z!i}0@$`TN8+kG8X8OE!g+c$5UT)DN)eDqf5$p4SE*hhd+u#4RH#HRbkE@l4HlbN@b
zH<HK&9t!H-SEL3T(zcUK;(sjYL>zNc#erPB(AxBQ=UDd)uUbLMr-x_!sc2%C?H2kR
zw)<$(Ehs^zj6Z(E{S$48vX*tAd3*(?x9F$zOxe@5P#Qwm4kzo{5`a}K>`t4+F9Zn=
zVBV2!&K-N8pahr#LDm=ou|d|^SacNuec3PKTSR~ZfBb-8DKfN#gapx@V!ICX6#zg}
zC>Xu(@?E|`NWS8!;n;%C9(R>c#1!{cnHoaLS+2ek>O#=|5R>j(T#wG~c=EWO>)bSZ
zT<R|w1>X`ufQwN$z53fvM2h?VDu0ALXnh|%WCD5#P{YZ8P>4+b%n{pUgP;67pMr(R
zFum?bw?iORPC9Zk@$^npn`&ZWN2X9~l2B`2a-j8oW_3`gF`d2ICo#@aoKJcH*PnR%
z^$5ekT5M>M7A3Cc>;;<S@dqL9A$k|+D41ViO@sq=%ORiG>oVPjRd0%l!2Fwit(Z1Z
zU=JyyZwC<-+M3ATdRcQjM_W%)J$J5!1jGDQ(WWT8qH9GbajikMy=zWXcE}=vLPDSM
zC<$-~=2oCpU?`56g2n=zsa>EDiRZ$e0z$CDi*D)Fmk&f{$=3HZEsQl|2roLq#}0V=
zH|}cJmu#!B1>{$l&dipq_cDWoP=7zPPp%1JiD7Qlb0@zv@|~se?tfG{l-{~nN}7z+
zlRq_1s@vDVavj%8+-w`VZ$v2@?tL@WrIQ$+Ie9Hg`9-Cu{ZyZ9e?=#UVxJ(co&E*l
zDQdob!5WiSstvgi$b^R4&b!L<xg4}iW0zI#4N1pzvYhb2g<cFuT3F><1MDYT57?_Z
z0-<W!asSYj)NNP3N`&YV(i+aQ6bUVEPO~-P?3j~cJlkr{nrQ<^(9+Zyb0<6dD=96`
zp?+w6o1k2@QLTzX$@3xQS;>WN<dj*#-o=1=;Q12Lc0i^PGm{#Hy^Ge<JvUIPK3iGY
z|Ddc0wyG==v6=KLF|i4j94d14vBI1eUjT(O5Wk2ojMZ7>CqO$S?s019E{VP`#h(yU
z0Hxrr(*Dxq1jtqN$mE_>F}{2JarwUgPN~;$<PJ{Q@YZOfL+gy1{>&kZ)c1EaLc@dS
zGWTUL?LwJ1vGTV3@Ws$nsLKfLL)3vP%Or(s*(qGpAIBd4_GZlY|M4S@$ks_w3r9g|
zDLH)V`%UGaggq4p`?m{dj-%uO)WeNJ>yA<ebb1yVk6QyN8~HipS1guB4t8!K&g9hn
zs%X9;b{6y*<uPG&`|R4)41F4?o_c#ZN!p5}18=sIS}FDwbwj!{>mFGz3sBp|Oo}3v
zLf?Wa8#i2z`h0eF-QAG9ohXd@`yoRHWiXb=jJuR0BEbBqv75u(Vv_LR+7{B_X^BD!
zGgkZ_m-pUDv8D-SFEK`e911E36bq)uYUd)$&Lex0RB3ngpMH4Kix~nylI}9imFNA(
zNVLB5dMd%`DGEybZ!Z%Rd6eq^Ns%`KTvd4N@R!Ujf@)O5TarjY7sJ|ZO%|TBkKDhr
z;aY0mY09fy1a=-Swwv>7`ILLJF0rzCZk+9;g{uoXpp!=ww8`<!P3u;MLqx?0%_-d;
zrO<qJHZW7~oIvT#t$|#MM4DxKp?e1sj$@rBZr+Aw@k6>Z`(8O0JKKURgW}TSE_^{}
ziRoA;ppZI>1{Bpb0Y3QJTkk(Q;fuxvk)a$w(GlB`tf$!dOLRKaHUf<!X`^e^)|_MC
z?K`Jkkv@IEycEssY0K3`P{cR#!%~f?Krw;!O5HBrMl?oIu9^Nf{9_GMRcP#N$ZiN_
zp(0oY6k(fzpoFJC8<zkML05r%D!d4LR$|rajs5$N32YJ+c8zP1`ERb_U`LE|`B-)%
zQ+c7f@<EICKA{2st%1xf|Bs0v<DR`wWc8!bUA-KQpO>q7FD9C$YI);5ifX?`=&s2g
zR$@U>Suq42d)-y><T@6LV0Fdshi`z$7Pnj6&(Z`&!aBD}R7&*Xvj8ba1XvBb&o~{B
ztg&4@f<_;UXdAwxI|T8C>Ro2e9`OiNs>LywGbf3|WEmJ@<Sf3+m?~n2wQ+2t8#VO-
z90s6jgsxyejp%b?sF@I4_AS!5BKL85aOK@^eMaKs#!%U3{obibq8q4@spmGT9r89*
z<%%KZI;<JlYz;B1I<2Fl^ZQ$F0pU4Q<~zZQ{D*qMab*42A!0<e7T|Z?xdHcE45o&5
zL_q2RkSx91XDwF)M8e8?aj>5!fKP;$6^QnTZ+8R~ZWxZsAXF-HTkee)1wjP04z_YX
zp@(}}@n!9b3z4~gXQk|2UNnEz5jy$bVN7?9JU&WPVP#GQ<mHwzDS2BO`2C<Is^+ht
z&A;iGp6MHzHI9e!d|PB10;me>>8ljReNr2KK1f)C`1N~qrpo_P#aQ+<1b8&eud42j
z5)C=GQq<Z#hZvX(vShJ^NnCNir83^5!O2NuGdoAkBoXJXOk69zh#-wZrulYO1r+$P
z(F8mg0UW5N-Z~LNu3>hZD1M>V`vCUw_u95wmbNxen^z^!swnnfdxOSD*VolHqcXvh
z#*@MdG;6G0wEA#=|9uCn*8+DN=5acJ;FS{+o;zK_NnJDx8^_rCT6$i#<T~upaJL`=
z*FC(r{5z&!wJPIEt`_TbRF4&sTSuv~`i7fe-AAG|5!gVi05K|aWGo{c%s0;kpzzx^
zavY)8Qxb~f^d;TcEc$plhDxBXpn=Bxoi*{&obl<hKh<=-;X(FNpjyopyLaiaB<cTb
zn<EYzXm^MMOVhtibh*6YJxl*{%%+)m^Qctin_wE(j@Kx~#HtmD4|_|jjlP<uzc=-m
zrrJX$$e>Q!(1C5n(ELP9&A$(1PVP3|^g4^36i<0JmsmhWTsZ{b0)&F^1499*A_^&p
zJz<-UW;b27Sqfuf{(7p}ofwv9EU|IZE`V&4t9i^YbmX{BFedN#Z|wPi;cn(C)UaGX
z7~7)HC%l&6b<nZy7kpH4%x7^^c4N<wG-JPUI4?=ZYPD1ux@3UH%%p*VijzEvp@jlU
za7=LkZPL%vceKfFHlb2mo}FJ5Ly-BA-tQlLf#D>U%&wTU0H6W5Wj6{OTpgfaFZFnT
zbBU|y#BEVK*SYYKabv96#Oo9DJBnfD2R7jpApH=;99GYqX>%+h0D^mN-uRfHYJ*zT
zw~~O>T%z1TA>~^;Rd9Ho9fkC8wvYVKV6Y4t2}@;8ZmAg_C}<;EmXGq=qqcYZcSr26
zt&{&>5@0MI17ubtg{0h*ZVMVaRZ0~Ya>r5XDrlGj8*3{X;|1kK5m%+v?JQG&x1o~C
z<j02<bJ;>^1stS<^EMT?Wax5;93yBfO6MG7GrhiP9ESSWp*L@almQ#s3Iu1aIHrch
zgs?j>m|_h(BQ8QPDn1_{%giy=fSvP#7+6n2`hiajcsacfysCrR&y#%p`X;q_IzxLn
zZY1lJM1pn#{rD#bsU~QY0+lTR5&kw$c<o5<%d{iqgmxdyl3X38jS2>#ew#_PrzO-I
z*{?8;V|7+OQ0XcNP^RaKXK@E>#cy`bEWs^>#u%1LM43di;)r5`Yu6He5cVMEx#5dx
zrqFkaO4vY68n$OcJB~tO{IH0OzO|OPQAq6f301&!E%sNAPTe<5NXkl+q_k!lBS+eS
z;U(MSr!)gi4ipCm<BJGk+|lp^eOj{Laoc}#r#_SdX#X#*716)ja2uIe%7x%LRgoo?
zTYf)p&gLL^JfYe9w@OJiuIvCWxFA^c*|CwGb(F)u9`nRmZwCC!DaV3zh=R>&ts78#
zP(dVliHJYW5A#tih>06Jow|~l6m9pZZq;sJw?GQW(8HPm+IXL_C&M}#Olt%K=@4VC
zczZQY<Z|<gub-*4o2hkerB7v_GnGT?$L-X~uc0fD!#H)}Jc-<yY3Sy|qCe~H<?Ve0
znE^DRw`D`_gH(~+fF2Y|vam@(2M$;TUK5!Yl4jTL$>y>l0Ad}3#0wJ`oCTPSqxeDw
z?i;Bt->#v!fOKqLy7*j_zi8-*v_e>Av79IdK0c`gB8?z}ifiHBMu?$$5ge{tH?DL*
zJg0X`IMGZz13v5Ki|a+C+WI_EX>3hyxZh%*_?NhhbFH}VuAE?BegyI5Zr1W?VVN+`
z#lwK5g+)WKvhnd9Uatpp_^Pv0%Bug92l3CGvchtBR7uC1;>MJ9puH=<>=rE1_p<AT
zS;~oZ*&(!MX@6Pm#5+RBE6-&a4_yBq*^L|f&+{KLp1S0@JveWXZ;m!+>;lzTiJ&M(
zjA+Q3!y~NqAL>Sox}Q!}j$7PiNARMdUv=7Yj;VaQg6C1WtP^JW_GDDLdZIi^2ei5o
z0*Fj5p#O<GM2Vwv^WZ_PRVuw!39*cmq!$TxYAT*&<D;o5KeOfK<vJ5mI>i<PT^|=9
zTcf*#+S1WCg50L30rr#LsWX<ugcMsx%YAlDVeobqT!{8}tqUef!nVM)9fg8EoIIDU
z4151$LKVm~LYq=jo!qo|al4t_^NW)aZCD7Pdn0Tn(tdOM&t<(gTq-Qaanr75JqAY>
z(x<PrUUfhHu%+HKXRD(6UCaQn*@AeyRRQ?ucY-h>gyJF4D6yG>mv88imj`}&UFSGn
zXrd~CqA6kh=n%E;i64(P*gL4oOt`jqCk@P7*Jd&r7GYS`Y_B<;$awddztiNOBn$t4
znxw4Z3J#a3IDGio(v1`Phki0u(snsrWYBLE8*;`iV=NOGWrRbGdeY3S(5h=+AhFUq
zx<h9}_~eIy&Tijs4*$M3KY<GK_lM59*T!5EFCmF6W?;`kUUu2$$ksI|Gf`Z&^o`^w
z;H^La#FUyS>TC*rxYI5X@p+>ml_Iwd^j;u2BVuX4Nm#Ht=b$LBw{=7`X+uoDY%(;%
za_aCjzN_?V>OV2fuH6_Y$Ra;t=Kn3FU}@zo2kFhG<|@&?+UaxB=i5o)s!2b4g!-=w
z9wRKu{i0bFW&cj9?7{e~7biHgu;xWe0%G;CLg0ecS4t+5o2~&y?stnN2T<^P`=zL!
zI))q3(nD;ez5N-s;Fzfq9t4wg3AJa&eoXE%?;&<zroIiR_lJx50_swGEWO?t@)%g{
zbUC5mK!gP&jf&X@p^J*mX%X$R53B~TQ<Apy-XRz<!nyF`L@cc?U?Fn%&`o_>(?CQw
zA27w}_4`g)fw#LOdE%Pc``0f&)Q<Qc_`k7sC9rqjd_1x4#V@l^>X&5#w6%(<uEE(=
z>kV>;H<P_BGN^);IacT!7mrkX41H9qnfX>R7Wwm`|Br`*v)lLJJT`Kwh4q`g*5^RH
zto<C#bDneq5LuL5wr(l4R`{6Gt)`iq3HWC*r*GB@_3=!FsKxI`ti|qn!3+fqCXO&!
z+2l-N{igeJ8~X<IpHowB<z;^xAyzN(ByqQV_I`<}khcmD38}D%SgmxB%W@9#M_j0E
z&R=d33HGl5f3N=p+bm)avn6bk6W`^S*rk^O_4XDMP*amEnZCMvB7R{W8|D_ZNVO%y
zLHTIm<RqgfTfdp*kO2yy+vg4Kd|G=yc63z!sV^|CBnBvJ*LI=z#|qSOWXXlmN`%O<
z+DygXhqZUO@UXE|#utDm_#WFoY&Syc(T4#iUhIscyDib#&Geq~mbme^HAE+Et3b?h
zau?STE7dqXqWJR~NNX%}aZD9zZ2jwh-qU{wEzE)=7=M8~cER<&m#Jy$@MSfB=c|_l
z1V(F=X|^YeRSwuF4}Q#iOXRaPUnH~Bi0+fn6myw*=Tce!rR{?tRcl!6F5|{yW!a{6
zH2owVRytO=CMp&7lXw0FJtog$gBQElq<ty4sPv-rv;nP{1^q8H6OTx^QLj7~;p^j8
z8y@a<7{Hqm+Y~mL7xpTKCofwZ#AvqBdjEPlx9=sc!~l^3fy+1OM;#v4Zj0!aBtaM)
zldoBjd-K-)B^WYbrK)aOatt%+#-SF>R;9^0O>q)c0+zpdZ7jG_?5(3?H`Zez!1q`Y
z$~jutYR_-;#M%XoW$cTFR*Tv@D|3V%O?&`Rm_P+!NQebW%NY-kAc7AC>^mwBWYDb*
z+a6*8P^*vI=s2>7`~xBmEa8%LQG0Sjt*0e!I9Ad@$9zpwd}2xH3N^|($WuD~TNr+A
ztaJ-x|F4>!Pfyei#kfhQ3hgtpvi$&1n{~}xuS}8dlC`{tJ~R=2QE#co@5nvL7%AL%
z@MUV_+TyPR?y!3<|CzZxucY$wxdyNI>-={2-`{@g0Mjl;;Swf*>kn0R?wa8p5ICE`
zbonuD_n~3E{_f2Aq^Pj&gDOF7&9PFeh|9!TqEFw(&$BG>h-~7%S2Zl&4XY(@)V1E<
z6fk0_Ruk%x7e*_!lhF?fOV1mP);8r0IiR&FI{w1>XbY<Z_lKfVFeY#9j8`Y;XlnJK
zexV=zWJKL;xR10AtIi_BrU<~N9eV)vDF7Ec&OnS{{@cbpj2IsAOguRM<e1Q~;<wy;
zYyo|x#F8E`N67HU>A_a74pb@QsWSTU>WQUP4!gMl2Bknj5+IE~tC+42kIxbvErChI
z#FC{I<J56%2E9=6d~&#kDlJnEGjQp+mGkESuZjU4kU8kIS@cb4D}Y7xZ6^e3*=|oS
zC9aRAb(Sc$8PG42jxvy@>xmZv<UXkRNv*_Ky5Gt4-|!j#z9%fJBBqjal!cBdcX8wn
z)8Bf1T;=zZ&Br2!NxxuRwO#GFhD(!D60uHA271eXq+-JVbLRxNzfi@5gugr9(9%6x
zp{Vx6MXy{<m!En00{<zBp9sJ*^uELl-IKi_Iqwd`ZH)B#DnK4G)b|gg>OZJ3xd+W(
z6^Y0Zp#DyXy2VT88Ph4YBh%C}xgMCs#0JR(7F?l;Xy{MaSx0A4MAc!R;<ov;#C(=O
zNpxdRYab?<TMg|L5;u2Q__0&5N)wY)^ZXEXrcSkVcb<P^ynXn_1{r55R=R;iGx{JT
z^zs3e02vTXv~LB$2_We4#GliFvRtqcNfQ4TGx{l#6YzEa+W)ZT57-3}`hnB2x=y$f
z#TV;(k`x^Dn|31ev1Y8M)(50lu|_@x=EW;r0jGJoi39Ua4DnYoFMX9cODk#L#DU2c
zhC{$K<UJ0_=}N91yE%}x$3QPJJ!MuB-RK`ST{lw7zMA`tSrrNvUmb8&w>P3yQGI`1
ztfAA_3$Y7?=_sF#VeUK1v<sw8kG&jhz)z+7?)NVL$<8}wj5P07RDiOJNO8@Soea?a
z$a;T81y2|#gPp?wof`)WL_V92;x)n$&;XBsMdA_rylDgr=+r&Nhj_JizNxbJwU2fX
zDXgt&k`P)RY&$Pcp~U#xOK4OE#Hv0Nt{H{L>gv_^$ty|iu7YvjBnn6(>9Jy5#d5Y`
zY_b=h4W(Nj^0;qd=SaYniuD{mf{BTJI%|xFxdRX(7&PI=gtW0$_scILi_ThJp!EfI
z1dtEpY~tDwxKhAwh<y=yS^}+Y#XJjPj^g0DxNUTDwrYDlG_B)hbG2BAns6wCk6Yn2
zqvV@4D#2a~xyXcP@Ex!o)&X+T`qn`I5M3D(&dF1j=ZLCc{G4>^%Es5a^Qzxz=d!qi
zdP?3GbSlb3Xmy2gu~tfBHY5xljFmi4$dSY~!k2vz5L&Zh%4YDg#PzNmqIuGj&}OU`
zGS76ONWJVUKZE58ZSy_y_5IWx(<W2%ra$`+`B!yL7j;(85=DY#**N6_CW5tM-bz6y
zP0XISII1iO55i{+j|WD>&JfOfBCl{w&L4%XcPh#CEDPZxHWqqYmVic!%+=@GbXr2K
z2feBoFyp4KK%P&X;0)F_ebbyYQmS(^T3U0Qan5^ZOp@~0cxL*L>2RiF$?>R`cW8**
zO0Gg7>0O^UyuwXP{{pBSV^&o9hpH|`c}I6@PMxjY!~rM-4*sXvg~<+8)#}%;qho{!
zxEYIwUOjG<@)LBJjw-sHOs^WQAynhc^T8Yttl{sgM{2>pf-yfPp!glsYp9o@NybgV
z_*~26MaTgRCN{V=N++ZI^c%rSuvPkE^fAU0wbaa%(8q~N4EnA&51CCn{cp*%nf?(F
zTU&B1^lHzUgHE0@>x)7v^|_te(o)NYE3#HWOSkK|V(%jVj~omRffZAqNSnP)_oja1
zFcxTNUGdk_rUeZ0hUD$NWq_#%m0df*)7?2SFvB)5wPo_I^@KA$e9W|@_SlWEm2B1V
zLGT?WaT}Om*xRZ=FureCoHB1p#gr0tH*`{9Rn8j4?<4Jv(M9&MJCn-glbt2k_nfNw
zR=VdX=hd{G>kh>)V6(L+eWt30>@~fWD&xXsDIGuH@Vaf`SUD81L9T*2xmyvo{QTA+
znSQ;|qFFwF>`8pA8%bE6H|!6(kC;9KWAbUX>LT(Di?(>$@@&7IEB@(T8@C0EKKZ%D
z8T^i`bD*LYDT>n*!>dEI@ofCySVUk0v7iA6pT%bbVuKim4ve<a_{M`2@i(0PD&xiC
zTW9riTeiW3aq&F9W0?Xgel~U-U^5k2?;8zwd7j!~iH27Pr5W@E|H*BJd?QcpSP4IE
zwc@Q{fxZs+kN3>Kdyx3ArDjY*-(^vET*L!Fq3ChXWEpKw{oTfOaPewzBmCF;m;XA0
zRAPNa&Z1URgiN)Fzl-@e72GvG{e7^4go{|#UO+6d*V@FvS#v+;6a;Nx_*q9mJ&<r?
zX?fGSRB^em0FjJLpOoXzYwhk7+LImU-jLJKS7|DjA4`oVK|T<S^xkF>`LLBd+R>#w
zXjQ&#(^Iw+qPx7L$`&OB%>pJ6Zivv2p7PJ?KitAoV_dh`xiAtS`+}gl@<@a61^xLF
zROZF`TLF_t)WNC7n$`F0tcXAGu2A8FL<UAT<8OOCKwHOJh({YRfN$hn1LfvEbeV0J
zQ`2SFD~X69YoMdeFS*p)<=ZW*a&y31QnE8T?=V&GKh3J3PvId`?t_W5eojgX#1g<D
za2=R7I`=FRtKB+UcLw#%Z^2sytI{PTjd@iQGnL7U2NW=(Tq8{@2qwsc)_yZHPY0$N
zr&s?LP@AYKP1gJZDMG0bwnxUi(OGv!9w@46VOr42Qv3jb9k>OkI{`t%+T9_YA%U^&
z%`qJJ24u7LsJGGK_{HMXf|P`c`ASA=Rs)gp8&-L9UfE<J9T#h`<h^w^-Iryk^lfBk
zpzz+&4M8t8@FM=!9)WKX*f9jC4~+#8;ZYl*a>Zf+=*@2f_8+y}a5OtEr9O}yMM4L;
z8j~DsW`R(Ej4ytN=Ao0~pJotCU{+)QhiSrEwzN!@>O$RtF!+Vxb$KfF<?*|C_<^dM
z54?Kf)>sBy@*h(vwc$qiO!Rn1r})%J7ew{0R_c+Ab7Rhp^z+uP*8|6IU^q1EW7d*l
z8Q!~T&d=(9B{BX&r}-Dd?tK=)r!fY)Uy4QvbW<2xrW}`<xB919)y<G_3^a-TV&{0m
z%(jmc$A5QJ{3_^vA$Rw&k{z(t-cfGxup<hjiasS-2;#qN74GT5{uY=~d@{n>1yrrn
zHJe(q=->>4H@0IrLjY-lv-A<i?kuULVX+}2Zi@RIK~X>lk>3~LJ0Ri&<Rd~Z0Bz_H
zpi^@ja`xb9#=X$I@;>YvW}}!~68vvu*gpiN*RMIT`!&m8qR9ha2E}N_6};WT+WA|f
z<5e2>jWX~b5;_XAXgKT<Aq4;h213B4FQ)YSM0u(I0HhKu$>Ppf1)S8VI3B;>lTcH7
zi;WrBF+aooi~GfuU(Kc4RfP&I&b0oLa{t9$3Z<vReNd+&FUFYysjN{<O%lZ4=0i-k
z*I!>o*VyVqqto{Srj=Z(q1l^GfzeChy$1@(q6~*shPR|@(m@zsf7Z#=B|h(}|LLi<
z(+M-b)t+~b6Zv)Lcyr&W<%x5=H_a*pzr)XxL4A3mN7DljPlfr^ic4JkQw-1KED<}V
zhXq;PTa6d6jk+Y!i@R|shT|IO`<O|Y#@9tcsugr@6egG<p@m}Yzh^<)h(;`pR*}f}
z!Fvz;S?az^l9PsyAtEn|JR)2JxGJsU4QwDfVf4eTjw9;C#Al_#3c$Y<__3lziUfE8
z`4JOKYzy(d#ob$6aZo5r%`NV5Mtp&2Y(JwtjetJWKW>5s1={6_GUS7Fv!g#>=lLv}
z^xWRHJ}j1^R`@-a!IFU-f6<Zb@o*Nn_P3AH0b;OmKVz6)^X<HB%WsS_CZ}}hcI$(K
z+e#c(sU)OM9*j{K*g-(w4Zr;r)7re|*I?UB`^?grhmkYZlNsae6}MAAt!~|Zx)Ai~
za093m=Cq>@mwMg6Vt~hg_&g6!xE*S>@XuJt7H8I5n!nQ@DUdZo*9#OuP*=4|7Ixrc
z2Qgo!2-EI-6F>v;XlH9RETb?X#BQAM%wU-ffl-Waq`gl=CITQUi!Z*kGq!w}OXZ=J
zju4K%3n#M*C-+a(2FI%#j+S`ZED|{mJBH)^E+F#M@)F2INuCzZJ>rWn#K)sYcEq0e
zxvL<?SfXWz`{g8L2&mwLF9U%q!6-5K9cB0`5_Y%6&TzIix-L`KJjthsX>0yt{%%>r
zy;e;@16i@9`2=f2fk9WwJuF?XCwS}cy1mrijhWfA>tqa7$W1wL@F~CF)2*cTr^BL*
z8^z01U6>=8WqAj#9VvV@gyOx{bW9?A-z23vXMDUdc6sc*$syX}UmIP*?{0YN%ay3i
zr|>sk<}3B^wtZkI*tug$i<vqe7NNk>;|(kpo0(&l4Ie55+YIJOZU(BHJYHXZYjSzp
zRzMY?mkIwpWGgJ9QNP8u?8`B0Y3fC}%$kj2fsU$-Q0bF}eh&1yRYF^I$leoCbq5Kq
zhz-!i1Vk998N>|KB)O)sVO1HWY1<6lot|%;*o}|D3rRZsB+{H_<2yw*g{1>AXtpv4
zHdfsS?_PP)lZWW5rI@S@ExvV(AyE<t3*f4<@Na5PDR&<)9+g!$;^h)udLXvVTj!D6
z#(%}&$S)7%vi~6sc}~0U_@STFC&yGgUU&9<^E<U=X1|dhZL`qoIBla;aY?mDnTE+p
zBPpCJFS>JR&H0JR9{Ggrr}?XOMRJ5}^kka<<O~~|-xu?NAV%Y7rtt^bIf36xUW?g#
z?+{(_R@+fOCTK+?9`d|@AqxqW0fbqU=uj!oGJpU(NGjrK#IC7#?;G4{s*_z6FPmNq
zdQQTm%}g3=3sMQpX2-`Z3{x?v#AbCPW%>Jp`5Q8DUOa{lt|?x01W|#Aub7FVnnWOO
z>G?xJ`uH&wM%c#+ZK=w;FI>k7`MjB99!qdxZa|Xo28vx~XE#y0V`;8~_PEg!%iznE
zT}c&7iB4kNT^EL6Dc3|7<L|^c;mv#6@XwE(Y$WdxH|o)64^RJxx86VWBCIUN2gaKz
z3DXaU&(Z1GO_X&@70WxWZg%lfc<bf36!vZ1(Mvql2_~K5RC0UfkSg)<&EbgfTkgcC
z8NXjTYytD?D;slN`kUfqTz>XtO>KMlb3{0XI{j4ek-*EfaE8sevf>pUZzb49COnwX
zBdX0~12g&gCNO|ddk_ppY>s)u2pY%EI`7mBBA4#ZkRLe7V!yB~zH#(X3hU7UF~+hD
zJs%-e&m9Zrl9Ou}v<!=61P<$f4uYDURR9WdN3U4E{PgTl47VRF0$`c&g^eBo?9^;#
zxfxi64l4~u;!%TrrW&Dqpn*_s(fJU$ddY%WkzZURESC_lft?*!#7F-D2%fkn+gWa9
zOtpc7YI{$`(^j*#X1ISe_r0IaAuFq+^v9oNuir8^=F}kq3})X-=(`=RNKrn2Jy}LV
zKKEs|W8(dCvK`|fh2nYE%Q1;*_oP|WzWumDDf$?pqnDQurcBJO2~%>@^~N(BSiiA7
zt^cAXv9aP=#V;l2L|I;yc8;glGE?g8T0YGj*>ofZ&&_FNjBq<Fv3oF?#cl8S0Z}iu
z-9Q3;YHx2hIeJZQJ{rvReY>^HqP;N5M?2!}fqH|m9_K|P_t+)+WZK~Y-JP2`G_#CC
zkH{{j&Bg-sC#D^!nNjo-p>^Qu<VU(&Gp`=R&@|%V$-81)9NgXGcQCAhMkj(Ut1Ch1
zC1eF4=zCY_kHr}tq`6j8AF#0xHrVdvx+dHx4hvC`oG~^7sEjzhD>)YS<#?^?txF@n
zXqErl>?U5;Opg9F;ef{9!jRVG04JWhn4+w6z3xR%v%P({divl#V3YJa9%Nr@Ongm=
zGpOD+)KAEJde8ab72xkRq%z&rF#4@1?SS#mg8iMJdg69DltgW5d;5HN!wM)yMyu||
z^E`qFCPUD$l~9MWw0a_0r5dDuuufWuadA9H<w~&W7sB6X&XEjY&7&H2+u3oRr;o?h
z*+H`nP(#2pq)KA)3Ev0QsHYyrwP-fs{=#$+$;`26;g~<fbO=MjKl({T-v<E(LKxB+
zeSXn>7ftk<?X^YKn>fT)hFz1DX{mGMREWMcD={{^aK_LB6iVm;L;sBs;ls42HK?0a
zG#k!cvpz_6u`gF}kOUZ$9mNDdT%}ZG34agph1q_`i{kS48kP6dyCVc2KnV8yhi`R%
z%A17!|E|QqTiD|sym5Tu@Q219+Wv8iZ|$m#NmbtXo(L9p%A8KX)be}DQoWVcy&diM
ztwHNuaCv@m>zG(a<@WF`KFvG!S<lE+`A&Xzr!nR4xST%XebUKt{4aa9DVuXh6f45n
z_L`yzU#7*b4OKO;DFPnT>3~#-C!rky90My3<au0Sv}1_IG-Vt<#kr(6RA&K<wh*r|
zSweIGt^!1^2ZtMnW|3I4WOQI>pa?|BA!KU4fi;cMwDnhw14vP<<u%l$Gyx+AI+um;
zltX<C;4Q=tiI;;g!V+0X7jM-Ad|9LR&g)C|<Zx^;fy2Ykg&zke9T|02$xobIWU-NO
zeg*+D<e*!mP|aq1mWeGvQFqm?OZHW#1I>Y01tCQI<t*7eD**w^ohPFy1G>rS21;YD
z|3UZy^u@*X98<Bju3ma5$AT9sGfQdJl&)%u8?~Rq*gh#!_J{9@J;n(!*^6lud*Y7U
zU6>sDwQAaXV#!RxiWwfZpWF`v;Vk}XX1K4S@JoDFFtkhNjewiXP7%x65@D;#hlVf~
zRBb#oI40%9r>+{**OG7Ztcl-GB-A&g7*WTkpwJe*3xnLntR1bxL(?3|lA2LpG-ZI3
z#Ot9y03*3hr7V)Xj_EJWo$Zmh17HmnfS2e-FpWcH0p`bBESr!AF;fL3_p=)wKk)kw
zZrjB0qbCRZ-6d3lO<*|sFX}=tY8mVUJv|rHQ=xB($`fBKu#Dx<S^SwW%@uXggA$;n
zKrgJ}7>Lxl=Dr&mv9+@{)IP}*TM_g}DJ<}Rc(+Ub#k*zPJ0LA#-2G}|TXjlT&GT?u
zw)>m=Ib~jtrA@uD_Wv2MbM(Ze$fdLu((@@O$LK=UQj~45yzCBsS~Js?UGcQB>|1Ga
z;o#SKRpXT&(UhB#N6#$;?2>@!qTa`Yv9n{Tav+BTqAOT~ssYX)Vw_O<Tc5j!%ad`_
ziRBdpk_PzUw~2NYLo|S80h*o9z}ds!GCNqq@3`E6m)2~?5z1n$79I^5nw*5(4hn7G
zmG#uv2Tp`TDj~i8hmecjjL@7#Yl8B_5w3Z#R#*yTMMCngMQMnP?FPg2z4cVUdGTig
zh0&2?f<cs;usC~S)f1|NbE~;JmZ7>^gIl)#IU4!HZI-k6(K-?>meO`m{-3VQ4&5OI
z@lQ<`bgc51Z!G>~f4;%(E6%Na-jGcwb#v<~AtNUdNn`imtoI8I^i8W{X+jjrNT{SZ
z;q!g+%gLG880Kmg*@L78L>4QJ5q*&QYadk@L4sO|IScM35KT=h+2U?DJ~JU6;%@dl
zhC63$H;aD7qm5bt&ln{$Q8h!{4!dKPZV?cSgksGs8EYx_*JLtn&GO*b_#UnC1?M_H
z<)mF_Y4ps}m=I`*O`4(D#`-=~Z^(q0Bok~K!1D2|44QqSN~_C}sH3)S@J&Unu!*A!
zXvbO#OR_Glic6z-`4eM#LXj9^<bd?r#4rWt4<1okz*?pi7_wXvh#}{!OG2Nqu@!5f
z2GK?qkCt~zCBOf8ab|^H_l0}P&xZrLqX&I%51LuUWajy&#KsYNf+%%s+rqa@wxTQS
zKb9F&=V8D6BkUAPqk?BpZ(#*urUwodA3M;s*ewAXjHxDTI4^x;Q*wiXC9#0Fh&mZB
zOp$;Fw1v0|hbF+Ys6~c@>;a!Z?t*FFixed8h{%i+POaBqjKcHcd|EFo0<^jyyRBHV
z|KsY!x2}S-ZWZ7#*x;~%DS~AKcy~({AUG2B0KU-^g8`)!%}xG}CZTartGKBo20C}3
z&O4f}LiMtgdDT>LF$w>h0A+NR!@;HR)G_Jbe8oR=+Q?2fSc>Pm|5pKB0(fSAJ+<tJ
z691#KJ`xKUn~i_FyD`}!MP=$;XTCq29)C%qv8VFJWrx%mAl0n2p`5Fww+_y_gKvz9
z5ep1Rb%G3=oqO~JqiDF3lDRf$ALb3wr|04Gb1jEwuycp<0M2UrqTvZHTxxG2#CCiI
z3P0`d&yR^oZus)H2XlSc*5QkF!o+)#8jvHP@^&;zjw3`=-*qg4#q2bjn^a3{YK+SI
z4&ivIIz&dCE<&#tv9QNm&lEr!R&k1H#GR0R=h7|{*Hs)vE^MPBoA6rZ=WXJ^(5LMK
zfhXltcxEELg_fz4c;%drPZ_~A!nBL`&?oVww0);4nhA}vZ*LaU{!J;7MQVj#%*HeE
zbUNQ?Aakpd_j(eIa+G>XbH%1pdyK`U`Yvz7P-2Z@&P4+&3AOpEaub$<2M<d-3fXNE
zdbD~jP@_N?$9%L)zcz0KRTKGZ4`Wt56@ngA`VjWa>}p~QOqy1>m%P6wi1})e6W3)M
zVb+ABve+5N0yR1dN_}^;2{|EFY<~$jea*d~rj@AvnwYxp0rg*$1T1#t-cgc+@gcwO
zYt-sTMb-QphhneFj)t=OB{=O6@=YwSAL1Bd_61<KtP+t&fHpgC0Dn(1uY?}*WBVHV
zyF11K0A&-Q4tChk$nxpb0p@bFm#Q%bKjA!P-e#=tK|k4Q*xK3I(SZWN%6(}wYMBKj
zTC%K!v2&$jtG<%=pr>pAtmw06&Yr6>9UvlK*qeVuzNH02gU^M6)(CJ@6;hY);C1nm
zE@|9lps20pMGYiLC6cf?%(zdwZqM8aPtDMoA6+v(9~yL*G1CvTe;(0-ieMy>rN1D~
zILT3%v3`DS+@gTAL&Rl`WR$4p@=&A^(mavscw)wsO~9VA=7Vk?j>m4!8P#S#H&RrV
zfM)mB-!(o(6?|;K9<%J=w5XA#oFB>3EiF6Stgo{&^XkTlV-PiYVLMjRQ4HHboNUC{
zTm@YV=4`)8u#WJ>GE4`<Y5$TfZ26yv615NND0in3mdB@55o9nl3|_I0@SX$5mRJFG
zI1f)unHdLRDzsKb8!jrSn&3KZex$vl104u1H~|Ey$P3`9t8)`G_$T<6ked4M;NP;+
zw06^G$~s$vy^4g;z+&1yE&B~)$16q#UR}=wDZV>7?e>W~%bt)3g&Fc3cCT+6`Yy*X
zeNe9zc~CRY*RmD8JkvJH&~(SsTYb)g4&E-RbB82k=BzJaW_C$mN|W_*R9_mPHvODy
zPX=UdacyZHQP6{R#r!s|&<Er^A%3=>z)+tTZ5Q81s5lkOk<~u@=8_GWaf~_TO=~*P
z5Z##dL*W_e#oTE4p=d)j6;=P8);oGypezGUfSz+iM(YIXNhB@D3g8^!B7ZLpYyXwh
zx>~y~^q+X0vnWSGf|yX3huj|KvP5(qKn;1Axc+XPbFWHB3tYZ{u`rUps(;GGH=mXk
zaJnyVbqP!_Q*mQ*WT_7R5-0N(s2i30pJXJBH@)R0<Z4+f9oQaSNXeUc=6GJqDAgla
zV$!C&oyY!2X{PKlzOFGXMvZCQh#k-y-us<QdlJix%B@<MIt4)}BLg0D-08d6#Dp{&
z*sCCQ?uft-1`}Zjk>L#K=hiC$LMA4wRxGMad?p7p(fwr9EST4}m152hBV<KOtN2Ei
z7U^Cp)yafMC5j^EK#mDuc=|SFmp^!%$~PHvUP>k0pgdl|UL>bn<hiXUw8s4Q007rl
zAu1z5$|#Tb1KJeh7?hWIkhU49DsjH_e0cAZBPg-<OK{ONtd_c2LgWh-0p!}MAx{p2
z(4T~iHf%lM=LZRc>qf!3YdAHEmrQ-suD%xV_&5d41u)i!N%njE77;4?V=D|y&M_?o
z*o4)oM~fVr;b{NSN~*Bv(yYz7z1VW(_dgq)tdvQ+*T&I4A3bFa`I{-Ro=V=bpMFCf
z`je4qYVk#@E<3v+JD);1seJ|j{W~h=3>jF-&y@LIdbLSoeNFq@!gL7<SiR$k<Xkv|
zN~c#0gro^yTq<-Jh@iRlCxR<)#N0)(7wZ%YH}e<)DS0A4ZPl93rS9$KU<2f5X~Cj8
zWFs`J_HQNBjBd$<<zcP-Uei!~<GiyWFYtGp9rdjM-B#<9cb0wSCcHY-&^7E+Sw>?@
zUvs;4*Sh7^B2a^fhu}K01;GE1IuSWzJwkbCR{H^Bk|FhQ5jxX5QYR`o`>44AWMrAw
z0g>4a&lK2x5*C$&B`a`!IL|oMlA&){r&n_Po4i{jJ2p5sScH%x*BkU#b-om+jnz{c
zi`c)^99v;1ivPp_;DPKaw+=eJDqdxuXYVRvOLO#>Pyan_|FqnRlDk_~yt$$&jG>zO
zj5<*U-=-G=V1MoPJM+t-fBdZ9=(BAu11tQjmtUP@_j40<?Ed`OV^Nzl7INrDy7C56
zjO5rsx*dPeFGABj_kveDYGmE<m9OEAgzXsm9EiSP7=U(UU=-d4E$3Nc04A18JqC4Y
zm>Fm&ULbL9C&lDYsf%h@3kQOf$>Jo5i3Qv2C=R(T+g2+XTCHUdS_;n#R8yI<P50!s
z=<o7)?X940QZW!SpdS;;xXGg3+ETIpR~Xb4=OZal4YXjoHcR9XTia0#K@m?9S#$t`
zTQ9!MPJA)+#Wy~xGwBU%*Em4Sr0)lGkT7}wcf&A3n*ZGoMfAkaH_?k?R#NxthEjus
zYJE9QOUMqOVSvzT!(D?F{$JA?i*0G3dFxRqtyj!rDeI2+YA`e+5SsO0IvAFumtB6*
zm6bJvjZUH)GokeXyV6*Wr6-wOcBDZ-ajS{<%e8*-?;or#@b5WGY0RKr$i^+O@isNT
zRi(Me+ZU^yq%sRX6W_vq@l1AK?;|*VjXe-Fsc}V0vEPcLJQogG5JRqOXL&tf!KmL3
zB`rineA+*9Lo61B*bZ0H9^8>+EubSDEK6f$+V0M7*?|0N3WvM&^ZUA$OcOg)6ATqg
zM<HOBjN*grRT>Tk*;%fMZ_5&r(tLNL;PlPT?2r>1rE1P;cz~EjsuG!JFs)4AVaYWs
z-+4(xokUaETsNc)f(Y7C4B%8U0KGXn3w*}^TkNLFFaHglD~M<pUb{`;QJNruW`&3_
zWX$8zC;u4Wp((MQV#6m?9G-~B&nRwg5q}ADpr?dA{pM{x-D;Ff<(_*V#23*bi;lFP
zy;?u^>Y?WU$6T>}C7%C$`>`D@F%${4>cULPd6Bxbr`K*P^LI`{BgMx+96$i?RZe=x
zlcKE6j3G_cBE0wz^n*fb$DKx@!Np!;{}g*)-SMak`$+r0LlTkYS_yzB>ENs(3i=?z
zm#u9cNuB6G6qiK@2Atn)o^LGzDgL{VQjA-8@j}dHh6-u#n@dM6kMQQ`&WD8w^ti)P
zWCp+h+@J7$m(<3-qlJR<ZAVJv@P{&O_<ye+JJ{bfEb5?AXE6T)(jb(T(3nEQQ4hB<
zq8qi{NMT&}umV_)a-&~wjx3x}!@pjum{Aal1_wHj$cm~yzXJ&GkH-V$m4qG3igl`l
zYa?PGX@8R%=8n$l{j2TVGO>70g-XSzpFd9%HlS@LtJT=2yLV;Es*L6Eit4xAfeu82
z=_T4F!6{}aY`@I^R_S8=<s3y>ZPioXOt~-^7QLu`um`jAl&(avB#E?4tC?2=5x=HC
zji%|F#xKdbFuvb|lE0bwF>^nyUL`Ko7|L<3zS7nEO6Cd?36g^1Q@Ce|Uf0by%A*Y*
zOFUjy=b*FkxI8J^ghy%)NwLONlAowpz33Kkvw;~xLer~m=+ci`D#xG9D3c@Ah=|tv
zQA55fOG)*8j+T?E7d3H6Xj#u7@sF#<+?E}6MC{($urP+&y+d#Q%FVm|mNN^nWB?Ym
z@3B1@`24D)t_G%xhWo@6=5_bJVtYgdZ8YjY<l(}Ra&(P^myN35<P|*2+l}&aa<B$z
z@r)gPgNLczuY^L>!V_r;H9Q>Yw>^|2JWq`MZNunxHsnwyt*B*o2kfdNcPm4T@!-^T
z7HBkAq3C~sMm^OYhU*)%3yDjj<A`L_T(K>+v>_>jLwSSLyr&}rlcy%hW`49Qsp8V}
zSiPHuIW4vK^KmOr6w3?7-GWel#B;=}JGaY6*y<tGQ?<W7H{rXC+TPSOPL?dMUs_M?
zhY+U+5h6r4ewC^h`Tk{SY;&)<@qzf0H{=vnphe49cBV{-N$vADOt{JLjP}>w6TfrV
ztcqm9|0wNnQaz)(|ARrZVDdq`4n?@eU4xkxUtdkqEhQ|X;-m5&YWfuuZ~EjO@f{=Q
zRIt(o3W{(Lbe=`QVH#oEXg-z_2M3X_GWTkG(6wM@=_&}h%5^O1tz(YAZ@DT?YY_1^
zLR|iLjP8FM2eR~^+^2f$I)rC;3GKp4W1F`+v|o#gTetIhb&3@2DMt+R_^ritrFnVV
z4YuE9B)`o($qGkBr=vX149_{`yZ4Z3A^k#(sw57e-7YwMBN_p>j(d&Yzr7O<q=@cN
zFkyA+!t;9ApyOL@8aA@kjx!DdQCx*(`TDe{Q(AjDui6J#$sa24SMVj%U`^m)>Gwt=
zkrq&)+STIu;eL=^NdaYG-#3Bf!FMc~<t$%M6Ls>tJbz+!V6_Fp;Xs~1lnf1Dt78|T
zau3HIG-I=L$3Wpvh5MmdzJa1?PjQvt58kP&x~GU1aDzOiX{~Q+l$fq;ytF^_PSxQJ
z2}9Q_6xEeLyPw^;pc>0KG|0gGiH||o3A1d;uls$>;FeU_24(S$0^JcmU%r%<(yyYB
zXdA9>KlaXXezM089Qv$h{QN<g6zt;TC*stdRrd1-tnt7Al==iiX!NMCgg{>-wjENL
zref$xv2{a);wnI=5aoPC0e)kLc<&!r9i@Hn%|42~wTiYk(;^Yij(uQc)v4juC7FjU
zW#2T%*)yXuMf$jn8Pi7j-A(M4&*LL^ifkc<U~~tlBe2M}814g4kE63gMWt$NOesG<
zADbE=MUp(wLw8(#fAec@TkUVSApx%keA&rU_B$7So`adMV=jY!-YQ>mpOIhE-DnCr
zAu=2iV8WU3pk=@zC-Cmzivzhfz+Y5TtmT6C*VipAa98TBdwM+u#*sHWGE?$gwrrtf
z`b_PrNd7+cY#{6!OeB>Z42@5I7q<*1KI+{|D}0XYu;(%^L?e<uixxdKhfklK*qMB&
z-F*@|ew6(fP-4wj`XS$<y4d0_X292QPY#jhsxb%*xC3+s=U~`~^G2v^)|M*gWsCHU
z8dKV;(C+46SLo7N_cFmZi$luT*pt_U?2_xJmTkPl_uKiHUxTVFq%!&pz%syyCEN~Y
z-B1ZP9IS8uzHUDGdPsG(D&eA%_3@&MA+H4#(wh%2T7XqiEVLv5pAik*vDV2c%AKiU
zyWBRfa!yVudsGUXVG6wSZt<4+Kz`-W@aUt`G_qq8<x*q9`nXi(#EiVA*l-=wxEX&m
zoN{6$e!ha!jwxG&h`_@v{q%rC!}O9a4SaC`JC~WGIbnZ_v$F6mrpK#<oAzS_hglq|
z6l<_}0k>ii)(N?@S;iq&ly;I+7I;cQP6BS>+&B0F)_5go*^wl9*E^U>zTu7<nYdrK
zShZQyHacUZ<(Q>HJoQTnud)7lLb$fJ8A4YW6a!>Jcz3`b_K#-`YCwV?4P+qVcj#ig
zs|L^uWuAarhoTmZoV%8S0O~!p$Bp8(bx##V|KAuuBV(hGl)QCW`kO+2nHU?t5j$pS
zmwG5Y@qp<xmr0^I5p6USx%hctF25|WK46rEJMD=H!oSP-8WJvIUgrsWI$S9+DL{<z
z2=E{<wE<reEkOEVIrk7RJb?l(OJ*Fx^K#=QFWs)n2Fp^JvOqUsfKZ-l9vRyZw~RQB
z34?EowvxCIaZwh6F?1Ma0mH*QM5hc(Fu<TOm{>U<gKgqKw4WQ^f0UzguLF%aj&U~2
z?lEyqxkHM|*fdJ5r6-@IB-F;$*U9^z&wBYL{=cy*Q3Y>2!~fWLm&WM%xQUYnB(rKw
zV=uX5@v*Ej_leqX-yDMiGu1S4@u62hFD_ay5T*==<UN4(L)!om)^YElCPFUMX}iJ+
z>AFaO*C1<zI0#9db|(Nd5N(P=UZynmpotc&xxoSKZP2GdkIxRK_ZPjijya+1%^@a<
zzt_&S_C;D<!9`&rXcAUMpo$o$sdi0H3c}|va9~1u_(G!@us$Iqb+UO17U+N~#m4r0
zdeD|MK5kxDCt&iw=J&m;^q=UjaylSuq1F+Y@B>0lm>1TZ)U!FK@bj<1>+$eBJThoX
ziM`MZKChU}7GjROh`hedBw(O)+2qSyh&KsDs$Scue;V4!;MT%9d_@M)kQ&{(-eaE;
zntBV$0g&HoWinj_!8t=*DB4`aTG0wxN?TR?^0Fu*c7wG~0CEQ7;3-}pT+3V0;a-Lz
zFWxNFz6eNCx0ec9W_Hod7%8o%)Ct3p2;Y#nuo?tU@F_LVbNiN6DVH}M9)4$A;2fC3
z;dJq(l5NO8RLuYTXbBs*ut?KD>Si%R{_g34FyBw!Vvv3EGunF##ybGoTqu~@r$@66
zP~Jgffs-72F$OQ}zl9==Oe8cY^BEqE=3zjOrlS}f2Iv)FL@HaKahv$#%9YfDxH4WM
z)XPhdxL7+Mc#`9c5{_O|qn+Ks>-?Vn8KL0<148}671LZ;chInw50|P7y$*N^r0mV4
zG2!(8U}&&xV5;zb!FKh%ABB{>FPo-$ou8WF9xX#7Hhnkl<TriWLui)?IWv>IFFnOY
zZf<jhUpG;*$1288${3pZy-WhCwy!|qZC)Q(%CZYndu!QYe#YTHLT}7_%%5Xx;i#~%
z7-w}63%~)r%We0owi<5xifthwq*z!uugz1b?h<j#ap!gKM7LhTx)q-Ycu>^i>2!Kn
z*iz!f5jMes5o_yv?f@SYkSVDjNX5X-nXuKttbnDg=Rgq72d3QFVkrl#zP-tk*!N$w
z#ek`hJs17>mFJ7tVY-9W&&+F+RW?=bBwAa=(Bry~>n5i$#AXiFJvxqn9Cf#b&Dg_k
z%kNUxQuLTH7)|qWY~tfssU59X*ODO0`^aoP#@nih2=g9bbdOaWK{(+@6a{!5(7(&;
z&zwEm<^A#CPPK(6@EBO1GxHK!%To`RvKXeY;3T~7#3ZfpA%PwQ$|W>oVa?E(6e09T
zn3`khgi0^oo>-0K_vI6`kcuJsH*%5%g}>rX=73m5D2H3bFDY?$V;-$f%_$uY#T5C-
z=JTp&-)2-awobh-mQ(R|i*(Rn+U*$JO6LEqVdDK!lu@%rY2#;aH`}W`NG<nBA3T{B
zW9xqsRN<%0Hy>in)=OOzkpK${+Zt#MJsYRD;pZ=22lXASdLiI(YG$T=BH>GQ(R}@u
z!(LoI`(n<g+zS;ns}a>viXiVvWQgujYE+9NPc1y^qoZ<#6glDVkzDn&cqCJ~QE7We
z-o`uM<H_2BXGxhaBz`}===E!N`NNAIdKtO_!na2f2Wu_oDk|RS7Fi%GD{HT>mvjp&
zI2K|LAT@>?a5cf010hp^qFlKty>F*Le8?5z*Vpe?y3?@)!QsFfkcGus32`8*y7$iO
zM@tSLePWfc<D^IiQz}#>Y>8HmdeSX)PZeVgF4$JpryjMJz3FDX&&W4kZ7ON-=#DY(
zOADpl$rF#ldEPc#3<kI@4zIjaG!x&M`tmf>=1}QhrxkwsZegC}|Kvm(geT#aU)N8Z
zp*rmo&sbWs&ecOzwW0rKW5x6YHQPk0{o>Dz5!1tS<TdIzB)B=BSUV)a2`#U!FMR1C
zY=In$p1VQW_df9rz7V`n1YyvE+kiU(Ji86wu0!LqAMy%KG{PTv_c5P{L9eSdH9Gu8
zm`!?lc|G;mS#|V2SZ7zdAANV==8W4jKelR2w0Y|dhcy7115fQgZalO2mP5QsgWjVN
zUax}%dhKV5E~dB$7KzMZjIK1A{v~#qVlBXE>7q{-ll^vh&p4;L>?pat+4>zV>k=i6
zES)^zD>pBtVc|TT`LP!&HEBmH+ZR_KT6flP@~~C&l`C%90c!NGKASRq__R&%VxR+q
zajHET!&Sf!ZTRK}IFZFgGDfw7;Fybr6qEu|dw?8-+|<h+_dIhMSbjGSL6!xYN$4FV
zjSuo&o>RGr<|nN+`%B=(r3WTFENR1VTqW#w<*my<4fQlq)s-e5%STeUUY{3_7@HGt
z=MG+N(k(=G6bO?)D_4j$-rMO%osj4}*4{HRbXTkN>S=|&r_&V@iXuB#R@O~~zc}so
zv2ce^j<5NXpwn*s9kN@~eOk`$5ZKAv-s#jon)TgfiLi>D+FCBtkF6Q9;(Ye1p=!q^
z>L=cDadKXNn%(Z(zVCMIRznn*(3`&H@R#RW%!Tn4Z(rBJyT`<gX%GVK{0x#>ge`&u
zw!Rap=DDlB3@PCYJAS^)2MgM;z%=1eH{pQ*JdU?{ZWXV+2<z8Q{m-Y=!>Z;s_wwLa
zBVFKFKYIMQ^2Ao+;GBCBzj|Pg$%|c$oUb2?y^;udE&vPUwL1&)^U-zFaIz;)r1h75
zQ4&mS7$_Qga*5JeZ0C3KOW}>ca~AS)DJcVQC!DEN^}YkgqZ(rh?#7P}DSz;L{Qehx
zb^J>C-5qo2zb^%INXh6Nwsk*2VXCS#egw0#$KgC0eU`6ZZ*CG0|Efdx$A*qj{+jz`
z+Y<8o@WSx`O7jb~6R%uR1RzRykzFqJgavP!ezRHXZLH7GIBw#&_|RLF_4zWIvK&H1
zNh@pa%eo%CYkti@+{?%3MP=nYp>^vJcz^!>eFf*8*yG<Q3rLt_FLJH`M2!n}hRp<d
zCAXEH=0B2aeZ#6CL8XDVW54cLibhF|&5n{sKj;<F8PDsgu6oga#IG}Y<&w;JC}FMf
z=TPrjB5~Zg$FDH4^ZSP6iB)qLl;GH^pfvq`F2-Av%gj$(6?kiDjeRm!+_@#~*xBFN
zkrf-yuZa`Zpmj8?3(eTw=7|?2yoinN?YV`M(=rBo)#QfsJJR^kRu|b>>+@(M!~m+e
ziQ^`npPQ5Th2xn-xhNbkFaZIn=3X;J32fQ0=SAhlGP$t3muSCSYxrYGMzWJGxy@5D
zG!U_Fo!{>{z%+k%Fs~Llyn5$!$@OZreuM7wOC=9%PnrCo&mmEvxTLbqyYchStY75t
zF8cA{jYkKcw0U(u>R`NhEVMa5YpnQi5_z;?`C=)xgj>8_^i4x1lS<F`QK+HWz9OnP
z!zSv}^5MDE?7jT1g7p`_WoVOwXfNV^es<@O;^utq{&@#fL$5XVFt=%$@ZhJ~5rJhN
zbjIOmT3#OSEPaP2u$=V?IbS=3{GuaFkNGSIehhC8rM1uWBtMR<LZ)(BBjnq*8fqrp
zQr@Gd739)GG24I1h7Y$D#R;1Ewu<R(kz+r4SGQG5o?gm7;iE|kE{jk|Ntsb}JN@|=
z-7%PQHqz{?yY$v0xn(~eHcL)+#xOryySsj4UD^jKCHGEK7YPnXi^Z?7&n59Et24Im
zdO-5tMS9{wVVV>*B9J^ti##^3!dW4+Y{l#88+W7)nuAyXm%lG;I=_pJu9lWOr-YDl
zV<m!Wejjt2D2FB@)W70XFtQLU|0v2`HHmfKX=gASp%Q4Lz6RplS8HssG$cIufU}#u
zuV`BF{D5_~3aIbCsil=X$KJ3Rs6pk(mR=7;D79FXD%$VY<??p#-Bq8D7awRE<u&r_
zh#9+^*mzlPV@gVAalMOizSCdl3it&p=qGa-Q-a118)+`as!op!3Orv7)v-rZ)K5!4
z-cr&2-er7dN`9to`;4jA30<;**LkXl;gsoKa)Xa@cAF+H-MMV7dZ58EoyN!j6$~U>
zF2~|ZTZsf#A;b<Tx0kQej-E_8k?m%=I<=<@i!M+l@dX6{mSt@?yZ>x($`C==?kI*|
z!)bs2$cc&b{ChzO<%_I2D9g)S?m;9x6%!uq%am<~7afui8kj@qwI7upjsC0L@4-Y#
z#MxImW~<ohlC6xby%aelddhye3P?$(9DaEv`~9ggA{o`4cC#}2km}>mMX#)3zsIlm
zak-=ft^c%}9OFgJR;dAleYRWDClBu>kCjd<J(1($TpO&``t|CjmIoCG$u;+#OnK*Q
zetqk3ZmX97KKwgeO%}A6MR1o`hsavk&M`D5SUzC;#LBkFg7~!9!=SbLa7uIkes9x@
z1wFIxH+CWCR^GF0O9I*OCswRjK}6x6#<an)q2uUA@y^j0R~@}Da&BXrm`H|;hh;F#
zPdVlR2XxCp!-NOB?a6NkDl{PBdp<NS`hf1(`^e04XTO2FDkY^|bQf#X9YxW6Mz7BE
z2Fn*UI&w&;?Q-G_8+cM3-6&Nxn%FttxQiZsPeIYWynST(8C_kas(w-h&)$zl`w?JF
zF04s1`NboxIX@E58i)3#4IgtrB=hAwrk2*Q;NUrMw?6M+XUl6liBbwbX0J!+sfRMg
z%uwONfeB<|HZYxGK<;F@e)iSU0MjK-EXS_o22{Vq3eMVZO#5=|Iw2u?Rgn!JjB|On
zujF**+(^4~N2ARXaA~A$!)D~2pSjjkp^SOw*DR1zef-PBHf199V`^u8!5zHeig3Y{
zM=E1Y&+En_p0s_rBk;h_lf7nRQRAJ~WfCr@dYj%++|Q|s^vEm?j?4CO`88-N<tk-J
zYY(Yt$t*kh97Wduq3lh-sa*H>agDp8v`U5yMaoPek)a8tGLy=bA}K>CTIL3GXj@Bm
z<|J%|qDZDnB}=homsZP=DYcO-Wytiu-*<KP+27wezv1tCudaRe(CK~N=lKlx@OdO8
zWZw-H%oGgBSm-BZEAwo4QkRrnSExs*;hZ!#AktPr^Fj<P<GS@`Nqz4MT&x)-fz12e
zZ)wLEdM67@!n1ViP1oWhKi@Ux0nKLle9S1pF4!peVP;0m*VmW$zKKfq#G9kwSi7jl
z(?pDSy~K~SOzENz`l_=150~kgqJA76?yfD$IViNP;Y-}%*_0Z6S0tOdsTLjgb&34=
zOV)bTWdi(n6son$z6S4;I$?ZVePaB+S9`edv!{&XvA&vplQ!EYK4h@3i}$)lvwhSx
z<}MEkbj`^>%eA^Hl>Y2uL@rx!{dI44`K=DRb|}MyPtWeNSlD%><tPajQVImyPR0ol
z5wh|ID-fr2P3M1@<aR)$z0nkcJNkHbsTD2f1ZdG+^Xb}+7l$TMCu0Z0Cqh(@xLgDI
z_IdXvR3)fjGNsw$(iDo@3NlJ15k+WT-kxSYToHnRuZYb%d?WB6w(l<eoF&rf%|YjQ
ziS5e>=6~aEthf!)N=?Y9q26Vb)f-0RT{C&8=<-2k>R4dGZ|aHsP9N@EVf!d=P0npN
zu+76*E4fNX_K+bCn9`(eoFQX!g;mA8?f<GAiGaZv18i68S*L<CwGYD&P%|NDXS=X3
z2zoxgoR^0g_rzVZXSQxMEFM6<B7RKu^h_}Z15n7-`t|(L3-D@yI*2w79q6_F@JivP
zFIOX@W;*JLOwp%$BAG`XZM+=Q?Mp_dhq9hwZ`ic9;AVCoAEPaFu&!3g^5y$lk4v3B
z)A~o%HHHp}RQ;)QOF3n#>kq2pkL=A$-iI?!dtG{!;!#3Br#YEyZ8Gu42B#0bjJ|<O
zz6%nB%!p0<Vt4WR4mB#5;~JMF_e*p`;{P2cc)^7z8tRdTMfaKVQRC04CC@qM7t=)4
zQ0vINBcHgBK;?)VTIBbYOmSY3GKf*5<d9Jm9!FSe8)%Eg#KedwTwbn;k^+AG7|q$`
zl&1PKwuyaLboU&xwEXli7F5@hqTAT5i`_*~z%|*`r%jqR*<3`l>3c9bEG#R#go;QT
zP}_dW`_0nB)mtwHTn%u$Uy<JT8K1L6O5i?kb!@VNM0>*bw(sW(u8e1Vc!_PBc$i-N
z`(L(oC-0^u3H7Jg?T(CC_GFFY8Abi`(bZd37AB0Xbw4D1fg!OrJ5ScaGQr+?JE}vr
zgq;->q}}R8fqE_JLE0G5h;WR_on<z-<>8xOWeoQTRYO{#IVb+=mOu5B-^GZFiz6yT
za&CStd6U}{2qxWFD5A9%?$y2bOH-Pw;6i9Vf%M{69D|(Z$Av-PQQ#aDnsx6Oy5Fj?
zW{$Sg_Ae=;yE@M^Y~JFaE*QOcKr6#$XZLA`!D`frDk%Z)t_m6jjl3<L$6$%~+6r~_
zU10p8DVddGs8_8v(f5S0TCIlB_swsw?is-u$r~!{x~gM)CSr8=@SHm-gnNGwAc>&!
z^5QpLU3#=D%}b;fTNQsd_Lz_j42U>^5Io{$a(jDXE;fu-c$!XNxGU9RDd)5{l?>Ph
z02co;G&I~>FVNc2aSYw8TCxM4y9WV=a&_z#jfgn0MbFgn6oHJ$?`uyOLH@F`(%Mp5
zm_9HQZNF{Id}wG>#+2BTx{j)+r`e}&q--TlXg#mgQ1z>krBVW$Kkl<GyLX!1d90e+
z+A^2HI@B>3;<Z8Uw#Za{*Ih$~#EB`h?Nm+6)Ko1>lb(8_Jx|0X<6{4*dXJgw11=Uj
zOG3@2G{}kZ-~D1xr|VI88&Ebjp#+YXBvWmj35Y}#y$HqVVz7^4qQw8~|GpaXdk{IK
zc#rFmezLMK(OKX}TF$UJrb7i2B?%u$WBRsZT4Kj^(y=kn%)8%{9ubtLG>ts9)p(ZL
zfUO-c^YNp;Sow{XX?2#Una<d{)Y_Lr<yOq7OSJRNPX~=g1MFRisQZufnwPlTs39;p
zbRoS_Hbvewk>XLh-XoEB=!E5gk|zugUsO(7=LhNEz27u4FP(kZy+hZ6+V^eW+@>6+
zEvMb<-67e;A0H@b@MIo7@tJ!Ary|b<&Vk|>(rCn->}DA}F!J|N@!bphyK2IDk&&V@
z@r5xSj_U|<0`T)Gb!D}6RP@1S)lgK~MmZ3(A}s&RJ%nfk!D><mwJ2J>eynT1Dv{om
zZ$r1uwcXMZk>OWEhR5mtS6gO+zU<Z*xn}=^@5;jmr`fElFFjrbs9Bg8pZdL{mJ#O1
z&He7`8WY8-j@OId(@;JtmTFHt?Q}e%l2IYoKWS67#eDDWp7`Nc#ReL}7VY2P1?^ld
zXtp6U`uc~8Alr<wFR07A-#_NPf1GLl`z6;$`z$R_ByHeTjl5`B@kX3NStUa#fiVQ^
z&O!M*hx4JvK;tkp2>=I;ZL=!MAG|gc5VqSszCXsvyLk`E5>5Uvb_1GQ<lOPzMEkGx
zO4USTPpkp(T)ek$e02A3wOWx8qgm2XS{{;f8QCt31HD8r+jzMs=g?yN3JtNhauZ|T
zI=J<;)lAyH*zTD}|G}I^QfsRJcNOK!(_Ze@9S!4Ee`Z%Cj97N%pN(Zqh<T~0yb5TQ
zvQ0HSG<4F8|5$2szXGKxc{p4!HQBgWb84h;whFs`FR`_C#Z2J740Cv4?R(eqQba@)
zC?oXqN!e(C^YU=~g$^FDgbv&=jf;}6zzJgCbbilU(t)YfiZ%puRfEK<6FhUddV0`<
ze0R<{hj>cTKw_Wpxe8juiQq2v6nHn03(>Fu6v6~NPGdhdCuMfxdxihmW`QExzI#fe
z$@|XM{Og!+Oi3A&UuJ(Y{SCsH&pQeJ4*!@H?`fffr@hp!OzouHEwa_0vS}YPjCC<<
z|2|)Wt24INCz)M9|E^K_aRsGvad1eJ#?d_!gP<2tOVuoQN!p2W*s`LaU-*@TtDv|<
zrbK?rDbySLbDCplwRw3A6SiOFa}I))WZ15Kp0mHT_r}l1Q7>lhUr2{g$VvT>sx;<P
zZseI0P+lw&=edG1xci=T^g^flB3N&n_DtZ0w^XAXqT_A!nZ4au*e7Z<^Kr7J&HFvC
zz)GVahh2L2rK~hlD*Ur{rkh`%W8E8CovPnCdzCIQn-pyxPV??Ketodc87)6!VS+tl
zGU_zDAjE6b^UkrIC?mJgZqK-Pu|A*t5V86MwxM4;2f#@I27#JNHhKf4>2^xCOr~^v
zsSzLw*{DouoZHy$7y3?3J~yS8yzSqN<BgB-b+~&sS<&Yopber1W_Xe>9P!C(q6#iS
z0;2;;g{Fw;1;2d3AXZ*<Mnof+xu>A!ShB9JM`>G@O<_y##0_x7_YUWiRb;T6na&oy
z=Lbq}Ni}}X9h~vj2zZO^Tv!AQa)B%Qz<;0MzZ<{}osJ(SN4s6$sVDcgEt4K%Oo_e8
z<PqUFYfqbco3-UVecgb&S?t|wuRnS`!9LWOtoMqr<jc63>|}gX86TKWPdKKId8=FR
z#39vK)QYS0O!;5}YaGdtUJ@_)s{ljN03)eTHz69`J)9568iZh|3pw~6&pf>R5tVuM
zoGt$5eqi_(&pEdiNp-Zfb=D^;jZaOez<vYS`|)BS9FXPnc{Z%1hyv}(&6B-IoMkUz
z!)8KOd5<-S<1BPbTH1e!uo1=Du=;ss!{<eVGEeGi3j<sAqn1bssC3DcS_i3Xpo>pQ
zNi+whb69=i>s@Z4v74v8He9~#=v^tUrc!-W<L(#J9lghr=;^nD+tUIHR-y*WoDNKC
z>rB^~DIqdfQZ_1ngZLM69nV$`cO9WE+JxF_l|t+>@PgP4^Wj#<)oci0=n9Z|wx(iI
z9f~pZPf)f1v8daP0^21Cr#&fqQx*pWuy9~@Cj#O%SOF-Y74Pp6JSkiiFmVCR(*iiy
zIZy!N9b$OKe%&*Y6cheh=2Hzx;#!~TA8!heB*`y*Uz=b0sK(E7R5i<d^slt$CZ_q*
zK?aXh@Z^e`y25|pmB-X4m=4hg1BNy?NT*Z`Ub|CNbc?Poe566AM5X!{4b?23ftK8<
z)(%>JxL{w39Szx4cBo^}Ka4&`&;Grls}kfX_y^3s!h_=DtSutArDc7ZEQL}zPa;J`
zkHV)1-!>#aNZ#m@k-woiLjQ}NYSCP*1QaVawzdFUrK2Q_(|``C8VXrprSX#|?dhC~
zf?)m}WUub-b+`$E4DG~<%#4;+cj8QxxYc*a=;b06F3ZWwtLQ%tXIG_Nc(w}{!OZse
z_s2UQ6y3(0F3REhaIdq++iO#2kARwmg+Cjp^oXc`;Va86C#L`6iM2l2VL1|%mc#Ps
z{e>QV(pabawDKL5jN7h>D?FV2i{lcNt31#%Q`1kgM=vpnk7aRTw)Im2`l;o=^q;le
zP4*nhS(uBD*o-rtyc@YWD_(LA7xUu~RSc~WHZOKP`j3E)M5$cu#$6({CVUGp1;Vay
zmnmNuvAFn=-OneFCVRY!x8*EKbB^qjKrs;uzz+?d<AGAKcKG>2)C~WE%FzgT{-VM~
zSdY|_*Ebdct{&Rg79?}m^6tS%rP|4DLrdn<I}C_DpZJIR?TYc1sf7(^2X;B1o&sk3
z{4~3`=408X<hQb3YxqS>8cNnjyLv>l9qMrXxMjf=F_luSBzm#S=f1i1ELvm!k|g{o
z)5gl%JII$eltf9gkq3#HxC)CO6da0Xl<1{q2!2~Pdmd$u1>$|kOL2Nzz*zWI>f^;N
zSHZhW2=j@2-trR;lBGS^_~xuF{#_Cia@>RIn@7pY4KtAE?^xT?t9MIRx3*c;FikTL
z!~i-|IeS}M61)WfcK7KOcHNCz&R<t>%2oQTd|q3c&EIFed7!j@sni-BADQl;S3R;y
zP64I!>Y0s^0Uoz|t_*8`lkXkvDQ$kkTHHUl<P)DOu?=PaB%7{6Mw<S5N0zQGD~rJA
zZFNF&))t>w3#dbuwrG80pEp(9kmMv>k{H&;e?44g*o??0UoY$b{%6|OhwAG7VU*oR
z@SqY^z6P!G2kc-Vd?;Pew_qYEPfg5sW+qOxh@~Z{m;KH3+Ez5HnbJt!Xq}JmnOJ(j
zbo|Fo4gWX<Y*kOqFk0i;at3#>+@l5Iw6Ttl6cxGK)Q9iDy6pYJk^>EKx3lME`j)v2
zrB#$*tG%}e8YZil;%G<a^bRhr=CcTrxgD@U=?yYT&8`65#2rMBr&xHw%F<FT2GILP
z9-{NiEMB2U!MI^pHJZTth?P<(O=6GXoHvdTA-CNxp4_*&J723lL;eN{1!}=j`7s9-
zfJp;=C5O4ic7~wnwQ6XlG=c2(>b9OYSCxvsqRe&hYA8Y?TIhJ^LrY8BMdJG3)Rp+M
zPqXEsWz-D|J{)~gr{tQOE8X)yPc8cp`~Dj?d1@0`X*Sn!Ha|6PmiBNC@ckH=(C1zC
zOWh3vs#P?5blqgn)2*yIPBYm!x%74I&pm>>LMpReu&{tJLA2Z{Ld;0Jz|eYNtgG9%
zA@<mVg7V>{^+tOpt1ux4B`N;o;Fci}?TK7vY1ymZrK`s2vQl|F#_k@J2QKildk&aj
zF93;1tDBEj6b+Y7dOg3BSdF9vBTo7OWRM_yi{w9%;19g7-rnY;d~)5p%Y)+wQ>@hw
z1(bcV>U%OTq3kyD)sg|1{BzsOE<Nvj@;(@;8$HM!g-t0beQ)E$G?E8j9S`uu>-GdB
zyt+M9kip&&OJMXx?M4TkrBe4zs}N8|JCJynB!VR-r5;3-Ry9146sBtUtE2(be7%Ik
z+^-4|BJ<Dl2-kXER*(HPTLz;N@D&4eYHQ=x+_h1}t)Pfo=!S8xGLQVNJlQqOVNR^d
zJ=bWf`p~Wr{}R3}J~FWa=(Y%lI2ekVhEkOBc7^+X)1Wjtne*wjpI5Q=y3ucJnoQbD
zE!pz7x^q@uq8p2uw69Cy;S;{%M?ygal4)ur3hXPnm)zGAVYKMgBRBJH$W&b~?0c$b
z(8gWuR=Mbb27caJ#@LRJ=S~Gq)do)#{!*Dzsgb4iqh?KWT<{F--B<6rzdrOIv!W^2
zFmM*79+h0428)|4wY28lBKJs^kJBz|{C%gDQia__zrntNo<oZlmr-%Z#p5c}*M2&r
zTLTC-45HA;<saxT_i<WCVBYvK+6~-AV2h<Cmaa?v>QV+iKJ+zskgp$9SBH12qX10n
z6JK<pD19!a=^5_GC1>H*$8}X;tP~QK5&?>J4^-W`+qgN!<{q-Okx$n6vr$KaRE_NH
zsc`ljH=Dfex45&Xwe--5?Gv__Xr+~ZW~IDhM~(Ijbjv5xeGP+J)@eGBf2dElrK9GO
zAp{qeX1fT-#AoVV>-0c8{)O3c1K(J3G1BtTB=|A_?PwaI6oqgVAc%7A6sl5qJ&?+&
zntMGZY$|4cf(b6`4K955tMqhE0dq|3ar>%($GFwm)l@7-A0??`{jsaMy3Y{Vam9?P
zA>dr<FGe*RL9v~Njk;O2KJEDW{quN(&pMaI`QvDdMGsDSH7LEwIbp05cbd)3-BVsh
z%XQ5yy%XX#F{k&*tU_vr_QaXV0WTKqmVi;v*>9>No03OfeGFXJdrjX-YmJExb&t?F
zjob?3JinHw=JZLw?Yd`{&wR-AR$XE<_2yb<=__D1UZa4=I_33D(NLlkg^2I2T_GjR
zf!zA<Wem-!W?PAh4VUTy2Fq^#-PP1iF%c_+`5^5}4z`><k+hIN+4Qd&UnD>Gr#3Vs
zm>GT6R-T_5WDT(UDR#Y*a>f+;&WsM?AE1_rJEe>w!Geck#(&+hDf71U5_DdEm3B<?
z+n3i223XVORvI-h|BrR43USW4g*DoSHwhnzO<yeKu|>u?IdLV|=isTt*~2Gk>`~o|
zqO|s9vfC1GA8&|@$cx-o`DuS((U727J54>8mkY4=yeWxKtB6r?vQ&H>-ARhuVN*T_
zFcp<{xGRHGXaLj>6UT-?YExbMu%Q09AX$G&Y*p~Q`}v&KbM%-9<0!ufC>#9<-dIaZ
z3p+&8z-;tPWJG)DK<G5;7^dr!jXWsf$hnL^qgp!apcJokAHb4sugvK*-@hBobqHhs
z(O~{{U#i-r$95@Wlg}^o(viyYC6OzoYb)uZd}d<PopGOQx6K-4-@2=5I&EV@wMtf2
z8d{&Z_n7{*hBrA(?U~*!vnHdevODI}=e%IGd33OK%6Sk0S7o_GNk;)vfUE$Zl1MH4
ziSV#rVzvyn!*z6xd($+?E#A~?&}?yaonzeQs16qJP{3xMzP{mzXZ1(mseW=!3p^%|
z^h!!<M5Z)~fG~Rh_pJzvXKZ(qV`XrNoQ#W=gBapjHqTBgpAK02*U!8)+Ov;as!wy&
zF8P-GA3xp*iIf8UwcDw#G3Lhq9-0tkpX#10F0=K`S(p1RGMFht&AX_KA2YQ>ydI_t
zmwA{y|HK=-UuR~*_)&A(yMmyG5}9g#lMNDi_c9AUu4B}2ASFr94|N-PkXLM5s=4LS
ziC7wl)S^)ca{$aky`a~uDrr#ERDvrl(R)IhIsSe(S)0(;>6u=Utb4AP=^%&(g_$-5
zLW<bYKie)>*7Kux3AE|cQ}C?>O2~~62<8vO)WZT$T&UlNcg}7R9IwCFqOiUs$$a}n
zox^bMbgAdK7{Kn}1UC^y`6Jdr{$6OZjs8drco=vf$e-=N3JlEuE0o47rkz}GNFTab
zH@RwsT*vB!xTVtbc3+~;5whT4*^%ZN(#N9f9if&RYAm1y>00cVdaT#?*FUi&x@&UQ
zBCp6t$l!#2MFvp#pq{FC<U;t#(QNj($=r~Jv<FLp9t4a|Z1<7-hjD)=2Q#hN$a5m;
zdHE6!!y$@*fUz1Id9F(@L8n>WQQOugO0-|qGv5ox?PvRgJ3;Oswd3~z7K@0UW2F17
z0Gs}B{sv(>Yfufx90f4sD>==bcZ;;AU!^*TJn(?YuG0VEXwU7^4dq4@Bb}hZZZ$It
z^|Ie6#$|2nnz{)$8lrs4Lz37EkF#xmK)rlBAid5|Id1R|FSQ3E9t-H+WoDzfD3&hI
ze61btPM^&>>KpB)mhp0flf*-lsa{$GD~BmMGtm-kk3*n$Z%>#*1Z>~gzw^lTKdfpx
zw-XIRLy3ntHh@U&94=}+i;Q6VbId_Zqb)_hi7yEUKY^#dIT<K$;H~OXu5-#^oT#_-
z>4GejsEg(j<a9?zH%iyw>3;iRu<onUw*?bH0%XaPjr3~PO&R0-6_nJdf0b9@hZoW}
z6Lb<tGg%_lfg2gwDmC;*VQq^1&B^c^MwC)Hm4C?*(4+pR93if=r|?DvpHWbi-l1R9
z8~G`HU7xP7ttebelZ_>wj1AVYb~uIXfaWE0=ccSFHFVIi$C$?@E-<(plZ(`au4Ky?
z^MK7@uJm05EXtIx?sSp5j1B;a<X4W{sdo>40PGF&W_)~j9dz%6yc<zW|Hd5Dd#5rB
zwuv4%V)7GjJqwWow_e|f1`Y<`#kRIoS2IFJ1y142Ry~ngt(iT#7AtI9Aj3R)KgrdC
zUr%k&Ct19A%lPn8amYky_J;=EO$@Nw&;7slq6K#98tF~dv0rU2vI|Cga>9#LI)rCy
zhh|T?S6}|P;_uKXlYFlsJHNDsQTO&@K10{S<?^vV$|hCImHd|^#1n`V#)Se<hy28p
zPw46Evw=5kHDbNX&s~m)$lcf_A;QHzD$+QjW>f+qnp9!L5=A+2pC2wXMT$UIiwEK|
z6oLVPj{gFn{UvPVgPfzf1#u?O*>?{g)ZwsYU%!400rSap4Fg{+9mPn|fygLZdI@=p
zI+$&g-Pq1hWUH=@LE$gD?TcptyY)$(hhVSXXE{$q5YzJ^?8us*8l3-2=bGm2&sJII
z`b&cEFITG<(2FPRhBs}YZkCzXv7!2wHdQm_rHL}Q-N-2?ji<6p{&@YWD93DZSI_yp
zJT{jFx_2_N0cgZ6PS`qkhz@^!ml1KTE+MxgfJ#!sF%okecVeoo{;g)?{I7#&<G}zD
zKbd{&4J6{kqe5_X@YA1Q1zsQ75+wrGv(9mDt*0P@w(PY<G8kP5u=0eeS`u9|b=iJe
zU)yN!^G=X0)a?`VI~`@(L}UM>5C(+Tt2#O8RZ~ZXhXrA@VxP^j+qCHdNeSnBwgt`F
zdoQi&;U_mk9rC1$=Ap`4OZ>{UuQ?E>;=(_GtFRLWYCQ8GS%G!~H7$BiTxw>pLg+Q#
zBX29^F#mZ+Na+seJ6;6o1c-|qbBc=6u=ySTE;46)o2zHq<N1_SEfK=bp)rDI0dq*)
zkB6Qe1QV)fPZ6wGdu2>SuH(Z3;-(Z1PW0v^8h^5?g0Y=CJd-x_v0>w~)>QlQPp6gB
zcy&`H+-py>%iOJ~ogNDwd^pXXVQC#b9=h3BK=b1%>)(yGU+e6dr%%<PPx+}+{h2ns
z%q*UOTNm5EdlH@AZoGtS$OvN<<6F#$EtJwb0!tZqRU|7i5i%d(!=5pKdC#ki9xQO)
z(bAH$voobHWJ&`=OUi~Io}Gol0Y9b?2q0YSSpK5C1*9ke#CCShL0Lb}?&U^(uqoAu
z183q~1a$O?wnjwk&F+8bYJy99fyN+J;>GdXbxm&>ulo`_R)*t0cJTIZz-u@4MlY7C
zzVmM#oxfYb!~deALxvm?3T&iQ@yZP?liFQAbK7EV)|A=OZv{`!Mhzt=P><6hWJ5RU
z<z48oohGQI%aR6;=8@RdKoFoIL0emO-v~L;Hc8R|O)P*Yj<zyWnnWh+%a@S#5rhc%
z9KE4h<OkLrGEK@Z($-ism%K8OXz;}1JUJR2lz#*;i39*D704c`e5o&IE~l7Mbt4Ud
zX0KO6c|eL*K-nC;3LH|dyLSJciRnwU^sLX;yG&Lsl1^9usnUTqR-lz5j77E3n4h?X
zEv)}eak<fTt>(O`L2>U&MyZxZf!9)*I3&JrNPLnTs^q{q-AP%BVvJ(iFynnAe9cLm
zXVb-9qtPVK02t$99Ri)DrF7Jj5S)8kRnJ;*YuMGWZh}}M#`YU8OlaxfjI0I8EGC&X
zL7of(VWbEG263ae_^t6{D~rxSj{+o@yjX~Ig<=7qyTh+S2Z90uF1fKV&)7kbgkv{6
z6@4mSA0LNmfZGk8vRw?1uiXP;+ZWI2MOyRdF#5axz432?CF)#k4$~u1*hnW}fc;$I
zi+!hEJg!3NV^=^Fzz&|8ss1PyR#%Q`XA_ZuWAldtO|H2iyl<MitJfE!z?fY5_!y~X
z#G3&!;%&;r^-#uw$ac^6He1*|z%)m+K*2t?Rtpmn>(`^QB{Cd*8-zB{%E}7!@)3+5
z7GUs*V3_zkhQ#fckFa9I%dhq<6#@*IY6J`bz$FlPMD}6CH9|amAZ`B!Y@w)3#7jwo
zFHV*?SX`EtPSM18Er762hX98T`k-}UHs8L~`TlYKmd4@zerKQ5P5!=JeIj(}rj&|~
zJ&0@C(KU(ho!{(RB-K~-YUexv)qPrfOEeCJfpK8P6!ALwem?yrYLQHI#;~`+z+ed5
zhs?!Z-#CJny{49ycH$l6)H}G1&{6jee(6>}P9r}1ZJY)ZjVK9h<Ce2Ht566NCsutK
z$W>6|D20h!54MGX<T>{_HJy0TV$!VYR9=KbM2P<IlPF<mjwFCeX=3vA-~G@=u{azH
zLa0c>vSbyVioBs?+tTZq4*;4|L)f+uS4EysC0RGeG4ItbjolH>%T4%M=cX;EzcN#H
zV>K4A+gc4Q)c&+S_^4zm1TjyuW3;Dvu(;mMn3Ml`#X9>7ua4_vRHU{U(%Z_~o&?D#
z$YxcfJ`c)Q^P5~evVT3ra!K%M(Xh!u)u5JRG<Fr)(*Y$7qbHMBIbvtdxut4IeH{{k
zndnK`>;_zCxxHBxVH(<wI>!rmAWnn}1ge5=tLh)CfV`i)zY`b*=bBf}I46->QYjGf
zGxa@G-Lag2G!gXGMF^sb5YFqNraFSZqkQd=rwBDiP0U$RVyhB@t6ibTC0sdkp`agF
zH7->fkU8)?Zgl$Bbrwn~6)r8O0v3Q=7BNyh{?y#Pbn3KM5X(_rLn+4{*cT@_n67gz
zNLeZxoY-RN)91sjAf#AiK%K198Ot(?e-5;A<kzmxk4}w-S1k#%jQZN-`hkQ2OczU>
zyPYRXTKfU|;n~P@toy)seA}VO2TsF2w~E|Y5DycxdA!Olk=O_^J%VF`M5p-J(YED!
z|5kfr67<PsgY}9JDCc%xK3_RS3p!(hcx`WI#KHPJ7Kn2CT$<E808VFu&<)@+Mby5S
z?SJbH=F2NrxZ+3v2k1_skEw^9TL@K9sd?LJcB#^)l&SasJ`tP4Xz8LQ5zof#Zukw>
z=}pUNop>goQVti8SLfrN+Y&n&PjL55>Z;d7Z7O?tQ*`ORyFzA0UpMEztTb3(Um}CL
z-&6C4>z-&(5!x%3+M~KbkLM{OWU{PgIUo#A5!_V=y`j=f(^=b~Kfg=x>T%kjRZ0dc
zFp`u|O^YL8;M{WnfUjr`i6jVi9l|e2ks<5q|DIIK9~sf`8>!t!Fu~?flMxP^V`+)$
zm4aqL7hOJx-g!GbSl34}G8#Ml*8{hnPfrKy2yL71K&6^hTgLZWD(de4me1eQMuy+1
zPo35~VWqB-%xYU@p(tN!{pYpEz^w(h-$SB5d+%VP%w~zaMKcc=HRk2YkB(}~(h#@j
z9Jw*L38jq5@#e)S_L@(1lXL@HA|0~%hXp(`(T#FZ7bFdOY`IU~Kg?{FB5F5@tp`PF
z5y?;p;i1*(E~MjXOr&?GX!g;&+v6oc(?D{ix)JgTT(b)~s&LFg^m*Vq$OMaxT){QF
zwCkHfA+<J~*0GUcI@+f1_tX_)OnWEca6&lIkybG~+5pAGnaSHzqdhexlRpl>0Vts2
zcOI(D4ZZmx?0sczJ1;VFPQWFQT_N{tV9}2g5}8_?%>AVou(q+&-zP9{?(bzTlg84*
zyW+AMB1LF|E`*2^^&OB~@J^tZxYkBx#`xU%nd=jHl!Bq~*ug1M^CUeVV{*qMU^>7v
z5XOkk7NimejDTIj!*qRqEK*l@Rc8NI?Ac1Y-8lB;ZDCG_gqPMY5l$RsZ}AQD{P3mh
zo7m*Not=)`=hJzkx25*X-dPz=ixwKDvlRZ8@BPlADJi4`jKAxRu|5euNP;=n??sw9
zlZR<u4v6K!i7%dyxeN^ImYnNiejjg$jgUQG9saTJ^Ma5~y;?DI1%l5WA9owc8~8n?
za`56_a7>6>+)Ch82mzRaOBfDrpTRxJ!j@AwY$R8Ki$`Cp>hb9jE&!t(@Fn0P(D!48
zH7-vIcMx>VAf>69#R=J;pWm&4niDJ&sv2M4(qV#DHOj#QJd}WVECKOgD?=8joY8vn
z{*goc5Pf1iq28FsS6W-gvENT#qVc|Om}6R~{4d4Ce+9J01dVh~(n~`brTaolh3Um*
zZM1vR^PX*y@sKLI@p9eNk9W)DPuVA0J<3p~q?Tp2=T`GJR+J1YC$y3<p3ap_Y2^(0
zx%B{ATFz20>`2MjW8ThlAtItZeW3pX_BP2ulVaKlDysxeTY{58!$xuzu-_8b79c0W
z=U~h8ZV#t=?^_$&9n&2`Uc#K&G1N^^?1XLxA&gMl1Gq*f2XWbm)+L97?19LLPYQT@
z@}?;LoaSG}d*LuqHN@|E?)tQFg@{@E$h!_9Yz5|XYjZm(%@uwS+y4jPRdws*w3cyc
z2ZL8y;7yuM<!LXS8|kQlK1^18@hSUSaMXcfp}604D8uKzw8t;>8~L>b`c7#Hk`jK-
zf+H_G{(<kw5aotM4TJdy$=U2YWDrl+GRGNUM;w8I<f;OLtVl6E2<e)*xtryN9t?j?
zweG=HAB1Xo3U#T_fwyOE+w0H^FH9t3`lqa->V8bzL06OBNYuLMXMl6-Elk)3g$>jg
z(osFbJH=EaCGz;2RePU(L+=DKo3!vTt|$+clvS^B9@EtEXLj}&PNsfWEY>c|J?*t=
zU7A-pB)oD;Gn=c=dy7u~!O`F&)(_kv)SvgGTEqKV(*pin<S|?2#JZARKk$X|*?o(Q
zzzUY^)~t$tgYb8v77aSX^SL_OD5mQ-%6dj5+BLr}wEF#Y<Hz3S0Yb?f|3JHPk<A=q
zHyMDw8=TTy*TO7i)M~K3LNO|A6Db%%p^IDujkZ=<F{&UNMPfh<C*2cczBnpa_S<VD
zHxVpwfWyLXU4R0(b`$<%Bm&fcu+<Q#DADTy1tMft7~EM+2pxp`S5Qpv|CTMoqFLu(
zt(f<lbfLo@YH-c{i%RVQ{l)$jYIeHn)apxCi0&M<`cIDMe`#%>P0PIuewMG7e#$r~
zUr*!Nag(y;dhO8?*5cDPvexqxe6^`1GAm6NeBEu)K3=mexcY&bSj)1Y+|BiAdv8Wu
z@#51%&k7=ze>a8B2r{~K4**s>%AVeSho4uE0RIEk2}KzYd}Kgi9UK`#C!|Ag3yuXM
zd0Quf4$8{|o$;_>L~<Kh52sp)5Y7ZnqyXSUS71PtLm)qB`T-C3KNRj~9~8$bfMA39
z#;06iXng%nMp8EJ6vMC*LYo3H!pn|4ic!NU&EKbtJz3{g)zs~d`!5zU=*z@!h>U61
zUtPSdC9|S`$of!Q*Qh&1ytfZ(S>rS1dQ^~F-Y0c+1^&8U?_oCct+DMN03b*yG(jF3
z9;5)<C5inuMyqo>94yHXHWwkSLxdpS3up#!j`EM{U_uJs(drwlNaG^2VkT`*vqk$R
z!)8@z-y!!$en8nzbbt8pVE7OxKyd1$*OUwn%|)qWp$#e9LZ@M}uG<%6+Y&v%RUyCa
z{n)!n$05Xrk+!f4d>tL3-yL(R(wQbZwDSc7{#gRNaIR}PP(MjcbFa2__xJh^h5uzc
zGj$C)VY48`;&RY|TJ*i3KrS*^kps4HlHE}rO4y-$CekkO(7tNBn@_U5>`-oKQgq&)
zg?Ek+rq~tMUo@A<l0Qt?JU=<PB*Dm8zY&-20SxZ{eppQj>@GlJyTUnK00>Bo;6=Fr
zrv~aZ5b5T=zy;TM>@lPmbbQ2n1N|j<FX;LSWk#wItW4<rZCVmKj{`he=ow&;!lL+c
ztyExuD#U{rf#2382_9duM8Ava1CRT;ZI7r6o!eSp8Tj_j%h8^{J!XHXgHMCWb4|(m
zFxp6~A9dZD%{zlSN(DQ@<9(+!9{bL5(!07?U_x77Hfzd8*%}n!8kWJJU`Uw94yP*C
zI*EtNCvE&_UB4S%tM2dXYaXaVTri54WUGyYIS@BP1AZZu!YEL6O)W-Mu_m%ovBz%X
z#%S(9Fe8`(qQy&wehHpk2(k0#WmbksL70diY2&W3U*DWuA1oylR|`@TxOkiXo{8aM
z-0=pb;)Lc>Zk~vUJj0f=xWa@OG2y2TV2{N1mSP|7V&pMmD+AR%LuT)aXjeZHx<ni4
zh+QNdTzTxYIP08evr+(P&K7-#zs2#ZhDD5vI~2G4EFoH}pG;Y`dN;Wq*dnaouDGh9
z$9t8~fdf@~WlK((E5eT@ms`=6RMD1QVJoHZyi;h8M5gZsr(~~!nIe&+s}`L2OsGx?
zdYhy^A)d;W7Xw1ke6J|S09)69`4Psm#fHrQib+@Pg>H2$#*XEWP|Wjl6K_*)i?<OB
zGaWTwOB|0(ZkU6nDxtIlAkKjU@GE#RLezny0i_dMt6Vj8D5oU9{FyK<<Ss}Qjet*q
z3*c8gq%*>VFLVyTVb8Z6)1lnUq=HBIJhw8Lg90p?kWr9Z)?Z<?>NTIcp44gn+g(bA
z2EGw(mLjErfANqs-vPBHp$V^2+#RBp0PloM@pRF-Cy!fd_qj2cAu1iiXoN3MD^@rp
zd(9BuV&gTz)&x<esupFv`DD^Uw5y;&Na}c*q|rhUy@U!mO{-}oWC7CtX;QmhVoVgE
z0-~tG6+IlQJz(5|{s3DtO=MmKi1g-UC>ls0zW)vr*dM5;gKR)F7q&PV148#&dw~H^
z3}SWcMLg)X%TGp)0*9jxd|nz<<a>$UxHiMaRefTTzjtRP#HSAD1Gf0C_?t!gFRgK-
ze2V<QN#pi686tC{rYefc_dK{`RfPVC^&RT#QJ<}vkH*gTx|$rC$D&7mx;<RmY+c!F
z8RN#7_~AR5Cf=(Z$x_YBldUl(nQNg-D>`K<j(D%kQi{0KaG!J(B-vN%Z@QY>p|pH;
z991E#nM8NEfmWw)Qf60JQ*=A#Q4#z5-Yub46=bStvSK}(k2_vGMYMEu02=OCNUsi<
zkhoX}4vKUrkZ{qQB0~ltpSm0~+GCWnu7-hF_4fL->m_=}4)uxToyE3F?p6P)As~eq
zVQm08qH)$ARVUTGvf|gc)z{w*u?kjnicZ+*ee_DmiDSO)9nA0&kAnD^#&3KyeM7Hn
zve&n>?#vA3j_;JFLtVVNrPTUWSu!D#3oJq>)p=*#e5_#n(wmR2Jl^V6^jf-05@Fn=
zi0-lBku+LF`l2vsx?*FlUEwiw^^mk^iQQ=aOi(7!-^Jl((lp(Zq%RkMC*;cQ3wGRV
zix?cf1k@2t*d)Im1u^o;M%;yl)QPDmMcNa{ACPMoX&h?peC8>lYbqHW2SxBnMi@a9
z0Y5^&T;6uojnNfw#X(Rj`%DPj2#1KE(NKTmrpx)d0s*!BijwU!$?_JaF*SdKNdM6?
zqmo#6e6@#m$puE#%Wm0@E%MI0a3;LOm7MiL>E~Zv@lwmJkh@zc3Lo4_sdGPqzNW6B
zn1<fGhEhzSM5eF1_*5b^elJR-qgerww>|+#0g6&O+GVg9`=ch}jxu41gEr0h2dGz6
z;3rgj`oChF#>P}57`t}AJg!1GG+O!zUtp@y?kc0`7UJqk!G&Be-U^d>)~`BV^08kf
z$mN5IiC7@W_m#5(P-DTM4v`q36)^Q6CD$$j>i{QsQjtUVfJVmgVwuJ{k-JH?<DU3S
zwe<E6wN`T(iGoH8pSGRn-fDmDzm?&C^UkZ!IH{LiF*RCgv{23ae&YiAF0SiEn#swo
z+OinP)V7#;Z`4Pf+K~!Pg-=ze`$WBqsTtQE^Q<Ko8D|c6aaocWA5@Pt2PraeuoIUm
z)OyO-%mW1kyd5wIocYI9$N{oC0gn(hAXvw*n2f`xr>wsANrP)S*^l|zKON$S0J!lP
zAs7Zehw2+`8RkqAoI-f`FE~I)u?(OUpr9obrY}w_Q<@gMR$UsE>3UBZBaNt(05q!j
zv((bh*|-F6V~#Zj4f6TlO?lvc_~pbC-xdGEMMs?-O!e?*e;z$FF<2MwHujw?z(q`S
z^wcME2ep_9ucnhK+n8zc%H>S<he1T48YR6gGSQ+%x6HOq_LTZQ^>%Q|e>4HB=p{;T
z*UfVe^_IoW5s`Ap@bQ2)&R_-dT}>U4TvVA)aZyf}x}tv`+EoE4gHcH#v`*?^EjVKv
z(2bLD6Wiw%Dvc<K+`J`-#;4TDAW}&<z2-;XkdE(kM6a4^M06$S*HMWLpOfEJkEMpJ
zh`3woY5ZJdq#G`<E5Y3%tiX`SkpJ@M`dBGgchpdF4(a%ob@oUZBzfocKke)}s;|}S
z0}D@IsUb=YufNEC4Z=uj_tn)#LEQ|E!HceU#!snP{VHVJJ9+ub=sICD9Rm}Mu`6ss
zim(k*1ZL`$e0saPdBY4<^Je8Oy=#ZIBjxUGl?-=dgbOcFocZhJSJp>Mrn!cHROX9$
zmUt5z1*Io-gs&;HC}%HRrHWZDurl<Xy0b_j`fJE|E7XHH2AR?(mbg=|1?4i-Qp8WH
z35uBU)NXY`{=R<$A}~>M6V^&B2l_Igw|J0xKr?RS>D{y=49YCUaG<y6eueR&rC<3s
z`PBpjij&*vSjAuQ%KqxQi9YkxCIyv@l<Y1ZpS#Hi?5x+$8uI7mK6k*1q&@50H(0#7
zrmk_B_0ul@!eKEK=`mhEh`Al=hpzrIv`EUC`;tB5Bmd0tJ`1=@GrWAHT*DJeSF##I
z;X(S8id3tk!}R5m88T>HX4Z1{(~oBBY*j<2UF&wCRMyQA2*bSw2?ER38-Pk9%W=rW
zP;od2GCp_#Zv|LbY3DpiI=PxOaRmj8ZlsPHbp>Grh|$MlaOyBo2rMC8I&do?V7#h&
z6Mu~h@akQy3g*qKzh##-5t3vA6K~sGd``<{Td=p=<#_^+;86=Ucimo=`~TRgTOq|d
z9*$det=`hd*}Js<Qt|ql@+A83$k1&575k9rB{D-xq&-TdJ+}EQpqE}`$i9R8>6`mq
z!-;e0$H&GKJEnFqn{)X*^&q+wc4HXCf1SSoxnXf+OD8$`0*hf-M=Fz`pHh!qw=F8;
zF~<smYH5FSPu+S>(X|RiR~w<B-I@yM;zq~NF~kMc)pgaz8&*zCUVQNGS3uje6Q(fE
zAeoa^DP~6N2~~{DL1=dLTw3|c*k`uw5^YtFg(kJ@|5tn7yO93ug>vF8*N<NK^zMCU
zO05&P-7IwU_E7zIyGq?JJ(cYXfbWz|nxSN7aYo3tD@Kt@b8ErC%YlK<%9Dkaa)&e}
z-pq29r%0l<HWG;Mu1mN@;_413@Ts=Ta79x8LjaMS8bK|Y$QfwMfl2SoX+94tuHi~j
zaTCUwS|clMXg;yr_uwHytR{fM9eV?m92)Xx5GDZOJE0n%&&bq(0=s=vOvI-rk=|R;
zqVZ#7dnkYqvf%jcmnz;nZ`!c~vmrDt(YCZ`Nz{IS@-d->|3xq@^dbFk0RDdvR1Pz8
z{31q1!+b{VioETL6wEadaZ{MI(cj{$S5@w}&t#%(ughJPr|KuqkF1y2x_sVJsf4+|
zKdB7LwYL$DJN3)9Ka=1biRaT}J`nxQL69rWC?bOOs@N9wp>(i>Zi-~&)NhiJJaW;e
z*(dMEFJ-{|qXQ(Kz_y4wyu2;ustDyEf8lV~4jd*RfeMD^m}~{7FXj$tAA#8dJA^|<
zaP-yP!-5d)K$r=Dm+a^!{1>7qR`ayWL1Tnsw9UOf#0wA?>W=&a-JaL6BxBsdlOpgl
zo)V?^K#}_C-`@Ja&40E+Mv9K~5k!O8c-iDKn_pkGDK2R^?&aI-P0QbXeMh1>CJZc*
z@o3rYA$7mlxaupBPSrylm|HYvYN=88?%2O>kN$@(ENs!*=#~7L*Ys>Wm3EpPJ%l~a
zupixgp!L~WVG|#b-D*<K$)_R4{YhJ)DF=Q7EiJw#{sO;(bMmWP)~pI}4t{)|nkxe9
z-H!E>Y+e%;j<a@CR9jH`K=m{6(<TV90$^j%VfYpJ4ir4yUw;#WV7H$C4;b=Ep#peO
zpom)(9k=jYtPNVW@v!E2*DLQeV<I(mudO?Dv{e2dyZOP>?2n%q$GeC3DKMD3LlTQ0
z3cSfm`M_FPufrTnWt-)n@d^KGD^6+Bd?s+An-KCa_4z#~_i(`@gBu)(t`xdfrwBrd
z6D4Jrh*3bmdkb3t`LZ`dh>WTf*UA#dW(}w20JDn+%SCRTt)fB}?y9TVn01Y{NP9|_
zUzsDY^n>5+qoVvmP{p#qLrK(O%feBsg6H6jM|wo2>efX-hnps$?&qm_r|vF!fzfpi
zS0G<35D>avX0z>MYi59=XItOf0Q0E76YaheB2^y<EiW$%B=FC=y;B~l{F!YNhYwP%
zioNI)Qw6SHP-X1C7RS<bu-cpa`x=R@ay*PGTd{2~n!kOXS<pHgKB4+k%L9?>0ZU7<
zN3ahfAfV2;2UUY^FXehK>6bxufj_LH@~?3}aV>a9;TuDGK@Z2dqlVqJ&Ja;mZrGVz
z1lc(0g;<iRj<LWs(8rO%mB2bUo-?cju?&vN7<i9J!?7NHGjd7V&^Q=T?h+XS#+Sw$
zM&}(dj}`CrUL-wK`c}@l(q1qtw|pV2(BC%PnzR{BB9rQ2O5^Z9bSxaOPi3)`b1`nR
zXM4Y@V=-G}x;~@_&w2<n0p~dEKiSxLUn<ACo6Q4c?OD&*$qo8l`{t}NzPIjH-V$te
z<hC0JZebn=v+D(5bV}32i-lJ~?8CAUpc4*bH{`sEi3qERkT49l^w+hV^*m+lDQv=d
z_1&Y(WIaU)){dVWP#8=}s3y?DLge6px>I#EXiY-Yq9q1MnfXve&y;B#4xtAq7|`wF
zy#V?UP!HU?-;^oOaQQSCl(c4zS%Hzdkh*FH=6waNv&5vG={*3u+VB3g{u1dg++h_)
z;}+_D`J}GUwXZD0OYL^btG+Kw<c3a~7TD4LF0vXo%OIyZ`#$lud<}sc5T?wOps1X4
zC(ReK=)6A%I_#w=D3I*^!K>>F3p`X;lV(e{<S$qXp^?Bfh<AjuR=TDhvNzVQ=s$`c
zeA?4$Foz5L6{wUE84~ZR+y^2*N&>!)jxn(PPzfLykh2f&Lg9@Av3~<`IO1w?eS{b*
zI&;qq4q>~1C&h16#2D2Io?<5bnJOkLxF7!IbQjm`F^#R~rcYw{SJr2hoJxBDTSYe$
z6DERB{U;^%Q{&y>{LTClE$)2-Af98wjjMivg|^+@v^}BGaE_#2iS)djQ)cb$LgHzK
z3CBTsewy1mJx@#T@VhtJiC=uJf{8X3gQ0<3C8L*pe*zPxS5^&;kPC5*yJk7)Ph{3d
z&5QF?4dIC<A?*ZLEBq7Az!ugZG0gSVahg;f0eM9jVIBrYHvoHw$TC9r0KgPMmxFHr
z>mhn!kldu)N!%blX6&n0&~9DQqYD9`U>J`Kx)XweIIIA)`QX(Bwk<FT#LZST#f>^B
zF#1-s1bS~5(rlpqt%jNS$IKeQJZaWS9z8eo)r&`av`fY9E|jon_VpjXF#bB;A~%)T
z`(!7>o<)BaYcd=ttJyl-hxl42dlU61AA8bN@jacI@?j8SaC?#^f-t68RWR(X=lx2T
zq$0~-r8Mx44T-3?`4+&%2=(9H?GXL^`ieAcaU^e0t|7g^UCJ3Q4cbI7?C}ATP0UP*
zH2G~@y1g6}96)q{KcEpo3t(AZl=NVU2#q*tc*MC9GU4xDfw;K=*wh9m&)WT_dyff1
zIqkfCf@|BF<z_nJA?!}?7f%!I5HpJMF{-J04FJKPz6a8w#(#Ib{gg!^IF;bgiNbUG
za6nmqkhT7ne8e~M&iJ^^`8(skaqWqAH(eOz_HV`XlB&Uo7TDK<+n33Cl*){>A%@({
zDzkTs4-z%+-1nsY-{K)5Nc~_3klWalk`fp}lhW!$adD<y_YYk1O&ZhHcETMFb^s9q
z`+$>hu^T`iz)8+g&B8~tNPc8KKhd@{4L=oqMS9o?H-X6Rs@=G*jwsGRyh3_{hvR4^
zwPQ#>1R|tMcQb8hS11N5rkpgeNzAW2y?wb^fsguxK~j)mWa@vPe8ZVVK&q&`nr4H=
zzVE%8n%>sX<L>O8K3VQ+Ix{gg>l1IE!M#)__<ri2?W^_%{cfO9ehjSrlHiSlg%ML5
zSoDrq1%}7E3$+p%uoNTP_vTlV?h*F<B}5OF0|F&K?S}MRE`$;fVcT189Jx?s5w0TC
zPFL8l3&9{J-iz&m2Un)o7a>5B?I`;a|3!|scXH*>n`a9`NpJwW0uPuuj+)qLAsx32
zp$s9ec*fj|V`sosfTaPS)HoawqP&YS!Ia7hSmxQL_3_HwJqc}6mEe~#*$UZ}BQO6q
zFvx~ugx^TfXP|pGr>U=S=vi0#>r!cF@3P5D38>DDGr#`3l`R~+Vyk(<0i#^F&Hm)|
zc-vq7YN#Lx1_s&PB9Bh0t|snt!$ZhbPIbSvcO|B=Z1w1B_=Qv%b79giPSC>wj=%4P
zekzJEY!0St3Cthqvw#Uvf{?z*qyR(#C?DY;TNmBQIYmRyJV7kN5e+#pSR`^32?qp1
zA;QVJzk;LaT>wcRW826<Q3_=aJUvA_%D`_xu!P0ebUJi(edE~1tSO(u1@y-C1<5U1
zubsb$-0pdpFz9JKWCkYXKlJ4{!)P53ycRsEJH*=2+gDV^tPW1Bn<h3&k$sAB6K)e+
zGb0#0HIi<jwZ<aWSIZ}qM%6tZ(Xc;?Z^^u(!aXXp3MbDzr;m-4izU;?5+E~Im}6X;
z?5dV5Z1pa>FfxPC&!two9u#NgRlp(!`-MwV#ex&%ZO<`n5|ax~s1mv&%vJ#?1ZolG
zy;}PkkOzpBLKdKAOXQ>fm`|P~|Ml(B&PuJA|JUC8bVLg=iO9|%W1z)E%Y)kf;^XaM
zSDlN6FfIceTzAB=GZh=JL`2NDZP}u@N5Hq?uQOZO4cA~4AJP7%21W|-5Gs3WgM-x(
zM4A6Slcyc#=v>nV`IMAypCY06ke&qYy;#g_rD^nPD%{;w;`_+^>}#gz!)d?9J0R~V
zqvKa2F@Okn8Irh1bMB>q#VhfpcgPPf|K+x>CO6smze$u_4B-Y@wPM%Q_j=vX!*x!O
z-FJ1@5^72`D1Zq|zJhE3Sfn0HQ1^p7iv@IurJz(ggV-EvH;6w{X$ScZG&+brKt6s>
z3BZD2HQ*=3gKCHLoj)0=5GX|?%aUgJ8Zz<^oVsC9cHmZB^vlnKLG4Xh#9%a!s_)VQ
zt3oU8o%X6aHgv!2uZ8*csgFJP4VIs5ZyY%T)iG(oLV8AE-m+Wih-M*3`ou*ZTjZQu
zG_TQ8#V6lqPuZOGx4Ovop(#I7bCTGq!t)|+anQ<KOe}fvc9rKXAIy~;WX*XI;uW9r
zZCV)bk{WX2OmbsiQIi|W%{F&H%a2IP9xjABF7M^@iyq@T6;OCEM<eNQ=2`lUp=Uf{
zgqu%bCg<2W&>NOWV^RQ7rNJwVU+*4#(S5ora0^24zNdQKU9pEGa^=5Vl~LR%a8vc0
z&$i<`iOj3;7k&8S=MyCufXIMYMNar$`>BPsN)zuf%LtRx(xR(e@+HPPHA<z{Y<_ZR
zKAn51)VyRvMwd9dq;Wxl??=zznd$1(Z;5}Iw~T9*?y}2E;NRgi4EW*1yqc~_Arb+5
zmt7G)pdkM2yo~2Y%sZJl+f?%63Q?l=?j?j%t1&H)NGEcvcq@`z=d^Z!gP_%4J|7BV
zWi4@|h~Y~D<hS=FD>-J87SPf3Mign7ONF@zgvA!KF3HhUz=1BT<@%a`aQpl;(0gg<
zba>Xa;4bR^vq-_dFZs}SjqJRaK3gU7IHo3=S4yCARTwR<-SKRCy&Y$Usa%&v+K-Md
zIBj{?lvj90Pg?yp8?|Z~UcQX!deKWQuH05+$|mw_oyV%6zpk(gPJ5Lc@U!y6(jZy>
z7OOpPYQsQ3EwJ_q`#k(Y*)AQnrj0alv?oI_yrWkGqmFNf0jrDV0_XAAB1&-4wH%cp
zib%mQ_B517eS^CQQO2iCr?>>CdG=jLgNK8t9Md#7C7gN=)<FbFsF@H>;5bGK@~dor
zw~)}?-Q%s<=;-ft>rH$d8shpQ$LOMhejb@MAEe5I`RDY~0{ZxR(k$|jSSSCteZ2~|
zlt8(t3@amLsjPExsnP;kF}0Pp)nyO$TllsOYi6f%h;c`n$6J;ib=g2k<(-k-x-#I<
zyV`$fRfvzi-0E#H1^ePtU}U!-q>}d!qo9DlpYT*%pDI6L_2s)C4n9hY5iNo+O(wjP
zpMLyaC`_u;wx+7bI(2&rhQW|n<(p=dL%cIwN%-=}mK$i11Va~&@(b)N6h?%{ATy9N
z1{R6Pr#yy)dM^5uS5nMAF6a%PV1DUSeJvDD<9_$Y|ItoON_2=pDH=Z_gM$cT#j1tq
zPp;m4ylr--Ls2tp%2KVavhqe6x2&^p0<{P8TGX<*Uh{p>YbgAQdF4n)xmu|4>=Jvr
zLJB5187d*$vY1b~0~-pdL<FsYsZV%-+xF?HpphWHA`o#m7W!0n)^F<KC4q?xiAF6`
z)|_tvPC%3dsK$*vuv{NumGG?Lg)ASb0HiB=RRSGG;~t}ra;qS*Qf@Jhnssj8LyDO@
z5_l;K%pEd0A^s%h?tFdl@;sU~K-*JKztp{_RWthA_S~PxfLI4p``BpJNYgPyN_}tZ
zg(q?;4XBL{gS#)&o46T&<XW}mUPk*>3e*Wuqd$H{NVv2=9&MZ_H##;FeX84Y@I(0F
z;(^~SjL7oaQ)J$`t&i4UU|e;Y6EX$$t=4(Ml6C&+<ebJ4$L#>=NW~>*`V(x0s2Bi!
z0+hjn<cP6|cwnUsMKumS8l^b%wkukUT_b8335sR4B&{mc#9sl4e?B9t)GmxUP+;m&
zZlcgcLx<l8i55l&Amej)U`2O<Z6KyLyakB~^JbAeIfGGrD9ONMgju}lu;%W|r=Jhj
zHD1z;)98mMqbd3J7jNdNf6SLel%(dvwJGquaO3PMCDjz30DqbH%VdXE9x$T6ni{Ob
z?7G<Dk`29PzI>D>#i@j|_BrEC3MG$+7^nU*<ys?iv+2D@_|p($O&3C?1i@`-`u_M!
z;wX|bxoMS)3{-1=&y(#~89`tT1VIRK;*BOi5|15~cNDhJ6c`-$){Q9}pm_L)m3AE`
z*c2(o(>moe)pf-F@fN7v1QN3UA)pW-8U)T0SBo)fWc|Pe;FO&94Tec*Z*2E(LxE_c
zq#>p{rYez(iWox_TupRZjYqF5Fq}ou%8L)T%>TT@x%?K!3kCf*n*I}mQ1GN2QZIWq
zII!OOYq{}4U(FPo!x~@fb}?>jo7yxz-nVUV?-`Gw+qRuOgTKU?EN>F}5SQtV^42`0
zw!j`C)Irz#D5*as8+=>_gWH*<qtuc}Jxh_(5TWD&GXOuTs$t1@d_Gi9m-TPNTwxOw
zWb&(?zW0MA6@w?xnj5|Z_8Y!vm-DWAXpkSF*U;6y`JzMwdRXuh>Qqdp0B))!M!-H$
za1)BiSI3EUiqXd+A=lzb>kDj2dN~*Rpso5ONqZ#@G^92CS!idM;IhiLMvrAOR&ps*
zZ8@(?&&niH=>>mJ9cmBu?_&|(PoJ{EI#T)i1McVDZhcGFr8fx){#pi=@6|Z(#Zuf(
zsxtE$Ta1R5$Zr$C435kt8it8pKfW)ys^aH=It;s}K5gfEr+M4E_#L_)y+z_6$4e0d
zvohnI_u;K_uDS)z<LVg0xGP{7*!4I^cu<AouTf@y|Gs%^YAlZuaSe!{HnnTNs&Y&J
zSu`5P8E)YT=U;K$4^1)%PUJiUNTsfKgIhw=M--mz?O-0DcCWhv$v6@k%>7kEDp#<e
zWh(^Ywj>);Js&*SZezn&(EsE4g3fPhlUrogRCyS)ZK<1EC+<sU%$?J_Z}I<@dmEBZ
z;nC{I5}6nAbQkr<gJSvdoQ_%6<O#ExljRf*YfVunv2q{&I16kRjpD=`DrfW7wBBZR
zG}>R?bvu(wp-tXjHbLUTZVCB6VP@&><cQfZ+weC%ER9*bB>uQaLKu;#Y6yN-s1=6*
zY6i{(fG5~BxL(NB0{$h$N<?V@V}vc_ZqqyYEEfg=uc2_WAeBN$SS4=2lA)|58ae%S
z9BCLTFvb-&GBTu?$8=4<*+vru&nAq7(Nlp|l9WYRg%RT`#hv`hO+R!Aw6q_qQ*k$L
zS4rWC7-=m`A0+zSf23tjK}L?U^#_Wy8S|<q-&-z|e@|;HS^nVm73Hr1dHyZOT4Ln_
z`|y92W;DK1)9^erJ=s%H1i0Q@+i_JpW-qdDULG9k)YJR4iUz}6gFruL+LB%%<$D9_
zu{`4T_Vz>-O3LCnSreHF29iC8E?uEY#1NPegPtuz3cd*b7L5$)_PbGW>ILW)kS+kz
zF+!l?ppvl30lv`M*WfdJ(YHMjqFC^w5Fzdz0PI?XCZzg(X2A2#9#LUGh1W?{uWt}_
z_&>&Gc3=J55u-v(d@XyhDtdZ$WMCeH`E^*o<N#BCe!C@)jbHny4JK6Or#~OdXAl!>
z@~)UyGqk5OUqrA^&9qfHV`R+Jx)qcq7?%TGzE}_L(iHonD{>VAq96R>07hY6!k0X2
z%hZvYX@>=>|GV1GSDB_7hUOeDdXksR6*UFcYNunzMOi{Uk8TynAkG#J6KRH-(O*kN
z?TrUe7^s&r8(5=b?ovgtX1FU4gR0FT|2e3^SrxNo-)MoeFRk)c8O_X?mpijtSqaLs
zL%+LXdgdG!n)k7e_K6aybocaxjx?M5r@ihTTeFS*ceI$$yg0O)o-h<0;_KvEEZ+N?
zP8a~WH%Pq`-5;dbo7zToo|y&snfl|tq9Ftef|fy#gOl5cX+@M*p_#;VLjF=~<7Rqk
z4n`DZT0QfWpSL=5J!Vjn;3o<T&ICfoayB9bZqSA6g&gh++dH(NS*wXNunLAC#0Tjp
zWNsKLv@nqqf$}FItAFEH9z_t6kPSjlj6I|tTOYI)jQ27}IDbxijmG}g^~IvX<m2gW
ztc&@LB0)BPN$$S4j7MtT`1|Wq)3b8_V-W8DRMJnxz$wl10h3~Sju#KsmG2EHRVleC
zFgH=itRcI+*yYReLTDs??H0^z6VI%Fwq{qM{{+AKGuPqIc?XVfP8%C4?kn)hwk?|D
zL^0y$^)%rQsSwq~3pJ^E%AIQZlP0+s1;WszvLj4nU`2ic{xFT;WpL7&cLL6yT|bNK
zB{A%&hM=-_>E-`oaM$FJLo2a2urY8HI5yy(KA5FZc)abuFB0ys5xHZ5r6r+Dhux46
zfsqqw4sE4LMpo?O8f=W-1&8j-fRC41w75l5+%6BVefnR^bPTOmKSMNVF&uq8S*WMl
z)Gh3FeowP!Fpt-V?3;X6ar=tbXe1%`$tXJK=-ckIbTc+ZSy8NZj(ts&t9cM6_lXAi
ze)n}UzM=fo?aitCNfp|F>R1mp3ko|B&BuDCb!4`SY=<Vm(1|f}GAC&S#J_};2=xV2
zafoNAJd|^>ETLdjjUNbwJ*K&m>k^IsLVX#Q0-*x&kc<s7A4kfj{4rMma$#f-lDQ>x
zf*px~M-4td#MKSpX62N=qA5VB3PDv~-rRRu`nx74o4y23H=}O%b^UZ@PrrGFOq=)T
zqkOCWuM%34bX(;wa7DU9xmXG8yfwWz^wCY(asv%k)YMhFuiBm4X?|T8EB`KgDn(hT
z#vz&AKf@Y;Fm4Mu%>~|IYV^kXB9vwrr(1t_6)_UP2Xni-DtE?97MU8|lZ_53auCGZ
z;BE%248nYg8)5jxRDK}p9O0loN7aB|ah}TC^hu_b1+?@j0R~V8CE63`UKtXl5n3oT
zbhKA>D9i|XHQF7l9tbTwh_Z;dpmo5LYk73<4BAw3&X{q`8RajoD<STu{(PqA*yI!U
zDt^%~&i*0A%cT4MNRtncdn&t|pdC)lVvIKQ=QSlzT>lGPj}Gy2s;M(kR$#n)+34=<
z0$J@+#?7iZvYHS=MKX_gmF%EenNZ73%TCVDE^i9`P<W?MX6$T1tG#GrM|I;DuBify
z1H{#j?Zg~3nKc-;Q<_hNmKMhjIma3X7cm$z6q4{>FPWkD#I<M~JpxEo&MmE|=s{~U
zmr2gFqRJ<^R-6xL#7K#yiRIZFn~B^L(O!y^Mt}qHwGXfnNy8fY9fF$SjC=**f4=A0
z(K)OVL@~&WF08#W<l*ZEwkmfIIXRXpq9-Cu5%7+Fa%~_~r7P3xeWNk#bHSF}Qu{ua
zm42&oqhkW*2<td_i0M*lRcBA;^8gn)c(wlJO$GI@oKnGR(QqEQU;du%8%irJzAwXf
z^G`0TL*l)vBg18)<HVc-(Y*5!xssP3>lA+5=lPY-;=s7*`_Hqb=j0)U5MOp6Qq8F;
zts<~|R61&Jnhd_bWz3>e5&BLaWbR;A4ulZoJt(AZxh})2l$8*}=z4o|(8oF~0hW1E
zK3if_2<0W_1s;5>>UP8Bnp7{jM(L`C5Of1?#Na?=YGf(2J~)KfL)bcKoPn31amED{
zV0`lrfD|P~?|_^)qEJcDZ&syBy(BS%BZsb5ge1p~%mz%Q<93A({|QDI&9p(|OXliu
zOep8f+5ca(0Gv>C)b*9G)m0y>&qD<^`HQ3yzlbE!%>&jLEla1m4jqI%DK}uh%(hzt
z-A!G3VI660oH5r^B~;5%i-xpC1Dx?q{l*^GOLpmLEhcO`REog=$Jd*HL%sj)<4Q>-
zr9>h{WJ#11LMUY4#bj$Sq%2Xg)1IZQgDfFgvP6xtN1G!v#VJV=N+t=Z5wid9&p6NX
zoac9apZ|5ubuQ;z9nE~^{eCU?>wew0CZJ+?hNyP7Ll3vTSY~~Kwkb&gD{Vg2_Ne;-
z98BPoV;0(2y)P`X<oRa7gU-i^QTd%ohm&EvcN^z1Gtf^l(?qUDxsSC?B_X1m1#AW;
zKjxchX==h_cC4CB3WquRywk8g@MYDK{a2^zW}4@k9q2a)kH0DW+;q&vJLPnUHcLmb
z&gkH7LPzocQDplc2qB=g>D1tj5=pj@;y>dPkEvO1No+~$ouyQ=YdW)P9<`LQ)2c7f
zKD3RtwisuGeOr3*UUiPk>)Tu*ryYI~%GmxU?eWT4UZx+VZ{h=e;{}F_QkD=#rd?6t
zG5>L953#ukC>h^3(JjMx5W!&REV&e{4tR8+;fLIWind$%R(39F0-LtKF!b{tgz4iJ
z!AJ;g&%a?M<}ilfcs!_P6xO54omaqU@X~8W==%mj8<<5H>jACtPs&HcK6JNi<PS%*
z|KKo$X_QVm+85J?8dD2+P3Pt8q1;vrXCSj<wGV){{X^|1FvUVaV~+HjBPY@R84f{Y
zc6W`{p|C>atuk%p)UmF$#v%D;B)aO2)uLH(VHuyTf|9p3C2LWdk8<}7whRi-F;hb&
zVJowF{c)|-TYlSFEP%W?v86cotf>+Ks{sPr=7(^Mxwj4jXt6`d608GF?L|j~9Y0=e
zglxhBi6FISeFwyX+rYISgxqLX|MaV`kBjYZ1Z`$s(hR6Fnn2)|P)EblVaco`H4^;K
zZ-dw%MlU$Ly>(!;cX21-74?WV)%c_C&nSJVM1~HT{SiOf=Fn6w2}f^2aFs4tUO`Oq
zUYufj$i1A%F?2s?PdcRCm^kNAE?L*Hg?zZwXd)-m`<$nWVk$@5&VhVl%~k*L7Dnfr
zQoSX4j&<tO+(S%P7JFF=DQc?$wgU(DAZEIza9+?TYsu~#SqV@oO!mfc>YDRV?ZEaw
zQQFza5XWO+6r0~G+I;YX(=%_7jrk9g<(?m9t(Je2pU0SBC*_|W)UA}!Ojc4BJP4T=
zgj?@+Rwck9vUWd|$630`{%rf-IBs}>{+j_=s56fSZ45`%_|5mW$8Kii)O}Tt99b!q
zuJqRSIM%!G_<w%g%0D;vsv-YbW2I;5-<eyVHD12v?F?pSxy3c>uc<wwZ@RJ7q21_w
zqI8JfevVnqCFNpE(T0%v(4!d)ywf^wElsZtwVDcz3X|1bmc1;LZK>--iKlr)-eZN4
zwbe5{AN=f8{@i*;w1e>OfG-1UnQ49#ghiLYctJ7(3f=hL<HNH!u@7_Qx_0JsAA1B6
zG<2XsV5|!&#Qa3hp+rJ3SrEc65LAfwBkCJ|?ET|UU;>Ja!`e5*kep!d!XW`_fXRl0
z{tNiKs2cr6LQp6uqo)ua#$RV1zPzB^F`8wbBRJm{{S|rd^t}z{vH%Ze@|ZWn*WC>>
zph|Zr*%9mTW5`9mwL^`09WI2<<NWTRXj;(Qvt8T0TOAnVH-?K1-FM$fTOl#*VTqwe
zQIKFYFFD=)!&*z@pY8|9z8()o-x~)-UXA!ZQ+ELGT$`5LbxlDnDU2C^I)$!0sZjk)
zeQt7?o~}b?O<iI!&ytMfgE|-h78I_k;Y19!RDd@b5cMsQ@t*BpRWOwsszPmxn}m%B
zC`cgbI8_rYGMuKB$RqO6M69|N{^`_ijiLe*Md9mNgwiLVCZedBAA<qCMq!8)#TR)w
zC>oxfcyPb~qfeqFZ(q%SoKO%!L|<N_t<g@E!Tw2)q1?0ML!TlQlS0)S&9bn0Te>-y
zsXh^vIxd;V{eQv=ToS`&eQDb1yHle(xkX$$;xo2RsimHiKP4=}7>o)oP7m0*cs7`+
z!jXBN5SNI{+vv<C&E7p$@9&?&Q>R`{kg-?*BU?9jP&m5-6h6RJy&}NlO!S7*=3~8S
zoLksUhHalOwuD|v4AZ-kfZ;IiL)ivQx3x8)W<vp|9C~bmrG?DBo$F2kjfJ(mPBvpv
z5UPO}h;?t=6Lk?IE%2Q20+#v9Z`%GCw)*Bbv;nmcb8?rhMi!9Sa~y!ifosbAcoq{{
z;s?AG*8i<=cTmc@E85YPBBz#?me7#==-*lUKR^lx0-vRo8-$h~VdNDJS}6IeWY&~U
zTTiSv|4ZPEO5$4cCJuO13=hS4$N1OqnkF_5mqbg?&gp8cP0BS%%BUINsL@Z-p=8vm
zzqA+}b3N<G+Nn5)1#uV`wGH51QT?}bg2xYjjR4V46YF)n;9aaa({mvof@7}hS*Ef0
zl*m2E#o(@{8sY8`R6U1W^hqGLVLuV$uHDkUdp029v-l)ro8()>Bsq(TK6d|K&>25w
zfjmDSuyNARmHrQXYoLlPy{jdBho34W0SdHmL`BY`8ZYnYRJ!schWRU`Y$|6FB{J&4
zUiA-Nq++b5{huZ~U_N@(ADx>k^3a$#H{r}q6E&EkBn2;%SmL+WWb4j{5N&a<8kICk
zHljT)zZtti3ZsLj-BO-;kIPRMSZ%*!s3or3{%-5D$9-Yl8GSSV7G+{7xCodwb*+8`
zRN6}j`fQI0hp?O1IOD=>gV50-0gtGHoCt?PR=_8d<{J+vjdiUhY;EUBfY{E({v8W+
zvJNR`$rCCk!rLf&V>df~AJ5GwzhQ}mH1rv$iRRUL<&!_55Wv%8jhHNFgC8DhT;6%k
z^ZHAcnVIsM65F296V|%(Gh0YECn>i^{u&7~9po_nm_GgCctdu@{~NY|S=Akko%)%^
zP9q!z9^O>EX7{j^`#_UoN3PI*GwON9;4zoaCCuRd5}CP~UmNEhK5+a)=K^iiLcJ}w
zX5g2-+OArYQ0t*N6VWQR;YUWwYII?S1ku4BD1CB{nND5u2vTDRShiu|wzph7D~*X@
z#ld64v56g&6|gAaw9ws*^MbwQz|?tu0#rWGbolQ`<dh}u<h+14RYt(0m~h~;=SZ}I
zh_j<OakQR=kNfl_j<M(tNQ@9ubST02ktGXJ%Owr{W1xS;ng)(7zz`<~89v>(?7=)8
z%r~}MSuN0T2(t%EZ$|`5C1|SI%T<?_Or|M4p0kOH5SV`cxi?tHkP*t<F#h*{-Uw!K
z6|nMDy<vw-?CqI@e+ngc)XGhCw&xbPaoo`6v%Y3sW!{x={U3dx2g#FNZn`1G>3ceI
z(^g1%X4jfb*oB2ryH48<E0;?yAD);RiQB?X^URW527(HVt}>HYO7s{Eq2WaV6!nF*
zp#TEPqC?N8czF)7=;Xu>C5vN<AvX3+FslMXeV&TD&19@~z^Bp<+c;RYWwCF+`eNnG
zVk2?Rh7Q=B;cM8zqVNN5!&ae}5U^L^rFh}F(cblfI|X<d?yoE^>hM%eMLM=K^@I#o
zH3$GY<v)xLTgh{CQ`_HO@S>bp8F1w@!~}W0*S0EwCiaj0Bw1;q{?>@yZ=;hnnAVv6
zGkf6O+%U1lu_GS(1HOqf(%t-Xd$Hlu&D&i(xP{TTp5mKROJ)i-FG;YtXM_T4%%OwQ
z@*4`Jmw7ef7Jdkv`dZWT54jH3ox$BKT_2fZS~MXHmY`}t9FR7w3_|f<n+JL^@Vs;H
zpl&05RPF%}7uX2U<$vNcVxW(5N0A%ai5<sNa^0+(K=N1<!8Wh_U&1ro%HnJ`;8(!)
zgk`^ly9ow2!S5jGK}eGT<27Yp*CGX2Y7;vpT<$4f|GK~2%j}_&&XX@I8m8KXdsRN6
zJqfCF@!aTMSK2w>q5PN9`~Ode;_zr>MlpL(dgc`eMf=LxyKE<GL@cJhsLu{mb4z&+
zzaMPs<d&%`+le$VV%!Kb8-rL;qwcowG2Xt7rhMkvx@>n60#90`L5M4tFc%wRvq{M%
zV*l<`{FBg|NrY*GmX;MoY$-b2K7X*9u^bFo5l|8fO*)W1Jpd_QY*_~FJ&@PZUU9eJ
zPxWzPf?6gc2oOF%Z12M1fpJMZb5TtN!QlY82M#6BHL1B_2vc&pI13aYXqv3C4St~N
zwU4MONBMOp<VvNYOurNN{1^K#oO8rWK>oyx)lUl@LXP~Iw2vti6UY#JJ3dvTT|i*5
z|8_EJIz7O+4v7L69#e==cbpSkz;H`E*y~C*CAD`Jq`mt1LwaVLmFw_JozOqWCVw@D
ze(1`VYqf~X^uE+a{Kfj>qSg8g2c2%?2WzbRuMP*gSBEqW2e_GsEGL$3^J_L#9Vha`
zh}KG682by*T5x}2-v&<`>XAw$zm6f=7Im-H_YkI9`Vsu$^l*LL0E=dr2&48tp9J3N
zOadXJ0g{YF&xLU@J+`3bYanUO8Wj_3WvEH9Q-+lxY^b6S!n!YZs#0H}9e^<f!WlZp
zpziMXQR_0$G1Ig(;Gu~IST387-vap-^VTbr)69|kA`FyV(Hnl{lm;ddi)H_ZMD_nP
zH(-#s%#{`Dk-oYlcuuqrMYjt<Yi*cx$(k18Qg(B;i6aIcU*oRkO+R|2XG0CO_%#;!
zT8TgUlj)suFi)HGXLmP!wrlXq?e@o>sn!51SHDr{tPrnMVoO{hIZ)MPN-PQ4r?CWJ
zAg<o$kHZFBA0zOx$?zC-HiV}N3&u_;C%NxV#l=EOWO+>B^IfF%7_WUyH@ZNa%p^&I
znE(Y0M&I+PB5(@k_o}dO24W|udIF}vxxrO{bO?)o1RLVj8TjkXc2Kv`So=2qA`dH0
zRpDDYSorJ4KJF;+nQbsJ-}}VM^2>IR-{*ZsM^``n4}H(VXN`}YHT*Z56b?4M<#=DB
zr#wL<!{4UX`63Md4;05`lEW4z_LEWKqWZC-Gl$x2AM(Eo7Qf@5LDFrpFPYHtybbNW
zxejjZYPzgu9U}rSyT!Hc>JiC~8N?=N{z-6XcXZ={q`b=IlJl*`Iw2M-f%F80Er0B{
z5HyXjuG=hnFi>nfG=g?PaE<T+p$f%2N&{3?SPZ~)RM0Uxi}SICEo2f+>Q}4EV#3Dn
zT&>NbSKWq25Jvcr+o%a#v!U-(43ia;6epq*D_g8RjzJN~uANXm`j4^)EX>v_FhN!+
zE$t!t8Mm*lj<mPwJ{kg&b$SBC@OT2l<9aqrw;%Tgqh4;QD{cH~Z$xvs4^HQ~yj3MB
z3r%H9OeUl**|&ztZylG~*#1@BYpXHZmx2@R7Q{H5mQcZFu!vH6w}EBZiQ1sS3p?Lj
z@4~j+M#qQh0KJ`aY31uP5UyZj!|Zt6l66G1KHl0Im?!(oAN@40&P`+TR!sZiHJ#7`
z;+6?<GxjS;Ru@M?WXZ3+`W*2!to9NHtR`Su+%~}M`NqQ5dT+}0`>EIM%C>tl)_b&J
zUN`9f7bnqQE-9g(t0ypZIbJ()Oy=~1KIV_X>XZ|1KRylJd}I>OY#($I4d|||_mh*j
zlP0PicW`FOQ28)DaQFXG19#X{8ULvM#_@wl>0Uw^rwB%h)&$RR@Ee7MC&Vrq(JRLu
zx#6bUu3!X~4!9ZX(0B=GmAMPH^q<*W0f?56j03R-aED;@`y!+wtt1V*_iMFz@P7{9
zeeSgxUN<+5|7nHs14|np{{{UaMrYX0II|}~^D6>SgIDcI1Lr^lEqiINThY0$5NpGx
z@t37N7rwDZg|ze^+V89NfLL1}_#bTqmUN7^OzrskzM0|LirB(VLO-$m$(ZmqD|sW@
zpW2^_FI1o2RX>pIb5!-Aq(q0HB%O4La)1cU4~>);W`4$Q{UCQIixN$=Q!e6*&ne7I
zE;~4r$t*wTJsSF!syERnF@<do_V7>FIJZQyAM4q4gHy_7|6byGS#2_&4Tr{nQ}-MT
zD1mh~xL!xrv6>eMgJToH_8N@U(o}gRR>MmgJuRUx32@t^oP39PoClo`6UbzNE@2cz
zIBUYy<ci3-ym51&al2F97f~=v0!5l&w!t>SUl>rfZviM!AwR@bIdIXEaH6nitNw1l
z$N$&qF&mU|s#jtz&J)|(_$?hKPeN(%=-`)b&71SE%eUdLwl1eW{!2&y?*u|;Ot};j
z#ttbRz0XdPV6j!}hg{qcZBMRir+zB+;uB$%Jm=<%9uoU=w$yv~#G}H3lmmyuzPZ`y
zn@;pBq!~)Fvw@QEORjF>Dg-D=Su<M%O|f?XuBmNSsTu1t9=ER0UX*Z~kl)2>>lKQX
z)#YMBauHY>Fv)S<(E$71l3*Z!qz6zDl!SS?31V-6g)vT48t1XI1PPD(5Y2W$1Z%Ja
zG6Pl^o^IEI0XRfySc|bx)!q$a7#<wp1Uu9X?SEm8*Q4aT3h$(At=CIUBB@n+r4hOd
zIB(1fP3{n=>`XwAEpH-9c%-WUXbY?F3;xCx<;e2ahyw0{ebJ3&oo#IeEG=RaNLZgx
z@myDVqU0|!CIDsI_R&C!)V7(Z%<3+6GvjEI%uaK)2=Q1l9N=W}N%YRkb09~F3s`cn
z<pgLl7D>HXzEpB~&*IxBx+80#OiRpWSBp3i&hJ%RpYR-zZPDEedaol)Sh5UaS3Gxf
zq&##eVc2g$HSC+Z3gduFx0I6K<4ECP0m6%<{;B4ab}><k_*DA-O2y@^CK^ZqW+VW`
zDY>Rrqt(qw0_-$EnCR?Sn?Q$Eg(2(?rwe2SdbmcZUq*tMNgv-@Z9)=w7^i&rvfqlA
ze18#dheb>F_sB(7TZbnWAe$Y;+8zCx|8I%HT9yrBlA=fc=U&BPTuBe{G6|QQ<c4A{
z_cAqq4BP}Q_g2Hh^cway+{;l@4sv!|TSDF^NLgx+Xw^Cs^)z!ysZ8nCZ<c(gjav*N
zB$oHL#V{vAH%9#aUyL%A#!*25vg{K`Pp}{98BpDXZc8iag!RRY<Kt_1wV&WqxgW@K
zCq0mg>?kcew5`3E7*Qgw`_m8JcLO0Pv|Ppj_4g;3tACx6!@1gX8C-{FUeJXMHihtS
z7AGVE|4ZIvUO@{@3w{71ffWdr8!l9&j&zi{nRRU`1I73^44j}6-0Ma3nZF3_>`H+a
zg}UnFVca7lLH8!YeODQW1irZQP3>pTSMQJWL#D=4Oe~(R?AT2}pie<EYZ61w>&zDV
zGG*S~mqIBoec-|<VlBFqZ!KK80^t44yS!3ymglpwO`DWNQ=LD=%l%q%a+$`F1o?A$
zyBVRgquohr*J(WE?_N4Q@+9%ilj&I2Ltlp!2_+zVI>%W6<ZLhadi&2*y{5?QgItgG
zm8AN<(olZi;#IHodYao;J$TQtpV(Wl6!V&HAn18Zkn(T_`w*HCmxz5J$U38l1m-Cm
zO<t!=>>j}*V@31GIxU=A+A0w;GM~`?`SypUISHNBo!My@o4bT=KO7laV|&#yBI-f<
z98cPss{zYbyiZs9$Bin(3bTQ(k8g4w!b{ZdV?_Pz9I5jhQ{YPcdCxYwn~O_nbRV@;
zu)bD(_Dfy5y$0(Ek$ry~yrOkvMBn%KYH1&G6U{2(?Jv);ba_`WeOFC@f4#{pG;!qF
zu&F2F@vw4hrp!M1s@dNMBMF+FXqJ)a1Bk=hwIMn&wT^*{3mEfBubUr9K{j&SPV|!u
z!Ch?ZsEzoWsuY{=x}NZlWa%kuow<XOi7<7?se2B*53@G#js~ij!A)T3VQ34K6F;C#
z*cjZM#Q*qY3RrAt7o){d9*|i)6LLHig0Y!0QA@)z<wM)lyRfF+zfuB@z%wf2-#bcH
zum49!G0ir6cm?5LE&v&Q7DIap<-g<6?~G1Q%Y#Z2D>}N?TEZvcfM2bevMn?4Y|{?7
zKOY#@zv7wqq!c9ROz+lNwN<ZGM4vv6aI=zClt~Sl@r+xTw9zC))Ip(ZDzbjw)=$Vw
z(8R0In<w3sg)E}9W@)4F42g752bH)HWw|yXbOVjUFsS{WOzk_2Z->q*euziq$0ePb
zE9E&Zy^r3raOdk{u@5)cermyBn#X;W??tvvwgUB2xN=BtZXJoAdjWNVUiQ8WGl}`_
zH0RP}yQVVtiBm!@#{v<P)*+z1<_0D49JS(hoRVDpqr=G}jG$?^amS?%U+bdjlf~gf
z&ii$m<woDQcptQRj9*%l<_tBhf95T3Pca3pot6B{ce<N&ZSgK`COem>Qp%q-x&)ej
z9^A(wB@%U|`&G*A4I48IE_g11m_bwe@ofi9jcrJma$-YH@`aeZ;nhv}ZC@@m_b<>+
z3N(LcsX(hxem)ZY-VL36(`HC80-nVV4b-sRZEM?B4_9jqc#j3!+<qOye?45qF6?Rr
z`<3Gd`L@yV8<yA-#70iPWujLJZcU7siMAHY#u(Jti6GgH2;GoYnkV|sb0<)dBaH%4
zVXbAgo><&Ga@J$zAHGYVU$Jme4ZyazhF8UTY*2o}-~{-L@8sqgg<-n7V*o^U`b2fK
zhtKmXCVlzG)ONG;L7R~?_c5$X_*iaz{C^h-+D{uFm-F9-#Ln)_wBSLR2?DQ6fP-ob
zfTkt!<-W|ZCTA$)OApxEObqN_5D~F8@?9^liAd-8o3t9b#r7$+1C)%Q|K%bGnYAm2
zsO9VVt%o9Y_NE)+E{X<JvlKiLq*!r_kg|jdCzK|LhE4|;GlTyKV=(XmIAxnSolH@y
zvR0!R?(gcC8vSB>PeoPwTR1$1%DEOjFbJL(i485>PP}j#K~+1vJ?3s>VXHo{4g)IZ
z)<_MetSVs<^ZR<m4JA;dwm~-CzDuuB!RT96Q+Ic?0=Hez*`cSK@Akn&lNtQt+w)sp
z`#M7DEBPJv5o#)gy|cAkT3YkBR&EK;D2IpE@0VSUFV)gePE8H&zi&Snn7DxaDF}3&
z4e6hzHq_i)*6m?QtZ_;kDF|+L42=tZB^?^PLq*0g)p_>!uJ~zL+xU!UyUCweeuLf6
zX+*gPzc$;*0cdFy5%I%}aP>w#RLr79pm;ne`u>X&H`*>t_W?q=SHjAo-OXcBqIMWb
zVY_Xj_(n)Fu}o61nA2Uqt+UiGD<1h^22)@bGyez{I>BE<b2}bPup_n33M~N~54|JI
z$gu#-+Gk|(v<b{&leAF_Ke26~^+0bCWyLc{YBGDsue01za;G1q*OhM^cBytep8fFU
z@ZSat7@iU@iCwj#neyrXv<wb=M!}zAFlH+nJPQ|hkKKaiQ45~alKaSotIR`O^E@S<
zmG+;y|D|xu^KzIjt!%^A{&MRz+OvI(jo$t5Hd4R5&duXH$=Eh5C)*V2@tc;uFDI<y
zTx8*&`h>}L4hEQNbOJ=@d1O>pucWscla@Idk4M5+L|6jFaqgD^3g8>(NU*(&aaNR@
zE#G-!wVBl%tLD3=<kqY+swqM1m_-LkjF3MTW2zkxYQRJx5JO;WWg&oJUYr_R-=}qv
zT5;|K)-~jlU$B)z8nWRc^@$V59aQ{ddFas%h)&RVYG@1j3S|UtAaj#>)8mZFhbWd*
z5><T&qr$|E#IU$e&{+1w%cL@IS8#S3WIaFj%&z@fCkiOo>VcZTpTyIzoXMrh?flDw
zo7FP8(tP3N{Bd@R)}Or(AP&!pUb0V%w>W(o<yV>A^fBXjVn~TQt!{uNl+XLRIFHy+
zj4$VE7UndFyELSPZ&J&XvO;2ckD3*1comGgUPB5GyRT4=CxA9tmWF1En{m@#ty7hl
zGvEcRW1c&YM#Puv_1|JoX{xWGDR)Ns`DI4!FnxIj(Yic~AZV75!^S>&<Ii6#DpQ;Y
z9BQx+j7!eSoY2Q&j>x$KehM4g9HL&rLg@tS1|qvZZ=)a8RhCe1UQ$6JhN0D@pY{KZ
z`aS}+;75my!oT_Eytkizb~;l1QV~T2RM9@?tNGu4Io>_@%dvUv+_TCnP`=7!79V8T
zub>M`{VG}cu4~n6ZIuq|8_V<;h)HEy4Q;UI-kY+5#YllrT$-*qA%PsymM!yC1_3mi
zAW*zsP}aRa{2cZ8o@-y$0fyt+k12y-0ubUO>j0VnF>oe(Ew_1{`H`^YC8fQ6S+)(!
z=PaKMz@CIqm>yxNRRHl0tXVgY^e#4L(F+dM5=-V+DDEA^svuq{Al!Ce1QZxfvQrY5
zO(%+#pfHsUt?(E-a-UN&vb(eX@?X=zGky+2J#}&iDzE#_LhT3+Z6h5HgqPUnU<tU4
zb?&YHIM(}}FjRQ-;dAe+*M(yz0v3FB5_vK*@&5Qg^L=Xa^#l>dbY4QnX>RD~!Gfxe
zr{hU)U+n^`DMcKO9RCb=+U>{njyDA^(gI%ye_I_28HU}`kxM*zh>^<~A)bR>YY#<b
zal(3_MFFZ{-GTjtA=5uzFuHt4j`5f`mC2D3FJlM;iZNaX3&k(K!Za1!rgW~mttY6Y
zWC;yq!KP@=6`mXGc=%Hrytr5vga2q_VZ09A$tqdc^v(M%LYDzlu;xDEh3yxJB)xLx
ze0J(sH8QaK&nzF}X~8LXKdQIj@#-4Cu<q7=CH3};AO9-Q*AQTffzLztye6&vzJDpp
z5egNpSbF1MYs#^EC$-}`jd{yW)*=}*<1T{fwMxaOx^~Y?)w2Ecc2BLfbKB5vEj4Fz
zoy;BI#Q$Y}1MK<IJ8q?r>7`mJ2lhnF@6|`euaJvOgi?i5%Y9ZKoRaJQacEy!4R%N6
z&&fP(0IEB5Ht@DPHe6p^A^&=)b6cdXRK_5E5t+aYzNv;CJoeffZ?5O*7O%X;ztH&-
z<8;)Q=;a_FHBg1p0MBB6rwYVXzU2tz6~(q=Eg<Cu@WTv&U&Lkj0_72V+@Bdot5#~B
zt-pHn)PZQT{8Qe}oll!S+Mj}AHT;2txCyGD{Wsr=2mdU>VT^wE2w{32CN%Wt2Fj>}
zNV2Z-l9F9g!;@Rk(n+xWiK5{CjyatCqy9g*fL>Pkfd`{(sn&nspXh1OL@WXFn9V+)
z+QEiC3)BmaI4wFTl=4L~#-}1aQ8_n8p7MU=kThh-(h#*pYohlWup9z1tR-wywM+>Y
zRvKKYSh9LQgB@$tF;2$yVNEizNj6W(ha@E$HkPWz*19e409Rt|%EunKZCnOC*I!Q^
zS@)NYE@|39HpCb4W;@r#1&p+5w~fN)#YdI=PhU!F?p(Y6o|VI&4p&}bdn<?Old^;_
zBVl|$uYL0m-_7o@66w8%aTllL*6s^%8~u|WY&wxf<oX*kTbw#5C^IuoT82cEcB0L^
z;WrfGD7PWQ8O+;TZ?qyafI~cnd_2~xk12PJAt8s|@#|5uCJ0`p99QcvJN1a9B$i&M
zWx0|T#dQpoxt&=3$_SHP^lTvxm(uF>%z3Wy4P7U0psLrB!lYW+r_V7bzY+`Ov<zEl
zf57S^w&BYT2xU3x5_E0Lrmh_ZHF=+QY{1=UdGBG?bS1uic)0!aj_J-RrMZ%Zx$uoE
zIi->++oC(pp5H0=k9}Ee?V1Pw(9(n6pLKLB3;<_eN1de`hZXaJM@JJu#|`eNR!fbv
z+P(8$&zG}a!|hk&FPC!tw|U8;AZS>Cq*DB#Yt*QH_Umzp;MHg`Ny88ICjRYeCd6~W
zjL@(mhK`tjw4Q*YP9q=rbr(TI(aIBFsNJKMeEbmKliO~Gl>-+(z$$p_3HL>@+SOo7
zYGb7Hgn<x)BLkokGgma^LoV94OO;${@p<?Ko}gPJgz^Jr89ssG5IChzzQDx-%u$e@
zruMQ3lL)!kZ<%K#Aakh-dtOQ@b#H869`$F^2oHOYXm;A8OdC4|O!R7O<0>?6a4Y^y
z`qMqzJU8=dz^YD7*#(y4pi0vJ&0(K3xwiA9NpVjhV{y5}auAYz6UHEYQmQwhi{r9_
zusK|k9KB$7YC`><^#ngDt6cZIW&T&ehsRiTbLh-Ch9KTr-?TsOwfi_*%~~CXHV_1$
z{)<e?v$VYo6%QD&Y%etKD+x&NYhc+8qfH~Wi(|aeeVljxwbxWZ8bu*{Q;%2y=eCo?
zVm|rr2pm5LzALv>|5~WJI0-s3*u{kl&;>&5cwP$ztU&N4L4ad9dIJh!$;$qhq~AG~
zQOuEb9&Yizgu;+CEO?l{C(M>1rF&vK-gk$&)x*L5gCFL`LT86<be?q4VSP4v55puD
zJ1%2COkUm5^uK^h?#7`Fm(cN}VLj!&`$pQU#}PtBewNa#Qb*uDCjQYqsqr(;i{8(?
z3jg_A=$Ptd%ES*M{xIKWPtQ(L1kCrQWl0LIpk>*6;ytrWA3@++Ls*__xyW{up@+rX
z3l@ef2P|Sm5+gyad`=OJyyjuZcBUoEy#&xL<L+^7Khgu4RJb;`-75I5bBe$>YxVxC
zct+VgqFPbBoYlYrcw2WxK3g@}5KGDElidv9tI2sGk1Z-o5-|_tp)sef1%x$45R^0N
zR4BBOx06(%(N}ldW%!Vm=lI+ohQxH8M8^1`G3X{OOTVD&zaq7JulYaGNKI1%ZVGp&
z&v}~>vdR4ANAJ&t9#qVnHb#$TA?B)*mbR_GQ}%msu_h@<HLy>}-1vUA5op~0ZJRLw
z50<>sRi}>cM>~R-Rvq~Ij{-|v3A0&n0<a-h6sLCls}i?RgH&J&@w9&HWZg1bl;~rh
zbR-XR5ijb4mi`yZTj7OPBkqod%-sa1=(rQ#6Y^`o03KL+c}*(=u2SWcm9G?dq#R#E
zNg&}JcNIk>t3Lr#9%H52VF*m)q5bb+BpAvpk}_DX%uy>hlaHkH{D}nBaHW@7NyQ&U
ze1J#$V{=`ngC{f~Xz-_QZnW-Em0fb+c!hhNka5%LZ~qI#g^cdm_p1ECEr+y!JUh)T
zIls~lf#LOo?za;jl&oe$@#C}NUUiY;^YWU(;!N*ln`XvCr`t;R-tlS@As^nEDI?te
z_ph_JnZ;WeLn$YRvYTNdBQ3Q!Gh3s;a(KnZ9sd?u=JJw<Y&g~t1uS{aui=r^YoT0M
zB_85her+D?m$0<9nA!Qq4frGs>!&n%9QZQ$pTJi@*7gRxVUDb$+LKgS>LY~D<_+z`
zq!vFvg0f_C>0-S-*L?FaYHVeNrk7=tjw6PrijxfF8(Z<trE}_mbIi>KiG99D9Z&Bp
z&r7XSOCRrgS;TRN^xLgsowvneo{wUi_B}pcPcmt%i%AiGt@h!qvE9pky*lSMX{-yU
z@b&O+U;THhk%s^2J&~()&z4s0XkFR)vgl<h&*$OtN>`q7-=2!g2R{D1T3J~cs(!9B
zY~b_RCa;~ReM?X89aHoD@BK+!C}P$0x(;TB_*wr@i2lCZ_gb7HmxTB(@6%jLlm8*w
z_P6|4R<pj=I~QTOpUoj74I6{I?&nBIY_#wAs&8~JC#`1dSZQaJwQ*O?;2;hEYUeFn
z;$b4GMp~umZ)G}c(KR5gtZXer&90HwCUuMGT~rnD*=E2qUb1tc>EoFlf9<r|$0@Rr
zvDZN0MPL5MuQiVXuZDh{)O+{p#k@?Sc-9L_w`daE_`c=hWc<-fF<-#$MA@`Udzu9o
zXq0!@r<~o*c|qJCjDmzLJJ}?w6l!|0OgOzNO@`4gMHk&lH(7S@(yD#TD<9vgVd@1c
zBlw^<ja$2yU0#Sib$vtZw15mkNZ<^IjH?9A&5TRykrFYE$Cw%Hmqw?o+Pn+{J?9JZ
zB9~Tq+koKJQI_N^_ogIY+tj$JR>Pr5ed^W5FQLL8)h5P|CB-NCKdJo0LvNHTIUSjp
zlc;*rBe}ViIx=a#)?P{?&zyF?v9bSMJhz0T1pByhc%iCIXQ0uZdyO$&>%w^%Gqsmq
zDLLtp{G|1irDW8T?#K+>sNkm@a|(IMvyW5x%uzP4nlZOkm3@U{iWEv49tZET#p0Wr
zop-h#+0qh)H6PdgOYuJUMNyQZ2nZM-86WmRX<U>{-(^ubgN2qk=4R$pRXMKx3B`j-
zP3BLjV!EWn9KP5@a)hnkO$DxtJ@}w8hOehI(kJ(#L4-sh;H`Pq8dxtVVZ>#|r-H>L
zvE)uHSAP8XX==ExVWs-?Igj$f-mhr5CxVI>_0=D}<*d00Z$&rc@BK&gbKaOoP}4JC
z4s^ZxW8FQoc($Z_c7j&7AvJCK7gb%}Bk9DX=a~2B9oZ7Yp4A&X1H5N@dpjxVp344}
zziWcmq~A&7I;MQb`lGel9p7b#FEku^#8g$YBJn4jesarzHpyWxE?E*siFc(jYn)QN
z-!%y1LkO+fJbTK8-XG*1<UcMuu%&kykA@cK>lO<6!w!LUoAKT_MH;T^+Qb#h^aUr9
z^LZOW>Pq5b#E(>VddR(cERs*Js*q_k^Y)P1Nerkc#WHWG=|x3zj!xSIY8abj@mDC2
zWcc*u6NmLCJR-i>gg?V52QNG@V)Oe~1Iv~IPXq@E|6~<DcRe`QL&WdQ9>2L+H4wEh
zD#EE;r5&5{MBjWZV@fXSSi%QG;<P`5@k@bVNPHVD>WgZXUQ$h3oj}<y;wDB5H+ye&
zj!zFV5mG<qo0B-38XO?c=zVn}3vk|t(yji5tGV}?HkG7trR8R(-I-E0^S6qi%2Z{_
zXx42sr(CzomfA?oE;?YYo{(A8U}y64aJ&gyh2MKKZJA-QT&XqtIQ@Cza=E^osa2@0
zn%aUV$1U(jnlDm%gGt8M4l4HqC6ah{m)s&`XZV&aW*g6sxpIWigpLAPHe0Lk=zUH)
z^(|#AE>1>|CjU)3j{1U5ePKulF#B}o@JhwT<#5}Z%knP!zQc8E4so|#u~-EELqLFN
znU3B!?bh}ZDje42i4J@u+*KDEP1wT0;&6UH9{)tpBIpj-Gn(%&@n0Z-j@S59bnh3X
zsMcf5M~0foc$bYU<;z5O^614e3sTQC227WBPw&W$Jl=o*n1P8y-p2%&u>7TE?ZkZ@
zImPIGKnp5PJQg3XD5p~G-4e+EBSf3lSg5S6?5{o3^EhHj;Z#W@FL^>u;BDbV$(1eA
z8~0xIc=;`kT6^(ez5T#bZ+UIfS@T7Yr%m}<D7c6<Ve5WfmAlq5@xd=}oO>>FOnSh%
zccJ~a7)x@n-j+5m(`&?Y?DdOr(t#08(@Orx`nV!HNnVU0mR-IOR}@h%>eiui-{umm
z0fRTD_vH&HYuO}flKkx!p;t{@oiNIDnzR*ooS^y9sxMz&yI9s1U!F)R)kfy6;X<@&
z5V{Y4284PhBvfh}NsQ4w^TW5Ix_UpB{Z1)a%ELl>A=_xMO0ps4({x-2-&j5IekR}y
zA}%@LN|S#U<KAJ=_Mn$BJ<?rkpJFshVK`}AlO0lx^=3w9jK*0-YZ`a?-^`CFro>xF
zF&_8!o*B?vLTfY#Te6jVpkNVME!#RVduqM-BF3M>jG5_ADtoyPnup9y@Oup;)#>Wf
z^vovmUz`lS|FG|nO+@t;T*<x0T!(qsFR#(Jm}UMD)Taq%NwkR9_wKuO4iUwP?;Jas
zb#MI-yorm?f9@?@!J<mFR^D<=K?R4}C6SEH88DU6W+tQ~#NF<_vx)O<w^PhP4mZZc
zq?>TSQ_K4{c83n?(Y$A+pHinCKYfxDxs<k>8Jsun_02nG_NF4J0UqVQzC^1RI5|~n
z=0ViAZ6M#N7uOi{?~S3;Jj2~}ppj#J&{p3pDoeN1X>)f8IGPbzeJQNFD`YOo`}SHw
zRkqAd?#30Bh}B=ZyJk1cR@Y5runA?rRjz$U&k|S3T~BrgYAXBWc&_Ykoa%6HNKH>u
zSN2!T>9aL<kX7^S_M^mGi*mckXIm%5k7}<RocS5F-}m@wMxAg`_VoEYb4t8dn@MtS
zW5vnvnx<FWqDWDaFRTx|%iA^exv<)oZB~;=rvVYd_f&22B;J_m6OWiv*|=BLM7=v+
zfy`#mcPZDk;Jj6ff)(=@U@a$kQd}hY)IoEK^@?{9X);R*c{yC#Gi3iLS!s2}<zcEE
zhxKd{M{Kz{*6e%X3d4!jG1}aT!&mY3WPd@@1Zxz#gVR5st&|q<S6Gk720*FZje`Ic
zKvY7Dj15&p$e9qw94R@z<!oT$FFBW>xL9urA!SuuC{FyUMJIGbW^(N-&wiEB7#Ja9
z9$S=&X;<z?B?tS*_p>E7UQA0nh{U>2*+1!_i#7A0xX`7{G)H9tB_qv=#Xb@j1L`jt
z<w}+;`*xFE=S}ddn|Fe~#mV0bkhm*P%g{GZcWq7>9lc3u7~hFNtG3yd_BA}{WKC;s
zpy+_e(<@;TiDao{CnceR^QzCbzts>VQYxOB(}_S$=T;zD28}fDAIDi@N5nm)Xm97M
zw0VBa-t^rX6o&7#$@UKK;uEzml(`2=>5<-~$ImShb~<T)&^Y2KeiOG~<?BPWw;h-h
z8ya??q6RsFRUZm~-i09OByScSex^ubbL1If_0pq!$2zXJ<Ay?n&N#uG2bJ==&IwYh
zC-r;oCS2D`LMx%T&h;!O$5I;7P4CO@dgO`^?1=iB9Jb>s%8ZR}vd1QZm^Qt=No}C7
zt?KQ4_1)NKrOQE&fZ?Z|CA!M~yYeI@N=S7->b`fKG`D<5W(XYq_T%K6!0UCB?ailO
zYkL13$DC@e3$v}8zATIy&U=X(Ii7C(riF5v9E*UXEN0x-m8hxkEcDy#xo)`(awp|u
zW~4yemA<Otsut+Cic1Y#JSSwyO1a7#;&ACG3VQaud{QL-`1t7lH}^SdqGA$@$+0P;
zTF${6q^vz2#MdJTZ5e3k-z&b=!fp%F;Q^1d#KnGpxDO@h6PF$qzXPL`Gw9_2J+^*9
z+J#g%`pNC<2)7f0<IA$dK&C(ki5DIkTW0MsvGZ7}M_UeCx!KitdV%M|r>|<?GBy1o
zEh?ErXXJ=6d^{hBKJMn~s7U4>Kh$@N*0}GQ(q-i{*Ij1IH=0{pUw4_SUNg2$Vz{Qs
zQrnu=m{0LO?Co7On$6DW<=<ReApCc=zUl0}bxTR4a9S06YN}<;#YD=2PcwavGKcU4
z&ok}}WE<5=Yb(3P(na6wj3+WhQT+`60j@7vwR^hCy3NeH*!qQb8m)nk4d=p#h-3s_
z*moMq_02cplI(lYpSAJ|UAYJIvom`ZpHkON<qgx;Pt<bqH#D()Hz(Vd6Yvz4LI$oc
zMP*c?Zi^j{y=~NDr*Fl#6iFbvNakB}!g(j-MgvZftLMXLsPRzO1*fOxhATxSnO_5!
zfKdJN^d#6(uoPOywS67eyx5N=(~DP|NWiSJ`MW;~LwIK3-wYJ{*y!Dt-^kL4BmPFG
z_G+(6^exeL3mHD_tF5=-AyryR=L?)Wca<dag8rv!WXNHl!=qXC(OB`g-yv_N2fgb7
zJOxLN6>j)3>*sGY5gYGVc92yC)~;0cFAR7msqES9Mrv$SDG4;%U8M3ZF<R-~Uv3j~
zCbY)LnrumS<rs?kCP8b(M(c1|A)kNOKHD4F2*ysEjr)Hjmiu;pstXwzFZIy*5p!NO
z7CIB>ho_KKu{O9jgZ;627E#~hJnTW?aay~w(Oo|I&H|F&p)whRrS$-6yoAJs=*Q%r
zT}{uDk$Nfrvru9?eb{@pAWzZ2U({e=i08hML_}7?QhKA7Gq)^4<5@T9?>J|u`&b2F
zZd^iMOTPZjs4s--H&DW*jiOQjGzjSyphc>}D4y^Y0~y8M0@y4G6*5^&go4olQ3WC_
zE}p!$gK00VNdLCHeWiSQmSYKRMS1;Ep2yGpSC(vcY-$@C9y4#gl_xPU_|@J>+249l
zMQ!3e?mE4x_@g1Ccb9c;TFv@g3D0Xk9&d@GTz5*LUiuv|S0yI7+2zMcSuw_VMkr5t
z$V4Qk#)5Id&gWk$P6i+7SCLA-lU98bonh~(Fq0SNv)_N%zcYHVL4Mtm!>-FbNI~7!
z>}Q0+=4<JJO+D+*iMzIM=SnywlSLKVG_aH|DoO7q|MX4Bmgy_59WHbt+n-BpK<@>c
zA=ecB2#17wT!(YAYfik*z*+2Nd{f*%kWp^30||I|9H$TG&N7q`G{KAIWUudpwOXS3
z!}%MZ0;e3+%(H0q*)SxT{xbz0MLeHL9=%A(;=74T=<1~->%Mas&IDTYIwcS3IBRD=
zNE5!iZvn)0Ev2P)kU|c<qrQ7OP(S*o>ZV7WcY3eWVScLVm^YI?kewF9>CJTZu`o8y
z;iWa&?pOBT`kNOOjc@K!TFY7kbIYh}LL9%p?lb2&hi7;H&nD@CI4W}FLoso(y=7Bt
z6OoNM)1q>wXPnN}377p!c6?PORjDaP)-&U)-2Y?DSLlFIDlN*8h~cZK+eEh+9Noc2
zlr1M@!W%@T2QGNYE0e^CEA8;EB}x29dzt&g7pJvs_%siF5LrrZUP@z*N2^^$G&t6~
z(|dQSl%|q@3-pg6#D^=3#DP_Q;_eDZTkQb=i;<PN;tL!wU}yveud^91{6LteVu|aK
zgQrwD>3i%joIEKXwTZm4%)2FLukE!E_k|j-wxcEmmI0}VsvTj$)7AA<8PJryD~l4%
z^B>4-C)cEnw=9_Tc&nE)sBqfUqLa>1w*B(8e?sQ?CP}n)WI4&ag#V|&ai+wVZ+XYJ
zO|M<Is-zLIDO*)>@1GLS(6JG_o7cYTn9rU|&y|U73-~?CRk(%{FGb@e!lm2u8Gc=^
z%_A+hW`2(S*dWf>1-hX=Ek<Sb8Nod?h?8eKdpOQ6R_Je{;Cb;`q`M{Vkd!N3^iPD0
zoeZCqJXN<L;1D%EyD0LN=$<5fad9cSe~Aq9Mi|{Z_bt{p%qOX`Y`P~%NDR!;g5Riq
zrkgZR?0a$lmZR_8k%`?I%?ZbgmOrbq6c1j_s;?czHyZ?aWB+DoXxoRO-H=<+k4gjo
zk){6jPnU9CN8elcy3W=T{v3~)ovB1kv8Z>wUR@WD9;t3`a&w;~UnHsvv^C#{M<&v`
zPpas6B(^SQz$bQU_RaMYC$-t48GC7lN4bQ2TzrzW&G-yDKkx7Q8yUSm5uFT+wnJNA
zx%<87Z?sJ~wdw*|!l|e4wr-=lCzyApWV(I(QRoK(yT6|4?gyv3%GIa*f8-7|B><Ei
zkvqE0$p6aJn1nmi<xBAsUZN+nUEOdpL!FUW@}u^N`lH#IqSV-}xVg|X9iGCLd+Uj{
zcqQmBKC%u5%V@7xQ$ep$Px)aQLfx58d^We<sQ?fc08Oh37i#(@FVHflkYmq~+kBF;
zkE!{Sk|^RnBonK5^w_>bi8;Ltrw`Nz2i~YJkeRU)ml^2KZ1mw25o-=RTw^1X_$Fp_
zK>pL`517Y609~@*58>I8fyny$c(#XKtWz)c0Yd8LB(*1Slu=_M#-bNMR-2I>ZSX4k
zj<2HGfuC6wi_;(?+O`jwm1}=Rap4;ZW6HSh%KGa%-vcS7>~2jKin7VMmAupD`Ie@#
zNv*Y?eY(B|b<Up6W8~zl|6yX>MRbHJg<Dz=MH&4Y`l@TZkzRB9=wn?msm;mR(=P&d
z9Skg~+-!}LSTi>rd1sXF$ouo&uXyK)0tIsLfyRAi=uThgJP(VV$$uS!_K*N5ICwUT
z2{d^LEz-RA0E~{K_x~I>jg3)wt#V3n!@75Q-&+@N5>RN%c(&j;zMtNo;a_ZWj?pq9
z-LizuM|P|0QD!v^TY^Io*bronC8myQnI|eZlj5kQhIfl3ROt$}ac&%+VsGRvr-@1p
z=)DYJ3M?sID&oJt_@NdT)+oMv1xp7^2t^}&Ztg~6>KG#AOIZW~Jaud^LV5sNO2D1O
z#UpD7%}<tB?bVJB+cSG^=HM@lq%_;ND4;7iK@`n#ia0Pr4Q85~bIkSZTW;j$FJiC^
zB9?s8Je2;*d2sII4&+uJ`$d|W#VQ6ltFbIZW*ECFOP&>%+8Cea7G#qi%I_j4QCHKn
z?U1wTRQJt$fQ0&da;2-x5tk|K@$oHpB{oP(<Xw=J96*n+#gO{JaOw=UnjPrKm{Cpg
zeaRTT=F!@u|KmaLmr6w%VuujNxNqh2?egkI`Wan8aXxFjuX3%A2omW(Er0VN7w2lK
zTwF`ufNO;4x=^=Zwfm<035;R$Tv=%%pNOe{>s#FTfXve)=A|sPo&L(tyK8In8h=tY
z+1`%(!eOFbfX+-|>SXUUsh6yGDDm_OS6g1~?{jR4&z&ahwOl4ZQ5L&2xMpUL-ds%p
zx88ByUw*2>a(YjJpzKb~Z}{C;27-09y|^&_7#gK7zifiLO!6$CX(qn!fY~K1A!B%k
z7Y=*&xr7>FF~m1!7@34p26qu028``<%$EXcp%&!~TX&RMZTtr6<8g(fbA5bbEnozh
zD%Q|^oK`K;H~+fp<5#XMsacD;hPyP~olADL?rd~<bN);8_kHd3%MCH<HFZO0Yx)}}
z;^TcS0%{q(j5i}^>kXBaN7sv)jssQVKI~C6W`BC9<}1lO@#g4#ob0?E+Qh*&wt0Ro
zd9?kFu%G&dp7M(-(_(kf6<j>PFTy4nyFcTvLvi1gHVCyVPX3b2!XCB3Q!1CCfwZ6e
ztyMPZC+tlQFB0zEa`tcHkxxkYN3k8&n_TM!j($Rf%$xh!GLe~s1hdo3Ldrls5oInF
zd>+Uv=rx%qik$$Q`J6oCYj2l$_WNAY>9sRnGnI{k_fQEY>Gm8oA52KqIFsEBxC)~)
zEHGn@D+>pE&&>i$25Ys5D#jDdN|N(=G9l|DWEp&Y^WG#dP6s2B399uv6V6U(bg>5R
zwl8~Dsbmi6vX>|JD2}jn!m_63W+;7_N+-O1l{CMegNOuNz4N8A5V!x#y<>(gmTeJm
zo#NW>v$Ke^$7yG+M`(HLWJ6qfTG6m8NpkA96Rk0|RqH==xBk)I)F@qA)rj;`+Rpt3
z=6_ur$Dv0spJ-U~f}3KePa9UB6ZM#3@{0-%s2DrO?Vga<OIACMW3sz<c5jD;u0=>k
z$N{~H%ip9*uN=8!siYUeF1$LHvz;Gx5eiDpRg#*L$@0h7U0FZ)F4}}nr34`WTduDV
zYCuG7>9<^QxK1pYfYwANv}*f&J~wUW!sQpo9Q+s$MOz0lY|?8Dl2WuC`0`0fH%Hs`
zvacPzUm>=C0C@aNcui5WXe=cuUXk1%ZHB&1EH+iSO}kFSa#62j#M{0E_YStq-1wxE
zMTiqP=7=oB*Kl=zLV)V}>y~fs2Q?XOS%aztub0lH=+^KjP+hvmG@q&bwVxI{DYcR4
zNS*PBrD(VdeSLjODk}Rc?XL9sOY$y5CPwwh_QMk->pceHG$)nc13MbL5bDiX0pCnE
zDdCPxpFZm=seT>MAWFoZK?Ztf{|ayKL*%{I^!tsCsYy5OE32iIyZaj{@f(-XucoDi
z%$Ot==Sj+i({x1j1n6CbKWe@v;_KLG^YF`8KT!_B1H7R|h#-Lki2u(2pof3p_=?%W
zfjL!Rf*kg4t6cYKg(n#g>Ax^|-BRtNX=9=Rcb%LIx|5f5)u=DLq}aa(RAm}2mI+<6
zLyzOp0oG6nBaHQLStZV|%}hBq@wWn3&ObSv8K>;yN{XZXK0{WMA_*sr@bXYb-;&j0
zNwn=o7B^d7bh~#r#3dmVtWMv%u6Q|}X%WKznk3xE`wTeP<n>f$B{R+kR9}4uq3b%*
z`AQL@qm>VE2;`}Dr}6mVLZSxHLkqJLJkRf?dd07`_9Z@G034g=4^0Hs{%&7!nBkID
zW4z1;M=xq^EQQXHZ)$2f<>f_Klbm?Cta*g(V`L-Xy4|U`h6KozxrD0heVUSH9lqV6
zuoV_I3HR)z6Z|%mi97nfi>H`q8b`IN7iRdJ@n%N+U(jP<iRQ|;R_TSmVs50_o#`Hx
zRBcl=pgsPM(lTvYa&o<)86Fpvy=RS{MaX>8<<kpovG}A_h02Sh7~Pwe%_X<=JS5gW
zHz*h(32*1JS{)`k5%&AG$<<;5CA5c{KCu+?iNkus0P+b$8gM>Z5iX7s>yO}4TuS4^
zpCJio#R#IPn1-aKk8qY8OGz1wGF+!KGE8d7?9q<0fMUaXnP`c{&dM9(n7wmslzTFY
zXT@{W{8f`Hz+PByC~>LxYHwiB8nLy?QC|*ZEMSs0H@D<nN)l9Ea&ZCpkgKn0uY>q9
zkd5y<D$OwaC6EtTUv>&flXc^$HNvr}qiwrWUs2X=EA1}&AWMlw`b%5y%}##gU>(P-
zuV0K{g@r;^M-PK)tTfx(e(SDu3O@MC`F3*!tX97_$R^QG30;k2CZY=Zq1|j?VBBTY
zK43K!H|cc7H{Qor@4J4t@SFJY@>fxcYRaML<n^1&F%1mpGUTB(CcnEUCM8HO9w^XW
z`Q_E4J83R@q;9X}G6s!1*({2pp1kw5EA}OZHE!VxWR7MYb_z95*b=(a?aNdnTc2@^
zj|GzRm|jY%?Byt(+d2J5l_hm|-!hey<T((vljp3M8ssY|bTZ0^D0Pwq*M7hBJacOA
z4T`a^+Ut{hGYX}`#Ge?fexzlbsQDY?7hH`u33vvpysKWYh45-i_gjcLl6qDD0z|ce
zJ`s^t{cR~dV0~$NmCVJC$?F#aCaDyW{+zfm8TRhN)rq5X^gYzD=S6wwExGw8RP}4i
z9WfXtYJ9^j2Bj~7(ui-P`@9X>YaQOmtW*@@PRjm5bfpUu>gYzdaV$>CR_?H*4UA6#
zkU|TqfBc<ed=OvbD}M7Ocvt&=#Lm9m&V^Nd2(EdZJ$q$j<g{g>6=<vzLEH^qSP4c;
z{FX}=ImL|ke%<LMFCJ3tmL%$nQ}o4AGM?El-JVv`A!lH=_ww)FXtj86W>bOQLq1w#
zb&*<iT)Fh;O^X=Qw8qD-%V<kPB?e@!E%DtYB`I;yFKMBF_v~O_<=&;QO0&6C`4X%J
zz80R{w3uiER6bMQQ0ChB=OWPnsECo5p3`@aH!ep4#F~o2ofl0;RBq%EZXXhjL0uh<
zQ8`hJII3-tk1mq1SogjEJT+Y`pz(ecW)hpG&g`*j;h>*+#-fl3f>np9&ipWkz%^N3
ziiQHjqX?_op&zH?v)~V0Sy3XExhGU1^@qKTD=x|tkEF=-+sbEdyHNUq#;g|GN}m*!
zrfXD&U71vld-0Z8!FgVE3vHsdXr;zXaLe0+`(FH5HDUecich7Dhh@z8**FO2qHl1H
zPd3~FR2nswJqn+Tzu8~xiqU;BX>=3tOoGS-Oh?Q!_CF3@|0r96zU#f8Dj^qG(iV93
z?3S6G5qgbymoMaVU9&8uZ`sM-HqXyIbQjA{gdBr;KE${%sm6i8jC3uxNFS$n{OWeD
znJdy5j9t&wGb5F$w5O?0g=Gh&cr&Sz6cD26-~S78$g8@#bE~hlc%O<|vNv=8k%^f2
zaf`WtUF*Wd$TES&rO!>&u8q$JHj|~a#+!d$$TU2QqR()n7>rW`?R9sQtwnBj&E7bT
zLJ0x!mQOuO2iYTq6vUci4L|PYV%J;+tsBDb`1SES8EE1(SIHQ~7fY<5@dn7drw>%=
z7w}Bzh>?*r#L!|DaeS-plqDM4Kk%%msH}W4H9p+R`@HI<O#fk&1S1o6D^dnZt`ikO
z-sLoQEwe>z`rp_STORgaQRlfn`n<PSIa+p$0gbBa$e%iTh8&z}*K}2Dd*bQI>}zQh
zO3o`$wSnfgC0~@uJ46O<JIc9<&PEn(ZQIPW5ZmnE{5|}e=mr#Ir~!N~egz`drg{6d
zu6-X?G*PahO29Ws%V5bu<3>p@CW$$dLRE#?&rJsLgdg%8$57+&cE5wB92#B`c#HlV
z4PKTJ;4vO}ct|_PTr=O&yloj<qQ`E-^aC-kUb{Sq9oQ55m$vzs>dG?r5$EP@6Fz;_
z3DFq^kP%;Xxo0y5GDQJgxLE^+R%de3()6X4h)BpKrI&kdp*3E$xcKY6OYn);8?1vH
zXEXxR#mS%iG2;3&aJ$_f$A$aiROE+0cT^0!x`N};co^I=wb2iX@282a2;Y#}drWKz
z-KYh_KCJW*MGM|!Y5Q{qBPS)%G*>_Ak_M`I(abPKG*gXxME?#my)AK3Oyx7V>Zl%%
zV044N1|*V+sWTBkfZRejcjzn(YYTICB`VxQ99o?s#p+_TToNjZB-XLp=<iE>;@&;p
zy5hX(M$8^_ddbRitNlsodGtm!F@52dGUUb)9G<>OQDyo>3cf!I3X7y9#_&`{`dd%G
z%}(8J9iE~C5nM8nvt3g*GiJQ`mOUU(fHu*4*xzKJ@frF|MQ7T>Yj!Jr_H01Wwri&P
zW3N=4{LfQsN+}{)rP-2Lc-Jy}K=~#U-aL7I1#>-thB3X)Jh8d38r`FQ&T8n$94cs)
z*3XGs<{je$+SX)Kl3?PMXP*yv@`*NYRt%us(95?xw5&TujNIVm*GUyGjH8ryWSNva
z^3?RNo*Yx*h^=~4e)V_n%%e4>CMI1)UC|{8tU258e&eo!^j-Q*9g5}$i7(Y5o;@M{
zd|&6g?6ex%>FcbhxKXRQG`lEIWPo{JR<h(nEH9Bl2GiE<I4b0s!VDdpIX4>=5l42x
z)MHblwx}U+G5wqic!3ZR7l#2)y&+TP$`KhOqaRhW_#kKxh~N$?Bnlxq-N4SlA?wgx
z*_<W%eN{s7k@fq07De8OH{@FLyFu%(SmS{HJ4yZ5k?{d5;s3@aNmXZYN>c@b%s9cn
z$V3CfQcEm&;JhUmJP{EbBt{KtFp6qHo>zb0KGAKF6-8i49GkobE^gszx!6d?gs7Ck
z!W*&%k-B%1#GB(&>z!ivFjNg`3D55-JSSKFh&e5ne>Em}Q`_neq0`<MZ@+w*wY+{(
zMn`nc;8C0hrY@ERzc4~>6OZ#tld)cXT_S$z<V@}AR7UPA7!9E)$9Mg(Sod#Cp!MFG
zH~Fa^TuL*)^6~LBFH<>l`P<2B3b+rPr0YRfzjvoX3J~Gbp=7HjDI<Z~Ps#Uh7uA^5
zYnY0$A1Ko>F-r{jv5Av%qiBvYqq+NwT)^;-Yr;)COqA6~E*UO=H*Au~v-*1ap-^gC
zjemQ4LTY|m%~u*O>a*Xy&m`6E>%4I>s=C{}jHa}@ZdWso-o4FQmLas=8J}IJf)<>6
z^b7B#tY?<~?Zz(hr=d=j=`Zh0s2(~VF$>IDodHZ*0j3hp&!}!4O7e0e#e1c=o|*QC
z#s$Ge9<AP~kPt#7SLgOEF-BzMa`8n27nZY$^U)vI?p2#G&|VN+b(*cf9cg=BS_Nwc
zjW5vLV~Turc*<ZYF@k*(Le!2iMm+5ry&61A#r!QjUcTWM;2Kvo!Jv9?cJn@wJ{i%q
z22AR7osA46!+k8WULF#;kWCSx-cK(sTK<w*kv?5iZhOgD%0<viIeB$q^KlXK&b!w*
zUBx^jIb|5%kIor%i0!Nf0=%mu(ylM;k{Gk@$U5mxzdbjl!)=_95U!#?dVZ8A%%!yb
z&6HO5F+IIRc}#*n42_w~Q;XypUj$@Lf}KH*RGPW()ayj8aZ@4QwQBTUgN1DK1guP5
zL@idgWzb#X-9#{uFuuI5&1L(ZdnQ*tSVt&1*xb1xR&Y-vaJ30;Hb$@5R8_N_l}~~<
zb5DKluiyM7UJ>5RgE^+gUF#82&7Q4kc34}X;`r;m#RYkB2D`{#$4FpDNiZ0_$0#f?
z*yP$^^HcNQc~X?trG?ESrumkMxzQu4(XMR0dQPW06<*~SpXQ;+$;o-)tAb$xZYRDl
zd{;g}YP&Y1%jSgLDas2kQh9&Zfd1w_7^v7`pO*!O+J?{)Hp(@4!{I7JUkWP0iMA{P
zOz(o&wTmmI^m~s~E;W1;=c8<sEA&pBETEB8*Wc)v>k6GPRHn2<Obu?4nyR%f@DZRV
z!%K}s@YO!agvOVBBNtti6_w+QOt_FO8=af*eVR95kBs(9x7YzNhtG=MH}EfQckAcS
z_wWApCL-kadCo@|a{?$*Ai^wO?QXffUnC<e1GNp~1lRQr$32yeW8z|Z-#29Jy=lBa
zRZL2Mg4r&QnE)oH7^)1NP536$9}#fKpQ&RKEa)Mv%tCB9CzJx>!1f0L6y<@PY@zmK
zKY7cZV0-12af0c+K^NkfUtiS53~a1?$3wXH`V2igd4A|m)yz=hvjmfpo8Qy*%pPAq
z^2uSde)QYcRy9uARAAqKLi{x-@hkJLNW8N-{4ld<F7$lsQKp#>8^hk*At5Wx&@_F)
zjT3akDP?xkMTajCL%Eo3;nAuF6ZC3dlA?&OZeJd_Fsj4<A7O6-4)xytkGCk4vbHD@
z5rZfpV@*+JY#Dp@Y!MYv*$YXOWoC@*k|j%IgzS|HBa}5+%S?94lI*|xGdj=re4p$0
z|NXCPE?3Wa&IzC8e&6?NyY)Cu^VChW3vS|-ub#iV4)PH9vWHMn;f{h&KQSdG_njSP
zi4IfG>jlo}&*v@m<Y(@L^a0O_M(J@-d+GT3eYKC?D)=jE>m*W&Fk8_BBnJIKRrAh!
z8tgGjE&9wKyk`oTngrO#XUK`Op&xUz4r?CrzW7*~a4)$$FT5V!+HofK$Le7>Ef%6*
z<%PGEze-eGC(PL?siVWvGTKeLWd*#jXz!&P7hCTT#ZGlKpxtMty^d<Ln+-Trh?r4A
z(=LyGHLTHdODa*<n-g(0p~MVn2ega?Mrq6gaYh`+;}Ou~Ya!2ofY3Ow7d*&%2oT>X
zg{J~21-y4G{;Rq#Y2i*iww3hl`ZK#x>J@+Iim^I*q=mkzIkTM~-69#=2>k^3U7s0<
zI{zrAdJz)OYs8a3W~QbbiJFgwjN${LFhy;Q5J8zAac5<RflZV(KY0XXPunVTb+P4L
z#J1avS^0B%I+j^qER;Q;7q|9?+)qrdG-PJv7tz<LmUIzzdgH@?-kp4SZEWs=ede+9
z*Lm;gxRhSvoge=a{Tvlh5fT-?N24~&Z2sco&kR3k#6y$Sgi`%4vA;U{PQhI^uHTdi
z(cR*E0n5f{d}6%Fy!ofd^X6Ga$UHE&vKoR6uYLRWjV&1E&R;&DIQqJeXmN984D}6~
zQJ5%Z7kyC{7q%~Gw?XFRxtf7tpFpHF^-<jyt1imr2I4+3j9lD{yzs`Q8&pNUI~>XC
z<aT`<bzG;u?aq5|@$KS~3K3gdX4@ZGl~ctdHJ8;qNYYZ-`kkN0+cK;yv^wo~t~y}-
zyj|klO=C6YyGq4Q$aZh9>MNJ4(wyTye}_h`vN+NenoH5eb|;XYPwV>)xZq0C`NBtX
zv&0JI_fl^YJcaIF2O<MFjBW!#Pm{Hyr`Vau1M_j#K(BG&&qJ#%1B3;j3ee8#YLiOF
zi|S#T5Zb31NDgi3ySnh%U}~nKa>3Kf%(6DeU}m@kwM`g)MM0&xI2P);8h91pD00Zn
zH4tSA9MlWQNffJZ(k)L-RJ7ukJfGtB!4u??#<_+u&d(#CL^Ul|=rK-b1?}qe75HOU
z@x#Sn^6HrRi8iVZ#V<c(GG(!_qxB2lGIxivCE9D2TqJg?>Q?uZF*Ha(C-a>fl{6X3
zzGDH}m-ker^YLjcR1;jM#s~9o3;;`*abJ>&mMii!%Ko^$hvw7YUsbE}tcp_3U3OY9
z7+%}MOrYc-Wk!&in?1KnHMphi+$tcXQ3t|;y2V8#Hu>j|rh}DUv)f~1V<D{ot3|TA
zYN7na=<1hk0wa(Mh0RENcL(rAVkm!F6C`QFMil_X@G-Yyej-XNwwV(i{CW%@f8DMW
z92^TIv%DBb)d3c7t<)j6cpH}(RWtgXtlU?qc%I_}zM)y_OTo@oZv{_N)7ZR>L+(>T
z!&Vqz%#yQ(`^P2;>LuN>X9`0Ty<2a%f4$1SzeH)lB4#jgWm!VV1xV8`qTh?FAI;03
zLYcJ&5h5GcKG}aC{vFMP7GTueLS}08F2mTEmOSlcNfqV|PZuXgGSzvPy(WJYE6_~8
z6itBCRybPVpw)UWF*EBTYv$cVo>ad?l3HF>-X8R()n_o<Z^qyfy1)u)6eglm8AelB
z0yn`E-{deG4FvnIGdy);m5#^vE0(_ra|z+K<VR6}6X<aEbRwND8^ybTNZ?wR`OqI4
zRa(QsLi?Dhe4mtxk)}U`{r^_6)IR%muf=lEEo_Ebi`UZA38Q%9XFhvKLy^|mK_L^2
z8XA_GN$=mR${yd$bv=M*QaLOLX5j|5y3;Op0coVYR{efwF<!guQcEs~JbP&ty<fcX
zlC!v8{p;#_NjP0#P6mf61o6+zxTDn3EVbwL@9}YzYr$p$ICgpJG%?(ub?mw*5ep$c
z8yeY|^wgawq)hWKtL+othS4F+%0>ryqD-vj4mPMoX(d-QET$yy&13B!cLGciC?(IQ
z@e<69?e)35`!J?g(_e~@>v<PR#P0CMyUm=iAa`2jwu^a=aFqJ{&2gwn#jNr(zroeK
z)YiS-Svp(pR1fmv2PW%EWvuh%UFu`4`u^Lp#Zar&bHyXLpMdWa96|YvHw-TMvpZSn
z%I%E<4o9raB;RrVAj`h+d!m|SF$;L&kuxtO4uroFcQ33Au{-cGnUa@vARkYhXHeUO
zy#P45W_B+w_NBcc`PEuQ^6C6(UCUDBH@qL@e>TCj2kbp`iU9DfnBH-@W=h<#xmjMc
zF88BzEIi#?R+lhKp@6BrJ#OIlo=Ch^3pLyF&&RBSW4J>@S~7?JL`j#_YU}n7?En2M
zQ+ODdOYPA`5+PQzC81Fjy?EW(<m?6iW@GIzih#)YC9;R~A?^G6{X+KE&{gjv(i~f{
zUAJbqPO~rcyjgk?ad%?C5iY-bc;XSBVdDnlThp6Xg6M!yAzbS})dj6BE<Hc2kLqou
z33vc%3jI}4(MgzWzJLF|Wdnn6yqC(NsL?f3Zg&9&7nw&X)5~LpAkDnes_=x+CXR4D
z!FCT4W9xO<1vfvtu^)Y&^OuS4b<MEUnUis^MaLg(Iu@|60i1$Dt`ysHXxB~%bH;jl
znJ{c}vXWvp{S>edqp!QSlV)oXyqDRidDi2XJnC5E6_V>=RH;DVJGVUCH3Ft$sieI}
z(v#x#TeCSyJ*q><BHDhwJUyv=B%!yig`fGeR7C{SyO{tBlP>)-^^0Ac)dctA`jUor
zTZ|ORHC}6;R0qg~-(ijdjT(R<g!YdM6umUl{nq*3-6<%vcIE~UM8RjAirSx-9*etq
ztmjymfY9z;+k-OPJwBUCo)1{hbA~k$@pOKiXD*?e+6%4V#=wZn4n~N6u(|oatWIR4
zcwuLb9wT#;Re*7)(_8a9eUcfwJUcwMc1IK_NMdlx4-uZIrQjTTvEkCK#rEP(?2UuG
z?%w}f4jHA~amwUuZwC|W)iKd)GI8nR<APg#B13G^jcvDda7KGErm->eDJOfwDK`|;
zQ3>_Kj+{_n9zJkO$*g?JZOs}{dTW||^-TU}p)I2xMM&frF(_X6A(KeN(6imMaGNXr
zR}U0eR=tG?I|&Ji>oD7bE+jqJLR%(jZ+R*hjTKw;$#1NWgAL7@;@&b#B&O+ZYg*Wb
z91FXX(qD<|pANSc#b-#wALYD9bHr^T-B7k5bA2{xPV#tSYG<0Qxi6Lm{`6NKN@*$*
z0c}0mnz`Nz^S_ywu5`AgWK5)|i$5p<#NOtj-L>co8NBTF>hT=}AN7ci7Nv|y==EGv
z2To`0?C|zm;_CN4P#_^X?m1SVM<Gr=3;*1z+-(UEme5+*#6k6F?7qFKr%;NT-2%y)
z33e5u!>!~hs^<giBWa}*=PjB$+MWpUo9?^qi8=pOVoRhlCq+MvEFlyAN1XW{py$Q9
zl-7m&{=TCERf@V_q2X?KOM7Z%7HXYdV?~FZ6-S0`Ie2h%wuvp|ThW<3{~L+^&1osI
zQS+Z#qCMTrQ-sc?sD=zXB+L$Zgw`A?N%8t%Q{PaLd*czVWWV$6Z@r2NVv3Je3sY`&
z`~r>;;i0%I8)Yp)k?pLbr=wG|W<awWd}FrfydfzIPkira5tHugL50dqX^1WvfmZn3
z64#wajC$Sgf>J+vA)X%>wTJspJp8CT*lNwz^pme#WpJU<(e#v%@d!$H6Ynn4cw6UK
zTOEGr{Pr&vUFz8eXUfda0WOtmT=28oG38sEuFCuIt^zkma)M{!PjyP8os9(%v<cb+
z!dthMxPq_^^lZT56~8zekJLA>v&}S5y_2Y!qisR<hy4w)g_aR{1&w75(zaEVRO=dz
z%~C%1nQ8^P$7XVY)z7Z-j=zpe?dTEcQWQ`fFKQHQdqn%Hk(`w^JSLS>pO5J_2nI{n
zd`Wy}nMcqwiHOFM=vASvD+0NQUee!_+alPcd#&|cx9rk*yU^v;GJ##;E8;Zlju*Wz
zgab>=CC8XO1aih(<?aYI7*jg8Sw<7*UtadgpD;#$m|K0GB^|!$*g7A5+k8%vcc2^D
z4{RZ>F(nTs#O+6y72^7j)wSfzO~zX45C8g#K#cq?M~2fPhM>sBz#tV1+fm?D14%Qf
zR?t29NO(P^s9QY9_y$isnw&188Se>}92oIk-JhLg{Bh5!GzF`tY=d5`Ejq>Rsrmmt
zrfq6}k4)w8W{O>!+0PnCPR;~4Rox&0L9NaoiGZ8F-3W|ELbUY9d1j;fy*nu<#WgA0
zrw9mu6;=F7S3kn|=Ji8cuk}oepb9Wj+qAY*g;1w6=e2jgBDAGm5P?urArUJm;vKoh
zNCGU4CIl27EEE`AuEw%9o}4)V@SKFCWDEl+_7IIIeTM=30bEvbIHaIQ^*&7*D!|9m
z4yeYe*+<uT#Nnu|hJm!4m$vAER{ck`i)y+TfaesVjJ_>O&d=4{>b>e6Um&r0yg<Fc
zifZqrwT>DNC+pz8=<FyEh#Kk9;oYaI9_-Y;cjgta;8bzumh#tPBiZl5-+$MWm)c}l
z5-Lef@;G4@Y_)0<WndE>JKvR+wa{?o^ZT3Kq@_asu2fvC1gdeI4P%mi$%0K6+c<Mz
zer}kKBfk_h*7WlJ?kS~hu6vg>DLy{l(9r_F87bn_s!2O%J^dBFI5M2nI(!k>;^#)d
z7VC+|fTh>G%TMqfrOCqGAAa4D;b1!@>Q}h`d)i@V4xq6BM}QgtqAbhVBVC$Pifc3#
zXXfPWbB;XD&fHoc+56xE(Q?-nGJ;S(y1@5e?x(l|^oe;_50Rxi$+kE*-owZfW2$Da
z^m5^HeKB3V^}SOvAX-bG^ku>@(B0ip_j*dYxD%!Ge#TFm2hpp`AbQf+js77``<Dtf
z{}W@I(6Q6<ROlZ9V&6EsUj&-zFw+IPuNEocz&(LA8R#dOZQly}{=lLo#THm{Sxetp
zqgFF;zO=`X*;(N71p_+&d?HQwb=U>b5rHXD<{KT;xXv5vu~O>&QR=N;>Lc1a-U+Oa
z*$ciOwM)P=x8SnsbI5`;d(-o{r1_V&WmmCS%98ezB`q0&Lz-po7Dq>$Fx|$yu~KTJ
zm?gDmF+gK`-j+=`mVs&%7w|q)B!SZWUeN!8F2JaiZj4~61dPPVZ0v1?OC$9JzboO!
z71Vh<+vv3UlJUR-7Zty+u1*!nnR!`<K}i!M{g^A+bJW6rtYOew^mpASLOg+|rx=br
zwlkIy$zR3%es(w;&KE5qIh*Zh2M+Xk;CaV-wVGq=9qRKWq!O%wGZ^gzk&P<bvx6dW
z>6M3SrbaN7tc2v1KXH<?l@Y2sI=LT*ouB+)+*rP9?&K(zFHn>d%^hw>5r^e2sdCqs
zrRP1VvnasgKzna!w8XWLMt(f{<39QJUb=ue0<FCWZVs4?W5It|=hiOm(0bzQ^g_pB
zuxvDy$Jw=L1gZg-0n0}h7t1(^SBein1jDeGbNq<yS=<k`)~m7!K(cA-tApQSZh1As
zK<?;MI#9gdOVrI&m)x<Q%yrr-kXS}H(-?B&J#sbQ2^<f<o|}bgTzuTWU1{D_u*VMn
z!ZJ4bMWXqRdv9^%B0=r(hEGf5pGspws{|w(uY4-a4s9$?&)a#-TeR?{#XkS6wo%-r
z0^m|TvD^?-iZSoJrqiKr!G51+ZZfv-jfulR^p&mA-t+BOFn2p4QcqVGO&Zt5Fma7u
zOcvWx!Zbl<wd)an{~gE#hLK_1b^_`<t}~wy<@NkDRF=e}OB1`G$|B;iI}lRNusQJX
zg$Gr+b|z>mV)r3C*&SsM9=e1i9;6TR#K+><ol>>T`ovOOuBM;ERST(!E01H%4^`#?
zKYIVL@HL}bIX-el?hES@J-+{#zEtK5jB9uIa1Ip~Q3PqMJVBFDT6z&L)%Otz3a0)k
zPT`O06coo7jzXPFbWh2A__R@f&0p%Z6>HC*j|C10AJ8^ko*xdBF07WLuhxPr+K$wd
z?;G3ByqgCY5~8vIhJ*k_1lE8~8SvriX+A&iV5fLk0tLex1aU)~sU#7&u_dZH-ZJ{Z
z5Hbg$g_0E(x_^H#QF>(5dM86mX`7efma90U&J4S<Z(5W+n`Pr(iQ7N8ox8IYE>g*)
z5a!~w$kd&mHxcjTh6{RM1kqJwx|x!?>ATBb)%7*FSf-k#@jYcG4x89f7t^iTV?FuL
z?TgiJGk9B^d{i-|CAM{$8lvO{T}WuNz?~81j-@NH?M|jv?+&hpblY(hC5Ju_S0S%^
zpjI-5N3~RunxJuDWccj+OL`9|p!{aAN7wmcLpvNFesX+iU9;*?uk2=Fd<;B^MP*+~
znr2@gTP$BaWR!huF^143Hrnm@LD$SOJ2&@3(Rv^AGKp}@(lIAZj7BxUqM0^Kp+OgP
zloJo!6c-tv%9+ZI;0&L7F@`7ew)%Jor<BDK2LjgAdzWaEMh_^u86u6lZ>g+!T$dDX
zyb=CtK;me`_4_bzbn~6Kqt5c!G-Wd5bx*}DuxY6Y(x4BA+X5fw;;_HM#&=@w2f%47
z(`QZR<Af}$yzdzoSV95@_->vi|IlGz2cM7+!l!!3S>QT3mmbUMz%$A=bNT=|n20)a
z6=$x4&7JX)GBp*nv54?{Koqq<%9$k8ZM@G%oukv<R;j~;!AWWU6zccbe3;-9&5u$_
z#i8%KcNDivo@yHvDAYJM6Yp7O=HEp~$XKyEGYiyo(RAo*@fzXMtZH4mo>8Z!R?BW-
zZ_zpZx-&Rf^e8i4_fg_%WyU6?!;8{JaqR$cF+?DN5M{wH*yb6YXwhXj?5!(=3QrUt
zCh#So0u>_}JT>s1`$ND7BiixR)1p|#sOj)o%jT}6X<nt0YpRQbR6E5(XXE}M)6B^K
zkZC_;741~L_yD$_&(m}7ua+U>;f^iD^OPluk4%l-z)*4pX>lbwQs!`GRe^QQFpykb
zn-bR2`Lu}Ou=IJ-1)7Cr;d2kS-8nO*6P+@gTXF5DZb5(wqK)FNEOs^XjKVS|3I>72
zxw)D1EJrZtyy+G1G~AKyZf;=3%)1OPeK>KmIZW}%PJtzlUb?a<yZsfg#BZ#lJVECQ
z&!i5Ah523&Q<<`nmd^xYikR7YP3SA{ML>kiEF?Z}vyLXrb2*|@+S~k%lWfNu-ash}
zg%JT`Ng96zP<v}a#%^{Jp+{5WsNp#7<%&|Azqm>YR$RwYe{5M}jIiD~ZelF1lX-1?
z{=#wkJ3y-cxZgYF%5x)qT%HCziN7i6#6*R$;V8wHZk9$e={Led!`|ZJi-yZ;OF=2v
z(xZ|?<G)MC7FIHav!3jhyrS~|!*9L+9a?eV+Yp9gt0<SNpZD8v2vGy-CEek@m!6k_
z8ZYefj-L@#MVEoYcXTO27_LHzKt%Lv_do<l(CD&b?zv*ugRfUp-3q$&^&1<cUpH)a
zabJPh$Cg<VhaiNpfldhMS9w{FxWO-<tYYi=q)e>ij{-jj-Zl8ZLg)DHiAd`kY!@k!
z3t}sK)$|_2ao;>%uWAWN&A3e@2CFKUMJ1GoO2>AK4<rUB3v-=*bKwGvt}E$Eh2czb
zvp%N7`+&Rq!91U)S6Gqlq=nB%(*U>M_)G64rgWF=8d!+g579Y)yDqQfcz)YG9FJ2a
z2@Q6mD}+4Q0wMaj9GFxL1rBEjiginTI=SAzZ@p!DeMdm&8p(ZThle`)|G_jRm(<na
zi9+gLIjikKx8HjyQLOuH574Z@T#f~PB6Gz6kJyCBGAAw}0%G=Fp{hke#8{3An^$Dr
z)ZXp@uNPTd=2NpEtqo*K&_Bm~hYqp5J*f4BSoV!9{9%5n8bQZRur%@vedXi_SPzFF
zT%#T$b94lR=Xgs)V?~tdycG3-XxxYxe!65!zJ!KmSmEI=uZCe@iXOE+P&e11^kHra
z?a`N@H(!N^hXX*mNuN17hDD|Ff{!T5T(RI%M=XPMzaVJDbg`5~%=e*<4q^fimna6S
zjv0a`ChjQhoD!LPJ@(q#*>c;c>>aOAJ8%8sZQo#VZrS}`d>b6}o>XZkAcHV=<oV=q
zidW>EmlWCh`g1S-Hy8@+ghv%tP$&K3O~7b-zG?{GzKPgB_0>B4y7&J~4oV<KXeG~F
zxJI59z&&&@4|^qJKMK`pl7l+P;vE>vY=d^3XpDh-Fpom{ezzS^e(7j$AQWb3O@ZM-
z&PzBRY3^Q5@y^Jj$A)KZRm|!jgxE(PRGmka&)ZR5$D+&UMqkf+m-s#RRuYiVyf~X}
z!AF0lD!DK?$WI4lni{AEh%8s)e-NIqA&~EFFmwRMfuS4}c)~@N6wEvOLES?&4ns+p
zy;i+020A#|svQsO*842-#~bIr1#zhwOS^QrYOfco<U3_~=IecNnD_U6E?sI<-teX;
zrkc)?WO-;!-AyVzB^)^8ubkTjg>y(2<DHx9YlzXe6cXjDD)83AP+m8-5j_<w_Il+f
z!_=+-uaw5QclBM}vpaA1_}@=qC`KSRu$*Gup*^6Na+xNcHezI?;g;ZZB(EFDFX-Cg
zYp~5L;tw~9R6ZR@gigZI(Gf9C&3nTG0YX?%En=B$$Q!P=&)uc(67nRUFQJ6#4`JR1
zkXnBgYp8;Q+jLyK5f?Zk$}byX>p2ux4XWGoevw*w{@lpX-ms(<2#R61#K4)GzNe<X
zu?Ec6jdCSll>fPAS7EYzD@$gAAB}!4m6A*Qg5R*!(>tv9el9_Z5f78o5aBE%5cgr}
zMe%b6ga;IkJh5u$U`DI?5gqABZQR}?tJ!}hr~(SB=#xos<GSN*SI+!fI?I8Xf@lH=
za`jBcrb%V>D=E+UIhpQ-VFk_yD6Bw7!w3EV5)lr(uAa_b*zq=A4qO7F-2^`^e2uJw
z=<^L9Bm5IOsZ4)H;0szWy`XPak=|n>g$FW5oB=>8I0N46BDWU<ZYLoVX1pAjMqu@i
zftZS3dY+nyZ@Ck17?sz;xAwz$q$#W7*c=m;YH-CQI<fM~|A%?B_8I@sB>$yCa2QtS
z2#Sn{Pvv1><j)w=$9dr}o(-oCGVAdNx^Hg8P#Q*D{O$|iD!-bE+%<*6YiyZ`^V#30
z{<=x<kO7j=WWbmufp|dW&)Ofb!e0~ICV)aMSJw~^^B?*_@-)aAJ_yV<T$1T!;0F;Y
zBV9@bmqfI|4$L3Y2kiUI>2KG@7magJKzd5R(>ZqSxlu0sH}I=Nt@X5s>8%KZ?--}m
zp4Kv0B5Dxmx6op1x|c$eg41I|xDwxV6b=)ZO(0Tz{0n})cEI<MO4(vh-=4?L@;~{R
zZ{1G-&%v!X{{>k0ly0U`YqR;sT?6l1TL4omBqH2KBtTs`nB((subPy+f>Tz(#bhjQ
z`_DHh#Ee*Sdde4_DM>^UGr)a!L%1HCcV}Xw)Ze`n*Thgji2*`(rP!pT#iNk7!2k;u
z;4TiyiNs*RMxc!z@!~b_YAP>dyr;z|0zxzJ8Nsj#79;rRWYz9Qx)t~?^e;O5YmdXO
zj`+e0t;E{n1{ey6Sxm&0QK<2|h_djQ_hY7Ieya$3Qzo#77<9Hh7y5j0=_4dj=uTey
zRkJ=BTOoG5*kjP;p2F60nmQT;e-1qV1>wR!D+*1|Q^+k(kjIFhbp346@pM;*h2P#}
z+x~#H_c=-VJ3$4Q{Ic!6mde*8XgmcEMtqjBkDqbSs7pv!2z@ptw>#{8+L(Aos;_yy
za;%O!2DdwKQ~tOzx?>O}u%!lG2Ln+^1b_AF)$0uq7X+vvh!?p@F9#dHsr&bllY7eZ
zcfwRG;5P;8hT{(bS`rC^#fqd>WHrK%*+ra+(!H;dM^q0Xg@ZvXW``qZFZ=B6yDBS9
zeAAi7x1yLSVmtbm;u1QmZFcFpB>3*2upS#2y<+n8MbnsH_eCe@dGvo@#rBITp*vPf
zEE`5UqIoO)__!ta5{vqcMzfp>jL&xV3^6Bb>g&ko%-qO7R{qS_gC_}TDbvWw&%_z;
zCZ{GoY*ya54q}(yo&Vi&2IF;7utU|(&~FS)$*H^7?{uv7xyMJNTtgwT<?5E^?mck2
z0Q>={btZ$NiVDI!LO?r(HuE%lGrev6d%LrO;N657V+TFnAw+ln$T_`WfQ}#-0wksI
z5eXxK$E`#$BwYMR&$*K4(2=B`<#_HlU~F^9PGRlmshYRf7;p9L!N=dJutbf57WYGW
zOEQ(ZDv1AaLTQl(4g==i`x3&}43MCp*UJ4{D`-R-b#5DNKG`p|E+Y6_zPn+T7%9CV
z2S=pW1uwE!-cdF8x&kp#K2Kj+==EJf0l4?olIZKHxCXHORFnZDh8>B!1_eOZIMjX7
z=ijo`81a09acA*Gpw|UsL<|-Y2gECkrqh%I4zx)F4FT^hTm#r62|jSQ8#jVLJ`KMn
zs0g<+%UwVQ@W0Q0{jaAjmRXWe0+0+Qm-wTA4q!20vX4oJv5{RpnuBQnp;RP{n6Ab9
z^qO(YtigJzI6f_r<+9D6cit=G(`#GSORuf_&TcVsnd+Oo>jEh(JLJEV<m>7CLuA48
z7zcNEZ@a|A%1F8eJY}WAkJWy`v`k{TELvbv<~K!D*Kk39evUe9-tm+sq8$om+$X}Q
z0{KmIhw`UyD`8A;uITY?f*<#HW1&52y{?~iq$SJr)DdHNvClk7l^=tW1m!8GBL{o0
z*frtp82jNvd9BeFU|cVs!Ud`U+6{W9Avx~S-#=lDqOTf)C3`~?6D*)YfL8`a5QCoR
zLp}do5x6#q+hOyNXJ1UwUVxDxFiMSP2;U8!6!<_qlA8TswJvTi`G67zOuc(r0U{)Q
zP}$#Js1)!g!OFGT-FeGY<H=)fSiyb_zYkVdC*?kAT6TV(cF)ylx~>a*Qc1?*#I1_f
zQc)Z2Y{khr>Q<fCet3N@;sPPjS9-MJOsgKGUkD#E?9-2#zqCoNlgobaHi3ATCQYNN
zMrj#z@^!J^d0N^|Z_1XXX@8yxmIKN2Gtx|@0b%AjEJ21kB<CD<a`br2=<udzIv4QQ
z)UdIm46Z#!Nk8D_R%30H|Is!x8RY?|jE(7Da1p>9liTsOuP+IUMEihx!x?+-Uv;wH
zO<#n|ARw3F3}mnwi0%P1M3<+Y_}&yPz3q2kMguJqXpQZ_ADGih%ouPhG&znjJ^Rj%
zJ<{F^8pL8gwF~(MqAX)y@C7{sm&(6b_lg*zy}DX@T_^dH<o#G18k}Ml&5xN10hF`Q
zxNSlD%X#PPM>EosZ&z71c)b(=a-?5(b-b|wt;*V>EuRgPwy3d;_i|B6%DdOUKlkU5
zz!{};VeN6=N3mV-b1Trkv=VEa0}!1qOct=(-RUd3%M;ivmRg7`sdJHS1mtb!+#C~7
z|EZuDgCzxsG>MfCw+QPi)+>>B9jXI?@fo->Fr|Xs0x^a`VHJ&m;fZm_|GYy?Xm&V(
zU3=iaMhfQypz6K@R#-iMh$?~Om@bQlqZx~Eb^F{JwvsLEmM@aF4J<%A(jguw_0*b;
z;qR4sUV0{H^mRA$t?{Z5$K<jss4YwZ-8(4>S;AR*Q;p6#3Nms898R~=GK;@qVCKzL
z&iY6Dc2Z1*xA8f9yGNMRX?N%ob<bVc9-)BYzAeYJ<F?#>JZVW(U#XQuK<|g2dE&mU
zCKAqr1Oe;;`8!bjyURcp$>ACPUH~FYKqW`8j@&ew9s}fwV?};0=~Cfs=@dq%!#(HZ
z%j~Zs&O2D}4dI}WtJ5<H-t&+i{NG{%e|;3^6Ue<{W9#xXT))D`0D3f%y^<KWrpnWF
zVF)?{9Sp=IfUv%!tKK@eLu0)ZN^+n(8w369zJsDD40a{S2-5a!nYt>sGH1shvN_Lt
zV2WSKE8qy%eSWwIs_{KPquK}y(1{r;T_GlxMwjRPMJphO!r_Wt^OjtFf_>azVTwHC
z@c`rjX>b2}-mh`szO+4@;7^N<-#e=~JrAO*(++X+=_St?-_pYQqn73zy>m{NnOkCS
zT8q~HO4X!DO(I}%QW4}0n20;{ee8u$Kv<~?ohiI309ycE3MXv`S43~}zp4k1N{Eff
z38N)5``>2}wK{q3bN(4Y+W(v~<qPU|N@0}T&yMWGRfLDwa7#TCFVeu%HF<R@b86l*
zzbh3tpIlb^e&4Y!m#_atZ){hxG90(AW8;lGdlb=V1=U5k5T&C0^Y6yzvCZ=Aw>ili
z$T@reabjYr?eNxIJRxOAvtghVOCm9qg{LfiWwQm650iDNWLS(uP&f|vE4|SV0@t2@
zIdm;9vyxpvG!~m7q}T0fk@@Ba&uHp4Uil)Z8Ovt!P3l_0lIKfo|3PikU?(v&=EWE9
zO)pQ(A!F>{_kmhpD6TpG+*ouEkPlEO;N^!X%0_bX9^ze}U*JW9R1F9lQnQ1;D?alg
zOKoa;<(5V=Tv_^VG&soMn8O<Eu4PLeKJnOIb1iLwxAyYyIZhns^;wkgz>)Y*Lubg>
z#(x)2R8||b5v(MHvtnNRJ#!}$@i*s}zi#(B*xmn}dacsrLi5EB+BIv0Xu@5e6zRBl
zaR`0zzkXqtc>mhmu<*XeK06$P0NAv^>U?lf@r?SJgQ#G1&?G_oO@HuCl!H+4;6;ZJ
zF9;h!5?IfBriP7RozN*G4N&&syey_K^RhNI>TLSB3+zE2;PNrZasSWMvOffDy`b1m
z=lbX3z-7`<{=Ajn15%RqhhT9KS`X(s0@zg<GYn-dbv+e<T=H69Wh;&t0$$4Np;Nr(
z>dRJRv*$*!cp?gQt))O=;Q88Fs1f6%Zz}8*(2L3YPbXd^&(%5W1`*Xhf!AfDPFx$9
znB9A)$+)*mwm!iQizr6w+uVu*_cu;C;j~EVv4kW0DK<)I9h%E=xbxVHmV0WFw2YA(
zc{AiD>$R!HFqokEQM<z5uX;VRH`OFd2{rnr-MS6CA3*m78F4U0$VJ3Ee{$pbQlh=t
zPaBl|Uo+_#2H=zOH9}9$;&RIT{?&gq2)CA&ioX8>vqrl3Y9I`?Hv4m|nb;$Yp@D~R
z3aC5|Mcw$kkL}c9_6KH5mxB&pSXTXg_oGO?^3Tb|@~Sg99YdXFStYL?$ASX^eV%3u
zV!~Nb$(8%ii;IY&aWt%qKCfpVwJgwg$3Eys6)iK&OkDk80tQsVtj9#2_1%tduNt^$
zJ_GH44vVP#$r1KS^P2n62=PcJkePOgEOA6wpGZiX!NgC%m1rF@X(bNeaQ#f94A4-B
zx*5yQ!<tRk9#Orxq_o8|x^GD9W~2-?5BgCU_qK5Rl0mrlmNMXOL{%|kpRzm<fA`-P
zBba9&D4Q1iKvQh@1((eLGH3C<RtD!F8HyenA*!W@CmJioia(9C*9apsQB>K0gZxw7
zqWtyHPk<Cz*D=Qq$@!qYdgiE%ywbD@^)Xjxv<{kE())u$y0G$8CiOX$`mv2LD<-@Z
zRo{Sy#-ntOKp@;@oR*GDOXMPE9X}rQBvC(_7>R1Uek{Tggnkr?cbGfwg3i}Tx-BBT
zHagLtC5?OS?81EntAJQkp1K!^Cmyy}z?&>1bFt9GF3epg8pU7_QH%gW?-ZiVN4Yxf
z58)t!eR^Y1+<+XR1qI~@UWFk4&lP}c(IDBdrqojs^LS!e8UA0t{MM}e`@DGq5w<do
zPl5&Nbr<`NKg{T6Yfyp_rDQT=6LVDn$bN+$-gs!M00AX$mPdbrr0>*MYGHlDLj9*J
zC9Zwi$13)S>3ijrd}yRnrAushsnkbU^l>!zmpUKo(g~xPme|C^>eR~qqS>1hd8ZTg
zmG++kC==We+T2<=<(QLiO_sBTMOz9JpCxj|7z>G=m-`%a4~oGQGK7}8!p5DzWQHE_
z5jYnEdk$`6!>{?F`GP8f3Z?ccjH=*~tdZ^rQsmVFWCoJKwU2xOwg~=RXcnFu0ks8Q
z_oP8zR)n+r`#j(SpwE6vwPaquZecvXwL0C(!U0WGHjmq-yBEgLw^Ck*(Tx^RNubQa
z2mXKv+0VP4C8MoMYqT*sB>c_QkFl|(SJ2!Vj=i1%?tNsY_<}jAG{w)fX>>9{KbDX&
z=ht53BO=VFv&&2VYO?A+1R6DDn__@lCB$O9#ciI8X4@KU`4OW*cb%hz$BlfXynR@#
z(~sFsUTfGI3iqoU|L1b^>A!|2*v)ZVn-d^Mol4M(2$_O+Q7!`Z$f1+WcSmP$21jx>
z$CAJ@$pz+RtO0C&1{+i@!fCDwMe__2AO7oQptAdYklzH|IFtc^U4S_QB_#X-{%476
z%`5Y&bt63CSmT*O6N7L3QU?WuX!g)!4atG54j`J1paj^>{mAuIz9z<taix4VY00nD
z-MwVA!AWU3x$n`o30)moj+)7re+AE6N89$?C>8U|<-ln}_eAB2XizHb`B!<kIiVO7
z`u=-D*p;Qh#|QSmea*(2pZzD(TVx8uIFuX=hbP)v`KV~5u{fu=u&5@(LIbZH{6aTt
zS7Y1Ue3N`V>jY`q07OL$9R^8cQV#r>C9XxBl8*Ij_~2bi4*<^=!o<1UXBre((!`Vd
zxv%!fW^w(iKnM<pUVugvx<{~L;$!R`WE#nq<|GN;gCTce(owUo(oGJWu}x1=VA`_h
zFg#fsfEH}Ya^&b`w<?_$qS;0h+B9@KA3W-NSUz<=U%~Qkbtt`dQ-8+D-TeooLBtYt
zWt7l;t-2R=wv+tyF?z}0_|aZ0wkRfJf1N^c5ukcXRaO_>n9f{rcYZow-d!ykXr_yx
zobXfgC{_x<GfY-G@4jPZBN&OE$@OmYs6j_o24Q!lNSM@BAum*`*D>aU_<l3w49&!T
zf8o0P{?HFQ8L}&kCgbHfCo4Z+Msn!Esqk`ln>1?qUq}W1<lLIk;1kJD^pew&0t;mn
zTg-A@s-}(hDk+(&Is9&sw56-kY!AQ_1J4a$cax0Fqt>_I-CU77^nLo=<RK?QcgEeR
zvpPn(R~xEncGh+C8S!tI-O=bru8$efUQcnmwRE=s{>A&{SeCeOmZV6*!bIgEg)4^x
z$qe87eKJGbwa$MPh{$I6K9-z)J6E^?m9*z5fw=FsGRCW8vmmL0=d`fjPvrgB_3N`8
zYAez`LH06|z<{Lroej46qj3R=NMUGVZ_v51N!0Ibis&b@_{~pKOx%xQ34u@+-T1r-
z<N)peelMbYZJ;qgll)&FW@2O5OdlnTaTD@S*aY}B`Knxdj%(k@ZUHb?P4C^>A_IKC
z>S-}m*lUprAIpMVF8k3~d%iz6;f3hijqMKSt$aFD5=BtSb9Rv%ELSPJYRb@L?CpMl
zR-N}_?zgrw)y~UXJ{{2Q&4~Z;Y;r`xv8W5ZXqJ!3Y{d5`{6tw!ZMR+tQd!+6Ol}~d
zi;JLCqDaf4n8oZMVHD~l21yI)>>c>!&Jz!9^nhJousQ#sRDk8%8AnZGh?|efP<LAB
zbSjh`s=GTR2Rs(q{;(+@cj60iO$cogcyP$u@m2M;|I|gOkoO<WrHmD|J>U^5;-Gvl
zp9)_#%SicjoTa0bj_ciXMb>^b6oZeZFqI>9NUj-+DZBVQ$KN!%x|^~35R}KhOOvsh
z4o*zhd>RJYAy^lCVLY?t?9N$X;To+AIzb-qdnd62Il(djRkRU`NLNqUlrmGhg_;U#
zAuZ5L;il{$KUddj2HS>{7JGul@<&DM=F)4g>z<=WpO&ZQ35$+1Mut+n)7+&(%yy>|
zRBpr1KM^6P93jbd6jcxzN&*lO;_}phjDFe9jor-$ME$-?>T0nOS4SsgwNK_|^6tW?
zUqXOe1AP`FXqD-GvXGdjw~WGvf45AGkUO$%PyNGSc+lz9QGMH9jgdOd=sV0v&;0v{
z9*9121nook#E(1?!(htjwnBQt{bjvs>ssGA3<ctGvmH+f8}*#i(Fu5Y*avNzoP2KE
zqNZ-Yc7s_%LpGK+9+P<f%^b0fz?vPBS9*T?yW^G<aXK$+nog^l{+dDSMr^gvA#u?R
z0keMper&78ZUMMYc4EAE;;f(iLeIBD{*PI%|6l;o;Vn8p#m4Xk{|qu-pm%mCGVRnB
zlvI;{p7FFtS{pjQWoQ6*jG$m`E$HetPPnI^PlGh+x*Ukd5eU0wn8zH+p8oq(2r_D{
zJL7P0oL~(|uOqc7kkDx=xcN4OQz@rp{mhxsEHqJrIgOtkN&&qZ?1IKJjzPQ4Zhe9D
z5W1Pl1^=dEvgTF`t7|>gX@(h<>0JQ<4GpM{xSYfla`8-DVxkQyd%9>I)YZ|~!x|cv
zvjd9E_UnYPa@(8S)CM25T_9w`2@{C(%$szsrY^NNvp|+vL(ut&-*zku`g*;>w_d}~
zi-~+6?VIb{v4(^$@5KbijJ67zqM5dEg{#MGX%zh8dl=PCvOlJc5y9~EMpgiA0a|O&
zDZpWBZf>3mOF4{kvjqX0DPETT_fx<pAhIu9T@I2UAxsRocZM^0J;TC&lkoN<EKV^P
z_!%JmMhAn2R79AdM4K)O(%|m}*Jb)g?#Ytj@nReitg|}f-I}@Fk^X%zLy4w)Rf{zK
znrS!q|DmoV@9*<VyBf^4Tm_E+?v_z{O-g40|Fyd2PVpM+U)oPuHKj0KG0#>saI(A<
zG{?!yN;<B=mf(qdZ5oq4VcmVz%Kp{3TQA(Y`rWTu^kv=sy%!+gZLMK(pb>y6HF>3s
z8|@{oz{Y?|1MpTzA)qf^>(mxvPOtSelrIo;-yiaXL-31g0&m2S2xbxD1HfhpX~$~7
zh60xI=$z@PXg2a$yPr}2c|PjyF`OR3y8}{_Z4d~lkczDN<7p=1`G)y(O;{CIFlq0I
z0Xz%ovYQ-)B~h9OV7oRnwa{uHR?Y@I5<L%4R-uXb{+Hw<{(b%r(i5M%l%7H_-V=KF
zwMHSwPaB8LR=V>yj?n_14-+rGp2LwT6fIrlhpk2ZygtLfcIuR8V(JX`5sb?-KLpv9
zYgDZ=e#8i&Y9K!hiTcG*T6h)tIyqJ2o<2ePhBGXK`*{*d0ea|Xh?Cb*GioHmI{`hh
z6B|BLYep4}6qm?rZPHD73AYRYew2&Q6XAk?16pj@G>+8=RTzLKP)C(w+1wTDoBrpG
zLG#XWLBKQ;fX2aa7f(Eugw|HOSbLm=C*ECN`|$pIx)NUK^A{un@RcCy)PY?<I0C9C
zbMuL$oq~Vg;_tW`2M9S%o+?${$&p)?7hJcxwv1WRI*}o3_<7#`x%QQVne>k0gAjtx
z`8RkGw2H90&GM+*Dx(eKN(#smqqrVuyqH;@mpG@D^`R;?4^MRao;;c@gxbn_bNL2i
z1cdkP<`um3%I$lI0>DOnoJY<j&Wh6l23BF*-#%b2S#_uP>8<jO;s}Xa5TOO&5lE4D
zkQnRrm)}WigTR44gDxfE-Q{81IN)jGd)d|55kmg5^v~6W0hrN&8mJCnLQ7G%!v@)o
z!6%h1=v0k=jvFc`a{#oUurGt<EL?yMWD_n*+jZ7#-Mvs?O*<PXCHUjV3I^PGidi>c
z<^wfAGAT_Ply>DWmX-$GSGl4hgmR4CULH%OzRMXWrX^S{A9PgepQEk_2~TIv!BF;c
z+@xr(a&fyBB6%YBAN3ht+{ZK2&*`V5!|T+z|MfeQ$v7NN`9w0Fcq~s+G-dA1S*WiI
zidyQ#Y5f{Hv>OkxybsMGXB}!P040EG1Ns86<c4QzV@E|utqr_MlIg(8mq1^fo5?qa
z;15MbH|s{Sd8MKNY63GfY(;_I5hw!M^)nIsxub~fN7#w>neZmIuX`13`Da3L|1yGt
zm^tv7!M0){1Q&XaClW%5Vu56FVBh9t5_gafr5AK8_!X~?&&zL+giz57>tDxP%#Q!Q
zCC1xi^r|O3wem?~UMNa<`VvB^d)qrgSaI2ma*1lcco3aXhn&N$gR~Q?-{bwI)w^E>
zOmrXbl?+&#Z9`|b%b_ug5WM^Dxz*YkmD@W^l<(jfqx}c-q<Wz|MkV0`)VYBz-6cfZ
z1#=0#@dA$_SS(?4Dm*gaVTafWD88AfVTQP)7t|oT&@f)gzI)EiVjrj{c9}yD7Hk7Z
z2GE}5tGxWwBY7wP2s_Q=lkH49Tj%oV@fh?NLgGEA#(z#E$fG>jjz&C&pb92&_%_nW
zSX{p>8_cl9`$GzeIS^5aJc)HD(nRFr{JloIz$^qnDAXHBv;xfo%q>EqjH)lV4y)d*
zEl}_dgG&%gPPsMwk0oFunhK<Jo57-o9W}icSyj}(v$OX4u*u|O9ci>z+Nw5>{!!Er
zgAk3iezV=6Tl@^eokz5wVRY@P9V;)KoG3R2(hP{=Ky_Lg`b^wCBmFfb)4lAO%m~)N
zvkG?WryPwsR5}V67jaH<It#XdA;IV%odEx()22W+_uCqQ(UrRDM@gvju!a&S^1w|1
zZol+~5R0#@n~L*88{jEux&SWLm`d@TZKQwBt0<Sxj@g`;GRRbc4?WWwX2eGFO{pt=
zU_R#{&kbSmMZG^+2JO`B8_5PgAujwcce%ma!CMAhFHfC~)j|Zf`pFg3WwQwkYrmeC
zS$p+KRkZ7L<Lw^-3dg*SbuF&!qBM*?e^uZbRjPXlF`EOqc!aYv;4|prS~h9z(l+=&
zPF=BMn^IY63$RJN6hDK~1Q8c1cPn!vV}$?DPuD?Cg+DMu!w(bQN>2BGE4c6K`@^{C
zD)wp=aXZvwFJ1tpAk}c#E2qeiO4T1>P};7RM$mCzRJlGdS{2@WV@6skH5JCeuXSK6
z3ru4<4xSiR%s6|uiJ(2VxCaj-jDg9+$NU_}(#Xmejs}H6v?Y=XxEp-@^IyQ}-3n69
z(6rFjzX8TDTRBq817aW|ufMf;T_WmW;yi<6T(J|lO{+Ki{m8V<0K{I<9*SLV2NGNW
z2^5J*Nl&_6L^03`I`s8E9U5=GRWGIFb!{<00Z+K+F%Vt%U%~Xs{}A%zTe1Of=o%n^
zhw5(g1S`}A*^1hK<Jx9F=mF^`M>3bg)c3{2rGb!H(0(Q#HIRkU$N8hd-fZ9wwj0lq
zutXW~rYGq{mT#SV9JSjOGy6$M0W_2fr43VTg~!o{zCne8i#RnkWsoBZsh2Pz10+hc
zt8*i9jDqYRpD5mdmC_iMi#Q-cXfavj-GlOih)@iCd1g=F_h}6qRN*5PCt_n-9i;yF
zKVWnHRUq`n%2ntgQZhr<`$G`F?7*VLr}rb!`mibSMzYuaYDgD>WNZK|yF9_m0nil4
z7?1;0j}9>xS}APv--yPtZJZVgxKRfR?`AWCySw~<C({FuzxhSqPD>i8`BCg*2IWLY
z1~uSM*7|qq&AZp}|H)Y2_R~7&{3{i4?L6f>cI*d}F%&2qIKlNmsZG9~kQp~RKkeA}
zu=lV#*<+K$-kmb8N#Lt0Uls`?UlsM@_Z|yz?E$T)X3`IQSW`r=2OE1YNUR#|Hx|Yt
zdDD;<LXSSQ@6$R{&slG6M_0<|2U*QjiAX_gxM-gSgjJ$aDrV3MkRIkCeukVb3`<|2
zC~j25zuCd}&zg`j(rNnKQU3uU$Y&hR6S!z#|AUWnaSn&CMVFs+P^)ZxQ!6~kFo6K_
z0w35c32)f&jgP~#Pix#vm+b7!LKAy0OOuu^KTb>BT{ER9iPG2c*Q&hXI(!=@FeJs=
zA|`PfF}$(fqd{zo?yAq?T=S`(n#<9IeXn8(?+<G4pm0KV)vy(y^O5f&^tJW2>lom+
zPpR8)Le@=0o!ooktyV9eqAbbwRSGBF1C#V507Z-2@V8Ax?KTADOgJ@^(mLuGf0ST#
zBhVwWQGg8VL7kUQZvx16f^_By4BTuX{0>Fj5Cs8=EF0oI2->+Ndq|Mq?#5gk1552`
zhz({&(g$IffN~9A;jj=oiXoGGS<(hZjts=&9VAmK^te!Zrthf0eA0C#K@}AWaOlA4
zQ#_K=jsMhqFyx;;BWe!X&HVAV3#&p-CLF$+8TfKluYGtj14iS^Uy&2cKcFQHseOXg
zSN?r4meBJA=6Uegf=Ujy=)#2uG#E1W4%RkY{c~zyeGfzlPZ9zK<~#cn`blZ2EEBjd
zCz2Njbi(^PAV)X|NO!TRD+j5QoU!gRHtkXKzw3khsz$cmx#u<=iUrYMTAoivxt09#
zOqNQ@bl#rxZyO>h0UCyugKKPtpL=i1rOu^z^OW-35fDwGE`#q&1N-hYBmy_mN^2EX
zj&l!wWF}w5N<=f5#5GF*JnI0tf*o$&6G>@a=Yr8~(lLU<NhOel!|X+7?S4{WW_SGT
zcGjZ=X8EMuNQxi`5<r_R6ZK|+oxhv^l*fouJ)|NDr9v+X#SCSGoF0(N1*2_?5hgow
z==SJIV=HRW;V=PU51|u`dE5(G7*|UH0fVofAu5T!YQYAeCS8)}Vw~zfS3VuMUrl~q
zoBe5!Py=Ao!PrA}-H@oOd?aBGQ~wBgs9^~oJ?8k3U&<YTEm^LCkfyQ{@yA{F@_GQF
z6z*lPEcKh*?%Go~Ki(QrUi&E-P&{+38$!ij9=1oD77|Kn|4}?GM-fVM93SClWxX}%
zoQrqDL{eYP;Jsf_nJYxyB@I|D5XAsP$X2XD#PxtzW1Y+TX)`>H4=C)fj=>Q+5d|w}
zZani%k@xRd+rCa+nMT?^SP1jNL>%aX4&qIC`)KamL%;UJSq&E$V_esdjz1;1i|h5%
z>`h=ohh1nk`>}aPDzcyrl-8Du2$G91+!B_aTK_T_G}ZD+4j>5A*~`tKrPC)m-L(;8
zWRzP>%!JJ~?P*zpnsI<R>@uJX7y4RJ9AVKA>-Memtni0W5L!F!50fv5q?(9Jbi9|Q
z@AX3FFuMf}mAmP{W7+Td4xJ@vxlc(<<M1&F&m4>5@C5Y_GAu)OOAtP7i6<67D8N{U
z7&7pJ;NI_rm};Wp&w+qtzRxL=3;J1w_ZILqWs+zwVV;8j3IsZ<x_BJh;_Wk$#ASK?
zj85{mLd~Ug-6fiCfR3B(Hm)lwsSr&uxnP(!lVfLpRTF+Lg(Dx_-p$oi`JBgiowbwq
z8s&)*M5-G|6<Yb!>jmhmBM#97jsi6ej}=T8Fk{4k38u-ewB8>8_gyaagV3PBOo<#?
z20PY+(>0~adXKZfb%mh(P+vHKVS0~gCUB8*J4AWvOhvce0r)Q6yDZwV(B#?hm=lGp
zv4p2f3#zA)mZ)v?*&B^TmS28Wm_}I1HH&r}hp71tu;b=~<jdl#-`vmMK&bQO&v~9b
zhfnUMG&O+aO_l91)0PfNYe+X=e|zF-7N*Nt2E?Kry877$);%R(jqFP4DER7rD)kt)
z`;PX~ToDAd<|eqii~c8#hu$FFTuCebj-Pt=C(YY=h9PpB!)_j>&F{b3tKk4ySxwdC
z^##fKmQ5`P5;gHqZu_Mp1Hy!B)71^-%w_60?tcu;o*#E_-zrpzd#0#G4;D8{q6vJ|
z&eAV_oC#6z#7_hJ1Vka0Oo;t6n7;Y2TXLBm`C74)eW)QU?DuF#6E~#5mEQ9^jE#+7
zm6yMrpO+b%FUN-n(OwzlzL~9zp=o|CtPGf6D5|d4K0Tw~8VfSGr9@an>~tLW&OYul
z@MB1V7JZSrt#ZojQO1qS2oEG9EK+&99p({G=68Wg`n4O+p529FdPWYmQretA)*g(>
zxOV!giSqC9HF}N(kq1Ic5yl3_BN&|XJSMvOmDYbO*SrP8-?kUaC*D+zy2fP)59?ph
z5rbeqq`-=yPF=Hv2qe0=k$kHaxVE~f%TiP$|E_geMz#~`i%ue_E0>)p7*n)Um9RRE
z`;NdSg&+q2yU`b6Jgmtgu2(NBg?whIHvNc59My!OBX+*)u>-_yaFD=pEmQ;8!e!nQ
z+BZ-gXQz+Hwp(HbtjE;6H8{fTP2myIyO<+h)IDLB=^`j7rY0b(%?TvqtReu4@PZA3
z@$>fTl8j}I|06KGe1qf_Sk_Ax7h==A*jg9Gm0^eVgGC5KCPI?qf#hK?z(*ka(a)&<
z)=%i#L}S9@f6qZCDWpBZqL7U>BN<||74-CA<dHRH_4^<+%31I~LegBW0gL(Uqiv7X
z@g!+`Qy@!`#ZDt4OH^Y_7kvP-b#f}c2oygI(XKwx#9PsJ+IOZu_^o`iS{oB-C9tlC
zrY0upOm@dqvTp1RS^PPm(r5>xbmf%HlSJQ5=|2NG)s^>LPhctS;0u3wPjRylt?u0`
zmBG0n36yxj61Shp0=0h~e!f=z4<K&nU-qKMmN@&9JWtZKKz@8~JYe=oh4ko+h3gQ(
z2A{T3YW2xElg~7&o#S(f`obfp4tZeoU~mI8TXlQcW?0P*tJYC^anq<Xt*V(<Iyk+M
zEI2qMkZaqNa0-jNT6_`yeSqWMs5FvGmOc^>`fPgkk`GcN9icHp5Gb82hz&v&W<z`<
ze9OM8Lv?%sV1;>uEWv;y5wdxofV2iyiX7h$+i`8>lSDVuUDxyqA^8+~f9@D+h>a<I
zNpnt(Ow%A^Bbj&J20Fxzq~&*RiOc?Kn25dKdP-c81P8e35WeD!Ty+Cc=qz)t*&C*u
zg80g%Mk|}+DN`>$>PHbWnq#YWM*iIf*<_$&c>_X4a&5e4Y>3J8zrQGqHa7gaFT{OW
zVcYS=pZdDE<Zo|;EiR7}gj3U39-EjY12Vf+SA{8j(+;pDE0?uwffsS~u$6|_+9gEo
z4I3WVxf^E{_lMjC84R2`AV-8XQ^@*3c)s2XaOSYGD}9rvAMYFbkXCQ?a$Co#HTqnS
z8ht@$H=uS}%kf--?43drk-Ld(WV4NkLFni~?1<b9aIqtid=?hSv)<L!1^T`_)633E
zUMiWts)?Ing(ytsHa)Tcv!NQHtAcYBRu92vBhL*HnnQFAax?yJ!2)`=Poo{S=s=Q_
zJFJ`fTd0*I=)!z(FQAZ77l0;{&Z@6bKZ;At8ylNC0XGrd1NQ@dUnHp%M-hW(qM8j>
z(mprpa4~MRx&cqkxu=Egx<|(GeLp4FeqUJ`Z5o)$tMqGZFe8#rBvgM$Dm`iVIR+`-
zJfP>fsx*2Uy?ETY-rL9SZG_d_4m9^|rk5QALYB!*HRA>_a!3)bB;Uo)*e2r%1jeN5
zFRx;W+7Jq{S7+|VT>xbBqi>|G_&<j|ov|Q^3)j$xX+EMGvW7s40J^vyS-6yh1FS3<
z0%WQ}Zvj`P_S*%_ReUQ-e<*tO0$`7FyMfXNNhyT%K<f<2avz2Yu}Voz4#XYI-6w$k
zq9>DrvEvM;0SS}<g%sQw_zz&{GVijuc*{`AJAM)}oFD;@ZX5Be#ncQK2c=70I1us;
z7ypUA56RF88JIR2I4^F2ZTD@SExoNh4hr|NuVxNenkO;buci}XrR_&|f-4qzr?igd
z9P<Ly>)Ff8uqW7y7rrC%EZn@AxpW!6b>QJtVj>4Pi1x0plwP}q>+GK^VQC|z<V=r#
zp$uv3HUrj~&^mjJo&&0DNaw*id+esuQ<B9)Q^oV2C6g2_AU8;{tMgIK%4o~1m-{PW
z5tNF9A@HFl?k^s3hD)Fh@^$X9S)?i-fggM#!d3$Z61zid^S9QS*J-o>5+or9dAy_$
zuLxkv@^^wvsQj)|{Bdmpe&#Me-i)U%r|qH|vJV5~lfStYZZCv8#RK31BM7WogY<cb
z*%h7lP~^Rl0^!$MZRQ8^#A2`rxejt~<wYtdlCV@@Rr8^vBPO0)49}xyUywKJSli>h
zblM;@DMr>OIZJt#cEd_}_8+8%ZRW9I;5AsLhdo9T@|sNgJ>enM_ylB($Swl)rf2Q2
z+mgOJ%3TuWlYAc_d-wpW0R9~$^Agw!dR8uUB_S^9?-VTuBmHM{lZeIBb5wEfspRLg
zpL?SIJw|Jx7oWWLWrAo&);p^W4c|k?{Gah&N|-x4FHPenaS$K*ChwnO$5M0&ifr#A
zVz*vpv1WMvzADw_n<}m|(lrcuEJ}_{lcesf|7cl1zp}4b%)Ek@mw-Pg$ov-8B{+4d
z-RU3*&Fgem-h9Z*NuCDv3Y|3oLJ6K}7}r$MlP_7xu!{}GF;zB1$p~vPfQ5!cBLJ#O
zT!pwf&d8n*wp?s>=V64&FU5C;S$TQl<VyeaDlh#yLsSZkW01Jp=X;}GL*OM-Bc!M#
z+@qfKhfqJRRXU+Tz6v_-jVIXs6VMsKgU~oHO#rS1ua8TQQ43+MLRuZLul9)$=;q-j
z!(W-CbN&J4>1UDHZ~UM=hTsM(Kwgo+w2cg&CI|h#3sZ8<s(o7iS}UKg`QIDZJzXL?
z1Z!6WSE)lW8{V8-hVIoH`XAsWh;&}UF#zxkE+PPj_s7DSD15ENN38f6&Npjnsu(XB
z?|3p#=#r{PAX$TgPj|5?CTh_an_$|Kbbr>Y&aUrU(+yY$5v{)aU5}Lq55@KKpJVk&
z49xEPL#27?sx6RxjmgChY<|Y&Lgup@1mY!U!4QWl5x?TZ$9|^syzW8#bRIl4<KtI+
zmqxSxH0b&VrmRqf|4bCMT!(UfoFx{H(ra(gc63b~+$$^=*<~_Ydz@w7t99~b?Ap11
zrny8Z?~#yx$9)W&K_P*?kWcSBph7>-wYrtntqj8VatXM`fQE0BD3~_k5A@JS_U%Ec
zQ~3)!>5;U^VF6R<Moa>7SZt&mfvE_d{r8tnjQ`*jT3?u6<6P^w6;SwV$tRk?nLAZc
zdNpb3`@5pqBM_OV^RMkCeCKunM631XSGeU<xnF1Z!L<%M60cz(tE@o@eJNFc@&e93
z-Y~i!V`jNqgAvjJ9^F^7=Ui}i-+Vs$E(VY_u?U*WkM-5HQdqZ<X1|Ptw}k!_YAYB$
zR177xYaGSB_q0UY!H8(EYtVNB^F1g4>8^=5NYgl%2JZ<H5;C)KrqHlqam_(In0kIh
zyNGC2FJvpi!v*~{(gx-i6d?JTFs}flZ(`(MAqX`e$OK@{0Df?h)ep*rP(-y+1PO-8
z1Wte=*ej<y?8-*?ryzIl5fGK!yHjR1Zq8>MMFF0b%idJdmFisuHGRPkB%c5-0SW?G
z6ZEg}XBnSX2o2}o<MpaDmN){Qd8B%PTYi4L3tIj&plt&53IXm8cdg3ar-;OVi(H#A
zKKQ(%bfsylt}Z}G-@GfF)Gq$-H72rO)&4VGuVxAFEKdY)C#-Mv+OnS$LXcX14U%}*
z9$0_)`ElmNc2Z<lWCp6@wm7Qs()YE~RP%o1?YvVk)&DKPb8P{#Vy8Z}AvRe!H~jy6
zV}!R3z**2J!#Jp>M@(q`Wa|K<oJSM9-4I0gG5}b6)E7GyLB`|MRj!aHc%o8^8P-E6
zI4{z*$64OMuly4{>>KlKD#FLFKHWcpPY4ByWYe>jTXB;xi-496f%2dQgw>?elOkca
zZQ1`8jVQud2K%xGV+XOE4(CV)KzbMY?~A$;4i<vG4vf;UIiS3Rk7fIq0~IzlKszN8
zwuxs=y8-x8RAslz_pveTe2+%TqmL8y|A(ya0LOZN<F+DOh>Wc49g)3dZ^@33jI5$$
zk5Fbj$fm5aN0dTQ*(1u#3dzdKUhn;ToX-Ee?|WU(<y`;szs@=OJ>TE`z3<Puj~2%^
zJm(}eslSXYF0%y)-}6`xA+D*~VyJ+IVwk4A&8hTV?Cm(<{sHi9Z^M)G@_?)s!^|o7
zz!G*YS#s0`snhKowS<Fg$IpZ3ndXE&$QI()AuX;u+zu;y$NCFJBI&*(Jk8c1iVI=x
zl=dR7pizxt6@~EyMpMM)3GE%M4Tajf<{*=0KlU|2Wu^+fr{!7xae|>82De>8E@XA`
zuoyfpMrc^>L<qhXOwoitQZ^PWfCJ-IGZ~afsK#J{M1UDcu|ag#u%QKH)BMTSqc!U1
zyDtC~l2?T%kI^g$Z3>JX=<R?YTlBOLc|BLh&)cB@%e}z1OW*Q8e_%R23kEg8sl){{
z?nETxrEKtOwj8_+6~>*#(H?g^b9CQ3iz+_<qU+EO=}>$^9aA~m9UV^b%Ybyke;?U}
z0jC*`%h?~3V%>V*&HW<|-frD<k`9>gSf|Q!?#$o(w2*apnT2mtFkrszE!Te*j=am-
z+g_i~aD`dCmXn{0E3#1PlYT@=4jWfdV^=vZ*EHkNeB-5tgLy_bw_W`EUb~lyZq#pv
zz&fMz-#NHt?K%*_l$i(p<?DXl+ghOIXJb)cctH+N3gAjupXvAjG(jkLjYj|+fyE0c
z8}Ji8Sv^33otkZy%hIldd(Wr{Sq%jfG^|^Y$5m);0cs5JB`6w5_Y6Zhk_`pT>Vx{y
zzBu2B9_KiB>In1OCSb!<g3ZRTVA!A^R~}=u2HjMHG!`0mkmdkz50Je~mhc<I^@*J9
zR<#?c?C|EiXb>7Tfp`9mxe-^p$sb15i4ybu(VA!|ZRd!9p_a@bfwB$W^v~04A1o#!
ziTK-kg=(@Vbp7GCK&-<+cIdw-v?A`tddhJ9o+Ufiik54KptSep&E-kI<d?tH#l;g5
zb(eN+=H2Xc-*Y0i_0mF9Zm=70n9z?!ZMYZP*;CIXlf%CdrHnMpjz;Gr;&Y2r)*ao~
ztU3(zfnKOIn9-O&r)_xbx9tP^5h>wD4}B_Cyn<;l^?a1}s2~w6a)9k872yLk4<5J5
zL_AZ~+9==%6;u}Yfw=YH?1)4dE#RwwxdOUkXyGD11%vfmK(6m-^M!&ebQH){h-Abg
zl~0Z@$!=Cu*~GQhx5g~#6adsM7a`W?MkOmJDF5IEKS7wM-0sI=KmDh%P*5FFJuXDn
zg)suQ3Ev}p0Nxz#2IFHc>DAKxS^FGaILrD|xbXTxF$a6BV}A1otMY9xf1;GlHY7td
z3*OHJ9hnaWS^kf8WDhBkZ~Wam(3`^7>L~prt<44aH}Feu4bSDWeUPa&rSf@j&C}&u
zRZ`ury|q{C90<Mb*V<<Gt?Ejf_kM1tI;@i(bMm-__%LSDDJ@q?9-AjVJNQ5>VsvUq
zyV`DV>dhUg*Ln1-yQfho#tm|Lac`!DC*wt|b_@~l2dkMj=C#8Bpy|VL<*oy=`q-Ub
zK#NNjl8<j)LKPnz9r|F1?QqS;oUflY`Y0^zZUEkt*7<cI_Qxzxu0a9-=0+<>ux^1(
z<mhe4ZFMQ3kU+5%|3U%6qu2H}o#5^P5DwJLuxkYObBueoh7P-<Xxf5sqb?`F%TV}%
zGKS9yN8)*b@ZVOBKvnj|wAf3XQb3Yn8U>|}E=fwxFI_aVt8CGu@zMkKa^IKweCtxv
zeJPp3qmRHgplQ0VTC+%V&XMICeA#sz0cDUhE|`wseqK}y5FM6g)X*D`{C!W!WBv2?
z5o6&0gL|F?p)W31+#@m29#DJSe&@aMSu4rDq};+zvQcUU$sZ0ga+*Wy8hFnl?c}`^
zb31>DAk_gR6-3jV+U66;PFl;5nLi%+4)9I_x-fmOasNCtq(E&tq?%mbocTg=ad!LD
z4NTBTH1Hs7b=v+S25cK54g*mH$V1?=dg(w1@1}t^zDW@4f@=Hg+oOQu0CNk+*r5NR
zrG*7>kWC=Ci2RgLC$FfV1XmSAxzd)y4x)cIy~DcCY2qt?&#*S`JfkN)9vonRMM1lm
z2k$lItm2|91crdxG_}%vPJ^)xXc6{=HvK^5q-Ek}vj5!!$f+GHyFPtzKo1F(V55Mz
zVFzh~_W8Ayks`L)e+KEdB|Y!2hzGqWSVcC@oooXA-xz(DU{vtT4JWDDFm`O`EW!%5
z9yi(xp-!uifPfga*U06EKh!e|K$3++yC%9aT!^`#y}NvpbIdAv3KYs9#rUh!Hhkt$
zVGNs2y9)D^dL^?UN^~JN38m%r=Vokf2CjS{*ayLge6#9BzQeovdg+nsYYH_VK<-W_
zFe5n+Jt&+KBQ7cE*2H9G0g5oVQK|Xm0rv`znqxBNpnqdo{76vv%43MJha2?MjSqCl
zu!%0O^djURq4$OJ#07mkq}rbeo03;=#{jtybjqoG4x(wyF*Z;0@e#oc)Yq6gd6Ods
z<ukx_(YHv}Zzao6y~&Y^L7E-_wjq6Z&k1Q3{rsYl0}D?W>KWV$G1E)Md^6Lm38?M-
zT07X7=p(+1y_SmAjhV{<EwSq-KepMQCD+nC)~A4S5*?Nd7vLad^wm~8=-mxhR&Z2S
z1!)yMb_7h(jBasrG%!R*yL3jv8c)#DJqbqQ!oSA7h>ZELxO@El=y-LoSyX1D2-@LV
zw5-rj#RZ-;w5kA3La4nwt4N$rE`=%1B@D_hBp5VgXY>#uDD1*rnLvaGHMfnEKA{X0
zF3nVF;8H;I1pOWSsgm;v>+9-F)u)FynamlWb`!sYc12{<iM|B^{8>k8qS4W0QRvo6
z+r{j!Y+sxh+GVu-=BjY|RnWyaiX=W;g4Ld$8_ZC$;aP$Xkqigk3IHWAeqT?3YEd==
z#J65wWiH=hq~(_Aa(v1@VAziKP@~{tU4wAsZ`ba?V+PO)dguE{KqvrIq>hq|!pB=%
zE32OSW}C-;1swJj$p#cCcepxA>7HB`7ayyv75HCZj|^MYi3JWal`!=0f8Zi)2YX9^
zs=mL_N3W}?>y)JWRmD0oSZuT3c-;buJLEa>en2*C2i|jfU(mBMhMw%iz}Zz052F|X
zK9!TpHAuCzFXx}e2=H*qb(YWiciVwM&8C`UHt~k5{RN8+tMx}u3v~YJGhD@Bra<r_
z19fe1iCp0~MX;URZ!p;+v&Y8IX!W0O(Cxl?D!-xrjLvL5Ygzv#El8eqF+L*?f%GD$
ze~FPGJF8C?rr-GsQpDLbSMq{@q-TQ+)>$PGHj|+R6<kwWy)!0fioqiE;u8!lG7shp
z8tdzo(Wxx3=%*QO0I0)gFO>?V^2=Dcy~hcMY>R)gKn?`IFw<!mk9;Ruz|}ea&O2mr
zaWT)N48SVD1rNW4zs`tF3>a+C(SxPS-VW<#2jt4T82|q=W=PLQq&W!!;wEL_u7FrS
z9RqFL;RVc^2N8VAv($2GyWW`+WC?gy`i4{$u+$TZvaQoh)2LXm!RUYXvYN3z(a(=h
zzSIHC6z5}PWj8%Zl?ywjVf;fo6R?!L?7s(FJ0X+4rS&K-2&bd2k>10C8t^+xYBll-
zUc!6_e*BOE78CMH0$ARO);IwEb(Op49}W{e@NKfmo6H%?tUywNL+ZlKh#M7!`TmJ%
zOtxllZ#D};UkKYkllxGAOG=!Ld}%B=-~EQsl3Fdb8gg2u!LQ-L=5vuK=TT8Wp%yg$
zAdXiM&%yP8B>S41B|U^82@>C$l<%5|iwi%K+Nz-0h1Ul1JkV}IM7RbycK}-<^~K#i
zCkX=7)bp_VFI`Ouua1t*xPGs}%*NIgY)+B-qU^WDO<Ij~^U-^H*f|Q!bR^(}M}dTn
zA*Fnpx_-`&)=Bo-Zccn*v4C(f6-xa60TxnIaEuo4rYHj+g;Uz$#ZJHTW`b4mYSBBY
z&rlQN^&P-sso=~4@(CZ{puFCeNJwI~qfAa_X2h%Wh2^<gs!%_7rF^~&3X`<O9oqZ$
z!>hoL24ZfX6=!2TG=_a^K<5F0sg$WBvx|cuMUZy>HSj4bK?e8w7|#nO^x7+6p<H;$
z)l6_%VZ#bKYRRi68p&`%KW2S}TuU`w5H{ZuO#|@3(Mz~z1O8L;+aF*N;Ejq1z>Pl6
zz`pm&S2#={WQ3l?IO!okg!+kg>`E52mSEBr<yJGLV%@=d&BzGwK3vD@3NDvu;r?iW
zXAA`oK4{q@HxLF6sV^BSWskxkPYG1=ejkKuhXiQyIaV4PVp)e}M3y`;{_iP!YyN*6
z_V+vvS(69Fe1O_{RN@TARE&^bWv&6O1}W!FUmSHXv!1fU-o`^=W(V%yPLev8@cDqA
zG)bCN0sf?tGh%(H0LgdFqzRba=}e3HGchM(R&-1Oz%vLK11mc7A2Xr@UDS=(hndKo
zp_+pGr!RB-18+w6x^KKp-&$G0)5qz&2EyunYC@Q4KV-#4Sn-WzdZ5*Se+9xSqah|h
zmR6zFgTV|~k<D>&JMe7SzxhstjEau2$i}d9XmlI;;_H5~JvXa}Own;~$d{6<lA{Yz
zB4`X0kq5IO3P_ZCp7|yZX!|bFr5&KsPzgDw;UC=`(_yE8hVg0suBuW}!?{EDTi!$%
zpmoB?Hst;XQd*b;QQHaDl<&ftoR+7o%f3I?c)TUsFljOhb^T&pIk>hgL|9xsJnH@<
zj=VDGH9nqx0+xPnYttDZo36;_?cmlEPs!(X=%7Rm2F9=Vw_;K{#GZ8vcyDr_40Y{M
zyN94YlIIEY3opzs_rd=j65w$HaR?GA=5dBwM?en(^8_r5fi~jgo&yx&%dL+N)52FI
zo_KOWWw#McgUb_ofy5fMEk&X8RIjNM+1m2J(ZF^Z3l1S$l8WBPZAx*P89$c(3NV4e
zOose<fo4KrCZwQ3CL&OfB9rjWh^=nopdn4rFasHa%L%)X;5!FkZtqYANCa7@za@wM
zY~L|^715(YXkdVYJ%jT>|CXHPiQ~*)5n9tIb?}#NNC9{P+zLQU-|%4j05aM_Y7MUo
zMpCHle8LyIUl3`K$__XnMB9Q3&oChRbUWaz5OHxyi%mVjd1sb1;GwLKs4@t;>Yf#n
zyOPMAtb!i9+P15$(_NNMBq+VBd!jU$ev*)&8n}$rfk31XI&k6gpN_1f5599<GB=>q
zXubPUux5SHy>Z#OIOb0#i*?uZ{d6HXNHi3TZI|28sf4w#O%B4Zpq+r)2*ejUl@QUV
z0@?Rsep*#D!k0IA5#)pCdtlvC9*@D#j!YOGMr)vy!^wsZ;FT~FgM8+P*k?e~_T%Gh
zRFb2?uMZ7hh-G{)!X_cBQ_^7LCD4ISPai7)m<p|kQ4v%Y5Y0nFf@CF;gkuE4mw(bd
zQr<f$LZ_MEsMwm}<AZtwK7h7BAXJK|PH-jY>FPe1DR8k`!2Jx}2_kNKAnh3f`)NLc
zaMABm-yG*521GzUYK5-_xuu`)QNphZAE-L;6YkR{L38p{&$X1L*a@w@o#zt=Z4?1x
z&q(<c<;53~Yy%(z47g_B&&=FiRKxpM!3m%UscqH*S37KPfPhxV6g^!_M)h`{`^US~
zh!#u`$<4vSFl{_|$YAa;>HoHCRjkj)uJ|%17r-LJ7h6{I*0+AA6{p4d{80kS1Y(p!
z2l%RAgC~JX6H<_Pwn;JxkOSO$=#&%nwutGc9@oQuVI$r`KtJFEYXphO;ZF>aS;cAU
zstd6HI-lbkUcZuh1He%5UZ2!f--&C%uzBY|%7?k~HeGbso(%G+qp!fl4g3`3Qz4cX
z?Sgmm=shM-`c+*nY8kZEoM@8tUoa6VqCN6O`lyYe0w$sI#kz%iPaZ+PiW?FXpMVz<
zlzPS;7I&jpc^CTozOkkWTqr~N9|Lg*MACIFP&kb0#lL`M0tpGEp~qX78x>q-OxCFL
zIb1Vs>99X{e16*p>iqw*T&{rqf!hmVxn}sKy-WH$#k~`=iv@JVb8NNXSbqRRbOc4C
z3h1t}Udz5Lz<6vNv!%^|Dzdd3e+Ke-TRi%ZXG914zqij_5#G6Z5lT9_v>KS3p`puv
zlEAeFA!p>DTaUO9A>(2tBWv_EYLr?{t6>zHB7BbC9|1&nk!`{UD>=2ebR2N|Zb+g1
z^nk=6x>kTt5mpN@3CD$A@KId~Z0)H^HUM39Sk7KOLiYLfUs?l5;9zGu_7`je#zy_U
zR#nI2XgZaO6HPj2e%h;@PNK`+2>o8n$JL8iif^+~fJ_Oh7!kz!E91;UNpjImN_$Ej
zc?zL2B$4z1WPKp$(M}~tOfGvfK;lD1>spfH<n^QFvNev&0crP99_uyFerY^l*g(1x
z2+Utk<yo<zn^_#4&4e5mE)GVFHwg)YrBg5d*jngX&IWQ*BV_{<uUQnzHtpY)^5-x#
zi)935*}3odAaEY7D@c$TMPw1LjAGU^tz7meZMg>V{k3+Hxpr{+Y~AY<EF@AfXDT-i
zpOWc)ujB<M4Qx=!SMxnSk{)jiAO%jzFfue93k)zP3oSuG_>Tqzx)Vgt2E8D>%CAc!
z6>`e|j8D!DJN~(Ch~{>YT)O!9VN%u#i}*xXT7q65ihiVTt|X(Bb*a?A$uv}RSa^I%
z2YwlU_40_$KryJiEFP5+Z7c#h><KDWE6RE}pb$TfQf;qn^}_zWe*+-DWv1bEdx*>G
z3E0ouYTY`)Mw#2%BziV`3-TC#3IKTb|82`eP&?}ufxX_XdAz&wCP`QE+v_M^9`t}U
z08dkIoJ_sSX&uJ&pNKVpb(V{P5mlGYi&sVez2%G_a4ItxVlibXNX*vNA16D5Vmo{V
z*O$XO#9*pt!^;UZKu#hI49kc%R>QtZDptYUD<T^ta4;8W_J0I2r2hhVPmrCzKkO*P
z&%@J675%fmP7VihXC^@hX)sQr`rv1%b!$%<@srRzXsY95;7_d94r-Yb;0=TY^_H;{
zvwsh#$D__6*7-)*cuJ(@fqf~ihc0VVbkVOK^kp(rP?_PHfDSJS%{4@UuoV7>i#e5v
ziNaaiw~x~4!<eDoRB@>X1hGnV=)@E2y?cNBTOeR_D!^<h99Zps7zFjUVAa~}%vh2y
zFEkaKx!9mkDr8<<eu7`X%z@p()eZ|56DmE{UB8Hy$asp4G|=drPp>OJFnlI7XAFo!
z<=Gb0^zIfqB1W1;;-Pf}@HXHOK!A&gWF=hbH_5rNxg3#Ci#cwBY7W*vvI-?Q$K4i?
zMbm5XsSot^)j<Q8aS%!`jFts}J?$_5_orV6QE8mC-$4G&`1!~9rC%LYWf0Z^sjUUE
z_7+H$fk8swk5D6QWahed|31BS6=(^AOaAacX9raiK)gL_rBdm98C-l?n1o6f!Gr;O
z1NNd?rY{0&bV1D<Efy)Hc?x~~W%0=D;z4<maM``CtjPd|@7C1wZ6@z1NJOM*;wrw(
z9_@CuAMTj$CM4W5*pr!=DOoUiKJ(5m;UDplxY835#MuM_B+{KKZEBP(J3GfI2scl0
zKFv1Ub)S1r(tDMi^!w@U(SsS8LFj6hlBQera2wwFgf}wOtfgG<0*%*aK_YnjVA6v(
zv3m6O2@3?M95g4;O9JKexZOuVBAks&G5=SB_DwbYn;afntavi?EkKj0e<VV3QC>K3
zaB(Ntuy`K;FAoD0z~YJ72go)DP$9t#;HayaPr>%Kx#l(^9H%8t%si3Hf>8GBy$II#
zNU=db(eN1Sb0@7X5}N)hvBQpSlszFLmIlmXI5vZxiRHYo@Ia3ez$2KO*(4y0yZ@4~
z@2MH_BgV#kot<kFho?ko>z}6|=-s#wcK1;p#5Y-bdClJB29pmQ!Eew={Uh0gCf00_
zDcGZLLld`${g}~Vk<ZQCJJ<(yQ{y7_@bzZL_0|uCjJf@1Bc)sdcE)g&k$HKP9J|7u
z{yIC9Mm_y;%#(e+MnzMAEC9=KA^=e@PDhr1Z?MKW94g6C<!Dk3GqyY>@gM^E4e0)1
zrrK|ymEuZNhm~CFTLrAD>)O!dJEX!5?by2r|0V>TMc^3Z33ABp;`^^BWKBXt3rPp^
z7sstCeEy`Ifqv$MNQTB0*a@fydqjP%UQ_n}@^*@!Hl#%MucOVZ5C{zGX(2Yo<g&mW
z+k9yYDNxyUJ?0F$0Fd_P<%H&$<AdL_#f|Z6O92KPq#l)&HE*&t5$1xE!<G2o_f^vW
zOL&dWHL~exuJ*d=6ivO%jyh!K5@V`5g+g7P<C6D&#y3scfwLW7c}lL7L38=u<b8#U
zh;B2;X0LB&;Wynu|KdasTqsv&gr1;AmqnqS1gaC-N$9?bnyntMXOgY$i0>LFW4Bo|
zH)T1bLdrBuwuao`Vn<WB&s81`h0dHsNAv)Vb}kG7uaWH`Fm55MV_e7IeE{XA<Es&U
z1*u_ap~+wSGVrh<&;=e7S|kL}*!$&WKKKvr9Uk>rvHc6FpN66KY%hfDfKC&3nrobd
zT_6cels}1IlS;AiCp6!Lym#vQ?Zpz5^#cz`c<Sz;lBc;C72U)mI}Qsb5&KW>_jl3B
z2LF2N{><1;R%90!aF!7~(+cbBe&q>|iwzzZhd0@x*P-nvmiWp?`}g#V4@s6Q687oA
z{ZNVTaWV6_Ec(PsQ2OMOC3<xzm<i$AhR21@;|Dlgi&Ujg1;<XU9ue_bnBahjLri)i
zhy|MqD$i*w5#qel$58e+cBK|?l2@Jue=ey0JuSfH3*{4W>Hy;cXeuz7@Rq{7u90<h
z{jKkD+!#}qWWx&fYwA`A0D|@!_O0r#L#N;KFH}HsYM$b(9^?+el@GNO1Qh@d#G?S9
zIV`B;_oj8pYMz6jZ_A?@BB=a>DJ~Bx$y$sJ^t!cy2wPg#7=IZWLI#b=1a(qY^`s)D
z2}GedYY_<SkZmtgM2vDht6)&rM@_(DSStCH5Fs-NUA*%LTb$;YkE{JTX(VBGL^!LR
zBF}}B6~v6nF1r$80x=*$S#7or6h;SheiS5Pdkcai_)<Z*3d0Bhq)?-Q5GYRb_JHXR
z-ETvrVEo6t=i`;UfX<Tq#I{AKb&rW{7<NuPgn*L3*@0o_lzhC|jE`c>3doM2Uc%!;
z8bS!|1IA(cUTfw_njc#33J(SB#IXKUX%=Q$m)Z_W^gdPuX2E+jevPjx0ZJPTq~Gw)
z-~`oSrusdfG2ZdqzsDFrK0D+2_$s)3>CTp~RSkrDK-DU+`d9Z@z{i2D6W-Ik%ETIr
z$h@H5lMuP$tJH|jfG_dalgYZ<S(K>ok(CQl^uL68Rf0FHkolWxaYEep^*=^%h7{cH
zZ9jMz2qg~eC8(wVo%Xcw+WU>KhMCechiKS9_F_b(H(bg4H~{bQ5+fWxNKbCp0K6O0
z3s?35Rt2@#3Rk=aq@citX`>=;>0M-X&&uKW$`nH40o(&`9ZbFSE#8&%499PkU2Z%H
zT|BvH5Eaq-y8a^&LW6XF{R;ly3jpt7=&e@IlJ7eL9$=sb_p}>*xF5mHulu5E!*cn-
z9?$kN5^MTBvbcD!90XVNCcnB_JIZqCaqsH-XSQO!!v?T%Hb{Hz%UrbN<t(U`wY;*o
zS(e3`QEOxG<aH3dU8QpmRvfOHSm-jaB^4;yoyllR5>)e+G&<i~ce{+SQyJM<bzjvm
z>EQ(=1Y<y{L1I0?^zeV4aTR%>F=3DehzueyMbqGetjLCoEQ#D|lWVipCC-63XkW<^
z5*DDG!z;=b7S{~>bw!@Snx>Pb4gWYqz%a77xpTz)=F=qv7Xk^)Kd=QNRYZ_dp!4Ci
z0z^;{kRq>vE@>0k5}YLu%;~M#QDCkRgASnif()@K3xDyV8LPOXYC!71Z4EmjKo$W0
zT7-Rp#F7v@JsqZ&Psgi=*QX%9sp$AAs4k2iITUaRtbHjhONPWlH&ybV+8SaJGc&V&
z8=~ahGk=xj9?6;Mj`I(8mas@I3vhZ%)eWfRy&s3O&YlLhIQ~0=p_r@j>y)w6?^Q6C
zNF({V%@_ylpNRl{^6H6jj*ZwGe;){=v0yRy9Qpk&%u=9bg;I`~KMk+Plep7^;id34
z^tvGGH2m>O(?^wr0&RAL{gvuzC4-Bv>+7#J32igB@x<3%FEIj~HVEN?kn@JF^vK>I
zD2Xfk^kSXoIx4|X1vL^NXQ0KwN(G`pGUYRH!vGZajI#-Ns@=b6zHLaT5I~|g_nhF{
zd}RvXrel~;&xY7rdU5Usy6rWUfL)H*<2YEovM(VVpTI0SWH4<!+}1XqlmUP5XZ!vT
zt~Wiltg<}f#fy^^7vH3fhn8Z|{HzS@I)Qu!uB%HJF=8&^^k0ieex3<r#@KF-N1^O4
z%1^`2w$`wZ7s=>$WxpJ4kHHS=o2y!0y1xlPlne5YlMNq$7VS9!i)$sA-HRfx%cgHx
z)Ay)Eg!Xh0>7hI6=NKLpfox@j5Ab2=;AC(+9IB9HUUis&Ru{rp)?R@K=VVYo*mN=+
zH;d%{G6piZ4B(#v{u1;6@BwXr9wnsBpU9-3jrK45+m^NApk>Oy)WPTpI*F=-c{NA_
zTm2m1VR%p*u*>sju0I?PeAEU@gqgHuakngu<c|8FX+_k3TpY5LNu=W9hmZyy0DNap
zj${Vxpsp)S*EGm<SkyNh$3&rORbo%+jBK%RNrLqbR;N;*N%U3*nVs!z2L7InfOhm7
zsJ8*Q#}gl=7&J!<_EMPIL4Xa?4k-o$DUBy$F=v#&JfE>mS%(ZbXb*pfU@L`c58G$X
zfjk?`R<i0zc|@w~f&AQbVVJ<oZ~qgSh>W37%%LD|?#|Tl8=y!*wewX=1#U$sgC{r@
zlteI!hFK#TO^FQucXEFV7VD`K67Ix=(o6=#Bmg#`SH*A7oU715(Gwr9y9Pwd*6+Rz
z#}lGNA@h*sRw~QVk1rP_OK$LSa;&(5EzT_XUt8Sy#$Sq~1!n7gl(W`1)YO6J9cQNZ
z{7xcXI2)MxV@i%neD{TNln+BE6e!rdFj#j}+!*q(U_++L373Sv*1oK4BP+(e-bMqi
z3J4!T5*KsE5(^h3{;NEzoE9s;EjgVOf3@{USj*rLl#@V~p`$7g<g8c)|5YeG3kxkg
zr$J9BT5E_u3cV@U^w8hmzZ91M1}ol<AKBKQRk<61c!#uKiSKaJ0gMYjVOBi3^?F$U
zKBg2vKtXD%1`0hEG`_^f(XnK^<jU0Y9_%ciWRI6e_m4p|lW}9v-|?Uc>qN`p+tJpM
zH@7?*^e=(?Mnk9DmULdGfS*f8hWEnlGx8=nhMAM{`uU@8E&xjiFPhr4f2hxy;iK}`
zBP(ipCgh2;fnWg8f!}1*o!4Fw8h5*~(C5fHGI#o?`HukEoV2{<SUZXo)%yB}psC4b
zd#hMen`^Q-p9fT)11SKT6zX>Zmvn%5@y?{x{L){p@?7t83G0jlQL}N~MG(7y%8ek)
ze~$}M=lZU<ar`511oI9=Bs}B%VKo)R#MQoD|J$1Z&I_~@59%{bfnF4E)%a)hVMUVP
z{|ab8e_&c7gWl>XsHM%TDQYv8RyDn!__c2RSegDpOwQg)LdT8S>5iVP7Q%yj0Y?R|
zGe-p71nyDFe06#sct%m$#rHnZ`@ZQ@-do{hEEhzR27(@BuB{&D&7FHB@q3Vybx~MG
zN&uJOvxmt;B|1XEz;m&Sf((`v7?ebXrctDHW|<tzVQb|L4~SwJ-(H`ymR6MJsknjV
zZAsrkT6R&dw7SHKlA%^{QN?U5FaMQVK<O5@j8eMP0snr>{kqZ?PY&+OL?R)j)>2W&
zpBV{!?=wmj7i8KLvSfY)dBMe5FcnIyXhc0JzGHIc`o|*6=7l<2EpO>{+CWlQEyY)p
z4M(J9*N$qc$HNEOCa5a5w)2)LZE_NKr{Wp{u3tD}vK>2qH<7u0+?d?(BX5wQ$Rl1>
z_JBcSjFBOhBA~@%7$a%%nk-|kaD8-dyLqu=_*N>FsdS`qKt^S&XxyA*V@t=rwKpTb
zY^B4>Mu%)YR3;S6gZnS6PgT?utLrn^h^v_#ztCtmkWy`Lc{XE9upF@*7b208YyPo5
zju+X!>vAK<T?YXkKQi)%Y;+|<+;JH)feHv5LzRkmvR$|b8nje)NU#N17OrVGcpE_o
zFdfe5F9*^;D?zLEA0$t1r=TEJPCvM^wH?*A9uko$)QhaCAn3Fbl%rIJY?-)3P)R^2
z7lt)Ny#5tN{WIbKp8%%&uS_|uy9X{}L^D1(uP3rj;9`((kNMsOTkFYH>zP3GcgU#}
zK&SIVR0}jQh~%s2n*w}*4MpNXuWw$X@ZgL<?N+qJ>+HxLjU7`@X!-0|{dnB+hIim|
ztdY;9Fng;{Aq%cNM^;j6wL`l4<zI^;-xevE9R{SDzB8<}<{jMp8u!`e%hq&RtDq5n
zX9o{mYWdrjx0;Qqo)iCCcT+I(|Di19uhp?BJ8s{>z1cS){9D=0&-d1wQu`U(l8Z&h
zwwyV}YeU!kt4ogpuz~2H7p{0<vG#m39zQayp_?N7*1^;x+^>FE9k&lwWA_rce_4v@
zv=3?UnUPn_Zhj<H*p62^FYqm=9aZCJOlm4Ay9d$@B5d3nHlNa@EZ6oQb4VSJ_Gf9W
zlgo~G$JPgT4QM29I*PTO88|7ul2>MIA?KSk^V_Le`5WFsYP)Et^m<Kb{Q6dFP)7A{
zWqm5~;oa5#<`8$;=E~eoY_LSNhOr8q-DnHgm^;336ghIXbj)KwOPqk!?+cI>+-IS^
z`={L=T(m}0T9_E0@+speHT|Bj+Nrni$|s{B;2v6LXkDOd2IqDB;VuJ9IN0xj13^2#
z4e~e}!JT-bE`pkiI5O)Fq9imy1OopQ&>K)uz7eZJhX@!yf<<VuYd9SJ#i~?OfmsJV
z2Xq(c;1y^vbM_dZZ(6JaebpiH4AYSDs&Js}?wTv0%Etj)M|?&J{9P-dM?5U!8@sQ2
zQcSuc9JCc2rz&}iM4l3IGP4%>VK!j#r#jtHyt<Fo+i9#>L}tUwJS9nXE-p~7gF2$}
zr4Rq-`}s1lY?t~A>y~RYlJ<G-4JN1Tzni#I{7Ly^)s2#~JU^l@-?9JQS)X;hUB7#!
zc$$Jodp~;TlkAaQaKU5Jw>;NyndVG2F>P;!+mMPn>E&jMax3ZMkQrE>iE~VpBGc~v
z)^V=yyGPyanN+f-u<H70LmcxtwGn!&rI1(A(XG7N>}2@4MO#g6`EM)cGo6#}T`<Hj
zF~wG&9f&=}F^A!qbKYB6_Lfe)Z>|KN+cfXeyVufBlDf=MU1C+QJ(`z?S1P3_9KE*g
zQfPZEJV_G|>S*YA`Cwjaa(MoeML>7o2%*)J!K1BPN7*(;^44JqjW;f)A217@M`=;d
z@(HP?$l3KPr}Sy%-@y~j^t*I1_h9W=beiabNW8m6Qy40(bu`%QtMvWnAnE2EARJtF
ziEi8%IL1F3yMF8{?aA2#g!Q7x8Z>C&fvOrN0_V+;OYQ5%#uk^_=9^X4Dnlw%@@>$?
z0YL;m;qBL>TsyT}rh1pT0#t$MYz7Bi?+Rl%u7{fsqL;lobeR_0as`YFMy*PWs$zBv
z3JUITE}TVnkWEk9K;H)qV1xWlBG+jwIJ1Cy!r!7le|~Cf%SiS<Wf=4Cm6Zz9lJe1I
zro8VHoROby!0%0k{tIIc=VPRAg`iQ8Ow*HqMeUE`5#Y@wMv-Om<qej3Y0FsxD{BD<
z{l}e}+$HhYffjPtX2~oIqN6_lVXDpBn61Ef^{#4M$Hh@%&QVwHV*q5LR&)4WbO}|W
zJF~f~poD6phPxGY-njJpJ2E#VNotRo$``rqlDA*k{PbQCJ;&s-e)H<%YqLMxxo^5V
z=!!2+5GqJ+sx#l5sJobNv6Jc>W6|PlS|fK_b<UqKvFmdm_I5shJO8}Kb9R(u>1T2u
z>y=PbBNt1TH&TBjHs?S8{A|utpXoOnH6-(}K`CzY)8VN#e5#%Pm?`S2Fw_RK@i0ze
zo9_Tl;=D~A=iAR!#j%u%?Tj2Fg*L1{=5hWCtTtvEvNPG5bmDJ0tB~h(HQTk+fA=f1
z6wjBUA8B3e&<1!}y#raZ#~U0B;9oDA?>_Osl*fVuZ3%dvVH4I7^reTFq!&8|U23s0
zmQM?#*P((UM`XseF22fGhfQLfkQqce?!LZJ0KC#OG9qRM<U0kPABMv(%U&=CktzGC
zO*+Nv6X;PQ7&bJ`uS^|O?n&T&StM*5%jjA8=SxKXvWI&7OtrDA4E(5&sea;01rzA7
z^ZF*<Kr$DjfwBJN@oK^G^^ZdQ@5@7vl859A34XaT=l!a_X4=5Za%VhNiLmA9+Z?4P
z%^PFiLy6UrvLTuBA$iH|rnXbLcOB^M;soR5Drs)<d;H@1wuU{_y^}h>2l>BY>!?lj
zh$atqvW&7Mxq8cRN4-hoDCaj1-|7ursy@}8%EmV|;+EJw%N!Z~J*w)p!+7+k?;84@
z!&NL%=JvzHx~Wvn#;oUd`OTeI!W%M#w1nkv$eX=d{X9K@i!qB`>i00?faUV9mwSD3
z0|K~yhvs_=lQTX2%|RZRc6(&m@4}V~oh3LE=9h5shu&e69WMd7=2|TkyhHs#<<~N)
zf8kBp-HRp^%W;Pp=a^{Buo5kPjOYEf5iIP4*Pxh`GDEK6wd}EsZovJ09^M~(Bi73=
z6uqg0EU4!_eGoM>AeGI{*hWQX85kJO-Ppq+2*!Xn7K%Gs(*)M55DUcymNPTdVL4tG
zKzuTC=I76!xKPjcZqP(-qR16~(S<-A;Naju6g3D$?9{mB1h+W2WfW^izqRBG5@ks^
z4`9aY1N;YYK0x@KdFAH(?)UD6hlk@84>~!E(tV$p5Y9$^<A1*MVx3a7gdAR57?%N+
zKvwRaBdQJ&syB(zqBF)IsWY15dsE<_lyM`FJ43xOPxE&nSK>#O{zZ;IYNdG1>IUjl
zvO9yaEsPTliciNM5fV)rJk6qnc(JB7MN>=LU4o<a_7Z`pKT1LUB%jE2qv~9;7f-eR
zAW0X+ZpMF_t(6mh_q_=f=JUK_f1}IKT|H?MG3yl`z2muP^{Sw6=*4-9m_7GwGYW!%
z4(9J(#D->=sf(|63@Es}lOCbIouSHh4(cF#Q#Y`lDj#%ifT-`=ya^MDpcB?B^`5kf
zgz54M!E#%yKJLY9j&TEA)71taBb*9sY-q@nUF=?E@Yz>76VPxv^FN_fS8OoU!IX>p
zW>6i_|4vmFKP4kCB)`hr(SSvqfo!S>r})TLLN~rI%FDW@Eb3gJ$JeCaX84>rQ4J3n
zeDWw!7q5V{YT3Ue-WWS#$&9t4|46`~S5jP*6Gu00zJaOVPd~p%i^hguu0{9G*v{2W
z)l6REw(m0Ai^}85elrjKgWQBB@lY1jc8uB1!R@wE9f9i?BAI7n4KWLy71}*fuDz11
zl6ZdT@9n|b-p>WIgT(pmOo=FaD7_Fl0;<nk)X=gQt!4vU5Qm4-&>_;;KCnH*YmCkk
z;M09$WP@?+bVj^il$KbPtHQEOWC`T#t3X==ErStmVP#*zpdn9|TejIz^9_VJFv<Y|
z2p7XY*c6;RP!R0NKN2L`#htpwE>R1Dgp;_wVj~_PkA6TxBAi#4q?E~UBBpm<G~4}q
z&UgoeuK|FF#xvf7Fx^(@!+ohM=-t1{^_GKyLf1p?({R^adAX!;%CkOiN@puD{c!{8
ztj1PGsy-7bGtY}lAB*EUJ3@^<AFpgl9yA^;WwEHA>BaDlI9F~xeIr8r;jKs}TMO-5
zzeIFhBu$3j_v3a-+11jWd&<v8_>D+2`&?7-Q_Mk=p|CtmYx!2HbRs6XY>GmQDorY7
ziQ8o5P;Ni^nTYK+$egax&VSM`BPNa=ms8e~x|ZrX-1%x5tKB8@l*Zh2Hii)EKv?J<
z?H<d7bg3EMyU8>{nr_9ty5)MW3;lHNwRc@`GqEf8FzWuz$Wl~hZS$D#wobm3JC=zW
zHj(J0EjAAOQ*kPj*uGBeZ2ci>Ow1M&=f1!XTD>OuHVHw!G`YO}Vt3EaVsx_;WyENZ
z#CVM)HgTr?>~As87hBdLz$)*rxs~}@j*U|&$Wh&vN79U=UE{3TVljrK(ZH9ck6Fu`
zM?XHft=qKWpyn0hgo4`X=BDd7C_3L^m)pqgx_^pWzksf07$3QcHC2`c51(qcm!?%J
zPh1#P+qp)I+~y<+>i(;~6#?W7g>YlF!9SsPIwu2#$rNJ)Uo^4M*w~oEHM<8_r4|M4
z%H}Eq3c#6mM;G=7_{tygkYfhn1poGu1mD7Pu4XpV+b@-w>hff_hMvZ30uc=r4BgCh
z*2xb!zWTGJIZ<4ua#s_%bolSRF-K-;kZvK~s?l0km~a5k3u*uF{gmtY$`rUWWRr<@
z0bn!;OcUDz4e#mZIL3JY&mLvr6oIP!DLe(;U4beYj$K0zRLTy?ufEIld!VlPrN&i;
zU9b>Kp+`ygRvHo%-48z=ZQSe8GoblSm+O|x84_;Lr@ym)m4G>tNR@<O<V#!yCWhI(
zYM_~LQm5LRk6b3h8&voCmR@5&k2%E=$+ReZs!Rm0C|Um8?Cn;ukaXb`{<bU6Dl)r7
z`;tOcqrA9ep3J5k61=TD8#&sTcpYobbh6qzeFL*YMr=6x19g_93rV`U@c>SJVlY{d
zo`6<?Yw~9!nsbEAfR*lU_-WKHffk9gP~Xv<k@}KS`lCJloAg{yi>4{dtu4>wPP+8v
zQ>wX=S08f()@ovY7gP2SNq3g3)93W-)TNr?<+XZ>Varu%lMsr{K`4xC!*FFEhaV4r
zRdc86(v2!6yFA_<Q|#6z%;bI#_UfhJPMnJuO0W|*ZEiEpKj?R87iZ{K4@z@sYIE7m
zjd}IEgw}g$*ChT~*5p<vAudXdy^x+R%TK$Vz2^;Pxt-kZ8gjU}A12mTWgVQ$ZX708
z9lon4ifQ+3)1e^y4s36LYX^_Zr*zHxS6Y#4mPDw19@A++smWR@yoS<y=D_0B_d9ZX
zdmHXbZ2_ax{#pkr+>nL|y9m>?P6Q6b+#AZfWh&LOy?f5Jqp*T-ggC!{{+4c>Ax=q^
zaV#HmsFzb20Q9d+0lO2Ek@-A74~ra9ye<l1+Ib1=J(wLo4OkESe<R}ihx^-mcLD-p
zXFb|lVB5^U-K-z^DfvJ&c(&-N56PzBin@N?#&1lhOn@^)u7<`o4^x!s;dqp<$F!l>
z4WluPs1id8eJ@P=#BU-6TDrr*i6x}(dMi=Ge~u4FWj`)IctlI2O2eg8RJltZmyXpa
zzCA^$8fL;yJQ3=5+juxEny>sEdwIHm{VNBB)~2W7#24+m+q`elI*4=T$lIUuQ%Js?
z?a-fSJUcam!;tSqkZsv>k<|^uYVdr@<m@wc214zvPdBkXh`L)~a8Q4=4bEW1;^=>;
zZE$29YS!asCfZ?FI&jnwgflSE63JAQOfMSmuE(XMIcqy|6IZKXm0Mhj5c7rYNSNUJ
zG1i^XD=4nlnv<?94uQQmyJ2Jw(ssG_fr6p)R?ZkrpJ|`qoFf#@%61TraURD);hU2<
zTMZ4m;$BgdWHwvQv#R#GDmoG`;bL#eC7@K&@=Yqn)QlyOV(a4NQLHG@DA)bFaYc!p
zghf%kB?WwG*_YP)uP2_<J~Dh}@dXFfCO_<{$F4i<see;$m*O$=tj>jd7p_zLdKgWq
zdXd-M0H%O<AgC|knXCh4Y+jayU*yDomF#ENjDtOc@bzL!<uib`07nW48!&c=rW>>!
zfPnzp*L4${i>+pqFGP!x84DL}>x;>RuXd?!Ke2V}2q53mA()`LZuSTFK=<wS_IxN=
zoIX```wi*L)0r!<?DDPz6dmM1AsSi;bO0IxZbo7*3QZl-XRE|bW!Du>L(c_dBcwKg
zjeGvi-h+C`KsXD2j&*K>xw8fTDPlLSp0QAIy9)O*v|KCOo0~sDTw_^gZ-LDw7gMI_
z>(^e&h9Mf5-sHqCmHyLV@P;t)mGjuX{hq|xVv9|g-B!1btH){%Gd`mJTu2Bd(#Fju
z;v7Eky|c-8w7Sr9Cs0Vv#WSekLh^?;L;g>yiBFP4B8cOjW?ORT6r_14Jr~+!TN1O%
zrQ5`!xJBjlhR6ulqgRCfwW+Ois!z>XR|%D<($&N@GdlYOON^n1Zjl=gS5Ycnnv#^-
z-Fe<4-S+qvI*JWGlQZUZtAoNF*JyY=)I_m)J;S|5s=k-=1dqbqnpc&p-tmZtDGZZ|
zqY5)F_>~@+gDwpYO<D`}83hG}CYJ;`Ype%naJW_%G2XxEsN~0!Ei5kRzw`W3JNss|
zP2}vyrD{cQ?vk|NFAv?Su?*udyYrod{1*3UO}nw%tCw(GJ1C6?rZIlm6kWpbv14#`
z2*t^>7poAtE#FDA<C2f>Br|$5-jyQh+MX{<ln_?y_u93@Ur&zdZ-|*_&}8|=RT8o}
ziP60OZQr;%>YK1l@d%DvcCGAAPU_)IOLfj$R{xmyD!5!}M$?*QU|Xdss{AETTK2Q&
zBf(xq)-mM-W5$o-;eZ9*%*0mD1o}i*7Z)Tz0E9(w+5mEgFS~$bqbP|C3%Yw4&mp7`
zyM3mxlX|9+B;~S|rixu~;^Jl<F3RZ@Ex<P@eUN-I%MI9{!F+j=0R!Vcu*=9Kr>&#o
z?%@&q{P_i##;Z+B`XX0~%M8@HfM*8C9dbJ093lWc3=2iy2oz4kBx(K1)Vkhs1Do<c
zXKs%vWsv+=EvGmq#4j%PJHFzTODRk6v#7KP$E46Uq9qq@@MQSbwZOJ4*6WVx>nv@4
zyE@~xjCZTJ6VLVJ$(aKou6pMOso4rCm|X!4@0ySIdyZ%pa7Oa=a2}gSSVUaz(kZ~A
zAQ@n!<F5&vMpY6j;R+c@;<s>J`tI<`UR;3A;iha;VBA9!(bn%z$98U6ihJVBigjrH
zZvQOwq4hGe<1a5FZx^>Dh4NO&6{KNI*YubWN;W?h_|21Sgs<H_NJwNnSYk$^8Tozf
zZkCV76QxJZ^`!U5FN-R>bLRaKP2|(S%?Ojn$cuMMY{z(=XgN@wt9pfRSP6%z^~L2!
z8Vg4lL1aPLZ`oZwU@>dT85_G*D2SF3^(K9fe33<Dsf`M3FdWB;KmRMa-Tw7$0yPe~
z3UZoaXT`41wXHF?X$5QcDspTo47=A~WNNnm#F+gvbgSvx-S2FxW@kS|ESX#74$oIx
z3f&T?qqd&Kb~$f7Yqxk=(sE(_2jQtC*$=l<bbJctir&R41q#WxOHSJaxAS+*SEd!~
zu!f;zpNt<(hQA$#gaNv8S8RciNhehdyPzd&DfF$hrI8m9CF@Ban>#v2E4#P5GQD#6
z`MM9By@##t6c|wSPLwSGG+)YnXIWM;x?FBrk|{P)W~Oku!@U;9cZimPpGHN}p7)Ef
zXDm^K!59$XMx@vwj}U3#yRIFTS8uy)Wt;`O_lN<u0nAB7MFkl&(=NLr>Y9_RLpXte
z3Skg&$X<gx64YLB{AN8KLE0qDl{RqdYDYny1vhg6Ny5MPX@&9i7rt_)5D<-4lOUc{
zK)Asj1s_n^#3{HLW|pb>4)^u)Yqk#J9i>i|6BFBeCSZw9roRbj+S~f3(@6Yu7>~|u
zqE3h6idS;@o(J=+^Kj>%tfPgj=Q+*=Go9y6a!Ed?aK`GI6F;~W_>^e|N0nX0l#aD~
zBv^1gN)-Q=sQiudjIUl084m}?_t!O*X7%X&AnE^zBh~%(3N~(fF@3VdVLhI9el-D^
zvzKO@h#rmhR&7=zsV)w70B8Er9c|GSW)$Yo@5;nMo}dIR;}NkdY2FQc@{<8P6Ac%8
z#(yRgejH*@z_!;~#K53bigR|r_f1zQzT%Dbh+@Y$*QJhJ;S?27ai}V@{VAs2Z2m;@
zd0YXlTS?ysm9xh}g{?%fW3^5Xhu(WZ8f<r!wh>=}Bc{b+LQPJsZhF<(ADb)N`F)F_
z=pD_lO)a|{8D&hHVw`VF@i`MTs#`<aXk$eC6VJ!$Hx(v06RP(%Bu06PGSr0B=qt2X
z$4Y-_C&4MH&E}2HZ#-?aTNIKYt3#S4$Da*jrE+IJKKlyBN7?<Z3IBDipW-la>hKIo
z_)diB(8<cTW1RHfFAHaS4$9VJKPw!57`;BZa`1!hAo}`#q|W*FDzczN$Yg`gesq}%
z>JJFvUcY(s0gw&=3(qJYaNM|h)}k8PdqoH_RSdZA+~{&jKKne14zUr$G+^EVYGm1~
zC__CDMZ{etlLG3i+R;yNM8GrM)`pLy#UaMpnHfaF3a7Hfh-Z$d>47CdGZ~as7$x-8
zk>PV<!Ym7}dKrfQ4V+YqR&OS0>+OmG2Sxk;k)QC9Vt}+W(?MqFJ1S>G?oM^5&7S^l
zaOKeU8IDK3(_H)0nnnw<;k)kLiC_GZ<Xd-l4mz9!5;Ug|hG(;WjICpe5Yy4gT@1w}
zQIw?VH`0k|VOLH6CelYdXM1|+c3+0f6Rc|7U(Q$^82TKxIn#X?Y5an3P+<iLPOph^
z^jC)^bXi8%;iS^O7m4zfY$q4n4cui~t<^idWE7atv7dj*lPul2X6~*1VpCHf<9mL(
zSeDkI0+y%uWmlI3?D-}#-76Nnvv22hq)t}$GI1xeMTn5823p#uX<FGb$(_c}V<cnR
zp1?6YyC?JD#}((LK?P&&HcX-SqLe|W2(e9kaVW9XYYVc42CXw9c}dGLY)%{Vf1;tx
z@YN_NJ)*)9q4VSJ=QA-`{_H^VhL0obk0sT4n}_Z684uNCAJJ%PZpKSIx=?kBWAW&r
z%#pgO?0nM@Mh&JAPk*!oW}-+GZ<J`KXUre^3rov#S=dny=j>$C;5mPP*45;x=QWnl
z<w~flEnBR^na?Uy;yW;9mo1bS9T(U>oJNQSHeTk=-5c2%lidy9Ia;L*aPVFGnj=LB
zrAE5|n=9`KY(Pkzd1LOF?NDL7>NF**J7cQF<OEd&XsqX7tc+KAu94x;YUdA{kP|%s
z+-r>Ao3Qt0)zo45W+4lntAtDK=xu1w5TNwWA4)`RgJeoTGY3{AC}$wJ;e5OkHYrRc
zzOAAc>h+eG83$NA=p_+#|2v;;64(IIGxpl*KgX=-F-m_R?bEM1djL0XU8<A;KNqaC
z{`w?g)7$VaF_VU9jwLpc*ObLZg6L$(6AL-xIXbtBv9NwwBDs;ku(A8MhQDuB%?5=%
zwRk^LT!gCUy*VsvF+Lev+Nh}bypiO6T8&*Dl_HH>B|hO>L1C@Mt*FX!W-DSUI=|+0
zBQ>I*mPuWb%ce2y+Vt%hnGdn^CG1LOahb?$XQu{j-nZ?iCtzDo+-buY^4O{2l)2;J
zY&;yDBUb9?=5Z#_FT%$Chum%br{DIgg6yqD^nU&@|H=@@SJfjv%x6<`fr*J(Sdpo~
z8H>?A&Yh{%LfoQd-b<&vNUl=LmaM!y=(_xET-^a|TmpqbjghYO;D@-6<qs8XW7-7A
z68ooPPYuwn^5x;%&Gc8k$I4>X-?p=>Be`%#GGA!ELW_2p(>A78pHnrZj-!C-F1gRs
z0fUhY8W!v<yOAP%RSniv4}FeK`joR|1*8vDhu+_PaKBcz-OUykvz??+bHm8bhTbZq
zRcMT?^Dgnrf_|E#->pNJ7-zZnE*#^3bn{?K3G{f9<II(PP&ILIyP;zL5lp8XUq=}a
zbi4f5ZSHVzhM)#97Unsyq<4~3&oQE}+IP=nTeGIyezY7PFiveZ91mszzV;jqkpAJB
zWMp7iU0p@geJ5}#9LFUu$*@!zIcV}A84W%sOw0+WYkf`zuG`o+GSKcqe6ls91xjT6
zfT<ISG6CE8z@i8sb}qF#=eZYY5X!|-#P0xzSwJ)(DzNu=8xB`iW+sI4fcn+%!jsK+
z$}a4X6edRh<JYg#0FsP+T38;f1&@e$3&nrN*NUx2*%voPkT(MDJ?<r4+kPMovq{6&
z?jyr3pK+Yns(@vDyG|!nYBsq_Y>rU}9A-Mt7h1vJ(<_rSs!l0!xT`u~eKN_MPd6t}
zv?sK0?tS1(&zYKSYrCw-b+GvD{83e!tct-Y8aI6GAzfT&O{>c8d-O@0wZxaxNIvqP
zR=`J5<mToOKEmNjcf|TinKgWmsq(w|^{{?n%ubgKU$dJ5yFxdk^0BK@ex~uOkK%F6
zjokZbao>bgSN_;r>`nG{tgDHQ^@>TOxWQGvd0SD*3D?ePJe*x^(Kqk8%AD>zTjysd
zC-mX)Cb9J_7cratqAE>H$M8iaGWoE#wEW))u+&*04w}yFy&w)&D!DxF_|32WUg`2R
zzwH*J(m!<cW>?Tqf2g@>)@N)_b)Uza>~e<oew;&idxYv+hREU=O?>ARhHKM-a<&v=
z+8J@mxMCW4cYVl%WUi>Qo$0^(fS%phgw~EyDOJ04uqLh2%ipL;^;#TNp{4`-d*@OC
zJSFWp>?b8od4_v@9@#=sqQ3pln52{Ec{cEwqIEj_r08BM(K7Zc51&f(>u#&&s+0*i
zr+@TEi0knDx|NV$=Ur-j(SR?~Y0q)cn8el1sv_5l$M<f99TarfsGSaiA_dMp@Xi51
z3nmTZEE@5^S;Xj#Y5%!6X8Z}51sG=GG{VIUAJ7B5O|5KjIrC>2=s9q7GsKD%`g(i6
zU=9ewG>Lt|bh=tnQX0qo!Gp(O+yhy#S>XD8sewQt#dZOmnVZUUe^wgyNhbWZ&%<Pm
zP=J<YyF(xULm{+3b0V~T4?qqO2Y=E&!3YiwDWoz2q5YPFvCwh#VjOl{35}gL?PRHT
zD^o3MgzXnKPO=)Fg7%a!F;Uts?yn+iD+;3|5?f6{Gxef2N~t9F>L{sm#5()&H$<~<
zN>jdhfE$;Qk6)yX%dd#(dYN?R`P+MehPQIZU)s3|#VLF={JcY@@KlOdDB13oLO0<~
zz9$x;-&SA~)oOYyGntnumYU=tfpPt7?Vg{Nw#!d7h*qk5)P@}Qv0;imVh|H2{X@%E
zA~j1TMz(Nvz%MX1C}Csyr@UQlIpIU<&H4aw-Jz`2KgrRi;_1#mQ(CGzZ_8>B#);uB
zUZlJkfTh6dtrmq-dh-iwciZr~?Wz#D?%*Tsyk*nX3bMLr!Du|%<n9q1^1^Gi7nXeO
zCYy0Sa@W5fZl-7#Q|&#YZt|+Djy*kXR^xdt`JYwXL^Y2*)t@)i$f~B&HM_q&4s^=H
zT<**;$+vRHIQ1x_(=-{kuvouPoITL&5oIuy!!WNGzDGd!_d+g<EK;0HVkOBPK^+de
z*#%WHsMbJNVy;MMv+9<swsTr{GCmBEjGcdfRnH%j(n2}wNZ_UY=4G=={8<Z>A8zFg
zptV9Dx_qZ+gd^0ldM=DGYw8COTh00|U^uD{XhaYhQkCb}p$m;=dczw?upF4_G6-Me
zRVs^uZVV!Z;H*KD29s;YEIBDDDace6>02P}aB-71tt)KInn(3?kZDOirrkz!-1>$t
z?Ao|QON;aF-Mg=s?Z6U-f#FUs>y9d(V;LDYElKjam8HoJh4B5_8nTfUK9H0THbgJp
zXOT0-$sAaT1^rwxv9ZX_4D+u|)$k-J9n-pS-RbQw&z>_yNcDfpqES#N8T=k6qd2?{
zvemM{jRNp8Is^JFr!%KK0(|VN*mH65V%^`{=1m$I3@hP&$|E-E{va4MS<Q#7o}}Is
z8k(0P;E-umQJlH#v9ybt<<fDs+r;4Hihg(CVQMo)Lvhcq5Ox3E2=7|tr?X6Kw4rM8
z{pEzX7f3hysx6Y_epV;pnco@3FRa$Pj&Yi3r)#N&KT*D)=A*M>Tdybas_|34a~(H$
zUz9~TMOxKS@#L+$%W$Qe=N&KTmZiBWO!5En;Y@eFxnN>P?bTc{Do&)`?FfilS!VzB
zIIO0tleTZmd@}_JaMX>xtB)OT$sSfdv9fuq5o9R;{sm?l?gNRCY!Q?rM%RN~Y!6H$
zBY&#<xZO9`^y#tG)iT(Lc2Z)r$Y`uI&sAUj93efNwImuZL{pd%)P^CA^Gqo_Hbb9_
zm`9OyRWQ3eK}~C!KI14^jJ@N}P?m&-#zW&gR{MpQKwDMBzes41BT7-|NGYn+=Soli
zY=hOVPa`KpWK2^-ScMz+<c+bD+wkEP0zS!ks#U&9ug#})cRgzSDdZYXi%~Uc3y!5~
zR9u@9XmVTsz#Wp%cFF71sNBFqRM%MCCO%x?1<4vqZzYw0C5a+UcQ`qWsD?LtY>Mr%
zyupa41Zos$eUm!api&{sCqRFBl;BmqS>?Up)<^)oBzSO(zFjgNLdW#Oh*GdU$Xj>c
z0?@d{&1_Utb90D)i5<+Q&%dZ<$bYxrFexUb5Fe*FJ@GHko}QkLM2f-c@xjG4o~tH%
zV&5PtgdrB8;Q;~yD(uE9Q%npP$N@SkA9JD6{qJFX0ZQ7FlpfH6!AJ~~4rGI%Gorp^
z3tivWv^^RSwYE)55DBuuzG7rVa(1(+#%PpeNyFV%&I9+Hflr@@j@Nrz`*A*}U^l^b
z{o|VHieaix_C5BV&SWVIX|t5yYbxD68C&zkw8Awg$Z-PSF5@<ukFn?y5GP{9j#LcY
z8ojHJOMhQ+^iyighZ?#nv+$9%Hy51_jD|L`LvB@#eAE6UisSgrGsl2h%7|s<y*Y*;
z-I+bio6}t`mSRJ0=_1p7FAZ=xd>EsK2sx7fu!WA-wolqJmMfC9zG3q4+Z+93_ZW5Q
z8ux_#7mpeF`S=IgMM0Qff9?rmT%R!A+?@SvgtEu#lOv#_s4ly_Pqc7W9%?uZ8Mm)o
zML)kteq3NH|1F~DWpSom#Hv_BEM!}2!^Yb9LS$_W*}Y%zpT|^n-gOh4GCpTlCFI4G
zLhmc$#?v2yb6$W#F6I5LQ+YTy@^m?PO<1?9S;eOm<+-1{7U;i7Pf+2dLQ-x(t}Y+a
zQ9+Q6(#)}a64*iz(SJl4*VI!!W%snZoxj*Qz-i~{PvTZ%Ndcd1yA+qfnbM^gj@gvO
zk6s_Vdfr}1xqH^6*Mr2kJBH3DWfseA*Zq&Po?=29tLiWwciGfEeRkcHO`4C5R5G!*
z3fwYK7OHQ8hzv`aSMqt|0tBNGRHT?X$F;P<Z-FfH;V&!;h+RW`?sDpnK%)aX=9Y`l
z7$Qf<oWEk&*)1TThY5lT4IQywoXGp2(4I(s(7k2?W>cB@-49V)82|<X6f@;k$o%C&
zy^BStU4_h*lg1Ma+o1V?pE6m<)B~y?F$sw#2q1(*USc$DLqsT7GX~Jk;)mz(Zy0k4
zhhGG`3YjvrY~=*UaVl$7MzA7)k^V;9gOKD`__kTeKx6}Z(uaybd{(Cj&D}mCD=M0>
zpG_@B`z9PTftnNSg~jsYoesuzWEF%g{nfmxJ%0a(tM34(`hWjEj=hDlGqTAzWQJpv
zWTlAAY}q4{y(<;6bIgp&$v9Rbq-2+zvR7G|l~rcXef0bPf6wz=*STEZD><FMy+8N+
zzF+I6h%s1t86nedW%iMnvyt0X+zY9Dv23OJ7fahod&CK&lGL`bFI>oHM)AK)QN$E|
zZ@a?K=@=J%*1punK(kfWUjlm1^`DVSyo9yg?<w6T?|&k@v^w``m18S3sj_7v<gwT2
zqIrkKYz&4ce{V@m6{A<EnPt4Q&!;dV|Kc<37m>|z+M9O$@k+O#M|0X?h;4Md1&eDS
z8m1Q-voFM)Rc<DEf8uslr2FOL^Qu(nJ2ck<IC>OE=HDx+lq;CX^qM;aMG`Y-mwz>i
zpm4wS1^FcHjFXcvTN75i-q9-kq1y=3DfHE4U8l+yhlxuiA3FKkGfc=(LakLXSNKTp
zcCjfErx!fhH5Pe?ii<*HBXmfUl9Et2Z<tcLbC!|lP96i-^e0Ais>YP`Ry2+7L%g()
z`YgYJ-^l%t<g`*(SB-9^zR1U}T;F%x-M4DUo#K>VU^d5h<3lF{*IlRs0=b`zgamVA
zovDTQEK@0s<bw2Au9WyHSEj5xAJ^>8vzDm!TBGUNtd3YM)Z<2`>KmCysUKbJuh9n`
z8R7R@{w{>yZ@Im{4EWB~)4vPhkZ6|TBaL;5J>hzVn|;?0)TND$jinU76Hcgi%a)MD
z-8^#1!jwTPI&`B}=wbTS@bjVN(UCe(5dr!*;(o&@Ya-R0y9~O`i=P}&pXj<l%LwOi
zz*8IghOOHNpcV+_E!=p;fCMB7c1(cU4U`1!FepK9pv$HAzqdJjlTZ6nQ6DQbnqY$j
zXCHip#-N#O)J4{sv&bFU+8vRkSO4w)r4J!8e$zh3{8x5L4$*k_&Ei9x>#8h<7*;iB
zt{0_}+M;TdOX!6^Xny)Y-dWsb`P%*dBL&>Bex|vmAI`Q3y?<TKrU@CDJwe51`qXX6
zKAHGrLjIKkl&{`P+2Tu*<m<P8jh%I{B-xq&I=VS#KF#U9>MBXDwh-Z@99ADvUh(MV
z#I1y-x8#y<Wd=#r!)P!ornJa-iPLw->aFidv~VGXEPhcGB+58?;z^R69S`QVC895s
z2dBS@t(*whUoIrMAps-(B||pSm`TB;`Da)RrDiFjQtEwT=STEVEz4!zo{RF`*l>Gu
ze>e5?^E@wtNu(zV)3P#2uvP*3!(T8u>4|7-j+-V7BUN$}i`7ABs;rZS+Q<W4r4nl4
z<4y~ao9YZA#0N$?-|k>EI(1p`&k}IIJfgY__=B%S3VWO~I7LrsWMG&g&G_43V_!g9
z_n|1OLH}KG?-61iw<L24<%!d$LOhBV%*Y+&@Hov!{r$MwbK{!ib7k(TR;vs~icfRv
z{6f3s$CUrz4y#+}(bz^xk;b*I5+C>EY?OyF)=(Vq@7I<(K>b-GUHE)1l^CiVFkm@P
z5v62J!qyUT(}r;y<BxhrxcA;$8H{W&0d+KnBMH@ufrLTB_O<oQ;o(U*E~Bx(wzt!5
zXn$N{Ah|sbd$pR4K;VL<-lAbpHT-fU!i570SuM0B0Pak9W>{lqLgTisgQ8%x;QOCH
ztpD-e5+)(2qeOGo6&>^%l?eWYB8VK#DNi+qGx#5A!msUB3t*+=QqBLT$vlxzdsUs`
z`&0(N=a9(La;djzZNy#9ruQZ#wO$<2z>+ID!?|mN5#FaO<;Z!RR~L3=M)?t=<fojs
z9=YOqM@ZsM*UvH=rC<4gh(5#Vj_&SM?+R$#Mm6^1gvlOXPNK~{%W9vQKsxUsO?|3n
zFppBi(~(BxID6YI8FLNP(jsVHTRS6#$u>g0agvhJDF?Qt)cP#FPODcCaXpBuOP60#
zdMXQezo?iH_4ai=$lt$adaTKc)q)bUqU4bMOQcx0WBBpuYZ=W=j0z*1@o170`Pb$p
zJk<NhTf6$lbPk*2D0zfhvFr24ziiaGdh7)>Q~R@cWJuWlD4(Lym)q5y2|QJecBXc(
z8}qi8+UtydzlJh@S@U&vxk4x5RGROuUSj>CG)vprc?~s(!t3^wBq9Q0Cg1MRSjk<u
z%*d`|@PQo3sQ#@`TpP*TL&8Ys_)URJEYeIPx}(7OP8ynm$FwRypAPSM=QvRZ<xkG1
zpV`cxk?0s~q#3#1#-+}_TJbXy6#f|Zh;kqRpR;?SMjX}2A%}Gj#j{H-STRfxVO-rV
z^I1F)CH;6kwe)n%_MKS^w2*?w-WFf^U;K@!>n#a5YLPvJV{_IX+q<Ooh2*Hbae3Z3
zCnfzPU*fyfs0Ok4X|8C(Z769eKXyE9b7QLG@1lIzGRXV_ARZYs2cSm)*);+G274ah
zhNfj{pkeTB_{?I*+N>ur2$4si)dGzjBn4&ZiJwn4Ei1NqZvN45R@nZPOCT^ScQ;_K
zJxHhSK5+=`^5r^-0Fnj{EJTC=Y6)%UG&sZwbyHRKT;s&KfAan6kq{qGZLIGnVp#}R
z1xOrLMFM&AXRYxE4CFuIpkjxX|35;e^1C`9Pi5SNxey4WCkVPO`0VgE_&AAq<A2t4
zt%>O4{In(g{guHlJZ!iC@gYZY@jAryi8@N+Q{0-@OuOHODfo2r^9*|%rk^_-3s-yV
z+y=G^^-Vrw%-=CinqJ5WWC@5*2u5WSN8&iPe6!6G<w&y&pJgMKcxfMO_ny68qUh-H
z(3BO{+I(!-K0)-6Kq%kO4Z8Q1dg^v}sGd!?>)w;Ev-du9-uAoYt}xOi<UDy_bhF0f
zdgz}$|Mh$MV~o?otEvp@j!9--)$E-geqUPCvP#d`N{)E5H(2F7<v(WJ9?EFzE|KMh
z?W<)~Snel{TWzKgm1w>9+x@rhTp|(ro09ub)y8pLaP^3$n0eDKD=oiH1;Q?CljSLi
zze!S(vARz3D;Dbhx1y-iOc(DO5$Ote`{sADW#xY(aheq2F=gZ<eJ~?i&cz&)nq(vr
zK|_18!*yh~sfb5=35!8?qpaw9PAeXK%r)%bw+e7=>k+wyiBHE%S7K6Pi6xv@Tc46Y
zc@c{5x-D(%Y-QtmOVJ|m%GKe6r{a=XdLP2uM&(>JrR(Qdh=fv#eo3o4`sxN)rItl>
z|EFl)nJndcdwwn3v%73FIGV(CvMu0>pxKT#m(|f#$#<~1+S1B5S;*B)6BzeVgqNgH
zcwj9D#&ft_$5u@>yMe12c9&ys8ea{NWH&vbL0CTkE#BCw9V~nSv=TbTqY(#0!!kwt
zdIFBV1=*aAaO?g<c%U3s48e;=q*_-X2Ar{wL#HbOD^ozsuoF_Q*3r?SprX<a3LLni
zf!BFq8Z2#~4a3xH5dsz}m+fk_18At3L<W2c5ZQ3|{2}5$8tMLB^~vxvaKir$NuU?}
z^ywIY05IIa9}mz{q44bhdP8TU7eov~HSSBBZqZ%FPDzy?xrSqcU%V3i5$3M`3RUJ@
z7pw7|%~XxiQ=IaTc#f=JR+A5_Z8i0=?0n4H)nk<&?=!e%*c~`Z43fo3gR>k=$#f!L
z)YvbTuDpt)WfhHKAbVhLZGCHC1@|F@^8F=J;q5cK1>U&qwOFs7#`5tDsT|MWtsP({
zNQr7KU-l&DrV?s2XkM#{rA~i!iX%caHE?e|4K|(3wuvDyVy#_8JQUc_vfG>j_EvQd
zU#u=vZrQM<Pcli~w2S)eI2KRsvD2#)vXvIxU$A;1h{cM9SLpWlnT&y<o7FQ|sS={6
zXl{FZ5h6!qH(K1DBCYkEQI|@Ekzj{U0dtJTF2hAe8%`m*mMUWk{C>_OSMyji5iQZk
zteMR>4~>$eeT~mW<~yL@GctBHGZ)#PS`5jLbW1~iW&_1(i(<Uv@>paNrO~->l{EZ5
z!yb<-tGIPObW?F<#I=+5jcsXD8EDS(I@O<5?`e2xt0dmo>N_}3ofPH4987^(Ax=%m
zy<w+<BOsZ4I*JBrHGY{9zQk-r<5Kg>i}6cwofKjJF3U7L_0C(SDa+7=7SzEc07zJ^
z3V+DbgrNpDI{|qDXH!|&tKI^nrNH*geEmuU+suMQ1L}j9e&%UGR27)k!QOB*iMX*O
zgU1ozP=nHql?g3tT&D>z;E*c3eANavg1z&h`KS8{#~OdCs{iETLOutYRsy34I(eYC
zha%QNJze$%nWWgTp~-;d`}Fp_zh)MiklPA;_9A(*T=s+J{J#I9o#H1D9!~1-X*S6e
zhb}1ID>SxUbxWcYC>0^gUx~npgox>9%xsRnqWt^V^9|+fa(o13w}8<w=^a0GHCc?v
z+F1u0w5GLU|463%^6)PwVKdW?xWW|G`&B6WNS&G9=&>tm^kza^sPl8~6AqzcMzYUD
zrCj+1ZH{+ejnzom-qvg@_Ao`Hl+|U3ZBTgoriE=QcHYjg+Ua>Ey&8C~oKA58eq%}w
ze<Tk%X3b6hqTjE!d-*TsI?Hfo&#~N3<O9vOw@gwOh_mQpQ??vzmDRFG;<0}%t;?Nu
zlqT!{QaB%fwah;i{>$g7ZJpHVlx{AI#gM{+Su=L}tfJ4pzIOcvyL>8#cojRPJl>o*
z-H)e86vTM?7gAzYX^!K_4I^R<3_8xUE|BYx5@lh8NEmd|qIyO0u~(LwF%caapC*gN
zvn@n$8wjGNfgJe)2jcNG1GX}8#PRR_DPH8VahGf;lQjmYL}TCk&5_Ecx-W>7-S1BG
z;c<>6K^vB8?*`u0!N*q73_5Lu6gVOi?iP#tU9xVAOMBM);i=)<j@@^5Iy)P8gu{lk
zgDCUkNVGhQrt}?{!->NZd?pnMWyR!a#!vnp`Kv7sIy+2vtiQsrBI|I$@S>L-2XHx1
zyrgLG+%Ki;##S%KAMo>KW}^w44Ho@xSUvTs9~;$=**bfiB$D(Q-c9ny+FDHlDq=kL
z%!wXmMI2n;piu<bD<(Jh8)2mca5zcpA*$zFriMD^Dh|kamhlO7DG<K#FZM1F0=59K
zgkgsucm;RB?bq`&5K!^IN3G`jwb<Z3HvpPs*<H~Pp`UME5q@(>*j;!ecW_!d631Sr
zKTkRT1<gX#b7Dp8g=|i)SSd!X=u5rqywx4^x7~s<?h^5or^RG!7Eyxx?j_kn6uq}(
zDdL&E<(l=M4j<H^?Vr$VCz<G8649}GPl92hC}SnrT4b=@kb0T&b=Gn;cX#g_V`PqV
zRK3ZrB)RKqfZq-@gWG(;qYOJ|>#YKrH@AP7)0J5h3BL}XFeA3rHacz)P@#F1%d3@M
zXgRc@rY~wvdhB#W#)~G)K9%KRZ>8;ujz@L(>_+eueUIZ6Y}EV>x;~h1?{vhZq--!@
z{B(^6l`2>lqy?(uU5irDyWYB-9lK`>mP>y6BGBy@av6|(4~ayWB(8RSc2$hLMaqo!
z+PDm~hps>n;4Tuy2{oZ;-Z(8DoHXBWJKA6DiBv9~8LDLa)J1E(bUN-)t#?$UaLIXd
zl(g(Ar>a+nS0XxQr&rQHxta$nDwro}R{M_Xf6Q4NrYP9-%ssj4?)1(qx5AQ9uE_nB
z2up)7T(xdb_CL5Rv(fZrND2z7zGhxm?qH`(TG8mHMNm8B`!Z2}wlwU~u2i3^SZ3iS
z{9~#{fB#Xs5PI*={1T{QzK#IVN5KoKX1K(T?nL~D+!Fk^BSROcCf+98wC1nJLl9h!
z7D4z0Yhi%mIdW7dI92bgSO<3M7Nrg(`GVmCF%1C-;QFfqpvY3<7ci53PHz!BWk#2n
z7d>1cqE9aSD+Ig{45JV&4B0E*{{C<`=Ub(;)MbZbOE0s=5wx&$@dPvo8dlgpyZQD!
zjQ33{|Ao#^^cXXcjA}Ok=%c{7M4v@i*Yf#Bt>n_E^QSF&W%FY~*&2008MaJ6`y$_y
z8R<<|`R6;(MinEfBr1AZpLuQbR?E_+sxKUSiXl#PWz*Biim+#r8kEW=k|w77?2u}*
z*qT*Ysd(Qh%U|KOy{iNJGGFv#b*Z{bHsKaG2c&u|%~X~TjWXu!l15UtNxXfJ_rH=V
zR|(~!Q4gmWx14`U&C%4S{vvo(GTM=dhau+dJke+R$du4u%<q@pJyP$Pv$rIj+<G&l
z`p0yqnvV5ycrJC?<#6pSRd$2PlnlK_t7{f5()z-sOeub%A{>GB2Zaap0yA^lU$;}Z
z%Bt<lQU{Q!r>7M{?X*Q=6>g$S{Gw6^g2d42+#QYMiblFwDilxlbP~sAUx^E7QSnn#
zl5|C?#HKL)WECY!e7i}+R={lc47am6jwyV_@Jyzf`2vSl$ESujBzN>*rAAoc)y*`n
zp>d={!2`B+8S@uF7b~5V;7cY}Iu&J^oksTHCdbJn^LsQ#>F3mfAXgQi;&+S6UhEnj
z`8Cr*PHyS-=IO>P)x!bVgqvKceTkH!UbY<Xos?OZQwZu=?ElCawI;Jq%!oPslvWg$
zC#spQXU+H^)aobHPpF4m>bEHktM*sGPBNKyARV?m3!HbVv%u~riz7gsVCn;)+(i~#
zf0fo1$2zLWQbFv<K%&0Rze;@>I2^fbP{u+=0-@LYCys>}OW48<{0fASG(V-@!giz`
z=^z2;f$JT*PeM5B-^15rM2l$|G$JkTI}+B&uBsvkbbwO}23f!<2}xL>&VV<3`0!zk
za`0>G@O%B{HdWtQzbtR@#6VO$Aq52<00PVp8X^dB_<v?`$Ghc$fJxB`C3e^o@<y(m
zhG|{v(x&+&&q8BKy1+<Kw@Chq7Kgz(GXBQ?a<cHth>LH#1h0rQb!N6!W=_A5%^Rrp
zVxkfwenjb8_r~d&J6b&YkxbHXO!%44f-$7BOjQmKBPq{hqpRa;^w8#x{Y=MEd-dej
zvv%D}4oni$m_|KbEk*a0i^>RtFq)y+=+Oyc{K<hqzo?Xi^}ao^@-naHACJN)z0XYe
za#I>e+<Q%vuG0Hrc!~^hho_V7o`6BKx)JhHI0B<Lk@Z_u^i9{~O*34`UL6-b2HP~p
z9oXoP?_KR4Lx&CCcuSt`yzsQhQ7O+MsO~aukKsNkPmfvNlhl!PpIqZ-!M;dTsjJjj
z)kp2BtlYMbZHVWj6b1$ZNMr^+AX+wk_K&2@Q?qUoEDPl@MxvQ4X)EcDL(?Kf@|?4f
z2>!r~YGzyz;fieHBi>=^ziPgfX7T}%dOfN$f|*aOLXD5KSSHh4S(*B}GJom8%SB`2
z?*4<Zn~&DB8f&WblmccjZY9pynUhAOm?o9Nxt(`>=#kjeyDwz9QYiJTTjNrPgl(n6
zx3gRN<bBi#O!!%29l15c_Cwq`q=<ad3iE9K3giB42`BH^g<Iqh=OJ!tK-D#s`7zg+
z?5DEYUy)`trFQe`zn~OSPJV(?v^z4r!Eb}(3IKW~yuh=1w6)lpGvY#pGXL*!$h%v;
zrIe)yY&cjXTe^q1aL+H^nf(IT&3WZVk1Lw)7YYZn=d|l;vv|fbK$V0U8Dts%8E0VS
z4<Qv4bhlu!+r0hyqc1*jYhTOr2LSX(`E&%&EJ5e-Z$r>zb7+DT4X846l`Q|S@g|h_
zhAcZw-QZp!m|6bK;SbJ%lLa&&2weS!N?h-94_nXkrkxWp?s^W>FOsh#s^r9%6-Y{d
z9uKux>!Opz@;biN9w8U7V8D%(lVLWto?`rYT=37#H9wO?@1U>SKaHXEWOLM0Fbl{3
zDe^32w>(GI@5;lzcpyxUU?CID+!DcTUH?t_l=FU**Vr!Ov;&Lpi=}s#pEusLdSA){
z4|3mfo5hs{nYUf4h#HFYa->ir*YYMZeeh~Ms)>3pv$i5P==DpfbiVH_luzBaoF6n=
ztV$`2)IUgl7-)_;j`j}4in!wal_$eFg(vHNKM1-$Wx6bPWkt@|KXYUV1dJDv_qP=J
ziPMUAe%@0(@6cZ<tM24C>3!ux_bA#Ii%L=qX6U$eOXvgIsW_=Yf_PLVYm^CnJ5SuI
zrAp7xh_Y8t3RNrgJ*40xTbqIngAOL@eX7$zHc?~hbJ-&CM`s;t9RIdo|9m;=imJz3
zlJMGPAKP9I9$egg+DZjK-I7Gzf?07j9yA+?T(GSZxse>Nvj7R&pvSdo1m|RLo5<yH
zUvOUEHP`oCoRR3w%(|`4pCI2Q5pG-&Xb=V%6Y%w{mi_zRlymaV*Ho{7l8q>Wfij<y
zmnexthi8s2Dx5svGJ*RuC4HURQs^1zw*P&hC9&l<M>xS9ItaY^zp)3Vau84YaBAuj
z_!otTzt$#{LDd0DS{5P)*wz7dl%r_u7=j(|9}EO258<@l>n9B!++cdMf1P!SL!PAc
zpVo)uL%aH5&vC1tMF|fecI6%)$aVoiTFu=j#JRvWUXTz0SO3EFqGri{EAg_XR>yPD
zzro}Vie8wI2#OE*KHy0pBLVV0@#=qJ&h`PX5e~X|z>Eklu+n4}e7T`ck*vU#j?UUk
zZBQ@S&^onPLlaW0cu;Nf_C;M!5iKvv=Pvc;D;YeFAV)0mE22H=?2TK+Vyw6&kq>Ti
zQQxjBaH=mS%cL|dag=FwY$ck&`V;5N<lFmkG1$3dk6E2WW<4pk(0If5EA`R2Gkn?i
zWsxLA(f-2m<{QQLYlvHCaqD3`HBVQ<jD|g`U)CN}dVNuC&)v7Q$*o84+J5Gn9#pnL
zZa?-o(f;xy*C<CGz5D5c#Mk!r?$g}FLRUrVTl&7;F7E$vQXyzc?rTqJFy|z%7~*vw
zV!Lf|&Fn{X%1uR^41f17-t>lF9~!R;O{sb-vLuL3HJ`2Ck?Xyc7IfC2r1<l#XQZd_
zzz9?rU7PQhd=c-#SgCOHaylwD+(TF~oh?b+<l7Cs1r<_agLJ3tROj4`R4&P+`O->U
zSfPNqYB4jC$tayFQYK(P!BB{poOUX1#AWH`OjMPtAz6)kE>Xc>IfQZNRmZVJdOfnp
z_uWLv)Xm2oCfV595UCp;%nhO!3sC1B4=%Ag$t}Ohl<VM+yqlKeMDrwYba=?aY+ShD
zN_pP+6m5Iya%iE+n%^Zsg%MI<e;V$*xu5(}Sk0^QD>WnVL+W1PKQ+(nbtSq!5j0ZN
z!Z9%v=_!+jx-3XhhV2T^Y<z8$CN#FW&sLnj<W(O|uS)(54EsfU_`c$xy_9ZlL|3F(
zeGX32mm6uX+aw+mfo7K)OkKBb-D>r1c;{68CZLft!kI`s{`$I1Mtb8>CdSO4KW5>C
zY<h4yz<2~I5jC3Qs$R({zRcP{j{>CrPh1BT_BtzSvjfL}1UMB^z<WbafnU6$jQdYM
z1{6~OGC>wc0lKvpya`*ph2)~&d9(l7SqW1Y0d#s{3SQaW{SO@y|1HV4vkK#Xoqv)X
z)_?OzBwRS+zU0Qt?f0TBDmsP4Z89!?ag3N(dAf>{$NAuCNKOqGGWJZSdt0mEQ}>9s
z>?pa28M9G|&SuoCwK<u<We(F|1Dfs~bW7?)GhL2)_RkV!8%4_2qHTZk`ljYXk8yGy
zg^?yXPj!9K$oBR7Hos~&Q9Pe4nh|8<<a#H5n@+AZ%ay&ep%j%6tuD<$^?#W%RITWB
zl)qbW_5%o%dh9L^49q02k8{1CAtrZa3gQV4Z)y)J-txt_OWhLLv=BJwoMRGrC14Ln
zM(g^dp)Ho-^-n~zK?q1ienhje)mm4ICstdDHWGVTG|SyXxI9|CLo-a@G#ueeMt%7j
z=GE!9iuB2_)~%yU(YP((s;TksXGMCP;b~F#?S)B0b8$Hs8F|O3$HmE>swAmf$%yXT
zx$ljf7D$RMug=Ym>zyh&|3NW4naOwie3W#_uEmHs=>mG@n6M$Y+-5f7orEn$qhmIL
z(;>gQ2rr=C;axCKgHG3Qh^YD5+Vi2~rgNN(m-?Vl)^gQP@h52#1FmUeUX7Rb>XzO8
zMQm>ItP{@?r%?jtIe+@<hmKjPXHAPmy}Noa{>v;9%=(CTU_#71oTW=%L7W*@zDEzW
z-Enj=q=onhT-4q2&9&b-5*p_<9n#il24D|BO3q7xTx}ID2B=b#fn=B57d#7u7V}?%
z57>IZLk}o{d%cC#nj1+jMgPM>3JnTqED0BW@YLQ(SU2#*%J3uj!$1iF%RD4eXXlsc
z%xy|sxR`E@F8Tvrpy+!3*xy+@^+r#8opHqw1ZBWjR>U3f&&oQQ_0+l|+R~fN*eMkI
zJ!BG6{1*KQmi#O|*eyViI>8@)_Uu_8y_VyX{^6uY!`fEUoQ4|=epB~BBoCDvTtx5{
zV&k(FL+;DIB;lwW_*pskJ~Q$?Td#XmZ_u5ran59e9;B@z{UyhTN#)VLWGo8>-+GS+
z<z4nA<G`<v81!}{JJy3p0@Nk5$zohWus4^v_APm*a`=-=7P?Q-D?IM!A_{2}$v%Uj
zcb(eW5cRGQ`OSV!QT$qQeFrHZRDn<3e2W7_w^GD#EKWBk5++_$(I8dc#f~Vie5i9)
zuTJ~wp!_jJ`CGiw&f_nru&ZWM0otNW!+TaEC?1>MRkt6g5OGY1H%9QQ87Ys%$sdIt
zjqEX21L<Cq;Xk|epGP@v^kx4V^+w5lc2h~RJFEBPqnd$Y(W(oNIfz5Y$*qd?K5C;m
zf8hMAZ-2O3Jg{K7akAnOqtDVktbKu>P$a&SxO-kv(c;5qo~+Gx@)Vcx0;_bZ3l3v3
z0n1-zZ_Cy3-F<HzEkM7Rxp3N1)}UXy_#%I$M<%k}n4hknDuRZw)blt()KnyOg*-dO
z5PABdu%tgG(9j^#83_YywUuD>g9pP~fl8w9QRX~#$(~9&V+a}lyxBqwx^q$ut$N~0
z0Fh%}w(oKD$eDX-1P8_;tWT{l3+dRGS%xG0@}j!qs5u{k)<IeO*c-2?Zo#5`Utw87
zxiT5JxPNg8JhZ>AmP`dNN#z!IiquVbx|uaVV8<sHG(~}=WlQKD0a&7VYt*uoXqnMB
zv>W09?FN)lf*v$C7uIAzKr=yK2jIYe5?}zl3O!cyo$ti*PFKtT7$FHVyvyK|1fd?-
zs|#9KAS?oM%^-^wOlu(W1U#OAQUG5H_B411uFGW;VncxK7h1zz{Z(aVqO~x`L$Uc*
z9`<)Piln5V5Gw(Vpbq#i%4L>!Sx^B`xPP1&;VZmYX`nOT+On;m<i#NzKQ4b2QH#t?
z(zqgV#Xa>R7Z2rtfUp8N_j1@H+yLTH!srmXiio4ppi3!^St>nltcd!|!l-0d#<@M+
z3)Gx=gxoqqCH;MIi-jvXii_Gq>x-ir5~1ABE6GD%7&S5f#6C_hy1M!c|E9|k{wIab
z5Be2n>P8p~g>z@F<OE)~BfnBtr@i9$UPY?;_02w1yY~%qJ#`+cNIc4Hki3OTz1RO&
zobl}Al?Qm6UZ#Zboo%J#3vr)W=b~0y8y`}SN+#eR^tCjycT^A~+~xP5mv_a|DpZ?H
zyIkJZl<BHc&947=JmA{xa-R=5sB|U$yafkUNyU2p1yhk?39-w<ihdrZgRiZ!l+I)a
z4K{S$wb1Mx1m*sN8vV9O7rlM^*%W15cSu>qdWJ}MDY;mxZ=HqGh_MLq_?hdLyv#8~
zl;o(`2PVjE(sVbR&OD{y;35@~);xxW#UX%qcGF7)nT8KyvW(-AGII<qUezd~Pi2+n
z?{OynB4z6&W6cv?_VvOBagT1s4%*1PE#Q)1?xX3M+J)*!ZUb=X-UXr*??TRlMjg%@
z%&hTfN1#GkjR_5J`c=u**`Ix1B-xA)wfIVhf$6l_8d~1kEIU{N-y;eR7x-}@4Fwis
zb9Mj$WFwou(%5+(w9DX?RE~qm3QQVq3OnHZ0V0WAR93PdW%LK7{nG|`e{foWa1Qor
zLCyuF{H|$y$!PMbJZV@Gd|G{Fv-K^()B%SQnq|VFgv<QPjm1~m^5JG!O^4=t^}rAS
z$pu1s&<y}z3f)7aFoYux_`&Md|Ak+3<9qAnMT-AHbw@grDpHX^rJTfoOLas7jtx{6
zh{|8=LEKW<o+6KHww2_&TSRT`4AnvtcX{0;=>^6NQps8!Yh2r%US8I9RMMX>aA9)k
zspz?}yqDxJQ&H(TV)Zl%3tYTOXN9O1X4ZtdA7CA}#=`Ql=ns2e2Q=~HJDi9mN2%ZF
zT)O+1iaf8L=bF>mrC}yR4&hee%Bm)${@PS4j-e;&FOyN9B2BZ(xEe0rrKbAS{N>!e
z-7%SR4F^Qs$DI+6^E^N9FO}m9eRgeGoV^qpV#`05TYRBD%trjSRxn>>3lFDhwpYke
zM(nZvi0T@p9Z|HoAmT^A-)T0wkSQyfUR7SfU3a(m<_FYrmQhWX#02^MqJc`EfO5Z_
zeAl4rUD<T*Ld|juvC_oo@*rhx{f>DaeV%@~NKbumzWr2{ofeBlBS<-nP9!BkT#FK~
zg=dee)dde^k{}w}Be3PigQDOwI4}4hXEa5y-yQjhY#$r(j(7A9t<-Lbqs|`VdscI^
zKVS24i$x$YmB<I8fBTg%A)m)0YHI{Up-lf1&*PK4^>Xp)3bLo>NkbCq=-*QE8Zb?J
z5W3{{iGE3n2GVe$a<M;%i^~BD5%R~Qcv^4N4piKm@~j6U(&p@ngAz;9SdlRm%w=F|
zgQEu*GrN&0I3Yt+hgeo%8O74pE|Vnb@-x5&;Jv`bc2jc!4iW;821qh!CVF$5jqm(z
zzsCHTE-+6LHhaT4gfM;J@k!KM$0(LlIzvt&bDmUys|dzCLU0OjT+ksm(c=F1Mjw>P
z2!4H)g0SbnDpe@uZu8m*c*R|0RiCSIe=oDS>V@uddi)t(Nq#l}lW^eS(C;ga#c1GN
z@AG6VAdM8sPrJSB`4OkrW5DouG3-L%pA8>%V%IAVN4qOJmX>eZ+;(KxK4z7ot$xUB
zyMDO7V|}o{{8g%|<zQ)*3pZ{uCergU|9W^M$%e`ajAbe}tBl0`y-aW4lX^F^Pm}q-
zsH0liSxQ$lOF$?p44+=&4i2|*3i7l3+{}9U>8B|tX<-a`N@y+J)VYgaXQSqRK6v6r
zQyrcyK2_qkoM@|;__l5j5h1pr&g7?hVO{MkF|h%pL_E8)K4|ddtV+|99tH<-0e+NO
z)`w&%b(h7+$Ct^&Z%7*CWLC2lt*H^Gt|pF<uWC6);2OizRWfHdY}^hRVhqe}+KCjq
zL>96mdPwF)eWeZ3&3Q&VDLCVj6zlEKj^$mOFSwGfN45pIu(4Yjpl4|q4VFqyeMNVd
z2jhH~u2ojWrBw%sR1f>DeQ*kyltUS1ox|}x)k{G47Ij&G%1sdK?wc`|cQEg@0pS%o
zG$Q|OyT4Y1)FzP>*MBI0qjF$h_52*-mQP$3VO$pdbnz_j{|~LQJOTg4h<izR2&h36
z9)g(ykRkM1o~{U_23&8PP`%k{EmAD&4tit+zuI%~Nx~&dNF0F4C?$o7klGH|DCj!?
zxF=w*{z^2$)d+_q7?ehD#;FoA;#LC|Nx2SJ_Xc1{f^0`9mY^>N6+OHQVaww||G&hj
zT28N)*?46$n^G4GxRrN3$l-?pmsJ5yfdHw3k{Q5$_)aNZ#(&GAS)PtRt=Lg22YeS2
zfFSbVz^M}EJ?n~6?x%<SVd6vQp4Ft-A^9#+E0hd%I0qSPHfB|OE1edwTV49@dj8`@
z+EdvNd)YFv_sxh}1I`G#u(@&}>wCS1P8$<<y=oqiDxuUI^LQg_Ln3sQC?&<WJ!!Ro
zLF(sMZ{JHV-d>-fnv30X<iG0m+kSes!&|?6An;4Twj%!3O{Dx`jcL_}Pf+k4&n1fM
ztqteS8_a1Id@nr9>u%N{9wv5iScE*Xi<?0x2}{lGo}E}t@@MN!7imzjH;&=<#H{^=
zW~<>Q@e?dka&97qu>*5~sFc05LtH@c=Rie}?)`q#xi{^w8^G(>zw_QlpG-yAKB^^7
zCY+4TnZBk-s9_V6Nz_Hvt&s892!)fU*SY&{hxHZh!F%GFv*~jpts&Fz?_BO?c8H@D
z9&7bEY;ww7qKiyPImRCn==J7&!_>EJ?Pb&m_R(3T^cxJURJsmh4=L&0rVVF%+%iXm
z-H@H;+yRY{W5^Y+6LI75khe@ek|@id2R)0nZy-NwicE^66cN*5v8`K|Ilu7klDV05
zaJ=e|d_CKB)PeZ@MS=Y!lynAG2u6HEG33L$>+%=EnT+I~g^jve%FktV79GtI#asUm
zq}s%K?0e0GXBE8QpjuZHqz=&az!0uM@y~J*lHAXwcrieU3(RZ(lpln(*g${+z62S{
z>}Dq)oP%k6)I{mtAHy4_0g(je9`vC9G{k355>gAmEz&c$PEb7E*ul&|Fy222KcTS%
zl=W(bHG#!~LJWgFfTIoe7t*qs#Rm%*HDE_FfpH}x2JJHV)5pihr{V1gy3xi)yqET$
zfvLFG|Gmy3qK%17KIzz7Z-vivgdOUQ!t&XD-ATBBsL%Nx^EB>C92aR>m0NMB4U|+*
zU&c8Ob2d^D`**q9%v<`uF1af?5%XUZd=zBUJQtD`*=~O33H=SNVGlg(9NI<YVNmAt
zZpT7CqpfJJ+w>$*SCxbJo>2)Us5ts<m_*LdlfyjpTf{P@EOmrjD0t*LwbSV-s_h2B
z1@-ku%~p5*Ackf`cjiMFhKF8tzc1UZGe14}+BQ&|b=Ur@gZ-k)Ws`5)8DlmTA!GYS
z3PIs|BQ?gk_QY0y`MpPbBr<;I)#k;US^YFEj*u?S`tZ*GFm-XKPd8!aUHc#+x?shx
z_z@~J|J*fi8AU<71yz2hQ9Ng|J+<72)<ym}v-gW#l<Z-0J&c$YWS6YX=Ah~x5%*tK
zc11-_#jD%I67~zD;`eSEDr;~zGV^;Fhz7B0KF6XTV+@=lq!Bt%iFhG-*?cUs1}(dD
z)zcb#(en>G&8=UJYO%Aax20x;6*0)&;iS;Pt7Ns#Fx#dp4mn2h7`-}u91-cMyn3E(
zBa-brnVLl_iq^F5={udr65UzRxHfja$duSNb|W`|cY;Ybgoq2wdyWck-$Nm0rsIZB
zKkH|4K7yzna<RsT-OBvSi6_gCEV;|t%K`Q86T=So{sg!D8p>5lWjQ5M9G4vY`Xofc
zfUX`WSJ;UH(m7}%AyhYTDl{H!<AAinWPd=seYx}Zy<%K5q%XpPkaJIty_zTwP*zb?
z{Ezkss3>n}D*+z`vAq-xZ!wADxoR_nQQ4Qx0B{rm=>^;W2st1C6~JkfTKSFClIYYG
zfrkV7+12-lu&9Z!rUa5bM$=w@xdH1JHmYK_%(WKY1LE~Dd^YI}Cmm$)_1~x^5Ef&r
zkFuLr5Fh`0hHZ+!zrkZJI#UVbZ(7}cxeHGj^sg}M15n%d`Dx@dU#Lk>#H~I`{24Q4
zMMg7h=Yc}>9T}snc@LJkrfS@3o%_($!4S?KntO8JDy<Sd^3BxM(Jm1R=2wU%<eF=H
zL~MJhRamW^b&x5p2A`(yVnc_Aa3B?-_gOqJf-{P+Noz8e`V+DJnqiHpwd>2+?k$A|
z#-^rHf~d4P?9Kh|`alNEO6rhAk*Lj~QN7dZV5752D)B2afrXQX^mjS}r$m>hh#AEi
zbEhsTwS@Td7p(7Ge?}!Z*5grE8_sdMON_3xEkF!w?-pi~icKW8N||U5-H4#&&30X@
z3HT)ReVFtC*PkKZvw{pGB&oDK=ES`uRHYs+3--kf9mZ-5PdcqegfA{!R1y;GM`2h!
z;vKm2F};G;*=xk|&B6BV^*v`DCxXlmtSvq%Db0FXm1Bi&MUrS3vdxf~#c4$NkDdAb
zgV}O$AuKsHo{0VOZS*hWSBz3t@iIdWu0_<Vw{8sHw0h)vO#@9T8hb~^#hfbb6~FR_
zZgvN`H%>hLQhrpD_$Iw+2lZAI_^kZ{<eCNCd4h^j@w3MxiggA$6@zGFYjv9IHJ*B~
zmRT{BDU6B)ZT67wsO^#etvgxpcWY|bW}SBIBdDVDthM&wbPHlvN|LjEX?XOOsim+-
zS(7Z~YPa^N_j7DQqFa8sh5rI3;TRvf{>u&5q`R$iKo!7c3?M3;QQ!<<`z0j8eEs@W
z{cSg>_l<SoLO&lZ|15`xf+4OT?Y#6OIPLIofRm1Je&H0u5r$p^WCuV6z}o<B3NLu4
z+piCWeQ>#nsm8nW4;hdEFD0~VD2vdO5re9lL_Ob&t}3@Qil_e}oOme81e$`dM;usO
zLWT%59>6<*3&m9-i)+L@Pt$*;K%S^y^Einp-YtX@Q=0HCWE@Vqq<M38vo;<2KYWAD
zy^wtK^&CC?4~^R#!!BpvLo4&oEoG~qRW@<%$s~fOQij{QPrXhGF~&@hF49@{wX})Y
z_v971b-PnmV4EjC-BC3{_Qi0VM<kKwIZdmvQVIVi&373UcA1Fb*`!B~C!+f8t|t=b
zD>_n#uRkz1mDN|H(IYeZ)4Eq0kx~}%%4A(PC+Bx$dAqh)beH4mTJ){Zn`O0IzifTe
zP~6mldwb%KrMnxaZxCr;qbr{uql;pET92e@lBZuICJG)j@2ayP`LUaRiK2k}&d<~I
zX<KPyEY9CT^yy1y3pe#lT=6y%HWfvEo63P$+3u4#i4e?7`WuF?x=819YUM2qR~m%f
z(GA~6MO11uM;Hg>d{8I9e_AMI7EXV}nVv?L>d3h)8{11{{=7F?F<7cIB3GYh<hE@`
zNKfR<(u&A3acXh5t!trI8nURUY9yRiJ1@MaT%G$?&~0v>*A)h?L&2kXacLJ0<3(wq
zKEvOh#l2SQpKOj-W(GTM8lo%MlF3ph6dpX{^H6yq3GtZKINkhqX07dTrb-N=TRo{y
zjMsR082dyPvRcH27S$h>UqKWXG3W}TDUu_*sdOod?<6o&=40BrdOn{HBC;*uk$rlf
z+4zL62f{tJ%@pCoP-dnhpd4tivt%=#b?RdDyIr8u3JfKYsf!Sj!|?}Xuf`=xM6CZT
zbkvBSjp6I`zYe~N91i^XM&Qs2xo;fh+Mc?{2Xz)+(33*7C3MrEUOf`CP7}!3qg0Y;
zLWYER#zjC;A*gA>voz9qmuR0CdI|8@zj<w1#(S@D6L}MReF+Z#|6*DM;}qBf2mzST
zgZL{E`fkVsJ7Q`R74Ku6!Lj=NTiUXL?oaO5CsXQ{MGXkChTy&eMJ<R9T@K)<39@Yn
zYlZDDo6f3vghXAr)?I9&y$K-A_(lIs#qzP$rA3I+fzL+v3!=^vSXAzxh%4BV;y?Iz
zyFofHqoYPB3B(R+*Tz6bJbVrpSr`Fcv0pj)po8_{`|n%3F14eutC<8L$AgWTs4Wmb
zPzq$J;X@A--S^NLiF}wwEh5yK?mt@K)|A!gyHvv25m)AGxmJXYw2Ky8f8hN1*qzqP
z_vg?4EbMspsp-UQKz*X)Y2pwy)|gO-A&2e?n#Gg-%`9!9$f4Q7ZR;2}-&j^{s*ii>
zBfbg^e~1#mcSNAmG6Zkgn<(&{2xR-++GWb2J?(Bj-Zmk)zJ+q0O^i?J=}z|^Aoto+
zm1>Y`X+6_G;Siwy2d6SbYSS|wDRbwl+Cw`<ri)LS`JElo8ibr9d;<;(Z?oxP6a~dL
z-f-f1vzFKv(81SSogEu3*=uilX0R6y+M}mX7P*C01wK+8&+d)m6rCOXoJB+|2Io4)
zpNRN7nNqdku}Gd66e7&gBakgse7%R*(56j3_q}gIjWv~+5ZZgbNc?&ucTX$kJc`!W
zoUBO4GDDvs3o-ld!HD+_!K?FVG-Ux}gyI8D;j9KlVg@5!if1v85CK%f8n(o|C>=$j
z)C^r)N?~6kYjys$E<_#3K6O`Pv^a$Y-oCi9x_A2Xp<1n3o*7JzufIo<F`258T%mj8
zd6aI6N>xew?YlJ-ewg?(c<L`WuijUmXb{FPAAYa>1?(4G@t~W5yB;tJ_-~-R2%<sA
zks#={{*7yUd%<w9p^qU1{nlF$UUHCqhyCif(Pg(RGQ)U`A7D0+JHQSiz;L%n<3TnH
zu_sXIcS1bXyMa#z{p}H1d=%<htIKl2T+-s7-WGZl{Ni_yiX#ysJZU;2waN(Mh>z#z
zrAy&^W2=OQpAb!~syf(_3D1sc*_@@Vq)+nos*~n|SUosQE-|2L<Hmxg>x-8%#GR?%
z2KCW5N7ZTdBor6NBJaiBiMD^8&erGd>1KomgDiVdJ`T;Q?NWE`dqD8_tvw){Vd{mi
z@aVbJT69_D#k6|XU;HY;WQ^<-NzNBbTIN$kDOKV~8GD9(&&%Z`>B~DJtBUld#i}_Y
z&3EJ&rAGW+Ohx_DWYyg^GCV*1u*SbI<NKR*EkLDoHn6%gU$H(fG19JQ;P+rKzrx7x
z`DERyIkWz6I8TFynT?J&UBwaJv2|2K&S{w;Qm&WOg1`N2_f2xc34LX__4uK(>dv6e
zafL1W2v0BmVkr+coz)*|EQUlFOR@*txWooq+?ZW4t61B6iO-hl3-srWzD9rZv0wkl
zF*GdpO3=df+__K5+KnXmri};#)avI8ku#s)nq_48xaTPA*E=v7nUX1Bzn>K}AyZo3
zu$9wauw^Ta@TKC)p04Vx&Bml$@6BG=QAjrujZUu>tCp}wcB!HjEe_k>Y<;}dAr~D<
zohORBwma-LDoYn|?O4cb8@5X93YEbe<O5VlH%L`;Iji!`A$K-^V<J8zL_4rh6!oKh
zc`i$t^P_#yJUAkLvFn<BN>ee}&-9WaJ1KwC6q7I`3tX-Ve6G6^aYbV#wUw^*93?h6
zs#Et4xd0=JBZasL^0ae4s*<ddl0QJ>nL?5%UA^((5aVim&TA?&L%z7w7Mf#(*a4zW
z!KY2?uQb^f&+soNflm-_f5+w6ly0@g-4iXpox|@>DeY_lxon;gwvT_X`!8&MHGsGH
z-HqBdh!X}h7#sqPfj<b{H(}Mrf0iDBMhv2;(tC;nu1o^LS24tXr6l{Hae|IufIyHS
zGwGZ(C{~xoU1Vj2b*9r6cgMk<lieq)qmpBA`cCC1m+y0@GJ8S`AGiBDzFMri_TpWd
zVC?v!<`{(f()<lQpn7Yh#-N@D4H9AQ1zR|*_yo9l>W4)Hs`9TBW5U)m{+NC*YB;tK
zS4#*uU2tJBp0_`q(0-e4x4)jXpW|Agum-dC?u*mv3}!8tziW0F?5_ULgU}$LwUmod
zW%W??{^A+fzq&1(d6x45aE6~&w?5r~$5@^xiosaaJy5xWm+gJc3`UQq@|Wl>Jkba;
zUH_$7MpvVR*fWOG>FNMS)2vf6L|xqx`F&@R%|%4UKRDz!8nJplM1D95_lc?bjQPj=
zpVr<Sf3r~dE&iq8)r8voDJl?{*RA*dIK27I@51ErJ@I>7b`z>6C(Z<Mv#z2W=MMuZ
zFt<Ao%_!4jor6V3ye7;@V&eGK-kk|jld|^d^-D-PoNT`?<?Z-$H>paFU3GK#((!bI
z-Qz*%5qx^k_mA8b`zZAxj9&#=+6%2GA!NeKZb*qcSX8}3asEfgZ4OEbMuRb#iAN9o
z?8WW(uJbN^yk&M?<oh;O4y%BRPv}$}QT^Sx!FFF<;B<@LW#1F6YeqXVIjG|DuH5zb
z&c>6!-<b@U6+idfPaFtTpWsXuW9rO8I0|Ut)-Jo4-z|E`7?IQ!m{iii=aZB{G$&w-
zidlVyY3=UEGeqbaIEm`ZN6wO%8nik*Be4@;ARZRwtz$3~v(1U<wkse}6t;gBW2WI?
zi;8%FtV0e{KKq8&Agj`PdD%rG@Og0KaodvhE&!)9c<~qD4`tHZ4!yA`q(vs2u=!T`
zNONG_6t=LS*8cl^se;=!h%Q+KqGE_#jH-`}4RC~JAPX64z1LLpm-22BQOOh3P}vD3
z-n8spF>zO-maj>=@sii8#g-9w|7#GU{jl~ZhsZ*60B(ub)`oANK#vYD7!KhI0bUs1
z4Zf3O(Ba!9m66!Kepez4uZkYyX;^|t@B>~%US!Kc0Yvh;%fa3>SRYWk*B_$VLD)z4
zZ|a4)0z^WDn0bd27sVPBb1<dO)og_O2cA0UTn^B&lw$xR4Gu}%*69tQNcP-spidMj
z-i+1(od$fk+H3lMA6$i0XOhU9SpKJRlXTSpD_wixpSx3e6g>?;Tbi`^n3IGEd5njG
znBo`@4YAfG+6im=C%;(Xg<qs047|Tc;8z|vF<oa>Cm{;>WuVNQvRoJ!R!l29g2vIR
zC0F~>U6ISTbKSQRys0L|bNUH}Ki=`#grGoeSy4_4xt=Ns?sTshg7m;6<d#}q4|E{w
zr<WJ}8U(qKy(a)@DY;{Nm|ETyzd18ViBv<ZMk~EuJ&7ne$)4AUo&Qm;X**IV8=lvo
z=XKUeCANvl%kfq{AI5!aLHgaZn$W%XduX9~_x*!VE@tQE!g|b@f%(}zJKf#WQ_g-f
z*8Qd@BWmY@zbiQ2UL{e<it6h8@IdNA^1LF`Zj<B7Qz<=F8dM}qmU@DWY_sC_&#aP@
zk)88F86V!h;%0MRFd<KUXiv#X*G0nL<3Aaf>4DD5Ul{*>T>Vw<W+C2x9@A9%>Cr0_
ze2(#C>%ANEqx&QoA@?$bGH~^$X(h$7G&<QF#Eny&k1@Dld1BGudNHp=OC@?pM9DRX
z&M7mpZPM^k+&f`RkycbpHn&La%*&I;ARVD_(`ZsO@&ThAapVJ}sVGtURR+7P)^58r
zjc&0tsZ<MuMZsqYDDHNh-eQh5DNj9$8H`EO$l|vyzypB4$i^L&+NF}^-jjiNj9`Bs
zZ<(s;kdO0LC5(8t3ksQfE~>JPuMPfb3E7MP^R?wy2H{~4<0UyVc|oijx_abuMVc36
zI@6G`8x!K~|HTWQ*2xO<|D+#|Kf@M5h8SF~T}zyXfSEX_>buZwKwq=y|J+Ki%Zk8?
zgOrCAa({vOVS!0v`)(bF@3NpvGeD6;%ZsaSke0gKnwJd89A#eyIvzvk4}c;8$NY!2
zgS`14+`eB)yqs|PwRHz@kDz3ln8*p+cw<)uFk>rRG?(K$p@T^j<&8P)vsHCD5eJKx
z*ahjL5ODMfu`>})8a@8jrSYPD+5Vdxj*yoLuj-NS@nToYj>G4l!}U)8qph$Vk8}zg
z@QR8-XfZ8@IG4{Hgn{E<2V@A2%@Wp!anSca{ld<m9$}`7mA61>5bKZ}Z=`D%jDcRB
z)@mdIM=2xbs%gQlCs*M3gqBwSLR2Hh_Z<6x(xLQoI>F!#>aU;qDr+C_w7>c3W&Zk}
z?Yb)KR`g665*d8YEq@&l$s2Q2yDtzz;Sxg*@*a}=jTQ9t?z!Gh?R7oS<Yyp@3{<Vo
zS0GDGC(h9mrS#SHdaN<r8Vc>KrOI4!+pxDvq4%oJKhU(}&FUDYw|>8_DsDDfeKPIS
zo$x8~xNk4wQ>h24Wre0LpL_hUgM<s;K;8Kv8g=mrjC)d5&E-iEZ%Npe$wN|mBOH*A
z1&G>bvC#iJ<IgGG-|%~UJ_1*Ci&KG3Bt+lCXve)Nbv4zpEo!ZCyMU_X;_^xmPnX4Y
z8}8#obFZz;@!U&wAqb3Bsx?A}nv^1~frrU~Q<zvbE`}jX2KVTS+|9JWrMNb18Ly)N
zDYMr64?PPuV<E?)-lBjjC+tcYZG)&IUy#T#*p+D^rB8hot=GL1Q+-AB1e)_59C?Aa
zIU1~DpXI7t#Ao<f8K+8u8DY<l;~h+@z&as*f76wRIDOgGJSr`EbFbLXm2l{ht);}%
zM-k{?vi6$AKM0SKZ=8?-k{N30nIb(jI4s%%8v7)~OSyfomRuFj{s7DKZU{fNZqZAE
zUA*nk>)1-^NVNUF4Ufc;IK${~>ysh*YBO#z<dmz$KY1pYit5H&t$yZiZQbP|ETe^E
z4gAsaM=0u${sF!ugEcU3HZy-}sDMEgM&<gkuWM^`P+{THX`h=DxKaDqgrL=Ec4&~l
zo1pd6r*UGCUR*9)pH>3zMb8^U_6U+ydFM>aNLvFK3U^?h=AE1KeU_7!(Ni3TaA>f!
zs~)kf9|KzJnnaIh_J1&)J$|hv3@87PcC=J<N{U<Yu_!Ql5+GU4mZ9_EQt!NFHNvIP
zr716k&)2dulYBGNc=lrMp}Qit^4j}Nr5&?*t|a~8%#wnfESb(O((KcvPW;nVCUiS>
z&%CHlVkvMZ=F=>@MWS8HERnmI@-A)ZP~9Pi`Vf)M2BB?H>5x_UrBE&^p@n4*d@#+m
zZX&!&71wM?&OzK5y_WLL7h#ww{XCxB=T&D<V6Ys<OSTF@c^9|Hl@<0j3wi?4XCoGd
zZC}jn|CkTm<8)Qa#JX3eIAL(c7jpFIf3_-x^ZJO^o$S73pr{mhdy;cHqTO7zf#12(
zmoudr2VBrb2q>=gf^8jS<ZdLKlTgkeP72PqXZa~%M&A4@&^Y9+jnSaZ&!ofG(mMm=
zVWLqTiQg8p-PYf7==WcTWE&(}Sxl(khw2N{Ej<QYTy~0F5ma%SqCHongr-9QW97<S
zn@h^z`7B0W`G&i1cC3k}D!SE_mnefMD&v?Yug|4)U08Je4qcN5eG0nmveV4FjPYD8
zJ=+krBj09u3Mw>RXL|NHBRXq5RbNLjZuR*;)KP>im1tV6dH-@qeOPPLr=TSx>n4Ev
z<o_0@ri+<_A<26k^qsy9n@S1EPqG$6&Z(7mE#BvG^woq7t5807HS=v@;m4Ctm3EM~
z^wuu)4#kbdBmFy+S76x*tc}E1T@gUz1Fjt2n~l!<UO??@tojc$`HwOb&`o%mmbKHv
zRw;#J?&lXiK;r@act9MfyCEVTD`~a)yTx$6L77Gqa7plfzzeAN1NrP7^sxgcSGnxL
zY*XL?fEfjMF?<F5#j62CN>fzXedjd0HyH%gyMaX7^+0WoIls2H=tmu_T{p@)qCicL
zLx2<yO|6e6>$$tpzRbKd8Znn1pVTIlW}5i#+q(n=CQ!+xcI>G<2gO?o24tg|#@!rJ
zj*hrwEnk&E#4}^5k*G|rPP6B|LB@^e+0j1pvlLe8(-Nz*!ud8Lh6kVA1Sp?B_kB=r
zJeix<rv20YOJ08>rO*$*7b?tp5-X*A7;&Th`Uc6|Qh)t`1k*)7o1DTL94e1fubhI2
z<;_9y&(+D*idzSnWA~y&E3Ur0Ooh;uw`@N%!eThWn)V=k<kg<{wERv<?JsC>nx~4=
z8+~E}fRV1zBb9#KgSD?*-$0jmVSJ6A`<xWLrkbEFfz4B~eX^-2?895)?|W)cqRaRi
zWK)uM=w5RjKqPMwE7D=GT$ziU91A^<D`mFmg+p^qA8yDHb)y0z9cQldzU~b}n-hOH
z$u&#nzTcNrYh_gT?VwhZ6C?K93L(5_`;|U6i~eMNqDble)&W0<BzEqU%f6_bI^LR#
zs8mGMtWNI0y}?KQmv|-?@H2I6aXN~w$&ZzV_9~6}c6dfc?e2W`P=u-s^*KMlXc&1X
z))=K{nJR!dzRG`t1NN`|xOukt58Yx0FDz-uNezuaDT?onr*<X7b~ZvE{u*68H1|gf
zG0koBm+|lGCx5?zC%I5tE#Ui?C=h}C4~Pc|o=_o|vNvW6-4#8Gcu)u{NgB<GW(H*K
zwbM7rDL31H0f(dRvbzSb;Lvh|LSMoOZNnecW+vKW{72{t0kHw_gW)is@ynRfa!<yO
zOkj=(;!eQbv-=3{JOaUgG(-~67W)&#-B?NeSqk<z-a_GqNEO0PHJD2~l%_x@GMFL0
z(I>pwc#1}oc9+qA+_Qd*L*hHpdqDl6(nJ3PimQ)1n@%*oX{F?+7RRm2zF>!#@m|Le
ziaJ41m}{>QN<AU28LX#$eGkI2dqGTv_+H=9|Kk{0+P_$n<03m2ygHmIP^eQM)c0y%
z`j*QXd#p_u9-9up-@~O^Z57b<Lz-n1>eVCXnAz#N(NqG?48AL29V2CP8=gHS_NU(H
zvE6yf#hx50gnmTCp(#*vuXFx7y7RNsmzK2cs4kPQcC><~#P((HyoLK;Nv7j6>0;}-
z%R)?~Muy%``rwnEmFnG1@w{~Vde(WD(t&_eT5*0VC`^-=YNcg*!q?B1oSxQfG?s{C
z)OR?aT3%6|`glHKAilLRar!alSm()~bq;Df<KDiEcE@TCuP>kUp0;Pv`(3`g@@M7a
zM9UB4Hcr@CwJWeYcx|Mpv;6;X_0~~QuiyLk%+N?li^QOkQX&$Hq$nWWAe|yehX?~R
zbO{JUw^B+E0@5Jm(2BH#z`#L5k&=Gy!O!#k<HwqHTt@|Qx!?Efd++OdT{)>=iG9J$
zZ(2{s{`akH>f|L$f!oE_6MJbxH7^?pl@0okDTUuw^{(`)>=KJThxQ%)$fZaK*GW+B
z|4d;plyc{4(i`{hzCPkRJCI5e+I<*~1k&uk(tjir{!*68ekl2v)PKJBqfY^S*?qC8
z$OnB{CCj^nOCt8t@N4qP$v+C5-G&v0<+*aOXnqKxsvH?-FufKd%{!7B2<FjOT0=WX
zwN_+uy9RHqIuys2aF?%_pUQ?SOpibIm8=4kxBI{*S0^BKH)?O6L{<+7pg&UR_{^f$
zMwNFyYp0tg+t^l#ZjM~E-6$RG{43$J2SB1%Pr(-i1-d&`KFWJzAaIUTE6KD|Jd+YQ
z-zy$hnuxQd;gbqzJ{LBaSAuE-_NrMIpabwj&pV6XvDLZ}o@5-)$w|xO>M@x;Xb((J
z-}3$dh#s!F_vr2eEuIcdaP9nGmH}R8`j0V+7qNgVYHHyUL(E>GYUHvY5qF8izwM_>
zJzjpY>2v(e3Sa2v;K&8nPSDi@MRoS;WSt`@2HZq}lXbpI6hsYzl@-e?Ad?nzGc6ta
zfETPENf6&>RiTD|`9-2ir|{prM3Ex?tqIIZ=h}iQIr5qTR|=H;7I64WE&R)+DO)Z1
zxll|HOhgvP^#F1?a``BffcFt!0c$7_j5X%OeIm_vVS4+8x$_HYNQX!mdM5Jcok+Nx
zGfEGMzEz+XsY^gMqb)^{rY>o$O3($P_-$mDr<(DaLEBBV+>7YCaP0=&XUiN+Tc&FM
zRP+R@KMN(xm1RoUkfN9wKcPAH2WOXk2&AF%-5x|{hNvXjWT!uY6N{#G1n;9w^D*9A
zb;e_N{DJ~LRFbt+BFdaRwD0rtem1Mg6Rxiq-afdsFl!n_RDn(I9UDyzmgq-~lpY*(
z^!v1Lcd-5P$h#*<QXABS{pDyZ8zt;iQK9pGcaoyL@nDg_+OwEy*wvkmWzcd#KsY4s
zww!TaKn5;{!b37#<2q?Ndry>IS+V*fI~ckgv#mkUtCxawe0hu{lBdhdqWMhBn*C{X
z<q+6N3f%)GdV3^ta;gV=%FD_S!y5^Yuj_MK**O>pKQXAZp4XRLE{kq+%URn`-PvSU
zvYWNdCp9scK1qYypT-gu5|F;;;)f*?<>{TJrS!-Z!U|YFypi>!le@B3QK0}OBi(#M
zQ3%E?3h&vF*$`{q4M>*)H_`*XM|)WU%QV#r!8hHXL(%nX5fB^~Elq<}Ks`k4HF6+R
z43Qf6=7ZqNL?qu;{Drex^nCwi0|-z{)nxk=lj=8`(=G#<R!VpIn^d!5U-Jcmq)}-5
z$&3s~V;cmbuTheX@EmHh&oIHi)_qog)qV60JSu>}2gWH7u%ySwr-5zL1sZ&=rH1vB
zu|>pZDd0QAZ=S%~2)F`r&i=Ra2W<|p_8q};TiyD4ImVfUJV-eH#aIDo5bz3uOXTL}
z^P!7^Y`p;xXTx8BPDQbo@F$SDZ00$)I@(?E>up4s0sZ0$wfsYE!a_80+!`qNP{%n|
z9=Y_P;}wE5fD{0l8lbp7K>W)jxv;(x@)i16th8OeIN4vyV65)u|Co#=P|Me~f|S4v
zGQx?%<Y-7K`6A?Bw%m!P2?xi#w>p*LgIbfslu!Nmzwr;6B`je`2WbnKC%I6+dI-iZ
zoGIi@yBYi~?B@5lA4f?NrQ9iD<*~EA!sF7{<FrZ^*as0J*!}ZSfIR61D)$=pp&CPq
zd#`lvmInI6_5(}IlZURdBy#WdkB*E)8p;9~M5aj6i^6X1{W6=Y9K6#&d!wZ?etE`s
zc01t^v2+EF2%&&>`KR8W53g_bjSclBnO|Lg;&ocf-|S@1hl@jCMbTZxb#eXsbo>Wt
z3*N?=cKrzv@q2xOllIv@=-a=X`gNZ$36@yU2QjpN<6rb#9c&=T6#ArM0wZ%+!z#)U
zp4lTYn>l50yE2qH*N!WDq+J$lkXua+(cxnt4m2|=cG4r&Fk2QyXSo4mW0M{Vt?>km
zb`R@(UtF7Na_V5xB5vy8yY~}Cg~L+yC9~9oXgfYHvv*FzfAm74`t=y>_xD#&rD2s-
zIy7@3KE#4L44e?MLQOMUbOr51%dS1k4SH+R%Zh~>*mh#AjJNcv&k4JTRiB6IT;;4r
zwQDGEB$#kDU5>inKYSz&tS<ctWPHC8I(l#3pu&1DjXr=0NsMX|jEv*lr{l=@@bt#@
zu;^R^mEDN?woiBxP)gG(Mkbz=o;Yt7yBoi^JJtiPot`TqLa|A=MdgEEuTd;YqBAOs
zB!uBrNqp;`QFV_!{+bKn-P-qo90mxpfLOpa-XfoLVdEl$4ZqREZ})+)1+Pp5bV)E^
z1k(<=+N+wNMJWKdD?mmCih_WEGH4^9OTtOPL$iC82LHYBJ&?#nZTvd9m^j|?k_Kma
z<{+59<FhE8#X&m*F2q1y8~D2#M0_#ffCK^bH0ObD>+5LTnrB_g=XnTe={`Ek(5$6g
z*J=R%htyR7g8>Z?_%G1S&giKAI67h}9s5c4FzupyAGhH@oZ&*!^mxH{=f$RYl-u%)
zU#}4D9#9V!2fYPH3~Pa_SBWEBiXv&fXt_$QZI8H<^XOFmXWn<EmN~GAiY)B%VRA}c
zmQ0Dd?U4oTH^?Cnbz<&wcM|ZG94hKhdo8Ee7HB{ow$<=3Nvbdf<sZ>oB|JMYPx8(n
zYL&+D8^1S|r)_P&zPI`rvxw9?bLf6KA>ps6nrHX?7c7hpIjFcZ4=Zk#Q*&3)=8n}n
zbdQ{sRFM!@EM#m{rm8<J9=evtc4(&}XQ$ajv|dgzBs9}Du<tSQizH;^7;0m2&qW~O
z(@v}ObolqpzBKo`<i!^r{fQ|<?PH6Z)ldBbRv9NwC8-@)aM4+|>b;h(9K^rhV|LhZ
z=~O*aPsr%$QN5~;a>hjy6!hruIGc1c8@QH~-&^F@RXbkh&Hjz7n{*EZ2+-Li5wXW*
z_EJn*8X4LwLgRJ=^1H9HJ$RZ4OeO35-@iEO4_W4zDQTfh_KG8r`=L6rlWz|nlXGb@
zJ)9@@huYnA;!qRB*d|x?+BTG7gwtIN@R+UNaYgy-s=T$KI*%cjAeZvps24cX4KDFE
z(36)E>T%oa{7q9KlEa4dr0u-}rh`<@V6-np;Rf;#3-%}~0o6x&!}1URY1t@-yTEzN
z3!0{hzY3-b%3Oj#03Mhz$nL16m?{K!*(`v|C@Iel(z6budokrDczwdQk=LS-nsMDT
zrmhLJ8K83p+Z?dL1YI<!IB?kn0x_SyW7P)|yPzI{QU*`b0ErQxj{yz^0710`$MKh^
z(%vNk<0Mf-HQ=NKGk2g@#5dN*2fy%O4#~LMo4id_RYqmSAPxXLAA2un8>N7vd#@Ob
zA9r@aK*kNoS;iEA;S5X`!E6Vay(<9-#B!ArAT#q3AYBI}2v8`V(j*=>>IWy3XBzW;
zb4~&GYCv&2QWpVV5m<Q-W}k||Lq}Xjz_1;E*=TE9ZDY7N?vGTJCZv^UCf%w@KV0>&
zTu#SfewD?N1``@-oW}v}5Hm5$11jCy5xlXY4mf&zlq0yrGE+SI&&wYyUrWZEXh5(M
zz>8#psB>Km8Rz4|Y_ts@-6R&0gx%Ke^B5_jU4F7?X(~P*hCoC8{4GcBkKO!+2oP!D
zEmD+Q>gCt8Y)LjC*Oz6_I^k;K{zZ^<Y346Pa>`;N$B<$1W=ImkTS+xN8HH_-EN#a=
zy#G455BIy`NoC$I#NPMq9l5Wq;(|s8A;-rrfci0uka8t{I1KKglEo4%{M5mssjtB=
zo7(q0@X~U}azN{C;hnff`<6EmPP+|4h2DcC?qac7b?OV%==4zx8E?O2Q%YXmZu3}<
z-t=19P$^c*;MszaEWItV*F8&pREmXAVcqVwS<2v~PSrg#M%Z11o9L@__M7vL^JT7F
zOjeDDU;Rc*jSEY8>&@szekNz<d$UrFYzY77bNZ#Pb1K=1G39>1)WJbA+Iy7fc^Iu+
zS!e=H)az3EN%_L8f=}d8?(2$24VqBgy?i&4!mY4R1ZWe|FfLjbk5YOcq({06S7;tv
zuItkPC^+FSjo<?-hCpBi%GP1(n$-1(o?Z|;t<pdwTd?8Jk9@~^F+*1kG}ve=*gM(#
zXSh+dHBaM`l;bNzLDozkSpH4hy93bc;7H6*2iirj^S3ztX^0oZoSwGe_}hQakM^zZ
z1EUKNJp#r$-qZz39QYRSt$I!lNUZ?PjL8Un%wBEOf}{t7S-AmTdVt@y0uC%71p_Vc
zZGm23306}erDT{`_om#XickcWb1-*7w8{W52w=B?RC^ymb}+82pNUe`CTAR-F$2+6
zfW5g!kWl25a`+^ybx@ODsV?L-8s8<Ui_qU|kRAPlN6HETwo<^stS=t<;nQ@-G(Orj
z=bu&LNc6#od!DQnjGn~mCo&61=f)c9bJ+flBNPfV6Y8h#?7C&#<}L(?-%9UlJVJTN
zwOxUuvLJ}s`)<kDRnkvjK$E#i1rbYAYfl!-v-8k5Kx*(j<%gnU&c%T$k;ksoLYL1|
zLGb4X{z;Rsp-H~XS*9zb>oKkt((!8<rUpJ7BA4iU&rE@`x1)a!=D+yYO+?CKo+DS}
z&vT1GFq+sL?rqMfjrqw2<IeORJe>&(JMB+TzO}wl?wjX?x%!dj6Z<HE+*A58{~B|c
z<MGVF#7_Kub-rx<)~&(a6QAPP=kIN2Z)(o4q*VB*?L6OEx&P^!$kbgnzae#@G*fZG
zWazKEZM|!(PX~*sw*YZ%W2m><y9Eiie#!gU%dp6^ib$9)A*+$c>BZFNx2URV?!3E#
zm#!1D+%)UC9r&%)cjP{QqjhzqSc8P2KD+(?!G|_=T=s{Vzq<?Dy(y0$-}9y}HK7n|
z>fMU2CQv0$m%}P_k%P$ZY}!ev^`)>{T5HOua(>W4pa4$%MN;_9jMu#wPNya1x|LzW
zqmQYkQ%sNk_2Ygrz<PK*aL`3j=)`?>9aZroYfJLD@-I1?NifhH_iq?J{-}i1v`Ct9
z-ARi%oaN*m;1YoZMQ~?}2bn9Nb1<*mbLzMPIq?9-h6{KR9y{~D@gO*n0oCK*^u$q;
zpQ5vvD!u#2^0lI)M~Eq~%i^IjhHXa-fqSJEKR|AZ|GYU~9G<HK&Q1WP0kHQbZfarz
z1Ut9GB6vuEWd{+r1yUa`u$aI~4vQ#69|fFquDZ{^atcPjd)|138NhIofnFULApvjx
z=(N?xf9s+v0t%J8D|n|ln6O-g9}qXq#+{qn;uZGmmr{tM%j-MB(wC}p4pvFzky@*d
zSX8OJoA$4+kg8F=RSp5Oe$MwwB}&+NcCh2z^dXsbAsO_!heC(YRO@NYlrIrGe!NBy
z?&Oh3-0q?7qt1owY+o)ov2ydUa%Xb8$9%2gt8=aR&8Evi@?g|GG^A85{b*uYal(c{
zxpiP~a(*=@jUR~(dtP|Ra#9X@i)`Q9X1z-w3d3V{;n{XTEVx~^m=@$#wP9>D4+#j3
zGs4O5yLlv5s{UC1d3=_GxHdQ;QhHllg_X!Vdw#6dvGn<E^(*u2pLF;MT=r4@L!IK^
zV##=$hB~!I+78xa-!u0Ta`5bYeL?d%G+ezz66-l!S=WS7n}3$q|Li|Wr%c%J5YB#M
z?ax)00Y;?y+jYOc%JIkSq^9w?4zY-vn>0KY6NLv=4)rnC-b?%?r>JGG4({}uhF@zd
zpQ<!0o)WmlU1i8F8pOyTXP?gu=v;GGH&EC_(>x({PHs0KtA^?v*rJkyO~@>vVI)L*
za5y{UJHbb$D5pO#a%vqcN&fFV(LBFb%^f$Zle#qdZPn~CVrn`vQJ`iOcJYH0$Ra+P
zPX#K<V;78e<ni?VAgUz=8O5)1_N<j;Vdh6F3l4_!iw9>L@!L;$+Rnb^SFx*5@YJ+&
ze<8{Pa|_T-fC2_{YP=}1)&#iB@eu*lMwShxGTt8r6v2$qyb|!W@E1nlb(lY%JHYE@
zLGHHI-Rx&oz!DC!`?j~E_Qs-J9s(K&=t}@41o(4+s|+tJ-^E^3a?#^zvSy&&`e%or
z+}0@8rMw0LuYgYne-M9X0hHH+<A3_aInNJ%*AwTSZdb2*yw@D@q#=Y)tblGEbP$DI
zfJO#hz!x{N2edUfTp@8a@F9V9EdEi%YuiD@Wc>`=e|eqY&jlsz$?Nw0bqHMQfCK<;
z)qeco9gyOPuv$T?jd+yoD45B@kx~?tFQj;%UjIt{2#lD(_bB$HeA%_=;cdV55SlfD
zojwmoHA@z|4n49mXP7pzSXgKNa)HKPTSKs=<`j8Fo*g3PQ`*CRyXWkST3H_q*ocLk
ziUrzA;(kGw727vc|I0AtTR}VzPyiD3g%6irXTYb<8kKH8G_&d)cKM>%ZhKqM^P$1R
zxwDrIyYDSA>c0?R44>;zuTyJ|fd$mY`f>6ruCZAArp%jDclVklI~#><Z#_9LvZ+pA
zfhCU*TdS9|s8r4H1t-<)Z?5RQRR1K+@e@P!v@9>rnn!)5U&S%!QQ_7j%_lgn#Wc8e
zG&JQH;{msI{3v#FQ0*qAI9>Y&tdE~fIC_dzX}+Kd=3%_myoxHiwr1~n;+%fZ?eI?G
z8z)|UV0A9ipSJVM;wXx5Al7RS@2rPl#h2Fm)oz~ocd^`rrnh4!`SR;jBg<(gA=ZV2
z<Eoagz7^=><%U|+5O#<(Z;Aiwk9@*;-(c;cM9J96JJ2i?CA$0Z7Q~seiU1B}==`yA
zm^eKO;!cIt@SG1=GFb5_5cAPxIcsEU5`NSq5XtSi_*X^FA0BFMpMK#R$~Y~i3d0Y|
z!5|J)GiZ-57Jc2;`OxIOmxJwCy8dQf7D6ZxG67G;zeEer(E{)*p6CIjBcP|YzunOA
zNP^h7Y!&Vi^b4GlR1qws;0}o&7gFnel;FkJkqtw5zDzx>i0S`v5jgR<&I^TN92ofl
z3it2~cN&;=UxM1f{pFT)`vYh~A4pJv%|5Vd$joVQ(}6t@7<WLflnEAsUDPtk8}NmJ
zaQ4%U&??h}J6WuRk#9_3+>f+j{FyGj+Gt#a#!Kd@DhyPtDQ*DkmC$gDM4ni_IKcxf
zhHAb%C<nK`he*;q`l=W%wCdc5+=$*{t8zl3h-LAO?1duRse{--_TIri?z6fn@SYZp
zuHzpwe-cjmf6I@!DS=|FC@e&vgVAI1+oJz@xL##gSblHm@#Si^Oqdf!xFiH(K415T
ztYaCmoh)B~w)n{Y9*NCKJ8=IwW|kB)Q*NTkw+Bx3j8wMOlhb7AiQT#?+;OJ9Epz+*
zcMQ(IUlZp^#x6X~(O0+rwR(Tg#EwG(;*~oPk^J1jW=38@c8Ns5EW)<l3@&xvu=#m+
zQ)%i@vq-3h_yxsXS9`bW5A8N~V-jo$WW;fHS;3#KP&~gcSztTO$T}%?kK-m^4W?1{
zKMBr{ZlZ!UIPzB;RyT70<T-8cS^7%%M`~?~XLHe~>=YV&k#YE;y2db@f`huRguk2o
z)L<QxV|RZ6gSeHK^{eG+7U9=fPIt{X&MN)(hw{!0ReCfxA&-+YYWO3GkVJ&+5CJr;
zFU>Pf`YotH=qCkU0ir5eopQLKq!ctM8JCrHD~Y^1_?o97k=!=!Lot(LpGBUImQ?=n
zZ5}%&jk$JW^k6aUhV#GuTp3g4|C5ml=Es+k@H-lOswsG@S0Duf@Xc^vko%24k3r+Q
zanN<`MWoCA9Tv)eR<g};<#{l&#$yt$!MY_hjLIstfeIGve*miH9~l`4%jbV7@`Zut
zKsXM}S;jf}U$&!$Ite!ukLSS!4pcwrc)?xqA)v{)0hu;f9X;bM0qR+U(lyTpvU@>}
zq8)a0RqRv|K*|cDiGVW$kG&O|2UZP$8+oUQN8dB9E5)2u05UQ`AGS7h7rH{vtC6=%
zqphciJ~N{Wzx>9eAm8@peNnhQ<?Tnr3t6aMLSB-0o}MtC2pW_~JC4wNyrOSyUKxao
zKqCtuWPP$+{`fSQ2RrYSUSmUCL4QvNM)Il5Ag~+5E-|8bv=h?apfk5GL#U=<a<gK!
z4eezlYX>&t;!j{^ThHIZ($o+eArqxS{W25Fs+)SS7R2*3#8yPoY)c`1<6y%qV{bi^
zsgUQ}{kUBIfl3PIPH&k&3K6N2xs^vq?AW5$blJ}gQ;fOiOu}cK_w88zRL=!bqn%d7
zCpW|Wu4k#bLya>*%|2{No`)i{-iSkSu{B_$=Xdb-&KT%3;b4Yi68ZA7#NOB%2&IU6
zavM~HV1m)r;TdF4!*wFxQ4$1*v%N0-5^<wlDlDnC2?G^ZOdW&(w+5IA;Gly{d3KoP
zufOsuI>c?v3BPT%jhp)1Z+!ZMH@>G${l#0UWs?{3Q+-HGrHhqe#0TvUa{-T%`qquQ
z6}nYP$x!9A&n`s*=@X4n2Vo)|@1glSFY_N%?FxK*aC9S6dx(}1)QG$=MZx+yy*R(k
z^?ER%R?-5o#%mu$e%KDG(^d|@ybA`@VL$O)k8F@I4W)eqd=2LJI5_f5mkTcT>#6)<
zh=~v8vHven34ij6d{Kvc96kZ}E^uiBEipiKy7~D<nUwE;vub@x#<}aH1%L`*HV3Gp
z;06c0JhlEyKqLZ&3dN)H^h)D{Dvaw=Dtux8=5fpvA2LD28!-fsWS)||z;FA;ZIl7F
zX%=)UV19;o^3r4g%@fdDnaBY9o2{rk&_%7<yeviO@I&2NwdAIfX%u8x0l)@iDdEwO
z|A*VVVC{jJ#5@jqJgW~7cg>U_89C!ZsVibQJ3As;xE>8$ZK&Jor)?Ijn~<G{+I)<5
z@?*+Jo<~-pRgC$JV`62_!PvWkVn3N(KqqW}8Ucm9?!NXQk&$@0g7v>fd1m=S>ex-6
zYIXu54_0-H!wM9a9BFcgk>1=P<==N5ww&+HJo%j7PTOUk_-Nx?v5KZi{hVu<Tap5G
z!;2ZFkUNH`fYP29ylJMj*$B5kHb|^XZ0n|(u^kmMo*HT8xh+jg;aqfR;6CMZG+BXF
zq4;LuE?~EQ3fhQ5c51b{32mXiPXbmnU9EO`%m*%`j>Z<|Leuw;Ke}%0OKmq!FnNkN
z0nPU}cI0<gdk*HGpN4f*FA0g;+?Fl!#}EfQnsoN4q)Wyn%~6yX_c7~SBTD1?a0x?p
z;|atoEA+qMKJs@qm~V&1rfbrtsXfXXy(=G0Y>dV1b1+IvnGxo*nAY83OitrmqCa#^
zw5q{m9=Ey5BKY5W!ovf|C<)O6Y*?CS;Ixd&m-^cG(0sHq#EPp*7e<yPP-Q?i@sY4A
zg!UuJZ^(iMm)4&aWMc>|HTlPgb_HH<YE)PvP}(rikJC<)T3)q{!*ZvwcCK6-%{Esl
zEWQoW)UBjjKu?URRa`7ktLVGBG;{J-&aS_XHMZp9ADRgA|0K79aTgPJqvXU>u@3|b
zplU&ksynD!;7G$0jH->K-3x?u3Hg%zughzLP@?~_Xh8AafscXtuabgC=m`s)mvNnj
zj1vrm?-WR9Yv3cjE(m$}lxsXvQ4KWV$x+`!_Ff(C3V|y9<SbspPLZKe3=}0ma(%>^
z4@P*Ptida=?R<ls9zbva7U{g@BPiJmXY-vh*4|s-0=M|hEb5=gqgoB3es74S>Q?CM
z2r*&K<X3m6q>&V)8f|ydSVN;m-RU{Wu}V@3$|9IE4Kp-bgdHr+`z0DRgo5%uNx+0B
z@fk#`-`MI~>(siYoTASMPa#*hQU98*eK0_&;mw3MkLTC;8gv5z2Voau%Iq&w)A5Iw
z(W~pd6GJu)+^JRh4QDY&?^q^QJ=80<yiO#aGsN$qZXs@@W$pUGvzi_Jsk3kc_aRjh
zt)lQc{<N%0haSxEovGA=h7Q~MSEf^n(_(?eDdpz{rtJUiD=r#eDMSZlR^P3eRYN?f
zXIq6Ly*rFnQBF1wBXs9nzJ`CsF-Xn4WQ)&^mP(s~CH2pE9sA6T?7focc9fe!p$pf1
zoZI_Ym^C^U2q7A?ue_T;uyw5<Cl^h>kz!X00VbAHx7|3sz2=6m-Hp>1))3xtTPPRi
zMLvbE-ThX(k?KBFMJwi^XT$5dwwQ@14uyjdTp-T#z4=Y4|1wlNIv4e!xR9gD<_$X=
zi^ADm3npbo;`ag%`fxXs*;xv{YE>nwMjDZAz`KZnxj^8xB&RUgfk%1%VME^I{jTut
z3FJT}<0TF3K+6bb6v=!^t~~V9-GpavY5z)vV~mxF3PYjrdyk73W+8}gch)@t7gyK_
z^u3)g_LDezk$Vr?>y~Zq@-_W);aFGeoZ|V6vVZ%}uYcpeD*}$SahuXFPw{C8fTj#s
zG<d-T*dBvzG8lgW;z#pH)GzBh>+b)DS^&{T>ns2BJx)zs0Ua=qchuTkSHzaB3QY5W
zeMUI?pJ*C9G$7;aFrOS$u>2bm_%pr$#okQNE02!q0`(~vE<bS<1)~HoFk|_;cHuf3
zrqbg6*Mx&BAn1%G<A4JKT-9E-grjNy#W?_v0{}prNL;(XpRhZ+O(0*~&YMuHS)IMR
z&!eLGJ({lAIyfSP9$w@Z{OV5LMtkoxXM<Z#k(0);740MVhZxLNe}H^uxC^a9b)Chz
zu#rHVA+ZVfW+cB12#5S`r2xW=c}@g^6imVvOtQ==#@Ja5^?y$=^{?W&^oH*Sa_%;K
zP(59ao`kQWEIlgDpzRe6XNUyniW8{ipVp57?tkTQ7;o_2WRb$-G4I!2pVx8!)DTw5
zSEobT;%xS}AGgwS>-fiVU%#cZ;;}XMtFy@Tz>Y{?!|$C0&mz&tw4eN*J%a~Gti3;Q
zkWsi0TgtT<UtLDV(tp2`9m}8>t;FB&XxyU{(k6Z*E9G{c{WICPLMY|q&G51;zjw3A
zuf$0OG-mo1>_@*$xoL=#E{e)c`QPs#z%u^%Fh#y=D5RYqQ_b|X+^Fv7-evuk!8c*D
z%c`-b2l^_Rt;?TIOJ6<+8>uHTHPGSZO5~aNTCR4Do-Zv+>dH-2_Sb<oMmOc%Feb1U
zNYEC1Byqb`@&r<^^;4t3r4PcW@WY&}46HaftH}#YIG3(-p~W5V44ROY7rnlG?eVan
zeqiQ;(^E}OV2lY!cN3N3sl3lIp$~%PcG$e|gI?64Z1ta`q=<vR*}3QR=b7hyL96c+
z!TYSRmcos=mSxxraN7a{*?$g40Av6?E*>K)el;-rz#)+a1~vo#mJ?uXHi{I5>R$K(
zR;OUv!}3912aM;1a~tnEUob}F$%_3RJI$cq_(vJXOEU1vfBbpN5Dmuz9>Kv1u7jXo
z02ee+vztW$fMhpsJ+kQoFgpck8il9rhYhV*xr?^r*8w=OKPci!=iuRSsaIqU$E%@+
z_VL#<>uP65h!n-`1PM}FxD;vP2N7=;D%xTzD@a?>6VX9}^cd0F%=S*Q5;Yokw1$(f
z?Pf(*zzoEb(u)2W39ihvge@a!|5~V3LIW>d0V^Q{^CHG}j#p!nH{G{s_mvG%@qB>E
z7-npt=Z*NjM_@za+4dog$dxKHw;k{1cZHkm#8(YT(o2J}5nWo8yZ4*RKZ*KJV+%Kl
zLJ;n0rwCAFyseAXk`x1=H}+v5go-W=U|yVV9+?Ho3go;8Se%hsD+i09K($GLXgJ5#
zZ)OYk@1&*B*E$xlHL7RqsHEDh*QhoH)|+xXa9-gVJ@n@woNBbIx$D#9<G)O~-h<AM
zsqn<8*}&VUj^yOK<3ccKve+`K1wtY5*~DbM$U@2+4huDF?O6+^)zCY1c@)XXd>)$O
zsAbkXftp)-e9tlj>ib^B*TBl$_fiE-WV;tOTQQ6Y=Je=dGym4E-Wt-cg3m@L<6_Vj
z1<@iqQ0fdi>9)2n)<OjyzA!B??oLLAB6Sp$RZjA0?j>I-;SPAyP7J#WHgS)6_YyQ@
zNej2~zqM$6B*JejV1+JWT?lev;n881cJeF7%mkC~mLd6P`)$BFKYOD#Wh3b5-ub2N
zw$s_5i0w3CJ)kc-d{a#GS-W9CWr~D%y_gi=7w&m<YrbFJ3p(AW-ewAS_`B|KKK5_U
zAOfgcgix>?=vk@`L>%}uDS=F+H>AJjS~@p31^=aL#u|oH%NT5hJ=1yn1?<bJSkd!<
zy*Qf`(a8zMv2P-S9tL5La?iJeBF|SX{&GlrGY>0(F&nsnflC}vBm?=|{yrX%cQooO
z?j95b@X;7>(NiPj9bWYW_dtNc17oz!9ME#?J)s0E5wN}>cKNTP9CH<5w7n>}4RL_!
zY+eaIwQ@GeG3TOjNGYcf)qpI*l^3iinMmZtHF2@h_p-C16bY|+@=<-F;Y1|$yH9JB
z<S&baCF&a5`pOW!-jrUtC)P%1nXxz?H^jhRFh}C>^4Ug9uGW~mghk!m$n~4LC?#(v
zL*nNLxJ%kZ)BC+z-{w7i)rAO~L*i$sni2YLf9FVan{YcrEgV$A2GFGbKe`?RxxNt6
zW?`WT{f{Hs4WBU24wfE4l4fy#P*{We+WE4RZ_enY-D+z%o(Sy_A&ecx;#%ERSCbu1
zQ)-QgQO7&caow$ZPKvMW)<;O_n50;n5GHxms5>fa1eyKQkJqKzJ-2eAp|rX(4NRHt
z4i$Z>6R%Yj?$H#FRC;A6(q-CoRbLWG3K31NhB|sY7~jJ7xn!F9+NAf$iC_qLAHPLJ
z-?@@HY5tG{wt;MDK<yO1%v*XhW3+W798z!)JY4wO?N*-i^SAo)E1FQ=)RK|SkE*kM
z>wJ8hrEekcLbX)o_&KdfmB|=xa{ACr@}UIbb>x%L-?a3hPEY}8t(NM2(l8PTwZbCw
zS~?TCuht&}WxGx+`OO<{e^?aCtlZ+y&(43&%S^UJL2<%Q0(|!_P5l44@N_!jX31D3
zKccp`qW4?mszhfH;C*^LhW>h=($SaUpy*%J!@r%fzvxB)kDhq~sSel?06;C6FW~uK
z;PpW<Plvp8Om)KQe%#X&EYv_X3ILx#_lXzZ1L)#dso{&$qdRyd9QYaFRR`LhmI@<_
zqO)<?<M`jXrl+()=fmfGXFs>w4!`{f*t_tigNf;8ku-pSf)~rH-vAqSq3H?5i{yTQ
z!WCBAt*4N>{B>Rwte=;*!8%jy0>bYOM5W+p218=-CjlY!KcRGg?P;Op#|W#E_-gQm
z0PJDl+h5-^k^Srgx@$02hMd&@YafaLKZJtfg;CJw&9sF@<yE@{41qXZ0BZ(S|HE{+
z(@j#MPwiWZP5i}fm?=ZO(kaA){J4(jo6#SxdgXEAerl~wkaFuv$$m68Dg6)+?oXHc
zO8?8rC`_JDIj66{DxBLCAy|J0kft?rd=_+VQjMKU+i%B41Y_+pw>;9HAGOc&Z*ov@
zXL{~x1rszKP2VFW79#K+y~X=7MxB9YUh!u#dlWxxkORieF;~=f`uzlDV^*l{i_nX+
z8xA2qM11&YmPkf7Gt?uKI@?@Rjmvrhxi{mjdh`c5?4q>h$zf2Hzks$M%!<2hDRqm-
z0OVH@F88=+Ef?B}<z)pAszOEMMn802o(fH+YoA0Wy^`h|l@1PKlQZsdXGya1un|bQ
zQZJ<umP&(mxSx!w4D^w-IDq_4N%71*(U=~C&%Q@`IG(%2aCv*4G6a7=ye!~6TBCM+
zo>((82<?M`^Og&(hXMm#YoZN!mA=ArhG0rC(zSIHr8Rda3*&-sfz?E)2~7z_6bY|D
z-Zum}uP|p7q)HWPXU$fx6>Ze}cmDi<Sv)WfY>L+}qk0JW>0(WZ`6r;?63SL-)ZXDM
z_Yz|+EMWRGGYC-B+&=$%>HL4gTsRr$*9pIez&rxRQ(CH4U_(}G;%JDsZ*<U&FGacV
zj(@$10&}m4^&dd1iH}Ca3w>ICA*ey3$d@X;a)X+7`|;G`px?7??dN9}XLO{HpdHXX
zpYLz9Aq_sKO3N60Xaj`{b`PNVPVpVA7z>a_uid0ttBnjeP3X4f0vVx4iqtaT*YRd6
zARhwIC_H8V-}o5tc0e<+r)UbwygSFAj1YV_*bM`q6nNx-jmFT>@M3-(qvx*qZ@I=5
z**;6bPj%^g%xw{==xdj5a9c`})1}%_a;#>hk}--e*!MA^h@>*}yjI%P))>rkIw|rh
z$fpd5l4^nu>WvF$%bf32jq^;{)i?ioR^^)P-Vc3A0xo5pUzGeJ(83{COFsd{@>7YH
z7uo2}R+Xc(v;FSfNdN*7KahXdy8k|;->zC8nKODA@6IPa8sLPZu^b<&>1TM=&*|o|
z@^AxNx}X<GlO_<l0_*2I?QFLVj*e+xMhx{v$bCJmJK1cU8E$>?T_#`@d%`X`1-*fo
zpgn~=OG#a)I~0?Au1tAsfBNj{6eJ=aBQtRKYiLUHT9%sDEt<!UARUJ-y%7L2rv~D&
z`BOdy+z!hWB^rzh&wfvZ7S$X&pOB$7er$Z&)|*rDHN*Y%gG_~VSO`BhpCyyb-G7b5
z@KpX4^&RuzAmB~?U|gj(XLr*$(5Z$-0i@i?mBFg@VfVPY7`3i)bpeYvq*e!Zle8L5
zmidveiX6^J{tQ~U#rYdf&!}YtdB%y?GbAVck>?eFmeJ*1xmqYUguTisb_t98bDMF#
z`uOgH;XGyAxW45gz;j$FMh&R@MG)!~QRDkS^MwCye2T<aoagojg>jtk1ch>(T@5<?
z*W^A}>249B_yS<Dxx;uK83?`s<0ey~+Iw>#CbkJJ^M;hx^@3LW@<doj=r>^9U-+Gz
z#YZg}s?HY$)*W}LTaTIkX3HKSPV0v@U0wwRoC=<mV>62>)=Qbpn7Jf?nJEm&%&hP~
zf~=;8Z}a1|3Hk9_x`dD1qal>Q5&#&w;E3(S;)9fBWO}gJ3wRDI)nz5^pUQ*}E?YDb
zW7^ojP67;o|LG~_TKpsGUFR2pD}8-0<KHG2yr)3jmUGmQ!VTOJYl@5brSBMKjaO~w
z8H@F}Mz6L(F5LcE5^IJM%?%B#{7OJpm!+iOJvmQiZse^o&97w7hWaj9O+2o!38i`S
z{r(im6eD_9?Qo{Q=%6wJL6v>HTZqJ}tlKSR#pm-&nin=#I$NzhB2Jo(n2-Kc_SIY}
z^hvpy#B%_-U6+^$uZ+LHj^QVySnmCuFbVbvZ*X3m3(ve`;-PkXu**+BZo&<%NN2wQ
zd|uCot@G<<Ne7C{-Ak0V69fAwY?!-VSns8}-fPTrq51vw#iljM`%X-(La?ahUh9c4
z9>fHVVEsG{Ip(cC)V&g=Lhw1<--A|hjt?N5Z20a3WXWT{LNh5zY||Tfbg%FhXo-3u
z_+#jmHNIL6&>^!Lxr!5Nh!bDcaXDKW3upF0?5?T`e)L)Y@iZoXN}Bu$6|W(|%Ve)U
zhFvOG=d;ELR2#$6uEJhexq-l{@!U1bre|+>`JE*y3F+b1TyTaeUDPDMB|+q?;A<1h
z$uNT`0*n!_CDfAawIuK35~4h?k${zu@YX`ANN<3VE*IJ)X5fxr2%#4DiaUvAF}*cp
zd;ZLf7@b@5SM}G#^<XrBOwj51$Ua%3E6ioa=0|Sc!O!jsqxf(-r<+0BxeFk-_C%uC
zHRmopXnSEGD-n#TfQt|KE!f%F!K@UCXLnrKyCFc6U<vtsnO^Dg^rkp?fo2w{&cI9z
z&%y)t5IoQJLIeyzdPd$?w9fwQ)}KuUg|;0FpQ{L+a#)-ua^fBVkr{YP2QNT6*z}nR
zPJRMF&WGQ&I6>bXwj3}6nBu^50zk_^a>+`CZxa5eow%Eg2YCP#$b}vvQ-bpFd!`C0
z3f%MnyrXvu)Gu&Gfe{1Hmi4;S*O=zvJNfP-6E9v{Rm&V2GG=E~r&Z`@!uYo)M-NL`
z-k1<y2W)3Toj3)cBVPHC72NkS1CHYrZD0JxmcH;bLt}|7*7d3!<B?79*ardbNdbWM
zFM6H==~Hs6yltKH?vJ-NZdin~4>^EXaPeyp?fr^CpcUJX<;UJXXNO%)O+M%d3EO6c
z==-Bq$=wmW?f;D<C`7J+Anza-5kJ>srE0&RB|Va){>AC~KpFZp!3fmjqmX1WZM3hl
zfWp&-3a`VbiG2n{8J-YKTh<*5V&<!kTfd&v-=AP!%B!wuzCY!jvS+-Um__p1%-3v2
zy`J=j&@<h>6IY{}G69HlvitCjRgu!{sM{}0KtR;c94RoSCK5ZB*{%dQf8@^G`Ov(a
zvH=@NAa^I{bg<(>tHECBXQhPs7^d9nz&;G3PYr*)gT$bxCV4iwp3}c7%zo`)T!T4A
znYQHI9H|Mbucr!*%Es{v7(xyiA-_4^YcACV$Vn^pJ5nW$oZ3!)gg`jK)eU5bXVWTN
zEmSRGKr+Bn0EA{i9!d>i)cFAo{+?J(PgZsUPiV)2Wc6p1g$7!aa_?TL8b5pu!Ct*_
zvdEZjT>4=9r|o?@u%Yn&OoTRL!X@wAnnkZeo_+p~CJ?z;PY7#ZHIopjQ|5Ib_%?S@
z_N4#Z#_QAczcx4z_@@64GzwJg7a0#>I^>`i#lJ;^T6x)GZ*)lTpvn%cBY>W<asrQG
ztW1*y;vGQq0wM!T5%3j)ekK@m9ldR;|C<@KOY)az;(R*u97x(^Js%8fZVmQ>MjBW^
z%(Q`M89W!@=>RU33w|_3H+}?C<ey3dyXaT|gJTKQF`nKF2EiJ|)AD{`f(LeY7i!I?
za#POnc8~c0$pGMR0QJlrfp~@Iq<#Mm6d4MnD4<gwv=5neIl%YUpuqKcIq11)J;ZQa
zq`zsE&otX!d2tdxutc`-k+Kq#NoPrNIk<VeGDSOGex|M7#5?f}>m?Y5?&KvxcPaxu
z9j_AzC{X*c^r4N_nS#xb)&U6*OL)C<nfLJ!cIrdbI-#G08cXGAkWj<O!r7kAre@V*
z+n5Zans~o7V_Gv)xDabPEEV?%mV}bFG=E9bi?ck4)l8==6>i}jK#Y`CZk=sf2B`rO
zng8m6q+1Yu+J@Ky8{HDKPuh;V@qkau8z02=Gthr{NE;;oYzWaJ(x`x4KHkY9)*qj@
zdVbVzoUV#X-BjfKO&!T~UeqRSYAP<i;_5r^vmx{QXVZW~N0+^Eu&v<goODT<^jJ*C
ze+u`WxcVDp5u2*HVRSwtZEnEbi8+cjx}^?m33O7HKRXj(b$S}(RQr2aoPDyAlhuw@
zrblVK$setLPvV<JCDrO;`b^(I|6`tY*EG}!<lU<`zj6kx#YiW-tT0tlo|t_YjYz+5
z7c*1wGU`b8qaeqpX?7Uzsz-DmxtReJrb?41fXRkx(WvsH*x){@mZhO41lHw*DKy|5
zC$mHmss=)S=YhqX!goch{0A_;n}sRY1WS?05Ca9P!q7>Dl5%C!Bwt7!c_hO(DzY?I
z0xj4z;=D<v!^0R(T%{uUl@iuY;$$*q0O%jpPWN$xWi#xbGuq+X>DzQReYhjT{;MF!
zN4EQXt8ITC@ly-mXVc5`EXUQ)x^UY4K_i&Mr)zh?X$fYsAZjFX$tX2n+^CyY;VDf3
zh&a1QIa^v@eh>r(3qW)^y;(=}xVlx~H1~X0ws-#I?YywyPED!n_QCl{?qVhTd1fx(
z`_)RYyaTPz1)1ZH+Y5jHQj}yNJ$M8-;=!Wj#!FJ56S_`8*V_RQ2>*aI%`f-J@eMW5
z(Sda|0ICA76hJ?H0T`$?FRjw;kT?1dBod$C2Q9boZ!uuR2gJ@`IrCqH5`blLMBv54
zdrEzwBVuTQ(C|1;WIN37_Cl(q9pyId=Tcd##W$=pTCAL^Y>C%{lN}ow*TjH8D#y|`
zW9U0tS6OP{=(?fZC1qkbYwt1z;q4S2wg@=I29EpzIW+p@MSipshNSF!jRq+7BA@6T
zKsY7gPMr{pc-~U=j$=x|`EA+xCmHg^_w3Xjo!Z1(TLfN+HU0MGH+57*KppXpiu?Fz
zu(sME;6Sk7e58TYw464RE`p3fmQf@-BD{2Lr`wQTlP>@aoiS}~r+#zWMQ5sY!y$23
za$ff7Ryb$%A4#CBBp5BO8c+Q>^`Ad$pw<u~kij-cq`Kx6w$8t`pONC3V`biS<<LlP
zep_iK8)uXKo-IZnMx;p4+2^4qj<SY@7505kmUsC~oWKgWk2}d~SsB@NuRnEG$JX8x
zmAaLbgjvtY+O>g<eE!4ritiO=ay{IgzD(H0bTj#f@PB+ys`!>~zRc9`wnLQrrM%sE
z|7gtVnnA;TZx5u0dq;pyw>SrZmH|vll{cA{Uq^tQKb?_WRfU0c^oDmap*UYwGptHa
zYdP@*vLJs$sP8~)Q3G;-Y<vV07JmIvfM?*%Cmr6hP(3GxNF!(5!>(t=qHw>6uBE9(
zrqPykVcQnGX=<;8REt{@q@_9rB!P4izDHHrXuny!rr}be!jUuU?Y_bz>a)FJfZV%s
z7QuAB_isiA{}RR%48Y_nKZlaOe*@!U&=>#zj2BAKBTVzfBkjinFJJ8J08SD}i8u4q
z+Hv5VubC?I@(kNM+#9WLo4mxEbhg!YqIpFp&=x2#TBzALIPl;bUtixNNp!Q&EqfA8
z#&x17_M&SJd!Qc&(`F#e#CuH^7ZsX#MhWD>AFQJa5*6OxyhynMSiiAnVW5nggU;C}
zCN4S^VEiMvz^olzOOEhiJ6LE}rX~bk?IajCmrQuEaFZ;;DK2Md^jU$o?ciP@uqZe8
z{Ho(N<)#j1S1bk<YvONtRoUV~viv<(BFmVVQN&4M;bka#YmI(VO8Zrmt^aq`NA(KP
z<)>Ai<9!kBSWu=2KinoTbPx!M_%WZDaJlCCJGu)r$a&h8qDDBx%_C}S+Xh;F-p6*!
zq#fx7WBq5&<-lg8g4p6tnJKHGT|2Wl<S(<O$6BM%^V^3-xxD!Ey>Uk<<Wf-Ta{~Ju
zoj&mYOP_03D_8Gw1W_d9$hxHOycZF!kIixHXER>)w;woqn5%>PrTH;#z|IYG?pv0<
z>iNRQ=?j839oEU6RR*<lEO7kKkY%X%O2GzD2>0S7HB7nfV0Sv*9t~y1-%d!Mj#d9H
z6;}8TL8<iE#Gr|dQd(EKXs6n|^}8xBvF$ADu()MP`OwbvjIyIKrO7YH1LQiZm`{$I
zG>*#A(=}<XxfhvUs!GPo_23UBFKIp+si&&K1w?3y@P~`0tT$g*MMO_3q6&4MajHog
zzdN=y`A)|53r$0<r3|+#VIVX>8W9+g=#=MIU2bQ**PyG=g}f$b{N8q2%upW!GN%sa
zahHsRA5gpxWP6i63IY(Aa9pCw=`v0?@;pMNOIQPj!thk!@zLnOimFVHhKtUn?Ur<R
zP=kd3zeD@7SpqOSkX|_Yz_kwWiRS+Uv3%x{k5sWDmP$(6c#0}m$$<?A0DLw3?GS@Y
z-;w?faR0TK$D`f}>&YSjighs8kO06Ds9oDS@iwbJM*#v~0>%Dqt3fg)o|O&Mo&fgJ
zU@E#W4D=kA#qI&C6D!r;Ej(f7-$imf7i{!2ir=#36`)Cj;l86OIHuDU&}@Ua6_Awy
z&>-9Un@|4dGz9At>6DA!dvQ6q`#fm>?2bXka`P|3^Vg}3hS~2u+gehKBtrpcqD(}v
z&neb<I77UeU_+JI$5$+7Pj2oML%e~K6_}#@oE!e#YP8N)#m0kDfYFAm>PN?Po>&Om
z*V%HmFNx(_MTT{}ZRL*9&Z-+OQ3$_bzhNow%2)^42M=v+?H|~WH;grOoh!8@UrFW3
zi5=+E)j#YHx)oo@nxLP(+F_#x?fsfBxo#k^Uf_uV+O0p~zq&3<DK~Jhw|kK5#aY=6
zA6!<RScd!P#-SX3eA{sFu-vW`v13h0q9o~W$GW?pY#2j7UMn8?h@&k$3gn$LTqR0M
zg$4{?s{4tnqv@`CCgt^n>(QvpJqwwJXs-d<;%^w!QY2P?Nk6Q;e`)f^tm&zg+|O%p
z$*Z1@B2sjDPc{1L1#=CPPmPGVSASB_xYW7ozmSmCl1g*#JHBR#`TPAT%dpIRKPv&f
z1h2GsWSN`C7nu)(HJde6WNP9KE<BFuKmWKtkLx!0D|X4Awy@n`Y+lPgCX-E;I=W~=
zeMP`H#ULUr$&QkCd^VQLRsgjW8{!UQUY8Rw+``h{;3h3zai(D3{^0L%GE0(LZ7pfs
z$ib7~$x=d`7<g1y2n0(^;zY0c9Yz$~idO_~R}w)mSox{(hN{i4?wj(?&v%AK>U_pf
z3#Gy7rExX_cWUk6LV3%7I|$R%B0t*%-?sfEbtPEB)XkDS{0b?ZNNl8tb$cXI7jhlq
z011a@-Fl-7QS3S7BM9%BB+TmR5($ZrinNclS#i&xQ(P8SZsevT%qv^4CE=R4kWxPM
z;l0HBm<dM{rHe7GaA0pt@*stBKh{dm|GV=(34DyQ(aJYFcu(By-}{u5x?P%kG;Ek1
z4u1ZA;ug7H=@>m)Ap0c<$$;=ie7#U{ZUqj_z{zCn#x5EeG{Hs)3H9&aKNL4MHu7dR
zLEzdOKtKEPwB+=Mm*=jq%wH<=fIs|!BKM!XpI?9UYjN0BrZuBr>-Dk|KPe12L%^a2
zZ&2~{r2Ex&E|VjJc)t+1y0NwF<(uqX_y>UxFI-<=pPZTL93N*C7Z<Opua5``A#iqe
z1s+;U8ykYFSFe(5rBoAsThh~~U4bEia}a!=4#PQC;1|I|ejyq}bQ|B5!haMdxA(#{
z_U~KB_tGBiEs7orOyBTs4Mm?ao7g+9E^lB?gt`wxmZi|Y-QjY}lY+*Qe8=R^#?6vi
zUYK<oxM0<u6WCPSU|{;A$TTwMd&0;kUMdp?H56&4=*)L6S45af5+7Z0{y6Xga`X4k
zM#;2h+v7KgRedM_bhkg=!C3Yk^7Yg~iQ{y_5?C4IE8jp7<6oZaBF2JDagHRh>IhHE
z%JO+pNPWJBMiA`Y>Q@0sSZN4$P3-+%|El&#oPO40+qj1487=f<_bl8{m|c)XA8SlZ
zyRBT~kBsAW@02aRRQsAFPxkPk+CiK%vQW4E7Mc#TfbxE1ZJZ;?{74%kWM!$L&0;>z
z8m~)*g+I`SNj$E%AMy)$Zr<~FTK7hF>g$j2^+V66GjHSvl*e)3Y5Ch5-ZcztR9;J)
z!nis=kI7qJ{CFt+TwaYYVLwQ_si0!|PVG$sG^GSY;FeZvEQ+++34=_!ZXHQ#)UNm-
zc91NMu9G5IBJ1*(*jwT#OX^2(7ET|naKa=M=XI7~5!=w)G?YqL9%g+M(YH$@`<pI!
zLSz9kqAC2INR)PxNRX7xS`d(lCQ4paw#1PUCP)XnX^zjLy|tzY`HtQ4F!&N~d3{p$
zL*u0T<<j5#Cx))4chB~D&c46b1%Vrl4v7{w!mT3BD^nZuxXhA(4bQ$NJs91)bj!YG
zihs1#ojh0$QO$I*tgNfAuWo9JIXy+Nva&Ax{!K$dLSkfLVHg+swPkwEer-bbi2ZEZ
z^}Mg)s5Ni{j_Xfsj=DzLWvBwMso?kZ9v$5c{2jspCk_7l`1EOaIw(1_y5-}`hxIz|
zXI*R@97rM}B3}Ico$sMBH$UGyH^&a{z+g#wH=8PBu;}=zf7BmoYS4lKGs|h9>|PAm
zNe!NU1rH<mJN9*_t;B8kk!uXy<v0#ZfF&Vtkpdk*#BOA5{faufTUOScux-am#z|t8
z5R$HMMq`_mj(E{`C`FuXGhqu&#jvs(5Os--30~r1LSehN|Fj6tyka6+u=cay9KN~b
z94-;#dRJCBOCVE#fIB8&K+g7ZfHD1!=t0u`6v2usD`mv&snF*BRTC=Nm{op_+OUOZ
z#x6vag$k*Q0h3<Gu#pxDwYvQ#c$#yZF*?T2C8gmC=bbi&-mh$_ua+3(ei@52^u4HS
zyiwF(b~Y*2JKz#;?%MYxUf52QUODTPRVcf6<hMy#BB9lt(e(jKF^Rf~3H1%Hy6isn
zj!%tERS>jakDwGTU_eFcil-aSM*w95<tUy_mRjWxcQa?4>GUgD(S`NBDR;9m(D+P?
zUb&H#cpHMT%%ZEeqv49{lb)ig=Ab}&j6XBU?EI<GiFS3G<al|^@2)hz(OO?W7>=oL
zk(QfUrg6Da<T3E3?s3OJ`ZLfc$9huJYM6H+|70<*bfqDat`rffLB+Y=kTGY%#KIRS
zA3+|*x)EF6$V$3pop+g3G2UsoJuQ-(o&L8p2@RC~GL4E{I`rh%J&`HC>m+>hFvRcl
z;jiNptq+tTc56kf3RxlrF&>mKt&}PZna)1vep6`D6<W(FEcjT9dmLE6+vqv3@?XDe
z8*25K<nO?KuVHpH{%lTO^#djlzJq1p1jO3Hh8m8pc~l#LNZ8LPR7Sx3Bj)hFJ^j1V
zGB0;YsX1i2IwPpJ7di+wHaGKu-uXxh`P1H!d2o9%$#8h~y!NfH@U6__n}>}FOh;hk
z4J_i|LJMA?ECFoK|5Q?HR{Qd0Fd(X#!bdNFG0_D*Pb<iR<NnJ~@U-mV9}<z*C*GvR
z{mF#yfQS`*D$ucl>IRzrM#rvM>BVL6;We8mA)zuN0gXK+l%#^a$RP|x)C#5IJZThr
zQ3kn0;4bg%X06fDC!e=R=jORUq7pFwI6i;D)&gM&huYDlWtj=__aEG@s~UEfZ|(<L
zC7eUFcDd%4zx^TDZ6Bt=390twYrio91}yIdP-eoHH7dvxa<B3WttAbwT3ce=Tb7wM
zIO4i1wRid}t_ApOh)jvOxX*hAW7E^_F8Q2VVi@yjHU^6lWs3P$R}!mlj65My(=fz(
zs?=ptkqg(%(^XA=6OJ-OXn5>d_7F^M=m|{R-XxdPt;wvW63T2}@;YzkbAbg-=V`1-
zXAMrElR_hKqN^;;r5(+D2R4-A+^`q)ojso#nX`DRZfP$L3FKK<$oF_ivmX?GOS`;E
zE@WWiF^iR2dQ;Z%S=8Z*?l{*>8MN=p0@FK0Ryk(S?^b~j({t8}%UX3?^IVpTea~a!
zyWiCHGc(VN$z6d&b|Xb&pjKUR*lH)yTjE5WXyzhIW;zkhMY<jlViC@&TgY3ZW@xhO
zJ-2N0ZY?o~?A#)?c_>C~LwAjqs*fzM87X3lqt)`Mz-owt&S5`Nv3W2OZvLN4*Ckmw
z82quaaU-9^lDICZk#P>3QCZIJ%7mGUmj^gxd9J}epG=MMy=BU2rhriC`FMDkq)QJg
zUbvq-!LbOY6DvCn!EW9p3t$2cJWmg_KYU}l?yC(19RT16KpWuAx~Zv2Qc|)VU#$Oq
zxz5VYZXFcF0SGhly5bHD#bZCac`6sEdxwWEO7{1d?AiU3t}SW{YXglxceu?!z&txB
zyPtb6`>@^M@!9%(fB^l)UL&xoOn(2a2dY*2MP3%VzOD{1n;>$tM!2&c^VUe^hVQ^&
z1%9Wu#!4363of3?R}2nx@KIcfrsBChyd1mHG5S54A(AzePtsWGDhbMn>W12ev02Lf
z`G`vSFMLK)zjcL%qV8cL;J=kLFv2?vhvqC8`)W_oop0iEVj@%Q!{0rxtS-C}pZdEB
z1wPZUm0SI>Mk%NyVIsA<o`<c^lhYXDuIrxc^fXe57?<Vp#Q&sm^Fh8J;rZfW>0vwv
zI_4U<_x^hFn~B)9l_+PKiA&t6&K4a`o#N&{&K%ND^=>`VddZRQIS_9V;5<|RWMlN%
z>Zgr^v7`NT<l=^CLU@#}H?4GS*X#AQqs8s7{ZnB9(t?yWt?vS-+&>r@KeD|=WjKPF
zli?$fjHk~MU`z3%pneQVr@1Fghk8USGX6cu=Fu`*(L>t2K<XwPCp7ytr2pMt+Q#t}
zV_%b@AuFFC1A;)xW8Qj~X_Ax@KicUL{>+e;a>gm?K&tQM8BT%iZ+=yJ-Vy(AT2xqn
zno&xhX;v<;kC`-9iB<Yq591>09Wv=w7zEAJ^^~HghcK2@EY1pd8A%oy9xK8be(QQP
zrycE3Unp6mrnClK*LTj#>QU&BKmQ+B-yKc$|HpstwMX{2%9g$E#mAl@M2L_G;ZhlK
zkrBE^C`4ITu2G4M%*<qDb*~wb8JQW`d-Z$W`u=`@{LZ<j(@Cc|=k<QQAJ4~n!pdW1
zoM2b~9HyO#@6RWhzXDLfkmaMO52e0!I04!g`-mDUVghJ$Cy%-?((9?3+c22f-V8;{
z<~<OPHr~H9LpL-dJ&KDAMR3jJKRPZtSm{$sXgYnOQM2p|a;R(RbL@8-*8!S!@bEYX
z7JSDcBX7q53IkBs6Wai*J|!A6E(`oi>yb&UI1p|@DLuyd_zL6W&jCQ}aQ{C2@$oS@
z)uvtJ<KrMGB5I|#@4vaQl-Be|<zVsA&cfr_hJ!f^)hUf;tv3&;td$HvxDNP9$i(-K
zLyF(OCkpp(zWy@Cxq@%!jnN}}4yWZbDk4xRFAsbFUUGNe55NlWZ;y<Qwk#|NfeU+i
zC9qwi?zgWxe&2tf;AU3B_u*l%ixZ5|%}OBQ?X~U=el#2|)eQCBg6evK40QX-qH{;X
z$g2_d{_^$qKJ{=!oR+gUN%bNPuhN&#c&!)&OMgZF?*1&6$u_kgx!nMl+LH?yrP#(o
zQq*`Fd1{=LzpV<MI+@l{f(tj{C1LyO70vM%$imOP>|{xU$ig`Xo<E;wZ-u*|S^Fy+
zCPD>%_(kBY6VY`y+P%>KZPe0EOZLuIw9Hj(R`PmPpNqX)<j#!>1UMz=c;JxGuF4}J
z@v&T@A)~OeFXZbD-Nmpj@+6whuI7#))X=H6pwiRt^`iOiMr@q|Zf8fXzdvs?z$sy-
z@|hYH#%=PQa}pD7_w27rmkLWyPQfmV*AB~b&v&86U(E}JxsY*pLJ($%inT0UaH_F`
zqqk2L=4l{Id1VdnopCV>Zdfa5h*kXkuKT}y=UysK{?y1P%nI|Q0jY=YWK0WHt4t)0
zRV`c~NpDy8k?cIu)ms(!!-^n;BX6tW;Sst$+qAp>Tp08#Ika@hw=RAm0$eeSt?kCe
zX6Kh-<y7#i%_EWU_-2HVqd*5D9FwT$0EaH~!L>h=Hpei=!x1PQ=8nT;#yp5Y=bR{G
zMrW4~rKdL(^XF7rdC@ul9y)RBH<Okkk2G)dj1p-sr=}*sJSHh->1)jR@JuLQn?bAI
z;S}xPn@mM0FdLOVj!EiIn+hel&-p1fP5m%!L%2Sp$hZ^SmfQxR)*uQ3HWxv1TbjXI
z0q_9^WqOSh2C%4(=;r_$`M>fEpg)3Ya;dH-PG5c>-AMG4p24;56Dt~5*VY{U{n_cc
zRf%_ac~esYIH2wx9=E}82OzE1v^w3C^QXWgs9o@Ois2K#dh87GN>2-&6P6atGJlPr
zWo4;pY<#x6>j!LBVaO;zrnEcW{=^HwgRiV$CdpfpiodvLWNhs7j~`j}OL|i4YinZu
z^PV*su;k1V<E@ay_2cEJs;a90z+9>NaE}(OSpfdnBz#+DL?-dnUqO-USM^OWy?Cmn
zu<-VK3%mjTD<!eQOJ$EWP69Fu(VA|hzeF0Az)Q(2JdcR^K3|DRN2RYHQwI$HbY9Bq
zHY@X9jqD~J>4rxv9xs22UljDj2lY~4w)CpNl=^I6Lp-Xk=!#4NM%3zwo%$8>q@o>G
z4=}=2V~4m*x)uwlPu$`j;=L*k&zj~jx7{*csE<~}hT}VS-N=)E?-~dNVurG;?oRLr
z)Z}JUdH%QH6!Ewqhi=H>XTycE>fcrnA>k?O3uEVxQ2Q1SXuFCe)W2tJBRJI4o1I0e
z*v2mxD@*73aoGERyE=q|Or0z?lafl2rB$xTJB-LRzRgOj9dX%~@yz$-wsg1;gA3EO
z4p^iEC}NU>()7_D=1deI<6GG%^HNRy&HApp&Py#7Q86LSbVrDe`)*yh=C#$;xQhuA
zITC1t={e^{Ss#~|D&kuy6g|J~8BNt6H0p*njhq+e$USBW)_qQqEZJXLu=*5&ve_me
zDwU`n+5|R5C>kUN{6<__wWP(@kuWwdv>_5QlZ2CNFD5`~pE$L_&pjDoh?FIdn>pfx
zG7F7ewf@Zdj0#I42X*q1o?$YzW4KM%&lzrI3z^pVGq5PIy{f2`K3_uk!H=3qe(P6H
zU(O?#qpCZTqxzMQ$B<YNIzY*r<4OlrQ!(oyV(P3I-XEWWx-(CTmg8slH|m6?XRS1L
z&5MkQh#^bg4nY&HDQTGB@Gn7y$pMc5xXFWezUm$PdXuoRvt$m=5ZHhu>iYjuttHI@
zr-tq0lx5)6fS?^}2d%&HpFXJp_#&ox`1Ot6$mAqgc5CI}|K_Vm!@Ic;L7xs5iqg)n
ze^(E_&~&ipc5tCDWZ7)?hZQ5^YmxB9K#&z6=msL-)8OWZfaq<ZttvRVo(XU#V4VWH
zpif`6E-d67jRyq;yyFHh7mx@4=U&_m3K?3O68~GZ_ZBh_zHj1AbFj7saFZabyNp{U
zV(>ZVB)X|(2Pr3HWVdPnvPJOKa50##9tvJub@S|OL|@(u*7yw;VpBRItW(Lf-Vwri
zWBBHIXTPkvN#Pw;7hGGM*i=M>=i|yhs8n=Liw2B#x!b^m*-yoEM7;u^IBZpHKF#^Q
zjPa(rc(cDob>E$5TQ&Rpe|0QgJs(P6Bz*F4Ib~{|V<Ekp8IJa<r!BdvSEq;tJyU&e
zD2;pnvxi0X46zTV*TmobEn<<sXwR?@b3hmNZnJ!|<S;UxH$JTQQ3mX1u<9p|rwMu6
z31p||U9>c)F1U6gxOV3i1e05O7nEDVF7$Hs?}ybHCuEacsS-$VUA-TRi$>l5<nihl
zE#}EBC!hA#${ktR+jm%-ijxyjpb-{dbe?Y%*P3bjt>~?hTCSg7tFJH<qgO5<D`@;h
zhAxc`?`X>1u8F4AN;>JRgT*N&nB^e9YyRTQ;Uwv&u1GyK8%L2>#UDeKPVc~OC%txr
zW49!?j@uDlY%;x7f-8i_Irig<N?(GWxH+WFRJ+r;yLVYgCf$0>qmbm17sniX2&_-d
z;cyoR_scLT?g#{bM-(sZlRG-J+6)(DUYk2Ak{}!-`JhhZMlrWx4njH3-I`d2d?k{F
zI40h7edJRbN;~AyVVe5=-Ps3dz&H|WjCg_<DoGv6cdMkKGa)I_U{?(vWS)G>n80K<
ztc9WyVLy!r6bZoZm1s{w7kDE4HxFwX!?F+C5=xa>h4U}G^#}W4?(A4*c4V?R8QsFr
z{Ov&p+nwuDuPlBB`2ldq8vdLe6PLLF(^l2ZfW1`?071g9&#@O@V=0hc+9}73lZk^q
z8DJp<B?bW5xVyVM`1@OhT~FXv`wY0lvQMAl!OI5hDTg&>dge853@dE(%<fC6tr<?8
zT_pmegmZM#YI`9`s*k`U2B2537U2BAi3ZgngqA_@;6A`_tM0}UeO+#DZbY#4$Od42
z44?@CkK^FxMgwd#n1Daif!s0O>_wSU_F0Rs{<dWoZm<2bei2u*0xb7;%iruoyZc}G
zFU%Wez<kH*QiD*ARV}Nq9~1Sn@rP8ZoNb<+1VN<-m-ZqwF@#qEY^G>RjSWWf8S7NQ
zG-wTIJTP&|qL^ZLGEi_%uUQv-hQFJ5Ga`XQhqH~LV%RvFeH`fNZ>_qf*i*D$>#99H
zz7l9xweM<qqn1TRzR%iQ!!X;YR*1cIZq!ep`cAQU6h|dvFNLu9MmcHq(6$7jlpqf{
znw1rK>^E8p_JrpV#j^H#d5425WJw-@U1^Ld?Vm0pI3D{gz~9!{QW{B>$ga^~R|w6{
z<Mr=(t}S4@q7XYtQ(@!H&T4t*jH=_YnXa&c7q3HdI*hg6B?n$Ek;+u~o?7WAC{@SN
zuehFrTS@Gbt>m)YP)vnHo(sJ9yOw;j&{mNhQ%Wu^9)sP9#xfhLhCNogyY{6Vq$7C*
zqFxF-HS_F>gkKkeMaSglEU>Wj{FiuhrAt7+(s9(-GvlR-K#se7I?^c9dkV94EHtY3
zg-0Z{8sCVot|Iw0j3ac2NB)W(A9H+J`f%Rf>h%W)w4t4qZbj8REd<U;!YBYEp$Ll!
ziwI+OVkCVhq4kpt?m}wRg5dl4Oag+Mj*IDSMzCngWI`2zuK@)S;6EQx{-hIuid3TX
zfyg>yWgVc;$ZYz<NfV1N6e#aI=hF}$7b|N8K$B-|k-{wH!w7!4lgZi_|MH7~Z(+j!
znb@6uyF>pYl&f!_ej}~s(Vd|JK#WkHd!u9*utgnusCPyvG_M;pE79P%VSPzwpdBn-
z1(QQA=ecaPhvZZHRnmAS;V_AJ!5M&62VJLS<>jw(euK71B9;AbP6*hThVJd{l>rPC
z^vUjotadQI|NGP*y2h<f_wB4t?g#Hs(3v%oB0E@eRDpf6x;m<-=PCeL#8ey6B%q{Z
z?e`K4HjAcjzM3}m1D^qSxZc|)fyLBAQzPKNk~<%7k^iu?9j?tl3Y@&)vHc%y`H$@Z
z#0{v6f7l0n4d6YI3=32<(dx|VWZ4l9b4pl}!h2^*AH6#xXeosb=QvUcJ#5TIo9>-z
z)(HwHqR$(p=2Y+;Y^d4CiMKdQ<6F8Eg(9fZoH0}Z);BIiAfc7ZD@+7QSQ0AaYrTlM
z;n1Ge+Z*%9BIT{I<7|2(@>Q?0y8D~{GF|mT(j-EQof7zV<^1$r!REOw9AU91b-cns
z#v!W-WlNlyd#|Uo42r;;C`o(2OOs45j3bjtN=EmF>%)J`eNW$(cp5nBfZpr4KEA?#
zS37-ox!*6eIQZG?tN$TJuV-hM9t?=ea)~eGRz{fysX#@nAW5Q7fmqESj{P$9Oe={#
z=_K!$7KUw&S`TcaQ)nP7!F5XXZFPC*G>6g!Ib1~8D-{uelJB_ZPPNw6v9I;#44?2t
z`0A(9eW8lP*%~sgb;fAjO*f}i6u_tGadHCX1+KSHFrT#wVU<?Vvc`@F=?aqstaudF
z9~JHeqQCbK2W$IUB*GCb!(@m`zLKRc^7O6|1(okh*iD~F9zTH?)9P@~LcASgn&J7R
zSL`5i!VYwKv{yQ9%fhrF+NQJ@N%D2zGOziZ+$s6VM47*iPzc2xBp?64dEaDkqPhHG
zgH3LqT@_~TRdE&*diJ(&syFtDvBG6=mfjvs%VxiQUm9Xib3OZRg6Kf=*7b+Kj0ub{
zRCtJR8MpQpm$4PN1Ee8na~Q$cn4vV<O5;E(^ykOYpZ7yo0diWjdxDGtPFn;m(U)>Z
z4)o8=#D+G)(m~&hYM&t#ztv^ctMvMp%a4O%d){xiIN?s<=uzXwCMPYtw$|39<P{WB
z-mG+VkORgLygNvIJ;NH$&iprKGD>Zi;H6xj9=_dNj|K2&=-zUbQo0dK?g_wQAX_QE
zhz1-@hGgJ1b_h7SKpZmT-8wsKL|^?eF#$^WG@GXUF<a*t7z_4;|1DqxQwS*8`y?!P
zda;zqswbIK%kKpr03}H6r_O5NvQ=iXc10Ql#7}Wz`(#EEuN*a^p8GkB?lXUPo}vfK
zHeoMLdBf8Du%q;m5fxv)8%BJ~N`v6f=Suvna8Gv^uA;@+(lMt!S~n%XEydhNK521$
z^yZ6XeeI%O7=9G{*fd`-I`{Wsvz@sS<9oPqYDgipy~9)_<D+{Zqn9;HzB_l}+(Smi
z!1NRMT_zFn$m?m4Bwg!GY|5U&wOeNnRl|bO-hTa-aoOtJMRV6a^{QG{XVP(4ktNO7
zui69{F@YB63&mHz{5eXfdmCro{{qQi(7}+d+~&8pU2%3Tp1Ee2H4I<zW7sY_yJt5r
zQL)M$t=!utqm`+=;r<}FHypoZK#?@)B_kIDznSUT`<mCGH^tD4MKA|JkpOkhLSME_
ztQ3ntyQP*3^s<t6`~3cI--Szl;qbRUl#q<+&%OT$iZwRS^|V7-^I>rw8Dr;Cojj-#
zzmpTJ?iP{kTa2gH6#aO>ijkpU_Cl{>jdB11)J969A7{YrWlU?H=m-<mkqgHpo1{l3
z51MW<=m=c)UGS!DIV6olVhTF%l5Z&Gn#+iWCORJBI$hc5FrhyU?Y-=!ly2Pshk0_q
zW9V%Wn$2uDTQC<#mSXU*R($_2(WAdM_elM8bB8wj;g+=Kft-5f>83^g?A!;L(}Ddz
zTxX*XO&}2alf9<4?bDEx^j!+^-sLk<TJ4>nPv6q20RuGZF2=65RLmL0N3>u8*X{Xu
zfF>@#`5GT-rP{~LE<PKZzD3gs9s@AE{zEwIKs+MKmO$kJCmamcAO}HQ-M6xfrXv_D
zm_ORz|HBH-`J)v=XW;K&T*12ukLWeZk6W$Ohi|bt8HIW4{M-LpYOAVoFiir`6R8$v
z$@k^5KXab~V6U!jsex;J`M&deOE3`#achLe=_PO%*b5}m5daRN4t@W9Mf?X$I8bUg
z#C|ey_+AxA;HN}XoShNn1%P`P)}gsH<&lXY8v4+VEFO|Co37r_i^tUZso7uEevdlY
z6-rpsvBa&q`4P0N!@8ibbB=o?*Mxad_7QK|Tla$z3tDjghfl@@V_RgXMBO&f54a1%
z@p-!ely%oV)@U3hnM+5`eNc_v4#&5fx`c9?ho&1*cM$e(ksGn0<P+;T2ddh^3!#f|
z5(F?)oTTZ;*mbSRy{f-w9Hqs3pK+%3NYZ=}dlSK=_~}lzL`G7fjbp<V2XtP^H|yGP
zw=D~>$rb70*`KoAPbes!Qi24&MkACydD_Xp?>e}rMB@BXb9QI{4qGC2Yo!{wg&ZX3
zJ3<twr|$*1oA>G_78J@tUMu5h+cv66<7QD1h6Y5~wn&LAqzNYsLFzo3iOK%2Op*K<
zIO*5BpA>DruB9!cZ*I|Ts+CTFx^QdcEMB6Dlo}Hpr{>pTGVKc=FQ~iH+9_mxdtsGM
zU%Ifa=CU+2P%`>mxrbOE4W<v(?~ON-U>*FnCj&t_@**y`5w)QRwz+aLj1p3wy)S%I
zzyT%{cZCB%_7zSmz;GGTtFI`MjWpKgr-jDBzs5iXIt<i}7V>N_;WsaOy&H#c;#Pg@
z00$&#*>N@SnJV$f`)MjE{k$lA;V<WN4BF4L4K=WIn}bcK77aUili5=BgFzzUvoyB5
zutY(yuw7QjI(rq7Y;|!`DNsE-A_h?7pzfEoY{0Vy9M;X%QCVWD=id+s#!SGaY#*Ks
z*rrt42eF4x-Ajt}Sl(5>`QGwGwiG%%Je+J(5t$6&22n;0Z7OZX0l%+H<r2>Jepu+~
zoAp~XITE`-b=nO61EGjttBMc&e(@(c|E0f8fPdE`1rfW=YR17d0ZMaDS7jC_w7^?O
zL1(K<{7>)jz5hTXB~hD=4*eCq@lPZW(+mU%ANZ7Ya?iL`-dk>|(rUKYWf}{Uy*O?x
z9s4W$Pu@3p_goY^0(R48k;5sJO*!YLXC(rPA+G0wjQ6RgJ7*8IAbs16q(_QS)hRs&
z4MRmGGF3OshwvBaX|NZwl9FVKAU4GQ_{pwi8e>;AmqoL)`JZ6AJ!CB3?;XX_#w9is
zWUt~_ocT0-I#50D9Qu}!9V@~*F8Oz*T&<1F1F$oa6r?)E`y{h!y30+ece9v{dM{NT
z`E#Y66G~e!gBnRat@~zcE@Ss|1%uSgJ{D}5n+Wa7<!iE+nbf>SDn#;RyTW^lp?%y|
zT^OCl4xPtWWN95eMB2?^4exq=qh(+0G%YXa?zs{1<QzjDmC<m`8<q$1bEpR{dDrc;
z2;#pTCBoeKOKekEbFV%rl3CnT2Ga>r{`p#c_aH`ASV#5B)=%$S)S3&(deLQpr!ojM
zOjdYY3xCkx<(Q+C^=DOnaS<;iK4JUTgKU5LVvm?d55va-j}*>sO}{{uS<T^UXV(v1
z)9c~Q!-=0GGn<>qa=Y0$Kg6`maUkUdqByYAXOvim5qL8ec|OU@Bntr*poh|O<jY=$
zGk!%xC3VIPV!byxJ6|+R)0XoUIv+jviXdemo^V$5l*m-5fx<i`Rj`DK*cP&<ctwi9
z$eBx|Gki-0)tOR4bbmPyE@m~v5k*qw$sUXRnK|G0N4Dgq<dND1gS&>I1q*>$uoURe
zvI%$j|G^Xxpn%g7D^h_Vr5bX?Vv?=i4*Jdj#|gx$yg^Zr`~knd7Ex9(64`$4^_ory
z-8;mNkw<oXGf69b{~ql3_@8tIv;i=X0*$E+T*DxLSq>~J0BnP!3c9asn|&Ys^sI>b
z$|<cLxO=lx^_CG)cik}5*h>1YympJl>GL(~tos1?1QBg)tqfWW7$3U6ZO(VAbG{K^
zR4`LuCqn#|2#SD{10)?iy>~!B3>a;Jyznj~bk1#L$wlO$7FZw%>a7Z*RoNeA`QKG|
z{nzX=={}mH(c#GMNG1GGIZ?t>lPSl`Y=J^OoVPmyuLTNMk-icdk5<+Qjf<Tj7h@zi
z6H_kuOjw0|(du}mM4m;mv?I}EzX3Ozeou8>{3W|h2r|)z%Sh^c&AkTxn96{p)Ufv+
zn|*1=6<%A1EpuYy0d~{Y(TG0wO4@fRZ+>}n<=k@1s5lZsmM|DM0$nT4iNdRLN*l*c
zZrpy$QDsN#KBzt^=}KO)vDta~#otz$o?NKpM#V|`8#s&fn9C-PEZaxPTzRRmlxEO&
zd83TdgWTWaSu2x6ibqBHTO*B^Z%AIt1MdvTINs>ARJt&cZti4*^Be5#K6<dhLYjfr
z$!kl$5#BjacLPRHW$LP+PXMG=OW!7Z(~oM&x!b@;^O+Na$OG%B1J3syhAJ2JRAq(o
z+cG)E8i%U2qLV~>p-chBTEEdemdt&>^}6W3tYvhEevz-K{*Rko(EFTI;?1g84SD=N
zjcGsKZ9&%n6vgk&Ov)G&!N<?b@6bZ!gJ4q;s-&VSxqqomsO=8SfwUjqoQNT+uJAAf
z`Ry+lJ^@O_^uZgdPJ|et94tFa-wRe8NepEhREQ6C@;Oa?z43#8y%`W=Q^g;c1I{M!
zk|-?MPv83Eh0lR^iC5dCeVqoB{VVbC+n;_jd9Qp<INxn>aq7q~c>Ca!?O{{UQOL<M
z&mZnHg^wBy;l*pEbL_vF%u=PI2f}KH?um}<udW(t5dg<e{dY~%-^M8;tuBxs0v{02
z1c|<-Y?`#PvJ<7irGw6@Z1bf4I651J^XCBLPeeNZTvotnMYL&+i~!KxIX3nZmWs*%
z!8==$OwVXw0HZ$hI$0g-JguUi>X2?cJ%85WlU=`8zW#LQyD#u6v^0FZBZ4pvCjevd
z1}qvl9Cok32pJg}fay1DGT!*-UAPUDF7Tic%?1h@P_HIquU63H@YikdEU5J2BVA(u
zX8DbGG{4qt|6UFcE$Fh_dq2lf-HGKRV_KmjY_T&j-58DLi7eRSrDVkxPb3I_!G*o>
zvqJsMl5lXL;wTE5pI9vjnoJqp4v7qfEKrrD-&N3*rpU64z{G}4EKj9Z4g?pr+pv8z
zx~#A-y=-&Np(o+mSU!W1O9~plXxbQ`S6{f@Vk%j>Z0cLmKG0lNb=4r-eH7?jqnv#-
zuncq&OlZ8rxf@n!`70wXedWgcUUV=^>*q8YA@(5v&V#ZiP9&kE?63i*vA$HBJ^$gL
z0@^d<d8Zgr4IT@%J@+e07ucI$1uCXfFUxq9hB5biZhG8w4-0)#nN53j>Q8?29hvWd
z+Og3HQ4uV2i%4oWhl%K^J`F>02(LerXmMQ1!r3Nhu0}~rCe=X_TV|_+IO$sCggscj
z;zsUTU+TNKG8dc<MfJ`KXv6x@Qtrn^4)$LVO9nOjCOoq99_F21T6SF!!VupQnn78#
z={81&nZa)0F&@QF9$u@)3v*&bN+|N)N)m1Z(w5NX7M2vN8s%lk4OJw$-4@fqYvjxY
zl&!BZJ}(b+m@i`yuQ^28sd{-8TgXUrIF)Lg^H)+MlGmds@9LqL->E*~QU2f#OPw4d
z2Z$#BVcvmIPq?&VCSj*z^W(vZ%~W$=@XTLN%wWpO%K9Jm#*@7Z!$*B;fn<QCDf|A*
z&dGA$MkiNlflMoh^XlTZ5(skwm;&TbZ&0%WBK<$T-to=@vmjAt1+^-mG!PIe)Xp67
zmjSy7R>fnDga~(u$jGOp8O8f&{_Pt2qO>OfB9XV>p@(i6?MTPABCdP&939oe-T|D?
znm7S4{=KSXC$KcnqCC?xA@uOLDdN$Y-u)8yleEUujDW^R7@iB9B@)r&QdDAwyw=L1
zz<&VfQNWD?E!nFVnGGD6CBW5Y&Nla4LFODek;R;zo`yVp_z;K*7A1J_uz|!sSoIx$
z+3E49iljKd+4>vsV-X1-APa*aAf!1!^)idD;gs(Jk<0lIgldtx5LX=0AVJQb-Z_w1
z0gHEnE77uIINMCo4CIupH>n+rBM489(F9#<o8ec^ecRJW`ffuNWm%!r-Rdy*izx5x
z)rT*E=O%VcibnWh8bpYF-V!h|d-z3%`~E)yZ;%_S^I&c1HpU2DVT;anB8#Z)a_HpH
zvyaRD_fvvB&tq*<rvh8%wrZCi99?~w1{0F@re*KBFNoQH(c<i66k-19p2%YHosV<w
zqTIE-^dwk`Tgg6~AK(Diby#mB=I=i#s411{eB9e%iIJl@FFtE#n8{%@{N1CZP3g-h
z1mzO%-{Z6}YqKKBgN(J!94eJsK70`wz^vdzB@8TMwMHzhok&VtvQH0}^@rF4{%2tq
z<?#ocj3(1}4rvmpg;F;@_}U0b@jCZl1fjWa-*}UMcvchpfkI9JHHf>oO-GZSv0k+M
zrOimgb^QEfg@+dHZ+Gvj@v+UP74R+?A#aBN`>#Up*s&MNx-t9nEGCRt8QS@BA8#F6
z8bmjhI~Z@bh!z}VTnpl28Im+d71l}_DXXcLz_&v7P8G!|;C22XF49RRpW|&{vr_D2
z5>!Xy&)dF_+2iP&Pk^yOA(w=(rSpb_kQW_#X@C#)7vet@Ib0j@CM<r}2srj5a^?0v
z&R+e6Kd7Ul<E8}Q$FUt*7!3{wzX^Y|BcKs@)mjPF)V3D_3L~zD6tL+s63-1VvV$-w
zcLMRDAU+JEm9C(*iuj?6Cbn~LZX!Vo7E7`!cLM7K_ysKaxL!Q!o{F|f?T@V;`SPuu
z05+#hSh!UkT*_Yu38rYFz-twMn_5I`Z$>AB3>A^B1TTafXO5{ll^;_7R|dAo5Bndj
znFDv)2;l+!p1|=7jy5>p;FyCs7JL^4v0MXclYr<CnhVUTzeWiGJ|6&tHp>%2Xn~_r
z^hVv(`Wt|LF3?T%Xw>L!g=Eu-cG>3uKC~rRoq|!+^q-+_Pd|DYzD;@^uS&*CLd$K4
zakduL5D?BOD^-up51X6w31o<(HfEc@-Er5tnGnd1qHo-zxstUm-aN98D%o;bki&Lk
zy$soY%OzTmvxP`KT_|wpmOd{qR4te4Zo!n#Z=23NI??p*zQy!@Me`gQKK!EQRV$j|
z&SvntMH&qLaBEuyVRg@k<1Zi*Lnv7_CA(Fq37^{XhFNU~)H#cJ>dCtVN_8mHU~!XC
zVw+o<4}|O(vBgz<F2#hVEUa2y57K(w>O*Pe3K&4B{u_6CQ6+Vul9Jb0^GEQ*kmA?&
z=LheL{0AtWY<z({nJ_yh-@18J?g6eqtx;X4Wp50fQie!`G>O%~so9M2$m_J1Sg07R
zp-mN;ba2A}rKm(DM+`RBR?+*M;xn#iN><I4z9!GK!?*S4n*(RcXkS2NO&nbKF>>s#
zlHafQCTkdDRHw7#3dvX5w5rf{G$daNSJ4I&*H`#|?Q>$#EkAK?Hf;N^F%?5x%2DE!
zya;ls9A|5RlSU53=8HKEkTTvVC3uE+lqd!n7T<z^wK6m_<S>#ZtCE@}ZY46tLyR=#
z$nvQ$9E#Z(;Dzw%Fn@+$)|yHGLqtZkn)xY!w)fX0s$3t48W&7vLEl&ol~dtso|;(U
zV-8$s==%H@JQT5^Q9T8=Oh0^cDuDluH2D#NE44c0Fk*fB4=DRRpvmaFrR`!YkEaR!
zTjMRR=mnx4GU+0V&Qziamj%eAiMO<fVNBp(cUZw9AYp=Zt|g<%673HR{TD%l78rBE
z1AwzarIV%t-V>b9f8Ix!``nwsrCc0#S9;;9hTtz>*<C<umMD`!n%(r)OOzk0@O5D8
zTO2|{xNK}}4mMipH-AR+i0NC0KJTlFr3d7dRVfgaOX)#Z3;<ixo3ot(Yjul6Km-Kl
z8DLQcAPLONl&y%}q*db@?X8oWMMK}}0{)>g0PB`Ra4LWPB1P`ETFNGh{GEMqOI>#<
zM+~7YRB{qIGi^vE+Diw(<U`94xhc&<rbTYX##_llN|{Sj2}^w&i=GKC?$~~DAuV0l
z$D}%<yiIg8X%|^md}M)1DhfW6jz}u{+-9a*g17Jy>k_xWehwkOtH$GN^S%*&{{8vZ
zAFu9}=+?PT_1aDRR~K-hGNi~|n*(c$GZ_jpr)%E?)fw*3x|(ZHb+^c{#c8YyFIQ4X
z?WhrryjWcPHKjj-yk3?c9?jpldEuv*`&xGFH90mJKd=w7r%t%c`+?1smV4(+<$I#f
z#2*b~%b!9Vxe(y^=fOw!TwEu8RhSivQ?HiGdR*=OR6p*0&2uz(4x-yT;6mYoF<e3U
zacLD($K3D{i8@h#&XOac<%HyU?apYQ1QG0)rJzV)QdA6!rFy<9&6TD#t!g&wb!BWT
z*$q82ZPeCMH&IQhvvAXfOtjPI)x;ZF%i-i(Vr0+GJ?=QHA^pNOI+jn@?4EtEzjU*<
zLg~IVluOuI+xlZL`msl4r_XJ3vo*X!T+EZ+HaJDIfue|>LrhG7j>Bi)WU8nH_|=w?
zNDQMU5*H`Sk85AUYr`(bAxMA5F_PAH-|dbOibGI*jq~2zP!vrvyK>MZA{86D320FN
zVwLx{fCg20fvHf%>nqHQVxH{5-MW}8OFDm(K-4i~|1GW&(SR?B&gNvO@#1$LlRjX9
z_ydbHnUx*qvG-FpHcrdyYua5o*c!htF|@e*#-A%UxU~&#zNQZ@X278XTl##ND(D;d
zR|$@He$5T=5_n4}iGYVgL=U(k@*rT(YFPWFljN7a=3t~H7+BYfDP&v)<FH&t4@)hm
z{st<+^{fF4u+*E6a+Sx4E5Ju(=P?kT?{hcREXzS4fb)+iJ=<LWDg5A3z)-<N&|xES
ziw_W+L8ljSR_6&l-~)p?h~DN6-tF}xnm1+$cHKks$Le4l_~$$ZHfEru5lN!?tOhWV
zXJ&F0mx6R{>vdKC*}mT|){lV<anJrf7-4`vi8%d(fGvp7T4lces|@JkFN9WlRjjE?
zfL@7XApqtJMy&iobCg6~ZIc8+3;C<sK*a~9!(9=Yib7rMEGHw+t>YNAzk1VA;kwb|
zExf}aeI^d5QK1KqjgLmAqOX(PlV%dA(GZKX`Y*M>1T0FEx8%UOOUlyQj_B<ak-Od;
zXYsw^l0CZcnzF-I%5Es6igwPu)a0`2qqqv&u7s2f3Za!<ROcTvCDc#nQ&|hLjT*(Y
zF0!QE`Bm4dBGRNvWTECP)A?d*A;XmWX@L{&DC*)oM~bD_n-zhq&^G1Z!lCm{&Rx!&
zZ?wIuGZ_!wUV1k`%D!@17se(@xSHa#UHxmoX+>%<sMi+S^jcyiOTx&?vBv3BX5?eR
z55>ceX=E0kpEs1CNyPf}s98rp6&xv!V`W;LBQ;(qmn?D5h7bxKaA;b?Ru9*8430Ib
zNQ#T{s;dtbQc|lWIgQvZh?#X-?N-+-($1mVK^Q?K;>kw^@J>h!Pf}}HV=RL>F)~Uv
zk8pn0vX%*AM#u}bte<nDn3tXLC7)(-5P)ct$7v#)yOAzB0&JV2aF@6doNn-%_W)Ot
zz?E);d0Br<#o*QFV0K7t%Y@%GRiY7h&Ay<|nI@RK+QN0=uMA+yGQRYY<}3SG`aJEE
zo%@eir9-v`4~po|b_xf*(e;TxR8Jv5A=s*O`279ZkGk+UoRpz5kuqvg(qqy6HxLkK
zg1$xSOWbtP0GV9YC%zNZvjv{=?}D!{ySC5Y0Kn>WqJcDd{BQLK;6{6`t<TyTB42{?
zXTBVXnyw`mFsnSGus{5vV3m3K3gCrU!L;Jl|Eokf#YX(_=KZRhz|;dCD#Gars6YTf
z&moS3%L%A)QYJ@_-TJnP3)ukFud90kEEs@kn#VO%R=yzs?>^YYBbd{<<U#<3h_fd?
z&wvpQ9C|14dI3lZP^)j(-~5&JJfqAQWSAjk(qVIO^*yJWnYI?i2ZqmXC8_mAmxd7<
zrTJzVLZ@kJ<`6n?N|90e6P8?Da;Gt&F@bl3b$a!%(MxOZzfR|K@-A8V80QXnGw+xz
zy~Ae<4&;Fit8vNJ*<2oIS%P-e=58rI?$3JL0Ma4t^^{oLR2!<g9Da7SPb+VIz-x=_
z=gO7-4fOPSU<Y#X=0ua%c-r}o%7~)x!2G#TyN64KCGCp!RsJZp7EE&w-1|Od7LrPu
zWPW;MGxUqaH5ogLTB}mW;xF+dARZcq9}bP$e)M&MWijhyEKBn8*V0WCpMw|#akc-h
zJh|!WXB>|}!sR^7l+h(6Rb_<-Yy(EPkgFkab<TviX|zxl)kVQ`Cn@7^F?LiEE0{8J
zenVcX&n*==J=?Gz9Hr_^lsce;IMV`GOC+Y92vvgo8zN8leQ+Ai!C5b3S#WW_&kQ~~
z=DYEpd$c&VFg+h$jXLe4V(O}UHzv>CBz9n{Nnu5vJR5SY;QgzLvb({Sk*fFLip?=_
zgPJ*g2;l~$CWkD8OL;^Ef(7NC3qd(heijYu#d=*+C;3X*97FkLL@RJrH<;96Syygb
zu}3raqhOc!<GQ=FMkG1*Cj;yg>{~$8UCY3K6u4BLfkx|HK6>?0bS1GO7yx0w6fr+P
z>ARKC$P%)K?>b00IzDjrcO!ZjG7Xz{P9H^s2_bmxtjTYOuJK5(DLQ+^XDwIZR$Tdk
zt9=OY#enHl=Nf6R61a1T0jiYz!2+~S3Wy8XNw9$jVi|OBv?y7zOSB&3I|}9!FM04;
z@-^$l1M^MRqNvx+OTD+@B|(KasduGKf4CYB7Q~B;1@Z^mSv2-DAH_Ev3;!7a28y8<
zrMLKr8z}zjw|X_>x_}>g-Ue=Xw4k({0mQ)0{kO03$d4Z#Ksv6jt|r=th;}<*^a1n`
zm>K{L1$+&nnxjj525Z|0pemw){SLf8e(zO+i+%81T6l<)m1EyN3#9~Ca)>?Tb0d#7
z{=9?)FA6Gk+~}+P?A3zihjXnrX*r`eS~Z)$Y@!)Rsg3F6^p$8yNJDdG3cZ~aZz`(M
zAgn#3r+;B#F9dNCQ0vV_5$J#viE$$ci>C)vH-I$|y0arW`ZgM=*Kf#FbvlWRDeS>J
zxLF-+Y%<R^t7>4~8%gcRhuG#_63l6ZvQ;kYLPoN*t%Dzi=8Y&rtdeccJJ@rRm%L0U
zn3b7c)%6i24M-zPGVGK|U@mp0kgXXkOwFM2P$N0CSaUU&e&lYU`hd6l>ia+}1eG*i
zexpXE)r>s+gI2B(`+|F5$$hqi|Fl|Cai|}<A(eE5juTo>{Nup$TyL2bt@jo=nyp|2
z&L=8JgOMBcEIuVAw|#mUHz~g~69y&~y)CP21@n2Gk>n!Y*DTjMkqr_*oo8&rzTLB_
zCAW5ilyV#P8VCJA>yB|Q?(1?_3u+e>WE_qdzCOQ~HeY1pWMjAytJw)Egd%I?%IShc
zF)}3H{1~M7jkYy>8$%c$e+R$g4Gd8Rqv&x$1XAZ6G$1Kw;g`5!aO5;?q&b5I8N&G@
zL8&zo>XhI_>(H%ul_S4^{VUE<GacDZxdk_6n{WsYKWcB<ek84Cgu|7X|3yLT;T2K@
zbI5;lKj)AbAtoXoVgT=jNbRA&#)foYY(OnP{^tPAE+rDDZ+Cb1AN{$6srW`?V%M=9
z5yB}Kl$Mw>=x1PeLJC%w(Dl{@>GJ|Mh!_P2F%hgHjp&LeL8gTCnIP>X&{vmylanfZ
zO~?oAUn}P<?5)t;51H*y5?KlQ{cF-9->WD^lNn|SUK=-TY+A?0&V%}!<=y&v1|lZ=
zZ%frg@j?)AK#&vN{lGlk+Nuo#TetsNgQ!hGOXV`({fm%^sflr^zU^<v#USYlJ|J!$
zTEI^U1MO3g_FhBci_ZAFFjd7%uIID+qyJ<rnEKwl53MlGqO*MqzHnln;a!EM5?_}q
zRDfb5Bt)&Q<;L)mLj~AmtGa|+rE+yHM^eSmlJ_O8i+A^DovgLUOIs|i@gvR3uPmoM
zwV)QZm%`%Se-uNEvY<Htj>#*Sm<zwx^U7DfC}IQETzdBtI^!vF_4uRW_X<v_3r53l
zo3aKc?wzLu%I~lfYhRjoe*0|tG4?3%u_s{xUX2y&6%5BmcH7>xr?~4-LydLsZiVf<
z?ENn<un6WQeA*qN51h^q3g@5u8T6(K`knukvHwvL^K$o~eOD>kiDbk5xZ6brs^9XB
zT+uq(Eq227{j?m%#PfJ1S9E-USu`nwkQNIlt|3F<6!~b<W=Xv)fr83O(pHfW9`%Ib
zZdn=np3u+4Zj`Yyt~)7!+VWW!Geh0M^Xx~*!R!(EJnL#s=42a@IK!wbJ~wEbLv;H$
z&FBbZ>hLr=Sy^u&RJOsJo1cU+Q)#BdV<8eT<w{h{UzNgK+OTr8ToHG=DV-n!3>~zJ
z%w2LU7@5pK9T4Ppz+#+!$r?SmO;F5>8*Gc!Z!_iuTYfZiC7~!t&Yy*KWyMkO9tQ)i
z9&l?;CJVk(jpea61#F=SFD_Y|Qq%epfp!M&Z6*BoG(VY}%xIgy|B(UA=);53A4h&C
zzfDf+h*Z#Z>ohVEl4iW)Qz8jufA+iyExL)b6yUm9GIS3z;0f7c)O{YL?_e!u_yrU$
zvYJq&yd^0rsfdjS^W^=kPz?K2Fzb|R=>4GrmgV1H8P&0L<reo#0alUhh)i+?G-o0j
zT+<*N$WU9F10KICOz?jTmcebTqM|~~bOE2I*Z*WI5E5{5(mp)}W;9Td;Kv-c#Q$ON
zcjAQqXHi+5Pj*&KW}tc+1kS;e5p+NGQ{wX9k!4p#8(V(Cn;>c3-2Q~70uulBqxUpL
z0h)^M;pQ=vij@L6A@K?|Bu)Mzr3lSH(ytZsl3du3cWhgA{#&=|W=#}Q$ibOh-)fAs
zLp*{mjl_4`r&=|@=G&Ug^xW%)7c=u=_mEA7;rJyj@q|U|iN_v48(t&DARqJUyGW8^
z6IxDchtMh}7nm=8q%C_HZ$^HNUFqQEaxCvq#R$qr#Gd`=V93Gpq4DGKdk@v_cyu|q
zo>qPh+;$*4?n>2T?R0ih6#yMY9h)ZkQZzzEXJf9~D|LD9AE^_r%4E;V^?#(s+6vNh
zls-L1K*-9L>k5^#{P>K4X0Rmz&%9^e@}y2NOEQ6!dU$-HTBS0JEZ?H{OQ-m<-uW96
zo!99)y))O|bt{>);6+EQ#09b%xnJd=z9@S}#*1>BeD1E-mQCQo&hKrCF!>0+xZc~v
zWW0DS*SiK_b;dX3`nGnMzMlIolV!PAXpcL;nRirG4=1F+V#e6Mv#o<&$sOpl>RUpd
zLjSp~3{*=<xgaSyKMMQshDXQ1GdkDsvW^V&4zRK|wR8k*BnBQ2zx{>uLvRO%aaK{F
zH#!dCj00xO+mXmwxyzpq@txr@Y%$RJ0FoGpqR8k{^_N?G^)>_0c`~p}<u+e71aGZ3
zugs)x?QaI$g!G+-e}T&dJpIMB&pY$~e-75XK>g2f=`2XcdKVz=Pp3Z~JYH{%YC0@8
zxpmJrzsQ_bxShX7fJ6JUbUpL7W4WpQ0I~!48(28!efM{E_nRXAY&V^zSma-`F6C*8
z{o}jHTx2{u9E$`zB`B#$LvOnOZ7HoBmGJx82LOm8s3QPlFsLU3V(Z`%acl&UMS_~O
zI<TIByB|2;l8Z{N4{vrYJ-E`?N&O#tHz0-f0^eJjHw7^2ek~mIBQjFk6)o1(6HYVi
z(W+Asi-&~&lxX-Uv?w)b8OUFSKVjA9qzq8}E*%vsn6qj)t>h^_fh8gJKITn6b;?K0
zci_U3X6+3IUQNg>POk1ro2O+~Za;e-Yw_wT>u4j7fxms|I5Ye5OsEx=P{t)G5+F8p
zrvfmYSC;pwvv~k>PnV&Q_Moc0@Z06sCW(^_op-5=6%J_m1++r}U3lso3Sv~ru=;qr
z03sxPVX%5iECQeIT)3mcgT--+C_LNyS|WyZ@8Qe5w7fkVxFFB|pU=cE4Em=n%N(DD
zuTNgMxp{xyZBY`+an0Ky4elzV)s(<tER2c_lbw0lUd-qy-`pMK?P-K-txC^taUPN~
zFZsxsLAb~(IqlH1srbP9nH-#BIY@_FK9%8vuch&r5XO^5^2I4Lb0cHI6+UN<3z$OL
zmLG;%Z88*HIM?wpE!UyX=if24746-=*2NNH-^O^s;qa+<n!C-;sio($o$9R*{FX6r
z{d!@NIG=tPu$v)-FUvwj2IJrvnA=yH70Hu3VWu64P;E-~c|MV-6e<BP^*zON9GxJ5
z^&B2=O!ZmGT0Mq^;ur2krssRW!|Uy|xT+i4*dM>BeXOTR*dl$aO;E<F!un%)lyg@F
z*7LeQ{EwUGg@uu9OceYvyLjs^*Sl3;bE4<wU?M54w0mhMcIqtcph@+J@x#f+l^K;=
z@rBI<d8Ck(-Al)1dJ5sGTOp^X>(lKD_SSyq4VS7wdfYzMw`K<fQ^3pzKL69K>W}LI
zrac722j~$1C1gQhJ?jYQ761V(Z!fPL&3O>P0@@Op+<4ge{hGsS{D;2Whk2^({>of{
zQY3B~BxZ4Qd)xowhO~Ts5e^3jZgWlP?Y7FuB)@$pD1B#{7XvS)HdG{fQy?lO0&!i6
zz;tssGnw^Zb)^rPyGnCMDMstgsNN9%tWlV|n`AKbE>$gu*OrWfjA`&IpY(mLd9ST4
zr3+lT=R$lWQzt0+FJB8y_tTDhO@}%z>ZTgPMUYXxSA0-~OpCx>MaJ@uF3;k4j2kHd
zADFUu%hc*ua7KxE_@)(^P{!+<d_N?xwtmGv82r*EVVimC<4dQsh0co?<K7;PtPIRk
zlYih;7+sH}Nl@sBn?mL{-=Imp8<|Q2ah%i|B^AmTTiSKL<MyJbz#35@)L%csYW2A)
zZx4@6SAub1`hsnl-Nm-kyu4s~olcA<1Sfq(Gh;kwJ&Dh-meS0%iD-s`GiP7@Jh%+o
zW-VGLb&a5!7x^TB+EXa&519d(yqiOXdMT;L$M}y5Y8gxNbx(Tsg`B!nz1(-}rJlJN
zk8RZ#FdXw^&@*GV)Qx`z0NLm+f&;{a!=6kU(lR317{7+YZ7{c`cv14u)<}2;Lo1xt
z2@_8RmvzvY79htlBAj&uAOe_<w!4%ioJqpGFB$HHkNlTT3C6HHea|D3GCe<nBU+=Q
zuOiy@QOs9AFpktXwG2IMV6A?!xcdW8)`{%_4A7ZYuC7M~a~~Q&@>SV%bg5&RsGf!%
z)iedo&o(^Rp*#D$;A5r!Y_L7WPX-_A?aLT|_x1kTLhv_XG7WvCZ0jdxxCHVdR+*nO
z6OYm_YbuEmEiqH;pp*rJB<?51JI_oj3g^=l+LabvAQlP$ovEy{GNQG$l_jSE*u%g{
z2Df8q*5v8$PveepAh87g1mvyqO|rqd;(P3-;E^E$<7?^75@dq17^F!Voso~Ri;)tO
zFve^3o&<D`mjQu>Q%9cwl4AH9n%a^MHlKxqbiji5ZqRZ(EVrp#Z{1ge%62kT5K|b`
zUa4LcX$}7Rnp4}Yc#Els<kRp$G4T4w>BMaJ7_feKc8V38_6V68J-*Ww1xGot|K1i$
zEGsZ>!HqVgkv{d9WeZEqlNoOuFf6@b#P+GErKW4)fgQrSR663Vx%ilWWJjP_fSQfg
z%u|O7J8jgdLvO?=?6o@&enJ*1KC$XBCPXD7O(WFB=%-n?Ga~+_^}3XU=Jv`>3P6C*
zU2o{FsFRqo>$-B9P4UuigrlRrq8utxJCs||_|}Pi?ws~&ZakX?P41Y3Yiq?13uSuY
zK;hK-yT7pFzm%WedS1DP#|fL3qn-0UJe0ZX`NfuDfFP+v!!GWMy3*1A)8$TxJAHY)
zEz>L{vAHkm7S10kg^8)kG@~T#`jQuNlSRQlZsesDM@{bgF~`9ZTZnSrMELFQ8HS}{
z(r~Y<sn>uD9YS$Krx~ULTntbwGlFb{r5Ovf8vf6e4g!vFD4bDvDJ!^x^0Qj{V2tFQ
z&x-LJ2ox?4;=)jZH-Fs*I*ksKyef`oI+l#T0931Ao*@$*@JB>zo`dv`s_yygB?7-+
z+eUDi?eR{#Qi!P#84H0YHYCu5x}-m-UlfCU#MAk->1<;AU_AT4hv;IY!a$M*H6INH
zo&LH7+BNt`8&`+EOVfCw6TeLku>UwBCk7`<l;eGtY=qaHSb8#7js@?O-v9>`NH!pI
z^e-k#TmcXM(C68bZ~d}dCeE`8G#7w+UofB%Coynl0mZhuzCH`o76507;nTf6d2pgZ
zR-*pbiYG(;v11%q2n()R5IQAVVgN88_LlazOxxD#N`msyd?JYTbxR?uB*ow!zu1k2
zJhrA4InMGG+9SiQ8UhaW3R@R6yez8BQd_c^@10;75D%wULQXZczGN-N9>lfJ25pc?
zlL@(QZ>ZASO<YbEcBtw(90YS&SEWbqg{g0O#q+v8SDvQz)zKPRI*n6Ea;>O>*ZSq#
z(c#s3ft8PeD$;=ihd8!Vf<oHx=a08Oc)s_K>r!`{V%*Sv8=#JjQrL<t{Qa?6uS)tj
zn^wz%jhgdV+_Dp)9EO!(7ah!3Y%<a$DU)h?`Cg`JW^|&2(kPYi5uS03RX1mAbYa1)
zqMmM#+X(unM^J6BuDBdnxxE^D^4|qsTUz?@ErZKt(9eP)vSs%?-L;G56y5z3jUC}*
zBQr}L6(wEDJK`<3jQL-ZJ{N4@Y;DjX<yMi5c=15>fhu`>qRgV0)S^KqT6=+?E~fcS
z?t4t8)4p?$Qo4*LI(N&+HoJ%Z+(N2S+jy4WE8~Us>31IP*lwrB5w>yi)$<0^6U(|e
zPBwI}-^T~{vRk*D3g*nxKnb+GcVBhSyh#Zo;dSJN%|iGcV|a7s$l%Pp$X1rPtNLaB
zt{3C9q8;EcInquB<U3jiN(ai<jEq8YPC$U6#UhPi0YU>kdhBBY;gIBQ3UI4RrU334
ztA5b!dI$<$F7lx!uLY8OzbNJlOODou%uD1JOB<SwM$CTuJFlgDo2<x_*2(OBE8TQT
z{zEZ{-Tq`Qe-pnp45;Y(&Z85Xrf1+Q9aZuBV_#Li%cl;`7gFAMfF9UYXxi17$^P>d
z1h;1|5l{ZfVNd5j;QDiV6p(Hp>i;i9O+46ti>6DJb%pb7N<pb8IP}1v4uaY<z)wif
z3IuU&KnxQ@K!4kNO?Le-Ta#(3^Z*_Z<;GvX^ttFkgc?}EDo48~5yTdk54pa-u1d&4
zw=P9I|3rEv6_o`GQ|lR3^j4t@?$J_uU~K$Rm_}czcnCmk#^FnmL}bg%j+$aOCU_e`
zbz?W@rA77iZ}#bhcXucF#CBa@Lz-Idoc(O@GeZFparSyxQc)wlSaebvMbg!^9HBrL
znGXX2x{v}<TaG(!7seYy+RKXHPf!cFp0|8F-8DV78UzbSZT={TENVk&z3bX(x?6&0
zCr`RM8z=$wd0tS|aE>je#!>Z>ONr!ZoDT}~blV99d(E45+xmVHJF_HGZ<Kb%$7Zx)
z8H47)N^~-c<or}kA$)=-rh3fzFh+_$QW6?e`EH!-0{wY%V{~)ZZ>U$!QESTfk5LF)
z=zr}d9<Cg7?vV0IWc;=z#DfHqaNBk%xcGAHcRzOL+gCG0XOn9g-I~2;>N!=!nZsV|
zYqi=|Crl<<94mSzqm7?sW=O8#<u$E6W%JwAMrnl?6KJsa;%Un#*oR6!e)%m<Ss+#%
zh$YD%V_k3xIET8FOq|A2s&#*pE#@pIO}5NLV8UXMu;wQQ`_4}R`GCAjt4!9K%w*n=
zgkL7b$@C}2WFn{IwCjasftsxazqFHqxpOJ5=Tl4{_}e1nFGR_Nw4sDQRUdg5ee>sd
z=Rb-*BQ@Q7u1Od$sLNAZ9`kvpcJV`lCQO`oCo4^Ado!p~h}jqZPz<I2Q}k#zVPim8
zBfjZq+N4yO3s|%iOf;H_c0?fN`M;k%8~&<tu*`GH9qPCqUFv5G3W6eD;@pqs*Z=gs
zrvQbEyG{VvM*S-nsu^zwT`^f|52FBTC8qcPh1$Z_kBf{4@k^HaP?ODe2hD~KaQOnS
z713l4NHW#X6F9Nt2m~#F^7pTB9q1rZS%arw2R#tSMTO@R)Q>2M6^&|YWYeCSK41(h
z6ys`i?jF>1W+SPd8SSd2Hxx+^tWbw0h1g5busic;(Tc8kp&6nGT6LDCAm0Z6oWPaF
zid&`8LoFbGzUYPrPccs|$#CH=!%QC<&9A)WT~S<r>fk=KRZN(kLp3{3&v>C3u@cK&
zmz6Yd?xaacS~Q)7)Izb|-N?eYTcn>X!y+r^EkNl-c~_^Ac7iU$JHKK<be#lZwA#K)
zb3<&pWXWnE6$&z)3y!SZvUP>I#zpqj^^)P$;tTd<M)LFbA4%HM8nN}4*}2-dY}#g0
z7#Z*PEoK6NZY0%>$yB;$z12O9G*}9jG+#j_GCaPNQ(Tal9O)MCcfQ!|I7N%`O>^Kd
zrP+9m;HPt}Ij$t863`zP(}I)Td*y`}Ue{I_WouKxx?p4?K2nEv43R|@kx2#-p`xDR
zP{)C0r=?n)pGn@xt@O@be1EE6?G68HsZxCf<t#dxWXV<KRfmQXxjoZ$x2fZgb9c4_
zMCa_jRneMOH>smL<4^l*b$%&ktk1Uz>ci8++75La>she0RKl`4kfb$yIfqCO1XN0T
zpTOIvHxr@4q{PxgSQIz#hX95FVg$cU>oDW3F4VCA>EZawnM1~?%b;=ZeU3BRA_+zX
zF#l47uzY{h|7#4ffTpZK&TskvxBl^$kN+uN@htfld;W;d*&6!g1GwIm&h?qMCZXp&
z_6N6&zvtK=oi()r4%qJNFIs6;U?!yB^X~ukUW9_!3DlpF&-1^*EbTSW32*sP`(FYX
zA-hs>#Wi3f1X8j^0-z9hz3dH?lq^717f7xFYZKrN0BP{>;e`YqjSinawdlpHhjhf>
zztKx5tEv*Tg-1YkgC@1``=(p4CSNlktbpyU;BM9fmfC2ZP|B(OzdB@*@pq8kS?Ulc
zI|wpr`Do-LXsjeCj<8nb5OzSrL=6|6sKmRPCYe+%=?GPbs29M(aqc#)D1|lqykgd3
z;0B4t1Ap>laTnSWWI1F(OGz_A(Bo|edK4#No;BnVo&>r#^Spt4(Vc2frr1DP+${z$
zE1Znv#&~K=t*#gP{Fy{_QW3Z7<FX=Xtz5e(NmAbh0UBXxvJpEOY;o82Kp^V)iskA}
z+oV{?&5MgA+jG}wIWiV)Sd$78lzs#@&lPU^dL2wCIm;p%qzJb+NKH#%LT)MEJU&0F
z{W3_r=|1<gd{DN<^<DlPXPxG#X340EyO;K~NYZus=W7pq+I$!%v#R?u^O=h@bGC0R
z{(S!ZUz#3Hcou)k|H|$T2BEhaDZ%T!v|-4^J<ORDNXy-&n$lVhOMU)-xccvSD*XTd
zA3u)GA)}14WyUcQ4l>flNw%_Q!#qaFCM2uOI7U<1BQxpPIaZ~N?45{^Y`*tX&)4Vo
zzWkn-bLkJ4r^~bF<?=X>$9>#x*BkCioM=cYD@VGZo~{S-ib!{7nfXy&EI#MOSdqNy
zo2fGfv$Op)%<K*2l)64fFQ+ObTsWT=9LnIS7oF>@t#IB~uh*uY`Lc4;mGgql;kTk4
z??fi-1gIFP7)zqyTVVBuuc7@?J(@v;yi{LSjeG(X&Ps(xD%ztoZJa2qkP*xr86hwX
z*0*tLRqW!lo;df0Pk?8xse+oBpmFwKd^zA#sW0}9-cs~O?yuC@TM|9NAox=vQIRfU
zU$Gdu-00}_@g@8D+~?hOaQs3+nb_QN-FG6d^oQM20?EVntMwrwIn(1=uOX?O+lP=L
zgrceTN+Z|}I@DH;jgm;pYQyg^RDd&^`V!7=gYm)U#xK^l-RpaGk8;k#by;x@aQU!)
zCVjxwUo`~O2{94-`E)P^C$Td(tf{~z5VRbS_XE`~sVoLf5*;>>WL@6a0M8H^&0May
zTIP~x)p04jUsj}_xw;eH15ifR;FACYCOCEKm*1`xK_{wDd5Po88jCc>{Vdx;{an5y
z_gNTv)>E<E#gq(egbi?;szg?#<Md!oO%uvtRP@#0-Lh5*=Kc(IJ<i2wZ2ZTsw9sV3
zOGxz8?yq0=)rX~6sYTl&^5eWdIm@<5eAhqx)~<7LUu-)?b@XfNZeE-@W#AjnuQ(G%
z?aCQbuTDAt>q1x=e4R>Kr|kL`SI5NnXvET^8bdPknN7_20Ag`=(o7iCceWQ1lshu7
zH>5c7QQCdCKj(hDwPNxub|A^n>w_X$QhrnCj8)mq0*d44z|2z5gwRrAPyWm05GUzk
z69vt8!BleT!+{Jgr7Z|g8C{9w#ssZ~solAvH~QsS#ZRu7`lp%2om@CFZs(D;Rg+>L
z{#<SDxs^4UU?MU)c}n)7_}t;J4=V0zv+k-CXAZkR$3JB5Zwq!mOy0)V#r&9M&i<Wc
z-+fauVzVucsmt5h=UmIYj}Cz*O8Kb1j7fggDMc$0cWwI_4wtJz9C7haD_eYD#4Br^
zxrdnzo*rJPQSW#Y8?Rg9!(J|d?+g!<X*z=-Ixz~TDP8VV%SE(;?&t=E)dw0)fupjG
zV(Oab6|KWrnKDB_t4Bi!-x*C}&c<26Y2KyG=9j0Qp}m`HvAVi)2!FS%mpc{e;mYG?
z1!2V1aB;v*h$%b_?$8Iw`R7ZfYM1``vh?pC3S~5?ciNlS<P7{Vw)=jiD)4*HcscJu
zIW&0ljJLdkar@dUKd4=&*z0$W?$5TVR)qRP1&zLL>9cTbb9Ut}l)ifZ$5Zxqg&b`5
zcH}(dRKTD4U(W!Ajw6CG%M*9)42*9<RSHr&QWpmcWRi;XKP}xqfi;-Wel!-}jUBO^
z-x=KqfSwKx`Z|0oYBJ7|3U~bKWo6L?^oOHS1!1jO*|-aTetQkd(~RG>K(SG~t6b%w
z;18A7b;EHSk0VZE7^JB4vVA~V{K>xG?gc&f#}lFEJ;ox2GpE~Sf8nlnbOgUB=EfG4
z{4_LA)Uhhd%C1O!^3zvif@}O*f|M-bY0V{!oGFdoQ6hRN-AOef?@UeW`U#v4Jn8Fy
zS_7W*()(2~POvrpoW;plyewks#H!m5{eeON-BF6G(`I>DXhx-7S(w_i(h<_Wz2elR
zf!448EI(Iv-;UPVHx)y-AQLS2<%Hhn?PKcK^ii0I02Sru5|g)jUW}J=68g;J^#im-
zr_5Z=ZyrtY>-cljYEbT2CbjF49Im&l`QD%R(r)vWG@QmzvTG62a<)CX7#?3x2}wNl
zE3M6~d?~6^$gPu=efoG-O6sE4=N1fEp_jU8WJ_;bOh90)IK|4(k+Prbr(N{}?|d7Y
zZL54{b)GPY<d*Ufd^70#?OV0qP_+@?j5d;p=a_0%w~C^XKx%@P8p%c>Pu_$gpkgj5
z`qDj>q0y9%QRfKl<plWV_Za4G!p#p+Sc>KtewKZgALl+LMLAg52-hMAwBw9ty=!~G
zF3AmcNld8X7Uf0t8QSpMs}wiYec9mqrADj#x5C9>bhqf%J;t{BiH2LYKO0C5nc)9x
z$WFyy^Y`AItO_D0aUdM~4i$^>{;dlB2Wo(MgI51MEZeeS*|z;K8%AYS*j0Lm`R5Y-
zkkbJAG+6OMeTs6Xxr6a-Ai!H*04XNWwSkg>wDVnF&LvF4$N^km;4w1@y@g+@)NX?#
z{3C*S3r-?b6DK}=B$WWOT9Iukmg?^@-xa;pQkGiUYuyyE-=2>=>h%h7GyLOE#@$x*
zL4z-aoD%(DUE13K#4FH=(dx|YGY$-VqzA_Rm@B?Nd9ZI*zzz5McG)X4bT-VcyoMjG
zXs<bL%a2msyw+9`;~Z91%%Y7Zexlkn601s<Tu#S(owH!n?(WZMEvDlzrcPW7pvMuH
za>WTJwYDEVDLnF_%l@VsyCGF^`HIso*K^XQXx6Kzq%j#3LX?KB?XyYud``?HY$?4U
zx+gk0)?W#{AoSaVGp%4l#yQbIMilRo>WA%3c>TxT3g<mc+~Sxth|Q$ba~{f5W*OS{
zkR@*c*VlT-qx@EtwJ8T%W{n#D2i2Se1*#;k4I5;LKvoj}ye3}ovklD$ldy21vrt&x
zwwH;KShz&}aLNu7FGt{@m5}tfAbh|PEIcVjV_w;{6hG|Ou#v9RY5vkq!lSUT8$w=w
zx<)2?v>EWn2L!4=rnAY2yS<}JA$=Sen|IVq&Z<+a)FGNe32bNOs9|K#MU%#&-#N??
zPD_4_Tn@yV?nh<fPBO)E<nm9!fLB_;ih@k!#B5xY17bFY*7hSTc=!9S&d|a>s0q#_
z^xT6o2p$Bl&HG50-Ny3|C(UH?v*>wa!Tk#Vkj~_f-^Z>TFa=Fm?p68BCuX+>dH>2a
z%8cuk>?0MVP{mq=G~KKyZzO3KP6YjVul@nwk)7D<{DpKfNys(BTzC6zk^}4D_r%7p
zIR#$Q4p*#aO^^Kxn0V8>9Ow)fZK|q<DOtg4Bhi}Xx98S9v7<^A7J^LV1$TMmAzqT=
z7;Gp|my(JD`HXJN;A>r|eSxO)cPB<7fRDILK?>m7mRaV55gkGJ%Ws$})wOVdqv4)%
zDI<Hw4By2ZJ8twJTY@5k?CqYcO7wzlf@upXaX^Eq@zC)PTmq;m*}^gN(8M*ji56RP
zy9}p<oa6(^2j<cahSH{oY{RJXW1Ptob=*H|JCObIy<gOx)j(E9vn!Wu=%MpR7T(fX
zI^Cjx-CMycWkcRCtjeBbdn(4Apl|K?l3Xo3v}(nabLNwbYgl3`c|=FU{P3S^qNDZ=
zb~WhtS4wAJCNs<KycYPS^L)L0TE~D}8@paw*IHim#t-dx%v?-S!PuBlcws92N32)G
zDb^dekQvU{8ewXCR=g8VOTI-8`TS)aYK4A|`28bS)7#fxwCssEIjZ0&3N93uuG8qC
z+D(~?-&<z(ey@y1SfcM@7HX!1%bl@#>es-$$scxKzV3&IqqtuWTzZUG_}ohQ$lbHV
zwEM!+$+C&G_U^~d*=e&i66ZfTdY*JM%gQ+Vo<>QPI;SAr%xG9o{qg4x6rVIwGo~;s
zu}uI8V#-jCuKU)Et+N!W<Y|aRP94Q23M;{j5pkb5!kNjBg;rBMisZ3wR*Y4ryWE%U
z_z_``pcjzXZy59FZpUKvKt(eZJ(2|T*}7mw&dVY*M?a&9wHalMqFN&McUDQ{SDfCX
zjd&&Pt<_Ug%=>-Ifefx(?;l77uJ$rcdS1Ks|8!+t&(L?!n}2?Zkx)7NquIM!#!2y{
zT6CVIkija6I<9Z8yk*q0(0fUL6^;`)N50;A0w3_$0!3%gNdTn@`j)T#4eoH7Sad;F
zOlROyHADjD@Bo9D)JgH~47B~)T(Ko7!9ZZp(9WpqEzS}zMYS+D1>#{?E<gwbIKKWu
z1w%w}v#O+{sfCe#3|L&BPI6F8((@_#7vxN+D;<qYZ*rduU0h)DK#6J*FeuHL&Vu-7
znjNMIn1XxP%pTmuk>TqkxXg^QoIOT2Uv8H%=bz~_QeUr~n11Bnq2A}jq5me&Z8eWD
z{gi7R2jJq47Zjrvm7fmO4h@^(xC}z<f0lN+=gk<(<a!X=XQw{OMpEWJ6}Q4!^WgaR
zyT{=d7yX2!Ip>#8=7mjn^xWU7c)yEhAsf`f4%*&}ikUk#R2Pl;&_84~E+CDeoclIz
zGfmc&hbS#YCx)QAotetG1^R~4CUmv5(<U7!UqpWB$!<XLrAAKdD7Xw9dhX(&v;AG6
znq^MLt)Pb0qwAgGf+v^ehgF{hj}G6%<Q=+<ow8Xlj?L-R$PvA)@sTC4yEU~ffg(kt
zT{RK&vO_kw%pN^OwVinN{Dw0*O}+MkgkrX&LW}&FPx3k65X{#n7KA2+4K$J~w+8Nc
zv|Owa2?JmgD*&6|0k;A()JbU?>Bl$dB4|l|s9DmIivV|IXRu6EEaGU$CJjhrXgMC=
z&?bb32r~_5PA|YBK?W=%(g)N*%MX_5*+g<<VAG}6%x`52KEZQ5y4|&6Jr4=Q&<mvL
z4i0x_w;|~vd2;c^gUw9WiP&cwFs3>X*zep91Mr6@y<E(!&jjco9hXnnv|jBw*gg<h
z_NQ0-Hr95~P!pky8KDEP(|-KHY<3~xNx-T93AkQZHq$jR{R53bhxBi7N&wrq;vAiU
zw;R4U!1!+uMRfNE&Fz1Uer^_tlRYwG4xN*9s~$T|zWQ10#udB=2r5C$3&<Rp+Jb8k
z6a@30Z|IwcGb4fcw^X1_cx)0E^vHrdaVAz1B$RxLA&Fre=ljN|ujyzILl`A?WIhy_
zJq!r9)~Vs8zc@%rCMZWH+ad-x#_s~fRX67iMaH}8>~C2fjqP)Vy+dZ~c(YS+N5YRJ
z-;;522O`))wUDmM5yhG)jrS-JbG1>uvJl4OGsDDtt0YX!C<d~E-feFVa^)fO4Q+CF
z*nYn~+i!RGy$^5ef|7}YhpXD54;e@EnUEK1yZhQ6Y5wpqq)jt(#{U?yrO50R8|ZO2
zoHc0t()p1(1*2YpIYn0Dt1#t%tfD9^gHu|Lo^@z)+AOU?QCMU~K-$#(6<hzX=b~pO
zBm1#TtMz)eT$6m|^4Ail`6K-3B{D)xJWjIAwflY}Q=WNRKu$Cxf7p=RREE5Fsj)R&
zbn+GoYZk*#r4`b~;xc>Oc14aulGm(5vyi^czS}F({`oil3WmpfnfIDw3GRN3&dIXg
zUV|g)i3WKM-S>QxgS-C>P>l}^kKi>muIt~mWz^2M(XJatOk|<h8p0@&9Gj(Kq6%3Y
zN69(JoRC%*NrcL%PfX3{QLy(4vp(}Y<pL@VAt&@PNv2ah`b9=NHS;AlDO@n56PUs+
zO8q~|(tWFl<aO3uMv;z4PXmh^0VD2zQCk#z`y6LbR!fX2`JbyIc_)hNR<5ix2L0ll
zdcQmH_l;-h`}<~-5;`KM9Q)-aXX<{aZA>2_S^nC-g=yWNgK>bD-%9(54S*g5|DVoY
zt?q|`%e%kKC6ek8=BqF{C7Fy!sWphC2f4Lw<I|ppge*nCbO-G@tOLMg+5$7FUl8g+
zf^fr*08B<89{&+q;%V-U6Myt~qV<m-MuK+y)vALAnDoKsl18;qn>Lo!UXhh@&#==b
zm`FtUoPG(0HI&vJTUz!V@EXxDhr_z-B^i;wu_S`;a1V<CTLLbak5Z0WEZ1(<xAd5-
zR_uZ=k4oz}_5JH*ew&x>mP$roMBSoKb?GF-gf+0m+SF|;4+gCb)4OF)oRf#?W;dQ%
z;U=oD&FS43qG|a?7@hRjiFKO#(S}W)KM{6k#_!>MVj04%zFyHmM*6x+_FUv0`qi5A
z(x$()I&9nL5<94T*LzMw>Qi2aA5+TUE4YlAzg!PsBsNT+NvpBKT6gr8^w?+Ql~hFm
zt5s_O^~UaThdVPN<EN{4>DQeBjf)nD%M0Damwt&u?!>|>3W}~c1`e}0ogs(dvNeR4
zFdBtzP%=Z+ygT~5^VBB0%;vEJzbi-gW2+RX9?ry;mf_bfHY_MfbFEx#(GH$zerVcx
z4k?(6PdsIU2}xwG>POg19vZ2+-YPn*5fv{;Z-0B+uXR5&Rx*M?y@(Frz>@nvqV!gJ
z=v7lIyWaFPuD+~H!lzwQc5XMWf#3o`CPfpJ2)RcvRH~uU!V}>t!X^B{{2}L?-RWAp
z)fLH`qBtTDt%z2>^k8e`M43t~$4j0zJedbo`rRZq&lqcsv!6G6$)Fsi#FfnNF>Bxt
z2OR(C&ZY}g0&b%e>2iwX|4#Gs<7#q>3N!g%+28L;t$<M*s?{K7>ex|IBW1m<{Qe+g
zvu1gB;WpRGs>Sk2o{4(^0e*jXKj_X)8Myb*mhX4>#v0TPo`E}}wPFDf!INA2WQ<2a
zL5o{3{2x2-Ax<v~D%J!KT<w+5Qe`JwNtY-zQQ&|64~hp3J(xpDVi;K0f*a`DPwxpz
zXAzOaf6IoyF_y6QBEV;GkP~npty3#``hR)$39Cf~$jj$Z_o5?nCCfRk#dSdT(Oj6X
zHiivKA4*<o{?Ka-Pds(wCZ^NP;ub<VF0-LKWWW<dWH7BgezJ$0JSps?t;9@G0dD<y
zuSXqs7PsfSLvFa5yOJrPiS!eyryZ`k2zgwm5}DJmx+^g>V83P~zESXVv4-uG9J1MP
zwhAG8D%rPS)8x&Jo>zdWG^c2o>CdSj_A|O38rW2nNH+CT$)Aqxwy#xxm{s{RLO+Cl
zX%=y!R()Z7xu)_$5vys>$!(lp9RQJ(jhJH?6IM(rGOy?b-z>ta%vp^gkM$=hy9;MA
zQVkKN0T(pxguv;~9}fN;@l(P>`Fx(lAsxrI%C8yQKNzCPR<zj@d`%gI|GdV#eh#sk
z_bE`fWI&}HOlmw6_11|<+Yr61O-r(jMz~#(R}yJk7yF1I%l5ipS2U@Q)ICvsRNR8i
zWVZCJhpc4_qE^+lh2UBl<ZgS+AnRuGy#AMZf92>QHhbU&JO7lTJDFJhEP-HPj7+>>
zgc2s_LNPNds+?tVTg9i@jJ7tzFp9?NLsT~cA2Nlq5<z5!OvI3B+Q?DZvf9WAwF>iT
zYorN06=OxW)1Qy3i(=EIdUBYVR`D^-Zim_e!d@g-05kAA>w8#_e~N3sPc5F$Sfa4*
z;O~zbGfYTN?>30%1sV)u{V`+nhY+H10SR@tpE$OAVRI68T-n1OMR8cdoyZG+ONZFp
zARiq;i>}+dz)*ssZTUcW-xUT|#w=rS|JnGh{U-lsa<COr`ajT@eC$kLv<G8rxMqRp
z&*6aq0+8=WD=!dq0rFZjnyy~@4z5=Dv(#_m49FC~^ac|FQg#ez4S^@7zW|ppssHj+
zqG@tVo>v~xyJsB*|00MPV%ls0COP#iUbPYMUO*}Xo8C3LtiNkJgBJ&t!f_y(z$+d(
z@k=4_&l|^{jR~zeo3RJx@suW9HyGRRd*Bwc%<@}$E}0h^X)=L4Uqf^_%~p5tfqA+i
zkCl}L%3SvNQn%GnJIUZ%B?4B<x5n*Q?P{Iy0@W+S?oNa?p~m)ed@n2XFJ%c7ah4UR
zyp;DoB8|yT<W?%85HSu9=*Ss!s$m$F^bkf9g|aW02DobvE%fjVWnDkw)1TC97}|c$
zH{D_M&$Z7%KWAitfaHy*jzbTORqpxD=Xg0_^D%-@qVHtAqtV79ZlrZopbc>!hsb;{
zOoV%M5y~m8aKx<v**?8sH~2q21#&Nup=SCaLw9$hg%6#YDgAQI#E=fmhu$n#y_6pQ
z4CAF81px-_zO{!XrLR(RN`99l2bTp0JPd9%eTC6)6|&Z<^(d2We4)sWw27P*9bHsx
z<3I~b#!-X_a${3G(%Mf|NZE=GDUL<7+#0){{=^+qRWn|rq+i4BreJ3(=5m#%WhKAG
z$m4Z4`MH~RmdZ>^?GopiOzM<51zA`-?uUdiQ>P)qPutgGsGo_^up(>)j)u6kA|}$b
zJ(1&`g0MOu(@^g~X?9Bs#OpLj-)nO~IN(b$YtoKf55bSd6D27c-8=m)tbVuLvhzJ3
z+?{&3ocX;+fluFr5QP=&M+`BD@tcQ8pa7BFO-<3Twjw%f&qqqI66g<_3nHim&fWRD
zyp6A*&d0XjJhz>@ypLAhn6~fS^cqgPyI#DQz0>VF0k`k|?(EbF-j4S6n-~(lIv;EO
z;_oEke=L`4FyzuZ8xC_Umnx03;BfelhNQavT@XZuK3S2Xxbv*o4bANq%drh+3KA;}
zc3$_uxufFq3jwG+l1jY0o3XFIpHvD1E$@b$c*9vNw)QxxA1dU(A-~D_dlz^E!a>CF
z>=_6mVOlqG>yqf8KDNnbP`?9t1~kKS?KsOfiQVmGvAnO|_XaZYfw%byKzxW5M?(iL
znHKaXFO(tiwh9jO#KjiAH^I*+UrAJIILCeyc3vNCY@Z<%=Sg&)`y)l3s4w#XcOuyQ
z0g%4>1vYyXJsI^R#(F?lz`)QYjk~uc8SFB%X*h9vr}}F@7<267{tOtlRcLmd!yE~j
zx+l{^Cj209Rwa77z5m+kVB4_Tkcgqh@AT4jv2@F?BKlnM&y)|R#)o<Q7R&B^Y9id`
z&`V_>&wA271|QvRc~07yO*^FnO#Y_!p1+gbCDh#+(dbv?AJQBHGeoJ)XYr32q+bj(
zu`-!WBb<+xqZZNB89xD(9pug1yr4qmd&DH&>q;D%J&(H}UIJl5CfHH^xy1z0_Phe9
z2biF%&9>snHi{VU(|!2F7LyOn_d|Yi7|kY4=keJe?`R)B7HlI+Uh#5@_)>yk`K5=E
zGNbha?=$<-LP=b4$7=QYwqIuYy(`9tzx_(LfxL5RQ$jma)({uZ6d>27=j=?F22dyu
zCI_oDG7fT%uuWLolI8L#*>gF@OocON2E&{di-<v>XysO$R}T~9=;|oRJZ4MJDaOva
z-%_GQuefg$djd=8-ho}20)$-(?TIgF4T>i7%rqyTFSd0o5u^TH)W>;sAFt`xGw!wa
zsx@0~8XrhnY=ep>n|t>)smqH^XKf=*!_<7|uJ=oSTdz;Ig!zNTUiks#w|Wm6Ph73o
zb|XM!6ynFOdmTN|e=tWq=AE#LIeV7w4~<QKH4F?uX$u!CNkI$8ZG^r3LA?f(=$#?>
zF;p3V;2MW2g9Nf}Xdr_U1M~+l--RqMkE`BFRUjh+%r2NYp+@$8``<S0>#Zdf&^3DI
z3t21D?)qhka6LSdB#|N{y67w6E&E4K8pFU$9kR#@0hk=EJ*Q@hBtmBV3gb%H8V^y_
zNf4%=7cYyfP#pf5oPF0B|D7Q+aV=+}Wy)k?T4?;MtWPCP^BeEHq+HGzx!9ZB3XHl`
zqYfR;Zh35<({5jFy5)A`W*Vx*l&6<#nQLbX>4r1xpSEAS+@q!OUdvkt4X#luT;SmF
zOS!C>Q~F2Zx((9S+7~<I5N3hhuSRn@w9%pXb^A}<kF;I5hK8rxuGkopC$^MvYE$L#
z>3;oqRO+Ussu6|{PkDb`<HJX(V<(nrG9yiNrt%6)2w8pmN6nmy=34)3`=!6`Y9jj8
zTvMn;>x&<fsa*00@4X&DYR;^o{buSk2pCfdZq4v0iCj902T;)wA5^eU68Xd45)Twt
zn+#<RS4&R0yxKJjvnBecU)(PY24rXOV-&opzE&RejHSr-%(z=CQstd#J}X@CE@zL$
zIe+>|zn1!A0uAc;@FUJSluTMD#aJtod+T$L)fkh`@^;6k%)u}b1GU#g%sHe1Rc`OV
zY2UL6J#wLlj0;h|7}xGO1e<EfW?5K5CLXN=J)Pxeejw5T_|=Y9@XDJxnP7)H34Z(e
zr=5Ip{B8g6Yh+Pszm;^izRCUn&BkaRzb*0aUoXxM<Rys<#&`lf8J?LR<ZnW_(JxYq
z_uydSVEbDe@GFS_$0^(z`>_f;+9sO*Y{uoO$+cCRRIsqa>GD(}ov#036JcHZOQmP?
zWHJ_e=Mu}|b(KTlzl5DDsjmamB#2}f85v97zb`gRQqGL?i~SIY#dbu?MDQ^|Cg)h2
zzpC-v*tJ67Rex=9AiWzy!n9w+jqS#?a%|Ot=jg`=HpnTgbBMcNH18-evbdR!-k95=
zK{BYs@lW&7faI&ikAtpihVN$nM*qZ|o^CY)t16{pMu|n=%`ROWFeVczfBsaOeXxq=
zcvp)Av5+FGqM_Z$q>7kU=5ncjT0rsC&FiDW8FllUqU7&&>|;t_a0PU|maI}kBZbLZ
z+*WHehJljx!uREce3};-Yt=WcaDMm)&R5tLE6}--S7w|-@v%c;9J&@P+T9^t5vmkK
z=hT~~1sPAiCDF@PHC32;KfKN7(tV_c5THmY=fH2pt%4>?EYI;8ZK$E?lw1F_))Q1D
zj*DVrX5Ah_p&eK6Szj+xd?jQtna(w*Wj2JM6>ubsZmvB=eiEva;Et=f(9HVr@rBQ1
zrl=1UCQ%Kx6Qii$YPKugtXIf7&Z<y~B&H-Xik!$X%~qKFQJpZ9FiIUE)UlH6V*d(F
z)_V1Ol+)Nj>$j(M5<B%iavo!il@SAT_a4)JB%6s+XC{!0k=7djWhh$dOLs<r@&(yQ
zan@E^9`feMG|zK{x=}?iM`e<EQd!ZE9gn9uZkEd2hH&8P<R7=&5W1h(E7YMu+_?BZ
zEsVW+(%eN&B8xVU2BBRL!SLKXiIe3br~Bi#mr3Hj>EmkYZMR_~i{6{MQ-4n(ZnQ}&
zd-uWS$%R9tRl|{a(iXOU^JkDRgJ9vG<ei?;yR4)=m@)6}eDdtXwC5ODDBp*HW6b{R
z&0*1ZfxpqD(g8LP`F6FxjPt+h0wda`N}r^UnfIjY#&FpF82n9H3u}tQwkS}@z~tAW
zm^A-|NHCDmLIbC$qS6evD{N{>Qe1$Fj@xaWT!BM-x$#ioqA=i2q_caHl_6fk4N(W;
z_CDvJ;ue$1iv`O3TH}f#(d1-|k^G^2F+nq$h?aUOu17+;z;1A}iR+0KOTc)aWAs<M
ze<WUTN4UfTsrJBU{~kUu|I9l>VtFmC8&1qF-Qe)7z)ng-`Ag9bAb?UP7TO9N5@xTo
zpO=y9JvHd<Bmp!mD%)7`oI9og^ZV5ZyuXEXjoi|a#`c#1eio6IJ0_JrGkXu_s#TT+
zq)iU#c@_+rM@x~tNtw!xES7ue>`y!^L7r%D$*o(tF6JS_w$P62b~dq~?hGzDXH~{H
zfSyvtQzq(9ajYs-D9k)-YJTY}dItSqIJJD`^NinF<?+nh`4~j*>W3>hmthr(tdLB~
z!TfKyxaQYs&u_~JNlRY|womvFa^WgcQ1GT8*^L=|0eR2S0aWu(!()Y6A|GWYW$Yd!
z9zC&n9()t~dMatb<H^hBH^p5$f+K<_M<@yI-JfdHMo%FYD_=U~y4+p+!L)L2a@Ho*
zN%w~rfk1ZtxWqfbGe{xwG}eKnFj9I2fg{GIGB);NXe-K=DFSiq<gX{9&-kOxGpVmi
z2WBx@hbP?^(IM-6zgsRPDSjmw)&P320o!BMP$z=FF%qIZeJOI{<enEOi!Ce%k5mo9
zoF0q)A#xXi=nNtDzi>UI)@bQFl|6ayU|g*=XiH|p<M*9Q%UCSjoyNa@P>;QveCX1h
zzfKw}Z0)P=pZ#ZZAJGb!D0Ee%o{8ZI?vDqq38zg=OOZ$8IU{67mxI1y(b<Q!@Qiw<
z9_QbK3KRAP@bHu12_(f9yzu`;8UaSc$;AcEDMGy>JM0dw>>{<wbeYP=<%=~&`TQtN
z>PDy&PPVUwGK!DA+<>Hs5Sc3&k)5nVYDtJPMj`sSDe1#=wbBJ*C5i=KDL!yf(Adym
zXwLZ5M-mQ=Z<Q*S&xujbN~3I2XXRoFl!dj}9aiY+8Z~N|2xeq>C!}xceGg76TxF58
z!#%UEk{b^1F1y$hep61^E%<Xx+}iQJo5j7^87}FXI(W!1uCVO2t;BK~KK|-Ylhe!F
zfY48%zUHjEBKu2RoKQVjN|r|<Jhx-cbMbm%QQZ|It+&FrF_wn`s@6k)ErTl|_3;%`
zBbuGtK^dS4^E==|IQww3)o?gAn7oq!8>5fdfWm1by8SLfJBCK#YabW+Me^quL>1z4
zcbxv0{#l3X6#VRDV_Tb-Li;dcZcddy=})EDXMb=`a4YkJIHk)(nxkfT4;?DS_(xt0
z*uJ(Xfp?aZ(Vh1ZlX*N$J1MtqX{teAQ$r<^SbsXn+Y0NPF|0->afR2hCcyr&zO3O@
z7ma*bUQKyUew*uPtuxIXbxjJbm9y5O<*`P?DN!K=4)z>b0UIYYPg;oe87qAoIkGf2
z6^HxBWP|xB+$6XkUVykc8(XTQtkU*RM<`INlPs5w;yKk%BFd!>7PTJQ-vYr0S_58t
zN7#c2z2a(M6whZ>>`W*tmh8AwWqRWuS1jxD$lc3sIjg4@NTx2IsZDcWEx(TnJMa2+
zUeR3Fa_`v~Nle(<ZsT|F*>I=vcaX=1{y$ph-4`Uq>;u>O=)d?I1DBZJC#NAPxkw`Z
ztuUDr1yQH+?>dY>T43fSY8b8C+WuM_0!-kV>|d}3^f5`z0u$T)eN}kb-Q3)QA(0ct
z0KlChZ0zC7^^YVp!EZp>da(v}X|SM$&(~YlN|^%DPm?+_;QKd{ygT$%?RQ3t%Qp;P
zdvDmB3>gy)9-9VEVbnSLm?-!guc1WyZ+$CAq-iqP{Rz_@Dd7*X8(SkxX58E53tsl6
zQ)RPl`-NMqmgrEKZ9#V-pI2(WuC$>RX*)vsDX|{1>%}f};>~8`PRtw$P5atbK<8s2
zp2nq6fYfJeNwKFbFr{>}*yk`4?#Xj2Lpn({oEogtnF~p@hp<@T1erOrI)M+>Hv4e<
zuG<S-<I(Orda{**HWh{MI9K@0=kDrG`=X-LLLzl=zVm~<I%kVZAs>$bSnA!p*DLa3
z=%U0Z``FZL?7^1f5RtRyQYr8DW$_KZy5}3guHrOj<nZr-7TuebuJXLko)s`2IW{NV
zt2$Sk;bI$|$UVfNKrTZ5(Jby^HA>6MH0(YlS3y{t``vEa3p1X|Gj0VK5v>Fh_11R9
ztA|^*gy1jBawbLQELy1qZ=Q;2k2Q?zdQINM*qG~`@&tW4+s-BFmUW<Wnw?TnwCByZ
z50rBu;-5O=M#`jb#c?QduySCYMg8@Ve>#lPN48REN0HbcX~B<Z70Cwxk9O9PBF1Ex
z4J8<?!*Syb>Lk-k2tFA=;T^Q4-5<qgkjq87>3YN7q4x;s+1F>$L*w*Qf^p=uIZ&PQ
zP7cr}cpJ^GfM<|2=$ef8x1#VV81K-{)#}gZ^=Ch5cik8CUjcE!m8HF3Ft5FK=Gl!n
z(l&t<-3>vpzph_cHd~2O*znqNjT>tM&MZ~nEUQ}m4+B_O-TRBW{O|4Oc-$RF;y77u
zO^r1jc=zu)zl%DaBhfGo$O5=qfl=8x2Me%p(dZJBYRf*4+UwvM2lOAvvmnio^yrg3
z<8Ycm39}9bF7cC*lIWP+#m@OZ9T4n!O%fD^uP+q$H~FZv+R0u;yLS(cLK2WDy{_TI
z(Ks2ZF?N^tS#z&zMfc~I>?gB~&y5#S{85P)yOAbyC=f-f<!DDK7Fm(xfiuY`WdVW4
zy98(QfIDR>@Mwud=A*}LHk*nOv8SoxTc+B4zsGg+K&U6G9%jH<+%)yMAeSL`(YwlI
z{^8FF)>H<OA-s3$L^pjy2i?{@q9iNKeK*=ows*FexGzqg=xi$qKHP5xcQ#M8m6N^l
z>wb~5!-80DfkYmkOWBLDcFBBGP7z0DJvmK=zU4NW4OsBj)B-PXR%uNq=iJ(Bm&Hqm
zEIEH;|CkChH!1S0;bv0MFZxH0)1*autk0<Y7Jbs|$;<txqTk;7+&kyewNg%P`PoD7
z7%?+X%!d?x-;~ij??z`d*ar-nWfML*j4M(LJBALad+67MK6pqVl<AQXkA`c@SQRwY
zgA->6N4t+BJC-jwDI3Aaq?I$cSSit#gdV(w*H)!-JFi-0_x8AFSL|Byw;eliom9>-
z<Q&3$Z8vx0oQ#>v&#JRS)lc^AqU?U^)f7$`>T_kxOMV?pAT;ySoaeG)mT4%Y-LR7v
zvO1f?Pk4TPm6JS^Y!P&_=RGf@G$-;p0gJ%Yl*HT`bK^3PLzMIe<n?&C)m!0bnf!;5
zgwRL!)S}85^Dxrp_nZl%K?;_rR3O32VvzFRlq-pb0*i`at~-W0J9;Wp;^pST`!%)k
z)cOP0mDdgYs;3kQmIMASUvFaUfA4!n^~aAHTCG-L1Ko54#}@mab}pUxo3y;M*#S7T
zn;;OlV{RFApv<V-XW$vJ3!a6zup}wb`J7=`)h?|6|8T2dqJ`7kIk)!ZoC>r|=rNKI
zmb8k1rLT(rCKG89JTbxV?d=V}1Mpc`^)gXx72nQR=f+|UTQJ56Wh#d@b^<_G3kX>0
zRseDicL`vzRTpV5W+?J6yw|?^I}3=&EZbkm6k9o-ieCFRQ5dZ)of`hI#cx;`Z7f|v
zMSU0GOf(bJlct#CTI!gH#HMT974Y$cKN&PQQX^Tk8THYfEdGDqtlSfnePl1?p8Y4O
zyj)@<X6XC)^sbVY@!_-VvJLHX$i#{`L+gnJPo2t;3d8YfajUWbGvze#{`~1TD3Cp*
z;fs6buF>RLK+-!yN_U(&TRDoZZ^zzM>5`+Js3ps+d2PHr`=sloOWBWxe{&4VMB`3;
zOG(mqPp;vY?RNa?d&ybz6DIig)(Psw@{fKj!VC^jRovDT>5#9oCKe4f*u9H)mYBMG
zm>lJV^wX!R88kz=d9lHu_5<oVts6F#U-0fRZjSERS5;jVGM&ksFR`if-R&%kIt}f|
zCoE}sIw)*%ZF%?|9@-Mp1fxe1H8>UQX?bohYt|S^kvSAqXf<_)U-Y<|V*Bf2BT+X8
z_vjAtvPgng-633CJf))1cLNt<td8??UdO11<4}6pt1nq!f>J75KR5oFpZbZj$v3s<
zyWIVPLRuzP!r?2&gD8FhD-G1i;4XqRrL@3P{$PEC=6O%LaSE$k1T<!-@t6>Zq9Xqk
zr6RvZF(*%X<9v*ydKArJ87%VyQGtfmb7!Lyrx2S5Q*#zL%QOF-??dbVvzarBI`TvE
zJqe?%*T-*9E8hQRRXR@Sps)Ib$luaf>~%K6=5PH`C8sb7Y+-S$E&VXfH@j{vgCke=
zh$Q#T-ZM24adQ0Ur%$Ki1IsOgUM^br2oNrjoR;RXTsinxL{7a?gVkqYVJCX#ieM51
z_HfEpXBQVz_yf2ZVWtbq-5*dOUl$7cV{Wl)0o!B<+X+on+h2(%iNImg22O<wzyT>A
zltaRzFjdT#SrMfP(aNBq{SQM1ND#m*16LekBrfNiB`GH#LgpbPDh`;`k7DM2q>4ZH
ztiF7P$nxzozvDCI5h1PF9DHO0B7)<h!)wQ}jfs|_@4oeJx_<&TL)^|{haYN&S12_C
zg1HQlvheEbZK63;)9YLm8g*Qvh-(;AS%%|rsSOkYttu10P|_#fIBe%!ogUBY>5`qP
zb;33o+M3DZ%{YqBpwICYOTHJ-9WZhYJ^H}xvev|OwAHK8-zrvR73L00pu*Xb`TXm4
za?M*aMN>-!?$H-kYmy!?GZVUAq&&{fooJZy0QZVc*qGDO)l6k>OD1~TU@)a1tn8Fg
zU6xm2Q7t}-P#5Za_NW}TQi!|d&7SZ@#W@4BaU1-#4ze!a?}p7<s)7RqyBhqIxW1QI
zdSAL=KR&A5Jen}Qt2kIaTc;Dy_0c2m`Ev{M@RL9F0#+|Q|28-p{rFF5^vjU9v^fPB
zmi%oB-PskXo8R}M);R3qs^kNMUR&8yp7OZ#t~L1N>xHJPNfxQg`pof7AjA#*!t$t(
z*zEm_!|1$U#&qAB9Li<XQfS<z{rXthlvMvHT=%SHGtN$#?U!8bBZ^b+JKtqsr$D5I
zwoZ?5o|PC4nT$Om65N)6JB#u_UPmOjtx}w2MQ9>&QHo^j!Lou%{2ys)a(O<<2;!NW
zn7A|%f=lHPbP{{UwITd2Zn}l~an<;v_(Um%M+>0T!quwQxYjQh|7I<2zvLSu7ljY_
z!jEvc87Xz;$KOMaJ%7I^dNLsJ;aYg_*GtqXWYm(DsjZ*5pvHIr_jm&w?nz6vz5Qs}
znwsLbPdi7m8CUuPcZ`1#4_1pOuXt7!Ny1^<;iDCOuf`3az2sYi<x{uYmK85<Ay-un
zfiV1E@c}b0MJ1(1@Y_T|jRL&M8H(@tq)eA*9JUb70E09@#11-<W)q~Q@X+4JLld^V
zI{?uujh`TW2?hr>$_FVRzNZV=tzy3XCOjE1dLw=c1U(Qbh%@&5((b~v=Y#1T)kbOf
zrkpibqNHwQHmT3PCczf<RCI>tu~%&(@b}F|3iyIX&QRl}^pUE`=w?-u#RQ21v~25V
z-Lh;I6}5<@Fkm=ER_KPbRdAeTY|_;rHeSVen=&rYH1b)nEfF<12ph*~G8F|frhR7^
zTdEka^-j_?(zBG6=g8i4OgoT&kPJw1*XdqYqo#3|P?N$t1=r?>OmA%(;_{Fh>|ayY
zH|qT340pzk_fJDa?Kv;)>qV$>3nW)&mgVzgvH%^1pE<6d@lgeHUE+KBuJ0Kc9QDyl
z`aZ+u$8C;$d{37qk?b{HQTxXT(vcr+P|U075nO1~@JGi!#yy&jktrKW>1ko=P{Q(5
zKi5{iZmIu{kvwC($G{~bK%YizLF0{IZ(G^0O77&14q}4|bu5FMY(Meh9dW9t=d3uC
z%z{FByUZJ}1n;H{`4oSt$M2XKlkS{Pskwad8b(snHL)|5<3au<?f3GX9wszXN8^Gr
zQgCOn6{mQ|`eq-0*y?{GbEc(QXvpN$MG1E{>*HKWi6b=e%rb7lv>P@Nu#R}-pYC=w
zD;VO)&+{qS5>6(7hN6u&Dhr%w=M^;<OHhqWq0Yy4@UW9>kVprf&7<=fmo2wUsgEmT
z3Q5w*+|R>FI#?pvs!O{P0ODp$hcC3ciFp29#_6@kw{&0iUB`U{|5?}|O)>Agu>LrT
zq}{?q-U7vE89Udm3zz?;HEn&lVs45ir9J)iI}kVOYktTX<*j9Lk16^c@>$Z;uQ7(>
z3zIWw>HdYr@&^27hD8MMKtlOSESNxH%wQ8kY8OFF39TP68bHwlhgUD2`VbE(q5PeN
z)K_Dk9Yen3BDrf(FbXEkasO@H-2MGyz%Wb-Q~no`V*WQ^Frh?mRP1VTDkBjTL-2u6
zwVE|8tHCuDudq;_D@qazE*%h;4rxKX^M|KaObYL2MXvbxzh+#Z)BV$nC}R&JW-T|x
zk32fv6UIwH#0;*HZ`K!EE%KqJ=8iV<eHl)o^e-TQ%+2F^5r08Ta9K1r;eq*!E{sSG
zyDR?nE3RAh`EIg*3zk~6;vmIM^i(=N`YM&96nVH&Fjy&Y$?iJ#FN&J`zo$ESC3F?C
zhjq=}OMEc(EoAuVRvr@c6WUnEq%V^6q1h*>k3Dz(p=k%b<$9jm(zz<3)8M8|S4}k}
zOZ+<aZBr@jh}nhD-G8>_V-u--x^is(EPFAh;dgD$1agUubSYj>y`w5$MNaYYR-sm`
z<0yLQ&N<A*euPuT<m`5O&eCBfC#pn4OU=_#0}e+%A=B9B9@A3_WZmjm|M@A$ni9F7
zUV==0Viu*p)b4IMpH+}!GstdpfTKOrW{YN{ugHAuK3V-V2-)^v<JH4&t#;WrYSx}h
zw4YPM{zSO($8<>UwDqaJjbCM;6CBi&;lBOSwu8-aa9KYUne?^(RE3bQZG?)ABJ-5h
zc(VybL{!@N=Wq|%uBaC%IT1uGSxg!t{)&bo<pc#+?%CaNcS_Otiz+A7O@Fi34H*Y6
z2{{(cPkU<4@XhQ<&Ii^Z##;`l1%A&d^DlzoF+)`1?%|svGn?Fv48qnhD1ngT@F*IC
z)t2Jk^%L}x2a)f%DJ2eW93Y0)|In4ct=S&^PD=9&gWI}=REoT9dS;k=c){U)eM;9-
zM%(hlnri^L+S+~I_mBLcn@rrFt=hk^yQb#n@%3hb0Xs&szG$z|^O^<NhZ<$T8BAKM
zfhf(9$eMZ>itb#nw}Z|XPV0X|JD44l^ueS+9w4Q{&)*DfZS*l>y=!mD2;=L8n#)Wi
zk{+yr@f~bJ91$-{NTZOiT8&p>RN`YC+WtAfa$l(hC=fvN{tMTUuN^50O`c(}Qb)>r
zbSJ+pqey-x+6et8&?+G}#Ye?xf%JL+bsqFYvuj)TG9ix(R--n@_V_+X>>$;lX3==n
z(0u7St*F+f;>VF>TN4b;i@Ufq1kswK?Kd+8|1KeN?s3qkaiFKl^OfF;8A6EUHoW`c
z`^3rxO(&AOEkz~*^SSPdd>WT@qI1w}YDl6Kop9=sYwyBqq~h&@6KQwfb=K*y@fk1Q
z=>j+X_V?hQzBth=OAqOxSY)y#`{RHc-y-scY_9bkutl$wy%x3|m@Qu2Qc^93+T*cu
zh{2_Yy@4{)c$Y-Qt)oXDPr7@U_c2$~XbWrdysk`&e;EImN~X8F(vDk=()DHVq-TJ9
zmUGPw<#KmBf8_zEu)M_4%He41c{`_fx{ZymOU3r`IVrjxn~*1axtad?#2t5^<{?ky
z8qmV%MWzZeaJS4@oTC;>+$U2C#;2=;kfLiG>ex8}zFPrrzov-j+YA_Q1z4d)n;)!Y
z77|Lu&NOZvy=19*)ztsA_xiG)(6Dnse%kfQ(;>`|k%=dSUzfFI2Cux62Fe367t&Y&
zQ4-G4$`?bj3{ylfheX5-h^goD1zsj_l1Ctv$!{F4N)28;9;p3()kCW5#NT1`Ww5~C
zEf@fckPmNtRJKq-fhMswY?VWu^`^S1$twGpHfc{XJ^lQ+>I($V+j)+ey*_9mgIzOs
z5eGeLn+IeF)xM)NRv4<(?Az<c8ROA%(X+_Qsw}#zt|N*jKONrZp&yMHj|PH)Xk#U{
zN-gl_76i#E?5%DD%~_BWQ{5xD&U@@G&F^gLkNnh%@_ygU5ytzrw>_Ir4TI4D*&2M_
zS=_P0_ZD9LBOs4}P-MX7kQgP4i;%3$4mu%{m;!btfJY@Yo3|>4OPAiTv`@*SG?|#;
zD1-O{+E<cb?p6ipESMtrU-rKKAt)ui(Ds(}9e5B04?xmM+_O&?%|ialokv`Z_XN&?
zmk$2h#r>MV{QEhxVAR+RNkW8Df2fk6ZKWi!^5GU3|JN{(qM%6<BuLnr_Y@!gouVQl
z`ca80iPA?xV@g+@SgKs+c@8`Ls3~col$WAW5<``!&sR=$V#j@vkH%IJqx+MEdQK&$
zj!v<L$e?wH7F((Oo^8${jK9!pkjv0tYv}2|5DfHErr4zDM2v$?ob6+h!`@fo(d1Y(
zGUUNH80(kIZ-W~#rK5pP_%@|sT+=ScV<mEYhL=ncYL}hIaqHO-)k$~>jP^9<Yfg4P
zJ1d?a+HR^lKsCJTdGB-<gHv~mDV1=(yJXJE;H0}{lJRyb*Og_vFNmWR`n3kWR&om*
zuvFQya-i1g?58+wO-7Xs+XdhL6uWb)FdG@3eAm6~iq{6<vDuh?4!;%2C+l5mzo}qH
zD|Mk=MrdM2``G*>UQwuR%ALW8DoG1BnS~;|pv8ot0*$@QLnR{y{oA;YSe-oipcJNS
z*H5}R%_lstAKG|1>7gtz_xi`>DWSsDo5fwc<RKv*d`OeCNZqW&NGPz$rG<of5W-fa
z<YkD&NWPdT>*E^sOx-k|tsKWfZ%{PPn1!?}<*0a&hJ-&8!jdv+?q@(~UjPjaMx95`
zNMw8SiL#jTzsQW&Wi*Gwh}~DVD!;#dCg^|1GdyT@sS1HmUY8H_bQ2v;*@@iLfA^O5
z_B%Ch@zm-5_|X>lDwO<4qVx2y)JaVqgt0dnd6VV0{U=({-={r;eggr`XQj&+62&ij
z4m!5Mg3&8smGU?@Lg-@eyyxMIg50V=;0ZhK#?LLevsWMb=*8Vt*jbZ6JT*11&6=8|
z&AfhKH$aoN28B3jaF_oCvZHA9QZ6~(^^I|agM%@D2bDFLKH;Fkd=Tz;IJW?nL_@J+
z{4V|$gwI}s=2d^``|5IX-6icaL?}}JCS4(+W_k9B1_~ApB6O)xuG8uDh1ky*s6aYO
zMwj0EaBF4bD0rcig?7?eaZ%SKz0bsD#)aut3xZEMG&%V?Zm|UaJ3cdIk1sC6+p*8u
z-Vi-uH`YN$WW*NH9c61i(cbG_-e=#`E_~60vwzwkhJ~Y?sj+|p>F8k^IPCOTXx+aa
zPPH@s>Gq;3nkq_b^+eixdx9NY>Zj*@PyTsp*FU`$+-{gM6Kg*g`pkK)q<It5Z!@%F
z?*8^SM@P31^NZQC!~Rm3em0E}r5vZsHiJxVY<FYHxrBVA|B2xCODbz)PTZRk$VBIy
za*Lbc*(Y6@Jr!$c{z&4yb(H6Yrw+f9dcEZB#4|*7I75BapolV|T=L3g_UX1Sm6|Q$
zFp}pbOQSco!CvA97|%trda+lJ*Eo3M^xAmlZYtQ%)Js@9qICsoT$9RiL>&y-Bj3eZ
z&`9u+C3#(N=kAtNawuO*Khe=y`a=HgiOFx1J?(@2ZjTGH-)hAfYDNz09U1o-9I)b(
zZpAP$+pYoe`aI9ek6elO#T0!nHlIhy9WP-f&(u&9*Sa&!EW<*O{REgk&bDy(^fPWw
zCqc1@nEl&Ko`#L$Y2(G05(m*IA;jBLKTBPmXp3MtE^>G7Z*btWr}gVYC9vpD-fr-W
zIsm7K<L*Y#t{OSwTOfzBB=2OKP1VfMtL50Z72xmqNP1rmyzBYqN0Rt=hxeiAlW#Y2
zHS2zzRZn3w@Wx%kO4V=u$kVU(HvbBIm}hrZdJi428z0`=7o|bM7zG9(l;S|t0U;dR
zl+;a_5tDZMex;mYCwUUa{Vx>lFU5p3CWOQOp9aoJ{*#kDGsV&g)7-C@k<8W{;>Jd_
zB3hAsd^E7}{z~ag`<8Uri^{qmmPw#ADTV)wGJx?p%*%hAFE+}5VfHbMPj|d5JRP(i
z!9n_Q4Lnga<%8NaVzA0U5}`haD_(;L2c2|VlT@fvQ?^j)8#ECA#X11(Ui_t}LtoXU
zYLT5^r`2ZcxFS|CuPoTradkX>KQ7fElcq96kz?QU7H`5_CeZIKP~jgxn3Iv7y}2Xs
z{z|8Em%$q)Mc+a9RAjKuo->pEJ!i4d#MFqF^s_A#C@FFg$STuE8_|>_)0`Hmjg(wI
zc^gBHJ;snT_mwjSam%9U%LY%iF#JqE08JiO+0|6js?a7r8f03nv{##_lsdSU$S9iX
za<^f^Wz6#1m>aCoxwK4E9lrJ>gdgCP33<cG56oi9)!k_EsA=<HXbt3zDrM*Voe+uk
zzUV{TJZMBB*>iQn04`-8eA46LxKnI>lfu`fgM?|mB1+jXXERol2(B&#+Sj=mQ-Y<L
zZGLg^P3Ec0WNDjKa|(JZr%2<tuAFHqiw}Rew)8>e9eU`t`{yO1fRrIxWDFrnU#%2m
zc6>+|!R46pFyjinflE|MriZOoFaO(S$L|GeJ~!{mI%U6-sg#WJT>8V+e4j()BQO9F
za)KqC>Z}6t!mVwH^L+9Y09Q>2=8x*8;Xs<Hu!8S|d5#sq8sAG34$9H9ax*B8_>1W`
zx|xsNV!3fO={3CKU5|Fo57Ux#AN~usEs+#XiyHVZ0a2_wUw(bOvfq8~-|q6vBVN^f
zf3aoH@>4Ohn<R5+almG!_|B&WvOp)dDBhp+H)>eF7^OWieSddV)k2L;cNJDa<J7$q
zdHUDRDC~H77!K}P`9);!|16$-K$426o!;y<mWzh1;y(nN?=A@ctBuTOI9J!F?i5RJ
z?>*YrpMjYO*ikfrY;r@c(NjsXjt0n5;_W@?zvezTzaVi24Jn*dS6F5CFMz#N9PwpM
zWn-&t{+oU13I8$UVE9_Q$Y&4-EkYwI>A%C2h@IV5yW!1QFjXZ90`$EN%r2yd1293b
zwWs9W@H-7(MfetO*1I87mh(NMYAlKcX+pWU@&W^uz2Eh6Cp(sNLi*?VXUwF^1&0eI
zWrG5T2y%3Mwtv6&`v8kGR>~7sa`h?IPF(P|^iYiBx1UKeBRxalfo<zJwQiQLz!Y(c
z!H98c|4C@FkiUer-Bvy~i;07F!O1JbX45i<B=)c4Ba~d(@J{4@`OPpgWN)dFP#mN(
zEoVW0tPL<`5RsDMoudp{-+i>cyN~`xt(WCUZJ@J<?l|R(LEO2&OE)q2LmK_2-{I5k
zUv*YE)vR~sLTUZ!V@H}k2^F=^g@ln^?ul$UHSopqP7wNsKIJO5sk@~unf78KQrOY<
z8?L9ZUskOGkL%a5nGTtd8_!dw&DeHaa9Y9Xy5|L{T1=YVMtXI9bToAQ4jH*3cIoA3
z+?0kyc1HB*1mra)=e<}5I(>yNqSqx#oFkHZr_#gRHC4f1Ps#B>Q;;mwchO&>BW9_c
zP92-d+^;b~al@#n6XA3C{#LI!fDIo{uFw$gV<;tV2jiYw-8!Bi+aGtrSyd;v?8r(i
zBxYu$4}G+3Dz!DysVbDsdGW4r-JYz?E-jT<${EEm#U&ByrkfRFjg0|Cw6GQGBamB>
z-Armxy_~_ln=%w`tw~|BIvmo;ZnA>u{7*Tj<Q(t&I#86>@#+3Pbg(~W8n+0tx@)f?
zK9+)yQoMYVe_2(GJel3i<)6`-Jz@7oJMFLGd)vynq;h8A0_}%&k+(ye28(68utM_w
z@$+`irP|}~1AjAGDEm=2eV5;7`Pda$XlCq967wBg4dM-Y`qoQFPzf`F{mXDYF}5=c
z-vD{@Hq7_Ed8V;r+*9uPK@87E%-bIvQ5qNH1;85Iz>qXZ=LRVxv^Em%SJ;9LqyDkl
zfC4~T^24PLqs0aYm?ahOkS}<$aCrxwx>}h>ZSGL(K#2oIIb!;ToL;c3mHRO;LDrT$
zKla{JGX?u$wY89BBY^e*radGt*I%{^@rfD@(n~XB&iD1PLtp^Eb!x3^l0`O&ufRzp
zgn7i=SCQ+b8K@q1!7SF=aY0cDvKx|Tx*U23OgV#tqo&@6X)p;yh1inu_Ujfa=;pVT
z;LMEjAJ~&g(~HcMu!+yvt=u{F5phF}>e1ui1ZG!WR>Z6D%j@<S_Io(DGR&w~V>_i)
z+2|Zh*ne2M2boh~fdh2Rmea~0<9cD?*f~IAwAq^;2q@OPaTUY+^F;NlI>I)pamqBz
zDSy_QJdu7!g6x*o+%9;)W?HYb=?%FK2DR*D;@Z5EOIv)dVJJME!si*<Vi^V5;;%9q
z4y~RFea$8<JtR*pEJC)Z?nV`Ss_ETh`3SR^t?PVNiHRj*8;jmYJ<1TBWrehEXq+Ne
zGWVq}xxe#IN&8#ZC9H6!SO;lLrjESSP@o}`tPr~Sa9FFbqG{uYr_P_zlUKznahSL%
zA;wOD^UbVx9DqgYZXrA~jdJ7BXe1`Z+J_}RED)(_FI%je6Ye$0xMdUe(Bnz;xQ%9R
zp2B9djwxetnSzv3k<&Bds7(CW3}+TW9U;Ktwu&Thp-$R3AtPvmnM1&;MOmtk$Zh_9
zINULY#27k%oM7L`lmjki8c+ywT2o<9R*rjM7(J)71sM6ZBh{8k#+M;s5X{dDM6tfD
z+p&v#V_r{@vOZUkzlvO{{_B^$rl<6()8teC2W*x3!T#@n-JRZzm4of3F)D<L%Q|U$
z3hzNM?@!)IKw`XzyHlAqLDB<&ZF_R-J6h3lblCWVxa{4D_7j`QSN6WporjS=_xASA
zSOZ8~%6`Zdo$C)^{0Hi^)Ljj|01lJf-b;J<Ef|{wclrk&a}~wacI17%1yljhu!69a
zWLNx8t_tx@K!hWqHKjSg5H1*{AwHM0=*fB{f*)WSFznpktM?<dpzPgq9#j7xS#JT=
z^dA2YZ=*v}32CG|6c`|afG9299ZEAmKtV=>q;!lh0F_dZ80Bb3sz`?*8;CRriq!L&
z_xJpN|MNVY9nQVn$vyYn&pzMx`<1-_WUWZjS=f*PWb5XGO9^83nS^p$Ad)#eY-~9y
zo81&42UsO=l!NHw1?!{B?^~<*zM3W#CAZE?FeQ<Iz1a_3{<Jqv$O&NT-%~<<t;(N8
zKxGw89RN+?Xq-F7kGRjxp}4~5>u$7EhrghP1y%MV1Xwd_uz*W&tme#3bUpGUR$~$H
z8F~!4u>)*0#x>X|`z&tZK{bCG{SGUI=|<50Em(*gD_e5$tT{p}Naj_o3Yu?h4q)aG
z3h2vXwlspw(P(_A8Mbz4UM}}hr)QIP$0{^4B{T=wS!WaR%_aQ$C0xhGuHjd~>rH54
zN@vs9_N2}|#K+4yXNTScUU7=5GH46rgSUuRb2L>IP>>EY4Ox1razFQ{Gf^g#u9j55
zJh7r1tGQeE$(8AAgSYcSVsoi2G!Hd;jV-rg27i3h+rh(T`OpONB*ks?fz%wzT_uhp
ze;94A(=1PO6HR6W!i=AYecw+=lBvFt*1VDmR@(i9O7@rCL6i=_T!1DO%LUonlx{hO
z4Kqv>&(>$jwQXJs1H{rV?16s0j$8b)m*@Bt=QIM9<8D+PUL!8~#Fk#6Mn)nNE%RhF
z(<v=P*Xf8o*&o=(E$EyxSV3yHu7)~9T>zC~08!*wjPXZFJ!ganCi{^BWS*riAq}GY
zy`LQu#c-$_pSi4eNvrS1$<c@94wN6FB^Dv|I`a!5Tgw}`zxMb_Vzb?)q6b>)>%#s(
zy4FX4aSK?t9_T#QgxoxFtsV>i#Fd=7OZLL~Bp5hjAP$uiH<e!X>JJV^58lY5ajtNI
z@|}avNv+7iC9oHWm}SocE5ZM&VxCh2OthwBXF`4}HhS;P$c=wF^QWC9HlKe4Ykj%?
z76`ZiEd*}lKR;fieKm+HAp^;DE!xt3S;xWuR#yzLE&^58SM{g?V9y2B;L<Cpt`%&;
zeR&8UTd>2fE`foaKvKC#5*k4(_J!Bq3iJDZotc_4)IT9O#^&ZU{sA@wAQQyYzwPL#
zvB@f1ugf}N%X}cqM}oEeMVik|4xHx%X)bWuY`y>IMSt+g1N5{3vU=`bPbq@vF<lg9
z@kpg@>iYEWUOgo>7IJgk)PZG-Ae94}Q+=R;CCrDLT0Mrrh~_5#V^#5~hdc~i=%kR*
z?+*)mKAid?9NRLKH+Dy#nv;N0M{8xK@UJ;>!}VI;Pqw6jgD4&3(nbr1hWN<+Y5yT-
zb27JjxH!<&?o<Wc?zJWT-k^>i=l=Q=IC2duN4nQ8is0?8k<Hez$e%yNGMOZ1kBBmQ
zST;}X%w&)BmpQx3Um@bwF&w4TW#*vDnS6b}wLZdHua<%$7wElMwZuIoUZH=_^Et0T
zt;XtX<k13yez#Vasq;`zvY{YYDc0~>{_JaoMsNLI$+-Ax2Sl3#Od+&$4oaf>F3NKF
z`1!}g=|lfF<h+x)2S;=e|G(iS#4&oH%SITQv+Sa<>Y;?Dp%17zw1&amAubGyZ(}mm
zViK-S<z00LPFVq%LRzQus}4%T6>Qy%9Goo9{_tJr0`~l&HCv(=X+i^0`}@<_CEQnZ
zsT1AHv}F;IQ{7R>lcJ2o$b{QnSR1sUY>t!*wXrtQz!fQK)?TE{5^I7!NcS*MJxTUQ
z$vig(T_Z8kh#?gHz0{by)+gN`wpCpOD_RNnzbpSXGBG{aWTF$7xO6ujnHdhTJ0YZO
z<3LNuZojJ(OgZQL=EpT3?EBywc_}-cUNM9X*gzCH_>F^e3C!<1Gf(}r&lZ0C`75_L
z6IgFgCePDa+jdG?41(Y!eIvpMg;Mk6|I_YjO6VX5u3fwp9{$6pdB4^Sq!=i;gYq0@
z&dL&+h`%N$E+KTpG|2%icQYJL!3hNl8G<)q5eI}b|2wY$c>rX?uOM^m4L^rD$WS80
ze}bb7^fiQB&;P^;U{hr#Wd|O(o^P~$EXt!qPyg@}K!3tR157<2v<7I(Owthmzy!0>
z|3Nar-+~nwI?(b2r6DOY0tcR*idm8*j-*p=!ySN2zLLKTv+{XQ7juocp`dw*=ZTY`
zD^&PcjT^K4R$$SYLPVz@&{X5&KZEe4?8JM|;6h3YmQ|9h6?ApF<`>IC*8J**r!9OV
zd=${nJ!Smv_#vQqUdgUA9k5RmQ*M@%xa7MZ>lP6z-}X8vMa2E)Q`julM~S`^AYL)m
zDyP*=W)f$M{(c8-ztnHgH==+np-Kirl3<u^Hu@^urO@W*AjV~xg`@`iFw`gFA}8Y-
zf<mZp#kLQh74%`a>F&C8cwW$uo)Q&vcP$!mA!H;DXxQaI5R>uYi>i(k0jcS)5~f;5
z?Dg*gRj*m+7D~FTP8-r5&+!=NR+h`K%u);IhLbqVna81*G%i<3blMv&D!Mv+&#JGB
zigLb}#}Mzz;h{g59+{wpmFq>;+l>lGW(RRw25aXWz4p%9lttDh#PSt~>GjNd&E<ML
z+<^FBmZ*i$Gpl0jY<Y%k=nA!+D)xkE3sWuc_HN#R3?u_gr}g6E4knQYMnnecVaZEb
zF)I$zC~Dn=d(;kj%0IHm1(ScUUO`3u$nsZojgci+2qkuixmv2MzEQlfNI2)B;GCCb
z`5T)BzvkX77kh^DUwImv<>^-n!Wh6>4ClM?Qs#(c{0$?b#YDXIhWBD!RcmX@wANY5
zKq^$S?T8Yh00QwB)^k6cD}_iuzCPzV9~1Ct7QmlU98Q%BcEIUqmYi6^;GP^AkP7l-
z2!n~Y_tJ~B9)rZehh)Ku^B2cKUD~WSAs!I~EiZM)hy2;$A<$63@JR5cdwBr`0aF*5
zzP`K~VOhNxB00Z777bN7*-;Ps^;rws>j|#%fTRRELc4uAf);c-P#WgQw-c<A;2;wg
z=lbc3sa=i<4{b56ATyLw$Uf-oi?{)1K;`nV0GMEy?#3KBLG_W_WYRV^TJ4`TTE_q>
zMVE=aNS}sbF$M|jbWIn4a;66kCs!wtBHOhgQF}FT%JLxRQ}EL#iyeXL0q>$Slbo5M
zm5mE-EaNna>)}<%7Yh?6X7u_UG?O?lWl@L~Owb&QrA)ev8FgL4@k&|x`Vt!e;GHT8
zTORg=d&`w0sdD9Mv`7BJ12lUM>$j{Ac7`0u5_twmvH@!kEE-!ss!fv$7B+VOsf6ah
zb@84-kt0o`QNu>UemO&iwHI_pI$x36DQx$7#yOiRbG#gWFS}v_nOzySqngV%E_8fg
zmm}&rTe|8Kz$n)@g3CE&5<bosAcO6Dr776!JTle?dD6U+vZdaGY2VMYl$u}}y>$qk
z35*LjY0QP4x~+mLB!22T9FGFcJjzcPe!E_0T9v*n)~|NiOu*ZESq>%WP_JBYI|m!#
z(KUud&<N;nmj@O`HP#K0vd=!SwcPmX?~v$SX}=;my0@*h8EV-Jfi3rZpa0RDilwG<
z@`4J4NIPXPJAe6#=z*s|<w);+M9R`#Pzw}Ua-kbxsgR|M<GxHUtYEq4CJR+@KBJ8|
z+isDJC_-&DgZ0!UgN9HpMxdSygPW;&Ao$BY(s-Vu4Z1E4_kWU@)%253RFFNNhZVEc
zgLBcrZ2F`=q>tSvQU-K&R!-MGMGolywj<~lgE<X5?yhn}{_mQbCqN?mTTKH@+kiwR
zjN0J#1%ot*PX^sf;Q9n!ydh8|_N;1<iM#YUAXfk5nD7seKr9Hs0Rb=>kPJ!&R*OP0
z3ZZz%1hYyKUIH5m@)Z6fO+kZ`$2zb&J`OF82Uz5pK9GSCE_I-t0jKz<?*PG40X`Gp
zmxFFK&^UkGfuNE=RJV)J)ANqdYXU|?awSia2?AL;ChwvfHe=h&hh(>NYgqV}E_gkg
z-~9f|E8$<FQ);gh4UFy&ATk1P)9@#wM40u59WA1{NXet=isiWE=G_&hX~gZW1M#Ai
zEs4JV=Ai2$E^5jg1}d`O_F0eu06gnj!)s~*>%#-khNJQ=u;{Es`p}jh6*gakEf0Cf
z_S{3&Z+tgtPtZFw-y6J9J*tmn)-kqj`7!HG2&<H3@8{L8GV#J6o4Ex8LOG-4)En+A
zjFSpf{Ow7<XH-1G6KPbb+8(+iLq(YziFF!^YT=FlF?0x0$2=F%p)yT6zp;m(`mW)8
z(sF$ZN$w1?=cdd$e0TCbwjZ6BVuoO@)Jzy{_1GL@TSNdw<S!&Z!m#S-+7lk6%E~vF
z#s6N%>bgRtj!=Qtt};~l$#<J{$bj_Xkf|fJV*injg5}8FHaPF(Ne}Z-tr|kt*tF50
zb9G+X&?~L7$)Mok`|1<v*|4hYLlK3?WA~*Z%NPlg#<y@Jkij@zC7DWQ0Zv|Cxwpyl
znOez?<POwM>PT9iKkq;!v)u+tm--6R=Y%g+(-%vHiQ`uYyXGTavNT1!i@$O46@TLi
z+pKuV<=fmqeSQTQRv~fTrxX0IAV;ub;K>cU=({KVy`2M1Mm}lcFAA?)%Zn(f%WNCO
z|M5IOG~bB`KT#$K%paeE5QAz%8biAjyBh%8*W_2j1GD+Pt6xV8nykd}ax8H?Oi8h@
zpI*#Sz_MZerR?rmPdFH%|8qqU`mzZz;RJyFKc_ChJ1(c(;t$pmtT5H#d;utba-;Yx
zu>OIcQvL+npOsy*bWrpl09*(&Nkc;!`=80lNz<dPts2B#2_manbq%0hr#n5LKK{*O
znRC)Xdva@{_09VH`1p3fNE!drE39_(xa5>;UJ_SlIHr(7P&@LZ*p%4Q$(x?eju__@
ziO5rTlGMKUoP7LGz7AZB7vQUYMn*!VWCNms?__?z9xU{c9Wj#dVu>RoKqCJ49D05=
za8g!wxhX>9mWzv<hM8H(R?W|?Kb&7*hiROAXMUP*I-Or~!n6uLDvlqQB!QwPt!z75
z0h1SxKu6uGB%y)WQSbxxDBHcwA*sUAw|SCpE&h_LFTEP#g;;}grD{2ntL>U4dMS&S
zuEj>AmN<*58;jgcOY)A-oQy+W7;J*-H^SN0@wJqK%!%U5%$%}psEv2yuGGuEL_Dr`
zmESK$Wd@WU1}(X;<+TkYJ{WJ&Lkko-+g+z+%k!#~Tv%ueH-3ag47IgWrn)0$%Vs`d
z9HEyh{@#VaCOX~N@=(K)g_(+)w%6iZQVS!L>t3adJBT;In&5A%^O^NAZ{!r<V;yB2
zs?n(dj@FM(l+AJPSG{0<a=7APajoD_!@&x2zoXyZw{DGlpNzTRH9Sc-Nqn)wx$q=J
z??9LKd*)x-%U)G!X*67dld+C;X=HaC;3lpNLCt};<eJ5`{N91u_z^50kMoGBBs^_G
z6EruPBc*tXb3X-NHk??(h2+bXYeK&FEEXQtnjG~Epq!mGE$eCrJf%$^AAFk$jYzAV
zVGamz1bq_3etQ<TyWrDgFGxbY?qJxH2{C+cNm$B>hZ3Zq&c>v=6j!02v!rz?yarT_
z=>&8y+c6AxUP84Y7l)Y=<}#tg1QzR|0wTRGQ!j?^CA(_4qtI=@*aVTQqN7-ha(o8X
zK8VXjxDNkYSVx4RqFSYOt@sFeT%++_@87q*O&^eC93}d8Bn6M!FV=tCzY}KgC~zNq
zh`cQr49^^-rvpXgpI85-)?S<6XD-xVvb)_v{PTvZUG3yu&>#N`*nmrA79p+`h0M;>
zc!VW76f}5)Hp->XFSJ6|tDANvtbLbSJJwrwa9BnVJ3<m5)7J<S3qV+DB}@O;77HdI
z@H)WUE;I5kL4(+!G9^x(A1}u}goTbkiX=fP156(Oo)?DwHUO;$=?^~-{+flQ-ClKp
z3daAOrzIlhAzOtFIN4F)wJ%-MCx=Jok(r4^W)p|l_znR=R?1%Nt&Idd@v|uRyne$y
zZ6Qqv0&NquO5taztbA8n`L-ib);2Gy?ai4S28_LyX01M8`Mh~~up&cPg~pduP&Te7
z?^0x@Uie@R+C#;Py8F*cC$@4mbF48ed3FXZP@C+UWc^ZK`fcM>$PCnFb}xjaY+!on
zyGtvBS7Rqu6TvV3Q3PpI+_!>vDUkS{!`3P&_-rANx;+CMb8XE9^+2={`f24bGbsK3
z5bhQV8X92pbZo~PCMI_>><|z=;e01VjziuqhxTcK)B5tgzs7OD4ie7<mfdG`l8q71
z$@9F5^i7`FB)^Dp!dOISl0=F@98Kc{A+96p-Y3sG2k1+yocu1?!+LERn4!INAF80^
zB|q9EpNw9^SmxXTnuFa%o6Qf3Z%1F8m61xxmLu{hxEoK$J)$$cBUa0s$Q%|r8ZYiI
z;hZ9@`?nU>K5AlC$I)~|f^XA-7l|sN(;O+|EDKv9mgDcb=pZYl3wNYkWT0lel}9H9
zbsQj<g_S4Y)5oyK2rCRHymerw>brmE%f-yw1<s_9u9Cl>x%_YBui}0oaonyI%Jg+_
z5i@$BPsCnssBKttX1uLC`v*KCtgQE<qRA4dKCzATV<norCq5shPMEniy2;q9L~xay
zFqibZe^|}@V_<MrR0z)deX3u~5$nvmjNeZ7%Ey}Ny~o|tHSI%Bq2<rg++NqMnQXoM
z#fJQ%hqkH)HcdYEY|n<vVfG5!AT1;ZtaqS|@7NJM89+h~h>m<|2i8X*mjM9!OkfKb
zc%EyQJwH4K<I3T=N5d05A|E0NmdZM(<k$9*uQVhuKo{^osZzVU@BW^FL>~3lc{5s<
z2JjT>`*6o!0+OAsB?11=fNk(Pl8YE-B>*GC@M_^as!5GF<v8<JBeO-H*6%S4;$a<J
z+|;4O771jJ+A;(eaYhvUwHMK1Nq_kz5e9P5JYR^Q>{RDOTON)60y_$2X+@|1*jS2H
zwd>lnnFPn+zFDp%h2(03^I8OF9xBl{^2{7sRo;SjsLQdi)Binu4PJX1_+B0;d~@90
z-nM;u&>(J*lgPRh<!zyoGy%<Yb)^up?pHQdQ*sgWcIj85Xm*2|sY0?_yfAO=?%YTj
z3AiT#t8<%~9Q2o|k%cBxIt}}4Hb!Efi$^p1nP4^;P(-SN<(vMdw%vc*I{H2dvL3q-
zWggtGs~R01qm!qa@>srQ6NKFV7@4O7u~fX=|1JdQF?6NWf-Hx>hZ-LTY9!Yo@KyF6
z^dD1d3f2Hle6Q(7Gsm!Le&AFdu~!u;<Hgy>Hk05npMlrj<=&#*hOUjVu_LB<hFNY}
zUAawp`K357qH0+RTYhGpLIgb#hJ}odL0t&yN+Sj>V4R}HWYG~}Rf+kMCP~P1xeIRX
z%R*%7R)`xNO8KwoqWe#bFu6U8viG0dxyZdTa+90}T&8PKtlEIS+!|6w3skCclV*8>
zwo1e=k^9?gxiLB5Ab_tLF;29|qoK#UQ}<;N*3KN>zwdeLJ@C!@SQ2rj+4krRcH`vT
zjl`U2@}jM$#rADI?!`BHb4R~@3ly2GsXjUpe&y~ixkJ>NCSTN7lpB*(mbp4Il8RuO
z1N068t-(S$$xJ4yH2<E(nUZB4WIG%@1ZS*!cUL+!H5FhZX&^EOY?fcw5A-U4j=WkV
zcaTEW{;|O5YJga-4#aH{1WMn*yaqybAj%DoC)~}#TpSz(Z8|{}=BJ=hId1epi07Fb
zuLyBhO{Exw<o4gl=Oh|1zNZoc0;<40@kb3=#)Y8``}3Im#tslAb{VESmmyg39+$Gy
zU+7bdA<Pf8`LhPpa2tYBRFbf#`Kz7#EXuVnxh+@z1*uxVQ1iDG;Itd!RyRic+vj8s
zY9RuH4;%lGF38DOC`-Y8bP6SZ7Kn*_qe0owl}dbB5t+p4vRW&_4e^B-Y^Vc5v7})U
zhjMNpzvwj}NE$a8S!BL?mt|-2`=zvO@hj+e%7${b@=Yl?hr(mW>5|Ps?^B1qQ}0W8
zx*jt*CH#{(vV4DKZ>=6hhq|-t_Ynsqb=(MgDa!(E!OZ8}X^hL~)Ut46jwnal#@c}t
z`34(2icEkF@?~yUwppi;#bk}fYe4w!WL<H^;)Yi^;~y{1Q3XMzt3K9lmx&+)dJ6Q)
zrv0Qm>swfaEerh0psNNSdtO5*g%A}v9G1Wz!#<v}d_|g8yG4+Z_eVY>x$^X_%U*Gf
zX?LV9&opb8$e16$pJB=OH||Xl-I(n*TX=HhCTGA&hoa?Q4HvKouzouDM8H|Pq4w#Y
zD0BK=Au9DpB!Wac6z?J3Fb7#=IYfeGDT76#(;s>(7t-ni!UR=~VTCf!Ss`~z7O&uv
zJ~^vg*-PZvfpH=)ri<D0K+cSB{Jo<22oE+f18@@yzOZ}zbL{HhuGPTX@MJp`;+JBC
zaDE*_%dUaHZ8idG|1+G+Geu$&JDqW=c=4yRl$s6mR~k&GZH;Bl7k`&52fNko>~o#p
zur*K>thlhwtn!ESEGgn(zfgbq)_c<uHIjpkIW^$bN1&5I(gRrVfaUYQ-pa{Z`|jW2
z5KN$=CLJ-`$k#QMC(`AZw-4J>!Mm3H62QBFe;FKGb`B2Vw^hKgY*)M5!VlQgYyKmQ
z>5uEDd%&QeSd<zy2?qXyzShh!btY$OaNdEstEeB4Kp+5^kjevCRYI(kib~(TtAHET
zY_MvV7P1sMqwqUj?Q%BcyRaIAijHUJp{0*we@sUXl(<XrP?){yz+*b#n<h7-p(BNc
z-9U21krP3A9u@oDJ|qmsFgHiLJLqdkqnBut-+dI%PxK5EYx&aulI<09ia(+)AV?3S
z{x)vOqSyn(djqWS1t|RrDU^*L%6sQti;d4Q3Do0y4)z+{wvYUr>7y3cSkMygxkH`s
z2hiv%mwv6t<Zik=dZbOc$lR@n<V`yzHwe1k|H<raAKsTgQT@62qb$5QOQZ?(Z+lDT
zYhA7x88%Hah_e|Bl2Lzp`2+P-kwi_lsN&=2ZLBqV(bzZ1&}KbzZMO0m?x~mizAAZM
zBhqSrYKXza^J&tR+e$q3Gs(oB&w#ag*WJs9&C_}1(l$_fQHSxj-r^BsuilMD=ZZ@O
z*=@<mT%+z*xWVfeg?Ob8!HYV$tYO4*{Yj*Wc>GCj2s+#w8E-~VXz8rFJQ>hHiqVE%
z&9G`Mcp>w=Q3zqow1s^WH=y||DY~DiXvWOaH3#p_re3DJsNjkPRkA*c6vnG(rl{PK
z_S=o^Fy33z(ZhVnx)A4#BVGu~0_QUH7N?;jB?cXWg7_0kD33jzAM_ImksaeyS^p8g
zw49?x?G;c35C=!XuUKXDhtd5Gou*=~;9n*WN%SApBEOK4+QPRF)Eo~Tz&veSjAM?C
z$qJIk!--3kHJOZRxG5wDbC2u(s{}HG8-^<3ZgsqBeDtzWq?PUV%HZnu^7)bVhXk&_
z-RGSLWFVYR^R$cX8^*mj!hh_(O>5&%*?W4|l^(rYySSCY;{3|25-|059YzTjB5>M8
z1@WqpGC(KRVt40umC<V<)zF9tE-*v<Y-nn#z~Id)V-eG~RzWQ{GEyzSd~N`#*Z*7^
z%X@o9;b+?NV4secSbHe;;l|~ZIK)D(D(GRt2ge64`GRg1vq}ghrqP!>)FxmUo>o}`
zos6haYU%4klM3&D<N(@pM-Zuglh~2?iWVrd;;V=@#0-Clw1gQXhr-NKx#cNrq(*fh
zSSchzQFm<-cM0Q)=pA79q0%2A;+4f4V=ofD-$du<xX;*M(HeW4KRZFGaMaR?(+y)}
zf2^qx3K-3`#)BYDn<VqecU)<es;ga_6~|6<l3(_7<R$`n^>Yjw-gRBGw3k@4r^0zb
z<M9l>dIb16EP3WIqAvr9Soj@xp+-;U!=1$m%BsshVRy{!*?=$xNK=zSq~90RxD4E|
zoglU7-x|u7$cGy0a8^Qmiq?l9FG6>&;r<8)==|!;YI|0bwPx<f;OrZnyYFyM=Q^AH
z8;l<bR0qRMd6Lq(+?)!Ue=_8tgVI%EoGN2Qsfj(cZs{w<RpoHaows>8A=0#E_4NwD
zI2#|`wPLYySNlK(OYi6(A&Y#!OG5nixaU|D?4%F#Dd2@<g3Qo_&EuTcZNiTWdL=|M
z_nYpD&0K<{Gw<3a)4xj?c1k89zf)i&(-}tZsK!~S?q?rBzO~MP;fK~j+uYb|JN-Hn
zB-HYxJ`rm$bFMN<H9~e-m*4Y|+BeI{>T|~8HGf<oeafH5o}%?+4+vEr8Y0ZU6z#nv
z8sH=@wfSO)|JsVmNv{hV=YhBboSTsBXxqyEstu0*Y^OD=WM!Q_eL)bk;)XGQ`=-=$
z?zsJ}WjYR!khdd1u?qc--C;>z#T3tRvv5+|yW6O6svGB<wM%5bKRg<HJx66)X$pMx
z1kC#6gq@%wS5oTk?Ii)psM`F$iS+d4E5R)dU_b!55o*^$_lW<SoiRXS3wV@@G5_?s
zvG%P0(t+g)L3a=kak0?EH@CI*IK9I-y@MFe5oW9o;4aP5<`1$nsx||Wt)QnE3G)0v
zTjBDLALKx41&DgnR2;D5fe8jWATLb8v!5PqOwG0bQP`mEd&#ef+;|>l+E{<i=>E$u
z#V{_rpYj4tb5!e*S?brgZee7fLL{$~JiV`K4HdjAU9vwJ9i4e_Ni<-}Lm`wT{c&au
zwh^H@69$yR4YoBvEKVs<v5(eMfcv7hl056~UD*zJ;m{$9N*LSN2XjeKpJCUB37ZQ?
zKS~a!JNOx7az2~8HQjCWy^}CQBT!MnDCIxn$Rd5afYPl%7t<B0-^v<qYkka=XBls1
z9*KSEl{w~r`RrQpswdpV=IOQcFQh+yVmzF#kG`e~7poWMZ%llqI&*1e;}G?s<6}#p
zL}jVV>-6|tg!cwID*l>Z{`dIdKroaMUHFwvTcJr?;!i`f7RRpEqp|2nzV@Js({9<O
zT2Nv(0piYKmb&Qtk?h<1`y{XTh4csDtm%G%DqhL2t}))d7=6&Ey`msjmYT`?_FXl4
z!(&Y4eM@rjsN#2GqY-TK?vLxq-1u}BDmqpphEBmx3>XG#FfI2W_b$fhkhZyL)Ix2f
zx=8$69Ab1JjC6Xej$I1F5HW}tbT5DL<4uU;d(W7N_5Q~_wpRgmHt^B*PH^2%e1(eM
z3&}8jvh1?VIx;Aqxlyf%lN9<{J1#c^=*|hVw2ubNgo(MV?NOwXcheks?Tby$>T2v3
z6XVhTQ=)33?4qrLvtRWe;w4Kyd|10$qi4N;6rlHa$h_qH@k1?zH%;<I14TK3J6=t*
zz?=z=EI4f;M|=E)*hy4Tk>;Xq^O{M~*6RU$9b#|czcCs>=LFD#L`eSfg@ZRHBBGAK
z_SVYV$D&=v08qh}c?k(g$xAO!{vLSR*x5PB<d@n=D9O84kV%g28@@InI3tM80c`R4
zGd=iKeh(i$85=VO5v>G~BIr8aPJP(^c=_ngsxl}1^5A}jA?%~Fi4T@R^)9l`On@35
zb|lJlISeFF!7`(z5_03E^QxK#FP*4CSd@J)X%rSmumG-n>X6b8VRz=)jV0-OE`;(L
zIPY6>p<?$0d%I!z{b>;?_*m+*(e*Cu-?p`>7WzpS@z-tA{x#URj#rU73ljGDf;JB^
zVN;dNlJW5AFYz$^#-bnfHNo5p(W-m$nir=!4?{x1pAx13@>?7=yB8GXZ{?NZ1KUg;
z>d>byB4#XbTHcXn*}?$Rl)Sx90WtS&W$+E(dEtHK{;53k?0Hk4YA7^p@yXN{Ho3;f
zCU>Ra(EWjftUzoIZR*-GxULv@Q|B|DO<b{(<Z<Mrys8Bma)dq1HI#L+9QKDF>9Db*
zbo?yDnEepMlAcFDPsgC1csK7^_xi*BME~LVlzQ35FLe=Up1N!X*S#ahV--si&A~w@
zcKT9kKcqi{UmY^G6DCFcwEG>fW#&MvA@rrXc`~1<V}y|uL=tL&o#eHufr;r7*Hd+F
zU5!NM^G~%QV~R-BnIPm5KeZxWmY&uFE>Im*wicP)0WI9pn|L15TyRW@vkDVFpaA_s
z;KT!+c~wuKDL-I)ar@|6#Ilvi>~$mS?0Ip){Z6N-^Ar11<+Iwf4_r5X#h*vdde33k
zp8QSvaF9e}xPY)FraFnO1$XX#dxRUfT`lS92k<iBfPwEU&fsbV-UPx6Fwtd3iM}WW
zc>6QHbm93Y%A?(2z>VwX0k2yxo+6A@06GD#*LFLtG9xuG#a?381#&Jkw9Tyst6uO{
zz=ngjj*N_ajpnVjYe&$0*Qd1u0YU%Z5249jK&68icS3UuxUMOUI8$!@WG@H*&_<$i
zySUIq=ov0uG+iORrq7W<AZJI$e>Y$WWlSmuIvrKGh&|d_^;ZU-rHQ^0qN@G*&{!eV
zl!HvLV$xO;g(7vKLJKe_SQB5A6PJVuS})zN9_HPwj>-i0-pa}5`Y%^Gsr_3(v8THJ
z*KaTUfu4SJH$!uMV{TJ8^|AE*Ag|en^U}%l>6z%2NUR9g!1Rh{ADh@%%{yPpI0+ho
z{xDH~Og|rc;TuorUGGv4PUR|9Evme#0ag%Lf1-?$gLb*i$gCD@x%BN?7EW}onq0fE
zGSt<hiNWbAw?C$l65AUnwx4Ft_8#0g*1x=Ccj4tsqfS@5(!<+gOI{q~%P3Ds-7Bs=
zskn%1+!iW+w{D|(;-5XiUip+4Kyfk2ejNFvyHlMtw=!H)fp7mu7|rcw!6deN|Jl#V
z7nQ3Oxx8s*yDyg&+S+}(z-kq=9QpvN0JrRUAHF{+I9Nj~SZf@cpqjk<Q4LP)%o8fc
z7l&k_!o20{3v+r;{FYn8Po|rlPt}>U0Fvk56(c=_xes&bN{~>f(CL$QEESg4RU3eZ
zLO&@N<dEiee*63Rd;xt2$o(&6v_AjwUVAS~wQGPg!)Jg&8)TpaB~;wQ<z^CR6=nWM
za;Jzse88LxNyFPy2U{8V@FjV)YI#)F&HBlJ5g>SX>xe4s@R3U5_Wd-u!sJZj)W=KF
zzKen#M|iyofAN_&qor)0a#g};ZyYWt{gEsQa{}l04wfw`Tbl&Dz|jI%9k3C*yK8km
zt5l^Eta!esYyjP5NdE>Tc7R_2MS$b$U<CsgGLZ5KWoy3y!F=JF#M2_#l;9HpI7V=0
z0oe$c6^5-P7^Q!1Xc1>Iz~+FEwROG^oM~_igELJ~KCOe9fB?A7&uapK>o=}4qdUFm
z*ZZuX%~XjE^f~!&gt4Q!L|2!mu3IVyS2()Sj2s|fHz%I0{!+vfzgjFfX`777jO4?u
zEV_77Tc0Rj@w~F}GxU(MnHS{+4H`qYNu#{{?b>N3U5w`f^QW#gv7Ctf#I2U5{Z^EE
zTbzcTmyZSq`s;);RTnmNw(sXnQ~*0uY@m5#$s{wIf|sXrtS3Un%g4IoA@yWz>;6y5
zdsA=Et1tUjlki%hw|{uA9Sn_OWds!9#=#_csA7H9iu-w{@OLZG*hE)S3-kBgvO&p5
z7#e!_#Tk_H7oVmHf4y6}n4U(9J_Wf?N)@%zEQQ1U#F}i*7o)Q3UGbFKLj_B&B#eU!
zGe;R)SH(t4yawrd4QU4HU4{-qU(Tf`QcVeF(Vl`*@q4y`H3Ew(&EEPjeI{?LT4@j3
zjyIxV*$s+UetW97=+e->Q6_tw7AZNZw~{iSzMq;M_3jRvwG>@wG9wY0Gl~T!#RL+6
zq*tOpXRuf!Kcj##A&)US(4e^PM^S<v<{qg7rM@vlmKUV1LqPupiQ|aEox{O+KLb1N
zsLUlm1qxPx1lw0i`2RBt+#I0(K1llceW0PlmALG65Qz;eg!YD@>FOUC+~QB)!!=C(
znPf#zFfbn9=fmN>eb3v}%{prcKxqAf=e?`D)7wXUXJeTW?qh2TAjgM_D~zZl^sHB}
zViyEycAxbfpli=%&XUHqHS0|P%@J%j?s2I|pA5W!Qyol|puGho;xID<+GBK$^LD&6
zP!xgl1_CAtihuAh2Xo0HxGe#IP^7Qbsgf)mBI>{aW&|PLo3Bm4Z4DGdND#f+Gz&zk
z6squTw-4ZSXP<2N$`$@JCv*^#MiV{){(*rn+5i0h{agBS3PD~6!Y?IqeesW?we-W(
zKGMfQPq;y9$u=2z6gMbsv6!ZX#n?29Q-O}W)YyQ!K&y%Qr$9?XhAIm{P>(^r*FyK*
z`@-Wg_qDxY`fgsQsJ%tp)lZ^`umE6y7hsBtr2^en2peYo#tRXoZNnr9lhK*`7@X;m
z_X|&i-K^%)Vo>Kyn?<5x&SCv>W~}Gfe)%D}$vyTAcv-=O$~pN<-ZoGH*MUl)UUru(
z@e$P)`Y;dEUzz@+i*AKI_+&B!#)qQVl((;;`Xip|un(5Oi_LA=vS|xJsG^{8rXPYc
zkNw&6J{H#8i8+<~NPm&vHef0UcHe)GXKM2Nh$UqxoXZstml|(cxi=kO1!w*K1O5|J
zPb0X#x%;(w_v1U3r)=+ruUjf|50Z8LI$$u%@skl=#j!{+9V?|dVcYmN*zp=K=Q0zi
zv{qQMMqmRy2S*`QH@gpSQ#(0*73=*)^v!JEM6kXpB&G`@T_#BFI+5(lP64AQij-a}
zX4lwOVRqPKpr(QfIp1;sv#nDePXQ?%{K`HBi+~uyl+;<)uORv!X;-qpZayUIim~<{
z^pD2Q(jQ51h~)`B$tMZwn2$WAsO(4<lbm$KCFDs|p?b5B<Wk2g!YGT9<KB(sx5@5}
z4IvY)y&`Y6ey>%_6I{A)T%1miR_%LQS0#74xsEgM!QXzmuk(+?*lvs7%i-LUulZj{
zxZMyC6gOTh0DmqBnevjm``YAU!m=v7q7rOPfLZ%>9R5FcTOls6F#)eF*qFecMSuc^
z<RyT3B#AG^vfAvfNHzve2;*tDGiaVJ^!W>7DTAf^z$pa3SBjlx{5K(y3Si1ooN2&?
z27W^D$pG~2)(ihlAX+O*W27imh>8y%28vIjKapJRIu`KcZ&xN==f(529O`1S??m>$
zM7kb)DP>-oVOE-%?vUp|p+3x!3o_q3uj$5CsydD#GCdI<UkaT2pnt(Y`(QRybfRK^
zRMZ;c)u>v<d+4?Zu&=$JA?tXD*uS+G2W?10!_oYcF70iQtjTqCo=?_9SxHe#R?jPE
zWpHA1na0nkopYgS=18njOTC{{5tLUix4yIbzSehQzC!<$;lylL`v#QmVin1R9-Pu0
zxo%TxP2^$h=K|x@FVBMn>wWwES0V2#Szv&*@as2rOTnKmDwTZ$S4(Q#kBx5>Ch?n5
zA`LKDO6pYGSdt_!KZ8`HJ>G&HI*4kH%Dj}U%izYe>i5MmKMNRU3xd+~MIn}AxvtMu
z?yUw(r;8nD$5bB7{R$VSkSZh{e|(b1$^9|4!Ml(8PfY+vzeP@RW?j@Xh5SJ)W`T7O
z4Q)W7;Xq8QLCtQ6L||CMq^w|G`k!m(8KsEwkWVA;5Q`dgO7ZJ6lHzleNh(;c3?<&{
zOjt>#Yac`TTfue`g^-KwT|dg2YtJXsPJeH~^1(c7MZmZ2uKCH(5e@K@li_2R_~${K
zs^#PR*^7jE78gjB5kW4iV1a%$_vhmHgS4+^u1%)cyaU?D=kSH=5nO<|F!)Xu2dIhE
zF`|k6q})G4`xSroj*biU@hO31G{<aX^#P!-J-$azyPW;{>?VjrkN_86`IxX<J!{S&
ze-yXhr&nXyCSbAmAsDZ~HcKJ24G;w|4S;07Z0!@sm<SSJ05wU{aYnCWj9wG--mITC
z(H{V!`EieNc%stv+dwV^fCi8g*y&IT!hTf;{S-`Ag+X%Eu-m>>>ZLCjx<qjy7Lr{h
zb5(z0VJ+3VsH|{%qMV@H@CT-5saVI20<rm{kby`8HE~r0hvtosmUQ||gNP1uE;^J~
zthq0hzJwHw<`S;w%Y`ni+a%7Z6tC#F7t|t9=rGe&)9_Gb;(>WL_DCvl*X9X<H-^_%
z-!weWcFrHHJPpyCba7^j+ssr-6>VH?@oNeSy()R5)jYK+EVtyKa;9O6=#J|2goK0t
zjAjPydpOl>sTN1Z_#nQs8ps!xr@SJg7{^|_pJ!}ZsIFesK#_4-_leP`2{9UzdFq*4
z?=8AqH(<-Fa$<YtEgK`g642Omf0o~g6~U*ZQMu|wbgzE-megjUBDpRld+NYP`{(Hp
zskBH}==rVG{>wtTHj3Pgs-R?e9_a3G7z1g^j5cy4VM0d%{)Ao$KDe(0dAUk)!E53#
zd%GYK|BKv>Mzi%1!ijFU9vkFKEx0P5K1J(gz-%#4$lUWinQp#UAXKE<Ba|Da06JKS
z`E-#6x|eU&#ro;=ff#BPQay;kT!vaMv7F#@3Uy*Zt(pM`3yspcQF1DX$`vaqeVy+%
z?ya~aPsw2E3BcQW!*5UDFZ+Ml2VB)zY&W-q^ID!*r^}JH`k-Rf24E+^SB14HwWt;i
z5sk&j#}v(~p;*DtXIDF}h8^Fz1P*MYDYj~2d&F+({KWGda{jyN%=Se7eB%587zeJa
zo^EWPDyr<CBmpKOC9ofy*w6L#{{#*R2_WZy00bBTIVEsF<$?B~6N>InfbTO~8?k8t
zvnd!%f%PvnEo~R{2r57PMNWu#0xiMdgoAVL(NGYbQN&p#lC9nK*)Bd?Qc|+$1pvGN
zsm#p_8ZJNwi(PF6rxP5+M}hh*f`bGw89)O8EWjYx4}8zk_I+@y9&DhC6lCz>(ZFa2
zj_~o|4Hd6=sv>YN+kQC<S;psef4l?w20>Y9uFE_J{0!|!!7EtGe8-4RHK8+$<-y>g
zKieX%azgK`q@jTCQDZmuev{+VWss7<O?+|-$ctaevQ|J8P((uo121he>(`TjP64zD
z{tcWPH>{HOfu;u^C<YYEuN9!Y1n18lzC--(zT{WeunqSv9IVKx^|Ku5ldc7!A{CR>
z#F^PW*Y?Bw=2W)!j=jb^B%dulufb+)>a?=1gXo1?I$czavq<#SQV(cRCc960Dt+6v
zt`^Ys=I!;xn`c&wn?l8Xf0U~!a6VATn(OI+R<%K|%mJ2EVDRsn;{8v`^JGPIdbRSA
z*tNg+6<9xE=7k=R>X(OD&b(G2O~2DizcrHOXRyLRT`}dFA3}XC2|!^#;2!cz#mVHi
zQ&XR5z~|BMhS5R9>t)q%bkok`HKsm&Ycq|Tu0!2(rO)o|h1i=e<mlM7zC^kRJinFu
zc2W9fkE`YumOc3h0;Ac3igix0Yf$lKd<c|;;&kXFSi578IxszmE~~R~LY?ugym#s4
z5b^eOrXD(Ew>#>XyxE|@y6#XPqLA1x`os79kG(&$Gr+MXa=>^%xd!kqjvLiP4!>eY
zvHhVrbjMF=ghq{OufDg5MP|*Pfr>V81Ft_h!;xi1ghk(H-r*8>e?oU$rSl={Bfj@i
zdJ8Y+`ORb~=d9iP@XETGH!c&C_NSZny{&+OsOLK7TktsicQJqB`?Z@7poR0D-zC0^
z*`n*IT!)iyf=4=HOU@4`ccPEKwF42vd)%m%S*7X0*iQm3_Ak9}C!A~lC~WM$qBok)
zM!?<(%)`gW$Aq{u&|t_38eRnriV2f1_&MOyQf_s_`xLK-WDmNU{6mI83&uf#eg%jD
zM03Ywj@g?2Hw~k0NS(i%bAS&Cm?prfe{T0B`~#4NR_uB-JQPRP5#Lrwi+EDVXukD;
zfhlJzh=(>ay-sp2;}nyI$|CyX_=-I{nmd^X&;8(T&K8!KA$8y-k$6-|!qP=_GCzdp
zwBEtk53!&~2os)&%-pYuEeo(0-2zwleW~e*1#S25JvV3nFi$nkJY-{eWWI=VIeQ><
z&9a{wM6TZHuFRjCNXVY-xS*^5+tqdw_bYJbnF3=!<^Ig~suJ}%vqlN}>85s`)waFR
z8m!U7@Uh2|$NB!^c2EH#VTEZ(>XT+0YoBQS%)yU_0(QAX)JepQp}5G;X#MuMld;Vz
zP{iGwsG^M-H<G^~Sn<q;ZO7|PW>sPM&aPgk3x-CA8HQy<UyL3dJ5JgvuA0y~rl(c+
zZhNs06tkYC@07WWRi&MN<I+7p$!gMO(qlqda&vd;0uADhp2>4wPbR$(C%{+1(;rzn
zLhIU}0HG<KXPQ{PCi^jQcf=>=0O`Q-6`b(*=lI@6{LMpCv&6gZ#N|Gk2q!0ID3Krq
zITiTT#HcoI2Z}hjYQS!)FjAPMl~~F#h6=*?{e*$DQ0h4a2n@N`M{Z0v%u)egzDkxy
zi}?+sY)dE8c&Q{P+V4UkFqlh3z-ntm$PyK8??~P_xUT}Z<FJGeu1s{tMwcuNL!bN+
zByxjzKtv#IuWAB7AOiTBNnpr*hzodA`osMTCuvRC`J;QeZBtjC9fYVnVvSMTpmn;z
zLpR&e?p(Py{l=xH#{TR_@UGxvy!+0-^={)c*KDAo{%?!(58GjkB}YGvPqsG3%Z<@>
zTUf`4>i-V{Y}R0<`HvS)1(!QOfDe9N0vPUHiPk|Y{Yrjly>?j?m^Hvm0kjh&d3Tuc
zz{L$1W-v{Ylvl&DLD*D1-wF4Pq}k2Js4T%u3IM@5y#p)Wi0Sm2U7~axNR|RDHuw%K
zxu6pU{AoZ|RFc^MllVKJ$na=*)8V=BYfb~!sf>`N336XHEEK5W{w`(397FHL9tX5Q
z<XW!yaxwNCd>wZJdloumRiu$P6AgQwsCIorGBWQ+#VaU>$zA7roRcS!xxPo3l{XgV
zk(o(h-;@-l>8uU9zUJhp%cYFZb`lG)x%oo%?Jfwgts1ZI^>K4ul*8o^?J|Q;#$V@w
z*nIieOS;3`g=LRD6m2Cs9)EaTWr{z@p2QbaNgBY=uN6dXGVA?^j%tSlr#F87txRn-
zIpTGwvytH9(-U3ls{{ose~k46OQ-n4*oYdiayYoB%E!uVduT<0jdSKsRmU9l+RxCk
z<7~>iEc%W56{;?b7AUfFfu$`M>e&Yw0PchP=Ss|a`ILyMzN`8`nI(tMPjOzb^DoI5
z#C0FaB=Y&OD~ST_1cODK^5}Qa%6-f_(EJk4mxLU+ZmBx5i!SsMe4xAXCM8{KOor*X
z#obv3Wql%`W@h*0)Auj>m2&M07~<mZNM`2`<(PL@$=^$9r}6XhP~483%l4~V3vozL
z`IN}$+$Ekb&50I>cM#*_hD1Z+S@+f6L!7@TAV_(!4isdr&#7)T^DQ!HFy3l*LGhX;
zx%Et;=!aF`|5LG1+=^u0K{CcFuN4f@M`ATf&zEYIgg|1#+C$vdih`{r{VR6WsnevA
zOHxWAI)>;-79$BY<iHoxf1wK4`!AlH6Ej8V%0d>d$7OFnto@Q(i&Kt%@Gcs=_LZN8
zcS`cO%-S)ZfA(edOsaME(7`>d`pM*nBdP&YwAUYD1F&#@PCF^I?<8FR^MRMmC$1}R
z)ao&J!XlO`?R!SUwo3@WhFO^r1;93xl#~`x&hIQ1$1BbMj0AT?gWO6~Ny#!G@rqaB
z3M0bi|AE0E0q|dDyznOgZWDv8k#O+?eKZ(t045D0RD+68oNxG4jerjj)NTD2R1E$i
zUzeDC4Qv2o*tHo`k!%nc1x&S|q3bPh<^B(Ex3goYw+$$ev9(tLPuD}5HDLoSvIrLF
zM^yi|qtHOl7`6_kL6fP95aP&EkNF(P5<?y$gmjnSVc|UO3Bm2$qI;{uM=?u3YRB^~
z=ti&&I|?=1v1!9ct1ovL)C(W>#$Yv@Jn0EQM}_mHS&7i~$zVOE2B)(AAk)*fIiK!p
z(=@*djva`MhrKTib}Cel3tDkh-kZJs-Z9^fEk*)zn>kQ#VqIG`bo0Too2;wu*2r7D
zm$2`)PNS{!qnW+;SE}g?G?}<f2gMxfIKR=fm;YoLaCc(mzHe%oNnEp!nOByY2FM@L
zUI9|u$1`cK6Q@zU)=>?4%sA;mjYOwNHo$t}4t4h=>t|>ltoBLUJJ)Sy)}_)2?F-o(
zt@ExrgW#45*9&ert}bG5Nr+?3qN~OjG2dfb_1-HqoRK{jPZ`K~;&M7E*p;L5HNz<M
zqi=T(FnFal)e}D#UPE~)tz;0PSpCPQ3I{!$OKdN39mh4SW3WFvoZLc3I5fJC{EDxF
zOjoQ1HIH;3&jOJYvBM6<kYGA(Uk)S(baIr!AxjBEyrfA#7%*A>HVlqkJ#Bu5q=nEh
zZ3mqdM;fT0WLx?%B#hUuI8bi|;!h?(C|An@Jjk!|41=#PLC2?l@lEAU<9KAoV8Xcs
zFW<BOkqK3&|9N3+N8Enn)t>DJS=pJnL9n|7jRv#wxMoq{W>7XqwZe>>7NW9o*^ECx
z_HV0XM-(9QFtiaH2H(G}bNMQAR6HvER$RO!YajlE@Du=)UiL2t&9Ud_{~rV`Ex=B}
zmO`qq3TAh}^buMR4;~V5xT}|6nTU`_IgG6-dcVf0eD!D`r0ReTayO(3QXd^n1f<K3
zAR2A+fi>JgoBM~Jqin_S56@h#@0+UzyR;qWqdtn+O4U--m#U@#>azVi`^70p^b1qz
zcagycWrHjQPY|IAIFqq82L*N~;9exN25W*eXWBdPApnmA@UJp@{qpD9_PTbkVNU8t
znA}A9wF;anQ64;P31^tSyc5lhFa^v#GDCBk8+xN(@<}aPV_OlGMLPW5zdo%Z{4IV~
zV9`Z9@{<3yEXphJw=D_qrXXc8`%FL)=hK0e2|d=6R)Al~s44vB4MfV33n=#L8Z>;i
zM#|7j<!}!by6K&ltKQ(qX&{=_?!$GpS%LL@xNrNXGjr|d`6BOrRP~8s4R-JPXHGA=
zpB_?hYeAT(W1J0C>7@rObAMp)zO#gx)C{PIg;ba*g?eA+{=qwfiQ>9Ogle&FpwLa7
z9XU)f!g~eD`WssU4CQvqHRJM9);c*{Gpux^4|{WMjeTF$YqxlAOS)OCUt00om6}Dn
z`D>Q#d@;I^T{;`0Y&BvDNBq_Fs>q=){FbVHGa!ENiOtuxK(7f@7NGcl&p6TLGa4vg
zETz37pKQ?f_&_eb*DJ!#<JJ2)?v<#%iP918B(qYK%fu<81qU<CaODd;{}Fj253#x`
zOy7<{9aoUd57m)!WVjO}OjHQVqrcZh&PeB$ECtQozv{@&NLs+~d1(1+v_n(Q%pj18
z@<4z)bLs^<3fzZ*nE*Q(V-rRguYe9a>|LO>ztEG8Ac5UJOpnRd@vHeS9>n_Pk5PXq
z-nSpGFV{Fx&gKPks-QYPxprjbFh1ouf39BR6+BsrDv*a4Ti1&>2|S(sJSoVvbR+cl
z_VR~Q^>dXoQ8ISL_Q<odq;EhM8bPna2@YPv0fung&sS+Xwz%=4+6xMj+1jr_5<-I&
z;nhS){veoE0RspaKOm!U9tx8kT<aXzZ;N4cYnts{SSUtxDFbz9i1ICv!2OR>yc@)U
z3JX36`1Uub?es(#vWQ1S-EC-CO5Un@Np%D$7J$|OoEq2?|CK?5hefUZhg+{r5G76e
zzqRoQR)Ca)#bLnz{MWCGgy7<|{>1Hz6hcJ4hwYtch*k(LRV?y5_b2wJS5}v~$&D~F
z1|6aw;UJaRP=S9P596d<by*u43dKEPj=U!Ywp6&nbjL{t23y|dORR4Pr7v8>Dbr28
z_qvp4BNi9hTCitn8Eop~OAab=tB<6r(6s+n)c!AzKc8Y@hXRL7)(E}wAH`j!KL_x)
zeml(u+Kj7AM4<q%Gov@CbJ}7<FU5NMaV`$cQn=_+Y5cB0p6LD2_UpaZEum;Pc9gg`
zkuXHz(v@kgfN*;CV4k?3A_`qIZ&tNAlg@^{QL;=ZdAb}3lKA6lX5G*E`kjHe1a<j(
z{#4&wOPU3bsnO=E)2IE*S;Se+zg;Lgw58x%5gTzs{C@27-@ZGt%&+s3h(!nZU2qYg
zVAp#49+vGIb|H*OHGo=2=anL9hO;~Vho0I2<(!`1z?$@vXDg=GJPTZ#uHKk;*B{H+
zZQ0%&=o<*qm)NQBeg4=cGC>+fw8|O6pz8*e{?jwXP)dSf!b$HA9)Kz^3pz*2PsvD~
zTX{Fz1d_*+74%pIp<?k@p}WEqW}a%&r5lyZJV>c$)XVgKet-@f+x!S1j!B-Qz!VCj
z365<3mo}QzeCwrkZI>6>v;V0u8nuHt%ocKN#z|tcciuku?71oTU6_f!x7WF<C9&`8
zHL3wLJkWS1XOrAvFavFST2^~58S;CBMCBO3g|8N-!K5C18|ojn0)~@s;k#@AE!2L9
zi9cmq6Ro|0x%%s~n5h}_1MS-%Sgj>;?*fZS?PRv74}ksB5r9o$iQ7?@^bZ(ZbF2hm
ze&B}vFT)b7tdxcepuHKSs{#xLOew%mNR+wVEbuIi>*3kO^B=!UDqXjW+tbePoy5!K
zjnA8beYC-f1VXHa{0AN)0h$dmJm?y%^w4Pl{x$`h>_Mq9c-Mz-<|9}FxMS_x9^0SB
zwG5VCkgouSscN%ztHW@R?xhUN3hg?YqKyif`Q7`7STvr&C6)cD>pVAZk)i@melr#F
zL+T<%hXpQ}bZ}8kVi$gKa->@q;a3-aBnM7=f(0%(coE0DSb2XMgx0iqKY3LyUSfK%
zk7xz%*<T$d!~!Gwg**#wA))E(BCFp_n9bOYJr<wYA2j&Tvbls4)nF;;zlO&+`PUr@
z&Lw@j_%ju--gTk<rYbH)-eXxG)~?&pl4uF!R5&i0D9;8KAN#@FuJzbtdv+a6!TQR1
z%o+WPuDZPYeK)w~+CRvpv1(+D>CaR_6U|x83}^o4E4(w1eVl67OsaVqpPqOZ5BF^R
z@(f6l6K=l{qI~aXvH3=o*tSpaha4U?=j*|o*A)aJ9Vp~Q$*pd7Q{G<b8dfN{MJ%x|
zWg`7DNd{Ufg)q-?@F$k9Y<{hBw@Z%|EIobB88ctM)5|aWxo54u_iAyrAl4UHPrk$?
zVYV<+_1I?h_BZ(}3i&JvEL4d++;5Q(E(aB;4qXm$0i$zNE$>jmJ60qsG@wt4WC}#f
z-_E?P4=9?4JYZ90C6}cuxdO|g53tsO@Z44q@{~MlI{$0`fY}fnXdlxpk>_JU33Sl|
zJUz?{FzHfdwJ(~M!A5fJ1fB+kYBW-;qJI11R_7sXuy9B;fI{EaT=ct-SBGkQ%di*j
z>%q_LP1ebBADaYZlRS9(MR=Ra6pj(bk%;f+DruvSH!}Z>5m1&u)E4sb!G`spA@dIj
zwFW9Pfg+xR9EZOD7oSs<=@;Tm$w>&dxq2D6Z7GdFW*NBL!N5|WlaiS^74aV<{vT(U
z&9?HNKM@>vz>&{vHo~Nfw#pyxDt(A<D!I|SJ#5FH1G>_?)CSD6PL##Lb7FFx(r63#
z1pX_4{5J{hC>zWtPjZ}t|595p`y0I`1ZA3tfDAQ4nZvgO+i$Ue<|GLCeA^G0fQm`#
zata`ZOxm9xhjw_r^c%|D@%xBgDQ0gEJK7R;une}K>4rqcq+bzqt$|`040BfRpaMv4
z#iROb9?j?s&>Xup?5T0(&?+(|?x!!hkX)#FItsctLT7Ozeh3Ds@xiJ6Ry&1h*_>;W
zqcxMQZE{m#>V60w$P{JI3ho}<ElH7?8QPEpYfQ9xg8ZEGc<!ZV9jASK64j($_EYJq
zAjjH5AJw9gG9>6X18N;Py4l9|Cz$v5rM^;*uc)=<$h>Ryma%7J@e1JAZ|1b-p4bR}
z1p0IFW|3GV2d^BoxqMd^x&XV{P?Cl_6eeL|>%V?iCFcUU*}9%jzlIf8U8QEb7Y+6u
zHuafs*08vur2A=D;xE3$=lddXJL&(2t@n<ny8r*j&$0K;$WA#{R^c2gG7E=NAwpKk
zK2~PO2$iIc%tK35wnAiP@1)F>bx1;HcE87|>+}BOd;49t*VXO1u3N+H@;r~%<9VM>
z`S%C{rV}`UYdcIYY1=twowzVEWV@g~cS7ejj$JG!vsup1&_1)t)VT@NxsN3SBC+-D
zeL3v&>r<FTUzc;oS`CB!OO}<RFIw5Ez0F@BDJ8Za9o_8Mt)F~y%hwq}$xexqXAyym
zYN{!dCa*PH6stWJB8-(O8<VC(x~#hCQcHQUS@PIp`l)vU)1#!(K`(2f8=QU;1w<w*
z(LURAb;}lG!i(^otdIp+=%rX=Y)91{oS+u7!eg5V?=!Ln{mu)dTBd)~-u6eY_sQvk
z&g?$fNuwaWsqf|%QNA8$3L`(B-Tt$t;@gmLpjyDY`{OMqO!uK@bR9{?rpTDZYj{PY
zq{<+(Eth9SFAGk)sQIzW{F)8ldo+3F#?p=E)*XY*>r1?kz6)b2&gI63|Evl8l~TLk
znSpimsBGNdULzX<EFMNHgUkcq8VE{CT!AeFBp`61l+VPn>z;63grWt+V_366^$*cF
z@E#z;lNuV$0H?!kGj}5s02uvem;L4ZM#A?<Ub_t8YZc!V3PXPHHXps>d*kpEUV4%F
zM-8s19WWk%cMFs7<w{9^xzWkZf(wlBz(R9<?K?Xe>NL0~G|^5#k^dEBCP1_zhwOoO
zsi#WxrgiPSDSe5FjnTqsB2i{y`gt^AkJmC81k|Sutv=mpa9Z~L`d$)qt|yG@#iMmB
z8nr6(xSECNE-hU2CMw~kBi9SZ7;%d~dX1L*>+cF;304Eom4@|MzNaN*o}M(Ba1wK>
za`<+UN2ILW*^n&5Xn8O?4ZUXyMS1n@@m8Y!P9Z_^CC{<igH9Uyeapqj@=8n*e&nua
zLr(aYVz*M3)gSnkR~{vcLk9#a)_i1eQXjBXD9(?G>p!YW1^KJ<Qto%U+n2g(%GK8P
z(1>60?YMPw^W8|YIY)WfH~NWW)$P!#0`)+9=^<vL$^)Z~L;4IHJ(Tb7LqMli@yqZ-
z2{bk7*rJa?sjDUN9u4U7P*P|WOFRDHCV$E^x|0L*KFrveG(}uZdP@Cks(WI;9jQfq
z=i-S8><{nz-$E=@vvO&7t6uLq6?Okf&zbW#);&%q4Y%qsO_}LCmWdOqDMXTk>dgi&
zYW8Vu54GQ)czY`83gVWj%ju*QUhM_Zcrec)!Fwo*C`LFkp59m710NUOh_3c@t*Q2L
zYI(8uN$e(ZmsvhA)koe8#p)XAA8v3u5loGsIQ_m|0P#n7f7`Bo*a1EYSQT;<$fXoI
zg%{PjyU$f`B9flp!F^UNX#BU!V*~Y<k8mk*Mszw{^Y6A){beleY3wA1l&Q+nd%3}>
zV2qU1Nb5d&DL^mv?ycuq&JPLIS5G?~Jm4!s`B+Bp%}RJbzpLALeZeWGkoCMP-`*Ex
zg)!psj{6HeyuwR@6P$Zup*NQ3LdphQIQ>RaZM)=tkwJUA?O!z}xuwZpF5lP-4ww<C
zoHzT6Ug|;!J-Af<i2`A#cUOk;Xux$UZ9G-a=URis1(Z)11<+_iiUPP?KygZzOaB)*
z5B4xfsfXkSc;J0<`uWG0ka52QH7Xj7qrH#drM0c$miFyB*(13mbm(2ApJm@_byMRZ
z5W>LH0}m|x@cTCeuowqaNrRdTvj}xm;4=dk4%zcdw$T7o1ez>JTBAsc5D<|$JA%A-
z=!1?KgL>Az@dxru4;gW6F?_$LpHmV1%`jY^9pStpOot3H6ogww6;4@rm&05-?D&>j
zl<M?YHFQu7v5+6zljlF>y~5Bo!223eUoi5Om*^h&<mgBJEVq?cRm%0|on1nMZ?KdE
zi;644JoB5}(S$qG_JOlEeS$n?rAKp`I!2=jCTv05uh*m=7nK-C2=v?Ih^A|Nj$^WO
zEAw~L@~;G24lz~!U~^Zvt`MMKo?;sLej|0j#$SGrf|S0E(*C%H<2xodJ9JqE^C+x$
zMzZ--U<5Xw!e>X7??jBJ`KSb|arwH~;_a}oYV)gM^35u<Y4L<(<<wK(nbSA;rR9Yh
z+D{EGu3;Y2L!T_Qzdf2+CJLX}>8mt`#NbSmOWxAx6GAuR4G_M5i}c7tMvtwKZD-uA
z?1%?*5|lh_;+R-SdU%J%w+Hh^9ryxm`wumSbE()I*0ssk3iA1rd@JLw?quOJJmIF7
z5+w>xDMO`W$fRkQs=?7n)OwHMOJz5gU!kg}#?#YrihoH$H36S>jJu&Lz8|nmD?@P2
zJbUYkhLo;V&}+r1{dNL!)ua>o1nZTb@@V_GaM*<Jilo3?b~S$UX_9>D;!|Qw(t1kh
z?N+NH&>8{)WR)g&>p(#YP1c>y{e55DbsceD*&e$_7wY_^<xaQxocbcuamrRMC6IGA
zXS?pa_c&lYfol&3Z{Uln_bGQ5zU`$T@ThK0&At=wTSE*sUmM(TZyW|6?*4K`T}*H2
zx}<DmbB?){(PD8OUybj^(cNA6@4{dHESluva+`V0dn!OR0Fm6Yv$JGWe`d9^3Uk>0
zk$2;;>R1jiwo(soQyd|~_5^=}F0u}LVF9988&=gxh<6P7xDm}g)1^P+U&HsgjB2+n
zZ~We+-fx_k$}s&QbUGFkGPfxR^XpZ??Ne>P!$>Ym({C33HKl;K@xRO{G}?^yHE;#+
zi31K7T$k^C4zLMmOIP`purRD%4Qi{5gdU#QB*J3*H7hncGzNE;C6d>h+YI-K``!%(
zb>u4%3S$Y<kN88yS8%&Or3H!ZgeU^@hi9Ooit;k)gcU{4A!jNLZx*sO>AiBXu-z@P
zVa7jMouR&c>+x&m_m^0)<p~jG!o!;-CO0ch=Vv_p<jUO*Mr$vBW7%@d+*g-9?upUz
zW0J7$$lky}`cq36jku8yBU6TiIoqy@T(2s9?vpbG2}g32E--W{!B#fzpA&vKtMp;n
zMoQO0_Jq^1P}hJ28V7L3i)XA@be^sEv)yyj+!{Sk@oqV^+z!s?NrBdC%~3DMe6A1G
z&^EzVN}T;TXZgOJZAvyXLA!)e@jg}Pv#syh#`}{uuZi7mf9A?o&#orZ>d{hPqTVaE
zE`RY;7K?}x;$qYc3G+(x(e>+EvIZ-g1GF9o5{bh!0}o%ZP4?NGxxm?b^wC%=?zJq#
zv17U$Yr0hWaOoAW@50<{;t~+R>v@;PwrQl9G-QV;nrOeJax_P9(p1yr8abrm<*6Ob
zTeg_)?<k_Xwck*+Qs7xn)}ui&`)M2$C~C9hlV3n~9TXa^w2RP7$%$+faX7>}Z9$ED
z`o;qYPe@-N^Rv@juAfs5MMZTOKj67Y-`Ci6EvZf4?@lmNV(;natq++x3yrMHm@A~x
z-WC1MxpLhrIHvCRq{=U{fadRQxWMH`zbod_Yw)6zHC;b<slQ@EB7mDMn@bvap!U<t
zmP}@a+AYx;tZx9H_1eMsElK$xGswE&l-bzWjQ{+}l#-HSDZ@<{iMZ1Yi$1GiJblU_
z5Ktg6fCoI1koiY;%dD-PLw0~R77(Ie>a0XxG2QVuV+?UqWK8q7-LC$GD%_5j08x?%
za|gf8U&jOF@4Rp<{M%Fvc45-^kLfoa5AZKS4WEBo%A>tV7C}Pg4=hKQ!KC(oddYse
z<n+U$PDlm=ky5|o2{StM9I-YSVcNdN^V%YI$YfVD3>{UfKuzqEP}YVc*+dE#4s3Xw
zYy_`J0azA}&fh`g&Nmg$zEdrXB9K(S$Pl}__LN8#T(wF^t*KE(ujM4;j_2;JJ6utG
z-TJxy%TxPe%|~{vM8`9*0ku>mG(OFY^sqr<FD#9$k7f9oU|V`O^eAfQzIc#)M~^0o
z<yX#DxQ%Rl)bqcD9-i57i?Uc1G3#Gh+sZD}FjCK;EM%66v2B6=cyWZ48)dt)Nk1NO
z(qo?M%{7SoQ*2lFyFOUllYYYnwRBmB|G0Zp?`K8+`ULmF3%r+nKb0-27<w`~V&_Vf
zNg3C4wA)_#zTh=W5Z3R%Jk2Y^y%}$R^BGRv7>dvka<_c-hI`+t4<?9ZQ9ZyU@xP`@
zB5geu;F`-Kc3HPz7b=Kutz5lhDJxvUkCIooFQQq^di*bA3yP6HF%1i({sl!kc|KkH
z_(S$-cM2ld;!ioY;#0?S=-iPp)DDDJdNBHm$ICl2rRW~NTz>>@tAps^q{v}9PU+|?
zzbQk<BO=3hzgG)6#dTdY#k-V6JoX;|l-xlEfR4~h`ZjrXJpX=Po5M0$3pe)bfnC;O
zh4SA!0TDOtecPeqYii%3dUlaUF@*~5_$Asm)#pcj@*Y$^+9ur#=YhL#IL)KIEK^kr
zl3HQPom&(SzG>WufJtxPFL~({jBpLI>EWT*F5_;S@ZL9&Ca4{4;Z5%rhr2CglG)Y-
z-%OFSn|yXp^F7$gZ0o{n=D0FFt#m;J77YR2FkfNefb04AzlGxhVg&BOa1lOz#HaEU
zg${ITsRhsew7$NxWxkuyZs7JgBp|#(NMPi`sbX^qauXFmIPfFjm={=MH*39a@`5N+
za;w-JKqiotlTp$yDMesi0|6y)zaVOWEb!}DDG|LNIb>Us@PtNrE;F2jy;E44jZpM^
z4NSaEQdFBLX5y&;4PEF&u}lXQgFA~3i%|19`O+C8N@mQV*r2+A6_wFZbTM#gVwz%>
zl?5Aa>0RoYdf|P0^1?fY9;Kq!Ci5S>ORif)J$sG!@Z^_ARm*+}fmEq9H20uWn84an
zT%OjLu9Kiji~CekpfPU;?}lE>VVlfHVu{qzBtM~4rXg)d>s@qVrG{i?Us*$YS|BAC
zW47qUWA#MuRFk6D_3h%bw)5A|DEIn}YH?!+>ZVrhKdR$LsMK?(oGQF(r*N|J%pLJ|
zIW=p%yh3>6WzxW&2xaO;T+d>jkMt>d4E2^2n^gkAQaFYp?S3UY;<k`fzhhd2(_Bm!
z=e)=hm#YgL5jsWEt#(llwMVv|B)?lM&)+!Fw{^TeleL&Xxj}v3M=W_I=d&_;xZ?y*
zblS5aK6%w#iYxaW6&-DAqE3Q(jPX_jt>ecJeYC~XOh$;+?@H8t@yHaaF5ZcgGL6wl
zPPR*kw+Qf*nWVPPh;-fRqhLa!L-nUakgRbEi`43mQ+p>7X!^5Rf7ZTj<}I?~$kROO
z?=*+RQPF*tODBieax#;S5N#+w#M!RC*yUI@WlNe^O#5q*#5hma3JkIzGv>4K(zPx-
z3gdgH;0Wuo-Jcn`@v+t4%ft7=IGh{rs^=~YoY)5mO<Iv=vtIF#pOi(i+O*0unB>Zt
z?=vmn9OZqEoDG7>Os+>@qJhav*4^So=#+`i`#oOS$dXMF7zZJ+phMt72eYp{ul@0Z
z0BTBT@p*f%T<W*GvBbCk?#61T#jBm4KdJZIEB+|$f7%zWoXHo;%FYObTR`1k$P-f3
znwxJ%5;6%j{mLNy1%cT=Gd(!zfp7vP1DJZG2KV3}z_gjv=?u>~sRP%uuI%-&{h~eo
z+ex<mGAU6YqWy9lV3z8iz7T?bJ|-d_Qp)UUy%jf-zyQy2TTIk>?jsR-<#2Rk^Xjmq
z<ukf3E>hM`V^!c{SN%<oy`aD_-GmdU=5xHTIcRBXTK^&W(|LHKj<|HDu1Ztfzc_lD
ztKIY2Q}0p_>--Agk$X=-On{j=?^`l`27$F5ys08raWb(W>rR<4weQX7D8}-4xF9Jk
z63wS29_t@*^(v&kbJ&*Dtl4n*<d165#IKeY9AGPdi$Fa&^G@2wHgxpewNZwxk&cW@
zq4x=PyBgPMN{>~F<rZP&i&?h7HWcKME9iLF@wU6C_zRV^j3)^+LDmgU7!SlF3yf^8
zVAi?^*KFb|M_dxNu_amEqrS%WSU4gc)!}`6ETO-?6i;xpOj~WkSoOHsE8gFb=H>9y
z=hBHKaAM3TG-v3Cc=7w=>0gtRpMiR(LpRFVcSJeiN`SX-W}->nb<d!W-!62d=Pew&
z%8ceS!z-d%(CXe4CmpyzhlgkhMQAF3!t6n`8@S4@AY4S!DN-rk+LJA6Vz)05Zb5`Y
zgO)i<mt{Rq(@>;B`QexkjFKWTiCgMUeFOST5}aCV{WqqHFn_o90<y%cP>(FXIS5-j
z6_S_ItWL9+tr0{Al{82x717$HSvI>*nhu7;mH&*vnCH`Gx1WwI-XD4Ovs@}I_IcgT
zq%??m8FzoR?jPQ-*<a4-_vz?6zP9}H=S>_oeD<fOnqpOBxNk6QH7R%K^~(leK_S3K
z3;^O^3{jMhf}AxxZB;bL$@Z#$FDZPr8eql#IkIRBZlVmwLsq8g$(r(iwI(pslR6hU
zzF+#Zv&u+jq4u|8muk)s$w=n!brv5>FFSJ=R`YlT7{wqi`=@IU{nbE=K5?QoRFlV>
zgsYr8o=HrCEEKeCO6RpdV_-0X!0;kC_3*RdjRC(BRvY=2e0(X(EkjyodQ_~aL7Uhn
znvTto*UI-qazBF^i#fJKn+ZjGnM;iBz^{B_8pae&Vn2pP<+g}r*=hvL=Sx?~c!epj
zeQIcL|03h<rHa(C#`&Kq-hJhblU28-^E1`AQPB_nlKR}{{%#=^B=EAz^56-p^Qo(l
zW0K!$$@BBTLcV9?OnT_^@<GJcJ0@2gM?0?k=IPLgS;2b><{V3wdr%}AE|i*ID&8?i
z5Zrv|{fe2&hMIS_xi>2Pl-np*D{tpyx(rY1C12+dm+MNxcpwq1>W6H;eJZp?F^kVT
zyHANYwvc2h4~`mB^vu*cAx-<0PcI3V@2(y8`1MZ4;R_G$N7Ghr`IBb7^te{DG8Qfr
z!m;lGB6*5-@NjV#CgN~c{8b@#_pX~e3ULmbyi`Ih*fuFU`XfdvP7fIe>u2bgEVN4F
z8#rZq`OsF8-A+T?S}i6RjG^J(pzvXtIIml`er{Yk(!Zbns3+K4QGqRjU7byVJ2HZk
zO}Ae&NCr_1Z%ZC3nK<xlVB%bKQ;&V%iiT^c7QQ|H>PezFA}8Ll18t&p{IY2G=~Dy)
z;VSFW5nQ6$33ja;vtGMp>omROY$r;b#rMoY^A%S?5qim##ONe`HdA^NcAXAal=FXv
zcZ?oRYgo0apzmqf=*X*ux98i{ZNkD<*!Ky6K(LOUSOcy#c;%K!?o{^Nqx}Z2iOknL
z*Wu<@yRsc+lIwdE7rPhmeHcClty3qe=Z-ICQ`w7bOoO@$NFx}o@PNl1j_#*<f0(JW
z(?O=7dI9Us0rdfQ<%hm2Q?2*NMj<kQNG^N@v$|(02O0|Ox39?6MUc-n?VWc1)+##g
zznjXM@SlFKtRc#W)mzs8lr?F_@cDq}7ZfW@PEJz5!ZJVvYC^f_pGpmcYVg&-R{%$R
z-M$lW7f>a`4h5kn+qH>3o%%!il;^LvHJplUFFA&5)7cDvEHWK!j&o5p>*<JKkv8Zg
zsRA(ZWcQ`F5T(>gOuQef@uS~Gl&IWJ9xDqj+|i5sQr?P)>(t{T)$0}1DE?M&BS0lZ
zswl9^+SzlKc@sdOw%Fi)^Ha8^o2yt|$^0T?pNHhK<zec{bL?_7L2q62_G236mhy@q
z4AR8kFUX)BDYVV%sC(>Inicgmd^tS7M-bxyIcwXA9^cF^?7X|!!Lr-Z=|29VR-O4z
zVS-l-VSUqz)`4HQpt4=uf_L>bIOQ|Sm5t{U=PUxwR~0fF5BSD0<T#AZBLisX)IV4^
zD^EL76PY`wYKI$~&iA~@k|?{8w%Q#cVU$8|sIp&h=(lj1&N<U$6#W6W;v!2LQkIOC
z!*Gs-Fs`6*gHa$`yW7V1lz5siL;TFau6$@k<}+}1;_pzhx+i?M*1@{0_Ibr?*W*(U
z+nZxbucb*EyPWiLSpSyT-g~Q>#yD@_DC?mX1PzNg!kL4~ob1+Exuqcdn3^8WdIYz<
z@mulMo_1U~#rvd)ocMm~`<A_EV5ubOj&VPbQ9~t06P~v}qj*rp6Fw5r4!~rfG0WM2
zD&?BpPr3AXzjMhBahXZdY<Bsok7YzjYyDaW;`aK-i~`9z%HU$0UF7`L5O5MHf|H1t
zz5j9W4^MT-2F0VjA(YR#C3n?K+5(<c#=qoi%GVDiJ{i+2I}0CQ9|NC8N;bbkZvUBL
zNWCU^QYLpkuL^45mG*nx=(=;{@@u$L!UG2$kP~W2Bdsl9hd?3|0rX|P>&|+k-Jc`7
z(<*hL1qy#)a8DG~{bstc*xp*Cw5Y?*P9~K4uHt<YwnE9F5byvvf3bk(5RUa0+T$U(
z=`R)}tJ;Cmo}BYX7Q|n_9tX=eTyx>FyKaAn*_gZEtkBODjY&#q*G8ORV&6pWwlNR0
zHouD`h{ZY~#c8Dy7K>%RJaj~-88~N$WZ|ZDnB385Or6H)_6{A&TDv9heR~&iRtg<7
zR(=6QG4mNM`Ieen<2aY|+zd)uWm}2M+c3g6J^Vi$O>R_8T<B1?$6h~>>T$=Bx|Ff0
zByt$!kRJY|Z^?7zw<*SOf<~TkpSM6bzu#1La_A+|p9JCXB=PHL#!oYi?Ln-2hm%nQ
zJa?Yq<j*wZXj*C+TR&P4__)|Q7x2KRl#HdUzHLkMGY_b6EfZ=N__#3gW5njkU9F|t
z3l0>`oR0Rx7H==}lgcO+59OwnUNBUvVbvK$l`V+$z7uIZ+0gLuXqr}{XNO`;bi%V`
zI`9C9MLv4wua3JoGqRlMZHhg5&+h&dZu)`D=)=k*UkFN;pi%*NkHvtArzj@qb8R`r
z?opg%Px%S!PCFGMao6z3R(zx-mROIDLnD#A6b|n6F-H{2MV2*b8zb>nR1&6$^a%MU
zaj_1MCGTISo#i}aT$5<Z2lOf<<Z7R?a*<+_V-Sod%VX@w8GddSRMNBu^SeYuNQ!tV
z2ecJ!5)b5~eP`qm{i6AUjGn9^NNl(L-;6zjtSBH&v94irjlX|^u6K3wXRGF|6E%Ov
zGO)ASKZI}WxB+SUX!4MyyXr$Xg4)jt)1|DsAwP!S`oLITERaHx{%yGci9oS{ac=)Z
zl@;P*0<U8S^=rt-0hxfh&=mQisG)H_Ar*KcIB4|VZ5})Co}KzWbz&#w4_jTp{{G#j
zKTG@Fa#K6Ov75SoHuWdJS)3|<ZX+wLvP#0e6d}RPcX>_UyPBT}jaB03q#lxmWAG;f
z$^x>n|1wTMits~;mkFGC$fn2Vd67pxDww|);b7~jZ)bljb15-JTjWa;Fr>newG(|=
z#C}>NZsxg~qD4Q)xtTu29%^DfBDoEd<~EDY!w;RWUNC8F9}Wd;QT?krP^Q=jU^l8A
zM)T>2?wt$4<I|suUwG{K15nqBWtSfoSbbG#xgznw^9<_i`NV>&j{dbEjIlZH)&C~!
z#!*$4<>0LfQk7Q40Vt2kku(r2w5C>?XI(r!jQ@PR>!sRcPUnx+dD&j{=)w6Jk}FcW
z3_Bh9O8z%y{w=>ws`6VUMz_%s1vM_z0Au*^XW2Wio%8S7%PUR+<w{*n2<sV8!i<cx
zVi@aQ>2Vidlq$J&cE?Fv^G(U|0@L5eZf8F-pF{XmGnMEl(5ZjI+d%=J?)_(q+LZ?x
z>G&9IXqxMypNKf5juB41?>J*ijO}Z5RW$Vpr<?ofj(F*>HN57H(5tj^9?Ef1S3j7m
zBwLex?~Y%{@UKBiRNnsUI9>5A?(@H%-s9HI{Stpto?G9()bww#w!SY?qCXyz;Zr&K
zsOpc9^NA6u5%Nb`T$qm2US=rf6Vs%=jJ0QEJ8Ic=>gav@9En|^NC&@|*TeMveVPT~
zU!_2vzKwL}dZLm{9<L}$@#wC9sqz17KcW@;{oLYQhTzN12Sc~{^+q2ToNkdcC97{2
z2meThZ~Wf32(7*meeZX|a@}ZH_1v4qg8x6buh3+%_?vAQX`MDTMl=0+2Q1U8lf-8*
zHpe{472M=>GT3uTByi2)XM@02K^IxC{VMR>HuE0fd@z<lzXDI+)$A|wtSl_W|NNO3
z@*YOS-mW*tjy?ey%vuY=`K8@<&yuR)*6jzaXP=)v!^&jH0|AydVYHupIsGLmv;1Vd
z0w5j#Hf%DXL*_kUQ77B<i{!~0J(a8;)k1EGW8BI2?OzZQ(u30?a61&f>e&%7^H@g2
z=>#1c?ku+bGH)}9!t8=~>3;usbo8iDqQ;UPq}Zp}mF-qdIE4+1Lkhe<k`|SL>x{bM
zv2Y1RB+h-i=wSyzDUYtyiV}$c>eWFmrCQj-4QKx3@_v%nuo8xOY-EZbr~B2?x%DXX
zhOK`WN57kA-kNgn#<wpy^zr45jcqb1<JqoT8OErF3+8Eue!b53qOhBC?Rm{s@HIgp
zm+QJ)B%xcYs8TRP$0w&CUfcJ}1<x8TqOs2tupnn<Tx>CESX;k%gQ>XUKvQ`=d?b#i
zEqD!GQC#k?oOhP7S7dvq(ivMnSv^VH$n}fycKswB8fyI{>YYgCOpM<78s0Rw#|Wh0
zM09IX*vy$A%+X_}_1y;(yVh-(fix+UiS<7~Bw{=tG~ur0FFcu4#3ZgNHMXaPu2FMf
zFT^X-F@e)lpQ^;o+d%{m_M=~-0!gz?^7jzVTl32wm4YaitEl7;Xes6(X5u%QqA_vY
zh;C#T#?eS_+P9_Mz<rw6gup7|y+~diexD()4rj=Rck4UjBWsFZB6vr7NmFw<LCZqq
z3tdvp)se9?<jnzBHHfqGo4=~2-x)38W#DqZT#7H(f9?O6%6hfUz@1hW%Ubao8)L4-
zU2-M_9Muf$+z5$f=iQUPj^xGb;bQFU(56~xLr!G?v7k>-&@IrHR|l^%Kx=9zi38ds
z$B6!qSH{l1JO|K-$RR-1NJ$z3EZ}<g1}K@V5El@T@U8?-|3aRJs(-HZkJ;Mw`Avhm
zKf5iXK>^`6gicq%i?0Y=8nF5P!@1!{_~*X?td$WmnL@8Vj;Egk%>Zi|Io%HwiDdB>
zB)Gy>3<Xf&+XGWx%cTSZ`_oK`-bzJ{?Ilq-j<Xwki12a8O0Gm4Z%a?R_PG;r<4#CP
z18k5@Ea7=b)=XQ4C>q26$Frfmw5uk9!066Mi<&J9gNB;oKU`p_7?q+yS&!W=^VggB
z?k|Y<9ewMJuH1Yhxe+|&G8%h3K8^WROS(rocs66JJ&4}g<rIy>&RQHJpB{QADdAt_
zeMxde_kEPL>yv!|{-*jmn<x`PN!M1IV8kl&YUOQXkz5*w!k_m<QGK+p^L!NzFhviK
z?LIxlt=T)1eLEi?W8#=T%f{UC{;7IReWD|`jPkv@m)(A0oa0hcFKZ1*&qGQtiX-h0
zQPUb&b6*y@L?Fnvl<p<MDmrN)Pseq8Rh1fE2oo2r{1;QaFDA2<siWfY-@hZ3KlbpP
zPbf9*)6{0yE-1HhiDy<zy1e5wIGmheVUz2Na~b6^Or&}&!|SZio+_}w5s!}WL5T96
zM2IR-u@kGQxKh={ios08F3v{mk7k!e+Hf>R`xql`BV5vvmWo)T7T1VGZjp<@#LyPk
zTbn#Hlx(NWFuQY)yrFa$G-qVw&eP`pvlR{q9u^6}V%4NNAqlmUEm84K&X`8v-zgk|
z`b{cND0hEzHfd$RsY2-%`ckZW{R(Q^py!tV&036xh1QB1P%K~9-1{ZRQh`se_^UDT
zlKLCd0Aksq!KkGjAW~q;T0CCKvnV=emHX}pq&UHK4uYIso&zVm_rLHla<uh-Y`VHS
z6ed9g5B7cY=efnX|3u<m&kM+(2-DiVqUGF;KWznX0w0mDg?XR3UVMQT?Q3OF))D$m
zrRqAcrqR5PK*(gxLO}=|>VHt(B>eW|p4Ht=g)H@;NLYly5GvmSUj<UDK~$9}Bly#h
zGDc_aaTg|1<lL_k0P6ncuTJbBlWXOYUBMP+%mT^oY>|YxQ|bXjQ|bXzQygctnQEuC
zOQ0Q_yZ@S$%8X6~!Nb1vnpY=G;wc%IL1Z9d$B5gztm3(xbgf*++b))e7wauFdO(u+
zrKBw0YGY7mrE|7%+R4VVyOD;5hvLcVJh$M9yo}<@{0Ab_83h~_)ZMMX&EHimTyU*_
z&B{JBwSjGvhPHM7zEz=4k@5@74394{mbqW|RM!U^aaNrEHB80Y9hv&#8jp;;JeKW|
z7CGjT84MS{wp)LIA!R@JhXE>-SW)4VN4M5#j*UG`=*0Xgk8>T;#CXK1ZY{*A&hWN}
zDKVPoPR-dXOUhe~&c<nFs|izX>$s`X$?S-a5io;a;-B;5*7A6h3OQrWg!r&9xLg>2
zLv{PvP<CB*2#_vEs0uJBw_iSAZuIM`++DqfuKfNj$l@kp%26D}(?~2qQ^wI(K*sp7
z+g~1FkOEAiupBdrYE2-RH1QY(83c)_e#D;QHjzNj!=4z59RE0#%67FQF6XJNH5_iX
zt?~QTim|;XSCB{@pkOHCX>Q(_MTLLbLjr~9=S6}|dGx{VNjrTxdTAqhtH5(uHOXUf
zuNlI=lGEDT=XyHBo*eibkiexUw{tX?`KnrM$6R_gWhdg7S6~Z>sOUgMHL(+7FqBuO
zoG~7=&d(z}K6VUw_=MVx2v%-_Nv@wXU>IZl|BTH&MtHpg)TJEkq2F2!gu|lA6iXxP
zi6&a%P&^>UpckL>%N#fds#i#iD!4!nm--h@l7#Pu7Ya!K$aRt-ZZ)<l6nfavqEU*Z
z^^Mx1z_`ZXoeQHiUXXlEtA9BA%bXwhr~kVApCi;7z58FPjybl|rhwoCcz9k%aAJ_p
zHd$I#@imyyG)ui9Fl;c8EFOW)9C1jV@4k*tn;09K{;==*AyTk_u55=knJ{OX*r~cx
z%6fwCS>VJgfvYSXqW3aW;X6qA5iwbP_HmuyZEFoIxGPbsl|tyW191;i?m!^VY?UM$
zBb;o~le?1~|3-){U+x(5eF4@yx8jnTb>CA1CDSPs9*hw^7LkO{AFpDDcf6?c0ep^e
z!v$Av3%t2I#bB}MFBtF9>Aqu2!If~(vZC8hcv6$UBgbLr3!z!Pt5YoX9;ArB74sPe
z&-2+ln>3FF7p-&M>z`>RN9+gkJ4<Oqa-WPHm_p5`)?Xnn9w(d@qg<J~3!8_V;{BO@
z6pP<bZ<cKs%)etdE?bB1bB%C&`FfC!xuk@(%rSh$nVnl<H_fRO4QkHLUSyM3aQL2j
z{DZ|OLP_?H&l~0RWNcSrPdoct8KgsDM{J+s?Ru9&`|0SAQbQq8R%e5!zD?-Dp<HwM
zQWgnwCe@5_r~*&hvV6FDP>KPSp%w7?>ixG{%}AO<G<bFt66GtePZiz78;f{Ai9jMj
zVo31|>~QAl{Z#CayW(V<)H4u%FTY|gQKSY(OKSNdd|56=O4t)AqPg(E`?Aup6?QvF
zN5{<wJ|<rVzmbo)bzi4xSo*!%UzG@obJ%wP9|Tc6j>x`K%5o%E+d1&W-@`tK?Ffs2
zvwn|@lJwQM>b%FPgV>5ZvN!~m4dHRGKZ|u#2zA$u<y-q_sAPqwd|p=!Q+ZiZ=<jo9
zW!3<N{_XwYHI>Yd1s#GLffoO7Vx?7kMk#pU-~rkXy4xX}9ke3Cp}znsV*%bVr+;ol
z<t)+DV%uaB1~IG?uvfqXOEnB-vZ@1OhGCxo^!i`ePbJD|&^h;AE{MqFCI)-H)qPf$
zi=-yYM^5au{#h$14|&vtv0#mdZ5n+2uuzlhY&0K9zwOtMnDpa^B?J_{WSwu>#i4KC
zTFL5iGE7LmreZJ+kQhi4r9EA4{7ao1+4uyP(;wkfq7Eg&Hu(Zc&qHv$m?K6!B!u>3
z?G9*+jPJ~|6t!6FR*k1<;X*vsF;JE-7g4KL&d;FbFS7o+cp-yz#uE^`&o%eFOX-cV
zN);U9Wo2ay`gXJ;%7qng%B0G}PUwmbI{9{euACk`MPgsL(X2i7Ify$7(*Bw%9KATO
z<&<BBT@OXpdZjuR^*q12<0_)6aqWWH5U*UF$>;`m@#YWZ2;OQX)s37KH~U8{^cgk;
z_zrst+2mPVqV66UF=?#7mY$J$>vetc6FI}M+;lAmv0v%k&0Hs#S4Xtm1jjEaCCmwM
z$^=$<bn<qhgJgRKY!bVh8FH^VI*jnRd!%kZ$alOd)gKcie7r2v4=aBDWY(Qih*Y;<
zC5J2hRiNg8djA3OUW@)dLyVt2&6%}%ft5`z67@Kf;qPgAh;DQiiMfrrmw8{C>%^C2
zURGxP3#{+%OUdzD_~2U5CTydc2DjU2P@|Jh$rm4Aa|^-f-D20%Ql#Ytrly1r>q19I
z*N`e)DbDyz?<qCjnM#83x6}J0yg7%prWy595r@Z}o=VcC#P?Ix$Fp~%9iiXg3<tu`
zFy2RQoI`)+%s}6pA-(VxBw0_$szqMRO>RAs#%~;XME&pz;s@dt+fHuu!?X#A-iE*Y
za;Iyn$>j#g`)f&?H=-&%m*g=XT1zIqu){FjSZek{AU3WYrIxt?m&+3aUq`)TEC!v6
zn_jQ-To2%`=@BmEEKvKo6sx)VId`-6D|-U*&TQEWAZCW${|RQvnQYLHI=V2qK0jzF
zoAN53GP%M?mF*m4{=kZI^|%?}&;NXzWCaOi0zxVaG%*3Z0n?kD)RxlO5i|@-&!d`M
zn(&njyRU|$e|8&xD}P)JV|OgMLoUw)e-RLr<QL7bJ2XhaK=1>-AS>hr{0rEE&mhDM
zl3xYzKjl-PLB;4@1POw&C4zazi*ZIX>NpJ+u~bV{k=D*Ak>HyO*#cDB$Bz;pLgF?w
z1qHR~g%D8<yoe@Cilh!-(^^Id6bN&8sgZRg*)r)7NT-Ux6MC}><PFL~;E*2dY_vHs
z_u-uo(LL7M;$C3MN>G>tg11q9V<UX{gy~Ac6Nh5ofV6^cDmD*^j9Y~7z5yj=Ci9y5
z!=*6_r_>f&JEzufkt81<Gneh=eBg&u(yD3M<Tm?^xRt(QIGaWj+s{_iVtcUj`zo<3
z(N~B_q`J3y*_QvRV8!L0-kBdVGyX5oxmoNVj;wATT#;9m&3){~>Fv>R<h~<|I&((?
ze{X=;8F5*+PL0puH10glB_g4J{m5YE9mn4e59&+(jwCz3jee-t=<?gi1H-oWD&E}=
z(R7)LKZ;I0-rnizmKpJ08z%Cx^H;9f18OLp1>aj!Dqg~R>dHZ*k6(Oj#mZ_h?d^{o
z810l6%DhM!&&eAhIxRHS#8&S}k50oMY4%=GUr&#eNA?{7ec)F)q_JfiDh{gHa_Ogz
zQ|9vF^(6QGgXqQYFdt}afBapLYCElN?r&;atY{$hw>jk_Hwr}-UJwJ!3l9aD%j&#E
z<q9SVzJON%oaQUjwZ#K{0rxJbsy+RAH^I6;fc0TN_xGq~sz0m3yWIP;CpM2FwDQh1
zptf5#_xIT~`_x)sdFSu>P{_*7*)6$69lknA$}t~#`$hCd9a&7yQ@gFV+x2@t4a9$5
z=VVj*+?D?sk;%vhtThg*nPj>IE}hzJx+-zRauWy*`X~DUZ8sc0ILqXY0}#VPiTlRJ
z2DvT)(l3D@2lgDIhI&^SD9>X@pWWD4;RkM}|L)8$;NUfDn`d=K^I=!|%erJX%`O5~
z4v|JPs~DHpX==)up638S$uI$R3R!r&xaa`dDA?u6%5+euY)3_G_DDQ*xZ;Ez@%!m6
z5G{J0_YfT>Ql8DbbTJwRmIhr?29Ay9*Pckg1(rs-i`ao99L0CGt-Si^5p8HaRWycA
zS}w)peS7g6;nu;KC0?RU(@hn|NN(D|b!P|;WHq)uF#5wOuTq?UevA#$3LXj09A(Rs
z@2dW>DZ}t{WTm9ZYU2GRCMlP+4MNEsP~-*p1?f$=dIbHfCKjZ`T^<W?(3~2#VZ)Y3
zR4K=PQ<vu})~N}?R!WaWNT%s1VxLmZZ1bs1x!>qty})x-tUkT_2rwmF*e2`-ZBEI+
zN^zCyd_T>#!Rge@o>7_uMSeErKc#b~f9^yHwk0Ev?dX;#iCU~G_ppR;tI|FWKdR%*
z7tyZu(=wm3Lsxww$2c3q(`na$AJUKJl~4XvvY18O{mj&-Ir{`(dcK-hVuskdrAwFR
zeN~)VB5zNNGMyyE);O%-UU09%HQwVQEqAh`gXMi<w~xN;m#nQRX-RPkeS1ZAM^RcI
z2N|kyg10HRcPY+P9C`SdGNnB2)zi>O<2s$HPVL=}zNGk`ifGE)M-|N{W~>c|UXp6(
zuuUp8BT1+D%y!=?#JCLs6i`C1!0PNbdMHlLaV{rcr-Kv0Jgw8Q-T8N6cbzABwsXT0
zlXO(gXINtCLjO_msp&T-2_IThG<U^o*CteuahK+%YQ#dTg=>Nhh9oVD3N#2i%P&TB
zU)waJ-UNKRwy`|gf9(8X<|@x-=G;~p$wwGinw^(&VVgq#tF5{FFM^2N6bY_kvUBZD
zGX*R+zz?=KDGvMH#!9l^$2ABr(fm*Xa0TcT09Rmzq1HOB(gpTo2Q=Nv;#_Rqz7fB#
z*}(c@WmHm;6f^~&7!)Oui&7!-5K89Z`K8{gAu#Dz@NuF5jy!BbkmwC)_kYj|jBg6F
z#ov9Nfo3n`>U^@raBb9~U5(dOKTS-FL_;5f7;4(AO47g#$BWz4M!-Az(2h3SQJxQw
zS(u#A)dV7Ys-&iK7fWx%Xy!J2IyB{&QE``~x(xMUG|_hg%5B3l6gR@639rLOppJ>5
zj@$j6rR~Tj5b7Q~H-j>nL+_?oX^*{^Z5tLiR^o{O%**lo)_F+nb`AN-CE_upd9x+o
zK126SGFTzb@K971h^U?u*s3IV&umPR+SjjDHct#vp?D&LIg7nZ3;7HO99G<G=hW}t
zz8rPkzkDcNI4xq-oj*qXEQUYA;=yC@Yd4?w&V`dA+u?{Sx6D=ddr+MEFgyg2GJ(pK
z4;MRWMAXg<pciL09W>`ZxKms%aO6ZbD9YXLawG9dd-L9B^r1MHAmUPKtKD7VrXb2J
z<4b*1I?zQI@B2w;_A--;fj{O*XQzN3X6p`r5~9En8sj?@<e5D}zB|^+dRO-Zo4fXJ
zyt@{X=J@`ce(069ZrML5I{7qC9mDvW(<RRlgkg^Oqqi>8f1m*Sr!iVw+U1ssHdlB5
zahw@=g3Q@WM7~~;$)>y3BH*Hh*;G5RJSafsY^JlxT{aatWTp4r+FzJ4lOXzK=cA&e
zi)$Z|&h(O__RfK*4)O_il_u3bzD>@EJs+^_f6w}RD|wa=bcRkX4Oil0m$Op;JX<(J
zCBv9Ys=1iw&8YX~o}FxNJ?{3mw5+K=d;4A#2>s7FmWxS!VJCi)j>cGre{#%QgV7pp
z{~JUmZ~Gsp_;>ydKmHf}3MU+_^HAV5zN6{#dibKs6{08fhLKZ-Mm@;(hJTwmEGW~{
zy4o%<*+B&k;yw~?1(!2g7-(SkZkNsL#Esi_YBz55zeFIujy+Icd>TgwAC|lNI)Zio
zgn;l;fJiwyHdV}}SYF2;4cS2IKqA*a=`0u7`UV%_zJ+yi%3|JtSpkEgm3AWlUf;3p
z?B@ii(#3FWBGX(E?WZi9kg^7=+HoQ{2NoqXG^S3Z=i9Oq-J|HtS5>G{o$D%$D9Z9W
zm+6m(9v7TaL83CQfx{sL?4LeUM5pS*<U)h)g@T^ubFBd~`2iohN9FF|E4Nx*mgkh_
z3n$WT+$&RgKK!b}`<CYygg%X@k+4QypTZp@RtH;)J@X%T%HV&+-gVypa<tugFrU^C
z#i5Z7XJSl{@;BYAv=w!08jj+wjn334WDk^*@mX5BHGU7zucXS5N*9QCCcZgVcJ;((
zRA&Imb#S(8h<E$)uy}A7X*c7gTaQl9D1Oloqj5DrxGnkh0uMwgm9rksbD3~DpUf7~
zVB>M@>}sho70rCT!t&4_gd~f%CeGI%yEhUy<;(CHEivnK2r0TcRFkH4tIBX1)$l#r
zw^WMq?A}1@uF1r7^|n*`#nmcf&!EO|f@Dz^&*fKisn^+g*>wzUr?Y;RAef9C+_6H|
zh~u=5cj&NL5tsGf8~ZRA{&rHPe8v}Nq8+NQ%VCoh5-ll#6o(p1BnU7wt;rF`<`>A@
zwg>wI=t59@Ni*>nX7+)F^LiXM$+?}Zzdg(0=MefYbc87sJX;Vf@K+Qo{pwgP<N+J^
z7@tC8l2u*MG6(tZ6MOIeFd%-t818&4WOONIr-U-;bue9QO;zA~A^jU3!2#gNp9fY2
zgnwtXR)DJk+3MZSP9unhCkt6eU?{VxOlb6)t4!v7mIARVd3r;G^%sIEgv0y!k%>eo
z6SH;IvT7ll*1lAP3&P2*nmc}60a$#8{!`>+<)*CJm$5pqa=_=~VgWZ`DyN@6pg5LE
z{Cdt6p9hQ)_%}xv9F89(tFp<KpRhH$RupPTaPcWbx|}V5i27$XvaPZm9Qp%zhlxmm
zsP<Z6-Olb0C4T1zE~PnoTh(`{PmkLqwZjUxladasP4bjux~yWdp2O@zmC9z-zsHry
ztrCg0iamC>gmLCU9k%6kWP3~SARYV?r<<XJ(S)K-rD5egaePsh9XmSc68hlb-~A3#
zk9bB8Ql1*>w)|Sf+IM>>Z>Npdd?9ygh<K~uaUp%z=lOhmdcT=xmRspL3JGPcr>^!X
z4^Cy{vfcTbym6XI*N~nbj#t2n9M;i7b5{E9a{t!z38$&+2g}`DGV0Y^a>5fQs?{dm
zx>MvFW%)c1o<AgeC`U3O=-YX1^Aclu-$S-J%%`W5159VKk+s?HI9m}4hu!lV+E+%7
zEX!Xz7GSFX$<|O{{4m0ip5Hes;*n%zqO&bW>cE66hMOL|{8EIk;XkuvCY(k$+RbeS
zs!gBkU)N)@7HghtE3NeGHV$Jj-@S}wFs1<mju_4~29XpdFB4aSLHZs_j|7J@rlhH;
z#_l9VN3^6^I(sf(+|2m#E3j=`#LLt6QHa`^F&&ZCO2dz@-dHslGo>fnYOKg)`Y883
z+ZDzT7Nm#;dt=)HC!}*&<>ITqyY=*&1GuYAsw0bi7gf}JPJNFF)v5@2K`ggzo^Z+t
zzY?6~qmhjBUS6MnBCzri8Kq_4q{Yg6E_?=2sgg`(`s%Y3oFVk=-?TS!szVF;wq{Xe
zJE8V%iv@Dx*<u!L8^p;`J}~$oDHgEsqCo`#7N37z24r2SN*C+{)Z&Bl{y>SF<SzM9
zo1BE0dr<K8EuDZSv-^`?M4;JK<TUGrJ(fJ=AH9<VTpfNd*_4SOavUgp1z^|5o#95{
z$-+g5u0N&yY6ZzdFCi=-V1MHeB9@kxcOl_}yV4n13p1%8`!32^F3KitHjFsqAn=WU
z<B{t>va->LCTrvCS;q}}IzIZR87N$*LUq37ut`Yq{Y7{Cf-GKo-1ub%&eWC%YkPin
zd)z5RLlQ`w*&so0TuH#dql$*2L%9wmo}5_O2L@H{i*LETwUsx-`e%a4H4AUG>OLzv
zwQ5`nIBGUje6Pc<%}P-1gqK^*_e)-={y}0%M|W>=fyqvbD#+<CbNidnJow=15cJCS
z%-}Ry*i+VYIFxgN5BVcsTF$k(VBAj?H9TxwIJ11Z+>!Q8zg7YJ(=ud>{=MOIi^-bY
zHwlt_L&eMsn3vlzy9Im2Yf7JAD6mQw7gA#zmzBROQH~$PyHJn&3hqBF>^L=UQG>P^
z_Q4{SUkDxg?I@&lq~1Z7?U)6dqD2Wh<*}mVZ)Zb^gHMkhpv<;+q!x4oUyUnX(X_F>
z51r47;>WL7dpO5mmLPNr&n}(FJ4xT@B#5P9cZ9~sdMB34Nxb?!?V7yY0;!b8!8s#u
z3|X`G4DYFoqmnQtaq>q#B(cvR=tcM}OGSw6Z1Pw}@3RT*XF@df`XBwtIPqg2K@NHA
zxKHMBtfRko9-Pc&?TE3UFCpNR)<0^K*r4u+OhJxNXq+p)76yf+mz3UpFa5CY&@rlU
z4r^Vc5=-#F*{^iZi3)D$PUdnx=cQwHqjLCvEnu@pE*-$|AU?V)H^P++RB%1Ly%0?G
z>&Iv5&Dux!nYD~&aCyRtTk8&p8>FNFa|6{JkY@N_S>8XK!yz!x!sF4|mv5A9NN<An
zCIiE1$=k!Xj2bZP7l-v$hwRrzB@J>p8Y7><8X|STSez4(Jtr;bUYZAn22v&TAeEW*
z?O&Q`lb1F&mKOqr-l{d(zrB;W!BTyeUPNS?%8U+6A<k`&BTG$x5`WUZKQ;V&@2aMm
z^FwPKm9VebG?!3~k%XQs;RD0ZeEn_Cxtwt)_K+IKWzG8bxd3p6C9LLnmlo_p+SM(e
zI&JiDxrh7IHOjjtVe+I@_o;1Y3d5H~x(<4q$d=n!mp}ccS3Y>!@M4~v=v3Dgn+Ol5
zuHN(;0coyZY3PmW0|W=BU*P$~k0MZ`_9KTyUZ%9qWO#NmZKO{O`Vsvy*KnMO0-bb7
zJyb?pg?i1-dUYsi3|xQ=6*SYR&65Sp6}2}Ewkj>J>+|PS=Qd3C6e4=&w)KZT7n?fQ
z|2p?|IXr93i*x*4*F&)hclXG!C*85^L>XaOPCV{KOBUzgq~QnNarJo7DXQ!7bXaZ$
ze=|N^x+A=T&3%HY&W}F%aUByf4HMFz{V~le9+1@)JUTSnB5m1B)sAA!w``iOj=Zmp
zh;3~;OnDCqexw(fkoBPs?`7_5`BG$XUS%kj)Z#fzd~kqh{9a=V+La){&_ro)nwrs-
zUc_G4i2FhG1Il4>*M4%1E_68`1?lKmIURgMXryw-2yT6O>^0=VyRQE`r>O$sB?;w$
zZwkeZRms+RhQtC(^L+*Z{kMHOqgZv{h;5p&?@Mw2gWi5D07fy}%&P;O3X~Q=;trfV
z2r3yKP76ND%?Q@2=sweOHLb0U>7c9;Sy*4mA)NPG|AT?yB$E*eh=vznfeDPX6bOdt
zeq)QjE>d`JAy7G!=0aG1Er|U7q!9jeFJ$9-4{P)Hu9BnIt{x{Vlt3H`fCNzOAbzIO
zE}k&XTr8bj+JLU=i-17$GR30R(SJ2}Fw*z-$gqW9mEbw4vbHU>*?3I!ohbSE@hRSu
z8)Al-IAT)RH=g>^CB{~gJQp4}og^c;)r5)0UwE5pST89svs5G_^6sD__Fk;($c2Eh
zp&(MX$a)pqtID2N>XX9)BG-+^r<=}ZSi2n#Gqym!Sl;b!Y|mNnyX3v{qudui;hy_*
z-Ck(t5W-dHc&#i1yM-fjNr!gyO6GQ}HkOZ7yyKQ|A38w_-t%3NGAla}gFV`hB>DYP
z=N*|2wfUqp{K=r6DfEZP41`y?UKnJVyJ+vI^YhEf#gG<tk1%Sh9k=uflX<pls@$fS
z#u@GM7MrtS^6O&^$+MWILotTF5pGUZ$y5~<C%Xg8EPF-AYHd-CBK#Z3)E1@Bw46!l
zG4WGXbGVpb%#KI!w-2H>s3kgARhTXHsrX}UTJH5MJze2&O5TyxBc0SfflAJEorIV|
zA|uW}TmnLtJFL`NB!ar03yXjF&H3<MdU?EW0HagOyI=m?jmxew$-`KSXoA%jr?rWL
zjqOZSEocRAc|$a2S|M;csxMN@2;bcxF9uF_C<7C*QEZV#a%d@f)5l1P8OME(^hrlT
z0PT!@*A=3il_8R`lIs5XjO<>{&Dk3t|3c1KJ@Ou5{rUfR7`92TCg-(9ZG4L>SqQI!
z=t6#arm7T2xU1NdQU|jdyj2j6!pfI@(GI%81~Z2FPX>&t8G{j?1zYPD95;IEE0^J+
z4ulwfs8N6u@s^HzlhJPPQ20#9pDs`@!8Oi{m$NVn3Kaf@LtWq7E`#jtZ>?bG0<a9^
zrBSn<XV!x+VNmdvAOF|fLAH*6{VD{T?f=~4>*GgryXR}>_{hvk-Adn*hw1^YEtjC*
zyL;AcQ~MAN{y!lI?9Mh1QG`<zD?Y43=z_enmGPOG+z^ct(?=<ri^@e~G@2*wdw*Tp
z*nV;08r^%u<+cOo1a$E}ZIw8EmiKL#;la!#e?u!7Rtn~`IYSy48Y8xdLkdh>+Fa5|
zXQ*@Qbv%EG7UkwMNQHkrA2A<ISlzgFzT{4Qy9@93HOP}pm7(e^q(^<ral<_wdrA_G
zu=SO7i+GO#0rh%C1}0#+nlS3<|21b1HD<TxWu)=DS~Gp*m`;h6#Q8tZg%PJj&e(b4
z2^LjzF7>nSLX{-uT_1|VGEy8|3S(jx90@X4XcIlWaWsbqTG0267(#2oMfI`OuP3M9
z!9U@?RN~F_*Jf97#MjJDzU|_PUxiIX+z)CgH9tA07-0C-td~pfz%k_Nwc=t4qsNJP
zs;b2@mnYu}B3V}xTSXM!ydBZGwo-l1cOt**`A%th7N_1qlLu<`*-+3h)!})$Vg3cJ
zbB-psOD-?oC|H+VGLEW#8hR*)UD;Of+wHRk!kG$0W8IH>X%PfkmZ$RIXCE(Sr4hC_
zqBvSFPg{b)^F<r2a6z?Q3e$BH-S<kENCMJW+i%XtSPM2_YQ`-HXy7Yrc$S1|D(oR0
ze_A7T=gwb0)-d_pniy(2PXm3<NRC-M^?O<p&rOwG9slFIsdPIhWnk-HWMZy`RO?Pg
znFqR^0-MbrEC@;E7NkTGmYxcYls{0ReoIO(xZXv1{kj<EypeEi?Eh>l{lG~;rUB6x
zRQzO6VJD2-YX{z1Xh=FM+flouX0{kJ9b`)*0tPN1K*(8`fExp60gYa-v`Bxo%rT&~
zLnLJ`ChzC|zQIXe8IG{35Kk>H5dgss^$kvF?f%;R+_(40rMc&XJs{SBc7!V+ToIXj
z;M3qtbzL<|GIzZR=#gNkC*_&i+xsq>;;0_JOZTisAtp)`GLu2Rgr?j7Xi=B^7(x1e
z2^6KF`IQoRf!}zXcZYm1(-hvxU~Q?QYaT=@GovUc)vgPY9?CJG=8YEYuWCB|gpD_A
zlmRun{#=4ctmyfK8r>WbxK$?A;$O;oQ*=~~1vT1WTO;fIqmY?_KY?Fue4i#r8jt25
z{E+%wZdg?r9dzJLR>I|Ao_?P(DzsaKDMTpPsKhAv7f_<CBYKa2mm2*PVb5<C@Au8b
zW4Pg)6NVAbRGRRT8QT+CtSUjICXjHPmTpmP>-y;4jm~LrVow7V_Z97*p;~BWbVdt0
z^Q1$Fr7Xc|MIKR_R?alLA0sO+^5u}n;Y;_Y!cOj1>4mPJ*$d`wp7p-=M~tH`PIN5k
z*|s{8!=Nse{fHpPJ)4($mylAe&Qs-2mv84#4+*@ofnKp5HRhw8>>nIXw_$|tOk0v9
z-_^}<oBn(~{4hsv1AkMtru~W)-&5(f`1<;WIDje*?M2n85RA7ZT@35nA5&AA9(RT=
zF(K`?$42^Z?FEX`ywB7gKI7RD=OSXvmn%6H8X<G<u<0^soDh)xv3Gy^UM1rlI2>3v
za+q30;s0T3ThW|%Y(JV{7M{vr4>g_A+intx>wvbnzX~S)!N%sCf&bMkpvN>Y=q=eF
zqMp(e9I}VY6?KO*>RC;Cd+f$f2>qFW>ZL13yl_TCIm7M_SFX@P-vRi*$*Bjh^!oVt
zL=3yf%=uMK#4i=a(LrJdM0fn_V*+>#{y4IrlYDyrbEZJTD{LGR51X7bPd_i^t|q<F
z*d8IpZhD2U{&_bcQ&_$POD~wqy+jV3<c)!fCRiIFLJD|5c*ue&)|)aZohG?I-jE(F
zKHLr|xS(_6Lo3Yje>FL`$r;Fkf=>pQ3|K2ZX;I1WXhzC^8lOgJw>1P+cm1Y3BhKd8
z@$r<ER%b1Z0xcc-o=s{aubKGS?3>%QrJ-xt6VtkxkyB^h^!P)LH`>CQ@UDsJ*O(@e
zm|{_NQd3LXv4)RGjK~z0Ale+4%Wx~LtoiI+%52=zSt3EKX35U*nEK(3dlFzGU*x>I
z);zhi;y>~xKJ9a#>VBx{>?MKvp7-s`vkG_Ao_6!_p(N$}f2M6;<-eED=)*1Eh3MiQ
z|4fBKjoVUwBHqpH7OSXE<GefZ>41HyjZ0LFzjxqyEJNR?dMES3a?2c?LzeZArhcm{
z^XOPefgK3syqeJyalIVTOZl>Y9poU;+-Az<YM0!%8+vykt)*%0!-XpP(=@kG{i9#s
zB%mgy;|V?#4hap8LKrc|HuMj>|BtJ;fQovJ+P-Io?v#*58l;688c{*P5kU#*6v?4M
zq!mz7X&AylP#UDW8CsDXx}`%z((fM5^Stl7zO!br^c>d6aow~3`@Z(Ie;2-F{uT?g
zz31EjFYb1ALb)UqebdtwkSMa-dTI4WH|{9@IHBgK`Y~b1k>%s4rwVQZqS`+_6rPf$
zG87W=v-HE5BAQjAh@)tNUByxR-0)Nh$}otzQYrA5OPDod3}B9gH-CR1p|KTl)3ALh
z2sZ)IjS-?GDFF9<tq;)Cf&dMihl0)@!WDdTqG3>;35PBI^#oUyqt^h%zF*z!FZxXK
zDRrDH^8$W0A(X#2%hJG3?0hNv&G^HKwzm4!oy!?uo12;F|FfL3{h2lU#n^A{DmqN$
z1wm0*FP2?~c#dmrA{9&`G=KD$k8N;~tBEj>f&10}5Q@POU%wKT?)9P}@IUkh;68!F
z5Bdw-Fb@paz_!)3ZY_u{Fn09Y>Q2WEo1}n(3>pWIHfCf0CeSnZpK(zATf4EcU(${s
zvLXjnA2cZ7(*v4r@WSaIZZ++0H7|Eg(~%?3dU-iyETe+sQLBIy4j-TG`VNxSV;|tG
zs*CY-o7<O?gis=MFcg>kJfFmhPkWD{m=hM=7MJfCypkufZYSFNUQgLgdj*|79pYqe
zVSfl%XEH-$pPho^hYuHODUqfG{8A}Z!_QIsCmSv><mf7}TG$pCu!BcC;nnm>!^Fj}
zuA)yaqB+g*&;@O6uOO>8=9V6NMlLzQ;^}Na_^cyJjojT201gfvYwmW$G}!*VF^_VD
zM9~csQjRX!+lNFKQhy~e=vS%0)C#3tj~RW!9#mkQr5}QM_~>3=PEQ)P<nV5Xcx7T#
z@@9IHGIIb)JIp5++g;0=+lh(G_cfL_n1q?<&d=R6FNZSSZyrhYA7I7GHI~HQ|6ocQ
zBokcVCjUn5`nK@uTPulg=Q3}&Kocck0P9Tfl8kA#EK=*}xr=zii`5W`Y+i}$LVR_i
zY3oAw`q~9a+PDwRJtYzErICa+@)>px)w1beo9Q3;g+3Eyo~0htL#sNGh~H`FgxHi&
z!?Y=zsj0`obfe{RFWepHj+VCh6h*}k)9$2)J<5cn1Pv2Jd=_~QO;L|4_b@kAVd6P~
zgbe<>PsEFBC8l43x!t!*FF%QjIN;JV;ahL2ZpJ@`c=D*)__~?9>55+Svb+={+-eyR
z(w1yaM|RV9yyzR;u#?*sZnbo7xV3Dx86-?wbbW(nw%VuLaiOl*)KTu0A1_HSxXk@N
zJ;(o~kltTnAcFGs{e@PXRiU-j$Opy?AjvAQTLB<2oE1ze@NpyIO9*QxSWykyuMJ8Z
z{&P8w`&)5vB>VFueMf#Lvu1Es7UV*BgLPLzJm9^@ArCS#=5cF)rVT;?y3@Z}Kp;Bn
zUugF9w!qO65HJV?cA~r1JZgcZR5csHLu7$}RoxlGrJpUi+PY^%lrp>%kUiLCF2$+z
ztO&;68&D`k2u2XaLdAG_PYS1a$H@4x8w~DFH<BO7x@CZvF7xMRbJh)FowZ5u+Y4__
z!jw#)NEx+H$G%Ig+ro3!!b~{$<g3Z+uw(APddK10fYUl=eO#1<x8%ff+Gsg2QL;xG
z9gpVYUJS6A=GPH^7t<j4SE^jg{ZmD7K7;xAQeJvAKve3n54BiW_iSjp@Wnu%+{bIg
zc=ZQ)$mBAKBC8&@2dCybKKE>v8@%Cr9X=6P=WCSv-l*+cW@x2Z%KXFY{Gp2yu62uL
zYwBLJ-<-@KAY*po4r*MQJ%8O!_@L5Vu8QXcvga2(=E_dpk7I-I#n<W$Z$4i!&@x`a
z)J)ENsJBxK=^^EgnWWuy|CafbYZdNNw9JmzV~qD{aWVN3)}aLxW(BWFE3lE;kg{s3
zZE4@Lc5W77pnyi!cE5CxjCh6rk|srj6@-hLJGL;Bh^Ia~1KiaTTrQ(A1QVfR{ec<&
z!8=Wy7I^f~A{p_A`n$(X$5sQFW?U&H#2MtGH$W-fa>#ZgSooxpY22M`%%{36ho1+4
ze7YdCapSxQWH#acQKft<@O6v*OJR?5{yW!NKirA@sWzjBUqB$xb;zEV?)G!~KXzBF
z?U%BN7u#>apcu>uaO}K|Ly+IK2lTMMRd?_r5ENq|+v-dXU3+s6_TXf29(1ALgyCRe
zV7CuP(*wJB|H)&3at+)S-vRF0PMQ_{g6j96X#k$`F-VBcUf#fZ&!lHAULu1%og_Hk
z0canXfdFFb_8~Z?t=^3XY%hL)SO2^AF&3!4(|g+q!j}QMSEKpB5B9Y}2pIDLR<9(`
z2h5f^K(kBO#x=P`c`ru?pVb}kS#x&O<I1Sa+%+iGV#Jr`(ASn|pRCL&5WWp1wxkT9
zhElcMP(k4FM_Tv$YZTT36VTPXkx_%yP1I`boK5L-4AUPG5aTou>1~8^qggvtYQ_BV
z@1_L0?^DA7qIv6$Z!I?vZ1N=3ERd%gd-@_S^F!!<r^kLDq8$x4o}%753Fl&_hz^tz
zod*9@eWWp{fG@7p@a4;pXFFWY>yLIyiCJvZ=_IfJoZ-Cavr;KJR6>wcS*6;!XFL(p
zu84RrQu-~m)+>?Sn=V3_k)_20FCw#h7GlYAl;eT?)k|+ZQ@nZp<U7Zw6E;CSN|wso
zwMi5*GQpV6#$Ppp5JZnoaMOt#1!v4C#y4&ottCVRWyXs@zZ;QYgAQ0G2EDA&lYLaw
z&$AhLqZ{E{aPY}pl30O<S_|byAAapH2x5b-cJGg&SWMU75ki$EBEU4~jskvMkfZ_e
zO$~N3C_YWfW}6k6r9m7@GZX4a7>1DOix9VIrq`<F9exSby(0|;V+z|Ss{wdcd7LeT
zA<ez)<-6tLDlyev<$v)5ikI!cXWSWfdP_C`vfy7q$o2<ESYkdkxT=-XNwu}DKyg|E
zdPZfN>{DK)-zS+QHfZ|~*%Fz=sNwUt^UjjtKkM&Za`Rz5t|rZzfk`t@b52eG3@O0;
zuaEA;l)US}75Er9dcchaygi_^1xuvAJl-t3y^f3|``Z}HNz7e!yKS&W&Za#D^#EW_
zIk|z-1xCZT;Vsy@20Gg73^71p2kvFyYXgIa;xXBb#^D=lIO`hd4t^&YHKe*kcn{8W
z1xjD2$<m&3foceYgm8j{(L2Rsx>USC;RRgOmIM5VWWo<<G?SHpzL4G4X7=!}{~UEz
zf%UQR!|F`2M-ohw-+y@xIaW3+P-!ZIV36H&NH1s54iQZI55Q?~&ZqUM&BrjhL51m(
zhg(e$JkWd%J42+OLo?=NJv}6T?6p4~%1tu1UguXCB~s@4>>aqF_s>S)ejw~MB|;nH
zi<mGX2X-D*zW83LpIG$(dt&*l9D*j-2=ck&h<;-+F8Y~?7RgY|?H^^ylP=YY`FwWB
z?Hg3DKveq2g#b;?xp(y8O28|^J6A2ojpOgf^zsGWm~9v*53%*>y5);U?c}O;|6fCW
zBglbdNq8L}Z4;jU_9`RGTA!0IH$pr0xdMyPAQu%<6qfOHjGy7Fr;NB^tW|#c079xT
zHgy@TNHLgJY52ip@o&mOvhd>j(bycHe;PrJ=zNE_k}d-^ZajL_5@XXz*VJvlzwe$l
zh#kDLiH^7}u~aaXP{BGZMsiP`pruBklb5N`39ljc;jsenT5#uCe7`j2+t;AV8s9)6
zbRXG*F$)>4v_%*$2?pwG00u1aB&d{EMG9sM5)8=cAiCT*f`%i^obZDt1f(;mRP#Sz
z{*HD{3oWtcx=xbt>Rm^1#px|#75cNy2j|tNcq6!4%z$fa3x{wb*&Wc<23VHdYW(O%
zA@Ru=FgEaDUIF4q-!7H08RKnDtINDmIjJA+;~tX8Bb7e!<!}9N*t^{;jb*0z@_-Ft
z4K1BjbdTHjygO3@kQBh|OPYqoV*lGT!P#pDSIwj5_xEWnnB-7J*-UgX$-t!s?2+Ir
zFeCt<F}SSZ8e0tJwcA7X*e<}3A+-}FwLv%)2mIY=`UWLhPRBk4X?*jX%d*r~S~X);
zPp;$6>V)H0B05k(VLC}}Sv6D-`2K>)Fs@k!y)H03Y<Up@?`3%SC94qtvnMksul6%2
z6!QdGn^p6ZX-qumI}PA8PHbD7|0wQ-iC3l?_?UIRLQ~zhTt+N&7t-Tf;$na3iPr;c
zl?Oeud}wrrX;bD~WtoHrvI}!db$I;PxhL-RBM3%9Col^h#@cxvd&b6|d2O9Pq2Mo`
zx<k<E_&BTlJTYj0y^If@$}1n2-elTNt{CY}a=?6=JRC+DUld~IN$n0|m^{us91Rr*
z8^)zGR<z1(^qTXZO!J^K@F<M-lY6zy%LI|nGj0&(bX_+xJ8Y4w1cx>+8IPYaxYTKw
zPaN?*tg`R7xaU>3flrw;W<RW`+ENVJC-vqWg(~`dawdCYULykil}xoBHVIZxSQ@~?
zp5NYnCTl}LD_ouOPFrGZ#gpHLMAo6%l~otkO}!eU!fHemeGUF(+jOr{82h6$L3-7?
zw=_OQ48s3ZS(W@YMc1?tZ*0>~zZW0)+CmMaQI^d7gxnvf$w5Lc1j+<jgu6fb;w&fu
zy{j1?E&AF576!-{op!JgND_%|=P%lq%!|@bxvo?1^TG*OAHHbeB$yP&e@M)EHj#DW
zy7~wg&gh|u3m}lWd-hU|7TD>99zcJ-#m6a^nI~}V%UGFbIa7=sxS{Z5%;pDS3Ce2M
z=huLhVOXZuDc9h0I@Ly3Q1#s4wqRZ#tqZfFe4wktmnrYN5qgc`77<|0dD)F*0D;4B
z0RscZIt2W7_6S&TSjCfRt-QL1fjQo}DnGjnmlh7RThPve{ZyQZ9pERxlN{hG=(CC8
z;(MYS{~Md(u(~sAnJ@d#e|Nj^&uReXK;dvV7(dYgJpV_8aBRf{Ug8{%f4*4~$Rka+
zQ?kF2f<Dm_R29!D8l0&<erE@AFz5<^O%u0+HT@kIv6t4*(6dZ>b6}o`3oJcNCE)Ef
z8lT+NK#*6MX~dNfKP!<eLq6x>qWwW9dqm{bT)1XntY5>ue2_g`-=w;A_(rWHzu8tw
z`q`i)peA+<M>>L}T=UMBre|>9krxl)$3j~(_-M53pN&?G4CB(~?{(HOu=dy12?{l)
z{!N8oQ&qqJP|uU;0t#P)7*BV@gH_5%d1ptln!fr0eEyBB*xQ{NqwzcWb7R1iV?<6O
zXKSUtQMlb<+bz|fYTKLJJVOp=9V>4c{9Nt3>iL`lmu%jES(Y10HOM4tTv<|2a!tOU
zQt4kSCc}i4al(s#ET7sq(WSoh?wJzwYPz{zh(%x2xkhJ6j5{Acown1gav4?u7WY~b
z8_T^54-4ScX6-DIq{*(~cW2YPgOBG<B6Ywd6SJ_V-T7u?kJwUgH9=vy733+B$f{pU
zh9*4bSh`?T8-4BM!WXT7Rn0nl{DneJmw!(84?xT`L`&$}kl=Px2`LE|AVkz55-kK0
zd|KqpD?F*3Va&w3;>IDv_}QA`#^Q6dFG--fmYcNAXqx*>xg>xt+Wd5>loxI$iBz&S
z!Z60pSyK`TTwcj)aaXb%Z>AEKF*n)s6x7ItGz})ye%w#ZS-GsUXj(Tj>f065*DCH$
z(5T^$AP^?E!Tcf84J*U|rP*}#FRefQoeE$iNjTcu-(#0R3gzm5)3TXA)Z*|HG;?Ud
zC?SMXTXuQ%0D(Kqpkn}T981YmfxHPQ08dez{!Rf_GMwbQvttD=Xwp~bfzch<8o>Y#
z>>>fR7ndfUI}B>xbK~5)oUGH)VjMKlrL+T!eGsttzYE)i6~1OX${e8g1yjC?C;j!8
zz)?~VoY5-okp-pFlFRN(AzFF;EY=s?&7$wnWjHf}61;d0P>u^+$<yD7fizb*N@;&s
zi+V_T6Ut4biZ7&&+ShT{1O7*V<szoBk9kg;9Pyq`r<s6G`u3TBa(V3YFON4i2Y>CT
zVSNheWQ)%p&s+OVFO;lTHe<>Rz`m-`SsZ}1h}#?3vRzXPbgR;Z!Y2;)E4L1tF?t%+
zsN?j|_$}1Vn?{DK{F>x=XyZ)-v7|sRMocT_7c2NH>yeOjEa$fRRQ3MB2Be@LBTXW?
zG547r%c|Cj<||?7b{DI>pN{<jj3(vIgv2txpIxa(>tzn3CYi?)h~i+T*8%G5EGkZH
zLAA&I_(sd{v#O16j^fgT-{R9`D-xv|1(OVUT|x)$`?gaz_>=z-cMniJ5X;Z$`p!f)
zd2EQ!ANl94o7m{UFM4<%-2o{Q32FPJVto$10Rdhdf;;Mv(hHUJH|js4nMrbOINkj8
z#NB9-d;=qKze>tB`Ey#FuTi08lm3+Uvv|7)!?t=F#bM3u5$YUc3*XXcQi2kood^MZ
zHDd7?01CyY`W9hY-mU0+@1<ab8EqRukS2sNYVn4I7DS6l+$M@|P8v_#3jNvKfV#vQ
zYE~BaQnw89_9GjFH6jmHie|;_IxTse;_7nc-C99SbsG`^u%L72w`pil{Bk~h15XOt
zOYc8z0XEmj^u_@=DD%sP<GF~0<O_vQDO7W}60yj;uOol0q|@lfGEWm*95p}i?caIT
zw0kB9cI;k$e16)g^E~3~BcU~~I!<nU)ojjXoKpw%4`97I9)o{y46x~fAh=>V<lZf?
zp8=qAp!a=PgaOAHH?sh42Xk>15c3R1(BNy$rl|6a=l|o?Y06=cJ;3|}+udj3ZCxzS
zPDM3|UVw89N;J5rf#Cz7MS~ak1^72$h6i6=KWO)#QDOxtw!*%<;CTD4tA5DvVrd8f
zk{7#GeN$}19@DflRkST#;=L=TU!pk4TS8TWyFp)E9E`D7BBsXYf>Y6iDZt!_%|-Cc
zUrY7>^erHe@j#U|Y|0&uzKXx^S%iliu=XaJTk>4F?>j?}-jon~lKdrM*i;EEV%l+%
zOw{L0hc@PUdLUuLUibJ=OTuz!jN0mzt_QnzMzjO7a$A`P_h<C7l0|?23A0UiOG$s9
zcoB2ZeYLNtwanT>)NQc3b#uU4Om$nSbS!EC9*5Vgx_dc1@C-855WSfEFI2DHG(5E`
zvZ!{}5$ng@2EHKA`kqtG(jj7+&pbHf(09=wWuyI6UFNvn!HqVre<=?3(#e<qpkgyL
z%-?>*x3O_+F}%xznz_d$<;R?ar3w@0OJkarY+{C_z3Mm2(5{C@v`R^+Qv7gU`u-a}
zsnUN9Y6KjtitCirpx>WoX&ll?dOOY0EqW8+R#P8!x*;CNZ>7aPh6<>QOImSS<Y}4D
zwKLIZz?q~NsN?Uv>f^uDS`x=i86i&a9Z<CdLf(=TGv|`2H2097(W3dZszn}<7>Ivh
zhNY=7ahy;`y_LTGQ~8#_I>C9Y^5NoktTyO5b15*>yr@YTOm6azI6$Oxh?tJ=o$LjL
z*)Z@DHRM)%8Z=yJe~B8B;!#`aKq48IXIZ@_m1?%EM?qt#!-~Hn8rq}(!fu({U~UUU
zn$I3D2Chu%K3@az7QlYf0)$o%xehz&){U`Z&|8!ki_V`aft~}TG0JAF>wN*5*@!(3
zCkOnB0D=JsE`VX%g283cfhn<_^iANUxWov6p`3%NF%p|=b~dmB{-_6ge&zaAKNnX1
zB!erRcijFB+~O1PK7zLc_cZ{7!ESvWH0xw)71x@Y@~&;N6!Fp)P1M{~*gMCqdkZCh
zf94HVT}!HDJMCPe=C?(|+DtrZ6@Ri&B;2#1=0v|m=!2=Cn3+=2mt#(HnlP?bVx1Vl
z3ZKA8==Z~U>xp9SCqzimt5w8kk<ZZ;75|)E`#m?ejK$CeV>8m|0@-8g`K8CmdKv!N
z{ASD^d=*k8Tv_9Ke7zOZ5oE2KrC*iGTD5RIk?hb*UW6G1)^p7`Pu>Kli1ZoPK7^*q
zgwANS^4l6mRk$In1f17>f0HrmADms*n~qAiP2!j9rahCR#vC0Hd>QGwHAZt!3v;xT
z-|hTS;+cHQ@a@-frEA4f@#;5GASYMlvp&1;dXR0)hLov%28pLTUO#o1^p}lq;tDqw
zpQ7$vSYcbWn{BL|`pu1oH(LdnKCOCvz$6kgXU0aWXT{di{QhN#?^#^hHd>_1)LC}<
z2Wz4(vNsM@D#Fq+%=r|uH!HP1ey3d^QclQ>LN5u+izmQ8Yo1mfJud`Us85gb>)SYC
z39p&pAlPLQpOl22Rt>TTcv;7q>txhMu$18MTB=>`d{Jw(*@BDJkX+3>9vHZ&${B44
zW7-!+i{QWd6TI21xGP-B>9t{;Jzeusib!}-dp9#rgMXO>KO1n$m%n-FBGr2SbMQQG
zcGI{tOaj(n0cPkc3{1jM`_u7Qq}SeGy_RT=JB_+2lA84<cL9H2@$9eP&fLF0OLSxJ
z-o?jxK%J$pg4q{Hl&h|;ZniAdwt=tu9Pjg9=9yE+=`uNsZ}bBMC)gtbTH11y0(-3M
zf6Eu3!Qfx1n|6kEyFnJFN5ukGy?3+Ddjowm3!N_9FD8y7@8_b#+L?5zWdE<)o0Zje
z<S*U{<m9>2i_f~AwFDBK_zA|MxO_R|BFJUV+=Qcrw{&aEJWVHP!m%t?)pg;{#j$tF
z35p-x)W9db2^DgW;e@>l<pjW5CeESeWgKchl}sB0(^yeWtSwpxhStr;HlwfWNSUH@
zqbn$MF(&}3^K_#y#P<FTHgx^PnMrW`ebla8iL;Dy@BWpd=G59(J&fawQC%99gDcCj
zM8NvCQCiHFrw4JZJr_Yc4kqs(7()41on-5JOQ4jamVBzVna(;L8p#zcMbE?yk&K@A
zN8M!-n}pDg`1%x_MWb&d{yEO}`D^^XPJ|r5i+8Uwy87hv<~JI>s^+FqnJeb@?80xz
zn5qgXwDd}cvAuC-l7YaGLtsQV<jAKlZ+Cq?#iTUahQi5xfmC#Kk){t-28+K^z`7&#
zo{~%**+(lnIvsCiXOQ%L)V;`3r}cGezEgIcxH}2gwUGC8PMxO~t@jOQ83e246>BS9
zuY0_{jpcB^$<mYZXBzFr6p_+etVTSgfKT0m(Q3s2RDOg7)Uq`@&SD_qGpEdQ^P}c&
zX(&mqb>}k}GCAsi1RpH|T}F@`+5E~BF>4vT_!+y*QH>i3BUzGxx2E#`OP)_683|7{
zY6UdKYiO^>H@n^S^PQ~2DZImYKUZAbXRiPwq|;+B;digulV*$18I>CU^}Ri!F+Xqr
zVTto#oG0I{^mi}@W1$WAe=Es*Sy=&M03CGWw#PeJZ}5Z$1_R)IxqME8nXP%$e}oC(
zM+e$3n5*FwV3AqbyAeGeR!I_{Iop7X7l>xy>=E$9-nj<G4Y;ltF!DjcH1<lqx)aIw
zJ{AxQ8(t+<u9lxqOpu+$yZm!F`*N;ahKrDn&``y3m4U!;sWCk8UkS5_z(F}5WjGrw
z$ah9g#cSeJGnur!kqsb)tQ`VV@7Db8F7fFp5z}VO(8kes$E{}J#1_Pu@moZzcpBBs
zzp4U3OPsf;iNy^YTgBb#q##ZJ&0rJw=J5E>9DoS<18b;nx1|dye{TvUva%S<07tke
zfQbg<V3Uu0W$8d@$o=8gWrO^t4^6j&?cFCt&}f134~mE{@>2YZla+TFkTS~-9$!W{
zAJZ+_dos%e6-|r-<FvEu*9rusX4R@=jHMh!$z<~q4}Hc~7a(D5rlg+6Z5qkV!ueUy
zXxC>*wTz@v&DFS>?R*uiM}SU#w#C9k_3JHX``Fx5XE!J0su}0Nn{ygp>@<o(yx5Um
z6iY<$5=8#oQo$#4zl9QCwKoCjgj_xeu?1bc<ORVU_E>7_U2U09l}(h&{257BRVz$-
zquKs2;=7zr!gFu`tMGXeIK23diEG2>Dr0_Q&If}k_4kmZKRQK&;Ce>QAN#7EnpTat
z^MxhxoCHU6NJ`K<5=TN>lq9hfM0a33ipE58h*u-1J8GDBF-2nFDQ6n@%1a(}{g9~=
z<loc4^{%PmV%$mfzO_9N44T=SJ<xZ^9$lVkFt@m>114nJqAmYXV-?{}gHGFZ*<fj+
zu*Wgly8B@d@(x1fxCrCKU)OH)C!Gk;Ayn|6Ksg*AZQA}xeYz^Pi`UjAOMPtdMj{OG
zE5SmUIQjRTBk%~lO!|3siAx4USwKvN%gf_I_H_4;=s#b>>A8on>`rSSVEs=BIoPEH
zng9?RaC;MACC#fY8epuzSKxLe5}1`bA3DhW+c1&O;%sxOG&~z{!3#M0)Zv#L{bA5T
z10@ZZ7kcg<<M0>2k_NgC_xFssr(B&(3SuV#g2@2l3Ebu-!_FY6_QUiqFt)RFQRwuk
zk8MFQW*`|yp0M>Dl>GL`=hBC`F|~9p<lterstNgpRSXAhC{$9J2Y$T?YA-W31&MCI
zYL}gca`bMV&L-qH41AS|kJLZv9i2&U#ek&6XkVcfCjfx5@uo)pNy#nLyx|j#+80up
z{Mkri+8>+VFUH(7E>R02gRmE)Ou~QCEzL{Ou~3`SKL%n?SyV(O#eWLp>+<XLELqR!
zJ*4_L=!Z2ZfCczY+KfT?H&iSoy=01%it%!+4T`JG+HW8RI`COcDUdQ9V+Z$y{Rsr9
zibk0pS+S7L89(GGiOmYpJM;uT#Nkq>tsE<xE``);e{5QDZqCsStWv5;7NvDDUEzRe
zxzo|l04+>nXPH~#_h{ozT_nc!wAN43^-cDSPx_u6i*S`Sgg%VsCo6QIXCx(mkKC8Z
z#OP(n>khTG4+WHxbgOgJHt<+3YZB^6R$lz;sx~CodacZ0y~ijw1=BxGks{%B4GKTH
z^a7q4`PGQ_iNF$=;YZXVT%5pC26EqmuB(u0XVO}M6D=;o8^+AU$JwVM4Q=D4NtfK!
zM5RH&@+KnU)7yES(D-CX&QyG$LN^BuXl(g&o!-^?&4C<)xU%g}B5d)V-qs}7G&Kyp
z^<@x3QjUumRF~i7ka(JHrVfmcMu`)8yo(;P%RtrStIF8Vai&ZW)1O7v$v~_^EninT
zjMY5_uzBQNjfu;L{@%es4lTGLftSJ{Aa1Ps_H4OJraHn8(vw;75K2JY1zsD{drDxo
z_CF38plAg&aX_Eqklc@i2KG*se*GWl6`|31CDnlejwiR_9YFCfR(T%}Bqk#7wQ2%e
z4@h&yX%rycmNFdN32_s)pLks0$H{t^U4);|phL2Jk9p+YeTDHKzbwUBKa6phgH8eb
zu_vL>CxvW}>>rD-STq`Q6W{y`&>mo0ZHbj@b&iWD*N2%3ag(ID+jxc=nRG5Vx%$}V
zO=taGICDu2vl#(WRBv$9#8x27dFjcVHKb3<Z`>I8iI!6`{ku_R)4v#oWLp5XP)~NQ
zbvd#Pli!Tx)-K6Rf7T>O$iR=)+ugo0XRT|<LXMuku?ZA{I;Mj%XJHAkL0N$1G*o$1
z&(G%QWK)hRV;0p4!C;Po`0GV8W?h$$n)>{uh$3bgQB-?5XkKFaaL3lyZ4&>24~k?-
zA?InCBmZkZVH)IHhdr@a*SA#44PfUq#s`Oe;ASTwJavDg`+4fQ!xc0zNDe@=hp8fY
z-4+zSKY%s2iI-}HbVm$q{0tzV<&?_l)z~$-wFHv8)jDY^g@YIEn5nxMYiQuFXz~gW
z3h@!lp`9mfu-a2Me31jV8A2~1FMf(!2H4<oTdb}8HA{>I5GLqg6|D~oU_QYu!)_d8
zLDC5jgl|#Gnb2x{)#mPOlHh(IEws_MD*D1K^xy<-Fs}ZHw^Sc^i8b3lhq=kciP0*&
zy=(~Ba9Jdk_x_jXE%&Ja5BQ)ATYGP`XO~e5Zm38N&s?fy+o@CGr<bbl*UcGD=I7<>
z$fKM8BbS*i*tWbO8<U+-x`u+x$T)ss35(Ck%xJ2fXJF4<>fZ#Ku3JrkO6~IoN`QP9
zPG%n{H5SDUivdoP;%R_8sF|dEjvz8nFCQE=oY*@*&(*nD`kW2y=qRv%{B8%K-d3Z%
z@T%p&na6M|^DJ)~39@Dhes3QGH_iOOn}|7IcChI4d>4=>%mI92c4u+9HyOMk$P52P
z$kS|173I5o&va?XEKFYrck=_Ct@sp3tYr$Frw(nQ24BxFs0%Y>dPamIn82z}oIAi7
z*IbMA9w12Rt3l;0q8DuF+tlxBOUCE9j*?Cmt{r?WyFITyd@?d;|GDx{pk5UTKNjof
z4mX4>{~3RM3|N4-9sk@xm3_$&;e0MTXU$^n5VcyE)h5#d?qW+I;l6V4E4ZY|)oWUZ
z#p}oDzIS?ej<&!o=`W=hBnM|*KYd4nlyMTHL#Kv$IjOkXreUqDCx<O3=ANW$x`Hh>
z0T;e2QYyc%%r8lcuvJ+`S`E&QhD#e~tRl6(EZ6d&(Vmb2XJN&F=kht&%6@m}?qNcH
zmQH18&}wjV(uSi<uFFH&iK3|Fs_Rmx-Cr-Ba(8$}aaNUr)LiW)A94zY&phY?&Vcq^
zu~iFNkfr3f^eyD=$3U)N-Ji6{yKXX#Ab0STLA9O+&fR7GE(Rl!E%GOl3CWOMPH5J<
zB{*-l=v8X0&)b{UUX;Wk45oT+1?e}{i!k^>@#?JRXaxpXA5kd4Xa<2)!BBBLacCR;
ztz<|YWG$pw6spCvN5Vv>EopbBou52~&Tb$=th?=p(Yx4`EAMsyr(=F;H}LEW$7~;g
z{bfxR?ge&8tUw67#+>@WZC{O>JwI;LkY-b6(=_xZ2Ylc*11TllxG9`zkADf{|KAGP
z<CH(Lu)lErnRYxW+Cvjn_ER1Je+F@E*p~C1#vBHYBpwa@FOTT8WH0SEU|kqMuD~)7
zPF~1BmKuo2Wb98$wh*4&mn7k#3Rn9tr3ql+f!P;l%>f}!K#*-PeULaYG4Zez{Ka;O
z!o0ww&6!ZU(de|@rTzjkJfvQ}#M(NRm<tUCp348)T`*4rFHnourN(}3V<?w>VLGC|
zyY;j^%iu-Yvn-ji`hLdF96r_I6UcE2cG)Pszk_BM*Y8$pe<39T7qWS9MI-*G&&g@S
z4(+*0f3ux{9(Ie0f#mDQJijsuu-1Ovh4td!=DAaW2*!7o<RYtJ?7++HR0ezB_r4*m
zbx(-J^=F2a^KWfyGM7cnhi6764aP6v<FC3>sN@X-=0-Mib6z`HMU`6kc31F_?hbJ~
znUnX*-)T#8X66$SOJA~_OLYB`WB22lSNvu&c6CN~Cy{wDAR?l#yE~$<!2K)ICgb3Y
zWh60?d9R&LEU^l?QhZS%44(*`?(~<A+3K{gxMi;5J?q(qaqcq#I9ie5_+E)BV*Yq3
zYbW=J;DYT^;8u1TZIMx#rAD)EtrD7h2zC;p+=icfLKl8qs5Cmwv|aoiZ&F{MvzB#a
zLr*QG8Hc!D3wlP9PzXf}gSDC~{U1q>WNP>}_H9bWTn0UlqFaY)f)Jxw+AOz~EZ`He
zW-(oq^5W%0HevH)-Q!=~{s|pEX~w+wr0bJPP&42D>hCaHndJ>Xa;sArhhmA+gq=D@
z9d6XmaM)ktm~9Ry&N{K#tLblTHRmiB?M5OIFUTnbjC_(Q;x_B98%4l-$sHZB)-rn*
zliIKRB->m~R;Cnd&U}Y|hW5<HGiZ!*u@1>W`1QNSv%=Tjs!aDhEZ*evY^X@2lRiFv
ztUo4pg}8wDo{KJ_Hlca{%Ns!i-{kkvOuBrTA<sxx$RLcI0mx=|m?-=`KqWT2^Yyh9
zOeb9ti}3AGxhF&vFSVxYKzJa$X?%9I?xb~=QN@sl_BG{0m|&z_uD8;~-NUsJ{Y_2D
z*mhroswJ*(k^65C46-*16mIPW^9Bd?aKF{?u;SrAxc_td`}!l>jC*%HQa&^%<@LVv
zwbu)v%bqc^^5zcA96x(l6+V%_<feG?Gk$FMyB|k8QP)#HyYrRE2;cz1f$aT`wk_l`
zEK3bIdnLuC%0$}iD?)TgS65fTD7(9??d0~|E!zn-0R1r+7YOdaF%0KlF}igvg;Q}X
zw1g7XxbHXk9bNh-7p<mh+=+miS7<zXEBwpgi-X?3C!;1Wsod|sI6S+5w%JCtB_AEi
zKw|-vzs*gNf3Qn060mUqTxizdQN{;5wJWF;f{twH@~t`nqt&H=2mkzrIVi=~s8Flh
zCiS~(w-ui6;epW3+X_DO1m`=w7?P87uDH0cx5MPAIvQilKSXZZo0k8`mW0+sX3qSe
zW(&TDzQdMn9HyCn?yw%HJHf|U@oQXW-a30iu|^R+W$dC4HaU1o_x_lXp;?u@f>3V6
z>rN(BkB?70se1$I;0PIKr%^3;_5md2(H9C4R$(K;2pW+{z2l_VWb;LVTc$HzeJx_L
z9oa)edjd;WgW_3Q%yxP-h><ZZ4NB6ieH9*?fa%PnVFb0+JpY<RwMrjej(C#0=1p95
z{bZTr>ws@SUuhmiTZ=^AHxx=HSV(BFbT~6v|5>1Q;fd#gB$TP1Qhdyp;2p(i{-;Mv
zbg3*&3PgN4^_l*;)=x9{{YR&(gWkH!+CYSQd*aGpB~lpldC3#SdHNdUT8oZ8+Roez
z>PFOWCSZEyPg}aNBQ`x&iETYOdzOy~;qdKw_ghXBgf|`Jo?ylDmGdO9?!^}oPDg@-
z!qkG)l9ffBFX^+dMBzy`>q3D4LYxnx%-@OMEG4ddZ4N<qKfa<?Td}Vt<>5_DaWMl6
z)h|!D-b)+W2uerw39$G0^}_@rZ*hO0zsG6wkl#kd<_PZ&Y%9!>YZ)pXr9dby8%q$x
z`|Y&MQ@1jvN72q86_w|$ehZvyd==zE`WNiSk<?AO_=q*s?cnPVlX(w*B2Vt^X6{-G
zDK0NGH~KHTnbc||`<yKWwijRgG#l3?tWq0&d_0gL501yf04d{t9*P#gnAvbN({`{r
z2%Z-}k&d%T^_rQO3mfu`Oa&+`+(i>OEP(;j0_@gbb}+d7HIBc{L~a_#uNY06+rp{#
zf~j6WsMr<2jQB1kqAsX>ZRMBj%rS$~QAV+o#Dc{?htrXxSy~NWvU~6F!Q%$J{>SA3
zm~UV{<002S_})Y*_RTj^qF&fNzI$;zFL}QIL;gPT2Wrf3BH#kWDKk->R#Sev)Ku(B
z?dPXdsG}K0+=r=)KbI-y4urQ`lHywN*d)R?qfUTDBeiCeb}iWbPq{rCdb4qp8@;(r
z#zEGehZqD+&U-IWGXAmPCe>=62Pn4!is1NSB4p9v^|Gdqpg2#t!1${i#AeJG`J8o&
zzc?~;Qy!I*AY-Cq{Y|Q3rq2JA2XlO+ME<duphg=*UpauY+wJhaT~4FAvR}K1vX9xm
z9OW8aWw;rMSM^=#@jl5a$GRGIQZ%e--gs*4>%ApH?2*-@_2ruO7HgJkKf_`5g^#X}
z2u9zv>%YC}wvoNA;N0dD4mlX=vY#bMwZgJHva-rHIex025#cn|`#SzK*rNP_JT0DR
ziDMHayDz!@08Uw89FRxbM}CLIL$81o&Rlt&9#a0=A|n9`#~Pm~*jv~PRI&Lqe)C!g
zT)eef|2fR)^T_Oj;0nXXlmb$862#;`%*E*>?N*oQUnT_8zq}H}D<x(z6%V!3Fno4o
zZK*)ybMsBs+g2&P7tme&W)Z0EkisBsfz^BS=Lx<L$c23H+uPi4*E;WlWBl}Rpi3DO
zYwFMYs&Sw!>>}<*sF%7aKD%ZDmkr~iZMDEW8d;mS8a#657))JZt)fmwk261}TRj?k
zEjXU)p``xSH(8Y$0%~-qKwW0PVjHH&H1G(UV&lMiM0ko^zSA=^RXXIHWCL^zM4toK
z?CH_Y;@KJNV5VYQ`O`u1tU=Jgmp&<s`}~Zg63dHPoyl{t@A?C#DAL|{mE`5yKx)+Y
z?+>@r&VFc$&Cj=m<4X1JLZsNje+WX1NpFG<qGYa(`a|28OZ~xhg@FCZqz}_k6?>h>
z>8Hr~JKdR7ehsFSLL2}S4lv6H$E+KNa^VlvlpkMTPl}-BY2XQe7tf=*P<?nrglo6R
z)vs!^gLP9yL|Z7)RqaOiBheTD6=jY7zCk4owRpee87AHT+bfAnYH$-z9g)o%u5r$0
zlQ$^lbwkCM&aOuAjuE_OTX)w+b+H#A>OzF1>~Nvruk&!^`~x;a?iUxs_=yf&rk$Bj
zDCG8@%TKmqL|JBTz-yj4ulKmXPVQJ`H)f?bV*=Ja&kH-Fj|BfT5*HOx%x$e15FHd%
za)iV$HeVMy?H$N<o)AJC6H#Cda+o6U>H{*lf^Lj5k~xP>PRXE+SJRBM%n$)36kS)U
z>Xj+=MsIc|zl=AjS{P4*QlD)a5`iW_Kxc5U?ez%pWB2+O2dsLbiL)#-7R3R1F=mZF
zx`vq2xJSQ?S3ZQayh<-)Ml?8~aYfnBpP)mg&w-J_R_0Lh)tS&|H;bo_+l2}B%$lIz
zSjw)qF=&KM=))COt=eYBOYuy`!0aeBFFbx@Mrxk)qBm{dT4yY%%n~N08XN~rkl-SC
zNZNW;par8z4FKUxDYOR!!p78Aq${iD#wvUGZ?c3b_>JNWS*Y9w4wgDYVLZ6<yA`L>
z^8wi#)Z@VEt@0-$AvU|o)sy_ZWrc~LKq2J9T<X^RXAuq*D7<SBPbp`2v+&DfTcnGL
zpnPyaA^)9AW96At2in+6$X@n=On%0|vtHvOT^DrI>nu9`!3+${14w&7vq0REy&4VN
z$gW;qVcs*oL=Uo*I)-x}>}^hzPd%Op>B%1<-8j5zm}JplswVi@?a2&7wIi3s#&5?X
z8^ffkw>)<Rj~GpTpkfWbX-!}3-QF_yeGvuLj1*q{rN+^|6&D?kcjwjW|1LxV``q<l
zXi@0CDAe)$;fB=g*?tmhLY>!WecFuAzx!t=M}b6v+c_Mw-a<sz{V2b!KHgXl0G$pn
znF46xSfbWXlV_X=01Fs2iNAqXKuyJm!<7&nfX^BBktQbQTj;-YA}Z9*1qdU|L<C93
z*<RHeMY+9^zJIuJmP~qc8zW1FjNvHWtl@*#<W-TuC!&Q>`|~ZBiy@#|hi5-(Dm;OY
zOig~<a2P|meakAGBrB-tu4|%Xi0Qh%)BAo-WorDy?kWd2#dOE2;1ton-Z;-@N1H$h
zGG^2P7KjJ85UFeuBQB9q=}(loR^qz3(*y4|{+&H=194^?L-#6+ZB2@he@rzg(cyt;
zyOBlGVzcv?Kp|;Lq~BX51IUOzv(izFFUoCM?yBtXvB6T?`UQ6bibp?5eiH@D*egZm
zYb(8Vfb|5&nlKqM|CPlT$nR_I>QtN{h^WcKv$Y;@UE1!v?XMT(Uhaytg$I=hx=bgM
zQ8N$B@{p4Wo%$N7)KH*VQ<m4+18by>9lj{C8>HVaWop+*Nfp3Tb+w#Z!gsJcTDn4~
z2f5B%Z{6A2j3LA0c@W{-6Gb9nrn2{v-Yjl`By`m{EdJ@-l#jS9>)Fo5#caPl-Idqi
zyhgtR=XIbLnn-6i<nv3xGN$JbR)6#^LT!(ZhQxU?gASC}+V}6`5Bb)&?L39b9W=v=
z%O$>7ynz{g5Y=3H<}PTPw}GD^2F1r><&3;_w>i;{ha*P>Ix~sV<7&VHbhSiIsWU^~
z8%(Eh?JqcDxCbrn0q?id5bize&*;*0E`uXYfS{(!d#Bse<-_OEBld~%p_AdBM&RHo
zh^rVh%??@_GOhuk8nh4qS%rHSvBtmLN`G6Z7IKnJThC9Aa7Yz^%g3Ez@9inf^!=Tw
z8W1n?h|J|iu*rwEAz8kig+0)hA0AXyQg7K7z3zHo+NHkJ5S85Z6g&^iZ{Mo<Sf^Mu
zQ|C=}v@_dfS=v1qy>a+__RAZ4>n8f)jl=I8zsbaFlvM)7n?db2hqsQO$%E9OH;W&w
zfIoX-bAWk!rY;uz2&`;bK%WUw)sQ-OpL~{fRRy7Dn{CtamAr_KK>47<y!Y!yo>{?p
zldBVS`ToD@0&hu*_B<$kF&#374G$KyM|eI;?Igld9B^%S+Bs%lBDy65l;+L~M7U@h
zre7QNsb400n|E$W#%j5%B2eyU3#QQAo{V3rDl8rwJY*XhnQH&_sr#2uw?i*a4s9xa
zeMmE;kHs_h#^43xmz%z|HaAgjENeeUDQ<oPr5Au^x4azd6G#VNrPonatrR%1IkK!b
z#jDwo3uD>Qtgxt!xh1*02^UR*m^#sJ)it=d*ZX7V9xg(@<aK?{wUQU(aQvjzaBA19
zR0&meZrO~4q_$P}8MVEFEb+*tn9zod%P?s|DFeYgh0pyH&7BN)Mz6Sar#3j*F1KKi
z<msB3%4&l0X3kNPXyYhd=#B6z9OZgrq18~h^UL^a8(|v3+0xls%={Xub`-5?s=^ck
z!yzqUP47W;j@_D1%bZ{^UtbEf59t7y2$E<1l1ovQ=Ra+F;|yJPsQ)WM`})=gQ4LO;
z@2YDVv4J{@&kqynSJcS*9^Auc=hF5i-m0GpuNK5rZ`C#1nT46eTE!<BJ=<yT-cnta
zeGKF0NVCoq<zMZh@fH(ObP`El)l*p(UU6f4zIfK}pNm?wv5-40KLi&sgn{?b_B5+b
zqx&i)K%$^PX<i-7T3lH8fG=#MJpo$pViVNF&5mbjgf4evmnXrE@W0H4-|p&~4TIYp
zv0c@EN20iTTz|IT>$BIv==!fQtYdq+E*#t&m$l10<GoDcZ_DbM{oQWrSmZBu;?hYT
zOw%8`UE2<Rw|Jm@_}V|$t)y&K=#-auV>BQj02deqK2E^v4_+d4bcLdCBI0-yf0L&u
zFp=L#P+<C?%!C`wn6vZe=2pAw%O7NAX3qcK)KbpMw0z+a)nZ4ye)yYlA<C+ZjQ?p5
zUIG;aIpgt9W`4;aLvDyE0@?%P*N?e2;~gBIUrc~(H1>NKBn#oMGrMOO63<e_21nrY
zbsZFTiacJ{4xCRc43D?sv?Ugg#%o`Ov@k(Y-;4_pve<w=iV9LgqDhbSB#TvIG|Se5
z?~#0COZG;~XCHUnNj53kFiJy4@++nnPV-Cm{(UWQuW4Z7F-~`2Dx*Zm%42R6ODK;u
z-w2|Gu!V-4SB%rjgvs&+7|N;_@vsCKLOO(sSLKB^Jm+C8LnsLYeXFpAR?*<p7Q@f;
z*7xOCRZtf>E;58<l-cMy{!{U98S*lV>B{oErO_@`H@<8xeRA|Bg{N*5WfP7*NY_-#
zDRrXa-!O4i-_*h0>cj6Ku_1pVRYGXjk<A;|h5<NlzBJXgPAO_z$mdVLrRZ$&wIF>d
zKBRewAcSF+btk$-w9AS)>o-~8KgMO5Gz}iw>6;{<MC5pKhT=-e1dRlIo!N^%?sqG^
z(3p@D<R43K1mE&{a?N&jW#;pLO)eE?Hz4iG)-C#RT2B*6{p9!!gECzw65@e=<s!Gy
zX&uqA_r>-9OwNXP0`anZegPdI(TMA4<royLiBvD0k%70sG-)-}6v%<=b&rn`nTmdT
zA&CG$wM}Wy0nlnqUy^JLNpuGGvns`6j8YJv0Jd^?czEiHuLAq(HpY8r1~}?T09XZg
z%#XHB6rTSk#@*v~LC>JzvkSp}!^DIkr2w|FLie*ukE~(S-au6)M1Z5!&|Utd-@7R~
z@FLV@A@DSq0}J|weDMB2sXyzb0nJBG##$j>13}X5>30_}O^)+mXhJx{J$7)tw17)u
zKhoU!C;_-ysMP_JsRnPVTrlZGAUXg};EuZb8WmyoCnDRMX<yG~Yx^$JyT1P3VSo7d
z$9W}mc_So#>*8-|b0cRq8}z~b=-NMG8<{Vzz$ZEzXwj?@(<bnmsYsJH%)INgBHDPY
z*uNQbf2YH1pmk)iLc)G-$!n`)z~ucoz459i%FUxtF5$`A`3SytS7!kU^3Q=iEa+p2
z5d6t(v3?lUV(D+MV;_HKi`Dt(-EvFz#?ks5xszEstkUvIc%J5-_py|ZjM>#>otgd?
zM){gIwhmP!%fjZI6FrLvA`ccTmN!%vd#*@{9qm59-bK}r_l+(j;#?vo@c`?z$6nGQ
zv}d^^mXHy@HCdwlP3hA;NXbGg0@Rg~5=q8wOpTRfTxjc?#2J@fgY1cR{etGy=`KV>
z-;5XK^>kLpfWEJ5u_mjIT?Pm&0Xpi#Q6h6*l;`xTQBKDxa{lDr_TJMSDQZDN$sirV
z94&)jyOx){#O^Yp7($Iq6?y&#{eM`SUy``PKAVF>O_;ONZco`t+peisAF|?G@Fn{O
zmlo9eKPh4>5%d<J-{m>}@hF3uOMZeEwUM_O+T#wfdc<WzCSas7p@dT)enK%MHUD#c
zi^Yg<Y#d}<rTQMW@1pXWQdROnG9#9Su+G`8cl8dTTCnI5#ZNqi2Z8GthreqLR12Yt
z<)GpTTf3;(7=|sorPlEX=iu7n)dAefCoX>pT&!>`)_z;?x)MHwYb-?STuoFadP0mC
zoC$+?1oeF1@!*z3LW!t$=UVV^tp(8Aar7XZ*7onhA3WU6L5&2~W}M)$I|tpKZHmQJ
z<RG`ES+NfK=@%w!+DhNbSG<0Y07j)w<EIpa_cLX1MZ~@3=7d!J>2_T;cq(Kn2gZO$
zf;v>{`-OGWw;vqjL$U*ykKdWvz7vPOpB^nfWVDF^;+sD>*&I*3x99#Y;j05Hbxh~l
z;fXn4UesF~!IUdyeHXtgC4=0cO0OWxy{(v9t%$Hf?q;W;<yxz{m+{koB~)q}&rxi$
z&H}{-%7TyJ<3-JwaD4voum-T)hh<H(GNV9J7AE#%vE;}`oyOyFhbyb&Lw+ADlx{Xl
z3WU$3xp~BPWg9oJjjuvKKMrED8EM8`5g1lBO;7iHdYB%m+ELjdp2tK~*oGCwH+~Ze
zqj!2kH_$Y>!|nVqk4deGYI&^13EVEBb)^#cr0|-}b)zDVuPh$dVpgsW&U!l+tx&Vq
zT^$``Ci}CFA7mw}|7WbgNLjCICsvF@Tx@l1OZ2cXg*Xup;|n{U6hbf<x#QcIo5PEG
z2a2ENO?9O7dqJ;bnJsKASCN>LYRT91LlUgohRdGr<Ea@(Z}9~3=Z^0n;!2dQN5oQG
z=)nJKEE9SY_48E(JsmwUbA%4x6Nfw|GU}DS6sQ88#XT)LLQxe3uFoVvB%%BeTXQWH
z192Z@pKefcyF$1Z(v)kF_0AnAA9iel@!~?Pb1M@l6dE3jym4h~L!Yvh+YPi4D|wqY
zd#=$Vdy3M_B(dN>^GtqR(p&nUt{KVGwx+@TLG$&0UGpz{9Rbm`9j!;tfq{<V4bu=!
z&$9Md-{nL6-?f0l+>6#m-I^=h+@MGU(P|4HaaC3D71-DVt9u6r?tQl9tj6rocWSKi
zbbz#-+-qa%++ezjD)W3&Iq>%GyCs!a9nxoMu%YY}e}lQf)_~Ew6R;k!`rVt9f=1%I
zU^Nw-ZiJTqix<YQgxghW4B7#Hhz9GIYKrcrCT5C(4zq%+1)B&bIP1!);OT9WQyWrF
zuVE7pHd0m3yYD!E9S&_gJ6dWnH%G^-GevhXMRzhq$AZtCyZQ0v02?^DjJcO5mwxf%
zHS(})JPnzhk@TU6ehYR+PC-bPxE)J~s4UFH=y%QihVjzceNQvk8`U?Ze=|bkuifpK
zThiX^ZN{jKUpc7-47QoASMeW<YMnRrklo;xSBa&s`_ivxBO4T-Ua96!5zTgRWuKxM
zlXGNqZOS;>tf%FcmwVJeAU-mt779Gi)pf2nRLIjPYWPkiHrS$!TB{NDYjs!kv1E$j
z2c+$S+9P_dUq!o?8<$TXI*AbSW4TL9hl)+Kt66I+Gw}#dQEqH0ju~!hhiCpRo(r3g
zh*1g1j3F`hFWt3X-alJ@CO*GmRea>PMPP2_ttE7-7Uw8%N9pr&LqDpF>|*`Ud^!Po
z>RS?#$V_CTMsZCF<xa^erS|Ok_=^CKsdv}w`?}eZgIjM*boMvVcg1X!LlfLlld!>@
zqxV9!CaIA_fAzXX*=om`!bx9Qb;P>68og14&`CfDLSPSZVsWAu@-so`Q}aw(iykey
zAbzbgRxL*J0LmePAd+ByEomShoA)mZWYQhHBXO`c(EU1EA`+`a^LeELmfg6k`CplK
zN}hU4&6c*#xeORXq$<7<W$|SMiGF#tLZBsnyOzewKYRb@78rYYfw70ma$7v8C48y5
z`55D{+3VT$YVe6fSi17I{gc|y#$JVEv*$7w%YL&hgNkQExKn)Z-K~G$8eFvk{_b)q
z{eDk1KT-bF=im9!Uc-(*l`^=!?f${KEaG%yU?Tb<-T$6vOB4a1g3q>7>P?G*{I>s;
z?Ap<DMesqu<3my6oOQrt_xt&Y-zc!jfqP!dbe*@RXb<j6NhVewROh_|(w_v1rC2F}
zG3qT3NG(&Plpj-l{2@K(az--c-ROmxEJFsI9{*o0a4vxd%r=w&C+;1OYLd;duLGFK
zk8QNb^+pz{i^5lw%&Sy2`kK^%F`p=bNR3P&R$NK0a}Pt(xd>FbYq;mkz_edA5af`^
zLFsG-ChlLqG<Sup#Gh(^4KjRDBIH#hc&<A_i}X7EMHG3n)bmi(P=->6@DIWB@A&*(
zZV`aumO}|PSPd#J^yaN|=b5Mj@=$II)zKhrDkiky`noeS(%Igf4qihMgeBLIKfO9<
z4UbuT|AIx;?w}T*zbYhMO(Ngd2tGl@(kr~p6s42;LLM^Hx&8`h%I$^(WE*bo=8qpw
zbG#~`33apbf9=9}?z#X+?SJ0xbr#_$p=iOVFxMWgW|K&U4TozhN$}6beJttS3!4tx
z$X=04DJOlVUlik7|6-;%O~&Iw>6@ujm)Y3k#u}!qdkyADuD=L;mXDIi0QnJD-MO;^
zv<VfZ_+}dhH5J$Yw0SBAP89X9zOr;n;^_S2BT6AgGfr`-Rk)DRxIvdt=#LTrmk6I*
z`cS<2yIlI=h-*RA^LihNf%zvb74(x|l;iFGq(gXowWry*`$wsBAUb^gqvAf$UGir7
zW`b{O`vml6)Vt(;gAilLdJSQ`1i=Bw^`k3{@C}Yi{-!5PkHqx&Hg7J9o&6Bo-YhG}
zwbPWEVDv}%@Px(e1TU6K5DdS#zQ&c23Dm`v{dX0-q3P^i^|(s?tL}uSQ0jt^cGnNu
z+ql`t7_YnMjR~?HiwB9<?Wj~x^p9rLMC(TNX9IOPmmR>g?Z3gsf2nyl(?0)wo%NRl
z<f_i$j=20zeJID-9MRv4qo&xS*+33lXEL~oh|{{oYcM#2(!CcP;TAF81M39WNw9r_
z_@wZENU`8yfIDVjVgfFiZD3S_lm39Z;b!aQvPpt|;rBZlQwJtAH(s3UAe+mO#{Re|
zU;)}&Hc4^k@br8t`T}$>h&;!^$1N<|9}4q>-qG=p1i!;4;XRo+O9!8n=jTc%4N4$k
z!ApmoQsP_O{$JuQb+G-d*z$LcX>kuU`=GCiAxuzTlsGwuzB7f1b4fKZWN!|-$nU<r
zILc$|)`Y^Z1W%4Z`DGKEzWTXk$UB+l^X2xhzr<eeh<uEqUzd8%_Q-T~`$>CoxUQ35
zK&I#7G=5;1S+XBWcW$XRU<VKBe0niO&fwhhg&paPA1b@%aOOWoiF7_RtvD4Uab7(|
z=iNt@W$DwQSzo=zQ*j~I67BIhd=W7mNk$*%Jk<gxpYIITXM9KTTTGJS{kRg^c>6&*
zd{4myjv(a2Bk(YCJ$HgKgQh?R?NH4wI})i@$)!=Jz;0R%K|9W4wO`apRfmCYpAS-}
z!!OP^Qufx>&WA&<*IfdM*mbb~)gsXR|1tFzKv95O|L{^vgGx6Tq_n`&sUWcu5=u&m
z<gy6TEFz_pEGbKg64Ko*APq`)qoi~Re$RUEegEIgjyjHl<D5CqIlnqqMyC@nHtB06
zNCle#c_vy2l7dN^;+sc$PinjlQr7_fr{QiKG}&n<;?KDT1#}$Jvvv5n4qAJ-;$)!z
zB+XT5lI7K3W-99I)Gh2nVM*PNK^1Ny8Ad_fJVXvc1`y*STd3V>QXqhtQD?X<ZkGPg
zPPd*BQEtt!(UG|7X2&MWPLdBlc);e3g`D5#!m*{SzU%u3{RgnvO3ppU?IS}z5e?&G
zI!|Wez^r0N8H?Wj-zaiNk5;s0<QMgQ2)?$QOt*}w>6Gv%Z#+wfspZ;7fsw9iik$Ob
z#SZ?%P>l74CdmH|q-599D;ReGL7OrGSIqyK_F9U|lF$5w8HE7Iz_5Oj(B4KnC;gvT
zx6WnPof%NKZbXx9y()3L)f8aDB?1<z+WLJ@nWEqv{Kri!=bbC2zJKQb1Zj*V94)?u
z_-<cT<AcH3#PipGPXo%z_AA<K{&6;4y<d+GjzND4{gjjL)5scY26A`7Pk^&NA;bgE
z0+!KJ4My*K7W{3Ll9X*9-h1z%4;(JT-NYeTIy*CzoYsy&?iq}^Nm)+Q{lmxai9#*K
z%)8MSVYU3QC+qrzA<!%nTsSlg7vf4C8v%)CYU|GA9rpUkjxmun+X-yx4!mS)=`O9D
zU%RX#zH4G%t~JTu+Fg1?hv#E@t<b4ywUqMbcPqVtNvGC;dWqI<o$5shH3Ok{KF${*
z7n&=w!0yr+ONv%4a8)5*3Ra6xpgR3>(K09XM$IfIoL-H*mcWOl+lOkiJUGH()`PF8
z-?@3|h&QSk?ky}NX*Af&flm?2z6}nQkWNT1*o%b|VWK^qpS`l3_mli{v*yamL42gS
z#l7_!exWvuBXPlcLd<q7kKi>wnWG%xnw<b%K^Zd%)WTn<BO*^Xy9w{^w2>o@HQ`;`
z6@;jxEGesUG4Y88v0O&q=b1kTL4@3dYk9<Qa5lan<cq+sseR5{q+ER0N^csl_b7a`
zFk<9<ienVy6^_q`T7#?&?R=vWdeq`cu|{gS{Hf0q{#LMM1NTF7H;#?%-HbDg7;2Qx
z{}<gVm+QYx@9Dp;8`1XAi~KSZ<QKh>`F&jL+*6c=Tyn(LvYl$I@k~l?eb#7RpH#f2
z=JCz%ino18?;!RwAFln{C`+81O~EmyiBCJ`47qwg`bKH)rG&D%*WdcnocX^{xxMdf
zQ?Gb$X5D@$1T+qya{|N}V9g7m{y#&1qvg-g;7$|xfR6jG0q^>s)O#NEOYDsqyDS3Y
z5Abj5JRpFG`EeCw9=rN^dw)>&1ye>)YYT)NgZG^FWmVV4ZPfnu5J1~ZKPD8A8}8VB
zOH{7He#ifCG1&3r>hj{u3ykE3m~EA5bTFDf&hf9-ucM2f07$dlDs8n0yUK`ewMfG2
z0qiwc{<{7xn;rkAiBb<eIjJ_gPb_ia-nGW5mgLz#1Bzd=p3i4owsbpB7PWN`OW_Pn
z6i!zSR4rI<XHSuN8Z6ari$ls5<%luNQyhr>9e@kFxj%t1483{sqXB<fNhK2DR#1o>
zsESB2ze|@d1Tm3Ia{+gPquSClt9S=;udeKC?IUy_9e7w>I(XP0FD}`eAC}s7tq(ag
zH@g>gM{G<`dzrGr4d(o-&A#o8H<69y2ciq#757U%mLMxz4`OxXf@KTCB+EKyJS=PV
zk%fWDmnTY+{`lg@p_2EGKktwJiFds#pvyhC+2Q*u7?Xz!7Z{5n2+`LWZ0o6_`MhZ?
zVasAn<983<Ln|YINpvT+QMBoRGw6Fq*PzOJ9=|vIHKRxN!P#FAbnz0ydGpL|1AZaM
zYj-UJ3?e)1KRcbe;+?GKJ5&h}el@&KQG){0I0m{;2*o}+Z5fVE0-m(jqxdNIM`@WT
zbRy>(qr!Qx`T6~`Wi2oas4xMxd6?s^4veQ*8`b3(m$@w0(^6(Eu=0ty`2#l=Kh}8Y
z?CSE-IocqnZ1b^*>*S_)8;-jcz=l6s7ZqjW!YOJ!9Tiyu(WXzTf6kH15&y#{M5F!j
zyBb@Y^se8xHjM=*3ikcj&WAz)v5d9r8QlfH*uxIL+tu@({~?^;XJhXm0Z>grMRMpB
zpfdp9(P{y)9bieWl)l42s99UVfh%_aPONXPyEthtuch$`W)0aj>|Ldt9_{|6xSG+B
z_uoF7xknDL4-no0kYg-d!7Ag40s^9&y&`?KqMu$czL8N$G^F4@@nKX0z|Q3rd+7TM
zzjBu$)c~Qd7T^PGLBM*Rtqyxe{6On5rw9&4bgI%bgn13Iu|M>djQLoDcr-?NF<=jk
z6YFX)RjB31Y$fLLvEjPo<;9&i{Y>*^fHCNFts?3_d%AbipQVe5hoBqVz;#Hig4vet
z`+`#={Kak(i3p~hO%T7$VN}D9F`?vFJtmxXfSONNO;<VQn-mjqN#K1y5>|HIK584P
zdqK{nXOorf&;8eat}5b^YoeI$Ll!vZMB!+#xN<FND}i6JT<-UJynFSfph7TYUh35n
zlGH}Kasy@({MKsP`Kg==Gv1?=g%uJzmdsbPdv9C2D=o?h%M&j24F~b8Q~T9r10o-@
z55h1exIg$Df)~wc^NYInU1>NY8l^@(EZJ-T2f|5I$1IP_L|8H$bMR6}4iim^tltgR
z<SCCn2$IAU&}o-O-9<SnST??}urWd~oVoTh@I9x{UrZ~Qk@7ZvG9>9T?pNa{Xp;PF
zx~SpAG*?n4#=Qo~;y6?C#@=|yr*|_UoL(haJw|w#k@&M_1@syDM?maZn7y_mws2-6
zNy3v+R?_=U(L;5O(?T;$_|E5Siz<tKW=42g7D&Av+Uee>f?=_Tx;T7YT9%L0BwVXB
zJyG9|C)>WYx7=FGB_qnkvLVLknKkEp=jg}(ih28!V8^Tp*I;{WfT;RQqFArG4<;}#
zRZkIx*;Fo4UM>mh$4x-`vkZpG8LXs?W*7>jHvmGogE&(SI8+5%GPf=pQo&*azC-Uu
zW_8=OnX?-TO>ErGMsGFBe*RCqQ+bzvv%_#U>`xndd~euCaCvoavU1h!hxMQp_*$M|
zJH+xHU^bR-0X%l8Suq$$!2whd3aij#97_;a=uifI9}2>@!B~v_ks<_uy--a7*2^o%
zt@R@?p@J_1lN&bwzz}t>WoGwCG5i>(G+mZdq4tfP!Tz)uW-DR#tL&N7;)*tlUjxNN
zM5W*TlNF7Y?s)N=k{1Ln00frur&?S&E^!sBHaeT|d0_}}-r2Fh)?sFhAFy;VP4jLS
zIF6u3#4d0Te$c|B>ZuZ=@tz4y7Bam?=l^!0fl~x_?>NxGqgLtNNx(Aqg*vN_<(2*M
zh2<?nwMlkQ#z9$=Z2YVT3oE=e+nwafoqcA{PW_6<c;a1Ah9^d3LIcr+miPOd9Kb)(
z@o&<6s5(=8ZGlpOVn%KDC1AhH{BM?&t_EKoPt|c6?(actJX$PCxz{fylq$|Mk=(VA
zL>lRmvTZ{GUJo|jXsf`N9nN?4Ek>%M4(<fv#)>bLhuh$+>N-BBNJ@p*FVjr$M%t{p
z&q@*P9GSUpT}$G&QUaiY3eSzFB6b9AZ^IKHf{>>OB~b|Jwf-cqR526xhYJ&ULbHj(
zaxD5FX~KKVEfI%;N%aHW8W28caVAS#x!pSUoxg35l|cmk_<zu^;BPS!iIr=gWL`tj
z+vUM={8+qfMF#taUq0-9_>#)~RfSailbqN|P1<YJnqKq&Rae$E%uEwk)4)A!T%OL1
z9Q-z%r&BxQ<i;1^`y7+a1K<J||ED?DwnY#;T&{Nza&CG5{inU(%WO?586w}k3~#>N
z{X2ge@aJp#q?eq8&e&Pw{4UT!l3~GuEY;}W7yk3RSpEZ;PscXJXx{}xdW>~rZ~r{H
zSR)m{!@EgImLc;M0c%6t15NS!v}8;{^#y<vgeXIAMMOs-e37rSva+!EyA<PA(q*^T
z2XFImd}#gS8?&#A66yFmHoY?cO-o@dAJPMsmN`0*YiVKQms&^9dRl`M*Qf<!Kvf&y
z)<PFn?59UvX<)T}7s98P+6yZNHpWU{w~>hB%9mCiZqE35R_kJ4cAPNpps?CJGcvdU
zUd!mR0zEuRv5*5c$^QeND>2nKAM8a|MC<|qhq-OtP8N2}(M3R;X6Po!Cb(Zg{&k@T
z^tMShWT;b8mVM(f0x8?D$xQmYR!3pFpGU-2VX{?{gv{`q8G^~r(?=Faaj5nXjy<NY
z8wk>$-p?G*K4EHbx=4<CBF(g3%b-oqE7<-47h0Z-CxEPiclUh(2gR82?zXq|^}O!t
zfedjPuEN-jwIu@vVHc1Uo5{I&ovL)lD_Yk~Yc}lxI(3AJ<uaS}tFJviKFdz0^~42X
z)Tp>{-D`#P75F8PXE@ujGK9kzV$o|_$-;P6Bo-teC5?i<LHd(D)%*0w#jF%tUI*S0
zO*BFG-y(>4{r<Au{K{uP6w9WC3-AZV-6zcZKUwdihu287Cxhb%A)Q00vcFr>ywi0n
zGXKU+yzX!}DtDg>DYWVH8?_(rv2abjTaqZaKO5KgW7OIOIbyQZRF``=;`PL0Eqgx&
z$Z%?Rsa2#wiTuYi1nH5uMxaI;Y>l&84KkNv%lu88q!?Wq1lMPKd;D9M0M^<Uakp0}
z6~1@)OApHecph&t0JXzG*~i{@VTW|aSSV{ub`o}n5_B~Hr#_(n&}HXq4O!Kp5Ln;o
z>C|~5&BON0_vM>byEm;CZ_J7l@p@?0IQBUrggzNg&52+fdz}FTyW~Re5`YkCWeXHX
zh|D|{EBPs~Xiaw*nV={(ru7;f&P_5%f-$LPiK--pNglJ#P{Zdw6{<UZadaedN^}z#
z7aW*$KcGgEMzxl?1L*Qioubl$13}rLWye16uHZ=~EBrWf-c$aixh1@Rb%J9^YFDDo
z>!LPM;b}kt32t><v$WY>4G@tNKHcmT|M}AKBy?f2rS8R>Y1!i%l8cb>j7!I*came-
zhPkozlvm>I-TFx~&`$}|c7#lc2M4{O7MnkfYR@KXpo|8`ZBN0mSW+4{x%IdDI9z}@
zUrTOo2BV_aPq7!N@86yEcf~UFxFqVoFr93O!}%fxF@<z@Bt6ldkx5(=l`F@=+n(Cz
z2&?s2#ZM@}6-JR|L{JqdelkT9L)F$8sQMw~x67eut~FsXN^&7lha`riAmPxq8xiri
z2I0DVc0nx_hkf5(Uewpm{8J80b6~3}ACjxd%?`A*VZQj}oNUQm%mlfP&sM3=8#DT{
zyloy~r^w1!w5_qb&nKyP)127a4_kLim_|P1X<w2%=tuz2$cq?bBDi@L%WQpobl!6@
z-ywf_CGk5v@e07C*SC#uIp1v@CcF4Eq+bk-GpctpBHbs4PY1S`cMGyT@5#InxXT+J
z2G*L^KA`?miv?LF%)5;ztaJVsPBMPy)Ik4rzj%E5)fd8BY`W#zJa>q_A3g#b@&$M<
zYx(vqU?YfqL<E@V-h<{_BJ({1{FX}b^UV)Zo$g?+yZP|OHJz(qp%&bp;~#cp-}&TS
z$hgN*nLG3qI#k*dO_~<w5bwx%%An3?-%6U`ATpy~!1ZyL5I%0=k_z&H^%j`$w18)!
z9rfZ2bg|5;Y8<ZWz=#w}Ab{-}_hP;o<rI42byoET&Hrbb<P$|}x0x^7ZD-a*`vL#v
zy9&my(i}*+LR-Eg{nm|Ge*bFi<`9xhGo&B5ixc&k-k*=VvYCO5CCU^>x8GrYPXb>b
zj{w6QVwH~&xv;^t(w$@rdTQL(t@u-n=w|5gXiMkljuQ>M$nD+nG-;Ib9q-JTdXJI@
zx@s{;GI}n{b$6nQf!+tq*~&*XDD=UgCoR6XTYgt5Lel_xs-Mie))80gSB-M*RF~#T
z8H`f@1>+U$J#%G4RPFR&cnC$-WL}X<u2XcAbu~_=_wIXfyp;RhC4aBVov-y+_})(K
z%;2pmzK7S77|34Z#sqTPDA_|+3BuNgEcWqR;^?VtTC8z7A?A=IAy3@S$MyU_a8dq3
zf?=iBOJYgmUkheqkCc_W2`~lpea*+cf1PH((*dE@`&_J7f#NAeuI3P8|Avj4R|qaV
zt=<3k_l&a3@V7BpW!TR5RPFx{S~`WhDoc+kj?S0Ry6cHNGDIUGtEu4>0clALnlv+u
z4}R)BGHX}e^}EK#oDRmi_i7b{m%Gj9cYH7Z6n?lNvDf!X%+y__>G;a{^pX5;E934}
z@ooY({xDy_h&KFgBL#fF918GJ*z((LIH1V>@q*y@wRk|=#j!s;EQiKbzC#x0;Mi<u
zIGV(TPH}j_Kswa-n0>@O5ZNn^+BZsMh!SG9=UQ!Ye=Hv3Ry?j6dx%tE=%d)Om5=VU
zP@y~}ZOsVu@9ep-L4TUHxdhTD|5-o{IY+Q-*Aw6MCOBDfZtv#&(9*qnDZvb%BchK|
zYSKDNeae4uoM7`SNv^Qda&+t6K%}gBSJ&sIl{yZLiTe!!Du3bMl>p#?5vlGU^j5fP
z31lqM%P<FJwzIzgSUalnNZVb^-@i}y;|{RZEY@V!@&)ta6oN?%Hhfc8ybP*zWHJF0
za*W*U@9%!KalG`VC`I35e&M{JsNY`tdaiUrj1HfKL(bms)8GSQmIYL!;=rL40@)=0
zO3yaOj}(&6EK;?+DaE=DDfbdM#_QvKB~fLijT%HP^@(S-T2T*li$5d3L$^!pv0pwd
zm$4<yO=rd8FaKcuX4wrBwdS&U8l1jQl*vD2!zLPRKl5>4qgP=-QNIxb62N;I1q?~V
z^n=O5Hsp8AAu>3NuLDUfO;AF&EP7NlP;^6g`$-n!)k|*P=Sk92*M}-7`XutL&CCo=
z81L@4{X@o`-(h!$24-z90&@D3OnHkZ?2`m>>qA3;6bfAvMqP0lQJyc~Qr=k#Q@N@0
z82NfeWtOE7ZZno=?ZI(oSL&l<p)q0ua78V$b<UO|#>^DaM;;9bJjK;_FN1;Lt{wOX
z(t%bKOD!FNeg}+%<Q0X(*!?0ft{SAwgn~P(*pOzdp!)FV^*O&2Q*4F(OK7?+$k&2=
zqGFASpt{|N56H-e-siRB%EjZrF{9dC%o67Y^SajtR#-$q;L74RDV}KZjDp3LMS!QE
zBk|mtbF7ZMWwn(c*C!Mg-lk_pw}!D&GX&A%U6RF9#^(oxOYlrXi^x7%?~~vK_@6C#
z(kDe!F5>k7?%Z(6<T@OP1@hAK0fdJ!Y$&2sq^oS-Cc(`7xymC9y0u-&Q?Cq@lybff
zt98FvrHG4q5(cY*1}*CjNRH%yvXxu;-~uGaG&L2_oukwczX|a;!}rGYUkQ}pRsN)_
zmjaj>GUZUK#}>oPy5!@Oilb5|LiaY?yP0{Zl7pU}SX|VBp69_86|J#Fzksv&S&}TM
z#@S~}KQ~Byx}{r@UZzZ(&lT6<lWvp#A0bCF7lA@vfm!F|A<tV2*^XXTq{Qv6j|Kv5
z1bULuYT0Dv8>n7DGa!0@S|m$C1u#+kxhA-~4A}*eJ9XSCPJWvD>??o7-eTTbt%{t5
zm0vK^pD$c(cnKkLpLWVVfQXh>_1&@+5fV{5m>iR2_yie%JjF8#GHUNoakEVm5ei-+
zrb@fM{2E0}#R(BmTYTLzm?^~31{MZ-jcW!Niw<I4ZnHai>S%r21#3Wzu4BmBJ*^-u
z#g}huus)XeKR0fuJ$ZTS5I6vmv5Gyhd!J^4%%+v8QTu@QVbo;BqfdUmS>rexSBMr#
z{<rlaZ&%ky=v1rbV<{Vk;l=;4?Hk_(I0H$Srf<ZjIY&^_6C1j8vKGDdZ@Aj2-h0)2
z3B5X}{CmiHa6MdQJa67}-S`wRgtU_0dIN6L!&R!w4wV))xrJdw00#scCRW9Pg;P|c
zSr>o($N?Foz&!vkMYSK)PZBtK%lvX?7+!Eo4i|Mo)pw3Arer4_jaxhcCj?`rf%jy+
z_fThzM4qe++r4?V^HFAGI=d<IMW!7jBlZBJf0}9_EHQ_#P0^YaSZbFg&hExi@idNj
ztKJqS`9bxXcoenA#@803YI}Z+BZ)Hs9D0w)+|3t+G7F{i0myn=8HnpjPBUA(Bb__R
z5y%FAoqTL1;-BjWqnF*xv2`f5@Pw^cse)#RUSGT7>AV*Ufnlone$8l7W&;EkmA@#L
zd&AC{S7&m!JztpK>lRE>$V@+U)_zXv#OSYI&aC|`z4lIfH_cBG+zDJ8BJr^kT!{OI
zSZgzf9hE+;cH`K`9X6>{nK`pBB|J0{P30*t`l2+={S0!O>B%hx$d=%%I!0;SHN2sW
zDp5F@f=%#?B2un$Ji>Yul7fbB@FByuKx=X?QTm>DhLE0QT|idNCW@6Bf3>K}5K*r+
zQeDF}sahwdQxixZS2|#0==^RM!uQ5<<SU`@4Iv=z5T+7=0w)!O!Fb);f?T^to6oLg
zg8pzXB9OQgCpI}<(dE&N4k`#0w@J(rsjb+Zj?Qn<hEklm>+h3UCq91;TrsC*#~K4w
z>feL?l_?9X5Z#Bv`Nc+RUZ^&$$yd4_vvI0J6nz4@1@eKLp);ePW^Ue+<f!^TqHy+M
z(a$aQ|Kft24_vcsxYiSxNVw`ns_(}z`wHftHSgE|Wn^nS5%FKUqTJ?S0}F8KdE~1m
zY;PoGmOONu_NdGsKHXk-u}Ua(onhL_4W7QWneDqfCyQ0%gM!26COL{L$KJ;75}at_
z@Aqb3eM!~XNx<0A*Us-cR4-p;RHQ2Rv}bxeVvSUv&RXOSO;F8pr4TmOu9c_%IKR7j
z%-#J$qQBR`u2(&crV}`q$|7quPF7%X4E9+RF_Eh@LC(Fa8!eOBQXrr9@>0vfF;{lM
zwzV5_c#$2e;nl^_4!o$wqV&F!Fv*R>#_L4jfQU(AzRj@xbZ8C{#-z}ke_TdlHAHDs
zYVENJAy*d%cIuisd^RQHEJGJ?fkYtjd7W{%y4=d%%l;e>3cQ4sHn{JUR$=7sZ#-Bp
z_}+!Qa9UV_C7BEq3PfAT%`NKw#WSqyiRvBQBWs)F$Y|nh{(hdvOg^4p;~shg>1L0L
z%pdabGEurb<Xy3uT~|;Keplh<_db3ZMiwHjF;;9!L_ef}#G%Yu`Pk<;6MGcZ;D{4n
ze&BBaH515Ki8U^lHD8J9?Qr$zsi%~V(LY(~g>gv$3iXlgTn*v<uWA`dYJuE(Dhxyn
zRG-;DIWAap(q|+YLioduibF-MDAsr~aBPGgk$sX0)&<@@YneNywhvmE7YOc&F5~tv
zptjhqz6bEh8{Wc#$qoS0`k2DGK5b*981Qrdnau^ga(lP(&l4p!Mno072j5kX78F)(
z|LoIQ);ze)c%F5U>VAKY6R&@}_J}F=Z9Q|s^-=VriyYcc;*xzsmM!#W$ZvH0k9ojm
z_3vo;bGl+4Ixx>s`mLpyQ^2^{md+fjr;ft@&lg$~pHl4K91W!OSg|~CiPV}bEiK(7
zL>uPQfHAI+Y0@dy<I&gvz=ks*v2G=vsK7~i=BU2%;ON4XeMq}vsSAvGqYPxqN{+}V
zG)n;J@gYM68IJp?Oe2die%v(dW;Opb05nWfwWYV84I8Cu0D%Ldynf3TfrN3r?aX(O
zh~`8fa~4;AKFte6XI(lt<~ODHx450v*dUGDyWgLI6*X$m%1<)`n)Dc_1_l9Ek7<B`
zK~#18i+GGzuq{o<^}Uz7IYS(|)WR65emVrbnHoJSnVW#KG5eM#G|*Y4v%l)c$g9R#
zR=5C8dtkIH5N<Nq7Y#grX>tp(Z)1v9_9KZjCDSZ&Bt*We{8e1}^Z7G@pbDLF#~)@C
zX9f$_vdxa0$Wh5!NKF48vygaH*BbpMOPY^J#^Z!rBU7c5Je%#E6fgVw?z~zpBQ2K}
zTov1q=IQFZ<GGVtzt7m>ZETu%#}b~Q5GTy7uK@l1A$*M8Omvv!4NaH<fsv+$&>D#m
z%Im=!NfZxAKkjSs#p$54NV~brAUqf&Dm<phz;|E-fZbItdn`^?K?hXAbvoYb%jtcr
zjpypnOU2&TJ*06Qs|-MJqJ+ak*_X90bGKk-E<q9WFz1*R@ss}}eck$RsC#Mr4^;b5
zL<;vva}K7Ts^pa9)~+>8100$bkPH8%032HME)3h1X0a44u&gmQ`ClcxTG1Y`8qeBf
zj4W>ao1oBR_!vMSkq@wPbs(DwBLZF{AZ75p`XdQ2Tdc~_-`x2b1d6?E4~{0#lYzO+
zf_-`MTg`2z&=o}rmS~d16~*RU385}RWwHmsZpdaA$f_Cc$J-<iL*|_RwImVk-HI97
zjXUzhBkkQ%2Hdvvv@d(RHos00uU{w3>HTeejW*jM1f6HHIJ9Uj4UTbiRiBV^8E&``
zaZUS8qtN^-Gb08s%(EPj!$wT7WDP{RUKQXOkb*^_Xk`neOtFwQ{cCjZqLNTPrEgUz
zTEnRx+)|NpZ(WQ@amy|IeFc2>A=bAptPu93uv%KpsyBIwdfd_9_aGRPjerG`aZtav
zw2SSLZezA;PLT7e;z@8uEHZy*+}-?pK4e*pqE)=WEbBaY30GijsO(2k4>ODbYEWe%
z1hHv-rvG)d?N?H~2z;1~WzJQ}fOj|1Z}8`NHYEi<IlZb)3uIV<;!%o!t>2fX#LZcB
z(58f1l+U2nDxRf+aH(Z>1ThcQ7<&OM6O@vs(ap>|Eu@@xAxQr_nCDPfns5ob1xY9V
zhgZoWLU%fu-w@D~f~`VW;mHgdD_yLR=%$ey)`}CMoYw(XF{`hNt1&5Meqf#-E{wta
z7XRN2mAskQIO5wIx=4a8oXyzUI?UZ9$KCZUA}s;uxE)UKb$6qTecLm?|A=uD_geQR
zqxP|LRH2~n$@~)V+B7tp%D*x#F7mlI2Maj(VBFw!^XB%%^ezZ6$pH6>wDhoo0;~w#
zTBH@=u>bJ^*(*k-+`Xw33w%vx%`Tk|VOI#$FanE!4ckS4$!Spyulv}`B5<`;A-dAl
zZ0F>{w%&y~4iHh%w;!s<xKL#<TSrXe_FPXoEkhyoCJi!dxOFn(eUL==qN?=t)~_vC
zSI+=fIz%n@bixS_emp%S*1r11i3GEiFWKx(B=NM$3V}3}#%0;e`uwpKd7duH)<zl6
zzs|Ht6}X=dO$<f{QB-_BH|t}ccuJt4T;eoawJWsMRk%hK;$W&>RokdW#^!yDdY&IM
zU7r3*sJcN3-?s28aZrgOObqu~{k9|U`9!*YHpADXs5Gpgaqb$++j2mmr%~PhEY-5*
z*9)i|ZpESFzqsa5SU-88H|i%>)G#iqYAL#Qwr@a!NgS4TyOt~&ir%g<=2~Sg@mK`6
zBQCh-_<91<14-`o)51O6f<tn21k59zIQA2^oN`Xohy;?aDl5A%-Ld542HI{BI5$BF
z?z0XR(rf(50B;=t4)Yh~vTe+LdUjPd4^*ZYl1oug^mCz3U;I8rqt{-<CBd=$OjcT#
zy0*FI7dALpU5Q$SIkqK*m&G7Ma35Cz{S@zDCWp-hf+b#*D@q9biJxO?a1W=^?f3uG
zgSP2CE_7RB@BT>$xs&tzAJ_GF%x#+TKyJ=Fc%q$5mOIlZ4AP(Je%**pZwN8J%#&}P
zeF>`4>@h)H{i~{}oMSNnuKwlpzr1J;lII#YG6Z_7dyIj}u3$RCdJsU40=5TX8xRPp
z28XXe_KshUH$Sg(T(}oFw;CV>?1!rK<f`;6EYI!%ID*+qh&Sp8xWWmQO>^Xj;C*n6
zS+OVU;e{I56M@z>IIw+4dIwYy8Qi0{qopg!V>OKoLOj|L6|K(#<RPpO(^+QRO8{P_
z{A@MZSeWr`9{1q;d2b|*l#TF!7%yXS+{}RxbfhI*E{XQ;(po!GuEYz^WD}<Ix9Y=c
z13QeAm2Ska<S)C~T8k@EB^4Ed5Iln^MOm=w5w6)A2cClvYNNqG$>r@z6k6ZCV1~Z*
zxYsZ2D36S5ysa^_2J4*dT35d!@H6gyn1}FRrC+?dFs$Kq7$S6$jVQYgr$9BvR|b8V
zY8jJjV7&NlLPQ(Y+ATS<Sxs}7)k|>PMC@*%v7TTkYS!Hu$3`DUzn)D@YgAQw*Ukht
z#+~K!V~V?YQ?d-)Eh~RypHGbl7ix$m4)FJcXK}oM)S^YzT_G!_`>vIAhE9K^PA7i3
zc(UJGz4SYCV#`sy*{@4Qt-wGfa;*@jAX%;c4wWE%lJNVG*l^&Cfeg0XBd{a+jY0AB
z6$gu!Sbw||_?-Q2xEF<ihETDZE5&WFPKssp6(lE`U(o!w+yb!WmRA;2=B`^cPw&>c
zB784tW+i*;pHy`?G7hsKuJa1)f63VW@0mJeRT<r<_hixd$!bugjR!{{n(zCGVBb$d
z=bar@!@;E^U_)uX*p~P`9&nmd9TN>i+lfuWUWSlW#%-T_{)^q4z+#ZNR|Xq*xeo$r
zK+XX0-u{oN1d=|Be?cA#v#;hm>??uPPx3e9>$CSk@6Y}`V?m0AWph;B{>Aa@1>Ne*
zr6+(Sa%W>zvxFZk3wK(P^H8c<MLd8$X0;2o-prWUJgL+`7C&#XW~rZN${63xOgxM^
zypeP1t#I}ofh@7CoykTZQ+<%e<BLG=Kk9i@J3by>Nre=Oy1muW^u=d?Y{8nB*O7#)
zYl`2j%e2`=6vG_qo$5^pgrN?6Wc%kA;=PsGES@AIuNGJ8JU*=7LCjoSh*7Lo1);4n
ztc_t$&fM*L5-?86F|O26g|AEfn?FpGwpKm^_O0;{jPqs&qB(+14BFF~u>g1k(J|-J
zO%z(=NNLD9&pAHvr5>4I6JITbgu;>&G8YJ}U{9LOzpN^`E`u#p^Kj!CyJ|<rI<S1y
zyA{Ua5v5FEc(_T0)-@8|8JT>w*(}&XwM$`o78hq^8OY+F04ZD_0nD_a_$BlBi%DVk
z+zPAZ+Rc$}E}2hK2DADTP}rpQH(P7SZBAvskG7CLO)~and<F%cFg;GdloPZ8*2oU#
z1t_C2gb0ovr)H}Rj?gGYZxW6X$_{7kj<6uHk#=?h#zISzkNcAaJQHV?Oreuem-b&K
zj7loj5u3vOVCJFMPnicp+WSL3J6UAB#%fRI1VKaVd@(QKPO0%f$gA#_pF6Mr51p0&
z8e;Rb{d)u7Z7==xp^i^eN!#kDw`9*0g1O30IoB8e*=g>?+D+4I|0)|t^4$n8dPI<M
z_c_AQ^dDJi?zu1J53X31?ml*o08>>|bhIlLYCZg}@gLoiQSmsSov;lKh(tipGoUQX
z?oPkzF{HYF<L2$~kRYCiG1{fGqA28`6!%HwyM4QeMMT)}DG`w00WE%`EUT3e^NX-a
z*nfp{hu%~7B_H0{&Yjr=1^52&rF1w4XTE|XQPeDXh6!OvNT>^liAv|TU?1_MsN;)Q
zoZ!WfJ)T7qz&)bfnc!ni!HN@pZQa`@37g{+Hei%L76aIWl^X-BR@nbXQrtl#i(vd-
z=WXaWTpXdwI~Wrv({yXME7t<TIMS8nX9;ulR2pe<$WkX{mgG%?4Xjp%^ZtmGwgG_&
z!kvvTQuP2a{Q}R%a9vU+%0EIJffR$g%PG5UjP1D<9(Fioal-ASrzo_S-!O+n@|igD
z6rWy<<mn7>iJP14Eqr*@HHCvRt*OE1;-e!HcS>=8mT428H{(^al-uWw!4-l={3u)W
ze0QR|$Gz{yoa}bWFeg{tY=YDs=<3Jo8@?Acx+oYg*iZGI6$dvWSvDxFUryEprG!h+
zo-2gaWmJ&vza|!iFe=$<Lef5r1ETWbUa+LNO&fEB<h}K>5$Em(OACUJ0MoaC^XM^<
z#q?qRKJ|odJMg3aVf9kC0tN0%=tPKo+N$F=CYlUI`*E$}kf_x@XHE73>)*#9Lqq2c
z){uW@RFDtw^3m~!Br-isEVkf$VG{3bUV2hTdIRQlOZf27!&sd7;#V^sKc;fh8bSj0
zxUP85#RO1*Hb^<|lW2}8P7k0&(h81ryg$Wy@4OP0a=8*g%(m?VFvf$B`~R4qkQv|F
zzBQT&62{|zbfQl$>IN%{?~-#n5H8<-c<2FM!(Z!W#Q<dG|0FZA^DS#wO`0b<8svNn
z%Sn9~=vfvHX}=DwatzRhbO!~#P_E9x`S}&{{UQEoBnrt7S=CxtX&}4yjYB&p+tJwW
ztZrfDTxysX>%SvTCBB>xK3V^|>)==rD|LE;00SC%s%ni%uGvo`U9H<@>$EN1DUfCm
zU7J!h4WO<g=C^rT(5cq$y~>wfuv(G%L=@T-CPt)}ut8|~#rvmWS0Qeu*0eCH8X)>O
zd04H+b@2Qa9XL&_if<h=4@$>8f@nKUKOh6vy#?L(v%p`9GS>PLy?oNxVCENZ@#7n#
z;B#t*lW#Rod5*`4F7gtl$@^_g=Nn6rR>8c6n0&t0{Ovd68p16-1c59<MuFsErjS;F
z3=>yP@zr<F<0Z0Sc$iZ~`pK0Z^;L%l&8k{*YfE3`qHdZIXnQo@(`F6w9pGy>aQiH-
zEne~%n41|NKm(IoAUvGh+vHj#pG+eZYs2SARQSX(BvfmVZ1xWd^>-er;q_42YAS*#
zinu^xdQn{&!V>mQl7a;Go}`%AB{3Te=VqT27agjCRgysAa;n(Z0D{d&x#6ykJbL_6
zuOnBA$M|$2G){GG@pTWaZWZNbwx+EySQaS^OdCq4wf;w5{wmuV<!2}<yD9PRe(IQ3
zIv0hb{_)BQKTH=n_%TKPQGHBfX*>jCN|@U)P3!;aoI2fmpV90rT)zB=2N>5pj+l0R
zNwGZUN$P`~k=1}7P2N|WS6QI}7V|Hd8)9#_g3x-934pz)M^6TRjLFV(wXnjA76yHd
z6W!gU*A%N>f@lskD~|pu#dYEhhOxtoAZBe=Mg=b+ddr_mJyzsg(GRR5LT-@(I|z^f
zLzVxrOm5dc+{0*w6T}4ySu^2o{kgf_vRsEuso|0c?5p>9bj;#)@O0Q$94vq(o`wt7
z0Rhx-aVBBdWWqEvysG;7#^yOf1hw&jtcnX!R_p}I%hi6|AZ70bTe<g)vjcY_2?*xw
zc9s^H@Bnpds&Q2r0;V<?J<|jo3LPY5_xVx?wz+2LdU3Qg2BLxFPJcgVp!c{~Tr*2A
zzo^4~C%(dh?}X>39r!M2=miaD)R+s^IN=+sWj#EJ_4hz@j*wSAU?4;v_~Kc(*zp@o
z@rQv6B8G}5PHYB8oh3K;zQ2<97c&#fc?y#p3jU(u?bYXY@+<>E^n4(5f%ms@-)G(A
zq8AI+n;!~l5D%;9?u+U>PQ#Af>Z5WnH6#`U9f<wq3?JZfSRWY7cF{Rm+27fKg{(A=
z-|Qt|w-mQ1gT_B})rHZD-uduwIHoKFjYG;M*k?sy0bL`-<2LTHCgmH<@Wk!-fuE$4
zZL3*v`*Q+hEG|zBvYCWgbD9pk=UK}VE9ie!o?8Ey9{^o9|DmY=^40tgH~7YY>X)3R
zKI)stG;Xgz(w6hdb;9_1Ul$cnfUjez#{Unj_~G}9Pvsh-<Ti<8cpYz}Z~x7vXGoXD
z(b;=pyW24<%;hvZLSbSfrA&{{&TW6abGZ_~8VUMKfQ2s1jZ?rQz5hQzb^aA&>M+Nj
z6wa9aQ5&4XRt4o+V66Dh`2$?hTfHLOpJQwofcgh)46H9d$WmVi06lnV0)!Nra{?_%
zX&nexaiHxu!0r&zu0z2`sR|0tDk9RYEF?~q(UO%yY)0m1*7$hFt5`ekTGgq_$=4;%
zAVSaseW~j^DNM!S&Fklou`NY9Wv*Vc%ay1m<=U*0*tBAXAIFk^4PSjUH%~BpoaN9O
z?nb(3y&s4SL~l+@UxP_fyD=d|GK%kf<@kwg>c7mq;WEDBJPj<IEi2!zzyQ>t1=^sT
z1vB4fU1rF+emGMg+6v`FFf<EdO(N@!1)yAAAQ=2~!jA)1OVZNmHba2v6Br^bipvW^
zPj12;P4*|q5oPjIS9Pix<kgNtPa3|PKd*;^9JPZ{59Q?0u3Nhr8KhS-wK~b~O5GOC
zs?$gh%G@79_mMp?=9FL~Xq3i6`7yZx)WG)hGS7ItBw@z;E=aknyFlk?f*Z~%?$hG|
zr_VJ54fR|augGqFkzU7F$5SZ?n?&_6+|YGYcm`K1=T_m-qu??=Xdu9Iv%H(JXubWB
z%p?Yi&ap^C2|w$MNO}%g`{YT|L8@yHzuzE6BJ@P~R*<VkRq9!#`E~ApIQ?uQc39&L
zIfF6)i4Q?kVM-YRdNx#cOqs^H%JAz}6}I`AA<L#zW?2gdtjCj8C?1hLSTN|~rzWUU
zG1d2fHc|VPslIhr|7ChlNB7L(EO)eBMdY~7FIuQiF{S=v?;B2vWFeMHN)7k5z2rp?
z^SFu^vrf!9v9W$H%x~YQ{PDc$%*It)Tq-PJP%{?mo^1A-VSl16`Y?7z$E(rIs@BA^
z^69fuBMUif5(F-wF*lFumTr%`Uw`lVNI+rb$|4tMyR`*1#VAsBK^VjVLb=b^yo4Bm
z-5nT?!NB&+8WVrvXWIOHCXospe|Y%X>^r~;0&($=+@czPJ4gxNTJ<%Y@@wzbl6(WP
zc?+Z6v|4m6+{~urt`0^AdhB!qAYHo_^eOFR-U{P@21fbt<JQ4OGH-+4HydxFOljay
z$OQ`t$eWZj1qD#}KEI{AL!6-Z%>V=Vn>{|9F)p?Kx$<BYfX0|X;Q}l!#9VyQ3t;7y
zkR}RMY}J;XPZkFrHppz9OaaIhxuz5e`BK%NUG?XjCaI1hB1*B>c?V}n9Gcg~ob%4l
z6=-%5hSU3|uk(-c(^tO$fBm^Wk@OrK)Zk&$?<vX^YTOLIn3oGiUgoo<AN~L&W<j<f
zsgkq|54V0uFyq-X@fO5Rk>d?({TePe3R*hnuAa_;^g?41Ltw|5O-wz$H)>+CF6MNv
zz5AQhy48-B9`lV8Re_fU-X_<DiAacfIG0TTCk&#=!p1a>g6s&87&L?1fsEA>LDKV2
znJWy!0xbnBaSU$UX*WFKB&649EyTDwG=3U6*cuV8j*uBqaFHhclLjfHu`ChKpjL9<
zEyYQFNqnAY%sS4OqMoXzNFLDtQWG$_Q(vZUjpl7h#rG|S>&!INq++48d`oJH$Uhp3
znVa?ae|4SjI`1_Pz0K{+n`7sj6|K6!lla_Q`Hs%y>@Utc^+zB1^C9uo!krf>pMj$B
z{}^;*ikt7bi(EA4usI-$o9pu+N5N;j^^ecm?eqAP=63<dLG!KYmnhhni=N6Gz^TuG
zSQ}sh#s!QKz-Ynx06t}R@vdX9A?J^7uv^%WlbwJY0?3RSmzik(Fd2G#?pl`xYadN{
zN+J3vA^MH7Tjfr@c67GnFs?>c?}s3?71YjDV8!Uo>m|d{bhzq4d&|40;zBR18nr)$
z1${0O2Oz!shFhd^-H{(a9aj8o$*Y<<Fa_nrNB!Z|nLbESjLD4u3*1ORV6XZv?WtdD
zZ0UZFSw8!kaSW<3zI5^?XrQ$t+E!^=2&6Au>GB20vSId;007)Vy5YP`D6`KpMQg{V
z?Kg=zLHQ!mcLs{<L_dvxxt$p?(-#d&uMIKb<z803J^+d$eR?s92fMg%kJ*UYXKB8$
z+J?C$x2FwQG&svNFf^IZ4>A4al~U0du~HHL!urj7hBqjiC>?scIw$6gb3Iyu!?(Q$
zPm;yA8rFZ8?VCgc5|&XHDm120`nu<^-1O$#a^GnwQ9Lpm^UX+^=K>u+x{%Lp49<?0
za1qDrMPw$4><2ARj~<LNE&smq6%0%cQ-s1fdm!*+?$tMAP;E1$65q81%?j(z*C<Im
zx0`@`dBZ?2in|7}T@wy&qkfwM5mb9du|^<L`388>k&}v&KPU&bZd`-?>x7}aK&B>P
z0DU~rr}6rVeO-#D-eRAqg`VJ(j0z6dbm-<+-<oDF;AL<QrdaFTnfo6mw^m{!aN#r@
ze{gMOeu&Cl|Ll&3n~f)$FD<cFim~o!h!g%>JYtgljn&Mz{Ay#I=l8XGTu8p#49?6Z
zeg5b5-Km~$G5@=mK8KZbd>f`~x^US$?cThAXHji*PL?UYza8)oNSRICyZBn_zsU#=
z!GmA~tW_5z(%jR^j8g(>ry#!LpU?xK!R3==`J?K1476=wEh-n<vA3rbYsXcK=OtQ7
zabi?E$tRV%#uO(Z-<0(@2@?T(5Zd7pU^kO^fp>T=i>GultPx8>M0z^(j4W2AJ<0<i
zQj+Y0q>7ore=^m;8@`ruges$Kq`g?K#w((NkCnWlr+3EVs_#;=`P0PAh%qrN214=M
zVEi){Pa7;+T<NzQTYvHQ33unzkA%e)+li^K_b$NGt7tf(o4nBkC<3?VPOs5cXQS~>
zcwC8?NIV4cXmX7u@EJpEDaooy;z>55Y~BE2#!Np^Y!wZrM-HLdpNRq76#N%wr!CJ}
z?vch#+rLI%0ylD}Ty1=()OX6}RPbf<e?Z|yhgsAgdWRg9oyF6K_J$nh`)5*tyu50U
zG-pL*E8hsGhgm|B4kEmMS+yB08C3LfHlbC)HZ_JJ>Rp-0LRsy>B5sd*Cc+GAF&n?Q
zVmE!Xk0ktZ=<k01^sj5PzX{%F29-X1Qts#M)|!~tc7<$4NF|VU$6y4hu&|)`jsiKP
zwtZoCBx{mDR>VdkYMT}~8bLkS+Mv=&O0BTpBcr>55_vu}(5=2;jomU>>VR@+)=n~`
ztg(*9O_!B`RXHv-fpBQvKs5a;0E2UVq{33_QvZk0C8u!s?G;?Do4F*|JJY_Cy5&a9
zFjln`BfrIp+$jW>ugZK+Lq}ZRS1t)=XR`s3(Q`fw@@{`ak5k{Oz1$S*A3F^OjW6AI
zm+1m-v%4`jDUuO#J3U9^Pn(2Um*C&wM<$Qts?}&8_yYq>l4Ph~*RnAD(IZ!ES`CmB
zVH0;XIgJVpz~{f(jf&8YN=UE3p>5NQb>PU?8b$|&Nf;-kSdt7#+GECkfTfJoNHR1@
zEO&;DjG1svaVHJ@;$Ga$CdaQM;8!?*h^u}>u#v>emj}iWrkrYa$o+;vx3S)(*L(E#
zzRAXt2^3B-I$SJ)qlDKD-7g42&wdzoMYj9|u``*TS9`zf><GE)j|kvJE^cff%+AFS
zkOCbAgFi%k6=9Ov5+eipiz`D?J=YufKrpk^=Swidr376F+K3{yM)LVLm0xEzkZM;5
zJOI$NeW(sYzPRLtQ-_??NUe1JB_LD<kQC9z)VYnfF<W&oL&aUsc_qKX9a6>Gy-I3E
z`{RSReV_B+{wO1!qlMcN9+P(n7D`fzQxqmzNC)nbRMO!g#AHlSUum@m7OmHNy`3+3
z^$4yxLhnx+i9?NBo+$Uy8bMz^v5imUTS~Fw%r=#l#9)d)FL_dkPK(df3KYstN6Bi-
zivzn63IW-X#|2_c9|t>^S#%)Tn>b=9Je!7!mi?&^)AJR45a65vh|Q!N$d(mFC-vLJ
zw_Xq4(tYy8>2R|&Q);fVK)%J*6rXlvj?47lHk6ybOpm~7G2Z6}V{6cpL0d971U<f1
zOKCxDfpfVnHm^15uJ17$uV-g8>H9%0;({Bj32Peg2PwHA`u5-Z+#lPx<xxHxM*!a`
zQ%|&?^e>Thgq<01BSg(!%6{LasT$E~XMX7uB6qYR_UE1ap5d^|X*Uv(AkvrlCs%l^
z@?PZ45&=!SkB`4Nr}IaOtL<a)<yQuxGK{Rd0_n!t|6GY6qwa%Cbl38YjDIr9%j0QJ
zAmc@9d&#9+fgur$ieP4JH1iqKVXml+V`iUh%TRgqhJ<U`46MTeD1QnbN^pi-7?dMp
zv_?r;NFr(CY9Q(D6#7O*wxFg!G;qX{;iujvzVYz^aJhO$?he`G#wmkj+Wzbr5Dk9f
zNrjmi@z33LPe0|=-lbSvS-2txZ@i2_=e7nvh73P*J0CFht9Y-`${F@4Rp38E4!8>E
zXfQLK$rEzpz}#waa07(L89w8?$S2q<8<o4?TZKaNWfZeT_jXm0b4je)eSZ9Do`Q!7
z&m35$=##dpC#wX-L)fY@0eC7*C?|<Pv~qeAueM9CJGF;WQFo&=dGPtRa&flr)-7u?
zc$DATVXxs^Nw;q84p-5bWOt?Gxs%$UQpxPT`xaHsowhMXG*Nv5sJ@|Q?IqlA{tTqd
z#y=$qlbxmqENtH57_2edS>-LAzsv(uUe|XoA(=G~M_lR3zUN|4`r|`*6~6GHJOeF!
z08;QonyXl`!n!S1(G6<L{vlAvW{7%0GK?%qQxuy0RmNtGA?`z9>Fo|uq2iVX$U*Fw
zefW)-2x>;P0Z7&KTb!f&|6wIw2HgfhFb3c7;(dnZ)s2;cx=5)@$_NsyVvuZ%x;v7k
zpe*f*rNWr%Q?>t>Cy<i&M~Qj%>tc3Gf?fI-6WVpl^h^278e6`N3|xuj`{ROsKi(WL
zF}*kbJDxPWdRYG>=M)<#-t*{BvR}^48?)+{g)Wy&e^LYf#{1=z8oK}0d{LqJ@z^}L
zIJG(a?_meqw2o4>9&ndFj0LA9u#OjOBG<29JlNnYz?J<s;^=G{8^!@tMA<)hzz6KX
z5a9HNd`pe5=44ZY_jxD*p}9)9hxiWV5S!h%iPzt5EvLya0?aQeGNsHzXJ9riGHBnz
zVqwLMt<j(I(v0mAe9K2JTn{Z-@dNg%g6L8_DuU3;`o%PGap_kCT%j9|xZH+%#CRXi
z)n;TRReeCKE2Qhm8UZpRX-Y10@j*auTmdu%c?r;BxHI0~Q^ScdL>Z;CGZ9x;>#SwE
z?kGmx#<cne(ILZYbo=ts>*Pgxo9Cmylf-dG{f$3+h`&K!(lfi7dR#mXj!4B+0cT9+
z4DW#vA=mo2Qcu6?m|>p6k!Pt)TP@+Y`8xN<Ea|5g-TlR6hoQYw`&v#Fy*wtZg66g_
zFI|5KAt+9yu9c8-gRme1O-D0PdZDN;0_zB7BWStYdWqbL5#8!XR^&AX$p$iV7CwRc
zTkDb=r&aLM!IARy%<uH2dUMI43)c*aye~rc9X!eKxTs1e2hv-$g?73%)f6j2Lc$Wk
zyk~(kggYe$_zK)M>OMUX0p-8OAk0aY_I^o{o>aHtlO{jTQ@t++aeE{L85kN$1CJCy
z*)^T=rAHRr!#=y=6na1!?i@LHcFI+*THFMJl416<ar`eGIjl)<U&NHnvML1FZDm2?
zDc9!S0(0P#SoEX+=TkT<qZ@RNN+}D_dsN|)_XoGA(ZX1t=GirO_-(h5JGBz)_r6-X
z8<ih+NCYrpWouVU;NsqVUYJVtgZZ?^Xt0lTxyW8N`sI`vvV5IH2Hn0K+&B083p_a0
z+@v1lWS^vgG5Y1pIJ|B^o`EB(J+txKR<Mzxb`%t_7kHxUgN7b(fg63M@AzY8G(hGt
z{=@IemY7JN#<ak=%3Ec@??aWhKVsT@ir$s^!6<qfo7^?VfO|pc?i5K&u2!)Y9ixYl
zQB~8-1C)sVUy?a$ZHk27I3sEW>@Y^u13o8%b6dv9xacYybmWTXofB&pW~<BSz~W33
zKZVKk{5nbO1wAY6PbPk{@(;p&$-C@PXe%cZ0(ex1<ZY0AOhG{d4|%mS4UilhvtkIB
zDu-oGR6UDBmkS?t^@caoZX#tF!20SMd41{eH8e`qIX(n#s5O3Ja&z=$4x5U;o}}C?
zpBRV^0x1zWAsoi?ToTrff80&0XYGAbK(fHdmm@GT@{{e}@nHbYoMdeRW-mO)S2}Wt
z80FNcIKk5y`_!4diLVt1kp+w~^5=v?%NK-!N3V5(#p1jSEuOkfHW^a;37g&-Y;~Z!
zr|1#h!|8Fy)I@jNK9agRf+#2FY-a#DW6(}fV1y{!`4Zaim0n>ba=n}YjbmfR*BD8r
zK%zvlGSCfl;6I%^``qXik^_mU1t6Nl_wWwlh`|~x;`xB#(ZjvhVDo5`dh(nv&J>&%
zv_|!dCIv<?Y-ps@{bvP{_*aK1hoPF6<drn{UN$H&JfM)|sdw)AZxF?>*~@DZ?$TL1
zQ3g38z(!rIdaNW!==_N_je_P!eX^wxwW`1fW48<oN`^>rge7^JM~xm@CfWDozt<E5
zmnm82vC{G3$Z>i9&G|(jFe%kF_CUz8vwTekrDwH%1sY)GXKNMVqYFQ5(7X`+iclFy
zh)^c~!+2uT63RPr46(6;@IUm<cta-R$i9^3UscrAXl5QX29@%_S^YD)#T4}**k>OR
z78r=WxNHhRJNe?>J>kZ=b0kpB;c1wdH-8;J8U$sl9(P%MwX0odh4nFbZ=d9W6!%`b
z>Aj|jZA79tq4n%o9fS)H=m<RkZ&l>u%Y;Bmrj3t-d%|Os<UAnT)W5#f*O?o`%=WiR
z$VG3~J}R@U9)w!0-x^7`T<1iTZ8~pk@FL2X@=(&7i__O2o9i@S&p_pxSItREl4YJ3
zbswL)w{|NM&Tmk!WZ0BMl{cGobx#wT-cqPwYsQsMd@dIgpf!(C%8b^$W4HlPGbPx`
z7w&OmMfpq$QU&TZN`^^sT=*3Pg<eO6-s(^mswMEt+5?lqO8E;4X6bi@kTBT^2}1wW
zll;jXoE2PJTN|q<i8erT{me>~!h;-^o_PQ(0LKTW-qav$#ItfqzR|;d>EO15jnjUA
z?Tn^aDF+=ao<FgkK~o8pyFLQKTxu$wYYv1~A0xyv94Fu35lw(<S2eep!npCMEYpO3
z;=bXb9!zqSOuKFcfBW3REU@%uM}kJTwmaMa<aa;=ejTJ2DlTGmC%2`1u}T1atQKws
zw7jLD9u4-t8D0!({{l~jE7jol{@-A?O{Huet{z57jm!eit$7(;5uQKaoo*9%6V^BE
z8%5L*sRg%Fb(?=LY)AB;KfsO^?p&a2@Gnc}SARe?80oTuVA9S>*0Yj-s|Dve?;0=H
zNYq&Be&me&7o#a3Y7<5j)C$2;VmBzEY7y<O#p6KIgmWF_w%yYzo{_`au_zn;=ONu=
zVztp}f!RKtIJNR6G5g$rjcLCgGIoEmt-JGntvs_OjGq3)TB6BW@ZpUg{<@@>oeS2c
z+5W0%A0)#@M8)oCVTc1E*R)8fuKREM%0eVuVU|AF^szEmq<F@c!fGCN2M2JL0D^?g
z5Ax&dcD7y1tz)%c*pz}z$~F;W3J!8RDOYbqtsi!1+{K@J;8v3ffR(E{O0G~Kj?-Le
zl&&V@QeF2V1i@!QTWq;O=u1Bkx!RUIPRP5AK%$qO3ZkpFxqtuiAhP-PW3V>zpwztB
zDXL>{q5*&ESYgu%pwizqtJmARcVxgRU9Hyqwr-A{Dyam0+VYLNEc6;KWLfZUOfaSk
zLtTf7i4C8Yez#jm45Nz3xXC|SwST;ew?$L#f=e+$Q-Z5F9s?#N6E*CK?6!DZ>w8*7
zJzpz@kgTf7`w%zV|BtKlfTqI#|Nk9U#-)%Q$|if0k%p}7k)54QM994q%8YB}63Wcp
zE31%eHn{ePiwH&b`oC|V@Avcj{r>0N<8(SY$2q;d->>m}KAuf;O}xHY>@34(_>fw$
zcgwlQ*FHKduV78<P)SgZN7_6PT?|t7Z5kiko8W3b%t?PqA8yib3IjPzqyU!Zt-Hj}
zNQ<N_ED?sh5)06|;R~T~uE!H2=^0%`%lQrXk^SwT%6tL!7H5h5Yr$ISCP0MU(bwk-
ze>U?|53IcmycBS>{sD%yzvQzFG?(j!q~3P*w5z-7ege%p+`i3uB9O#JR4^Zy`8ivE
zN%hE#9z5S`!G=VvDtcw$XM!EK9g<|{zBrtWex3agdQv*u8G3Lf)Q9l8%IFZVCg2GA
z9dG>k(P}UrjaZ6lP|bB8+6F^r3gx3;)E45LpLq!56acFHt3M&+7!ZI_VvLktfP4b1
zS(w+a|5u6kS1%CVc~$Q`C#SeaFW8!$Tu7VRq)zzr7&z*gbvG1*foPm0kqWK&2ZBd|
zQjd{YeY|t%1o}y(_uTpG9*x47lGm%Y--hGTB)0OP_=I=-H2unN!De^3l_PB&P(6}G
z+;R10%ULZ(vg3k`pufF;6qdB(Kb0)ElqKZl!@9rY2N(<&$Xag8$}I)AQpl$FiEItj
zc~y}we*I>APCXWx3m!|mxK`D<uj@3g<`)44SmQmp$`qw^4`5K6%98M!w2!><qk%Q4
zxkGnC%2bv&{2_B!2P{7g)Da1rOuk#ve|Fc7#{c}@3ooCyZmmip=XiTYT}pxhjikLc
zyCS1mKkkheldf^e3#nYg9~hgSduEL3dJUo(AscHg?s<O#`@MDQy#FXMf|@_ykIA>c
zNlf8aHQO^*t2)=&OcsVZL}@J$mbBj}LD>K;9oO5mY!&tN1e7yXF1iD_;8kiIyJ6YH
z>KPN~0fvl;(ZKo9B0r-6faCxf3N!%wlo;x(-!k{q7*g5}CMoaiGY`C~#LWaD$6eMS
zq>oV1O)C*@H$QO?tZ*T4D}xLHrsXDq$;fF2AN`cQBp_320y8`h-hac0?DN{y8C3?T
z#x|k=VKU^C)g1luA`8b^esSR&@}UcGL}75i;pEw=AOV?5uFHGDU9W8)=$>Rk_v&{u
zyJsHKEO$~RPW_=h!~q7=MYHVtqKtFEyKI({gImLUZGm+y!2J@Sr%ppz^Cx?=J%p$M
z;AkeOH-QlqeERSg-i;KqRcN6VB$kD-BwQQ7&`d=D@`4W|^Gbs<;{)AsAk9zI1V&V3
zhfU!c6Dj^w1wQ!Riv3PGsm7-U7V|V=m{%Q<{t27juK;iZRRpWtfhU;Nmm_pW1H0^=
z&f8nQD;wZChi5^_n>SQ<AI=i-PNp@7C9V0CjlQ>%mHYNud(J*vF@9YIucA~BK-EN3
zGtwA06HWY8T-B-0oV`Cs%a_yd0j-_};CM1T^*oU$`v5xHda%|AuMzW-CIWco*vtjs
zsXr<hyI#P#@0S5C!jEj>&LJ{3>^4+dwt}*a&h<-vR2U1Fa@eW-<oUhtU{XC_S+d~@
zI*(c!-|J;DF0zi7a3N&~JJABLOKb=(QW4$ru{sQu<NDbno#53A9XId6z(4JVth1ZQ
ziI$k+kA5apW4HXux2oK_t(AWs<3(%k8?-?(&l91q61m?Wm&fe2Dk`zG_d?RG7-iFe
zpxi~2i3#q?Q=)1rW=lMrD8*8B<N7FZAW1iYiAOod&)?s!NNjhjpVl!e@jE}6FYF?r
z(;$8(g}xV7QHK-<&A`*`yU%2K|9Y+JG_o~5yVJkf=Z`%uP}D_YDOO%@0)G{|zCT;C
zrKw!7kuK;nq`#?7eOr%dQUG-H5)fE%eO0MY3w8CgU!T=XEraJ;nKKrYZdjq`h1CzT
za*jjB$9)6d|2r-5{*uIg4oIGI>n}3=z1EhRXTi)k=(8&349rTZoc@$&wFMSE&h_`<
zq_~XbbZB-5@fkn@1xUV%04e=nQngaZ3j#PVsh0?RzW$I5%^F&6yyE*RUzY??*?y7x
z+y#w<Ka_8$*4`*HI@64+bTP|bx-WUjRc09=b~}a@A2xP)0sd0?9#i3X6YpdI$g6{6
zPm~hh7O<f|Lj*=4MNL`(x7h-K?M|THeb0Vw3+Y3De@EuKZlU?lRj(?SB4W(XdWOt3
zjpm(=-qJaHPp2zx%)O0|OV}`!Xw|RMi1(F>2r2*p;RfP2Oo&LtOxF5T=GE$Yb6Gj^
z8Y|!;{I1gg-c*{EwH_*1Pw_oo<Iz|KW`@}x{<OeAFSh1W{LPO!O73RrWYmzRiyJJv
zR{5FjoC4&1F6H;p(Q+5u1_kP8Ap2$49N2MF&$gn#bX1(3WtLvl7)!oQ<6PP;pWjRG
zE(mXBUBw=Sj~KQly4jtke6IDXnv9{?UmE<xQ|6~ef)p~44iAo-UMQiXqmyiNvrMB<
zF>4cH)SzGMd6t9P@<j-+?AbTp+7h?gzz06+aPid5^gb6Gc$0{aM;d_W?+&IJb*qr)
zqR;V10m%iJ7(*}jg$qGIiY(k0$|rI6jxBkY#QiMQA||@Kv|YZ={`MU)+)%+U5dou&
zLlA7EuXq}<7QyE7l8gp8{erpxo59b(N>!~#lg?{vSn<~2hmC3^cz$XB1Zk)M8OuX%
zwvVyXI)(q)r^OWNOcN#ZS8zA!j8ja_n=9KADmt*R^}Dz3)Z41i{E)=+-E4;P2f})+
z?VvY)6Ih&AZ2n49fVwyi@EQO(dL580XGoq~;Vn+Pk37G>4mo@D%gOriw`LEGR3GdW
zsOi821zd)JNe2LSgqIf3M!mhgQN8wIguC!yds$Xlxd+gjyue=qL@$sX6D~!t<^_9<
z=q?zz7r|HXad~yW)(R*N?9C#cUZ4_Wk-?wZq$z+L#Va6-3Lr#dfFLxkn-K`ASYyde
z9N?Gt4uKgc+qVsOJ9PN=Yy$!9eFw1{Vf>bY7e1g{-1g?2t_N*A_D>FNb4l7ldHLGI
zy9nsOh~p~ib6SuOh^p$|QX@HPbC8uAa$@lEiEao~zWlpSPIAcGGUMR4jZo6<FH?2x
zy#u~qh-Di+2EjHYuCd&MLdUVKMGwRT5lPlRO?tyoPMSU1$XE}^Rw~kHU*a!u$kv<A
z><|PXIP`qi&03#Z<S?`xuI`Zb^Y~qB2%KIV8xMY5_M$s(1*^La1GnH#6Dc*{F>;-_
z68AO{V8AJ43w<x0><Q<^$z5$miUj!yDVB^|1#NgwB_zXH6?7|!TWhO*923T2e08GL
z{;I;w;??kr5)A9rCWx4++VpKiI%~4GyR@B2f2OzlpaOGx{TQ)V)$McG2rZkSmng}j
zj&o)zfyBA=>dlwJoe?&JPvxSPNVKxfoA245fJTC^7O5|R4M4!9B$o3Zxh~tL(GEfi
zMNVI+^b%T&3kL+_sd?9uEvEzSP9TIzpcz@Umg|;@KH^hxIC3jD{0mW%9_^SKn{NI$
zy=S5rlDtRNbRe^P1IHh9r@rCSl#cjaQjWwxc}jOB&D{vf&6DF0Iq-G{g%LIxq4>zt
zPlQebq4W9xWY0Hl28YbL{Jcr(T{p?m>z>&bI?vMfJbo&ONhE%#?!sNePsCk9B`%x}
zYuUko3kD(`y6&UK-Hc(0>vqx^gmVbycmUaNC3MkLC^b#H07ypV2MG(dSOeAse&Wki
zf_tRQikD)$pG9wZJq77IfLsfZqLi_AKC?*d`zP!`;e4mOAZPKT2Z5Sfn2O{R%eElW
z-R;0qLGbsS7Ns3p0>?-&T2K0Q6<(gG?$>a0Q5HBiYFra#_Pq<-^qxR#E~j^NGWP9S
z<GGLP;^3BS<_8&|)vfL*(AU1z1dYb~4UpxdR#~{wMuaJ0>Y!nqg1m{qO1JllL28JE
z<jFgF^TavH<Y0c_9G3ZN3oEB4QUI1syCb^}qO+nHqgK&MJq5nqLfB$l2(=1jtfzD|
z!NW$LDBkGdCEAr<HhJ#QJ~^@AQO3#Fl%Hu6?+Qd!$G?uc2Zkb(qIWaL%#0b0S7@o}
z7zTTUbt%1EGAARekd*?f8G?hlE*^g<@)+GLQ!E3<Epm*zL&ldnd8ipM{?L^RUhaJ#
zvP_GSKF(H}xyPaX_8%lNXzTB*pb}6CVtV9MJOx^&Va~nniBuR$8|>3`6wus@ixO|g
zpo;vpmzb(w@ut60{hdfB0x6-ywA|=GzwHW_fQ&PQ;;W&Zmf&%{arrz47$f^@=6D%c
zskO2HjKmUSb+8hEl)skm#(q)XpB<deZ+voQZn^)RpXBzbqg1ujqxOiJx;(|B{Y6Mz
zk0M8%tNOpHM%iZ})J!tCBU%-z+P3#b{t8#F0O3mHUF$Kz1Vwd)CFv1Z)Vl5;9_;OA
zVr$np2EACg8rOYy0f$`VB=U^<)KmGi8t9bC`>xwxgP$Ns;Gz;>ehlK!HsS#L1l)i>
z9^Qkc6nrHx>%i>Taa(SwTJB;WL6>a&PHNQk<P-XOiUac!#}{2OZaJhPBjPhVrPw9H
zCM`B;OZP)>IP@@NJ#k9&aZ{PMrcPp31Wytdso4E@Ux?9mdDoFXwEw)q+A8CtMdAAI
z4xY}dLSldI8cipKRMd`HQ}gz$c3V`1%srX}LzY`iU#7@(;y#vM|1iH_>qn#$moiNX
zmK|-v9lb=@^1`R>i#{j1)>sFCCL!&x-)skxdfJEPGbVoWWeo$PKyFY4l^u7NA>+M0
z_XX5#tZBmE>%9Giclb$ogF+CEJj*APv1aL=lpiXrCJ#=(a2&Yb?6#ry@}VCqjjN@W
zNYwx{YTWy>vVl)hMu_lkaM!g2t4N7H?s6GlT7hEwDUAiJ10n4iAG${yWkrU{7AiwT
zidaSKeou?bw}yC{CXDk(Zv9S-mSDv>kQzr^1>XMy#^MW4-R$soy1XapW2zt~<BJIG
zeK;?RO1eHz33MIP|CBzwW?SqttexnF1dkR}YrKhvcF6{4L{XKHfNnywaUvg?gi|?f
z5g+2O29mx)kr1nYY+zi)w9l0Ip5LffB}D9@NTQGUY!>Yxw;_Z53kYNoWJ4jI6tW@j
zpo4kb+DFhV-}L)Cz5^6Qr>Zrr%en4%W<P$Fc~S24ctIxZX0yo?vBz*lv!x%98+p8X
zPPNla8Z>7vyJc&HbEkY@7wR(W5_KaJ_dgPWg$X~J<yz2r;L{Izhpax9pZz{t82lbB
zoup4lF9UBU_<-hO=(TbWpo0_Cp<nqgZr^<3CZ1_iYhhb!Wd{~!wfF2A32U<}1LF2e
z1+`4WULxG;iGULu2K1vHHjI9E>ue;_A0JXHv-^}jaN6iCyk0jiPN!(<UPV3<VShGL
zS~q!VOel{QWu)d;N6fosm(k1%v{QDCrq5W@?7ZB_R!iFlrM!Ij3$6l{RbcXAwHi>5
zC?`5y@v<1;x2ik2#~|A{Z7K;Qp9fd>@`7yurjiUsinI;bTLP29$O-<<HYHiP6qf6D
zNRV(8VXl@?pBi~!&8I9gq#OK*`FeW;khRQ@z4G%K@bW3L7rondoX_DEVA;@U+_6sc
z)_hQGzt#C={4_(8&uZC+FQ*_<Y-id+iK8qIWdu%_Wi?{B50Z_RC~%uwgz@-89oBr@
z*kY6AvOHPHOdW`Sl^D*qyP@nNrkB6rtD>tc;iDbn?tQxv)4^M)@Au25>TlVfYFw63
zT%=@)GJE~~YFf#_N|ZPWFc?x{TK+C5fJar%RrUbJY?yUD97|j0fo*5)_tk%($|R!x
zJ5jvWzDrQQB|k<A-QN-c)G5_Z1|jx<$OfuA^=|||u}3X1cZHKgM*;>{#qV=S184OE
z;o|JAYjoLLMS+;KLpAUR)v0EJc~C@7uI9gndZy3P4xXP^jy^Yc@_i>9ved?yj9Iq_
zg6Z$#2-+weSDEUEpR%{k4%FDY3ZDGDyQx6+vM}kDqbq?k`0H-#UM#5cO6Ajh$3d0X
z!*p1Gw$K)_Fk`d49wnV5?(wM)|3#Kh1f-LRz80x41z%m9%|7%B3YiEBrS1kkA9)<B
zC03>N!<0lGJ751clIwHX;GH&VjYL*09ZI$gt~*fE^vAna+VqDvpB1`N@X5&k1T;5Y
zApt%$k$9i<I&f$l7yVYo@;7>Owrs<fU+te)34FKFtI8*@qy!wx(;?uR;P<Ck85N!e
z-)wowzeNPj8)+?5-wbe34;CVTUn0An-?5MN`5(EODfPaln>wsxCb-nG2`Kr#Z>pCM
zUH;Bm+5@1}BTu*omRvUX8Of;{NkPHZH|)tFTp!C8<)lVZ_WGHMXPc$WCL<w!aSF)%
z;&@YiA;B^Wt2^yw?Y*`sZ<{W5B0BM`YHPJrX{?*K`QL1b(Z`!No-+5fk?1%Y`Ms%g
z5Uj-Py%@N<8EeMcPw9ZN>d^+d5*rLe8q{<fJ&i#=rxV>(VunyQ78NtRb<4XZq2=t#
z5WYw6*Q{Yr%?JPRlJUYx+z^875|^oXY1_3Z#ebZq1Mt=tQXC~J%#0+TU!Bu*WE4$b
zf?VX#VoXChW7~<zXdw(FhFM~tV@wCqopoNM0z=M)BH)QC2Et=H;si`p^-11PsPqiy
z-&NQR`klQ@q<YQW_uIek7K6#~^<j2C+LhN!E2?cWKkbmP`~I(wbj-~2mlYXFA5!1u
zx?mZ<_;%R%zc%|DK-tbwG^QovPFol?YjbhkYWKcmb;!uSrQEuI%GT>@MHStH>e&Nk
zrN^5N$Gq`R><JUQ%r9+^H<jlI0dqmy%SSr+d*E`bph|2Rtllyi0!iyi$8g+eR$->3
zs1}&i0rxbs{Gd41W%C03-v;5`2y-+ZbPC|kBabFbQ<&4S6;=hNr3NKm=`{Ek)SZ-O
zX_=TuuNN;O3k-vG)DyyabCQ3+v6VY4aCw=lw8FPleMTZ<GawFaikHjPdM4U?n_}hn
zrM|_!+Sij!hkP6AHJY<me*I#y^6XJCr0F5`D7?dazxE600_vG{g!f7b=L{4uLGLwX
zzEyd=Og-lfx`#*$ctw5?tCz|L0KN8zHUVz^Awa`?H|HsapdbF`Bw^m(>o=5_@RD(!
zRc(PTJt@V@$1il`x7@c=u*R@r_?>1kqZh}5e6@Cn-`t+0N-ZyY#-YB^8rR<GK5tL&
zTs@W5*1F(5+1}exyL^edy|>PSQinuK`FIb(;3X0cB^8c9MYvluckeI*lRytNeSq{c
z$H$+F;f;npv806%%GmEG%Znh&WMIdgMMgPhwdq`}6qyap6{tj#Qlbhf3d+Z04*&)H
zQKJz3(S)rRDDE2HEy@X|!0vX+7?AmrR&b*M*-1>s02zX4y&Q(M!_FC(1l%e5;?djB
zkJgaOi4k8<q2#bTDai)jC*8k2tFsyy(OvjKokT<>u$Qfh=Prg+QSe7{ETOjVopnD4
zk0le^NA2ms|Dvp8R^;tPj4v?A9G69y>e^`5xZ)Wy$HzVR(+l7<z$unFveU2oq&Xd>
zMwF}!1SSv~T{AB6L-W}~nSBK?xI5BWARMPs@K*Mu@4LeIAL+cBw%MkgAb)|fDklYI
z=|7^32(e4-^D6<ddEWZqf{yOeZ{{yI2FgOa#&11eCDmRzd_sRg^yO!E+9E0_q)E|w
z?bl9M|A48Q0*P$X<iK1d^>_r94u3rm-S_j$H8~RM$cWrQdOunK1`#s9R6iEX)mmL0
zYcj*+B7vwm+~`d{)_t}3zN2@3->>mTHw&GI1S~hQV_d#M7V*1#qEQIL`j|~{(nwaW
z?v{rD;Pu~lM8r#)Gp5<utH&V$nIPeGbZVnGO_>0CB-V|)4X^VOEC$6$cq5VKmidy-
z_$Nd@zR{K!br9)4FIKpdo%(h6MaGAGnTT(`3AD)MsYk4FeR*3JKOFg<)LeUtHSTMm
z>KWB5a$DPaavc?{5eYx}g=_s<Mv~$>XK&ETw-)!BB+Mv94wp`$GA`oD{`e$$I-xJy
zXQ`x3G#jI!806sA_OtSPP5VrssA$vSB~l3}<$V&m3!>yWd*bUPxiJ*Wi}Y|o(NW07
zC9%keqyos@C^5Kbud2hGTpBOrE(m!Ojm!lp+viJVbLh@Fd`d{vXG*-m-6sjp?*GYu
zMDvfoFt?8Yn=ZxuhmA^q)bT}LU=LD~t^5tq61iEZC>j4U@U3I$zK9^0dA$2)V+u;H
zzpUQewUqb!&Hxw!&~Ra3)N5Jv&=vU}ca?klAtd)pQsLL88MCV8pHES0u;g>?3Vr;S
zC;;&n3Q-~;qkU%&4qLxANV9!xeDHy5&yVDE^&kUKZs}4SWmr|l8`uF;RSxOx2#Mjk
zGrUGU>;(A+=e~CGSoEV$<Mr~kTjeq;<;v|7vH}GzFAVwL7R9CV5uYfG2nXvg!uE&D
z+ib1s!d|RF@r)SOxSJ!HI_uigNhej>JBgx3tVC#_1yYxJSejDd<}E~7aZ&E}_h*S)
zSo9<{q|5x_UH@%2V&3|wAHtLS4c$4@8lh!OI>PG=vU1Gj{6iV!yhk)^Wpnn;W`41;
z^|W!8oN|6sBnczaAadAl0jYbY{oa{Pd712(34YJcrh4c(*qNOe<(jhSoB|pbqGo5_
ze$C!|!>rJ1N5d9q<79E8<tEmZJ9I?p;mD#{D~{GSBAKU2c7cZuAj)!rROIvFRvuS4
zzQK@-<_eLrM7B@bnBwF@-cVHcdfLw8ZJc7Yx8dO0q|`aSWP11XCXMkrHwC_q?xw3$
zB5Sp4^*{Qp&s{f`^ZNOu)o;?vn)H@Qv`vS71W7V|6kVbKrvHX(g4ioNp>$wYz0e+(
zwp@~#Bm&8L*+~pQAiN*bD)ebZ38}4Q5`0<bZK3&j%)6sgALPv(`60Ki3z#{O?@C6n
z>N-M>2`^tuz~7DN1*sSJ*<m+xC-cFp_eBkQIUW)NWrfO0#Wyc`v$wzhw4Kf$<hDv(
z1}=-2g|7ccj>M#AV(<Be8^w~ElZe0ix$NWIv4-QpS;n2E;)TDbYT`_YJNpIP3P;zs
z@s<6Q@f1sa^3rwSCfDkVCkvUL^U6oBPshR0y6AUAiFXda(fn&Cq3tAzz2|kvkDk5E
z?$T6ZRSOk4Z3$9cM|%Oa6i4;1lxso*9Sf{VI@MFEav)t|l;8Lr{isWXq@`V^<70)d
z4o3OehWV<_6ah5JbMy=wLjI#4Dsr~8X|`_xGd|7WLW9@`fdX7v+_hG!y+ttguYbb$
z$t^2he4piDvKg@ztCC)i&_UETLNS!z`n#ZDvQQ@v0jnTF+jnp>T<7tUfpoWaO*7C9
zl;}O}V93xb+|Mk((NogxPjN9Lqv0o5qjKkSq?y#})GFi1c!Nm6gQ_HCfqqBoUQf(>
zge)ir+fD)dJvY0XHcPo$HlYreI@?1kt`%reT6%{*?~FBJsdorusG`EJ;XoMX$*#pP
zPTBzZ0R!*1x#r=qp;4qkiSK`T4Z(4S$%wxq!8mHhWMkYIRAm)Z#^O`io%|c&Sa><k
z>?@|+j8wPm+TB|`==K-q)c?8!!QScb=%KhCpjEoE^Pow~q6jQy|DE5I7d4F2s`P2~
z72ykErrJ3H8X5|G@U(E>Jv{v%5w`+JRq*PIk4zAvf2Oa_3Kw`CUKY}*h0+Sc`-h%6
z-^LUF3E~E!mu%>s{3$X#b#(~q@U=@o=*ZTa(fUtLA0e;0i6<yX)!Um^B6w2tk4BAq
zWje*??r=94ctM`Yozq+XmT4xg{NnjrHMz#(zEwh5^b%0NKNN^qm@8e@WB5bx2!cUu
z>!xJc&1pr6zzL3C;r3qX@Uzwb8GP-&8gfqs3gvX{{Gw!Ef`f+Ir*snZ*BWFJMdYGT
zXl6=WDODQAWlZj?uDD=NQF`nen_k?tfEe#zH?L%~qLQY?i7^Z?o)esHRqE72UK+|@
z&|MN&N#4k7pnB!;`p1QC2tMKKLy154h)v(Q--Hb6&S_-q=sJql4_nCFF@!lp@6fEO
zy^Ti*v4$t1m`Q&KcJ^*mN<%gY`KPUbOWpb~?wj_q{ZXFJ`|{xaOCrhffM+i%cW<bC
zp>}%l5=)riRS0ufQpj*MLZ>XOjW78sTl|tyDQzEa);hv6`kH(D5^&})rdbr5_lhBS
z75ZJWt&Jc7fIObbX$iudJQZ4=Ts}Eh5>2CZ$(<H|+I2`19h$sFe48b}vnGmdGBsXI
z!FO?xlU2sf7?kNjL=zn8(R+<>e)EP#A|}Cw5;uqwP=^xzaWe3qPLjHigI+J9omJ_@
zvdL^WF&2ntmtcHqy~M{i5(Z0%_|*8X11(Ul6s<m24Xna3mmA3)bx|~13{D|Bxen&|
z&jUM|)wA^T^WO$Mz^*(7M(4Q}ScT*eKk{cZ?DF4kug3RzQkDARrv`p=j&azh+YlXH
ze}h)ibxbgOJA6A}g|*%Td*xFq^hU1vo0D1CHLm}Hzt6LUksUfx>OIlSf2dWX@MEoh
z>G~G^_fKG@)QNdo=)1Z^c^>(egl{5p!f}gmlG@I|M4_ND>W1YPf(I*jL%|x@{S-_d
z4c?LhXQ>QheIoQ^;~DjQ>e&kKZepLi$*J6XB0d(z@#|QZh!Odl52llF{Kh{f+*ruu
zuG-+gp;Dx#N_^DK_@DJdh6TP(_FG(DVeM|vB^ta@q7Da!cwW+1lTATY=5^<EXNYu0
zo9}ykvL$0cZMm}sp5G_;YAhuiW+u0I?mj<1-O=0bF`Z=R<#SS(8vfpAz7z$P<l;|5
zLcDyY{iPy50Mp6ANbLvVv=WvF>308IMSVNKg{;y{1vcY{6QJzbDK$F7R0%u(5CHC~
z>~B0^ImG<NjtP~d>9uxo-kvJJcyIC$ERqD=s5S#wb%Z|*&s1InwKoQaEdA;)RWDFt
z{z9+X3Q%QK{EpSuo6B7n9!K8S`P`%!w!qexPlfpjKcOFf4eip1_sgm#N%Wf>{hg*Q
z$6H4HN-sG7*|YS_t=$Go_M4N$$o{gg;<DCj7O>YZanT*Ufm8P(TWR!6vSwmGV2AL0
zm@0o5jME{P@?0rcNPZ?aL?tb2rHTCsXY@rb+kjOhZx=8tL$@*FQdhcsS3*YR0ufX^
z=A?Lf<e%u*XKu#YrhiAEpM|-8y;V0%2|Mk^&_mVxKM<!`^5QE4Su>Wo$?Ng0MYvc6
zE}*96O_bT^S=|MJ<UnR|MV@XvI{D+%r~Op;5MI9zwO3`C5*k-qh;)=V{2rw+BJXG8
zP;@fdz$4{3?-{(^xBo5v<VWV>l?#41XVY4u363h<P#aMB_!lEE5wB`(6D=WVJEr5l
zM{VJ4d+X)j%TIo>H;)E;G=n$HfP`aEQl(|kcA--$uvW)cG>|Za(g8Vj6pV7(wQD{p
z@?e)cZ_i>K4&dEPg%B)#iZ`u_9K^KAA^4N@Tk;r2R8$hlC+SfQ@arA<P;i~7m{(P1
zam)80jkN7n`|wufp_qT)Ew+d?j#)rfBSB{}R*J>MYSQVDtg@xUD+Pb&Lf)l*xCj>A
zki1Ei&tg3*Ji5}XGBe8o{ktJ+@)*|5zI9&^MUXC*y)8f(TBUeZg(er5g&KL)kn@rr
z!4OMp-G!aCct*k!Qso`a95QoLmKPCYQtR0aFjxy#@U{(d9pUmEQ82<E1~!y|cib4(
za8x}om45d$ZS~KH4y-yb^J*aC6U3E{bz+MP7hv7_)(&aoyGE%)KkJ_7u>$US;P`#E
zcP+8s=Ld@f-uG;+&ARopc<CeJ=<#O@$1%B?NjrIB_q+ZQ^}wltn~GNu+4c_Gh(}$d
z%nBE^&1a&`hvaLS(%6x7!Wz}YhkzcV0-``o#U-NKz*`P2>ALpU_>{YrEDOTBvj7t9
z1vQts4Bign1la&?`B|Cx$fe(~nm1e543@rqV<A97S4bY5)vHk7{I}naEVvKl9+p8g
z<8gt$0Qe4a)HH91-;K=qkj72Ldp_|RK4`@F?3De)cmM9ClTXGz=bty_{}*lrDx1NP
zooa5DrlR$NG2!3|_lyqZ<4;2i@v~=hI=<&!#307x*oFZ!6T9_bj0-p)+sh-5OXnAj
z01^88%QDbVXz>>csojGm?o?yZ9(^@yr<rqN-GeAu(R%{=aG>(afM5a2Os(`|`@(}u
zI~(2fJPh6yxY<G1Zy)1T>-<7ho9#f@%J{})UeFdGzPCSeY=vQdmbPin+fNPwM9YH1
zhi!9{4e?)|xcS9k1+UIYXI8KM!QP*#)E_~EX3@Cy?Sf8R*dX91^u?_D&n?o-1%if+
z5bXZgeRW;beySy66*U3`I_5G&zXamK>vQC_y~If;3d4VDBeAEqkPUPqlqV4rFG|bx
zr?6Aq<--wiMO1ybC3JYRymA+Lin)jlz7r+;=$16%6Oi$f@(CIrSeWhEi)M=(ee_Xv
zS`wgf4;_J1p6m(@3YJVY@%4ypRx9!v%=k-^@g<$I!zslhdy&2NrmhuXKyj{u2EKF?
zj(g&^&|%CuUma0WXOi_SMPg@|4fz}ZayUCJ67$z`bW!kGUDys2x6&_8(QUchrP=;b
zWDc}%KT?VQY2a5TKzyq@T>@Od90xE$Ucuy49J8p1mz!^pXm(XK>|PG%k_6ob_6jXv
zyHC@spIuzxEc#C_Rm_U$&E%_L@6#u#0v3jot%XNTY>9PRsV=7%d>Eb|xPVF+-g%u~
z|9K|*4Q=q@2X<Orvxjda|F2=GK>lepH|+;;TJ7at;o#P3$?(^$XR5~&x1T)WZHyv!
zDGc|0WS2+xMY7K_0f5bjxSEk$4>H61*8E-w^*yA_%T9aAG^fU-JAR#uS-%X1RrRQ+
z%Em~hdUt%z?Q%9ShoV1ByO5xPQ_@c-1Mm+%ij08N@@9NEji+4(0QJ|gPA>E6FWb1<
zxe@Uy(67!df~7IT+@fhSjJMv@3Dmo|SQ*mYu0y>I&4YF&!fnH*288$sYe3X!uLDy{
z%Ke?bCPc*uqX)SyK(|C}m^S~$@Vyi(g|}>nm_JMU(Acn3PR^S;SsK<o`84T+?_8{A
zMXV^tPoi^+%!7gtYb8l|Jz=oO84AB)FRPwSa-8+&SUjEQ6>4(GC)sL-1VOJV{bC}4
zE0IsN>+)l##Kv84Fs#5faxvJA9_=&A==*~LE)W^*$A&X%grJQseSMbHXEzs<r7Zk*
z-hPxl=AkXjbYY^Oh!`9CS>7%e5VE6E`c1<EI^O2kBl9`5k|PoeI9@>-kp6nD5K*U2
zOpWW-3w<kS$?J67+W>7QAOaNpYDI1{;CQ479;p)+&3_&#oo};`V3+l7=P7?)qkczp
zGSGDC&k(XdrvuL^8SDmKDjL6(VFx$h6{8j(0#^>Q_RKsx^6kG{lMnR7bu~Zp)b7L1
z<EH%S4+N&a+ivdAbNhPzZQ-lsr<kf_1}9~yJk%!L@auTGv+=V(Yx~L}D{ub3Tvu;h
ze7^f8bY~{Cogi)4TN>KS#11~BancO)W`n=DYh|ljF8JD6B&myKv+a54?g_8~gEp)_
zE1s;;+ycSJvbQMImoRbrl}pc}AHDkSYn-EkcJ$%TPPRI>g?j};0pAx5=Oa-9!_46?
zSU8rXfo_DAo&_a7a&ZEN1y6=<zhx6|r0o})7?xK1HY!;rNMfR<H~-{{lJd(wki@%?
zBjva?!2=@HUtgeMgXm2+Rqmg@+PsXJ8cP8--`_lo))k((_$^Zt&N<Pkd%_0K?ms%m
zD-&LCG**A<6Xo7Dx@vli1xogv-S@DnjP_oKvTCY1izH2FVX({{3pi*r+&<UCKM^~!
zJ?ZN^L8((#f>N38t~KNt6tts2&o6R3?Hf4(Kj*K8$i1$m+za*M^{BBx&bEQWcTK5a
z0d~Sf6GtJ(;9J;XO*>1BW+NN-yF2QdQ)n0ePAl+axBb3RNjhu9Ap6bxyAtn8sVm?&
z8N%^Ph7Z23*O#gwVPY2`TKz-|ue2z`?k`~aRY*99U3Ak_yQk9_IQr9_NdR~e-9anL
zbEErOI@p)!WQqPJ%9evtKs1p!h&Z3~O8Uz%xje1BVRy&CNfn-dU_dvHk_eC3c|f|R
z4~uwl#D`q8bC$syIM**8@lB!G)0atyh5H_9HQG5x1=`<#4sibuvsb&_ht>XTDf)+Z
ztt|zlBW6$T9k|D0rQCL{Ja3C{urUzki=*>HiH`APy@;c#HNC@60BiPxI+T)KDa5wt
zD&QS|o;@6G>zF;*vmOUE^w$my|2bVR&uj+LU>Uq{@Qe2kNSsd2pB@YyJG3Y&O=>7z
zzUrVtnOF11JHN!&Cqk&NLPx^h<;946m%N`vlF8bIk`@a>cCSw=e&?~fMVf>i+@#_2
z`0Gz~b3TZyc9l0dHzpxirKrH0Dvgvd6I-mtTjsej<?pbD&+^*s1q=^C1GHjFrn_r9
zXroweLyNI^PEq{ljX3RK!w9@;M9gI&Yl?`sS<adPV=-M`F9Rf_>otQ?iE0ferdFxQ
zP)|<G&&|n;7z<LK$@R6TNegrCQUyHVUCW7>ThyN|BLg(qFEs5xMC0<_gEJ8lAOqHP
zsOc?N<y2j>X0}VqUhK!L;7)vSrY8r?kcizYH<pZj$jdW*-tSu^M?Y(y@tV@MGC4ZW
zzNf)fUS6rVeS%UGqnV`g#PL)M^w^%X7D0{?+5H;$(7lT3a-paFay1^Id2gcdVO=_A
zM`niJtxI%*5+2BBSw>HCzwL|#h8`iSlniUkV+sZa3^_z;q0H)|NO5&els-7(KxM#Z
zH|o(1G3F?NTER?3!M>NW9nCZ$#*{|0O-!pDbpVT@ps`l@O>}Oc-P8-%cuU(cu#f1X
z82x1KHKA~q40i<b{?bqOlhx(_Jb=#eK<C0${P~TE^_}dT?@E&P!B1g^t#B#<WzqQC
zO9z~iZ-W`y)A>VT>lKQkl_k|+11t~e&8YpK2>{~mZ<golS4>UETbnGgBGVZZ+qmBh
zpWRs&!VM|8&iiS}xTcVaLT3Dr{HEXhA)L3rsu$3iT(GqI|I9-Yz&!MYtB8h#9kOV>
zw9?5t>hAc<S0tFM?fABGWmi<ad1F$gzAo1Lu8zc8v*PI_gA+H@;*NZmPpx>l0nk&)
z1~yvRzSpSKz^UUvoRNJxHgl<_Bs}wEEmwCK01VYYQd0Uva?$mqtJzaz&SuZ)py-qL
z=BOk%N(PRegyligYC5hlk?p-~xh{s3BKf$=NcU9qgdE6A)dG{uyL;uN>$Fb<lYjpu
zSv4e8;~c4o-UBkMuyxy2RMPue#<541URA`@b}B!G=NHqq*TA4@q*VIbP=Fb`P8_Ng
z)Nyj(TYti`Gdg>VVJ9T3#-s3{o)00d0H`<chNOL33S;qA*<5<Ju~~fMD)9L#?oU<{
z)`NVnn-u4UZj$99+WV#b*(BZ^SnORZdJzy)l~h^HCzS2BS2<nRE>;bv1Ej$%;p`G?
zg~1-T)@;&(mH~d!#DNw;Vv0+oZNmxm(JdBPyboV%Ub$hPV=t<T%2s)co*86B1ARA;
zFvDtK0mLWoMg&)^%d8;acUGgHS=acO^`fD1F!UE)Kf}i~mEUN(yB(&*s-Ce=N1NQK
z?(8|bIuv};VyWJUI=tO>*tk#hcM8DObdj(hk@|$hhB&te0DBy}8vN6voJ=&JCS;;F
z)xxn(e3;buY_{~s;kjyA5H-$*JM~0eO8i${iZGxXYkE*K2^DHiT`%4K;CIaZ(4c4g
z)^mcgMdnconNh*;yROOj)6UH!QKggl`1!XoV+JIb^%PDP>QC8#9hXfod0innbA@9t
zMJKTsisj%kzjoWoxY_w^TX|>sH~Y0?xI$ELko~9dDLFI#<YwLiFn-hq5safQFLv%J
zr)JwPpe=s-vZ4JBw0;CIBOPgA!?6^x5f@4!Ine#ej%#v6blt{|#Mn}mw%u~&HY={I
zy(S>agE6R~fXIDy&1Y)YOPo{ELuq)NsIYtDVUMGo?069!wj*TyuId$x+kJookbWj_
zDnm~02?4%uJNL$oUXy1@n(LM@9S;FPAYhLLLOVbV_%;Zfx))d*oMgw(mW6<)(e~D(
zlIehrT}RO1Ekytg%0*!Zqh95yCzR;eAxpiX&D$kz0~1lf2t;Z4L{rDk4_1$Jyth$@
zG{0CfdYEY%>!?w|6X-T!Stga=uFS$@{g3a`=t%FB2KgXp;lys3GjY;~as~wfLv%LY
zr@C=LRiu384ED+F27~fG3ya~pVM){8{M`dAMwczY|0MycVoyFHWftCxhN-KXs;c#B
zZkw(+Wr^7#sQB)7q$PY`CP001gn|$%YP|%BGMF7rx!L6p>-74!Zb`QNIT-5ZHU6i*
zvI2fu^>5dKD5dL>fl?Sr{7|;-;t`n|_K_CTTA`b&15OWgetud<tp|aoged?|^B~t{
zc?&%oSB9%;ld6WZ-#LYTVAqKUML4aa5w<ZN?+gEMt(JK7P;!&>H)?n`i|V(XtqA`;
zG)}}#w0^q$f`38Rf0o(+MMXi6VGtU-l~_WGX48$W$=Fh==`%jtW<PZ~wtljB_h-ek
zjH^_SWH)-v=CfKr%LHD%$L|MK@gt8dA=n}~SBBk>30b}ZrY|KBwA`PsvzKfbEJUb~
z;(E>E?o&x9@v$HE)+Agf_s@%d6{HnpcyXRMiUs)*7-dxaSP45@tM+T-r!O&Wh^`zC
zuk@Z$RMd&a6NCvUT@u(On3ZRSM!l-|ADYR6y5*v*+1%ng5Gq15bnOz#%W|DuXBRaI
zgani@5a^z_8y=W!&_Id#*)tb)<ntz6N>P3EP5_X!=MBA-hQ`JDea@PVKMW)kYkt&-
z(jk4%P8yHJ_k=|HrAovNL<%)(uS0TS!6-y<WN{|9+^=!+z@~63C<RJP1qHImqfeYR
z;lmT|dE}&c7<mNr)wC%<dH-g1L>N^Jz|ux_u#YX99irn7HhS~KhU%wZa*3jVCC8LM
zm+}S^EeLy)<D}xDEoCxAjj8n3Kq`<CN#K$}fvmQIw*j!c{I}nT871I+Lmg>=Vqw(|
z!CKSN{L}P@14`#*JxxNrgwqbBoFif;JYAqkstTEs>e+m?#hYJ70dAFC#19USpZ?eX
zSHmxFxG!%`?PRk#V(dxnVKQRQNrfK^X;Km$*O=L5Ghw@Nc{R-gCuDiiaN|4Q2}9_w
z_=UFPm&%Qsk~eZK{y$SrQ&Rwf46A84($IMDj8^yDuV#ehZCB-xv+(^|;3Si(dAPj@
zV+7KOSC<E>rl5iDw)Pa4GuCTSiB~Z#8U}A&d%WfP{<u<Hb23Z*1TCZFTaLVkc7gMm
zs4Ows+$&;$N~rmn!?2)%OYMvNZY(I>@x>j22PJt%M%(&k^iGHJBO|3BBkjFUrlhky
z@TQu8N~erF$D6v)975G%LbKt^>VK=Iq`g;aVJ>Uq9`5S0VR4X+m$<@C-+|$F6NtQo
zMuNqhz_Gc(=p(MUMP?RdwpsYo?FGKgHUNpZp7}(XuHMU1ShZ#3fO-TM_-vBs`0<m_
z>w!=EjuzPNRhO;itZi95lLyvBLD3e22JZn9U&>{R6UCwAst%<FZ)t&eje{>i*^es{
zoql>IjGa!s98>A`7*xu}@KcBdZS)S8_2h{VRiqcND*1ETE<AOMhj%-ScVY!R4e;xH
zE8pFg8l#LNvfM*0$HvdZuX}&Fy8B?yo(Ggh3EtP4qJRjFD!YZivDTy!QLQ2JX%85j
zI0tMH7yu8usH=EoiOE5unOiU`={tX#6S|T(X?^=b`xas5$cz#CM>3KLaQ2r!lRO55
zIyiW%h}L6*Ve5&(49pt4ocN%nv1`tz2Oqz&WH5p?$O+fW6&}r*sDF{xw;l#;dD{D%
zZ@6FI<Jof9Dzh*iEI+C{YIb-XI{h#t>xK$+B)X||Qb98#3~PC*NcmuHZ{hC>ZSMC8
zu<|$I54$|uP;CdzZMB<{{<*@h;XcHP{h^3PZQGT*$-UgIAHA6@_tH*fln=a*$5Wpi
zzcRn(W+t1S9Ht+m<+8UZIZ&a2GcAuX-+CO2e-JzJ_}K{ddhc3dbCsXv;1{|zLwIkx
zK3KP<kH^XJ;W~TswOX0r_E#LP58GP0TKe9NOWDirlHXgrZ+(sRhX5sSs@=YO*OuR8
zD<y_fvVk|qmk$2O{qxwXs7Br-lit@~^|S6Yyw=l|B`4#3c{1-0i2IRNJy?)}G~7)E
zBy;N?a1csa1ZYim<3I<(E%GRoMYfT|3sK9|-}Ty61Y}nl=P5-@B5K<}5TWs=IOy?5
z7L0K2p-(^vQ%A4g@I&o5y)oH+IVDQo$SwYv)R+6_paQFSB$<jt>*HO&Nb|S?MSEX+
zkDIW)gt__{MuKi_YcNPY^tG4|ZsAxnKts+Ov%r(0L|t^;5VT=O8W-{G%{7&Aca}L;
zv~cVdhAYKaXjF7L_>iXO#cA?@JObt1drlIql_p^ey-acOF4J(_r*=A!zK}+12hF?q
zQL(`AM+lP$Wg$kbcc~Imap6CE{13?u4KHqF@WT^e>e<RPd)c$MM;*qSuI#lq;pE0s
zqJv|NfV`*x@jc(e>0hVT!?YK(H#UkvsOh4`q~9jL%no;(|7w>45gS%<`a!KlWNwJ2
zBT)6m++y6MxwTrKb$aPk=4^?5?(doTYrX#w<&iUU&w?N{`kk@)BxttC%`(yp5ykDY
zu~08)Q9P1AHBuAreNx`~J7b^F=xy!0QJ1o(_#m~!&{%S?=;t$kv`9EB4~!JeZ#yht
zfaF;UQ_=yrkr*bEC3%(f#E3RrFqt?J`)%yU`9&|+#r7{xuFx||=*n<y#mO}G8*iUO
zd!D;`v8UJ!*D-V|TLbzpF*}<YYmskdMj#BD`+GC5HDH6$_}vBAuMz(%*}u5K!tA&q
z$+y<(YRq~qua{5XF_Dh+Z-8*s9&Ako|9;pFEaso7zXNaAw`n4s-GId}(O_p{BW$I7
zpHSa;L?neLxV3xVOZ8ipgpfTV)<J~Tj(8)iZcfVqBJ~DKXYc%0o}<7)#8I%6teo;|
zH0-d4vE*Uci)3HzW(b;IK}C#}%uH^R(}7BUtbtR)!pVnC1D`E!9?(c6YR>D3l=B_E
z1?6k5Bp++`@<TweSPQ%pQuQN;sLBgy*A0F$rEJAWKTV!~$SY_XXg2_`p?RROdJ8s3
zkL0EkX*W&$&P})GWaDtj^+@PHZGGfbZ~cqFI0c%lN?m~CI$tBP#ZTU|7{u4iKTG<H
z=GnU{KSL#PvB$WyR&_4%fA~oMlVDXe@3@D0clEK4deigmhYJK*Z2@=nqDFHB2{Psy
z-q+|PkNTABv^!EcLg`tkh9%cT-Cs050m^W*Y3h-H`G_h8&YWoY3mUm(bzZU`{nMFf
z_gb0qUT5U5zR*9kL4&zLRl~Q=ao7Q*4+U*`EX&qJ`={s;fk-ri=KIpj<gcvyIs;b>
z$X)WYt0=LM%K)5GwGPGRO`Ai|;mA#Z#G0sr&Q%YBS~%@ibWnn*E^%OAPv?ReN$iO<
z_<z#_zv1F?Apnte!rGbOGG_{WDS87ray#3=0n)r~Y{y;e2ksD@JFce%c0753r_9xS
zZgHz}>-R}}?`i2Du$9-F+Xq%oArpF(59ssNeibN_<|s9Yy{PjCtfV7C>%UeG?0C=V
z2@JUEa;VO3n*#mQq>Qbt8Q+)1IvJd8rTYy%qs0cLIl*~M%ML`{Bm%<Ulk4E+CnsHF
zvggM3Y)+NLkI({DmBZy#DwWVkHWf2|;TbqjK5^2vJffup^B5zd1;mce^L+IO4~_WR
zWHQ}J7;iR1AlRYjb@U-m)!VrEi20D>H2M(!DC3K|B)>7wC6QV{!dNVnpd|_Qf_Bo8
z(<b=q@vOgXdHS7RtN>k1PdEBDfS4mst2AeR5#8bH_!csHwqm0ED5{=@Ae#jWGQuXc
zrcr!FQ-uGI%wCCk)!N&i&4TgS)9CdNmo`<K#W#<hL;;NumuExWe@(|%<Q1i+R$4im
zTuf*@w=AZ~^nTp==F%(l?4u3esLXDX=#r1_;g^iI*z(tH&+dlq+wG?xDc5X1ydh&q
zsBQMQg3csBR0l1HSy*cQI$n%eVjWI!X43VT6(spn?&lWB(Ij4Wwl%#uw;6KuZr@e;
zxMw$0gLg5^E;;i4x_=4UTUAHuts4XnE*SiSXI(ze@DE(3&^u4ovPuAGogm$Gd~(6O
zQXKNU7^!9mwS`a6=pruNCd+k$?t3ul&V0Pqh~U_T;vM%Eb*&c~pF%!8i1A_DYrMxZ
zWyPklI?B6k&4Dy&yF@udG&)cZ*cp}^-ln6P*GPEP>X}&jq!TKL``_sLdaX-Bsy|o#
z4xGM-`MLCkk`A{n*h4ETM;W;R*FkjoT5kfyBirnGjD@W%un0#&oTlTH<42lqJ_$6Q
zIjwkCJ9}D3pch$u6D%*s8au2T`wHVH2>g$Q%n!XIdoe6QfY5cWCbq`XX_P_XHl3#p
z9Y^JOAHuC9&_-CT_u~Bs|C@ueeY=?ZP`gb%<VWG*7=X9uTlyGsCWaR7Bl~zga56|)
zVm!yGTD$2jDv3##QlG;fBJR9wPeCoBY6?*#6}_@V!ShPgj(q7HlmWuSp#=qhM_e6|
zh9agDi3)rkV?e`?)W4sspRZ5iRALqrGhRbw5BnRRr~YuCAnRDUKqza@N!lwCiGrU;
z>6!nX*6>cf15{uOWIg}5{U_MQ_t40IZ=Cp+%#sk9^RKn!gF!UQbnf6Cf|?)f!#LCP
z>Q<R#=YQ!rl%B1{pYk8?u%G@2Dd23f{l5iFc))VK`WKL$2&6g}CI-_h-q``+{nqyW
z!k%)&rsQrh>m>t?vl1x=gpWqi)DKxHq|Yly({h6abW68SW9R)w$%HKpIF`i5svMVZ
z1tyC%;Z=Sec5sno2=&TPGSyu9%s3_ckwa&;$2k$IOB7vHu(aCBKKw){rZpGoGw0h^
zH<fha%~x;mOrUSwlLTBcKQPzr7Wx~NSKV{;3Y=+V-NU?m6XY=UA=)`REu^yJOXpu*
zsb9CTes7V{=K<2&Q>LL9R;=pW;syb-7N~X#5`U%^Kd}g`!P7Z8WK&XEt`LoY(B<Zk
z;y%1MP!o+U0sp!vwHzz6;>GZ!NtFR}5Ek#_yD$Q1e~*@EXjltZ7tPElJGOFy1^ih0
zZfC$Cd`~Sf<s`hg>&t^@M)`k?5flnuRiwu$wl)$o#mJbxO%}L_AnAjM(+ig{{)VlP
zDfgbd@w)OCa@)4cFdb=8U#ke4UoKzOV6OMoQBkUgq_<Sws@hWVm#cu?$1n2sDMd#J
z52Xdt8wwVuT%i?B9Nole%N(!51Me`2q$R*W6UO~cNFux(^0jsQkT4ePj3%T;0eZJ8
z&?mlJNhBPcYtE^QggGGPT)*N*E+xNmSOl6K*5DcUI$hq*87NdQtgBnThmLFikBB3m
zuY!BE3e0F!2v$sEaW!}R6`31P7b9Y3LytSD7d(Ua?E(UW>oyH<ke2;lE3i3EwPn(r
z2b3|lq0j5I-4_(c@ZKk%lSud3pDc;IxhpkTBV&PuVY`!c9GHzklbgbCZ<eBP*_^>)
z4c%huq8`|2fk$B9)%xiPw=W5OZIAL{-5^7hTp+r6<w00HJi&=_@S*0jIHOf!m6<v(
zaR}G>{VEr05UbG;tQaY(Wloy%3OACK5@)j3)m=Fb1hBl5iW6XWhWFu@9XE6XWoXD4
z(V^eok*mj*1Oi{6=sjC!Z{?G9aHEU>Ao#m#UfD+2uZa7=OeD6lQw*{baJRgC)F|uk
zo}OxhW&1H0TKBZuXOr`qn@UU>|7Img1yGHmGyJ`mb)9RwST7%KtGqf9P*e8^q3<S0
zxO_ST=j(SCwIoBJwb5#~mBMrU+CYo2igg5RIdsemL7g$fPx@2FdkkGK#)wq`mhyb%
z?Ny&FmMRKGg~{FE>;;DKqBVZ%H0eN=3MPn(TsBCq6Q!jiBfp{H!YBJkC*s7&rIUP)
zTQaTWHIo5}<1l24SS$;ghCJw0ysKUrk_9tGiaGsaK1?k(;CSs8)HGOcdutV(moo{)
z%D=A^Ffamt$Bk6<LcSt0A_lxpFNd--6F0$e>6t>uYn`0d`m1m?5wC6y^lTLU2eucp
z@{7xBi&Fy#3@^wYAi1i)-ST^bW=vS(iCHuYefsxZ`lzW4V0t%5i>=R2Hwo3Q_e){0
z=`!{c-Q1*tIEZGKLHNwMyR{F*R}UK|j!&sY&TNC1cu#Rb0}j^ro#(DQjEc!g=vEP<
zYbgkKOo3|&?(b;taqNgdm(Tkzx?4UDQ%O5(e_JRpG}ed9ng1G}VfmeAbn=aeFUVD_
zLE0SMNSDnD4`i`pzzZ*AvBIP5{G?Gyeg}Qcyg1aX9*4}YXOK24DUX*Qq=7zl8h`}c
zmu)<2wHgOf>^?(y!BGLIfenhz`XGRET)bzlGq?CkY3j199NJ{YgMPR)#TP3BP76l^
zC}C>mN8hzJ2w*J2f@F|M(Xu0r<%_G!wcN6eWtS?LFjGV?NfnZpjfAa>X$u*`k~9Zr
zP}0SLDHqi$g1-ihlwT5N*B~45NoJKbow@XG!Q$22qN4qiGVgmbzq~e{xKXgI;7SYh
z?!}%8<k@gK0<2kDqe{3YH3g^nvICsOJYl>p$A@gZ&Sa3~7i*^GaLo@vF^ThO>8c+o
z`FKPxV9^nY5)gheepi)1s!sBLO$Y_uMz619|MNgtH)TJ=+kIF&u|qJ_7{ng}t&gbP
zQhHH6k_3Z;j9MQaFITNB61?63KlbthpFUSp3uBiI`A($Y$$;{aZ3xjaT;b2%EY^&5
zu^RX`<S=9mo{$y^Ko?<6PvBJw|EstG0*$D|MtR7~{;#&`jzzH(F^RBnQ#GAPrD6~p
zb=GyZ_j&<zO+)*CB9;B$Wvbm=m+Twe2SVrl)g#aO*ln5Lz~{Z8kXO3KF@s~_zH@l%
z^diIo(wH~O!~ZZu>*V<KxA_@@y%FSdp5kARX(}C|{dGs<CHu(v<oXiwgtOh+?cw}{
zejAJIedkwsH}$GTszkC&s^dZ;4#Wgw7Zt!d!Okm$Qg`OtH952sC69;C-MFv;Ddbk!
zG(e%y&rm`CYB~;%L>;+PotteNWWMPL%8nDcfFVld+Nh!PbR$j_BZZS3t@OX%c%x&5
z#2K)8rVMv=0jSXeOlzhB!DZ7KWCBXxvPJ-nvdb>ZFrm~J<}5@pD1BS48z-7w#o2;w
zm$b>@>iXsdj0LQu@3w#+^ZX)Nn#Gf{R+;0+r%km@iXgMo<RoZB&eT8&O;e1AZ9dZt
z3{nI#pCz$0Sn8}YCFI5SPV$Gge3+n71OM_e&35CL@E6cU4{S+^rqBFCOsMMhbot@m
z6b)wVL#%+ph~D1!KWUni5WuhR3lJksQBm}!oSa?RF(|qlt~wPYpP5r0aR{YX5PfBB
z4?{8uW`+4K*^=AnP_AJViP2At6&J*>*h2B#$Uk7?ZV>?cN7_8?OsIq(5;Mu*FOkdO
z#cC9`_-hQdLbfY(u%p@sN;O}%n6MsLIKh~0dB?I1I)3$kJxD~lN|FSo7|?^!^a|CE
z<X`biVoiq5_2^vses(&4M*la6Re@eF^3ULd7v-&IDI`h2uH&g38Z^tlu0{IWYKG$u
z44*Ju2XV-No#%9UsfFvQ5pJ{n^5N{4L@1MTaPZm0)_6UggrVbfl0?bZeq@&mf~UYG
ze>&-Ef?NGpy0zpdR;6G_!<On`AfjymR#R+fs1xGy|KsY-<DqWf_VF2mv5ykjDb?7Q
zkg=1J#83=niBQNk$i59ii>-!ilclon3fZ?ZMz)fWv2SHJmcd|tZ}<KAJkRs{{&)9k
zdEIqg@AEp2<2=se9LV+Sl1lQ>ecJ@sygCwX=&WMDl4?a-J?J*wz@i?2>s|I?#_kdc
z$S47^dnuR2i0v!`bqAs~V^d!qjxDBi4_&Qrw+`|5=8k+)jPY@hv|H{2Jn?t<4*tIJ
z0IczHv`Dr^Ka&$2%W|#t7rpE~JW&7#D9c|ymIK6T))-wK2QVy77Ta&=s`X$#KkIm%
z0=Fp#UH=JC*%$BGr^M*xoCPnzWBr$8v5Ixj^xRTVac3#QLgS6`?M^A1zG91s?$Bsp
zO#&F+P(Gq^p<z*wnHdV?tlWpN<W|`=j_g{_+rJxW{c2UR&TrcH;L+P!AZ^{pQWn{@
z7TQO-S3)-g423EG>^CqLzVqTPFiHI`_GTmjXUq7GlS4|ok=7}^T|J<)N8Zi=+>`xD
z{<2tRcmOgj{qk09Qu=9owy#|JH9Jz-jR!v#HB;!}xBo>jrHLns9_P-VA@jh2Y!W-~
z9gS%EgmtaRF7j<qxAceL{v>5ZpOOYIAsRp;<||%S5^l)-|M@enS-<HS1U~}?4EyvS
z6_l1;zH{@I@p|GmLFmY2uOHYJMt&Pz067Q?p;BY;lm4vzzrFzhh$aGGRQ)3_NL4sY
zKXs$&cIPv6Cs_*)*VG+7hqee&r^5DI#+;?CU-@KzyBwu37?0=Dy`L{;*2x&~a0;*!
z)`5~0V{vv(SI{4yzZXbVf+oG+n$rqxyR5OoD|tsdDv|>9%;!8jc#!b|FXw}Rgr&)|
z^hwLp4wM5xMspi<n=GP%2hr86W?I4A@ykzh%w<+tPmJ-zRfK7c;BMSbwbaQ-PQoY)
zRri|@Kl0ITpSc@_lY7w`P+^cIKs{M~1CzcF1?okh3>a>1XAO{P`&B+b(QB!+Ketz-
z;+JEO^8XbA)E{MIZ+V0=4O_!4Ra)}F+h={+0to)z!aqR;hh@_aVNF&*XFh;$ZaHF$
z%zOhhZ`ZP1g-bg|V5-0g$uoAiQN&rDG{>>~U{M#xhaXGNfTHahLZ7VA*wVL(Cpdqe
zwa^+(;{c4`OgPW0zm@Mnx+Nf@GO56#Li%?!Ncu;%25qUQz1hS#ih0h+WczEUSzzR#
zTM=g(;t3Ldmj5ojd4BCE1TK7f93=MV7J>(JX4lbpJI0-&J#q9GEBDEzitcsLfbr#=
z&v_1Qj=+@aRqN)Hx)Jn2CD}Uv^(lc&3nG!r2YK22U;o6Lo;!SQ_DQKcFVOF`=FLON
zHnT2(pA?Ygs2|CGay1IeAws+5it<4%lk1`V${AMzQjK3b-+f~zlJ`}FCovM_%fhSo
zF1Yn<r8h?_!I0+5S#U^>)@D?MR=Ljg8rR!uH@IMoqNiX!_99FZ8PCo#+C8Iv!XfdF
z<>HIP#MgRX7jfzzH5w@Ui|^{|i;Id6?0;Fv4}Hv{D!cfVuNCDV@0gMU9rMDrdRm?e
zQKkkLYu*Ra51l#OY}pCxEI<4;8SE}$KJ<9)H=S;X*1glh?eh*Iz!r!5&R=dhKX0LK
zx9m};`VTd!CWn^PFD+<A&5NQa?NR8f;-dDyKh5<tP@M7kf_KWqnJ)~ZEamDv%XtBO
zu1|4nfM^BKdKs1tvg~oo8Kk<trg4v{@EFj50&yw{prjPj9D!vlBq^n}f0OcrZb{*(
zYOH`z27RbE5TrvrOV>r2aJ|!8Pc|8PY9g2K#npcfs@ETZZ9*nQVaMxkwGr$ovvT><
zldcU{fs33Kfz@8`&QvYGmI+{F)i)c3z#507zn#tWILUjhJ6>onrYga$tf$x}r2wJX
z<MF*`UO&HDGr%eK>U20YYtQ}HXc9Xu+hSs~?(8)+I8MzXO&J`DCtr0yFv1gnofx$}
zfjg|xi@H;Df+cGbS*i@9)kffMyPMNjWL7W_EJsJ$x@N{B-G=NBveZ|$ma1!WMj@Tj
zmky5EO!J;dH~HLxbboIWQDEm)pmlLKV4b=a$q}pEzCHB%0Z+6PM*X=Ii!VrU<(6c&
z!3xX6{%9y9Q&^1#k*3R5;d*V}ts#Ui2_>GapL+*9px-S%ba8vPSVpO$b<^c8!@kbb
zp`@0DCaa6{GQYh;73X1N+<wN!_j1!s|H6d7-5yO7@3y}E$aLA{Fx#k_{;x`@mhP~#
zzz<Y<@vTSQtr0XbjHdLMes`PMup=YttvwIwEcK_i<!$wx;{!6aWp6fYxqN!vmHKkA
z#tT(C<8EzR6@yi0|D#eCA$Rn+J9lxLKdfc1|HyAcCiDm{y^woG*(vqB(>0Y=e`B$_
z5P@Qo1|K{Ufl6LASE_kkrBGw%WpDptdkCwlrCiAkEQFEU+mpk1)mM4^^rbOg1Ak>{
zRGE2D|6&ty4Y#fOBUBol4ouV?%@r|F?xjYgLybLIM9%``{iqI2$~Ii`zjVLPNZe1u
zl){Zrp1czFw#gKsJn{N_h`QhXLN$22R224zI+H>RC%10%Cn9J)Qd$Wn8;_Zo))sk%
zjP_zl=BIJEWmTL>py90;?5utlBvDEESFlHE@jW2j?CW$?X6K>RX(Y1a>mb@|H=QZc
z5r0e9kNo~A8LW!%_b+|+DGU~U4nfrUTGlS7-o;&!9N)1_MBSqYIe*~m7}u#u_fqAu
z8gfu|H@{xjkTQaL=qvjq(Q<e=M1kA5mA+F!AupmB=ZbzR7laAoUUBY$un2;-gy~_R
z8q6q7FlfWJ3(BsVsQPS03uhp5gWDvdzR;k9spi?()b)4t1HtPJEr~6yrvF@yCZ#)i
z_j2no6QMh~;dC!-a-hkY(NPJSI<SkX#-mNt#*mN~f~LO7Z#>4#|Mpv$7Qs^{&fGF}
zuX*8Rmc%opqPPhJ{NkRxH4}cn{E6+ac<)g4<%bxqUqJybn>U*GJP(&OkNam5#7Yil
z0F7td2H!&QJND>zn=P{VDp240ftogJ`<mUnK9gnOC_<MI(Dbxc2mnHdS;t*ecA;^i
zweat)ms%kBOQ`6uPn-#gY)q;w#RaQP-+k*!H*XR(OQ87%Nc~C$U(JwjxY#_6WBl2>
z@9sDVf4mQNP5T~uW%_Q%62UFCVwEAA|LO~6kmBkCF&s|Cml2`7qd_OyAD*j}VVV)3
z@j7vp1~D41$to%XEUC9QLHjjwi46n}kR#AuD-Wb`{M#9{X({tJrsu95GPe=r)R>cC
za3ixVbqKN#{P8{7OEt$Eu-WtvT+=Z7Jq+TFhRDQtH`3#X6qbasBW~`#I3BX0c5f(~
z9J{=h-!MGDEw?=KbBLct1oj4aGd2MSAJgcM$kSR_F<&R_Tx|~NT;@Iu2L3OX^CB_w
zWOtCLDi8|+)%qJgUFOCtGgyu@C};9ITP&1e3xz|RavaYVL(>GF5RZQ3N5+p)2m#)k
z4W56kxYu8-Kfb1<kQeSO!OQ3Uq3zUE-_`CB3j`0DW(%|-z_JG30+GO6_b1Y$VHe1G
z?!K}M3?v8MOal?c6`KsMk(Mi~HLRmT=34i1?-T-e2747KCQ&EF#KTrXxW(P3ES;2J
zAxz@i&L*8*4I&?uxRK=ojrUP}9(T=jas!P}qz*jE)VJn^r+MtDp}^!Yh7xg;dZa`m
z<CP-2TnPp9;l8A{Z&68ak2lUveyN=}L>v#C)r|IN+Fy7XzGK?zA+)xMr~b0(C_j4N
zLLr)!<)Mga%m$rViZg;AY({=0T@fXJT>lY&Jd$ucxL2pSS8jz^8Z3k$en0j;A5uMA
z?>z8{GpQJ`n+l{7X$bh5$5oM7BUD_`_%EMv{HxAlUq_Wzk*^2$nT)S!;L=%6>7@M?
z+EPhWM6lOB#}mRo@W5_`&je>(bi;$k*3^6EROV;$g$Jf6+fnGMmN>o-3%IDrxRy(U
z10K?Ace89`{SOUx#x8tn<mopTytAI5Y^-Wo9YBLaAp%B_({q5S^<LsG&y)lG@g=o~
zc6xp+Ihl+|L7JMv;JEENcAVNT8I3StYG>{BQ|yf*c^eL7UpO%(5*xD`?u<|lGh{hp
zff;rXLal<VW3tqrc#hEEpLlNyKY$_Vn~Dh9VHVkS0tA!_mp-kuCCu7qj!pkfUxsCh
zpz)(A)=qHu6+3c(?QbP|DGfJTdRAA(*`g?{5r_=3K#aPAU2ir+HrU)<Y=BA<J@T4J
zGm>6Enh^~J!9kpXqG=OvfNq>sH@TVOM)&I9^5Ih-;tmlXVt!4z+DL7m46O{sM7{B=
znA(&W30kKQsdfmP4EuX_T^v<R5>$OtR3(o}YIxUE@CzP=9Tv4NZqG%NUVJE4xM7r>
zt6kl4wxH$sXtEuUXgt7S!_DSS`w#zU?nA@)!VW!-6p6!l;633iT__8_kWJK#3jHl)
ztqgq{WpE|0#LDK^%eY-3>e1vEP3nHaj16@&Y~kJatgCEg>@{IE7PN5+1uoj-Ke>$E
zCC-~C7bjUwKY4s^zUCdNvI?)FRabT!<pYH)#g%ru<~TqwZeLe(@+%*AjBhgTjX+bq
z-R%A#$dgOiOlUTXIDA)I77uLEOf9}tDk|e+g$oz5#u0z$dfB#gSxKWsA&F-t3k>&j
z6TGdyFE;Iv4NVeahO72_YpuAn5k@^bK~oXa4rfnouSKLC&G5^6x(2dMge}`J;BY(X
zv{qxOCW1{eVGHd9oegh>GmF8RN;q6(6XkAUaPPes)5iR8tbkhj<FTE4AUIjpS*=e2
z*TX1OCpV+~p_9i7Qn-6FG2oBXQc`2yjW8e2?R32$(nb3^p5oK1ojYqDLV~`-6;Fn+
z<G<#uED$5m<nQ3VasSAa?M0%@kIJLs^JU^yvUmB$el-Byu$RKDda#FA&|M#==&x!$
ziz{_-Ery%RF3@g(Y+<QyBjeAJKvNU!WZ^J|nYWcd{K`|9uphg7+(SOL)o<W#MM>uG
z<7eERo5|xg{=Yrg{zW6D`b=WQrYTs8yDU_EGAAZ!y)i{~_xG4z*!k-%spd^&@xKr3
zN5xf5uGdL)$36~uA+%liOV~P}vp}|0FBx0*!%7{MCyCiPF1^JfMSUd`%CC9!RcUk3
zqxm^dGA?N}ei!mfF0eM4Rtnv)+>2YMUI{&dwnUxWDfQaiVrNdjc^a%-u<j=qIco!N
z=G;*BXl={!u!=4E`|%f#W8ToE-h>%)7%A>${1ZcE#Jy|6Y{u@qKMf=6%We;?))aXb
zNi?WeXM6cPelQQLMqfqNSPzYKe={7+KE*5wopK@@72K@w=P*+VeUiZgYzIpBg`Vd9
zdB2aQMF)0kdTyH%O!c0cd*T{&<!ryeEy&>DMiHL~<D<T(SQEjSI@fnEMomD{1Lo&5
zmpIx-I|5h^9!)u<PAgt!Jz_ihiXr7r&lT1rfH8}pul-DLKkBY30B+;_fID>525?pw
zXyC%s?~a$VazQXcCGdyZYxywZaNEkDtS0u5^IRQ-qv<E{f)Kcfe9X_!v&P>3hPhi=
zeYL7@c=@`5`1nLLku$%0oTPEM2Ko}pwZpGyuj6z<O0zH>JNf&M%wS+I@3l3zp!M`W
zWe+#YdfAb?ZFY!61SS@&Pbp4|6&J@R$`c*$Qqq2GXL}8o(<QDtrczcm#l@lC9A~pw
z>12g%MQcWyL1*Oh!nF;#zg^<SIT`biv?4`lRY5@*nCwV|Kj`<_{XPcZPPH+tJ5SHo
zgpj>3Bu1?riIxK4Dge(Gnik;+u_51!?-aiIyJ~Jz|DuacBb0Rt0uqAk(Y88*G(2oP
zK*8WCm(}H$yCaJr4L3^gNavlx2~~BfNT*^U(Uo()L{z@W=kZ~9O_ACOyuCD1@FKZv
zz)IaG?~3e5V@x6`=~yD`h@p9BA!E;sSvqWX=tjTW-=?4InKdVus)i05wlTKXKXKIO
zZCPJ$=BXqtn?@rcwXzAkHF(L4UGI%Ludq>1{QIMyi<&RikG5NUzn?0<NF#-lbW5IX
z4)hwL!J|-Q&LoNQ*IuaUpKIpT&`Y(l-cW=RYkD4-AlT?lf4WfZtnrmo<LQ=CgM-hu
zir}?w*wxE!c?+i@<xN?!!%mtHUEFiW_KtwOg=@^UEJ!3dgBh5~<LTcF2L8$Hr{*qZ
zZ*beHIAujP^`d+HLpK|}vg<V`f&2H965NA_v?i2h2D>5Bw&Qmok{JzV$joRgmebRm
z9`3l{FNRPaR>C<7J(Tsx^UmNC;e@k7HGxg?9fOs%pF<wU9wN4;=7NzwCL;HqQ!Y8N
zuU;xos4(fgRTeUKN!o@L)-I6A-Pqfi$q4tu>&eK--Cb!TRNzCpMo=nwV^B-mG$~V-
zM8ie=4(9{E1fSVg!e8EK`raq18_>=oG_TTJeFlD@#rq6I6B&&eXeV$g6@X0Irf4_x
zW*8#Xp9!=|J$rjj5afO?SoRCoT;!<<9e<b?^E6k;{G2*Q<UqW8sQhq7=I39!+b5t!
zKQ|p7m4qr{UIOk5?_Irp`jFf?nC*jEZqrZ*^QExypb%E|g2S6*i$fep?c6}$(wQ80
zIdN6X{}iG0Q1(*sTmJzq?{a|$tN+zS2y@IICc|ZN$o6$h=Y9M}sZ8_pti6q1Z_B{V
z1Eb@%$xb1vWf&`YXS1$U?YWt!khDTtq{LTs$uOGU>R@@A$onQK0-m}Gki&$qnfwh8
z>YPy1pRVD=XW7mYz*F?Z))RKoTp&1@Bs_1A%265JVtZ895Ly%A^%fW?t>_=dz34F6
z*gUUQ7s)y}uwL<QwPHNC7@h#`HbHw4A4yx)mjrr^*`1l|jZ29MejEHqweP9+ry7xs
zwUHn5pm`#$q&y&m|EC7GCY>WNXYP#$z$(teQX03in{@{1J}HU|4nQbF$2I7YeP)cv
zRzlsOH89<Wza>D%he;!pS&xx(lHX+(zh7^<gv0f&vLKX?4+4P7K}vez832GCjg1t#
zXN$d>^Sv#TdB30P=o&Z6ex3|8+NE$q!DQu+sMy>FXkftzN|Wm>w(K2=9dl{yareYg
zuRH%{qu%;hWOmTr@8Sc4i0BKkceT=3S$KI(yzRwlIOCLQ)P79qMM7-jD-t(<C0c?k
z8H7NtacGy52PkhE!aEwVEddv@opzx593<NuMqO&b{KIw%O}9QIR_XgZXgXT@#wz(?
zrAWNxDuZ~~i{^CFFeq3gu&$^(t1Y6tJj<6}d}zy=6XyHU_<z_iww>UTREC<qg9x||
zaN+lhJ+IAfhc<@J?clX@=QTI}jxE-F3S3R^HZwRr_Lv<G10JBAb?PtRLYyvLV6(D)
z!m`Wh5BpwTkQ)%P)Zej)!G;{|hwTD$Rr{;!4FxZ6>34A<-6f3yI`f$MyeeY<iORJ9
z(t~*uRIQtMdAnC!9^LEZdTlAzrFz}LyC92e?LIaPq7=>q?{jttW%(ZAUy@qNbdGs5
zj;*tMUBo3q^Q39@^uF8R@ERzB6+`%WHz0KwDw#hu_X)3n8(r;xT0Yx7ESH^L`bdq*
zIzgwPiAu}=mT-^sxpuj{mwQizxL&?&l|21<(iBJ$Y9$ozi{jLt%Jzq2O|(`O0b!0>
zI5qd#&RtsQ=APo{?k1X4LE5NTM`t#FQ`5?bw9jHjuJyB0;(A54Qvv89w9q;Xyk8Fl
zN&7}=T+fmH80?MNpz%8{3I`z_bW?MESz+oyL(j+Mn9q=hgIat)CC9Z13SQEB#b)fA
zSjl(WD%XBDIi(@>iLCs2hP3-p4*r!KLs+_y;6w#o1_isj9ocW<3T=Cyx4vo(=UV{=
zqT4&^eQyyvBD%@!7Pj|3%eSlZj)o&rb*<*-*!0|OI58ZArJ%{98Z%?Ff2xGfsd0^3
zaPkt}3(z}{t@<pv1aOZ~+c8ga!a$gaaZY)Unj>*YO~|A$_-|ChwfRTIqK?<oZkL+u
zLx+kk=PrMu;jzy5(KB~0%ak{v57@Y_SURcfzG+UK4vTL&fKk0ct@zsVjZ&qWA0(sN
zsjA0TEkVC$l$xnyTn5K&9<!Q3WL4miAymImMTZj#^}iT_MK!PF?^aTGT*Pdy9QQaq
ze-t+3LFEOVVHEe{OtSnu($>kQtC5LBWFl9aL)QC-3<u5bm8o)%q6~J6FKf~p3r>}C
zV7!)sAEko3)Jg<1_B?j!!1E3}b7w$|7jp2p+|8Q5_R3Db5`?Z0dSvLY!9nw1hp-&U
z81GrfbHp3%1pBc$ZjQviBb&et_$$W|+O$^}&}<5<3~so9bDdd#X;^>6j~MmGcOK$5
zC(MBopz=rzrv{QWXM?}|f{zL4(*D{{2hT6hH!X2S1o&rm@W&q*+MIR!P++v8B9|4L
zZ)z#rs7XwH4sP1jNhtI@T*IQh-n|Xd1lBLUKa9l;*`vLbYEo34rzZsUc8g9&wrN;s
z-LnAkR0OPw%Y7{W;5t3MR#vr7?!#lM)F%>FJ6C4wuJk2*0Xp(7FfA!QGmv<=@^>?R
z4F-RVoGQq&b>szsZ$-=XK@Z!UY7iU1bcB7`1x1N`Q%=0f*)gw;K;o}CF~5VH=^mbl
zXN9I`v1WJ73d^!<Z9@KrmaSyZsgAtp0!S~sA~~rAQ<`~L6LO?SIbQx#EEYz~e7V?E
z23hQ1U>V!pYj(Tu#pU-F`6O5s>Y@IyGu`DjT~x`FsD+}Z20YwE(GQ1Cy#!ow)!_+v
z%{rdR*xOrJTWZ$-MX?Z5&P5Rem<^CQO{=pMJMhE&3<mDT57K--k3KzrMBm!7&!HI=
zMuAiRjK)i!qe;{i{*moY>oabqQ?^`l$17^~J54v<KB{BWNkU9j51==-oJq^|e?bqO
zX<4O`RzOalHg4o!D+5ph|DOM;Y_<1~H_l|czH^iIt?Iq16~KmD32MH!;}90hzHML{
z%GTKjE;Ye;@}|BWqylBdzC~+cG3@kqZz~~kmjg|zl9NLSC8$C(Sr1x32xZ0NixTks
z-!AiWi!s=k`y!yNt7uZ<4a4dXAh$9t>pI%&u~f7@8zr~X$l9n-xlt1MEuE(kqq7eN
zSAjpx){P!iIjP}e6Ym5-8h@cDM)O~_TLpgcnaQl3n_@Ti6|M@JRQQ44>WdMsLLBp)
z+3dvj+4y<*9vx7IzItcfEDQ9a1`K3EwRv+M=tM@-bF#)iRsL=ZvuC5NRTycT0(TMz
zr@*o>*@}v)sMqVP`yj8E7QVXvvT^*vYCJg-NoKFMa~w8ndI6@p`dKvNpKBmK^gcHk
zfDjUY{K6I-88IF$2R{TZuYS}1R`!DA0t(7L8kE@e(W2&~kQJCMsqIiylkoq8B7k=z
z?nkw_6%)K%uU_BV^v;_kqfp$N539p1LJ^<CHU@?Mic5tpzx0?{El!Z}s1SC}oy#%v
z9<$~(-W58q(N5_uj%t-Y(=&*LbjwdU<hN{)kLg26vp^y_^?lfsr1@f(_Sy4KpV7Sp
z9trDV#1Lmu(VFdC$L7=7&8HoIe?q?H7nNpOjfT2WLCCx(+0V3TOM7UO4E3a<SK+_s
z-h7F9wq@)tXnY@tFk0=Lnsey6Mj;-(cF%BED=1NgAZyk+K)VdW#H@au^5eUUfYE?o
zoqoM93X8AVL_iQK;Pi#;d9gn}Oh{z!_rTwClolwu%BA*81_ZKbiWH11PsYjs&@$Y7
z(GggjJ7n?R+0jcPa;qlpWr8vD)ST><h~1;jhrq0+qacny>>+o=#|INXvOSt7z1s-R
zcY^|BcYA=1l}{|63@lIi&un<N6Ra4_U<dhL>t!xA6&2)r!rPcxknx+f`!#Z3&2Fq|
zGN`eSM!p`id2NqwR_~kstUeXHwtUwS0ZKVnn&P|g=>~%cQ43^hXWTY*EfvDb2J)jF
zTk2W>e^CXIJIxH6W1NdGS?cYFpSRRmi`@WCpSl*Mi~Pj-`sguXrs+@9zY<8H^rg}d
zC3n=rPbAdhyhEm%%cL714Z02b8ZVNPZLuho!S(vQ2WXGiRdL?PRzfxm#Ed95E&lab
za7Ck`yU+CY+DVnj$`jHk?#7l4F`dzThT;^r&Q%tdx<%jG>H1;|RaN$?cyjA|M)nWV
z`q7_WPTc^NqZU6+r*H>Lsr~h}&<*077eB<;YUwU5n)3TrL|OeE-OlxXY+E6hQk)xn
zFHHWOp#tG;Xmwp1brVMAGuq|32C}9@>}o}SN>0eg(?%%)uo5*Y-pyvbtM!msl$_{=
z8g^RJ&>f@^_|ay^Z+H6n2T5=ec2VAq-rgbn`==C1ZUKBIDCwiIYmvT7@#k*O_P%Rl
zPiB>euqo|;0R^{I`zf#rrexGcIIdJid-WL=^`610VcvUZn&9QpYjx<Vmz#k4zN0a`
zTZti=v=It-flB}9dscA$x}pRe!uE6v{QKB{ctIV5H5qVJBcGmfc))x0J`;TLfPMm^
zxg3t|@a;HX?yWTi5S0~qAPIac?FEcQ;v8kQoEZ)q%Wv!fV9GM~4q!6&Hww79&>R_T
zJv769dJVr@1aH_q?RKQ>V>78wQ7rf@6FO$Kvw263a$K6?e)PNiZQPFZ1!vqk*TXk8
zIpKQD?1LLW&5Ycq9ag|<Ab(7oI>t{plwm3cG*OKQfahq}ue-(pFp@2=;B6)Nws92%
z%g2)AK+}3sKPu-@|1uB@$rA<wzr!rv5<wTQ83Bmw(2bql9WH0ReG`yS$JOrg9L{8v
z#E`h)64!gLkEo=NCrVl2W=ww(>BIj47y3O(;cbTT(s$?ryCs`KcH}hbLZUoEHVuQ9
zdzCh0n)l=mzr=WC4zyi9>ZMG!c^ppFZt_jm+LG4!Ygc)iYJyA8^D!IrMPh@dW_Wk7
z<!OU__K$i!4N`WL<bPlLSSU(uKdTvUb6iNBXsqmG2OC>Bb%4JB?U@0E2tWi;I+bmO
zk^%%Ye>Dr^eTRg%MO=5vTokHHvx%QJtsbi4^=X&W^}5*SdLkP`pD7D_%Ib1)_XZKP
z``bca{OR?7Yq=Q!_-+VG1b19T&^wa1LMMPN_;JLzl<fzP;?zM;O=QLFQWdg<!oq-{
zf4mK_mW3m7wy+yN-54Ee0xAh(&9xoLre1R|vsT~#jjmD+yNOUf5$0K*9LAzapgka}
zD<;_~rovr3rgdj}&MMhfs5P!_{Ac=3rOq>@m9w2`OG<Rj0Ixx|*G9TLNW}sdG;H;0
zi2Wj<;6I`Jvyb-rjZZ^+t<`5i;P6J7DD1_`W^2l>%Rs1XP$}7d&$gn;?7*h%rG!`z
zWd?=Rg^7~Mr`#$Z!Hn0m7JSdgv9npxQ<i^%yJ?*kAGtpHt@17V6gQKAf`OTx@hL>>
z6oc0dbAi!j*C7W|O!OYOONho_w%zZH`-6%b9dA#qf%q8&!KU11hItWP|I(JsJ13%<
z(FXv7INO3}VYb@y3E*IQW|1&fK!cJIx^rCOMv;^J02H5hUiF`%^--y2o5r==$!|;T
zh@AG<s_)MW%@uVgl?^gEM8p-{{maygBBcHPG;|d;D}GLgI<joEG2QESiZdmwDe9G5
z?%ZU1P2oGPiw-6H|6Ue4zYxSeuclnH@BRM$;Kq{LwC|VFr8c)UkD4aiTXrbNT@ixc
zfWPonh4YEKLke7yWn)W(YDiKs?@(cTWf58>6ODL_RzV>EdBiPho(W~0Q<h7Ylp$S{
z&@4%p1VoC{ZpDXxDSVL0VT^3g;s}MJ;%`XjJs9)cyFyoeFw4@A02$z#GfbdM)ecfZ
zE^0D=SvXh-A0b*CV7#O?H|y9p1mgeh;)Nej_e=okl-gAj;*L-rTH{4~NjJ9>q%)O#
z!-0<OniQ~01;1>6Lqlyuo;EoqB*y#Y`KA~&iC-9@7`@|b4e(n1MfJE2!r&Imn4iZx
z4aTfUhNI^faJXsXH|>Pbi?eIIr=zn11hXYCOko~Hqf#0?|9}>`^mfPe_juZVDO|5F
ze`#$1)uOb14124Zl-7Cqm(*JkJF1f*q$663wIV0#xzU9LSp;b06_LgtR7n>A`$t2a
z<e7#E2SZ6<Rw3JBDS*44=olb#?K(5je>*Qj-umYrCND7#0rd2HWOo*gKZN;0m;iA8
z<>j?_&6v`9K)=94lRYw@(1;IREuCp|KlQ&ds=lVMS+W$MxIYBW9|!^K{mxtUNsc}P
zg$NN$&P=N<*XYF8`kB+E)b23097j$viC<hq1$Jhd6%#sX-k)es2uO@<_r-0~0L6-|
zSGS7AY<^9~I&l~ZE-%U)n^U=uzm`+Khy*3v($f>+y5P=hsa)q=EEF_^Efz0%x8jkI
zT%2v?<F;I+J^mEqRc5z*VuWmRi=0|L)5;IS#zf#-f6KbF%ec3kL1s?-vHXcUk}k~T
zcjb5~CrMNj(1eLDhEK_9jmBgExI+x4^GBVPZ_y62^<cVkiJ<I{ple>}!~x<LJ`U!v
z*p$Qo&yNSxiZwFFY}+7FIqh2q;mR9Gc*pFqGSGPdEU{aE?H6r{?1Fu!k1|=|cv~q5
ztWrN3JFCIwF7Rd3OPxg?cWZK&1T0*7B$kzc!E#)2>>%*0D-VqA^TaBWj$&wAJ(h_S
zt7qVcIiQ(r2U?7mk_=RwLzvd>o;_yU)~WXTz;N?#ifQ|Z#)WYKd0WV7+4J;Qb<V}I
zmoS=iRJgazy_jPB@_Y*1nmH%f`imU~Qq>z}hY1TAe^Uh_2`^XZFpJh6B~(;+(M^3o
z|3~*ceWG6Md<mosrTG2PK0y-~TeNL%k3y4Leq~ZvxSWo;&lLnSmll+W8y5#uP*(o;
z0(5JBK_yE7I(pY1l~Nr5epbIR80Ws8WMbisoM=Aa_OJ>2sr_v-aLe$pYdOJ@Onp21
z>ncGi@sRqNI?mJd3s<@9#L08dmJ|X65K0y1ouG>MaSE!#{Rv@vJjY8y$4VRDd&kpW
zU43<Jkfz&b{0T5jF?@f=0V-yW$PB)_|MRVPc2yL1`HGpFy$QY6x6=U42v*B)_Y%Ky
z-uSY>BL-1?uL03lEL#nd*BAa-ti>eJS=@ci`yy@9AZY&e5+NZ53$~Be6gLXbtW3BC
z1Z<XjOYkh(4$f`H_Ww++yX6Y#T?}IORo30#OdIlnn2tKhm&v=jd25VeB>6>$b~Yxs
z4GhU9!3u2e@UCzNDPQvbj@Po2oKiu11@e7O1H<t)%+Y8UHVY)UC{%jE`c}v0KwGZo
z3NFDk1K&>YOkM)shfV#i=aurp-i;n(j^dgLU)S^5yUN?gYmMmozW`e2VwtjLXjm3`
z_=$NbA_5!rW8&ND(TeaaNH<>lVL8*~-%JEo>+0dBpC(0L?Up}h0Xb}ESHVPsvgIE_
zX!4w68EhZRIawg~Y3;>%aZZ4U<>Vhx{8W+%9*GDBZQ4HNGT>ShCIY?MXyVoSB|;Dk
zsZm`ng3$;6hpp!7ofL`ZJO7wHFq=@21^_e!3J&S6YK>}lXVuCGnI3~$1tbef?r`Ry
z)P73o{;!PjH>R)1>qN#7!!@2hY2~M1erTJMrWnv%bN{h)=uC%_K*iyp%^`Q_`VE^W
zi^pqgloKu77kY`>bOhtl4J*7Cgfn{;2P|0;IBlDiYeR>u(a5~Phrr0KXDc8Aw3BWg
zRfZjsXTvB^YBKY67x2SV3a3)7P(GIGb-F0l;!99e@-pyZjGJ*TSr`vj$Bm9?R<7#E
zMY{OYGWBV9c(=|JgSy4g--9&iMB&q8XBc>^>$q^l9fDPkPSC8d%jrfInppFQ%pzL>
zNOwYErW!pgUp*$EH(_*L6%+_GLP(6;fIi=u0i>bb-TGY$MzVF@LEM{mh{8e_f3d*x
zwQO}cBsr0N)*j;{FQ?|7QVDyIm!L=Ogu0-Mukd0<gmukE7uC-k0-`J7t>Z_+D{E5Z
z-ndps8bq`%>XmBtXlcfFg_mDe&1iC-<rpo7h<}2w5783+Aa}v|TQsX=w$-V;i0S8@
z{LbhL6;S`|HZa?j8Vsp#goyxib#Un_b^fdvw<1jBv4cHefeU>%L%P=voxn&n<+6X;
z@4(l8=kx8{z`#=&P{D(in~3(%*-a)c72l~%L(1<@<r#nGe;1hcm;H0J4;_m|{Z14r
zqrcGq9BrEIdvUY+tKQYW3`&Q{ircTfH_5x2t&?L^P}qDe^%7``uUuoP@96#D+N;|(
zUY&1TI7dIT<A~Fs?cM&%R~BXi<~rS!qZJGSt8WWU&i%+do(<c`*jt=D6g`Gc9y<2L
zU$5az;<HpAZ6lP~c^Rwk-<|eX$thC5Tld&_2nAIz_le7MKS7et16@Ag`IUh3IL&Kp
zNctbB)J0Px-%Y$-gO2CAm{!2eow>Sf=OV+$X37X<B{P3DC9@)$D$RE8yJ|2beIPrq
zg3Uo(Ip3b_ZA}A*&CF&0Vchd63M5$}M#~n_rRA)mHhf6mO`z8m<Y$4tKRuVi3eRsH
zaRFpZHMUN&+QlY~)5u+F2IUOei_JhVK1T3boSoxXLuwi^A;<{S*oN>tb7?_+dn7AJ
z4X0-HEgYNk?`@C<!{I@+nO+-V{4Fi8re8DG9Bf?fiz4UG_4GGCuJSjQi}b!9#`fS*
z2FF-Am5v5_#h(4+@|b|jBwr#uEBF<wc&n*YcJR~b@J{}(?Svrik0KS3ZSe|07)#fl
z%n!xy8Wmf)BHot1i-kNBc?ykSGw?MgF}_hR04p$w7XL#!#SQ<#gC&zkX<74Q<;@oY
zjYp3N$pS0te|5TRCzktynjTXo+CuwlUO2iGrHD1{EflR7aL$(jU9TPIy^4FN<|n00
zDS!jxpJv8#-;ern8~t^AE<xBO$hu=E$-?jo0NTObtM=Pmr#}m(x~i)s+?lOUOD_)b
zkQMoT4ePS7Y;?F#KAlB5ys=5q-0ysUzyZ2G@J!1hdjy(9c|m~lj~&etn%Dcuw!k}3
zaN^NnKgEy%Zaia!e@^?gb^bkQo?CxsX5n`aILX_1IospiI3O*FwTM;7Eh<7`_|-rg
z9jB2vPMs7_>S5^8h%&m0QEhKArcjBn8L1%xfQSSe!ibO`=mxw9o-aJK`=!E-CE>;v
zfC6CT7!1kV)CjQh;dq09mPD-3m3u7k2K$r<tbkr2+KXETg9g@b<A?ljMPX-Wfrg2Y
zd^%o#dakYM73lG-97m(Him>8~x4cTpE&OWy=H2mQYzOD3!v-|6I`;f8JB}*-rFeX<
z3Z?V}b6Cd=zkq@$T9qy$+kMrG=PC>jUUa>6Jol$a;(hs3SM9~!hTOGGXjk61(k72+
z)SK%EFphKK$2grVKHkA+59m^{AUOLC*Ux8dVP3FEsWvH$)N?6<6pJ6IrXs2raCAiw
zuYQIE$hJeeIIhL7?9&Jtl-Z6(=8YHqt3PmFJmHm#r@n{1pDGfgXjR!rjW&_N^PiWh
zCcYKHhyNd$AsUu}n!j}-M%E*j6M=TRbKWm0>z3Rfv0p8&>vE;?muIes!&?jx0j{n!
z)YO+@0`e!UFu!ZGWh<j4y5%tDxcZ)3P;=LPiSsq}m5`)mt}7aEm$>vark_-ey@y^B
zN`Uhh7nRpo4Fk44V24Ru6xQ0O0WfP_i-=}_Xm|d8#w}2HWKrS7W+j<0a<kRjAC_TA
z&w`Ad*Qq6F@D%uz2*|R*DP!!7{4o~TMDZT9p|0@Ylf%u;dS$mC1B=RN5>CI<y5ZX4
zg=gtc9}U_AL*b!^pTky^P~#_li^*k#v6^qs)|)zaeIV>Ij0xUy4N+H|_hQ-~$fYz<
z{xb=%I@mqkn<YkwERDzh!Kn?1N3Oy?NA-ii(oOu-bP5|B@6yNuOfy;hY&qMg-wnpC
zRIg6YR0Mj#xKxGjsW1SWNfredR$CVkW}wTF?7nLU4ZpdcCCUni>&DVW-!!FVlT~A=
zfLGG>2u53i$n0tZ8{t!2oob+w2&oSSYZizAn61Y%EI01Tc{Y9D;P5d&8N86ymZjFe
zigVG8n{B$uO96AuH=PEkHM52@^TKngpz^U{=tOg5vIh`<;a0SBOXkx>aV5*^$|p97
zx0j!*qX<8IEW!;cZgWal=hXc$D7C4NzLm?fRsF(MQdXG_*n~h*+_4IHEAI~7chDOA
zmVg*v5hKla?(u0JQuavehc=rZHTQ3PbQAc)0pU3xp(}ZJMO@TUukM7vsR>e;y73%E
zMeWm0ZSyAMG~LMn;a6b2HbT?A%c*nu7HD1!5l|9WZtCOUU-7`b^lHUr^$~#BbDOQi
zCxE(d6~T?bNp&^9DA!Q-v;0wVovo|DJ#cpNWntTCVzu8}_crzoB?AA(@z?vm3PobQ
z?V}S(=#uPS1B7y8c{c{@?cIkb9rV876okWPpx7);j|&<J(v0x@e%72GGAj}pk<OBb
z+r>s=8uoZ9d@tdMc-;XFcPk*l6w<7`kt~(XBf&4Dan`_fN}{fl2ctOeMhv21Ukhe!
z2ie-K=@=bR0yfZ%3u5tm0BLkrkL<cBBbKd94fzT*z2?V`mYcbJT(s7k(KhA2S?Z^)
zU`GdJRQ&V3rZ|yyw)>Zej&~TYP;Aa}tOm8->C!Dx9L${He!<|Wb8djW0BpjDrs;W=
zsQ$ppXAXK*P;{hyN^nY*VJ)tWYYM_r^K_Jl`I3;<O5BDPiJ{{1hvL)~@rz0m%|YP~
z@hJ0u^`k&Du@jjw>)ElAz{yh&uzuemtU&h4eW7<$shj3LfQ>W#zpVcqk@KN$wMN>P
z%p!;X`WxPq{orr<B-!+7zvnvzvAFf&->adpuVHR<U!@E)VyotOmdEEE!DLQCvD0QR
zC=8;xv!5_+6H1}bFG{XIkRT|dN6d^brRQl6DimZ~0wq~<&d(N>=LM_0MHekKzw<)*
zVZ6o_*rD*C*{98e*?A^+=M@z<A7Z?Km-eL28$QbBzJ%n_2X}X}gpcDUT~J1SsMBoW
zBMP5X=xfrsSFW+kP6(9D1A>gWE>XrtQDxi_xie?wv%vpMrA5Vq*Eo~~$4`+sg<M<2
zy_nz05GW2?pPC~oaybb10ZC}1a%Yf&@hNml@@Owy20w1a9NO1LP?jBCD~BQbfFA@H
zf!eJprnm>lgQG9%iV#7hSvN9XeS0*gfd_-A&e}#_M~wD|O#aC7{ymr-67-p(^1XqJ
z|1Pnuedge^%Ov%>>fi83{;1TO3>{MZjl_C(^IT=bBF0&V_v#7AW34>LD02P`hV=^j
z4+Y@s+0vT=p^ef7x`2$QGen5L?3B9G+RXq*43wQL`v_|5(yFes05qPuI&raGZ3-o}
z1F05|w`DKC4>>@SZ1(5ZKb%GW>$_ay$u)kjzoX_gxw+7MQ+D+m>UL<9puw(9=*5Sn
zc_sDf+wJMa|Ld>Aj+i7%|6jU)tM><GJLzw3+4RoI!KVW?L4^?x5y%dfQa8Wf&y#Y&
zsdJ0`=QccD4b|M)xo>l{Y6HQ53syk$x~{qQ%%ZoT@L``p)a{z&R{|dZBbZPzQ*vao
zC1T$cRj|bCGepyUUh0D@cs^$x^NYE^7n5LUm2)NenV1%^`i0{h{h|<9G$G=-Jb1qM
zQVTHm)s<G%<2lBEfrlom<{`g5lk3irG=Rc2fM15+=WQ+L`J=L<Ve?8!z6Lf7pL*aS
zKB^X-oN+^o$cRwh9wXOX$PQ)&-YDF|$^*e+3|5Ne&F9efKDQCdgJn+;%8*DEz`Kxt
zp9$Ib6&h`?2}Dl&e=#18#zM*0X^_bIX~4YCt#)XICe7_@1XrzJpPFNw^rWn9wSlSI
z8fS?|UoK&=6qC?-`|=qtX8D)JZwxMkKjga+C_Eb1dn~VeDA@T7^d<W0yHnY1aYrtM
z56*f&FBs7Rx=tc7a)f6?yH#qWy(<ZN8b(j*V<$WL!y*Nqm7XIvj`QCpoZ|ETZK~#0
zwD7dT>+;d?3B2zq`p;{g498NWRFn06LI8`xt;lAts>oNGv>S35BXA>lgAP*;@gHk$
z{@&a?kxBiux}`m4lUtu^wq;w(kepHa4tMg^Bmd{C|F3C31t$iH%wXx+zO!BB=*Q~Z
zX;*q)gxR3EA5F^byizeR)p%F_2F_{sjI!nB&VO2qSw{p-rq-)x4WSqFr|0S+Nx(L;
zVNRqr-8|qIl`!>*94x$wGA~%B_mN_Eu9ur}@kSfMwfb*s1<<#yWObR{h}Rx-z%tyA
zo$fCfN4vJT(O!QPn}2wp#yHbN9Wa%|f3>kt50KUU^d`XRiuG57J0G5*Un&v{4e2N}
z=rn3CV9ZvEpB8#4lY9n;ThGeu&+gt5nV$Q}odRV4+TXI@d2iIob}8PW0r=@2ozuJt
zO)>kKKnBPb6Wmd&he1mq5_{C$;x;^CLVbjwKTtGAD94opTZ8&EV{Fv8keNm4I2Y{Q
zZsg|CR1hOx4^Ngjm#|M7qmb{R-4J@*pz+~^6{>zLOcXC$8JT!9(ED*h?U9>-QrhRh
zs&Cg-quC>BysE+(gcY2efItf;<3T;vcc&?56C|FQUVza8krTqcw6zz6SoT47%-yOC
zLMN1+9Nid$X10ad8f}24el`<r?v#DM+Rb#~Ur2A!a0X~IGYtS}{z-OKaq8hI$$~@W
zK)roFK<KGmt(37nH=po8dw#g0;-kLLn|}-a{~URorkYtY{6yWM<wI5e;7l(jpEHpW
zxD#3@-|$Akz8PTUbeDiJnUInRjjk5)^p#+>1Znj(`b*X~{r*1V=*>KDlOQmguR*<i
zs7n1ACZM_heOLuh+px>^x~PV30wD1()##G|EmtMS3r!-lee<#LsVh6@W?-HZz?h9I
zW8pIjrO#FaT^@8>0Ck^hOm!EcDRA}x3}!eQckHZ6i%5m@uUp+7eOXY^{<6#!x_<4k
z$YzMg@#klKJ5aA7thcQP5zzYud(Lr$u^<Kc;2K<83*N=>#VH5jB2uvSiJ1Eu=o%?O
z0kiZMXVD-w9DmSlfV<BLcgh3N2$F|~=qe#U#NwwGhenAGg6s58guB1SgiuTj)aJQS
zE1{#Di2t1hjOGMmH~7(B15qsHuy;paulNsZGl(|045Bn@wkj4s6z8qX_kjD>tj?;U
z)JL0YZRJ_SFkPJt;nC5GafMMOb=trr&eXLI4!Ro=a130ZnGdU)IL4(FDw_mlsCAl~
zU;r$ztd&vd7(t{42+>Z{T}^&`2KleSWarL_fQNeayQmv*5ZqW@mLGXcu+d=<Yk1sk
z^T|!ac?|fE+6vBeRMB;duR7y)C%$<vb70WoK;d}mE-LT;vbJ@TH}}Oq9FiVAadYkr
zF1^IZ9NQ^uO&K8ueJ)3neiYDifqJJhy|3cD5A0MN`E_h{fSEy~<Kd0TBYbEJT@vu|
z&SHRX`*jGSu1JF4C^_mi0QSoGOylS>fMXw5aC)pQ5?9(sq&b&E_O$iB%j3KENIDz$
zl-(Np`84`+k3zOueqhxW4wCk9_Q8ihud3B2fU)FS+IGfP3g~LbR);5W?C7eUEJ)I#
znGv$1SUaO_Cpk2$$bmoM?aQpl+l!!zP0z5Kz;s{(Q}@lVQU7oFKk?TwBw$SENr?gu
zN2KQfK7IRX2ft%}gtAjP+G|p2MnVSglW6rU0)|=PO#mqHI`l?X0KG%me1nM}6Q?&H
z0Y|-6s3;UaoRGH5;jiD;|6JyC76&rZ+f~v%osp;EP4uuW(qw}Ff-BilW}<pj=Zl8c
ziUyS5mAHFJRco{~X0kv0+k@b%yye|qw4#2SlFA}>209%UEaL5^3I%{3*LJg2<T->j
z{=zOOK{(!lS>WYehBRVCu$GABAR0Q|_iTlwGDbEU@ZQ$#Px!TrJ8=9*yQ#i<0^vA6
z0Jd`hSJWKrLGgvK_y>v)h3)_@M6oayW^v&7l`68U+FoRi7G@d+zUo2!kNa$J<6K2i
z@L!AnnZ<ffmIM03aRJIL2?rWw(*gfwa0Y<<HsT#3I~wROXzfUpUFsZp(fv~QtrN*~
zU&W6fDms?Z)^ZRRMnm12+=}tMkzHhgsPjU-E)JbUp)>+0aiu`owxNw;-Kn2(rJ0gR
z;4{0X7}8xI2YZ9lSK@(o>01z=44r6x_DvW0S!tT26f+~X&K^+9W~Ei`bVV;3XNtU&
zjb;k0US9;@-Pcrz6J>pNmJKHedEZdeWeU$EC5L-MW!HcX+oJrY;#~FgoM{7~ua=Gf
zRLr95i)7n6U>$op(b<>)jM*3zqDg_%b5q6g<&hTv>Ci?Hm>;$Blcvi<noouBBGL$t
zn+MTe8qkgaHF~6ARPb!&#X;@_fIswfrcckgU+|-9tuz5s=<i#a4~p<dLFNKH+vN>4
z<$C0AZ`ZuFuF)^MZJi$vp<()h0r7@Ae}1KJE6(kGW`nlF-X2Zs$sP#_7id-gZe~Q!
zCQRls=$rtohHJMKVQxEfk<Opx<M=`Rj4Gm|ZpFJpTK^ScWlU@P-L!J=>>P||?cWOk
z)Oy7gr3*?Glem=!ZL_d}j&DMr%2l^f*l~$tvGHDPs(Go||Kl0T6XGxKetpufd-C?z
zIqTfIz%2ReC1J-dWUAVal6{NoSGZDpP{`fsOqXS4@4)@a!Be9lm_B~bccUhAD?~XK
zkj|&Xs<!A>*6@~H7}aL4cb%$pXw<xKvIh|gX^O-q3^C&E86?b6QAK6Itw$l><Y|l0
zFadupU19DBSpc@d5)TAoFbmmne&|nuBzw@j0p(0Kj-&v(3b@4hOUB|}EveF&{2oyj
zR+_|XhL$c7X3<sQ-@4BLi#~WDkMIZiyOh+9X3s+hpPBD<hP@pvR~EMW`xj4RNEfFa
zYN>aD{m2Bm=9&H|o})XPz>J+|a1_>LwSBWR258sB9suXLJBEY6q0K<Tb^^E=5Z^EK
z1!^OBZhGhly@~{KDqj=akU%@gK67gpx!74_*CUq@c&plRlAFeSlv^;H7}3e7HQLq@
z66)kF5u*9PhG!IY^U|2_#IFR{=TKLpdj)Bo3<O;k-n&}+*Jkt-puFC~mC(9;xu8T<
zZdb+?$r56CcM#-b<a8#zr2>&}Y@y<mpq3zG=70{t-SRy&DL3R_YIdgyaq?j&!^4ks
zzl213Jw5W42?AN(MeHjcydWs+jZWrGJsb*>ZW4Y|u;R)&TgEL?wDb5Nr?N=$zvS%1
zeBQL4wj{J!ro^%Ab8Ph^_TNM1$AnbC7Np1<B=G!ltm^5Zo<rKVP8(-{Pno4dBRUeR
zMC(rbIj5}@L~R-8qOMIIw@kI{#ASU6n|(R?(jEn<uRy2XCy@jgr;@RJ3UQ_S@T&?x
zAl(|DPD8q9If0dHXFjB*E_W8&@nzi_$SWlBUYoewy46a(1YYmtG{qNH=^9pXrgI#n
z04%;3NI=@#pGIh<Vt+ak>j3+h{}sU?_Z^2tDX3?D5gaxwIq7k_Z)fe*EtF9W=hOT}
zO~+W^;I@A)D?yOHi{*@P=yKxczH(L|@F)t)2{dE*4pfDZ$Wm+n371+B64}`ZWc!n6
zr>p=AGfr}o?in!fvIzQS9)s;zn7w~l_M_^BA5ZOHhq!ZKE`J?Q__4#m^qYR-Xla}x
zwPCPn_o?DzYv^qQ-XEVl4K2|;7YtM{2otj}-DUKonbV4lY+V7bF}ONch<t|B(e<cz
z30!#uosOpu(phsZ7F##$Q160ZBG1L!8S_M9|K>PAFAY~1Xt92SZU8pXU6T;Lg8zu{
zA1>S52p5|cEq({^S|c5hkpx(N=r_bLN}y)YM)mH_jxU7{(+>{*)vV!62^r}899f!o
zXw{E1rPG&*uKVA%9B-KI$vB)@)HeXQeFt2{BE5IY<K#w4WJQ*}4~ZwY;6#cI$ts51
zI1-;1to~)y)zDdt`dr_~g9pj7&Dqn64t`3!j2;8k(h#51y`d6d6H_+FGvq0q(JA2U
zO@|7QoJI9lQI?3h$4g~_RX_*+x@)q^)!&*eM^8@b`rH{Df9+pVGAPo-g*_54W*b-%
zNaDt+JzIaP@O}Rx+H1tap=4x|MN64c7*KMwCvO0soSc{PT^cx~^-)Ezv#6i@&RpT<
zVMj)s<+|9O?*!u92ba7tDgN0?VBJF&GhE~Vcb5fT>+!}6xz-s(4{R!W$cYr({0JE0
zn7EN<t4$2J*n-JfV0GC<oLnThL=_CD)R8Q#05bp0u*9W<@OK~CL9j>Q`4ItUh-_u+
zJhKLvo*7_G!iWZrxGP|$=ez%R-J>4#WVqPc2ntuV*}ao2=$nU;#Z)5o&fPp6{fSlV
z!^luByOSiYhl3@aX3ZHq4+UAu{nLaOuDIwAvtN-LG`n*NCd>wk*L6sUzA>n9gV)3Z
zY50)WbLcfNI@qMwU4Hz-?^DEoda-AvPHbPvV&u5z#d*P)BdlEcKqS80c72*}-6sR#
z3ju!maqihRtK_$Z|B++jG-b__#r{tZHaCswx*h7~y;nn5Kk&S)ei17vn-+tW!^oB8
zE_VO*>3sW)_pNW&H&6vX<JLcqH)MpV`_mQx<V&+%C2^89>A)<b2c7>#`Y`Ho%-)b6
zu!?sXEA%yCBGf@b)LgOX9~f7kl6}fjJ53;4?Y_3Mx!dP4rM0YATb@)nZITd5U71&t
zKIx>Xp5Wa|9HDS;XQ<RXzd61aTtsC-jCS_D<s&PDZg_8$Q`ch~foyuql#Xg-YiS*~
zvaK4dLmCnjY2UmIwEWz@I8a6>tXf!buA2I!W?(yq{SiN47tiuuo{mA<h^T8VPLS~p
zR~Qj>P~Z2v>vA|;$;y&|=2>KooEOH^ZUtPB5q|_c3Y|S2Ek<BCtdkEgaJ4QgPR$LN
z%$2)kZDD!s8F4GEzmrm*B^Qh0wl2>*0)(L<FEFJa1Q%)a)%*1DC8b$=XO7D$&yP3}
zxdM8qqQ~=8;YZxGm-On+@9=J6273`KCj41AT~c^nX|<%Rlin5D$EP`2k+cbzQI3)D
zs@%3W5ztgF<GiRh+Qk2ltn&`1y8r+9894SxBzu%2lw=(tGpjme7uiDALH0VwN@Zpw
zPDWW#LT1)Mc1B4yM^-YU$n5t%)xEyIf4Z;k`?|WXtGCbl{Tk2b<Jp37B5-voV8fvF
z<IZUH*(mi`$vkO-qyD9y7{|*4hL>FTU1)?9dt^fYxC{D(NIw%Sg+Qm{2PaB7U!0Az
z6gd$>VS06k4tP=yyC3fzt%PlAxX>abizfAd>qGyZ;c6@D8vSQQpZiF=n_zTHR6#L!
zPeeaW%A%;F4F+=HHU_3U`yLK;$H_6d&)(L(vh&sIOx~qvD=e!nAiL9@#s?0-I>B}D
zp4eNI6Nn_-7+94dge<Ny??1E%xmQ^DE-I^drP8sEx@);ep{?&2%vVgjrWNuk<P0FO
zkJoxWiWMwXG)zEUke1GLW|<9-6V1Xq%h8RVXDHoJ!t9z9`g`mLLxch__Vf=b$4Ubi
zpmR+>(l1}At3Qb;i;-0*PB+y(CHE@!6491J;_<_7NEcQ-y{M^kb^qzSqRIA7LoWzZ
zY&SgbE#FtZR?S0H%^Uc|bUn%lfUR!4$1Y9VsvXoFkZIOZ0{81QGv=Il3OwXz8K!LS
z`pvVI&7C*)o>fjPB-Enif6{vcr~<!2|9DgDq#w^^Z;OD?pop9PXFVx}nI%hdA3&K@
zVT@4rY)UJ3tW*<)?G>Wh6Tvw$qS7Y_27wU?MPM0t*T~0@f692n`i3=fJ7>8cLt_&Q
zp+Af2q)28mP|s0mp{D4W|A0c(1iV8VOxyk*(yt}|>eoM#?*FieXvJ`#>5CvwR%~a~
z-O8Tp0}36UO7T@o5#gdqf&b9~Yb$6PB#VM;z=7N5slPW6V<q3bbzLU!r~b^zLNohZ
zZHalXbv1E#Kk*Xm{zO{_vPq{69)BjU*>d-ncl+MucMP3uAMPhGR!^7sZ(K1|*hvW@
zx1%{#g)RGrEgSnuDdQWk&E#s;|Eq%E(sA0udZ04>l-#NsA-WlbvG}l{D^FYsgKnM9
zBb9PdWXm*=ueaOf`<`_z1+YaY<NQ3^x%CzNXmvA;!LMd$+_Og)dYi`a<iu~j0U<GA
zvS-30Cw#wt?Xmr7H<MHBK7$)nb?~{lPm7nYyb{CQ&d;N`1{QW>7Szb?1HFcprA0lA
zhYwi7W@j>?{2-neU}$>c&Bg;2lx*;aL(q<g_y(>#m;s2G*gZ4wQMa&z`@^LiQwOZ}
z&%D?-2evyi(_<wjLonpgv+h**@d~;NGrQa3S)xgh3)}{VTTOG7uvexW2%-CwuRp_H
zGVpopIPh8vg*%eH@6Td-LZSPv1a^%Hb=8+Y2j;FsBa}FNo$Z?Qob3kmp!Bp_rTQ#1
zlnV>*fa92H(P#4n`<)B(R7|8EeV(=z#y|h{=zj@HGBPrJ)|CP&fZ`$JC(P5gyx;+f
zQ_Vmq>}0W;I%V`d!{NV%A^lzD(n9CO>nvxOoeeHtGIorWytHYT9lPKyZh%Xnet7fk
z(5<r4jsje(nE%-w$9J;}Adw4P`sX(#$3o7H&n13-Fhhnd8(yKi0zpnyV#?r2YYss5
zQ3b0zLaZz`OmQP;q&$3g2<35GU8hdZ^7o)M9NAHWSWtXk89m{tNPim~uNnEwunP5L
z^mc+nVp^Z&nRh?#gUu@|H}sx<|7S})jj@jODeTyf1;O3Ck4&zkvZ*p%oq}|C(u#wA
z%&dy{*YE^T7R_&Py!9l@c+4(_isZW;1l(PPZoA{)`4&~D-tZ!M4Hvl7IBpzc4x9))
zjT-4$e1|FXDFHT-9WZ#yu)v^zGHVFe-1!t!7QoZ%10&q%uXJIGO^4lkdLg*+Bxb`P
zckTgkwrR*$;6|8P+jg>|D-6BLI}6cuZJGi8+tEJG0QzrWySbu7Bf-oVKdeM4NjJ>s
zOld%;8&+Vcd$;NCn*<7EyoR$cq~|UngBUwI8PBT8>IeA>*}%Qwe($WjdZp!bxSYT^
zGhn0~-{Wkg=x3u5L;#ma@LKIuaF-kWw>|mUJrdx+aKxk0?ppSp=$VJ`6gR6TTVp9%
z@>8`hzK}afF?eU*u(L1xTUc?ou#BXc$sM6LbO-gDNY96JBO1YTAN0<&w)_8i)g>Ky
z(rOho_P_b};?0AX??%O)cQLl#eA9T@Z$$jEvD66Ac|hkXL(9S*fPwr_L&!;rD(<Yn
zh0*@z?y57~bcL@eN}K2|v){(S&)hltGK5<b-ssLTMsF4HllL6(%Uro7#6#cvqLot#
z81S4`OyLX-;7}NQuT*d#g7n)Xu@0XCQ6&prbsHO~EO}1kn%Fx<2Z`rL<Qt~kUd^aG
zIoG+>&HDl(PJ3G;3QsFhExi!!?udG1>&s&|YQc`~ZC|aYYp3Zug%1Qt5n`3Q*3-6m
z3pO@+O=gYS7*4rJ^MSJ%{*Y6jKgI>vp=3?9*k)|N){Il``vsr3r@qtJk!Vo3^~rAy
z<~fWoM@6+rtjG4Pv2Z01p8KB+O}rcTM=D)xOvwCI9kG#?SHiTh+63?+N}`y`&z3bY
zy}@yC3f<J{fWF}-a=%F-q|Ej*l1zWwS<9DC%TY3@`||qoV!|U5i?$3~E)4mIJ@`=A
z`R7S~@pkxFsc`a_9g-x9D|{I;%K#085%nRNP+vHa24Vrs!ecAPfN6Lvh?W?(<#i`n
z{-^%w#4}%x{4-HkSb>KO?vU228_stMg?`=GU%rmAF%!J{au<9`q)7uflB$}Wjd3e`
zci`it*r2*3cO-v?&uel`8J&%>{f9j<F)6Slw^k!vZt^6cedQp=Dr7Ae3QhGJ*-uuV
z#s<zUl4CbIc2<L^De56vobBB&G$5<$S$5}A?#n=KNbHo|{(^Zc&O^oUZv7fjdd`nl
zny&XwSb+IJmMfc_*O}<^`7n3kmqF=YWpno-Ruzl42Z?MOT$`zx0T|KHbulU-`{`Nf
z&(@46`1Y9$#f4t`9lWQL3c8hkA$6nvD0-_;0cc6vkD|RjucB(U(_@ddbWWR`N*x@F
zBiizF+Yih9Nb?5Xo!y6s4Z%-KJ!=nl%Px4i-41HQT54tDY4|Ou!CbwJ)92EBZp58~
zy4PPOD$kvRD?I{bp=uO7ZcO?{vaQd{bI%)=3(su30DbV$_iWkqmz>x?oDd!%ssUe?
zHHOy@+k~T#dw@717aH&&fRaA43|@mQ1@5I9jnk%Ju;!=K^7NUV&U!eQt~Q5ug1ezV
z==NV%v^L8n5ulTv>kcwr4Bq`>tP?9D)JU<;BpSWXP7wLVl6)gzpdqB#-l&k~(lL*_
zUA;zsycdxSQHFLTQAF1(p2ANgAVpg4=EYCfn4dXS8>ESo>MoT;R^7H&u&%!3UoY|Z
zx&g@bAE9I#S%aQlJJlbHvV&YGok2xaWsRH5()NJYJ!!fj5uW1am_09-N`KoMyFs^C
zs>e)KxR{3>e+2Q|z5vbLePdtvejW1T3N4HAncf$jHxo1Z*s`3Z@7`-xHPu3maOvzk
zh6}|h=bUQvoVJy@gRwVJ@B`>hM0d<$<F>N(R9lI*tN-dL%i*|mGXH~MV4@lK1(O@u
zLo;8L`$4RsjyWgKqU_toNPo!L>g`tt?-%AFo2O96*G1ggOIu{~dTYS%Mi5>ToYP$T
zDDX#fhQ!<keu_vl7IX#!(ms3o)wHd8O=q+xP}1DnYiR1!8Xjg(o66{%n5`~P!c@5p
z>AvoL-~8r6)_7AJ___?$T@~&5-Un|D`<5Teo{)jiL80WZ09lYAY^myK*-vQ#&3gbQ
zrwo$wyuiH@1}D^0Rzj-CyiaUHywA#GhnY~=ap<_}XJHW@UiquDlr%zM`{24noMp!X
zZqXEX4fP+_&7_;-NxS4*CrMgZ4wLF1EWlk8XC)#GifySKCrAU7s4==SG&<vE34D1U
z8q8ZlyXj4gUYNYbdK!A?{TZq!`4&7AHV56eNS{~1Jph;BbWAx@#3AwA!PjMx{Y>)j
z2bxe&i{R=P=z|C;8k@hASXKM5iAGl7wBRN7U&@$b$P>qWo`Hi8<yOm^XmrBO$x)4s
zzf<Xv0pUV|5IY%jYkqw&D8b4)G-^qG$!@hh^Hl1R;hWfLq5&Ry6c1sqHF;DWyetd0
z<PB<M+DQ3t2NILa=3!q2A#+j4Rr%RZ5vBSI%~<_-F-F7l=t}Va1>-(UNxEt9g-uTs
zIdaU4{myM<d+Nq}{lO--L#Z+{H22u<J-(@D!8LxstUHrB>$?p4_+>|WJZ*RuvV;~*
zUwQRs=~w64z~^U+eJ+DkhI6jmVqop^Lji$C&r7+6kc9AL3-x65p=ZP$(o@^no~3_1
zM<F2x8o<P_5~W~TSF?wC8_b%tWvyk(;&&9NW%jLx(9Uv>Ps_!r0_tD)@i{)5U)@YN
zaj}28*GMG`XLXwO$OS0^6H%FVi_9CcZC!G*8d*YK!s0<=?ZjsbX3z;-SOMq3S%r?o
zXJ6|1bewa}(dkeY34+>nvwkhORx9V~?^U?=oE8wd-nw{8c}{LvyOZAFkd{^LkRAK-
z>7As;SZ+P=TZhByl|AYpuem{a0fX=FFPc92m!SsH_+LYfr2fN)xTYQ!!s~A)g>!#R
zSHETF6Rmcczx<l6l5TSUvX?+GPt)_Itj2L#w(^3b)p$%+*j=}=!VXsLRgsgXY_eCC
z9|g6}nWlmbBSI+s$3vSVi8o+E?ZlgyCmdqQDeIqYpmS{0l@=j!s*N`82&@M!mpyS7
znzN&cl3kSImh%ZN34><IQ;;n{tlq|2g$4`0Ym1YrN^?JQ?@Dv$LBXr0>7&ExWY(L8
z!N_*aUUByqZy~sYr68~iTf5^@XJb$3>~uM6Iqv4VJbryP2;NP!{h8X_$z9E{BgTZ@
zQXU5cnog(byF^=_6DtBqpvlHf)PU)^>s`RIYv%UnlJoIG)wpDON)%))2W3Q9Hb%sZ
z>GbwiF`UdVLOvZs1PqS^v1(m9JrE!BqHz4o3%*k-Dh@)<B0`BMDifx8NJlnq1YY)3
z+;p_Cli`Gq<V}qY=o|QDgwPtVJVaj4SF#x>8uew6p~=r?$+DzlK7jKDu&f6gMUCIM
z$o~NAp4}vgm~YwLQnoX`vLj&f<A<jS@6HlrveFd%{z|~%z}9y@(E?B>Nt<-OQb|7Z
z;oLt@b&TRS!~bkTK|$tKQNwHUFtK=A@2_{dZ9QxI>_J(EU=Y}YDz2)%*+h$ZxSuHq
z92xhWvVYg$IhaW2GC0Gr2%R&1Y3*U6Jo2(+8QN98$Dz)W1emhEDyJZu;-F&=2EDML
z8l}z6QCcEVIf`u7DRSF~crx$u)9N9zq&q~HN&C7?PJs(N$GesD7l$1o9~!o^hDDeC
z{ZJ!ua^%SEAc?_1BypDbBN;LW2Bk^uVEKhv`koH;0`IAy{*b${_o%5OXsdHE=bm~3
zFZCW_MGqxgoc99b!H4nqVC8#@6Wv?f!!kWvI;>yf?)uySIDu>-$N(-*-(})#toRnx
zPRm%wdnXSeV2Zjn&O1?Y+GVNDyubO(i=&|`X^}~hk*r5pGaNAden&lYEEzDD!hrse
zj$27Ss)@cbbOTO6U;~(zl~*C$$|gcJDFWdvYQh7WcDEJ_Pu3pmq}cLyKI{?;u|M(j
zPst^0K~krRd?2;P&~^63TU}F17Fs(1$Ff=3WA75cW6*H%H8HC2fBmuAim8VG@u(!+
zTy(SrCdQhy`wMMGrxX8N#|~6L-8uLD{0v%6VeOmsapJqI4ZC_Z#xns*egm*J8Odt*
z<OK3+CxHDdDi~Mp2E%Tdl|wpbay!+kpz}(yk$UyK?|cPZ&MmYTaw5znvYgIMMpwg$
z67cU7!!8uuT>hW85y_d%lgHp)RK5~(VuM47o)-N9UH7$~?;RCdVZEo$4k~i1uJ3#S
z*bFvYfX2qW1^uxX8Fhe%6V$fyuz3UsO(%kHFOs={2^g2$iVHQGLe#GED0)>Sta(X>
z!auUaminF{s%Gjc%J1`cS)gpF_!RSi3tkPR#^DYRfugZu$sp|6-co&NBD5qds@|!q
zc7ZcXRHc8evgaa$(0w2=S72Z4ds<X}x%9LU<9jwfAr(s62sPCbzH89uS_ZtrP@No3
zN>APzf$Ebr5lwI$BA!9gmzOK&En*b-26Vk6<sYER5~wLEn2ZY2&xa=FRv;nlM>dRu
zcjVr+=KP*>=cYX%;oh`Kwey;#@{9r4RqAem2X(U2D{I3^sKBfaXiv6B{&`U)4++Mk
zHTB;-KyZeO6^_btm$Zg3&m?V1jS>>b1@+f?--TY$Ap;L8`p<)!c@i!POBP!?87k};
zH(qilQHI+AU!3K<u8kebo1n&sP*jDlw6kLJ>Wgc;-cl*anojt@-UouoRG8m^^s3o;
zH8R({UIF~EbB(hS4(a4EcY@|&U0?!b$*zmWeGJb0up^H3<aFjXLag?kMg(7Z62l{c
zIFH<nqs-vOC3*&fEwm|*svEHRBpeSI(RricPY-lO_5t@6mV;KSHB#bmiF-AEo?N&@
z9ADBi%(?3fz=Nt9C5az8f|B($?>)O(L7ZJ@I_P(z!g$nKoX6wKZUc^+7lY5LdV_)q
z9e$jfay5IaVRPE{#}q>;BI(r1OLu>)M}LOPQ9gD`<(4)|D7B1w_)!=g9S75gvjtj4
zXP>M2!Z1e@n6dD?3PN&F@5nmj>wBH1*)MWvF*+WIY1Y|Z$({@jPqiLqcG;#9>5307
z$+pI)u3=LiS7OhZD*2r#`twZ|;$1Vxu-du)<?Il^y}$}!6|ly?IBk1-tOPnytaF{q
zmwaIRI+pi8t`KI17{gYr{|JUNG*(NAHQ$rG4!8w+(q0<c+Tc3J6)wtiW$oi->-UG5
znfl~{m}FB`Xw{YHy8X2XPn|`OEkzm2Ep7B@*YO6h3Mq_u0E!|y7pXbAE62wAdj}9n
zuP|lyQZ&PYM{F!9;;k@~2*+2azG!qcwO@kH0h5<_2R`;VP=@+TT+quDSN-hsdSkPs
z`RmPZfz-syyO%=yYAW!>8S;>88ViY;KEfKVaugqSgC^MgLL5tm;vO%-!Ip0n<538B
z6@FzFw^kpSoy1Mcme$MG_HnhF_QEq&Ely2=6Hi<@26oZi0igVa3<C&gT(MPr<-WgN
zJ4d~FSaD0B0yUa`D1aM}-(+sjOfBM3)G%jwQoUsd@Uo@(mNwC|b{HPPPYZ2d#x#aR
zo-Q@OlgN*5L4cBghMZoBV$WjvGoL*|eC&k_9g{!~juF^u30bURk&-Mj&owg~eEHo^
z_>z&jpk@U^J;YSiRSSHsZ{VGenOvzxV8f~<GW>FEoLAi&9n+Yng(Y5DAq0{C45_re
zGRhy6X;)ZR3?^B$80jCf5$sD06B`{Y-1?36yxSE)mpi_-$H;mM%&ZFSKmPI>mjM({
zt#x$4$D;ifbHhjfRS_#R(t^smy*T1uuV?Ip)|+lb68Lg8Kiq|83NW1Jmo3l>xh8g8
zGv##Fc!R5hl}%%)hq1yn>{xNA2j%IhdyY-Btj@qF>jOZU8LwVxEBV>RJb)~Agox;4
zJqlN{o-9#*a5{2f-asb<7VJ5rkmv8Q;6z)#1tlt*NKok`Aa2$f)4`M7C=5rh6Kxjx
zywc!Z8v)&_DP+OCJp5yjC?5Ys4%Dt&wTu8@1lAc_Rm=13!J>G1ip6<QVtwWayBimt
z7`8=?3~<?#7NbM=e%IiX+e%^=3#hlwx&$nuXjgpy_F3=qhi}XjfOn949_4-|Zf-N8
zxpQJ?^>qR<BN=k8vc1+xcX$CG6k+?2!_;u3NKHqD_XJ&QJXr)qOl#(qD*`N5!cdV=
zBpY50L*@s@5;aK7*=DXLu808Dkq5(}f!KB2jX~QH+;AcaCZFZHl{cS2q*5X;VKr5_
z=700`PfEmJVxipENx#n0RSI%z7|4PEU}P%$<&)(~naQs|#`A(|f(9l2mr$g&{7m~H
zKopTkBjvY~=QUp$#8Rq6wPuz5IX;p~2tA?fvhq@b?uy?pfymv0g~ttP*Pp{26ph8!
zPIA3~M;(WCDd3%_7(|2u`axLh81-AeEaxI24lFls4b=hJwlcL69+DJ<8rh1;=XtD?
zC&YPjxUauVeEs+u;7_@KpB7h=z2>+BxMVpC4-sC>Hjm_QFO4|gq3Mgm^<03Yg(l9F
z=GVDi-t|>Pzox!TUCs3)og8U8^{^lV&tXsbF}CIPH6=d~yS)ooyf#yZf$dE*jyeFc
z8Zh0501G9?-3%J!Ui&<ub3!~KaT;*-)MB+m6PcA(CFRHadRqOcQL=^xE2EeVqSJ0^
zU~lS@!@jtf08kZ+HA$sMWp13?YLNSQL+(>ks}k@Yt<<lDtse#<RpQ@I*n3lE$G8e{
zJV;WHc64^JGmzAidehn5DSrf|2Zl^J2?|&)a2cEOZShuREo(WSPI`O#qBrBnlWEd)
z3d*lRz;k9@)+W#5)2+gk72#=b4gT2}6c+s|YbW@ZGt(R)29Q`F^kLxN6Kk~gD7*rX
zNXoqUKbB2x#m}UA=pV~w?jvEH)h19E-pgI8GDs^mDeCBeC3>zOvYFJBXNcqv9rxk?
z;+LO!nZ9M%?Cn){TQ34eSNR&2gPlPg%okm!in7Ly3>sL7DS1U1fvQ9b3K{DX=yk|w
z(d9a=9jd5CzRPKl)F2m%_{H^}qhL@d;T}kpB@au}F@(^^F(?3{0?x{OzpZrp*}XO!
zPqxbKJM3FvLeaionc2y<+Pk}6H=F&@9U^rsL_&NUQ`Zx9=M9R6XuEyrjFLUaY~f>b
z@JSZ@Az^;DfG#p7OO(+2BDU*C^mOIgkDC6G7{6*mv|S?t!I#WDh?l@+0#3J2u&}y(
z0|T<X0aQ;j&!Omy(3pX*n`d9GxeO~R4Aqq0k4PLtg?~R|{gQ8#LQCmU^M!2yxZFE*
z?Qv}EuoEC7$+Bt5`A>wmeNd5>d>&pv$w|doGJTcORBB}Sbo9v8YVvA<QmLGXB%DHo
zspc!0iingLqS7?<iYEkfPU&LTUCWDQPfkNVn%rXtZAy;r@BOdvU_~y_@>K*=Ez9V$
z0K&)7oaM=U-nr_uGF-5<{3C6@&z1E5Zc9Y}2sHoLZ^2obT4$c|CZI)cG`4HL`nB@0
z^l)8qhcU)_K<pQ5HvD&fO%Qjr=IY`hboLe^Gw<pImiySAgeAFAI-7mB^~ZRG6<^=D
zVS{P+UNY2au(MFu34$jjVppO|LGhDbJ^C$v{EGiwk=k84ByNz&M0WKRkn*60m!X8?
zN`7E#AAFEEpMt<1%&7hdVCURw?+DP0u3*$fSI8YhbKhUw=I9;-c&FAhin#WL^z==P
zYiZTGFNSup@|kC1yemvSNE;;!abrQ7ljAwzh0m%N05NI^zX_s`2ul&f*%Q3cOD&z>
zJ_=q)@JI?z3@{Ztg1-4ANINXx%z$*AePrgHQ%WNfQxkz&0X?_HIn%~%R^LlA+Bk~<
ztGs~il`=s~s9hMm2~M|Fj@<w<jZiQZG(B~k;e<DD6`7B^t&RsnLcQbYi}fgVdR`q~
zVW7yA!WMG9geQf3f7rl$8vh&Jll<lSxo2YDhQfpXzbWXtU%5oX_pk2lk`eMD7ZB?d
zgkz9~$%clH4L><a07=)qKnPLFA9Ky?Wb(KE7lDAI@b?P8?*Fxyugh`Av0eX|Z6s^R
z&I8WdqFBK~*YACf$LpLWk-3xKAMFl1P+(6B-=RlWLfmPD28ekP32wQFgaFMEF!UF+
zP`D<VRhLbkB;#Uc&wJxwXfFdpLEywo2tu3M<Q#pad==pr8pyKn2=ssxfJ1U0`p7^T
zG^PaJl3&n2d{LGC5RCI&+KID55$m4ds6AZqhd(I8)9QmS!&12_m!=jV<KZtgnmeyy
zWj4x=qR~64fN;^C4&Cj2w0RVNxlGXq_-6V9)<*8wE)bdtj2|D9U7Q)FX}hHSiW<qV
zAPEA^tokR%82i8XT;N7$nA~9@bnfi<4t3Cg!U^kEct8jkzTjF-iRRucBUDXYc>TOg
zd$RG-*1-Z}trY#*BeJYzV;zdo?N1fZaBDFwo9Wpz=at!+aE@v@d?xRPKbgWj`R6n-
zIyPG;0IKPorG|$_NyLNFuyyDxee-Gd5soe#9YdQ+r-r^1f++(6e3WJt*fRJNj`$vu
zI3q1k@Xs2@U?`r{xzOm634;kR9{^HIn6T~8e2XuGCr_0xTyPl+rUY_J?QUve3+0mv
zz9Odo>~wYzRlo=BWLV9=u4GvD>lH0Gce&1s+(gV9URyL=bCMb{vApeg(X82svUXPI
zYNp7`zph>bSuy(X2~m%9>?K@r=tkscqAeW3Hb&j0kOQV9jet39>=H4@nN`P!St8mV
z<2jv$hoWAHF|gqwsbLc*J}mmYZ7B*nh`@Sm$e)Eo%>X*gx%y%_%1?7uus#~-Dg7GB
zz<>x@+u&J<yfc7@*X0e#(5&<Vf7I8Bx=5$Qv^z_UTAY`nH^E40>s5jOh6-xrJMcT_
zSD7p3?DfFQ1~$F=I1r6O@tq4mRO{8t#=I}%M;Q~@jsrYT%lqJ$+m~Nvg8Wn4x(Sl4
zs^|3jZVqN6{O9Tc<M3(VjaL0w<3(3NhOW$yG(Q7S3fsC3^_l})o@@0rIYistcHsOy
zIIAYG8b&C|t}PS-@G`z(-Qg%JIMKF<W99=yLy7(bJxm@l-js^ScBXoj3)&cS><cDW
zW!Djmy->i1p2-jLp&rf8WFI^S<WI=0P$|nIsg}d5K)PzNF?Rm<{#|(RGD&k+&sdSQ
z0XWuk4E<qUAj!bZ(&WkxooL6t<Jh69N&g70Ygc|xMR+mW^j}QjF2xi7E3~*?MJxKY
zRQyuH+ik<N+}{stTn#fNxOeP=BX|F9YYuYKW?C)7dY-@h{{HlSVJC$GfJlPftYN%c
zYyEYlno?lNDp(diHKje(-nI^~8Ks$Yy4+BgE4M-@0+*{g4J)`ofQ5lh>~yEWGJXp5
zHP5zs^-m2ZwwTEFORWsX;X}P!4kT&QIodOg+2j->i?TSmX;gsBwjHp~K61qjPQ|~P
z*#qyW&X&ZCVU*|QBiDh9%?$Cj!Ndi^qz=&40S^Rn{}djd(S00fx5Su!YE93s53_^Y
zIB@K7pg+p*l#q)KU43)~7ykGT>nzY{t;p?P=74m)YcQFu=EwnF7Z`<eI5vB@JApqd
z9<V6h4Hvcn;q%cZT%<;P04x9|_kaO*0b6~iDVCT*00P+#g-$N^5q6I_b^3yIfr!ER
z4|wo*0p-F+vs>D<YrvI)zt>KE{PNdMu1|<$Sv{6VlF87}9GBDGu>qkJbj{jb?oxAo
zzXtYOLF6>2<aBFt>@|uW2)f?=ho<+Qfyr?0-!wgKiHhI|##G0q^EXE7BxQbk9e4s(
z<k&Qg=!GnT$L~fhJPy|D*?!-g;6gq?%b&j1(_mdT=JS$+orb-eQMbXVNU-tN_8L2l
z(A3@1m0C6eMY?OWYdZE&rJcJDMIzFq6z@?&mh3yiV_r{i2Fec#8cq-YSa@>4)*N$(
z{hQXj;Jrr>#Th=m$vcDb5}VziFzVXA64oOge_rzO%Z~KEZApyt5vK%mkr1%{J*p<h
zDMmL*(GH~QvTov8oNqBuJU;eOTF8Y%R0(JM@yzbq9PkI7J7<7Qrc@kk(KA95R}SS-
zBNcSJagb+9d)?Z{u!&(d^c(spS$uRAP|bV*34$AEc@}E`j%{Ea!!F&6Yx{WdlheBM
ze%Z9GAi=Y7g}B^N#dC=`%dJNL@)gdFK|Ak;+Fp0s``g=f_f_E~c8R$zj#GM+K!FA8
zlESHXu(Rr3NC0K;k@2gflb(%zlR{MJ2ngMVRFYL%PtPgPZAFgj+q$uH>&j?Qis)a8
zd0h$u^0GN-kgB84X#oH0_v>rOM2eX3$Niy}*fpjBK-SDH0Hb)RW0^hAFxF5E{!Ma4
z;Rft~9^1c-Oe8!r$-d~A^6}+3`*rWPCjB#~=vc&hB1yxC>`u>O-x|iOeKWuM-fHHn
zeR2o3X=?(|<jTe$;>af!3L}Up88Q$c+|7EHm28mo;3nOWnlrX0i6+b2T`imD>0DX^
zjTJ`9xf3B)MBf`=Fps0KA!~TgP147Ro(F1oNVnP#EhRs|wlh}Vy3@T7LL1%-?Xt01
z0M{_1dtUtrZQY)W2gajo={EWFv<dJj9mwVv*EX44nqC0k;hGiqXE~{bXVTIi`x(>^
za^}-a!}>h~F-Jdq&#JFJ{~5TN9e!HYP43-~Oss!w4PNl#=1wsH|5t7RV)o6QjA(^V
zt@K>24QKK4(nHE)4mT#)=OA`aC>X6(-a1#mc{XLA+*v|T6Kl<6mz_;H!U$YR`X;qs
zI$@+(m|^BJWNGe;*nJ3%kel*2ybkH7V@=Z$`vxtIg2>0uV+&>3*E`gO2TIQ6zlQSi
za<ZL<kYmd!c8r6s{iSLz{7s)x0RSRkUIF;a2V6;<s$FiQL-{23B_+*y@FC{~C2z<-
zb8vj{FW|tb1ISgw%70r|(sI8$t+L-Tw9DTPI?Ogoj3HO=t8dGVfb{NkF8(vxO}v%y
zFi@m>CT&Y*2RT|nT#p13*JAz+7aG2)Qu36sLa_gBaPr62grz%T-~%oF5ZZxc=OCA6
z3aT|eXSwjB^=smH-gqCm#aYgdi?gu+gCqm5iFIAjw7`-OO0LwpXhK!#HV7eDPhE3F
z4OuRWibgvqPPo2x!ml}VKOQ(5l}2tQU^(;_9&k`vOotrvo!JD=wvn$1Ull2&;;;=f
zEDzgf5@_c+zVUodOmpfT>fZ!<#noZ5c=ndOGwPIQFZl$H+f&{c{kmy0RtaJj-#zg^
zE_PTd)zW!p3tY)RM7n`x5ekblN?4s#RK(-WHZNcPIc&y@8oLm5*Dq*eC;0ip;Lk!^
znq#wK%CUh1;-D@!uRXMHU4KnawDpHb)E9Q~OUIV-AW!ff#VvN7q=}he$%2|BJ}1k2
zVnx1jVwQ@4o1N?rZO@Tc0^2q_0YE&>suS2znpSY}np2P-R-F}k!^D9Xl7FO*o;NXe
z5vDS&A9-}FQs$YF&TrRJuJ9$&Kwr!RFfTV)M4k7=MX~!7tlHj#NfU?ZTi^Z0R&7jl
zci=h7c82OF6>f`I{KJ%KK~(<M$NYuHmDFGQX!EhJ*}8aRYR{LvF3z);BzCjhI5PAN
z=*d0jzirs)sL8K=OT5+aa507TvgEMsXd6qC%xQx_)vy?lrZ8fx)37NlG*v+;BdHtn
z<kohbDaQ%7L$=<Q)bBD|+N{ZidI5{V32YriSGvR2Q!tA&EJ1mn%JQS5#Cj^+>#24t
z)sLEO&KLmF)}MTl*$(dl{-MG(@Ut=QTRH?v=++)6nCnL(_So7mr!F-uw9E{VCjp?G
zpb@P6``Ar|+nYR_k1j}4Ah%ue7M!9Nc`KI{P>aY!)98SkNA1Cu)uh@BBt>(HLV@7<
zO7sCB9exJxrL%3XnmSkRqPA7BAEy*~Kr0lZ*b4|0oBi&i3$o_B*-xDuXE~yBF&=?#
zv!nUNH`KO2^Zv{^ST2WG3`LWbcs!JYk2anictFUZ6oRR$u*0NqKrTbHweUESD$vB-
z0VgPg%RK-zY~H{u42=iiVaMORAGwy30P=uzy&(+zC=B2VWmJm~f|EYs1fg>%LbXid
zR}{mW|9YG4zHHUd6&Ms26;fbA>xbN+b@$ADK0|G{Uw<{{ShE{_A)Nhy`*rfmfA*>0
z;D&}Zq@iKtGOkebx0><5f0?~f=c|8Iaqg)3q14w6`)ajc;?;-Mp<Mk5WRy4e!0hRt
zr}<ptNhpmE8?6#3jh6|ZOd0gNL3RRiOt|ad@Ua?8Qmc{GB%$pUFNXWUioK7l(um5v
zmb$m(NuCh5zzePt6hisfDb6fB%&QNUBtcxua%25e5#wp;xa<e^bGJC!J1UNErkT-%
z`^Dm>e*}IswO*}UJL^i1?)9rt|86_v206S*oHckZ4nTrdoF=g|(KnbK6)((6FB~JA
zpFM{_wl55GueaaNehob3JhhUfUnua-Gi~d)npI)w-dIx*>ijrF5aq{N$+aoJJauu+
zLpJ)m9&z^BaPS_9xBvYp-U5ak3md{EGB2^vbKJP}s8Vgx2Tl5K6`a+ZAH7N__z&zE
z06Ar1AsZnyfaO;akId1fY@)a#;9!}<ltp23S6m9K59fw*MxY=b5DafMvPOakk_dQ!
z65RBGslf&$f$Ji-(z4YUk0z%wH<(3YyW&|oDU=Q#{`}OZc*r6BSLysplcbXoGb4S_
zK&5O8)~XUf9@7}LrRZjn3H^}+oW*+^**9P0jTh$Jah)pyu9e^Ln7{Axq4*tL0Gh~^
zeJ}dL;-R@J&$Qno!bW#mPt7mGLjBP!V1HLs|3@yPQ|M>Ty*3OPE+-a3N3A_sFH3_t
zjcG}BShOh-1zg{IRW29jOt-Kbk%ZIH5m|8Ik+H{Eo_5HbD3+F?@bcsB>_OYIjrFev
z>6}Wm!~4kQ6G}uiZbl>NEsi(!GB;60C}SH-dsf{H>?P0vn~j^bj#uk+#VyI@`=9Gq
zZ2~G8#mvoHrukMwzFh4@+s2iOY1>9OV;6r=d~Kje@nEv+yWh<8w$-hlf2a{@>MYIV
zJnIj5_@qEEu6=(Okj2z;U*<e`n5mIE&7H?fend2NR$2tc<c``}zhp;qXE03WgFo26
zOb!jD?eEuKc0=7~q@z<)q7j5kHQj5&YjU@)QxbWJ89<_#(X2{?T;m;P-@wHzUmlRp
zN_p&9Z8xy2%t^VdJi_b%wUcE+;hsAdA%{YC#0~=!`?CH_GYV}<(w8rDmmrc5;o-tA
z4M@PZes{BrvW8biUKR$1ZTWTUqo1geiYe!R%m=wVyoFsUz)_ZTQ_`SS>z5dHX{h33
ze~7%(=a|7Diei@Q=J`ubw|-IH^o#|SHE9RmZmr9|{!>%sq%F+Jw-?KF5?{T4nNi0C
zrigHm$o7cBhwi-Q4&eKN1-T5!u0xz>uEsUd4P=SL!zB+YnXwN1rvb@4osw1h>Uh%j
zNvsF;hn`jjreU&@bVyA;WbPyw*Imet`{4WRpdI+^BKs@v_O>6kS+0u{hcflM*6?4t
zZEO_4!O|e(40ST-6a^-N>!>}3k<p2hZ8q_oJR<&a?>KYk)4CPqkMGqGx%_B<?*M|v
z>@Xf1c+z?0x79-M%iEps0v#G;`x5gTwsM^wKQeSLqzNqc+UNaaA65r_*^OR49h!Lj
zh76$54taes35*%rn@Z;yYvddVHm~Wscn=ckyu3iKtY=BtqOO!JUJ)-vCjtXDV3>wp
ztaX1_Cjm|1ic+BUKCwksndQtk{$O0W8t%($>$beZEQH{r=Z$k&roQpr_9N*?cM@3s
ztXcJv&yYHzhhqF>$)52Zw4(WzcH8GBm+A)sWEVAx{uNka_40RMiR!A7VGbBgfab?l
zVQzOPE{v|l{+YPxkyU!-j0Hnkw06I6St()mWT_u;C=>^i7UmmZVgBb~ZLf_DIIq;f
zSS?P7q~M@xvPSSS2n)&Ny}&Pzrmohlxbe-=@gRg4cHU9$<Do(bmT-Oi3xsrDSz2h^
zLiurc(v0@Y!$r<F>;dRjo==rv9SiT<dK%GAk(^c^DGS+^CC&~Pb(Vj{4V0S2!CqZE
zJGR4iCqUokMy<aG>n&HSF|&PHjEDRkaUJmTo~Nfs3oZ2oti_no-FAxpas_nWLNeS7
zY;lTRFd#E)c6?>$hZ*X}9^eA?pADN~o?!vVuY=saht$ZpN)6v0hP^q@uWVvXi_O!s
zKYTHL_c>Y`8vsRxwz;`$tKrG8_tAmLTWyr5R8rL~Z>fMWVKXNcHx>71f^8btDO2`W
zzz$YTULsKQ6vHt65hBO89a)*l#4Uuvd7RKqK!x63_lyAI+`sD~|Kq<pnjL=89J3Z9
zy{&9p9+fSTQ(*P6N)FT~Q#AE5%=J+~UMRkLPyEUCmyryrf8>Rm#|xk2fXuVBXWy4(
zhvMI591OidjtLxG>wYY6{N1kpOVBSFsrhc9zv!QSM&K%vLUeIAZEa3|NHvP(Y<Y4b
z9aYpp_Q+|O>aQZm0LG2S3_{G+F3UB>)~~;Sig=QMNRk<TbIgfWmI8ZoTpFfgqa=Nc
z2XN0ieQa)`kxInulM?I6g|YJB*_^hO8wLFJcQQ2Z1c@UJed|B;lw`4<%v70hp#|b>
z&cfCKKrfc+G-@1S&qx?6u|R!1dV96bf(DKLPFH2q?`ozjPlt?iX|Zv~yGISH+_G`e
z^Q2qrG&{osN46iRD^Ji1bK`E?5IxAy6p^~5k<*<1wC!45b0-bR+q?`UFdDL#%jFo*
zZCMAHp#85laUv0k5<@EeGGBKjQqfO>ieIde%mabNd&@^dX#KY(_$aU@LXRLqFm|P3
z^B4-UQ*;+Cd52T;n2gTKGohGJlyi_Ug-!+Bi+EEk16gH`JOP8SzGSV@0v4)R;Pj)#
zobZ*5;Us@fJF7+`WXvmqWx=sGr&&J5{x#X}P9V(=yN_8F2@QP~In}=IPJ3$Vgn4mj
z*%<%V!`3gYI_uW|Solcjr9~3g{oh^c#>ca3QSTbECnQ{C?hr&D>3?zGAbVZGyW35Y
z{1ZOULG6@M=Md+(Gc^7l9d6d)7f^#^B?sATr*pu7aKPxQ^a<C`(9Y9I&f8Z)-WKI?
z8tCTL&&!k#Cd!;`;UVl~-6T8ukX{HWQnB04Jh?!48up`fimxdh(kFpkVq2DT$Yd{?
z;+j1+gWX_r;51O=zAQ$*8@cw-P$#x&P4c!}mH!6gA=jn??8h141*1Xc+q`v<06xt<
z8$en}=ICy6%sYTNQ)9VnkqdLB+edqwRsmFRHMR$Uv<{|_s2W9X`okmBv!~WtE-aeu
ztRS!j&7Byc(|K3w;Ty*QHE(l&41DKbGGfN^Uxs~MGx)SO8FXf6FcFTbsmHZemY9@H
zJc!J}Md?yfOXu-jO?+<UdxWStjf(^p+y|!3%bI?Wb>24$K!L<?f*a!O=*wG*^drM$
zP2-5}Kt@)h{7SY=rR82Tg(Y}QWnj;gJ3#DD=HFq3U&u*H!7p?7K~@8YiROrMr#s7B
z?bdf;|C*U+!rh@;3+p=4>Bx`w9}l<wqhU!3=_nF5CZ+wn^SXZL&n25&6Wnb^9#_BU
z@5y$GzI%3+c(vFu&YAYQv(8-6TPLCSu*9LlPQY7E`S!n0=Cer($gq))It}s?82PJo
z%BYMVseamYW58I~J8Pqe-dviiJW%8eeF+Z7hq0lRF89G%jShqF&><6h9kFf3PZeA-
zv+k(?Z&kN>)9?V~b9}w*i&8L}Kd`MO`#c(xd^cg^0wn3mcyP7(sbTNWAjFFvoiQqN
z*Nd~;BIFijG0fa>h6**Zw$KN%FP9fK0B3Jn>U>qElN@YI{jgo+WV>nKKnEK{9wo~o
zrZ6;pfC|r@sm{*#jRxgzAmFPh!SuM`2F$$f^l0xM0C>M)7LjNSFbp2tTUrZUZZ84m
zpNp@?X36~|N_HWr^5HRh5c>3b?}MR!CA<Robc+$K719cmqT#y3gc82~_yVYrwFn1~
z=12>M!CGSjHwzaseXz~DCPR>`jYGhf*)jIC-jXPGq8ixj#7Y)WeJDuF8(|wUcn+ne
z;=4#GP5d-<@()*C?}gIMFElrj!7A<-GKc$IFgZ~Q3snAOL>ovZ><1?S^r#WryCBN+
zedPhrg|aIU_CFP#YAon9yYWw_`sfY;<Wc^Mvkp2{qen}|c_J1Zsq*7n8ICO1K%06U
zSVkz#a@%A9+4Lcg5C12yZ#CGg$1j{83Q)amm?Q&pcK9`DBBlfx&dNc}JM$tQoGuB}
zAXaKYy2xDLln;_GbCuwfe|B?cmmbySEPG5<FzA*EF%_c|LRxhcxOgX*F}hs1=~@St
z=m?5*2Le!ZQ^l#9>o0kM0aP#EbMSWl*qy|C?OcSBNqio`%J`zc*4(T+qYtW0h?Dbe
z0J>2qxV92}!cK}bsq|;EOJX7t@_9ck)&L;{hNHqIZVu?g<sYwM#(3iR8cM%Zfrs`+
zEl9K6tfT;TouAji>h)(wK$rcL{b#XdW^<$`fTY$0$sD6}4XZy2qAAalBXxQ<!D6uD
z`1sGmh5IeBV*=Kkxbdny9M;z66`GIZ0UfPV^9M3vJ>hsHbzJawTgocl)r$alF1Zm)
ztY>O?N`yMEYJuolG;Q2?>Qe}do9<QLQG+X2Y=j_gRr$R}OV^dwjf21TiPSy$W4VTr
z)~(+&K9dsPsJnpZmyv)d>Hms^Tu^2I&CmE)xtFuD$G5B9T_v>SH~92Bzu;Hfe?q*v
z<RM0&WBG@X1C*h=hLh|m3l5jTCVWl2pzhuyF-gBGN{@JLDgK@&YM>w94(i$V$XGIS
zgA)vONlJb?*I{F2Z<a*2Fzrr45v%5E-}g1H;XnaW{ZuH4#QXLWbHU*7M7n=ES-ga$
zvjdW?e={j%#Fxe6PRj|K`8E!6A7)DozWqe_Vrv?BC~bDu7l5jBQDQ-WS<Li`%veLF
z!L4zqq#QGvqL_gcNvppixWKs6)ztY)!E350jb5|2{n3o9lyPF68|XO}oZ`dY4l8kg
z1mL5ILpJ<)l^;K<hKqO-e7ls-gz<IsDA|!aJM!PJ2Hc=SbJsM!FMT&@%=TFn<u`jh
zYD`fIkGDBD-_mJ*n?X14p(a8o@jbnkB=6YI=TU5wT71KdjG-~{Qna^#ti^;$9txxo
zC;T8axgD&mdcjy?MiOG61B`}0Y|zi95K}7OZ*b3M5E7CEI?|K|U0aR(ZaSH+lX_5*
z=YO3hH%SZeML`cZV_8{6gXsYN1nlENB0&(>F}BdI%wIXgFPit+6$<Kh8%s^!hgFsR
z>xK@lE)q8Vk8A$@uHig;%Hl0;3P4vmKcR>&db9+4I%KX411q+lexLt4Rovv%`7V7K
z(nWt>5HkeuC_e9vwY~c{9z$L_z?62psldeOAwZSc?j6OZLQ^Nk_{KRH*xNa{Jty`0
zZTnBalJpzHju_)Wv>LgOge3-&;`(0kizHjyV;Z5o9_gEGA@o_xS(7J&C>WQTw+@24
z;u_6o4)j>UfNmb``c~HFMAc!`D}XuSo$e0p3Q8>R<5P3CvK5m&m-UqlFRw84xuH{O
zQXg~f(KUfJOZty2i_=|zXx(nzNBThk-ud02TOh1>@rJ&ZgJd0gK3a+#85fiIMFQN(
z#nazV0F|8&aT+^M7G?pUR~C(5G44Yi#96-RWx##FFFW6(bQB8DE>}7SfR!)sq5w<(
zMEN37ZG)^Oj53y&A)J$UcpH1}x@HY|(#6VxIVi!XgA2>$(mDt4gcF!Mm~m;%yJA5L
zeZon9pXa4t=b4BAfocrGA+$t&QXcT#p#@4{$p~*?-{f*mo;M^76YPrqCefbyx^8S(
z_pk71;DenI_B`miU7^NmfF=Zq=@z9<!xD8$Z*FBP@Eup$`CE*?t0mRf?H9Z74Grxo
zzFuY_KC%M%=}=(5!L)PhHt?M%>`-uwL#ng6kmN%4IBPXVH|q0&6iTv&c3#xF+y*{e
z^vtnd%1V9&JtV}Axm<@(P317a<3_6+%eTO}oDPshy+Mf)arvKe>cI}+BoPyWQbgnL
z%MQkX?On#|vDQ9NP*=X<{baf5>9P@=r}Qmn`<lC|&P448c2FLv7Wk5Vh{^6H`B#U5
zK54QALk~>taPe6#{|!VVE6B)-W<k@dcbK;;XXb+x@l<N$o#4o)>R+E$o>Dx8dh|d>
z0yyKxFvVd3q01puuNsjnOsae_Q8>yR=;hOv7fX(0<xNu&w4#_}NHX72<gC&(&>uBF
zCQqtSZ9#zSZ`Uh>@~P_Ag^q({O}(uX^e5sVaVK<+8ig#%g-9m^{xu8Q6C|~s+q$O<
z0ujFeM41He{RuqJ+ovE5jlO^<D_*1|afwbUw5NS2d<!CL{tmXfY^?+K<ByH9_$0D~
z^#D~m_tEvnF4L9`&c_0Xb{F&EI=qp15xHRI5ut}L=p4zf{<UsEr*Ox-*Ha;Xn^S4W
zFpO1Gy4uO#a>t0)oax5ktlS8Q`XijCb9E?b8h~~^CAeD4AYRPJJ~B%$IPY;TlA-Y?
zt03LEe1$7bok6vM`@?GhOv9$qGcOsc(q1%nW(?sE(9u)SIxa{X-}7{s`KJ3zIO+b(
z-=jNdNd<Ci(Rm1eZ1Mfu>6`QHEuEX6A~VKo#Tn3H)Qy)pyOoOJjP78y?Vi%o=@>`C
z2yJ?me%b^hf`b`A4<yM4j{#1+P4y;-T}`xHhXE{A2p=eV>J=6HG?q_(JUKnv>Gw$<
z<%g-)IQEI?sj*x!W~@)4nT-vHY6;MR@`p8xUzf_4-#4e}E)honvO<D6KauSXxG|eg
zchobDyqfNnER%>?e%Ha>tZ)i(1VMmyc^sLhQp+7o{7c+`Ce7p5o`J-nISN)$k{~;@
zByA(zdx_U>dzY5g;E;{g0Nkm<Qx`_adUdrJCc2DAMufX=T)k<ql=zoMkbMIN3Z7qH
zx9>%DQlfqDP6DBJ$-{HqhHt@6E&n<f;v9Iz{%BbICl`?efjkSc83SVwR!&#GISv>H
zgKh&eu~^*qB;kgs>`pS<xRklq9KEhIt9A;jZe&v)NyxGf#c~?=fKhusWY!7ZmHuXD
z*m#_u{OV2UT*2$o<6k8yynSYJ3Xj->iTyX{%dBYcnOJBS-#)uP;7{8ydQ`G+r2c$|
z+Ntz}%zYJZKF9&pimM#NQKOv_H(@vV>ATjCdqY6B!Wo?aPI-UUa;K@j7UjBj_pJpD
zu(FNkfdrT_R4l`I9+>d{1l@?oo+e1AEY4g2y6vqaJON->)UJox0lT}x_iG@Dh@jbv
zRG?FN0b-pOA`NpruS)=;#Xu<B`@$KSl?6QrGRz1hQtAp>4?_xCv-mMbx)KN(6dDf#
z*q1G<k4>YoTsXu7Q!EfGF(N3vbHG7Do5LJuEyEF#9%=#_QJk81kan|b{1?ny@0s;4
zB||%DL0Sc9e{q)CGdfm$voe$2%3T&&FAMgMi+<P_Z@$%aF{~?B9_&~CrB%68LDH&J
z{1rV29Og^LkDN^MiY|5It~nJK*0C?pL|-_09-RaI_fZbYi-DsY8#9t}owmAO@&ka2
zqnWWAMR|y%wKqm!`VN+CLT+T$ILD%M+t@P4DIc5mb*JX<M}U3eF+%)G$)ds!7v{U^
z;3d@knzaUrzbq1)O!froOktDf`LH+=e|RWQ^~3eZfyALTBY2%^y)TI@Y^@wAaccVs
z1_~p>9YBqn-_{AVBJ!-UrNgEl&H#<dXfG^JV@PhrUN*|*$)4~EpR8Ld;5-%AFz}^7
zkPBn(5GP@<DF+LZf_!Y^E2VX6U{7@7V0ieY5_sOq<BA_AWu9CB(*l|NRSDLtQa7`J
zTEzQ9_HfEmlFzk1NET>vlY?a_P4&qV<cLZqiT3dZ-V^{cWnWcZ)*S8XXy#&<J)SU-
zH?0N!S@uhU0@>rgn($ot!>?;o`R;F8BM1|=XQ5**1bdPq+<{ZF6qb3}zWK(t*B32L
zj?~9laFq9jj1BeZqpf4L{XZKu{Bvi%XHfbrUtouF7@P!S*y6wXj3TFJL6$Vlq0&xa
z$;Q*(xX%1yrz?y{pz3{(0IXTzud9=L%^$4;hd75+w{G3J@eS;NT^*-q<&<fJIO=X;
zhgM4E5W=q_63Ht=&jp}r?oY%pLnD|vQ0t4CaarS8I%Ft88^MNPb6`sww0+!UAT_1D
ze;kZttKU}8w7;#m)RH9&jyMAv4eO$?a{c5DzB=)<k_e%N4Fzna26ptr@YkJ~k?0?(
zaA^L*a4-Lu47m`gDw(_qR$EQtz&R|Xu)u~k|AOR@NwxR2=n2aR&^fIt_bojtG1T3r
z8gTIryxlo}W<|XXO&ri&KkI#@+^fU})^9*!Ia*DLGor{Jr7(aW7$EokvYSfgJ;0s8
zf$p_x9@*0G_ja0n4;18qA(zQ$jp8bw@#SzmZYyV#&@AH<f=Qlz4%Hu6KQWs?#LdzZ
zQivcJc;ZtA0Upi(nL56~&1y0^fMHlyYl0HpRgF|WkU6n^bT;8AqcDLxs|^z}+zz*=
zb{LNS2?Nd5qst)${%!x-l28Nzi4n<O3Nt;mBQadq3BJoGr1)6JzL0z1?sfV`+;~;=
zYgh93v{v|FUEi>nRNesmL#wv21L>%Z{hUlPo-V|PwW}-_$R{tq7=R`ACAi><&d>jM
zB?cM}*?PKuXx9xd8yE-s_z{p@U&P-IzEM!%iZGZrFpaGl(Spn<M?O1M23lJXLWRv<
zbNpO4L&nUD=|_e#C5kg!u0z<aoDh4SBPW>(zp1Z_u^!ESNCWMX{Vah%<`=f@Li~iQ
zo_*=;G`VVwI8^rz4fYd4-FV?oFVDfP!*ze?-HA`u{n88*H;F@3sbjc`ZV_sofbOQw
zhL^zDRCy(qX=+Kc$4Nf;R<Yyf5MkGVP=VlMgwCc?00F9+-eW`4&#1Y$gET)#G!;>A
zdR7uIPkk`I3%=p{U{BAlxgfl2JGiOyMT9HucK6Yhg(DL?tY|?k?RG(ZN=dSavr-+}
zv`&m*wob;t4<b+!>X18@*nrx&fcP~{zq7M)54K^;4kh^*5G%-L05f3x=)f+XEGZe^
zQ|b!QcDlH{c1Q1@<D4TMr*JaC$N(&^-OfHwS~k`VjwtCW8+*-v*4=Q<{_1}pO8wpG
z{9p8rVx|IR!?fQzz~JPn$T?YmbcN1sE9VzVG#6ONL?3>#(F1pE(0AL~Ha(^9vm{ar
z@Q0HYUcH`m1v280BsT|)DgwC;(8=-pg-~-@hniBBB#TlvpLROeC~jSK8p!K|YM5(@
z+-*e~U2dF%D%<T;7|(^+gPhTDGYq}+l^=ISDg@g!Ti;eJ*({9uO(vfGa(4BI(g+#%
zodZ3Hzt0(Xz#dEdW^!!3^doV7Oi?3GT>A2mI)eh8&CNZA<D<$AZM(T1;`A%AJu=aX
z6v#V=ZS9#|E`cxmaJjGC0ZJ11)Q;Vsw#`BW?HwHY4hhn{R(I(G`SA?XCIXl;8729L
z??F5QiW=F+z69XOoT*0Bv+ir?uVyU`Vl?QC0Q?GD^ZEr|%_EEhEuE5#P%2fXCrxdV
zCpcg%P1KRdBS^kD$kx$isPn}imcWm%%E+L;PXE3R?&qeq`fhy^!Vp;|;LeMg1!<ka
zuf(P3{9YBk*=2;~)Vr)$Q+EDrvG~EDUiOzQ&3}rN#VMR@ab?6%<&d)S{4Y7m6zQpy
z3Qq<C&lNN9xsI~=ln^!(72zWPmH%lg=KNPzpK}eAt!6FfZX^rxXqCzLZYp&bnY<tu
zyi|~}ZzlVfrs((n>rw!ydc1*jY@8CKN-SO!5gIeLP~Pc<r>tF9+F_w{Z6yO6%<NJP
zXxDMylBGnb#!PA=)<Y0ibFa|`yFuut1hm|8XNr~38IWN?+Y0)Zr<QKC<^E6SA$B~)
zpn>cw%5V<FikF|yr*^wF-sI;N4bo89(Xm6G<`y!mqyf6ixUsJSz$2pF>PwDO2GFMT
zw$aTyhT?!9!cWI)8~_+AL3vthbM?6u-Vac5_ewG6ER{Ro6Kz8-pYWXBT?i7vM;Enp
zy1ULu;nVKF1h_@_I>6bx6O^~WzQwhRuHwMOB)$Xr%*>f~vN~}JXvopZr;)NLDO#)p
zU%U7@qD&stGL;F6;%QshB{leTRDgO&>620dR}@!-scMZ(a{%?E@+_2pEqdL&<8;Sq
zt16Y{Qy#n+*tLY`!Hazdi-MWtfBKK)TfZDjxv=H!W*{q~Ae2r_mXDc!V@a8z?(%U-
z|1lT*WXCqk<JY%Y1B%#dq$2jidQ&#|HqZYJLj|*(*M{@E$%};s<Z-X5*!orMF}xFA
zn>F_auUl^H5NH37us4r}!hQS42V={g?E8|jgzP4JizN)nQrXH<$r7XNyAox~kZmkU
zW#7s&b`fR>sZhu^V-i_L_U(61@7~Ynd7kfi&hH$~iNAF2xv%@WUe{}3)eWd<cK8qD
zfD|yA^vjKv0(aaRG_kfPF;uB7I4tv(R_Adqn*62C*6KC#5z?t7X6_j>?7O67KvNC6
zCksP7M1PD-B}IOOJQJvU0$wbQ!9IzBt<!zZtZv85H#??5oBnHtb!2l7plq$-G}jWk
z2W;Sif_zJ29P(4Qnvpg1^IDZF7PtKoZ8!MNw<4#v$3}kX&>(U-z<Zf70Uz@HemN|$
zo%PiRa(3n^ARFY5M$H}ib`XI_s`Fv=P6Y(G-wW0yU~y~Bt&Hk{R~~M;ff(+)UUEPp
z(Fz=xj<LS}mSp7<c2=FoJ1}lu-yfk%MJvhB{z~D*aj>yuyrWP?ce!I#Jc|7za4)^?
zeH$7aK~AB`&QTak2@qa|06^q~c5(;=Xh&ao0}3|MY#KVV))c->{29Xe*MGsX#C7;0
zPCD%vWR+~J5+A}~OTR+}NCY|$%M176yLWJ57YKvnszNWorR1$sTxY6alurLkrg!=$
zl<=icY6&ilypj$ZsFX(h_W`opw3v~X+XLw`^{_`8!C<1Tg7y$vXme2y6#dy1anjzA
z47u)Bw=f<2#B!Rqp7fq8k?{YL5~)$ATVtcQV;y0m+~B@<l^lAd8n$)cgYtF`G*zq0
zpAMH*6;5OF`^v{iFxHQ~VKE_g1Bw1?8Cez)%QjCtBlUQ$%}mNvSmd~=^lQ&0ZdC)i
zjd=(`T}P^Y><wSwF<&L9>!*BoetE(Aym8jBoTw}z21yWe3Kk33`=5rrX`f7m6%w=h
ze8yfSDG*@9z>wVNuEhCdyy5pAV#WhEu`z|1;C~>xHSjo67*cBb3C8riYXBfY;cIAe
zu&g>pNQd7JJeCkmE(F3+skn7vaY<?h&p(~r43x9WK<TX=MrQ0}E;<*UD?utYF2b{f
zn)aeO%_a1dRmR%$CCnKDD8Z0fpkNRQI0q1Z!z=9>z*ggO11(1Boj|Vu)b`>VB_<)0
zP64uXiERP}OzRmiw1H5c-Q~PJczNzWmRA>FHWra3XfG)t`^$s-Wz7da`*>%}X!r9X
znvP*B3P=fC$TZN-NmREqTnt*ZuKQ1#Eg2eU1e8CUU7j{T3X=Zt^rqW^^j)St-?yp>
zuF1j^O23TTaiv$wsrB=Ynb)bGJ2KbV=zQ{egzSGso;AUMGU#u{Q>XLxF@#{tf@pOI
ziVfVaCmo)1Ee5vms}Z(!5jAgh=rpLzRl6S$<ryVe_Qj9$oNEAj!p1Y+1!moE)*@zo
zk^{K!s(T&VHjQp7-t;+iHtU=SsJivecuM}4R7~+@;nMIl#`aa{5Z~@&XBbI|Qd0gA
zYOnShFsEmH%Ut;cejs|p12BW{W857fPpL9Ktr^Ve6P?xJT%#2AosM!DxnkFEG@j7q
zjgE3=1=|tkx$CS3NKb=lm;7?oT`*^=r!T(4u1$cMl}cBWf=h=My~Og`8yP0&OI-Wx
zO={k}F)p>zYE2D;A#*+RFeA5WJ!yZiJZih5x3HpCO&~mYr_Xo?q7uoD1-6w;0Z{MT
zV@o&sbGve5K$dxw2Je#OSjY9dy?~3$x)eDNv6201WiPP>WjP_}HYH?~OD?u<|K+P-
z$(0Tkn`tGVRu20Ik3q=B#2tgS5zM-(VCvkijT`VRnCm2o-YWWEFx|nB72t0IoM%QU
z%<AbsCRhJd+dHOND7~$W39Y$0#&4vFN~c=RjAm637e@bkllFh_Hn%5s0Fh7Nb_mRG
z%E~P4WL;2A=-0<sf#!H7ufktu(WbU~U*E3X2>Ef^tXYmCwW%(19g)o#f&Pu63~@Fq
ze*6wRnDu#(RFtqMt2_5pYm)BwE>`$9iC-5EalY(bxtP}=8LF;Koj@s)kIpjIK?cdq
znR_n1PDpEbb=K@teua<@hGwjPNkc{7?jWeq@i-~5;(!~0=)rCqO|)IE`h8p-qoL}Y
zhyV>S@Y|tHO6+(YIHq!E?*Uki5NN|sp*>n2v@ZgxZhknc8vhmY6)p27wS!zx9;R<e
z{nhSAUEWA_=9#I*??j3B<_|m|vD}tW=_rHaJv2yR69yh%mg~ky25;k;=+Got5rE4j
z2a7Q0^FwlnR;$?H5t`LAA7iOgjB@GF!kTPl|7l(A{}3=sBEz<Bh1G?DV>^7_@VMqE
zh<bV;IPY5<#hJ04Q~z+M|0@xIN$AyIx=&a)xwc*;gLCybyEaApa-$^d^{uf9z1Sa-
zmBvXMc<@4IdDmRv9{!T@=zp)ZcG2HysyI?JX#g%TT~WPP=?u)G&0lp0cz$tZkg^3o
zn0<e%Lt%SNa?9$ng8H-Q$G~ZdxDOy4w;`qP3y*k#?0X!<<FQaoXkN*FOy3pZEqc_E
zA(n<zav0j}w_?wY@|$eH&Uqf~hbW9~t!F$5_S6j&bXn>cjb3{dn$UsuI#TdPPK!Nl
zD^|&&z40i{H2VEa(%0A1_*Av)+GzedtZo<RP<_dwn1lD;NWOj0p62Jc%%izqD1)pJ
zmESAbppYhR0qM%17%hxAHVQ%`sG^!Ssjh^Sw5<4lYYNw1)Oea*y$#e8X5B1^c_L4}
zYxrqi*#5*19a}@*<m+OaTTbr0mg2xJl7bn!1@#jaf_PDALmKLn6_+Wrub5|QPyxee
zw@ADgH84F+F#H19c=eQ>$^D8hM&fE@5&cM^Wd48E)GT00wgfE64*r@)ChI^}*&@s!
zAvDvbd%M^6b_+r=b9q$TZ#|E0j305FD%+R*zlibY<j6Fjhh!SiKa<I|Y`L5c2!rCm
zo_qX#4q4?uS>3Oqj1_lC+k?{oO{QeB9*QP{`^fu^o2*g-FDS?K2td*=)^Im(LI^eq
z9qtNsI7QDrK1#KT>z&$lH&_C%4Jx!e?!d4X{r2Oc#7TToy&4zTeWkl0yg<qYLdbVx
zY?XkFUu9K7{)!2syj#2`nm~c;u%wxOt@d3)XTUB)MJ>I$wfbl%FT8(Bh&EMV0*L6B
znCGQ6S%{guz96};Z0v8To$)-A`x>+UwtBerawz|_YFRt~mfouN=J8If-%$`48oTYX
z8KFS#y*XaagV+Rd20$~d{&iH+V~OBqy(a^Z`TfB-q40w?PW7*M2M=|f$~Jv*-EaKM
zwL~aW9<Y|VnX?xe7SjQyosk#));kTrj3o{HLIGAhi#LOl@zZi{etNVnTG*87MnX)2
zTcOC9hjj5n4_#GZU?kc;pB{cV$S5!RZwLGHESU~ukx)XN=GghJTd4D0PhRJ{UPs{v
zsgsMoU=Pw$l}vIU!{KJkhHe$5{^3*p8^(tf-?V&bJX+#?@;bz-_>cedznd)CppvEc
zj#4*!R6?KJvi(X;r9+L@d<fIA$*uUugG#4(J`g^*c5>c5Hh*}HDKeWef4>oUD>PE4
zI;zoBcHQ&zvUki9M8r^Wg-$(P@4_hP2EJ{J)A1}=vq;AGuTMT_1c@D)bFVJ-tM>!@
zDZ9^syhR`SMW~fvt*}7G{1CbRti{E8V~UYZ%{^l0(JX%Y`^yL<Z^~qRqJrQQRCf;@
zhUM399p-3zJtEPP>+EU2chzUFDmwVxU1LQSEwL{!tqPWpuOcS7TXV|1L#P4XO5Y!i
z5(wCxox%5h2KX!Y)RQKCa7my2%x>*Ifk>Q%$(1V0pVqcH3Mgq2(oyNQ0@7}>%1lW!
zng2<QK-Q@N6{8_@#)mGz=){~&Gg6bjA6NwR%AP6=VVQO~A9lMpm~u-vTz8h@oqr4o
z!TWVMLrWx>5=Mb=GXIqS70ru*tDH5L^w80k6iwZ{lj?Z5qw}IY_FtW7$)A%n{Rx)L
ztuYVd5AJ|PGlC4l(xw5$ihssVNb&!C8{M{9tw5L}z<6FPzGVh|&VN=|5^3lT<~+q%
z^O91QGjA5|PXO_9>%bp#44@(-b1BoaDydLv5LL>FAz<NsDudt3JnW0Ld7U?jrjobP
zbad+8myBP+eGmdgDsLz179|ezxH;_g%Lr0;S6VRp2Ym>^^5qpiwYmFHb3{ke<u9n7
z_TlZ|lpC(!=x9Ea8{Eu!#BqOJd#=^_dk;{l_%J!Ot9ofvUR)ajtCv7I6R!gJhsakw
z(|E6Wsi_z12@z8+%d;xHNmCmO<&4X5Gb&Fy*3?&gr(Nnp!BlZuBUXuWVg8#7m?E2?
zKsqYVf!K7`f}zn_<3PvB8ex`@LlXgZlORU%Z7wqcQ5jmI)k%e2yM0Yf+zs;1kU^4x
zf`Kub37gd(*G`XrijM?KVrD$RU+Gq)_MeM{(~$(<`LiCfiUP2QGDH|d+W_+sCKwjH
zs-p2Y`(x~O#0d)eebV63LfqHCX-UPF>7Rc2@|hP~mL|Wg<kh8~rn}C<AO{>fowtG>
z3hK8~c}H);S%)~%F?@Vw{}Y|Nv(0+tPgnE*qI0!wOgu}gth-erqSw8GB;L$73X1?|
zkOK_!(>!z%@gt!pCpLiGqlU;6F|D7`fCO6?M2Z35OfAQ72@jJTRcI;;sZ(8@OJQp5
z<^M2a-&V6xM2!UFwW%sz4fF`F=7*OYD?M^!OXp2I=rX@W#Oj=muZEf@MiNh)|JK8F
zW7x7OmOdAn3aqFudyYqbs`8ifJ08GvZg&Sqfs@L!&t^Qq7tCvNlDgKZvBu|^1@rjB
zLyrk4zow$TCW2c==aG*i8=eoilC0?;WT@*03C!;?5i_Gu&eL)XZnviK0on==G-&df
zrGE|pUyO2pG)7fv8qfUJo({Gby0_&AL}lFmaP$N%oC~D0uCGkVpqB~(4^O^u?qPMV
z5W?@#S9^FNw;1Gt@L4K~Ay$vBcUTTn23RyLgqBy3Zb|p+@dmYb{^CNqtyDYb=!C|U
zIqf9sp#H-3kS$SQ24;Tc;9>ZR*_Sz`e@U>#xz^;pRta1*f`-#VyW=d*AL93GQ?>ej
z<%vR%<n4VafaC1^htP2?DyHb-)BoXi@%G<Wa3re(!sC_~$hXq(eRD0^=acnw#wmHz
z*f{-PN%EI9^F)@F=4S%$<@mq~a{YI)%2PdoiSqC%Zu3^Q=yAUzM8r7fMP?aEpk=FH
zS<%$pc}7$3^i9Znqw1%fZXlyB!rClxH;+M*H~>hWQjXbO%Fifq(T{74roZGEcl~S*
zxY>U?IxNrh-S*hGonNLd5;j^qKStr3O<x&jzNC(dJjlvDCtVBr%*$T_fn?!;Tg<(!
zuw?bPu=;9*diaf{wT3wD%=_HyiqtoHylW1aUamDrNlrYPbY_8{9%BN|oD15a9=4uJ
zOkXyA&IZV9`SJdY2ho>gD<YLqVibyYt{((tVGI^uc_U{-r1RWj?*m0Bl{e^th}}tf
zyf45YUDCoR|J!7uE;Rh4MLb(d0=KI0&zLfLYpAq``Nwxy1MVe}o3NN*IzNcNxFF+>
zE>0_w_S|=4u`PWU<C^lCc&*-?*d++Rk)WR7(Yv;-*YhELfBpGhT4)uM-LWlw&yQWi
zG3{Q&F&ly^>i-c=E3JXaLw|2Lupnzl{u5XC5H%`kU;1C!<o^Z8l^e;+Mz75s6puWh
zR?2bey=y}AiYme4MP>(LB&Ob|2NxV5D23Ln2z0<jqU@U~@=vwP$_9jtTbs1HGdk4o
zRl2ZST7!Qs-`)dm=+6Z6m%1}4B+V##eJKbO5sL?#jx|B6(ER4OJwOJjt63b9NclBf
zO*=F%eCr%wUv1<9`^}2%QSApedS+aN=)z1k@ap<r%sk7%$ITIeh97)^%14D8kOeF`
zmYqz_ewYKRr^VD2dL5v%Oif=>|Hx14jfBKLsy=hqFhR*<@gBW;rl1x4m{z?lk2{Lu
zh5@i-4KTM4pg?;>E+|_v*ij%hqwJ{=f&-TTlX+}+8vn3r?cy~4qe|?zS8t(+%}UJI
zHTF;tLELnCNzrUrw<|-xJtn3$8Cck#1r}~m9hV*$05E4sa^mB4;13m}LFX@CFI+Dy
zIJBxKsc}v_cJ+EMu(ZcgVYS++;>3DUlwctI^ZJ!okO@{Vv&0G>rMiPqaz_<-OzI@C
zFWMur9B6-Fy>9JlH7tfR{~I`Jp&RH)7X6$+IK=&jBI(+A4W%4yEn>)TqcE^h-Qq%d
z@>M;mZ_7C^R)@H2AGQP(<qjhxq!w&;_^=9FnNQ}$Jxq42WhE`r1ju$%Vx44Urc~=D
zWhm9at2iAa`+Fft0Vr_!`Xll#g$q-#?}Y+SO=r;;m<*#><ru}w>kX&!`4u`AFgv>L
z%rvI8Ouxjn@FmAs=9rxedlVPYg-(fQ2Sm_6S9-h^S*?E!bX2Y#I#1)_&L&Y!AN}Yc
z2Gt-dvgB8pbdN-$!pN-<FZ;YMehTp!7cMm>WY64hBI@*Y#?Ln)Q+5yMz}yf&?gcnJ
zVQ7jXG=%*E$rIR=>uJQS3p^1TglE)4DPkN>AZS2-?YMIax6DKg@zW@vOFAEt;K;=N
zu&?fY?tDE1iF?57I*!>PnDT@Iu%?fyHJ7G@@$yq^dPh^mfrDY%6800XhYIO>&@aaW
z8JC&2nKuJ&BqBI=>47>oX^3*54yhO*?o&{;n{*iZpXr`(5pZxWsu*;!TC`Gl>#|7g
zRn>qX=<%I=0IN!LNJ|x=t&EZ+&v6jSKo0bM9B4gH{z(D`W<h2DbBIumZZX*!ED%hh
zd?LT|dk-_RdS)*1d=7V2p?hF!`nsK?`+z+)f%A0T7t}beH;U5&*voWJTag9Sp8hhx
zG%x`I5OOEy@C<_eHb7(;^X3P1fDs#I$qd<`*6Jy+mPbUEhK{cq1ZJPFou`P8n?Qpb
z2Vdc)v!^5DX~o);7!yG|^o^JE+Tp$sS6YtOBE*I>t5t2jb3M4)`+Kd#9!i|5if@VV
zBX^Z=62Ks$+u&_N$Y9jV7mGKF!QJEjX!iq71$*>-H~Iy~Q+vKHh%YoSF{)eZ2ljk6
z9MCvpADI6HRJr%9!obXnu$7<{G%^56XAkiIie7uMv!lR)Sf-LfgAre42ACLsS{^@A
zV;W1o4J|W71Ea2fy>=~ZIcqSt7yXj12lWJH0AyA7I>U^0dyB}{33$X9&x|18=2)@{
zN|qVIDT*N0Cvpufk3O~?y^i5Z{M&I}v`YR{DN%o-qFO=bV*zcNKw(G}b=uwJm{j|d
z1c-&|LD<iiKP~;v{B^B$IUZaxO!lWBXZ_0lm0%IdZv-ZeNPM_1uQYb`^WxXvcFbT*
z2!xW>EA=bMi>QzfJ6^X-VdLc;<-a3HGPIIlM0h`K)pS%+UJ(MTJCOb5elD8Nqe|qu
zroJn`d15Jr!O%jt%U~^SlWE0c?>kdTUj=fh%wJvq2s~T5R9+5JrJ8>BKW?7r#Jmrq
zxFSCy1{jZMMm%86A&xuva)ew~7>=kRhFAG5G~2xhB7wz)pw5&$DR1?J-J+Ow%yoq*
z?oXlJmGa?<%vbpAjM;~LrHQRJDMojJ4W`|=Ao4>b`KZ*Fg&hPT<OXzpax3p4ibsfV
zw-K;C2ExZwX@urVK4T8`vc1?VM}KGi?)fz5z>Yu!UIq%zpg~YI?e==s9p*Q&6>QGL
zPgLha4U>Td&~3WUl(`Tzo!|rYc5h>(F^PhPrRp^49Qg1?H$Pg9#B)vyGHsm@gjmOn
zF$7M~nZ3Gk;o^lF3Uic)?uhuES8{^?4t^!4$;6G~dEBE^wI6z}TR1)3@^%9K_-?4$
zj}BL>88^wllpCJF*V0!1jZkj*1~e*|n7`HX7%crX5S671&8;DU&=u+xs*5R4*)J*2
z-<PILbp(NU36T1U>z_oUm#H-cf&b!~#9o0=SHtGo0=;e^Dnj(2E?Q7Z(jO8gvFoCX
z%fAX;_EIbr(jC8<k=B8B{sE`vptW7Gqe0M+^4GY5d&oG#s`b}55h{H#Xez)_s*b9D
zy`Uf{L~u(yT@h)>CQS?RQFeSp&$`P|ISrN|J!ZM*9J=R6eg&K3^FKh*6A66b)^9xp
zKma3Ov|rPf(hU9{AIo=vcvSQ?h1+xOQxzZ1s$-H4!hp}r^-Zk@s5zfTU`k;S<^Wuz
zZC_)z<1`(35q=ePtTZRsp6hicxanFmn{#>t@6q;mSiBY^_#<EeX44&XEA-#b1(-QC
zv3XG=mAeE*4JcI}zBZMGeYRltIF)aba$DEJkU}JUZ7>X;!t-CJ!uaP@ZWfonI#s3P
zSfKz-b2i-JY?p!fzj~B1sTz?!o*u@|lwW3$`us7{8iXsJaRoOjh0OhyLN_1(pF^A*
zO>pLN-UY2(dwE*(u&HUhtE~K}ufE6T8gQ1DoHAsQobuBp=bY06^AFZ_rbrv*v7xb_
z*r!{kR?kp$4N#{hOlfob?`tK~q|QlqUrKh3Ktd~D2~SrP)9&*Y65UP40>LE`ncd!}
zOU^=`rZrccnSr8nsO%NY8ymd^3r)|C0Ah~B+^v996WzJ9F;ubY^q*hNtG)O@*R0Vn
zpFDsCZ`a3z`p-{v5*@#cb$DC`07+^S^Se5&^Xf&t@yc%(6UV>1ZQo;T02`&to-yri
zyuhXu7&*@=$Wg-{onD@s!E2&JPpV^70!`K7H)^SLDp_E!4_eH;N$R)J+G?CfVEbUX
zueX=hfrUbx<EFi)-eL1&ll6gTv&5B;tLJBhr`g&Q0f<CP$ZcmySbYRp5ec{jtRh(S
zw{@pGKG$$`9>l00*JLZ|s(QwEnJ;U%?xu6;$;*<^uZKS)4kAXS|EmeH>WvTWd;unc
zYCvv#smH?GM!Cw{P7BedBxm4`kMdD9>uv8&1cx4*M%7KxdX{Z%bHOeKorhcg)mrHb
z4wpPmB?sq4C6rwGt4992c=*p*4iaX`_OSwR`|dke{*b%fKYWF7n%`tN`gpdxzCyI-
z3hPeA<YqFGmkf<D)2Jz*=C?i!Wd#03bEcc7?;YqQS*E`O;tb<HB^#bUFpB}Ga0G}X
z?&#+}XtuU&=#Paq7M4tpCI0C<J&f-p=CzqR=l8$DJyJX^B1k9D9<SzS1LL3FW&_z<
z;~uJj>!Rs*Cp+s_%?C(+9=n?T*ASesTv9RomDBYqPM$#?z^f5V%!%?l*it>qn63EW
zw9@q@ZZ!H1Dq0uxkbaLrWKRAHs*L#nVAW5oj-i+Kupe4<D!dg7>OfDb@hzAUM%*uB
zL)@GV;#r)Ia1r?!jg5hxKTk*X{NTwOmEOk6+;}Q%3&p0mkfE1etJo^DsE+XJ0GY0@
z6-%p3POg0_+d+;;Lzzp{-JxW{#-!3fTG!k5rl2>E436F$J*0?p+-C_#=&}60h+m`&
z$qnVTGGOz9(Q{M*UOYnYsL~Rsp8_ovfHx_DH6_5tJf}uI%TjW`YDMC5z5e>~lUQz(
zNm3==OqmQ;)&HVmf2jJN!;G7X(K#1da!m&b(|<jH$wm?D{Eoq`eBX4|T*sDsB5JrN
z;yUbYFYMH^oT~3x=@M*Wp9Y`MI*!!!b}xT93XxsnAa&~HFDXO_uJdQ_rFgwB({9OQ
zfo75Ek(J!j{BhaQMYQnPc_1ynt4g1t#unUMb9YS2MkSyb>3dVokl-^m`oTGXsc?Ey
zLFe3cf98BmYM6iih6J>{2RRws+u4LwJ&mpNy}mxSXCYs_HVf|t%tMfm1ENo}r|i_>
zUVP^=dZW5!uk#6*CFR8EUS7S+2-J;p07}Wvi4T6fo)V3^9Qb&y;mVU~yj~fr<fgM7
z17dT^krG^9X>h*B%;u+O>Xtq40sD3Z`LAO(QK<1cak@?Kn%KzLa;J%*FuM`uB0Xzy
zkZlFs6bAnK;#=ZE@vFJ-{2^0%+XHmr5P-MPs!mY8_3RMnrNW9~kDb4ezw<C}4|Poh
z;F*824L1t@9AZ!^`8omtS_H29Y)@W(`Kx&4vi;i420#x6t;*C3GCw$%|KC)UIdFci
ze>{2`5sS$c!**`3M#>dOD@|<NQ*I6cfarL)mHH0J84deHiBki4RX{mWFhSYl)OZYQ
zPr!3t+@v7bK)tdbMvx#*jDwGcoIZ1XET&gi;D=E+$LM!M2q#weEF)^@6T~S9;`svS
zUdGiS0K0;_T_a<ZE_HmTcjLphF`zkB!GWlzd+{5KjN<G0#WxG&JYw7=e}KQy``H`I
z@79QAgfXqmJI2W4L4p&;t)rRie0wJ!g5;vfaDI?@V6fRM!j|MK0K$jm7W&qT#JovG
zmGtEN1m4aSfQWG+nD-s%5Su+oz|lpZN2ve83_hOguJS_^`_pvFt_636P>--BOF`dX
zKn+Rtix@@WCakOvWu@TbR)%`IIy*~4IC?J%8?e#}nRCvJ(A5`u=xIj;(9V93ojAA6
zoS+FTI~urfg6+T!*0YQ~_UE9?-!vlN<oi6x?7D8+#rxFs5xlpb_>KLN2%4m9aB~Dk
zq4MDn2JxnJ4dPa4hx7uQYti3ia3-N?5McB_i{oFyy#y$Z?^);i1qUmLNS}Y52UVu}
z0xR;I2D*Qzp#U57c?SS^Yz+gQ^$0Wp1ZkL>Y=%i?6mtCN88F6&;P}`NBxRQQ@-FI@
zCsC)){p`hpLgMDFNn#z8x$2`DqplP|%75&;q_C*QBKhd_kJVccW9i-Q2vU{BIxAgb
z+0BtVG$!sQ%IUYYl{%J-r9V(z@goOMtRNL=RuVfVoIRZe1$VU1jCLIG)+XFA{q#Nd
za%)e;NtcS3mK!cYF@hDL>PP%0XK1#yIS~W3^ev1nOy5p{{v*(V65$uxGk;kW(1Z_8
zc`LHroLos4>=Be<LonL{Nh6}%RUNGHjGOUu4qOPYy^j<)r+C)eT~!iS*^6$QT!?1x
zMd>D<u8OGX5X2|8Gbea}#laHew(fQJ*kHpquARZGwnIxDHw{)Hvxd0HX|}Got_7~J
zWwS)thJQQ#+B;;~{>6>kK&A!)JSbdf`NQdKd3KyZhCyEaI7p7X?dW*7{_eN+ytlt}
zBFcM19@XR$?G0VcivNZUbw2X{$4n*j`U~l6V5Rp5Cn59g%T2R792D9c@<-qN#PHkQ
znpmg96CIgKceDq5_r=#eX%}W(rmanEw05lH%M&N^AEjnjAxSyFwWW?KYws8?8)(gF
zl3%R*Qn;ed<Uvk!`W)B2=_2}NYP$y%$m|>OfCf5d0sL~_VI6J>9XB0NhzRQ3cX95d
z@(^QHwb6nvu@gRQe26FZm?S>&0SFmlz~(;lXnT?t4VfK!;L|fb$)Bu6^TeP|aWt$#
z5&0vOIaM3b8-j8&97msprPCE6`sHmf>MW%a?O)go0b0c$Y}kPvKrw*b9ka9O)GJ}Y
zso%fojr4LjGzacVEi<(1AYf|CiWBsd<2TCE$xd{470rVps^s5Wna32ruX?U$x++cs
zYv*_KKF)AL2;wVT1;S)7(d8M<(cT86j)>r6{`w8ky(7vKlw;3l1f8C{97}?jLCsjv
zbe3(azTq|4EUSZ~uX}G~jZ(z5*!PV*u%YkK@dl4Wk^&<zMwMPBd5L%O8~HMRpE8-t
zr%|U>JK$wYIuZErnA)V(b{+#K=gh+^=RaG1u<Sy&p7~j$z|O8zYWuP{3~_@yhWr%*
zDNCuj!Ck386n=)iL|P4!$Xx={xl#0L|3pg|jlA-s#UY2&AvF7u=LxQ6&X>N*O7l2}
zW}ut(Jt{d9u~?aRes0Wl7Z-131~>GVjYO4ykB~U+Ftq5Ef!5qxYf_X?-{{5a!X$lJ
zQ(NLswRrDOV6m-Wh}8RfA*u#du>#_190d8(CD&WO6&Mnn67$P0R$n@P?MOdOYL0s@
zsz$ZaqCi&6EvjZaabuqZ?`KWz*@Pw9Z$5eXUEIgYOLkB4T~&?WVc@fLQArl7xHm4+
z&L`IlpoH`JcV7|k>XmlpBQoa<)+=kVPG+@nzUFf`7YnS#D~|QgiTYHB86mD_XOiyA
z0p$|QXXq!3TqaLTIHqX`I^!_(VYz1wXE#hl%~m(r0lE{$91hRXNvR7Q3J@$2ZdHU|
z73#y|Ut+O07`2{Zu_`HBFCu<V-L1>XUy^~L4305JED1CKJ^ytBTw?nOy#9iX0#7?b
zuRuFfyDU}!J^s-LmZ)(&SY*rF%WSwOMk7)^QXNiWRJZ9HJ<Zrj{?a~XZS?hYeChV5
zt#Hq20^T5?dYh+Jlq5ebBYyNM_<;GG3L%_oI%#raWkyE!h^CctUH-ITf6|o^{}8U4
zrO#YxlLaXXBL)OxTj!5s+U1by&}+1H#=g~Q*VOtZ+FWneN_bFfNx<Z9c%V>sb-I~)
zp#%xK?x*<^#=Wwu;Gmu(Nxq9RutyjBlAcp4QL|K(=AWW8dhpJ>i~*;@R@>ZvO`LT5
zY7SiBwWws0VMDuXR88J2M-);d_L=X{oDFGk9^Kc!X6GOySqe}5O@V6}h7zuAxOh&u
zBq~u8-kM$gay8Ab$;VeOFcDq$ezXhkF3s6vIc_Z@W)NCR>6_ITP;sLf$J-x!b@27^
zH!mbo=Tj;d_iJr&*xl;>QuUiil1uhubXl!!HelqtB9Zs3CTz3NOTHFa8-ezsFyhX{
zWdy=E6GtpfLs6X7JuVBCy)b|P!WT;Kdr>V^e$aYW5B>&fmrui<Mjj9kh}VPd*&lUh
zy>yqNGZ;ydiL~lVgbBgv35u>be5Us;R@peEg#pO;(P7?UwACfmzo(@6=ak&(^8_E<
z!BNVC6k@og1@m#=qC$QfYi3XJ;O9*`haWzOULCecm3o5nxP;|iEZipzTv4qvlhI9y
zuD;^&Xx>StHtbpbqrSe|o@2+D4Uc**Xe|%4)SQN62KMc&g(Z8Af>S4HP2=+ULBg?O
zO}LTc1>jQ_MT%zEG~cJSd7Q6F_kimdj>q)E_1uzN#vhdHsrhJ|h=&sRL^|BYvG$i}
z^h*mm9A+%DyAq%G0aI4TpLc6`HFg_qEvUb7wZ-fLxAnX&duVN~Q4bmh8|}IteyV5a
z;r$Du6^5+~6ETZmQGiLc&D&~d*RayU$eot)Eu-t-*?-GzRc+fmY0X|@e4{xP^~u>!
z2Zic<qp+Z=Y1ZZFV&Q1zl`tsdN%O{<LmZ`g!GRrFV`gWM3bFSRf$h~*+-~B8!x#3_
z`mk;e5-_4?`YD*Z`xW=_V)TR~eS3w{%ey?SgZDh}CLNj6)$FNT%oC^@(6-OSFw2WO
zN`_EWv8_^f#o%M`Ty%ckV=0>3cd+Wpi3Y>~PLknFu?V72gZ-ZOZGj)UXLd)&vfKZ?
zSd0=<d@(*8;YaH_t7cP2Du-7Lk}u*|ii~lCm$D^-2~$yJ=Pez;d7XTM^CF*DCmyfj
z6kO|Vf^c91H>k?$noi$%C&Bftss}Z#Tsq3}%glTTKRi(Xa~d5^$O5t4h2P#Dqwm%;
z5A!nC!?baIyJs-P<$!oqnSUxXzs%}wLR?;H6!PBUd=&CRK|&>JiJ@AZQ*sY+q0G){
z<oq{r8s*FYJ9BH9L4992mqp1`-QkWmn;RE<H#Yb*D(TDC!Zrg<<84cqgNySgT-;T<
zit)<l5F@10flz!#m!R9_4;bk^>Gdwxf_tXzg4vF(emk~mbU0VS($kELXDAAJmU$#{
z!W!k3(+8dT8PGTlmw-J&cjNJ`rm7<+ku_&Vhn7{qg%b1J^F!8f)-ycoQiks=K1FG0
z)itzP!{Kc<+N0|udr3r#vdG-)yLK^`i7bWpjvT1+j~vsyq1OW7J0=}$V80p}B_N^^
z#b*#kvGWkm1(6qSrCX&8(Bqlb=T{SN$!8SIdiKaS`VEyC)7UWPvM=&hl&lf$J9u*b
z3*W=nclP}U{yj5Arz73L)A6>euUC$w4NGB1jP3G`qw*B{pb`qU3)PISQ0%MTSud@a
z8oczlpTSki(d_3b>81-Ooy(F)Pc)o&#vJ~u0{%I_iO#<j&;o7udC_Y^Cc2(%UVcAm
zDlHIWm%i5YK`kIMCu8CsFP17YrggWvve1qzHBTdMr+C9E^)-}my55!^SD-cMGL=<W
zM&DnspY*$0x*qb~)O_H=Hz$X$A;?V3ijwsEo4y%R+ORomC^<vons$n5NOKU<HSg83
z)K1SY{g;zxR|R-KOS78ITf4omxjf=)^YV8H)0?g5+DFP;QkQ~GB^B?uh3i19oh(NJ
zV$e&2?j#Snx$fdZk`t>yCbmc=QZ-4#1|0qe0)E>exvVk`HFw`={6fKUm9h`f3BlEe
zy*l@@7wbI7{L2K9h&hiK@RjntR)N;!+&d1(Xun3bxnr7N1Hm@iy_Lh|yOFVr*UoHn
zX}~em-Vy<#-ogP;46KT(YGYVI6%sDG^RUb8L^EVxw2J|oAf~p8Bv}LRRjO-(^k^yd
zC%(6iFVBqSwxl!mL@pa(&V|e!9j3QsR{h&;+%~aLWt_X|Qb1yFcHuuyaO_~{ZF?^)
zaU^o)$bM_79uk&dY01oyA@to{F1@<2Qt~oq+E?-`X0$`dsrK;5v({bE;(pTQ=O$NY
zv3N6cuO4(A56bfm9;2QFO(q22^xmt6i8^d;+h^wWExrk>-{~$m*>v>QS{~+)tmcCP
zI_yh4OwG|eYzgZ6Jl*viwQI}SpRr3<*iv6^Hj9=@m`V@!Kwa%UWF>}sj6&xgGYZ@n
z$|Z{4)xP;s@wW^$fwd_a_4NBv>9M3<tjeW;fxz;2Ud?&6Ke1~KD)eexIi1Z0>&5gv
zS*Ip{%zAzFI{Ur6_9K)aMWHdEmtdQgv_TB=+TB)JvwOnU=5I9OY>jtOkx(`YXkh72
zN1?Qp2>5FgyKf}DB#KHrYpREj4K1diP;k>;?3!IE(p5QJHm|zX7}VKT30O9|j7a*9
zL{1^5(;OpCtU1-R-;Y9O<FF3qF_q{LNVOP-x`DR(c_+;@SCqW>HhsI<*{z4n@(<Ec
zdR_umcL%&=%EwWfPt~b3**#iPmwp&abh6>YkBCRnkkzC8@B_11_WRpyM}4<Y%Ktod
z@s%{FRHn`bshpzyi04_P!RZTgj_>+<+9J79!jtqY4@3Rso1HH0q}#{e3b+)gcWOZf
zm-61FKI1$vq6_o6b&7BNHAgZ%O}9n0K-QD;&%0N@3(CpAsJ$VhE_@)wwb7ulg`4&4
z7Kl5eFs<m<sic#>SQC~GPuG^u5|(Is9z;3^<I_1at<d|YJ;5CYB|j5d7(CZTs~RBO
zh{Hw~z$GM0t!~JlNL_#2?Mv-iyw9>RN{4&MAM!F^Q`$df@58E)9((F~V(FJ>7?+w`
zPE|=y236erOpaI9eta;Km9KfS_)Zk*Boe7J<z>I_^Esrj-@PmdY5Kif;B^U&85VnO
z`DF<IhF*CiX<zfsY~X&_XJiGTbiO#=7M19ZM5?!)M{E-3kVyV5UN|O6=^P@_R~Zif
zaI0*^0*?<${O8b+EO7X=?wQCJbQJ1AaLmb};U`!a-2fD^no?7l%eb`A!0v6_BO5E*
z&Lp5F?C@aUs8?Rt;l~6+$bEu9tZdGBCq^rl43@k}`tU4~Ie+hN-Vql1&k-?p#-1iO
zjoF{#a7n99Yf2X{?bI7yAWU`i;nvpbSI_zu6@hC<9ir0^aYFx!rQef_-B*MNT>jmp
z>XvKjjqBG>9Up%QACoT6Bm9~nDPH5Ni=z6!3S_^Hf2f$=)zu+^y^FofEwTJA_}A8}
zZyJJX_@pJ>qmVCf$*xYm)C7ghx&Up_&)uVL=-)d<LmKDI6A3fwD~e{J6>F3FacRa~
zQBMP9*4^$bX4Y0(!}aY?{n|SJi#@JAZ7<Yy_Uwe*R#0+W-&WFjhGN3-U?r=cjAEID
zpc*d7W#s^r)EAxwJ0qn2;$`}7{M=fz``N~4J^20e(=rlfUHojHh`rcDO?Fcycy;ng
zXDPK5PG9&{J~Nc9@7g7fBp4o_>>mP}VA$Z3HMWi5U^sl><VIX23Wf6HfWyBpR_Y^`
zDKpV_b8|Rst=5Ne_iFhxl+n9Btnh^(1)iv9!&WAo{31wk;OJB^O@}n=0pT2tCB4FR
z*KJOIy&;BY=;d+)FQ0y|EPuWDmN4K&3nvb&(GB$t7K$w#juXT0&OZEG93?XfJ-EBG
z$q(GOb{p-mcf!&~ln$1MOfipf)n)sSeryE)D5xQs_6CE(2s7sV+urXGN~=RaVT=GR
zKGol|b6uWMd~=PxE=mxE0-GHywq9tCrvF5$P)>Tnqfv9M9VyGpeo@4c`aWs`?e1e7
zOKNKr@@k;khXM(XwnzDP7cpPg8KADU(#xQow`L8pFwZ?6S!>q~B@`<9TB(~7pIYnY
zKJyM`z7HQg60}OUT=#Jlle8&4LeSw}-m|^fdacR06vpED8-uVitEezSe$FdZkd9jY
zy#FpV1-s{`eDmquP%n|qRP+QfnI9$Gcm+*gtdgnHnwx^E<p+I;_lgBP&;xDThYcJz
zCg4Sv`9NK>mR)gQdGYq>0PkWtrUDO)RX9v*0{^!Z6#Y;i*87B;hTcv;O>05*w4ppq
zNO(+RbtvwAa1B?g+F36nkB1ZOKBM$FTlIR5M@(4HPmN5G;N9w_I#@hE3tcs=qW*X|
z+IHB4)A;ze<zQ^LIiteZ*#IZ^i48iTN28pR{;;pkzxL!KxsrtxVn6749r8dP;fNe7
zohhvaCRZ(QlCKl20L^3Ez3D~`#^mE3>Ezzt>lr4zum6tee|q~Kw$wIN_)g8U`o+NP
z$SCxL?hq6+#x%R;{G`x=^ah2nT$Yx6q{@hVleDI*r*zp|5EXli0r&Way8K8(ei?m7
zoU*2fWUxY=s#i$)DAny6%!xFl%-J7ydRGgiimV=6E!3yp{K792dT+*-t)r)qfNO9T
z_1f3?$`6OfUy6MrHtWNhYB2sF<W~R2aVhn?uj|TjJ(u!I+R?|}&57QJx1DcZ<n&XN
z&?lAkVi$bu5X{7^0KW6*tyry{@CK&V0*mZK+qZhTccs2NNW-?OQ|}C;$PMGbB|hg0
zbpK<?kT|qmA8a`s_|a6_bDvp!fluEsLWa=7ulAYpo$CX$b1e?Lv00yt+LI<y9`|Nd
zKYR8)9@vLa8U*MCnoUnX(1_BY1O#=;V7lF(SZE!k*N;m7*>AG^e%=$26KaVX#Fz6_
z3kc!Focs_PFx0(rb_bD(e<o)}XI}C{6=2LCAuRnugN6s*EZ8^-2zk~HW|MCo6&-1i
zTMJ`O)TGmKg^?#9kg^jTmm%^r=aXMUb{;QB21%RVfjs3*wjC7Y3$(n*gm+zvA!_0A
zi@}a+u@F{DruX6MC-Ad|=3#fD<2C7Up&W#wD3Wtge!r!;7oGBO>d~FU95Cb#lf1v(
zc3O?i=VxdkG;qZEpEZBhu;j&gCS9`!N`Sv}7qK*0Xwf$x_+&I96EtQyBGq--R{%b`
z*)Sn<K~ONEI7tl4Yqr5K7a!L>qZPNH$`PrCMx7$sKMn27iZ3UoqS2MuiJg;!^WB!C
zjoS56aGs&UFD4mF&>m513E1M?ON(*|bm9h<`W+m+YlGk(t*I}Bu=&d&LyW;$Xo+vt
z_F*SYc4>r+5ETT43E&z(`0^RJR-;%f#4wWT<M9+6ww<CqvORKoBvOObw!MwNimKrp
zYf1-KipJXsI?Rf+)XGc)7b#*yfZrPu2*SgBHlq}6A-4=hqwXF3TGz?U_xJ~F;#T_e
zIH|oTeRXQJKr!OHMdDy|S(y>eO!6e*R`br5oAK`B8kfH)Q!Mi4DS`3Qhho@{E6V@R
z)K#dCDH@$@u%Wq2rXnv2w#lp`F>-VMnJc}*Ob}>XL$H=UI1mZRqf<{&p{|50=3%h_
z->@`~IZ+^+O#k$uZ}zdY_o+^R2@*^^hP$(Y-`|NH%4SiuF_$Y57-UWTatFiC;cm62
z(dWZ$FvxZhZK*ooJ3^eey(5;BCw;O<bSw2iMf2TmX;zy(X~iY<Pi%VE;MSd`qW1$y
z;?J$P5s6ki>OWu5^{3-M)+gXcW?R(NZa%I2F4wQ4P3a)nPjD&sG(|9bI*7q~eB3dn
z*o^mM;K|A3YpE_PGYwj)0Br=3su#QAGi216wq@t(#Zf(qcR``1E+ZsZ5A&2G9S{<o
zTJkWBv;2H$3bds9C2w7ulbP>YqL=8rC_Lyq_^U)3^yH<SsQ}bSofsTptG=B5j5gW8
zZ88#gau>6wn%&d<WMk%KbY{2~isl<X_0Q$$JbnC6mtIaUB#@SO$cQ9m6oQ_?+y@@Q
zyRTnBN@>kKb@gxOu$%s%y9VG&6p!O?oMIb)i5a_Cp6$9k_4oCWLdgBRueO}lF-{gF
zH}<UHn(Zq3wkPLpd3yV>Imu%H4FsXKVQ2>2p#27(SXfZQl_r!k9d~KNg*c}T)#K51
zI0t{~BySi0`Ti*xVPeWn5s|*$H_I5gmS<~oK2BLCy`R6)&#Fj{!UTNn%)K5Ta?GfC
zU0%q3E5w!xIx>;7qvCTHTQPXM3O(2Uurgebqm<-eNiIK8i!*p+yD@{0_}S4}vqH|}
z)>&Vx>$&L$$%W1x<&Ce|hkdR_@_xO)J8xWg?HOsI5X~5Y8A5M;SSHR<ruSmcb!ACr
z6SZs+i7ECeXnx|Eld#u&ne&1vO%Ks_*-cr&cGQ?3r3umJ5rW-@g6Y0Wa7<Q@ECHW&
zs*JwXsyg5m+D@o6WGaD>(i46eezl<Ku%~g*Qa5(nbklgMK#yVjCl->ZXYPkjj1rL4
zpsSx<oo(kjrV4R@0@OpS`WKXtll(`$<1?W_M9c<6aitClpCjOv;-1|bu(>2*q;5bw
z+|xM{B10X28jNlemnAOFPB20bSzX?oVhd#3p>H3A>{GT^LO?kAU5b4$MpNu(|JX2q
zghN)FDn7_Jh3Rw(%TL9g484$9Q`FCy7GI6{El;07aN2uI=|HsQUP=6a>gB>wLs8Ep
z#)h+AXzR>g08i5ln8fa1ZR<nJL_c78@W27V3EZ@EDj)0{PFHWFkbEU1b!hc48WK@8
z4AsyC<=c%(7$Lb6sXZW<{wzR1WZVaD(cy+u;dWw@uZN4|D(8*xDL(83%U*2Cz%^9%
z-nxg^TB{W`p&;qRIz->)+6D6wJpS60t+m;*vvd@x2(77D8F0+E6qa<X|67>&yRD3n
z_*326{8k@Z6x*cW7*#cUXl2)VsSE@bdv5#&^F5A8cxkSt#D=x&G#ecLHct$gQWku>
zSr$YJ?!|toZAe4g!PVgKgj>j(+&x*mi}WqG#I=_3K5T6$Zagv3o<UIVAZcuM;5BO5
zKzLeMIQCg9YfAJq<Fv^%w1cf0&A0^lYSHj=jwXtWQXcB48z$mH<xOnY>d0XYemkL(
zNb45pJ-sutlsHUc%s)BFbKbb-{y?n*5K)}5;whxZ{d1f#K0M^3UGutEgJez1quA$n
zICFS1VygW~Z}@p>X<DP5t;u!A`;~c}PjE?-i;q_(p8e)<#TL}qS}Jz$O162cJWKfX
z=~z({XZLKCeO7e&@c$aewi;!&aMPY-G34Z&zjK)2sO!Gmx&V)OZ1q|2>0Iq;4k9_j
zgmnDVQiCsHeo0GdER1V@tlvt@pk{(HX(*&U*3GtU2uuWO)hT^;i8-H(fdLoIpXsxS
z%dCY4+xLFZIFT01;L~`*w3Os6a{pC4J#Ob);>ETLQ)}9~je={t307vyL<N!S#941`
zzp)kMz{OlE_!S;1*M~&#!PvyB0JTBa9eecPrt{e#LF75kbB=r|sKC$oMHiv@f%4?y
z4Zn8bX-sd%sRvUf$xqBVl`%pCzHs=j(!NfWoi9XnS(kyQs6DuYB#bKJMPIJH9IQ0$
z$%{ZcX_(D=gDxVGXuYD*s-u$Wmxx9aosczHtVqutKL#>>CaDpX?11n~MVRZuV8IFi
zDxvWf4WV|WU8Y!qcyvC&({w!piC*%MwulfPAkRxs_Yw#pxTb*Gs~0QAZ#LY>PQ5&X
zAAs7jR@|?k27v;HV2Udy4g4Lo8J(H`obK}e!d{zpyNzC0Oyy&KY3WCIaN+I3(zI^i
zbQeb_Wnw|yEHk)wv@x`Eim+&ubHTN~$R&Q`29m@}p64WBm304OBw(lu`a913XF=14
zof(v%!!bHlr#}%oBeL@`kbb9F1N2<2DWuC!z_!Lt5B#oURfIj20q*}Q-|kM*C<87o
zBBVSYZJg17FdumDXk{+=)M0N;#(91ViK$JtRC)fJvKT%gKYxkmu1UL~fKuil;Y-4&
z037$#iR$f6yAxkk`@-!q5G$pg$plT~pn=&Jep<N(S9jENY=>_-d41F2rgOi-$G&>p
zK@Rq4Pgi>FRnK!D5oKcN`W^@Tsz6(Ede6iDD-bZ>L|Uk}S|Sn$Q7Zl2aq<@Q+Z)Il
z)uU@i&8A_M{iRE{kjT(@(W3xy6iTLCyAI750MDefFnp|jjGw|F|6%^i^pwG52NP|S
zdR1wDJi?sQ$+(dzB7qfnOC`%g9Ce!me7?3oYQ!+Fr>EOrsONUgJgC!RyDc=J8GR3$
z<2&%w8*F_34059X4ssI_J;AW++})!+Q%CzOo;Frjhd7285?NMt(gh1_0uq-qLPB^Z
zlN^<<{I`|Bxud9|e^vq>c_Wd(hRJQ{2gR+snODo{ZS|5)z_>OV`L_C}H$gT<^YWk7
z+M*)Bx2$P!wG+ZH<7AY*xMY)wxmVvsoEdX~hig$`yc|tdh4dRa$)4GpvQtuefw8aY
z8_va_)qQp<UDR(^XTJ7lmvdB&L_zn~oX?Yf6_LwNd=v8djo)A<cza~JhM$|3m-w3M
z%XXXZN(9<GSox^=VyE!Si#{l?O54-gh!F%=^!lypB@Kzy#JItYJMUcwzHB+0vW|SG
zlB8a_svtAy4%|<?D^?!WY3r4g<2o_{XA<1M1O+)5ivnnSSlGY|%^h5Up3phM;ql@}
zJV>OciKpDTqj4~EjDdyUrUa}F0fdvXiOxt;2s{%ina`d=Nh>c||Jw;^JZ;iJFACen
zVQHfbC=FsOqb6GFX18bC%OT^!;tUNB=NJSWp<~L4d=QN@^(R4NdSdqcjXiH9%+D&j
zCBuaBp12{=1o3x|6VK;NUW9HHPzS-`%}z$}XKBt);mKP}2R5K?;tCYWb@L1xq5q2F
z|Nq{m;h<sk;7(io<ZAyQp|Nbrbs$ZTG<66yEP<)z<nNEs<C46JlF@Xy%(5i{-q8GJ
zQSL-s;!<}>?9*EG<>yLV)*Nefr2T7p@N}Oikr4au{n@H+CvTaWC(5Wjy5j6J94NW<
zcEZ`sIy?B~T~l-GJrit|Sm+{0#_RPE20cvK4~<JXftL5I=dWSXpJ3LucS087Kju;u
zkwIF|dQ#D4gsT@nrDZ7f2%`K^izYQod#*LNP07%e#a+`Y0w~lMJ0&<gi9P9g+&fHu
zvm_i7eA)mZ=VT`gbFVf<R3u%(3njTAkuSd>2`&)tWmTji1PafLcwuO)#{izW^BtM(
zPeoyL%;GFi<`ew7AXG3yt5-Hcpd%q2RqYPoEPrn=T7Z4IVMH)u7}@~VS@Eu?bXKU3
zHE&Q9>Jf!vM~jX}Iauu9r*4g=cT8Rt{BReW*LvhmHBH~H6ahjas3ROHwwWTPD7Qnm
zcF%3?I^SFIx6ht}Y`bm+>2C-*{sAEWmc&bYC8OZzY#0o{%I!87%deNd{qHL`ALOOV
zCZ9^i?>8snibl1<Y~~3_`|cIRDTQ=ZcVTX>RKQVaiC?-vAgWfBpy_c?L6O_)vXXoJ
z;$~Gaj%een);G0ByNl>?(bu!zTX0&tWW+msQ1dK%{4%g#AA9nS4a$GcZ>>cFRq$xa
z=Azb_t8X;DE@LmiW3HDZ^Cnz7He`=}98pe>Yb@ilNg6mFyL&sOR*pj^w28lGmWcFx
z(Un<}t0{e&h(!7;NGp#%Sep&fOsXL%q%4YeKM1h5C7wNz1aK2{3hK)y2OM+vMNw>E
zyhV^clBtW*SJ2ZgKKJ--0EJ<>hkO%#O^9c@8ubCI;SsIzGgRNi3%J$1J>Hf4#9gUJ
zqni!%ET$dXGeu=1-ms%Zz+kZiC5UyjP;IixOMJ3HXwzXRyd%<4^|MtJ)oe(afnQxd
z=bYs66>gaO1HUMi=mmB8_0aUWhIQfsISe`-Mkz-dN9%2wr~e{d+;?NG$a?^<{e!`y
z>BHX#j<yiky{W+bvP6D|6oGTKgw>#T=7E?)&(H7v#VFI{O%pCS2iH<|ZMq4J$$jXs
zb~`xb%aWe?C98_#x8eVBe5zA8-7IFlt{bKR@yFwP5V~QA#P&Drnsx1Os04GSvG*;M
zx{hFVqI#Hx6Df9Qe)D|MIbK4KyXN|}+4obt+<7Yx*>S1B7>P`iGVD8_;gZ-))nXj5
zp)9pB7<f5eU@CfSbhJ!Dk|_E!;JxdG-Euk{79%QII_q8iz0oe4HbYD@35#{6{0*9w
zTUOiy-|I1G&5_rQLhm>BH6@6umzzgMc`i&z1*%m>*r8D8@~%-khnuWWIsCBNbD4w9
z%Fo@NbNZdC%g68;3?1JB0M=EzcVo%h*91|qne@Qg3gY+_UACnH$2>Ct7xVqrmZLl@
zX}$pch18Ia*3>qWJ(EK3-SAtdyfA#e;27=}sFpCg#1s7XFi{Oacc$%NZU++v@K*H<
zk?Z1Y*GG@P1L(2<zaPuMuY~|_JcWa~c9eRG=2K0o+d00nKW^<D?Ij)Eb^@jH&*f}p
z|FcWIdu@9#H@c{d9y{D3n9qQl;dn3)6AChNy!7gn^c(ULj{Sh)k~8N2q#^vpn3bZ(
z$v$)BNpauKA#bDUgZ~6pc2H4or-On8Oo+mgI`=67E8!*nHRtwKQP_*p+*1x7G^@HN
zI8rI>c`N8~-@lkH_05>Dajj#C&PJO$BPy7!4`B6nu^PMWI^HmO<>735Q9gf;s-cWb
z&+u(272~(eYo5iM^l7>4=FcB{+uwXv@|m|guOxROiDUD%s3<Y?UU9%R(TewGbBuu_
zrM#o>C$K-UZwiSP`iSn*vYvo6VoeZfZ0{|vuglW4jR}PrMJ`|8%zEi^(ezC+in;Gy
znTmQ8>5f!9O;_)W#~<#<5by<ee4fe&E!0W4v{{2KLD6PgHQJ8vO~(2$0zSyg5rsz2
zvwzD%Ys$gcAua*#mm1rA6g^Hyawb2Ede+X?8ia(M?2FVh+>9b|wsRKh?2Zmhbvi`I
z^7}(t-kR_~7^9zjnrM*nNEjBa4sT?SU^g<T0U122IdT8iR3F+uLw*JM3bt2dDMG9e
z%uYzEBPAI+oY_YiR4Vn43j{)ShP*i!Dz@6;xBR((#=cbfrG%4C8p>R@#T##$wd@@9
z-8-$lb$4~-e){+?r@aT0{_*=x?Eo&j6(%?8dLZXp!DzAf%Pit+6zQ_eAvZnF%rjtK
zMove)gPl6*Gqr0+h+d$#9Q#If6W3(s|Hs&S$5X-n|KkTo=52<Qk$G%U#>t9E*0J|N
z84(FbR<?#c&#_Oalrp1)vWgt%5T#HO;vA=tb?o`Obl=_Y_vgL7zrPwDk34dn>w3Lj
z&-G%<LkCFdTq()4`Xt*$816q-eRluck78Q9^qVgG^u&jEovH}_N3{s@>73+kMs2vu
z(7DIgut*-AN;wYGJ2-`cnJvsZ;AS_*_~7=bTCm`pGC7w1j8M>Mgeb^q47(F|PQP_C
z3|)&{RiLadW3cKev2o2wb>a(a!{)cX0f=ne<~=(lsoT=RSL)UDdkm86+jC_tlQ=!)
z+y?}_!BtMwvcF3%x~E=Z`|`ScCYtK-0+sdANPRf3fhijOfh=f*DEhKz1HPSrqL}XM
zd#Bk-4<^4?`EeR}kyw7AZG_nv^{4keczKBOcc{k}rG-t68ia6At&4_0S^fy~Pt#W|
zI=Xy*5jv^6e14gEa_)QI{<2_eMYG6X*k(s4>#viW|NHTra$NEhT+%0K+w@6Gh71qr
z{-kBv=4~jIH_HVRyCx7Vh=96>BNDkTzC1<e_~4(<?N8nmupF&>f6BE8_zn>h=G#2+
zpCd0!oi@lH=)tEaVwu$&fG!A+Hw{>IzsG}lz=<h5k#>n+m@^s%B@Gnbkv}oq#rg;+
z8jR4^;K#2n1W53}KD`e@ZN}Zcs8^we+^HknbwQY7VOacG6SHcYvvTEjZp|&i7sx-n
zz>tnqlT^>0`dG1*!x2#1=8gUCEH2RxLV)^J-d?O52v`&7uE8xo>KVBHlFi*B)JDu-
z{_AB0AJ<v4RU4k94~88tL`Rd~e%=-4j=PpHcijOoG@a3d-`sgfB;MQQAi6``?G3g7
zR~k7AP|YQ5t^}AV#6d5kk23G`4ohKpUHv@6aE@&s0KALXZ)q{r6$Qo#lsZMe1nx7*
zhh$?oV>w@OCi2sUpLu?mc`yqPyk(5aI?ZNV;f`!-@~|lhrw7EYi>&0d<EH-VN@=jH
zD4AuxmdfkDUrCu5J%$J%wri2NhdOg_`0PrOs7E4L>rwpW>C5Qh1^L_+(vLd8a>wtH
zN`A52VCQQ~t?s{)2iL(-zHStUjYp*%Gvhsf?x<*-;ted;#!LX}KG9TBurs<vA%EYr
zgh{AWZ|8nbk(LM<ydM5dp<jdxV`*#4b9t2<q}?Q?uitNdxj}e!dZ0#X?Y3qTbK!?l
zbkSi<;F5AaliB%_fa)b=lGF>4IF$2_1Nurb<&ov6hI9Og0MOw{El8nI1AaQk<#fLq
zT92cOerWcBb9z&l6<XiiRl4T=U}~?*?4{f|tD6F)tpZd#?x*6htPW20m#-Tl)>`t<
z$2w`e`~HH%C=<;o8LaDX{bt<w`A1aHwd_XoV<Ao0%Dx4DqPyV9(P9{ktN_?k{C*k=
zl@+qq%G;InWoA<>pd|Sr5UrUahEVoZSiUG3#%=^c{0u!G72Og_u?0I<0YiH-ra{u6
zEHPImDqwz)D`&3W{N&1pX`^4#@&BTF1q2o?LB18G!60Q1Ntx|GzV&jxvcalldKo+y
zb!795`x(8dwLff=pZU98327Jg=s3ZBLs)8i;hx!fLOP^q$EYa7#wzEL-^ATyO0~?7
zo(~?N$f0X%%E0Rg&@pgQ4F9UNqbuE?x+f4qo}wqX6k;Rnd%mnTzdd8wwo9wslqgD~
zqXr7Qo1?*5C^q!{a(sEo=xX>GwxW~T@Rab_66CW&HjEE7<w4x?@rInRha3-c^ff%|
z4Fi>{niZEbTyjG680Dm7J!wQKy+f6csPA-IqbzPtyT9#uaP)CO;a5&yt+R6Pw+%k$
z>fXM?mQb-Sr5M1jRg`<wQ>H(k!%y6uj9Ru63`%Rlr2a6(&D1XMfZBC|?fbL!9C0f0
zYd<d@!IQ&J*=+J96R&5xIQ6!i1MAsYL*bwG@uenu#zE+NPc3Lq<GG%a+<o~c)ys%P
z?|mjBQ9am1Gm;Yw5Q`3AqKMXl!@o+m6Nxo?2$r7`$8V(xyNklWrZJ!$IqNUnVrHDl
z8KXf<2MLGBf0u%VkP8>R`II58Y@>{|T&_$g)@2AzB=%(YN&bGXuaCK&v%;Q39Q4WU
z#|@tD$^+%X>E|VXjSLr?Pb#h%C0;zzrE#F2<YE?H?x*?_>*Q`<fb39ZXi;ydyq`_~
z`V@>0N--|MyHP`L7_xK5Z@r0#eQ_XJa78VTJOR*l7GKJne=B(4{~(sKeVrx@0Gwg0
z@Xf8U;Xf3~4Q+O2^U@M(Add0<CIv>Pe)?75v19@eYL?CK)S@wh^y)xQRg#>XwDW~*
z9zDIFr#K}IB^9V+wW6y^kW~WM-MS2;k|jG1m%|i8S`w1_?U!^0f?oD_q*ay8E0c#=
zX~USoj1TG`e3NNO-mC6<X51e9FcNXTt1&@ct`A|*>CoELUO_78UNWCcE{E9?=gk97
zCo5QMgc%}+x^y#dWV!Lg5{aMgI$bwj0E&jY=~9)K2GO8jVJqOGdVW~SQV6Kr`rJ18
zz&$^_1rF_PNQQO2665Cgz?M7M%CBueO4#L6gdO#$2WDA*kz$ISCg7Rwch-(OBIZZ3
zbmYauwVNJnl;fce2ciZj;(LYo6n*q)5KgvR9`YRGd{f(OTHMO?P%W4paQskE{H_6g
z=JT{_-@YqUr`gg<-9lL6`rzyG`%7w^fX4ep$GTm#6FazrzaW(zaH1L^xg0G55Qc0F
zW%%qzxdoo_XiM`j1D6sxX`oQ<-Ml_xcNA1T*ml+!82kQJv-iK)qX#;V8hAM|9kY+K
zL7a>CHxG|X*4b(XpV!wdP-o?PwparUW3P4J-|HPuhLW58^af0}fXFftS=dwjIkez{
zjJ^FgN$sXy`C|`-vA309sAH3h2}iQ|j{5Rh#~>&li4IEVh97@d5te|}dBeW~244MA
zrqOl37cZXf$wbM6*A&}YoQEeE79(F5+wJS@+{plnN`r531MTQ>V?*4^q+7y1IG3Ch
zCL!z0bIEpK05GU>8@w453)lUCtHR?A3IUK(ENI9cWPn0p&(`nX0pLn8e&;qM7mp8E
zS<gi$?`(&Q!)zT=iEueO8ldZ=YTwNAWTCbY%|V;am4k~%cV1qCP9co-0v!srf>H#4
znEYggq#tDNdw&C~taE469o2aT0kGr|dp<<`$I*5w(OuT_`z(R}X=%qd!O02Yc*g$X
z(NkVt$Ogw~C*!0?f<d%sh~&jVOf-#)>?Z41X)E(R6(C@KL9PcX1i#GLtX6d=8vsye
zWJ<d|{(nsi|1Q~UP|Pq^qBs5F*vpOia^51XlW69#DanBmyiqb*G4Nml?rtA_2$Fg#
zVCq?ocxP8+;ZpSQo=sd`DZlj(B>|(lSf9xt3;r;jO9-7ylfD9`H4&BPTMWM_;2Xae
z_dbxE7Yoc#j1OgT*e}P*+?!@k=Nn_HIhYjg<NC(c+yW|=L-zJ63Ei7}#IO{Sb&tH7
z&9s6Tj2vhlDAVa{q6!G0cXFC6*|+a%rt4YSW@l2|uKOXQ-ns^H1o>ih+HY$Xw`zcM
zV*u*RONJm9ZkYVK+`YUfcO>s};c5@w%`Fcor3C<NAWdm5#;WIpYn;v^Ca<%cXFStx
zK5=~Dg_)AQ|GkOt&DaPq?rOwpK-C~>Y-&Qg@F$$E(w#gIN1;eD50ue`@n@gtn|N^i
z-`4J6BSgcXq2izGj1uR@F79S`wC&KL|7Ozhf8M(uGxN{UgAn;6@wU!-o9wNIb=Pq5
zbA!4yg^_fPrShf+ohD*BO2b!%7Kpr#1IkwB`Pl=h`9BF||E>J3w_eRKirW4Fh5$F1
zl-B9cO;t`zv7@}K*2!)lV@s2k&^ogiW2OP`nR`uploJ!5N4f!o$ORCGQ~hu7NnY6T
zV?Jdzl($!V%Zm!Na#8DmQg7s_7GGf*zFdS}_Es22F_(xR$>q2z_c3j~rWaPMZ}N;k
z6J5V?pR4347FWTMjvFL}J*ZpiZZ0c__1|$U&{+#H&bMt8f86UCG51bh1IUtbk&}6#
zFG*LDgOU4}W6^69UONRqE?Fo98+Mx(zB;<F3a*={(EPpGePwWP@@;-7l-T}W5gwmy
zBNA5af<QbJPU-tb|Li8<P2xZ&8A~HX$H=nB7w?QX^c<wo1RuTvCbpN`?a3K?@LDdM
z)6|m%Y1obM7tRQ<|D4@R`}xJR{mtlo7vuB4(aP+LzwVTJ^wu7QCdLz@6<?FGBfVjW
zZ=+cZ-(Vy}cbYjJa0DSvFUQ+$E%)FNdL;ePiI?qSk+Cl-|D*_hRkwel2>zk?%I(3|
z%=iatreW1ZW(=XE8f}Dtz}pNoRm|f91-h1MDHmqD9KY9GQ#xm%y@^lEMe{F8{2`!v
z{+(Og@+S(Cap<8<MOFk*Cwut@23T;W<CasucnlYuP|JBKC4u8A0bNb1OSQRIalcw>
z{@UeIY_v(rTQ^Kyy&{fX8*U%|V80P7LvYTS>QdZ!vvfZK{VC5y11_i*HBLg+2^ghU
zy?F-?jH5_@x+wk5dAv#n`o@Vu8>o#GhxDYE<ZQ4)U@h<C<HuZV-K>GCxkeaPMllro
zs)I<=IhePfCR3_N)@UbmI1d4@7p;d#+Xi^_EPU4I9w`I}SEh3b&%|N(LxII$yBnWI
zul4N5{&6@`ZQQ|u(SaI9S0tgT<6qJ}#;^uiTR|-gwBMd!X+M`B+l*-j%xufMo}Sbx
zmd@t^AktKSHVwieIyHlbWqERt_h*auetr@@`d61yOa04zlT2QS-ut-sW4qrIqR(ud
zNioES)f1a%P!Xp7Z~EIFM1(P3xcFaW=)Yvl+*ei){aMmZz{A$2LOh)1T;jn;v_MEg
z3m9!`?u}Yo1`ikz>}g5o&8}7-#wbaKR+p%Yi^FQUsH_dDrEI6+O8N9Mh8NiNP8G3X
z>KR`IJeab$B*25gb<P!(AS)XAT!x37ihDxejQg8&Vv-wc4r44C?06#2?wG^j#A7%)
z_fgwUkF_FmH?Kj={bzOS{UC-IXysb+MpF-K^w924?q;_{quwj?Y(FfA0Z=S}*2&~(
zpU1dE<wZWd0S~Dk&eAZP!ff42-CwQawbrn?Cbw2+L>gsN0;Oo*g+RSHj}TY}?H2F(
zig4+chk@pO1odHHxA_`KCG<7iY3q%sUUP{tpmt)OQnJzMi*d_iiF1<l*;yfw=XbC>
zaoT2M5TFI&@773Y$xvep5BP*u&zhIGgJlM*hYm9k0RH;d$t)@SlB(4EAKL0)qP>yA
z=Yk_E8NegZ!e3Xf<oq?RB}#a9WgVy=K*!JWZ^PPOJ%;C@2ElOQ$pc+E6XvUmz|<kD
zL)QoWqVMF%NYIjCDpsaZ%pk)AtadymN&+ooeH_m2>g@?_(TWmvtpy`snaL(DCAB(i
zvqcFH`9Mi25Ahz8b=GoDE?`}%!S$@*lo8xO059jD=H{+~c9UGs%<d`eCUKalwNUlg
zhiU7~!~*~s#pjs&NS^b?S-$Ac*{`u#36xKqC{(Djq&n-#sy_ZBM55|6z$lqwZQ0LF
zxjAbzW%!+h+b5f&(Np0|N0!=^;Pz?{t#6;jY(=0zPgq+gfLD)K^Am#&>s~QI^exD1
zBM>7V3rbo!MPJ@6?2?F$RdUaA7VH5g@>H=us0R;WlM=IJ^R8ZZMUmwjAbe0y{pW0a
zlB(}*jmIqA<EwQPWB=%urF9fiUd&qNcD5RWe2N=os*RV|xpp}I&NKD6d5JWAHcvJ(
z0s*)>8uk6?{oOVVTbTD>(xR@JjK?kyfeJR~HQ)$#=>OyhXg0)RTxd3AW5%gAG-9-_
z)o7io(Yn?j{l|o@(|B=EfOy+o$AR3ko&(ul&DQmiMv?!2<j!YY&!<27cYv5>%V`?P
zORR6(H2_mL<mj*g0n-YIaVv#Q1uL^)6M1Q^54JXIgXma&Mgd7pmeQJDbtxg+qHL~M
zBcV{^yq1;gKrl8b!zFZ@pzr;Xef2RT;r&lLGj89n_Xza`ht<U;OcTq_B1xK<q#h?0
zw?TCM>L8SaMA+b7o-@0;ve^~VG1B(@I!aON!B4ROrY6<VeiXL62<bdLZVC28ttr-t
z#+EP1VbGR8f*Z#hBukCBL5(WQjk`f)+709T0JqHy3ZhaTJqCO`o1nXRTL!Ek>Qmth
zvaon%fL31H@rqK(LB}hDwvn>5-8iFZLJWEUb~xTW#RLZ>p*{jY(dma>+bXck5XEkM
zVxEk@C|?YBTn;h6oi|P-<_0GpWRB`{@IkJK3#9qx?EESR9)Ao%ClstB@aEh3#Zhb+
zI8{I=Xg?Eml$Ura7O>a0_hbC|ApY-X((u<a>GS&Hi+iZhTZ{zlD^Z<h2Jovq<uVmA
zKbSrq%--q4p!*o*1cEWk(==VKwg0ARc!Sq#neS>*3@rqaYriZ4xjjL=Z9NhChcX`@
zs*ln(yOe>Br6WrVY6V+kibu@E!q_k!K`z>#gR;?QSKX=KGHb(!5ok~F^vPE}WG;Q}
zFC^jFMC8h<kEr|d?PByhDaCXgul+GS=A4U-u9~1F56rZkwdE`TTIiO7nlV#OO#Nzs
z<yY0V_$#djYGO4GJMh~2)$+GN&a3WXhwLlg|M_Hc3c*{8GOCRfS`^Dz<PVh%9B|-E
zQzEL*`ipsVPS)iazxi6QW6ybETb+x_jwd$o?h}kkN0kwg=-MnZ4CJz?Q47)5Y|tt4
zyZ%_OSD4!5-c0HHsf=3AaLl(A453A#iu7fRYrwr752;5^ep-+O3J4ywwqXYPjro~-
zoM2@DFkJE8!S+@!E6}O|J*GbJX~@RNvo&AH$Wnnk*J)-OrLU!V;={KNVS)BvQ&(Z}
z448-BqkVTsJHM3XQ<lHD#BbRVuJZsqvFQL@IDQwDxFuP<v$kCn_<kL;ypKyqb$?Rn
z0khSgmr!uTwykRbNEfJXow+E(`G2$v2Xf~BAt-xfB}b0u6wbs5B5k$vze|a1&vf6w
z$~YTdZNcmu7M%tuv%oF4>RKnd@2y^}Ygx*tAuY%)>}PViHMazL{(%R_u6D)9vB^&O
zw5#Ws;29SKTL#NCbNV`^$HCnyty;vJllJ$CnYdoDK}~1zo-Q7}w4@TgH9spe#Bw*;
zWkJO#N`SMt3_2O%Yt0+(jUa`KrkOjNc6WXa%1AfRnRwH<7;z<XaI^35{Xm0Hf|=k;
z_`9YS4EgzsqYct-mzVvE9Jj7`RagfBQNvVJK%pw$DirYrs8NPTRtJ;03@Ar|c15eD
z+T{jzZsCniu<ZWM9xelY_KQNBN=?omm9~f?bO^Si9;HZiX6M!Bezl^I4&w*{5e#*R
zlWob1p!TQs7xjk_Arp`bOc#v&nI`oTco<txvB<C71MysG6Nfq<3N@s5AQD<s1>OZ)
zgHOS`Sg&st%4+233DuZBh3#*@b(~%CI&;wr$8Ogu1M-e+r@O}+<3hc`b@0^k$Bff#
zh{OP6>i+wdy?0;teTV05AyMyFS+jGpYg*>J58OP741`~j+yA7H{snS=`X_*FcyDdk
z3Q8(7`vcyMGp{F-frLyQ=c~iz0s354048_<pwcNQ^t7zRx-V)$CCXV{nl!2YRpFxD
zTi{IPv37&GdpVi)4V7g$&=nw&siKoeZ{Mtd4#CS`Uc5tcdaNpV>&#(5lT(DW5w}^+
z7n1OyJZ5Zgn9E$!D@9IaHUborh)%2MKZd3p^~<g4?>yQI7#%n!@Mw)H)~F#VuK<t#
z*6Q$9bv~E>14Z@HSa7%agw*RRCQ%5~TjDnbpdxcD2Q*z9@LzoqQ}okA$l-Px!CrU;
zW5kdHrm6H`NSGgCr&i5a<U%vB>od42I>dQ0IzV7a*a>NvU~@@|#)Py)^3<V&!ni@b
zsFxuBb(QIrYy=nS4%;zCmbZ!9T}x+9&=Y)N5}Z&qcs}Ga9>3AQzrAk}7V`TUU1EP?
z2>P4Kw6^_CG|d6;N?xo@*2Nt^p(PA9CdAy$s&lV-2ak%<%2_d~uC=J%y*?7^cVI1;
zS3Q#45$tu~x$+DPZ){E(x?MF9vl5N`I5j^ioeYd~qF^V&|1F6lK<@xSY3H|#h-IN8
zX-VW@)!RUt;tMY4o;@4pEG_A{go?~Lz3H}xVb%hoS#G4_0)TamoUBi(8bR>*-$&q|
zp3Ba&R0u?fU8*t>4)qqk@^MWmZrN)tDMno%Oo#G5am#sF=2D*mCPIClhq_pZ*j0;F
z`xT`;K$Q=9BKq`5Q{{)%ZBGxa1>ROySCZG`Hr@7_J%LO&c)fS@2PF}NU!qW6r)>rs
zZSDkD>eN)w4*`}na+?=bXQQ#vLUJE>aV&g+#rAINz~Q6OsaR&_@KO@{_z^U^CrTVP
zTsjgVzdX7M@Gnz+gxd61iAAklC<{bVhuSwJG8hbRoTxZ}ZPDsqR|u2uWN|hdivwy}
zn{J+VTMO<<+P&-2fO6ID=69Hf@X;52uTU&z0C*-_iO6S@kHu{Jo5g$MQ%rup8=Uxd
zDIoZ(^*%T^Gm}m7B3E`aO-sSDpn{d<p2^DAtqrg|-IXFTFDWJe;|tR5)_<_+|Ksrf
z6co9nVk35c%?q>H=y?5_eqnmTAC}sik_larTpqmT;tj2F%YiUaa@_KL8i5QOvp!l<
zZ?s8ZmR-qP3V)|@m7g#rje+o<0{T?%Y*MwaOtbgPETh5l2)_EqK)**Bf_@Se)AQ$N
zPaKR)oYrVKS{E#kez*QI>Kl@)rhd<SPQl)R9aC7Y?#S3MY-KvMb&7K!5VW7lS0z>T
z(-kbU<aDPicRQPx>#O=7jNXVqpyL*AV)?e9;7+}N<K)CwRK8rO$;5H(=t<zU)T%&v
z36%Bf@x<ZC`3?C1dGV>~erX8|tqZpZ_yA2eAlV%MNhE&D$Kypb!~-{H05^o0FNV{f
zXQw+jslH4YV;^WsQNjoM2G)apLE3YoS?F7zu1n^F-V~xdxgktE)LtF%38MOL2_i^7
zvzTCKCO#6qhl@VJ`|2;qs}=QU!$M}3=Nguh(CRSi1;zt2*(UaAIRcZ@%XK${apF;F
z-E-9`4-VLZzcdYRY`nk91n%ejsc8VmcQe2Q(Ykv?rQ`tC%Hu5f8h<<!$cOM<wIzH~
zVp{P@c%UZw$bj$HHogQCPiG|5F6MpB-4My|0ibe>`WmtVp*GScF2SQiepXV6$At+1
zKU~wqNJBmwCfVbW6;hz$wtBF$)wLlgDG7^RAHMmi33afXyS}lZvZylYeF3sMRn75q
z(HA+F;>0jX#gG9_N0&@(i{2@!7xY*|@bsw(!C<&Xms4XHHmm^pN;$9#M<00Rz&Oxq
zffyp2*Yu5!htmP@f7<~bHrFf;TmOkcp2`n4K-FpXWf6&YTXMob19I|5_hlS_RWxR3
zJ*Z`4hGIRbJ(X312hr<?`LMYd)-x{fFgkx47v8WB*5}8lozEZoY|sN}0Zvza;JI=O
zWmj)!BFMg?XhpBHW{2GSP2IR0JJ2V9OS4Bd1+a0k=@IDEaVcC4j2TsTe#GXU?Qx{^
z{@b_5o|gKdQnOLXfWKJ>r$L+W^TTN%iv=#l7xvS8vL`E){*Oy>4McbSQsMpOQM5hO
z1V*~1lAU)guG(-blc*%!1IK}PEX>g5GMpQHz)LF08b)RUk6&L;1;5Q&9fp#Og@4}a
z7ar5qT*}M*{+QtzB+0wC_5|O0&#ez5qdI*`+F=MKqUcb{c74)|8ZJ!5lr5ALO6Ltm
z08c%xsmw5&3dInz=zDh`b$gAbx*P+CcjeM#c*;sYl;mh-%JSMH15JS=9S@*?7845N
zYfGIc!CKv5(ZMPs!Jt@Yf{*81d=hRdP-ed62EaVjYl~&39{LV!yHT&($t7o?UwLWN
zk5X!Lz+C?G8v8NBa5S2-Hr$Ob$YYK<%yO63iR2E1u*=MV+~#c2ZH}LqfQCcV5_r%*
zNJRdaUmU5^En-!MS^#Jk7dnZNQ%~+s-AxXQ(JG=Zdc9QKwjCerYW~}}xIy?eaSu|~
zwW3jX9*wp~OK>C;?%rD#iJ?oBk`&PpL}!oR7K<$OsboW=hi?P%vP2|}TjnR0x8UCF
z%J{M7=Dr_!!o>$55QeGNu`|wHS3_G5!NWqYMRI?;k#^hJq6h$rKx@vWu0T%!!Fw<@
zw@Oe!+)oQjcv1&m*Ger+?s(Hk(9$y<xvaB98;9Zz2U~E{-L&m{`pLzbe$PhBBaf}x
z->+KFv%`+pq}raZ*xXHADj}7mTUwE-y72*{jSF@=4==BL`rPaZjBo0!VO)y!r^_7X
zMktSi?Gc0Ul|NRaM#l9akd`BnH1H;^nr{5JFq6^2@}t8yot$KXf2`gF45%!Ru6-Yn
z_}I94u9pSg7`PgG0#q9uuVR=MoY%e>KvH4Ok)v0lCxZ_hrRw}mrss3fv~y!^d$7c7
z``q`@XoeyKpxd(WzZa_gIq_x^{+MCN3JUcI+nku@7^)3Vz=x{}fR9HyXgAUU0DQ<_
zJZz5~eM<ZDZ_J+Kz(BRZ1~dl6Bf$E$$-MvVytB!sfuRsQ7ZYd?=g#-*$!dM8WRn2g
zQ=Mzx-@t1n3?PvIq^bjI<hftg4`al<8?c?-up0&a99cO#yZBD1LB2CGs&s+!&Vaz`
zwL>Q*^Ry+q1j(5Fm{_>h)75gutlFS-v$H_v`|(im8%P<!+(G~P(GP0|&L)by)+MPF
zlOoB@u31q+)-cq|<ceJ9H{Q59-Ca4rNw@GE^)=4tQTsX)pwO6JxXRFPu(oiwU`0&~
z7T@ILsL_<;rZxT1L64*Qs8+b?Wnq6Qp4q`$H@3loW^iK!iXOlp;7*jv!EoeyL)`A@
zIikC4?tzEOG_A5jnn>Jewvh}4|0A+G02h6){Rl+VPrFvz@07m%1Ar-7O8Uep%uhmO
zs1|@xT+*H*f!T`Wj6&_5H1g33*#Z9$Onkizv8BQ)sl%p?2!<dq2mx-e4T~NO%T%8%
zo=`1Pqi8PByyCvo$ULE!0Kkz1r>RCCJ{D0HRIq+1`{u`%zC~+vKcw%tL)<@rtBW6h
zU88i}<Zm~!XkzqvZ<6fJ96H1f-|dnX8*NK{)O_&Lab-nczFbKKpn9_xhq{Epu>5~s
zy6+0$4!Hr2dRyV@8IO_|hFgS-r^6a+uFt(TU+jM7U6+dyYg#|aD4-UwSv&gycV<fZ
zq}xV`M|Z273MeH7?Ugo4X$3#-Nq-bGmO7qJ|0J?f<~ArpSwV6OI_xpPZ)K>5th$bp
zm#BC>cYIg8)|*mowPg}~qTs&Nn)v$D@Fr#^S@M(TFU}w<77P1AkQ0N`8a5jCxtY=d
z(vi2ZkNV4)EYjFD`aI^7x_4{_(cyXP!&6iu!i$(|bJO)qKZvd<rzce;jIGxYlzg2;
z6pg}Gh|733eWX<N;Fkh<cq{J5p=7=Q<`ZamDBJ5f=&8-tTu=eSq(Qg^!0$=kzy{9C
z--5*&;lC=qEH=KGa6_ICktIt-H^Kah_Xf5Z=KozvoQ`2LS@o*KhHyI*z<xd0vhh6B
zU%HqG;T&{XW$`D|7sp#}EsQ%rK&(Xqcg!r5M0h#y*A-*#IGBqyxpmiLQ<%B31tIDe
z9~JDdwdgz`Z8#$@!}qTKTs<&b;gxI%Rc%GeV@JVhl|;O{3Vc>~4}4bt<j@}24=_)E
z8CWB8Rck(IZRwxP^Pn)T`bASil1g^iq=Yz~>WLO+Yx=JOZ9O&>I1ifehfl8nTkpes
z6+siTq8Z2YwV7rWvTnjhTqGoXTo-+y9_FXtggua3e^!E}uXM|fe7aGaP|oyr^V54u
zk}8&Hg`m8pB^6*RyNdH@h;=jg@@}U)J`di$8CVK^DHNqVW|#J6kHr^X?XYaM!vwl5
zV6^+>D;gM}>Qnr+LeM<jJjHxkON^b2Qk=Lo0S6j{v5RTz7_<+$0E?`A1(b4Ic@z>c
z!^r95AGn}yJYnVD&=0;Gaag)hE1+gjKq~>fi8tw~F2Y3B=uRi_#PdoWV@g}hMu$Mw
zAkSNF8n@DKa>nr7-MLwfQlwSOaAXYD0c?vwPYaEy4&SR*?>eTM`Ld@JF6@u<FZNg3
zc1{0=E^aF%8Gv~yLX^IQYyYl6M#_{S;oE7x;%~ImxbGWBRv9MiF0kdwgI;d>sB|YV
zv|jvE`0@Xahyf$Ky7~(zZjcZ2UG$()xVE?R{?<(Q<I0`6X(qx5o%n}1loR(I4h$#U
zqDmL(g1$$Ml2Zk)#3^OH>tRaBpx&isls3HeDklBmb!{sKtZmO9OFmrTK8jW1GvZ6x
zLA{nUU%Cc!rGYa0)Jf=MW8~>^WC{O}b_G2WeU0|=7tkOGpu?2x)x(h8^YwgElc=?*
zBUMGl=Z?7)8i_@TOXS8q5GJhE`TA{&uUA;-JXKgJ?|N~@u41bKU(nZN4sR-U1V8@(
z3i4U=>P|@)Ls<4Mrm0m2QQiL(gGO%;N_||qo@Ug&sxXVHEA|xs0jxIW4sCw2Fw^vl
zDAebBs8uQ=u}IrY9O^%?n{b1TT3?>F0Xl*((+r7QW{4Bnjdr>QTJp?TV4XR=xxqy1
zaudwGFd=8Gzw{Y4%(`gsWFyZcWMb#tG880x3hgwm#wlvQd(SSq|DkQW7zexh*W7fQ
z>Q^Z{b5P1Q-mVw#Mvq8wB_AVgu{&F^#G&LQ$=TUc!>5ck#fR-A0UO2+5JccMNcwGx
znih~dSy6wXLgn}`0P!{+|5UI|dSS_ELmvosPYE4L4AnJ@N&*FKh=~AFF)$^gX1P~m
zEs$?LnY}ockHfvchWlED74<wzynwZiqBVTg#3%ByE1?9LKYJ~RC{!cdc@r4AUq*h>
zh4yuv6ZXDV#n2kCZ|@oyQY8Pg5n6TrIcFS-?T*oA@V(*`+|Yx0!$Rb{w_C(OYmXU1
z<FJh>u)oNSi|x_rJ3XuJ>?aae^1k*%KReG?JUcdW?w*g(nc;H2VN^kuHyuMLGLGeO
zna&eAuJkwWhmDhiQh_JKXZGZs@n=}R{#Af5Apt+u_vOtnXgE-zpy!s!1>(RpUwX*l
zeawKh33tXBqXF}G1GdJRRI2Q#xCz~T7p|{NwKR>iPpF!>CiT8ci!&_^sP0r`$V1!c
zYa<u?*F-_~=VBIx{d)y$Jx=j#5_E4Zz1Z0BFDjFH%GG@)7Y2-p^-JJ*ZrlXey(c7!
zY9bdBMLVGaiK3fWhwL3Wz>>Coho^^~wu?g);ySVW!$y{W;uOYa&c^_3e>|5f<(BKq
z17&K^)81XjOEx~QZQg;S<mlQPArgBW#;VD_hptzQ?@iMav^CRcNyW4ZP5fXygTJ~V
zETQ!-{ccl}wSr3F>`z+KC+gH<$vBekSohSM0F-3CaLdEjS8}f8*eVo@M}<hl1!0XI
z)LoT?>2qAhmI<@O@0Fh#WqO^B#Y8PN1`^%hZ*{F7B{}O6f0C=!E#1<4DI;K`o0KAM
zVPR3TfN-~xFLmXG8q6#){?U!BajTuVmJ?;yQRXcwoO`EUspMK$3<T(nn27a6q@21W
z2OFMN>h){{8BH$O2I0%;iJ#8CT)BkQ*aQkKl)y<M(WpWRMxm|Se!odgA~x?fD~thg
zmMc&XXB#C%CW2o&_<$n-R>H|~P=a!(_3KiJ#FKhyE7@q)6~%yz<Pe=J+E<&gbJ7wt
zpOjhTAwb_z{91zpXa_Rl<+^H<IiVy9d123$I`;QmrP6T#Ml5~XSn4kxRT+=g6p2_X
zYCXEah`*b)%=r*#8(r(<01i=zbbA0=L*;|y9y!({OmF@xio}FQ&o;+nBO(`O4~6xN
z@5PmjM$-@A-<SY{DEtS}-37i0EX&0rV7&Te<-Q#k@0Egu$DyoshOVw@h;jH#X}+gi
zJ7V41^0)*%a$Jc@t(i7%a%h8TTl{^$5b&2=$vx>!9FO6CTH@=icT4#C!w{>)&^B>t
z;WExavK{B(_RhfiXkKqcmFZE(EmLI!@koQUthO^}bo<XXG)8uNy31{j7d8O<uu+WV
zFLpEj`T6Qt3oL(^gxiqbM5QSZl{Nuxb15`nRJHFKqI=|lihM3QUa6o!zldB#BA&7c
z9pd*Bg$<9%cdYsXq@Ed#UgahX(qYhN9S!$QmH6}k!%Y74I+*}~h?{)h`hkN68oX$+
zBoNo4dX?b{^b>&p0DSvF6`a_OXU49q7noe&@LY5{1OWyrqo$_c0=Of#t_Ko<jVBS5
zUC&nsfDv#EE@o#%C*O@jfqcj9`PG;6Vc<8()qWZIH<~M>DQ-u2mU3K7kNMsDIr!SV
z^@MUx($4|DE7|{?hwxtzc&BMK#cs#QjCbL;`G~|ufVI;IsBLaEWzQoQM-Lf<vE>4g
zg^*sI+#$yT{o|=N`^{c<xXSe+tz=sF)h1z0TRmr!%G5w<*IaAY<X#G))-cO5PAaBV
z(eF49hR*&}&tu_4?z(DXd4M>^=GOFbU?S2?K<Y?Y;5(>?^@;g4(22ABEO*1rT5@@1
zZ+s21e4N79kT`6Gp+u#P96b!TE&_0S$La6egX9xi>XGuzL90Yb_~*gWN#v8aPy@1s
z_BAA5b&69az|*Zb1hl%AzMjb7uzh@3BWqs+oLhVV9L~Ehj;b5-3`3#1%dS}Zu~s@F
z5XGrDL6iy*Ea<2V6_swBI__*1?Ea`AVg^cbgGnsh36;!`<8Qv%4Eql5mZOU{Z(-Mx
zUP6`X)nm9jY}*LLm_BNm%Gpr!H4^~_X0=kP-yZxiXvIMsOgssv6)hZN36rMN@gUL}
z+ghYmNZ=IcC)is<GcGoBfJr#v=O827m$IM>e~G($HIV=G^eMqN*W5;92xt6Ok6oDL
zwSkh%v^SepmzU5qMK6-@{a}fPvl$B)1`c@W5*{=F9V!*CW|ak>ToL5LC|y>a6_h<U
zrdgAL#SOVnyk9Ekp!mCh{MJ*Kw?~c`B4$R)h5f}m^<?_U0_(*h5<Y|RZkek&Ah<KP
zX%J1*R1%RqgID4WH=cf569`yoCsS*M>bDgFHGXD4wEU1I7;3j4@>W`MqK<l}Q>58^
z>Smsu$95UhzDLHa;f`yYTB{<qN@-FF`0IR5VE^<5nT94*UB6ORGZ%a%7`4piX1v#8
zh_GQEJE8TKch&DHKqvAiDxN3yn;<~C${?zaRoWjYL{;OQxznV?JB;u^j=--1ZD{v5
zP51+V=T)3n1b}B^O_b$0%ki^sh$4m{xNc%c<!{|#N9fsKYI~oFl=*NVP}*R3HtZ7v
z{A(Yr$G!n_i$5)q&?S9BJIwb$tcFJ0#{HXh6OT7Kvh9OL%lm}AENbBW?nHFI7N_`<
z_2i2gT)H_e35U>3dqX4-*;+)+NQtil9l(~R*$dY5&}Av@CgC!c;L;IWR5Hhz>5)qy
zq^C)h9WxcxlmIuD@G0wgfj`2z98hFs{!KYw$N6$o<?$-gr1W+gQ~Oeoq-WWtS9X+s
z1#5iX=Ji{vzRwIl8LD|)`Kb|Lu<G7olCytv@e4WUt!@9~Wu*OwQ{zDd{EzqJ3#1W4
zHTOE#3}+@um07{QAiQ$Q9j1~d$en4s<i~x|fYEP!fkmo03xl_=GZ2u`K1AY-I)?G%
z#gDa79((y-*cO+M&C-Bnmc#!+7?IIMq%Dk}1p{?%6G-R$(sm0wy6?;0mbWslz`o}@
z^<5fPEa@!pu;v;ckODB)-n}55CV%(}H1zHTR%Ty`(YQ}a#38Eq+o}l@P+Y48YUOlC
z4_?;T?C<u0|JT6F+Mgc1M{M2PqwwA+<x3_smF^s=HxeVQyu7n<_84b&YW`nkC}y7X
z|4C+)<MAr<kzbGdZGLf?<mG579kn$Rf2Ru^EHGiR>vG_=ef%HRIcHZ%4=YMW6IQGX
zG8t+OoKFH?US?>kj~zpaO`fU06@bOfpX`m7dSgs-(e<YGhwq+ew1JWpw8M;d`W}oU
zGfcLm6)VIgaDhWoyW$EMb)UGp=eR)dpdq*)WCaRsd+%P$RJqoyze*%7{{cKq-XFWy
zrJ_x}qKoTPF3KG`otxNdCOT-WUTF>7V7BQU8|SyayZVASTmEfL5mcSv7jSWw!h~w&
z&L9Nl+nf>AsVd6T`dR3}rPUmiqFJk-cau*Ou^$ll4d7IIzZI62=F^=0b_}HP1YEmz
zSe3pHf&l3{B^UY;>LAiGHRBkY%V~u$kxnLBd7Zh_nE;TbbER`Nb(L>VoG@@Gn$ROI
zo~H2vLaBecKdhj;(*WIF`fJc8uneN3e|2^=6CmqvxFu@5Ws(8}*s{>J`K_IM;cu+-
z1Zyi7!c*V;2MhZ@^l5*(#UhkY%d5ZQ8UgnyAOPgaqs1YU{vFz&uV20oI=Xt`VDER}
zoo{I8qyvg<13?iUOwfB9^kuHhKwY!>VzRhLw2Nc%w^*rX;%?bZDqdOSFb6$8>w;_@
zwzr(#iZ2~9`pd+zIoWsh2=HO4joXDvyh_<&@f6MRB=-uV_L1w^uK_~ehpT3oH4rG+
zK@zj)Z#>rrB8@`vn(I#U=8Ah(y)$EjX_U!x(x;9B3wKRb5noSg{R|Ey25KwILcbaA
z7hw51&zdl%R={w+-W&mPFrR@6IWF62zJA-P%7Plw7d-wl3#r)$<<)vSVBd#CTsAn>
zN0)=Hhi?o$V2hB4f7a@khPT}TzQUR-4aRWZ7#>WA)mC8J#$krSX{^iWafV_cRrVGj
z+!OYDF$@C9gjyjW4Unn%(GG3<ZQ*d(DAgy<PstrcaUVK9^B;}pgwB8K<&I&JKSzdq
z9dF%s&x-lo^)1Bx@@@JE?-(tKM66C7KceOO=*j5u>P8toYtka$`}%9XIV&wSwe;>p
zf}PZXmgaxC+5ZtQ0R(FxUg9dHidoe1kE^G!5Hz@JOzNmdSv$n$k{vz<LO&orfV<Wy
zqC}XH=iZ}7{YR?;bw6NE_BQwJT$YTGqLzW2XLsbmBok`$iO7_@R=IC!T5gpf$NEQz
zk#&vlugwgivf*BJSG^oKF;gEDfUCj@;MqF>DMVT|*${u$HUcuRs5tn}!X4+Ib}qx-
zJE4Z&&86DbcEcm&sXyv`^?KiqI7z|cA5Kl4Kr=q`NU;)?tB{^m_^>II$Zi^(P*@$y
zP38bFpF&#k`5bdKx~gTxr0_rz(fx`wV!o(12mL_-WcT!a$(;=cO}Q)-;A<_|%2ksu
zvbKo6+q6(kM|SGT@p5W1A3qdi6~COpf1N#lfz?GLjOKz!CpVEtD2jTKw%urIu?Erx
z31xSkf;z^8)6Gc$px4VmKtffkXnO|#&18!5-KF$z^Yq1Lm0wX3O4=LIVq>qPA=l1G
zO47kM?>6%E^hgj%c6Q3l&0mK<bUFS7U;MjoMRDB!YWhn`01$!#KB|)R6NF%OhvV@Y
zT)>yfAn*gg>WbanH!^lcgIz(kH~d-nHK%A$tb}`Js4KNhi%l5`_VptXL-*tAPQ7o+
z>PaY(d4$IW>|EBiN2e<YW2ny*7`FJfI$wV<^Q=k=Cf#q5ui`7Xj>ip&O*}ea%R5Js
zMjc<~G7)lZOj(3$-0yUi1MKI-duxtjwn}|B?nz4Mc2ht+*U@xzTeOR%#%H)~!V^nW
zz?BvF^sfEN!geZOc&r_Jc*_iLPn*(GLvU~Du>VQjOGjUFHz-FXp!@d$GRtj|js6fi
zwL2M>o|cFi;j`zhDJ~o$1L`-ClXP;(`RlM(s&`W!7TL9V0j#_i6e`I$91xQ+S;39I
zChBlXl}6Q)1HK8_A7PB7^$3Wd0wg^+v8?JWl&wkFS$`u}X=g7@b+B<;f8(_}Y(rS%
zCm5+La`t~0vW`ZPUk5d~I905*RJfo8GF+JVsmzcXP5Ue?4@$7fy{Q8#oAVtT4gj7#
z=r>~Kn7n{W{$B}@MkFH2I=+w+KqT^->kEuW!w&t)gKW!m2_p^+n6@TOYSHsj(zv)~
z&<46{e>MQGCDFrfO8Nn#Z>F0swZW|~o4b=U_`6O*<a2dboa>dtoOz8v0LbUi+a*{n
zLCzLW27*O(m>gL=@|3}KUA8{Ko7*NH&LbnvK;DbGBH5KQ<?NsV_e0m<-=O?Gc84qu
z##MXH%YnSL;ym$m_2qcW-N4#YQtcJ=$&b$?iFVlumT-7y0<541KXbG<1N~uV%n~6m
zgc}!p4AD3XYFBPrm>L}3&(&QzNf;w;UOsp$ak~fO)h94<kDm_G_d(UAE4T=N&JPF>
z_ww0VFwCm-SP{p2AP<$=|6VtHyHFM&b)Q3?A6Iz|K%2{aE;mD|o`^Nx<;f5E>cg}Q
zkp~C`Fv0(xp=7vzVMinWi*cxw{hJ1p`$TIiOzHNXDSQ~}!I|CW@W0HJC6n*o{tQwv
zMilLWJlb54Gsna7qc|{BEDpGLwN8VXb^^|kRqXgfS_&3dM<XB+3LK$Aiy4^R!8i8N
z5>Hd5MaUM5&q~yVU#H>omh6-Q;DXYq)WF!MFY_?9KK<4P!MbKPdQ+L|?v<uObaeUF
z{Teku)eAuNsya_RU@W|x^5n!0f=z8S;E5HLt4Merh~`kTF-1tFke9Eij3e(wnyk84
zIXt;uwKA;x<;7uOI37kFnFrn5sSA>JXH(H#6(HCs>IWbAIa5^2Gt`7(tnlgeaa+Wp
z+e)?qAQ=-c78UFVhx=ILYsXZgo=d@bp|pHNceq;Jt(v@uv;!A{2yJ+K8~q&gQ|u^1
z?cE95_GGqW7eaxm?grR{P=n>0T!~u-cVQq+=H~B8)Ad#1Ajh;Osk#8Q6>GBQ5<N%1
z98i^oa<&G^9D-w44|Odc*&ccSKZ?vj-~EM1{E{9(BSnL^y77f2_xWAFVURPVZ8+ao
z9Byy_V!QRLNK47ic^7T_?;gF+%>4Bht3=hrBz#aai*}h#*-JigfpR_E*691uRF4^-
z=65nwrGBRrV%!JeR?v^VomGjAVa-Ad0oqLomYhz>&)-Mt(H~4fZp_(_dp4acHS8TZ
zt58d!l>v!%$%#nFInC*xK(?NV{t)(Yf6s*J+2d;tx%sf0qlv*Ez-RX!L)8aGf0?N?
z5PP{jGV<{zK>^;Bk~uajHafpgXmv;dqvwEWHrUFuT8~0)iOy1hmCJt|bcZqvwzJ2y
zKoCANWJS-94+^5{U!8>NGa^=&7h{F#sMQ$PpwBZ}QOnhw*(!&xv|wc%W8@Fdow<64
zCvSa`EB6LAqSNfwCRAR+an=3z2+E!Q3;qt;*&v%j{r9Hh`@dOsFQSfUq6rmZIE!UB
z(vDD4ed$7Cs>)PVD_|%Xe=!tPzW<A%0L>-{_T{1{EXct6Q;&cRE=QmweOM@1KHwSO
z>RMM*FbA|kkDTNrRM<{U;*Ru`Ken=ylX?m)Rx_gyp(KM>O^Jnl92osuZdE37Cp}S@
zy-tdQxij{qS^oEee)$>O#NzHo&drY$#*W$9JKfb`f*u;yyvQcH^Fx=s4vz(vN2muD
zS!O41ANkDpnv0iYo@igG?|}e%2ke#mUOpCk{C`lpX^sO*h;k$>$jz7eW;y;5(JnJq
zDe)N(MjEvox>KOu!VHHy*@~{m^JQ;y*{k*mW&21k(3XxmAgcRj&!JH8Ashyc!vQK<
zbjx232x4F5WPof-Miyr3+6$)LY3;2QV1}KzRho-w!8XtKH9(Yg=CJ;B@w+!>Nkp1%
z%aABtn0N@;^B;}09n+uHVR@>dcm!d^0&TKGqk%Ae{*ABbby|TspQ*Edf;X{%8P6bI
zkU9F7dh#D_GoXnAYF?s#g`CVX6So(DNB?)Mm+r%(qfy<@_~U2rVFxp*AxOw3k<ApE
z{vhw2eil|5t#K`G`D~dYV-VF5vzy(Sgi6ael;$~e>)_J1F%=GnYzBJQAB}uZyYGuX
z5g&LazI93GN1hjw6Z2HlgtKRu^Ys!yyenSIIPHW=rDE2%;J5Zf>iou{jS&daa#E+7
zBKiFc{w&NXuBL0i?KWlW$a}G6h1o&03y4RgbdT5EG9Cw}ps>9M{#b*8M?LuK`xXv^
zrTxtVekP24BzGl!B|1ITzH->CMZjql4QD4{0bc}O%Z2=!S3ba!s$L=oD_gS6(!_6T
z_dhK@J4E&+5})lJe2Jx1DHB)2gfq%Q$8U(kc;3ec#?pKPjdRWb1EF1@T_BD>mOobZ
zWjS7FM|Y7ahKI`e-Zp?ng}^Flz(V-5ZFu*Z^oYa+t}~*RdCwS?>=yoQBe+;e|Eq$O
zZ!Uc?oi7yB6pST)=s12{E4Zd?o=cYm!&qaK>_6+0SmXZ#kKPpD)9Wq!Y37eV<4ykN
z(QS^y(wVRMfmR<R1aV^o0nmo0GR-g98MU^JV*dG!)<Yhq(m}_2s|3rGLwZ_E=ytks
z@2g!d@-^+DMCFA-+X-4Ju%?4!mLpJxVR(cYetD`B<rCR!VUs=6PkP9cfSbAnWi(f5
z>)TExbiCK_Xai}x$a6OH)<j3yJ)d$Id7;gr4~#G!frpztR*CKr*1?FQm~b|HW59L+
z7Ax>{+v%mYxE!hbz1st)gOE#L#D83C6dQ&FPMgbt$+%`K1cD6Wbpo|NYZm*N)ZNKK
z$6AfF=pqoq%iz><d^pOM$3AwMX4{)sz1^16Lm<9TTW06<c;;GNDk69y#$29ajbk!0
zG#z#Lbo-!v!0B-pjc$M{FIs1+fY3F7k_u9T7VK6W&pz|P;|KU6ojH0wA=Ot)qAQB1
zy?a_O+9KC;gWWIw)}Nf<j{o)Ct?q36ghf$0$d)0zP?DWaREl$LDPP;s@2{6TUYc}<
zy8T5r{O8H$U%{TLsO^`h3j8-R?ghLknX3@Z<-*(mlCAn^$P2d{`U&Y@-w2{JD4q_q
z-@ac)xfz;O6Y_<x$BSN)n#%b(FncaDxv(*-MoO@3#CuFU$C6^$6Ht9W%UAhIl|zly
zWT#L$7uu81+4Q{%EPdC^ZQKXxzP>7Q92GbMh3z{x2*`uFbW$$@L4x~^cXO+DbaBeE
ztH%D0#z||9&s_aR+M8K3_&&Ht^Qkfal+rJT*0v^xf)TkF;*UJI;>7igOH%dDIJhUL
z+CU%|{?m+r&Cm}WAgGDES_JF?`c0Deyx|Q?@UmgI1$GZUG`gaxtG~-*Dy*r~cU%V{
z4xyh65&yh@{{H<05h624hSCKAh_US~2X-6UcAkJ8j!t*TK21C5l67niJ~M>q%t@k3
zCO$nI<y2KkE5n_1gj5?n)~Zu$VEk>%xmd}5aGy??ab<T%=cWtZQ)d0FQ1ghdI(2Z_
z=1=7M&&QAZ-#GFDsLK8S7EIYp@{7(r$IOMH%`Y*QKc)o?`6j@dW6MZLpWC|sN`%Zr
z&_mR)VWjU`@P@dju-zA*o{qfz;i6nvH&SybE9J{a)bf2y-0~H?4c_QjjrpWln9Yc-
z9_e9`OP=Ltm&B8beR(B(qAQY_c9>5it!?1W+>4B!r)@g^0v6^;Im)k9%!wht(iY#@
zmLAF%%PREj9t0>tOQJ20#p>Q?1)p5oi+y_il0fvx&lz{)iG7NK`^l`3(+xmI5>O+2
z4VbdkmFbQH3Fr1yoquDlQ9GDr2Y{#!<oIsxAxM-W)Us8f7#`08Ek$QhQ4iqp{%n^I
zT=%b-2>KA<w+Fh^WYs-U;1JcKm9+2P7wT+d1ZfFpUDqeJ`5^*liYBIRCE=Hj-G(kU
zEbJ*hKk>UoK9a2eaDTNN8108URJ0<Zpo`2CWJnal#4sd^W+DU<#dJX5y=#dQ<d6PR
zza2EZKz71<ZbY-yl`iC@+5=GTUsp-}t438&A5p}<+QC5hB~fKMa3MOIRa={Zu%CX!
z;?tpq@T(%)ym55Y=~u3AHVMjaT{K5$?Ct^^VflH!cTzm8KaI5EH5SK;bHK8dR_kd+
zDT#hA@B6xe3NOVa(vt}uh}ee)!xo%r=q9V1w$_!_%Z25W$H>y=SHt!%bvcNy7l>c3
zNo>9q;mE57J&G<W6h|kTCqGU%Z)Ppg@{^(zyqvG!Fq?d{kRaLgLj3t!+`_RQ{KZ}i
zMxf?OFayD`u*eL<rUzEYK3VQFOiPVmGtv5O?0|9&XakiCysCOH-Uf+mDj<XYXWfy8
z9Q5jFjy#Me*IGpxZlA8DVH%n8qx)-%VFYza1_oiw9U#j09*;j`G?9Wp9P%##D%C|U
zObFP+!2`D60FvCG+kX$-t~hy7Fy}ZG({!KZp+F~AKI{J7*J++OSXsm@^5)gP&+5sY
zHs|q+U|CUXwoe@ecn<fF_x8Fb!GxLmUpJ;4?2H0wRR=qx9x<=jv{AJxJ3K<EOGz|*
z?{t6lbhhu#%Ijq}c?l)~s#BO&5E>$3B>qJ>_l6%#0>?2AHWT3_13#9ITRfNRN;o}O
zI&!K^xE|zUKPs)V`KikEp!aU&V#PH7YN6wxs-PXG2af38%ja1%=fp^~7E<j!Sr{T<
z+VgFC5KU>R>4bfLlT>4AZ>M_*9w-&%o0_!xsFKz2+fb7+V-@}5`n?D3g=5lR$T-9&
zSzU6{#IZM($vp|KA4%@AQF6fX=_C#-`zg#Gu4^{5oR@{}i1f3W-X-Jl^-Ed7_RJ4C
zt;r9o^~OAnFGba3v6qnHTBM)119~EE%YZ(XRPbCjup$*|+)v3`0y?6|@HlFV33^o>
zFmlp6P|J!|V`cOq;f%l}sxt?fKXdg)f=p*J4`zwO8QLJ~JW(MV{y2ZGTw)!CQmdo?
z`$7p)|9avq=;bAgPb74G<D#DokdvS@Xd3F*CA|Uu+TXri=)dJHAB2^!0qWjmXgOHJ
z@8r7yqWcA;Srn=ra8X)n6OXyn8(x&`U>FFjLh}GnzMCpeoo-Z*UI&iQz>5w%Hid~~
zhajRNtBKjr^e5Y))XB#nLKkJShLAGH`8;GGm?tzB_$!pLYug_S1REWvxpBsUO=hBL
zQMnN&t3->r7URV`KyP{#`#9;H10dKgh8tHS9$eHjYcXFXHhZJ0H~=~gTR-t^(0u^>
z(F7sTuXe{#-KOc#S({a5;CZ$smvn=1Gnat>e*bMQdT=SGbR5tN&Az?@tDY31dk~)<
zLVz<4M4bWRbzT-ma|G&fFDNc&0!SJUE(hOdYm<3EaWRC^xk_iq-jav3LY$2cHU1zH
z-vN0Bk>!k>M7n(C-%}>ts{;{MN`he9Zl_<=#@qxTuIb-J3-d->@_-#qT(%Z{9$?Pr
zN&G7#9y6YJ3~ZnIpH7H>A`iyv<4~L}%b<*~0Q~hPkz84~jOJ7iLZM_NR4^81HtrV3
zdEAtK*hFgrLxV?O#HeV4`_25Hs3jY`e#I8JQJzyoZ~)>z-8?M3tE8^7YoSdk$a*)q
z)`v8+aj3kXAa;C}@YXQu%~mf@;@N2H2sw98CziCp`HT6pc@yCVtHg|sjX;wV)osPk
z!g5hh$5<8>U8LHkCw~MDqQ7=hkVzmdIYZy8coYanMviFSmNQ_Dyd6fW3*>FUpzA+f
z)Nql84ZF|X+t0h;bpGj&(;v6q^7fO9x+eS;PdB(XpL)eaB%X)^2DJmE^V^mZP|bmX
zwE5>2(Y=HdlQ_gKOAD0Tj39DONY#u8xv<cj*G%(7?8&VQOanc*IDR0_x&jXA6C5Sd
z`eNBRqV23WOmJeKspy|ZpN)b8iicTb3g@TQ1da(Ag@Uu$TCiB4WTXLSl1NAf>Z||0
zpbNG1^R;uxR7EDbCVHp3x5uLewY=zMk5Xh~AiSd`mDw5hNlAo*MJ995bj0GPRaRG&
zkc7(8!m9_@P6O4VPuwH!Kttg2DyRP#87+2odAnsqJay1G750Agqk3t#T2zkF<IB}a
z?+TL3c~)fmaL{po7hb=#5+r@yyWtuDy>t%*%n6s(j&i~6d>k3uoe@5z{~u-V0Tgw%
zc6$>gDUuN-3p7DsKnX1wL6S5`1_??KCDRC!lOQ=ZL7*iGB3VG9L?v}2SwJ#EH!Ywd
zIm5TXInI0DdrsZ^)l^wERyj50=D)*w*7~iq5$D~e8k(nybj^9*aG?|+TGp3HMsMy%
zE~6@*+|)0vOZB9l*4|KE2z4uMaJhNgJjgh;>|-e~bubHqeApMY=sQIH3+~DV0#=_+
z|J_wi^S<Cd2)kR~$q(@T2+XPlzJe+f!5tEm;Gj9WtN#SDVf6ZR{sw~KTTlBc@>Mn3
z$d26u?_X@o<dh=tA+G_ZISC?hJc;{;^h{qu&}c=V6GZzFerJc{d4^a~CGJeGky<F!
zi~ZvgF28WLRnhq$Zqwo;>5bC>$DbrySZ^~>&wP~sTG%UBiFILIwAlUs*Nj8m50})A
zaAF%A@dD5UzF{z3>^xxK0*XC3y1#5`;zIB`;j{0i<gs>?>iSc9`@I4O%gO%LFT75m
z!Rtm~$5E(9<vnoh12Kn~yhKOG5=CvNC&r%f59y7K?926(oJ3J#s~O>KmliX;+C!1d
zv0@?;^@e(cnZh(XdrBLP@4Vq+4>RpIrB-|lw90yh)97UQBnrj?twoE3c$;gHQ*mji
zI_3oRmu|~Pxy!D=hKSQ8Z>UE`ANQ|$?tq-r$rH8(e-N9U@I?9L2Qced?A+*kfkf6`
zJYbPI8UXQ=1)~Fi3cpqC3aYm{e>rr(Mb&PSn?xc(cZq6SX0XhAJCwQVHXyDDvrAAl
z;(0Fp`~kWj2~jVv))0ITHA}X}#m>6~pOm|2=8iF$YxkWH{LAO|z}NG1RQ>VNSMZrs
z&mR67Zwolb5W7};|6;PCxf#@dyd|+kz<F!y-I~FF8CJL9|HilPNd~X~K1Q~i%CA_X
zfH{-cMgcgbc)?gki1+D^|6+RC1RTQq{wHBXnRjJ~$UE})p@GJIZ+V1qGtQ-z&UIzv
zz2j9zczx+Mdt5hHqNSgEeded=%4sYv`edbeZDhZ5ex=c_`%O(@M4hY9+6J(NQV@f|
z=TFN|g@ALqz737Rly$e6+$;62WYyi=Q4Dy~{*q7CKG>B{3lasr<rI4vxDOQ=zkSwH
zj*Te?0qz9JjIBb{z)>BefmR>r(SkjWj%+>kX1ltB@QIAj;S3B0TfMjg+GRw48kKFW
z`5yLQFusjcuU0nzVJyv%4|3qT)yzRCcx(#(fJKc2w0`uM6)xJCJw<bhdzhn&0I_9Z
zMgN@c+srM%hBF6QqffGESVl{!fB+XBvr=RBcOqMSt(b16`xNYreRwAOZnUSbm9r=2
zF8`^bdEP4c@3Y2tS%~=K)9{?&N{?u6ERVMu!$(h`6`xtzvJYjajJW;}fXcr$4JLH$
z%ti#AkvXPiqcSMa=YuO#5p1u<5E{hT<Hq+Ev`L}zgbIkH7?#dHn>jG>%JTJ>NapPV
ztt2LE-Bor@-sW5URkvVbdxlNT71^d&%D;p~9O~5@;_<p^C@rW0ETyIv{{pb;-v|SM
zZ7mB>vU>sUxP=i>*i9Khtc=9jujt#S&X6>!zQE0|UL48>SJ*MT{Bcus%zVr{iUrWY
z+2uF1*6GG?rAM?Vodrx20#X8w&v$LU=a?C{)tTCxUDL-dTAplrbsgO>zSOO3%5a2k
z6sTOa2S(v1^`tB~1MjpD-O1-Gn=-{eI;46~<37A`7ViQez?UeVTqWrtICePB4PW1t
z_B*)OgChEHQOn~%U0ouHt@Um@1-g|i7_e3vmQkeih@coMxu8R`4~+6>hTyCL{ktBm
zDZzfXn7_=C&XaasjnG{HvkkvTOpHyg-($}vpEqhG|GAYhoNaXve*yVz3OsaVJyI3^
zidDuM={RYpCJ_?-`Ya<cL-*T`^8wtRmH&-VYw~vFO8h(~*Ze7{k{myqS-(K~egzcs
z2x1$`<Ty3}V8}6nTd0lW_xDkt748Ff;yEA`uwS^m_{`gGB=Xirop|NVmJ?9^Cr)}B
zKq1^M^krk$ozRWk=^8?DlpAT0B0_a`IBZqIi)1%_Z37}B!5#nJTq?0PBif_`?W@7-
z+RD&A*KmE(uUB(D?<PPG?iEE0yx<^#sC-)(Bl>EWycCHi!>!si61M>DS_54=R?xIh
zjEepBJm+`%h2LZ%dr-|SoO=e=bAPcWt-S-iTb3ZKA}#h*FcWx1=TG5aoB9}Qzn=|A
z<{gx>vF3A_m9Qa92<O&>G~KiSo8XQJh(eSB6omYm=F8o{msFqn3_Kj#bBZ4VkasI}
zYeH*TabzRtR%!`f(E3%m-XczEffpdyuJvR}{w@7|tv3ANt1MdcE_ri@Pv)7LqNC$A
z4keKP*tOTh&Jf<6cXA_vzY`dAz_{Qz@f-bLclR{ATU<m(1A#L7w^Qmrm_)(Y=(kf!
zbY!mPW7Ex)C*JB*AWhB%3A01Nej)XA0Bu!wg&1oK@ywg!08pu$3VYO=a<`>}6cf+K
z3FtVTlyLF_Lj5J(;7Z~9Qi4F0wmLd$8*sdg_1cco5aQLC5%2f9!|OEU)$U4@qZeU_
zuCV!J;!{ntt~&WEKsk$gT%swpPZ0VL4)^s27itacob1CIR$w(mIA+rgxoF0hu}1BE
z9<_g@0+;-7&m;6wCN`I>0k!oBxCxjA^XDccKikZKSnD&TD;kCQZ{DEdt=yOvbIJ+v
zBpdF6zx-FY9Y825%>p*XqxHrx@$z`YE7K@hieyAJvgZ<@2uGJk0b9MH$tN_o(+R`b
zwogJspc6vGtfmt5tQG$e8Hg1D{|yNN&<sHQ`W~L}{4OzQ76c->#pSN7kS_yKJy+Ho
zfu)K)&aBaG#}5A;uq8&Rb^rO0=+6$RI%blSuft@=gr!~I>8&v$YD~Yf>O<XVlbzA`
z{A3&(=}!&4N3fCqj*&74+JPxx%nDE?%^Jh}5#gJOmF%SGT#C~SbS+3WXdCZE05-pG
z3kd-zq#(W+sa}HWIJzETf?ano%{Polh-}OMoVM9(DRGrPa^bik+mua!rHa++g@@i_
zNyLML=7`576pj8c&7q6Qo__v!S#^_Bl~LFG^1qFhCh+hU1iNcCj==Q+w7tFhi|g$o
zQW;K?Vi5r!34Ix;|11vLeur!b)Q-^Ja-Z;J>0#ay3DRNy)X@q6R(()?*=@g^88)u_
z%9r$uAO-rK8c=9#0>!Y})H3qf0CRQ!XdAJs7|?`39Jx_#&l`Y?08vTC3#jFjb~MEd
zwL*{s_rX=r2jq@)`Amd)B%ANj><{(2(+(DK=C~B>pWpm*sR>l5agE5S%B3Qni|}<a
zF!mt{ytljhM4Vkh(Ko0(<*;0_k#rA1W$~?!V_&`J)bA_hi8@Udoq9>7(%<~+q4XU8
z41pP~!EsJ)51d(WQyF9aa-JweR>m7&G2oiN07gIe|A%Y-PqU*-J@T-KQUbTojk=y&
zrlfs*8N_b~$uZWR(0VoV!tBErK*mupak6>y=1$uy(9rq3^_C$N>U$|68p=tiu}P>B
zV34L_9%?|192Qv#uAEU}O#Nu&6>FFO0BbyPCBEb)oO*>oq27esD>w12ftiGJ7P{X=
z2zw}2=yhjXbG1Xath-sAdudR;>*c3=fLtV-*nDv&&wE|z<XxFOF(@CBp0X5;Gm8D3
z=O$Qt&roZ$hzB?SSg3InGws>f(3%mOL*8edKU5~03Y<E%g`fnoQ3H#SmU!JeBvz*l
zeAaQRR^N(G6@$YDt_QihfSol!gvkKSMqMxT-b+H6?#O<XfO<VT0hX(eXbMsV4b{m+
zK_c7zv50tD>v#m;myb~>sT>EXaZiDG#z5gf#$^0P$@^C^QNRysXuADl2Z-C$&$6X6
z|2$it5B}chi2s(VZAnhDe%<6KO*i#FFi1_pSIUyVICfF}yw%_K(Esy1`iGhR??Z)m
zJW!@^BAX|ct`J_5aY^8`)&WzXpT{8l%?FyAIC&*PTYc(5N5tEsCiQ&7fU=0@wHJ}{
z$agF)5Ou|h0b|;vt>uob#w+;^E_9OKuW=@%6<618D0np_!o?T64^==`XbP!TsSrI1
zimXMt)_Y@VH`GSo)N~G-*e{%NY)Z{I=>$Hw3!dnBWFlV8P-=(FE3yjO%r(w}ShCFs
z*b}XbfQlfnENX8yz`|v3-Sucny4Jx-cdw}coHIV50FLoq{lH0Xl^3#e`SGZ@3k9@(
zs^`h(E`NV*c=C@HhL%TJV70!0EcguW2)t5DtdG-8=eHSAgL_-&K_-)4a=&(HoUES?
z$qtM)dEE&#{1B@y_REA)9JNF>R1Tqj$rwiRn(yt_9JOG6$=0wz_>UX$V;%X~JwXVf
zrxRP~m&!Lz0gm>Gf<PXAi;%ym7yl_+{5x?M;0Ry05p*8g5GQN`d17a1c_Vo6a3>(x
zt^(68aKD3Rkb=>}VG8uI9PV>AM?(O0Y1E73OrblkcjlvZ-wE+Db~-&3V2siotj-KG
z&=l8FMMSvmdFk=$do=LyYHVxVof2}IwEo~ml_UmS<7o)5H<s0)0R&^bX}EaxIL?^n
z9gT$JE1WF%M(kE^#|ryFv*8w|5IPdNvL>cF_9}Byun027WwUiNs<c++vuw4)g0zm8
zVZT)QBC@^o6Y!xgm>w{T)!(=RNi9c+tj`UMdB!45S+4GHKD`4I>US1^5T-vuHk0fh
z+@%2mL(iV6Fa!ei14tUW`bSyj0jf|YI!T&XOd6*n1avq`YM)}(_KHwa<Q_ocWB}3o
zl=<?|FCQt;_r~<q8SVp|F9$qn^vBmw%vDh+lRS_E{<UXyVowV-pSyD06~h|o^|cI9
zl=FDYSdZ;()XjdJ?3=#|_Ny}~ZE5Et;03Y#w@TU%7|hL84v59<L4(4~dTq-NRTA++
z@OsrL_zpzB5;VCWgcluhFVfQm8TBY9M_!M(4nRM$-)wqKYnrOFku@mpubd>B`=2Lb
z!|Ny9Z<cRLQzr?mJ*VdrI2RXD+8h2oy`QhQkW;XjKA;CltkX`<6!(fX8Mh__5|T2b
zA;27yvAUKNUV^#FVLDsv4d;X4^q*^ZL7^TZEBc7*6>fRnl{}D(dj)Pxv`W>?X!1`S
z&IOC~{em%W^+@KfIDumxx68$i(z;+K%Br}$vK;CEI9l%*c@=;umAcDnY(KkgADIr7
zjXHPpIV1WXn<R~3F{1wZu>rTBQ2*mhI9x`!AGCrypLoDUc`%?eg@+V(Xc1CH3%%vC
zZYc^N2}o68ZW5-nxpS=)L>XlOEY|!K10#C>{0rGzMoX1nys-)}S+LE0b>|s4{_(yg
zTJTRm|M}hD5}e`b#hNtx=+tFkpm_&O-Lx(>>qd8A56zjoc)=hY$kBpe#$*`sUm^*=
zVTEC|NdO3UF>Vxue0Pt8cNDZ=r^d6*lcOm;zk@!tpWKTKpr|!rnuJ71HP1Yd0Wbgk
zJw{cl8%s~avsF`FJH9Gv_k|cL-;a&6@wYkWmBF>6L^+zbphsC&&4SSAd2to$o~k+g
zs5XkfK|#CBa9Y1JCjX%_tS^6T-dul9kQh5@s8zO^7OqAUb0gn3Gh)KNBoj#G!&V-0
z;cW@`(|Y2-ZBBP-dwm%dTNrMVZ1d_1_>F42fsa#1A`Ai)3U!xSgtS^f2?r)rN3?H@
zXY&-aoY@3}jjp5fH{Q*TU4>{>uZqnZcBACYNKa2w>%x$wtHH4j`0wakiD0KpHJUHf
zUx2ul0fYGml_ML2&Kx(#m%l)PU01l3wK;bkKbC}`;JKuH9<k?bLoq;lI^Vh^@GsKH
z^9P@PuZh4em|EArVUAT0^$(AY8z|<{_$<NmP67|3@s3NR^^Mg(vd+KPae%4Lzo9FD
ztW(Q!m9!m{Zj(q4M4>KHdd`C?2WVRrVdjLk0+Xjla2Er(RO6dFfe?Zy8*lT?s_^w7
z1tF!Bs$^A<kp^?nsUd2QwTiAQD!u8?<}BR|IyV+JX`Gc)&e7#~D$y7Skl41UcC5kg
zK7BK9dPNxVa6?QT1|-r(aj{bxo}nAld9(ap>9nT~u(KsO5iZO`?u-ymgLiJ|u*rM$
zV6TH=cvhOUw4Q8PN00JXS8x*YUoF<b)qxY`!rE;CDV&Z?=0PZciJWui(+A#BH%(Y9
zx0R+i(1n<1(!3^JWK(YWuL38b7*Oa`Cgb{@xx8zh6McwVK7M(nY_R2rYPh6g9~JUz
z=+~=6KBtNsodd8;Gsd5?t_jVx&jEAi^VoagsfdmqluEn`0ptdpri+#s7dA>U_)AXq
zrCh-?If+vd;kTr)ko<dt<4$-ZeK=;&!;h*~N4riMnNz-abAIl}zZ^MWn^XVaD2s2#
z0N3WrrLgx;F5%G#$cnxJu*ld$;K+g9W{oX;#|$K&)}4)t#_Vee+1Ie&@YL~Zo$uZY
zG5X$Xqo*7nGs(SPQHg%cZ-Slo{L~PbL0)DOElHtl?u4FNaU-jjP|o=4w$M0b|K@?K
zPZNQ2nii+RQ-XEn5roQOLBu<AuGq;u2*(cKRO5V1P~~*#M^LpJn%DWITb^!VCJt<k
zF3Qw`7)dc4KE}>SB(@~UU<8PRxV=zuFZadWFSGqzeM+)Ii+bu6%KEnM$KdA0;W0kO
zwtybAn@xkm;TB-Rc2J&*u=hJOS_Qk^8i0JJiabw<GIbJ!T-8KjFubp>piq8SfE6z%
z;9eA!`FS2SlAvedgf6e=h`>LpbMkGht(46-*R|Ni@mffKg$|MR0MMr}r7yLmKr3ZC
zKEUH7%&CPV;&CLnV(zEGl`F^TA}I*9^(??ZK6~l>UM)}VQ}@n;sX&;Xtb3}$jXrKn
z*f--Tp?9TgbX$99u)m!dI@b@J{sYk>{x~kmd+KEz`1F-Po~ILNau~sY1E1O^BzT~{
z<m7#73FqY3^`*Dj2loSzDc<a=xJU&+A;tal3iSp{k^ww}7onwgt~bkccA&ILVnI+8
zg|W6H(zjNHYf#U(_x+Mi6%X&2(litodZ+mCgE$ba%`?&mKruE52mrcmhJmJJ#?BBZ
zQA7`-Wt;W4FaXuk(^&74PSH{Hf?1#DTzph+y6L2a3XKcrajc#eEM%Dsa3b$AN`cjr
z4zQ6;e{vZ!-*({V1Fdgio9g^#K5o(Te(IIzZT>(I4HiSy)L3{OZ0?nULMe4=%@<(Y
z#!T!VgEY05<AJ35l_3xX3%CkKaMWhg$dRIj5)pjOb|R>hz-QUYh5#?mLJ9;p{O|Y<
z_znuqq-&wU!hzsrLkv)i&)d&oG_F=scY*#XBJfRd#L+<|1pDdUU9{~?Mtm$8?;TG=
z{tImO+*RiHI-O30p8NzoiZ|{;%Tka%pG1`Qv#1Wx{QMdJ_lfoYqACf1#z2d&mtG>7
zxv-|%9^lP0MGyzTJ>QA?%j~l#(7^RMx|nB(i}9UUH})`FVT;M`?n!naSEzRwyRl`u
zd5Dfvn&<DeBgVo8STbynH@x3jI@M-_x23&(x`D0Gq#cN*sUYmD?S^A?6uj@K_Y>PV
z%A6Y8d+Vvvb7UbAw=d@hB(<xa@TnaE<*uOf{03h!D{kGvP6MQP#wPN+9+W(Yehg*$
zrL*t?kEQ==Z-tP@xx1ola!&2djsW0*Mhz-Y2xZ-xXc-mm6?5`_3D#w{s}ExxFsna#
z(Q2uG)89TG0a>($!$WNV)=n_gT%94FN+p|nTxUSXg%U2xVT&I}6dU**ED`t_ZD*J>
z2o%Qir<sq62hWNH#(gwu^7$9i>D$aeHvm~-vR0;FmZE_*2Lg-7+tU_zZvQb!dtUvm
z9|#JngRpX#I$9)=g>VTnc`O}{*4S@8?EGf2$<%XNzx`jY*$e4^gDu$8>|7?l&P20c
zB!SkW_<4cO(_hM%l*IxnkNcyEY#^Q?<sYW9*F-?*^r@GX*HggE<t7R{4&tE=A%4v1
zMLZ?Cn8a)*s}(xzaL74wN4xA(HJ#NkB0Z@{pM`Sz$)KU!#Y$JrxzWzX+uA}(#x|^K
z$zQM4^QG0YwZzA0EbWbqAg+7)ER7)Sb8lp$8m5mZ!0j4%nOl;S0gw7R%@vYCk4xsX
z=HRUQj?^U{H7ED50Y{+Q3&U`F9K+6gC74jB74&KK>K!QbrER1CQWeODz*%ck5l30+
z+9e(8vldIzvd@vsnZlsm`RHnphMStGHz9Pq7Q#mwTzNIm{l3QBr*pQaoVinTlskM~
zAzTJPSUs}KbFZ};gkK00HrMW?KwW6eYozWr?|!-ZQO1{kk=9#*N$@6g&w2L9=F_h~
zYx9>=$f<?uk^}}!Al87}O=|bz``{PMEtZdsa}9<5>dJEI+L3$v=J1bI2kd*?Kq&n;
z@D=l)?T+{F_t5FhTQ(Aq-axEW1B4AugIH96?`{G@Nd<Mmq79cGHU0vyY3VI<fJib5
z(VPI?J*4QuOGRqsYwl*+qc+Qr&O25{JVMIxb5Hsr)3dKNbx3`(EBWAb)SS{!B~{*i
zW#_Jd;??wm%4YCcX$)D|Z#08znSIN*=xx6}(2r0k86WimOj9GRp?Xw(+_u&`_FWII
zfrnIp+?86Mw|17+=AeV?8IM`lksHW&!D6xVZO_1FH{NyL>v66NboOg=bLN^?bM0|6
zaIO2zXux0xjWC$d#`{ua$+n*@)OeCsZ08!IjQ^g4GF+7RxZF(`f)EU~CGVpHI5Gfp
ze?~}2#e9>)OREjmEeK4b&jK9r9Pk@k4Pi5+^~Bb*k`DN+@~mK1WTfqck6bF|$@phi
zTZ{dDQ}7A@DMx?q<g4MEH_Q~BUz0U@!ehfTJ9gw=ym9#ZF|=_=wD9x)Po&@{H?PFE
zA_!<-_IJvE&k8_|=GqLx(BUZb1shXs1I)B9K0N0gfVaGlQ^gZ%pHZB;^Ilx@IHB_|
zgGX`&iEcvBuDFawL9WMt84T7?C1s>Hren(vW5qxoUNUYh-z49dIs(wIx{4rS+Ywwc
z92=b#3ymuI%kxcMIjSY3GYd!bIi8Ajr)Sb7v(X=!F?w3D)AFS>uADRk=#BJW>1foE
zTfR43pUWVBRH3qrqPO8bbGd?ag(?5&Iv)n}jngqTPD%>~<DSV`0+XM9pOT|^H`e-$
z$BxrgP3i}acHTqlL{rOw1gB$Ci8SrE_XE?pW^YkaT(}!2giuA15|kW%C;L7Z=R*w~
z)%f3eJ*k0Ge2V)UTqAC+Sqomj+bZLW#aQj!bnm(j^tca1z;y8Y5(ml=PHQJSe~=}v
zGZjW3Qkmh(>(G_fNArfh&e;Fpj!N^X`&2R-Tsjl@o8ZR->GwY!YIk5T6k)!)`xbKo
zXRJOTfd1SEu267=3loo{^bFYzNVubB3E3z{m1;JS1BsfSN1z{`R%P@IK-|O)yv@jS
z(lJPe;ob$lk-Vz4&tAL)E~We~=`li0%k_MFV_c6g_T@b_9g@i+Hx!KNBj`mMn6>)d
zU|ZH0x4<{`u@+aC*hD$q+UgQ*mFt>;l0&6+i`DC(0`U3)iTo~h(G-3g>oqRhjaqhV
zU+9Y{=t1rK9(nPGltk92iI^07eqX5L#5E4SKz_GM;&iGP2iO;1^HY;lV`8@<DA8NU
z>z@IOFdvJ3MU*ZGdD;e?3HCqa5<v|@k5o4b+%QgJ30(RrLRI4tTATB~@aF}%tE33P
zO9&-JaF0~{Lz(3|JMSt>J?%O>uo|RrhIQ6Is{@oqAST%26~JKPe+Jv<7ykZWtWGD}
zEqh<xcHasx#wJW({IEu-P7Uquxj4R2*H2y;;AMA>5lElxHCijJ9*<7}nGe$rYQYMH
zlx73}9d|TPbbfTz;U@=K7qvM8lf=Gj8X4NjQ|382`fDTo;Wa&&6SucevrNayP=B51
zM)r%y5IIH4+!?KP!hsvu7Uwek!;#F#md=vj5*?@Bp&AzHZ8bRFp=|7xVWT0l>(Y!?
z_8Z+}Wvs$4!#Fo@pIGW49^wqNVmGpAlEU~0Y%*N;>KIf~y}A~)>64O?(n^gxV1rW$
zxDx4MPILPM7?~G}BveN+2H#JmC_a!X@D|mTdK<3-{u5;^K6`G8uyc({(q=LBbTrT2
z-RtET!UN~h0lIVHwC8{^X;uXm`zo#nbv#w*xbRgMUm`u9(}&TmJ^@$^`qqam`M@nf
zc<h@Fl<At7`H!QCT2Pb*{b2=%ubVy}Zib7>ho9WVbu(}ZURg_yCRPEhP1~@y1>FqI
z{Ob(hj&ROgY39k!8E?3?2j#;-srd8CJUiukfB?@fPB||OIo02u3OAtN{W0-I|L$sv
zmSCqzzL%#6_h$`RP{|X8`+qLFMp#UnJiEN|c%k&P)(5-u8i23<i|X?<6*?<0A9fbn
zJ^<A|>nSZq(+4svJtzfUfA(LhO7w|?`gc>zjbL%XD(GJi<f#xK-u@=h6!e6OPaH)W
zO?9C=2ij*$wx*?@8%%rN(KZ+^RU1cLFI{)M`P#t!5^w0Y*FDBj1+&8WUop+mwLfs5
zlUA}L28MO&W}HBA*bEOnU{6!cbmc8Sf_EH`D6Q^`B_6-QNS7F=3(sNSA2-ml-y+8B
zzNpb+;D70c_1gu2k@Lwx5u5({aG^=8XtBnRk2>Og{E)b~;)rS@ikvDsY5R%;_ihXZ
zj`V_y#@7yy^2Si_jPTQRjotM?F>f?KZqb4DEdaxPK|dO4rL(p#3ibNt1~OO22UN2+
zpvEl$_E2cwSAHmz0YR*gurs+rY@<Q~=0d8#+S^9<D_B%SPxGhwsc}J=n_^9IDGQuX
zYM^^dq8%So;VOwI9ktM0oauWT!bA8-3R-{Q^t?<8ca|{yN3?r+X7yQ!`KE*wZxG|=
zvloR6@!MCzjW0iT{SVvWAG2sH7QSA5a|-Dbg<QSt&JFwv>_7&R%>;$d3iT&8NOK{O
zJzn5P_y4|lQ+hrLE~xS8gPlUZ!S#+eZRV|@Tbj|1BC;gB(56J8{d4|T(0&$O5ISvi
zZIpZRYf!Gb7Aaaswb&~$6i>ir<m>*6$L`C3Q&8zNA(85{;HJ;stgSsdu?adsDtHPR
zt@1aTmA5czgYP9@FXUW$`bdHgpyQI87#DWbYHY2ndP_4uwNYKepfmtWPEzm2&nwG6
zM1-ic40OxF0rza$FiW?QK_RmY3>;dxLHDQcus_VPxI^pe*iH_!21sN1hmx|OI-7SU
z(I`7vi6lhSF3^O&ddW1SP$;)0taVmlVBoGGgXGPdRD#5UM1EV=cqAVTwShg3MY$1q
zzES=ywdJig3UDw91`4uR;C&DI60v~adY4&?9TxMkP>tAz5rqN|t$%uWK|ONUnP2Yg
zf=AuD2d;QsSq?PQEi+xe_AGwbcN#rPd1EF?(@M)A@b^Orz=H)&^Y)J^Yt&b4f1h4p
zZ!DOfKF!9Zu)~}0eNOI1wN9Cyr5ZL9u-5m%EdX|BJGhcA8IkoKFe<>E-0JZjiY$}2
zSq_-#rb+N3uL<lOm1w@w`^Z9%0x&UJb%aMno<MaRn<ZJ@F?Cm7HnR=gFc*E;Fmurm
z|IiiPL<g9;W4(b4(<`^j35Zvln|7V<(D9v-$=<@ulG2_R0|LAY<{+<?;a2f9*$d0*
ztuM`*q1AkwUgQvbSkL5<=?L9imutLZn6K=^iv;C7j9njBM*JYL^a&pEk#~VD2=&0+
z8>u5bvRKejNH)LL3(VR-49gJw5S8Q{d;6qLwuhHy`};xlU@F0nI_dN5r9Y>Mg_6GB
z>qZ@SUcg$H`U(C776xyjRc^?RW8M{mM&uoU(hN$;yE>S=RLqnm#Rm*R_^r)8d*J<K
z0IIaizY$}Ff9WTI{gDzxxWy<lo=-iK3||<GLP@vvbo*hg0qHRXL1u*g<NS^0esFf%
zkUAAxCr+JozAX_7ivIG(B2Yb}GJT!cg%I^m6d#DN|6-Z_A7Nl12?#&<{&x=8d*)$w
z?fwsEwbdndYT!^M16zRVX-zV4)qR_(qP5TjLkl)3j{2P4IwDrfJ@GBM;$V0}K~l?&
zJbjPrcW><TeiaLb^QA9xF2~yl<!<jJuE3-9w96to-N7xotW<5R=mt`MOxf}KPtI5w
zeVO`8_FyZ84`Pf@475kf-WVY}Sg2>BkZxX26_;b;xLy_YtL_t4`=f7SR@>0&jbM0T
ze&;q13%}?pVC&x^o>ZLE$IKh0WTxaUSaDGrf*5hIwV+29E*vx(Bw_!$1&g)frcYKO
zveDevToloT{X77cI+rq3&Kmz`P=?^>40gv2LQ|8?`lK{^*Bu7<`d^4^H9Wfv_Oce>
zHecU@G+mLyTo+^x`{K+8DYlf9C!Gu7nqyZ4sTSZ0?lZ$;Xd7?%a$;256uwZPr2?e@
zGU-aIp*F)lBT)S|N0N48oNGZbQ|l`#6%JU8ReU1AlB66X6=32bf0RF~w6pvCn*Y1;
zY9(P!ZKvD=t6aFGkdmJGZy^Z}+hb)F@qb+7!+jexfRN;40}yO4cj6J{tG^)^vxzgq
zuE`6XUKP7R&E)WSM3%|sNaBGydeDo5cX*|Zrmw2#UA5^A;G=(((N`cuI%ZHGFH!u%
za?i{3exHuv^*8#euzve_*O_t$MvJi`Na~2U##c$FY!s^EJMhD{WsNSThWDZ`ecP@8
z($WlWZ5idqul#dDWV8OYccqa0by}qT270{=D<ztB%2%@fWHRzqf&KP`5W@f9i)x1=
z&XnVk9l`2=Y_hCakbe2=@B7;m(m?mPSpX2~IQ@k5g)dJN5yg9aaJX!MyIJLK(SQuc
z<>H~L{mgV>$bl^OW@lX<0)-ke%`UenJ&Es!V=-kvw4ugxq^@+JcfuRGcfKtmn3e+q
z%2;k;F*1RI-KA7gfqvE2PZC7~FR;|&)x?xn0?4iWI&Z5vGwh;ZoGR8@MJ}0v{bYL@
zf}M8!<HNf~b~XZ>iTL9xmb)YMfGWm~%(&8a%}-M?wx!HFe4_dKf1u6%nVbKn&2=U!
z^-09-e<i|hfnwHIalrBHt}xz%YSpz6Ve*7u$Yg1J%S?n7r%z(Lbd=i<`~Har<;JIU
z4?Au|jMdQ=!Y-I@3TCAS+Y7cd#+WNWW6QVLlm^oum0($jY)Z8_UMj}jqYYn(z}D&&
z)^hTux($TPfmVZZ6HTts?b1hxKA&q@zSHYaQym%Omsi64uCTRaIy-94<a)!e88TQd
zscvDsx90ts5)b@FFpB=HmJldCeC=}`xF0^!Mfg@5ab)t)<3>VNfUI|IG0xh|4%!b=
z_Wrc%H(u2g&H1GvrPFo6_~amIO-BWEGGTlV!Eq=pHwOaT@A@l27jj800ufXP7iIR_
z!i*+>Tx(;tZ(esqEARpZ5)OjLFX0y&1z6{~Rv}VgSPk{eSvc3a9rQ)fh5apwCzq=)
zd#q*;p|AU*x%N+t^;qu&KjiOX3IM_<`n^sGV4VOt=Bry>bW6-Ikxfu)ti68H%IsoO
z@f*{V%FjkZryz?2+GuV-mpkjv{exJ#`nyzQ2&__!_w*S^{eN&Qq^2^ii<i0VVx<_9
z*xoO9$0?{&Rsz?goO^%pCuV+OA2s7&@=Wt8>$aI%`YFmh+WfSG#l)$>W&FHa`kSG2
z1dk?o66H0zYC$8DktU*FqK{9Ga?LKdl7|XsWmmuHnCwmp<?(uY(%fhi062zv6tK}7
zYH!lrIMx>(lbRwzXO`@`#-)?onGxPw_SQ!v_l)zHk`W_T?wV!gndhxJ?KJT+?*ilT
zL+KL2D)IF8jGHe$P2k|Je9f;F3khG;^y8?z-so$Ma*dnv15f3<-gn57(zq122BF89
z@;rk5Nr)iTW__XsPABCOP&~M_MT5mezL8p++1%NZ)deumNieQwpU$cTOtj0CT;Qq!
z(_s(>t=-7;NMN37g<bopgk=F`b$Io}Z>cRrycY7U0kK)5i=|t0p}gty9o7MT%jb*;
zt!aU76eVa&y94_pHW{&;DYcE)U%t8ai>-$2ki_}IW)UiGu*t|M1$`IsVm7`OH2Qc7
zoE8|MPy%iL`*~(t3KxBQCv8KM(_a$0SD8|~;*}933k=p`9>Nee&ki6m8FKHTLe<f0
zu2d;(e4Zia6GbnBYEN~AdJo|NI>quCpjT|5D<f|dlYM%ep+nbaCC_pzDnBMoE=p+_
zot|&$I4;pX0!}N8-ORxL*hbnt8ECfJpjd<iV^X!3mOe9V%pk!BREChHne-%EpJc`?
zv;ZzyHam{KMl5F=NVRUzdhtc740WW)ni!4yvuu+e{cwN!rP`8qq<V)u-*GJPx-Pyx
zU6P5~Pi~;%lgmyX%sX5WP=0;iVJoNs$7RL?$!Q2{M$xC@IBHxW>hpx@2P42?%1ZWo
z^3`+ZGJ(A8bZtDMr>+N-+)RSo$Wc+a1eVNGjp-4~Ac3^53y%w@OE6>7&Z)$~_l4lg
z-M54lv~Ut(NDJN<d{1il7z%Yz*4ym*<pUU6?UH|j?5?T>zW_+-Sd2C8ziI?Q0ZjCn
zG`Q2&79dBbgk|UGN_6$rD$YY@b&Viqjs6^5YLx$B7XR1$O3YTB)M*kd+z!4TbIKpN
zcf4JT*?HFvKP4xqfeXhQ_-VG{w0rgR(s7H8{4$9&s1GayTNSU?A*(h5Ax2gc@S3!B
z>5#-rxA;p`!oA`33ZCP|1&LQb5I$Gu;W`O*z`&Or=+hxi@mhaiYfN)p+DSJ$PLDF~
z95Z;2Wj^*vFl_9@rI?0!Jzg<pde7FvX}<$i<vI_@tT;oh?I1LIX07Yq$2D+g+}vj&
zy}eqY2kR^5ZBgE`W-~+t&EAInR9*(Qq~f0j`v#VTdYx)kBk-jKgLUF_YIWkacy~u^
z&hG#U5NLfuqm1VO2zTRf$UpzrO&$pDM*+qf17KfO>y5Kr^WstN^#(B5hH*2}G{Bz?
zS|7l3YxtOjoxLHlgNqs5_aTv~2)<L_GMFa@0ZW`SnS+88-ni~tdduOaU}Ph4g+24d
zn2)Tj)Om80fDNAMdv3r34yPFSAJdQu|KDOep`8Cg1aOo?eLtmZbgtQ}D?J|)YWA;l
zT>O7I!j$_mRE}JN*8zw5>!2YIgiszrb5TDD)J%Yd0@RqQ8rq~uFt&Qq9CxC{Np@MC
zt!jryg~l%ZbzOnsp0ZbYK;Eptp+!E;+x{Fndqh7meD6kvo)B-;x^v-2X)!WM26qkW
zn>~if&KiaK_6QtJZ6VngY~$-4Zp?L{B3^#_=pIPje#?j}Sj;4EdIg?=BuUm<&Y%<b
zjyG&}Z0L%|r?G36Z`!~21O>m0|8yfv#C0FoTy|(&WIIFJi}h(>O4HJ{f0>(qHA?Wa
z&vk0@uIwpW5MxS3AnIfEjKIbyjdQu08i3#WcxW9c@!<<zfu78cFYUoy#Ne_$3#hN&
zAje{I`X{?lCMUZyTX6Wqu^rU-`9mVr2W%jgH2)2FgtTSGqfp0nPdR;di%>Z_!0SRm
zBqc8{Z=pQi(-V6SRK#z`d^8Wg3N}%89soWD*vI<Na^!a~AAw+B-v4v^_&Xu{vpsIs
z9}d>ctB71aJ_byNO_H&u2IrhO`FpyPEA&oxMUVm@{Erzj@aU{%J1we9EIe1s^0PNh
zstnIBf8mswe<h3@Eku)aZ*F`~9o?J@2NUAlHbOsTER{1=(n|3ueZ<Q)ag9B7@kix7
zuaUX*)<RPBFhLb3jV*LcX>Rv60(Xhx{hlqoScRmQIXT*9%UaXZAmpp*m%dTI62(ns
zsPSq+M_=Bo!-0jFj4^7720R1^ZddqUw=@{DZuEwy2Q=TtPE1vLSGuL~M(W(ZW~rrr
zBlFFv%i`nXYet40lbkgg6G?7En5hP23GQec^h`*Y!B`h_da4^fnM4kt{F207_h<MH
z9>L+^`3|mw{wA<YD#wkYnZXV)KC!z8TF|OnZo|&uso%(JE%$|R6S5OihFl##r@#a%
z4+uy|w}Ie0PC!bteh!oR^ANyH$Ukndw6R2(5q`&K4w0miCXnu7=LxfVgPgK_gLHuZ
zqi&F#LBFG=P9YAYbBDbbvlb?!&DSLS63ryW>kT5(j3*6${h5w*WAUtV{@ZvYzXvH1
z<#~GN@CWRyBO$l4^XmpcrE+5vW_r(<1S$?L60t6IU?33BpT#^re0%j5Dk~WgS;)DB
zU=>x1(E9Rj<h>MQe>*Wuq&x-QP+oufu3-R()$Zso;Pf3J>uQ}k#5T%2y$78m(aCjq
z1FHMfDnpRv&vb5l9rrb#2y)qCkGV$>J-d;-QXGVY&n}rGJa5VUkP$P$N>YbFFxD+$
z8}mggH*R?gSub154qJ5s=0GCAaLV)B7=TGKS@|`a4)cmT6E+psgLa0NQ9`s4)zTzE
z^`9W1ZA-f|8mE_vtg2>CKrGbj1K;3VvxF2Y1+Nq&lGsxV%EYTYY4mo)dIInUY7O1)
zZfm^k2}epti*F8Mti?{ym-#hr@I%Jhr>nbBZWE`apIs3|Hj45PE_U!7a|uaB0Kd#>
z>ZUsfX!Wg>@iKuIz^4Uzj_G`tVuckQs6cj7xZOOEFz}Tmt2{D9Qml^x1-{cdI2>?>
z{+P_VI?fak;?oy(@EcQKl7hy0+dwemaU?#*9fjtu|FJX6CqAn%atK?r6sY!3sSWGt
zdF1$ooKp3qtE1^B`mC0&{CfOq%au&<s7hUX|LLhU6_`&RaLguYbP4kF>i%ji<AnR<
zfPlILfE0aeiN6TT=R9W%e|UHYPD@{Zo_R+t%ib?Y@36OcIc34V|0=%Trtgif;$8-N
zpY$<vL+nBmk3=dLM~XfaMi)oN>4=wgmBu7@4i}Mk9oe}iTI~ow(9JQV_E%Z2T1L^Q
z=>vaE)N`jhf8FL2S=5hn%W-qs!ZZi}QVqqKl9xtxjGO4ew`;5fXFX@`j4sc>;e*0#
zZv!$HGfLEMPu-@C8RG)RKAe>vup&n=vq`+sw;vS{O!fNqs>vMI=jIKNL9qc~_H}z~
zW_8F67|S>)Ib#C}=a{{28$#2}wybxTJAunw-UAP~WbnVYujRk@r=F)6&_Dx!**XKD
za$F66_D-7A&y=GtNt<uH<&r+^+#J{p3)a;B-0&|Icb=^db|5z?eu4b7RW<Be^ER`p
zTosb<KTY!g_jJXc8tP2r&J7p+OM(%I!OZKLUcGe{Nbuz9VrL~08Q)vbm74N_e_V^W
zz)B|FpdDfCYq};9)&?vFaxL=OKSO&qABrZKY)M33+3CeTn0=^z-=KYx^QoD&=Hlz=
z44TIjBI4<(H8tqHN3aQ#;#_CrUTAV<hIeD+`I2c};-&A-Yv1|1yrRF~O#OJvXQ)MM
z6qvwY7y`RcA(3<~n6cck&n@Fwe@(-S(E1?+R;=steDaWkezQT|Z52b+_a8m6$W5;T
zP|5>&_r3{CP)D+o{r5ipwc{X%5W9OxX{|OH;YS%Mc}Narx}Rx~?k3v_QZ}~^Rxp@r
z?)tlB)guclI}E_I85l@65EK%=zz(U6-L!-c>x(il<1^z!y`fA8KdZszz3W2Yh1Ls!
zrP}5pzSQ%O;46|2?3K=eFY!o%yPEm0yE=t$=U|y$GP7_ku>%ta*&m|AB{sIRXBfi=
zaA3`!CTp%9HQvp&I*#PVzJ4DYYV+g$IQ3oQ|7%rJ#s8>E0)8{+KKm{f`lR<`z+$F2
zf>Avy;~~#f>;Mil`GF_NNy@&558kSpj9A+!k7{Nzx{ZP@#wCNKJmUT1S)RwQ#G<fz
zehy}dtXJe1t*gV7#TL1=_%a_Yq`DQU_K;e|_H^$zMYdmYG!PuM-oivj6+=i0pFUS*
zo@L9JR4L8;&Izfai_{>iF%a_#FRazDp8nQQo-UpE%GppoClurXK2&@!AsusV05)W~
zr>>XuF)X0?owW1gVORSy$4niPc`=@(t{#bmTrykqn{NcOq#Jpe&YFrB{QyhFIvTu^
z%kF9}`UwF1qkvm|8uC2DQncYKf3U{iR$~Ug71jMrWz2zBw&KqtpClvNt_Lw(fZ@UD
zp!ky7R2U3lU=x)T-o5yH9OifqA>U)a;7tV|D$Zlu9xMn5Zpi<VZ=>eKk#qqR<l5(`
z?y`G>zl1qRi!=Y&WY+Y53yHyJGCg;)mA^138Mgib2L06kW5$xm9Nl3lbxLI4s%n%m
zHNS2D(ZEH1UruP*_30>&Tl3BTHDW3KAu~;ZmKSCqQGk}^P8n)B1r=y3)Ng^CJqWdh
z0jFR?HqtQS!j9g7#JhejD^1vsFxB)Fzk{_T@%oZ`i#a%ncD~hAoUt7D!}vIjLP`49
z)pv%sE_c3B`#6EzK)=^@lfSW3=z61~@sseXFfY-xL%_ON2GAAbjAuWJJiZx`Q9kUj
zWPIWyB6Pjr8Q^qSH|P<S7-+^4VI@;u33PyL5hz#vZp!}?xT|P$3=?2ts9K(pZT<AC
zQ0UKM+)Fg-Rt9z9P1I@)7_pRQ46#!eRLKkvtn=^dbv)4I7s5$ndh(pbqtKV_`YnVZ
z1y@G_yXXX%&`cZx`Z3nZuyeo)<3bm*j1oj$Cl&-9J!H%rUuchs?+Z^S%-s$zWlU_!
zz6N}haKID9EVg|mZe_(#wo;qJW%2z+tT_*j)MxPxpnsIY1%b2r&FE0-2y}fsb5Cc*
z3sCeI37VyEoB#R&;)Yji3@N}57yX03rHwz!GzW+7>9aWXO(^vh{|Rm+piqH~;YY%w
z!JQn`Nm=`;n_aeJ`kIvb$)Mr-WzwPiQogOAMD=C(L-c6qr22^w0Sj$X@}-z0{h)<w
z^8lv?FHK){$&AfG)vb8~uOQ5rgK_-GRC^S=bR$)L@>==s9=K90xT_bNM>X)&g47u<
zdp0hBGs!LE=PGw2&*T*lgMt?11A-mqUdsFeh`&c*kcQPP3D0j)${SDD^6U>ms{=Gm
zNAez6TjlA;xhq^}L;`?>;ObxB?J*cxbLDq{HB`OJ@vzC#849`Y2QKGs(w*T*Q`*>(
zLzT(gNA^IrkoEqh_j#3a{wZ?4m{zy_?!e9UJD~Oj^js0E$SVSMkk}`?IQ{&0)t{kX
zJjU)R1&emGFxSl2Ir`><a%sVa<sLlPuUr6wiUQIU7;r_A{Xq&|dVcl}*2tqxW;qP=
zw%xtB2UmkKCM?*Yy|}U8M*{o`{rVqWqnj-L*HHQ22PZ0#h$!#r7_mrXd29m>l<7rA
z4><;t5OfFDw-3r!L1qx7+T$v+r$E`_d@+BR_{-0Stz~jogk`G^mG7nVu&@;l82gDs
zQ&o$4+dt(xBGg(MZp$DegrH-Vbow6jLG=a|Q0jsWX@U7lJ#dY_bj9aUxjvr|_A9)k
zK@a&(JvK){dMy1LKni-5@Vln(Jo)5#xO9x2nrB}H!6%j#z!d<a(D&aax%})^F!Q#2
z*ZV;k6M1uJbPqWA58|-XQT>r(%cd!`o0m+sutw?E7QVqnvrDKqmpcoSpjxn492KsR
zd<>wxy@HP<=^q#*yb61PMQWS1fct!&-XinFlmcvfXSs3>ENJ6w69xnQQN7doXmfY<
zv1D=%-$$)W_<Sybk_5K6Vn`P0X^Mj@97VFbK-rqo$^aO6l&!JBX4m*>r#?Km+HTHV
ztH><>b5P&xd1?}{o~@I~AB^YynV6iv_1g-KO$;E&=;^A6M;hw-hiBRdFDcuu{fG{&
z;tKls=07@YK?an}>VE_}meg?>7j}fEZfkGT$FVUVCImwOB9%hA){Gb!%^vF6q|FoC
zgdE3|XWwD>aH!@-A?BSm(Y^a(>PWrIvDPrn7c3E_2Iak-gZrOv=2iX`X?uhOa1PTL
zrc0xqMH6r)+s0B0ezQ%%h0(E!csH?xr;0><ap~E<;4XzdqU|WzQ5mwJeP!>~0H%_@
zXZAon-ac;aln)O;#=5w@4rJfsRf=_<gwGmk!VU}tZe7=V9=nLpd}|ta<F;?6;RW!}
z&9&g*eEJj=7=(#;u;vrSmaQZqO+Tq6yc;%6k{wS^{orN>x=qK3zxse3?Jxt4+OtsQ
z{u!2xc!keAcyuwGorf^Vf8k|nQU~=TmhlgUnzGcAfGd~T8gvR-Ys_xss8cW$sYf9E
zLapzsxTsjpdOiTvnXGfH?bPOXD9hP-vp~ULs+)W8pFClw;Tfjz5(RhBYi6pv*)Dag
zKWyjj>kyBRPbKnq9j=9QlQG{W{*Ss6`w<%>(3N)3^NyV?5r32=6v~-)jioCel&XEp
zpNEjMGXMJG5e&al#NER#@U+Y21iYZK0n5%EWU6mH{<d9E|3sRfTB}`OV40u5?MEZ9
z(h}hs(T<O#cy8XJ)`V<i4Bt-3w6}cHEUVKViMtd)bR9epcG7pRrbZshm@Ikr&G5Sl
zmv$p5&>##DC7mcndi$l2TEW4a+6cwl70;%adp>!a%U?U8B(T-b?QKV=yD%4R(G*g7
z0Y_RC7Srcns&6s6IEqV)_|&xdM!mzE)nsJOpOPM0UzZv+d6fQnqT|+cF=1RW8fEs(
zc%>-2Z|NC8gNN_!TQBkM0aa?$vIRs%u*I=irWO!zxeT#bd^KHyM&&*6-qIFJ!!t<~
zBrZ)&JfdiwA={Q0W^=}O08Xby(zij3NfNiN{8ad^z#Ouv($y5N;@TAN8*Oe&P;kmu
zga0D{p1=Cx_e*KxPtaqc;8kwpI$*st_dcF$5x#({_;N@JXe(WbT$H!k{?nl(|9kh-
zk3y#2{+FZ#&Lm|dJ1rrVH^*KM$Vh!C2EV*i(oYJ)0k>@{VbJqn67_Y88iTu%Wdl;U
zGy^Qd#HHNYJ8lCDF$Rh--IB!{hT2Z+TFkeJW_3Pj&JMa23&ur!RK6F^x%k>I(UcB!
zW7z_mlVfuzb4j0Cs#k#}7(AtnbpX8D3~t<WZ_?%otiwc)Ymu~6-HD^v+z9Z88Mi_&
zTY(3Yc_2%n|M^GyQB~W?$MwJn6*Hf-l3*UT0ZM^W9(G}7!4z%*3X}-B#cF_`X2-*6
zZ=)Qhbpi!(n<h|WwZiVbCXBTVczmqvJ(37(AtAzB`WYTztBWVkjHeQG^($!&p6`}5
z0IMamX121%3kJ$GhBp%L5wsbEGvW!;HR7EXy1>32NCIIZSd!d}NNOn!PV<XsZmQ|f
zvvr*4j}z^@73G<rUK05CIAnwA_)5j*@FT6mA7(r1Pbd=<faAy-EYF8TBFb06zr`cy
z7yi;Lo1b4rt4vP&xlaK^|64q&pROa|@K&f_B3cN8iEqQ<%GyprxElK|;Jx4j6~I_y
zB?4#tn|<$jPcDN8&9ME*0MUTbZI{yY4ZZvs@b|Qug-bx^jV~CPmZ^3udENAG?9@$8
zaue|`X?P-S3-E$ERR(58$=NDI*a7nl4L<sdB53sIL_rzaJ#VlCYb$S>L}e?XBV3E-
zu-9WmM}8;Kcgn-Ycf*cey!^6Mp0*i-GeCdL|11d<K`K8tGKo|T0G<R@RXc7>)E>p0
zZh$#!e9;W8N%TqSur$1Du%S!z8vxxfGhQk`X2!kfd?^gU9VM;-HmKfY;Zcv=j3(Ao
zGcjiLA+S&3*@@VB*)EwvAgOF-z_DiNR`yQQl5tN65OndLv&Df-E>{SZBmrd(T-4<`
z*U+yZa6$nfORGjV2<bXBtfazpjtMOEy>7J`FmK~gvV&W^<_b&^c{%JlF|Kp%d^YjG
z5L^FW|9ICv*5RE!s0<V#{8@kbW)}KgWeRwO`u;r6(OXfPHhH!CZ2kVjNaSCY&LtL$
zNT#)qC{${7qgnstc_67^CS#__5&?sqN?#sy0OMfV{%AmK^#FkfB5WCf8!eFyZ?I?)
zWkmo3MNxP0&rT7#CgV*OVQ_7jdR-XRb}L#^$;@FZ(Js@3pLcBCi9D?)v0pWF$<K_4
z)lupm@QE%mftD+00LV90?>h<DK6a9VGpmiAvv{SE;6|c5!l>sZ*YwJ^S78;oFil^z
z2&Tqw_*<&VMKLY#uwHypn!hf6dxA3KygE~mIy9};dLJ*@m}Lmzd|6~yz1MyTWjLI6
z<&>j^v#+b%_+I_h_oQ~)Y;{#;3FT!1zq~WEi7_qO!p!>$%2n5Ku&jKkl5moY(xxaa
z-Nd9Ina5DVOd{_&AdI+R4|Uh^r~ZrAvDQhPWe~_@J=T2L%fpZEfEUMT#Rw>)`p(IS
zl^aWknN`md(gNt9UVXby0rY_VNRAW4eV>sN<RY0B@4#el{VJjfwZwxE#|AR%za@Oj
z*g>I(cmd9US!FREzFW0LU(}cbNUuOO2C9f;Zc%Z~0VAVG6BLc|6ZYQQo(#CF%@abU
zNXijHC3;029YR&f>3w%E*yQHV015tECY9&Mipn*`yCQq&lI322BQP(BPT0EQ7g=v8
z1n5sY=TpZ9C+5gvwPEHT8k{BKQ`c^AvT;CK0=5qrypoWIFCy9MojgQu9=CbYdK=dR
zS3JS5C-|Aq=8O-<WHtBph9LYqV_`&CDjZ~OeC<gP+2B~@X>(^vA(yq5cG*y3j48P@
z(sRdNWTelT44oM+)!34_xU!uZ(~JEK)?=*Qth`L3*>jf4roei<Mq65pXM``Ki1<D`
zo2c=|g<v>;0r`_ym-Funp0_*fTu+{IICM4Ktx??d8Ag=uPWTevw;3zk#Q6R&iNEbC
z6M?dF9)`mk7|8d&aFOl|*=mfkZezL|Yc3jgyAxqWj_S=<SmBjpa&tW>Ww}Lf|K7+d
zV6-3Ezt&V?0IOMe%{5-zjnY42g2P2@6+diYa7&w*8l>rnM~73v>P|PxxJ4rwu~AFT
zOGG+!qN|L_^@3N<(dY@sbBOH`wcX`_;GJO<@$Qi|+@eE}a6mP5j>`s0i5kN#Aw^R1
zvE(xD3KsOFR3QZj9uboR)_2a=@wCaos+}#MqEl$1C`U+1^vTNDNyDLBz=_Yvv$oAD
z*xe*p*b89+zJi^l+M8zL;(k0AAF9!ixMqX4)!n4UR$UXh9-rquD~}w15vk;PyA79U
zSxs}|V)SR99^6VB>j5K|H3lrEP(AmE&DZ^K;!Qgj#$3mLl#5#<DT(kGHyIk=lUy`U
zxPWL878KL{E9{}LvB&p^CLc<z=Sn_*HeGGdw5_p%SJX8+cWQ>0nzQXym(jewm+L6r
zrIu6UUhGFV_TG{WVKt~frdziu7!0qDrO$#8eajd^+|4`^Bb!82M2VHDqkq&ZTf}BJ
zKHs#y;0PV@XI1)=wuH*u{GJbCM18cmjtI8DSJc-D6*H}`WdmN{TWo#CFcivot?<cw
zgL4(5X<CMP43%BJvyaS)^xol>c?YoqK2fQD>)cfM1)me|%N0L9CW9XZfmFo3g_ZrS
z33s{GR78{Q7;DU8GyOLN0;u5%=b(Gcdi{8P{F1j^2rzlAcZq#AgDZKr2NYm4`1_of
z5eR$-<%-bdr$p0EikF`e>Rk+ErKU~OwS(_V-8h(jwml1rJXt-#ortZ`oqly^E5~;B
z<tcV5+3*or7wRI0k4*2t)X-#NS+sP?Z0%sD9g^Qgj^bwxnR=djgZrJ+%u2~J?}va|
zy3u^3?}^w?c`J^3@ww(68Kg=#YDt$SNq5<Gi~_BN?>QtQvbU7jgL+kWNN0WNy;@?G
z>fuNC{jpFMzUt+YcPT3P!_-Olmfm?>5FoTbggLHP2tGzU&Jc?<?^*5SPuq>$JpT@a
zhC10)lHg8$tR5?c550IYFMa>~Nlt=j#`1y|SAXXiZYa?y>YQnzPg(Y1$|fdPDDUR5
zw}qiy+RP90^)2m}%0jZK2X)*sP-~CJx|U8&54VQhg=L@?8z$bBnwpsznbO_8$#-7H
zysy!+Wi$7#gkTt~_Py8yAs4SRqb|LGK;rCy=0BdTf<k%t`pu?}M0rIeAgV99T!hw-
zceU?U`Zsw@_s?T6p2peyq=BXcF8D5#162g&{GOrR#RL#~5bL6sMDt!=r3Y!^s0dWU
z4`Qq%<1d1d(l<Y;PMtb_NeEZ@B&1e&;yQM<#V<eb&fm*};n3FM35g%!jNHA|Gq(9&
zI4PoguljHWH5HbYdh709>@~GH_6vEQ98sZKkA$R8YXfn$pUOswv|N;)w@lqw7&dPC
z1g)CC2!YIsuOC;<)^@)FNHV>{Zub@bm)ik#;Q!=B@HJ1JO^aY0MQ+{{geQP3*fu6>
zxTyfof&EtG`nl=&ZfPOs(1DVRA_2%uKHj)$Uyd<Gs|}s%OD40mJ+;b~!V2T$Xl#>*
zrq-_GNUxkBjmJrS)xnb$PeoVXUhc<QlB@q8XKx-4b=&uiBTie3ENM|F%pjz)OtvV5
z%-Cg#Ldnhu*_V_pk{J|ZNcL?)ma(>3W`rWfK4OOK*~ZRu=)5}b>$&gi_j-QM^PgA#
z>h*OT$NTf%J~;*{#p}+DKeu6zkVufWm^Lr<Tk3yc^(#flvW-yLf()E%@iH#{QLgkx
zRIo-+pixE%79Q^f8>vpgT13{`?mWc#wNs3%Omd)4qbdxZgC|Xk-aHWITW#9<nihb8
z-7T6amRn&Q3@(^k^PM~SwT)tEyVa~1^&PtgS^{WS*xDsv_<+^V!<)W+DbM3kG1yaJ
zeGqo)LJQZcIpyf())wDoS2U_TR3`~F`3!`YpXuv0K!u?0y--NH5!e+8*SLvL79bM*
zjfWZfbT;`=iB*~PMMX7+cbI!lbNfdjm*N|q-PGINa<i72Z~k+*;uTc2UU8_I5cgKW
z=DK8ZBmJUEQ&L_qms~?<Hg(%RcNNjxLekw+ANjC<&O$)_$f#k;0ffWO`GpgbxHlCa
zxcKn5;X&QuwdJlamYWO2kD`Nty`)YZwqc&AdUM`dL4pe%!A%VjRNufO;Jch5ljuB*
z0oot7!5VhhfQ6{PT@O;;6_m`yDZJOtYi3Vw?|D#v$JVZXdpYRm2d%wgL$?{_Iy+0Z
z+k=KLDrbd{*n76wO(d--ly<5$(6)GTCoCALHb+E+`%*ze*FKiG6X4{B6jqr;zRP&D
z^K6BpTD=-R^z`Z`cEt_>r#Y?c+%_xhv2_~hd!A_IL|@kT{Io6A=^Uhr_D~m&o$Y5S
zv)x=&g!RvKqRcEfSKKQ<WJ#pj0bJQ~rVZwfjVSg3-Iy&8WLi$Gfwyd=O};?_Qstzq
z2cP%1+Fv^R0gX8bWn>dLh%B2n6BdCKj%I|tQHx{a<CYiDZdHiYA^=*LPJw53tCLxO
zmIIZzrZ+2mM=JU;qpracz5Y$ms<hk=|MP{0{=WWfb|W*FK0kM%e`W*rspigX97o+V
zysEzudtM!!FDL}ZqMQx;h3@UVnEnGSRbp4HKbk)>ba|%xR#mI7ZO?D}uN4$RuBKeW
zlWWF;mKdqAaW+eaQvAuH#4xGkT|X&Fi0TIGY1znxWdTJ61IPXtpZa5umz#C;zrVnN
zRDAQxTP-Np<t@DR^mh);hV~JNYX9(E!L-c;&o^_GMV1!*O-YU?dY%8GD!aWvxzvM{
z|CVQORhdU291pQR3%|B>rKh2N%~X)ve|*{8lro_$Qn_^T0$U_qO-ZfT^+_2Ed*QQ*
zv$wQA9WTkQh_NcVYR+0u>P9i=_p}xkuh@w*SM|~;W1@Gf68vtTsNS|DbZm^2UBV=w
z=mi9fNEvJ3VOn6nTm2$r`#ucMZ+AwQJRV4~Qy3u<im?{RGKYJkwt&o<j2Zy_#w|h0
zU^ocLxUu)e@N|#=lh2K-^o&;}#i|*5Os~T>^1^I6X6d!2TK#XNM;@GIT#ti7-9mSR
zZfeC{EFTkyw$1GmU*Y-s=9vL<)6aNx^%T>+zd;6zAU~rp_}L5=yuE2mkBKds-McKC
z01++RU?BQV1xm`HPW-+R!#f>L{}SB?impi;+e!UBDWj|!P~iJ@ArjWC#}B{DYIiy?
zUpw>(07Z<oRv0q(O%@;1Jwg}iO$NpQ)p+nDQdCl<5G*g(&{Q?M%!NLE<y`2*urU}G
znB_i?-_-NwO1dQIWqzk4+z4?N%YoL-D9IO39CdW6l~Ft)H$#s8HnJIX5m~Q!=jJVM
zu9Q>2GfX#I<itcZuP*M)S;@oYLbgc7eHa;OnUP+A@9z6uob^R-5nA#Kmhs1j$&nv_
z$jcHjm1Oe)^$l1fi#1zXDwa<zFMP<UBH`<m?W@Jo9mFOr&XTTJ_I`PE$D4EV$@TJM
zPMmp#%WZ8F2^A(vRQn%opYL@xg&tfD#IL7doc+sfXVz8~6{>_GrcvMeb;TiaExduG
zop#W=jfGq%Pv5i$7R|oWw1Q;hw$YK*a?m<(AdV&I-nM-!f%?+M7k6v1G#SOiz?c`8
zd67GeE-@pF-NtW8J=By1UXKa5ijG5t*FAO&h)uVS*CKMNQyw>SQ{RZeQ$x;W*fT1z
zQ7YIUsc%b3H=HFB?DhTu|4Sc!KXUxxDnUDb8*K~ZS?YGx)u}cVdhXG}2p3e$`kod8
zvx7hryV>~wHP0VfU~#z8C#u&w*+*gLo+MumZq6+adgXb)i60JcaG~tjH@dw5Z}s#1
z^%pahRJ#mr^fE3dgD)$%7<jC#;K2JK_%oU*w%Tmyq<TNhVv#z5jFoSPJ??mxpAexV
z*wY)S^`2{aSvG6_146OH&|g<4cE+l>l-|GD_==nCw4YyD;5bWTKW2KPLmT1BPRxHQ
z2f-=4dU`MwYso&{U54)_U-v`epKto_7I36R4X+Mq@Vj3(Y|n0;JsS|^3-?35`f3@S
zT5K-JcrK#1%UN8hpGdg3pe1|wQ=gMRUvIgRQ0ZlaKyT4~nMSnka`~fon1d8GH)Yb%
zJE=0w%D4Ng_}zj35KZ{jzG&vworxoR_ADGimYG(~PClkkf_@bvk)Mi7W~4w};T1ie
ztLi9p(c+$=>Q^-Oygc(yCQvF^el`bkKwf~@s&LgRWAiq*1&_1(Vi;5;n0a$e-%6QI
zUi-bOQ77jLPqZhmdmOUND#R*s-kRPL3g!CON!)k<73ETEJ#t-2EV)szsL?!gF_E0Z
zkddcI0NmHuRnQ*@4q#~RyMLS1h9n`44j+fj{=PF=h$`SdAIZc76i1uNu@Xg9VXQ=H
zEj;fFaZ%qo8h414*dv1vB6O!%YLs>Ho7a}ILM--P)cjJ0Ol+<WF1d<_Rzq#GzB>kq
zMEI=%!oI?r#E~ujLzcfgR>h!|G`BPp_T#*m*6}9krjzq$PI|cnYv^IM5i~pdt!G7#
zW_mp)szH~IvDb}m*Bpqwm$6shmXRN_RBDlO->}w01eI9<aZFV%!dNz>%xtKmZGK|F
zgpDqGGWL{l_C|TMpXF|}O^|4l-?%sozc-pa|8x{0pSuZWF|P=v3g8)d?4u|M<?V7X
zQHiSW27P)EUK{|INsi8vo0VOtp6F&jYgm8W+V#F9ck7iIMW;be-#H-WG<5L6eDmtT
zmI5$!w-Oflb;4F|XL9Jrs+AR-6#&iHMyo~dPlTD$X)?hhGZ$5v{OP3hoYhwqLR9GM
zA`kD213piJnBUHX%^P`yj|!<G*eq*8+9obW66o=Khb%W6%SHd|E4u&v6`T1RgOv8D
zYi=W*dfjdTS0>F9k`-27>*qC1NT%xO!)!$vm|y*C4B&gTAA{+1|9cF$ULYsezInM<
z0y%Q)x6Pm`nWvS&#jX#>n&unJWkV;!mcSj_k1(8PqQ-o7>c_-_{?bEsMJ}ZCjHDH@
zY*@@p7*ZJ5!GSJ(HwcoS6^vBTjeZnH>8PB<0Cz$C0K7djSbnk_74z+l{7ma7te&ON
zW>The5Lu`C3}<}oFN;`$^{KV)EijU+FH~Af{irTTpF3nQw`#A2=tv}1>OCtkVbQqX
zAgzQ)n>hWbj1PA~y4pThM^r{wA&WbO-3pUpiUSlh;gS<|*WmCST>{~gSMm2g4<*F9
zx&@pzk(E<~Ie1yfqWhZJGN86DbmW`Gxd>@>CZM*y>42s1Dvijg8k#1N@czzb^Yk?&
z5{)v}kiNyl8ynSZ&thfl9(HVs;7{kzTBCkbD~SGAF0{e+CMdR<KCEvZr26%Ah^^1J
zbD2K_=1JYN#$M}+d|^RbbB!X{lYd6ErQ?5prASoXLG;H`9|>0CDNkKJ?Y604g&;qD
z-~HPW95~_s_~Ol_16u*O74MJOzc%*_{E6iM7}#4<_~j^?Ny}iqJN}aD28_;H(id#0
z8e;$A5XW$8>TA9fbdvRbK2Eg}Jrn4UEVnQ%GMR1n={-d;4t^%CU_T!1G=bPE#&3MY
zVibj$CrRS3D}`B8vE!@aT7;2`r=Vou3~vRWLv4}aLKF4_I5>6WkG5-^acLV?n@gPu
zSd7j)(vvZB23M_e(!EJ^iZku*=RRqf{#L@ZSaau!GMao<k#UhzM=N-c>YLXAeX+W2
zj%`Q@+17yd6U)K2G=u;)0X6qX{G&+)-w%@*BeyCE&m0#~dEUMVU%6I-)sc<ri^Gr1
z;WWd7nB*GL_A)bpF#fWjy^x*piCPGqF^drs+YP-aLet47lo$ryCk+hpPUt9v81S2N
zG#!3_CR!;&fo@PpBxVfp4Twd&bP51WzGZs*c$yv*eLC%VbqsXwN*e2F|DmwT7=M4H
z=Q?}o9-mGuMr7b4&v%%LtDoqYge-65j^z%m48)H;Ie$#~HU3u?SV)civ2?}!l;Ve@
zCw7tD&k&R4HK2ip%UBdb_Wd^oYd!!58`cR!-bBPXv?A)A0lS6>L@p&OtHVS@*|C?|
zhzK@kfxgHL_OV@69|OHg+k56}k}X#ZitujTv4JmS`H{L;B^UfTT}vBGuMt0_eV??Q
ze78v;)k})W(I`ZMfp+<)K<h{>T%-A!v?rGA5>^H=1l7A1WOief_KLwUq~ZO_$psdo
zuGYzl%a*N$8|v$vo}$0tggO(bB*_!?axw?vWXs>bo$=v?`K7&Y{kF!J8q@oGNL{Fy
z?-M7GWlA{hU#r*0@B1MwK-<F=B(jn_#Kh6V0wBcE$gjTFz@n--a}tG8lsL9iPqo9X
zf#8`Ht{Vkm8uI~G0<8LoPGCo@@)=Nv7d)7!9o!X`!C-vj>$>N93|j{q=D-TiSsa@$
z(^CrNTu42X#x@*n^IS;afW>uHU-4+(yoSi>7^wHj1^$d0&6`$WdmxUkBY1C4T;1*u
zsEm2K&s}X-Lo@RNG$hV_!YAnv@p%WAm2v6Kdza5^fy6EvWy+Y|_514c?R57&U3>^T
z8foO_nzHIpWI?GbojwAA;=m49H+^`6$&~qRdb-n0)OE4PMb04N#yh&f5WspnJ&P}!
z8q0<is2@G}us*5Zp)#(*`v(irRLI(=ArV&Z0oLy#Q+8%qYEWzCe;MvPH9B~%Y{FmX
z;cZ`y21`s^;f6A%wBON_kxG;KPU7n)^;UH{YF}9ELM?lBW%rsL+^VQQsk%`!H1xCo
z=u=#p*H%wJaHO0Z*;W*E*#XM%s8(e=IT=}&<}abtpR?E?s$>SCed{bTM4gttOx6J#
zj%hyt(%X1-N$mPekQH=N3B+mrmxRWzyaro6VphfQMi8o<8(cX^`=*>>kO0iLh8EFA
z;93O}$$0%#78RURk2*(m!V6Bc)Z>iv@`a-^SFMdHS@B$q5T?Uye8R*g<~-MZ7R302
z6}QE84_`<$?~{4;#j^zpJ<1M%8P=qW*qvb+T?H&i{X4|tf02tUW{1&eRzV6g^|^K9
zv~|$lgnBT`CYf|++srv3ei?mFp`{kP{zIL*&l0QdDEO4!_`@v*JZ+nG)cwVH4vk8F
zD@mnn;XIg6vJ)!k*KW`MP}GKF4J|mbDMDPD(q?a2-y80&Vj3jhCx6X_IMvaFS+G*|
zt><l*01Kjb<mEBF;RfyG>I6fsv_x3@j^+G2wfx(iY{VqR;cB>KNA{1hXH{!so;u+!
zUmT+Pu$PC>uawNpwe_otre^P|!8>rm_yHopY6m|>pq|k#T?6#fTKz+ieG@)YSlKJ9
zBuiWBU9rRU76W22L-Am(ppqH0G+t@m(XZ^Ni^UWnzjp|X-;sqE0%fT4HMwF+;JWNF
z-`P^ss=gT`yLBN9DXBN<bn9+a3K?C7Ya?u2vE^SKY4+L<L=Xt8kHk-Hk=v|rUXAxH
z_c}(z{VG;Z_0FeIDSZfw=?2-}xW>mAFzp$K(y9KWQaq)g=?p8zTc7Qi4vWdJOo5@A
z*?5?~Pxb9w6G+=CW!P$$*^u(*)k`%2QbfgL8sNv7sNNL0KW|pqg}>jdF92B>(<L`<
z?@A;mcxs)1GEqdeU@!JQZ^LSLuSFAeIOT5~AQHqw5f1-H{wnCwLiyL{A`r6M+#m?q
zqESk!FcIk5(qeWp+Bg%fm<7=UkC`y^l&V=mMnHCzn)4j7)<jKm!mPgs(2P-48#VWk
zE>@mJ!wG&f%3nsx*zpt$xv&dUFKuwljkWuB!*y~M1kzGUq8aGaf^!C&A~+mb?n1TU
ztpI4hgK=v^AUaeACoS6Ai%EBn<~d3T*r$ysD8Yi6o?W-l!a>80#UaVIbCZ6S65Mi*
z#s+RM*Biwy9nVP--|Yrd&voYnRst?0)Q`c_yA6ekSTkQ6)vCe1-985Moh2k^q<-|0
zCQaR2=UbXG<A2a8P3SM)q3n>xm5zEVl7JcCjPz`V-0#0}#gl6HUI*Er3x8V9KZr|8
z{WWrIJjRJ%2-4C?C7YT;?2<jfZnXk|yt8yjlMhVq#climDq&)NGHd*{8-&me??CkR
zM$WReN$qSEuZR*jGegctA7F@M@w>{Mu#6+EsH+0r>NiQ!I*^eC=tvcMFmlV#vD6nG
zM5~}&w21g3XW@HH9Gke64OW%-ygH%K$CvYh$v2O&rhnaMitYcoCt0Wz;i4_?1HLB^
z=a!Blm$v?WKm-4Wzb>kpjq_ir{mM>cQxv^;^u!bN_T?-{%VjDgmV}MJZZA6B+=P_w
ziT>=Oir40lC9y}9bG~dF82Rv>noCOGw44){{c7scnW`N`O95l@Y>AipJ8aT3e_s+y
z<J8`C(6bChQI+J*@>`v=`LCj#CoFXkG|H;w16oJs2x`t$!)zluBcOAIveXm#N&TYh
zHJgqs11Xo|t&}hLif0-5G^@)zF@d-At8bjazQAJGi&sacPtMh^+h3|PDRb-RijV54
zv>zw=sV%&9D{n^I%_>i71cTs9eT_;j7}LnL`H5bD7jOACw+Mu_$7(V=DrX4B$DrK2
zOkvR&!AHiU#;+l<42xFM(F8gx#=e4SxP~aq$`mLsfLt^>a{K*jc6p}J^ONq3?#)o4
z^l}p5mdXD$ikS)iz8oQA;Bv(7Ep5q(^ZV?O{#p#?mF=$phi%Wn-EVlsy72!+cX3K|
zEf#+t@2w~Vb<#%sg_YD-mNY;uuMOm$7k?XihbwFE)7yi=>#^*yNQY0bl9aV9<KwVP
z{HBR*j8w~G9aC&S-tGE4UPj+d4LSHst%CFceSB7{E+4LQxI=x+s9Q%A6#J|3c?;VZ
zd0FhVYa!sIl5EK2#fN4sQ#e=)3WZ}KcC42vhwQ4L6Vn%r{K|rZi?V-pdT8sprafM>
zRC}q7D9IkvR;yqrd0}x4o8Uz0yI-t5ty}Tz-5`g3gv_^}bI};*E&7l<!g`v$6`bzk
z!+kS9&Qa{D?x!rw<$kl);Vfwko?d)E2$Gt0a+PqZ<aCWqb$)1>x6a!%aVQD7#Pe$7
zF_l8tOc#K(1c%Yq?$s6PJ&@ZSQrJBmN_Ru=!-=pM2FMdE9mdH0%V$b4)XL~>Vg}hb
zGw+$cK@k00REAV$U(7U|8v=!&T&ZLAQ$@d6Cb5UaG<VXwKlA;ALF|}*lOXBOJ!#52
z1ya0;E-8S$YooEpl8i;fr1ok4ZE`=V=|k>iUhMZ8vx#Mk`=(A|kmIjGQ>7OGjlr=B
zUvJ$gIaIxXJTj?|T(WRHB1$~o{ak@bRxn6ul@pne(D*<KYQ4!#-L>LOkqb<VWogd%
z{N18f0dd)Djlx1yN0ugwI8RSrV~*-k(9!gkVE*m~X|kGJ2G35;*C7tyyPLGG7ZJYL
zpM%^5Eh2#QtI`nEt;(8+3hbF`+4ZaR_78G8)4UO_e>Gh`;`N?MaY3$!%}t9lnvfp|
zKTj!<i!G~TGu3a!yUexLPEQd&aH&uBc3o709XER<rUYXxxVMQ^YWxMeRv)x$P9Sjk
zvuZS~GIjxt^zx(xSQF>EQO-ieI5zkPY9BmPQkgD1ji1t(!OR7Hk)0QV<UW56aOB#+
z_gkYbNTfbk7>8V~|K?oVs3LKxQx|R`+Wa6gSY_tdb4gXY>#tcBcc5%*$R&LyP3YnJ
zj5_9Dx4EVCb|k+ph^<SCSpbp}L8qXOM%89WiOIW4)+Cqbpedt)4vS`=s>r)DMA4&v
zK3rS)Z}4chy*OjK1IQ~yeHfvMCYKt$y>>@tqkfxor|@WEvXFcHUj}vYO&=lqpKsG_
z<dUV8rqg4Hd;<PP)tB{$y;7bXMeD$@P!(IZH@haW1chx#^Qa()*V+gUaT&ac6?<|{
z70Iy?S$3mSU5$;`1vhj!0zm&Gx^77IYyO=BntC0ak?gN~E#HPA``sm^cx_MLt-7Hs
zsnj!z473%aKcT+Es!HnDBzYMJ8X}Jyu6|2Y(q$!@6u7bxdnzs}RS)q_JH4U|Ya_52
z5~p9GQpHw7M22Mm?*5jk@x$AuYck)iZ3G6ZugmWAIDFr|l3(SY13ALl0?io%4usdj
z<tafe5ASg=L%LC?&14r^MC1|KkQSk7BoH$!KqkrZt4*B`yb8d}l-D(XU0?26@Rg5s
zlmh>7t&?9Ek{X5uMZct|uR%vvSM#wYVC9Fx(^9!fKwGK=Rf5e6FkwjY@ayL{02#$V
ze>NrS4iDoEK>jzVaWLH*r2{K~=%rORm<psa*SYj>1WhsQGOj3OVzzH{gGzY)VJ~{-
zLw=0zA78#K+DJZ@4u-5hJeHZl-<M|=Z(?`wmHBXKEm$D6vD{!8uPzCN9FzF_=>3}l
z@Xy@#0$dq#y44CZ$6#(nHJqv&_JH`dK%;QRzNW*V8I+ZHl{eG5?i*BC<;hL$20Ck_
zv8|hTMi`^=+r;`ZVzh|r!5T?my_N^=7}_(cpRM@V3?rDkYU>O&8Rnc-%O43V_<wxo
zxd1R>ZG578;U<(^MZGJ%wRru$W?rI2Z|fwCv!t91(B!p4rS&gUVi#TgkmG!rMV9HN
z5|2{r?PO*iI+Qm$1_r13XG0+Lq?|V<Wj~*ptLt9g)Hc6Eg8`mn9UvU8nFF>7E}*j`
z<`xV_Mk%FYk}|egK)1I0@Crp+OxHicUnv>YHbe8zwdz(_z=Ja72HI5@q?d8Gl27?x
z0|_oV7uS@Z@0cX#-kazl68gz{v5RtGaIe0wJDkIz&%<cU$kK6=N-1wnPi+)i;R$=g
zpt(XIWP&K}HLt<OJ;`Vl^~KvpjWJ`{pv2692o^IO^8318zP{qM0ra$MJ<va(a_RAJ
zJuS$}=FW9dF}F=zk~1cz)ux0RR=#7-7vmiL446TV255VZiULoU@~3;=6~`X=oinJ%
zJM`w3{;i<BK!xJ#_23O2ga~A#S77O?QV-*C1GM4EB||!F*r?gjM!ypqF>DQ%&oYX_
zAWX9HVAAf5ULBNmTO_>DxI14pag6odlkM5oArq)~%lCx2U--fWs?5`s@uUaOE(!t~
zQrOY{qaz-BPDzd?4^ir$w|CX{v?%hT-EYa{8`mZsQUw%n*$w4J8MmR9tnJykT;n#~
z>$J5W(Sxz|3~>Slc}*Q^BB!?w3%&zu8gDj%aliyl>%9^4ET3?HyJE!-CFV@+j?ZO#
z!#5MvJA$R-_L}y>>{zIp&D8Tay}`z0;{1m1wri`^<1o%^ceh%zmokIPNw3KYpn|OP
zek>38G%o=~Vmi~wjo#Oi+(^-dYLD~I<_H(6G1!<dsWb5c*6LTmd#8jb9;6yr<ygbp
zWb)cVfA;zdAu6waUF5V+Uc=SW#Fu3ObZ@&eB6b6{^@4WxrO!D-fj@?jp!2^60EBir
zFtzTXkhr|;v293=g!GBypbjLX&iCuFk^h6cAm_ik3kU?C!eX#LP{a23*ZK=~;?aCI
z<z?0oJ8{Q}8|`0uA}m%z%h&K#*{fSKS6X*nvr+~(oAaNTSyUY~@(v%md4|RaHf;5C
zMb~33w%%>k2rJ?FRN_(rlZ46tOxF7L$RNaZp^3{6*I7F0+^a4jk6>>x8}kqbp~i00
z#qpDdg|8wNd&|gD3t9+C<x{>g;iJz(&s{lCG34t~FC+qN{4k?-MR7Ak79>KCh%t6V
za$=c3;Q5k|EmACN?ZqL3SEgJlCepyd^9sdlTLBR>&0fV#mg3^c$P1&GeuaGyLiSn2
zU-WKYyCJ+%DLe=+$Pr}_m?@2TEf`!66j9m0IV5uXFfXV+&Wz7E{+R;6;7KkN0tBof
z_(zav32p_jZnM@(FRl#9GAMW#7AP85P@`Ft$WG*fsQ%GCw;%ltH(vhHT!`s(nreI5
zJB7k~bIH19w4m~SnUdh+`{UW;B_Gm%@Lje3N>05WaAhVLW^c0n1NiGSM*iJ;HUAiJ
z;J~zVmRSoHq~wsE84%?_ux9v0J-aRp7R?%RIDd_4=XfZ#)2KQ)|25GGLg`VYClxw2
z+4va!sOY^|&J&T!@T<aKM9^2vYi&T@{~S40{Ov#kpK7lT0-lLWULTw-EY;BV!ot-$
zRvqm5=P^ru7HC?_A%DybMCJ1Ar8S!Vtg#HSSuO`MxSA_8v+vo;8Z&^ltD70wF0*5A
znrHFYS13iDpu61H@MT@YZ%Lr5a+<aUWwH<Emn|i6+2z{xJ{-xed$nq~dW$4#h;7-u
zPax#v0Vm{)#nB37<_wG*)l57zd3hiK*vOszt0OQN1YSg_O4*vrt8mfn1_DUd7@SuO
z*lI7>&Xva&#$f-%k`Cjqx{>u>^y>2vzOc`Xc1$#m+1GvaH3xp3)rZ4HiGXR$sNwAa
zRMrg#LiFPT5z&pWjSXR8fTQ|7lf|cV?|XKZP|Gbi>uVxngE%3v$#>T&y<P%2qQL`u
z6aV*&6&1Us75tWuYA4D><pAbL9d7iSw~An6-M}FO&)lmKFEh_$InnH8pRx_Uf5G(!
zZrZqs%K5Xl$!_#<y<AoX@FJVhdBG=vy{qA;<@dbaGc6Pu^D<Avcv;TM+m}?t#G#Wj
zyh=A&h>su4whs2xMNQ5`4X$D~8+h7tKU-BV-<=>OAo01C8M8`ePp*uWD{W1!*!fF7
zerCZ^{ObHvYhlmJZBG$FlZ!6bgj|K0<+@Ri6ALDR_o`muqjS))rp^17=AYS0o3MTk
zkF+&n{r8T;@aA&uPD*5rMn|C~qfGdEd-F|<7l?CGDDopiK_Qsg^R-Y||J;r)NXx9~
zgXVDKP%(G?CUtdqWIjO@h}2&nN6#9AagN1XZ&*-7xRLG={ac<MUj13&pJxh~^)c7|
zkD7TH^l)XOF`p@1A(#GZAlry~y+1ND{w-32iLB}&fsHs8@S(1XqT&!VPj(Y&Ca3-8
zs>YLZ?Hv9HWg7>R|4XU9{kKwm`B)}K5!ZE&hU!Aqa-g2S!C0~p!7<9-+g2C0GW>vI
z!!o>Om5$nAJr!wl<9=nk923>=!3GSW)qzt0IZAO}^l2p+?=vM;d>uK-+aka5Qw=t2
zFd?>Z%~xe4B`5BMmEj@8knVVn-9QrGyv|p+!7`d}h~x1btN}b;R-OL8x=<FA@sdUt
zJ)ETaH}V`+tTO3dHm`)a^2tR%277uCmn$;I&UK;K&3bx~T`1!pE8Rakvmnf<N@u38
zmLX{<tff}>lOc<A3Fkw&+<+4C)g;oi^0*=*zFrbiCEVq)qjx(U*bEMel8njVUD{`=
zQah)&TPV?<a#hj%fk~fKwBeGQFVAu0cD^oorCnzRkCR*G4UT<|LcQkPa|(@CRyCvW
zgrV3Dvtf+9$PmZ^-EcH9hRupY9dK2}&0A8LkB&CM!3)NB-D9@V>a~2~w7B01eDLW8
zZu#Ih8UMJgAis&=_|W}JvBMzI^U7zkr`HZi#NVwE^p%|38Hy;|r#XOkkrVKXA5rFV
zZ2>|5#xK%30XZeRw9iw;Tb^<!Q<cNFiso*-_Q|*Ca!ATJ_(9}Z*iy0()Y`^zX<0a$
zI5p#N{8V=`YGNS2yYQ%z_>>IqjkFnKeA4AB(mf(`n_k34-@VElNajF=IK6jfqtLY~
z;oFTq;&)BA#2nYdkjCBp_DUkWsi!LH2F!3KDJ3c-r(;`RdI4fjQ@WOGPe1njjv`^;
zXnf)13qsq3Wx=k9WB>Q*9J@o}qR&ZV@PPcjQ=U$KOGV}yetbek=L)TK5Qk4<tn}J(
zYkJy=Bu>n71sL;`3bm0~QOMxusUJHBfty__S+7lcWE1pukUE<$0~1K@K+JN$(u;fT
zPXk<-dB*F#W@=p1QEGtiZ@A(EY7M`yL%U3Q2b;EFT5lo*c5Zv<3tXiXy?5Y18?Uhe
zN{Xw=gb5(`!#yg3>xW0xnFfW#Vr<xZc$p5ThNMB`LK?K+vtHHfVb6>YP(8R?z2am|
zXW6hr>)8-F_Yd4Svkx~|MAyejxa6BgaFw0*>~TxAr^^k(qLU;A&jrcf#t<U8J>h>2
zVP`*CyQVBo1(@(YH}lC9qf!()gUDx}C!qKZ7tfJ)j>Ap>a`W(2kdnc$0;F-%=!_w{
zyBZ4h>3bk$Ua)bX2ofc``HC}n8EJU7=#@Dk!&@TMZpLoSlkiZ@xkSiR^LU4Um2H7v
zT!uxXTV!h@$DGhf`&c7n$&2d3@6_f2S%*g)u7#WTaM>Kr)7oz&KYxGu({#E<*LR33
z{?yrR_eKwq;)hdpp`ua1DL^eeck_yE*JM=|q$>QUiQh$zZr6IoiQY3$N2G}bUQ+kS
ziNA)l5xsuxrE%xX<ifuv4JnA6^7kZSmjf$%SA^UVAmD^PIC+;gmaQ_m4BGlg<HgQ(
zz>+mTK0S0K36(lyAKQsSwR{BYKxI1gWA|)vVtSu1+<P*b8PCoLie{}0lo@*@LT+o{
z`0LkM)jjT}bg}iS2Wa#bPGWO0vyWc>A$`kyCWMaHSjl!9cl0psN`vWtIfYsF9Xx`E
z_vyrKpg7}pG*w@r>r`K5IU6Q61yHj(5!K~@PK<gi|8HRazYHHH&9xqXz`P13qt5DI
zcDQWKIZe8&Eh>*(8GdCg&a?>#fl%lq`>$+--mW@RXR?)^$y8%A_k<nz8{E)VgOj#a
z20dJDF+8_j3~;Z+CZ4w@KAJ8Pz8}zVN4Wi?c4>A<%1`r3>1IDsb>Hfdp0S6@CGozo
zzZ{drga%GA-it{_o`)9)c*v>xJ<XWC4MdnZdE@qu5(#Lo6Pr3@*ek!f;F~<x4JFya
z)uU5{Pm@tV1ic^mwA74bjpKG0Oen94c$bqo+s<W$dN9PDR=;AW$<fi7Ie<EMci__A
z0~lWMzyb{YOlepI@(JUQ;TnS7bL+1F0I<JS*Ibghl1iDXspMZ+_6K_xlkp55<dRxb
zx*tkNWZlLs7+zXcSXZC%>bKRBGJ&hYc>-5~C)MPGc!()oM24Ty9p<Q|QYdhKFpSnh
zchef^cDUXHfd38nh!4nL^<Z$Hrc!kFWSlQ3rMWc*rpcf9Lv~OmAMdjr@yDXqANO7g
z;c^u9U01p#+ftiEE?*=6YgqpOG++Jmq$T`gz6u}@<dnoA9kdNb9!QC-F)S@JfpXn=
zL9GqZ>!KU5hk-!(yfgpMxo<f}L0s?%f6>YniDn|D+B{rcN6}>%2ez&0Ii|z04TGHR
z7Gq+wH95?eCd~bi=H<tt$LW11)7&a<zLX_lmXYe?1xe(>RF=7wM>>9Gd-&2I6I0`F
zC-C^)-J(Wj@z5)KRu_<)n%R&ewxcYy?u@-r<|gkB8IQRXKDE9iP)#bv%+g%9v~VA+
zxgt3iyh^>fAJ$`ucH?Ee#xNMl!Xj)51PN&`R#S72PHr8d`cm@@6*^eCJVmm`6^Mi;
zCtxkGbv+u11192wRJ-6Xw##{HhK^+dbeU|H+|?PM%t8f~PP`Y*J3~J!*mt0heqB-(
zmA4Ja%-YV)+ZBxtUK?A2>$MV?fexeFwM<>$e!=<&5BBe4d5*Cq00&p48iRE*m&B*J
zsn;iORyAh#%+{=VUHX4iS-=Bp=<)(85~D#_hpiA=>(PAOC?@N$Q5`AK7Sdts&2VaL
zQdn7{WW=i2>&j|}`P;l`3(?jmiUUQS)CRJNbR%cHdfAzq#PUlyFOyddSaB1GZ3jzH
z|EjEeMx;POHFsov!_Y{Xr!%415ad{fp?~?PgD$Ck%rL9VSkM}7`*Oop3Bgj;`L0Mh
z4lk_KUqT)@FdBugc!Xj9U~=Cj8^Wv}IL!9`&^O=sl-PB7VQ6xuyM%%%XGv8<U<Ke9
zd(Q(!*;Go}XS&V;?6s+!_5jb;7ayT<HKq2+icQy6!hLN~6S4rf1a}r<v>POkwWj;j
zPt=1Vzte6<(j_31s{l}}@rX1>G%vqAj=KuX9YRnJI!hxzuy4EIx3ZFeI>N!n@nqa(
zzf$bc(b$ACsMYmV1*68bwFC@?k$NuVO|u^kjr!+<4gSW5cF`eFQTArAQsX>)VVS?9
zGU&O><m7_=-n{?5qvKAHgLV?z|CN{$RD{7C<2c=oQV@q)+ueZgs{2cCoPE@ATouM5
zE69cZcFcd7G^-)4aWeR&2P#B2P5VrH3}xvp1ge1aC=EZCX|h$TuLmD1Dz-<3&%FeS
zy*{qQScD;&RD^Nnis|MvrMR^AYzS=K%;}2zQ8b<wDDK;bUCz?Gy=v8=rDmUy>q$RS
zvIC+>HpALqfN>E_YNXsO8(@qhq?-1&30sR79W%%}R}&`_U{-OfjjMQ0xqOv5qDuJU
z+PFCD$-Eh>7ZJr70fiV#+$)+M2rKPtUMkG;^+q$VJm%^V>_nh^4NeY%dD!)AU%3b4
z*xfnc5g?WLB9WeP<>jdqyYFP)?!XZs#cxJW5mePBuK=$hi<Jx;QIjQ(US8GXE1X70
z%LL4mxU}Ggur@(o;}`(-&aZ3Kr7j+jmcb(s$_3M`@27wQtyJNo{%2s07yXSd9J+Ev
zA^QEh0w(Rr5s=IN2}}3kDmn@Bkq4+09>C2rW4r#1+&$ZqntDI*<Zp&>vtf{*SZCEM
z4tZV$g;*D^90lv!nW#ddZrqPR5KT%I5cA5a8=SxW0?9fNQm?)$=0ICij3zSAe~q_m
z-&(;ZI$G9U@f_a?h?%=mKCck=Q0HpL0zz0_uE`AH-62wSebmO)c6fW%Il-;}8omJn
zO<w)7dB4~OK7zP*UvfIS9FFyV=Dp-z;r)i3nl*O9USXw|voLLyc#nt&igO|s6Bu4_
za;k^dGJ^E<=fUf~l?B1q{E!7dIy<MiZ9HcV5_c?sP-$IQTAZ?bWqf4rI70TYYnKPs
z)!FgOtxt2EX6K}2D_VLt7ha%d7D<M6?RJ+MtUQx2mPb;-L8=FyFhR(HNaRIHHazrK
zK^Tw((goxTM!7@kH9kDTou=u{UiZp<4-etJS{VW?Y_X!mnuzJcDcK?iw$8&g!q=C7
zk#=MLcmf37elzd!TD$A3PV4*3`_cZ|`%IK=m4i+w^fI|pRAg7q;kR=%06r7)-}L1R
zz+lHZV*qd1JL!{A`ZroX(*eQ}TfJB-$sJW`)@Fq2CJD-~vWAmZf}s*0kHUD-k5`ao
zpH&{fg-`PqTOLKnhC2@6XvW$_v>?$C?}|`M?Sh^a1HF=Y--YrZ3(qn0Lxd`Mw%|E<
z&FkPn|4Z4A)buLPx*KmrXUrCNgR<9#r5diVG8F(C{Ksk_JcB6}6vcUy9xFGNZ)U}t
zMcYcBH@QBRps;+;?E0}YtvD0tVUzjOPVKsgI4?B`7C8#Vqth4f?T4HdcFOvxf(pWA
zLAE|3Qzj#8CltwNA(P9*<?lHdj5eOZ$=kJ|55wR4u2oSOYKr_74lHeynE(wB?^VAk
zZO?s<zPayyFrb?PT}(g!#c}dtx#`|FB+~LtjeOy*0Q7(s9B!@onL@e7!Ka3PR>d|e
z(5LFfd?z=|B|QZL5q9GNl*Qu}wTmIP>6ic_g>nfPJNM%d5bqoRoFS**4;3;I%vS|;
z9Y9L{oPLz~naY^SWMJ#a;1-N+6d!E8z0mbLbQH~7c56L+f6Dr|gqYoWo9-p$wAD7u
zCGlN%<dVy|1(mJ97+~ZGy4m#Ha_uqIt~}(<O~ih}Q%3}9UA96ISQl=M&)P21()&wR
zOj_v#Dcwi#tO94*>C|qXi-mj40#c{@QOLoyK_w%GvU)j9JxRNtWfrMIme!p!q>mov
zRsB_{Uo$1vl2#IzL`^2QuESbR4Wj+Z&nzn;gpc@(_L$^Ueq72ja&|J_I_)INeU{-#
zrl<NGEyz9>0tx*VN0=}va}f{qxok|oqUIr4c0a*Y;`pcHyTU&J6AAnv!&hSQLps8a
zr(k%q?kRhD97y5zHA<;4htFo_hdKa{s7o}}%n#UKZ%1|Q%5RO()^5OHt1`k6ixUmp
z@7f3Y*RD!`_KZGt(Nky1h4r(<QKi*_jy?;=s?U>n8!x<umj^PSSL5VqZh2560xKBJ
z3Y_-fP=dZSPCP9X)w+55<UKo{+3Vv@brJOpwYpz#OF>%DcBi`%5CxmjApUZ+>1SIN
zxuN&SA4dR-pxkd7zSS9=ycsG$9O6Nf$$?(|q~^MbwPo1Vo$;mE^KCx_&ip6WRO=ja
zDOF@B4Ul4TCby$9wlB33#wEZVUJ#Il!Tfx;$96SWnsq~z4q?ttY-IeftnL>IXx<h8
z>%(2QKV^%smM6)NPhQcjbiS1Q5Et2W-}+jndk`#@<VBh%`FRQEdv63|@+72kH%x5K
zH2Z!mP0VUPp1P@sF>HWnF14%F+SYsEZ#o%9Hu0jB^!wO|K?A?g0<)`jG`G$&V^T+_
zpFS9$&+!bBY#OQfe(8M4>jJMhPPJt7*RI<v+4&*8&+#56Ct!=Tgj}OrT~zyZ5GB{A
z7riCw2CVq#<@flOI(>gQ*CpP%?OXMN<4}fnrsRmzv{KruDN;e+p&xzNd*2MN9Y1{E
zI-lMx*jL4uKi}Ro#5J<Qqp_RQSy<MzbL0!ouR0ww%i03UGi3^5P~+vYT@RSt^*^Yq
zTY+Q`-5p$}0+V-vG|!1lvM?BzgV{2Rio#rY|I+sbG5gYmuroGn<yfInp7%`e&5Thm
z12nkbNJY}~aRU$4zbHh7tNX#ER$ZW4l0af8m`Y(NOH5$OeusH;9OA0CTM~Po3Al2?
z5DteGvscT%*jIy883V%Yf7*R{i3B;x4ZZm)SBY|EZUo4L$Eta7CuSoI;v9(T2}q6r
z{V3esx?OHY!~i=nFioH6CFBfwZF(hm&4g~X^6V46A<mb9s{OSpiC9YrkB&qSZCr7x
zYTfzCh_huObL#dCDGq1V^-%cgV9H3j0}IjW2-aD`8-8tD*S0QeDxltE(__SG7k_49
zp#n(;lq*eWVsq=Ype{KcdvMgz59#D8lf#LMoi)#9Nd|qEB;dQTT3)PQ@7KOv5<a~v
zlf8Hnv0i|g?L9d1@?8h$)JL_%oPjV>f#1ey&Jqw$u30Bd%bZ-3Wj~LZ_?7K6RLe@y
z6kcEzb>L@GOR!`M3gwk;_j{l|hc10tS_H<A!eezOKQ-dMkx{#YBD)Ga(N>2|KBx=g
zIQV$vSCC6qJfQlX@sa+|80aqJFE{-FW{Q6jScf!yCi-@+{7Tr;%*3-dYLX#XW0|N{
zUU22ESpRj9Ze<YMrEv5=ub<~{3`vp5;vW&|@5>SP`GrEw-C9}jT$AC&q>-l+5Z97~
z2jY(bH4G#sQtv4@0yRwJQP@!;LFZN~a%qQDeLzBLWguira?SmWFoax1E256^|9ZIn
z*eW*?i3IBBQ9PKP4TPX=@$c&NGuS!zGV<_zGpwb(4XYOOeEJmw*Nh7!h&4{8Yco!j
z&$o9UvaQHBoi=5C+i*UW^SQWa->}o0WFIyVLUX}MgC}!KJbbqDZL0NgWhYTS@;SgR
zx(Pq@pCVM6I2|nT8X&Y?m4GKDcu3GDo{l_RD$BoJY}i&_-i1<vRcn7e4f;o2ZRZ^M
zzitMZ^l`mu@vXe??cH-`MhGlY8UdUJsnZYHTa1<d&SMIOtgC6^fCK>^VT-`SPD8-D
zvFL_IwcNEW=GY$W;Pfu4a_!HR;yB`FVK4vf6K0?)60D%<1nwL$0X<+1F5q~gA$lu&
z?^XF{BU`38F+V+^Fjw-T8?UW0z4xI9pUs{#@n|;Hg=6?YU;d~-^Z{Vv{@~#G4UG+$
zVxq42+(<~|p@aUTG7GN#jz`C*ZUB0UO+Z*|k^;7ZKt9zW(v8a=M$ckxw#m?FVtVX<
zOt}A_losYTLM#(f72mxc^ya+?i|nfm%SfUu3`W;^vIU^ausgN}t}!yF#){S&)<5N=
z{LZ&h2F|~(ELy<1-RqiIk4oL}OGDAk<(%=C{@kyb%7J-auJztn>CO!lHX$Nn&0sBQ
z*Uit>7l-ONVMy!o4HV2iC&&1sxg0EHp2X3<)b05Bmm#I;uJ#lUZ(_x+uZUmt*3t83
zYS(6zwJpQr!}I%<#6SC;UFun}gM{LzW!hIM<lX^|cdOxd=A^!uoofATdG11sf9xeO
zzce2c2~}V16&i&SfCb$IiyK?o-WB$vUlu9m3q!b$!p{%)6q~Iq;__egmp3@qzc$f<
zW2I&JSIUV5hi3&EqWH%QnW`K{M+I0ofY1{9V^!uVf`gB4)~NLH80aG7A&0IDNAFo&
zwlWL{RQ{DaxUd5PfUIJs!vzj2Dw*-z)cIrRR#Rr!ms!+Cu4x+uH9;W|)f4?rnd`4{
z&p>JacvHZO=THArUh~CtyKcF1?C<Ao=kMD4e?4zWn1kN|ouMBW)r!y978ZtDf6awz
z>Tkr*!jMvOK4L$w3(ydW?u<q^*@^wkJ{C{7W?6{Y83Q-&bWElWg1A29#&esJL9Bl#
zm-Wo`;hGT0;IY-QrcLw8Bi806{;AyP?4UT{RxJLkATet)McAP^g4gCvno8ViuR}IO
zGj#BK35`;uu;Q@UO0&@F(tt}^J55z``j_Li)QX=Bmjy|#<;+wl&gfK_(9*OKa4&X(
ztSj-he_sUZV=De53W=PS;kEQqq$^K0fR{{XE%x9iW`;!eR~M?{Z-N$s;kb5{Dhg$L
z>!)KeD~PBb0dP|GN6pXoFM@NsP=yO0n&n5mVR-E<Z%04QRgiNjpea!)W$E763fN-v
z!L(+?B={ix{*bz;8S47^N5((JL(YXf7NgNshPZ-BR+yDA70A1Sh-D>8G0f6|s#>8@
zmY=NeS@q${dplLQ<oj>_(7`Oe$M=05;dRMEV(CYNe-6IFeg4($NC?ujZ735|Ke;@i
zT8}zbcK+MbcYhMM%O@nu3uQiV?RyA5)}T=fjm@VYoWBn9waNxH@ynfQan`5iP%k2+
zJ|v5Q(O-RhHCNE%L|?`y6w8&CP{*j^)oKIU!+SlB({o`+`<V(mn_N=<^&j~suRS-e
z|DG*n&l^?$iG2G$0?u#chIe+x-PS==I#HI*!^%>xSQ45l<t)hjB=nn0s%FODv`k|@
zKuK<-w@T0I3v;#@dh}BO#aMXBy<-MM+%wFdJJjD_5dVOGHJ_^0Wnx}mgFQKgpDo8)
z&o>Hd4PJ6AR$7@`v80)@VGNOFJ|ZJE8&BEkYvNPdVoWaRE-S+Dt0Os9Eh96F#+@dk
zJvEaSetGk9cH^D#;0;k|9b5p9+TFfWR67gOR>0hd1(0&_83C;C)!$$OK=lH<6+Pbp
zcg|5@!GX@exuxg~!j{_Bv!%_<O6RB)C{#QyGQpD*ZEd`|<Rc>T$7&;({?~8n*KOio
zF79I-!8q?`C{Hf=DGz7acY~#0>7*HJt_~bhR4kYUfJcV^SURxV|66Q!n+QDiH<1|O
zo?dY}bXC%8QfjK$6y$J3K3}Iy3Ot%dC8U=#f<XWzb;7Au4}!w(JcdF{Iak6*Emmw2
zqO%txuwA7tF{Oqefz;}k|9(S$GKkctssBch>=&r*`-z+kfyy-vBjNoK-81M6Oz}7E
zv))@?cK53^Q6Y-2u(Q#PO(Dp-Dyi=?M{tR^M*3p>NAPtAGm_1^*D7nR*FU<yqD^P>
zJ@>>3{{q2-dwnjV=aN5ONRjC(<gpYdR8E)A_M<F4`>QKl`y0>6nJzt)a6d*eclZvN
z_hoEZ3>Gsq;Iyd-@|s)bK(Xwp^6H+HYa^I85747gTWf%-0V|`}ItF40V)CNyMj!OI
zR={gj0PKcmK)tNQ=4Z;Tmw5E)%g(}g*hfR;eU3D}-ovyJCL8x?t3%F7|6mDzhTk7%
zA(J<8E?5llCBpLNQ{VPR^90YZ%%wNKf<kqVvHW);>c+&W;fv$l8%2_5%IB{e6A3=p
zod$IPBvf+7G)uNN0D5YN(|dqzit!0(kg9DAVttjtCln;uvB2GDr=6GhtT^0q#q>sg
z!;P722%_gzja!px)pM;<U-RbG$%ng`nkc^#CDP*4ayYDnJEUz@{n_WiwoVdP3NG?v
zhs@DX8+5IBXl>&o8KwT&HgJ^2x3OvVS6*N(dmKyTjJ`}f@Hq4~(_QptTzYK$tWZ11
zYR<=e(Lz*$1Xqcm8U$ZaVm4o%lv=EJoBfL^wNG1FveKf+foikx46ta8M@%V}FLw_5
zV+0}oHok?=2|K@qF3v~x#7`r5Q(H(l8GH$|&XF~|T*oPkLl=~Eqb997!G-5wRk(kv
zm8NgV0N1q?GzJ($r<*7!5PLho=pTD=3C`o6H*fq_Gd=ozG+10EU~$zt>}QGL>^(FJ
zlqq3yA&mvo;x-vicV5Z-F+Qm&)BpVz#hw$(B`5W-*x4;KS?Vko<(<EOP@U=z$c+CP
zx4f}Bkz<zLT%+D`s&M!3_WJ!Q6{o}`u=G(MevG0;)Y6Ny*4K`+wwC<ajT)cnz5svw
zLB5e|YQ_o`F*9Q&wjiIw9X<ClNkAd6aa)hQx_|CwNeQK=Ifd@4*De-`uSi)~Wz!Uy
z54h`hak!6#*du>#H3zn8*1KBo`9N6Vh`$fx4+|!0UB=Ft#SZn?uY_D^7Oh{XK_zqf
z+~`xbUxson07b3h`m^Sts$2*&!m1$wt0=NtnKfC&tl3;lqqiXQ$|_A=evn6{?8YX(
zjF2`S?u8}R6p827*pz!d?T^Tx>>coIAW!>we*s>mLgm68dH;7)1mHjWTGL!hp+uE7
zwco8TGp)_VBg=}Zl`?{mL1nOqU}4uEY;-9I;|Pfn>kwN4qX{!eI@FlHIKT*ID~{RA
zKV!55jsSWi?Wce^AU6&?B#Y8A<s-)!PiIjCVJ8CpF4`bH*J*<!afsGG1jqA`-||T9
zo$FN7W${r<gTFS(7O9*jdL24_8{CNtp&L{gGqL{&$&>$v<grgE|2rfXA?bvbJ=p$a
zl^dpeylH$Aoym7W1hS?@bogQ@CF8caajpGhxYzZsvlHk?irYFVzOl_EIh=7bTdVyy
zJJPig9nJ3?Z!{du;541%^C<~*%OY(DPOdI$!0;6siF(eoA034B+$!=@jN<pA&t#`E
zcOCA^K&>6;f{{zU_)V*_&P)7*%E_8JkjZrY;x~`44sv6cS3OV2hie-YHx2piT$Uqu
ze`R0VydrvMKv~mFj^BEI10-7@5Z;|D^^}h|DP5o?wxx9YcdS`hk~VL8K810qxe7zB
z0e?&)s(wE(>tEPBut09Ee$C(K`6vT3bljS@g-0T3YgzYIKAisRFLgS1lPNBA=o=cL
zg)P<wDaC(|NRXnd%#6%L1~fi5D4%EQ^U}YiEYR!6<esqZry&8crVHPE_~)~5ESddH
zJjnsVqHtFxnrL!S!&@S-VmGyJ!?L|$iC<EDY@s-r5>v$d|1*m6yGrHnC<;(a)0jlS
zAJfl!V}hzM;BA76u{T(1n86ynfAbb)-1qLBJ6$to?IR(qJ9t+z2E>2c%o+lQk}~E3
zaXi5dc3eZ;_w6Q!;*kmplYhZrB`=CqB#qmhdVY1skijC2UVf?SurgK!+Klm1IoYi|
z{ia?+Wl-V;B|&IaSGAL1jeL~N?S)xw!lSz0MsZ=H=&R}kO($5j@Jf=wLs7%H;yl3V
z24AwS9MX~DcTF1=Iy%^CTi{_kw{r603!k$}FtV)&+$SC^<eNQ7>IxbDY-3M;yX8tC
zH1(Oks~|fW0^vliZ@&13i<~K`x(XzseZp_MDZZ0Yw5YvZC}UqhlRz$E7JEF2yB=Id
z^r82WWsZEDVQSxgWX4|t$*(vLKBI@Z?~m!L1%1mHw1V$SM)T4I)AmC}wF2O=0*T7V
zurfxz$D`X>kLDOUA}=mAy+BEYOarTR%>IG;`=Ki1_)Wp%#rICC-@C*8Y|dX`x3WE`
zqY=4;PY4{zkWk>#*{C)753J#S^TTk}sn)hfCtswif|Oohb<XaEAa2;kKngaaj#c{5
zhE^yfHVkBBTGk%#`#u4TQB*&1`U<_Bn%)^EUF)HW*okC?{JrC5to4UfVNSV<5gOB_
zX%|GHh!ug}XI`6h8ZThUUgA-`U%Z+IK#5I)d+z6Pn$=QDqRGh5Tx}^=mXA|biHep{
zjov_bq@S87PVY6L(^DsTJ;5mMIIbNP-<s(K!LY(a_Dh2OIiePVFkUO&z`Vm{E(OMf
zI^D(Nk1c0Q_GyP_NA8W`bW1A6R?!d5bb6_HI{l>8EJK7IT<sk=P|N4%?HO=B-n4+_
zBOB$=$sSSx_GEq&?dEbl>1?@=Y+ecVkg*aFyaqS180W_4OQq$Yx5HJ~Tgntp+p4$1
z$&A$ZwzUcdlDze<8ij_eKd+BzUZK!^V3JnfU>xPXe&j|zAu@`<vUyuTYG-m`N{4$;
zpyt5a1#V^@46xTo%CopLxQ|%6_a5dAhYA%sf_<HI59qG`6DVRa`x}}Mr3LQ*3&84{
z+j}V2q%}<RG|X;cnM}vH+_O1Hr%o09Pp0ett#)N?cvo*?RSmY$EX+^b=`9xSjX=J^
zVCG}uaA|Rwi}4u&QEE%cv#Y?(d$0_dURD&iRK_AJ7O3RWG;9)a+mx6;azP3Pla#hv
zsBs8|Swl{J#YnY&Hr0gF9B<~pa*ghFx9n$^!+9Q9$ov)B*V^Daw{31e!4R@ye^FFA
zl^t*s<@G$7UN1BhI~qa3)NPUwU+O9@8}f!%aD1q@ykh3in}R^_8+^5^&<MfQu$JSQ
z_&0os8B6_@inw~L4|YO9>%;x_uX*X|hPw?nCuC&O81G9d^+$K%Gc2*{!!2N)^MZKb
zug^>0QB{+KOFq4?27H9Kj+K?;Wx+${F|zC^e2FXsd6SfW%@!kWUASiZr4%A})K4vD
zl@mMJN-(SdJytChuT_l&^?3AS4Ul?KQBA5Ip`#2Z5@I{);A~0^@(*sRjxapZ?^)pd
zb@sgv&v5iu?E`K!&*s_aVius4i6IR}jkzKEfpZL2g4NEFDHIA+>xha<?^Wr47-2KX
z-_w?2`2HuQV+r2bv5@4SR5^ca*%2qK)vx_U0xITz!}ott_TKSS|Ns9#BC8@JX^5g@
zgec>fS;;!bmUWboa*P}+TN-APbBs7f_7*yHaEu}$WF9iJx9rXDq29e-ectca`}6sI
z|9Sn><@L{bcs?HYal2h_)s|WR=u!U{mF4=s>iI(|3lIi?;;afG$nAE9Yg~ij1v=>*
zd26to;y&u+ZnFPm$VrTXES*2hD$@Y6m2SXOh8?}?>7N}x2inorZWen^roNqgs{R-v
zNP6LEK{Vf8OLsO&d=mEq^wj#~Y@Yeg-%s{Mq=kIz%`xcHP-xlCZbvHxuh^$SUan-c
z5<K$VtLj`wGB_ooJ&Sh&Bj-;SOd{I_P#pM(V5GatrtsR`dI>T+zWDb%-%sIGPvW>B
zFTKN%lp8+A6~Is1<m6~&b4e@1rOCY#i#`9iJfF}VQPn}|;ETBUa8}(Yt8RS7c5des
zAn7;|3np3sk!nA9d4EGwNYF$!r(y^U6Eo7kfaK`+P!tML?2dRbn}R*|?6VSAV*`ze
zF3>6RNnlB_%#SF3&`ePatv#X}CK*<G0)V!w^2=&qfNyElwMbGD=eNi!9th@&7Kyc9
z`ZZhAxdk^~VNG_iTVR;@w+DiQHsEL3qMQ;ad+Su~h1(fJ+-{$JD90;Gc^aqrE}(s-
zA15#0VkK8PMf;y^0P|rR@Lw@wvbSNbr`}r;q~yT*Oze!#tCXk5q&BDzC;P?g#O(s0
zu$TKmUURhisMeze&8DJ{W#|!<#j(`EILuH`{*Aa7F2siYuX<c76GPLYOOH^pc82f0
zaI?K0mGpD@v$?M+X^p?%E@H$9^JAB`Dl4{2==v;h8#eR%&!dW*<0G!@7guT%y(gx(
zBPYMA%7*!)BBJVN9^^s{Cs+9J#274I+=i<YW77u;CE8XMI2*;Jwl?8%laY#e%n1MM
z_mQz+iU=d%Q4cfg!;HBL^iU|*n>r1jN*f!xJXb0{xr^va6|8Fqm6up&7)aJrf&KK@
z-F(9iW7)+#E=b=Qce?d$U5{m>aZAa{I<RBg_h7&rQ51LsSJSV7lCV%9+P<e&Q-`BL
z#?OtgTmD5eCrDK^eMo7bz!{yFS?TQf<wWo$Fa%*tDyR-`@gk-7w@C{Btyg@S%lNY^
zU~8-b3$??Kjv?}~dxY?pHBNq{ED?<%JyzAX?+zSd|A|_(k46t&|Gx0cysk<<{_|Np
zG_Q}RjE@-;)d3P|3CTzvhtTz5)5cgBEk~1hO+OnK^ln^sN-Ww<B_!o)*)OnznN^ri
zCv$Rv%6jwSEurknsGzh5L%2>8pG@tLkRDS<;=9>zIG`=_P1Li$^(k+`l&1$5bk?Hj
zs18e*AF>gj>}WV7rQv2PZJnJzZ&j8K;qJ56%o8M|_zJ8khCJ5!;?`W6_YWk(=F>Tm
zr+Z5yq?=a9q}<;}CP;NQ&+0Ct__-l%?wOHkqb1#Z$h8rNw-Gss3D}LV5#CO@;-+24
z?ssQ$&g_~fuYei9l*zS=1rQg5weL+0euDj{R9d%+8xlg8`$Bw%=Pcw<s52|vP-wUF
z(-B>WdLrJfp#3vIBI)23c_5_=Rz%&DB(FAZ?!=$o2oeN`IL}xrV2{1@V`CSWcO-e}
zXI_x)sb$O2vqs(cQ%t4eH4tJ>d0snOhsG(fF_dIj9{Tr{Df{cn#QuH-r&BD=^tDTh
zP&X2zUJ*P-JBUuc2a}AEJpYgMF-FwfrAHaqW8(frst3#zV;OR{!u*$==%viAUgm>9
z4O22JR9wnV#3!?0G8E0SVg=CF%CAdi_JWhHnBP-zuR_ns_K$K>O>WGt-tABxw9_$Z
zsqb8wKcQ5%T#5c{ydldE$+)&cakqe2a%-;JZS0O@JtL;6I2Oj%(kl063HsWnPq?@v
z+BB!Pyp*M!xV@u?lQK3_7meV47r_8HJ@Gc7<}fbm4S#z9%F<MXU}Q<YNH^bPH%=P5
zdDu8iMVYU^p9Zc!jGv?dTYk&6N6UF=@%g+rYsI0`8X-Abg8?{Gz}TvPg&u&$OnuxM
zd@J1}QgQ8qM_f3XAY!!aq(hb8o_awLVl9Zq<#z+kK`X1Lp8ruMf8a2-e&N5&Ljs`!
zHL2d%>Esx%uQ9P$&e12?3nX(~h}n*aLcP>6%Q>aOMY0ot4Gom+aLrGw00#JU(+~5P
z0-wIf=VvLpMK271wOZP+_p$}`+`N+(64tr(t(SO-RT+?wr~jllKpxosTw05*$ov5z
zsC1Y&Y$70~<uE~#%+4Gvn3GClY98n7#yIZc!L7-w1-qG;l$%0w{5Gj5Ba~0BHxxJG
zCh%NzZ1CNuPT5go>e}^$=mhLkS-yJ#&#BT+OfEJJVvAN`OnV49cZu=$`3~d5cg+^g
znQfmZH{+Qs9iJ+qF368gFj*J#;@ubxY(OVFX<dTDpR=MHa1X}@qXLon_=2lN=9fjE
z-?bfl5_w|<Pvn~5^g;kGi2n~vl1VVp)#SiOJ_?5_D&Gf*XT^JC<_@uvkqQ0bXeUF-
z46lis+b(qjj`mm07s5BaAW%GF#}&#-Lbg2VXbpwhS6}Fp2)XGfcyjwPbW3#}3>XW1
zkfxI!>hPmvH|4gN{Lo~|RDe8iv<8uqRB`0(lRGbCo&&ZRLsWp^Gy9WP;~@Wq`$Vz_
z$Z0kRMWGllt=S%0e|}zDjz4YdV;>)aJ5lSVJ%MgQ@5?Sv+3V5npp_8IBE|nzS^#t*
zD}TB$L63Kqw!%e2#VodV1i<R#cRfxpD2>Z-Z9&ULLE>Z&1b|q{;^a54a1HDS55X_L
zQe9iDNf(-5{3Ormauh@&AGhm<C+^8KSL_m)qY1L3tV`X@nDac{`F78~y-C;<b}nc=
zMMP;4HQ);~*C%aS^))La5K{Tb>=^HoXCvC=&v_Ng?(IeC0{)8yk6mD!4lXsbodE-6
zqkwM)hU@dQU^6Y^DpfAr(EB;2#Y1$D7+jS0IV)YP+|~GIiT9-Ui=bv4&&*{h7VDr9
z91k5!kyx-wN{lH)(g69WmBvc&l*pw1ny%@~0)A+3{1ZpP$P4FYoMN{7BhVK7x??Dm
z-pWlV>-x?X@b#>e%(%}OeMq-}J8IlJk-a4wC}M2Pu@^+eRIPN9!ioxU<YA`_KlH|i
z-`r!6dFq<oBVVX<pCQUESxC?DjVJRti8{8g(tauAfQ#yOmmbp^h+5_YdXet^s&t?R
z{ymr6D*SbZ3e!qfa=UWxVhe`7TB?l5iubTRt7dA;+Jkier|Rk>b>nR0zkHd7XslF{
zU*;-LBjHYNb@4g_N>!ths?w6W(M<Q|V<=Suz18^blxRy&g=_}+O8$#4Pz2aR8KLs(
zJ8?iC;$OrZ5VCLVUxfZ7rL4&|d&@$-0+S)y=dl&LUa7Ig|4AeDoN-%E&SDw__&6t1
zA$`Q<E65u0dUSAx<>c<z3SLFuzAF|W$hy~__JZea6#Tu%W?kmryxit~)1!zvAtct&
zqbSkm*vib@B}VPp+Pi2@yOuPm5N()1w+O>R9&lWj+V~o@5LYw^){=W{qli4Ovoqkw
z%*E!;yF7P)v8Z4ynZGzkw|?x*a>0tXDT~wC*nZiycSef=Xf#Fk$Et&`i}}T}o`>W<
zKc~QvA6gUv^j4$i6-&SgoYH~~2H(la7K{jL3VhHAaL{eHDT9wN74V9?l3D{fuD=j;
zD-;oal>Tq43x_P*&u7f*=rm*jf5*14YEt8sg6X=<tAmo3^1(H;S>@kb=)XOgzqio;
zKJo$@0tIbvCctNXwNG++iB6wjvYQ%cW@xkQv?RmDV@=%@INh{Fci$hz8Qhv#*O5)>
zgh%hbe58NVj<~Gt)VF~IVwojR{zW&mxjjFv=V<$BzN4Om@lm_H=~^qksaS942fCiJ
zANwmx`KBAD+c9I{99x#i$G6?LXTcuYUC;;Rw4FPSy&?@=>jJm~)#>*Uo_tyQd^N&^
z=k%e@ulH~~q?0blEU}@%w_%{X(<;H}*qL1eq$5T14#~#OG2k)7QyH&KbCP?S+l%JX
zW8$nuPdE{&Zl2q?HDa@-9xoHHrw%NASgD#vOmAlkiIxby-fom+_Kw~M)F|{+!NVB(
z(X$G(d)I@_&WfqQ>{rS9ftIm7tr#;#qIW}AiAngVJg3m<YQr5%3uY`+nE_Z#pQxtw
z%i}Hi#g5Q_oL7r9e(?Y(g;lg=RH{{ye|Rx~2Cl&LD<_0fj{vYKf<rz#5(o5ujmL_3
zf4yc8_t$0(C?0OjE)J>ND8GuyZJC_0-k<Q871>)H@KCb)huPRq8Qq_c+TMLI$Lw|;
z>nZPyuNotCGlot#M39e@*peM31t{o<h^7JvvsBKp2Tf;ROgF{ThyrKD-MpJd4;L=&
z;Oo+y;~><YxcN<+`H=}sr*UF2<gEv-M?>wp!s|fnd_@FOT7`Ia#}#lu;0uDVkh-w1
z(qdr#|7pEH>Z#hvM_4np=Ir47oB~kFuba(I(jTu8dI0_Oyy?YsL*aAg$I@9nKKa+J
z;AHlyO}Hd1WPclXPQ_W?Bmb!_pJbf(*er$h(AFKnv>pw(>$+B}nEf;W59Cx5)e03Q
z>sSXXwS0zpIL+Y|M+Z3|)}th5Mf_E6GoIih13agN5K^Us3&1Ax5;O<Irxr~jukrL4
ziZjjtnNRkH1(R~R9|BQu?2fVb(w<Uz40Zt|f;4kUk4RUQ*agox)865$4gs2cFiM%#
zb>l@bbv+>m$v<k<$P7RTsleC1vs8pasoOaa9Y5wv4`O|({@QgKp8xd?o;lcFLz2+%
zZ|{_zz@;9;Frc<0f;{!_w^Su;6Pr^uTnzpuMT0mq>-QC5-ADf-MgR9r4mi-cI<*&z
zia?sr@WLV(3^)(WbuhSDH0H4ez3o++#`u#MG`EM{CB2IU?u*_f#0p&FPCTBNKfsW?
z(rO+uwlSBuk)KM|3VJD?PuIpQcQzb(X1W=cbQASFii+>?i}V$yCE0d_7pQ&cdW>K-
zBZBpFetnc@mQK9!h!ZD^M_y0%r&Ijnrjk`lv3p(J;GfxK0F=99He+LkcvDgue|nax
zQKB$Ex-3pR82Hw&f~m*WctS_Oh}SZnd)5HCzLMY2u#odwv*FXA%i2ba$J$dHx7<s)
zbU<>9o@aBqTM@%}a5FQXpaLri`NoEw9M3cLdfqWqyZL<|426$)ql$RR6y@nn*o8z4
z(J@!F*q)E1`WxJU2U35gBoR_FT=I{>?j{Gb-0z`zOJSMwH80a%sqiopLGF#R);X}f
zCL3-yU1h!ry9s71ntLGrKuG%UmyY_^OV_bQEHL|M=6i6o_JYxoy~XP9O`rS{OQa_G
z_7(@1e|~|#C0TzT-ya0VwE#v8SV;U*7{3Dr7x=m^k{UKKjZczzwXMorhRE_9mD}Q>
zO3*&;{PDs%d8CtMK5`uHR?NRdWQif*(k&;u_w<c^;1F|b`W`5uqaB`(jL+iMDt!8O
zif6}Iz^lfBM7(+uG8oWe$$f_LWjQcv3X<Z5D)Bz+2lPFA{TC&OC}%^x=iX}t$G2Dz
zs*6wAfwI`-<@pVR%j_A%uXBa*3D}tuPUAkf;m5{?toHC@AXl@cv6~lGvuY6UG%yQ9
zxZ4uJLi-gSZz0+&#Ir+?EF##orI!d!*ob~IDaEy)3ClsI^_vGmKLS)-l{1*(C3a)b
zafHrR%V4Om)B1+bnLJ)98hj>u=CB?`H1{PCH#(A9twNzpqI@H)owiYPj&wlI+^8t~
zQ>W#foUiATc3_3fbqFR&YL6+hFcR2}s!!n&p-I`%`?pb0_sTaKzwgTME`^`Mxr?6_
zt?D)<kuP^W1bFz;(U@j%ziQ+z;XiAiy~OpuEqM4H|4nnnR>ab_>EPx{I>pp;KOzjw
zxfs!j!Q|sw>R}I>2%T5_kCU4oK#DZh0}^;%z3+4`UV43cwxKTa>H^zWOj3U(;AK||
zaG$-~qY{u%F-ULQN8A-kAhGkr5^h>FziYt)e1yZ!p5vZNu4ZQ(IEm@DVj@-*F3_#O
zqXloOI^IBhr4aX^95#SVcEZSeZ#j9u7R*~%Fa&S*KIMe>{S1!?ie`KtPiecRAbfn*
zu_QM%<|hW#z1h{ADlbC%e580vxvrK1q7Ga+(|&DPA3!8hAfa?ff)1=iB#+6qwU?%y
zAL_aMEz}dtB9Gj-1>vRQWdJj*;oS2rgwWkHTvkw?hsTl$*idgl$OX`3U#SLhwt!C+
zJfidln4vk-FE5t+5y%3XmFT?V4MP2k0M4bc;rI3Tx%K$Z?RCf-mPmB!v3FdfS+m5?
z6jN3~jsGAa7!Ew<i+}JB#NZ3yXB_%=;R-~OR!a5Vsa4<G?VDv<(@NvqZ1t&QoR{m7
zJaStP!XrJk3`a~<%wEWD-frkrjdxUku(^Hlt>BTJjlPxgeEXua#|<(LfNB@kWYur}
z1@O_4voNsVZfMAJiH*qb3m7mn^puF-mhmqlMw?cBE2mn%aC1|o*{%(L+bmUrl8+la
zT`;0?vV9~k_5cU9u1IJ~ie)WHxDijJ;t>CkH;-r1F-9$GmSXr#gGx`u3az3y;;QKG
zKCeJp4uQ}zfL+X~9dxc7pMdO~&xgkoJ;%qy-L;EcHi#goY1`+8L|qS4vy9I>EH+)E
zvB6#tkiAzt4(G!!4yekR{ELQDt8-|=mI&v_R>`N+fSP0R`kerTJ{$SCA0e}n+nKfw
z?5Y(V1I;`zekL#(j>d)p5&;4>ckjVdU=~Y@`D;40JN9$yR>XZ+R_uYILSSMUlWL2`
z*kz-xTCBvvZ-cnueYU9oL}LC%tH_37rBY^U=UaUlF8zcBZ$H%7ySvMMSYoVhTs#Ae
zE)->)XX&ULPcxtskNn8>pKOXrI;<H3wVf!FbmTJHAs4KXL6<IiO0z<5FC*(>@ct+T
zKi8-2Gng~eHDHS!-{*##R3Bce9WR3s^ds(NFvR5`!Aj|C#G`paS(ItPDW|(04`@C-
zfUOa}FE-~}>+{d#%&=qO)Qx#niRta=`q|3s-o_p%Uu${Z{PA%UHMXJNGL?#W)jI+c
zr{ovJ3%dHmOF8OodeNXmN2W&JF`{nnxM=T^f0Jh2a@Yo5x3%qG1eLgSvUJ>1Ji<|{
zRsDz6Y6*Y|aIYstDj0vEjU-5!!C?Fb@+c1FAw!s`g>6WePtgc~=D@qwKAD|8`^k}p
zRF~e|L)c$lfF2888a;{(gytY6kO@-)RTUu5a=X&lD!mObr@d2<y~O5BH>J;1F2Dpd
zo|lb5MZa%d?3?lTv4tP|IaEBQoS2D~$=)g=N(r%Y`x>xMc<)nqMP;i0Us8k0#q|K;
z0+!#h;Cyv24-)WM9XSo)>Y3*N6*E$COi0O{De~H5Lo0Q}^Q4aa&5y*l2rdTeWq~v1
z`K4&yx)PC3ESQz81;9+D51;omMyhJI+}v?FyOyCH2MozERXHnr>S+*zHOTfOxW}@{
zD|f)YL7MRBoDC|=yzVkM*eh09a89ll?Yjft!<ncyPnL|2sOq?;jg+<A{unAmUEAz(
zUal8#pXHj5_ll$4i}xYK<5n+r)t|6RM%vr9y4~J+fhrpC1e8>eA8{_&yN^!kn~O~C
z=gI_j)4A{ZP}Vg=mJRhpfGss~UH3@HmR{ey0c8!I`F<xIYqExtg&ICyeR;aMl5(_M
zVE>#h1!=gND*R>y`f&=)3DO?Ew;kM<7sJV|?$F+WRI8PxXM2*aYf1;s$T95*+@Txh
z&ITt4(_Tg%>|;TQ70~N+LA`#PFrV(u{ra$UBb-6Q#nbi4#gji8yV6Q?=ZVbltVVn#
zg<;fv1*89c)e_q+w^;cTfd6yVvP0h{kyXK?;!cnG7?TQ2YnWt{RK5-2mCMwP1b^PW
z#9EY4uWEu`w1KwsEHZPjRf8gba!1%c8QHxy%RHJ}9pWLVl*YFsK+Gfp4Wb<#(WY;8
z%(+>5==r(1maN(zV0z(w=aZO6Sy|;J?oy(F?_9Xr;%HA%N>q|9SHC*+Ts}Og3}6-o
zfN@2_SPm*ayP3FxAD_LQLI^N{!6J<aWPZ=J$6U~{%}@_{o39`TDJ&EK^;PvXto3iz
z9-kJa9_ke_fVBvWxvb#%q0j=u)oa)%ruX`-bIjZXVTGz}H(h-|Z&~cvRf*lJP=eX#
zNg9Cv5X%<Vqd<6zEYCtV*|XHd*Q%P+O5Zm<RQ*wYY)qhcVW9#847s;Qo)@rClrT6$
zf(i03Zv{03E3iBaPpTKdXbazjg6fWJIJtWKY{y?NKl=7xVi2X}0<$V!za~)CMMXpq
zYAtsLgFk<k@!XJN<KtCZlK;=0%71&Jenahj-U52Xk5|bq5$JMtQJ==Nu@SZ40?51$
z2t*>6%OKNweE+h)K0oGz%^97*EQ2v~gQ@)5?jR=16^HPN?FTyN5^ksPK;mS`w-WN+
zo3O-21;#rXmd2lSS}7%tFHWzHIevK*W}x$m?^Wq3Pqpvwg(}-Rz;sELJsJ)1#0|YX
z%{8Da#)8Sio>RUjY{&Y#Pn~-`bwrLtly+04>1yH%u@I*AaNVrFa-cFlFWzz#4*+4O
z+-Hj)J~eP^NuZa)zJ~C_)2>;s6{JHXL1v({i4R{Vh}J-QHWSODv9_(vvqy{)HE^z?
zZwy<z*KM-C@ecvXOI;ia)otB!1&W^<Z$SzBA>?lcdZ3ul982Xl%*}=Tl0ecP5dTs8
z>u5@6*RERDz8m9b61<h@LXRm;?vZz)?RrACBQT}KR4*`0UYC=ZQ0B*|3WbWJg#S1+
zVd`JS#DrOuPZvTLcAEoiFZGQ|%$*SRfWMl!D9EexH)HB=TY70zgG1~?w|{Za{g+$W
z|5b%Aw;jG#*wScDMzFICNF7s)PsR?)QXE4qA60Qa<(!p%o3!iTk^!r#NzMw+HWnew
zQHrE(!9V1E0hWlvCN_3lRMt(VY7l$XuMw%!9<q6s(>XxkMAex8b}oNft^On1<=ROk
zupeJJ#=hU+z4|7;vTx^1evW?YJGR-4YwK}OE%{+QENH!y&)`EdSw3IYDAa|NaY|R8
zyL?fMAs$?*^-Dlztm-E)F}aox9AqnvJNO$$K{YLvggqG$A|M*A_9-9RZ}!$DgfPf{
zhS-{0A*MW2c^Oi^d^Pu+w<&C02Hq90#ioWFR0E_c0{FB{kptzxA|rQ1mtI$qgDTc%
zg6GLOucJR`YP24sv)P!%*;z2m#w3ClA8FO;Z!21nfMA_007>kNiOr?ds&HkIUTbK;
z$DAXA;qDZXj)d&bBclBEK*6K3=MLCXmpq@$Ra$}ow8_c+Z;T6Ef1c0(#sJH`0ZX_I
z#$FYwvSnE8&{+n|M{aHwa}-#VEHhaRt5CpsA4Q2fq}<3b&4suguX@s=n&$ulAG|XK
zdu60fiX}eM(~NR`A&hfRoh{NGmk%TeFdFQr1XYG%=hQk8s*c8*32f;2$DI~-HZJZJ
z+VdYP^;`3J*h`T;GIQHb5iw5C-t!$pr409}^zT>juf6Y#)uN6_>o+Fy)HOL0jIOTB
zzKm`%GN_wUEh=fZH-tQ$8`9W#YF!X@(r!fpd`7uK9Nfy&xC+WU>PjGj1lXH%5^sw!
zjoQ*BBzeK1_yptiL>BBN2#d%dJH!CFG3$v!6~kh}fn4bn@L}FKSiDAlJo|=UpbZ!a
zoi+o(VuU0wvWMDJq=(63B%VodRsD(^htrAdO#ud-NRcVhY)Pth)9Mpn53fEy9hQfH
z>94jQ+0dX8rJw${*<tL|&v}yJ@xyt<CMf(`KKl%HY&NPS+I=Vz&6nO79TMb?Gt)(<
zefTGuCC*nAx>fROBhEeXnU~e_-NRh5hK7wiX@`bZhCGX9#|X?Q5v|I>>r2g0#q{We
zX<~Z+cb=f;&kPuf@HzNB(+A1-95fQcl1zf~?p*^xw!1!)oZ_Ar-x7LNg<yL<oOj($
z7~y7?64bFTU&kf%XZM~FTb-J7;g#NsO}a9D@a5o6s|K7+X(~5Cd=#rVHj$uAaW>oC
z9Ovd^CkK_Kgd453qdSmXBl_kJqQyKXqmf%8s*l0{-rKDx@Sg1L$OA4?gP?M}m*jf*
zqJ_G=fYKo6EbKB1F!-IbSRV^|w+IUHGNraRT!QiVM}we*uyuG~at((sJW4{|+){__
z1w$v7*6uI$rms-Oi8<RBfLNzVqf7ii#>Hk+Cw)UY7qY~4LnM<|5y07i>W0JHrsy=(
zaE)p>=^|w&5W{7eL#p-Z88L6baESs82Te5xvWB^<5lo(1(l;ERj&KGO1*TZNwT1nX
zvZo0%Kd$L^IaKW#c#i(=sQ{nj=SV3x_?Mu=I~yf%*7Jj-z{P-1(S+!OB=0BOyh9T-
zxBjo^OX)v#!QeKm@D>E2a2)9zvH{%OvXnK^CYmRzSFF4aFqL3MNAG-*{yY)<#&%^t
zVs2s0O6Bjn4|Us9w+zPS`BcbA7++I^3lzNEXu=s4<r1Ucy9{1}Wca#w2bV>DM%Jyr
z27_EPFa(?h*iw*PbsJ*blqI%)oeN6zESeNK(Y6m1JgOqyk~>lZB~hk?)Dcu@bb^#~
zS9qWT73jiphx3Urc$mCHypM_6mr6@pis2Dgeh7;b<mufw_FUK2;{eSOZeF1w-m=j?
zL;3><|ETgcUzpqSN5nZ^Qykp@Q0{~;8zx;D1nUVpkwI>VC-V09_ZUI}ZDUth|K#GJ
z=*k%s%C|L97=BOx28R|*zYDA&$*lmU*^t7TxkCO&_7iL<qu&w56Et&&L_~k#Z6+&H
zgyIs}owL_L9Il)H6K=2?lXIl6%c2GL=_33TtI8o5^CEa8WzEvCe|2EtrN5>XyoY0@
zRzfCr%v^2T?(RM^p_{IwtG~30kO2Afy8Qb&`xBY?CuJov5%_>D+6_Qs#hr3X(+6At
z-@OwOat(t~*=bq#_`5%xv=|^6mFKh@YNhBhUVubU&GJvElshnE=4I+T2ir?<!j3W%
zdHzL&dlch5K^3u-1tqfeGQQ6sr-?!J00J@YQJ5VWJW18)a&h?)|Iyct!xtu~n+Bok
z8IX-j=kQ|Q_6^Ar2<Nmxl{Tnx-)w`L30p)UWg-rEsF7gpqVU#QvxRRA(dPhmk+9n>
z<KOjB4{r$kxVI4-vGxAunipYT+P%lSJE2yp(ypmtyjoSaRDKaS9XenlzL`P_jcW|i
z1BYMSWOPQT@$t2nXBIgjM;MyNlCo*Y1Mep$cs)Lr)c{r^S;OR?Ro;Ms5*0Wm(z%mA
zx!SX7W9I>}s3dtaEJ^ZykG)@f@HrCT`U@K7B+Y16f(K52aYrrxxTZ~UXy@bk+Zr)8
z@T;UZ0=+WvXQ^nG?Njx5*^O?!DGtJDW@l-beEh+0z`(>clm7+&>oe=~Hfi@N6rX>+
z)Se+&5+o5|(sxr+FuUh=H{f6rZI#=`;j={xgKT&Z%UZs;)qd~Z6XqgDhlRBEn&asE
z!qFVhhGT~cE&4qS7uUDB8Hp~Fsce0+DH9gy6ys$tlS_I-@?I0qe+=APvE@=yyT19{
zdW5<WuZztP6XS9i+u7IvlK#o$f;>C4x7uS|6pHWl^(xADN-#n1QZ5M=<i{^dbz>$I
z{Hb5F4?O7JddF@hfMkhJN;XBs8@r*>P><82U4ov?XplJ|uV7@~*-B~DRti&ct)w#v
z!Y|-9O?KbqM8(bOPG|%1FpZ7<2kaBVWGG!it)vbtzx(ZE2%mou>*7@RM>01w+M;k9
zp$7PxXAfZQ`+D?D4D9c@AGRShPqb%T4mro<q1krH?+Cbd-bBBmkks-9Bab}pb%wXk
zW)Sk50^gdxRFh^a*Iu6>&dnYh2IdLh49wr(7v(P-ZMIPIGUm>l3J^=h5Ko!9JVPz3
z0$UL++-Bg)$`=3MwxVycaagau&pXW(t@5(5Sg$JFPkq#F|NHUnVdZk_VP<er!qZT>
zTl);;t<iuVP2)WxyiIOJ!AadX)7zt;9JQ^_b=tE$udWxrA#F(9k>V0AeL870l}B+$
zkOErO78N#Y`mP<#7bJdm^ztPG<o3=I`t|JEN|{y6JIKqv)~hQ!v}e!}#t|Ex#pi6q
zYF(=Ik+ygQgSGVC(+?E6SU;XeDou_gK;D-qPA+9vh1qaKo<gns(C#i`aKgd+%v7&c
zUQ{-Q;#~wjPJY_1>E`b~jm%Iq=c6P>;Ii6oA5rPKslvf6xzcDkW^V4Zn1G#IGn#8o
zr<MHR;y82*uY?4t!w$A<{qfsiev#3HUeA;86}-1G^PM<8_PtyyywbXzm-M0>uo*3I
zjT=P9Bpj>Z;0D>VH{R-ioIhzM&-)nkn@5k$12`dkC8#z+<^t#a2=q(Mb0nN{01=!~
z5R#4H{^PT1`}tWdUjm%^Yc^1jK47^+F<E>a@tBiW&HKrflmCxlER#(Q8^$Z%o(&Y^
z!b*l7D$wB1W)VqPV0xjbL6gCGpA7fQy?uPlB?MQtqzYN9T4ANcSq?5~B={p2S<K(M
z$sXk3Ww|mvylYjD%at5lnmPw=W--9VuvqZ`c>}66mF+2f7eWM$i7It@dpN}Ru$+!)
z$ofD^b{fVsFCK*HSb^gTi@u-xN!%`jO^nl|IAB!*40S8+`Qm0flM9)j2jq`#vYI>d
z0Np0kV8wC0D|BsJgP`8fAYmcvDERUfx>-0Mi*0QlA2VNMOm@7QEHs>`ROlm;k!uRY
z-E1TPsB$+$fmjI_81$>peaw?|)Xx*z8V3XTy}f&8u3r0l91VsG)>OKBAPAaWj>4J(
z<k21xU-MtqdYnS$NwrQ$I#<OcDd)>y>3edO=)b^uhweBfb<Yv-VAXO%E(oPV%s}Yr
z%8>8Q%-PAm3l6tm?7glP&@In^UU@>L@dgh&<%(57-;n98^)2`JWX|IS@(se@|2-k|
zzd0^>_i}-v%K?PG$s9(DwbFPsL||5kf-3jGt99JmqiD5g`E5(&Q{!ryFE8SCFFq)r
zu$yimf{{X_U1-Ci`*cGjeV<F$?a~QFUZ-o{d-B9wp0BM%DcU9~@ktx%<w$RP#+|~{
zcaCgQ{?PwWyg{%+8f_60`py{BHd@&@2wW){AJm`8izey!v0K#yp%Zak@wD^fde**9
zfL}RXx4;9R#`zby3Fwd&0;DKm+yYWNL@n2AngU|FJ!xEIMY+l!8h6ccsEizkLYv*v
z5KhvIAJHE&Qu~2k#=%R+Qq%!}&o@T<JFR`4d1qF&L1uE%s7~I}1!Jmgn#b}<ON2}n
zxR2g2c@0l+gNJS%+eOMy^W_jSW2zFeIp!YPw@M(8N~=pDG1lwpbHWrmq$-&S9YAiS
zM<Ah6hyP`S{j4nibLl}>19XNShwI_xp%X@=TkXE?$whnEYu!wlwXN`5L5-G=uc!nV
ziT;^G`#(4<<Bo_!D&lX>3a)=d5!nU?{h4>r4Gm@jK=-foW4ztB()rOLF8c?Cf5sIT
z_Y?v)+d@_!iwi1o1j}SeF;)jf(vowX%m?2xneTZQcIS(6LCw^Tx9Hn<HqMHZJyf_S
z8V?I~$$w4Qob`}4dL_={*HmNBf$s3S<P4Wy8QIcbtJlgyG?l?@-a&_Wpahr8e$0>R
zk$dIT_|kL8x-L`;X3BQqxRZ`{(IqOe8DN!(N}Hr<)caOc{PH68mMR!hbr?MskCA5H
z8zcG~HU;(}<{j;dTURR!TXlWCb|?A=tfkbQDNJ{EpVz$bjAgrs^<+e$aIIcdt}fPI
zLIq2E6E9t6Bt8@}x6J!gjJQRki-laVVhESb3v%-ZmD(iPPm)6YU}PX9AW<zrFIh3y
zKP4N84B3P0B>|dVGC6ejqCblsK(J7MrTqfv%qOaas9<qP%FnPQ-jQ0jU?2(xViuF|
z_e0hCYZi6k`C&c33SO&JTPrIUnGC@sU~Qdcie#yeaL%;)_$Sd)`M<DMc7Ja8_wPV^
z4etac<M_DwelT&WX-3H?xvRtW+cWqbu7}ND-<RXg$|Or-Tv5JsY4Sw$wdt<(DgBM}
zr0=gx_q&~1QT&$Ax09D@;xxecW>2VVc|1mS*zToWLqo!iE3OO2<hCYO$7YT-7HYJ;
zXYifheISS5I+niO>Y7929PcJNFOvsMC{e&-7<wL58hv;o=XbeGk@j?EpA56R6W`tq
zsaI;O(A1PgNe-k=JWn$DT(foh5<LsU!xT7)mA_D<%!*?|Qa@8t!L8hyEY<?Z-F_BK
zS1PCH<O74^hDP9b8B}OPD2VD$8p+s)Vb6K4#uEb?8!}t3X*jg}NZ1G!D)DiY4gS0o
z@cc}bC%fj2>o-zju29{YZU7K^^5IM&X?LK?0DsuLZ=ZoytctXPWHY~7Ke?j9pz8?@
zu#sgBv&+E;l;FJYor6IeeN#|=!z`K;aw#36^H(gU{O62c)dgMMp}d!*B@864SXeH0
ztZ8e9RYf1u9)jD3`^ds&PO=^R&%l-6h?NV<ummlak@ZR0pntQy_#g+}1S4wAFM;lB
z`kCc$lM=8U(KIftpvhW!A9Ztb*BjhoO0e?~zEgJ>oywPR%yvV9Lwu4Jnz+n#PqKg^
zgO;@{*HJ?=Uup6a0(;h=zC|4_fHa!!-pWu^FfYOBF}E)Y%3jbr9}$t`Br-&)SB4r4
zBx`3rl6`HxcKYsK*8cave)`O)L_0GFkkTg`omD^HW88t(d7;~pevCL$aEba7bv|y_
z({5nd=|@4iwWZ~#>UEg+GxG~uNnhtaeooO(UqJR^CIO_Q@Fvw7&er6I=#yDg5zsXH
z%38TKvmV=VdTQ3EfX^4~N1)LI1l_tmRMEV27m~W3@Xa`UF(ctCeZKYe4h)zF8;p-#
z7jJC%21LPuy!2pVB@)!Yax2kN>hjqOdLrqUo5HabN|;w>fR|X4n7o<q&%#}j4X_U+
zLDNl|(yG-QR030D7qhJ+<zOdBKuexMJ_#fm1<jW({`I`MU;o)p?}h`MNkMeg{{C@p
zicu3`SOMA2PAEt3gZI(IC7%sBt*wsX!q%UunLlB`M#IxD{<Lxa+iu0)*vOlT?-hFG
z3@~!4G=e`q9*~G^xFxLG1i!0p=+kb0>rPZ9f_M=0fIfKOS{B8)O{h?46Zu``=eP~K
zc)kbau6GxU_-I)yY=ru{&y4jxe>)G6G>b;_QI{9EgOD&u<0>D+@+k`Qbxr$p9@UUM
zVhnLA?-ulEc<AUw!!z^H(MfH6d+J7F?SM8X&O(X<v|)W4ky9Y~l1Tm8>V<m7bte$v
za`WOkxs~7+o=fY9R!qrc#L?r_ZTDbZW>jmZFF`!_QcDpMPKeSnncTMQtghD+v+lxn
zTw0To?oY6`<3sQB5B8!8hRS4>VI{&|hB^7(Vo=tH?U@>hjLl<1)`Rbnz|niNee`?;
z$kw^u&_IasdOVz~r6~hS5*59dWVP4FVq$lbtM!?TwY-$o;XbeYd_zV&{eV0}uZFE!
zD3Go2l>s*_S)4j4E{TQ0*(pdHwWuE8V@>L({l~{>{i~IxQxoKc2;tRL^0~%WRhkRp
zq*d^{&5xcYF9&=`N%)6~|2IfrEGL4__-EwUPqi;U;@zniW<nIbjF=tHarRB6Q6wfx
zQi|`bT!4n4C>_{Rwx1U?ToLn(Q%Os>jeWVII8b4U^bjC?!$)wj_hpMdaGUzriKAE+
zsuMT3ybdPkbD`IbF(@wcsv)mkOwt@p+s&(`d!QpG)#+Q6*sUpB%I(KklxK{1W=Zi0
zYV<BZxvoK^42v<3k$B?s*@@|fwcBQ^&PtQ*>UK0A6?h%F-rw8%XeKtKAn@I28rSp0
z4Ho8w6?mpeo1|6Iyq=h6)oMGH#qq;qr}&7SgwtT@&X<5~pJt<DB@qw32O={AT3u+C
zxgFyYiZ>3CINRed9Gd|NP9yeI7_#@hY2SrKz@a!VTXOw@?$D8v%Dg8S?3F;X@E9yu
z5r(*W#}*!IXi#RxCW(`Y%a0{hvt=94b6JsAhKwjN2{NpflNgil=`i)303i}b#4~To
z8Twad53d!yS~(b{e}45j0@hN+_y>Q*Qs`GBpzyoAbQO)R^LlWwqslzFp_R#HZG2=x
z+WoZD$UkN?TJH%@ehHSv{_NfU2O9Y2D|kG4jh*?(l<wQ<j#r5wk+C&eNM)-jJoU<>
zZf+q5vLDx}S{M$R)s;P(r#aDn<=Z~|S9ko2HU?W<)3+Jk<vpkut>5xTu}#gXe2F~$
zVqL;Q)J08VeJ8t~^;N0S$HfBv^JdcohG1k3{?;7<THV&sSaJTAeoj_+2IOV2=~C~l
z#*XQRs}@Z{_!uCyu{GmsJFHQRy~0vL_8~x<h>OPsJHGI~EeCar$`R3-zk=^6!_{G*
z*Pd4==*#9~ZBrL)OgF*hNhYlfvriMFb2;0J1jy}AZ%#d$f8SWsh4O(`zP~+)Y^b2a
zpzY4!qDsH?l!Za8GYSEYadqh6jL01&Cd?CgAdh15cs!B}bfU~aL;&+U$LygyOj2{Q
z8sSIab!F&?F!O-A_fhN#1m}@^lfM1P4O!H6p&O<eeutmpOWyU`I@qf^=-AIi|CRuA
zSQh>|xQ`tnjighI8h(u1LmeAc*6^wyx+{_Sy-X9fjW@REcF!dA2XOIdH9h<PmpI(0
z{IioQ^gMY-T~nX6JzNZsW>w+)41R*q;2Kp6m$MwakCBjM4ool87Kj)$5+e4hb6gG4
zJ=?PAHn6&-yWwQWT*6|uu`xbYa#iY!P=uXKOLUN3j!WVO6UOQL`%{=c+{zN-;82dS
zXA===yh3t=W6!CTYvlX2aI=S&_a3hw3>GpCnT-ZLVknMI0!q~NPwHIEecOm&AOfIj
zRM}MbFDh+$v{ja5($KI+9@ls_^p@@H7;PH<K`;cD9b0S7#MS8e!udY`#E$>6r9^R`
z0YAi(@UC0;TH&c~jh9()r$H-JiOwPnurV*VaBN_+PiYi1e=%RPQKxDIaBD4@j(g3y
zu9_67zUtmKs6gsQ>t_idj!aPwk3p}I)UauZ4)VR#sYCdkXO|4DB<TTqoX7E*_9rRE
zxgkd=73H@4Hl4n3awi^4`W_Te7svg5XzsrbtypvaR5BeRLQ5XT;ItnZ(!4IAsn*;|
zBD%+dKdaXECh9my2fFeXc;(NRE?*ILbYi`T4C-zU6ecUc0+E1_yROohlC;Zo=mXT;
z=OU>xAqO_HXyAy~4lWv2A&<LyZ_h}GWy0tcqMtrS`R*Dpg$~+`mfn`FGNF0D?lbD`
zq@H4Si=Z$1CO$zbERKsCp2p77bcU7{w|n+rlh?(rbB7lxZtkq0nZK^ve4##*w=6$S
z9Z*_4RKD;SLS>l^yecnp11M=()Ti57C6T&w1c#=t1o7CW+T#sf1#PDgdN`)A6=~gw
z!eUAXc&~MwzR89GXy(WY^NgkFJex1GnxWAv1PSDr?3LKK1lvlbSX5Epk!1oa@ttFs
zo}KcCwMvKJSeVB$;m{Fr#to2AyPX~A2;Cj$p^^ihz}-sQ$N{*9Odk5ffOt;^VFq*w
z-8wno3;YJTm~)467zB@k$bw{$A}2tyBVhVcc;%#p2QWy$!Ky2*>n}64{O5>Y0I>{q
zHJt@ts44v!GYQA@+!W*INEss+tN)`N_!*%2AAtMctz7BYH^o^_ZU=faMvMbw(sIDn
z*@k4zwu$r<6Oj4m`15`m2LYM#<{~PeU{_Js?0U*C({K1(gc}N+B9!QlpLi*3M@Wdg
z+bryQTv-5SA#qKVOV(Jfx8<exmWVZVW8@7xF4IdhMh!haFUyd|Dm}Tcy!G{N5xcBs
z;!P_N-4Bs=g-*iXB|UoVvLx0-A=2*WjKdo0y0TU+cZ`BcU%WGc@A6Z$ziYFdV?Wn-
z-JFlOK5FDG3M9V8)|ic+Y-kI$#cokn2S_<Z+K&J}PVfFzh%_xcyLqRGHvQdfAq*cD
zi5@*?24f>W_}owzXIpWyD<^hhP6|p?t;pjXjf!~@Bs4DbZPa+rUD(CFiXXDm`cA_)
zfoA&r9ybpqvk4Q5bXCua9z(|~o;f4O;rW^Utb<77XP^KMXTa=OS{*MlsH@bOQ*vag
zlS>KrdOAa$c8CuCR`gdFPGv~_tA@OOSVPhbFJCy=yyu&Tq5aI#8L>=(OP$+Wmx_)`
z1V!YFf3@NNDf0d4)4cfafO{|KmJiaWA5n(TT)3XjEz9#G^)&g|B#D$6c`M5>lraiu
ztCS?AvJ?whtB=O-1e~c)bota-Wb1l&+2_2ftztJU+q3fO95D<1%v-DM^k^!7sg=2P
zbJ^?cA+Qx=6_b?irE1LkCMw-~cDBo{UTnf(hfg~!B%TRXL{t#qHdq#}{i5`YFDIKB
za-IXaV?vz>JoZl|BD;f2@rp>L4j@VBv6d;}(kQ3?+O?;BYpu0m(W2ksmCBrIejW6l
zzpIZ-39-`1y?GkCuBk3yL-N|>1%rroscCU;+{WzZ8pD?*QG6Zu5}z<XTUiQw7Dw9-
zmTuye`AhoH-W*>CD1PA$fH*XL>_LKbu~AudJ=}$YSaNX71t4HDe5=Q5PI@;MW`C-+
ze0BaN)SY%1?9@OpS#jebz`}J&Gf|R;`LXH#ps52*c(MmW9d$L|)rthYwcU-O%0ElU
z?0<c6qlvydXZA-g47Kbji<s3^7zN=YQ*9?nmVB)xuQaU~ic4ife_;NMh|U`9zdqY>
zyY#cWMb`zGCZYf<H3<ANc{Jl=?tE$F<FLmX3>XE>D{xL+&ID|K2FUqawVVfnxD585
z`#>vi-u$ML+4DtcoZ5$*H~A+`@DBC*D;W!G(#~sZ`Y6Go*=KM0)-^b}B*ZB08E69O
zHFbN7c0%pkE@KKvJmQkw2x<bSw~nB0zSe;bzy%-Y;nW!nQ5>7Y>I&VOO4Cg!m{(~U
z9jq9edK04BQ+G$)=ykg)+G@GzRp<)x{;f}j@kT9cb!KMATCB#5hFvBztE%q#!U$f?
zPc>|x8(K)kc@QeXHSW2NTN-e}ygLD^<yz-W(ax;b1;(B85MyZXl?UzdKm>B+CbuYh
zV09w9Yujr+^&Z(H5!VE8MP?>G5v%MPiUS7Qn`$L3wU+Zk$b&CTWUJ;yWDHF2Z#;ZF
z12{=H5)(pWo*o-k^2^HxR%z}rCBH>YFP34Y$Nc2R$34PjTv)5Mm<lvO(Rt)iBz=o%
zpf+k-yX%@CEyZDE!LMd+m+IHsrqdqMSisF?85a{e=Ob|G2+nwTSC#4p4ZCsFV)2bf
zQh#KS{dd310wrxD9U48W`m(?NoE+3wW4ggiH+Xi5PS8K!Kt@T&`;w1qi<3?2+{A{|
zcTv-V&hI=N-G)t<kHhog@8OZ-Ed`G}GtcDZ(}8ru7k0T~A_CHk4ra&Y@UIg<3pVKl
z`f!&CkL)x<gX1T^<;_xku4C;Q&6_iA!FlpUInXqU#_QWHJ6z^2*LJw%-h&R!o3dr}
zar%x@r9V+;5`$ku3gi#+9x77vbyD|Y@l!|&D?W(lWCU%-xO-CBM0_3uQfgC@pd>I|
zLO#y>>NL(PfvAQwe(a_SbM1(Oj9Hq6`IKXAf&SiLZ$JRT^6_%FYeq{XdWT3~1r<R=
z(~s<s0NE6U7L$wufT?S}QLe2WTU|>ImIW3+md8HIySs#{jbwiUNx;|d+`tJ<WEXey
z56MJu4xf{tHvQ@(#Xv~f-QxU>2$BCh^RI7g9DHNh-pNH8@RA1&qDs(xQCG#YW+XrQ
ztKiAY5xa>L)uS>0vs=s6gB|rDf9-Uy?g9)}X*afaZ_xV9AFAM7{da(EimW9c)6(>z
z0q#Gwh*#)v0K`v}m(@QkOFT~Mdwh4F;Be8tJZ7ZI<F@UUOJaH}7Z$sCHJt-8IXXvB
zb*%(%_sO`BggK8AH!z)vz~^|t%)o3i{@ecc3zyt0UMWd)gwg|F#Xa>;BD46sH5(|j
z(%ehLd<OFlS$9|5Li<prEzoB4y?s4=Y|x~efBlt84RGu&v<~PX6DRh*zPz;YA@?rD
zh}Xd&z8A-2gw=B|z@!y7jPBoQwK><(Z!Ng0ubuCGVjPvfdX~OXX!N4R%SgZ3MaW{Y
zdeqH2XH@Y<f8%1uh{xN#=4f<|%ehzI(4j+@Ad|?lnH6sC9Ge4jMJC{eIJC61f*(Kl
zL;(pxZ!NR>WHs9%y1TYuF?mx*`>^DA;=|Il$Xj&etwGIFMGbOltdjyO#zNrMO*C`a
zG)(z45s;|Hn*_{AGsgXYO-eZUenm_q#`^Bh=a9YMt(UDNO^Z1eTIIRs!}Cx>yQN?<
z8D%}OpZbI9pW)D8>{Y29iCT{TGiTVR5^UQ<WbX>z5_a?&$h^bory<A|P2E@>O|r{C
zZuai1)%e%%JhgcLnA$5><?W;Mh!MU8ASADKpj>c%JMF4T_fqPJSCB*y{*?2IBYy)l
zUvK@Ia$t#5V=NGbH1rJOd$x5ffF3x+XPL<Y{b;WH{%W{dXLhha+O?N;>K8`KvcD09
zgIZoE*_M7&g@+{y(Y%0_J$y6Z@u2(}w<JF+i?BVH;MpL3+*B9ZCBQ{p+m7mHpbO`{
zjjlPN0>f@MOT*$A=0NF_JyqR<i=Lo5bj@L5hdT?*8gW!)N`t?z*_SM5tlN6vZHvdj
zta<&Zmy9y74Z2+>@A!%V*_u0pd9jEPmw<1s?I6PkL99%8o+PKlbW!dxeJY&)lnIl#
z!rHMKKJpCk<OLljYSzIXgYYEo$|IXPOunjLK}Mth>KTfO^GT8vCrDs%*I+jizvRZ+
z;dkSv7n%N6JuV;lxh$Xx>M>iPQdQ~QQ4Ind|E9}=ver0Xl-3qxaQpcGeBAh$2gLvB
zabp-CD=iS9k$S)sn)r?dEm#duF462uPV`<XsG^&1@CpZMR=mQ)W%ZGPl3d)%?p=%J
zF220(cpk`xl62wwI86MWk$t?>crW_lHAiA(*bp~hQpFI)f<E}53b(wJG}m|7BulcI
zL5nu9)X*@S=zXmaDp|nIr{?I2Pnw%2V0^SH96O#BZ*(CQc=I@om7VjI8aL1Q&(}2O
zY=P@hMO{>T=9pZs#v{>0z9nc}oGH(GKgNHccx`4|xX|4>|7%=G(SmrV^~p&nnZvp_
zquYa8YlG|S6w}{}As&4$#X(&gHE%_EW07KGmPs5mao&^2)>r5d7bEX7qSpBy0DsAY
zTCGj^vIK0>MTF4!tE#K(*8l_mZg5P3m$s5R>=cv=)F1`bh$P8ShGk$M;~a+Lr#29<
z!8g${BIyE)CmCblB!~k*{96%j<Nf=Zq@(>Qb0ini+LGO0;@^9&^d$#!xs>zpdN+dy
zW;M#Ps<j{XXaByz{tJxBNj%!`h7SP1gsn=PHGe6h?OXm}NW)nQ31%sZVV+td1~5*k
zDxaP(6R3_38IJz?9T;q{hYj3MtPa&a1s?iKLHg79X(LD3IEU4Hp0ZqX>4j1Dw>s>I
znH}gnMPr(3egEgLkaz2$NL+mwvgH&;!`HR4G}^gA4Oy~NXg$fLEY{k1GE9$j`W&?U
zf<x7XPX}i5NvA1LaLp3yNXzj~vuGT0d~7p#sf#A<1-U-o$f<}zorBmZsax+jbT3X@
z#y)ht^rE@Wf=_HRDxL3+wG)%oP{{<w@5LK`X0PTRPH3FSa^bYea_kdMX)VtRw(QyU
z*N^syt%Wj+;w_zyN~1vY*_ANogv<Jzz&*z$Cbpia+Otx)13o&_dwDJQZUdHA@49HN
z;fp0%6F{AoBO}kuzIbDdAsOR!!vC6|0dob}c6K<LCXD1nIQQivA7|Nu&nvnTGAk&4
z_=i-kwgX45=HLUbflHUUFVJ(16A!i)4=TP1{C$cqX?~t!Dugkz^|Oz|GoxE25hwv(
zxJo0`|Fr%(<6F-2HoZSmyPJg*hlI<G7^P}XvhJfgLgD{0(E3x1kavIn>w^Tpb5^6=
zY$<THDDKOYHG(XbDveWB+t7<s>>m?vS2&M6l7`Q@tS73ZE$kKQTKMpkiY{ub8X)Y?
z5{cyFQW|CI0{Nb00_S%=%>S_2-e`HCstM!Dz6RXSl^69gkkcvhZEuh?Q$VMMo>Avr
zdjCx!7KxM;5$-i9a_QJxm4U;<h(f9{u(G$zH{(WlRo@h0LYsv2mQft1Q?R4djqbiP
z8}Y5c2Ur3&0Hq09Ol-ubtZ}rB2Cc~)qc9L{dpe_J-W|-gb~c_(?^YD=7__4Uid`8)
zXA|sDOBgJ629zOoF9**gazQ-T>>OMx9<*+Kz2z^8X4a&-o_UnY{}cImF%3cwz^rs@
zRNerJ8I3eak3zC#e-q%A0b3G#kAkP6q039P3d*TdTK5L;9<ENu56?(SLrm>=xk$e}
z`g3qh)v!4^R0(9@zPqEwOSANy<|sAomih@)kvDVg!TJ5iO5<G9noZ9kIG?DEW6{L6
zZ^P7K{Q{7x2OkEu9LL63^&s}X`{>hKtYc$K-2uoMx>4m<=*7|rAW2X$pZAkd!NV06
zbTEmOtcE`hM<X70p;SK81;85vLN=$~oq7<khQ$mm`*`jSN5c9OLA15kylcE_STlve
zD6jDwF|&-ag<Bu;eD(@;Ml1HEYf7(7Q1|5(j(#&re3du(xwT_7CoOg~gu^}ior>&2
zS(!3loY=-O_NOMo0^e5ubxHnE7q)$sicA}f`f8*2O@0AT)Kn2ZPm<=+FT&@$vW%M=
z5Q0k{0?-tmBQF!}E?&E;nt}F5ad=ugDO6sZ*_pu86$~03hq_OlA9Tm`y^cWVNFTj)
zwy*Q9GJC>ITmIcd&)&C}01pqf;BGAo_*aWB@0sEpJ|Qlzjfo>f_G#(m9w+&$3369u
z4~x(=i1RuZe1mYQB}QH8u~AnTB<rEMdDqa7kj?g{EMSDBMvKXl8G7rhG=1IV+$?z)
zG0XKg9-m&`=a+<)<HW&kAMpq1v|SfuFZYT^M=+<vc(qh6zq8`IK8a(9+n=ShP*Lqf
zedqaafy0aCdvCtX0!7dZH5trK9v@>OTQ%JkNQMXnKyXrss%YvTTLy2Ml$m+(vf!JK
zf@eA~iGq6sK(C#2;U&DA<-c6#VsNL0b41Epe9>*$hoeZp#A&G{@^=4BVPs`Lu#_#v
z`v6K!UKxJMSfyfb4_MG$F5J6<tz>tReX%-r*UJpT#b(g%rZC4gqvy$8kAKki(S%QP
z$Gh8n{*ll=sR>_Od-OeYMetpFLt<$2({^5W5JqjYeT05naah$xpRef~o7S$b+#A>#
z)L=%WZ@hHeH6Fxyp`ONXTkd^jc*WfGEg*W--_GxLAvdT#`Q$~CEk$n>R_Ykf?$FYg
zbn-iW3SEoLfUEO91v_T~!hPntd<(G04h4M#UQVBK%FOKH<Us#J7X^6@LZK(?4&^JE
z<E(<;s>ncWBQ4-@g1jF2wIsr$ex5LfDg?85@C-Gm!OtV~0AcvB$%aXE{aruG(*7Ki
zgdZhjdQHn-ZtJO>;Aza`9N#7CuMg_@+C5n!64X$^_Lg;dzE}T4s({q_$4D4lzlYLn
z(20no{xv{G0ujVIg1j%Y8mgmdpSOZEPCyI<$&TqEIP-aJs`Erq^E}qkhNe2NR=)&Q
zYk>X(nFUEEYQ_5!`b;H6X%%s^4CxX?s~Y?{li)qN@o)A3{(y;Rqs<ay#H<kLmJ1Dm
zy<&E6bBEz1w=;dalU#3bU*yAi>??VWcBoT*tG+Hj#kf%ke|)oqqd17%GB1Eglrp#}
zbMyv(e(i^>>tmg6c=l+(qxn-AlJ5Vb?7QQs?*I1@NlG#^BjXq$WE?Xg86B(0mYM8h
zWUrzq^O!kCX0pjRcA3X4;*iZbWM^-FFL(FnzVFZH^ZorE-``*T)uZz6^?Y8>>v=t|
z>u{qlYiw!w1=9dUtSuB(EY8~RWv!p1fN_BXe*{m1t5PPUB|#-tn5C@Q>H77z_4CGC
zUnfiIN7qDn(=ki1<@!=9Xz$HIU=3Z<3<P^NP|wG?nl2yR-`Qp#8sCG>J_0$UgX95t
zAmV@)4=5wg0Zr-$`kk+=9Q%roOjyc-XTY}R3`dep5uI7eX-I!)bH(!iv=xA);Ugtm
zZE3@%g5HEUop(CRi?O9l`HM%!LjJJLCl~rG3ml8j2(NfjbZQo|k@)Lgc$Y_?N>|_U
zJlVA`-xd=~70dr@c71;(wEy3l-Cx0ue{63mGPdK$6kL?S{#zY@O*Jsw{Y1ok^CxL|
zdIS(mDyyI;6i@wiM((gI>XVjJ%3*2aPN)ZW?fy}b%Uq?F+a}>uyn?+}P*=<3zWGOv
zgDWY~0r{V13sMtHcivG0k0qzk1Oc+h!}4QrGq`WfD?6H6q};05`RUnuT~Riqin4;}
zXRDcdRxc5b=a{C#&Xx!>Gx$2$e!VK*s}PT&EcIPz$7Os^eT{~)SILV>&c0>Eamrm(
z3-FbPO<|>}dyFRKpR8R3K)TxG@qH7I%B6UhFkcacah{atGl$j%3kBm*fU%KQ{;`7t
z%%%7CuEVaN0Zf$E*7HY4JLPUZuHfPFTY<S6boafw_)^2uPt+8VL_y>aA1Rsz1WXkO
zj_f@mf@n%V{mR6i!$JB>T0wBX_Jg19RBg+W{|%K+gJ9sCbk#}YbADJIkCV<+fAF4(
z#!^sDQ&75KuhgkP{?)&zlPKmt4mUce4-{~Q{FP;8W=D#IQxNWX7yjPu+>`vJMyf7&
zIOk*K<g`m?i-9#d{^HDv&y<WPBbeX7%i8ka3GP>8{m=gy`WOd{@_zR~_Z}?`m~v+m
zCs<3M)KyL>%COk3#E(Z*8Qn((z+BFW=9*wd%!Bf6%epIS=&})C-bGjz<c(|V*i6Ok
z&^MbXPA*;FYjPOl1T!_1PHDo%&Lb#_;>EZZP9`_zO@0l`B>47Qu0uB?-Ov%|#2vP!
zEsImbrTXx`8<S(7(hXN21C<7cup9ZWV>9eI*dxoh%SIn*IdO6sI1i~(S1LCuekqBT
zNG53$emNs_d0*v6Ddc<=_G!68(qpM_uYfq3(R$hASO#8}l5pRAE(>yH358*}>vAG(
z5#NW1x1B`sL3EcA*8*qir4P+~=Y0%AYYh)Mf!)b-3v+)%@1rw@A-93*8WuaV<{rpH
zj`xi`pAh^g3Mha+T?vE=G7`962_@;I$6;k9r_?o%Lcvbl5x+9y!12RsO#sjm1%SiZ
z19od_>HL-oF<K`7)!eIFvXh74MUFE3%|Y)g|G}@k?4CN~SMd1n=`j1x3w5^LNhw0+
z=4G7{THG>eG>wf5+>+c?EdRAbL-7Bx;b}0TMPH{@5Y(d*ysYQFF0KlLm2E`j6mVOj
z{96;7039xO(M`+y6|<1s2+H2VYR!-D4(%?z%NuVa>(xPXJ7E-|L0Ip8qG4H~82k}u
z@A>g}sdkvYcJS^d5`7E>+jcYu=?{B|aPEs`leN^bws}U2A|N_-tEiMLiWG$+msI%}
zq-N}{6>0m#!>9Cy<E{Q@{AM~ag>mme9?Xs*pXGcAiyg<JN`!r3$_{myyz>A8+@wwD
z4yuY;+_y@NcX?&ZNH<2Azhb>x$FM1MecbTM>Ruh~(unZ`buzaHy0BHphj}g6lHCxQ
z(^9*3J!QHBe)1E{LyDETVjXDb<xj{<+r1KT{flKEc`VL?20tuvylzXYBI(657U@Jo
z?f{`HRn03ifhI7UQ-Q=>?EpZwRDJNHZGEok1ESK-ZaoCJuHRPa%*`!N#YdhEeh7kp
zEPWXM@R@iyb|>wloq&P!h#wBD4siY_f~S3V<{FoymTjNhEXfZ3Uqu@as?^&<wi|-i
z^bZS}Ovz&{?}E}9WE~!gR(^ca-_@y(?-4Xb5)|HcYN5)(WqudNFQdL|5Sooy1Ro?~
z&la<sMU@~OEouS@E4PfSv|_awU8=mU4s++2tzND068$^^%&>^YdR0b+dyN%lcC~eZ
zuCmYu1THx_q)y)gscFe04?^imV%D81PV?!Vzh^Pi7T?;)7H-Q!a8?oNvEk@hQ6g18
z-9pmGRQP}eRxzt`^F5xPT;j`d-;*)-X<x*97qq<-?#4K0&l&1FydEsv?w2TxT`VnX
zr1dtod@$g%II2nyOM_ssUdJ}j{bwe<?^?G3*4f_+v9mjs2{WJMhaA>#LvCCEZWAQ<
zUYb9luGM(;$eT&XOCWPZqyd!1N2S!JAgYqe|L!`$lw=07Zbf$OQR&M3>Y)S#?;oL%
zAZUd6k7*8j&r1_}9CYKj?Ka4}sDGgCpxo)6N*yN_InrwIIEJjk&O>X-+OyBeld14>
z>c9L9Hua1&U=EvC;TwnlO?%PuK-dn!Ehow&lS%Zf$K8v#3qSP&o2-j=2gR(4sM{*X
zZZ0Gh`R%TW4YEl=NRs{sP`d03$p=9TJws2?H4XReFb-Yl?Zcg8!_?G|dR<Mri;~u6
zyVEOr-S}~-ZYfwh#bqn6vJnG5J;|lyB~P!i@v1mPmoCd`rVD?bSHB&{glLStEq(=!
z&E|oQ4w&0|WplTt;>Rwr-Fsz<n|24kp`7=&^<kBnSSM>M^Cx=5wyZp?Vd&Kge)y8k
zuS&77EGzXrH-#;v#`oL4p9pb_hXp!rxiRmH#~qbhj8m##ee`fCfXK`fJRi@s&obEz
zLo&_}O)~5pl=!8+y97!hfgzqW@n=1vTz1a<DZ)WE7UDZQA8THj5}T4nOF!JbF}%&)
zDX%XS)bM(-vawRTRn9w)4oCs^WKWON=67=b8tCqJ82)kI#_IS-4Ln_sbX-WF)=}x@
zz_N2RS`}OGm6gTC4>(#%PBj~mG{?k#z<tsFXNFytm&-tby7~Bx`S3^n;&Pt#@C}bQ
z@Qn%{AQc){bXP~D-+W<Or)i3SaP69HB-~WqSH(-Mcw_#sgru~M(?EkUx^gj>^_KdC
zBVF@cCUtgh%ZQ;T5Mk-WI&fU~4k|OU{n&Qxl2po`WSp9PtNCcuYgsG(UYh3x3B3gu
zU3C@AfWfe@A{C?LLN{}k#FG1u^6$I!3+Q#>sFRLm<<HsW66$Z6+H|(tpq0eD;UKr-
z{e6smS5L=KMMmkU?#CxXWSl*iutG%&xOrd0+vTny#kGiZyWO-2EWu`l$jVXr`-a56
z&7GOBwnFPCOBEnQ1oBWiVII5JG#gY2aU2KY<@A(KKAA{H+i_i3<!;Dk?1{BS5P7<F
zupwUE`Om<8;uwP;#s&<fN6roBFX~v{e=s3enxH8)K_GtrD_=nUo!#SYUkyg`R_vC_
z-*$1V*oZ$?Gj}KMbMhD{1Hak6V~p(%_|=6Q>L0gEx_4XCQ(5Al!`gpsX6S3?GUL6)
zcMW!5WKX<8qusKzL6(#Nm;6`!7f5hJlX%3K5qkvYYoI$&^|`Jc3!+1)lGAFqcp`1p
zWmQAtmt7mfUB}jsUusBdeq4RPV<R@$5&Wb$QtF$ya%;sK$*{7O`b5jIhjUc`4C>Vi
z6+e)0r+8JrEv@I1I6$5>;1;~Ibg-RaYPiZL5g%5h-UB}D3nbD)<EksAw=>7AnubF*
zX{}aqG<&{aQo*Q7eqpf1A?kYm+~!-9P09{4IM87y5NppzyfjRCf~p*sKN+ER3r}Ar
zox@mST|?YMd7w}eNdzLlf-`Hm&VAyDQFC~x6b^Ux4?!BPnoqpLF?UylAjeiKq$f_K
zW4-Ja(ZSCPcC?&*4NVh4!)N9BW=-vRxddLeGgG?!uta+><)S-z-RqgNov7qz7Vhg{
zxZb$|k4ZikaS;i=koG}F`d>O^a=6Lwx3iW`8z*qh(p+7mmonBwGlyH9gWEtx<Dv1N
z_01}&&oRlUkR;XAq7W|hi>>4DyvH-^%-=3!zct7JzZZ}p1?$sqUUcOFu?)Yy`%4oq
za9Cc-04s<zY7f|5m$cNIJjad4KfB7Sfcx{EA}#-(<D&f2-m--TCvV%=U$lKiwF6b3
zR2r|nbG4gfcdm+$;M3vkZTGbOq%bqBf2!#wHL4Q^x1C&3Fvk;GwI9nhYG1XfD-Qnf
zc1PcnqWoFKJ%jE-B1-}MY6^4SKoP|22SN6#7q;1FzbnOkmvnL97LK&WfkdAY>!~-5
z+kRa#KY--_O5}}C(9GpaFND{CY0ePi`ZJm(YXT?d2+yE$#6!c-nLBY*hUhO7%yko-
zJLZjPUHUruJD5+vT-T0+57^b6<z0xhCtEcL2#q#chb#fiE^yJ11+Md%o23&?zWlyM
zDK?^!jg!ss5lAf)MIcP@>?8+)wYJ7*Why1jy0ArMe(}6hmm1v<|JYJrKneezpPz5^
zHTpinOE}$Y7vtrgr`Ukju5xgBHw0Oj+*-Td@~1%Lzq2|2AL|zk0IAVIiK$3```ess
zvUyt>@R+#3`8Su=(g0FmN(B+PWOuB$M+~Ky`1h@`%9ZAHHWRxoX2N)QUmxB5BmmK^
zm-yl&(jM))DWY$S+V~|BO^5TOamgjLT(_K<p`IOv_i4Lz)q9h4XWzE-=HswCJ#3YE
zv87kMoJlha?yVDU`J&uafNc4P{_Db}aF+rL2CwJ=%9J;5NK{;l<hRrBC;fD4p)IMu
zK+>*vQD5EIo8%e0r3YKuE1#Dp5uOYp1CNu>OCd9r$sNLedY*!CrF!cYgWV&mPPBP_
zWCCKkz+n5`t`jj7I=+ducy#$9A-G{GXi^qrhpc^5MH2uaIHg=8@z3Y@#w;5IgS7Y!
z^6NL6-rS9(D50HV4=d+_!M~9b^h%s`I~*0K<Qd}`o}7uRbe8_+H?ZNqtK4b}7>a?}
z`aK=kn*Fn~<7}l@^n%5BB@2zrP5=9*qw+`7@qVbLc~5{x6?pDL^@vQoxc61beZY;}
zD#zRdie%Pd7EU2rUXhoE#gl;7Ic#`d9PJ_X`je4$QPm+x7Bk=$Yt*R|hzZiHt8(ew
zMplFfx)mD>m61@V#-<f2*tI`~9)e7ERNg_7PTJ-)h^Gtf&`bL$xPnr97Qc*Ktf#uW
z>`{7JAg{kE?vB)L_|g$KQ0Abi1ZlJKjhHROGf+W&{_c^SUU2eb?om0X(G!p~^SPsH
zpE*Zi;V9|TLD1Z8=fS>&rSzp@aN4veL9onXV6Z>v*}L!dly7Y=6tiQ+f3J1{t?1=a
zD(xaNn6Ld-n%m1~?cuW&>Tcs@4<$B`6S%ZmPg)@*GsK+g$@ulOmW1^llsHaCn$F{L
z5XkSoyQ=|W*#RFlReS!psOj|blsEAa$&}<nlbV1vJY#HzKt|1rsbL~U{JM|#==9pt
zzTsqxNncC<JUv=F9gK+kdxBzh{NvDx9BzHG4<gIJ>#EvPjXYBA(Y<jTkD2IeBnS|A
zeL7lr|NYbFPe8XS`|tF|zkD-)T*S$PZUy*^H3RP<_!4ztvHLB@SWsIlPW|_&hhK7W
z&1V3rig*6(fq{q%*WoOEMd$U2_$3qSsM^p&E%Zm1Mg35bT$h)VM=ty{v=xL!i3TL)
zsjkZ;o!p)tcVWE{@573qjYMWf%ZX>qguiyalE-$jo1<3~L^{jAFD4CTxT8*hA#k>a
ztpM%rjHhuI2)keM$UKz_D{r&i2DwUJDdgTR!b2;T1&=rg#TLW4t3YOISgjj9oI9j)
zmG%qhG~yH$g`RvwbRRGLYO6DXG4Bk2Oj`W0cEtj?&N})yTH1wkdbyon0D*ue57dkL
zlb3{hC7<Iggl&qD%R8GoP?X@R5xfkK{-Abz&!s5+BRSz~;Qs;gPXUzp=!I88Tym}?
zct-KjDb-Xh*9l)!qvE=w<}u}&-y-`@8bN!{D`4HJObs!;4}-mu@vGe?Zr`%_o3?2G
zz0vf<17tAp$2^Jj_c`YyD_2&m5ECsh$J@A*eR}lWzh3^oyVt+adAZ*|bOmSl{NHs2
zse_-lqY3cnO@O{nD;7kj-lK{-{DfualC@1RiVT8dNP9CHZjGvt^tQKP?iEw+rle27
zk9jyeOQIfJag*J-$8^E2;76Tq;;|^!;_AF1Fx$^J5VOFJODzmX7v*DT+GKd}#Z8)r
z#QHKR!XCW0x*?R&jBzvi7R4)ErS$<(ICffrSglE4?x+~S-PI`S_W<**l|%{KuD)D3
zhdie2Q=Fu*t>@3?N$(PGLF?EF;^d*xsK9}P3ZJ(@4~uEum&?^>87%Xe<S(LDxz{AG
zub~?i1@dHtVfC#c(h1xB+v|sfJ<!2T_(9jMCMoXLu$NAN`_9+p^3}f8%n!Eh-nPl4
zgNbvmN{Rb^g^ylbuf(Gth<QYfAC!IrzlkpB%^y0)ZQx))<Q*Ec@#dq$dq5`E@|;k$
zrU3>+6KHt61t4o>#q~U3H4XNieZV=r)h6;UBJrgU%kQQ}eAd*YPl2X@WB+lvKw@7i
zZ~?tFpSHQAjwf2>R`FjP;=fyw{mUr-ulyK4+ahfkGdMK>xsKrKiV7iiQ7V|7BNr2!
zdP(XP7r(6B`8<@i9E1SfEy(*r$LJAvtz|K04YRE;Mru>d%BB0o$)7EoEVANRo|s8v
zMLh3u&5iQ!Hyj$q9(utw`^GjEY&O+D=1AFLHlA6StCIY<sXxdYsiknfwR%ZAA(s#*
z5m>WK`oVJev2XQP|DxcT_^~CDPH;v)LZ#qXLF2uddLH=u`R4agE}-{oGOI_gU6CxJ
z3T@w{k1w{s#tto!6g!lySCr~h_FPx_wkeLGOjnv7S<AhK5s!>dXRMRJyRFJ!#BN`1
zRT?pnDaJn(DDU3OCpUaG*7n=T=m7QpUJ6L7mYy1e_i<f4IR7v`jElqMAkEW;)_*QL
zNLb5DXe+(xYEV6aB)<E5^?NiRtym!AHQ?#q;7uxS*K!$00ZXhDknjk0pm@D!bj5%#
z&p&#d`ER;ODiMf5ma^3!rhZZBw4N97KP6T93_YezDK4QpTX6!hoKV^{SaFg{)&E;W
zOk4-LY0Q`lGFgq4M@`m5a2?+FGlhDdybITS7ZaIjh>Ssq$(4X@0?Y=j4D(wpmfc04
z(ZIobug<bXAh*h#GASkRbvKd-yEVom46b+Az#fQYqqu1>1Wq+Fajwf?MyTgr-_w7<
z{nBtD+^Q=6sz+{W*U^^6fU3Mk?-H<WHHl(wV+&sj>8cpBo4f6mZ(+(;DU@|Ox0^$V
zI}@nB>q1W?2CxspE|d=%mKU)B-4+jOB!}<HEd5~zvCbib_pg40fk^%FC6P6-tn~in
zRXXFES#>1uv=g{|^CK#!2*Okw1UE-QsL$axoqq85IfKl1ljOM0F@5i9nnl||2FH~=
z<42FcX;aWlMtv-@#R9CJ&k_TFlK2t`1u#(=JQ@!(C2iFSq}8-jTsHz27GG$Z_$gzr
z&?687n1$1$UH=A&m47g~>Op_xGHu{r(}X6xDNH;H+e^!mdQXMebj6V#O%l5uCp7=h
z5@Jyt_ORHQnEpTQ?+bN5HT?Lz!j?@Q`z@CxGa;rh5qzA$+^^I@8z9s6Cm6M)lf@0i
z3n?jz#^BdmkHJkK2ZLB{HHFi+u3;5lX_M!7HB#9ix(i=?d$`w!NBcl@k52IeImzO0
zD&h{``4uhGyXL9ko7EWh+f@p-VDdP0Bf_b9y@lIDMOT_~PJp>RFRNH0GyFZ~HL@qx
z3M9eL=}$j)yRTaGOmg-GwJXyT49NQKb2IA+Nw>pc>zHqawQGX|-c`%`nDr6D=LHab
z@4L?hcf+pP&3kBlT4|LTC?#XGtGD?~e`j4<tGnFHO&;W9+Ydh7T0o<n$v(PmR@9FM
zo2FtdekwlV=kZt}!ey!eLo4xMIi7gw0KNdGhXAuHS};7UqCmtx$8)FHQhfM4WB3gX
zeB<i6^;0E<^`yT;3GQ}$`OWV6Fe=~jQ#|5Zx<*CaXlDjLVAROrf#{8#82?0n=bMaC
z{&xZ?uEeT$&vi(^rMF<O8I7(AuRHw@@B0)Y9bWyu#i^ZW`F>{s&Ri;7E}os&{vL`p
zF73-{TD`xX3p=`~;Ir>{{;QwWbi%!2lKN(g`w@-zb<XFbKh8|PcxD`S#Itwdm}On6
zJ8_Gv`SSBS?pG8P9u-||x^(MY_wGzC{+&nIj;lmc3l5cb$^NUrcN5b79rsY9l7$N+
zTbp>;9PT|F<$H+fvW@Th7(7Kk|A{i`kOIEsd6mU5QqFcImCD!7PRVOEyi0ogNXC`W
z^7NfwoC>qUmfnO9fhhK{>*b7=xI@PvkoRVerDg%|i|dxSF^aVuNINu6{D*RFPxgpm
zORIP#q<ko^-PD(_XjETPI+_HXeKdMs-J#F~-GlR>!th2*OcGCJPgh<V9qBVXjsVmm
zl;1#ktt&4G4*2T6Ty)1sS$1q{5vH1UVGIs)IW9~1cqQX$c6DmUkhGk~J+eM8#|xc&
zP`cK6xciDz88=jH=<Sk~xdbLC-N(r|L594NG-yNgLMH<OFTD;(1E*+;5k;f%jJr!*
zTiCe}w&eLNT*cv!5wllz9j6=HD$kX$m7G@j@0r_T{(>yzGv}hSZh|^mN7?AJZ*a0Y
zf9i4{cGXxkLWzevw+v2|Xe0#jp&cd8=1}sHmVT}0ny{KS>V3trIMZAHxzh{~gg>I>
z@AZ9ywAQJol-oCmLga!J)yPMlj?jI$tzR9|=iuFDW=1+A4x5;9gg1q6tl@X*?grvq
zDqR|60Zqs}O=Nk>`{-u0v1H0^-0X&-$u)=7Y;yfg6~qVqH$5u{2y)3!eYA@DO9?Y}
z^;YBNI=1&<GCa)B0q*dEJLv0f{NlQt>6|7(;5;~><^`{57neDaFMre+lWpU7q_O3s
zpivY&AkkVC<3%}I9m{FmI{MLpRT13H)Zd;__6omipL)v&I5TYbPuQwzta4H(>}I>1
z!sV(2uCz<{pwXUpU5y%1#O&TbFEsURVSmls{#NaiMt@u_p(xIPiOPMW6NU~Bh}4(y
zy8=bkx|jKBYslPyB-o2&J?;0;*txw-GO7fno1;XK+`PaqXV6qaSmHEA(9$1Zf`9Wu
z9o}R5lH2g$s93v}cUBBBqk*%I%%j2W(#@E@la=cy!x67&{{B?g!`Ll>q-Ta{AJ;Wm
zGkJXEuLXtrjcz^G7ebs%e2IO9UH{xyDk1vd;aJ#hPUAlNhF4W1gVOIMf0izy^~z=5
zf*7g30xt+$$cgU#Mosx-LlhP}z@rRnaHTE2POTInYjcgrl2!|mG93!O5*!!|BjU|t
zHrU0lm>23mtkz`}uOxl-TJ}2B-yhRmG<h?x58rV8DBTFzo?9Ax;m!TEe8$;U7ILA}
zCdQSUFt?(!eupBn%hoeIbxJyXE3{p{JZ)KrY2W9&ORo3bxga<<<Xo`0KDwfH80*F_
z-}|OMZtMr8o92b=yyX$<@LY`N0i1s{e#54~h!dhN$&)>HAYqlYvT)+IGw=>kN4(%$
zfG8Rb+9*Bn^oRE1?(SB*qpT#dC|*kLHq;S^LvhCf-gfA&qCstVMyy{K)lXM5VU&dM
zLRSStul>57xaW#4$O1QL6*&)`_%>xK{?$&&6qvhV@P>MGAp&<Antl=d;(bjNSbYAb
z>EnPsm8sT%U?@L+0m<mKYC$tqGwX=a&*5feZz2iFeq9}+hz<S=MLJL^cY^QMBDvX>
z(8v)W9~Gzl1Vxp=mThZ_-c-0>`Dw`9??KjBj5ufSyy#@(=<;IfSEjA|u!i7@MiJ<K
zU35aSXHxL;a6lLTF8^5JciVCS0o}cf3X{>#miOrKGatx0cd)Pmi$Dmm01vU=NSd&e
zSrfRCJKVRLHaT}8$hIb4%o<U#)M89ETWfaf*2Ut5v7xk96yaP$x|F~;?9?9)GQEv&
z4<HyI1HK8_iQE10h$X)tT8<6dHfsEow7yoti@O4G@460$0SV&pNd3<F&BK23*m9bM
z1WxE`9+i^n6Cv6dgjSN8ei+KPv?{Sb%5O<w9x4ZoQveyhMWYGtfP`R7UZ&XDcPBOL
z?lWsy^rql;HR}~6=e@-AWv^{-3TN~1vR8ftF7-M-<7!Xtja!eHmm8a_xi$Rco&>=L
z!0_vZv*RNR_=L;a0Z93@E!4(Y5Vyz<oMeO#C|_@7BD9nwAO2iI60;G~L=i$AK+9%=
zcXYA6`1av5<NG@cXyeCzf=+_s{CEX&c26r^n%r4-&XqEZ?a3dLo!Gm-U-+fDSdahr
z^P0*0T?u{2&nh9Y>1b5kJUh<S?mM|x*(PP(8Vf-cdqgP4*?Vnz#m&|6CmHPaSb?^c
zkNEY$R=1&)JRa!&m-5NVjtVSx>7h@bo2)UcVeDO+HY_4DlU4%nn&EvdcTF`iAt7Hc
zxSF=#ML?u`1cYalQ&$Ae1}&AooCRhd+S8TsQKK#!7^WxHtk)SLFre#*5zSX3#}@Oe
zVEn5zHw4Mqv0Ew6Ci~z5mdEj~$X%i5()$He>iR-%QnW;iLYg(P?E@7^7dHpnxDy8T
zT|`izmQd<d?>GL+)*C`$?IG#Q6tyC83aWXZAR_9lpM#scGSZhn)@)bTC4}thUuaoK
zRgpuv`M1$0-5)D5>O|K|>+I&2pE}={hdT_?6|+vrK5VhTW|}$XR?Wb*k1Wl77`>ID
zs1aS;6}rH~htMRvr=k0{pTF(G%g(*ME!i~nlQR7IFZjm0ri)D%LpVB!0%)U5@Q+3U
zOnPKbFYaww1mp*L&@s|{V&OVhOWe+M*_)p16>;^-rqkXN99bgLCg-?I<L@e1ZN$z^
z3aa4sZUK`McUlF|_#-Aw1U;U=zm1oXZkI@ilXuR_*bqzk=%YH5=)qLA%RFJMv>UQI
zrO+HGVLkN9dGf}5(j(cFbRQ50>Z%OiC{>9;2#mKbDQ#Q*NDPt6hRF8+`gydD4NxLJ
zl}f!5v~dTVStmx6x+}rjx)V{rbKYyIe26ACd7)KlXi2$0Q)*OM-65=*kEA+b2$_<s
z^Ps&4sH}42gMmrye&SBb%F65XvBp=yAJoj$WQ(vkAIEJ&0ZIL&9b|-@s#Z_hr+zQ>
ztFhi@p#KzUhi)p(!#$~xNy2Pguf#|6t9uLki!7vfwB8D=DT6?*=UEN@=0b3X{R7(+
zUmfqI@BBn_a0gVQdrzAio~!-Hr?@Xt4R34Q7Zz%6<4ZP*Lm-&r!KEBHJl_xHhcxCL
z*Nz@K)I@c*&VSH9B!I&UR_chmxgY~aHcAp$co}W%Wd>q9yqnFXLU6gX^dssP-E$|G
zrgYi-h<@GOuw;qHqrn@xLd>sRFdaP46jN+}e0p*{^62zL_H^;9-Q&M^Ut>}+X9e*!
z!&HKRf9>gb-)TD&U3^DJd|1a}+q(3k0*-hrUdhd^A+N_z%VATxuBT{~WLz#1i#2$x
z3*W$Ds_{~OSHS0PMH~>5${ppd@#3expI%gjFZR6ET9?<>*vUipk$K2mI{CJ*fU~}*
zN9U0~9oTe7lAkL<XfY`WUR>E@RPJLcTv5$76ptxGDQShfnC!hyXQYfw_U$Vk0BK}f
zSh&C)Zi$2nz6)RWk#M__uA};!70Y>fAk!efSW3?G!1D(J_$?jSceEyAjZG@3Y&3iL
z=G)~WikJ;L-XX*dFg@OU3hMLESWyrzU6CcI{d*`Ah9+B8&U2^XC2d}!BuIuw_-lXj
zy0>Uw_|4n6v!d0Y-{^T=h<;Y;Gq)j!!Z%3ck1pQhIs1|Viu0{uC-h?POJ6-(iDw69
zM)h7C#1yv#I4GUAgvwq_Pj#FlW}G6L5d^n#qW1|c*=St_<>*LBV>a#_jdYmE)dd*R
zmj<XC86zH3ny2i!6*f1XG+#gLx%ih7_;syZOd04?&XVNU8mEO#Dc+b<2;sAF>WxtM
zf$#@$8uY9f8;?I_@=?p!+~M<NqjAcb7nGtW^96L*dyl)25S=4e_=YE&fY_!GsV}hK
zgxfZjBgY<3IVdh6Bs>l~;O{=4+D}}kVYfU5N#TJIFV=lDD`OsVxwsYDra_rYTFVzY
zlBTrlrEQn$ZP_!RWbreCee}R?ODl!axtg0sj3RP0!_DM|!$fXUg;&3$&jd~U=H)mu
z_i<Ih28y=~(&Q!6)j%RQ9DYnZe19^9GB*F^gtHne9c0^!yAv_0NBZ&6jg7bIhMnD-
z9;q*9;V|zjd~%)+eKyBK(w%KIht7&-POC*jp`8!pMnrl=7Iz78TlIZITtt(LzP!7`
z(aEacs^iGV4(-^Fcd)oQTix2yqBx0^+i)8w)F{`@UjU#MaC`HU!%}JH&eT-Xxj-i3
zWqi8@1fnYRVcq$X+h)r1##4xBS{CC=kFTv0%mia~U3>)7QtJpxGSp!aZk5yGM@zCt
z>%&*p&kEV^2J~9;?(grycC>+Wr0b8>)*Ly?#P_B1JfKcip1B0>A!;r1)Q?Y39hDg=
zi)VKxnn#y@PJCrTMj?Gjml`JSG=x8I-eiVvj9NF^N818>|Gd4_r8E4E5@`fVSsxYC
z3uYVUKC7UtNm0nbQu|}mf{=`9vYcAw>gGG#opHA$*obvk^i;3)&gX2t61oA+o2-7g
zZ5sOX9j0Tx3X3g+*%li2Z(!GVa2Bd3X<K$);Kytiew?)-tbD2Y>j&v^+A^A#hh?sQ
zlwLxf6=_zDEA{F6!v4<<Zv?;7)7B_znWodLf;I59<wORzNG7~SrWUhn3Mrf36W!$l
zGtt=U3QE_XUNXh0$XTEax}tznX<6++>jgR2K(caUMEW;--nDMtbF8PDgRiK3u~$ap
zkaSQ~jh_N+bw}6NVf3z(URsB2J5B<Zy>;?v_VZjeh@ttJ@ER`U!Lod(c_L7fex?+^
z1YB@b%H5YcUqXOlLEWC>4Gi9Yh$9N%2yDI6kT`qeLZ>Y}Hn|E7B3X33&YsHeZ`3kB
ztF5YX2I`pZdVC(vF+YO29Q9i8@FqmQ!b-$Q$ivC2aC$~B)vMk+M*Te?efbEn?m4)%
zYJin0UmWFc<jviN|7YJ}N)}b%1_}Z?k4rP_=4#4hcQ-Zxmb;BpdY6!75CpHu($gKb
zuM}q!7Nd-V;3^EPi(b`P@ty7rrL6FUm5+a#4$&C<FuM3eN5kPlmnDD*t+UXP>Qv#X
z>jF?r#&Zs-!_RjT4JKYKa5U!`B4I?hjfpJP1d0zX!FOzYqbsSbv;jl1Gn8Sw{GP_$
zpx^!`va2FxLMTw&zI0sz;b){<R@7Eb6I$!b?rU|(y<?Qd3o_;%xm7dc(*|xVc#p>*
z2D}F;VpUP1GJFFeU9nSIu=Dg8j@+N0FIIG8Qn#2fc0(#+QA1C|RofDBYj12O!G*rU
z>}51zJMTUvAeaKBd&mgg$;SGpu9!YBeNavvNZk!#ePI}A81Ojp3I6bx`vTmQs^s~B
zhVN#7yta%&41C8j<y^0pz5fqju!p>_bp{h0UYj%=6&`|yLySi@Z7{en9X1=_UIUL!
z(iYrjh03I49p#B`UWO;HP1?J=ssAMNVuwtAsR0<Ev5n~GO^f#PhVH0*7e45IS%td(
zs-NOgCR3gLcW&))v`Ken!7}Ii-Zw?okzXw=oUd%vFQGd%H8_7obyPlh?|v_AyU!Wx
zwa;9Eo3SbcZc@r+aoq;5TL<agQj8bNM6$=)_}?w`$<$XvAi6<98{QM~g@^TSM3$Ry
ztVUQfLd@{30CT?d9OWDPms{p(zmVhE3L5;j=iKz|RDCRO)VIBNuGiW4^-xR?g7Acl
zh!<&-e$>Q1V6g4z9~?7$*^@O2U8wqIqFS~Sxo#aQ`Nidcejiyj^l5U4d9&sYYpE#v
ziq7#vs?`<XBB$<VixNF(%2pVq3=>YOV>@7`^iIZ<i7+g=dF4NgHQxw^#UlbE?3NA9
zE&P4dJD3I3eWT9z`*)D_mb@C7aE?WYed2<cD>laf?9I`^5kT~rIv|+NcuG5lc8j~y
zR=A+Y`GJ`^880xEh>dBVJb&+&pYPwt4tIU-{>E`)A+&M+1VAw_<4Dq}R+K#(sILLb
z#zDs7pKtT5`kk8;ui8Hux6<8x2@u@XRFB2CSLRPFdntMSB>otb(vB+?&+q}Tt0_9{
zg~QDQYpDy>7pGR(A|Jf5X}@Qh?ukxRjsa=DM?V9U5^{LjKjESpE;YZPdt32ZU&b-T
zHkj%<0-^f3;!!?CWY=<RfDmWwn0IPW^k`fA0;;f~KcjpB-TH3d_{~AM-^Wk0%aznI
zE^b$HJuQ|dc8DwwNUnvAYK@n(Ec_tH{mjtmx92ZxLrCw7B|I*Fh*jyf?<lFa7*``p
zT@>>+`!;|~AES(Ewxv~^EjmJC3J%s@x9s~m^^`6waox`q`Qb+T#El9yT2XHNW^;sD
za#v=?&@U~gVQJf|&wLyTrAwssHYjsAq3gyx>#=z=B+T7)qHu6)aOFbhd=4aG?LnoL
zxt+(Bx;<4a(;cPeKoIW15d@HT*e|$HoyU<GmNXC)x@kpAJW_H}0$w(gw~rZ0>Bz>0
z0}agM-_>1d$2W#Wn9RNt%l~_pZ82g~)&oVbTF!5OBNMzUS<A`IyL>#k;4Vd{s3Fqu
zt#?pD!dip@MSIIqb^Y$I9f=w-S7JzS0>VUOdS(oqqFW70UX5)ef_m1lL-!xY?ZzVn
zs^+S=AqG;{1`rTAPl^~yZci$N`9~peG#?O5Wo2^i=>>G0f|O%&RZRHiv^^9mJ-GOu
zLp{OropP`60(zkiD!2a-3iUQFSfdO+oN?ji5^L}4m~mM^+l+fnwqCxuW3pP#cpWzf
zqBVPbiSkCim{@iG$Tn5Ef#7o4uYizag5X9~f|Zgv1twduo!yCPqZF~`LYv~n+LEiy
z)R@g0rL5WE2dp~;0WSN@bA*-{3fE@Egls6SZwSH&=zZPVBidc~9~}26%(Ur`EJ-NW
zfK-y>Pm3H&RY*)-W6i0Gl*9cm*;L~cbpyF}hN|<aypY`?3yHRzTn8_h&<P3a6LE6!
zCOdUg)vsM>IgRt+Uyi?#|BdFh57B-Q3s3utknF3xd!Duut5ek?R%UQ1!y6Dua(>Z&
z?1yC_e?n|rf`IG>uncv*4~qj)Zqgqfnnv^)V92KTn44V0nWyC^98wLU+gg@TJ-fd|
zYinp-cR35Z;6nvhW*ZcHh4Uvuk(ZD^-e9r933WUP1{f=2qd@hWkhF$528pHWrt^4P
zk>~7G)ZIf~v9PFXl>zIB@`t#n^tYeibjoRfsIWurih|L4W8Vq-8;&fhGkmR6@mz1Y
zx_ft{Y)ZmKVm5>^>XfmOVeOaky?>ZbI0`?giGGSKVtVCywfg=|hb)M75L`IAhNjtj
zE%bBk+q{P<lis$gvRZHQ?_)bE4UsAZ6(y2`eDjrABgV&Sw;p)&TvsQLt!I!3?;HI|
z+9~)`Y%N=adC=DCw4itm6s9g?12-2|$tSUe3<N7z=}=y>8@TdA{f73^$ec^k4ZUlm
zQE$HUftI(Zv&#_f$rmWagpg6UoU;>1p)ZaCQW=Kb;^o3xD@IEHi7$9<kvXSyM6^<-
z#OG|>@k(AiS?k~nxO_R81yX#AlilS>i|#e!3x2`Py2Ae9pcIPyB*Fz6j;1oORlaC;
za56XB@SsAY3GdL#UicT302JQ9jo)wT33yX%_{;Ji;qTIY?no*_kK15Jy>PF>*f=Ht
z`K&B)XU}T?Zz#?TX!nvzpja<<#+M)+DaB7PT3s|$g*Di%AmVFHz8yRzPN4I&(_j3i
zPZb-=EjL2P9IC0ttc@V?(Pih6{ql<(I~O=Du$V0_iAc2zVZ$u#eu%X&Fi1gUbs6+i
z7!7P%dck?{F`>=&r6?4HX49%I9uRfZ2@FdYJHrpZf@?eP<=w~QH8V%UVkEb{`TG+Q
zDB~y?A+F3_lQ%5Md4kGklr~ls-vg$xda=|}!j2;;XZuYmeGuHmL3@?YtH4XZa&Esq
zewXmqjB{#Qavzg6f<ed$)vjh%)Pj*1sX>W--|faYysk;vtxWMs>E$J|#4rl4l<Rw`
z-Es)~0xSz4M!Mde2c4Gl9`(t1<K)$PsZV=sw?-KKJKwE`dO}Ay%*y<4E7hF4OeX5b
z_f)+(0xVJRhC2EYD7<yECelj+PJ7hYH+07LZ(g{0-c+%?=T1@g6Gb?&sdbBk69^4^
zp$<!J2oV8Pdcf)C>4^c!8JG7bP)}q0&FnG#>PonNMegA$37V@nH?%V^(eUJ5kJdxK
z3++8l|3N&w3_pe1w!LyW6`BvFmQ6wWOmuYh`HkMd#2|{b*QGJO)=*S=)A%ZGaLLFR
z#yri=bLI9WGLQ7|j15X(vooqZWqN<A6>CkOO3rS106jNtiHo?yh*|3_9CYI?>Q?PQ
zIE_|YZ67WB;DjMU_MF59oLuVjZ&C{ciwB<T=kqc56=)YE4JE6pM>Kayu;*^gC|n}*
zbLO;chcDP>+%GNKUcI6G<Ku1)Mbz|R*^>`N*0#GZhq-R*ZmiX%BUP&|Csni7z}!?n
zqXf9-Z*aA{WT*X5H|Q<SO!WT<JdOFvjjfU6KTJRowpfiPlq&jl9*T-fMfR#ABD@Kl
z%EjhP+CHpV8)&$@95YAzd8g&ABF!nlaecLHk(3CVlJ3mrTjZF5@F5l5&znBob#Iq1
zxJ0bt{klVHCIs{tub#|zT)xDhwt!Y;qs5EQOjoPBwjPJoB_}=C@a0W6FM;i~s=4aG
z=DF!RCX)N__kMA^;f*yOH%G=q|J}tJG=Ia5=U)U&{9G>s-Smm8swqDQZiB1cKwL&j
zb3eXCa9zS-O}ef=V)gR)=g>_YR`)@*_%0tjRc{;o|8@j36TdP{Mn@6YBh`8wf6&gU
z;Aw5Ve??BikLB032nap%cx=`Yl9Oiu!YwP&{9@DPy8LA})8=(=L*>sS75n?mw;s|O
z<bAd*G!CCYY89_Q=&Oc4DlR$bZ;lO1eGy+{Q>h82Vtf_B-IBM(*jq7Vo6V7nZY})&
zv#c-CSZesY(ufN`<8=_mPC6iaE)tw72&6E|D<y<0Ysag4LaAd_*UKr)D3d;ooV%D*
zO)fW{_;oaObo_efCv)BJ{8&s@<R=)vYTxVXU?g&-c-UICX%({)+Wr)2J818nWpUiF
zGxxCm0q6d=iSpdw$72TtSp#zxGpXrPz;f7IQV#}OWQs#r4>9*uP!tYq`v#SP?Ngvt
z#1v~5WbLUOOmrE~E=65xqA6#~f-T{UC}N>$V_`#(lADA-zJz_VZ_A5k8)ZOY;IR(=
zu1s>o0jE1xWQThGF#|i2{rR}o!Q<MmP4f@?Qi?wuqPi~Gz5bd6ofn1l1DD2mEsXOr
zpHG>l<rQ%M>}n8KVrd;|sO`TfzBOPBR&?{BvBl0PZ4A;!p9AD$t<G#`!jgXETL3~b
zF<oscQ0D3gX-9Cn&n+oR4>E#H?WiPkc$wgLyEe9}?1`Q2MYB_Vp26M@G_KVzW|MtU
zR68nPqP_tJyco+$9_@DTeoq{@8atq0=&d<~{=rrz4pQ)*?~Z>^Ry)08%OQW*GRHci
z>t|Bn;bO4s6%)>|VaFP46=b_ZUmj#7QWfjXU_OJp{h8M8t<6?^Z10X<4xeQE^1%o=
zbtRwYw;wU#*Akm&4}$wOhWWl_qP7*W#mpjYb8lF?bY=Z??h7`3Yj~%3sdc-vPXTaV
zyC@^lkskO>Z?1_(xO%Ix{tcJyk(Jk92Kkbs>$_wI)_Y_aY;$HW1R&CV(x>lC-EmrL
zk|tgbLe_DVgc5gG;_2Y&vJBk=Rk6E-z4Tf@Funl0?p(<&0f@j0WE}wtzl1fvSdb|Q
zS@7NBU-j_rg=p5_txWcU0QFk5Pa`A30;$qoNQe{P4bazunaIwhoEDvxMDesypHqkO
ze6acc$_Cx189EulVmI#@JIq<ql$oF7uN|v(s?u5u4O&SiHTkq-D+$;n^8tXpY%g>I
zQ#V4(DKvf`e+QaXaUYF;h4EhAeqhs!%OzTyDq8Sl)-OuhA?nnzqZDv;n1dlb=?+jr
z^gC6!S5E5RaTu&<LxH=>9+7k6Ciyg>lu0Nx#WUP7kB|3KZ)5X)FyRW(ta!ib5>7g-
zSUCjzvat27I=2_4pc6r=9%Ik_kw%_J84-%7-_@Mwpm(td&2)p(hPH=3@BXL?Im~v@
zrpxnzYp?n^r!6fUj+z_x*3EM0meZ?Z?oxZSGKT0L<eR8kl-bVyp+DTStF^Nxgc4oJ
zX5)H}#n$LFJXSq;&^i&%VbJp=PY*-jVp3Ur_$+L~vd0YWatupZ{x)?_4jc<rKA(zc
z;s@4g_ys9X&Y7;fcDt``{K63b@s|>UUuG>7EpQeZ7M%#gcg_7BgHt66Hz4PLQ{U{B
zCQ2zh71-YlI=TIyrSM)%j#~QOsyiS3^;;xp!3QX*c@5I8yy~OJ1uo7G|3NdAtjk*W
z>^aGNe6v&_sr?AP5&Yjg69+^r_iH2_IpM5=s9#oi(8lJTR2KKE+2|MQnmOv4%rWmC
zUA#rF<>)?MwuB>lhVNN3hCp<Q^oO~?;c;`L9|!a#sbll~@C4R5f>sMOkVV>W^JAjF
zoyDx|2#}bSuueL%o==!p2f3Gr9~@l_-kv`yqzlp4aA>$bi!{9N=0L&W8{;DX?!l01
z+IQvF%^)}j@T_t&(cq^H_nvW<wTE7R9+r+&$&wza26nzEuJ*QeoiA`c#)P1aJUE=n
zZyswwGjBjL;(4VtwH&Od^m+<`A!Ay!&->J^QOcCr?@|@0ctp7fgS0Eo!j#hnB_e*R
zwmnqztaV=I@i=KGHZ?@L=ZwfG$$lbls%rY6Iqbr#@C7vQHJy7m-##XI{3RDsjGTo#
zJ(iB*-Q?~{4nZ;k?Tj2u=O0Yd%S6TBRNyN>1u79tt@)kUm>51&?#wPh=I|<AMY&~7
zT6#BajRp&}@r9btxQ$JCdw4UAGXSFYpY`v5T5+TZt;DK)TW5M5m9=a+%MB=}EcFL|
zAUdo1M4{Q7Dp`=EQ~q7I0mxK2he}88iuP39OG#xxUTmvp?pC3z8COzW<xAB;e=hNX
z<Y|n?#do=YoZs9OgNAl5HNEoqmjj2XJN{b@*RRI5w_u%$62=TU@8$-<ZPk0WrWF?X
zaFZGh^aexENQ&FphoEvOm@{lCCk9f6TKQyYQK7LcYt%&zcjjo}bg<#P@x0~d5Hg2K
za+P;d{bSYKu4r+%UE|8ZeurNVTwp<FakoN%UVY8_gc;AX(^>njgHdvHD55K_+ZY~(
zL=G{SfX6i<!3UKv@~#s-){Hbsq?5+6KqLvgFI7THE(dK$oRt;8lv6y$Go&xcdScUb
zv4x<Chfx)eSEA_B<11izPBf>AslpRw-<ft(gUK{?p;grX@s7%W6Hcp%fN=Whx4WtS
z3j2_UJK>A7SA!U5eSFSM*(Vfn`D<}$qq}^{XWwbg+v7K({s)RL`WK34`UVlG9~(r)
zBYj4GK=__FW^8{Wn-E=$&CAsh@(>)k9ujyW&@K32KU$Q7TQ~>e4TXkvMqc}MK|u77
z(6aHsNdeZIJv%hNQf+F`_>;5^7aMP3sNAop;)Jpo89};UDuXet_Vk5MBLIopqjJrG
zq!BIEpUZ81OdUyh>#;#6%p@TtF5@}3X~LSG4|Q<+G%)srWLc5aqpO89hr5O3P&5uM
zXm^-KueM{ZOM6U=+{3%iQYzmryQyk0?;0;>a&U>^ey|0V@ji<;NAeylU+CQSi=k5-
zx4J8=k;Y3jb9Pr(=P-!wwH9*sYLLyv?Al_n@hnn9;&8{#)UH0cdOwpghj3PC7@81I
zzRqCwy5vi-xF+V1b3N{a&k03u7lg8KDlJ?y40_yDf_L<!r60m$ZnNd`_Hp#%bHkTT
z#%QL7>X3EBC2Z|jzzh~sY6kty+`bei`VA{mz&8H+mBz#U+&!VVaMk^4H@-s|eUG;>
z8_cK@!?4Z}Zliq*HY@;w&N6?KPX7~n@2Ts)n+9Tl$9}Bcgox2p8+<#+BAbxe82;o!
zXnE_Tb@*<O(66aWSI9grq0#Eo*AQ{gy?MgaW&A5-*Mg?(2${Q0ku)};)tTnchVHAw
zvIbS|^j-|#j_Q?*jarh$Wv`7<W_we(PLpBWywG2SeXVEdNfY(hJjLT&T^4FMo`c|Q
z`DgsAo&wY;gL00iec`0n?~tCd=g&cAGwv96qW8@~cCFOaYY6V7UT#HAS}n`bej9FD
z^$IT%bh}p}g^x8ME_Li}>F!;g)Dc^rrrwKvTx;rn4Mn3W@(!e(v-L$PgS*0`N?v3i
zdRq@Qw#5?e`eIT3<3D$QA{&r~MoMl_ReM;wg)NnG$8uQrKEj;)yJb&iv3)FJ0k2;O
zIwI(2)OORa&}zQ!R-Lo4gmLf|mg5qs1pjnj<F$3_rl1fG@VPGD-8?__<27|aVYC53
zZQ$4?{`+VdX-{>hNURuK`JWwYp5+gWu}4U#303K^A=r4oUgKjNF^)e_ynd<u`ScHS
z@en6+V?&um_?f7}xP7JmRTbu2{d-Aq)-$BxTOa7duof~ejm1{|bMiX&I9p(;z+^~5
z(k9qA<&r{_+5pZ7y?{nv8;CvL#*e%$)zlY??vjs6QJSoXer88=Dm-x7y*+fKRb#K*
z>??<&dV^RB-+ur_>F2X9hj>{EZ3)VuWF3CE@Y10S#e5&IFSIb_dBQeQDPw(al=nn-
z5!$6q88)2anEPXA?{luw!0kD<X<SrLl_EJDTnRz894u$rBvj@b{flUqAR&g0&9~`l
zm5Y^Y%8Uas5vp<{95Kt_k4W;8mXS4r!FDfv3ZBnQd!_x-Obref-R(fe+eM(ZJ16A(
z%HeSC*fcIku5H3>B=8R)v;yxs3Sr|E0F5I4u~+d<I^f)8Nw4=FFIXVWK=CqVs0K5p
zFU4CXn$m5dvNR<b{t+OuFc7mcC3@bj=ZP%2<U1z3)%)T>3qN8Ie~(rz`M|&$jK@9{
zX9VH7sk2I`$?TBL*~gS!!pMQ!yPPK-V6B$1ac-kg^%}iH`~RzHO}g(JK?@TO<NlP9
z0pPyCZ!EhQYg3$Oai0jMEaZ0~Xams<ZqQp2a~KtI_>sZir3mFlUN1b4%UasDn7E;Y
zqI%YKT%iE0R+Ce2C0~6bxqwEMj4p9pbz%tZE-bZ#yCm(6#UiW^_z_Xb-ds3ytV?Gp
z;DZXf%+ztlUAwh)X?*)$uA`K4s1!zNl3ZiMdpBI%vnDJYINW<Lb}kI-JDsf^U2(l*
z+Jv~9HxnwllM4;LuZDFq$w6~?6S%*RoFtOAd)AxTab8GRBh5VqU$px`UbKiqdfwY+
zdzdB!7&z@V3PQY%SD1RTX4gGOo<IKJBb&e9%8rS7WxhfQ3U%>@F}Pg7yZv=vHYDQw
zd=OIY*PU7X5b50F_T?$cSg~7n_-W~vO!47zPSa`TPeM3e1841Hmz$R#f2-y|F?aCw
zklnJICc+&ze}-^MIwnM#upI#L-}=`&<-h5SyU2jwnli-M$Ypf@Q7UD$L1>re{fS_6
z1F2Jd8O?pVCT(#;55gQ*90uw?GgsZSy9Ox4vtcU+;rPS|3{@tx1g<n6-5T@x|9e$0
z;kF2@Ak9@Zc7!%ENi*QHZahM=89=!k|5|;e5q8Z%R-j$9m^!q6TdGUPF^3MaX`p3~
zy2AqzYn>Syj!0J%Linr7p;YJ`kZdFjAwK>Kb5Bd&`y&d)ja2&B1uw~OGt+bCeG7S?
zl#1NFT@T6pFm{a=qn7<$ZTOQ-w}<q&gHR>M=N-a9l?MCoiK^<U$)hS`eA@o1`=Fh<
zx90FhUvqosqF=?4EwmC{;WV}=oP$@DwNhtnTff!7NvhUd8{~TMr8M@1{OKKM89f1k
z+cGZ8ecT~)WuEFG{NCq39zKdKwAk4Dl#+Jo>!59w(olsp4r_tj1c&Q*Dbsh#<zo>f
z%R%W<e)esF$-jK0nIlNg(Xg9L4aCsiy4e$MwxQ}_NLO;BB+!L{z~I`>i`mqW=Gj#5
zm&Mf7K|*n<XAkR7@@e8v`5)!RpVP5KI@vz-k=Dc)Aidy&KLMa%GPqPdk?`l<ZrEps
zE+)$2shx?YKi9;6;`ddPIjPc_Tm+(z!E=Aa(Etl1d~Q4q^ELn3$S|vd7dz$2V~GhD
z#MwDyLAY}tnO@ii+q8CU%5=17@AYm77cKot^oeA@09@Do!%TE8{&pSP3_LTb!N<o!
zjbm$w!buoi9d|Hx^?l)T9thd1B@R!`mHfK+9;Oa!k9BLuScq2fJxb%`s+wF>e%*UL
zaEl~bHA&5NUZts^Wjw@Hf(;moegX0@iktYv2|fAqREIa2Gha)h>sFX&q`8T3w{cZe
z+PTad%%kT0HKn=wQtquM`OKtx5`3$VF}zGaoL96=`LUz$w=UVu^q8AV5I0p4k9)9)
z`8QR<-rp!=lS$e$8ODusrN-64QGXxq-8Q>y8)@gOs8k=8(&uf`0Zf-A58C+x_(Cs-
z8>?`No<#HF@e1E$MBO@*0^+NB$a#{Oijb2=brHX!P8n_TPl22a6=<r3Bp^b!#Nbzt
zs!k8vlY0Il4}t80{x@Z>-U(<Z_JC%9?YX5>{4nCuTArS|jNaJEmM|fgqq#?yBvt(>
zUE@F8Tzxg_J>c-aQ`@S*{HT#98<H10&F_{`lXMCy(oLh()%^l!W3S&*?wd8quurrX
zw>W|IbzSbcP@Tszn&1LfOFda;_5$Ck<85+fI_Vd1dU|_=awrMna~*>F)q}c@JwB*9
z;*5qg%cR>$HC6LlZsWvH(DOGqWvz`WhFaJQ+bo{?lNsf~DHNOYtq$xY`6@ZMbhRA0
z7kV#r#vu)hNOfHLb-)@swaoB!6{wJn5QKt3*pCH949c-^YsRB^)V2(EDE$aV*4njs
zM$2oR<7(N~373#YaZ}{FA$CPm1`i4p4QCs0Y11d((J(1pSSn>ivT~pBY;@g~+tpD(
z4&5|C+Btl6xW6R$iK;1O1W0nOk6q@N<+#KvUT^^%9FPmJ#It}Q?DAs<6g^`+3x<8*
z`rXMV2T$B6oRBkpBn1D~(K!B80>VK@GpC9hc!(S_m&$m}*owXVe<*wJc&gh!ew<K5
zksTr#$DSF-%!p(hRFu6XGLAhWl8DS>c2HT#-j1w7I>!o`S#b`@KKA}y?z%sBpU?09
zKEA)-zaIVJ;oWszuh(<hLglAThOrL_j#e1v9o^21&A#*mLp^z&|9>rCC*YIL5`YR*
zHO&$`iX6pDikl%x2{7~ZbLn@=m%d#L5{pr|J`#xrDxORvpe7-oI5>G*$T2aNV?^~>
zxmSrh%x#14!{`lTr&0){-O+Y>MO4XG(hb$CBB4O%4+s!r`F0uki#05U#p88~U|Ulx
zQ>Dr{GAxVz*l~NtEx+updyj8~zN^qvyQq*g5sEv92WHrPB5HJ1BKOOvwUTdQ&XT$O
zXE|sjXX^4<F7<oS+;K-5W8Nq39003)_nnAVLDY#-)1+5j1uwK>zB2K`Usb3rDZkIH
z3&dJ^nA6Dg!3BAS`nGlgX#EzLs*on(A5T7?de+-pS^_jt9X<D|oXcXQ^Gzy#3hHDO
zBy{J;XW929wOm;vhY%perQ^?>ueR6Hl60E>9xidt5u7Fx3%dBY<g0~B$D{R?^;fMv
z?6BOW#A3p$pw0wNH;P3M!wT2V^gx{lYPMaDFN>bglE1G<lQX|&e5q?!_|*GWDDs}Q
zaB{6M_uR4^A#^nudft=e@22UT%2V4VHTTE!1$v}@JzqcXKN~XoHzsA2Yuf=R--v<%
z*tFimc1hU}9-p@O2H_NlTsIGAzRxJVIY9kzU4psqp|yJvs><VAiV?KS*T}sfkoOzN
zIxf_?MeSTYN{29>vAVemYUuBcpkl!o2i4|h4J&u94KpDYo?L&o0$;LW(NOHpy_pZB
zOX#BuBQrdwTZV4hj*Zc!GXT3u6$g3DsfXN$TcOzU?HN$pwa~5Eq^*Qv$sDv^lpnwE
z#`aF4kL;n6@6CHcpEy^P#2sW%gRm!dkBaYRqCdTl_<TtT6`+{+T|A`RePCB#O!y2U
zepM8D@xJCq2WpRT`=F)@htJ|yJg;1dVS8|f>a^8CCOIohX}iUlFsuy|*uYpyxywQ<
zug~yta}eDnFw@@(5i=whBrx1Sqsci2pR_PxPe(IzND_E))~wEV%uv3e0WzoK1M7!I
z{~)}iZd|{xr%6(!+d?^5ki^7=)4$mrYO5J2f8Y9u=3r8sRRoQh{#47g+vENl)QD(u
z@BY8A{`SG|$2$s!-kU;wfy#+0L;W5!F^(;LcLVDq4UQM-6j_V(Bmu>&`Tm|d*gS2J
z2j|}NH%QLf6))+UEO1IBZ|1G1OPn34F3#NQ*I{wq)>Ci@Z~rd$fnz)8hkST@s_r==
zC5r1NZbA|;>vl(ReJ=nih0fWSY%5H8f~)PoIFB`pqa-&3t|-cwj!JlCLXO5c=Dskr
zs&v1;bal6eDqvq(xO`Xdrp}56{E93L;fS5Vc#6+gv#ZsFPA9!uL^?)w?DHqxf8IC0
zP-zEjvvvKo)2yxQBS4gBt|L^E{Km})(hW_@7&Dcn{XsSk!ayd`4cf;$+LRuWj~bq4
zKMB6N=m0z(o(f97OONRyz}3f?M4aq;_e1G5TEs;*6c9X>vX@>AxAif3GRBBv4nK$9
zd0AGSg2l?$zZxqm{b!;Ntv&zinQ}SN()?h0Q?AK+iTi@D)~M1IJ_sBR)Awpag{$J2
zzb9c8dwz?d?`~oX@3DRQ%L#Jm5rRL&p!)&j-u@2A+E4rio|sC<#a_niwjC#}Ab=x*
zS;1oy$mFg7gwU7Bi8?XH;~OG@`PAZ$Oog>(s$|okah_oot-IhOw@RJ;pobXZ@0Du8
z^1faso180(V6e7Yxt%;?^Q3B9B*AOU<RwbLbUaIqSst7$muO{AUY?;Nd!e(2ea^hi
z)vDiY0wuF!z1J0Pj#LZt41E->sKR(IRN~4ff!ss%Vl9*65=l19L1^W}1)vICwtmo}
z<Cw)HH6cE`Emkn#Yc^P*Ky-<)ffjV9uk!*}F$J<>erG1-_Bj5jXqN*TeMb?3K`)8E
z@YmgP=>P0mqoJ}mA$o9Gs%zRMYqYnNtJJNTsX2aCSuB`q-i*Nb%4f7%J>dOCHGI!0
zy2wfwo*O=Y#yM)333Z8OE8<oCv*d&m$@J^StiOjpaL<{r{l4Npf!P6jGC6Rw=Bbv|
z-qb!_TBBhh)Moq4$RxV`*4tgi=pWYMlv9!}2mhvU0ve3Uo7?=5p+tpr<XDOy9rU6e
z_f`m=$oe__hOSq0Rtpa|fKlwNfJog#N%xaTt_Y=_kHs3RXJ2@QnUi1drO7>$oOoAt
zcyS-Hi1kU~v#mYP^t~Xr+^^qf8f@4<amI!>?jq|}l4!@knxu;5Yp5OQznX+i(Hhr^
z(%ENy=`_ULo;6{=D|5CZpGOXwsGdE3dwGepArX-pRWu~%h`oa>$Qd#U5%hbX{i-uC
zW<cIeP$}{A({BeFh}4(*&hATIE90AJ0Z_-_Vnto>ZTmggoTpv3eP>t^yki7ndhgA|
zVX;E{z~G9L*BmO8$I1|*pv9p4dnoNnl?2X4{Rv0jr2h4njg86{n#8@srEiks7>LcU
z30e3M)?OVxg+0~$5eYMtv?8;xZl-VbaveK0Oc~8g@a%)7I{e__U9CXz!KR}g#azW}
zLVv$NKpFV{7puQ>nk)j-%*ox(p{>tqYx6P`Yw9Mi`NN?;P%Wd7<+^PSUlT_S=aKf<
zU#g~8>TtYZ;b9RPJ@=x1tg}Pl>zCbP6yQQqs}}fH1bNK=BKL>9R=48Xb4~qXQcZ3K
zQGCDApOBTZNIAv(8Es~<MZJEDk$Fxr#S*l#s2MqX2PG0YK}`7zwz8+Mj@NL_R_`t4
zvKi+H^qj>!yddpZ$Id#$?4;S*26f!fBgHHc&($sOq@}3h-Xz0c0H5>Vm2k_?$Mdu_
z2oBK*pZLUf{_jmyq+^<m56Lh^h?#sH;8)F_HnA(>n%7T}SJ4&P(tZA+>k)kqztV{3
zAivFula`jMkYC20(|A^$@Rf!d&mL!)krd;4Aw?p2C^U79`ne{1-GsS1y>p@X%g+H}
zNMx7&gywVW7PPHt<3P}$p|t?vPNwE%hs&K1?)R4=G}2#OAKRKsjmTKpH>Uxs2i;?m
zW`+<zEfNY?!OK68V?0b4%fu7|<)xhnmhVeH9;O6TzS7ODkaBB$3zl>Igbt?u!=Tgp
z?DwZqBYOkq<+a!NoYbU4lW^+-D)ALELpUPnIQH$q`LXzeIn*yYn2h<@eGoiCS#`DD
zb>eq`;LK<18~q7noBn(SwcOr;PBy*;h~c-ywgg1VOk(FKHT>{Cp<l5(e(@>KVjG=&
zM5YKO2R9t6a99KCk{at?*2x=(4+o_sg#p^Lfi<gheWVaA?58X=ocjd00gMkT7?UxL
zoT9_A1NeycU6g^9n;=gGuZ>~6Pb@HqB{O&X6-{pA#g#;i3uqBD8PgmIn{&iC{f)xV
z&Nt4`YLg~<AA8~%gC&~-B<VMh7V7^vO6EmvtT%JUPf;Z(20ellARCe3n$O+9s1<EA
zq)oqhdD0=!`Ez2y?Zz!0XR=-gwj%$i)m;Y&1MCz$HFGh3|1MLjs=NXnE+MCa)&Gi@
z;fd8P&)Ol7aBohC<(bM(u*-RN`HRCXrM5TZ!(ciSRCj6EFJAOv-vzRZU<M+yDYj%{
z16LCYKvfwKSF05RyGNH7<(b?oa0GBEhRg0zChh}_qU`mfKW2ut#$Pyl<1euc3-vuP
z(j{r49jM`&w{^@{Vj%+RC&HPO@6}_KF!a*+)bEw|J{Na}=t4`aOWb<ryG$M3M)Tha
zw11djFCZ-wYbO76>p46=dyxY6eT)+VD`A_Fqj=rTDQ~a8(u=hz2;0O?9T<7C=TF8a
zMDDuXyrrvSGh#LhugHt`7^^e8`LXLZc9!Yf$osoBsTn5dRzZh44ju-5uO@~}50({2
zXt7~UuYS5!<h-X=TzX&Pw~r@dpnjtuO-v6Bf_!)T#vUGV00wBj$*al9%zazX5Imx?
z(@&>x{704`&)i|2J5sZ|ixg8V%1|WlHOSi|E9tD(*Ny~rCKxj*ELK1s8pWHL)aM<k
zk98h!x>u(?f+~AfHsrvN{o$!1=Gl{x;HhBA=V;lzl4nlg8aiq>l(}n9m5EpcU*=&3
ziWoxzO)Bzuej)>+va41PyNk{|6AWmO7deZGHcQdL*&IK6j%_-e4B}V%W8NMk{%b-Y
zjmzOP-mzog=1oP5Cg++oMtfcv^tqjee~RK-bD92Q(*6w;ZDcq9MfLt0+W|c5?d$Yd
ztmt)L&=Cj7E^gUSRol#r1A-Q!j{!b}o+%YN9m#^ezG_!>Xkzyr+MdQDBife+e|iGr
znx?KZx~_pq{$d@IH}E(mM06M>)EnRYF!SThQ_<8E2QhbZUG4}=o==uWO#|-c0U6!f
z*uD)<UdK<_qfS1F0xTKHRYC$hLwG0AYS9|5qa=SMu!kUU0p`1c=sgt$hPJ?I7WHHX
z-aT-~wz3Q60WriOqhU2`SS2{oJ$GDe;jX@H(WFy2cMIyeiUwLoZla&^IV)}|EYXLZ
zXtZsYSwn`+rJkoS`T1i~8OE^>Q#Teb{b;ihhPYS`qf(HIG2mMJkVx3Ex<`WT3kO{p
zr>CE`ehdfN+RqZ5l9c{6r%Hp96Z+YR6$xIwJfaefOJIlo`9*#~3#5b=T6ZRnkLEJ{
z;>F_K^WU1>?2yHli$xNlXE^FB*e5;<Mg-CP@`XxU&IEO*t$iPB>h!g0gFzsfQh2zy
zaK#t<1fbB|lI=jhAXzr3RLFl)6C)?N)aNf}(38j7PHVe9T?U|`2IL|QA{eNWM94ZQ
zSKr{NXm6S4CAm~=+7aUnemXmHfV=*@BR6-CPru#Lv@P`-Is|U0!(Di=+ZO$5@w(+g
z3~#rnZ1M(0!GmWILOb3bYweEqEt~n<w?UMdji6ooo&#)tFDIO3E@w{Dtr7tU{le5L
z5}ZzvQFzAky#Ui~JxunliJ?ge$-J~9$RI+G$}6EHO6p^BUYTr-{;ZH-eLL6ocuGH3
zC=$h~AswW-kk$_J05cFo(`gZpO`Trzvc~jO7!RS2(RO@s(g`ib)2{Z1bsoKj#yZZM
z*9knLOIO{tf>gn-%CyC~)m@rrso6pw2H>FJa}VPa2!*Z{PLbfaSPV~t-WA(c@^F&y
zHX;-!>U0?sixff4TN0dnfKhp5&64@?92*zsKX+RAOZC4fUiO=B!t2#NF6KV@oP*tX
zY5U&QDK3xieRtgR4$4}=XG_Jk4b0xZx1bZWb$p2VbIqEw&xDFj;QnHbARMUFW#}f|
z$3dW!a7`{2t0`(IE?uFdv(>aW6P!YwKm#3tK%Uvy&Lt$Wg?y3w@`-l#1NS{&A$eb+
zbRV;~OXQf%`Oc7s#X~R!oxEJUlY9+zKu7k}gdb4So{N*!R9nuasoC%DmRPpEBTSR&
z2>6b)!$+xpJloEtx&k0jdix#1<-?V>AtS5Roi}gehCBpv#!UEhZXv&c74>7%Ji02y
z<azt&Pjinrg~@$~QOemmE0lPJzV<%C=xsNlM5$bsn)a`6Jb3$Z{e=&|fv)Amh?YnL
zVGyS@R^Yz73%|Bl<^!8+omuo16WR|*lE0LY5SfvihHGSc8ZT*7?q22W7j!PWNj>|R
z?P2?62$A%cQtzYQNOYKlC<YA_N5d%@^~7Z&EqxuI0m~~U%w0ktaH2;#8?B_%-HH`Z
zO2r-pY2N&U*7Xkm3ldkhIhgN=IXLW5yso$V*hK#{H@cP!k{7e~(xJg-7Qe~yMdY+<
z^0)W*eiKcW0|^y&j~IXH#U3V4CieTG{@)MI)&v}0zYMsbv*j>Y1}7F&aAUEMQ!y+D
zj;AOqMSXoAGw7TdeBi{_eO_foJ|hmKAz}ptQSTfEQYzJWM^F1#n`P0ZBL`!|gd;wL
z*<}HZLqjGQ*|x@MP+6*v<m92r5&c)O4rWJ9weoSkOdnho26?v=x4+OJB9vDJY`^pL
z((Aiv>e{h84d%D)TJ!y|i@x7yD3#-_rrWJ~i2-5FW}VIvf;ai-!gZ;p)p&?Ld~Q!*
zA=kfWAGXK0ke83nzt$08A`!HhLysufe;NHwfn!(x^W4V7WWEJ2WayDW@&dHTW(JRR
zq<!$>p~ggbP%+i<VjSm>8+Unt@ciTxdAUv{g^hzA3lLkdc44)LP=KPEpq8Li1nvM-
zEKIF^;hWbAuN6OBF<3QMk5E4_!klkb_>6uwYxn1hO7sg!cypaLd>id}IXR)j<;_Fp
zIj@xX<c=1um>$a)XH=6_JWyRP-T#er-@MaWAn>nzMb5VG%?2JG$|$QchUhGxgC7AQ
zc=2<<Xb30mIV0D`Y0Jb)qvZQ)mloF=?I?}90~uGH*?>KAb3g=Jzu35=*;(Z0P$Hpd
zCu7b{Vm}j;4QUUZcaq7gvt235n8jL?+rP`U7SWU`{&}@^Mo)g0GZ{G0Dsve!+%@Fr
z5E@swWxT<3K)g(vl|^ApTF^N=TV)giPq&H+SiA_7xcOcrxY9|pQP%;NK6C#iT}3yG
zg8k%f+XlOWz_bjL%@mDRp>OJruo-01UDNJ|5q&pEeoQOWEUSyf+yQY8IlXdhj=-VJ
zM}`_!M<tihGfCzBkVtNaAre#?2JgqeYJ;y)p4AZM0aLdu(|H1~^TUK=y`?XJe*js3
zxir*L!y>|!v6KKzbRSugXG^bIJZ$xIz58VBynl__;bwvW^y1^bKgq%IU+w`aI(F@g
zDcRWD`U;m}y5N3)iS<RKQKH8-P@2&YtvB#W0cW)LzN_<!-PkX??!WQwe{#1cs@byU
z@#ljO&%aiqN1>p*D2EJVrn}Y7aSkSW9g4whl*6@PTUL0HZe{|x(61B0ZOUs7115s)
zTO|yJ?=~jtqVL`ITz~xtiZU%=8jla{57fDUh?MV2SGhr^lk-FG6zi>*%<D!Co_jxU
zwu|p?#rSsp%qV3v-$><_R`0%mSVzoUbgMDRV{ea6K6_!8Rw?y5M~;*1(Tv#H^6@m!
zrGS7D>$M-xC#-`~YcFjXZGDDo3H1RlZnfR$ojVnJFFhaNhwbOTNfO&8lf7SB2XLNW
zW}knr+0anYBCc54bR}P5k_J&>YMa^rfNXLIXp~I)Ly%6ZR^`XMgmVOEgn9h+hMiz=
z6$|o~q}JzQ=P1ZekrN0LGhC(@F+O#dv6Q&>>NNjny6_yB_9wf`HPWu)(n<PwpIDZl
zTH`+@5<3*XW^C02#Yekw1UvgcdpViQIT{a7-MbZ-C72l7Pm9~|^pgv-=81MmGmbx&
z`+e=XH6jn*kul8uHWIu#5d8cb_o=Tr@g9Or>A4{p21IhJ+bdjxa&yW?X<%U8A2(fH
zSIfmHD3!D%Ims}9s#dt1yQDJ%i%VFzRD=Cqmt4&_{%E3bav`6KFt5DAmO;bf{#Swa
z9M7~z2Gc76E(5F``}EYw&y**M7rMzYYjkKW-D3U3IC7+(b4G%l`0J|p$VheCc~qh8
zs7Lx@0PjxHqo+n%+yfzSm~`NK5cts_7hSQ|mK$>xlV3kh>uh>d_kFUuFK55cO{~dJ
zQIofp>v}S^SUiY5sA^YQ@+|Oc^PhY_2!0=)Z7jA@ZNz*DB4y*yOpfLRf`@`d50wL^
zr%Plk7DqMeP&j$Z>DIJXCIw#K>(fs^kpu$`?3YpzJc+2~h@~a=WB9<e!l>b9Hz}6d
z>o;zE+8`Peh58=&kr8f}{MVWa`Q_GSw<F8Pti2!3e=GOu=#x=uX}!9!n$TSKP{zl#
zaNavmg^A~9ww=dj!(4#8Qpwxpzg&muAlWLv;f9WaJGual@_aZuK)2Hh$Im-9zOHB#
z_80BWcOZx*#fZx9e%KAgreqvk&4m!bl#Bvz7AsQuG*2D(rIBKYL`DmXJg^xQV}awT
z-XX3pU3|<0oHcqIU~>+qT(gdoe4=RDU*!FKM7^=y*q_M|`fNld3C85lk()gW7ivjy
zzT(~s>5N3&YB165R<&r?a8eWMG7xi8Ar`vj;kZ_T2-icxh{vN}b?S@AU)NpMD8Ih6
zwS!l{%GUt*^`g$1s}G_YWQG_MhDe^9t@ojKWF%0+mmupZhh8#p-ooUeZwoIR?cTUH
zI6InwjCZ3MlPcYGKa++zQ(qa`)F8Fsoo6uwuCCpY=p*~HEE;nZ(IRKo%`+_y(!v0w
zLlgzapKXp0<N84IxI`_0WR*@qewU;<9!uRk`Z#Bo7$?C}!&$Aith6lh=tZk#d;$=`
zQat{zFJ=GBzsnvSAC3t<wo9|TDLvsGz4JnnK4v~HI<gPHjb4EGC_(XE;g}i}`p?J@
za6|%Mkbk-N)?lZ&?tgV)LgGMGq`9(QwMp^#DKU=y0>b(xoEFi(ffdlW=o1Q*7L3Q`
zObG~;Lk=MKI9)S)H6hSA=yW_{(xh6uePjQAG}@4H?h+t7;pN(=*7TI1oQ&JqR`HBs
zdkn>h5IDbH9X+DRtp_hVRv;^(oxOrXg5jA4O^j1s_K4zb2t+5n>Qqh=#C>-s#%alE
z@4d5`paE`=k|9#e%pOlk)=vr3?Yo&pOlxKUXtL@g3lY;hTjLH6QO%8r;vu))9UAbQ
zF$q_T@<Wn=OWt?3fp=(ez<cRY;<`Wl$Om&kHHWha-LN_H_`C;*ixA<4i0?ks)6)&J
zB(Wm2xb8|g-S(KI9&u3MdvBKn8xls>@VWj{IM=12%hg0>kEC50CQ4QA_w+Vx^|Yw(
zpA6dTe@Mzn*T{Z}Q^FH=`jgor<c#N6PwuvIyopTtf*h-CM$2&Cn?#r}-{i#XZ<h7r
zr(>kVObaR~b-nd@rjCw+3l{gM*0mA%-S15~Z{vS%Dr6!l2&|urg}mxi)t%*pC_-Dn
zq~J{2C#D=y{?vTe@;hV)u7rvb=)OLjWHXjzcv!U&M28X0XL*_4J&Q~XF!9#ijw3_4
zQ@dVa$XkED*3dZn6L+b4J0&K)b<QrmyvAyhQ-s1mV=zZ9z2ThQQy%9L{~u>ZDrpg#
zO2a4(BMB95p6%>05W8}yZpO{yJ(x)oG#Bp&{Jh2Y3eBVWw=XQ%mzb*0%Ad%?OIUAB
z&Ont?9b{bDH@C895Vvo_)~xS!L9gpNV;V0(eTxka1Kx>jt^Aa~llc_y$6lIO1;)>x
zAiLxTqnC{mv7|K9k8q;S^Tb<MGm%;ySgdwzf+0Z?fjO|dn^S=}f<TEra?I4n@Ejro
z=hAbZyQXS$?;#Sb{mNmQItte!kS3E<|6%M%*%iosVF+svLTUqcz@UA>(+KV>)G$b$
z)<Ha5se{g=O>}Wl`M3GHp9}*KY*;cUgstETbhpp6KJY7Ve{orR&DSWSB`SRc8-X10
z?r;a1^x1b`t=p$FNab9cInKqDIvu_JDcPS86ZxS_p^GnHYVvyCCm!Zx$0|WP{_JmX
zE#Yxrh1azKrX{@-{!MmiH?+P)qBUY{0!w0Y^z)J<u=1*cJcd=z(`Ghx-zH}u8zEIJ
zFDD~p98ywg)1KFHw#WlRT9qm?CY1)EsUhJgk#G7uXI%nz_x(@NdL3%-Z`=7)hW(?+
zIC$y`bbJm8+%p`83j^DblH}zqQ6UX$XY!AaB5G;F3M!XxHYy153`M&chZPSZ3&w{H
z)9a%)1`UT=P_E1%ZQ$@&K$_|NdFd};AOOtwsbLvcA}ewWx@fNEoLw6%!we6ki30nP
z#XM@~fi*_VfC$BUe(Z9#DA8C*l=*@y=WY`h#~&N13)NrjLa2{;(B6qsM=VwoZ3{Hf
zao*qEn4?YMLfRJ5o`r4%_kEbWKR<N-7ih<W{={XG1EgCEH^H%S)@qP|q6%lO+nzW?
z6r<tz$8)i-4zQ2y?R`9+!0vgO-DkzwS;X6R92%pn{2r=_36)5_chRVw51+Ae^I}ao
z+8Vz&S<?FhwZ;(qmUf#aren`w$azszJ6@-4c~D_yeQc;Yff(m-Ud5$t6Q9YaPHa2N
z{orWyXXuz%RW)sZDV29BQoJHs)nrh|X7&oInx7}xq3zt-0D1M&BP0ip{iCa)bA?UO
zV0kjrKmk)sSJ$Vxbu$nQxL~+q4{zk4oU3`m+7AjjPH=&H1tt}Vu~E@?$Bn#o&ce0_
z5HSVKYjv<wp~2@l&z)gn2_~w4Ri6jbp6D+nDRrAl3p*DsLE%WiAEL<t64^M*MEo>-
zUGFlDe+<V>l@gS4mR<pj-04!Kz0DWk&mw$0(?p;C&-Wl*d-%&&LB+tH&9@SGybg0X
zi+}CIdL3tqXUxpgz|3E`j>E4u-P^A|KvmRseL8vi|HWC6)=yPZ`WKtvZ!}AXO0(D}
z^!MHLD_LB82j9jmTuh(^`wXdEZEQ9RP$ImckaK73Ph^{Km!nqDxTxe~J7KQ5czvcW
z`N6lLgADS4T3$SN^2Y>^L5=*6UF(Bm>X4}2`|t3aM}zLs5nt}`xw!}37uz?Z<sAUf
zwX5p=<-*KHVfx#!Q!Tm;k*HTDXWzR;y|n31p-$JaPzikj&O$1W@r{Noq-^fBZ4z8z
zlU5@zGuv9Z@OIHZN_KpYZ)CZEC~#Xdvx(?fD$}|?LqlvkH|qG>q$Ke<^6P`5IE^}S
zkR8Fp%C)H<t`Kiy5?K8+V{{(faC;8km9J*|VD3;#!2VMSfyaXwc_W|qhwtY@vg;x0
zP+xTmm6oI_;EfKG0C59(>_m5&$~af)n&G7Jc-JY+XB?Q&?h@UVg+DdXiEd}b9+@At
z2g%=l`A@bb)ikr}FBnG%li7IS!=sAR*B7Em>;N#lZge~dTe^LlwYZ@~(_Iq}iXnd4
zqMZ(U?ujcPV}4G{EOg&aL<W5SQ-#GQ3^8}s2BL)NHz9C2#=h7mJ_oIYFeRO%Zo-~Q
zJ%hlr8Xd5!a;F$A7(071E593sr_;p()`jo81oPNiy@C0t0+I=n?xOZx$cZ@4Go91v
zt$%t#!p!E6)<LoIJG`G~YHvHK`Hy?BDhc}-uBhAHu-|@_DVLt%6vh7(=NC~Zd^tJ1
ze@HMw5+=4BnTZ>#^AaS>bE~)hDa!P5?RNc7;f(IqUQT)FxO#!h@hDw>Qx4+g#^{fN
zBnNxWX=N?voPC3BF_gFbMW%raKd`sb$TjxtYCCQu4_b>hnNf~BsJbu5#ZiNuGVAh%
zqN)_<m@Gfy8sGJJy|0qXVZIidaL7gj`rSZW#|2r&YT{}jFW{$gWJ=nw!pE=Xe<qC=
z>!Ls1`1o!JNV>V|DnaGQAxT`#DYXiRXJk+;{uuinbljuJC;ta98vaY-sBD5f*=?GR
zck`}Fgnzch8N}n+eFaGDxg|)apmIU28_d-!SMKa?G5vM`k`-hiCH^n;!2%qAH$mb0
zE+N3Io3n^ASd`)A$#D&@76EFclp5I0AwY`hp9%0QugV$vq~jj~7wp|K^C13alYYCf
z`_sL^oI4QzKviAVp^jN&E;^~>&ZxYU4O6*l{Wp;cqv%e8&t7NY1<&qQm@tH;cp%SW
za1G&Tw8918!%U8Mgq_!B>nh~snWKJauDhHo092-l(*7W7ac(xE?z+p1eOJ%{5iwTe
zF+b7Jv!b$A$IA-Yb}BaT6OP;)Xh^+)ph>PrwKu*(=E~Fz0cx+iYPKq8-lMAANL(J8
zIy~yX*84dxSoC4)Opl(HU#*H3?6jAXu9?45sw1@~qk)J3VSTI#(Bli-#Au$<BuWPM
zBcR>UD7@yxQhKTW6-!fe90u(l>mR>=e5iOlJY4!8DPt(rq|mRo-;M5Q1E;<XX0n^O
zX`jV72yZgz;G>;2;B%)<WSctHp?QikhA*l)m*+Dt{U;^Q^Y4-QbM(2|MY!Qf5TUJO
zX64x7^Fei4pKrjMIT49uSvbE+_L=BhjFM^~7Lg(cf*O$rH>L=m9Fsv$k5p5Ek~`?6
zHjHVjr;zvXljDd$K|5mRK?Le(Ej;7Ln5LzU?d&~0X=C2FsF&8?sNEco?>um;dgh>&
zc4|DKTEnW;(o7^1nE!@u`MI0s(K?)a7(suZEKi<&Kv8B;-?h43?d<A3pcT-mA;GQn
zR&DEPHsR7dj|G*W4KV|aOjnBdua89EMxx4;`UIaEn+ubn<*)F#GwCjtHEwGJ=jV+Y
zZ;9K5{&3UcO_h3(M8697o|OtWdE>m9pGo@<wj%HlUa(@2*9=|*?AfIjN`?MX_g4C_
zDvlwM%vcHw%jS<zA&Z>{&E(+>j0~l*Z?y^u3W0EVNX_Y*6LQ3vVA~_gyd~v4{u-RJ
zS>ul&6{<ICzizeSPvBO&L9jo0d~k~wEDbi?>d<PbW3pXu1M4?<ZE0_*Ue&f=3Im<t
zoy-alg|{v2o^f}r|1U24007kZ5nysDI{cqpj`tg^C69XEAP=aBac8V<YJ<QVPX7Iw
zUcwVoi(=qS-3-J@EhPVV?vTtWc!1vXb^zO_+%};vq_yJD#5QcI;*vU%$@kG;#AkHg
zGT2Ugn8<{YI(&NCIySCS=InBwkixc0AK`q6x(~7_^%!8oSGj|<WIvZA!`%~&OuEoR
zJhz-I&*e)tc4rT8qJ2#irK0@zr44n9{YVP-Q=T_EB=S@Yo?6$+?2_oOlHU4KAn}5K
zNNb2^(1jE;he6{~DxDu4SG3_93;cXt)JyCaDow|EN^AT&`9?iPcdy>>`@~L>$dZM0
znsKokg<7b9$;`GnB;nP8a(fBUYGSh?N8`s;WvlPaITsB?j47i5PEF-VAj{~b#=1Cl
z-KMn!7JG)^O#FE>U7bbGKPG0?-|tOvH~2C?2%LhzSYQht3Zsv3UXGqi`BN-3iNTax
zE6y56>wm*&CxL3)o2T&)>zOnD@-R2N&kdANi2NIz*}i}{LlK&cAd@?D52~#nn5+1v
z^1z9R^^%&BFCz`&+Z)HaxLMpSVqEUEZ>X;~B!tdXhkp0d5+vXD!}MnzLgPy>Q?t<^
z+FyMIYhjkP6U)7*vM&jDwhOgXqUSzU#rw;1arTg7=4G^lE4khpyjiM8jM5+$RIxzB
z#>XhdH9j%()zsfSnF?WVvru-tIMIro>291+;D239`#Dlxa!)#`b&MKe>h(^LCuY8D
zV8iGjohHoMbnf}yIiNPAt0A7T9vUy(wH3xa5>6(%#$n2%Mt(d2(O5SZwve`{2Bshs
z2Z~FX2~L&p;_j)Pvv@<I#*m!{t(k=N(yJt4I06BO;V-r9U=|a#kd<C^kf!keR06Fn
zRhp~$2^0-~z9b627ZzvYd%iUBiK#4cw=bn6Bx<5k(Fb>DUcUtx%pa10Ki?0R_Xowu
z!GDj~;Ky|j8OnG@6nvF5ibY(&T2IB%p9f}F*hYKSx;3lHG@dA65C0hWt8g%w#_~h6
zg7_CPF$Mk($HKbYyOPAX+h3P*Az5yZB;}!Tq5bZOx5wT}aX7-Rlm#hkIdl|V{ZUzx
z5t}|FxTUvs+YH9FsA?nOA`??$T_TSjDCuWF&^$xEOKs38VbEWYruTiJNs2+)7-v&5
z!1+2v)wiix`$O%uuDDgL*t{NW04Y7i{VM)V>24{lV*wFHy50Tnn63FE<nv!@EG%um
zDqeW**zbRjOSWg@!ouu!qe-p89FyntUB(~RzSI-2;A}qT(e+PRc)5OeJau(ZWBHmh
zsZ$sDuSPaaq9E$4gfEC|Vp=O86m~tRxDa{9f-tkPUWNkbdI_DjU8hbrE97!o?@fip
z2(bUD{tWN`z2(PP#AxtHHt(}a3-PY>CLyvmqE>fkS=pF8OMqAMZ{G7q9LP*_1iG37
zCWcoPv6q-#M{vd602S{TRB%A;(-p;dtyb&h=`?I8lsuZGAXP3KQn2>JKOtgOet^$<
zNuAI>SLz@Td`6sN_PRZX2<?0$$Rf6Kt!Ni3t4*{Rd)Is8RPj&SCC{I%+A!x5+^y#`
zIYxbgT$a?Ouj!xphlr6;s0gXemAgAO<!s2AYA$$!pgc|q)DKEYGshzHvzSziMQJo9
z1&}czu6!TT2I5x%CKFMd^JD+C5;Wm%$D@qPJbCktAd8;+wtqxkg}+fasv-pL<EnLi
z%1Aet|L4if?0>I*620nYn!Td2-{6oV5#yNLWvy1<ZI-;Eeh@(Gn0%OV8DcoYbl8&%
z=~t&Ex?KNCi|M@OCt&GsCJcK-h44NPR8Q1nA4>^=JweRM(o7*xbeaU)aV_u?OW@NP
zLm%E<E;*t3oY~DO@K1<ppTO)C4myI4nIk^`p<1iI@O!WHK6&WFuHtxDTy)9X<~QXZ
zn@VHm{J+M-X%jgN84_LoxwBNKS<}V;F9^&(iG$Yt%i21uL$3{QK8t+=-uGUBV&<*N
za6{CUc11s9VuWsVrBJ7w*|~~ydJ#=)V&)j+H9GCOCxbzaJjIUkHeh+tcbI`t{du!Z
zx~`mQv7_XzC}w@PIy-|hc0KTw%Fh4X#CR6g+BdLvFpc;V2dzNjP!s;VJb7127C482
z=F>~zY<22}Ohlgz?+<Rg6Blc{mk#I>`>-LWG}4O`bAz(t9$~h{mQgiq*qEZS1OenW
zG+NtF=DKe2c5GC^vU_)>vvHEp%EKy!c;e}w0kRRkex@DhW{;r0hU|t-;A7pX+};@T
z$b1LyH@db{K_b(4L5^zm*$81+`Z^jd8_&#C4P2^HB8A_t&m>;ZWTIEAXIxc=i%K^(
zyIqJ+OtxGEibL?o?38FYU1;oH1+5oD#RCO@%wy95zv#5-tjqPG^(vq3HM`$qN2{g1
z+Sae?LA!0;<fd*0t)Tu;<$^l#k~CMhU&(gqq?pMNq?{>0017ewCt|4e8638MsNNL1
zYb0S@(As37;2IxgEpw7)q8yJ;5SKgS;7gn2;M4k8-RHwD;VcPGwY&g3SpMXq(k?9&
z#gc61ghp%cN*U)-r8Cd@3rcZKr3U)L;evGFF|m!Yt#fni@+GS)luGMZ@?;8u$C!n<
z@Xt#>N>7>E-3<FGdzhT6D)XY^U=u_^l6_2c9lL-~cN@KPXt&ST!=UVi3gjJ4is-!8
z1=fF!8o}KZ)t?cWoIDj1DlF}9dI98lUnO~6-L0YWIkThaeCYx5D(BLJpckwk{CZ}L
z#NlvgVBzFl|Kb`Sc)czCq_X$p@7F9~7V6@{+(gD_oXMOl?)HNP|0W5}Rp5mWdno~m
z7<IXpC7>LCR^cJE`(73m;-vrGZ}P|Oe>Ci*Yp4JI2(GJ>*I`X$j9k_X-Rc*htj3iI
z2ue$x!d@8)GlNFbnkz@R6ZK%o@<iwzSa;f{<?ooSxP)MS@<fKz!>#L_kW2%O|3sBc
zR3POPtam(_h*?8UKo|%&94?(<^}JkM&aIh)f)e;tXRq6mOPXpncgaZK(^;O}$UvIN
zJsjeF@O9zq8~0`58j>juk&G3f9e~1YKKXP%chB!MdEQbg-44wBqce;P5gqL$xXMXC
zld=BW>rP6k-|4KI+$N%4jopyT*h|SCxy@8;GjUG|x>WPS!fQ#K7{?<mg&SWd#<>Nk
z5a=vGie25c24Ia&gEKt0+2|1NasJ538$LRa8qT%a8?tB=Pk$W)<TJX+%<)-WN`Gw3
zLhGv4wJWCLq!@Pjkq@Qb&Cy|(x+DoCqsRIp(T+Z79}T!h3w0~CBnIu+4B_oZ<q&qT
z$HhU(&@Wwu_AT#|{SoV<{M$bex<hqxO&tOGjMbi-oYh&-))Gw(=i_6GcL%Du=Uq24
z^3?J2x^h$JqC5ZhK?!pFPcrI&?;FE^4Q3k6+1?G@erV(ne&!ry3N?Za{5n18bsho5
z-VQY7>blh_aRLf)ebFyT{_6{ywdDr1E-xiy7ZL~RKDy;KnVAD!M%7m{PN>uagL1cO
zwQ=`)HR={{pSp*@8_gP5_z2&-#k4nuyrtlOI&Tsd9b-o?_uYiJsI_nGLbqz4T=3wh
zdpM)rncIb0^-E_l>Z&uEWEksd_^4{N4)4{xZEmTg8E3n;Z=$b<>8y2M0FRY^oi+(h
z#u5d@hj&ZqOPNF9_^T5o(;GZqd56_ohIv==5@eLU#2Q5_<@4MdjI&~kX`rc>F9j2I
z<w06XpN@nzV}eCDeH>J&5_ai8s(eJq36ty=tx05HclTmh2t%yY`Qh_sIyyn({jAN=
z*k$vh{mDOb9jT^o{^BNEpRMHyD^fRq-&Q=P1?I2o(C44OO%FqQZu$CVcFJFkzPo=R
zp(Y+%c4_VpS^A%giQfODR{VkFI~(dW8g!xVd}<o&CC3mw`pfASA4!a}KO@(#7~~<U
z0VTzht_eq}Ks<*_eAAGNnNq5A8Juz>Es4YaS|IDHUea@TseoTF876aObquNn3LD9T
zRg-*7&>cl)_REg^bSM^ePF}R7)23N>_>XHGYlFV2p0tI)ip^(Rd-Lr1V?oXz9s#_;
zTi^%hc9=rM>_2hI&n_UjdGg+EB0=KpX#8T-D`bQEIUsQ-0e)iBBRh|+AqE6Y8`kY-
z-|LRQk}F~O;?0qXoqxx(>N!M?zp|HERH?)J?shHNkm%Xtb1=XY54N@1;M*uPPmy!b
z1ERQ*31D84gu<)S_~%p4H^|UxLKhnCuC;W3*ov{)Q~Bag;2*zs<`v~;^dI62s^s5n
zm%<6A(o5(XQOMFZ{b-FbA=w!3cCc(8ka4r;+4LAut}4zayAS?J5FM29Tb22P>q>_7
zzriYW0QUNo&$E?^#couwbF+U-$v(h7=g8&(k=(>B9Gx9#E!DYK$ZDck@K;k6DlW-s
zk%5c-yk4~3r2KTuc;MY&O0g6P?o;F4PX3a)IiNw3fW$>C(9X|maEY>MfS?e=r(-Nn
zCsX?5AUy5+U+HAY(L-(Kc>UaO8C*OtqO9go21I{e4ynnOU*jv2-2Fkg2VdTn?YDsR
z>XQZwIb1DAFFjd%-dW7dzIbu+LkMW{m`mJprVZA7D&3IO1>2$|SbQ-r-!p3$Nxx0l
z)@Fhgf^zfNyBlRz$#D4=w$2ABTzlM|=`A81;%~Ggcr)x$uwXEq1%Es_fw_{vX}lqm
z83Bm5;(u&zX?ELd>TEO9)Y;ivU~yqC6Sq>7W+jMJ>|AWSFAUi)p@{`&r9J53qR}5u
zh#T+kC&ccA@1i_GyY%&=)p?ooz_!Jg?9hojAbJeZiooa@oIZOJdes#12Nn4rT;*+v
zvY9PNK}C6L2IDa%cS;}4HLxsAJp|!Ndl#|T!L%9J>t4uZMUiaYGZ%NEEU@q4J#0@p
z<p5s{;t#(3yk;tVne9T@Vb#&I>YI0T+Cdh_g9fcZ!EbLJbNr#)kOYlBSCx0pXM58(
zun*XZ8dEncciVE2EV%iUP7|P*S-Y3n)=<N9?V01Ppv~GFynM-5;INOEYDTCKn(Pcu
zvyC*H)2sL|NSjx3aSw(3talBlel-^B)S;mEx+d<b3D{A_tW^*XLBdC6QR^u7kmu=A
zfXS7;I;)uZ^W++}ychEvyT(gq8w%(J`!d!Y=jDsI_5;ogsK4KF3fXON*o1L=cYgZd
zk)3rsLw{u2Eg#&FobZs`m1r2aIIyR%kr$enRuj#1j@GY%Ty)RNM_PE2=8)!}C$Ii%
z%Z1a&-%?Intq^$oh$y&R#9@85r@LMepLBQpJs20q{L+|>(E_%GgHRRNm143UQ*>&c
zU|&}wP_&8G<#3z(zC3I_=c+yIe}~kLPNw}89`{fDgdhrh9qrP?#|`y<4TfG`cToE>
zn>zRMYjQ7$FrD7v#GPScWm#oZ#i}#0rhkXB#ALmqkZWG0ntRPo{Vi_OKIglGe&W$f
z<L0h-pqzz3D(>bXyg7Q$Ip->SKalEAh^(r|j=gh;xQjIB$6=kiIwb(PiB2)dRJ~$q
zZ+%vW3UTYhKqna9TAx1T5(PpgK3f!(Ka3jT4;NPeiZa!y2a4B^i|QKWy+O=&MrG@a
z99p_0ec<8d3);bb-hKQzoMU1Y)BeSMm++ybZISfPKh3mfj6UMO4k4q)<l3=9j4XsQ
z>-lJBEuAezEiDVm;pk@iuyc>k-3Qc)^!BMKlGY)@9r4d4L~8@#<}Y@buvi8ndIEaR
zAtJEQ6PA*1dJ~*KX{8b*{Bx0KzV-{^0**kA6{>NvOfL$pK6qLzOZzgRQvyc@z{2xf
zqbBRMT5!cRMeen6lFN+aAO63Z0DLm{y+$){YDNe5uaUNJ1VhfqL4k}}`Yh&XPlYP+
zMc9WalG{G>BbnijR~2M6p07p?nRK8f?-Z0&d9ewB<;}m`<4)d>4>1lmJ^)|Qe24xu
zp*HKxpC*xAl2=uyrK6VoQ%a0C(vYxQ$whikE>k%0xeM_~Wp~aaB~-uY9|+`_iQMR(
z&mJPfpiCk=Wk6cNiM?%oFnUY6n>_&~?(oH-OQfaBX8aZj?lzAQ*3E)C!5F^c7Kfrm
z$gaS=oNq#P+G`h2YcXINF`P3_=B7%h#!1j+P_OSH+#}Vaz0XbocrTBufPsv56<H)>
zA^1kK0T(NM&44f@ki2YonZXWZpAS4P;S7=x<_UT_LmtcpBoo3sB_d=PV}kcVd;3ij
zhJWPJ12*_qXG&<T#;--uwuO_8qh1L7g4aM8L8Q@bK*BV8f0riksg}FR$ITz~ipED%
zX8O?^S?T|hUj9i91FO?-YFObtlBI+>IG7%5Bg_h3lPL<=^>{qZIT&MnsBLDQT&s+T
zFi%b2<Cz8U_AHvrSZRg|J~OEBgTqybb<DRGzWOgj^yuYQd6-!@*iE}}Ttp4nj~G_a
zlid}hPG}a{Ai+63931+Y5lxQi*L>ru2YFoy*`!x?zW4;>|0jzrfzl`Od8oq{2s$5Z
z6NSE~LiA@(B|ZK&>RuHS7M?y-un*f#itH4c^!9@NR4}sc$6`ISxL6?x9ospyh;Q4M
z`M>^%>=GT3Nwo-k^9s2#)PYu0Z-#=1UoAj5=8Z%*N54u=2tIehIL2$nIua#`2q^7v
z+GZvl&|%3Vp8ARcbFGFI+3gVgRwe@ZN31}r_Ae_8m6Rq`BIWAjDeb23DY=jpx3p{%
zZr&9_md+u*1tK;+zRCD#pUnS_%l4={>l^oP;EC2D1su=o7#$+^QE7@D%Gxpm{@77y
zj7EI<OJZDbvjsi1=LLQ5{dhMYJ(Gn7p0=rXKNffW0_OtMJ(i#M0wXcM-geuJKHrwc
zheI2zZ-i{gz#hfj5u*Mhj0AZww7$vHMbFRq^V^&oFVB{~aKjzwd!%Pic%}2C<ZAQ*
z9k?(xgM5>cuFlh$BePL|o$XsCBg<tq_5`j%m-^`u6_wLK<+?F3fOp{y@m|QDIg66=
z=>Z^$BPGsge!)u&zajTteV|rt{1&mTVT_v^wQFc{f+eYb`NR3l5;AAfLd&~^ccGaf
z^#mv}3p!#P2*5r^9xHsAVYUs+11TXaeDbQEtbHr<=%}RT&Yw2E@Rz?eUCwEodeEmo
z@_5vFp}BldU2OKud6`+gCAs+6L?SR;ls<+*^GP19`>Niq|0hXG4>zC-m%Nh_n90YA
z?myB$<HF|a1Xhfl!0p9)#l+NCEI<hn{}r2u_2E2#CWB4MZI}J4Z|R32RfM&D===z+
zO@XSJ>~-bhEqekGudyd*MD12?=GDl~uwC7!W#qgelk77_^;Xk2t*x;R@8MM*8y)HY
z)6qy!B&*S$V%-Su#W$ixINAf%)^*mX+YQK0a^QMlNA_81ap~+wOA|lJ@(DkE5!1lN
zl%NMp8$j+5`5yh~2N&CR=;~tN##1YX?a<JRN+UTJhF|OtYE$d9(`{!Zb+0#MFAQk3
zyP3=c3<L>6YOphRCG<Nw*7Y}hilvu+Tzd%|$y?`#<A=WhLo0Wg*QYzN2@KI25oj}2
zvmNz9GJ&pAI2MGgJqDkMW;k}3BI-DI4uqBOS^kk*$liMHw=`7>rg9tjkj>KI2f4Nf
zr)Pb}xuM`9ERu%PU>}VA`d_bu6oxcC_}I292gj4$u@?$DkQH4!kvish4}C59WFPli
zf4`=%y>!5K5qLzN>FH)rBWBjHA;?9Fm{oIH|0s6?x#q|swao0XmGh(*I(T5aZZA;o
zaQK{&K^`JKvaakJA5XGlp688d|4xr_XDfHOkm`PQtM!+<po<PUvh99)x1DoTleguy
zWmu={%Uvb87`nO&Wu3>=)G?PzGB5MkD?6xrr4KHUW3IhVHV6^x&yZP9!@CS>3S}Ai
z1Hx}EKFOSeC-0OSOxLSxqWm0gJs#?7<=$jwRxC&aCPZF{{ZPATmBms!YHK&OG3}JM
zURBe?#-!Pj)eVqi@+9j7<!I^F25PxduE%sQG>Pav0$zUy0<Cb84-8b~j7h6v_~+Qs
zaISMNgYQCzmxIU7Zz$fU5IGvmzVuig-J*EB-86svf%@+`2uwtOJBi#FPW1M~t|vku
z!Xx8sz>^*3;;N<JAOY{0pAAg$^KM+ibH7i7l}<O*^6#Y2U!g$^XtceM?;W8U=Ts{{
zTZOWd<~f%iop$DCp~i^`Gk2|bI*j>vWNmmcG0r#d1coZID*NHDO%bZ>kY~Qw936j6
zozQXX&QCauUWhx_&<_sW5RNepB?G9*q7CdP)ImVOit<}YP2y#Pl_k#zEadc@AHU?S
zXrS~FyvRnApvfy`#+*sCSq06n0d=ip_2|+0O?MCg;GCp@OfA1-wTuEQ9d#fGZWYv%
zq(LMEV!lo-Ru%NegP)j|$suj3?2L7tx#(i5OefrZ<(C@db2+G`vNyxs+x?b{hI|IU
z1>sTfY)~CeVoM;)<aJ&FdT})Z&B<`JzIlx#3>b!@cFs=m-w@_!PKpMpc2C&ZM@l~t
zGOIhCX>TxpGSW1ou}sKvwA*y7y~~{Z&ny~BCH33!Fh&Qa(V3I|TGa<7t$j-<<`VFf
zsEA5-r#I3dH8Hrg$C0u!HQ5{!kh&-TokZQB^E6=de~H-IP7vTJ)X=m2-X$5ZOi3CT
zK!MR_()hSC-b!M-?^F7r2aM?M`fj^@?@Jznv<akBk>)wx+SE^DaS8rbn<+X@OQH{z
zP<J9V6tA1Vu(6?LWTJM1eatY(zdnzT(@4Gf?!4l2>Bfw>oUB+)>~=0Kg8xD#JIj9m
z?n1tqf4zG#BWU6gAE~DG)g;_a48i*dM=v&{mIeTW*blt?)dihX?q*HPS+s~4G&)a?
z&EK<OyFy37`79=J)zoB#6q6rna<*+Z&9<ju6EBK8_-Q7_Gj=+j^Ne&D8-1=N!27I-
zEJ!k+*AuVuqZv5BD$O5m4or2jM}Tz=COM}mUiiJ%Ev!h~Ny*7__+0D6x!H^@=YtPc
zdoPy=r2h+l8l(KphkKu#g7X9?f}isf?DsW5y@bxKWKi^%h`cn~x6axuE^?^mT0b&^
z;<~h^(0d8_|2NpjMO(-D&P9UEa4k^pL_8JIoS+h$pppi^aXXrNov5Xx^D-FcftGaq
zD<=j(5pW}8;%@)^qz4*J^R=MR%DTu*+1=l%&Oj6CJRtDEn>ww)=sK*yW^Ohmn_6Qy
zcH8JI<{BS*kA7kO>D$p8^_zLX<Y<g{;reNJp?zH$Zmg?g4)+o+xd6JCepJqKS}|Ay
zyO^7kW~lRqXgAxfB<2ujJ}d9ZtnT`sMv&Li{u$A!LbYMKCOv{Q1sy$F@8{qqEXIyQ
zoBQg=w`eHbZ*bgSTgrR#fkf1ovL&gL7zcXb)Egh~coG~@tQMT?Y4%j+U2BE}K!M%d
z(a9gb&|oRWSl#UeYdP|5UtQV$f~mZCv>tT)(vSa-TPghIZxVbt=`Te*V|-$Ti@_-8
zmAs)87Wp#YWR^*yo%3e513z*KP~HgK-*EV!hXVDN>(u)SPZ0U<nlWJJ2CsD&uxf!6
z7%v>waR6zID_XcnkOjLI_;@#yhhr7n$#zvvc4}nT-OPejO>Y<6O`7A`6<+<}?F&nD
zxQJruvHWj!R+rGVFEhp*gaEr#<b72hdN)#KQ2C&#@Vi>WJKR0dy2hNMiP1i34T_e>
zCZ98yE=Ro9%Zg(mzZ_>%#AtY<;SStY&)aXj*j+;j^=(Fa^K-3E+A^@DijrJ>vHz6{
zQBXMz_JRqfaCyW+mFh(HN;gy21OWcHEFcXatis9pma@>T>V>15^U~3d%ncDDmvWrR
zZO?gdK8e*L=#D#2be`haj2na=5SEhtrK{f17i$K)S;YnOGt-6~M_O>gDUuYR<ElMX
z>ig<=dz+{pc>oJKHl+R|usd`|^*8J{PGf^>SqVDg#h}4ih}jrRsf;qwn&6HxCI})F
z<_n?}<~bRdD5dWiG}4`%iSd`E|J!$I`x<}cfBG&xF<*NwL(Jh_?4%gVP}q(05olQV
zDvK%(V(*h1l#$Cz>6G^~?cRX-^09&8`=LFKj<O}DI}phi^EVsSiE(AQ=|FEgXWs&{
zpf8#Y<+p#AfpTz29d#PJ2?|m5NbG^lMXa!z+8Mg(siiB_@A8PrUB5BT2Pvos0WYvq
z-PQ|I%$iwS<2Q+QLgYmNA#N$VJ8JPwMKk2Nf$_E0e4a7+=EqAQj$$ubVsu+<sREhc
z@hxsV&#$OQw;|bM{FDEZsk1Oodn=YGm<SDu!3Y*_eCq{&^wVWwmfEgl7G%QgK>(c~
z2(S`c1wC%@=HJ_}B=w#3SISP6s_jVaC>|b~FTGs)ryPuweVLWN=?m}4+u+u7@O>_m
z!t8QjD<bgV(qp!%oS8|29ZiBJ?dT?f=*cj*CNlTH8__Et*FrT6G6rT^7kd8>6_$Uy
ztQ<5az>r$`q$BFYOu8S?wC}a6n5d_tlaSCv-o8%wefKfj*yFj4G-kDXAltb`#DdhE
z$uXA)a>YgVgLj<)ImRG|!4U3d=2R~^tlRnWbl%}ce5PQhFxSLwP?PgDzWk2F&0eX*
zH^w1&5$)v?!-uqYrrny_{aj`i*c+ra{RwHQ?NxM_BMSBI7Sp_l+*1$5;(b%!_(y`}
z?4GpN1FnN4YWH;bSgrav?X*XLv2O9rPEw3Ks9kBbnSJ>#BiAu$$XMB<&S+me!MvoT
zZO*n1ZdR~KckI#aL+M0|pgiwEd>*8*49i-bX%Uj!!`gtAOWV~JB(e~~$)?*5)&`mx
z-0w1$iQwD9NgmUoZLdJu{fxOFf9gI_dcU9v1-j!kz(jzLaQ?e(R1#kT;u0{TNkXTk
zG=|WkfiqX56gjINQxSs(P{`kR59d)o6o?D=lcBIeo`0>QWGhWJEb)F1$5%b*E`s-N
zDqmZkQuY%IhDcrNQBpgKmOP8;DZZgseYqIu3d+BGg_L-fNF{_tmf^xj{M!BIk@Gna
zseOlR8N>C|$fn1~nrmNT69ipkWYwu#M6gR&sT13Q+q!eM;hlcG;8z}-hvu?dtPkkk
zIM)x-AP9v^bMt!uRiUWu!<3wRbu#4D;`TTCB;$xSd8HBH4Sn?VRVVf~DBM>4Ua1N~
zCj1_NH#lIN8qu+xXHw8;MvlSxrx%a=v9>2^e2wFaK$(=WUw!7rId$Ll{20w}e`%KG
z!sB!zNn%M6z`*rJej*naAUtM@CMN)s>Q#X#A42N~Vuqw}vB%nR8TG+1-K1B4Zo-h?
zuH65JvbT<ka_!sx2kAyoL0XBSL3&65m6TLKI#ohqkdg)oX<_KxgordqDJ3D!pr9fk
z4KhRLAPgn^PTb!2!+xIk{=M(>r+c}UdtKKx=lPA}_#6r>&t94Z(#xi#8mPlkl8rua
zVL-w*w-Wdc&{vqL2OYL9fq<K{2`YgyO8bEr^imQUZF`p!YWXH+BtgbL6OHy74e7YP
z9Yo3jf;Z-R*|`)wHf=;dm2g9Q@w+T%7;Y>WwyeiqI5LDa^RNkf4a-a!j4EZ1T*&g1
zVB9eY-FCnCR-sqqt%>k?e}Igt_@d#>J;y5eb^)XCiGrXG*F>wv-@Ki^mz9?Qf5<1$
zI6pv&vd?igdzS@rd<R2hafMB8Ab5JMbdf7{gA`SIG)|lM{HMvY&ZX^$$GMulgDSi)
zjehmwN|T-aS3ShzbdqEHWCFmEBlW#0fk$m!?_L&BX9+i3Xnso}us-izFDG$+X#53C
z)(D!eRtL|9cbw}^Li3UJ-4Xy|rw-JIdVDGB{*$+v%qNp9CzNd`CG96=6S-&1waWe)
ze+JYBv^Ow393^mgW>JeuPEm`xg{o3B0uj<ZyBWE(an}`{`m{gc#wi<xAGVKmG7Iik
zHqyVM`0l|$PVrP3S%V<uc0lt%9#>X?@A)ZH?Y6zlR7DmL{>~Qae{=zadXvZvmhdXd
zb{VMY1?JD<&2v3qdVc{k&j^;cb*&jIrEX69;Tb#2xhRvKlv=wh<R$0e<{W_6C;sLF
z!r-!$&MhIQ-9v}Tu+gdLcIMkcaQBl;r+Ybu<LAR|Yq4OQld(RY;*mgtx|@@G$3(5c
zPW18?n_TmvNpwT;ib=c<Q|3j)J>0aJ$7)6E$dYesqXqy*jT;bP#LvII06aV%_o`U?
zmM!8mFW|&C%3AV%{n{bGI8OFI;I8^?DHeN`GG0Mxf2M<_<QrOo#ZX+p+W1|`6=8@#
z4gh&V{0Xbso5Zbd+B^cxYE&Q6c(A!1-wXd&dn0md`rEA4zN^!jVRJY1mT27R=vhJV
z*~@I&a11`T9xNrr;F@n9Gu1vW@Kbxko#4RENBhsbLtN7TJC5e(=zoy#O>g9yFX`Cp
zm22p6Z>r6>UqBGc5ncmi5rgR3*>@BmU~BDUdVKlU1-#~fNB1Xcexl>E%?3*CXKd!b
zFs>bbUUa?1t}tw%smtyE%2VR~L7~3VXx|r=Nx$dSt7RfOvg=Xr`oYmEt*6Tx=}jBi
zTMu?Vx!H0S?O@bTV=?#3N~Y#ZkjJMQL))sjJ7|Bc$r12Yat+NuBGAs>%lz!<?-;+i
z<(=&|)FDh>SkVR}|6w=u&Bm*N$HH+E$tD#2ZH4<m<g#uXQG%3Ar&F^`K|Fs-I2J6!
zI6-vd-u50D951h-rD2mPY9jcP+spyPg}82VriqCezyLN+b`>D81Cyth=m^`TvyZ#v
zk4jHL;`9VRCw+R67N|{V>YolX_2bobvZ?r2<Rjg6AgpetZ9i@oFp2*GG5Fu1AGzoK
z;uRWAe%9*xFHCXUg$<ACDOGLl$w`5mFft_Asa{h^fMbENMTCNg0+dvBr9dW{jjTY{
z-xqZ&A7>zjJeBB~YVmJ1EHiX%HO$$~v<|Y5(}!r2F}yf@KAgR-r>d*HVf`HL3~pD4
zC-8cUtGda@pTvtcRG+kKlBV87JLfi-#$T&@ztfh@oHugLP3VW=Hpty$rHf|jT;W`N
zLEa8HszP~Lm3M|cDjzmXSZeBe#kt4fTE+~Vv?UrATWLzm+Q`Vk*1zqh7*71&OOrDH
zc7|Ky(zZ|DI+w3Lbc28!GiNS7LXgr4;<QkV24EkMIQs)Mw2ytR&6N2n?T5$f8CwuH
z`>Mdntu);pbWowEGGMk^TKA8nw(){(9h1&npH~#*|7i4XZ?b`;7o|Jem><7Bx~ox1
z$!BWzm#BZ7#@l1HFbsT3pF&JP%>CNBXT>$i^~pA21-eu0-~Rpo^AMR6>fIMF95)Ek
z8G$o0R#`L|45D&jGo%Xl_DPj=BXq1JAu6iDtd<W%-89M8Ppb4}?@6;wYUu1<UFReO
zW8!Zf`qnafvD<AjOCNuU&wsb}Q(!!-a57nhu_O<0kdotLi)>X)zY85$*(MD6je;Dd
zeer&9a6@Kj_a&Q#fj6{dq}a-YlP7J&K&d|IzTcrh`0{Ii5M=XoA>-+kMbcx<H@P}F
zx%0(1PBu8m4qP!R@bVP~4c7YrVcc_Vp<-+Lobex_{1|Mk*-U>`^_b--Q)m=38ojl8
zWmPVl!<qIeDYM!-Gr)Vq;t8-co!dM&1DdFFdt$-#H0_M7*%ZJ|On*vhv<dK;>A&pd
z($W8+mGZkhvzxSBy5H329&bm-9IBTJLT7s^o%p}EUH)R?`YCLQeIMew(c*%>S~~G_
zyFi*1M3OKa&Jm3&Etli$6>N^%crVqwMD!23@S7yEeNu(N4S^jCu#0?f8gu2!1S)u7
zHrh^^QqkQ_0D-eXAWS;1=xA|<E+s7eNUckSCY5b5IgL0oi1{la+1-X@<%u^_cl^Pb
z&4O!$l^8i)c;*)%VK=Ss2K?yxc!uRt9MA=)A-IPoI+r#j(m%Gi^|tJzq04l+rU6=Q
zyLy1F*jn+!t*)f>_L3Q&fDZ04b5KOSTxEiGP;vf~7x`ufsG%met*^jT>G{Tl{$uw9
z82^~a3E@VP>p}aZQ%$vDYyTD0RnBI-k%jAGS)*!dySC=omKI4^dw9Oqg3;rjRV)<o
zGkf@roH&SAe|30T@(9F`04VD>)sdEzAhs$`NmzgmtX4eT{I68|yZKyc!mGY!vjIr8
z#1EvIS>5-nKjAxZOm+tD14of-5zm>Q##Wqlw9x4_(A|h)1N6A)u7p{i2pyGU)(%U^
z9%b<Kxn0mvk<KXM@9E5LDMos<Q_dY<Wp+On3x=X=;ZM5j+euy27mptPv9SEEGWV8#
zXSf6c7}mZsh;1GY{BO2=Kr92K{P>cBd^r*BE5w&GA|4R?doygdiXFL$_y;j;hHY21
zD^mD)MLamD!4K)p$7SBTKNg?)i0k`&t*g<S+x-+bZ+P%+rNUiJmHP)4^dAedmCQY{
zd7SZHE)B#A0JC@Q;IEQ7o1EpTc--eiGL*wF@!M}&7QksX7Iu-1y_0-El@Jrb>@>I;
zA!7ESyx3~uVxjCx5F&p95d(7R3eC)RSqFB&U%4?c=H%2KH#JyNspqP#yUW|t<I`<Y
z>D%^1IAPLHd{<(3E^}o5Yx1D)s0@)zlKAJ#!1UF{5lei|ud@$H-~|>iEH#o=grN0u
zxF&IV&k<eawkO8iM%5j2p8Rt{p^llwJI#LCBw4NY@DIP{rhlchePKT{hkN-)#Oj*K
z&Zp|WN1E}{S&=<CWh-qqssPT%TT+0vw9>c}s1e_bR-;RCMf$by<ayav&G^Ii@6rkN
zFFa4yc`4(uKR5mC+5b`K-R!ynh4v0+fZU_!9gw`tqg9H5{>@1qE<e3}VqEYB)$viC
zm|WLpw>n%1j2IbYc)N>@<;2kz1WFBD%s4_!1H_CBHEQc#_?6M!EXQ%8k}zxe-nC1?
zI=tH2Cho1vg2~a(Kk|-z3STk%)=zFn>ro>VZDAS;PxS&5H-J^M1PkJ2ndce(-uygJ
z-oCq`SW<wxLS$?906|5iR*?|VoA<!WI*R#pX{Aw-3BzhW{7|$?Z4N{du!Ovg9jlzj
zB!R)jK0n~0Q^XSqb204_0F^3!L$Ninm8Q*YQ2@i!J%#pJgSxcpR$Iqx8SKch<7-qn
ztdbVy%MuTAF;BKOj+Y^4uIs1ifA%qLzQ>!}I{18gqXl+HD$*vkr_a0=>)M$ueitTR
zPz?^Jk{_o<>HGhg=?grlAGxYfnBn>iBcmC*K4eMNenh|XF6L&F6Kr4Y-CtLD6hCiX
zpW)@?1SP3AaTiVq3>fx~MP}Zzizp6iq7uaI_?)gBr*>|L3rxSoyjcI{WOON4@*b+G
zki;yOKRKP)>Ez-yvbNTr3ojna35Mkx8rk+9m$iBmtZMP~r+(PD={Ydym*A|`qL5$D
z8oh5YnQNs1XCBc!$g%M5o(l!KyHZnB73L`665_IUO2KNj8W7d9m@l-Y27Dp~bb7wt
zp0;SyLFTmGyTh75<#!*vZ!Bx^hPS1~Knwy`!Uat`%$)y*^;Xm5K0RvyW~YPo{Tki`
zUc+Aixs2cS9#(k`g*0YtBB+VID$0C`2u!7WqI#}_+?#jJJlw9G=Ls=_E%|aivS-eu
zD>ov06I`yJ3Rr#+u^&~vylMTh$&Gm?C~l2)Q~l0t)bz{S2E~%d`gW4Jc4`Tf>kkDR
zl*wBzYV2+Q%(#~G|42hL*9<=bXV8BcfzHgIjSdWaf1K=^s44!YVLSM6Zr<1~0P05%
z21`YvXZz<PR2f)jZjSIdr4@6UN{ZKQT&0w5zXJ$Z4wLToOhkPI<M2T95<!;ar<A_a
zrFYDRZ=dKp_SCO754GmO+f9_EGrziDjg(rqH0k{k=f@%yVs>0{V;0QO^t0=C_X?57
za5odu9#54X{|i-zK#ODR-5WQV4aPU<nbhxofO{mJHhC`8<zQhDRIv>w&Q^j@Z`|zi
z%3ONCUuwG%W$9&4Mh{Kh{uS=LS)apSYzZoAx?uC`(b50Tm-ljmR|GJds+Ox&S{rws
zWO-VnlRCaqXvzpEl~Wjlp9N%?z=(51<=AyR@xw=!;IMy)QMyb2OxmOdg8cWN2kq@o
zhp1nTe?i=J{w9Wpf1}fukFSK(K+h(At|0$ok=II-6A0#|@BV-A$)8nepH@%()VA0%
zZfPzaP(MAkPzx-G0~TWh8>!qosBlU9tAfU?v6N4$yHCgUwRh9>-5XQAk#bz4eZ49}
z#7JKGW{&f?gY{>hNB0X`7c_N#d1%<VJR8%xfhn>;l!J4o<!(l-*^ZByZoG1dhsB*r
zAV(Q(c%d=G<>4z{8s^3uFO}7wB9Fm3Jufni74G?#x@Wpp3>CEm)YIl`a&?U@1K>T-
zKHbio^`uF@*FwC{h#9Mt$!i$Y?5!-`u!;Nm<iQ#+<<nonz5O2Ru@j)KNX&U1pH!NG
ziU6b{L#sKOxSB$&DZfVpNWtS4&d@{(0XzYS5T3vXJrfJ2>Z>)PZiH_3-9+doR=XRc
zEWs?upVJLz3MCSE+P{m}E0Ck!G({}6;)&nU{S^hd{$idDkz$^-b|UXIW_OqUPZPu5
z1>^05hwq!Y{@N~tS!xFkJ&p00zPgZccjJccF@Wg(J&HJOj3+XIs}y9puAz`4uV9TR
z)elJbmx|5^MB&S96mmcawn&{<#B6{l%y(ADY<@&tnB5OB)kHlSdPSR%RpNncJa`u@
zT{@}K=(@dYlSz>ice{Ae#$t%B0AV%^b!JF%-#-a`aMcU?3-<QgQJp~)D#F>)p3Ea9
zaP`TL#mC$MEI7~F{8;c(oEnOTTY4J-l=Z+j$n^rE>!6q>yYZ)tQnRls`7LD2G7fO(
z7f+Q3K6iAzDgxLw?M_vWM^KBJ);n_Rq^O|<hnCeNO>)=gAtKTT1Q_{u@|33MOo=N=
z!OoX~(5{QTHSs8k1Vv6uey$7;K*|}riDuq^kO&t1ur^cO^K)ZH<tmg7vdMO`US)EU
zdhyJokK+FG<j;k_^oM{!fzN6T$HmmXpJ;5<H2QYW0)FrVh|Xq^u%BA|+Zn_EehZv3
zR#L)IF&_gDm6MWn%Kj}pF;EQ4_TQ2i0K|@k8)|o`ez;5LhNd5(DF&D~FI2(0=!j3!
zc+cGmyH;gFRyoT_VuL!k>DPkcKsnt+P8DxGy+U6s25XkK)EqLHwq1sQG19#60zQN=
z==4GQ2eXr^rpuzwn^R)JS>9V@f3RKr(ws}2n6J6<f_`r<7O(UpQ-blKnbPQZp#ymZ
z0b1WAceINi(D`cJczNEUb6;0lYUoxLZCwVQ#I#b8$>kE0!YcqaUra0y%<`ckB~BFs
zhLcQnPuGU0OaflyF{!>EkV&p573S*NKM8?Dm;*wwLS!h2B-I3s@mDBR_@NITfTi~Y
zr)@q_g5z1?SvmYvv~%hh3wew8ovPz~##-zIrQ|P3_9vz?$A?3BZj~&DX9jRjY0Uo~
z4k>e9D5wpi{t+{sE|`!Sm96oqV#{7;+Z!NidH59zc1%&OQCZH*_=%@Jo-?xRPYWp)
z{{IedH^sV~hPRvDN7p!{O)VTwHuC=}^{>dzW(E@qMOIR|^FEI|Jd(<fm|1^G=XJ=D
z)vop-F`1wG;+g>^m(7BU^N?qz+CuQqP^zG1c=>I{hmSVx!)#eQcSdepDAgo=yIAg7
zrJ)4Vyx*{!Zu7j{PD@Ukm+NclcSduRaS3!#M(S5Bo#s4FKY9)cIspx{3ffd~N;S;#
zU294c$lhVbC058%`o!e>N59O@>pQe}*p)oM!`}D-)@&R_&6Aq0g2RV>nb+qOE`-8Y
zJoKM^Ls~GCqz65~%S+YzTEcw@kJr;Bq7qppak(5=Nd1oT*+oin$Z0AnISMoj8eN?q
zsyI`_=n?b~!&~L*z-UEequMovCo8>91V3>*WvV?rY2?3#MD7zCozak9(e5y5chvcP
zDVr{w9zx6I4BQUV8r^zA$5b8Ev%i~#y}loNkE8?XEKY7p2SPRj^!ep&0AMV|wX?Y8
z8Mh0;A7Jk9tac6Uw$1;6E1V}@1&OzlmD&CknI2^mq1nur8NMd7D8ofJS1G2O@unkt
zH>e^u98RmG(NRtT*{;HUJD?omAfgace)QMe{K#B7`bMUSS%T)SE*F3AnD6qIbmbNC
zUeI8*)@TdC(rjM0Po-u=!rkDNDY?!nbIgpmvUJ%O91m;(WQ(SujB#gwenf-`%<{BX
z7v^g{6s^5Y_;1@(Vy2|q%8jZ_D?O@!gZ;=XrWcc~Kf`jUc$Mev`v8?Mnq`rO!kE@_
z+}#0k`s6_F=K&K;!9N~@?Y-KVi^*so^4NUWM%0a+8<ufQZ0apS1iQ>6Q|SQID4l)1
zmU#YLTfhR!V0pbo<azRJ6N7&=Hlm|HN{}M*o%=#?*I=%mc6{Wt%~j|LSW7Yma44O5
zH|~;u&M9rqC;;6&=dPu2OjZ99BvSyEe%=Xp#&!U_8|Xx|?J?AI|D#wfp+jtd@QnW!
zW2>&co*FRdmY40UAOVZ&1wBDPBBEWfdjY?Vd?YvYC1|DbDX2FkVk088m)$cev!C@8
z3uHZ(rLg~NSGvw7?gr}9(R_hLw%7$_4`w4W)S!t&++T_fCc@GEl=q}3O*5LZKYVFD
zOo@KK^di!l=fL&IRjq|uy*-6Mi{P0L;C7O|tQZ+G*)NP6>0Mr%k^#Dv`K%RmKzB00
zUyE^S;=ugB{zaR{e$!&<^U&;FmDjz!$;H`byNSE!AeAlY6?bF0fL^Tmeqz-MHavgg
zd}+=`ECa?)npkFxlPj_p-`^2SH4`xt(ZOgT9^ZQoIHIfA!A#>~2l=UR<pcdWCyumR
z<wQ1O($0xT-0Vq_%GzF#&4eJF(ZP$5e`|I@iF#Eo@G<`HTLae#N+pooeX_aT-YZ_Z
zcY8u*>%*qp{(I2|WDRMr<<nla&V6&eu&5Bv!L9X)q|f`$0S7<)-vaAJ4L&q<PpiG|
zo8R4c!}l#(sf|wRzIo{jpSq{$Ew^KBJXCC>u@EZ~hY!?B3V}tl&my2yJ_WO12%lv{
zKnav!YDj#K%T2Q#+`j;T7W+iANM2o6$RI^lvpy-j&%J~M6|6%Rf7J~p*-fg$+f%_g
z_dYYjIPQt-+HseqZQm2E_dt>BGgflBWT{u(!H0C{MH*<6^*`Rmd1>k1N_=QQ+B7;|
zOn`~_><1z<<AitbfEI1Jl~oMNBLxtgJqrZ(U+ry8=ANr{f`6-0Pb`;*YZ;VJez|M5
z+TedA&0rSR=47Gf^_rarpO|qs{`(j3%dY0ACTu3GKIeQJPHF}AVxw%`G0)pW{lTCw
zEx9J2J~>UqIo6n(%1P@)auQ!!dg0c!{nOhz+U()kmaU0GL;v?u!H4j)T4hTt;ZwBL
zP&uY-31(&+oNJ6*X?HL?B=_+J!aBv-oL^O@v@(iGeksWz1ML50P(OvO|6iutm~b@O
zM_vO-eFKa<Pmd*77qv?Vj<}2Z%I$qwsTn`2!k4QCh_DHYurDPXpuN=)Sm$IM;qBVL
zuL1M^=^uHO6lS6GzME8D$ye*V_R}f-h{^*VwwSiBt@ET`RGeSeG)6qlCL4=89z<a7
zjJZ_aTe~7FkE)(KOeD!$vbDj*Z@*xqRBFB=P%;Nhnl_Fpou;Hs!<E|_!20|?SAgfO
z*;r25vt@-LxEmkzfu}aePi?jDi&Ra^mId%2WlujH@K&*`mt3{#0*6gymtgzlqS46N
ztr?JJ0-S>nlr$urNmsACJp3MP+|~DiA41NNHgAAP1I@F$bw}TIe~HL&4x_NHa~(3w
zTg3LNO0XhOLFvNJ#4FPL82h__c<RG-`ZkEa6<|Snfjc6$kK<x;-$D?nFdyinsW*RY
zdl)Y)x0C*TF8$p-oa*VGze-9LmOFeDbDCqX-Gej30q<a61utx0zhxlcZ2$}wUvon@
z5SZ40rtUZp>=};SNUFj*jtIJvx3`HP_(a0Iy<-(PvF8C548|`mK2UE@Rr_#r_KH-*
zcxkV_qWL?iY$lXMM1L;)y!2WtqmysN)DsX;?5=%9bYOB|@AakfidTB;724l&)(wEI
zQ?y1$N_&Nw9?0c91}-A}MBVN}eOb=EkllIVzLe?N30HZuz8h$!>osUIS^=m5Ui}z)
z-D!Ts%6i~-`{+D0fB2QQvd?qriEWg>DOlxSdNtSN9v{NVa-_OFLqU9wSap4=2c5HR
zgN=xg=m0EXDAI?}#u|`Lhz=IIQQ4-9)gm_5ROHf?ANxLTN#SDQqR4AX8AWF!ik{MJ
zydUDCKmAce4WPhTRCw4ma5D8)3qovl<k@PfHO(`1LG@^U3-5SjjdrxLr}xR$7)Rxd
z)Fe2HzWk@G4nFmNWp#)nk(0IbOss?$JO#Y{5R!u0+mZOWi}rRWW1Tut#qiqw3-<im
zQOWhF8|eO>`-Ek^x6^KUGzTbTu!i3e=bSJ1`VyKEVkM*kwQ)0uu8agRyfPxY!?XxW
zB5sMn?s?nLTH$zhU!3B_UO>!uQzD7OaWK)b=JYLO3-Zuc&G83O3kmZa24F&6QmV~R
z-D#j*IMSN9Ii<r4GUu76!$g&2HVU~kf#|V!Y^m9JaA>=abvr=H@d6w^3PfE)8{(0d
z6m9I(bQHj(qC0m+hy*px&QnDSdWA)Uq!Dq@7bh8CuFiLN0HLq69b{;-4l<!40Y4?*
zxTEQ`@x*lIlM3Fm=6{l{^?!ZlOnR&ek1KEL`DQVfW=|v`=HC6db}^l8rZqcePqmTr
zaUkda`|w&pLs#K}c_EwPXOYifU-?z}RjqyPsW6>+GfJ9hu8H``-G+KE%^5)=pfT6d
z&Mm!*YScqLDj!Pn?CZYYp_$Mnf_P*XCct7^#G#}7*f#Fgw80PeHyVOgC84#`j#qWk
zcH;dRP2|CLQ4Z6%$uFxfjJ&T=vvK%XX|yV%ufx`Hn+crzB<o}-O5VMP8b5iI-WHTR
zW}>G-Fm(b@u$RF4?O_*7-O3H6DOW8+p@oG*UatAtTLG?L_ge#c2EIs-E4OZLNQFLq
z_r=YvJGZ;vSIL*$YL0>_R6!7Q?wSzI%%HWXln~8x!j~#9*F2qjBJD<LW!wEkTI2;3
z=<XmmPOr_AhzWc*PU@3A@1VP@`e*8$E0r;S?+&7*z&_{rgW}QAhTc^ocmijjiVyM3
z<>r^!tg}o=g-Je`8P-3Kc>t0Nv_M}pYS3Gps*Ar1p#LzY_)pqt9++=nc@F_vOPYuk
zRt!}nRPv=w?lRun!ao2W%X#ro0%x^CKXSa$7a>bc`b65{Kg91!Pug0z-O?305s;#d
ztHX^l+G>tAU*T!&o_*ZTFjCPoryk#rQ>VQYzU(f$lS|5t>`NUgHE<lS68r1%n2muh
zz%~BR<8%2YcNK7bT(rb;?Iv69`<MH@1;IR*0-f$1NC3ta1KY-JO^{R9((0F-_R=Zk
z+4tgnITZ4y@F+rC1}-B<HzgYiABxdYCt9kQel74Q*4T?Phx~+<vnJAdp+hj3@*QP}
zM2JK%R|pqy{|Q&K^U?8xO%4nk`Kh<cPmxz~=b1Pf=F_>L_!IcjiQhSnvP6q`4bF06
zX?}40J)|bt^3qrL&C5<ma!15?_tXgI_HH?3f{AHw&2(1`DglhFYb98-P79we{|o7K
z8yfAdt)|dVXfF&^d;<`~cA;3?)LQ2Ga}4P}X*kCxhALL4kL!=%TLlc3zk;tX1hJ@w
zG|A_NlbDlrjPBVu!`dWjw4PORE%|G3+YBIwZsO$~nUVA_wG@=pTJTNGW8p%J^R*&0
zzx3)NDCFkH=R4sDRNw2lnuPMIhWzZ;pJ{SeoQgkB&?j$?V47-rHPG6*mb&1G=dxYy
zSw1#z{)P;NBSQLG<2KkBw`G<6bQE@7bq?zgLwIF=;<-3aHQLu(jT4`HzS(tr`Ep*m
z<ER~<)S1+UrAg^3v<u2345HlkG;Gv;ltz!zC4vb*Tx1?*m%^{PX7wN**)Dkbq}7ML
zgw3aT`epK2Y0A?*p7Mjvs^7J!bk^yQ;98XsSr>KO+<6P3K%z~c&+6QwhXj((2zPDe
zL5>zS83ulF#~$KY=pR!V6ks{9cQ8%4T*Sc<(%Y2VZ?dpk-kZOPy=`mQw{R2Tq~rlJ
z>?<>uNsIoalz-D<H7>Nuc?&}G?rk2`t1P=G#<Bs>Bb$m1wr;x@*T0z~=D$GXAGBH3
z?v39zcgpp<y15#WSd^Wb2MB2WaU&cmSCCXHe%be~Or|EeWm074vc2fnNN7R?m*tWp
zWo}F4j_2*)FU!2G1d|(RWa*`h^Vq)=Cc!oTm)r==`#v-fduM5M%a^{^vW%ajFeF7y
z-k4T4ZUdVela&hai`f0t+kRV@m2j>g&-sapZf;V9DbCFTKou<3E4buGS|Q2l6G;gM
zux2Ul+`M&FLj{emQ+t%}kOa(>oQn_L*r_>zI^FI~VH%<_-UZ(Z-wH!$SH6fr0}1Il
za(XV1fLq9>QW*<u%0xA+%#yd(iEMDy3zs_Ztd2rP@M*!7eAwH5$WMH>G~KiMGr}-S
zWap>ODZG%<PTfnrNeepT69Yi*JGo>5{Fvasc5(74ayubN`zo?qU1|SUgY3Jo(E&z6
z`)&|b{2YYN;9NSy!t(t|lHiDxu<oOr?Pn;RD|biJw9#{^H_X&}4Wz=^C-%m_>XOv+
z%=a=@{G5UP)jZ;1pu0{oRAKkPiBsJO*;6=^{Zr9Gr)VWUU(Qfy{dSA^4JE%@x<ri6
z-Xr(nmc`B>L#8Zd5b+{E+fr3wBQ6wffNebwBHXoIaLKoIcE7DjtyFOF0EK{jCDgl-
z5Mw6LkC|J#^bzndY4WJu2pnoS3$s^Aq#3dcWk0*ImrUoYbj8o4{#-&r46TxKR}+fy
zEv~n9tcG`xcW@yuQmlx$+7g3_g^HCc+Qn@`ky$z#2^0>hu3iR6<k`uc>7V^gceuKK
z%3vXsFDSAB+zKbW*$ED)_IHR*vlo0Ny&S>Bnl~_IE=N@VOTqN-N+^waEL%e4C_D=2
z;}ZHewVo1S{!%kw-8A?4r2_Ot6zV8y=2zlqvxZNEn6HYvSQqRQ2JA{|e4qV{z8fnp
zjgzME3i;vSBZoJ<M))^hx%+E<O;ORA+t=vKrAKVndwh9An@gtzV@lqgq<IYpF*C)V
zUn>jsYX2zB$PwUa#LN@~s_L$j*4>M%9WVtoemA=IJ=mS87i@uz$~aEC|Dr8WZBH$5
zYCBG5!Px@M%Cy_yG>BhHdFiVV)wz(4kbIrnRFT`r_ofVzh51?nN+1dIoN<z#{`|R)
znDfHGFbRf4KSM!9D#+XbX~1u0WA-ATPVi%eH2Fv`3?`qh&0_vpfJ#OFxtaN&e?VOq
z;!i7pYu$}MhjkSxhMw4C<t#J}QVL%l$#XC5%R&ZHqfoO=T)o@o|GQyzD+&ojlj^#G
z`Dt;5h-?kiw0rtjz{P>9nFSUpvWnQLcg)G8bU4-LXGgtH?gpb85a3k#1(EB7*PK#K
zae>zLsu4ZPwcbLaQkq))Yu+{W@6Ec|{el}^j}<jw&8vm0g64MBf4zyccYM@^?-E_&
zu=OfJaTF1zSvpKsx|^P#ty$R&F)fyq@>?Wtb3KnZJdJW0B(N9uAMuktt^FnZg#4oC
zP}J9r0M>qBNddd|V5Bb47&okCLv$UVW%h`JT3nhQW7n07W(98V3j9U_0BZrb7F${@
zssOQCEjrT!<=ZCgaAa>gXZ96zT}ct3C-~Gr1snt5!-s8df=`fOrc=-zlJn0u(1@B-
zmHNFw-nzG+ARNRQz43<F9d^4-rsG6oMrCeC0p@=-P0<TEr*JtaYK1FWjZbwtdSTy1
zJUi3l-&z-KwA~C!8FTY~d%zW7gSZPAE`LXEm*y#*%Ib_RBF4pHEgtT)n{;lf#ddB1
zx;}Caa9qr!>AxnkSus3hXt}g<5;}NP_mv(|w?*kL$xW7qY%`4R4gWAvHmX6R-CKWS
zVXypJi^gM@AtA0g<6yf~Ir9;8`q%_*-QQpPwK!jY5F-nSk=&6cIX+AyTVeqRSb*FB
zeklI^dgUkUB&hI{af>=NPa!&A{%g%1)bmMkb-opCIzX<M@dn#bCsRU->MdH9-Z0h&
z>3^!Q=DW#OUh*4)?K%!$YU9JoOs-t_QOXdPR%NAnj}O`~KSnQ3Z*gbf)5EKN@lG+x
zi2+PWuL~?0An1t%t_#Fjs7b<MH007kamJmRpO_$~Nq!L7Bw|P;aWx{jm*pSGJmm?|
zKXbXM=j=jkV3FLe8j_LzwK}SyqeV3d6|Qcm9gdL?Zsh`C-3*~E-{ZF*6?sp)gf-yD
zI3-Nqx43^rR&u7cua(@Jf5!rNx_y-PmxEF(xxFo`rg-@$6S61rB#ysbIIoWgNP=_m
zyyd;OkGvA~@pwcjHdJ)`=afTSn}AuasOvlK+G@;*jNfwQy=R@r&)JEXmdQT_{?M<d
zFjF8dxFdgOq^u%A*?sbs#UP_ZiPMlwc<`s!F1~_MQ^nn;$Y#AKp2)FqX69b$H|2GJ
z(WcXI`^FrvBW$Atc!SSv2$G?`+`r8)K7zai6XI!1I^DZDv4Ig#9qvud!Vd>2VAM45
zXPpz-R;cG-R`P~Wi*$FRd3an|fFa;Y6B!(v)B(Ww9bbblwidN81an1&pVv(63jsJ!
zQL9J9w_qUqDaUrJ>t*aXYvT!+!C7*0pIsQQpL60e{oNo$t!zi;odg8qTv%s<d{$wn
zjtsmG%cpZv$T?TYQG;}#?O=VoT0EO_1z6o!$2Corq@<9yFMPYq*01V$1=c*^_W|je
z_p9M%CBR9R{P3$h+&WdTAhP{=rxw}*gg`VHjZg4w$Q5}QNWdN>M^3X5`7cl?N_Tnp
zE3(<!6J(5D&K23?ywCVzX39*vGCADI*Q>>lq<Md5X<zrKq7=${Sks(u8vW6})*~}_
zA}MKHxhBZ`1^0r=(GE6;_b|qGB@@$fWm?CC<Fc%=dC<|Atw|k}=YfJ?oqD~GOsls2
z(jW-*t8lDv$j)dr(G|jcpO^Qk$p5(eOLH)3kPvwJBJHu)=z*P8g0WTPvgsqqah;lt
zqJ%Alp5sT^Nh#8a#OL}i<A5mrB<wE&FR-NnmPu|P-339e!o0$GdASsD`)*WJ@p&-z
zxmC%=d<nls66h$7X&lvb-bXMKVm7nji^n5D$%$vzG)<O>qThesQaasJzDaz!npgrY
z8L-WSXTFeNm5fuT9(n28!Ff9N5rqBaT5x{XyT6=z?svucAEnYRqfQix4iWE8>=0l2
zP{$&9MXRCyT6^O53kE59i-hy4d+yZYa^FQ>@EpGB8N2D|nsUcX;(kTRP}{4Yc49P7
zM|E>t<?~FT7ZJ7_Yn-<1+=^+~4IXNG(6z-_1PNMaJS6>MmrDlJ<ezklRUgn|7x_#B
zD8a>nY)c*CCp1ipYP3!A1Zjp`Q{}@x#ACH0U>fn)mF~`H2}^#Sw|3g5$Ab2=nq;U+
zcS@jEFCD4v1sy{IKNtFSF#r5VQq+#$wTp4cae|q!zT{ZUqOa+M40Znu$-2XLy6*rg
z<Maf~7SweuI$N@}Z17GWe>sWi<tDF)A^atby)cI`-sxF%Gcg%&l~&1@dS<izc>ILx
zY{VCdJB~;FH^$%x5lJaZk2U@oYT8=k?lG$`(5ZYZ;W^<_&V*yr6%st!HJ25_IWDHm
zxqJWL=gxPKxoHcn7<5dP@2wiA56J6Rp-^5B(y@68>)z|X+y?4<#XSNH8(7C&iv|8k
zz56d_@luO;pkwZ0Fc^JU-)l@w`(G_5EJ{s`<2v##To*=Gv_f4ow9Lqc4K7|l4y=t8
z1Fe?oZWTz7N*OcbRxa6zPF!rDh&Ovk2z@TsYWS?Nd;#}0Hl`F=Z<ixQhT6&8UvYO`
zYKwV_fxD&<OsZWDqzCM~q?OY!94CK?s^t8MCv3GYIc5F!5`5|{7Q$sLY+b7FKsWrp
zA~85DBwnrq3f*970_!g+0IdaGJf;JrM$Rb-Sv9bO*qWO*r>pq)AZn;GZi0d;E985p
zyDSY8{K)#7Aad{M_rt%NHGQVgBylhK*VC6|_aAe2gwhz$!nKCRGJ@JP2i}#q6({(i
z=e$4=8dY=o5y=J(C4db7pwF6S#qmM$^OWZ0YJQX?Cl4_BHfN;WTdV`P8`Hxa4Ep^!
zZ+?@!Z|}xcS+l&B<lCm@-+s+Qm7$O&pBPV+m4qv)2v`}|u0{7QZE1DUy%5Lv1m?#D
z>0K6K_YMb-(}P!5kDKr?H{hHL|IMaGT<Xlq&9SysV4D5q$*7{7=`O@#n#An91MIMB
z*&dbo6OQFjp1M0)Ov*xCN1%}6ejD%b-r7Z>o;IMcnn|bk&Z2W`!NLF-Ek}F_c$IIq
zz62&kcSq_x2j%NQef=b-b!6k_F)Zw)@EozIoDK+6ow>dV>3SF9w}(cv1bbfP5r&kL
z2+bxZTS5dV{P9k+f{u?+!4kn0=d(g4DA1R>hl5!@<iJ*yfBl^H)zJB;JK)dB;?4n7
z(?L3fv%PL|SJZkWx%f+-q1tce6djGDqik)jZhIr2M}wm42L$sK`{{X;^uotIH-&_3
z|Dyk2@w7r!HnS>-fKk(<Qb!F??e$acty}GdD5%4@`GRzSc=Uokf*4#`#%j9#fzhwj
zNOlL9)VmF#G^!Dod7E+mCM6c+_eH)fL>5K<G<?79c>thG6=JX7kAz}t9xnuTV#Nhk
zIonfr6<%}A>$ArhiMiKHOT{5q&{zj3Y16B`nAe|xrE>iN9l}75{w4_8s&$%}@c_y?
zp2?;9)#6u73vYk;!&~<zwB$L+P>+wxX7qFcb6J5#S{hF6rqe~$O{F038>X&dao)!r
zq?4SB;0gxD-|IpTjT4P&Kt<0~^TJw6dYzTIQGAuyCQ=?p-QV}2W6y`Z8M<_=cv5vz
zApGs@b*jMfr*3hHW#?CXI`^eTjX5Fp<x$q=Q6q9?`chR8NiXh0IOc(t=uZ{*qFU^2
ztFt+0|5+z(VY`PE6Y$jn1LTfEF*#bWx*FY|lTG`8>_2j6*fi(k?dr6=L4jX>AVHnl
z&DW?OGB{N@VY)N^JYmHo(D2D4HQSuQZTnBsRfv+VXRjKiUvpih|J9R;57-!)ESKLH
z2wzN4efK8uovGgrhzsacOf(o1*FbtF8Bh+N4<R3wz9(Gqd<STtD|_aO@;HC&8xq<t
ze7(k^2Sg^L68$NmaKk$HDqG>5dXr-p^8We`Dl3hCp>LR)T>k!rP*Dd6-og<X6_L!$
z^^<YOUGbxn(Zf24dryAK(3!FMM@6-s*zFLCm;QB!(WP();i7aGkxAQgp-1n6dWx}r
zN@lUbQXRUeEAXINI24I~^mDLFz0u`*bJqf188svwPbB@V>Qg^0zgEt#A3p_C2j5&p
zo{ShV%l!%&)hT)g#}aiAftQO4<EKd=p~YThWm9UmQZu&bw7p3_C!p4^%0V^nn@&Tl
zbFg%t_^~2yB<t)J8`oK_?-d2Hf7PSBCx`7v$ZpYlmjhnGX;auflYdY*cHt5TLYmN~
z7`%b!zRcE7G3c|$T!`J{R?`lo-Ui$HV`d{Oy8@=t#Jm~5gOuW_EV70w0h2uQcYROH
zWqytw`nN9Zg2?FYd3nWG{fm^KVbv!SO)dhsiH(An{z0(?wFCNyRzmd*DlwCzK2;h-
zOncNFFyH#pq##@w_mdD?^kY_g#l$@C)}uOwaE6ypmo<1?G<zQG&M<lun3Lz|t}bg9
z7fsM0-g06hXmU5|Q&zLcQM2}Q@*nu*8}}73;Q*z6Of>R6(^H2G)${9BkVc^7gqAMr
zhu(W^h7xa`=)o1TcV7k@9*PB)oXOa=68RT2uT|SzTU6cK*|JDcyz;QsvB4QqW_h=w
zSv4TdybcFHuZ~Nxy<khAUWq3ZoS5X?eNInT60Yv%0CG49832R&)CaYJoa>%b3(H&g
zy^)waS8J2$5f%C3pMtsvf6T+(vh(yOSm<m?Hm>4QO0skI4UC+1GZ$~;Q%q`kwn^ql
zuu|}BnbqWpoI`ik_LhA=Ie(xGY~Mk?O7Xxx*<TX2|HPxc7KC)qbVU!-SIl7K2Pl)X
z<xF-n&GW=x4RL%xF6two1TVz3NItfbC&9J!+a7T7^<Kcpx~9Lg*07>IS+?HK9_{u#
z=l=X>jLiAbF(((2j+@GG!={W{Gt#%_akt>t1d;=Z;XmCx_m7D@U;P59rxhT;_kL1h
z7PuR18*@2q^E3%C7T9_%<%)y*;S?zZ<17Hd*^}-(Wx^f!Du3d(IcXaR+#F`5OO6b6
zG5kRoT3a$8ESa2<C&ZIb{&vjPuf2rOei*09i9K){a=o@D%guvNVIwLp7=_H0_8|vu
zyjEgwh}XyDWLdD2xL!`_!&w6d4}PE+PczAh<I*KM1WV1Wmvw~A2_82tXQ*pykH@NN
zX#eBhp#RryX5r|E<uNB|r|mP8x@h2s({H?t5OXvZFswfbd*aZneJAi@t`JnwsblDL
z_egkGF)IH^{sZz?h0(}2*He&3l~f5D?X)E`u(pjt+O|?e@uFJu<^BSXbDnY=QN%pn
z>b*4)&WYg>9&kvv4dvy(Jd8q59T^Kdvh#FVK^ulkpd(4HT(A)U<}S6LedE%8Jp0D9
z(@nuLpjAh<e|VfG5rF6OTy+~YO)6KcU+WZH@(uGh>uXf8zGx-3<fBC(X)Gd}rQVYU
z9D6!o8`k-gm<}<Zus~0OLcXYw@)A@UO_+od<F=bYEwrkc$7&B8&~iS2#YC_R*%e=v
zYd+P6CVqz)hokMv@hTzZamH6UxvH`kKr#=W=C;|rrgode<gf{?b;#<RT;$IVwdtzt
zj+6bM1D4jy!e<oedzJs1_TxnrV2umlGB1WT3|)!^7HY97lIsd(L$oQbnJ*_cLvux!
zKka?Lx<FSJNCUg<^UTG24Q3GqU|dFEoUnFRu6=|{g$tS|5b&u|htd62+La<6Sw?1?
z&|2dAdfK^SY9G9BdW4tztI{8IiTp*$T|sy+Uk5^Eu30MGKAeQ`81?0$lDhGbf-b@7
z@|()64eRvv*mu9ac-9B3-na1(yV#ucm*964bopYI^QtEgwqWbh9IsNd-_35$dAh4^
zhK0Nrjc8q^umGQkb0KZ4_|*ysujvs$HTUjdzr{?KRZbI4k>|Z4hymu*mHm-W(QD=A
zz!3E^a-;G+m#@jg++#@B@x#+PUfEIfouJC{R%S;8SQ(;DBYV#SvaU}hbB%aHF`kpE
zfFLJNRor*cVrViGq&dIkWcqls@n^)_GddFG8!ud2|JI<pUmG2GB;=oD>+JXR<a%!7
zK`*`Qk9u>jiJqFna>JvVZu6`j+k?Zop_}$d<c75i5;>ALM;5E~nExUcRFEST0bd@o
z8NF6`8@lpSPHOe{$?R2VnIe^vEX?Ft8Hr~e4HxE*<isStm~f;LW4n1@TI)ReKK;t;
zBD17V7ceq{K5l|f52=#U2Ym&vLLC}XT+CEz>HM++`0$!Hj2>0iR9eR!eKoekzehIt
z>Gga>zH@ao+sLBPT3>Hf&-Iy4Yx;m%_BGhHl|gbmpaNv@K=L;*lm3(=7Z8U5L-qW#
z4ZQS5RbMKH#!NMdDtpcNcox2`Ku0YGNR>wY@^-p{-N6jm&V>Y3zdE@{4iSG$R+kV~
zXt~zaD4akLpDXvd8mFW+^#y=_o`qsFI{?-h1i0b}T&6=Pm0x!$h;4aljc<bwSGQV-
zl8_&4FMboGb6%Q8?x;lZ2t-<(E#II2^=Dp5t*m7rNuVUn#iV1axr8$KiA3Re2A{8R
zVp;-jW|U`_@Gm?=86ZkgMqAaj9i1hyJ{}5fAAPM)6Ot7x7IswIos&$#z`k@ZbGz+7
zozBZiZ(2|BOHlB6e-Ysl!422TGOow_!r^)OayWO2V29m%FP~6$eVa3xA6OqM?Y(!B
zETR4J+T+9uUKt$RRq(=t9WrWIi(BOpv$2{Sut0Qm&s)efI3K@bo%6jSTKJ`XtK897
z>IS<J?(;YO80-E0>m>03pPYHNa@DjY?ci(olzipB;9AL1zLN%_a0P`Ty2e~vTk3tG
ztRL5iC=AlYNS}O`V)a;R=1okqwCsG4kQQsbGDXK)RPqjA+Gi6WKugeeG5VrS+<+&5
zQF(%;%_yRPld^w6<~rdw+$v|Okr<#1$?<i`>AmzTQUw^6SSL=vV_%!wM4sK?KK<7j
z67u1Oy$UZm)E%bYCq~fRl$w$pe8llH6fScTZDFNd3m4)Wps~X^NE&4x-jg%e4jc|H
z+pf9U4>kP79*V`WLvf7&YFvgLgJXB*{r!L|7zWNQ|FX=I`={O`1IM>ox?Uuq*l+k^
zjQ6KR9$Wx}vSdo!l;=oT%c_XVqN38nihWbjP2+;vdk1;P76#wrawsvaAW8ztxZp8l
z<F0wBUq}kQG@IQO&~WIr(l{%o`RqdpEuyOli0g!x11JgXV@b;DX_%Zwunq>9H+dW8
zmqyvMYMlkSOB8tR_x)%QOOXRr>tftI)E6*NGE^?BXYU(7jgLZYk=i6EPCWd8MY`?0
z3>V)}){f`1(09x&*VMX{bf+N3V#!H!zDddSkTgi@ErILh6yG~M@U0$n1j4%+(fx9`
z`0pW;I(ab2cd;U<=Gm|g<tVy8SKpr1odS5$)9r0e8fd9>T1xUuAWc~D;?slGb80L&
z$;41<=Zc>_rJHJWeN(;&=cxhDL<%si`&}H^@I-z-rD6p1XOW>sPRN)b95YsK1TU_s
z`(}dpKmE3Z0&J_)$5#IJPqQL%$UXy^JAqrsZ?+J(R4!qof=9e`je+M)%k`%C2__xW
zouWE@v2OB3>E7bj`%izG?I>nNXCuE?ATUp~N0QRZPS0rV+b*pm#stEOO1kTZoYu{`
zZ(q4zRcekgtsFVQBgAkLNYNpPotPAyroVT25ZJQ~CIX3wV%2ON49b&Cv2cls)lLqo
zzCqeKqt1%Q?4ZstpW8tdAsA9mEJeYKjPWr8*#b&#gfm})FGg|lRmDz(erK$f`T^PS
zj+d5Z>#-xcETjGf5`Qwzc06poXLsiEqfzGk_c6UCe%z;cfPV~X3xdVl=Mn^Hs9VeB
zZth*{mW7O$dn#r!+@m!frW&n3u%Jy@*s>7hgId4g71RXfv`3ID96R{L!WHew?YQD1
zm9>6%uUdcO6Tx^RJ;L?aY3V*3x3+99I@k*sVoW>PZ<_P7$2jRyH1CaCr&gZVQ(F(>
z>?Rog#Z^!?o%o<ng1TqI02gy-D&kw#!1&ImG#X19mC|M|Jo@?-J9Mg>*?>9aBd+V-
zwm&93Et9Hw<QsD<u4uAq|El#h33UoIDLJ4j@VJK1S$T}E)pnZFpIBQTl32Fo;t|WT
z;vwjjXvJ5`xB=6dEP0+Ux4Y+Y&4=BKUP6ct;Y8Z<-F5gmbH)Qsfvz;t&qY1Leo-U&
zP&vxn;K(qjF94v=K)P~gE|DxDb0;9i@2byL6Kp-~=<E`vvA2&|LS6!vGN(kQg_@oB
z@*y0^IZhH3F+>$K)<NDZcs`-3A_)m+D>{-d%6|t3(6FVazluG<CL9M3NvLGJl)IVt
z@(42H<XpjtIodJ$`3!|yq)YL*=gVlS;og69sb6aOWI=x_lIqZ~tm#t=^c84lu#0K8
z;Q%8bdC}5isxj@qO1#@}xRY7Z??d_^y&$_7IsjY)IZ_kXSh{xD1&Muql|d4ot>Y%R
ztnQYva&j(|cGnZE-qKIg;}$J63A;OuFF5KeVkAT%<%n5vL85NPnK`nmr)no*_lqzQ
zx;Y1TdHb8q#$pC2eWkK5%!Rs{50?^}hnWlT(0d+`5BWn0FmK))10Y5@^Pq{Upjbjw
zUqGunkadzP-qRIwlq|+N>MRT+MijYtsBf&jmmBi?d_%aR_gYan(+!c}XBOJrHh|;4
zep&Icx|^ySJ}|)%cF7cTg!ynbLAFRX&x6Sq#5kiD^pXq&_5{uQBBZ5ARZ&8iDojTF
zYI!*Oub_j*_Cl%Qvux)d6n|Prg=>1o{q`7Hf-igI8hZP8G?TC@&Vu(~&rr#)V44M3
zP?ah+eAw@Pm8Y$H+ysXQX~MD18cE4jt<ej(H+Jc0^q1Pk#lWqov{mG1C{)%Ei4>pk
z4237AY_xRAo2%X3jkwHwU)?4C)5G-GsMh_K*6OUCPr`P7bte54=;^k0%B18FeY1Yo
zddH!aS0j=Q>q^S-%<*Ey^R@EgU*r2cDf7T48k`{(1J|dxY#g_JfqevMwCP8dVYX?S
zgqT(PMsBedF(u!ojR)kYAzN4Z5^Fy3hyXsN41u5PgP?_EvE<<y5%nD5@NaFA`brlI
z#E21L&;)U2rFA^Oal!xoWKR%Gt|Xqt0kQ!nkDpMi4?#6#?;#1;$QkzA+IN3F@1&D=
z@;ZYf7{9BN!KnXr1rOBVAL>YDvp??VsEuIDqrN<8f@jjjY>N38Qnq(hJhdDExJZz{
zq<fw>xRlFb6m<hL*wRXMxCO4g({#8Va6T{II@Jp22HKb6UWBgk#|%F(+N`xmqc|mO
z4lVdA*Wh@dQ07Qp)HV<AKtWM9En*|NJe$-0jkZ><$hY*2uxrJ5)Sw3%5~*ogwe0Km
ztaaO}r3nT(?Yujeer#QLWIdlySR}V-$wfb6YhJ(mBsQh7)>DFDp=rt>R5Vy9r75So
zuks{4|KsuwjU?ezzmZEUm<&#225G$t024?iBT|$)qmZL`-uN~l<~|%Qp=2>VNKGCe
zJ1`>39Zd(J;8drkA=)6_04jz2*_n5MSJWYIdrD!^(y_odR=ws+X#=9)(WofZULGFe
zE56p<WZjwQOC6b3LOhWK_5*WlC$p|glpbe=7EQE+jTXVbt7@N%QJWmm)FhpoRK?O-
z-RSD?UNyxhZtY4mh{k(|p_^>m)b7r-`K^S={Q0LZj}#t%PZ0L7_?=Uwv*O$Od?e#^
z8atA8m;$D;1Z;+{+a3}MZ}_@=ugC2@8Cbp5&gaSpwUk_5wxo-gXHZPJ(M52n*uD$V
zA)u*|``J30liO5LXsAxBl=50jG4q7k(~vETl|X*`=Bfvn3YY$^jS$g{=GF$2XqVlA
zq%dNKfr?f;QFlgACPa4Lt7y58s~!{egeE@8tGJf((f+QNdQVT&i+q3O3+t6QsP9*o
zF<0IJ1xwxg_G`OK-D`|G=Mh=kh7YEsfWK|sSD<u{FSHhi8Qb2EZl)Po>GPCixQNh+
z>dOgcX=|C{;Dq{h4!V5+ZYfGxop$`{;_#qI8Z-Qy^cJET5i5J3eB{d&CJ|z(0W)zc
z!U2iTL*Ki%S)Sns|6)6N=lah!JBq*XPYu2C^z>WSf@tY-^DQ`h;AMevaRn*mHJIfE
zRa7cMd_-obBHorC3p9{|*vlJBzOtvYS*0Pm#_3)oj!i~5w(lQRbZfSMD-3MmVn%#j
z@{F_t1J|}pgt*vy=O!XAU|^IUu1&)Fh^~DWmifo#W3_;xTym+uBGX>$91(>7>(fZm
zp71@>;)`$3yGyt9FB^8cY5MdhKXU>R-senr-bnPLvbw|<%e8@hGS5!KT*#^5UR|nL
z2wL2fw|*D@)mEHI*v}b9y8U{?PLTt<dAZ5``A|`v5_zc+ja+~2H3VimR5a$9FdZU<
zJ&AtcZE}PT32HR`FI|P#x!YT|(~7>+oWf%M`OgU$43b-k<8MHDHsqQL-8HS|RDNtl
zbY#A#WGJ#F2SJAH#he~%ohxy^LgqB-Mz@^HFCfBxYOS6}qw)VycuWv*@&2CD`YX}^
zp?OqdaJCw0A29-nTq)-|56kgJ+Cp|yzal7mjfG*P!w*L*8Zovp**fJ1wc}MNr$sMU
z-ao}MFrtl>`2<#JMyPAE3i`#Y(<d|`N5W^5Rp}zO6Qe<cQeqAl6$!W2NaNz=$S$Nu
zKrcBW+Ez2F)UW(%2%mmqX%WMj&Gc|!v{m1c_4ciP$Kg(}VBcZ@_hkOGPl%5q{Yw?C
z3jX0~)<6xBosaq{D+(V^P)3$a^oWdO=nJDf@YDndo2VF$SHU8#u|~dDkZr;56>gO4
z4t@P+J20}-Y_1&1T#FMa$~olWVJie7s$j&%0-gyEPZdp4qLg?!11B16C~5eg7NNQ1
zMu^Jez6Vc8CjhmgsxxGj;drf9RZV+}<5+e+;bcZCVa@#Pfct3^5Lf=b-rCrbiYDHq
zc4J_jLDkdjMSReEzTtr39?ipJdX#WjtqoUrpLzq+avM(z;<0FuGW|3`7_RUa*eCGE
z2kzgHI)xA6e)4O!N?mleym#7ir|rCXyjYLhe=@MPMu=JX<QMp@BG-CPMULb>RRk#w
zX6`fh#~b!}F-R%{2JTqOV6l$xeedj}-EFC%Xgk_sA0YRXa=GXY@K>C~4j)@L<+9Tl
zxVaMo<Zm6`N%9b4TCi2?Fay9udy)vh2&GS<bPlZwtQCv7<V}xqyzc1s3wdGPO`s8>
z*z3Eucj|m8cWVeRHh!R{;HyZFiyYNbbs_2fRHYI&Z&DXPL7UF+FF94xKYV+`!RkVC
zg2?KUZ@#sv4f!0^>@+KX5<f#KXR4C!JZ%TaOzF6~BW6WPaAj1k@zyGomFcm*D2@OJ
zq0-Y;xc(14sB#OL<R10kJl7wC*qEUG<O>IBH)cs?e4AwRdL)V6A`?M9JMukHWJXow
zh@>L{a-{qwUcqypJao~Y&2dy9i{n$y@Rp;$f5vuaG{_AOG#OY)2pr{X1nLNbEs2GI
zu<luz)t`ppRVIhc{64+|9$lQ!gw)piBha`v7_rA7GPLdv`ZA@MYr0<-%iI^H_13N&
z@1~NK%1;jIgVwjm^9&1m#D7?{u@KNkGW}Y)Z{y}wl^R?3^oJ-7>{rUm)sc!qfv<>X
zF}`)=l3&{`((ZoU`fjn^Cq&o$<qcR2OP%lDSl<76{Lri7jZt4M?_aRDcYYH8e|)`n
zJk|aGKYnCqQ})a_RAy#Y$d)6@&Wr|`$sQr2GS4yNkVwiddq+6OD&pAl91@X(IOg}+
zb#-0u`u*PT&%bW|dY#wn`FuR?Q+~Se_3(hC0oj_Y$HQFectPPVsVf=2!ooE_f-k7m
z)nk&I$<?<_NN`Dq2p8cTZSNAt0cJsw>vbzGx$AQ6ud)MAOcQfQH9=_we9IH>8+cr4
zk7^x+gH{e!W)kMhrw|V{!QBWkzoImPw1z3kF4lQ>OjqnfaBV~D3@VrXdH3nAA8qg)
zEE4?=aqzjt|MRl*|Fy83m=!<5!6@!ic5yeI&wA3Md(99G91vMDS35pkHW6I~9muj$
z3v<GQXqyl)>IvvTJ@=f0T$46q|HVZ`L7C(0#_Ipy0p0VJ9pac6VUp>kr%9`e0#4BL
z4I@yILK2*{k(kr0{d-UU3LSc_ob!|091IAnDn<XQuV$0y8K3kJ@O-^@u_7Oc$Li`z
z=P9KzLnRl=Qg5s*-bJG<mL4nFr(4vR+-Zi)2Jf|T;x<uL*Vm(;Qxt#A0=Zc;)&{bp
zaA0ZKuFZAb9u1ffKZS4+Nyqh^9j>qFs)vcXE>ZT-cpd-~{Sj!~5cXYo?@p_xSG~r3
zXZh;T&@9m|e}n|JgSH2V{(KCYH)vcK&m||21zl(l@8EBJL<D{SDiO3kaVoFbgAV7D
zu3Qz~+=`K*6Mo6s;mgoQPSjdqmtW#GEbX@%U=Wtb|4Z~46n)u(!0P0b(5tF0eh&7~
zMyvY1+oQ;=!S2D(t+LF5eR7gHe{NgYLi){j2J?ex^Y(M!e<2*e_IR;A8;jk!SeEVQ
zL;N$pn2;rIo{=5of7%Avxc9Q;P9tV)4A=VMxxI-mmFy2BsW*bgxdQ$D=U?Pzb>8=}
z-azH2W?2gf)`T=&50{dX+`ajD!*5E;&+qc7QmS}Co`rY0O4hd=3m(C<zI__gbPQOo
zYZ+v`(O~lOmiZL->9A;rd(%y+*_#E~rxBYMH}cjS2N9oLZjn0OWOTQwF*NEwy`mPE
zTKpsc*@-+CO8|Mso61r-fwk*PC($CL{<R}_RkwGZk`iaksVr2PCnLk4AQn%j{ozK$
zQ-K<YEXfE&j>MEVl{Zc6fHO*ab64v-1D}k4Y~riN6Xj9~r(>%4A220|b`y{3q6E^9
zO6C@Rymv_sJ&10(9*q0*emv=T=vM#N?6+G=_|S5o%lHTid6~+XLgP#E^964{0i-}b
zj~Z9i$Cwbvs>|x8U^O|!*Fyijm!7#{ESiZ#ZSO<#u+_=BveE*uA<lnEgWZ3TH3A?$
z+nOJmL&4Uc_Et}p4sA}ByR2_+BPZfQu1{YxY<B$&^byNGAoK7sDk_667Uj(|&PC(B
z?y*GUJMEFvzg!<OBxXPM2qq{aJKKAjiB5BVb!9n<G|RhB8S;%}VQq9qc<*zb2V0Q!
z@Rz$5Qw3vX&t4>Ga`!6ICoRxmCiQIsg@uU=8Yf~0rkYlj)j6nHg-)5hf0Ge+5>5O}
zjwfd?haRyXz>jwey7%M0nnKw{L{|F~7b<kg_EI?B2vNEel98))V41ovdiNbtmvcdM
z7lkAdJ=q3P{z4SMqgveUF$o9Z9mI8{wOYHeqURXTLG=B^=#7&@<3SC1-%mp%DiKKy
zD1&ER>~k83z<``kKgJBhDFP<+fV`zhuI-Oqk>(e;$EQUo05l^>lt{Ra60rx*1PdY$
zBs+-ep{EY5uQd;h-%wJ?Wl6mytcg>)xuN^NPkSc?|C`}-4QaOqpb8dV_|5XKlk*c=
zw9F=O4Jj`%kr_z})TV)<;d^}6^S&<*Xeo)eb37Zm=~dJH-_|WM^n(mfaqp*C>NR@W
z7HC>ey9h8_f%yKDU2zv9jtIy%PSkg02A^+AFM`kk=HiFX8=slE=+?%dJ`)Zl)#@(x
z-5m}_=g;!W+G-p1?|%-zH@*@G9(PsESJ@w1>S`H$ZUOH?Koi!uw|5JKb^7{CsO}0#
zd^C_y37FNfqiaV3EDt_OG|7ubkrnHb5JtTz?dD=05hd{Q^leIlnq<hzsH}jj+WpRW
zEK%#6>d}!q=kGfW{ruFg)_3(RY+A@3Q#3Da^ktllM877M=Q#{^lsVOtKzF?dheMQG
z1J5I>eeW0NqDBZzgzsAc9shMuZuvLi((YE#_3u1gh$$g;0*um7dcG1Yxze|GCrV*o
z0O!w`GAFU=$w>_#-SbChJX`M0_-y^8W-)39omtSKh<F^WxtUhV{QAX#qK=<izHp%>
z5HP67*6CTTt?#^lun)@+5CKG2Xj#iRBVxjM!$_sn776^i($0z^!Wc%k>BI4GFZsR#
zyLbDhmEu+dE|=Z<F;k3i$uG9537%>S+YOhEoSWkvMES$LHjkFqndp-F<R^HRm5E^O
zNwYtf@gLHh6l)Z&x~mXZB4A-ZpD$SUB_h|g^(2yt-;~!Zes_`8hRK`{je^K-O-5Rd
zCQTl9%^b(Idp!E%W}K$_bu)Ho9t^@l2LTf<r9kM>=~-9JMWk)60D)#v9}|f^;XIEO
zHppR^?hKT;?8QXWc!;(Tps*|i!GdV>VKeZC)YgS=lv&Q^%W(XDGJL4Y(&zzz_@xAd
z2aLU>BAy1>7C@m_{k7VA_f9!!uM9rXhKHYfE>F==@~9dfsWD~Wbj{m9!tF=j&G|_0
zg@w&v+&*a8-Op^_xFl}RGG0Zq!oPBTeF>3kt@mNL&rh{U`YmW7dwy~<CmsZ6?Cv$`
zQKEfX*1m8Ai_s&z73ypz6u=pqlMEIQ0ISiEv(u6M;08q19<Q)FMKN3cJWXpAQ4cSF
zbmUDI5?q157|keaMLeKV02A@5B90OrnA~u@yWcVweeTrn4{QTjtF!kvXLTZsJTV!~
zx-FP<BLPojDm;|zBI3`L&>+7CnN1(YKUZ<~Ed6vDS98kC_J;o9{eRKU`m3PT58xKW
zMLW*n-zKQga>rHoFVyWn*cQuJUg$=p#RXhZsCx+(gv78(K@s8$WNiMKmv6E=Zv^NV
z(jE@UcL{U&ZqjK^x!t=y?)gIFM|hlyh9}`>(`)zXXx?HyFQG9+Hf;Ee#Y-u<%MNCk
z3r`lJ243jsYKdv3$uz<SmfqzWka$<^AkJm5h9!uvUM@)m{GUkvG80H>Q)4SPV?ub}
zRjV6bmtjOa@u_$ntZww9xhPI<M@mR|XFCbB8IvcF(27WoF@dyw<Ya5PS|S|tRJWT1
z14)IXl9-YyvDeNg>GR9Rn_qNrR8yZmBOp<@31VHkxj<lQ+8_QMdLH6mD_zVJLP5?^
z2a&8AeYDB^#tqla)dTJugI-DL2O1+GB(p8CT|AX4I~Li}(N9Vi7NC7L=GVRODgR?#
zHiV^S1!a$cKnqPZ|M0{RB|(We5!8h!1hyePD`*L@ANKv<0~Tp62w1SCw`qPF+fJ^m
zn*%6Z<#pQ`rr56M1GSSudO@--0~yY=>R$fCb#r^0GFTat8xb_U%OL|}y8$zIf;5b@
zs&~{&#irL+G}oSAQ+js?b@8DB+=^qJ-{!0^%Zh`D$CAq(N_26-q=3PM&Y?pQZcmH>
z5gj{#TLb@4=;_b3F1b21YPO(ns)Jt5OJ1p)u<bJO@|8?!3rq$}#bq`BB@`D3bKwZ*
zQy44X>Npt%Y-+975h-a}=hc7$JADc5Uwk7GAs0U>85HKJ?LclfBjc5@OVo9|J#+Ae
z|Bvw?h~k%uj1L*cz_*=Vfzw4RM&`d&P`C6-PQiQ`g3DFhV~8Z?S*gk#h}V58tFdum
za&z+;{wu&G+*{73mEac8sKcL=?@veUkF&^`m1O&5xOUHU-u?%Ew{7YiV+_42-{tm5
z3K6vAX@Fe78oXJ+i!4H868R2C>A?P$jq@AU?B01)?XW$)y!B_7VQ#%%pOOmcMIdL0
zC$HSp3HOFBm|k~LV}qC<YTj~fB0P~^mx$qIvE%UCh{>gf6gA8}9X<IT?TS+?krYY4
z$tP5}H2(U6&_Prn4T^cW?j7j4^C~+dj0qIqnIS)>{d!H^N5Re7s9$sGjAq^ItJk<0
z5S#nD-akf;vuEH)wQIDD753Vo0FW6#KJ3hwZ-S%{W?p6qF`Z)iL3kQ<I6fJR4vYH`
z%0Zu<8zfMnz3Y2vx9FUGKY!SZKh!OeyXL<bCwi96<Nfu}91shxdI+~RGCa{n>+~Ma
zsb)NFDq4^$I7UbxMz3r;oRFyc?u$XZavk#z2xck-F!3FgxU3+DzXV2xhRtr@p1j<z
z&D$P3-B=8C0zV^jwiR-WRxk9Rib8cvXz!^%5k&)Ihtic37tRO0j#1M?wmyQfdLBe@
ziz?I!Lg10!B{RNW_VO@BtxqlV2uiG4zsB-t{R+;q>-e%f$XjRFR<Ru1uuGWQHt^<V
zD1+b4v@LP6zX0y2rPps@15THX$RMF@S$&`9&nnch6b)S@UvoEc5G%W-?sYw9n>%XF
z-La|gtHWrts}N_+*Dc2=->*K2!P?A@AUtGy^@-$ImE%)4XRj}!@v;4N<Vw^6vrUD!
z5RZN$$dUrkGRk5#F$R-l8o#ZM7Y`EzxQUc9?fZbv=!TM^k}1Wp>?qM>1FukE<!<i0
z<TyxftB3ynKF<@|w<AvfJtv4Qfw2G@bt;v0T)G-K$U1(I0)U7d(n3EG3a_$_*=g=<
zzB8Hq$g;>S^Notl4bzbA-pOCQwU*7Tyq~UD_cb%n7{nB0{%J4obYtUEojqoHVMX~Z
z$KmV7qO72?S3BVD3v)D3UAuX+1A*lgobd-6b^6d;KgU`^lmge<Ta`^=nP{w2>!&uF
z$QivoWj+$QeY0hh<Du0us@%RwqSI-~aofY8*W<8CDSMAPawY2is;fLLq_<AZ)=s#C
zjBxP=`^#h5g86yNH{4Ww)pHnH<0NqgP&V(@RA$Ok?KF9pu1rw41~c{QLSp-VUGzYH
z=mzlHHSYulBa$vG$caVdKSXX(C*wa|I7Be5v8lM$KC;*FisR)3>Hvp80mAK$8Uf+1
z6@9u=8uKJ;j*15zsjY5bJ$d~ui3?w-DyQ~iJWxG6a{qG0_4oN}H|(D`+l>EYk1i)N
z4`$M@iQ{qDX9zpy3a}@zR0e|d<qBbuE?I5yvo)5PJ0J4!_ik1!adgi-%R0b+5`Ts@
z^}^pf+q~H5E)E{GAs8FxeUKANDexTI{RcijHF9uAhf2BL;KYQbX`eiNaAQc<_W&(B
zT1?!tj~D0a(jL5(88;F$d;?2+Ht#MdoVH!rg!M!@VLX@XDqb3%=Ps6MEPb?4ad|NS
zW_m+{8nf_PG{1R!x-&V++oE<AlLjU#Lrjc{X8@=j=&adO(lBx>gHnEK-%cpOW#vxp
z)Me+c9!gNxW@Kf#zZ;(s43}zatGaI(CfyoMf-84i*SK5xh{W*(Io+bFq#y;l^&%yj
z7bKAKvXBK4GsJVxaDKl{jjt@UB{nA>4Ij<bH|j*%$BPLPLJym4G0%qo7$Rc-Sr1Cl
zlvfF~3qZpg#w&UZnI&J@SL&#^bHEv_?j2&F+rtDz7_h}A6pCds<hqSUO`yI)U5lUs
z!TLXQmw^}l_L$ZE7Qc=_IGjn>7|EFa^5#vR!Q>?s0g-{GxoO2oLHmp-#rnir&Sx(k
z!Dv6d=Rv1jnq)@wU}yKLPsgIZ<u6=Q<bnIDiYS}8oZVw~{OXgMci$yg_of>Vm%4QK
zQ$^QR6xuMmQ8mt`7x2<C#$NE3c@wH<v}PkWqI~LJ&kiSzY+}6aSz!$Iirn;0F?YK@
zsT0p;j7x=lsPd@(r|pa3UWYPwDbe;#m(x;Rc_BOl01uYnrr=ZpCTy4;xd%CLnG!pY
z4hV+lH%i#up;oUXqS6kE=yEhLw_&gbRTLcbR=|e0rh2dv>d&e~|NGlXpL6mTB|880
zX$;zX3();3@bm+I5(UD!y&r<rpH*Yg%jx~=Z}%>2<>5adQ?b~gAa(^p$6w%Oc2>~T
zILnnCPtBFL=lm?n^rLM8bINumr$H<4bmNkClbM3T9EE{#IOsfX-lX-mG%%_b(<Ok9
z{zaqGtuGf|*fS2nO&)utowBBEnzIx+%v+!4p%U@iRmfM$sJ+p$nl{rziklBumSuqt
zGfF~g&*1&!Z)M|^Qe!4x)4Un6q@~EBM?9Q;lleS5U{VxCrtYC1a}N-kcaaEBa^EPI
z(1!rVsCX9%!z9RA?Q^2L{Bx?!53Z>xsFrDdd6_%Oj(Hp-w$`;YNQKruF&)7tGv+5y
z9u4la^$4TJ=blt{#8MP!H7%R>-g9A(Hg)pwgG#BohFbJ+42idzf3%I!$g{ye-cByw
zzugq-lTU5;mUBQIn;_SK5&A<m!d1=GkP=Oh1kcki!Pw>~+}PUC`v%sfDZqIRX!vtR
zHuqlG%A!iJ*okXFef~4No_>kx0WaR0l^jm~eWd=J7Hr<Gz_q6EY&LIyL`G49{7V}{
zq#P*rQ`w19WOcru-nl^*@#ft)Ldd#U8GfI<4Gpno3$~{0A(TknsLg864EItF^=}dt
zrW?6mIa2N%VhlZ#K#5+qX|6HWamslO;NvAZZ?-zYR?F?ZtlS`}D!-3~*7_B62-HU)
z>~|@3p+J9Vy4#!wo2nUG>aZTWxU%_hl{$W{t}69NOV3nnU+S9jpvPA0rwDSMWcUur
z18G-E*I=D1Hi3JfNRa*<@d$5fHZY3<X`2HsNIrS;Oo0$FN4Q9^0ty9sf&Y|OZ5;ml
zG!X{Z42#egZWf_OdOB^dqPy$zt)KMRN*dhgA4)B(h|Al;dNz~)`oL2#ssG{Qe@ZR~
z$`1E@xFBi3=cc3C{{g#qfTZ;mQ5)lUN#cCp!`WliXWnw8Pn?WUT~%W+T2e53*B~QR
z=~PVDUOehO#Ax@k|Dm2h^$~XKP<lG$aPKgS8L_$jw%Y9t5x^!r;v~#XjQX9XJ}v=0
zS?v8qAdLz=_r~$Z6a{(?cjpcvD}Pe8VgBPl5-?YPp$Vj0qg=f+8DOX1KEmX}4#JqZ
z`4}?PF4|b?T(x1ZlW<@o#Z_&ZD>~2&(1gb{l9mgyc3=~88=-aN9tXA|>zG43{@-6Y
z5W)Y?8rp}tjJuy@ne!U>wBR?h`n}Z@+*NOW)80s@>ML?)OnG9i0o<7WG*gCN_`vL%
zRET>oG!-tMvq^gHjfs7Z?ed_ATL(?cy4zXdP+(OXzAtT%oZqXYte|gXxOBfZA$YOk
zIFi-`sNJ7uKXvV=S*(g8ncU%FL0A=`2;0UsKv5RlH<v?O#e1XYy^j<gwe3_W221GA
zUo<`5^QV`TspDH8c?v#)0>d*ZV@v+wUN#XH<KB7{rTVmA-z`4I-)|cE5kNK}?lkSt
zhdGU~^3G;X$Pp605Eykuq+!kFy8u|8q*t5l!C)WvjKAW!J+Ymphgy3wHaSX=o;_CZ
zVO5g;6EA!u&?j^<aQLlRr=}Ws(Am+=)8nwj0)MF4JA6^7`wD-bg{34!oJ0}<Vk*rl
zXO2d1krdfS@UoV1+>1!;t=6FDqOCBTPR|}#@O;dxmGH_yTn&iDK+zCC)0A-G_7J#@
z|C?g+hyM^<H92b%LwmdgF-Wfw5F<8We@Sr$9~s6k5o=ueX_l2SCYuIxQRfu5g~k+L
zn$=J+qleqUoh$<rw|uFD(xS{7i<z$*)cjf`P7ywVuT9b#nvnDeqA=uGdWfsbw?eEV
z<oh)7*af%kJF%Ux(};b3s!x8r!e)+7Hc}81%!q{_RzCd{=#n}YHUE!=`f8<vN&DHS
z4|4>C_d7!X#;v^kB6rnZewUYiy|F8*vkGc@>8>^VcbX3VpNL&7Y51!~DUW*{Xa;=<
z$EG>cD5#4*6WdLm+~pVYbV>V<K6Uk&b-+aY#gWCB;oJ|lVAPW33~Jm+K#2{!UEosB
z#1ltRqE%0Pr`kx*Dcrv^d81gSM&<fg<&P81a{sE+Cjn}D<pU;bhlVlQU1+qq%?7EQ
zgT0AD4J<sU64k$|WpI^jaoQd!2%UKgGiiRZvu2kb7;Rx-i+n$8m;U*}so2gE+p>($
z6Qj42lqU5=+<J%qdYDTl)gy@mQ(sM>5_F4P4Ds-MYu6VXAkFH9m#fAryhMsTOTy*`
z@$+hSt4Gqeyi?BHOKZ&<29dHmZ)R`jAn1d^bxl{;G0mjgo5HH(5oCqximUsI`1zrK
z^#!TkA+;}4b$|)*t<$4>tU7~C-+2?bmFPit41a}CIlfbdUf^ap6=a42T*`8m^eGG5
zf)7u3(}Y#=kMPQBtqOHabFoF<{1_pBCd4lB=8oF?W86;24fjjGbB-OpkN>90DK?xh
z+bf{ZCM3E*HM3I@+-79lPl~fNJ__d*nX#fzS+A^z%jal_pIcgD`QAi{<}ou|bJsOg
zG1CMMJ!TsRK;_Z!f!lVeF4#taK&YB2bt`SNG`NhUPxo(s8dXR4?V*?sS|ttP$fKpp
zMr%T*F!m|YdiL!Z&m-j|L5dnvi+nyaA4_kAG}H&;=e37T=4E_24e2i|qg^OIT@u$!
z5R|mKI)vAdbT%+D+@T)+rk75m?x|8OD!jCFvaL>V%^6c{J^q#i_b$frhm3i~{ay|b
z0R92M&CBV7`Q!`ix;=#?PBYqCO9{M&YJ=ffy@}3^g^zM9=VN7l)<Z@{VznVuFctRi
zaooqlVeWGmStw%B=fYf0SW!`;pT%}Dx3&%W8hcqIwXKOEI)6y|DgHr18>x;uVaNHY
zDeLs0&zfXomwc-dpTI7vi?%HRFBX`?3-cFsN34B9)J*{xa;Q++iYf!6k9UenX_iv)
z7D%Y?{Ew&+s0$2_T{cAdqJLIsbn9w0iwiilb7?%rEB(j@YW|112t-^|tYXt1;!IWx
zcSuXx4tcY8s;p9Cg!dgk-cz@VdgHWB%4lv=4Q%h`!P1rbG0a6&3K!$KbVloP70(FM
z-KKcU=Q%b{069oc0pOH|Hc1B}8JmGzm`_OxGpHxoJ}4RjS94LIzL8r`SKLzDo<Fsm
zjF9k9>pPN}9uzW{+=Ik}*N?ZN(DP@kIhEaqbfjx5e~dO54CRPJ)?(b_fxhc<4k(zF
zsTHgsE{FP&%pSL|AL%#M{ch8xclcGQOh12PLk_Hsy4R3D^GF=hUn|{a-n)MXEet5X
zvjg*-{`!@bPBb)+im>2>{6`X_rBLF&=W|dV+b+lCI%6k3fLzZw*HAPW{45}?cgqvB
z!e|96Db$rsAF6fnzONS+hEet|OKfK3XOTF809`{v+DiOWTDZC62@PW(Z4<<ko7g$`
zQ^fvtisIEDhuN7BpY&`xjX1eKyHbAsVv@w6aaJT?SJ`&G>J{SVV$|qVG>*2c3VS_q
zYTSiS-sWjV7eKGmtQ7@GqLpo7MsDkGPQ{ArK6Q3HhVDy~;+|i&E+a6i)vcPHV*d=K
z;JJK{@w@a#Vr1uwTZ!;H@?qc;2gv*=!)T&4@hGe7&y09EdMf@FwhD@+-NhJp0O*+k
z?eJqPGe9o{)d(}kYY7RzYxXJl#O(G)sQ&&_UO0uS)E<Rhg4A4Xb6S3E72m<`U6NRZ
z0`a9Ta-bY9>T{U7`B300n|#4zfFS)J`a}FZ=N214nIg+!#<X1mqS4oIBJ)X6B3#ZC
zOJZj$Xv5yBWYirtDV<?T*eY6wht)lpQ9Blsk<S`0TJO>`avPCexr@_?O0GJ+vvRS$
zF0f9a%(mxs?;d0!96vk#G431b{L!EVTOG5PBE#NI2TA(IF^N+1<2VxB;$9Ajs9K>e
zo3b?e@jW605_}4A$e;uULhQpRAkpgPYHK2|mdGc;(!8HZGJ~*a1(H-i>Llhr#XQ<H
z?@-YX*H2s#kmWxJR%Ah>`2De|%sc!dL}PZe7^4y{Evev=@eHj*t+cHpQdim3O3iZ(
zOpYGg*L&R+DN;y7EL=-}TKXNh3I9C+b^kp8|1vx=Pl<DHYz)FDT0*?k#)z_Wl>p57
zRv0>P0buQbgVDu=&vKIB?7ciymuiEu#NoWwZfE4GulAq>=B(YsF)ceZEl}Og&kGwe
zI3Tw)ORRY2*!$j+J<oe3KumaPdqNzt{^CoZx=}w2DB@Fth+1=YQFp<>c)vv7KYssB
z*Dxh|8WZkG*zVp}F*D+hX)DAognXYJw^IeS9L;Z_z#wE-<{oNMK$!*^wZNce@MCs5
zLh%G>dmvNr@A>XKFY}P5JB7{EZa1)!D1n59)*Ph#y==CQuzr3VxXs*@+X#eie>}s1
zUmf+4ehGdgzJh{6#KDj8M6GRk`A=22)`(ge+fU)jDh|VJOph2>I6u9CGGAaf;YZ%~
z2NyF#Sr+mD6gRDmG~cdI0QTa)oO}OL7t5%bge>c68FQ$x6;v1&EGVcA9wTNu1CpJk
z^zZ|Wk%rVM+~W&w&R$B&zVxWse_TAqG{Eans`v#TMM8IHu%3i{|7nfPO=mgLxHlx&
z`2gAKZ3E+De1QC3x`A~!8fxs%mKPE2<*jkn&;=;9BMNS>2B)S@x^#W0t$1gbm+E=$
zxQfq=MecFaH5o%x*BE=}p7hI$BEktm3WW?Aqc=+5evx@)3l_bFIUQi2CxDU)s!*Pn
z_I(0E#QN-}ycmgM9f7Mr*-d13#lc?y1LqZRzx1-j%w6YY%dPcbdIa6E+Mhy|m|J!w
zJj=p=0c#yr_@C#Yh_Zij)t$R!z4HX6YD7%dII*5CA^LT&CZD*Ko|WM;dei(kFL65I
zZ`;Rz;4?WL&gKle{U6IHz-VlpTt=1c^iHcN6Z|)kJU%_Qc1?l77*;K&?5}=*8Ra`|
zZ`ckb#h>P;qg?<;Odf5~i0r&_$K&dizTgFcHxIg!wy-wy_hK6Ok38SLwAiR$wT<s_
zub0%VPks6(0Ep!qlbu#XoV&|Fam}laxoIcSp$C1d`MFswU-tOZ0w(gl<=O(6$UAQ5
zVTucfm6?jKw`%+KRL8LA2Z35D&N8pY3N2d|fCx4fub`@L6i?YR?I++N;30n^(0y{S
zd>h-zF+w5IeeT4$D7w{>+vXDaXUZTT9<klqsrAJ42$RD-Y$t;YKS*#r6NBPQ${6Jf
zkw^rdWLjCV-%(0+>#%=Tn?M0Jc9p;@D{#aP#XO!_%^b<s{h>(kDuK%QqV<92aK*Pz
zf+D`mG;bD(3>eh;p9KMh{QuGe8p4iT0S)ZZa$`!rz(r#hK0qIdqoe^7mA@WHVOsO4
zmF?S;|B?6B!uH%}_XBqBeA-(}u9UsK0&_{h(gJfw(;k_7&S{mL1(Oa_7FUw9yRFYR
z7+Sa$#|ccv<}YPcY}rcXGXcjJ<m0|;Y1{XQ6j_(Yc~ugiM1>u;Z<nahBt1trk3Nzw
zM2U20)>VcIT7XoEU1a^O>+2kcZ;Pmg!~HBmg#&>BTo=fST^BV2*}ps_rXfF}7;Xh|
zNd}!(#HmQ@6Xvb-@vY%l5H*R-RJ&$lr31531nZDX>ooBMTLrm4@Mkx)+~T3{_`<)}
zlJNU58A%V*`Um#8@z-mMD)3@XsIYj@Ym#~k=+xk^4g{-tAO|PHcs^|M5dC<uyQe+2
zEl%OJ$wlq^h3Q$LC-U7(0RbQXhc$j=X?XVfDXag3+)#Oum9=yKsuEH{r|SayqDns0
zR|GvhCnj`mCNR2Au6M5f2%g>ndb@7i>miv_8m_%ZQ^T0lVDk3mkk!+oi;?<{p_W?<
zU+ep=NA8(l*eWWk8A_Z$^&DDS9d?1AtC8N7BiKSnY-z{pjK{@Ampif51oJoR*&tl(
zh`56Fkp1gni?vbdzI5cOo52g%8`<6SAlccFCupX3PjG>DEYygW7jVPh!D<dsq;sPz
z{s~u{FlP)BD7Q<5xA+6eyJ3hx`Ar?mX?j3(N;-VIZN<N{=<m@H;)P{sP3mCy!qF~u
z7`i*fqyKk=@IQC_hU-71(X`<4%<<6<K`wL;dWl|uI=Ei)<Gz%b%Y{e;Rx}*wtKn}y
z-{d6f>3_nUyg;o;(ng?F3qW>_GeMYENSEKmn+Bf#-}FX-Mp@GUCeY1YTv*k8dh>+}
zGV7r6>l75&jS-&;1#jKc!nDL2;{`Y{$IsJ3VOfWLK_8*8bF^HREiE$nO9vrWk2|n&
z^*(3u<;c@&t%nESR<`l1W*;qUV0XGfQt&Ceh|Y^&dagm462$2dE1UH;B0Sm;s|`g$
zp66zjHP~(@T2Z18M>oU0-n>uGRidvO^Gp3C0&PvtHDFgt&Nm2Pk0(h0F48rn3PBJR
zqn?p^;7!a-QHigi{I26bG#cJ|@@cYEg^ZoA^jh!!RL8=lc26(t^Jj8;D2Ngf56jG6
z<sarRdX`_3VTz4X!ZW_BY^*g;lU-!ga)PoLQ`Q-hlUmLQK!IbqZ3D;y>G$UJIn83w
zBbYVL5-oifg{o<m&3C=MnDnJR`0ooR`EsG$*`VR^0dTJcA~}Q<{NJt0G^pJ9ESfVH
zr~~WwBoUG@XtLL`GsF}K&-kAkgs;yx?IUVe!NK9FaPWL(wz3A%%bx5dv=&d<>wm(_
zZ5L_a6whVp<t4^Y(@8U8T);_ZoM^RD{3BUl<$3fZrc@B8q*e>lD_twVG?7mBeWI`<
zYvsM30}{ceaRFZ|UJSYg+zU<ubAV6IJLHc8oV+YjBik+r74l>pO5zkiTQ=<AQ?cr)
zHfB{r8sK;0yGqbp{8j3Bkxn%uvhvMVveqc1{)uytgy$O!t#VRLoi8>I$n2JB^daZ`
zToY&&6BXCX<p!oAM+fNSPkaXrI}H*`<+6zfbjN$PU#~d+F*%6;vY$upp1Pc&pxXh=
zcJk=XAEatJhG+0}ns4WAk`X-gnzvS*aBO?maMET|$;Cn2*Kq4<zqU_Y6%M=72LhNa
z8~eTY8-Ja@X{JaCfinY+>iFy$-ZOz1p!zp&z9oKkAnz3?W6BK!lj?MO>#~nzEK_w`
z3ujUC!)z~E(ntf6>Y8zt-fz~NzIrzG3vj&PhGXvb4B(hMN$$7^?JUB9Ut!<=X%!kt
z*bOSxe>7EUhwN4o66R;|x$vo8kf%%F_U9TTpKw~cWpLEW+K5W^j9&|(plFYJaV70)
zpX&Js)V5MEe&yw+HYRyS`9Tj7fW3k3DI)i2Vt6B2`H4D`a*5fb+4r|+E`<<-WaYVT
z_bqe77)?Zm9OMCsFNBu29Y#Z`sun3gP(AKXI6TIQAC>)B58desh%o>CUGBK@51XNK
zI&`xzbe^7Mp3P48TF9lugcj$nvhuZq=rzmQ<1k*RYA)=0v$nx?EwI_&du%_5A6BZu
zeEcQ=ZNd0|>v+TQJr*QfF9o%s-BOi;Co-3u3G(5USga&xc)pTg{Y`8gupkLR1KXPh
zE<4p4IsgZ$lg?FSdfTU2t<-BL-xFxH2BY2xEYz=GG7{UX8hTQ4^+M>qxJUVQog_Gu
zs*>-O6X8-Hdp>MvhBi%~xNBcioD{XQ-Q_gJ3?;=SDcrT#yLr*e;J)D@AnAJ)Y;29(
z`aNTFt+Sp&r2~n<@RXoEIOMUVt3yT$Fk#??(OIP@b)x4WQS*Pq%deWpf=F8=^bzT6
zpP$tL-N7>I6O5C@quOc&irMelEr%`v7yf*1Z~j^3gDxH|!j7c@%1&>bxK1Cr@Ud7C
zrtXytZ-MxgH@c5`Sttx=zW{QaNSwVimWd$kg^!JOLk#VK9sajb)+YeT@!(r<^E7xj
zmr*=0kQ%;?2vXvch%4&e6w@7><0}vWM@qwaO@D42_71zd8<ws{>3Z$<ep<?V_X~5Q
zWHnTuFi|tl9KAfU?UcR@qdJ!`MxiuW91IoOizIjN$iw}tc1pfne03EnjG^o+3CZS|
zvK-2&SpNJ~g+Cm>yluMa(*-imJUd7MEV>WYPQ{j}T+Sy9vkc#-J-*xBjv{X3q6s_^
zNatoZnmlT$b7}gcBeC5*>|mH+TLFRhtfH5F$6a+L_6mo@49r!dX<iDn6@)zU6mEg%
z@JAU}G)(*V7ca6_^k1(Fy<8YXN!?+RThf(*%o(Guej_mfBM-`=PF&ElSKw0Ci|+ca
zu0~i?$6eR@-KVTJ{~h{MEOOA80j!?hKWg4g#x|}V+ntO>^Ii8>Z$JOIXK0P}=FO#t
zl+^KIjtAH{z7H$$rQV4_liPOkE>rIL>`<NeULv*ZRzfEe3ZZnJ$r5s$UBdK$k<g`Z
ziyaja@dp?82eUV8UEIqIwjlLCY^T|LqTJT+%9d_FfIMHgPVAg(){2;Jw-hDFXJ1Re
zgXcDUd^a3lv_V)M#uOUHkrA;sLgS}Bw!rQbXsILr(q6-ECC)1ERGe-x%_P0`Gkf3+
z0?8<YcKY=CjyeYM3)SG^ojKt0ee!*0$`S0Z-<SV*aP&35DxD*DpGs8oC+`h%7;8sf
ztK(8<^wKFhSFms&NrEe*nC~*%!gr~R=nZPGh1R*!z^Z#yt_0v?byuyUS781oD#ai7
z|NdfIyznJ=k(oH)7Xm!R1D~?PHxdVPr*19?2*IR8{0C!EYKF*GiRVwkD<4%{O)v3^
z>LEx;%K7(3WP@alxDV9Z=_BwvLSn>M1-R3m*|)0%H*IzK1|~Q+B~I4o%(zgZ=K>Zi
zA6%f}U`dF^7w<fEc_U|quAbUM@(R=KS9~#vR84q1$WMtTPO7uMApbO#yLh1k`IeM=
zO@<WrI^A6f;r!S;<NTGI;Zh(>Ot5@&76cWvc5+|@fR6lq`I+*c^>|l2;=b<;?6RLD
zw_>l81>J2Ufv(o{h@I}Ec)dO1Fk;GM)#Ld8&o=7*EA5kF!y>jKTt{{Mq7x(cr$Ui+
znsh!Lu4kib3|N3AxmTuAp`+!sk$IC2;Z$*NVf@gX^uPSY{|$=+vHv9w-6jb5E6yW3
zJ!8Cd8hoyP<wk%PJC*bccVMxPn3Bd$nJlBWNs~R6_E>iI2wy!3?~@@}JzY~-!`zv{
zhO5z!z5wOnzKZk6sqlx1$($n=*I0X#4B<x4dZr7MIJ)+p*927LW-SivJ>6~Ud-5pP
zdb@}er@rV`AZb?V+Y&!le>E0`OZ#%!0dJ?6!8)x9Z)_-hhy?ctZ>?&$;t3b`5Dh!c
z_Hkl;#&fVz$axA7oXTi!K@T^n#FNc}SSysUoTHqA!%7U-`x)JzCW#-se@5{{`pL0A
zJKp}Q<W4;%p%2KFRIi{qqrN!e`t~+O!e1$AC&EqiW6`$YaUOrXZTohCW$R9_Z^fE)
zH8x-vO9;C~UA08zx~^fU@@cMqsOcn+tmp1eMO`7(R%K@Y3M{1Q)?W+;5=g<m5_ieT
zrlGq3vcMPqlx#G$b8&GuwrA5}031&nogvpt(MxZf$_3HJK&797*G6otwpl`|RFLBt
z4O3!Q7vlZ>Pi5R-O2pM_DQ8)^)Xq+NFg1l)+lsFff0LeXgoVz8F@@uK6<GO;`mb!%
z<}85km;=~X`H}J@R1=!vUiG1-XVzFOSLYCb(7kb1*lf9b*^;@w*Sz#?f$ji5%`_Yw
zjjRtyfVD@d7SAW?uj`=JE(Nk(A?G2*kn;?sB}A@jr9`V)UE>T^#1FJq`!*#uM&BNb
z@R27jxBGRe?tQaW{DW4D^mF`YdRMu?twybK<}p8WDOb{az6hnpfpOYk2fSEA$gA<u
zs%s0}YC&$X=&0viDfwRyEcbCt{->8Wfb;KfCfCNWc{&StSU<QFLO~tH;;Oq@-Sy`%
zBg17>Y8CH!*q!wEdjd}EFFn%F2#N3u2rnIDhTSR9#B3}j@37u?x@Gqs-xLHHuggp`
zGvKNyBJ_-Z3vn)>$Ql{XDO9;VGjMTnpg1+Uy$bO<2spjtr`Um+>=LqL^3{+I&~`Wi
z@rHIKGlIN_dkuF2z{W2A)9k)^4;;A!AjEWh4==``wBC2`7HmS4bm`jLn8;awkUSu%
z0lcKOgibdVJ?MY(6U(xvoaKlW3=`D2i$B?#3XGH^LA(1R50wMd2;E1}$o!U>fXYD8
zKeqQqg<q8_^f{U7ud?!I4e7PSNuhueKU;VWrx1SoAS^NQfV3S8gvbWw-1;WKE<y=O
zD-l#(T+vXp#d}Q~{~8|u2Xg1W{id7|!GK+0V}K_CM_>AolP)@nT$NG;4dRn8u)oxN
zT-gSUz&c%EgB-r|A=VV{ai~1bm}PaRD`LFw_M&7(z=2P4W#8A31r)mKQkG7hm%KX_
z+Uc10)Z0e$r*z&07Z<|4JZ+xd)BV`Kec$nI)5uCQ=s@0ku6SolWWB%>=k1tuHrsdU
z(nU8x<y*mTmOhr1@GMb&yaY=g5rhUtS!Hl?8b3wxLWxJ)YWYA|>51<;c5U|sB3@CS
zS5u1e7Gq0>W5`y&R20~|cT5xQ$97Febd51|$=4+G>}+k3C_|3+X9yM3-+wQbqu3Dp
zXKHNbI5HJK4;@^g47dtNA#W1YE>Y8~gQ~k#@3QpKbe=(Xy&)H<s-PC$!*#^P$tgLV
zZ({TQPauz}5E^<Js5T(15xJ`lckxf7ERqxN)Na=0ufCU@&iSAMPp{@YO3ai#HoDu>
z>BED`P8;+Jlr!B9PBzc#l-9!}<IOWGEppiJ4etVeYzZH4O}Dh_gfk^aEO(AP>-*5;
zVD88PP|yR!SO5%MI+Ov4-3Qca3SzV1o3BeX#cy=KJ>Mx;ygF}ueTJ#{r)0KC&W~Q3
zJsv`fRAMdP>1b1*-=LLLV*~dPV=KQ#IG(?o^aa9oBlA<LPPkMBU#)LPeCxwj&&7&_
zGxzoG2^ky7c84A-s+tD;UNNH{<NNRZGsT3%rM^<q(76nJ9CTkav8nG41YPLHMir^L
z5FfMd!lL`6EC|bac&nGt(7z+J!0_`kZvh5t4r<<chCBeK-fw{M&CCj#7#A-K(k0by
zZIiIqySnyap^i?J1wKTW<XHT&AH7kUhI|&jac8W?=e$eS5qBTyl1Ww93f+1Y;O9Nd
z&sNomv#!x}A>2!7;PDzctB-tG-dxnU!)##ap<&b?=uFjc+1RZ>x&|{?FSlwo(|Fma
zzwCw;kYy>G7zthCu_5p`LOql$b@YF<5~*=dGl3~OX_RO*s#7#ZMjg}_>KY005S>4t
zFDxEceK|3ih1bj^0kj#SxWbg*D@i`j&`uUn$bjtrHFG=uQQ_9n@XroNpVYE#7Mcm=
zz<@-oYyI%#mZ$ITAGd*5(A}Fl(n)_w*WA#huKVs^iYuSP>p`nq*#AT%j4z`=eKYXm
z*1XvV%kkHxQ7W_vyK{_}lxV(SV044uwBXrx`!%jbX#>YSFEL>b2!_%3Qz}9{C7c}9
zRT3Ci;kew3+ud}0a&%p<G9}f!roY}9C=@UzGx(;OANw9as1i~}Em2Cc#3a*YjM#iq
z++o`eAo`}3MAmzn$7v@)rovZ?Y(~To0Z!^ftURauMcHqw6*vRPcceLWRzWD!WRjYJ
zbK(;@vSKx$_G1FAE+;BT2z@_OZK)hU3i?ACCA0X^Tt<X0H;5q!)G_!YXJVu6pYQW>
z=gc0snngNAfmF>&9kAedFQ|xBMr&5helv2)sR~|e{RI62WaO1?j>nqG{p|To8z6+I
zOr7`NtCoTFd2JJb8Ux%KrYVR#pu~P;|8<iD2jC5m9GAyw`Q<`^y<W7)`Ol9*27<RZ
z>`G}_%^H#LV8*k<Vnp;xouSZ~iQR(;9{?YJD&sT8oN&wKJ8`ShZ5kRXE+!zB!-8lR
z?5wIL#VrN~0n&jmT53A-=JWyBmD6@CdngC;4J^f8WCI!HkZT$RI-)A^MYt4^kXpv9
zBoktXdy;9ib3q8{&d@JM5sPx+$3RBxfX%SVU+k&5yu;5VgL5Y)C@X%be1_2DZm!Xj
z;P}-aS3*Vp2%zQbh`yZs>%%-9x>9!Z`QsC`$do)w%4RB5pazQ#M4a$es<HOy@0xk8
zZ?<+BM^q;CIR3xHRsH6j-2~sMk^<r4aE?ivrozu%q`D3qHc^4~Yw0&}{*AKHPbkrR
zkH+|P(50yltXxVW=e(EEgd?r4(}fz8d-v5hcgO1!Fy<Bw+-YLz!<qNPrbhKx-6*Q<
zk5WHBD`Ylf@w6QLcKMF<9^>_y9zf}|WCR3hk-2><;~sqSv7`!_e(SQ4+ja~1I=hI&
z3dwZ_FP+%xNMdG0Py+c1%#Ni1%M0Zg5vU;<A;E-^vSnDal{@I@*%VxJ(mvk@LLkN;
ztypc2LPoo=03XZ^iopQe+sxre%g1Lme@rPV=YJvOk++Qmf&Eya;B28IW7-mAtt2_+
z&8C2GDCv24%9DM6v{$Bh-#UHY>ly=Cv%TR8IST?v%yylZTikPyL4=?I>kKTm5b=)(
zHfUfOC0-<<D}t8pb$E=wp?9?(hdqB)A-4Cez|TdPvN#66er;-R2YFVrPLy`X!DFS=
zyL7lIeLwl#qO^8!%eu+*edj(&!XpV^|As(liJqPtXj_e09R95AoiP#aHLdWV51#&H
zqHkkNON#_I7j$o<rm7Hpn?<%REqy+w&c5E+7CbXEPkuZwr(3W)fxKh@TLj@F(@CSB
z$~Ul}=IUV_xfzM6bSV)?SXV3l**);y+p%>#EY9Ce=-ugueUTu!R3-s<C)o0_A8W}y
zo??{#lmJEUk{HYc{BpZCny^}VS$b)r`12>YAgiL?9U`Re5l7k~OUZcV@_VakHNr1v
z6RVU1*e=GRO{ns`py$j!lZ^~Fwaw8rw$LWSZIy6cQ$3o0JL9Ju63R9wK{V{a!MMBG
z&V!bdIkF#F4R0Xgpyy3e3dacNr+YqNx1yiEH9CjjhQU;}>#JTkCji>!XmB}l@o3ub
zUS$fF)4|(CwVoNVzcFVCQY(mG2tQr((al)2!k`+4yoI@dZ;z;TKdOD4l}BtlwxiwO
za&y=z+SJo!X4}Rp5agLz?bm|Wu=Ii=J+G+VM`p1P+ll$)U+%~U%v0FG+W88gL*+E;
zQX?d((#2}$Gq5-9NXHqm?5)YO==mA)K9Ckg?(hv{>^2AsufKR21o1Ek_i{z%AHSG6
zlKE4q$ETw7&#dPV`SWtP+nptP*@88=e&h4|6_LIBxQ0>VGk9V<|71L{Qs(JjUn?!d
zh+I(+lo=>4vn@Qkq>eXlphls@OaBuu5JILv`071<%?L;^;Ac~26btMD#lJM|(UqPX
zbl>m0q;dwIu0J<ZYD6P-uU-24N2+$WH|R1M)W)hl%o&roJ=rSixy1fpd*_EyI5!+|
zL%F&lUM#jQ5!vIYZ&b>B+D`NXEmPb=wy7_D!DyE6SM?CMU3TmTr(r#h{^A+KQn$Vt
zL3TReL217~55Dm{+ot%N!Ez&^XEh0+FGlaGA-TH<eJ1Z0qQp6_Pr|~{bUFT@o>MV$
z3tIw}BN>^lJpq>E@2w|WlRFqYQ$D2%q>&X<mYxSI?Tu3iu6s){*$j%nKt+t&vafMW
z&sokq_HEe~SNdayj)8^$(*5um(re-@O&^Tt_t0uwRe-06!c0^IMM6R9x+<jhtr0)$
zn&GB_(aW<9S`F6MxR*WuFYD>j71)uV7vAT!t%jkYe4f4@kPwltdxT0)dWSa{UZV|m
zmhNq|Pc=^Ko4xJf7~sVi+;!QJ<EX*}Bl%_s&N28sb_-j9`6iPRoV{Ivs#>>>!Ng5L
z{i~aYqJLDf%bki`5twr;iR+jxeK`gPv)=eCiW}Bvadth}Ina>3zJdz(3i=cvc3w_q
z@|I&ld+=7fi9;n_fK>1o9Hk#?5F$K*`rc1%b2K8?0Mx6ok>E%l7C@OL&Nm2v3LP(}
zxmrm(rw-@&3es}&3g3>9_uow}uqJ2HU5$?~La7KsdHw*e)q|vdX$L4gqYSReo+F(#
zMT7Pn(VvNCSv}o#S=LYyRr8#VplIU|WF8^~I(ESHZYWhdPhw)f2}I^Rdt9cl&LKb>
zI=_VlS!Oe@7mF-@L0$)WC;4Y4;+yUVX#@Sy^xSj<?^!5NmM-20Zg~4%Akrne6-7Da
z@%SOD65`QNoWR%WAM*6ERz6@y+pgLR0qIj`F0_*YC0V9^uYymeXPQ^<*AN-+b87w`
zP4L)gkoXzvt#SqMS7W&t^8io(GV2My$9NME%Wwdm>``$}O=G(`LQ;8b2+)MOz~U$=
zlhK+4ib04$2I>Re8VINcK~C}tfH3&;_U3*7t08jGZ3oqK$aBP|WFYWR=$(&_!J?b`
zD%)i&Fy`zi@#YJ4ZQb-+S0>v1xfTy`y!-RY5&Y-MsZDIjMxi&1MbE(?`gva2uR$2C
zwO&Yfe+y*yZg4W57zDK-xvWs=64OG_oviYO&f&^5J_W)DlaR?K>v#W&bi%CeVXyLt
zD-%pxPTHz`WJm|#<))rda{>tR%h8VYH~d@V_efR3;jx$MlF^y90qwy=ZR7NwFh>rd
zGli-TGh@4PM^;jvB8CZt;uCh?oN|J;3mUCjLY>=HHuU7peYPXZtqLT7uhb`#rp3Uh
z6d3h3c7Y`_C%U507y(44X@`D!u!Z7e#F(5Ex6&8oS!Mrx&8w{PtsdX8+^yWL{n1-k
z(0CSB#Q-Q+;#6*MV?)aA#6`$KZ)&d#go9sB%azM@Maj2+Au&0VXg57~m1ysK=F??V
zfTQW}2x3R!zjL%W@8cfGdCE3j`hMdSN$o_$`llY0|Gq<0YgTE&5g4rpmMp%FNT#I>
z)z3>!?AKloz;~s>qyMW44P;vw<9b^KSYQ;rchbAwVXciqx+8J^r!tnLduJsId}Q`u
z^P{?Ibq$Mp7PV%Pu%}m~JOpKzYQS97)N8HRH#OB(Kf%)?c`HJNzzG&KCDK|>tCQ7t
zN|s1)LNvNHhuk(+!r;VV@$$Aj1|W6FJ->naGXu;B$7n!NT`CV{n{;8^;@%=@{ONJz
z299v@+MU_>*8Z?;n3q(Yq&a;8s)|8B>>S`cNn<r@qXc8Bh+NhWGo!d-N{FlDN3SS7
zxC9l}=2PHYSX0(LYM9w$3DEfS(fL07=QJ$sm^pf*dI7Z7RtcO0;hK*kT@U^Bt^-a@
z0E+2SvLG~HT)fj`FFF$oVkYWlfRqB5Xnb5$eSeQ(`gcrp7lamy=^7!mHsl2awM7UZ
zm{~w%1_X-%iloinSQmX~zM3Qp<`xk-?{!a>bAmp`MNJ{6LCm^mpdVMJzr6c-91k>3
z^l`i`4=~`%CInUGT4U-j8?6mR?%(=&3bE<kBc21hD&`WW)3+TIu@@H@XF25^6rOAS
zbRZz1<e_9HD>K&C{X57f-ICyp$3ARDGIwARnY|xnfrM-v1csfCNg+Cyc#cHGAE(3u
zNnO0>q-Hto=M`Z-z|cThp`qifaz+Pz)qr+hnm!LM2>pZa-<|$vA?ZjVOoVRV5^0i$
zT|++C)2udD2ZJ3sbph7Wx_^D<*1}*F?^!G8Sr*T<f74PxEM|Z=PyjKPW&foBTEc`h
z)VrUBs@j)3?w&#;qC_6n0WX)w5*N2aCceD_cCEP(fd9Qeec?oQ(zWmt9?jL@roqd6
z&hvZ&6W6FQ$(+#b@j*$zG(xyR#c=*oRDM<q%rpVSs}ce_FWWi~dQQjE5ULH0Xu~&i
zJ-iQUsi($mHO`inS~yOEqwl~$9|Y`{e>D~rS#q&!{L-GP3)rQn;dsfD=s@-#BsJ6J
zf^~ubaCvTj#evw)2|IrKT$sM2_*vlE!lG}V(>|}`N@K1VWeNE~vP0p8Cj#0NXA$((
zD`ZDg?uW0Z!~TR@Zk+jL>WJaks5$;R0GjXE9MBcpBHe@v=J<6EjCX9EN~mzDRnj{p
zP4O#=#Kp#Cma1<LQ9-T>{^tJ;yi!J^#9w>OwdUhqp2D$3+*jJ+6C|jdMDIvy657Z;
z3CO67n$qLQsMxymnstaLFhRbWJ&cvaTX?vX#2Jkx-7YBb9jv_$wD=a_=CZS5{DdRQ
zRg|ZO1Q+Cb<4tPRek2yK@zZ62wEQ5Z)hx+o(tJS>EWm!rPVc2VyR!AP*E+xVvR6JU
zHdLly%{!u6_f}(*RLB65WvXx~FuQP1ek>;|7pgc>PF4pDMJ3eM55Y8@Nq_D|N%qn>
z5r<zfB=VHA7o8moDULWml90}bI4+z)ApX#bszcOFe@P)G$<7EYvxP-$V`6p}S{WkK
zxWewaa%_3wMRW=JN;A$f>X|uw1$$E;7N}c4U$huA9`ugC-nAEm_-QWuPk4sS$V&3_
z98e*(m$lBT46+=)kzIsz@|Hxwa?BAa`<HQdlyW{`mXS|lQDX73fcui1tF@ik-Em<w
zi;n%koE-MVv%O;>QA)|QG3bWai?oLgpRsy*V)^3?Y6`=`;#-B&JDW%*MD&to;hIX@
zbJi&X`u;M~3XC1#IPsQ_$Xh4TeVZc}!9FbcW$+ZDAvoWSY|l^cBk`8B9~~#<s<Iyh
zXuiR$=yM8@1{onS&B$$(vZJ(n1dN$a&0#SqFYfhuEAov9JdjlIPZNZ$e}aPQW547-
zJGj(i{xM~#PsR%f1ajEGlJVSX5=cIZE}!>a7gZqUT?;}#f-i}Q?0Df--UQydky+XI
z5vmBiI>s(~114Qk1-{)fxg9XE%ovmh?{5KOII!hkP1y?&G+c6sN0;$I3)h}&swoJG
zz#5#>_l%8@gAxxP%^a|0zG#ySv3wL+x95G1kk1m-LhFhhT1IK;%7+9w=O342jiero
zag#AtapWj&-{~8J67xU)5b>P2tu8TMu|a{wLxEE+uZmgSeRX}A!-dt>Oz#pI`qu6>
zHNqs!fnT9=;Hd6_clN84qxncyI5pJ|aeAQdZ82j2N1w817}!hPDP8SH87Bwx5Y&Jm
zxo!T?+%ohYjh!yY^kYYI@XPj3jDPWsAN7C0bD{?hH&N_dJpSf>;QdF79|hlKV#OUE
zwqNt~z*5WM8`Yi<GxPytJe#tb^eML}j6J8nO5?;oeEKefEOe|7b#LXeE9yVre?e2^
zY|h;~hon4;%7kzK3Kt#(;BNt^mhJwxugyhS2OVvxTsI6<kT-uuw&Y_K*FCh`n?MT2
zGHdCSe>5AH^Np`k61sQLqD7-<_=uPOhCB}fQvnC`?xd^mN#ZbX+11HcOPN(AS6>tw
zO0B`8fiAb(>Txspwnwb_)jgWYW%u&<PO`LhZ$qDFqPlSAZ!ad*sVImFr}2F%2+E>D
zLs!h+KQQ}nQ8EiDKtu9{$S>1>y@}l8`$oGB^*wz%Vh;^F4G;r$vi1F6F;HnYbqs|-
zh^nL=X+jFIx>sdaXWN5*@AVk2vEW=+m_%^vc@(Nk^}y`dz`gk7A0ssM-wPxT1%^D+
ze!*Eu3k8<c^{Y?NBe7^=-LWw7u!19zB!%H;4#MY_HWLa&Lp}aD{lFYux%E>~(*eRI
zTVA25nw@EbPvQk$3J%;(*S<>VSDf_kxs3a`NR7F5^)B+gk8bajwXT+1Kyx8HX~<zE
z#57-GLRXFLwtgqB-taRl&)ZE2vqMJNf3GOrf{XU)0`F<tGs0%dmCf6_bswRW=nvHG
zRdKrL4=Hw{wwgx$7}Z)IEhA^Buy0+jz!x?+kZ>ofo%#VlQptB9k+Y2^JgD(ZEm=8l
zEpIsl2$M+dmP_=nPKSf8PRW69k)w*^)BbDF4&-Xc*AGc{Nm&Qc$6s@}$mz6xAMPUQ
ziT_zOHctLO!u~uO>i_@$#}!$!rb2dE#=eWOh6<6jkX=Z!4`S?5mdZL~sm9n-AyW1=
zL^D%}vS-N*WjEHbe(!qL^Zn}i`JMA~P96R8JkQ7DabIrB^>(>lFgf$%B4KUVWdLRZ
z(A2B)gb>S&bPyy9Ky*xodK(2q0M^66{MC%XoDZ4M#-(r~nEw0!BTA@FMx0D+p$KCp
zezb=0c0bkwikW9^YFyK=$LS;mg!`iAQQbqyk9I-|Jwk84ZoC?L$(S2H#6E7Lqph`Q
zUw?Ab)Rb70qiQJ9Z)a7!qydD=zyoy*OeTH}ty~+i)6+KQe(YM(ZwAILLzCZ*{v6aa
zr2I+4yez$}8_cfS!XGUjy6k+}C=n)GEs%<8pwo$Y>kK<GaiZmDdsIs*b(AbUhirkA
zN_Xjz5FJh-bzKa?DLegR^7l}Ri4$P1Bk03lxx~F&e<u9mEClqg8Ga#FXg|J*YM+Eg
zs<18g;5zj#gqIs$tr8FiWtcF<ri`>D<I6qp+kVWCi>RYu=$E^<LkHU@pB5f8g#`n(
z%giTGK6vt`*_Ob1t)_M06BGVB&57P*kin)RW@Lcrnuv}s^Z^#MirO5pyh=t)%#Y~-
zlSW9SC}if)G`#DYE&B~$byufXbsY{|U?X0YS>up6sl+Eda26&ZY;RSZcH1N1wr`-g
zQ>J#c6(@Js0O%C=<?@CZ0@k!bzuD4UJh*tf5hjL|V7khU_ijETIRh}$O9)8~IjT~w
zZ!{KZ$zO0*iVl>N1#06u3y&mGs<t|5whc$Slb;(aUvKZ?*x&^qX@f>Cnr+ev9a(Bj
zoJiNoYtld4vf%G+Y4k_X&9rn+Sh~JsJ+Hzg50>`QqeUC&G6YqGQb@1ZEg)juCRkh`
z*|hmHiJ3g2dn}WEA=VuS@T9BYa08;D|4I<NH;b3I$-4utq!N)apu3w_5Sk$<Jl+Rx
zABaes<E6uN>`raL=yqk`*>^Q36Lnag7V$FPoBq<7wXXmZ)jK16Y+6PW`mNX<?|t{H
z_EVa7Usdz<+Qer7K5a&g?i?WYPrY;?K)~S26G^?>UZs%cj{X~w4=G$?R_L^3G06q<
zYLodnq0YL}3^3X9O*~5s*BJud&e-m*4Ro+IQt-oleiq&gz%>s+fm{CcJNih4yWiCN
z)z~*rCN(N<#j2^ArTVo*3ux|-jBm{ZlVAI5j|TnWH{J`1b#5h6yo2ES5|_wE&-hST
zLd2Xh%TceY1qj>05cLsbXE3ZU;w2zDF6wLf@BF|&nB0aJIpR=~&^51R1wY`FD<a&9
zcV6-_iVfi@(}BXdcMRP!KRas8?K_;oiAtJ1@LU^vF-OU<QCV>a7%rAgCC=OU`S%AZ
zz4E_@L*hZR44q>W_aVXZhMfCxEMYEJNi~O7#H&Esbtg12$0qRO<~lHOV=6(-4orQV
z&^t?w9V5T@jwha1h=UP5+6Tl1RORb2Tp0RzfGR)2!3<_?DPKLR@@eV)pu=x|35?vr
z=-jYm{d8KZKpA+L(@qAF`8dOh>KB+Jwm2?d`KyP)XZL5|jBiZlV4-`z#NaCB&Hh3j
zYjmg$L|h|Bf`8niG#$ZHkzORx-Md=@d-?nH^1w3wnhyB?C8^pJ09fG@E>kwf3^#~j
z=mO5N!?XVdJW5&;gX!?qI}@k(ft(L2$jP{~p>~O4MvGl?hkpCqwCMYv{h3ucqc-RF
z)GhT@7t82gJ-u#b?A74$#WTL`t0_R?7r6t6VA!i(0m+X3iyzTa!ZUvD;pz-)7XU#Y
zke@yyI(Y@1mOk;Q0#I4Lqdu5|`ji~maA}|>QH$6cmE0ih^HbIQWYYa(eDDNbVF2C1
zNPylp^#lMJ?Tc~$eQMRv{GWL0yTFC5*Hbq&pS}M0{PJcZ*~ayeBZGNJ1fmDkj#qUf
zjRng<3y)7Pu#Jd!WtgTx9Mx>K1H#i8pDOLfgN7jhlN`DY0K(R%q48J5hoYlfb2m*E
z0;nri?!~yB%0M)z=HpTkrACI&-VH)dW&=Vpn_;{3=*?i*P*S<*iV{5y`ZjaQNQLhc
zF{3M8^&2!TMaKv^9UV;pb-we)ah-&3?*f+0K;jjc9yD5b(jJ{-pwZCBjC7ZQ3ry72
zzIw=~Y5oEm{Sp}J@MQb!@|+FoMms8?0VazP2%+FUMh>(v4lt4oIVWGDTN3%D<{S{2
zlRLPTD=;x5j+8i9^2cFDf;>4!B&*0H+m;Tt%Q&+)jQ=WPQm<V7_fp(<WJ1zYS1_I^
zm?RGtfyeMp6yT@oXEUw%M69L6A~bcmMsrW0c`D$uU}_0_!KJuMr}71(*jo6FLvsc}
zNW{&5rWO|$!cW^pOz4|G0S3-3pQZWoo_m-W+{p}fcwlF@c+q7p!fAR-w2KDow0Bz*
zO22UNd#<iNQ<n`iN5)9%B4d$KfYi68?y%Qgqc7IS0#^xv?ysw#XOsfGFXH3II1BPA
zrp+ihUSG3@PfoEvz~lj}vFermW8P1%%s1vgY_X;V4V*VRdWome6;#|>Qd;iD*y<Wj
zr%$A}JOP~vdZ3gz9WMyvQu02>92PKXN30+A%F<h_z&zj5-%YqE@^~Kfbui1DXNb#(
z%WL~H?5_pZF6LllX!@H6*&rG*;~J5Eszl(3%9>sWLs-SiDNSB<Zbtm7Q)>EBi2)e>
zT;8UQ<RJ@9SkU))VUNbD;-&+x>C@ma>T@OKb9xWT#Mj5JmY2FT56n!@zWiVOqb-=F
zEx?LlFmAhtC;j0_$)UWb?epZj*gH#Y>nchg!?K@`#($!Ys(3RQe5r71J9uN*<AR|f
zZCA$^W?vImyqrXI?E^{kH4Y;HwK%<+=H)|bvRoQYq{R}}?=bsi<%|l7RnI1XqR5=`
z<Jw9MbDGZR05ub1$=PclB~2f*7P>VriJ;TzAzyApd4vFglu8tb_b8Jk!ltd9ZpM<y
z)*7hsZLTLHt4}&OwRfDbUL93V%8FiiG^l*TNET#Knu{lT5sZIS_y5GL?Zv;nS*ZEO
zpa<CzTOqqD8asZxeouRfL;$&IN<d=f1y)bMu(NhO4^RNTnVj*+b2VjFsiBJ|LZU`*
z&S9P;>6|^8rs47bcKZLV8-en9_-9o(z0hYRPH=$hUNNNSy8?X%F=CH`m69_e5MnDz
z7f&axdD2PRSIuVYxpPzZg`M|St=1g(c>HZ{{+da*>(8z>#>F>TyrEjglv>PwO^a5)
zU@KhPHY@4nqxQIL6Ts`8!%Fq{5)~+1cqm++;pdrRliQt>DUDj9H4Qgd!j+?A$7OaK
zoydKT-JpjZb2wArEJ-G=s|s3T<#C%9=ux)vm6oy38v=I{2hWr%-dKuLYaLIRz2l+&
z7psOzYn1Ny7&rCG-7WCGR`HpYR+nX7^`!w8rn!*lxKrjK3_jvKFia4vw}*%peVP8L
z`;^S5A0j?57k2dj05z51Yy<e3U+u3e1U!VHsb9{ALdfJ28c2|{C&oCM+9?RJ`E41Q
z1%N5$r~DC^f5^VcWkvSYY00kW2E!T|XRrhC%diuVZ4%r+bPy_)ufQ|Wqc+_%Zo~6$
zb3R_KkZ`9v`5<;3yHdLPxwcrr^4hLVBWQF(e3IthWaXg(-flGNgNS0k;8>*!g;4a^
zcbTmTouk|nM_TUYE>JJhcsSUQgUmA35`?ho%(`F_HVU9}1u}s0$+DATy7WkAe03u~
zBZ$Rr=`eD-a#N*pYhG55T%$;gSe|RnQ`G!lY^7G5#lK};o@8M$a^XffIu}CxI!&j@
zHE5y8Kpv5GEfuER5B`khvZ7bprwwhY?996coY;=zT%TX8<5>Nr2SuWGxeo(biNnxh
z?7X32rIx6ybbcF(VRKAbhV%|-qS*uq!m>a%_RjGr8Q4S2)+$4rUml3@)d(_hlnky}
zeus&OFzzV%T^u*7xr{PIeqV9ZthMphp5Xf8o+LGH!J&82tAx*%t$+h`5Mvd#PZ3nF
z1sfOf32zd8uaA7}x{O8NgHV&xv?91EgL~v+#N>{*3NTVIP~TWdjNy7G5m|Wzh-Jz2
znSnS=n+ZW5^hgAo#6m%Mp90E8>oMOx*95-JUsXLifm8}^8RuGHyYI%>qN6$Vx62IJ
zn3TV#kU)bbuLZ^Y$j=QPWp}m-XO)_SNJa?|%u^4&1f6G8bKzc%HL5y~UyjE2)#zUR
ze|-ypR|K@rRsRrOSARH^l*pp)s0D<Hqn~|v-s5APghY0DrN%*49u3s+(swyNV7>)^
z>NrMtBl4lminMY;B$NKwl&T&Hm7{vjLWDcNZeWqM?1D)Rx0&4!!3j`I%T5oqGVXH!
zQUP<900I*=@3#rXaAJz5X6>tt?(B^vJ!$bY3hf&(Btq?nqf#)D%@k2fRZdmTUrt|o
z5@`Vpc&wu|FQ0Qba#*8$D6%k<You!<RC}cS8<aaSM%$8#L~u~|;-gUQq)PA2`j6a`
z>0hZVpSsL}EmJz;`FxLlo1Qc}KO>Rn29Wql=#eCgePelrAwQw4nf3lIYw=x>3xxk0
zE6nKL4DSPV;pWg^11&7be~vE5?K4NH25350Gm)X$%*gzL-HEnJP|27nMIa&a2A2xP
z5h7T6O(>7tC9!WoO3uP;GwwB4V_w-;J5MTK4;n06ddhsSbB7)bWX@Q85`9<0jMN^l
zS)6qzf*~tq`rJ<QM0ivVYFS_q)3V0BtP>6KxwkU$jz2P*3Ruelgl`onZ&S8BndpdA
zh$|#dU$$gh{t<USC5fR*I!T8NkYtmA6&j@p{2{OE(h*Lg+U{7=!Hcnpzrd1wV|;&F
zQd<mh6)-U%G2{k40r*RpBxKxCSY({M28YMj34&>AF(VUz;X+pd@2CHwwEk3es+oxN
zkun0jMsbG9XgLAFaUB^FpS~IBMiEglXSjqYgyd^3n|cXx71=)efFV?xo5ECvYC@sN
zhwf)K=asuJij$ug9GWJsh-i{bJI1x$3r}0n9yxAZ`|2yMtrX7Tmk{@P20tZPTPstL
zjx7FiYujh`%rO^lFlxd%3AUO1yjm@!oV=zG;7z9PDntQW1)#Aed`No-IL8b?S>vc4
zhns$VLDw*m2Aj*E^=)ryZ!4QS)=2NKEQe|7#J}4M0>C){)P4AuSD`4YtgPsfq047T
zjs=9DSb!;&cz{heO*=Ab8qS$L@djWt{~rlb9jO27r70p2W>=h&1L*<7<f85+G)X{Y
zQxB^LJe+87U?>%xikwV_C3Fk&J+)1ZM2)|hh6k&GQANAgFS~qnAD55X8AC>&<O&KW
z^|2s3dmpXVVxdM~*(C*dMu5Jgy63W{<aCJE1Ar*{ETz(4zqmoygSdChVu&1w|F{Qv
z-l>5J2o)nhZgm?a8)UM4UquND)Upo8U<*;@4X{H}3BN^RmF$fLt??^Jq@sZ7+135c
z@Li+7Tk6y+j{g?30y`o$-)I>9eD72N1CCqr0Hl<TLP4Y>jx3)pEI|z<R8$n!_C4_u
zD;ksE8!&7L=28c)WpirP-9YlU0*W<P51zwdtzdwf&ftul5!}kOX_H~Ii50m7I&O}w
zcJ2x)X<9bVc<oP_)3wy*4vwPdF2Ll-W(*IWeP=%!kQc5YmR+SQ-()Gu-N8E`^09QY
zr@{vx9Asw9{ZWDKDnvFz@#DM@IBt|S7=L0YjUs;>BScdHL&n540JdZYjoqGdKCPp&
zk@u<6c>zI`tp5_wt3{704Gy1pDlFmsapWdDo*;p_q6C-^R1TXV8vBsolV*RdlP!lo
zgu3r40?$N?*1p>SQdOtkWwgb>biaKW=%@)*53`=Cs1Tg2)xTP0ca^)k*RJ(bZv^P7
z9rBcK0+{XZ<fVaDbLhqcS?Rm)-cQ~8K;aTwFDyE40Rkh_JGoL^&n_e*=f+pNKD1qN
ztf{nm^K1=ef4fj_Q$=EjwbEm-T47)?<a(vFs*o%w<k>5`N@#+PMuWYzw3mK=WsYi1
z%w&w+Sz%Y%b2Lnp+mgD_0?0|FpgR~0N0d8zzwPY`-}k)28qAjMRq`xAw1e+|2bzjc
zX8Mn)tX5?)fw6fq73L5M&FlY>uC!BK{G&e`7cseNSZZi)Nc2(8wo)?cg{3~S7Qt#<
zl{lGJYIqH8{rvItCo^>4uobFmp(EH%e!_GNsJlbEHpTuC1*rcCd4S@lOaoMv?~+g4
z%50xP0#nr)Afct-1jspx%`9LZx5@{H-_|tc{@8!LyuTq_?*UAbE-A?!_g+G?_W4&Z
zXIb0T82jE}BMiG^o7H2D!#!rRDSM$n)n}Crs+_A8De0rkNb3<f)hnin5;pYFdSnHz
z^9wYbPAW`F<)@Gq@+D-FG~&9Tc+PCe97M6~m#j}=0>60A-<iLfj6u%ZI+;pm9#y?}
zPYF1{aWaA_1BQd?&Vx0j(bB&oY4xN3lk9qKMI5w6Tp@c@A$uf<taFfoY=auC6+3he
zYld0xp&2k)$q^5x2=Ne)ok503ml-8X)sAxs5})LRM6L4)n;muuw!xI+cTuBk+h_3j
z;%>OvXTYMp^0EQAzBkHlbm|#sp>KMoSH4jE*wBWIp`lNt<e#{~*$6^kUYZ6^QgyRN
z#rG8XDaY-K)0>}4_&62&(-RslP<o|=)Hp7_4i6w620Qjn)>i>ddWvWV-X>IfZtmxf
zg1TnaMi?_pv55sq3mqbR#{Jo7WAcYvBN*zt)c`5ackEoA**h-7_c%s})|A-F#*91?
z%1^q*2VNRT;`DFKIRJx&4~?sLC=NS~DRdExe3|>|dH)$|`A+_4IF*kg^6d2?z945k
z7H)kjP)0%)piB`F_mcB=`CEt=2vTHCmP%u;Q;1zuwS0t?c2iTRt+pcxOkvR9A8;+5
zPzKJt^&9H;OSNHRYj{8murJyM-GAIHNfi`;SqRSUV1Gdm8Syh*i)=|N>!=u&m?^q`
zKt5z2jpurN-As9LX`pN*ELIQO7T)zT?d@e3)^8td^p?30_lB6ltXJvB$n`ftx7sDo
z_oe%_?3DL!Z{uy*=eR$7{8^Y~XmS%2>zp25B?v=ydJvD$vh-^OI=d~VmyIX^MuyR$
z6O>PcUQ?DIx>wP+@W*?bTm@s&(i0pXI)x?55^O@+`lCa<mAX_LAbX=vsV1jxtI-%q
zH$6L0)Y#XUP0;_Vzwn)$|96j-`ZIoMqZcMYRdj)HCdxrohbghSou$zEVXACXhr500
z5<-BmQ*wzp%~iONAYWA670c|`=CC6URK=h%c<qzf8{L0qEJznmkKNA+-45wJk$JX2
z0L06E(wmxEO{*~>OUfp2v>w)vZTis*th51jjh1Qn6Tq=l0esEiPY(GnI!FnUdbP7Q
z`mNDBpoPlbBPgVpnqsY>bBr*kxYSo9!^mn@*BVyg2mF8Bs|tQ-$$AOcf%vI!OHvKm
zuwjr|A!pPZLe$ldl4T^@prgU6kXfg491ibCmmXT!Q!LYSC&loJa4c{{653ZewDJHZ
zhP@@-VVi~)-A0eNuoXdexErr}TmD_}%_jSEr^h`c)q@00;Mk0;@8zrE)SpcAU>P|c
zVZe>RCNYf44ZAv@`4ESgwR|RF%_l})lkTy#0V5JXBR`xG9NN=OKP|RC&5kWvOs-Qk
zrJ+#NG#~gh;pv4C5bo3>kd3Dg!R~n=*@KWhq8!+66l0C*j>C4@N(xuMvKHiF?D)E;
zGh2D@+xPu5fJ%hxO);Xyl3e6C+Bx3vu?5Dg&{7aIV=rX|hbpg=g;5)Aj7`u48FAlQ
zLn}#Bw^OkLG$7^3QJ`?;jq7kv71Oakdl{R!b+f(oR?;O^_HwdXEnO=+V6>?1L+fx*
z-@J{Beb@R|zD2#_mQwJ$-y9?xvFRG2t|>+5npZ&pV^2K*C$&y~8F)dObtOr_fjg*Q
z`-I7w8;+qR&T5m>P~X4o-KVc#hdDyS3-ya?PAKkLeY@%pI#dYJB?Cis#U#E09^iYX
zy2-GbiQb){-}Jwj3QYVy{*dO^0xgb{v^rD|Q(@)@H1QuMgLn3!uWQqKJ-+VUh`N;h
znd4LStKpxW>yf+Y#|aO*bG>14Qmn|;H>ajG0N8pJNz?5#cUvFZwpSQ(6EMaO9g2Wu
zIDEMSIJd~p)6Cx(USW9zs@{CH6`64!Exorm$ho<Iua-=%1eti7ns|OsgBIHxO#_qt
zZ}xjGjDKw-tx$_Um${be!{(hg8jAjT6@tE73?`~oj<;cwLZGl+AG-UY(axdD)TRpg
za`(m|W`sx@^FEAg-hhFIU*t}JVFK#ue{uw)=b;$RnB)Qg!|7T4x$FZ-fE7g6ja1@j
zY*U}O+R$gm8=A)F9u=?-tXfs}_b<`XuLL!q>6wvKVKyY|HA}*p;Wf!omQ9`BB;04G
z7oE{w7q3JS-qaWycg?Yt@F{1=`aUVfU``QyHET1eeo`CBF|Sc10faZ_<R--i;7dE`
zJItiqyHP?`lAj5DJ&xAO#>^#F-DzP)i&2$udg?jI(L}*gpAI})58t$ifRBtl|IZDk
z@;^726Y*d>_Tb6kg7Li|{VY_O9iXw|Jd=3OdomRi$($icstDymSy3Uchmup-5D#Zc
zj7-Ci{2t#r9R4_DrY%0jVvIO<?eO?lQoyce>x#$GwtFr2T28$>ncC+&%!KR<{kC$?
zax^_9yL{s2C-u{VABT66w0vhn3;`B8m|-K8wieAz@BpP+`6hI|q57i0588eubI7<m
zAb7Eqm?qEFP01ytugeD+Ezxi`uAJ{^cWz5x>2^Kp4&pv@vH^i~$`Wev_{C7q7Hn(w
zjP+4>xst5v)9*^(rJ^_sU1RubBYN}xvteEs{|R)yzu*0qz4yMo?kYj@n^GZnx6@OX
znXt2<ZU8w$qKzyYS~(IU_2WIbqqs{9EYTgn21V-2;;&f85Q70j{o!MyzXWIGI7~c1
zryTXGOU^^}A{yJG8Nme2*d!WkoC4KC$D`4pKkt^Q6bPeta>Fb%-KjE~vy*#Yd)bn^
z*rGu@t<nCgeh++{xL2Zejr)q4q<v=Ab0{%<=oT6#^S0z9bM~BaR)Q)UYcEQoe(Z_2
z>cyeh`hv+H82-Eb2>^(4)cO$Ep;q(E5rfS1uw<$_TO$wjjYk~Lm>~zK^N3zrBdval
zd{qop>n-709w+YM!JNxGuE@Qm{aGoVr;Mh5MZ!$u$Ns38wHny8GrZ^kLRbXko_4%w
zM`4+2Q)Kg5L0R@YuRGK$MN6TTR~6$y_N1Otu8N$qd-}AvxKeCu_i~1zuQmYSHeEYx
zNRH4qENsr;njp3Vbcr$-&$!%xT{cDsIUH}JVCWNu6d5_O*(^t5p2dfXiqz`vzIWq;
zTUIw7<lW3SQiDiK3K#8@=qhtiuh`~q0IYjx`WMxb$KBt8C5v`CS-({-7rvNQPT}aG
z#l^XO2Bio_JE5pd1EyuNBrt7S4mqv~=Q>6ZY&BA{jI?1ZUv57-(Hg0f3Mx0Oi=|{9
zx}Z_9@?&#){?^;96FW3MWJ7y!gL{7e36KAHE%V(Ak~!#(*s9+R(b#z?APT}epr4SW
zD|$;183ym+kV5!ngJqV>_%c?{&d}|XgbDP1!dXuGp-QHC=xG0yZyk_Zkrq!UhB;CJ
z*4m^xG-m0#aX3Acba7QRm#9B24N!wJC)1oR69q7s*H`jUs*W2!Y1wUdzZF*|>I1G3
zJ>$hu^><3mUoRVZ_4|9hHIftsSMH9!OiB%(Fhn)%RPdb2`a_#F0&lHS#>3V93&SB2
zCQdyO^Wjzp=X<+Mk#mBGrjRT&2TgAUPDJg@8d;(|r&_sdhQ}ibdNC#->ZAk}Q?`|a
zc78f=3CVt2t&(=QD=f@WzS53vIS~U&TFOb}-(yp$x$FH}qmh5RF=$Z_4n~~+U9*py
zS>6oaV6EMKYEuU$3}X?-p|QMd<DDDFV#>(EKRmtfPpDIKo`M}R9L?~Kh0?Jw+pg__
zv=5;i>c2ZehyK&Q7)8K$JzHt35r;Xy`$<~g0K_jUOUn~r{&o`Q8zr@}4kRZ<ZKJWy
z1k82lkR%JTe!WK7zbWfOXnf9S06=9a_-ZqBiA2RyMJ-9&DwVuQ&DH@~)lH2%gGDBJ
z)*H`rP=#C>0oYRnUd|-o?7bk&M^2{pI11TkB$ZRmd;BE0uisloTC-PyL&uyf9Z>!A
zkBpyJkJ^GpmjjSXI?F2wu5y^;$MklFmb;o@YC7~oeBum+Cei+2SmQHC;$NMO|5;63
zqHr(?+MNs$wa$39d#n((3zTHgZ#z-yaA=_n=u>mrf~!lWPT#*OoSvbronT|x#I9&z
zv4F=FFXw`iGE3<8KZj5jn4Hh*eCe0(c-aBV%MHqWD(u{eng;vd^$%gw(<ZmjxhxOf
z45ufuhHKp1+g{xmv<mmycqur2GFA4CIv6@3SjEP7k_L^dWxTbIcR@bBpA5|^1w?uO
z*MmUK^{!@gKynE|8MK|8&RODM6m5jTpx53mLg?nfOcQ;Zk<efy7(Y2`85!i?(ygh^
zzVhkTI45u_=Es0OPzx}7jl6OS8E9g>L1L!)9vHEnu|A#m_Z9Wz-)p-;*~IdWZeE4s
z13UR9s%2@gSqg6lq{6=4j3m6kn$2}ToCBf~<vr5-(3my+Kneo)Slt@vf)R)^vppvq
z#$|n`tM5fKok;Uf#SCB}(=eF`PpxPhm@3e}28t7bz;|1#&H2LlS5nz0(_%M0{S8Fq
z>?MUa)ASz%=F1K#)+5(F8v_;2Qd>B=Ujz(jRQ4S=Ov2DZE7CSczys1H(=4jF_&oU|
ztu4U3wlyUBkc?8W!jy`D16tCi^C?91EueWjHi(TprgZ}>m846O;8l!qPc@##oQ^8s
zbO#Ol@%@>DHebBV->0wEqdz)J06E=l%-*|yXg~sBw~fz7nNwHS@Lxb@Mx^s5Y3^!B
zdh?~n4WXCK!&y=h*@_4}-o0!8PMZDINn2{{00}*>zK~3VWr);&@j?D0M=FtZY1xOa
z6TA&JK;|@N_xhwdEnIyOZ=Odi_lDRo?oNZygWDR&5i<2D2r>_<Yhl;n$%(24>fDc#
z1g(<9Sdo5pYKaTEDS4w~{1bY;=cCY(n=$#2L`}mEz?C2iF<NvOlmYvM+?-+n{4ZHJ
zoJxxa)^9rEQ8^y^(Gp}OoGO}7%%|F%5=F-U+{)sAdZT<}sOHUqh{>KJ&h8WUrUZCr
zWQmrzd(Qf;uz`|h`|;wFnSQ`|-8CMYK_B#A$wt{Vk`ox-Uka&6%bBNEViTeR`Hk{N
zFu7Y;02D_dhjwDC!tyE`#V{|A`%_0(p_<SoMc^ub8*nug-I^d@lhh!>tnrw0;L@=Y
zxdsQqQ;`3;2EbM))G7^?4OL!DheYKVq=Gm;*od=nQkOW95R$o|R<jx-fuX+(r19~>
ztbPDqEf|@0LZ7y?Co*l-hTVhy#zNfbNc#LtI@m*Fv%mRy)O>dBYo)BbznjG@UnCeO
zs3#)TL*;EXcrRQ206GC>G`YEbbGPZ^x%7gWFjemFqpM2erwYHPw;D+e;l2j{pLbGT
zF&_W)D%=%DRG0xc1{!hBe1Fx13gX+4Z`Cv*FNWBxECL18YGdfaq|y*No1beg9W{^d
zEdoOC>ERaP#LcM#_IW+PYzyiRUyK1$^a3~q^l2n+ov|ZH;`vjl@KSWncIrJU)pvji
z!YHsgR>SCq84a3>X&*+oA9Ys(!6{jZCb&P&l%Hox8~7#KUT=5uEKh(7oPp~*fYSUD
z)4s>(e|sS9`X>JdWwy(k+Q9oz3Vk$F!KzLj_{bO=N)$o{Okv-e+S4KD#$75`p{o}N
zv|n3fe#Psc`fR|Xb7&#=Kil8E%ZWO*BKqACAAnrr)&1bd(~ke8O$;+I1P1`4x^19s
zvj}}a5RB=jQCXdYNOdP&FXB^2v3389K`?A(-%NYWCEt6<6&Sy`4-~t|NX}Wr_r51y
zvM_%ptmxuOCwW@x<Gk|&S-s$vK7h;DI111UsVeM>+?F9e^lhyOv7=%~^&|D4;hhKb
zPtdZNCt9+<K6k^t<<9roJLOq$e_7e76L{`PZI#$x=YV6m(%(5u93=0L%D+8?uTW^9
zjFI~2SBk=R)ng^;nKS)GVgoQ@Mv3QLyPgLN$(8j=fu6~NCbd(*M63EG{5A_9a-;SZ
zM4O+t)-xf!*VG<=H4wpGnlnuS4UI4~`jO~GsDMP#7Z<9QA>Ah{k1GQwTeGgJi~*Ot
z>>r-+<|2Tp2p*1t4b8}Eo|%6?ewV_QzF^a~d8K;^+>fNsPZ}TI{q~}WXYSb1F%H#t
zpd8B$7_cd$y)`sivKf@(Y9w&*M+(LXIsi8u`4X_br4oMGFzC!TI7U$1F}u;Z4`Ic3
zHqiY2*bQv|ruLf`Dyma+LNBbX_RzL6l-pRMk7QZ_a}WU#koS|FyAFnJZtnzz#&w3Q
zeK$uxA1xHMRzd?TIWTO?R@`*qSGNl$28c$0vFi~mhU;S=5|Kk^5N~LVCJ~hi(?G}Y
z{`Cm0)(STzT_&%PHlvfsr<ELfUC~!3cTdPu5P-31QjCPbw%yI}Rz`B#N}XAjULWQJ
zcgH#jVa1sTf*kL#E>?2eZwW!@N4JkIqxW?uerVUNB(^9Y$GU@-4R~y_rZX#_BE2cN
zHN_pl61mqYyKx;l{A8l*gtfPlf$9cD;@DK=vTF&|k2WWVZHL5zEc-p4b*F*9Wh1_)
z=f6Za{|TuErk4)Z?FBxGT@0_ASIDDGe`QpX5nB}JhW#N(2tm?d?cc2OR1EygESTOY
z`IgI?Q{>wvAR5>T>n9SX2hdz~yZ<dXlB!lIw!ze6R3?hk1%Fsd18|dRy3P>gQ8=iq
zZOVgue>Fx#tF9tH0ozim8w%jK2|Oo}RO=i7!!b}q;gT~DiTI%^(0tD3YF3_2qUuUG
zYsmE)f$V$(f3Gg6IQ%wuFfABt%kFd(pE(j_6lArFITBL}IS2f{20*o*g@To=>FyTg
z_hu_4doo*V<r{93HI$FlANldN#d`II#?|&Y(&LwQXAeF!Z#D0YF9aUEmzqn^`g`Hn
z(Eqz|7%}4qa|fu@M`m>^ry8H5ibNo!E!vn?;=pXcWqU9jqe3Ppb`lJr!qYZ9YHLv0
z<V6i1EOd>+xuCVEe+<;Nf6WCp$Dnre;Amq~WjL3B9<+d;UqA?y&lm#bM**`jsY3R!
z?$}AQnAfiQ5JV+by9pgxHudhqm?=}3wIz31_9jnSVf~jW5Iy<AWVUSfg7L<pi~ZID
z9M$??RP+@9CG5H&gyRu#i1~PPyJzI&GPT#niI#UKsUky;T1SCWS)~49sf_KS^XhIH
zSWe5UG%J2REFA!WO7Fmhzv*#)@kIFLmb>jM)%!yl8-GVr_ku3|d#S3c9!!)S_%Yu@
z6rdhr8LxAkOz*1{5Sh{YxXIG{7|d5)SFk{)NznAY@_qYqy&>u!J=TAsZ#fKx1}Gqn
z4lnT!44X1~AaWh&Lv??K2*;kf$r{c-8Q-lWuu{{Y+;nb+a=;MGv8aMBag?*!@x?{O
z?V{mQZpe6D-FaXEc?ZWXhDgAiL`z*UUfhmNOTH8wMWM>saOkxMs;Izkk6DG{age@{
z<{ABaczodz{h$zva-WHA`a`Dl^up&O`Z4=rfSX3^FM<sriyhxhizrb%_<QH92>lyR
zhO+O!vp<+oom7f{4?hb-rvl#$#_P?CxAsV!pvBc#<?lIs3sFMAU_m5hzVVBS{A)ag
zn1#d8=+HKw=#gIM<047wn}jQE%mkV^pyhP2eTRfamwd<Pr;`f!vWI-L3PR&is3bK(
zk(njA%3xpuNZ=&x-HDJ3sQ?CfxXm7B(9l4q=NWAA3WGBG$uW)wDn^Q=x4ewxz*#V5
z@aSM~`H|YAujPH*z~PzpnLEKU%H~%qScV2`!{I^iaPnj(=oa$^Hj~5-wgzq9{{6+<
z#bm<fj{S}&+fVb6@r)dCx$RrmL{<LKlhz^H)VYb=$8`_x%LRPlsRqNQg3&-vKw5gG
zXXDhz*HEQP3cfK7`~<5n|CSK!?T3bFLO*%ZeL9lm1N=0K$ZpCwLcLEipM2_|#Sw4s
z4PH*vr*Y?58G*@Bf{}27RBGxQtFLEUz3-UYD62iI=k5S6fdqF4h@Rx^3)637_0{W^
z8X5iQH%8HNCy^jS4{rP1w$wOz-XVj%rIbohdW5Z9sP^38(T>x=fZ;x9X0m|sOmQv`
zC?5;;PmG=+L#m&y(b(O=y?zbEe{#>bf1zYd-`{AG=^*Xw>+YPF5`kz!k9TUySM3KJ
z3Qs)Jdo!0lC!?HJS2D|QeUY(#ipQ^1fZ@x6KiRm+{Xjn9W=1IV6VH=(?<Q_A69tZA
zC7|%jZ4|jso(r~AWa5E-myHd-<iX^8hlX2y{WR!Bpc(1Km^b0R=-@Ed@QRGhj8d@$
zM3MhgiglB&i&v4b;`>`~?w-#o?F@{nn4Gu^C>vlrt9(^vC3CVysGJf=?vK{MT3REI
zMS{hiUrJVT{uNg_m^HemQ4bX6<mLRe<mFVp$^U)i&3`AO+g+o6^dXbJR@#`N&kAsP
zf0w~v<d7mCYFkivyny65`BoIKX!S7PEB=r6lzGY%=a>m|agPDE{}+Tr3|d>V+Yf`C
zErh6$CGuSg(*msqX~*>E25m?Nv$Znq<8t}mIU`nU&JTOO0=919m>`D>@?N&0>gm3A
z#$CrIO-?AcA-;th9`EC?k}WR4>-V$b=khcTp@%?08FItOXC?u3YkBVBPv1Qq3H+hS
z0F^UUOinxz57V+{vh4GN-Y?bk;(RS*x$ScWxaPd(su0q={&#Y;^siPXuKQ92E_6c~
zG~TbGd8zHp$jm^$O5VM;1ck%Y=7cgjB=CwxXi7a2)Dth|r+9I*Desydc3?sqhcOc0
zTC4q~_)|e3yj8mUbZM|10%Dy4pvPk*dj|f17h=Rjag#NotsM9R+_BZHy>?Ul3TYGt
zukfjn?v>g%y6n|daC3kA>agFe;j^$WCyV-CE`AslAp!rNNfd=1GcO7rrE`?Zdyprd
ztEDvQ1qR%5=8r@$6d5bSsfVn>LxjpvJ_RVC6Y^I_F7%yfK~He6bjzpeka5qev;FP5
z_&=nhdt36m8vBgRZ4t`@2hhe6K^N9Ct{z<L$u-Xh#DMhY9`HOEGm@tVV)g7QPIjpG
zfj>?yGAw-%)twe)da}~_CV!^|QdR?0x&J3r2Cj6F?ZGu+t|9?~0Ds6FFxq+PVPeO?
z{6;?tcDoxWRl~5}>=7>wBxk-yLH+J#I4h3l*xy-H>n?HpF;HciW|UC6d7|G(3(zhf
zAEytOd$y!;L+jp}GxKte2J>Ls&=|QjfGIr_1Qv3SEL-MJOz=k`8+gmhg{3*lVdeLa
zxUY_~dAd(@Ie=k_^PJS&)Xx~(w(1fJFa9;Y$)weA?2o{ODf{$UU28oH{TW+nk*E?6
zYA*-$9nM0Rr)PEF0|T>s1Cy~E!Xn;|gLxTgfn;}B3Z?~=XE2{CpW8hA;CAJGScrm@
z8E})k8Bvdwdbun~LD&?5%(%bKd@>(7f9w9Ed1dHO9txF;d^1krA<S-Hm0lsGrxbWQ
z^a*62l0*TfhhuV|qc)WVPuMFP?QwQ!JwB8Fw6p!^>!TzbUZm|jd*4Fx6O<PMIMH{I
zx5b*=5eP&B8Rs*{`@B4$w0xpnk-X`%hn|BbOmpyf`><9oksVJNvv6XRqns)zv!~aH
zT=!F_PODd^tLe+X^JG3dj=qSN-=}`)xnGx>8K&*New>~8oxNHx_K@)$NX7)%HORJ7
zE6b!q7rLC@?1!g5HOmr-$%9qJXl0_EXS%l474iQ#LIKj^uber80t&U!>s>*{oB<56
zpye19W!0cqsm>rg#pYpe_0aT)Qt`$L8WG!BFe+c7HbxmHr6;al9h06jYSs5by=bQ5
zAVOucg8zED{m-zy&DRfg+zL9Go;sU-T6D8tqri%$-e4KJM|O>+*lJHnbS#Ie$JnUl
z<#z)n0cz$R?N17~uhx9~)Mdv$-}ot!OJjQR#AM1OH)$_YyPijG^N~T$Bl{8^w4xQ9
zJLc1Xw{*a|>HekK5_gRGg*)-lA$rh}>qHaawIubza=2C;;w?~i%{<k@v?0S^iWEkl
zK0acXo$c-^xS%yGI@fc(^Krn{hpD&=u&PmW#J67j<kE!3{^mIGA`jW`pCE-e<Pq?D
z==y$f#BTOkfxV)GX$`wQ#+>o`#K-wJ{n~Vc*hHoL<I_e~_$w_*%)$q{y-8XF)$7ux
zXvQuzy0q9%k1bS<syExU=JFtX|3mb4A$nW0jMs2KRBE>=oT^O3IGu8JI&pQqD?=Cl
zv!_)L#~GZ_M|hq~RDnoj$#>`0_U1?IOuP;X5;aB#Z?8kSDX~e71AfigqVAZai|wYk
z`~2Ud$q5@hYwl&_1fjEDl-RbXRfJ^QHoRTeal%+vYOsEzDavCr)i9M&!Z`hHrPN)R
z?p(5YmHA{4kAkV;gPmk^(G+5Tzhqhse4J3x?oBGtWU}RVIUttpKawK;c^NHFH4Eqt
z|H0q9nbcf&T9nZuEhBaD)XYV~s>_ebu(}J_hwo%~xGYRyjW+G;ej|P{?l>v@oUKuK
zh%XU`+Vkwh47f&k`d0XJ30F5OFr_P}9V1+~HL<HJ5gbp@CZPN?^a~n{O=qmFKSXb$
z;$pKogzdRQwz3n7qNW*Xkg;VE;>KPxTn7e5jRT}f{nFiuQN`{PzH-Lt1r<-`Jw!Gq
zQH5i$u3k*9Uat(UGnrw1V=?s0c_{uGvJxMl)L&uF_YK!%^kApF|AlIxdkDvgkhC6>
z!%hwn-B_@naex*+^XFk3Y%gFBfA?uy+M6ni$tC&>1?Bc=WmuB33hHYTQ^pe{W5uxg
zw*8M|-x0UAWhIwnLTw+WjkxZ#Uvv46?_OQBU;LQuA7gPsxOq|O`V$5uqOgD&(k=dJ
z?g8;^>>_0f@f)vklV5MI%f;GNIo5aX$Ph8NaPWN1pB>qDgL*>6<il_6zI{PPdR;iR
z5@yuY8E_(Ob=0iPn`chcxay62CQ2JBmR;fh@-Fcm@pEn?p|K#kwDB^qt#RY%73G{u
zwn#mvXC>D83}Ie1kmZzCP+3rF(|U76!dQJ+{Zjj9Szp;d54$DGBtj1U(ZWwzj1|0^
zSnNV5#P<4Lzt%+Yerlv`?qR$Ww22et$EH+`b9_Fk-VOh-PSW_6k%Z7#Y|K6b3txuY
z6TdZ&J|5oyozo=5)2b(&qA9SKv{+2<#|%M2AtD|6Ry7u@V0y>egh2w<Yx6Vu=I1`S
zfn=WW*~!%@%)nqbbjd$QH8eYT2|v{x0KO<bc(EvN)|$};YQ5;56AGSECgVLmx-C|G
zL3~NTGwizY*yZ8Dv4V=s2E7N@QqN&tGq^u3@wA+~imtyF|8z6@Jjr?IdBw+3u!^aF
zZQ{eHZ5g<$`|BH{eHhZFg0Dd(UoT5bb4v*%k>gkVVjNdxU$x<@+ErE4^EYq5^WU-Q
zw!D_8;}w?I&0fb0iAaCbP@zOxZR*B}fa)@?rg)&9U9_9?$xKfx!E%~K)c!lZo?S``
z*4vufmAiOGVcAEGRO|jB!$`FbBLCVybj9KM)|qaxYM$@``S86jejQ?+isMeRl?tY4
zJMX4FsKiii3A?b0|9l!_dLq>SLx0y~So)n*#q@Iwp7ZRS`qo44dQY88Q{OVUtN4{@
z58p|&B|>khe>p~|LWLUBAdITOc(l~vSUx_M>d)5=ivH&in>Yg;^T1{m5P9$NuI6HE
zn^Wlc_Y_=VL46%+*N|}v&+|3kmZwa%di8C0ue#Ht#93S!Bw}<&9Wu77s_rel^V9V}
z{X5<$IGa&v)jAIGv;xgJgEKU>7(mRdTjJ+*Ywuw)&cm;h6Npc)h67v_jm#N#jmDTz
zF<_CJExR7@8L?9CB_UvMp6IRS8amMSoV0-mbIOHpy&1~jnn_0K=b%o*%oioHIF+rG
zo#9J8>elt*-dw}it*pm%tz6s|>T<;0BtkfPm9Iao$SLtuy5r~0@j%;|M}X_>ciwDr
z0z&dG%>25s5a=ua{;@xO*5&Zmpl9Pd>t$zWBX6s96=T<bT*|0EJ6(Kinq8Q&4S#1(
z=cG<ZDBIzIHu8ihZEmPQa;L{DDLncNEa5#559=+zh}zZg&q^e6djphe>8gvJ+Si_%
zH<EWNq{J|RP`9C0yh%=j&&x6{fe$B}lYgEktq9^@Ag-u|g(m(id$2oE+f9p<D)a1y
zkL}p1&vjB@n;MGwwIsVe0!OM!fu@=s8TMhLE4TFQouOeXJC(Pu8`9qr34^-BbtS?}
zT4Ms8LIt)JcjP+NOSCOJyOdv+2xkj$39eV|#3|`vtM+is+i&(eq!|A^_r#|sVL$)*
zm-qXK?ZAWlgQ2nrLA|z098051+WD;l{RWgjO|12W(8Tn7o2tPr!`@fiO9BgmT$egu
z+bsrgiAd?p@Z7_GzA<~Mj~2JUrg2lPKQVmcv>iCyU@Tx6Y)r`Q5T%TYjEk7X!4`ON
z;e#LJVzDNy$oPptQs<K5-Xr~%9qH}H6+LfBNyhFK%bZcPb(aD|LN=?g#?qdD`26fl
z$fX>cgJ_TYU5>DyDJ~cB#H%ygC`92{?>pky##nA?Vdb0Z^mE5c^C&Qz>E;!Sf~EX#
zs~TzC=7{5XPh<c=8#WL7A&*n{2R?Ro0SebM2zZbB>Uo`T_@5U+=FP7E>mXmgyAR%g
z0z(KwXy9QHYsOTtZSBR5z&blKuTGzfq~@MBL4vhAHtW<vSbYsc5sym8JEC@(1DJV#
zkC6gEZ!U;^_Fm~R>{RztPJ;c*x*Sy%DPf2NIL&)qWNIaZrC`X;QMVMT&oDLnM`Oz`
z&XcyAI%9jXIc?0{B6%pg`^GsEvQ_r>de^Wk1iAMoQ;39jFOoD8k=8l$x2{!GSkH;O
z;D*e3a_vl|NVk68`kB+yztPp((Lda4MX-dvP4~A?N+YBdTnXDZC))L@i`?Pka)mNv
z1z=$s-7XT>C70alrJB`e4wfwrY+3$1ev8GUf6hB~E&8;=5nOv)GW4hNDKvc|3tRPa
z&BYH3M4fymP7x`}4294{7T{Jkxf77@G~47@zH4&rr}aaG#!_vw18VoK)J-)~t~JN|
z7abT+D15P}q2t4p5()PB?Q=dq8srFE%k8^ukrQ4=-QD2eTht6Cc3>ujn-&?{9VTil
zSysEg*SEkJxc8)P!X`UH%(%-@A6skFUA4_Fws}0uBU?O6OI;*SqHoE=`)Ydfz>w8@
zK~qT_rZ;4yIC+u(hGz&zqAkNxzPJnzVk)_$Z`2pWe2rLalY4>32q^<lTtd(Md)|dJ
z{XXxG(dD`PH2s#PO{ftUt;Vs?S1S-4pNAIxbRqE2FJ{Lf^GoHbE2r=5Pr_E0n^vZ#
zl56i_|K07(;DeJ&-(z|gx>5e|$PsqpXhtPC-UY_LDv=~160=SZw?~I~4V&ytn!*SA
zX*D47=IzlDjvDgD;j;l#GsEGqFw%=9^fMMz(NV(Bf%Hz;b>pt<SO$Xl=FlwHtVEVL
zmnV-PmjpPzt{o+K-n&zEj<3}AC}D?P#Je@vUdn>6RC#W!_Z+U?d4b>m`HaRVPMxpP
zwU@Aja*vUA%%4v@5gYUS6X%ahtq^@oNDU=6Nu&w3OPve1aG16Lf@4lY^7q?%Q0^Ga
z>bt3j9Gmn@&Fxs3*Uv-SVs)OkW3mb2W#WeW4NI$V)eX3|DII8>vvWyt3L1Cm1u7VL
z4P?}oK)T%vR8I?QY*ZgHtt0hJMAOU7jOc%K>a(Z)`t>{2<nmT!hRUn#In0de`k7%U
z`Z#iA%JV(A5x{>W7eP?hFB<*sOuFS}KZTFR+tz+D>^1Md^dKO$zoI|4qJXnsd3G<w
z_LViqh>BharZX6AXId8LnB$>GVvQgbg|7U2pof?L{<tC1acLm*3NnGS(N*V5h%rjV
z-X%i466cnK8hTY#$IUiBowg{(ww?Y#H8eG2^;Zmf@Z>vw(l0fC7iS$kRh8WEG2<8^
zaIxz(5o&tp78>V196`~_IFQu6bh>ZATiuOUE#DYKW=$w{A;9<uGleg99>XT7PEGF&
zn=-iE577@}$f^j%_R0*r$BT2F8g>nla&vXF;3wEV7Z=taPM3!EJAyl##;p(9pIRjn
z!tvlrMOA8*<czYe`U`Pc*Xpm%A|v41R{Ot*j4b6>a2A>*bc0a!T-8pUo2p|<30{Lz
ztxOG_Tdz)3(WAYe9zDRIO?dn&BPUy~S9ucm9+SN}YA<cL-@+Qf(;QvR0@3e2_~^M;
zvnpt6r|g!Ztsgr^FbstL{N`W+hguq=t4|X&QL5`>QxWfk4(`{lcu)FeAbG+C{pVEj
zR`Q6YGqzIF8tYkoU)tKz@SR5q^S#0<D2XhTKb9z?4+1d>E*Mf8I-&w@eSUZ`v!Hi?
zok@R{IHXfZZ?5aL-bR&~hx0;erNwpIv4PWse5|OS)X%HxiQ*V>$YHqo>&NDI=*fCG
z{dvt=(ChP(G$nVL7-Q5_u@Zb1IyG0_pSkqvRp*z>t0bE?%i4#yJeZ2SZ{y0uD~L)@
zbv1fd)-`-&4x{l%sa8#K_5sWydg}3112S}`ATaD@3e48*RSGO*v7|{I<<+!VTC+8~
z4>?LG>RKN?U#qtCb9t++$v#-zSkl!m#HjkfjHg;)23B0{4>qg1rLvOS)6%f?&EuYa
z*VV@?heaDwE+fU=CEQtE6yRcR2+pU;Rc}f>T^F=oC)woT<CJVzkO3uPSddMW(I5^p
z$6r4}pPJTw9Z!)w&si!7!SI&@ZzShX5Ir3|Igt(a5{}M=OgN`V(m-l#r-~kqUAVa~
zpk39@#cOFKJ-hoMwywQDmuiI=VE$rGNS@mU?%0kQA77dPw-L`OyfdOj+q43iSHd4C
z;b0!0X_l_6pCy?Dx!*d4tU(`~!F8VKGeW6PO(^8=T-sFK?-0G~GrAi&^?qvmvHEhx
zJnnHxvUq(5wyWy&jCm~uHWyKtPiQP`)UF>Jvl<h*;SNGBF*gY}3HQ>TOZ}H}vo4o3
zRK7KOkWgT&Tkq`o#-DDd=d1A0#%DrvXJ-us_VM|Dlc<Sv>AxcH30l&g{lTo}l9KZ6
zB{t9ZLEIUYgnljc(_+;VL7c+X>v$GnMjj4H(dw_$q#unP*#SXcvIU<#=_-zcs+)~$
z!=s0cG<HT35oO)`+Hr`HxywkK8&i)9Yg4BgB>b)$HQ{mtYb8W7MnEL?v*5lq=0yzl
z<V?gO6|Gv_eeYvfvzdeB^YC38oZ^^!d8BIar!U<K>({PUxnLNI<v%FsN@hUyp;}cq
zdwjWnMQFv_>bHKkT;1H^p76PzXBV);x>jjBmGR8t-r@_b!+OK|!vTq&y0Mb(*#%A&
z3?g&Sw^=WRuCwZ(B33dYhR1i`|NG(3GymSlz6$2)+~OxYq2}D;Y>Y`PvfY#!ezsIN
z+63Uv`0feFAmL$9Zb`W$ZNlWtk4xDhrzGsXItk%F@2d<9G@y6&T!T*EC5|K@3XK?%
z`4WOE@@BnfbrO*Im#~6{pdtVsd=gf$Jxnzd(M83fR&*cRO@s8_DL+plntY#4>9iX_
z_o*)>oF8HkEB41q8sQQvB(oP-gf>mICY!{C=W_f|B?*KCLd6qt&x(fBw^eTq1FnFp
zX2fGbOWATbFfqD7+cMUfOJ(j$+WFej;Z1e>zysR;|6Ahd`~O@*b}7u_vHo1E3>j&$
z8B?7%pV*m8e)gr)B6KFP2ulUfAX)3|ZgIbY4s0f_gt^TItI}oKgMIqRmk7zauS7zG
zVO`Y^j#MH~%_8OB$ot)z>`{e$gG%aS!<5TJ#t8)~VCB?mXZ3w`B_*%pMMR#Rdt6lc
zVT+k9!X{^Tu(R&8aCSD2jIS3|Ol9D!GY%U2roiJxcZTuSVKR6-^if=6K~`bbZIDQS
zryn@xqUZkA2U-&FU~qUi<}Kq--noe$5on_AhTF8#<yqUdTWRN^`ip`tM}K7zhnqQa
zhTrt?Hj1?Pu^F+acq00U#g*8`sU4lq<u`Bp`r4F=Fgg)@YZ@Nm_S%$4DaOzklQn|@
zT-QuD^FUI9_FU2>)NXCtWHpCybNh@0Emn})FkSinO$BB2Oiynv!2wFDC_ugSuZ=(>
zZZZV4jT0c0Sc!9quBcm_XEZh*fWT_X@7sRxPhw;ksi9xc_zq2~AA>8du~1zn9`Pms
z8H)6O8&Dof{oby3j3{~0L!k6c1H*geg0#ZQ`O1QXf`GxN$^8{onAeqW{eG}Lh%Rh;
zV|YAkIxHboKW9|yr#STByVQZa#`eQsC)eR$E#&`{LY6P@y(>F-wjLql`)DZ}q|cn9
z)lN{~M^pGa1{bky#<iu8Y9+XOojO833AaaU8~ahha{sw5Y3I0K^QK(r1nqUB@E?Wh
z-lPmM<Zb3r=ofl(Br90AJ+hr}C@D)y42!clI$f(j$eG&vW)451q5FtCNc<}L(W3>-
zfI-;BwW|d-*=u<6Hb3i%vt4&#>D!!8Yh}XoVoa~hh>kwheTJ${(PB(U+G%`NL1{^X
zR<SWC`)KBH_Fvi%5;>Pu`T8wwXVaUFbcWRJxF3Qp^lf<9w^v*WbN_}?WIVCIS8=UL
zi7H|8CCo-bl#zeZF|+~sHtwen$vnp<DTs+b39V5!gH9SZ8gT#Ec8%aV@U#rBMnZHg
zi}oYNy-?zfJ~^DZKFEj{HfIUf)mWmzs}UYgAa{XCz-%sJd*%9*c;MjN%U;TUkmyZ;
zt+rdu6~kr>z`y&!*HEy>f@1SF77uW@D~!^Op#E?D--dqZHR&De1u0Q))qPAfE}<l$
z;F|8-sn>e_R{cT+rSrC;NlUJY38m4OQKA0O2pXgy?#>105e38a!y~CR<L18~GduF1
zCmc)G*jgWNKUiJbZ^IuX??l@lT=`<^)9`ZC{9)1A@9|`CSs^c9V$c=%S`hTRBc0l5
zO^_X1Ki|DovioETZ=_5Mc5KM!ZjcKV_=xNy3Jo((6X0GcFdAfrq17~taK<_wW>15(
zE(qOYQFKl9jwpeIV<(~6$B~_7B;<M0q-!XfaJ`)pN`m0w8oMo?eM)({M+O)2RT`Rs
z%PO#~(oeQ&#g|?^i%+hqs><@`=?CAiqDpW0{#$*?*vpv>)d61uaz=|OvFolYQu}Jf
z|MwN8exF56S`o`no84X$zvK~Bxi58?>3>*we%FuA3IAe1V#+bD0qL(;)6h4S3oF8j
zzMWlnd~zmuC1!*66{Pur5@~zhTf#W}r&6t6FC3KU1dY*Oe*`@S1<EyEgRb0qc5x|6
zzSWXf(i)-lpTiXE2{Q-C6D#xQN&OrV9yvC4A?i;HWz$OP&^8$eKkrS3Eb(k{F0L_^
z@5zKLZE&m#dP*$We8+<gxy39Bq9f1LDvTB6{kiKS<r}$!h4b1M;t_&eAidP<#SkhE
z=9;(k!uuJI|M`%SNwPb=|J13Mhb&Fa-6%uS#1@+D1iz@vGSeir;e*dib#h<8$~0z<
zM&TO^0@wFCbX4pgB3<k3u6mtuEf#k2r+c3;rx^O~u&D7dn5dmCW~iP$0*={?+fB!C
z3g+gwZf23=yK)V4@l_2rc71jl?>7XmhTlDZJ+)%BZHtt0zSfKS|7+?>1Cq?zw$m~-
zIq8>gR#T{xM(#Szj3q8unJJNoYoel=QJJ~nifc_8m5I4Qi@TUQZfS~3sFe!=rAemc
zngWsGilO5Aow98H@$l>1=bU@HuIqjt%4}WL_~NfV`2mPDl{b8z$kAbI3No!YnV*~I
zAv2*pHcAh;pwi+1MGw@J?l0QF)ZDrwCAS#b{e~x}#x{sKlX>EH#3eR!8{Y<s9=BhI
zTke8?xADpO)}+-kIPYgK*0$~<7B`k33^O?Msob4vWQ)Q12w6>_D9%{)3H(*b?PF^W
z9weHl2ZAz>Q@EEUgn%lgkKSdfjs5zVdBt_O!sgaGS8LZ!dA=FSsZx)h_9+aHbtv^>
zO}k_m7X1_uHrXm(;*4c}FkR0dS|@_(NBxP8ntae6u3LSBH&l2cXB1!UjWc^|Q--Dm
zzG%Z4V|olD9#%f0`(N<#^!DOjk3T_;kV!FhGLK6s0kzS-v1E@y%wy_X^HMM7jP=#V
zU)?49Mz-Hv7DP`aFB44mez#K!F8>Xq2x41TG#Om@w^(j|Ml7NtAG#z2q#3~h{H&$p
zK2>5-tME8%GzeoqAPfs4!1aB@S9aKfm>b%+g`7VgP`eMM2wK*Stjxbq4PgZ(s@-f9
z4qB_dzrE)kKA!B{+?T`SbwK7+OkTaYCwgP5q@;!v(-VxNmC-(gDcd+!)8ThXy>3@R
zYM+zSN=DJ-KJ$X=z{<$VK)(yVo)e<M*mwXU$Cp&)rH++QQ**S`iIpdyHUrl@?o%Tj
z7w%S72i}9t1nJerGMC<qB9;WjKg#@wGRM`lS=OgSHscq1K$e4Te&xRJ@DRjurz-G<
z*M!WmT{>zj7XlYFY)I@JkxRG#=T5;LU>Rxt5!@OJZ~9sQqS7LqlePy*LLJugb||@N
ztAU$@E2Wgqf^1?)Jf~6g!fFZK^v9g-dVZK7*JON7m^B6SiDlt5YhB}$^d5g^zDq*h
zDTeoAI81?_hk_mVY^s=7axv>tZ8MYa=<HO~S8d5L0I}vEqFh*<PHlQ&5J<|U4k$J=
zgQb%nQ3vq8nqP8cjP8@Q18g||Ek&-k_P}}m=}2ObZzfZ>cPCC>@`H7Y?}_wn8eRuw
zE#^MuA`$Ewdekmss}h)%!pKbIfyLk8EY*z_Wy<KN$Nm+cmwsktSLCGGbtM_5>JuDt
zxpHa%RgN_cxhf1Gl2L+GK8?QkUUo1(VcRy{-D+-ahGBCz4in}Gt@%Oe$Y9BFCb`1f
zu+zDXJby`?MB0+RPr*eN(R-UlrX?K$G|vf6B&}OR$P{Xo+3!fHCezYA6mDV#Zz=LP
zrSl|ON}3ln4a(oGejf1-*ZbdEUV8890V>@QI>J8j-_(gFX?D9ALtO6fHD)bbi&|`S
z2-mR{hahAVv+>Seiq|?G-4JpwgzYzbZ0r)NxZ+|J2HLb{CMDT8k~Xb0xm!L&@9&H^
zP(A_W)FscC({LNicpUp;gr3v)X1Zn4QgP{5|A42Vtdg40PvgNog=JTKGR{v9%k2}e
z!5c2Zh>ykwY<x#j*nDf~3An`oviwn{?TwT}v_tBwGkA-l+XKBlH5=Go>^;cGYTKq8
z0xoGOB@f>&Zdx=z(*RJcV{&GsU}au+d7lZld-M0ByxbnMnjoW|EKeSaHq=rsdv<o@
zi0wN?-wt2t%g%rFDf5H-wxmwv9B0*$QhND*mIMg2Jy79~gz*FbAqDZzx3JuhX^VtG
z!Sq}Mm$!ceoTk=GMT~WWmlsQi5?4nxD#{kQWZS9mFC>coTE}NIqUHmlEwUn_p8v0i
zWwn8&cO0&(+WndbVUfdoBX#1K=E(d@O7`UBS&i(eQY}@lJshX52iKTFyBuD8@dCQ|
zON|#_kwbq}*)H0vKBI?%cfH}kk3ubUOfK7R%J9l6Ze0FvvEK0aDk}aY)nxJIVC%|~
zZ!}>#T5EZFaM4>jl32vjmhtg=&UKWaps0q|VvZ=(jG$qoc%0G1*wH=P(&?eJU9)lQ
zONwk}D$O)x%c+O_tNFYOX5h0;gvDAm7?uHK6n|mll~FaHkIsK}GP|JYMZ%JwKldr)
zC_mwslsxuL5y4(UxKg1IOe?v5N!>6_w}=UMGH`^S6K1_?*vJyUjj1A6Zf!9b4Z`0@
zVUNjas+vcV4vB+;Zb;~?h>KS$I@jSem9?BrR^l=zH}3p?eor(_S4qmb3wn+|Qxkv4
zDYi1c_?+u0#Z9cHnmGrmQAFGOpz=F4(J=oERQFR6%#Nf=PE`_Xtm9YjQAN7J(LUY1
z^Zo*)AN@JbjF6Z-{<9`4Bv+nao7+%Iha*lTDlWbA*~p?O@rSsXzy(oJ3mY5g)Qv1X
zltU&Tu8hJMVBqHnz9a8n4Rp0Z{}@2_03fK$RO0|1W9_Qyl@1Z^^Hwf$IX<dY{g6<G
z%!;^3p^4g~=ef&MncKfxB30Hhij4mH^yL1zRQuQH8^82MI$(^9TcCelViz=B4~0Mz
zuJW9(h-_~9Z#H)EgI_eT1IS5!XwAQi^wW1iH&uTPBnLN(BxK|Lp6uPa-E%c5<KEr`
zzeZE0O-&vHF*BARw#(8&1gX#A7xLEeXYPfDc1uQZ+V9@AVLmAsU;r6Y5oz}ky&e=T
zk1u&^k7@blwlZxl<utR3(8Z>JmpF*HkLro>*eP=0w9O2N;L6voh^)npDzy2`mCuYt
z%kb~^Tt4*8vnC6j3L=6GFS-;6EeK^JRlIlc!Y20aedf}MjG`CZ1=+Bg&WNb?T`{aW
znN(u;egCaSOK+bGwh{+@vK9ey_vl{Z{#YB!>xNs8p2a$ld(VhoSYPGK{T*_1o~W9>
zB4zot`E6G!y2b(lGBbcD+KZidcs3TN-n}q?BrBpau(*i1sW+TRlkq;0H5#86f7BW2
zj49f!R@UScTW?9StO!o89D^A4jy3)g&fbI|h7BH2?uj;>pJ-U}&zSp>w<lS#`74d~
zUoA9Zo2<MXl<6iWQ%g(#7b4!JyKB%%FKSdGt^ru+fd*rGqk^`bYjY5!f8eip(#)fe
zi;r5~&*z0GxS4nWXkAxc)kHjyT}Re>BLxGrf2?U7#L2+(uIU7xgGJS`vvHuTrkX{L
zpTr?xX1{g;oURXW;+cy-!S|wP#I~bnN90t`BVlY+q$+z8a$&dHgtxEu1jNbN$+?_X
zNv1QiEiLvS?>{_PSgc<5_zP%tj)b)tO%8h4-<2fWC7HJd?Lz0uV!p{xoaT>1ySW+;
zCz(zKknVEP<gTBeH?gTgi@D>)5N$2R<E)9#f1HduR0rr?Kp4A%E7%~f$;i@5cXDnp
z7#7io_|l<Y^4f{+<Ra>B_3EIr)rV`3J!eDoP>N<Oy)zkxomJwqj9Q=QNx0H3heO_G
zu-3BZbJWp}dNS27{Dg6uygRb~XgwcJxgP2U?%~nLTHZKctukmyMP)^$O*6};)}%7E
zDzVb?ruW<4w(?Ba4AWr$$Q=>vINS+dw!fMUE~4s^`BKGqqogJ;u7ZwIuFXoRXiOa`
zm{`=z-6b*D(BR^m@RE9=TvTE~fM_)CG45K&AIcH=VI&R(E;`&Jx6=0OB$uuqE4+@G
z|A?Xh0;GZxpWHkB%qAZ#kb|^uLb7lNV5RH$k9Vu-25|ba?<Sv1vfG*BhBtec+dK!+
zFi}DKSe2iR?X7K-#g^<RhO=SHsHFKzfCO+#)AIE8R1LRxNUll>%p40xXcEV@z2(cx
zE34>RTTq-Ra@E$6|KoTjasdbfD=LL{Th#WqZTL!WeYMSv_>Qvt_1ERHrWLVov?pRd
zcW|-9ZkVylg(l?>k&sjdf&*<T9cw^eOSujBKH-5|DYF{>q9Yu5X?M!bT%j~}-N&K2
zfc!fy4vM<RLXvDwn)nzm!t!K&g1`7*50@wGj(H|lBIohuH)WS<hEzHDM(g*8@OS{1
zVNW)!Oy)W-XQ2rYH6(76q%fdziv*%VG~SZJ4#TSf{!i{}NU4atzI_wMkVtbck#{22
zCsuALjA3>y+~tPIc=L9vJKu-&@UKT!1Xsw~jG7v}#D9hXN<yVbTc0At*vxdwZvB3i
zmk-^5e#K^<oDcGFmn5?OyjJcT7wdeI4uLei;Q_j5CuM>BKl8X$L0t@M?Bn*ULN1Ok
zZ0~Flj|Fn^8V*tG{6+kj%tL6DPV&2S)4z}b#2qLn-YaDySx3lZcO;3PA>!)li~<Nt
z0VeYSNBQ?l#@~<M3B5X_XLaK?R);SKMPl{}kh}V3MTjrhDK$4QcNCyZ%E8tS$=&WH
z&q}Z0`&36<F%!^!bKoI}p7f)J>;*{AdC8sa#5RKu=#^YJj-s2>MhvQ5n$BE~(JlOL
zom7MJYGHYKqF-t0gKw_3?qF2gU{y^B{XyRuuKS%nA6@RyQWKvyB#u2%U_|+QPct)M
zM8i(;XSOyfBP?Q$KIL@{$xis{g0wCGo#yC3L-h1B48a}c!_bN`EIqU<kBW&;VXU0a
zoS#eh8a#C;0wOaEQe9Ysq7o|%vZ0w--@DBKCsC8UuY-@vsZBt=kCt0*10i!z6<|vX
zXZwj<c&(NOvgd@8b59{;6Jpal24JQQ3XgSEluSg;do<iy_uZ(upML*ZNlMW#Mt{EF
zxaS&6TSKu-S(l5Cq-XV=;rXY1@w7WeN(zEJevK<uv2nc1qX{hm?VS-`C=U1X7fp(p
z*=?Lr4I5B7IubBtgKrMk#<IP=YXJ_`F6vWq*RWf@4Y1EBzL6f94Ao2vz-Y9=I-LQc
z(E->fTXF(<X3Ln%Iz9jzEP5Xvw9erLRX8BZKNZ_2IjA%0FW%Ae1WFs1;Sco}vx?<Z
zXOI0r7fen*(p_3bctd|iehTW(JBP?u+siuwH5e$;s->Nl$f+a%lt@`tx6Ji_vbp?`
z09U2@Qk4(x1xD9X2gk!i&LCpIo_P-#=&YQE@Rd^7b$Hb`Y_!cU_mYn?DLsoGpjX|%
z1QwsmTI?i@&Zx88%z$Q#*7tY=OBHF3WI#Zrbe2%|3Qmc314I%4?Lm#Zxx~~}mBHdT
z=^Vl1E5VH@pT->_l!+#IiTXbfd1Cyxc;177T98l7X(C8?w=%h7ZT9P!?)&<V4ikl~
zQ@@{7)K$woMe(>8DG5uJlrzGNUF=KQulz}=+%YlJ4UrXbBR<(<yBae6XnOLwo70in
zXq}Zwof}icW~QmOT@;`|mBre!y$+5;m(-m<3D1Z+I8QlGv#%o2bV{X|@YB0J&d$U`
z_)GR(iHG)<c>7Y{I`PYs>G#zS)-A?fc2A4A9k@8&>x)vF?frgjfOcmet4HSRV_7%w
zhKZij-bpFLka<^yzbrR0DGURUPn!g?r&4xWlkZUKivN`50_Wf|9*iqBzw%#x;a^n{
zp~o_$N7(-~?MzC(zfawzEBWzmrFMlc0v;f|s|(;Lt7<ENJp1@d?zPVRt3CTJyCZU#
z=;ikE<g!7AOiCpD8RjNHF3c%!c=dT$86Zabf|%Q1@bJXSlwm1J3InkoK%s)@eHfhX
zK`HKkQ7+vA$1JEF5gp+#GQyUQWqsAqeKF(mwQr%zxuYxDSRXX$HW)iaBAdUt>do+u
zLqKcw5O{O~zBYpAc*|#S#Qjc0zMMC$H$c7AA)PCdI2ELQYL_@xYFTc2VcQ;V?Pi)5
z0ngO+VllPtNOfs0KC0a&yJTLL?nu!UvZvflSk$3Gg=%xb^qZD9^<&u(8Vipncb>1M
zMBvlb4^N-L8$00kVv24%Y$^F9j?c49f%4Pyu(aUxs#Jg#0Sb4Y5>>xuaUuAYoLc6m
z4(YX%DOm08rRcw{h;i1KI@i65*BU7QCy`TSe5`E}RON1DHR<slhn;L;RgY(GzLQ&`
z-p&oHCG0wxV8YdRP-HNzj7y&at<9rIZGU|7zP!4ew2POG_LQm_RssqemMT6=B#Woa
zWL^sY5xMxoyBYW9o%ZLTRJ!vmlA6G$)pdnTD5{Md7E2;0&=Qh@Z(g_}>d8)}0>MPy
z!tDWPCq!L)n<fRvFUd<kL8VU=M+)8&J3~me6{x?=;}XLWs=blNJ|}%C@UH3JU9-#Y
z_-Zc^YJ0Ij%UFJ_74}&H9x}HWQ0#IlzLzuQCoe(0J|gOm7v?VHlxXP=(+eV8y?qh^
z;E2ChV=R9lL03td!e}I_MtWZn5sV2B?-&Jufp8#!yD&lp;a8`6Cy*RyqcaRWOKCNW
zQ;DLIva@aect^z;W?dzlmZI<pNgcIyK1hqaU3m&)IdHuB1gD;w%9|C>*HdSc#>eoz
zqONRj@(e6uv+xy?5zD=x9zu$`t5S5lW9(r%PLR3MVm0^AFKZ^9TBXmW++qXhkYxYy
z&Xu|N(3%3HZk(m0+(J@*m|)H20ZEt|Wh{+rLAYoQGxm?9Z!m6{=-NF!IXhKQrj1g}
zFoCgjy=!t@x|2cm{ikaG)@Qtm2WhP7i*JqgqzOFWvmRF}yfx1<B!!?FI?H}%Gbn=q
zkqfbj8IB0YI3oZB0g(El?IR%-6Q=Xnr14~KNX01)FP2wo3y8^02O?f@%4IX7I`dtx
z>d!Aud9x4xha2}<MIxjY-!H`tE@P#moDUoQnri-&=na2zIm|6hDNe}vkjw4L5!qJ1
zj$iPNebICp6Gq**?8N2GC^pIS@0wzzQ@n1`ca;FyPAxVgxe{GFu;WfLT*oGAv}jjv
z@N-P5JN_&&0CUnI6PXr)@8qAA{9ahB>fEbVK6Q{9<OozL$;0r`*>ui=-Hw#*q%K;w
zc>Xv5IS~fW4Ne#=jrRJ^%Pw}TZTn>)aa*m-r0DA$;tvP)zo0ExX?RryAi}ZFxI6Z>
zd!^BDS6&w~TCa#M4bS*l6K1Y3TwZtJk?ybG&!4qg$Q$PEjGizb$c0$Pjj5GaP}^(-
z=u=#N8-@K&5*+TM_C~+<SbL@_nwAE|5rK@()Ssnyk`_@B7$^y-@uh$%%3$mDICY1c
zavDWIZ_92t1@6*=pIOeShb8nJt&zz`zKyWR_X!)<7HxER_4*$^@M}_u&<`Y6?vqe(
zJ!+{0qtH?2VFJCtD7<yLN9-&FJ4{g9TFSVpsXLTfqUu(kAf#5Z56G*wA2*LnKCjW;
zq;n-8KH2S0G}*go0}D_$pS*5GutXjF_QAc#O-RGoU;uD-&7iWqL6ZV9GQu_XjJ9Vw
z-!iF->_Br)F*2GF>o1PBH!_vxJl6PCspMw~%hAuju6WLUZfE1t<F51vPY<~4SJYNu
z*#8ZAT|U+6B(l3~B(E&1B4|;mCeHS6$vmH%DO^8G%ADN7<qa*wvTx#@O*0JtW@p#t
z)h6!%lrje>bn%>u{@^eCBjo~N;2FQbi>36Cy4c*U+eT9WMFmL)D0lE5-`Cwq<AYj{
zB&UMxu^4zRFawajw*iq=K2jXy_!zXtXDG!bCG}M7v06JIDOd(3jF^bJ*OGwL+AiyD
zPGywirO85Eng_q8nVqoVrR?~gYt39Uo_io2f<<AK6uGAfq2nl<oqUnUwaZFjEua*N
zY^rlQnNf{?mu(iG1i}fBBUbuxu#uPM`Y@4^lBUGc2Zc)0Z%^q`#Aj#q&~Wv@=b`~l
zTQ(K|g4LKF{mL8HQ`r|qPBSMcW;{!a-rK<ywDdFgM<YR3u$Kw~Wv_t+MdhC#tiF_P
z!fFk1`HFRc(M>WJqRjvpk;;+jvrmk`=N3!%%K8F1t^chr{EGid=Ya$yP_WqhFo=Zp
zS1Rn{E&-;<&Xh}tPuQu4a@F_5sPJ(z{k1XdX`^3L*auwDsP;vLlaXSZ;j=i+^T_(V
zWT0yYQb5*4QQa}7E5^w%Q0Znk!i()SLNBH#S4w*-bsa<y+x%}w+%uF(=Ux!1jqFR;
zwHl0Ecsa5$>do4Hev<mb3HPrE8YQvN+E!N5_NwPEHBMdAWOIcH+6O?v2FL}ORE9LG
z>Hi23k*9Qtvk7j|x7Z86QqqYB#FNv=ZtRpevfF*UhEp$ozjGv|3|^DRlhaT|VCx4)
zDsCG+`g@O-4n@q@ceBzc^AuE-geu$7pl+pPE1P9jC@meDUeT!;_%N`NZu<+VdvD35
zlI>1F0x3GtHNna|wXM7@2mxa++@492$<~jFNQw9ai6qaL9s`26OulQdo9~)-MoO2o
zT0ux<j3h)a%yMssgkJDlnVj-}krr70!01-8pw9J}&(zdh1-9{i=qN;;A+Bvr1WkNY
z>j-qjhC49-u53Jk;UQ@EnLDnnHLms2g`Bt(_IC*9S+1A|bO|}|fgLH%HAZ67Tc-RS
zD(AiCUdCaE-`8aoyNU|0G0jtJLfv*Ib|;OS)=@qW=%Hm&^<l+>v5`s~7H7Y5Fs27q
zX>!eqX^^3;Bp?mgL{V%BZj{+m5(`XH<HAUS<jflIDwVl$RVyKt`R;cgMI8;<?iA-^
zMlSZQC`ia5A2)4vMw1zTNYWcFtILREQP|g=i#lY-<3^ria)p^MJNHyt5F98;0+=Vp
zlpm`EBoS8`f>Vj2K!B@GTrX<jxtur+=qKn&862~oeRHaVV+}nCkEyJxjFVF<bIO9;
z9E1w%srG_7tc50GwxV03z5hgyS+gtW2tQ4g?R`M9_@+H%Fsg}ic<n;X4Sd59CXIh<
z0ida3sYZK{f@byeIaGBMf2jOAGQ|kqLpYy1tI*0qG@XW>2_2l9Dq`2jr2Fdr8U1rY
zYm<ibb8j<3hG~fGI(}Hat4=!L?AqtwZ=>;cs{u5Jh~IiG90?#Dra2`mG`}K?ZhIsC
z&|Yt#0sriJbCx-e1ypLo&n&NZnuz;hmxM5xB$;jg^GeH$jW3m!*c;ZC>bm*c?hfW2
zDKH|XGK4JKa*^L}l25@{E5*`36Lft+@01%sl2c7s|DmveutRp|T?W8Gg7`b9I+;fe
zh^^BQr<&&HaiHgmU&S}l08DlF=*Vww>5dh-#rp*n_xjFDcyg*r3I+&+w*ZoV9Mhux
z=Rn~GRxi5vs1u-B9HBxuHAvuS?u^n&6a`QAT%~f`8<swN?r5}FTk8ARZC|Gmq$Xn*
zr+il$Hqx7L!#fsfS)sf3$X?~fg6u<YtoMQYpNwTv7F0fj?Lu0)wtimfE(ba{pKdoT
zyQuMBl2}P1x{gZq??9S6SatZFwjC(TaAuO<((?L1JyNzkf43S4-xnmOM@|FGC134I
zIr5H66E}cLm+{07^Be(8OZC=22o`S$#$xh?hig}^yQ!=#a`N5xqmMO~(;U1N3;1Ls
z`QH=V48hx3%lt)iF0%hzNNxF081$jA+l2OTu_IeVY*4G!Qg2rP=t4!Q>L7L_c)3D6
zx!w=JL~S?6e@n!Z!-xw51~4G6JF;yPaRxI`yMbxQc1G->Jg?DKDT8{O_QX_eZF%52
z7*m-Jw1hH$F;E=W8CQxv!B;ZZUv2Uq<@=MOZ$S7H%}GY%(!!2$fKB!ua8b)>36hG?
z+^lGF@IF2AmlV}p-qZ41!d_eR@RuH%?6LC;S!nS36pA771NnoFG*bZO*FA^`sU5?~
zR2?Q{fLs|iT=J+>FsbfrP9$LloO=u_?VGzq`;mK?+9(j6M}RxA|EZvT;p!3PmR&U4
z(WV~g$Xl39g;f%XJex#HB7xA%eQmycacw&<w_lveGg*DSy_^s|-M!2mT&~}TLzJ46
zJG;5ND0;(%>>-;eyN?kHy5_%9$c)4(eU$4byw=Z@xQ?Z<nMtF_`u=6|=aZCO3z)Gf
z#?LIlaBnuR;$I*%DTD}mWhiG1`Z7-W$iT&>U_uB;7Bi7X^b$tTj8nFK()gpQ8?S)(
z7r3yT)GMtq#|KnB&$sNwJXS+8!5k6RL07V`u(}VFE@SOK3ZzMm^3@9IiRPb<c5q!y
zMo{PdYYCQpCgM3LnzXCzFRNIlG$QUuTEw9`NGr=~;|p6acGKywxxao5KWhcZlkIw{
z#B&bV<0vPC$`^<)P_K+G`(DKa{=(hloif^={=o}Gaa+t|wqHF&%Ns>!@;$bugph;D
z)#QWVY-YFgJBR5mz?B`yAOKefX0mcbP8^z+3Ik3x)&0VDRam8SEB%1dzv~fyN2%Zo
zfF0E~@gsrE$$e|x-pl`{rSZ6xIp(`lRt0>VC!wEmG0n*JRE8{RDopgp(I_9PQPsa7
za0oO8%>n-<-+e&T+5R_uWnP9dcuK7g0X-Pp{DPxJ{D8R!`#V7NVjET_h*Vb-5u|Zm
zox?8+sSfgT{-mw&=nI!jQ1Num<<e3jl)5I)Hc~#rix;kgcgV<hEffG*4~?XMR&vLb
z@o>HkK9jwSB>@1Yh+b9(C;qrDr*3^qe6F?8n66Z%zrcpq#9!Dp8Hxcyz)TgsTDPy+
z^8_=NwDGN!{G<FG>eXn#kt&y-N-Z-LvT+NNHfhF?-2QG=L#>c4FM-bNx*5sw0}#gs
zQj*E2@)z6KkaEA9u__-(AB$^_vHh2!b<NM}sl~54ID2~yk<O{_L$y8gD#(lQ?GvV9
zO79w_N8sw?xRSc-_L1YQ925sNa|BGGjhM90FS=#ay#`V4TI$&~1L&TsxSrJF;YD`z
zJaNZ7O~0n)*>sJots;>%CPK+l5*gK$u$mY2FM|m~_e5HyJ5#|x)X-Q!F~lYY%#v*<
zrUJMK50Hx?v#p5}_?tMCKSgC64;HPBZ|+MxNBSEGZ&7echj+PtQL$>~Q3FiRd9Sip
z$2rHLDA(8^B~IkllnIpuIAS813)gWM87Lpr@Spw?QLH&5=Le%+N6(wA>E5K?2YrL>
z-fud$OOt`KCe-{JQB_(jAU>bI!7g0rS$KJTV$gV6UsqFAGxjI~|8F^%YMJiK25Jla
zGk4i|QM#?u2Mcx@r!VmXjMLW<MEXeWR3C<@Ib)z`I4m$F@OI$Cs@J$90MFvo7uy4}
zS_9W8>-h%eu(w1wEOLNe1S4M{BhL(h#RY^VMn7EJPvC&x<uFna6IzJ9;;X%ameUAT
z1S1S$_iVWS!FrxEqhyj$DyLelMwp|!7aQ$Sy4qbC#J`;a$gB=nr+ObI14hUIScMbs
z;|`b&#tP**{_&jHLJV=wn~y8cg8+2bRrHiE7<|`L6w|7Wb|ntGax{3UF5{ZBUg>$3
z8rC9L1IA0x;`S0JHm}W`Nlm7nL3~BvE-Ph4i>+4Bo36GOHES<=Q<@HkmJ**%*|{v*
zT>!;zzT%$7_BY|J{Dv8r?M3P5q>qq$Z*qSI3%ZU&dEU)+HsCxfbpFtpYJtHDEYekC
zQ<B)arZ;V4!ja|_G*F%2&->Co-M2f#3kQ>8MO4SwN(3oLBf(co;T*{_GSaZ!H7HT?
z7&LkfFH9`gcJIB|eQkj&K&jgYzUH@_-T&e$v6#qa)w4o2tEN5)-nG{Xi&6k8bb5|a
z)vJiAK1kP5&PnivxGn0jsh^VY_qL?$gl}R!G9&q%ds!!^7NUnHg>`1W%OBk>PYm8N
z256fS?GjuW9(zItIE#aZG}D>c%X@pmpc!fb{}h%!Pi5cKTU*qmz49td{C?9#Wjh~t
z{J)eY3kj7tG!+VzkA6mU#au^1V10>-+eW4s|LG2sG7T+6o=^w+)cdeA<JSqKO26i3
z9oqQ7X-I$H0ZnPvf%`BUe4zIcjNlJLFblhP>=B((m3YP@+{;s*bAsc<sT**0?kOnN
znlLSAa~hIKOwjle8yPc?*TxT|8l1j6?&bKS{0C;0T~NpIpkHq1lXeY~h_jZU3<Lz2
zs&20sttAZj1XGUHLppRLuoZQ8Q_q1FQ3)u|jIw7u%qfx%2CNl!6t5Q6ThGg-*7GAR
zoRMDYzDLU);#07vQtV0bd~E*P+Twi^$$_H~RktYLCMmkY+6Fy2^$p9FGHSJgXxCUa
z^@1Jms6FHNS&XlF+c#n`s-Od#wQd`9LqW|5s2F*~wjp#y3dEZ2N#x(gM~*X-6l~f^
z)3DKG7HjGV$6+WP%+q*|O*#SE+kR)tp`?Q{V@YR1>;{O#1;x4B!FW#u#dhJho{OT%
z*&kUj;>|Zq=}Fvgr<l6Y<D1=D+N{fBhARg8;Mbrq<%;czc?}^W$vE{`E%RS8O&MxG
zfhg0Jjm22c7$Cif2gzWtE8qguvq<iMZvf9!Una?_ez?Z(b3N~IgiI3F#8&#TxW9U^
z(r*pNmYUP}RwQ1wnt=gWM<zRT@*wo2V!skj9YdC9VH{{F+U6<h=a4kz^DqYPh(yg(
zP(h_LIiOHs!0cRcgmhD{Zh=D5Y~-BZW~E@}5{;d|GTRSFa#m)_8dZ+1o%s{l9p7r_
zxqW~qQcSDg*9azoYNEW5!Oz767%_*EqLq(!aKu{-?NFQD3*Hc8T3#Jthcyol_3}(M
z6*j~$dmFB?89Z-t@Ip$+V_s8qMyAZl;BX-4wEBNc%9g&rGPwF2y0k1<;YUXv@}H$E
zA`v0t&_nv;`?S4vq1txK&w>wy*4wwHCW!Sd1`xXyd!LHyb&A;e=(MPru|pBHYfOvO
zt%s}iiPS(H?Q~kuFQoHHHjze1gTm&P{`bo>b=K74rH{K>nS#R@xhreqz*2tezZo3L
z25=MtdK5?}{=z=v2Xcde%ol`!+KsHGkiq3U{{aPRz75t&(Sx{e^*^~|hq>-q5D;Ox
z1wu2v`@;xGT7Fr(>W|S(DaCK&<Wd8F3d7_H5Y1{v7-*rbt_s2@7k8*<{!AHLj`E9o
zl=B~~M8|k(RUmP>n!S8G<Ic>kHFpU8<(p@&{Uasiy8W9WKmNPC)9l-x{P^#^+0qKH
Whi2NcdqdNge-UP=W2Hw=|MUO9fC+B^

literal 0
HcmV?d00001

diff --git a/studio/frontend/public/whitelogo-c.png b/studio/frontend/public/whitelogo-c.png
new file mode 100644
index 0000000000000000000000000000000000000000..ee1595509211068f9d8bde0da37c841e15bcd4b3
GIT binary patch
literal 139810
zcmZ^~2UJtt)&)utP`VAIiw&h0As|gag--;O4k1(tNFX6ZdQ*|2G|NW~O=;3WN)kHK
z1VRy{CbS?4y+{*6d53%N`_(tb3j-a5F~~k=uRZ5nbFTC1p}sa7GcPj@4Gr5rI(H0d
zXqZqmG<5PO8G(23<UUp4zl#n!hI%wKe%EMdp1q`@*#q8sHb+C_DNRGOU`s=DE0KnV
z+dZZFkqYn&MtfcDJ2XefUm10|?}2wtx$9T}f0E_w@i#3^a>@nZO{S;+=-p*nIKj@!
zAfrpsF9hDAMR@2N)AoM!P6qxw%|CZEjD5zI#=TRlJnN3uhbeehU4=T4EHfQC;SXp1
zRs}RPibqbDG!VpxB|pt=S(o^nj}z57%W(em!y7q${vsd3PV_Hsk0-6J#>$ls^dzN@
zCykH(4Ej0pcn6=fs^kRqj$iUuPmfiiaK4?G+0;hFyeMdSG@Vi)UHAnrDlLjHJoAZ0
zbbV6j1S@?3gG+ZI^QRML^v3jqiGx#C^3cKCl^d(?YwMV;PDET8+M-EmreV3&5Mv|5
z-!mzJgRwfJ`WP$dB&L#6GHIPpBqNJYJA0X*#oi09fqBK;qpTW*qm%`{)9NV>zd*V5
z;J7GOVlfI%6<D9z)>eqvx=(?b#kc)oEp;xbosAope!u!na-_|&b`;%a5sWjWX1wc3
z@?bJ$ke8L#<t#!=ChW!F1!Ge9=gJF}uXfJbNtzs@P3(-P!?7KN{DVw};N|eTE9Wkd
zyxi3jf}#}rQUXvtl{9)&$pSZNCD@bkZ&`yD1-^&C^p}RGQVLKpkg0_flxNGhK6mV@
z2hvOk5xML<T8(Odz0ojY_0<6ZU7oHSV9AX5{o15sM&0%6%Znu%Gnmd>I>b$h=HFv2
zh%mk<R{Q#OlODP%=8G#o)PeOLP4k)H*b{uuW#}$MHx!*2D{zvSl!4HKFRHY#HE5HI
zBv%7vHHmKytfm4XZO{bbxn_<4j8Z$7I{^tVK*0+_iWIBP#v&^7Uq^o%)8Hd4@L@3W
zHx*hP))h$W3f$h^W#styApy^#Za$2HIu&tL#I1Sy_iNts9QpO;Gb5pc?vo@q|A;xh
z%`Xi!6nU@B;-+j*3g05B%uaHJak{YW-TA5^Ydf@a_=7CD_N4*gVivVvZlB;roB{=7
z18)<z7$>k6pDu?+-s4-Vl~BYqDPcxi$M5m;N!X(s1N%%TE2f*8)#}A22Xo9;_QJnM
z=Kj825l^K;ahI-o{LE~~qYaxPUbKgonwdP}dv!0ojTfyAz9QR$Z=-c!1+ia1Kxbsy
zo}1ienypMV)_FtN;vAmLd=uZft3~{-EeSK~xn;NGEoqRr<Q|XvrnG>BU&gZalzp34
zpj(vSg$VH&*Oe8OJ<%;J%afcKGc~{uwkCrY9N6?cW7Cgja{^-T$<CQHiv6y%OW9dK
zrRj(iVUg0%f$l^&bO7)A+IPZO>P#wg)YCHaXblEV2GQ<M8FhM}w|pX40<(FCXVvwc
z-z+kp3U|0I*0*pbc*dgOZXM&sCBBpyH}2Sm&~?G}U$+V(>dR-cLBlGZ)5E;Yf=NrA
zkfU>3bbo#}Bp3MX+aiUvPu@(o*()%&U+>+x=XV<MScQvk(Rbob={n;K{sM~vZG*hW
zJ#+MH>Nye_mbXN=`nEdag(fit_0RYU-3r=Q5-!qHx0*tjE7eoP2pcNTh0OJ7psmE~
zyIaQ<hKH(^hKSYK$mO<ZgOtOP=W#VI-Qs`Tf!2qKVrQm)X(RF*ao3S%8sgTlG*FWY
z+Aa+=UOEp|VC&7{jgkw_DRP>6ewUWRP#6wPF&pFLenCH&5_q3zh}#f`j|EwPxF!v8
znWo)MFvf7K5~k&hNv+Jqm`AJpi^vrLCItPeU)(p4gG3QRzW(5loTEqJor&%)D%)!t
z8l0>tmY@6ffWznf%In`%qoW2KvVMLHDeOKzC&{5Y3DgmFitG#A5x1B}wwcnPHvBz4
zg_G|-xIyor`F=1io1{#raDz|{mnfubl2;1-Fnykosi95k<Kl!G_&t@{v+dU?M6Phc
z(%p~fI*mbe;ib<Lwh$t2r@LGpe7KJHmx6fQ1^3N3K&|)e&Mp196opWntgXPt*Ag)a
zeF^E46vi3i$&E`!D!t^SQ7bvLUE7`JdsCPP+xUpv`4#=QSdk&yqWZzGJmR1Wk=TfN
z<6<>DeTXYc4k`J1_3*y&O4W1N3CyU{geP*sOY$(x%&79U&;G#>giZMg3>G1;Ff3nt
z5H!AcwEjMRX~0Yu`ujwm?f{^XF(oqrLP;7uA%|d>?ovV9lcVL!WylxY$oo47oFEf0
zEbpO0O^<@BBz)Vgbq*m6_wH@gvjN%Y&VCU^(r@?HYq>N)DMSS(@H0KG^{Pbf(I4)M
z?zNOb<NCqsO!(IO@L<2dzVbH$3pvF`QbT`K3w`t#tNEQ6OJwuKkkeCPr=7Br^rU;y
zX`dQX9Y)UCCh`n2o}Z7ynCBM{T{nnenN5o1drF*82KHsVF}`yeg=As&P{!!zdvH55
z@TQ{BsVG=VWJ%kGbu@cXAfeTP4+3`e2L}!e?IlbQ5BJoMJ}FVx*_gf0|G6KmtjwtF
z$y+X?Ta}HEqb$-!S>@1BvYvuNN>cO*peV)~H?#=&u&*7gMLRFC!r<4(8?v~L!%9y2
zVo93BDC$$S7CjcIa&6iY^VBoXu@n^gxgL8J9UroGCIvieVp2PAUnXCZ78s~?(1nmM
zrp`jjeyE;(UObz^%j|Xc&i}pk<U(=%2!7npu=^<Z%@~=+3n3k3-*UWbv6I5t^1Q~A
zx{<EkhNoS<<l({Tkzk#x@DZ8&T%>F44_zbB_3(Qa7eAY^Duzo|7_+m*%5BULH%`@0
z=t&9&k1OpkwxO*(2MNZ;0aKO)g-ZBnecoZ{Ab4`Gw?G!Ul$^xvbHh)A`*&%~o1G)y
zPxo$ly{L+cYLn^RaFxPy`J{E9XuF{B1%HpG64;lHiwepkSFg0N6}NOx6!;vtNy5#*
zq7=sZCSllK@-CS~UP(|1ox1jo0pH5z+Q5%h4FrE(f2KegnYR^*al0VI8C!DAUh4F#
z+L2oL0Kv1jw!UtDd555Iu)VrI*Vn{F7_?Q=&5p_B_WHfBw?Vh<iZ}lWFR4{~;Q-?B
z+BmAp^O@l9br2p|Vbqc(T*<Ax-IHI-(2(sR+M;)mv}-+y$RkRikY|O?4H)7|IYFp&
zdre{tnN<Dk>?u{asXSa*7(OMRE9J;e8POruor$eUsOy{qn;Pv3_7p|R-)z93b~l7-
zsarM`m8)r^VM;R|e-+Rs1Sp_hrGQ<}W>zSl*BkEj_+(%IXMB#*UrU>AUy)0xtl$I%
zNJ@U<#eA2DL+z~A$9IO}TRkOFx=xm&B0IexZp^dz&eWnqH-_P<4BC-rGrkn_(}*X&
zh0{2b)tmM8IeBi>Vjsd#YVpea@T*~vpKwC)L$m*@n>$$?hKXW<U#7*1)9LWiiFPC6
zy{^?s-OY)Jhi{`?jh5fbfkmBqTf5!zQugEyR70`CtDa^Ny6tu@JwtD}@Tqw%%TImy
z&yaHBn!~IS)N}^p(@w?qiWD!L&(is+!LM(c5EgfTESsYi!RvcD7;B}4{;m0TU33B3
z)5Y)GpH(9UR85KJ$m>NhJWjaLbz{l4x0ap_?Am60@tt5F)O6$;=G);TKkpw=N;r#k
zSAz=U6#?Pm2FIzMRyGz7We_*0^x!i~@JkOZVkwNxr6I-3_~?Ckhmz?u@qG_Y8)CFE
zuw~G22;UfsjxOS>s&b{yd{%t?vO53!`QIfn?{K^d6tIf0BnJeZ&>rOrRTyu+#`ECI
zr%>aIgi#zeFOaZ<_cK5t2h<m0aM)A}424Vz3&r9@E({ss-r%sI0&>?C*>U<5cdSH<
zZ~7;Uq-@2u-pB2CFnD0qdZo9G@8Xx$rTZ(%i3Moa=0e-Qb^{>6x-TEZO86efoK^&N
z;JMJg;mM;Kf(r8G`3iM`9=G4{DROh4%xhcLLm|7)lb|bRdz_#H!BjRRD#o<MiYQrW
zEP1g-qREP=Lg6kg=^7InQ-3Q^p+vF1w31fZ`MTXfdC>+Qgj%jP4;<L`Jlc({-X1@Y
z670Ig5L*2Qz}7{5vAR*Cm0*KGJMlAnMI?2D%Ku@^WaCTD|12TilFcF!*CjyGv`LCw
zL@f2hcM61KcYB4g*&aThl6SBW=R9H@4!dh9N--scL4w<492C72iChK6=8Gm~ss=6b
zpt`-4PK^3~5!*bafD8ZUC!>K+j?^-<!tlk+Ga7ZfdT+Se$PwE4FOO-B6((WqgyDN?
z|3K%3;S0U^o$rQ92bU8(9fVEFJ0h>UMWB7pWG6;~AL!0Dnsy6I#K~>5or*D{GD4gl
z-bIq%pS@)>%-{-YESqR08CcxiL>A-)jLTz+-Nq}HQ@wi*$C&1w61kS3f{!Mz{5~aj
zI`09<k*H;A<)jG0^P{?#F3Kr?RYAi>E`&_quC!C4EwV{PTi6l66p~l))#3(iB|sMu
ze!j%ws9vA5J(`5!<am9Ixc(&XxiF|uIdJn>YI^nob?Zyk)FKj5t~6Cjy+hMPvBaW`
ztJ~(y10_XnzSM)TA&#mFd6jM1G6`kzo#Kwgeh|y=VttziaB%f_|LpKxHZm`s5$X%w
zzyn}Q=J|e2%xwTHI-mi-y=$J8P6S_t&1zC4+P^_R9LuwFiUWBN_yT>0QUzjfNGpx+
zRFVr~$y1XA%J!D^FDoLQrD(YNwnl83&6(P7xL;z4b+#b|W_Go+$zZc#DSvo(7{ZD`
zw=CzW+8-G~5MmA-;QYO+`5OSH5yT-#NM-A|PKauEInX%S$f0Q!`fU>d^jvxC3)2<`
z_{XY#$7*fYJDV5N7vDJ#i!TOh$I-R#);MsKeWIuOKsNI5<GoI^OBa${JaN_^qcp<q
z%4pq=Yq!FyPE<pi7pk_vRUlPNHL^YRDB`LngvH$SL-8NdjRQDN<D@sU!+-`kRhml~
z)s#bXJdhE8H$kVAoBbJ}nWb+M5^-mR{{5Kkari%cFLfar_O`lBn!dncdBBl{ZEUc-
zlmkV_;-p*#&->+(H71JI<;k69ukKk#xe<2p+k7j5{uBWYCFk^9m`3;-M<(LRALTr*
zVORVwzW@BNsYS}Su<q6w9XaJMcqm_#v^aC#8-??IIc5~#Z9VuuM5u+3SW9<}K~L2s
zuxO2BAv)OFb;U0!4x0qF;vPE1f~4;)V?iO`2a_9m0^+bu&pqlnLA7T<8<x%anR2nK
zG|2+4S68gP+Cs>_Vq*&Mz?#`;Pl7_=%$VQTlb+}2{*aQJG@&@dGr9>7<#GJBy!qp_
zWjkmjYfcqL?;0pW;z~>Rr&jr9UF96$0E24)WxaC9U^>Hffjz-R5-uE9X7|Bm?v{WR
z!X;hv#f-j{lZd6LxA7}Nz2S$Z5*CP57z$xqd#A<!^wRuDyg+Y}GKCNJo|<TJ1*Y^D
z$czJo{B*QQthh*_T^fbIS+3Px{vPjoI>u5L?Filjz*c==a*+VQymlTSmw<Y)d`{~^
zqJaRCcHN4oDQbCeKsP2xIfV=vn*XFJU%m%2FmdhiWVC?+cZRZ02b{tguY6w6bwFvX
z>T`H#iG4Qe9cCo0oSH)P;6yAZ4{h}p(J%~1W$giK097!<?D>bPs!jkz=DX~!5&<Xo
z47>2eQERAzKq5RozkI8}nI$C>{p+%`B-~|kTeW;*5uoO}%rnE2{qymil8;kMsR@cL
z-Pr7@_)ZP8@1yaZ45lPTekn@w#!Kw;VVj4#vl;D5m@}ZsnaOO<VaJkW>MzWsGVUAV
z6OokS&sPpAl-)a*CV}wXU3O9k*bc6uHNSF3SkR!#nqBhq5Qz8pfqN_YC{c_-*{8J<
zDlOP99#W^;t9(YW-!?+1nvBR}Z8e_R2JksK5}b^~KK*Wp>yySlb1%<~@65{p?aTBC
z@iQ%<xw05ldr~=&azg)>+Ny5++6CWa3M{S+c3#GmT;>XHOj7-k47PslHdbG}iyH)c
z?N34^UinrF3V$f3gZ#0Sp2&FWiLBK%{>VmsnHh6A^b={i!cK9ymQdr{Xf<x`JwS_(
zL!t>=Kr4N`H%~wEf?U<y_L7^5!sHM)fV$((C#%A>2eGMXk$eFwfY{NqkrHp|Cgz5C
z_QyhQ%fn@a;XSCgd$wmzue#dinJLfRMT|^^UvdaP@|+8g4eD0n2FWqq(HYvzk#L0O
z-2RhYQN~GPzE{bl9vJ|T*cHmP*r1jle^eEqkirGo9?t<vU!Al8DjvRQ3DAe<m!5(y
ziMVD_=Yd1kJm+q<@C}TyWUCc1%2Ufyl*#Oftj0CnRMurFb-<pQ9*M40+Q5cZ8Tag8
z-ru+gU{^i$TEiS^oq`n4&HuUnD`&Y^UiDxPO$c_^y?om~4em#EthDC(ee>q8gyPP)
z%>RUi^LoAHp(Kt@GtM+!iq~<IvlPWJRdmD-gbaPJuR}^ovK9h-Ynl^Q$Y-SiP6Nu#
zyFEp^I+GLgW?^!FnCFZO)5}_=_9bK8*(pwpoJ5>5H;G)lthxPwKcAr&tabAzD5@q|
zFo68ZZE|;u{p+cx@4X)_I{seCTe=(gIDXv(8)e#&;Zl0MYp@&_A1en?@B$<{b13>p
z#wadpnfn|mC)jv}llyf+8-N&W%J<AaCGBT`0Vt@_l%G)p%4vYU=S3_*Bh*D3xR08|
z4Aklh4m%oaWUf0~{1U6wtMLjbg7%)q)^XrSKH-y{lNtX0`SVtEr4)R3>ByDtWMXcC
z=%4jPcO4(+rE6vd4socm(@q+_0Tq|Iker{gY^m$m-O5&DFf9yk^OyX!Ut<`JZpi-J
zB_yk<alYs<$T#1Z#YzNdM1RId1!~0%A3D)CA@3<Jy)YHc%Jv}3&||_Oun!V(8)09c
zU_94+$v=r)0ks+6CLZ)NM<1^SLiJ)<&UvZ&@{P{P)ff9k34fvK4?slIh(m&wSHc9K
zZ6Baj2~Faq_q!4!(GkYw&<QTU4=7u`of1h90#Hm$tTfwWZ&9~z4^k4<Z7DM4Gb-kF
zj>KpktO3XZ7G$$15>;IJj1x0+f2vUX;>~Bv_E+JJ@{fy5wTVR^FYTq68Ev@}{Q1mo
z?ym^J7J_%&#ZS-q-A;LC!1oup0zAxw1;8uSek83?IL)b@hK032r>Ax6Br|dYXsei!
z5)wbMRfHwaas6Q51azx$E7uv)nSMjucNCI<@%)hO;ch!#2ke{{5r1~+K5!&58cqP!
zbQr*<*!BHiqM!ZB6UykM1h21n6?{MpEwy)e>)~H@q<I}Lkil+^bZg=F_Z#^qQNX2*
zKO`?`*3~q-$T6%`up+m3cT94iz6dB%D%pyd5y*TBDYqk*#f!OvLe4+NCLtyNTSl66
ztpG3g0C2Euq~mU78-cF+v^bS$xA>!e5^0-?gCB0)A{gRc4q~4xD?2yiFweLF0Noj>
z?&eU~4;m%76%4gm+W3QZsh2wZ0oBj^^f%5w&M-<;rjY=)kF}XA37M6Y`dAvEED~dD
zd2%FS6tMm-ScZ;<{D*kMnsH`=D&7hKsrHtlPARA<hs0n>xR-e<Du`AX-Y28%w=YNc
znESJ;|50sRqlJ@^t?n#QOpjol%n71SbW3stKX?9tFurG2EwEK0zrS}V&*m+G&@JfK
z3m!SY@SC8&KFgK%rm?6dRg0p;6=v?pDW4k|FO`WOVjMLv63>MkYnO?IpA#jE05^|~
ze>24SlSw(NJVO(T&Z%k1fZGvtteyO(HJr^XMY*|1P$BiDbq=Z(i~q`(mGt1~Qf)bJ
zO5ZcZt%+Nu6d7tbVQ>wT)1bCx*_9~;s;uAidKd?9gZ#Cm+nK<U%&SS3y(!)D=C088
zMC7?ru*Zs<-<GRrhF+C<Pqpf9JwrG)kE&@9j(uqUhG0V`F{YL*{-}O#QpOYeD2?+Q
zLxrloO>X1jjjZlc-4tW(!FC8o4E=UDn2pei!#)`vvuky|9f989|54P!7K}&rWHk2<
z$7m6!%Gu1j(sTJPuUI#`CYjt7dW|UXFP>k~9{N7+ZiCV`sXp9Y;7j-|{*KKVDo!ij
z#*ASJ);16IdTHmM_$XBn55*Ykn@o^<s+T`ydkjf*pb4w!dJ~B_qt&P9QqEgOqszxb
zOL)ysFZ_Io=7BNVJTgD-Xs&*0TQ-hBmj!?ItS+ToEdK~vbx03<sTndU8HTlswd-@^
zAt--pV*~my!H|sdrH86({4xGkx{H%&A;0a)_&BLSj0!bOTO9f$442>kyXSmMd)E%G
zUiq}u^K{5ABNMMz6c5u2y_;Prj<8oJFgH=i4}fhKip}%>$>5~awFPMT`i6$$t{VM+
zw1}FPI@}LEt=zA|$3AMrVdvXf!6G{c8JwUi*H;{c;cG5Q%G{?(d+(5L#5%f_Qw8}}
zWZq;!*#5z1&>|dL(u)81%cL4fm;dxnG}8&rmBt9&dfe(M=>9A`715RlRbc238OGly
zW^g7va~WUa1mRX2QzI9HzyH5l#~UzzeH8hL%t`?9xtk9sc``qd<2QYU4mv6}J%XxR
zCN&}ZL@nCID?V2RcJDWNx4F@3!28kyodDB#SBe!`YW7?i5;%~ayyCkQR}EWzqQml3
zAnze)#{Ex4mepXGNa*FSChh$*r7v4<#HQRB^{8WRP)HNF&|wZY3Aquyr9LwXNz>;I
zVcjX<2%xO9P_OSU*ueCuJyMLQ!+UBthNp)OU2}#66%>Ju$K)d+J*srENp<=YoS2Zt
zTQh2R!?AW0!Jm(qG~ghN@k$~sa%g&#na?WFVDWv)fhVD1-<AGZl1atthd-%_=i)i?
z_Sx{l+WI#Ba#nqX89F%v{g(pd0N?^Ez$D>1@^IVR1dReh%8u3Hl(-Y_3{Zv**`?Aq
zQSFUu@{dy+5465?S6nG>UYH@as~G6cMzyP45%;u<HS>u;A0(3bYk<`)mp)%S{}VT4
zcbmV$5COC%GYt=0k??Gjdk7e~avRYtW^FaRIh>n3gRT+@mbzMZ-|YDx4bdhTI1|sc
z%&gV>r=EtqXCnxezsgI(=Hak!xK_emV&|z>yuY#{je?=|!$$y*n?(>F$-_SY-Rqd0
zTiAe_sBQ4kxt|rhJ-)23QzF5Q>&yjE_OXUz=e=7%?GVC&+)zKHS7>2Vk;jbXiwb>C
z?rcy{-z2J69Z?3scs%{bEV}U@^H+kHG$z0QZm+YB9wrjP1wQ}u>$R;{(*GGXAWab3
z&$<k1(F;l$x~DX0WwoadNQQ9ip<r<{bqi?jV?>Rbha?8Atb!X<2C)o<GalY}wh)4p
z($eU%pYH6Ew%AOfZ^(BdP2~(H&>tzYg6Z>-^dX0}w61plp|IsaVfbs>7y(7h3!NVm
zBgt$WH;KQ72lcwrVy{BydGicTnaux&W1WUfX-w>`r?XojcqEpGJ9!ka{_D0MVFUox
zfv8zp9UA52-m}-8)zl=GI=l=`?%?>z;Etnmn@59f6^VJK&CA2wRyJ17@^GIBK2A>W
z>TC~lcjBwkTL$uQiWsuFxN=`uQYjW9qDibs^}qWD?fXc=r>ck1WRl|beX`$Y_@`5H
zLfy#s?KaC;VpSJUZvLMx-kxkIF0ju|tRx-g{?abjmj8*lXP*`Eabxnhp^~%E7{(LG
zM?m2K#giD9hX*-9Gj%J>$PcKZXt`}EldEvrcF!ilfMI1!2_|D<GGoRQu%&C1o~9Ca
zB3Y2AH2PcldxBHK@LLUx-h9wt%tB{F%P<0Bu@k9YH@ewZj+86VS&AF*8Vz>cjMo3X
zuY1>kWqk$QtP<(=Ga&=`@)<U;Gv|$)Eez@rY%Iv~9laXdLH#_~-zb7FC>9Tn<UrC7
zc{EXkc)x5-SrLf`?{YZ<?%Ks+H5Q+qBSBqJ)3Fa_H<lC?*#WuzKf<NQFsZc(m<dpd
z(NYn5L_yw0_Cy>O@{L&~)G7F>^1>>j7S_7Bh1swBhQD<CrEw_yUdE!j>0X$~?;brc
z@(WRoRC<2c;HDxTuP<L7!gh_))7s63G-`sA0k28Gj6|0liQaw$_k{f_`>aJ=aw}$8
z<T^|8GL8nDN1****lEDM0Zrvkyv1%9iOAyz2v<<OcP72odxEBl2`~s%;&+D@9PL}i
z3+`MNE0WzCju){G%sNbNaCHXP4Lx%vZJK2N5y}Oo0ZW=%J3rh@r!U>xBfCTslAn=;
z>A+!Mbu*aE<n;FG4TJNM{xIOf(|ba(^E-2ZlOI^230ODXt;>eGvns-HOZy?aE#yBG
zQgz`$NQ}h`PSB<rpi;J}b(dP$d{}Q;79<d2P-1xzL>`-lEum<@j<cRg7jkRncq*Lr
z|3QxP%qV|cvzc{6iJb^~be&O`HS#w6Hs6q;5>DGmi?~Tms0NIjU!V3R;^-$(0!&V$
z8gT6o(6ueE`fLv;QRGHwzHjr4%5=`+l%Emxsny-3&xYCwPLn@A7Kb=6!Z~V&{-V;2
zuqMD;X1mp;R`3ez9~I$U0)rp;H+u&A=1p^mS>g+2k^X)egrK}Xg<O^tfFrsB^*_US
zLk_wXCrjV6UAqPr&1EVBlr<{flYq%)s?YOxtWV$}#eY7-%^Vr)=O|yyVj;4082U2#
z3FxKiIAf|Ft(~(=!<q|`1ax^7fs!47E_ETkY0nICAuhp|u)HAb0+yPORIW!*{RLiP
zo1BtxD0Nn3dC7)+RLv6@Rci(!mIlKvH{BcV<8KZ2MSWWS{9Bi7Y5;fg!(~C7op{W9
zykdK^@`Ve@5J$9YUdrjEz17FYl)W0@{B_qCnTHzB!~=#9v1F#SGLsWD7eF7<qOuqk
zS|ebZSG9Upki_MoI(@AF0KK=r3wa&iiHn8E3}QnA&Tk`o>S9OfxuW;$eOq<KyaN{Z
zx}jVQm$K*}+C#bNe>Uy&MF4YM3Rm~gYT-w*#>kXMvV!@~^;xQ;BU#RIehw4*02T!z
zkRL2lvfEPMtcaU|%0P}L-lE5$$f5=Z>Ip|7kxhGUxpUPDumr*VD$W450>}QR$HH)J
z?FDa3q{_2(!bq5#Sn`Txebsub82UN$xj(!p9ig)lJEwGBHysLk)hDI8O1oVQ$4>8j
ztsR`LGkB;xy>vkL-+}^g-h@Oh9N)zOw(3;_VFN%WUU<r%A67ijw3V|x4{Xbu*E5@2
z?{?NQE!Oa7o)=!-o4eL<=8Tw#=y0lTg@`vNhYGZ>{H^~|<3?xD2~GNuegM9ehc{E1
zoFmZ2&yoDnH)Nk|ZfbDzqjCX3Y(EgvTSXHGXr%QP^wq<*Gg760sS(g4N4S)G%jE_^
z7xyfso^5<g>jnBJ5^-9Zz)gCm{5TBoujD2dJK{UD*^vjT)j(z+T6os$NHQFoL_K$w
z6lO9LZ!Za-suq@r3n_?b7HS|)DbDe}K<tPwa&uTUdcNoVHyK5&(5m?n{3-)9=i5d7
zC1TbMz|PajsdfpUqv9}Z6L?Qa&V3!zxYCC7Zblq#`y6x!7s!?8w+N1&ul`GDT5H?E
zlLU6ZmS`u<+?1n=$g$I1>gYSRKRW6qmVM|wCE*jCpmmKKn#7($KJ)XIfUf+XZJj*e
zqIl6+5?Jivh8VE*VdZaIMBK@0xL&iV8MMgSgL-c7o?x-1a@-Y(xEi^Z!2)HPn!-za
z8L&vmq|YKf$FWFHO)`1cNgedE_{|W05|tK`nVEvSTX#5mT9;Jhw!>ozkW4QBKSXM!
zQ!w3QO!C^pN<Q}Y0Ox{5z4=Lu#c!)j_f+2r!#@{!kp9&i{u-#RSjb)*OV4oxlNx<D
zs?oER00eps@l#;Hd{c^b)Yj(h3&Rc#Idn%Fv+s?Z3JE`;FzhH(suC?2g1TlK4Ulqj
z41a_nhVXp6#QRLM2}OY!tB~L?iDm)E)(Yt0Fi@t2Z7jFoKc>T#EEZ*bDKyY}?Y{%a
z168}N&cYl6X>>^Qt(=GX0(phMNj1Hbg*J20l3ek?ESw`S07Bk=+>xVOc%F$QDfsIY
z-?o*Ldhl*OG9=hSV-S0lEDg3h{|#*^7`!#H2Rq!?Ab^D~97@Swb(+i9{KMJuJ0FU3
zbl<!TT6wZ|Its;#7UU&eJ1i~aR6za5oHWv%-Hfd^lt%}h_SNd78<M!zvJ7O3xnz=;
zq@ov3T-9j%{j%*0qI)6_d$XpfxiULauB04LPCHdg^6#~uN-nMzQiSJ%l$^dzQ((hD
zHzsjrilHE{x@dWsW2KOqu55~_{}oSr{(9dZjg7UA9h^g%{FZNrzDG{|>v)krg#=wS
zQ3=JSf`=L0B(P^ds7RoZSwjNe&{Bc-D|#>EK;okaw9JfxSintw$wh9KlrF$+;0S5f
z99HB+AxS-_<(rj8Mu#5oYrAE=+cRf*)SG$D@U3FidB8wX^%{+>Xg-aaUQU!#Tgc$W
z0GPQSb)Mw)nzpAdj2kCW?@1YFYp6dO%%P?^w*e#v`p)nd68T=oFKe``_|4cc9z7lb
z#$EupO98VyF#-eF>RWI_`J17VGLe-zJcL3DTU{+7RC0n`kCwEriDfg_O>C=c(ami-
zWFNU`j8Lp)e{`REMIV#!oxDF^=_aXfRLB-CG^EgCN!ZDYHa4^qtg)Byf#L;JUPn;Y
zy$%#|Sb5uGhM*IOv00{)^$U-rI@dZ~@@uchS!dN1d9+<V$6HShTDdYwQcQ_-nkWX`
z=JzPXndpgS-sGU+wE}hZ!CwjOmd`p{iaKs{pZmi~(BZj(U#11F%0a?Cv56rp)fU>E
zK_!!*lcrG<Kz9hf1&LnN96F_Z4+KQ11;<8bNON@zKz^p2;V-c~i(Y^e^wWnYJvU~X
z{53wPejZj$l<yyo77;wE@~KrVbc?8b$X7&X`!MT8nMueidX})L>5RAUM<>Qw-sqh7
zXX9(1(z&{-uwv_;^@3h|>E8FqwS{6me4aq-(9k{8K{q=m*xA*aE4I7w^0n2l%V02i
zWT>B}=aE2B)-}VX3>}6zGDASxOA{DIo%Pk?vkOR)<i_GdVYt)}qQB2JH@~;S;fhe~
zAOzK`FIR9yUctgG;kUVyRd|l<Rqh~Ihm2AJl-&kvyuE3vweNqEx`@Tq<NOEm6&UA|
z%952|X5{|t@-PCNAKz(rzHzi2U+ltG9n-^v%_F?CE3#PF0Lk|%Tv@_bt3pt<(cKbk
zOD?dc=1-fKzTXo|Qt~-*S<J;;HsF%3w*H6yNa?JPuk+-aBNE&D?lgj<qPw>O`f4M&
zz2HAq(G#TZy&_k5F(oxj;3h`l+1*chpP-K&<<E09%~JXA`=`CTVT)HH+1Y@Yk^UU*
zTq=j`Uj7?ey~Upz_R@;hi<B?_GH~=tEcnyj_WSn>5&m>36Qth^_@djV<+eG}BGD!Y
zG}k+Z)^eOgoGu?p6VRzRO(Oo_7CTs(85z<8D&+<BR@+U(7u2SKn8l!7xLD{Tbm;l5
zl%rQ~k`nh;Eyixe7=KDcs=zp|zaOj!;`{M^dlRM0;a=V=QOe*sCe%T&grrCSx563r
zQC`p~z(gW4T+h~P`8pRnPuIU1%{&l`<`*9v@=8ytx`AP5rB;ZIKP6y3<#u9H%f!ZV
zWwnf=>{FLchZ(a&@7(oVSm$6R7?BjmtnwK=0aD&QXE<W^;opa<J*ql4J)8(Im|!ur
zBNaxNuwvr8>K*o1{4g)WjH>A<izVLdntCAD)9i`>a<%bfV4lNAk^Hl6SVMO<W#(TX
zqi&NBrcz+XD!od3i+pC)qOEg2u8e1=_m+P6Fz*@D_JJ?fx(y{_S6V_9xE)Zs^aL<^
zoi*%kfyGT%m5s0{=+M+7el-_SmhhNN%xmTL*F`Y%s#(ayr^(KrMKB*2<zAREo4v7P
zw0a1rmy%B_!(UdYRIQ<0z>A66pM*NM;j~}*U@GgjCjvc>yoal+24QRw8GmkO=N&sZ
zNO9#Gnv2dUd32PK<ra{NZaI<t=`p+@gzjt$Q>bzGHK~s<VR)LtmsCJJS__~B_O7XO
z_c!0}vJ}DYfu^cio(PRMIXFEbd<I{msF~a|DvYkKf_hcT4_rVvI9I&B+uVj4`Sj%Y
zBofCSEvdTF1rEl?Gs3N@bT+}u8eR%RGWA~POwzQ)vpSV`op+xXckL&eL=dPKI8kB|
zKHe?rBOXURLKE9iLc!__n-7aeL0St4=Oe=bkm2t%VP23Kr8|e!M#$Q}wWFYs3eu|O
z*&a|wM?g1f+O0n-VMYG;00QbF9B{yaA1$X64)`sh<FF}SJ7Ks!ntbLZ_QX<peYUEr
zeuZ@q(1Iviq7xsgzI+lACR65jk|`zyPkZ@V`BB6_o1oR3u9Qfrqd4V>z}E{4AEHVG
zHDM}TKO=I?=BgjR1&WHV_O4$PEpmrH%^T)Ao})f^u-O9wr~41kp3f3}`u>CDUn~aC
zlE&PL@!!1GJ9QDS_Siky?mczS$|~a-u5vf_A%GOAK(JmG{0gvNOTGboF~vMd94()Z
z3^O|*mmGGV3V=*qPT4Cj7lwb8ud2SFS9SrN%=ZcAMyfcqq3E~tfL%yF8dsV5Y0WQa
z!cEEQU}YfSFx1}}4zpW$BU91p@UpvAo(bR&XtSfq%=Z1D4C-f-JF8J{Biq#Kqxyr{
zc&Eu{#u)%C{?!SgZ{Q`K&j?YY!)I*eBxRS3ZKD9kxW-@y-zovobsV<)v|zPeKfhHK
z<nXvzcI+y5zo61V;;1m(#eXG}XVWt!ut;+1qg*p(Ul`t^;#DDER-YI1;hRgCnQ^)M
zoOukNa%5!gQupnx&SJfk-Ik)-dN%>1FoO(<0!?n3gu<RJ%1Q7_cR%5mO$B1c7t;z_
z#b&OEGz4eQzA|%oY&ut62>MH588V|HpYA&}$b@Ji_unU_y0q!c1)`At+7XM08=etp
zR0^E+I8HJSK<`>v^Z<i-l2blBn3okG_@rrLC1WU7hSH1Nl2+O1#dfzMZ=Wbp4Tw<+
zcPhnYzun>UmZ`6f4^H*Hs2!cw)UqHPRZ@J{7C6>awSAU}CnKHzjy<YqJ-damkceYD
zLdH^3%f#wlqY&V`n;d0DHShxJ>ZBUsXG_7T#9z6O4(3rJ_~ltas@}PF!)7%X!CczQ
zR6x600~zGH1_B+4CXy{kqh7m1C)tsx*t71Qz(7pRRJ%${^VR$pXnVJtZDSYs^K6~K
zas*4hv(($!+#kjw`!67}Zw$S*`+l!L+SUxLYkyuloBQetzdyS`dzx1J`h1@52yDdX
z%5Ye`U(UqQ<OS*Kd<C)6t)mBw2OVBr+55K=;3V}YFyMezA1}V<XAcD}<eSH)-sgED
zy!x5uAA#Bpz-ExVHC;A&2R|y@c$bX)+%&fo-)U_-4mH1892{vWnmcjBM-on=8CL}i
z)_)F?Fw4CrV=BZx_Ynpmu??RmLb$079aB0?tHaH92OsiuiUSOGS%zCACR1LpT@Lui
zNpp*{&6u53*Cy~l=w?c0J3gzmHA1=l-SpX{hpLGj$qPO~6vi;^f>Ge2E<tM{4bi#Y
z0gZ{BP2#)?F)mV_&o7I}T{nFGV~5INJ-g&MEbTRN;v-Zkou`8$9skMC6tE5zcLcwE
zg~0vCB|Oz}8#r>D)HSeVIAb+1x%+>;MwU514pS}d@Zv^f29J=;jB)9DYnAZ|2hc!Q
z6;~RwTZt^yRuc`C#9y-Ix9L?FE3B>c!&$1|eLw7BC{9MEm*TKjUd-h9M75do$#CbY
z4QXnD#X>UUqVoCg0BRt@cS+urKhF9@v!ci5Q<qw1OGp#|u;t*bhZ8O%Z7E}qjQt;F
zQ3HR>E8}L1v`5`WPS{IM)T>f=5xO;cdnYLqEgLIQh{-MLUi#5|DA}<W^N#tc!-)#3
z?bRGa#c#K&U!EDIJb!#4H1Wyn!p-_}&Z|nW_}B0*8E)<u1@@@z8%W*R9y3zTa>BM<
z-tjCYVOObgRHi5?!RQs=Q$4`5)N3#R{6SN02bk{VnlVdfPnkB8LD_m2={W0`WpS1G
ze1Q((%ek@+YbSXfij1{C;xF9%*zBa|r;k!;`$0X7Om*vyD$c*kwU$)0KvY|HQ=!%Y
z3298V&B1OqIXumc;p5*JnqgR{uee@@&LlJ8C_G-W9TzdN_}0+1ix?0vTsSg0JJBEp
z#P2)ZC!HW<g<e*UW*o7cImHQWV$biTLT@jrj}mX|Ip(fPTgK*(g8)~ll)GrlWJ517
z#U|qMO_%f5E*bm2Wlz6)qd`>aOYDxyTNc4xdAGqz^RL4%!%s_1i5d8%&bN?qM@yTE
zD$`X53qO-)bExUgUrPF-r<khe`JW{v{bh@Yn~=HYvf2^btnn`(&hE_8l9+4{U~;L?
zx^><Lh1_7sIS0gUYn+(6Y}qLbIN(nM>7>hLeF&Jv(@o+#rc(BbVF#e$p{%@3L6@(|
zpw@Jt9D=Q1J~+JXcXoBn%E`$&n5n8#`1mbtS0H|u;~>REc~~yT&v}aSG9N`%q*%8u
z#lp1@(ickVD4{tGs^hPP$bYqc_wU8w7!>E%(cM!QQvye#jHKH<9sh6E`gow$)mxw>
zNi|nNc~d?rV%BtqF|65Fb4D>nps3pw;r8*yMB9`O!vGVTlg6_7Hjmm{`g{7&``r((
z4YO)7V*;sOV`F0x?K6eeb;<Iup*}#aJ2+fk94d04ZZF@-d<6V<1cwsQ)6+9HF?j@m
z=s7xGZ4P5|Set5M6;w$a_nu6}p(wI0->);W@f!DU?DrRj;O5FEr}D$&slxA9rbbs<
zAT43+F|3E*Zy<;Jx9vC7$DHQqTFxT=$^WyH)Hg~5(=>0(v@`gmJQ=wV3`PL++@G^s
zP)OA3uqvQz<&FotHOZuFlRIZBhk8O&&DN}}o|$szDhWFH!_ID%ndaABFNbM+ER7^R
z$QS%Fp(Rr4^Xmk_X5PP8PktX9EQ62Vvb41H*_}=BJ=~>?0+$6FE_NKuH@OJ{4UCO-
zj#a{x>%8vXW$01LFDOV-@s%~o)t`Mv&p!3zy#ci2-PNl~NAKIZp+DYyTY6n7eZ|mX
zCEVzKH!T$9zVe`;=oQ=HuUAFuK<HbcwsVg!X%Y;iTYqIUPhXFTdr$q_4!1b5|K8V2
zP`VfcWT;m1a5Er&KLFwuG&bTzKC(ud8|0ae--w!;T|y%yi)oQ0fJH{W_utmbdU5*v
zAYM@A^?~+M8&GBzDz0>uj~ILfp18Q!FOOBa60vV*yAsM>21ONIM;gm})w-(O2obIJ
z6%`dvslNC`nGhp=rBGd|x_>Roq37!+o_CeZmKDJ+*r;B4ny$dT+VZ+zObNr>3y&-s
z-={(*sojfih(H>^GVvJ$-|KP&p04@7m*L4!oNvm{5eXJ)N$K*~3rmcFsN>NwtOnJy
zGTS3T40)Dmu9y=PSnRf^7IxL1xMYgA5c?(XSy8RvRiRK`RQWc-7ig<?`?7XZWP=3X
z=Z8zz=T|jr2V}rtFbALrRn{8mRgU&IYIn)#LMz<6ZuId+S*{E?kXl|5J-Q_Ly+-Wr
z7NMF3ygfDu2GVw4=lXCsx1Q75C=~VP5lGqkihaWq|JiY6@1PyfY71yJf==*opSUFb
z*S?$slz4=ZV1uJ~2nxS+xskuSwHxEr3phNW+$vz)w?kc^qpGus^6$huo~ST$mQ~QQ
zy-vkDi^|MIIiGXk(P?b}Y0CumxnF4jR=6?Z5W7AEMj$5xch(mQN<PEKtF2E@F!F%l
z{Fn(pF>QY2orGcRdL0Ef_X^BorYyQ<rULl0ml<2e-#4;J7?f1!t{koGU97i5hYQWM
zt!qL4)KUGw&<B0WUQ3ajd%#E(2j9{J7%aNe2|RIdJdsvIyp5{?4#^QdC6@20^&#_V
z6<+(ILfF_pFqO8ioxD9sis?pey^h~X*Y9*|{q$e0+wu9?A%a^huJM`APVia5Q91+f
zr^=$p4+MjL*?AG%Fp}d(Th$Uv2eCE}9Z%<9#$E1wZwF@$S3~b9US6l0F(vtes#Mm;
z0&9ZYyO@gWZzN;odC1p}82?)IbLVy4O{rWp<l-Gv%1dO7T}|%f5I2xyOkUm?O3QEa
zd@hzN`LWN%it@!w;;AHjFYwYUiUg{r8Sfn3J~}$~C5(LQp7Hm!^&9?jh_4s8IMN$a
znlj()+m~I~nedkirV=Kx8$HSvHC}EwgS@VZfDJu({VueAGh51Y{&0Yi&B3Kx8fe~l
zFOXjx`H>j*&%6}%q$X}po>^zVwqhx|lQw8h4J{Y^1Q=aa*qr#2hpgilg0kUzy0LO{
zMj3y^HX7&1IZDf~z_4y)6ps)iK@JBIh6C})1kk;8tciIvt@<Yuho6~J%)t#I_ocp6
zXf!K6{{C1^!BK|FnjZ0VscvUtW}ASZ>@*yy#;R_c;|y{irv?1{CpkI!Q*rUiW*~Y-
zCYdC2N!WzgVLF*+acNvPkgd8!o7t1f%On_X`^|i?i6z;7)otIuSuePNUi3p9jdI%9
zXzriQ<%{chJw;%ZZ0*9(6gDrKmq5(@9~pFP`h;VjFTNDG44CUR2%W{p{Q&yBKgs-3
zB8#D_Iaogxwd4we1{$fFSihvdx6;;X#EqQ$^u*YI@8|WFgPqA<0Q9>ByopV1(R^Kj
zx32E=>q>ni5{cQ;J8=-S`9Nv(>9~+vM9={@Tgooj9@|*LH{ZH{LhUv3Tp9*P{vLAN
zDg@NdPqd?-zphZ7$;y8^^H*c+aii$IwBm{K$Y+`I+?Ws`594?ZG+;roJ%D+|jD&ow
z=GvzL$+dJ_HRIKD0d?u@40FIFe+WPhHHq`6J=eOYx4ia9VD1>~54MJJlr;ZegvrCU
z5Dqo-6{7{y%F0Ss9@#A2oj^Y-wFS)ksW4yS`kU?vk=N6BcSRCA>6NN+5ur<|F~!d7
zGn)WWW$ueZvOQmw+Tb33!PJ9({4s46{h!tS?jJiz_r0^2<;e2bWaOELY_q^aB{sn9
zh8U6sk+&4QWM}^KA5>nToTgQy+(@5sdruXYSBXHgx}HUgjDwSOj)!}Atc#6_nVI%Y
z-t6k?WB^6Vy_PNUQOU`?3kwS)rFQg1MMcJDW@lgBadvjjC@zla?7Yv&A@~5G<LKyU
z*NNJsXg<)-sSu9OuA{QNyu6)(PgF@!ay^Y3wUXoBtwZv*Zzau!s$%V0_eZO*gg`27
zSGM*=KE?a&pF?bxrdB4TjecHyjka+HS1N?5tAlU}+P4XR%nP1=nAkOMK&+xsoQyJh
zZaEhWWB|Zm2mG;>Gcf(cQ=O#!ly;E;=wlnjO+pvPhC5Y)IWh+;%Rb=JmW%WN3pBY`
zZ!*320fDJ^Pn*B$$-uRVz#Rx4n69o`|Jfx~?z;iC1cQB#4t9VeEKz38D(A34?k=<_
zXOweji8y^zUcSt(nUO|RRP=&cfTE|T=S>9#x`z-*ci#sssw2y641e(cdcfX5`bvMG
zL(h8dRzXH(8Bl@MmhMT+p3hsN#ueEq{K#_)+m=1gW!{<Z<LBTGUW11l7M?ZrJLti(
z9n3SKxocqYT=h`>!9Pgu-DCP)N^7n<k|3zq1Da8>HSGl+zwrEztvHBv1njqZRk@bz
z8P4ITA)lz7WUq(IcSoH@uEfVAf6*3MZOzc}%N>~LRnk@wbOH#vcI^wx(pY8W*5dHS
z{zmWdbpso0Ug>fk4MITsGBlQmxN>4Hk+NLYF1Wvp2tH;yNOds2b+=X7V7+F?W_sFF
z>XC(9)b#nmJNV3VM(*!R43PXXCJWnefIuQ`K~?peLqdy(R16HLS>Io~0j%nMdR12X
z^&kEO2u9aLf1qUbK<(NxJ#r4#q-h;$`OViwZ4xFCd|pvYzbfR4Vu8_t|D=H^Evx48
z@9rJj2Og4m#NC?WCnDYJ3i>>t*kgwPX!C9b-1LtmlcClD>vygFm*oM`(Pm_z+S(=6
z@E|Zy6;Z!?4JhDQwbr2JwPp?q8I4%+c!OS6E>XqDPM9|p58aEXd<2$@vO7Q6VpTg=
zX=KB}?CCGB-WSMU&IeAn2%PK@B32JA7aDlAM`Zqa9dxW;rFWu4;C#_DDspNuW!Kjq
zZsKr&1UYZ`3XQV?Vz&)r;33Vd48X+!=9&7P#w}NZBo8h(%zA)TCIm?<GLOZNZf)F-
z5mGK_)-ESb_$^AL`7PWymQSA@yUgqS<iS<0cEHxns#VfbNRd+HPO*~5Yg7F_j}9}Q
za2p_wp7|>jlG=9CGCu@~S$9j{GaY(pXZ@|Ry5k{;%)j2^0$fR|0%=8K3e(+X^9we9
zL@fJOS#ga*Y165MnEuHvjWn#NULR5w@07Tl+9`fb>dmhHIRpw>GX~6b17o!?sM(Ak
z)F+Fn4ks-6>h*YKbWnsppm$Sbg1nC}prsc+U%R(ZPz!LP%Kloj%Hgk>+M`2i?e-XK
zR;`HE4Up~oJHbcZ2~Iz}+KdR3lm=5NN-G5I<4IwRsFbaRDgxc^U9&}AmpWpXJjsWK
zyCH8H6dar}J;dZi7kA~ekKuE;?B9_JtMNBd!W<Z{cv6vR$@ZwZfBuyV5e@+MuZ(j*
z4|N=bVhgD`>S}4bas51i6~A)&61fbQo#xHw?f!jkKEST+IlrvS{qWew1_S`<3d~B#
z_0*M_%3<r7^>!h%K{qLnA8)mPHXKnK#I3NU+A)~kwO7RY)o|xm?dCX7w&2k<Ms>@&
zsAWB69(iX)fVO<N%axgYE~N3ey%Gh)Q(eTP5ah`of??nS@L;=XN#N(%ykUpiCxXM^
zDR;E!`spBlhk^9H&+_rA8y$pgMm4r=yAa`QOH6Ore+3p3@NAyO;x)Co>n(bd3s(6r
zq<MY-kA_K>8Q@zjAPsRHgIc?5aF$&P57TCTpG$YL{C_rrx>Qa?^{N7Z7DgtMJ<{W|
z)>UFx11QP?TLm@}J9%qsWba?WQWISpgkie2tftGaYJ6x*IGWzsXt`ZY>=L_5?(Gfa
zE+^Q-6eg->V8gXslNHU)9PSP$dixlRKm67wvVE+H(mQw&1Ruw?Z-@bHrfwM&cszKC
z&P*S2rwrJSN~5Yf1wW!VF$dd37a~WpcUgA!^Rw+e(?7>!AKQC&eBS+d+zhr0`bR0R
z*1k`$2}&9)FiW8Rd!^-(wm_?60;KIZlORB(9qyB1g8}n#4eMXROpFI(pvI1DJvhnG
z>W=s>HvmKHVvg*0WYZwSn-pzC0d*dN{-+Wg^PbsLUj28j`Iv@}({3fnU_y;fv3bho
z`EG0R@WJuJAnV7EdOs102OdUQsTTko1bAMG>nXmn!a%xw<Ynm1HcQ{kATgldTVIvq
zAVr<3@&2Xbg#m~Y1<39)`&Q<|wQ%)PyXG*D^mk9EGD6txM?X8>x%Ac}{eFfiH>JFH
zyc;4^+c^|g#42#}h48n0$=-o23T69fcu%dJm9(;2T(q3OS8%v#J5U!3A_4A!-Je0k
zTUyuKFL6$)zSPmi26`%YjcGdY{3KBEK&L9hJv_~wbN7~W9;+?1C3U7Ktmw{~qQjS!
z0JokGQH6!`Yu}b6mA#gx@^aHi$4c;<w$N;Lg0(0Cr$gyZk_$)mD1|eyaFtS5YUF|s
z13rG60BP<Y<Usa+6t(}FUOal{6NRkIjJQ%S$n1EVRr>*OuVZP=<-UDQpIKBmcuY+Q
z+Hwv!7-HV{-PTx+pUIH^&;5Q(qbOi*yz+kT<q<<w0`<zBJ6Xd@0)gshZ@dZ;BByQr
zSZ1O*qD`^D-=bFnqI7K;iMW9s14i5%9g%07y{$j|A8m|`jWzD&+={uNn%ZkHnGUGK
zq3O&~h_%l&1JF1Ku2e5d0!p^yb87g%`g(e7EL`GKK>KVv-=8gRU7N@w2c7b@#u*fi
zhoC3Plh)O-6om#Y&n?b-ChVfv`(}1<DGF1yC)av0Ja9I3(c9W<ApaLx;L$q+$k%<V
z{|T;xF}B9LvudyO6WMExR^D@-erru(+&p-Yb!E>p(SPVYX7#zvTYtg6Y*~F^@B?cm
z@@0B+6dR=LxVFFcF>rDT=;rbvax=?&MRz2=BtWHR&l+(XvDbJllS+aQ#E$#3u6PND
zfgC+YC60NX_t*67*Ga9%Z6<%fS`$89JzZGKul)g8sIKhYk<YxEvzz_nWoMU5=LNz5
zE0RjnpxR`KaPRKc+fnn{>RFxqsUUqF31jW<{ZDAl;|#y8zRz{~ZT;&1*6jA*H8bv?
zUGuedby~G41}0y27B&AnvQ*2=hCB;8QN;H>Mrrrg6)h{dQg%(NH8HQnv1H%!AZ2QF
zVhhK%VsfZfyXP<G__2k-0vu2@xcMCPaVO`u|I084xD*G#yWkJ{HGTXoZeE%}%x&3F
z9^Tl%q)nM2)R+~+23d0zEO~erc>8jaaCPaEc~j?8s1`GjuGRdb!E_AW`Cn4e$?5+A
z8QD_ev!GuRmo5cPvdvMGbpF}wlHMs4K?udn1VkCo`Fn34`ns4IfvLd!J_?-13G$xA
zk5p#fCvG07(sQt+0y_TalT)oxLu$NFih(`^KTH-Jni+iAt40+E5#8w<jt*8WFmMQm
z5X7(G?PcxHPv;HwRo3sYtM9v$ds4we8@R;~Go1aPY}zF5Ie7?rZ7}B23W>+YfUyNx
zktd2NiJ0;2dRDL)^po0f^I*1&G2<QHw2Nm0rSab}SsnGK^n25oh=|QVP{yj$rK{go
zzYxpQYE@s%T!_+r_Wr{fYvu^>d|)WnuKjC_y7IuDqW4h-p+*e>l5S~5#HEy_v&3|&
zh79?nn0dL`Pxm3M)K+x>xz|5^+44Sz+OLblRw%Y$K#G1UDOvr0xcc&NsK56Aq*B>J
zlD$YumL$eLl}b@mLX52x*~T(<#*#wWLPBT?C4^)dV;izFM)qy&%UH)g2E+W$^gO@k
z)92eCg{!M;dcV)P&wanP({8Cfa>;r8dZY?TF%r7x83_iyV0K}%VuaVzc|P?WD$48e
z1b?`Y!bb_2Ym$relJB{fwPdEzpBx!HVD`Jk=H0Z(%{H`ZnQ|n}g-;pE^ItK;x*X(_
zuwPi{tq~8#ZNIY*ICo`nL;MX5vg!xIr6<+@xb3r@ft!g9-kA&(qUWEyKgpOl_&{k#
zkwZ@5a&LZqW}yFvpP`O;mE`K^qu$N*!A`!{q-M!)vrI_k>pFYTA=W3@UGfANB1XIP
zbe|0r+B{Q=QgmuU_>aeDzXvVoJ6ZjrMTdC7$P&HSDfJqe46`6>tPQM!RGeatwQ}?I
zLB5=%OuW5f`1p@L`4-mY|2+7-KMx*x6rU2+UssiUjU@LLI4&IL7k~n7hY1W5r^XH6
z7me^K@bEn7F3FxpU&{18XP$&OXMHE37d&l=DCnM|p_!K$n&6@xC8)QZ_nJ%GWqURQ
zG3-TjZaf|GZc4y!1Fz@1y;b(PC^}!Go)c(GG|Sa~%irll_(0{HNJdXq%;~EHFoe{o
z!QWI9I;j~QIC*Hj#*u?#b_afCr0u#^>d1d;y@i0ar3s6`K>Hrh?U9@5N~%q8G?}2&
zFeUF3!UikzT95}iM2AAXzB^`FYHN#%M<%sSl?gWiLP}6Q>&r4uu5Eb@MN8&)ERXK#
z=@C-;$|3-2%s7CIrpnjw6A6})HjeVcdS?AauVt`QLz3B=WSo4>NtmfNOm&*w$W7-y
z+<er(xi%DmwtHThI~KfSps=^qe(2Pj{{Y+3fusNDqnrQv=w%@j&%@jHi|^BayhF}=
z-)sudRVOSMsKtyT<~h4d-hbO%i+x4D3>no@mHY5Smaf^*YQ0!_dB;NivJHdUSc=E+
z7-)Pa4YAvCR^wVxitE2MFkqNbA|@kFD%d}Y650`WRBy&&U4<QT?uH9rI@A_k62>L!
z;a;?U%@4J`sJJ;ud7lF5-raOWIeL*I6i&UMJxF44jlBw2^)mc#bw>`f&V-DJh;(cY
z1yrJ5zk6v&;yM?7u*#<D@yi34#b?@H#aBeqS+-<7m(?ygf{}fu<Z8pMMV{r4+C`V}
zkHqIIGk^DvB{M5VP5LK7o&vnU41(##JbXK!$RW#q5qtf6MV<>ejv{;mXK6Yi-J?L7
z)YFS!j%S2+h+ZW{3aXS)S6hrqUC%@JR-(8kwW5rQ?Rin#T^W?$A?me`{yQ-D#er_i
zhP_`nMz85m?LH42OV5%?Tk-pwsVBtVaO4+Cc+aK?jP^T?o^zc1oyLS*csV<qFt&(J
zuP_=&`b#1zhg8R@%{L8-k1C#+IBap9yf`F0zU?!~?#$)<4dHdo1X#F&vaNy!g^9ov
zob7Bn7rLck5?_p^IoGnt+>e-tx>o}^oOiF1rw#Eem^qvlQ}wong*0UWgxcBWVLjwI
z5h36?8tA(_Z*DzQ>Ne!38f}kPB4KHM>g%7`oo7CWtkRcaeSN)<68YEX3D5B#x`?{e
z&gZsvR*Sh2fan|Jkk|bTcO(iV++~am$0vI(k1|II+O^+WojrOKv70%iHZ%zPANzK_
z^e&1G!L=C2UZcp75R3HrjBp0pa+T+9gvVm=xis9|00H@8fA4x&BhR?i_awcDq%y1R
ziVh9)HkLM3Y>taat(%cyj9+DNJ7-jWmLGFQB&w!@$@isYJzjEhR;syE7upJ1rsf+w
zrnT95l>>_nT<2>+Q{6y>(!6w(rjEnQ7K??lyLLMkG&~o6v|aSCJ=5c0u9;k7ZWLv1
zx^%v4F-5@Q`m((vF7mrcsYTlx<g~#)hnI##iNBbLS<P+NgfL6?aIOHx72@Hu3Or^J
z{FSbIj3KSrtvU4o7`BqPTTKBJYJZ?JVAu$Q2y{r%Lp+=FVAOkqh<T?+seC+C;MiOw
znAQWZiv61=62>gEVA3X@Aw4#CBjZwe>erf#qLcIvFxz}rf=STzn8ygx11+STeqnq?
zkzI9p9}(rRo{{3tw@E<gMhSReIbuEc8rxQyltO+!ZaBne>F_$wCiKbmR=d37c(Yn|
zS36mi-?mw7IZBJ4nudQ=H+||6h5QwI5$%`W7W}7M^$afvNDHu?0g)$qr{T^BXE|}g
z2|1qLXWs=0UzJ2Z;*UBrV7=MbGzh?e`ZQ-=;QDf3pGiYm#NgH#)p848E~>R!PFxFt
zQUiC%gHcP!C7(ak!pikG8^1N!oyq;yYwb~f@3=1*xbCcooWvTt@(CwZymL?)w$?^4
ze)M`>ek|3>^XnCK?~1>Q6{oX@Yu#``6yLgTLlmPi7uw2jx;UgX3;&PCaCT?gK7m<a
zfAWz#61^rSjcRb<?qCMx%OzYI$nlLKz-p;8!))bOC&V^>8rT|p@7jkaG-PeJ7M7mu
zl~j&s(-}m+B^I<OzzL1k&>bVZ;slF9j&XNR1LygouV2r?OJ~p0e^$Acvh>5UArcOq
zb_0TY-Axr#v1hnjnF7a($f9|tUtI2h^rjZaAlI5smT{?5-V8}km3jqI)_v*Q&_Cku
zTiAB-8hSs0lm&buR(yK=M8|`E%r9e!co$$K-vtcTiQf;D?6N~f43vz%0IpVAyvn`2
z+>Q|_jV9Q4P;HqL#1q?v?oCo?!Y>9Ufhrlll`yzk<SmcEaxb|N>NKR^pSFvc%gQd@
z^^C_W24qB1SbSJejishZ%Q%|*Tk1g6*Vc|Wvnv0HjnIb8;(-rowospK>da(5Xr&X4
zrNaJKhSs%_(*LPK`-v)w^<P?>7V*F(3L=6kJxD%H9zTHwQXgoLL@25*FuLnNjJTJ(
zyV8_ff6#w8=kaTa(S=irK<V@IhMhnIbD4t6aQZ8PPu0PN`zl}MbR9D2IE0C7`wibo
zQDIO6M@YpJq3eac=Kd+#(wLWF(xsu+C}c4#s_6NQ=SSsn&a*6h_?^L?{Z9*xBA;CE
zxMp@<j&}QG1R}&xGTG0ydu1o`z=vB!hIlqN{gi)3-j0S1%*Ed7Ks+i|r4p%)N(p0r
z{3j{|@Cfj*NrIE&Ze^lWLT7fA6z_x;ZoXA@5Z3^CaufOe710Aen(mA!aTz-z=9ODx
z;=j8=ubBi;JBPdlqFqW?>iOx%0hof70Dx5|b27C)Y?(5KdOTaJ_6TdK#_Ev6%aF@-
zy|w1vr$IZ<VWHeFi=&{KO3;(FMjt(;BT?1*@$=srzI66^y*~r5-!Uh)A5|kAy}*TP
z*y;{QT|9N*JK`vil$49Ho$T2=63s8o6jF6xu18_<DH5S(jzXNo8u?mfeC)(i0Mbio
z)Kvdf$77Bwb(=3;R1~SWp`pQOPTTHP8V?mM0iZYdVQhBDPwHFU<qb96r#?eg3i7>M
z1wU`=4{NrH#uyY`n3@!+9pAZrHbN<%Y>rP{iV%qL<<(_(Eg=3&yx(R0iy((8M_zV1
z>%>{|hGd(WTqbXa;smV^$Ty(-_vL`?+B-DqLlnR7P`^}Mtj~4a#KyK1OWwC>^iFb(
zKTi4E?~H=ZJt`53_<fbhxkU=?Iw*VHJ3H=zp$olc3@!P^11#8W(794>>6RacwpCat
zUr_7^yZMRvSoQH4@}a#jvSPSiD&nm0eK$&~@u$R5g_eKSiJI-}2|m-YX!c*2v?@W?
zT!kI-Fz~Ky>1aN-vr>csm+p(+yLPbxbavHVH0H3&8h~_m!%saCa)j7hqvt*oD%+F<
zF1xtAM!A{W*<Bc+)I$`f-syP~Q#?wiPKqF}>gjQr?+o<PTOXrJUoW&Ek={I1okm$a
zdM0l-58+yQkk%ylwA}1gf2&JQzNj5u)fLrThp-J$a2b>elK9Jzy!~CBWBLUw+*#rx
z#L`ZN*?A2-%yR$j8?FHGsQ&=c%LX^;ZhYGmzT4E4@4tc7<@=6!Z{DqklRAQzld;0<
z_krl3eacz^MMr2lbyOl+q|3g;dU$5T%)DpC@M@bTe#=3^zrio52D)4Y<T9ZW%~XQj
z279Itm45gP^IMeZ$VZAyXnwp52o7slt(~a+oT3VVlm6aYz3bcmfs?>4+^4y(`?of5
zJF3rsPz*ry4$1RCwod3SEQ$Hm3%5P$oqD+XRanMWD4W0d{?v_<$4i~OoH=u=h@>{K
z`SH#Tf(?ND8kCw3TYom+n@4Oh-@C+P=6Tu?yX%e?uza#{OvubzRxU*M$q;#SCBt?n
zk2jd6EqK(m1;;QQ;gV0uZ|{#K=c%s*R9|6EL?95;|30}@ya5CNR)GFE;c9{)!<~R(
zAR4fQG37)d@Hv`5C6gQPJ7OZy=w@``X|W}~-2Z3M>W?=PB*LZ!&@GiCbe{)SuEym#
zf$0H=1xn#Y=IK~EtN5qjZuG#)p6#BOukjq4@t+=DK^S4}M5Q?xzS3Mm_79m=VG^&G
z*Vy9K?epGLSQ$*vB6WTf{y%8(hcZXr$zww}<Cs(g6@SP8Rnn@7$uD3tbEikuF_Lu`
zCKI16a4YaYOlwt#Z3y+D&9-|#X_|dLeJ^96LqMr8l^&}z{^mmNxvc?_-R`e<lS(M#
zVO!0HmA^pWs#UpnN~ZaZhI;kuGO4G#J)6s_w85J{^~?4TUD8H&D&vwa&P`q=3M=sz
zC;$l8{aNm{Kcmhc4QI4^dM{`X)V&5D5ecW+B3&W}J|Rm%O2FD_=>xCm=Yh^p*A~e0
z=vly-5W3D&F5JM~x`DacAbPRU)D-(lOQtzx=VQaW506c~wAPF(eU(ClywCY>wU}#F
z7T@yk?dq!KHZ-X|$Yb)`!6DusJ>n&KQYCB?T^|3u#GW@WH{)8{sxzA^e@K`xPfEq$
zu)MXPv9^RG`v+9^+~1!v)$q@!ggK~^*!HO$NJv2X0PX0b3loLNYgtIdui<qEX~@hX
zaBjSH)1Zqr6P@!^L5tJ70bRxUmrgJM4^AX;Ut750yaIic^jPxnLl+uUjs^28l1e5@
zrCxIKp8Hdysu-A-8=!mnlXktH_l2MfasqN*GU|Ppk0*20ejfy#*E`ntQH<m8Qhfu-
z9HptMSNb6&{U8t!r4BV*eE*LkZ9Bn?ce=DB$<j%&e(y>lQsoN>(=q}1OP`!BV=~jW
ziZYu`>6+zVJ&QoI+;AGcQ-SmUIk+_M5*4>{@?=7F=irES(~sK(kE(;X96{4n;x>KL
z2=DYY0DX}pSP6{nfQH44zF;m{_}QXWN&a-}9ae~{o2=@^!H>@Q(p}d%P`p!6B5HN8
z+(5f&OJ*)8^k(GpO*XZXxo>r8IbS8b?7ZFJ-rxQ?PDLtA@|ge75VlNV7hy;5=)h$z
z*Ghp5jpYi&h<0VH8V~xd7U9XC2MmPQS3<Rc^~Ssb(err9(fcibSc8{Cua%AvBwmHA
z6dxTYZ;I4n?C$MF07PmvNSOL5ef|wykFyft=o>`u13=x6^W80DryS<NtG{ej8O}K0
z(_<D5X2Z1C?xdrxJ_!sTA^dHFP-VpUdb1cp;&v-O;;jnA@Nb^(&xdw~=T`ewNzM$I
zn7F7*FPadlc7SbNnip+t(DYSD<eQia-Yui(q0UGr?R-Vi2vjnhyy;d_GO=L_qvNxL
z;<kR4NQ2{B=~e?J&Z6Qd-%B0@XU*Ha+|hRoP2IS`sLQ6}K?08*BCQmbNAkJ^n^)*O
z85RAZK83~AX9OMZC({odNwCvkS~NtC@9C9uXoM5oTmFxxbp1iaT*jics`KE1(Ze{h
z$Hmk#kia*R1{fa9Y1=;EVLaLv-_OT6KAr4nAfH=<im*E=6s&!xnKirq?(q273SE#Z
zXT$DS)O%jc+e!~Kvvu^^TEh-CSuMJ8rk9ajwF+R$tzb4XTrH9r**zzn>~#$0@x9H<
zLH6D!7#lvY|M*&YO{&kGLG(0E<7RYEdf>x=BwzD?hB@#2;f+{#;<QqMh5)Xmfa?p;
zpDWE5{{R6F#3XC!+%!3$dLWK}#Y2FFJNK!ytT^Oy?cJn=`BQHil2~D2_`VDsTuYCE
zRz#_9p95@yJ^H1V2q&pY95tySynPG+Wfo1nmCh@?^sUw7=)*2&r^~c^)bx_^MQwK3
ztWZxMqZ?0ePpKosy+mzeWp=LWWBzyCM}Jfh*zZ3%lr$0?zjKu<IxxHUC5Fz?{!m<C
zmfW9(*qkv5<sW{a1pG%}Urq1o3A=)fw)ziOPs1A!&IN6!)c|`_p!U(tg5k856*IBx
z$o1BV#b5_dE2`B>qFCOyvWZ9)#g$qvugxJ3?b6)P<7bVy{ec)J&UX<i=)|z2+n2&#
zjAd5}%pGy*AAnV8qt~?4G@lG}I{x}rk;OLRbMlyILAML}KY-GIl=8E$yl?y=UBUUp
zVdZ}2qLfDz9C(f?pFo1tYsSg!+E(+3B#f22YKkKW9MOK}(&Mt&UAl7p?flg$Sp-Z`
z6F0F~?-2g&`&cETw8ivjv>4JauZ*NjKkSPgN6b3hZloiCVvho-4;;Hcr6wj9ws-@K
z!n%m<E+oaG!CP`M{%VX=`~}?doj`chYJwwoyQnQ**){LYje@KuR#$nIf2f)C_<8Qf
z(igcGo?D=Bh-O<uLS@kerX-NN%=LJ2Fo?j`GU560L`MwXSem%08^@S&ZD#pcFE}jV
z=4y*2O{gvv5RUg6sTOosYNJ+=P(CR+&CUz!5GRX6w~h$YC=J@BUqf9}5Ro#-S#94%
zJM#-$10IY1>YX9Fle?v@yxqG^z4_-7u}?*Y8W|q3eq5Z05Gpm_XxFaa=P-%Kwc4Qn
z$`-EB?5`noqDoq8Fcq?7#+zF(#%m9*?H_mxSkujOmup+gp-OQj={X6VtC%1sMQgh&
zQ*ci)IEdPCGLgxtaB&A~KvwYMRnzZr5WZrUZZq+MPqkM`9@rhzy%E}uznH1sb*+iA
zkzHwX>NpRG&b=^O>aRK~Ix3L#VLtT0fl`jOho1*61*+hR<jg+1>GC%fgV~7AAi}~?
zQ+?t;DFf>w&)5;p3#{%3E`q2cX(eF~xHR0j-EIZ|3CHAprQte7PCb~C01E}Rhj%~v
zfE1;!^9O(&%~TvCA8Cg&<kdBopOB$J<<ew(>dROMx9WbyDm^14$EfGf`_<AftGd^<
zh}gY(Y{_JdwpQu%uhW)W-Ic`OCg?<|){W~Gw{{#7xj{BgS;|P(=lu=W9BbbO9rC9n
z@D6`_eA6Y_*&sCO)2Y7}0nd56KWmot5L*pM|M>Ca7<vt)D?mXLtw%B;OsX9PyjP_a
z076D#KYLk$H}d$?vf8$VSccmJS-ia;RaHc%$IA;5qNt=|8DUguNqXBnSehErWeY!J
za?gtlO#q49v>a;d&h1|6<hwd5`A6ID<Brel5J(?%fGY>LKUpBE$#Xcc-79u8D*>9r
z_goA;<g)5`<zLnRhSvo|N>U^yNC0kTc+I4d%{eR~Y#yZxceC%o4PNUN+p$*$KN#3-
zN(mbj;$#B8-6B=(u-Hy~`@x0HRs3CF+W}7Lk=<`G>R(27Ngi7r@s$rd<Zq<=?5@UR
zCu336GeBn}23t#h?X#SHBct;k`c=<&o~23qrO*K>4=9a+yll_F9bFd}A}J6?Eb>=5
z<q6My@mKd^mTLc)1SB-dzsA@Q7okWWYY?0R&Z5d#-r{ZkH2_tD2U+3xc+9EZ#3Wv*
z*~(z?tIlbVUt@eYiCE&?;kW$}blX0yhwGEvV04ob)YS9IZB!?nM7GV5?9entw95PL
zt)aw)qT=ma>C_6Z9J@tl<i_=m0V$2SjYgW-@mW9k0f&`?abd&-wV8#Qo3YJZo!KsO
zzfAsKqwmc7f0;1tJaIpIQPcreFSG^9JgI%fgz(mu03a7P-fXMfM8A=RWAVQcjlB<)
zJV3EJC)gn7$LVMjSHeIf3JCoGoU)?R;&4XX`WCgC-LDFQ-D;)3ZngOVyS%AR)SfGN
z-L2ztybxEG`8o8le1fA=wAsbdiV<x;xn+9~ki~Mnn<LC>aCA;_Z)yKCw2{uClA|vF
zbXMMhKdHxe3&r6PqXf=Ti%&o85YDC+p)ioT2VxPx$utd6j3dsb<UJq(DX6hn{T}-r
zwZ;iWcw8mMmly<0dxoR=i`Ssck(T(;`(i%o{wrbT%TqG`n5XlvZ*7@?hiT$H`$Y`q
z2t1D}WEEj!zW)B6cd}535ufqnKl{hhU-IMPQiQOQ$|bMw3=R~U@UcgU%>O&2K+!*s
z7NQ;5w$UZnDx!J%hS_Y(F3|@P>uV;@LA*toADI2EKy$4E42eluwXMP+Q6heHLODdK
zrh|lX#ueJh%b5*hqGejmnh%yO8vP`$0p-ke0czL2sfR0ib*I10uig-hEHODAjg9Ah
zecEjnvIeigX6S6Gnh43gV6OR95MIf?6k`8Fb9%v~2=-4H-R!gGmCm)>Y&!1soU|G|
zNbNS5m<HPy+-C5VepF(r?Y&v^S`_0K-YV=Kcd_pKDy*hbE4Ew$5cYiq;KZ8E0b2hR
z;~}d^0hOR7oJ#6b-APlAp}Gcja(+yXxJX3<z3GqgTThG0*^E#MaTzE?_LocqC66j3
zp7pl+fEpZ1=D(v_sXNv2D#TE@pzYizd1vkCZhwW@ywi;e!Qe5o34)C9o8aIM0-8F|
zb_)IoL;T55gGn$EV)(~P*{%`TdTz6W*v!N&eT?%94U>SQZI*7kC(b_Xbs3B)&?c+q
z0{O76?>JXsaq$UkLg=~{ZAF+?-EC-Y`f~$r<Z{9?n6idQGdnY(0Mt_Md~e}n@mDoN
zZ<1HDve_c<2tzrapOvf9|LeoulIci$Ov=<H7|%MbRWE*o*1jSr#_P1zvBEP==;d6V
zo}tkFKnamUMu0S()rb2sq>_{4l2}}7_S?0E7u{9KeEragSV@$<Y^)zcT%qe-z8B!1
z0Nt~gk?wWxyf!2WLGYdj-`RmR81mjv05=g7ym-Ap!j%;p8}Oh8KIxzK%{?b)?ZUJ!
ziaVtOCVwm~sQ)kV+5;iA6Vo4oR+xi0zQf3*(h#~ZSTn#Z4ta)th|kSp1)&lq5rYV@
zbLu~CfgTj?au&JK_4&1UaS>XSHIczyan3Tw-rjcfYn}C_Ja^aI9~*#4z-hoH1E}$W
zW4?oz3<F0YY71xW3HSu9pTpGCmXsC{6>jJ7)EdS$+wS1vmka8n98+$QLkz@<(lMZ|
zJ0c49O9-v(YE!extG_(>71yI)VeoYgN)7oIni;~7?|hKmdZ62y-MpJuFXawg52Zt3
z|9k?`dT!Q+6JH@o?ut<AbN29{yzXV7U^-iQ$>)4X<Okyz;t!yZ43nu$(9brAd#c7U
z%kNGiGy%~1u|b}hY0QQW7(ZJ!mE%$XxZCpPh#(zdau+97*(-7^eVZYiXVS-^3?iT^
zAXHR8`yUl#^rbn>MF^gyhHYQH2H3I&@<}H^2DSXW|GiR734o=8@CBelvbpcTH2?Hv
z<!de+$b|TLURwoA=oYy5n8R$i==GPhud5rFkhldUYFCDO9cb0>{oJ^ozVry|r)Hhv
zb^=Hn#f+A+3*ucyYY*D@?J-0LY+*QB+ZL@3eQZTPe~N9cu#T#hylL!wIkEa*HpcT|
zLPrp~UJKJrqwg*5wD5yaDiEeYo&+I5`t~c}9nUy3f}|9OMpaZ-L%-DRPDVT8sTU1p
ztSZ1}nUHh>b^j-E8R%r)o2ISa+!|R2Qa2ZXgv8lbj4b*{*Dys{nC=>+nvtl4(2bc#
zn}K3my9GrKey(|mKCIiz_tSDjKCmES9sA?{dZisTrMu!&8LEETUl0pHM8PqO7_5B7
z^gHt1t?6E?1?E?6bzi{P159}Zksr7!-5*QCowTPO&hVGA?d6Lsx?yIy%QDv(EDd3D
zDgz%%+?tdY4CD&Ww2jgA$LPw9oS5%oBFI6tO+pR5=Xw7$=XJM+=W3{S=!b@{UzJr`
zsL#0BL%$8VY_4FP@2my0DQAJPcTGg@>=|ofa4YU#8=8Hde6d#vGR(WPkLg8r6PqTG
zJ^EY1Zw*#jfEbF59vp<HcM4Z<mp!-u6rQ?@$6%7^ZOSFE0zA|N66s$NUE7AAFzWr_
zCn@(xADigds+KHPBmQo3%LnM!C^&0%y_llw>Hd_JD1RKi^28KbvGP^3?j&m<b`^}r
z^TSu$$XgijY0*vr{r5zoYeX!A+w%I5wyejr2_INoYhL{2>zidz<CxZwvlJ@HKlVFS
z5wTSMCx8>^yFEb4$LLmY@x!m5+y>C}bMP=SMfhS8uh4t19NvcgZKEmmivqIVhiK~a
zpJG%!SXo)=eS`IX3(t0xE^<y!b+u7c%xGRT_Jq^k!WaADq9ZM~ISb2!QWvf+=)n1`
z5E7e7kRE<H)oWQ7l?z*cX8q;E^S+4=#)w1R*F2qk_IDyQ+VSXzXNlwhuQG;tD*(zA
z>fq>lE<&Z>N6?-(+#Qh>XVD=g9nL+ie}cOW9E6eY_$?jm?QAA@d<0Bmh<K$2FJelA
zT$`$14A?6)rKR#UuwRh3drECq5^(~uuchvmxx5&hq=G)ior#I92Gbv%$i>r@?EU-P
zgOMjOBQO5OG#{KD0#%VpX}wPzUz1%r#Z(YTa?DK5w*vXT{hyx$TaiEZ(-p<+R%d+w
zluhOx_{n6zRqh=SdxHY7jQ0Xg)huTahn^P1$2|UGljqz2EZ!N6iSbXRIA&49K1*uS
zC8IB5zH`e?#LVt=nxK!gwdCM8md)9@XWl>h*{x9vn_<mWN_RnF^wK&TG@DXVl33gZ
zGnN19x_8`Q3G>PP2a6%~dy2L}L5<)EvL%JhuYOPD$PX^O#>V2D!mL(#!6F4*RbaKi
z?^Hgi`+EXxqs}E^`KVrZhskfUom+DY8rrLbAz<S0T}v8~Y%&(<z^VaD%^%HSld*8p
zPe`n0n}zAniBzBylZ|@`o}K6Lfz)Wm^7}Pb*NM7$A{!3pzDfKGXi-b&uQ54mfvb_t
z2jo_8E2fprnb{%pPLDA?C@rhL6-aylDl{FZ$RaAlIlc{j22m*L(&NgR9JrAYH|Yff
zidq`kk!Aq)!cz;%@MdL$`QB=LYA^5o`>iQEDeNmJf%DrX6;=czxyNRHPcX~3*uEu^
z+QbAc>;CsLyN7;>5mzH8ox``f_qj5~?e%a1ii%b~M5Tj%E#JeX#!LtS-~b_Z4(lit
zZlA3B32(`zy48lR!6(@KDwof=XA2pxuk`@)*A@}gwzGsaUpMa9ku>`G5@09s`IZgu
zcOIq-1O8vVE|``7jhQuSjEh44#nrEmR>~mjExs>HM0B6XH{xFViu-1ADT6PR!z3kX
zFaLiGTc~Rq{0L&g*-FL!$VL8So}TT#k$ZrOc$)B``13Lbq@UV?2p%$l0KQg2z&p`a
zMGr}LLYX5ypJ*ItQ4Cv<3-=bvTO7dIOc0&PwQ7KUMFOwO^lS!-6sddUe5~_Cy)&rN
zD}Vz}WD@p$%MULJR3o2jC5GJsyim^yXc3Fg*a1!aXPlr~3i}+I^T;`B_*q+buo&MW
zEV_=ZoETUhxS_8}QX96li=jomh1CA5DsG<lO7I!CgcvgVBsQpzy0^ey4<^2E<`s4a
zCo=qCj{nibqF>T>8$ZBSo<9isik#KO8SmJ{pXP4U`|(Oe!|$ihe(hqJux8ekXj&uF
znPa-fNX$sP*>R7$W#ZodOcl^s^hPJywc|F6pz=H5ZTOIs)PnTx&>eE?H8t^?iUXyH
zMpivCU~KcIL9STboV~z$PV2SjuzlD~a#pgOmu+47dHpNNqo)iD{w_pU3MVfpLk=Sl
zf(l8s>|O;Lgk*=e9KK&2a06`zLw?Q*V{jCO=l%i=Ky6yA+R)lR!cnQ?0%MX&k7W-N
z-H2WB939>hzQ5*a*LySfk*6wGmR*?VPhYxRaNeBe-9CTID?&gi2Dv#D&n#UMkQS-@
z<_Oo>$%Qq_MARDa#l-DA{2~k}qsn@qWNu4VLcB31&dACs+M-{n{ve5+Wfj!KeKF&X
z&5OG)H;}CN;gJLz?py7e)kXW|$wo?a8^Z>LmiKl~s#!F<>D6D0Dk>%f<4ipJL6++z
zV+vpli|p#n9&=aHZ&eXMJq6*2gO@bABO{aiaKGL5N5=bAG(iiZ;+2>$T^QAQxLDa+
z7^I}pu$tfG<FVR*ggmS3Z_HpW?2eOda9XZrLQE%G|8yGrRlcLXhLZzsg=|8G<ZfTC
zX&I$fT)i5!ELU#axN+ZZ_j#0M-%gO>x2U3fi_iSZR*6JDj-Z77L#fZmcde53uUqR&
ze~|w=yC;q=>nVN;Q0BKu_v7rWN){YF|7Ri;c~BSw4hvQ9&U+T-I4hNJHrUTxr{SkR
z+uy8BEb`y31axB4z{mSnVL)!x1~R~!SikI|-oW5-F2+U9z^j~Fe1g=sX;N>%O4^Q%
zb6RVKAm($>Cn_S>lfnMI<Uul_>+aZ8m#@5+HZ^v7PM%4NdcU*&Y~LfmQ;h79(HJ24
z2{X0^jyPZ6>8mC?+ckIi)X%3W^H6*1ZjH;!B`RnUC)KUkERO%sta~xLmaW_8uQ!1*
zARQ2yJcQvh{Rf5WXw~AIR_<e*xWXj2%X;un++2~4bUG&liT!i6%&%T&Bb8~^BxV4n
zB5$x^=M+?y?_&LHRn0O7Yw0_@u`#nw&8Q04CUEgs>QriPG)W(rTuPiL=mbTdU0#La
z%%`9}P@GoE@vSUFl}MEfhPjfPQ;qwa>$nY_n$Kta0D6C(B0Rf0^g~jwYRh}zn=@cN
zD|C-Ko2aRJVJ41xEyxj?mGi4$$iZnNH7khcJKT&+&83h_LFlXaZ+SVrzlR)~IAvVQ
zGpjHsX>{Es%x5bX%jz{71SwVqWY)!SRh@7G0N3b`wUt4j44{I1a>%Ztl?{}`1G5Q{
z1IFh|s>U_kWx}TGh=a><dA_fT?0Vlnl{%mD{Bm=Qn06o7H;Kw_5&#{R?99+*PYxX9
zy<qTt6d25;Hm92QafH=T2t@UUSY7qkM!j}cn)*FX05d4DsADbEXA{{mJV|wLIVP)D
zHnpF7{iK%yhe(;`+3@7a2%%z$$_nXV^A!FQGcU6KqFuFepFSq7fb<xgQ)l4$A@1?7
zzkUPI(i{T9Sn&eAe>Okp^@tzKYB;JvOo?7E=}~V<nB9q#W?(=M^Ze>d7kARtI$-Pf
z-VouiwX-uSe9ke^5M}-A<q-kzna_6;FTJ`QdsfzuN{JC#0_}j*rdWL4XmwC#UY^XR
z4a+TE=<eka+TP|?8;#nQ+OyS|{)+0`;VWcBQ;rReWOhIc=<cm*V`R$XB`D4-{z2#7
zV44(Mm%EyQiyK{$59i;b$$4H$e%LZK+3`2nFndUh8IcmU6xeTX@qDMa7WE1bqH!`P
zzzowMGYhW#h%~(AB+$T7U9|-jWcMw<2-E;_z_hBa?3E|2u5muwAZL7pwmTALrm3vp
zOK}1o=*+CFC&T4l^#2m5$yWH}CSc5H0hTLXsGlm3$}IM*^;IHw0+Qo*KC>f3^ZYil
zC1J(kz+PNP$dEgQ1YO5uVW*BIsPn+3J27m1anliMrK!!RAz{1U{JiM&)PahQZl^(p
za5<Sv=(*53Ib$<F;uOfrv3xz%VNLYD@}1Co?titS!#wT`PK^}J-=MT`MJ|cyAwBe(
zq<7~}gzY1riK+=J{fR@lu<IZT4LhNzT2<o`*<FcZY#R;wn&Rs<_PS18f1fXHXT-W8
zQV>klpZH{-8~{4k)So|NfH?i)liI_{2w)9-nJ8y{2~JQ{ZL<W9>HmnngYVtHUmeOJ
z+U!yZgqfGX!;;vn=}D_xekHR^cc$zE)|6w*u&|{+<x=5OVNntN?0x^N<vn*19xNWp
z*hleIN8P>m101m~eA{OSrBsSw_}^0v)x0OumhGb&-{muBMS*%++HD6Jh2m_B{wgdQ
zTKb793<bzQ3>y|#Xb5r@*s`5f<Ny%|Zi@n;z6hUPqnOk4%F8&<0>2B|ns=vA)(htA
zTy8*Uu?$YC(iyiyA3&=hYmFBGatmR=8G=o>Q2pW{bLe%E1iH@s*IM~Fd)Hkc9{9r0
zZL6r(cz2t=_a>WSm4FJ(1`e0rJkNHnC0>R|2W$Z#`t<%hbv~T2WU7Mb<u|BY-^Ie}
zJG9;f=Y3O34)^iCUfzEqlc079UK?A*1Ol<sEAk9l{`V?H#0+A9X6H=!D>lewV#s`o
zjYD!@-~CWn2biDvT_n6arBnOB$|c5A&wHx+3>CgDmF>A{`u&ygg7Y0`UCw*v=579a
zME_J&KF>4|!&tSa-jFmcefLE>{?gONW1v|xAa%~g@R<AN6Ec2VN@FHtl21WqfslFX
zqYr5eQ0Wgo4c1idZR8<9%{;Egyp8(0&`RvZ2D9=-GC$9>ztW{TaOip26z8iFn<NJo
z4#%$$Z_2IgtkAOQo=JQYdj0*s)F<cjQ)>v8!kGLHCCFg-%2;=a8o0Qt;Ir*0dEFuK
z&twL!tWt)CSes4zNh3g%$oH&~#@=ax=BuLOJX?3FH!C;QfF*J$wRg4+xV2>Hn)gWc
z?dQNB0tH2;1*vso5LA1x_FOQ!X_^aI+8!0`R`X$Gr1IQ3Kd{rB#$NGsbag!>=?Pt=
zyVn5Zde;wd@F$A!;AmNGldtJ(A87pj+|9zryT*Ul9<jopmjYWo#=|)zx*S^9peY{t
z;Q{6;X>?lC%yz%Ttyd+C6^D)g#|WmwxVw4CeVUG|<I<HUEp?gX1B4BN4`8Gxq*l!8
z_ju9iMOSrllX#2IZt4_6NY7KqIX2882LXInHQM(jFNsT_tZD*N$fX09(}tc{w1IDL
z&nL|*KA(Ilo#gSfU3VQwmkj+VHNqz<bo?)acPL%He2rATrX%D}b*|j3SaJqKG&$@P
z&nVE4OhiD2&u!YcmeJc^&?AZ&#}hrk!bWU#8|vM^f1lJTbPp6G^A(^;<Swe8{P5V<
zj>OT`>sa$jwYzBx(^Qy`osT>P(YKmEB_w=XzfLF!Xm#A5QX9`@26KKcK;*oYx&9y4
zj{f`Xz58*PND8<S>R18Nsq3Zwx6-PdS*Zj-gP^sg;snR$&!EAiD4-AmnIQzV$siqM
zCbkQ}C}yd#dqQ3InX|0!{c&GGx{A!f!Qn<;H;P_T>Ck90lH9)lkiRMQ6!;o=*K5^G
z!tZu2mk#k%E*~TVS@?@hEFkxv+t`dv2-7?<y2vYkBp;$tst>xtXB}~B;Yq!fRWP?C
zUOlxlmw#;FM2DjWr98Cd;p7~)i~J7Jjj3<?W6km&DNYRaZRe=0&}gU6HM-gO|DyuD
z75yd)S9d?mQkR&pG~mCNlqJ0Z)f!Qg8pR#GRCB?()6PL}UgO?Dr$^81Z5yiGVjQI*
zhuvp;W8H?UiKIqW7r9M&mbT~$kDW>@0Q^S}WSvfy%U^bq!0GyL>%EsTH(OWUts`I%
z7$9HqT*z;rOBJ@=-ZK9HF%?hpfdM`Zwg_J8AN%h5u5=~ki66ND6}Pj_^sYhJKN)C!
zLH$yPTKW(0c|18mb&xbl;g_BF$b7QxH1J@kb9Km;CG}U5b&-5Q`M?K>Q#U>J;X9mv
zAC2eD?mvdB;?Gzs$YFGx5<c*LJ5+Q{$FB?j=c)qenTL1H#mj+`VTQl&fzpL{Y~Oyk
zohHVY;sY)Yh{61)gPWS56V46w)84rzo+2Ra(7cPQ+^Uri;7^X+MK}Frt#Bu$!NwHw
zn&hA-4#+<K_0k3PAj%W0MxRMV@sFM&@n+uVC+Z{SSL}d*6Wi1nEn4O}LKi<=$SaQl
z&VM23Qs5jA_gH|xbR%bl9&_N;bzA|mh~u~C>v23(>aH2eEciX~&GQE3wzHaFBl)%V
z6j}tO|GLZg)x1lB&w8`D;RM0N!ruL2Vi&l|uIZk9#uEiQK|p^1=lnqU0M1FlKQu{S
zDoO8#fXQx>R|@7d{L)MErN~(Bi{d0~=9j=Njr!I6^I>oNUeGC#A3&V}YEla>oc?Jw
zg^};*r@SfK?sjTVkHN6t&LUi`9-NCMK#)oo9a6wBLGSIn*ZJ6$XHMy;EEcBrP=M~Q
zEzZy^lebiDlI|qP%FdOrjrefuAnnrH6=Ob+YquDl{~U4jzjdMvut^LM<?!hq**G73
z!h`&VuB((lc;zWHRCVE><KtQH6dkMMviw{V@qDCb&*R>O$>n1JZci*!)4T92xcV(`
zyj5FfD`>QfwT#{DE5R)}-B=r`WVu{9NNt@GONf}U-}FOq@e>gMDXd?V6qLuzh;JrZ
ze;nJM&y(>^sa$xg!8Qh*j-x;`%qwS&m)x9PFEK51|Dz0p`2&~X)ZWe#k<MK16oW~F
zJR@)BRDDp)h{%Pa96*owsm!qq%H@{MN%Z8J!n*iSiKXmm!I74Fozkkzj>30@;YEar
z?vkq>+DJKmAk|mMo33T7huyd=uXJi-`q>C9=Pw=6n@*6fFr(?Ws6-%n77Jgzgw~1+
zkQg8=b07X*E2i>L44HzvCl6#9%!h+IqXYp;uJ*7M=e33#6WjIdZ;iRlY$FJC1lsi&
zu^=;le@6A^!*yK)?Y_^*pKm}hBU^Y6I-b@(zkUOR`%Gv;w>j9=){oJXN<cb=J<t~5
zf!?h2*&-cN$L=mVE|e`*E)LR98K6D$Uf4b%V`k#24=<I&D+U1rS`_f%ko)9W9`Y;L
zkAi>g{(ZZy>#4pTQE4`I@Ke*jCXDmgPrP=w4z)|c@{d;8r*=P-eyzg~?GI)F#)8-j
zp0B?D)Afl6hPhAataz61w=b51y+E{wAC?pMWo2_d0tj7NO4iJmAU6}+&*5r}y{we0
z{`g8T%=c|Q+v7el{*|nYcfcKvFX#!KKtI}>@FU87E-k*Gd$){kZ>4)YZsfEZ3i>SM
zQ+xe)d3;D1YmDxf1)vnaWol}AyEpu18kN!jod%P9D<<dj=g)(`CQb?O&S+{2m4o&v
zpGrn!r%<#llouG8y%&rRrJ`t5BAqwHeUQ&?vGetw<E+GbD=k4t7OqpH$wvv5J+_#V
zHA#?zs3~Bf@GCXC0zhqNVfuAxl$`fzlf_1)>{q`<L@qg94{Wz_)L8N7%{=r?YcTKp
z%5=b@jpJZI%2egy=zTST(G*mtMu0K3ZFa5}RxC(CjW~m^7hhvN&g{XZHV?kO)vv!l
z?N&uG%P)T<@4FHy7)#&5%V&n&_rn6Eu2^DZZeD-7FSDM8#vgb<b3A-89EUn|mgfH>
zu?y=~EI$098OagT@dfO$n+Xjp$^6sGyyMF87_oC_q@Je)9E1GecDV(vN^!he?M&V}
zXv(N`8t$TZra}U8q1G+HnyyR$6!m`K;24a{_+pAc$QhbA1IS?gMqFGR9rE!7rA2Zx
zbDa;E>p!QazUb-cpz`~!PBxkNEEWkMaI%4Dh=Kz7QGyDHmcj!gGTdkfw0YBXlP1uN
zE3gAQV3bMQl`a6?vHLAE=2mXcfmy1pKV*4U+%crN`t#1oX^&gyHg6EyO%cx8K35=h
z{AMa*%9c;`7lNcVt{eompe6{DU-mO%@_sXL#M=Y~Rs>+GBrSBtmzpThchUyHZ25?}
zoV}d2oUQyl#m5IDyl>VcYXyoT?(YmZU-i%eVj-}{J};r=i(9Pn9%Zy(+ylYxRj}7+
z$O`rZ;B@%z&iPESP<!KVQ}U#^ZtpH*kMpgV>-LBZ57+q%dWo)(%`n0&9MS4lCx4uz
zx%n#W*=^ZVdzT-S`#ILpXxJi3Uc2Ks4Mx<b(s;UL_BKB0gkUstOHPkF6<kcZb*rjB
zPRrzQUbx#;nCVA1&ai{}&!16^t~mYYwwT$5h>7JGeT-L6C`OT0DuOznog1dUZ(qQ^
z+t;r?#4)Puyqhcmxj%<Y^hxDjYbBuTR07~aGg1OvaNC~7D{p%oFatO8ibni)=slLK
z{~q->l8>K%3%m&L$pLzY3g(DjhxsTz`8S{-BrW;Vc6^t7M&_gK_(I_o8wT|0=!Zr1
z3+X)PjuNHoS+<!`XxYY=-D0bNjdAbxN)SzKLeRIoldW!rYI1P=j+85Y3<Dq@WcOCZ
zdNdD5w4ar_)5#y~KcE+*q$9Z{?s4z2S6II!?5QtbUOuq=aC|9e%I|G@Tfza!+vSU_
zl2z=Fyth5LLd1usr~0j>_GDzs6Fhy6cpm(=G=|`grn1mNcC`bLfv>>(u=H6Lux+GO
zAb2u+qper49+jcL_H%G7(;NzToD}!4UgSD<cv)v-E$lZtd~jJ~s9}XRw03;i<Vl%p
z5O@#H<7EtC0k0o(=O~@FHMpvyxfl^dzrc|>v%VEvc`Pv}rV`Mqjs;g*`JT?DJ=)T@
z5PHuw`E|+eq@%n&F?DvwX$)S6KC}mhM1yrcWD_vb)yVj7Y2OS#d7q>=>t@yVaY#KI
za}u1X(sc2msAeRtY~;NZtPfMLqk;$61739v2Z!o7I7rbID~XqFIO%`VGgQiU6BzG;
z+zlvD2+$Aa((?NBxzhqFR%bVJksnWMG#tdYZE(h6DsFqcEb(%bB7dn;!ZMgks(*3W
zJkM;W)93Fq#JcGq#y#py@i{XR<<`Oo?$Q|NXHMQHz$)-+MEyG?F~L(sY$c!3Y~LQ6
z>H*d8U(&Y|!)QSm4NgbBX5_V$AF%PonL-C=YQUx}_2RTM)#)abz%{EBSLIjr0Bi}y
zl|g&5l}11ir{nEb3Pbf#>Qc&u6%BJ>zO)d@^j{N*7`giy9Xd{X^84B-*K3CjN?tcR
z1N*+VG?A#7*@p=}wW7T~ieD{<If)4*$5tRi@&7ge?xXz-6etbaBQgsK7^2|u?Gr!*
zh*82w#ZaC(@swDn@_Lnnw&Pu1n|h&~uI8sQQfU~lDjl!TF}JYtorjH#^N&z&f-|6(
zPucYx-Te&$9862GL@yJ+ttsEB3+Csa(?HS=xzSA!_(TLeEmrgwUX=o=)uVDxSsqhY
z4gd_y<xj93hYY!L#hbVsVdjxuSmiTMcDo23ePF$3xK@kguh|G%!|(VkXE{YCjwCA$
zTXMJ#@8%nQeYkt86o*T;OJ~$vmo27Og`XDk16gOja=|a%dKc@0x*|TrvW{B_$yGVg
zmqrk8Khk@DPbuFrP%aR$?0<16B@X??ih{k3<ym&T!n_O_l^%tEvfk4)_73A4JxLYD
zrv!{hW1uU`<|E{N(mM8b``%B`0+t?wcQclI%jPbl&Vfj#$SPFQP_I4%llXAuOIqp5
zIVb2c4^`sgF09~U#y+_gB&UsSGjhM_ve9<K0?~ie5Uag(nM09Tk%v~42hojaQ8=l1
ze&vuKQ#|y{@Lu5FNZz@WqX^SvEn;7*>WKjULn+{9&q{gSXZGM$6))Hr2oh2Tx~9BO
z#~|NQ(Yv}t!xxV+fi7h44F{<^xkAlgc1!V6c3J4l4+6dN&u{YEN88p)K-56VOllat
zSQ%gDn@UFD$AIE>aUE%mOQ>%1Bg+BMm-1q%0Dx3Nm|QL$#m_L9B$E<|BcQ{+^iz5Z
zgzF5ZHISC(=I6JXbp3t-_jBizJe#mu1lQT6x{7%-t6!thN?7HIjPNN`(jbWV3YV`K
z`rZMv5VZ|~-8e(~u+j*Hpm`ve_b;_qV6kLwVU5RdC29#l{P%74&Yc<pJ0o(|-->p{
z>lfkZ=)+Mx3N~Umem}ui;`Z(*(MVd-JY+PaY<Dc=WlZ)V=u;&oMP6EnDPz;A(UMyJ
zD}sA=8x>Wkz}($KKSOa1w9hnNO3Lw_vU8czhvFyNoTwV(i;t;GnVbs~RW=|HK|L3X
zd%VK~g5(t9s%EP)K<khpbFgG<A($qYp}nGsC%mCKe#~e_K0)5k*gL&2zH-&-&H;#8
zMXQs$Jlxwe@%w|ja5rfcg4^lOwpQ}F+JjhR65{p57eVOa-eR@$6%a|Fh@ITuZ<mh!
z0&LnJJ?i(Q+bugDSSMo~=kDN?6d;fvAeiN$cXaAU9Q`;55%wzquj@s7nkE7ir3;Nx
zzv&VT&`I`W?wELvy#iR_cfG69S5$?>InR^UN<;ySy0q>U*W8zHiJAUj2Lz3k)xSSi
zRH)n^Sp_K82M|QUQ&c~Nh^i_-EG1o1=6FOo(!<ET$u^T@V_F8w=b3n{Fvu@&{~cmg
z5^bB4JPcPNvV1?THge<o(^H9&G{%C>fRQo7OPlehAW0$SU%cdg2vH7JBQnz+2Y6^o
zyZdN}Ee)3PQ;=Iye&2TGMop;-lOs%QpJ>#5dg)~g!RBAKRsERP9OW*)Ae@Q~Y@47x
z=rSJlUncGk3td0h8F*pY91CZPOsDPn|GqVfsEVbm{l*52n3*5#0uD9I5qw|k)w?&V
zZc99J7(wqYJ^W%jH$D2P5NT)pK8fp4?1@P$lyKshjb^p8PUw7Ns2Sp#iDo@t`eLH0
z*=l(#^d}n-fSXy@s*dNd3Mjg`+}BM<MAG%Bpo`^M0MINQd`*i|FQwES3I)gr=&`7&
zpO&4+e!U`qL5=QeyYu04W@iH^ZqIL9txp$#qdnhqWgM<{Q(YsNWeli46zG!+Vj6dT
z&}$t+2P@ikK0;k3Ep3B;X6Gmo6_~DHR4ova;#kBV>s9tg$2hN0ZsiqQxomLwL=`u`
zndB%4A_yt-gNDkQT+1$R$Y{>sT`Z`*#ac&;pz^E?UPQgDM3HaEQ@HQRjJAdEUcP5?
zA$QO5^zt(*Idf@0En+`!v@16y?a<x}^CrK8dz*J|CkB{1Q;zS=FA-O27&)Sct#n6E
zQ2p0@kDSyaam_tu4xou`Xe)MtQwMpq`Z4KoOI}X3{n?BCUy7{B8Uwgv5Yd3eQ?;#q
zAKq6UG8m8H=Zp3G{&rMR{4gBK4BA>NbL%<2Sjd)?s-Vz&H)Q6Tu31Ln3Vg!-L_{K9
zi7Sjh9&nHz`7Tq<sXa4`bo7*N5h(y;PW;Zp?PMSnk4$o`R4SBT0bLY@E_Pg<YN2DM
zWxmN-SrP*SrRa@a4o!jfoOzx(zzY(iHHF5&L;+HN*7<?IfC=a>=2v&K%MTNEjjw1A
zoe(*l!Y1-4YWH$snXv?nxMX0=7FzbrD>_qur!~o}Xo+w&-Pq_^ND2L-n)G{2`!+`n
zd66e)hTUb5HSLr|Yx63{vad@S6(i57#VES~yJbeL+480;>^RdU$8keJz<B)<dBw-<
zD@b#j+N2<gH+DZ+G6!A=bgL?&$j#rl-Ym5{qDoBjUt_7Gi6ujp2W8qMHjivr9i(%P
z^-pl_R(b(i1K@@b;0PMvk6rPAhbEal9;|c(kssalif0t@>Spd6j`u?6J5GRGWADN@
z2QT<A1iDQ#j?*Mnb&z(VBZIr=Za#EHPWu4MB6N3Fx6GGTY0;*(Ihk8VdQu<GJEb0{
zK2J|w1))7a-#RizPoa3pr87FRw>VM)9{m<@=T!4xDK~;yFOcHv`1$xsH~4IAZ53zJ
zW8av6yM8l*ua1sBfw{nTH|FXQdI}buVg}gCQUP14_wr!{1=6V}i6i4NHL$WWWwc6`
zvkHr|&jM=oeKOSK*?q?kkS8<iZ-z41ZmakWySMy-;~A#MX-?bssa~lub_{!xL)Obg
zCha9TKCBi=a5VdM)R)Uv`$qN+*XJ3~j@{EI#d6LvtOd+3H?aCLoiqn)H7zSJERY!B
zKT5*}@0m)7sNJPaH_7ZbIZ%FxdX&bqb4FWyKyvUA&b>kQ0#aJ~z{B?7LI_*39zPi9
z&B1ljyhWG>NndGcqn)pYKzdPk7y5aUtXPOnXgwTdy6jdtQWz%MfS-PuUl8Waxdb3$
zKeY=U_+%vl$oX6-9)!~$t0VZHo(Vt%Q`IDBIGLf_?NN$bld-%i?)FWQ?J05ycBRQ<
zR&COx`7bI!F^L+b;~XHlf<CW#0aj85gw4*a?s{YkHQ(Jy*0!_bYF!zXSRrM#E*b(W
z8+*LxB-QZ+M`|=ip~Slw;@xKvcGGkRzBK}{<d}K}Lk5`%bi}(zsr`G0sodvsi(=n3
zTZ|@APLtw)OA<&=Dcy7EG4y1az$;h>bTI;IzQa9gdF2^K_nWZNajXbQX;f8N$?x9v
zbmn!(kNDU7PzTH!X#NYihF8cM2Z7b?X@3z1kB6loxW(`Du}f{mVR89a!Nsm|B^aq>
zpy=u#&j1tKLKOt_oP2qB!SZWJ^(-;@n}OC4xEl9G4epy`ruE+2ag9Qd(UE@~#az)5
z&MQY>a*2x0PcXVSj{X>gm&rb<4TZyKjiOPKn_|HqKvXb2w*~{;!unUuUP6XFgH%W*
zY2W1WAAhLuv?{)Bb%VKdXoS!r+Je3tZs*RgU`p-<AA^hre7sxKJR|8yd25`V<;3gJ
zq^j|%L_{P8AD}x*>)EirXP2L=cnG_(s=Ew}z1=29O)PQTH*)PXT$ew>oT1`S2?vzQ
zIXh3T${X-Lq{zIhlBKeEgmAR{Kyr|f9mjdK-J}dfc?vCRSK?^(Aoua<QqS^31VyS6
zt5LP&=fwy{VwY9XS+);hVF)GYQ{BZR+aQ65yV~TH4uz*cvgMEi?w7_{er4*$5K}NZ
zSrQOca9tzeeMtaNF&^JIum>P%`r?qWjTpB5Zz3-)Q7(g!A#(78f&*XvBVb4sQuDs7
z9WM=7Q1_W%%rcWv{!~#O6DNlD=Y^DmOrPfSu$43NIsy=86IB3Tf2WvNN#%AGh8{y}
z0>`-kt_amy+xYFVs3_OX8@HcL_2;~o!<*2du|tJw0ne(stCn3*cZ7Ku_IPPb?Le;c
z^iaa9DlPDwvRI_@j*ecVF3+n|m0U%Tt!GAiDS5>0K)1F5L5kF#U1R%{Ixoj81gxTt
z#5c4s9glsM&B#kr27S5(Df~}9LWZk9eF7BJV#H~=b+J>UEI=GB@NJqs64Be5yT^Js
zu>jny)b&8g+NZ~jZ}Q>9w1R7Js8@NBot*VTF%MBE8KA_TU~X>j>wDD2SJ6#R*8va9
zoYeVNXSJL(`=H=y6XyxZNB2lA7a9Q(R{~O-C_sR2v@S3nI0N(|L?H5!wvEuZsq>if
zw!N#)*afcb_F$%T{2HP1A`iOCs>>MyqP6)9fk4XQ5?0WP2}pl0{oq|*sw%j$L|ZJc
ze0IqR>cseowz6QRU$wacjbfMFY`PUWu}u7ZzALWs$;CZMLU)yBJ2t7=hGXvz3A?Xu
z_mjc(Lzlq1*-4HtpYrv?=%yzh&BoQ8RkD0j!wNKLNgg^527i*!rOqglKvWO#0erUv
zj@RknpR;y+7Si)4PwQ6!3D-$|`2_O?H_)XF(?dGw-~c#4b1K0`LHs%*4p@%#*sQFq
z(wvxW0JY8{s%PkC_U5<?bT=bt?2%d;8XE06@5LXoKgNHG?kZA<Und*_d-ob;gkYQV
z^NCjXa6yBQXvob*TyxO$)Ev`s5fR2VPpZ=t_BR1ba^>)Q)*j=6x&<vd?8!;FWE<N%
zD5fz4-DkTdR?%$)Af7$t4@`YY$@^5<sqB1~r$kb~ndgd)u9Bi)Yn4xFCcB=T7Ajx(
zaM}~+UNU>dq0fYXvncv;jOQ?rxT%Fe{I#cmYU|js@lr6m0Q!KeWSyI+OpE-rI4L~O
z@8MEkJH!jqClCoa=wU4+5((h1oAha(E_Z7|@^u{4`+Dawh*c6Z_W(3AJ)CQ*vRpb5
z>aR{89%*Y(TFtxjbw&EZx8T4aT?`*G?&6EJ3%;tY79EgVdC1f1vqM4<Sw>rXN~kmQ
zBxJ}b%)3BTuR%A{_;T5yQ0o-r_6XfEap>cuy<rgpiB~Q_6P)GqJ8+wm%Z^nOl*PC&
zONy*mhyO>@RR>j_etqB)Dj^|)5+Wc73KB|}fRqZ-(kX~^gObt`5>f(!0@B@mrMr|)
zX(Xjn-t+zKynpQMj0?`*=RR@Hr_RPlZxT2P5x%QOqU#qlg<4e#Qdm#CJTZ0gL?AB?
z@TtcaORE|bm<k*if-JBo0?VzKCFt+`vH!GiFb!VWnfJPlb&EZ$1KU8+;PCTxQp#{~
zH4wPnuA%3tOy}ayIr(}Y^L9S|WzbW6wvf4@QkOtWmysPZPt8$Vd0{$HYDVTf?P`jw
z58Sg4AkhhKdx`4yZc3rby#~q>o8RIHzzpZ4v*E0(VH*@YJ4GMeb0;$fnISbva1sO=
z2x*e_&WGabSPRPf!5(wsn=I-^ehz=q(C(c#IT=!>&vOYYKdg6ZUr6=zkdxfFRPS|x
zTEi5b$P7@$68&+J=cxLQ+?3CTd7DASU_f0J<Kn6d*%9N(9Xyugi@s_1Hd)2wtJD|W
zo4Q*z4y*66z6Ba$;=j+N<V<;Q7<@-6bV|=KL@||#>86pz0AnlV)Ac)dNytgAQSnH=
zU>QssSvz++JUuna-<aN<-n2O0To<v{%W$qy{t~~t>9icb>`EPX@xa4JLzq2M&wAE%
z){*fqyYD6W>D?7hQtq#U63%o%zZ?@@6|k!M8->5GUqX4a8l7j`qJ-r~>}>_hd5Kid
zC(LTTF3pu&h?CTQ;{`ufepF>Po9j9U-3wCn<2$zGY<Hga65uXfm);M32l@9%RCdTe
zUKv`DuHLG9yFW<dAJ|AD-RvPW@HK!|PCo3s;K}~RM64%BgN4rbe!?p#<;*9$1(#CT
zoAWH;08aCMgY_)7t-ZuC{86;G^@v~C91Kr$jjUAsHFsjwg8sdtASI<|YNnXaAllYI
z%B#}#>b5U&77LlnwaR3Uwv)h0KKz%yuAG{ZR@b)gaOgPG(KE?$cs{7Rx?Uhi8R2w*
zBhr{l+BZZK5;*#BR&dsE%KOso$Pcy40}m*6Te}iuqRc~YUr1A3{4?@@Z<TW5)p<3h
zd1LhH7|2g2ifv`)2Fu+7%Ue5n*{EJ$w|+NSKh!|;v1C6`AO_!ASHOj|LiLF~>U7~6
zM}B3&qP3=m#*GI+Iq%J!GwveICJ6afh$JLe;d1)18nQu|1(m-1p+QT?&(#IU@?6{>
z(K9_*Fa0t$o!$~mQDWHgA+ww_Z-is}A)Z?=I+y9{QDu6aMq=>9VpI|N6E>^ajT;X3
zd^Ey|Eyu3@@!f<lb>N3Ii!ZTw6#eY;?kjv>XMV>y=RSQm+n(q<mj5<zQ%tg(CY1l`
zt{OTWmg71*P7Efn);f$znA?;l(fKMVmu@czE7(<$*@szmIuBkxo>O%dj@RGAkaHi$
z{rRPE-#TM`=E{wgU-DJ+Cp22DWQJ5CX37mtok?>_Ny96(3gioa<-01>uQ_{Ym+8bG
zBnPim;A15E`+TUj>%HMY@Z`qWZ;y`xh`#BT`(DK2e$)Wcm_zPQ>bSqsl_m~|D>s37
zK^GxfA=;}G<(2@p&{*<bL-LJxcI>E$q_xUSb#!RtSmfm708z#he!G{VRbln%)2IHr
zcg`PzgXafD_{jJF3<ZOmr0C61cASOvO8-{`{;$4zZ|UT4s}GrAus&9TBpM!p_sJpP
zlLO=>$8nnyrAlix{NrP+z~-{Ix%Ys*>G0nj96pZi^0V@=wr7*!eN231H@dK`PI<`n
z`8ZaVR@6p&G&I+D*6*7CvGM&uTHVP(Ac6DNo6eWGG5@W8Ic3SAMaZw8w@Pl>vMx=k
zlZ>aM8D#vPu)<$~Cy(3v!{Wz%$=qm+j|@ZSStd6NvG^$BV3XeVy}t8>nLp{O1D*+{
zW3^9EY=4J#`UirH0j<Xdg^X6GM)j-x!*3?#C-$J24Voy=?@O(}VKlTlb9Cpmp!$fW
z{2Z>#OXplZr&2~dl|V+TwGqLaIX5(a1&D^lQ4jET%<T4w-WodTSb^*{Mh1q|u0;@o
zWMyUjnE#8g=Lk~;Efli$FvkK$bo}0bJ67#@kBZl-dC@K+K3-Ns12V=J7Z(RgOchS{
z*Q780o!=00;YgJVHH1e85j(A0tX)1<8O~6+MI^Xh7<b2>`B|KpmDL|7dZdN}*m>c1
zz^x@vtnreQlUGZ~u_CdapiWF*Os=rKb55x-nlSXJW;No*kv~Cz!E9Ng^0R+(lP+nt
z?&TR6A*88v6$-6^n8(j&a%UzCio|YZH+lLIi>LY1qQz%1ro02a<h-1{l{p&}=nE4#
zw$>?Df7^6_=(y+2?1;yp)ER-cM~lHvRjy#x5>66M5>z_kG5X3-y7om)dSG{>JP1_&
z5HZnKzrJU**UItso{!t57!^y;@4~S+?>`>5b;`d`ox~uoqd82$vg}~?d^+2lyQB4e
zjnQhNM29h#5dEUQV@iJl^~BMguwllN3nxF9qd_>y66R<?7HHQFtUSb%FGJzF=W4Mv
zT?a5E>SJ5e;f!Y=?~I!S>~18t`*>&Z*RNmwlat*Ld_8wv9UysNKJ|f1wfqQc&*`x8
zf{1Sxqcx1C3(3%h^~^uL#*_rRzQsaowg2rq*!9p?dH#m{KZJiiJoaRtVuQ$3M9$=K
z=JIsSl#=M6Z>!h2W@Ns*vu!@M^bv;XQ(eyDZo{`~^G@TM?<5vU8Jfm;?H5w4_MaS+
zPj}uNG9ZtSy$2U=-leKc13cHbjahlj=>eK&q}BLG(VKLVjf+O*iwPkb(K-c!bO!H4
z+Y*yU9n0Pq*60~9ZT4Y1tq@6FB5~gN(ncHBvYu#z#S+&~__<qnu1<RQ(__z+0`Wyu
z47g3?5?N>V-G>Par?z3XKz{SSPOEd*j~v??`MGittJ1Tat~8GRg}>p+(S03t-=5kb
z3Af0sgYt-HTwGkE@ODF(4nOeeG`Bp@IA6Vb1rg{`0Ls5mRmBl=ceks}$j-}q1xtKi
zuRamDznJxPi`I{~`XEW5`xA>Y9~c@O1LCeB0dO9m?fW@8N;%{N+V@t#S+Y)TNKH=m
zR;g{9N&dIMApRPlksowfj8?zTY>#QOPt%gf)@8Q$*0!v>eIriMA}MK2m9_g>s>=L{
z5=IMOl6#VS&0+Us!NstDpBhj9^cRj7Ij^!h$;AVIg$HXS(NjEHzfI;sxSPNKq}oT8
zKC+#B88g-Er`Ih9JO&FE{E{c9PpVNQmmLYV?<{Q*-FQVO`6u#F=)jJ0DP%V*?Pb?u
zioJfV9JyiPNAV4xZz=Yw=I<U^nL~X|S|#TwKZ={(YMC>sTL1KT^J6|)6^&EMrNs7q
zQQVnDclsS2X_gQS^>T8LQ}eMGuJ)*=E7jbL1<gCWImN|Xi`jNb^30xR^S2OIFHZO!
z4?^kp3GEWV`PE<UP5L!5!u*_`H{9PJ_d2_FJH(9Pf0s|{R8`k$f%v(6!TTFT0xKz*
zONr$IpgpGKH1yqUlFgj1d$*e*dLD!%)G;k{kRx&9is!U^B{P`MR`^4N_oFeNN@$rb
zJL#BH)fPT9?l8kzQw%NlG_303XmqBWB8W|GU=XdyDDnJgF@8afE4NVYPU(L;x?22_
zfu8)VYqwf@qlldsfBRwHkaf6ZK1V1{J%&=lqfdcl!{rH)?avk5y<6hAEB%PGq^EaB
ze{#lMo_aRr-5kkqq?~;z<e&Z?ih4s<q2bLWovilI-<i&a=f!&TaYU;IL!7HrbU_n4
zUT^f8OmnU*mWU>|{Z%M;dRFa}PjR3yU(OtX>ccw7+uYCF&a=**uiW&pk+cnRhx}i+
zqu&9F9$isS;jVpmus%Q|^~_?TJhTDigNZJ4LE51Ea69|kRxL5u%0YlgUNS4V%?25@
z+0}nd6s3Z1DGDx(<Y4y5Lm#-AjX_eYXlZ14=sjyia6&Z-PcBKC_Ctp|R$Pp>L9SPL
z-*5NCS84DiormOhD@n^5#LSqKCXztyWa~`9@!XlXF;cwSRjJ7l@7vfN^p+#u?MOM6
z@I1jw2KN{^?s4T+M3QFFksV#N!Q1(w9;m{w<+^UpwY0VL>9~$sC_d+IBgYBL%2Jd?
zO9g3#eYAN%vl)e~_V)9+lX8`Vf-}pDX44p}fIET^QM6atx3h8jT^S}6$!gDJRV9VB
zJP4W)kSAVTLVwd$bT5HK-lqT(f10r)v4)FI2w)`3JpY`W9LIpPmLTRYGEU|Dhd&Ta
zNoH!;pgW~z^ruhTz~KiY8E~9Ntt&**IZuMWtzo;kef%-o*_elwVcs({^l!Dlzp8CP
zX0BBv)STbVXVbX4TJVEiiC}5|rH2pHBazMG$?dJTLL@GkPGurqq2f&UIq}jNtPTAt
zF!njH7F1+j3SK|p?8(kw@8+N6;?@ZKMF&e^&9sd-^|)EC-#B0C8mo`)mFVTDX1^5K
z$*{{q`7mu-CZ6?T&l21$-de1u<vIsBS2j!Cmeb5rvge7Ge`F-tr&yjarv05_y8EWC
z$Jlt$-bV3O*G%Tl&Q9jNpOS{rv9b0by0rnZA{S(LPyJ?`grW(CNfj^dLlVksGww-4
z==o~FbO7(Kg!A2TNM!bc^V+rcv3={YtrKS1m+vWElwPTM0wWLu9m8m~?nLQ+B$fZ$
z&cXTG(&CL9y<wiLs5=f?x1=@M@NfHWgkb%HFQ0@=f>Vp#^hp&>&Dd2Jk`l`=#57ZB
zf*1x=qFB?I9~|%iRXUE_UA1?_qxTJjDq5C?*U)L9sLrcnQi9^F2mNA-JUh)X{)N4*
z`_i(L+#D|ObpOQqV>)jG&S8-?Syjx_tgZe2jxDOEWusDc#!<6C*g2PV^T=(@6t^&^
zJTVxLy4%}t_?|rln`ZNHo|e(B4#nktNdDq_cVO{vFGF;ozduDV(f70PyYfZH429&d
z1($I`@cXFixkp3er{H{76N#RO0ajYN4bp0~%bVPv$Nguccb3*CXMft)fS8NpFUTxW
zHz=i?P0ehJQnkL4Z4dHCP`y1WBo+@ejwP%mSXd>JZ0^P$%n^Td)di<ysFawdjxcR2
zmysmMo+HYUAK4M+tR~S7jHiBu%uq(WrM-<KMUS1|{YZtAD3#*F88l98SzUB&(OL~y
zGgs8o-IVLRBP(^l-IgP4*Y;YItA~ax|8Z0$tC)6PMg|*x9u}kRS0jOuP^<u-l9MOu
z21hXcsbzbDoir~uH$?CJ4`^Rn$YF%ZlZM9$d=MhU2Q(zTuJuBDL?Wc6S2rohjp)8(
z15+arIM5n4UyxdDPF5-AiGH2Hcri`a_xbn<Od6OmmB@{uE6Q&PH8@(r_aaDX)`(R9
z?1-MFUU9hT<A>C-BAeKQTTf*E)DTz7I1^vq;q8{gzkPPD>5S2My%+O~IP2+?kuZ!U
zR12(uxa70x-85Q?ZPI>aXS#t)aowu^1L_8QikY=Ga0)zlWERac<+R5y$or})e`xaH
z$;~;9>YQFDPa#NlMA3+fX(*!@KAhMXcZkF^$byRJH^dK>*4E3=BB9LC6*O%)tr6WZ
z*aTWCaHI+ppEd%Eb&4Dy=oCU8Ay5*Q!Y6$Lw<~<oyA_n$V-HUm9RxGDL&N)=_9fk#
z>#y@3{HF4^yZPdoiBxiTQB#fJj(uY3$FSHx+P8fR8&bX39Pn<>Es@OX!+9QiHY+ss
zI56!Y-jN(~3nT@{w>wJK+>YQOLW!Ugl^^&P;=z|)eS+0zG*4A5P=~?4SJz<bH7@=2
zgtdL^C{cmj@}<4)-{*7L>`QamGO*{TGcG)uz67tG7I;mVPrM1hs}(WF)~(3T3S*Ev
zxc$@M{2~I1mslVrwRH6uV~=;2xs6&0fP`=eh$KeBqj;yT{ag78gVz2+WWzJ(QCVv!
z9}&RfSpv^<=ew78!RYcAAVDo%Pi<R0Df0dA-?GB`t0Yl)9neT0lvQ%}e!Hn(>Q?Ft
zW0bVrbx>~)+9}6*$#xZ9wN1&9s$aUjzV>***@d;MF_9tR)`owiiuLSPwGXkZEM~(1
zjjzmGQ<MUIkTyvb%8NH5rb~R(501TMo5B2udy=m2-E_MbdoA56%MySgNT|)XTC!WI
zvqhiZ`dFJ@tnf|qn90t)Vor~6Qb3(zn}-`Z_8D`ieSz8y9$71^C5&+UcL;IB9d}~W
zo_8%G!y=N;*G+ZQCcXfgBXqKoYW(WeD$&^W;^JZ;dnx#~YWMDPWai||Ly&J-ZT(Lh
zj<V(Y%7i~Y>_)vk&;8#_GAhgsMmBGp>0f4-ww5-~%}Oo3>oc1d6B>5$$XKkLSbX$x
z(I$rE<Ocp^s$d|WD`i#)skbnmTZZ7Jh+pzJU&KhO8mpVUT*dM?#e9i%Schr60&BFq
zBpmS!o-7;Zt1s$Po}*WC=rWAv9*$~jUzE50lbw7C$tw^OdEMUDTmI<@v+eov60Y1_
zxyRzcTXb~GQRKgvhR5c^xxZn0RN&tV9Y!Che7*oTpy*6_c^ANJAEYEjjDNYg_X(Ql
z1O*c&C$*WLe?}sTJ7EUS6UB(hT*{!tL~2_1Jw+sFcIqo*GjQXSstG%aIKn^HT1e=e
zSCj?DA-$|ei?!`e5Y=lR6NyFfrqvb81rta6Ym2YSg>P1Jja)Q{U%Ktvzs04W4cwK*
z6`Rjc#VhS#Gz`Ia{f3M~i_2t%NOE-AW%gBfRVe)Io-O7Drkw+x;T@Y>o%20RO=A=i
z`|l4no01|EMf8q^NmEEG=8L^qZ3RW<K+2wT(AN0?EzkBva6@pL^T>izj7?Fih3#LE
z)`i?FWmkKKCj7TtMGl32)}gDg9EesqW`v?+1CBUilu7jbcU&zY8weZTcG9-3$nDq;
z=}#X&#$=ER(T6EcmD)RRhqIx4N5-9%TZ}b=PS_!UO#PPIHYVh%rRg<z#0)3D6bKw_
z{G!6ct2En*9hsWMRUQ{awWLcZ?aUxp(i{=Bp%+heHv$?P@xX;$N)FP@Uo<iV##mQZ
ziNqQ6n7zu7{cq05K|!UvP7XKyw!$p^d*>oj@k=D0r<l%IBWpyAq?qIP-*JcLhHPH>
zfCYwI?I&x36cuITJP2&2#VQSgl;W698vk9~tnZd%$ur5yGpUM}u_>rBqwo8PESb){
zoOBuC1OfZ<_o0|Btj#YBfN%RVj5(5mUGXzj`YMycmu>{khJi1X!f#6%?TqMnd1K+B
z8?uZ1$qr+zzWq_-xLXky^^P$0`&WUzU)l@@9YeE`LJ{V^3GJ3zA~QBsA38d128zD=
zS-WtF!ddr~(RJs1V3qbh*-$r^q*0ICCIG(v!9RZd%cYWloh@p6Ibb8Npv@x=DQDt{
zl~1o6KvD49j-1s``B<e0S01VGtn^r~vqekvwpjMWX<B4a^94;*r59V%PZ;ZCj|VGI
zy+NU;y5kzBV;cC8fa>{=qG4maT8tUs|3W10@xk}MAy5I5F<M+w;ut`x4kHA0Ro1R{
zvS1Ym+FTjJcvaflvwxv;%iLdBlU2IZAFI7}^{&H$aap;>!Bqv)O4yf@J2lh^y^(+D
z@a&mlh`rBfM$_O3e=7F6z(PTxqBLynJqBN+c;Lh3W;fy2b)?)s1aSUX9_(%Z;w>*F
zzO3DH6IjrssC3Kd%KH7BYzeG~lU0>@T7TkfH{zN%O2b;JD(kzj<&94z*QrP`tkTh|
z#&#cX!)Ho%+qn!76kI=rA3;MJ4$qK^GbKF^wd+CzGN`AzDKH1(NMm5iBot)#Ht~C(
z31v4CUc*!hAyG}0YQHY4y++lIUSqR4*itlJFc!dkP`?q4>g|0L;_}57N_26_cJCf`
zD{LDU{qxN?5uG=^ME@Wg(-TV?{+(9a=571mCv)br3EHG<T-<H7m>lU_czZM9bbfyU
zSKiA>#+-F#;oiTlaz-UVW(~pIA+uh~%$UyX^qUpuwm1&?EgO@){6`iO{cO>wJAs01
z<i_PzQ9F@Q5fT@P^j$NFPlB&&=&FzQ0qxKUrd@}C+e&C8GaaZo#{d%Er!n^p0CS5O
z5zTKS#=X9mGB_2tEif=TGWW(Ri#iVfV@3Z^L8<9*oIw?~RoVKjdPY0IYXTSSFWJ;@
z(%rYiyg=^TQd;ctHTYoZIbOT)3=fL)Y_s+QOf7yXV#x#V+6oapwTKUE!hy;jm$sBB
z0mDqghappT_WH7tq{zyonC<7oq>MpNGZqEYf}$_xlS(q{A>K!It~{5flyt;-&HK=B
zmT>T0$^LHv1#V$3`;}y6%8tX-9)Pekij9NP9iKSZ+1XL@+0elyWYlop{Z|f|QLCm_
z3=`Fcb2Xslp!$|gKwx~hIW-BkDue;AR`8a={3x_@3D8UufXsf<$aPmuS*hh~cR-2G
zmY2NTt^Zn=Q}kD~qWEmguO;h~mpq#CJ5`m<nWSSvIyb~3BBF-w?V`UvAxzG<XkOch
z>yLKoLat=@Ga?tv%n=$HUwOO(cz8|r#Vor58tA)uerZ(u{7S*Cl!lswB@}PreTYn&
zJI6(<Jf={!4+(7WAyg$MRfy?d*(FvL=O&vy#qe8q;Ye251{7d1x(c%2$YHbs4>zOj
zx#&m6AgYzB@;;U2d;R}7hj?L{^P$ZZo#AOJ_<+v2nq{4X)s3Hj_~6<9$;#kzKXCxO
z>8Y+Rt$fB$(|Ay#@K{fGXRGG8zz-t>a9JKa_^3eudC?93^K5{J!y%c-z&<@a9XSa9
zH7B-3`ai(Lzf9e=c1vyWqo1JoMVMs|k5p{$^C-srt?S$z)KC1ohJM0^8P)hyF`=xR
zcS$#%>fG!)a>{GhQQCx)d4l1Wbn*nfursmiabr+3DA7{4b?i<~t;^1LuCCzjHJ<5S
zeSJJiL5V}qn#dd@w~2EAeH~+x((IFNDrG;Jw0SQ<uNipQF5Mm+-A6Mpx-MT&HfH~@
z*_m=Ryx@$hfaYgfQns34>B9rp%=re@@V4`cq3U!SvWDwX$St72>O8L0wYV~Tug^PQ
z<lZ%xDe5VA{pOi)H9e2xJ3~Wo5-5Im_4F)W{M)*y+#7tlI#Qt5{t6!-ACW)<P&&8w
zQ(+m2Ekousfnp+f3|B}FbnM7<wMqMrlXvfb<Av1*J%!L0j}x=0ug#M%OZkPIfn)cZ
z@+KTzYf25t9M{&;XHZqlp;IVhOEOZ^Qu^Dwc<vThB_&Yha;cL-M+>*&$)lySNx3O+
zUo$=vKb`C>oN4T7oN1Jy?V+>z;-<-LFh-v#76XTtRfR{$k9zXDMf5rH+O?q*^~N8J
z6J9nLDDd`+YnXxa%Bah}tu@O9f7cv0_sE)4?A?|07sp=l&8pDIYzS2hoBp#52r;FH
zI#<5X|4T5{-OG7c7P48r+#Z6=UK1BTet&aTt53J?1cS#BEC5^HJ$n0Q66tLKFYVfB
zoE_rZ35}1of1e%qd4l=jivbL&<P2Jc`3y*aRkcE>3<<&<v9PfX(MWY88x&d{=9rPK
zq@=4vrrN~;Madr$KH1xzf8#!@fB1ngqa+|AR$1Tp)lpx|qxX3hFME2Sw<V_k^1-Xi
z5s$<h-vN(U;>yu28(b6o%J{gTeaW)FZyIy|xl>nS@nPL<&1I~?;r>f3k7HAOME2q<
z+)xhr*x#WhNPC2Gxf!1f^@6}T_6pREFUBLQnv!dWR`2(25$#DukidC67}$BR7;9pw
zP@63;JK6j6^p3d?ldT}Hpj!64n)}FVWttMJ3|6`O6nahT>0J4ny|*2m?FJWa=9fP~
z$OfVfTWP4FPb=va2u6?|D3194SoVAoPH^2_0`#{(vjUv&?0}Z!obCl)L$G*cJlP5n
z!xEGNFk`R`z<~{sz}#G+o3%gM2IKfUWICT1CAWs|>uWRxI{I0A<;Bz7-?61k51Y&t
z=i(+nuT{M7?k0y_v%W|X$E~(8YcYk{?1krYt)$cC=Ju5i?utsx60V~|^xrtk&k9(y
z7`Y8%9~+X;dI((9`>&l~(MAR9p4Ty2IWF(ey3OLLD~%1Ojb+zz1*RtkZPPl_Hcp||
zA?B#qt=U`uJy<b>XUUs%;%4S0+WP}`sVkKdayN%9H*S=s+Xd8j{$RCLHe=6k=nYyy
z#~yW2t;O|QVld$6PQ54tryRiICT2P$U!eE-nDON;>-PzSvxQI}=QyrcS75^gztKW$
z)v-D;zf{i~rNFOCicDb?IRoMS*XCaz$nR~xt>dE(3ml<|Q|=ZS!?Cc+q@N(!aO?oU
z^G-fQZ*2`c4vL6Zd~IAwjj0x+_P=|%&H>NEKfds(2_EL~GR>WUWAr1&{1YC{ee?v|
zEq$i8@2Ct^z#pP~pBk^5Nb@+?nCo9^KC$gGcF7If4c9IBxaRKljrnR!1~cm^5lnlY
ziphluD_k4erHfctdeSYf?vapp4tS(8T3jO<pSE7ek+Z1W#+Hd`)n~VAD?s(NY$Vc!
zG6v!M$}C9nQSh#KV2KX?Wf_`mU-_nM;Eoc=%fFWFal!&mSIjVlP6D4TJz{LpE*Ey$
z)(3+TCjhc?5<dG@%HX>;g{GQ-l0$67jFe?*vu4W#0tcX6$wJB)7`<6sOz*jsH1A+*
zap)~n|C_tL@Jn~7^c|~b6|vil`F+L5#+5q+8L5wCsd8^t?AyvNT)FKVD8;`mY_dP)
z)XkbQ%=<gK^diX?6G!awlHZ~TS_EAf(<W&%f^~IZ+LqzQ`AcUE#mu%`+#|sQ51rC0
zyY!<vXbgQG6Fb9Ct#LdI<ksIx^dhu*saKRPccZ2;u4DeqW@kZ(SR5ajb{PR=+w18&
z=9kj285My|+h}3%KvhIMP>ctaz~L#qj_rg|CT8E|skLZWzs%C=*2j%u&vRBknlTYZ
z<ce0{)N5l(DmS*;%)E`WS)5Ie-Zr~48IwX7C-CSZ?>)q&A8$uhTK3_{EZ}N#sgPFW
zI8KPHaRp6`EAgWjPs{xXU(n(%T=TooS<#2R20rm$gVo>6ad2>iz+!vt+O=lOTV%}g
z3*bte`|>W=`OjG0JKnAOQ+5ca{~Gp_QK|pv?>rJZ<#Mw6tLlgSNWRYcbX@{iciRR}
z@U9(TE7-Tc>Tc~|ot-bbzFSp!!qCmdrCk;FUYjE6B26v<p}hxQ;<*vr$E?RxBTcv-
zmX7w0BbSiIXe$+BdgkL5#c5i}ltbo#r$0d3E!j#6c=jk$w^Ob&%E)V%b2-BCsSUL!
zo$I(t@Iv9OrO9OnRs7mZLin{G<FWg{VsS?ZUA1RFq}0E@wDrNGS|O6@NgUg6g`YHm
z#Nzk-#%O*s%qn*u3gvib->op$XL6E81(dl_`^7h3``*PV&EfKG>P2qZ9eeK6!<C`h
zielo7rh-l$8&Y-Yr;D$lRvWnreERmTSjN2VJzisDTgjlZg+_q)QmqE=y+4J)4HQvT
z<vLXmqVDpyh2j|Ahx^tN|3rz4oSF!Ew}H)4*3ifj{1i7@Zb^l7dL^cP<)t2B45*M<
zw$G?o2^o;lj40ipTrefrj8bnnTDvl?kmYHPEZVnwi)gJlRf#Q(nz_<sVRU26RgMe!
zP)zwB7Va*En`9B#`HZenRaJ2mY3n8rmY$RvLr^$=m!^mnR&TR~k+US$K@HJS?4}Qe
zEX)^^O$ytH7H!K_@+8R|rg?^e{2b<Y{oc7GOg=}v7=3PzsH%ciml{v(Fa7Q|Gz*v)
z12mGpy)6|d1NmwzL}J*bG`r{W<{7>&+GthuRkmAs1*_cNO~74x9uf?=(?6v0a*bM~
zp7K7j$D}SivS#=D<I@yBGo(FP?&&<BG<LYVkQj`wqduRV5({HKH(Z{nZZhF6$J$Bq
zns-bB>Zz=zhTM|l%|T>UK&y}y7Yq(6A*mbSS1|3GID+5>pxndE$%|wj;8ELPATqkE
zLgDMFy^TMCdtB;CU@K_Q^KiC#HK%yRG@KyQdRP%@!D1S{+qA-q^e&X76{qE`U*&n`
zEQ(#I`WsNnv5^0xMcCNPH?w!TgWZA$>`hCt>fan6(JbBX$M*{MqODd4XJXbbH`U4!
z4993i-Kh2nd_lJ4SE8b@;PeYaXl%#!t_{it_is(d*5OKyz_pH5@$O#y)dkD$E5|^6
zU5>)rWy-Jc`Vq&r@_@ch(?`VD6{U4QYBTwd$X!uDA`52;yFb|7HhjgqO}+H)%Y%OI
zH>sD8QWoc9<9H`@yj*w1tJT*3B{Du+62K603_ZRp%_6s0!9z;)sa>^{k;6UPN$X;x
z!L2JsWmz)+%BObAiNLJ`w~DFahVY@vAEZ?tT1rV15&;SA>93{^9LGh?9eBCuG6F3E
z@b67~UbsJ93(8Ld<0oZNpA)ttBwI#A7@TW1LZ=QHM^75($X9=VHqZZBSop%3l5-cP
z<jwzmSyTS*rL5>d&x7<vnm6@j-YC4H+VX2_{~G;C)+F=yFHMXndD*d~e@LvWj)rv9
zPoZ1M8mYk-aN)}6-7?bUTk+F~C<$CJvn{()NwRCT5iDDPt$hzw^4y9`K-(GP`lR>q
zg&QE1CFy@`y5IbG$*iM}rLZa@6QS~fBZur~i<|y^8`1;Wg_8-dF>*tUZPRWl@xjM>
zv%JVd-Q1H9$Gf+W%Kx#R7=GW6@F|C2<{rw!n?II!iu&ryoZ%3Lwm@fEu21TVhW$D>
zSVwq+#(&>RX0&=IJ}hcGevqgKn9zY+jP=v(6Ojr2s%&va{b&WD59lDZOQJzk91#=6
zHied`S*rzfqi#(7jYp3j<?1GdjiBZV0KY-#u**}+5O=c*OrgqzOte!ay(<B&5MR0p
z=L5)9^T##cTHHyDr=D;p4h?mD9K&IQ5k~o)Lb>WDFZwePYKMkjqor4Os#s!O=6_g$
z5PCPx*;@hQ%k8RVEgp=IC_H^N7megRLxttYM9+q+gEV5kY6fbZmrDjuxw^BrJjrUP
z_^eR1)mP;iqh~OM9q}SC?s<`EiPR6$JPTR>kxMyo)#dYo?Azv_`*Avtfgk@$bR@Xc
zx8H%*j;whnbI0a^$;9HCm%|%obByH`qQ+ziS0SPo{GBOOAKWgf22F7OnCRv89_J6D
z%Dg^Ok!}an7&$#AmG0A57mQZRuOt|4wXScv2uXQ;*88Y5Yr|RPP?&Dr3dX%}Z*H)@
zPXL?V>3Y;?0t-QtKGe)HBi659>w|dXa;S!+6}(OsGxbTgJdV6UQv|MaY7mH`VKcUX
zE>Z&`LV<n!nw2H%oN1X~yCVk(OzO-FQn(bpSC79P`DztE=1^a6RC@OF@7hY*n@3g@
z=2aXaZd^J~!Wt*<GYE43`qSa^_$9O&XDr?2hNq|!+4=)d9b2Q`FI1Rk*I)mldGVg0
zG4bzb3t;Bx#GQ@)i878`Hd>?$M6Q@VVQ;b+VwcQ0=bT{rUaZ-TziGsgW2Zd8H0#!F
zMF2%QndyzVp?2|0@60i1(a%2UhMiZR{z4rZO*<SCuiB?h9#^BbmzAt43@Wwx#dvY5
zL^>23swy!}C!Jhq7caCemgi2`$!dUjV<2=O(iG*!vdWQl1#Rc6hWCSo4VW}+knfTt
z=&Y%!?(5Y5IH*CkCNneei*Vk|wuY`Q%$t4QIDh>|o=!~+jiA#u!0IJXc!FIOECl$V
z#t>uiGobtw8>_6Rr(j~j)Y8(jw>}mH=|x<x`{{5AZ#IKNrVW0p5B6?H#{l={E5zcz
zUedP&Qrr=Dxzuv!EnSEZ>jd3_g_(QCZ;S$p)Wz#G7a@izZS)bH(UrI3J*jmlrM^V`
zW_4OvHs;?cCRC76Hpd#dNOdgf6q1imfF{HYuDh5PXlEI5|8P!#nfnai-pM>$5YIke
zt+>6p7Hv$S8pq2Oh#tnsK(%3BHB|c{{YPyu)lR{8%L<(Zu=iWMqZ@^a&UBY|_-!tk
z;P=m9W!bj9H5F|<(|J9@GbWzKeRb)b&xF?o$dSD5X5Y+ebW>>_{-k-47}{LR|0aG@
zOtb1!BS)fCn7AV<I)|fCp>qy0{Z{{qt-XA<z}nEKg5E!i32`N@&`FuN`bWs>t?|+z
zZw*~;)Rr|&@{JiL)1l(C(QiDAPfB|3>guX8(GK%&T_zhx#>>oIA$#p549r3_1^~Sn
zl<ix-d1HtaOdtxgt~YCt($LoKPI>GDsOUpvN(^KSU<c4Ry=?plAiIXp{x9XqsRMT0
z%&aT}sCNORDozuv<oe6%#~}H=l&y=>`z1whYrVrG`Jl<L;F}9n?50H#QPF0`IC<TK
zpGKqctpR|l9yZd&b^$QTw(trcZe{vn(^tV1z7tLpKds`)cv0eajwf+XPgLxrjYw=b
zmM+XaF?JfYx0Ji>?@Q*Sa+;ET|0dP#y9AxmH*cy+{5T)}{2x{u2ThJwIfz`7@B1dX
zVQAyH0j$CI^8NA2`s3LQSm}}*jLKXTr~svDZ5s!?RJyY_+UBy8vzI4hmYbf<;l9n^
z$l+tM)q>#Ix3QzYRN8!}PZd`>uWYTKj#b*GrteFRygl4Z+ID{{L_1udXYK0V#9<?#
zeEfZ2Vezkv3#8h)&IK_ejW}%Qr`_c`kVb`!r-$*BNS+w%<`5de7^9EEF)Yg8IUu_V
zv<H(PZ96})fNs<OLH!Z^=IMuu*H)k728}MEcHb4(6!}Mnpm}$0V)z@V=>-(^l;*V(
zR=!vC`l-0y-tlMDgnw8k11f`2A1dw}yVBiiMp>Ocu7~G_v#(MloyBYisY+TFnX6G|
z2A0DSBDYqF&{C6e6!?00v^d{=i+Zgc@Yo`Ib*fQ;QDx|7kt^aQOgUbFT#)L06^!HK
z!{$So(LC^_S;x$8$<=+nNX7BT`zc4%@I~`_WYmZ6MoykGSJ-igY8P&P*c5Y0gZ8dW
z*y<hoAGyiiiLmeK#aH&Z7h>Cqb{-O^$#u;y#cd9=igk5$0rL<F(BSJEH*Rp|+2rNq
z7%g^2gFJC(cekUnv+=jhtPNj#XXiISfHBVY3p_(vA15ayB&gV{{cM3~i*}$eyXRVh
z$%-33e7J&$VjBJMkI%O}`z|vlq5(eThjCyl@UBfBtl97ZpS7z*AAbtTq4R%T?oXYC
z&v)zFSdobSiotol^j<^lr|MV`qb-V7@-|WGXnSWn{k_!r?;*suvd#<v&L14AK)a&V
zt5LEkerN-<kbc~z{@g+vqXHN2-@LeQCF0H)Jf$RjB$dS6RaCihhwigFQ$IeW*^y?A
z{qtZq={kKg<;@s`J1f^cgQ<Y^c1g3PZv%agQ9+vEwxnqRW-Qj~L+M?XZmM8iRBjLL
zb1OV7ADKuJTz1qo{|G_{^%w;B3C|8waEb^FHD=GhuowB*25+!S2u-~c-)d6QM{)!n
z%Tb@OJSY<eCa=h}?+eIUtfz&(xGvp*J=(W9RfA{^i2{6JSQeN(ukP&jY}K{Rf32^d
z&M3<ts(do%wl^Ss264v=p}c#JXXaK`OAy>N3&6DDWTkC)B7fBC&kSzk_VCK}Ed*o(
zi&AhkO{JCS*)~e*T6|yAhtt2)VY3aMJyryH%8r6jj8?=;L0vr5VTN2f8es(LZ9U~Y
zRom7Dlhq-SRaJ2Gxi;}Up1)C_$1S-Yhfy1(p&Iq(yFst8Jk+Q7Tq1@eC(LzsXKiOC
zx?{5O@Ln!4O+EgQW}mQ$F*l1jd%;oj_>$fOOe&73sUYj06rv1X8gUJ$zQi5r*GtlO
z8sI+YXp2iVnZ`U>C1N7-2sw6E7o@<&&7^*nX-H0|XzoF4Mq05Zr%eY5oG(@9a6kO?
z{4En%NqmImT-WwoO^p`s`a$CdK!W7RO%i_%4;O=iQ=acP_S3Zjt%@*+VhF3UM}3^k
z%E<{fqncKOg^i3<+31WB0*f*tyLE+dPDl|Pp&`<rCN2c`NX&hWM&OWp5QQwbp%E#Z
zQFXG%`H3wQvVTEm^Yrkxbj#X)v1uRe+k<s8vvZ3K{Mu>ur{XB#2N0n@8s2o5Lp%RI
zL(}R-$VI+@+LD}KKAd?yLK{pIa#?{1!e_WkMrwb)M@FDU7^MvdtvV)Ir!Yq+Z_dih
zTFrhY`q&Wa`dBegt<9+1SAkr&`ydHuIK~clZ+@eL=b-NV(cEwCMG=13Db5)62yGUZ
z6gW60r%%64n-$wH{Do0=N#K)#F-yBL&A5WYvtic738kt!LVK4W?I*Kuv@e>A{%mXT
z1x?OQY>z3=N?LU_{yQZs%ga}JG;E_%fl1I+azMKSOlo22mzB>-ACd&~9C;BRKKS;p
zd~OfT03sGN#L3&|-=kpB?FMC;k(_1_y^YKOF9QC3yf_hefrg`&2QQ20%W1}tpwOuR
z$}*(<vTi8*_c^OQ8iz<aHb}eabwN+)e!n`yQjoSeNvly@-q&_V<5$M~a#yrHUbWGX
zRdY$enO5pz?QVWm&1|Qo3Zs_w{tz(|F|87%_)SGt-Ka*%V1~xY=Iy>NIJ+p@OpzRv
zjoHL3n#JjrCawBCjzx{{_FXWXQUHTb!x`uno&tkuzR1m>Ni}m$vg)j16G$&k7rcZK
zU`Of47b1Qwf0)rkORfy0S<8Ew-|l2(kCPSAhzhW0eYz6ir7EHqAXnfI)9-5EOo}Pu
zW?%#uVAs#Bo6R+UkzjKt1Nr6APM3<CUa}C}ax#vN#zvCJi3YZEwA3srCJv8)5)E}w
z*}5PNGQ*P^Y^s+NLaFiysd#LF@3TXaV&P3f?DcULhhWUq2PnAu>H}5~fq@eCrS8Ro
zZc}?7d&-Y_qUsP&h1r*!AulUb<-dT!+39cgGX`G4nX!t8eKve)b84_|GXZfWm(^jt
zI#jSk>1CJLT3!@0+9wTI(|`Toh!-yY&&M@l!(h;yy<%njYg`hE(qnV_!y8lHqt9+E
z^{8Zh>W@xP^OYfjYH2~Pf?Us>Q@V)Jbm-F~zqkUHQu|q(Ypm_+Y$Mt|jQKlR+Nvas
zwm(2V;g{4*s2V!)inIb&aoZASu;{{QnVE*5o2`8BIh5uo$Z{wK#aSQF)KJryjds1O
z)6H-A_qL0?u80E?GVSc`%Qi%#dy)iWvy?Mh(}kUCRaIc5Ly|Kx@*N@y@-w5VVf1pa
z>#2#(a!a;JK}MG-2JS36(CqZLio#rZZKfDYL05aneTV(~EX@W-73w^fChatUQ4+4Z
z@(i=n5ic^$2}@ZvFIc2wa4B8J8G|IU#>r^SepU{FEA5}}CCJ>Q0EI^mE9Zrp`&ZaM
z@fg(K1mSg=EC80t1^Ej^@Op=6Bt!1B^cWBkT;UAC^5Z(wVokW%NTgM@jPJ|OF_VCy
z6(E5qh8~a+v-xX!aRbvLjyjKoU0x6GRGB#iXmS;$ZddPYH(d>=TZa^eMhK$){%9Hm
ze6OcGv){r8=7XeQv)8W!J&bPVfrJYgLz?I~tO22z9NCsh&=gi*Wqtz%b<(0^R{~Y$
z{=?sPd{`p4cFUV=&9V|R%c`eMGB(~uH??=)wYj!*eU8K2alX(TAfOXr{O)gcQ(2cR
z$W0tovsVK6dDSIN5;rk?Wo+xbxPcmA;DzRxsBtNqbAqg3h(;zvJ5LY;sG>e^QTStA
z-8*HyKV7CUG5EyDS<`o3SU!1pdwGs^sQyz~Ri&60W4<+@@7xWRVzZ3-kRRe*K19At
z@3`DGvlv=NtyX+B;QC?N-t(6gRy1X0Wii9#zu&E<12kc+8{1_$R_q66(C6m4PzF=m
z1x~)EFb}49umoP&XXD3}YKZChS&~3BLJRbXlnnE)fS2SpH@EWPSIJjEK*o&F#EgWx
z+g-`peRwLgnH=|!Lu*5;4)bm_D6op2XZ*HXyfLP6eEWLT0l?V?)_sOdcZI7_ML{E8
z06LRMNUso~{pI^<yogygxs;4}_9f6$)jl#LHQr>hp&DVPNb`GJ@@6&GQNkvj#RAhT
z83?7?4XRNXHg}(}hvV~+!VLkVs_V7F3e&l~$(p^($%$|%A1^ItFX0lgY+4>C=s&__
zpm*+hG*Sp}d`chh!-pJDr)OrL85)3=DTJC|$($g|1VU1fF~6T;GjvB9V5G9H%WoV6
zw*t7N?~Cv|dQDHb9-bxgDIOX~wyc%uu8*-X1h|<p^{`EZS5-z>R|MAy*a+B#5~#~}
z#fPWm0U#Aa5h0mRUV)*vMno?a`L*+xm9z9M90E|f5VLe*o~bw!a~5J$$~zOYm<V&A
zUWYkJe13%`3fwEP*%H$%$<*V~5WHRYZ&SJ%cuH3Y;LlM_?W*B9FQo4Of18=7g?|?l
z<W4szQ)>4*J1a{6!T)M0Gttiu=wsqRFFoV7`#fX0{@;nYyvVV!$g$)FAp^j~X_b{P
zoheyTi;5z~N=yTuEo8iZp1^}7B-ui?;m+A^UqGEM(5`^cd2Mc|78|t@9c)gSo9R^S
z49+?R4E^K#=KDI5(YB~o?cu=Dvv}v5^3q`uzdJ289!DZ2jDQGsuLNn8+Cwag@SDq<
zHdC1TefYZ#zmTu_kL3EyKX-19gkMKg5f&o1A4M)?*)+W;cKUmD_G7hAVOAL$DsK{o
z@f>@aE71jjT5Nr`XsfE+xcL5a`3J=HXYUa1saDEr-(!oOk7%|W$`uNaXd0Np)710?
z^A3snK+uXCaKqH@djpUw(J%{q-W)i-mRhN-7a+U96k=AZn^NB$ZYJ8icbNK+K4IOE
z1SW+)L<|O?0un?rJ0ZA;=Ah>?XnxSE4u6QIieY5T3w3o<vne5UrZ$ILGVaQJ8x>75
z!8bc#pj8ow7(~_@e|JC9?iET)Du%B<a?5*FENWBC!s3Qa0|=e;LpCuAmuW6Uail<T
z5O~e_kWrn?2A7Tke}36uT{OUj&!%*f&Q}Jtg#}MY;}UJF+l-O3kpteWl7B?75`Vu4
zH}4D5irr}Sp%$SDU<#5do&~bu=)T0!=5|sK@#*Qxdu?Sr(?|=LRCxMEOmuX*@x+L+
zI<#1jN*UbrJ1|?=TnLggg8t4&8Gd(?gVPV8q(&!GLPP!iXmIHCX<65`Jm;NihKxg5
zd3pK%#g7omdgaQM%*@Pwn(8}N6;uID2CnA}H5*M)YcamsQY2D~R^3naqjE@j=S)?*
zqKFtMG8lM?3@(GdEA`3|iGCiN|H=x_taeK5rr5U^*aR3nW+7i6AIcV@FAH4`sW3Ye
zXG}Gmy*Bk&a`n2{Qc=qcc^unTSw&AsV-*??a6o!LOLT2jYn~u+f;~zd73_#ddS`>z
zK)_3gPTkv}EWNnAs?u_(();^iYsOxiXM&Xi8LA}MMzi*P15~Au{X9>cbxFGTjvXd4
z%_1{WsYpd#J)mv=$J)LUASe<(b@`l?Q&#ZEu}>MiR)w`9<efwVPvQzBfDgO+%a<}9
z^<_?fJ75zOTTbwFt*JkJBpHnfSM2R%J13bXxmQqWruheR-m`zQ_r`1H4-q}C)SKxG
z`6*`fHy@~Z4lE-WB<=WV?7w8{xMNM`2nIe>^x@y6J7y>Fw`K|liubnCCn`QFSYeqD
zM;LMN19w9QX}q4YL<fIM*S_cQ$Z~~MlRV`cPkW5G+8*x2=dH%a6*O&8udq+Xr<yVS
zgxRd^&rkM$jRgE0S5EVR3`aZYvEPLL1{Llye5<DL9}!tU0vc&+KdZ2sy35Y~-Fl`z
zgp%_;k}VIw(NxC$G$cz9#DV9}Kd{FJ1JfM~|4_ts`Auwm&(8jU6q9*c6?0C^GlW`Q
zF>LeSg4ySw(OdvU!K-%SnFIwGWQ-^cKtnq~lYOVgdCLS~zz{0l-oX<Mc?!zwR^I|$
z1s7vNnJn^Nm)~fm?Gf#iZ{n~c2w=zvj<5R5hYinU(ILln@g|*8cLT?A1`)i1*QBtl
z%_bX!^_oAM81%N(Vwh>MOP2Ny(}-RB<DE=rZ%4L@ntgzN0OCN!%YY^;gkOpXvBHUw
ztRo*7u=X|sn7>hs?$)je9h>%DJC63&g3rV1phD8=`(=6B{ascYhYw5J$ZBf&v0ojV
z<p}HwRn_QELAN4YplyL66ptq+2SL>wil`&uf+%*{7!U8dPC>zKG0O4%S$re3EJzwt
zX)|>vT)Rkes9|*vBBO-tZ95RpE4_T#4logP>sW|c#bUfP2=0+>P?HzG9-xo8FPq<R
z3=gsp%;HA0^q@DSIor$FNc9C<;p)`H_*StojJ_sTO+WwPS7%#Y3?9LJ_Q+~nT-|G7
zaq4qSZ=dH(gt&YpqL$(zl6HD}UaYbN_`JJ2K0<OUg;O1oZC{=8mVA6j^`53$LF|)>
zkuKfm&}pY3@lg#H4n=2T#Y)ub+pkS|ah580t3=J;xWkR}iHi<6Z7SgR`_<7xv&QiI
z2H4wG`AIcmukMB151i4~UJ;0(`UfJzgKMJy)Ke1bbVIterLKL_9M^sRoDjZx`4@Zi
zV@bBHHxeOI^770UB+GOp|J!4R9QnG!k*v(jS=gA!=3GXiNs-tq(3+<M@P7HDXdGgy
z*k2oIGuA?oHAtag$Vm2h+yCnA{+}lWx^?5M5|EP;2mW$T$p1jdcM$yCny9c|j(p1g
z8v-mIz+WI7#jJpfxHPLJg5@6awi*lSQ2L|NRbq8~2%8cWi3M3yf~ex)G3BCOG7sgu
z^&WbaqrX@bSfPMqC@NqypmHnR2hGEm&b5*&B7|B|dZ-bja@}(KHHF&_aZ<T6*>`$1
zyRRrlxa^}Jq0OP1d41C)$0KXyIz7)Mx-Wdh;O`FedQvw{NUL$a#0_JV#^1zXDmTH1
z?v{-=3V+F8UpK59d2VY$<YWKzzNI~2jEX%j*RmKyMU#=D^YT>2$l4;3P!8!aVRs|e
z3d+l4kkneUpWh+;Jkgg{f}^LWM+oA16s_A<9qhoXQ3wf0;X#vPfTr^U{^)wV7+VF?
zzFF?x?RVD{h=<D_NeVrNa3ueU4U>iT2wCoeu@7*jHFXb9KXXlPN@*KVd#-bZvv5&V
ztZfJowl~`t-{!XWBW~>d_r8@k-z??wQOzdZH;$LP|9B&BQWOV!w|Vy!I{SSk7U=GW
z6PYhQp1Zauuw4%NI#$duq>5o#pzXGjq!*cPn_gC5dq*44*sj>G#AfteTB1xv?QxDR
z1L#n!*isnf=ET?hc6%v`*FLZDXQnXM74O`)tbj;OfXbP+Wh9H=0%`#1_Rks*DnNsR
z2va4l3STR+^ni5YM5l2xM8<SkFid}`FS!p?>!d?GV3=z{0x7`o+}_bqRGZY;D2_xB
zfPN+6)8#49MtXp>#~#kyU!Zr*&39?=VHZ`@7h~Qf-*$GF{_ygQP_CXa-{9{UM};{h
z5d#(8m_a@hsCoE>zX^Ih+*VlE!s)+dV>aoDDwH~u4g4zM^#mP&R#|uyNvzX>8#Y4Y
z`($@O&|W((zqOthXUr8how#Eu8CqzR6D?aRZ#;<md#s+%`RBLpy9>RTn74_l!~}QT
zb$NeAD{0K9L7%Dkzb?P%nz(=Oo)O%+pTlGrgYA3y1MnZzau}RWRM|5Fw-S<llb166
z8f3(CBC3nVu(srpwO3%h0BA;YTIwUM@XJDdOsM!mG_;Ws6$OwbG^kb&ISm2bNrEH8
z+_kB)lDB@k)79<DW!eTfb31%vo*aJ|H1Yd3169-O=l?x%qfux7!W!7nHBg?l2gUaf
znfYx-5#y0@m<GF}v~ge7L}>^9a=xa2&Tdg>1w|f4H`V)pse+e+|10bzr+1rYcH=g3
zH(rPLCJDktl3q*?9+=m%|6`jFlO=PIDq8psS^71;B!Z4I5hJ_PgN=tgo$Xl;Jg+*h
z{X6b!p58`+(Mw>TS8mS3S7r*q|CfdH)n<SIRSEiYcIJi%*QCOzZ|sgD!Fj$FKf^vL
zbsS9axUt9X8-@>dYs+yJo9+15lT<jGSS>^r+~lQwM^OFUz}P1uv{VC8`&Ao`;CEoS
zKis-7pJJv9lfv?BdUhm=9WqQ4PAt2`IEDFg@8&jQJ587|>iaXPC?-1f7>KAiC!?9+
z15ImP$GcB>mUK!LL7jGYuBs9{e}}L%Pg2>dX=T#oXK161$>6<-lh7i?GHmj4=l;id
zFq~OIa;TA4=|SBdDFEDtlZQY58#MnIJkO0`c|#!7@BK5B!&Y#SFmvWlrXa*=^>@l`
zw}*g==l<{IJ2XH^;1`T&nl|=zRuos8Jn*}r@i>+<<UrEu$TK#QImn#&P0awaZRiUH
zfp)I(dvKmXLq54SdMNKW)iL?NA=wN2E^a<{3@X`BZ7pVnNG_03f8=^s#Tt>=(M$X@
zEob8E&j+At7njrz+r+s4=d1m4o@I>Soi{(hXu<JgY4L3|K+;Mpogds%*__r#zn*UU
zGM%3uF2iiNbSN_}Pb?!dcR*ngTX>$QQR0szFJGR}1F3njhw>T{x(#Z@`I~x9PCQ7w
z?l*;}nxkdrv5SsWLRPi3yuBE&DrllA`;T$IqraIeQ=3a&>uX)Q8%jWJ-2A&A(a)<W
zW7dM0H~1~oT>)YOW%U9CC7RB7+gCVHr{(eF^f#ZNy<ywV(xkES-Z*~()HDs!Tl)qA
zRmyNqk)@=0#mLd-uFY{@%l^e{_x#k?TC5cXQ80;)+=PCFgpMi0VD|6ZOgk;lf=YuX
z+^K-O4=Dwm?n5#a5;rHi;-Tkx?x>b4vrd@>FdpAm`>%Dn)+%jBkX!dYGV*9<)Z=6|
zgIgc%<(m}i0M>T4K1Pw^36T)?O$k?JiBlgbh#=Fc&<GpL9m~991G59OUy(Q2&{@X7
zR~2Yen=XhE{MKyRlV-bUGb-gVlg-U-R+EY9W$&L6&G|C7q{))s8P0D!p27_DQp4H>
zm&lKzY83qDrS;}73`ztK)s_V?Ow+&{qdM<Dp~MgQFUB>ER6F15lwy{js;R0L*)Pja
z1u&DYBk5#NnV4+P{<u7Ih(x+?yITQ|3&iasxNacd#C#;btaLL|F%<zy$Vit(PS=~D
zWQi@@jbz{1$*rh+(AjR%aHjNU(u8Y-gLH(NA0{@|c<{uSm#gtm>APnZ!RCw5ogSj`
zdQu%#r9Wd&anBVUkM5E&R8RYO8DtdtY#a(8GS0*U7156uk@H*Ni1Y2rFPPZBV^7Y?
z0REPqHss=XtJRU8T@~@;c#6MtblmM2C}k*=^v!e(MB&^ZTIt&DM*L4PF-&e3NPIWc
z8Y6b=aLAMz56W!kr4H7|jLfEx)0Azjp9WcNMpYkFu$TV2Waj0SmFFwv=G<fMiXL-Z
zw9F8F*%8}JC``d?8{H%|A%{!p&xjnN!0r2Z0BspO_TI$s;&s3ampmh@d4RcWP#!==
zA5HyT5Ci`o5kObB-@nBl%@n+ZzkXB#?VaA%8WH-Np^l&v#@N@JGyyPO72>IklY`Bl
zDBBAIOp}g|280`o3ILZo-tw36f&`C3$d0k9{EchbQ3K^;AyBw%r@DaUB=A_uwbw&Y
z>en`3RJsX$3z@4XAIu@M6(}nkRgi@aH@#>;bM2H|Rq&;9O~72pdOiiudk>QGW+vOz
ziPRy+{DIHSs+O;M;|;7j!`KFq<MeU|D;XfC_ilhTb$JZ|6eI46z5xnL%MB4MtHv+(
zQhqD#7*L%{`qrZDp-H>#>TQZ_JH9DQxC*h?k38@mPNV<4SZW!%o`ksYzRw$FB{e_7
zz>}JNS!ad2bWLO03H-L>DofSc?6^eK?Lb&70K@{dN(%5Z$nZ4Ke@7+H{QEMq>JS+f
z<ukDVqB9Ky&7zxC%d|)p>C1fO+;XZ~b)x=oeXI!z#LwaGs_N=ma!X5)RPg3krtg>7
zkPreGz(!Wxr=qT<_4KiKSwqHHFM+l{3PxWgMP}UO)wQ=RDl~aVI^t=GB48+4Bv}NS
z1Repb!ugg^!zs*C7b#~9*<Q>Lyv^oB)dSS_B^$26FpbeZ53#SCbo8afqu&KD4WwTH
zfQ2Osh4;EKb9mg9!qLH)Zc8)Q74=xEl=R}Qk=yJ}_bwb&+AZ(g!{H-DF^F~J;@*AI
zK)qH#VQy|-rYgrY4k;{sRm)$V!h9O&zW)4q1Fb7&dv7lov03lk1C4tI^k5*5goKj)
z{e2Y`l|m?j;Clyg4-UjzLkNqqva$s1e3j_9zg#x6V*f|eb-+`(|Np}&lr2R@BH5W)
znU#?U+1VsQ_D)vGjH1j)*?VsX$I1+my|*%wJ@fxO-QU00>%8uLz2e^UJkR&}%=ZBL
z_Dw{!|3--e`nzaSRaI3S;0c(`jNqrEIRx<Gro+qozIUtG<|D=}v<&D>AY;IlSL5T&
zuKt{8VIhxmIZ2PgR3T;I?Hx0pZ>u*i40VhK^xVJwQCtxjcKZ`d@k2IzM`RO?^Otx$
zAg+|sfEJ}cx%rTk3pI%|i8r`Rq8!8xiPQJ2V^gq1?wLAVl0o`=XHv&rk~P}aeSLAm
ziuKm}sxL6eA{3>fI%0KbdbnhnezhWY&XcGJ69#VpLPBJ?e+Qb#@b4)pK3)de+VY@p
z19t!}xWBg-4BXB_jsFB_5=dutL$4fpH99&v5GYB0e)lHPw|lVj1Jwk1`@u&R4xrb?
zYZkO~Dm>d@wKAgfkb|!t2V-b`%>vHgvO41@g3wDh?l1rCyKQ2^2tW)ob6snnsT<eu
zF%7mP3ZGm%4D28VRFdTawd3QTcv3DnEUATdF8Ic`^Vd<j&v5<{Cgh8e(RSdZm%Z?P
zbc3&c;W5r`!UC7t7!ItE-%&wxPe5KSI?Ne2M#Xm9Tbqz^bxv&5$Zhm>`FDP9tm^Gs
z*N?mt4Hfja8l9#Udv-pVA1+)f@$t@DSku<fSe9(|M-=)+)=K-6KT*NP7mmyF3fi{f
zBNWNSE-PA)eH)++W#i$ofZ_?cCEq>@m%v|u^+EX9S~$li(CXlW))nGi6U@XqaPaA*
zVaV6mV<&RByO!f}8eHwrVAWSG|Mm(@MPT`^G`rz04#(}?$B&^f2*}STNuvb)8!Rso
z^b$TBP`ohDC!~Ia16%F(`^ul^g6k}j9Ah%J=AV;HQnq4DjvY#1D<K|P`L~DdZuVm&
zyb$D`jmNRJk>9?TeyMivYWuhjHR`x@p^5NCc-*iT{{ESu^#^5S-%IqKtXoJwFa4=V
zyyZIv1M)B5v8MmkT3WwIu*&-P4-8+2L7DnD>wH<zTB#QK*;@gPZ@7d8Ru|U{tPdU*
z%tV*hd+ymz1jh1$e6<<ek?oZsp7kTgDt-8ug8sS>YBz<1)Zk2cc!&aV&kT-Hu2KCb
z7=sqU?}0)MC`2~S{ZWPB@#M+%`|L;qYH7(1FDNX}1pu9ZbQlN-MAR{Ly`T2d#M-As
zWs5tzj|JoOG&N<PJV{GqHkL>B&vuwUc#u1vz#bp;DqsvVK4?15OVrWX2~q<p6poU8
z58$U4Tx!{ViGTCns(znAGZX+LMRD`Wmy_48BSFdMN>_H<7icuIRQSfuBxQ=dqnt&N
z*d0`!&BUF=HAGL!a5VT|Ir<5JC%a!H(lTc3g|y4CN+DPcMAFr6(QT-Ep4Rx4<q_SW
zUz!cLC62!KlwY`irnz6IGw{jB8>~+KmWJk<nit_Em<461flIGUsLd=a-UB`chDOw&
zBo4yHdymqA4IUp)10+#lVXDBU+tBetD++h?El@`%GWv>>lbHaiP*%F@TBUz3hp4-!
zr*+0&X597w22yXc9WixvpO79GQCYCmbV+&!YdkitQlU;vZ&b@$&)+>KQ$EJ!dL-A|
z7^<2pG|J^E0Oy@@A{aPf!~$W5EVvj)VNq`G5_24_%4=+HG^y|gj=AZc%c_To^6*=~
z6L0xCsF=vfc^1QNA96pHvo-qvv=skPFS6}RP$7+W*-2n{lFQ$Y$vXz~yKFp_#Uh!7
zF$-(IM@r=P7+HRdSG7w~q(Wo?pt|qhlcRpnc|ydWrs}1j;45-rqJo(Z$#h{8Nq_Ng
z0Jj25EBc++qde(J@U}vzID@b*<AnX~EQ~im{lgp<D`sXozh&HwO$x{aoWuxvT|^BB
zD*xP@*9{(es>i-D+E707$KmyVJqS}N)TrcrLQ)(-J$@O0{A1C_I$%$*&y7VlJdNr8
z)GR#=^6GWJJ@685niDo?L0-9Hz61*V+5{%7@TM5EN>XWQc1k6tJ_HOT-&yrF+hhwv
z;q+~4Q_C^HIo`Z^6A2}U01DHoam`!j?bCv^gD}~ny+|1(6Hay&m6lRozI+*!<Jgef
z0f4fKYle1^p8u`F9u+e)wzA&gWM{w4zTy*BS8xL@Z+4?$huI;o!lyj4S@2D-bzt^a
zVmk$RR(H-6sl%l>%GL+R*5JU1Q-+3y_!!Sga{2i*_p0!i2D6kk)x!3p5K%&d;m`^0
z{o%DGC=e&dW7k$?sp)!MD5!8*9T;~aD`zsDg!z~&V%OuUq6YzMSop!dxCA8ZH^FHN
z2q<|5AYSsX+N+1#o}8M>%y`7X#g(d(BpE%JC}5%7ymAb3$4fA=RB9^<&R>|H*Qt9b
zhqL8Ya82M^P3OXg5EVgNM-~T04T>Rx&DfdKAQR(z1~A(tAkW&iTB2kS^97!M#3CL}
zePiz62}#LG*@W&~$oz*`59Ok*U|8f`=k&b=b8M`w6cv;Zzdah~2}aZ|pIYfpWvGh+
zAGUgP&81(g^Ep0yMt_`ZR26D#@moLr9bMX+PsyOi^S-kuir9tK?a%tJ{l=n6&46kg
zJ?MoX(J<}~!=0z=ZuN6%F^2bt;rqiw)XDEU!I*LyX|JW~Q3?N~_Ps`OHXZ4W93P;e
zsW!V2B_`4Lsq?`a6)}7tnHL`*G{}nMU~<I>aIgf^B*r()X;)9txf5ZWXNX{Y;~x!S
zLuVE*`X>n!u0)ac5%56aiTN5h=f78IitZrasuaylO<EYBCy^ih)^%3kQ90Dn6DFx;
zKqHR;6b<V$A7C0V9<7$BH15U#0J-XLcxA^Cj!$6ODhyD+0$?V-aE@Zep=M%!QFkSF
z!(hO6{2Q^EKO4h?Kr4N&QjK+tvo&5tlM~1CFZj;>1o<HPz&>w*g&Rx25Qyu8CxmwL
zyDS~B{@FiocumeX3L2g0JeUUiukMxrF4+!b%#_x&^R5T4D8BaQAF@1=kbC_2u`Y{~
zwUrfDS@Fdpgw=fI#YThws`c^yALLA^?DHunh<x9yh9of$ZhSz3KtKf7WAr08<|qsR
zk5z)Ptxvd<kgqhJ?L>I|L1or7FU@f}EiAlB-{+^Gku-eMAj7bAl-fN}Ui!>F`k>|J
za|yVZ^UoR*3dD5WR2wfn{{>;$+8O^UR|!yTY^7-L1Z`^-h@ooirH?B0r+(bHU3Z_9
z@$E=l`&dox)DM`T8UmxrnIZX7U$j;|sK<T$Xj{#me*~y4F%^j&N(e8C`+pw9^5VpF
zKJhp~q&Pt#>6Bo99L(9;k<1R_-50#|ldZpZ`Ae1`Ny|}grxsV$p7ri+eoE;>ITzL3
z(im?8J)y>20BEEngj)OSXN;I;A}6Ifl90A&huD0_!l{&&!lD0yQ{DQlX&;ztm8B4H
zJJ_@)SDej+4Q64NZM(kM$uqco%+k5{=-1jHK20bpF!a;Z^VQC;-xD=%E0W?ik|*2h
z0RR0g_NLSghP960WvN=80F<5N3HWJ3sjtylL<~^}%ncZ`Z4gr(B+MLx&g$cqkc%&1
ziaGAw{{XhG5kP3x>U0Rsm+ilUnVd^+UuiI*HptMpVW=Fkq^}PKc6W2>4Jj%XBBrA%
zZMuZ6c04$&%SFn08Alz01Q-x}zmDD~=Oxt%KahBTsjD$5kwS~s1^;StF{DQF;{Kih
zvO71SwBXi1Kdu%`>&vtsW$7k>3~24DtE=N7yIEQF6<9~N&dtDrYzFp-L-hC69@V(s
zFnaOlEL0R3kTK}sY#|z`4iihDtHgs;2;dZUxx_{}Niq@tin}ik8n5~2!`Z0B#BYFU
zL^kjmV;-O@0jl=G2&k{Pn4)4VZ!K2F9hkvf1vR-=_)|mC-&s4K=PVk}#x;kz)2FPI
zHQ8p2Zzica>}&kyxI4=Ak7IZlgK|b+`A#21(20Xp$0lOwu@`(Xq8E(+(v7u{z<3~Y
z^#}5%Y{Q#E#-tP;egE>+T26<P*Wy%@<|LdyGsI-Zo`W#_NS#;BI&b71NbH~(PY=xG
zIRrlpMvTOW^`AVw3nk2HkhbIdZ)Sf`)Z0VEKoT4~*dN+h0N_L%@LR_qsWA}WX5)fP
zJ^&e)Spoe?6O>KYBR=A}-W>cq{y>CSeuyyeAdo)`<UY$aNQzuPTi|NVmYdm!kp++M
z?!1t?fW&}%6h3J0e5jiv{$i#4F2=4EFD#_B4bV%U>8p@6vKkJn<&GQsX>k`Gb1a=N
z{^|~u0&B%kXhg&}zBKd=ySwbNE@eAH-(0mdHCq5!nl47P2EA1rAf|xgC<R)8Yy*GQ
zmtAXsIWI3V_eS8Ul~sr}-9LR7i9I?}$^EOy1$%hRXtP{#9X9zVzXrs}X&BBAx^kZ&
zN>Jb;Dg*5MhUB;?Otm>Fo1MG1|C{_I<@l;fXM-@>>@>#=`Rb=xbX>t>#Td+we@Bu(
z!arYPjT93y!td$7ZtSn+dS6FD_EoJ%q4u3gWe5-ijE(bZS8YhI8B^nRtfb|(AL=i{
zF|1@=j+_8FgKlI;s}Q{7I^eno{)&HqNf)6150)r(dZ04CI9A@#)3bz7!H^H^xBp4c
zfw2OpI|M+53u7n%Es`D_98~IaBkcfRD_s-t$J={)hP*h?)v$~@IGQ@;)||Fvj-O&p
z`e5_bURElgQyE^f>t<pSrv8Q*C!*Ju%Ri`ni|j@ioHCIIw5mAb!Chu$AA_+&Sc%vh
zv5+#EDe|S^^}06|68_^Mdiw{xOze4h>EHzJ>X&LPhPW=aEMf0D83f=M3imvF)XiQS
zpf~Cb)$sb=AGUDzF#EHih!t7&-S3SnJDXrG3E2ePArPy32F(EzW<(<&CQ^tr912!x
zs}nU;O}tR5)t5V6$^zML`vFaDhyDk`KS^Yn?@iTt-6+so?%2!<f2#rM{*@i0Eim?N
z6>)BoKWC~%9*sXq56+nHkT26jLr{<zN*KoiwS=3rLUJ$Y>byDn=)FG%ekzYHmP-3I
z!^K|?O}C?JzHgRuf)+-IP4DO1P=x%tCCIl0ndCh*?66gSXlU3C4pU9Ku1G)w3z1RN
zZ=IcwvA#IvARAW=Zjy+4pi3Yl1tPA2Fael79ScmRcU&KdWTOhyY)0DVf!Qn~G3LS~
zTwMjS18$`@+!8)B$Xnk?GKn9426}B79e%of>dLeBzewbm{qMppkF+k>waAcEcRf0Q
zZ%mGd@8}@tXh`~A5ns=+HoyCj*aM~q7!Z4h{hMfY&z&E2$D4`><^}oA+S;w_6Ek6B
zh|mq>?LPYA8y7EKas=$tPm;UL(y*{%XDgh1lH+O4Ekehb<3!x$W|!SG)DTnJqdyL)
z_T@zo&+XuaKOb<W%iV*J8%q{kNFhS$bS0;dfyy3#k!!*b!~+~GDujxOA&);SQU54g
zbMgXykl+<ME{-<BynssHoa90^aUq5{qyI20nHA(4LCMo>!+52A-pmoQN&3R^^w+O)
z5GorDz2Wv&-!0EmR~Eg#E)e+I8)kyc*P0GvOMYALzHAL^BKSn4=_~Kil=-#ntCb+P
z+(y|okjT-Rqqr5ZQtwZ0HW%<)pab<b)d3liN6csQ>FpZg>)5qkcfN(EbDj=bd!(c#
zU^H#<>*^xDhhwObvC<u8DBKcPwcTBAh;zw?pQ<Y^Ckx@hh{F)By6Ijegghr0^S`b=
zhM`p}@C{u8ZS(-X$pDzc0njQ=iHwZQ^seBZJVFNuxbs&6^J+c1!>?j&r!KctgQPf*
zL_t5rskq`?LB)txSQhgCkI1J4J#^lUGzc5lz5GahfmF^k`&C&p5;cO{M^31R&aCck
zAyJ-)q^st^^)IyPECqqKmq(m?2<o$o`-Ql)l1b=9-oa!MrK+x;a-`9WaNAp9cHkTs
zp$b8P{M+XVfDR^k?Z&Ma+T;`#z5~AV{l%s9!j~^`i%arqrSh$8%^nSqoacXaYPgqa
zEZjSsgTy)05j9b1iXq?V$sgI|vX48p)ez459Re|(b96Z4XMU`Z$jNY__k|bf3?iPd
zx_*3`4pK<EMPogIFdE<$@x5i{dUu$gUYR4FmfgDiuVMGEuJLQ(A};}h#c(w(G70cz
zbMR`Ftn*TEM|SA@Lgj4e)>bYA{(8vr2bp4sQIg}s$)>z`@sX~Zmw~7lja8dd<t>RN
z)XRF(Vh6=1*PIm>+7@RiKr|*oTk!J-{WD1X@}H>mlfpeh^T1csam3#s#XhHHU`&q5
zfo4)L?k;i_IfEVgac3|1p0_`dY>jL9bjYElu~v#P091@+Z4-$YQA!$fPImVFy`RBZ
zko^l@K<x{MAQPP5<1iRZVT{ko&d>pbYSSyN$+Dr9@b1L{aa_Q>YZ}&B{}v|Y<yw2}
zAyvpFL$V!l>{DSbXl!gOpL!b+U++mz{B6ffWVGP7GmN7>bU9l2g9)`FD;XD;;@`5P
z2ZGep;{M(3n$a^|XTr})en&*7dwetA`P*g}^}pt;0>MGxF9rD`jb;$#ymS^GR(ERZ
z=Dg;4I+76n+_yE<yvEYaD-gsG^g_9pji6;jE=)qI4i-PUz|lbClbC_Mp$X3vK!yNT
zL|bxFpkEb<Tp+l`Be>)eklI-6ifcm76F{!IVI~b)@OA6`luq(jJeUD?d*u~A=^%Yb
z&=&~KI))X46F*${ls`A8C6@^aR{_raHbu-vWToHYES@f}Duu9&wZ-e?vTp#Xf@hR?
z1Me$Z-A@HXb(tK6B6}YQ4vtR#{I2MRV&dS&9?Awbm;9^af_O!J)1$2}ljAj^?s{16
zKEGjyz`e)|;9gUcfH4T_gBrP}^Jq(gsM9`_Fd1TR{NqKP%w;;IV6*Y!&~kf-2<ZZl
zD**7qW#>2`wr)Y9_lG^}!#&H8DnO0Gf`P{#r~xY^vdwK+KlF_PF}4ck`Ei}N&@mHr
zIlu%m&Iw=w_~YN~4$)+}2uC89(Dz+0cFXxKt|6{?HJZZ2^1-sMbqq^nSmM{pVRmx1
zTP0(AJ^!Eou?aDpQB~f$1vih}bX_l#e5i@juF$h{`}mReAoC4SWZuQaS$;LAX4j4z
z`<4FDHo1t;YyCzn=p93OP3(C9kukq^SNDUONx_9Scv_u6zpK%=1&CY%a7DppiHmXR
zF`yn1Gha*V9epE)`k^F5i3X%Skt+P0s1UgnN#M0?vvuvLvEbGnx;22?=YB2Yr=n^M
zO}W0aBAdmskQO|R4njgE<LdG?<4e5MPC_dz6F9-#iNxbjb(3H4bipU{t3yK5&ny^J
zjQoGK7VEZ>s<O+4%P}ly5%`|i<zjS&XB%T6V3IcqV|u!}%%Cojd^_;t$Fih^Ly>9A
zdBjJ1!h~7r;X`t$a(b@mTC_)A*LV6R7`^vMPp=dDQ1abP7)BvBQ4Ug6fQ}$VHr_t_
zO$?*D5-J~s!T|Z*Y?A~54uMV}AMz8Wka;y--TRi7SqnQcn(Nqw4x%C%vxC-Dc#2O1
z^LMH$<*;i9l6{N^E1LHOKVyjLFOj-9UZ3HtR?9@Th-8gD=F=6i3KQ&Dr+Xhe&0GcT
zQ(!;*WA#HJ4G33~M=>I^-$ZXA2q6Ahf^nR7{7mFeG@m6&dWnESO_SkYsIJYSZkBut
zu<K<>LWeDI1&ZjAJW>J`E+^2XD+B<pd}%k}9h!j*VXb(iS8R3;$R&4WR-R2_;?mRi
zDK~}&2l+v^4_Oky6nJK99Q5rvOD+3{>%2VmcPD=3mUPd8LH8UuHl@LIB>#w_Xb4*2
zkdP3r0;4%g=1%gH<=mS|p<}yI-?IX!-uh*?uH>~U{YW>rJ{NJXkLDp1c*UEzguFRY
zo;XXU<z^s@sN}pUKTQV%s7z$Gfa~Rl3vGHQQ_a(ir2k!RJI)VO*}dyZ!~Fck`}xmF
z9~iDYb^vpvTXf<11+pZxLb>XisW6zX{$|VXR5NkBY9W=9&n>ae$HvAs{gw7wJ^McV
z^ejNoQ>v?PCGP-JgyK&fjGDnN(x&TeEi6S4>J1=bK6fI)_84rNLk{Xr*1$CA7#L6R
z7oy-%etrGuLQ#LckB?$W8;;O`c5&X90DTL&ptS04iS$8z{i)~<0GAPeXttU7L{(5h
z^0q?Ay_1lUNiFx^17;!;#0p?Fx|-Kzv8V64pSI8QO_L#)>WWNT`$)R#cKiQ-1OX{G
zuYLR$3@6{+aj6zDh>Z9a@XLT}{BiBc3XH0{!14S}&!>h~ZiLVVaItXc2lvG%hlfF6
z&T6W64;DdA|BBUXCf!n}FT%lqCupb!4uB{i1w0^nHX63xA8>-h1KXkSpM0<vu!o&M
zv2nvo?wFII`oArp8Esm9Y8)=QNR7l{@GCI_5cB0t3k~K#(0a(QNq0|O{H@#CJz7lG
zKHh6NdN0ReEV8o5Hfwz0)(o=)arLuWy6Hb;3gd(RW%3>x#8>QWIu(htvMImcU!%GW
zuesX8%NusaG^*zl&SHE$9@qA=(z=sho-x+egBSfv1TvEeLk3FaKS^LVBC}e>J0)IP
zE=ftx)^$bUuXbXh2zCN1|Ky`c)>2(`g%1GH9kDq0yg(C%#u*=0T(HrAAu=N43@5m)
zr{{O6p|W!LLi6iHq_=#-AoU3-az9C~uM;&*Bi>I)Vu5y)Jnsltt@a3MD23r92+%_I
z%U=Qy$5FjMe#k&K3$ssT=!|r$<gG(+9H4A8b;ZDMA~3bCBq^!f1rYs(|28qi&#*^C
zU*;DV5e2Grm85PMUiTnch#(Nz1Yz_-u@8A1jD`aTeqMcNm*aN3^MP2M&l#~)R`jt;
zX&<Xd!u|1$E5f6HFK~q0kFkF|<-?U2e?o@?u-zDkxj%zh=5EBBx_6u9gpujAKiR{~
z2BrBrlo<{ys^9N|pLg{uP-*s+=hZVM^c(@S5kVOG(;&TnkPeIeJ6Jsy5(f1o+=SoL
zty0Y<X?oW&ZcNG6bjo;^gzDoj&Xmbx|8o~&c};L(7)>2O&g-$#10(o^jglOMo^_W*
zl-Wc<JP~R8Y$_(qvlbj-0|EinXCF$de;W+|asq5L<@0A>Si>Q7VNhiNIgyIAVbe$F
z*8jN2@;#Ot4%SD-bheb+Jh$aS6i7tjGlKCD0A;orpmRL|P1MNY2Nb|!1mgVg=HZhE
zJJa@ORp!gRvtCM$rR7Ife))7R)|8ktn08fQ>THt&P|dFG?;ghppcWn)+9g`K4(W-%
zzO%pb53E7PBuC>{b%OzgZ(9CD@?0#+OlLels7o}qeW&WX?LFoXz)lJT%!&>_-qu&z
zn!>RjWJk|i^-iescHGeKi+=fH7G{gZsp78m222m4kx(lkdJe(#BIas<3A3)+>{j+Z
z=rlrl%o0Fnc8(2KI$Q&#>o=DAMe#=ZkYN!0IRHst{Qg{2R1^##nzv(_2H+qe9uGjQ
zggvbVV)X_?6a;>bD7}NO613W7Ut_~TS{d9U0@GYp#J|1JtV%q7Gnl=F$TV`K4^+x6
zyNo6gvIf(Bb;P+s8v9kST50!rRI6S={J6{Q)6Dbamw)>3Ks~Q{12#UiPPtgKuKD#~
zuoP%Ws|@sW<12W3&QqrVjtTvnkx?;;Yyq`aawP3ccyl9ZryWP*%ZE})qC4bfwzlTn
z)*TfaedTr+uBZcOSND1tJls}0pCdiUt^Q3j)9@03h~@mR{GHiQ9yKkkO9MheHlxkk
zKl>3>^fg21R`(Ww1M=OAW)yCMJ%XUiF**r>f;A50W4FL+zzJwoP95wSya>3MF+B{h
zS?3A!!YU7QUl4L;yQxJ+JSaR)wmQ&Gt%xk`34JiGKhC{0QEt<*H<CZcJNow=Mf{Xw
z;-#(G-5=Qj=`^a5T^EaZfj9)wM+R2|?c}S-ce=YqdQ9hjb~r_>lJW<`KQ928a+8AQ
zr+D_Ukbs-R5lw}`|6<S!k&W#8lmoGICa)RL8*BsY{cDY=riPOhsR!;#E-o(6Ec}!d
zJhuZQk58gbi?C5(M4l<wdR^%s{Q6zD%JDjMn^#r_A9+652NnTR_92z7;tI4#BSdtu
z7~y16nwv<>*{GR12}b}4h{7s9!;5zaZwB_RKS_Af0M^>$NW07MD+jyw)x<d>JM8Mt
zuqK7@@FFgY!@0!foe8a=`DbJ4w^$u&K80EP-#hQ7|CPX7pQI^|=3|byX4$}f+2?Yy
zFTp*6?2D~$S&azNTOH{V9}|n9OB7q4PIF9ae%5!hXP-Uwh3=ZnL>;ROONpo1&}GT!
z$2xsmoA>+|aDqtj4K;})yVgQ{T=J=8E5=PFC(EMUTv^Ls$pY)+(IZ!yAu?LT`WL`W
zWzcPzBBH-wuKn;EZde(j2j=qYM^NN1*lI2;Nu7#jBsiW2G)i&l6yx21jjg0p{k@sM
zyvoX0*tjI*<X4A4{}_TGdo4bB1SoP>`2!JD#m$}FqW)wzMVrTZNKZ>EQ0i`V&frD7
z8PGDl<uST>&y%|%)aLcWu$34q;sOqldoRmZqB|tV_=(U&8Dn}=Iun?h<nsj;IUV{x
zDU#)30?$?GN(GH1K?Duu2JwdWwDV`mlGiIPSZnF1b(9SUX(b?@nV51wgzXD(y<tg;
zYg}PPQ*)w(3?1GpdejOAG-WnmZ^Wp+B6PNLq*G4lCh9(36%|U@Ck)XKRwx9#3Uos*
zNs<0zB^IQ}k$`!cd0!c0cp&1Si6nGIp4yw1k;4E=l(0x}z(Wd903(~jwgIszfu3jD
zu?pyk7ygHkN1O+cjI=sQ@7>sC<n1j98GEYz+qVI)<!CcAGo<({fsYde6_6lmhyJ4h
z_}q_xAPQ<)8QJ&qOJq_2W*0qlQu$2a^5qc&p3LgZHxH)u#|t?e#)O`nc_gesCl|El
zQk&qTaULDm`F%CdF4JbTv_#7Tlc&PL?C?=(RV$9rT(D0YQC<OaB7Z8K!A%FL=`&wV
zyT>K2P?OFtlk~B0<E%?(C$7!Xc3HX9P!q4B0q762=5|E<hW3AliptRLn#eESC8c|v
z)u@*SHoI)HzZ00swQL^`F7a&cMl4<DN(ZpOM20M4SZxpCWD6`|-o?aZh$U2tK{<@L
zvRA?_L39F~1JS;%b0#y5uYrI^21C7p94f?o2Tm1$ck51e^*|Tnad82p<|Jr_mL=z7
z**Q2&p+8mnBpwIccgchK3zObQHkUxj9Tp-kffkBdV|Kr7bDtO(h`zquVAp8*Kq>pG
z&)LSTX5!y}Krwx-=k<G#ZhlsIu|bQ|0o`Iz{#eUcHG?LeW_%&)%7@5Ho$c-6QB^6h
z%7HhJnU8eCqqEgOTFCy4xuY9}y@+|EwsSQ8F;S%+NZE36A|STQvBCyH^XA5`l7;WT
zNv8qD2h?o>(M_ssDOax7uiiZ$`F)3tM}p8(z-Mb{yvW?)&K<xVFd_W9%ik9PkGIGw
zbr1}1U^--D<ZK{B0YvDvy6MlC52nz}<?0l@gh9105EQ<0MR<Y%DRN_1B%3SQxw*}M
z{m2^kIXyw%UF0{B4U2*Vt;h%^jPsz|($U>J8s&Et9*kVGKl?(`Oii;%e|+OYl@{P~
zu3L(IYq2rlezSio*Ba%n5$5O_zu_$U?KX$YP8O^N&klO;BUyI)2MU8f{&Ga9lsXpg
z7@b8%G_QsQW()ZxF&3wuA$nhePWM&e(Ux#2QaE0XW_2Q(+2odtkK;Od+1S|@DilvU
zecW=D+a&Y4gOn~7U^YYY7Bjy#BF#7AoCKDGcO4vx*6nA1K|MSTeyOkFc9_GkHDN$;
zqZNhnMJdSK)&P{{F}PtrhMo{9M?m`X6>KF}NN!{i)7)&^)c^UmL+JaszDmH1XIOe?
zs#S4!i{PvTRmGMgPRgd9HapG3Nsjo52a_yr+FOpVC0(EGIZ!)BXhLWx%*F%B#>>47
z)aY@d7cWPOMJI^yV0`CK(78AK`}tzSy_m~lg^&AA)c<=881N0#RI_omT2xBbeIqq9
z9T^MM=0ybB+HTAq^B5-QRx#!mnVFkMCM6XdeR#T8QGXH!6M6$^_XHtduGgP>12#(l
zH0N!BV?v>6TJQFI5}HKGYSPYDq>aB(F`g_&WCKzg-w<_XKruFe82L9?4$}<i4{MY|
z@2IM#W@<Un!&&avG_YGEp}JY(Qpj1#{BPkGBW~l-OaPmSM-(x`EDmA$F@*8WJBst=
zFAPG5^cHNaY}FpP(sjlQt;TlUkmh~~Y%a2*O+kBB^Xs}i)2vuU>AZF9&3Guw$jgCw
zW8Cl>N`A%#REaabi66bsrByjON&?1Ghc6%0?hjnPA7!G74k+WCWADOzE>LAWAA-66
zU7+S&UYY0R)4x*ge+o-P3VFQ;N{XzWSi4nBxYEMna0#Nq5BmFepkIFjG#tNB?&=_z
zz<@p3d94CNL_Zg{RAR8bL0C-&24S5G$#9QA18;ig`DKvPMVKQn{+$oH7H&L^_4}ut
z3on80Ti~7qA&Q=$+MvGo^TGMouU{kFRv<n*!lr!_Qu_v!FI&*6sTvs-h+$HBO`8c4
zl&uR=+;0AOH`77%h$S0a(03DrwQ(5u?!=_iFiLGrX+(Cl5SOc*ozBy`qB)aiX9cfH
zy$;0~%nn0YuH`kuD<7mYh87>#j6f$4b)8XeNMQZQt&HYwO_1V>9LvBW47g>}>59}n
z;tTDoFS=Z}UAN-aQGL=lyI!YU{)M7ch6Q&-fARW|58y_2m9?c{R^J<S{K`&zW3WxI
z7bY2(6F_$*#tS-}pOWLL4zTiwfCVFB9Skz<`T$x2RD@|?s)Bp1q3iESko5F~!9E(h
z=Bf^KxXuMraBhMU$b?!hqhlQg)iJkU;W>=iR6Kn6Fk)ja7K~w$k#&mv+d;4U@6}+5
zQZ*l`h<HH6in(zkP6=B}@R~=vUSLrwES-!F-~d=i%5c=h7mU}IsTdi;tSKsN#0*#+
z#H6k$QwX;?Hp~jVywNZBB{O|QeLc_S>GsW=_d*xu%V2GB=>gt45GRl%-uTN;TUXN2
z?<XjXoJ;*%`Dz_IgKY)Z2ZKW>1us#esZgXVBr?vXs38QeIuI*77qs7jVAYwtsy?)?
zv}FAv8h({5mp&rfy0+6>@cDxTv7r9Rg5nGP@?axUDV9aeYJSE3p@kSDK*uc>cXIdB
zhX?`j%dqzOBdeNKdyz`8Z>N@gA%c*QAls_k>q5ZA0KP{)!IB2$A3lEU1lid3%N7X@
zFwq5N#SBE0xPNmo9B@G-;v3S6D|IE-^kP<>)cU<CPJ|~Ungw{%vNKZ4f|z;B&TH+l
z%{HWYoeNwn4h_7#*gWxUUBQ}8i2)bzqRnDcF|7eS&-XvERMD_XV}iSq>UV2;B>L?n
zRu=Rmmpaq4F92bS{zcHE+cp)u8AB+ayNepEAG5H~Ko(9*OyHyj=PnDw`#UIN?}>58
zxUiK`Wqzku<=?9R%<!glCCyXoiUi^FKga25banelbVuLot5#R4j$KYNUTD*w2=>GN
zVnvc(p=Hh0spX>f1~>(P*-N>Q%P$Qixgz2Kl|6kDASeU-4pQa_Yd^3RMBKrlhnPn`
z;4RDysmx91oS-ol)mh>{^kbD+oIpbQ$h2V7YqUdu+;|p5>m?ZJnQtbgF~x<_PJeJ2
zweGCGqAHl3mDQ;JaAMi!*?B>4$#JP7CdpOZxR1mMfTo5^Vj^Bc{Y9;Yjs$r4t^3ZJ
zB}0v?W99n<uH2Z~J{?*B)(TkXCzAp;n`Ks59Y$Ey)5D&V446s@7x#fHW?FYu&zIcc
z)Hh7Cr^fTzRBU&x*1>q9Pxt@c%|G0{_!YA!EO+C%Kf^E18?AHpsn6w8+B|9vxocMW
zueeX+One#E1>98q;Xnbxq^%ft3|;uuDh(jP7UVR*T3iac%FarMIcJ9awL}=FA}q{k
zJ(973wuoy$NE)0l_k2t&lDJm?U2ik$ujPY<BWG3^5YT4>RKNSM=ab+|F0*|b_G7E@
z?nl~91lLArpW5LJ5ogGjY@!A}jRq8_+j!ET*Y7ZgA__%~T>niHEV~FX!J-A*SEmIh
za9W0Eh6_do3~C`*hkD~zg2wr1Ldb!Wo&0YEv@BNsn+6$T${af4WX4xp-4Au0a~*Q#
zs3BjY%p_2Gs}}MsFmzJ%HMMW!=db<EZew+G-x8;sMn2!Y@p|@;yK_=ls>@p0keY$P
zjf+VSX+R7MV#FdO=(BdUJC_jcQ3U-6`1{X*p2)QUEN;a6_y1PSphngM0+CO+ZRxl`
zPY>KjR!FXW0Yfx9%UZdDzygkP^>+~$3Mqz+8kc7Ag4tc29Kv|ZFCM$&CcQ;1R4-be
zduyUPf=PL%GLaodT)O0YG8V*Me;U$kjVecb0Ufl#`7=SGcVGfi(3+ZvE>10^_yzL5
zuW^kD8?oIH11~=ATH<jK>?aH`E$8w9>m^+L4bO?&I@=d_F}ExD4j(X&qmVn=7en}Y
z)nc#O(|f;HS#vH5<_WRE!^_pu+0i~rxAcF^Q${Xz6k_s#T?Pp1r?$3(URIHPkfQGZ
zmeS|2Kgl`J)DzN7SG-^-3@b})CvG97t5%+F%kJ7(R(5tW29p3O4*_rk?ie;KK+TVZ
z`4<VWxpt>~d;nc2Cl|1DIwfn*sg*mO`d&G?n7?+Dx+##fqI9w^WGy#$<ZeN(_*HkS
z5+|?x(>eMpd@m)IPAk1&$Mu;)xHrDx{de%W9PgOP%{HpFEXc?%zufl4X;zD(vi<S3
zU8O~Hv;_{o1!G-0$b-V0W%B~{2EBg9J4-B<T7*DhakPO|T2S?~7#X2Aq4*F*eeco>
zZ2kbgYn~E*$%E&hrqA@RH4oYz80<*#I5;$j!~d<fmX{r+m~*?dT-Vi9#wBFDBQzrJ
z8>8&tq6BWVg>3qJSU})|4a;xuJuwg<aS43v85b9aP?E{7$)Oe#X4BJI5|mk$Gv0%-
z8+c{5b#`9k;Zbp`v^0SQ1!Ql5HHs>)!)+5R%3Ante$JR;Nd`_#3a^=IKT4bpbH<K+
zT$7^Wq{<$oUz3O{IEFz+Pz03B9L`+%v{Le*X-3Y&ie}L6oKWNKvj_UiMdTuCKy)$)
zW_FrmrxHaRN|^eI#!)W_#ls){G1q>~d%=dON)T;_16m!XAW|tbed>%K`Sm_ZO#4+A
zpiwBZ3$^)$FIFoPFh1)rp*H7fUM#39;M*+bj3CT*SmLM4OqbIdTrJ0hfrUy|2XPZ%
zwy1!Uj!Q$kHe7YBs!E?5ZUZ9k1%{VNNeCMPpfXpXNnC{Wyh6=x2MRXEh?pM`aNGtw
zrCp-Nx4o2FfrLA^m9(LVyEIFacxhAWWpo;iMoCS#y5<T=VHUJ14U$yBWdLw`D1wSj
z6$_$t#pl%lWghsNKNY57U@v2t3UPDzZ;qHRoBpeVc&+6>@}I$ygKHFr3xs1SSG5v(
zDLk(yy(VK%Jy~x{Hy1Y44EB(JLx9C5_|dLiG|w>c-Q4IXkBY9Cajk(1=<~Y&Yz9iv
zS@23T1*N^}yW~tP2;>IVQecf(w60o;>^=Cny4{S_>Rxxfd6v}L_~29db%!wEKgw;a
z(I)m6eVHYA$F?&OO)fGr_T&XQivw}}>z6I&C^pTM_;~B6p%tl=tE!FV{Y&@R6_!~B
z_aOfDfn>zH<Z#@4L|o$zfu1Lgz(Z5*BRJCjHdoAF=hKVwiG2Zu+{l3NGv4XIvd19o
z9P+a@3ygHwFG%N`nz3t^-;Z-&^E1qpGR-c9a(14_+V@Tu{iP`ZXxH}dDWkogg<-(b
zdkN)sVvON)1vfUZZUdGo!|8tO3rPNeQu%77xgH;13nRW)eM%*3&oeu8eRnmIXWnCo
zn3I|F$y1nZ1ACWEt9&269#9jt-yC*(gg*V+*+C~@^o1rqTl0O*WBwtjh?SPTuncDV
zodYd+cJXK44AIMk{KOLv)z_c|Fh{MP-vZ>hd*=Eh^4E7i(ck*?BUZqM^&nO_L8nGf
zTojWDk8=%a(KhSU3BJhc@oRfTX|rpRSP4&mr`O?E&s;w3@RXp_9q-xvg~lVgdhOcX
zy{y70K-BVE4|4sN-v~?;zycJiJQ9F}_d#g{{>r8>uPG^A5BX)FadYt9g1xKg(xBU}
z_0Wx|CdFCTPlB|z)=Sf4E0s#mKqC+o19QLbj78%FBwu=)(KXhqGiRf+c*yE_lH+ha
zB4?;#_|(&$bhnP|y%JH^<v_=!&q$qYkFM_n=B$LD60DXAB~su@Cc;HC9x;FQ0(ZCD
zGA3fidxSEsN{h-sBO_yyDAQhijxoe!<F+V&mUA<1=P=j)NUuEjW4#Ph4ShjfogqB6
zG|>8q2B8ymJx~CqWNaJ+rD4#Cg3SF14kV4REgfX8^3wI{QGQhpGqu5t7dMlmJNCU6
zYg9#?9MCY211IiV!DDq(8^CB9HO+u0>0G~N<jEN|R5PSEqyy&?x9kzgF3zqB+o&cz
zNL!LUq#;fC0<YX3>*>~ila&ayN!&W1Ma!xsS~ED|h#9T6H!Iy+0<AIPE|hmRoJond
zwR&9WXZ`Tt=J@Qk!CJ%Lv9voi$Lo0=RRIUSYx0Lgqy1(BuJrxO`AT*9<CX^@F6!zR
zBs@E&h;u&q{Pj@2^^+N?%qG47F$qFv`h$<ryTSUtAoQTSet$z@#5|$9dj;=)a<;PN
zsB~wmE28k;2zx)x7xa(Mxpb`^?HFY`(NuTJlr}0xIp#>Lhs0ai89Xi&>I9dm@}(z+
z`9d((AVcrryoLi<6^Bjy@6sZH?TZQpnX9XfT7&>H1{I4lf79UVg(=*@%8c*b`BvlB
z4t;!HMMM{Ddv4>PV}YwRCzW6LOo>WrqoT9d@rNPBFP#0GLZE#B=DvmN{U+8RJHZG7
zsIzMoYvO9Msny{B(JguO?BJ9Vu#&~%JQZIjnmghpD4&=eiBDGjbnrZzQxDJ%U0?`i
zfG8_PRiYa?2o>}yn4mv=&;5K~u*OYzjVs4dd+@O0(R!Oy73cP$k{gPzmvsLXhzqUq
zPAkA-4^W4X3P1~zDbC>uPu4pEi3|Hc^G(e(H6~fa=!+du=s^6ShIu<Gbk%U+hJPVr
zl3Bi+N{j0?&c}q2UBvB}S-J&5ize)a79p?=1NQlMFOFw+K;ces?o;@mjowcZc>+YX
zzydI$|KO4UCPkcJ0zrZ)%|%#PqbR2+Iw87a+*hH^H!9PnCtyFf32g#?mYBNks!9;x
z4v=K^#<^SaB~z6+Ozc<PBfT%<wx=EXmKg~c$su~_H~KgUP~Xtf5MyxK2$dLRIGk|a
z;)0m9g1U9#(hiHIhcyoaDhFa9+UJg8|5sg>4LV*Wx2t>`vFj*PvqTuD-XbWO0N@4l
z8x?-Hy>s<k!snfQu)M#V@uOEigsgKoKQt%l<fzzkSy1jAvlpR@9=cd}#bcfvIfRp5
zdj@a_aj;-8+=K{ELVE;`ayLmNwpHccP@7mz{LXfJIH{o+-E<iPmOz}C8x^bmAY{fE
zOlzyS@b28GZ}70aE=QmgmGF3U#}`gtwCopNtc^%mNcx(ms|NQ$I?XBfRhElimze&j
zec`Vdn3J{xy`;9M#*C`j2vjSmBndYL^_$U$)p-LKgjpJ2LWLXz0w?gir%{Gy!x6<5
zUf!H?b;jF<VWN7X7oXwA1CQ~00W#gY$CJoi_}oo<oxhmh%-ojXFV$gQ+fBM3)wClM
zW+xhZD?=)KS@+lxvdj;4l?8C=1x=Y>B=V_6h=U$5*6R)yX`%U!a+v#SLOtlc!aZ7x
z^WoQqJdgWkNmR<5;~9mJuORMUE-{7w#y4vij2h8g^1f@Ay=l*DDAFlVRCpQTQQw?#
z))IU+Za9;bztwwCCimH+s?S_hRO20uCK1ZZmx`COQ08fhUj>j%$e4DXpoPm`NaF2N
z4;#6uT4sXw=B0Z>=Z_AAcXM&~V$jE=g|6Bq%gd6}`iO#7J)+2mML+9>fRup?1`^P6
zOWtC2%CTCp7tP`Gi3=gabUbe?z*PC25I%R%&)TV~fW7E|*LgK%aEwMUzzD^qq(?1F
zT4A^a)FntE+JIY+7(8bU;LC{2608U_7v$Zz)*oj%7<O2uL2~bV@w42e+4#3IQ)*Q!
z|J+I4NCx}#$EDj>b^~dN-e&ZvYYr_}Sbg@l$!Ptc7Wrdf#S1J!difB`kBE7|G@Uv8
zg_!72O!I&C2`s>W?3xz)MXd0qdKM6Fe?j<R(mE>+QJ4BB2?k9xL?0A3aaJGvQm7xM
zPI@L>L>eU!AA@B77L`XD#`0k_?4VKpjEV8zoo&;+Tf$>Arf!dk^=NaG7IV3NZeQR*
zSl$Rg3q(C11BI-*iQMutkSURL(#c(;v?z>k56Tn>8%%NK`3F-sb$XBPuFp;mV&yt`
z$GTic+{2%mi4DTWJQ!vG&o?nTT$s>I42gf7nfr{yV)&Rw_UPCT{dLP{&sf|;pJX*Z
zFeHTK&f^$dVUQA1_;zLy4bwHq183S;?ZujLyM)r+aFr)+r=JruYy+ih-@<tuwG7oX
zh7EFa`HSR|=#m9jj<O~%`h7zPy1QlTsAOJp5g_qH%<Vm3${FkTHKk=+Y0~fa_qaYf
zTDFX<eaX{$kKf{mqw}bzr-xNa%4pB>-XjEO2Yb*Z<ZQzXdXPT+o{5py&PDLdp(7RF
z@LgAp9EAS<O~2*Z;b@?z&vZ@p1mzQR{0G&p$M?^^NGJzNJ9w<xl&C)!YOJc1OlooG
zw*CB=xTMHw+UEJd6+18WxQy!jtfr=p-}TP>x)e3-*9Ri2P(Ek4q0*snF#YtN3JQP*
z*!R8ES6IGWC2aDSm@ZYU_oiwxJm^#)Iu7CjRF5g(KA(>YeL4t4wKeB<*CFxCH3pPW
zvEA+ZymoXyL2Z&z=GV%HE=^lEiNcFKdbbrg<H(4xIl1Xh-^#o*D<(~z0wy;Dc*oT$
zcRP+?<tYLn0nwG*5|Aa1Tep`9@qpD!9GIGb;Z_7(x0bRp3E;yq3p-NmQ+Hm~d?E_`
zorJ-zy}Hf5a8!vCU@})#xu3gU;sWYqX#XF|SsV*H>)V4uA3Qj$Kc0NrO?u0DZhXF!
z_FjwM#M`Uq40w5P>t1Vc5;FEF@@;hTMvLL;kN}_@%ibNnm7MAl>6NBCn?vbN8k;8e
zj(Lc_1pN&le8EL~zSlwB=oJJjvNW|WTKwFbzbWqo6}l)d4hjpQ#H4LlT>`RN!$zzg
z$sUjC;)n<g>l>MPOkKZzy)RtwHpY;L2eW|-E*6L|ct(D{vfI7JMc8c}fHzdT=u*O6
z&8YEZ`i@kE_tb(`9~O2hgtYvcWLqERHe854Uoh2Fo)k*^7Lclg5Agz$i7MOzYqHBR
zRhSNX;}VXxCBM2$hg|@QNf5^!2f|U9@>k5t#Nh-V=>Mq<;Z8K_m!q~`u3GPR4_z$!
zjxI6CooHLVexw_4u$6G%(U;11>S}w6w48DGT`4WlzqYS|2WlozC<`^!t#g%{$Z7sD
z$R4-ysG|0JTl;9ZnwaX3WfI<z%dwjJTKUqn-9&W->`ghj2KBq{;?66V);-n82|)*Q
z7M?C*I3s?t-4AcFP#LWG5tB>AnF3<9L|{|Hk1TbED*foeJvJUAt!Ej$0<;7g7NehL
zoqIn$c$OGSYbyu^Q*L+58XY;bB{kqR=S?Bp1?q#I+H9KZHm<*)4r3nMogC)TTP(Wf
zHWTkw-Mf<I+K`hng>%YWJ4(1zLAOOKZS>#Cj>0{)FsW7di;BHiJ7^dbBFzHg059>3
z3t)Jqp(cp4>uF3fGnE@j(v2^P9{_6WB$KVVT@N&~dh^$Xj6r7ybpLbI`73@+Qhl%(
z`?MzKI_N6hLE1Fkl8{L2;F4bxaW~{>mN$3ep0=7=Bajg@k2d=IKykYgQ1zm~c5sIk
zCNO-f?@|bXQGLrd;i)bpR|8usU5ri-8GRzWz=fUq-0|+>4tnRFwsSg0D@m~$vF-i#
zOS7&~pLbN6DYPhtf?#WeDsrKh4YE0;FFD{LrH-0RPp20&%H!VB81i_TP(7$Vk#%32
zJ=9G&SVs?cwU1tg?u61kX8Yc=?lX1PqUKO^$=5eyvRXpKr&VIt7}jF_mE@q?Hf{hg
zVNj;m>qqVJA<8Yel9KnqCi^0)2c=^AcU@UFOOj>bE~JRI_pwr7XGHEK*{<Kj%VQ#p
z*=jLNVfLX?%@LEXDRkZ$SxM7VEQ`GefD}M2EK6>e4nPxt=zt=Zj6geJW1la^9MO#6
zrsmt<SHW5qxN6l+MPbMmJK~zWuAz^C&E*}CN|IYis!=X0u))|Uo>AWJvm{*W#JA2`
zt^Yvotk({j4!N$~i!W`7IwTeZJ+12g)(K(rUw-R|I%B)}2iFNoE(>{+Zdq|zINbac
z0wyO+eJmOT5p{<s?N(zkd=w$CML0w1(dTE+P>56drQzD04Bm<JslRCd{EfZ%(G6#i
zhktY^IJRJ=2PpM)waQlA9c#--qP&N<!5`Us@lLbfwcb~Pq&*IEnx1@bIOdtGV{*%N
zg#d<E`OI45sJo})`TL@vG-NCPUinsV((ybi2F$A?b1&E+xYyTIhHZdQgHyue9Qau5
zU%0!}i89hG53e|{<}!Nttn*cNQ{P`)ZDj0lJ|oSsm)Z6XOzNy}J@9R|&rGqL!{dl|
z)7B9BoxeebX>n|AofnkbeE!f4b*`6Q!1Bbbb1BQ1cQLnh0n3cM3;yQyEakyr7x*}j
zKjH2Uc-aeW({_c9yOyW{w3lTN<pWuPPSno`htoH>`jx`FD_9?1yVA?4Aj7_i5(b|}
z94*v$17?*gm$5nb+?W|<-iW@Gocp6?Qb_A08MOQ|N~^Qwk;e<QU9^$*dLd>CO)5^-
zWctyZTn}Vn5xxH7WqGOD45<@#P}w>^cg0^5xFpElR4JC%kM)!L9XFA3;N{t|sD$UI
zOw3ArWb0@<`%+_+#+Z8TT*n4n`Ka<fc#cbkZt3e2I4dy%DjSMdQ)vlXm8zQT^zN12
zQtt7rDE>%nS*V-e+C#gRE8VRatefa=VyFo97Vd)8G;1u?HjK3siBa64O|Hb#;$2e}
zJ?Io?hRg?aeqKCQJ_xT_paLBQalgdNPc~Me8+2`g-z!Z&@0w}?lMfagU8A{-R~3lK
zobq%Mwsv?=BAtw>eV^G$tUFuX*6m`54Vz}Z*u9z!IgwvHtuu{dI-SQ8D_p-zd`|zi
z1vSj~4G*^gn7bVr&mwpqz~=oEe~i9I2E#h19*I`uy@0X*a8c9YH=3We0BeAsYu!AY
z(pWRq{C1_<B|b(m@J#l>o5s~bGR(`tV~%tu2faChYN@FcJYBW~u;I>jKxPGrUv2bS
zS8i+xRqa)CTCY6Jo=ZRGbUKmZRjEJbQ{=xg{b=Ggv#Rhkc~{6_34u$~V@FY%t^D6*
z*tFr2Sp@>d*_n#wJ-1}E$c25p*jEQfI%l+eM+6(Wee_Ng_=R0o8=h4pkaXi{nUouH
zgHO^M5#4ZWLGzbf^o!kV*S-EGz5a23vAp_93Gw082|vIt^Ef}rSJP8S%6^kc9jl#M
z+bg@KCRPB}C!00{ww*^Vw-RB!xGedCT>9=^bV2Pd4Io)qs0^IRVPAbdEG(?LiNfI~
z=*laL|B*p@RD}09F*7^~Q|=U9pN91v4WWU^g>6n%ftLdXhg=k=|KFd!;6}*0UVCKE
z2W>M`&K1k1jo;5&gXk&oZ9((F?MvF^lHA)zZ`V;bsd2;uA7y!7Ts*sM)=yhWvQMK!
z)P=31YI{w_bY-_?RqY*{)YoILV?leqcPU3_m?H*9{6G<)IoW(#4`wN`R<#~H^k$%A
zH%V|zz9pDS|HSmK?a8IBllH3<IM;g%{P^LP_v)pZ2lk$et#{*=pR72?m`t`!6_lNu
z*Zgz%8H=-K-n#S}mWwZBBffOzvw{cLZztFtzPTcyyZ0rO?1LK8$vy!K3<RW!cxeO5
zW0^#;xflEh5YC;{;rF+QbEiw-dkW}`MHXWpCy##3SkSW<6&S5%q4vHJ)u1GIi`U1E
zv+)@Z9gsD0fTUWc5q>k6Rzc9#U15B8joSBNKE^6EBimNHDV?5}%o)wkIb;}(^YhlK
zC)>u=LbQ|6?78K<=WF=-xLxRSln)Ncj~C(aVm+Gh`{cTXkNei-@29^4eJ@2XxYKOe
zWg6VTnGaw6F^iWE{PdR>afEScedYZ?L+QM_(KJh!+4f`X-)i*z!X95Aphh=X`JTc*
zl+W|pu^dnat((S>G1-7qf>75frtjgq8uAu%%7JQh85&(ODwNFm!;lp17y>UCNNP}0
z0Vu<n9Is%gvoB2%p;U-MrC9|8FhtGJut3}GX(KNDC13%*0h$mgIpLheufJ34cqmWb
z0%9FxfW#ZpemLE+s(O2@YQ}z}0L}aq&1~9aouxFapkEE%oWBJyphXN~DppzJyV7Ir
zH=5Qa$16RS!bj}}U6d9%?v!fDMOM3R6vBDxY)J^)7b+wy+1Zwj6|kDi65}N<2{C0i
zH=TZH%)w6zTJU!+2aj+YW!Pd*P0Ee`5;?ir6)nb6<WX<or9y6|NP|+8%;66ss7&5|
z`Yz1FX*wo$t*HPP2+tiVB)k{g(ZfXJ=X8>4&5UfjMCX3GH*9!7mKynjJ&G;KQvo2!
zYcg_zTfQ}iUrw{vd5&FkO60$54Gimtd5t+RYpg_B;+q(LalJL7N;RP~J6XG+_gEMd
zMrWYio&nm1FnF#|BWFtFXDU*sQj!E7)2o;xuBNBQTL^U!kVC3&BJ9?%`xZ$Y+`5Q_
zDVMb?A^gQ+X{;yjqe8la`;?jP&AN_;k{iwr5ET2K3PE(j+wUMK7k37qB+J_9zkN0~
z=}{c&Ve1bxjq)bv^0Rl0wM_jF1u|d0;EwFsiXD)hfVuLS5Wf+?r;73M1bBss)~tAT
zg({2jj`6!*zp#a?Yt$9}`RK8h13ijp$g2?QV0E9X$L;dLy!i`r-d*#}im{sVg4O8I
zls{O3R5{Kc70}AgF5-33E+PHyd%jqMr|Ye`(mupnZvAdFra#fg1Y1WAxaNJL6BB1s
ztrhD%q_3OHdT@Lk-YlZoXI)%*_2As_Z#H9o^D@uG_4il(gOspl`6ko>YWJ(^4YhyX
zayfQ|`}{>rRSaxtnh1WMJT{l(PKvhs(lK%G3ukBFUu04P0>KUwfj77^3ksHig7xo1
z>a&jqMB5^WznIN72ZIislAC+-t0+$GO(+*OK{}-9`Ht1ZOlm2tz+t1uFd%d(<^=oc
ztJVFI=P9~lT=<U$SF_1Xx$~#cTJJ0U_88og{L14r6!hy~X=K(y&>OR=*E0(JotUN;
zXYE5AyH;xpqZ*1|XA{OZX9_6T><;YYAH`n8_mIw7M}*EO1ui*x+wgbbDq9~s#bR_$
zC>CE}&rdY8++(yCu$RdCE^ErO_&|?IP!s$_p!iCTLvP`|`2!$y+pdm<x4u1mVnof3
z=GKyn#BTzPyU7z#;DbjI+2w})?SlX1p@g|`Y9~D3t4Ai2;wKhe;YtN)hR2g;<Te>U
z#HX)L8&4BVyN`zM7`8q$;r1^i*Un_EEhGLG>(jQkO!9-92pifMO0MN%o8ELgEF-RC
z5wEj-{FB7{JkF59WLO<&8+yD3$SLph?WYK4KV0(gDcm%bsk)!fOEL{LH8s0F3crOR
z+I!e2*K24iwjRhJ>T{33z6rLXwYKiW+9!WFAJ<HL=n@Ojafd0dnYd84Q(;`%`s|qb
zQ0oO}DY-&-qPLsbwclzwI^x};YR-+bu<6_xdQmJm#~nFvf*oix*o<If+WE~w1-|QB
z_!Bt(>>qcV^JxVfHl?~!`)DQ_9=$uKPcR<tk)odM9=!d`Azt}!phy2~3k(yPBz!M}
ziQq133Rem>##ht-tQEtTMzK|EaB)#d6}|N}oNH99cl!*UDT~9=iX@hK39nn$+(qjC
zAyg$aibqP;JcSfnE)8<V5{o^CX!FR5OHRzHTAe<Ge%`TC)+c;oQpDHX4el1Lzq)uw
zhTDu^dqDX42ko`9<tBfV*?Q><4cO|xCN|8A=&#ylnrnVDJk(1_9D`B~Xd>MJUvOb4
z55j@s<$pdy2%{UI?Q=kyega5AIuM6~6Z{6W!X`}8a{*U~pxrxp>bnrW=0DKj2Ulnj
zDF?VV3OL+0V6AsKR&QMPqD^ao##qb@izGTth4|VqKVvx+?+lHHplvdEczzfSZ7s?1
zu+z+|NzYSM(T=KAi?Ze!^JC@9p+c}gX3muM$*^Oy|2}E;rzf^Oa*x;IXKkpcfW)I6
zJZ@AzqhPC}1gz{M$c9YH>F86T@z**}hiFGTh%JiGIad%lEgpjc{`WS&5eg5DgmR<Y
zBh+iM2*S9oJP+smUk44Zp0IhN%7~kWRot+y#;E)38q|5wi2>+%+DGaoG3*k#E|)|>
zboesASrD_6`VSzcB2y0J1HRD=h>RgZdxu#x0T6CeOEs!*$|<t9ASxaRu?WdfkkMTN
zKa_(!cr4&u7k2h;4mhmG6sRuQnkD7Fh}S&&wjH3d-m7F4e<ONqD<r_mr|x8`PPDZ<
z3J?xH7ytISj%NJ6BNtdnRKIY%r+}4rx$Z!`)mwYt!$&pj+t|lFt()$&A7AA$2{^9(
z=AO<-l^c@*)JGPprmh;v8SlPL9P{hl8L?DlQY{#)K>=UgC#=S_QrAdy)?u}=#0I$1
z%iPgGTc8=<Ri8=`dhVn4ftD57=7mVs-7|rc3yPN9#{3nJYwl`3s)->QB6S!4$oB>u
z9-U>VFtNet45P!VR|DU_?eLxsX_(_$<Agt8KW@SRyTAnyw_E@|6~f}CbMC#3*jBwF
zxfKNdK>`5bd&9dVm{GI2=~#FA*K_ju$_>PY30NZ|pli8a*f$BO91CRoyWAIYuN9b3
zi`P66omxBb0|ooO)tyqQr5dxq0!>BM*ZE6hx>iGU#&7AV_*Esd9-PiQEJt_HN4DjS
zsh2XFjMd!r-=byQ)*rvxN%Q$bww7&=oSV(b;30$Snd(AnE5rQX4IyR~<nHJHkEgc`
ztFr65hUrESBqSB3rMpE+TImv`ySqa|S_Gs)TDrSIx&)*<H_fKOO?+!#_w#-pz=5Lv
zSm!!tj5+2QHPJa>CK{(u@U2<hSVBkCtik7~&!*{y(rj^xsMBgBVnu5_*b$+J!(5jt
z!ZQYM#`hh;7Uq_#mEbLk^GeO@$zWD*4tpl3WJ~t5GtQmyEl@M(F+9h79gGVs%Xn~l
zKFAgQ(<DxU=ySW46_-CE)=;;+cJ9eb*k5wy3WzGz1FB#OWrZoig>_d=aB?1Sor*b@
zAjSkN61YbRpon}1yINa7yAOB<0}Oc>>`Kf7#Sb_H0iRFO)62Lw%4UOj33$Cg08j*O
zXaFy<;h(Mg&%Zn?8yQgmI~G8{F~6vYvDlW<`>={XV+VI4F6yIgTlo5BTiW0dt3}Dc
zk=1dTy`qgKz-WDlYd}A8ZQYb#Z8aWGMoG{?siXn*4Ne=dg~_QiHMmRMhaRPs{3tK(
zY2BSoRlKq<qF|3>K6kR+))<K#<P-uGk|gIX7?ltVphnr;^7ouHIgR8Un|&rRbNX4j
z-4R_&?`_S&jIaBcNn)t6dx6l5j2w{N`B>w=;`m)}yU9;UU3qSvyT((qWeD>bK9UKh
zBt8Ka#;afqq&JdJkkVYD5f?vKj`G})rBT?je~#)R#UJcTroeeeFtk8nZthMGIW8}D
zKBg}0E=oO?6{-=bfr;Fuf<G%4T^16m**9Og7W(L)(t+dCIuVH<BXELV-bpuxx~}ES
zjNPKq@ie#-Jphl#>t#RVQM7)*3k<G8?rFqVS@)Ux=g}RaKY!G_5bSaM%$<nT$9X0;
zm>HF?PZ-9WOF>G#k#1UIe(3`Qe++x{4?t|)-Q9hGffV%OTQdZWT~KfcZkho;@d6;8
zHo#y6E?W%XlvwyD{ClvZ9|DHWDol#En*JYES(A#?nC|v9JK(nKhl>rd@PWFEo0}7G
z+>mBFY23^>@u3HoWDO8}g_f6>Wn9ax%Zo(zZJ1UI6O4Wej)_QwX)Q1DmfhAAeapxr
z^AvP;JP^~E7XIlmmNL>?=GmI!;qx+G*N2NflRmD|)ep0%U`ZY8Lv8`tA^F46m$0Cf
zuCIs49Xf-HI+&jf9UT#FUO?%i^}oKr2zv2*3E@sZm!#_L?pHPva9EM)<47XW`oRt(
z6LSVSt=PKmtQh?f9VqQnyHamHh$6S*&i&mDp$Mzx^Fj=IF_LcioN}3qUr(n!dsuSf
z<^qcxwK7XhhEz^fjiup+7OfBcX8FTxsp&CL`L(pQ&DUe4u9>~Q=@-j17lX|gvD;W#
z<y2I}g1DRJf1BhVm*;>yp;H4Sf`A{#Fle`YJI832fjcC;9|Z`OXspNYS94$w3*enO
zxw&7|OEjun_9ms-e5Hd8-{-3snwSKyf{kAaeun@+J38-pgmqk8T{*3FB8G;BlC_O&
zroz{`#qOuX@@s11!AjvJIPHGXt13OGL0SYZIR$#agNVWfuB-XEMTvh+$msmg-MWh%
zxn0O=97j=Zb**u?sw)d_pf_etX7S#o#g63{Pp!N*_<XMfY5VF3+J3#`oKw&f=UnA`
z9Ail6e{X8B1s(X(B3)Iqv{5rFzl{}{)PbmvS!#9h4n-V4n2-wfPAZq=C50yGVY>AC
z?>xN5r#G~dlCLPp+avEN(xM#KL^|oo+P(p=2W{eDsR1xkF4Eye?lEmuTqd24>`yuS
zczE$)jyokK$7&v#J@kZXrR$t8L#zbib!bdI40Jb#0S&3e_-tY2{@DFIl&>`~k@<WT
z#%I4W0tdo?{>2b%<yv&*svam2Ut*|o2@8(^4VVG=Cm4=A=>*2^aBCf~0J8z;Sw_pv
zu7Fz*3f#KDi$Sa;u>A~RYRWP)GCIDC7=ZNEf73|W-{cE(g9|p!&5^>z>)}@DK0fJy
z{>9GAJG0wx8@%Z`?oEXNS4)olmjsLg9=sM77U)C@%-+o+P~BC#Q~OSH%<$9&j%}I$
z!)PFzpTkAlY#!FLa_in4kvvV6QB24;+kb6$WUG`>(}ksx$m$(*qh1ul8v>I;6*?7g
z8eCa0-00RbNLx|w`-~IzWhFasD7zg)cNQXQR4u<=+D&7G09%~M`hSlh6Y`2t&Cu?a
z<g+(;8r%sjfk}u4;D^DRnPr`kl|>HOqK_DN9Q6+d$wCX{Z3yD{W=fw@tfM%XVvwX{
z{(YfaJ?KlOBy~mn!+T`T-^CHox2F7kj0ZuH#fCh^Hy)f1I(WnR=_>Xr9`-A4roatk
zjD^LE-4ZH^T&mfDd8%<WtKkj~E-nW*_ir#}q&u{>v@`?@cLM7zp66zO7_KNUkDad)
z{y(P%9Z=Ntg9q?ON3}?eAyKXrI646iIib|k`}=zoari^Cu(mdKaY+Dz1O%Zk!FWDb
zr_2euCf?lA!qNAEH3_6^5)ly*i?Y4LL%4pVMOk>x*RHNE=?+&U-lO|;$lS+IGO)<k
zU`KEZ5Y`n=Nb-+Nc!p_BDL2?)SE}9U^ckrG;}=68!>fjRe-AEU7W34nTUEPQx6-N?
z5>g*i7+VQf&xn}Qh7@mw53VhoyG5UG4KhwTnI?)@bsH;E?&%uOR7^Ba1y7V)jdxmb
z3m(VOI$NjL?cjV$Now)9l{pO4OXBr<8T1098&e+BL-W7aBq31D`hq4HwW|Zk{8hND
z;lnO!dH+LBDZ9W+sK%OgwElpjG<byI-;AVajGq00bx@SU_Rw3{cMNg7KkUav>)#l?
zBmcrDCD>5*cj#)uubDZw)HdbQ^brsp8uWjF_>Ub}!=2I#Xq^Z%AMYSFK#X?fG{$}8
zpgFjyP^IWYk=pLjktq;~QKjl|+Ll{bT>L2au}lLqC}3trL!(qH@B{`bZjdHY<-j)7
z2td|Bm@RYvA-C1rtD5ysjq19wACZ^CmmAKN*lWsRNL5>DhT*)G4Gg7x<?E85!zgWg
zZn@;*tdv1KUBE=T<h9y$Fv|rwVJr~LD$zZ7Z=fpd%<8VdQ7z;s4gJlROFESCE7UsQ
zorttHJcn{(_a;V8bG)(~?O0bw(rlZ`-(roYGh>Tz9vozlL+K6j+LAQeJO1*U7nKbl
zZxZ92bYU%%fP?OKltI;nwAD;tYfIjtLk&84;HMe<_YHP8jfq@Jni`8tOLL3|ec8Mm
z{*k!jl?xTK3H7*g;@BsBs~zZjJhA=do^##|J`j3FjH$RzD&S_3X{W|34_6K6U}^ci
z*)A0X=gE^XK@MgDcSk3ncfaY$8`T6BymAF+owHziI=Z?XeLw4@8Pzp+Q^~kKiM7BQ
zmgfDx)%-9u7FLh8C`q#!9UqQv_?sN5+d`~O#5czr&&Hc+)o1Hi!j#F@K&;AdeR|KK
zv_I!bxa8U5>W4x5%2Cnl?Nzd1XnoEX2lMR@t1)rt?y>w~GpR|sT*6V6b4}<sx;ER@
z@LWhCpBm`S>-DtTQHl@B2a;grPxemvz`D*UvNqE7MseekME{xHkT>rh@EfUN_6X~{
zmi8A&DB{TiD<(qA|9XbTPsmk96@|JE(SxMf?mT*s|2*v=7v?^2hF1w-RnmH8#$&&z
z@C@WXKaJS?KVhN?1#GC|T6gfD(~F*lBY>OpXNn7!dIL1Qv%nmlIN(L4n11by+fJkN
zgpTsu3eeB*xexQ~|04ETQI(kr&6FDpSLX}wo)m3EbQ^>5N!pg^DX%I&oqM=y@t9|+
zq+*PEQFr>8Uh2(vTQ0g1Ke{H$$8k>2VF|_C)OL-GPm{BH7zbpD?V&bU_J+7v2`3?R
zU$36rUv(`@boe?@EcqRz|Kb_I`MH_WfpBblLGxeIib+B&<bj`4@4>fbRYMx}H98ye
z=ocfX6%zYF=EH6_M*7L*{}(1>_?D4nM9(R6V{s_*wx|*M$lIX3!QC1x)^sO7f7=_)
zYCV=_L4Y4y`mD<_k)|HpGuP~L4#4qA4XxWO=k>he_%STX9^Hh%<OmyK4y-1=diAPC
zG%0OJG0&DIPS=`T?)2yjp0h=5Y~84>g>7k}g0Cr^CfBvqBy8b%URwGrQThCNoSc$s
zLh0CT^|`WemUsEh%c>WOCt3uC=@uVw<x14<$gWim1iq^_g)bpl1+6L0d~Sg3U<)NF
z>=r*beD*LH_kV1b(fo3!wG#fi;<s-iFHUJeviOTPE=-*)b9l|*R8ZEnDp79E|MzO4
zL8rktT9r7+vxbC$K@QxllsA%~${O&bJ{u--4*B0ZrNha0TgkKJ>hPY2?+E$o6?Mnh
z*Fy=|zgpgg)0!1u9F%Qsv2h>;CT3r;{xR0;??;gbo(fOgG{Ea?xWo){IE0JrOf7<l
zfAetIX9hHa_Jj|#$B}H98W?O2!u5=C6o<YHILv+~u^qJyrZeikTS(p0W|miEVso6V
zRlrXk`sLw$_H%1#@RZ57bm&c4O@~K#NYR8NOO0y=7Nz}EQMAV1KZViAA=()1@!Uv)
zrj-Gm#CFX+*mj<~GuPL6R<qoOBs@22))OI<N@J3U>^01ddMYNf@($?`<G!}eEcnaR
zdJYcdEk4T2o$1a3`0?27BI_^iN?~AU+N~9J#z?bCYSrGMs%(pr1kV}Jm2-oN9r6Dv
zcJqZny@1`f?my)o{A!!XS%|o~dKHeHv?u;{cg5*HG(2{eX2IPu3X6&?^=JGSEM;a*
z;0pNq&>)hL;RO_Nez*U}O53{!#p1-0a7z^}oA>;1BM`Wm&>@tX$fMya(pF-n?MUhA
zxnJU%>n$!IrXSpYKc+LON@yQDX`3?OB4u?Xx2@J)uu-osQo}4c_}X5j;xNx%z1%J(
zNQ?FT(hWk0ca{hF?WI74w!lej8h*KaIe}n_Iq%0tF*3GQU6ILI`&w4RN0yKBea#iX
z3RUrGcFV6@A)8_;e$nTs`b*$6f?gmD2mF`YHX}5VYO-8u)8D%wYa%R?wc#61oa(PF
zN9+G~{4LQ*PTux)6M-6b!X%DQ{8V*F&)v=r`e)^jq!*jcYB+~CS=&u-@KIKJtJS=O
zxSHws95^EgT;wn=)@LidV)fdW*H-YKL4aQ`vp&Z`u^ISZ(8m3)YJk|It(X7NLS4cz
zm;Ey>;A9U!Cj^G1VV)4%Hti-IYbcSB42)lo$}F6TD9?sYU425QI0vDwdeY)(q>7d(
zyRWWX$9U3rjN`O-r2$f{KbJRqe7n@tsO|M*={(IS--&Z|)>mm3Qt&;s)u(Und3U*<
z(tcX>IT>A^2R6e8QiW$n#Z-G8?mmz}#IqWwgE)-kn>Tebupo?OsVGE!q(d7}ka?~@
zBD9dVVSqPoP4BZOyeQzi6M<60!}TwI1ictFR_rOr(fh~=A1LOUJp9lC)L1|YGX{4l
zKY{6SO^I-+?LNTDt{G1uNsVPZYEO-2lkfD)@uwQUnEHC&%!iA@@!JJWu!;w^6Jq`e
z17L`kZ^@x2z2Zcx5*XG~f%eTA>?uMo3dJ%&pU3MZ8{{xD4?6vE-(zRQ(hsOWncvyT
zz4>l}`u-`XpI+niG<Vk@^OlKs#WuT7={_&S?6^u__-$xaeN=@p9XBRBVMOG_Nlp?!
zl9=J4pif4xoFG=tkWBx6M=!M@d6(t={F2+Do4Zm4v+uHr-AhUZ#jZ55zQ&Wl?y<E?
z;eh*|Ya8#zYncwppY0<oE?h0r!VSJiG%h48Upi|7tH2AEtlEz4Qw_2`iul)C76;lt
zi@PBhk(mYv{*&P9MSjn7x@`arVda`9mr9v%Rx4TBB$1i&|J6~+A4jWI<Ssi-Urqn=
z{#<BfyO$~$^-AG0R?^~RbwbwVk+CqA+5I!HPd}K>=6$(8?XeQouD5?Sq}lE?Hp`a<
z*J{V}SwmiOo&Dr<;=Q`bm34~)SINHZ<7@A-);~{`W>dzGVVu=-#)ZfE0GqxefG2sL
z#b*iUp6j_UyS4(ov3}B?n;Q>BL)`<TsFzVR8r&?|oSZ}c6h2?;@_7|hqV-CY1c<WB
z=ciq_tXkP(%9+YG&)5I*9hAChSws9dLeeOW(;=f`qPo_iA=cFKqGKW}L`@3{9*Pk@
zy)MV50)tDnRs!Y@*wcSF23tRE7jbauFVCLe8&Kn*%e<m{5Xu1|6?b0|Kik{KkqwG;
zW`NwiVM+o;SGcXv1dW{W(0HJH0PMe)brMyWI{7|~wL4si5Bh8k20z&sm;oe4Q;*$u
z97uYx_b`T$m@9Zx3{K%-5rU|(FbAx)e?l%UGsR0@m+lDY4Mw_nHos)*M<}vwc>G*Y
z_yE#$HpJ}t`xO;OgSBSF%1YBM>Mp(TI_BNV$^A-txqZ6_IIa=+Qsl}-0bdnp%|IzC
z?(ED3B%|QHtq;)9g@lRSy#!O;kGV-q82!NzVTgc7rc9lVl(i!G!q^KI8w8J>fpk`s
zd+X+Zo7Zu1am~|1uIAfF9*V8^4=2v6pGgi<lYZx@7ZOaqww%`bu&`hpUm~Qm^=q;h
zr9{>AFtbGZY<78vDrr&qX!Cq;h0JoPkyo&RIQ{bV^uX}}XAH*7El<&sAh)(6chk_0
zUYSdCOkT!<i-z|<ydT$;#Yo3Lw>DyVKh2rZ326L2I_*6gz{%YV<XDsb*)f6r6h-`3
z7wg(-X9s{Px^s+iJkb8k|96!?u3ZPRfz;6*WP|||*)bq*`$ibN+*uPKmrGLpJ}%W%
zT+3vH+saZbfJd58x`ks2TQb`3X(X=1y>fY4V(b%UAB))vbKeKhL^%Up-toOuJW9de
zz;-1)y+n=aG#w&rt~AR>efMzkDB7^<Cj{gV`ttfeffC;5G}1tiJ4grLY#829!%KXG
zZuZ}#63?VajT`riOB%rtw7@kp*6$)$&+m3SVR-vOIG)yN**7jmYoa=x+GS;2#_c|J
zlx4VVz=b(+nZb&6G?mt&=6PLAR6c_d<7;0<@rV64i0Sgj#QSmd>Byk8{rhdOssD9{
zvZ+G)vrGoRseh)Qet9a5i!0sigdWjN|FKIYNf+Glc>gvp%!SsYaY1Ow_T7>3BkuYG
zKeL<;Um_^xXETy?B`2FAQ}K|gtw2kayOB7TR|guWccv(f&pn7f*j*R}ONk>ap)5^<
zBksplU(;9{hhK2<N)lM4^!GW@`a1#-j>#uH|F72TEA?_##1BPtBarGqMkBkYAtqEE
z9pxcan<6--I20^~z)jmAF3Y~YzT7-Kqw}|Lxh&AyiWFL>%V%EI7M`9wKAeD+e<Pj#
zEcX@9|Dr_@Hmk+OB&y+x$Dq9aJ6+vH6ht9WQ6<n(p8ya+tHw-mxt|TF(Zu>58n*(Y
za5@YmX&D)7xg~~({S~}ij8+1%y?J?jHSPVx!@H%%cUsvWrNLTOPB72H=bT2-=bAtF
zgB+q&+w7#ORn7Gjj&@X<2R(CM6P0e-nqV%FCA|sJeSb&XK=QiMacoaRfdm3=mer3R
zTICNZ`F{OM_mv<l+_YAd;z{hXSUcpiCbx!?iCx?t#<I@WMv(`PSI}6Hnb?^hp53T`
zT!j(Dt|*EtOzcq!E;BJV8T0Z3=(nE62;o3yJI^Zz?bm;Ak<2|szynvAp1}Z7e*wU5
z@BXip?yEH|SFIa?qPOetNEi!2@6p-Kv_yk4o|{EX7*meQ&k0QSEX%xf0Zu@Z0*Vk;
zuzRRd#BB}%T!nKpwPCL}o*RT&1Dfq;L?A(#Uo^(VJZV4AIB@`j5x#%47~SdT-W}K8
z!m#;qw4W+PTjq12FM+@~{LaEf=k$Cpr*GP`yzk(R@^$;Qh(2s#mXnvaZ*A>GX9Abg
zvOK9&Ev!tfAMI9ZoJxa_jW&WaE|QsvvMc4aM)WHy#a1^y!!+t7>q*<8`5&Ik;m(eq
z)=R#tIedrmY?+_!hbJw)`K0=7gmBB{)Kifza+%ERlgCnUgvD79|I>}4Ev2!-&R^}I
zwVhd^UU3|dd>Wu^i<x#OZBu4ts@~|9QmF_oGZ}I#SzJ(LbS%Qj|C$xE0c5Rg7>CJN
z1N+lxMO)dUBK9mFqDCBDWVUoIpJyIS4tW#U0$sPl4!ZFA{M-Km#oQQUzOZrBvERQe
zmkL<ds)$9@o<y4Jc1KD<7|nx^&-d6uU6HKzA&gc23-(z3CP$6&O_(T~*F@wEP?{2@
zKhOuFIz2tkLNvEPx^P2i^>!>h^W4JoZ!T4DQpcSs01eD9Z-A!@6fja#n9Cb!64iqo
zgyDwm|0xTB`)wjcHF$DEYWz!$i6Ou7uzL6glpCWJewWN}L)RRJ4d78Ctn9n5YH4q8
z?%=>K!mF6u*9c*@lef4`^O*f2wni^=<#1fhW0CO}fgW$buGr;E#7Vzv4}ap&U%__K
zlsN7&;|=qVy_*TREd`5onYscGlX>*Dh#ExpwYL3zo>AVdfW+divPto)I!v%hOI@v4
z+Vmt=g<+e*0s$|f&@1d@zTW*}UF6kU^4tBbjWoWDh}^;N&E2Qs!HQkn$k_hvLD2C|
zYS1VG<e|od!h;fOYQyzgJ6iu+<qM2T%QAmUP|pztc~^C4ouRJ%lV+1ncJHe^gJzMp
z0m+5Uxp5lNzBh~SmD&H_{kO^aA<dW{#X+~^zV&wRaI|G)_og@6?`354P0eX)mXWpb
z&%Z-#-q+OA&TwJ%yGw&d9^eMVzkeBOH~>KW^mcT3kx`;qQP9xft~4&B3&$^kq@H~Z
zFrUm790w?ch3HKe-r4u0*1QQeS^S+~0CYO_;PN5xCLU~H%}q>v9NOlDW1;Zp46pY9
zwhv$OmCNr2NRLzhxa5|VF~8-Y6BX6&MsAB=&%d2~Non&SZrQlva3p@%(Q|7NS59g^
z$8US<gdvQs^}o6)w?*}|WlL1T6*6jyYyxa-G!v^;cwCz=Gx_GtBd!F?O_g51K~rlb
zFt*=3q?u0``b0$UrhdmCe}{;(Z{^@%<37vI8>f39lgwz7R;uF33k&X*{hDEUBfGX8
ztsnIvNg1I+=HUS|sPg9axx-5+)!OKse**?R-R9i>qCFqTnV3>yP}tmw*ds}Di>GP#
zyA8iX|IPPhXWg6eo$OG7t&>%?|HJ#Ep5On7VDx!c4huy~EIgJ@ql6!Sc+t)dQbMZE
zoQSA{+IS7{ZW17j09`sb7!#${Fd)f&I$|#BtTZm5%WJnZ(9nL#!VB1B^pF;i4?M@5
zUkAUYpoYipo!Tj-To8K9wu#w3`=91^bYx{_QlX)v7ipBZxVulSt{S*Mv2&CzpGjqK
z7kc;nOV~Kc_42So-AIIrYP%J`oY~)n`9?W%H4`Y!2geY{W)T5aPLH^*DKzDSn+5$t
zM}CqPmWI{e*PG7swsVjuhvInpvEw=t%JHq}UN_Dj-tvHY^@&@i9TQ!hSDy~PXVRCL
zC|Tu44~~^Z4Nf-N9l02JMaT3XrACya>6Wy85dxzXrYo_pR@1cfG;YoB8g6bD(B-Dj
zPT!)m65PDaC87Dd^<_>?9E?mrgE#Wjbgis&v(ibE|8y6v{H<j9A5t8sdnX~nZR^F&
zX8eH9Qpi|RfH~qT0K6f|Sp7gXjB`|z_;8usF7ES4BKc2P?tyW@zEy$T7{~1Uc%ddw
zUA1#%r2j!{79`8%0knCs54*89Tl8YG?(Wwg%No>}s(>fFfIoG4>zSM=i0fmwp61;^
zU<8a>Zs8lcK>{KtVWJl={Xlg%1%CaRrN@xzc|~=K0U&9!D2}M;=={n0<UH5-`axIH
z&?&h!qa_1-4$A>z<39@@aFKa=*)|V^{QEnO8zcL9D-1*Iy7;gRJh`(iRH$f$Qxjwe
z{d|trnemjxW#<?D&U`g{V?T=Of6r5=qTT+ef~D-qkl$G&t1lEW^HUf>>Z^FJZM{bC
zE7whfsqiRiV~i-ex2n@|ZMHPsH-dkjs^?A}1<t6s3^YDrJP3HJTlk&MeUqoKcZ8|L
z3r#mVo%T!3>uQxb6DKlCjZ;0nkdJ-s(uw&A@z=|u>aA>?z7rm|D{-)TG4^_6|K~9K
zFcV9=jA}+bRY<#okx@Vc0<4PKPWDKBlcO`sNc9znS^clvk(H4$@mlft^Cz$gVlc0f
z#!;@|NGawxWoK@{BYJE7MC8+$)GwXP0VnW^wDyM8K>*6~Sw5Gf=I#H1gU@SzG&xQ?
z{W)FdTbg!R-@8se9h!=>N18Gce46|H>_x=QLUvfJ+zv33`~iFjMq5ER5-aAh|IhSm
z=fQjxNCfr3RTqG_DCBNws_WXf9VqpmClm^sj*Yhgv}5J6-2F7n|0WFA(>92m1eEpT
z>s7J7ZzLonf(<bgO5c^ccVsrA_W!fAE!ocKFW56M#>o9jdvkbrvZK1_KZ;ftJatFp
zftH%69C>EhSJ-=Q*_SJoXmKg$)&iL-6_(3XCcIjZ$QBMMG<te?>*0u71!43V?^mVD
zG-r;TdSXhX`Jz9EJFbDjTeOmZuN&VG;}42pj?KHFkMFg^)Lh>8N`f!i^s}dagq+0c
z&oN7)NTpH`%5VjXl+dEkaut~L;-9x*g~+WKZ!bw33OjjfCweSvt<fcpV08{Z;dBJ|
zoGTPoEP|}t(g7P@UWZzI&G6q>+6YOZb&6E=iJTr%sq^tU$Ln-kGYAARZv=CM0ZOx$
z_TP-Nn`3P?U@`C!`3|$P7CgqC=yBskx*a#^gcl>3CguBl0m?5s)T*^^X`s@d4M)NW
z3x-AJo?Xrc%9FPxqrU%tO>or+PakQbUw8E0V-Mbw3fLfeL(%tD)9znw6Kob$?9+vR
z7f1zvl<nxEA8#IWWp(Ci+NEDQw;j7yY|0z_BLKY6)zb0PuQI|xfHpdE3xgvGKyITt
zba2TZ*5IE;5|;)o)@+OL|5xb00S0Dv0|$h`x!Jk7ugb95l7>6CWS2IM)cA?Z<M+21
zYF&puKI@&mjo2QTE8aAG)<Er(JX5@pPD>8`$nD81wHPcD9}(X;R5*BUsUkWmH?eh_
zkf1y$PteRnR6Tg~K2U?Vs8|?dL*I1VLjx)pE9(bk%!i8Li78d+JTJ9S(STZ^>X0K5
zPFQh^Gv-*)yem^3B>!!>Tpa!@m0fX1gNE64*zm1D^)>Uy_v$igt`@4T(u8D+Utg-B
zQDW=Ie_e8UC`ZiQ5HwYJYI|gvzNH?c7yGrhHoO7mmc$X(@9us~?P+mMs@+@7<-#w+
zlC7(}fJM)`ipbjI_si8Rm&6&Vnj}rDx#7Q#k3TdD*&ow~g4xv|_Wkl^_{TNZPBD<S
zf<A<uAVZM-QSV+(s?=!Bp|_IVjKjeWp0G+5N`LIu?1@P2&+kX8U%cUrDKHbbf2(ng
z>*~dvB>J_CMQdi4q)v-{&cAIMX3{chpVPtFjqdgk!KQb22vAo*j=BY;`60Ly1wKlR
zU_Yn~c0IkR;lLuam(9;{Wy~NJ2*zx}>x}b%6rjs)y>2_%2AVNKI~4%U_(9UPE0qq{
zPsGK|j!!A(xZIBqC#(}pxwWcQOeSNt&Sl6UlT_;oV2?Va#ln20NGc^6M~73kgu1BW
z@QX9bU_#h)Z}28FEUY^gwODbENqBrnhpqhml}HH5>>FlOIZq9>0?J-0I^xEm$ZPs;
z-nfR$i1^Y$0nVp3YBOdwM#Be;A+d?A(&F)F993LAeV-o-OQAfx1F3XAxE9>v&r~px
zhNhXetwQ34_$8(trHhHAT|%mUk!f;-C4~QCkm#*ZnT=THZ!6i@-{uc2+0-#hd?T!0
z6^;-?*p&Jv<j`6{Qkh$0MN{N2UvOQE-<^!8F2@0#=kmEi>R9yRTS*jZ17QTcC`E>)
zrQP22w@Xi2DZV|&4|<XQ>jJ^PXAJw1@G$J6ELuOoL2c>FeZ%h__{*x$g~7XgE47w8
z_xoohJ7xghXMfE1-yb5g5Mb^n=AB|R1YFr|6WI7DhGrDgBTE-O)9ix5Djm`Xs;8}y
zDf;o1?5J@QZw%@~E8JGbSF`+CtWy`we9PgZmDUsAlg&|kzw;>S7MDGh%+o0Ik;Q{n
zP+0SWg()x%ooyA7iR*y1bZi2UeU!tq9KHq#L^3l^d>}&ohnJY$6Ha~4%^d*pr;wbO
z_OGs%e!fWZiA1sZd;HL<ylAI}T&+2}Yl$N|M@1i@ADsat^ai>Nj$_|>+*WJ_S9!~Y
zm44?IRv&#kI#+vRwYm9W-d1^jVOwpf4<-U>)FRD&rvJvh3v08~OW;QryWyv+*)nxE
z3Rx@O6kqGvu!&O`$%k+zwFG6<SV!bnPosj_2}q@pdpuBxa1lk-O@A>cN<9y{A@CaH
zaZ9tu5dJF6?5Zd%KRd(GZ_Ra#ckDq{!7HuVBQkv)XuAo+!;KpqD&rxXRdAwyoyh#x
zjMM`WD%KemKpU<9$5G+AP|c0;|AHq{Ha4v57*;8pq4|p1ij|kC)4S5LWd44p`2~0A
zs*$dh{*zV`t*Kp9Oo0dTn^#*Tk})#vXm@N-%eN|gXD3wRS}r4UJA&%N2nmk|G3E8T
z6}LR0*n!S*%{h!!SuV5IMI%}}d{aCN<EzJav?t!b)nzWz?Y*Yb;Da9Oe}BTvE*h16
zHVB1xIAT*8uHmE)xMmuh*pbx$akxF`csS6OJ<47GbmEmXplA5e5#B=qVR45Ws9_F!
z6y#T*`herYj79v+Yt5hlAabr!RW$hm;}_b?M635aII7RnHMsg$|5cxDNWKYzK$zov
z$7}ip_OudPDCy0P5JP<VIr&5@ATxCRT=tvB_x#!NOpf8Sc0LN#(rh!e0-ouZi<19#
z@@$<b=sppCo*J6QO0-I}%9qR6%T{4gW!3zp!4#Z`rO%U>PnL}&QF8T%JG+WTq#M6?
za1~!-ekzJlv6m~#JG9!KA?fEKd2!~;iz<~_d6izYO;1Tu;q>2*CSJK!XFa)i#|wfN
zEoHz7&rk+!)~K1FA$g1_cpO0t@d2|D5TFE}G!ZrurG*lMS=5<}VSKYcYAcQhp@UT}
z39Z85cz`C7E@-R7lWEDA6WWnlDPi)olLUXXs1R1LeBg0*CsAq5YmKNACQub(#{36I
z4gR<B3;I9OI=@tegi=!f%SBJ4a_!^g(trNW<e0+9;!|}H-pLJ-zu!!}mWGvRag8tZ
z3w2&wTgG)OL`0PePFq+3IKgQ*NxQ>}!FGJ+MV>K@^Xl1;255|XtMN)!;lL)<HNQS+
zwxCU*l}Z+CcrlUyJb)cGYcf&$wBoSUN)Wf@LY(qb!xyc<1S#0j*Xf)5n%TKSyn){7
zGlp^b-jcLRMXPUwhzGszc;sunEII2L90^ZTMMV9|wT`_kGpbWRq@)Y?Iir+9dV~0$
zHuKjwi<o~hA(&OQ+~NH(c2$oHrHdWhnF*?LcKz}vVo5Ght`0o90{EQ&H)n$mEk-fQ
zbO1XJ$HLlk=1;CmF*f%edT}<N8L5U7X+ESD2NiRoWXe{&LJh<-rDmd)$j|%rFV(6h
zM+a$>Q5a8wNASb(p8v0FdiP^IbQy_5u`@!@sPV<o!H)6<QOZaD5&=agWi<UrMcKg)
z+StseYvX}<$RBe_USR}Pc0TW_mjn<6FO+Pp2n0&@K;%xLN}js26X|u2&-gHGT59wa
z+&jd13-a~3N^0H0JM9-X0JAd=>U!YpJ(_b1(<N*Bi8lF!!V*5WO{M~csFL6;a?U>6
z3AKgaou~Jx$M%cLstLA3k1m9gw4^@*F3GbiFv@m^(`XsrG4~GGGTckS2O!H^{*TxG
zpw(@<hl}CDCFATphjjrH>zu>$KM!sAkOb^ir!_!4_T&wQ9%25EC&{guI<4!t_G>^u
znJG7y(@sLG(Ia!*S|6svYr8UX+y)~hr!FK5j!GJiiq<5pP-E0AhJUYF9GrcLY*|;S
z8;En$&L`f!Qyh~xRHzewzjf`M>HS0{JZ&)vV-quIu1re<mGMyEV6ZrTN)Dp7;MFv%
z&ivA*!QA>wBNG{0yEQ5+8V%(~T%@$$Wop0FY_S^j>S-Fl!!1Lr8o~XV!=9;1kwu@P
zZS&o;Rd(^qvc@DDwNzx4v)7R^(WwMRqG2gYF72pCyiqGAlVXINY|5Wm!(R%@rmd~E
zHV1+CJ8zZO#N!G3UGvKr^_Gmyn9TE2wg1fFO>_j_$L^DGdx=LT+kZY-zk*jQFEApb
zFm9}Tz-A5+l;E>%E{R(ZSJ1s_MF%pNiIa*aqhA?y%-->6yb;BFZpgPv(LJN*PasQ|
z`HKQgAIAp^cD^l{Zrnx?*A!G~!M|_9dWv3Fpf}A>hgWveuS);r`{P^+5|`^RCjz_>
zE#HXr_8(3gYfOmLq-taK@50vaoXUD|vFo^ZtG8Q@fOsaz(ZZW?uFO*$UOVA>(e|>k
zXrQ2Ow_<W!qJPOG3N}xure@W)aw1tbfOO;*aC|qCn00*PIi#A|4c7$(ls@e?FCG99
z!aX+Dy;ouSIt|Qk-joANy>jn9+(XenQp(EIo6}%WSrER+km5{kS<6vHF=`bxB~YVJ
z8JPB$OyaIl0m+-tDo#B*@|BrLBCXkSLPZj$I-5ve+KoqIQmrfvzd@Vcv*FuCSsXqg
zIx!NG_vToIb$<nObjlXw^Q}}H6Xo<O^r{w%iGuiVOI=N9ZQj@}3-n%<j~yw$W{bJ&
zoiAYLey&CQE|ODn2;UQ*F{Jy~ho_0I9u}1q{vGoxTsqnlRkdt?r>jn@u%;g0W;3`g
zXL<B4_HJ4c@C6R=#Snf`S_{#tf{A~AFcYEhUqZa2UMu`xa&+u`Vza%Ig*e_F@uQG~
zFDn;g%J`X3%=)8|m<vhpX2wzHXs1xss=q!CH;>oH{^vqf9sK_U4i9a`Zv-3|9bR3L
zFpb$>1jW4k-?)1U^6TN%Yb6L0`&D$-LEWnFD(|$~zcS2jY^dapQt18f1-gQQZUh#6
z%pj~qX>W++4pJ#FsL{34=2s+ui;KeZ<&NIjConSe0qnn?Yw}gKbI?@mHzqZ~d0eP$
z9*A%N45&LX0=cz;BWfzDoveqwEDwm#>foP0A99ntyC)|n<>alIM_E{E>+7qkeQ7DJ
zXB<Z_$4yGf9M{$m`}&4a48L;ZegCTp`gf1auoKUs0G0wLLPIRKoi~5uYdi;q#FKNJ
z&YeXxs|yl`!tY#nHk$Vgl=Fj9f(GiQ$BRE4l%^-x=_^TG32sW8($l6d^GX{dP!8K0
z+c5FrNTdyvY1q&-=*e0qGFSaB(?AZgN@T~`=wA@4c^z$STE^^rq)bRStV4;t#N9-G
zB5zXRl;^-0L}%<zV2S54*u%gx$IaPpf49~t{!gaZBeTM%LSu5~S^1|W<HGpL#AoFn
zA`K8?k?lHX#y-vMqOyE;GlG=e3!S-xLf^2w$Oxk2hpL4vj1H5RNqRNdbZdlvLkdOw
zSqM7)V6;kHp{aAWfLUX6$qK8B9AW{n{+Yc;bj~2j<gZ<Rwo5Vh0*~~aQqlT<Iez}H
zQ2qBSm&n|4x|*}kBEa?WJaAY6|A7OFv(?aF@LO}G_VJ*iLU<(PD6MgGu)^-58WIf@
zbC2#uSMNsCGfuwCv#tZf6b@iJV0j@g@3}CQ*@x!`{UXg~JSG6ld_0e=4Tj;1?LeHQ
zK=@%y=wu6c_Z3uBR7}W=Wn0);8jdz!+RS+n9PrwhEQdyy>a_i+&zHA6+5HUWv?zw}
z@llE-FVazL1Z8HN^DTIVG}iOg`^(oxtYRtm<X@id)5cz=TN%p?2qyOH|GQJl;q?<9
z+!TS@DDV+i^Wu&KLs;c@J<LvmU)$@;(O0ku@Wz$KO;}=SS}%%HSSsS9faEDxXT_Yz
zOk^%?7h?4xUk4K-h{=#NCt7JZk|$+=r|4(hxH)C{Shy~@6W_;1^WEY%H=5cP&VMCC
zJD~$m`xwCG?O*@o4OC;@@o=Bu#v~1Iihn2z1>0V8WJ?%A@(fK%cJ$YE&PD1y>D;^|
zB`J~Dj4Med$C%M!Woh2b&|PUZGej`5<(m`ybDMcuH}~)$LU!{hmxKc~3)v1#3Cubz
zJ2R!(Hc&PhJ7dY(zCLFk2*1o;BR(82Fh@>$Y2?m2hw6hQH0<zmBGQsQOTteY<GFkQ
z31S5*D;SXxKQ->`K%_=H{%>3sf>hoWQ{MDr^V<B*xvB$SILqhBttXwr7(`3Y;MGar
zL-jO=dHH^zf{H(>&2mS@UeJ_V%ETE0r>2>X0$)t6*PxuAw+DGeBShC{=VQm}uIV`T
ziPvn)&HT7uADiE4WUlBee98z1iUf$7#P%1*_sW?0^KG{QxUewWe%1KBiG_vb?v(nG
z!`^3;90cggTI%cMe&Nt@pB*mvY{%|Gp5Kl!G1UTGg4>A1H=o+m0jIPkr!6B!y+pgt
zjHP6z)bMh@4*Y;({G7dyUBCRNt4VxAjO1vc#^9%T`EeEgdTFGFoi|I8tptU(@fX%V
z9u!rZ^)sCNyXXGtuo2bE;&KmqQerB6=Z>*&3111HoNcc+*|Fm!o++^7>DyIXBeTtC
z?ZcN6RNB-D*DYGG+RoExQmL9A68~(S*u!4C@A|HICX?js+Sr{Odnr78$wY|m!7~1}
zuLuLIfGh`3Lpqe2bYJi;Edx&kj5@AYoL^udN8g9M85hMcH?jMKDfDSVwZu&uzu+CR
zK@6KR+6szr{N}Y-ap4AiCoB-iCWV12nXK3EFsiMi$M5PC$0qIfwALw&0v@gp`3*nR
z_w4cb8A=ytOc})^&K1dwet4PjvBgC<F~cLa4>4@@i3%qD2paO8xsL#4T7o}gW%dMi
zn(_a}6b0iShnE_y1nyrbUgO2C!SD=6loHffltb^gyi>n8BSslsM!4Q>_K0m2Su1UZ
z&}XdfCb2_6tUU{1^9h7EoM4B<Vg>?qN}O-N*1%9GuJbk56*D^Z)~(nr%Hvg@CUxRS
z8<+y|gSno^S|Iw)MVmFstiCr3XmPzgFLKREzBdDAnCDDQW-nj+X|!}l!Y0z{^-=!c
z-__g0+6{(;X}tD^_o0fV+8voh%$k#v@Ij$uQBk-`QFuWCfndbfZv2gn>+c4|jB+oF
z6g2kgQopsAajo)}NsgHu2LIdMpwKHcn?DzZQW)xft2#dS6nIV{Z=_d6-a!lzh!<$*
zoj4F6;FHH7EcC|xEK=6>o6DiDih!>~y&l%Xkf&2e!>72`H(N9iZJmhRpfO*hG)C*B
z3g82m6s@Aefi)L>R<hh@aTV;OsuSz&T@PJCPpE}G%xA|wEw)@!?jGz3aN7N(^<*CS
z-iTRV%Oba;^Q+sq@ZJ88kUePpjT4ib-+I~U(PoC<+IeST06M(70{k->J`D;VVrHi-
zEIlHFn2tw>Eb?QN2O?dAwhnk~m-@ZjY1>I2cu-k)GLEo7)>esj4vYwIs%zjFFS8rs
z2?8*iO+9;O!z=?pLAPfVeRMM&fDa00y1qM1DQN^N4c?8HgR#wv7S8DHg1B0%yCh8?
zKK{<CQl8y5vN)Bit#^f@cidw^Sao}PnPVFaFoLF(gy8b%8E~6vAocIx7b^yoOKw0t
zH8^jcC1P&<@8gMn=Hu{qV4ywK%Fwxzp#2O)2UuhQNnUJw@5RNP%_AYY+J=VnG57X(
zzP3F@8Q*d+;Z3A*j{W}Q2eF_=nc;M5olB{HPekHTsaDhXM9LBigF5gVN<j2TS5YOL
zKUw)*%~lRkca*9=OINdU^ec0UNvm95-q|K7lf|vlOb$Qjm9Q-9yiAuYcczkAYsS=P
z%Un-W=WT&h_z#8~(Ho|>ujENLvp3AP8@nl$#TYqSyU^Lk!>T+KA)6W{0&!_cP3j<v
zTjBpccg*-}`{px`QpTAYXVRT3Ckfxg--R#LliQA;mia^l>s=d5VC~3@^~)Xi>tRo#
zVxmLmsO1Mcu&vZzr*&Wb>hnv=NhbQpJi_B$In8_0SKw+8tSCvT+@YY=TaeHmb&q%C
zluPm&3|kZi`Jov%NLO5Ol82(tbywYKi~?E$>}jf{@OOfku1==-syZZ{cpP4794dA2
z+MebFuKuogJm{Q}W_!<n93Jo(&Q^iQ+W+9|E$CpEOTvfC4PEUNGln2*lC`}6^#8t(
z<0Bz1c6Lsp%Mb0EEVk_zt!X1#x6=%lH_{ygAyJ4v`OuG355X;4w-lrYGTAprs`@Q!
z!N<pE?1go#V*BSfI5^O2&{p|?uGz%F!9ge1a~<1Y0JyaH0gbsl0dJAEE`Ia-Yq3An
z3EC|tw9C0|ce~*gTa4C98J7YLcFRMAnNSojZ*PcAMt=nGQ5f{udPTi{cMjPh83_vu
zgYVuqf|Vrl`$=M4#nI8)+Imph4OMS5Yt^_>TF=^>^z<o~E?35FGi&?-HD6H?W`|nN
z7K|a^W<z!Ugj=?pIHG7F{5rD6VL8Fhj>>LbugF1B%0|C=hdDl;Y}=!ttO6PVi6pEM
zczLi*TMMHol5{IS-{8&6#%}758|jN)r7m+#ZH=bUEle+Ph?c!3j~~#NAf08>ELLZl
z#+sH(acYvQ(+NM(#4L6bwR<LIA=ACT^hR%UVEJxoF;7JxEu4IH=1LST3O&2<>oY@q
zH1x1gv=dREwayHf`|;Amtn06iibC(DIT?@Bq$Fm`cr`moKaFxaG3EvjJx6^oC$!^O
zK3H+90QEphQ#Us>fF5DIdyLo1B$tF6yDyOXGh}TgCPTal-vaQs%?8+S%>qN-lm7v9
z%#nbXc@Gy_sz^V;_?oM;(xKg)#bn0V{7^siKIM}1XWc|(D!o?~QQ~TBMJ&9}!?=C^
za)cO9Nb_3fkS$Ix!qz~LYA(cf_g49^BMH#LUWk94#}R%Clh;-5d~pBeu4bN2E?woX
z*-mNjgy%Ev*||b;MDG=kN`ZKXg%~|2iAr`!6bISoWjry5aLk-V_MKabjz>eMzodum
zZU&==Fz>;00rttBvmIVHgL^X#K9;CB;SNgS0WQBg*>?Y4TfIgjyLa<qck6h!v;Oy*
z@Bv?qGejxEpa$%a6p9>KwgPSMoltVS<nvjV^;pq9-+Od^&t5_HO8eR*newFpWX&Zv
zHHhz+b7l|7-i-@yR{qU2obmae^C>ku+j$J@u5_b|LIb5cT2W(9VDJFbd<n`8JCPEN
z3PV$Gp<#_ulindWje7S=mpx8lG0Oqkna?SmWjyG-X?SS^Bi%N-=6$7wHp`BjtGr@g
zG)&K}=w*Ks6E<I+>MK3*4(Z>@V|OH~Rw1N{S7+&0;rES^T%z0H_hgy>Ct~^~7fnT0
zS9e7aNplf~MmW6k^+igoZrKO7_{Kymb^2)XV8oBl^1l;Hya^fnUFNp47NpvOgO(B9
z>H57{_HLR%>ZbxXnfg^sRFp9Hoa@a)Pm7!VVvVo&>>6~9;i_NxtoQqS&1;osO~<sQ
zRja;w9HZDZ@l8$kH#E*6+Leg9uD5E3VLOm&_J`aB+HMWS2fiz@TK{lG9T0dh=+PAC
z@O;@&sJ8VgtS`FFzIM%|KXA6+F}1Tj`Vj{AO=1E54}k+Fl>f78E(s6rQ0Ga2Q01zs
z-xCUpiyRcg)LKLTOal}>gmg-SdiEpFNG`h0hov^WG@VD<`X3@T7iljsx|i5zIY}z1
z>&tJ4T9j0dh7E#eD*B#ak?emC_<jaWp7-Cr8On9O!HCsy<?IlWtKuiQ)4W!)`!`MU
z{AhLDHr9`qx(&ML?DY^BhzM82rSjT+d3881(WJL?cRmgQ?w?yg@E|j3_9929qb~{Y
z(^!tu{?J`Lm75Uw(@W3Rl%koz34`3j3zYNB>sC+gYrPtPtBan+#X<pl!bE?}AL!{h
z04|+s*^v3Hl#Y(BcC(9sJn;2tGbTFV%ls{?t7Hc6Nh}Oxot?pI6Xq)0dXik#U#=z;
zTN<uFZrn&_Yi!FUwL>LNY`!eFORKL7Tc<H|mujNgX&Bmt8uQNM>wnphui7E`&R%Vx
zX0Uw6M>KML&=sLK-u|3$W<O%`y^N)+6Nhiw`{noK=K6M=)S86SA9YG7o)sqoAPy|;
zOzLcFEp5DZa71MlkQH)_VSVdXwz2T==gZp^hNzt5iS>0@tP546^U2H+9_60!A1-cw
z{NJ)mTk<rre~6{*-s_KrhAc{j>#`*Z&7(^0VD)F3&gFF*S8?HtN3&7(@7|Vly4j3w
z*;f%<yU(pA!@*gUB-c)X&Xm*b!<&EHpG}cxO#{AkN^~NhZa>_3{R%uFKr#Gh(;ZQO
zUHM2Siq4O%H62i=7D#zYd+X;!edQhXo?+c17nL6sCf}d7A&*{<YDrs(1j8nkWgnx|
zBtm-i*579jS+L3(a*fmy*TEI!pvPs-Ny3>8merV7&#P%!lv|;UX03N}!Y4Y1cWBv9
zu$W!U{}pB@<XEr0;LC8se0*qx@Y&pFExYrvwVbaa#wN=D;L@iwhD_U=IUhuXmhD+!
zMDO$&JTk9}I^fZl&3%In?q$!3xqKDlzgK~Shk!`^@Zbkm02dSj`NxUg6sQv1fu%P>
zKu}#8@+2wafkP1POO7(R2FF2NY>B&Cb)nJpwr2MQ@mg(Wl+2ptCQ{z$lqH%qHn&c~
z<LD4?51LD->8`%KN$zD0nUR7$4bHB#EX#wDuHL7$dmNW$Q*?{1+{<ZHC!S+fA(y?d
z@6YS^O7*V&loU4qa?BP-)ejnOq;$|F);bnzSl7ll5%3B0jSWP{x}&n;TPh}`#agR2
zY18n<HHHu1*65h>d8Vn+AH4BAz*?-eOVK*smpogz?qL9spc*>5OuuvkE~Vhv+WN9L
z^-7EJ;sa1Y`?&Ge$>)>8AL97B?3(2=jq=19p1;c55hFyz*wz7D0r4BdW`|+Blv|IR
z3q23$yjzbwUe91@e~5mf^W(VmH(VQ|9>?6&tMOuj7$P%Fa5!F~IA+RnWv=eME?Bg0
z_Y$VpcmG_YOXKd#{kKLu*h|L>=Qip&5`!n7{dd4N$lU~Ddc=K$!`aCO&)oCYlFfVD
z)86~e@q3^-w%WHjcfu$!cS5^7cS0Xlg~S9d(@&@txQ*iYRv%KsoQwJQxZ8JhPkcWl
z-WhxzNnvmm&owxFCykuVk?{Az9D?*2hg`#M%NlmPF_G)*v!6rmp$Dt7mA~!rZXcDY
ze@Om8%d9oqOks@kCGq2f)8hEeR>&;%RNV;C_`lJl+p(2ZCzXEUXH2a6brv}CNu^tm
zHg7-#c*8teKG_C_IuOz8rn?@oJBeHayHBjmyELM}v(dQ=^&B=kdoFC(yjMV|zf;eu
z{Z4mL<p-;7-%!l7&0KAYs<K#oypkyGj|Qyq>r)YG1j_l30d|wzL}`O=6(2lt#?lwM
z-OSr<=v1*2hsT`fA+Hh(wV7szWtZ=w^ym&X_4N@5BUkfy=Oay>bmB^}h=%#=hIw`#
zy<$d`f-Ok*-vp*%>0Z~Te5kOg4L64fo;%jj;k;uKPy9(x=vb1N{-><3#_>z%3=`yg
zeR!T?i0$`_5VZ|K*CD#a6gzAYi7jyxmy;Xlz>Ea!`q6Fnstk`}-v6h!iN!A~b~4>n
z%~I8JdMCn{x(X2Dsg^-L>1#bHE7`@tZ;xL|$fT5g0N)uuxVMvE`#Zo=49njW_`ClP
zN)CFlm6Bv)=cC_tcge@g;&~O%%9prL4*iRJ_Vj3rv{Mr+D+P1;b)cI=Bl;b|<H?=1
z=WvR0^7W<Lk1)@@41>)Mp_5$WTb&>OMkV{|-O;3{xR!bvvM#X74g587S0=Sji@n=>
z3mYEOqnT%>?`xom{q2ED4;$U)&~oQYKu7;~?A|eaIm@Q74rrm8=WyOGu!Ouz7G8x@
z6ybcMzW)B$2a`yg@oVimFWfZVtdW9IT`kajGB0XYfwy`=vn<DUPvr%n=iv2G&SzAt
zbBBZ14x4ft>A%WowWsd&K_%F1#FO&PRi%L;p~7iUm(B0kD8z}zcPgdT1X6tI6;b0<
z<7LW5?{vgQ5{EVS-r7a&Sz}D?sO*x~N<odItDR{1j9|IilV2LlY8zIj%#I~$03}x2
z3f&|7E+%j6^2Q>vZZOn`jBiCr<C_hFa4g2C$A=>_QYxQc5GXKLy>dO?U8pH6DXcL6
z>O;_=k?2&W-C)=2<fvrFQlzV})9mbYknb4D<x=EK{W^sip8Nm*k8aGd^dY_XA?4kE
zJ61D(u*XcLL4P8slMPQ8JEogoc^P}gURG3v@V{c?s90Qw%N^|7EvfOXzx7w1+7QU~
zYERk;jI}M-*G8|33NP^h=OVE``&3+~jljFY7rP{tSZ4UjyWu`ctnI+M^khz<22!a0
zOX?YD6)Z2)=6t8At6KkZmO!wB0{X$^bP2dxM*HmvfIJO3g`*5RJM;N5<`Sf#dal#z
zKps)d@?uzwuC}(eIw$uF5%Z$uiB@l_>w$pfFr&?i6BCHTs!WL+0;3ufbl-{7f7?0<
zHMzZJmeY<*iqMg9QAuV#Xfm21YSm{7j0Y%QY8kD@r<8rqt=APognV)Bvdx+TiKDl!
zy-s|xb96kTmMXJ~wB9MDp|#Cg)_^8YsM1j0$pi2p*|KUOwpo+JA-bn+vekUAF_3gi
z^vVE;Jgv2hHqA8s;)Q~BVjb7xOiCQNf9!<P#teV>GJlUX<OMnGK?bWhb0%;?Br$H{
z(|yLLYnztG_Li&T`Dr4wz%HyM6Py-i&D`#z4cCuL#E|V7f=;fdO*GD^|Ajz!TK@0D
zY{2L5n=kiJbws))zn=|zIb0{4n~5@Op2`wtpn684F}4}934O=MnHsZ0TWVpo4a>N8
z8vIOhB6y8euI0=GT6=3CB<t|_$Iy?<;hVjy$Ozw)rAQ{8ntWfv>=-&6|HvpaiAA|5
zI20QNt9RsLUfuF!_cKa^X~&Ps4~7TI+KcVsX6go8LH9`Z_E!1MjcN97w%d61di*62
zAeOLETvD#{f%7__&6hFj-`_%p#GQjh30t{*0{D<z86%~|$W$mzSu}BKToC9`h{eZD
zTU*-M<>c2T{+Q7a&j0FSfyaC0ElwllX!7Q-^7;-_ya*d%ORtWRW<hbQT1F$A(l#9d
zZ(L2L(@gR-r8kdCUHhxpvXM)=Qbx9Kk$bxK8#<dj1>y5rt<r|i=&hqwlMdf6c|5Vp
zb8S}<P~7U{oQ-SLu?}Ldt)6+>G|io;J1S{Mir;kZ2-ZZLCwb%Q921?=yQh^#t9?ze
zHW{NT7|_N4Kf2yJs>%NUANRJ!KxqL%iII}RXm1rkVnezHh?LY|jM0bzB1lLK1U3*U
zi4EzERJu`OBL+xL7~ReH;{Ex3<G$bLcaEO3zxZR<Ua#x<e58(*_cqqwdyznEecgS?
z8yMJd7#PUgI7QZ|16qs+m<#)E5S#YPGI1jCQEI_soG+nl>+=6OY5()#8ZxyrL=@{B
z@e8LzY&Ab>0jK>nYr1pftcwA04?(+r@p=BG**@iN7kpc=q?8yb4KQS3&d`Z^W)7dm
zVi{QNo^63jZV&d%1z&eUO4SAW)rM3H7mhjWYAL#Qn?Umm{B9l-xu@bMfB(%|7}Urd
zu=6GdyyuGR;2zsnR$2QFTyh?~9hw>zo`I8XnG4)=v!*HH)3ihia97p>^_c{qRE+(Q
z<<v?gxh4uQ;_+VfKN<%T8DIgcO`_Q=qmDV)<`XtttY!mvzgcK{dxHV`b$KU^``{HY
zqi+z(vpRswlNSK;pC1By`?wDS2iVoVs}dEToD2||(AQ?U;Pi*`H<CZL!PJZGPm_IJ
zFa?IT&&lg+GL5Mz^_#_hGnab%lj2g^%nGx$q52hJr^F0+^6OQ4*GHCpgdvN5jCw7p
z(j2UR3-va-GR<L1BRy-m<c)KbthwL`6d9S=Dvig;6G;Cy{waQz5AJd4?W@~Ye(6rP
z4*RTm-HD;mS!f^0?K`;gdp!mOg!2=>Vq&l-CvkD4a`EWa2=`bt8pkgcW-cOWUZ_X3
z3@NS}#pasRFlD`jI73k#uaTN;A1giLKc6h-eP>SRxs?dXUu$6H3gYITR+=VL4P(Du
zE48R^^2Nyo`^7Tq;k$aNY_oC)SEq?D=oG?>ne$LrXKK-6iqApbUC!JHkNtF<l2UKO
z+Xi()N!^FrY$eZgiRi6tBeNPkm~~!u0Nfu347KjZe~ksMMpmy2&USROF8l3-R#RUe
zKkw?55na{Nzw8wH2z`UA%;QX5U~j(t;XyMSo;b^Xv|o2zmA-sZH{Glb1opkd-?{QA
zbuB(;1)xm7q|-iK%fWiFSB&|IhzCquKJf`^C5knzyKK5;=}jzEEzR<BA4*sqH-lF-
zAxdipm*v+n>K^lR4!uQwS8J{6*O|_X&?9J(&ahVj${r8ehi&r}rAZ{wsZIh8T@RTl
zFh^+1=95O#lt}B<0aC`y<v_!iNmk2<aSYv1DZ9LQ^kKb%hQp>G-+c>hSRkhRJ1W=}
zD4j7=-?fDcJ6SlUWO+v-%~Q0e-|5;qj2oTKdjL4(+(w27)@+&Uh}YHMA+r;eb1<An
zsjFT-rn6~)E+Xfm=RK)=_g4O<E{)a=@BVy5zj;eu=_$@AZd?vQi~kT?MhX})%N?2Y
zQjYA<=36q*V|D6)<(m9w|M}l{?b(T7>ZexNtFq(Gis<j3$fqk#Ek?ea_;60~Kmrf4
zgZl|S%9z7NS-+wCb3NHi-gU_7{2Qr5-kRULPo1n$NDfG<)%8z(aRjkcy@<z#d(G!}
z@-O43bb2i19&XRn?Sqa>H**S1iK()6ZAP%3-(_gj6L~t~aX1;FInr)gI0ew*_X_<6
z7D85jo#_38PfV~^{a+@YcohNXxtrsL6ljM2=r{e&`u$_%N$hDUDIz)=$n*K7P`@3R
zoRS<CG5G^Gva=Kvn1-#|`sUPSMsgc$4Bzrn>UpGYom>ymt6=04n|4AwYfw}Ng||7O
zcaiw{Pwa2XB|2ntWy_`A;%o~Oy9*78o_PJlg6d0~olL#ej3S?0*GQ1d`>Vd+asFFR
zzY_N3SO~+PGroQ-?Xn-8rxc91r2R_%`mu2UiuOqEmKi#fXq}xpAJz)#B|Jbo@iP*w
znXDONoQtRLf>Nh0v)vlk)UR6AhJ9~`KmL3r!#d01g{@D^@0g`f|E;C*S~qIcnMc6M
zFnLXTv}a>2ylSRHA!*LBgWAWRE0DF<w``?79OB!{mUCKR|L58L=L1w})=n^6vbpgg
zS@6U!L%QI9Z9j#4G-%M!w^Bn(YNw<7I~HFi^_l5IbjIPmDh*$k7pCX$c04s9ju>*K
z53LWCK0nh~Z#{_|IOa`Rpx3uCF>`uQQMT^j9f(bxXA25CLPfrGCT=CBQR}Xa_3Uj$
z-t&KH5u-TdY(9zpb8~LW*@Av;R$44&(+poZFMN{jJkbyuy}Q2xm``JYFvo_L178|q
z$S$!cVAtbaXedWb{H2rX@)L;?#cRa%9vCbwA@O&aS<$5p!ROV0@aHPq&084`&)yfU
z(mAC1tK^g1?)2VxDT>JWIc;BCxi$cq^o!~@gk~Z1vM`#zfhr5Ur(sbhB<pU}v%3a@
zRIbE!D~ESD#cfg`U=L9dxB<IPv=mK@Ynjlr6o+1DS$-?|?y0(}f+k~Gj2B}W`>iGo
zGh7%Z**4==k~3v-RX&`#L*pUq+o&rKu3vv=_~tg?#L~kVeFc#bGalCmon|C@i#|j0
z;@dJ^d63A>_EjYqV~^H0Q+n@<jFx7#mMu?!DWRs~%$MZ}LCX`XzkZz(L6}d9x0pFn
z{L0$fP$o83#a+s2A}E2B4~nL}EDrv-k$>K}7whkRO0*T5?j_QFp3z`U&JqvWJMVUw
zmw0CRREeqz39AwY+l=$XmIPLQ6{@IG|CBiwH>63Kgz5+Zg72Oe;$WATOliIwkp+&H
z9-`xFr>+^R`F6V<q0_vTJF~YSh{(-GZx4k!*UztNpJAA@jJzy2&nN7Z93Q?1`sdL-
zK!s_`Mh#&m!8~dg2@TCrni2Y+5mm|qFkRjZ*lFON1$4H6Peeja;#m{s^esof3+Lw)
zdbP%XOJvwgMHmYk2>%r$8=oA&M90|k9BSu^ODd6gFvh}<7cq(RCG7Iy6XU9Uf0;4m
zBd#KOye%XbT1uRCeE}HTISKb_vsj+cG(S_SDK8#uk+isLdeu|!5vHp?Qg0$T(}Nc$
zxkNE{{RtB0Ltd6f3Oq5-u^h^^l>M8@@y&OD^)SGr2}P~A;YI7~*HzmBVEbjy;|Sda
zl==Xe|H$s7Uigy-@_E;;MO|Iyl<PfoUw-{oF=}BvXr*-m|5t%FDvltnF~trd3F{c<
zy3AA3;J9Lupd8JaH+=sdRp+I|+hDC($iEDB1-4&&b{^GH@rlkoYc4taMXhe9Np(%_
zUGry&qVGk_on;c!cA0s|D~)aAI|*KSf0s$#^X1o+Mq^Q4FKk|HYPrB_quiG2QWW>-
zS5`lywn>9A{+(kjg|`Ael1KM%U5nl<%CeDe{&T6&6%>OwQviqCjQ`D=Hi*37rV#GA
z1PBlUw43Gw!YGK0A(`F=-wtxu9_NWSjWxcr1{i#H`g2s8GPpJcPn<jk?QAx{8MAS>
zjh~3o^#h{+=eTWsY$GJXJ#tMm%5Yj<`W5F)FUxw#>rPfOcR(#9<rv5v!FNLp;1B9S
z#RGJ(9ASBRH%!FlWobl-JaGs)9aqhz>AO>#n3j~%`4cru&1U|!TB+PqBGcG>{icn>
z?_@#9_%M^JQJ6~vD7-k{HK7{iFlj$$uCG1Ao-!uWeP#5XhVwmZHelU0!ohHQxkeX-
z5x8P$ZtxY{3om}0B1Z*jahi<aIrm!1z;n=w>*gr9fAsL@j~ajOB}x_b`_rKD7o3Be
zMfdYRFqD1z9BJ=kp*2hzRv8_tXj+u083!}O{Bu2gBXMaxp)=&t9Ly#{(z$k*?BLgt
zSWPi_llVLLYHT<5R#-vVrOJcaFA)WXT^LeOHdV_LZnS&pKfW2^?VTh|idno;T3|-t
zDW&Dib)x-f60B@X8){kn+;XhY!sal*W&75_Xv6(rJ&_#I#oPVAepR2PN`~Xtt~=bd
zYB)5e7#V-ld0b{T12n@<0XE#^#O?1hC+Q+5;dIo4Y9Q&scV0*32TH)=y8zc_glXDI
ziq|Y6+MiRG33$55rW0Gr{iCV-9=#d@zeYy1vS6Tydkq;GobyXlPrbR{@@5$18P@+K
zof3%cH)8kS<QGFcN9XdvtwHeOx)nBY=GHKY&ElH`r{|Hy%=nk1rLMJ_Lt5twm|KTL
zE$(AWXib-3>Ico+qj~n;xbrT1q+Cre>Y!R!W@NZDyhI-8@13nQ>}=K?-KYk|5t<2Y
zF#k*SA0;#<&l`%kkB&Ux@p$w_knfr~IoBwvuB3Tnkt;-(6l>%uA*ZBJT|-@&2=$})
zTw0(h`4B29{P}tn_>PL7eff^#f99M8r*$=9@vT@WySLQP<=A%S|Fa<e=YyP6R92_q
zE4cqJ!=v6rOYhv=^KMQ0b>dRnf4**R@w<7|94;Z^TxABQ3k$P_%~PEmTmxn#xz24w
zTzz%TwA_~e1d#FB6pF^iZO;NLi5rk&u<W_~TR_>LEia<Tug&8GmkE2Fbhw68ZT=}k
z>gh>65EipStrhgF1HMnp2)xP_Y8j+_w9s=DQFqW$hgn}ga;6?u&w5u#-wO!7Oit?(
zPu-wP*zM|j8L(FKMCjyE1DEC$5CJezUCXME=!1Y)=X8RS0TXa#RkVbXRQvChi-&J{
zH|VeZ0?Rf2USWH{_QgO{rbeM{MAZpo*e+u~0sX3vzk}$z2UjwdB`ym&f!qzDz4i}D
zqdAkpb~AP|t@!UK>X=_tD<s(MWdc4B`}IvVN)>T%0m^ilEErxqJY_rUs+WS%`mF}d
zD()6We<%jyjKw;;Ws**jqW;LReR;wd@<7ztCW!`n^I7xh)yW~kU%HtBopoLCmc#SS
zyOpw^mA^j!`pNyXl1gpR*Q))Aw~9dXP0%*}(1y2xdFm*&Th-dkjcdtZ4hS;!6`PM2
z$Txa7e42dt&RlC2D&uTyi~oN>8sNPrAi6G$yd&UQv$I=2<3`BEE&3W)n?^^}UA0M&
z-Gj)beyWKtDD(=P>C#~L?$sB3tYa6iiHdLrV_OWJb*RHA2Sj)AKzuwHx6XFOT-|S1
zqsMZIL!mPWwQQ@tZ21D7QDY~hbqG?d=trDIemmGv3O{@lx>K@VI3eB>IA72sEhpD{
z($4|ZR?oh$yLn<HIX+ybUUl<nz@+V$>@UsQ)$J*{b{Lf)+BChkmZo~d2=L+Mko`cx
z-GtQ&z&ljEiF3oQmq8O*YgR%64j9v(0WI;I3-ix?eBzgm{<^lk(RW|2(S-Q$dhk{v
zmtK;i+as?<x~^&GQIXaW7y}vixSRjQs7;MQ@@z%!rOA7c!PtwI^gTV?iCSZR&~TE*
zq~8+FolE)w=g1RxS#Ge(HM?UB3YNk7xCjPeDAARrCzEc|>I+5!c*t~=15V04lryNi
zO(u7m9}Fg+<RF|)XD%DK{n-pq_zZl0?iTu&wahoN^?Q=SVP)i*irnuqUQgjbfm`XD
zCqfTzU7?STT9{u{S^FJxxbMBetq_p8{WJYR<fXQrMgUkE-K~(;_zmTEcMcnFn_e^+
z=-71n^r1q?C8ybI+duffM${*>7PAk9J(+*sw0Zanb^my!%)a!1CBj$gXa67Y^CD?7
z19_AtLLXh9Lqpm<xTTcU8!V#xeT`p49*29YKQc^o>Y9U9I!QZ-Swx0Dr<!nkZ+I7H
z+X)%oTHC~j4oMETEEWBAzWMyA7tR*V^Rroh9@2TICg|v(BH~7dhxYo(+ylP<u@nPg
zCkQCH97yK)4@Y)li0!u8uHz2)NIl*X5fzsysU%)?y%P9CPDBewIqoMHed*WRE-I@%
z=9AjCmRl`4b?VhA4b=w*|5xSb)=gGoK3OVvTh&#xM|$m)HdbW!Le<zfxza0wJ|3mt
z$b0aS4mPecjYB85ISXo1{4^+{At`CPiip{}#PCV8b$5$&ry0C{<bsh9`N?>8udD^P
z{A=2No3pNCWa;vN2EG3f@@-pJS(nDJoG%Bl#Cg=u4(G~FCQ#k;KV5D!cSsA1s~g%X
zFdC8@=+I@l0VsX)A3WoSD+yGXUHy9A@QcLli!Qigk|{fw2+fa=OCo3hMLDPK1XG^9
z2&fKtmRaF#zFm7rzdTN~>G}Tm@%7c7rX4+JsH?J&VjtfUfCX+&yWVwFycKhKlZJM-
zV0F_Wi1Qf#e{kLZT+6|XNj5^rJ6<oE*=qLI-NvWgOpi}X8@=SO^@e7Qzi(>)rJ*4+
z!*ymyb7HEq3ro*vHbrdt@-gO{=5?dL)k{l7?K)!7g1`UveZRUUxWiwinIVKWGu3sn
z4;wc?LRVZ8xNp7%i0Tc4ApI*p<tYt0T>gxON;8rrFAp0U8cq<L1t9PkI@Y5lUK4*J
zv%>({6-W=2#QVN0b-u@sjvLzfMNIoBYHT0Y^L}j@c2HI`jop8QZ<dW0vXA@6x@zQY
zu62l0fo|z0OrgKjNgU>hTXzC1v-<s!xqNS*j`#Q}8#Em-p*UhmLOR-Hg+xN-ryZKQ
zeAsBRG$Jl<s&r|Pxppkr6JAs+mxNAXtZwhP1T#N}O>EvR#W`f%HEk!@^N@25z4$#r
z2#x%Rj8=5MOfX)0SvI^NAd=xu{;glMeu1j}GG6D!bpW^efNmci^evB(47UE8Ef_&Z
zCJ6$F0NO<se%*c@(I;$Y!5ibukJqKceHXkKiI0kmUR+#M@vtJND${Rzp>DA+Epy`l
z%yw7p;EOiS8=+sJiw)G*pv<+7B!yF^Wu7{iBNY1+D$ABWbpG@0IuV@R|B7m$DP$KR
zPnf-bwB|4tUM}!Sf+0!q@+@zpIWk(YX_f7|&ZkVkwb$kLuuQb+l;JJJH>E3V344$N
zv)Xs9pX8JaK#{&(8;5#-1E%61K0bXu-=a_VOs!v7QWj3C0(QZ%-EQx-N#Gm*I^JPi
zJ;{8+Jps5(cVNn{^U^L$TEI?dffX?<D=XVz`pp;9U4u7>m!spR`m<1whsB9`j8i|!
zf56aqb({M@LVYiEu-r!AZ9~SDE0GckD%Ui1*8-C@k5my_u)CP<V#sW}eOW_0?ztB~
zoaec~6If2=C%IoKn1=P0DOEQRVl>#ra?-vYoK}z0FEuSDlk%wc6U~J}8j~Qv*w<XE
zKT{A1x>7P!9z)9g9Qnm;ZaC~YxHl;n(pV^DBv6G!M*jG75!V(45yo6~Q~>81lFS&J
z*AWdSzlZPPnF$XtKd&~JOgqt@^wHTpGRoz&65s(Fyeqz>-Z8WGBc}Ff5kToi6ug&*
z19#kw#f{*5tLL09lgv=<aarykDg;tW%yqiELtC*fV)$0me?kDkjvm^vM2S0C+D)5!
zbdXm^GHJYSL8dZjI_vvf<V$Im%PrcIp#6B9nTk;%9FJYq?#X#|F)sZ1nRq8d2>F@z
zg=ai3H#3cqTfGD7WlJZn$pFAWl00Aob{%d6RM6umhP%$Lls^(|MueH}!c|=N{!q1x
z4F60E@_mW7*<b!@e!Fk=#bz(Fj(>s|ZXafoyn5CN+hhJjVDkXdGmu9L;jrxu+U|$#
z1OB|+knyoV?lX^gWv$r@n+nm5=iKevK>#L+#A)*zW3@rrIlPF=>+XGK)6dx}j3SGH
zO<ukk`+?@m^^zwP8D=fM4TEQuP&I$=1U#euRI~onRReK~K*bM$)DZ&+F)WAHxEv}!
z&a|?OfdFlnmG71lqS%h%NwRmHql-o&?xdFFUwZo!<Q`{cr#}1c0s8h807jJvmoQVD
zOfb0E;=KUGJGH*OsE><?UZyE}?9{&X-%=VZbL%-Onwaold#t-=2gH~tV1c`lr^1K<
zSWJ>Se<%-Y&iElc(#uzHsQ*S#S2#;-@siDQK(_8#`ehEMz`d;d{JlAozWGl}!FyLt
zTRW*RN}CsFzD_UIY^r<KC9HLzs1;Ii6Wjd=w@cXhjm~26qUW40tT>Dhdbts)uRNL=
z@aGZt^A|7Py9Mo!^u5#udU{XN;(Yc;#=C`bFj6_o4)tZg+;Z%{atCT!c1}lV;o`4@
zD%V$AB)bEGH~pChNn6<@#~cKs>8vl;2_Ak@ytD(nBPGQP`6o=l%8AC+d$NA-nultq
zUaB>>8%-t#Xr_j`KI!f$D%)OE$*T4d6Uz<g>Zy)otkiHVlD-etSoiZWRbCJFC4_yC
zeC%u%$ho|I8Qm_NZ@%ROLX@oULx;f{)7gd{-8LWt(ToC#fn<sB0yz{zg>LLORDOI}
zT)9kT^OjFDPRK}t)fQgeGo2TY^(7{IRzc>5X@Xa9-y_8boCZLM{I>EuG{O4CV>m`O
zIwk22ULX3>`$76oE(1?F6(yzOk~&~Atv!QP^sB0@jN+8<zYsjggl9eGN5oYuxf=hh
z%%!vmqXo1!p$Uyx909BS2kPZtZ?&qib;AAszyHofnM$#OGLdLWklAWEtC%Bkuc#-R
zz4nHWOUC%ekS7^@_z&D2dAq)W$e`Jb8^jo1kI^_^Ci@F<8nDDwL0i6pl@)YQtB}@k
z^gL=~dwU(wh$j>V9nuW)2P#{D8n9IgRV9&JJu3tAo6ABTy#!JG$U!T+qjeMg%pBSG
zBvbVkfFP5&=)h}*3Y$fYjf=M1k&^M6Tiy^LG&_0~U^C!LRuO9^Lu1{2X{*}sm%nfu
zS0le2|1HqI4&)LevfM>2u&N?$hpL_m`xap|kh4WmvXMj%5BY<`yaEF&J(_Mf6FP*`
z+iC?=TFb1x#MPJXvx9!^N6K4k(c5}q=AS}n{otJaCwc@0Y$(TL-(9CT!Y;9_t!PP!
zva!>e`&ca8D$ar<r`VKZD$`>$&rW@**^vSaB-yS2F)wl<suc%f0MmuN`6w7ZSt>s#
zUwnrMpk_KZg!Qiq&*0OSQvJ7G$6s!zVn?<O+NYSMd{Zt`CY2~B$G^-&60JC@Ff(T$
zpu_JLcgO^k5KNQujS~$1wMNw6lgpN3^>|1=Q!3e9a7gkO*K@_WNj=zcd1jh19v{+J
zgg3xf^rhK%t!R@6|7hB6OO~m^bxF;>O>$^s4jrMbYh06yREsYEo*ECRB^ih$m#hW&
zpJE_L8_>`Qp*4ZPECi4p>v6bAMlP0-yYm}y^Y1xc9`YZRtgApxiIPw|=wU4v8^?64
zVaje3qt!Lr`tsCPRE0>^N?@AnNcn1}fO1$vD1FZmO$hi~v+~i?xyDv4HbKi$%^ik&
z`xl<3elcjbADNwYfqLZaIR9X-vJ?wT$-n27RmiOcTSK?a*-UBL_Tco4QscUDx><m)
zQXc4cAriaI0vwiHV^<)l(Ue`+t0jvm<w5tq$c1pow+&`W5e2;jIvZtHB_ErnfZPQ%
zpJB665WfiKsv$3%$&pVkvOuO4XYd2OnS6oKt#pn?9`F-~Ltw^;8zO*=b1h`d=p(PW
z(H2z#mSY0<%QsP|Tl33{H)UnRXyKH$ylC10zYe-tM%yw&Y2aXF2!zN*+|M=1{#G+~
zXjTD$Vzu1w|7*a=iID<BsO5Fr=*7D_zWvgu<5rQC8UbM6GrcGpd=LKSuK*cIwBNl#
z_#F+H^CLPS0uL%Tx=RFZtyn5zI=MnGIm5SMYHG?8g+h!6ZcD1DbOSl9J;2RZ4<rXr
zcSbR-j4!twU;6qe=HMx35j9KkY>!^tvzU5xFyB@<9UIkEBq+iqt2+(aBmf)VDJ4Mc
zLKU&vZP%@f%IZic@mo}Y${%WTc@HtlO%<p=#T(q@=EyKuTRm`@b>&gHlMnS0K`W~-
zn8l=|g{EkdrdNsb<-0d4eluq}TlUWOmEt!Hp^F~H*Te<<``TxGmI_m9yXNDtGW$gf
zk#+DmTW2QzUfK3bz*6p^F`_+KxqWq5^-n7^q1)G!DaUoHMO9->Hii(w2am0{dp=OC
z0lYs{6~gUYpIkS{HF8B4o9A+#y`#%ENVYCL+yhjXTOzMor44=p?pePVrFB8H7-j$E
zAwU=v_2b9uI|~(Y41HRY(1dcV_HR<YZmI@5YqmI2NXfoZ8X=%HP1<k1llw1fAy|PE
znRJ=4m)OjfVH~)n0~K>MJr-tsCk|xJ>_s{<UFHu{T$Qk6^ix0FV+U%qeKCLW*!(O~
ze`22sRZ|j<(-YkQ?T?fMWv0+uc=87M4D%!RlB4<4f5A38dRF5aM&wbu?{<_wr2<$5
z75`f7=-?pVw}%(6-50y{agPD~K<F{>PP>0Nr}Lj2N*Xy`DS7}f!jXv4)0lO!>HfZ+
zOq&6xyc2WqWKs!281;|=oMgwLHRc1q8`k=GMmQS1Hf7U)HjHM(N1adXvw1>`U#|zJ
z+HeRsA{{N(=FH5A7@eIxqCsJz{}Kh+NPChO(jb^np(}gASkVKV_6-tCJg31i9yK`v
zf~*bSbIn{gY;W?!BZN~THB|A?nZg=KWuzRI<MEvbYz#(IW;3_*4X-gKaR5~auShY0
zU5H2zrQ58xf*>#K*u)dJw>LOzX}`y6=r_0)l;jj7B#f@Uk=SMUP$-apU##7%!i+dp
zK>32f3WxcZN+Xh<D{9{b<Sp>W({%su3kgUudZYI?ZTvd<ZerbGe)mjJ!1}2DvuW#R
zXVj-;Snsqbg()@&$mBL-7KJ%@5u;Nu#p_KPlnuk&IjnTtcRxVa{8ZG<^oRNKu}*5;
zoD!)_6x&A*68)1OwZ3tDc)CudIIeG4r}zs>IL7>IE~)t^r8u5@bD?PYLr(8>rgC*{
zD}*7vc6t5eHUpy6W)7wR5HiB5WsO-x{;&WkH?vVP7|@=JDV;GGP7uq|%l!-qwhFhq
zo9Rdn8)$E%n5qBFvjw@nGzY8)ke=)kRE^xO`A33gR@k_9IZJO@YP*1^bEcE>xDgMv
z5uZ39kW+}eoftUY(bW+71yDVhKK9^2@Zx^ysxmb_mtiT_4FK}U#C^+BZCT(&&xgjK
z)Y0sRa`}4QP6-;(l41R*UZOP{!#nHb?^iLJGo?oISHiyhyg^(g7ZwGP%pN3tl`<0I
z=H$LPCf$#+btW%aj40ACD851CRJN$B+7?+bt!b!HiPDe~WvIIzS3z%|`2BA4f`eW@
z%q`{pzYopns~WQpu|q__Ix4?2cKO%dV?mT!#yG}i%d)e9m#R4l(yp*1km7E2XwKML
z*)vNWzR>m>H|#69#Ho||vrIr#D0;L(-KTx8WcSyU8+OXu`sFV+vE8;D!g?#T+oUYy
zsd}G}zINCBlvcK_X!@0Wf!a4(#%bD2?>-w?#F$rK{bt+7a198~+$3A|05((^9Qh}u
z_hCz(3}&-+U2eJl!_c=?z64Q=@a6qKPu@x~_7a9VP1IL}W$v^kcO|2G%kdkVhn`|^
zvqX%6aDKtj;C3;!tE(LqWju%Q9m=0PtmSnK%t;v9rZj%jJm*%<6&;&1dH*s-ct$W)
zgQKTeCdwevVNz{%@$0+z$oVJ3m`>EzP2&~zw_%=mk0HV(iouuyz=Czc3Z6tgG{<tj
z`|ClOAQAwIZ(n@$MWWNGE)p2y+DSsr+6FiydisrRWp|(B+Wja`uO;rKWe%a}=nE!c
zb1(Cgw0~E=DYp#qz<n^!R~kLLWWbZCv8AK_8z4?uD)a~MFW!S6Y+I=QYfO0fI^;I8
zZ-}U6<!+a|IHMDEdSGS1XMX#&UF6HRvzyAFa@4!+#g+`%TwGlo&m)o1MpGn)k~Wxm
zoTB|Ci~mL=XKvM(4A;n1K~HDQQ=zSG@3jin`zgoBW;^xm;FkCP@2;&ESJnLh1W=fA
z7l@L&r$;~ExVlo-AVAApu22dO@?Ajc>s~DDBP{U`WygO3Jl}xrDD6K72{0Wcc`Dz^
zH$tAw2JzC}&pUE+O5lfT$;l~_W;WWkaf2F7?p=KNaM}myuslkhZN(Q8UmxL=4>ZMC
zW1*DhP3%x*YFt8BcUl%aa2&FvP~w43WGsR|2UkFd3}lUo?9=jbt;5gs%XH(z&2HH}
zlnr05VvtQqNzoYeRKq9vxWxXB<S`yohy}-#E3gnkD($8^Tf1`HHS7u{uRkuAk_Sq7
zB=YsjAA&We&Ien^<GCl7xiqbr41P_RJRQ5QB=Ifq8*ODP^$d2ulK$30u7SwR&eP}x
zj&1`Didl)f03*RKcAXSg-Upei?|;Y&3Lh~W5#oGgZH^m>b7c(sFN)@B9Pz37AFqQ3
z8aZA}>ChLTXVUv^*bmQlb<k3QM-|$LDw>0ZfJ1&_MZ%o#zGIe)!~Pl|Frwj+>phQN
zdu!~<h$JdAHCTt#u1VApCuO<FwT??Cme74{-6W3w2yD&d;V0@La9jeyiklsuV!P5V
zx}GwGn&ALu=d;$Q>?bJ{Ga#UxY^ZrcU|-`_>8u*t6g%2YI|j@%82?IAd6vtdCxZh9
zCFXE{EkM>~xHGm~Rz!?9viMeIE}LPVTA5V;^>SfIeRizTjDA_W{!Y6F4Y=l#d{Fk3
zyf>dv^YMqaAiSOoAKv^IMmaGJ<hAYy>rv*WyOm?|=vI2hMekZyvs}(Omoy;|LfVmI
zZV%mts%Y_$1=PvC<uTpD38PGKkVnZb5u_5$Otd!3z+@vB!#K#oq3Jlo@;X23y66zD
z;7?9vX7Wl>5`@FnAKe$OEsj(<f6H<BW`+xvsm*gX<kCr~#3i<4wJAU`gCdp*4uC_6
z6@8F#Qoc^AMwZ9uB9F(vq<SESA8nL!2bCXId+$=X9ufU6*LpJoX%=q1n3^$PzYUnN
zu+BRX&$#`ixLaL){LzsTt5>w7sp8g4kMYT0JZ=We6a~4wlDX@;P&*G1UFA^D^W!;u
zkYr3bgdEw7Ta;FyzEu(<-D6eZd$OecqjXIE&c{*Snt&ZZ1G^_J&R=kzFPQ)^i4|Ff
zRc2GZTtLOjN#DvzmC<t0@jfMRdlV>n`l9QiGIOO{jtVovnYQ%Wi3rM|f9f*j__t}8
zfF1K;lC(^dMcG~(+?6f?lynh$PM|XVI4=$YjarQ(5>g)u-DfWHvUdeNJKW*4R1V8$
zN~J2UfmPH+J&^1EQ%vyUYq+b{Oha^b*ev+EGy-M01UgMtE*Er|Qn1fJCCoxX@VQB?
zU0H4)00obvZLbl>e%UNZ%w}Quiv-`5`T!m<Adqbkw!Y4wW{UGjXf!;lI$=`QD}PUk
z;CgbFYP-7w)vtMZwP?Vi4=U2;`h|dL8=KpRg#!UJl>^<$*0T)W`1yF+(tn>HWp-g?
zdihV+pyPbSWNu_PvwgMSW&Y)x-|I`Pqyw?*E3Gy{2#YJSh<TxWw`jGA+5Gscem8)e
zGyhm1K6j_xrst<Mp5lJ(zR9z7WYBO`B6W0?`C9ZAVl8}gN7K|)c1R>mjUCKPY#tkt
zCiP4WgI8%HmRDoa84UY8muw%BYgA)!4wjXkLV$n%ug|of)?SN4ElY<N{@h|y15(qX
zs#D~aWfPl2V0$u*#4^Upp+f@0P4&-yh&9<8#qxjUWA4F4%?=CHhOl8Y`AZ|PfnsV)
ziF=KwYb|xgt|XlL$B(PFxX$GxpX;Mrrn$-Crg4cS=S<BpK^(+~vcMeifw?2>Ygf;Y
zt|-ZDK|hGt>wGML#8)xZvH)cB^{g<Mi2K{m$%eugOxejV&d~7yJ88yS!3C~cY!ols
z<D>D#@fw?!C&%BPU>YIzla;9UzN9d7ommnCv1$EQmo^2^bj&@W<n|K!2YdKhbg7>X
zn$?ao9{mFxKWiQH%U#1MLNe;-W1$&z_ps9Mw93_MZjVQV*tkMk9Q(NXE5FO7^5OTH
z6j%zgRNu+(=j>&8^yMoURSbVJcrONkV*!vDzy;9A?&Nwer@*A;7)Y%CuynwFy!I(&
z*3Y_RQr~}hntu6E%<z^+b9-d48kw2cr!i&U!bC5FskS{=H#gG4>NVbUH^<`210}K#
zLrhASY8SV(tU~?w@2Ac*WN^+c=2QiJ)Xe<Z4m9kX0%09!AUNW&AW$G&B-dp>lZUZI
zIRLc*XaImT&ncK3D&5tbF=^jRYD7^7g(=7?AS490;4hpw4G}?Oh3&+T>Z5^>_{k(f
zRO3K%I1HEJGv=D`sjJPx;U(USvC{FDM{FY?rBZ{VJp9H~$4D#$vy=_zAAv4Q&ObDq
z1j^dWAIkGxoB||p9~p_dO!hp3`oS-hx#Fwdoxf1NbxWZd^=BofoA<&NH@kRr^l25@
zNE&|-HmO?ESIXCXc@SXd+koe_Z@rd8kWhXtf@=Tg2z~K##5)0`^}~}e9#TGcOELKV
z_V)McYYio@%hd{zYK4xeT=L^cL{6RN3<U&{1NqOl0b+g{Nw6ohu`8^Js4#I);#DD2
zt+esr;Uww%Q4AN}#GA#+Zvi><v)B`AEN^i-!*|ZYifZLG{v*5`UEpl0_BkWh$``lH
zO2?D&KuFVPPO>dsBbIC&{&vRND8XzHF~IS{8E)qp&>6m>(Mt&HZ!&q{{3_;`WYor1
zpYNZG`}ADa&wLtlMZFrohL=pi{bazb<c<M`e1hK4dbKDbpH8NA)mNyyDzdLKHK%|A
znFa((ae&gsb(<o?r(xy5g<syC^Mi#sXO$#&uWsu37D${2oT1^vf{+e1M2LqiolHmA
zM-)H)R~dC;{$WUQN<BmsZo8&I;j^Y)w93$+j?H^=;YY17F)5>DRqOctr(O<I=dERM
z7SQ>!NA#zAf*&w#qC-QyuX^6{d$vsqU8c!X25cfUYB#5~PJ`&kW*)N7=WUmM@gZ_V
zkV&@pr3sM8MW!t+hJVsYXD8T7Zrh4wJ_OgW+0-+;`A>uc4{6&|A79I2rR-RgiHvT!
zhu^m?OND<S7&r=(m@Afw+E$S@2R7(tT&RerKx!L&MA7p-q{C>v?F0PxO<GSX?VL{g
z3D;LQyv4XQH`Fl8%<*EaI@ytODPNq6k^5t2Df3fJdm&a;_#65Kwm*}jREq{haYFdo
zpj^zaI8E5Ta(g4XGkM|{VvFm~a~%iPnlq2$MgF3DyFB7F#TVlo;+)`x3Yz<DlsACY
zO)(N~`2tSyPieF6sgP;TyX_HZW#{a!SH1)@w<7kW@%b<A7Pr^}q@JrsvTPLDnf%Pk
zIo6$?R|El(T9-{AXgnw3)T1y%gs{e|kaQK|Z3N>FXG&lc1ppB)uOm=KSWKZo_bfG;
z;H8)pma=drrS)oq1PigK`sy3Gbn{4kqL(Z@Q>MA31kv#z%irDo_m74PTjOB&KR;hj
z937>n|9EX3+q@I;#TF+sB|9W*ie57i0`9VrTtF+z`pFE?A5N0?gv{L(L)MvVEkr*C
z!Tte@zPT;x!IJzxf^W%R^F4Kkwtzs**t*kIP=TfRr*bVnTshL5LwcfUz3xCwi^8KK
zjH|laJH5MbbCEav^jH9iaknlyfW0gAK3V)WzeIA-us`eg@>o-l=UDVw0Mw24jR&Pr
zSpBXQ65AXjnOnNK(i=5=sRR|Nc9ZS$nQQ3K-5Ex!A9fr%4n8j~#^;&l7nDKsBh_1)
z=av?!zbD{Z-l#rBi6@ydU0Hf&A22B)wD%_G7;SOrLT&BwFTJHW%^7FeWuvT1&(H(n
ztPGV)ofN*i3y_b5A_64qTcYNOs4J~=6*W8U*Iu%gTmY31$MC`RLOA(zM_guckY>;q
za61Dr4up`2NukKVAA4B-NyZGK7^<@c-BIOt<r5ALJyqbM9u03=uTE+L%MmvJ5#-rD
zW*zkb!9Qa3(F<Hk9w*!8C_VRJ3GLPE>x!1QcRP%<CrQbX=30Z$gnsNofHaz*&+kUn
zn$*PpryuM8K7#tw)<0TdpFVbPMo9ixBP$`>Vp%OvSvR?Y_){xcUcfgL1Pc69xrh?N
zXooNE)Jemh$$-nMlPuhZ`5r?}#$3GC=%pAguBxRgmVRp)@@~6kfVf4fb}f4~qHoGy
z-D@T4!g1(0YWY<*zl$U7bKClR`=n;bd?~N?9hu8qMgdSCOgU7u37Oy>y@}LYr{k)g
zr#J1b4z8VenIeW_B!2Q#&d8YJl=BR}<*zt5zr=s>bbO9r(;B0L7(T_PYyxKZXxd&H
zG0F!QHo@*zgnctaNRfNmo~7h%Du*PJxb|}gH*HL5^3upYc85A62IH(=Fd9%QVT;$8
z*b1h#^W%X@n8nt`fY5F021HI7XpgI%OE=Ge4nZcHncH-yJHMkCDnU@gsu5?hppLeK
zdU|=iEFF+<#7s1-mw-;byOaJ)l`hZvC-~k^@A34&{g-5-KQ*f8NBS?#ve@@V+gjB|
z4Bo<FibGpwfb)pmTNc&6&W}j1B{2VcP5E}p<h>Lt#uH!XSY~>(#flTc*6ePFb(*jC
z4~c%1;T~rot4zFMAtF^+y|+XPw!ii=>5)Kc$_gEi4BjHD^)3VX^)Z)?3#bw0*HXU$
zxngs2VX}3O!_GG^wjY#wi{Gb>w)FUIuhEXq?6gl!m!lbl3k59d_O}7P48g!8GG-v^
z;Yvp~ng)g0#$g)R4E%)stD{@C(i<q>=3)Y-E>$oXh+HvJ(_NMSZgly=gYPk8Dqh@w
zCV!`M?VS$=au~B>Lznl7?9<w<0B3n;G!1S{*TbtdIvb1=@pg3rgD(IDO;HuQwVKra
z;oUOJ<UvDCT}@zJnZsPP6ptyk#1vx%5hYZJ@TW!J&$ki06;eaY6}rMOeR&R|9ys5m
zd*CkBDPhqUg4Y~Emo6>9rZQbi`f=FAl=qXiP`H$UynK{gWPV?vZicI!CtHqN13^tb
zDcZS@a03vLLcW_wynNyF;)Qt-^~VK&MMZD!B_(&c_w`b;A+#2&rpIk&T9YKk3V_?r
zot-3UjdrU|*DnN%p~@lU*tYTyjsIk~AAGW6iSG=-{M>whKLFgUny<25UZR;(5LA`>
zuHu<??gR}JsxAf~Md!p2rUXaIS;Yu_UA^$6gR37RvMT^oSnzGs3n$BZVh8D4E=rz?
zGijO0ht6(f-NKMaKf?7cpBho~8<ZXHIQCqgO-*m$<m9%eJ)sXaG&37E8L2K^f>A09
z?|?UCQ(tq|;U)0)MyrzT@;F(L|EN5m81DnEu{Pk3pkrQ3h6m#<2P%)=9vXbP-lNav
z%Elp%%&dsE*uPesvwnSQCY<?eZCbX0p%N=O->|5F!fuNWht1F#WqAlura4W+Wuaru
zZD!0l02Gj`ALT%<l=ckl(P6A^^QAb=s?HAg<J<X>hGs+zppB1GFZ%7t4+kK*{_@Jf
zY**Jk*>L@e7?MnEWyl6zgX=!f-}7Xom5AkYk87!FGpl$$#3zloV^snJj$Gq&V`m6k
zh-z5%Re&=v0&EOGCULL=&-*vOV$y+V?S*mgma#>1?mx;Nnb8ZZZ(>Qyei}oq#MG#t
z*vLj~2QZ@*`4LHrld_*+3oLHZ170zNf6Z$x??Ob8NK{9M<!(fImUzvUBS3Hu?2Msu
zO%+JAlu-5XM(mGW7D6NBQ{Ho51*(DIX+}vV1GccdPv7WXhA~651}T+qK8%1pz9}^-
zzYPU!4Ud?9Cz=7dnf8x!&sq7(=o9p7&#O=88C03+I!{XHc|ze^exrPDmsYfk+n_&X
zhUI$lSMNnUzH^~vZ#TN7<`qXzz!l)(#Akeb{o{_;^8MCY;Lp;yGda)3>)n~Rn;XHD
zqTm8=WYJ|&rG}PE1l3HI8Mv^OGX3KKMd;LCdq-?-mPxg+c@c*19TJJCkqgW}fVG3g
z5z<MtoOL9#J+N-*s!qQjZhfI`knw!TS9{ukUtQOzq+m4JPG5VrIaW6HZ3@V<@2nUb
zqr4s&<cbnSjcVFmd6Ds8q)^1>N_v!w>xqOGzy#itZlDEi2~u9R?mo}<zfUxn?{9f8
z%`1C;_4jnKy?4Z*9dz@38+8y0LeRn}ZF)8Omqs#j_3dN#C5Vl<&B(uF|NTHdygDa<
zOt&8eV%5%yj34$!C3J}U)LyDTemfZfXn~@)HD`z<!ukY^wVp0#|7%h^We#KfSB&IY
z&~s5bw!EmmX>i}K(qNXJmNa`hLTzK@aZl2Rl-3d3T0D=|bT$u}5KGFO|H>3(I5$0*
znQ-%rLcZ-1KH_uagiW;Hd2!hOlX9;4=RgD`=O2ZK$G;_-UnFZnqBzSODHP0z>Z~7-
zlqTTi+}R9q{>a#Nxt2Q8;%Cy9+%LKhYzp}<DQoM!GLRAB`vPI>y|TGZM$iEdKre}I
z(qJHi;3%dU;n;O_dz;Eha0)y_6p^x6p)my01E;2?6dD7=k_U7;Hro}2DS$YmQI><*
zn{66&WQ~P~UW&VWlFirE(wQrubYQxNDUFe*o4+_2ToQh3^Qggp@0!=Bzk3h<0=>5v
z*QJKrx3wlgR*16Rk`l-yDeecn(px2#AdLlHIx_k$4#kT+h<`yY><w??i67}k8x?o|
zwvG7Z>=v|~VKVH;F|cCKBr+R#Yki|NB(@p5FS-;4X02f9Llf6GH)W>O8+ph_>>=TC
z@gNzsIXB}*$My~D$1(YT*>Eq5n9I0Q^6VHtz_k6u#@{}g;x*_n94p;>{!+*L+wv3e
zqp7J^Vb18w%<XL$SgWy6&glftF&706<6NJr(8NrKxM>%$tbDx~S{8FV9XEZ?I<jRK
zT1+zH+0UJZ^h^k}Sx_mC4B!!?aB_4QEglsgM*vd)qxkW-$G{soT*@&ed@K__l!T-i
ziBK!#g~sLVI2_~h^}D<3A+yuT3qBQ2Lp6VrQSC{sk&}LZ1UsG38Mpb~FYre^Y!<dT
z6L=<&ev?c7;Hew+hw>SV&`KboMy<#S1)w9TKG*^&Un!!ZX{9&N*v<No!FWFYUtcJT
z&}EC+H*wwP>dRSSLzNro9-m?%W2@-NJzt9kB_)saC$L<D)MIkiDJx&p_N@-G<R-87
z-m%7AICS!%QN{wFN8fsOEO_0q{Z(ivD_I)Q^`I-8uUkK5S9rr9(sHFEREE{>R)GZM
ztmc%z(e#%m)XCZ2vc|R%7*FCxE$)fYT!9%{(+JHEN1|F+WdPhG)sdE8-MYyblhRoY
zX>@6ej_%4q)J*O&t}X~^=DspQU_F?OxvCstaaoSMB90k?z&*WMYcf9%d__>#VlRiu
z`GYN6I9`@<9__lv0>IfN9v>LI-&7AE9#>RFPQz}T$R_j(IUoP^fJutosPl<kfrE=}
zCi2gd?X3p?Z57WK<73^we#J;qwMqmL&I8=!{HIYMd9Px>+1=z1mywE$-o=aB<raMZ
zg4XNnpZ#cueYL8nK8Wlwt(+dx@hiG1Qg@?i2x(Cv3Sd;0o1#4xt^5GFtJ&Vb>!!t`
z%?`H6I?Q_?@gQgWa<uq+n70a^xkOkwEj&M=OxCBJbT{QqtIR4g_o9(;Ae76^UMnkU
zK$kJKe9p9PI-|9>v|WES5J)fU=*lrnRpc<9;d4q3hRjZafUHELiUom(a#%#lr6JYp
ziAwZ!)6BDzdm)*4W8RvS0agE0bs@a3nvkb0V})*fXJaG?`0qp#atO1q_zI1RXdA6S
zcO(Tzep&EJ_$gx==ph<_`^k^flg*{vy-2x-)lxw_dt#c)sei?B#bqKIRHxPMpd~)D
zpNkindUSZ^a3l%{Evc(2L2f_Q(HVCS8h?_B`Wu9ZD{Mc(i)N=tafHoS!lvz_3qQQZ
zRl68R`u{ocEG$`S;?K^r6^w62N@h6MZl!Yj<{M+|n?))<_rD|hoKE_nm{{oNB>)o?
z!F)gs-mDlV9dJ#NC1-e<R<D@vqi!&C446fy8BE89!_a5xuClrIWR=}%w0#K1yOG$E
zc}&d35W_ywDI#}EQyBP_v^8EQ8Vcux!M=(hl*N-GY5lzZOGhSuzN`dtXDwWTWgP&q
zQ*jHCkMSa2afh|bS&kO|OMXa1+`5k`kBLw`cgJrzNzR+DkL@vyGP()%krXC4&kLk9
z!gW{{G6lM15vMGcYXUv%!V?%NIp?dVk&^B1Y%EN*g+CxA-wV@Hcm%e};moO-`com9
zEYYuO{c%7f^v;xS$V9uz@ygHNsXy=y=Jd-WRXFG6lb`8?uGDLqv@QmkfWI3~@iXk$
z>;{icFT%7axv_+gSw6@ho6F?QEh;$t-*ceJN6{B21>y(=GA?Z+UN3tWL4{i!34o8Q
z%*MPJhLt#8U>Dar6JNyLTe{NijmboOU@Q2!ITehr3amA{H(IC|Qh!FoBQi-Lk_JbX
zi?m4I#P;7EaYoa@_GP|K`OTC+Evki0My#^P8m2gA1m`d{>wnaw@LhH0PI+Ql;Uo`W
znsy~_CHL^W9({`{?JeP^;#Q%Ui{#klf!rtMj>D4$Bige7#I{*1!(U9CHb)f1hmw+S
zhffj-03a`lfPmm7syImRtX_F8Au<HcG<35<VI)msq`FZC&R|SstQfvgq@XM-j_9-D
z3`=SKU7$u3^oVOLw{T_?K;&CL;m^~<$E%;eu;<MMOw?B7vC)zJs%zfSSFT1l+VB9l
zkC`JdrQv1hD=dY*^Vnz1-jv^9@Gf@F9_+Rqc|7?qT3XbF#k0$!vb6v%pL8Nd%<Hi9
z8jZ~ouRD_WbKkJ@nfI{2lar<A`=5Kj8w2=fsM?vC`&2!eFJx3%eU%+Rrivjq-U&kf
zRDO^k+Qd0gJe6`0)IT5|a8+SMiOyTu&++M*)^-C9HpIrZ8GG6DU83p1Tdz!}0eKOa
z1B^|iP54gdqYV=<o7Dvm<%xyyp@mGTX>AsYA6V4~n`PQw#`bRFFq=~?RQqW^!Er5z
zso}+!m7wvY(3DJMV2`@cK39kWPej3283i6--^cSlk8_$4^vHC!7K3L(F!f!w-=GZT
zAjUe!V$w8q78uZ)fL}L)mg64h+PK;NIEF)uVrvOMAP9#41fpF-B0>${b=aI~;C;J)
zo}PQ?rR{cV==Kt&*3Hw_C-|Ux<BQT=kOPh1jh5vhpfy2?CG^Equ3O~5C;*r|Fq}3{
zV0*dqZ=d~T#U%5$_tNv@g?G(KhGzKc9oDhoEl9B6-Bbm)N&$sh{-ydt1C><ZURhTI
z6*|V1I==bwJAGxcfQG3#kA~%H_cEB{=ib9?Ko@7jc)&NRL8<8})h|NI`}7>IEd)^R
z&e%h7rQZ}>>bI3dJg|GVfYJ|&4;1eWpEMcUqw^opps>(B_6^%+xeOrIOaxq=MWPQq
zS@6t6OS7B>5V9{YH7#GadrPrr(SKEGOtj58;?Bj4QmB0H{ixEYikZjo%qcoMa9K)s
zB`OM>{jpz$h|DT+hddaoENfRI<Mg~l;>vDeocjh8#C@fK8-YG)N9mcwFvtT)u4F%F
zfO=A9c9;I@#<BvzyT`)6+%53y@k;twQA|uYNukk5NPwXt!4?Om$>t%I;{fyTV^`cr
zlg(@u)VG@tf0!@wFYWqS>!NRhvBZyBRM1k}nJ=?i-7+Fa2Th4Tr|UQykguXNdZhQ6
zJ6|qKG7*t!P1vcKo?RGasBpz36Cp5SzU-D-_&iL$O|s9dP6TmBzypP8#k!|v!sQ)J
zpmp~L@5`az535eI_wK=%fnUku3t-NA<a|*e5{#Jlk>+^IG7Mp?U3;(el<TY9&HSf=
zA<|S*Eu&>@KT5wsv&pbI?0Z?5rIq{|Da{0p!8GVFR_-E!4nPtgpij;db#xX%PxZso
z6Xc1@A|f*xf;^slKr@#FOP($hGanonOeaZ6K`Dm3ia`A59iFtt0X6DbSF!Z`%M9S$
zCsD_p5gJi_Uf1xv_I=ZMX<o9C<H-H}%zF7<soN~wF;Rk{%<r1j_xCUCd2Mse506#U
zA-8{6C~j}kZ*0bOHEe#`)uzl*@aP3B9@ygUiD~QnH&FHFn~$xS_Vr53>Z4~(DQOj7
zHivYoNmn9kPxtr>8(q{6I9N#YW$}-izvNJ+mTLg-q$TfRr2W+x@7I6Gf_t2ojw@kS
z#cQgX9~^7&fR7s=$C%3{dU+Fgz%OA}W1N4AUSdjVhhUUR?FmjIcXwWJ11z{fa)jYm
zg5wYp2&_QFmEsb+JGy2^hlB9|w_0=Wt}`R?<;!kd$+(tCl1&Va8dq69JJh3LK8|!)
z8ul#2>6h={(fC5roN7k37Pa0r5_Ra6t&O^3HsVVF!pwNQY5WMA3HHbE;qrHlcli*~
zVJVkPae%V2yR*#dS{tV=I6Wnsc=fD;tx}%erMIiU($f!XE?lG3E!F}uqbg*}Yl*K;
zYwPYgdHShg)i~h%42bGDhymZ;AF0SiZ4V2jD2=H6vbq(r-;KkMtp7B=T?c&|bV{&T
zIIEMD93lBg<am<*zT?ZubF@qK1A?fc{i5<G6r$fiiBj4ViYZpVr9{0%VRo;?E3!ma
z{xn-W+Ak~sCmmzVUz_u;wd_lId8=U?BhAI1$AkBt)|n>9sV<CidGMySbXU6$bxIhm
zVItyDau%+DG_qp=Gv<d;HcM1pl^DTEG3J`~Bf=re`>O>b(?zws9#i|!JXwt?VfD3y
z>H-D%m~tYKP;}AgcfSA?kX?n1m`t+^U;#LY+>c4mw;^)e#DW|~lPb^e%f*3wMoo!~
zO-pCPfLk~+NrD4Bx|~@K5k(<+$h*l`BJDG!RL1&{)9Hs*-ecZA<73K?)GPNZqdD16
zJuC*WgI&>)_j6eYXTUL7VC1WiJ2IDvh}(P=m}~Bw=5O)8?^414A{hwJblU*;XILL`
zDYRcxXy&HPwZfXin)_d--5PDqe4G1ZEZhO39PCEfgeh+NQkU|3_H^S?GV|j#k6FIL
zP-U9uzPUak&58DzQM>KOW7(~36yvb*@ZLm@Dq$O1eBYsCQ!#Vk4Wl*O6(wlo<E7?r
z=dze$xQzu{fF+@%9bKhi;en9FQ{Pi2BVhF68vTAlzCfv$2!D@t7S5_&IJu+|t~6CX
z4GO}2-t3(!PO1_Pi~u-KVMSJOIH+~VU{(!|^$<n!G75N3B@trq*#AY=x5qR6{_jia
zKq7L;LJmpJDn`zSghI+8wvpsG8ykk4N;##GQ;H}_CpN-tbI$pEC^l>(n?rJpm_zuz
z-rwKj`~CF!eD)9j?%sVr@9TM8*Ymn5Lz=mejGE222<)Y)yAH_D)*~<Ci+H|jMjI5}
z)I?>LS&jM5h`#sL!9P=(Q*a+z8E#)*evr5Ski_t6|3e_KXDJVspR6$aN+&8pZ|EAG
z8S#(2+3_|&*&_U6aA&XUlXEQ#>Zzplk`Mpod}763J`<S#u7Bu$iEaw;rUJ@lF1Vg;
z^2R%TWyfouqSx+GpfOEf6>pPf-?E^0tj|`+jeLSAItV$fKO^l9KhrF_H*@te5>puz
zeMwT}v9yR?&G_4Lv|wsc;;oY#FZD(nwONSb$Jt*Vw%_zYYko}a1p>e5J&y8<A(}|!
zA|mP-eg$Yx&uVn~NK@aT(z&NhTc1VLtMO3FFNEZFPENiAO2^|AaO?1;DK*Qv^a8tH
zGhimlF&9drXe3)q`n_Es-x$fUJ7=Q-G&r*OUO=%org~#5f58<4SO(!YO_*Zt^_*DF
z1tYTak?`$%m4G<RT7>cGFB|rKd8k)l@^EP5qViqb#M;-g^o~(y+T)dwPbU`b0#RPT
zH}}&yWs<1K`JY19#@&Gq==0M-?hid`KKDgOSIl;&FDe84@wRuMSlvB=Z(E_;^1tco
z8Kl6HfSJ%uU;VLUK?sx5bX9V4NZ(Z<e5g6{2)^AOg($1;J5B~z2=B~AM~kVlO0>3}
zn8Eqtl4n|o;@AtFxZ>n&HHjZVnz$VWj36m*<>ft(vp#cE9{xyu6_hhYRo0JH<xzh+
zg5gB6&_j2uvL+Gd3mhkkHij)^@k_K(kL#wFq`46Y^rWm;fgPo_92uNpGb6WcRTrf#
zNbUu)!g_m$0J10?n%x^b=~q>&#z#bQgI@%nI$K6a{>C^l!xT7+`f^kM!DE2_N#W(?
zU0j;`IeO1~1|N^{kMy>2^=BLY@o+@f-I^=(O1iV|olZ(G^iR>m>M*uX|L5T~(&8Sx
zvvW%88Tt0*2aS(889}|m<;*_Uh-@N$$FuQ=L&hce;|o1^`s{!w`-Ut9h(;ywpMesL
z3%$Pw6xgTj_nwPImy{Qze<(ypi8pGG9VaVTO~xepf6_|jIVhdbFY2*Oa6^!ro!oIR
zosk}~K4sC0f{8`U$5XPoMP+0X*@;YEz?)lrdd9lvZjc9fOjX{m2Egslc+n#J+Eh1v
zIJFohHP?18T|n6@pdvGFco!hVz?P2?O0a@k-NFi)H=T0qRC7Yt5eoNZ6rE+>n`!I2
zN_#9Q(Pc9sMb;y!gv3Fd6e?JETuYGrt8QaXI+bi;Ol^*);G4M;GNc;$eZ`+VIU0;v
z03drVpFR*{0QeTK96Z={X*0&6J_h}D5O-aLn-2Ipyc7f8L*7HL4*!>k6r+X^#&yS5
z8H0Umf2!}UP{%jSopX?1mK9EPosBxB5{R1l`H06YHYrkNu|gtnPypQe{0}8t8k{o#
z7nq7QI$1n~^?23l;w(LUajf|yuvB)4<jeTR(BX^4(wRg+QUi3eSN&F=!|%4a6Jg`*
zvk7!)PZ^{DIy_yS*%>7l^6j^O_B`S12)q1}r~BzA(v9+E<w$JDG&9b&&TQmIXcQ8f
zdFrwQ)A!1xSbB-GrjxNF-2F_zjc}lSFg6B{xcbW>6+B)%ZMJGbO8ZJ>{b+d}>VUT>
zBI(&27wa*<E+KyE<mh<FM*n)A-ZtjcFxlO8rjAj2(0KkDeR3OJJ3=6sB5E&9y=m9b
z=$*~W5^Ec>WHwG3XZiS*1k-9&7QNC~z6+{<)?m#B75_mUJvvz~4}T`;OHnrY_~)U{
ze%;_bz!x-=v$5OZy>o58*vypK%Qzy^$t`kIEM!b}F^HJcm?tLf_+rvfD2aY+K70PT
z>10g4@^wEYFp6?ZG3TM@v*!XEn7B_mB94Ve7^e6%2a)$}<Rj3&S5N!b{=CJHbWJnw
zCH`5Pdkh`_Xx`hySktrgy;F#;Qnmgec0Lgl>$HZt@$uyul#F5kR_-x86`=q||4dK+
zIn~$=W!Wj7wV)pXM#|2j0tJYkxoF*03ht)$G&B1SHx&qw*&Xrr`0yjaSL2kVBxqq?
zZ*cAVKcqdK=)V2nOWxwZ>BDa}6hEG;!=G__pI!;Llp!|EUXR@K--cc0D*yMMH9*Dx
z*Ucc9d^s1`WwTI^Cm&U!doE-V^p?&!X8UejI$?7a?g7vB`g&ITU5U9uCg^w7JN=o)
zG?L)DR&*BJrc;Al^rqeZBc?O?f_7>Vi~R9+hEU{H&Xn}lt1q?3IxUQNt}d(RY3w96
zw|vyaczsuJoj8{hYE|@%(K&f*_+DME#~fLjl8eomzuMyGKz!N#O*t&KZ3f7`l6$x_
z|2gW9ZYt%Go93#fWA1$W@CPjsN1g4Lz=UihS`M));8<<nY344Seg*^JWs81KH0B<F
zf=FDbt8OYP4xRAZy2W+}uN-8rV?z9wfFbxR)ZW}}zWLD+kf~ed#x{QSvZD=Px^GZ+
zP`MJ@W+R94x{(j1Cztb11HpMThZ?<+%8im4={+Ome+j$)qkD8`|FdG*vryg*%a5M!
z-R+OTdhE8V$k4`1H;+*6-wVR^b%~gXIm#jnT&ROHu-Cv#-gl@4KP8-8Cl!EqJ}~;+
ziO}4<K6&2J(u>Q(x?{~RSU5j!h<B?=_%A;wFksWux`y_F7S`A?J<~AE8v0vfr!zOG
z$IH)5<^M6=*m5ePq|E&Mqy?J`e<8mf2_&hf`+6FvhNcIK;y%XZ=RV9ClxXXKD7ste
zV~bdwQ|%ui)~0xM;9(<Z*vodPj*cS@kyxW42UY~=(9yTE47=&7%#P2+I1Ld;xOiVP
zK|sX3Fks;Yi96%X9pg^7>-dbW_(S1UMDgLzJG(GGJyQ7VnRah7^w{Td|Er6Ryy;iz
z&PV>CchY;l<obwF8OLsR(VVw$qhCPi*4VZ|6NSwGJf7J~_kj6@M=Ha7KlKXj($zf!
zh7RrhsLz%Np60&bfNXM{k-egPG*L%VEF{@A#!f-9oY4E}#>n$60mqgy;kI#WX|z!5
zJzDcmvNV;y&swASgL7*5J?-OU{Zy~(7KE~yLP95i@MgL@gQr<?YwexDEU9~>s>P7=
zmrogv&r{|KV}RgPOP`*y%<4L?mXf8m6~Nfx^5<0D1(0JQp?h8I^wB0D<E=t&@YhRv
z``uZ(tQAPx2j+gB44RtxF1$3?{x&cOt8QxHftJdL%Rxc!9h}|^gcyrCp1t62;eoLJ
z-fFD1Szfs{#=mb!{xCv--A-BD0m|G+bqPQ`;`4s3+&cN5>%r)o;L3k;b%B-*klPA-
zquvc4Oy`05p86k_!If1~Vg4UB_5)wg(01sd5DYS{A922Q9<q~2j3@}ASj<`R*Iqm4
zK;04D5q(S-1PVY@?AfYj9Px@(fPQjwtQ{T5WUeYXjs8X^CAZuEdUyEY`Rw(_edkzK
zFP(inB8QCy6@<ws`?AhYPwmsH?762llj-)GBZOF4{7kqLH(0pTrGprV&IIDUa@y?`
z@Yj340x@~kMK4%$L6GwxaB6i6V5UI-%*cI<urr?#R>**lrKA%tbxzL<dwm~u2Yl#I
z0r0DS+*#Dw!E+!c;2%WoOSe1!zB~Z??GsMT@V+PXkWt|ceci!t)wsZ!<n0jsfbm`#
zSnhr|Q8`!Tjy`OM=P*T(oPLCH??&kPL>*wJ$yqw<Qd{GsK$Z{vee15X`G0|t=fZ8+
zZd___tWy4*1}wBSFPU9G4YB+f>>at^k}zX^I<s>VK)#v(I0(E}fBf0jeRxp?&|l<S
zhWN>|<Wp9KX}sANKfe$!>KD6sJWsO%pu0iv>ECF@q5@A#OE=xJ(Mc^t5%+}vdnLDp
z!!T7<q`1#qiv`WJSz+pJAj{dKWOAbAj@FCVx$P)9^tZN+*i8};`dK*C8TE2hqxCv3
zt7=mID#P;*@czrHP^YJHVOjFUe2R}yS>%&1&p)Vvh^PI(4dNiLiUS3}K-QxyQZA6?
z&N6`K0vXIe{#+WVgBoj;pDER!DsfU=J7ChBYLZ)x(GR-YH@2SvJ9NnGN_^KZN2^*{
zyv_^;xEfRwbzU?N^*S!tNfSerfI*fQUlyFH*FM)$)HkvJC2Ik*M&R>AL-l;x33~Mf
z(d|a|lJEAz_rb3RFManh^!e<lkOUA6t1878k+4J_@~69C<J^wXn)aUNlE;GNgPD2K
zs%ZO~Bc-#~FV<-mf7*C?X{>X0O4#9wRcKzL>^aEPlq~YRDitVO^uAA6A&HScknx)i
zU++2bmc6&jey#lR*80{NMKr^Sct$ZKeGsV$rpjS%8ohiQ<(|mHZ^mT{y6n(?=?JX?
zJ#u%JCjZs6m7)R%f63`CSC+*j4Rw6{U?osv@snb)SY-hv-XS~WmdSmz)f807WF+lK
z+}*2x?5X?7u1#^uq+n;KI(dF%_51qvBCXQDFIj?#dAN}lVLw=OQ-zxW<Z0X$V0#0j
z25cjj#a>x#*F`)q2z1j=<UdU(r>`Ym8q>d;{Eckrr&KnQEb9JAJbH6wP1#X*E6me-
z{_@68R7YKyOQtO9pIbF#?6NHC%W*Q`GA}m}B3pQz@03GfF~4tU$~c~5O`a@QaCHS;
z`lSCpg@A4i<q@|P3<bhLy^i4({oG$%RkHb(v}C8ClL#HZ%04^PS@&vCYTG9U4Bo3e
zjks!^O<e`d&ITXh>ld^Ek*ZFpepIwkO_(tHco;}p%kL;e5qfVBZ1ZEtT8F?EE558T
zj3C(>hQ5J=zKkyt|E!~!m{Wh(2cXSgt9x@D&;C3xumf10$I1g~4}pXrdf07JNwi-?
z#KnYqZu%0IppXJDM7;;(lQd#^ZX(40*y3FQk#}bU;6rBst?Jn&kAI(#Kj7w~wU|b}
zIj9WY1UZI;!yTgD&%W&GveNLo@-c8a<T4L*vJ-ecu#)1`*Al#R!W3D1;UuebeE!1`
zY4D5ux#{80Lp9D$snu0C9v5Xoe)L<(yXNZ&_0?OMs$hiil2@HG?PSl|cSR`$td^s8
zg5Kd$QX(U&l&z@uk}K9$fmcg+3V(|l=#7QUEN&@WWr>wpdf*070^rBj6oL^jG*HLn
z?of{toM+%!!IN*1b7#=Jm-KtpKHt&4%9=g{7;ayhLyso^4nBZHk8Oy?3N0Lu`=xd`
zt~HL$_dCdcX{<bCZWlqrEREFGf)0N+5Y+V(13Flw9)Qp$s{j+LCDS@Y1zDCWMsq_0
z=dvbu@f!5w&3})AG~%DmWvRT#sVf~l0+T$9RUfD-*&Dh%_GwBL%*xEU?#t3GzU_7u
zn!I#zF(x}HC%j+j8|mq+ltzbpnb8L_oa^<fbe2}ni{5T56OKse4^3{@E2mplBaK@U
z>3;h7EKBF9-UfSxys)|MkM<G@!k;@+D=Yx*{j#<Hb{)>eK>{fJv$TpgBMypLX)`vS
zlb3`#Tbw=k5}4XVhdQ%IsWQGiWtzv!*Eenrnq7V{*FB^$I^y~g=)BIJ^<i8a_RQ>T
z-qv_q9%#Y#B6Pg<fw-5oofQC&NX(Qas+n>tWTM%I?9@XT-OzD=$vXC;_`}*-cjvQ<
z7k1J<orM5v{!fxLO)dwkL$maFt{bb9^Mcf}aLrvuqj%yDLDxFE|3CDO!>hfhtZtl<
zdOqZa<p6qCX36!}^vrG&(f4xEk%Adc#hIT+JXDY@7uw(q^p+3Z$AaFNT($k==KGg&
z=iMJmgX?aP+eR$M%z>Tt?gsP8XHRQo;#Yi*t%T@%Yv5_IyvCcmrXO?OdIb7FiTY03
znVr2*h=KLcJ!@aE^~mU_scW2By5sM{YLyD-bo!>+VunRcH1Ot>R?6A<*(cD2L51ov
zaB5XH{TlsuorPlNw*^~=oZ|EYxSalA!O9lI-BmQ)LE-E<=#^#>E`Ri?gBddqZUC(1
z?Vx~Y_Jj!^VAU_jd5zcY9KtNkpL6W=m{%o*_KLIyomsqV0S%$u<qW9=HdlJ1o)~Zj
zA-~q|(wOyW|6SV)SR||I-M#h2c?<k}o!~b6WsZ~LCytBv7tK^Tss<SINKCm%8QEO6
z<+4?(z`3ud60--qTQ6jv)M=HV@#TqJ*=X5|5m&c~S}u~2le=pZ%G2Y0(gK=O>o)%B
zJ}0;Q>-(cw#ogh(eS9-7!)?Z@*q#pg94GECUD@Y3T?_X2*e9<R;~E)l@VY3MXR<m_
zh(lQ#Bs0F(`F$gL?}}>{X2xdUAO41ASY$0sg1Pv|e2Vd{c_DH<UZc6ED=qky$)=$u
zw2=8~PODvcz9sgsU4;zQ@HMvO0ro%}HE#J1{P4EleoHvDeMrpt%R;k#OG9P&BQ&KG
zYbFlgyL6-=C74O8O7x$vLbdY>#aGtlg9ymr!JoVjaLJtCn_Y`u-!ATW2&q2==JT<l
zbZWrD(FL`ncW)EtOd#$&3p8U8_11%1-^7__tY=&Ubx;%w#|Ta^cAi^Z@BbRy*7(Xv
zw5wQF(yaf%WylSmkkM7eRi2p=Sk3apY)Vw{Rx6=sb@vTF(;@bg2hRCnj}+3r9Rhw6
zrWhVD*Eq@~m(6wdilpqL@t_n_E;OH!{@>4=+!b?hV8IUAaMxfL&haKg>_ERo9(}Jo
z)!<Jne}&k-Wf;eP$e!h?kAGg*m5j0JFe&8Fi%xz=p_q}fB2x}qTNlV-bW<cpf_y{D
zguOHX^Y|y7lOBSOxZDCYf5sZk3yDlh!x_@!>7dqg{@m!`N}G^C5Lhg-cqNyV__ZH<
zM+ZDI*owI8h72hOOOCv}`=G?I-{Pidj^_l92fZ4%@J<*>|4K+pad{ScqMXMCtHCd`
zxN~DT1r%G2YSIL~_u~>4zQbBgSfAYDpn_32jyntkH^w$<71c@E(+1rtuAm^tvvakB
zSu2yXpmf2TA@551l$E-1GLm4vj6=m9%Csm9(}932GN$9^TQ0Yd7c^0}fi?5GzP%xm
zcN4Cf{@+JuJ{39bIewSD3vJ-465}v^Te}qfr|nRD-OeKWE#cSM{w>fCoqmsm?f+S<
z2y-?wPDZK=x3SqhCMYL&wPWp0wV84fM`>O4TwhC%wK6_+yK(fiiyyCBJ;Hd~3@Msf
zT$W36+e3&FU&dEn8cWX!B6`C(a)JoHEZZ*BQ(Ax23^MIdRid>f_6}xgEgX<}aeG2Z
zgKtY)Q>>HQWOF{`!+WMrJ66$%VQ?xY<yta^HO)KrAq>PtMkOdFU}i>jpJw_^qC*57
zBiyNtoVE!(l1$@sqQ{Fzc*`q4M};GNHdpAs*8;Qp^7M63i``f()&Vq7ANGL{vMpW+
z75eU=H{R(2g{){l*fpk?9aOn4F)bWT8vc7z!DfJUI%sP`eJ>ey)$yFhYrZFavpeCt
zZ#Jis9+qMbTSoE{!X!-Sht?w3LdLhLt+ifKNQ0%s7<zGjPHxvhmHq8a1|i)kuwJtZ
zJQ=s<$IS!$5d5lsN8Reh4c}-K`!nMhB+J<qIi%xvkoc-hS~;pS(d-R<)hz8UE34_b
z2afT4D&l2XwO2tZiT{?5Y}AQBP2|X-dS{<o@Kk4348(o)G+19a1yUJy3(j)?W$poY
zYsOyfU7vjR#BOmc5ngQq?Py{sRS<so&Uw}|`9r>0Lqo0^eKG#hAWaa)Vc?j>-g)wD
zSC$Rz6)9-5cWrJpX*Gm`%V~RQ8twkrsK_YtX*gJ!?$Sf6swM4QL=UjSeu-N4QuVHs
zu#Tb<6aM$}!zElPn>+%hG5UGy1voUrrtgnz8|-{@$I963@VEXk@!r||8b1FRSH1?O
zE5BNH^JOh&r*{6DXzY8Zpug%?9KN{M)N>;)?Z@+Gs&m#<)m?theZ?(fmGH^BL9te&
zZ;vRQ3+?uGusIVtUXMJYYFP9({4Lsh^3%#RLIyd%bzAd5jRzJc!hKeyeQn8hbo|Ph
z>)htpj*zkJMfJ|G0MJrz<{}0u{9WvpVi-;HQDrLule7<i*R6!dhn3ptL{h}uOrSAZ
zm3dIC)6_#NyM1E~cNjrmGqHq`%-Gn(*g0^oYjx<v*UYOTo(|8loCjPpN=1mDv>`%t
z;S`~!u}{mD;}it!Db0UBaoms?dMcL1VFLZ%0wpayS}0MZCL+%p)U;vyEn-1_eV1=<
zS!n0?u4L1&LEMd!%S(4b?0u20wM(0?jp;vi=)oW`FzoNqR~DdJrAhC!L_7t#KuEUr
z;b4JNFv*6lxhAc<2cO<G!o$DBY2vwcd?ODRWf%*r&&27GoR-C6>)cM2Swq9ADYg_#
z8J`VU6%0#jCZY)bzkQ=!3XYqB<Ja+Dk=>uw%_9?X;`E48!-N(+GUlcUSjTI%mL&J6
z{M~Rfp;T@>5pJ?d*;pQ$Zd*A`t*;C-2N{pA@Pql#6DH7oKp|H!a|Ypbl2!vn-jzW0
z!xp~ACX(l<Gc${Hmv|ni#>$ZcJC8`e<_--sFAOzuZSkOz+V20nO`O-hC+_ukSz^=N
zMW#7(IAP*?*<Vy^-?i<ojYr>H65l+ew%&IAXDnl^H8R75KE{qba7AJJRmWF;@XHiS
zGQZKf<LGoa8aAiNjltbzACrk@S7;>Tc)w3k`133e4J;RA!tVdvKT?dHYn4-6Fs3is
zcDae*I~&JjQAYlU(XcKDDO-!mRn76>xkf0iF~LJ*-dA!<Gfn=)`T|Uz#G$F7_Axe<
z{|Dv}vDJ)~S^m0`p9w7axkr_Cv2f!6xUt&Yn24Et5<#u9+9xN){+Vt!uMu*vl5vV@
zN!6Nqia`z^bD+%;v$C?}M3CK3Qq(VI>{Tyo7l@0C^A&c*gFB(Vicvy2q!y}x9z_F9
zEEfr>#?s4XTHHNK_QjF@uW+0G=D+j2nw(EN`3DOYW9@CFj19Oq85sHQaX9Oxh;Jb7
zBvhxf88Dg55I<F9CAD&!TPQc>Av6>BAdUyU4}3>^zJs9h?FO?E8q&`7IQ){>73#s|
zm92l4sekZ@tkGce&_EsCySVH|p-iS=bnW9dS-9laI5o%3+r<1owSv@!wS3%qLAptQ
z5dpIukgk?DB3@3B);Oq%6|1c5?_P65j16KN>GRKnU~|UK=Xe9p+!Gfg3#>EE5cK73
zkTWSv2K=agZ6AL7VPJIGIi=B3^KZRhSD1~>YaJaas(Jb-<nDCOW?}3muluz_J7CnD
znWCcxdK<wU>*63fCP3H;5^c62FQrtD|KCT?JH_KU&+lVF@pWsP(;M|$12o^D4N2p9
zNyCrtm*n~x0}X*~Y$kF~P+wmrb7{bZt`bE>OdKmT!lMhG0jHgSVMJ#fBy$>kbyTC<
zhDOyUI#T0NT0o5!IGu=LDv)eDrTVQR5{gy^I#4{0@#>wI#y^8F^s4)s9gr%EC`8&L
zD)z%b5na?Hmvq_@{v(bjOiBSh<XJB}o}RNMj6CfBt<l^AEHRRiVyB{sYAR9lb@J1c
z(4Oz^aS<_7h^OEPn!YS)$s#|+82S%YbI8$WV9KHrAhP%N0ox_g!xz2MJ^N87uskVX
z&-%=;vmnv&pZB@f#c&a-#?aqdagW^}w0QnoLC&Xy_Dg;@I)41u1LPlr(Rj7jBU?$X
zA*WXcHlo>V0gdy+zh$!c5NmU#*y@|?n(7h~{EC8?qv<)sNMuB%?1cURL1EH5_B54o
zlLK_dSz6TcftRz@DSzGxp9=}^tZ&;7w)5iXA0mq~Fa?={nXzRm4ZO-QgG<`n5$SQ^
zHfdP%byj*XGi2;jL>pc*_vOVu;hB44SQ1uIO|Hl6EQpmukmK{hs_PFDMo>9eNOBB>
z6$6t2g8#ys(VChg(+_Z&AMzC$tn}6I!kh9QaE7E&B?#<49><O|IK>rvBBoL_=!uh)
zt5cnB;8f@P8H1gT=ApzSFa%!I>dG*cU{bIKDhG-7IQpFgh=vPUshGjb%6QO|uk?3W
zz9*|&>85i1<d3*gV1c^$bj*L3uy?@*63y-3)t>C=aU9wrgH7nURWk(=%zG#;NR9l7
zfXh=lfla|~O8%!NJqxnW>_yuw?{ojg5w9txH73>^ID0x<0i}b^>})&Nrh_hjhwp_D
z_5FH4Rc@-t!LJB!;vNNzSj{DG_CglgRKY<_L{z~}m9vZw>He*KYOSZOaC`v;sU1$G
ziUYrZJBx~h!pkwp5E<edY|cUhfq|g&w@N>8?tSPCIraI8?)S@sjReiNB<}0n2OEC2
zZUGqp3#(M16bH|(OinJ$&V?YyYU&P);N=EfQ=-hyu?3mWxX2Lk_caWDksqr43p2;{
zk_#5K{$7?*oW$J!+*T%k3in-v>`gqxpl{DRhVOQ;4^DIj*7JPmG-TWIqLy0Nl<mA~
z_}dlhI1+3o;wdBwXL!z#+p8-E+CU64QSsPuh3`{R5>V8t7-+tVlhUP~Y~&X!<&NDU
zxn05g568K>Eg?>=%UMvmt%AntfD~QOh`cAYW=LwVM~q}KMn}kxw_5%AF?BFab{q?@
z8YTYvF?kV0d<hP)a->xj=fV;8GLB_t_{1x+G^W;6e--X9x}1Y5Zi11xPx@5hAa*e3
zSx~<(afX`Z@NOCjH3*y?>`0%nj?x{P4Z0Mt4<83-A-&EbVF_@>MWJl6;nHb#R6J?o
z0%F%_Bmy?Fx%lw!bLE{CKHo%-(c&)+km_vsxwo5Dq_zSM-yK<!yD#>refzp^^sk;2
z->2K8c#^$WB~gT^X0DAKc?SJ(VAIc|Stb*Fh>Sc}RNpHNnt?ZCp;2!b4M{kontq8^
zc`<xy5syq}ra$YGEDt8Vv3)86jPDF#X|BZKDPp|28yvR^@<IK)Q~Sqc;omibI2q<;
zC8_e%R;#oFmgnAnM7d{zV@Yyigi{3M&HakDujgo0Diu8;Q_sLC6L-hkEqC`J4u^HM
zo}=MwiHmu_p{kdpjw%fkQ2qE5PdJ@(*E@4*<?mHpwYDdwfWJV{vWkkbO2x$Pj*nS1
zXz^VlL}+q-6dP`^&g2}DQfWIU_<u!sz&*uHrD~LH7?N;%A{({8Xm7oQk~VW}*KecR
zRV>b&voH$O;BsGTv6sPCKlu?#5Fv&tjt}!jpR|8zAY^>|mk~1bmAUsM_iLqfqQEIv
zwh{i~X|ekb*QK8Kv?{da;~pE2$q%ndA(;Eewc4!!(7(_(RF|5QVsmVP9!KfNs>O|@
zAl6Ivz5X_PR|nOU3~~2kBk6uMWecM@)jp2+bDI8EV-?sCX56n%eY#MzE}ZFE@C+so
zzxohr_|!Fhp`GhB*Y!(Z;zWxpsL!4`K6@vOAJ>Xp__O)V%U~ze7ZDlEyEvY+T;ykf
zFfhU!1vfw_jmmlZxA=mWDif2k-Q~4YvVr#WKYA^`lW}Ahi`SsLyo|!V&2i-qUmnz-
z+*+@;5~W>N=@(~w>}%j>??z-*%QY)cTPlcE&i5r|KtUL$SvWfLQEKOA9`r#&<DDAZ
zuf{F@WdEb?uMiD#_dD0e6h3$tqu&ss7g~WVWl;~Dy##aXHki+sq1RN~<UJL4tWQvO
zBw<?jPw)feerH9=y6JEfZdM}2uk1DDt`Wj90j~69xbeBi{?7tfJIuaPy8~qDsOGX*
zp5TcTT7TslE2EGk;#dYFJ0*ZDzwzI~-Imq3n#7ZZcgfrM&g`(UW{}XEluqW2E2o7U
zZf^*hEv@_2yJuLXZ2M_cFf5<**d40!m|?at!v>*_n7ZxLf0vhg=IY(pjQKw7ehywZ
zsB3S!4>|XFqFyX?Ci4XMLFY)3hIUw_Td*qUAVY9c|C&rD@dR@#eoQ7)ah-vzvY)Un
z+Z_9(=HdkXMR8q{F?#jD>DK(LHvo&9lRzfJ{yByM@dP(je9R3u)3TL(L>d9bTGqH2
zoGy!GO)o%lf9Tfq#EdpeI!5&&>->om>{-s33iqd5*EO*rqn;(J7-ZcAs%hV<diO6p
z6s$ioZbR4DJO&l%jMF0nb!mP4aV%k33b!DEdqT=h@{32FV)`zuGgXLGN8x9}DyrRM
zRTsm<e$q3`z1|_|B+aG0(nc5h!eaBnTGe8nVYJaA#)Mu)H4MJp#h>-REmT6n;rq-z
zO0Cmizx{yfLUf`8$f9j;C>j&+W6nQPykR{4#?pRK+8A~8mrb<fcrch<(}#c#Z8LVb
z$h6@kNQCeb**?bvI@R>$n&?Nm(Bu~65%msCUBs=Xq~-RffV+_#C3rs8i=e~2g;G(b
zWyg4(rZ{gE+6njn>>ssy<AC>{t$M0C-veN{CdOF{0r2^eLnP$2a`>bsa~W*$Ck!iV
zi~Rlhh)lW_rLHHNFml+V{t!{iOq-SjTIlPEq3c%ka5DJmai2S+`Qa-wBY659TBS}m
zVYjapdp9}cq@w5<@2igoM=4udqu7yI_$#Px@X9wukv=QcKGlk14S}6Z`M7tqkwgf!
zStvR}Q3pj7W4@UrV3YR`xgken?wz{AK>ZsU{&4#O(bU33r2XL$L>4uzFnZN1P{CsN
z+xILNn}L4N^`GI_hvW5S4#bhzGv)vEo3OXK&0Lx(F2`gto3>4j$n8MOu-bhs@w8Bq
zQNi*xIEQLw+;YZA)0<~YosHm=Mo4dFT))PA^H5cI4Vr51F~sP`8q>_A28S#Q27ux#
zv_4s-iI0mh&USv=x*<-|<VIPY7>p~WK3i(RwyIb2Q@EN?+Y~eE<DRC`0O(u&b^>L(
zAY;e+4je`~r*GGnc8#~V5HuH<v?>^a?0s3U7aV-Y6TLF}JF{bRFk2HU2>!EuL%1~i
z2dYR4uS`Dscga1j_jb4&H&BqzqoMUQd1rN;rn9m$zQHU1=fKnP2RGt>ax$)H|HBP#
zi%c%|@{4d7Ye6W(nHyg;Q2Qan>F}Lc%h!IQ3yNw?PDT?>&IJN$9avX0v)|8r2r!5+
zl2A1&(%m7RL2`3Mdb8p<z$lq#>D~tdfo76H=C+WT^<9xm>=h$R2Po4o`ZBg^E5j~;
zkA?C+$&7e{t^t9u#<01PB+U0|UYNm+k$RhAp1aB`_rZMrFhMHTW*rB84VOfUzAxsu
zb;f(-1Uz?rr4`s1u_KUw2*J>MEAs&2B_dTo?n*1sCY{-&mgP#7W(^@KHI{T^H9|+3
zP70<e4l3(kNpWuwHi5DOoA%Dj{9nWU<_cd9S>v8YkVgUI))qLrQoaaRFTVMUdSl41
zQflMa#XQ;x?Euxq$nE0Vb*4iuxw@8Q-}h$!4Z+EKKxc5`A=Cil6r0+)NLVvk+{MJm
z;PwI7rCDC={=p1PIEwM(E>RIY>pQ(=hI~U$@DL*U7f>yjF{844_nKDCLh*K$BH#nf
zShE;eA4A0k=cK_~qUl@rH6@Eq5s}qmRm?swU?1GN>s}Ct?`V~;F4gp%7L#0zqaeH|
zISUu){ELlMt+*aBvgT-_qS~&b&O&jYs<QQ1m3rjX^yFmANRgxeSQRSr^N_Q@3?HiL
z)t#XY)~D~4NHG3)Xq0LUy^8fgJ1%YY0~tRbF)<Un@9&0bRCId1MAYmS#}U))u07gb
z`Q7QPJ><Mieb%QhUxbTyBr@(vLI21TQ74%WdUJK9jtQWpM*9(0jVbn|G#{D;Zw3eL
zYv5Koz!Jy0&pp%{J=v<2ecgG}0l#<*VLw5j)Y*c}6lRQ&V-%~h>%{QxvR>yZ=X0{y
zW!+<F!Gyz?qTZsZQ8cZmyfdt)L{9^xuIGqZctCkB8R)@5@l_)~CO6bC9g`dzRns4l
z<cY={jNyIjU6@N)BMiZl!2Y?__a+)=uX@>7^)RZ)eydMy;|<*Gz+Q7P^gF|Fg~gi@
zV}0gR3&}U4zOynIDMugv#lzhDt=_#L<vYX4kTGL@?%EIEYa7QGd9t#?qgHw<#Uw5E
z{i(a5RayROdVmi_TwR3e+|V9N0&zXa3F0Cz7lhkSKtrQ5A`=y{il~MDeFe+itsE|X
zd=sz5?^DDCjG^^s!4=4A8mTU<=nd6^L`k`QjOs0fwZ^z1gnxKW_%l<0RA`=C-v;B$
z1s0kKjK0X*IRtdMyCz@g4ib7(G=DzXLeqF;!~I4Xi2B&D_!lqqZZ*0*wNuI@E8>v`
z|JNHyytI*9&Mq}*xaaizT9@FAzYVNbKQP!gpD^$^hm=3$DolD7OWs**=gt4{W!;z_
za8zjKc<y+w*3`!5PzeocX0GZc%I$CJC511J9QomJ&_0oO94h??{4TBDxqS6%z1Vh^
z!Nrb^yc2h#BrR(nRVPbeY!YbOb=5`K>Oy|1!ThxZP%iEWIw|^NhUpbt@Y%AYH@B$@
zB3qA7$I^1P=Ygde6Kzb#9Jb1A@HT|s{Zd2`vkXqLKPHn=?updG3lPCXcsDw2sSiMQ
z>izZ>hr<MAdD;5zN56XSA;uaAFgG0GppT<Fb#ejnjnN;h3_)4JfTr8+cbU+5122|d
zDg)+5B*|p9a5ocQDId_Eij9de%vU}~u?lyOEYK*A*zS!5iKymkZO>j2@gIBc@_1J(
z{YT2hN2MM*eo1ZsltY%}l6ERl$Hi5{kNF*oelP~~%eQm^EO#Ac#Wwo4`$qK6bK3iC
zL+}BqBloAbqW`F6uC2$j_FCA~bT2;Km(f>-6PXK#s;bYnM?N>Ef4ac-dRz7>)W1XT
z%oA&eob^dFkuPVT;D7q;gh^gAK5%IChzpXJWq#?&J4&fj!kn2Td@6=+dN#@*Pzrh;
z(res`{zO4x-i#_ZXh0t-p5kmRa50E8*W2*^MGbR=81}NTftfaUFsp$gDr#^{TcYJu
zibL-ed76AX))+d+W|>u^aN{9!oGvuNekxb({YQIUt&T}ouVyv?Zybzv4!xKyl62LL
zT$!9F%;qE!y&NeN2gt6ClEizuzaa6|4k?&T`+0ck&Ql1bc|qe2c|Pp_FHC2zzy6oM
zfs2Iu(LmyB>3(bzKkw^-t=W!bn2QdXUB45)F#Y*GrKMBScwzy^Rdp0~b-du-xSL}v
z>5R4=CUcSi84`-M(6n~?P9ssqVSel2-p91`>iXR8Lu<!REj+nxuJP2@@bVWw(a-b7
z&`b^g@_x-n1gYm%^zbxh6ciUpt1LYWzBjh~1uv91N_+=5Kx#T@Ke?(&!P-32Eww7h
zB~=FSfv{G`+R6^%%_SB+-7^t`b;`+h$zoL#Y-bMK#Ty=_Uy%d{iUVG?m+|6x^`*#(
zuN{Ij_-igTvwLsFeju6af`A@Hd^f#SjbJ)(RRr%tmVCz)Ndbz%@cF;w0_jir8}(WK
zuoFKx8P$Qozv&nAqQ5Tgltlgh^F7OT@4?&Yv!&~L96b^t<6~Y`j^lHBTc1)UUFgi<
z*jG4wk!Fj#0L&}$BsI3Sb8-YS)=5yiqn_p);fWCSC$yIGS>3|Dj<`LSO&XAZ;{J@$
zBKVMLvYYyhAHi=TKjdg&(rZ$SpdUafH!(6<g7R4WI)d_!*FCLesip_RvJ{@22F^{M
z*8?0tGT}O6fUYmoTM~b+X59*U#EPL*v$n5<FW?L<QPCXO9mAA=5Zc_gS9u1%X5&R{
zUXf}lYra~4OE!&u*|J<d{B}WzNzp)cane5lm*#_iSZ|yA)q?-t5guo7#gS=tf<NNg
z4%{DL6?g*oONjSmNZkg+TQ-{W^O8v?Cu5ryP+J|Y?;?}AJ(>hr*O<dhS!Rqso@5K=
ztexSyU*=|m40#H_td`+>3c<_7QV!=`Ffn}NMlcq@Fh6NETdlBczg33Ym38J+Vh`|+
z|FlYzHF$OLsjQSN=7GPA#Zy`Q>+E=v?GLeJr||}?u?yU9upL{HltWB|iGb1Namb4O
z;77GEnT#$#w84TJL%D1l?p3Ass`k28p%6bOv5ARE#P=nA(43l!UQ(a{EBf*U7h+Q-
zK<kzTI4UfDD?&hohACXt0Bu10*ZByMoQ(dz))@FF^Um{dk*%XaT!!%9ZLX5ydm`g|
z2#x-YwzO>jeWHEil|3cVs&((fe+w;LPN*)F!$<!RXRafeRq+zxsLPfHS2XWAt=K6(
zu{#zXAXWGK97l0qm+5JT*QU2kpWgcZ0xLHLm24RZ65W64!NxQ$+3f6$5k7g8gtRMj
z0DxKtt=d)x%!7$s1_9rhn#dPHh%>)vrhTB~UHCXcQ$ZU!?c8Au4e2G^L!6*ze?Y*m
zwkloKbsmW)Q1+t8{8Rw4wp8&mY)HcUM$_KNwEFo~AbS}#!9Q1KZ%P8~pvVkBZ96<o
zxj`8|HHJLfj1zkgr8X~w1ffd;f3QRYBms0r&D%TkFCyF2E><US$++bhENRitX1}P*
zPRO1Q%z?TWG5>03EN!gb>v)VL=P;3=SJm5ef{C1A&=5?de$=aVrZ9pVT@Z5F2(KgT
zahp@#b|8-~KS{VR+>iiY(@&KPt-Eml{ru@TyZx<89lZPS+&xesV<?w+t74OQ7v~uM
zH29K(-?Cg{v@0{(M;`ZwG^q=)0wbPMNa1-*itXZ9&fF{7kFZmd*^8e$-3;)Ljsw$l
zID#Avt0)bV$!f~G^DD5LX8yUfEHrF-iy(M&rtejqzB&rAINj}A{cVv$)KuS^tP+aj
z@#Wp(LnX1fG!IpHxFJw%#4oW7=>Cj&(g!UE-Q+ZO#_mPC9lHA8CFCd8Z4vhAaL@x?
z-F?v4I;8cz9ba+mqzG$o<%*R?|3&^yJ$c<B|JU0emRh(lBy4p-I12n~L2s1IzZ2)D
z3Ho?yp@Sgb>?y_HoAXwfbAOVa1B<-QksM@JdL<j5N?TLpMn!@gw^KXZexoU^_F;gg
z6c6aNg)hO8Fb?{1b<n=kRz2@Hz>>LsRgkKv4D?}tY$AlrzyiAibBJsk3gVp5E#@lT
zINs#|s6k8JywRGLXMQx?)<EPqghygLkh{#atAejww0?DvKXzp;Htf*l+97K#r5f|C
z3jocgYfRqAE5j^V6992qxpMi`nKoNg-O(!w|DuwPGtR`3bzaN$V;?>{(9l1+x3-%6
zb?bfj?#7N#<n0Kq21&9ZQGY!2`6!8x<wE}&z6L-v{bxJcpVGEB*KyTq^$J_%B3Yxx
ziP0YRl^3YSYVzFi;B2i?iPm%sM<S&(kq=@284Cj34fb;nhn1SKtx9JHN^~cKv16ui
zS%vWw_~obBAACS5w*N>SATUp8fqIu5DC;FLuw1;VvN{a3)XLj;Mn}?^7BPuFHT~G%
z7{Vob_{rN_a9r<Ll}4cN&#$q9-$YcPb%Cr_W@jh5ZBIQel2YqT@-RZeE)Z3kalW*5
zZZzVz(-KA!X`cWIF5akFe7-ku`)^cMVSh9*#5^UnOH5^OfJ3ZkZ}<IN<<M^EAJ+(j
zq3u?O+f!Hif3D##e*!dUhBy-=&X$SjGAP-8*z?B4b~C(epo?ite;VcyEuvIhX0Eh6
zX|}(O))YE=EI(gx_<)nULkof)+FdqYKpO%mL|xL_Qpf_mIlcyUle*`7@pXl<?-Tz4
z%L-~`cm+DF9da0sU<ZQYYdmz&1tnPoWI-kV)?FPm*L{HpF);%T=+U3W#klTbB^B!)
zD^~+>@Ot-V;;NTF(ywB)J=%-~=*I`h9@PO|oQ$e)2d>J0?$PxH%=$!G;$Pl`3Lf=t
zhLN6`O4~`-z@0vWAH(D+o7UamAz9g~2H-=A(HvOfzN7p2922T@LyGX=h04@U$Wc@|
z^%FBnzgQ?b^8{LbywM(}o@3ZtOEu<D9H#kvzZHJC?}giw4+s6;ypE7@yo{jtclR{X
z8gEU>c9_ulDRG|o$0<bzs*aDP8=h-;d+?(_BkrA_rv8X!g$P)(pyph=5KsCjUcGbV
z!?b)UZx8HvN1H-CYq^K^AwTIuq9VUv!ewrE3YFqD<LRGUdns&o@ON|JM;$OyNvXFF
zC*$88UrTb6MLUT?gz3T{a?9M?wpEw#fbdwt&OtKZ<2Mj4xgyzSi%$9TZ#3DP>USIU
z#Bq@A2atqD73e<qeV<d2maThJDS46+>`ecL;U(jbB?{R0+l&Gldxo4;uCvp-O%?$g
z5Z7u+kj+z{HUi&PfvTA~BpB6c?uIGw5bB0?8Jom*5<ULp9d{eVBotSBD1L#zg7UP#
zop)i>JEAN+cVh-=3Pc}jOT2$6YaSG1s>s$oP-`jWBkL{8>y6uKQyxiyU(K$37oZ|W
zRQmPYqy8#Uv01rJm<PV^SAMDTd?P{s_pxvk7bj2bGoPGT!qjXmSto1yT?uff&VHSo
z+$_HUhZiYN8bnJqD}y4T;PcuCNWt5IK>LQPL>09v`oG80`rsKaRh#}snLi}lpNFTI
zx~V)pbmf;SFO~b^y>SVh^_05R=4GP(-etY>8`T_jT(Map(*2QlO`v-LssIRwE9Q`X
z^<DpUVRN7>4Eijdpa4(ZY7%H<@XtHHy)#hSkV-l2;-A9&aPkYk*hn5!q+Q{|2R9HA
zB{+3OW&3mICDa$ZHI$|apYoLQk1-R-brci6p)TR{Cony>RH}v`z(sCtG}S@j)=I1f
z8+v-L^zH0ISTwTRBQ)SBdgPm6c-lM3sOz1lySqE)S8py4RZb=ErEd82*W4F#NnKa3
zBK(l;Sqr7gJU-iDb(ugfQe!r8>bk@v=h_36>yke?hrTHPO`X0u^u+zyQYydS#&z%`
zdDL@|EXM7!@~-OMVsdfuUeM$Yd;0h8=X3IZ^fR_Nq2H@L_-+E_Wjb%MrDw4`H?{N3
z{!DwP;kn%lDI{3~70!g1jf;-4Ppli^BEFn4JGlpwgAbyHkGdT`B5ON<q|3M4+o}~o
zs~xs6NK`i(K<bJ=OoL$s6=pK1Z}H#y6V}}D36tBXGnW=afSD&SvRutksaPFW?O#67
z{c8*+Iqw-BP~G~^LPk}TZ;UG*fAHXTQ_wtfs}oBm(?k}p!E_8RXoq<K6%*3cbrmi@
zo>-m9PjtLroy!k?RAw*p%oZJ?$ja!-YfIVD{}0zm!lJll;L^j#&yXX-9y-gNtLL}k
z$F+3-^n}#$d~Z`*=ea&i%zU;;k^YuT^6yzo8FZni0S^&WD%ba>a2%tyxM#VUoEH|W
zax1}yFSIH`p1cz)sW4kM3qN|ugWpo|^bw=t-jFfRHUa9e{UuP&DZEvXfJQ)5*c65p
zC+E-1Cf`?ha{ZaNDn8MfT&XyZ6RUn%D?DDu%YOPJY_PRwlq^M$jTN2CA>EICEhCIj
zAND5BJe!$Ofz}57Vy&Ll*N$HOIS4HxNHmVPurk82H}IES!ItD#;v^a)ktZX*_Cx4h
z!M_ONA5yJqZ2gZXtBY+92icdWChrt1of6yG^^drWJ+^oFDLa4JuEdZHm}~Su0L+!!
zoJ=O5ZjcTlW;J$a6#j&M^rM%ZwhWCdo+;N>9N9c}G}~-qV~c;xJ7vO2=ZPnCebTr7
zo8|$cRKidT1}XY+-o!cg(K2}aJ&2QD>P{Rm=h*i|U5(kQm~=}t2kgyQq=#NapO~~{
z&*49JN}V{iz)+IyWwRbPyS4%SS7>qBUe{3Nxo1>I|M3?&wOSW6g4ZdNQyp-@EF>@v
z=3>U)D{po%p-08g18g3y7_wa+ksyfi_0fO6d$q?*HQ!48#edP`35%dOGTSyK4j%*h
zSiG}IY2S7g9o=@_+W-&cO-M>x)gQsRC?`2eyyGYS?8ao17&E>x^lE?_2k^o6$T|P{
zt*^*But)$<nnprB?pgs{jYycuT0nIo(K&oMZRDZU(0wb8LKDNHDmDwSX59+}0o|BI
zox9_eZFD3*eR%?(kWq<(omR9Bg_osvwcfh5@g4Im3i{x9G@oFr0#C8%c>kK)=fkqx
z39xqyQIhnfbYsv#Q02qQStY!(3al>J<7YEqbd0>iY#agQVsz%4l;KQdAr_)(Ir6$L
z5u#~KcD%rLi6sT6{6L)wZac{o_*Wajh^D^?mrP7BeSp+GuQ67j7?8i%Kl&>AY3EjJ
zAN$X*ZDLN_D>C;7Nu!C?`?60NVoZneiPX^~noN7o;@QpI%;7~^i!zLOfRb`TSa-nZ
zduX^QQ2GxBa}ID@p7Souj{t!y0?MyQ_QuUsSihOG!{`IQd^TC^I=7rys#1=b5-n|@
z6A-O$`hDLzO?^x&%psyy?+<H46-1IolA-mg!|`KpE*1+E6Dc$Py-)+~unBs{oS>lN
zwEvj%%jyf&+#;2&)rG9qd52MwN83HF#+@fO4^~P{bMA0e?dZN~Lk<4BBk1>cweH$J
z1y06k>+OpG<1(J-T?k%$3H;`}0UqB2Rmb9PT;QlH&bW6T^sTyMhX)<PcA@LzO(dm|
z>?=jJH#}IG_QL(fQ*9=OA0re+IGR%&9XLggUo38%s7Q48Jh83zBMn{n)SF9alwcW5
z3?%URWv(EgnY9#!iXfqjb`|c_9nKR^)L|@a(w8E@8CFZytQw0arx`5yZeX~`QL0F-
z=_be#sj@wuT*EK&U$ST0ec?OnfS+Y*<YSSSoK2o?<aBczRq5^i*GJfaUk#(Jpx<WZ
z?g$cnB^PT?eLVVGgt;#EzB~6XkB=#zda_fLp)ck~YeAOap8UOt!6#9Bis9>jHs+Sj
zgtPzD8P@T9`Am|!=lIGCGH$X{E!o5GIQ7MEq}CMrbk4=J4WVhK{lxKD!E-k*?lcX%
zSMwK_DXpUV>B&Hbj+*{+D*@gq2*F~onRkIM%k7=Mp?)!-{Hddig_yVOrEy!u8{QMG
z`S>2psI!bd32QtSdI+EDZGcy<eQ(aD@?fE0tn=;Dr5w%+*eQUY&wHsAxYhLSRYVYS
zk<v4TrQGxG@(qg5t~I8YvMQ0PNSL6brNolQ{Vut`DCpzr-H}ffIAN#x0H)+BN03?F
z+T!xC<lcHw$Av@DH+|Oj%We@lKH#mF9Q2zi^+T3I#@VTzBREW`Dssd=_SFmB(gg*U
zbXRAimPevW*)Csq)#@E(&5V|@i(hCs+nTYFq*`y|KY4V(KKYZ|2ez#_mO0ntQoo-X
z+hE~H0`+=qhk?0>n;+b18v5GvwQJP@R;N0*#7WE;m^(BHjQsH!`_NtT=avK7Thf2{
zn6N{Zr~i0fZ2<h|&yg329UWaWeZ5M`s`!STQ+N0Hj(za&qi7(BUz>5<sO^#TaMjjx
zRKlh$TJIn8?PT;x{(l|gg-Zace~|qQsXl1WagDXS@4`A?)b@9v%x=bsiESJ)jS`G`
z2-&TI?7plWXZZ7I>dph+%bdEt#^8azu6H(PkmjY2@i$TX!8~Y~JNSuX+%3_=6+$(o
z>Q7XO>l12mQi(C%vCSzKy%?5U=R(t(5i;#ZF#!a|2;x%Z_jHdez(lsDtNjaL1bsVA
z7{a+;gc@|H?w#7D5v&6yzvQ{Rtnh=+twQMteP`{xu?Ll+ehEfr7nD?0m9DwehE7o5
zN9}w`{Fs<j-uiX|tR-RPBJzEm)~F0;J;UCt%PPoN@_c!|R6E1t?^fn_!bC3}_S!QA
zAIpO>xt+mb{uLB5$lTjnn(>Hasm9l@Q><#E=ao)G9ZHPi;f$3S=`PT{LH;?4NU_83
zJoz<J3K9u>$`ce@&mTQzz<1~<hX<EU@WB8j@m75e4xGjfF7LwE6_O<{Cwrho9`JbP
zT&pTv9!C$3I#q5hAkQoETTaY!pU+P=%Qyl;xr-R7c>%8qnQ1M^!0~ujx1F8ALQ@;P
zJvv?`@l~2cJy%F7#1dLYbjoME-k0RoqkNcO*Af$Rz+QM&nZ0evsJ469n<|z5MR{$2
zL;XI@#vA1P8FD1(!7e9*seb=&bJ~D=d=DM=wok9hZBU8ht%>W#we!0+QA?TASNkGB
zmOH}tL-GeFBBuS>yu{rXKgVm+rI1mJE(<NeSMC8{-FyDo!dZv=EA`ZgAPj7<)Bb+S
zpJIbdBARyRbDMX~fLPPaSJZqH$7`XCgB6x5n)uE(`D*tg0t@RipYQ<Sm2o^P#=Fp}
zrWk8gkc_)U?IwKHJFgvuII7f~VvV8ZcwfOQFSchcs`oHK^%KN1LP^2jTJOEDt*I-E
zSfCK#?^Sa3BYm0kvvWLwA(Z}TL1OU(+CO05psvkflo%rTJ+7YDa7m3g|KA>>H}QvZ
z$kpbTO~5(yq{aJ+`<_<lYk9rRb)l~w8ownGI@NxzHsm*?e?#ho3wO?C$u?k@r}(2(
zk-tLK+p*&yN*!lxWEE#ZXQN93R|XH7p`K)Sz^W)fUcBMB&XcpvTx3o7n8J9CeI2%_
ztU>dM(=j9b*G-MidK*RRDAxg>pI1(8KXu^7j`-azBZO5`HU>JJWT@aEz$3G%hwUj5
zQRF{d%yCfoBjFajq=f$BRnBuVYGJ-7u`3JjS{39zJvvl4n0;bR3aMU$gMW9>KqUdc
zwjfbq5HPW#A1pRJ)A`E*o#5Ry+|4MgEWP5mpI<}QJAN#(@be5`eOSbILP0%mDm#CW
zX8F;?Lz&T8^8f1k@^Ch{uKl1|$E<2rini3OYO1-E9z9gGh6+(qR1tHG3EG-!?4YVj
zl%q9AO(Eu4sj&!=nCBqW5VP;;`~JS~<$T}opX9o7UHi(jpS||lYu)Q!_f9#~$Gz>s
zZTbX#)fXodAcWe{ipeP5WUM$pxx;P4tdQOn0Bb9}hREMJs3g~(d=16n1J^$CG;x0f
zzzcVu&rzh{D;{^1_@f2i1~q$%^gViD){G^VO;JCRt3{-gxgfG7ik)N%cUlYCJhTf}
z#*#-BEFa~MM={Wy$)OwSX96fL^XBGeU$|g|YWLmc<eywUpRkI-f$%O%=hw%F(&1lK
zi9l+WMQ}@~B3f6}JRdD~FP9uXy`?|ceyEDxQh}>zr%8QmouuJD?x~|}_Hx-AW(Wuz
z#2IYjX#Lni3K_h4riOxQU67vNadCjgvxNDt>GK#*D!~~nJfE#0`@WwK^aAQ`n+K~-
zh?&+<>(=y}vgf}<5=+*G$ZY_g%MkeXvPauHy?*z$+n4E|9BPcRu7=-lNAedtuk4*I
z3|_f49+Y2fp$U6LNV8tVqe-`H=pNthWqT24k?&UeetwQ@^|AK4|I*&fY^U#+ua0P&
z%IaP@r*pgG(8pd+EnNo8Xe$9^#xxMV|A#aAYV&Qu|6#2lH~u)Hx%VH`a?=?4p{g_S
z40iHc{^vX@i@`hn>%p6<`+DQYjikMaS{dybXI6hc%Zq1={F7DzM%>XPWVxjP3m2tK
zy+>$7YUOm{o@lCSARj#17^=}H5bU!h>%$nlCiU2qq}Kj&1i6(_rmpbl=r~(Q0Y?4#
zVJ!<T7Hkx>uLxU$lL3MpQ2YTXTV?J|-FUxdPySl`!l!F7hSG6R#j9r;Q&K9X>h#$f
zuq%(9psr5#_Pa|5vB?_`=DSl>Hr6AO<~LF+LfjIeUhcf-itV*st&6}I8zr@KiV>}&
zOrC7S5?mSq>(OzN?H2izJ-m`7N^C%i!8`i1^t++!YeVZ#)>e|*^y?0Hec$Tr*PhF&
zWuCCt;!Ri~w-AGZ&QYcuC?l1Mh2nq01Is;198El8b>F^C9M1vBzf<EJw3VR8RI@4H
z>}b!KXhKHfY*D-~2dT3Zftp|yAj%!J%3JbDx9!d4P!n;6CzEYg?{4~7CpcT(Nl|b=
z^l~nd>-29YrBo>bEhP~RQ!ggs=x!>ugfHgHNK0DP!tH-GXT25eAz<Mv+k}FYJie7!
z#b0%3;G%9Y<{U=yhZI8)sa1^p5a5Fv_yb90UZds7`kw9?&{MSq7Olg7k!Jaanyd6e
z)s*|~3h+A%x3KGdVLX9>>*Us>1u(<+&j_wdJK~qfIlAg3J-Vde8YExQq%A;A$n9x`
zjeK3gM}d!2tl%WFXkXaTefx&OW2~)PAxQ5`liB1o`H*>gZYcMf6oj^b;Afvj<hJL4
zNX$Z<nPT15VxVTd?hS<uPh;<1cGXaPq+2X&R&0z@aw##!&5`Baph^_!5lt2L(aMU1
zqO;L_d7aUB8O-M#Vb-6LKUDjz{)`uOv#Kkf2<AF}%pk^MmDe~@d=>VH5A*Dkr{=!f
zZ@`i0oUZoC{$yqu8DJ+*5qiK~#pnW?hdo=QuCI62m)jY-D+1Nr>GI``U^!s5s}tgs
zq3(|n$v6kl8ska5HtfFOK#$HbKKKOZn5gh-+#`O+LxR0lm`6u~3&Ha#>6T;+-D9b=
z<s0jIj{N8mkLCM@e_Lkg!Ke!+^`sm{-Xqv?@sJ_wDo>&cXvAWMb&UQOv0`ep)a;C1
z#wA=hfCNq_>LiSE9E20AKqeM{+r8$60?a(LkxlZ!&ZfQ~GGLcbxtQ3}0<S8u%^07s
z^RpakUGaVA`N$Yt-DdW&!e{<K&>aWWRP{VXqu2FL;-1rZX|TP}s(PohRa&CrQm*_#
zVozMmaUaIUsdkf4^(=jzNo-3>jGs|^<_p@?=ZPr30t}>*@IfNErQjI(aIi<uoPQo7
z1}b_}W}2ASx|grC{KF@7HSOWFpW_*HiU)ry-8(bZtb~_y1#aU@&1Ig}etDnM0aE|{
zAN1{wBfZsKI>c2301+^07czwCB=BL}HYk>I39P>2VR06cZux~hQa@}eRR9X&kF2N-
za~VjrqQ~OFMbI@XQXNbbHi%{&`l2vTqMk(ZRrtyq@GM+hXpK)f8v4z*`|Y`!4FB!9
z!r0pqL7+=Tf9h&K{Qb+O#p4hAta)@k(YCPl%03;M)YonZ<dcJYwHv#08~2Wk@u}#B
z51Gc#Y!G*Szm8T^J(htzievz#Af*Y1mv6Y1gyIwh%Uq}y>EXn(4lIBSNPXl~?&oPt
zSyYZ2xzD%+z^p;adFc}$%WYp45B*VkO8^g~JL0%u!}F;tjXqLVGnwy7h6?H%9aKH-
zABdP^Ei2EHf0Ttp?hC}0StMO>h2lwh5;zC65lUlYQ6KebXYEB4&=RoKI9>vnvgrq~
z%LK2BR5P&~LO_mI_-MmMHvXI0K-Opv&i_weQlx+?qnDtkIyDhM!-by8zi_J^=0_aY
ze2o|n-0FHVK8)#e31XqwPjq`ThiP6VM1%J#%T-J&ceXOfHa-`-gCSO4=Aiq#%Ckih
zIE-f#Ig3N-jUVC0+Q>HdnC!H2HgKOzNN+R1mP|C!=p6){la5su$N9ude|m3=E;fR9
z^a{5gq-H)=#J`phE^7CkF(@_10Ys8jb?I0$R<V4zWoin>Ee-Xt!rN|YAjb;>kj1+R
zA0To=rIXu};dNd$dY2*|myZ)GOeyEzMO2z<yN<|Rh@;t}l0FFv3S->>gQf-y8X$hy
zXC^U)45bMkv<{D<f=IEz(QHGzX3D*PwG^q5i`r(JfgxmIgJmgoJf>cC_&ftm8+$&b
zcK9sZ=E@{RIC}BGxsK2LAkb9N<dKyQS0U|aY5hj3P*C@BiyTi&$WYN5RiPO`(Q&y~
z_^MS>KeUh`Xz~R?tiEgrQQ-2!bF9A$*xb`^v-YiS1dPOnnT4FsPhmF07YyO?OFvTi
z<Q=aXr5(&n+Z-(eO!ZXz3Xqa^ZAzu&lh~%=;&dR&SW#+T5v^xxV@uWlURE#7;6bzz
zJ)-%4#{4e>R)&7qryFs&%O&3TY^{rHtNv~eOB`||HR&uQXwJ9S<w)yV5T`KH!-I?d
zmjH^S<1lzTCm^gM^g`S|m+;hc!ZPo8V<HfU0LD};f!k@*aiT#Pwyo#JK=}<Lsnc&6
zlbQ&!b}D00DeAt^q(qZPor*!yRXMLots=e(p0_iI6^abW1>cGiTyB)Ig_>%$eYvir
zsp)V3YpQMDC{0-%u|5^MJ2J90q?gBjn3~J$Y6Sp0&+<Kq%*5Qj4y|KXt9PIDCpEOl
z6YmXAP?LnQXkZoX45*1~-1xU134G_OsUJfZHc&kH<DTHPU-hrQj@^$6;4*d->2tf4
zqi*}s-9`xZ9;KlxB>he>wR&-7*U$d`5v6`U_<24eFJrx9#Dnc`JGrN!NvfGUpLp6z
zxNiga%yX8f9HqZvt-=l}1*`HWZ$SzaK;QMmY22q1T4+m5hT0W0>cht4Nxn;#-}$gY
zL1v(vLw2=g#kED%$~Cg8s&ZZ)Q$KduQGZ!gyRViX7rrti6w7t^7!t5@odA1tlpHw=
zkpe)y(Sq;_E&s=-^Zu<c+f|X=j4XzC=PIoYiF2>NK3Uj;(qxi@Cr3i3dl0wd84+BT
zHD519TzabU12|nYqw8#L+*Dx(0>(6h3TYNk(}513Yq_X?N$MTW{PWQ|i=eSPZR@XM
zY#C`-kE<@ir^VBIn|+{(i6(cvr&`7Le^@sQ=a=|s7Y3HLhOKbg1{Jfo3q5j|K)x&H
zpo-atJg{*2a3M#*!PRwZZnG>;RQi+J`uCOxqUf=-<VUV$Lq8WVh1?%GU*ND;VS|;a
zKz{#+S6`dl0KfjEGCdBUrk;Z5h4JUZE8~j<s=NX3`Yp>Du*j<;6w$hkAH69uwO`u8
z$4!iFv@*s+`+Kl_cYW1-xB@-)ou^kPAcSJoS1$A(B~8|BC<d<$>&o*A-bA_C7e60L
zZV#V9!?!1V<l{u;Ug|nUhL0oo;M3UqokA5pp9?in6-BtcFK%PqfE!}ttt4i?yPW|Y
zatQPyT0Er<t*^JsA{xxU=R}WJss}D^4-D*fI(9>9{pJR`5(g7Gu`iCW-`ad8UF(Ek
zBQQd)T(^&qv;p8^gkb1-F_pN+U2O1QDkjM%fY|aF+f`_~LlQuNS#bCaD-QmPsRk*@
zv%56U%p80)h<RHRSX)@SM;#&p+SaiZt6Q((%OUvk`AUc+DqVd0i>#H=Yw0pJ6$`G^
z(7PZwL(Q2(m(Q)1#5`Totx<g`llvi)1#PyjwvNuVCnAZ~MR_F&O6$<~I$|jtNS#Y0
zeTjKh&kc~d%UFG)?5mMr7ClMJt1SWe)qHdTu^QnucQDWr?I`o<ri#aON8)Z)R#@CI
zbJO@+ySs|a;Lh#wp_OJPAsEqbZY9Q8dbRa-i;qvkiV?ZFK&|a$xhZn5L6#zSzvM0^
zHzzNX6!T26h#X&A<F=_ZriS<1o{-4x^Sc-oV6O1iH_>FDDE(*=yZfcb*w2)=`~BrA
z%8h0wH}alJdtfb4Au`BNyLYCj1LKMKkfP2-2ZTl8F^$ft>VCN8HX+7_tMnNlCS0>=
z2T)gDvC2Y3=79>*b$2f*ohfslPC<xpDVBN=6>Lk3CFsGC-(T1;aLp8!E8X1ajQ)v%
zby-!Y`0pMJRBo7GmHAXvQ(9Uv8B8;!xnG=K2IGQ_7!k1dJAe+%8X!vRp#d{^-1lJ2
zW`=jBW#Ue;sdY;^fTL1gp~^e2Rl2Pq?FLFEvH=&*6kBsV=B0xcK!D?!ga0r_(a!j1
z4akFd#ST*Orhy)JrA<aE7P@pYZeQm8#?Lt@-04a+-n`ZID5=<zo9?`XU!nX<r7q>T
zQ4hvvE0&eH!;L;|-T=kjvMSJ0>z!6q1z4X6%tnGw(qAgt&9;~DZ<nl%a(<AXVuY=&
zMLJK;6?g&3>E7l{L|v^nX>aove(70TaKz1^uM5luK(9MePrOZg)#znK%aE&@W%Bg9
z6Nmr8>qcM+P2~eB3$s$lKh?Lrp*>0(b#4o>TxmRhlWPp$+k3&{e<InVknHu!#x7M@
znT>kNlpdG8{%wOgm{S77EDabJW2zy`ROO;mR3&saQ*U}B72(&k+NAt3gp;mxJc2>%
z+IYvHaj)@+fVFO@Yqf~d4R5M%iu+hjVrJq7iHatAs}wqyG>5jXLmvLv>U^67GqS1;
zSReSF+%nt38FGeL^(OB*uHa<_;6ZcMD{JuO_HChiuRDety)LA=WkrE^<Hb&&^?blO
zK@l;K)xiOU02f`oF*S_}3Z*XK$%2;*;?H=-nsZP{Xqn8dyrvKn)@7|oKK0QYRwe#I
z{&#%j4yMItL29Ko)_%FVAd+QetECz=d@W^#Mp-%##1xq~u?I-Rl*IJ*nm&RmF}o4H
zXxWl%@qUi5;V2RxcH})6Bn2({SpFlR3K8hP+T9}YwFkKwUAIemfGhk<niqPDMhP~>
z*)#|V^dR9YOl{rKbFOF**Fp_T(m$5w_wF(Ws4fMuknWGzSkcz4G`HF+x1AChe*V25
zueo^7Z6i6lS~2hX(b|9W#6!lmf$ALvW)A?s>ZUDs7MACQM`<NjQ(}Fm8x?$8_N-*_
zHmb1m)o)fCzz%}Ub|AfoOSt$)073nzk_BE;Q=uVpA-GPq^-%_^(sCmZbcfc9*IP97
zFZFjVF6N=qjyB)kE42Ceq1I;w-=C^ll6M6DtC!eZFzN^fKzjC>_X>dL_X%Ap+J?kG
zC{&Gm6F75jmqq+Pz%JX$rSQr#>1Y7iZI{xZ54oV_I72iKXFm8TSgm-l_YIw`_v?MB
zDvD~2`uyR$Dv#PsxFcn1`)8{Zqw{v~ai^Z7&O^B2a9&D8_0fl}lw#IUHU^7^b?*!0
zFM;d%G3lSrAFE~qN(C3JM9gQ=#6u2H3gTiC$<pTRwyHNGU4oVCEX|!3JD`CxajPtO
zb~o%hP=lB^I+)@TMIPp%thdJdK~|A`FfgulMZrWxAj_p!wVr6ihZ%otZqQ9L3!DI|
z9-&Jim<y-Pjf;EEB<Z6ywtfzT$<JW`IZTY`d|#W7V_~XWp9<}hsarRij#_ui(+#E-
z)^8csrrjwjt>y`I(-Y6@YST3!bD7Ox*?uVAlr|m`?9>ye3P)RLpF8^kX}%gb_ead7
z&_5f=Qt1z;S|uagfpm~L^P}}3m%m?17<TnY`bU{DDF0Vf&iQ;_MNXJ?-TGpbP4HUh
zTS26|2NDiP*dhz3gI`1ot-|5cX6k6{y%F4kuMDiNF64;XD(=D6|7iyh&WJeQWQG5`
z=RM3Cj2pZg$_y$wW~udmOx`Eg^edN={fZqo?dc#709RmH!^)epTQOY_R$&Sd6Gz(e
z6c34%8r#=F@r4(PwNm0Ox&t$*zZDv$SWG1FG|&G;_;tlm88nD#IfHsN+PC&S%R1m|
zb2dg9GJpbuHJjMeq(kw?Qm#;>kJH}va)H-RfRJ44IY(&zT6Z+KwH3R?1TqB0E?_cp
z*#KMPeHc;YLs{S$*2re{%`+r)oKjkdTtGhla5r&}&c{cvhpJvNjT`fwefs**mhUzh
zNNg~?5!=D03U2`@%HCoG9m2-Okz0f~Q~5i0^+DT(@C)`JB`-4|bQqLM{aeDFrSiw2
zftOq-2EADzWz&AtFN(yMcTIaO=T*H9kF36Zeod-<f2UMbX<u2$%Is-Vy;y_Z*ODUl
z{p}nE(R&BK9D}!#M5S%2W!+k?0&O6<S7CC0DJ(EU0e{YkCN4n;`vCZ-{=X}Gx)k;%
ztEC*fP0s^0X1C7mMQT9e?uB6)`J4@(V8GTw!)K9u7hxJc6SNAwK{~*6a`k>GUf^wq
z_pbIrtkP$EYB3l;jiY4a$&#1ToWy;vL-c@>myLn`TGr+tY9dV`W^{et!?AhZo4ZYW
z)~G+{NPbO5R!OcBv|_30u1s@7Ne-~Fk1B|*gD`gUo;1`4W#8@X0<b*kh;lRPh;a&0
zQuDFYlxh{;dTpG<ijmGgjg|;pWAw_A3De>luD*wn9+0Xn?-h3yz{>>I9<-(*#&`SI
z8@QN*<#>1wGUy=B6JHoiUK3N8f?a%0Or`S>Dc%j@_RA0YUBGFQ1FVVj*ebvNRZHaH
zj*BBBIayTp%f|(=&|fIaU1EjheCDSO!5Xz693Hmco>ys<oz3fi9c$zWb9Gvr+ROrW
z!_Ec@88Fqv;-F|L70Q?O$z@M*PcX+NfIsOKWt-j|ko@@F3!>V6Dz*{099~&^Md35d
zuaKcFS{1-8$_p<d_GcTup=-<B))s!AJDV2xtFm>tofPnNLgLOFOu9SW^DocGW%vet
z2EWAjjr%!J60?M7V)Y~)1D~1VBJJyeaJ45ctn4{e-da9SGvA^zq!jv8jFw?M*K~Ch
zz~<w_ZCwSXa(lBlLdEKL+<fh!E8PXEC8A%iUe&1aocqy|Q!)=TG_;0njNcL2Q>O)V
zQRIwcX@&u14esUlVf8?<xHbhiO8)PTWnnxM_gsJZ!17O~_!>B=wM`O~5)<2=MuG5+
zQBq~pRX6*+(4v`5Xuy*A)6+~1E{2-vrw<m9%i*zquBtXOq3MIk8y^%{yh`BGSjw#M
zaw!MM&bHYO<up~`&5zES9)59E@{C4PSU<i?J+Ag|2KOce;mC?s>q3zW9$s04^g`Ir
z{vvE8KbY6t+i8xd+Zh>D#klz-#6K`H77XE}v3TGIVRXdE8~NvnB)Z1;FaX70&rjD+
z93bkPz!a=U?6ScwOYc_(6B{z`i9&6V(4Mu<o_H2D#N*(NMVhwFg^84a^j*$U<e%60
zK?Aa9XvD!RRl7RxE`Cq@b6{-+Z_U6}NF7qvF2g0LgfBG4y$YjSbGi9a#&>)Aw*1-q
z)R~0CvRBb1t4E4KPHbYc;o`mtjat2m4z@_Aqs`6PPC>&o6|W<FSTqJ`1+F7JbK{;7
zN&L3>W%l#45p?ly8HQ1}$5p9j($64JhW~@8&<AUlO@cQw-|(59r`Fo+4Y|R*Y&va&
z0It9p514PS|9P)EVVOte%g3&H+gYBhHR@6%TYY-RbBlLDF}LbT&c*&kDVo`9m+?sq
zJBJm*#(<7gsPj<V@EMsb5x=h`EcZGtnat{4p`qt3^pq$v&S->N`SKS=RY_kn>mMo@
zOdP!XyGJPPX5i{(o96LuV$S!w5;+P&2hu>Nq9%IO4Xotwn}j-FL;X8N%~m-(5CL3}
z{jW|biuMuBoqxk-|N5-r1`dt;h~<G5-viG}AlZZD%&ynCI3^Z;4cY7NI}>A*(>(%C
zYTqbc0Y$sCSzCAMhh-9M>B7_Bb?aT>(){C>nG2HAOOB1ZZ42;e@*8Wdf;$>BY&p^v
z8K<_qS>jUeGYW^AzFGOsL}7%!Poz%2U_xhNQ5u^JNQ0<(9pVS$yMy-@7jxtt9Sdzz
zEo~lI|6xejq0ohmClwo)<ywQwB``|}pw_;|KWv#0EzoA{!KfZb1u!BHQ4+`N|7nL4
z&ajTPl_s)t7!ud2$G^b*m~^!>o?R}c@Sy||Q&l!d<PH#dDYfdwXFE(6t^D1ilf(6H
zNI!jmee&%2S$}J7Y0#Bj{b<T8T)}G^QRzNesQN_C;ReJtezGK0Zn-VQZIXI&l!fvz
zBfZpls{C=$Gdu=oN#c4BBkcJtZ6?1JJeqQRz}17Y4tJgi_vl!UnOh5FJ%+rTHwC1)
zx(V1dE-@Au&`DhdP~iTX{W5Wa{nFR;5_tXxql6qV<tAT~+RpOlCktVYXWytdeXlRM
z7q@(fSCwHB&{!)Mc#1Srcw-JylNw^yd2)wJhi)=El020m@k=7SG8bTejy<Lw!g!-r
z87B(1l_7G_r&<lUtNT~nuS~|B|8S-#VQ>h)l(X1reze+Tb96A>`PFFrQ)*=&#2U=;
z8?5qYS2`-SZ+K>%4lcWAF5XSnjRanNXhB#dz}ZoclW+fzE*TJY7)@&jg91uhs4U-+
z97s-?^B%^0=PIAS+Fotl%pduXW*z37+Fd4XD6@KqAH|(Tsc&jSgEozeeK(!dd)0E^
z1yDt4&=<suWwlxx^Svrh*{HD#<T~<loRqf8Rk_T&@4_Bv4`LFFToTAz-kufzo8F>?
zucTAL*>MgwrUnSw%tXZp@BO^{)d*1xqR0<##y>J<auHC7{pCK8t~M;C3gH|ZBd1nS
zd_G;LIf+FDeE61pmtt)Q#-P5y20cJU0Yy%4*Y9WRH8I!iEKf8DA`chZ9pi$&$Z6~u
zT)*MXPv7<=(@QIVgyw!=`T%!`ZV)oNSi4D)!Q=3`)ytSZ%cL`cHw=TmsJ}Arms{4M
zxq7BDBSBnSl<IO{-kXmbKxOWDkL!35eY~Bdc{sYbJMzf+FT55&r60D4GY!kM4|O3l
z)8<kp#-gvT!fz3<`kTse*ndPu_v>Yo1R|wY2Z#o}qa1q78eZ9Qdvm~*Ci7|dpWK2M
zd4jfk*>&2OHBm9lb=l_+>3P3PU|&~*8Y&oE5tt!-_03OVl|#e<`tX9rZ{4yQqdZNv
z@{!^FwaIxxyu%6J9{#RdNjS9O;ndXkw~pvk)!^yo2ti2;N@aKgpzMM?zHQpzWIbAr
z+EC?~cpp}wbqpZvYR?KKRL2ns$LS|LeV+(m&FaQs#!9MIzZkM%i<~rLI|4%OH$W!p
z%IU3jzvgz>$C2cb*XD860|X_w8V`>ebynJIqX6UlWtt}_>TQW^ArMTnkV8u{JtU4}
zS?LcAj9j>qDwI3%>`lvFljr5{Z&O{h++e7yUYisRK0H1esZ&=+9OBQ0*n((=tJ#&1
z7rf8)18pmV=^D#r>fLTz1mepI{;z+X@P4JVj9yN-Y@RxP0StcN_IGEh$0-rpq9Sdm
zi_cd1`0ljo0bv5t&U;2)pM^yOF_HDeZ%rOM;>Q!)8r{335_!Y)8Pzx0B31@9oIrEc
zN{zCEUczNwKHf#ks;%f`XWUWKSHXi{IQFkbT}C#w1WO3ye1%6x?3*5H7uamL;2pl2
zu_)Pc<$%g@ubNgdTo|@qo!NuxY~_hZ++48xaJQ;k@fOkDDzmnQPmo3E%X&Pw|7?5j
z5xUOn>CCK1k0Od&Ge06C%^>{ay`_fGubW!aPALnbrs7QzLy7)nHI56BF{^^cycdTh
z^6qv9`))L5p$-pZj;51YG<J3`R&cJehB!=maR`!bbzyBxZP8Md`an*b&Fo3iU<go?
zK?b{=<ms`wQUQlX84#g2LF+`4quYGeA%M||ipU7^@CJDv8(+UZa7QmZocCUgpv$tL
z=9?JCi$om1|CDK&+xYz$4Y?RSXibTniqF=s@q;Cqqp9hfd-n>YC4WS4!c6a07+5Pm
zW5BHpps^r~xa8S@bJ%t0zm2SJ_sntMe=;87jJ7ktgKRz9UJ7r^{nl|92W4*a-li=x
z2{AD?Hpn|;KzeacLt$qQOkNJ8FxUimj04qfjOxnJpv2p^!AHTzsJG@a*3TNsJkHTm
z;x(PuH*m3+%dfPR-(2;p+IJ5`AWl_Jbt&4veO(4*idBXlPV0Y(H;v{{0iHEug%o7F
zqE`$u3%L-I-}3|Z7t1HN%t0gl?HEJMDBoP`)?yal+oT7;(lgD3IsxS!jsWUe+iUdm
zuW*AGM-ALZ$%mb3IK&W^T(ulDi&8(ikEX-BtOCRmOMN<@|14M@9PA@aZA_IZ-W&oY
zq~BDNMt0{W{50c7GZ^hVE#WodX%1;CPA?Ww)Mc(;Cx?6UHUNJ1$MRLw&!sETxf1P0
z&AC?16+c!E1F8*WOSdc^TA5Sm2FREK=lP6y=wXn<HGk{v>ZQPCjrLQaH65tQDnR!o
zs8<P|hyOhJNn9=d#i;UdSvwfJzCQ(*p`Q;Z^URu(H$}|l&L5V0ksTwWU;S=2T|^0;
zm>HW2qu1hVlVTf8?v^DZBpLwO-@}5WGN3IEg|}*U#%j6~Yos-{*%zp1nZ~d<dmHq>
zTYJ%m@+VGx;S7G?xqhe+{u`J-nO{>z(JeZ@<sx7)rV>ib;hh7tRqF~`KI!<LRRR`T
zlSbL&x9=gvRk9znMV!EUvpIn&<?k3{To^qc0qRt~)3O-J;R5<sZ0h6PEChI)B<1E1
zNyQ+}91#aWscN<dd!HB`xz{MyxvBiBshPDr{QmcrHovd13L~yncEP;SWH3Cxunc^V
zzfOYK!ic0|pkK^q1V`TT-(WvQM2IZ^?J!qgHNg1IT2;NLfvb97ttMb#1J@1Gk`Un#
zlviGrSD?C#fJ9*In+-%T%Rfd(K;JKZV_)f^?RYV?mfaWF2CKul?Qtv2hwQDDN(a@X
zXknI|6EC9o@^`h5rf;3_Fy7`sP!4ru0QwG;+?qVBJqm#$*9a*W{oo6Vt+S=AARJQ@
zmxzeVg|41CzAtoQ|KZvAOa6K%F>1s2{JX5(h}}jSamT+RBDccL<DFL>DsQLSJYkgd
zv3SF$r;l_>en~hZwM91RX^C=>)Y<A5ccrlcFhyfi8M39)0l~#u6b3Apv168wd{Au{
zPbquqdf$`%ab@YATsAEhT#23eG*De9-MdmIgW0LZF<S#B3`apwqMD~Ib|G!5+bUrB
z0vsUXu9OzsEFZ*y``i-1lwp#^2rftcbn3Wp^jJPATl2Fa7}r}6U|p?WXM<ky(fkZU
zL+x!cxBoGZ{b*o+N^;;U?Wyu|m2vlUFiRE(8rJ}60xutQR?kMwKb5D1GtRvLR{H%O
zcNxDRNwk*<O<H{b$fiUTjfcbhY7$A#r<7lXg#(I|9g?##Q+_jkGuL~oTt_7?5Mdwq
z>fV2(oNCV%4xj@fn)s|Y4${3<rd5gCfK;c{8&`hlDIUZE8}Q-<z^dE}t!?{%xOR?>
zx&>lTA@I9G()E{4Q1K(5DouT`&lsEX=xFKdXmNg3j&EPVq=Mm}PF3_W^yyM6H!VhQ
zsbZE0W`!$RD8W`F4WYy9w0R0JyUd+;y_(Hz7u`ir=3c=RRHgQmB<}xX0%wYV3DmrT
z!{AsHn9vgVLQ2VHzb5=7-;k}ZDauxw4Su<|CWQnAABLW`kIhKnxSZJnVAm=JCts5l
zmea+n41Jm=tBgB>p)mL&;=lrNnDCD$C<V+*{9BkYhUDtIYmYf-Y6qQsDGPnT+skfM
z)FB4GI;l9p);Aq$D}5g&@UW<J?fnVgu&@rKicilG40}6^PW~R*ad6z!uxsRS)&~y)
zfv$Mpw(x#p>urA%>S+&rfv(G5SCo>yAtiU?q1^SGvI;ll<t1cgZ_3I(j=%BX)Bsnv
zCr%Fj|NjAO{V`-<fY|^1gSV5by_dJGtNZ^Rb3^&xe_}+t_7fNb`tvq$jmWLX&;JYB
CMRQO9

literal 0
HcmV?d00001

diff --git a/studio/frontend/src/app/router.tsx b/studio/frontend/src/app/router.tsx
index d507929758..13ff8a5cbe 100644
--- a/studio/frontend/src/app/router.tsx
+++ b/studio/frontend/src/app/router.tsx
@@ -13,7 +13,6 @@ import { Route as loginRoute } from "./routes/login";
 import { Route as onboardingRoute } from "./routes/onboarding";
 import { Route as changePasswordRoute } from "./routes/change-password";
 import { Route as studioRoute } from "./routes/studio";
-import { Route as apiKeysRoute } from "./routes/api-keys";
 
 const routeTree = rootRoute.addChildren([
   indexRoute,
@@ -26,7 +25,6 @@ const routeTree = rootRoute.addChildren([
   exportRoute,
   dataRecipesRoute,
   dataRecipeRoute,
-  apiKeysRoute,
 ]);
 
 export const router = createRouter({ routeTree });
diff --git a/studio/frontend/src/app/routes/__root.tsx b/studio/frontend/src/app/routes/__root.tsx
index e1bbdc03f7..69aeb11748 100644
--- a/studio/frontend/src/app/routes/__root.tsx
+++ b/studio/frontend/src/app/routes/__root.tsx
@@ -1,8 +1,13 @@
 // SPDX-License-Identifier: AGPL-3.0-only
 // Copyright 2026-present the Unsloth AI Inc. team. All rights reserved. See /studio/LICENSE.AGPL-3.0
 
+import { AppSidebar } from "@/components/app-sidebar";
 import { Navbar } from "@/components/navbar";
+import { SidebarInset, SidebarProvider } from "@/components/ui/sidebar";
 import { usePlatformStore } from "@/config/env";
+import { SettingsDialog, useSettingsDialogStore } from "@/features/settings";
+import { useTrainingUnloadGuard } from "@/features/training/hooks/use-training-unload-guard";
+import { useSidebarPin } from "@/hooks/use-sidebar-pin";
 import {
   Outlet,
   createRootRoute,
@@ -10,7 +15,7 @@ import {
   useRouterState,
 } from "@tanstack/react-router";
 import { AnimatePresence, motion } from "motion/react";
-import { Suspense } from "react";
+import { Suspense, useEffect } from "react";
 import { AppProvider } from "../provider";
 
 const CHAT_ONLY_ALLOWED = new Set([
@@ -19,7 +24,6 @@ const CHAT_ONLY_ALLOWED = new Set([
   "/login",
   "/signup",
   "/change-password",
-  "/api-keys",
 ]);
 
 function isChatOnlyAllowed(pathname: string): boolean {
@@ -43,24 +47,63 @@ const HIDDEN_NAVBAR_ROUTES = ["/onboarding", "/login", "/change-password"];
 function RootLayout() {
   const pathname = useRouterState({ select: (s) => s.location.pathname });
   const hideNavbar = HIDDEN_NAVBAR_ROUTES.includes(pathname);
+  const isChatRoute = pathname.startsWith("/chat");
+  const { pinned, setPinned, togglePinned } = useSidebarPin();
+
+  useTrainingUnloadGuard();
+
+  useEffect(() => {
+    const handler = (e: KeyboardEvent) => {
+      if (e.defaultPrevented) return;
+      if ((e.metaKey || e.ctrlKey) && e.key === ",") {
+        e.preventDefault();
+        useSettingsDialogStore.getState().openDialog();
+      }
+    };
+    window.addEventListener("keydown", handler);
+    return () => window.removeEventListener("keydown", handler);
+  }, []);
 
   return (
     <AppProvider>
-      {!hideNavbar && <Navbar />}
-      <AnimatePresence initial={false} mode="wait">
-        <motion.div
-          key={pathname}
-          initial={{ opacity: 0 }}
-          animate={{ opacity: 1 }}
-          exit={{ opacity: 0 }}
-          transition={{ duration: 0.15 }}
-          className="flex-1"
-        >
+      <SettingsDialog />
+      {hideNavbar ? (
+        <main className="flex-1">
           <Suspense fallback={null}>
             <Outlet />
           </Suspense>
-        </motion.div>
-      </AnimatePresence>
+        </main>
+      ) : (
+        <SidebarProvider
+          pinned={pinned}
+          setPinned={setPinned}
+          togglePinned={togglePinned}
+          className="!min-h-0 h-dvh overflow-hidden"
+        >
+          <AppSidebar />
+          <SidebarInset className={isChatRoute ? "overflow-hidden" : "overflow-y-auto"}>
+            <Navbar />
+            <div
+              className={`flex min-h-0 min-w-0 flex-1 basis-0 flex-col ${isChatRoute ? "overflow-hidden" : "overflow-visible"} ${isChatRoute ? "" : "pt-14 md:pt-0"}`}
+            >
+              <AnimatePresence initial={false} mode="wait">
+                <motion.div
+                  key={pathname}
+                  initial={{ opacity: 0 }}
+                  animate={{ opacity: 1 }}
+                  exit={{ opacity: 0 }}
+                  transition={{ duration: 0.15 }}
+                  className={`flex min-h-0 min-w-0 flex-1 basis-0 flex-col ${isChatRoute ? "overflow-hidden" : "overflow-visible"}`}
+                >
+                  <Suspense fallback={null}>
+                    <Outlet />
+                  </Suspense>
+                </motion.div>
+              </AnimatePresence>
+            </div>
+          </SidebarInset>
+        </SidebarProvider>
+      )}
     </AppProvider>
   );
 }
diff --git a/studio/frontend/src/app/routes/api-keys.tsx b/studio/frontend/src/app/routes/api-keys.tsx
deleted file mode 100644
index 5846690d7b..0000000000
--- a/studio/frontend/src/app/routes/api-keys.tsx
+++ /dev/null
@@ -1,18 +0,0 @@
-// SPDX-License-Identifier: AGPL-3.0-only
-// Copyright 2026-present the Unsloth AI Inc. team. All rights reserved. See /studio/LICENSE.AGPL-3.0
-
-import { createRoute } from "@tanstack/react-router";
-import { lazy } from "react";
-import { requireAuth } from "../auth-guards";
-import { Route as rootRoute } from "./__root";
-
-const ApiKeysPage = lazy(() =>
-  import("@/features/auth/api-keys-page").then((m) => ({ default: m.ApiKeysPage })),
-);
-
-export const Route = createRoute({
-  getParentRoute: () => rootRoute,
-  path: "/api-keys",
-  beforeLoad: () => requireAuth(),
-  component: ApiKeysPage,
-});
diff --git a/studio/frontend/src/app/routes/chat.tsx b/studio/frontend/src/app/routes/chat.tsx
index e435f090bd..49c05ce219 100644
--- a/studio/frontend/src/app/routes/chat.tsx
+++ b/studio/frontend/src/app/routes/chat.tsx
@@ -1,18 +1,25 @@
 // SPDX-License-Identifier: AGPL-3.0-only
 // Copyright 2026-present the Unsloth AI Inc. team. All rights reserved. See /studio/LICENSE.AGPL-3.0
 
+import { ChatPage } from "@/features/chat/chat-page";
 import { createRoute } from "@tanstack/react-router";
-import { lazy } from "react";
 import { requireAuth } from "../auth-guards";
 import { Route as rootRoute } from "./__root";
 
-const ChatPage = lazy(() =>
-  import("@/features/chat/chat-page").then((m) => ({ default: m.ChatPage })),
-);
+export type ChatSearch = {
+  thread?: string;
+  compare?: string;
+  new?: string;
+};
 
 export const Route = createRoute({
   getParentRoute: () => rootRoute,
   path: "/chat",
   beforeLoad: () => requireAuth(),
+  validateSearch: (search: Record<string, unknown>): ChatSearch => ({
+    thread: typeof search.thread === "string" ? search.thread : undefined,
+    compare: typeof search.compare === "string" ? search.compare : undefined,
+    new: typeof search.new === "string" ? search.new : undefined,
+  }),
   component: ChatPage,
 });
diff --git a/studio/frontend/src/components/app-sidebar.tsx b/studio/frontend/src/components/app-sidebar.tsx
new file mode 100644
index 0000000000..8264f329a3
--- /dev/null
+++ b/studio/frontend/src/components/app-sidebar.tsx
@@ -0,0 +1,607 @@
+// SPDX-License-Identifier: AGPL-3.0-only
+// Copyright 2026-present the Unsloth AI Inc. team. All rights reserved. See /studio/LICENSE.AGPL-3.0
+
+import {
+  Sidebar,
+  SidebarContent,
+  SidebarFooter,
+  SidebarGroup,
+  SidebarGroupContent,
+  SidebarGroupLabel,
+  SidebarHeader,
+  SidebarMenu,
+  SidebarMenuButton,
+  SidebarMenuItem,
+  useSidebar,
+} from "@/components/ui/sidebar";
+import {
+  Collapsible,
+  CollapsibleContent,
+  CollapsibleTrigger,
+} from "@/components/ui/collapsible";
+import {
+  DropdownMenu,
+  DropdownMenuContent,
+  DropdownMenuGroup,
+  DropdownMenuItem,
+  DropdownMenuSeparator,
+  DropdownMenuShortcut,
+  DropdownMenuTrigger,
+} from "@/components/ui/dropdown-menu";
+import { useAnimatedThemeToggle } from "@/components/ui/animated-theme-toggler";
+import { cn } from "@/lib/utils";
+import {
+  Book03Icon,
+  ChefHatIcon,
+  ColumnInsertIcon,
+  CursorInfo02Icon,
+  Delete02Icon,
+  MessageSearch01Icon,
+  Search01Icon,
+  NewReleasesIcon,
+  PackageIcon,
+  PencilEdit02Icon,
+  Settings02Icon,
+  ZapIcon,
+} from "@hugeicons/core-free-icons";
+import {
+  Tooltip,
+  TooltipContent,
+} from "@/components/ui/tooltip";
+import { Tooltip as TooltipPrimitive } from "radix-ui";
+import { HugeiconsIcon } from "@hugeicons/react";
+import { ChevronDown, ChevronsUpDown, Moon, PanelLeft, Sun } from "lucide-react";
+import { Link, useNavigate, useRouterState } from "@tanstack/react-router";
+import { motion } from "motion/react";
+import { useTrainingRuntimeStore } from "@/features/training";
+import { useSettingsDialogStore } from "@/features/settings";
+import { usePlatformStore } from "@/config/env";
+import { TOUR_OPEN_EVENT } from "@/features/tour";
+import {
+  useChatSidebarItems,
+  deleteChatItem,
+} from "@/features/chat/hooks/use-chat-sidebar-items";
+import { useChatRuntimeStore } from "@/features/chat/stores/chat-runtime-store";
+import { useChatSearchStore } from "@/features/chat/stores/chat-search-store";
+import { ChatSearchDialog } from "@/features/chat/components/chat-search-dialog";
+import { useTrainingHistorySidebarItems, deleteTrainingRun } from "@/features/training";
+import type { TrainingRunSummary } from "@/features/training";
+import { useState } from "react";
+
+function getTourId(pathname: string): string | null {
+  if (pathname.startsWith("/studio")) return "studio";
+  if (pathname.startsWith("/export")) return "export";
+  if (pathname.startsWith("/chat")) return "chat";
+  return null;
+}
+
+const NAV_SPRING = { type: "spring", stiffness: 500, damping: 35, mass: 0.5 } as const;
+
+function runStatusDotClass(status: TrainingRunSummary["status"]): string {
+  switch (status) {
+    case "running":
+      return "bg-blue-500 animate-pulse";
+    case "completed":
+      return "bg-emerald-500";
+    case "stopped":
+      return "bg-amber-500";
+    case "error":
+      return "bg-red-500";
+    default:
+      return "bg-muted-foreground";
+  }
+}
+
+function formatRelativeShort(iso: string): string {
+  const then = new Date(iso).getTime();
+  if (Number.isNaN(then)) return "";
+  const diffMs = Date.now() - then;
+  const s = Math.max(0, Math.floor(diffMs / 1000));
+  if (s < 60) return `${s}s`;
+  const m = Math.floor(s / 60);
+  if (m < 60) return `${m}m`;
+  const h = Math.floor(m / 60);
+  if (h < 24) return `${h}h`;
+  const d = Math.floor(h / 24);
+  return `${d}d`;
+}
+
+function NavItem({
+  icon,
+  label,
+  active,
+  disabled,
+  onClick,
+  children,
+  variant = "nav",
+  dataTour,
+}: {
+  icon: typeof ZapIcon;
+  label: string;
+  active: boolean;
+  disabled?: boolean;
+  onClick: () => void;
+  children?: React.ReactNode;
+  variant?: "nav" | "menu";
+  dataTour?: string;
+}) {
+  const isNav = variant === "nav";
+  return (
+    <SidebarMenuItem>
+      <div className="relative">
+        {isNav && active && (
+          <motion.div
+            layoutId="sidebar-active-indicator"
+            className="absolute left-0 top-0 bottom-0 w-[3px] rounded-full bg-primary"
+            transition={NAV_SPRING}
+          />
+        )}
+        <SidebarMenuButton
+          tooltip={label}
+          disabled={disabled}
+          onClick={onClick}
+          isActive={active}
+          data-tour={dataTour}
+          className={
+            isNav
+              ? "rounded-none pr-0 pl-4 text-[#475569] dark:text-[#94a3b8] data-active:text-foreground!"
+              : "rounded-none pr-0 pl-4 text-[#475569] dark:text-[#94a3b8] hover:bg-muted! hover:text-foreground! data-active:bg-[oklch(0.94_0_0)]! data-active:text-foreground! dark:data-active:bg-[oklch(0.3_0_0)]!"
+          }
+        >
+          <HugeiconsIcon icon={icon} strokeWidth={2} className="size-[18px]" />
+          <span className="text-[13px] font-medium">{label}</span>
+        </SidebarMenuButton>
+      </div>
+      {children}
+    </SidebarMenuItem>
+  );
+}
+
+export function AppSidebar() {
+  const { isDark, toggleTheme, anchorRef } = useAnimatedThemeToggle();
+  const { pathname, search } = useRouterState({
+    select: (s) => ({
+      pathname: s.location.pathname,
+      search: s.location.search as Record<string, string | undefined>,
+    }),
+  });
+  const { togglePinned, isMobile, setOpenMobile } = useSidebar();
+  const navigate = useNavigate();
+
+  // Auto-close mobile Sheet after navigation
+  const closeMobileIfOpen = () => {
+    if (isMobile) setOpenMobile(false);
+  };
+
+  const isTrainingRunning = useTrainingRuntimeStore((s) => s.isTrainingRunning);
+  const chatOnly = usePlatformStore((s) => s.isChatOnly());
+
+  // Chat collapsible state — open by default, syncs with route
+  const isChatRoute = pathname.startsWith("/chat");
+  const isStudioRoute = pathname === "/studio" || pathname.startsWith("/studio/");
+  const [chatOpen, setChatOpen] = useState(true);
+  const [runsOpen, setRunsOpen] = useState(true);
+  const effectiveChatOpen = isChatRoute || chatOpen;
+  const effectiveRunsOpen = isStudioRoute || runsOpen;
+
+  const isRecipesRoute = pathname.startsWith("/data-recipes");
+
+  const { items: chatItems } = useChatSidebarItems();
+  const storeThreadId = useChatRuntimeStore((s) => s.activeThreadId);
+  const setActiveThreadId = useChatRuntimeStore((s) => s.setActiveThreadId);
+  const activeThreadId = isChatRoute
+    ? (search.thread as string | undefined) ??
+      (search.compare as string | undefined) ??
+      storeThreadId ??
+      undefined
+    : undefined;
+
+  // Training runs
+  const { items: runItems, refresh: refreshRuns } = useTrainingHistorySidebarItems(
+    !chatOnly && isStudioRoute,
+  );
+  const activeJobId = useTrainingRuntimeStore((s) => s.jobId);
+  const selectedHistoryRunId = useTrainingRuntimeStore((s) => s.selectedHistoryRunId);
+  const setSelectedHistoryRunId = useTrainingRuntimeStore((s) => s.setSelectedHistoryRunId);
+
+  const chatDisabled = isTrainingRunning;
+
+  async function handleDeleteThread(item: Parameters<typeof deleteChatItem>[0]) {
+    await deleteChatItem(item, activeThreadId, (view) => {
+      navigate({
+        to: "/chat",
+        search: { new: view.newThreadNonce },
+      });
+    });
+  }
+
+  return (
+    <>
+    <Sidebar collapsible="icon" variant="sidebar">
+      <SidebarHeader className="group-data-[collapsible=icon]:px-0">
+        {/* Expanded: compact logo + close toggle */}
+        <div className="flex items-center justify-between gap-2 px-1 py-1 group-data-[collapsible=icon]:hidden">
+          <Link
+            to={chatOnly ? "/chat" : "/studio"}
+            onClick={closeMobileIfOpen}
+            className="flex items-center select-none"
+            aria-label="Unsloth home"
+          >
+            <img
+              src="/blacklogo-c.png"
+              alt="Unsloth"
+              className="h-7 w-auto dark:hidden"
+            />
+            <img
+              src="/whitelogo-c.png"
+              alt="Unsloth"
+              className="hidden h-7 w-auto dark:block"
+            />
+          </Link>
+          {!isMobile && (
+            <Tooltip>
+              <TooltipPrimitive.Trigger asChild>
+                <button
+                  type="button"
+                  onClick={togglePinned}
+                  className="inline-flex h-7 w-7 items-center justify-center rounded-md text-sidebar-foreground/70 transition-colors hover:bg-sidebar-accent hover:text-sidebar-accent-foreground focus-visible:outline-none focus-visible:ring-2 focus-visible:ring-ring"
+                  aria-label="Close sidebar"
+                >
+                  <PanelLeft strokeWidth={1.5} className="size-4" />
+                </button>
+              </TooltipPrimitive.Trigger>
+              <TooltipContent side="bottom" sideOffset={6}>
+                Close sidebar
+              </TooltipContent>
+            </Tooltip>
+          )}
+        </div>
+
+        {/* Collapsed: sticker with hover-swap to open toggle */}
+        {!isMobile && (
+          <div className="hidden group-data-[collapsible=icon]:flex items-center justify-center h-9 w-full">
+            <Tooltip>
+              <TooltipPrimitive.Trigger asChild>
+                <button
+                  type="button"
+                  onClick={togglePinned}
+                  className="inline-flex h-7 w-7 items-center justify-center rounded-md text-sidebar-foreground/70 transition-colors hover:bg-sidebar-accent hover:text-sidebar-accent-foreground focus-visible:outline-none focus-visible:ring-2 focus-visible:ring-ring"
+                  aria-label="Open sidebar"
+                >
+                  <PanelLeft strokeWidth={1.5} className="size-4" />
+                </button>
+              </TooltipPrimitive.Trigger>
+              <TooltipContent side="right" sideOffset={8}>
+                Open sidebar
+              </TooltipContent>
+            </Tooltip>
+          </div>
+        )}
+      </SidebarHeader>
+
+      <SidebarGroup className="group-data-[collapsible=icon]:p-0 p-0 pt-1 shrink-0">
+        <SidebarGroupContent>
+          <SidebarMenu>
+            <NavItem
+              icon={PencilEdit02Icon}
+              label="New Chat"
+              active={false}
+              disabled={chatDisabled}
+              onClick={() => {
+                if (chatDisabled) return;
+                setActiveThreadId(null);
+                navigate({ to: "/chat", search: { new: crypto.randomUUID() } });
+                closeMobileIfOpen();
+              }}
+            />
+            <NavItem
+              icon={ColumnInsertIcon}
+              label="Compare"
+              active={!!search.compare}
+              disabled={chatDisabled}
+              dataTour="chat-compare"
+              onClick={() => {
+                if (chatDisabled) return;
+                setActiveThreadId(null);
+                navigate({ to: "/chat", search: { compare: crypto.randomUUID() } });
+                closeMobileIfOpen();
+              }}
+            />
+            <NavItem
+              icon={Search01Icon}
+              label="Search"
+              active={false}
+              disabled={chatDisabled}
+              onClick={() => {
+                if (chatDisabled) return;
+                useChatSearchStore.getState().open();
+                closeMobileIfOpen();
+              }}
+            />
+          </SidebarMenu>
+        </SidebarGroupContent>
+        <div className="my-2" />
+      </SidebarGroup>
+
+      <SidebarContent className="gap-0 overflow-y-auto overscroll-contain min-h-0">
+        {/* Navigate (no header) */}
+        <SidebarGroup data-tour="navbar" className="group-data-[collapsible=icon]:p-0 p-0">
+          <SidebarGroupContent>
+            <SidebarMenu>
+              <NavItem
+                icon={ZapIcon}
+                label="Studio"
+                active={pathname === "/studio" || pathname.startsWith("/studio/")}
+                disabled={chatOnly}
+                onClick={() => {
+                  if (chatOnly) return;
+                  navigate({ to: "/studio" });
+                  closeMobileIfOpen();
+                }}
+              />
+
+              <NavItem
+                icon={ChefHatIcon}
+                label="Recipes"
+                active={isRecipesRoute}
+                onClick={() => {
+                  navigate({ to: "/data-recipes" });
+                  closeMobileIfOpen();
+                }}
+              />
+
+              <NavItem
+                icon={PackageIcon}
+                label="Export"
+                active={pathname === "/export" || pathname.startsWith("/export/")}
+                disabled={chatOnly}
+                onClick={() => {
+                  if (chatOnly) return;
+                  navigate({ to: "/export" });
+                  closeMobileIfOpen();
+                }}
+              />
+            </SidebarMenu>
+          </SidebarGroupContent>
+          <div className="my-2" />
+        </SidebarGroup>
+
+        {/* Recent Chats */}
+        {chatItems.length > 0 && (
+          <Collapsible open={effectiveChatOpen} onOpenChange={setChatOpen} asChild>
+          <SidebarGroup className="group-data-[collapsible=icon]:hidden overflow-hidden p-0">
+            <SidebarGroupLabel asChild>
+              <CollapsibleTrigger className="cursor-pointer flex w-full items-center justify-between">
+                Recent Chats
+                <ChevronDown className="size-3.5 transition-transform duration-200 data-[state=open]:rotate-0 [[data-state=closed]_&]:rotate-[-90deg]" />
+              </CollapsibleTrigger>
+            </SidebarGroupLabel>
+            <CollapsibleContent>
+            <SidebarGroupContent>
+              <SidebarMenu>
+                {chatItems.map((item) => (
+                  <SidebarMenuItem key={item.id} className="group/recent-item relative">
+                    <SidebarMenuButton
+                      isActive={activeThreadId === item.id}
+                      className="rounded-none pl-4 pr-7 text-[13px] font-medium text-[#475569] dark:text-[#94a3b8] hover:bg-muted! hover:text-foreground! data-active:bg-[oklch(0.94_0_0)]! data-active:text-foreground! dark:data-active:bg-[oklch(0.3_0_0)]!"
+                      onClick={() => {
+                        navigate({
+                          to: "/chat",
+                          search:
+                            item.type === "single"
+                              ? { thread: item.id }
+                              : { compare: item.id },
+                        });
+                        closeMobileIfOpen();
+                      }}
+                    >
+                      <span className="truncate">{item.title}</span>
+                    </SidebarMenuButton>
+                    <button
+                      type="button"
+                      onClick={(e) => {
+                        e.stopPropagation();
+                        handleDeleteThread(item);
+                      }}
+                      title="Delete"
+                      className="absolute right-1 top-1/2 -translate-y-1/2 flex size-5 scale-90 items-center justify-center rounded-md text-sidebar-foreground/55 opacity-0 transition-all duration-150 hover:bg-destructive/12 hover:text-destructive group-hover/recent-item:scale-100 group-hover/recent-item:opacity-100"
+                    >
+                      <HugeiconsIcon icon={Delete02Icon} strokeWidth={2} className="size-3.5" />
+                    </button>
+                  </SidebarMenuItem>
+                ))}
+              </SidebarMenu>
+            </SidebarGroupContent>
+            </CollapsibleContent>
+          </SidebarGroup>
+          </Collapsible>
+        )}
+
+        {/* Recent Runs */}
+        {isStudioRoute && runItems.length > 0 && !chatOnly && (
+          <Collapsible open={effectiveRunsOpen} onOpenChange={setRunsOpen} asChild>
+          <SidebarGroup className="group-data-[collapsible=icon]:hidden overflow-hidden p-0">
+            <SidebarGroupLabel asChild>
+              <CollapsibleTrigger className="cursor-pointer flex w-full items-center justify-between">
+                Recent Runs
+                <ChevronDown className="size-3.5 transition-transform duration-200 data-[state=open]:rotate-0 [[data-state=closed]_&]:rotate-[-90deg]" />
+              </CollapsibleTrigger>
+            </SidebarGroupLabel>
+            <CollapsibleContent>
+              <SidebarGroupContent>
+                <SidebarMenu>
+                  {runItems.map((run) => {
+                    const isActiveRun =
+                      selectedHistoryRunId === run.id || activeJobId === run.id;
+                    return (
+                      <SidebarMenuItem
+                        key={run.id}
+                        className="group/run-item relative"
+                      >
+                        <SidebarMenuButton
+                          isActive={isActiveRun}
+                          className="h-auto flex-col items-start gap-0.5 py-2 rounded-none pl-4 pr-7 text-[13px] font-medium text-[#475569] dark:text-[#94a3b8] hover:bg-muted! hover:text-foreground! data-active:bg-[oklch(0.94_0_0)]! data-active:text-foreground! dark:data-active:bg-[oklch(0.3_0_0)]!"
+                          onClick={() => {
+                            setSelectedHistoryRunId(run.id);
+                            closeMobileIfOpen();
+                          }}
+                        >
+                          <div className="flex w-full items-center gap-2">
+                            <span
+                              className={cn(
+                                "size-1.5 shrink-0 rounded-full",
+                                runStatusDotClass(run.status),
+                              )}
+                              aria-hidden
+                            />
+                            <span className="truncate text-sm font-medium">
+                              {run.model_name}
+                            </span>
+                            <span className="ml-auto shrink-0 text-[10px] text-muted-foreground">
+                              {formatRelativeShort(run.started_at)}
+                            </span>
+                          </div>
+                          <span className="w-full truncate pl-3.5 text-xs text-muted-foreground">
+                            {run.dataset_name}
+                          </span>
+                        </SidebarMenuButton>
+                        <button
+                          type="button"
+                          onClick={async (e) => {
+                            e.stopPropagation();
+                            try {
+                              await deleteTrainingRun(run.id);
+                              if (selectedHistoryRunId === run.id) {
+                                setSelectedHistoryRunId(null);
+                              }
+                              await refreshRuns();
+                            } catch {
+                              // ignore — next refresh will reconcile
+                            }
+                          }}
+                          title="Delete"
+                          className="absolute right-1 top-1/2 -translate-y-1/2 flex size-5 scale-90 items-center justify-center rounded-md text-sidebar-foreground/55 opacity-0 transition-all duration-150 hover:bg-destructive/12 hover:text-destructive group-hover/run-item:scale-100 group-hover/run-item:opacity-100"
+                        >
+                          <HugeiconsIcon icon={Delete02Icon} strokeWidth={2} className="size-3.5" />
+                        </button>
+                      </SidebarMenuItem>
+                    );
+                  })}
+                </SidebarMenu>
+              </SidebarGroupContent>
+            </CollapsibleContent>
+          </SidebarGroup>
+          </Collapsible>
+        )}
+      </SidebarContent>
+
+      <SidebarFooter className="border-t border-sidebar-border">
+        <SidebarMenu>
+          <SidebarMenuItem>
+            <DropdownMenu>
+              <DropdownMenuTrigger asChild>
+                <SidebarMenuButton
+                  size="lg"
+                  className="data-[state=open]:bg-sidebar-accent data-[state=open]:text-sidebar-accent-foreground"
+                >
+                  <img
+                    src="/Sloth emojis/sloth rounded.png"
+                    alt="Unsloth"
+                    className="size-8 rounded-lg shrink-0"
+                  />
+                  <div className="flex flex-col gap-0.5 leading-none group-data-[collapsible=icon]:hidden">
+                    <span className="truncate text-sm font-semibold">Unsloth</span>
+                    <span className="truncate text-[11px] text-muted-foreground">Studio</span>
+                  </div>
+                  <ChevronsUpDown strokeWidth={1.25} className="ml-auto size-4 text-muted-foreground group-data-[collapsible=icon]:hidden" />
+                </SidebarMenuButton>
+              </DropdownMenuTrigger>
+              <DropdownMenuContent
+                side="top"
+                align="start"
+                className="w-56"
+              >
+                <DropdownMenuGroup>
+                  <DropdownMenuItem
+                    onSelect={() => useSettingsDialogStore.getState().openDialog()}
+                  >
+                    <HugeiconsIcon icon={Settings02Icon} className="size-4" />
+                    <span>Settings</span>
+                    <DropdownMenuShortcut>⌘,</DropdownMenuShortcut>
+                  </DropdownMenuItem>
+                </DropdownMenuGroup>
+                <DropdownMenuSeparator />
+                <DropdownMenuGroup>
+                  <DropdownMenuItem
+                    ref={anchorRef as React.Ref<HTMLDivElement>}
+                    onSelect={(e) => { e.preventDefault(); toggleTheme(); }}
+                  >
+                    {isDark ? <Sun className="size-4" /> : <Moon className="size-4" />}
+                    <span>{isDark ? "Light Mode" : "Dark Mode"}</span>
+                  </DropdownMenuItem>
+                  <DropdownMenuItem
+                    disabled={!getTourId(pathname)}
+                    onSelect={() => {
+                      const tourId = getTourId(pathname);
+                      if (!tourId) return;
+                      window.dispatchEvent(
+                        new CustomEvent(TOUR_OPEN_EVENT, {
+                          detail: { id: tourId },
+                        }),
+                      );
+                    }}
+                  >
+                    <HugeiconsIcon icon={CursorInfo02Icon} className="size-4" />
+                    <span>Guided Tour</span>
+                  </DropdownMenuItem>
+                </DropdownMenuGroup>
+                <DropdownMenuSeparator />
+                <DropdownMenuGroup>
+                  <DropdownMenuItem asChild>
+                    <a
+                      href="https://unsloth.ai/docs"
+                      target="_blank"
+                      rel="noopener noreferrer"
+                    >
+                      <HugeiconsIcon icon={Book03Icon} className="size-4" />
+                      <span>Learn More</span>
+                    </a>
+                  </DropdownMenuItem>
+                  <DropdownMenuItem asChild>
+                    <a
+                      href="https://unsloth.ai/docs/new/changelog"
+                      target="_blank"
+                      rel="noopener noreferrer"
+                    >
+                      <HugeiconsIcon
+                        icon={NewReleasesIcon}
+                        className="size-4"
+                      />
+                      <span>What's New</span>
+                    </a>
+                  </DropdownMenuItem>
+                </DropdownMenuGroup>
+                <DropdownMenuSeparator />
+                <DropdownMenuItem asChild>
+                  <a
+                    href="https://github.com/unslothai/unsloth/issues"
+                    target="_blank"
+                    rel="noopener noreferrer"
+                  >
+                    <HugeiconsIcon
+                      icon={MessageSearch01Icon}
+                      className="size-4"
+                    />
+                    <span>Feedback</span>
+                  </a>
+                </DropdownMenuItem>
+              </DropdownMenuContent>
+            </DropdownMenu>
+          </SidebarMenuItem>
+        </SidebarMenu>
+      </SidebarFooter>
+    </Sidebar>
+    <ChatSearchDialog />
+    </>
+  );
+}
diff --git a/studio/frontend/src/components/assistant-ui/reasoning.tsx b/studio/frontend/src/components/assistant-ui/reasoning.tsx
index 6b2c7a05e7..e4306df4b9 100644
--- a/studio/frontend/src/components/assistant-ui/reasoning.tsx
+++ b/studio/frontend/src/components/assistant-ui/reasoning.tsx
@@ -6,7 +6,6 @@
 /* eslint-disable react-refresh/only-export-components */
 
 import { MarkdownText } from "@/components/assistant-ui/markdown-text";
-import { AnimatedShinyText } from "@/components/ui/animated-shiny-text";
 import {
   Collapsible,
   CollapsibleContent,
@@ -151,34 +150,6 @@ function ReasoningRoot({
   );
 }
 
-function ReasoningFade({ className, ...props }: ComponentProps<"div">) {
-  return (
-    <div
-      data-slot="reasoning-fade"
-      className={cn(
-        "aui-reasoning-fade pointer-events-none absolute inset-x-0 bottom-0 z-10 h-8",
-        "bg-gradient-to-t from-background to-transparent",
-        className,
-      )}
-      {...props}
-    />
-  );
-}
-
-function ReasoningFadeTop({ className, ...props }: ComponentProps<"div">) {
-  return (
-    <div
-      data-slot="reasoning-fade-top"
-      className={cn(
-        "aui-reasoning-fade-top pointer-events-none absolute inset-x-0 top-0 z-10 h-8",
-        "bg-gradient-to-b from-background to-transparent",
-        className,
-      )}
-      {...props}
-    />
-  );
-}
-
 function ReasoningTrigger({
   active,
   duration,
@@ -206,7 +177,7 @@ function ReasoningTrigger({
         className="aui-reasoning-trigger-label-wrapper relative inline-block leading-none"
       >
         {active ? (
-          <AnimatedShinyText className="text-sm">Thinking...</AnimatedShinyText>
+          <span className="text-sm">Thinking...</span>
         ) : (
           <span>Thought for {duration ?? 0} seconds</span>
         )}
@@ -234,7 +205,7 @@ function ReasoningContent({
     <CollapsibleContent
       data-slot="reasoning-content"
       className={cn(
-        "aui-reasoning-content relative overflow-hidden text-muted-foreground text-sm outline-none",
+        "aui-reasoning-content relative overflow-hidden text-foreground/85 text-[13.5px] outline-none",
         "group/collapsible-content ease-out",
         "data-[state=closed]:animate-collapsible-up",
         "data-[state=open]:animate-collapsible-down",
@@ -246,9 +217,7 @@ function ReasoningContent({
       )}
       {...props}
     >
-      {streaming && <ReasoningFadeTop />}
       {children}
-      <ReasoningFade />
     </CollapsibleContent>
   );
 }
@@ -481,8 +450,6 @@ const Reasoning = memo(
   Trigger: typeof ReasoningTrigger;
   Content: typeof ReasoningContent;
   Text: typeof ReasoningText;
-  Fade: typeof ReasoningFade;
-  FadeTop: typeof ReasoningFadeTop;
 };
 
 Reasoning.displayName = "Reasoning";
@@ -490,8 +457,6 @@ Reasoning.Root = ReasoningRoot;
 Reasoning.Trigger = ReasoningTrigger;
 Reasoning.Content = ReasoningContent;
 Reasoning.Text = ReasoningText;
-Reasoning.Fade = ReasoningFade;
-Reasoning.FadeTop = ReasoningFadeTop;
 
 const ReasoningGroup = memo(ReasoningGroupImpl);
 ReasoningGroup.displayName = "ReasoningGroup";
@@ -503,6 +468,4 @@ export {
   ReasoningTrigger,
   ReasoningContent,
   ReasoningText,
-  ReasoningFade,
-  ReasoningFadeTop,
 };
diff --git a/studio/frontend/src/components/assistant-ui/thread.tsx b/studio/frontend/src/components/assistant-ui/thread.tsx
index 8f41987fbf..77fdf28736 100644
--- a/studio/frontend/src/components/assistant-ui/thread.tsx
+++ b/studio/frontend/src/components/assistant-ui/thread.tsx
@@ -16,7 +16,6 @@ import { WebSearchToolUI } from "@/components/assistant-ui/tool-ui-web-search";
 import { PythonToolUI } from "@/components/assistant-ui/tool-ui-python";
 import { TerminalToolUI } from "@/components/assistant-ui/tool-ui-terminal";
 import { TooltipIconButton } from "@/components/assistant-ui/tooltip-icon-button";
-import { AnimatedShinyText } from "@/components/ui/animated-shiny-text";
 import { Button } from "@/components/ui/button";
 import { sentAudioNames } from "@/features/chat/api/chat-adapter";
 import { AUDIO_ACCEPT, MAX_AUDIO_SIZE, fileToBase64 } from "@/lib/audio-utils";
@@ -70,13 +69,25 @@ export const Thread: FC<{ hideComposer?: boolean; hideWelcome?: boolean }> = ({
 }) => {
   return (
     <ThreadPrimitive.Root
-      className="aui-root aui-thread-root @container flex h-full flex-col "
+      className={cn(
+        "aui-root aui-thread-root @container flex flex-col",
+        hideComposer
+          ? "h-full"
+          : "relative min-h-0 min-w-0 flex-1 basis-0 overflow-hidden",
+      )}
       style={{
         ["--thread-max-width" as string]: "44rem",
+        ["--thread-content-max-width" as string]:
+          "calc(var(--thread-max-width) - 2.5rem)",
       }}
     >
       <ThreadPrimitive.Viewport
-        className="aui-thread-viewport relative flex min-w-0 flex-1 flex-col overflow-x-auto overflow-y-scroll scroll-smooth px-4 pt-4"
+        className={cn(
+          "aui-thread-viewport relative flex min-w-0 flex-1 flex-col overflow-x-auto overflow-y-scroll scroll-smooth px-5",
+          hideComposer
+            ? "pt-4"
+            : "h-0 min-h-0 basis-0 pt-[56px]",
+        )}
       >
         {!hideWelcome && (
           <AuiIf condition={({ thread }) => thread.isEmpty}>
@@ -92,19 +103,21 @@ export const Thread: FC<{ hideComposer?: boolean; hideWelcome?: boolean }> = ({
           }}
         />
 
+        {/* Small overlap and extra slack so the last lines can scroll under the composer cleanly */}
+        {!hideComposer && <div className="h-40 shrink-0" aria-hidden />}
+
         <ThreadPrimitive.ViewportFooter
           className={cn(
-            "aui-thread-viewport-footer sticky bottom-0 z-20 mt-auto flex w-full flex-col overflow-visible bg-transparent",
-            hideComposer ? "gap-2" : "gap-4",
+            "aui-thread-viewport-footer sticky z-20 mt-auto flex w-full flex-col overflow-visible bg-transparent",
+            hideComposer
+              ? "bottom-0 gap-2"
+              : "bottom-[140px] shrink-0 gap-3",
             // Compare: pointer-events pass-through so messages behind footer stay clickable
             hideComposer
               ? "pointer-events-none pb-3"
-              : "relative pb-4",
+              : "pb-2",
           )}
         >
-          {!hideComposer && (
-            <div className="pointer-events-none absolute inset-x-0 bottom-0 h-4 bg-background" aria-hidden />
-          )}
           <div
             className={cn(
               "flex justify-center",
@@ -113,11 +126,26 @@ export const Thread: FC<{ hideComposer?: boolean; hideWelcome?: boolean }> = ({
           >
             <ThreadScrollToBottom />
           </div>
-          <AuiIf condition={({ thread }) => !thread.isEmpty}>
-            {!hideComposer && <ComposerAnimated />}
-          </AuiIf>
         </ThreadPrimitive.ViewportFooter>
       </ThreadPrimitive.Viewport>
+      {!hideComposer && (
+        <AuiIf condition={({ thread }) => !thread.isEmpty}>
+          <div className="aui-thread-composer-dock pointer-events-none absolute bottom-0 left-0 right-0 md:right-2 z-20">
+            <div
+              aria-hidden
+              className="absolute inset-x-0 bottom-0 top-[10px] bg-background"
+            />
+            <div className="relative px-5 pb-2">
+              <div className="pointer-events-auto mx-auto w-full max-w-(--thread-max-width)">
+                <ComposerAnimated />
+              </div>
+              <p className="mt-1.5 text-center text-[11px] text-muted-foreground">
+                LLM's can make mistakes. Double-check all responses.
+              </p>
+            </div>
+          </div>
+        </AuiIf>
+      )}
     </ThreadPrimitive.Root>
   );
 };
@@ -207,18 +235,13 @@ const ThreadWelcome: FC<{ hideComposer?: boolean }> = ({ hideComposer }) => {
               alt="Sloth mascot"
               className="size-20"
             />
-            <h1 className="aui-thread-welcome-message-inner fade-in slide-in-from-bottom-1 animate-in font-semibold text-2xl duration-200">
+            <h1 className="aui-thread-welcome-message-inner fade-in slide-in-from-bottom-1 animate-in font-bold text-2xl tracking-[-0.02em] duration-200">
               Chat with your model
             </h1>
-            <p className="aui-thread-welcome-message-inner fade-in slide-in-from-bottom-1 animate-in text-muted-foreground text-base delay-75 duration-200">
-              Run GGUFs, safetensors, vision and audio models!
+            <p className="aui-thread-welcome-message-inner fade-in slide-in-from-bottom-1 animate-in text-muted-foreground text-sm delay-75 duration-200">
+              Run GGUFs, safetensors, vision and audio models
             </p>
           </div>
-          <div className="grid grid-cols-2 gap-2">
-            <ThreadPrimitive.Suggestions
-              components={{ Suggestion: SuggestionItem }}
-            />
-          </div>
           <GeneratingSpinner />
           {!hideComposer && <ComposerAnimated />}
         </div>
@@ -243,10 +266,6 @@ const GeneratingSpinner: FC = () => {
 const ComposerAnimated: FC = () => {
   return (
     <div className="relative mx-auto min-w-0 w-full max-w-(--thread-max-width)">
-      <div
-        className="pointer-events-none absolute inset-x-0 top-1/2 bottom-0 z-0 bg-background"
-        aria-hidden
-      />
       <motion.div
         layout={true}
         layoutId="composer"
@@ -284,14 +303,15 @@ const PendingAudioChip: FC = () => {
 const Composer: FC = () => {
   return (
     <ComposerPrimitive.Root className="aui-composer-root relative flex w-full flex-col">
-      <ComposerPrimitive.AttachmentDropzone className="aui-composer-attachment-dropzone shadow-border ring-1 ring-border flex w-full flex-col rounded-2xl bg-background px-1 pt-2 outline-none transition-shadow data-[dragging=true]:ring-ring data-[dragging=true]:bg-accent/50">
+      <ComposerPrimitive.AttachmentDropzone className="aui-composer-attachment-dropzone chat-composer-surface flex w-full flex-col rounded-3xl bg-background px-1 pt-2 outline-none transition-shadow data-[dragging=true]:border-ring data-[dragging=true]:bg-accent/50">
         <ComposerAttachments />
         <PendingAudioChip />
         <ToolStatusDisplay />
         <ComposerPrimitive.Input
           placeholder="Send a message..."
-          className="aui-composer-input mb-1 max-h-32 min-h-12 w-full resize-none bg-transparent pl-5 pr-4 pt-2 pb-3 text-sm outline-none placeholder:text-muted-foreground focus-visible:ring-0"
-          rows={1}
+          className="aui-composer-input mb-1 min-h-12 w-full resize-none overflow-y-auto bg-transparent pl-5 pr-4 pt-2 pb-3 text-sm outline-none placeholder:text-muted-foreground focus-visible:ring-0"
+          minRows={1}
+          maxRows={6}
           autoFocus={true}
           aria-label="Message input"
         />
@@ -457,12 +477,30 @@ const CodeToolsToggle: FC = () => {
       )}
       aria-label={codeToolsEnabled ? "Disable code execution" : "Enable code execution"}
     >
-      <TerminalIcon className="size-3.5" />
+      <CodeToggleIcon className="size-3.5" />
       <span>Code</span>
     </button>
   );
 };
 
+const CodeToggleIcon: FC<{ className?: string }> = ({ className }) => {
+  return (
+    <svg
+      viewBox="0 0 24 24"
+      fill="none"
+      stroke="currentColor"
+      strokeWidth="2"
+      strokeLinecap="round"
+      strokeLinejoin="round"
+      className={className}
+      aria-hidden="true"
+    >
+      <polyline points="16 18 22 12 16 6" />
+      <polyline points="8 6 2 12 8 18" />
+    </svg>
+  );
+};
+
 const ToolStatusDisplay: FC = () => {
   const toolStatus = useChatRuntimeStore((s) => s.toolStatus);
   const isThreadRunning = useAuiState(({ thread }) => thread.isRunning);
@@ -594,15 +632,13 @@ const GeneratingIndicator: FC = () => {
       message.content.length === 0 && message.status?.type === "running",
   );
   if (!show) return null;
-  return (
-    <AnimatedShinyText className="text-sm">Generating...</AnimatedShinyText>
-  );
+  return <span className="text-sm text-muted-foreground">Generating...</span>;
 };
 
 const AssistantMessage: FC = () => {
   return (
     <MessagePrimitive.Root
-      className="aui-assistant-message-root fade-in slide-in-from-bottom-1 relative mx-auto min-w-0 w-full max-w-(--thread-max-width) animate-in py-3 duration-150"
+      className="aui-assistant-message-root fade-in slide-in-from-bottom-1 relative mx-auto min-w-0 w-full max-w-(--thread-content-max-width) animate-in py-0.5 text-[15.5px] duration-150"
       data-role="assistant"
     >
       <div className="aui-assistant-message-content wrap-break-word min-w-0 text-foreground leading-relaxed">
@@ -701,9 +737,9 @@ const AssistantActionBar: FC = () => {
   return (
     <ActionBarPrimitive.Root
       hideWhenRunning={true}
-      autohide="not-last"
+      autohide="always"
       autohideFloat="single-branch"
-      className="aui-assistant-action-bar-root col-start-3 row-start-2 -ml-1 flex gap-1 text-muted-foreground data-floating:absolute data-floating:rounded-md data-floating:border data-floating:bg-background data-floating:p-1 data-floating:shadow-sm"
+      className="aui-assistant-action-bar-root col-start-3 row-start-2 -ml-1 flex gap-1 text-muted-foreground data-floating:absolute"
     >
       <CopyButton />
       <ActionBarPrimitive.Reload asChild={true}>
@@ -755,22 +791,22 @@ const UserMessageAudio: FC = () => {
 const UserMessage: FC = () => {
   return (
     <MessagePrimitive.Root
-      className="aui-user-message-root  fade-in slide-in-from-bottom-1 mx-auto grid w-full max-w-(--thread-max-width) animate-in auto-rows-auto grid-cols-[minmax(72px,1fr)_auto] content-start gap-y-2 px-2 py-3 duration-150 [&:where(>*)]:col-start-2"
+      className="aui-user-message-root fade-in slide-in-from-bottom-1 mx-auto flex w-full max-w-(--thread-content-max-width) animate-in flex-col items-end gap-y-2 pt-6 pb-0.5 text-[15.5px] duration-150"
       data-role="user"
     >
       <UserMessageAttachments />
       <UserMessageAudio />
 
-      <div className="aui-user-message-content-wrapper relative col-start-2 min-w-0">
-        <div className="aui-user-message-content wrap-break-word rounded-2xl bg-muted  px-4 py-2.5 text-foreground">
+      <div className="aui-user-message-content-wrapper flex max-w-[80%] min-w-0 flex-col items-end">
+        <div className="aui-user-message-content wrap-break-word w-fit rounded-2xl bg-muted px-4 py-2.5 text-foreground">
           <MessagePrimitive.Parts />
         </div>
-        <div className="aui-user-action-bar-wrapper absolute top-1/2 left-0 -translate-x-full -translate-y-1/2 pr-2">
+        <div className="mt-1 flex min-h-6">
           <UserActionBar />
         </div>
       </div>
 
-      <BranchPicker className="aui-user-branch-picker col-span-full col-start-1 row-start-3 -mr-1 justify-end" />
+      <BranchPicker className="aui-user-branch-picker -mr-1 justify-end" />
     </MessagePrimitive.Root>
   );
 };
@@ -778,8 +814,8 @@ const UserMessage: FC = () => {
 const UserActionBar: FC = () => {
   return (
     <ActionBarPrimitive.Root
-      autohide="not-last"
-      className="aui-user-action-bar-root flex items-center"
+      autohide="always"
+      className="aui-user-action-bar-root -mr-1 flex gap-1 text-muted-foreground"
     >
       <CopyButton />
       <ActionBarPrimitive.Edit asChild={true}>
@@ -805,7 +841,7 @@ const EditComposer: FC = () => {
   });
 
   return (
-    <MessagePrimitive.Root className="aui-edit-composer-wrapper mx-auto flex w-full max-w-(--thread-max-width) flex-col px-2 py-3">
+    <MessagePrimitive.Root className="aui-edit-composer-wrapper mx-auto flex w-full max-w-(--thread-content-max-width) flex-col py-3">
       <ComposerPrimitive.Root className="aui-edit-composer-root ml-auto flex w-full max-w-[85%] flex-col rounded-2xl bg-muted">
         <ComposerPrimitive.Input
           className="aui-edit-composer-input min-h-14 w-full resize-none bg-transparent p-4 text-foreground text-sm outline-none"
diff --git a/studio/frontend/src/components/navbar.tsx b/studio/frontend/src/components/navbar.tsx
index 2d310a81c3..56b1e0346d 100644
--- a/studio/frontend/src/components/navbar.tsx
+++ b/studio/frontend/src/components/navbar.tsx
@@ -1,647 +1,20 @@
 // SPDX-License-Identifier: AGPL-3.0-only
 // Copyright 2026-present the Unsloth AI Inc. team. All rights reserved. See /studio/LICENSE.AGPL-3.0
 
-import {
-  HoverCard,
-  HoverCardContent,
-  HoverCardTrigger,
-} from "@/components/ui/hover-card";
-import {
-  Collapsible,
-  CollapsibleContent,
-  CollapsibleTrigger,
-} from "@/components/ui/collapsible";
-import { AnimatedThemeToggler } from "@/components/ui/animated-theme-toggler";
-import {
-  Sheet,
-  SheetContent,
-  SheetHeader,
-  SheetTitle,
-  SheetTrigger,
-} from "@/components/ui/sheet";
-import { cn } from "@/lib/utils";
-import {
-  ArrowReloadHorizontalIcon,
-  ArrowRight01Icon,
-  Cancel01Icon,
-  Book03Icon,
-  BubbleChatIcon,
-  ChefHatIcon,
-  Copy01Icon,
-  CursorInfo02Icon,
-  Key01Icon,
-  PackageIcon,
-  Tick02Icon,
-  ZapIcon,
-} from "@hugeicons/core-free-icons";
-import { HugeiconsIcon } from "@hugeicons/react";
-import { copyToClipboard } from "@/lib/copy-to-clipboard";
-import { useTrainingRuntimeStore } from "@/features/training";
-import { usePlatformStore } from "@/config/env";
-import { Link, useRouterState } from "@tanstack/react-router";
-import { AnimatePresence, motion, useReducedMotion } from "motion/react";
-import type { ReactElement } from "react";
-import { useEffect, useRef, useState } from "react";
-import { TOUR_OPEN_EVENT } from "@/features/tour";
-import { ShutdownDialog } from "@/components/shutdown-dialog";
-
-const NAV_ITEMS = [
-  { label: "Studio", href: "/studio", icon: ZapIcon, enabled: true },
-  { label: "Recipes", href: "/data-recipes", icon: ChefHatIcon, enabled: true },
-  { label: "Export", href: "/export", icon: PackageIcon, enabled: true },
-  { label: "Chat", href: "/chat", icon: BubbleChatIcon, enabled: true },
-];
-
-const STUDIO_UPDATE_CMD = "unsloth studio update";
-const STUDIO_UPDATE_FALLBACK_UNIX_CMD =
-  "curl -fsSL https://unsloth.ai/install.sh | sh";
-const STUDIO_UPDATE_FALLBACK_WINDOWS_CMD =
-  "irm https://unsloth.ai/install.ps1 | iex";
-
-type UpdateShell = "windows" | "unix";
-
-function getDefaultUpdateShell(deviceType: string): UpdateShell {
-  return deviceType === "windows" ? "windows" : "unix";
-}
-
-function getStudioUpdateInstructionLine(shell: UpdateShell): string {
-  return shell === "windows" ? "Open PowerShell and run:" : "Open Terminal and run:";
-}
-
-function CopyableCommand({
-  command,
-  copyLabel,
-}: {
-  command: string;
-  copyLabel: string;
-}): ReactElement {
-  const [copied, setCopied] = useState(false);
-  const timerRef = useRef<ReturnType<typeof setTimeout> | null>(null);
-
-  useEffect(() => {
-    return () => {
-      if (timerRef.current) {
-        clearTimeout(timerRef.current);
-      }
-    };
-  }, []);
-
-  const handleCopy = () => {
-    if (!copyToClipboard(command)) {
-      return;
-    }
-    setCopied(true);
-    if (timerRef.current) {
-      clearTimeout(timerRef.current);
-    }
-    timerRef.current = setTimeout(() => setCopied(false), 2000);
-  };
-
-  return (
-    <div className="flex min-w-0 items-stretch overflow-hidden rounded-md border border-border bg-muted/40">
-      <input
-        type="text"
-        readOnly
-        value={command}
-        className="min-w-0 flex-1 bg-transparent px-2 py-1.5 font-mono text-[11px] text-foreground outline-none"
-        title={command}
-        aria-label={`${copyLabel} text`}
-      />
-      <button
-        type="button"
-        onClick={handleCopy}
-        className="flex shrink-0 items-center justify-center border-l border-border px-2 text-muted-foreground transition-colors hover:bg-muted hover:text-foreground"
-        title={copied ? "Copied" : "Copy command"}
-        aria-label={copied ? `${copyLabel} copied` : `Copy ${copyLabel}`}
-      >
-        {copied ? (
-          <HugeiconsIcon icon={Tick02Icon} className="size-4 text-emerald-600" />
-        ) : (
-          <HugeiconsIcon icon={Copy01Icon} className="size-4" />
-        )}
-      </button>
-    </div>
-  );
-}
-
-function UpdateStudioInstructions({
-  className,
-  defaultShell,
-  showTitle = true,
-}: {
-  className?: string;
-  defaultShell: UpdateShell;
-  showTitle?: boolean;
-}): ReactElement {
-  const [shell, setShell] = useState<UpdateShell>(defaultShell);
-  const prefersReducedMotion = useReducedMotion();
-  const windows = shell === "windows";
-  const fadeTransition = prefersReducedMotion
-    ? { duration: 0 }
-    : { duration: 0.16, ease: [0.165, 0.84, 0.44, 1] as const };
-  const fadeInitial = prefersReducedMotion ? { opacity: 1 } : { opacity: 0, y: 2 };
-  const fadeAnimate = { opacity: 1, y: 0 };
-  const fadeExit = prefersReducedMotion ? { opacity: 1 } : { opacity: 0, y: -2 };
-
-  useEffect(() => {
-    setShell(defaultShell);
-  }, [defaultShell]);
-
-  return (
-    <div className={cn("flex flex-col gap-3", className)}>
-      <div
-        className={cn(
-          "flex items-center gap-3",
-          showTitle ? "justify-between" : "justify-start",
-        )}
-      >
-        {showTitle ? (
-          <p className="shrink-0 whitespace-nowrap text-sm font-semibold font-heading">
-            Update Unsloth Studio
-          </p>
-        ) : null}
-        <div className="flex shrink-0 items-center gap-0.5 text-[11px]">
-          <button
-            type="button"
-            onClick={() => setShell("windows")}
-            className={cn(
-              "px-0.5 py-0.5 font-medium transition-colors",
-              windows
-                ? "text-foreground"
-                : "text-muted-foreground hover:text-emerald-600",
-            )}
-            aria-pressed={windows}
-          >
-            Windows
-          </button>
-          <span className="text-border">/</span>
-          <button
-            type="button"
-            onClick={() => setShell("unix")}
-            className={cn(
-              "px-0.5 py-0.5 font-medium transition-colors",
-              !windows
-                ? "text-foreground"
-                : "text-muted-foreground hover:text-emerald-600",
-            )}
-            aria-pressed={!windows}
-          >
-            macOS/Linux
-          </button>
-        </div>
-      </div>
-      <AnimatePresence mode="wait" initial={false}>
-        <motion.p
-          key={`instruction-${shell}`}
-          initial={fadeInitial}
-          animate={fadeAnimate}
-          exit={fadeExit}
-          transition={fadeTransition}
-          className="text-xs text-muted-foreground leading-relaxed"
-        >
-          {getStudioUpdateInstructionLine(shell)}
-        </motion.p>
-      </AnimatePresence>
-      <CopyableCommand command={STUDIO_UPDATE_CMD} copyLabel="update command" />
-      <p className="text-xs text-muted-foreground leading-relaxed">
-        If that fails or unsloth studio update is unavailable, run:
-      </p>
-      <AnimatePresence mode="wait" initial={false}>
-        <motion.div
-          key={`fallback-${shell}`}
-          initial={fadeInitial}
-          animate={fadeAnimate}
-          exit={fadeExit}
-          transition={fadeTransition}
-        >
-          <CopyableCommand
-            command={
-              windows
-                ? STUDIO_UPDATE_FALLBACK_WINDOWS_CMD
-                : STUDIO_UPDATE_FALLBACK_UNIX_CMD
-            }
-            copyLabel="fallback command"
-          />
-        </motion.div>
-      </AnimatePresence>
-      <p className="text-xs text-muted-foreground leading-relaxed">
-        Restart Studio after updating for changes to take effect.
-      </p>
-    </div>
-  );
-}
-
-function getTourId(pathname: string): "studio" | "chat" | "export" | null {
-  if (pathname === "/studio") return "studio";
-  if (pathname === "/chat") return "chat";
-  if (pathname === "/export") return "export";
-  return null;
-}
+import { SidebarTrigger, useSidebar } from "@/components/ui/sidebar";
 
 export function Navbar() {
-  const pathname = useRouterState({ select: (s) => s.location.pathname });
-  const isTrainingRunning = useTrainingRuntimeStore((s) => s.isTrainingRunning);
-  const [mobileOpen, setMobileOpen] = useState(false);
-  const [mobileUpdateOpen, setMobileUpdateOpen] = useState(false);
-  const [shutdownOpen, setShutdownOpen] = useState(false);
-
-  const deviceType = usePlatformStore((s) => s.deviceType);
-  const chatOnly = usePlatformStore((s) => s.isChatOnly());
-  const defaultUpdateShell = getDefaultUpdateShell(deviceType);
-
-  // Warn before closing the tab only when training is running (data loss risk).
-  // We store the handler in a ref so removeUnloadHandler() can clean it up
-  // before the "Server stopped" page renders.
-  const unloadHandlerRef = useRef<((e: BeforeUnloadEvent) => void) | null>(null);
-
-  useEffect(() => {
-    const handler = (e: BeforeUnloadEvent) => {
-      if (!useTrainingRuntimeStore.getState().isTrainingRunning) return;
-      e.preventDefault();
-      e.returnValue = "";
-    };
-    unloadHandlerRef.current = handler;
-    window.addEventListener("beforeunload", handler);
-    return () => {
-      window.removeEventListener("beforeunload", handler);
-    };
-  }, []);
-
-  const removeUnloadHandler = () => {
-    if (unloadHandlerRef.current) {
-      window.removeEventListener("beforeunload", unloadHandlerRef.current);
-      unloadHandlerRef.current = null;
-    }
-  };
-
-  const tourId = getTourId(pathname);
-
-  const openTour = () => {
-    if (!tourId) return;
-    window.dispatchEvent(
-      new CustomEvent(TOUR_OPEN_EVENT, { detail: { id: tourId } }),
+  const { isMobile } = useSidebar();
+  if (!isMobile) {
+    return (
+      <header className="absolute top-0 inset-x-0 z-40 h-11 pointer-events-none" />
     );
-  };
-
+  }
   return (
-    <>
-    <header className="relative top-0 z-40 h-16 w-full">
-      <div className="mx-auto grid h-full max-w-7xl grid-cols-[1fr_auto_1fr] items-center px-4 sm:px-6">
-        {/* Left: logo */}
-        <Link to={chatOnly ? "/chat" : "/studio"} className="flex items-center gap-1.5 justify-self-start select-none">
-          <img
-            src="/blacklogo.png"
-            alt="Unsloth"
-            className="h-9 w-auto dark:hidden"
-          />
-          <img
-            src="/whitelogo.png"
-            alt="Unsloth"
-            className="hidden h-9 w-auto dark:block"
-          />
-          <span className="relative -top-[1px] inline-flex items-center text-[10px] font-extrabold leading-none tracking-[0.12em] text-primary">
-            BETA
-          </span>
-        </Link>
-
-        {/* Center: pill nav */}
-        <nav
-          data-tour="navbar"
-          className="hidden items-center rounded-full border border-border bg-card p-1 ring-1 ring-foreground/5 md:flex"
-        >
-          {NAV_ITEMS.map((item) => {
-            const active =
-              pathname === item.href || pathname.startsWith(`${item.href}/`);
-            const disabledByTraining =
-              isTrainingRunning && item.href !== "/studio";
-            const disabledByDevice =
-              chatOnly && item.href !== "/chat" && item.href !== "/data-recipes";
-            if (!item.enabled || disabledByTraining || disabledByDevice) {
-              return (
-                <span
-                  key={item.href}
-                  className="relative rounded-full px-3 py-1.5 text-sm font-medium text-muted-foreground/40 cursor-not-allowed"
-                >
-                  {item.label}
-                </span>
-              );
-            }
-            return (
-              <Link
-                key={item.href}
-                to={item.href}
-                className={cn(
-                  "relative rounded-full px-3 py-1.5 text-sm font-medium transition-colors",
-                  active
-                    ? "text-background"
-                    : "text-muted-foreground hover:text-foreground",
-                )}
-              >
-                {active && (
-                  <motion.span
-                    layoutId="nav-pill"
-                    className="absolute inset-0 rounded-full bg-foreground"
-                    transition={{
-                      type: "spring",
-                      stiffness: 500,
-                      damping: 35,
-                      mass: 0.5,
-                    }}
-                  />
-                )}
-                <span className="relative z-10 flex items-center">
-                  <motion.span
-                    initial={false}
-                    animate={{
-                      width: active ? 14 : 0,
-                      marginLeft: active ? -4 : 0,
-                      marginRight: active ? 4 : 0,
-                      opacity: active ? 1 : 0,
-                    }}
-                    transition={{ duration: 0.2, ease: [0.165, 0.84, 0.44, 1] }}
-                    className="inline-flex shrink-0 items-center justify-center overflow-hidden"
-                  >
-                    <HugeiconsIcon
-                      icon={item.icon}
-                      className="size-3.5 -mt-px shrink-0"
-                    />
-                  </motion.span>
-                  {item.label}
-                </span>
-              </Link>
-            );
-          })}
-        </nav>
-
-        {/* Right: docs/tour desktop — one wrapper per control so flex gap is even (HoverCard roots can confuse flex spacing). */}
-        <div className="hidden items-center justify-self-end gap-0 md:flex">
-          <div className="flex shrink-0 items-center">
-            <AnimatedThemeToggler
-              className="flex h-9 w-9 items-center justify-center rounded-md text-muted-foreground transition-colors hover:bg-accent hover:text-foreground [&_svg]:size-4"
-              title="Toggle theme"
-              aria-label="Toggle theme"
-            />
-          </div>
-          <div className="flex shrink-0 items-center">
-            <HoverCard openDelay={200} closeDelay={100}>
-              <HoverCardTrigger asChild={true}>
-                <a
-                  href="https://unsloth.ai/docs"
-                  target="_blank"
-                  rel="noopener noreferrer"
-                  className="inline-flex h-9 items-center gap-1.5 rounded-md px-3 text-sm font-medium text-emerald-600 transition-colors hover:bg-accent hover:text-emerald-700 dark:hover:text-emerald-400"
-                >
-                  <HugeiconsIcon icon={Book03Icon} className="size-4" />
-                  Learn more
-                </a>
-              </HoverCardTrigger>
-              <HoverCardContent align="end" className="w-80 p-0">
-                <a
-                  href="https://unsloth.ai/docs"
-                  target="_blank"
-                  rel="noopener noreferrer"
-                  className="group/card flex flex-col gap-1 p-4 no-underline"
-                >
-                  <p className="text-sm font-semibold font-heading">
-                    Unsloth Documentation
-                  </p>
-                  <p className="text-xs text-muted-foreground leading-relaxed">
-                    Guides on fine-tuning LLMs 2x faster with 70% less memory.
-                    Covers LoRA, QLoRA, data formatting, and deployment.
-                  </p>
-                  <span className="mt-1 flex items-center gap-1 text-xs font-medium text-emerald-600 group-hover/card:underline">
-                    Visit docs
-                    <HugeiconsIcon icon={ArrowRight01Icon} className="size-3" />
-                  </span>
-                </a>
-              </HoverCardContent>
-            </HoverCard>
-          </div>
-          <div className="flex shrink-0 items-center">
-            <Link
-              to="/api-keys"
-              className={cn(
-                "flex h-9 items-center gap-1.5 rounded-md px-3 text-sm font-medium transition-colors hover:bg-accent",
-                pathname === "/api-keys"
-                  ? "text-foreground"
-                  : "text-muted-foreground hover:text-foreground",
-              )}
-            >
-              <HugeiconsIcon icon={Key01Icon} className="size-4" />
-              API Keys
-            </Link>
-          </div>
-          {tourId ? (
-            <div className="flex shrink-0 items-center">
-              <button
-                type="button"
-                onClick={openTour}
-                className="flex h-9 items-center gap-1.5 rounded-md px-3 text-muted-foreground transition-colors hover:bg-accent hover:text-foreground"
-                title="Tour"
-              >
-                <HugeiconsIcon icon={CursorInfo02Icon} className="size-4" />
-                <span className="text-sm font-medium">Tour</span>
-              </button>
-            </div>
-          ) : null}
-          <div className="flex shrink-0 items-center">
-            <HoverCard openDelay={200} closeDelay={100}>
-              <HoverCardTrigger asChild={true}>
-                <button
-                  type="button"
-                  className="flex h-9 items-center gap-1.5 rounded-md px-3 text-sm font-medium text-muted-foreground transition-colors hover:bg-accent hover:text-foreground"
-                  aria-label="How to update Unsloth Studio"
-                >
-                  <HugeiconsIcon icon={ArrowReloadHorizontalIcon} className="size-4" />
-                  Update
-                </button>
-              </HoverCardTrigger>
-              <HoverCardContent align="end" className="w-[22.5rem] p-0">
-                <UpdateStudioInstructions
-                  className="p-4"
-                  defaultShell={defaultUpdateShell}
-                />
-              </HoverCardContent>
-            </HoverCard>
-          </div>
-          <div className="flex shrink-0 items-center">
-            <button
-              type="button"
-              onClick={() => setShutdownOpen(true)}
-              className="flex h-9 w-9 items-center justify-center rounded-md text-muted-foreground transition-colors hover:bg-accent hover:text-foreground"
-              title="Shut down Unsloth Studio server"
-              aria-label="Shut down Unsloth Studio server"
-            >
-              <HugeiconsIcon icon={Cancel01Icon} className="size-5" />
-            </button>
-          </div>
-        </div>
-
-        {/* Right: mobile */}
-        <div className="col-start-3 flex items-center gap-2 justify-self-end md:hidden">
-          {tourId ? (
-            <button
-              type="button"
-              onClick={openTour}
-              className="flex h-9 w-9 items-center justify-center rounded-md text-muted-foreground transition-colors hover:bg-accent hover:text-foreground"
-              title="Tour"
-            >
-              <HugeiconsIcon icon={CursorInfo02Icon} className="size-4" />
-            </button>
-          ) : null}
-          <Sheet
-            open={mobileOpen}
-            onOpenChange={(open) => {
-              setMobileOpen(open);
-              if (!open) setMobileUpdateOpen(false);
-            }}
-          >
-            <SheetTrigger asChild={true}>
-              <button
-                type="button"
-                className="rounded-md border border-border px-3 py-1.5 text-sm font-medium text-foreground"
-                aria-label="Open navigation menu"
-              >
-                Menu
-              </button>
-            </SheetTrigger>
-            <SheetContent side="right" className="w-[300px] p-4">
-              <SheetHeader>
-                <SheetTitle>Navigate</SheetTitle>
-              </SheetHeader>
-              <div className="mt-6 flex max-h-[calc(100dvh-8rem)] flex-col gap-2 overflow-y-auto pr-1">
-                {NAV_ITEMS.filter((item) => item.enabled).map((item) => {
-                  const active = pathname === item.href;
-                  const disabledByTraining =
-                    isTrainingRunning && item.href !== "/studio";
-                  const disabledByDevice =
-                    chatOnly && item.href !== "/chat" && item.href !== "/data-recipes";
-                  if (disabledByTraining || disabledByDevice) {
-                    return (
-                      <span
-                        key={item.href}
-                        className="flex items-center gap-2 rounded-md border border-border px-3 py-2 text-sm font-medium text-muted-foreground/40 cursor-not-allowed"
-                      >
-                        <HugeiconsIcon icon={item.icon} className="size-4" />
-                        {item.label}
-                      </span>
-                    );
-                  }
-                  return (
-                    <Link
-                      key={item.href}
-                      to={item.href}
-                      onClick={() => setMobileOpen(false)}
-                      className={cn(
-                        "flex items-center gap-2 rounded-md border px-3 py-2 text-sm font-medium",
-                        active
-                          ? "border-foreground bg-foreground text-background"
-                          : "border-border text-foreground hover:bg-accent",
-                      )}
-                    >
-                      <HugeiconsIcon icon={item.icon} className="size-4" />
-                      {item.label}
-                    </Link>
-                  );
-                })}
-                <Link
-                  to="/api-keys"
-                  onClick={() => setMobileOpen(false)}
-                  className={cn(
-                    "mt-3 flex items-center gap-2 rounded-md border px-3 py-2 text-sm font-medium",
-                    pathname === "/api-keys"
-                      ? "border-foreground bg-foreground text-background"
-                      : "border-border text-foreground hover:bg-accent",
-                  )}
-                >
-                  <HugeiconsIcon icon={Key01Icon} className="size-4" />
-                  API Keys
-                </Link>
-                <a
-                  href="https://unsloth.ai/docs"
-                  target="_blank"
-                  rel="noopener noreferrer"
-                  className="flex items-center gap-2 rounded-md border border-border px-3 py-2 text-sm font-medium text-foreground hover:bg-accent"
-                  onClick={() => setMobileOpen(false)}
-                >
-                  <HugeiconsIcon icon={Book03Icon} className="size-4" />
-                  Learn more (Docs)
-                </a>
-                {tourId ? (
-                  <button
-                    type="button"
-                    className="flex items-center gap-2 rounded-md border border-border px-3 py-2 text-left text-sm font-medium text-foreground hover:bg-accent"
-                    onClick={() => {
-                      openTour();
-                      setMobileOpen(false);
-                    }}
-                  >
-                    <HugeiconsIcon icon={CursorInfo02Icon} className="size-4" />
-                    Start tour
-                  </button>
-                ) : null}
-                <Collapsible
-                  open={mobileUpdateOpen}
-                  onOpenChange={setMobileUpdateOpen}
-                  className="rounded-md border border-border"
-                >
-                  <CollapsibleTrigger asChild={true}>
-                    <button
-                      type="button"
-                      className="flex w-full items-center justify-between rounded-md px-3 py-2 text-left text-sm font-medium text-foreground transition-colors hover:bg-accent"
-                      aria-label="Toggle update instructions"
-                    >
-                      <span className="flex items-center gap-2">
-                        <HugeiconsIcon icon={ArrowReloadHorizontalIcon} className="size-4" />
-                        Update Unsloth Studio
-                      </span>
-                      <HugeiconsIcon
-                        icon={ArrowRight01Icon}
-                        className={cn(
-                          "size-4 text-muted-foreground transition-transform",
-                          mobileUpdateOpen && "rotate-90",
-                        )}
-                      />
-                    </button>
-                  </CollapsibleTrigger>
-                  <CollapsibleContent className="border-t border-border p-3 pt-2">
-                    <UpdateStudioInstructions
-                      defaultShell={defaultUpdateShell}
-                      showTitle={false}
-                    />
-                  </CollapsibleContent>
-                </Collapsible>
-                <button
-                  type="button"
-                  className="mt-3 flex items-center gap-2 rounded-md border border-border px-3 py-2 text-left text-sm font-medium text-foreground hover:bg-accent"
-                  onClick={() => {
-                    setMobileOpen(false);
-                    setShutdownOpen(true);
-                  }}
-                >
-                  <HugeiconsIcon icon={Cancel01Icon} className="size-5" />
-                  Quit Unsloth Studio
-                </button>
-                <div className="mt-2 flex items-center justify-between rounded-md border border-border px-3 py-2">
-                  <span className="text-sm font-medium text-foreground">Theme</span>
-                  <AnimatedThemeToggler
-                    className="flex h-8 w-8 items-center justify-center rounded-md text-muted-foreground transition-colors hover:bg-accent hover:text-foreground [&_svg]:size-4"
-                    title="Toggle theme"
-                    aria-label="Toggle theme"
-                  />
-                </div>
-              </div>
-            </SheetContent>
-          </Sheet>
-        </div>
+    <header className="absolute top-0 inset-x-0 z-40 h-11 pointer-events-none">
+      <div className="flex h-full items-center pl-2">
+        <SidebarTrigger className="pointer-events-auto" />
       </div>
     </header>
-
-    <ShutdownDialog
-      open={shutdownOpen}
-      onOpenChange={setShutdownOpen}
-      onBeforeShutdown={removeUnloadHandler}
-    />
-    </>
   );
 }
diff --git a/studio/frontend/src/components/shutdown-dialog.tsx b/studio/frontend/src/components/shutdown-dialog.tsx
index dea738bf6b..dfeafb33eb 100644
--- a/studio/frontend/src/components/shutdown-dialog.tsx
+++ b/studio/frontend/src/components/shutdown-dialog.tsx
@@ -18,16 +18,17 @@ import {
 interface ShutdownDialogProps {
   open: boolean;
   onOpenChange: (open: boolean) => void;
-  /** Called right before the shutdown API request so callers can remove the
-   *  beforeunload listener — otherwise the "Server stopped" page would still
-   *  trigger a "Leave site?" prompt when the user tries to close it. */
-  onBeforeShutdown?: () => void;
+  /** Called after the shutdown API returns success, right before we replace
+   *  document.body with the "Server stopped" page. Callers use this to remove
+   *  their beforeunload listener — otherwise the browser would prompt
+   *  "Leave site?" when the user tries to close the final tab. */
+  onAfterShutdown?: () => void;
 }
 
 export function ShutdownDialog({
   open,
   onOpenChange,
-  onBeforeShutdown,
+  onAfterShutdown,
 }: ShutdownDialogProps) {
   const [stopping, setStopping] = useState(false);
 
@@ -49,7 +50,7 @@ export function ShutdownDialog({
       return;
     }
 
-    onBeforeShutdown?.();
+    onAfterShutdown?.();
     document.body.innerHTML = `
       <div style="display:flex;flex-direction:column;align-items:center;justify-content:center;height:100vh;font-family:sans-serif;gap:12px">
         <p style="font-size:1.1rem;font-weight:600;margin:0">Unsloth Studio has stopped.</p>
diff --git a/studio/frontend/src/components/ui/animated-theme-toggler.tsx b/studio/frontend/src/components/ui/animated-theme-toggler.tsx
index 24f3c68ec9..d83e278401 100644
--- a/studio/frontend/src/components/ui/animated-theme-toggler.tsx
+++ b/studio/frontend/src/components/ui/animated-theme-toggler.tsx
@@ -6,11 +6,73 @@ import { Moon, Sun } from "lucide-react"
 import { flushSync } from "react-dom"
 
 import { cn } from "@/lib/utils"
+import { setTheme } from "@/features/settings/stores/theme-store"
 
 interface AnimatedThemeTogglerProps extends React.ComponentPropsWithoutRef<"button"> {
   duration?: number
 }
 
+export function useAnimatedThemeToggle(duration = 400) {
+  const [isDark, setIsDark] = useState(false)
+  const anchorRef = useRef<HTMLElement | null>(null)
+
+  useEffect(() => {
+    const updateTheme = () => {
+      setIsDark(document.documentElement.classList.contains("dark"))
+    }
+    updateTheme()
+    const observer = new MutationObserver(updateTheme)
+    observer.observe(document.documentElement, {
+      attributes: true,
+      attributeFilter: ["class"],
+    })
+    return () => observer.disconnect()
+  }, [])
+
+  const toggleTheme = useCallback(async () => {
+    const anchor = anchorRef.current
+    const applyTheme = () => {
+      flushSync(() => {
+        const newTheme = !isDark
+        setIsDark(newTheme)
+        setTheme(newTheme ? "dark" : "light")
+      })
+    }
+
+    if (!document.startViewTransition) {
+      applyTheme()
+      return
+    }
+
+    await document.startViewTransition(applyTheme).ready
+
+    if (anchor) {
+      const { top, left, width, height } = anchor.getBoundingClientRect()
+      const x = left + width / 2
+      const y = top + height / 2
+      const maxRadius = Math.hypot(
+        Math.max(left, window.innerWidth - left),
+        Math.max(top, window.innerHeight - top)
+      )
+      document.documentElement.animate(
+        {
+          clipPath: [
+            `circle(0px at ${x}px ${y}px)`,
+            `circle(${maxRadius}px at ${x}px ${y}px)`,
+          ],
+        },
+        {
+          duration,
+          easing: "ease-in-out",
+          pseudoElement: "::view-transition-new(root)",
+        }
+      )
+    }
+  }, [isDark, duration])
+
+  return { isDark, toggleTheme, anchorRef }
+}
+
 export const AnimatedThemeToggler = ({
   className,
   duration = 400,
@@ -38,14 +100,20 @@ export const AnimatedThemeToggler = ({
   const toggleTheme = useCallback(async () => {
     if (!buttonRef.current) return
 
-    await document.startViewTransition(() => {
+    const apply = () => {
       flushSync(() => {
         const newTheme = !isDark
         setIsDark(newTheme)
-        document.documentElement.classList.toggle("dark")
-        localStorage.setItem("theme", newTheme ? "dark" : "light")
+        setTheme(newTheme ? "dark" : "light")
       })
-    }).ready
+    }
+
+    if (!document.startViewTransition) {
+      apply()
+      return
+    }
+
+    await document.startViewTransition(apply).ready
 
     const { top, left, width, height } =
       buttonRef.current.getBoundingClientRect()
diff --git a/studio/frontend/src/components/ui/command.tsx b/studio/frontend/src/components/ui/command.tsx
index 6181d55ea1..a2340b62ed 100644
--- a/studio/frontend/src/components/ui/command.tsx
+++ b/studio/frontend/src/components/ui/command.tsx
@@ -1,6 +1,6 @@
-// SPDX-License-Identifier: AGPL-3.0-only
-// Copyright 2026-present the Unsloth AI Inc. team. All rights reserved. See /studio/LICENSE.AGPL-3.0
-
+// SPDX-License-Identifier: AGPL-3.0-only
+// Copyright 2026-present the Unsloth AI Inc. team. All rights reserved. See /studio/LICENSE.AGPL-3.0
+
 "use client";
 
 import { Command as CommandPrimitive } from "cmdk";
@@ -39,12 +39,14 @@ function CommandDialog({
   description = "Search for a command to run...",
   children,
   className,
+  overlayClassName,
   showCloseButton = false,
   ...props
 }: React.ComponentProps<typeof Dialog> & {
   title?: string;
   description?: string;
   className?: string;
+  overlayClassName?: string;
   showCloseButton?: boolean;
 }) {
   return (
@@ -55,9 +57,10 @@ function CommandDialog({
       </DialogHeader>
       <DialogContent
         className={cn(
-          "rounded-4xl! p-0 top-1/3 translate-y-0 overflow-hidden p-0",
+          "rounded-4xl! top-1/3 translate-y-0 overflow-hidden p-0",
           className,
         )}
+        overlayClassName={overlayClassName}
         showCloseButton={showCloseButton}
       >
         {children}
diff --git a/studio/frontend/src/components/ui/sidebar.tsx b/studio/frontend/src/components/ui/sidebar.tsx
index 898972f1ee..6d9016f8e2 100644
--- a/studio/frontend/src/components/ui/sidebar.tsx
+++ b/studio/frontend/src/components/ui/sidebar.tsx
@@ -1,6 +1,6 @@
-// SPDX-License-Identifier: AGPL-3.0-only
-// Copyright 2026-present the Unsloth AI Inc. team. All rights reserved. See /studio/LICENSE.AGPL-3.0
-
+// SPDX-License-Identifier: AGPL-3.0-only
+// Copyright 2026-present the Unsloth AI Inc. team. All rights reserved. See /studio/LICENSE.AGPL-3.0
+
 "use client"
 
 import * as React from "react"
@@ -28,10 +28,9 @@ import { useIsMobile } from "@/hooks/use-mobile"
 import { HugeiconsIcon } from "@hugeicons/react"
 import { SidebarLeftIcon } from "@hugeicons/core-free-icons"
 
-const SIDEBAR_COOKIE_NAME = "sidebar_state"
-const SIDEBAR_COOKIE_MAX_AGE = 60 * 60 * 24 * 7
+const noop = () => {}
+
 const SIDEBAR_WIDTH = "16rem"
-const SIDEBAR_WIDTH_MOBILE = "18rem"
 const SIDEBAR_WIDTH_ICON = "3rem"
 const SIDEBAR_KEYBOARD_SHORTCUT = "b"
 
@@ -43,6 +42,10 @@ type SidebarContextProps = {
   setOpenMobile: (open: boolean) => void
   isMobile: boolean
   toggleSidebar: () => void
+  hasPinMode: boolean
+  pinned: boolean
+  setPinned: (value: boolean) => void
+  togglePinned: () => void
 }
 
 const SidebarContext = React.createContext<SidebarContextProps | null>(null)
@@ -60,6 +63,9 @@ function SidebarProvider({
   defaultOpen = true,
   open: openProp,
   onOpenChange: setOpenProp,
+  pinned: pinnedProp,
+  setPinned: setPinnedProp,
+  togglePinned: togglePinnedProp,
   className,
   style,
   children,
@@ -68,33 +74,57 @@ function SidebarProvider({
   defaultOpen?: boolean
   open?: boolean
   onOpenChange?: (open: boolean) => void
+  pinned?: boolean
+  setPinned?: (value: boolean) => void
+  togglePinned?: () => void
 }) {
   const isMobile = useIsMobile()
   const [openMobile, setOpenMobile] = React.useState(false)
 
+  const prevIsMobileRef = React.useRef(isMobile)
+  React.useEffect(() => {
+    if (prevIsMobileRef.current && !isMobile) {
+      setOpenMobile(false)
+    }
+    prevIsMobileRef.current = isMobile
+  }, [isMobile])
+
+  // Whether pin mode is active (caller provides pinned + setPinned + togglePinned).
+  const hasPinMode = pinnedProp !== undefined && setPinnedProp !== undefined && togglePinnedProp !== undefined
+
   // This is the internal state of the sidebar.
   // We use openProp and setOpenProp for control from outside the component.
   const [_open, _setOpen] = React.useState(defaultOpen)
-  const open = openProp ?? _open
+
+  // When pin mode is active, open is driven entirely by `pinned` (explicit
+  // user toggle). Otherwise fall back to the controlled/uncontrolled pattern.
+  const open = hasPinMode ? !!pinnedProp : (openProp ?? _open)
+
   const setOpen = React.useCallback(
     (value: boolean | ((value: boolean) => boolean)) => {
       const openState = typeof value === "function" ? value(open) : value
+
+      if (hasPinMode) {
+        // In pin mode, setOpen controls pinned state.
+        setPinnedProp?.(openState)
+        return
+      }
+
       if (setOpenProp) {
         setOpenProp(openState)
       } else {
         _setOpen(openState)
       }
-
-      // This sets the cookie to keep the sidebar state.
-      document.cookie = `${SIDEBAR_COOKIE_NAME}=${openState}; path=/; max-age=${SIDEBAR_COOKIE_MAX_AGE}`
     },
-    [setOpenProp, open]
+    [setOpenProp, open, hasPinMode, setPinnedProp]
   )
 
   // Helper to toggle the sidebar.
   const toggleSidebar = React.useCallback(() => {
-    return isMobile ? setOpenMobile((open) => !open) : setOpen((open) => !open)
-  }, [isMobile, setOpen, setOpenMobile])
+    if (isMobile) return setOpenMobile((open) => !open)
+    if (hasPinMode && togglePinnedProp) return togglePinnedProp()
+    return setOpen((open) => !open)
+  }, [isMobile, setOpen, setOpenMobile, hasPinMode, togglePinnedProp])
 
   // Adds a keyboard shortcut to toggle the sidebar.
   React.useEffect(() => {
@@ -116,6 +146,10 @@ function SidebarProvider({
   // This makes it easier to style the sidebar with Tailwind classes.
   const state = open ? "expanded" : "collapsed"
 
+  const pinned = pinnedProp ?? false
+  const setPinned = setPinnedProp ?? noop
+  const togglePinned = togglePinnedProp ?? noop
+
   const contextValue = React.useMemo<SidebarContextProps>(
     () => ({
       state,
@@ -125,8 +159,12 @@ function SidebarProvider({
       openMobile,
       setOpenMobile,
       toggleSidebar,
+      hasPinMode,
+      pinned,
+      setPinned,
+      togglePinned,
     }),
-    [state, open, setOpen, isMobile, openMobile, setOpenMobile, toggleSidebar]
+    [state, open, setOpen, isMobile, openMobile, setOpenMobile, toggleSidebar, hasPinMode, pinned, setPinned, togglePinned]
   )
 
   return (
@@ -165,7 +203,7 @@ function Sidebar({
   variant?: "sidebar" | "floating" | "inset"
   collapsible?: "offcanvas" | "icon" | "none"
 }) {
-  const { isMobile, state, openMobile, setOpenMobile } = useSidebar()
+  const { isMobile, state, openMobile, setOpenMobile, hasPinMode, pinned } = useSidebar()
 
   if (collapsible === "none") {
     return (
@@ -190,12 +228,7 @@ function Sidebar({
           data-sidebar="sidebar"
           data-slot="sidebar"
           data-mobile="true"
-          className="bg-sidebar text-sidebar-foreground w-(--sidebar-width) p-0 [&>button]:hidden"
-          style={
-            {
-              "--sidebar-width": SIDEBAR_WIDTH_MOBILE,
-            } as React.CSSProperties
-          }
+          className="bg-sidebar text-sidebar-foreground w-2/3 max-w-[18rem] p-0 [&>button]:hidden"
           side={side}
         >
           <SheetHeader className="sr-only">
@@ -210,7 +243,11 @@ function Sidebar({
 
   return (
     <div
-      className="group peer text-sidebar-foreground hidden md:block"
+      className={cn(
+        "group peer text-sidebar-foreground relative shrink-0",
+        hasPinMode && pinned && "w-(--sidebar-width)",
+        hasPinMode && !pinned && "w-(--sidebar-width-icon)",
+      )}
       data-state={state}
       data-collapsible={state === "collapsed" ? collapsible : ""}
       data-variant={variant}
@@ -221,23 +258,45 @@ function Sidebar({
       <div
         data-slot="sidebar-gap"
         className={cn(
-          "transition-[width] duration-200 ease-linear relative w-(--sidebar-width) bg-transparent",
-          "group-data-[collapsible=offcanvas]:w-0",
+          "relative bg-transparent shrink-0",
           "group-data-[side=right]:rotate-180",
-          variant === "floating" || variant === "inset"
-            ? "group-data-[collapsible=icon]:w-[calc(var(--sidebar-width-icon)+(--spacing(4)))]"
-            : "group-data-[collapsible=icon]:w-(--sidebar-width-icon)"
+          hasPinMode
+            ? cn(
+                // Pin mode: always push content. Expanded when pinned.
+                pinned
+                  ? "w-(--sidebar-width)"
+                  : (variant === "floating" || variant === "inset"
+                      ? "w-[calc(var(--sidebar-width-icon)+(--spacing(4)))]"
+                      : "w-(--sidebar-width-icon)"),
+              )
+            : cn(
+                // Legacy mode: original shadcn behavior.
+                "w-(--sidebar-width)",
+                "group-data-[collapsible=offcanvas]:w-0",
+                variant === "floating" || variant === "inset"
+                  ? "group-data-[collapsible=icon]:w-[calc(var(--sidebar-width-icon)+(--spacing(4)))]"
+                  : "group-data-[collapsible=icon]:w-(--sidebar-width-icon)",
+              ),
         )}
       />
       <div
         data-slot="sidebar-container"
         data-side={side}
         className={cn(
-          "fixed inset-y-0 z-10 hidden h-svh w-(--sidebar-width) transition-[left,right,width] duration-200 ease-linear data-[side=left]:left-0 data-[side=left]:group-data-[collapsible=offcanvas]:left-[calc(var(--sidebar-width)*-1)] data-[side=right]:right-0 data-[side=right]:group-data-[collapsible=offcanvas]:right-[calc(var(--sidebar-width)*-1)] md:flex",
+          hasPinMode
+            ? cn(
+                // Pin mode: always push content, full height.
+                "absolute top-0 bottom-0 flex w-(--sidebar-width) data-[side=left]:left-0",
+                "group-data-[collapsible=icon]:w-(--sidebar-width-icon)",
+              )
+            : cn(
+                // Legacy mode: fixed to viewport (original shadcn behavior).
+                "fixed inset-y-0 z-10 flex h-svh w-(--sidebar-width) data-[side=left]:left-0 data-[side=left]:group-data-[collapsible=offcanvas]:left-[calc(var(--sidebar-width)*-1)] data-[side=right]:right-0 data-[side=right]:group-data-[collapsible=offcanvas]:right-[calc(var(--sidebar-width)*-1)]",
+              ),
           // Adjust the padding for floating and inset variants.
           variant === "floating" || variant === "inset"
             ? "p-2 group-data-[collapsible=icon]:w-[calc(var(--sidebar-width-icon)+(--spacing(4))+2px)]"
-            : "group-data-[collapsible=icon]:w-(--sidebar-width-icon) group-data-[side=left]:border-r group-data-[side=right]:border-l",
+            : !hasPinMode && "group-data-[collapsible=icon]:w-(--sidebar-width-icon) group-data-[side=left]:border-r group-data-[side=right]:border-l",
           className
         )}
         {...props}
@@ -245,7 +304,11 @@ function Sidebar({
         <div
           data-sidebar="sidebar"
           data-slot="sidebar-inner"
-          className="bg-sidebar group-data-[variant=floating]:ring-sidebar-border group-data-[variant=floating]:rounded-lg group-data-[variant=floating]:shadow-sm group-data-[variant=floating]:ring-1 flex size-full flex-col"
+          className={cn(
+            "bg-sidebar flex size-full flex-col overflow-hidden border-r border-sidebar-border",
+            "group-data-[variant=floating]:ring-sidebar-border group-data-[variant=floating]:rounded-lg group-data-[variant=floating]:shadow-sm group-data-[variant=floating]:ring-1",
+            hasPinMode && "ring-1 ring-sidebar-border/60",
+          )}
         >
           {children}
         </div>
@@ -310,7 +373,7 @@ function SidebarInset({ className, ...props }: React.ComponentProps<"main">) {
     <main
       data-slot="sidebar-inset"
       className={cn(
-        "bg-background md:peer-data-[variant=inset]:m-2 md:peer-data-[variant=inset]:ml-0 md:peer-data-[variant=inset]:rounded-xl md:peer-data-[variant=inset]:shadow-sm md:peer-data-[variant=inset]:peer-data-[state=collapsed]:ml-2 relative flex w-full flex-1 flex-col",
+        "bg-background md:peer-data-[variant=inset]:m-2 md:peer-data-[variant=inset]:ml-0 md:peer-data-[variant=inset]:rounded-xl md:peer-data-[variant=inset]:shadow-sm md:peer-data-[variant=inset]:peer-data-[state=collapsed]:ml-2 relative flex min-h-0 w-full flex-1 flex-col",
         className
       )}
       {...props}
@@ -374,7 +437,7 @@ function SidebarContent({ className, ...props }: React.ComponentProps<"div">) {
       data-slot="sidebar-content"
       data-sidebar="content"
       className={cn(
-        "no-scrollbar gap-2 flex min-h-0 flex-1 flex-col overflow-auto group-data-[collapsible=icon]:overflow-hidden",
+        "gap-2 flex min-h-0 flex-1 flex-col overflow-y-auto overflow-x-hidden group-data-[collapsible=icon]:overflow-hidden [&>*]:shrink-0",
         className
       )}
       {...props}
@@ -408,7 +471,7 @@ function SidebarGroupLabel({
       data-slot="sidebar-group-label"
       data-sidebar="group-label"
       className={cn(
-        "text-sidebar-foreground/70 ring-sidebar-ring h-8 rounded-md px-2 text-xs font-medium transition-[margin,opacity] duration-200 ease-linear group-data-[collapsible=icon]:-mt-8 group-data-[collapsible=icon]:opacity-0 focus-visible:ring-2 [&>svg]:size-4 flex shrink-0 items-center outline-hidden [&>svg]:shrink-0",
+        "text-[#94a3b8] dark:text-[#64748b] ring-sidebar-ring h-auto pt-3 pb-2 px-4 rounded-md text-[10px] font-semibold uppercase tracking-[0.08em] group-data-[collapsible=icon]:-mt-8 group-data-[collapsible=icon]:opacity-0 focus-visible:ring-2 [&>svg]:size-3 flex shrink-0 items-center outline-hidden [&>svg]:shrink-0",
         className
       )}
       {...props}
@@ -455,7 +518,7 @@ function SidebarMenu({ className, ...props }: React.ComponentProps<"ul">) {
     <ul
       data-slot="sidebar-menu"
       data-sidebar="menu"
-      className={cn("gap-1 flex w-full min-w-0 flex-col", className)}
+      className={cn("gap-0.5 flex w-full min-w-0 flex-col", className)}
       {...props}
     />
   )
@@ -473,7 +536,7 @@ function SidebarMenuItem({ className, ...props }: React.ComponentProps<"li">) {
 }
 
 const sidebarMenuButtonVariants = cva(
-  "ring-sidebar-ring hover:bg-sidebar-accent hover:text-sidebar-accent-foreground active:bg-sidebar-accent active:text-sidebar-accent-foreground data-active:bg-sidebar-accent data-active:text-sidebar-accent-foreground data-open:hover:bg-sidebar-accent data-open:hover:text-sidebar-accent-foreground gap-2 rounded-lg corner-squircle p-2 text-left text-sm transition-[width,height,padding] group-has-data-[sidebar=menu-action]/menu-item:pr-8 group-data-[collapsible=icon]:size-8! group-data-[collapsible=icon]:p-2! focus-visible:ring-2 data-active:font-medium peer/menu-button flex w-full items-center overflow-hidden outline-hidden group/menu-button disabled:pointer-events-none disabled:opacity-50 aria-disabled:pointer-events-none aria-disabled:opacity-50 [&>span:last-child]:truncate [&_svg]:size-4 [&_svg]:shrink-0",
+  "ring-sidebar-ring hover:bg-sidebar-accent hover:text-sidebar-accent-foreground active:bg-sidebar-accent active:text-sidebar-accent-foreground data-active:bg-sidebar-accent data-active:text-sidebar-accent-foreground data-open:hover:bg-sidebar-accent data-open:hover:text-sidebar-accent-foreground gap-2 rounded-md p-2 text-left text-sm cursor-pointer transition-[width,height,padding] group-has-data-[sidebar=menu-action]/menu-item:pr-8 group-data-[collapsible=icon]:w-full! group-data-[collapsible=icon]:justify-center group-data-[collapsible=icon]:p-2! focus-visible:ring-2 data-active:font-medium peer/menu-button flex w-full items-center overflow-hidden outline-hidden group/menu-button disabled:pointer-events-none disabled:opacity-50 aria-disabled:pointer-events-none aria-disabled:opacity-50 [&>span:last-child]:truncate group-data-[collapsible=icon]:[&>span]:hidden [&_svg]:size-4 [&_svg]:shrink-0 group-data-[collapsible=icon]:[&_svg]:size-5",
   {
     variants: {
       variant: {
diff --git a/studio/frontend/src/features/auth/api-keys-page.tsx b/studio/frontend/src/features/auth/api-keys-page.tsx
deleted file mode 100644
index 2065384a7d..0000000000
--- a/studio/frontend/src/features/auth/api-keys-page.tsx
+++ /dev/null
@@ -1,426 +0,0 @@
-// SPDX-License-Identifier: AGPL-3.0-only
-// Copyright 2026-present the Unsloth AI Inc. team. All rights reserved. See /studio/LICENSE.AGPL-3.0
-
-import { DashboardLayout } from "@/components/layout/dashboard-layout";
-import { Button } from "@/components/ui/button";
-import {
-  Dialog,
-  DialogContent,
-  DialogDescription,
-  DialogFooter,
-  DialogHeader,
-  DialogTitle,
-} from "@/components/ui/dialog";
-import { Input } from "@/components/ui/input";
-import { Label } from "@/components/ui/label";
-import { cn } from "@/lib/utils";
-import { copyToClipboardAsync } from "@/lib/copy-to-clipboard";
-import {
-  AlertCircleIcon,
-  Copy01Icon,
-  Delete02Icon,
-  Key01Icon,
-  Tick02Icon,
-} from "@hugeicons/core-free-icons";
-import { HugeiconsIcon } from "@hugeicons/react";
-import { useCallback, useEffect, useRef, useState } from "react";
-import { authFetch } from "./api";
-
-// ---------------------------------------------------------------------------
-// Types
-// ---------------------------------------------------------------------------
-
-interface ApiKey {
-  id: number;
-  name: string;
-  key_prefix: string;
-  created_at: string;
-  last_used_at: string | null;
-  expires_at: string | null;
-  is_active: boolean;
-}
-
-// ---------------------------------------------------------------------------
-// API helpers
-// ---------------------------------------------------------------------------
-
-async function fetchApiKeys(): Promise<ApiKey[]> {
-  const res = await authFetch("/api/auth/api-keys");
-  if (!res.ok) throw new Error("Failed to load API keys");
-  const data = (await res.json()) as { api_keys: ApiKey[] };
-  return data.api_keys;
-}
-
-async function createApiKey(
-  name: string,
-  expiresInDays: number | null,
-): Promise<{ key: string; api_key: ApiKey }> {
-  const res = await authFetch("/api/auth/api-keys", {
-    method: "POST",
-    headers: { "Content-Type": "application/json" },
-    body: JSON.stringify({
-      name,
-      expires_in_days: expiresInDays,
-    }),
-  });
-  if (!res.ok) throw new Error("Failed to create API key");
-  return res.json();
-}
-
-async function revokeApiKey(keyId: number): Promise<void> {
-  const res = await authFetch(`/api/auth/api-keys/${keyId}`, {
-    method: "DELETE",
-  });
-  if (!res.ok) throw new Error("Failed to revoke API key");
-}
-
-// ---------------------------------------------------------------------------
-// Helpers
-// ---------------------------------------------------------------------------
-
-function formatDate(iso: string | null): string {
-  if (!iso) return "--";
-  const d = new Date(iso);
-  return d.toLocaleDateString(undefined, {
-    year: "numeric",
-    month: "short",
-    day: "numeric",
-    hour: "2-digit",
-    minute: "2-digit",
-  });
-}
-
-// ---------------------------------------------------------------------------
-// Components
-// ---------------------------------------------------------------------------
-
-function CopyButton({ text }: { text: string }) {
-  const [copied, setCopied] = useState(false);
-  const timerRef = useRef<ReturnType<typeof setTimeout> | null>(null);
-
-  useEffect(() => {
-    return () => {
-      if (timerRef.current) clearTimeout(timerRef.current);
-    };
-  }, []);
-
-  const handleCopy = async () => {
-    if (!(await copyToClipboardAsync(text))) return;
-    setCopied(true);
-    if (timerRef.current) clearTimeout(timerRef.current);
-    timerRef.current = setTimeout(() => setCopied(false), 2000);
-  };
-
-  return (
-    <Button
-      variant="ghost"
-      size="icon-sm"
-      onClick={handleCopy}
-      className={cn(
-        "shrink-0 rounded-md text-muted-foreground hover:text-foreground",
-        copied && "text-emerald-600 hover:text-emerald-600",
-      )}
-      aria-label={copied ? "Copied API key" : "Copy API key"}
-      title={copied ? "Copied" : "Copy"}
-    >
-      <HugeiconsIcon
-        icon={copied ? Tick02Icon : Copy01Icon}
-        className="size-4"
-      />
-    </Button>
-  );
-}
-
-function RevealKeyDialog({
-  open,
-  rawKey,
-  onClose,
-}: {
-  open: boolean;
-  rawKey: string;
-  onClose: () => void;
-}) {
-  return (
-    <Dialog open={open} onOpenChange={(o) => !o && onClose()}>
-      <DialogContent className="max-w-lg">
-        <DialogHeader>
-          <DialogTitle>API Key Created</DialogTitle>
-          <DialogDescription>
-            Copy this key now. It will not be shown again.
-          </DialogDescription>
-        </DialogHeader>
-        <div className="flex items-center gap-2 rounded-md border border-border bg-muted/40 p-3">
-          <code className="min-w-0 flex-1 break-all font-mono text-sm">
-            {rawKey}
-          </code>
-          <CopyButton text={rawKey} />
-        </div>
-        <div className="flex items-start gap-2 rounded-md border border-amber-500/20 bg-amber-50 p-3 text-amber-800 dark:border-amber-400/20 dark:bg-amber-950/30 dark:text-amber-300">
-          <HugeiconsIcon icon={AlertCircleIcon} className="mt-0.5 size-4 shrink-0" />
-          <p className="text-xs leading-relaxed">
-            Store this key securely. You will not be able to see it again after closing this dialog.
-          </p>
-        </div>
-        <DialogFooter>
-          <Button onClick={onClose}>Done</Button>
-        </DialogFooter>
-      </DialogContent>
-    </Dialog>
-  );
-}
-
-function CreateKeyForm({ onCreated }: { onCreated: (rawKey: string) => void }) {
-  const [name, setName] = useState("");
-  const [expiresInDays, setExpiresInDays] = useState("");
-  const [loading, setLoading] = useState(false);
-
-  const handleSubmit = async (e: React.FormEvent) => {
-    e.preventDefault();
-    if (!name.trim()) return;
-    setLoading(true);
-    try {
-      const days = expiresInDays ? parseInt(expiresInDays, 10) : null;
-      const result = await createApiKey(name.trim(), days);
-      onCreated(result.key);
-      setName("");
-      setExpiresInDays("");
-    } finally {
-      setLoading(false);
-    }
-  };
-
-  return (
-    <form onSubmit={handleSubmit} className="flex flex-col gap-4 rounded-lg border border-border p-4">
-      <div className="flex flex-col gap-1.5">
-        <Label htmlFor="key-name">Key name</Label>
-        <Input
-          id="key-name"
-          placeholder="e.g. My application"
-          value={name}
-          onChange={(e) => setName(e.target.value)}
-          required
-        />
-      </div>
-      <div className="flex flex-col gap-1.5">
-        <Label htmlFor="key-expiry">Expires in (days)</Label>
-        <Input
-          id="key-expiry"
-          type="number"
-          min={1}
-          placeholder="Leave blank for no expiry"
-          value={expiresInDays}
-          onChange={(e) => setExpiresInDays(e.target.value)}
-        />
-      </div>
-      <Button type="submit" disabled={loading || !name.trim()} className="self-start">
-        {loading ? "Creating..." : "Create API key"}
-      </Button>
-    </form>
-  );
-}
-
-function KeysTable({
-  keys,
-  onRevoke,
-}: {
-  keys: ApiKey[];
-  onRevoke: (id: number) => void;
-}) {
-  if (keys.length === 0) {
-    return (
-      <p className="py-8 text-center text-sm text-muted-foreground">
-        No API keys yet. Create one above.
-      </p>
-    );
-  }
-
-  return (
-    <div className="overflow-x-auto rounded-lg border border-border">
-      <table className="w-full text-sm">
-        <thead>
-          <tr className="border-b border-border bg-muted/40">
-            <th className="px-4 py-2.5 text-left font-medium text-muted-foreground">Name</th>
-            <th className="px-4 py-2.5 text-left font-medium text-muted-foreground">Key</th>
-            <th className="px-4 py-2.5 text-left font-medium text-muted-foreground">Created</th>
-            <th className="px-4 py-2.5 text-left font-medium text-muted-foreground">Last used</th>
-            <th className="px-4 py-2.5 text-left font-medium text-muted-foreground">Expires</th>
-            <th className="px-4 py-2.5 text-right font-medium text-muted-foreground" />
-          </tr>
-        </thead>
-        <tbody>
-          {keys.map((k) => (
-            <tr
-              key={k.id}
-              className={cn(
-                "border-b border-border last:border-b-0",
-                !k.is_active && "opacity-50",
-              )}
-            >
-              <td className="px-4 py-2.5 font-medium">{k.name}</td>
-              <td className="px-4 py-2.5">
-                <code className="rounded bg-muted px-1.5 py-0.5 font-mono text-xs">
-                  sk-unsloth-{k.key_prefix}...
-                </code>
-              </td>
-              <td className="px-4 py-2.5 text-muted-foreground">{formatDate(k.created_at)}</td>
-              <td className="px-4 py-2.5 text-muted-foreground">{formatDate(k.last_used_at)}</td>
-              <td className="px-4 py-2.5 text-muted-foreground">{formatDate(k.expires_at)}</td>
-              <td className="px-4 py-2.5 text-right">
-                {k.is_active ? (
-                  <Button
-                    variant="ghost"
-                    size="sm"
-                    onClick={() => onRevoke(k.id)}
-                    className="text-destructive hover:text-destructive"
-                  >
-                    <HugeiconsIcon icon={Delete02Icon} className="size-3.5 mr-1" />
-                    Revoke
-                  </Button>
-                ) : (
-                  <span className="text-xs text-muted-foreground">Revoked</span>
-                )}
-              </td>
-            </tr>
-          ))}
-        </tbody>
-      </table>
-    </div>
-  );
-}
-
-function UsageExamples() {
-  const base = window.location.origin;
-
-  const curlExample = `curl ${base}/v1/chat/completions \\
-  -H "Authorization: Bearer sk-unsloth-YOUR_KEY" \\
-  -H "Content-Type: application/json" \\
-  -d '{
-    "messages": [{"role": "user", "content": "Hello"}],
-    "stream": true
-  }'`;
-
-  const pythonExample = `from openai import OpenAI
-
-client = OpenAI(
-    base_url="${base}/v1",
-    api_key="sk-unsloth-YOUR_KEY",
-)
-
-response = client.chat.completions.create(
-    model="current",
-    messages=[{"role": "user", "content": "Hello"}],
-    stream=True,
-)
-for chunk in response:
-    print(chunk.choices[0].delta.content or "", end="")`;
-
-  const toolsExample = `curl ${base}/v1/chat/completions \\
-  -H "Authorization: Bearer sk-unsloth-YOUR_KEY" \\
-  -H "Content-Type: application/json" \\
-  -d '{
-    "messages": [{"role": "user", "content": "Search for Python 3.13 features"}],
-    "stream": true,
-    "enable_tools": true,
-    "enabled_tools": ["web_search", "python"],
-    "session_id": "my-session"
-  }'`;
-
-  return (
-    <div className="flex flex-col gap-4">
-      <h3 className="text-sm font-semibold">Usage examples</h3>
-      <div className="flex flex-col gap-3">
-        <div>
-          <p className="mb-1.5 text-xs font-medium text-muted-foreground">curl</p>
-          <pre className="overflow-x-auto rounded-md border border-border bg-muted/40 p-3 font-mono text-xs leading-relaxed">
-            {curlExample}
-          </pre>
-        </div>
-        <div>
-          <p className="mb-1.5 text-xs font-medium text-muted-foreground">Python (OpenAI SDK)</p>
-          <pre className="overflow-x-auto rounded-md border border-border bg-muted/40 p-3 font-mono text-xs leading-relaxed">
-            {pythonExample}
-          </pre>
-        </div>
-        <div>
-          <p className="mb-1.5 text-xs font-medium text-muted-foreground">With tools (web search + code execution)</p>
-          <pre className="overflow-x-auto rounded-md border border-border bg-muted/40 p-3 font-mono text-xs leading-relaxed">
-            {toolsExample}
-          </pre>
-        </div>
-      </div>
-    </div>
-  );
-}
-
-// ---------------------------------------------------------------------------
-// Page
-// ---------------------------------------------------------------------------
-
-export function ApiKeysPage() {
-  const [keys, setKeys] = useState<ApiKey[]>([]);
-  const [revealedKey, setRevealedKey] = useState<string | null>(null);
-  const [error, setError] = useState<string | null>(null);
-
-  const loadKeys = useCallback(async () => {
-    try {
-      setError(null);
-      const loaded = await fetchApiKeys();
-      setKeys(loaded);
-    } catch (e) {
-      setError(e instanceof Error ? e.message : "Failed to load API keys");
-    }
-  }, []);
-
-  useEffect(() => {
-    void loadKeys();
-  }, [loadKeys]);
-
-  const handleCreated = (rawKey: string) => {
-    setRevealedKey(rawKey);
-    void loadKeys();
-  };
-
-  const handleRevoke = async (keyId: number) => {
-    try {
-      await revokeApiKey(keyId);
-      void loadKeys();
-    } catch {
-      setError("Failed to revoke key");
-    }
-  };
-
-  return (
-    <DashboardLayout>
-      <div className="flex flex-col gap-8">
-        <div className="flex items-center gap-3">
-          <div className="flex h-10 w-10 items-center justify-center rounded-lg border border-border bg-muted/40">
-            <HugeiconsIcon icon={Key01Icon} className="size-5" />
-          </div>
-          <div>
-            <h1 className="text-xl font-bold font-heading">API Keys</h1>
-            <p className="text-sm text-muted-foreground">
-              Create keys to access Unsloth Studio programmatically via the OpenAI-compatible API.
-            </p>
-          </div>
-        </div>
-
-        {error && (
-          <div className="flex items-center gap-2 rounded-md border border-destructive/20 bg-destructive/5 p-3 text-sm text-destructive">
-            <HugeiconsIcon icon={AlertCircleIcon} className="size-4 shrink-0" />
-            {error}
-          </div>
-        )}
-
-        <CreateKeyForm onCreated={handleCreated} />
-        <KeysTable keys={keys} onRevoke={handleRevoke} />
-        <UsageExamples />
-      </div>
-
-      <RevealKeyDialog
-        open={revealedKey !== null}
-        rawKey={revealedKey ?? ""}
-        onClose={() => setRevealedKey(null)}
-      />
-    </DashboardLayout>
-  );
-}
diff --git a/studio/frontend/src/features/auth/index.ts b/studio/frontend/src/features/auth/index.ts
index 9962f6d431..75db92432c 100644
--- a/studio/frontend/src/features/auth/index.ts
+++ b/studio/frontend/src/features/auth/index.ts
@@ -1,7 +1,6 @@
 // SPDX-License-Identifier: AGPL-3.0-only
 // Copyright 2026-present the Unsloth AI Inc. team. All rights reserved. See /studio/LICENSE.AGPL-3.0
 
-export { ApiKeysPage } from "./api-keys-page";
 export { LoginPage } from "./login-page";
 export { ChangePasswordPage } from "./change-password-page";
 export { authFetch, refreshSession } from "./api";
diff --git a/studio/frontend/src/features/chat/chat-page.tsx b/studio/frontend/src/features/chat/chat-page.tsx
index c3153cdec7..32c3b6cd16 100644
--- a/studio/frontend/src/features/chat/chat-page.tsx
+++ b/studio/frontend/src/features/chat/chat-page.tsx
@@ -7,36 +7,16 @@ import {
   ModelSelector,
 } from "@/components/assistant-ui/model-selector";
 import { Thread } from "@/components/assistant-ui/thread";
-import { Button } from "@/components/ui/button";
-import {
-  Sheet,
-  SheetContent,
-  SheetDescription,
-  SheetHeader,
-  SheetTitle,
-} from "@/components/ui/sheet";
-import {
-  SidebarProvider,
-  SidebarTrigger,
-  useSidebar,
-} from "@/components/ui/sidebar";
-import {
-  Tooltip,
-  TooltipContent,
-  TooltipTrigger,
-} from "@/components/ui/tooltip";
-import { GuidedTour, useGuidedTourController } from "@/features/tour";
 import { cn } from "@/lib/utils";
+import { GuidedTour, useGuidedTourController } from "@/features/tour";
+import { useSidebar } from "@/components/ui/sidebar";
 import {
-  ColumnInsertIcon,
-  PencilEdit02Icon,
   Settings04Icon,
 } from "@hugeicons/core-free-icons";
 import { HugeiconsIcon } from "@hugeicons/react";
+import { useNavigate, useSearch } from "@tanstack/react-router";
 import {
-  type CSSProperties,
   type ReactElement,
-  type ReactNode,
   memo,
   useCallback,
   useEffect,
@@ -45,6 +25,7 @@ import {
   useState,
 } from "react";
 import { toast } from "sonner";
+import type { ChatSearch } from "@/app/routes/chat";
 import { listLocalModels } from "./api/chat-api";
 import { ChatSettingsPanel } from "./chat-settings-sheet";
 import { ContextUsageBar } from "./components/context-usage-bar";
@@ -63,7 +44,6 @@ import {
   SharedComposer,
 } from "./shared-composer";
 import { useChatRuntimeStore } from "./stores/chat-runtime-store";
-import { ThreadSidebar } from "./thread-sidebar";
 import { buildChatTourSteps } from "./tour";
 import type { ChatView, MessageRecord } from "./types";
 
@@ -135,7 +115,7 @@ const SingleContent = memo(function SingleContent({
       initialThreadId={threadId}
       newThreadNonce={newThreadNonce}
     >
-      <div className="min-h-0 flex-1">
+      <div className="flex min-h-0 min-w-0 flex-1 basis-0 flex-col overflow-hidden">
         <Thread />
       </div>
     </ChatRuntimeProvider>
@@ -223,7 +203,7 @@ const LoraCompareContent = memo(function LoraCompareContent({
                 Base Model
               </span>
             </div>
-            <div className="min-h-0 flex-1">
+            <div className="flex min-h-0 flex-1 flex-col">
               <ChatRuntimeProvider
                 modelType="base"
                 pairId={pairId}
@@ -231,7 +211,9 @@ const LoraCompareContent = memo(function LoraCompareContent({
                 syncActiveThreadId={false}
               >
                 <RegisterCompareHandle name="base" />
-                <Thread hideComposer={true} hideWelcome={true} />
+                <div className="flex min-h-0 min-w-0 flex-1 basis-0 flex-col overflow-hidden">
+                  <Thread hideComposer={true} hideWelcome={true} />
+                </div>
               </ChatRuntimeProvider>
             </div>
           </div>
@@ -241,7 +223,7 @@ const LoraCompareContent = memo(function LoraCompareContent({
                 Fine-tuned
               </span>
             </div>
-            <div className="min-h-0 flex-1">
+            <div className="flex min-h-0 flex-1 flex-col">
               <ChatRuntimeProvider
                 modelType="lora"
                 pairId={pairId}
@@ -249,12 +231,14 @@ const LoraCompareContent = memo(function LoraCompareContent({
                 syncActiveThreadId={false}
               >
                 <RegisterCompareHandle name="lora" />
-                <Thread hideComposer={true} hideWelcome={true} />
+                <div className="flex min-h-0 min-w-0 flex-1 basis-0 flex-col overflow-hidden">
+                  <Thread hideComposer={true} hideWelcome={true} />
+                </div>
               </ChatRuntimeProvider>
             </div>
           </div>
         </div>
-        <div className="z-20 mx-auto w-full max-w-4xl shrink-0 border-t border-border/60 bg-background px-4 pt-2 pb-4">
+        <div className="z-20 mx-auto w-full max-w-4xl shrink-0 bg-background px-4 pt-2 pb-4">
           <SharedComposer handlesRef={handlesRef} />
         </div>
       </div>
@@ -322,10 +306,7 @@ const GeneralCompareContent = memo(function GeneralCompareContent({
           className="grid min-h-0 flex-1 grid-cols-1 px-0 md:grid-cols-2"
         >
           <div className="flex min-h-0 flex-col">
-            <div className="flex items-center gap-2 px-3 py-1.5">
-              <span className="text-[10px] font-semibold uppercase tracking-wider text-muted-foreground">
-                Model 1
-              </span>
+            <div className="flex h-11 shrink-0 items-center gap-2 px-3">
               <ModelSelector
                 models={models}
                 loraModels={loraModels}
@@ -340,10 +321,10 @@ const GeneralCompareContent = memo(function GeneralCompareContent({
                 onFoldersChange={onFoldersChange}
                 variant="ghost"
                 size="sm"
-                className="max-w-[50%]"
+                className="max-w-[80%]"
               />
             </div>
-            <div className="min-h-0 flex-1">
+            <div className="flex min-h-0 flex-1 flex-col">
               <ChatRuntimeProvider
                 modelType="model1"
                 pairId={pairId}
@@ -351,15 +332,14 @@ const GeneralCompareContent = memo(function GeneralCompareContent({
                 syncActiveThreadId={false}
               >
                 <RegisterCompareHandle name="model1" />
-                <Thread hideComposer={true} hideWelcome={true} />
+                <div className="flex min-h-0 min-w-0 flex-1 basis-0 flex-col overflow-hidden">
+                  <Thread hideComposer={true} hideWelcome={true} />
+                </div>
               </ChatRuntimeProvider>
             </div>
           </div>
-          <div className="flex min-h-0 flex-col border-t border-border/60 md:border-t-0 md:border-l">
-            <div className="flex items-center gap-2 px-3 py-1.5 md:justify-end">
-              <span className="text-[10px] font-semibold uppercase tracking-wider text-primary">
-                Model 2
-              </span>
+          <div className="flex min-h-0 flex-col border-t border-sidebar-border md:border-t-0 md:border-l">
+            <div className="flex h-11 shrink-0 items-center gap-2 px-3">
               <ModelSelector
                 models={models}
                 loraModels={loraModels}
@@ -374,10 +354,10 @@ const GeneralCompareContent = memo(function GeneralCompareContent({
                 onFoldersChange={onFoldersChange}
                 variant="ghost"
                 size="sm"
-                className="max-w-[50%]"
+                className="max-w-[80%]"
               />
             </div>
-            <div className="min-h-0 flex-1">
+            <div className="flex min-h-0 flex-1 flex-col">
               <ChatRuntimeProvider
                 modelType="model2"
                 pairId={pairId}
@@ -385,12 +365,14 @@ const GeneralCompareContent = memo(function GeneralCompareContent({
                 syncActiveThreadId={false}
               >
                 <RegisterCompareHandle name="model2" />
-                <Thread hideComposer={true} hideWelcome={true} />
+                <div className="flex min-h-0 min-w-0 flex-1 basis-0 flex-col overflow-hidden">
+                  <Thread hideComposer={true} hideWelcome={true} />
+                </div>
               </ChatRuntimeProvider>
             </div>
           </div>
         </div>
-        <div className="z-20 mx-auto w-full max-w-4xl shrink-0 border-t border-border/60 bg-background px-4 pt-2 pb-4">
+        <div className="z-20 mx-auto w-full max-w-4xl shrink-0 bg-background px-4 pt-2 pb-4">
           <SharedComposer
             handlesRef={handlesRef}
             model1={model1}
@@ -402,110 +384,19 @@ const GeneralCompareContent = memo(function GeneralCompareContent({
   );
 });
 
-function InlineSidebar({
-  children,
-  side = "left",
-}: {
-  children: ReactNode;
-  side?: "left" | "right";
-}) {
-  const { state, isMobile, openMobile, setOpenMobile } = useSidebar();
-  const collapsed = state === "collapsed";
-
-  if (isMobile) {
-    return (
-      <Sheet open={openMobile} onOpenChange={setOpenMobile}>
-        <SheetContent side={side} className="w-[18rem] p-0">
-          <SheetHeader className="sr-only">
-            <SheetTitle>Chat sidebar</SheetTitle>
-            <SheetDescription>Chat threads and actions</SheetDescription>
-          </SheetHeader>
-          <div className="h-full overflow-auto">{children}</div>
-        </SheetContent>
-      </Sheet>
-    );
-  }
-
-  return (
-    <div
-      className="group shrink-0 h-full pb-3.5"
-      data-state={state}
-      data-collapsible={collapsed ? "offcanvas" : ""}
-      data-side={side}
-    >
-      <aside
-        data-sidebar="sidebar"
-        className={cn(
-          "bg-muted/70 text-sidebar-foreground h-full overflow-hidden rounded-2xl corner-squircle transition-[width] duration-200 ease-linear",
-          !collapsed && side === "right" && "border-l border-sidebar-border/70",
-          collapsed ? "w-0" : "w-(--sidebar-width)",
-        )}
-      >
-        <div className="flex h-full w-(--sidebar-width) flex-col">
-          {children}
-        </div>
-      </aside>
-    </div>
-  );
-}
-
-function TopBarActions({
-  onNewThread,
-  onNewCompare,
-  showCompare,
-}: {
-  onNewThread: () => void;
-  onNewCompare: () => void;
-  showCompare: boolean;
-}) {
-  const { state } = useSidebar();
-  if (state !== "collapsed") {
-    return null;
-  }
-  return (
-    <>
-      <Tooltip>
-        <TooltipTrigger asChild={true}>
-          <Button variant="ghost" size="icon-sm" onClick={onNewThread}>
-            <HugeiconsIcon icon={PencilEdit02Icon} strokeWidth={2} />
-          </Button>
-        </TooltipTrigger>
-        <TooltipContent side="bottom">New Chat</TooltipContent>
-      </Tooltip>
-      {showCompare ? (
-        <Tooltip>
-          <TooltipTrigger asChild={true}>
-            <Button variant="ghost" size="icon-sm" onClick={onNewCompare}>
-              <HugeiconsIcon icon={ColumnInsertIcon} strokeWidth={2} />
-            </Button>
-          </TooltipTrigger>
-          <TooltipContent side="bottom">Compare</TooltipContent>
-        </Tooltip>
-      ) : null}
-    </>
-  );
-}
-
-function getInitialSingleChatView(): ChatView {
-  const id = useChatRuntimeStore.getState().activeThreadId;
-  if (typeof id === "string" && id.length > 0 && !id.startsWith("__LOCALID_")) {
-    return { mode: "single", threadId: id };
-  }
-  return { mode: "single" };
-}
-
 export function ChatPage(): ReactElement {
-  // Do not set newThreadNonce here: each /chat mount would run ThreadNewChatSwitch
-  // and create spurious threads when navigating (e.g. Recipes / Export). New Chat
-  // explicitly sets a nonce in handleNewThread.
-  const [view, setView] = useState<ChatView>(getInitialSingleChatView);
-  const [settingsOpen, setSettingsOpen] = useState(false);
+  const search = useSearch({ from: "/chat" });
+  const navigate = useNavigate();
+
+  const settingsOpen = useChatRuntimeStore((s) => s.settingsPanelOpen);
+  const setSettingsOpen = useChatRuntimeStore((s) => s.setSettingsPanelOpen);
+
+  useEffect(() => {
+    return () => setSettingsOpen(false);
+  }, [setSettingsOpen]);
   const [modelSelectorOpen, setModelSelectorOpen] = useState(false);
   const [modelSelectorLocked, setModelSelectorLocked] = useState(false);
-  const [sidebarOpen, setSidebarOpen] = useState(true);
-  const [viewBeforeCompare, setViewBeforeCompare] = useState<ChatView | null>(
-    null,
-  );
+  const viewBeforeCompareRef = useRef<ChatSearch | null>(null);
   const inferenceParams = useChatRuntimeStore((state) => state.params);
   const setInferenceParams = useChatRuntimeStore((state) => state.setParams);
   const activeGgufVariant = useChatRuntimeStore(
@@ -515,8 +406,6 @@ export function ChatPage(): ReactElement {
     (state) => state.ggufContextLength,
   );
   const contextUsage = useChatRuntimeStore((state) => state.contextUsage);
-  const autoTitle = useChatRuntimeStore((state) => state.autoTitle);
-  const setAutoTitle = useChatRuntimeStore((state) => state.setAutoTitle);
   const modelsFromStore = useChatRuntimeStore((state) => state.models);
   const lorasFromStore = useChatRuntimeStore((state) => state.loras);
   const modelsError = useChatRuntimeStore((state) => state.modelsError);
@@ -541,6 +430,27 @@ export function ChatPage(): ReactElement {
     return Boolean(inferenceParams.checkpoint);
   }, [inferenceParams.checkpoint]);
 
+  // Derive view from URL search params
+  const view = useMemo<ChatView>(() => {
+    if (search.compare) {
+      return {
+        mode: "compare",
+        pairId:
+          search.compare,
+      };
+    }
+    if (search.thread) {
+      return { mode: "single", threadId: search.thread };
+    }
+    if (activeThreadId && !activeThreadId.startsWith("__LOCALID_")) {
+      return { mode: "single", threadId: activeThreadId };
+    }
+    if (search.new) {
+      return { mode: "single", newThreadNonce: search.new };
+    }
+    return { mode: "single" };
+  }, [search.thread, search.compare, search.new, activeThreadId]);
+
   const handleCheckpointChange = useCallback(
     (
       value: string,
@@ -601,28 +511,6 @@ export function ChatPage(): ReactElement {
   const handleEject = useCallback(() => {
     void ejectModel();
   }, [ejectModel]);
-  const handleNewThread = useCallback(() => {
-    // Skip if we are already on a fresh unsaved draft with no messages sent.
-    // Once the user sends a message, append() sets activeThreadId in the store,
-    // so we check the store to know whether the current draft has been sent.
-    if (
-      view.mode === "single" &&
-      !view.threadId &&
-      !useChatRuntimeStore.getState().activeThreadId
-    ) {
-      return;
-    }
-
-    useChatRuntimeStore.getState().setActiveThreadId(null);
-    setView({ mode: "single", newThreadNonce: crypto.randomUUID() });
-  }, [view]);
-  const handleNewCompare = useCallback(() => {
-    setView({ mode: "compare", pairId: crypto.randomUUID() });
-    // Clear activeThreadId so compare panes do not inherit the single-chat
-    // thread ID as a fallback for session_id routing.
-    useChatRuntimeStore.getState().setActiveThreadId(null);
-    useChatRuntimeStore.getState().setContextUsage(null);
-  }, []);
 
   const openModelSelector = useCallback(() => {
     setModelSelectorLocked(true);
@@ -641,30 +529,26 @@ export function ChatPage(): ReactElement {
     },
     [modelSelectorLocked],
   );
-  const openSettings = useCallback(() => setSettingsOpen(true), []);
-  const closeSettings = useCallback(() => setSettingsOpen(false), []);
-  const openSidebar = useCallback(() => setSidebarOpen(true), []);
+  const openSettings = useCallback(() => setSettingsOpen(true), [setSettingsOpen]);
+  const closeSettings = useCallback(() => setSettingsOpen(false), [setSettingsOpen]);
+  const { setPinned, isMobile } = useSidebar();
+  const openSidebar = useCallback(() => setPinned(true), [setPinned]);
 
   const enterCompare = useCallback(() => {
-    setViewBeforeCompare((prev) => prev ?? view);
-    setView({ mode: "compare", pairId: crypto.randomUUID() });
-    // Clear activeThreadId so compare panes do not inherit the single-chat
-    // thread ID as a fallback for session_id routing.
+    viewBeforeCompareRef.current = { ...search };
     useChatRuntimeStore.getState().setActiveThreadId(null);
     useChatRuntimeStore.getState().setContextUsage(null);
-  }, [view]);
+    navigate({ to: "/chat", search: { compare: crypto.randomUUID() } });
+  }, [navigate, search]);
 
   const exitCompare = useCallback(() => {
-    if (!viewBeforeCompare) return;
-    setView(viewBeforeCompare);
-    setViewBeforeCompare(null);
+    const saved = viewBeforeCompareRef.current;
+    if (!saved) return;
+    viewBeforeCompareRef.current = null;
+    navigate({ to: "/chat", search: saved });
     // Restore context usage from the active thread's last assistant message.
-    // Use the thread ID from the saved view rather than the store, because
-    // activeThreadId may have been cleared on compare entry.
-    const store = useChatRuntimeStore.getState();
     const threadId =
-      ("threadId" in viewBeforeCompare ? viewBeforeCompare.threadId : null) ??
-      store.activeThreadId;
+      saved.thread ?? useChatRuntimeStore.getState().activeThreadId;
     if (threadId) {
       void db.messages
         .where("threadId")
@@ -672,18 +556,12 @@ export function ChatPage(): ReactElement {
         .reverse()
         .first()
         .then((msg) => {
-          const saved = msg?.metadata as Record<string, unknown> | undefined;
-          const usage = saved?.contextUsage as
-            | typeof store.contextUsage
-            | undefined;
-          if (usage) store.setContextUsage(usage);
+          const metadata = msg?.metadata as Record<string, unknown> | undefined;
+          const usage = metadata?.contextUsage as ReturnType<typeof useChatRuntimeStore.getState>["contextUsage"];
+          if (usage) useChatRuntimeStore.getState().setContextUsage(usage);
         });
     }
-  }, [viewBeforeCompare]);
-
-  const handleThreadSelect = useCallback((nextView: ChatView) => {
-    setView(nextView);
-  }, []);
+  }, [navigate]);
 
   const models = useMemo<ModelOption[]>(
     () =>
@@ -727,7 +605,7 @@ export function ChatPage(): ReactElement {
         );
       })
       .catch(() => {});
-  }, []);
+  }, [navigate]);
 
   const loraModels = useMemo<LoraModelOption[]>(() => {
     const fromLoras = lorasFromStore.map((lora) => ({
@@ -771,9 +649,9 @@ export function ChatPage(): ReactElement {
           });
           await selectModelRef.current({ id: targetLora.id, isLora: true });
           if (canceled) return;
-          setView({ mode: "compare", pairId: crypto.randomUUID() });
           useChatRuntimeStore.getState().setActiveThreadId(null);
           useChatRuntimeStore.getState().setContextUsage(null);
+          navigate({ to: "/chat", search: { compare: crypto.randomUUID() } });
           clearHandoff();
           console.info("[chat-handoff] loaded lora + opened compare");
           return;
@@ -851,39 +729,18 @@ export function ChatPage(): ReactElement {
   }, [modelSelectorLocked, tour.open]);
 
   return (
-    <div className="h-[calc(100dvh-4rem)] bg-background overflow-hidden">
+    <div className="flex min-h-0 min-w-0 flex-1 basis-0 bg-background overflow-hidden">
       <GuidedTour {...tour.tourProps} />
-      <SidebarProvider
-        defaultOpen={true}
-        open={sidebarOpen}
-        onOpenChange={setSidebarOpen}
-        className="!min-h-0 h-full w-full max-w-7xl mx-auto px-2 sm:px-4"
-        style={
-          {
-            "--sidebar-width": "14rem",
-            "--sidebar-width-icon": "3rem",
-          } as CSSProperties
-        }
-      >
-        <InlineSidebar>
-          <ThreadSidebar
-            view={view}
-            onSelect={handleThreadSelect}
-            onNewThread={handleNewThread}
-            onNewCompare={handleNewCompare}
-            showCompare={canCompare}
-          />
-        </InlineSidebar>
-
-        <div className="flex min-h-0 min-w-0 flex-1 flex-col">
-          <div className="flex h-11 shrink-0 items-center px-1.5 sm:px-2">
-            <div className="flex items-center gap-1">
-              <SidebarTrigger />
-              <TopBarActions
-                onNewThread={handleNewThread}
-                onNewCompare={handleNewCompare}
-                showCompare={canCompare}
-              />
+      <div className="relative flex min-h-0 min-w-0 flex-1 basis-0 flex-col overflow-hidden">
+        <div
+          className={cn(
+            "absolute top-0 left-0 right-2 z-30 flex h-11 shrink-0 items-center pr-2 bg-background",
+            isMobile ? "pl-12 pr-1.5" : "pl-2",
+            view.mode === "compare" && "right-2 left-auto w-auto bg-transparent pl-0 pr-2",
+          )}
+        >
+          <div className="flex items-center gap-1">
+            {view.mode !== "compare" && (
               <ModelSelector
                 models={models}
                 loraModels={loraModels}
@@ -899,92 +756,90 @@ export function ChatPage(): ReactElement {
                 contentDataTour="chat-model-selector-popover"
                 className="max-w-[62vw] sm:max-w-none"
               />
-              {loadingModel && loadToastDismissed ? (
-                <ModelLoadInlineStatus
-                  label={
-                    loadProgress?.phase === "starting"
-                      ? "Starting model…"
-                      : loadingModel.isDownloaded || loadingModel.isCachedLora
-                        ? "Loading model…"
-                        : "Downloading model…"
-                  }
-                  title={
-                    loadingModel.isDownloaded
-                      ? `Loading ${loadingModel.displayName} from cache.`
-                      : loadingModel.isCachedLora
-                        ? `Loading ${loadingModel.displayName} into memory.`
-                        : `Loading ${loadingModel.displayName}. This may include downloading.`
-                  }
-                  progressPercent={loadProgress?.percent}
-                  progressLabel={loadProgress?.label}
-                  onStop={cancelLoading}
-                />
-              ) : null}
-            </div>
-            {modelsError && (
-              <div className="ml-2 text-xs text-destructive truncate max-w-[28rem]">
-                {modelsError}
-              </div>
             )}
-            <div className="flex-1" />
-            {view.mode === "single" && ggufContextLength && contextUsage ? (
-              <ContextUsageBar
-                used={contextUsage.totalTokens}
-                total={ggufContextLength}
-                cached={contextUsage.cachedTokens}
-                promptTokens={contextUsage.promptTokens}
-                completionTokens={contextUsage.completionTokens}
+            {loadingModel && loadToastDismissed ? (
+              <ModelLoadInlineStatus
+                label={
+                  loadProgress?.phase === "starting"
+                    ? "Starting model…"
+                    : loadingModel.isDownloaded || loadingModel.isCachedLora
+                      ? "Loading model…"
+                      : "Downloading model…"
+                }
+                title={
+                  loadingModel.isDownloaded
+                    ? `Loading ${loadingModel.displayName} from cache.`
+                    : loadingModel.isCachedLora
+                      ? `Loading ${loadingModel.displayName} into memory.`
+                      : `Loading ${loadingModel.displayName}. This may include downloading.`
+                }
+                progressPercent={loadProgress?.percent}
+                progressLabel={loadProgress?.label}
+                onStop={cancelLoading}
               />
             ) : null}
-            <button
-              type="button"
-              onClick={() => setSettingsOpen((o) => !o)}
-              className="flex h-9 w-9 items-center justify-center rounded-md text-muted-foreground transition-colors hover:bg-accent hover:text-foreground"
-              title="Inference settings"
-              data-tour="chat-settings"
-            >
-              <HugeiconsIcon icon={Settings04Icon} className="size-5" />
-            </button>
           </div>
-
-          {view.mode === "single" ? (
-            <SingleContent
-              key={view.threadId ?? "single"}
-              threadId={view.threadId}
-              newThreadNonce={view.newThreadNonce}
-            />
-          ) : (
-            <CompareContent
-              key={view.pairId}
-              pairId={view.pairId}
-              models={models}
-              loraModels={loraModels}
-              onFoldersChange={refreshLocalModels}
-            />
+          {modelsError && (
+            <div className="ml-2 text-xs text-destructive truncate max-w-[28rem]">
+              {modelsError}
+            </div>
           )}
+          <div className="flex-1" />
+          {view.mode === "single" && ggufContextLength && contextUsage ? (
+            <ContextUsageBar
+              used={contextUsage.totalTokens}
+              total={ggufContextLength}
+              cached={contextUsage.cachedTokens}
+              promptTokens={contextUsage.promptTokens}
+              completionTokens={contextUsage.completionTokens}
+            />
+          ) : null}
+          <button
+            type="button"
+            onClick={() => setSettingsOpen(!settingsOpen)}
+            className="flex h-9 w-9 items-center justify-center rounded-md text-muted-foreground transition-colors hover:bg-accent hover:text-foreground"
+            title="Inference settings"
+            data-tour="chat-settings"
+          >
+            <HugeiconsIcon icon={Settings04Icon} className="size-5" />
+          </button>
         </div>
 
-        <ChatSettingsPanel
-          open={settingsOpen}
-          onOpenChange={setSettingsOpen}
-          params={inferenceParams}
-          onParamsChange={setInferenceParams}
-          autoTitle={autoTitle}
-          onAutoTitleChange={setAutoTitle}
-          onReloadModel={() => {
-            const state = useChatRuntimeStore.getState();
-            if (state.params.checkpoint) {
-              selectModel({
-                id: state.params.checkpoint,
-                ggufVariant: state.activeGgufVariant ?? undefined,
-                forceReload: true,
-                isDownloaded: true,
-                loadingDescription: "Reloading with updated chat template.",
-              });
-            }
-          }}
-        />
-      </SidebarProvider>
+        {view.mode === "single" ? (
+          <SingleContent
+            key={view.threadId ?? "single"}
+            threadId={view.threadId}
+            newThreadNonce={view.newThreadNonce}
+          />
+        ) : (
+          <CompareContent
+            key={view.pairId}
+            pairId={view.pairId}
+            models={models}
+            loraModels={loraModels}
+            onFoldersChange={refreshLocalModels}
+          />
+        )}
+      </div>
+
+      <ChatSettingsPanel
+        open={settingsOpen}
+        onOpenChange={setSettingsOpen}
+        params={inferenceParams}
+        onParamsChange={setInferenceParams}
+        onReloadModel={() => {
+          const state = useChatRuntimeStore.getState();
+          if (state.params.checkpoint) {
+            selectModel({
+              id: state.params.checkpoint,
+              ggufVariant: state.activeGgufVariant ?? undefined,
+              forceReload: true,
+              isDownloaded: true,
+              loadingDescription: "Reloading with updated chat template.",
+            });
+          }
+        }}
+      />
     </div>
   );
 }
diff --git a/studio/frontend/src/features/chat/chat-settings-sheet.tsx b/studio/frontend/src/features/chat/chat-settings-sheet.tsx
index bcfd585214..2e03591a8e 100644
--- a/studio/frontend/src/features/chat/chat-settings-sheet.tsx
+++ b/studio/frontend/src/features/chat/chat-settings-sheet.tsx
@@ -55,7 +55,6 @@ import {
   PencilEdit01Icon,
   Settings02Icon,
   SlidersHorizontalIcon,
-  UserSettings01Icon,
   Wrench01Icon,
 } from "@hugeicons/core-free-icons";
 import { HugeiconsIcon } from "@hugeicons/react";
@@ -476,8 +475,6 @@ interface ChatSettingsPanelProps {
   onOpenChange?: (open: boolean) => void;
   params: InferenceParams;
   onParamsChange: (params: InferenceParams) => void;
-  autoTitle: boolean;
-  onAutoTitleChange: (enabled: boolean) => void;
   onReloadModel?: () => void;
 }
 
@@ -486,8 +483,6 @@ export function ChatSettingsPanel({
   onOpenChange,
   params,
   onParamsChange,
-  autoTitle,
-  onAutoTitleChange,
   onReloadModel,
 }: ChatSettingsPanelProps) {
   const isMobile = useIsMobile();
@@ -743,7 +738,8 @@ export function ChatSettingsPanel({
 
   const settingsContent = (
     <>
-      <div className="flex items-center gap-2 px-4 py-3">
+      <div className="aui-thread-viewport relative h-full overflow-y-auto bg-muted/70">
+      <div className="sticky top-0 z-10 flex items-center gap-2 bg-muted/70 px-4 py-3 backdrop-blur">
         <HugeiconsIcon
           icon={PencilEdit01Icon}
           className="size-4 text-muted-foreground/70"
@@ -753,7 +749,7 @@ export function ChatSettingsPanel({
         </span>
       </div>
 
-      <div className="flex-1 overflow-y-auto px-1.5">
+      <div className="px-1.5">
         {/* mt-4 matches the Playground sidebar gap (SidebarHeader py-3 + SidebarGroup pt-1) */}
         <div className="mt-4 px-2 pb-3">
           <div className="space-y-1.5">
@@ -893,7 +889,7 @@ export function ChatSettingsPanel({
             value={params.systemPrompt}
             onChange={(e) => set("systemPrompt")(e.target.value)}
             placeholder="You are a helpful assistant..."
-            className="min-h-20 text-xs corner-squircle"
+            className="min-h-20 max-h-48 overflow-y-auto text-xs corner-squircle"
             rows={3}
           />
         </div>
@@ -1174,27 +1170,9 @@ export function ChatSettingsPanel({
           </div>
         </CollapsibleSection>
 
-        <CollapsibleSection
-          icon={UserSettings01Icon}
-          label="Preferences"
-          defaultOpen={true}
-        >
-          <div className="flex flex-col gap-3 py-1">
-            <div className="flex items-center justify-between gap-3">
-              <div className="min-w-0">
-                <div className="text-xs font-medium">Auto title</div>
-                <div className="text-[11px] text-muted-foreground">
-                  Generate short title after reply.
-                </div>
-              </div>
-              <Switch checked={autoTitle} onCheckedChange={onAutoTitleChange} />
-            </div>
-            <HfTokenField />
-          </div>
-        </CollapsibleSection>
-
         <ChatTemplateSection onReloadModel={onReloadModel} />
       </div>
+      </div>
       <Dialog
         open={systemPromptEditorOpen}
         onOpenChange={(nextOpen) => {
@@ -1224,7 +1202,7 @@ export function ChatSettingsPanel({
               value={systemPromptDraft}
               onChange={(event) => setSystemPromptDraft(event.target.value)}
               placeholder="You are a helpful assistant..."
-              className="min-h-[24rem] text-sm leading-6 corner-squircle"
+              className="min-h-[24rem] max-h-[50vh] overflow-y-auto text-sm leading-6 corner-squircle"
               rows={14}
             />
           </div>
@@ -1268,9 +1246,9 @@ export function ChatSettingsPanel({
 
   return (
     <aside
-      className={`shrink-0 self-start h-[calc(100%-0.875rem)] overflow-hidden bg-muted/70 rounded-2xl corner-squircle transition-[width] duration-200 ease-linear ${open ? "w-[17rem] border-l border-sidebar-border/70" : "w-0"}`}
+      className={`relative z-50 shrink-0 h-full overflow-hidden bg-muted/70 transition-[width] duration-200 ease-linear ${open ? "w-[17rem]" : "w-0"}`}
     >
-      <div className="flex h-full w-[17rem] flex-col">{settingsContent}</div>
+      <div className="h-full w-[17rem]">{settingsContent}</div>
     </aside>
   );
 }
@@ -1348,29 +1326,6 @@ function AutoHealToolCallsToggle() {
   );
 }
 
-function HfTokenField() {
-  const hfToken = useChatRuntimeStore((s) => s.hfToken);
-  const setHfToken = useChatRuntimeStore((s) => s.setHfToken);
-
-  return (
-    <div className="flex flex-col gap-1.5">
-      <div className="min-w-0">
-        <div className="text-xs font-medium">Hugging Face Token</div>
-        <div className="text-[11px] text-muted-foreground">
-          For downloading gated or private models.
-        </div>
-      </div>
-      <Input
-        type="password"
-        value={hfToken}
-        placeholder="hf_..."
-        className="h-7 text-xs font-mono"
-        onChange={(e) => setHfToken(e.target.value)}
-      />
-    </div>
-  );
-}
-
 function ChatTemplateSection({
   onReloadModel,
 }: {
@@ -1391,7 +1346,7 @@ function ChatTemplateSection({
         <Textarea
           value={displayValue}
           onChange={(e) => setOverride(e.target.value)}
-          className="min-h-32 font-mono text-[10px] leading-relaxed md:text-[10px] corner-squircle"
+          className="min-h-32 max-h-64 overflow-y-auto font-mono text-[10px] leading-relaxed md:text-[10px] corner-squircle"
           rows={6}
           spellCheck={false}
         />
diff --git a/studio/frontend/src/features/chat/components/chat-search-dialog.tsx b/studio/frontend/src/features/chat/components/chat-search-dialog.tsx
new file mode 100644
index 0000000000..eca3b4bd9f
--- /dev/null
+++ b/studio/frontend/src/features/chat/components/chat-search-dialog.tsx
@@ -0,0 +1,120 @@
+// SPDX-License-Identifier: AGPL-3.0-only
+// Copyright 2026-present the Unsloth AI Inc. team. All rights reserved. See /studio/LICENSE.AGPL-3.0
+
+import {
+  Command,
+  CommandDialog,
+  CommandEmpty,
+  CommandGroup,
+  CommandList,
+} from "@/components/ui/command";
+import { useTrainingRuntimeStore } from "@/features/training";
+import { Cancel01Icon, Message01Icon, SearchIcon } from "@hugeicons/core-free-icons";
+import { HugeiconsIcon } from "@hugeicons/react";
+import { useNavigate } from "@tanstack/react-router";
+import { Command as CommandPrimitive } from "cmdk";
+import { useEffect } from "react";
+import { useChatSearchIndex } from "../hooks/use-chat-search-index";
+import { useChatSearchStore } from "../stores/chat-search-store";
+
+function formatRelative(createdAt: number): string {
+  const diff = Date.now() - createdAt;
+  const day = 86_400_000;
+  if (diff < day) return "Today";
+  if (diff < 7 * day) return "Past week";
+  if (diff < 30 * day) return "Past month";
+  return "Older";
+}
+
+export function ChatSearchDialog() {
+  const isOpen = useChatSearchStore((s) => s.isOpen);
+  const setOpen = useChatSearchStore((s) => s.setOpen);
+  const close = useChatSearchStore((s) => s.close);
+  const navigate = useNavigate();
+  const { items, loading } = useChatSearchIndex(isOpen);
+
+  useEffect(() => {
+    const handler = (e: KeyboardEvent) => {
+      if (!(e.metaKey || e.ctrlKey) || e.key.toLowerCase() !== "k") return;
+      if (useTrainingRuntimeStore.getState().isTrainingRunning) return;
+      const el = document.activeElement as HTMLElement | null;
+      const tag = el?.tagName;
+      if (tag === "INPUT" || tag === "TEXTAREA" || el?.isContentEditable) return;
+      e.preventDefault();
+      useChatSearchStore.getState().open();
+    };
+    window.addEventListener("keydown", handler);
+    return () => window.removeEventListener("keydown", handler);
+  }, []);
+
+  return (
+    <CommandDialog
+      open={isOpen}
+      onOpenChange={setOpen}
+      className="shadow-border corner-squircle w-[635px] max-w-[calc(100%-2rem)] gap-0 p-0 sm:max-w-[635px]"
+      overlayClassName="bg-transparent"
+    >
+      <Command className="rounded-none p-0">
+        <div className="flex items-center gap-3 border-b border-border/40 px-4 py-3">
+          <HugeiconsIcon
+            icon={SearchIcon}
+            strokeWidth={2}
+            className="size-4 shrink-0 text-muted-foreground"
+          />
+          <CommandPrimitive.Input
+            placeholder="Search chats..."
+            className="flex-1 bg-transparent text-sm outline-none placeholder:text-muted-foreground"
+          />
+          <button
+            type="button"
+            onClick={close}
+            className="flex size-6 items-center justify-center rounded-md text-muted-foreground transition-colors hover:bg-muted hover:text-foreground"
+            aria-label="Close"
+          >
+            <HugeiconsIcon icon={Cancel01Icon} strokeWidth={2} className="size-4" />
+          </button>
+        </div>
+        <CommandList className="max-h-[420px] p-1">
+          <CommandEmpty className="py-6 text-center text-xs text-muted-foreground">
+            {loading
+              ? "Loading…"
+              : items.length === 0
+                ? "No chats yet."
+                : "No chats match."}
+          </CommandEmpty>
+          <CommandGroup className="p-0">
+            {items.map((item) => (
+              <CommandPrimitive.Item
+                key={item.id}
+                value={`${item.title} ${item.preview}`}
+                onSelect={() => {
+                  navigate({
+                    to: "/chat",
+                    search:
+                      item.type === "single"
+                        ? { thread: item.id }
+                        : { compare: item.id },
+                  });
+                  close();
+                }}
+                className="relative flex cursor-default select-none items-center gap-3 rounded-lg px-3 py-2.5 text-sm outline-hidden data-selected:bg-muted data-selected:text-foreground"
+              >
+                <HugeiconsIcon
+                  icon={Message01Icon}
+                  strokeWidth={2}
+                  className="size-4 shrink-0 text-muted-foreground"
+                />
+                <span className="min-w-0 flex-1 truncate text-[13px] font-medium">
+                  {item.title || "Untitled chat"}
+                </span>
+                <span className="shrink-0 text-[11px] text-muted-foreground">
+                  {formatRelative(item.createdAt)}
+                </span>
+              </CommandPrimitive.Item>
+            ))}
+          </CommandGroup>
+        </CommandList>
+      </Command>
+    </CommandDialog>
+  );
+}
diff --git a/studio/frontend/src/features/chat/hooks/use-chat-search-index.ts b/studio/frontend/src/features/chat/hooks/use-chat-search-index.ts
new file mode 100644
index 0000000000..cab91da44f
--- /dev/null
+++ b/studio/frontend/src/features/chat/hooks/use-chat-search-index.ts
@@ -0,0 +1,156 @@
+// SPDX-License-Identifier: AGPL-3.0-only
+// Copyright 2026-present the Unsloth AI Inc. team. All rights reserved. See /studio/LICENSE.AGPL-3.0
+
+import { useEffect, useState } from "react";
+import { db } from "../db";
+import type { MessageRecord, ThreadRecord } from "../types";
+
+export interface ChatSearchItem {
+  type: "single" | "compare";
+  id: string;
+  title: string;
+  preview: string;
+  createdAt: number;
+}
+
+const THREAD_LIMIT = 200;
+const PREVIEW_MAX = 120;
+
+function extractText(message: MessageRecord): string {
+  const content = message.content;
+  if (!Array.isArray(content)) return "";
+  const parts: string[] = [];
+  for (const part of content) {
+    if (!part || typeof part !== "object") continue;
+    const p = part as { type?: string; text?: unknown };
+    if ((p.type === "text" || p.type === "reasoning") && typeof p.text === "string") {
+      parts.push(p.text);
+    }
+  }
+  return parts.join(" ").replace(/\s+/g, " ").trim();
+}
+
+function truncate(text: string, max: number): string {
+  if (text.length <= max) return text;
+  return text.slice(0, max).trimEnd() + "…";
+}
+
+async function buildIndex(): Promise<ChatSearchItem[]> {
+  // Fetch all threads newest-first, filter archived in JS, then take top N.
+  // `archived` is a boolean which Dexie does not index reliably, so we filter
+  // after the sort instead of using `.where("archived")`.
+  const all = (await db.threads
+    .orderBy("createdAt")
+    .reverse()
+    .toArray()) as ThreadRecord[];
+  const active = all.filter((t) => !t.archived).slice(0, THREAD_LIMIT);
+
+  const itemThreadIds = new Map<
+    string,
+    { item: Omit<ChatSearchItem, "preview">; threadIds: string[] }
+  >();
+  const seenPairs = new Set<string>();
+
+  for (const t of active) {
+    if (t.pairId) {
+      if (seenPairs.has(t.pairId)) {
+        const existing = itemThreadIds.get(t.pairId);
+        if (existing) existing.threadIds.push(t.id);
+        continue;
+      }
+      seenPairs.add(t.pairId);
+      itemThreadIds.set(t.pairId, {
+        item: {
+          type: "compare",
+          id: t.pairId,
+          title: t.title,
+          createdAt: t.createdAt,
+        },
+        threadIds: [t.id],
+      });
+    } else {
+      itemThreadIds.set(t.id, {
+        item: {
+          type: "single",
+          id: t.id,
+          title: t.title,
+          createdAt: t.createdAt,
+        },
+        threadIds: [t.id],
+      });
+    }
+  }
+
+  // One query for all messages across all relevant threads, then group by
+  // threadId in memory. Avoids N sequential awaits.
+  const allThreadIds = Array.from(itemThreadIds.values()).flatMap(
+    (e) => e.threadIds,
+  );
+  const messages = (await db.messages
+    .where("threadId")
+    .anyOf(allThreadIds)
+    .toArray()) as MessageRecord[];
+
+  const byThreadId = new Map<string, MessageRecord[]>();
+  for (const m of messages) {
+    const arr = byThreadId.get(m.threadId);
+    if (arr) arr.push(m);
+    else byThreadId.set(m.threadId, [m]);
+  }
+
+  const results: ChatSearchItem[] = [];
+  for (const { item, threadIds } of itemThreadIds.values()) {
+    const merged: MessageRecord[] = [];
+    for (const tid of threadIds) {
+      const arr = byThreadId.get(tid);
+      if (arr) merged.push(...arr);
+    }
+    merged.sort((a, b) => b.createdAt - a.createdAt);
+
+    let preview = "";
+    for (const m of merged) {
+      const text = extractText(m);
+      if (text) {
+        preview = truncate(text, PREVIEW_MAX);
+        break;
+      }
+    }
+    results.push({ ...item, preview });
+  }
+
+  results.sort((a, b) => b.createdAt - a.createdAt);
+  return results;
+}
+
+export function useChatSearchIndex(enabled: boolean): {
+  items: ChatSearchItem[];
+  loading: boolean;
+} {
+  const [items, setItems] = useState<ChatSearchItem[]>([]);
+  const [loading, setLoading] = useState(false);
+
+  useEffect(() => {
+    if (!enabled) {
+      // Clear stale results so the next open doesn't flash old items.
+      setItems([]);
+      return;
+    }
+    let cancelled = false;
+    setLoading(true);
+    buildIndex()
+      .then((result) => {
+        if (!cancelled) setItems(result);
+      })
+      .catch(() => {
+        if (!cancelled) setItems([]);
+      })
+      .finally(() => {
+        if (!cancelled) setLoading(false);
+      });
+    return () => {
+      cancelled = true;
+    };
+  }, [enabled]);
+
+  return { items, loading };
+}
diff --git a/studio/frontend/src/features/chat/hooks/use-chat-sidebar-items.ts b/studio/frontend/src/features/chat/hooks/use-chat-sidebar-items.ts
new file mode 100644
index 0000000000..fb27f990cf
--- /dev/null
+++ b/studio/frontend/src/features/chat/hooks/use-chat-sidebar-items.ts
@@ -0,0 +1,82 @@
+// SPDX-License-Identifier: AGPL-3.0-only
+// Copyright 2026-present the Unsloth AI Inc. team. All rights reserved. See /studio/LICENSE.AGPL-3.0
+
+import { db, useLiveQuery } from "../db";
+import { useChatRuntimeStore } from "../stores/chat-runtime-store";
+import type { ThreadRecord } from "../types";
+
+export interface SidebarItem {
+  type: "single" | "compare";
+  id: string;
+  title: string;
+  createdAt: number;
+}
+
+export function groupThreads(threads: ThreadRecord[]): SidebarItem[] {
+  const items: SidebarItem[] = [];
+  const seenPairs = new Set<string>();
+
+  for (const t of threads) {
+    if (t.archived) {
+      continue;
+    }
+    if (t.pairId) {
+      if (seenPairs.has(t.pairId)) {
+        continue;
+      }
+      seenPairs.add(t.pairId);
+      items.push({
+        type: "compare",
+        id: t.pairId,
+        title: t.title,
+        createdAt: t.createdAt,
+      });
+    } else if (!t.pairId) {
+      items.push({
+        type: "single",
+        id: t.id,
+        title: t.title,
+        createdAt: t.createdAt,
+      });
+    }
+  }
+
+  return items.sort((a, b) => b.createdAt - a.createdAt);
+}
+
+export function useChatSidebarItems() {
+  const allThreads = useLiveQuery(async () => {
+    const threadIdsWithMessage = new Set(
+      (await db.messages.orderBy("threadId").uniqueKeys()) as string[],
+    );
+    const rows = await db.threads.orderBy("createdAt").reverse().toArray();
+    return rows.filter((t) => !t.archived && threadIdsWithMessage.has(t.id));
+  }, []);
+  const items = groupThreads(allThreads ?? []);
+  const canCompare = useChatRuntimeStore((s) => Boolean(s.params.checkpoint));
+
+  return { items, canCompare };
+}
+
+export async function deleteChatItem(
+  item: SidebarItem,
+  activeId: string | undefined,
+  onSelect: (view: { mode: "single"; newThreadNonce: string }) => void,
+) {
+  await db.transaction("rw", db.threads, db.messages, async () => {
+    if (item.type === "single") {
+      await db.messages.where("threadId").equals(item.id).delete();
+      await db.threads.delete(item.id);
+    } else {
+      const paired = await db.threads.where("pairId").equals(item.id).toArray();
+      for (const t of paired) {
+        await db.messages.where("threadId").equals(t.id).delete();
+        await db.threads.delete(t.id);
+      }
+    }
+  });
+  if (activeId === item.id) {
+    useChatRuntimeStore.getState().setActiveThreadId(null);
+    onSelect({ mode: "single", newThreadNonce: crypto.randomUUID() });
+  }
+}
diff --git a/studio/frontend/src/features/chat/shared-composer.tsx b/studio/frontend/src/features/chat/shared-composer.tsx
index d45004cd43..f82355526f 100644
--- a/studio/frontend/src/features/chat/shared-composer.tsx
+++ b/studio/frontend/src/features/chat/shared-composer.tsx
@@ -455,7 +455,7 @@ export function SharedComposer({
 
   return (
     <div
-      className={`shadow-border ring-1 ring-border relative flex w-full flex-col rounded-2xl bg-background px-1 pt-2 transition-shadow outline-none ${dragging ? "ring-ring bg-accent/50" : ""}`}
+      className={`chat-composer-surface relative flex w-full flex-col rounded-3xl bg-background px-1 pt-2 transition-shadow outline-none ${dragging ? "border-ring bg-accent/50" : ""}`}
       onDragOver={(e) => {
         e.preventDefault();
         setDragging(true);
diff --git a/studio/frontend/src/features/chat/stores/chat-runtime-store.ts b/studio/frontend/src/features/chat/stores/chat-runtime-store.ts
index a041452a48..9ff7ded18f 100644
--- a/studio/frontend/src/features/chat/stores/chat-runtime-store.ts
+++ b/studio/frontend/src/features/chat/stores/chat-runtime-store.ts
@@ -173,6 +173,7 @@ type ChatRuntimeStore = {
   defaultChatTemplate: string | null;
   chatTemplateOverride: string | null;
   activeThreadId: string | null;
+  settingsPanelOpen: boolean;
   pendingAudioBase64: string | null;
   pendingAudioName: string | null;
   contextUsage: {
@@ -193,6 +194,7 @@ type ChatRuntimeStore = {
   setModelsError: (error: string | null) => void;
   setCheckpoint: (modelId: string, ggufVariant?: string | null) => void;
   setActiveThreadId: (threadId: string | null) => void;
+  setSettingsPanelOpen: (open: boolean) => void;
   clearCheckpoint: () => void;
   setReasoningEnabled: (enabled: boolean) => void;
   setToolsEnabled: (enabled: boolean) => void;
@@ -243,6 +245,7 @@ export const useChatRuntimeStore = create<ChatRuntimeStore>((set) => ({
   defaultChatTemplate: null,
   chatTemplateOverride: null,
   activeThreadId: null,
+  settingsPanelOpen: false,
   pendingAudioBase64: null,
   pendingAudioName: null,
   contextUsage: null,
@@ -294,6 +297,7 @@ export const useChatRuntimeStore = create<ChatRuntimeStore>((set) => ({
       activeGgufVariant: ggufVariant ?? null,
     })),
   setActiveThreadId: (activeThreadId) => set({ activeThreadId, contextUsage: null }),
+  setSettingsPanelOpen: (settingsPanelOpen) => set({ settingsPanelOpen }),
   clearCheckpoint: () =>
     set((state) => ({
       params: {
diff --git a/studio/frontend/src/features/chat/stores/chat-search-store.ts b/studio/frontend/src/features/chat/stores/chat-search-store.ts
new file mode 100644
index 0000000000..3719d4f10f
--- /dev/null
+++ b/studio/frontend/src/features/chat/stores/chat-search-store.ts
@@ -0,0 +1,18 @@
+// SPDX-License-Identifier: AGPL-3.0-only
+// Copyright 2026-present the Unsloth AI Inc. team. All rights reserved. See /studio/LICENSE.AGPL-3.0
+
+import { create } from "zustand";
+
+interface ChatSearchStore {
+  isOpen: boolean;
+  open: () => void;
+  close: () => void;
+  setOpen: (open: boolean) => void;
+}
+
+export const useChatSearchStore = create<ChatSearchStore>((set) => ({
+  isOpen: false,
+  open: () => set({ isOpen: true }),
+  close: () => set({ isOpen: false }),
+  setOpen: (isOpen) => set({ isOpen }),
+}));
diff --git a/studio/frontend/src/features/chat/thread-sidebar.tsx b/studio/frontend/src/features/chat/thread-sidebar.tsx
index 62246cdad5..b15c16297e 100644
--- a/studio/frontend/src/features/chat/thread-sidebar.tsx
+++ b/studio/frontend/src/features/chat/thread-sidebar.tsx
@@ -21,48 +21,10 @@ import {
   PencilEdit02Icon,
 } from "@hugeicons/core-free-icons";
 import { HugeiconsIcon } from "@hugeicons/react";
-import { db, useLiveQuery } from "./db";
 import { useChatRuntimeStore } from "./stores/chat-runtime-store";
-import type { ChatView, ThreadRecord } from "./types";
-
-interface SidebarItem {
-  type: "single" | "compare";
-  id: string;
-  title: string;
-  createdAt: number;
-}
-
-function groupThreads(threads: ThreadRecord[]): SidebarItem[] {
-  const items: SidebarItem[] = [];
-  const seenPairs = new Set<string>();
-
-  for (const t of threads) {
-    if (t.archived) {
-      continue;
-    }
-    if (t.pairId) {
-      if (seenPairs.has(t.pairId)) {
-        continue;
-      }
-      seenPairs.add(t.pairId);
-      items.push({
-        type: "compare",
-        id: t.pairId,
-        title: t.title,
-        createdAt: t.createdAt,
-      });
-    } else if (!t.pairId) {
-      items.push({
-        type: "single",
-        id: t.id,
-        title: t.title,
-        createdAt: t.createdAt,
-      });
-    }
-  }
-
-  return items.sort((a, b) => b.createdAt - a.createdAt);
-}
+import type { ChatView } from "./types";
+import { deleteChatItem, useChatSidebarItems } from "./hooks/use-chat-sidebar-items";
+import type { SidebarItem } from "./hooks/use-chat-sidebar-items";
 
 export function ThreadSidebar({
   view,
@@ -77,14 +39,7 @@ export function ThreadSidebar({
   onNewCompare: () => void;
   showCompare: boolean;
 }) {
-  const allThreads = useLiveQuery(async () => {
-    const threadIdsWithMessage = new Set(
-      (await db.messages.orderBy("threadId").uniqueKeys()) as string[],
-    );
-    const rows = await db.threads.orderBy("createdAt").reverse().toArray();
-    return rows.filter((t) => !t.archived && threadIdsWithMessage.has(t.id));
-  }, []);
-  const items = groupThreads(allThreads ?? []);
+  const { items } = useChatSidebarItems();
   const storeThreadId = useChatRuntimeStore((s) => s.activeThreadId);
   const activeId =
     view.mode === "single" ? (view.threadId ?? storeThreadId) : view.pairId;
@@ -96,23 +51,10 @@ export function ThreadSidebar({
   }
 
   async function handleDelete(item: SidebarItem) {
-    if (item.type === "single") {
-      await db.messages.where("threadId").equals(item.id).delete();
-      await db.threads.delete(item.id);
-    } else {
-      const paired = await db.threads.where("pairId").equals(item.id).toArray();
-      for (const t of paired) {
-        await db.messages.where("threadId").equals(t.id).delete();
-        await db.threads.delete(t.id);
-      }
-    }
-    if (activeId === item.id) {
-      // Directly set a new view with a nonce rather than going through
-      // onNewThread(), which may return early if the guard sees no
-      // threadId and no activeThreadId (after we just cleared it).
-      useChatRuntimeStore.getState().setActiveThreadId(null);
-      onSelect({ mode: "single", newThreadNonce: crypto.randomUUID() });
-    }
+    // Directly set a new view with a nonce rather than going through
+    // onNewThread(), which may return early if the guard sees no
+    // threadId and no activeThreadId (after we just cleared it).
+    await deleteChatItem(item, activeId ?? undefined, onSelect);
   }
 
   return (
diff --git a/studio/frontend/src/features/chat/utils/clear-all-chats.ts b/studio/frontend/src/features/chat/utils/clear-all-chats.ts
new file mode 100644
index 0000000000..eff87eb95f
--- /dev/null
+++ b/studio/frontend/src/features/chat/utils/clear-all-chats.ts
@@ -0,0 +1,15 @@
+// SPDX-License-Identifier: AGPL-3.0-only
+// Copyright 2026-present the Unsloth AI Inc. team. All rights reserved. See /studio/LICENSE.AGPL-3.0
+
+import { db } from "../db";
+
+export async function countAllChats(): Promise<number> {
+  return db.threads.count();
+}
+
+export async function clearAllChats(): Promise<void> {
+  await db.transaction("rw", db.threads, db.messages, async () => {
+    await db.messages.clear();
+    await db.threads.clear();
+  });
+}
diff --git a/studio/frontend/src/features/chat/utils/export-chat-history.ts b/studio/frontend/src/features/chat/utils/export-chat-history.ts
new file mode 100644
index 0000000000..9c8b8fccc8
--- /dev/null
+++ b/studio/frontend/src/features/chat/utils/export-chat-history.ts
@@ -0,0 +1,41 @@
+// SPDX-License-Identifier: AGPL-3.0-only
+// Copyright 2026-present the Unsloth AI Inc. team. All rights reserved. See /studio/LICENSE.AGPL-3.0
+
+import { db } from "../db";
+
+interface ExportedChat {
+  exportedAt: string;
+  version: 1;
+  threadCount: number;
+  threads: unknown[];
+  messages: unknown[];
+}
+
+export async function buildChatExport(): Promise<ExportedChat> {
+  const [threads, messages] = await Promise.all([
+    db.threads.toArray(),
+    db.messages.toArray(),
+  ]);
+  return {
+    exportedAt: new Date().toISOString(),
+    version: 1,
+    threadCount: threads.length,
+    threads,
+    messages,
+  };
+}
+
+export async function downloadChatExport(): Promise<void> {
+  const data = await buildChatExport();
+  const blob = new Blob([JSON.stringify(data, null, 2)], {
+    type: "application/json",
+  });
+  const url = URL.createObjectURL(blob);
+  const a = document.createElement("a");
+  a.href = url;
+  a.download = `unsloth-chats-${new Date().toISOString().slice(0, 10)}.json`;
+  document.body.appendChild(a);
+  a.click();
+  a.remove();
+  URL.revokeObjectURL(url);
+}
diff --git a/studio/frontend/src/features/data-recipes/hooks/use-recipe-sidebar-items.ts b/studio/frontend/src/features/data-recipes/hooks/use-recipe-sidebar-items.ts
new file mode 100644
index 0000000000..7cf70dc8eb
--- /dev/null
+++ b/studio/frontend/src/features/data-recipes/hooks/use-recipe-sidebar-items.ts
@@ -0,0 +1,22 @@
+// SPDX-License-Identifier: AGPL-3.0-only
+// Copyright 2026-present the Unsloth AI Inc. team. All rights reserved. See /studio/LICENSE.AGPL-3.0
+
+import { useEffect, useState } from "react";
+import { liveQuery } from "dexie";
+import { listRecipes } from "../data/recipes-db";
+import type { RecipeRecord } from "../types";
+
+export function useRecipeSidebarItems(enabled: boolean) {
+  const [recipes, setRecipes] = useState<RecipeRecord[]>([]);
+
+  useEffect(() => {
+    if (!enabled) return;
+    const sub = liveQuery(() => listRecipes()).subscribe({
+      next: (value) => setRecipes(value),
+      error: (err) => console.error("recipe sidebar liveQuery:", err),
+    });
+    return () => sub.unsubscribe();
+  }, [enabled]);
+
+  return recipes;
+}
diff --git a/studio/frontend/src/features/settings/api/api-keys.ts b/studio/frontend/src/features/settings/api/api-keys.ts
new file mode 100644
index 0000000000..73d7b50def
--- /dev/null
+++ b/studio/frontend/src/features/settings/api/api-keys.ts
@@ -0,0 +1,41 @@
+// SPDX-License-Identifier: AGPL-3.0-only
+// Copyright 2026-present the Unsloth AI Inc. team. All rights reserved. See /studio/LICENSE.AGPL-3.0
+
+import { authFetch } from "@/features/auth/api";
+
+export interface ApiKey {
+  id: number;
+  name: string;
+  key_prefix: string;
+  created_at: string;
+  last_used_at: string | null;
+  expires_at: string | null;
+  is_active: boolean;
+}
+
+export async function fetchApiKeys(): Promise<ApiKey[]> {
+  const res = await authFetch("/api/auth/api-keys");
+  if (!res.ok) throw new Error("Failed to load API keys");
+  const data = (await res.json()) as { api_keys: ApiKey[] };
+  return data.api_keys.filter((k) => k.is_active);
+}
+
+export async function createApiKey(
+  name: string,
+  expiresInDays: number | null,
+): Promise<{ key: string; api_key: ApiKey }> {
+  const res = await authFetch("/api/auth/api-keys", {
+    method: "POST",
+    headers: { "Content-Type": "application/json" },
+    body: JSON.stringify({ name, expires_in_days: expiresInDays }),
+  });
+  if (!res.ok) throw new Error("Failed to create API key");
+  return res.json();
+}
+
+export async function revokeApiKey(keyId: number): Promise<void> {
+  const res = await authFetch(`/api/auth/api-keys/${keyId}`, {
+    method: "DELETE",
+  });
+  if (!res.ok) throw new Error("Failed to revoke API key");
+}
diff --git a/studio/frontend/src/features/settings/components/api-key-row.tsx b/studio/frontend/src/features/settings/components/api-key-row.tsx
new file mode 100644
index 0000000000..74b4ba396e
--- /dev/null
+++ b/studio/frontend/src/features/settings/components/api-key-row.tsx
@@ -0,0 +1,101 @@
+// SPDX-License-Identifier: AGPL-3.0-only
+// Copyright 2026-present the Unsloth AI Inc. team. All rights reserved. See /studio/LICENSE.AGPL-3.0
+
+import { Button } from "@/components/ui/button";
+import {
+  DropdownMenu,
+  DropdownMenuContent,
+  DropdownMenuItem,
+  DropdownMenuTrigger,
+} from "@/components/ui/dropdown-menu";
+import {
+  Delete02Icon,
+  Copy01Icon,
+  MoreHorizontalIcon,
+} from "@hugeicons/core-free-icons";
+import { HugeiconsIcon } from "@hugeicons/react";
+import { copyToClipboard } from "@/lib/copy-to-clipboard";
+import type { ApiKey } from "../api/api-keys";
+
+function relative(iso: string | null): string {
+  if (!iso) return "never";
+  const diff = Date.now() - new Date(iso).getTime();
+  const days = Math.floor(diff / 86400000);
+  if (days < 1) {
+    const hours = Math.floor(diff / 3600000);
+    if (hours < 1) return "just now";
+    return `${hours}h ago`;
+  }
+  if (days < 30) return `${days}d ago`;
+  if (days < 365) return `${Math.floor(days / 30)}mo ago`;
+  return `${Math.floor(days / 365)}y ago`;
+}
+
+function expiresText(iso: string | null): string {
+  if (!iso) return "never";
+  const diff = new Date(iso).getTime() - Date.now();
+  if (diff < 0) return "expired";
+  const days = Math.floor(diff / 86400000);
+  if (days < 1) return "today";
+  return `in ${days}d`;
+}
+
+export function ApiKeyRow({
+  apiKey,
+  onRevoke,
+}: {
+  apiKey: ApiKey;
+  onRevoke: (key: ApiKey) => void;
+}) {
+  const prefix = `sk-unsloth-${apiKey.key_prefix}…`;
+  return (
+    <div className="group flex items-center gap-3 border-b border-border/60 px-1 py-3 last:border-b-0 transition-colors hover:bg-accent/40">
+      <span
+        className="size-1.5 shrink-0 rounded-full bg-emerald-500"
+        aria-hidden="true"
+      />
+      <div className="flex min-w-0 flex-1 flex-col gap-0.5">
+        <div className="flex min-w-0 items-center justify-between gap-3">
+          <span className="truncate text-sm font-medium text-foreground" title={apiKey.name}>
+            {apiKey.name}
+          </span>
+          <code className="shrink-0 font-mono text-[11px] text-muted-foreground">
+            {prefix}
+          </code>
+        </div>
+        <div className="flex flex-wrap gap-x-1.5 text-[11px] text-muted-foreground">
+          <span>Created {relative(apiKey.created_at)}</span>
+          <span>·</span>
+          <span>Used {relative(apiKey.last_used_at)}</span>
+          <span>·</span>
+          <span>Expires {expiresText(apiKey.expires_at)}</span>
+        </div>
+      </div>
+      <DropdownMenu>
+        <DropdownMenuTrigger asChild>
+          <Button
+            variant="ghost"
+            size="sm"
+            className="size-7 p-0 opacity-0 transition-opacity group-hover:opacity-100 data-[state=open]:opacity-100 max-sm:!opacity-100 max-sm:size-9"
+            aria-label={`Actions for ${apiKey.name}`}
+          >
+            <HugeiconsIcon icon={MoreHorizontalIcon} className="size-4" />
+          </Button>
+        </DropdownMenuTrigger>
+        <DropdownMenuContent align="end">
+          <DropdownMenuItem onClick={() => copyToClipboard(prefix)}>
+            <HugeiconsIcon icon={Copy01Icon} className="size-3.5 mr-2" />
+            Copy prefix
+          </DropdownMenuItem>
+          <DropdownMenuItem
+            onClick={() => onRevoke(apiKey)}
+            className="text-destructive focus:text-destructive"
+          >
+            <HugeiconsIcon icon={Delete02Icon} className="size-3.5 mr-2" />
+            Revoke key
+          </DropdownMenuItem>
+        </DropdownMenuContent>
+      </DropdownMenu>
+    </div>
+  );
+}
diff --git a/studio/frontend/src/features/settings/components/create-key-form.tsx b/studio/frontend/src/features/settings/components/create-key-form.tsx
new file mode 100644
index 0000000000..2166abf3cf
--- /dev/null
+++ b/studio/frontend/src/features/settings/components/create-key-form.tsx
@@ -0,0 +1,83 @@
+// SPDX-License-Identifier: AGPL-3.0-only
+// Copyright 2026-present the Unsloth AI Inc. team. All rights reserved. See /studio/LICENSE.AGPL-3.0
+
+import { Button } from "@/components/ui/button";
+import { Input } from "@/components/ui/input";
+import { cn } from "@/lib/utils";
+import { useState } from "react";
+import { createApiKey } from "../api/api-keys";
+
+const EXPIRY_PRESETS = [
+  { label: "Never", value: null as number | null },
+  { label: "7d", value: 7 },
+  { label: "30d", value: 30 },
+  { label: "90d", value: 90 },
+];
+
+export function CreateKeyForm({
+  onCreated,
+  onError,
+}: {
+  onCreated: (rawKey: string) => void;
+  onError: (message: string) => void;
+}) {
+  const [name, setName] = useState("");
+  const [expiry, setExpiry] = useState<number | null>(null);
+  const [loading, setLoading] = useState(false);
+
+  const handleSubmit = async (e: React.FormEvent) => {
+    e.preventDefault();
+    if (!name.trim() || loading) return;
+    setLoading(true);
+    try {
+      const result = await createApiKey(name.trim(), expiry);
+      onCreated(result.key);
+      setName("");
+    } catch (err) {
+      onError(err instanceof Error ? err.message : "Couldn't create key.");
+    } finally {
+      setLoading(false);
+    }
+  };
+
+  return (
+    <form
+      onSubmit={handleSubmit}
+      className="flex flex-col gap-2 rounded-lg border border-border bg-muted/20 p-3"
+    >
+      <div className="flex flex-wrap items-center gap-2">
+        <Input
+          value={name}
+          onChange={(e) => setName(e.target.value)}
+          placeholder="Key name (e.g. production)"
+          className="h-8 min-w-[180px] flex-1 text-sm"
+          aria-label="New key name"
+        />
+        <div className="inline-flex items-center rounded-md border border-border bg-background p-0.5">
+          {EXPIRY_PRESETS.map((p) => {
+            const active = expiry === p.value;
+            return (
+              <button
+                key={p.label}
+                type="button"
+                onClick={() => setExpiry(p.value)}
+                aria-pressed={active}
+                className={cn(
+                  "rounded px-2 py-1 text-[11px] font-medium transition-colors focus-visible:outline-none focus-visible:ring-2 focus-visible:ring-ring",
+                  active
+                    ? "bg-accent text-foreground"
+                    : "text-muted-foreground hover:text-foreground",
+                )}
+              >
+                {p.label}
+              </button>
+            );
+          })}
+        </div>
+        <Button type="submit" size="sm" disabled={loading || !name.trim()}>
+          {loading ? "Creating…" : "Create key"}
+        </Button>
+      </div>
+    </form>
+  );
+}
diff --git a/studio/frontend/src/features/settings/components/key-reveal-card.tsx b/studio/frontend/src/features/settings/components/key-reveal-card.tsx
new file mode 100644
index 0000000000..667351570c
--- /dev/null
+++ b/studio/frontend/src/features/settings/components/key-reveal-card.tsx
@@ -0,0 +1,71 @@
+// SPDX-License-Identifier: AGPL-3.0-only
+// Copyright 2026-present the Unsloth AI Inc. team. All rights reserved. See /studio/LICENSE.AGPL-3.0
+
+import { Button } from "@/components/ui/button";
+import { copyToClipboard } from "@/lib/copy-to-clipboard";
+import { cn } from "@/lib/utils";
+import { Copy01Icon, Tick02Icon } from "@hugeicons/core-free-icons";
+import { HugeiconsIcon } from "@hugeicons/react";
+import { useState } from "react";
+
+export function KeyRevealCard({
+  rawKey,
+  onDone,
+}: {
+  rawKey: string;
+  onDone: () => void;
+}) {
+  const [copied, setCopied] = useState(false);
+
+  const handleCopy = () => {
+    if (copyToClipboard(rawKey)) {
+      setCopied(true);
+      setTimeout(() => setCopied(false), 1800);
+    }
+  };
+
+  return (
+    <div className="flex flex-col gap-2 rounded-lg border border-emerald-500/30 bg-emerald-500/5 p-3">
+      <div className="flex items-center gap-1.5">
+        <HugeiconsIcon
+          icon={Tick02Icon}
+          className="size-3.5 text-emerald-600 dark:text-emerald-500"
+        />
+        <span className="text-xs font-medium text-emerald-700 dark:text-emerald-500">
+          New key created
+        </span>
+      </div>
+      <button
+        type="button"
+        onClick={handleCopy}
+        className={cn(
+          "flex w-full items-center justify-between gap-3 rounded-md border border-border bg-muted/40 px-3 py-2.5 font-mono text-sm transition-colors hover:bg-muted/60",
+          "focus-visible:outline-none focus-visible:ring-2 focus-visible:ring-ring focus-visible:ring-offset-1 focus-visible:ring-offset-background",
+          copied && "border-emerald-500/40 bg-emerald-500/10",
+        )}
+        aria-label={copied ? "Key copied" : "Copy key"}
+      >
+        <code className="min-w-0 flex-1 break-all text-left text-foreground">
+          {rawKey}
+        </code>
+        <HugeiconsIcon
+          icon={copied ? Tick02Icon : Copy01Icon}
+          className={cn("size-4 shrink-0", copied && "text-emerald-600")}
+        />
+      </button>
+      <div className="flex items-center justify-between gap-3 pt-0.5">
+        <p className="text-[11px] text-muted-foreground">
+          Copy now — this won't be shown again.
+        </p>
+        <Button
+          type="button"
+          size="sm"
+          onClick={onDone}
+          className="focus-visible:outline-none focus-visible:ring-2 focus-visible:ring-ring focus-visible:ring-offset-1 focus-visible:ring-offset-background"
+        >
+          Done
+        </Button>
+      </div>
+    </div>
+  );
+}
diff --git a/studio/frontend/src/features/settings/components/settings-row.tsx b/studio/frontend/src/features/settings/components/settings-row.tsx
new file mode 100644
index 0000000000..a08b7beac8
--- /dev/null
+++ b/studio/frontend/src/features/settings/components/settings-row.tsx
@@ -0,0 +1,39 @@
+// SPDX-License-Identifier: AGPL-3.0-only
+// Copyright 2026-present the Unsloth AI Inc. team. All rights reserved. See /studio/LICENSE.AGPL-3.0
+
+import { cn } from "@/lib/utils";
+import type { ReactNode } from "react";
+
+export function SettingsRow({
+  label,
+  description,
+  children,
+  destructive,
+  className,
+}: {
+  label: string;
+  description?: string;
+  children?: ReactNode;
+  destructive?: boolean;
+  className?: string;
+}) {
+  return (
+    <div
+      className={cn(
+        "flex items-center justify-between gap-6 py-3",
+        destructive && "border-t border-border/60 mt-2 pt-4",
+        className,
+      )}
+    >
+      <div className="flex min-w-0 flex-col gap-0.5">
+        <span className="text-sm font-medium text-foreground">{label}</span>
+        {description ? (
+          <span className="text-xs text-muted-foreground leading-snug">
+            {description}
+          </span>
+        ) : null}
+      </div>
+      {children ? <div className="flex shrink-0 items-center">{children}</div> : null}
+    </div>
+  );
+}
diff --git a/studio/frontend/src/features/settings/components/settings-section.tsx b/studio/frontend/src/features/settings/components/settings-section.tsx
new file mode 100644
index 0000000000..683dfb25e1
--- /dev/null
+++ b/studio/frontend/src/features/settings/components/settings-section.tsx
@@ -0,0 +1,30 @@
+// SPDX-License-Identifier: AGPL-3.0-only
+// Copyright 2026-present the Unsloth AI Inc. team. All rights reserved. See /studio/LICENSE.AGPL-3.0
+
+import type { ReactNode } from "react";
+
+export function SettingsSection({
+  title,
+  description,
+  children,
+}: {
+  title: string;
+  description?: string;
+  children: ReactNode;
+}) {
+  return (
+    <section className="flex flex-col">
+      <div className="mb-1 flex flex-col gap-0.5">
+        <h2 className="text-base font-semibold font-heading text-foreground">
+          {title}
+        </h2>
+        {description ? (
+          <p className="text-xs text-muted-foreground leading-relaxed">
+            {description}
+          </p>
+        ) : null}
+      </div>
+      <div className="flex flex-col divide-y divide-border/60">{children}</div>
+    </section>
+  );
+}
diff --git a/studio/frontend/src/features/settings/components/theme-segmented.tsx b/studio/frontend/src/features/settings/components/theme-segmented.tsx
new file mode 100644
index 0000000000..1061995346
--- /dev/null
+++ b/studio/frontend/src/features/settings/components/theme-segmented.tsx
@@ -0,0 +1,58 @@
+// SPDX-License-Identifier: AGPL-3.0-only
+// Copyright 2026-present the Unsloth AI Inc. team. All rights reserved. See /studio/LICENSE.AGPL-3.0
+
+import { cn } from "@/lib/utils";
+import {
+  LaptopIcon,
+  Moon02Icon,
+  Sun02Icon,
+} from "@hugeicons/core-free-icons";
+import { HugeiconsIcon } from "@hugeicons/react";
+import { motion, useReducedMotion } from "motion/react";
+import { useTheme, type Theme } from "../stores/theme-store";
+
+const OPTIONS: { value: Theme; label: string; icon: typeof Sun02Icon }[] = [
+  { value: "light", label: "Light", icon: Sun02Icon },
+  { value: "dark", label: "Dark", icon: Moon02Icon },
+  { value: "system", label: "System", icon: LaptopIcon },
+];
+
+export function ThemeSegmented() {
+  const { theme, setTheme } = useTheme();
+  const reduced = useReducedMotion();
+  return (
+    <div className="inline-flex items-center rounded-md border border-border bg-muted/30 p-0.5">
+      {OPTIONS.map((opt) => {
+        const active = theme === opt.value;
+        return (
+          <button
+            key={opt.value}
+            type="button"
+            onClick={() => setTheme(opt.value)}
+            aria-pressed={active}
+            className={cn(
+              "relative flex h-7 items-center gap-1.5 rounded px-2.5 text-xs font-medium transition-colors",
+              active
+                ? "text-foreground"
+                : "text-muted-foreground hover:text-foreground",
+            )}
+          >
+            {active && (
+              <motion.span
+                layoutId="theme-pill"
+                className="absolute inset-0 rounded bg-background shadow-border"
+                transition={
+                  reduced
+                    ? { duration: 0 }
+                    : { type: "spring", stiffness: 500, damping: 35, mass: 0.5 }
+                }
+              />
+            )}
+            <HugeiconsIcon icon={opt.icon} className="relative z-10 size-3.5" />
+            <span className="relative z-10">{opt.label}</span>
+          </button>
+        );
+      })}
+    </div>
+  );
+}
diff --git a/studio/frontend/src/features/settings/components/update-studio-instructions.tsx b/studio/frontend/src/features/settings/components/update-studio-instructions.tsx
new file mode 100644
index 0000000000..e97ef9507c
--- /dev/null
+++ b/studio/frontend/src/features/settings/components/update-studio-instructions.tsx
@@ -0,0 +1,185 @@
+// SPDX-License-Identifier: AGPL-3.0-only
+// Copyright 2026-present the Unsloth AI Inc. team. All rights reserved. See /studio/LICENSE.AGPL-3.0
+
+import { cn } from "@/lib/utils";
+import { copyToClipboard } from "@/lib/copy-to-clipboard";
+import { Copy01Icon, Tick02Icon } from "@hugeicons/core-free-icons";
+import { HugeiconsIcon } from "@hugeicons/react";
+import { AnimatePresence, motion, useReducedMotion } from "motion/react";
+import type { ReactElement } from "react";
+import { useEffect, useRef, useState } from "react";
+
+const STUDIO_UPDATE_CMD = "unsloth studio update";
+const STUDIO_UPDATE_FALLBACK_UNIX_CMD =
+  "curl -fsSL https://unsloth.ai/install.sh | sh";
+const STUDIO_UPDATE_FALLBACK_WINDOWS_CMD =
+  "irm https://unsloth.ai/install.ps1 | iex";
+
+export type UpdateShell = "windows" | "unix";
+
+function getStudioUpdateInstructionLine(shell: UpdateShell): string {
+  return shell === "windows" ? "Open PowerShell and run:" : "Open Terminal and run:";
+}
+
+function CopyableCommand({
+  command,
+  copyLabel,
+}: {
+  command: string;
+  copyLabel: string;
+}): ReactElement {
+  const [copied, setCopied] = useState(false);
+  const timerRef = useRef<ReturnType<typeof setTimeout> | null>(null);
+
+  useEffect(() => {
+    return () => {
+      if (timerRef.current) {
+        clearTimeout(timerRef.current);
+      }
+    };
+  }, []);
+
+  const handleCopy = () => {
+    if (!copyToClipboard(command)) {
+      return;
+    }
+    setCopied(true);
+    if (timerRef.current) {
+      clearTimeout(timerRef.current);
+    }
+    timerRef.current = setTimeout(() => setCopied(false), 2000);
+  };
+
+  return (
+    <div className="flex min-w-0 items-stretch overflow-hidden rounded-md border border-border bg-muted/40">
+      <input
+        type="text"
+        readOnly
+        value={command}
+        className="min-w-0 flex-1 bg-transparent px-2 py-1.5 font-mono text-[11px] text-foreground outline-none"
+        title={command}
+        aria-label={`${copyLabel} text`}
+      />
+      <button
+        type="button"
+        onClick={handleCopy}
+        className="flex shrink-0 items-center justify-center border-l border-border px-2 text-muted-foreground transition-colors hover:bg-muted hover:text-foreground"
+        title={copied ? "Copied" : "Copy command"}
+        aria-label={copied ? `${copyLabel} copied` : `Copy ${copyLabel}`}
+      >
+        {copied ? (
+          <HugeiconsIcon icon={Tick02Icon} className="size-4 text-emerald-600" />
+        ) : (
+          <HugeiconsIcon icon={Copy01Icon} className="size-4" />
+        )}
+      </button>
+    </div>
+  );
+}
+
+export function UpdateStudioInstructions({
+  className,
+  defaultShell,
+  showTitle = true,
+}: {
+  className?: string;
+  defaultShell: UpdateShell;
+  showTitle?: boolean;
+}): ReactElement {
+  const [shell, setShell] = useState<UpdateShell>(defaultShell);
+  const prefersReducedMotion = useReducedMotion();
+  const windows = shell === "windows";
+  const fadeTransition = prefersReducedMotion
+    ? { duration: 0 }
+    : { duration: 0.16, ease: [0.165, 0.84, 0.44, 1] as const };
+  const fadeInitial = prefersReducedMotion ? { opacity: 1 } : { opacity: 0, y: 2 };
+  const fadeAnimate = { opacity: 1, y: 0 };
+  const fadeExit = prefersReducedMotion ? { opacity: 1 } : { opacity: 0, y: -2 };
+
+  useEffect(() => {
+    setShell(defaultShell);
+  }, [defaultShell]);
+
+  return (
+    <div className={cn("flex flex-col gap-3", className)}>
+      <div
+        className={cn(
+          "flex items-center gap-3",
+          showTitle ? "justify-between" : "justify-start",
+        )}
+      >
+        {showTitle ? (
+          <p className="shrink-0 whitespace-nowrap text-sm font-semibold font-heading">
+            Update Unsloth Studio
+          </p>
+        ) : null}
+        <div className="flex shrink-0 items-center gap-0.5 text-[11px]">
+          <button
+            type="button"
+            onClick={() => setShell("windows")}
+            className={cn(
+              "px-0.5 py-0.5 font-medium transition-colors",
+              windows
+                ? "text-foreground"
+                : "text-muted-foreground hover:text-emerald-600",
+            )}
+            aria-pressed={windows}
+          >
+            Windows
+          </button>
+          <span className="text-border">/</span>
+          <button
+            type="button"
+            onClick={() => setShell("unix")}
+            className={cn(
+              "px-0.5 py-0.5 font-medium transition-colors",
+              !windows
+                ? "text-foreground"
+                : "text-muted-foreground hover:text-emerald-600",
+            )}
+            aria-pressed={!windows}
+          >
+            macOS/Linux
+          </button>
+        </div>
+      </div>
+      <AnimatePresence mode="wait" initial={false}>
+        <motion.p
+          key={`instruction-${shell}`}
+          initial={fadeInitial}
+          animate={fadeAnimate}
+          exit={fadeExit}
+          transition={fadeTransition}
+          className="text-xs text-muted-foreground leading-relaxed"
+        >
+          {getStudioUpdateInstructionLine(shell)}
+        </motion.p>
+      </AnimatePresence>
+      <CopyableCommand command={STUDIO_UPDATE_CMD} copyLabel="update command" />
+      <p className="text-xs text-muted-foreground leading-relaxed">
+        If that fails or unsloth studio update is unavailable, run:
+      </p>
+      <AnimatePresence mode="wait" initial={false}>
+        <motion.div
+          key={`fallback-${shell}`}
+          initial={fadeInitial}
+          animate={fadeAnimate}
+          exit={fadeExit}
+          transition={fadeTransition}
+        >
+          <CopyableCommand
+            command={
+              windows
+                ? STUDIO_UPDATE_FALLBACK_WINDOWS_CMD
+                : STUDIO_UPDATE_FALLBACK_UNIX_CMD
+            }
+            copyLabel="fallback command"
+          />
+        </motion.div>
+      </AnimatePresence>
+      <p className="text-xs text-muted-foreground leading-relaxed">
+        Restart Studio after updating for changes to take effect.
+      </p>
+    </div>
+  );
+}
diff --git a/studio/frontend/src/features/settings/components/usage-examples.tsx b/studio/frontend/src/features/settings/components/usage-examples.tsx
new file mode 100644
index 0000000000..603edb3984
--- /dev/null
+++ b/studio/frontend/src/features/settings/components/usage-examples.tsx
@@ -0,0 +1,146 @@
+// SPDX-License-Identifier: AGPL-3.0-only
+// Copyright 2026-present the Unsloth AI Inc. team. All rights reserved. See /studio/LICENSE.AGPL-3.0
+
+import { cn } from "@/lib/utils";
+import { copyToClipboard } from "@/lib/copy-to-clipboard";
+import {
+  ArrowDown01Icon,
+  Copy01Icon,
+  Tick02Icon,
+} from "@hugeicons/core-free-icons";
+import { HugeiconsIcon } from "@hugeicons/react";
+import { AnimatePresence, motion } from "motion/react";
+import { useMemo, useState } from "react";
+
+type Lang = "curl" | "python" | "tools";
+
+const TABS: { id: Lang; label: string }[] = [
+  { id: "curl", label: "curl" },
+  { id: "python", label: "Python" },
+  { id: "tools", label: "Tools" },
+];
+
+function buildSnippets(base: string) {
+  return {
+    curl: `curl ${base}/v1/chat/completions \\
+  -H "Authorization: Bearer sk-unsloth-YOUR_KEY" \\
+  -H "Content-Type: application/json" \\
+  -d '{
+    "messages": [{"role": "user", "content": "Hello"}],
+    "stream": true
+  }'`,
+    python: `from openai import OpenAI
+
+client = OpenAI(
+    base_url="${base}/v1",
+    api_key="sk-unsloth-YOUR_KEY",
+)
+
+response = client.chat.completions.create(
+    model="current",
+    messages=[{"role": "user", "content": "Hello"}],
+    stream=True,
+)
+for chunk in response:
+    print(chunk.choices[0].delta.content or "", end="")`,
+    tools: `curl ${base}/v1/chat/completions \\
+  -H "Authorization: Bearer sk-unsloth-YOUR_KEY" \\
+  -H "Content-Type: application/json" \\
+  -d '{
+    "messages": [{"role": "user", "content": "Search Python 3.13 features"}],
+    "enable_tools": true,
+    "enabled_tools": ["web_search", "python"],
+    "stream": true
+  }'`,
+  };
+}
+
+export function UsageExamples() {
+  const [open, setOpen] = useState(false);
+  const [lang, setLang] = useState<Lang>("curl");
+  const [copied, setCopied] = useState(false);
+  const snippets = useMemo(
+    () =>
+      buildSnippets(
+        typeof window !== "undefined" ? window.location.origin : "",
+      ),
+    [],
+  );
+
+  const handleCopy = () => {
+    if (copyToClipboard(snippets[lang])) {
+      setCopied(true);
+      setTimeout(() => setCopied(false), 1800);
+    }
+  };
+
+  return (
+    <section className="flex flex-col">
+      <button
+        type="button"
+        onClick={() => setOpen((o) => !o)}
+        className="flex w-fit items-center gap-1.5 rounded text-xs font-medium text-muted-foreground transition-colors hover:text-foreground focus-visible:outline-none focus-visible:ring-2 focus-visible:ring-ring"
+        aria-expanded={open}
+      >
+        <HugeiconsIcon
+          icon={ArrowDown01Icon}
+          className={cn("size-3.5 transition-transform", open && "rotate-180")}
+        />
+        {open ? "Hide usage examples" : "Show usage examples"}
+      </button>
+
+      <AnimatePresence initial={false}>
+        {open && (
+          <motion.div
+            initial={{ height: 0, opacity: 0 }}
+            animate={{ height: "auto", opacity: 1 }}
+            exit={{ height: 0, opacity: 0 }}
+            transition={{ duration: 0.18, ease: [0.165, 0.84, 0.44, 1] }}
+            className="overflow-hidden"
+          >
+            <div className="mt-3 overflow-hidden rounded-lg border border-border bg-muted/20">
+              <div className="flex items-center justify-between border-b border-border px-2 py-1.5">
+                <div className="flex items-center gap-0.5">
+                  {TABS.map((t) => {
+                    const active = lang === t.id;
+                    return (
+                      <button
+                        key={t.id}
+                        type="button"
+                        onClick={() => setLang(t.id)}
+                        aria-pressed={active}
+                        className={cn(
+                          "rounded px-2 py-1 text-[11px] font-medium transition-colors focus-visible:outline-none focus-visible:ring-2 focus-visible:ring-ring",
+                          active
+                            ? "bg-background text-foreground shadow-border"
+                            : "text-muted-foreground hover:text-foreground",
+                        )}
+                      >
+                        {t.label}
+                      </button>
+                    );
+                  })}
+                </div>
+                <button
+                  type="button"
+                  onClick={handleCopy}
+                  className="flex items-center gap-1 rounded px-1.5 py-1 text-[11px] text-muted-foreground transition-colors hover:text-foreground focus-visible:outline-none focus-visible:ring-2 focus-visible:ring-ring"
+                  aria-label="Copy snippet"
+                >
+                  <HugeiconsIcon
+                    icon={copied ? Tick02Icon : Copy01Icon}
+                    className={cn("size-3.5", copied && "text-emerald-600")}
+                  />
+                  {copied ? "Copied" : "Copy"}
+                </button>
+              </div>
+              <pre className="overflow-x-auto p-3 font-mono text-[11px] leading-relaxed text-foreground">
+                {snippets[lang]}
+              </pre>
+            </div>
+          </motion.div>
+        )}
+      </AnimatePresence>
+    </section>
+  );
+}
diff --git a/studio/frontend/src/features/settings/index.ts b/studio/frontend/src/features/settings/index.ts
new file mode 100644
index 0000000000..bf9e8fe21a
--- /dev/null
+++ b/studio/frontend/src/features/settings/index.ts
@@ -0,0 +1,6 @@
+// SPDX-License-Identifier: AGPL-3.0-only
+// Copyright 2026-present the Unsloth AI Inc. team. All rights reserved. See /studio/LICENSE.AGPL-3.0
+
+export { SettingsDialog } from "./settings-dialog";
+export { useSettingsDialogStore } from "./stores/settings-dialog-store";
+export type { SettingsTab } from "./stores/settings-dialog-store";
diff --git a/studio/frontend/src/features/settings/settings-dialog.tsx b/studio/frontend/src/features/settings/settings-dialog.tsx
new file mode 100644
index 0000000000..0e2083b31a
--- /dev/null
+++ b/studio/frontend/src/features/settings/settings-dialog.tsx
@@ -0,0 +1,142 @@
+// SPDX-License-Identifier: AGPL-3.0-only
+// Copyright 2026-present the Unsloth AI Inc. team. All rights reserved. See /studio/LICENSE.AGPL-3.0
+
+import {
+  Dialog,
+  DialogContent,
+  DialogDescription,
+  DialogTitle,
+} from "@/components/ui/dialog";
+import { cn } from "@/lib/utils";
+import {
+  Cancel01Icon,
+  Key01Icon,
+  Message01Icon,
+  PaintBrush02Icon,
+  Settings02Icon,
+  SparklesIcon,
+} from "@hugeicons/core-free-icons";
+import { HugeiconsIcon } from "@hugeicons/react";
+import { motion, useReducedMotion } from "motion/react";
+import { useSettingsDialogStore, type SettingsTab } from "./stores/settings-dialog-store";
+import { AboutTab } from "./tabs/about-tab";
+import { ApiKeysTab } from "./tabs/api-keys-tab";
+import { AppearanceTab } from "./tabs/appearance-tab";
+import { ChatTab } from "./tabs/chat-tab";
+import { GeneralTab } from "./tabs/general-tab";
+
+interface TabDef {
+  id: SettingsTab;
+  label: string;
+  icon: typeof Settings02Icon;
+}
+
+const TABS: TabDef[] = [
+  { id: "general", label: "General", icon: Settings02Icon },
+  { id: "appearance", label: "Appearance", icon: PaintBrush02Icon },
+  { id: "chat", label: "Chat", icon: Message01Icon },
+  { id: "api-keys", label: "API Keys", icon: Key01Icon },
+  { id: "about", label: "About", icon: SparklesIcon },
+];
+
+function renderTab(tab: SettingsTab) {
+  switch (tab) {
+    case "general":
+      return <GeneralTab />;
+    case "appearance":
+      return <AppearanceTab />;
+    case "chat":
+      return <ChatTab />;
+    case "api-keys":
+      return <ApiKeysTab />;
+    case "about":
+      return <AboutTab />;
+  }
+}
+
+export function SettingsDialog() {
+  const open = useSettingsDialogStore((s) => s.open);
+  const activeTab = useSettingsDialogStore((s) => s.activeTab);
+  const setActiveTab = useSettingsDialogStore((s) => s.setActiveTab);
+  const closeDialog = useSettingsDialogStore((s) => s.closeDialog);
+  const reduced = useReducedMotion();
+
+  return (
+    <Dialog open={open} onOpenChange={(o) => !o && closeDialog()}>
+      <DialogContent
+        showCloseButton={false}
+        overlayClassName="bg-background/40"
+        className={cn(
+          "!max-w-none h-[560px] w-[820px] p-0 overflow-hidden",
+          "shadow-border rounded-xl border-border",
+          "sm:h-[560px] sm:w-[820px]",
+          "max-sm:h-dvh max-sm:w-dvw max-sm:rounded-none",
+        )}
+      >
+        <DialogTitle className="sr-only">Settings</DialogTitle>
+        <DialogDescription className="sr-only">
+          Manage your Unsloth Studio preferences.
+        </DialogDescription>
+        <div className="flex h-full min-h-0">
+          <aside className="flex w-[200px] shrink-0 flex-col border-r border-border bg-muted/20 p-2">
+            <nav className="flex flex-col gap-0.5">
+              {TABS.map((tab) => {
+                const active = activeTab === tab.id;
+                return (
+                  <button
+                    key={tab.id}
+                    type="button"
+                    onClick={() => setActiveTab(tab.id)}
+                    className={cn(
+                      "relative flex h-9 items-center gap-2 rounded-md px-2.5 text-sm font-medium transition-colors",
+                      "focus-visible:outline-none focus-visible:ring-2 focus-visible:ring-ring focus-visible:ring-offset-1 focus-visible:ring-offset-background",
+                      active
+                        ? "text-foreground"
+                        : "text-muted-foreground hover:text-foreground",
+                    )}
+                  >
+                    {active && (
+                      <motion.span
+                        layoutId="settings-active-pill"
+                        className="absolute inset-0 rounded-md bg-accent"
+                        transition={
+                          reduced
+                            ? { duration: 0 }
+                            : {
+                                type: "spring",
+                                stiffness: 500,
+                                damping: 35,
+                                mass: 0.5,
+                              }
+                        }
+                      />
+                    )}
+                    <HugeiconsIcon
+                      icon={tab.icon}
+                      className="relative z-10 size-4"
+                    />
+                    <span className="relative z-10">{tab.label}</span>
+                  </button>
+                );
+              })}
+            </nav>
+          </aside>
+
+          <main className="relative flex min-w-0 flex-1 flex-col">
+            <button
+              type="button"
+              onClick={closeDialog}
+              className="absolute top-3 right-3 z-10 flex size-7 items-center justify-center rounded-md text-muted-foreground transition-colors hover:bg-accent hover:text-foreground focus-visible:outline-none focus-visible:ring-2 focus-visible:ring-ring"
+              aria-label="Close settings"
+            >
+              <HugeiconsIcon icon={Cancel01Icon} className="size-4" />
+            </button>
+            <div className="flex min-h-0 flex-1 flex-col overflow-y-auto p-6 pr-12">
+              {renderTab(activeTab)}
+            </div>
+          </main>
+        </div>
+      </DialogContent>
+    </Dialog>
+  );
+}
diff --git a/studio/frontend/src/features/settings/stores/settings-dialog-store.ts b/studio/frontend/src/features/settings/stores/settings-dialog-store.ts
new file mode 100644
index 0000000000..75eb53048e
--- /dev/null
+++ b/studio/frontend/src/features/settings/stores/settings-dialog-store.ts
@@ -0,0 +1,52 @@
+// SPDX-License-Identifier: AGPL-3.0-only
+// Copyright 2026-present the Unsloth AI Inc. team. All rights reserved. See /studio/LICENSE.AGPL-3.0
+
+import { create } from "zustand";
+
+export type SettingsTab =
+  | "general"
+  | "appearance"
+  | "chat"
+  | "api-keys"
+  | "about";
+
+interface SettingsDialogState {
+  open: boolean;
+  activeTab: SettingsTab;
+  openDialog: (tab?: SettingsTab) => void;
+  closeDialog: () => void;
+  setActiveTab: (tab: SettingsTab) => void;
+}
+
+const ACTIVE_TAB_KEY = "unsloth_settings_active_tab";
+
+function loadInitialTab(): SettingsTab {
+  if (typeof window === "undefined") return "general";
+  let stored: string | null = null;
+  try {
+    stored = window.localStorage.getItem(ACTIVE_TAB_KEY);
+  } catch {
+    return "general";
+  }
+  const valid: SettingsTab[] = ["general", "appearance", "chat", "api-keys", "about"];
+  return valid.includes(stored as SettingsTab) ? (stored as SettingsTab) : "general";
+}
+
+export const useSettingsDialogStore = create<SettingsDialogState>((set) => ({
+  open: false,
+  activeTab: loadInitialTab(),
+  openDialog: (tab) =>
+    set((state) => ({
+      open: true,
+      activeTab: tab ?? state.activeTab,
+    })),
+  closeDialog: () => set({ open: false }),
+  setActiveTab: (tab) => {
+    try {
+      window.localStorage.setItem(ACTIVE_TAB_KEY, tab);
+    } catch {
+      // ignore storage failures
+    }
+    set({ activeTab: tab });
+  },
+}));
diff --git a/studio/frontend/src/features/settings/stores/theme-store.ts b/studio/frontend/src/features/settings/stores/theme-store.ts
new file mode 100644
index 0000000000..1879459a76
--- /dev/null
+++ b/studio/frontend/src/features/settings/stores/theme-store.ts
@@ -0,0 +1,96 @@
+// SPDX-License-Identifier: AGPL-3.0-only
+// Copyright 2026-present the Unsloth AI Inc. team. All rights reserved. See /studio/LICENSE.AGPL-3.0
+
+import { useSyncExternalStore } from "react";
+
+export type Theme = "light" | "dark" | "system";
+export type ResolvedTheme = "light" | "dark";
+
+const STORAGE_KEY = "theme";
+
+function readStoredTheme(): Theme {
+  if (typeof window === "undefined") return "system";
+  let stored: string | null = null;
+  try {
+    stored = window.localStorage.getItem(STORAGE_KEY);
+  } catch {
+    return "system";
+  }
+  if (stored === "light" || stored === "dark" || stored === "system") return stored;
+  return "system";
+}
+
+function systemPrefersDark(): boolean {
+  if (typeof window === "undefined") return false;
+  return window.matchMedia("(prefers-color-scheme: dark)").matches;
+}
+
+function resolveTheme(theme: Theme): ResolvedTheme {
+  if (theme === "system") return systemPrefersDark() ? "dark" : "light";
+  return theme;
+}
+
+function applyToDocument(resolved: ResolvedTheme) {
+  if (typeof document === "undefined") return;
+  document.documentElement.classList.toggle("dark", resolved === "dark");
+}
+
+const listeners = new Set<() => void>();
+function subscribe(cb: () => void) {
+  listeners.add(cb);
+  if (typeof window === "undefined") {
+    return () => listeners.delete(cb);
+  }
+  const mq = window.matchMedia("(prefers-color-scheme: dark)");
+  const syncTheme = () => {
+    applyToDocument(resolveTheme(readStoredTheme()));
+    cb();
+  };
+  const onStorage = (e: StorageEvent) => {
+    if (e.key === STORAGE_KEY || e.key === null) syncTheme();
+  };
+  mq.addEventListener("change", syncTheme);
+  window.addEventListener("storage", onStorage);
+  return () => {
+    listeners.delete(cb);
+    mq.removeEventListener("change", syncTheme);
+    window.removeEventListener("storage", onStorage);
+  };
+}
+
+function getSnapshot(): Theme {
+  return readStoredTheme();
+}
+
+function getServerSnapshot(): Theme {
+  return "system";
+}
+
+/**
+ * Single source of truth for setting the theme. All writers (the Settings
+ * dialog's segmented control AND the sidebar dropdown's animated toggler)
+ * must route through this so the DOM class, localStorage, and React
+ * subscribers stay in sync.
+ */
+export function setTheme(next: Theme): void {
+  if (typeof window === "undefined") return;
+  // Persist "system" explicitly so next-themes (mounted with
+  // defaultTheme="light") doesn't clobber the choice on reload.
+  try {
+    window.localStorage.setItem(STORAGE_KEY, next);
+  } catch {
+    // ignore storage failures
+  }
+  applyToDocument(resolveTheme(next));
+  listeners.forEach((cb) => cb());
+}
+
+export function useTheme(): {
+  theme: Theme;
+  resolved: ResolvedTheme;
+  setTheme: (next: Theme) => void;
+} {
+  const theme = useSyncExternalStore(subscribe, getSnapshot, getServerSnapshot);
+  const resolved = resolveTheme(theme);
+  return { theme, resolved, setTheme };
+}
diff --git a/studio/frontend/src/features/settings/tabs/about-tab.tsx b/studio/frontend/src/features/settings/tabs/about-tab.tsx
new file mode 100644
index 0000000000..f25afc6d8f
--- /dev/null
+++ b/studio/frontend/src/features/settings/tabs/about-tab.tsx
@@ -0,0 +1,101 @@
+// SPDX-License-Identifier: AGPL-3.0-only
+// Copyright 2026-present the Unsloth AI Inc. team. All rights reserved. See /studio/LICENSE.AGPL-3.0
+
+import { Button } from "@/components/ui/button";
+import { ShutdownDialog } from "@/components/shutdown-dialog";
+import { UpdateStudioInstructions } from "../components/update-studio-instructions";
+import { usePlatformStore } from "@/config/env";
+import { removeTrainingUnloadGuard } from "@/features/training/hooks/use-training-unload-guard";
+import {
+  ArrowUpRight01Icon,
+  Book03Icon,
+  Cancel01Icon,
+  MessageNotification01Icon,
+} from "@hugeicons/core-free-icons";
+import { HugeiconsIcon } from "@hugeicons/react";
+import { useState } from "react";
+import { SettingsRow } from "../components/settings-row";
+import { SettingsSection } from "../components/settings-section";
+
+const VERSION: string =
+  (import.meta.env.VITE_APP_VERSION as string | undefined) ?? "dev";
+
+export function AboutTab() {
+  const deviceType = usePlatformStore((s) => s.deviceType);
+  const defaultShell = deviceType === "windows" ? "windows" : "unix";
+  const [shutdownOpen, setShutdownOpen] = useState(false);
+
+  return (
+    <div className="flex flex-col gap-6">
+      <header className="flex flex-col gap-1">
+        <h1 className="text-lg font-semibold font-heading">About</h1>
+        <p className="text-xs text-muted-foreground">
+          Unsloth Studio build info and support.
+        </p>
+      </header>
+
+      <SettingsSection title="Studio">
+        <SettingsRow label="Version">
+          <code className="font-mono text-xs text-muted-foreground">{VERSION}</code>
+        </SettingsRow>
+      </SettingsSection>
+
+      <SettingsSection title="Updates">
+        <div className="py-2">
+          <UpdateStudioInstructions defaultShell={defaultShell} showTitle={false} />
+        </div>
+      </SettingsSection>
+
+      <SettingsSection title="Help">
+        <SettingsRow label="Documentation">
+          <a
+            href="https://unsloth.ai/docs"
+            target="_blank"
+            rel="noopener noreferrer"
+            className="inline-flex items-center gap-1 text-xs font-medium text-muted-foreground hover:text-foreground"
+          >
+            <HugeiconsIcon icon={Book03Icon} className="size-3.5" />
+            unsloth.ai/docs
+            <HugeiconsIcon icon={ArrowUpRight01Icon} className="size-3" />
+          </a>
+        </SettingsRow>
+        <SettingsRow label="Feedback">
+          <a
+            href="https://github.com/unslothai/unsloth/issues"
+            target="_blank"
+            rel="noopener noreferrer"
+            className="inline-flex items-center gap-1 text-xs font-medium text-muted-foreground hover:text-foreground"
+          >
+            <HugeiconsIcon icon={MessageNotification01Icon} className="size-3.5" />
+            Report an issue
+            <HugeiconsIcon icon={ArrowUpRight01Icon} className="size-3" />
+          </a>
+        </SettingsRow>
+      </SettingsSection>
+
+      <SettingsSection title="Danger zone">
+        <SettingsRow
+          destructive
+          label="Shut down Unsloth Studio"
+          description="Stops the Studio server process and ends your session."
+        >
+          <Button
+            variant="outline"
+            size="sm"
+            onClick={() => setShutdownOpen(true)}
+            className="text-destructive hover:text-destructive hover:border-destructive/60"
+          >
+            <HugeiconsIcon icon={Cancel01Icon} className="size-3.5 mr-1.5" />
+            Shut down
+          </Button>
+        </SettingsRow>
+      </SettingsSection>
+
+      <ShutdownDialog
+        open={shutdownOpen}
+        onOpenChange={setShutdownOpen}
+        onAfterShutdown={removeTrainingUnloadGuard}
+      />
+    </div>
+  );
+}
diff --git a/studio/frontend/src/features/settings/tabs/api-keys-tab.tsx b/studio/frontend/src/features/settings/tabs/api-keys-tab.tsx
new file mode 100644
index 0000000000..e0fa6d87d2
--- /dev/null
+++ b/studio/frontend/src/features/settings/tabs/api-keys-tab.tsx
@@ -0,0 +1,159 @@
+// SPDX-License-Identifier: AGPL-3.0-only
+// Copyright 2026-present the Unsloth AI Inc. team. All rights reserved. See /studio/LICENSE.AGPL-3.0
+
+import {
+  Dialog,
+  DialogContent,
+  DialogDescription,
+  DialogFooter,
+  DialogHeader,
+  DialogTitle,
+} from "@/components/ui/dialog";
+import { Button } from "@/components/ui/button";
+import { AnimatePresence, motion, useReducedMotion } from "motion/react";
+import { useCallback, useEffect, useState } from "react";
+import { fetchApiKeys, revokeApiKey, type ApiKey } from "../api/api-keys";
+import { ApiKeyRow } from "../components/api-key-row";
+import { CreateKeyForm } from "../components/create-key-form";
+import { KeyRevealCard } from "../components/key-reveal-card";
+import { UsageExamples } from "../components/usage-examples";
+
+export function ApiKeysTab() {
+  const [keys, setKeys] = useState<ApiKey[]>([]);
+  const [loading, setLoading] = useState(true);
+  const [error, setError] = useState<string | null>(null);
+  const [revokeTarget, setRevokeTarget] = useState<ApiKey | null>(null);
+  const [revoking, setRevoking] = useState(false);
+  const [revealed, setRevealed] = useState<string | null>(null);
+  const reduced = useReducedMotion();
+  const t = reduced
+    ? { duration: 0 }
+    : { duration: 0.18, ease: [0.165, 0.84, 0.44, 1] as const };
+
+  const load = useCallback(async () => {
+    setLoading(true);
+    setError(null);
+    try {
+      setKeys(await fetchApiKeys());
+    } catch (e) {
+      setError(e instanceof Error ? e.message : "Couldn't load API keys.");
+    } finally {
+      setLoading(false);
+    }
+  }, []);
+
+  useEffect(() => {
+    void load();
+  }, [load]);
+
+  const confirmRevoke = async () => {
+    if (!revokeTarget) return;
+    setRevoking(true);
+    try {
+      await revokeApiKey(revokeTarget.id);
+      await load();
+      setRevokeTarget(null);
+    } catch (e) {
+      setError(e instanceof Error ? e.message : "Couldn't revoke key.");
+    } finally {
+      setRevoking(false);
+    }
+  };
+
+  return (
+    <div className="flex flex-col gap-6">
+      <header className="flex flex-col gap-1">
+        <h1 className="text-lg font-semibold font-heading">API Keys</h1>
+        <p className="text-xs text-muted-foreground">
+          Access Unsloth Studio programmatically via the OpenAI-compatible API.
+        </p>
+      </header>
+
+      <AnimatePresence mode="wait" initial={false}>
+        {revealed !== null ? (
+          <motion.div
+            key="reveal"
+            initial={{ opacity: 0, y: -4 }}
+            animate={{ opacity: 1, y: 0 }}
+            exit={{ opacity: 0, y: -4 }}
+            transition={t}
+          >
+            <KeyRevealCard
+              rawKey={revealed}
+              onDone={() => setRevealed(null)}
+            />
+          </motion.div>
+        ) : (
+          <motion.div
+            key="form"
+            initial={{ opacity: 0, y: 4 }}
+            animate={{ opacity: 1, y: 0 }}
+            exit={{ opacity: 0, y: 4 }}
+            transition={t}
+          >
+            <CreateKeyForm
+              onCreated={(raw) => {
+                setRevealed(raw);
+                void load();
+              }}
+              onError={setError}
+            />
+          </motion.div>
+        )}
+      </AnimatePresence>
+
+      <section className="flex flex-col">
+        <h2 className="mb-2 text-sm font-semibold text-foreground">Your keys</h2>
+        {error ? (
+          <div className="rounded-md border border-destructive/20 bg-destructive/5 p-3 text-xs text-destructive">
+            {error}
+          </div>
+        ) : loading ? (
+          <div className="flex flex-col gap-2 py-2">
+            {[0, 1].map((i) => (
+              <div
+                key={i}
+                className="h-12 animate-pulse rounded-md bg-muted/40"
+              />
+            ))}
+          </div>
+        ) : keys.length === 0 ? (
+          <p className="py-6 text-center text-xs text-muted-foreground">
+            No API keys yet.
+          </p>
+        ) : (
+          <div className="flex flex-col">
+            {keys.map((k) => (
+              <ApiKeyRow key={k.id} apiKey={k} onRevoke={setRevokeTarget} />
+            ))}
+          </div>
+        )}
+      </section>
+
+      <UsageExamples />
+
+      <Dialog open={revokeTarget !== null} onOpenChange={(o) => !o && setRevokeTarget(null)}>
+        <DialogContent className="max-w-md">
+          <DialogHeader>
+            <DialogTitle>Revoke key “{revokeTarget?.name}”?</DialogTitle>
+            <DialogDescription>
+              Applications using this key will immediately lose access. This cannot be undone.
+            </DialogDescription>
+          </DialogHeader>
+          <DialogFooter>
+            <Button variant="outline" onClick={() => setRevokeTarget(null)}>
+              Cancel
+            </Button>
+            <Button
+              onClick={confirmRevoke}
+              disabled={revoking}
+              className="bg-destructive hover:bg-destructive/90 text-destructive-foreground"
+            >
+              {revoking ? "Revoking…" : `Revoke “${revokeTarget?.name}”`}
+            </Button>
+          </DialogFooter>
+        </DialogContent>
+      </Dialog>
+    </div>
+  );
+}
diff --git a/studio/frontend/src/features/settings/tabs/appearance-tab.tsx b/studio/frontend/src/features/settings/tabs/appearance-tab.tsx
new file mode 100644
index 0000000000..edcf3a9e4a
--- /dev/null
+++ b/studio/frontend/src/features/settings/tabs/appearance-tab.tsx
@@ -0,0 +1,40 @@
+// SPDX-License-Identifier: AGPL-3.0-only
+// Copyright 2026-present the Unsloth AI Inc. team. All rights reserved. See /studio/LICENSE.AGPL-3.0
+
+import { Switch } from "@/components/ui/switch";
+import { useSidebarPin } from "@/hooks/use-sidebar-pin";
+import { SettingsRow } from "../components/settings-row";
+import { SettingsSection } from "../components/settings-section";
+import { ThemeSegmented } from "../components/theme-segmented";
+
+export function AppearanceTab() {
+  const { pinned, setPinned } = useSidebarPin();
+  return (
+    <div className="flex flex-col gap-6">
+      <header className="flex flex-col gap-1">
+        <h1 className="text-lg font-semibold font-heading">Appearance</h1>
+        <p className="text-xs text-muted-foreground">
+          How Unsloth Studio looks on this device.
+        </p>
+      </header>
+
+      <SettingsSection title="Theme">
+        <SettingsRow
+          label="Color scheme"
+          description="Choose light, dark, or follow your system."
+        >
+          <ThemeSegmented />
+        </SettingsRow>
+      </SettingsSection>
+
+      <SettingsSection title="Layout">
+        <SettingsRow
+          label="Pin sidebar by default"
+          description="Keep the sidebar expanded instead of collapsing to icons."
+        >
+          <Switch checked={pinned} onCheckedChange={setPinned} />
+        </SettingsRow>
+      </SettingsSection>
+    </div>
+  );
+}
diff --git a/studio/frontend/src/features/settings/tabs/chat-tab.tsx b/studio/frontend/src/features/settings/tabs/chat-tab.tsx
new file mode 100644
index 0000000000..d9ebfa457d
--- /dev/null
+++ b/studio/frontend/src/features/settings/tabs/chat-tab.tsx
@@ -0,0 +1,129 @@
+// SPDX-License-Identifier: AGPL-3.0-only
+// Copyright 2026-present the Unsloth AI Inc. team. All rights reserved. See /studio/LICENSE.AGPL-3.0
+
+import { Button } from "@/components/ui/button";
+import {
+  Dialog,
+  DialogContent,
+  DialogDescription,
+  DialogFooter,
+  DialogHeader,
+  DialogTitle,
+} from "@/components/ui/dialog";
+import {
+  clearAllChats,
+  countAllChats,
+} from "@/features/chat/utils/clear-all-chats";
+import { downloadChatExport } from "@/features/chat/utils/export-chat-history";
+import { Delete02Icon, Download02Icon } from "@hugeicons/core-free-icons";
+import { HugeiconsIcon } from "@hugeicons/react";
+import { useEffect, useState } from "react";
+import { SettingsRow } from "../components/settings-row";
+import { SettingsSection } from "../components/settings-section";
+
+export function ChatTab() {
+  const [confirmOpen, setConfirmOpen] = useState(false);
+  const [count, setCount] = useState<number | null>(null);
+  const [exporting, setExporting] = useState(false);
+  const [clearing, setClearing] = useState(false);
+
+  useEffect(() => {
+    void countAllChats().then(setCount);
+  }, []);
+
+  const handleExport = async () => {
+    setExporting(true);
+    try {
+      await downloadChatExport();
+    } finally {
+      setExporting(false);
+    }
+  };
+
+  const handleClear = async () => {
+    setClearing(true);
+    try {
+      await clearAllChats();
+      setCount(0);
+      setConfirmOpen(false);
+    } finally {
+      setClearing(false);
+    }
+  };
+
+  return (
+    <div className="flex flex-col gap-6">
+      <header className="flex flex-col gap-1">
+        <h1 className="text-lg font-semibold font-heading">Chat</h1>
+        <p className="text-xs text-muted-foreground">
+          Manage your chat history stored on this device.
+        </p>
+      </header>
+
+      <SettingsSection title="Data">
+        <SettingsRow
+          label="Export chat history"
+          description="Download all chats and messages as a JSON file."
+        >
+          <Button
+            variant="outline"
+            size="sm"
+            onClick={handleExport}
+            disabled={exporting || count === 0}
+          >
+            <HugeiconsIcon icon={Download02Icon} className="size-3.5 mr-1.5" />
+            {exporting ? "Exporting…" : "Export"}
+          </Button>
+        </SettingsRow>
+
+        <SettingsRow
+          destructive
+          label="Clear all chats"
+          description={
+            count === null
+              ? "Permanently delete every chat on this device."
+              : count === 0
+                ? "No chats to clear."
+                : `Permanently delete all ${count} chat${count === 1 ? "" : "s"} on this device.`
+          }
+        >
+          <Button
+            variant="outline"
+            size="sm"
+            onClick={() => setConfirmOpen(true)}
+            disabled={count === 0}
+            className="text-destructive hover:text-destructive hover:border-destructive/60"
+          >
+            <HugeiconsIcon icon={Delete02Icon} className="size-3.5 mr-1.5" />
+            Clear chats
+          </Button>
+        </SettingsRow>
+      </SettingsSection>
+
+      <Dialog open={confirmOpen} onOpenChange={setConfirmOpen}>
+        <DialogContent className="max-w-md">
+          <DialogHeader>
+            <DialogTitle>
+              Clear {count ?? 0} chat{count === 1 ? "" : "s"}?
+            </DialogTitle>
+            <DialogDescription>
+              This permanently deletes every chat and message stored on this device. This cannot be undone.
+            </DialogDescription>
+          </DialogHeader>
+          <DialogFooter>
+            <Button variant="outline" onClick={() => setConfirmOpen(false)}>
+              Cancel
+            </Button>
+            <Button
+              onClick={handleClear}
+              disabled={clearing}
+              className="bg-destructive hover:bg-destructive/90 text-destructive-foreground"
+            >
+              {clearing ? "Clearing…" : `Clear ${count ?? 0} chat${count === 1 ? "" : "s"}`}
+            </Button>
+          </DialogFooter>
+        </DialogContent>
+      </Dialog>
+    </div>
+  );
+}
diff --git a/studio/frontend/src/features/settings/tabs/general-tab.tsx b/studio/frontend/src/features/settings/tabs/general-tab.tsx
new file mode 100644
index 0000000000..e1ad3fb58c
--- /dev/null
+++ b/studio/frontend/src/features/settings/tabs/general-tab.tsx
@@ -0,0 +1,197 @@
+// SPDX-License-Identifier: AGPL-3.0-only
+// Copyright 2026-present the Unsloth AI Inc. team. All rights reserved. See /studio/LICENSE.AGPL-3.0
+
+import { Button } from "@/components/ui/button";
+import {
+  Dialog,
+  DialogContent,
+  DialogDescription,
+  DialogFooter,
+  DialogHeader,
+  DialogTitle,
+} from "@/components/ui/dialog";
+import { Input } from "@/components/ui/input";
+import { Switch } from "@/components/ui/switch";
+import { useChatRuntimeStore } from "@/features/chat/stores/chat-runtime-store";
+import { useEffect, useRef, useState } from "react";
+import { Eye, EyeOff } from "lucide-react";
+import { SettingsRow } from "../components/settings-row";
+import { SettingsSection } from "../components/settings-section";
+
+// Keys cleared by "Reset all local preferences".
+//
+// NEVER include auth / session keys here — resetting them would log the user
+// out, which is not what users expect from a "reset preferences" button.
+//
+// Explicitly EXCLUDED:
+//   - "unsloth_auth_token"                 (auth: access token)
+//   - "unsloth_auth_refresh_token"         (auth: refresh token)
+//   - "unsloth_auth_must_change_password"  (auth: forced password change flag)
+//   - "unsloth_onboarding_done"            (session: would force re-onboarding)
+const PREFS_KEYS: string[] = [
+  // Appearance
+  "theme",
+  // UI state
+  "sidebar_pinned",
+  "unsloth_sidebar_navigate_open",
+  "unsloth_settings_active_tab",
+  // Chat runtime prefs
+  "unsloth_chat_auto_title",
+  "unsloth_hf_token",
+  "unsloth_auto_heal_tool_calls",
+  "unsloth_max_tool_calls_per_message",
+  "unsloth_tool_call_timeout",
+  "unsloth_chat_inference_params",
+  "unsloth_chat_collapsible_state",
+  // Chat presets
+  "unsloth_chat_custom_presets",
+  "unsloth_chat_active_preset",
+  "unsloth_chat_system_prompts",
+  "unsloth_chat_system_prompts_migrated",
+  // Training UI prefs
+  "unsloth_training_config_v1",
+  "unsloth_prev_max_steps",
+  "unsloth_prev_save_steps",
+  // Guided tour flags
+  "tour:studio:v1",
+];
+
+// Set to true from resetAllPrefs so the unmount-commit effect skips writing
+// back the in-memory draft — otherwise the cleanup would re-persist the old
+// HF token into localStorage after it was just cleared, and the subsequent
+// reload would read the re-written value.
+let resetInProgress = false;
+
+function resetAllPrefs() {
+  resetInProgress = true;
+  for (const key of PREFS_KEYS) {
+    try {
+      localStorage.removeItem(key);
+    } catch {
+      // ignore
+    }
+  }
+  window.location.reload();
+}
+
+export function GeneralTab() {
+  const hfToken = useChatRuntimeStore((s) => s.hfToken);
+  const setHfToken = useChatRuntimeStore((s) => s.setHfToken);
+  const autoTitle = useChatRuntimeStore((s) => s.autoTitle);
+  const setAutoTitle = useChatRuntimeStore((s) => s.setAutoTitle);
+
+  const [draftToken, setDraftToken] = useState(hfToken ?? "");
+  const [showToken, setShowToken] = useState(false);
+  const [confirmOpen, setConfirmOpen] = useState(false);
+
+  const draftRef = useRef(draftToken);
+  useEffect(() => {
+    draftRef.current = draftToken;
+  }, [draftToken]);
+
+  // Commit on unmount (dialog close / tab switch). Skip during reset-prefs
+  // flow so we don't re-persist the draft after localStorage was cleared.
+  useEffect(() => {
+    return () => {
+      if (resetInProgress) return;
+      const trimmed = draftRef.current.trim();
+      const current = useChatRuntimeStore.getState().hfToken;
+      if (trimmed !== current) {
+        useChatRuntimeStore.getState().setHfToken(trimmed);
+      }
+    };
+  }, []);
+
+  const commitToken = () => {
+    const trimmed = draftToken.trim();
+    if (trimmed !== draftToken) setDraftToken(trimmed);
+    if (trimmed !== hfToken) setHfToken(trimmed);
+  };
+
+  return (
+    <div className="flex flex-col gap-6">
+      <header className="flex flex-col gap-1">
+        <h1 className="text-lg font-semibold font-heading">General</h1>
+        <p className="text-xs text-muted-foreground">
+          Global preferences for Unsloth Studio.
+        </p>
+      </header>
+
+      <SettingsSection title="Account">
+        <SettingsRow
+          label="Hugging Face token"
+          description="Used to load gated models and push artifacts."
+        >
+          <div className="relative w-[260px]">
+            <Input
+              type={showToken ? "text" : "password"}
+              placeholder="hf_…"
+              value={draftToken}
+              onChange={(e) => setDraftToken(e.target.value)}
+              onBlur={commitToken}
+              className="h-8 w-full pr-8 font-mono text-xs"
+            />
+            <button
+              type="button"
+              onClick={() => setShowToken((s) => !s)}
+              className="absolute right-1.5 top-1/2 flex size-5 -translate-y-1/2 items-center justify-center rounded text-muted-foreground transition-colors hover:text-foreground"
+              aria-label={showToken ? "Hide token" : "Show token"}
+              tabIndex={-1}
+            >
+              {showToken ? <EyeOff className="size-3.5" /> : <Eye className="size-3.5" />}
+            </button>
+          </div>
+        </SettingsRow>
+      </SettingsSection>
+
+      <SettingsSection title="Chat defaults">
+        <SettingsRow
+          label="Auto-title new chats"
+          description="Generate a short title from the first message."
+        >
+          <Switch checked={autoTitle} onCheckedChange={setAutoTitle} />
+        </SettingsRow>
+      </SettingsSection>
+
+      <SettingsSection title="Danger zone">
+        <SettingsRow
+          destructive
+          label="Reset all local preferences"
+          description="Clears theme, tokens, sidebar state, and presets. Chats and API keys are not affected."
+        >
+          <Button
+            variant="outline"
+            size="sm"
+            onClick={() => setConfirmOpen(true)}
+            className="text-destructive hover:text-destructive hover:border-destructive/60"
+          >
+            Reset preferences
+          </Button>
+        </SettingsRow>
+      </SettingsSection>
+
+      <Dialog open={confirmOpen} onOpenChange={setConfirmOpen}>
+        <DialogContent className="max-w-md">
+          <DialogHeader>
+            <DialogTitle>Reset all local preferences?</DialogTitle>
+            <DialogDescription>
+              This clears your theme, tokens, and stored settings, then reloads
+              Studio. Chats and API keys are not affected.
+            </DialogDescription>
+          </DialogHeader>
+          <DialogFooter>
+            <Button variant="outline" onClick={() => setConfirmOpen(false)}>
+              Cancel
+            </Button>
+            <Button
+              onClick={resetAllPrefs}
+              className="bg-destructive hover:bg-destructive/90 text-destructive-foreground"
+            >
+              Reset and reload
+            </Button>
+          </DialogFooter>
+        </DialogContent>
+      </Dialog>
+    </div>
+  );
+}
diff --git a/studio/frontend/src/features/studio/studio-page.tsx b/studio/frontend/src/features/studio/studio-page.tsx
index 07a045685e..d8a35359b8 100644
--- a/studio/frontend/src/features/studio/studio-page.tsx
+++ b/studio/frontend/src/features/studio/studio-page.tsx
@@ -14,7 +14,8 @@ import { Tabs, TabsContent, TabsList, TabsTrigger } from "@/components/ui/tabs";
 import { Button } from "@/components/ui/button";
 import { ArrowLeft01Icon } from "@hugeicons/core-free-icons";
 import { HugeiconsIcon } from "@hugeicons/react";
-import { type ReactElement, useEffect, useState } from "react";
+import { type ReactElement, useCallback, useEffect, useMemo, useState } from "react";
+import { useSidebar } from "@/components/ui/sidebar";
 import { DatasetPreviewDialog } from "./sections/dataset-preview-dialog";
 import { DatasetSection } from "./sections/dataset-section";
 import { ModelSection } from "./sections/model-section";
@@ -49,7 +50,12 @@ export function StudioPage(): ReactElement {
   const closeDialog = useDatasetPreviewDialogStore((s) => s.close);
 
   const [requestedTab, setRequestedTab] = useState("configure");
-  const [selectedHistoryRunId, setSelectedHistoryRunId] = useState<string | null>(null);
+  const selectedHistoryRunId = useTrainingRuntimeStore((s) => s.selectedHistoryRunId);
+  const setSelectedHistoryRunId = useTrainingRuntimeStore((s) => s.setSelectedHistoryRunId);
+
+  useEffect(() => {
+    return () => setSelectedHistoryRunId(null);
+  }, [setSelectedHistoryRunId]);
 
   // Derive activeTab: auto-switch to "current-run" only while training is
   // genuinely running.  Once training ends, honour whatever tab the user clicks.
@@ -61,9 +67,20 @@ export function StudioPage(): ReactElement {
         ? "configure"
         : requestedTab;
 
+  const { setPinned } = useSidebar();
+  const pinSidebar = useCallback(() => setPinned(true), [setPinned]);
+
   const tourEnabled = hasHydratedRuntime && !isHydratingRuntime;
   const isConfigTour = activeTab === "configure";
-  const tourSteps = activeTab === "current-run" ? studioTrainingTourSteps : studioTourSteps;
+  const baseTourSteps = activeTab === "current-run" ? studioTrainingTourSteps : studioTourSteps;
+  // Inject onEnter for navbar-targeting steps so the sidebar expands during the tour.
+  const tourSteps = useMemo(
+    () =>
+      baseTourSteps.map((step) =>
+        step.target === "navbar" ? { ...step, onEnter: pinSidebar } : step,
+      ),
+    [baseTourSteps, pinSidebar],
+  );
   const tour = useGuidedTourController({
     id: "studio",
     steps: tourSteps,
@@ -86,6 +103,14 @@ export function StudioPage(): ReactElement {
     }
   }, [isTrainingRunning, requestedTab]);
 
+  // Selecting a run from the sidebar only sets selectedHistoryRunId; auto-switch
+  // to the History tab so the main panel reflects the selection.
+  useEffect(() => {
+    if (selectedHistoryRunId && requestedTab !== "history") {
+      setRequestedTab("history");
+    }
+  }, [selectedHistoryRunId, requestedTab]);
+
   useEffect(() => {
     ensureModelDefaultsLoaded();
     ensureDatasetChecked();
@@ -106,7 +131,7 @@ export function StudioPage(): ReactElement {
   })();
 
   return (
-    <div className="relative min-h-screen overflow-hidden bg-background">
+    <div className="relative min-h-screen bg-background">
       <main className="relative z-10 mx-auto max-w-7xl px-4 py-4 sm:px-6">
         <GuidedTour {...tour.tourProps} celebrate={isConfigTour} />
 
diff --git a/studio/frontend/src/features/training/hooks/use-training-history-sidebar.ts b/studio/frontend/src/features/training/hooks/use-training-history-sidebar.ts
new file mode 100644
index 0000000000..d5fefb283e
--- /dev/null
+++ b/studio/frontend/src/features/training/hooks/use-training-history-sidebar.ts
@@ -0,0 +1,56 @@
+// SPDX-License-Identifier: AGPL-3.0-only
+// Copyright 2026-present the Unsloth AI Inc. team. All rights reserved. See /studio/LICENSE.AGPL-3.0
+
+import { useCallback, useEffect, useRef, useState } from "react";
+import { listTrainingRuns } from "../api/history-api";
+import type { TrainingRunSummary } from "../types/history";
+
+const SIDEBAR_LIMIT = 20;
+const RUNNING_POLL_MS = 5000;
+
+export function useTrainingHistorySidebarItems(enabled: boolean) {
+  const [items, setItems] = useState<TrainingRunSummary[]>([]);
+  const [loaded, setLoaded] = useState(false);
+  const controllerRef = useRef<AbortController | null>(null);
+  const inFlightRef = useRef(false);
+
+  const fetchRuns = useCallback(async () => {
+    if (inFlightRef.current) {
+      return;
+    }
+    const controller = new AbortController();
+    controllerRef.current = controller;
+    inFlightRef.current = true;
+    try {
+      const result = await listTrainingRuns(SIDEBAR_LIMIT, 0, controller.signal);
+      setItems(result.runs);
+      setLoaded(true);
+    } catch (err) {
+      if (err instanceof DOMException && err.name === "AbortError") return;
+    } finally {
+      if (controllerRef.current === controller) {
+        controllerRef.current = null;
+      }
+      inFlightRef.current = false;
+    }
+  }, []);
+
+  useEffect(() => {
+    if (!enabled) return;
+    void fetchRuns();
+    return () => {
+      controllerRef.current?.abort();
+    };
+  }, [enabled, fetchRuns]);
+
+  const hasRunning = items.some((r) => r.status === "running");
+  useEffect(() => {
+    if (!enabled || !hasRunning) return;
+    const timer = setInterval(() => {
+      void fetchRuns();
+    }, RUNNING_POLL_MS);
+    return () => clearInterval(timer);
+  }, [enabled, hasRunning, fetchRuns]);
+
+  return { items, loaded, refresh: fetchRuns };
+}
diff --git a/studio/frontend/src/features/training/hooks/use-training-unload-guard.ts b/studio/frontend/src/features/training/hooks/use-training-unload-guard.ts
new file mode 100644
index 0000000000..78c1e3d5c3
--- /dev/null
+++ b/studio/frontend/src/features/training/hooks/use-training-unload-guard.ts
@@ -0,0 +1,40 @@
+// SPDX-License-Identifier: AGPL-3.0-only
+// Copyright 2026-present the Unsloth AI Inc. team. All rights reserved. See /studio/LICENSE.AGPL-3.0
+
+import { useEffect } from "react";
+import { useTrainingRuntimeStore } from "@/features/training";
+
+let currentHandler: ((e: BeforeUnloadEvent) => void) | null = null;
+
+/**
+ * Mounts a beforeunload guard that warns the user if training is running.
+ * Call once at the app root.
+ */
+export function useTrainingUnloadGuard() {
+  useEffect(() => {
+    const handler = (e: BeforeUnloadEvent) => {
+      if (!useTrainingRuntimeStore.getState().isTrainingRunning) return;
+      e.preventDefault();
+      e.returnValue = "";
+    };
+    currentHandler = handler;
+    window.addEventListener("beforeunload", handler);
+    return () => {
+      if (currentHandler === handler) currentHandler = null;
+      window.removeEventListener("beforeunload", handler);
+    };
+  }, []);
+}
+
+/**
+ * Removes the active beforeunload guard (if any).
+ * Call this before intentionally ending the session (e.g. shutting down
+ * the Studio server) so the "Server stopped" page can render without
+ * the browser prompting the user to confirm leaving.
+ */
+export function removeTrainingUnloadGuard() {
+  if (currentHandler) {
+    window.removeEventListener("beforeunload", currentHandler);
+    currentHandler = null;
+  }
+}
diff --git a/studio/frontend/src/features/training/index.ts b/studio/frontend/src/features/training/index.ts
index af34b306cf..efce4ca921 100644
--- a/studio/frontend/src/features/training/index.ts
+++ b/studio/frontend/src/features/training/index.ts
@@ -7,6 +7,7 @@ export {
   useTrainingRuntimeStore,
 } from "./stores/training-runtime-store";
 export { useTrainingActions } from "./hooks/use-training-actions";
+export { useTrainingHistorySidebarItems } from "./hooks/use-training-history-sidebar";
 export { useTrainingRuntimeLifecycle } from "./hooks/use-training-runtime-lifecycle";
 export { useMaxStepsEpochsToggle } from "./hooks/use-max-steps-epochs-toggle";
 export { HfDatasetSubsetSplitSelectors } from "./components/hf-dataset-subset-split-selectors";
diff --git a/studio/frontend/src/features/training/stores/training-runtime-store.ts b/studio/frontend/src/features/training/stores/training-runtime-store.ts
index a57c444c8d..d958078330 100644
--- a/studio/frontend/src/features/training/stores/training-runtime-store.ts
+++ b/studio/frontend/src/features/training/stores/training-runtime-store.ts
@@ -41,6 +41,7 @@ const initialState: TrainingRuntimeState = {
   evalLossHistory: [],
   resetGeneration: 0,
   stopRequested: false,
+  selectedHistoryRunId: null,
 };
 
 function sortSeries(points: TrainingSeriesPoint[]): TrainingSeriesPoint[] {
@@ -171,6 +172,9 @@ export const useTrainingRuntimeStore = create<TrainingRuntimeStore>()((set) => (
       sseConnected: false,
     }),
 
+  setSelectedHistoryRunId: (selectedHistoryRunId) =>
+    set({ selectedHistoryRunId }),
+
   applyStatus: (payload) =>
     set((state) => {
       const metricHistory = applyMetricHistoryFromStatus(payload);
diff --git a/studio/frontend/src/features/training/types/runtime.ts b/studio/frontend/src/features/training/types/runtime.ts
index 1bf319a5d1..3e4dbd3b79 100644
--- a/studio/frontend/src/features/training/types/runtime.ts
+++ b/studio/frontend/src/features/training/types/runtime.ts
@@ -99,6 +99,7 @@ export interface TrainingRuntimeState {
   evalLossHistory: TrainingSeriesPoint[];
   resetGeneration: number;
   stopRequested: boolean;
+  selectedHistoryRunId: string | null;
 }
 
 export interface TrainingRuntimeActions {
@@ -115,6 +116,7 @@ export interface TrainingRuntimeActions {
   applyProgress: (payload: TrainingProgressPayload, eventId?: number) => void;
   setStartQueued: (jobId: string, message: string) => void;
   setRuntimeError: (message: string) => void;
+  setSelectedHistoryRunId: (id: string | null) => void;
 }
 
 export type TrainingRuntimeStore = TrainingRuntimeState & TrainingRuntimeActions;
diff --git a/studio/frontend/src/hooks/use-mobile.ts b/studio/frontend/src/hooks/use-mobile.ts
index 9f505b5f3d..576d2d69da 100644
--- a/studio/frontend/src/hooks/use-mobile.ts
+++ b/studio/frontend/src/hooks/use-mobile.ts
@@ -1,22 +1,23 @@
 // SPDX-License-Identifier: AGPL-3.0-only
 // Copyright 2026-present the Unsloth AI Inc. team. All rights reserved. See /studio/LICENSE.AGPL-3.0
 
-import { useEffect, useState } from "react";
+import { useSyncExternalStore } from "react";
 
 const MOBILE_BREAKPOINT = 768;
+const MEDIA_QUERY = `(max-width: ${MOBILE_BREAKPOINT - 1}px)`;
 
-export function useIsMobile() {
-  const [isMobile, setIsMobile] = useState<boolean | undefined>(undefined);
-
-  useEffect(() => {
-    const mql = window.matchMedia(`(max-width: ${MOBILE_BREAKPOINT - 1}px)`);
-    const onChange = () => {
-      setIsMobile(window.innerWidth < MOBILE_BREAKPOINT);
-    };
-    mql.addEventListener("change", onChange);
-    setIsMobile(window.innerWidth < MOBILE_BREAKPOINT);
-    return () => mql.removeEventListener("change", onChange);
-  }, []);
-
-  return !!isMobile;
+function getSnapshot(): boolean {
+  if (typeof window === "undefined") return false;
+  return window.matchMedia(MEDIA_QUERY).matches;
+}
+
+function subscribe(callback: () => void): () => void {
+  if (typeof window === "undefined") return () => {};
+  const mql = window.matchMedia(MEDIA_QUERY);
+  mql.addEventListener("change", callback);
+  return () => mql.removeEventListener("change", callback);
+}
+
+export function useIsMobile(): boolean {
+  return useSyncExternalStore(subscribe, getSnapshot, () => false);
 }
diff --git a/studio/frontend/src/hooks/use-sidebar-pin.ts b/studio/frontend/src/hooks/use-sidebar-pin.ts
new file mode 100644
index 0000000000..71c780d9a4
--- /dev/null
+++ b/studio/frontend/src/hooks/use-sidebar-pin.ts
@@ -0,0 +1,59 @@
+// SPDX-License-Identifier: AGPL-3.0-only
+// Copyright 2026-present the Unsloth AI Inc. team. All rights reserved. See /studio/LICENSE.AGPL-3.0
+
+import { useCallback, useSyncExternalStore } from "react";
+
+const PINNED_KEY = "sidebar_pinned";
+
+function loadPinned(): boolean {
+  if (typeof window === "undefined") return true;
+  try {
+    const raw = window.localStorage.getItem(PINNED_KEY);
+    if (raw === null) return true;
+    return raw === "true";
+  } catch {
+    return true;
+  }
+}
+
+let pinnedValue = loadPinned();
+const listeners = new Set<() => void>();
+
+function subscribe(cb: () => void) {
+  listeners.add(cb);
+  if (typeof window === "undefined") {
+    return () => listeners.delete(cb);
+  }
+  const onStorage = (e: StorageEvent) => {
+    if (e.key === PINNED_KEY || e.key === null) {
+      pinnedValue = loadPinned();
+      cb();
+    }
+  };
+  window.addEventListener("storage", onStorage);
+  return () => {
+    listeners.delete(cb);
+    window.removeEventListener("storage", onStorage);
+  };
+}
+
+function setPinnedGlobal(next: boolean) {
+  pinnedValue = next;
+  try {
+    window.localStorage.setItem(PINNED_KEY, String(next));
+  } catch {}
+  listeners.forEach((cb) => cb());
+}
+
+export function useSidebarPin() {
+  const pinned = useSyncExternalStore(
+    subscribe,
+    () => pinnedValue,
+    () => false,
+  );
+
+  const setPinned = useCallback((value: boolean) => setPinnedGlobal(value), []);
+  const togglePinned = useCallback(() => setPinnedGlobal(!pinnedValue), []);
+
+  return { pinned, setPinned, togglePinned };
+}
diff --git a/studio/frontend/src/index.css b/studio/frontend/src/index.css
index d051b24245..8ee4f9eca6 100644
--- a/studio/frontend/src/index.css
+++ b/studio/frontend/src/index.css
@@ -15,439 +15,501 @@
 @custom-variant dark (&:is(.dark *));
 
 @font-face {
-  font-family: "Hellix";
-  src: url("/fonts/Hellix-SemiBold.woff2") format("woff2"),
-    url("/fonts/Hellix-SemiBold.woff") format("woff");
-  font-weight: 600;
-  font-style: normal;
-  font-display: swap;
+	font-family: "Hellix";
+	src: url("/fonts/Hellix-SemiBold.woff2") format("woff2"),
+		url("/fonts/Hellix-SemiBold.woff") format("woff");
+	font-weight: 600;
+	font-style: normal;
+	font-display: swap;
 }
 
 :root {
-  /* Animation timing */
-  --duration-micro: 100ms;
-  --duration-fast: 150ms;
-  --duration-normal: 200ms;
+	/* Animation timing */
+	--duration-micro: 100ms;
+	--duration-fast: 150ms;
+	--duration-normal: 200ms;
 
-  /* Easing curves (Emil Kowalski) */
-  --ease-out-quart: cubic-bezier(0.165, 0.84, 0.44, 1);
-  --ease-out-cubic: cubic-bezier(0.215, 0.61, 0.355, 1);
+	/* Easing curves (Emil Kowalski) */
+	--ease-out-quart: cubic-bezier(0.165, 0.84, 0.44, 1);
+	--ease-out-cubic: cubic-bezier(0.215, 0.61, 0.355, 1);
 
-  --background: oklch(1 0 0);
-  --foreground: oklch(0.2686 0 0);
-  --card: oklch(1 0 0);
-  --card-foreground: oklch(0.1281 0.0179 169.2764);
-  --popover: oklch(1 0 0);
-  --popover-foreground: oklch(0.1281 0.0179 169.2764);
-  --primary: oklch(0.6929 0.1396 166.5513);
-  --primary-foreground: oklch(1 0 0);
-  --secondary: oklch(0.9596 0.0275 167.8295);
-  --secondary-foreground: oklch(0.2868 0.0649 159.9823);
-  --muted: oklch(0.9702 0 0);
-  --muted-foreground: oklch(0.5486 0 0);
-  --accent: oklch(0.9596 0.0275 167.8295);
-  --accent-foreground: oklch(0.2868 0.0649 159.9823);
-  --destructive: oklch(0.6368 0.2078 25.3313);
-  --border: oklch(0.9208 0.0101 164.8536);
-  --input: oklch(0.9208 0.0101 164.8536);
-  --ring: oklch(0.6929 0.1396 166.5513);
-  --chart-1: oklch(0.6929 0.1396 166.5513);
-  --chart-2: oklch(0.694 0.1395 136.6059);
-  --chart-3: oklch(0.7014 0.1193 197.5897);
-  --chart-4: oklch(0.6926 0.1112 346.5775);
-  --chart-5: oklch(0.7497 0.1003 85.0057);
-  --radius: 1.2rem;
-  --sidebar: oklch(0.975 0 0);
-  --sidebar-foreground: oklch(0.1281 0.0179 169.2764);
-  --sidebar-primary: oklch(0.6929 0.1396 166.5513);
-  --sidebar-primary-foreground: oklch(1 0 0);
-  --sidebar-accent: oklch(0.96 0.0279 166.55);
-  --sidebar-accent-foreground: oklch(0.2868 0.0649 159.9823);
-  --sidebar-border: oklch(0.9208 0.0101 164.8536);
-  --sidebar-ring: oklch(0.6929 0.1396 166.5513);
-  --destructive-foreground: oklch(1 0 0);
-  --font-sans: "Inter Variable", ui-sans-serif, sans-serif, system-ui;
-  --font-heading: "Hellix", "Space Grotesk Variable", var(--font-sans);
-  --font-serif: Source Serif 4, serif;
-  --font-mono: JetBrains Mono, monospace;
-  --shadow-color: hsl(0 0% 0%);
-  --shadow-opacity: 0;
-  --shadow-blur: 0px;
-  --shadow-spread: 0px;
-  --shadow-offset-x: 0px;
-  --shadow-offset-y: 0px;
-  --letter-spacing: 0em;
-  --spacing: 0.25rem;
-  /*--shadow-2xs: 0px 0px 0px 0px hsl(0 0% 0% / 0);*/
-  /*--shadow-xs: 0px 0px 0px 0px hsl(0 0% 0% / 0);*/
-  /*--shadow-sm:*/
-  /*    0px 0px 0px 0px hsl(0 0% 0% / 0),*/
-  /*    0px 1px 2px 0px hsl(0 0% 0% / 0);*/
-  /*--shadow:*/
-  /*    0px 0px 0px 0px hsl(0 0% 0% / 0),*/
-  /*    0px 1px 2px 0px hsl(0 0% 0% / 0);*/
-  /*--shadow-md:*/
-  /*    0px 0px 0px 0px hsl(0 0% 0% / 0),*/
-  /*    0px 2px 4px 0px hsl(0 0% 0% / 0);*/
-  /*--shadow-lg:*/
-  /*    0px 0px 0px 0px hsl(0 0% 0% / 0),*/
-  /*    0px 4px 6px 0px hsl(0 0% 0% / 0);*/
-  /*--shadow-xl:*/
-  /*    0px 0px 0px 0px hsl(0 0% 0% / 0),*/
-  /*    0px 8px 10px 0px hsl(0 0% 0% / 0);*/
-  /*--shadow-2xl: 0px 0px 0px 0px hsl(0 0% 0% / 0);*/
-  --tracking-normal: 0em;
+	--background: oklch(1 0 0);
+	--foreground: oklch(0.2686 0 0);
+	--card: oklch(1 0 0);
+	--card-foreground: oklch(0.1281 0.0179 169.2764);
+	--popover: oklch(1 0 0);
+	--popover-foreground: oklch(0.1281 0.0179 169.2764);
+	--primary: oklch(0.6929 0.1396 166.5513);
+	--primary-foreground: oklch(1 0 0);
+	--secondary: oklch(0.9596 0.0275 167.8295);
+	--secondary-foreground: oklch(0.2868 0.0649 159.9823);
+	--muted: oklch(0.9702 0 0);
+	--muted-foreground: oklch(0.5486 0 0);
+	--accent: oklch(0.9596 0.0275 167.8295);
+	--accent-foreground: oklch(0.2868 0.0649 159.9823);
+	--destructive: oklch(0.6368 0.2078 25.3313);
+	--border: oklch(0.9208 0.0101 164.8536);
+	--input: oklch(0.9208 0.0101 164.8536);
+	--ring: oklch(0.6929 0.1396 166.5513);
+	--chart-1: oklch(0.6929 0.1396 166.5513);
+	--chart-2: oklch(0.694 0.1395 136.6059);
+	--chart-3: oklch(0.7014 0.1193 197.5897);
+	--chart-4: oklch(0.6926 0.1112 346.5775);
+	--chart-5: oklch(0.7497 0.1003 85.0057);
+	--radius: 0.625rem;
+	--sidebar: oklch(0.99 0 0);
+	--sidebar-foreground: oklch(0.1281 0.0179 169.2764);
+	--sidebar-primary: oklch(0.6929 0.1396 166.5513);
+	--sidebar-primary-foreground: oklch(1 0 0);
+	--sidebar-accent: oklch(0.96 0.0279 166.55);
+	--sidebar-accent-foreground: oklch(0.2868 0.0649 159.9823);
+	--sidebar-border: oklch(0.9208 0.0101 164.8536);
+	--sidebar-ring: oklch(0.6929 0.1396 166.5513);
+	--destructive-foreground: oklch(1 0 0);
+	--font-sans: "Inter Variable", ui-sans-serif, sans-serif, system-ui;
+	--font-heading: "Hellix", "Space Grotesk Variable", var(--font-sans);
+	--font-serif: Source Serif 4, serif;
+	--font-mono: JetBrains Mono, monospace;
+	--shadow-color: hsl(0 0% 0%);
+	--shadow-opacity: 0;
+	--shadow-blur: 0px;
+	--shadow-spread: 0px;
+	--shadow-offset-x: 0px;
+	--shadow-offset-y: 0px;
+	--letter-spacing: 0em;
+	--spacing: 0.25rem;
+	/*--shadow-2xs: 0px 0px 0px 0px hsl(0 0% 0% / 0);*/
+	/*--shadow-xs: 0px 0px 0px 0px hsl(0 0% 0% / 0);*/
+	/*--shadow-sm:*/
+	/*    0px 0px 0px 0px hsl(0 0% 0% / 0),*/
+	/*    0px 1px 2px 0px hsl(0 0% 0% / 0);*/
+	/*--shadow:*/
+	/*    0px 0px 0px 0px hsl(0 0% 0% / 0),*/
+	/*    0px 1px 2px 0px hsl(0 0% 0% / 0);*/
+	/*--shadow-md:*/
+	/*    0px 0px 0px 0px hsl(0 0% 0% / 0),*/
+	/*    0px 2px 4px 0px hsl(0 0% 0% / 0);*/
+	/*--shadow-lg:*/
+	/*    0px 0px 0px 0px hsl(0 0% 0% / 0),*/
+	/*    0px 4px 6px 0px hsl(0 0% 0% / 0);*/
+	/*--shadow-xl:*/
+	/*    0px 0px 0px 0px hsl(0 0% 0% / 0),*/
+	/*    0px 8px 10px 0px hsl(0 0% 0% / 0);*/
+	/*--shadow-2xl: 0px 0px 0px 0px hsl(0 0% 0% / 0);*/
+	--tracking-normal: -0.01em;
 }
+
 .dark {
-  --background: oklch(0.24 0 0);
-  --foreground: oklch(0.98 0 0);
-  --card: oklch(0.28 0 0);
-  --card-foreground: oklch(0.98 0 0);
-  --popover: oklch(0.28 0 0);
-  --popover-foreground: oklch(0.98 0 0);
-  --primary: oklch(0.6929 0.1396 166.5513);
-  --primary-foreground: oklch(1 0 0);
-  --secondary: oklch(0.33 0 0);
-  --secondary-foreground: oklch(0.98 0 0);
-  --muted: oklch(0.33 0 0);
-  --muted-foreground: oklch(0.70 0 0);
-  --accent: oklch(0.33 0 0);
-  --accent-foreground: oklch(0.98 0 0);
-  --destructive: oklch(0.6368 0.2078 25.3313);
-  --border: oklch(0.38 0 0);
-  --input: oklch(0.38 0 0);
-  --ring: oklch(0.6929 0.1396 166.5513);
-  --chart-1: oklch(0.7511 0.1407 166.2284);
-  --chart-2: oklch(0.75 0.14 136.5572);
-  --chart-3: oklch(0.7554 0.1285 197.339);
-  --chart-4: oklch(0.7503 0.1199 346.7805);
-  --chart-5: oklch(0.799 0.1196 84.6633);
-  --sidebar: oklch(0.24 0 0);
-  --sidebar-foreground: oklch(0.98 0 0);
-  --sidebar-primary: oklch(0.6929 0.1396 166.5513);
-  --sidebar-primary-foreground: oklch(1 0 0);
-  --sidebar-accent: oklch(0.33 0 0);
-  --sidebar-accent-foreground: oklch(0.98 0 0);
-  --sidebar-border: oklch(0.38 0 0);
-  --sidebar-ring: oklch(0.6929 0.1396 166.5513);
-  --destructive-foreground: oklch(1 0 0);
-  --radius: 1.2rem;
-  --font-sans: Geist, ui-sans-serif, sans-serif, system-ui;
-  --font-serif: Source Serif 4, serif;
-  --font-mono: JetBrains Mono, monospace;
-  --shadow-color: hsl(0 0% 0%);
-  --shadow-opacity: 0;
-  --shadow-blur: 0px;
-  --shadow-spread: 0px;
-  --shadow-offset-x: 0px;
-  --shadow-offset-y: 0px;
-  --letter-spacing: 0em;
-  --spacing: 0.25rem;
-  --shadow-2xs: 0px 0px 0px 0px hsl(0 0% 0% / 0);
-  --shadow-xs: 0px 0px 0px 0px hsl(0 0% 0% / 0);
-  --shadow-sm: 0px 0px 0px 0px hsl(0 0% 0% / 0), 0px 1px 2px 0px
-  hsl(0 0% 0% / 0);
-  --shadow: 0px 0px 0px 0px hsl(0 0% 0% / 0), 0px 1px 2px 0px hsl(0 0% 0% / 0);
-  --shadow-md: 0px 0px 0px 0px hsl(0 0% 0% / 0), 0px 2px 4px 0px
-  hsl(0 0% 0% / 0);
-  --shadow-lg: 0px 0px 0px 0px hsl(0 0% 0% / 0), 0px 4px 6px 0px
-  hsl(0 0% 0% / 0);
-  --shadow-xl: 0px 0px 0px 0px hsl(0 0% 0% / 0), 0px 8px 10px 0px
-  hsl(0 0% 0% / 0);
-  --shadow-2xl: 0px 0px 0px 0px hsl(0 0% 0% / 0);
+	--background: oklch(0.24 0 0);
+	--foreground: oklch(0.98 0 0);
+	--card: oklch(0.28 0 0);
+	--card-foreground: oklch(0.98 0 0);
+	--popover: oklch(0.28 0 0);
+	--popover-foreground: oklch(0.98 0 0);
+	--primary: oklch(0.6929 0.1396 166.5513);
+	--primary-foreground: oklch(1 0 0);
+	--secondary: oklch(0.33 0 0);
+	--secondary-foreground: oklch(0.98 0 0);
+	--muted: oklch(0.33 0 0);
+	--muted-foreground: oklch(0.70 0 0);
+	--accent: oklch(0.33 0 0);
+	--accent-foreground: oklch(0.98 0 0);
+	--destructive: oklch(0.6368 0.2078 25.3313);
+	--border: oklch(0.38 0 0);
+	--input: oklch(0.38 0 0);
+	--ring: oklch(0.6929 0.1396 166.5513);
+	--chart-1: oklch(0.7511 0.1407 166.2284);
+	--chart-2: oklch(0.75 0.14 136.5572);
+	--chart-3: oklch(0.7554 0.1285 197.339);
+	--chart-4: oklch(0.7503 0.1199 346.7805);
+	--chart-5: oklch(0.799 0.1196 84.6633);
+	--sidebar: oklch(0.24 0 0);
+	--sidebar-foreground: oklch(0.98 0 0);
+	--sidebar-primary: oklch(0.6929 0.1396 166.5513);
+	--sidebar-primary-foreground: oklch(1 0 0);
+	--sidebar-accent: oklch(0.33 0 0);
+	--sidebar-accent-foreground: oklch(0.98 0 0);
+	--sidebar-border: oklch(0.38 0 0);
+	--sidebar-ring: oklch(0.6929 0.1396 166.5513);
+	--destructive-foreground: oklch(1 0 0);
+	--radius: 0.625rem;
+	--font-sans: Geist, ui-sans-serif, sans-serif, system-ui;
+	--font-serif: Source Serif 4, serif;
+	--font-mono: JetBrains Mono, monospace;
+	--shadow-color: hsl(0 0% 0%);
+	--shadow-opacity: 0;
+	--shadow-blur: 0px;
+	--shadow-spread: 0px;
+	--shadow-offset-x: 0px;
+	--shadow-offset-y: 0px;
+	--letter-spacing: 0em;
+	--spacing: 0.25rem;
+	--shadow-2xs: 0px 0px 0px 0px hsl(0 0% 0% / 0);
+	--shadow-xs: 0px 0px 0px 0px hsl(0 0% 0% / 0);
+	--shadow-sm: 0px 0px 0px 0px hsl(0 0% 0% / 0), 0px 1px 2px 0px hsl(0 0% 0% / 0);
+	--shadow: 0px 0px 0px 0px hsl(0 0% 0% / 0), 0px 1px 2px 0px hsl(0 0% 0% / 0);
+	--shadow-md: 0px 0px 0px 0px hsl(0 0% 0% / 0), 0px 2px 4px 0px hsl(0 0% 0% / 0);
+	--shadow-lg: 0px 0px 0px 0px hsl(0 0% 0% / 0), 0px 4px 6px 0px hsl(0 0% 0% / 0);
+	--shadow-xl: 0px 0px 0px 0px hsl(0 0% 0% / 0), 0px 8px 10px 0px hsl(0 0% 0% / 0);
+	--shadow-2xl: 0px 0px 0px 0px hsl(0 0% 0% / 0);
 }
 
 @theme inline {
-  --font-sans: "Inter Variable", ui-sans-serif, sans-serif, system-ui;
-  --font-heading: "Hellix", "Space Grotesk Variable", var(--font-sans);
-  --color-sidebar-ring: var(--sidebar-ring);
-  --color-sidebar-border: var(--sidebar-border);
-  --color-sidebar-accent-foreground: var(--sidebar-accent-foreground);
-  --color-sidebar-accent: var(--sidebar-accent);
-  --color-sidebar-primary-foreground: var(--sidebar-primary-foreground);
-  --color-sidebar-primary: var(--sidebar-primary);
-  --color-sidebar-foreground: var(--sidebar-foreground);
-  --color-sidebar: var(--sidebar);
-  --color-chart-5: var(--chart-5);
-  --color-chart-4: var(--chart-4);
-  --color-chart-3: var(--chart-3);
-  --color-chart-2: var(--chart-2);
-  --color-chart-1: var(--chart-1);
-  --color-code-block: #181818;
-  --color-ring: var(--ring);
-  --color-input: var(--input);
-  --color-border: var(--border);
-  --color-destructive: var(--destructive);
-  --color-accent-foreground: var(--accent-foreground);
-  --color-accent: var(--accent);
-  --color-muted-foreground: var(--muted-foreground);
-  --color-muted: var(--muted);
-  --color-secondary-foreground: var(--secondary-foreground);
-  --color-secondary: var(--secondary);
-  --color-primary-foreground: var(--primary-foreground);
-  --color-primary: var(--primary);
-  --color-popover-foreground: var(--popover-foreground);
-  --color-popover: var(--popover);
-  --color-card-foreground: var(--card-foreground);
-  --color-card: var(--card);
-  --color-foreground: var(--foreground);
-  --color-background: var(--background);
-  --radius-sm: calc(var(--radius) - 4px);
-  --radius-md: calc(var(--radius) - 2px);
-  --radius-lg: var(--radius);
-  --radius-xl: calc(var(--radius) + 4px);
-  --radius-2xl: calc(var(--radius) + 8px);
-  --radius-3xl: calc(var(--radius) + 12px);
-  --radius-4xl: calc(var(--radius) + 16px);
-  --font-mono: JetBrains Mono, monospace;
-  --font-serif: Source Serif 4, serif;
-  --radius: 1.2rem;
-  --tracking-tighter: calc(var(--tracking-normal) - 0.05em);
-  --tracking-tight: calc(var(--tracking-normal) - 0.025em);
-  --tracking-wide: calc(var(--tracking-normal) + 0.025em);
-  --tracking-wider: calc(var(--tracking-normal) + 0.05em);
-  --tracking-widest: calc(var(--tracking-normal) + 0.1em);
-  --tracking-normal: var(--tracking-normal);
-  /*--shadow-2xl: var(--shadow-2xl);*/
-  /*--shadow-xl: var(--shadow-xl);*/
-  /*--shadow-lg: var(--shadow-lg);*/
-  /*--shadow-md: var(--shadow-md);*/
-  /*--shadow: var(--shadow);*/
-  /*--shadow-sm: var(--shadow-sm);*/
-  /*--shadow-xs: var(--shadow-xs);*/
-  /*--shadow-2xs: var(--shadow-2xs);*/
-  /*--spacing: var(--spacing);*/
-  /*--letter-spacing: var(--letter-spacing);*/
-  /*--shadow-offset-y: var(--shadow-offset-y);*/
-  /*--shadow-offset-x: var(--shadow-offset-x);*/
-  /*--shadow-spread: var(--shadow-spread);*/
-  /*--shadow-blur: var(--shadow-blur);*/
-  /*--shadow-opacity: var(--shadow-opacity);*/
-  /*--color-shadow-color: var(--shadow-color);*/
-  --color-destructive-foreground: var(--destructive-foreground);
-  --animate-pulse: pulse var(--duration) ease-out infinite;
-  @keyframes pulse {
-    0%,
-    100% {
-      box-shadow: 0 0 0 0 var(--pulse-color);
-    }
-    50% {
-      box-shadow: 0 0 0 8px var(--pulse-color);
-    }
-  }
-  --animate-shiny-text: shiny-text 8s infinite;
-  @keyframes shiny-text {
-    0%,
-    90%,
-    100% {
-      background-position: calc(-100% - var(--shiny-width)) 0;
-    }
-    30%,
-    60% {
-      background-position: calc(100% + var(--shiny-width)) 0;
-    }
-  }
-  --animate-shine: shine var(--duration) infinite linear
-;
-  @keyframes shine {
-  0% {
-    background-position: 0% 0%;
-    }
-  50% {
-    background-position: 100% 100%;
-    }
-  to {
-    background-position: 0% 0%;
-    }
-  }}
+	--font-sans: "Inter Variable", ui-sans-serif, sans-serif, system-ui;
+	--font-heading: "Hellix", "Space Grotesk Variable", var(--font-sans);
+	--color-sidebar-ring: var(--sidebar-ring);
+	--color-sidebar-border: var(--sidebar-border);
+	--color-sidebar-accent-foreground: var(--sidebar-accent-foreground);
+	--color-sidebar-accent: var(--sidebar-accent);
+	--color-sidebar-primary-foreground: var(--sidebar-primary-foreground);
+	--color-sidebar-primary: var(--sidebar-primary);
+	--color-sidebar-foreground: var(--sidebar-foreground);
+	--color-sidebar: var(--sidebar);
+	--color-chart-5: var(--chart-5);
+	--color-chart-4: var(--chart-4);
+	--color-chart-3: var(--chart-3);
+	--color-chart-2: var(--chart-2);
+	--color-chart-1: var(--chart-1);
+	--color-code-block: #181818;
+	--color-ring: var(--ring);
+	--color-input: var(--input);
+	--color-border: var(--border);
+	--color-destructive: var(--destructive);
+	--color-accent-foreground: var(--accent-foreground);
+	--color-accent: var(--accent);
+	--color-muted-foreground: var(--muted-foreground);
+	--color-muted: var(--muted);
+	--color-secondary-foreground: var(--secondary-foreground);
+	--color-secondary: var(--secondary);
+	--color-primary-foreground: var(--primary-foreground);
+	--color-primary: var(--primary);
+	--color-popover-foreground: var(--popover-foreground);
+	--color-popover: var(--popover);
+	--color-card-foreground: var(--card-foreground);
+	--color-card: var(--card);
+	--color-foreground: var(--foreground);
+	--color-background: var(--background);
+	--radius-sm: calc(var(--radius) - 4px);
+	--radius-md: calc(var(--radius) - 2px);
+	--radius-lg: var(--radius);
+	--radius-xl: calc(var(--radius) + 4px);
+	--radius-2xl: calc(var(--radius) + 8px);
+	--radius-3xl: calc(var(--radius) + 12px);
+	--radius-4xl: calc(var(--radius) + 16px);
+	--font-mono: JetBrains Mono, monospace;
+	--font-serif: Source Serif 4, serif;
+	--radius: 0.625rem;
+	--tracking-tighter: calc(var(--tracking-normal) - 0.05em);
+	--tracking-tight: calc(var(--tracking-normal) - 0.025em);
+	--tracking-wide: calc(var(--tracking-normal) + 0.025em);
+	--tracking-wider: calc(var(--tracking-normal) + 0.05em);
+	--tracking-widest: calc(var(--tracking-normal) + 0.1em);
+	--tracking-normal: var(--tracking-normal);
+	/*--shadow-2xl: var(--shadow-2xl);*/
+	/*--shadow-xl: var(--shadow-xl);*/
+	/*--shadow-lg: var(--shadow-lg);*/
+	/*--shadow-md: var(--shadow-md);*/
+	/*--shadow: var(--shadow);*/
+	/*--shadow-sm: var(--shadow-sm);*/
+	/*--shadow-xs: var(--shadow-xs);*/
+	/*--shadow-2xs: var(--shadow-2xs);*/
+	/*--spacing: var(--spacing);*/
+	/*--letter-spacing: var(--letter-spacing);*/
+	/*--shadow-offset-y: var(--shadow-offset-y);*/
+	/*--shadow-offset-x: var(--shadow-offset-x);*/
+	/*--shadow-spread: var(--shadow-spread);*/
+	/*--shadow-blur: var(--shadow-blur);*/
+	/*--shadow-opacity: var(--shadow-opacity);*/
+	/*--color-shadow-color: var(--shadow-color);*/
+	--color-destructive-foreground: var(--destructive-foreground);
+	--animate-pulse: pulse var(--duration) ease-out infinite;
+
+	@keyframes pulse {
+
+		0%,
+		100% {
+			box-shadow: 0 0 0 0 var(--pulse-color);
+		}
+
+		50% {
+			box-shadow: 0 0 0 8px var(--pulse-color);
+		}
+	}
+
+	--animate-shiny-text: shiny-text 8s infinite;
+
+	@keyframes shiny-text {
+
+		0%,
+		90%,
+		100% {
+			background-position: calc(-100% - var(--shiny-width)) 0;
+		}
+
+		30%,
+		60% {
+			background-position: calc(100% + var(--shiny-width)) 0;
+		}
+	}
+
+	--animate-shine: shine var(--duration) infinite linear;
+
+	@keyframes shine {
+		0% {
+			background-position: 0% 0%;
+		}
+
+		50% {
+			background-position: 100% 100%;
+		}
+
+		to {
+			background-position: 0% 0%;
+		}
+	}
+}
 
 @layer base {
-  * {
-    @apply border-border outline-ring/50;
-  }
-  body {
-    @apply font-sans bg-background text-foreground;
-    letter-spacing: var(--tracking-normal);
-  }
-  html {
-    @apply font-sans;
-    scrollbar-gutter: stable;
-  }
-  body[data-scroll-locked] {
-    margin-right: 0 !important;
-  }
-  h1,
-  h2,
-  h3,
-  h4,
-  h5,
-  h6 {
-    font-family: var(--font-heading);
-  }
-  .font-medium,
-  .font-semibold,
-  .font-bold {
-    font-family: var(--font-heading);
-  }
+	* {
+		@apply border-border outline-ring/50;
+	}
+
+	body {
+		@apply font-sans bg-background text-foreground;
+		letter-spacing: var(--tracking-normal);
+	}
+
+	html {
+		@apply font-sans h-full;
+	}
+
+	body,
+	#root {
+		@apply h-full;
+	}
+
+	body[data-scroll-locked] {
+		margin-right: 0 !important;
+	}
+
+	h1,
+	h2,
+	h3,
+	h4,
+	h5,
+	h6 {
+		font-family: var(--font-sans);
+		letter-spacing: -0.02em;
+	}
 }
 
 @layer utilities {
-  /* Heading font utility */
-  .font-heading {
-    font-family: var(--font-heading);
-  }
 
-  /* Elevated surface shadow (use ring-* for borders) */
-  .shadow-border {
-    --tw-shadow: 0 4px 16px rgba(0, 0, 0, 0.1);
-    --tw-shadow-colored: 0 4px 16px var(--tw-shadow-color);
-    box-shadow: var(--tw-ring-offset-shadow, 0 0 #0000),
-      var(--tw-ring-shadow, 0 0 #0000), var(--tw-shadow);
-  }
-  .dark .shadow-border {
-    --tw-shadow: 0 4px 16px rgba(0, 0, 0, 0.3);
-  }
+	/* Heading font utility */
+	.font-heading {
+		font-family: var(--font-heading);
+	}
 
-  /* Fine-tuning Studio: equal default height, expandable when needed (md+) */
-  .min-h-studio-config-column {
-    @apply md:min-h-[470px];
-  }
-  .h-studio-config-column {
-    @apply md:h-[470px];
-  }
+	/* Elevated surface shadow (use ring-* for borders) */
+	.shadow-border {
+		--tw-shadow: 0 4px 16px rgba(0, 0, 0, 0.1);
+		--tw-shadow-colored: 0 4px 16px var(--tw-shadow-color);
+		box-shadow: var(--tw-ring-offset-shadow, 0 0 #0000),
+			var(--tw-ring-shadow, 0 0 #0000), var(--tw-shadow);
+	}
 
-  [data-streamdown="unordered-list"] {
-    list-style-type: disc;
-    list-style-position: outside;
-    padding-left: 1.25rem;
-    margin-block: 0.5rem;
-  }
+	.dark .shadow-border {
+		--tw-shadow: 0 4px 16px rgba(0, 0, 0, 0.3);
+	}
 
-  [data-streamdown="ordered-list"] {
-    list-style-type: decimal;
-    list-style-position: outside;
-    padding-left: 1.25rem;
-    margin-block: 0.5rem;
-  }
+	.chat-composer-surface {
+		border: 1px solid oklch(0.93 0 0 / 1);
+		background-clip: padding-box;
+		box-shadow:
+			0 1px 2px oklch(0 0 0 / 0.04),
+			0 6px 14px oklch(0 0 0 / 0.05),
+			0 18px 40px oklch(0 0 0 / 0.05);
+	}
 
-  [data-streamdown="list-item"] {
-    display: list-item;
-  }
+	.dark .chat-composer-surface {
+		border-color: oklch(0.38 0 0 / 1);
+		box-shadow:
+			0 1px 2px oklch(0 0 0 / 0.2),
+			0 8px 20px oklch(0 0 0 / 0.25),
+			0 22px 48px oklch(0 0 0 / 0.22);
+	}
 
-  /* Flatten code blocks: single border, language label, then code directly */
-  [data-streamdown="code-block-body"] {
-    border: none !important;
-    border-radius: 0 !important;
-    background: transparent !important;
-    padding: 0 !important;
-  }
-  [data-streamdown="code-block"] {
-    gap: 0;
-    padding: 0.5rem;
-    /* Wide lines must scroll inside the thread column, not widen past the composer (flex min-width:auto). */
-    max-width: 100%;
-    min-width: 0;
-    overflow-x: auto;
-  }
-  [data-streamdown="code-block-header"] {
-    padding-left: 0.75rem;
-  }
+	/* Fine-tuning Studio: equal default height, expandable when needed (md+) */
+	.min-h-studio-config-column {
+		@apply md:min-h-[470px];
+	}
 
-  /* Chat thread: code slightly smaller by default; step up when the thread column is wide. */
-  .aui-thread-root [data-streamdown="code-block"] {
-    font-size: 0.8125rem;
-    line-height: 1.55;
-  }
+	.h-studio-config-column {
+		@apply md:h-[470px];
+	}
 
-  .aui-thread-root [data-streamdown="code-block-header"] {
-    font-size: 0.6875rem;
-  }
+	[data-streamdown="unordered-list"] {
+		list-style-type: disc;
+		list-style-position: outside;
+		padding-left: 1.25rem;
+		margin-block: 0.5rem;
+	}
 
-  @container (min-width: 36rem) {
-    .aui-thread-root [data-streamdown="code-block"] {
-      font-size: 0.875rem;
-    }
+	[data-streamdown="ordered-list"] {
+		list-style-type: decimal;
+		list-style-position: outside;
+		padding-left: 1.25rem;
+		margin-block: 0.5rem;
+	}
 
-    .aui-thread-root [data-streamdown="code-block-header"] {
-      font-size: 0.75rem;
-    }
-  }
+	[data-streamdown="list-item"] {
+		display: list-item;
+	}
 
-  /* Chat: use the app sans stack for UI + prose. */
-  .aui-thread-root {
-    --font-heading: var(--font-sans);
-    font-family: var(--font-sans);
-  }
+	/* Flatten code blocks: single border, language label, then code directly */
+	[data-streamdown="code-block-body"] {
+		border: none !important;
+		border-radius: 0 !important;
+		background: transparent !important;
+		padding: 0 !important;
+	}
 
-  /* Keep monospace for code fences and inline code (not KaTeX). */
-  .aui-thread-root [data-streamdown="code-block"] pre,
-  .aui-thread-root [data-streamdown="code-block"] code {
-    font-family: var(--font-mono), ui-monospace, monospace;
-  }
+	[data-streamdown="code-block"] {
+		gap: 0;
+		padding: 0.5rem;
+		/* Wide lines must scroll inside the thread column, not widen past the composer (flex min-width:auto). */
+		max-width: 100%;
+		min-width: 0;
+		overflow-x: auto;
+	}
 
-  .aui-thread-root :where(p, li, td, th, blockquote, h1, h2, h3, h4, h5, h6) code {
-    font-family: var(--font-mono), ui-monospace, monospace;
-  }
+	[data-streamdown="code-block-header"] {
+		padding-left: 0.75rem;
+	}
 
-  /* Align fenced code blocks with the main chat column even when nested in lists. */
-  .aui-thread-root [data-streamdown="list-item"] > [data-streamdown="code-block"],
-  .aui-thread-root [data-streamdown="list-item"] [data-streamdown="code-block"] {
-    margin-left: -1.25rem;
-    width: calc(100% + 1.25rem);
-    max-width: calc(100% + 1.25rem);
-  }
+	/* Chat thread: code slightly smaller by default; step up when the thread column is wide. */
+	.aui-thread-root [data-streamdown="code-block"] {
+		font-size: 0.8125rem;
+		line-height: 1.55;
+	}
 
-  .dark .aui-thread-root [data-streamdown="code-block"] {
-    /* Streamdown `pre` uses `dark:bg-[var(--shiki-dark-bg,...)]`; keep one surface on the outer shell. */
-    --shiki-dark-bg: transparent;
-    background: var(--color-code-block);
-    border: 1px solid oklch(1 0 0 / 0.07);
-  }
+	.aui-thread-root [data-streamdown="code-block-header"] {
+		font-size: 0.6875rem;
+	}
+
+	@container (min-width: 36rem) {
+		.aui-thread-root [data-streamdown="code-block"] {
+			font-size: 0.875rem;
+		}
+
+		.aui-thread-root [data-streamdown="code-block-header"] {
+			font-size: 0.75rem;
+		}
+	}
+
+	/* Chat: use the app sans stack for UI + prose. */
+	.aui-thread-root {
+		--font-heading: var(--font-sans);
+		font-family: var(--font-sans);
+	}
+
+	/* Keep monospace for code fences and inline code (not KaTeX). */
+	.aui-thread-root [data-streamdown="code-block"] pre,
+	.aui-thread-root [data-streamdown="code-block"] code {
+		font-family: var(--font-mono), ui-monospace, monospace;
+	}
+
+	.aui-thread-root :where(p, li, td, th, blockquote, h1, h2, h3, h4, h5, h6) code {
+		font-family: var(--font-mono), ui-monospace, monospace;
+	}
+
+	/* Align fenced code blocks with the main chat column even when nested in lists. */
+	.aui-thread-root [data-streamdown="list-item"] > [data-streamdown="code-block"],
+	.aui-thread-root [data-streamdown="list-item"] [data-streamdown="code-block"] {
+		margin-left: -1.25rem;
+		width: calc(100% + 1.25rem);
+		max-width: calc(100% + 1.25rem);
+	}
+
+	.dark .aui-thread-root [data-streamdown="code-block"] {
+		/* Streamdown `pre` uses `dark:bg-[var(--shiki-dark-bg,...)]`; keep one surface on the outer shell. */
+		--shiki-dark-bg: transparent;
+		background: var(--color-code-block);
+		border: 1px solid oklch(1 0 0 / 0.07);
+	}
 }
 
-/* Minimal scrollbar — thumb only, no track */
+/* Flat scrollbar chrome */
 * {
-  scrollbar-width: thin;
-  scrollbar-color: transparent transparent;
+	scrollbar-width: thin;
+	scrollbar-color: oklch(0.5 0 0 / 0.54) transparent;
 }
-*:hover {
-  scrollbar-color: oklch(0.6 0 0 / 0.3) transparent;
-}
-.dark *:hover {
-  scrollbar-color: oklch(0.5 0 0 / 0.35) transparent;
+
+.dark * {
+	scrollbar-color: oklch(0.67 0 0 / 0.5) transparent;
 }
 
 /* Webkit (Chrome, Safari, Edge) */
 ::-webkit-scrollbar {
-  width: 6px;
-  height: 6px;
+	width: 8px;
+	height: 8px;
 }
+
 ::-webkit-scrollbar-track {
-  background: transparent;
+	background: transparent;
 }
+
 ::-webkit-scrollbar-thumb {
-  background: transparent;
-  border-radius: 9999px;
+	background: oklch(0.5 0 0 / 0.54);
+	border-radius: 9999px;
 }
-*:hover::-webkit-scrollbar-thumb {
-  background: oklch(0.6 0 0 / 0.3);
+
+::-webkit-scrollbar-button {
+	display: none;
+	width: 0;
+	height: 0;
 }
-.dark *:hover::-webkit-scrollbar-thumb {
-  background: oklch(0.5 0 0 / 0.35);
+
+.dark *::-webkit-scrollbar-thumb {
+	background: oklch(0.67 0 0 / 0.5);
+}
+
+/* Chat viewport: solid track matching sidebar so the scrollbar reads as a
+   full-height rail flush to the right edge, without a separate decorative strip. */
+.aui-thread-viewport {
+	scrollbar-color: oklch(0.5 0 0 / 0.54) var(--sidebar);
+}
+
+.dark .aui-thread-viewport {
+	scrollbar-color: oklch(0.67 0 0 / 0.5) var(--sidebar);
+}
+
+.aui-thread-viewport::-webkit-scrollbar-track {
+	background: var(--sidebar);
+}
+
+[data-sidebar="content"] {
+	scrollbar-color: oklch(0.5 0 0 / 0.7) transparent;
+}
+
+.dark [data-sidebar="content"] {
+	scrollbar-color: oklch(0.72 0 0 / 0.65) transparent;
 }
 
 /*---break---*/
 
 @layer base {
-  * {
-    @apply border-border outline-ring/50;
-  }
-  body {
-    @apply bg-background text-foreground;
-  }
+	* {
+		@apply border-border outline-ring/50;
+	}
+
+	body {
+		@apply bg-background text-foreground;
+	}
 }
 
-::view-transition-old(root), ::view-transition-new(root) {
-    animation: none;
-    mix-blend-mode: normal;
-}
\ No newline at end of file
+::view-transition-old(root),
+::view-transition-new(root) {
+	animation: none;
+	mix-blend-mode: normal;
+}

From f4422b0a62d51015e832495f13016e6ec0217c71 Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?=E0=A4=B9=E0=A4=BF=E0=A4=AE=E0=A4=BE=E0=A4=82=E0=A4=B6?=
 =?UTF-8?q?=E0=A5=81?= <sharmahimanshu15082007@gmail.com>
Date: Thu, 16 Apr 2026 21:20:57 +0530
Subject: [PATCH 127/216] change torchcodec version to 0.10.0 in extra-no-deps
 (#5043)

Co-authored-by: Roland Tannous <115670425+rolandtannous@users.noreply.github.com>
---
 studio/backend/requirements/extras-no-deps.txt | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/studio/backend/requirements/extras-no-deps.txt b/studio/backend/requirements/extras-no-deps.txt
index a16571567d..23c61baa44 100644
--- a/studio/backend/requirements/extras-no-deps.txt
+++ b/studio/backend/requirements/extras-no-deps.txt
@@ -2,7 +2,7 @@
 descript-audio-codec
 descript-audiotools
 julius
-torchcodec
+torchcodec==0.10.0
 snac
 
 # peft 0.19.0 causes export subprocess shutdown issues in Studio;

From 7ef65bd2e5cfcba2e852eb946d3f57be854940de Mon Sep 17 00:00:00 2001
From: Wasim Yousef Said <wasimysdev@gmail.com>
Date: Thu, 16 Apr 2026 18:58:10 +0200
Subject: [PATCH 128/216] Chat first onboarding (#5063)

* auth: default to chat

* settings: relaunch onboarding

* onboarding: return to launch page

* studio: stop auto guided tour

* ui: soften global radius

* cleanup: rename onboarding exit prop

* fix onboarding redirect safety

* Show real Unsloth version in settings

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

---------

Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com>
---
 studio/backend/main.py                        | 25 ++++++++++++-
 studio/frontend/src/app/routes/onboarding.tsx |  5 +++
 studio/frontend/src/features/auth/session.ts  |  4 +--
 .../onboarding/components/splash-screen.tsx   |  6 ++--
 .../onboarding/components/wizard-footer.tsx   | 18 ++++++----
 .../onboarding/components/wizard-layout.tsx   | 26 +++++++++-----
 .../onboarding/components/wizard-sidebar.tsx  | 10 +++---
 .../src/features/settings/tabs/about-tab.tsx  | 29 ++++++++++++---
 .../features/settings/tabs/general-tab.tsx    | 36 +++++++++++++++++++
 .../src/features/studio/studio-page.tsx       |  4 ---
 studio/frontend/src/index.css                 |  6 ++--
 11 files changed, 131 insertions(+), 38 deletions(-)

diff --git a/studio/backend/main.py b/studio/backend/main.py
index 8a40791c06..d146a8ef12 100644
--- a/studio/backend/main.py
+++ b/studio/backend/main.py
@@ -27,6 +27,7 @@ import mimetypes
 import shutil
 import warnings
 from contextlib import asynccontextmanager
+from importlib.metadata import PackageNotFoundError, version as package_version
 
 # Fix broken Windows registry MIME types.  Some Windows installs map .js to
 # "text/plain" in the registry (HKCR\.js\Content Type).  Python's mimetypes
@@ -78,6 +79,27 @@ import utils.hardware.hardware as _hw_module
 from utils.cache_cleanup import clear_unsloth_compiled_cache
 
 
+def get_unsloth_version() -> str:
+    try:
+        return package_version("unsloth")
+    except PackageNotFoundError:
+        pass
+
+    version_file = (
+        _Path(__file__).resolve().parents[2] / "unsloth" / "models" / "_utils.py"
+    )
+    try:
+        for line in version_file.read_text(encoding = "utf-8").splitlines():
+            if line.startswith("__version__ = "):
+                return line.split("=", 1)[1].strip().strip('"').strip("'")
+    except OSError:
+        pass
+    return "dev"
+
+
+UNSLOTH_VERSION = get_unsloth_version()
+
+
 @asynccontextmanager
 async def lifespan(app: FastAPI):
     """Startup: detect hardware, seed default admin if needed. Shutdown: clean up compiled cache."""
@@ -140,7 +162,7 @@ async def lifespan(app: FastAPI):
 # Create FastAPI app
 app = FastAPI(
     title = "Unsloth UI Backend",
-    version = "1.0.0",
+    version = UNSLOTH_VERSION,
     description = "Backend API for Unsloth UI - Training and Model Management",
     lifespan = lifespan,
 )
@@ -198,6 +220,7 @@ async def health_check():
         "status": "healthy",
         "timestamp": datetime.now().isoformat(),
         "service": "Unsloth UI Backend",
+        "version": UNSLOTH_VERSION,
         "device_type": device_type,
         "chat_only": _hw_module.CHAT_ONLY,
     }
diff --git a/studio/frontend/src/app/routes/onboarding.tsx b/studio/frontend/src/app/routes/onboarding.tsx
index dcc3593b1a..8d1cd6ff5f 100644
--- a/studio/frontend/src/app/routes/onboarding.tsx
+++ b/studio/frontend/src/app/routes/onboarding.tsx
@@ -6,6 +6,8 @@ import { lazy } from "react";
 import { requireAuth } from "../auth-guards";
 import { Route as rootRoute } from "./__root";
 
+export type OnboardingSearch = { redirectTo?: string };
+
 const WizardLayout = lazy(() =>
   import("@/features/onboarding/components/wizard-layout").then((m) => ({
     default: m.WizardLayout,
@@ -16,5 +18,8 @@ export const Route = createRoute({
   getParentRoute: () => rootRoute,
   path: "/onboarding",
   beforeLoad: () => requireAuth(),
+  validateSearch: (search: Record<string, unknown>): OnboardingSearch => ({
+    redirectTo: typeof search.redirectTo === "string" ? search.redirectTo : undefined,
+  }),
   component: WizardLayout,
 });
diff --git a/studio/frontend/src/features/auth/session.ts b/studio/frontend/src/features/auth/session.ts
index 3d3502e073..6012174077 100644
--- a/studio/frontend/src/features/auth/session.ts
+++ b/studio/frontend/src/features/auth/session.ts
@@ -8,7 +8,7 @@ export const AUTH_REFRESH_TOKEN_KEY = "unsloth_auth_refresh_token";
 export const ONBOARDING_DONE_KEY = "unsloth_onboarding_done";
 export const AUTH_MUST_CHANGE_PASSWORD_KEY = "unsloth_auth_must_change_password";
 
-type PostAuthRoute = "/onboarding" | "/studio" | "/change-password" | "/chat";
+type PostAuthRoute = "/change-password" | "/chat";
 
 function canUseStorage(): boolean {
   return typeof window !== "undefined";
@@ -80,5 +80,5 @@ export function resetOnboardingDone(): void {
 export function getPostAuthRoute(): PostAuthRoute {
   if (mustChangePassword()) return "/change-password";
   if (usePlatformStore.getState().isChatOnly()) return "/chat";
-  return isOnboardingDone() ? "/studio" : "/onboarding";
+  return "/chat";
 }
diff --git a/studio/frontend/src/features/onboarding/components/splash-screen.tsx b/studio/frontend/src/features/onboarding/components/splash-screen.tsx
index 7f5e3619c9..ce828ee7b4 100644
--- a/studio/frontend/src/features/onboarding/components/splash-screen.tsx
+++ b/studio/frontend/src/features/onboarding/components/splash-screen.tsx
@@ -7,12 +7,12 @@ import { motion } from "motion/react";
 
 interface SplashScreenProps {
   onStartOnboarding: () => void;
-  onGoToStudio: () => void;
+  onSkipOnboarding: () => void;
 }
 
 export function SplashScreen({
   onStartOnboarding,
-  onGoToStudio,
+  onSkipOnboarding,
 }: SplashScreenProps) {
   return (
     <div className="fixed inset-0 z-50 flex items-center justify-center bg-gradient-to-b from-background via-background to-primary/5 p-6">
@@ -65,7 +65,7 @@ export function SplashScreen({
           <Button size="lg" onClick={onStartOnboarding}>
             Start Onboarding
           </Button>
-          <Button size="lg" variant="outline" onClick={onGoToStudio}>
+          <Button size="lg" variant="outline" onClick={onSkipOnboarding}>
             Skip Onboarding
           </Button>
         </motion.div>
diff --git a/studio/frontend/src/features/onboarding/components/wizard-footer.tsx b/studio/frontend/src/features/onboarding/components/wizard-footer.tsx
index 81e2f2738f..4588c0a632 100644
--- a/studio/frontend/src/features/onboarding/components/wizard-footer.tsx
+++ b/studio/frontend/src/features/onboarding/components/wizard-footer.tsx
@@ -7,10 +7,15 @@ import { markOnboardingDone } from "@/features/auth";
 import { useTrainingConfigStore } from "@/features/training";
 import { ArrowLeft02Icon, ArrowRight02Icon } from "@hugeicons/core-free-icons";
 import { HugeiconsIcon } from "@hugeicons/react";
-import { useNavigate } from "@tanstack/react-router";
 import { useShallow } from "zustand/react/shallow";
 
-export function WizardFooter({ onBackToSplash }: { onBackToSplash: () => void }) {
+export function WizardFooter({
+  returnTo,
+  onBackToSplash,
+}: {
+  returnTo: string;
+  onBackToSplash: () => void;
+}) {
   const { currentStep, prevStep, nextStep, canProceed } = useTrainingConfigStore(
     useShallow((s) => ({
       currentStep: s.currentStep,
@@ -19,7 +24,6 @@ export function WizardFooter({ onBackToSplash }: { onBackToSplash: () => void })
       canProceed: s.canProceed(),
     })),
   );
-  const navigate = useNavigate();
   const isFirst = currentStep === 1;
   const isLast = currentStep === STEPS.length;
 
@@ -41,7 +45,7 @@ export function WizardFooter({ onBackToSplash }: { onBackToSplash: () => void })
               className="px-4"
               onClick={() => {
                 markOnboardingDone();
-                navigate({ to: "/studio" });
+                window.location.assign(returnTo);
               }}
             >
               Skip
@@ -51,12 +55,12 @@ export function WizardFooter({ onBackToSplash }: { onBackToSplash: () => void })
             <Button
               onClick={() => {
                 markOnboardingDone();
-                navigate({ to: "/studio" });
+                window.location.assign(returnTo);
               }}
               disabled={!canProceed}
               className="px-4 !pr-4"
             >
-              Go to Studio
+              Finish onboarding
               <HugeiconsIcon icon={ArrowRight02Icon} data-icon="inline-end" />
             </Button>
           ) : (
@@ -65,7 +69,7 @@ export function WizardFooter({ onBackToSplash }: { onBackToSplash: () => void })
                 if (currentStep === 1 && sessionStorage.getItem("unsloth_chat_only") === "1") {
                   sessionStorage.removeItem("unsloth_chat_only");
                   markOnboardingDone();
-                  window.location.href = "/chat";
+                  window.location.assign(returnTo);
                 } else {
                   nextStep();
                 }
diff --git a/studio/frontend/src/features/onboarding/components/wizard-layout.tsx b/studio/frontend/src/features/onboarding/components/wizard-layout.tsx
index 940ac333be..6df873ffd3 100644
--- a/studio/frontend/src/features/onboarding/components/wizard-layout.tsx
+++ b/studio/frontend/src/features/onboarding/components/wizard-layout.tsx
@@ -2,7 +2,7 @@
 // Copyright 2026-present the Unsloth AI Inc. team. All rights reserved. See /studio/LICENSE.AGPL-3.0
 
 import { Card } from "@/components/ui/card";
-import { useNavigate } from "@tanstack/react-router";
+import { Route as OnboardingRoute } from "@/app/routes/onboarding";
 import { motion } from "motion/react";
 import { Suspense, lazy, useEffect, useRef, useState } from "react";
 
@@ -19,13 +19,23 @@ const Confetti = lazy(() =>
   import("@/components/ui/confetti").then((m) => ({ default: m.Confetti })),
 );
 
+function sanitizeRedirectTarget(value: string | undefined): string {
+  if (!value) return "/chat";
+  if (!value.startsWith("/")) return "/chat";
+  if (value.startsWith("//")) return "/chat";
+  if (value.includes("\\")) return "/chat";
+  return value;
+}
+
 export function WizardLayout() {
-  const navigate = useNavigate();
+  const search = OnboardingRoute.useSearch();
   const [showSplash, setShowSplash] = useState(true);
   const currentStep = useTrainingConfigStore((s) => s.currentStep);
   const confettiRef = useRef<ConfettiRef>(null);
   const hasFiredRef = useRef(false);
   const isFinalStep = currentStep === STEPS.length;
+  const returnTo = sanitizeRedirectTarget(search.redirectTo);
+  const exitToReturnTo = () => window.location.assign(returnTo);
 
   // Only redirect on initial mount — not on re-renders after markOnboardingDone()
   // which would override explicit /chat navigation from skip buttons.
@@ -34,10 +44,10 @@ export function WizardLayout() {
     if (!checkedRef.current) {
       checkedRef.current = true;
       if (isOnboardingDone()) {
-        navigate({ to: "/studio" });
+        exitToReturnTo();
       }
     }
-  }, [navigate]);
+  }, [returnTo]);
 
   useEffect(() => {
     if (isFinalStep && !hasFiredRef.current) {
@@ -67,9 +77,9 @@ export function WizardLayout() {
       {showSplash && (
         <SplashScreen
           onStartOnboarding={() => setShowSplash(false)}
-          onGoToStudio={() => {
+          onSkipOnboarding={() => {
             markOnboardingDone();
-            window.location.href = "/studio";
+            exitToReturnTo();
           }}
         />
       )}
@@ -91,10 +101,10 @@ export function WizardLayout() {
           }}
         >
           <Card className="relative z-10 w-full !gap-0 !m-0 !p-0 flex min-h-[560px] flex-col overflow-hidden shadow-border ring-1 ring-border md:min-h-[620px] md:flex-row lg:h-[660px]">
-            <WizardSidebar />
+            <WizardSidebar returnTo={returnTo} />
             <div className="flex-1 flex flex-col">
               <WizardContent />
-              <WizardFooter onBackToSplash={() => setShowSplash(true)} />
+              <WizardFooter returnTo={returnTo} onBackToSplash={() => setShowSplash(true)} />
             </div>
           </Card>
         </motion.div>
diff --git a/studio/frontend/src/features/onboarding/components/wizard-sidebar.tsx b/studio/frontend/src/features/onboarding/components/wizard-sidebar.tsx
index 70621fc1b7..d22ac19e51 100644
--- a/studio/frontend/src/features/onboarding/components/wizard-sidebar.tsx
+++ b/studio/frontend/src/features/onboarding/components/wizard-sidebar.tsx
@@ -10,7 +10,7 @@ import { ArrowRight02Icon } from "@hugeicons/core-free-icons";
 import { HugeiconsIcon } from "@hugeicons/react";
 import { WizardStepItem } from "./wizard-step-item";
 
-export function WizardSidebar() {
+export function WizardSidebar({ returnTo }: { returnTo: string }) {
   const currentStep = useTrainingConfigStore((s) => s.currentStep);
   const progress = ((currentStep - 1) / (STEPS.length - 1)) * 100;
 
@@ -38,10 +38,10 @@ export function WizardSidebar() {
         className="mt-2 w-full md:hidden"
         onClick={() => {
           markOnboardingDone();
-          window.location.href = "/chat";
+          window.location.assign(returnTo);
         }}
       >
-        Skip to Chat
+        Skip onboarding
         <HugeiconsIcon icon={ArrowRight02Icon} data-icon="inline-end" />
       </Button>
       <nav className="mt-3 hidden flex-col gap-1 md:flex">
@@ -54,10 +54,10 @@ export function WizardSidebar() {
         className="mt-3 hidden w-full md:flex"
         onClick={() => {
           markOnboardingDone();
-          window.location.href = "/chat";
+          window.location.assign(returnTo);
         }}
       >
-        Skip to Chat
+        Skip onboarding
         <HugeiconsIcon icon={ArrowRight02Icon} data-icon="inline-end" />
       </Button>
     </aside>
diff --git a/studio/frontend/src/features/settings/tabs/about-tab.tsx b/studio/frontend/src/features/settings/tabs/about-tab.tsx
index f25afc6d8f..417922139c 100644
--- a/studio/frontend/src/features/settings/tabs/about-tab.tsx
+++ b/studio/frontend/src/features/settings/tabs/about-tab.tsx
@@ -13,17 +13,36 @@ import {
   MessageNotification01Icon,
 } from "@hugeicons/core-free-icons";
 import { HugeiconsIcon } from "@hugeicons/react";
-import { useState } from "react";
+import { useEffect, useState } from "react";
 import { SettingsRow } from "../components/settings-row";
 import { SettingsSection } from "../components/settings-section";
 
-const VERSION: string =
-  (import.meta.env.VITE_APP_VERSION as string | undefined) ?? "dev";
-
 export function AboutTab() {
   const deviceType = usePlatformStore((s) => s.deviceType);
   const defaultShell = deviceType === "windows" ? "windows" : "unix";
   const [shutdownOpen, setShutdownOpen] = useState(false);
+  const [version, setVersion] = useState("dev");
+
+  useEffect(() => {
+    let canceled = false;
+
+    (async () => {
+      try {
+        const res = await fetch("/api/health");
+        if (!res.ok) return;
+        const data = (await res.json()) as { version?: string };
+        if (!canceled && data.version) {
+          setVersion(data.version);
+        }
+      } catch {
+        // fall back to dev label
+      }
+    })();
+
+    return () => {
+      canceled = true;
+    };
+  }, []);
 
   return (
     <div className="flex flex-col gap-6">
@@ -36,7 +55,7 @@ export function AboutTab() {
 
       <SettingsSection title="Studio">
         <SettingsRow label="Version">
-          <code className="font-mono text-xs text-muted-foreground">{VERSION}</code>
+          <code className="font-mono text-xs text-muted-foreground">{version}</code>
         </SettingsRow>
       </SettingsSection>
 
diff --git a/studio/frontend/src/features/settings/tabs/general-tab.tsx b/studio/frontend/src/features/settings/tabs/general-tab.tsx
index e1ad3fb58c..a79b134e97 100644
--- a/studio/frontend/src/features/settings/tabs/general-tab.tsx
+++ b/studio/frontend/src/features/settings/tabs/general-tab.tsx
@@ -12,7 +12,10 @@ import {
 } from "@/components/ui/dialog";
 import { Input } from "@/components/ui/input";
 import { Switch } from "@/components/ui/switch";
+import { resetOnboardingDone } from "@/features/auth";
 import { useChatRuntimeStore } from "@/features/chat/stores/chat-runtime-store";
+import { useSettingsDialogStore } from "@/features/settings";
+import { useNavigate, useRouterState } from "@tanstack/react-router";
 import { useEffect, useRef, useState } from "react";
 import { Eye, EyeOff } from "lucide-react";
 import { SettingsRow } from "../components/settings-row";
@@ -75,10 +78,24 @@ function resetAllPrefs() {
 }
 
 export function GeneralTab() {
+  const navigate = useNavigate();
+  const closeDialog = useSettingsDialogStore((s) => s.closeDialog);
+  const { pathname, search } = useRouterState({
+    select: (s) => ({
+      pathname: s.location.pathname,
+      search:
+        "searchStr" in s.location
+          ? (s.location as { searchStr?: string }).searchStr ?? ""
+          : typeof window !== "undefined"
+            ? window.location.search
+            : "",
+    }),
+  });
   const hfToken = useChatRuntimeStore((s) => s.hfToken);
   const setHfToken = useChatRuntimeStore((s) => s.setHfToken);
   const autoTitle = useChatRuntimeStore((s) => s.autoTitle);
   const setAutoTitle = useChatRuntimeStore((s) => s.setAutoTitle);
+  const redirectTo = `${pathname}${search}`;
 
   const [draftToken, setDraftToken] = useState(hfToken ?? "");
   const [showToken, setShowToken] = useState(false);
@@ -153,6 +170,25 @@ export function GeneralTab() {
         </SettingsRow>
       </SettingsSection>
 
+      <SettingsSection title="Getting started">
+        <SettingsRow
+          label="Start onboarding"
+          description="Open the setup wizard again without changing your account."
+        >
+          <Button
+            variant="outline"
+            size="sm"
+            onClick={() => {
+              resetOnboardingDone();
+              closeDialog();
+              navigate({ to: "/onboarding", search: { redirectTo } });
+            }}
+          >
+            Start onboarding
+          </Button>
+        </SettingsRow>
+      </SettingsSection>
+
       <SettingsSection title="Danger zone">
         <SettingsRow
           destructive
diff --git a/studio/frontend/src/features/studio/studio-page.tsx b/studio/frontend/src/features/studio/studio-page.tsx
index d8a35359b8..5c5e11907c 100644
--- a/studio/frontend/src/features/studio/studio-page.tsx
+++ b/studio/frontend/src/features/studio/studio-page.tsx
@@ -25,8 +25,6 @@ import { LiveTrainingView } from "./live-training-view";
 import { HistoricalTrainingView } from "./historical-training-view";
 import { HistoryCardGrid } from "./history-card-grid";
 
-const STUDIO_TOUR_KEY = "tour:studio:v1";
-
 export function StudioPage(): ReactElement {
   useTrainingRuntimeLifecycle();
   const showTrainingView = useTrainingRuntimeStore(shouldShowTrainingView);
@@ -85,8 +83,6 @@ export function StudioPage(): ReactElement {
     id: "studio",
     steps: tourSteps,
     enabled: tourEnabled,
-    autoKey: isConfigTour ? STUDIO_TOUR_KEY : undefined,
-    autoWhen: isConfigTour,
   });
 
   const setTourOpen = tour.setOpen;
diff --git a/studio/frontend/src/index.css b/studio/frontend/src/index.css
index 8ee4f9eca6..dd559a3765 100644
--- a/studio/frontend/src/index.css
+++ b/studio/frontend/src/index.css
@@ -56,7 +56,7 @@
 	--chart-3: oklch(0.7014 0.1193 197.5897);
 	--chart-4: oklch(0.6926 0.1112 346.5775);
 	--chart-5: oklch(0.7497 0.1003 85.0057);
-	--radius: 0.625rem;
+	--radius: 1.1rem;
 	--sidebar: oklch(0.99 0 0);
 	--sidebar-foreground: oklch(0.1281 0.0179 169.2764);
 	--sidebar-primary: oklch(0.6929 0.1396 166.5513);
@@ -132,7 +132,7 @@
 	--sidebar-border: oklch(0.38 0 0);
 	--sidebar-ring: oklch(0.6929 0.1396 166.5513);
 	--destructive-foreground: oklch(1 0 0);
-	--radius: 0.625rem;
+	--radius: 1.1rem;
 	--font-sans: Geist, ui-sans-serif, sans-serif, system-ui;
 	--font-serif: Source Serif 4, serif;
 	--font-mono: JetBrains Mono, monospace;
@@ -198,7 +198,7 @@
 	--radius-4xl: calc(var(--radius) + 16px);
 	--font-mono: JetBrains Mono, monospace;
 	--font-serif: Source Serif 4, serif;
-	--radius: 0.625rem;
+	--radius: 1.1rem;
 	--tracking-tighter: calc(var(--tracking-normal) - 0.05em);
 	--tracking-tight: calc(var(--tracking-normal) - 0.025em);
 	--tracking-wide: calc(var(--tracking-normal) + 0.025em);

From bc9ddb3af690194e7fae25d7740a04d16184431e Mon Sep 17 00:00:00 2001
From: Wasim Yousef Said <wasimysdev@gmail.com>
Date: Thu, 16 Apr 2026 19:11:35 +0200
Subject: [PATCH 129/216] Fix onboarding followups (#5064)

* Fix onboarding followups

* Rename sidebar studio to train
---
 .../frontend/src/components/app-sidebar.tsx   |  4 +-
 .../onboarding/components/wizard-footer.tsx   |  2 +-
 .../features/settings/tabs/general-tab.tsx    | 38 ++++++++++---------
 3 files changed, 24 insertions(+), 20 deletions(-)

diff --git a/studio/frontend/src/components/app-sidebar.tsx b/studio/frontend/src/components/app-sidebar.tsx
index 8264f329a3..329175fa9c 100644
--- a/studio/frontend/src/components/app-sidebar.tsx
+++ b/studio/frontend/src/components/app-sidebar.tsx
@@ -330,7 +330,7 @@ export function AppSidebar() {
             <SidebarMenu>
               <NavItem
                 icon={ZapIcon}
-                label="Studio"
+                label="Train"
                 active={pathname === "/studio" || pathname.startsWith("/studio/")}
                 disabled={chatOnly}
                 onClick={() => {
@@ -511,7 +511,7 @@ export function AppSidebar() {
                   />
                   <div className="flex flex-col gap-0.5 leading-none group-data-[collapsible=icon]:hidden">
                     <span className="truncate text-sm font-semibold">Unsloth</span>
-                    <span className="truncate text-[11px] text-muted-foreground">Studio</span>
+                    <span className="truncate text-[11px] text-muted-foreground">Train</span>
                   </div>
                   <ChevronsUpDown strokeWidth={1.25} className="ml-auto size-4 text-muted-foreground group-data-[collapsible=icon]:hidden" />
                 </SidebarMenuButton>
diff --git a/studio/frontend/src/features/onboarding/components/wizard-footer.tsx b/studio/frontend/src/features/onboarding/components/wizard-footer.tsx
index 4588c0a632..399bf115f1 100644
--- a/studio/frontend/src/features/onboarding/components/wizard-footer.tsx
+++ b/studio/frontend/src/features/onboarding/components/wizard-footer.tsx
@@ -69,7 +69,7 @@ export function WizardFooter({
                 if (currentStep === 1 && sessionStorage.getItem("unsloth_chat_only") === "1") {
                   sessionStorage.removeItem("unsloth_chat_only");
                   markOnboardingDone();
-                  window.location.assign(returnTo);
+                  window.location.assign("/chat");
                 } else {
                   nextStep();
                 }
diff --git a/studio/frontend/src/features/settings/tabs/general-tab.tsx b/studio/frontend/src/features/settings/tabs/general-tab.tsx
index a79b134e97..874508a6fb 100644
--- a/studio/frontend/src/features/settings/tabs/general-tab.tsx
+++ b/studio/frontend/src/features/settings/tabs/general-tab.tsx
@@ -12,6 +12,7 @@ import {
 } from "@/components/ui/dialog";
 import { Input } from "@/components/ui/input";
 import { Switch } from "@/components/ui/switch";
+import { usePlatformStore } from "@/config/env";
 import { resetOnboardingDone } from "@/features/auth";
 import { useChatRuntimeStore } from "@/features/chat/stores/chat-runtime-store";
 import { useSettingsDialogStore } from "@/features/settings";
@@ -95,6 +96,7 @@ export function GeneralTab() {
   const setHfToken = useChatRuntimeStore((s) => s.setHfToken);
   const autoTitle = useChatRuntimeStore((s) => s.autoTitle);
   const setAutoTitle = useChatRuntimeStore((s) => s.setAutoTitle);
+  const chatOnly = usePlatformStore((s) => s.chatOnly);
   const redirectTo = `${pathname}${search}`;
 
   const [draftToken, setDraftToken] = useState(hfToken ?? "");
@@ -170,24 +172,26 @@ export function GeneralTab() {
         </SettingsRow>
       </SettingsSection>
 
-      <SettingsSection title="Getting started">
-        <SettingsRow
-          label="Start onboarding"
-          description="Open the setup wizard again without changing your account."
-        >
-          <Button
-            variant="outline"
-            size="sm"
-            onClick={() => {
-              resetOnboardingDone();
-              closeDialog();
-              navigate({ to: "/onboarding", search: { redirectTo } });
-            }}
+      {!chatOnly && (
+        <SettingsSection title="Getting started">
+          <SettingsRow
+            label="Start onboarding"
+            description="Open the setup wizard again without changing your account."
           >
-            Start onboarding
-          </Button>
-        </SettingsRow>
-      </SettingsSection>
+            <Button
+              variant="outline"
+              size="sm"
+              onClick={() => {
+                resetOnboardingDone();
+                closeDialog();
+                navigate({ to: "/onboarding", search: { redirectTo } });
+              }}
+            >
+              Start onboarding
+            </Button>
+          </SettingsRow>
+        </SettingsSection>
+      )}
 
       <SettingsSection title="Danger zone">
         <SettingsRow

From ee86530e55ea98e060fb9c86f0bf3f5a48f61238 Mon Sep 17 00:00:00 2001
From: Lee Jackson <130007945+Imagineer99@users.noreply.github.com>
Date: Thu, 16 Apr 2026 19:27:30 +0100
Subject: [PATCH 130/216] chore: switch helper and no-cache fallback to Gemma
 (#5066)

---
 studio/backend/utils/datasets/llm_assist.py      |  2 +-
 .../src/features/chat/api/chat-adapter.ts        | 16 ++++++++--------
 2 files changed, 9 insertions(+), 9 deletions(-)

diff --git a/studio/backend/utils/datasets/llm_assist.py b/studio/backend/utils/datasets/llm_assist.py
index fdc4f374ab..4c66d2ebf6 100644
--- a/studio/backend/utils/datasets/llm_assist.py
+++ b/studio/backend/utils/datasets/llm_assist.py
@@ -26,7 +26,7 @@ from loggers import get_logger
 
 logger = get_logger(__name__)
 
-DEFAULT_HELPER_MODEL_REPO = "unsloth/Qwen3.5-4B-GGUF"
+DEFAULT_HELPER_MODEL_REPO = "unsloth/gemma-4-E2B-it-GGUF"
 DEFAULT_HELPER_MODEL_VARIANT = "UD-Q4_K_XL"
 
 README_MAX_CHARS = 1500
diff --git a/studio/frontend/src/features/chat/api/chat-adapter.ts b/studio/frontend/src/features/chat/api/chat-adapter.ts
index 60904f3cb5..53a935cf45 100644
--- a/studio/frontend/src/features/chat/api/chat-adapter.ts
+++ b/studio/frontend/src/features/chat/api/chat-adapter.ts
@@ -455,13 +455,13 @@ async function autoLoadSmallestModel(): Promise<{
     // No cached models found — try downloading a small default GGUF
     toast("Downloading a small model…", {
       id: toastId,
-      description: "No downloaded models found. Fetching Qwen3.5-4B (UD-Q4_K_XL).",
+      description: "No downloaded models found. Fetching Gemma-4-E2B-it (UD-Q4_K_XL).",
       duration: 30000,
     });
     try {
       if (
         !(await canAutoLoad({
-          model_path: "unsloth/Qwen3.5-4B-GGUF",
+          model_path: "unsloth/gemma-4-E2B-it-GGUF",
           max_seq_length: 0,
           is_lora: false,
           gguf_variant: "UD-Q4_K_XL",
@@ -471,7 +471,7 @@ async function autoLoadSmallestModel(): Promise<{
         return { loaded: false, blockedByTrustRemoteCode };
       }
       const loadResp = await loadModel({
-        model_path: "unsloth/Qwen3.5-4B-GGUF",
+        model_path: "unsloth/gemma-4-E2B-it-GGUF",
         hf_token: hfToken,
         max_seq_length: 0,
         load_in_4bit: true,
@@ -479,20 +479,20 @@ async function autoLoadSmallestModel(): Promise<{
         gguf_variant: "UD-Q4_K_XL",
         trust_remote_code: trustRemoteCode,
       });
-      useChatRuntimeStore.getState().setCheckpoint("unsloth/Qwen3.5-4B-GGUF", "UD-Q4_K_XL");
+      useChatRuntimeStore.getState().setCheckpoint("unsloth/gemma-4-E2B-it-GGUF", "UD-Q4_K_XL");
       const store = useChatRuntimeStore.getState();
       store.setModelRequiresTrustRemoteCode(
         loadResp.requires_trust_remote_code ?? false,
       );
       store.setParams({ ...store.params, maxTokens: loadResp.context_length ?? 131072 });
       const defaultModel: ChatModelSummary = {
-        id: "unsloth/Qwen3.5-4B-GGUF",
-        name: loadResp.display_name ?? "Qwen3.5-4B-GGUF",
+        id: "unsloth/gemma-4-E2B-it-GGUF",
+        name: loadResp.display_name ?? "gemma-4-E2B-it-GGUF",
         isVision: loadResp.is_vision ?? false,
         isLora: false,
         isGguf: true,
       };
-      if (!store.models.some((m) => m.id === "unsloth/Qwen3.5-4B-GGUF")) {
+      if (!store.models.some((m) => m.id === "unsloth/gemma-4-E2B-it-GGUF")) {
         store.setModels([...store.models, defaultModel]);
       }
       useChatRuntimeStore.setState({
@@ -509,7 +509,7 @@ async function autoLoadSmallestModel(): Promise<{
         defaultChatTemplate: loadResp.chat_template ?? null,
         chatTemplateOverride: null,
       });
-      toast.success("Loaded Qwen3.5-4B (UD-Q4_K_XL)", { id: toastId });
+      toast.success("Loaded Gemma-4-E2B-it (UD-Q4_K_XL)", { id: toastId });
       return { loaded: true, blockedByTrustRemoteCode: false };
     } catch {
       toast.dismiss(toastId);

From d56f98045260e29ef3f671b3b801a7ed771c0419 Mon Sep 17 00:00:00 2001
From: Daniel Han <danielhanchen@gmail.com>
Date: Thu, 16 Apr 2026 11:35:02 -0700
Subject: [PATCH 131/216] fix: multi-GPU inference crash for bnb 4-bit/8-bit
 models (#5068)

* fix: multi-GPU inference crash for bnb 4-bit/8-bit models

When load_in_4bit or load_in_8bit is used with device_map="sequential"
and max_memory constraints that place weights across multiple GPUs (or
entirely on a non-default GPU like cuda:1), the bitsandbytes loading
path in transformers never calls dispatch_model. No AlignDevicesHook is
installed, and the first forward/generate call crashes with:

  RuntimeError: Expected all tensors to be on the same device

This adds _attach_bnb_multidevice_hooks() which is called after
from_pretrained returns. It infers a device map from actual parameter
placements and calls dispatch_model(force_hooks=True) to install the
missing hooks. The function is a complete no-op for the common
single-GPU cuda:0 case.

Call sites: FastBaseModel.from_pretrained (vision.py) and
FastLlamaModel.from_pretrained (llama.py).

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

* fix: align with PR #5053 final review improvements

- Add hook call to the bnb quantized loading branch in llama.py (the
  primary load_in_4bit path), not just the non-fast-inference fallback
- Expand bnb detection: also check model.is_loaded_in_4bit,
  model.is_loaded_in_8bit, model.quantization_method
- Pass explicit main_device and skip_keys to dispatch_model
- Use logger.info instead of print for the success message
- Use kwargs.get("load_in_8bit", False) at llama.py call sites

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

---------

Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com>
---
 unsloth/models/llama.py  |  20 ++++++
 unsloth/models/vision.py | 147 +++++++++++++++++++++++++++++++++++++++
 2 files changed, 167 insertions(+)

diff --git a/unsloth/models/llama.py b/unsloth/models/llama.py
index 999711efdb..d39f2588ef 100644
--- a/unsloth/models/llama.py
+++ b/unsloth/models/llama.py
@@ -2478,6 +2478,16 @@ class FastLlamaModel:
                     and not _head.weight.is_floating_point()
                 ):
                     _head.to(dtype)
+            # Attach dispatch hooks for bnb multi-device loads.
+            from unsloth.models.vision import _attach_bnb_multidevice_hooks
+
+            _attach_bnb_multidevice_hooks(
+                model,
+                load_in_4bit = load_in_4bit,
+                load_in_8bit = kwargs.get("load_in_8bit", False),
+                offload_embedding = False,
+                fast_inference = fast_inference,
+            )
         elif not fast_inference:
             model = AutoModelForCausalLM.from_pretrained(
                 model_name,
@@ -2490,6 +2500,16 @@ class FastLlamaModel:
                 attn_implementation = preferred_attn_impl,
                 **kwargs,
             )
+            # Attach dispatch hooks for bnb multi-device loads.
+            from unsloth.models.vision import _attach_bnb_multidevice_hooks
+
+            _attach_bnb_multidevice_hooks(
+                model,
+                load_in_4bit = load_in_4bit,
+                load_in_8bit = kwargs.get("load_in_8bit", False),
+                offload_embedding = False,
+                fast_inference = False,
+            )
             model.fast_generate = make_fast_generate_wrapper(model.generate)
             model.fast_generate_batches = None
         else:
diff --git a/unsloth/models/vision.py b/unsloth/models/vision.py
index 90c93ea3f7..df371e00c8 100644
--- a/unsloth/models/vision.py
+++ b/unsloth/models/vision.py
@@ -75,6 +75,7 @@ import functools
 import os
 import gc
 import math
+import warnings
 from typing import Optional, Tuple, List, Union
 import re, inspect, sys
 import contextlib
@@ -97,6 +98,144 @@ __all__ = [
     "FastBaseModel",
 ]
 
+
+def _infer_device_map_from_loaded_model(model):
+    """Build a compact device_map by inspecting actual parameter placements."""
+    device_map = {}
+
+    def _assign(module, prefix):
+        params = list(module.named_parameters(remove_duplicate = False))
+        if not params:
+            bufs = list(module.named_buffers())
+            if bufs:
+                device_map[prefix] = bufs[0][1].device
+            return
+        devices = {p.device for _, p in params}
+        if len(devices) == 1:
+            device_map[prefix] = next(iter(devices))
+        else:
+            for child_name, child in module.named_children():
+                child_prefix = f"{prefix}.{child_name}" if prefix else child_name
+                _assign(child, child_prefix)
+            for pname, param in module.named_parameters(remove_duplicate = False):
+                if "." not in pname:
+                    full = f"{prefix}.{pname}" if prefix else pname
+                    if not any(
+                        full == k or full.startswith(k + ".") for k in device_map
+                    ):
+                        device_map[full] = param.device
+
+    _assign(model, "")
+    if "" in device_map and len(device_map) > 1:
+        device_map.pop("")
+    return device_map
+
+
+def _attach_bnb_multidevice_hooks(
+    model, load_in_4bit, load_in_8bit, offload_embedding, fast_inference
+):
+    """
+    Attach accelerate AlignDevicesHook on a bnb model loaded across multiple
+    devices (or a non-default device).  No-op for single-GPU cuda:0, non-bnb,
+    vLLM, or already-dispatched models.
+    """
+    if fast_inference:
+        return
+    is_bnb = (
+        load_in_4bit
+        or load_in_8bit
+        or getattr(model, "is_loaded_in_4bit", False)
+        or getattr(model, "is_loaded_in_8bit", False)
+        or getattr(model, "quantization_method", None) == "bitsandbytes"
+    )
+    if not is_bnb:
+        return
+    if offload_embedding:
+        return
+    if getattr(model, "hf_device_map", None) is not None:
+        return  # already dispatched
+
+    try:
+        all_devs = {p.device for p in model.parameters()}
+    except Exception as exc:
+        warnings.warn(
+            "Unsloth: Failed to determine device placement from model parameters, "
+            f"so multi-GPU hooks cannot be attached. ({type(exc).__name__}: {exc})",
+            RuntimeWarning,
+            stacklevel = 2,
+        )
+        return
+
+    cuda_devs = {d for d in all_devs if d.type == "cuda"}
+    if not cuda_devs:
+        return
+
+    default_cuda = torch.device("cuda", 0)
+    if all_devs == {default_cuda}:
+        return
+
+    try:
+        from accelerate import dispatch_model
+    except ImportError:
+        return  # accelerate not available
+
+    try:
+        inferred_map = _infer_device_map_from_loaded_model(model)
+        if not inferred_map:
+            return
+
+        # bnb constructors reject _is_hf_initialized; strip before dispatch.
+        _extra_keys = ("_is_hf_initialized",)
+        _stripped = []
+        for _, param in model.named_parameters():
+            for key in _extra_keys:
+                if key in param.__dict__:
+                    _stripped.append((param, key, param.__dict__.pop(key)))
+
+        try:
+            # CUDA -> int index, non-CUDA -> type string ("cpu", "meta").
+            device_map_int = {
+                k: (v.index if v.type == "cuda" else v.type)
+                if isinstance(v, torch.device)
+                else v
+                for k, v in inferred_map.items()
+            }
+
+            # force_hooks=True: install hooks even for single-device maps.
+            main_device = device_map_int.get("")
+            if main_device in (None, "cpu", "disk"):
+                main_device = next(
+                    (d for d in device_map_int.values() if d not in ("cpu", "disk")),
+                    None,
+                )
+            dispatch_model(
+                model,
+                device_map = device_map_int,
+                main_device = main_device,
+                skip_keys = getattr(model, "_skip_keys_device_placement", None),
+                force_hooks = True,
+            )
+            desc = f"{len(inferred_map)} block(s) across {len(cuda_devs)} device(s)"
+        finally:
+            # Restore stripped keys.
+            for param, key, val in _stripped:
+                param.__dict__[key] = val
+
+        logger.info(
+            f"Unsloth: Attached accelerate AlignDevicesHook ({desc}) "
+            f"for bnb multi-GPU inference."
+        )
+    except Exception as exc:
+        warnings.warn(
+            f"Unsloth: Could not attach multi-device dispatch hooks automatically "
+            f"({type(exc).__name__}: {exc}). "
+            "Cross-device inference may fail. Consider using a single GPU or "
+            "calling accelerate.dispatch_model() manually.",
+            RuntimeWarning,
+            stacklevel = 2,
+        )
+
+
 global NUM_LOGITS_TO_KEEP
 NUM_LOGITS_TO_KEEP = dict()
 
@@ -796,6 +935,14 @@ class FastBaseModel:
                 # attn_implementation   = attn_implementation,
                 **kwargs,
             )
+            # Attach dispatch hooks for bnb multi-device loads.
+            _attach_bnb_multidevice_hooks(
+                model,
+                load_in_4bit = load_in_4bit,
+                load_in_8bit = load_in_8bit,
+                offload_embedding = offload_embedding,
+                fast_inference = fast_inference,
+            )
             if hasattr(model, "generate"):
                 model.fast_generate = make_fast_generate_wrapper(model.generate)
                 model.fast_generate_batches = error_out_no_vllm

From 0b57884120f68a9765dba63b28f9f2d85df6ad9b Mon Sep 17 00:00:00 2001
From: Daniel Han <danielhanchen@gmail.com>
Date: Thu, 16 Apr 2026 11:42:42 -0700
Subject: [PATCH 132/216] Add Qwen3.6 inference defaults for Studio (#5065)

* Add Qwen3.6 inference defaults for Studio

Add qwen3.6 family entry to inference_defaults.json with the
recommended sampling parameters from Qwen's documentation:
temperature=0.7, top_p=0.8, top_k=20, min_p=0.0,
presence_penalty=1.5, repetition_penalty=1.0.

Without this, Qwen3.6 models fall through to the generic qwen3
pattern which uses different defaults (temperature=0.6,
top_p=0.95, no presence_penalty).

* Add Qwen3.6-35B-A3B-GGUF to default model lists

* Add Qwen3.5/3.6 presence_penalty to thinking toggle and small-model disable logic

- Thinking toggle (on-load + button click) now sets presencePenalty: 1.5 for
  Qwen3.5 and Qwen3.6 models (both thinking-ON and thinking-OFF states)
- Small-model thinking-disable check (<9B defaults to no-thinking) extended
  from Qwen3.5-only to also cover Qwen3.6, in all 3 locations:
  frontend on-load, frontend refresh, backend llama_cpp.py
---
 .../assets/configs/inference_defaults.json       | 10 +++++++++-
 studio/backend/core/inference/defaults.py        |  2 ++
 studio/backend/core/inference/llama_cpp.py       |  4 ++--
 .../chat/hooks/use-chat-model-runtime.ts         | 16 +++++++++-------
 .../src/features/chat/shared-composer.tsx        |  7 ++++---
 5 files changed, 26 insertions(+), 13 deletions(-)

diff --git a/studio/backend/assets/configs/inference_defaults.json b/studio/backend/assets/configs/inference_defaults.json
index 1b4b5381e8..1b10b557e4 100644
--- a/studio/backend/assets/configs/inference_defaults.json
+++ b/studio/backend/assets/configs/inference_defaults.json
@@ -1,6 +1,14 @@
 {
   "_comment": "Per-model-family inference parameter defaults. Sources: (1) Ollama params blobs, (2) Existing Unsloth Studio YAML configs. Patterns ordered longest-match-first.",
   "families": {
+    "qwen3.6": {
+      "temperature": 0.7,
+      "top_p": 0.8,
+      "top_k": 20,
+      "min_p": 0.0,
+      "repetition_penalty": 1.0,
+      "presence_penalty": 1.5
+    },
     "qwen3.5": {
       "temperature": 0.7,
       "top_p": 0.8,
@@ -369,7 +377,7 @@
     }
   },
   "patterns": [
-    "qwen3.5",
+    "qwen3.6", "qwen3.5",
     "qwen3-coder", "qwen3-next", "qwen3-vl", "qwen3",
     "qwen2.5-coder", "qwen2.5-vl", "qwen2.5-omni", "qwen2.5-math", "qwen2.5",
     "qwen2-vl", "qwen2",
diff --git a/studio/backend/core/inference/defaults.py b/studio/backend/core/inference/defaults.py
index f3026dddaf..53718c1294 100644
--- a/studio/backend/core/inference/defaults.py
+++ b/studio/backend/core/inference/defaults.py
@@ -10,6 +10,7 @@ DEFAULT_MODELS_GGUF = [
     "unsloth/gemma-4-E4B-it-GGUF",
     "unsloth/gemma-4-31B-it-GGUF",
     "unsloth/gemma-4-26B-A4B-it-GGUF",
+    "unsloth/Qwen3.6-35B-A3B-GGUF",
     "unsloth/Qwen3.5-4B-GGUF",
     "unsloth/Qwen3.5-9B-GGUF",
     "unsloth/Qwen3.5-35B-A3B-GGUF",
@@ -27,6 +28,7 @@ DEFAULT_MODELS_STANDARD = [
     "unsloth/gemma-4-E4B-it-GGUF",
     "unsloth/gemma-4-31B-it-GGUF",
     "unsloth/gemma-4-26B-A4B-it-GGUF",
+    "unsloth/Qwen3.6-35B-A3B-GGUF",
     "unsloth/Qwen3.5-4B-GGUF",
     "unsloth/Qwen3.5-9B-GGUF",
     "unsloth/Qwen3.5-35B-A3B-GGUF",
diff --git a/studio/backend/core/inference/llama_cpp.py b/studio/backend/core/inference/llama_cpp.py
index 77b58e22fb..2e26995309 100644
--- a/studio/backend/core/inference/llama_cpp.py
+++ b/studio/backend/core/inference/llama_cpp.py
@@ -1514,12 +1514,12 @@ class LlamaCppBackend:
                 )
 
             # For reasoning models, set default thinking mode.
-            # Qwen3.5 models below 9B (0.8B, 2B, 4B) disable thinking by default.
+            # Qwen3.5/3.6 models below 9B (0.8B, 2B, 4B) disable thinking by default.
             # Only 9B and larger enable thinking.
             if self._supports_reasoning:
                 thinking_default = True
                 mid = (model_identifier or "").lower()
-                if "qwen3.5" in mid:
+                if "qwen3.5" in mid or "qwen3.6" in mid:
                     size_val = _extract_model_size_b(mid)
                     if size_val is not None and size_val < 9:
                         thinking_default = False
diff --git a/studio/frontend/src/features/chat/hooks/use-chat-model-runtime.ts b/studio/frontend/src/features/chat/hooks/use-chat-model-runtime.ts
index 037d3182ac..6c870d74a9 100644
--- a/studio/frontend/src/features/chat/hooks/use-chat-model-runtime.ts
+++ b/studio/frontend/src/features/chat/hooks/use-chat-model-runtime.ts
@@ -289,11 +289,11 @@ export function useChatModelRuntime() {
           loadedSpeculativeType: currentSpecType,
         });
 
-        // Set reasoning default for Qwen3.5 small models
+        // Set reasoning default for Qwen3.5/3.6 small models
         if (supportsReasoning) {
           let reasoningDefault = true;
           const mid = statusRes.active_model.toLowerCase();
-          if (mid.includes("qwen3.5")) {
+          if (mid.includes("qwen3.5") || mid.includes("qwen3.6")) {
             const sizeMatch = mid.match(/(\d+\.?\d*)\s*b/);
             if (sizeMatch && parseFloat(sizeMatch[1]) < 9) {
               reasoningDefault = false;
@@ -462,11 +462,11 @@ export function useChatModelRuntime() {
             setParams(
               mergeRecommendedInference(currentParams, loadResponse, modelId),
             );
-            // Qwen3.5 small models (0.8B, 2B, 4B, 9B) disable thinking by default
+            // Qwen3.5/3.6 small models (0.8B, 2B, 4B, 9B) disable thinking by default
             let reasoningDefault = loadResponse.supports_reasoning ?? false;
             if (reasoningDefault) {
               const mid = modelId.toLowerCase();
-              if (mid.includes("qwen3.5")) {
+              if (mid.includes("qwen3.5") || mid.includes("qwen3.6")) {
                 const sizeMatch = mid.match(/(\d+\.?\d*)\s*b/);
                 if (sizeMatch && parseFloat(sizeMatch[1]) < 9) {
                   reasoningDefault = false;
@@ -509,12 +509,14 @@ export function useChatModelRuntime() {
               defaultChatTemplate: loadResponse.chat_template ?? null,
               chatTemplateOverride: null,
             });
-            // Qwen3/3.5: apply thinking-mode-specific params after load
+            // Qwen3/3.5/3.6: apply thinking-mode-specific params after load
             if (modelId.toLowerCase().includes("qwen3") && (loadResponse.supports_reasoning ?? false)) {
               const store = useChatRuntimeStore.getState();
+              const mid = modelId.toLowerCase();
+              const needsPresencePenalty = mid.includes("qwen3.5") || mid.includes("qwen3.6");
               const p = reasoningDefault
-                ? { temperature: 0.6, topP: 0.95, topK: 20, minP: 0.0 }
-                : { temperature: 0.7, topP: 0.8, topK: 20, minP: 0.0 };
+                ? { temperature: 0.6, topP: 0.95, topK: 20, minP: 0.0, ...(needsPresencePenalty ? { presencePenalty: 1.5 } : {}) }
+                : { temperature: 0.7, topP: 0.8, topK: 20, minP: 0.0, ...(needsPresencePenalty ? { presencePenalty: 1.5 } : {}) };
               store.setParams({ ...store.params, ...p });
             }
             await refresh();
diff --git a/studio/frontend/src/features/chat/shared-composer.tsx b/studio/frontend/src/features/chat/shared-composer.tsx
index f82355526f..1c69ade6bd 100644
--- a/studio/frontend/src/features/chat/shared-composer.tsx
+++ b/studio/frontend/src/features/chat/shared-composer.tsx
@@ -557,13 +557,14 @@ export function SharedComposer({
               if (reasoningAlwaysOn) return;
               const next = !reasoningEnabled;
               setReasoningEnabled(next);
-              // Qwen3/3.5: adjust params for thinking on/off
+              // Qwen3/3.5/3.6: adjust params for thinking on/off
               const store = useChatRuntimeStore.getState();
               const cp = store.params.checkpoint?.toLowerCase() ?? "";
               if (cp.includes("qwen3")) {
+                const needsPresencePenalty = cp.includes("qwen3.5") || cp.includes("qwen3.6");
                 const p = next
-                  ? { temperature: 0.6, topP: 0.95, topK: 20, minP: 0.0 }
-                  : { temperature: 0.7, topP: 0.8, topK: 20, minP: 0.0 };
+                  ? { temperature: 0.6, topP: 0.95, topK: 20, minP: 0.0, ...(needsPresencePenalty ? { presencePenalty: 1.5 } : {}) }
+                  : { temperature: 0.7, topP: 0.8, topK: 20, minP: 0.0, ...(needsPresencePenalty ? { presencePenalty: 1.5 } : {}) };
                 store.setParams({ ...store.params, ...p });
               }
             }}

From d20b3067555563a001d047d829db10e21ef98036 Mon Sep 17 00:00:00 2001
From: Daniel Han <danielhanchen@gmail.com>
Date: Thu, 16 Apr 2026 12:06:10 -0700
Subject: [PATCH 133/216] Versioning

---
 pyproject.toml           | 4 ++--
 unsloth/models/_utils.py | 2 +-
 2 files changed, 3 insertions(+), 3 deletions(-)

diff --git a/pyproject.toml b/pyproject.toml
index b4c0122f4a..815c6ee119 100644
--- a/pyproject.toml
+++ b/pyproject.toml
@@ -88,7 +88,7 @@ huggingfacenotorch = [
 ]
 huggingface = [
     "unsloth[huggingfacenotorch]",
-    "unsloth_zoo>=2026.4.7",
+    "unsloth_zoo>=2026.4.8",
     "torchvision",
     "unsloth[triton]",
 ]
@@ -578,7 +578,7 @@ colab-ampere-torch220 = [
     "flash-attn>=2.6.3 ; ('linux' in sys_platform)",
 ]
 colab-new = [
-    "unsloth_zoo>=2026.4.7",
+    "unsloth_zoo>=2026.4.8",
     "packaging",
     "tyro",
     "transformers>=4.51.3,!=4.52.0,!=4.52.1,!=4.52.2,!=4.52.3,!=4.53.0,!=4.54.0,!=4.55.0,!=4.55.1,!=4.57.0,!=4.57.4,!=4.57.5,!=5.0.0,!=5.1.0,<=5.5.0",
diff --git a/unsloth/models/_utils.py b/unsloth/models/_utils.py
index 820a111a9b..34450549c5 100644
--- a/unsloth/models/_utils.py
+++ b/unsloth/models/_utils.py
@@ -12,7 +12,7 @@
 # See the License for the specific language governing permissions and
 # limitations under the License.
 
-__version__ = "2026.4.5"
+__version__ = "2026.4.6"
 
 __all__ = [
     "SUPPORTS_BFLOAT16",

From 7d0d2f256c27dffd5437b20595adbe66f9d6272e Mon Sep 17 00:00:00 2001
From: Manan Shah <52329525+Manan17@users.noreply.github.com>
Date: Fri, 17 Apr 2026 03:21:30 -0500
Subject: [PATCH 134/216] Add qwen3.6 script (#5084)
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

* unsloth gemma4 support files

* some fixes

* Fixing cache.empty() calls (#4813)

* Fixing cache.empty() calls

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

---------

Co-authored-by: Manan Shah <mananshah@Manans-MacBook-Pro.local>
Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com>

* Fix/gemma4 mlx (#4816)

* Fixing cache.empty() calls

* fixing for mlx versions

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

---------

Co-authored-by: Manan Shah <mananshah@Manans-MacBook-Pro.local>
Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com>

* removed bidirectional check for 31b (#4839)

Co-authored-by: Manan17 <shahmanan170602@gmail.coml>

* Add Gemma 4 26B MoE support (MLX) (#4844)

* removed bidirectional check for 31b

* Change gemma4_text for moe

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

---------

Co-authored-by: Manan Shah <mananshah@Manans-MacBook-Pro.local>
Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com>

* fix(gemma4): cast RoPE offset to int before mx.arange() (#4901)

* fix(gemma4): cast RoPE offset to int before mx.arange()

* fix(gemma4): use zero-based arange + offset to avoid CPU-GPU sync

* qwen3.6 patches for multi-turn chat

* qwen3.6 script

* removing unnecessary scripts

* displaying errors for not installed packages

---------

Co-authored-by: Roland Tannous <115670425+rolandtannous@users.noreply.github.com>
Co-authored-by: Manan Shah <mananshah@Manans-MacBook-Pro.local>
Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com>
Co-authored-by: Manan17 <shahmanan170602@gmail.coml>
Co-authored-by: Théophile Lafargue <138336683+eauchs@users.noreply.github.com>
---
 scripts/install_qwen3_6_mlx.sh | 191 +++++++++++++++++++++++++++++++++
 1 file changed, 191 insertions(+)
 create mode 100644 scripts/install_qwen3_6_mlx.sh

diff --git a/scripts/install_qwen3_6_mlx.sh b/scripts/install_qwen3_6_mlx.sh
new file mode 100644
index 0000000000..5ce66d29a6
--- /dev/null
+++ b/scripts/install_qwen3_6_mlx.sh
@@ -0,0 +1,191 @@
+#!/bin/bash
+set -e
+
+# ============================================================
+# Qwen3.6 MLX — One-command setup + inference
+#
+# Usage:
+#   bash install_qwen3_6_mlx.sh [--venv-dir DIR]
+#
+# This script:
+#   1. Creates a Python virtual environment
+#   2. Installs uv, mlx-vlm, transformers, torch, torchvision
+# ============================================================
+
+# ── Output style (inspired by unsloth/install.sh) ─────────────
+RULE=""
+_rule_i=0
+while [ "$_rule_i" -lt 52 ]; do
+    RULE="${RULE}─"
+    _rule_i=$((_rule_i + 1))
+done
+
+if [ -n "${NO_COLOR:-}" ]; then
+    C_TITLE= C_DIM= C_OK= C_WARN= C_ERR= C_RST=
+elif [ -t 1 ] || [ -n "${FORCE_COLOR:-}" ]; then
+    _ESC="$(printf '\033')"
+    C_TITLE="${_ESC}[38;5;117m"
+    C_DIM="${_ESC}[38;5;245m"
+    C_OK="${_ESC}[38;5;108m"
+    C_WARN="${_ESC}[38;5;136m"
+    C_ERR="${_ESC}[91m"
+    C_RST="${_ESC}[0m"
+else
+    C_TITLE= C_DIM= C_OK= C_WARN= C_ERR= C_RST=
+fi
+
+step()    { printf "  ${C_DIM}%-18.18s${C_RST}${3:-$C_OK}%s${C_RST}\n" "$1" "$2"; }
+substep() { printf "  ${C_DIM}%-18s${2:-$C_DIM}%s${C_RST}\n" "" "$1"; }
+fail()    { step "error" "$1" "$C_ERR"; exit 1; }
+
+# ── Parse flags ───────────────────────────────────────────────
+VENV_DIR=""
+_next_is_venv=false
+
+for arg in "$@"; do
+    if [ "$_next_is_venv" = true ]; then
+        VENV_DIR="$arg"
+        _next_is_venv=false
+        continue
+    fi
+    case "$arg" in
+        --venv-dir)  _next_is_venv=true ;;
+    esac
+done
+
+# Default venv location
+if [ -z "$VENV_DIR" ]; then
+    VENV_DIR="$HOME/.unsloth/unsloth_qwen3_6_mlx"
+fi
+
+# ── Banner ────────────────────────────────────────────────────
+echo ""
+printf "  ${C_TITLE}%s${C_RST}\n" "Qwen3.6 MLX Installer"
+printf "  ${C_DIM}%s${C_RST}\n" "$RULE"
+echo ""
+
+# ── Platform check ────────────────────────────────────────────
+if [ "$(uname)" != "Darwin" ]; then
+    fail "MLX requires macOS with Apple Silicon. Detected: $(uname)"
+fi
+
+_ARCH=$(uname -m)
+if [ "$_ARCH" != "arm64" ]; then
+    step "warning" "Apple Silicon recommended (detected: $_ARCH)" "$C_WARN"
+fi
+
+step "platform" "macOS ($_ARCH)"
+
+# ── Detect Python ─────────────────────────────────────────────
+PYTHON=""
+for _candidate in python3.12 python3.11 python3.13 python3; do
+    if command -v "$_candidate" >/dev/null 2>&1; then
+        PYTHON="$_candidate"
+        break
+    fi
+done
+
+if [ -z "$PYTHON" ]; then
+    fail "Python 3 not found. Install via: brew install python@3.12"
+fi
+
+_PY_VERSION=$("$PYTHON" -c "import sys; print(f'{sys.version_info.major}.{sys.version_info.minor}.{sys.version_info.micro}')")
+step "python" "$PYTHON ($_PY_VERSION)"
+
+# ── Create virtual environment ────────────────────────────────
+if [ -x "$VENV_DIR/bin/python" ]; then
+    step "venv" "using existing environment"
+    substep "$VENV_DIR"
+else
+    step "venv" "creating virtual environment"
+    substep "$VENV_DIR"
+    mkdir -p "$(dirname "$VENV_DIR")"
+    "$PYTHON" -m venv "$VENV_DIR"
+fi
+
+# ── Install uv ───────────────────────────────────────────────
+if ! command -v uv >/dev/null 2>&1; then
+    step "uv" "installing uv package manager..."
+    _uv_tmp=$(mktemp)
+    curl -LsSf "https://astral.sh/uv/install.sh" -o "$_uv_tmp"
+    sh "$_uv_tmp" </dev/null
+    rm -f "$_uv_tmp"
+    if [ -f "$HOME/.local/bin/env" ]; then
+        . "$HOME/.local/bin/env"
+    fi
+    export PATH="$HOME/.local/bin:$PATH"
+    substep "done"
+else
+    step "uv" "found $(uv --version 2>/dev/null || echo 'uv')"
+fi
+
+_VENV_PY="$VENV_DIR/bin/python"
+
+# ── Install dependencies ──────────────────────────────────────
+step "install" "installing mlx-vlm..."
+uv pip install --python "$_VENV_PY" -q mlx-vlm
+substep "done"
+
+step "install" "installing transformers>=5.2.0..."
+if uv pip install --python "$_VENV_PY" -q "transformers>=5.2.0"; then
+    substep "installed from PyPI"
+else
+    substep "PyPI install failed, trying GitHub..."
+    if uv pip install --python "$_VENV_PY" -q "git+https://github.com/huggingface/transformers.git"; then
+        substep "installed from huggingface/transformers main"
+    else
+        fail "Could not install transformers>=5.2.0 (required for Qwen3.5/3.6 model support). Please check your Python version (>=3.10 required) and network connection, then try again."
+    fi
+fi
+
+step "install" "installing torch + torchvision (needed for Qwen3 VL processor)..."
+uv pip install --python "$_VENV_PY" -q torch torchvision
+substep "done"
+
+# ── Verify installation ──────────────────────────────────────
+if "$_VENV_PY" -c "import mlx_vlm; import torch; import torchvision; import transformers"; then
+    substep "mlx-vlm + torch + transformers verified"
+else
+    fail "Installation verification failed. Please ensure Python >=3.10 and try again."
+fi
+
+# ── Apply patches for multi-turn image chat ──────────────────
+_PATCH_BASE="https://raw.githubusercontent.com/unslothai/unsloth/refs/heads/fix/ui-fix/unsloth/models/patches/mlx_vlm_qwen3_5"
+_SITE_PKGS=$("$_VENV_PY" -c "import site; print(site.getsitepackages()[0])")
+
+step "patch" "fixing multi-turn image chat..."
+
+if curl -sSLf "${_PATCH_BASE}/qwen3_5.py" -o "${_SITE_PKGS}/mlx_vlm/models/qwen3_5/qwen3_5.py"; then
+    substep "patched qwen3_5.py (MRoPE position reset)"
+else
+    step "warning" "failed to download qwen3_5.py patch — multi-turn image chat may not work" "$C_WARN"
+fi
+
+if curl -sSLf "${_PATCH_BASE}/generate.py" -o "${_SITE_PKGS}/mlx_vlm/generate.py"; then
+    substep "patched generate.py (mask trim on cache reuse)"
+else
+    step "warning" "failed to download generate.py patch — multi-turn image chat may not work" "$C_WARN"
+fi
+
+# Clear pycache so patches take effect
+find "${_SITE_PKGS}/mlx_vlm" -name "__pycache__" -type d -exec rm -rf {} + 2>/dev/null || true
+substep "cleared bytecode cache"
+
+# ── Done ──────────────────────────────────────────────────────
+echo ""
+printf "  ${C_TITLE}%s${C_RST}\n" "Qwen3.6 MLX installed!"
+printf "  ${C_DIM}%s${C_RST}\n" "$RULE"
+echo ""
+step "available models" "unsloth/Qwen3.6-35B-A3B-UD-MLX-3bit"
+substep "unsloth/Qwen3.6-35B-A3B-UD-MLX-4bit"
+substep "unsloth/Qwen3.6-35B-A3B-MLX-8bit"
+echo ""
+step "venv activate" "source ${VENV_DIR}/bin/activate"
+echo ""
+step "vision chat" "python -m mlx_vlm.chat --model unsloth/Qwen3.6-35B-A3B-UD-MLX-4bit"
+substep "Use /image path/to/image.jpg to load an image"
+echo ""
+step "gradio UI" "python -m mlx_vlm.chat_ui --model unsloth/Qwen3.6-35B-A3B-UD-MLX-4bit"
+echo ""
+printf "  ${C_DIM}%s${C_RST}\n" "$RULE"
+echo ""

From ac2daf8b7a4df093d597a9d1e5e6ffdb5cb0804c Mon Sep 17 00:00:00 2001
From: Roland Tannous <115670425+rolandtannous@users.noreply.github.com>
Date: Sat, 18 Apr 2026 12:53:23 +0400
Subject: [PATCH 135/216] Studio: forward standard OpenAI tools / tool_choice
 to llama-server (#5099)

* fix(studio): forward OpenAI tools/tool_choice to llama-server (#4999)

Studio's /v1/chat/completions silently stripped standard OpenAI `tools`
and `tool_choice` fields, so clients using standard function calling
(opencode, Claude Code, Cursor, Continue, ...) never got structured
tool_calls back. Adds a client-side pass-through path mirroring the
existing Anthropic /v1/messages flow: when `tools` is present without
Studio's `enable_tools` shorthand, the request is forwarded to
llama-server verbatim so the client sees native id, finish_reason
("tool_calls"), delta.tool_calls, and accurate usage tokens.

Also wires Anthropic tool_choice forwarding: /v1/messages previously
accepted tool_choice on the request model but silently dropped it with
a warning. Translate the four Anthropic shapes to OpenAI format and
forward them so agentic clients can actually enforce tool use.

- ChatCompletionRequest: add tools, tool_choice, stop; extra="allow"
- ChatMessage: accept role="tool", optional tool_call_id / tool_calls /
  name; content is now optional (assistant with only tool_calls)
- routes/inference.py: _openai_passthrough_stream /
  _openai_passthrough_non_streaming helpers, routing branch in
  openai_chat_completions, vision+tools via content-parts injection
- _build_passthrough_payload: tool_choice parameter (default "auto")
- anthropic_compat: anthropic_tool_choice_to_openai() translator
- tests/test_openai_tool_passthrough.py: Pydantic + translator unit tests
- tests/test_studio_api.py: 5 new E2E tests (non-stream, stream,
  multi-turn, OpenAI SDK, Anthropic tool_choice=any regression)

* fix(studio): surface httpx transport errors from OpenAI passthrough

When the managed llama-server subprocess crashes mid-request, the
async pass-through helpers in routes/inference.py used to return a
bare 500 (non-streaming) or an "An internal error occurred" SSE chunk
(streaming) because _friendly_error only recognized the sync path's
"Lost connection to llama-server" substring -- httpx transport
failures (ConnectError / ReadError / RemoteProtocolError /
ReadTimeout) stringify differently and fell through to the generic
case.

- _friendly_error: map any httpx.RequestError subclass to the same
  "Lost connection to the model server" message the sync chat path
  emits. Placed before the substring heuristics so the streaming path
  automatically picks it up via its existing except Exception catch.
- _openai_passthrough_non_streaming: wrap the httpx.AsyncClient.post
  in a try/except httpx.RequestError and re-raise as HTTPException
  502 with the friendly detail.
- tests/test_openai_tool_passthrough.py: new TestFriendlyErrorHttpx
  class pinning the mapping for ConnectError, ReadError,
  RemoteProtocolError, ReadTimeout, and confirming non-httpx paths
  (context-size heuristic, generic fallback) are unchanged.

* fix(studio): close aiter_bytes/aiter_lines explicitly in passthroughs

The httpcore asyncgen cleanup fix in 5cedd9a5 is incomplete on Python
3.13 + httpcore 1.0.x: it switched to manual client/response lifecycle
but still used anonymous `async for raw_line in resp.aiter_lines():`
patterns in all three streaming paths. Python's async for does NOT
auto-close the iterator on break/return, so the aiter_lines /
aiter_bytes async generator remains alive, reachable only from the
surrounding coroutine frame. Once `_stream()` returns the frame is
GC'd and the orphaned asyncgen is finalized on a LATER GC pass in a
DIFFERENT asyncio task, where httpcore's
HTTP11ConnectionByteStream.aclose() enters anyio.CancelScope.__exit__
with a mismatched task and prints "Exception ignored in: <async
generator>" / "async generator ignored GeneratorExit" / "Attempted
to exit cancel scope in a different task" to the server log.

User observed this on /v1/messages after successful (status 200)
requests, with the traceback pointing at HTTP11ConnectionByteStream
.__aiter__ / .aclose inside httpcore.

Fix: save resp.aiter_lines() / resp.aiter_bytes() as a variable and
explicitly `await iter.aclose()` in the finally block BEFORE
resp.aclose() / client.aclose(). This closes the asyncgen inside the
current task's event loop, so the internal httpcore byte stream is
cleaned up before Python's asyncgen GC hook has anything orphaned to
finalize. Each aclose is wrapped in try/except Exception so nested
anyio cleanup noise can't bubble out.

Applied to all three streaming passthrough paths:
- _anthropic_passthrough_stream (/v1/messages client-side tool path)
- _openai_passthrough_stream (/v1/chat/completions client-side tool
  path, new in this PR)
- openai_completions (/v1/completions bytes proxy from PR #4956)

* fix(studio): default ChatCompletionRequest.stream to false per OpenAI spec

OpenAI's /v1/chat/completions spec defaults `stream` to false, so
clients that omit the field (naive curl, minimal integrations) expect
a single JSON response back. Studio was defaulting to true, silently
switching those clients into SSE and breaking any parser that didn't
also handle streaming. ResponsesRequest and AnthropicMessagesRequest
already default to false correctly; only ChatCompletionRequest was
wrong.

Studio's own frontend always sets `stream` explicitly on every
chat-adapter / chat-api / runtime-provider call site, so the flip has
no UI impact. SDK users (OpenAI Python/JS SDK, opencode, Claude Code,
Cursor, Continue) also always pass `stream` explicitly, so they're
unaffected. The only clients feeling the change are raw-curl users
who were relying on the wrong default -- those get the correct OpenAI
behavior now.

Added a regression test pinning the default so it can't silently
flip back.

* fix(studio): reject images in OpenAI tool passthrough for text-only GGUFs

The new tool passthrough branch runs before _extract_content_parts,
skipping the existing not is_vision guard. Requests combining tools
with an image on a text-only tool-capable GGUF were forwarded to
llama-server, producing opaque upstream errors instead of the
pre-existing clear 400. Restore the guard inline at the dispatch
point, checking both legacy image_base64 and inline image_url parts.

* fix(studio): require tool_call_id on role=tool chat messages

Enforce the OpenAI spec rule that role="tool" messages must carry a
tool_call_id. Without it, upstream backends cannot associate a tool
result with the assistant's prior tool_calls entry and the request
fails in non-obvious ways through the passthrough path. Reject at the
request boundary with a 422 instead.

* fix(studio): harden OpenAI tool passthrough validation and error surfacing

Three related fixes called out by the PR review:

1. Preserve upstream status codes in the streaming passthrough. The
   httpx request is now dispatched before the StreamingResponse is
   constructed. Non-200 upstream responses and httpx RequestError
   transport failures raise HTTPException with the real status
   instead of being buried inside a 200 SSE error frame, so OpenAI
   SDK clients see APIError/BadRequestError/... as expected.

2. Require non-empty content on user/system/tool messages. Per the
   OpenAI spec, content may only be omitted on assistant messages
   that carry tool_calls; enforce that at the request boundary so
   malformed messages never reach the passthrough path.

3. Role-constrain tool-call metadata. tool_calls is only valid on
   role=assistant, tool_call_id and name only on role=tool. Without
   this, a user/system message with tool_calls would flip the
   passthrough branch on and be forwarded to llama-server, surfacing
   as an opaque upstream error.

* fix(studio): normalize image mode and passthrough JSON verbatim

Two Gemini-code-assist review findings on PR #5099:

1. Unconditionally convert decoded images to RGB before PNG encoding.
   The prior code only handled RGBA, letting CMYK/I/F images crash
   at img.save(format="PNG") and surface as opaque 400s. Applied to
   both the passthrough helper and the non-passthrough GGUF path
   that originally carried this pattern, keeping the two sites in
   sync.

2. Return the upstream JSON body as raw bytes via Response rather
   than parse-then-re-serialize with JSONResponse. Matches the
   passthrough helper's "verbatim" contract and drops a redundant
   round-trip.

---------

Co-authored-by: Lee Jackson <130007945+Imagineer99@users.noreply.github.com>
Co-authored-by: Daniel Han <danielhanchen@gmail.com>
---
 .../core/inference/anthropic_compat.py        |  33 ++
 studio/backend/models/inference.py            |  95 +++-
 studio/backend/routes/inference.py            | 439 +++++++++++++++--
 .../tests/test_openai_tool_passthrough.py     | 465 ++++++++++++++++++
 studio/backend/tests/test_studio_api.py       | 367 +++++++++++++-
 5 files changed, 1327 insertions(+), 72 deletions(-)
 create mode 100644 studio/backend/tests/test_openai_tool_passthrough.py

diff --git a/studio/backend/core/inference/anthropic_compat.py b/studio/backend/core/inference/anthropic_compat.py
index e7b40a60ce..400e6053e0 100644
--- a/studio/backend/core/inference/anthropic_compat.py
+++ b/studio/backend/core/inference/anthropic_compat.py
@@ -114,6 +114,39 @@ def anthropic_tools_to_openai(tools: list) -> list[dict]:
     return result
 
 
+def anthropic_tool_choice_to_openai(tc: Any) -> Any:
+    """Translate Anthropic `tool_choice` into OpenAI `tool_choice`.
+
+    Anthropic formats (all dict shapes with a ``type`` discriminator):
+
+    - ``{"type": "auto"}``                       → ``"auto"``
+    - ``{"type": "any"}``                        → ``"required"``
+    - ``{"type": "none"}``                       → ``"none"``
+    - ``{"type": "tool", "name": "get_weather"}``
+          → ``{"type": "function", "function": {"name": "get_weather"}}``
+
+    Returns ``None`` for ``None`` or any unrecognized shape (caller may
+    then fall back to its own default, typically ``"auto"``).
+    """
+    if tc is None:
+        return None
+    if not isinstance(tc, dict):
+        return None
+    t = tc.get("type")
+    if t == "auto":
+        return "auto"
+    if t == "any":
+        return "required"
+    if t == "none":
+        return "none"
+    if t == "tool":
+        name = tc.get("name")
+        if not name:
+            return None
+        return {"type": "function", "function": {"name": name}}
+    return None
+
+
 def build_anthropic_sse_event(event_type: str, data: dict) -> str:
     """Format a single Anthropic SSE event."""
     return f"event: {event_type}\ndata: {json.dumps(data)}\n\n"
diff --git a/studio/backend/models/inference.py b/studio/backend/models/inference.py
index 4917a14579..956b99f75c 100644
--- a/studio/backend/models/inference.py
+++ b/studio/backend/models/inference.py
@@ -11,7 +11,7 @@ import time
 import uuid
 from typing import Annotated, Any, Dict, Literal, Optional, List, Union
 
-from pydantic import BaseModel, Discriminator, Field, Tag
+from pydantic import BaseModel, Discriminator, Field, Tag, model_validator
 
 
 class LoadRequest(BaseModel):
@@ -338,14 +338,68 @@ class ChatMessage(BaseModel):
 
     ``content`` may be a plain string (text-only) or a list of
     content parts for multimodal messages (OpenAI vision format).
+    Assistant messages that only contain tool calls may set ``content``
+    to ``None`` with ``tool_calls`` populated. ``role="tool"`` messages
+    carry the result of a client-executed tool call and require
+    ``tool_call_id`` per the OpenAI spec.
     """
 
-    role: Literal["system", "user", "assistant"] = Field(
+    role: Literal["system", "user", "assistant", "tool"] = Field(
         ..., description = "Message role"
     )
-    content: Union[str, list[ContentPart]] = Field(
-        ..., description = "Message content (string or multimodal parts)"
+    content: Optional[Union[str, list[ContentPart]]] = Field(
+        None, description = "Message content (string or multimodal parts)"
     )
+    tool_call_id: Optional[str] = Field(
+        None,
+        description = "OpenAI tool-result messages: id of the tool call this result belongs to.",
+    )
+    tool_calls: Optional[list[dict]] = Field(
+        None,
+        description = "OpenAI assistant messages: structured tool calls the model decided to make.",
+    )
+    name: Optional[str] = Field(
+        None,
+        description = "OpenAI tool-result messages: name of the tool whose result this is.",
+    )
+
+    @model_validator(mode = "after")
+    def _validate_role_shape(self) -> "ChatMessage":
+        # Enforce the per-role OpenAI spec shape at the request boundary.
+        # Without this, malformed messages (e.g. user entries with no
+        # content, tool_calls on a user/system role, role="tool" without
+        # tool_call_id) would be silently forwarded to llama-server via
+        # the passthrough path, surfacing as opaque upstream errors or
+        # broken tool-call reconciliation downstream.
+
+        # Tool-call metadata must appear only on the appropriate role.
+        if self.tool_calls is not None and self.role != "assistant":
+            raise ValueError('"tool_calls" is only valid on role="assistant" messages.')
+        if self.tool_call_id is not None and self.role != "tool":
+            raise ValueError('"tool_call_id" is only valid on role="tool" messages.')
+        if self.name is not None and self.role != "tool":
+            raise ValueError('"name" is only valid on role="tool" messages.')
+
+        # Per-role content requirements.
+        if self.role == "tool":
+            if not self.tool_call_id:
+                raise ValueError(
+                    'role="tool" messages require "tool_call_id" per the OpenAI spec.'
+                )
+            if not self.content:
+                raise ValueError('role="tool" messages require non-empty "content".')
+        elif self.role == "assistant":
+            # Assistant messages may omit content when tool_calls is set.
+            if not self.content and not self.tool_calls:
+                raise ValueError(
+                    'role="assistant" messages require either "content" or "tool_calls".'
+                )
+        else:  # "user" | "system"
+            if not self.content:
+                raise ValueError(
+                    f'role="{self.role}" messages require non-empty "content".'
+                )
+        return self
 
 
 class ChatCompletionRequest(BaseModel):
@@ -355,18 +409,49 @@ class ChatCompletionRequest(BaseModel):
     Extensions (non-OpenAI fields) are marked with 'x-unsloth'.
     """
 
+    # Accept unknown fields defensively so future OpenAI fields (seed,
+    # response_format, logprobs, frequency_penalty, etc.) don't get
+    # silently dropped by Pydantic before route code runs. Mirrors
+    # AnthropicMessagesRequest and ResponsesRequest.
+    model_config = {"extra": "allow"}
+
     model: str = Field(
         "default",
         description = "Model identifier (informational; the active model is used)",
     )
     messages: list[ChatMessage] = Field(..., description = "Conversation messages")
-    stream: bool = Field(True, description = "Whether to stream the response via SSE")
+    stream: bool = Field(
+        False,
+        description = (
+            "Whether to stream the response via SSE. Default matches OpenAI's "
+            "spec (`false`); opt into streaming by sending `stream: true`."
+        ),
+    )
     temperature: float = Field(0.6, ge = 0.0, le = 2.0)
     top_p: float = Field(0.95, ge = 0.0, le = 1.0)
     max_tokens: Optional[int] = Field(
         None, ge = 1, description = "Maximum tokens to generate (None = until EOS)"
     )
     presence_penalty: float = Field(0.0, ge = 0.0, le = 2.0, description = "Presence penalty")
+    stop: Optional[Union[str, list[str]]] = Field(
+        None,
+        description = "OpenAI stop sequences: a single string or list of strings at which generation halts.",
+    )
+    tools: Optional[list[dict]] = Field(
+        None,
+        description = (
+            "OpenAI function-tool definitions. When provided without `enable_tools=true`, "
+            "Studio forwards the tools to the backend so the model returns structured "
+            "tool_calls for the client to execute (standard OpenAI function calling)."
+        ),
+    )
+    tool_choice: Optional[Union[str, dict]] = Field(
+        None,
+        description = (
+            "OpenAI tool choice: 'auto' | 'required' | 'none' | "
+            "{'type': 'function', 'function': {'name': ...}}"
+        ),
+    )
 
     # ── Unsloth extensions (ignored by standard OpenAI clients) ──
     top_k: int = Field(20, ge = -1, le = 100, description = "[x-unsloth] Top-k sampling")
diff --git a/studio/backend/routes/inference.py b/studio/backend/routes/inference.py
index 4246f0056b..6103c90915 100644
--- a/studio/backend/routes/inference.py
+++ b/studio/backend/routes/inference.py
@@ -29,6 +29,14 @@ from utils.models import extract_model_size_b as _extract_model_size_b
 
 def _friendly_error(exc: Exception) -> str:
     """Extract a user-friendly message from known llama-server errors."""
+    # httpx transport-layer failures reaching the managed llama-server —
+    # raised by the async pass-through helpers that talk to llama-server
+    # directly. Treat any RequestError subclass (ConnectError, ReadError,
+    # RemoteProtocolError, WriteError, PoolTimeout, ...) as "the upstream
+    # subprocess is unreachable", which for Studio always means the
+    # llama-server subprocess crashed or is still coming up.
+    if isinstance(exc, httpx.RequestError):
+        return "Lost connection to the model server. It may have crashed -- try reloading the model."
     msg = str(exc)
     m = _re.search(
         r"request \((\d+) tokens?\) exceeds the available context size \((\d+) tokens?\)",
@@ -106,6 +114,7 @@ from models.inference import (
 from core.inference.anthropic_compat import (
     anthropic_messages_to_openai,
     anthropic_tools_to_openai,
+    anthropic_tool_choice_to_openai,
     AnthropicStreamEmitter,
     AnthropicPassthroughEmitter,
 )
@@ -1122,6 +1131,56 @@ async def openai_chat_completions(
                 )
                 return JSONResponse(content = response.model_dump())
 
+    # ── Standard OpenAI function-calling pass-through (GGUF only) ────
+    # When a client (opencode / Claude Code via OpenAI compat / Cursor /
+    # Continue / ...) sends standard OpenAI `tools` without Studio's
+    # `enable_tools` shorthand, forward the request to llama-server
+    # verbatim so structured `tool_calls` flow back to the client. This
+    # branch runs BEFORE `_extract_content_parts` because that helper is
+    # unaware of `role="tool"` messages and assistant messages that only
+    # carry `tool_calls` (content=None) — both of which are valid in
+    # multi-turn client-side tool loops.
+    _has_tool_messages = any(m.role == "tool" or m.tool_calls for m in payload.messages)
+    if (
+        using_gguf
+        and llama_backend.supports_tools
+        and not payload.enable_tools
+        and ((payload.tools and len(payload.tools) > 0) or _has_tool_messages)
+    ):
+        # Preserve the vision guard that would otherwise run in the
+        # non-passthrough path below: text-only tool-capable GGUFs
+        # should return a clear 400 here rather than forwarding the
+        # image to llama-server and surfacing an opaque upstream error.
+        if not llama_backend.is_vision and (
+            payload.image_base64
+            or any(
+                isinstance(m.content, list)
+                and any(isinstance(p, ImageContentPart) for p in m.content)
+                for m in payload.messages
+            )
+        ):
+            raise HTTPException(
+                status_code = 400,
+                detail = "Image provided but current GGUF model does not support vision.",
+            )
+
+        cancel_event = threading.Event()
+        completion_id = f"chatcmpl-{uuid.uuid4().hex[:12]}"
+        if payload.stream:
+            return await _openai_passthrough_stream(
+                request,
+                cancel_event,
+                llama_backend,
+                payload,
+                model_name,
+                completion_id,
+            )
+        return await _openai_passthrough_non_streaming(
+            llama_backend,
+            payload,
+            model_name,
+        )
+
     # ── Parse messages (handles multimodal content parts) ─────
     system_prompt, chat_messages, extracted_image_b64 = _extract_content_parts(
         payload.messages
@@ -1151,9 +1210,11 @@ async def openai_chat_completions(
                 from PIL import Image as _Image
 
                 raw = _b64.b64decode(image_b64)
-                img = _Image.open(_BytesIO(raw))
-                if img.mode == "RGBA":
-                    img = img.convert("RGB")
+                # Normalize to RGB so PNG encoding succeeds regardless of
+                # source mode (RGBA, P, L, CMYK, I, F, ...). Previously
+                # we only converted RGBA, which left CMYK/I/F to raise at
+                # img.save(PNG).
+                img = _Image.open(_BytesIO(raw)).convert("RGB")
                 buf = _BytesIO()
                 img.save(buf, format = "PNG")
                 image_b64 = _b64.b64encode(buf.getvalue()).decode("ascii")
@@ -1933,25 +1994,32 @@ async def openai_completions(
     if is_stream:
 
         async def _stream():
-            # Manual httpx client/response lifecycle — see
-            # _anthropic_passthrough_stream for the full rationale. Briefly:
-            # `async with` inside an async generator causes
-            # "Attempted to exit cancel scope in a different task" /
-            # "async generator ignored GeneratorExit" on Python 3.13 +
-            # httpcore 1.0.x when the generator is orphaned and finalized
-            # by GC. Closing via a finally block that catches Exception
-            # (but not BaseException) suppresses the anyio cleanup noise
-            # while letting GeneratorExit propagate cleanly.
+            # Manual httpx client/response lifecycle AND explicit
+            # aiter_bytes() iterator close — see _anthropic_passthrough_stream
+            # for the full rationale. Saving `bytes_iter = resp.aiter_bytes()`
+            # and `await bytes_iter.aclose()` in the finally block is the
+            # part that matters for avoiding the Python 3.13 + httpcore
+            # 1.0.x "Exception ignored in: <async_generator>" / anyio
+            # cancel-scope trace: an anonymous async for leaves the
+            # iterator unclosed, so Python's asyncgen GC finalizer runs
+            # cleanup on a later pass in a different asyncio task.
             client = httpx.AsyncClient(timeout = 600)
             resp = None
+            bytes_iter = None
             try:
                 req = client.build_request("POST", target_url, json = body)
                 resp = await client.send(req, stream = True)
-                async for chunk in resp.aiter_bytes():
+                bytes_iter = resp.aiter_bytes()
+                async for chunk in bytes_iter:
                     yield chunk
             except Exception as e:
                 logger.error("openai_completions stream error: %s", e)
             finally:
+                if bytes_iter is not None:
+                    try:
+                        await bytes_iter.aclose()
+                    except Exception:
+                        pass
                 if resp is not None:
                     try:
                         await resp.aclose()
@@ -2339,22 +2407,12 @@ async def anthropic_messages(
     )
     stop = payload.stop_sequences or None
 
-    # tool_choice is declared on AnthropicMessagesRequest for Anthropic SDK
-    # compatibility (the SDK often sets it by default), but it is not
-    # currently honored by Unsloth's backend. Warn once per request so the
-    # silent drop is visible to operators instead of looking like a model
-    # quality issue to clients.
-    if payload.tool_choice is not None:
-        logger.warning(
-            "anthropic_messages.tool_choice_ignored",
-            tool_choice = payload.tool_choice,
-            note = (
-                "tool_choice is accepted for Anthropic SDK compatibility but not "
-                "honored by Unsloth. Use enable_tools / enabled_tools (server-side "
-                "built-in tools) or restrict the `tools` array (client-side) to "
-                "control which tools the model sees."
-            ),
-        )
+    # Translate Anthropic tool_choice to OpenAI format for forwarding to
+    # llama-server. Falls back to "auto" when unset or unrecognized, which
+    # matches the prior hardcoded behavior.
+    openai_tool_choice = anthropic_tool_choice_to_openai(payload.tool_choice)
+    if openai_tool_choice is None:
+        openai_tool_choice = "auto"
 
     cancel_event = threading.Event()
 
@@ -2392,6 +2450,7 @@ async def anthropic_messages(
                 min_p = min_p,
                 repetition_penalty = repetition_penalty,
                 presence_penalty = presence_penalty,
+                tool_choice = openai_tool_choice,
             )
         return await _anthropic_passthrough_non_streaming(
             llama_backend,
@@ -2407,6 +2466,7 @@ async def anthropic_messages(
             min_p = min_p,
             repetition_penalty = repetition_penalty,
             presence_penalty = presence_penalty,
+            tool_choice = openai_tool_choice,
         )
 
     if server_tools:
@@ -2750,11 +2810,12 @@ def _build_passthrough_payload(
     min_p = None,
     repetition_penalty = None,
     presence_penalty = None,
+    tool_choice = "auto",
 ):
     body = {
         "messages": openai_messages,
         "tools": openai_tools,
-        "tool_choice": "auto",
+        "tool_choice": tool_choice,
         "temperature": temperature,
         "top_p": top_p,
         "top_k": top_k,
@@ -2792,6 +2853,7 @@ async def _anthropic_passthrough_stream(
     min_p = None,
     repetition_penalty = None,
     presence_penalty = None,
+    tool_choice = "auto",
 ):
     """Streaming client-side pass-through: forward tools to llama-server and
     translate its streaming response to Anthropic SSE without executing anything."""
@@ -2808,6 +2870,7 @@ async def _anthropic_passthrough_stream(
         min_p = min_p,
         repetition_penalty = repetition_penalty,
         presence_penalty = presence_penalty,
+        tool_choice = tool_choice,
     )
 
     async def _stream():
@@ -2815,33 +2878,42 @@ async def _anthropic_passthrough_stream(
         for line in emitter.start(message_id, model_name):
             yield line
 
-        # Manage the httpx client and response MANUALLY — no `async with`.
+        # Manage the httpx client, response, AND the aiter_lines() async
+        # generator MANUALLY — no `async with`, no anonymous iterator.
         #
-        # On Python 3.13 + httpcore 1.0.x, an orphaned async generator (e.g.
-        # when the client disconnects mid-stream and Starlette drops the
-        # StreamingResponse iterator without explicitly calling aclose())
-        # is finalized by Python's asyncgen GC hook in a DIFFERENT asyncio
-        # task than the one that originally entered the httpx context
-        # managers. When `async with` exits run in the wrong task, httpcore's
-        # internal `HTTP11ConnectionByteStream.aclose()` hits
-        # `anyio.CancelScope.__exit__` with a mismatched task and raises
-        # RuntimeError("Attempted to exit cancel scope in a different task"),
-        # which escapes as "Exception ignored in:" because it happens during
-        # GC finalization outside any user-owned try/except.
+        # On Python 3.13 + httpcore 1.0.x, `async for raw_line in
+        # resp.aiter_lines():` creates an anonymous async generator. When
+        # the loop exits via `break` (or the generator is orphaned when a
+        # client disconnects mid-stream), Python's `async for` protocol
+        # does NOT auto-close the iterator the way a sync `for` loop
+        # would. The iterator remains reachable only from the current
+        # coroutine frame; once `_stream()` returns, the frame is GC'd
+        # and the iterator becomes unreachable. Python's asyncgen
+        # finalizer hook then runs its aclose() on a LATER GC pass in a
+        # DIFFERENT asyncio task, where httpcore's
+        # `HTTP11ConnectionByteStream.aclose()` enters
+        # `anyio.CancelScope.__exit__` with a mismatched task and prints
+        # `RuntimeError: Attempted to exit cancel scope in a different
+        # task` / `RuntimeError: async generator ignored GeneratorExit`
+        # as "Exception ignored in:" unraisable warnings.
         #
-        # The fix: do not use `async with` for the client/response. Close
-        # them in a finally block wrapped in `try: ... except Exception: pass`.
-        # This narrowly suppresses RuntimeError / other Exception subclasses
-        # from the anyio cleanup noise while letting GeneratorExit (a
-        # BaseException, not Exception) propagate through cleanly so the
-        # generator terminates as Python expects.
+        # The fix: save `resp.aiter_lines()` as `lines_iter`, and in the
+        # finally block explicitly `await lines_iter.aclose()` BEFORE
+        # `resp.aclose()` / `client.aclose()`. This closes the iterator
+        # inside our own task's event loop, so the internal httpcore
+        # byte-stream is cleaned up before Python's asyncgen finalizer
+        # has anything orphaned to finalize. Each aclose is wrapped in
+        # `try: ... except Exception: pass` so anyio cleanup noise from
+        # nested aclose paths can't bubble out.
         client = httpx.AsyncClient(timeout = 600)
         resp = None
+        lines_iter = None
         try:
             req = client.build_request("POST", target_url, json = body)
             resp = await client.send(req, stream = True)
 
-            async for raw_line in resp.aiter_lines():
+            lines_iter = resp.aiter_lines()
+            async for raw_line in lines_iter:
                 if await request.is_disconnected():
                     cancel_event.set()
                     break
@@ -2859,6 +2931,11 @@ async def _anthropic_passthrough_stream(
         except Exception as e:
             logger.error("anthropic_messages passthrough stream error: %s", e)
         finally:
+            if lines_iter is not None:
+                try:
+                    await lines_iter.aclose()
+                except Exception:
+                    pass
             if resp is not None:
                 try:
                     await resp.aclose()
@@ -2897,6 +2974,7 @@ async def _anthropic_passthrough_non_streaming(
     min_p = None,
     repetition_penalty = None,
     presence_penalty = None,
+    tool_choice = "auto",
 ):
     """Non-streaming client-side pass-through."""
     target_url = f"{llama_backend.base_url}/v1/chat/completions"
@@ -2912,6 +2990,7 @@ async def _anthropic_passthrough_non_streaming(
         min_p = min_p,
         repetition_penalty = repetition_penalty,
         presence_penalty = presence_penalty,
+        tool_choice = tool_choice,
     )
 
     async with httpx.AsyncClient() as client:
@@ -2969,3 +3048,265 @@ async def _anthropic_passthrough_non_streaming(
         ),
     )
     return JSONResponse(content = resp_obj.model_dump())
+
+
+# =====================================================================
+# Client-side tool pass-through (OpenAI-native /v1/chat/completions)
+# =====================================================================
+
+
+def _openai_messages_for_passthrough(payload) -> list[dict]:
+    """Build OpenAI-format message dicts for the /v1/chat/completions
+    passthrough path.
+
+    Messages from ``payload.messages`` are dumped through Pydantic (dropping
+    unset optional fields) so they are already in standard OpenAI format
+    — including ``role="tool"`` tool-result messages and assistant messages
+    that carry structured ``tool_calls``. Content-parts images already in
+    the message list are left untouched.
+
+    When a client uses Studio's legacy ``image_base64`` top-level field, the
+    image is re-encoded to PNG (llama-server's stb_image has limited format
+    support) and spliced into the last user message as an OpenAI
+    ``image_url`` content part so vision + function-calling requests work
+    transparently.
+    """
+    messages = [m.model_dump(exclude_none = True) for m in payload.messages]
+
+    if not payload.image_base64:
+        return messages
+
+    try:
+        import base64 as _b64
+        from io import BytesIO as _BytesIO
+        from PIL import Image as _Image
+
+        raw = _b64.b64decode(payload.image_base64)
+        img = _Image.open(_BytesIO(raw)).convert("RGB")
+        buf = _BytesIO()
+        img.save(buf, format = "PNG")
+        png_b64 = _b64.b64encode(buf.getvalue()).decode("ascii")
+    except Exception as e:
+        raise HTTPException(
+            status_code = 400,
+            detail = f"Failed to process image: {e}",
+        )
+
+    data_url = f"data:image/png;base64,{png_b64}"
+    image_part = {"type": "image_url", "image_url": {"url": data_url}}
+
+    for msg in reversed(messages):
+        if msg.get("role") != "user":
+            continue
+        existing = msg.get("content")
+        if isinstance(existing, str):
+            msg["content"] = [{"type": "text", "text": existing}, image_part]
+        elif isinstance(existing, list):
+            existing.append(image_part)
+        else:
+            msg["content"] = [image_part]
+        break
+    else:
+        messages.append({"role": "user", "content": [image_part]})
+
+    return messages
+
+
+def _build_openai_passthrough_body(payload) -> dict:
+    """Assemble the llama-server request body from a ChatCompletionRequest.
+
+    Only explicitly-known OpenAI / llama-server fields are forwarded so that
+    Studio-specific extensions (``enable_tools``, ``enabled_tools``,
+    ``session_id``, ...) never leak to the backend.
+    """
+    messages = _openai_messages_for_passthrough(payload)
+    tool_choice = payload.tool_choice if payload.tool_choice is not None else "auto"
+    return _build_passthrough_payload(
+        messages,
+        payload.tools,
+        payload.temperature,
+        payload.top_p,
+        payload.top_k,
+        payload.max_tokens,
+        payload.stream,
+        stop = payload.stop,
+        min_p = payload.min_p,
+        repetition_penalty = payload.repetition_penalty,
+        presence_penalty = payload.presence_penalty,
+        tool_choice = tool_choice,
+    )
+
+
+async def _openai_passthrough_stream(
+    request,
+    cancel_event,
+    llama_backend,
+    payload,
+    model_name,
+    completion_id,
+):
+    """Streaming client-side pass-through for /v1/chat/completions.
+
+    Forwards the client's OpenAI function-calling request to llama-server and
+    relays the SSE stream back verbatim. This preserves llama-server's
+    native response ``id``, ``finish_reason`` (including ``"tool_calls"``),
+    ``delta.tool_calls``, and the trailing ``usage`` chunk so the client
+    observes a standard OpenAI response.
+    """
+    target_url = f"{llama_backend.base_url}/v1/chat/completions"
+    body = _build_openai_passthrough_body(payload)
+
+    # Dispatch the upstream request BEFORE returning StreamingResponse so
+    # transport errors and non-200 upstream statuses surface as real HTTP
+    # errors to the client. OpenAI SDKs rely on status codes to raise
+    # ``APIError``/``BadRequestError``/...; burying the failure inside a
+    # 200 SSE ``error`` frame silently breaks their error handling.
+    client = httpx.AsyncClient(timeout = 600)
+    resp = None
+    try:
+        req = client.build_request("POST", target_url, json = body)
+        resp = await client.send(req, stream = True)
+    except httpx.RequestError as e:
+        # llama-server subprocess crashed / still starting / unreachable.
+        logger.error("openai passthrough stream: upstream unreachable: %s", e)
+        if resp is not None:
+            try:
+                await resp.aclose()
+            except Exception:
+                pass
+        try:
+            await client.aclose()
+        except Exception:
+            pass
+        raise HTTPException(
+            status_code = 502,
+            detail = _friendly_error(e),
+        )
+
+    if resp.status_code != 200:
+        err_bytes = await resp.aread()
+        err_text = err_bytes.decode("utf-8", errors = "replace")
+        logger.error(
+            "openai passthrough upstream error: status=%s body=%s",
+            resp.status_code,
+            err_text[:500],
+        )
+        upstream_status = resp.status_code
+        try:
+            await resp.aclose()
+        except Exception:
+            pass
+        try:
+            await client.aclose()
+        except Exception:
+            pass
+        raise HTTPException(
+            status_code = upstream_status,
+            detail = f"llama-server error: {err_text[:500]}",
+        )
+
+    async def _stream():
+        # Same httpx lifecycle pattern as _anthropic_passthrough_stream:
+        # avoid `async with` on the client/response AND explicitly save
+        # resp.aiter_lines() so we can close it ourselves in the finally
+        # block. See the long comment there for the full rationale on
+        # why the anonymous `async for raw_line in resp.aiter_lines():`
+        # pattern leaks an unclosed async generator that Python's
+        # asyncgen GC hook then finalizes in a different asyncio task,
+        # producing "Exception ignored in:" / "async generator ignored
+        # GeneratorExit" / anyio cancel-scope traces on Python 3.13 +
+        # httpcore 1.0.x.
+        lines_iter = None
+        try:
+            lines_iter = resp.aiter_lines()
+            async for raw_line in lines_iter:
+                if await request.is_disconnected():
+                    cancel_event.set()
+                    break
+                if not raw_line:
+                    continue
+                if not raw_line.startswith("data: "):
+                    continue
+                # Relay the llama-server SSE chunk verbatim so the client
+                # sees its native `id`, `finish_reason`, `delta.tool_calls`,
+                # and final `usage` unchanged.
+                yield raw_line + "\n\n"
+                if raw_line[6:].strip() == "[DONE]":
+                    break
+        except Exception as e:
+            # Mid-stream failures still have to be reported inside the SSE
+            # body because the 200 response headers have already been
+            # committed by the time the first chunk flushes.
+            logger.error("openai passthrough stream error: %s", e)
+            err = {
+                "error": {
+                    "message": _friendly_error(e),
+                    "type": "server_error",
+                },
+            }
+            yield f"data: {json.dumps(err)}\n\n"
+        finally:
+            if lines_iter is not None:
+                try:
+                    await lines_iter.aclose()
+                except Exception:
+                    pass
+            try:
+                await resp.aclose()
+            except Exception:
+                pass
+            try:
+                await client.aclose()
+            except Exception:
+                pass
+
+    return StreamingResponse(
+        _stream(),
+        media_type = "text/event-stream",
+        headers = {
+            "Cache-Control": "no-cache",
+            "Connection": "keep-alive",
+            "X-Accel-Buffering": "no",
+        },
+    )
+
+
+async def _openai_passthrough_non_streaming(
+    llama_backend,
+    payload,
+    model_name,
+):
+    """Non-streaming client-side pass-through for /v1/chat/completions.
+
+    Returns llama-server's JSON response verbatim (via JSONResponse) so the
+    client sees the native response ``id``, ``finish_reason`` (including
+    ``"tool_calls"``), structured ``tool_calls``, and accurate ``usage``
+    token counts.
+    """
+    target_url = f"{llama_backend.base_url}/v1/chat/completions"
+    body = _build_openai_passthrough_body(payload)
+
+    try:
+        async with httpx.AsyncClient() as client:
+            resp = await client.post(target_url, json = body, timeout = 600)
+    except httpx.RequestError as e:
+        # llama-server subprocess crashed / still starting / unreachable.
+        # Surface the same friendly message the sync chat path emits so
+        # operators don't see a bare 500 with no diagnostic.
+        logger.error("openai passthrough non-streaming: upstream unreachable: %s", e)
+        raise HTTPException(
+            status_code = 502,
+            detail = _friendly_error(e),
+        )
+
+    if resp.status_code != 200:
+        raise HTTPException(
+            status_code = resp.status_code,
+            detail = f"llama-server error: {resp.text[:500]}",
+        )
+
+    # Pass the upstream body through as raw bytes — skips a redundant
+    # parse+re-serialize round-trip and keeps the response truly
+    # verbatim (matches the docstring). Status is guaranteed 200 by
+    # the check above.
+    return Response(content = resp.content, media_type = "application/json")
diff --git a/studio/backend/tests/test_openai_tool_passthrough.py b/studio/backend/tests/test_openai_tool_passthrough.py
new file mode 100644
index 0000000000..ccb0dba325
--- /dev/null
+++ b/studio/backend/tests/test_openai_tool_passthrough.py
@@ -0,0 +1,465 @@
+# SPDX-License-Identifier: AGPL-3.0-only
+# Copyright 2026-present the Unsloth AI Inc. team. All rights reserved.
+
+"""
+Tests for the OpenAI /v1/chat/completions client-side tool pass-through.
+
+Covers:
+- ChatCompletionRequest accepts standard OpenAI `tools` / `tool_choice` / `stop`.
+- ChatMessage accepts role="tool" with `tool_call_id` and role="assistant"
+  with `content: None` + `tool_calls`.
+- ChatCompletionRequest carries unknown fields via `extra="allow"`.
+- anthropic_tool_choice_to_openai() covers all four Anthropic shapes.
+- _build_passthrough_payload() honors a caller-supplied tool_choice and
+  defaults to "auto" when unset.
+- _friendly_error() maps httpx transport errors to a "Lost connection"
+  message so passthrough failures are legible instead of bare 500s.
+
+No running server or GPU required.
+"""
+
+import os
+import sys
+
+_backend = os.path.join(os.path.dirname(__file__), "..")
+sys.path.insert(0, _backend)
+
+import httpx
+import pytest
+from pydantic import ValidationError
+
+from models.inference import (
+    ChatCompletionRequest,
+    ChatMessage,
+)
+from core.inference.anthropic_compat import (
+    anthropic_tool_choice_to_openai,
+)
+from routes.inference import _build_passthrough_payload, _friendly_error
+
+
+# =====================================================================
+# ChatMessage — tool role, tool_calls, optional content
+# =====================================================================
+
+
+class TestChatMessageToolRoles:
+    def test_tool_role_with_tool_call_id(self):
+        msg = ChatMessage(
+            role = "tool",
+            tool_call_id = "call_abc123",
+            content = '{"temperature": 72}',
+        )
+        assert msg.role == "tool"
+        assert msg.tool_call_id == "call_abc123"
+        assert msg.content == '{"temperature": 72}'
+
+    def test_tool_role_with_name(self):
+        msg = ChatMessage(
+            role = "tool",
+            tool_call_id = "call_abc123",
+            name = "get_weather",
+            content = '{"temperature": 72}',
+        )
+        assert msg.name == "get_weather"
+
+    def test_assistant_with_tool_calls_no_content(self):
+        msg = ChatMessage(
+            role = "assistant",
+            content = None,
+            tool_calls = [
+                {
+                    "id": "call_1",
+                    "type": "function",
+                    "function": {
+                        "name": "get_weather",
+                        "arguments": '{"city": "Paris"}',
+                    },
+                }
+            ],
+        )
+        assert msg.role == "assistant"
+        assert msg.content is None
+        assert msg.tool_calls is not None
+        assert len(msg.tool_calls) == 1
+        assert msg.tool_calls[0]["function"]["name"] == "get_weather"
+
+    def test_assistant_with_content_and_tool_calls(self):
+        msg = ChatMessage(
+            role = "assistant",
+            content = "Let me check the weather.",
+            tool_calls = [
+                {
+                    "id": "call_1",
+                    "type": "function",
+                    "function": {"name": "get_weather", "arguments": "{}"},
+                }
+            ],
+        )
+        assert msg.content == "Let me check the weather."
+        assert msg.tool_calls[0]["id"] == "call_1"
+
+    def test_plain_user_message_still_works(self):
+        msg = ChatMessage(role = "user", content = "Hello")
+        assert msg.role == "user"
+        assert msg.tool_call_id is None
+        assert msg.tool_calls is None
+        assert msg.name is None
+
+    def test_invalid_role_rejected(self):
+        with pytest.raises(ValidationError):
+            ChatMessage(role = "function", content = "x")
+
+    def test_content_absent_on_assistant_tool_call_defaults_to_none(self):
+        # Assistant messages that carry only tool_calls are the one
+        # documented case where `content=None` is permitted.
+        msg = ChatMessage(
+            role = "assistant",
+            tool_calls = [
+                {
+                    "id": "call_1",
+                    "type": "function",
+                    "function": {"name": "f", "arguments": "{}"},
+                }
+            ],
+        )
+        assert msg.content is None
+
+    def test_tool_role_missing_tool_call_id_rejected(self):
+        # Per OpenAI spec, role="tool" messages must carry tool_call_id so
+        # upstream backends can associate the result with its prior call.
+        # Pin the boundary-level rejection so a malformed tool-result
+        # message never reaches the passthrough path.
+        with pytest.raises(ValidationError) as exc_info:
+            ChatMessage(role = "tool", content = '{"temperature": 72}')
+        assert "tool_call_id" in str(exc_info.value)
+
+    def test_tool_role_empty_tool_call_id_rejected(self):
+        with pytest.raises(ValidationError):
+            ChatMessage(
+                role = "tool",
+                tool_call_id = "",
+                content = '{"temperature": 72}',
+            )
+
+    # ── Role-aware content requirements ────────────────────────────
+
+    def test_user_empty_content_rejected(self):
+        with pytest.raises(ValidationError):
+            ChatMessage(role = "user", content = "")
+
+    def test_system_empty_content_rejected(self):
+        with pytest.raises(ValidationError):
+            ChatMessage(role = "system", content = "")
+
+    def test_user_empty_list_content_rejected(self):
+        with pytest.raises(ValidationError):
+            ChatMessage(role = "user", content = [])
+
+    def test_tool_empty_content_rejected(self):
+        with pytest.raises(ValidationError) as exc_info:
+            ChatMessage(role = "tool", tool_call_id = "call_1", content = "")
+        assert "content" in str(exc_info.value)
+
+    def test_assistant_without_content_or_tool_calls_rejected(self):
+        with pytest.raises(ValidationError) as exc_info:
+            ChatMessage(role = "assistant")
+        assert "content" in str(exc_info.value) or "tool_calls" in str(exc_info.value)
+
+    # ── Role-constrained tool-call metadata ────────────────────────
+
+    def test_tool_calls_on_user_rejected(self):
+        with pytest.raises(ValidationError) as exc_info:
+            ChatMessage(
+                role = "user",
+                content = "Hi",
+                tool_calls = [
+                    {
+                        "id": "c1",
+                        "type": "function",
+                        "function": {"name": "f", "arguments": "{}"},
+                    }
+                ],
+            )
+        assert "tool_calls" in str(exc_info.value)
+
+    def test_tool_call_id_on_user_rejected(self):
+        with pytest.raises(ValidationError) as exc_info:
+            ChatMessage(role = "user", content = "Hi", tool_call_id = "call_1")
+        assert "tool_call_id" in str(exc_info.value)
+
+    def test_name_on_user_rejected(self):
+        with pytest.raises(ValidationError) as exc_info:
+            ChatMessage(role = "user", content = "Hi", name = "get_weather")
+        assert "name" in str(exc_info.value)
+
+
+# =====================================================================
+# ChatCompletionRequest — standard OpenAI tool fields
+# =====================================================================
+
+
+class TestChatCompletionRequestToolFields:
+    def _make(self, **kwargs):
+        base = {"messages": [{"role": "user", "content": "Hi"}]}
+        base.update(kwargs)
+        return ChatCompletionRequest(**base)
+
+    def test_tools_parses(self):
+        req = self._make(
+            tools = [
+                {
+                    "type": "function",
+                    "function": {
+                        "name": "get_weather",
+                        "description": "Return the weather in a city",
+                        "parameters": {
+                            "type": "object",
+                            "properties": {"city": {"type": "string"}},
+                            "required": ["city"],
+                        },
+                    },
+                }
+            ],
+        )
+        assert req.tools is not None
+        assert len(req.tools) == 1
+        assert req.tools[0]["function"]["name"] == "get_weather"
+
+    def test_tool_choice_string_auto(self):
+        assert self._make(tool_choice = "auto").tool_choice == "auto"
+
+    def test_tool_choice_string_required(self):
+        assert self._make(tool_choice = "required").tool_choice == "required"
+
+    def test_tool_choice_string_none(self):
+        assert self._make(tool_choice = "none").tool_choice == "none"
+
+    def test_tool_choice_named_function(self):
+        tc = {"type": "function", "function": {"name": "get_weather"}}
+        assert self._make(tool_choice = tc).tool_choice == tc
+
+    def test_stop_string(self):
+        assert self._make(stop = "\nUser:").stop == "\nUser:"
+
+    def test_stop_list(self):
+        assert self._make(stop = ["\nUser:", "\nAssistant:"]).stop == [
+            "\nUser:",
+            "\nAssistant:",
+        ]
+
+    def test_tools_default_none(self):
+        req = self._make()
+        assert req.tools is None
+        assert req.tool_choice is None
+        assert req.stop is None
+
+    def test_extra_fields_accepted(self):
+        # `frequency_penalty`, `seed`, `response_format` are not yet
+        # explicitly declared but must survive Pydantic parsing now that
+        # extra="allow" is set.
+        req = self._make(
+            frequency_penalty = 0.5,
+            seed = 42,
+            response_format = {"type": "json_object"},
+        )
+        # Extras land in model_extra
+        assert req.model_extra is not None
+        assert req.model_extra.get("frequency_penalty") == 0.5
+        assert req.model_extra.get("seed") == 42
+        assert req.model_extra.get("response_format") == {"type": "json_object"}
+
+    def test_unsloth_extensions_still_work(self):
+        req = self._make(
+            enable_tools = True,
+            enabled_tools = ["web_search", "python"],
+            session_id = "abc",
+        )
+        assert req.enable_tools is True
+        assert req.enabled_tools == ["web_search", "python"]
+        assert req.session_id == "abc"
+
+    def test_stream_defaults_false_matching_openai_spec(self):
+        # OpenAI's /v1/chat/completions spec defaults `stream` to false.
+        # Studio previously defaulted to true, which broke naive curl
+        # clients that omit `stream` (they expect a JSON blob, got SSE).
+        # Pin the corrected default so it can't silently regress.
+        req = self._make()
+        assert req.stream is False
+
+    def test_multiturn_tool_loop_messages(self):
+        req = ChatCompletionRequest(
+            messages = [
+                {"role": "user", "content": "What's the weather in Paris?"},
+                {
+                    "role": "assistant",
+                    "content": None,
+                    "tool_calls": [
+                        {
+                            "id": "call_1",
+                            "type": "function",
+                            "function": {
+                                "name": "get_weather",
+                                "arguments": '{"city": "Paris"}',
+                            },
+                        }
+                    ],
+                },
+                {
+                    "role": "tool",
+                    "tool_call_id": "call_1",
+                    "content": '{"temperature": 14, "unit": "celsius"}',
+                },
+            ],
+            tools = [
+                {
+                    "type": "function",
+                    "function": {
+                        "name": "get_weather",
+                        "parameters": {"type": "object"},
+                    },
+                }
+            ],
+        )
+        assert len(req.messages) == 3
+        assert req.messages[1].role == "assistant"
+        assert req.messages[1].content is None
+        assert req.messages[1].tool_calls[0]["id"] == "call_1"
+        assert req.messages[2].role == "tool"
+        assert req.messages[2].tool_call_id == "call_1"
+
+
+# =====================================================================
+# anthropic_tool_choice_to_openai — pure translation helper
+# =====================================================================
+
+
+class TestAnthropicToolChoiceToOpenAI:
+    def test_auto(self):
+        assert anthropic_tool_choice_to_openai({"type": "auto"}) == "auto"
+
+    def test_any_becomes_required(self):
+        assert anthropic_tool_choice_to_openai({"type": "any"}) == "required"
+
+    def test_none(self):
+        assert anthropic_tool_choice_to_openai({"type": "none"}) == "none"
+
+    def test_tool_named(self):
+        result = anthropic_tool_choice_to_openai(
+            {"type": "tool", "name": "get_weather"}
+        )
+        assert result == {
+            "type": "function",
+            "function": {"name": "get_weather"},
+        }
+
+    def test_tool_missing_name_returns_none(self):
+        assert anthropic_tool_choice_to_openai({"type": "tool"}) is None
+
+    def test_none_input_returns_none(self):
+        assert anthropic_tool_choice_to_openai(None) is None
+
+    def test_unrecognized_shape_returns_none(self):
+        assert anthropic_tool_choice_to_openai({"type": "wibble"}) is None
+        assert anthropic_tool_choice_to_openai("auto") is None
+        assert anthropic_tool_choice_to_openai(42) is None
+
+
+# =====================================================================
+# _build_passthrough_payload — tool_choice propagation
+# =====================================================================
+
+
+class TestBuildPassthroughPayloadToolChoice:
+    def _args(self):
+        return dict(
+            openai_messages = [{"role": "user", "content": "Hi"}],
+            openai_tools = [
+                {
+                    "type": "function",
+                    "function": {"name": "f", "parameters": {"type": "object"}},
+                }
+            ],
+            temperature = 0.6,
+            top_p = 0.95,
+            top_k = 20,
+            max_tokens = 128,
+            stream = False,
+        )
+
+    def test_default_tool_choice_is_auto(self):
+        body = _build_passthrough_payload(**self._args())
+        assert body["tool_choice"] == "auto"
+
+    def test_override_tool_choice_required(self):
+        body = _build_passthrough_payload(**self._args(), tool_choice = "required")
+        assert body["tool_choice"] == "required"
+
+    def test_override_tool_choice_none(self):
+        body = _build_passthrough_payload(**self._args(), tool_choice = "none")
+        assert body["tool_choice"] == "none"
+
+    def test_override_tool_choice_named_function(self):
+        tc = {"type": "function", "function": {"name": "f"}}
+        body = _build_passthrough_payload(**self._args(), tool_choice = tc)
+        assert body["tool_choice"] == tc
+
+    def test_stream_adds_include_usage(self):
+        args = self._args()
+        args["stream"] = True
+        body = _build_passthrough_payload(**args)
+        assert body.get("stream_options") == {"include_usage": True}
+
+    def test_repetition_penalty_renamed(self):
+        body = _build_passthrough_payload(**self._args(), repetition_penalty = 1.1)
+        assert body.get("repeat_penalty") == 1.1
+        assert "repetition_penalty" not in body
+
+
+# =====================================================================
+# _friendly_error — httpx transport failures
+# =====================================================================
+
+
+class TestFriendlyErrorHttpx:
+    """The async pass-through helpers talk to llama-server via httpx.
+    When the subprocess is down, httpx raises RequestError subclasses
+    whose string form (``"All connection attempts failed"``, ``"[Errno 111]
+    Connection refused"``, ...) does NOT contain the substring
+    ``"Lost connection to llama-server"`` the sync path uses, so the
+    previous substring-only `_friendly_error` returned a useless generic
+    message. These tests pin the new isinstance-based mapping.
+    """
+
+    def _req(self):
+        return httpx.Request("POST", "http://127.0.0.1:65535/v1/chat/completions")
+
+    def test_connect_error_mapped(self):
+        exc = httpx.ConnectError("All connection attempts failed", request = self._req())
+        assert "Lost connection" in _friendly_error(exc)
+
+    def test_read_error_mapped(self):
+        exc = httpx.ReadError("EOF", request = self._req())
+        assert "Lost connection" in _friendly_error(exc)
+
+    def test_remote_protocol_error_mapped(self):
+        exc = httpx.RemoteProtocolError("peer closed", request = self._req())
+        assert "Lost connection" in _friendly_error(exc)
+
+    def test_read_timeout_mapped(self):
+        exc = httpx.ReadTimeout("timed out", request = self._req())
+        assert "Lost connection" in _friendly_error(exc)
+
+    def test_non_httpx_unchanged(self):
+        # Non-httpx exceptions still fall through to the existing substring
+        # heuristics — a context-size message must still produce the
+        # "Message too long" path.
+        ctx_msg = (
+            "request (4096 tokens) exceeds the available context size (2048 tokens)"
+        )
+        assert "Message too long" in _friendly_error(ValueError(ctx_msg))
+
+    def test_generic_exception_returns_generic_message(self):
+        assert (
+            _friendly_error(RuntimeError("unrelated")) == "An internal error occurred"
+        )
diff --git a/studio/backend/tests/test_studio_api.py b/studio/backend/tests/test_studio_api.py
index 9cc17c89fb..521c99e126 100644
--- a/studio/backend/tests/test_studio_api.py
+++ b/studio/backend/tests/test_studio_api.py
@@ -11,11 +11,16 @@ authentication and the CLI's ``--help`` output:
     1. curl -- basic chat completions (non-streaming)
     2. curl -- streaming chat completions
     3. Python OpenAI SDK -- streaming completions
-    4. curl -- with tools (web_search + python)
-    5. Anthropic Messages API -- basic non-streaming
-    6. Anthropic Messages API -- streaming SSE
-    7. Anthropic Python SDK -- non-streaming
-    8. Anthropic Messages API -- streaming with tools
+    4. curl -- Studio server-side tools (enable_tools=true)
+    5. curl -- Standard OpenAI function calling (non-streaming)
+    6. curl -- Standard OpenAI function calling (streaming)
+    7. curl -- Standard OpenAI function calling (multi-turn tool loop)
+    8. OpenAI Python SDK -- Standard function calling
+    9. Anthropic Messages API -- basic non-streaming
+    10. Anthropic Messages API -- streaming SSE
+    11. Anthropic Python SDK -- non-streaming
+    12. Anthropic Messages API -- streaming with tools
+    13. Anthropic Messages API -- tool_choice={"type":"any"} honored
 
 Training, export, fine-tuning, and chat-UI concerns are out of scope —
 see the unit suites elsewhere under ``studio/backend/tests/`` for those.
@@ -266,6 +271,250 @@ def test_curl_with_tools(base_url: str, api_key: str):
     print(f"  PASS  curl with tools: {len(chunks)} chunks, {len(full)} chars content")
 
 
+# ── Standard OpenAI function-calling pass-through tests ─────────────
+#
+# Regression coverage for unslothai/unsloth#4999: Studio's
+# /v1/chat/completions used to silently strip standard OpenAI `tools`
+# and `tool_choice` fields, so clients (opencode, Claude Code, Cursor,
+# Continue, ...) could never get structured tool_calls back. These
+# tests exercise the client-side pass-through path that forwards those
+# fields to llama-server verbatim.
+#
+# They require a tool-capable GGUF (``supports_tools=True`` — e.g.
+# Qwen3, Qwen2.5-Coder, Llama-3.1-Instruct). The default test model
+# ``unsloth/Qwen3-1.7B-GGUF`` advertises tool support via its chat
+# template metadata.
+
+_WEATHER_TOOL = {
+    "type": "function",
+    "function": {
+        "name": "get_weather",
+        "description": "Look up the current weather for a given city.",
+        "parameters": {
+            "type": "object",
+            "properties": {
+                "city": {
+                    "type": "string",
+                    "description": "The name of the city, e.g. 'Paris'.",
+                },
+            },
+            "required": ["city"],
+        },
+    },
+}
+
+
+def _collect_streamed_tool_calls(chunks: list[dict]) -> list[dict]:
+    """Reassemble OpenAI streaming delta.tool_calls into full tool calls.
+
+    OpenAI streams partial tool calls across chunks — the first chunk for
+    a given index carries ``id`` + ``function.name``, and subsequent
+    chunks append fragments to ``function.arguments``.
+    """
+    by_index: dict[int, dict] = {}
+    for c in chunks:
+        choices = c.get("choices") or []
+        if not choices:
+            continue
+        delta = choices[0].get("delta") or {}
+        tool_calls = delta.get("tool_calls") or []
+        for tc in tool_calls:
+            idx = tc.get("index", 0)
+            slot = by_index.setdefault(
+                idx,
+                {
+                    "id": None,
+                    "type": "function",
+                    "function": {"name": None, "arguments": ""},
+                },
+            )
+            if tc.get("id"):
+                slot["id"] = tc["id"]
+            fn = tc.get("function") or {}
+            if fn.get("name"):
+                slot["function"]["name"] = fn["name"]
+            if fn.get("arguments"):
+                slot["function"]["arguments"] += fn["arguments"]
+    return [by_index[i] for i in sorted(by_index)]
+
+
+def _final_finish_reason(chunks: list[dict]) -> str | None:
+    for c in reversed(chunks):
+        choices = c.get("choices") or []
+        if not choices:
+            continue
+        fr = choices[0].get("finish_reason")
+        if fr is not None:
+            return fr
+    return None
+
+
+def test_openai_tools_nonstream(base_url: str, api_key: str):
+    """Standard OpenAI function calling, non-streaming, tool_choice='required'.
+
+    Regression: before the fix, Studio silently stripped `tools` and the
+    model returned plain text with finish_reason='stop'. After the fix,
+    llama-server's response is forwarded verbatim so the client sees
+    finish_reason='tool_calls' with a structured tool_calls array and
+    non-zero usage.prompt_tokens.
+    """
+    status, text = _http(
+        "POST",
+        f"{base_url}/v1/chat/completions",
+        body = {
+            "messages": [{"role": "user", "content": "What is the weather in Paris?"}],
+            "tools": [_WEATHER_TOOL],
+            "tool_choice": "required",
+            "stream": False,
+        },
+        headers = {"Authorization": f"Bearer {api_key}"},
+        timeout = 120,
+    )
+    assert status == 200, f"Expected 200, got {status}: {text[:500]}"
+    data = json.loads(text)
+    assert "choices" in data, f"Missing 'choices': {text[:300]}"
+    choice = data["choices"][0]
+    assert (
+        choice["finish_reason"] == "tool_calls"
+    ), f"Expected finish_reason='tool_calls', got {choice['finish_reason']!r}"
+    msg = choice["message"]
+    tool_calls = msg.get("tool_calls") or []
+    assert len(tool_calls) >= 1, f"No tool_calls in response: {msg}"
+    first = tool_calls[0]
+    assert first["type"] == "function"
+    assert (
+        first["function"]["name"] == "get_weather"
+    ), f"Wrong tool name: {first['function']['name']!r}"
+    # arguments must be valid JSON
+    parsed = json.loads(first["function"]["arguments"])
+    assert "city" in parsed, f"Tool call missing required 'city' arg: {parsed}"
+    # Usage must be non-zero (was 0 before the fix)
+    usage = data.get("usage") or {}
+    assert (
+        usage.get("prompt_tokens", 0) > 0
+    ), f"Expected non-zero prompt_tokens; got {usage}"
+    assert data.get("id"), "Missing response id"
+    print(
+        f"  PASS  openai tools non-stream: "
+        f"tool={first['function']['name']}, args={parsed}, "
+        f"prompt_tokens={usage['prompt_tokens']}"
+    )
+
+
+def test_openai_tools_stream(base_url: str, api_key: str):
+    """Standard OpenAI function calling, streaming, tool_choice='required'."""
+    status, chunks = _stream_http(
+        f"{base_url}/v1/chat/completions",
+        body = {
+            "messages": [{"role": "user", "content": "What is the weather in Tokyo?"}],
+            "tools": [_WEATHER_TOOL],
+            "tool_choice": "required",
+            "stream": True,
+        },
+        headers = {"Authorization": f"Bearer {api_key}"},
+        timeout = 120,
+    )
+    assert status == 200, f"Expected 200, got {status}"
+    assert len(chunks) > 0, "No SSE chunks received"
+    assert _final_finish_reason(chunks) == "tool_calls", (
+        f"Expected final finish_reason='tool_calls', got "
+        f"{_final_finish_reason(chunks)!r}"
+    )
+    assembled = _collect_streamed_tool_calls(chunks)
+    assert len(assembled) >= 1, "No tool_calls reassembled from stream"
+    first = assembled[0]
+    assert first["function"]["name"] == "get_weather"
+    parsed = json.loads(first["function"]["arguments"])
+    assert "city" in parsed
+    print(
+        f"  PASS  openai tools stream: {len(chunks)} chunks, "
+        f"tool={first['function']['name']}, args={parsed}"
+    )
+
+
+def test_openai_tools_multiturn(base_url: str, api_key: str):
+    """Multi-turn client-side tool loop: validates that role='tool' result
+    messages and assistant messages carrying tool_calls are accepted.
+
+    Regression: before the fix, ChatMessage.role was restricted to
+    {system,user,assistant} and rejected role='tool' at the Pydantic
+    validation stage. This test sends a full round trip so the model
+    receives the simulated tool result and responds with final text.
+    """
+    status, text = _http(
+        "POST",
+        f"{base_url}/v1/chat/completions",
+        body = {
+            "messages": [
+                {"role": "user", "content": "What is the weather in Paris?"},
+                {
+                    "role": "assistant",
+                    "content": None,
+                    "tool_calls": [
+                        {
+                            "id": "call_test_1",
+                            "type": "function",
+                            "function": {
+                                "name": "get_weather",
+                                "arguments": '{"city": "Paris"}',
+                            },
+                        }
+                    ],
+                },
+                {
+                    "role": "tool",
+                    "tool_call_id": "call_test_1",
+                    "content": '{"temperature_c": 14, "condition": "cloudy"}',
+                },
+            ],
+            "tools": [_WEATHER_TOOL],
+            "stream": False,
+        },
+        headers = {"Authorization": f"Bearer {api_key}"},
+        timeout = 120,
+    )
+    assert status == 200, f"Expected 200, got {status}: {text[:500]}"
+    data = json.loads(text)
+    msg = data["choices"][0]["message"]
+    # The model should respond with text now that it has the tool result
+    content = msg.get("content") or ""
+    assert len(content) > 0 or msg.get(
+        "tool_calls"
+    ), f"Expected text or follow-up tool call, got empty message: {msg}"
+    print(f"  PASS  openai tools multiturn: {content[:80]!r}")
+
+
+def test_openai_sdk_tool_calling(base_url: str, api_key: str):
+    """OpenAI Python SDK round trip — the real client shape opencode et al. use."""
+    try:
+        from openai import OpenAI
+    except ImportError:
+        print("  SKIP  openai SDK not installed")
+        return
+
+    client = OpenAI(base_url = f"{base_url}/v1", api_key = api_key)
+    resp = client.chat.completions.create(
+        model = "current",
+        messages = [{"role": "user", "content": "What's the weather in Berlin?"}],
+        tools = [_WEATHER_TOOL],
+        tool_choice = "required",
+        stream = False,
+    )
+    assert resp.choices[0].finish_reason == "tool_calls", (
+        f"Expected finish_reason='tool_calls', got "
+        f"{resp.choices[0].finish_reason!r}"
+    )
+    tool_calls = resp.choices[0].message.tool_calls
+    assert tool_calls and len(tool_calls) >= 1, "No tool_calls from SDK"
+    tc = tool_calls[0]
+    assert tc.function.name == "get_weather"
+    parsed = json.loads(tc.function.arguments)
+    assert "city" in parsed
+    print(
+        f"  PASS  openai SDK tool calling: " f"tool={tc.function.name}, args={parsed}"
+    )
+
+
 def test_invalid_key_rejected(base_url: str):
     """Requests with a bad API key should be rejected."""
     status, _text = _http(
@@ -464,6 +713,73 @@ def test_anthropic_with_tools(base_url: str, api_key: str):
     )
 
 
+def test_anthropic_tool_choice_any(base_url: str, api_key: str):
+    """Anthropic Messages API: ``tool_choice: {"type": "any"}`` must be
+    honored (forwarded as OpenAI ``tool_choice: "required"`` to
+    llama-server). Regression for the secondary fix bundled with #4999 —
+    previously this field was accepted on the request model but silently
+    dropped with a warning log, so the model was free to answer from
+    memory instead of using the tool.
+    """
+    status, events = _stream_anthropic_http(
+        f"{base_url}/v1/messages",
+        body = {
+            "model": "default",
+            "max_tokens": 256,
+            "messages": [
+                # A question the model could easily answer from memory if
+                # tool_choice were not enforced.
+                {
+                    "role": "user",
+                    "content": "What is the weather in London right now?",
+                }
+            ],
+            "tools": [
+                {
+                    "name": "get_weather",
+                    "description": "Look up current weather for a city.",
+                    "input_schema": {
+                        "type": "object",
+                        "properties": {
+                            "city": {"type": "string"},
+                        },
+                        "required": ["city"],
+                    },
+                }
+            ],
+            "tool_choice": {"type": "any"},
+            "stream": True,
+        },
+        headers = {"Authorization": f"Bearer {api_key}"},
+        timeout = 120,
+    )
+    assert status == 200, f"Expected 200, got {status}"
+    assert len(events) > 0, "No SSE events received"
+
+    # With tool_choice=any, stop_reason must be tool_use (not end_turn)
+    stop_reason = None
+    for etype, data in events:
+        if etype == "message_delta":
+            stop_reason = data.get("delta", {}).get("stop_reason") or stop_reason
+    assert stop_reason == "tool_use", (
+        f"Expected stop_reason='tool_use' with tool_choice=any, got "
+        f"{stop_reason!r} — tool_choice may not be forwarded to llama-server."
+    )
+
+    # And at least one tool_use content block must be emitted
+    tool_use_starts = [
+        e
+        for e in events
+        if e[0] == "content_block_start"
+        and e[1].get("content_block", {}).get("type") == "tool_use"
+    ]
+    assert len(tool_use_starts) >= 1, "No tool_use content block emitted"
+    print(
+        f"  PASS  anthropic tool_choice=any honored: "
+        f"{len(tool_use_starts)} tool_use blocks, stop_reason={stop_reason}"
+    )
+
+
 # ── Server lifecycle ─────────────────────────────────────────────────
 
 
@@ -578,10 +894,10 @@ def main():
             print(f"  ERROR {fn.__name__}: {type(exc).__name__}: {exc}")
 
     # ── 1. Test --help (no server needed) ────────────────────────────
-    print("\n[1/11] Testing --help output")
+    print("\n[1/16] Testing --help output")
     run_test(test_help_output)
 
-    # ── 2-11. Start server and run API tests ─────────────────────────
+    # ── 2-16. Start server and run API tests ─────────────────────────
     print(
         f"\nStarting server: {args.model} (variant={args.gguf_variant}) on port {PORT}..."
     )
@@ -591,39 +907,54 @@ def main():
         base_url = f"http://{HOST}:{PORT}"
         print(f"Server ready.  API Key: {api_key[:20]}...\n")
 
-        print("[2/11] Testing curl basic (non-streaming)")
+        print("[2/16] Testing curl basic (non-streaming)")
         run_test(test_curl_basic, base_url, api_key)
 
-        print("[3/11] Testing curl streaming")
+        print("[3/16] Testing curl streaming")
         run_test(test_curl_streaming, base_url, api_key)
 
-        print("[4/11] Testing OpenAI Python SDK (streaming)")
+        print("[4/16] Testing OpenAI Python SDK (streaming)")
         run_test(test_openai_sdk, base_url, api_key)
 
-        print("[5/11] Testing curl with tools")
+        print("[5/16] Testing curl with tools (server-side enable_tools)")
         run_test(test_curl_with_tools, base_url, api_key)
 
-        print("[6/11] Testing invalid API key rejection")
+        print("[6/16] Testing OpenAI standard tools (non-streaming)")
+        run_test(test_openai_tools_nonstream, base_url, api_key)
+
+        print("[7/16] Testing OpenAI standard tools (streaming)")
+        run_test(test_openai_tools_stream, base_url, api_key)
+
+        print("[8/16] Testing OpenAI standard tools (multi-turn)")
+        run_test(test_openai_tools_multiturn, base_url, api_key)
+
+        print("[9/16] Testing OpenAI SDK tool calling")
+        run_test(test_openai_sdk_tool_calling, base_url, api_key)
+
+        print("[10/16] Testing invalid API key rejection")
         run_test(test_invalid_key_rejected, base_url)
 
-        print("[7/11] Testing no API key rejection")
+        print("[11/16] Testing no API key rejection")
         run_test(test_no_key_rejected, base_url)
 
-        print("[8/11] Testing Anthropic basic (non-streaming)")
+        print("[12/16] Testing Anthropic basic (non-streaming)")
         run_test(test_anthropic_basic, base_url, api_key)
 
-        print("[9/11] Testing Anthropic streaming")
+        print("[13/16] Testing Anthropic streaming")
         run_test(test_anthropic_streaming, base_url, api_key)
 
-        print("[10/11] Testing Anthropic Python SDK")
+        print("[14/16] Testing Anthropic Python SDK")
         run_test(test_anthropic_sdk, base_url, api_key)
 
-        print("[11/11] Testing Anthropic with tools")
+        print("[15/16] Testing Anthropic with tools")
         run_test(test_anthropic_with_tools, base_url, api_key)
 
+        print("[16/16] Testing Anthropic tool_choice=any honored")
+        run_test(test_anthropic_tool_choice_any, base_url, api_key)
+
     except RuntimeError as exc:
         print(f"\nFATAL: Server failed to start: {exc}")
-        failed += 11  # count remaining tests as failed
+        failed += 16  # count remaining tests as failed
     finally:
         if proc:
             print("\nStopping server...")

From f5eec8a6f2834a34062e1b5e2760ce5160a6227a Mon Sep 17 00:00:00 2001
From: Michael Han <107991372+shimmyshimmer@users.noreply.github.com>
Date: Sun, 19 Apr 2026 23:16:36 -0700
Subject: [PATCH 136/216] Qwen3.6 and ReadMe revamp.md

---
 README.md | 53 ++++++++++++++++++++++++++++++++++++++++-------------
 1 file changed, 40 insertions(+), 13 deletions(-)

diff --git a/README.md b/README.md
index 7046a2af7c..29a2a49bb6 100644
--- a/README.md
+++ b/README.md
@@ -1,28 +1,54 @@
 <h1 align="center" style="margin:0;">
   <a href="https://unsloth.ai/docs"><picture>
-    <source media="(prefers-color-scheme: dark)" srcset="https://raw.githubusercontent.com/unslothai/unsloth/main/images/STUDIO%20WHITE%20LOGO.png">
-    <source media="(prefers-color-scheme: light)" srcset="https://raw.githubusercontent.com/unslothai/unsloth/main/images/STUDIO%20BLACK%20LOGO.png">
-    <img alt="Unsloth logo" src="https://raw.githubusercontent.com/unslothai/unsloth/main/images/STUDIO%20BLACK%20LOGO.png" height="60" style="max-width:100%;">
+    <source media="(prefers-color-scheme: dark)" srcset="https://raw.githubusercontent.com/unslothai/unsloth/main/images/unsloth%20logo%20white%20text.png">
+    <source media="(prefers-color-scheme: light)" srcset="https://raw.githubusercontent.com/unslothai/unsloth/main/images/unsloth%20logo%20black%20text.png">
+    <img alt="Unsloth logo" src="https://raw.githubusercontent.com/unslothai/unsloth/main/images/unsloth%20logo%20black%20text.png" height="80" style="max-width:100%;">
   </picture></a>
 </h1>
 <h3 align="center" style="margin: 0; margin-top: 0;">
-Run and train AI models with a unified local interface.
+Unsloth Studio lets you run and train models locally.
 </h3>
 
 <p align="center">
   <a href="#-features">Features</a> •
-  <a href="#-quickstart">Quickstart</a> •
+  <a href="#-install">Quickstart</a> •
   <a href="#-free-notebooks">Notebooks</a> •
-  <a href="https://unsloth.ai/docs">Documentation</a> •
-  <a href="https://www.reddit.com/r/unsloth/">Reddit</a>
+  <a href="https://unsloth.ai/docs">Documentation</a>
 </p>
- <a href="https://unsloth.ai/docs/new/studio">
-<img alt="unsloth studio ui homepage" src="https://raw.githubusercontent.com/unslothai/unsloth/main/studio/frontend/public/studio%20github%20landscape%20colab%20display.png" style="max-width: 100%; margin-bottom: 0;"></a>
+<br>
+<a href="https://unsloth.ai/docs/new/studio">
+<p align="center">
+    <img
+      src="https://github.com/user-attachments/assets/35ad5370-e140-4095-b46f-dcb7a5ddc933"
+      alt="Train models with Unsloth Studio"
+      width="49.5%"
+    />
+  <img
+    src="https://github.com/user-attachments/assets/1429d0bf-f8ff-4366-9da8-f99cd02f525a"
+    alt="Run models with Unsloth Studio"
+    width="49.5%"
+  />
+</p></a>
 
-Unsloth Studio (Beta) lets you run and train text, [audio](https://unsloth.ai/docs/basics/text-to-speech-tts-fine-tuning), [embedding](https://unsloth.ai/docs/new/embedding-finetuning), [vision](https://unsloth.ai/docs/basics/vision-fine-tuning) models on Windows, Linux and macOS.
+## ⚡ Get started
+
+#### macOS, Linux, WSL:
+```bash
+curl -fsSL https://unsloth.ai/install.sh | sh
+```
+#### Windows:
+```powershell
+irm https://unsloth.ai/install.ps1 | iex
+```
+#### Community:
+
+- [Discord](https://discord.gg/unsloth)
+- [𝕏 (Twitter)](https://x.com/UnslothAI)
+- [Reddit](https://reddit.com/r/unsloth)
 
 ## ⭐ Features
-Unsloth provides several key features for both inference and training:
+Unsloth Studio (Beta) lets you run and train text, [audio](https://unsloth.ai/docs/basics/text-to-speech-tts-fine-tuning), [embedding](https://unsloth.ai/docs/new/embedding-finetuning), [vision](https://unsloth.ai/docs/basics/vision-fine-tuning) models on Windows, Linux and macOS.
+
 ### Inference
 * **Search + download + run models** including GGUF, LoRA adapters, safetensors
 * **Export models**: [Save or export](https://unsloth.ai/docs/new/studio/export) models to GGUF, 16-bit safetensors and other formats.
@@ -40,7 +66,7 @@ Unsloth provides several key features for both inference and training:
 * **Observability**: Monitor training live, track loss and GPU usage and customize graphs.
 * [Multi-GPU](https://unsloth.ai/docs/basics/multi-gpu-training-with-unsloth) training is supported, with major improvements coming soon.
 
-## ⚡ Quickstart
+## 📥 Install
 Unsloth can be used in two ways: through **[Unsloth Studio](https://unsloth.ai/docs/new/studio/)**, the web UI, or through **Unsloth Core**, the code-based version. Each has different requirements.
 
 ### Unsloth Studio (web UI)
@@ -133,7 +159,8 @@ Read our [guide](https://unsloth.ai/docs/get-started/fine-tuning-llms-guide). Ad
 - See detailed documentation for Unsloth [here](https://unsloth.ai/docs)
 
 ## 🦥 Unsloth News
-- **Gemma 4**: Run and train Google’s new models directly in Unsloth Studio! [Blog](https://unsloth.ai/docs/models/gemma-4)
+- **Qwen3.6**: Qwen3.6-35B-A3B can now be trained and run in Unsloth Studio. [Blog](https://unsloth.ai/docs/models/qwen3.6)
+- **Gemma 4**: Run and train Google’s new models directly in Unsloth. [Blog](https://unsloth.ai/docs/models/gemma-4)
 - **Introducing Unsloth Studio**: our new web UI for running and training LLMs. [Blog](https://unsloth.ai/docs/new/studio)
 - **Qwen3.5** - 0.8B, 2B, 4B, 9B, 27B, 35-A3B, 112B-A10B are now supported. [Guide + notebooks](https://unsloth.ai/docs/models/qwen3.5/fine-tune)
 - Train **MoE LLMs 12x faster** with 35% less VRAM - DeepSeek, GLM, Qwen and gpt-oss. [Blog](https://unsloth.ai/docs/new/faster-moe)

From b24f3f61b83fab129aa67e46ff9c2d430434401a Mon Sep 17 00:00:00 2001
From: Michael Han <107991372+shimmyshimmer@users.noreply.github.com>
Date: Mon, 20 Apr 2026 00:37:40 -0700
Subject: [PATCH 137/216] Update README.md

---
 README.md | 13 +------------
 1 file changed, 1 insertion(+), 12 deletions(-)

diff --git a/README.md b/README.md
index 29a2a49bb6..e1b7e448ef 100644
--- a/README.md
+++ b/README.md
@@ -17,18 +17,7 @@ Unsloth Studio lets you run and train models locally.
 </p>
 <br>
 <a href="https://unsloth.ai/docs/new/studio">
-<p align="center">
-    <img
-      src="https://github.com/user-attachments/assets/35ad5370-e140-4095-b46f-dcb7a5ddc933"
-      alt="Train models with Unsloth Studio"
-      width="49.5%"
-    />
-  <img
-    src="https://github.com/user-attachments/assets/1429d0bf-f8ff-4366-9da8-f99cd02f525a"
-    alt="Run models with Unsloth Studio"
-    width="49.5%"
-  />
-</p></a>
+<img alt="unsloth studio ui homepage" src="https://github.com/user-attachments/assets/53ae17a9-d975-44ef-9686-efb4ebd0454d" style="max-width: 100%; margin-bottom: 0;"></a>
 
 ## ⚡ Get started
 

From 9954781d30fb568520204cd581878281b8398258 Mon Sep 17 00:00:00 2001
From: Roland Tannous <115670425+rolandtannous@users.noreply.github.com>
Date: Mon, 20 Apr 2026 21:06:59 +0400
Subject: [PATCH 138/216] fix(studio/chat): cancel in-flight run when trashing
 a thread from sidebar (#5067)
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

Trashing a thread mid-stream used to delete the Dexie rows while the
model kept generating, because the sidebar has no access to the
@assistant-ui aui context. Expose per-thread cancelRun() through the
chat runtime store and call it from deleteChatItem so trash behaves
like Stop → Trash. Covers compare pairs by cancelling each paired
thread.

Co-authored-by: Lee Jackson <130007945+Imagineer99@users.noreply.github.com>
---
 .../chat/hooks/use-chat-sidebar-items.ts      | 31 +++++++++++++------
 .../src/features/chat/runtime-provider.tsx    | 29 +++++++++++++++++
 .../chat/stores/chat-runtime-store.ts         | 17 ++++++++++
 3 files changed, 68 insertions(+), 9 deletions(-)

diff --git a/studio/frontend/src/features/chat/hooks/use-chat-sidebar-items.ts b/studio/frontend/src/features/chat/hooks/use-chat-sidebar-items.ts
index fb27f990cf..33cdd88081 100644
--- a/studio/frontend/src/features/chat/hooks/use-chat-sidebar-items.ts
+++ b/studio/frontend/src/features/chat/hooks/use-chat-sidebar-items.ts
@@ -58,23 +58,36 @@ export function useChatSidebarItems() {
   return { items, canCompare };
 }
 
+function cancelIfRunning(threadId: string): void {
+  const { runningByThreadId, cancelByThreadId } =
+    useChatRuntimeStore.getState();
+  if (!runningByThreadId[threadId]) return;
+  cancelByThreadId[threadId]?.();
+}
+
 export async function deleteChatItem(
   item: SidebarItem,
   activeId: string | undefined,
   onSelect: (view: { mode: "single"; newThreadNonce: string }) => void,
 ) {
+  const threadIds: string[] =
+    item.type === "single"
+      ? [item.id]
+      : (await db.threads.where("pairId").equals(item.id).toArray()).map(
+          (t) => t.id,
+        );
+
+  // Stop any in-flight streams before deleting, so the model doesn't keep
+  // generating against a thread that no longer exists.
+  for (const id of threadIds) cancelIfRunning(id);
+
   await db.transaction("rw", db.threads, db.messages, async () => {
-    if (item.type === "single") {
-      await db.messages.where("threadId").equals(item.id).delete();
-      await db.threads.delete(item.id);
-    } else {
-      const paired = await db.threads.where("pairId").equals(item.id).toArray();
-      for (const t of paired) {
-        await db.messages.where("threadId").equals(t.id).delete();
-        await db.threads.delete(t.id);
-      }
+    for (const id of threadIds) {
+      await db.messages.where("threadId").equals(id).delete();
+      await db.threads.delete(id);
     }
   });
+
   if (activeId === item.id) {
     useChatRuntimeStore.getState().setActiveThreadId(null);
     onSelect({ mode: "single", newThreadNonce: crypto.randomUUID() });
diff --git a/studio/frontend/src/features/chat/runtime-provider.tsx b/studio/frontend/src/features/chat/runtime-provider.tsx
index 348a0a85c4..5747e17970 100644
--- a/studio/frontend/src/features/chat/runtime-provider.tsx
+++ b/studio/frontend/src/features/chat/runtime-provider.tsx
@@ -730,6 +730,34 @@ function ActiveThreadSync({
   return null;
 }
 
+// Exposes the current thread's cancelRun() via the shared store so external
+// surfaces (e.g. the sidebar trash button) can stop an in-flight stream
+// before deleting the thread — mirroring the Stop → Trash sequence.
+function CancelRegistrar(): ReactElement | null {
+  const aui = useAui();
+  const mainThreadId = useAuiState(({ threads }) => threads.mainThreadId);
+  const isRunning = useChatRuntimeStore((s) =>
+    mainThreadId ? Boolean(s.runningByThreadId[mainThreadId]) : false,
+  );
+
+  useEffect(() => {
+    if (!mainThreadId || !isRunning) return;
+    const cancel = () => {
+      try {
+        aui.thread().cancelRun();
+      } catch {
+        // Run may have already ended between the caller's read and this call.
+      }
+    };
+    useChatRuntimeStore.getState().registerThreadCancel(mainThreadId, cancel);
+    return () => {
+      useChatRuntimeStore.getState().clearThreadCancel(mainThreadId);
+    };
+  }, [aui, mainThreadId, isRunning]);
+
+  return null;
+}
+
 export function ChatRuntimeProvider({
   children,
   modelType = "base",
@@ -762,6 +790,7 @@ export function ChatRuntimeProvider({
       <ActiveThreadSync
         enabled={modelType === "base" && !pairId && !newThreadNonce && !initialThreadId}
       />
+      <CancelRegistrar />
       {initialThreadId && (
         <ThreadAutoSwitch
           threadId={initialThreadId}
diff --git a/studio/frontend/src/features/chat/stores/chat-runtime-store.ts b/studio/frontend/src/features/chat/stores/chat-runtime-store.ts
index 9ff7ded18f..821716fe0a 100644
--- a/studio/frontend/src/features/chat/stores/chat-runtime-store.ts
+++ b/studio/frontend/src/features/chat/stores/chat-runtime-store.ts
@@ -146,6 +146,7 @@ type ChatRuntimeStore = {
   models: ChatModelSummary[];
   loras: ChatLoraSummary[];
   runningByThreadId: Record<string, boolean>;
+  cancelByThreadId: Record<string, () => void>;
   autoTitle: boolean;
   hfToken: string;
   modelsError: string | null;
@@ -189,6 +190,8 @@ type ChatRuntimeStore = {
   setModels: (models: ChatModelSummary[]) => void;
   setLoras: (loras: ChatLoraSummary[]) => void;
   setThreadRunning: (threadId: string, running: boolean) => void;
+  registerThreadCancel: (threadId: string, cancel: () => void) => void;
+  clearThreadCancel: (threadId: string) => void;
   setAutoTitle: (enabled: boolean) => void;
   setHfToken: (token: string) => void;
   setModelsError: (error: string | null) => void;
@@ -218,6 +221,7 @@ export const useChatRuntimeStore = create<ChatRuntimeStore>((set) => ({
   models: [],
   loras: [],
   runningByThreadId: {},
+  cancelByThreadId: {},
   autoTitle: loadBool(AUTO_TITLE_KEY, false),
   hfToken: loadString(HF_TOKEN_KEY, ""),
   modelsError: null,
@@ -277,6 +281,19 @@ export const useChatRuntimeStore = create<ChatRuntimeStore>((set) => ({
       }
       return { runningByThreadId: next };
     }),
+  registerThreadCancel: (threadId, cancel) =>
+    set((state) => {
+      const next = { ...state.cancelByThreadId };
+      next[threadId] = cancel;
+      return { cancelByThreadId: next };
+    }),
+  clearThreadCancel: (threadId) =>
+    set((state) => {
+      if (!(threadId in state.cancelByThreadId)) return state;
+      const next = { ...state.cancelByThreadId };
+      delete next[threadId];
+      return { cancelByThreadId: next };
+    }),
   setAutoTitle: (autoTitle) =>
     set(() => {
       saveBool(AUTO_TITLE_KEY, autoTitle);

From 9c8a079d97bb20ac62128321fa8b889ab6dcf2c7 Mon Sep 17 00:00:00 2001
From: Lee Jackson <130007945+Imagineer99@users.noreply.github.com>
Date: Mon, 20 Apr 2026 19:28:02 +0100
Subject: [PATCH 139/216] Studio: Local profile customization in settings and
 sync sidebar identity (#5088)

* studio: add local profile customization in settings

* studio: add local profile settings and sync sidebar identity

* fix: adjust profile card margin

* fix: move helper modules to utils and use single-letter avatar fallback

* fix: keep profile icon visible on sidebar collapse

* fix: sidebar account trigger labeling and profile reset prefs
---
 .../frontend/src/components/app-sidebar.tsx   |  22 ++-
 studio/frontend/src/features/auth/index.ts    |   1 +
 .../profile-personalization-panel.tsx         | 157 ++++++++++++++++++
 .../profile/components/user-avatar.tsx        |  45 +++++
 .../profile/hooks/use-effective-profile.ts    |  19 +++
 studio/frontend/src/features/profile/index.ts |   6 +
 .../profile/stores/user-profile-store.ts      |  24 +++
 .../features/profile/utils/avatar-initials.ts |  13 ++
 .../src/features/profile/utils/jwt-subject.ts |  21 +++
 .../profile/utils/resize-image-file.ts        |  53 ++++++
 .../src/features/settings/settings-dialog.tsx |   7 +-
 .../settings/stores/settings-dialog-store.ts  |   3 +-
 .../features/settings/tabs/general-tab.tsx    |   2 +
 .../features/settings/tabs/profile-tab.tsx    |  19 +++
 14 files changed, 382 insertions(+), 10 deletions(-)
 create mode 100644 studio/frontend/src/features/profile/components/profile-personalization-panel.tsx
 create mode 100644 studio/frontend/src/features/profile/components/user-avatar.tsx
 create mode 100644 studio/frontend/src/features/profile/hooks/use-effective-profile.ts
 create mode 100644 studio/frontend/src/features/profile/index.ts
 create mode 100644 studio/frontend/src/features/profile/stores/user-profile-store.ts
 create mode 100644 studio/frontend/src/features/profile/utils/avatar-initials.ts
 create mode 100644 studio/frontend/src/features/profile/utils/jwt-subject.ts
 create mode 100644 studio/frontend/src/features/profile/utils/resize-image-file.ts
 create mode 100644 studio/frontend/src/features/settings/tabs/profile-tab.tsx

diff --git a/studio/frontend/src/components/app-sidebar.tsx b/studio/frontend/src/components/app-sidebar.tsx
index 329175fa9c..032244ac39 100644
--- a/studio/frontend/src/components/app-sidebar.tsx
+++ b/studio/frontend/src/components/app-sidebar.tsx
@@ -55,6 +55,7 @@ import { Link, useNavigate, useRouterState } from "@tanstack/react-router";
 import { motion } from "motion/react";
 import { useTrainingRuntimeStore } from "@/features/training";
 import { useSettingsDialogStore } from "@/features/settings";
+import { useEffectiveProfile, UserAvatar } from "@/features/profile";
 import { usePlatformStore } from "@/config/env";
 import { TOUR_OPEN_EVENT } from "@/features/tour";
 import {
@@ -185,6 +186,7 @@ export function AppSidebar() {
   const effectiveRunsOpen = isStudioRoute || runsOpen;
 
   const isRecipesRoute = pathname.startsWith("/data-recipes");
+  const { displayTitle, avatarDataUrl } = useEffectiveProfile();
 
   const { items: chatItems } = useChatSidebarItems();
   const storeThreadId = useChatRuntimeStore((s) => s.activeThreadId);
@@ -495,22 +497,26 @@ export function AppSidebar() {
         )}
       </SidebarContent>
 
-      <SidebarFooter className="border-t border-sidebar-border">
+      <SidebarFooter className="border-t border-sidebar-border group-data-[collapsible=icon]:border-t-0">
         <SidebarMenu>
           <SidebarMenuItem>
             <DropdownMenu>
               <DropdownMenuTrigger asChild>
                 <SidebarMenuButton
                   size="lg"
-                  className="data-[state=open]:bg-sidebar-accent data-[state=open]:text-sidebar-accent-foreground"
+                  aria-label={`${displayTitle} account menu`}
+                  className="data-[state=open]:bg-sidebar-accent data-[state=open]:text-sidebar-accent-foreground group-data-[collapsible=icon]:overflow-visible group-data-[collapsible=icon]:hover:bg-transparent group-data-[collapsible=icon]:data-[state=open]:bg-transparent"
                 >
-                  <img
-                    src="/Sloth emojis/sloth rounded.png"
-                    alt="Unsloth"
-                    className="size-8 rounded-lg shrink-0"
-                  />
+                  <div className="shrink-0">
+                    <UserAvatar
+                      name={displayTitle}
+                      imageUrl={avatarDataUrl}
+                      size="sm"
+                      className="!size-8"
+                    />
+                  </div>
                   <div className="flex flex-col gap-0.5 leading-none group-data-[collapsible=icon]:hidden">
-                    <span className="truncate text-sm font-semibold">Unsloth</span>
+                    <span className="truncate text-sm font-semibold">{displayTitle}</span>
                     <span className="truncate text-[11px] text-muted-foreground">Train</span>
                   </div>
                   <ChevronsUpDown strokeWidth={1.25} className="ml-auto size-4 text-muted-foreground group-data-[collapsible=icon]:hidden" />
diff --git a/studio/frontend/src/features/auth/index.ts b/studio/frontend/src/features/auth/index.ts
index 75db92432c..af629cfb9a 100644
--- a/studio/frontend/src/features/auth/index.ts
+++ b/studio/frontend/src/features/auth/index.ts
@@ -5,6 +5,7 @@ export { LoginPage } from "./login-page";
 export { ChangePasswordPage } from "./change-password-page";
 export { authFetch, refreshSession } from "./api";
 export {
+  getAuthToken,
   getPostAuthRoute,
   hasAuthToken,
   hasRefreshToken,
diff --git a/studio/frontend/src/features/profile/components/profile-personalization-panel.tsx b/studio/frontend/src/features/profile/components/profile-personalization-panel.tsx
new file mode 100644
index 0000000000..994115bc19
--- /dev/null
+++ b/studio/frontend/src/features/profile/components/profile-personalization-panel.tsx
@@ -0,0 +1,157 @@
+// SPDX-License-Identifier: AGPL-3.0-only
+// Copyright 2026-present the Unsloth AI Inc. team. All rights reserved. See /studio/LICENSE.AGPL-3.0
+
+import { Button } from "@/components/ui/button";
+import { Input } from "@/components/ui/input";
+import { Label } from "@/components/ui/label";
+import { getAuthToken } from "@/features/auth";
+import { toastError, toastSuccess } from "@/shared/toast";
+import { Camera } from "lucide-react";
+import { useMemo, useRef, useState } from "react";
+import { decodeJwtSubject } from "../utils/jwt-subject";
+import { resizeImageFileToDataUrl } from "../utils/resize-image-file";
+import { useUserProfileStore } from "../stores/user-profile-store";
+import { UserAvatar } from "./user-avatar";
+
+const PROFILE_STORAGE_KEY = "unsloth_user_profile";
+
+function readPersistedProfile(): { displayName: string; avatarDataUrl: string | null } | null {
+  try {
+    const raw = window.localStorage.getItem(PROFILE_STORAGE_KEY);
+    if (!raw) return null;
+    const parsed = JSON.parse(raw) as unknown;
+    if (!parsed || typeof parsed !== "object") return null;
+
+    // Zustand persist shape: { state: {...}, version }
+    const maybeState = "state" in parsed ? (parsed as { state?: unknown }).state : parsed;
+    if (!maybeState || typeof maybeState !== "object") return null;
+    const state = maybeState as { displayName?: unknown; avatarDataUrl?: unknown };
+
+    return {
+      displayName: typeof state.displayName === "string" ? state.displayName : "",
+      avatarDataUrl: typeof state.avatarDataUrl === "string" ? state.avatarDataUrl : null,
+    };
+  } catch {
+    return null;
+  }
+}
+
+export function ProfilePersonalizationPanel() {
+  const displayName = useUserProfileStore((s) => s.displayName);
+  const avatarDataUrl = useUserProfileStore((s) => s.avatarDataUrl);
+  const setDisplayName = useUserProfileStore((s) => s.setDisplayName);
+  const setAvatarDataUrl = useUserProfileStore((s) => s.setAvatarDataUrl);
+
+  const [imageError, setImageError] = useState<string | null>(null);
+  const [draftName, setDraftName] = useState(displayName);
+  const fileInputRef = useRef<HTMLInputElement>(null);
+
+  const sessionSub = decodeJwtSubject(getAuthToken()) ?? "";
+  const previewName = draftName.trim() || sessionSub || "Unsloth";
+  const hasNameChanges = useMemo(
+    () => draftName.trim() !== displayName.trim(),
+    [draftName, displayName],
+  );
+
+  const saveName = () => {
+    const trimmed = draftName.trim();
+    if (trimmed !== draftName) setDraftName(trimmed);
+    if (trimmed !== displayName) {
+      setDisplayName(trimmed);
+      const persisted = readPersistedProfile();
+      if (persisted && persisted.displayName === trimmed) {
+        toastSuccess("Profile name saved");
+      } else {
+        toastError(
+          "Could not persist profile name",
+          "Name updated for this session, but may not persist after reload.",
+        );
+      }
+    }
+  };
+
+  const onPickFile = async (file: File | undefined) => {
+    if (!file) return;
+    setImageError(null);
+    try {
+      const dataUrl = await resizeImageFileToDataUrl(file);
+      setAvatarDataUrl(dataUrl);
+      const persisted = readPersistedProfile();
+      if (persisted && persisted.avatarDataUrl === dataUrl) {
+        toastSuccess("Profile photo updated");
+      } else {
+        toastError(
+          "Could not persist profile photo",
+          "Photo updated for this session, but may not persist after reload.",
+        );
+      }
+    } catch (e) {
+      const message = e instanceof Error ? e.message : "Could not use this image.";
+      setImageError(message);
+      toastError("Could not update profile photo", message);
+    }
+  };
+
+  return (
+    <div className="mx-auto flex w-full max-w-[640px] flex-col items-center gap-6 rounded-2xl border border-border/70 bg-muted/10 px-8 py-7">
+      <div className="relative">
+        <UserAvatar
+          name={previewName}
+          imageUrl={avatarDataUrl}
+          size="lg"
+          className="size-[124px] text-[3.15rem]"
+        />
+        <input
+          ref={fileInputRef}
+          type="file"
+          accept="image/jpeg,image/png,image/webp,image/gif"
+          className="sr-only"
+          onChange={(e) => {
+            void onPickFile(e.target.files?.[0]);
+            e.target.value = "";
+          }}
+        />
+        <button
+          type="button"
+          onClick={() => fileInputRef.current?.click()}
+          className="absolute right-0 bottom-0 -translate-x-[15.625%] -translate-y-[15.625%] flex size-8 items-center justify-center rounded-full border border-border bg-background text-foreground shadow-sm transition-colors hover:bg-muted focus-visible:outline-none focus-visible:ring-2 focus-visible:ring-ring focus-visible:ring-offset-2 focus-visible:ring-offset-background"
+          aria-label="Change profile picture"
+        >
+          <Camera className="size-3.5" strokeWidth={2} />
+        </button>
+      </div>
+
+      <div className="flex w-full max-w-[560px] flex-col gap-2">
+        <Label htmlFor="profile-display-name" className="text-xs font-medium text-muted-foreground">
+          Display name
+        </Label>
+        <div className="flex items-center gap-2">
+          <Input
+            id="profile-display-name"
+            type="text"
+            value={draftName}
+            onChange={(e) => setDraftName(e.target.value)}
+            onKeyDown={(e) => {
+              if (e.key === "Enter") {
+                e.preventDefault();
+                saveName();
+              }
+            }}
+            autoComplete="off"
+            placeholder={sessionSub || "Unsloth"}
+            className="h-10 min-w-0 flex-1 rounded-lg text-sm"
+          />
+          <Button type="button" size="sm" className="h-10 px-5" onClick={saveName} disabled={!hasNameChanges}>
+            Save
+          </Button>
+        </div>
+      </div>
+
+      {imageError ? (
+        <p className="w-full text-xs text-destructive" role="alert">
+          {imageError}
+        </p>
+      ) : null}
+    </div>
+  );
+}
diff --git a/studio/frontend/src/features/profile/components/user-avatar.tsx b/studio/frontend/src/features/profile/components/user-avatar.tsx
new file mode 100644
index 0000000000..62e37f5133
--- /dev/null
+++ b/studio/frontend/src/features/profile/components/user-avatar.tsx
@@ -0,0 +1,45 @@
+// SPDX-License-Identifier: AGPL-3.0-only
+// Copyright 2026-present the Unsloth AI Inc. team. All rights reserved. See /studio/LICENSE.AGPL-3.0
+
+import { cn } from "@/lib/utils";
+import { avatarBgStyle, initialsFromName } from "../utils/avatar-initials";
+
+type UserAvatarProps = {
+  name: string;
+  imageUrl: string | null;
+  size: "sm" | "md" | "lg";
+  className?: string;
+};
+
+const SIZE: Record<"sm" | "md" | "lg", string> = {
+  sm: "size-9 text-xs",
+  md: "size-11 text-sm",
+  /** ~10% larger than `size-24` / `text-2xl` for the edit-profile dialog. */
+  lg: "size-[106px] text-[1.65rem]",
+};
+
+export function UserAvatar({ name, imageUrl, size, className }: UserAvatarProps) {
+  const label = initialsFromName(name);
+
+  if (imageUrl) {
+    return (
+      <span className={cn("relative inline-flex shrink-0 overflow-hidden rounded-full", SIZE[size], className)}>
+        <img src={imageUrl} alt="" className="size-full object-cover" />
+      </span>
+    );
+  }
+
+  return (
+    <span
+      style={avatarBgStyle()}
+      className={cn(
+        "inline-flex shrink-0 items-center justify-center rounded-full font-semibold text-white",
+        SIZE[size],
+        className,
+      )}
+      aria-hidden
+    >
+      {label}
+    </span>
+  );
+}
diff --git a/studio/frontend/src/features/profile/hooks/use-effective-profile.ts b/studio/frontend/src/features/profile/hooks/use-effective-profile.ts
new file mode 100644
index 0000000000..3b64519e92
--- /dev/null
+++ b/studio/frontend/src/features/profile/hooks/use-effective-profile.ts
@@ -0,0 +1,19 @@
+// SPDX-License-Identifier: AGPL-3.0-only
+// Copyright 2026-present the Unsloth AI Inc. team. All rights reserved. See /studio/LICENSE.AGPL-3.0
+
+import { getAuthToken } from "@/features/auth";
+import { decodeJwtSubject } from "../utils/jwt-subject";
+import { useUserProfileStore } from "../stores/user-profile-store";
+
+export function useEffectiveProfile() {
+  const displayName = useUserProfileStore((s) => s.displayName);
+  const avatarDataUrl = useUserProfileStore((s) => s.avatarDataUrl);
+
+  const sessionSub = decodeJwtSubject(getAuthToken());
+  const dn = displayName.trim();
+  return {
+    sessionSub,
+    displayTitle: dn || "Unsloth",
+    avatarDataUrl,
+  };
+}
diff --git a/studio/frontend/src/features/profile/index.ts b/studio/frontend/src/features/profile/index.ts
new file mode 100644
index 0000000000..feec20607e
--- /dev/null
+++ b/studio/frontend/src/features/profile/index.ts
@@ -0,0 +1,6 @@
+// SPDX-License-Identifier: AGPL-3.0-only
+// Copyright 2026-present the Unsloth AI Inc. team. All rights reserved. See /studio/LICENSE.AGPL-3.0
+
+export { ProfilePersonalizationPanel } from "./components/profile-personalization-panel";
+export { UserAvatar } from "./components/user-avatar";
+export { useEffectiveProfile } from "./hooks/use-effective-profile";
diff --git a/studio/frontend/src/features/profile/stores/user-profile-store.ts b/studio/frontend/src/features/profile/stores/user-profile-store.ts
new file mode 100644
index 0000000000..5bbb4d11c9
--- /dev/null
+++ b/studio/frontend/src/features/profile/stores/user-profile-store.ts
@@ -0,0 +1,24 @@
+// SPDX-License-Identifier: AGPL-3.0-only
+// Copyright 2026-present the Unsloth AI Inc. team. All rights reserved. See /studio/LICENSE.AGPL-3.0
+
+import { create } from "zustand";
+import { persist } from "zustand/middleware";
+
+export interface UserProfileState {
+  displayName: string;
+  avatarDataUrl: string | null;
+  setDisplayName: (displayName: string) => void;
+  setAvatarDataUrl: (avatarDataUrl: string | null) => void;
+}
+
+export const useUserProfileStore = create<UserProfileState>()(
+  persist(
+    (set) => ({
+      displayName: "",
+      avatarDataUrl: null,
+      setDisplayName: (displayName) => set({ displayName }),
+      setAvatarDataUrl: (avatarDataUrl) => set({ avatarDataUrl }),
+    }),
+    { name: "unsloth_user_profile" },
+  ),
+);
diff --git a/studio/frontend/src/features/profile/utils/avatar-initials.ts b/studio/frontend/src/features/profile/utils/avatar-initials.ts
new file mode 100644
index 0000000000..926f57f3b5
--- /dev/null
+++ b/studio/frontend/src/features/profile/utils/avatar-initials.ts
@@ -0,0 +1,13 @@
+// SPDX-License-Identifier: AGPL-3.0-only
+// Copyright 2026-present the Unsloth AI Inc. team. All rights reserved. See /studio/LICENSE.AGPL-3.0
+
+export function initialsFromName(name: string): string {
+  const trimmed = name.trim();
+  if (!trimmed) return "?";
+  return trimmed[0]!.toUpperCase();
+}
+
+/** Default blue background for avatar fallback (readable white text). */
+export function avatarBgStyle(): { backgroundColor: string } {
+  return { backgroundColor: "hsl(217 58% 48%)" };
+}
diff --git a/studio/frontend/src/features/profile/utils/jwt-subject.ts b/studio/frontend/src/features/profile/utils/jwt-subject.ts
new file mode 100644
index 0000000000..9c7596966a
--- /dev/null
+++ b/studio/frontend/src/features/profile/utils/jwt-subject.ts
@@ -0,0 +1,21 @@
+// SPDX-License-Identifier: AGPL-3.0-only
+// Copyright 2026-present the Unsloth AI Inc. team. All rights reserved. See /studio/LICENSE.AGPL-3.0
+
+/**
+ * Read the JWT `sub` claim for display purposes only (not verified).
+ */
+export function decodeJwtSubject(token: string | null): string | null {
+  if (!token) return null;
+  try {
+    const parts = token.split(".");
+    if (parts.length < 2) return null;
+    const payload = parts[1];
+    const base64 = payload.replace(/-/g, "+").replace(/_/g, "/");
+    const padded = base64 + "=".repeat((4 - (base64.length % 4)) % 4);
+    const json = atob(padded);
+    const parsed = JSON.parse(json) as { sub?: unknown };
+    return typeof parsed.sub === "string" ? parsed.sub : null;
+  } catch {
+    return null;
+  }
+}
diff --git a/studio/frontend/src/features/profile/utils/resize-image-file.ts b/studio/frontend/src/features/profile/utils/resize-image-file.ts
new file mode 100644
index 0000000000..3f829ba975
--- /dev/null
+++ b/studio/frontend/src/features/profile/utils/resize-image-file.ts
@@ -0,0 +1,53 @@
+// SPDX-License-Identifier: AGPL-3.0-only
+// Copyright 2026-present the Unsloth AI Inc. team. All rights reserved. See /studio/LICENSE.AGPL-3.0
+
+const MAX_EDGE = 256;
+const MAX_BYTES = 380_000;
+
+function loadImage(file: File): Promise<HTMLImageElement> {
+  return new Promise((resolve, reject) => {
+    const url = URL.createObjectURL(file);
+    const img = new Image();
+    img.onload = () => {
+      URL.revokeObjectURL(url);
+      resolve(img);
+    };
+    img.onerror = () => {
+      URL.revokeObjectURL(url);
+      reject(new Error("Could not load image"));
+    };
+    img.src = url;
+  });
+}
+
+/**
+ * Downscale and re-encode as JPEG so localStorage stays within reasonable size.
+ */
+export async function resizeImageFileToDataUrl(file: File): Promise<string> {
+  const img = await loadImage(file);
+  const w = img.naturalWidth;
+  const h = img.naturalHeight;
+  if (!w || !h) throw new Error("Invalid image dimensions");
+
+  const scale = Math.min(1, MAX_EDGE / Math.max(w, h));
+  const cw = Math.max(1, Math.round(w * scale));
+  const ch = Math.max(1, Math.round(h * scale));
+
+  const canvas = document.createElement("canvas");
+  canvas.width = cw;
+  canvas.height = ch;
+  const ctx = canvas.getContext("2d");
+  if (!ctx) throw new Error("Canvas not available");
+  ctx.drawImage(img, 0, 0, cw, ch);
+
+  let quality = 0.88;
+  let dataUrl = canvas.toDataURL("image/jpeg", quality);
+  while (dataUrl.length > MAX_BYTES * 1.35 && quality > 0.45) {
+    quality -= 0.08;
+    dataUrl = canvas.toDataURL("image/jpeg", quality);
+  }
+  if (dataUrl.length > MAX_BYTES * 1.35) {
+    throw new Error("Image is still too large after compression. Try a smaller file.");
+  }
+  return dataUrl;
+}
diff --git a/studio/frontend/src/features/settings/settings-dialog.tsx b/studio/frontend/src/features/settings/settings-dialog.tsx
index 0e2083b31a..cdccf2fc72 100644
--- a/studio/frontend/src/features/settings/settings-dialog.tsx
+++ b/studio/frontend/src/features/settings/settings-dialog.tsx
@@ -15,6 +15,7 @@ import {
   PaintBrush02Icon,
   Settings02Icon,
   SparklesIcon,
+  UserIcon,
 } from "@hugeicons/core-free-icons";
 import { HugeiconsIcon } from "@hugeicons/react";
 import { motion, useReducedMotion } from "motion/react";
@@ -24,6 +25,7 @@ import { ApiKeysTab } from "./tabs/api-keys-tab";
 import { AppearanceTab } from "./tabs/appearance-tab";
 import { ChatTab } from "./tabs/chat-tab";
 import { GeneralTab } from "./tabs/general-tab";
+import { ProfileTab } from "./tabs/profile-tab";
 
 interface TabDef {
   id: SettingsTab;
@@ -33,6 +35,7 @@ interface TabDef {
 
 const TABS: TabDef[] = [
   { id: "general", label: "General", icon: Settings02Icon },
+  { id: "profile", label: "Profile", icon: UserIcon },
   { id: "appearance", label: "Appearance", icon: PaintBrush02Icon },
   { id: "chat", label: "Chat", icon: Message01Icon },
   { id: "api-keys", label: "API Keys", icon: Key01Icon },
@@ -43,6 +46,8 @@ function renderTab(tab: SettingsTab) {
   switch (tab) {
     case "general":
       return <GeneralTab />;
+    case "profile":
+      return <ProfileTab />;
     case "appearance":
       return <AppearanceTab />;
     case "chat":
@@ -131,7 +136,7 @@ export function SettingsDialog() {
             >
               <HugeiconsIcon icon={Cancel01Icon} className="size-4" />
             </button>
-            <div className="flex min-h-0 flex-1 flex-col overflow-y-auto p-6 pr-12">
+            <div className="flex min-h-0 flex-1 flex-col overflow-y-auto p-6">
               {renderTab(activeTab)}
             </div>
           </main>
diff --git a/studio/frontend/src/features/settings/stores/settings-dialog-store.ts b/studio/frontend/src/features/settings/stores/settings-dialog-store.ts
index 75eb53048e..d1fd4a1d0f 100644
--- a/studio/frontend/src/features/settings/stores/settings-dialog-store.ts
+++ b/studio/frontend/src/features/settings/stores/settings-dialog-store.ts
@@ -5,6 +5,7 @@ import { create } from "zustand";
 
 export type SettingsTab =
   | "general"
+  | "profile"
   | "appearance"
   | "chat"
   | "api-keys"
@@ -28,7 +29,7 @@ function loadInitialTab(): SettingsTab {
   } catch {
     return "general";
   }
-  const valid: SettingsTab[] = ["general", "appearance", "chat", "api-keys", "about"];
+  const valid: SettingsTab[] = ["general", "profile", "appearance", "chat", "api-keys", "about"];
   return valid.includes(stored as SettingsTab) ? (stored as SettingsTab) : "general";
 }
 
diff --git a/studio/frontend/src/features/settings/tabs/general-tab.tsx b/studio/frontend/src/features/settings/tabs/general-tab.tsx
index 874508a6fb..6081e90605 100644
--- a/studio/frontend/src/features/settings/tabs/general-tab.tsx
+++ b/studio/frontend/src/features/settings/tabs/general-tab.tsx
@@ -56,6 +56,8 @@ const PREFS_KEYS: string[] = [
   "unsloth_training_config_v1",
   "unsloth_prev_max_steps",
   "unsloth_prev_save_steps",
+  // Profile personalization
+  "unsloth_user_profile",
   // Guided tour flags
   "tour:studio:v1",
 ];
diff --git a/studio/frontend/src/features/settings/tabs/profile-tab.tsx b/studio/frontend/src/features/settings/tabs/profile-tab.tsx
new file mode 100644
index 0000000000..2ae283b767
--- /dev/null
+++ b/studio/frontend/src/features/settings/tabs/profile-tab.tsx
@@ -0,0 +1,19 @@
+// SPDX-License-Identifier: AGPL-3.0-only
+// Copyright 2026-present the Unsloth AI Inc. team. All rights reserved. See /studio/LICENSE.AGPL-3.0
+
+import { ProfilePersonalizationPanel } from "@/features/profile";
+
+export function ProfileTab() {
+  return (
+    <div className="flex flex-col gap-6">
+      <header className="flex flex-col gap-1">
+        <h1 className="text-lg font-semibold font-heading">Profile</h1>
+        <p className="text-xs text-muted-foreground">
+          Update how your profile appears in Studio.
+        </p>
+      </header>
+
+      <ProfilePersonalizationPanel />
+    </div>
+  );
+}

From d3215ce11341178eee6384435603d8a7e7173622 Mon Sep 17 00:00:00 2001
From: Lee Jackson <130007945+Imagineer99@users.noreply.github.com>
Date: Mon, 20 Apr 2026 20:14:49 +0100
Subject: [PATCH 140/216] Studio: Show LoRA live logs and update GGUF quant
 options (#5058)

* export: update GGUF quant list and ordering

* gguf: add Q2_K_L quantize flags for output and embeddings

* export: add live console logs for LoRA export flow

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

* fix: stream q2_k_l quantize logs and include subprocess error details

* fix: route Q2_K_L preset to q2_k ftype with q8_0 output+embeddings

---------

Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com>
Co-authored-by: Roland Tannous <115670425+rolandtannous@users.noreply.github.com>
---
 .../export/components/export-dialog.tsx       |  40 +++++--
 .../frontend/src/features/export/constants.ts |   6 +-
 unsloth/save.py                               | 111 ++++++++++++++++--
 3 files changed, 138 insertions(+), 19 deletions(-)

diff --git a/studio/frontend/src/features/export/components/export-dialog.tsx b/studio/frontend/src/features/export/components/export-dialog.tsx
index 401be3ff4b..176b10a52f 100644
--- a/studio/frontend/src/features/export/components/export-dialog.tsx
+++ b/studio/frontend/src/features/export/components/export-dialog.tsx
@@ -272,12 +272,13 @@ export function ExportDialog({
   exportSuccess,
   exportOutputPath,
 }: ExportDialogProps) {
-  // Live log capture is only meaningful for export methods that run
-  // a slow subprocess operation with interesting stdout: merged and
-  // gguf. LoRA adapter export is a fast disk write and would just
-  // show a blank panel, so we hide it there.
+  // Live log capture is useful for any export path executed by the
+  // backend worker, including LoRA adapter-only export.
   const showLogPanel =
-    exportMethod === "merged" || exportMethod === "gguf";
+    exportMethod === "merged" ||
+    exportMethod === "gguf" ||
+    exportMethod === "lora";
+  const showCompletionScreen = exportSuccess && !showLogPanel;
 
   const { lines: logLines, connected: logConnected, error: logError } =
     useExportLogs(exporting && showLogPanel, exportMethod, open);
@@ -314,7 +315,7 @@ export function ExportDialog({
         className={showLogPanel ? "sm:max-w-2xl" : "sm:max-w-lg"}
         onInteractOutside={(e) => { if (exporting) e.preventDefault(); }}
       >
-        {exportSuccess ? (
+        {showCompletionScreen ? (
           <>
             <div className="flex flex-col items-center gap-3 py-6">
               <div className="flex size-12 items-center justify-center rounded-full bg-emerald-500/10">
@@ -460,6 +461,27 @@ export function ExportDialog({
               )}
             </AnimatePresence>
 
+            {/* Success banner for log-driven exports.
+                Keep users on the log screen after completion so they can
+                inspect conversion output before closing. */}
+            {exportSuccess && showLogPanel && (
+              <div className="flex items-start gap-2 rounded-lg bg-emerald-500/10 p-3 text-sm text-emerald-700 dark:text-emerald-300">
+                <HugeiconsIcon icon={CheckmarkCircle02Icon} className="mt-0.5 size-4 shrink-0" />
+                <div className="flex min-w-0 flex-col gap-1">
+                  <span>
+                    {destination === "hub"
+                      ? "Export finished and pushed to Hugging Face Hub."
+                      : "Export finished successfully."}
+                  </span>
+                  {exportOutputPath ? (
+                    <code className="select-all break-all font-mono text-[12px] text-foreground/90" title={exportOutputPath}>
+                      {exportOutputPath}
+                    </code>
+                  ) : null}
+                </div>
+              </div>
+            )}
+
             {/* Error banner */}
             {exportError && (
               <div className="flex items-start gap-2 rounded-lg bg-destructive/10 p-3 text-sm text-destructive">
@@ -577,14 +599,16 @@ export function ExportDialog({
                 onClick={() => onOpenChange(false)}
                 disabled={exporting}
               >
-                Cancel
+                {exportSuccess ? "Done" : "Cancel"}
               </Button>
-              <Button onClick={onExport} disabled={exporting}>
+              <Button onClick={onExport} disabled={exporting || exportSuccess}>
                 {exporting ? (
                   <span className="flex items-center gap-2">
                     <Spinner className="size-4" />
                     Exporting…
                   </span>
+                ) : exportSuccess ? (
+                  "Export Complete"
                 ) : (
                   "Start Export"
                 )}
diff --git a/studio/frontend/src/features/export/constants.ts b/studio/frontend/src/features/export/constants.ts
index ff78490599..e9c3b8c95b 100644
--- a/studio/frontend/src/features/export/constants.ts
+++ b/studio/frontend/src/features/export/constants.ts
@@ -36,14 +36,14 @@ export const EXPORT_METHODS: {
 ];
 
 export const QUANT_OPTIONS = [
+  { value: "q2_k_l", label: "Q2_K_L", size: "~2.9 GB" },
   { value: "q3_k_m", label: "Q3_K_M", size: "~3.5 GB" },
-  { value: "q4_0", label: "Q4_0", size: "~4.1 GB" },
   { value: "q4_k_m", label: "Q4_K_M", size: "~4.8 GB", recommended: true },
-  { value: "q5_0", label: "Q5_0", size: "~5.0 GB" },
   { value: "q5_k_m", label: "Q5_K_M", size: "~5.6 GB" },
+  { value: "q6_k", label: "Q6_K", size: "~6.6 GB" },
   { value: "q8_0", label: "Q8_0", size: "~8.2 GB" },
+  { value: "bf16", label: "BF16", size: "~14.2 GB" },
   { value: "f16", label: "F16", size: "~14.2 GB" },
-  { value: "f32", label: "F32", size: "~28.4 GB" },
 ];
 
 export function getEstimatedSize(
diff --git a/unsloth/save.py b/unsloth/save.py
index 3c318fab02..22ae5487d7 100644
--- a/unsloth/save.py
+++ b/unsloth/save.py
@@ -122,6 +122,7 @@ ALLOWED_QUANTS = {
     "q4_k_m": "Recommended. Uses Q6_K for half of the attention.wv and feed_forward.w2 tensors, else Q4_K",
     "q5_k_m": "Recommended. Uses Q6_K for half of the attention.wv and feed_forward.w2 tensors, else Q5_K",
     "q2_k": "Uses Q4_K for the attention.vw and feed_forward.w2 tensors, Q2_K for the other tensors.",
+    "q2_k_l": "Q2_K_L with q8_0 output/token embeddings for higher quality than plain Q2_K.",
     "q3_k_l": "Uses Q5_K for the attention.wv, attention.wo, and feed_forward.w2 tensors, else Q3_K",
     "q3_k_m": "Uses Q4_K for the attention.wv, attention.wo, and feed_forward.w2 tensors, else Q3_K",
     "q3_k_s": "Uses Q3_K for all tensors",
@@ -153,6 +154,89 @@ def print_quantization_methods():
         print(f'"{key}"  ==> {value}')
 
 
+def _quantize_q2_k_l(
+    input_gguf: Union[str, os.PathLike],
+    output_gguf: Union[str, os.PathLike],
+    quantizer_location: Union[str, os.PathLike],
+    n_threads: int,
+    print_output: bool = True,
+):
+    # "Q2_K_L" is a Unsloth-side preset, not a native llama.cpp ftype. It
+    # maps to the `q2_k` ftype with `--output-tensor-type q8_0` and
+    # `--token-embedding-type q8_0` so the output/embedding tensors retain
+    # higher precision than a plain Q2_K quant.
+    command = [
+        str(quantizer_location),
+        "--output-tensor-type",
+        "q8_0",
+        "--token-embedding-type",
+        "q8_0",
+        str(input_gguf),
+        str(output_gguf),
+        "q2_k",
+        str(n_threads),
+    ]
+
+    if print_output:
+        print(
+            "Unsloth: Quantizing as Q2_K_L preset "
+            "(q2_k + --output-tensor-type q8_0 --token-embedding-type q8_0)..."
+        )
+
+    try:
+        if print_output:
+            with subprocess.Popen(
+                command,
+                shell = False,
+                text = True,
+                stdout = subprocess.PIPE,
+                stderr = subprocess.STDOUT,
+                bufsize = 1,
+            ) as sp:
+                assert sp.stdout is not None
+                for line in sp.stdout:
+                    print(line, end = "", flush = True)
+
+                returncode = sp.wait()
+                if returncode != 0:
+                    raise RuntimeError(
+                        f"Failed to quantize {input_gguf} to q2_k_l: process exited with code {returncode}"
+                    )
+        else:
+            subprocess.run(
+                command,
+                shell = False,
+                check = True,
+                capture_output = True,
+                text = True,
+            )
+    except subprocess.CalledProcessError as e:
+        if print_output and hasattr(e, "stdout") and e.stdout:
+            print(e.stdout)
+        error_details = ""
+        if hasattr(e, "stdout") and e.stdout:
+            error_details += f"\nSubprocess stdout:\n{e.stdout}"
+        if hasattr(e, "stderr") and e.stderr:
+            error_details += f"\nSubprocess stderr:\n{e.stderr}"
+        raise RuntimeError(
+            f"Failed to quantize {input_gguf} to q2_k_l: {e}{error_details}"
+        )
+
+    output_path = Path(output_gguf)
+    if not output_path.exists():
+        raise RuntimeError(
+            f"Quantization failed - output file {output_gguf} not created"
+        )
+
+    if print_output:
+        file_size_bytes = output_path.stat().st_size
+        file_size_gb = file_size_bytes / (1024**3)
+        print(
+            f"Unsloth: Successfully quantized to {output_gguf} (size: {file_size_gb:.2f}GB)"
+        )
+    return str(output_gguf)
+
+
 def check_if_sentencepiece_model(
     model, temporary_location = "_unsloth_sentencepiece_temp"
 ):
@@ -1305,14 +1389,23 @@ def save_to_gguf(
                     gguf_directory, f"{model_name}.{quant_method.upper()}.gguf"
                 )
                 try:
-                    # Use the quantize_gguf function we created
-                    quantized_file = quantize_gguf(
-                        input_gguf = base_gguf,
-                        output_gguf = output_location,
-                        quant_type = quant_method,
-                        quantizer_location = quantizer_location,
-                        print_output = print_output,
-                    )
+                    if quant_method == "q2_k_l":
+                        quantized_file = _quantize_q2_k_l(
+                            input_gguf = base_gguf,
+                            output_gguf = output_location,
+                            quantizer_location = quantizer_location,
+                            n_threads = n_cpus,
+                            print_output = print_output,
+                        )
+                    else:
+                        # Use unsloth-zoo's standard quantization for all other methods
+                        quantized_file = quantize_gguf(
+                            input_gguf = base_gguf,
+                            output_gguf = output_location,
+                            quant_type = quant_method,
+                            quantizer_location = quantizer_location,
+                            print_output = print_output,
+                        )
                     all_saved_locations.append(quantized_file)
                     quants_created = True
                 except Exception as e:
@@ -1880,6 +1973,7 @@ def unsloth_save_pretrained_gguf(
     "q4_k_m"  : "Recommended. Uses Q6_K for half of the attention.wv and feed_forward.w2 tensors, else Q4_K",
     "q5_k_m"  : "Recommended. Uses Q6_K for half of the attention.wv and feed_forward.w2 tensors, else Q5_K",
     "q2_k"    : "Uses Q4_K for the attention.vw and feed_forward.w2 tensors, Q2_K for the other tensors.",
+    "q2_k_l"  : "Q2_K_L with --output-tensor-type q8_0 --token-embedding-type q8_0.",
     "q3_k_l"  : "Uses Q5_K for the attention.wv, attention.wo, and feed_forward.w2 tensors, else Q3_K",
     "q3_k_m"  : "Uses Q4_K for the attention.wv, attention.wo, and feed_forward.w2 tensors, else Q3_K",
     "q3_k_s"  : "Uses Q3_K for all tensors",
@@ -2203,6 +2297,7 @@ def unsloth_push_to_hub_gguf(
     "q4_k_m"  : "Recommended. Uses Q6_K for half of the attention.wv and feed_forward.w2 tensors, else Q4_K",
     "q5_k_m"  : "Recommended. Uses Q6_K for half of the attention.wv and feed_forward.w2 tensors, else Q5_K",
     "q2_k"    : "Uses Q4_K for the attention.vw and feed_forward.w2 tensors, Q2_K for the other tensors.",
+    "q2_k_l"  : "Q2_K_L with --output-tensor-type q8_0 --token-embedding-type q8_0.",
     "q3_k_l"  : "Uses Q5_K for the attention.wv, attention.wo, and feed_forward.w2 tensors, else Q3_K",
     "q3_k_m"  : "Uses Q4_K for the attention.wv, attention.wo, and feed_forward.w2 tensors, else Q3_K",
     "q3_k_s"  : "Uses Q3_K for all tensors",

From 0a5c61ffcc0675699280491c91d70367b06ed898 Mon Sep 17 00:00:00 2001
From: Konstantin Azizov <azizovkostya97@gmail.com>
Date: Mon, 20 Apr 2026 21:18:18 +0200
Subject: [PATCH 141/216] fix: prefer mainstream clipboard copy over deprecated
 one (#5109)

Fixes #5097

Co-authored-by: Roland Tannous <115670425+rolandtannous@users.noreply.github.com>
---
 .../components/assistant-ui/markdown-text.tsx |  8 ++--
 .../src/components/assistant-ui/reasoning.tsx |  4 +-
 .../src/components/assistant-ui/thread.tsx    |  4 +-
 .../assistant-ui/tool-ui-python.tsx           |  4 +-
 .../assistant-ui/tool-ui-terminal.tsx         |  4 +-
 .../settings/components/api-key-row.tsx       |  2 +-
 .../settings/components/key-reveal-card.tsx   |  4 +-
 .../components/update-studio-instructions.tsx |  4 +-
 .../settings/components/usage-examples.tsx    |  4 +-
 studio/frontend/src/lib/copy-to-clipboard.ts  | 41 ++++---------------
 10 files changed, 26 insertions(+), 53 deletions(-)

diff --git a/studio/frontend/src/components/assistant-ui/markdown-text.tsx b/studio/frontend/src/components/assistant-ui/markdown-text.tsx
index c7974db365..ea409591f0 100644
--- a/studio/frontend/src/components/assistant-ui/markdown-text.tsx
+++ b/studio/frontend/src/components/assistant-ui/markdown-text.tsx
@@ -272,8 +272,8 @@ function MermaidCopyButton({ source }: { source: string }) {
       type="button"
       className="absolute top-3.5 right-20 z-20 cursor-pointer text-muted-foreground transition-all hover:text-foreground"
       title="Copy Mermaid source"
-      onClick={() => {
-        if (!copyToClipboard(source)) {
+      onClick={async () => {
+        if (!(await copyToClipboard(source))) {
           return;
         }
         showCopied();
@@ -306,8 +306,8 @@ function CodeBlockActions({
           className={ACTION_BUTTON_CLASS}
           title="Copy code"
           disabled={disabled}
-          onClick={() => {
-            if (!copyToClipboard(source)) {
+          onClick={async () => {
+            if (!(await copyToClipboard(source))) {
               return;
             }
             showCopied();
diff --git a/studio/frontend/src/components/assistant-ui/reasoning.tsx b/studio/frontend/src/components/assistant-ui/reasoning.tsx
index e4306df4b9..4d46fc8d59 100644
--- a/studio/frontend/src/components/assistant-ui/reasoning.tsx
+++ b/studio/frontend/src/components/assistant-ui/reasoning.tsx
@@ -322,8 +322,8 @@ function ReasoningCopyButton({ startIndex, endIndex }: { startIndex: number; end
       .join("\n");
   });
 
-  const handleCopy = useCallback(() => {
-    if (copyToClipboard(reasoningText)) {
+  const handleCopy = useCallback(async () => {
+    if (await copyToClipboard(reasoningText)) {
       setCopied(true);
       if (resetRef.current) clearTimeout(resetRef.current);
       resetRef.current = setTimeout(() => setCopied(false), COPY_RESET_MS);
diff --git a/studio/frontend/src/components/assistant-ui/thread.tsx b/studio/frontend/src/components/assistant-ui/thread.tsx
index 77fdf28736..36713d665f 100644
--- a/studio/frontend/src/components/assistant-ui/thread.tsx
+++ b/studio/frontend/src/components/assistant-ui/thread.tsx
@@ -714,9 +714,9 @@ const CopyButton: FC = () => {
   const [copied, setCopied] = useState(false);
   const resetTimeoutRef = useRef<ReturnType<typeof setTimeout> | null>(null);
 
-  const handleCopy = () => {
+  const handleCopy = async () => {
     const text = aui.message().getCopyText();
-    if (copyToClipboard(text)) {
+    if (await copyToClipboard(text)) {
       setCopied(true);
       if (resetTimeoutRef.current) clearTimeout(resetTimeoutRef.current);
       resetTimeoutRef.current = setTimeout(() => {
diff --git a/studio/frontend/src/components/assistant-ui/tool-ui-python.tsx b/studio/frontend/src/components/assistant-ui/tool-ui-python.tsx
index 6aa590ae11..bab735104c 100644
--- a/studio/frontend/src/components/assistant-ui/tool-ui-python.tsx
+++ b/studio/frontend/src/components/assistant-ui/tool-ui-python.tsx
@@ -44,8 +44,8 @@ function CopyBtn({ text }: { text: string }) {
     };
   }, []);
 
-  const copy = useCallback(() => {
-    if (copyToClipboard(text)) {
+  const copy = useCallback(async () => {
+    if (await copyToClipboard(text)) {
       setCopied(true);
       if (timer.current) {
         clearTimeout(timer.current);
diff --git a/studio/frontend/src/components/assistant-ui/tool-ui-terminal.tsx b/studio/frontend/src/components/assistant-ui/tool-ui-terminal.tsx
index f233f951d3..1b1357ab3c 100644
--- a/studio/frontend/src/components/assistant-ui/tool-ui-terminal.tsx
+++ b/studio/frontend/src/components/assistant-ui/tool-ui-terminal.tsx
@@ -34,8 +34,8 @@ function CopyBtn({ text }: { text: string }) {
     };
   }, []);
 
-  const copy = useCallback(() => {
-    if (copyToClipboard(text)) {
+  const copy = useCallback(async () => {
+    if (await copyToClipboard(text)) {
       setCopied(true);
       if (timer.current) {
         clearTimeout(timer.current);
diff --git a/studio/frontend/src/features/settings/components/api-key-row.tsx b/studio/frontend/src/features/settings/components/api-key-row.tsx
index 74b4ba396e..f121045e1e 100644
--- a/studio/frontend/src/features/settings/components/api-key-row.tsx
+++ b/studio/frontend/src/features/settings/components/api-key-row.tsx
@@ -83,7 +83,7 @@ export function ApiKeyRow({
           </Button>
         </DropdownMenuTrigger>
         <DropdownMenuContent align="end">
-          <DropdownMenuItem onClick={() => copyToClipboard(prefix)}>
+          <DropdownMenuItem onClick={async () => { await copyToClipboard(prefix); }}>
             <HugeiconsIcon icon={Copy01Icon} className="size-3.5 mr-2" />
             Copy prefix
           </DropdownMenuItem>
diff --git a/studio/frontend/src/features/settings/components/key-reveal-card.tsx b/studio/frontend/src/features/settings/components/key-reveal-card.tsx
index 667351570c..ab258537da 100644
--- a/studio/frontend/src/features/settings/components/key-reveal-card.tsx
+++ b/studio/frontend/src/features/settings/components/key-reveal-card.tsx
@@ -17,8 +17,8 @@ export function KeyRevealCard({
 }) {
   const [copied, setCopied] = useState(false);
 
-  const handleCopy = () => {
-    if (copyToClipboard(rawKey)) {
+  const handleCopy = async () => {
+    if (await copyToClipboard(rawKey)) {
       setCopied(true);
       setTimeout(() => setCopied(false), 1800);
     }
diff --git a/studio/frontend/src/features/settings/components/update-studio-instructions.tsx b/studio/frontend/src/features/settings/components/update-studio-instructions.tsx
index e97ef9507c..4d0b87981b 100644
--- a/studio/frontend/src/features/settings/components/update-studio-instructions.tsx
+++ b/studio/frontend/src/features/settings/components/update-studio-instructions.tsx
@@ -39,8 +39,8 @@ function CopyableCommand({
     };
   }, []);
 
-  const handleCopy = () => {
-    if (!copyToClipboard(command)) {
+  const handleCopy = async () => {
+    if (!(await copyToClipboard(command))) {
       return;
     }
     setCopied(true);
diff --git a/studio/frontend/src/features/settings/components/usage-examples.tsx b/studio/frontend/src/features/settings/components/usage-examples.tsx
index 603edb3984..a2dc2c3774 100644
--- a/studio/frontend/src/features/settings/components/usage-examples.tsx
+++ b/studio/frontend/src/features/settings/components/usage-examples.tsx
@@ -67,8 +67,8 @@ export function UsageExamples() {
     [],
   );
 
-  const handleCopy = () => {
-    if (copyToClipboard(snippets[lang])) {
+  const handleCopy = async () => {
+    if (await copyToClipboard(snippets[lang])) {
       setCopied(true);
       setTimeout(() => setCopied(false), 1800);
     }
diff --git a/studio/frontend/src/lib/copy-to-clipboard.ts b/studio/frontend/src/lib/copy-to-clipboard.ts
index 7c97904d73..cce9ea7783 100644
--- a/studio/frontend/src/lib/copy-to-clipboard.ts
+++ b/studio/frontend/src/lib/copy-to-clipboard.ts
@@ -32,50 +32,23 @@ function copyWithExecCommand(text: string): boolean {
   }
 }
 
-export function copyToClipboard(text: string): boolean {
+export async function copyToClipboard(text: string): Promise<boolean> {
   if (typeof text !== "string" || text.length === 0) {
     return false;
   }
 
-  if (typeof document !== "undefined" && document.queryCommandSupported?.("copy") !== false) {
-    if (copyWithExecCommand(text)) return true;
-  }
-
-  // Async fallback for environments where execCommand is entirely unsupported
-  // but the Clipboard API is available (rare; kept for original contract parity).
-  if (typeof navigator?.clipboard?.writeText === "function") {
-    navigator.clipboard.writeText(text).then(
-      () => {},
-      () => {},
-    );
-    return true;
-  }
-
-  return false;
-}
-
-export async function copyToClipboardAsync(text: string): Promise<boolean> {
-  if (typeof text !== "string" || text.length === 0) {
-    return false;
-  }
-
-  // Prefer the async Clipboard API: avoids focus disruption in Radix
-  // focus-trapped dialogs where execCommand always fails.
+  // Primary: async Clipboard API
   if (typeof navigator?.clipboard?.writeText === "function") {
     try {
       await navigator.clipboard.writeText(text);
       return true;
-    } catch {
-      // Clipboard API rejected (e.g. NotAllowedError, permission policy).
-      // User activation is still valid through promise chains per spec, so
-      // execCommand can succeed for callers outside focus-trapped dialogs.
-      // Inside a Radix modal the focus trap will block textarea.focus() and
-      // execCommand returns false harmlessly.
-      return copyWithExecCommand(text);
+    } catch (error) {
+      console.warn("Async clipboard API failed, falling back to execCommand", error);
+      // Clipboard API rejected (NotAllowedError, insecure context, etc.)
+      // Fall through to execCommand fallback.
     }
   }
 
-  // No Clipboard API (older browser / non-secure context): still in the
-  // original user-gesture frame, so execCommand can work.
+  // Fallback: execCommand (works in Safari when called during user gesture)
   return copyWithExecCommand(text);
 }

From c20959dbf4cb984889aa96a1bae2d2269182509e Mon Sep 17 00:00:00 2001
From: Lee Jackson <130007945+Imagineer99@users.noreply.github.com>
Date: Mon, 20 Apr 2026 23:20:45 +0100
Subject: [PATCH 142/216] Studio: Improve chat composition, fix scroll
 behaviour, and refine sidebar UX (#5089)

* Chatbox, scroll, and menu fixes

- Fixed chatbox auto-expand height for multi-line text on the compare page
- Fixed chatbox UI to be consistent across compare and new chat
- Fixed scrolling being enabled on pages with no content, which also triggered the scroll-to-bottom button
- Fixed scroll-to-bottom button to only appear after scrolling up a reasonable amount instead of instantly
- Added shutdown studio button to the menu for easier access
- Fixed pop-up menu width to match the user button width

(cherry picked from commit cd4e390dfa84fe311fae79a781b96cc0ef5970a9)

* fix: correct compare scroll viewport and clean up chat composer UI polish

* Dark theme refactor and sidebar/chat UI refinements

- Complete refactoring of dark theme
- Replaced square rounded-corner user profile image with a circular bordered one
- Replaced user profile icon with 'U' initial and renamed label from 'Studio' to 'User'
- Chat bubbles now have a pointy top-right edge
- Sidebar menu tab line color selection is now consistent across all menus
- Tab-selection color animation now also applies to recent chats
- Removed 'Compare' menu autoselect when a compare chat conversation is selected
- Fixed UI consistency in Compare to match New Chat
- Removed sidebar animation and tab line, replaced with rounded selection for consistency
- Further adjustments to sidebar UI
- Further adjustments to compare chat UI

* Fixed sidebar collapse/expand for recent chats and recent runs not being clickable

* Chatbox, scroll, and menu fixes

- Fixed chatbox auto-expand height for multi-line text on the compare page
- Fixed chatbox UI to be consistent across compare and new chat
- Fixed scrolling being enabled on pages with no content, which also triggered the scroll-to-bottom button
- Fixed scroll-to-bottom button to only appear after scrolling up a reasonable amount instead of instantly
- Added shutdown studio button to the menu for easier access
- Fixed pop-up menu width to match the user button width

* Sidebar, fonts, and chat UI refinements

- Replaced logo PNG with real font text for 'unsloth' and 'BETA' label
- Added Hellix font and applied it across menus and UI elements
- Lighter scrollbar in the sidebar compared to other areas of the app
- Adjusted chat font and chat bubble styling
- Adjusted app menu design to stay consistent with the sidebar
- Adjusted text style for 'New Chat' and repositioned content/chatbox
- Adjusted model selector and top area UI
- Fixed footer text from 'LLM's' to 'LLMs'
- Fixed active selection border color incorrectly appearing on page refresh and during general navigation
- Logo now defaults to 'New Chat' when clicked

* Sidebar, model selector, and mobile UI fixes

- Further adjustments to sidebar UI and logo
- Changed right bar icon
- Model selector adjustments
- Collapsed sidebar now matches the content area background
- Adjusted Hellix font spacing across pages
- Fixed sidebar icon overlap on mobile screens

* Adjust sidebar icons

* Adjust sidebar icons

* Fixed compare chat UI and scrolling issues

* Fixed inference settings icon behavior and context info positioning

- Fixed top right inference settings icon to move into sidepanel during expand/collapse, matching left sidebar behavior
- Adjusted context information element positioning

* Fix: textarea overflow in system prompt editor

* Code block redesign, font, and chat bubble adjustments

- Redesigned code block colors and theme
- Changed code block font to Fira Code
- Fixed scrollbar disappearing when expanding/collapsing tool calls in chats
- Adjusted chat bubble background color

* Fix chat bubble background color in dark theme

* fix: restore textarea auto-sizing and scope prompt editor sizing

* fix: add explicit textarea field sizing for prompt editor overflow

* fix: generate chat nonce on click instead of render

* fix: respect training lock on logo navigation

* Refactor compare page dual chat scrolling behavior

* Revert "Refactor compare page dual chat scrolling behavior"

This reverts commit d056ec09f295563b2889404acf2921091e000bf3.

---------

Co-authored-by: sneakr <hauzin@hotmail.com>
Co-authored-by: Roland Tannous <115670425+rolandtannous@users.noreply.github.com>
---
 studio/frontend/public/circle-logo-small.png  | Bin 0 -> 14416 bytes
 .../fonts/FiraCode-VariableFont_wght.ttf      | Bin 0 -> 264848 bytes
 .../frontend/public/fonts/Hellix-Medium.woff  | Bin 0 -> 58448 bytes
 .../frontend/public/fonts/Hellix-Regular.woff | Bin 0 -> 58584 bytes
 studio/frontend/public/sidebar-logo-black.png | Bin 0 -> 9170 bytes
 studio/frontend/public/sidebar-logo-white.png | Bin 0 -> 9057 bytes
 studio/frontend/public/unsloth-beta-black.png | Bin 0 -> 160502 bytes
 studio/frontend/public/unsloth-beta-white.png | Bin 0 -> 156421 bytes
 .../frontend/src/components/app-sidebar.tsx   | 188 ++++----
 .../components/assistant-ui/code-plugin.ts    |  66 +++
 .../components/assistant-ui/code-themes.ts    |  30 ++
 .../assistant-ui/code-toggle-icon.tsx         |  22 +
 .../components/assistant-ui/markdown-text.tsx |   8 +-
 .../assistant-ui/model-selector.tsx           |  13 +-
 .../assistant-ui/model-selector/pickers.tsx   |   6 +-
 .../src/components/assistant-ui/reasoning.tsx |  44 +-
 .../src/components/assistant-ui/thread.tsx    | 103 ++---
 .../components/assistant-ui/tool-fallback.tsx |   4 +-
 .../components/assistant-ui/tool-group.tsx    |   4 +-
 studio/frontend/src/components/navbar.tsx     |   8 +-
 studio/frontend/src/components/ui/sidebar.tsx |   8 +-
 studio/frontend/src/components/ui/slider.tsx  |   4 +-
 .../frontend/src/components/ui/textarea.tsx   |  35 +-
 .../frontend/src/features/chat/chat-page.tsx  | 407 +++++++++++-------
 .../src/features/chat/chat-settings-sheet.tsx |  47 +-
 .../src/features/chat/shared-composer.tsx     |  42 +-
 .../components/recipe-graph-aux-node.tsx      |   4 +-
 .../components/recipe-graph-node.tsx          |   4 +-
 .../src/features/settings/settings-dialog.tsx |  15 +-
 studio/frontend/src/hooks/index.ts            |   1 +
 .../src/hooks/use-collapse-scroll-lock.ts     |  63 +++
 studio/frontend/src/index.css                 | 117 +++--
 32 files changed, 808 insertions(+), 435 deletions(-)
 create mode 100644 studio/frontend/public/circle-logo-small.png
 create mode 100644 studio/frontend/public/fonts/FiraCode-VariableFont_wght.ttf
 create mode 100644 studio/frontend/public/fonts/Hellix-Medium.woff
 create mode 100644 studio/frontend/public/fonts/Hellix-Regular.woff
 create mode 100644 studio/frontend/public/sidebar-logo-black.png
 create mode 100644 studio/frontend/public/sidebar-logo-white.png
 create mode 100644 studio/frontend/public/unsloth-beta-black.png
 create mode 100644 studio/frontend/public/unsloth-beta-white.png
 create mode 100644 studio/frontend/src/components/assistant-ui/code-plugin.ts
 create mode 100644 studio/frontend/src/components/assistant-ui/code-themes.ts
 create mode 100644 studio/frontend/src/components/assistant-ui/code-toggle-icon.tsx
 create mode 100644 studio/frontend/src/hooks/use-collapse-scroll-lock.ts

diff --git a/studio/frontend/public/circle-logo-small.png b/studio/frontend/public/circle-logo-small.png
new file mode 100644
index 0000000000000000000000000000000000000000..8fc411695d90ed1170884241b327bb7d3f116eea
GIT binary patch
literal 14416
zcmWk!RahHq5Dn7e?ry=YxVyV+kpjWBxVsc67OZ%&;_gmyr-9<`P~07E?!zVz`+d9r
zj+`@RW}?+q<<O9ckpTbznu5HvCiEW!z043lLVqdQ;ST@+WCdF(DRnnZIY~+d87W>4
z0bX8KPBsnzKt4K0%S%7`9$$3h-V#Gsa`B=r-5~@(O;d?Df@x5I2?m&jV`7UBMtnoo
z#l^$Qme&~wokNz-`L_L&q19R<=9^)xjuHPE#cD}YfbaR~>GQ&K=Rx+t%aqSFMlcpO
zUDAGm9Y8=cLRfH-B0)`lV6-b37EgW?o)U9^ywCc>AOLUy=i?pdUMuzjQ{xPP1-Q(y
z0kMa--ox6}>GS}UNq|ji>;n{-L`}d5oAwAUV1y7*tJMHP2BZK04nF>p)PQm{z{!n*
z=m)@JS@J3YV6lL-3=t3q1IS>I1Hy`T0KVviXtKce_5prYic_<}_g2FM8#Hsu!3KAI
z0C>jhCt`?y0R(4AvP=Ld`C$Q4<T#UXl*({FKh5-PmYKRlw93N(ddE{I)0Jkpa9dOV
z5PMx+Y;IF+_c_!_nh`K!S|ROHXQyY;;K(1qDNmsS0K|`w-*5b4i3j_;Tl-cA4(B(6
z1#k8j&eWn*mm6=L-(lean5xs~%cgsK^KdT1Ftr{%_GO;|%enxYjYngy6{Mer$bWnv
zV}1Bv#L08>)klQMNRd%cPHD~ayFTs2ioSDgId8@W-a+rat@vI5zGjMNRP`2yy?!-w
z@~qa3KNyFX<Z!w_pm`gP2zrfYglu!y8L{BST<S-+s#B{B`?P5$NmGrcZet$>ab8lS
zy&@%hp#dr*)p}K%5t_ywdIQg3U}MqsXFLJof!D0xiBXaOnAY5!KKB4Xa?fLeI5P}j
zGQDUA0O&j<G%8IYP#!`A0HlA0Fx5&Rp#H|K>_yN1P1e~<0BiP9S{xPFD~u`maW|M0
zIiE01f=Ihpa1p+#7LNEQR^}FF<8Karw1Z7-qmW}4f{$MabNdK8Baz@GMi6n#s06=Z
zna3|uPazUy#KTjY$#a8)iztu2;c8G1j<^=VmPy(r8OzgLvAW_4#2JY{C9aH6a>l&G
zbjoq0gw$)m{e|@qZ!4uBjv4yt265-cFb+;Du!j&gl8Xd47nDGh4n5of0Wt9w!4M8~
ztM5x@M&F}kzC`O2(+%TW?CJC9Ch%C0>GOHwk@V)8og)HaV}o%{5KY1{`y4_O6Xo=k
z^`%N_PiPvk+Y$AU!ooTF=&6%S)f#Diu^$J#%?0Zc^<^vR&Zx+#a%oNI(Q!<Z<$>s<
z;9XT@D%wO&1r!B)`4xHABCR6P8BE)3WxC@;$I<m&#QNY$%Jig`M2ti=@bhT@ZVRs9
z;J0mMqynI3Q3|#qkEPfid~qJ9WX8{Ojmfgyzli1E1oI3P9DjPMvuj0_;c$Co(oSaK
zON|$?7S8?2w6(Uqw_!hWZ%0}SBbQ14lRjgAh;Vdp6n5i<iWZ3$HY{(>Kx9wkM^r-C
zLd2e-{OiMPdd7(Mrv*-u4C@~lKb&>+neLeU@sh@{Qw39XQv*{E%C&Xebj0R-%2Bi{
zbTBGRbOV0L>NKeDlp+1aub3+%)9EPVt)9~K(o9gF)-WyXHhm3!vOuekjF9~qDPO2<
z(K197M}<GcX?dwHIq#|qcg$k{L&f}a{o{$U=Z9w!Pj)|LEh)Eht4gaRtDOC>lQ8H*
zb(<FgOWBH3R8weEeynh;Vyt!Zc2!nYfmM9VzRTUq=BKWw_@~Fq=N>LO;yEliw}Os>
z_Z~>?yzLt85$!7;)Bi+gEslMTkZ#9r(T~<3a<`5*g|`I%&gah|6nvk!5qT4d960zb
zE+!H!4()%~wn8q#w@R_bZQ7FtYziR?h4ig-XHnKWJl+JNR+Y}A=9S#*maF!{X6Ck>
zD@6lEgS*{~lPUI@R_&0+S@(&F5th};#mdo>c!9ZW`&_f`rOuU?xQJqz>_Np?-q_)^
z<uw2C4{0iC_mw|%MRk32Z*><c73PhOS&kj&^&Hem^s-R11hdAQORKV~20XJIH*mLc
z6BuV0Z?&A3IZj?qkWVsCKAsBPncdOeIo?f^O^{-f%?r5;`MbJ%On5rDbRHLNarc)E
zl=UC!S(8}jyX9Q-5d2a~9LG-gOefHCEQu*`%JUsh-6KA0pJ|%CEzIi3TCG}g{JS<G
zG8j<ttiqN>GHTo}axQUmaj(Dqbz&!{B&($FI`M$r#M8vU<hhGnts$E(cm9*($jZRy
z&8~63FM^^XmQPbhaH2z^uz9|DQQg0~sW(KrX#(d1GXfD_a-W5s#$OX2dbWJ7@Gn(A
ze1Z85vj@`^LLAZnXAegKp8;Qt!1E#QL&ryUR8c%hR3tP53fE-kUV<x<M+>f1YnDt1
z-WXyIo*~K8rwKY`@_DTI$Z5n4%s}T~kIF|A&4}MN<s81XQrHq@5IhFpGBR<Nv;EU(
z)N=poe$O&rxohxj-4%bCQ1-n|XvMDC)ziAt&Oy_`;v}mzxV5}B#5?Li<}w}I9S4YX
z8ofKPI3N<e;|6=+kja(3o2i}VZc)9rTF^$Ye>63o>m_W!W02_2AMpb_y<TiDj8ama
zoltGB#7Tt==PEoaLQST-Ag(~8uBBB`7`&gOn|jB;CKMp`WD}_->oLTIq^8O2lSBk=
zqq$bmm7ZU(oL>66#43=IoswOsiDsMD$h`2Wg`q{wikUeXZP#btaf~E6YV3ZO=F#nX
z1<xTu8B*A&S*?zaugqS`?ERVWvv;d}m%QO$t_`cVe~&aMq4@q?iM8sjwGE4{-UgSO
z)IBmC_%yLv1YT?$#+vOQZ`+HRfN`7g`m|T|ZGBChD?6Xnt9Z?@nc`w8=1@KE)UuS(
zai{7V!!`>%vs94Ixn26)XWbwsf>hH=m&(%Whg#Xz{cSx}(>ecJAIIh8b;Xvk(_{JL
zqGPAZdxM@D#V$LExj&H?XjLSaaeE}Dj*9E<`M!3y@so{{W0P<^W;~~#Cz>l)uJU;c
z?|gHDjK_6ft14Q$m*|$~d)B=CuF=ED@PyXSQ%*XYn7#GvyJP}xoo3EL*Aj@$N!ok%
z1J_#RywnDB6a0$(5bk8okL}!@9ad|4dg_rJHF@L6<po{#vVxtLvraR;np_N-(EB9z
ze$#LK^7q?OKg^%%U+tgV$Ovk=MLd@GVV)7}@u*}ibo?oj-*tZYy)cK&giNlx<vsA6
zW5>c{m?uFm_h$}+$mQd1X!1@8s3bQ>RZQe>#;VV~@=J!jv4(M%wbXR37=inla*OHx
z==J(Rg7HPC%MJCt`3h}|Qk6)xh+Lq{>y~`YXXj(f`|(pf;cE9Am17~u`E1aWUGJRR
z!ovcaah^}t`OpbTixb@%X)A5}?StptDS;RAnewW!Z-RHtOZW2YKZGZYe?g6Jb(X!3
zn-5=lzfAWVV4R~PMdiMwyw65$&B#s2&C$rx#3kfVP>`SzcU^&|lDA6gOK_&g0(IYg
zj@>E_HeR=%m8SJy7B5g9W2bwneafy~&nwDm1czycDLSWoL@s~5R86e}oVT90Zwz-)
zWH7vlzo}nCo-Bq-{;1uH9{9h#u-}JYKeNj5PeW@<Bo}!DHvj+y=l=@^keN*g0I*gm
zNK0sYXV2yM1mo^yyo>3~-HQ-#!Sz%93`Has$L*YE;3$ZWDIKLV$&xOl6N7jx0MFX(
zLfbRrIF^8w9hdMj61ZVswBdKv&z`{+k!C*%OU@hqy}oJ~&#iHZC{N3CKSvZC`&ayX
zoOkP-_M=oWzNLN*p@T2~gBi>OZ|GzUhY3p+9#{Ymn;6ke1p_XsQP>YmOiFZnwlt8H
z0E3FOSUr!+BxfTFqLXPAoWu)c#qkhJFy)`t#3_MjN%+KMu#jI!RR&sfA~DkqH5NNl
zIaT8eHL<5+**33-T#ef8rspq~y;rI0#+*frnAn_PPK~5J6g-1mN~J4Y8Suw^WRQU@
z13(b+aY2CkMhuo49`+wMyd<odhI|B#_i%jC(+rnXnafSGx5fC=e8NHi>6*_MAD6))
z`K!&OnzL6W@uzZbc&bDo(uYS_18}#aPRcPWm{UB-62$0L_NOMI^TI7C_8TT5_qE>P
zG!=K_s+%2RI2~q8_=$&ES(!MVA+JwkfP~FtvjD4GIty~Zg@JOZVGro5@MSSJwJQ^q
zk~UMlYl{vz2$Ymc97|nDNk!w)kudVgEFEYJXRL+HBZZg@2vH5C3s;bTOJ8k34xfxE
zZicN)QwGhR5eTJM`n2!dV(?mL8a~q|?}V_&VRDB6M+;fOH|O{=k7+u(wep;t*fF3(
zRSsz^tRJCl^T<{5Ldveli2pK3Lw?yJBJDD<yB6nZ`yD)8_}P=6J-oiOaOUydEAg7b
z2f0ad!(-3r%YZ2wMz2!SNU>soR7xVYGPYu7h~4_k$dlUaee8lg13sA_RY}q$ct94>
zr&|K&8yk<$ADrz^w*p5?#U8S-3XsrxYviwHbqyZ7Wi)F?uwx`-;}N4>U*Tm;dn6KN
z*Y&=c$TLs_OC*>wRcmcw-3D4u7^Lh+Nu8d7oPH3oArqr|1Ain)hU>QD#=nJ#Ul)p%
z$A(lv8^0KsN$V{;Rdj2aBCQUs6#r@J=I3Fy%_xh7T;9&^q_bm|N}9Hl>x~LLN{WGn
zdOl%(Ld3%^B}#Qr!aSM(J>{JDEVWXk;gNynw=uWeAKF6|Ul@t_YLBO_4}5g2U{R}+
zN5)r>oU)COu9FqV2<*dDM>4~ELTVb(w*A#V-+nrdqkDLP#>x?W_@uErg_YSmqaL>F
z8D6N?!^`@#8jA7Ecy97_m0Mbh66q+5*WmGQHG`C6{f7#cEV}ZF5(XLRx*!J6ze9SC
z4hLeeQYh_N#eD%xgL}9c-312_3PrA5WpE-02`h$5lc>Y(pTR~$Jc*eK`^t@8lJFbd
zNvn;#td0By1+T(mv<z4SMB~duHspDa<8%GDOEl@-SBDCjx->V(Ll==VQ&G0GRPqp#
z4a~PQlo(kRv6I@(D*?JDMOyAsHqQSU{!Q^VY?-0*41UG_g7MH7UMfjR!&*0TMXPHt
zS}d;kp>C~-+L`XEk57T&6Z=Q32;QDVK=2nj1vgGf#YyvI)`u$elbBGzves8{86g9{
zT~44wy`*gEV?sf%UTQ>T3}-3v=*|i1>rI;*j&o3bv8g0WNMMPoD^Y@7EHKC>HI*Z#
zNKQR6@6%LB_7g>+D3vP&jCJb%`h?b#jKze+qa-=`V|ghpK3TgVd%J2YjcUi%<$-Xo
zY9C$?kZ#`3G)gM(UeD|OBW{ak!f`cRrB2h#D+$ZaMCb{~^WoM-$hDY#Gw5jw&DC-6
z@8XHf`K?}B#9?|QqWG+I7>hzhRlV?^QZ;v4>T}Y#Q#Ob)3|6GV*xct48)6O~6*)l(
zV3s5-tS$l`F9Dt|0yd_`tnv-U?iCD8hNi%e;5o=~1H0|oI8TL%Y%)+7-*01-A-tT*
z$5=md4lCn!?B>w<VEQ!3Oz^3MDpW$(z{DbH*SAdiMqBVZtNY)z*YLWvwD{TF_?_m(
z(iT7d22aeeoWRHqCc?cktDpT!5tGp@3g5DW7$Vdeq%ACH>h!-qJ^6mbAV=*j=ArSv
z`df3|e~aQ`-~gbVv^P)(Npq`us^P37IA@1Z<GfDubz?H(7e^5=MP&a$P{ht-1RJf)
z3EyUV6{tV>9Ax4%A#CS>c;d7I;2x$nXIssC<s;jeF80H9pF*w*guO2)ulFY{-d-L_
zg}oCC<r1(Hh9`HD=~T>*W{SPLOsXq|GV(fJ2L7U|4#sSvBvOxm17)7XOL<fL;E)^8
zDqGB3H~6!$%~GsMFdR&+^dci$&f^KSb59Mw-B?sZI~}D2e%!Q$1gGTA=i9=zHZ#&>
zCSqUz0ubkD2pocqg9DHdb=zIg*1=)-@gUFo_TSvF{{@fze68aO@09ShvO`RS3#TL#
z(I9^5Fe5XFyXk#k1e-dQs-ptyJdAI0VOi3eexKcRDX*ZYLa#^jZ4}%nAIgk;9}SjH
zLAOAYtCHuO97U`;;@_wg$;kW8pRCmh*S_X_f40VAP#+$FLQwB@ZenO?h#2*ptA`8^
zJCVhg5X;s+vU)GEBAUC}5hJ6L1yg#heL&sDo*1gfv&%#Q9$vzJi5gZ@s##f^S)eX*
zuJUPGb+d-yii@khMH;aeH(O`S!(R{wk08f6Zt2RlGe6ETrn@-J{VB*g+33kD$qQWO
z78!=q`lW34JM@FLo}QObGwo_0D+zvWvn{i#8eU*WR^_j|cPj%t%Mn21-No{qJw<Mj
z$IU4mQ+lFpuRs4sEEzKg(Q}|{6DStw#*Me5PCei#MSpQIda$EixsRq>V=^g)Au=#i
ztn_>X8zAA~!J}syz$5lJs>Z*Uq7@V0wb&ZnF+WY8f~c`qGv2gu!w!o>gm?VItVneA
zB5Jd$;jNB+Ax&!J%NOQNR>$1-C#1<6g2;?|MP$E%Q+_<=>F-_HO7^lpBKUZDUGLYt
zuJ^N?@AleyKK)Y78N*Q%){vfnXn#MlH`(j|#p)dZ=fZ@)wsDIf`Gt(@`$$lyr`FBF
zwF%pOy{ly92I9rP)JFexMYUDD-4xe5J)9pr-a3f5pk_J4GB<B;ud7KJ(91qwkJ-iM
zZ&N`cKrBA1`NR*7)vv>3Rx6W$jx3C+uGEfUBSqkhB7d^vnH26a-qI^;t8>m%rQ^xP
z8G1d!h!t7>O&$f}Kx1_qg^IiJ6@01WIPEPr3wYv5ovY;W_~9`7GvW3P9A>mM|MN*h
z+w)n+^K1PuO3=#=#pR<vi)r`j>iuz@*5f46Lh^SHOhx)9IVf3De@)y&p6*$F>2PJ&
zlHCY&R2zns{A3!8qR5x$X`fEHL1&hX#~b;Xzs8sFntm-<W5b{)P{d;ByV*u}KtRxF
z57x{P_q5o%@RGy)_W?Q`?oShZ!9LG=g^ao<iemr%)Q2O=)a8o0Hhu&0g67ZH71)!S
zDq9FrjK~SpX^|(i*17i@C1;T*lHaoMi04U(Os^NNhlpk;i^hb`JNq9lccKT)1A^Mv
zVG=~27#2)}y4qg5s7!b6t(EDqjwy?+l$xK@;ci_>&OTQ^6mym?%YJoJ=VK^Tv>yRs
z_Ke$0e^oLzk-Bkj-je%nW9M8M<{Op%A=zqkOh(hW>lR<n*h8ns<ZyhT*>Od`#fgl|
zY69@ZTn0`f|H^FE%1G6D10@CctwdcGwUVyTvnbu!WRRm#nwP60QelQ?o8E3Jx`HrJ
zf-9mLIYd#F%heE!M=XgbIP<k$a_7FpSRr*1sn_)Fir|+@);v2*0G@uU3)O6i8edBu
zsZi?p$GAe))lh2Y0Cx2x4UERkY|p!{w4|_AQYodB<n~5F<R1*18GWCc?Hz?`upa5O
z3L`scT~kbRU6J$_;$Xi(?Qi706#?SR=<XTO`&RY21uPG2y_r#hY3;lNnuvDB?Mo9T
z%t7vn6`8^~D|v&<BQ?i)t9~M|9^qTsO0Mb=4wW0v$I&W@wvQpB$$*#eZTB*tATKV9
zL;z5OLHFBj4TWCbmJ5GrRn_V1z)rA1gB6EYb#7->U5C$EqjP8`Rh4Z3y>yaB)^e?>
z;H7SH{*WYwA+>$x&o4xk)ZJAmJ;4C(`j3piXci{5dA-&l+9`$|S@gq3R``Jn#$I*S
z)472$g@h(G=_F3AAyZSTPHlT>JDOGkt*6Z^{I1gH?|wS3Hr;cd(xz<e6KvltB&4_X
zKh9<EKHd579e8@oVk?=cXv_PnA+i`JhsY9;77{hpCTC{O#zWCMm&h-|V8Q$KcHyz+
z{i#M&ta0Pq`f7I+l}s=tGV<fUfBy)DJXxSDAvspOvJr5<RzF6}2|}FxrLyDI^Gvqn
z&^$d2>zi|Qx8iKQ)fWMnAinkxwOdiMri&gnKh|2MRp2zV#~dES=_YwT_j;F$9q6!)
zb{~y-pVI@tbZ`$u8xp)jy4X{!=A{6{IA*dw<py7;SmnXeoEe~-LkJW_drqRd9{01_
zj;lLj;7yTLR8*j6!_}Fj2XvXRmubhix3?0E*w=IoH7%HoyG@G>=Xk9f+%=^1oh&t|
zOumc`MF=<Sz}Yc>Oq%;^qL75V5Z>#n3HLki?QVPEGxPiMt&ssibtF?GiuOQs#BH?0
zS<4^kBRZvp@0Gs_?1<|eGND{Dj=O7-Afn4lghQ_iT4{Fdyy!#8_Sw%E2Eu)TlBx`7
zHSqbg^Kyt}ZG@1|iFCoFBI-u$36k(IoA;(Ye|-}^OcAuIT=(L=i(B5w6FxJ1=AyY6
ztoQd?tin8x`{}s2I!Prc<cHl;9ITke3G>zQk`aB_5a`G3OOp|w3o79EmqhElfac2d
z&g&^5gAPwtD1EaX`v>Ew^8Ypp+6+N>=%;v7mre|!cs-PpS!r4KWzGwHvK__s7g%ht
zPOuG)>Ztf3dZ-PKrgV+QC>6p2t4haXRndHRd{NLF3G`~Qx73+c=b0=(=Ag-ool!=}
zW&Kte6TgRuWDQ>Bx2ZJP+({5q{X}?j`+>s$;&**C^S~&5-o(@G@lvxRNzcn3J50MR
zLNA`45yWvL90|w9(Q$8x#QEep045iD4!2joFg3Y4K?RTK)|iQ8jh6Gi4E%TtG7g){
zoo%E8_pYcUK>=gDg5TU)nIjp@5Y|jhw_$F|Ocky33s;XWlLN%4{+H?Z7quF8PiRJ5
zE?OjHtVkl(FIBFOwxMPRJ{`#z^0@Af6zmX~`g2G{q8?p7rJMLh^KJO!l*}5;zoMQj
zH^mOd62Q60?tPi65DjtRgvGU!$we&F*Kw7jEi(FC2jveZ)40lVVy5lYK6WwZA>{cc
zQpUCbNyN<U93ypsaI<RUYnjiV{U}acuSV&wvo>BLucyWGwA4gW&H5uDfBy9Tln5i_
zvBTxEo+@f<d#NLWI?1Xp9BKb9g;C45YyV$G4IOy`!Cv*4gqhZFp4&pIEa{}8?6e!n
z@H#Cp=5K~(t;ifIgV^LWS6!$l9R1frVmrC`$h^Dr9CQiEQejl~LNm9l%-r@Tj&37B
z2K%{wXIp0-=T=MgmJVy}sbL5hGcz-NNi=ecZEo~xqK`m01oYj@VG8!o>aupS8n|bz
zJMr~3SzOlW&YhRCO9wqr5OjBU2e^3dVLI?)5UGsEKI=;-E&#FA%!#x`rZ_>py?xh~
zqn1)K%G}|3tTW%YPb){P1B4nBBE2hG>(7AV&M2nM4hA*sMYe`ZzmD?WAM+j`r-Mx9
z+n`3+eS1E)K3i?$=H^cEhzqf=Zi}3&(7Pwzf{`xNn6ET|>wSok$Tkgt{jFUYuA4&F
zCruZtxmDKmoshr4r6pSxC|6O>UH%EoIhyai;(#Yj52{@o7mvdD)5GmX_*v{roqp^6
zx}VSEUV7Wh)%fIVg76h037^x+t#4HS@0!3{=I-07x7+<ZZd>f2x9hxdL9bCT?fw&?
zGfycaaMr=2xZnq(!Wf3{V8Is>x4y;D{*mZHG7D~*nWo-v<s(_;eYz4fG?mC9BS50P
zDh`yV-DI^VhEQ`B<MtG2VV6n7Wpx3y{?!h}`=apO({W8ue*O61B;UpaZMxCSv5%1F
z$)^RioBci;4O~5#K|?ca4R7v3bCRV`buKjyWmOI%MKh(DBPGjdDu0k)<RsGcD8+RI
zCQT8<sgu48#^AkThQvcLFCH3XJvOFPRI;FJ83@$B&*weU*#d447yTGYcbC355IH~A
z|Ep=Mnvu)5kVOZlddWimsj)&5tRaZ@$E5>eUGKkOoreB!9f+f!1Ut6d1exq`%l2FA
z2xq^mPsLh?*X;bfZlt`T#b$4(X_kl|Z!7`(klmyg7L(59F-qUor<OHOlA5kNLr7rk
zV4sej@gPltu=$;62Gp<zxqfUEK`&nu_L^E+8f<6jrUZ{!J}$03uzJo&eo2a#fyc%E
zNC~!0Q8W$HJi0C;ds?7^dX>Av?gN@?q|eM1O+0FHDh*AFf@`QV;6NJ<?#p-@JFjpb
zBJGCv4w{w=dr$Vcz^snnQ)qKDud2{#qZ252UHAeDE}yI>2(2c5NGQdwL!fZXe$uom
z6@?CiSu61M^GWFkfhzDmQdy$~hBs5HT~KPBk}ein=AS>@%GBD)h{%zFc%AZezYXxh
z%0`+H@$VK7G0@3{M{3RbTW{w-E0OnHjWKte9v>cVmX+k1-<>QCPwcMy9OPJ;Odh$r
zu6Zor%^9F}KK-lsJ2zM7eQEC0{kZQmth7e)bLL;s`Waodl`jzXaDSH;phLOjfCrv+
z!ZwG5mojpz8E)yvvq}H>P+{<KNo2qKbfx9~3ASsiHw2FMe#3Q~RVKg{M)d8$VoL1Y
zpR3&Oq;6Qh&isRihlk98<3{(I!xv&ME7b9Hc6v-iz|9YsZfNNL?}!lb;Hjx8^1PN>
z$fUK$&!hgQ3FLy1YTXKE5ioF6{vpoM057xQk8m{Gei{AWjtNd1XJ@gY3?Fz{7$zpB
zF+qYG_$?@5cAWP7&$E&z0nc}rf21%xgq+tqefFOkrbX?+W+@`JGezKg;|8lqj?v)l
zmyBUMDT;h{Dqq(VToP8K05$>;k^=Z-$Oj)SpAdcJ;S^~fYj6lj^>r^CvrDuy1WH~B
zA5ifB%*;so`3c3v#r+Rkug~|htE&@`Lb)W%jNf4h*U*Ie5sgFd$t;Da!Rcc#mOw<I
zsH1~to$K?<=e6U!TZsjq1{XQ?cC}cY&_;X7=;u-RE+Yi8_65&DVYhw|oFNAJjEW{X
z=8dU>sLrCjYy=qC`C0(6F<`%hA`&#%WgQ(^o0^)kUB?)WJeMrHO!Xf)!2?iHIvvZ7
zk1?1f?Bm+vhK4i0zCJaU%G~y_9h)KGMyDontDSddfE)up55S(bk^m>Q%hzW4aBfhU
z`lQ^#S^3qngpW+dX_0eN)s7)g8L4$cMLrOl@pxIE#cKJ+^%5GR7b7$Y?ko3k&i?(Y
zxf~X<8g{F_H%GIR?jz8mWfWS#_!*4=M7X|ex=@E3{b{Hx@)@f-lmf&7iz=Ou$-Dsp
z^F^XOy!MA8IAm?;E;%_lRw-`V(e#=+Iv$ndkN+c4`pRXj=}Tzd`^&-F483#DD{r)P
zC^RoFXO!4eJ3SX`&8!N(9!%xPi0;Mqoz9b_hhv5}Fjj)AIfkkG2nm^$z}-1{*Mqy_
zAGdE{U}NY$)aqZE4#n}UFILR1*P6jiLzVi{J=d}b@-Y_l{i_}e#+^_UhGIM}?FX79
zc@6lR=Z*L7X{$O?LJAWlCRuyIcr6X+_;_9TNHRU{Hal{p%|j*_DBahp`f)6`N(q()
zEXkexlQF-9OjYm*s4DjA&er{4&enCKDf<3$Z8H8i$R4QS={Gzbc*iBKW=oj)$?ME0
z@Ofp!b(qWxcTQq?Wd*9+tPXhLpt!Z4VV(PX=hCI~3y2{8mqj?ILZv{@>Y6YSp3G>r
zwgj&NI5DX{IHNxaYg@eK%Ohc}JT^$H(~Em!!)~qJ0}Ac9QT-aanKY#6*BI{?7?x}2
zl?_wo`^o{pH7K%wUpeJH#$aV-tu$;x{-ZJ@aB#?45OKHGD{{XoWpB@X-g%9VA^JeM
zGZe>s{8o&K2tT1czOIjulL6#4II_xBuKF4LeKn8>Y0xeN)NUM0r(B$<ks@W~Sc{6-
zdtAPG)Rk8Yl#4Or9J5vXR0|r#KI^()ub;}{vBzicdErDuLxV>0|0TJoku#yt{i+)@
zV>Ak7>nw)RBhksmDr(Zm@%CN#e+YUoC5YS+Ldg&WRX;B`C1RGn!S%zlfKP!h-u|Lp
zOJ<I>NcBVkp=8n2)HedOC{2Ao0(A2H2F)7^JY>U`=KZ@Q=0mpvz_TV#QzE?<bx>{F
zeL0G<+PuYDV9*5y@wdBP9-F>rLb<4?(#?9}$M1Zps6!}P{>hg*S?frLifW9k$w*At
zin+0g46db~pr9pK(XDVC?|)?_Hnz4LDi#9JQYocv{Re8uAhH1Kp<V9xQpHGfH(CV-
zRUtdSLvdc*xWudh3DufKU;Sz%_nMu10MdvJ(;PZG*xc-H{(f=ZhwzWilG8a!o0{;Y
zbl!3kN?0U)u6JK00&TK_)nw*ei66-~Y3fcWo?GTQ@AgqI;T}uLmDUP)oF)Q6yJ!21
zeAfEwQHN72+qE=(UoCdQsFa^&F_wTcL=e4(REV}ezL0mh{GPIoqoZSszovlqY16|0
z^_}1S+39stOu?pSN7CL~+=taBdjjZ-%?+x+AOEL6fNnDdNa=wgl{eksp20R3xp5FE
z2iB<6#a_Xv9LKAG9jPB~QWKh~3lZ&36wd<yc>0sHyUqD+{`?e8wfF)}&hMlGVAI!s
zdav!>V%`TiUQkVuo#DxjfJ3MJiR|eh8WVByFKlsm9s^LGqe)g$7q>_|b1_Zivzv@P
z!1Vz_keGdk23|9~f>D%!bz27hW$`N@+pl|ida561j0k!v2{Pa84GE}*+<zDafwP?2
zqnT?u!P-o7CH5$V{j6K~prz0)F~vuuWJD>gKPqYFEd?J75t5(LNp9i0*3?AYQ%=_V
z?SZBnr1hVa5|>EF78tu3r~s3blgfFbIb&mrZ?~rJbEN1OGLZTCA*g1InmuB%1017#
zAF7Xc4=t{GF7@!9SZ-mv$+jBtIsNpG>+rdY9Gu7JtW0Vn3i~5o5C0PPs7WYooZOnd
zAFD$O2bP7GIQencyezOC47?qmLI{%}4}38DcYOS~ywE)ahaueie*=u0*nT=!MEG$x
zIooYY;P2nRH;n1D<D7GIbNg-yBz#kqJ)H~f>0OtUc=&!At;__~T8FzFTJ+ePr5E7E
z%Qg;+ct#%-anI|6&(cg?a5BU?)ou$U7Vw^fL;0W|@$T7*S+P(Gt66DF5PRdt5WWEX
zA9+h=PI?ZDwIOIEyr1mnS^PyQpaLYnxj7BWWm;NVVbcdRS(DhPA8qaI3d+i&p;!!o
zo~-6MWW76ZD}o@C?9Qom>*Mv}*!(MqiwFdhNq5B>0Uoi3r_NKJXlhZauU4SWz{;On
z`EGDH;U1<@Sa>xU^q8@?8yPffIPOd{;-OK@O1^!{xUom#K@aDv2QlvO+>1$D><Q{w
zTJRi8VH`f;#6^V1l=RPUj@gNbq6qu}RrXFDr@X@C7=Cvr|DUxVf|<@1lq*jcv0y&?
zYWsWlk5vFXW1eVIoCncomEhgCeZ1oyk!#p|JI@>cf-@KZX6Vfr>-nfji@>#2TOuve
zU({oi7kM+zX{!c24?9uaF%Yzg03Y^0I>$1Zn&LrQJ44XAICOk(w`E$iBEb>&qGXME
z$<v#lOdGuK64f2PY>V9dPsy5jlFYGDRxtY4b`h8DU+s@c6Vx-dEGG2%JfB4|%hPDZ
zx}rgEa9>0$3Gt>kdK9E_EM(f9`+Jr(Ii@Za8umje08b8RJHrVGH_OjZ50q6_!f#$|
z^?xVi2nE%9*Nz`T3ZVkJ%3x)q&O$D4D2~+G$n&U}CGcDXRVY!JPQZN=Gf6shxP_f2
zXY|vmPBe5RWx%SB`@tX$0{Rw%`qiM2HEYwQG^aL=&N6v)5yYa3fCqQH=rPRBj#j<;
z0Mjm|w&w(#uwt)j!dW4rA5ViwmBgr5YL^s{Z`BDduKE1<`X(s*{88Q1bQ&rc|8I5O
zZv+KtSb!IfEuZgB9o9RSR`sAaGkNS|mOCpdDvk~hKRrY4BB`VzH{Autow`oYFvPCT
zQAN$*&x2aC9XswXm-fj{#8&&>=6UG^jvr+AqQBpxfy}>PT5F^3Pn1(-8+*YX%*X8(
z{4UHmp-`6k961(dMjT4)VJ2ZbG*tUU)fjNIB;;q+bD1apA+hL7`xteZP)UyQY0x&p
zbG;u%x`<^6pb|Wh<elhN9(i@?Q|y;}?5W#QaXeqeHH!hQrTp)ftgRwkRcez$(Am*c
z#s2ToP>6Oy7u4X`(qj+uf^v^*dUA(`WFs*s)@wK}!9cE;qGSQbKHwC;8z-7nlxxu*
zgKq4H^x5cfv*9E6O;>&wp_4@iT?bwg^eD2ZwRo8%2=fBATtaib&n#UgI+-L$+lZIm
zBq3oc9$mNc<zet%O#kKBMvzGQp3Ka>B=<w3IgdK1eE)UDZEO=p?60BV*bH`moe(ZF
zbg*P(D3*Eu%fDvDhc8etz2cv&n^f3(%LA0Xpxq4wqpyQ_fBzl^gfB#(b8W2|2=l%X
zlM5>(+}E`26Yo#wnO?2`-v@%qZW+;kP~peE#8^#_N+t@vW_uGX9L^m|%<sQ=h68aj
zQs=8W=%P5K<8;hLt~m`lwd^#mxU^-rlA>4YGD{+A9&wksr%5wSLI)&^?Ur1Qj#aHj
z<eeeN%b{eIkD$k_inM@>o>Bc5tsA5O&`&aBDb5riyvv&VDB?!{#!AN8T4{=XGrX^(
zb`owTjU3S#S+7@Q{DHPJ+Df5g=XztRipmP1rbx;h!Ar29oY#j{LeV}Mis@f8ivxUe
zxs#u**WeGOewL`_ET6tqRJX@KYjgLNjoC*4zRVUx?hOqTv9Mt{uzSr0dMRosEiDCN
z#U?70)G}~-W|2i1zdz-)nYPSE51;lqHBc!es#<@e#SSmzz#iqU^A4T9Keb$_HimXc
z%kAutA_eDknA5sC0w>27a3G5%!FXNe5F}|xfE<%tq#+nHcr2abDeN}ItfeAqW$cdD
zYPOnXHt9Q9Rp!Q?r*^Rm)fIss)7C7wf4pYccoZ#6^}|oD<lTs&IN?@H4h|lroLM#M
zEb}>^7Ot_+VVI#?{ZRyQ|GKtgnDrZNlG<Jscj)X+$=S+@RJn;Qp*8T@Q_LuKr=T^u
zR!w*Op8Y&z84^vcP=w6^LYlqdKk>{)5BMrT_8{!E%0`bL`$0!1njY*PNfX|uVIB>?
zR4L1kZFXNx)<`N4x<Wvit62LsJ4X@F>^fL+DeukNzS9iIC&fadG_D0f+zbjzGPFGf
zf2p)-*`q1!VVh?cBZU~(ZP#~vHC%fjCHLnQD*sdJ;(aJT_`~gjYoBbb-(F?K8+0Hy
z@#@b?-jQV(g@I%fhfr4A(OPQ74+d`CKdU%;KWs(C3M#Vj*M}Xaz0u!x8?q67!p@qM
zb#rT20^4jt=ySJHqP4u=iUr_>gud#7fcenb+lB8M^qc%m_+V<%$B@OE=2EqF;9Zu<
zpiUB$GRyqDV&i06*whs^<&t7Ut)QiaBjn{kPpp(EU-IaFI%XMU@acTPui5F*jV{Ly
zPWZ690OwnNRw;LT#@J7LS=$nmZ({`H{(l9({uQh6gHF(&QTZb#pEc8SbQO5dOXceY
z!aNiww_5}+3{Kld9d6Ek<ST=bV{<$RUJ_y2*eb@;SZOd3ae2C?9GAy#bl5}E_Ua?G
z41^y$g4eJA91`-`jC=}onSRZ7nPEUDUzF3K44yR`c52yj=XVjf9Xxg4^ys*J?p!?e
zE!1Plr6fd#5XEE04Mo#p58x=wWU!QJXWrd;hW@6M`<zf~eGorKB4*4(zr6Ga-DzA5
zM)a9YnR+FD>*LlU<f6wPtwX6a71Air7IBNhkIw^ckCLk_)P4Izsf4Zg`zC(qFR3vK
z2b&f!V#tPMfiHU~n@f?;<aM^0_Qh^RlgSv*y9(V9n+iYYmW++`C*SE5blCVF_~=pH
z?)lNFC01RI1PTfz{rPfAst8UcI4$GR)5lbsZWb*|Bjv<N@+tA^`t;OkG}O7kB;};h
z^3rle22!hy^|h15?#Bk!yec1o<)t>{HH{UBiq`kegu`;I2<aa_%h(0Xk@ww43%+M>
zclH@7eiDqTsGg^!j2;`d<Q&u-Wt&qGT@Q+><?@_yb?8Hn$F0u00@=m_geB&CHIC1Z
zd)1Z>M-JvcJCLIaDB(#_TigtP#_JHQPrX$=>t8=o)x(-^sj}36U#XIlWh=o+SczWQ
z>pQ>2yzu!d=upIEAeqEK_8W6cKE;Mbl{=R*S3XgmPN8^4zCN@_f3d@hk6F~H-EOv|
z!bE`{o5T&axiG3W(txw^_XU<QPr<iu^@uQ~{+Fh2U$R_gPM5Ne{-IzEw;QNQ|B$AF
z3Xpi9Bya`^wi<x1@QW3z&(=fz{35fk6IX{O-=<MIbd`i+>Owc;C%Cx!7jlP+*A&#i
zEi2K(M+4{AnnV{n?#rW~dT0*2!%Kx1a1_)L*6D|Ob1G0L0m;=UnUUE{gK6Z@uNUsu
zO*;OC6}!zu4;WraOON81h1Ug87o>bw?DHK{m)jc`TcI`|#VX)3*A4XLR5LC+^ljy%
zMs7h#c4O4Z%UZLnql78{6@tmY)TpS?$&q@Tpsh)uRjb>I^w0?fmQ^+pquX|mxKdR3
z3-9dGa#;q_eAYz;hU|rJ(o<?cY@|d`X9&}Bm?Mi{&}HVH(uOhS>4{MYn+v@Ug&Dt{
z1*2u=7_qOBF5+JM5tuP#*x58`=__J=`TAN3YSp0u&6S)ImPWbDju_iVaYP^tn9hb)
zLPh@~XWvdW<yHCx8Oo&z;}OHB1IG}W(()KUU><=17*Wom(E}bSvQN6;Yr!S|rFW<_
zr;Kp^!J=>ovS7cNZqUo`rv&_-SG(HVLq+%B%aA52NiXWBO042y&5th_LFeT6UhF}1
z9RWc|L02tQ1&h<W%LMRrM2<kBzl;y-NictxItO0xg;nXv(@|=MBNldk&72pOa|ZWY
z(cr7i%wrYNq+H`%!q<6A@~K3_OX9S?;z^rh%?GsG9h+`sylbv(^-~n0{Ea@NX}1gC
zs_enUDh4eumz(<9G{07AIVm`c1%ovBbN9=3zZR;5gtBGAf^>FDQ$YGctG-2P!Q(13
z&VvW8TVFdaeU0G#P3)MMFa1P7b1aTzIsbpSoP~4&hxykv?m-+RZk>LdQecu|o$sDp
zRyteuy{4SfxShL%&>w|Q-wug+$Ox!?!IK>M+K;7yuN&)ncFb&JXZBM`SC*Jbv2*Yv
z<v(>4Ir$k}LEa+|4iBqsV>!brG7)*1V*t;r92f2}NV8{XK4Oy1cW8<TCsC5CCF<}P
ziHb#oz$(WJXOElg&eOK59nS&o5+uLMN-{k@=D=G#N=n}V3r)<3OrQ%0dj}P?B|G;2
ze&Z+O@|*0J{gmpd%Oy!+w+}^6Dq<?S3fmVfH#V6G3`51h6UE@sNritAtuI%!_d{?b
zlaNf3tooL$VEy$Ou85OpKB(Ppdd6$rzs%?T&GL|Vlez}WIGVwIUKZ`S6%(R^9uR8z
zApRVz7gVQs$FDYh;wV{fI5rcb>4y^isrj1b`_(Q(`OjqBW}8e3@@k83G~&z7{du!7
zW5Ibab2|KFo_!EPxfMag$iDDGUwFLOL4^{o)&Q7##QCa6(8l&-P1mBo%lq(Q8JhLk
zg3ElQEtC7a)l~G3vNWU5Em9iJzZ|%?```Dha250=UgX>Xie;e*GSYn4Kq6Vinmffq
zbYZG`U@}s7P-`ZVB3iVfG1rR;##}0%w%$j8&FNuL7mYkWHVtVkR<!JBQA4YL;7s|L
zb%ox^;|05OkI$F;sh)y^^wk^Xfej;?_JWOYfW94b(Lb>%{}o=(hrbi50h9p?<QJCD
zNLo@??;OP`D9$`8J2t#SJaR8asbunvW04CxL+96@e%536Oe=Aw&NNYz&XalZD>CH_
z?Oi5~fF9iEu0$wUeeNgWIy^fQRKpS0fDw$mD#kgt>5|fXk2f8FlAO$q*PEUk0=0)g
zPc<6qjE@FqveH|+NqOSu4boItL)7j)kjlXyf2<Df9x(qR)1%iUBljd^jv;5GM;X*0
zMihqAXn;p?dDFABdHa8%9cQMo`0u1ndpUd#uSSZHzQN}q8c&a8SN;~#`=0d{*4wr4
zswy%QKfjD*{82)3osaN220@${Bro%esj}Ebu6}h^AY9P-L*}$<@rIm-3_wN$>OXj)
zTVEOBq~5bD@R-$7ZQ$vajPq20?R)&G5@b>%-<FfRZ-g(EdRqzdt(xB{9MH-}9%Oin
zBj)@0u9X<Ngh+#wFQ;fVI3*wDt`<jFG;)|30?^U2Rg-CJD$KoRy_IIJ)PC~Ueh9G1
zl(lA%wB}b9&K>yYo`h9vHFp&jyFZ!H5VR1V5RC3_odZ;pHk&y=k3@tUPQ?Ti(@`$G
zTv6nCnFj8AHN;ifkYNF3&i3h*x3oeDZ=Qy*c!mZNlY^8o=yzInZv8HNgss9fSaN&M
zT$AxGmfD+6QB3{zQcsV?KEuP0X8l5Jw{k}_?ZKviki-j7ee9f_>iYT5wEMcCsoGWD
zYQs$4G7)4zkrjs}fp>qA{Ch%caE@D1c(Mz!h5A+J=i{k&wkmYb17bs3G90VUQfbn(
zeVA3#&F^*Gu9gnVdiaYmZk^7|M!L*MYhn3z<HMU(IDi4q#D1=8K}7%SjwQw@Zv~Ir
zCL(Y~KE~(|bGFkml_gImmqs<x!iA$PgGNv2q!n$}=(elhZ@Zo3R0;~8<0d`)>Sxd(
z^-rXW0+h8s->RhvsZKT`Z2utzX-T5;562N|f1O=8$MC;C5vmkTpmD4Hz?vqVwg-zg
zN{=SB1Pp22d@ok*J9bZMQiNw>N~+}5tub8Ze7_ZNu6jHS^9t@ilm;=(e?xs-Cw-^?
z_UWP+)4W%}?lAb@d60$GW<hkzGQzxe=_%FUFVgSMwX_Ntu-XzzbYmV#hTT*=iB;EI
zcTbCrX2!!6?85K=#(LNSP`bpDF^o4Oq}}5(zh~ZVh+uu})|N!ow`Vfq;WOdw+OFpc
z+7$IU1jddlG*=}%$o-gd&;Hjsz)p%&R$gVXQ_cFj4f?Eu`;9<*QQyDX5o{NIrDfU=
zPhLkfzWsWV-+bH*hJ?P}9KE@zcC=oFBF$#gql@xwCq6iVC<~h-6NIm~-KGN`(!KCe
zbnM|i0+2;Y_}lk0zi+&2Jt!hEhJ!R}6G%8bA9kZ>hk$EmgKj3T8!p_gh=eigbQ<#F
zjMlF~nl$a(=Mk7Y0>~c>VBr6JaQ43>E`H{D&7<$@eMeZq?^rd?AWw#Fa|0A)RHdsW
Hzl8n=HTqeY

literal 0
HcmV?d00001

diff --git a/studio/frontend/public/fonts/FiraCode-VariableFont_wght.ttf b/studio/frontend/public/fonts/FiraCode-VariableFont_wght.ttf
new file mode 100644
index 0000000000000000000000000000000000000000..d7077f1d6449cf260ee1d67c358f24c126aae7f9
GIT binary patch
literal 264848
zcmc${2Y41m*!TON+43X-LP9E`Lx@Q4y_Y2P7OM2nONi8ffT4GYsPx`JM4CtkY0?B#
z6h*{_sEDY5fC|a^-@DHPi1@yq?_A$?lIyqklzZyV%+Aj2J_`jPj0Y3V;_B9m{i4%l
zT8t@xekZP3)0PRD?oI$2w*_)Fifh@bcJloLbAg9NftIJ@Ti0rFX~b(DAnFh!Ip3J}
z*`23uV%Q+YTDNJ-5~ZgVo&N>x3AA@<nNTZXa*KhL7=I}3L%I&@I%sI;6sD)P&>oXG
zplkB0w<pX4`mP3QUFu2yDHuk7y+gwH>N+@?l5wl<lzzSXkLWo$)5|5GUsw&b`9?45
zS3m#fA*g2ep#nDc?$Nb-w%=p^qW-7UFW;LEMrCsk?K5bP?mb{g%Am9TjsPXofSkws
z4@~S@X2<EbfXvN+z@!0PQ<6Qa^?2&{r+(g~t^<0!wfjbHz^+LGlLrnS(x+PKcY#f*
zK(h_WgL)+Qwr5PFeh2D%cBzbsTg;iXLS<wiD<wSTwT&)FL4eqv_%nD0#DCwzQGkD!
zTSe|}@o)B&4m`!YE()GQIOGHmLf{*b2!A6l^0g|7;&0SMO>vjlEsXE`ZRY1XXRZ1>
z49|IS{@lOVaFJiM+I-+vkq43f&HTI2YD0@F|J+>p!9a9ka_=GHGI+?Kf#Q|YyO(2k
zbTT`~a2^@LNmF8=*^?Qjxmy)w>M-w^bLcw9mHG{o*28!<Gx_<`Q^WiU`j>3#kH0kX
zMoJ2G66)4wv9e@jxb@&S#Q32Lu<Xl=j4R6OM3uaZSMnId@Cu>KdBE^qy@xP82znp=
z=%J|NxcMpOOY!_Fe`WA^m;@-3$2%mGl8mr*tIW!FL#r$j9DW2&{A?P~P8t$*W<;V8
z%Cg9f91W%<>9D>STxm(k$%BN3G9*2(o+&AM&6cF;HQQc&1KH>%&S%IimajfF3X#Gn
z!ecRX=CLal@VE%?@VFcA^LP%Ic)W}sc)YFxnV}$+jmJ<G%44_+XTG9THJWRvI3DY(
zhCDV_ZFp>_I`P<5b>%TpCGnW7hVVE{4dXFIrSLdXjpcE?V!CRPTF>J~^#+e`s$D$p
zQOtunr@rU$M|GXY8|o&Hzp3ANysg=)waqrKw4Y|3>Hr<gV~Bo^$LIC)JQmRvd91AK
z@))brS!q(4%FxDm9;X;nd7Np?gfgd_(|KHLuH|vP$()!UnIG}^yZJkhf0%zj*}*pB
zw_|L^Z$ED{e!H?=na8SjRUWI`HF&IL*XFUV&DyZ*+s%1wX}9FDi=D`0PrE;lNj7W4
ze$gJp<7N9YkKfw2Vc2)=yHK7GPbhm-q$iT*C{JD<qdn0)#&}|Q78K8&v>ItG8k!-0
zW4{RTgg9e>Vc)WEGrY8ry^dy2nu5R_i(C4S87d<vBRnHlMv0748Rar6WK_<mno&KY
zRz{tSdKs}94Ko^LG|6b5(LSR?M(2zk8ND+GXAH|o$ry#Xn2)8nj-OSWYM@%GwyGx@
zGj7J3I(GbwIj=8VvV7Cl9lQ7KfA7Sp^PhZn`O4Mrf4p(?ms@v!zjy!7hlcH+DRW@f
z;A|nG5!s`1<jj>jPrm#G3l}L?vRwI!RjSvDZIaNYW9P)4z1@t~WFL^fJZdCQm@@ro
z-emmXH7n#cPI=85fyO?s*+6Myn%8VTp<^M+NcOth$Y~^a%^vjDH@s#)^w+6gv;Pz3
zOvtZ0dEGNVVa}rRs_I_%K$KMhUULw8+I_D%m?QB`uQ@x<!WGxdD$9%yy=L`Thoi}(
zvaq3M#&NHXN%J1B*<w5!y=I#u$x5%;!+7R<&3=!|sQi(aR|KaHnaA8{?=@$lxvJO9
zav{oV&O#Fqi#`~{@t_U{qC0wUe5g%Rf13a4(~|xJsQ>R8jj7R{8bhh~&+skiJA}ST
zOz+vgy%@4TeFvchb1{IXq(`Yeo%)k$JsEQwrqTz!(3@#Jol2vB_~d<5a?u}xWK>28
z{treXbJ2(PAv6~!eVDW2%+(;qT>?!RcOxV)l|J;#OWQ!^bqGx@Y3fPSaHkwS=;CVh
zr%YttlBh4Wl$T{qa_W2#%|jTI=(gl|-H5U&Qx@GPW0u@Mj-e?1T2p6`lQ&5{FG?|o
zLf$8G_V;vsx#@~w%%7C0E4uM0X%DAe@{yNrl5Q<Dcg#arCwZOv{C`_NzB(Q3t<@n;
zEe%FLj9@MYFx5IvEe)c(TT9Xo2eIsO?8@BqfRwYIQ%_QJ$<*t?@KSEKeMq@$A<ofj
zLf?K)DWn9=s56lEflhqG9<_$N9M7b_22#gORkVjVDM}xZcn2|DUTXB@QO>f|kAxb@
z5dEE6b6d`ndF|uGTKrKukNec)cK7rdE5~{$$^X)LY^uUqNkCn;{3`!rT+-?v_ZVMW
zf83(;vL*R?g5)KcwJvSi?HhTSW=Xb|vh4F>5+0ZJS+TkG*_Wv%vbLnwhdS*=&Xx3m
zLXX;r?|k=SYSN;HvW*pIDCyHB=pwc6wt^D>DB<&RIpDeL&-OSDGdb3+!e)+&$8a9^
zm8oLY2sKf?q86+5YKQuYvw++>M%T~{bSvFO_tV4lSUp3})2sAueM>(y{EQGIw^80`
zWppu8j3vf;W4m#{IB8rkzBH~IcTHs$FpHaUW(%{MnPiSICzxsG2J?b>&(c;FE7EFi
zb+Bex3#`@FTh<=yn04Oz+%9OBw5!;u_B#8VCyS?|r;ewIr=4f1XM<;l=b-15=VQNY
zeg*w%_%-lr<=4e;oZn2p`F^YXHv8@NJL-4N?~>nle!u#g{`vij`B(I><KM)8fd2&l
z+5QXt*Z6PoKkom5|7HL0{croH2lxks2IL9o63{PTSisnT83FSGRt9Vi*d1^*Q%t5(
znR;Y;HPg~e8#3+4bTHGYOc|M*WNx2%e&%zTFJ=BN^RJoz%3@~;&XP0Bb6G}ZSreE!
zFd{I2V2QxWft>=U1->4*H}HeN%Yokq-U@t})h}yE*0xy_v%Z*hRMz!bk7hle^{1?N
zgMx!f1yu{GACwT(C1^&_k)ZE`ZUq+&t`J;1_{HE+!Rv#!2OkJN8GIr5%i!z5zh|qG
zEjC;8Y-_V!3dtH$H)K%A=#VKPuZ4UZ@>R%BA$K{6%pBS-G&OW<=>E`?p&y6d3)5j)
z!Xm>8gcT2~6jnE^S=b9<J;O$Y%@11@wmEEf*wL_W!p-ok;ZflQ!%K$8hBpuI5Z)_%
zaQK+;so}4MFAF~uek~$fL}~uFjp!1wCgSJFe35k`8%7R`91}S^a$)4Q$eWROv#acx
zvzN_2GW){pYqD?2emg2_R8&;8sCrS2qgqAvj!KS7iAs%H6}36)r>NUGvgat9qj8Sp
z9Mf`a&+%nWznnF4Cg)7axjE-IIj`sZJ?Fz*1#>0lnvv^Z?o7EW<Q|@TVeV78ALOZ$
zXI7rMd6womn&*7psJvzK4$Hef@AkZ(<wL##`D*9uo^L|FBl*6{UoQXn{LAyN&wnES
zZv`R?#1|M|U{Qfn(NWO_ql-mXj2;rbGWvY<t%CUqwkg=H;DCZd3#Jy_UGQkZa|JKO
z1jpozc`l}OOuv|6F=Jz9#LSCX8S`e$iI|U`^M9`Hb1j}5{@l?*cA*M|Mikm!Xm6ok
zpD+J>((`Mc|Fm#u;hcq|3&$6pSNL)fzar&|Bo^6I)URl{qScB{E_$@+g`!^;y;rPc
zv5Lj&6l+p!Sh1bOJ}LHfu^Yup7H?F%P4N-MR}|k}{N3UgieD~%wfMc_86^TrB$nt`
zVrGerC61K1#-S>vWV4c^OKvauUCEzI<tf#?RJ&5$OZ6``q|~NT7fT0~jw~HtdSdDI
zrO%a7Wonk0TxNEed1dyMd9Q4ivIWcbEIYsKs<Nlb-Ygedu41|N<z6Z`zud+0G35uA
zpHY5I`D5j8RtT+7vqIkrGb`+=XjKfV7+JAE#ZDEcSKM9k^Gcpdl`7S()U;BYO5G|g
zsI<D$Tb1@y`mNG~%AU&ED#uoCUb#c%UX=$`9$k4#<&~A+to&7#993#oiLbJ}%H^to
zRhw0vT6JI5+ttce>sM_>wePA|uO3^yN%bMs->qTQC|;vqjRiG+tXaBdr<y%$Ce<8P
zb4<<GYc8p|rskV9uhqO+^KPxowL)v<tW}{_jaqTFn${XnYiO;}wRY7yRO`pu1!|Y5
zU8Qzh?UuDW*X~z)c<phuXVzX```y|n>*zWa>$I&iyw1WpXY1UnYt#*_8(X(Y-LZA2
z*4<n8MBR_;eqHzHx_{KO>IK)!Rqy$F<?7X{*SKEWdWrRt>W!#3q2BCz3+t_^x24|R
zddKU1Sno=`>-Fx&>e#^8sMwg;(y`TJ8^pGb?G~F9J2G}s?3~!8u^VG|#vYD68+$4C
zTI{XZ^tgby@VNYO#p5c+#m2RW>lD{FZdly7xS4VD<5tIQiQ5-<BJM)mS8+Gu?#G+)
zLGd}`3&od>uNmJczD<0$_@wxe@sr}`#4nBC7{4?AaQxZ$OYzs@Z^ftA52znrKY#t=
z^()tptv{{)NA=SiG-xoj!HEWU8<uI<vSDh&O$~Q6ywIpnqmqp}H+rSfnZ`!rl8xIo
zp5OSr#(y-a*`#5U)=j!L>EC2{lW|REHksdKRg=w4b~pK?$=#;;n>K6Ov1#9?OPYSs
z%+suDvq8;{H;-!Gw)yhrcUzQcF}lSYEw;DV+u}%zpIR1dS+Zr7ma#3Hx9rfeSIa>y
zN4K2P^3|40TW)B%qvgSt@3;KC<&Q1zC0Gd|3HcIABveain9wevd&0nkkqHwM(h?RW
ztWDUOurJ}ggbx$GNcbV)c0zh9|5l-`^0X@4s(h<@trA)#wi?uGY^$_Z%UW%2wZGMw
zR-d=J*6O!b4_bR#XKS6Sb)nW}T32sfzjZ?E&aL~l-q!k$Hf`G^wn=I;qRoUhv)e3e
zv!>0KHhbF~Z}UN$%Wb}IbF0n6wtj6x+U9QieA}{ZYqV|9wpH6TZGUgqp<S<bgW8R5
zH>KUH?UuIN&~8V&gY8bW`?%fL?S5^S(LS(!uJ%RSS83m%eY^I(+YfI)sr{_>bKAes
ze&-8#q4EpeUYPX4fery38g%H}VReTC9d2|i)v<BMaUD-|yws^grvaTdce>uWMCbaQ
zQ#$YJe4&efm!@43yX@+6r)#CIwYxU#n$Wdl*PdMmc1`Izw(GR6uXa7y^@FZ|bSvAf
zX}6cV?eBIjF(R>1VynbXiM<k&6GtSDOPrqgTH=z#wTW*f?n-<&@l@hRiI)?<OT3wQ
zuRFSD?OwQh&+c#ci0aXx$EqGb^sL?Ut)35iHSKk<caGl6`$Y8_)@M$ixB7h2*XUcR
zZ`;1Z`)=?1OW*tb!unO}7u#=4zZv}&_1oO<{r>*_gZpQH#($yy75aDX-@E^d{pa^T
zI6w`kGobZ=J_AM!m^R?`0UHNwAFyx0(E(=$To`b9z|{da2HY9&!2M71PYO>ekW?b6
zLQ>tNMo9@tU6Y0+jY^u3G$ZM?q{YtvuB7V&a}8`ZaKON612+%+aNzd?A0%f>&X@dr
za;fA>$+eQZC67;@o4h&six&&LSnb6LFRpuW$BV}X*@Hp`bsjWz(A+`m2Bi-!H@NoT
zri14W-ah!^kOD*I5BX-u%^?{>OAoCvbiuH|VF|+)4f}d{tKo}=Uq~sLQZ=Q0O3#$x
zDKk>$r7Ta`kg_x7NXpp}K_lWuj2*Fm#J!QBBTJ2JGIGes^&`)Y{CwmOBY%IX=}V(t
zdi$kQFMa&d7cYJH^#7<Fqk4?mIqLT4(xZEho-_K`7&WHCm~Q;vIp*}3PsUss^W&J`
zUdGD>Uat9ax0h$W{PxQSUViW8Z(ja6)h{(9wRCFJ)Sjs`Q`e-vnYusqZ0g0-uTuXS
zTX1Z<u`|XV9D8eA__#Xb29A4m+@^86$9*(jjSm}Nc6@{JoyU(FKY#qL3D$%X6KYIo
zHDTj~dlR!vjGS0wVuOh(6H_P7pZNB~vy;@M8j~7N>NRQnq*;>|Oj<n|lT#;8ojhmq
z!pX-bpP&5I<ew(rouZ~>o)R%7|CI7m+E1A|W&V^^Q$Cq;eai1s!>2Zxnml#-)VWhv
zOx-Z`z|@mdKX~f@*0hjm<)^is_R_Q^(+*DidV2QhO{ce?K6Lu>>2FMbd-~bwpH06y
zBma#0Gm>X4nDOq6k7wMU88WlP%my=4W{#hEdFJ=C0%p~j)nwMFS;uCbpY_LVb9Tkq
zgJ;j0y?FM9+3(CgHv6O5U(fz!_QSMHY1z|a(#odQPHUFdDXo9n$h0YGbJJF*ZA&|p
z_CeZrX?I`oyb|_GkyonD8T{I3uiu=Tb#A@6FU{RB_u@QjUfFpK=5?7jZQgtHt@(B4
zr_MjMAbLTk1rru*Sa52=PYcz;h=s)#)?V0RVZVjb7cN-1ZsER#A1wTO;jKk_QOKgA
zi)t)tv8czQltmL4y}Ia)MaLFhTJ+1}pvBRPt1oWLW;<c=;>B+-KDYSCC7G5KS<+}p
zmnAcotXpz$$!AM`Tk2VwYia4FaZCFyow;<;(ydE>TIRQ`$g(EO#w>ed+2!R0mzP-{
zx4h%>!ON#CU%Y(J^3Rt)Sdn8z)fEj^v{=z)Meh~ER!mqiXT`D=n^x>u@&1Y{EAFhc
zR_0h)c4hp^b}I+2OkFv5<(8GlR(`(n$CVFP1+L1os@SS3tD3AzUiIp#6|45G`gGN`
zRe!F|v^w|d6057NZnL`I>X%n9TD^7kd#k@%eRoafH8E>yt!cET&6=KTMy#2&X7QR0
zYxb@Ae9fP0J!|u<t++OJZM(HY*G^x%V(p%_=ht3ahjlsDm0cINuH(9-braUTzV5Ac
zN7h|h_tSd4K4g8-^)=TgtWR1$ZvEW#Ti2gn|J?>`2;Wd?L*I?XHpXx4yK(x)?HfPd
znEu9dZ?t*i=%!MerfmBB&7e2iygByGgKu7a^Uh|!%{ezex4Hc0d7Hm|OTE?Ot^Hdv
zZK<-Q@0K-Nwr@GHRd3C)wZ_)rTTg7gy!Gd8g|{Vc8@_GQwuRd^Z#%s0%Wb!}d$#A;
zUSdbB9nE%h-;uIo#*XDXcI-H^<HH?aza95>>$hKgd+pm-c4pfdyEFNniti-6)A^mT
z?>yX9VpsKDEp~O?HFVe9U0Zh@+I4Q%m0dsYR=dM?Kfk-|?)cp;c6Zp_Z}&*&|KmN;
zdsgh(x#!59i+lU*9lUq+-f4Rm?tORfH+vuK>%MRDzPI;X-uLVNO#6%PZ@GWU{*(Lv
zI#BIE+<_JcIv$vE;Fp6v56(Dv_~4m?KODSu@ZllPp~yo84%Ir8bSULr``wQ3E<IfA
zaKpoG4tF`+=kTDzD-Z8K{NCXoj|3dabL9CWb&oVX((lNyBh!y8JF@-Adq+M$TI6W^
zqw|lxfAsdT{Kq;R8+YuTW8WN0KOTL&`0+}|>l|-<yzTMC;{%ST93OXl#_@T_R~|on
z{LFhL-<x|P^NG$U#-7-E;>yXOlaVL$o-BMa_GHVGT~8*TOgWi)a`wrkC)b|bbMpO@
zKb^dN@~>0Iso+yNPZc~>@>Ii9El+hlHSpByr#7A1d+OAwuTR}M?KxfWbo0~0PftAk
z#_4mXubjSi`j^x9&v?#+pUHEk%$e9T-Omg-GwjUEXC|MSduGj<H_x0nbMefz_e0;$
z^?uCzwcqdg{<QZuzyIOcOlRwyeevwFvscgFIQ!Q*<6Ne5As@8=@Q;sP`lQ+?qd!@5
zG2&vui<K_cz1Z~PfQ#HqkhM+jJ#kNmYxTar!k+Suk$zl=<zh0I`?=1~1%B>T`Eh#%
zT!=#^S>i($?(~4`vRtx5Fk}fn#EB{l2nQm-<$h%64i}KaS<=k~F3KSfc+-r0;G!G~
z0MS4}AO>8+<*hV$A6yt*v**GF@6dVs%sXe^qVtZOx9TVZ-netk0L)|spd!E>vC80z
z0T=sF4X6&(fGn=o0v8gvqQC_OuA}gFn`<SA2iKX<0B8s<63F^)6L8Imx7}Rx;o=+D
zp|}KrR^Sp7+5p^WXa_DS;su}s&=KG^N@t)8xK_fOd*0h~iH0i!TrlDK0&n?w=a0VN
zDhhApdDqVye%|$S2?8&I%LTmq=Q04-1-J~w`+P23a&3Z(D|iVQ1&jv9055aJ85j#L
z17JKb0hkC(0wx1ffT_SVU^*}ZTqVFPU^b8jyuz&_;8ox?;B|008}q<L1nziXA-Ly(
z#lRA9i5ts+<-iJXX_adXSPiTJ)&lE*^}q&j`4w+)IU0BqT!O$`z!qRDunpJ_>;T>d
zc7nSo*ahyRaFK#*@z@9K2Mz!SfkVK%z+vDBa1>m_;8HK%15N-Zfm6U~;0(9}gR|fs
z43{|Y0k~w0kAMrn$G|7RMc`B5GjNLrpM#q;_ySyF#+SfXz}LVxz_-9v;5*<N@IAPV
zgCD^~eEbC50DcCSr|}E$EASg|3%Cv30e%PW0{6g$Z2STI3H${-;MOd65P%Hn7mE7{
z+$58AWpJO18(hi*#Z4H0Zh>*{i<@3h-0k9)5%(s8pty6wjiV6A#!(m)H}@i-xW$nj
zDhetGR8FW|P`RP<K;?zX2bCYH02H@A3PN$~gL@c-pt$Q&7^(<VQK({2#i2?-m4qq<
zRT`=cR9UEUQ01X2KymY@5)`*+sz7mzhWk|2p}0#^6RH*zH)`rYacig^R4f#CY~rEn
zLp6YE2-OIxF;o+%rrh&|;!aKrDDLDWK(&Hu4aGgKwovV$+C#k{+p?-7WXr2FR2QhO
zP~D&sWjk5*fa(d=3yQlNeW3b6^@HjUH2|_LHV`Tq>P0AS2MvZA0yPw}MK&C=MK%K5
z0#PqPje;5tH3sTss8p!2P~)J+Lrs912sH_6GSn2PsZi6Prpq?8nh7-vYPM`qs#l=q
zK)nj}8r17hbD`!z&4=u~ErePGwHRs%)KaKrP|KlKK&^yY1+^M#4b)nwb&w6Z4Nw~)
zdvu$i-sH|W)LXK}uC_vLgW3+Y1L|$4olx&U?Sk43wFhc1)IO;FPzRt6LLGv7SGG&l
z5vZe3$Docwy$5vy>Lk=DsME4hDZ77Xq0T{_hx!2OL#U6SE<k+@^$FBP*?L!>L0y9S
z9O^RE7f@HAzJ&S;>T9TPpuUB=3iTb-HOMaF4^TfsHWGh=x&iewWIOQ}s9zyFinpL{
z%hs*>9qKOBJ*fLoe?a{S^%v9wsE1JL+-{eFUvnRh+iwQA38^h;ZgP9TO-Stz9l%{^
zXl~DCfewVu3e64WVCZbnA<&`FVbI(-jew4X&JG;~odcR1!MUJwL+63c3!M)-KXd`;
zXy}5_G0@LJ7lM8sx-fJR=%UcYpo>HE8HJM2rJzegmw_$|T@JcDbOmtpQ&)nn3|$4f
zDs(mI>d-ZyYeLt8t_@uWx-N7*=ve4D=y>S*&<&s)LN|hL4BZ5}DReXF=FlymTS6y5
zw}NgB-9|RIbvx+xd`JSi19V5|PSBm9yFhn^?gpI*-5t6IbWiAB(7mDiK=+022i+ff
z0CW=cK<H%X7oi714~8BBJrsHvx5c4TphrNDgnkKn6!d85G0-nVr$Uc~9tS=@p(j93
zgq{RF8F~uzROo5Y)1hZT&xD=@JsUa=`W5Io(62(j2K_qpT<Cew^Pv|&FN9tMy%>54
z^it?$(95A$K(B;e1-%-24fI;*b<pdfH$ZO$pEl5&px=bv4E+}L7U-?e+n~40<0SfR
z=$(AV1$r0sZs<MGdu8KZ?}t7BeGvK(^t;f9p^rcxg+2y-9Qr-z6VNB2PeGrCJ_G$e
z^jYY0(C48)fc_BrBj^jzA47iveG&Ro=+B@pL4OW?8Tt$8E6`s;e+B(D^f%DoLSKdc
z4*DAO_s~B;{|J2@`X_MXQ~wNolTS%N{|fyZ^eyPy(08DJhrSDa5BfgzAJBh7{{{U3
z`XRW5t}|e8;-g?_7~D=bVOZR|hT(z1M?3ssa4R$ujLa~yzzBqq6-E$@U>Mn8guvkD
zb{LFs7!fccVPuEFt;ZZNa>B?3BR7mZF!I942O~d>0x+Us6oe52<2e|GU_1|_FpMHF
zioz%cqd1HbFiOHG1*0^KGBC=*C<miFj0!L+!l(qJGK?xPxY=0^2DfZ$z^DnM7L3|3
z>cFTAqaKV{7;!M-Vbq7w07gR?jbJo}(F8_Q7|mcbhtUE?OBmcdZUv(?j5aXZ!e|GB
zTdFU>=m4W5j7~5*!{`E|D~xV15@B?Q(E~<L7`<ThhS3K`Ul{#h^oKD3MiPvHFt|DW
zB8)*W2E!NvV<?PaFowh6#_b3gBVoJ*V-$?hFvh@m8Ad9Mu`tGgPcax1U`&KD3C3i3
zw91$Y*#e&qV+QzKgE0%nY#3=UUV$+O#;Y)1gYi0yxiIFzm=9wCjD;{3!B`Aq35=yM
zmcdvKV+D+rFjm1>4Py<AwJ_GfSPx?ZjEykffUyb2n=m%RcniiB7+Ya%gRvdP4j6C4
z*a_nu7`tHXhOr06UdSf>ei#Q}9E5QQ#=9^M!#D!tD2!t;j>C8l#t9fFVVr_-8patI
z@549?;~b3hFg}3sA&ie;T!8U0j89-(gz+hi&tP1F@i~mkFus6s1;&>!zJl>JjBj9k
z3*#z`?_gYm@jc`Lf*)aAhw&4P8!&!`aTCTbFn)#c8;n~pZo{|(<98T$VcdgpAI2Xr
z{)F)tj0Z3t!bpda0h6<S1(OdM@yRL^rUlc6>4E77(;sF4%uF!(s8SY~fiSbe41yU9
zGaJkhn4vJkV1~ntfEfuhJIpAUIbi05nG0ran0a94g_#d#ewYPdM#C%!GX~~!Fblza
z9%f;fMPL?%Sqx@zm?dDAgjot^X_#eTmW5dkW_g$uU{-`#31(%ORbW<ySq)}&m^EP5
zgjow_ZJ2dn)`eLQW-QD&nDH>{!)ySvA<RZF8^de@vnkAGFq^|{0kb8{1emR0wuadT
zW?Pu;V77<(0?ZCDJHqS)vop*tFuTI+1~U<6cbGk3_Jr9BW^b5%VD^RC4`zRu17Ie>
z90)TR=8G@~!5j>82+W}{hrt{UGX>@dm?L4n1alP3(J;rrd>Ljc%&{=X!5j~B0?dgp
zC&8Qya|+C<FsH$s4s!;~nJ{O;oDDM#<|{Dgz<d?vYcOAjITz+UnDb#SfVmLnBAAO|
zE`hle<}#SeVXlC=66PwHt6{EzxfbR+nCoF~fVmOo8!$J)d=utom~X+{0&^?OZ7{dP
z+yV1#m^)#<19KP5-7xpS+zWFb%>6JAz&r@^5X^UB9)@`Y=24i(U>=A09?TOkPr^I}
z^EAveFyDuH7Unsa=V5*T^Fx>)!Mp(TW0;@7ya@ACn4iJC1oLy4mtlSZ^9sx_VSWYk
zYnb1_{1)a_nBT#?2J?HEKfwGE=5?4q!Mp+UXP7r({sQw?n7_fi1@kt{J1~ETc^BqA
znD=4+0rO9of5ChJ^C8T1m>IADOTp5x3|J;C3ziMb1IrJVKdb;)nP6pxl?7HHtgNtt
zU<Jd<1}g+sD6BA8;jkiLMZ(GsD+*Q)SUF+kf|VOq9$0x{<%5+URsmSiunNM8f%P1$
zLa?5PRTx$gSbPYu7_8#3O28@!s}!u#u*$$H3#%Ng@~|qvstBtRtje&ez^V$X8m#KD
zYQU-qs}`)<u<F373#%TiSXgne;$hW?)c{sQSdCychSdaCQ&`PlHHXy#R!dk3uv)=t
z4XX{Twy@g4Y7grLSRG(>gw+XFXINceb%oUpRwAtKuzJAi39A>Z-mv<>>I<tMtp2bD
zz)FHO5LPm*7hw&8H5k?qSVLhAgEbsh3ak;ZM#6dt)+kt`VU2<HGOScsV_}VhH6GRk
zSQBAQf;Ab|6j)PXO@lQZ)(luPVa<Xy8&(>uS76P7^(w5_V7(4&F06U5=EGV5Yay&f
zuolBw0&6L(Ww4gRS^;Y%tW~g9!&(DtEv$90*2CHWYa^^TU~Pi+Cale{-h#CS)>c^C
zU~PxB1J>KHcEWlG)-G7PVeNsn7uG&l`(Yh`br9AeSnt9*4C@H2qp*&_Iu7eSSSMhe
zgmntmX;^1qy$|awtaGr=!}<W$hp;|^bph7Lus(rx5!R=$K7(}$*5|M;!}<c&6<A-w
z`U=+9u)cxyEv&1szJqlQ*7vY}fb}D+>#%-;bpzJVux`To1=g>yeuH%j)@@jKVEqp3
zF06a7?!)>6)}OHcg7pB_Ls;psGGGI?f~{d2uua$&Y#X)*wjXSN*a5IJ!Ojdj3+zDH
zSz!mk4u+i#b_i@fRvHF79CifkNZ8q7N5RekJ16X1uye!C13NG5e6aJwE&w|kc0t%N
zu%Cln2=?=^3&Sn~yD02pu#3Yk0lOsZQm{+IE(5zP>~gTn!>$0kBJ4`AE5oh=yDIEz
zu&cwa0lOybTCi)wt^>O+?0T?cVaLIahg~0b1K15=H-g<5b`#i5VK;-_9Ci!XEnz3X
zZUwtF>^895!fpq<J?s}?cYxgyb|=`KVRwPu6?Qk+iLkrF?g6_e>|U^Y!|nsSFYJD>
z`@<dpI|=qc*vYV8ggprMVAw-o4~0Dp_HfuKut&fi3Hv43qhODQJqGs6uv1}=g*^`T
zc-RwQPlP=Q_GH*oU{8fT4fb@{GhokzJqz}1*lDm|fjtNItFT{#{W|Qqu;;;^4|@UZ
zg|HXFUJQE)?4_`m!Cnq~1?-iuSHWHldkyThu-Cy}4|@abjj-Q<y$SZ4us6eg3-%V+
zTVZd5y&d)r*l)w$3Hu$`yI}8zy$AMQ*!y7bhkXF{LD+|2zYF^?>?5#`!afH3IPCXe
zpMZT5_9@t>VV{BhKJ2rw&%r(q`vce?!u|;M1=t_M{si_#*q_4w4E811pToWk`wQ4t
zV1EhwE7)Jd{s#89u&=`Y4)!(J-^2a^_K&cy!~O~O4cI@!z6tvm*uTR54fZYAw_)Fb
z{X6Wtu<yaX5Bm?;f5QF?_5;`tVW-2+fCupKDRB*t0gnlf1&<An2Od9o{NV|JClfrG
z;mHC|AUs*&34$jWo^0@hz!M5j7(C(dM8FdXPj+~s;K>0`PIz*`lN+8q@Z^OjA3XWt
zDF9D2JO$y2f#*4R3c>R{JcZ#Y0#8wRiosJHo)Yksgr^idrQs<9Pg!`%K>^6-95Z{$
z(+n~5GTn?+{cmLaU~Uoc8A<yW_Rzfq@k$kFZuw6<$1r`Kfy85a<NJ~_I}!Q6v4%Rr
z&n~JE<!t7K(g(WJtsapyo<%>p|3@<JF<uY#Z<9pw-Ba|X{aL(0_x8_V72W<THL1&b
z{s&FyR{I&Oq}!9oxs>utStXC+x8%Q(affY!;B#{9Usyo*xugf#NVfb(j#F>`Gmx~3
z5lUtM&q$kkQU~W4&c}~53liUye_^oKUv!w3qbp_~Yia(EWTZb8?&&u2Z|D1ChTHTE
ze0lgUat6Z}c?RR@#(wb+p6*|AZY3QltK?JsPrG&N)Th*+j3-hiy5CEp{DhpMPCH5&
zW4<LEI~4QbL+bR&@d3H1(}q0VA0;lKIH~9L@8?RV-$@&meiZK|52dvKp+w?J^dja9
z#2|G^7V-j-c0ZQ%AilD%VcY(L_MW7**BniAD3LHyrp}}bk@mQa<Rp@>I$m;7zD|x1
zU)g+Rk^GAPF|wFcCX&CV#8)0)|CO>x{YrU!WfEN(8wJU;#4m0Yy;AzI=t`a?uSCXg
zsc-c5;>*)g+9j+Y`S&69;j2@Tj8D#;Q+KJ9F+}o~OoB-TvXBIk0Yt|1d}IPCPQr<V
z3nOw)Brhktbz7ch=26aP2Gez}H4<LVk&H(&cHE+W^T#;l+Cq09631xrWS=-q`!z38
zo^D?MwO%Q@at>Z0zP?b4=JDhZk@hms3w2cp%~EF3m;A3MtBAxweB+JNFQi>aent^7
z%XyG^r7Q{LWm1kvds^+4lCP;=G$s3xbIp~o5|^(m9ch-f_%3;$oF~=EOT<^6$utX+
zrnDa!N2E-biIgkS%M?nfyBnkzkvjRBh_2*)iHkGVOI=GI1z&wnr&+=XK0lF?4_}>%
zWPEb|oVr_1DfKFOlWRl-$xfvGh?IPYjz5vU)q_YIk~SgrF7ZoSWZTd9ol?rajIazD
ztf!2dG`Ax%o~$8Kf1fdZ2}85f52TOy+Pw6kpGjR(&+ER9vY%^q#u=%H_M|qEJT@UY
ziRds+880%JRtC>m#;1hsDC1-LN!$`f@+9ZomnPeelD;7E)*#;##-Gv1l}@}er${4`
zCrMAtBBeZ%W?wRZEF#-UW#aYo={zm}ElkFJhP`5LNq>)gKv*9cHn~QmY;2<$)9B9Y
zVmj+MeF5RQNoN~QFH65UMCzL5`(JfWQs*M^W7u?ZfY6@K@~5|>JL^K~(KjADy06oZ
zWlc{Z%uD({>d3ifn{@nn9@6_0mjB^#x^*Qy2flew2HSiF>)Ofx7)qZ1bmld^2kA~f
ziEA_U8PETw>$KbVsoRUpCDKOa{4su~ZO))POp*x8hWg|qBJDPioF<=n?QJQ)BXZuy
zkr#-ZLy5zO`1_DJ#9#W3q$_Dk+J(t^BK3C6D@B**MFmkFB#p^r@+Mh9q(9{$6N%(W
z!kzO<>Ca_IHzH~M-$u&vKOyD(zlE=!|4*gV^Z$g@t>8O{|EE&U`~M{r`<L)2ereZf
zgky>`W=XqABqG^&oNJa`N2JZZL}a|)O12Pjmuuv!gnbn+dbvUQ29bCt5ovFt{~qz_
zQ&$xvi%BI?oD?A(BOP;XO14?Y?f(}&;Ne;3H=1aU6}QQDnmLw1(fl>pK+Y3h^AO^a
zaf|lr<SyY@tKKH7$Pu!XIt?hlbaCo~b>-+t-ErJg(l&>X;a)zV8`F|&p2XqIxj^~9
zG2adEq{}f(F`sgMQ!*AXkE%2oOFEcaw2WJ?krOWJzv@4k2gxVTlXGnq_qs&l_33cz
zb?(QX)}!60qjuB$gO~3Z-(Dha@3=_aI55xZ;WOO5`F?Q7+;Ts=x#dAi!nXWy0p%vb
zI(YcIx#f@VNiM>)AGV`A&+|i$gMU!}ek+=p=Z8Cay<q#h|1JH?^L~fz_~B5RdF~#t
zAN=(T-9Dz=OuI$cUmwAIJ`m{k7u&qZdt&cJI7fQGc7BiXJk09itM7~(ESJG{L-{n)
zCVXz_$ap1gNpsy3?yTd7tao3275l85U$u)!+x49bN6HeB&&_MTN_!kxNPTJN@#I(X
z3YkMv$STr}NZ7vAlm5+i?X+Q@f6Y2`?iu8IDDYa3O7xE*BVC;N?Fz3y%OKYS?e7iC
z^Dl8cA}_;?_9AI{<8#I%)`P_D#3xehlBSGle=<yfD<93L*`6SLPk3HM$GKih+Z6M9
z)+PI`bAPp*QsDV@U>=<@vjXL3MDihNo&0-Vo$|jxcOgG1#C-7FIUJ!~^2zh!<S)n#
ztJn^laEItFNW7xw(-nQ8p;t=XEhWA@NFCD8$@_Gg*{_{C0Oh|RV-z}*Dz5y8eH!Bv
z9VxHqRQAF)?3jD`cfnl$F6uYhnTE{I9?8PAzwO$6ZqAr^onfl{3(KD6{y5%$E60Cz
z*KN%$W|W&JF+VHZ-`u55`07sT_kX|%_dodmTUZJAkFadJD(B<$#H{^XoNEf(=aW+W
zeD){JJ~xr7_hTKOyXcC$51;$9bf5H-aKe*%|5m=I-K8wT)9(LY?7qB-lsr9|FQ4C&
z`IYtXr}dr=BYjESB;3F0|8H^l(*B!dpUBulq;D`E62IWnk#v1-Px?z*KArz6CCszJ
ziJN{R%_nul{YgJDJNKbYyi)qZj{k+K^nU^sPP<$;c>Qybes!L#B%y?3i*w(So#vq=
zk?`8@T!;EnK2K5!^>luckLJVVxEGE!i1)&{Grn-^3MJ`wpNM}oFOq+s*()=;F@1@%
z-9J%GTfk3@^AnnRotN_><3dIk&JTWYrL4=z{kAhs%DI*E%xg@>QHJrR?}U{!#ZHuz
zo$)@R0L?zW@|q`7E+OgU3nKX!WZV~cew^#dXv&ei?lZ3szb4-@Z!Fh+*3<8wnp^&S
zj&AGe&-!`LfO8MV_t&+@^Mr@*(agB+?K8LBDnj!>hGkwJHszeiFt^;{{OdmR`XG*y
z*NjL05+`*Z^rw4XO3`DuyPwmX$u&PnU|MW54;c4@Sxkd-i3c6%w)pRPm${70Z$ze_
za^;hDG0XXXQi@rSIn$HjA55nk+l%=9CwI>Ko*B;*?ovk&k1_A#h_`G{&b4rf=R?{Y
zbrk0c4(BQ9rdo-e8*-%yA30Y;2_M{7c}YGmRVb&C{G<)%V|<og{B>K(afHvWtMQB{
zka_)xvNZXXj3HA9+puE&I7pw8I^aA?vHn#mS<Cv%LirOZPNd#OkijGu@rA8M^JJ2P
zG$RsrzrghKkk@IxNxmdCh{Pw{BCIdf(JR?LrR~djugVk7Wt{uHp_D%m$+N`4IFZ13
z`S`uXHu>-p)~fE%bOlcE*Kn$-$~4LDW&moUH+S!28zu19%|;CBkC<MAdh|m~|E@!l
z5Hqq*Qg0qdj2M9!@Od?WzY@mhHW^mKNKEFho#h|Ubx=R#9MEq-KZMdpi;trr@e5|o
zefR{gF)bTGbO|5|$%y8$I5NYCxB<UFK1ipBaOXZCA{)bGN|mq%f0yVN?)C-&Iop#4
zUMXglz_I(>d{X>`!X%0WICkdhZ@+Qm%Za2?98o;xWlj{!Xz0G_Y1wCmhSH|E&CIgo
zm_Wvm6uS9-8%b)z`3<fdXS4GG;2Z(SY$RqMmp$b_(K|<r?7}$GHhcUtsYt&<*>h+2
z$-v0;?EeME8F?r2>eD!BNqYa0$V-ezI2F0aVN2xl$frq*oFwr@4tqBLHLYjsMfT4=
z?lF<wBHKQbCXw-xPx0*d{(ISd;UlZFTu)P$ZciZRPRcJ~ocxJj;irgt%I@=L>x>kT
zqeaA@5qF-2<99XUQpBl<LlHZv_cWWH;kGhjegsFur;)gPusyekV5|NcL!Pqx{QJ_~
zOC%z?2sySfe6xtyXCZz$Rz}nmBFaY;^_d+heoqjQ|L^u39MxT1ccvBTq!k!pMm$0I
zpW(kglWXCZ!=K`DoNn9^ydpg9_SpXS{QNzCkHh)m_{aI+dGwW2@+bZb)k5ko{6p3k
z%ixX?Zkq`|9KOv*_HhxBf9vCHQ{n5wm-^bHPlmrncjB}ELrR-up9r5uJ>v5ZPj&O+
zvrE2wW}lQcEaq@;n-;T-13rHzk92!%e{2r#5BS`}J1~w|Qa(J!7e>nB%XfIr@N7PJ
zM|$(@>u+v65s{K-!58n}CF@J_>iB!dUKv+8PKNu@ojCr_s6)q{W%SjNBPA`K5Ah4T
zANE7om9Pt9{yT1(;wEG4-~Ils%io&W<CgJp7*}8H5s}U~0j9MjY^{$R+ae+#wJG*(
zY5QS|!{$8MXTqk?jXde*Gs_s_n%%bQ>px**<lKi1@#)FA_nCcC+N+qux_jHKj0bGD
zGG_R6-1s8n!`eS~^Z7Ypn@PG6kzuh+tt#UxN^*u}bKF=CU%tcqLes^~Det2^yLLCe
zh{(`8p*JPpl78sb&`UDb`1~ZCQ-3nH1~8xOXQAhT(Bm}cr+KGaXJQs4KcRfxGIXWn
zTXfv|37t<r*1y}&AE)mdCm*Nxtn#{JldI$Ne;kLed`@1SJ|u1QaT}9%D%X-H+mKvq
z#qPVNJn84S%e9Wzq`$j|&T_^SnIAB$4_8n87~em_NE>8*Ob8w6(`T(jh9=X^i_iU8
zGPHg6l+XsGEQuze#LS)&a)(@b);uWWY-oGZfRrWCB$SvTXG89gE95|AddLp4n!H9P
zkRgxfJt1tDAqk`|VZDV!lh7P3vZa%okIS8XT(+wr-AMweOUjaH5}NI5wsdlnTws_}
zWDnU$mXkD+N|K+9%Qa`~Lic8bV_LTIB!+}P*2`w+Xc7E?+#r|98Mn@T<C)0deWA00
zHwCW@o<li~3?kh~3sPIolaJtv!TZRj;Nrn^D94dOq#J2LYLn>TD7P=NAFzJ}2L<y<
z;E2ee2SK;Q9`u9T9>wgoL*~bOEtU2fbmhr*`gGeBzsPurBZ&3ni_7gJKHbQ8$%CXT
zY5udH_~Li-_heol_ZQJ|uZ<#oIzdN+_5^JVjtW{%(nu;vCS6D~Qj_UQ+W#P^d{7L@
zNj8$@B#oq!WYUE+BQ;58qJ!<Mx5y9V3OP#-@CRC1CL!yNpj%m2lZ7OWq>^OPlXM8W
zCFOoh)<#*YlG0=cSxpv_G?Ge^Nl(&&6e78Ew8+Y~o7K$GBJfXgop6i@JWr0hb?(%^
zoP%c&xQ}|Qv%r;vV_4u6lHxlzG$%zw26iX^q8FGzod%EfAD@%Iy9Mrxj1Od62wX|#
zlPM&HB#;J?@zPEn6Ie5_JSj@{k!@rpnNOyW6p}z1ko<u)opD9(e>gq`Mg|5-Z+dJF
zG_%~vay`pcX%`%$yUREZWVu8*&S&8`pM|f(XW>|$Wj&cs_@aIwpVSR3%DXc|x3tFx
z=GJ^=Oi#5_5ugv#HGp@Ph6?5b@t5S~LT%Y40?ul_0;cEKB{Hqmd{s<sv!gT3(0r+j
zF+>M!(R}$!ccojX=1ao5EnmGb^cYn+Q*Pez(qymB6w13=z+d9lEv-&j)@R{MYg%&4
z88l7B==NHc)ikxWR%Pj{_<ESGV;9Jhs^yL|(Jm0MOYsFUeLX!bQ#QU(VdxHgDqU{b
z4ZY0CyUlpAIq^r>(Vo7#8GSY|cNcUlb6kSu3D8w}M_RzqX~5_C12lu9rep6wyR4HK
zYMG;%jXo6}pDcEC=1bI+bfN<~sQWZkw#R2at|rhl+D`R!;TnpesD}W>@}r;=u8^dl
zmNTBZj=i#7BJ&)TM$OTV<`()?6Cd@H9qr$rni5+yt|-1VsGC}=FbPt_mF)tV`+_gT
z>Gph2Apmz6u9y>UV|p3_%o()j6T6&E|7a)7T;{4ZH4PQxq*a%>ig9#mIk}Q^5$g?8
z!tYnGX4PCf+TN+67?0_M38GGAM~6OUSuRu<s+stJFKr{g<I_)K1KSv?JND6b3IEF2
zM9o8N4Lfj#u@!g19H!25j*j^B1z%#+&snSZQiJ5BhU1gXE|6)t(_%`qq_Xw`G;n;b
zq^DVDjFk*igk{w9;d+GQlVqnFQyjZf2S|5POQ3(W<DcEhA=f!j$FUc&OZXjjVvZBL
zN^<n0>0iaMU$&$9l8m%k`uMGPG#ff0m!yxem-_|KXPp^^fB^nz1M)aN<*lTET^?y~
z-OQ+fD*=2tQkSP`b-)Ft5Dxeex-~VUJQD&A(bS9UPF)<g5oVO1UjS=NH(-IsId0F<
zEq8#l@!!(ZJlFL`ntn-73mBr8^R>uDe3{8T>z-7`nsL<X$9Bda0zh`Lr@zSepaL{s
z(|2>1k-*oUq$c^7D*D^Kb_HAP=^r}!{Fwn{6}#@pw+<Rn*~mzDQp!)WpZK5w-=qrA
z97(((ABw-u$&fk4|KV*Xrf#(77JJ5uM=3esGtO8^x~;+2OHt8@sS`DWL^FM`6S5HP
zxtx%#ABD`}*x!6)4|Md;SUuD@>4_P4B@cXE7sxC=Xzavn(r$`9BhB%blp-DbZmBIy
zp<PzcjPzNOvf2!wk>j6iC96Q%43v44rs??fax^C}q!v5Nkff%`YuTtFX|iW4M2L^R
zk)FmEZrJ-W`Z+c96@9{;I9_AwLC)DIOFQ4~5ZClRPRxY?)OWNRK1w-@W8e139^mLZ
zXN-MTx#L~NPfq&MU)`KVId+LZ%+cR3@#7Tz<Heq_>`{Dfo!O7<&iI@W<LEDB_+Zg@
z%C{4Ix*16lGx#GF2o)dwJ^SSlr#y}wgIQj;d?o3x5aonlAa%}hN+Cq-8GW7jD+8$K
z_&0u(zUA0gKC&xE|0MO@SPRq66-Gzvs?;sVV8$qR6~Mf?TBT{Hp1A5xd{t1|v!PnE
z+1{eNq2IJRnKwNZX?lUpWSZkvow31~LAUxeoiWDI)ZTGRp{bcQNb@TN)U4pd+T2Pq
z_86V$)7*?Qw(un%>M;H}PF&5bBlZ-&qN{XW$6lD@g}iiR=*CRrSI4a}!(HP`Oolo~
zx66*(bs1Oe<up63UPs7kuZpJKsTci%8KpmEr)AEZ7>ChRfu`Ji$;(hLS&^EmbgS>g
z!!>*qNfmo@R>n2Qr@9laI^9woQ(2liP-U3o+sw?Z@`_2#W`1qQZ6m+gu#Yd<IpY&w
z4V3tjSf>GeX%8rsD)y4h;0ed4q}9nfz?TCJwT^bFFQ8Vc=%hPgj?!Mqu{X3j89{uh
zlGk>+<t8*yC)8Tb(3@y?Lp%M(Fr3hn=<19Y>JQI5-g8|$m5vDH(a_Z#tuC~eaO@SV
zPI?nxDwT8EO+TQ!q3h8ti7%-dx<9q$WmZFvVy+AErB{y7tSPrH+R=9#U!pZse};4O
z66M5tL{j0)x^P>wj-Y)CU-mWh&+LG6B^C9J6=|*&lWxKmcE&Mv=aT?)%)|6OOFh$Z
zn{P&`Kh4e5YnwiXX>Ssfv^u3}>Oq;{=<&K@_B1EaxxHh*K%c{oDT42I2FU)7JNoOM
z^jYutTw&Zlm@ODj2e#Jx(tmXawz`{qIo;5)(nk0aJ5W4T^6AIJ9VgsX#t5~XZUZ^O
z)rPeHfsCsnU%oeVAL@iUdVQRfQ)qTZ2`%l^^@(<TSi8=ctF+XF(pyYvMQxqfYtuA9
za$szupL6xlam?i$r>3i>N{-@~*Ku@VTV)Mj;8mV;Oq#ZEv{#zFnfBJS!$_iuXAajL
zw=Qg1?Tz+cH#bL3sTW|x(!WfqgquT?b5>MG$EPUm<s6@7G`VNn89P;9`fqdQ!4k&x
zaj%n0B&3#_cSdQwhxQcx4P&48NHcG$r1lzppr!xMdgS9==k$0dj?Hu};Al$iN<H%Y
zI6kk@T-5O?=h!FE9^=^e(B$T(pku${_;;skZpZ$aV^5&ny(YC}4$9IdXFzgOnN#O#
z97_w)O}<T12=Z-`!jNy1bcB4HWG>{}B-<h1COHB5HpvC&+azBrew#$u@@*1+kA+_)
z;kQ`$RT5QPzDlAh$X7{JW%(+JswZD1p&P$Sq8iCpN%%Djew9RZl&_M|pI;?Wo#m?}
zs+W9~M2(TJlBgN-RT4Ez?NR&HLUmBRua?TUN$ARNlc?kJZ4z}t-BuauwAR{GpUPKB
z)aUY55_MU=N}{gHS4q?#{3=N~^{1|&D{4!=O`<(IR{yTE>3jS?tpCvI`iQ}cnm#Gt
zDbc6oJ0(UV`A&(^#N-!BjHdF15~G=Xp~Prz@(U$K3;9Bc(NexpVkGbjB{4=T`6`Lg
zTD~u0w2?217;WX7B1Sv;rijsAzA0jKm2Zj|-4MdM-a%G-rI=sy+I?<5DSpBPGK6&J
zxoS@m9Cw8#Q0fz!dAG%*q1*5%e(V#{uc!;}9KDZYZ@{of75`n1y)ylu7rRP#?1gDB
z@A$WI>`}BwJNCA;yKSuy?e0F|ZC+F5`zgT^EAb-%u=pQ94TH9cbdFLRFp)KXR)tZ^
zBnCTRBd<By(G$DjwB%^+FY>R9v`GG$wRy<m!e5&J^17Ie8Tnj{?<n)T@Yguli(L4N
zBW^#6W;`=#F6hGFY+(O!;k(_uBXKdyQ%bteQ=fTcKXBmhm+%_uVsbAUGENZw>Iv_`
zT%6h|PBSsSqAcNJ+@$1{5tZnkMkzJU{dVIU%Bn8rQOasA{G|%sMY-_zDtJ%iVth|o
z)5Vs2)$*b!rH<+lsi(SLLMiKciKLWKFOEn(#e0#m*7qXiYv4udp`n+Glya^F$$Mkc
zgw!N5+6oeHGcO!rz?lYGkXA(OLVF@DQ|L;>Pv}Wx6qlC!B9V67jtnOfMi@;vM-xmB
zCFd#ve@g}&r5sE@N{(m(f6oIP#T-m2Zx<K7VC|+Y=}#ivNFtFjy}OqMls&vGp_Kgf
zBI`(RFK<%H7%gleeZB0X?C0emWq&V+DF=`wa)J!>a-K5T%VkO_{~&UO4EFLZ<q$8|
zDTjKIvJWFtPk)jWGD12b|3<nfoAM<W)1n;Z!rv`$>y-7%t2r>ng};UZEO+5716bi=
zT%%m+!ryuUR=Jq3QLc7z)*sf;Ow89Q*SgqcDW&cRe?b9Q@4{E!fekJ;`wTX^s0hk8
zT=*+H;3|Rxf2IYHdMEt-9$>Qz@9TlLTs#!c`QSO>FEjvKUCj3=w~_5^Gh*K1!nf^#
zw_VJOlsjGcV-L<VF!CIX@gvQ<T=*6~cr|n|j#KV&QIV8;UCh#y(hi7mfpWi#@d4!l
z7vp0}X&=P+f%1?SoAO;Rno`=0V3H$V0w|Aqk#l;?i$A60NytQ`oI)^>HYemEa*l=U
zMC$Z3$w|(5k^H~!Mb6h*a*jwnOM4Z<h?GZ&B_DceMEQ}IcuFb1(11vph34cFFAXWB
z-iayo@u`=ll%IL&M|sIhFUrr!WzvUy;bjD+^cf+6eCg#!%CEdkr~KN>WXf;6Or?}Q
zCd?*Ry-cJ0&dV&yYhKbQzxOhiQqG?s{Xo(amXhn_CnA05h8MOc{Om>A<4y7l*+G8w
zvXJsOFXt$4d6Bxi?M3SSju)x#-@QmZNxu@L-tW1XhbZrRkvjgvi^TD#7fJsw@_<O1
z4_$Z%0;GG9`p$4sY?IO*1o?#uMT3ADlnmp9Gbp`a3e6IRaGn6gwku#JrR{>OloA(_
z8Ick_!rzvF@^`^J$^aLcNvTXOm`|D6h4+L|SzPcsWuObfDJ2cUxh<6F5kyc1yI?V;
zq(_jKQqmwYuBlKLJ_iMv%Q*9yWt0*xk^5H}znn2_6=kH0%u^Kmn;_$&lJEpMD7|1g
z&FqJQ%u`e@7a13o#7pF!K%Q9>Wb9OVU1a{EBn=|trONLja~xH`MaEPm;fc&^l%zpq
z>{OyhWZt7BE+XTrD&&H4l;THZoKq5>pbTXZ7a12-Q5Ts5DUKb2jJZne1Z62pxX9S4
zO1hvFr572qoo5wGQ(xkfG>DABO2QLVpp-lj8B<kx7nx%z@vBJpLzH4CGT%~_U1S_p
zVka`^QdM1K993c`GA~orU1Y3PHC$v~rfRy#IIC*8$Q+G#$PO~*syZ$*FH?10WUN*7
zTx5=>VqIjcRdFsdhg0z`G9Ih?E;2@|1}-wEQw?2YY*vk2WKO0UyU2K|#7<;>r<%IR
zIIWtw$UIIpcaiZ}wQ!Mnn`-GIW38+_2r@rYtz7UGWos9itI0E?&U2&RP_}iEd7KhI
zBIB-V?;>+J<weF?)xkyPZK|V-jIT=UMCNy@vx|(us*8)v|5R5O8K+e@7n#SYL>C!r
zmDq{Q^HdKP8IM&@7nuX9UM@0TtKKd$FH?P7WV}>;U1T1o`nkwBE9*jn%ng<J5gET#
zl8emslouI;mBdSAZm3>#k?~s%a*;Wr8tfwDw;JLib3irJMaE`XpK{iyMp6#<Vo;{I
z$h=RDaFO|)5<S5~#E&4Aa+DW;%F!-(nNq?Ceq@e|nMV1*Mefp+<b&W9%5)dG4^SB{
zat+s!E;1k2*<Iv%siRzEuB~&p$n{WHaFKbF-tA(uUFkh8_E(hqT=?5d(Cml$J<j+3
zqCDwhv%PVr(!pl_wB&=>tWSN(#b$o>=PvwdC2*(2!Dim|7cTanoS(6HhGJN@F@t)7
z&HNkmFf5vHP}=_wXWs!HRk6N5GiPQN0)*bnZZ?EolHF_?2qYm95fLd?6haap5E4w0
zV#AKu@Y>6@S8Q0XU9Y_&_AUtahJ|Yv|L-^7Y_dV`^ZP$HdER}`d}sR1nfbms=geW)
zFTki@g8c}Lx+B=_;53Fs9}7A#?5AMVBf)+S?#!?|z}<l!kbesv0u06R_h5{9K^~5G
zf=d{9@>c|93|QkLK>s4RHC^<v08l@G(Z>jMZzn(>BPfce2|(YnC^xME6Vzc~^eciw
z{|r`poCHSS366&RM=<&bK_3r3mVs3z5uC`t?o%S5FaY+T62Zv~H4A(SLmvlT3#>z&
zv%#k`)B)f#fwLe#5WJqDsGia1f-4~(2fiA(5y#WO8-SZ|d@%TC;8q+{9p1)J)OH&g
zv{D<~4p6^&3H%&GQCmFEP+x&xVBk)-2wr3;YR63sMeRl50fn(9*bKakWAvrq72q`-
zSA$;%-T=^_f;Slo{V+gZC8&8|^izWR1^f;}%?G2e64b9?^i_hY08^U*>NoHQ47C8f
zg`s{2<C-DR8au675Y!)Fsu$`P)R(Cr0BSLq+M9rW98enrmeTx*p=!WCd!T;@lvm_S
zEdf(LD9!c&rGYw7wdfzXj<m$M5Ctm+&11CALLmJHo4Y#D{7LpIf#y}RQOUli`BNia
zZ29WuVeI=#d%wCl6RXOCK>CR8#321er!z=b(484H=j$#&clhrI?g8|~aRl58=!@gQ
zU^)kou0u8__J?&kj;3>2IL-s(+9Z&kp>d58NdM53_HY~*fhj!lMLLBZ#UMRHk7kgL
zpvM4X;b$*!HW0`025>HbyvvPXN)PGCyTBALAnyl{15ju3K`_OIdXi6p%Ng=n@SY6$
z40sBFbmVj3sleViMqk#`7!rM4@57L<fu{re!q3~_8Nhxxz5_gyA>Re#dM404R6Uy^
zUj@%$NL&Ycf8ap)c^!NZa4?QHgAZXyTmyP8Lp}gLlp*g0AI6YRf^nS?<O|^W4AO%%
zt_OmA4~%PoKsu4GWXSiyRSeRNbTvcby3q?6q$}w~42f$<FJ_S5q-z**D|iWmbSS-)
zA#q*lS_bJ;dKp7*1J^M~x6<_txgETmK{}Q`oFP8~H!w)&(v1xHF}R6A`j>8I$WOpW
zFh~#6D;V-q@Ja^hWqK7ueg<C0Af2ApbO`c3@aYVB4fqU($_1avkk^9GVyFaoJ%el#
zeKteh2L1<w^gn$LLw*iEmqGfVK93>40H4nw-B4e^kY9o?WRRYyFJj2Az!x(}XVjN4
z<k#R!8KgVve=_7Z;L8}KOX|xRatHVdhNOOUB}0A-zKTIQroNgXzXMaf0HkkfstZ7V
z55A5;x~IOLA$NjrV36*qZ)C{-fHyEm57jp@<d5K+8KjfyTNo1KF76IGkbbJEegQoB
zAoNBC>8kp6hWrJ52ZQuheJ2A?P6>S%gLGJZH$(mgzK21&to|2+bUl49gY;TW{R1GK
zPv6fVJy+AU3Xl${A7qe@s~=)00e+Z4I<J0&p(OZG;NJ*K{rxe93c!yuNdMJOFq8&A
z$sk==KgCc6{4|4fV*Lz5S@5$A(vfKmk)T50=NY6g>lYZR9r#5C>CSo+L$wFL#84=V
z-pn9<TEEOtY2a5Fq+9D(8L9*LH3sS0`gMlt2!4Y>dN-{<5>zMfTMW{__1g@U4t|Fr
zj|IQWP@Tc=G2|NX`wZ0u`~gEA2j0R^UBO!!^lqR2kfFMPw=v`i;Oz|69sChPo(TS!
zp?ZKnVaSufpE6WW@MjEpGWc_b>IMFSAx{B+$xyw)Uoq&tMEx~`><0Y}L!JuW!5|w#
zf6I{Tz~3>*p3vVj<Z0j^7-U=MoeX(8_&*Fa0Q@6Eo&o-ep$39~X2>(azcADw@UIMc
z7Wg-Y3WI-V$o1eq7%BqZ#gJ!%|755PJZ&HtWLFu<P*E`Y7D1i^4j3vEjDAm$=YkQK
zpaz4{zX|d@F#0q>Wr5L;3G#d}`Yu5Y0i%x+<ON{#OM)5-Mqebz3&H4r1T_qdK1Yxj
zfzi(hYB(5uiy$usqdyVU2r&8(L0$qzzaglRVDuG&ycCT7K~ST>=o19_PcZrcL5&8Z
z?FsTSFxs1-#(>er1bI0a?MhH%!Dvf@yaJ5&BdBaJ+KeEt1f!h@DhG_VA;_!1)E<C}
zfvF7uc{P~o9Z+#F)iprt2!`qxkUxN_P64v-4b>w+wy&YO1LSpJsxLt0fvJuFc|Dlw
z1t5FZP+b7>1~8Q$Pz7KrGazpSQ#k=uDDZF$gKViDXioy!QiIko#1Y`*7_?qH=y--`
z1fRg5b>2ZIGDI`@BnGYZ4nn;X#C-7g47ztT2xTTnq>ri;;SkbA9}2evIwNnWhcNm@
z80{ONzQSn3Ftr8J2}c0bPk?fUQ8!^K%jsaWQy6s+l!8&_@I=Tj2TuaZAfE-E0_+3%
zP2d^80gxXGK8T_20w2OqcZ2C1pzZ<BW2k?DQJ(N3gk1(+3?RR>mJnVJ91eMRa0AeY
zW3*Ga2|(QhxnR_3cm<C42CoEG;h5524IGK%YB1_3d^C;`CcFkfn*`{i;kCe7NQ=&|
z2QGlT2z()cvMRX4R{+;Tj<~`%0JlN@G<YL$JLJ8<cQAs!;5!)s)iISH2>OAk%s_xP
z2veDX0R1?8FK{2?M83oK0}tSM1(?bYC`$JsMu_5nm_c@Y_z^~k(t4B;8V3G1BSi5(
z#s~_)k28XN@Dq%n5d0K_)-S?THh|VJ!p{KDA`k1p&jBys_+0Rd4D44e!kZY_wOWK<
zVqouT5#9{Eig>1j-(UpjE8#a8!A$U5z}tuu*HQQ#;9VS}uEXy$LIc1bFhYaCTNt4T
zcq_0S@h=8{1bmF+L%^Rff<@p@89^=hGvIT?Pj&nS@FkA#1%Jg*)NWq`-(Woi<4$-7
zBQzL{YmE@B0{_4WR)Tjjg4N(38MH1F{)rKs0R9>H1!<wLgntEo$1&CYAB@lt@GeHE
zpAeA{&<=AF!bWfn6Z9$IZVY@EMMSy-=tmm$6Y0eWP>x7%2HC@rK0sgi84IR(0kU-?
z6b5ZYHf{uUN+8=ef;x?$+!lQxg0Yqmpq?VgA0a>;N3t0jbss4NiV!vqp3Df){*kE+
zTEmDeX9Q1zn*h{bfc_L&%^<xdg8oe)y*P3dgY@{w(ZDf~cLE>FkSI@N4R9*rxgNZZ
zLAp}pGzRI~k<%GuA4JXo&O|&Cd^3aWq{wXyvY#UNF~}Z@+|MAJDe@45Y^Dsw2#|IL
z@<gC@u?(~Y!6v|H6GDKzWT0IL7WI^Yd}pBkY5gn%bx)x6vkcTR!J=L>1~X_)En^JB
z7Jz2~v*G_0@EqVU9KQjc#|SoqDIb7EeP>YGfUO6wXOIq&f$|VYFUg?t0Hlj#pv(l)
zKQeG_XPk%Qbny8M(m66NV2~b@aUp|rmyC-Tq_1S$$RIr<V<T`U;^_|lfPt?fhzwl6
z1Unb}5rAu!^p%WH7^Ev?d<uLHIn~)042yE3njA>)i=r(Eq>o1f2HAO0%^+PeY8YfU
zL{Vk}>Hblafk5_O6y+z7eHlf02xM19Q5OV#6&PhBkZv1AJrc+ci=ti#WaCCVGsvEZ
zc4d$~5k*}P$i|BHWRUF}McD}YVleszfozWGAclScJQNrS`6h5KgY@EPf<ZQSG><_x
zL39d(Y^>;f2H9=V3I^H3(FF{BAGnG^wpz5BK{j0Ucn0bC(GwV?%STTHPC|a}2cHa_
zf@8E{bS-cyj_(GqV~|Y{J&i%Sf0XJGAR8izYmz{^fAkuL{sBz+14x&TUdNy{hUoPS
zvX!GZFvt#w-pC-EKDvQHwnFqKhQ0!PGlSL@qPH-}4vyZ+ptXnSZ49!%GttKgHU?G<
zI~yD@>?E*ekS?2P7`7UW{zb4=VDxDM*<qP|82BWJ$c!*(jUh9GL2D10gBhd~XJ#=h
z=CaHo40|9L{efWj2M=S|gTd${1bl}?WS+vXbRO+OurtAE3xeGPjQ&a>J2mrC0R2(F
z4MtrM$TrPHePv#O<M+T<GH5*`6K$ADZHM~HyoNzGVkX)x^E$}C0HZw!v__J71B2F1
zGO3&Zt*K;E`2kuJ$|N^H_Gl*g1ISLzr0@XQubH?82>7Oo$b5-m4-x{^=3pzq5<r{T
zC1AAIV2$IcV6+jz?gO?AI}{vZ*dlN{h8+%W&%ifc#NafBoe%E7unpjj3_AndiD4&#
z(;0R#7-Is#E&`)35Nt6R;{(B#fx9v6p5X2bTMq8Qu#>?UHwd-@+>2rBz`YrE8W??y
zU?+h4GVEU9ehga+?hg!rpXp%qT>`#lBL)v**oELQ!yXPsUnSTYa0bJcfTIjM9-PUr
z<G^SPf?WX4V%Spf5QbeQ1jG(zH5lzhFwcR}ZUl2A7;Q%|Z-UW=1oI8pFwAB!+L2&(
zfYFWw^D-FiNHAZ5(T)W3A{gyRFkgYujs&v_jCLd#ls60QNHG5fr!&k0V6-8@JPSq}
z5=<kwE5kejMmrJ=#?>seBf&faMmrJ=`b-wukYK4CXh(wi6O48wm^Z*^LxO1nqa6vB
z>I3aaFn@s2js)`>80|?gE5T?_f~E4IO$lZf7;Q>04}#I21WWaVHYJ$Hz)^;!`p#sS
zcff-gW)&EXN3c|vXj6ju9Xym_?gtNJSZb5u4D%>>1jAB0k7Ss)!J`=F2=Hizc@I2>
zVSWIQWtfM-*$i_pIEP_=1jiU=3pma&PlIzAW(7FGFz<u&7-lCppJCA#vkDmIH*g`t
z`~)syn62P(4D%{@Ji|1D_h6V0z!MneKj4WBOZ{;Y!~6;^W|)V-B@9dbx|CtIfhRM}
zFW@qUc^6#HFyDjsWSGanQy7RwWbMT;KZB<-%!lB;8RjkUG=}*Eybr@X0iMn<pMv*g
zm?yzA80HM{ehhOQcqT9l{oy?DY=$`%JcnV91@F%=$AAxDnB&0*GR(Q)gBa!<@WBkT
z7JLZ9oCKcBFeic!WtbDdhXM1Dhm*nc8Rm3w1+W0}qrjC6vj$uRR6~9?cp<}_23`a#
zhWspW4a2MhFJYJq!AlwD0&p$EoC#hAEXOs{5qvnqd<>@RA21(*8yRLhxQSst2U8yc
z%(q~=9su(-cm>0J242Z9&x2Pn%q!s44D%BBNQQX|Ol1bl7vQ5A<~uO80bpJLQ`-RM
zbuiTnV5sjO2cYgv4w%{!FvVc13&2bQQ+)wuBADtLFcZMkKLA751l0>*#)GL&0Ylf&
zdWNCCPi=rUH`MQ_9{`4~x$}U_aa;<%0=Nms7lCgEZozRK_*RBl2EL79&<0s|Fbti)
zhhgaa{R~6rA7q$~;71web}(H7fVmS)_4qW7sr{d2n0esm7^W6X?E#pX;7trOAN&%-
z><8Wqyb3?F!Bp3Pp+5cw@FtGwI(Qqv^=&G^?*UtIOnq}J@F9+=&u#<IkIda*>RW)h
z3ryGQmpHx_yn|t8fxiWQfP4dZC-4)FaXk&`0(3)PC5|v`cW^e44|xwT`oR#in;ih2
z#IWQSZAP#e;Q0(21>^c6*i3LWfV|tmVAS1^r8s^^h@nV>V26O01BXM$>j1uzVUYe%
z^!1^*CiJo3Z-MWSXT(4Bdw|N)8;r6H{S(KCf23g;<Y%M>LXd}r7=u1U(2$QopCf4G
zKf5E)3CHb)*eZYkYd*ulR{&Sxcr<t;a3_wDpIu0YK=)?<TmYbqx+iXz7dp^0GV%xP
z3<wB2Md<W1u@>7P5m+>cPGYPm6vyE0sei~nRe!wibiaB~J*}Qq?+5LIjzQ<3M=&(V
z3rd2rU`23jur4?~SRY&&+!$;Oo((<>e!#nhnYu`q>Q!cJ57Q%U&`E<X8}vgs6z&}E
z86FrO8Xg<Y4^Ir2h4%>`5I!h;Nchn3yzruMefa3`DdDri7lbblUmw0Ld{_91@Jr!W
z!|#T7gm;E_McPF=MEXSrMWT@*k?crOWO8I`WO-y&<k-lmkuxJVM{bMUA9*N4Wpv8u
zp3y6#U&g?UXvW}-F&VQn4$GLAaazVj88>F!nXx6~izw{iXy<6RXwPWB=%DDxXkK(m
zbV0N_dRp|hOxQ%3T{8P*W@HY_9GST`^R&!cGG7`zBx_C9+N^W4F3-9t>$a>rvhK;c
zKkLD)N3)*DdOGX*te3K0&3ZHI<E$^Ue#rW1NVg%8A^Ahb51BNiWXSv>l|vQ|Svs`)
z(8GsbIrQ_PKMehIWN6H|>`q&QtygWmbJw4LR$_lgfjW<fY*8exK%M`t`k>D5L!Cc`
zI)4jql%=80y9Yypgsb!A!Rp{d)cKjg*}>JphT!(#x!~Jir|z!@>v4LrUgzq34eI>+
zFjmyVUBZ3Bk?^o^cDNusDLf@SqovO0hZl#Jhu4JHhS!HL3||qxA-pksPxz_u=I|Tg
z_ru?Ze~g4s=iMR$A`w^Tg}%<4BC8{7BI{h8-`7&-Ju~{Y)cLH8IjHk>8S7E!8#6x0
z_y~2bQ0HAy=Y3J<L!-H<^ZC)L=<(63Q0L0kc_ec%>iiVc`6Zc~21E1CIxg$ftn;$2
z$hx^roj;ZJY}Tf%%~`K!y`A-G){d;5L%N{OvxgM6)Oq!gC8+b|sPoT;emivMNHeA=
zyW>_|FIzXF&KLaor#=d+Qu9&sLhYj8T_@uT$<ztmhg>+fHF!UGA$TNsIl!1mf4iOs
zHtf0)SJNN>Bf=kDf#c}hj^lP<w)*ffJoMjT`=XC}Y_I*O<MyhLy5qR}_R5bsZ9ioD
z!P^hoejtu!;{}0Bw(Z=$VcVbE-`$RDp8mFDZsWg?j`|2WqrdGBf3$J?uiH=Ce%<yt
z+vmVN6xhCf3w~eUcFMM+KgRg7t#RAoAE$3y`f=K}W49f*4I>lI)@;Ysj&r{NPXd?|
zx81t!di<^-IsI*qA-8XAFL%eIaOMst#?KavB3o|Vg0NeD-}1|rpSSGX^4*plTdtw=
zTTXPp594UjR<Y%WEgx;!zGdr{En6Pj^5~XFwl3e&{k==xKkK~<gm@ogkPxr_F?9Hl
zGlr}&r(rDwgNs}aoUSlV2b+S|@cS0|0>2h~5PZYO{2P1{><rK@x*fUcS7YSY69AM(
z-$Cnc7`ZUAk@ft4{gGAO`ezTc^DXMwo^H>#_gj<($Isb!><6uWoD4a%@42I`$ul$`
zwEeR$Sj?Z~lfv82fZrMT{gsaJc}*2UnM0^&8l6L^`%nQPG&giuXoC>g6$F3n?x)|@
zzjk-DgBH;49+J1)+$#UOBj@K~>_gJ#PoETgtP{c4`V76rj?{TN&^q{1KNNh3H=(x%
z-vrx&?fMD*Z~b`isU3k9>Lk)d579?Nv8Xgmj1+mI0Atc5Q7rZpv&0;+zc^S_h=t;C
z%n~caN_^!j6ztIL^z%Aue-<Z;e~5F$MdF|0dU1oeMQjugibup_;u-O**eu=?ABZht
zyZA!v5WkAwu+^)SRvR1FRzIsx4nEPR>2vkz`aGL!N8`QMF|xPL(a-7rc5F!49Q}0g
zbnuM*MZalB1)qmx@RaPMW9T_A2d@YCJgQJa3wmR;z33piim>P_28e+oCPrg6%N9LF
zxtJ!Vh^b;f>`(rP*jFqSi$slBB36mn;&XADI7OT)&JydynPR;-UtA+D6IY6>#J|KH
z;x2KwxEOCPy)IrBuZnl@y_iSETVkj9R(vmhkV?ES@hMetx(LKt(N3HpI*W5fH*taJ
zE-n;Z#Cf8ZxJ2|8mtus!R`e5>i$UUQks+=V{lyg+;x>rE;zlu4+$@HOo5Toln;0%`
z6$$Zxh>QEg7;&e_759rAal05L?iKmsVa&~sihql7;&HKucv6fPPlyTPDKSwzElR|5
zVzPKql!_O`^P)^_64S+-VlVNE*hjn}W{P*k0Z<<f65GUq;zMzW_(&WkJ{5<GPsCjD
zv8WPXi%Rj8SRlR>%f&BZnfOsG7T<|_@w2EE{}FZKC($JS5UXV%j)Z;?5^H38aja}7
zI*PN!4Dq&@Cq5I^;v2ioZnb~dpX{!XvVYoN?Z@^L`<ea9er><8U)YcAcKfCM)PAn!
zsvp#$YNvhQexMFh{|Qb~KjKa5pY`#oDAY65Tg}qPso83d+Fu>04pLvJgViDGTlJm#
zUVWy%R-db{)Hmu&wL?{@Kk?RewOXha1u{^04|^Wo#-6W!RTb(tg`FByrTRlH4np=l
zRTH#ROVrY!eUPSV)iPBlqcRikln;`nvIK9Km*dUzNwQc@kt5|WnULf0)^UG1Kn|1<
znIQ+uEIC9Dl_O*x-b^o)1$Z-koScC7)%TE-@xJ<0_s03nc(;6`yj|WU@0Rz-f8ovZ
zr{vS}S-fTboP1tx!rSLB%Gc!E@*R1Hyi>j_Z<VjgSLEyR4f&?rEMJyy$p`Sp`jhen
zyygBL-hY2zPM06ZedQK8LvEG($q(gBxlPWJ+vRNek(?txmix<3@Rt0icpv^VIb7Z+
zuTi=3TD+|umbc-}`OoFSc+>q5`K6qTx7`nwU(3Vr&ig#ML(Z4q;*I+6<O2CU-m2dz
ztL1;>Lir=!pZ`fNmOslH`HNg4f0aw+Z?abYE|<wa@b3LCxm@9$10|s|HmX22;jQ^*
zW#kdc$`vXkSK|HoRjR#Ga<xj6N2(6;DAiFOtvbnLRJuG?b(U*X7kM1suRmUOlP9R|
z@<i1`o}_xplT|Nyis~)bsy^~m)mN@l{p4w?KW2>q@(eXlo~Z`Ovs75FR}p!(%8>t1
zQF)Hal;^6!@;sF#&sRg_1$blsLN!cYgm?BYRwLvkYNWhWjgtRVqvd64jJ#Zpl~<^2
zd8NvcSE-o1TE%4*-v6)14glAwJbAs!mp7;ad7~;6M~FW~qxfAki(TR<X`qGm2u1`W
zgIwrhd*j{ygRnD1b+9mK4vq+V1-*m5L7$*s&@;$XtJG?>LakIshWZ94;tl>`q0FEl
z$PWrhb5`rrY3f{co;p9&KQu5D4)zIVsvFe?b-lVFG$1qxZ~89|YJw%f(%_I_Zm=k*
zg%19Jx))mEouQ0S6ng-y3RXf7T!DQCZdZ4xC)AVbDeMZ+PlxpYJx~t{zJfM0R%h#x
zdbA#+N7+1^AAF&=={xi_x|2?qef5X>c73&$TIqB2*}-YH&@Q+2rlYZ@ok=t4rnBi}
zLZ-dxU=Ou(O&8PE3^097KhxjN(QoPNuz$ge*va4qeWkuvU!}j*|I(M}*Y(5tV*Q5x
zQD3CD>&Nu9`ab=Xeo}v<zti{Y2lPk!V|};2NAJ+z>#y}r{e%8hU#RcYKkMuDPx=ad
zxqeVzrth+|?IbhMR@*AQLEmTw*+VdE?{5#b!|Y7`u6|p;qd(B^>G$<UeXG7rf2F_B
zU+P=*&H5&Nss3Dlra!gw>_N8DF0o7PRJ*s`#}2hccDS8|xxc~AuoLZKyT}&XGP|cO
zx07v!t+UhY1iP25wbSjsc0aq&9&T%F33g)`XBXH~yUeUM&zU35o8}v{+3YYcn=j3a
z<}0(w9B%$?9x%_EM)Qcd&pcxqtTBI@H%yZa%pc}8v(k1jyRhd?JM$POym!nhn`VAD
z_uG(p)GG6~Il{bWe!wm__nIHg7W1@OVcs`8tu()xpUhVCs%bVKnE%-J=2!EOwPu_7
z#k_02$Bsb4{A@lnZ<$Zb6XsL%q&dSJXU;R{n^VoP<`{FlIoF(H)|!*diRJ`zvN^?^
zZjLfXn>FTa^AB^HIm@g!>&%7b0&}MAXgk?-+u3|<J~G?Q=jL1UwfW3EZ(cDknWxMb
z<~#F(dEFG6Vl&B1G!x7oX1p0^O3ZY#ubE+rOr5DW%gk~!&(xZkX1>|a9Bd9W2bl`9
zz*L%rW|3KJYD|@>Hg}tQ%)iWCW|rAtZZbEUTTIksn%<_5$uMCP(HUgzVlH=oe_o3D
z8!98aT7N@%I36sX#oYTeW~*CdfEl<Fv*vnfFryB^9QCKD!j2*#%&VWvc9^AxiE8XB
z(gC{2bF!m&UUrh{m}f_bpJiw4KGIc;#116g#VFZBydZmG2HgZ5<t5om_LhCbXxUec
zk^RJ2KMxGT{FyCsFt*Z+5QRPz6Pslww1Y+BRhl8h*?0%?Ht0Wb%&55{f%#$>o{r9w
zBQWFshIwU_94+#(zsXpcEpwn_y(VMOE*6V7WL)NA&cWP*S-uFfOCilKnCr)jFXedg
zfZRjuAt#7WFyGJ|Rf0KdGG_c5Sq6QkRxF{}3-k0c*pX8(|CfthsJwcbm!QFYfmu#M
z2U?C<Wglp8jdHp;9J9&{Xb4lqFPK+m%2{G>%qerAE4?Y2<o+1{R>%WHGv<|pFozr>
z_Q4EtsF;rVW1iR-vquGHj!I~CGsJ!vL#r`&ERu_%8O{`|WDT^0qo6G;kxRvuvQ``o
z9cGsJ2Kv?<%p-NOUhEHx;A8Q-JX|(F`#L~2Vn>%|c?7hRW8@04MjQw2Vx?S#eO-=}
zM?r5nSRRd4iDTs&=q_``SMoS<zdT+XDo+rf$`j>D@??36Tq{qN>*Q(J_2mrg`Er(A
zFVBWXdKv8MbL6@5JbAvnKwgL)VlKuGF_+4JVlSA><rVTud6m3cUL&uS*U9U#bIgq`
z^Z2cpL1`|hnL0V2--{Xce$3GiVn>;Wv7^kR^560?`M7)nb2iQN&tR^nnf?XL^EBJj
zoc;>t_}Ba#|CXEk-^JYiKIZi;nA1PRe7+rX`Nx>YKgAqQJJWoLx%z9&(>pLne~0<`
z2h7j^!QA{4=H*{7C;x`|_z%p*e_{(2iJfb(&m#7vvDlTS9p>IN)j@Ssom9H&tgz`S
zcB$#EdZ?bNm+Fo6g}$mER(1xcfohNntBA@_QI)9%t1LA{4OPRi-^~c@b~8$iR%6sy
zm927AOvP2MN~k=y|4kuwVHu~!t3A{NHBn7c#i~S=s>!NMm8(6~6t$O{s`ggX)IMsu
z+E>j``>B~~me=JDaM~Q{a&w{m9R^)(KJ>H&(9x<`hf|AH4e4mCpQ(DaTpg|&RHJHA
z&FTm@tJ3Uxq&i9+t&UO0sx|63b-X%3ov2PyC#zG`T6L<MVNX|Qs58}BYP~vJ{X?C@
zGp@QoU8pWn7pqIurRtyRGIhDSLS3n@Qdg^M)V1n5nuFb3e3QCa-J)()x2cV8j=fXe
zrS4YusDHV6nC9aL)kErG^@w^@{TuuLJkGN)R?42iy3KRydG&&NQEgH$sm<zT^@@5`
zy{2AQZ>TrbTk37~j(S(Ur`}f|s4Z%%`cQ3C+to+vWA%yp6x!|Q(00Fse)}~v+#O!$
z{Xy--ZbME(hIacaG~3^y)$StQm$cp>2y|cq8-%b*2kkcKfVH|#K|0osx&&Q=ZccaU
z>9mwS&{O&a{euC)z+g}i4kAHD5DhYe!9f;QnT7_#g5gff85N8U#sp)7>>wwI1#zeK
z<UxlifEH5}j0?sGdju1LiNU0x7%Nky!DOrmmIr$VQ-ZyMsZKxICzu}W8_Wpy3uXqh
zg4w~GV1IE*a6oXN)0{}RIW#ydm>0|sDuM+;Wl-g`sYNZ?Pi?R)s0-?Y<-y@WL(mvB
zIZbN?be~nwevS-|3XTqr!5ZY6;5h8<bpqDKP6|#APH{UgtqV@WD%lxWCp!!4lV=D2
z2+j%44bBVB4=xBU3@!>T4lW5U4gMKi7F-@&f%VI)u$p#FaBXm1aD8wCRx>vQHw8Bb
zw*<Eaw_#=T_TY}-&fu=#?%<x_U%|b>eb}$>f#AX5q2OVx5<MFHJ9sR3Ja{5_GI)wC
zRjh<Q7d(#@q!+Q)@lvoEcIqp^tFTvJ58epg4BiUf4&Djg4c@~l(+9y8SmPg(H6DBv
ze2le^PlL~b&x0?5FN3dwuVJU}!1~^I!S}%rSQYtC@MG{(@N@7>@N4i}@O$t_uq*ge
z3#Z2gT5AJ)It1IZ2W;M6usi#}67C0Ea{%o0_Bu^>&>czt)17q}-Bowf-E|M$Q}@!n
zbsyaqE4Tewe}o=MvvCw!<Y1kphv=btm>#Z2K%X4tG|I8iEpv2C$8|1N&GK}<F3^Qo
z^&O|jiwE@{VyB*<C&Fs{QBTsvx&*7ilXaOc*L&(IdM{Y$onf61g2k(04M$;VKMo7L
zlb#Aoz8k(Uy|<pG_tDe!zIuk<PtVk|^lUvx@2?Ng2kL|L!TJzA7w>d+faQ6Zm=F7R
zl0Hn&)AMzOUZ5*=m9EC_c8m04U89%irMgxx)7VE+*Tb5>Q!m$t>jvG3Z!I_JW_^TS
zp;zivdbK_hmh_X(+P?>O|BbNBAIACxcA(Nn>7(^A`dGb2AE%GkC+HLPN%~}cie9Tv
z)$6csaXMC5&eUh=_4;h=$afCbG0xNH>kFXSUIcCU66mD=gjRbw^xG?;DPIi@{95SG
z*F$%{5&G>-&|7bDdhAAM<aa<DzRT&z|AJn9pVOKj)DJ<gendZtwVKED<5;tK5?b}s
z`WgMK2(U)-ynX??_$G0pcu0iwOL{ZD@B6ZTMZYRe!WR@@gI4~AeiJ(T+t8ceb$as$
zdW+twKZHH7U6W4!iT)H?^XE>F{tBA;H_*_(h0gvxH1wU&^nZkQ{xkL}`xPrhJFxO`
zFIF?o!YW1>>l^1{)#rMw3tfYCqJQY$^zZr)y-WWoE;Rz{CKriojT9FeWdfs(fprnW
zD#=r@G1|k*=m2}8le0C*zUXGUn;xd8>E$%SzGQ2_<``%Ok!}bLF$!IAu*ot*%uwvF
zINXddBh4r?+Ke${O}5D~F%vhrCSmeSzA1pdS_Cb1Jhasb&{`)!do3}gX0j<W<z`Pa
z#q4FKn!U|5vyazf_k->_%gi=&%>L#8r@tO-4l#4hq1erF9<<sDr_EMDi(Tk7+9hTw
zG~Q)S|6L9{paC{O6ZG{Xpu?{;t6(u4>2&U6ppmbEc78lG@e`qulQzB<dipx(;ip4y
zKhtUB(96Yr(Am#*TKomj%`Y++n@h~4=AY&=bGf;~TxqT{SDS0hwdOi=y}7~M*rG+>
zYHl+d&F$t6b0@TG*0S$~HF7^JlLyU1ur3}kkHSuQ%sdWj;Yn!qPs37q78cF((CuG@
z74s6TnwO#RzY2@xbyzQN!m@Z9*2=rEPTq$FvIVxthp;}j!$$ZRHpi#15I%=(@Fi@Z
zuVEAIfYtCF?2;eMPV*mFGC#pO`2{x1Z?JX#fc^8Q6;@hh1FNmE)`o05+uo+x4oU07
zcEK)q-E4Q;!}hejY;W7g_O<<Ne>=bqw1aHeMr?+S+Dz;!orRt9hQbycZb#UWb`;hN
z$Jntp+veDqjbr^VLDsG<fJIvb>utQ<1D4%HI|=q)2`stEu=>hj_f3JdHx;(wG+2kz
z?Y>y^Y>ySZld%@K4y$=5*cr5*W%m=ui^uFtSdO#p9N3@-z!E(OmgpgNF6`pNU}w&^
z6?TEGv{km+F0_kaS=Knav(_$yRap<K`f%8sjkd`)+av4>yV9<*tL>5YD0{R$#vW_e
z*yHT+_5^#PJ;|PIPqAz5sdk+`4eQfq*fZ@}cD+5@{==SQ&$Z`a#rgt!p}ojnY%j5w
z+JD;1?B(_fd!@a~UTv?j*V^ms_4WpPqupR{vNzjX?5*}TyV2fm@342;yX@We9{VqQ
zue}fJ-VfLZ?L+oq`-pwi{@XrgAGc4~C+$=AY5R<Q7VF{9V`ckAyUD&}H`|x(EB00U
zntk2AVc)cG*|)J;{w`L$-?tywEm-UO5G#J$v99+q*7-igO5f*L<@?gF_wBIX+VAZ5
z_6NJu{s-&2KiQw{FIW@&&Hj!x!d>=HtdC2qkOv_hG9eoZh1!MMhtfhFLLEb$Lg}H-
zSdo1XE1G9u_4qBUTD~jZ$4cWnSWVs<>JsW2>PGtpVzsbWs5e$w`eK*F{#Zd7h&7aO
zC=$vDMMIgP!J(|skkC-9!3+<L2#pMl3XKkp35~^Cd`>79iidJTiBMiBA1m^Op`y^Z
z(D=|Ep$Vahp-G|QP>C&G&`^Cub*Ok*MP);MU8uNzQGH$Y(ok_(MP+kSb%)Z*nuf~e
zWeaPoSEiL#)i+gCR#w+F*~yg^@C|CHuV}JmE)HAfglT0h-a=(Q)-sN@!=A0Ng_`SX
zvU6gwv^`tA>nRH=8ahmA^`rKhYWL#&?d4<Ii(|5THPzHsRd<-$>X{>q=h<m4=XRRQ
zjhB|_X-LLS=Tp;DPUR+ovbsffUzbd$eHS;^Evjf}URGPt+!WfE6Ya2XV{Jv_Vh%gg
zQ%V$xJUh!tJIvZGJF%Ps%0Nz8VVRzd!r9p=!SZ4*bFrLaJBL%B(<b#fzQ}X9$YxGM
z4Kg~%72F=c2_DdzVEY5AYN{Km8*3Wv0gD<cj;QW%Q0qy1kQ3R1U9hx+Tk1T$)JewF
z)>qcfn6`I1*HDXd-?nGDm@Q{HwxrWma8p(Id{_9cR8h@&t!&LsT4hTHY?Z4LTkQhd
zYUfI;ZaE*S_R&@QKD4Mc8gATJJZ({ncfFYU(Bf7<w#G%=zQ$#9_RPIi&5}@!kF&=2
zrW#kN+#K=z4og~ta}x3Vj!W?!!|J-)in^+rN?YqnWNTAeGhV7|kqcYLg{`YtUf<Z%
zP``X}wXI9>mYWw;qZRA9!Ry;Jcs=K;Lw!njYv7a`Qld!Y+eS{LakqTNatd`5b+e`v
z?|E_G%}Q)Dr`p^m)n;E8&Ayv8yKdIZ-E0LXw4ybk_A6St*$VDvt6ERmRou;1yI^Un
zTbd=k(n*qCC7o-j#aZ9>EElikEXS5~rmY;athu(SW_j(Zw3;!E6}3(CYU@{|EonWh
zZ8>bWVoc-mipuH++c2iFF*~PoYuGVO^^FVDR?NfynA(~}6;w&F?EG?9S-Cmco^w3&
zwVIp5m6@CExiIN3*ZWH_XM1^eGJKvtpYPd?3)#6jh2H;U=7M5>-g9=Tcb9lhy1oCL
zGB5Y}j`?(BKHZqlcPyE2??2}A8}s>%`FzEEK4U(Av1I;|@%!{+K7TQvuk571WWJO2
zkW9ztH`~Xb?bFNl>16wOV#)mZ^s;^a^L)Adn3<dB<IVH&<@tE>eENAlpLssMd>=mF
z$CvNJ=lk&aK764+U+B*l`tya!^FIDUAAg~bzu3oL?86uP@WmWHk*x1xA5XE5r?|BJ
z91If`HBFVPI{PCUHkMXbuHu$0@iCV87)yMPN_<Qu98)6MawR@zB|c}RK4+!AN0j=W
zQ0iN#)VD;bZ^2UE5~aSyCj0P{eLg1pd`$NFnC$a0*_UXtPk*vczs!d(^Wn>U_%a{9
z%!e=Y;mdsZG9SL&hcEZx%YFEAAHLj&FZbcgefV+?pYT^=!uO;^Hiu96o|y1GFp=%+
z(O+?iY#-i_1c@9@Kau0Z=WzN7p5e0-IX-+2r=RdwX2M^QiR6`7;PY4D<M&ru!e3E|
zm=B-465_u6ao>J%Uk`ELj)~+*lsxb2C+^!f;ln4A=Y74!eY?kf{l$IzCz2zP?+0;T
zuL+;;cv7Bpd;f9YUU7fr=K6GVeY&|m-?_<rd;hsUzqvlYxjtXHKA*Whf4Ry0CFA$$
z=lcBR`g|pl{*w7l)<ZHKpWlR!KjG6$_;eCJp4?>qe0m9=|2$tFKdt8G`S|mEJb6C-
zJReV<PcP5Mlkdal`+CUt;q!g?d>_6r*{=S4p+8@kJn!Q#^zj$^_=|n~#Xfv7hfgF&
z`eGkXv5%+N$5TAHeRIo5<d1e6iF}MDK1U@!rV<}h3CEO3&Lt&2XC*#orAhWRS?XJ)
z)VD;bZ-G+Za;3h7OMOdB_9dI_^D)`yW3tc3WS@`8zC@FK`jdUSWj=hF4`1fPm-+Bz
zK75%EU*^M?`S9gFe7O%_?!%Y+@Z~;yxes6N!<TdTgx3WUUUx`jbNGax3ln}0P58Mm
zk<H-~**<&@r=RdTNFv9F&*AhFUYAJZ_;$$g;d6X=uM;Gax<G-?UqSLb=Re`+^hC^u
zkCnGOsJ@BR-=(IwcKPB8T~ggtVfL(8wyeU<QE|+cWy>3DYU}It!8jF^;q>gqIMRE;
zVl;bKEMHy$vuxRdstUDFvzp$l4y=K3RYMl1no$$XTwHHv)ht?85zMY=4$a}Hf*Ffz
zf>K~cV+|)$T$tw4Z>q1WZ|vw(bdGjN(H*p>1o^y=pB+9<=SbUkS@k0CiM%KE+0Ncm
z(#?r_%^~hJhq%`m;(p$X`*|<!=e@X}_u_uui~C+5_w#4m_ujalN8^4zi~C*__q{0Y
z=d-x)MR7l$#kn`-l=Hj=cC7|>R}I*;KG<D7VAqOZcNKwMYv+`+-T?OgS)YL0$H#g^
zPI-a5D#2dP8Y<jA9M5NP`*e6-gWIR$E#PwQojK(^PlA1To-g6{;dvCuDQA5or<~_U
zuuq@oNw|G@o-g6{;d$PK+o#W?3F7nqc~r?MXN?Z*<1O^{SLpq*hKKV$Ue@&B_VMz(
znN!a5CfJAP`4Mhk|8XCl=SRqWc<whj<vdz}eLeFimQ&929@yuP=Q+52{&Ic%xjuf@
zgL2Bb|Kyaj9+gwh^C8%W<M|zKpAXi%;NPETeGG1&E^F2~<vg!}eK~o4gWKnW=Q+52
zKjZlh`SJdFRD|36XH7V#oaZ61_s8=Q+)01Fe8oQA;&M~r2B8HQU(6yJgiKX+ZBs?4
z+7CgiaU!6TO>PLPp&@9g8-i-x5X9GYNvW!9Rx4{l^*jy*4U6mfOlizCy1}RkgHf{|
zjFw{{ss!LP^)!C8r+9gAp`+xW(KXe<r+pO8(b+qa-h1rk<`6zh&XML)<3m?)_Tmfc
z9PLn8ci=}Xo+sj7!-(hS_oXGihQ{iu+UB~7CM*$QsBK)eY(af(Z&>I}e<KekL;RQX
zoRXh$<9^1C`x!UxXWY1-apQi5j(b}o?q~A2pTXnamWX>@KkjGRxYzmPe#VXa88_~A
z{<xoM<H<h7x<1A@R|8<DX@K2~4R#s_*v*_^r+I+g44PBUV?Ef1XIl$nqSIx;Nx6@Y
zZ8OMyyanE0f%jM7(=YJ;3zGhP`UO7y0-ru>R2U0=JUrvW?c?E@A8sEHYXHcvKhG_W
zvCoI+86R$co@aizeR$Rh;r97p%?<JS^Q@a<?DXf^c7)r<#~K{QM4w)vFE?A!ko$78
zH4V2fH(S+k`*O1lim}nh!?q6GJ{`7o;P&}q+Xil*FSccnZ|`PZ8e^(=bH9Vz=a1)N
zxPAV(e`0L)>9frOw@;sK7Px)-Y^xwWKW4Mejxp1pXU!60tdEavFSvbtY_q}b{U?0A
zuuTGavOoGUF+Vvr`|_}Ef%JTN*tUk-mxnc4jKw~Ewprlz>9g$uw@;sK7`T1>Y|Eg0
zehkh}&N2QxUym5eeSECLV9fUEu?+&ZKhL%Z+&(?FN#OSJu}y;blGg!i{PFzaZj}{v
z7+Wjq8dufURxGHgY^?vQsKPeqOiu7b9LwR09_%t5%i)Uz>?Sm@t5mQ*AIojOy1Jo$
z4E^vCtqi8O$lO9<JKPlEgQgYrymZ*K7#guVY`?I+xxpXR9O3;n)~sZIjTn3D_^`TW
z(c&idSXbjC<b)_mmjfDDW7#~AfL#NFT}6Uj1A|>ff?dY}yNU$63Iw~32`=OkRn^rm
zb473mR0KLm<08-@7lDtc2+kiB!TIANxI-?2^XQ7;j$B>UH7{#$X}AMQgARNebjWG&
z5vAe$Q5w!4r{NAc4d>CNK}TFKUK!8!CTF%^9Lp~8^}+)emT`T(@PGifuNS^x;r4xj
zm7rJ-Uyxv~*Brhe;pTeH;R_OOuGbt^s^BhekB;H;ySSsPO+Ieyun;Bm#~oeKyo1ZZ
zi$}5S+~UsFD=VRvRMoGj8&g@)Sl#vT=IX|#n)<pHNyo~17*=EGwp(>Whl)o0)-*01
zQ(M_-IgaY9{0g`C#>?GZ8mpnzxNt1!#L0NSos)?q-7T@Q@2;rv1yC4j7S$zVcef<G
zqsxMd%B87-^b|p3c3zKGXVc=E%B8gQp4X1E(_HC$IA49)Y%avJ*<1m$JfB-Smd{f$
zTFm>;FH2uu(SRE<)eD<kg>`n$26tsq1*R{+EtBfj)6Us?x(m6NRyVbH=;{TnUMPd*
zeyjw?T@a3utEXFP94zYICKi@aJlWaBKI1uww1qW`nj5OCa4XT(K^dD0v9fq^r{=mU
z)OTe)PA{nKuy_@28*zW==p%7u?ND7uq5bi)mgDw50Q!VK^6l6ft#7(+i<)a{Ypa*l
zw^l&+7Ex;zIB$)(-3Y(Fo}D)*YW0@V6f_Q`HU+u6ac1TvbBDVuecOmq&-C(VDsWSa
zZ^yaYGOn&&<<G;7Ehp-cQvEH$0jUizFQxy|y`64+rKBd86!~k2eD(Fd+K@}JPaDy^
z>cg8WYTb$6ZBNuzH#Tw>y}ac*qxideVQcU%?wXj#R|a|Ml_L0W;=crM74&Jj?B}%{
zx%_nF%YU991zc&{h~{zLoae5|Gp)r;m82BEN3wjW!F#m`lLs!=?nzfl_+H6zXI}Ea
zc?~CBE#j5~cfM!Kc`Qo7p5boMy+z7HlPhFvN0`?-Z0vr9x(;98bf$kwn3SHzDW{xE
zX|)t-N)7i(5hn}(msnaGuV0Gyl<StVH27~%?N*}huGnpBsvFBvtEmf%TI;A+ieO%=
zxCe_{hjPkD>&TwsGi4M&J{wX;0djZs7xuE6s;b)RF^$!Sckz<Oriz9p5~O<phBtEh
ztH4#wyw+g7+w{U#dAF3>@iFvF5w&{kks@tL+gBSWmdpT`efJFXOcl38(3hpD=X^8+
z{(7c0GyVU1vL(uHt?fTAS)vqCYa~5Vq%EO)`FctjNihNEx4o<(DQH`v+!IOeW~Y2!
z2#V#G6}GRat8c;uQd8lkw0OQ#&tutaHiKOVOm$BEhr1-rX|dV)W11GK(wP*hyo?tv
zW7);!x~#dut+m9mc@+}uXUUSnPF3~L^l`i1-M8<MbTzW=9*eQf5R0)57{enEF1s<W
zIK+7!3365eW4vyJaL$$imxNH1WzeR4xZ-qIa!hRtNuNq;nx!2et~sKjwz{sex<mEK
z8Yo_MO*Iv@9IUvo9UgVSGEQAnS`9QaIBFNP9yTp^-r~Hv6^rv~7nl{%I9o<=v*H?O
ziwO$5v-F_uFvq<ASg~8Ji^aTRALG^7SS-6VjoYTRKzXtD5H&V0Xsq@{^j1Qwcyhby
zWy_mZ;Y3<({i2%63VNKSy+5pP=tu>@GEpVzp9YPF=fS8UWBy)gHsAS)#d&QM%&ivZ
zHB-1bmvP?_;=E>xQnAUEozEPL=XGqTrbo&+m+46dCL!dl6V{%5o=^~9k3J>eig8{g
zME!9q$9=DevxSLPj&)p)hi?#HYCROUYizDu+_bo&DLr|XljLIh%8#`~!i@tpy*-u^
zkx8g$>@(TPiVNDKf@p26rC%gEp(EAQ)ikYgYFc`7?DhJ)QxW($w7j+%>Oy4)$K2mL
z`m4}&{+6)*T5w^#W|w+#JI~#mgdfH67f(70M@e~SsG^~venm6)%$&mfH0O5OoGVpM
zVL^K*BGbc5aP3nd(gM9Ck0lH0n-+KQ5w`m01&~+{FM#05Hg9L~0$MDG7tp|L&*pf$
zEuR+>V)?wVfbe{o<oGcN&(Cq{{Lnxw?gxW7FJ#3^xPyW@TnR7mAwC|F@_9iK?pzm7
zJ})G~&GubBFFe5Q$39*}hnqWWJ})%F&4Wlj-;{xyN5cH_^3DtJOxV(yWQG(|jk$CZ
z-hNMb`#n(*Z%>s~f~W14dLP_@6W)SP6y%!)e0F(?FJ3^4CHSTanA1(*rc3QY?;jUx
z!uQ~Wzu*!WCK{Sj&KD+NUshB#RMa(1t6x#wFb(b6t&PN$pvBXi<>*4>X++Xfi=?=&
zswIl#u?sxtzj}42|28HUD_34hv?Kp6roVc1;aW=3dcJK&{*T!Ia=J~KLbXkJ{<Esu
zxZhCKQf3+)6Q$*ruDV(t8WMcBIF{fwM=&qpBzVmcZXQ$;yfz6pci9B5g~t-Uiza**
zP53UF@Le>)_r;KGS^Jv0BT`f5J4%R==Z*xg62=m|N(ec3-UP4G!|gi}uht_*z8Dg`
z8XHR#m%(C%0a(?58G(DNH)a!lI!dtB9`g^yfcbj!55>e1d=o6@A8LV{ufGJ}^n={H
zdEq|hAJT~>c)=3P<3z%*!zBE=OoA6MV+meZ1^Y_mn_F=EqVi2IxP4LmdQT$3*LA|L
z_ayk{7Si)ozze4c=gY%4xp3a6!#Bm?_W9=pRQTs{D&g0O61?CGxo>p7X@+pc?YU|f
z)UWh4n=^35;+m%FS?Kc()d$oxEuL9V%3W1yeM24Q7w^l*<7cRZpZ*el`b&6qA>q}H
zgdcwse!fchX*1!+*MuKm6MlS6@XbJUnUctExoNrX_w(b&?jBkOgQSOy-C}Sa+-3IP
ze6(J=Ngth?8fq#QH7}nH!eWHK#7goylAHuBQ%|BayHirwK%;QT1+87!+YQXEQ75Bf
zd26PU@?Oq=OKIAQTWTRG=4Rph7U=%IyYhV$bZ6h~_z8?yg6}J!+q0gO;QKpp`)=cP
z*aY9Vfq&oK_#Q9ZKAhJM6MSC;>H2QL_x6#lA9MV2XTmG&{s|MD_alVY$NZBhko)xf
z@@T>@Q28fHVhMga7U_9^d|wUu^6~LKd6dWJhwqcX&DvYS>+A`>CywF3*B9SAM?C(r
z<ohj1&&R{}x8U~W;(Ool@9Wnu?<V~6Z-VbzpgcUC`)B(Q-e2Bceell!LhdJFzsQ*2
zXU~vtA3xt`LVlBOpAUb_A;I_Pksm)C@O>os_v!QfHPo9gA3qxix6cpX>xX||ul$Sy
z&ij7B_wV5L`Q-aj$X_ykKTPqn1(17x{uW1q@6BV_@%iU_^C&lm%k#r_p6`!&z8~lD
zeZ*Lv@85ZRUk+|=w>;nf^8B!#=i4{W*H@k|PhPyRedDqkD3Mrafst`Uiz8@kuCojA
zgnw<drXK_CwyIiJ);2FNi>t9Z5W>BmWmpV%*I9l}yc?9PI?_k07FAc)*WpQ7w#VWH
z`MfGkx^SBlern0diMOYP^<`Ku;@EO>@;WT2uB}b=o$GsUuCMWUK}jdvfx_yL_tg^*
z*3*zwJ+r#0xuGsKY&<*Ov#O%0Vp>hz(z3?Nisi|Cbi(5a)lErP4|bK;H!PdIqTWZ)
zr>gphn#$@@E>SsdC;1b-+nm_9E*VSTHuBkYE9!4Uv?kGQQEmMKH?iykGuPcx>xQ+3
zeHZRiT~*V(%!lscM6;_`HhDoG7C>V}qct|dB=jc+{^f+zSN+-a>N;4@4XeDfQ*~;E
z^uXL#J*yfEa(HYZS<=q1u9lToS6BHUJr_09H!t7am+J@df`a~w>l<oTqq-_;XICs}
zuB{-G)rT~T>zf;^ozEOUp~d}(9QPws++UIY>8x1XUrBL)<>YvsD97taaej&^7B4O9
zT#Gw*Df#SIM-P+11}y$d&rGkYUO~QnzB<%3x9x3lZ??p<6TP5g)i+e_UflG?#Wf3?
z_DYtlvvcm-oNSzq&KOKqm|Fnacj1(3EKAfa^3rZzn$o0Q$tG^9X~OGW-cz4eN!xze
zt8rCb<zhUghZ@}rt9RZ%*R3};ateyE3{}&#d*1uPEN-GThQBmWmx^U%c4O_qTc&BW
zw9kfZddj&*Tv4sg1<wAadr9$71#H^J>_mHKIn%9%v}HAQFs9uho6vOFrzP~fmLs<U
z+Z_qcbIZO?SKw9Iqzre5TBRMG)Zcl)5QAV07p6~&U|!PF_Y0-$Ggat<rwY3B>Q<YO
z=~Sk9yflCn*X}LOR#{gkOO2?DD|70~dWs-L+|36{5%oxlQsW!2o3Kp|2hlo&zm?!`
zskFJX;!Y<Q<6SlNYa?nCStoZ2)#yBSO}bj{>-0>O{Eaj<*Ik@nzhX$O&?iOECe?Iz
zqNS31wK&`O>g>eaohUlaU22LP4|#Lh@9qM${5V+;=S*3yp_m5jCTx>zc=waL1?FtI
z<ztsZZ|9$vlT$N?yDQ0CX>HmkMU3l+veP?7lo~_76nX2R3leR0<Ca59I;o*KeO{4D
z?QnioitZV?)hWsWU&(a4z+cJkkgJ{!TP`&2*C{LDbOr7%?3uhQ`2=Pvx~<T=t+Zvm
zyl+zemm3u>p1(+Y?dIP{-;%}M?{0J@|7~*Oh98!+<f0q-YI`39cUxLSE&g0?Q<nW(
ztHKWgt%6Q%679TOf)U?V$-eiOTI`ebzyVw19hf|k<TmMqlhP#bmTvF=BFBX9ZsxVE
zs*I#G$$x2%zW?>ipyZh(x2@j4RKV^H*HQtypXl}X5%MtSuEyP~jEAw^rF`Y41n^y{
zuM?%nTUx(ciqL!J0cN*K?%x{2Un-18yT6Eeuyo=4IiEmEv39Qr9`M?F=c4ZJbGMlX
zvv(|YcqMnAznKkbn$P7JQ^s{C9q6+9U&C;Pv<^)aCcn)*2|+<G_LlNX{t7xQYi=af
zh3>7SryS$q4AOnF<NoclxPOl@?%(2x`M2+4{;j;2f2%I$A6Sa}TV?U={Pylm*~<E5
z3);JvU+IYB_V0hh{Tpj>|3;eE7PI5o#qHh8xhWx>-H9{K?gP7Zg6z0|QzY)+S@T+N
zHnd@=h4hkPBh=`odMIkO6|1_k2+tZV@{;!ME#fhF!`3~8g5d6!MyKS%%Y1{Q<sDz=
zOmoHJLw>2(J=K)@#<4rundb$%3B;Z`E%Kb^-cokb_Vg|?pLa!ehaLEZWp~`M64ndu
z7dl6$Hg6HbRD_1>P9$F)=KA$-qQ!F)h99eTzZ^rdXf2OZ<KbRcnP9hwfrk{G5simn
zy_=U);r4Pi8Q}Ke*p$G-nchFY*ogCP5drdYs$F821<E55naslS$tgm9m9s3HO<J&v
z0qo3iuv;!E^Y>xF{yg7oEX(G*Fkmm|yD)IaL$1mzS9SF45~VE(X?Majb*t$p?Bw!k
zvsGQ?s)kxshw*Qz!prbYbz;csn!3fRZuP2Ft5ltuCen(__l~F@#Rr!~<kr&S*%5hj
z=}i2t+`IVTh@8C7zSAPI3w9cai1%jhyH7;iFdKeE6JE@ZU@vBY{o?U80i>(2<FLkl
zJe{%2a98Y_^f!AxxjmM`e+kp+e+bj@e+bjze+ZNIKZI%jKZI%bKZFVW4`J;85XSs}
zgwf~&)E9cC^o9SsCqybRR;m3la5u>lX(*A$kQpdfF2oZ)vX1t+kc-_f-4m;obKUR$
z?sta!o$7ukyWa`!x4`|zNC}mr>Opb{4D~uWh;|;3edw_P*$s0>j_iovn6&tfi(UB5
z6+h!QA-=_Lp7;#E`C==63&cBQZ;RI&@$!e*)Py+2GxP{8b`z(bKhXWAyI*_>!(9vb
zm9ufIu%qREj`zkcm*k!z_aZ;o8IxR-u~#O!O5JaX`<>-}i}9<lrx?DoffzfxU#ew^
zYm0ud+dqD>lRJL7*6|Gh9QSm;{oOBpJwW>FllG_&u%8+2cSTZchlSrR_+5mZ&m!2t
zl=fI3gS|v!*lE28yO<SYFV8)(7w10MrF;%{!JLa7h!$Xf(HiXTxeWVoACA3<o3Xp*
zk>Xf!0`^Ng&Fxlx4))->0Q+oRiapD(z)ot{U~lmouzT4p*n{+T>}Pfl_7Q&oJA*xn
z+Is>!g<}s?><_*Pd$PYGUc<iNZ(%nb+MWGF?Bn?{cJTfJdw}o2zTZD!|L<SKZ{iQ^
z)gDMAL$V|GP3?pIum{Kt>|#1b=3uwdeC$t3^@g&^Inze$rKdbG<)|q$rVO6aWzV1Y
z+_>je_^;ZtWY5T+9n0S+pHw~y|7rLiRrUg~wydsfHaNSi*W|U6mz2I$`f%w+{0}PC
z#3gG=>Pj~je^LAveoroLDvlO+nRLaZf6%#!yC!}+aU1>*n>b@)$;6TgPfWO{bmN3E
z6CxA0?Q!8A>n4_re{lR|<JXVxG49uKUyYkGu4vq-aifZs6wNJ~UNpV%vBDdQrWedC
zm|T!oke7dA{*r>cyr=SR&butnCVob!^@+8Kx^Z9S?#kVodtvVC+#$K`;&;W*k4NGv
zwmLR9=aZZlavsi^o|BvNaP||~_u%;9Y`k)qJtBL=*!#!cJoYmDN5}RUi;bzq+%V?i
zF{h6iGp5Jr$41{Z?yJ#rN0*HHcGL%>P9N1U^7oP7j2kub(vioH>^S245yy{MJVFou
zV)&Zj3y1waY}>GP!<G;IWawK((}&I<I(^9RL$+mY%X+OKFRMH&cksP~H{@M5c<|t~
z%zHDh%goM<j$a?$9DO)i8=V;y8J}d#%_xa{8~Gq|dZYp0@38ou%EkEl$CyFy4SH<Q
z@<IC#(gVL3c;Ucf296roZNQ@gt{X6AK+%9v{om>TO#dVL&+VVq@AsUC`)%mAzTbp?
zgZsYH_lfKgeW&yt)91ZD&-Gc^XGx#Ay?6Egw)eK)wY?AS-LBU+y;k;GGWN2b-}ihC
z|MPp6_xPsAd)>Enf35pd4wrQv(K$W+v-HjB_oiQwzAn8f{jl^Y=|!Es>9o1ift@CF
ze7M6l{%`j|yJzs_gG<}3ZP(Ope!FS-EeKs2Iurlnv1ij5>^ijx|3_hGtvc)m*27xt
zGj<qub<4w!aQ*O)&Bd@E^efnZZzc8%oT7VRpTuq0ckwRl>o_|o!rq#jraYl;#+67{
zV>|TRbfIZ~Kiby{5>5LJ(k@&nKGK8=e#RF=Fz4cn3eV!2#z!cy+kOmJaXPMX#8h;q
zkcH<`7&#LXSrmc$c$WrC$j=gdbbWatzJlR=91ICiIG=qXEae7i8DZa~w6NRX-=>xK
z2|^8XY0aRR664|K5lLV-UzeY}$B-5_WaPB+?hukturnnsiQG$97D^fMV4O>ibpm?>
z(`CO4zsNbx%dT%=Or(;tZ*bRVI5s)0#h1&!e%7vl1UvrOD_sbMZ!x4%3I(;05B8y^
za!w1fA)}N9b}2@h2-A}0@WU_`CSBZ87w|r0kZOGfc1^>+XemBAiU518;UBU9d&4!_
zW}kz>E(er*E+wT`4wr(jX3oioh(F6`B;>7Wqg}9n8%Zd{k-MG2{%G#ZsrG`FGgP+;
zxFmKiOU5HdzlME`TU}4#iwv!<dokuGU0jo^?LCkv9kp>+4l9nqj+axhXY^o{!?pD2
zlZ7OpJd?bnNvMfyM>jdyM93uekCvn7`+A??BqJa}8xs(64=3vdcM-x)$1h0{3j0f9
zXH4XRV(B5|#4+&GujQ<A`(`HdEJyv09Vc5|-(Z$%b#2A|l+=F$%y@&abEnJws7*o=
zkXn&TV+h<2`kWRfQysMdQUJaR@T+JC&(@f8HzK~jxh8Wj#hkRph03FtWSx+?>wMe^
zCz%AvN}QSN@+GM!=Prgjn|&cHdVZn{b_U!%eu+KDWiGz;;^L1ZHXP-C0oV6V5_5Uq
z3fWt5lRv7#ui$<LOy6^9t(iM8^ZzA&%;R#@Wj_9Fj(^nY2=xyNDaX3_*1*qdxG(kb
z<NT;4aMKvm5`W?|j8cD%e=<t;nvZ`p$Deo%p<)zLj&kuw;AbP;<9+<2;JyOxE^yNq
zS5B7Dldw;EU$IVZ!1of)R4)Z5$w4?5g0GWsuAjV7oP^z?H_4vZh4&)40bikMC-1^O
zjd6KD?T#oPp!5KYe&f+Mu@B+P_U;tf2*@7CK7_ZY$a+EcBz7jeD@6u-$Vj8C>-QYr
zxNqR^G0YcK2A_s{10zQ*_B?#i-j*VL6w+n8hgTaQU5XtNU$D2j8qC1AN48;q=e5{}
zdOr5J&c*)O?XWlZ2iWiX9_%50CiW^{h<(&2V2}5H_%g~)d~f9i?3Q{BzAAGJc4$8s
z`>%6P!IiFNQ~Gi+OUAu|zCk0Bz&`QL&$#;`$!#M+d2^#}B$o;`?$X>YPBxs{Y#d@9
zw^qpb?@l(%$5jV^b@9(y{pDQ+e+R?gYi<0^guj{bM_c`kQ}8zd{%&sZCnfSaE;)`5
zrBo$CjKf#zoFA6l3JLW(DiKR?Jr;e#fNTiDsiF-Mc`bU!`^$21H9_(yBk3=(Q_hFv
z<~IJOL4vD<kn~qHn?{JD^=<qOh6FQ0jIK_uZz)k@MROs0qg95hR2I#KjIILj5Bjhy
z!kAc86hnC+&7@ox?u70vix6+@LMM$<zVW>yS%h9()Gl^>tKUQs^l@4EJ^a?D_+1a#
zXNc*u*#5143(pt0nhOzkY*LEf7~CU#i70(HbI1CCdw5?o8}_4}fiX`=kxqLayS+be
z$DW^$qOa5Lp0uCmOZH_Kz8g}hLh4ocMWHaFA_Sc&oXW_d?_!a=$Uzkr<=o0SfNzu$
z-yK+l9pEQo=l4?V_C6i^yN~<-{-20&ssF`@<^TRKMM{77zZcR^{ofn8Y5PADx%)r=
z{RlZ7V_no~bqe;`Oq6zkulWe-gX(HJMz<cY)~Gjx@hkO7u1?V#@U;i>Aus~6-zZ$D
zYe)tnltK+1f`PMMZNlEk<esGt@@^V0DJ<%ibU^yn5nKu-p6+s{kfPI}u%maYmewD~
z5<V~%)BgI1>i~5CLe;C&M4)a`cL>V0{F^YP2xD5p9O%Pn7bXKc-9LcuW3=W7vo(Du
zLsD4+_hqlge6EK1T;WVUeF=@q+84hXwK!5#AOxk~4`3^r&ng!O#dW>={zft-nfI9d
zjp6`M6C*A|zi!4Zwa;7k1vxKWD|GPt?ql_c2KZfxJ#C-0x1_`{7t%)Ta{HXUqg4t$
z%LKpM`*x2*OcF*Pf&FnG#4fmQyyFzn6O3+x`~mEbds9k0U0gi(w!~wCZ|Hk!{yOxf
z(COHfyH3~Z71(|Ee(bTk!PQ+h_9~yL3$Wk&YuH2n2JBCN6uu-h13T*v4!U4(|IhHf
zfJgBSf(!9ogXQ>E!W7Il5qwAKS9~Ah4SZ{2Bfej;7GKPmkMDUD$ieuc$u8LLn=lG(
zz?a8P#=3#)gCp)mez_0QuYoo6G35~>g*#sWt!M;~VwhE_<`YvOVU02=$%X`@GEVmP
zenvpDz)8HH5i^7;7~)5yt~jHpE_@zxh06WC>m=vD3(IofL;FT@@x?svenA8yy5zpq
zbwFyY`J08x-#q+RC+i8HGBXW54I}OFPn@iW4}C6DICuCft^RUnV6I#Pe-F3uw-Ejo
z4&Tt~Z|ueJHxvHOZ{u$&{7oIs<G4#Vjy&Y%!5``i^+4%Lg<8tb_PxHllc1Jz)7nT-
zOY!g9NRYGmS8XI1P4Y0)5D+(AGb$QYsN=k(hSt@^-izm9KNmZH@x@9(dZCZc`IDGs
z<Hs<P@r}7dz5~gUHvXQ1gz^U@{l&Y<J0O|f#@}T^j>P;otikysn^($psEhHC9hM?H
z0y&OAHr2_frKLpNv7aGB-60K@O^56Q$f7B-BFNr=EImaw1icYGAhCHUY9FCHQ_W<w
zquD!gKh%nkoYdt9U+j^w^+Ltg58c%2H#QaD+>x=>@JrW$z*T}g;VO}_!{GO@p;xu~
z9lai!nv9|M4?R7_FMPx%!0&{iC|xoSIbUIqa2Xp0zw=Z47Qj8#*Bs?l;%h{5D5U<n
zpt6PN3T!h@^l+mCeT@L$B5-?m--7+R|K<1Trd_&eU+!n^^Df42Lgw5cRL%|fMRs71
z!Y7@{Ig>Hu1<1*b9ssDEWBwm^?*bQ9)%}g1+nizGFfcION4brnAc}HP6vqKk#841X
zltfXyhZhV*6fZcYS(as%W@(m2Sw`VQqDE<CS)@hjLy|;V6ygmf5=F;(zx$jsI2iRj
z@Be*&pa1{!_qd0%_t|@`wfA0ot+m(Q`<$K?NMXd>b|?Y;0*wGT(EIqx1Ag&^1kChu
zU<E8dDuIf8xN&<B!@NeC|8wv$rr>LYQcDsG1U)I&&jFBWd{?7bqCiYN-tm;+WiSia
zNm0b|hp&~i5ZW8_7b^r6?mQ730blwSrOkIG|L%(&$j^$;ibaSRUzj6XA(9#sjnp;7
zF24*SQU}dSx?85$+A0jbfyi)kS+i^}jKW;J!wUpoD;+NJG%r$uHV$_d+Y^orz@kx3
zvJ5tZqi$m^C-xNHF;C-LN%R0B4If}ad$ISpIU)_$*=8AIC%@Bx5AZael)-!yt=a7=
z?gz*iNeUhe04~tjQ^w_v6yOr)aD<{}JmpY)9Yt}Qlnl2HZ(`TryQQvpGrB#2j6kds
z*Z}ypGrsPCILrP*w46wh0$;95g5?tch$6lodKk|bbtp*IsPISsJw^8*{3fM{`bd9V
zgiIQlQ1J($p75kzZUZshicx}ciemJed&%XH8$H|q0KP;@W*I#6-9&uxD(Zn}r2jTN
zN5E4Hfv&4Y9JLfv?n{*QUySjwI1n63as)!A_0JLmi6dwK6g*GCulT&b5h+r@GZB((
zuc4o2B8BL@&${ZT{-8DVx>)9(NJ~88bs_``^9~VLSOdj`R)tncxWdj-Gzp<c<y>Mf
zxV{1I%VzKWU7f@BpatiOc-jdbpE|@R$`Hg9Juru%|M6F@V^|K_H3GlZ_!PwPPmI%H
zNyv|&*UC`DcE?;++tC6vd{U3eweCC~$S5R%!Z$?a{>6Ba+O7@hG}|WCo$?cT)DO={
zKFD`Dki0beD}KoxT2X&hvS#F1Jyu3Y|3Bo9)mI@W0-pRSLi`=^!y*g{vvP@_ic!yE
zy)3KrFCm8%B>KemrlEbf@$8*0am<FH56uFoEg}w#;_h>093%yB^l=meUpwgm)Js1=
z3f)J_xkMa99<z7234p|~jj)*{e$K)APZDM$9iG0xTNLdl@z4zsA4ZFxm4|MiGz~g|
zB?Ou&@D!e%02agp4<S?_UM9vtRp0@N@!RLOPk#O?QV{&MA>Pr>-a7hZSGD#=ACQ2&
z*9mlQ^p)2wx#TsZ0EQPh*{cD+6v^QQE|$Xs_>zCq!fzn8#`kTy95r5`O0-`oe3O<y
z+J`vlFM^$<5M0Y)M{Wxyw38g*eG-D#%efx3@Sof$f(zeTYV4|Bx&ss)utS90L?C$p
zO27sg3P}Mj-IK7(@&fWCT)O2*K?s;>gNxQ>;Y+%t2Or`{9mWs|c+x1sw?lleT=I7r
zelLpA_0si1Dk#uF56a=8c4PnKhD&wfyt~5XbAh7$VGjXL1A*j@cJ#OE7Lo#ZJ{7Q}
zYW!;?Tt28h3I1P@XJUY4b62Ci{>1m1x~Pv&A-FwD%=?RoV-Vu9)rc=7N;ojD>>krS
zM#h2m_d)31#|8%tNz$_eBP2-&*ZrL&xj?XqGQcWHh<}|{C?uTJKS~nrfiGU<K=v*-
z9i<t3B|D(2TDtc|>-7@p`-$LngB)>pm+2=dXf1b)Uw3>$ORA+C+FcB-Hno(veHu&V
zwix5=wXSfvRZ_H{yNK(&hzqjCk3c2;K~iw%wwq#nTYSlUf=~mCj|AUe<XjK9{Vtaw
z;=0C>HR!)^quWH3!f3*iS`m^Oabk^FQwXjBz_r$QEm8?C*fk{hE|zn3;Nqsr_4=*r
zdTB4C)DjW$4+N6-4aBo#C?o~Av|j+%SYN_FBnKoG3BCq7R|hWiSC(8aNlrdA%CxX~
zeI&X6ZxK=y;#Rp&QV6aLit#D-DVA}KMhb$@OAq7XGUa;x@;CK@{ME&X^*TdvxdN{)
zScXDUfXnqdMzkkJu~;uxSjr^m>>kF2dQFk*h0mSi=|lB$ErwjXEaDP9c&<|sKPN*W
zDFoL#NZl{IzQ8@91}#p4*C9EVZ5xWdCGb3!8u9dDT!SgvffI3kOCY(lP_#YaIKmqw
zh2V0h82bzM7jRE-@lvp_CeOqm+eQ3t1%nh~7fIIQL%8(6V17h&7lp5`0+N0k;`K5V
zk^*@8PbennRFFKY(U%|vA*e#m)lo10MF|(QN!PfRfy%KWE^#KM$GCPSvWZ|PDFhe%
z->$)~q!$S_aZ+&gmUD@?5Mg+H(X6Ys(}O4UXbt55ia-cFfnb#cBn3DE!4rYM1d=gJ
za)2Mi@VzY#ViWM&<@$Uo*GCEO@O=2Lh-4(5775xOiuODwVmv9rOGTXU0HJ}T5R8Qs
z<5`B44-um$#(iK8a_tl2jL)m!eL4{Q=P7AXQereptn(z+GNBP1L4xNJ0!<9o{(V8F
z@tLmMJP^I_B5A$P38a8B#I0J7qySz(J;k{6a*?%Oz&A)ia3N!+Gp=VOT<n#uaRs0j
zUZ@w!eJ0|9uJOu{p^y}UYYE24D6dhtC)CW4g4ZB9m)O59N3nLz@UL{$zX1aP6M7Q)
ziEoScZ$NyfTw3;a1wd<hJ?ZtNgv0+nav{J@c~BqPI}P6(BV(_Nd5bq>3JG|6jNtJ4
z5pk<r`9!LRw<pDTd-{_$B2?H(HW}<P5(gfy+o0h#e#PfrmyjZb2Q)BtHL$#T<139_
z$N;ay=xy*W%2$XqOho)3fh?yG9Pqh$?(y7%6hh^CDR@37=Xx+MH{lC_kO9Q&DA#$J
zI9s!stNEeF-7$7h_9%g-;}A!rBL#PU?@)}}b<7J$DldCcBws5^xB-ISQ{rE_4g6Lh
z|0izpN`~KjJV$UV5z7-8Xf@C8F`}e?AoliX+xesCQ9MUL{=<YOPw=tl+j8l0q_BPg
zo-;hjxB=%O5WI}%SkF<2trLSxi6>^6o{RA-wv=ZUo=JgFAyLA(7$GHZbK^v6lvpiX
z2zok=bkAUDs_SBj4+%XU6^Pf#^oS|iDnN$x!C!)j^<C#tNpgAILfTO=jqxb)cpE9d
zAf;GL0e%n6a<tpD(&(o>jw1y@yIzJZwv;x9{z9b95#N8j01w(x>@wSdFXl;o84s_6
zA8-PicBB+=|1UzP%d3b-$gmE|c>*X_8{l$D8-R4Flrv~>b|Kn|lqp6T0t->Z8J9z@
z&=PWxOF)w{uJtZ*|GIpE=imVZv4r~<-8Ufii5N%<+da>{%)JD0j9&yP*B<wsi0u%A
z748PW-Qc=G%t5;@!!s#>9;9{fHTM#4LX5%JKf9d0IHtfK(=L|zfYixx65^Lc2o%;o
z#9BM{p_pzp-DJt^s6#G<ZePf`L|iUdsbg<(V`M&^3k}f8&*4gBYK@4egy3-<g!l;=
zf|%lR5UI~0m5gAtrrSGi`;f90DQ}1=sFmwNwDNN<Wb_gkQ>Eat7IAB>Y1fhP^^noh
zr6zLigE9zset^Jm`2cZmx%|6g%9lvNs6&w3+68jT<+jUVG0wQTxCtnCNetE;kWwx+
zsGmp|>18=Qpi8o@<2d5wBbL0KNFyhXo(|ufGwN?YQ>1P`p@qyK>{Dgx_9ITCyx1qs
zSR=3>Y(H4S=sXclYXp0%1b)!}o?I-!$SLLBISQ=rkH${U4o{eK1nf%4en5K5iqq$+
zjPorKr>Be)`rhZPQ;UcbHFIi5{DcqbMbPSWSqeU7a@hwocyLCM2F2geKu3Xw6(S99
zcF-_ahV!OKgNH-|EdXZ<J-|u>{8EgM(Z?vp#l0(J|B6qTT(*M-e8)0Rq`|I(hBPYP
z)+6YB3-oJ68uk(zyuq<fj0{IiaROI(e-HQk4G|k!z`K%S)ECqj#5nB@4kkfeEth(L
zy->!^cVH*IVB?SUUJlF|BKFrs>}V12hQvQ&3h^~TBh!%nB$b9;EOLUk!PYC~-50qK
zWDP`fJnl(;*%o5kIkLBl!FQXl%VYH~K1pYc0Bm0-<z6GSIM0QyyDrm0Qt-@qALa~o
z_!awJcM2)nkV~?E$@BGiUZErQFM*IJ1^6!{xVH2_;t+bemQT(x(AZMD>>`ky>JYcK
z3rQikU}tzmd69M@xX?=^c=fcwMLXS)y?<TUGEV0J7XkTS5pkd|oUGDqr-Z|)jAES1
z#Bl)}#AE3M>4ZRVf(9qp8BW{qE7IV!9w}0IsJ{Fj1<GSQdUw5V5_Hne4I(W&L|VXw
z&fgOZ2qcBzIz=(g$DNPkp0r4X6r4XG&%}VfcIpY+{0-nDlJKGVjMHXG1LMg{B7_v^
z(pI@NNg+7qQVe#^{(^gg1ARn-hn13!{uB34lKOnb2X~3f4>ZM#bZsM$)TC}?^b$%*
z3hvYx!;~K`5tka{mjo*Z1Fp4JwE7#Y4|Y)_^-+{T0EDd~meq*AO<*HDFJpNLWA!EV
zOEMPdU9}XXTpcyy-&AlYM=ZgwWn6#p@(6+mc?-csWVA!OjB5+xG|G$hLX8-*V_3pS
z#xDsDQXA5M-cc|3mqyvdref_%s^woWLu!yx&sZ^cc&gE-SiX7u0cK$L>x{XhMMG?G
zJteMWkv71X!TOGmh*<&uU@BniQAVl@^bPFc*<#73MC`qApDG86+JrQ+=R}JXt4HKX
z+|#tH3_*{-dhrtbS@e#%po|MHf}h4d{*D&$T#mK12eJ7$Q9xQm0WW@%EnS{zuv*jb
zOY&|Ybg|EZyNyyn+7pB>b_KfvDSbpd6!v(Eb@GH>^A!If9VLlXiokf{v(9o(u~pbZ
z*hk$#3;Q<xW!JQ@TY+gSeo5YSA}u>W4|tITq!o#@tY_CFWxkAItwaku5$jbG$%S-M
z3yc{AhK<MX0}iG6Vro{`*!kx$;-AAW!L*vdpusDAp%jp|ieRUsh_2<Wy0uWk&Zl7y
zVj8(5>^$m=z~EcdaW>c`Z{(j`0&p<w1$l=1mp<XqhCJ^IO^C892qn~UjEWZ6RJ+Br
z0zwJwuLDSVN`_C;P)|h_c1TvaE`v2Iwu9m@(hg&;d)RfRoKx%*#crE*Un^ZBWx6D<
z+h2HC0qaq*(2npdg%vqM7kF3!?P6yT)0T;Ji9Bp)wWOBe3C;c~Q`ZO7^|5odbC#*|
zMtN^Q^;X`up-x(dyo<dvM0T;T_E=8x+$u1Wn`Fzs6z2L=jy>83<@2mGArEbj){)Z~
z>`lVBvfm-&$P;NIqlE?rX?&8udi&+}%aMoF%u@^)UG}rZQZy(aHMxdg!cW8=AgiVK
z_}4^_1xr>1J9_k?(XY9HdCzstS+VRA0$&3@)BcFR66GWfqhAAV(^z%<5)q%gZm8Lf
zxzqv8R?Sw4J`Lm*0;AojSu4_GLlbA=Uy^7dYc4WPENoUkLNUp$NddNA8jKOd7K`+d
z@@OMXn4Cu1=3|K}=iia;E@N3tsNyVQu8vl~ENPpJ5%Z+3Fgjx%NA-dxCKwkIjLwiV
z1f%mOavFh4<TBbpw0DSYAmt`cUJ~y+FtY9ZvvRw`^69`B@t}^*ITWLebIuaW<%k${
z6s?49=bY4$#_*>^duNc0E2lFS{#6A|I3SHBIAn|CFXPpj0j5HyTVmPy1e4Q2V491+
zn1KT_Nn@13P8hvTWaes3!yW7-R%vuP>U0$8q-2>G7-e54tNxYhjJX5+Rl87@`Fl#7
zmP2y1@GT;SIU+oaP`+LYz%Z9k!tWt=I$tGILh=zxVj#l~kxQloW6voVx%ctbo*<Oa
z%GJ<8M7}GbFY>T+G6n0Gz0i{)6eZ@d%9;2jb(}>gPy)6xMG8opDN=Bizls#AUp8AV
zi+#-aetR!<FNp%RJMzE>t#rpPX&nLywgL7o;F9u?x*|xm#&eQ~P>DLbcEqH)V+wm4
zDGYmo{Y|ESGo<V6&h+!}KOsCNmYq)M=LaJY?pr3NWr_6n<a;7EL&Qc=B#m`!=Lho<
z$O|cnFjl0G;q`oXq)kE^ky~;t_;|E+Jlat&rqKKV<QafG1FX87y&z)i=zDmLQjtQ=
zCP~uS4(9-$1^q<Q7^q;%9W<csip$vLS;CZvr8AL6D^_4b#!e|9$5f;-)OE#dq--E1
z5X5}U667sG=_R-$bh3`U6mu0bkoqc8vzRO~l~JTBFn$ypkT#kbEv6B?kYG#{xx^G&
zk%hD@<k2&vjvf8N*i*wE#OQnSN}w8NHV4q%X|kKXjyeIa(^IfmX2M#F!c1GN8GRPo
z_bh%1tVc=Bi1s0S(n&yChFCLE`_NVyO6F>`LT+_>E5*@U@jD5@ng&fzucir4lN8Y!
zr6D=!9Q+cxIZ3W#-YH5Za%KvFp~vEvJZFfH63=i?4*;wI_?7K3_$aIpq2q8?z*{Uq
zb4mh>vv;UM8ORPA7oMQS9S*U#*c*U)o%9hdl1exhngVAy8H96U8o_)5&rXP!MEOCa
z9YmgkxX5!7th3fz9FcjTwPWW28M$Mj6MBMLiTYGfl8O;cN~R-?#>r8dHsEiTm<9_A
z(2wJyVM9t7#5C5(VvNujsR$iBjykaj9Xt+?B8|WxSlj1Q5w191aau-e6RX5|SROd+
zD6-FWy{PqVxE#Cf8H8f71Z*$-1utOkBA=Y0#*4H{*7RzbRx*BQ_Li7ik9$Zegqe6M
z)*s`DMM#yYV~E5jP?#Ca6yzR++=Cv}4)yvE);KH4eyMf`v9HeBBQ&)S^A%D%CH5Dt
zfJ~lfD?!2e41y$qQh-IO8<9Se1WG1Up|wbRStSs-5_N0G6VflE@OKn7N?0q_jTROn
zEed6#a9L~npXtAXQUn^LDdxH8VL2vBQ+}U35C5z~D?ViNK(b=SV|PI;m()=Y^n%?f
zG3}8D^{_iE!ltN4@Pr0rsU9MYcDwC%i&Tqa6TJhQ%x<%op9XF#RW$XxO&d$L*=>Ha
zY$N;ze}S}KsH94$oscaPDI=t&thMZnz*vSSlH4TiM=P%b_HJO?fs0@vsSMRec@D94
za-gV0q_UKba#vO&O_CWTFX<1WZ)mQ!SPnS=vkvzukZMG3l4qoMY3x-I>#IBpPeWgD
zGP%SttptZ7u%i8{mCJaGWq!l2+&jwE^uJ|(S`Kb{Xzh^bViJ&hBp?xuN-I;vvJ^E!
z#*B3drID7TIZdOlm0@C#xkZVYxYCcwLn(<;`0<s_O4*7RH9Tm+UqX>g6<;UIb|qT-
zFI4epEfSD>IH8J0Tj2zkR6b6^!lU&O%8*LhX&-VtD+V!66RT#m7-XuTb)k9jD?aC^
z;+Yg|sY(Kmk+H|a3Z04;v00xIwReO_laXLlv;abYOw%x#Ce)mFru-xv#2*1*BoG|L
zEA^?APjLzM+9mvov?#zca<HLA@dY$@M|&v_(={?>OQ@{QvRSbQ(DvY$<Q^*0mI8>I
zrGPZpZpaa%h(XG7#7Vv(G9@!H=bEXQiS;eHofMFC(Binn6k1_KDR6+oh+kX!ygSh+
z$n$m{+`pJgv(+73E9%~gUxGK9&`0DkcR>nB8${@nWEOWy#xW4N@X7%(w+nJYcX0>t
zODJIB&wNIle;@|ipT${kTzCf9+<GxE+<M&E(911`ULsxb7YiNxWEd_(+$}@lye@%(
zQNWFq0!l;>`Z!!j86d+X8lPbA4(#2@g%~L*h=Jh*+&#d~(D%`T48KLB;31D8djX^T
zN3nJK6Bz7PgfHdNBn5ET3h3uU@Rf*I*aJu*q1YA&18)t#4!&?w%ZG75JFtsI9Q{Nb
z;AD2HT$-d19MJJ>5}Sm3LPeY$tai9)FVN5wsK;}E$BV2MFx0GC#MFntVNM{lO97BW
zk;X8Gm=BQhyNsXY1AgWlti<ne5gw3IF#lnWlS@pYnKI-mLwRDuqJ$V=#n9+MzDTY)
zk6HgiBZR@shglBGm6RZJd744}8OS{bG5~3!6lVMwKcq;~S4yK9v;q@@(lN5_O6UXx
zCRnyz2`o=BFpMYTVoMceFSgb5z<^nqyf#73Z_mg0FklQ=ZL15gg?=QLNJ|ew9}QVX
zpFqt{;FnMsgj5E$=yIf&NCBxmk;>9@am_&54jDUv06J-$lA_m17vcnJo)}nqA-Uw-
zq7H{`7c1fsdznzquSI(j%E>t)!ckpO?g#9CxV*)(fdmJITF@v@+tZwwN-T6nv_xsD
z6)D%nRDYzh(5RADJtxybQb9T1P{JX6iT9OMB2J)|iGk>0qKQc!5@bIMzZE=5s1;}8
z9jibOdxO9Q;Ef!ytgi@_*oA}xCUvCA-XJPo#^H-6I8#i1fkC1!41a@dY}f91+8waT
zi3&m$jnx*CpPa;`VOcz&745?A#HdBP;uORysHId3;_D!mlRchr*onko?+|JGFfPg_
z&uF9(dy512Fl>{}6Uck{$rs`i(qF{7wihWTvb8UAXb~t_3!n<}nbgHvTdDuZ>M0Gl
z<as56wS{#!r%?pCZi21-1p3uMd=HFTNzQnYH$26`qInd2y$$b``HpI!mHZu>(s;WD
z-eNNg=V(dKbC^%ggf;fSQ-+-!xCl<jgbtjeD1Y`Q+rl|<&YXsG;aoW_r{{WbJ-Hxm
zAU6mvwiwLa#Y|tJ<dsULol>Q=S2`#KrK3`<bb|HauI!-<0z7}5uDb*eA;&2=9*zM!
zXki~t$93cSa|5_=E`mD+pU5q~jla#`;oJGU{5{^x-{&n#N=Yl-lzL@P8>;Zu0-R!h
zh26#OW?y6Xu&=XkuzT5k?3*lR!Qgo(avEMd-$LqR?OPr?2XW41(_^GB#J=Q-Myw_}
z09P!6wbjvXv<Kd{gMd^rmJ;AosTRkgwQp(B`m{<N=pIm!fIMWxtVZbkghqKQe5@l_
zZppBEihLqj(0xVx6jH$RY$+>yo?yd1fTj|keZ-y+^V00MHc)w-`6MNA(qD!o)swee
zL@pZZqE-qRtcN%O!$J06csI+3&~D}IVfF}nlpDgu@XPrEeuYim6y*JgtzbXKNy%gE
zarRTTlKqT5!Oi35^KbEQ^JM+r+H!j2{ha-RJ;{E_e#KU?U$dv!)9g2#Bd6wmffcb2
z6mk-Ugfl36ctSs+1nyJVXKEWbI0ekUg>#wv*>~7?*#qo*Yzg~5`vFJ87P!cj0t)+A
zDB`%1>YPCZv!1LAD=_!*)_~ucADM5MO6b0KnBB})W&_?Ju$Y;RbDd+EIHo7wBcNs&
zob7C-uhI?lDf&2lnBGV4q!;5oOZHSdWao9N0e0F!P#BD{au?#RtFucN;B)vD=gS(g
z3Vn^msiB9sKY-Xy+$X55=(Xg2=7aG*lW>e#2IELWxp77=me1o?^K1AmSo3&ZJdbvO
ze@{G5{sCXgAH><TGMp<vL~-Q(DtOz<O6ZGxytjo>D6od2<c4vDSi5|M-^K66n`|^(
z0%wMVewp7Xo@4tGC)uiSqL-XU{{|=0t8pgXkDJ8J<Zz}JdZ!3hjEUdCKgB=IKg)0B
zxA8CWJNQ@m*Z4Q!Z#;@KxD`0%`w7nU9_K&hE5)<zpHo_HB$vn~abw8)d$<W)7Pk+Z
zTUU@b`S5rv3;!m+ALrx_<D?wNm2wBcUvE<?^y4qwefY$e^XvI1_!s%NL~C+8{{qhY
z26NN70`dJO4v;;)uv2a*XW)jDlk(gGelx!Zd9|DuzY2d{IZy20WjGhC39*n^9^9jV
zz<}eNxfpIf+JpxGIf1t`uyQdEoMUG1b8{$1_FJ4Bn#|4OW^;47$2oEaaG7`ta5qls
z-sR%B#p1h8c(#>o$NP-ZxorL!aPJ+w*(i;hhP&HbGB=f51i9bFjm6o*Kd@7X;zo0i
zaGBgf!1$Ak=SFZT+<4r-$$r2dWlyl*;C>%?`3T0)Y25DxXCKChI)(c;z~|){Uti<?
zb#VP5jJ7J=@4*=O5M%Ev+`onqQHGKDCGK})eEbXJ@+9teVYD2?D8)W8=($%gdP*^B
zKNlsV8-87B_<^WH<n*R>w9<2_Qk*G<eu11L8iweKCm{9g$!T-w8c%2&E$vA32d$>v
z$hml&q94Y@GXFn$?*RKM-b~en93=XWwLGNro>Xw$UF-)S?>JhDHz9q`|C_Jj&+z#j
zy0B|a$WUQubJFTgq9-H{TASz!;<tDdmjZe@6Z(MoyvW|e)zD5GaH*ixw?P|J;BpY{
zF#%e>3TG*+q3hLnv(e9(=QiNdK&$@_jdKl`E8Z2Sq_m<wbVEN4r`*XK87L3&eFvW6
z`wx5xhfr(4A1A0HXzAanr`b#FC2BK!pS@2#11;T+dKOwbf_k1aaVF|T?rY2rUsAB}
z0PTR>d7OG#d>_+JekJ73E07o_>Q#*NP1G*&olv_m=C@OO_}3wqUKiIr-hiIJP3`5|
zl{8fh-R?$}K)>s$_mw@N<*oh&v6a;8l*AiGsn?TVSWZ4v04(-?@CXcrZTJX0feWZS
zcpXl|4|*P_U7DyH;L>6ANSwvbq-WCe>0Ej_y@uXEKSOV$U!mWjf51Cp8t5x@3mm%c
zjE?EcB%w7jm@H-v-U_;uDP&B{Q%o`Q9&?QO0w@2^GruyInHH93Loh2F&t|Y$?DOmn
zjK-7fH+W;fdC2)D_68(V2rQYI+(zzgt`>5ZQ`jpsicyMW#Y9CW=Jt=lH&CKDrubZO
zO7XqoCf=52!7I`P-j(;}{rO&aJzW%^z-RI^`T2Y<<RIi4G-?Utdj;OZbB_O+|DC_a
zxAJ$D46Ljn$~a}BGDVrLoT{9qT%gQTu2h~-eywb=^R)A|>uJ}=F5GUg-3+@o?B2Ee
zO{G@3sdTD9Rj{hRYLIG}YNTp|s#w)-PuttsJKKBMhucrEpJHENztR3#`xotBwcl(1
zj(w^9W&4}<cN{2(9u5YF5e}0ara9y~Jnis;!z&JNIK1Of<?yGV5j=$iVXQDgC=^UW
zv2Z{*C>$0(5l#qS3*R}ijw(kd$FYt}9ba`k;ds^Y565;jt&UI+QOBth)hX(9^(*S5
z>Ywp8x@Pq)CykSbQ#YqTr`}HeoT8j!oQ69kIwd<zb;@<hcUt9Sa@yqdiqk=-!%m+#
zop7ph`pK!m>55Z}Q@b<vK{-1)yEz9r$2-q+HsN~H`CaE9G>)1G%^1yi&2r6Z%~P7^
zHE(FDH8q-_Gz~74%RrYjmkgKbE<0V`cRA$py~{b5dY1;58!oq8?zu9q-CcXQj(45x
zTHyMe>uauuT~E1Q)jDg#wCUOv+O^sP+J9@$Yk$#Ra#OiQx{Y^Rg`Lg6x_i10aG&75
z)cp<jU)`JBn>_>%SC0{RBi~k!4?QY8&Uv(XYCL;;#(Pfle9`kA&+A?xUPiBxUNgK_
zd%fy)!t1iP(mT}qG4Cbb%e_~7mw2D`zUAZQWAItv^SI9npS3<4eM)=|`FxChh%G*M
zbd*l1>!S<T4bjEv5_KuMblrSiuI@G6DcxD!AKm(Po73&NZeMh3>2|w&X!n`jpX+|O
z`_I0DuZypzZ<ue6Z?W$w-)nwOeo20t{NDHb(C@mx&cBy`KYyeDJpV=h%ludQzwG~=
z|5g7#{O<&W1&j!IBw%*HlL6ZUb_MJYct7C7fX@QH3aAcr4fG0(2pk<aIdDnf`oO(`
zm4UzM9rXkBgY;SYP5LeRm-Mgd-_Td<n|t{7=+R?LkJUZ4_xPYkb&s~5{dz|B9MUtr
z=hU9tdcN7SE{F<J1_cJi22BZC7W8b;+d&@(UFj9xYe6qluNQi~-K(<Kh2BhW|K3A;
zPwV|k?>BnC-TVFCzXsa{2L~qx=LByL-W~jA@VmhWgX@ET3%(NkM@aV&eMoRfSjdEs
zM?+?W%nex>vNU96$d-^7Lp}=mCFFL<{XWJ%OZ&Xq=i5H_LVJXc4}CiHXy{LUo%^Qr
zo!ECu-=%%u>-%#*w%>q$Y5kt=_f@|u{ciSa>t_yQ!?a=EVSZsf!^VV-51Sk|FKkiR
z(y)zT&xE}Y_HtNhSb5mTVOPR#_IJi-BS-g7?Vr(qM*n&JbNesvzo!3&{@eN&_dn48
z+yHKX&w#%8-sj^3UKsGofHwx59B_KTtpWGKdxVFEM}`j#9}zw#d_s6;_^j{+;mg8T
zhi?dfCVX4?E8%a1?+^bVygdAq@Xy0fg`WvOAO36jmGD2p??$i@4iT;qJ`sTtAravb
zLn4MpJQ9%>F(qPV#QccG5z8V9Bi2Pc8S!kywuqe(Z$!KkaWLX&L}f%(#P<>BBYusz
z644TICz6g-MQS3wBK;$K<Lg=jBZoyML?%b3M`lINiCh%9EOJ%k6OqqGz7)ATvN*CN
z@=)X_k)KDNiaZnfW8}riOOefyw<66^T$Dr9(5S+w9Z_coIt&~-aOS{g1{M$eV&HEB
zuMDCGjTrREpqxR^5BhY_)o7>au;^*gtD^Tt*G6B6Zj8Pj{b%&u!FXHKV5h<EgS!pZ
z4<0;t+~CItzcP5=-~)rt3{ej0G31dUj}0juaw8@nW<t!)m~VzEhK3C7KXl;Gp+mEV
zJ~i}tykqpWp~XYr8(KE>qoI{UzZzOS^oODKLw_53b!h9*dj`%R7_<hRL2u||h%m$$
z;th`&QVo+0S%x`=g@&buLW9Zhl;L^94#R7PV#9leGQ&rPO2b!%YQs6h1;Zu74a4nW
z)G)hY&ci&1`3>tetlzN6VKKvo4;wvf+^|W*rVX1tY{9U+VJnBN8}`w#R-@6FU>s|l
zV4PyiHqJ9HGA=b18coKhjL#c)7+*6M8xI)&Wjty;Zv4V{%J{wUT&zdz#MpOZ55^ve
zJsx{9_M6zVvFBrdiM<qiJ+?Kr9mWA4=NPAr^N9<H>mAoGE-G$l-0-->xa7F0aWmsy
zjr%(8m*MpA(Bav`UmRYFceY-M*Tg5pXT+Q0KZ^fqglfc~5m_Ui8S&AG(<9nO29Gq1
zoHz3Ikspk_k<cySk%Vyx=?R$$*$ML!79}iASea0iurXnO!iNcGuqfp^YS5@Tqn3`^
zKdO4v52NZw{Wj|As6R%vC(?;_iB5^`iMqsq#9oPg6C)A_CmIt+CMG4OCT1osO5Bn7
z-DscDgGa9z{m$qsj|_ce=_Bta(MiUnJxOh2qQ;CFlRd^X=FKtHV;aZYAKQ2A)UmV1
z7LMII_T{n1$9_GwX6$cc?<Mz5j!YiX#WgZ{Tyk!5e)6j19m)Sn{x!ul#Wy7+B`PI0
zB_Sm_WkSlMDbrKtq%25zJf$FIP0AA~Ps`V~lwB#sDJ3aoDIceNmhx4~cPZynE~Yf5
zTub>Q<qo;VIgN`LH+kIZaofhdJMPnQKaBfr+#jh-s$;5as#mISYL8SyYI<sR>XOtQ
zspnI#jn|Ej8Xq%0ZhY?e{PF)9Up>BNd|g_%wD`2qX(?%@v=`IfN&9`m@CmCYte^1I
zgtsQVH{sI6K@;apd}-q6=|1VB(;rVSOFx=^D*gNPbLqdQUrTRIzdMPYBusLd6gO$^
zq%D&UP5Nol)k$|JJ5KI7Ibm|v<hhd<PhLLxv&r?7?`5bmoHE=qy8Y!Ekx`iOVaCso
zY9Af?Xy&6&K3e|h#YZiVTBaDLjGmG*W#W`6Q)WzAFvT=w`;>R5oSJfD%B?B)G95BC
znf{r*GKXbO%$%FKA@k+T*D~M9{4n#Q%uh3`Gp|nVHg)vWg;QUc`u5bXre2;#PjjB;
zH7#V?*l8Kl7EUXi_SCdJ(+*AhGK<a9XZ6b(nzbscJnQqUQ(2d@ZccZb?lV1fdffC0
z)7MOYZ~BMRznFe*dj0f<>6d5po3Ulai!)xC@%oIfXM8uKZpJS&F3q?;qjg4mHkB=8
z_s-79&dQ#h{X+I@*>7fllHHVTo+-?9o#{2xcV^$25i<urbWNK1_{<k)md^Zc=G9q_
zvwF{(K5Of&9kce&`h3>cv%Z^EH|v*Kmu71_u7KHt<ZJBgwAnAretGuM*)6kg&*?sA
z@|<aNHq6;Pr+m(@|F2zl=KO!*vdpo}#a5F4@ylVZ=6~qwHh19MR9w^NK0fz}xjW|`
znEUD6nz`4zyzbBA<~hvM%zJd+&Upvt-Jd^fe&PJR^S_+`+x%<u?>rX#SoC9KaXtUo
z_Q&=;_U&VBkC}7W98Hc#PPd%!oamfkIU{l&!8I-?Jts3KJ7-?bqMW5UD|3o+PUf7p
zxvu0~$+@$DTHv-I57&wXYZq)-uzA573qD-%(SlDGoLq2v!I=fM3w~bk+kz_#ZZ2qB
zU|z_ISJJ|#7hYa?XA!eVy=cRt&5O1!+OcT&q9co{7yY`p&*Ip{Qx-2<yl(OHi{D)Q
z@#3FyU30^7({eL%vvOzWuFpNaq}!5#OC~Lux@6IkrAt;WDO$2|$umn{TXJ~GCy)1i
zyx{SYyoq`9^K$c+=dH=xkoRidJ9!`Eoya?v*N}HP@9t81snb%orT$ApmkwMyd};F1
zr<T6E^!!rGGPh-em(5+aaoNFT-z>YZ%#yFjcgWY|d*pY^56thK-!DHhe@K38enNg`
zes=zQ`S+GjSpMqrlI35B*Qw>-FaL4*jpc0xRDnZ*rogixu%K^2R6%?}TESxlj~A>c
zSXc0L!Se+#7rb6jUU017%YyF;ZeTOBdWF}DUMr$j#IJaC#p)HWtk}1ra>dUpF0Z(;
z;`T~vrNc`1mEBkNSsA@DW#y!mSu5wRT)1+@$_*=@S-E#*$;u-uFRr|_vbj)E=vx?F
zIKFT};cJE875=%(Yt{HwlUL1LHGkFORrwv)o>d1|UBK0}s(IC)tA*9BMrtkb9pZ}(
z6um#wNIk=rY{^SAS_<gdyVlbc6-Bfl6fsdz+f8CVY+j<)Qddn+*u+>l#^Bi#Pe|~j
zd@Y^w{Zq^e<Dbff5*lj?jMTxk>Lcc>`H5z8Mp$cW`NA`Ti9T_nNF@k+_eMuoSI?VQ
z!~_TLv8A|LrX(FOdT-kFc6EJLR?V5CyEkp#{8?>QR`o}lQ&Yzd3lz3*_v$+$B{em5
zVtU%Raj7QHVEU&sXU-HB7B0;m-f(E!wrx>SQEH9SVp-=X2!e`dS$jtnRs`%NY~;QF
zYwcVjW19ZB*Y1-MheZrr&Y60??b6lg9$d>0AD5Mt9lvOD*6Gyzt`~Og+V%EX)sXr4
zQb#OtM%f2AwzRa|qEsqZoh|(rI<;hLrx@QU*7@1*^0Q%}E7r2uvRL)f*SD$seDj}Q
zy=1g3c8rc5GBCuYy`-c~=@%Al((vEEwgvyztXZ=ndjRj?XwR)vpq}=2yhM$-R;$xB
zG{jm8ECuXqXK7U|7-B8U-~awKaE)VJT)Z*Dhb=FsoO?&dncRYFYJ&d(s4cgsWLgOy
ztzR!a!JZB!Pc!&OV>v7-FX}*23~xAHND0C=+p<kIWjphrxl`+Ojx~3B_TR_Uayc~?
zDHVh|npPA>6_$nvD1u7}Y2xS@hkCgB{&D)mi4$M_Xx0tL$Y5Fju&A`h<HQYBPYI6=
zP~SYfbEk(YDaqhv3T$b)uTt5m-E=w~&+dP7@7}#{9lpqSACr@lGd1bt$*3ryy~UO%
ztYhY{<E}FgSjSxlwG5-!lwXp+DOFKS2}y#~jvm<FrW9M6?Ej08w-i$AEfolbMQiyo
zN{APNo1~n!_*|_m0GNr{iRHV>sZB;|{aOTilMHzP+QGg4^W7UaZcI%TgtoTQ(zZ5(
z!O(-h*=n*gJN|jH?AWp7<RXXU<YUJK;mp5Yd+p_oCfnZqzb|jBZ6e8=yiR;zebx_j
zYm$?bgYW)WU-qy4jo*H7ay#lE4kMGR5T&|wxvj0uY+krLKR<u4e)DF*ws)+f(g#qu
z)J>aGQ{yeGu^Z(U6>XVmD#EDZ1;N8sj-HgMAZuufty?oPqVxbL2(i>gYNI?JBzkS6
z0i1_M?1vL<<Z+!0ha<~VG(N1sNNq`Wj40|#bcq!ev2c5@r?bYI_!4t$D>&8`hY|VQ
z0|GYgfA?P>oje^@9X%{@%=onY`uYrNzljoxMp&j;=37!M11v)=i|Lk@5s)iJ%MeS3
zCC8Fh#0o;9AopH0AgMO#`$WejBqWrS7^&x!b>AL8a^y&As%b5?J|33X7)zWb(h@|}
zOg!rG8TQ+rqiPXbZ&M??PW_{kW43%a8JN5(-+y((H=mX6-M)!5cdFChV79NVt-qQu
zE<HVcRD2Pa;aB^=L5X$bCjg(5!O+)*y8eA-W##wRDVM&6Sj#fYGUgC9S{~rcNb0E3
zvdrY8YKKs7X}KqO`(kkWda7<<aJLA0jO?DaJoA5p&pS&oD=$w2y=!|Zdtlgs<~!#_
znrp(N8e;;2s?G#ltf69|f~NGOD#77z{9tks$+RJxs&9_9jHA|J6-l39q)s{22nP<_
z5h5mMXJ`AVE*w7a_w}&S@F-~L7BDzD>tc28_<o^vbzyyz7gkkejvE}{gl2Ko>XYNZ
z0=Sgp^es@6-*H4ux)8Pg)-|HoeP=<nyWK-mItxmTuB_~$5S5XU5w2=IRVLvldxv5z
z3oHwiFMN91V7U6tHlt+$A0FO6NN14yr@<xM<lOYh%TRM@jUk?mjc1M>J4X0c_|t}O
zU6YgR?V#+V`+6iNuU~%}(NCX>1IOZ`-ZR+?9v&XNx&7M3ix-RRg`GRIva%+GP$zdv
zbP#^@jEWi<9y&B_c8<CIK1*HuxxT)>p-tt}FDlkD(=wC!R5d<MJT`TJNmhMkgjCS0
z+`IV(1O^9rQRdsNt*w7N;FxQB_UySX1U=vwqS0A{AvKRMTe784S)ZQ1WlMW^)DoIm
z@z%xk^p*-}WD{54aOF~c5nWnZgenU6>&sxp==FM4S3PDsrX*Qo9c{CKPoJ6^<H>$m
zQbH+}try=dDk|DkT)Ag&MMcF0XWxM{=jnS#M2}t)hd0mTQmmeA$efXC3iZ{P(PK`U
zdGm5}b2CHP>)#$O{p5#+db0<5-H|%57galWuFauOeB#g^-nLZT?lkU&PBE*sAW_-%
zv+1Cjs}mwdjWSZNs<Ts4QX;f!MeFsR-j>THahB!eQXju{>#5zRulvuQJ^Q7XUMe*?
z1qA4H=iFTbz3tlZ&%(sRgnL3#*_tta&8JRz9xvUoW5+3u=Wi<mBzgsiRu0+|-)NY<
z>x?0IExS<=HcE20lMPcDD?aBN_EHaW{I%6Rx{z@Zs13L>IVqI=@o;HrDU4Eg_BzI^
z$wBbYGHBzw4jx35e@wl^fKg&Gls=}cvRG=T|2YXG!O!f}N7j)qJ+sEVP!Eb7VR;<u
zS}$O%9;2Q!F-e$aK<Z1a8HfzLHuab)q3o)WdYSd832Qc59w*9&sw{%u%p?D8ImKG9
zzqX_mYiAmKfx)h{fZCjqpIAg0@hyoX8%@y31+eLhsEtJwD?ukZlgrNDwW}l!T9}1V
zlxm_{ikKEbASMn7wzX4!?Ns}=PBE*6D$!w$`2)U@+RtVr9ynmMjAIWRh>thLlHpTi
zq&9Nv*FzE7*0a77%+7Ud==7|!rQ4&^v(973t<$r+ax4Rs1<#)SXll6TdR@)Ol}GpQ
z-+%06d3ky5)$3<#N{^TY<LucUt>2WFdyK`<8DR>}o|BW)*4A?MDmc+>w#VSEKLkCu
zdyfDqu5)hA&CN(SdGh2PkA&P#t!A}jrIz_WjmcI|Jk@5~GhIV{5_R-wZf|aGz8V!3
zB;5b=YJlFM?dGpd&3D?`*7a<>SW{E;W8E(|+iDu>nlH87W7}(lkr-LS0+`xQt7;k@
zLZh~9tuPJK>D=1xHaGhwCMFJxGT%l0ubd^VK(qE5ji&w#z|_@w_@EUc5(nt2s;X|P
z!qd{yLfG3Es;k=r0%)eXQ;S%2cs%Iy&7CtjDPbf)C&U`YXJ(d`=JH=cckTVIJ!npD
z?hJpEV^|;{d%LRG+rZge<fQV-$?3!1t-|E8x;A=DjtqM^scnGMA1#KV5kh+{rX_Wh
zCS*u#yF+(9Ocq##;OHEqNmo#?JfDSO*V0b2j-G(+rry7M`}Xa1tbcmJ18#g`E!Emp
z?BOx^clkzw9=|Qg$$R%6I=26VLtw3n&rZDi;+FQo&=XUF&EM?YvgOtDCfZ9?q-&r+
zP8%(_dV`!$!=63AsGQ>#E?j7gsz1I*u3tXjQCs3HX_jf0)n41nOAnNk^SS|0@`HWN
zjbFm-{OIFjB~KJVDKf8Cw8dEpajkJE+WM0^U?PO+K-D>;rO?#hqxRIt<sQC$qYdp^
z7yE!9CxZH}*^DhTZ1-^Om!-`OjW=6QY~8xG_`9|Mu!F1v?0}wbOoL^sk|}x&DU6p+
zXS7VFYir4`mzQXr&}%ovS{}Eo;MYQjZg}R|ILlHHJ=yLAH)P0=o~<W$42Oj<8P{y;
zO_MSz3bsJhe{KfRVjcyY{_}D(2kO23z0xNP<S!mQaNt1M**2#^u!+Ydw|39Y-nRWG
zlYgBm0poUdrkY3n?s7b<M$@`0+S(fGDk^RT2me;li6=Wtv;WB$=|?^+bABHF8L18Z
zR;*Y$D|7mcx#^&9;E4aCY6Igbjkl}7t5hln)rrqw>z-x(`$ha06&72N?8ESE^BtcV
z9vTuh0{xqrkdsqVa^_4)Ny&RvCbgHBv#|Gd46Y9=eSJ$xij*ZKzP>dVEM9VZ?MC@w
z0fsTju}PVknTfqT!VK}`ZiFE?b{vNNK<c{##l@KO2MV`8+qrY+W#0GZ>C>mroH+qX
zYuZ$8oUb-r_8jpO)eRUJznO-mxBc1F)Yx4AGYrm)bu}062|BC|2kSl6Tw`P79Tm@e
zN5;p;59v<TVPu~;U32mJ)z-UaQ0wK^dvr;O>NnO0$b0fXmvrh^Z71n(jmc8f14p!N
z=3&;SeNjVgZf^dh)VQRiq|KW*pKSF=&CAU#Ej1~1hJNTLZ*7sX_-$0K{A^88hYd0q
z<-CJ}avK^-yDsf0a(dK1!5~@!62&Y@o-oAR5)$<K;^K`Pi;K;^u#bGrKNJ@i|6pQ-
zoSY)(oSZ#-gv`wT>g#CT>+1fQa_y}a8^dr3`r?hn_!v#jo=&xG@6<B?r?GjUv@bY%
zaV&M@{I6fbe#^>gZmzH2wX43q-48?4ue}Cj*JuaLn>TNe_6CgV8!&i`*esN(ZEM6Q
zpqG$Y`^)UEU8t5WCui5LPSxsEMVmge(oJ?~ucg~3nKsz=sP$<qo_b-DK2TX(@!4Cj
zDXXhhA(&c)sD1^dzlwyEaSQ!#H|7tAG5jpH0q<8o+^{0n-hkxT8B&*}8I*pYJ=RiI
ze+l{0DwP%oKS{V{wIZW3JA3b56D??p7?t2v1kq|!H>;i`J{Bt5NGp}Go99)Fd`JuC
zBBmKeC74ev5rQll4CbVm@Sa|=16<nv1+jSKh{;ivwFK_Y)df76rC_44k=eCN#v+e`
zCd%B=W}?64X-d@v6>@G3HqMGq2LhwVf#9s1#<DK(x|$Q4+xAocfSveS1A|75g#v@^
z_1@{r-9tlbYioZp>&9kgR#wKRKAM)87^)A6G;ul|c54>#ZO7pGc>m0U8+$>}hhjK-
zhW60Od{4YuZl$G}^OlTAjh065wvSsp#$))J*4ha|a4%{2+8LCeVEi8a>7uPC#%f1R
zAg$S|4h;?M9Tj!8<}_x|=TD!84;#GUIDO%SxpU{f{r1}*)T+Z5k?G(w)H5beUthYZ
z=DQY_tvG(RsofV|r~eL@WEWcPSz;4$NlD$#9f>6h+<iU7&*m3Cu?is6D}_iUlS^uU
z#g&HKg$t)8NHrmn=t2ED$v|tZi1)z9tQm;W7_FtQW1?Gf@`^bOy%$cM#;p022{xR&
zqpI}^Sf%;4O5>&5v*!m@ILV+oXSG9Y?VtbMd#SEgFOzL);yvkY_B2SeHCBqWRmUe>
z4GIrCQqEVOHV4BMV{oR<Kr>Y~6Aw;jN<CW5M#~Jhn5)eiK5xHn|NeTF|7a8mRbBgZ
zzcfZ{@!8&!la>P;ZZtJDU1|*L>(Ae5Zq)hmt#|J<-!yyJ-y!ycgU8*L*4wQ@6ck;Q
z@H6<~VjUM0x@mK%X&^C=+x~2B_IY3*|NF?d9Zjgy#U#h*cJ2C26)HC5ue(H}xWgtU
zJ?@*B@Ca(2I5HVFZ)s_Ec22gLn7^|F+0W52#b-?LtXWOqhPFmQ2nz>Dn7@e-r=l?|
z_qYfxH+N=AXNV6&^9qI+z!svaOK74Ed;0J@d-o<MZ`rc@NXds~=$OM_ura7gMkssk
zRoJi>O@2CEuW&(VY&&@g)xOOVJ65z|W0B3)h9&THrKMZ9mX_)h;zouW$ESI8^YrLD
zJT)~bskW9%3o@-0a&q?W6$GyUc%WQ_5f}w4h|338p&cX*wG*zaot^l*^E|P0%ql~t
z0CR}ydkpy(U)uHF$1owjzN!rxo~q{W=`tW77iDH-Ooq>Nbg*Equ28F0rjYFHzC!!?
zFRQ*h+tAo-hvsq<N=hKA&b2rM^$+XYw|77{1*xtoYi4#QdaV52ncJmUXHQP&oYr~k
z8q!KF+VGf|n1snW@V?KR5E~vI9_A-FYcw(8%+*7~;h`qiD!ZLWYQVQ=%U+1HtR|QJ
zrt&6lusvp`ADaevc?srri^9&%PNng1cUH0#nM2)Zn!OMcf<%F8{_W?V>gs;@sj1Ck
zzFm#Foq$?tH+xCrg?MB<lHz){eNIN>$&!+V3sd7UD;Xc!+t177ArQ;jH+|<$ykHED
znVmb+vWTWBl-t6H(%lROxIjz(7DhJ@*a05@2@-U>sF&7~r2rev)Y&6&QMPCL0Iw^i
zSdV!84(=rLulGJIeIIJxq!RQAFj0nwL2oLXh>(M?MW)tP9{lg#ckXA~($bSGnKIia
zw)bwJm`%;!T4KbeeRdo11A)h_9w1^d2+7CV4JCDXzBQp2x9CkpR9TTy^ya{!^Ut!)
z7LV;cu_+vK_HWvrMQuyYOW)dND&m1OSs)IgBN9wK@;IgEOzYKHRZ;%YXD5pG=eBM`
zE2@4v^;1K7N`!`xnf3p)1xdZ&nPTIkOG}NGIc!OZ!C;~b3yTzm+qPkmQKCU=!T)AH
z#)({l4fe8q($#pgjxgz&M;HA<+Qq&$EH$i_3au%r30?0HR?CLgl-7i@wPBe2lHCOV
zXqEd)Ryih3D_17>5iVS~AoNLIDRbF>v_pN3)L-n-he3%n{H?tw{KekexKZYjA*h?j
z-$~zp6ub-suVg0?9<r`pkHx)yTXe=5bGSy(w3#){W}(F#ReUDL<FvlDCa9%Gr>+eS
zI<0Slg$q-)tkZvum4U&-@_)INW&mzqHVV_dctecNj2#N+&U{y0U0qd;InM7_TmEdT
z|CJ63YH6wDBQeFt`pCDHCO<DPr?%wOv^2BjUi)pVTyiSMfPjG3Us0y=gom3~T9HE<
zmMGeVzL-SCN6WZ~zFgZ~(=#kAEXajw{t>$&&R_TPYG|NBu;dX!{TqVx-zH}WQdY&@
z1=#z%O6d$>LRw|zH5HZ&QuV59l`=FkJ8%v24GIdvtoEyR*AWQ`H8s%&LokLcvyRtY
zLh`w~BDSrKclc|28cAs_m7X3PgQ~|G3|qD|r>D2p?UQ^th4h=}G5Xf8hh7K_2=-jZ
z2o)8>$?7StA}&3k^((7S!?u^~VNl}yBEztHZR#dgSl?+iwNf4f8fm2)ugJk!NVt&1
z%nBiZWS3-aDyW0k^GH%X`s4e`GB}g?)Y(fOpFJ@Zw~Vjv=2$488B9gv=vW-&c#>Yx
zXmS_Un5i`dImwz@0Ta6Dxhq#gMI1R&#Nba6jroZL+m>EwY+Chp=NOrpyER1RHy*3Y
zifL%j=}zp@$9io#i7gu@y1BW?K~L6zq<{J%q{#bqW};j_UfUOQjo}d<5wX$H(S01Q
zo`zv*zTWUYoK(AxRn@iBefQS-_3QV2Q-6oiR35)<!c5=OMXk0sw_Lk)x%m=N7_F_T
znXqXhRo6~~t)P^9Dx1E37b`qv$29KNIS7#t&Lt&qM<fmo-<6t*nb=2FvmraM2u$~<
z-EhR`5OshIiQ)H8S$K=VGJpMpcIdpn!)ndU1xz~Cn<)MQEGo?L`wM*j$I8z?*C_3H
z&fGF=*oeL+BlQDX7>@QI{t5G1CpZ%vwO)ZRw*tMijz1L3ec}g<s?dxKvQ)3%5*>#l
z)X`XPqbtQFis)#}y*llNvaV*48JJ5dwge>Styq;iA<%r}n~T4ktH$&}D00}esbFDJ
zkm6cX#;nzGcwz5J9QCo$U{=Xu%MZ5qA}^C&94vNtq_Uf=ylI=~@9qg*R6;EEI`xb&
z*TB|Re*X1W$F`IdwiB-?@kBhQ9<Oe>-jbCu(z_W`>r|O8i36>3C^8DPdc6*<6{^?Q
z)t!O$^-En{i{HFB>IHIfgId(;uj0retz`kd<D$vmx2;{F>gJ|tY}9ykQz_cpe2<iU
z(x3`UO*@j78m4Oaq^yfkKxWsn)0!I|(DM2wth5U|k9^ZmeH<neX<n0y;NcI3_A+nT
z|0xz8zWqaI;}7sg=h=~#!kQiS7Ln~RVY{{vuk2~DuB<?TH>fod->Ie1PqHg?4O!J;
zd)9KoY+RzQfu?q_CtyT)5D@KTp?I=oD8F6GOxTQHLp>+~i^fig_ZEl+lso2*JW2@Q
zrcOO_gfuh_@019c_#!pg*?eGZY^)*7cjwO9?ChqRH>~pRjc-g!0iqBs?Xu26IC+xT
zmy`5&&z_W9%z*OFnVDy9sMTuaoo2H0W>043NX!}X;c!kAuSL{OD-?g{b&fDstE*vC
zFG|wit+lR85IURy=LDRORc9I->*0ZbMU7=_%S_hG3oaU7fHPr4kZPSusI7%{IU=Z5
zbiPRS7=iMPhlknZ&6k#Hf(H!fA+*<=J^gt_`7RYWrL_!`_#dzU7a!kG-+i5Dd3kd?
zb+@@Gaacr%uZPgsh*@RMB>mM!$-W@-bS5aMeVx1D&{kfKrTO~g<nnSMC`kC@BTNU+
zftslPz4Reo>qyJ11p%9<Dl7By@>gc)FCXuMo5lj7N*y>fv6a&4!l#ivZR=?j_O{{G
zSI%?#YvEo=h;pMY5utYKDw)p-Z1F1)Ylkff2^AHJ323mKc^(`jpuBlcsyFe*J2kmM
zoO)lbKJ{sN>&^PQ-LYu=#dK2>*-4bIe7m}_jVk}<qN$rsN4MW>28b&y>5l=RM^#mH
z^zT6OL4`+$|HgLoN+ZuI0d1d<oj3Lg;q&HY4byjrwv`!1b)d<A%Dj0M6>V)$l-svw
zj>Qg#AaxNR&tZVHU~aR0yF98%i|ExZ=o}AUwN|S}*VNX=#|wg59Wfhhmfc_7Tme_Z
z$2Uy^rhZ!0A0YgP6V=s4Zlt+;=qf5|%_?`Y=rG7#b-&iCdBxpeWO&9DmTcb*07vUK
znIkbF8yB^SNI<gZ)T#D|FB>rRKv$O3wt{@Q$3~OZM4e*9It9^ZtF}G6Za8rRk;~yh
z<r2b8p4&6Wi<MJd`ec8x78TK0E6Jgf!gd*E0bqKvu3G5CJHIuYX=%5pbP7nhmX=GG
z+VbM@PPCV$O(~Ac<MkZ;^zeB#o;<W|1U3(dn?ay(mQ$Nd1B8ZOIkg7^arf()U+{a+
z#aqq&+8_wd0f9nPHs)1^CYEiC#?6Y{c#YmqEw?nWdHM44HEXmDFUw+4_^5r&8ns6d
z^k<|Yv4|-T3p1^C6$G^V^08j0p)b)W0X`nyjzwGpGBrr!A`6gw)~w0GG_$%opqo#8
zpHRrZ0fW;A4-C`w?w`IO4j%=mqb}27VR3lz-S5=<+>*>NZx0ivjf%n^7=P`Vx(*H|
zE{oHjVjE6ndvtVF6&DA0@ZfN{)rfU%Pj(8IQmHWDqKTFDG6ec+QHd8ZS1B_wc2;xd
zP0h`l9*kNeRgP75;=VBAr7bl(*TmNfu}w{co)@TnmI5@S{gET>?PQxE9E3%xcFMeW
zulYz@o7EB$S7(ZgV=`(Q!^5kpH2Px|73S8OfPiRotT@+;i7BGZ<{~F7*a=4jAx036
zsIjf^m`r&JC}->I6B1(Ye|vv}0)NXA>#(^%(Hb1w)P(89K1Ee~dpLWnh)aymzj)DP
z4`O0s%;t8=7BV0L@-g*kK{##0Np^-CXt8}S@7Wq>Sw;5SVux*r&L=2#CdTvX!f0{x
zpj}yWMSSA;=nCx5DRek_GAt}!9UFG?P?OojU%VHQm@;K$9JL!J(3?RElf(4=6Oy8X
ze0qe3C&Y&(w?hD7v<<A}H*LZIsm%#*YYWfGl3LhS0@&UYdyChF7?Y;ut-($hx~SSn
zy{_~qDQWWZ@)k_|-dz}JN9)KA>ey-pCIx+C<T^<b#yTq$_XJ$Lf<^1<3s){)yo81G
zOBXMGh-Hg|7s=|OeOp~aV`F7Qz}0G2tGRu}<cmG6@Ih;dfUs9Nb1jL9{;EIVTe>Cq
zB{sA>aKs~<g(>y#-<_8+dWic_@e)c=H|pZciS6vp#~UpxX)Lyh=D8$W=b%<hur>p8
zV|pjHW#(C?GR3cKinpw^EV4`|7dbuINc9d035m1JAeX~)B|nE|PDt$e=%ZV=n$(yo
zKv_N549H_MU{qAM)*}1XoA6+^Miowoda!~plyDBqzW5ACasTRXjBG|S-n%rlMU5`C
zsMUHWHWxc+(3{%<)BfLoH6~?nwkD+rr3X%mJ4)6DF+1>&;k8j)qqH7ovduIY!!a1m
z!h5J{C_%5%!|aE-UnB2{2nJU5kt3UUJ}511^X9ZP-%a6h*n1g{JvKbdg-mB#gi>qw
zL>ZMnK7QA(hKBffT|Bn$XsG>^+A_^VS5+0+5ieI597k50*Y6>lhzZ`liAcATmO0FX
z%@bm=i!Mwui=<7+5|#jn6_wR4Vi@E7Hb$=s!`?V%Vevw2o1;Q8E0A`$?V(_SN~N+v
zn{ez{Zf<L9OpF&<4~-MOfAdu@AN#3iW=v17y|C<7!H2SE+jHxOLvoAWK^7TLV&o9x
z3@$n^FTc<tt-1m1dEO(;&5LkGP+y4q?p{SqfK#nB#=xJj8^snU)*x&~2oLTwF}DdI
zZcT^HO}vyY%{MM#PJQl&-`m||;#*od-e=r!EaUAXmpVV7f&ZgbiP7Pr<S$=Yr)j@=
zt+}DP<iG(o0+Z{6eti>6gPWRQ3n`p^Ff5hr#c#j|`R>tr%5!j1QrQn|x6ts2ZeU9P
z2)(BLdQ($Ul0FRPFB_ej9!=Lv6z77DZnZ47{L%HlpFer=r=L$(V-L#3lhl84PC>7t
z?6023fInI7;RO=B)pu^fMEdW}E`SFp$N2cty&Esip1o_=8ElL*i5yL@*;vHE?ILWH
zN4>4IB4ffnK0X`uczEa4G1-aB96Q!&6P7q8`l1{=WXRh)jh5A_a}RIs67lkDI&JZi
zxF?*HJ5w38rwEJa2X+-5+p{MuOk7+7ubJ!wct%SFGOSrx@Yr65%~^OLc_)a!o(%&N
zU_4;rAnu|0yXlBXf$N&TnUqMh!~q_s(U{HV|1h`l2v)1(gZ~j$e*#Mf4@-jFY7Pgg
z?j{!G&Wg*=FDWsxe=q4oDtvo^W8$#a8<*quTIUG`xw&biq)HWsjm~o|NmTJeKyt@J
zT(V+YQETL$T|!#O!$Yef6U*Hmd4mJ)T)yf>T~pn>gS}CV8A8KD9kwVIuR(p?rRdpm
zSYCBCzP(}81YNzVQR2iy|7%~KIPvE|tUaWp3dAroJ8J#OmM5;Yy!@gzAi&|ra=pHr
zuYj!u;85`D?M~V&(#DHvJAbzN5hj8{MM+71zVF32tW>^Fy};+^Lm^(CVx)>)mp~kz
zKKj=4wck{fyn$3iu}2kZD;DMk56S16>`+w|8#^bJvQ;GvREbOl*>|N)N?N}@abaG5
zPD;17vain9eP6XRNlSM5JPs>z`d_C0@z$w-hl@fQ%fv(J^Ot?@n@G3h=aXIUrcS%t
z6=ZjNaWOe<kd$QQ*<KKq4z{`j0*omvD<pqnhOome<Dn`iT2d5}gthgOz_&|gH2=(-
zM$0Vr%K-j1{&qbDXJE(L__opj(K$49YuSe%pFCTa*pxAS!O|6lMrt?B^Kkepy~^T+
zZF{$GC5;p>o;?_Yvj=0?tt~BE<1G^`v9>1<9#Cr|S*$(%_wyy`iR~J*@3CWW=s$6^
z{?|Kv`Ke20Z_L2M-OOh`I@bQ>(?$6?gH68bz=1I_^_ZXS-p{&&AMM(2685zB$;t7o
zsnJ{a-EnYC;p#Yt(U}S<c3$RD;yqEBfBEdgGTzC)34Tp+&5GH$vzBb@eM*^Tz$|;H
zH};YhDu~^kUu*JfYP#d7nLQ~xJG;Bk3iFWXc~5_>*3#52nTS^3Ygc%!bH~oljGzK5
zQ=LyYP{*FwSXdZG4a+as_~fUiZVw-7$jNc7sd2{AWLs!x+ulwi(27NFiJHzU5Z0%i
zX^>)8J4uQ)$*rUVeGYTy{`=&+3b<^Et4EZr>>3z!4|W@MhBqE%sP|*3FQ_ks_oD)x
z+v5iI@Nj^j9Zr&Pndrx}iHv%kJ$J6400uIn)z+G5u<Boi6hVX6tzyyd2~qQqIYlh-
zrMRS}ohaS?WcBC&dh_L{pMDw+P7O@$9gOe00QeF!cMs+l`{{GVrnQc1)^w5=b{MH$
z=q9<vSiDYN#6(sco!fsp0qgWPIE=9ach@f8SnKA^sUfB!SCwB>A`ru%S+fS)MJ!Ph
z@~pBr2c^T+_6isb8tTl+lZ`s84j2M%eOAJo-76||x|**~l0Q18{`>bK>$YucyMrz8
zT$?G}Y<36;>)#g!H5LF=QSp$*{0YI$94ZK*aQZiEG_}r18z0hk-yRG|x8A?sCi4K<
z`Q%Vr{_?K!v`hn6^U3CT$^0U}_F8v17_{xx)pc*mHoEHN#-^qQJ1CzNBW%(GreG|{
zHk4j$8G_kwP8424hfbqe_HIk5udgpxe*{a`n7mW<f3!4|N>Y-@C{^v5Hh>NplO1HV
zq;chSbyeHqEaS+fSbL;ar^61+m6mB(%7ZM&sR;3qgd>*SaWX<YB0)5MhZdUwh{0{w
z>gwt+q1IEa-(#Jkyxa@hL3BO6+Z+Gn)sCuWgQ2c2Av0@6M!dI4*VcBcsjEh5iXHLj
z3Zm`y6H6A`oNg&gOAF>D^t5Z1+e`Gk?A)1{SY4f%xO1mgd-{tK)KRO|w%1g>MYg+a
zW))~-^=%VVwP#O}FD6)UUsRzRZgz>%MiN@}Ul!`4ID%<)m@bx_P*n&wPs84lcO9|`
z>ehYabe4Q6b=}d_{h~wQGB3l?G(p(9m8`?zNRwxHEQMF3s)`-4?mfG2Ln{u9-f6sf
z`&NsLbtW+NQw0w-G~XA322R7W`lJ2zzN)C$Y;<R`)|qWj9NI@{{Q>@`V_!E_RX~9K
zaEGicxNDA;LAD<Fx=Gm+)&SDKxh_OuZ6PzY^%lhJjX%t-*PC0~+uO{Xi*;oo8Q20I
z){jI&JnU=lG^SZH`2NxNPR@JTY1gQBX)@Zkco?Y{n1KD&#5?$zV=qJO@F6|!)t&tJ
z@7J$2kwXL-0=BDB2iC#wSWzK$?So#II5ZycbO#RTb*?aif=7&tPfQGRJb?jsS)Z7=
zRc}53LGc0=Zdwaxn&4*FXl^@?A&)h<V8^D<;WR7aj<pFwki05FPMSJ4RbIw6VWoXb
zMn+XtI8G;U!BtiKC~|HAm;HpQDi0oKrNT2Zo>*&Fd$FqiIs@26826N~FBb7C>+X>~
zS>y=ynKOBLl6`Aym)qV45sX2>xeJHPox5kx+`0dYx%UBRy3YT{-`g7ogE1IP$6zp-
z!{HR2q9}^U6rCtZaVwOfxZ|=c&$2A@zFvv*^14r#<;Ocai#zTxj3kVr2&2$NVJL>7
zBuXSUozv-bPKU#>-{bXuZ@S^1?DP5F@AqRO@7Lb%_w)7d`Fg&d|6dc!b!6w&A&8>J
zu|(!CJ~DLcDLkL%G}-C!FrJmHlboXBM+&u)&gM@mJ6d+v6IW4C8ru)}yQ}M*8Zw(I
zRe^5|K{tjD3#KgEw^#_XB?G&sh!K`WD~4a{rfx`X@Oc%n?SBIG21$dUwZ1EfFcJ?M
zTUu7Gbh%cpY-xeL9m?(7ONGIZ3Ms5W%sE+s-q|4BbMV%!rKQeRqMKw$j^H1?rKNUP
zI4F`-ky^&xiZGAcDfG;kX0vzo5IwG#2q5t5`@xcx#ozM^Ololf!EA;|+a7N;o5=bT
zam6EU2ZAep9HaTro*kD?ckOM$mIX!Bq*weo#^|1&KHGdf)O!nbG+jKh=O&KvX|w{9
z(I%0}jcpwglc~p{)A4Qgb0@kw+pJbTQP*FiL&LR%q;oon1ENKhF6efbp<v)P>h*-(
zD+|`^8!24J>*p;jeqsq|wqD=eU9#wbMT?yxSi8xux{Tv=mQlXFgUGI-DsfT*$I>Fl
zn)pjlMMakWv02m>s)70-auYD*>~}4lon6(hy}pTFGK<;`!TN7J1jg+sFFOWT{8d`2
zoO(ShSIX>%wY6ec)w-brtqKwdjNOiPA;<vmep?H6`Jl!|FUe;xmds*MNy(Nit8<`>
z+Xlty+tga&Ju1`HZ~@oV=B?m1Cc7arkpIeg9B?Nn3a5`!DB_c#>%*|i=yP8|rF*ca
zo1UJZ|9D<n+AL_zr%UMXc4EF74w$T0!V#9xw7s&fq4A4v`xBL-G;n(bBkbQm8?d=G
zZFv#!Dr(Sn_8gIpqGSp2!$d-1v=@g6=c5wGr-O4GrJW9IQF;QurlzxBFm~axW!(LU
zCHact!*8sIvE&nFeyk`sTompR1bNm;L=ss`%icDKJquDfR!|u11jDQcT3GCg1nl6_
z%1WiE)IhLq!v^+ymiO{`7sui${>F_;rT0nNl$FJ}@ZfQluOT~jR8Ed7*LHlr(b!tK
zY{k=0J@bD^h*jl}+0@7U7nVKzV`-_WtL-oFlvq<1wdT18#|s2-As%(l<1b)OSwp68
z4cW0_s$}NRf6Up|+}cNmXd8xsTQ6*Gu1%LCTY<W-y&xy2u}`f=S9*HN!l4@?tDF4t
z%J2<KS(e~EZcA&8mqnM?^z>YC)$RGt8mfl1XEd#tgiT9+&3xjk1}ehY^B$3zNqF#%
zt|r46Fc~5C1DM=hMk#Dbvf)Sy7(!k13_346Foi!^au;HPCyUNPnDzFDcx^@I=R?=}
z#kcu&bwHchsj!O*%Cw#-o?mm`_!kMEn?)VKx_FKA8=ub++jF07mW+u{79#sH>{4(T
zF0U?VXaFPl$bC5#&nLaS8J5%qk?w<yO--SL9`}I5lAfNKOlq2qnkFFVyO2~h7gZg^
zrxDf>0kxZ~?`&%Koowa2nwn+HBn|nH?1EC)!x!7I7DEf09U0cvRxkDRZzJh5CpDD`
znTeQJTw1a~$!lnAsM}Eu`9e@8+`L(zH$KvCDS7m%r+%I9>obWm5NBnSHyhhJJC(#Z
zq0<$&8jIa<iX;-`G|xh!nTH9^;mBdG7JC*T1l`|POXz3s-MFG|#m4;L7D}M?>KpxW
z>xRXB)AG7}dgISo++o=b4kF$r+@MTlNkvPLU=y|J7H|V+ZVmc0y+flJt4f_R0obFN
zq)Jd}G#wq?Cwn@+16%riSI>#=4#gCp%@jqLxufGum}wwEqmg?o`>?0B7^X|bDkV@Y
zUPKdzUAwv4YVI<VtE;q#@Ue-?y1E{EJP<2e+4~K)oVGq?CQ1srr|uSOip)(+Nl8UT
z)#t{K!Kam!l;kPs4=NZphNv&vAbMC4B`KHy5E-$)xsfv3Q=&L!G^kUz7DDk&%gQ$z
zvORNy1(aYlvxb=!>^}!t#)9^=8ra-d<HOud7@}Y1ghIVcH2)KW*@qdmE)s=Ye<S)P
zn|prU@=YO(%^i<1P>FmcDlYCjGYmdNqs}#!ms^U9dz!q9ocbq(1<*gWO4RRwp}VOG
z?C5-{-DlgJjJIP83+2;uL1haH^R{ehZ|4^lN}TnGs5pXm!I(q@@neMR1RFOB8XE9+
z=?qbK14}cBRa>OdScQmNijI?1R=PAA{ty4~nr?$WgZaG^hI+6((Sj&G%>A96kn<K!
z6rJt#v$PXVHA$mUW}>b!+J**A`jqJzDdIj;&$q_r1}U^{KLQ_mVIju@p>)VPNp-c`
zebs)}qA$2-PP$sq)&}>l$^4UTY^#WMG7-LC6G!8Dxm;XdFA{CtDiYP#tI@m*yRm7U
zM9*N4l!S}pgzJbg0E{n|bypXrzBosEvWtb}ziNoy6c1%)xk4h9(^kI(g)796^^ptI
z(jZ6HbQOK^5Y+I!9z!D0HP=?57AuK<kb_Q@6gmk8v%!g8%w+ZtcaRd&S@NK{bzh5!
zcd{2ASbyM_@%hjPhn9WPdy*$=JARNrj2v|JUj_&#omU8evhMEG)HA4}rU@SXx4`wl
z-ORrq{?}xmja8&g0sA&tqezGyCd$b{m`09RGCWT4pf@K(=&JdOt8I`St%eAw{rUW>
z7toNNePEifzhK@Hz&u~%hmFN<JvNMg*2s<>gIQHo8TVkTTcWF~%4zrpd|qG1k|i|_
zIq;l{Ms(v|D~Yr1=x|*4skeu)_8dnCyH^#WWug@;5P7xYxd)h-t-SQeZYF?Ym1r#y
zC>wn769~<!Gla`X$e~6BaB*TVU7}yBIEbbA-BNv;trgfqLMeg}l9=wbVEc~(nb|lv
z`fGD@#S>;?E4=@)qE}vNX?f)pcW$oWY_%7isr?3lMlSE|1v(E87bZ#83443xaz%49
z7R94l$qApmb}?zYO67K|R4|Kh`Ym<)uStIO65@2A2OtvIpCaeRaer*Hdm`QE&VwHu
z7&HW7ATgkF@G=sq^$<iy=oh=Yg9-1Wa1LU?uLY6FR(5PtPy_UgL7FN@Njns4AvHC*
zx$Wucu<K%7@u2Ae2PA{R%JbMCS>A%Y5pA)grk0oUV7ce<+G+}O?fg{3S|+mtkq6vv
zHg}dc2LNNW?rO<OEq7_PEiL?#riKQ-+Xme{pSBXq5)XSIGBiZi)luC>CO3?+yA49U
z7aQlG#u1g3LSa>vPzVi-T;M_?01401W3PNEy*)#xgPEwvuy#1%qTAak6JUUwHkvvF
zf(}X~D&w@blV27e9s{&c*9DBrd$sTV(dWjqwXPlAUrl4?H3&N!mOmdHjP~Aw#lHj6
z<%pB-&pFoCFEotlJz}P5W$we6rGpSf&G1_kYzEC5K<?mh$BFNMw0P9o96r=UIhi-F
ze1BXaycZTHw|p6ri!Amm5^QN~Kibo4xnTDM@e2)WqXdE#b!LE$l5SWBC&Z0g*!HvQ
z(dW-jtbenpGvG(w3NN0f5#=vaIYgP+4`fs2@uK)J%r-w(C?1JrnJf{xp*V313C}gs
z5M*)(g@u&CHVPB8t+x;Ui)>JjFLXvC8RV1bUjctm!uBMJ%|``uD}2VZty>DW#0K+P
zq*4$w?@T|6)&PTQOxQ}7k$fg^QGW_q`uuZ1>$Wy85r}SEo^S2``h)&+$J+Po*s<fu
zC!d@I_+8NVGE5@>jfzwS)H#mRJ!P0hYNJ4~kufgh6PHPC8Mi3WYmFrSU2))Em$WxG
z<*?jqGmyLn+^b5}(tw--zXRZ#xY*#H9x8xaW=GVgi6!$Al#uX`*VkuzbT_kYT=Hrt
z$-#Ald1H3ZiFhxprlz1E9lpuxYNT`7MWuS4=Bd%7bdvK2z5tj_{0IbeS+Lpcbehdx
zzK`$|NbM%lZc(mT)6}$Pjk>M0IV0nxCTy+@e^?gImy-H2GqK1sz3YPbp<o5BTBWRp
z7PN+vd!FaXSFLKWT37jTm%P?i3LCDMTE!3H9BAb`TmCECvoN@<w6e5J%km9=+#Yg8
z_}iq3_wQpIp&9>kb6?3NhHMr8Re|7D_7#>w1X_t{^^Wp4poOGgdGy`61b)z(>wF?3
zote!!IfU|cwtigEePB=1b_it<>8$wkLAxtCT{UimSd1?)m!Rtrm6eR{ICT;noRWoQ
zqAYPY)!hIxUf(GS0^XczVb{sa7#*qp{(O*`d2_PTV8NKXV1Z7rx(3IM=!{=`RYSZ#
ztXQ6R-@{9DsQ-em6ZyOmu~M%ac^QUl&h(PszOdR5Ro`|gj2AIrGMNUgmj|47JEJum
z_Mn1if(U97#t1KUZ+(-^5%3^$FWheD3i&*;C5gq}*?JP~8Pn-bOiUa&a{Ta~W8ZhT
zwh?Rlp1a6Opt+SH`YtjRj6|J5IXPc7?8x>k4*|;!mbInDz~C^w28zrwE)<&V>K=|~
zLn|w*U!u1Z78e^r&3yvFGJ(M4(?*QIeZ;mFH#OaOXe}6~HfrdT1TTY9Q)?@bf2c4p
zjD)F~T61$W30f+QxZ(Ri803(i`phkqd}}7CBrQi~D>%7^A;_v0M8V4{v-n2rf<aoX
zvHsm`&y%-i%b9vPYXoJtV9u}ezQ8d)|CU^Rkj!kB-STC&1%zgVK$`s8D$6Zom^uFQ
zAXd4#_4O-P!rr0$3C-@Xzw`*`@I)CTNEwTxL|O|{sovfC8uFjS7z~Epr-skIZ_SHO
zFI+f#_G}VT#m-ga8tN7x*t(wr4AAM5J*tUd+tW1N-LhEwELd`PQh$!UKXb}txn7@_
zDhwrBk{AURP5R_CS-zI4q^h}Fww^L*5F;;ehKR)@So_Ag9?xV@+6VFkA)x-qjSzxp
zN?8O+jbJohRjVCdu8gIczU*=SI~#k)CwD)L_5~9WsX(Gdn@)rjAmwC7M_5q~>`ryW
zp@4clTr_z0a6&X?KZQHq6iU=$#?mZ@hTjdXUQF&Kk|2*KWi3bkn*loxgjeUt<yyG*
zJ`$OcuP?|N?yTPN?!VvOx_J@ox{qUbdD4@5E72LEyH|o|R|3JO_>Rxt+uU=ox$Yg1
zA;<)=8#j{J%M~j~$!ll!wjmfB5``3;V0?&KoSj*4Nki?-zvY6&5^o;mgS^>j%WBVZ
zk0zKUoHv@lVB-w5KuR$2oiA?4m)xBd6`LL(BNs8o;4mi_+H<8EQ3Fr-oXb}f2%ro?
zU)X-!Lq^2q0Ie3j^6D#PwhBmZ^7Pv3oYrhlE=SWd8}k2)P|-d{24xnb8hDOg?{KW2
z<te_Q6u8ke7AiZtzKHd}xxN_-d|EKZu0zMp0A?*1D<I2Z=g#_i7|jYF8Yk%cq^|Ds
zZyk>F_S#QQ86c=f@%V9|p>}vpr^3P0bn3J<1;V|?lzASq`OZduq)$>|>-Xj9BK7-p
zJYMUe)n{+b!@WpL=a;*iTn<y$4y@LHn;Q^f7P_K<Wayh31!ud*f{jLvCZoYH!fZZ&
zwnz6c1p8F#1dKzF{}M*X<0-R<gr8MFkHuo0e~q`h&PL4%LnEufD5$B<8u!>m`s!`p
z694Dn0kPF46A}01)zwy+>>N@LGI-%%1KV?vpMQt1=Qls)izM~GL$W{x>2!N{)n<EE
z1edaEOdoyr(`Nd=GYjt|mOcAynWOK(r_|-24uA8}M;|@=?6aALTfhqzYPGGcTbL+4
z*23hsNQtlGh7Hmdu#4Wl5G#<pwy?0a`|FR&PaZ`(-hA`TMT-_C1;ibakiVOO`3EDi
zBSUt&Nb+WpYb%LZbiMiJGtaP5i#qUK(2$<eXil*hOnT(A1^yn_(|Ql?e7mX&41a{R
zs}}qC-!{vWAes&9ce?CwX1?q+fAx87)7NLkS@2&j$x;Qk)-)b)hLBeb)11x-bvopZ
zAfR@^Bmqq<z#)7-1wO$(C>lm`_GD%{F9V1A`8+jr+moHI9&p+fV*V2ENc@_#U5Myi
zQPkD$WB=X^7u;IJ7ckC8QGe^P+B$;(Bx7loYM`eLULUMsatU(grFg4yI{WCVDl|Ew
z_Y;VykbiCF6IN3i5%7OAgS3h5`5e?l^LXX!W_j|+NA$-wr^|Q=&)<_?W?Egq<87G5
zL=b5VvDN<IE|^`%s;O5yeT*}Mjm3bbmbC1u|LDWo`izW*24(7`@rvQru4ah%5Pqe$
zA9jLzzV<59H{qZ`V#%B`BaMQ1W;DXM`Q@>bV1juPDHoY;M3#u51IV_S*!Psx)p>bs
zZFzas)oOM3mpj3BVY3SDX@3X0rI*(;7Gcj>156WbW#~*MypvrF<R#hEncG6n1!O*7
zqR}{=VTw@?cpV9%v-yJuYHQWv!a|)!a2}@FqgP3QeK@b?9Vno8wp*ciSn>#x(khjd
z(`vPzvjs`K(lp^`1>Il10~`Wn6Pxz`jn%O}4d2fvBE=F=k@%SYEnA#U0^R_2)fFm*
zFSbV?*OmaNdQs~><hbcQ)N#`3aw2G&1WtSTGoo{+Gt<=fd*vwkVpY^7cW#DC&Nsm`
z|I0(}@6gsBj%?b@e4Z#Z^&zi4o1@Qw&K>*eSrLgexJPF>3Cz5B(Q*08Rcc^>7ew+;
zj*uF6fhHfYhbWa^tkk1EZtYut%Jw`Hv}-4nlwSAzU^s<dxs6^rV5QR77jLVlCi=ek
z)=iD+wyn|vX}K+5-rn|=@w0}erRz26(<h}QLV^Ji$|q}A)JRz}r|GnOAP)3Y1Z})+
zd<alRD$^xfJ31<PXO6*udHS3+8rG$T2C-O?)YOY0&+uc-AH6)Ab%FLgN4*Uj3SFi&
z@}M74C9!G4dOB@p=%1#*DBUNvkA&uKK{9=Le~q`6Rj6fnh9DJFIuUtN<lWyxzuW%3
zbG^MQ6IL4e>gOg1PJUTyJ9UL}U51g!a@IXS7PHVf=9!hRZv4_P1q!Nfok5uWjt;B+
zqSM+BB&8f3RF51R%ze2OF<o7BxCe<JIm2nSwDiklm%Vl-e|qJAO*AoKa{0=YAAkJZ
zb0c;y%BHplrT-=Fx#tYwU7`|YfQZUcNuv9hklNFIsH3f=O{Z&a&dkZ3J!NdyH=kDa
z9Q?BRJ*4$QBW?KeQHMJvLp=d;fB3=>dVT!MMBar<MDm}N!5&0?V~ob7OU>q`ONl!0
zE@V6`VTk=?3E!QZHxqR*wpX!QEx7Q(`t|FbYPC%Qc~2!qq7Rt9sr3{ra{x-TYuRFm
zEI?tE;G(T8NP)+^8O#X2N5YuUBNE1-&ruPQ)C`0yAhkq<TAikDIX^BVW86ja2cVrK
z-Nb6npWZc)Sg6}3-nb``q@nZT5W`d2f877c;HAFKRxmM$5}Wo5c}iln8gI92Kru~y
z_Ev!a;SXZ@DD{|eG=1&DsovfmxqRsoi3G_Yco_vEGgv1`&D_SE9L54IhhZ-#2WFq*
zrH?G2a_)U{_3!i;<US!Z*5nt82BdSP`n)ZE2DO1b;`R0A7d^CWO*ZvPSaZWpD0iU2
zuv(d%$>Fw~9KHTtO_&YtKeIVvXgqepb&L`GaycTAK~hqJxT#qMk3LyZvAZ3DTg5J<
z1FRtG*^-q6gdyk&ePc-p(>`uVNu#%JA{TKoGZ{bRX#KKhpMEe+-12qH7oRr3?WtnZ
z%iv!L&k9~Fzy=ne)$v~`cXLhbv_UKNwHt9R>^9abl_Mx;oZ`H#x*I4ASJSA|H4gP4
z+^_+;k&$K?X^}S=<Q4><4D})8Ux60hB}fF8Cj3l|(6CUPm3m+zMN@y_)NW+7wH4R%
zJ>9OQumZ6e<w?jymNC7jP-k3=Ne=4B68JKf-|k)i?^URuf}ZqkxPNzaAhQnL(Lwa4
z+bE5J>wrH&$SNno%r<ZKW-9Z4*Z;ZzU=S}G32K?d#LGfAa|-dYv9K7h?&M<-FMHS*
z@%hean1BRPqxdewI7twd#2*bUY5Re_z5J;e+IDTm6#fri`y!J3>tQSi4GrH~p*@2>
z7jDuTO)n?~ST&VS2hFkN()rUifuNh31-g?thlV?%?OeZwmF0@i%C#I794rtJ386Y$
zBr+NsKY|9iwU3e`fIll=mz8C;?)GXhId2^`@DLqI<t|zU-FjKn>eYW(nVZNY0!_?a
zxn}k1GJGBU+@f6SQ&<ccUW!NR$o9Ns@M_t;<wZlvMMKJ^zA`HfyzZ3-yl-EBn!Wp`
z_byxV={wny&$|?M#~_**XlEm&trO}n4P8SCa+x*)dwf~)x!jPg?P;a$IZbIH8wDFF
zYfHD0dB%U#^U<f6%F3dmp=;rGxxy(+napz2vbZT<IDxBBn7;+XQu9&s_ico6xgwy8
z^0o7S4F>%oUEaL=;Ic8(5N*-TQ@JgN_O&*@U)OdfBof=MmwF2}HgsMdIW=v9X0$x|
zBFkwi0XI@JI$5=8Q)+sFexibSz!(IT+re%U#}CN%>u-42%J6jf*7aFMolPq3>ZZz4
zXn0$ispj0==A2xpTiJPqrIr%Z+FYgSa6u~(cFSQE;KFaTEI0#su5COP!^l1fEf+@&
zulN@!9dUf-Ky8WTEr)Jj4)efz&kE1Oo_U<t|A^edo=5S`0`j5Z7<mvmbAgt``3^1x
zl0c7L#ucBL8DfQ58%B#HBQtYWuB@qP&ZhnzDro%RqfYxkn7|NHNm5b2-{=c1*+7o=
z20M}<55&;t12N8rMMV2zh#xx9zmVwNgh7Nu|G#^ph(CVLBhaHrFj3y@(L(ppR&gxf
z^+ToC-W^)i=RQ9uXz6m27~dF3bj0y8Lh%yTZX^i#6Vv&J-h~P1ZxuWoSHjMA5Xdq<
zIw~)qoIIrbXX<12KGq(|UF3P1s%QG{D@Re)0p*Ax<(MJmpnx(}k?k4Jybn6Sz<(BB
zur+!q!;8LI99m|nflqhGmV9%3t9yY@{)bCUUrMQu1IFgjA?3LNV*$^KA?4&DW#3qO
zN8;HbBXgYs`{rh7eUA;Pf8G%IL(8fm_lE_PDOjB}0dw~dz6hAR=Z3WKI+Z85eJdT%
zf8u-JXe=Mn`=OW|7SOwF&*&j#&5&|pK$$|gY2au)I&?ICGh{S|?1+IF>BD62I(^Hx
z4!z}acykBfmOh|0(9sVmKQyGAccV<Lr8Wlaw|*rCkH`Pa$j!b1i+O&bx9nTftFNzV
zc3t}5d-3=>R<L(n0d4W^_TGM1`o<v?7{qa{H>{=LT^BGC2;B%M&lpnv?T~Wd$Wcb$
zJhFI0Axx2-*TA;3r7Zz%^~EKz_X10SUX6kG{<kHvCy%yVg08rIl|ngQs}oXIQ*U3-
z_dN%}`5m!}5J|Y|9#PQAueO)JyVrr_zp`j$0nF(Lb{Vl^=?X{xal56X8+*nufp9o7
z9-0Ib!P-CmK#IF<TLsH;tVA7g*0?D!$~>^VRG&QsI>`C=qz=2-D~-R@*#Ko66RXRs
z5$bgIcAUTg+lJ?num(#C^aj{_KL5!u5oX~_4xxM)FEm0V7Dru!!ZJadw*;9V@1L6L
zg609h)gh)Ypu0M&p?f4r*E33tL+#>MbBpeOAg}S`59{jQ-SO!c4a*;%8tF2KMJ9>B
zVYgnrh~$R0FYCLcQOWZPyf{HDPN{TkuFlRIE>DNrYT<arc+!V4u|clW#0UHRZFSY)
z%HBU<8;3y}=iKAt;Q!i6(hf$5>PZUe&-)NMB97vR2rijOYDn`L_sRWjT|IqBy(ew>
z@F&q&jMoI=<s)7_$Zt_EH#hBJFCU)xK!!Y&my1*<A;+Nr+i(6Bl6~hHccLc8El{eF
zMwfB^=iZmDH>jcivM`RrIq9b3MB8%7d3lfy6CyVTm>j9Z_%Y!kaV*Vs{TJc->Aa!`
zd`Ei7Rxi$b@=54`rb>ERz$$%uRy^<{q}e_1Rd3$`Em<C?xzjL;U$y_FLy$7(p1i3e
zlhhL0S+HQN=$iTawxg%7(1VV)&2PK`?TRl)dJgIkloo(l#c*uv;HiE?<Q{~i@PT6c
zI$nBd4T%L@NB!CW%}^XNA>YX68kU<J?h-|-rf5d{xDUe5B;90Z(Cg<UQO7^4aVO0y
zEUe!C39_F{rAcx0$rBiclNLH*q_n5UIR(==#aUhb9a33_e&@3{5Dk-PYjZ0!V$;<r
z!}uC$gt>X^Rs)>i1cKtH5yXPj!6fCsEZvJ6gv;bK=}BzQ-=bIEm*FyZwlowM&1bt<
zU1f-pP(1B}t6#F6bHwI21e0dV%V;qt#K$iaNlm=9^{aiIT_$06@zZNwc<PaY^2#sI
zI3cD|t;iWBxV*QzN)*Yzv}aS*CAHj6b-lgGp=8&SKX<19q==f@Pro|Ywc?>nxo~SP
za{Pu!B%HHWS=v+C)T`W;GKbgyf<&t)@e(Jb+k(fWKZ<UxT$!$Z>Hqv5vf|@~#WRwD
z6|nWM&G9VpD&L=t-a<UmF-yBX-DmFn3Woe6PD2JlFGP(yAjTU_R@hsCl~#m^-CK~O
z(@s<oWYCUFl15NhY}}{_kMn}{!c~eE#!80`liu9^sc4gQ%^Jqf@jjjG8lY-4%NHnw
z0#{J<)m8(wwQTs-^5rkSu*`lop<pW%#arLpw2|V#(BhqoY(Pj5i~!0m%qx8CdE!~p
zbX2eaoqAd$I1G%dWAg6ESU0>gW|BD;d1L!584Ex?6tweepZ4kjK;lbA=4#ynVQSW6
zB#57U{v>&!=vVWgLW^}Z6HWYUA9VfHg)!;4Y)F>Lg6WF4pa^feEKV9bIcw&ONvUz?
z`?Ok(IC#KrK524_2YEisCAPVdHYXf3i~8(I@XTMPp}xhtpL7&GissLhP-jV&(T_U}
zBgNv#u)4b1+7G|{i5iaw$GHbu>Wn*gn=B-G&^bHBi}ZP1%pk4K%d6V`^#N<=$5^6k
zBJsk)%*@fUoUtQZ9nfVN;z*vIE&r%OjvvA1K#cb0n9Kl_;h7}o-ei%T94i4-9?Nre
zkPbs;qt+W@1qlK+cWN|pK&0oQk|&DB^ODASHTOibsB$4dTQy;NR#s<cJJiP=9lCqb
zGi@mR*-ki|iMWx~hbi7?fa4*b$LG177NkZJ&@RYyiLpG%)Vy4wG!f2tPEAJMeJc8V
zi+7eMlJ@Br&PsG>AE~fx{^HQr)fKgcT8r+Fn;nMXGEwk|f*wy(nWgrvU44D=XM4&T
z>=qv-pM`f5)6*{#ALhy$2~KN}rNFPIwe^dA@PEOoAKBV!NEGDGPoZ2=sN-XKJo^>5
z&1tq>xfY735`k*Y_4Cet16bpvBvvj#o-mBMbjj|3KHh9u`oiy_>wI*MD)c0%CCMn^
z?)&OrunYZpV}${>-@|+hChl-&C;q>{Km3zGFd{iYJ_5%MEYn=$_t+((8MxGqkjs*j
ze3;Gh0;vIfg3rdu-U}?ry-yPQa4Qa(eaH9Itx72Yb@}Yg?32^I_ZV**<cW|%L^One
zieM*fX)}a0H)myuA|jCxH8&y3#jbQES`t}Qv~??OJu?_1rfF50^qL`3?lj9c(Lrz#
z6Kupj@rHMfi9R&8z8;u2N7Gbj&QnAH^uD8~%s`=g@&OU92XF!WB<7{x2n2PM{rp9$
zpW^h@36}rig;ib{I>6xDkGxw;T^=ws<ERiuiyiry>8jIF`%y0?p8}KJSfPQ&aaCf*
zw{$q9+`=Tf3k&hTH&ftmWU*0vA=i5jqn8h*$`PXC&nr?>@CCywlHE;SRD_cW?vzzr
zE&$ybYW==NX7NV=jn`LF^3=*D@YY(7xFqs<HDO8aukx^XVqHxY(CEGSPWk4luYq{S
z4t1V)UFb86mIz!I&h~!){aIQuHB&on>_}zwC5vv(c<I&Cm)s%IF+mPNTvBRQo^BjF
zGt7DtDI}Q>V@m|n^u>9e$L^NPsj$(hWcz`q*=v=wx%smbg1V1k=+tO5DLx&^Y_u_c
zS@W(#JY`?*(x+EG-PfSXOK+&%zHNI=Veza;J4RPA`d3<nH+~)?Yj<&juhj@N?G}L`
zh!Z9@{ZLO*@a3}YAUkPnOD$jB(IpUdxu(N*_v_*!<kyAUbO*^<u5_2hsHK!lGeKx|
zx@=}w_<+@EJ?Ntg>+n9O{%tKhO&%J*@xT9BRkeNlb}T=8L+dbY`cL$Etof=Bb|3x;
zjNABB7s-56I6q_LFc=XJbQnZkUEQYK1@L^)s->lwye>PHO?|_W2}!7eF!ZwR;vN#;
zr}d<I=7tHEKmPa<nL;X;-7}kd1BiHn`erR%@wYd9nYam`JH<wg0DF7y!o=|l3W!(U
z<OJGf{T6%DuI8@3(0K@<EgVIC|8F>E{JFM|I~iwc6h=Vv_O)S(R2t2(ySP`)P!IQ=
z6T6t~%vT59R|dk7eeZBg^6V09So0D^?oLd|-X4cGO<-?p@WtU1PJV6pxkF$N!Hv!m
z^!>}L;)(rR2R$BdN_thQ-eWNRnELrbImRf>kb+!xB4jP6#sLd*qFsZXK!t91dST(>
ztIjJnr#k{RqWBp4ymvJaOXL_`Tp~8ZF)2AZ7uM%x<i-mFW(u1c&3)+Rci-#vnVE(9
zs%oRLs>(1%5SFawJ6$1k-uz+wh?Z7C^PWLEDleC&Pya+q#33?aQxVT;<`p#|SsJ^-
z)}oe)Cga!p4s?RF9`D_@4<mBi?DSZEIOpa9>-pn~9)ALu^1u|Si$oP8^P#8LxNWZI
zC6aIwpQQ*AdC`zj;c!7+h#0bpifhLXS_v!2QsC3pj^4q+==iuB6xec|0!!mW*Sswu
zUR@)b+GRQEr&vt-(0Jq&gGtmd)L(pGwps@E=TMh{<HZ~v!5sV9K+NGSU;X^bJ7Lam
zQP23EAC4aFJ@O54j9v3&!2?gOShni9sp13Rc=ihJ#djFo9zkL@h~a&C_ZBUC-Z0*J
z!Qrr9;EmJqlV{G$o3FIdy8AHq$*!~5z|RQyIsrMNOD3Y6j($Yc`*Nm`_4|M8lqj;-
ztY6R0+wHCihr;RT1B)*bY*M9eZEpVhV0TWlcGkQFx)K8y{1|cdhqibYd;}sqj|OjR
zmP(udGRrf2&^fq=GO?6`WCm1ptJPKAm-*S**;6KG-}}g;kG}KHJ6~R&^ys6HE|y;Y
z;^miL-r8n>M+dfb^!w*IK6Yn|4A}L2r=zj=0#8*7JV!0?o-czM`(EHq8t6wB_~iez
z<lDy?O$diC_nrE3>vH6f`vXJD$wSJ+hLnfiA6g$a2k)sfZ2i0;<r~kl;}fW{vIWe5
zG@ukXK7se{Ui+a4fEma)O%1pgICFvbZp{&~Jqm1D;FEVM`LuQHxXJ8+<2bae7;;}Y
zq&#&<Ib}$B=>4Jarw@5vI;4E#dA8TS^P<Sw_wk(U9d*E{#s!oD?*-0AAio<P@Lu4P
zzr6G~p+s3QyF&RutxI_@D;cKJ<YWC{=lhTC0SCAzRWCk#3sr!G<QO)A0-M}YY<4%U
zVGR}+Df4fuDIhgQ9CM-q=dsUG!RJ|vy|v30XC<NbBx?@$t6y4Uz7jn3ZFCbGQRB>?
zTHk*A?be^@v17;5%FtuSj)f|<2tm;ztsP$&CKR@Pdp=kY$q(iSb9Fk^Evir2yZ53n
z1;?qT2rusSzaBn3IwB%M6f?149&0GZNl^%IWj?QqpMG%oEgI(2sZwjNp8>_<h!6MK
zQs62HJSD_;&mYfjcg;eJ4#1qjfKnhm{@=BNt<6V`*^-Yg-L8dq^F84vCu;72srMd@
z1P*wT_A_uGIMe<SSjUg#b9t?Il-BQru<=0H+Y9sTUe46sj$AV$OZ!gLxKpzLEm@$6
zK}%wK(;W@}^V)&DKUaGN`krvO@wL}pYdm}zu_Tv=yTE3-hNppjPjg+mG(cS|Q#S26
z?i}O{I0u{#cwyeLJ^wRojt4gXTua&c{{J&Rz3knW3C{Vhy{%sVvfs+I2b5$1B|$*R
zckgD1@_^^u-je@w<oZ_p;xK%_ZWai{S`TI5h6s3CPYPwaRbKohcv9Kj-9KVneEg`9
zVduMhdSXV$$Bzi_S+eBu$A0tB`t?hnS-BkY%pYbU1=BX<@maNGP)q$4pFu;cnuDV>
zI>AjpS32AD3AAG1=7pjVk7pfmo8a?kIxL*W3*(0+BzS#P?niypGg?k4Cp7GtyUC|1
zo;vl?rMu0a8Al?|*i3ySWS*b+!M!K${eYTyfZ_?LcPK&Ls~8l|tJGR%rY4~&*`9lX
zM(K35wUAUpk2E$G6ukM!vMR)-4n8#4NYXZFIqz^D;X<f`eund>KjD9`G($W}=m#HM
z^d!{6!sW|bTEfOGd@v;?ENt{>g!62m{)@ACu@+eg-w0kj+C6KQ+YsDqJ`53$ge?m$
z9eeG?7hmh!{wL_PW@cs0OZJpe9+orz4d9MIR8Hm*{G|WV6VW%^<&o$tT_W}h%AHNU
z7yQ)QzoaXi4|~ggzB_Ck<52|{r!;3%W!_Vpe-~79I5Pf6oHr?0fBooY?#t`h0=14E
z2cijDo0tUQIEm`Pxb;x+P%h?<6H@2)K<W9#nHEQ$ia1$AxdBlW|H~>XyZNEXL=k;Y
z5}(&uS-nA}^UfAl1o2`+h!~g4ThnT*=qPMwf{&l)I>_9vXE1`uay-%G#gImq6UO3g
zswTR$puPR$e{O{?r0P3Iv@)%I?>Pem!Ke^H+@yuSgWp=@iU+k5M}*kE0z>k4h2@gh
zuba%{IR40)i0E3VOP_x){?15nw|!LY_M9<)NJ2^Wv^90~9Qp{a?K4@=I?Jn$7{)<g
z7A*`F3Mr>+(CH#)Yek4wtbiDrDRy*$Y6CA)!)?2X;=a;J`*h)qUr~i_wN~5MxN_y^
zjo-quXvu@Sc?Mm^$)mkvz}R|nanRaB4``?5kf?$XYun$;Y5VcEcY%lRw)bA+N$cwk
zNziZSOo5I{C6f*#$D~KA)~rDj@&#v+YsUDs(dg4H5bD8KP07pSMJm#YUYbq4%<zJ1
z<<of`P({|6dD+jedFlT=n}6TEaG04ln!0eP8R@>i;~8+sqF7W@16|EeGgs(JKwzdj
zf7}b`5t%P7grB~Y&qnijLR@{PgMIykHyOn%EMH$=4-Y(^P5tVt0)fxo@I3BK&B|I-
z{9BCCv?O=mH!!e#RQ*0+Hhy-LR%Yj9ZP>7DK&X5GD#~pBp$#im%op1Bq2nEN`U=BT
zUbtAy86Y{gkrIeU_q*&oxBH44Ne5(+oI#<jv(;#9za*Q3rcR?&nbWn?M>#uSVCeOZ
zt{(NPL}TuOm32Y#g^t=S>;7Ev_CMMNg$tq8y<d9nv)cFGsxy8=T6MVJppwgFBgRf0
zakbY{`amZA&5?%fFfAN)CO~@@a`4?o$lRvWCUcNaOE?$M*CzX~hEYKC1@Y9$dXU|^
z6HrK8|1ltbBydqA<g#nxqOo!#yDIM`@Dy^ApD;~HnaGLT`;Jo4bD>e2Bcg1dVfJ>N
z?`X?PH;j--B6xfeIhf89#vcrkR8{Rd#ZSq{8YxioPwlFLT6hYTPrdFPhXMjwJ!&!w
z`G<A}G7>0W|Doc3g&m_nN}aBS^7VbQAyOg<8Dwh+7LgkM3t4eU9rs?Z!wWmce-#&1
z^=__J8=)BWJQ1`-B%xkn)-cg*tPPD)0)NM(YF9#}2+mZuwY~Rp4o;C+$;OL^)OQ`H
zM0It&1|%4ZplJ^sGvPtFSVpcW$`}!T#nOT`SzcogmU!!BG<7Z!$q2PDeac)c9i57Q
z`~##{a66hXRG;@5yKD!sC_m{vcgC^xU6WyqL=r3Fc|1@FIvtlUg3sfNiNbN|Sl$U>
z0E~lIJCQfJqP)r;<f}!FmUEhmahwT~Id*q1U;cjeZZxo`v=qm6AgN<1Nw9>BvLr@t
zc)F^n=>Ed2326|Zl+Nzv4X@aaA8NpP2+YwAA%Fj>=NgxfV@8so?aZ8v!59@Mjd7zL
z5sKx{!ML(OaHb7Q14!+)BmTd=mI;qSYqc3Lh^@vM76zWArW-b5Z{C30tY)Y|?{YOq
zN`|cFAh+Ae1WQB(d47I1ll74-8h0;N^K@57?O*>~UjEOPE6iHfORT#-+WGennp=@(
znXKhx9F-wUlbq=`m*_8TwS4|BY)?J57=$K`GBqDU2>6NKo>N@kEdM9h2N$hg|1<i-
z|HL!I))bFU)5#dGv2;(RK;XGbhv8H*C;q?c5g3H-bJ)GkxFf+KM!U^t*9Dz5yM+<!
zI8nZ=YmtC~yaSHbj`Vc+yEb$QS$;YV-%G^e<eWmO(B(S4hnQEoOi?n+DYrmu2<G9`
zkBp46@bnRUr=<((>mwGyI;x*tDI^X|<pN<BL-h{hI>&YH$39WyIm=hrcn<V+^jZ(#
z!gMA~(0lZxp)7QOa$UvAtFHcau>(%USZpK;u>%uG_5d4D?_dw@W;ulfKo(|ZBJrxJ
zNt|WlWoF7`hV`MTZY%QoNK`8JDJB<?LJ{jYakN(uX18`Ah~uEm76o5^xhN}ZMxruN
zVc@v<>mqqB#duUFL#kSG7I{}$NJj^c3zB|~?FrdkQf$}a)%QMLs-GrHk`q&c(0RdP
zHk+?W6eA<~*5r~0vg2Kc(4h`PqzTsIjsXRXuRJb?7ZS8C=+cSai-YSTcmiZQ;l#pV
z5$Qhd^LuCZRdq+T-HreZ+9hjkEu%WUw2x@*$B9G=Is6nvB9{f$gA=AR@EAHtMPdiH
zi9ij<(Z>**I?8$g`(Y<Froa&57K-WMASz@%FNDL338SvAlep;%CevAAJZMq0BV?E~
zHpFono5N{uJ#$dch~Djnt#9sVY3l)e>s7u4HQH?avmch1AEriw{Tpu(C}p(80#(n|
zD|W}gx~P60?}`<1mjXLFY<5L>BXfcPZ9Nt0h5er^`8c6F)$pHKad!h#fH9zjc025|
z7KOs%9Sg`vRG5OtQ$US%%rt+#A&lG%ZSEb8LtVYk5FctvRwkDnKFmGbo0FqZq!m1>
zl7)s|`4;Tkev>^>?R134DGX75@7{4G63yL2IZext90@+sr_+VYl9Q7Y)#{X#l<4rm
z3wFhbL08wVUAum?_u1S(;^Oi^e1h%lm4MdW<pO0noIj(+nWXo(qaAlc7ifks#F1VK
zhs(!nwV|O|DI(9s-fzGC_V`6HMr_^P5e>FYMB>D_4)=}g{P|9V{I)6b+~WE3kw4LY
zD*TBF`!?AVr-?3K7P*BaF$282;^Knb7fzVO(HE?H!9E@|b@tf~;=+6`BC_v?(=%sU
zt!8txL~G~IoGF(Z#xS!LN{4be9*>(#q4~5&Jqd0qYS}fg2H;90PFOs<P2s_zH`h<H
zVvUvz3`jaU+#TIBXO0<@<Uae8d05QA<!;Q_(f)oZOrR-oI^A4Vs%nfu;D;P{L*rJH
z{)Jkt5P#%Z_r-)d949^li}YznZ`kmkjyrkqXNx}HbLa>!$a<;`S%CIgEMd~0ev*by
zojN5p5!tB{?l4c(-LzmUTEO7U-QWg6ja$+7Fn{8n`b&oVXPU@j!DDFSgm9ZJ97mm<
z>g#J1O}Em}*df*2u!}!_e~ceuv$TT@9{{H(J9SDXoIZV8d{VsJAZA;FIiDQt7(X8A
zsmF)@Ooa@$`2^}q^IN%8#MtS-UcK_s1*sD?;PJ=GLpiSg{{D!Rv`Hi4-03SGyKijx
zNuWcYfmt?Q!t%@ZqqoRBi2M7+DJhAr)5lHGQT_HVnCZX2e0g}X&6X_5$(gNGD^o6W
zlQa)3Us{x{P$k1CkR*0rv7I}2j+>A&dRSz*eaw=F<|a``ckbNTWtd31&lREjLaurW
zP+rMBxqtuu!@a$sp+}A&$9r&a(A_qY8JCr4V}>6cO-@cvO9fF&7P=woT<IUg=6zK>
zGI{rIIvFzKEQxJ5ksMEejWCO6E<vEt4YK>ocgqIf2vH(Xff%_wAt51zx^&Vk8FtCG
z4?R9)?lRdA;o{7NsA$vivl2;4vK(YXLfI}~zI^)31?=N-XL-XUAAWdQf)omq_<bKP
zS~S*v2n4>HH(`+>?U!KgFX;d?dhht!|1&dut7czswdfYjR^P=u|Kip%b4~UVFOT#~
z*&$}Gc`2Qp0>O{Jq;ht}D=}7)EiGLRS~M2Y-dGXs=xQ-Urp{OlC%45jQsqQtaa~iO
zz|(6ZMnK)95-7TQzrnKUGYFS0o2>}i_lnVIu}3L$mU;L2bUYi2@8~$djiW8V&BONV
z1w*W6K5Ce+kpjMyA3GT^NANF#kH6@BpYZ*0G(xzt65zdLBB*`+Z34;=S5Pofe9ibu
zXQ#;_Nh`SiJ^|*1!T52)laG!jj3I4@LfJ+!*OaZZqr*l;5?+<E!H-j-h6|7yN&GAp
z-tDfvP?<4Xl54E;w#N##+G>atAf@86Wph)9qesJ2=laHhc{;bGs)<IT@GKeaeHy8{
z+Z2ic(|+*O`%ME1FYE`<>lA0A7TSQ^IXF}~3-Vt1=08K3(QH1;6IE~DYYPTt47MEx
zyMNe_0!OtF<j=g~i9xP<<ycM1@P1Ph^iFTp@`7ZyXbl73#sdf*3j#9J494{{-f`i(
zy%Q%#pZNSuSaaU^>sENC7*hP3*mmOR_djBa^PuGV-6C!<c6B>P7!?y8e$D2~jmYw<
zKK`EceXa1~5$yX%E>ck{RHlkDM1l*j`%2WsZr{5EyM#5JR9iz$I@PAnO{Y@z2B9){
ziP52IX=+h9j7xHr-f<#!5zgDc_1X$^P?GCQG#Y)8u`x5JrlPa6wz9Lcu1cHN+FV$w
z(db$lvkXCIr4l-F4$+f0H!GE9b6y^M7EfLtZkf$q*bHx5V5~*Lc}uFe%l})WJ`0C6
zb%|4b952x{P{;_jCTK%tCGyDg8m#0<hDuhlV&+~~T-iv)K>&-R$Rfa-9c6+Jq*Lf^
z#pJg3UiN7qiC<XCE$1fAZadUAbZ&5v-whC*=T4qD<BlI+IEylnPt>Z?@W?Ayu3QU~
zKw~$I*Wbyuh@Br_PyIQ4rq`zN35N{yNp6Dcj=lchu5JH%{jbMj1Y+Rhe(#JdV=HvQ
z#;;qASvq*#BdMzT!6o@=o!?&xryTs0)b`J+b{zJ$k$G$V`qwKPjagHtVa-gNnq~a*
zgEtX36n%I{)o1OgDSQVNe&zelwEQIxs=aS9D?z22Qu2p2%b%I!Y<K5AvwY1TemzyC
z>ieL+t*z<vxBk`FSMm4Ho7(n%xU<iY0*Cf(4J}2JMKm>VrnM?_LGh9$4hL77meq3L
z%euP!{H#e?&F!b5k6<Y)1HW5-Z{C#joY_V9%H{VJ&CW@my6`@EclS3H>(+02?Xy9N
zz3wlY)~~DB+YR`FL36tiK=P4M{L2F^+O$~Y#9p$bctK{>S7!!ko@i20OT#uqJFxJ>
zD2k1YrW^8><<a8uQY*9j-S|GYeseu-T~aP)z7K`K#wGSEJkAxDV__0~-}P_tFc;6V
z-}=u9Fow&apzFUyUTpv8{?pey(hgWzy3`zPGt+k8Hx|jzZ|p+fHzB~N4Ro_s+sxv>
ze=B|OJNcY~4hEwCUH@wqieuF(2MZ+}Q2%Q!jZ;nOGK#(#vj4kxbFBb8VQXuzGZJD_
zG%f2jA89r4mCSj)%BUZWWn7yp$~*7bcs8N77IXu)K9_~p+Oem$r^g{sfWnNHq&eK@
zdjeh+;8h=_pKgYP(<mLsq1rlX=@9V5%7TJv3X<uJCvjzHW)>LuJ>P)g|EA|sQdXHj
z6mr3@$VJh=O(aQqKO8&<ZTz-QV;$2bxNxhyZA!3GghRA`zkNe%Yr~#HLyBHV%t$~r
zCUC(0)kc{VvN4$fjqt0LP*))`R-*3;g@}^%nhv)ZIGv}~(Wg3hBU1C}pxWB<>M#5t
zgD}_CRSL2SEKaAT0G2sJL~nVMBFk;_{_#FL5|DW<&XR(H5+`CRg6j}lAgwb**xXr)
zrt)6zA9mgu{Z>I&ZEaUUiGgP);4Hq=gP42ewW@*wRV{2v7V@^;TZb5Rgw$3^MOIT0
zlv-VUei!-m-est9drO7}yUmSn?*^CrpGiR-P;<qPl)HL&fzjC2>ynR*7a2s#$@3m4
zk;+D=DrHGyQZ1Lb!@Tt}Z`f=nf9T^W=E8w>u7cP1!$}+77{<M1Ngb0UQ>KoVNlPA>
zH`)8f4SHkOH415Mvv+kOa|17!wU-B%+q=}l(qMAWdv_LEmT1dJ!C9L@*RD96&Vj+J
zwyv(u_Je(wT>VG|%;%mnBxWiv!ws1I+28U-GuCnTY#n$b=EoP~wFa;#<b<B#nGq?Q
zv}|eV!yKcX_@|dnX1^87n*R=Vsc!3Es=B(~`L8!WsH^|@?Pe`bXJ|m)Y6DX8=6LkX
zr_{4Lv?wW<8aUfp1(W07aHuLiBCN{HhnKhP-`hk8ea^J>EadX#t*!6rL6(pk^Xcm^
z)sW-KrIz9MbBYcol-bJCU_BbD^5;{!x@s%Cx&rwWlBIL=X4p#o_!ZeL`4zF%N{;gM
z#TUU+8zJCuX_BlI|AATxy4B<`pLe0);5m;Aw9O|5_)^VGjns9b^9H5&c>@y6@%f#w
zE!9Uw>G3;MJ{VdysEWJ0i&a%ss-bej(6V=|NKL+4m*aYE8cFd#j5>ua|J2=czmxnK
z++WHGkAUDKB>3q!UP8TF@AexpVRvzr%1j9pC+AE~^q7y_dCXowPdFrb5Pm1zR`R*+
z4MT=+6vgA<&Noi%h6dYx?wS-bO3*om-@ok*@3`iW7LWse^m;>Rigpofp^LOB>%yWn
z+WVHVh4o3kTKsuZDry{@A;v{fry<MuwC4V^<bO>pmTT31Gd5Hf8d^4t@t>%na?;SU
zw=bZbp&=CGRDtX+asN13s;|-4gly`_6_8nB_Uqw+*`WEpr}Q;^@|<J!Mv+m+Jo*0>
zmVJP+n#p4HeBJEqK!L93s4X#Ph)8!U>=)l!_?yY1{G|)_foMkho0u9uz({|uyxwRT
zBLCfLd5G^MDgF*E=j!#VVdMrnG5NKAfncifVcD(fy>+$MIWQk(@TW%_v}RaGYHCNv
zhK}Y8m0g-8ZE~nP>&#TUfmfpJt<I%u%LFQAQ@5zAvPaR=l8V&2H$hwmC~5kV^z>WH
ziRtM?PQ0c3;YW*8Zb^%&+sh|g2oC5=)aqN;s#cR)e@+q!y&;PDwPnUF+j94!3Nu39
zUFpuF($J8Xw`^Hn-p}t!R<XFeytLHn9e)OTT~U4&`SsuXZYiS+GBSus6vtuSovhA4
zFKulpE@^Eo`B~_+LV@5Fo=SC>(ErWO-0k!VO?)zQw_Dv7N)pd++Vf>Z+wlOMjo<Xm
zyyf)Wc3Ho<qurc0K^A*MvBSNgXJSU8s5C9Du*leyl~Y6bit6s}hAM4AdvoE^^z_`;
z<{X>N4L=R$BtMy~p`na}U;LHHES7xawq$=Tq?n6;<c@b1=V2tMrDcgtw^Xk%1JCGK
zf*~V0pkR^S)H7zt{j@A(x1^Rrb26gpXMmEF1JRF+6X9UQ9Ut#Dw(KQJfklr$Z$M>s
zQrWoMS4Jf-T3(uG?+=TOxk+!}zo#*lON82qJ2rGOs-Lh_Wp8ILUE`+#@_*i+Ho<d<
z;yt_Kuhg=QOyaw2Pxkf9cdWj}rxRdHK1(QD;(14yl`%=GK31;qNvB;DN7s0rlCAf8
zU=Zzs)Mz~8RY9Hs<8;VQ$Be@U6GxvhmT`_8DZ|;hDJf-~l$1ZNP4Uw|1@84nAA0EM
zwo#+D`TlxaL%bEsI44e&(I-wMCYEvV8+p^VF896U`+L(?&a|6wCE@;Svy4+)Sr!q<
z_4?D5J8I_$G>*QcEW}4m8E7FsXk|2mhklFfuAmL`K?|fxzYrSH25z7U7h(wajR|2y
z*crX^U54Kw{3v1HLxxiqDFP?tp`)X=5fIEs66vgrbKpQ3_rQVE*`8%VA5F;XkYrPT
zHAMPX$5hGgek5xH^9Ft3z`S{7oRK5TsK7TRzBhe;$=iMqMw9XOzRE>yv#7s%PJ5Pd
zKALcgN|{y;)t3BR`7fwp`h@BOg}iIw$?2(+Rnel>BZ4%|^wek$pL4}#%TSAXce-bw
z-~M!_G8bv+9z~}30?&Nrb3YQ)6$!T=cEW<l5eyqPY|w0ebC%}?PpRk8cpPwB4`X)2
z?k4_}#c=*8T@EMIlH6R^(XU$B*<!Scdi}Lo<TvPxni}%UOl{0VM~^-<YTLF^zQ2B>
zrNGgNkv^^!lga!<lGjW%rW%<^X41>_vPkmA(TBE;^8GU77KL%({pyQ<{p*Wg{mQVG
z-rt@!$?MrpsEu#s6RJVE4PH9Hlx)7zT--uX^9-Gij+UzyKu^C=DdReIb7rKawYTp>
z@|~6g?d@>8l*R<p<#0sY(nF`>sppna^cB0+>f}e$bWjAcw6fUqTfiHy5Jm)3F1w?b
zFOP$_Q=FXNi#*X3GG8j3eJ5cA_^HoHDIt+7siP3Gj~d4FWO2|s#L9R*IIw{Vi4ZB>
zrqi$qo;JCZA_C0e=&7mcrbfZNXB5?4<E`y+)HZ?7kBC=Wto?lcB%D)|o`OVoX7i86
zMt9I9v$+lirvoR=IE4tB^TTE?s9mJlFhMMpDR5l&$F&)wM<gjDU3O$iw>!=|a_{}^
zlKPhKx?JL65=h|%&cpB3eexYGo0IRYaRF*nv^VO~A!(U-64a*|m!3ICmorPwHzR!6
z!5;;~QBL#vk;r<%YqgR?eJ@y!9fUj3;ZC6lu3LlVZ-GXwU7aN*78Y+x05eqgaA|2q
z#-f7NTB1MT<ltatWC!46sY!Xe_{D<RnfdwS$7YV898SK~;u1Ipq$JaZKu}W)c*hkA
z!UgsAj)Zl0QIW$6joj&8M9P{>a7XgrlanOlo>-PM>z@2V_<xq<;XmE5lU<wt`ybHu
z)Yjs^rgJ2}PB<UdobvLWbtHjz)>!`eKGTY)o_a(_Lcm&%P*U%BOhMa+)zqw9zA8Vm
zw_SAgTw+Okol@IYr_4g<$dQ(RAs_CG(=6GZM>rb}mhq~(1wHL9&3XodiT9{z!-l4_
zP)&wWj{N+Z8mqOYMhivC-?;fL+jBh2Jad`Plia^oAyGsE9_+V;gut1PvZz!Rr(TbL
z`g@f2wn|_pJEHdkD&J`Ad>0NITN>=y)cz1-pISq_-?h}MRDFoiC=`-ka++*4^>=RZ
z7ML%L>*6oj&K`nyzd+Es@6)Op$0g(D&d$v>D7Mx@Z|gk?Z$6+aM3eLM)~j)9GJt9|
z{wpojs%vVtldMrKE%>kbtRY4w3v$%%KqK}u)}qoUAW@eIRz3Il<Ezd1ruSlmZ)Uty
zz_<PhMdxrL3)bWo=N4BtRX4p-yNsqcMIw&^-9oj{UF5Lc;$oJ@KZo`}4VxKD@2#z+
zLP8vUJ8;_8D^<pdN<=jtkCIRM?Na6FNg4BfiJ&50T%1+>8yxfW*=P9OWtc1%AuG5&
zQnYj_d~u6Y>GR)yTKC2Grd~HmEfqo4ki?oF9&v_`&B>jS948*QD!yj1czpqHX!H2`
zc{TK$uRP(*%=%NHySZlcmpA!`#Rkgjp$Tgqyf$6```+Q%p5ODQBvv((`n)wlw!b0&
zQ>9wUNc}5l7HL(8?@6X**i~<2ferFO<4o6wJa~)%8&!2<CzPJM|Ne8wj=41X_urpi
zuoy9Fiwnr5#&ygfojMg~2g&Ed%M_z`1*&~SUkWi)pZGFR&&13?@5j+-W+QV!0Y1+N
zw|}kCe7*0`p+oz~#dfXd8Bc*h5&-HN6?`4g9mmpNV@h-#`kG)r4Tk@yW2f-x8n5>_
z*l+B}uCN{<S!~afQ9{n&)p=o-@3wA*QLSY!=%`_>aCsINE}oZNrKRZf{l)AWR^sVM
zl143S*8KdZ@EV>yNTO1pJSX1A_4Q@k$^$-3kH^#Dix)3W2nzb*FPo6j!?2d0S-@Dv
zVkc&?5E(C9Jk$T_X$}71q1O%N`}e=K?sew*kRk=Ea4idO+5ZizVa-VQz6+fu6;9Lf
z<h`n@H(vc5=Lv4zy0ZmmuKU^A{KqELXXo0!c=}pSsg)gr>)%#QUwp&*<^P`lwf{^I
zjN|Kr2{MhUV&*RiNz%go_jqa!Qgl;>V5J}Sg%czV#K1-9OG0{W=R>!~z#WB!I}9A5
zx{#`)-lS|CAt!^APd-#1^EpXbDH~;{u2OF1vymd7cc>1|^5tGWArr~UVwjeJktE?r
zF!+%VH-=Q3%&Xf`S6f+Gi$A>N`>Rn`K2ui*pltL)wWAi)$JCs`5w4|{GksQHhfHAl
z!HyxkP0Wr!*fE5^X`9;jNd+*U1~n=vnJ;JHWRo!<uiRlJCXlxSme8$M<51mSU$%TV
zyJ{Fp;<s`L4=g2h-g?{(Qc~X?SN?Sf$;rHfbvrPM+p!Y1V<30b9poi@A)vbzq{Cq}
z<tDj#v8d+zVr(E^2Ao^RVhm5oVi(h5KMeE7Kb>CJRMQl!Gu9e)h0GZGBtv}-$C$-@
zLsns!6;E0zabr!Rczt+|_lmOS=-FKqxIxA1^))qKz3(Y#dQbUZ8{!K(Itt`+S6>_a
z;M@9K@-=G;vqr<=Y;;!P8ZRxmu|JTprmt&a69HXd$&$9<82OqS@{1!y-EHn-pB~Wr
zK4TG+F>-c?fu~_)>{|bOrqorb796zA(=^eAoxMK47zOEJZ+?DnU3r~O2hS8OCy2Vl
zSpk2CgUFKlHnpmZTIGeqJgdwv&Br0iOO_z~V>X`sJD%FlDO{0FUGjhSy3|GX>X00L
z4#pZ<W<{mEQjt!v=MXbmLz{~Z5s1Op7Uxe*5{0<0I;^K`w#vHJ-rL_Xq@|}%ng|YZ
zLb~Amm*DH({{Rf~2Q^)e#PqDJ%;~U>&d^F+--5z6?!xB2tD$#5mF4XPgL6Jpn4hH#
z2EEq9)_#MOv4Q2~rTZJnOeS|?cAgi)b(*JdtKrsl6tI+#nSr9B2C4u_f5@S#<u`JO
zKaWwK>flBIiNgm>=JSqzu=yTxhRBr=xpcTN#M7U-5NN$1T^W3~*VLcT);lOsB!roX
zC(Fl%sX`%dkfKQf#`dir9>pIQhjNqMZ(`{^$-?^j!_+vmI&GLR^oM$q{@&4_jF>d~
zP=mL{^C@bCoDdfnun3Xyk}#UJ(U3BJA{@4}XKfB%G)Zv4>(U~J&{EFpI3)p0W_U-=
z>~rcln#R2$od@NYU9<qQfjESAUA|bFXx;O!HBsW|f9HW@n`rW{wkeslL&k%<U^S<V
zv*CRl#Jp+KHgW_ZJ67yI`VVgjcoh{%+D85{q!m|@lHaUciJrP?p%AIKzOY2Dm??H|
zcDrmg7ygJJS;5r8Lh@PavQVeYe~%rqH)F#Et9E_E9kNf~*nR!?EC!QV7YPnGgwD;a
zEGemXhX^)0&Ey}`3hZ1Izo1}^TD_*Aps5K57&6gOh{+VVSFI|em+~sTx(X75Lg(dW
z7LeTd&vG@rRR*qj%NBTnWI-E8iCLP0`J%+&mh-gb)s2l`TyiH&P%i}2+1lzoFM+q}
zzEo#Jb+sW%kUlkua=Cvn+raDF?Z@855u5@iGL5;0=^*4}sC{GpD+064Id&d|!<;zV
zH`^h%yt%ci;tibMxGeJZN;(zddZO#(2MBY0=SM^GiWeT4mZ(q)FbAAir8*N*LCKPQ
zfxE5EBbW#qaLFX6<bwD7Y@&&v2`6FLz>J3!WpFRk6XlifR)ZN?jEtO9A79M48L<&U
zgt7`2F9t!Zez#JdG9!Pn!4P`G0f~X)KMS|JUJE$0Jrfa%{XMl|O6oL-4%*Zth?q(3
zM={lX?R&uF?`d!S5{kSpTiXwkBYnO<n5r5H*M^a*)CnXR-joT3i4uv$($w^gT@0!s
zvma=(*(jwBz06lqHknMVQN|`AsaX;RA{k0tr`vw&{JP-tr|eF<&tvTqv}C4QWplV<
zr@~cvTAb@>BLtkrqb@#3J>Nx=nwf>E`CyLk8z;Q@EgJOgMImHJX_C;5&`-BhGBpQI
z(iySNBcFn%eR{+hiWftXgb^<qX4={%=@2JkqzO~8wx=daB_4}fJwh6VccY}S>O0-*
z={ZMB_z^0|^(i>7^g<6}qi}K=ghNDNdspG@M%e#*$g$BPhE4HXs~nD~a0kQt=Oe={
z_<q9nbCy&2K2FQaXIa5x5f1A*f5p-QfA}{jYkCd0qs~iN{h#>XT;JGZ17$GA26h`B
zX07;6k_%$Rh7Th{!O*UNF&S34Ls9j^gG0xKUQV4JbszOdYThp&9sj;hCw!x?<^aBt
ztv8TwWNOVPwNiKBN!j_hF-k7p!Vl@}-EV$((`qET%h{gQgQuwx)Xwl%Tk7@twl<v1
zTmPydB>qwByX5S|vfwpq+S{{;f&<=gd3a;3!q2u%-R9aba@sWUfNA&c-KGKYwA)>K
zarQPcDnwaUX!C@#ee&n~Zk9U&r8@sn_P<75R8Wsb)6<jfd5G#U@CsGERk?H%+duy&
z{jVpIesctsIAA_x=|cqtwY58t6a9HkvBSX46jaV0{NBUy$UWoBW_!kZl%Al$qa17%
zY^zi8E=Sog8@Y&Iz!9F0gx2HGk($PaZ`uuE6{7O;&6_uhMCF@i4<7Zzc~l-1+mdN`
zi?(Ii*3TYn9kh8SdGs8kF$V{U=sZz_M+aNsoJAK<wE`^;0(~s_-;J=06wIQ^z{Y(|
zIlMKG0}Q^h{hNly8XR*{KO5P|pGF%Vqrgz4&Zf5GQ%_yuXKI?7vNAF>nkMgi>7?$0
zf`^u(yFryW`m7jzrMK>=1uLr0;f&&>a|$?rpg!d0N^^6l4>`Z*+{>B38BOW^d)$+Y
zGij(l;k?*?h307+8uYr{scD9=)FzzRnUU4hq{*BOUbCLGg6(kz87I-29LRMl>H-d@
z+Rni-RX8ZSog?C8aMy^822O!v7W!9Dl?4~T64dx_a?<@Yj-Zy6q?!Kqq)pA$=^Glf
zd9$g1q4qza6}bkgVm29DPu$>9{-$z~XyfM1<>ex8dmls1By*p%c9tg!eo}wnl$R1m
z;a2JbC$EnLNq6DI?h8~ewVi&txblrzm<EL>iqoh>%QxXXcuWC}DPWruWZOh7;d|Rj
zqn*Ba;ot~KW)%@h*NaWT1e*c}I+zUf4fM(p*XRb1d&1E8DbSMEQv0YYL2BbBqgsdY
z+Y61}zknTMJrRRHacX=Vlt1o58*(H6#<+jiZTslNN4&#hCR6(wbu5OCAf8baag8Eu
z8yC!_V5zu%W4%}vGh8Bxanae@_`i}<@6&l-e9`pTXP*qP@A=o_f9<3H*`BFo++R`e
z-}z}^n`AhV%E?bKhZBT9gcPDR!Q6KBKJYcsUw^Y<0~I&t_m9nqqx!EnuJls^PT2Nr
z&w`k3Vca;*h7DWe<~;U$eD4kkgKmu(Wx~0x4LV;)y%%H1_LKy@;Tr4mW$0Cqdvr?+
zm01>&>Erf=DJeBI<Tp+z+#wXkd9j0dJFXQ^`teGV@jznY0oeP9&F{pCi4#v?@%L|p
zCnaZN|9{^^T`|k?aU7nIIGfQN8s5d;AMeFlE~n?8Fz|B!*;4{v%ixkZpVd)yIi9(L
zuME}=UhxcHJI536;SYW{`1ar{p1EashFikC5sEilyz8q!=-*`V&t@0D$8qy$la<8{
zgk2T@`%a{U=@(fc1VZkO)=_6T;ZTFDqIPlfsr&`hF3u{>qnvTbWPH-U*?yq?mWZI1
zy1tbCmk3DLP2)I=j6dhK*UExldky-Sx;$vexq}0P*1=;qCK9Ka4IabwfQRduYmmTP
zdvw&Oswy%##?Rh=M@4l%sL#prQAFak6W-OOR+HcJGtL|4duDp3dPaL<J^%Fl0l%@h
zp6Z#&HHg;Cr*?w^bWvui1T1|C{&u0fn;}H3ZL6wk)X;hg99cQ~2fHcNJMVvH^wA~%
zwUI-5L-g{muD$b=gxr7sjvd$M_%;;YA2PEyGmy*fa<rf}+_B?lMr<KzFC|+~n?C+H
zaeeR!itI@`>`8r|bWjA)guy;$HySo?-fzjzym^g{Z;>Blbn@;`573p3x?(N$p5NX?
zR#kARuC%VPu8b;$n9UsY$>9ldNE!`_&zv;QzZo0+drYqVy`k27t$sInpN!UwRvNz<
z3|5ho?$>g<Rv^ESXM^X(P;5l4NYs<jYM^k!I#T3^*VJ@%{XfiocR*A}*Y}-!w}2E~
zs$fC74FnMoL<Aca>;)@g!-fcU6ct?Ty~c)JqehKJW1=ydsENiz#b_+C#2QO5MuVcj
zz5D%UW_MYT<awX({r>o3(A~q_GiT16Idj^~1&sb5!TYX{XB?LRvP4J+o^-+djH4NZ
zUz;D|tIyI>`7l<^d{O8)_FtBU#j?8#S{{=t;8B$_ZL}1Ju-Dw?RBJXK7aBD);St3w
zD`Ttff;Go=D()Y!0aq!uC651qjkhvm15SsET3S=uD0SwRr{at)5!W-lSKT<;O6IoR
z1sfBUOB{@Jx`NxH{qg`@#3HWT+h0$s>tKV@GGx{>opDnQ?uYiCNfzd#mA&=50^$hw
z5N{{9b8@uK4le#PzwnRWP&a=|<JbvtmS8*i_lw^{08XF1_4;v<^`j>b3YsC!J-n5_
z_U2ay4}SZQMj0Bm>k(7EdL=kn4=!E7neG?(^!{DJ17vWsy$fp5cac~B>xnhq6O4O<
zF5R`QQvKk;*9CGtr0Jr3{7-&y(721xyj7OAZkxFLg6bgyjQ)T1lr~|us0xYntYT}8
z`tEDj&!7MGSENrL7q5)&-vSG7DgC3K9$)?So|XEB$WQ6ce}%N=#YZn`RsY$ukM7Xd
z13&%C+EvuR`4n6|YN0yp>~kj%f7GL;v^U;c13%rS5|tG0YY;K^piFDD=fizHiF>b~
z?R|FY@{MQj@iq#^=f$A5nMTt%&cnU3Ph&XHy*qUZ1<&BJV!#nqi(Xtf_2~JF+ZPbn
z+W+}}nlSm|=$DAmow{f3N#DKTP;l$Gbo}^_FZ3wGbay>|43WLdF9qjXJ`Dq_ySO+x
z*l2`&%*E}yaczOIj&<R~+qW-VzWeyWqbGm8ep{*fnG5-{&ZBNbLl;T<>;6?E7c(dq
z>*(qr6<tRsJiW|L&e_{I*gkvb?iGztOEWLs`+QrCWM53r04v!lD@>HtOm!i=@ApI>
z>EThO@X=-Fm|uQWSjD5T@Xz}vKZBos?$H~1vR|S%eO<Wy=mG9#&wuv#(E}<lzdd;L
z_*p)ez%)=0D!JnAvJru<(hK?$=L@%9OA@$GFWn<0kdG3A*I?RL_pyJZaY7Ng;N&DD
zJO2c!M-QCv`1Ldj4Nm`>UtHPUBR~J;U-{4f0C9gj&;RRXzAP8W6a!)th4Q0Ez@jaB
zWE?x)Io#{|oV|Vfalz}Nyw@)kjmCBV$C9*g;ql{zkM1J^Ran?uDS=*2Qs?Bv$LHiU
zHzXNCmG?G<g?JGKew>`fn=Hf_V5<=k-3@cwEjl8&2D+*d9MQA~gr-N+h#*{E5~Qzz
z7o<o<ukJ%I@BLLsbZ%$yzUxArAC5=XEV>4*$z-J#trXW1H7yjUb`G_7?CaMrxN_yU
zf~am$&rk0={oLM8xGLDys(<=9hwvv^ys3YGT$RM*O{IFf|F-_yq}SGk_N)hLFZ&f8
zf!yytA<>Yi_uqF5?04?O?ec>B-lHpS&PweV9ftRWggWbSYg!u^^RBi8F&V2zm(E-%
zato>6Dc1GXy(?h6e~hrgI%zC_7mjQ?ZDOgf6XFwgAKtC@Un#8g*RcM}=~-*EU~LZX
zwvO7Ja!t6FvO9{m%mk&4Vq>lC?Okd&M#=cmqo^mmdkzaDC<uY4$VOmYKyXOiV1#gj
z>xOhgFtA%lWdx8NX{?!KXSX!;g`uxjUMqjnD-c^T1#pp6CLY9r*kYd?obC4J#p@T3
z5gU8*^7UWugpWS@`LzR<ro-!>KQhjUrIv&>%+JBW==jXgF-z({vj5*~RMcMZe)WsB
zZ~Cy_@o}vY7>tXJ?$*CQUUsZGhcmL<E*JR^OHUu<sn>n_l$;ClhBh?4J^5vJWox&T
zl=}3Q$+@|CHzPWv82hF*Hr1QQpf>4c-iZ?@E<7pJIM>5s3G%cNUfuZN?8S>SNyCQv
z>Q+@lnm2DA<Y0Jt<(sb~!u{=qf_s;+n6Ca&WNgA^#BwZ_`#r5H%u+e-5fc7{6>hVY
zTH#x5KHMcNpQkp5<a^_{z!>4AFcwZhD`BOcvKemsKtw_IkZSny?>H>+mH%wof6Bv6
zJJ$acZi%O&&w7lb*^gmiZdV>xSA{xM`~8+%SlEw0J`1b%xZq;JqiRi_{YZ$R`2G9$
z#Re=&l~tumm8?`md~*+(_<L_edrzF1AfER6nU(7K)c^T=AuQnP^<VXs@7%i@5H6h0
zt6#I8@a|oaJX0g8achHGoarDJy{lip7D53GMS);jt9~h;Of=Fc$xIPNP<k!uo(^=z
z{hzVI&xSy~2)hthU1m!F=V#-SDhgiLfmnF*hmCRrIYrK%DiBc|B5atcw#^il+VUtf
z-|0;|DQpC5EBK4tps@N;2#q~i^FhrA>d@NZhr=Jm7>1Ly0w&>yK~Eq`XbPWP$3A!c
ziCY~kyE<;xSZ@W^#xKTS7200Bc=g_sKZ|Cni~fA#=AglKO%86g>NUiF^=i@YYJ8`^
z*WmZ}u3o&Tulw)#Ec(S$cE+dvQ!gTVu~@|>&X$*)ir`eb?cP?Z^p^ezd}WZ#Oudwo
zb98QVgNO3M;33TAvsN8#t;@EXDO_rf7iKjwBny}Rfdb23|LODpMAmbRc6#|yS;8$p
zSJc;Bj6h@M8384WYy0$8zg_zBvqlYr!>h#@rq;S}_L||ZhYub+IQ7NRvp64j?GM|i
zapT4fZ(HZdmMvQzI(KVMvthz&RX4xK8)xgaUS95%E2(7xxtrkO&HUb}-2569oyfyg
zSH#A_O0?BE*lILRUIz~Ru5t|MkR0!+`SrlouL^3loib%gtmziD@fH-(=ux%*o{)wn
zLjL>SKS%G)-d1+C`U}^C{d(j6wd=4)e|Y)ct!mY(?%HS9&YVOn_V_PCt)?yU(uvsG
zg59HgHEMVWcG9z-fByNIWGB3qU7Z610~>_}INYt&uxZn#A+~@0{O!@B=koMcf8^i%
z;p&?=uL@_@5Q+RJWZOzB7gr4n4-aopReX00*_oX;MNO6V?B-)3tM;_<w$50f@_SjM
zbFfkrMQvM2tk<l*)$<D%9+k+}IL{XBU+3bv3tS#8_ul}IfV}&8YPUziBR9SWm@@iN
zXUI!8wz5ktg{ZSpXQcxFyWCQE0xnc&Wji_jM=RT&TiL?@&PIp1`(NAWWghYUZ|r)y
zf-F~x<!8%PV_KCo7HuC9aq=Ylv9_k45J3?l^ma&I$?V7HE<QC!Tvjr4<g-I7q$-7l
z9X}pn`Goa9w5GXHO_`Qz`9qv^{Wi0<wz>I5Uf%szKvlOz3kXH4*7a#c>-9U@fB@~Q
zmzFiHPmaU>GgY<XtkaU4bA&v7GpFu7Mx-b0U%zJKx`+S=LDYGuuN^%6?Ah)kC%@R8
zFm_rRuW@bIxvNud|AeldH}CAp-O8$5aV~03OGz-*SCqH%XkAI%aavzuO>0Z=2Q?;4
z`cTbbZ&&Jf(~ZauEqQzPgx!8tk5cBTx)`y8@r+hnT=(vAc;V3Ay{Nx%FdooKh>JUQ
zYXAOIr{dxgj2gI9&db)5Sgw~3;|r|X<nA*sC&@l8;;s`eJJQ>>>3H>$LA!X?=MCyy
z{nUuv4-~0zeo}Wc#?V`R)AUKWskaOBzk3E*5#v9`_O$-Bg7#&ebSnF#2?68VAUt0#
z!hig@Il}LP2!HqPb?bg`j-ukIzYAiB^f<@|hbp+vVA;R`RLkH@+y^J(`lEPe_g$}w
zPsFk61LKqbaHP)8h~Ph-tg~E?6@8`_lkV)<NoUXI<wXPpL`2MqVEzzHXavC|VYTMj
zT`v!xyZQ7bs*+)}WL;QjZJdq&Y6bmHtHR5ypfFp?oB!Pk8gY>FQ7}~@t;T5n7i*eR
zgxfdS+WucHZBCfhi!X803TG?OIJ0F}pm9DcSb?n0sH`-UdDN|}C{%#KQb$(>hq9~!
z0))Ig+*3bEZyoD%_v~G-Nm8z4oK?%#|MuCW(Jk8`+9p&&g_!@Nyu9fBxS*$!u}bWu
zWWJ}THkCNzNiNH#I^%(^51+8?;ToR)AhOo%Un6VJ1<%Goku72(>NpuMGccBD8H=}G
zzAPjLlI>t$sicSo-7I2o|0({~tbF|KVgB19r^X*ve8D}Y|Aj?$O6nIER^O{ebsCvF
zc+komsWWnaA;wofpq7nL`1_r^e+mv>-tN{y(ewMyZG(dwghqx|6JFlE^ZOgY+TGjB
zK_K}*oos~!<8}PhK&WMkbm{f$OR<J2h9seeJ}4{(#;N6ht=q-NMSuC-#h<=I*}dg|
zN3U7`TdX_@R0UYnxD|;SD}FCT8CJr@gPUf&93sGy3#LmVa2FV>faC12_}h3_1c^gW
ztqe575^e}WT62JU)~qc3oS8F_XN@;BG1NnPq@kfXKKFpHFvgIILqzdtS=~?z@YeX;
z4LO+<)PGa>Keg|G6iZbUpJ$+5FSKi62m#)DXdjQ~Qh_%a_~X$}u%R}719)c;kcr%N
zGRn%!J}|)afVkk-1#Jzrl)=h0<jmgS&(HWY*icK~EH;+0{^XB8U;N&Z@A!P?!gWOW
zu3e~LN=oOimgOvd@VVltGg%6Ptw;CnXJ|QLpEGB0@qmifeCB53eSSOr{ZV4b+Nr^J
zf;_}5_Ko|yWzRuY{0-F%sI}&vun5mtoD(A)%f_qwaK4TopDMI2wB~)_RvwL-wv|LZ
z>Ay`)TbO=lxyFIB^u&NPL>M9%ak9g9rTEIU%MyodFt<AzcQmfFowX?fm5pAdxT6*I
zsbk(h8%LIaPyri*+7gQ;h5yr$u$~2sm-S4saZZ$l7fiAqVdjvqKx;ba$8BG;1(qe&
z73xa#ydj$-Gh9OaPbh`c3}ZG~w21xC_p;8<W`B$K8r<t(m;i~ci}YwE!!k(uGW=T&
z--8VkNG&q3RdeA0sfC|b1NHtUDolsysE}DEG=KfIpsU}0TR%=E&G62VJkvanw7?My
zyh7B_Ui|sM&n*n?@o|)4tZ)RWQRA_^@|ZY=1WI?TazVL}-G%;*1w96f>2oDr8L9MD
z+9{WnPw~?ZpGN}rIjc?KFI28!qP09``rVY#rB)R<fvPe7H+!DG*fWO&k@OzTg>Og@
zetM+qoh=C!3d{rLb3~&Y+2n^GvaNskVZi2`mjhnrY}QvcJz{Rm<|u}>y>ca6jf)MD
zwrq!NaRm77X4s65H{;*B_>R1~-h9fl{1A-mvnwzON3-!isa~Cy+2o-F#ifeNhCep6
zH3Tc;mBq?KgT|my9^%t@L$F>)3vy7Q632N!|E6Uk6=gz0fm+YlQBhh|OR-e^I2&UJ
zJRPHyWaEe7=tx-#x$14Gq->HO$R$b_kT+iGBA3Vyludd&Rw=|L=3lS|0xJT>JB6KD
z3~Nh@Rk%*@##b>VIYnQ&EEaRc5v>Q!(%ihsvcz0EtV^46N&64uH8uZ`U2MFTdf@+u
z&t4d@E%mjSpXKFs<%-_AZKrE_Uj)^?b`YgOcx|)r(kMli^v<Wx1q8flKIgN$f2(|2
z_L$Z}SkL~<3mnpR_-NFZ#(CQ#V`1aqV8}m$+Ub=CFMP5e5%NE()r}uEA|V!6(tQ6t
ze!kN;^SQf!qV(#@SG#e-qV(w6zaDf$O|YMz`EVOwTdP2m+yCG|T_1X&j-A1Kb!=8F
zHvEB(EA#T+Kc3`v>d;TGROX|0A^d#Z#fs0@1-7E+>o}3B>H&4NcfG>bO3mrp>eu5P
zS2L~B&E&C}-wo;zh2Y2dEYJ6u$Z?ZQ<kSZ_zyD+;=Ys*Y9z9`rm+*jqw=r`*duCHB
zNbh6nX!E9nIQUlqVL^*62=g?a8?+o(^%^xjynOB+8Y4YDb71FIF6lX-IQ-vEX8+so
z^#AV;Z1dSIS5Z*Xy+pNFwHMnRHxw9;KgieHSf0x^8x{1Ke_f4;*df(|oR(4Y=_;`=
zazP%bc1Uc&wI7iXyQ9Zd>V=PfIdNO3+qw75&A(iYZJ8}}GTLYhnPPu=LF7I^--Aqn
zAe}uPg}dKw+uGi)U~O^x$#jz?Yj4}qHBT=M4!B~f^JdU||NDmleRjok8;*%?)3e~2
zpb3j}l3pIiMLONx?&sfk>u&6k)+%V6;=(Q*e<`6*VUE<v$*B|m6*Bqk`{4JLHNT!d
zdq!8;N<&`1jEZa?sIP2oUAybxlwqiFAC@w>Yi*M^U%qELhSGEF?6<OBFwte^N&qS1
zCnzj+3(c#>C|mV{v1*q+-JX8sl2AGu^}Kv;U!w=#F&p;wf8?EBJr5V^?))n6iM_F{
zsb^DtxvA%qam2P0lK~UtnUe|92>rX>{|;BHeRrQA<4L@*CEV0_HKPf7z3Zb#c6D1K
z&)l-E9bWKJ`q|hc%}v-yz9^vZmF^%je@AcAzkjg(pGe{SX&>C**ow`c$Jy6-{kkY+
zn&M>@&#DtJYR|}BZBIRlu*C~9BAgyc5w1^+nAjXYIeI>%rIW1JT(NK2(*BA`s@U(G
zT85lI`UJmgX?^92b;~H@b9{bZ<efRAv%~zpeH|SgT0>tM#T$-cQ86fjh=~ewe4J-&
z9R(Vy$?lP`3L@R*2WL<)apr+s2{i|m<X0#Tc%`p~%zKDel|SID{ZYj$q(_RqeN8WK
zWR1POYMN$)Mzz$&_eR>4eQ%`R`P8Wsn8~kC-c?W@K`(lZj;gLbW$fQHOILzO9>4Re
zA#k~zKmX(L^EkVJCmQHp-p{-61PAGFoYzYQ4{qPiuHx!o^D_S$IDh$yP6r7Qjo4Up
z?a|8;lnH36+S+OhU;bTC^bF@DZWQXB!*DUWqb)6nu(3WvPi$P3f?7oypJX$1Gb8E?
zz2s$|v(5;$lwKxr7_e!~<-aowg>nAZ^iq1Kf+9=2*0IuUy_6vpbTUHAKg;Gbq@_;C
zKYG6WdWH`osC9I%^K&#QA%%q@DM=P9S-SqsnNzsyLE<YpKwmjO-zGG!6Fm{#u2Wp7
zP5wXhS)PXS-z~?ExUL`_j&nT@g)(0Zv5hn0WblJ68GN-EUEL(_OiM_n`z_eIw#D(l
zJV;GlFOTX$zRr64n>P(^g@x6TR;XG1?zNki{Mz5$ZLeNRj(%2#izi#U9oI`Cw#QAf
z+Zt5aT#NR&b~q2UDS3xiH%A>g{)nqPjcP8epno)@It3#u6s)f{qK`*i%NGk<em3?^
zsm`HA*-U@%Z?VR2Yz!K2ER(RhKDY1E*9tw?+b1LhyQyA2$<M!X&(1B>G<p`#%NXJI
z?Ce9x;k%cwc5hpQSEujMJEXhm+72Ds+~57hy`rK5*}*@0s1d*UnY<UM3=#_d+_`J>
zvSrH-o1SUr?;yV^D!Tt?HUAc6o_Uthw&=rt+Unf?r62mtmA{Y>mkM9){$v|o7xxAF
z!83ygH}$Pj{P(+ezva96#tt^p))##7K6?~5JA8TV6`rzp(p<ZA>H6&(w{Ne>J$~ia
zJJ+vVzmn(d{6_C=YwZ$7U#Z(M%*)x!7FU#{rKOFZ7!SMYbKwR1^GbMQ{PVYPMK~d3
za9L)3=&qdo5$9W<NHnnFBfovC;5}^8$ukd&h)O9^iVE>MB)zk>wTH7u2tBaYF2uq5
zzvzSB1W+xcGivENhq$A*$vy7{M)M+%TCqpG+FGC2|Jz%AMZDFQ)5`Gt+q--J9S`2m
z%khi?J^$$~yydm&|J|E*MZ9U3_i4GRna1+*Kff8PD(|?-CbDc{`MK<xZUfrA;^U7W
z|Mk~zKgDi{ZLv*wZ9hL8rl-e<_wM{s(j1h!%6h#nK0d}>5mYMS*-u{|Iehrgw-5Ac
z(JgZ4;UHKgB*;{0QH55G(vy%{k*Gyty-?0fy}zV-_4NB$Bt4y1T+ma7JN&UgBpLH}
z_Uy$g0gYh1MKlVya`Eh00kv`YICO?dwJxx=Ezs}ud4w9VYmWpkKU9_k_<7y^m4+OD
zy=!iUM_#|?(Jd?BR`vAgDO|~!$2EZE=q*3{pk-z0`bk_~qJrKndKnefyq>=D&Yjoo
zoZP&zb$Yuw*<Clq(9C?J_=V1fHL4`__sF4Wh1vov5=T#+r?1XaN1Qa?UVfpWVPT<-
z{Jgz1TK2V3SXf9y7Z+zIjaF(zKjIUbbyig={!7|ka-Qs}<>2U%(yl>!j|T3QZCu=&
z_Vn*L(7*T4$+c_N#EUd))>eQ2{^#N9#G<XD%EjH?WtXN^rzV|jY&6vzt9e(69T7ER
zd}ZY=ZR%04o*)Xa2UO$LiJ%UVB;qM~lMv#Kf2B$Ck&MaY>~3-{M;=bzyeDInWfU!W
zD5(qv{+#BW=qC`2Q=_%2B$%NO%}`s}4AoXLL(!AD?J+Y{tujNk#0=GzLN(jX(8?UD
zw4k&bp<2ih_D&OKZF6&LI}037%ustwlxTR0@t?x@D+}0@+4w!Ro{pZ*><8b)T%|?Y
zT(VufuT!kk$ZTa<IvJ|eQ@ly)W-?k?p-!Bj*i4Y$P9RSty!DvlV|+n$5ck4Z2@CMA
zgez|Zl@I!{5uAiN=&6yPpEo3-VYDb=`n`1_4Z|Fr>_rBvDPq<&+M1%YzSoj&N+FR#
z(gsF_Z=9KLn6dbqUds;lQFfEip1bCCYd0ab!?Hw@F*YMmzVF<4n0PyDlu~1&qDo!e
z$6uXTtK;mUeKOnH+qD7_L|AIbR}E693&@O<hAysJkl>|3Uk)zlJ=|3TN%Ijy8#(xh
zKk4^(?$}~Qr^%P5PyR7;(%P^#k*AkzQ3^L~B5HMdY)tQnfK<is+f79o+cKbE<8IqX
zlgCd<<Yvl!)(hhRwH7D`ma<tX54XXqVp)47m{Ns#q3t!=Duc{en9Oi2l}xSa7&P~d
zt+iSWLYtUJ9Mck1Ex}(~0lp(?6eu#DMvLK2_7Y<%j-g9Od(6#k`}H^RxwE=1Xg6)+
zC!LjR#CgKK&qs+zzImADQ=A<%{qg7f|IY2{rA`c8ro6$t;oZfm9>6OJRRrX-iL+;@
zx?=2NuDo1Y|BpQsj~&~p;!eObMm^&h^$|2$n>&;T#=b?3;0!3LSwTx>CPPSHLqz$H
zmMj*n<ac6ri9vj>Q{^kx6Xcg;)QMva2L1%YOEZ30rJ=xUgwj>Yp|-V4p>#!as7*aH
zl+6Q&Rw{*R=-TB_bpvx-Ha8rqDTT5X&7s;-C|l7STA4$!qA@}vUaV-Gc%dWr$H*ID
z<cOH_k$dYrwPc4_U%onBvcsU|;xY1Ej2Tis0utf_DMy4>?d3=1ndww6?)GMJ*LnJT
z>hNmBl(kQ1CT==AeCDn$J(?}=9lv5?T%Utk9hbCFesLh*_Brpgo<#on8L>W)(579l
zsK!lZKiG9aDbDouC8s9I8G$V^Ppn4`<tL2_YqYiz;0A{>&T(j^C<c{pQtC`?HLqjL
zZ5iiyTTKiGN~J=2%jX+&t%H)l-yu4v=2J6fh|x2|t#YC|QQj+dFWO4i02G>PFKdrC
zX%e(IJ)?%yO2t|7v8*hyH3=^6#rB7d@*q6@so3S&#ED!6rMHY;nw!;Vp`}5?Z=m6_
zQRcYBg@Tq)>9TT;G?wj2W96JWaddI!=+Ua2(P-~xuvJw<dsMNgqs354Fgg#>dE+tY
zH!WFo`F?`@^nm=wjT<U4WNyj1MO79*Un|8I=T4iZ8a%zkXY|$2CfI`>)-#jcdW;^v
z!*D3&HHY#YhCvza9LjeX-j+%!hw>eUL#YPgPz~E1I0m}!a46q*cw4&fa46q*ep6ad
zo*1E!QY1F$Eb`WLhvBX14nu?;hCifB;0P*(31tDcbN_G`s&c`dQ0bCy(*Ak<PIYcJ
zV148Do0QS&#F7E=+ZHA?YudYs-}gHu?=)bhI9{>wh4Hx=MCbbj<{Lr}-Gp`F3Gr~m
z=8fGuWyXw3Vy9gBD{<7vnBgojOP(WFqZGb?^x+9m*qGr_MS>`T`BwP?4<cSvzOb@t
zXNl9Qs1u88X40pY(BHF}Ky#|m4Tz0EkMQS)0UL$GVsR~@Q1k(8W~eHzw5<uM)fAHc
zauUWc1h}i3#^C5rTnVFKOH=Wq82KJ0gFG5Q)+d%^adL>z3Qvr}6Ob_C=q&|G{xDjc
zG}L19b_{WJRMqIH4^e4RZEjejyd)|qAgkjq1<#nsziG8$*+_>HSLvBU+N!lpWGisH
zh4i(>T7wnfNi^7j>2f1setw~J#Rbdnh)S!T{)OlwT3Jt5U9U29>6Q+1b<pJLu=@5?
z@g7znIK@^h<5Vo;h!v>_YE#Id@=dzB!6)7p!b0~kC5f&1%7PchS1Kx@A+QOg4=5Ag
zv>oLU?AJOhy5)=|VP&aQk^-wHR;?D_an;iKv&rTPi8j+rrIyoSUrgy31ufbNTC}o&
zTncd#sTIm}DQr>)Z52qdsH@@-($GQGD)rX(ZMRcXZ*SXvD|PP9lP7oXJ$m9}ao+{y
z+s_Y@sLK~g?3ag>Z+|9kBvd)4JW-x1c_f%}fc3$+Mf3e88<Iom3g=M1!WopU01o9V
zoVTSboJ08v=WXd)=1{(tIW*Y}wH{DP16|>~Emk<z(<p<ts+FcS8$GqwFlJ{99{9RS
zTB6nZhli7JEg0cMm$QSek?1G(+Hq?%7()NPO;WQvc__QB;<7t<kPdG1j&E<jA>;V=
zHtW^tlp&T3`o%+0!ixJuKdtq5bJeVY$;uDPvuij?5Jak8C*oL6DW!?-&z(Nt7M(p#
zKeuWNF?+vZ_Nu|>V74BNS{D~r*bJ~M+23^jurs=vrjZg)Hr0x)CGOJ61<~!TNGs)-
zs#8m`i5%t8wARnp27Wz1XZHN8c{vf2)ah0>6I#6adcql{%>-*3l1P%ClBjKylo!fr
z<&08*V<J(PiE1+Bh-k4|J7Tp~V(XP!z0B4P3x|b^kJzIcs-#o>bCd308Mbx6!nuF0
z{W6QR&L_^&*Ar$($O}Uk=yB1i{92v(1ElhAU~3O-b^@Z_#5ts)idxDH58-&>Op)|^
zFNw|EKUV!bolJWFcId`Fr`3snT)MOQyFSW0b)vFMPK#JJbN_1K8?5YvK1_!d<3pvK
z8Qrd07?1xX>~Z0l!~5s-t(9A6<i?5XuBWBn-M0Nsdd{aOPwqc_^uz&i->SFAnmBfx
zIjH%l74e^v$WrNdhxGrTJcj&VhV-NF#~8~9jK!Xfh4u~AZRCj7<P)fT6E_bbJ2xw1
zuVs<8g+y4iZQsM+wj7e4j95R|SzaX$lsBmpzxiVERQb0VR7x0DCM7*Bq@+C5hV566
zp{YRuWQVs^6@m%ENXSl2@Fak9q;$KYj;f!3AhdvmL8Z|+L;787imICFH#W?Bnb5uU
zv6;O#4hfq-E$iIa5ht?d{m`Rl`+dDT&F<QA%e)~+#*vBRJ{#JzTT+WIQJG0?hIaDt
z9k6Un(z-q!+a|V*3LX?6)4xOQ>Iu1YEQ}I1mOgwe4i!OdtQZtt1gOSXc-tyW$1)5|
zYj9{K-WKDwY|F>}iQ$cEf_kdphjGI@awg0Tr+Uw{T{rUc4<*cLqPs(*FtLOF=`dyX
z@6U&QHNEZlI<<Wgn$F+&_#w&axHoz6#>uzUQ#&;2s-#r?{=n1&1KgZ@**UIUx?#EP
zAhO-3;i@eq^`#r9f1w-)ZK^Ji1gsTiwS{_W#g#LZR#Kt8>S=LR)zgU+Aupy@5`0L3
zgnBxg+9H6GtNg{=i<c))7G3!>;2?X3ad05xpp8O#)d##nf8Z5|R<U7FaK#d;QYdZP
zAPsh)jax5RYhHXybm9h#$}a8unK`FNP2V$r;V0Nk`!2XNZbIH%X-7$el`B@PlrB@u
z3<skk*+M2cl<Ij7<$9h$*^F^0*YmtB)$<(6^*nD&^*o1iJ<p+3&vPi(^Be=!^Bl@8
zxoz+_Fgi0-{K8Bxiy!c|Jbu72P)*CB+`5aw+`-urG=kvELMI_&U$Sb3;+dw@T5wpk
zr)VpF2|iNajebN7j!51T{)eeTt(C3!PJO!X+{M`k4}Y`wTWX}JY7{?&UC~<QUi`c2
z7^R0j%lHR7)4WeIDB~1|a!&EKlv5naImO#jPH`wW1)4g7rh%Xd^5h7^k=c&WK@--J
zcBI`xWgPitt+G&=wSXLl6?sjpORY$8zC2O9L!WMrr&IB?6~16O(8oi>wKJ7^i^&4A
zR4=;7&qXJxog5-wz!+F-HU?}9bPSGkWDLs2z@Z2TnxM?P<WRMp8Or=f4%OOGDz<_O
z9jJf>?2An<lz7sW;V(YamFyN@$bBXM2@_IPQzwi8Ue=b;FpbmDkm4;5wV~dy5qcjS
z<83wdDU`)9v=X4;C@f(I;RmKkk5NyJ!MP3P5J_8ei-Wf2d$U6e3gtS9La{o*35WLP
z{ul<{pTj7+N94lpS5yIlhfo*eb7C<Yf3^srghN7%kwd(&lST>&f;c)vbfA!>?ezn_
zq7og$cMRX~+50~?Et(QKyQ|ND^=*%Stqkb7y3zixR;FBvX=2~kx^<hj9S*G6yeE0;
z;Hou}y{mt|N>1tCp48W;j2=VjViah+LG5fxt>*p9QmEDpl`Wvv%+TUesQRKA`nD9R
zu{J~Bl|t2DnxI<Tq)KT}0x1`aP#P8Z-0+8)ww|We8pQ%fu^DQ3iODe(K#zS3P6ohN
z!5oo>FzBdFWR86K{S#kt-?+rNO56JRA&ac_^z)NO9Q;_4f3Y8(7TvE=Y}AY{BbG$O
z%o?&_*S)L#lX`_EHur0`C~Cr_YBlB<2Ifxc*S391m|YbzaL}O2FvWYR0+bie8vC61
ztbXm}291-3rS|@G`L=z7(g#+nF}9}9&>jhc<^1W%sbhu@ZJa^wZ24~gmxwL0F|fHn
zM2*e`!a|~O3nOz?Ulnwa5`mcG$I#UNT5nxAq=N3-G{UTNaUls&>k}96?=;}cImbuZ
zreqTB^7TzK`VLwd5U@a+*Qv+cH`_M8T{J6QeOgzPbL;lPQ^^yW%lBGz!AkqikS`U3
zgYf-)P2oc91!ar@Y72<D-Iqkhw)&>~+F!<u`DNXJeQle^Er?t8b<d9y8`bF^89S?B
z`-XQ*qnh*%4!-^Cim&_98D_XO&X}_md}O^|o1u1QXbFcZ!ITaYR7*ORVkm7}T#7+W
zbvDOa*y}YowiYA#1srn_YJp>}I&VJp=CSy)HN-?ihl%c;nNp?M`mPHO^&I%c%%r7Z
z%KPQ(n`QPNvLGOEp8Be;sOhpk?dQGOoLji0MXcMMU+14rPHQ20#dSnat1&m#aG=jq
zfXPQ7Jk*qM4R_VLIOwS1N;gv@1Mxd8qQd@cs^}Vqw2{%DHk#7DX77)h&g+)FZ^wj7
zb2H9Oo^UFC<jzR<rNj1+JV_OPFr~x%p8IF7Ij$P;MK!-Fsm>wZE85PUqD)q9?S7p#
z_IB>d+fzEm&A!|Ifc)CKsYjp0j@xE_{nHc5v!S4F1Nh)4to?~<Go+MZ5lF%w!1aKu
zVHf*>ZbWcALw)}vpMKAUal4R=|2KB|PMY-TzjEvV`W7K&?%<g~YKo=n)5Iy395F6J
z5-Rmx-m_bFTx{Df5|&*`O*@;Db!4DOlrb~**@!(<8_BJl-JOGi`!|k;WN&`EAoF2v
zwR%;^-NVO@ehv<>-q?!D<#Ojr`NE-8yK$&&0j1iFLrW@z7MDVq&f{(0T0q~JF%*_U
znI`0I-<3j{CS*`8brK}u0PGdY9cLj(2**0q(s(8lRIO=O2>XvrlN>9J9PJknTdQ%M
zDjpGaQpb1uRQcpT@Ue2`;LfeQ>NV0>P1IPWwuv3rmE8Vk&O%#!kA1lrWT+};(VS~=
zcE%P>eZEjnD)v%D{Pd`R;IYx;CX*3%%Jxj9gs6sotiD>^eW)w0Q2JwJ$xf-y$cF>w
z#9=M1Q4&>gstlp2&=xppwxZNPXI@AM&zz~)stRE~IZv(ufbOvR*jkGg)ogu3vY;4M
zZh>)YdiL(sXLm-EcHVB@vE5D$-7!!xJj^}TWo$^!)QnA?*QK1AJ~<*Hp=a`u$@8|Q
zD@lv>vzKLLWvMdeI$713({uQ?em1s_(Oz}J+D-4Yc2n%Ul;%C_)#)D7t?R`2p#FY=
z3x^KdnkedOty-O)_R+%3tl3n?*l09V#&$8ulm9G*QW@jWVhd=j5n5UET9vDE|Hj;2
zO(lxAFZ>g%5pF?LZm2II(*(<m+XmEMp?d%}0hnqDO^+vTT$q_WyycX{YMJUw@`0Yc
zwl9;dB>TiZQ3-WR3f14xodq5s!=q+76{-Y6!Yjfj9#e9aYx^_HV#QM5QvqU(ozbTq
z1FHAYu_Mwv6p;d0(2_xoGc7dR!J_;u(hTY0IZ#ZVc##ClS`x3EAZ>#hHx3GD+*sva
z(lsSTYMw%FMb@d`q)Gick%(wATq<{SzJOsWecJpGc*xr_9dMBA02&WALh1Z)sB8hH
z^TVOVrBJroIrOasl+F!@7M4QUYUj{*EZ#IiNt7O9>@JA<1=7*eaDacbWEnqqv|xm%
z#b@c_ygSpqVishySGwGkpPv85fb)9+fsGrBjd43nU|@JShFfp?aedt()mrwM)KMB&
zaughk8&=9O6X#6hW_pLwr$thMLJ{%C+!)8=r@>e-Z{o*hPN}KA#Ybeo??mAeMl=#5
zvJ>3+cvKh<%?E^9PN+0;z4F_6=XGzFC=ui^?!gKQVB{umzCF9vK-Zy%=1JjYW|rf{
z%u?eYTry*|#DwvdK>#Z0b1G*Iow=_u%UVq_vu8!&8T3>#+w0&#Sc~4)A{L0MwnD_F
zhvc*5jr=xK<*xY3--=Di9dfLwzWOw<J%g`|r-1DwlbP3*nI#5Xy+E6q${lbKLY1A$
zO_s&xhk@)ZN9He|6eGxecz5m?gfJJ@$UXIG*l%=176vJ0me_7Y=Z3ulef%f4%iO0N
zQ=cxb*R`dSLz1J%oT;i2DRgA)dF@Z2<rEv)d&5bx+XPpe;aZ$976wu{;?C&L*j}S~
zJ<W(k^_>}t_`Il<M`9jL)TvpN%GXt^_WzJda*;G^q9}hwk@zy?ipAge+OVs))T$iS
z81*>tsws9d<hvjcE-awJjZ~wiqKYa6wl|bIwZJFV>lZV{EhcP4gZXYTjfXXuoUE-g
zx=U1gwK^kJ0sd2_mAn!Q(}uK&FF9lSSLHsVNkT-=m`)~*$o5Sk#DtZ1X;g~0hF6OI
zvDcP)neILNwVv9)n#z4`*FIDwisL4)Zq>AR-I5~8C8oKUuK0oL3Im;Ys%;*sZ@?$?
zfhWzDrdI3C_Z!CwyPH|Nv7=AdZm|xoQ47oUtiBOe$FXULgNm+>OT9)#wf~h8D$ksf
z5(A?Tb9;l?#vJ-uq7rsX8Azct@9)bM3l<{x3D|yMWV5q$BQj%PaI4<++!7Mm&H&QR
z_S?A7Z_&=70ew8<W@d;zzzw!T7|Am6gsp$xCqUrIPS|mk&?nCy@RAm@z3XbR0gHW?
zeluXuw@a3OJ8;nHr7PF2Ub%AhDpkPjf}I=QE}8dc%a*qbMqm5w*X!4B-?;|V3@@93
zJU#==|K?DttvOV-fKncDXmKf&X=@IBTMA`*nM2=MK&O~76vb06v)-7FWNi_)g8&FS
z6@#jh+SMXwKsPS)n0d}X@^@r80~R&q7KoqNciuP2;K007!v>uuZN~PliUOSu>HR&3
zwQj`nHLY{{FJIMqW?7az_f-D&E7{mrdGAH4wUh=#_WR||x)VLG{kq{qPf8)Bg>mUd
zc`mV;;$y@_Varu8Q&%ukKHR2co-sV!4_#xE0gmNpX+hyL^M~i{-H1Cm?%R(@jq24X
zw&|44X{-GFW@e6E9@lck_?g*W-Z@J9QT_Uj8r>s-92-2yo~%%HW^Qn`1`b~zVt#P?
zumKr-$|n!%KW*gQU+-N$_s6YURjB_9nYslXPGyRFx@MX3*J;@zgH%x$*TmH&i*~k2
zIJDr<sLW*h#@jj#Tp3V*Vb|qL8c$FKa4&b7qS2|3ukV~VlKQx@9j{S$m+5iZ5B3D!
z54U)Xb4mN6zDH?4Fs3&5HFrfemiD8*Dt{f{xwJ3sP(>)$(@yM?zhFe%6NRP=57GIf
zD#7F!@`b|o1bCui(2$kk_QNxqZtAw=lXkriY+Sk|a$?^>O9SgKQU$baomtdUCwlkl
zIHzD6e9=+xM86%9V)8_32Eakp09qXWdk9a+8?`I10|F179Ye!Iq{p9pLMpjN?^{tP
z*2mjB)N96Pjc@F-nr=T_YCEDNYniR}gi6-?zEY|V7xOMmQ;tB|7;YwQzj0}cp{s;L
z>FVRq5)Q@A2|A2Wt>(315}sjkX6ZUOQIACzD$F#{-~tj2#zjRvX)$2-TF=qx={x!i
z-Zy&c7txWCi$^a=3!gq}*t(w6&Wv1evRQ}dt^J2}Z`7e_wdyTX`zB0{ZQh_?ov>al
z8np<iQ7wMTz%jF$bPey@s2S)*d{?^8M)8o3!j_JLL+L0uw1h(ug8;QAsFu8~sBJNC
z3!QAyR`Z%=S$@H)>kJu#7svKLlqoqo)x-~*d~?PjP!-w;W*}GA#;im4q*XnG+O!Gk
zo;j_1^3J5r(?cWuM+B$OjvkmYFuGBP4vl>Kcb_^hGf#cB_4MvhJ(`6@_=UyC_0AbG
zF1L%1cY<rp@fj@=<C?YZAKRouM8oI+zl3IUcNPDjs(<w=or%5RLy5W?-S2%FF`+2l
z2nh{GU4W<2K9Ncqt!%dO(7Sh;WUNlvG<ZT(a{Y#0KD%a!n})xxtQdyNL;4P`4g{hZ
zo0?UfF&~2b6V_SEdS%N)vDb&l#jjFQ<Vh9ufEnWSp2J)`$GjnmflF=sQ#j@b<1oVs
zijwIp?Ztdpz%{RN0~6aNX;cU%8P9a^HHW$tV^|PQ(|}=v_6BASw~zU<-<mZlcUhM<
zY)l)ea`EpgIwx#eCMv*<Se*0=lOY5hX@rjT$f0z1;!wUjaVVvaL;3E+plo;IP`*3y
zwlrGGp?r6u(AyfiJ8>xAoj3*>LuF99J5kDD!DDnrD2<`+x1f#2P&p1HeN5wEI|YZ*
zoq~X`s9^+$VDHc}69oKKK1#+CWqLO0Na9$&!o6gOw1Kw8I#$g>TjV=DEz%t!(&-DU
zx1IfQ-nyfQ4y>Iixs*JIwcb!_Q}U&Bk@n1<V;lrO%pBxU$}0}#ykbzsD-Pwn;%zCf
zI261BJ$ld+0D5dFJuKJ3(;Y4=$U<e;T2hHrS*2_v3$`dfi1o-zWd!+1ag{HTt9W*a
zGC<{tXDicjqlO8NOgJCDr!voL$j(_Vsng=B*HPxoChbWK_;6?7{){RqPMtG*U~rwK
z0~i;>LB~ai#ke>W<1#||xEPd<%LwJ;;%zZ5Bb1K|x|K<-HCAhB{$r}x$5E~!_D0s`
z^tf?@rPztdpn<hzG++iPZ&{>?Ln#d$%HvuLN<BU!l*g)AoK}4kyruC}tyU+jH_%A!
zEl_cT#aWpRsMR&6bII+bCuq&>Bpu6tFpbLEGL5R$b)s!~e`w45)9R2MA-H2+mVQ&|
zSjN*8mEx!bBqW?ja}$t7Z;+tmDd!WASN}1AgsF51$|d@#k@6E@Bp6N1QH34<WmG{r
zQ?e^1Mf`Ilxo`4~q06{WyW{`Hrv<{}W}jAkXo=V40Ve*>Jix@_rDhLMDw;^SRJynS
z<IY~urvAZ(PUqoAVYl>~^tGrVm2sIeDNM!8-i7SLvmqfoacTKCPhwj9Ee*;%k+<y7
zv#HsyXRmxdf641Po6nw~yZ-2*k9Qy4`N@~LJC82<`OqG!E9lxU|6|q#R3BLUfC$`f
zAV{~dV|u`6q%jisS=f$fS*KA&m(7ULN?lT#w-H<9Zt?c48Q#t*cxn5&A3r@mIc-3_
z?6$QNR+8H#MPfbuNAr^V#!jxIzPfxG3a9GZ^wu_R+_>rDiL>U!_3Gr}KFG_h=s-@6
zv?`%vcWkMk=RD|n2zn}G$M9hEl(F|&|CWAEuZ)8O2OXT8@!8<Phe>fsG0C31c=6=P
z3l^xajyN+n`};8|-_6pWNu|B09&Y*g<L%o&{&))%PH(Y5>Z9Jt_9Z&za{jTnD0RRD
z-!=^s-6xDt*BPVUIeMT{oh(c4)6z$A6+i9~6ogrgg7%}8@)}4B-IcNOQ~}7E+Oha7
z_H|Pw4bCN)>uCI>?``GXU%nwV=Ki%oIZqPr-Wj@OK;jm4V!G1t_^+G3L|u=ZE^lfu
zJ1bq^7`+&4+|di7GoYMm;t2Gj#upa%RYg>mvu3ChGm2@6DB?=21<lA3_@*=~JOR`h
zq0_))Xf*<h+tK*6UQeIbOjGVVkr3rRN23xn%5zkZ(;6d3M4RYb=n$re79GMJ)t-_@
zL@dn{JGtcT?-`=SKa$*_<fo?fLwxIahKrSzp<dN|{lWvpIj@Gh`&yNB`_#(UeK<1k
zF5}4fx?_%#2G2-VhI0Q`<<8Nub(M+4W&C*MIi=nn<D=Ee8p=h}cbcIl$X#GGmX*%O
zpasziP}E9e7HZ<z<)E3yIzf^n$^j6WKnZ7vgrY{20g1ZBcjqX-&HH-s;>F+0{!LkM
zYtdqHnLI|kEH09diLKesIJp6Z0bc^_$@-|;p~#J%k+zJjY;=pfy~FXzoDxLpSPWWn
zF|E_|xK1mRG9PVN@TPa~ZYR_Cf7N&GfOnOzcV5=N=h9vo`@3`=c1xY!XY+_2Q(9YV
zIxSDh_-<TsV!Jk-w=WnxwQ)wnQ9Zki2#JsB(Y-rpVcBqutqMF1jEROE2>6eRnQ(G!
zwyKhq==%BWDq8su$d7v5u7&MX`9)hLVQF#qVo$1r*z+0%eDq@+3uQJ5gEH=ND6L_T
zP%rshwFj@WQzLc?xGz}g9j-3lt0NL^Mw>d^)m~KXSv(-tODwr<NzA?GQ@dFWuG%v$
zZa~#RRz0WnR+bT$-n}cJQB$*auQpX%G-}kMYMWlQ>1;wbu}WN8-9byZFl<&%Aqso4
zvOXo}Rw<Q5*PY^8c^J@rCU%rDadf2YIVS(e?^F4wicJVKyks??4)_9Oo{~Kcb4uZk
z(%TtIQiNM@Xw!B9O>00uRR?*f7<r-uAD6cbZS6D6J$lOT--=5p?pB})b+VAom{R_+
z?;sh!XVnR)n9{K7gwk1q!N{x4J$a?M6s3Ah*DI@VHP!RsscsAltB7D&uo{3B+cv8r
z(y-bhBD8^APUf-z+exypohB=-J=!NP8SFefE1^!L5;BW?O!QcG&)SSjh^#!eO4D%z
zrQb@XNV9>JVWAcF1ShD#CtCzdCBeh+lWLiICH74>fo2PxkyWiCXq~-7{R9-JhKJJh
zSQQsnN4W4b-X-oRex=CI);(90*G1BV+>U#WY(BE5W3GDTC`aXc<+^fS`N3w)7%S2k
zf9jA($5D61q4$m}MSt6?Diy!A|C?xz-y2Ft*%6~e<qWtm3G|2p)E1jD9bM0krYJ6}
z^+o|PjRL>fFjs##!o9<coSx$k4wkZ^1?BgP=Jjos7S+yk@cyx)E3Xuy6~NC`8Q_?f
zYEim=XiYW>JD|Rf(W-wv)l%S-PW7^A>v*P*ATxtOQ3pH)`xdzq<5QkGRfiInIi%if
zQjbm`o~o{TM1Mf@VDyuRC)CvL@MZWNEuYMz{nR%-&O1^)FM|YFKrI_Nt4i~Qg}~~!
z(Z5(V%YKNml2Y7M*+*PV6yq7VnW`8@s|%&pV()?A8I=?1_V_h%nfu}+Wit+4y(i+p
zO^N+CfN5_|-`RY!J3h>n=LavCHGWPcFiOCPx_0A!W`U6w1eM}k^fz$Md9hZRuPEXU
z<uUQ=zq+>tR@J@f@+^62!}&AE&x)WD))IJAfVUDNG@#WHm{~T;v`vOqp$qh8uHmKp
z0y|`j{YTbpCEHeky4N?qyLppQxnRX!5++V&(<=&ghL@@X!05^(R>#pnWk_soV2zjV
zzv8#W1%^8x|J<mGZ-b__ySUczu4&g?)Ogh#xS&0XKIJ!OZ@&0m^-x}X?7{Rn&wkZg
zq<_6id?n||s{-fD+LuG+vM=bTy*bdR57GH>3#Ig6h$B>$3)>CE<AnS@>FUD$$~Q}w
zl9(_2{1WH2&qXhn#gmseei1L8m&4e%2J#5`V^U9_vn!jf1Xj(&)<0H0<S-lSy*u`y
zTKHnl`(pof_O!_58&8>?$U5AiRoe}nQ%_8{7c=GQR#U&5G-*R%hbG@I*{KwLw3%qs
z>CK~igf<v-|J(Gl1?lmDiDA9XN6iSUU#N|~U^|e8f{qtxb2abah#a%%N2^}w7U%Jj
zxM%;lea_wXed9JKo&M2Yo-IzP^25H2O|hNY{H#t__A5L7RKD6VxI>3lt?(*6_Y6|?
zY4^~svBT&Xsc&+bt#;Terthpi4d3aGPIF|Ya~{wMVU!aF`2MOPY^Is@Q#KDvvl}@4
zQP*BM6tQZ^k$rQ<%o&+eceaAbnxNP>Zu8T(iVL~CaX}Y9d7A1O)Qc!2gR|KftrN4w
zX+?S!R;TQ5+;AKm5G^zvZ*mAlJ-4J;vLo)$xn(<ANh_8tS!KKTOt+zd4Tg5@Ff=42
zS)IN``R*wOzHA=}IwOz6<bF&7NVCc5is8l$MM+O1<i-tJdyc+)8ugnUtKz1#!~duG
zbe|AcqmAuJH>d5&#TBIOl*vRY%*)@qbNOOYlj<EYLOv_@2|qjO*e|JbS1(^qxns~U
z?l5c3$Qz5?1f4j>A{<7!6tOnx$iBAAMtSs3@-gvA3bt20Q|v=JfKylWyN3IW_31RB
zFL5!m3_K%(4_4b^%t1qZ!2v*H4L|c2XdSadSa=AIfYHDrHIJaRX#NdpN-)n~MO<1m
z>aZ}W-?UnuV~2;O^^Xi|>AWDQg;!WobY$<9Lt^|}hk5w5sIN|px}jY7?sw(Q*wMp>
zr8b$L|JipW<VI9_zVc?{$vcUo7Lzkdv-x9vhMnEEh1kCs1AZbi&qnV;PghWFZM4Y{
zZ-s{8s}v=^sAOz4WBbURN#AV!`TUv(ne+eNEMLGY?#@7&I)pCB+Mko!=g5NGZ+fZf
zjTyt%J7gpmGD6Q9Q*Gr;WrN4t*l=kwjfZ<JmgIxqf1jgj-e~s3<@ct~e7Gz*qe<rm
zU0d~EGt_|;C>onV>yvs8sjcoO@1818RQD@-SFfpQ4c4#Mr$&tdcTQxSdYKyR27PJt
zGY}!FL@Okz5rkliF9t}$Jzb;NapG0V1Z6yFu}Dd)Ew-t{?3bU48xe7#JYG!9%tXR@
z0JsAMhf&yby=$_mB#tlKh>z_4V#SKh9Z2t8r1uKtf%4$aOl4dhF|Ll-Q{D@LC&=(|
zU>@Br_8~X}l_mC=HjQfaB;fB1{PZj?Z1NCCPi}HU3lkDb>sXx;Xz+ASx+li{p#L~$
z*~~qphH`c|I3JOStLQp7i%-QTvYR+YUdDcAi<6V&dE%rbd9H{HWq_X6PN-h8F*ri^
zLK{+|JzYI@uK(0qzzQ+LRZeUCPV~69;K+!LBaRYx<+qW<nba7BoA8EI5x)>IsCkp+
z6=J#GCk~>$50dAL!_(4e@1H6CBt`QMqw_-LXfULo7r%r<ws9g><aI-?gq<@84+(Se
zH5SeiSGBiqxGOoBGA@2~+Nh7Z%!y80(`V#_z_xK>jX8U^pPQI?Ctn^+4$HZcjiMo=
zqm{zm%A1(CO1}n^JID437S*kClC1iEG^Uwa<66hdzrDA()oRC_ROz?WbYd7cUOumC
z+2)5I+H8cz#Ga&DN%gAOrL09tW>XC+w7~B-tH(eFDE(9=@TenC3af@X`?%C==%{=*
zP3f_K^d*ip2X(FA(o6N;$|gFt7z(YppY$?0$;F9kXNF6A0v056nV{9YG|VC`AqzDS
z+uMc}L-Uld!A&Pdu~3+{@W-(ps%Gh7CxV?MoY!qj_9E##VRqxjvnPz56B#*Y?BHh6
zLxwbKMSgVXG0eHy{<(7|E#B?WIooLmiTv$0X|{Kla^~Jc<-2X^qy}U-gm@~qF{c-C
z9z>k#GIuTy6Bge|+e&lpt1xbA0jilb$7ANq1C|;n$CW+TBzx)xOM}LaB_$>w__Q)e
z6^myi_`u$RKRBlK^#9=em}UGR99eZ}2Wg8={;aFm+2S18UweL;F-NceyqnbSn|Cl&
z)6?IpERK??o$sDh{K$>j8%8HD?>C4%F<GuPVNsr9bVI9gi?U+hR&|Qnma_7phs(`O
zu|~yyF6J=-XBXh-Vjg`lk2oz}NcG+|^gNIC3|TStjKM*%g4x7dHIH_v_*r^m#;$zJ
z_+*f}YvBGO;DU~(#vMaqMg&bn&WxBfYs9EXlD1eGMZQ|4q=u6QBw2>9C4WKsDci-i
z)V&i|DeFKP^@nnRO+~G2$IuW@XX@OW@%7n8#>IoG{bb83C7it6sj4=1tTMekO|Wp-
z$zJGGrVC-*m^{oT%VErtg{#TFwMrm)zgPT79zpHc&&Br4%LEPIfjd(fcg&W)7M4Dx
zg6=)4sgybGjE9OjTgI0^Tf6}j90U%DMZpnimMkzxdJ<@>JkM0Dv$9AbBqp`Einv|w
z2afC!d%|&)gh<HeE<8^Q86lk}57V)nM8Z)w`stzU#G_bMIRqVhWx<wZJ=?dP+tof&
zKM9f==GQzly5+J>s%yn{sfw4f_AQf7i1B^9cN#FaYww~5fmM@T-QqjWpRsZgl@F6o
zrKUa=mk-VXct{y^^xZIeDatr@v%iZcI~jab3!#JD=owd&b_g@?qpbeYC4>YbN5|9>
z_HMX+;k34p8AdflS8>wE@8|Yeb7JU(^_|)_pWCj@!m%wzd_A26v|BTyV~<q>#w=~s
za@M(=N(a8$me8}sn4B2hR?_76eMEbpchl&kum+K{?|ppHFhMz-cyRy0#34I|4A_*^
z|Eu-=vu}L<NS~U#b@B29I>x0+0K476k?MP2W<(&3Y~m&=i@|H-=5MkR%iQ@ee@gC*
zQLXz`^K**{Z`@=+&%}SZ1EN9Qje`eF_Nvm;u3AiZo1CD3y%Peo^??N9Y!Nh?ng5QE
zKZr374O9Nt=f{`$1gORv4?SVMH;3HelnJcr$yo2A={qpdwZL>E3X}Qi6KRX`PMLV-
z`_eNg3nxvQSb7YFKHm|Pq?KI@R73D&kT+4KQEEyKe|QAONZRvq2VvlO0%J16Xv1Q(
zT2F6yov<amLunj}t}yWg?p#?y+f8xvm4`1Q=awM|no;b+qls#-;ttfAr?$xiwBowo
zmD+|zT~<*<feWr&uY6#^OlSq+1>iSY@+w&J?p(v^%mrPR0ShrBP025hTUD>*ThfC(
zCm|IDFR;!qTaCdB8WX4fUum@3xZ8+F%j_-Eq>8o@9@YEQh{vsBOpL?+;@0E8WG!Ah
zWmjt{`t$oUq&IZ>EU|9vHh^zXMrKnUt6(LMz@G1posxP6I67|WotB<+Mx_US6(|x!
zE>Upo>=Xiv+RINh0&kRQpz6D#$FK<v+63oLn|CvF;-#GN$3`oId(P@=BX+OUZgR=f
z@R|c1tBu^cC`T`zx2x}d@{f^CYW1q!^&=Ae`#sY7;0ommi68&`v*S-wls3(W)efrN
z+HdHT>5zF9s=PrT_Ba@G*yBWY!lAqu$|A@tN{ZYJT7}|o%DF)*M{uqZ_Y<)JvW)`f
zgD0&34y}bbDQ^}Wn{ks2U#<+Tqr4LPZaGI(j}(K2IRGubS{x}4Y9Z@M?3VSDp8cSl
zPzG5T1aw_BhTPIPp98h1>AqPSwZwTh#AKy2Gn5T9b{XB)6uXo(EDXzR^bbOqMa3-s
zN^xUM#xwC)5#@=eF)PzfR^~K(72}3hhE%e#%*G}PTR}HWMOa$c8~Lup*K9baI!3;a
zyEZ&#9A>0iZfZO}Y+SSE=pQJoF#ad8%)<v!z?sHmLj}vJ*oCGUM8O9X?goV}EQVvc
z@qosQKp`@3M)scXn9~rq9;n1eYhK~)hbmS5?2f<O$|GMaNsIL&l*i(mR-2Y4sd4V~
zs`aqtrMDlLI1R~T{u`Ak*g-~_GDB$wgF{)o!*p{JG&5q3rtAlkM^FANf+M0Z9BVBG
zYpp8H%Q#{kl}2Sk%rP4gMQ4tl4lct*N$ODPn{CM-wTWx@?Tj}>n0#l=@;jO2&)g$A
zIl%riaiv|b8}a$Od1%*WJ(S(bo`=e@Pgjxv<sjbJ=Ahhr18P~1G&)13Gdk3T)c)tI
zg+lpi;V~PQHDPVJHK|6bN)$FCYTy8lSEYU%_I!&?)Y}v36Yij+x|bd`)QX|~my8c@
z9k()b&#sT=8-#9YN3Qglcum<!F3XE0o%|$0S1UTOcVO(^d2?n?9d<A+`^0dm4!Htf
z=ye1rgU*!R(wlIM*$#d(l|w1LJlnzPWifk3FY-J*@`N>SjO0V>y0BQ7#Z3hu)`2tE
z!O=@Zot{|sz6Qnd^X+OmR<rxFy*up0V>a7AxVNF?A*oKZxQ*|jYOV6NxR_J|&82iu
zPt6noqvwfgYp9QfxG9Ccp*Mx`*gYK|kKI?UO6P+^>G(KQT?*AEo1xa@Xj`^tu~9eX
zcwb?bNIk5J2+R`ml8pN#^OD5sbQ2JNqE3=&Gc;u!c|@1tYG$Gwws{t`Ci0`;r7?r(
z3q>!LT2IUt<vC1lbC6$V%$RARBgIc?ZwT~I{|xy-OZBMXPMV?`JC<q-KlDY<&C<TG
zCUB_2xx7v*W0ByNF-vV+l3IanQoh4-MhbXn%2@cZj0K8y4DE%b-9n;LAn({^kXffX
z&>8fwGig2{bm#DeYUO9)#H9K6XA#}w!Wqg8A{w@yOx)D_^Lal!5UUq$O%sQ#{&-Aw
zI7Xy1Jf<E!Z2W3M2wywQE~eY2@k|@tNF9NjTKf2h6};1Qq<@HUS+(x6C?V<mris(%
zY#SBdb$GB};M(PDrjK83yCx>4e#c<7`XiOP?}};(2WG9jI(FQFqpLF~rp*|aB0V9?
zmDG}I8<&fUTxaz7RtGo7O^k`0((>zPqcFyB$R@R`T#a77KSJ#+>Q|<3MBnIl7KWo&
zOK8I-uZ)e&KU=mUG`U0MfC#mEmP*}keM#ZMEr&ii_4${Dg#7lmx6}M3WU^M-J2AQa
z(VZ<P#Y9f0or)92WPP-7YTBvK4}D$Jit3ehkl0S3!I@fij=1=OR{PPI1{4YJ*OOZ{
z>qswL^+0ca_;@p3a&VP|Z8@}kpR!o`oJuym@4<BYL+XsRi>DUn@GOuhR53hJEwLD-
zDWu6l{3Z_jvRoqFHIa2tNgXC^S<}pdvn7|3T>U(2>B30SnqTVJdHB=|cH8xtc(puJ
zt%g>Moss{&I89a~xb}F*z2*I3tc${E(6A0P*swS?997fqOp~w`5+rt3_95xeZRzxm
zQ(7-I(y~mY?z2>k>(9K%<6_H_XY%HD>n46k2h|Y9JHN0Od1F<Jf+uEsIB3xFEDQBB
zJgLOtzD3-lLxa7g(~eNiN@bC>(Owz!dDlffvj+ERk=<Rb&Q_^AO<yL~5>I7j%F*&q
zoi~iVwr=C%xMO>pj*W?&9KGy9RuuNb9Q4#1`y0=@n}Vw>qKq9M_9O3(ds@aI9KS+C
zoR)PrxpS9&8NHVE?7OV%v&#92>-ru)Fy@;cJ$lLu*FBh$w5nt48J)&oRHt|E-XkWy
z(TJ`+MmEfdm^yUcw$5$Z4H%kwa&qdbZq}9BW^_u~(i@`=G}}pD71>FaHdz@<X~_p!
z<#PI~!W6S8xL}{8@g5synb0C2gj(4Do+z_0CauYaH7}noRBn;ZYnA6p;S$nSu_uuZ
z8){R@YFJ|clQo!_55$`C4B1<pj2P-Yarzwj6_Y4$E@c>97JQ4POt!I?Lp!l;7#<4^
znX~D{H87qi4g(Qc*XN%~Zj-qwA!TH@MLo9V^b-G8Ry8EC%az##M(XIVCXU>i+IU>!
z`BT=dE_M_*_7?}s{p8kEJ3|8+BQ35}mO_kyQ>a6_TnPSQij1&oSY)+u2cI)>$eEgY
zbJOi*JBo?J;=#+dEXUhpqjX9#iBY~Gr^G&uzRS!wG}xF2IUqJ!x^U$J84>77O6AFN
z0lVGDcoPjIxCe`c!n>vSMc_L&4UGOAzE8pE={<xvNldhA%P<l-(O!&hFTb^nZ>i2C
zDaiwtmkA|FhWFTgC>I?tI%sIuGP<DrJ_JS0P7P||y@B7(y51($@P5rzt{Oz8?HG?p
z7LHlhE-R>YY>Uhe_K_2MWv$Vf^aG^lp;135odc_maB=U@arTVCS<x`YO%a=&Q{^of
zTNtR@1L`VcXN37!-h(V39l|V_wTqOwVB3lhV?--9DfUL~@^3*Rkiz3t&@^zw`Ms#4
zLCFUh8mg=LPUVeU0SrKg8TVL*MtM!tYk7_){~AFAv0RF#{^-$y#Ofz-=MKioPsSlZ
zf(t;g<Ioh4<)uhsr;Q|YqXTY6x^e!<XRBtt$TeMzbV6P(4x2m=mm@9Nzv#q3dN<N}
z++;db8Bppa{TrtmTiPubOkUCe3x|$uDdQz+3{SChA?d_yt;&dnc~RMsbkwKuL|p=I
zNb|=b@o-wB;>oNAxF_Habw14nO}eUgss?$8cC~_>?yht@BiFIg`TBa=d6PLx7g9IV
z)w-H*5LrBCZp3VT?vLxXgw9o-lblri<&vso;;+=ykkPp>4~tcbwx+5kDnqvav2!BH
zpN}vn4Rwx|L&bUWByk>S906(*K&=&gIuF5@N>T_U*ujw#$4kN_x)Hcw0-1r&-7~)X
z?$Xp9GRQYdbd~FiEtO*L>TNo<zqBE8-b50te7o}eFtHAE+N3|}W<1I7EbhOuA}djg
zp-NF6Jyx!MF<f>2Mf#E!<4BjUx0P7KizD;-1`WBgGg_?``N=e9<S!k8iB;B57GLvi
zzalkk3~OKyG~N{gjlr`g^c25JeVt{0Q1hgP!{}PN_tCA1rp8U)48j^bA&LuWdq>yk
zF}`E;&<+i4RJqeTW_O%5yv5=v{-WQxMdNn&8+|Hk+^LDpwyum%Nl;%;)dhvutlBhr
zx4zPo?{Bm?xb6J5#;X=?&iZ8Rl@+`4r%om6OSeeS>9FecgE0En&|+@vybH}x@O7j$
zt|m`Z>98X_+y=w-xJIi;jG1|CNY|`+Gbf%Iugq!pN&go8!i$%$>^HYLDmWa{=FL+D
z#AGCNnpDd+b85=m7?q8^(^##x<ojL&6S}%hi5u3htw2Z@^sGZ2D;zve>=Rl~av@Xv
zri%QN$~_ep1phs&jowH70?qrb(7GI|rkPDvi3m%bJuA(bFCRBPZbxhl^eAU4IuN2T
zyRPWPn#Dn|PoWagla`C45Rp!Rrzi5|v_=Mhg-e%P$q}tXTGqF771hdUJ8Sp4u|Z;D
z$*ZG|E!|V4TdD3X9g!I7HhOfutf$|T?rAqZx;Q7nYc%9ZuS_E@qp9U%ye|Q>p3C+P
zes=+_5<kesVR@SXy@Lg<Fgn~v0NVw13aSiHORauN(Fa9{qU+4t`hwdFzRsSReQNfd
zf(19GkvEEs*ogeDbS3-QkGuR6g;9Lda2|o)Rzo#6pF~KRg^+10C`k17CX2pKo3%*a
z>0tY2@i&L3em;6a_t?qtlP}F$cQvkA+|g;dxvLgk6bC1^@7F6Ns7qw4=xGCcrMhXF
z4sVzGQGDOXo?$H~3{2`n`@aR9L{D&3#uw}hwA_%|o-pHy^h>rzqj>Mi)~LueT74{5
z>wmaa$r2pncslSzh%}P&l0B~x!Mh`rOWH>4_G%8L9A!{sRjHluBW5~I1B)9w1U&Sz
z{h>o1mBEH;3Len9L+RL5E0yB^@l=hq3a5#~)03OGa#D_$D?H@nExHm|l0eDpmYY#s
z$}NdjmJ^p=s`F(^4~g(Q=5RR1jLJi%ZmIc0uaWyb^9gCa4%cxU+M(n;#u4Vu<bXow
z>Yc*wKx5A_c}Jvy$~y`MXhqFS={2w+7Dh*e7#^+Iqm_rRGnE$AEBi(CY!>R%B*@3k
zi<NNiQBLf?p4G{>i%WRQciYHOienHc>IfV(e}-rkY+i_#<n5v{#;a7M!?a*rBKmN(
z!AexCy{EN5_U;KAb5NJPBXMfsVV6_!@;MB>rue{ORv}`R@j0ottYl`2VkUX0)WG<{
z$ob;m=`6BS2RNIhVW$pCr9x|Z9(9eq6k(cJB<(BdMmuJ6&(9X(soQKFv6BuA%L?#^
z2+J_4dKeY3@Hl{HsH&d5l98!QQn{y<^py6{ds^8U4V<#k&@&pQ?@Xe2<O~u8*@;D8
zV^BHni8}Tsz~@@#h#}l!t;G>QC}>x$*?Gn(Ku=m4!wXg#`C)|y3~7`+sZpz$eF75e
z&a6A6exDgF!ln)jNetK(-?eLe!#06|t;mkBj`eHD1$XY<R(;h;5gLciE!ev4?fl3n
zrHYfNZW=lN?Y6B2bA$d5bMFCG#nHtN-`U*@mRJxFQ9)7RT2Y!*kzPfqN>>o2_htd5
zSg?0v@4Y6L7>!Yrm_%a|H6}5|m}<!zUt^k4xx4&-GrJcqh~<60|Mz^~lYs2m*_ktE
zX3m^8XIzAfzy9Zvn=7}_!D+aYSnlj``S!0D(T`g69;au}FAxLNIEQYW1~rWFfO(=E
z=u)(KXCwY}#fT$mt6!~1PLF!ik&>8v;_A0~8CfTl6ix~nNxsEi@HtixXa|bICj`No
zI!dh6GG~ma6b&jl6|s?dAmy(hvCQ;{3ZX87H@>7yNB)d{Y9x4TZ7a#ZS!5t>k$tg`
z4h-7LM_V`WMN_P!v$6h{H)uPSy3cc2L;S&Jm7tby5$j0#OH-_)f;EeEB=$1Wh;~#D
zlRw1{0tyM3o>-sYvP98ZK(EFKLgcGl^4qo(A^H0g;7(4xEfuW>RixJvx1>+mG6sM7
zN=BKBrp-kA)Q01Lx;n}>_QtIn3l&ElbTLHa>b%9Dmsfo=uU>l0S#R4kZb{4@RCB~{
zUG;)3KK$`XlE2hRlbczGI2E4F>t3jn+J2~~g<{AKA@((*!GW@7MaP&Jjl1}u@Y0-5
zzfBmMJSS)g*P|A-?%e!F_2v-#;U@JQm@~0Di+n)4PBoT}&<@w(&u+2~F4qe=w`Hg!
z>@D0S%lGfy`)Konj%{qb5=Yqga30#PtvT0gfNjkp93qqUY(4hY5q`h))q$Jy<1}&I
zMijiaksOppNZr#{R8b4o9#Ccr+A?7|+MB2h$^dmbN==>Nu3FT-my>h0@_=+_!2)8w
zx1VjqRG+zMCU3;TsnZva;H<>&=@aZpycA)ckXju}*D^b|#5lswYz-(Jg0<M_S*3xF
zY;Ec*{M?S@gQL8IXNAu^)0O-swR`1UU1VMGh~U#<>%S_@`%)O<<Q8i;@WdNMmrmN+
zjrANMXUM`}H0Io*l7vp{XtYW?LqL+lVj#zHL$-XJnt5ei!aIRpe#^!zIo@9Uo$K6r
z-~6yueqO#G2t#Tg?AY;S;Yhdf4i2Y2Dmry^s8hV_2=t2>U>eZe9tNt{D9=)!sjMJN
z5G<RCa`uUqI?XuS5PXFkGk87(YH)P#7baov@2W&6%9Dt0e~O;!$u%8jrp6cU@ziY*
zqqkO;uMJZ#ud7>T7`e7^>bSn~QDc+(ju(bh|GgjSn&$j8U&2l!xv#iZRQAR7ndh>y
z-l@8N4eeE+{bOvz;cvpeAah7iU&D+n4ge7|o_A?bmDS>Avi?%FUvbv;gJg>K0%<$z
zF~2K5D=(?UXwZh7O{>73ZuXMALGK|nQvx#Qz#9NNFSL6_@1-GAEPV-6hgoPMJrhgD
zk%`hq&$p0Fr;mSnfL&nhz?58;IpNSD_2qr#(zm0>B8f!LwE4*aR`FKBv!-*`veV@J
z35Uwp(^=&Kx|LAQ%H3gfq?~r3+E|&%$d}X?S#BF8WY%cYvOG!J8UMR>(dXr}Zq`ZS
zsZ+%E$idjfF}u~qzm@;#_v(y=8$9{f%fA*Qq@LDUlV?#oDAfv(UIAR!hOKrMA!z;l
z`Ab|bUR=z*7qi$u$T`<<cC9hlFGU(|Ec2T@EW&$}U)9;Hlyh{D%IR1cJG&(h9`eq%
zqW9jg9h~UpUi;T>q<aIWiAK-HW5kr$KxQX`wFt+waOk}+w~@TN0+*J~nor&%5#p^&
z<=}!iB(_|e2T7m-p-v)17G5<IwX8rblm~aABd@SvjFl2?$h|#P3|_?ZS}}$A?j{~f
zq{q@fQ09IGitQ_;aje=k+6-Eo6Kb1=+Fle1==1|u_ubU9yXNh9i^NHvjwNs-Mc}x=
zvW{Fi7cCy3G6?(8l53%o8~Gqq8bS^h7NU9<gZ&oTFodjVLOn`UPNQ>+#TpvRv4)03
zFDx)&5e*|E7o-P?MD-*}`b#>?b5;go-=E2c-_AWSYHr9u^u>9)G>|+dUxp%X=wc}T
zK9ZlWjo~81!{kdm;fJov9a0hwNcF&~q*6Is8#D+I^O%OlQwNs9@O2@9N&w)I>M4g#
z7Oo$Y9V2Z;wBJg;uheJXj*I#4&B&WFG&zQh^k#v70e?wJuCqr^n%vgVI(k(0xO|db
zI$5fmzj<<9P#Z(@0ljS_2N_NZ`0>YpV|2DBK=R3EvmJwp=SEqL>&kfQJP_=4<oKc^
zOC?#h{1l@qWa<WK{#@ctdh`t+XX)K%I92C0MiS<pHl25wlwfKyx{bjQd+iw1%xa=%
zqX}AqeEa)$I^G<j{Z3%^>+DPHBrnUvN+bHHeYi1-$G|`%+vH_?-yNf(R};JWeY}75
z;+^NstFB823TL^v&MC~PbsI4&d#saFRE(RmG=7S)PbnrjJ@i}uZ_TMJQZ47RkkgsW
zynOA8D_5>xy-Y_2=Ldxd%nJd&ddh8_9I!`0G38$o&wL2UmMW***up{)rzP`H>T6hc
zAr-S()Q!gXGKapNg2e<HBMm}o8n*~Jh>9PT6_vTzW01p%s;GGpq7f<|-zi;Z1sYYc
zl1wj4%M1#LstTT$O66IsyVo-_pS*6{Ew*d-pvW~#>*^@RRJ#Q-kQ92*)?3pUnIFy>
zQt9=Y`_~tqfZlj|?pOC0d{)UBP3-P(@+|F9mbmDQsKlv#mg|!or{iWw1lsa#v<zJh
z@*8G`+y;%Uk%Rjn`Up#?04H{Cru4{_=3=Sn>ptM}m=R28f`-C$Cinz&HR0?Kbf;M<
z@t(wEK4$)=MN*w-d<yD_yM3GNmMp}J<bHj<1T;8R66%SD9RCW%N>;m&aFb$XjrEP>
zO<Tw@XkVu*D$YO`)7~O);FQrF&QaFFkYZE2n!NT&IgSv?5#dYG!&j(%*GjHMu9C+p
zB|R!WBcs<#HRSY8Y5L^~9EhnE3{3Pp2EmTj^A0FAiBSqVoJ^@^v3`>%p>h?~Nuf|P
z4GML9f_@!A9dfK6CKQ#jWibktzD1Q}TN`Pz<XTZdKCIvm<>Y81Th?#|SQIlXAlvI>
z0V%zhVZanHrTs=d3@v)X0=ykxDsh68yHg>qrDa-Yn3gSArbeFJ^d-_Z?#w6}8|dfY
z8@6!GD<S>~rS59G`ME{Q*p%L_jiY{z30=)Y#?{STynt*X+oIy)!p6s>tSXj5DfOl!
z+RqdGp%=A{s2fwA>tgztZ;<hWu4dq+vMZpggJBcnFym(}J3QKN_JlcWL*|_=6J~7>
z3G@pJ@m)Vl`biV*ICpixyh%dchdF7dYyFCyZ>;{|WZalVal5M$c2|dHj^8pNA}uRs
z%%=161RE1mWJi5xLT2Uf`}aPsGw%)T-j{~346Me;)YEE&AHojx<xX&Rg?Ca@Cs)=Y
zz|Z$td0JaW_dO$TD==Dyai&^vQ(PZB6Kuqs_>pv+$xc~EmPeXDC~L}CgLAPH48WId
zs1!n3y#q1gdPq6dB#Mlxk#eMD@)7^1b}?V1?H~owvC(}Yuoyy9>|e&Z5BZ$y(xI=K
zK3+O~y7+0!?`wgpoSm(4WvC&4u%=-$5n>q&DStX2gfdH8q=mp-rICNs(m{JqHDMO{
zXQc=vh)RTwk2}~_`fHZ-Up`{C=rNm|5{tY0wF5uTQ*=vNrA6-!psq+(7ql>W#n1@J
zmYTH?NwMstuU_j&+lrEnde33oqnbJo&o?Md^|+^N<&v_<Mg2N?UOdSK60_FpgoT8^
z4k?*M`8D5#OtDs>|0m?f=Pq#_Jp4%Q(*2^kf*Us{yC5+jbJoEn;v)6APEYKB57qyt
zpcb{PdFr@?>~PYpYg(Hyh4wk8d6l|BX|3jNka2_AE<b>-6dHAepaW|AX5o8ky_0CV
z9UsB|*lVzvLZj;D*ek)L$8g8}i|S^Q+5EUgQDK=Q0^D}Q)?F&dJTvpuXUH!*y?j##
z&Q_@74tb{94jm9+<LH`QHoa<B#31WLTbgSY$;+hQe^YYJzJct`s&0tt16H^r1REQ6
zHjd>8uY=9>2jtx|pxOIS3VcVH$}C35SSi_JbPs#~VT@wSs6MLsgJ90oH`kIy-m92T
z1}v9$boz!|Dy(a_?;Yuh-j<IY%>@!BKSJvL>!PeFTMtSPDRXSL-=|}Q0??|FZtgVM
z1C^k&H!Gwni%uHX7|-GTJg8NW(=wYT<F_r1gNg~0uH|phAK6k1Nunj<#(KLAcUceT
zZpyaR3{h}#eFGq;`Vm-%0_$f*^m$7?v2*=Wv6x}1|8Xocu~Pib!O|n!x0p@j=@r8>
zJ3j1O9>*p!J6H6Z4)dAao_3vFcA?g;SEcfzNkN|E8`$s>%7=O@fZJx^CVPLFbrf-i
ze4}lYkyXQgHM+*mP%!vp)m^oq?pqyt>gx~5jG{arFGE^8H!R|Cso078BjJNaJE@Mg
zm_c&Jc383HjO~Vw_vpN0P?jra4xLv9EDO6XD>`Q1Qa1Z!?*X-dGu_|Nks5KCl->sG
z36XP6RtV@22<9`jGBJ^zN={Bqwurzj?<eM6ocHbz9MAnGE?GN&^?JkQYsSp-_w?T_
z44L)!p2L65B%SVe?{A#1Zey_T#Hs(ZyZzBX_Y_x}hl}b2Z78N3eBcz*$BpZPLJ3<J
z1`ip&%;TW`3{WO58azZgRZg`4S=F9^$}E15O7$A$zAY>jn*KmmryJw<9LBG!aEJ?B
zSFmd5o`c(?N3H#;sQgpuUGkwgk8dY7#7W7Th>^abBX@0DRg)cmWLC|Y$-FaXT$Cjq
z9u9n&e}HBKRJ;a2x)Dn8eM9aAgED=G_2R4!3QnkKLExuy{GLwVSHub;Ta*;;)GS{j
zIVhfcWy$Q>;5BtC+Hl^6%bKSN6c4?B>J+#8#Otq74%FB>)3!iaj$+E}SPZAnWt{$?
zP&o$??Uih~mV4EdndrO@e_dIoYspYK57jxk^RRj^$YmjR!=r{3Ap7B9y_I>s8B=?>
zGRoyU%X*K(-KpY(`j>$x_F^?baiTK<zXjD6NQzW{l4JH@A!{vWPgbdnsdh!gii-Jq
z+@!zUR0FiF+}su{1r1C4puY349H$DdN-K!gYpo3k4~<d;U-aRJUV4p)$*sr}En90T
zS#UagJ9rp;Syp=WTT_i)&RMk{UgwUl$~mh(#jRn(Sv9d_56fG1w_7T^M^njlj}^Wa
zRD%?N=&Tj$FR?a;D<@i755u$8lWFA>qf~~ZF$(ZwIUAW7@#8|vL)+Dj@^Xr_RSWY4
zb>tFpclMH?(W|p3W)C*)pEN4DU{JVo&~)#Cc3BxG>dd{Dmvg^}u7xxGU)vmn$h_%(
zZ{N*U*Zs9=%i|>zN0zx~etT{HnTh4AeXHj?Gzh;L*E>_*V-9>PLI3`EP|RS^ZZ$0E
zhi-icK41oGLo<dQ3+sYGO`WzQWzJgv-0#0CtXs0;F=_bV+Zj~_&iQ`s8D46Edmy(y
zZ~FU_l1g*NXRKeJyS(4Ul#u~PwgELtCmO3a812%G4%nlN8zb~I2>)Qh0)lW@RtD!^
zh(8)xF0CUZ`-5+<zeY6D`!h)hlPAf+45fk9D?3P0<hCGnl_xj&6z$uqtPPlInnmL9
zG$_FdRBjQgW?&HJ*k@?p?j0=2gGE1>8(BMqaqkAs7`@0lz<!*QT3Cwc#5G2VJw_CL
zvY#xH%t^8I@WkCkv35R-3P|_>=}K6UKJ|e^I&jzq99oaobJB|#(1j0jBc<bu7Lm}!
z^8%~F7yASaiy6)uPFPLmkCS>quy7)m!~!V^ZQdL3af=p-YD>cdmgy0z3A>42#2lSm
z$hSvt<)jBLVTs|UmTn4-9#J`V%$kY2mV1`E-RVGzeb3}wxnGNnie9d<c0;01ta^B%
zTTc>DD0R?`AYF;^j{7y}S6lQ8so3PqAkUyCOg`)gl9ds6Bw_j3ih@|5sbM@nn-c;{
z7vTbhb-Z6E>7aS!x@@j^fOu_AW$=j|A$e}jrD2OM%y7re)#q?l53_}+j$qGnzz#`m
zY&7am-V*h7F<tig#vaomyhGbdi&kyeIQ2}BXTZ%@XMIvSZkl6cdDii~j2ojyh8_*C
zSuneRn^+b<A*IyMd7N89*y`Ny99LqI7n3x@%ik?xm_ux0Xfl;7CPAX2p4(2$+tR`S
zL^X*f*^RrW$DmfW^y@a`gz7h83Ti3Gp*3Wv5q7azLoGpIe{$>}&m-*;jR7*Ft&#J@
zra`sh^!u}<CuHENpKeR$jy-Ixdt>sZI!gDtb?)alx;g8U1{)=I8=m*sF|sIM`h!kf
ztVHBShecor?8fGd3-E}Tax#X=TIXVTI{z!>0ye_zhs0##t<S}OOY2<6EsGsD*=BWh
zM@?abe~#U-iEj3Z?)|8HjTe6=cM8e7#tZ-1yyx%PzRps6WA*8|Uh(<Kk>cN>F_CX?
zT5~22`%OoT@m}Cz3~89%zM#u42BkuFIt+KPBF1Js_!Q%lOm|Z(iu(EmSNBbERoRsE
z)A1>p8AtNTIqj<dR{TOPw^^JqHa~EL-QoOchZ2+a79Q9IMdSG7iuUBRFo{2sS&}@l
z(AZ{4R{F9q#bZIf{pf`4&<N|8C66|2{(FI<aQtxW&7&vjzNuuF1@FJfqz}wanv)UF
z_1FtIn%6>rX$v%l&56bo!u!``QzMUp$VWk<MKhZkUzseOND?oBaQa3yzTqG2la!Ab
zVgA{*<YrGx_|3JvflhQ@bhPB!E`HptQ$MHCyiI5h3_1SxEe$!1^8@p8m(1IwkmuQS
znk-7~!N)-XIEJtqWF@_2lw3=z5;sWN0#-L+@zrP8mB^1ehF5=^E}ZAtmNtXh-9RA|
z>Sv{X3)#79MYl3}D${sELy}N2J8EIHbnpCyi&wI*u9Xr`lPoisl%xBlO}Lu7e_77D
zNWN2TEk^AgRz^?ENE)?DWpf%!*tjg7u_}m2!wSEW6BZxItw@)1BijyM5Sl;3!MT9E
zAubngH6<<9PmnzN%q}@YasyK2HR*3Hsi<!)Y&=#DD>F_e%Mly&V$w!MlWfsf!Bpyo
z``CHUf=={Y2leL;#deJR&MeXpo+5(bqoGtza<Cx}5dLe5GVF4#L!bru&OE5&r{Xun
zFt<zlt`-Bgo7-4+YulaF){<MR9Cj`|_3Os-_Nx#`g`3(CU^o+=_f@>nzptLk@!Bxc
z)H(gc3bF}n>_H8rDJoIXz?lb9XYv48#Q{IMYTAJ6N^~L4Nzm{v5JDGV4k-;ViX+xk
z6YhcJ59uG{wgX)Jmd@TAzM>+Z1V|IP!P?7~!Tt{Iq$?+T`kV9JyZf1)DdYPzRK8gq
zP38t~s@|}2L0UB@)ZZ@a;8u8MsWy)+ETedrNhNe0j{`<&Z0#vFo+e%oYY&Yyd#ggq
zwd|~MiSbmTclbe6+HPG<k77nC@DJtYV(<%CCGhNl8H`ghc!o}Z=dbbHk3GZ63(qm-
zFF28dC=m<BUzGBgp9Ywn1P)nZ#T_6Q>MxKFBcyp8^1sp40hRUh=-1Hp^t0*7H?XnV
zD_LvoL==05EQmUFr=FB^G}M`;ZgKKS|DIajQ3@cZh>`wnwTrf%J-c=H$urxy%Gojc
z*RituYR!l<?DfAQYW-8<Nc#(Z*(?vxe}OTwPJtfuToPeti?R@R%6bc-y)Qv;&&-l?
zrR-_sEIa1N&a2b9{PAt;+1b2V?{?#TTNWp5h;^W{rr9j%53aZq+G;m(_rxOZKc?DG
zYHMq`X?1m6N_91ARcHc?0-Mu}zzzl)BYKJ6#)|h*4C9S>rl&;_J%;h2^%E!<R)n{Z
zdUD@r=hI94_%g8%xywxxpOlny9-RHuGDQ2KEri@PWqQN251yM&YV(xK39q8=#_VhY
zHbEsj8&dB$JMrxBJj>2T_iJt^!RioyU-z5z_^$@OWYd@f$G03m5RN2Nn>t8;{wYfT
zwAsGB4MAW7G3u~MdccLezUk|?+VgJO58A(dee;oEiWQBKY0At-bU8+7#Eg{@4EkUe
zNE(XYI5I&0<_wKEqk}iIR&0D{&D)z-{4J5$?Q1U2fUB|d#<x51aoW8d-`a$=Pu`L0
zyGT<p%Gqdv-i+KAG?No9nNjt!>C`Z?3>hLC+G#jT!o<ZEZ;5<$u2k?x<@}19D^`A2
zv3UAlQt_NsQCkFUsZ{^<AJT7L6AQeF(VgFj$JVXV`QPqHfBF?n^d=^Md`;A)v{&*<
zDpO8i8X9($U{%14aF2mYCpCqmf5^w%RYK?u;5#u>A2K*yD}ak1FgSdUn8CC>)+&Sl
z#^7H;w()0h%o_$r&8P=+5o1crPf_5A?gRV_MgzdPNCgg^9C)%7LQmJR`gP|5+00;2
zS_`0H7esX*xV(e;fjLQeItIgve@8rsE9#@TE#d?m2|0*W9*vf8$vsNRx-b>ksqY-c
zX<8u;kw-#1$z6P_lJ8t8{5*Yny#=2@c1j5xf)rvXPK|tn*r0Xr4XK!opN?;!mgn7X
z&6!t|{rXI!@rr9)khutTxjn@=nqPEHRC}a2b}GIPdJ4lfo$TxyT~7}CJt9}Ax}apX
z=RR~zoHBH3-kOt9o=g9IlNg&zzi+*k5wf+qWL3~2DQ>*=<k(?8BL{o#{?C@OACG4j
z)!bdT>U@pE;JCc7%;m7Jn1Wt?L9e#dazK4lpcixwIfE*n-FvjR^Z^M9jIv!u4=i%&
z?K_osG2s@<=OaNK$`Q=)J<li(*&b<^HCATU8b}vMNPU*%B;{G~Nj<d>GRcK?+*{r>
zlUjMXvloZc4XTm&x)5Ir*d4)Bz(|nl230uYhrcOJtX#rb)_+Rwmq|lDCboIFrNWjB
zLs<KTuJxnUyyp-ttfYeK4rEDsmYm|yf!$gH5!(tg#F*Myo;sJwpFMd}+)eMi@4ZL-
zo(nwSuG4RDCwd{!LFT6_dUq+6&>Am9#|)$#YzI~{XeN8}AON#e62=%A5KXA+(A#8?
zS1;*{8Yyxy$+$aURE$e+2Q%Sc-oVLQT&QH07{&?lDg9n)k99GM)kxO`oUvF(8Znox
z3)T`{0St~6(2mVSw(6*rNzNdQp)z~U4kmQc?&Ze~*Po03!v=<x@E5QPW!xXD3jJ4d
zcgp$mbeedpxj{aDPA-t==Kesd&2~DpIvTz14XQ&%fc{UjEYTa=6ca6DnM%|+d9?4}
zxf`TM+m^iR9yEMiS;eI(Iq%kl?i%Yo_LRDCi}b<WTga#s<&tFUS46n4zb7Qm)-~SK
z7i}s%1HH4*(@qAzs4~PN@5^|jw2n79@|%Vt<&Df^tMJC1ja@fsZ$mJ?km1#h@b1#G
zh7gQ3YKAm~v&u^O9+FvuH{clYX7HX;X9i#20&XD%G5Api&uFa2Qi@}6HwMpcg!hs<
zOStv4LB=Py5#C$sP2o@N5!O;Vn|m$y4vn;=2~KmWGW?NKl|{R=()2XC0+mmrdq_)}
z<fnt0usfq<D5Y3FC}v=&52f@fXo5Cmu^K3~3iY75jnE=J==w%zu^#j&gHk#{=+(8>
z;KrboN~L;G88b?wGCk-cP_&bp;tM{7)!9wzKu5R%+Ho^@ckv#DOCK^g#aS-jLoA^q
zgS_&zd=tE<0;lqzKHNfF$3_zJAm9{pxgJaL0E6FVaJsw7ySKF{(UEy-SJx=J`VKIS
zhYWuzBLaW%B*TAx3wU>N4Z|PQLwSm)T)v0co8b?6x*5Eu0$<w#ZXpJ+@<$k);xE@@
zDaJCm6N6LQ%kW-edpZ*(M+T?Vm*Ks|Zfxc*XK>nISt_v>_p$!AlvXsdw1T`JOMz!{
zV?8FS8diP>gVS+b$mpPh&u@e;(uYgM3{L5=xEZ{%5snoBdMfouYa8M4is`|RFgT?H
zJQ8|vCkCf<Ko9lcjtoxeutFc+a1nI4%Er-`!Mlk!**IQjaJq72cz5v@<1}p;oZ>IT
zdx*yw{V`imPZPYSxRt@rw18WP*$gh%Pw|)Qu@oyA9J>?hq4>-2Ug9t|j&Tf5@t5Jf
z#Sv^A?=UzWM_G!r78}^OeAEK25t($UNQIQs(0X*7=6hP2aT*!QI8DQKwkOiD4OVEw
z)?&jd2B)-<;oZgUlr~bXyoWZzdx+gxpQf~c_Y`|D_>LBE3o(MhPq%<uim4RdK;xSA
zC1Y!G73(YYl<LDZBDey|gT~OXDE^@ge>#Sg$7BPZr>6KrL%|&BE>30mV;2IPj-d?i
zDRyG`P(GZ;c!mtO)c!=_bY%ce$3%wr*6w3tlEUD0Ok_#jTAa%8!3sqACb&jq61=Rf
zG44<6!I&6OSy`6UDfgG5o#@T$4WFY9f7Yi+WlVl$eWJ2rpt@QGEr65zw4e0}e1}Td
zs6{TX<2y52z<Y{TtgqWzz%9fv41THw+)|vt;E=7F)!$2WXY)0!1-!Qy#ApM4Cy$ka
zk8TVPwSa5H%amr&5SBKp-$XT=(eiT!r(=7X(Lo2_2sr2f_}dD6E-g>_j#3X+B;yOL
z9%)7+{A@G$wnq55W^nL9rJi@1!6!As-)#m@YlOeo1c#ldw&6d*FxU&bs|GOH3u*yw
z8-VHo2cJemlYcR<iyWs<^WZw$avVQyR-(CUakQg>wT*31!YiG7Mpq`hb|Grt-16%3
zsr_T3LV8+`Efj{hudE|};%VvUK7))3+owbphc1XC0_n2-$>Qu^uYY~DPtxm44xaR<
za_1+A!gW($VY#;yd!=E^tFKt7UE^H^)TM$`QPd?gD)qJe_{HI&tEYw3UdoBx<KMw~
zeb(5;F&EySW>PV_I&5=E&FtxM`_)%Qug%Y2HmXgVA@QpVYu}pa?pfsT7EqUR@c1Ox
z%E`0JH*9_4(^36-)yXtqlrAMAHpdS3Mx4e%zFbNXTo{Z%2*{;KtjvEjQwkl>q0i)A
zW?d|LO(?B&uMS!`ZLnb1Q`;ux{kbM1w(JR+Lk5#>U)c{CjU;pjF|hLO-?&gjWv%p7
z@pUz7drG=B^h)s#8SBzTFnh@KNlOEKf~E&17CGj}&pr}2qI6&G(ob^ZSB%XXH*wqG
zjq~=dTD5!RO7-XNZev_+B0?OzhFd$&ySR4zv86GK?#!L}?(E*fr-a3odx#gkzTJA{
zi;vG#t_Qnm@>UJRLu<~eR((|toOpO?^|l35-wFv0c(7*r`^EF7X66;XRgn1AD8HcH
z5j8973Qv&##->gv3v^C)jUBZ-b830#u2Z9vr+eDFj2&VVn>adwwth$Q7dqp<Q)uJ3
z#aX4N{L-VMjmmKsl{4sL<;-hMzLdUP@4vig!t$t<Yh0414a-ZObu@7$vHP`qZf+0h
z6e>%2SNcY(%cvY}-P1iOB4>}^MBB{BK@Nj^56;?NyAYQ<^*`C}6@lK9u@SK$*D%sO
z&^nEp!lYtxE9^`Mvd}nNa%PxidvQIod2F)pwB()@{Y*;RA=kFg1!2gJW%cDof~oEj
zOzvp$8-7Mz{DZs2CaRP;4vcpv$a7Y#eyZKTnJL?b5XaGldWO2dY{b&aH4o)DDjre$
zRHdZM^p<`~tML)INZz|j{rMZxYj=K-cE9^J8GHLD68V<)>r2w(nKOwYKXaP+{`=xI
zO4EhdjoLB`u!>kY>#k`tlcNW5<a1>g@?~cDpz*f-EEkTPy`A4xqWvP$ANR-_S*@Oj
znfh+Se}!n=4)_WpQ0NMP3(L&VmeUX(UN-2&APm1X8~$XEpbX!^PGdu@rO_j7I#pPj
z2}6?*3w%1*zjJ2W{;s{rJLIR#?;^z++>V?hvA`@3JSRl`Evwf7H9A}JW^_dvVsUAR
zhG^Y-MAqeX%idTR*4-_?Cb~9sR`D=tajDQfZGLo@T%&-Zo%s)>>C)XX!{gxHExDx~
zZccgh87YZ>Kz&&Pc3iXrsX%Z7Z|eZ7L({sYUAbqKm^;;LPSm`jAtb9>imihYq0i4Q
z4&!lM)an!bQH%Nq`2HGZr7y<L6_qvf(YCbEuya7nf`gQwj78Ej1Kfp-cOrd4_KsPw
zyf&X86`2@Y6Bu=1_KY=tt5>@OPZ-uWE-L5j@+j}LOpR-{f5|)Q&z(IYh7HNfnOrhA
zq9(YgI!t5ZpBR#hsoU2-!og;!#=Q5a+%3fk)zL!?x_VCz%XyXElcmH8!Ge1Q{pOih
zQjUUAZd&eZq^1^5Ffb=0SdG~+f0#2!VJ&xTh^s?(Fd3$uJJ^1JRUdPgv;_Ww)`h=7
zM7t>Oc6q&ua=Dw4MSPof=^Z*iY2k3nIayGHs-0wQy$ij|g(~Y<Hl1m@dDF&&s5V^c
zkMG95u{dF=bH!)xY?wZHanzF49+kM@^CwAK`e^OQ5t1R!g~G@9O2+S(rCYwEe))9q
zdy_!<ufcmeqYZ0)%0G)MfYL{E*N*e|Egc`Xe!Ts{xs#WVUcAO5dCGv|{F*(HGY}Rq
zWJKKH0V!|In>;nf&N?_XENiJxYTx*XA&&N%q2-6F4&Q@-B*(Nx3H_KA7vpVet+8R|
zXsl1LAPu5P*eOUQ3Wi%yP@g1ScTCw;;lFm-(pBSwCq+)3cy__UchgF~h7W|FYQl4t
z@%tCt`0DGM`#u|P(C_Nu{ZH!VKiszU!EDem7Ezu!e}wqW?r2;Y6&!9+0z8yJ0VgB=
z7jcvG=m^iQVcQaqriTxlATQXhIAUyIWR%mrOSov=xNf<#=*KOr{Pw6B8e~&^j4sO(
z%Ve+NzwXw3?$yFqKd6|R<x-sDVQMLKuM^jt`ZaIoF79V>P-&TF{JxqDVz?QYj>3qi
zVKjQuEWOAa$8PVzaMM89E;L&{+r8*cjN!-lGmg#M>{wU!#nQRoUk~oT$j#NUrqrn_
zZ?t!Zvr9%{0z8!dI`vz5duf&_S!PuK(<x$}wXY$&@>pgEuE*T>G7GZyBuO2+e0}8x
zvY4o9q-?=X7^Lb*IfgNqu^Fa}#+KJuvp?XH@hT_{j3lK~rK1s{*|9NjBUDTKBc$C^
zNp-j|XvEqd_J?X4jJdtzH+?i~l8bF;aZ<LJ)PehF9H^z7Kax+0%G6QPjv1nlP3{41
zW8J)AJuFt3^+^4A%j~u9k6$qB5L~cP9kX7al1@yQ&XG1O{c&!4X^Ih9+pb{$1a6a9
zQoFDVe-LdbC;HeOWz4)q7Yqw&VH8v{2GfJJh;5c*s^ncy=-kh1-J102?My0C*Di6l
znWE{}wr7a1Z&1%gJ%fCFLo7RJ5HTMw{e0m^>CcLi{Awcnc7a&Nm7P62A?x6zf_>wN
zfsu48ZP&y}2eT#|K3hilB-<GbTp*iLK51ZR%JS?nD2=F-p#~3cHkLJPuY($zF({j1
ztcIWE8d%&igR&YJl*Z(tH3nrh$YsqKl<_QDgK`g`DXW3eiPZ~zl48c7Oh#o;$l?^*
zme$MSYZ&yJ3{9X=wwE*L7c$h2@+{N_enKP6u@_LUAYB*o9wu}abGa2_fcc<`-e%;N
zQu3A4k-g%0?r6=nV0<|cb3Y$zNj~?22|+IrV;N!$47fEFlE%D`gSR6;T|0fikU{*J
zC$Af+Ck)PgrM-~P^KJ3{6|BX+Y?Wg*QnxgXDR(d=YPxu^RnXx!yHc%IlN;L#B^zYe
z4do5n8x{_+!x6{36JGP-vyN)lnpmxzIkal+o^eODH=}&IVn%%g?s1>3Nt%sPv6M0a
zS?}~0I5XEd_0=#pzuef&b<um*B~SJA8TaahWuFz7UpH}Zj<D{R@b-!yzO>Oqy4%)0
z-n;*IX2W|Qb=T9VbY2Y_QcJ0m_K5FBP*-UcvjJmE-!#moZ>XFDI?;QW6}z<>lrnCN
zJD`1K%z*|tF=ooGsq$^@j&G^hh-?z<TD|ETy<RTGAx^B4=_+REoqRx1@sl)(iY)Oq
zG7s@S3sbhHrmu?d-L@!s<=7=_TqouGl!b?{s|-mwToPFnXeCs(&#Ck=>FQpblR0(F
zV2hCxBh#1pjUSL0?C##TYgezyo6AbKPwv*G_uwGRBhnfBk|{=0mWh}cpy(h?ZWYd-
zTrrWnCPr2S%qM4hhx$n)$Y+5O1BLFn5h?XWM#B8UxgKKQ_MLK!p<3{&v!M1L>Kn|L
z1qqbSFK0vyArKAxO|607freKaa<g*UnhzWBZ~EuMJz);E<_!mm)??pu9${&a+tz&O
zKocXsl8VvfinWxOy)2s>RQ6r*>w%;uXKA(r-b^2ZmckIB@f))LP#+4NjgW+xFGo<x
zEL{WdWM@vL3F91MkdRG9M#V{YaOkEJNjcQDpXt2r{gz9g{2}C)Xbri?>0+Mx`V#3?
z$y)j#_vepCcP=CLM@Tnvh(s@gzh<s9Q5cB%(22#NIAGN^&Qd2DL1;iUZ~z$^vt__*
z*fL<7RY%^8%HJBkPf)6rOCOJ+?*5BcNZl`$lHC`Jw-$sNv~4SJ#W>Nq7=O2EBXFfd
z?4^lA&7)1a2ao&5?DKg`$TF8n{7UD^+5)S&v-{R<UzgWmve<5ne^>r{tL1b0)NNm%
z*I^R(Fv7Ph*5^U2PYyC(kEUCfF|CHf)q~BRDPt@J65ZK!UJ%BasXxNd85yJ?8E^;F
z%j_qu4YM;(S~=D!dvoUI(~<G3k4Go$&hnU%nq1@CFDuixEHp=*bNy<d`twmmZb@si
z%-Thq-5I%KZp0ok@`p>r_Jh>)SEWn$f03T#C2UB^*b?8*#<|8e@y#WEFu)MiR^Ztm
zc*065M=mOfHM$t#UK&MVgA=rm5?h8%rZdNxM6XP$$xNO&FVKHs_T-A)Gszdyps&9k
zzkTw~%PDtFMz72uRnnHq%EdT9i#Q0?tg4iXjFR`4jb9gkSh{#gDv+qTk44)Bv9mpL
z1~w&SkF1$#G*Bt9uBmQ7SGh4FR1J*{c@1_lZpNldDS00+Qg2<mc5{@lW#!5(WW&V?
zc^~~lI5LfJykNr3ySsebe`XXKetzcYccdgw{nhD{UvnkGyt^y5{yKx4l6u%_H`+^~
zM&tu~exe=u5@Xg2w4jlO?WkmK+{llupTDy@Ia4u34zPl4!>OxPa9Ii;p!M?f;<pU-
zTd{J|jCLk7$FHaxIW(<LuQomXJw5$;F7M&z;T354O5eWxx~z6>`3$a2WP<e1Yj>pw
zY_~Q1<1lF(pL29qg1p<_G<hiRwTZir<x<+s2Cj4+s9{IwOIP>vXFWNLiN5Ef((O%?
zTf4C7?^o!|cFNGE^;xdW><qC@q))3e`;s)4$?2V09!jWL7$0ZqRL1|L>F|SWMd{o{
zZ^rX&JN8dbE?gN{F{HA5VTGx9n74T~apmgy3%5<LxO#rkTS*S-8#2>2Nn?BTGV5d<
zI6L1sHrdwL#NVC^n3t6E>$#(!{&`*MVa#=fmfg1EyX9$T*2c%sS*+|VGy=o=x#to$
zKI<IfGtM1iZ<24~ZOTTW+r<nB<O96_mhMpM&!HuKhrNra9Pk{CSp+7GNLBiOUmjG#
z(Hpw4@D9`8`u^d&(KLa!0pBqDZ^1wamy*?|wKw@bIByf78CN`7dxN)h$xI!pwZKTH
zqP`G}w167>AlebDv9BU5KSMoA<)%tbhdBS%l%xe^YTFm6N$8U-898G=mL;QZ)HeGR
zv^$NZQlQq2WqX`%4cc>UYfuYzUyj@wa=t08S^l@FEVY2jIP|*J1g&X=_Wr5~io98J
zS%XKHo1nM{TY)}qg3|lOm3sR(L2*z{R~F`3UEk<=8>OsDhQj6v+h!xwq=c2FP@3h6
zm9;VK2ejd*Z1tVZQ}1-NwlM^)TYT5_4is<kPDgheL(u)1??C-$zQY)_c&8f)8$*ml
zi+8&5urb7Vw0Nf*6&pj0O7nM(^tLfH*6X`0Yt#P8vUa!el#(r>-M?*u(*DY2dvt6K
z+Vi*8pca#xptN^#4VELDpmgpqsKFz4FRY9M9Zebfm_cPNfZ{7_0o|Nv4K1PFUvGj^
zeC4t|erXNb^PSe979LGdimzOQrD+qCj)M&C#nz^b104Yw+MDqv8A|)cpazdvetH>7
z`zAvlH$mAr$u;zEf(mv@S@<m*y!2|;)q&amFY`J)<E>3AZ*^3}jCXqa+ordmCiQ1D
zt64`?%=%xR_ZHNB&bOfQQ*U+D#@si3`mK)YnEhC#EwR#3AM+nOL5sJIl*I~YMh#hO
zrv1H+uM7;k^`xWR654%D6O{H|F5BZ$YtWv1T7z2jZGzHX$~9POo1nCJGPKv<O;Fk^
z8QS|!6O_&=1~qupq|Y<%Cqo}ML0Qk`8u~Xu8LySgicL_)Yvr=^Ac3IJ;X`OKG|&4G
zY_KeVy)P3;t@QpjHa654KV^^R{v9xU_<;USPSQ?@l^Q&F|KOrM*V0I>e%)EqfB5kJ
z8fV?Vq^tCg(|ysdneqC6!pF=|!#{-&DWPcecXlTGDPnjqjjng?@G|&B-ggrJ(+nLQ
zb*?e$451NmgwX{u7c?O&y}x-F7pZE22B|;>X_eXmTfuMI04Jvbh%6`R8b?PB9%eX%
z`Po?6bxW|C=r=sn+s4AaXS`<8-z_l~esRzYaC97?anSusG<NpGY=@4Fv9lXA#CD9g
zxV{<knB%gohR%E{Ibr5|()raA+Wk}$l+Jv)Y!4<Q$Yq<LJwIx#tc8CQl+Jv)220B(
zD4op=N;7UY;y}5Z41LU?%tCEI@kKlV#W#qSZ3*puo<eUyK5x&;%GN2lZ!Mv6-()E5
zg<MwdTT7_ix0X=3Z!(njO|C)in+&C^T88%O+JqTh)iSg<vwF#8DIFM8p@R&ibdaGk
z9T=3YYPkmJj*UGwY%D7_LD{O7%YtxNMRL8mRqgovRsBp`vw3T4o159v0{zUzz4IBC
z>Hh^ktnz2!M^?AC*2$5V!4HdFN`3}<jr7y4`+)y--NO&BcpREr@5zRjTm6?>cHzaV
zU+VOX-eBg))>*o)Wb3TFR$4;kwIW05T9M1jYo#SrUMnr3@>-FhbgjrW$ZJK0(zPN(
z<+UP1=~|JY@>-FhbgeL`vQ}g$T`MwFUMmdB){0z%yjEDuAUuk4Sx5@1Dz;h}ob95t
zG^8s0e)d~<wV!&cqavg#&*!}bHJ|!cM^#8wko#M<+w>OHea>4@`Kh-$YD22BeEO}9
z>X547h-ulbj{1<Q+CKHRk+P7g^r=BsAs)`6;W*VoC}o>J!`9=W@<NoUCH{gf+L#by
zBjO>r)L$sz?y4=tW#TfyLc5b3q_a~g8OYBOUQ?;gsk|r$Ta8~hH-Mej1Xg1zmr&n6
z&S4>x1xfbVU&D>5rH{g7%a)-fM=y2Yug5NllBSRvhk_Ij=S?-zDl%i{;&ZXfUXMpE
zy<=nd%!>*s^$uH>EIPvXKDB77t@xv<OFDO>DqFHBlZ3q0@fvk<@6mH6#a9G(?BKtE
zA6k#j<5ouOIjM+t=2=7lT59qx>uij&_fNhPovg<Vns(2LwifQx$EoeSz2aPlAbkG&
zuO+F-Z%Cp`l<rDqYKL+I9m*IT<a_wgTit&&z?~kp;2y#nSS6j8sz?>_ko>B6k1BRY
z*mf}q0rLmMNkUb975zaUFfC~uCbL>2rk|ZACODI4q@hQgc1kryGl&aWYdj-e>Re0i
z@|&esq!@LwHirC7PCVI`EzGMgr{gmhzi|2p7WK}M8+&1Vs5`kY()dAW?8<J5rUQmR
ze|X1u&`KCxcYm!b0g@lAzg?w0S;e_0ZJdx2mA-BK-&IxoxvgiS=<Nq_C$>^trB6%e
z;8Fyq1>}YpNZV8H)Br^cRlQY%(H^Bg{h?ML_!eQt?Axd5;Ew)rF%ph}Sc0Dn=ZcJ^
zSqZ&24#?V6w(fdC;mz$kzAKz{`0Uw3$4;JkO_1u-gbnp+T<EHYr`(N4*QEMnuNZZh
zxLy5}c)!Wck##3d(nZ|#dq=uT;Gt48IxGMk=($@H(4j9dL#lG)7f^#ssw2C$One?m
zxRiF{su$1&ciw?GMrgk+^^h6Uql!3HMy#w581WxoBbks}M%umPz7)^(lgz3Jj_Kvs
z&ne9lrq?eNN`N(up?n2%nx3eoqo{XBvKd0iOe}GFLO%Qi@2!T5z5d;I*T1}Z^UK*=
z`vvER#Lo^ISzBTy4Ii#9lvYanaE;qal7m#LV@Xa9i1gn2pOyz!khWJZleRPL7iiP+
z=sNlwbMRl_OtXiwJXUIrH5}<q+~t|lTh3$&St+V9fj<&<pHFyH-yu62R=*m2^B(Yo
z1rn={+4v1<S3Lz7@)gOOGn@8$ir?x#4smOZwD-3Dur$2aqUpJF{1c<jstZ#O<fp#%
z$>YsM`Ri^UTwD`8V`<>%;M7=DhfHzz_-|3St*&l1qm&(7(`a(VY(~Nv<~e9=u;Rp|
z&?)|-7A2HknVa!_ShCxmm<{{G>!J%fkM^G#xuYmyPGAr{;gGaGDRN$Hdm(&z=9DwJ
zk>dmW{dX+Ln(C14TpU|j<>?z84k3e6ZNX@E0VeDoKI-P7$M~?sjXAf;(j4xdb}@Hs
z3JMI<)@Do~o54YMkQ3;^>*&E<P?Sb0A!nPcgtXEDXehWP`t7db7rsTlMY|JnKBy}_
z5fSdYDtYOqpsB%r6TESsYL;Ye-DBd}d3BfaBZ4EmJP#c%pX=JKE4P`_ipI*e0fw@s
zNJCRV3wZWfG!zZ^($j)u1gLwV%W!V>s2VqWySx$Q+rl}HtlGI`ncuYVsjoSz{a-7~
zdOfG)Oy0CT=|hXZRu}fSNwBcoBK>gAS<>Er@3rze*Jb&$XXeGfy<)?c`H6?Mn|~(3
zM=36!H~hi>#O4v~^Kt^zlSR0CC?A(<_YuC!OUL|P27G$~Ukk)HyWk_ZaT|p*>T_Z7
zarB6gCEZqR5IHIPItjJ7>&BYDQ&Tc~Z}F$|7G26&KQpgn%=GzgehbTk$7CHZ^vv_y
zw0^^W@o?now2Zu^HC0Q6Zj)DybU0aCbU1a=p}d+y?!#6WPnZ*JV-q<``)^uNs7?3C
zstGd}ujkEU$Gi4+Da$OZnT_78X!w@Dug;TWH7JgVhk`!YZ@81QOja2&XHyz~AbV!H
zk?_Ta@0A{&>U9yBkOlwQsjF5ksV&?V;P(2|k|Sv)@2d0WiLYIgZk?RrlIFJY?^B7#
zK8ruJbJx0A`STK|Wt^YC;=L3sEIlsJoyGFC<N|W6n-g4CIFycI1p=U56E~>a9y~W8
zX-;s^oFv!KP*>;B&}Z<2E7GGGlpkCrZ7QXqqlP<&g*gu&1q=!?P9ud<_#I^)s@&PG
zW9ckFMb8a{Ywf{|7TuUpbz=d3SKpW)yCp4gQ)=Aiw8V|!@seAlP&!}IzHwU8&R#&&
z$IlRMR(`^#2abIblX!Xm8y`mv9juM@Y}PKzC;;BlEY8nZue-*|={|)A*l@A)Z&aiG
z1vjdyZ_IWH3UV196y%<;b<(7*3GBB}I!ltyP!whp;WR}-8WHR?JS1eeQ!t4JHv29o
z0h<G#B9@X<-NV{96H=b$iubFe6Ru=6S*6tqDfM6Q8$V8zwC!_pFlY8-@a}+W7MR&G
zSLC1OiXJmgrWt$9x%C=1qqFwzvEP>jnyoHLSQi!>byg^i-#E!By<pDnoRa*?W3whF
zIp$OaCx)b^)AhBqp(p<-Y9FMloo=*GaTS>>>T%(LGxo-;%80EE$-6ke@Z6{vuO;#8
zw{BfFv9sUQ9djxZsssHZ4hp5ot7A$wb!`{EB(vnrtnk?IV6Po>7giV4OiwLw@fjOJ
zV>sqC*a{=i{?pvBccwIlx77x7Hw(K6rMB8BX^CWx_774>HUaPVfww43cu8(})l~e<
zB`a{s;qmz&&aQlYOt|-w$qToQstONCbSw}iWJ>|P%u`Rzu6;i*JUGVJbN{ZQS<dFB
z90Z{Me{sVP8!Y*6*!)3M8lo|+ocM2~js@g4?vPK(thdn!2lMG#|9^oaW=(k!N6cE~
zuFnzeR~C<-6Rpui*JvL-k0W+PtW2pAO-MN+7-Bz_k+RLS#t#>rOI;#edfbcjBTk;R
zS*unpot?kU%i&0Q>Cs6g?+T@}wXgpoee!WgS~#&MZ7kmSBH`G=RXeh#FPbn3>9A`~
zPo@-V;zj?T@&ssA)>08)M_%C!c^K&>+$T*oNHcg-q5|ZI0&xYTgq6Xj$jaE_t6By*
zra%UvOf50RSH~5|NR%mL^}M0?RSo;<gaYxA3iw=B=1m3C3tuf`Wlk!PLDCd{0fRuZ
zkY*s(#=7Bez7D6es6<J%HV6dDNKu(s*}15Zly5seDk3#GCTjdlpSc<7b-uoJ=^1l<
zh;MmDZc*;k{PLK!DJg4XVpk_6y{aN)1BS+ws{uE_>3>6wH4cn#`^baziYLDpj5{FC
zKx4!eWlGSt86K!!z$8S-9~24frD&B;^;=&<iaUzm>^L6m>*<(1G=BQt$~y9<cUqcn
zPSWJk$f5SmL*4wBmeBfa(U)4(r@;f;31r~`y=W2(b^K2ee7LN#I?u?lF&mbJZS;&7
zTJAPBJUle+6>r~BbuRgiz5y$^#p9zKeL8gT35hFq_VO7t$Zd$*(4hW~shU9{?o<RN
z;J0Y?Rm4pqS43YYh=69igEJCaoAMoINE=AE^xr=-oLW{;P%xHU9V(R_KAeZIMgrGD
z&<6$pYqmq{p~&Pu5W|#7hJ4}nl2OY?j2V#OI(kg7Z;yz+Ddg+6(gDjTtBjf(Wktih
z{QLL!8|mR{(uIu6UeKva7WOAbiyF|PC(GO<3wq52O<f3!8Q%x{1a?W#oj10Quuq%F
z56h1jkryzgDkNiJpx@$LmuZgI@^U&33iTWv6-q>tE?Id4><7dJ1!cG-Zk!OkBEDzu
z232w&{~>{nBizt8pN2pATzTvjNssF_1vfG#aDzHJkiW8!TZ;5T|Ch0`@53Z{ountr
zD)Ko9Isp))q)bT2cugW+{#f_1Ne31!-yJ*3(`$1{_3EVA8FS;4;?ksjF_|MqhDJq%
zl3xS8x$dE!i8;CBoX2$T800l7IYnAJY;b4mkb!otWVH2wE=K*Wdk(^p9NcN#@PuCg
zeAqn!);Jf^Z1!UtRU5-g!03qeLCR0Q+uJ@p#IHO!dT!YAU4Kaphg>oTk<|Dx<HwCR
z8#}(oz`=3;K@%O8)xLSCI~ilqH8RG1xU)UAy;yT!@^$J0nM-g#Q7*yV<W1ERWG=zo
z1jH4P5_JwB;1UdC%YDYzG6=W?gA78MT6L++B^YER$`rDCz$F^Ls$pM&OE8EJ_bHFt
z74a3g1a|{%^}<)nSQ&5$rRJ~r1q=c%!64RL7H}5u6*wDL#&9m<O@&8*fU_}(D^~`b
zX&H($En|zXY8iy$Od*3%rdHhnWx&}OWF*QIvU(`ay02>3R}^Oo@!>M~TvmqSOd-AS
z)iPFw;w;x(0Gw&-6z4J(K^p2(&Yr)99YTfu)l4#Q5p0K((qvAR3~8kwDMbdE33v;w
z))?;*OG&v@A}yueBeN=2_|-;^g<A|D^}VKi)u-8a&tPDqSanNTb$s8}nOU<|q;H;P
zZG6LD4Sx$~ly>c#wX1$fM(55DS<VpQTwlQSWGd-89{zj|$Tu7XcA`25wCthsm9cvU
zIbN16S5Rz4$5!dZwRtJ*iPckt6j#5XCp0Izuvek8c{4hb=Bz!Mk#n(3&LetkzB%c{
z<@#KohYYF`4lzm%qP9Xh?_QQ9EfC0L2<&+QDR}><P$U07j|laGKEk-3=tqq0#|lNj
zDD_qfhPe68+I-tGml2d_e$I2Yuoe`PTHn$5hP@4cs|}S|rlVc!Qo{={i)WK-RI8JV
zcX}bFtxT-?Us^?)Z@2>LeW#3f$p1Lr&nGW#j02x~?FHj6^muW1o<AQ7p3$>&=-C6L
zNaB%E8>Uq@FPblr7mPv}8-)RS^OO7Pg~)33W~aEgRn%a8HU@6XlsT)R_vPx%TAT==
zqTdTep-@KOL-GQj{}N(RGcNw)3uGi2#|;d}K8REZm2rH5B&ejyc_HqRjI<v<kD7O1
zwnzT05m~(Pg8r!+{=^u(ql`fx*dHP5>yN_!Chpm@!B{8r633Si{hAHNKQAu_svG_T
zye}#%))zU=;$^&FnoIw09rI!xJN1H9%wKxhE<c~SYb*a3N<0K5jw>_N4Y`COUT%*5
zZv?PqQ}x^DOjR!VWjo*cRK4)O^!X9`d;&IVDi!L_TE$-dzx7RK(RUQhhZA15AG?{3
zr2YA&5hvt$!#(aXG7WZuHi^>#R9{yte>UX!z#!XUAp-}746_XyxO7PHz=6R-Y(obP
z3WcV&UE0n4itnJiDl;5+5i@Hw7wZ|-w)5^e$>zNx$*^d5ixIuy46HHn>gOH}0h_#s
z7JuXa3%fx#)+U;iEzR_RV&emOFWS>PCdS({+QxT){71EWY^?jpv157jfj&0To+Dyo
zQQIcmAVh2Hz)oq(A}z%PW@}?z#q`(%7vmk%&S>wv@oq6AboCOKaqhiF^p&QQd4tAy
zjN^{8N=Kk}DCVjV<V&8cx(htbsJGAAu?aa)su5BBze56?Tmk}IoC3(wU?-=504Lag
zSLk554naW<PQk%WE&io6`5rX6%w`Y90T)W}`nYrJ+E3{a>g4b5<lyTI7wtHHZ~Y8^
zJBPqP2Rp<F;A?3bcbWeVIMc6<`x-a<dXecGV;5Q&Uvs9~N&I?p!Qa6yFwo9{ey#ct
zKBX)0oXGmVVqG@Hj_l}(b6Sc;r5Q1~xT3v@L0Q>BbAoenf>-4QZSbBD`>_wj(B98a
zT{v$3sPR*xBl5<LjvYTUF(M^ke0orXUwV9fwjS2Vgd4~I1FYLKJA>X>8RK9Nub7e$
zTqm27fQu8iz%d}e5lzfc+TG}dcJqq%Bn4$>1@YhMdL_9uZ?_2_szo2?+HjBgQPB5(
zK=*rME*Fyb`H}oGMm;^d6<r5htWU;>5g9%{nI0aQ<b97!AD;|&`8P<d%2izxd~x25
zt^#9gLp5(^Z0yWiYt%SfAxP6Sr0$>eKmJJnhtz4LX@XEGrB+X^l#YdxaOt&BGOBWF
zH4NFnX`#4>?`POXr7>`zIm%Rq1`aeb0;2=-l;Dn;A$x6kB-O+0w!#lm5VIc#-U=B(
za3i4=z%`x+-yWPhR61*$J8a*ZBM-hcG{>I!+2-0G5R-`aNa<~Uwd9;FZ6c}kC!1W%
zVYe3I=E59$BO!WJ-ry{_8jQwVLUwy%b~Bw%nsjl9!G9c@zcWVW>GDZ@Qo2D8hsRy1
zlTdY4%IB}*yoLcD6t4G04!I@P@+v7`qwZXiOXup3_;nStt@XLOl+IQAz(9KkKR<r6
zHih3zKJ$054+yZA=jy$NyZk@+`v2CPQ5WiJmmHoqXWmmy=6(~>IRpMD)ilJi<f>{h
z?zpd-{|z^=64japFTu<pfKrW8W{n*`+QDE@kH?)5``nDCPtgthP_+<sFOvBo(c>&H
z;)lfQX@-co^oh8GyGbr^hJw3dyS)O)j|_r9ZP;!JF_iw~0}bY|8?a2e(8LkQzX11b
z7kZSDBa7S$c4m8QzcX#xo$VfZyYij6)4L?;^0BypSvYJMar+L5Af0Lh5|4e-fc}w8
z)V5ct4X#6KGwnjXYaGqH(1q{ooWCp2W5?}j({AtZ$lh7t#+}}cgfhnx18Su|q{DY^
zONXUDW(C9@yG(F>wCYdkZ+?csEXEZ~;fgae?#lf1hP(j+ZcY2x;`rro$l^|L(Q9}c
zgbB?Es|qq0(xzviyL(WNHbV>otHb=3lzX=uY-m-tYSp~nhJ)LBmn|`vmDMZ6OY#k!
z?EXr7=SdNguXl)LR(G%Bc(TSP#VOTsf)80ASM1dtJ+cQ5XA~UZG1fKBkU=Pt(ll&+
zI%J5w(&CT>bLK1<BDd%lRuyaryrM=0ME5ovVi-Jgw81RepfuTMf@7*vijR~|8_co{
z@g{Fara86iFk*5jdD|<b7jUSCpC{TN9&s=#HM@HhK_AuRnzW51|77S<`uIRGrF|kw
z4V6m~nr9V1{7_tEnk;QIh<|jx$RMUvMYv#ak)9}}R>7u@U;`sAnAT(vU;2a^?^P=s
zEV*>`E&5)rg-fS(sBb-XMGfFkstec>Y2-nm@JeI!7HNdyJH&oj5psbHmOe)a5q={i
zDMF0k4s(Yohx>%uD%I!YS;_2fR{;5uK`<*B#8A44Sy=>sIP#?9n@l|KWzPbA#?1Hu
zGot|S)$F}Jp3ksn1NL0q&<+#}p#F6BoQUV4?70nn)=>fP5deq!lwRW*mON15D(Av~
z&mCZ^7#VA?;*76L2^cVe<<$i?UiOCI4W1{nSI8gDxd=H;-}njrm2c1|7vVMejS;@V
z->3|&`f+vmLW@>=s>0D~Fy{cyvP;)$umGESaWTbDsR1S7_ig!%FVmzo!l;H0x-ZlC
zyV9D{QhbItk-{kSiAGVWSa1mfOoUO|Zv0(sw~ZU6-#60J;#avk^p|4NT@?;`TjO~F
zd+tG>n`6q_uR||rc?<UbD&oez=MvfX;DY3&w37Qr^?>;^bbQQ^iG}2wvCghB?M(KS
zj4?IObR9j~wcFr<(ppm7$qCrrl~(fav+vN)MtMs9vc)557vtg@-OhMFD(XhIk9Bs7
zXlq0tTGVRE+8Acj)CT`4f5BiNaHBqSW+9<bA=o;*c2zfj<VduuM|d0KJ>}u1-N{Zg
z7}-W{P>7+`nhYCA(xpY6oLD1`Pn7Pr_z%!ZU!|4iG-%lqLsT5o!DLTKm~l6<3q4V~
zWv*W58XoRyHh6%v1_T=3v)L1!jc@Uva{sWtwCo9YEy_*CHr&*_1-A6dlX_n^$Bul9
z5n>qNOZ7I_F=Jf24PryoaX9VS3Ds==q0j@nJL1o*4GDZz*hbht+~hDSO@!f^)V*PF
zTczQz6ZT|HD%oI<JVP0ce?C<YK?ogO*u3)3c<FW+Ns~5&mC!#RqJ(i{$U1u|EtOT$
z_$TejLg3Sn;iJQb?-wW?CJ}+spRWyKjSrS4v?!^x$E}zC2%;ib5bEDDnAJZBvPmNU
zAs>w_S8UZF4Kn+WOtXB`;ysHKmi&iAD+%<6)++KJa>rO%$m<R~*WKcTg<OKJ)Dx=j
zL9J4(3_DePvuAty4D7!L_Vm3&i)TmJ4G8(9!3I8>5aeOU2-EGE23KLbp}VucrLRl$
z@UeDwV~0n(_*xR@+zkuI+KgFwq@wc3sxg|e3*kT|<O*u|QK^B(J2@+%v)GZjOj?68
z&XdGA%QfVo2F;jNM=C3htQ=#*YT&fz!1KHq{{j!eT}t2}R;KaX?Z;KzZOY0hXXGE5
zqm;8!sy!M;83*O(kvSi{l@Hm<2g}9r0ntX0&-ep;XO_(cz3GeXyl)@;i^@HD8l44y
zBJZDWK4wa6z@*6VVMD#$f`)5n@e@acxCMKR$_Nj&^_tB6<6@gP#v|OrLStD%dTA{1
zXF6t=cA|8P>t(P)i6jQfeVF(_!&#8W18G5UmIpbO9MPx~0*sxEJvka#EB($TiaUyt
zk(e79=kDD-znDHJtWWObHnX5$rh7+1lGgDHatu~X^i7fa7nF5j0)%a9VA{~CDP*@@
zxT|HVrE|C)dAq#K=24{KAr&XC#hz-xc4p8;w}aC-TyZUbKzosQ{whb=#|=Fmoj1fL
z$ZHQ^H-8V`P-$NSYU;`RAdoTr(U<odcl>sGa=*BqD;_D1=FYX0{lz}AT=0AyfIF{9
z4V?B0S<D~g<FOu$fgDDaWj4^W5_s|debP=(L89ZlubGa7?3e#Qw`~hcfrkG<>xDfD
zdRIedx}QD1({PVu%Wvvw)=89*isTZH;Hy#zk1e2^a{PF<BFz{@XrLth(Zhl&h+wYG
zY!^JIeUlH;zWI2fz72z))$D0}aj~?lq(s`HXWFpU`<vo;71(D8YgHQIyi`DO6a4CL
z;9p*h!yB53oa!Qu;E8$~EG#0hX+Vj6o)#@$aq{zXviJArmpo6MzM<2Zg;h;@Y;#>j
zU$5r1pOOmxqehz1%vFqHEys@ksQw*cWc@qRK)nIv6_TXu4<g3!TlBRrtSlP-rv?``
zLO*F?y#alqFoyN=n}3($4NBC9%O&V2s|+5+sN(o|fzBdB%p)4p4U4W0t7q#wy6p+0
z+e{JS<DWcH8$5X;T0LbT(*E!iNY^1rdP-LkCq4yI38G5%GpP4a!Pcq?Q}xpbDRwJR
z@|@sRe@-lm8tEM!?L9Ke#%JIlUk&|B*ZvfxEpE^NAC1Om05x8Lu5ZA4CDZk1N>8kf
zpDBHjhZH$9$oG)9HT*4Q)!btAR6qI|V*x)r=3OVoClE!_Z0@zN9<{V4k-wE`_wNlo
zvyyOL@1beR)igIY2`!$%v4%HH=08#|WAvhYj`}5)87Jju@k=L>E01yvS3b&F=cIn<
zAQ5@VVQv3Ar`eaVT<L#H>(H9f?}h$<oqokn{*miARa7Kem6VWtNXb}-rJzUt|ATc%
z9x4P8eg7}6LqQ9h9#12vX-<|PdMAd$6$E{*fW$$_QKN6Kdnw*QmH}1Ah=d47#RO*-
zSvY@;ZIcmB^n6}ln2ByoR;X8SdKheUm2nXhgU6?LbgM1>P2WoQavVp{(zyD`=qX*{
z{cz73UEL7upozW=ALJ+3U)jyis~@%Vc_`}^5?}oiT(F^`{&!gBj9{75@W0jLOo0lP
znoqH={{2FAji0XG%i!T;F#Y%647T!qW1NkvKcuqYR|@wb>B4wp4EYon>{vVS$n^wy
z7nUt+XZ1LKg!Vn|nm9zvUEP^H94NIVQI*ms(yz6|ReU_0{4Jc5$OiGzQe3C^;Tz(E
z*OxDo($mOkX>=@Ew{+Q&h6b&=!I@tHI~&Ds4QOV_-*14&2rJ4%vCU8!03Jx5w=$*n
zzNc+|&_f{-sGqjqyy3~M7h_tr19IW@CuKw6gO)a)=Zgj3Jlg=oaF%U06z8B9<4j)4
zuJeW`$b7W*Zv0tq#=4>Y@uh4(&+i*||79&h7#+nbqey5k8XZVzu(nv4zSNNXB3pN!
zk8FLNV*C-b3qffw>;!b(^<~nL1DHR4^yS9Z?8IEk_k+H-#*cVRPjGD2U@Uk@*CefU
zjeGOChGP}OhQozfdf4=Zu!^HJFT}=w#nhEMPurXp<6ndgxy&zlvh5MK>AA=V7$w=K
zK}X436-H;|vn=OUIDKPm^$gyOn=$oy^Bd2?kMxPvE@2Z=INgFX_sc6CZyIY^yl`QA
z#X<7?@g?7a!9B&BEy9+S<m4@mw9=RLg!&KBvrj;C+D~}~JhPv&hd_5HnRMFv;M+rU
z?WJ?}xeoi@v^n^ee@+nb^UDo7AeIqt8|iI>Gul)>7e9v5y)3Dm%%eY9#3YCOCYjQo
z9O)i0iN*1+Jy~#BaNt|8MgbQK8jr8Wv~PuF(LunEHQq_DtJgpNEuj@x#Ys?EN@LXg
ztCahw6;DMrJQp}xwoqEbZ-d>Lwg@AI8dsP;zfH9Pvvngwy2eJabw43(-JBkpy#S9k
zTMVMszMnO;;^zKI95?M(SjD$yyFCZHZw^T}{`x4!EAG>J;{Kt&y&Swh9lZY&#{2Ih
zk1^Yu<_`k}mjM0<aCTmfr<Ren$hT}}FANkrhH}@mDlvn)W7N@N?gmMdLe6v5PyDHG
z26Ya?w_hlA9%FUh-$3`f?*T8A;cS1?@b{Jd?mm;#i3rJy`vc<!zl|}@hT{JiZbPlO
zTwGSeZD1GR{aycxIvLWcu4JG9Wm}!>r8<We$S3fp-82}iSaxYt>J>nKWDqE|+)pY3
z3C%?~FWW+?9%&+cMO8%z3WRTZLv06y=ii~PaDxH9;Sb`QKq05m;%v}xPxZ5K9`bp6
z+*L=%-J0c^e5P!!c0w>UesE2v{O|iu%jlBk^y$<51pFeci0RHy5SK#QERg@jXf@o!
z8K(0Jr=wChy9tW6(yXNxa=6KnBS-iW{1U&^6ScVq&8z2cYMZ-vIrR7$9>CTt0?j;t
zxC7B^q*P57wX_60p987nHfawjUx5{(Ndjr!rYwP=O&m-v&~4ODwMv|{pUX1G^PP{d
zGGa<F%0AU@^IfdfcAITqFWn}O(VQ$qHfb<)ezKW}m1ja_Jo2}6hbkhvp%QOXO$LmY
zeR=*FUkYE<_ZPm>@~wWUN<y1@U$XXSyJ(Ypm*d3V+z+jOD%5*GYl>}blluGh@a6G+
zy$An^)`7b(BYpYl_Mfzo@i(A#>0v?VF+ESAt55eY>eR27-^6LPHvBh_!f4ga>esCi
znL@gD1nuq??KHK4x@n)`n`Cv?x21JHKKV?f!AY8Y9x^AHAp1PzH|J$<h{9u(H_T6>
zyrFi?G(q`>7yTN{xzp<JA+z*S!CQ(PBxXj;A_<Qp&08tkBbnWme+K+S3Avp=cGJ=t
zw}6=wx~F$9I9L+7Vfk$LfWjC&rEe*G&bvF^3=Wyqd;7=x4;?eK|Ipkq{t5mgl7?A3
zRNy+q)}Ua!Ei`U9TY#)@Ipa_uyKR)xl>bOOC!8OMUk}2yU-HxNi*u!f<<pw89I;5A
z4~9=^<a9p3J=jt&CyIVg|CM$0`lZ@v*fgUJA2t71*4GW1x2KyWl>65-7{X&$ME4%G
z>`dUJ1Eo^3AW(YqCo_IlDnGlMb|T$n5>e6?B?E<-y4?WKH?h-b`BBjBe&uPYqtZ5B
z1#6T6|1<Nv{~~?Hf2@vx)d_x0=H{mHEKRW`hA_v{Tmmqy8*w*1%U%8b^W7J0-m+js
zvEN9SDV{UVoQd97rH+uUscmgz?fcyL=Emjzj>&dI+DJEuwFxo5we;@t&IG4<INaX4
zQpgnEg&Z~%J1GvFI3P#8vzdI7$PoE$SSX9eP$M5van`7Hm>|k#?+J&QL4h;uPNW{~
zJ77@X<Ef|YrUwSka5#~AqEG++y^nLoa~3X~Gk5-c;oXzHaX|gV#N&3=Az@XHN7CM~
z>f6`q#Kboot3pF&+MTT4zGUgPZA+JI2ZvIvldcIGgALHi2dWTjtD}L&Nbv^H?z1CK
zW-M7eC4;dMJI^EMjOAW0Kj9YZHg`<I{(`7ol`|J!@{Sp?b<*~8p_``7{3%WRKkU5;
zcvQu<FTShx?(QICNJ7F)5|R!>LLd`_5CViG$P^$m5JCuZ7!tx9AdE84A}C-~R6GY1
zQ9*G6#R<`?I2@016cuMK$D@cqdhfj7sy%df5<K2_@BjV3_kEAByLazhwQAL>RcozU
zwW=y_VctUfe(gfjalH$Ar#04s+E00CX~9EfzS<Y8ZKHW9V+R!Y@X-lRZ_L;-r7?@F
z`KNj*^6PP{_`&8=i-wl=kYwhA6>AR*UD^qh154nvNcj6eJ0lFuYjsnAuN;g#tj!G8
zT0M2wg4I>=W%>O|I>~{Bee+X>rROFju}x=rj_l*Q;fmnH4jn%JK;FLbk-bV{`>v?m
zuyECC^)1Y=W}_e}16n*3UJLTGTc=0RO0oc4`qo}z2{w0P43mH>zBB)^c=w>835)t}
zI@zA{)>>0`E=U{Q`jPz;cSdyGG4b4Qot_W6M}8{ka5o0ePuIwj&Yg~)pYr;&cU~G8
zzNcMx*67Bx*QfjnA0v}~)+Wj?VZUJVFj}Ptpy*IvY<40Jtnh`>6QOJWNS%<}qi0c<
z{Nc>j;<t9mGe$%X&h7HpV`*z_9|pE74)h&0^5_iLj`F=zhKA*ZWUN{&@0{`Q=*o2B
zfF01NV?p1hm`g&Jz!~O#w62+El0r#Lh|V2b#h1;jUJ^YoWbMUAwDnzhWjFP%D89VA
zt+1v|%*y!zOO6+@szTOvVqgh#P6SA4Cq9Y+SGRw#77e4^UW#b5P~R{^cQ~mia5oqh
z%P@OfVFBx}?tMu6lJ8Mx;8^LGraV-Ly+Pi#6`uRs;j>wq{Oi=I?2b8e<|OP`S8-P&
zN?C+b!cmGpVg)<UuC0DJcwL7WoDdU8T&y@ujU9}M#8CG3X9ueX>|3M#GUnAEa?AVF
zO20iPk99pPkJY%XF?;loR>u3SSh07)vi@^3nhal_`}bFyv-_q@VO__S*wqQ@17Hp&
zq4S8{4^X04>tHO8NKRw5A+-;=ruma4#|IR|EgfFIfF04Yx{u&To4cMLmD>EV$8xJ}
zWkd7Qk`vc1tDUO-G-gNld9&JNPU*u!rm_~3tH%>36r(KKVTE)r!x_Z%hkh9xM}INs
z4>3#`q0PIcU1FWIOY61EtetlGnznE!o9x4;Qk2EC6H~O0Shp!Brtv4#`Fs<#`waB_
z4(R!j==X7ej951JnhQZe+HY8&H8FN~Ah}3ewKHEQq(G>j(xL2={Tu$7o0WEIRo=mp
z_>Ie|-YlQ~Qq_jHvU7$W%^$gLbm{{eN>0qvcCEnvI&8tb<CAm76b%}cR9TcUd1U9X
z2|MN#-CHnXNMUMH&*EVz<A+DI%RQhxpj9(A;pRr|C>w)0*LBhad5W?a7(0L^@eS(E
zUSyAHn@f4GiVB`y^DV*?Fo$}Og(&{Y0BC^g9-|L4nhzR){N^l-b(1&BCz0wwx1Qh~
z)x&&;&-a*b{X51+ObW*uu8sN%FN!TMq*!8Offu;$rP@VGKievSlQ_YctQ5gzz>WUo
zE2Ffjqk8eCT14sTQrjy3TX*>LM@(b|!m@x8WkI16dMp^+MfqpTeO#`3xlw+BG(p*E
z&_plapbM2PX!IigNG(FqY!d&7|Af)WRw|S(Ri<$j#~jNWN)+bD#hKpNFDBSYZH;ep
za|5~!4__ib?HV><-kAk_sQOIymaMET*>tzy%sl?^jQE9T%4g*7%+B7KKcoB=KxN^e
z?j}%I7ToFum6WOulNItv^@HBoyT`MR+Lr~378TI<_}$qAvT#@Sj8z2%t7c^HMr$pp
zZ7at}i@3wSgd)@EiR>rGC|Ki2Pp}$KmcEr<lg}wZMr*idj$nPr=hSxmvVJK+?r8yV
z%LC{H529D4OF<?d0K!Y4Y?MV=i{FT@JlJSXH`ke*hfyq@qzaz;z~nse&~dGhFL%5r
zVU~;0i7W2z6^D)0`^?T9oRt*WYiO6`G{<|tuNgfnA+Yu2R?UZ|_$)*Q-UzWcn*Dwe
zbLy7=z#}$Z4oaKYKQgwU``G09`|S7ox1Sv7n|<f?b&XbRMC=~#h}g7hABsz+a`lKf
zl(^%*iq!RGsXpKT-my4+M(Y9d3VhxHmf!1~FWQS_7tMsD5OV??`TFJc9niWQGu?CM
zE>CRk^Zl*!W%I@ScFkp_u{E=83nQ%2kK_js7e{5ohYL`CI3z=UVBUdpmRhzRA$A!P
zZzMQQh%h=tT-E7!yd%7)KLdBfi`lRjRd4|BSk5MV)#%*y$L10IPJsV2=D+ngG>TK}
z#NN{$P!IbYpa1)-)W4V2ZjxhaHSot`gJYJ-tKsv!u0J7;@scTosnL7L_BSL;0(;Ee
zZA6!`Ar51aveW9Xx?SzjXU)DRpA~)AqXX*wU!G8}8_H6j+D1wgcIKyE*dBF;tPwNU
z_4*<8P<D)l60XAy^~NFcUtBrAW4I9HI_1?XwOP{ueHF@C=v|K3nF=bz)WCF$wfN{Y
zdO6>s81)%GRPXe3hfY7lDrCVc?yi4Et(r!q(5Q8xG4@1lC_b3rA$-a%jx0BHy<XBA
zG}dV68YchNP(fm3pr+555$jbDX|6sf#T<_tr7$l^BlMIZYeXroh}%jrvz$7jOp(5=
zU4(hdIx?OSb&N}w(R`{~!szX+Y<9~TlCu`Jt#yht+`a<On(rEFS{+3{uT~`q?xwZ@
zx#1QSki%|IV?S>K>)Oq&m&k+Q!R6FPJhfAT+@~&k>l%h3J{`v^;ruxvhtNO}D`<IN
zzMHMnmMqg&u$4*=ZT=i>F}rIH+k|4ITRcWSfpAbnNfVXj6KX1dqOuZZ&mAmCegt}j
zU<J}x38N%uLZCc4>&A_&Terlgd@|R#yOmo57`LPa@(JvNLXqSY#7gbv5rg6Bx2KX=
zA;pz@T>Ma3nfA)yp0uW*=U`rgyVO+N5$(VAh@sFS6TXJFF{dI?^0(4p=%#(p**@4I
zPf(J0RWOk}>rQpUQ1&uwWhkh9;WZb!@}JEaa3(!34U=ymxA44@^M}=NRk(W9NM)ZE
zp*wlm8vyK?7VSlmck;;rY9#+)sM<`~H$_X<dXHy+-w)isLnk_?GzL^ewK68W$b?cb
ze?%=v<m=QWI6!Q$(pWolw|1eFHD&>>%ZnjQx%4%1zG3Aw;`^xLTN06JU7}jRPg2e@
z?OV+?k44?B9z{lumXOdAPP86zlbG?+d4bx5?_=$?uUBbbv-a$MHC1tHpUx%QZ!QaW
zja;yR%Kd<4%Kwn>hm3?<jp;rYF+kB>E;JbOYw#(1qfgG@r1m{hf{I)8pO|X>@a<nb
z(E3sL4R?lg4@ruM>6qRtq+3W*mzYi&z1X6V$dJCcP45}pJ-Ba|*iPw$id4cbD8mq~
zh91~3UTnV+k{0d`SsPUBoZ5EyBDn=crKK5>YLheKV`d{H7rlUWL@zj`>o}s5d}68k
zGCzc6Nd9Gbe^}|D_}DiSSCBIv|M^z7WQX>J_W5SE#J*YG$^NA^=d(fd+UwHK%CoWt
z-J<V$Cg$69Qv~BX*mtIdpUf?90|l+)2blsfy>OC`ym{8@q`p;CX5HN<Bxl~pS!??C
zt(rW0N59TF<?27~96oG#a^AoWLCNEIbaMWH4nh6KrQI`B)}}o<XYR2no41ar8aD0t
zY`WbtylPm*+AW*c<*ZBVxhP-GS(6^UB%ih1vKE+1y|r3pku4BDM&Tc^p=s%SE%LC@
zIX+HPO&R7XG>e1bkQ%I)NZ`41U)z(Fwdb|DWv|os>vQLxoIUl#qJ<}>PCdbTm5*tq
z{n#S6EZ2{<YUzh$Wo3GD@jZ9`YYP{ioj?EV!iBHRS55_MYPxH0toF|k=G#}Ei9JW8
zYH2pk9@E#Y81z2MCau=q*3PYBlW>AumORk)l6-{5AZFTh<z3i>>r*fT0tk--0#_yp
z8;j~g29{VXOa_uKYRU8R@0~xe0>(A8ZN<R(_bQX{k?*a3DG8PFxYP&GqF`;lhcfmX
zWx(1L_F_BYL|U&Q99ebkPA}up(*$+fkvCs)o1*sQjj*|~Gpn#~xt%AA(NO|UN?8nh
zm!xq(C{7B)K0FxgDvMvL()v}Mu@#oxT;*%uCO(K`|BeQJ59vt)!i+&!0?8XlsQ1)|
zut2Yvk5?P93tE(|aDirDs7VW#@)#bs6gd&`U>j#YO}i&h&QJsbGut@rZf(FO?Ww6O
zf<0qDJ@4jI@>ZwolD!)eNyZtXB7T&PW&<G%$C_Q73`@r@wStcz!B>8wTdT!>9c|Hv
zN9BzVWAe&nt+VoG?v2J6smruS;##$t+j(SRR6_LdrrKjIZ7-@P<~)=)bD)f2D3{*C
z{P7iAE?P7Ug?8Xr%_4f0)t|zpFq{ahI}+m);=;pO_K}(gN0dCXWZKEHsYj+1?}?JN
z4>gV5F=N~A1-Vt}ENz=DH}<Zx+7D;<|8w1xGwWBKnK>c#6Be1FUOlwy-e+>3+Zj;|
zatIrn#;WKukfou1d`CI+`TpAJ#wqkp`$fq%Y=6Xk=*xC|WB87l=e8w1^ws=w{=z@B
z*RC%x*6G;0>o&ELjagD^9b!jg;*;+vtN04@n2*iFdYoHVZ2q2;l;A_tb_XB+-5f!I
zSH^0;2<+08Pd%Svl(8LUq${79Pa)qzYX70EwXHNXbT{&_ldWM94xS;;d#77^2|9z%
zO(&rEt#m4=sm+y>k|&NIm|R$>zF^!WPgp&8@ahSL>(bKJ@e+?8wv}nO9!z6biswx)
zpIT8dkzboKdq(-x`4vU8$L$(3de4Ljdq!vPDo~q@+m$u??g<m_9-Xs`*4sqXexs}r
z@k}(c*8_MDm2!35NpO;z#B^){XRrn&Ed=c!y+b{3n@<_sv*w!^*1K<VtAOOBe*OIY
z`}Hfm`|i@xd+y;o`}GfOmC`S<U#nLA5~tsN_w-r!-mBf&FF7C}xnELJKtNK`jNQ9u
z%-Xk437f7BpYdMo=*VuleQ<Phx6$!=JC;;#+g7=3YhHs-<qk1OoYN<P6369jTUxnw
zYvr;n7-LmvpUO$n3m8$s|0YJ|71-$~NqJg-BG&2{tJdIAfyh9Q=>r*J`5O`-0OSV!
z4XbDzGo&Fic-k~x=4}Im+g1VULU=LKmCF`id23K%cf!7Vc`58UL?o;Q{w=g?29N&L
z%%$_Q{3iNNx^l$*lr^XEF-ElM%2D@I58i*={Zu-Qh;Hhc>B<wHPX+H6;OQ~*Dd^$p
zX%Nx#)c1Z99v`=Z0rJfB)4(_D*C56EG(@brq@qz%5$L}Hea%55O=<z0O?aEprf3DW
zf5z97)qV|upEj*Jj#b34TCWBX?eT`u9uJsx^?G{Md<r~P0i%TtVASAf(Tx#X+R#|W
z!pM6x!_hOVmB~sAQEx%#Fwl9Z(njl2D(|nTXyLjhHzBRM3Vz04;Q2{c{?o#d<J3Ne
z2$K7Sgc753zhEaE*%-#)Vt<*#7R=%5qrUt{S1n(u-rzp9zK$({T1ok>2j_5K+QZ}^
zYY(qv5_F|lCGI%_x$S*`5pc@p*S+CQ&1i4l!@N`9H}B9HZtoiu<1f@5tMbWrsv%Hk
zuZIVfd|7mnE#}Fa(DA}Af}HfNb0|iAb=)JMpydSv*Ggve2k5^@nJiPE_!a00iwLnB
z*IA14tTF=9;k;Pmwn-YM^l*=KU>t0c54pz4@h*&>w6J}ea~V6#4x^NnNEwEd>59v2
zCp_IF6)J78*37QFNb7BDe?(cY+qz(#6>MY&b~kYXf0(qiSuCGDT-(s5PheH+-fgsv
z!);yEf8?_q|32Mn)+&{eef+h*=A#A^q~ppc<tRZ&a5B5yiLg$biWnZBa57JBnmRgB
zE4RzO-IulFyvgxgxm^=32Q2UI`ozJ`cANlGFldETWxjHW#yk|JgE<xO3=XJcd@66I
z{!+q*`1fe<+9uBl3JMI=PAZqw<LdD&HltNk%Ze5)+q7Xb2+9O$GVmxOJp51(I$0FE
zWXIzOz>{**CY-lh*4_0_JKLH#I@R^7Qsnw7X+lT#s@?TT_vHaHZz5y;I(8m8uDq^X
z6x6_;&^W=67COO~@5}cF_G!Zw4o5W4&f#o9o8GOJi#2Py_4Q-f`RYG-XufvDzfTmx
zWW6x~02~KvbA-tWSbXb{Z|1pv1xeUeyX&7iNtB{o*JYh1a)e#$SDz+;fq5*5A>e@~
z;zLTblYxOjLGm2ew)Q>z*^m<T7v)lxwx>;-mMtn;MzzxRWbqN+Efa=H&>#S@xAt%x
zY=6`Bt4<Qx*RGX=By!W^atBaEOLE8&Iz?QUL6MD8l`UL3B>Vy}IH+xOGz_e$nG$T_
zT676(?Axn-&2_@auSAPrpH0?&2n-4gQtW88Anl)%`A+p}Hj6~75sn{Z^On42w)O#G
zEm|=W^F_5kYL6(d+n*BSM1vo_+My0>)@>A>Q|(Ww|0n>bCU(1-*)7qJ<rGj86=;9;
zON_K1k0Q>cwPXoW3HYu`O1Jr%Jp}ruFP<*OQ?V-@?U&z(lMIv9%1ZTPZdY$;GuR%S
z4RBe0QX8+0XVb+jL1OJ$Wr>oHnL(^R@UxkNFidQjwpTFoWgqiqBUX;U|9kJ9`|Rx5
z&(2kUVWsK~_8=R%dc??;!zyPzS624yEcUtU6nh95CZcx^vE2&{q17x&F%?f2;AwLk
z*23BfPm|R^<qyHg!6rrk!X=FISR&uP0=xYBpp5-YezoQ!_BR4ea(pUs&^ZZ!ZPuRT
zL+!sDb8K_)G^kOLgi%=vptFC~{>$eap4!A5ujmwxIC6bs!C3(9cu$MTtx;%z&-`p{
zx|TnNWeqI2|NVuF-oL*9--~wCzA|gpE44dzYOl<a?-;ACVap1%4}z~P)ZdGA%gh9`
zUeR`5qnb&z-z&%LRj|7KNJi1TVJIeC;(W148jf$+?0Gy-dpsdhdm%DGdmQ(xK4zI8
z+pE+gu1=8&36b(8_Mx_}=9H3#c&uYm8=It@)1gtz5+ux$2PeS(fmfR!;aNBUVj5-S
zdka*~HMT@Kryf=hv!aAZ77>}S?sM(*=WMx-TPj-MqJbMMSVAh8;Mt%>;Bnx$A`{qX
zk$W_9->Dt@m@Q|^Kek;=h;(&QQJ8!wG9j<#l(vp70Yq>7-~l3-RAJvnm^Skb<_h0S
zsHxM`BkUGs;+o2pb6Pd({|V|}&GXoDZI!m_bB2~<Qdj<{vQGI=;Ul3jVIa>_FAQg2
zvM-18N7;YoYm?Z4eE2_};V&sQw(j712T4B)R13pJXI$X_P<C7!A@8Qo=iMF@TZh@R
zXIHwtD9~nBTY0yXcN2Lw4dE!&npC6_xtoUG#PcQ@@_En-<1sg-m?x3ENrO&gX*`=m
zCqRg~eA0Y4r5SzQ&*KsbQ2j#vV)zS{l{d`)x)%IdYmvXtHBnhya~EB-!v^{k^v9FX
zAD^Q4grJXpDm)H~)VKC}qcE(4AdiicPGA0aJ{!WqTNBFFYrcD8(xfN8TeId{`u=v!
zmhU%g_<qZl?>27yZp8BC@Pk>J^7k)%I_jG84oYZdpzq}yzuRJERN`y4E9>g(Qb>Wm
zQfFX%8ubXbqyXt@{b@Zt?FekBbmeq?*l2AN3Ovnb9qVZnPk=WLn@Wy1Gc=zg%diAI
zN-*^vA7N-h!!q!yO&6;+vV{k6JR$~Y-iZEy6V?&+=T0YVA!lF5^&5B4Wxpt@xTSSs
z&y0EVGI}Pq-h%KiSB0D+D!H$HLfeYv%PZRU#=)kFbidTu=M~UThwTKjk{<_`H}tST
zhuz*BXF-4ff$-A(?dEa&h?~c4-RhO)8eigBG&DX+7vZ%0d_4rRvi7y>t)r9BejjWB
zFB){WXh6D7Z&9%>wLaJNt*kWaEle;sV=28{RZ6pZ&FfW8lvNITfK`qI_B~pj$|Co6
z0~D{U`~#Y!;5Gd5+tA%zA7Aon!PrTlKNLdKAPy(BEm<&dGC6S$_=B_FdH0=nXYn`G
z=?8H0&H-(!Rol=G?b~+<)lTxivRUuo->ixQ^mTx*H!2<8p=CwO4&g98z2uOmM)6i9
z-a3S&o_bh0rhNz=A?XDRq$U%7t`8YJP5L{DMi>+gB;H|coM*m48joD8yQ@-OW4=Qw
zk9Rz;CKKA-fQVEc@8mVNIzx9E?~uylopS8P4eX+1yhAFFceabapZAG1-yxO9JBQEp
zYP37@@jT^N-FGG~wu|-Oxoa-%(ewF%MfTv{ga{4X-3^g8M)7+0ipRTGv}DhBJkR|O
z`tPvUyQ?1WuBt6P-^pv-@8mVEO3!yn8}~aJ5Ss4@zIW~1<-uX<@h%kYYF1CDCaFo2
zngqveJsqa|opQ|mjwoHfekP@j@@zVfVNMZgYcd@^HJLS;dOH~@;H)Jsq(4eS77J31
zt{EYGF$N!?#ug;PL8Pe*sa}ff)jAA<Tc0ulQ?7Hops0ZfVWZdEtSfrRXytJazVj%l
zCX;AjloTzMP^}Q-O7d5atE2c^ePvHK_#tDUh5KU9&q0W487hsG#=t8%79+?>(o|`t
zG)I~zEtHl@E2Pypt!k6BP1+^hBi#q7{*d%2P7rxQdJ6keJTJWrnqghK*cBMxvT<*I
zoA1ou=C}FW{Qf`j+$@8>J@1ucC9cX6*1e>pgk94=v?ls3e*wSsAGlRhOa8|ml<1(0
zBJJgS_WXW+5B{<4d-iA_y1#jz`?q>ne{Mdb-}0J0d;U*7C)~`kJn@tyjFPNg{r@eu
zy(+ySy^RCkKal<=eI|W@S?2GgpQL|Eze>MLH4+?fFk>5ITQNV3&D*k$u-l!i8|#6v
z#{{Um{%jCShb2FfjlpqaW7z~YiA}|>5p&o)wva7lEB^22=7j&(8%5{+FZEaN-uUMK
zlLvoFuG~%@K#cugEX;7at0BwcFrU&7D}Dz{L$K<1l!U!Ur18>3X|gmOdp4Fy^QFbo
zGWeL+N*kmth>gEn+9w@AtmwniW71LSN$DBsIq5~|jF2e-Sg3m2D>N<?y|&Kf>xT>p
zxMW;5k8kGjo9^lT|MWA@G6@Ih$lLyVv!q1-&$f#@`ce-=v*|y1o_P8{^#ny3fN|&W
zA)VRqzeYtjit&cMVt(+e`kj_c5C3<6x#7VI6SEjfZ{d|QVozufD)IvIyt%L5ot1U>
zt8?d`y?e}<yU!}UJ$}rkm*hQjLkI<*)PYg={JMHD=T1K|b?TAn(~eG^dbB?L>ZwPk
zSqc7sE2~`t_3|}*!}Tk4yQ-)@dOf{qX2FI_+$6k3G#&V+pSHI$TEL)|PN-c!@3Ny+
zL_-*ax(#fKebU=hfDT49PJO&VXMW>$jetu1Ti7hv2n47Hn_V$0U%!df%c6lnV&Wu>
zEqaSMJz92XFrmqOHE)`=MHELKJaZLBT9fD8<+_XU#@dKmqX;W+inz0SP?}Y`;h04e
zp=m_#=K5g$<VZbk%Yy0~OyOjD1TX?}9KrQUIISi5&A4wIC_8ffZ?o<XVx8-k_7^QV
zXZ;B9O~HLL{e!AYy-N;isANBy*`XsFaULpmbZe-oJ82h#?w^%8_P&{Z>}>Y>kqtGv
zx_ef2Apd#f`fQ*%^S-gf0RO($MpQ?jFDCxrLC0F4x-x_~c*Ob~_Dai{_pu=D^U{O$
zYfUsBci#*hIW1PZ7+C!$$`&V>ia?ZvU_amB23Wk4Ma!02ay9E5R9(O9zW}cS6GuYN
ztK?7#`>_2x8RszPibNDl!z3_8-n68WB>W${vTsR!T)d=THInr!>AP|)zTum6^B>`a
zIJN&GRuBYwv=%s|!A5ET!+xzpLt6)RZh%j3i)nS|AU%^O=aQe2bOn3^bz_7)Xs5?G
zRn)x6+pT!Nvhw{E%imwN?EU4V*N?{kG3!T-T2JSyE4?%E&`MBkz8ixZ(f4^fAt##9
zYCpuBn1P6)zBuoxRX{LeOq2=(L?&;CT<T$htX|fO<L!RNGXiF4I^GWRL;@Bjt<1=$
zP%|7(0lt|!t_CW+oqE}<fxvnUNjd{7nhPW8<wncWy9P<CExjxB$~0Z;8hRVJ5Cl`2
zS?DMgrlPLan~}Hk5NmjZOoW3aZE61@AVE4Gyswu@2pFZo=c7~c4o`e_wxMJ?>k1uV
zNM>M0X$dph96C<N2q|<-2rYpFNqjx6AJI}wKwiX4LYKEQaGC32r@%|uNTxjMW%V*O
z6H(V}8|VY_6-kX8W(wX;dz~e+-eyYHTYHn9AjckPkTznj6)_);u=<k);)-GXA+}K4
zuwUE8%Cdd5cUX)Mi@~bGG3rSEI6pQ~J*Ga*hhbt2np@5hB??ZKbJW*$hBUDSUci>P
zoVY^g0~1s6&hU}Z-Y}Fx-mCiX8+y9xi<qx<mtdyEvGP<7yZ6>Dy{u}LLPfccD@Zwo
zz#j6Yi3l*|lzO#VZc=^oi2WM9=cURi$9sm~4=MOz-clXmc&{d@+E#*0wbxxi{ICIm
zoceM6aJ70>ZgR7UJ*%4D^HP4;f`Ca(ei(adwsgFAYk#$J)_@?&peR|7n0HaqRluUe
zC=M89dFDWwR6Bt+_GBi!sa!psfNS8c(*UzCI`%b520#L4PyqOusZa-_WTVu^&ziL~
zQ|U4JqMiimT8568sH>nRwI0e-PGJv;Ce_t?DS`^3U8#g>$_LOMutog^>^L1)QFp!m
z*D`ERS65dnoiRrYipv=S7k3Wu3ZVK_Yg<`4yCi@Mq&&)yGeEIMC=e;>5ApID0I6-&
z+e(nv#0EjyMoM6UKclXo?~xm&0uzdvH}IgDRIy%xVCv8X{)CM@x*D}$<-GLf5#+O#
z5JG57R0MoF&43{Ko{5*hN~eyFy-{P}WD}9p0(VL+YCz0H1pzVpwQB9OD2J%T&I>9T
zc?csxTVL#E`Zzx>uyVY2(*>Lid_?<+JfiNDBL;tzo|RXFrh;Q@k_@h8=S7Q}X^l1`
zbo3?zXwhIEJZ5q@O?v?wozEyefY)3^@>|2e5`r$GN02|!>J!vvat15cwj!a$n}S5h
z!Ok1~)XEVo`VxBZao~T6m8(}p4eSa^LKTcsEUF5OEEzDOS{{vp<RjHLn>coBr-7M0
zi`v{O$7pk^3Gq65F6c%0QSGUZQL~F`DQ7`Ai_C&=&`b2zB)$<P>l{gtlvARH+E(Q(
zz?6!TMWeD5bR|h9t}*JW;~?_O)ViX^Mt;4@Mt+0uM8$R95E$ruN7##UscF$;kDybW
zMo|iJxJ6Z8u`0n6S5afHGPSLGwY|zV7+)}%PI?mqcOMQ$`8pkmCy09dcukVd6iR0z
zpDsxSHf9SWht8j@ye3KKP!xi;6)Y-xji{y0MFMlPrr<dBszqOPOr5qunhOa=Wq8!p
zY7;}g>cRrLY?>is^<ID+CPuEu4JM`TC3+0^F?h#JD|&&|2S}$-7oa$3Q!nXKNS`J~
z`;v?&0ux(OUvcav3=9oMP%VrNeBIbWmszbzmXD!lbgpw_ZE&7NOsc1$cdUMI@PrA)
z=*wt7`t$75I<i99P{nQyGex+m=iN2{X>r2_aQj6JsR&yaOy<_}9{Ssz)=%Gkqs~~;
z49}t+x)H9$^NxH}-la?VM@No)T-K#a*~do~-!rm(`;qr7Uc4u>ef!KkN^j4Xl(-sC
z7kj4i7}XGdIxh+gWy4<4t&wUi!$K4>SsCnEkFkfIQ|Ju6c7EW0(FZMSCf7AZzSlLy
z4Wnr9UhTcTd;bUuw)}uM&^N|StG#+A3(6>hDc_5|$E<OXiHp8x6Y4<JHX+&S;blPN
z9qZs{K>oqN($)Zc;yM96!GRf>P!?wkU>>;Ul)BE%<efj!S}RN)@QH<UV5uBm>gvJ1
z|M=m*@t2?j%@Ld%a|?19rhc$(a3hYbE-X^Io{|%0y3ReyU)EZy-?HyMQ3tqz6YO(6
z_)8yaLA*8lP8}#Ag0OxG`uu4_<zTy^1}f7)nev6(Rj?iok9t?xw3*#@G1;=FUCe~F
z>tz>{XV|cI-Bz(zt*n~pWI`+=TsG7Pg<b6KS43m4v=|rMJ}s>@q&8*^c2`fRW=h?}
zr!M0z-t8^XTi7a}8FGM~r)UZN)5^wgkVWsAm*U&3yogLRa?5)WXA;YAz&`>gMt=H%
zM<04aaF`G%FI~DEGRs#X(M<^IZXJLY`$srHKwG4#R|9swTExS+x$ibLpz;a-Iuy_W
zqRLiFBd!4Pm^7r*#Vf_V8qm4jb{r8lUTGU~r9vMaQgW}fPt{F}Q+kqg+EET~IEcI@
zwxZWy%iF$#_^<m;5Qp;BdE>`|jneSS3unf}AW=sJxpXb(9vgw+!lKeTk)dx4jzlUk
z9<U(`H7mvfdJ28mro(aD!W3w$Yqx4jrVpS*QW`^9I+&nQ6GKyLTeVJld5GdQQo!CZ
zbwAR?2wh3(Ol#JDji6b2+O%`5vUy8NZ5D402+cYlG5ac6g64Seh}MY|FS}&W24hvS
zJW*>(rH@s`n8r%suC=fZkcvhmx;1;#r31t8#ciV)G}P@A*d|mWOp|)G1|)*0+sMK?
zx~L-ulc#R2MOvbyk=it-&Dx0ZlavXIstEP5q?D+6k(L4NOL(wLbcvA`c6iNj+;FF4
z0=Axn=ebhk)3$2*)Ug)4VfY}3GsF<lvY}9-Ar;+iO(QT9Ig@)#O6&Cy{AH5b<Qbza
z+;A-Nfg)rtirNVt6{BfUDS}U<FZVn0Dd-i}-x+o<aifQoNoin+5k(YfD4<Uk5Xg!l
zTLdwQUe`iB<ird$aT8RhnxwF4>J2%=CqqzGzzCy?c>G~pCYr!(3ayhroKK&)FBGfW
z#qPR*bIrQQUM(#81GTVzv6P>A75(={OpK5OUV9C^Bwcdrc-6ru2+M&l=@+#;FBRPH
z(egBAgrmo34buanSCM#=Mjm9jQw^zCc#OMs$pQ|h@iPsz^%T~qU+@t12<j7P7b=R3
zZQ)k}p5}<wg2T#PkLzVokFnr*VJpVNz{0T74g1t=62e%QAtu)D{!Dp9?LBx~mx`2@
z958PD#W>6>oeo#WPK@H+I0!8)_9HV`15(FCgSXU&Jv2xZ;E94_MB^U&i1s#oQtDM=
zVLdf?#~P2Da#cuQJ+1C#qY<ImGF~)6O5!7P!~y{)a3MdNSyL}eblfq<(Pg`~m6hv#
z6@I4bdYF*5h|{H)>v2k}Q-(&TW(~nho~>h~uG>1^K0oBqZ5s4yw{#|Ani87oYqlvO
z@}E)HDm9Ks9&WoV!Rp7!d=#gDN`H#@_HUG9l5|>1AzD(}>yAkD8nL=snX<hmbB7}`
z{rdeG@V;hZq+0~Ns>@o83qo;W15_bvn}B_*A!j*O+VI0Zdd~yjVl5ypw0Ctizfl9d
zW2>f56u{wm6@F8}y&{bvWQjiHGZkf#1cUbt5&=b+REDh3d5>s}_E1Mq<56n}EFjx}
z>tq90L$+bBaM~~2xL94;CMuxkHtOIh)hJrq<AR^j_}7p$BaeU^cOlIPcgQji3R24>
z?gOnTF1IW*5P=l)q9vto00ZVFzsOr%&&w*rmxr_vloJbws9g->1E!3(?O|BXLT-S@
zZd-+TS$OM6*%9lLCid>L=qcz~ddukkYmbg%wjnrK@3Tfv5q(y);~)1F0a;j1!iS^F
z4+9r>-vWo|{tzJ#^nR%KED!N<8|k6TNL^<Ck@Rru0FOTFAwEp1SbbK|1?_G0bl_j@
zmJHNq8<Y;HfnH~COUQdZfOrn&=)EGqEWyxCkO(Ga>ezMy6H{{N%7>B?=7uIjUuDAb
zHMN-V0pp==^IEMau=KL+bOa4uX5gnoLt3hf&}AO8C?38&UDo20xI(9{yP<7{9<pS=
zhaRHz#{7ndMmF@C;5T8(3&?Iu{v!GGE$w6YwRK!YN?nt9>TiLo$-hYB?jKg)aO@U6
zL$_`~Cl7e=;R$b_)`|3Z9bFCT;7_!SsU(H=a5uQf;iaiXYO*gWr8ynr(MBd;0J3QT
zpqCLuYX-)f`VDDD!&tZIMDp9@IWxVXMJ;2&93`0PsBWg;-=dVjf%Jipo3s);0DCK&
zTi(Pdr}kj_vzjJmp#(5#O8cQ3^Q%NVp($w9GtCDBI!#oE#%-n)lzNZcgtAg50;BUb
zKaA*qTIWZ*z{ba+{b<7V4yz!z&1rTPTd#|Ppsk35^*oFMY$yk&U1I{2&Q?LZlnSCu
zq9EwZ5mHN~SP3_DCh9IcnZ-3%<c%sGD&aL(6#oUTPk@M#_G^iD+aelZLKR0`p!11L
zd<2s>YA5b`VjW*;W@jJBEnK<meb&|lV?qfR7Cm`ru~t=_HAHJ44N&r=iP|;RWTN()
z_OFT9;^OCttU15K1S)U#dchh~;z!H^Xmza=h4l>a*h#kwA|%tXc3>KI+h32e^ym#o
zfWr~&bYhVaZ5bug5-+i<NrF>Eb;MxLA)MwE7#QF*b~{N(F!nrg1SAB~+9Zzv3#RXW
zO*^AFjQvrZPT3=YIIzrt1!Q9L6sIE&i`RI;xSd&TI>hE^T28NyZL@bi8N1ouJ~*h|
z9c|in>)6^qA|%z>$<cdCQn+7ec)K<&TCp1~x;0Kqy`#&+9eecX7#bPLR%{LJ>er@S
zm$vPibO;J=Z{L^^5)~EFDLR@(1qQTe(>5%mWq9(G-aeg__{@!Y(WTMTI`7%jc^Yo=
zHfE|JSvfJoLs~U$8y(y_pnY>+_TkVjoo06pZ`G;|GWfOi?f7_7hs0*h5<4UviVO|z
z*)upaQd={^w|V=3*1^$1En0;PkI7;0nXm5X=-bvWUg^}PRjcrjDV@3u1-Cqny-Z!V
z*~`?WpI8Wq{jBF+rWo^i=hEk%gx{I0EOA}aPE$Vp(oTEkXLHNT{~ZjBHsfTYe3nA6
zykX#=HCsI_h)>P(Nq3p|1}-KX`u4<U-n>795kV%5bSP9aZz$UKKZBEnA8_)7R2L`0
zh>n!|_q8-V>n3hGR2{l!Em?WFwv%?^wKqUFtlP9OLp@EH;+cV|C)|42>ah72XjovZ
zdj{i0j8PA)4jM{*+IvQYg^KOM9HF$~TB}y*^46wB2}TH;8M_zg@vdH+6x*8TH$EqO
zynkM!VM|7iTrx~c?;0Q971z6qUY;}O<wXl#E-QOE+<2zMkJ)X-(jsbe_ZZiwBT7m}
z3|TT{$P)HUY<NU`d_;Kc>4mQ?SoHdW1+On!@Y=%D=F>l0GYV{k@QHy`$EIKZjhY!X
zYLw?cK09Z;pSI_}R5SH2qE<twek{tY#BHi(Mzfi1M*CvbYcU~zouPVErRO$T!5(t7
zehcl%@tc~HJ=H&_QRecIBbR4taYj8DH4|O+n?<b_&9XQnV=>!g;%<JM5j;u<oFNcT
zoQrr%I@TZ%gEG4~*T8|{P&jSbiH5+|5%IKjCoNb-RHi%TZhd%ahX*qr?A7G+&znRY
z%sAM8c+sf7_zvi3$CK#ljOq@9ozX>0mJ}6LRw^AnR@w&~%BW5rU6`4;Kcl*9lZzLd
zbgj<Vk3@x;Db*Q=0@^DdYZt1Dik2@gDyl+UKr5)@z0hz$IPJi#)hLJ`f%@1<U)OA2
zaez7bJ9GC}oIYK2`t<w*IXm*r(&MwX<u5+$`$*=63z>&}|FSrLTh=&tNn%&}-txOx
znb!>^VcQsIIQEe7-mHuiCUe@o)gH$N@vZXikQ*u5C!x;Vo%#7Ya!_n^ei)19-Fv-x
z!azWNx0RNqyxy}1*DiFO+Bbjms8O5qfkfLt7Ny9tkG9{!hDn>@7VIqQO*_@;3e^us
zVIeBiP}K;@-1_d^83nsDJF(uGu4$R9cc;wV1vBQptHiH<`&dCyIkwARwTd-cQ(jbX
z>}`1JO_&jal7Q(D$5}K4oGL0v%kxUt4C><S6<JigpkUK*kD}HWEGRCD?B$F|U8BSo
z7L6(h_HA4_d;EN})cNCQS2p$yE*Mo*h_XvipJAvE=@$}4oIAofCW7OsB$y^8vzM&>
zdh_P5*DjqkZDC~OA|-y;<oQ~y_W2`^FzjQoVDhdR#l<s_S$ePbd-)O|!B;^sbZou0
zNJDb<bM-?`>y;lwNok0Pv*QF+vfEe;k6^nuYGHS2FWt3IdF|GN@?@HO@n^3p_se^*
z0uw7r*|`AP6RsKCF7BF3z2#1HCsvV5%6z~}w~vMRDolZghIS)kugXu0+)br7jXVXw
zhTw3ib?tCv6ZV>mkp@zIJIJUonk|YagzBO#u7#+$zOA7?Ar3R!;f(LBcP~8x7{_<;
zxcfg|GdL)+azV<tK0&gNZ_3OuGgc+N_r##`?Cu%eV*3Os9mhVgt?<XydFxo4fP2e2
zjTkh1T(4%KskXA>!+-m!+@+m-V8Dc;j*(;gEqHG&Yjmwmr26!N#oeYI|8mi$npftg
zR!pv#r0)Ibm6Gbh!TIfYhn6r!xP<dpv9sqK?35^Cg{{3Q$<QWumr;#a2m}gokB?`N
zVCH6c2|OzEzA>3`3r5W=PhT{I$DL_YQa$g!mvbMyw{Uv5+=tl~?aMhE2lU@Ccl@3x
z+nj{4!;+Jh6!xDxc#it_u)LI|&rjU@b_ZL)xQafxcaKviPTzOuog4G|V*?r{P143I
zccL}Kd_5iX;_$T_8=@+AF8*clU%#1G@%5v&!kYoD?3-?GY9;@|XFN51)>AW4r*CSP
z$#0@n;;`11xX|2h6Lr#!R{c~L>R;~jNy4%B0_vw9?-DQTbp*$XnDj-nclFOmT##Q-
z(RJ43$2LwF7(b}!kr^`|=znM8^1S?IT}vlDwzX(LLdxWalVcP69@r6+*mu9JtZ$FP
zE<Li-dk*Ro+Gkx^=E{hO>4Wo^5A4@7zgxG_LwXHN2=2SKV&uw*F4G4UtQe?%9#I_C
zKP@FbVq$dC;DO-#bk;<fZhIFr=t{ftpeESKlK6`by=fT_W$&&T>CZsJ>WB1pw~=0f
zg2O*mR{nJOk*h0~{pZo)Q{(%0&FPxGu)H8KKRhv_B4^`sV|R_*ac1JgGds4QDVliN
z_U<9=r^AnGKRx^~3wZP}3pn&cSm(UZ_A7U<U*`<X3F-6liDi$B*|GV`j%`;qZMw2;
z$Cb^nQfIKMN-FSeA$5`>#km#k*mBS|nL~V=tRv8J^djLD80v(S_YK4ZI-L%}bJt&5
zyY|xhyDqI8zW%An^G?o~adKY8i5W9aZ238}?~_aKPnvc(WpL6U+q>((TEG6%x^<V<
zPkDOdh|y&)mX|+2ckc7$<u8`4esti7=2tFW<h@2e(=|31{M3c#Dl2T`q&A?VU#LvF
z4oAF&V&MT!XBG1SIGx>3nb2u+Yo>5};o_w>oF^YJw*8M2-{-HF=Gw-&YPe7ExGw4j
zzA$`ji255Gq<Q@Cc=cIMp4uw8jQ@rcVVcwaeP}7N(Ws@wfXt)bU?+BJ%r@wQF`3&7
z3btoP<@V{58<nwOf$hWe4aLP9($Y7SOuQ?-q-%c0ka69LXU?1iPUtJ`#I8~2aVKKY
z#U5d|FGJ9_ul)HqR|q>(d2aFIbCs3n7A-ngIezchv3tjl+c$RXzN_N3d~xyHODo@A
zwCL^1rEf2uGj`uND+#;^Ay@^!)!cCa9)zu>#)bp%IS1D!@GW*46$w9w;d9<nT3Tu!
zs`lq+YL3z;NFlv~-B_OiWx{C>8hzs<QVh*_bL@JO%oF12OeY9PgM;~r%0~~apT9DC
z_Q`eAW)FO8|FI#Al^14A&+M5W)!Dho|E{HS(uW^!nO5+`6VDW9OuF;QweOsnxAE8s
z*X7Zv{w<6AI-<<iVNLvK%P@TX@a`BM%)42(_SYraUs<*dLy}TgiM+p*r4W{84$@m9
zPAk~TMSdYJa$30!2!Acrdd$?GVMAxK_e)te$<wa2-z!^fOR$$%u+9fqU4pZ#sB$qn
zLpZU}B`}1x<%QHA_skC~0$kN`d#7%EEv7i}$5)R0xIQRg#SbUXmhzma`)o_9K3uL{
zQ?GZ6nL2jfeGA{IVy@*MR$;rDmh4J+H2PcywBTBlM`!GalkAx<*g8b_%HXu1R^UJp
zE9qCCU)<*1h<85oVX~qW;aj%*@by~2wW5lcUh|44%#VwoUpS~&uR&?i2rR$#pmHK}
z)xk{l4je8tNM8Cu6~|$|^XJ=aGhbW3^7T2TudQAE`kb-H|9a%uu_J#yPWm^4ol};}
z+n{yHzpd-vAWQ!y>iQQgP@?TqhSe@5e!@9H7+J*xl$6U#xwh}CIdi^3gVz27x+)Ny
zBx6NGC{odR48pQ8I-3&_R2W1E_9rb1x(D>;74h18N%s|QeSGAEXYYDpMw8M?=Gb{(
zzr|yVcSlBUl7A^=-3zsuBeU24_RzuWTh^A@UiH1T=Cg~NUzsw0pn7T0D6}ee>V8cb
zXln}TfmZeNSSyua4%Hw~aKbc#xY(CXrav%o+L5U_yT<V)ugx2F*OOCLe0WzA_8(gH
zPdw@ho5+`s-;+D#$n*)-Gn>w7y5Yj|DaY3hoBu{-OICD+NByMTPgIMB)iOue9psnN
zsm3g-4X~`18?aB_G18g#NseKk&`w-mE9<b+BKbjuPYPivQNy*4T1V`xsQKly1KOl~
zl-3Vzv{1B>zg1eIRa%f$8l}_Q(j%SGg3TqfRbjD#E3r$QkPetdiGgtj1C0*jjybzV
zeteZWZjC$*`)7Q+O0d-LFS@#(RuZY+{ZQ{kqCV~tENe~P?SY9&pN7)y@Jcu3w(xd&
zr7OH`x}8QHeQS$rzp{<6eGEC=1|yXyDHf}lGO$)TAM2E-OXb+XV>Q+$-y>Cnj=dCJ
zVJo2_opqJnS+>LW2n?5L*&Eb!c(@H32=~2O!AFiq34vIu86)0cTT4f<b!XiKV<^Oe
zX&47vOynh-jG)ki5P1Fw9)S(Osxn49N`g1TLkJvR&@+)2PkG)!*7E)N^WWF59z3A^
zC*3i>GlglXPfeY_MRVB(#CIJZ6VQe&PF&nKeAZxFW4rdQJ!1K~OnXyD>a4KDrD|FH
z02@=}HUTl?<)YvTy{A1nC?s1Q;uj>dUISP2A=>$Y>r(s!HTlBk&SN4&3+21CTd-ay
z=Jp<_Mz^pJo)zAAaU#o0SvQcyvOw)}-=&FRvr-*RAxFpcLdndo4AfG3PY9084jJ^M
z?c;^-Ygd5kPwy{W_&y6baFF@Gzp!R&#~#CG2Q6hcr#w5NP4|}lH~$Co_bqs6DCg_f
zE$VDf)wc7Y4;A?O^S6gRH~G$uDJ{FHexr9L1<f87<m8VfEl<552Wsa(WPvUFR`J@{
zgF#N;F+0_jLpJvhjT`Z-9LtA1T+rNK4r&?1FQhJKXSMCJUyFi=hj8t3>uxPmHr_e;
zIhNCJ=NMmS(C|_kGX%49=#wPzf~+w^iry!|;rZx8JB_zy&z@-YwS^W(=A$QU5gZw0
z^~rQA-9(RcW}kGYbHi)ukuJ{|2f5P?u;68Tq<a<k(Rj(7uE0vy%uL6+SR<nltCq&d
z2sG#d8ayv*>3$nx(cmX5Z)n}Tkn<nczNLI>I{-_nCw2pnaIBhd7<?8s8;%wU^Myme
z*9Lb*SQs1?HeWwu>Fw>#BH&?@+4QGkpVoG=Y0t)-((a$E-F+<fDK<;n_C)MaHoMpS
z^Y`WD-FJSz`CWBk*|G~&E9ld9;A<`OVrlm5&p(@!GxuYy@XA$IeQ`$4tbcqtGkf|M
z>}bvFBIi5z<>ub^j({*_*@YEVwDgZ4fDSg$;Wa^r2&<+O4f%}ueaOiojByn)b4rs1
zV#a{yU~b@NZr9fF8%n&^OWn(-VfpD_v^;hgX(iYgv^%98mbSmXOKLE(V()%qo60y^
zq$PXqRH<Im9+wOC{VLhx+A)@S*u7<?dVxo&cl=0uJpP#VpzXA_PCd_KN$(X&y=|Gc
zfv{?^e;SSFt&K2gyGfx^#KxRVxk9;bVK+$ZeOawt)2^*!O~eOlvW{jxYy%S_YhI$)
z%D~8k@3mi8vyIwyk7QVY$F=W`^6rq}53<U`sRH$iavdV;n^*inQG7jL)nGw-z1~E<
zL^+MPxgn=#F>)s|@Lh*Qjid&PQJ1TW{-{2R)B!EEQ#k><unF3VY)ckw^J*nMR5Y<>
zvyzM5Ik!e8$QNC!<mK7`Z54MZC)A8L&32KmXx-roshuGWv#nBoL7R{#qd#_%nt<J-
zR?rGTi(-bI&B(MSD~|?NH1Z(tBmKs?QsQ^r_jP{5ko3-9=Xc%r!RC9xDSEM(8nkr~
z%yEo9S*P7Cf`VG?_6-b_d${(}y*v#+TobJ4>>lfOqIK8s!<8=(0N=o%7Nvlp`~nh&
zhz9!{+md1jWiz{W!hO#Vy6?3k?t7NyzQ5gbks{Dnb5XJEd*48`Ed0O1ZzR?w_#t8$
zM`4J4O#8A*wu41#tF<`@9X`ocYs+9Xx0tQ`JbO0uz%#W?<Qf~!I<$`mg!jnb1bgY3
z+SYh_7JXaHNFw5w1}_AZMkUX%huA~ffhz4BzoPonCBH=51J{gK9#Zd|%}&gwUi+Oi
zN_oo`iPahKWQkMegiVYcY)y|Y8ikt5-~{o*LHpO6V+3>vF>n<RjAKgl2f=0%N+G9?
zN=zIzDp74opXSA{>UVQe%LWfFOU0+!QvV)YK4?(6@r++Jo|UC4F=~*RnZEh=p1-VQ
zZg4^M!BM@>N*$Cd*f-EujPgxsMIx0>`lRh~G2zDIf|d2eBkFeA3in;GWSVwNTQ2ss
zy&yJxodV`L#Itw~KAT{tTy^W`Vs~58?OiRsmT1Xyw_YRr_<kYnNpID)+AUwB&<ZDv
zbf(pwXlX~Xfpsg}4P}a@fz1(<EOlikVcmBSRyM{~=5C^4!qV@h46AvXE*@@~{2gZR
znLWxh`#~{4_`!gHxC+XTF${L46?wh#t+aB12JomFKG7VD2A1ZRNVnfB?1uvDZ8xiK
z^!7}hx9MJQ&(?W6#(EoSjfn~@>bfnW2-7l=D4lM>xYM~UBDaM_>6RO%nXTofv)gjI
z1P{g!@F%9r-zi4q9{O)X*etSI$vr|_Tl+8h8`}b@xx7m$vK5n8Svr7cpWBM@Y^U{1
zv#2`7qMMtl|FG<cGV85rN60w+@Hgd}Z7D{@(T1mxhQtoqcoa@S>-cKw!Z<&Q+_?Ho
z3XK_{7j$nY`6#$lyeI3SZS3Bz1A}G<XT{yfxis>l{R>~3bKlwl1wBUJGj;0xj>FSd
zu3mOv{^BR@T6ScxZRxOeIg207)T(lxo7HLgJGqxrFBDc39GZRCb0d<c&q>SKm|3ta
zw^w%2p!7`zlV_=0)AmnWdUsJ4$vPA3w(wl&-e(<ahtqPM{V8nur+*epHRsP_$0Z*v
zRXL1SPs2V`lFjbn)7K|6$fqAD7P)Z}rglu;v^1Sq<oFXlS|L*5wQVLd@_h<N5KcD=
z50^h0I=nrsn&t`PGscz2KC`#)q#;eZcI)W$cMN;u<c6mvcNz7@@oww8#ircb`>}8C
z>f86OZyx*0<-7Ve>(Db@tI8kOiAClON(*VyF}#2cnvmVXnHe|r>56Z^2zF{Lc}5D$
zTDN!5H8%O|ypOa`pZZlhcsi(KY0p-xe`O=jzs53aHm|*|J#l{YL)+%`ms3XUD_eD-
z*PtNwNplR{8LN=S$w@d(isB!|9$NYo9N&KA$gC%x;K`@vJont3Q<QQUYl?80Dzqh1
zV$MHCyOUL6Zl7ag@dJ2sOUNeij^@enKAk;>>sERBk4Ppdt}gKD_#5*NdD!WTY#6s!
zTO3OJZ{{HCW%zfK7V1}?lNLByy7eE-TJU#06aP=ma3}}=^jt`7t@@?*jeNkqOG;Bx
zBv=V4kRa;UIL&4%&RR{wnaaj9tr)x29)!QvX^(LsrR1>3)V9EAq4vfrh9?(qYB1)Y
z6xXYv0POY{{8E1Vq;@F#!8jYVseUFdMVH}%(+0j>#;#{oD*l!76=h4!Jk0Qh*8Ib^
z<K`)Q8eMAiel^^lc55iVo8KeJx16;@`EUWP5$s-AARcgTJya?0s|&;%fW+0qfbb+B
z=qC%9IB1<cGkRxITqW8E@(ufb%$4-QK5{8kZU-J1fWsnuX<iP;SzzRdIV5LX2TnbZ
zq+kL<rD;~qkboG=1W?eH_EzSzN0KX_TOXIYt7_z+1$(nsjM=ombonDI20e6o=)ol`
zhAz8vx0<&wZy_&}`>Flg@+39VzF&LSCnh~&{iwVMho-bjo3^8<c;D>QX79gi_f1UL
zJZx|?J3o~0^u~;hC5>6Hnt!U>n-5w)Kl6Tmy!q0ip`|_8T8-`31{`IDI5d*E?vx(K
zxr{x;sKM}G>k-6I)A+4rh*LV3$upOUBRhGVaaN}`pKUTu?v&W?8la2@6dJz~6mLi{
zjC5$~<z=qBEr8xIz2LfG^O|p^_VRA!G;GzDupU4<SXW^NKysSr%Kcog@(<KV^<V5q
zt<^HEE&E0}jY_IFwzISBjj@{V94&zTG>83D;;#6HPwl)>>ezjw)O~8_@9X~PMCo5&
zXp*X5p7A`AtiDk@p65Xe(|iT-BwWWbT?s6}slK74Mb$j-r=&;KoEq9g863sdvrAL7
zHaO|>iz!-L+tE_3&DIFu=p#)<jaQ(?WN)FyIQ273|LQEyb?r)^GjVWgu6e7h(E9Jw
z1{TxtI8)rWlKlG%fD~;&f(KLff@OeAvEX=s^tinZbkBk$<(F=-UXF^HR9bBezd5Sy
zdA5rU!Cwq>VVEGGk*Nq6yw=Cq^`7rgXPH~!+wN=p-LG+t7d0IkH+XPd%%DNa>ADBN
zBvw?&;jUcyv6|)9u}9oBM{2K(##sb*+KEUGpiTxyu#9tIgtf_Vcuy7`hU$9N)^AS>
zcD<|Ch>cmMy`%lGn)Oj{#<Oe6nXPn2>Spe<AE%T5{N&CZ+SC&Ef;J+L-Mf9qaY?=r
z3SNvBWi*v&`J+*aD2j!77gkfGm*#lSV$fOtwp>>0X514Epa7XX5>hJ<QY+2A%9P1B
zWk}X8c$RKoNs?EW-q%jn4wc>%Ppe1{(^J&pC(Nl$vpozt(HRkS>L8=%y4JYf$EbO#
z?F;#!t5}Y8oxf75)NIuhP1?*;_VYgK8~fE4cQB13lJ(|JJ4S<wD;;w2v;VGK#HfH8
z?T0^-iwBmv<)ZppfO-$L=zm`-X70z}OEf*qup1Lln`n>Pq=QcX8*<*L4=P3Cp8Zj^
zJjK0LEywQvLAmd)p<pso1wh_V&uOEa0Y^Z>nmxxu!kXRYj9b3CdkPNibjw$>hY+l$
z;-n+*9^wJCf$rfUzdhi4^axK;j-GHB(zXwzZJK?9!QH3};(2x}*P+uf*2I(~M&gZR
zdAF<7HAmhdcXfU2`nz&neS5$9D(|;nJ&snnq50I-x=JA_Z$MI}**EI2V9#_i<Z!xu
zle<^BPD-8SROqAT_6NcEmuaL0xiAcJVV~qw&i+0PwvJPIwYE3D9rlOR!(<y>{|vca
zMy_=GArlrT-&p%rF0+pSwcwbPVHn`dW=Q+Mn>e;v{`*juk9>Kmyk@dEyjk5lo-f|-
z3X62~g*gREBX6|GEBdh^+ih+jx0J5pplFxm%J+r`K{!i4IGV4_$GOm+ph1f)_Mv=K
zZYeca`s+26c+JyB3{oEUk%#sCCu5MH0mOd_R>J6oPH83&I*kF7W&HPpg`~TPXiO9T
zC&QD17Vg^6wfD%_7Hvi@%uAa*Iy2fhPB~9MXnC@;NqAg(x0r*EZ5|Su8gZ`?#q=jb
zn0(szi%a{<mUdCIN8)g1{Y++YO0)mOmFpAJM}{@3$)~s`BgP4>_yXE?2->zgop6bH
zub5uRGOCFS-vP}qF&70NW;0x{3q4QjP<w^V?$d2ha{Bm`0prq>2X*V3l9W0wB__RZ
z|Ey$sGO#<n9XDXeT@!tk6~<9C)(IXzq)muON{XN>B&Ju$Dj#Q$?(Mtv@86A{WKIj?
ze)W#@fW^0!U69LhUUejJ4Wz&;Y8vy~qlv7hk{i3vt&ScwdP-V-l1B~8oIcDr<(mIy
z9A^!pc@j|nIaT7n?~pnp&g8T-4s)5{rMg5iiDiC+XbE+QY7!UU7&W+$KH)K4QxXT|
zCr76y_8*fxV0;>As#CDvxQqd*UG;m<=S0~BVY3pu4eXybet_NzdaC5HB8PaMG9Iln
zp)UPXiRv31VUgY1nN$yr=@l9lpAcsLMk^$_TghxK^P5-JYas3#2McwI?F!aoz=EXH
zs_2-i7&sIIqUAlVw%P`9#&;>}FMk7jm9M6AzH7E4j=Z+EW<O5T9gEX+oyr$AOCW#j
z$`>EwIjfKp<s@4>E>3IZpX13g`6oF>Da4b-7@^UVCGt`{`2<gplYcJRYm#acl(VoD
z$eN|Kmij(2WXO45hP{so+S9bk{`w;Qn<tfOGfMf-`dL0~uKulrk-;F02D$c%7ksSP
zF(3n?KdrUFYVvay1U)C}2&`B<uu8KZHKe|xo%4c(Xg+aXCBZYWyCt7<#p9~*M9Kph
zNvT&`;ZFeh#lCMeqEXJe;<YBOc$-TAIZ}at{H9eO?o}VX<;<G!2TS>p3VD{OxK|yX
zH0mJNse>rp&%mBaPBshmC=%;^K|{V3)JV5KWqJ*<zBI&(9+K5wfQQ(OQgt`3x}*i4
zctCr`g2ug|>99;VfYBSqxLz;##;<65EsoLK6DwQMmg@M@Qf=@2`HrSHaeNWdu`1=P
zT%!AJ%yu{0++-Oe9W+thbn+QjfBB3@It;Qr;90em>m+p4Bti0=h2$|gT0b|%O&@A)
zT#(TT?vt|6{b0U18MevFsrieNW9#KQAWw1~P+QkK%u4?3G40V3I?ud><(mky%eOtx
ziZ<NIQe(B~ELMrmUVFs@KaUk4-uPg1Gjxrb2=-0=@=I=86mdB8)KR1Mfg*y?=yBfg
z!>F>Tt?j{Eg|-K^@9UMq`{|PKJoef#`ga3|g=Oio?H1f~lFhZrqh{vWbZ%<XR-=N(
zs9R&@x0=RkC#wkg?;bR*b2b<QbZ;6e3@37VwXx_))T#+q8R`!_Q2k+h18`mS+QWll
zmW1-Amp^}9P2{g@|5=|<x2FyaD=%)Os4vKjxL}evfKMZS55?BAW|p!``Ar?Gjj;%$
z_0eyKSOx+5Vg>iN=x(&Y4EKp%=y?`(;Ikv`LS?~4MXU3e=z;91x?qn~G;lbWBzy2v
zBQcW>?1eV=Mx80QdfVQ-?+?-bL!>3ygM0LS5mAzxMhr~zv(n;*ezqKTPKxDQtkNK#
z>H`-%$vrp)V+CfyRL;V^e8TnFZ9tIpvicEdXy!9%D0gz5mt(bh#+n-scYBZ~Fxn%{
ze##toSZ$PUf8LbBdP<qBrZoGDW=hu>EB6cLGl3nYOSiw|&dp>aWxD-k_fwD3PkZJU
z{U_ai#yh1^`m1J2kJ4W;pLvvi)}7m<^w-R%!dvzr?`yL`8Vg%Tc+1FBMqB0?-m)I#
zCyT0i4BjMimC09yk4!CIycpxZ*J>Zc_|FE4T+qu0r>rvkv-&Am==+c>;hn|kLj;8f
zisdilfHH#eMI5J7pROXKU17dX+S<ZTzM-ZB@#GuYp_`|lo;h;e=+p-`l$@AX_R0PY
z|IE!wJGCnBU`hPOWmRvMPk*VZ=;*QYHEdwb11nanU|fYW;Q`iV$N6=`JN7IdmNI^L
zM7!JrYv&!GoHM3q(5R%!qKwHSJBLl!F{kL>f}q~xH@z%}Xw{5OxVaI|z%l5U7#V&b
z=fDHe8*BGcu=d2<Hs3_loWAg`R21*rIV=vN$W{t0W1}K$kte1b-pbAydNiNvQ@v5I
z&)|wR$2AR$@m2nvue#gCN`85uy!a?pWg_7^WrT_Ah#`fkNmQc=!<<Hq>>Y=0{H%IE
zyL+d0bl2M}(>prseZA{~k*e-ZA+38SdW?Y=ZOD({fL2zMcwtZDLAcDrJ)0)bY)y?3
za)Itj{QW+UO5AF!E;TBV<XQD8r(V49JL~zfaHXHuUah_OAsk0;x+@<yT($jMwFmX8
zc~hf(|4Exp&fBjoJIX>9zr1C9B<2)SStI^}d;^|voQ6v4$!XHgW890W(q>$~&2!c@
z=iv5LiD@f}+DvUT_h`lF{X5t8O)JaBuc_5#fAEeq8eA~A$H*QLJ!Xv@Ja%xu@zG&1
zm1Eh%4Lv1qT|BFj&lz@=zKe>Xi)HisoNJxD#I>$k|1#CCX~tt@@%y?uF!a%Zk)QOf
z=T+0ol_ei>WjStmf7b<U6~LrV;mJ9K6@@gXU@WL{#)dncMxY2LRuDIW<I4{f7Gwl8
zPmE5D>OXhL#IhaJLleOL$w@OZBS&@$T3Ft#gQ`S~>61Qq==5eyFz#hiG~%VAl~ssr
z0x#+dC`@R!I=_`@H&Css`(9b)T5bX5vlpnn+`rz0B`v)Me|U`&$?fbz^8c{o@<guM
z3c20>mHu11&Q+Xm*4+M;@`REF@ypnIN-+P6?IW>AfN+yL9HD+<3W*PMCNvMt89m51
zFwyo=eAj@4VK1;ozRp%%VgWl^o+wwr_wKZRRl6LcJ_pKJjc<kp2!(RK?NvxviS(45
zue!KmE2Iff6C?4?FGgm^j~@I=p2$>9CVYF@0_CMxWn;Hr7x@07oZzQzhz7L(&Nm_V
zeA^vzPkX4yjYHts(`;E+ZMHr1gS3b7c0Jyv8gBz_PFYFOW_PuRRy>?WusbPh<!yEw
zVBM-ESi#EE@>bjHk{xIv<`UtT981fRZ(dJfjcu==%Q$#2;~ZfetgJ)2y&_$pEfmq3
za%Qsjd%pHt%5~e`A#c4k<bI^;tL%{H+g4$`ivDN%4n()(qX#Y><0x*GmwxBOR>h9o
z70Qm-;?%&zks}iWQ;Ug`8<m-InC(korqyo3?+{yze2=XLziS0ma+S~JakfE{-66@q
zTsn#0F8ns>>vBCW4^}>>6@jSbE!1)e^zbcH=hfEYbb%3aH|*8qu>T^>vd>4lRmupS
zjVI~$Uz8O4eBq^M>?if_{H5Bt#0xl-fx;{tq57&ZUgS(@+^i%g$=5$&dt5}Iv(v>Z
z&ARHGuv+t%qk#>AGUTs8i)K7RKFII4@H?o^<=c70LXf8R10JDVa=%k9-TKwmRJ`L>
z>o*oVLz}xh01Jv?6de!uY<riuxGv#wacrtHHr9zBw!j!?m)O`Y&KUC>IQ)!5e_yiw
zCg%A)CVzN#be}%ak-dA{eoO4qy>H*{T@tB{mn#?LZuUk3W6aoaO{>ae<CTlgRb0l~
zKj15ggFkM`+d*;CGgyBWC@0`z!CO}z|LO3GcWseuc(V3tcQ$gc_8MS?DA&+FZ3Q<u
z!(zaAY+4#?oX;8!_*J=<_T`spPt{3f@Z5+2g6GnHwYQ~IPa{=d<#%}=n4%$`%U(Fh
zNAO%>t4TiTOO@X<=fwpLD=Zuq6gMvuSZ`7;$>H`K<RwYx8Te)e&_EYL74rPV@$vEF
z6B7&L;|mjGlapiPl9KE>J#*q?vwHT-ijB|dnKmFkcEEtx_yNGFBY1Z=c$d!QKtB-f
zS3UTS+^-1O;qS%_PYn<588kL{QfmCr0paa?wJUH=w#D@96&=;1e|FEF-6Eq#Qg9f{
zvW<ZpdDh~ie6@z#Y-5BRsnUANky0ly0ziykLJZ<hz#SeY7Jp{^@MaSyjvh3vPoEj#
z(d~u>$ERsOCI)|YkE1wlz-VV+t1-=6&@y87h&D)eN<9hZFz87N9R#jeZJ<vW#>>rv
zX3PqY9MGwEmlhrSbeUa|da!5Do|%I~y2aXk7TA4eXC#$R!kV(CN^3~cnF8NH*~#N8
z#;4e3Hr1YQL_O2&TXyt8`W<DL<%O`n9ro+BF6j!MT(Jl8Q}!?M<ac@k%of{zl%pI>
zKqPthA8bN*ZJ~o*82ljK&coYS`whK3o0I?6Uz^m~eq;W=!34j(?OS=XV+x?%GN6Un
zZphmlcX&#?b*98~+~JmZA-3<pqc=Py-Y8Sz*>7aL^;X6~l-|$wi#*@{bwk{j4T<M7
zDIh;RT}Zs;wqNQ<yt{1Y<q*d};HDXMZ)@*}Ui2NFU@AaRBh%hr&bA+t91h8NVzfDj
zoDq8FMP)hv#rB;=@9o9EC)>Uoq<tDrt@@(vG5(uFw%+b4zLo5d2WsDk6Lm~JwcD?#
zl=^P0d}v?$jL?FRHdvtLZ!&E0gxM={!+W($ZZ}rjJ}qZzNsgcIovnIMuB*yXe%9XG
zDx>%21x5DWf!b!8;$&aQKk)a(ygY0P@I%C7YiHRO1ZkUS5X|~3zsa*5jgeRHgGhlh
zcfi9s*C80yy65BfmrY*ljE{FlBqXqXUE<@rgeN388V74H`T8U{X}1bzg840YO{wND
z*%Lrb@S6P%)b}UlAv{Smo}2@AzaY(PNJDF|bXkwZs=D6KbSC)wCgqegYn)(Di0#s<
zYqKjCJ2?X*;@}-)W0b{mx;+edw?s}qhwRjITbknXqzTI&a%&5FpthnttLpBlGziZ4
z+rY7~AUMZ7FfrgXgipMM52%RUfIvC7=6Z4QvI#pq@#&ydwP)Mv;!^;8;_XhOUT!!R
zJ_P4MPkaQN9d+^Pinz=B?NykG3<Z6}UPLe%1x9`iOT?C4<9bJQiT4lhT-Km$RF*Fb
zi6;%`mo#_)OROU(Icuk^ly_iWUN6yTL{X+Aq2UrA&g~rLA0N@Bcf#m~OI$f1wVz)^
zTx^%P&M(#}PiFDp+=ccxpj&AdbAx;8bFCk?pOAxyStH`&w1<g(@ngWH-Y(mdCbOE~
ztd;g#+*S07K7hUCPr$}<8MfE#5r3sZ@FO6rrcXn#M_&c!$}CK|EZ=MYK;TXDP##u=
zr-dOmNHey-uHCV`VbWbTtq}8@IlU=}-e}sfcKbI%l6?<J@u?vx(4+W{A+y%n7HF$m
zFtXK5`@?SkR%i}Gu0RfAB}qM5W|OjNJ2c9)U6m@?Y_^phVvl2H0*)eMYRw9T?-@;l
zVERw06Emiy@SHR!7Y;c<np-@>%DMod3TaT>Ga(v-ApoP{CBEDru}F-GR`CK~wlKVf
zx)1OVbif>@#ptlH#n9Urc|*<ijHX1$NZYn_;ItEpGob^=W5mYN9aam*CB%luQE3#?
z7UztT;|OES<Ki<uzGsvSTmc@?L_v6uIN+E-J^0s&WmFwxWB^FmGWw)%V5lA-N2TFE
zB8UYe#ADwu8a5N8P}CF)xG|R+7-JbR34$uX)+R7X0P-<$&V*?A5>PQmK%gU9r0pr@
zn_KmCi0X>c?Dn2E{36OG1g;6)Ix+jLua<;QpE<f&^TIw^h2i#sg2~4gG?#4(w>53!
zbT(_`-zcO>K*o@ufVfUk>*BkP2)Z+^Qww|BMm|mM$mrhd{+A}^u1K2}*Ie<9?-(<p
ztxw|~t@ik|$qoC_-oB)N%<$2j?41g80|(!eWOuGw+g1CNecK_v15X^-%%^Q@A4T!u
zclawz4sY4~f3<flv27e@STnObJG)x#lKbFOe284`Qlun`)QU7I$$C(NELpN8OO@rs
zkY!tbfNqUcunbpyz(owybsPt^dkE4Vf&wWVCk=w42O}_w7A>0AEqv&qhoC8P3XGyC
zkhaJnmFWIU^^r?~9tyMv0|G_NEN5qz^L_Kpzw>`A=c7^TP8dvC-AKxB(yK?_eP=`)
zP?=uqHO1&qufVi=U1Kzb^`($p7zrHWvxSrqUf!HCo;f=&q_-MIY&@K;o;$XBI!JLP
zjP<%E3=HK1RFZ5y5fvMBWBk?EUanQQ7Boen5<#-o6N~A$&#b>V#8$Z%?xb=u$8{R}
za#|piw0*B|=GxYCb}1r>+3vaf2V@fOz+Km;j=pqZTtX|+_2Pb98a2z<o$125E0$?t
zDmiidme76TnAIblMPVwNmBb)9B&jhaJJW47$gk30ovYtE8MUH(aDF|iM;D#^$WXUf
z9@hJderfjT6a{Ytf5Bo@baJvYcW$BUn4-B933|0Quyte~L5gXcK1=c0i{bOip^K?2
zzVWM76to=GS3WV>5w3)k($@LJ?Ds{PNVyvem0o6yn`%+PNi3OS^;|qWIT`Nk>+96p
z;mIX42MVfWOLLQ}InG#2Fvaf7=*HMYfGSSRENCw5bwaj@o#9-ElqC6y*%XaUS4E_S
zO``W&9Lbn|xigupXDG7<gABuMj0`N+O3&&S$LMI(pA)apT>0{a!-}d)t`!|sgJM86
zM+15yu0_gL*K!dzq=Uss!bcdVW*x0BqE~w&63b*`J~cla!NH8g<kaN)QT;(EFVYkf
zMQ)OrJ!uBRh@GOMC_e}-)g{FxQ>AcJEmx)fD?Z9)x-h6~4t@OOv+QG78Gi&W{thYv
z?|Qfb!#=<apwxSt2H*j%grJ8d031LV%7ef^o)!nwCMbqKkoO)q_YT;bQ;PsO-~<Oy
z+|vJo<57)iv`8)3TDn%ify*O-ekuT|D8zepyaR$Jv=2R~7y(lN!2q=H01H9CTCI8+
z#imzQf!F{r8ZaZ2dtGY<2wT0ZRDu;~z&k^0qOUk`QGCC5QGsH`8H74uvZ4SDL96f&
zsvTgr)>S}Lqqx7tIRfv%={&&+yX2(t3ojj=y55C5;zT}hf37<#N|BQN+}4`Oz2dWz
zp@5d#XLNF$oJ-w#fm>xmFRq`tt;b?Lnv94<6-8Uvs@7h9?bY!O+7P1&-j*a9$PWz&
zT3yHC3Z}u+tH;h&v*GcW(b!50^JkwirZ$(uMyinI4+TaFa%f4%DY^%Jp?4@Mn!Poh
zsRP=`JMSL3O5c=|8><R4CR`eg=F6&P6T}%3_NXNsWy(Yi1K*&v;}glL0xgTKj=N@D
zTsB8bcz<y{x;>bvm@*Bcvr><B>_oRfXv6BB$tp2b;x6e&*C$gllQKaZ8P2P|ST@6?
zNKaLd^l463<rwiLBE^GJ0~eROta3z)C-lImsRqQLIvTZHNmrG_7cQ4`^op4CM@Q+{
zMg7@QZE+w1_YAY(AYeRGPbNEMznLMN)vFUTtRb;F9iyq)i9AV49l2o#n>N4?yIR4F
zISLI-j2U{S(^gnw(cp5clXC!FwJ=~Zh~BsB9;!Wb$In$oQv$#%Mb#WPSZ`(FMouEK
z_&^p#dI})Jv2>5gCOJ`LDJteeZ}m&nvKkGSriiOZ^`+2u7I$Sh3abQ>{}BCvJ>qV|
zzA`WKD-Bus6^P6qY+C^w!9?dJ6!hEtfM+&Ciw?}_Fv4KW!tjIf-d_&79=>33+cr2`
zwuUzN|G=bdgKGC0*ekOkOU9d=AagQ6f5Z!Ff*Z!@gQQ{@_cA()8iZdkl9q<QBlAg;
zfxT_bhjNSv-RRG8E<uzyniEh^71Q`VNn@hQA_089-4Hd}qpis-5pM82byWYLL4+b-
z>x13mqD^G~7u!Wp5V)^BL0LG1cHdwE;$+$C#RL(8q)u6HQ+D5b5~siUw4Oot(T~wD
z(67<^=tJ}o`ULZEHvnr-70&;M!hv59Eduuf6Z0@4;ft~DAgHB5H=TA2S_IkFZwG+`
zp~zEA8IN&W2el+&#Re8dPyuEVgD#9z)7T@&rt}{xr|H@Mf^vHP+Ee9ZHNJv}xr8e4
zXqYEP6zQ3JhM^??f;B67ix4qGlDyn}NDz>CtcKH~8bm_#r@pvC65LfI5Mipww5Rgx
ze<8n`mn5x+YNGL(LBHxyo^ON`OfJzHUlfX}UnaH)c3^VZ5ybXB(rd{whXj0pN0~6A
ziK&DxcS^v+uy$_y?(SXwV#_b9vn>T|3!Zf-1L;McZ?z}gKuW>(ED%SnZ9D=heH|5$
zHd<uYItr&&el~V_?%L#aTx_1qqlf61d4c^kbc&z-Byb~n@=Md`ht0s!_S%vBMpymf
zMl**VU%GyRe5?7VY4nBW*I*kwMjKDQO{VBV8{dlfLTZoD1<N|`oCDs&D(FTUZET_Q
zVzcKw7KP8>dqmXDH|WA{{q!G#XReMl!Cq%~)%xD&P-Eu<Xazs{CifN{18=}yaGjoM
zn}=)imb@SyA^8!+q~{|5+HMfi$-Q--X9J%0*j4wtU^!s=(ck?jyzFZ)R<{>TsI*k5
z1Kyr!|61QdJZ@R>_(C84V_|Xc{bJ$x&CPwAH}7mmw$V)4-VN9;1lfOYmz({ym6h7i
z($e#{Z<9;#vBQPk-+rnIt8f^aK#w+W-rQ`K(EHol&GgQ0#&t8;_FkJmSsosmr*wX3
zc-h^7%V}%dy-Oa$cZR(^>4-eu3Bhe&@HT?C>bP$}9n74tdj@`P`E{@Ub@IEN)ZRhh
z$^AAwQT+MtUG8l-Zd&O7*Yl*d#6L@m|0sos;s~FDh%o;55#i9v^z@0)DYo)uK9yZP
c5URUwJ#=6-3o(oujgw8<glMn(WVdbSZ;V8Vw*UYD

literal 0
HcmV?d00001

diff --git a/studio/frontend/public/fonts/Hellix-Medium.woff b/studio/frontend/public/fonts/Hellix-Medium.woff
new file mode 100644
index 0000000000000000000000000000000000000000..86e46d94de001b27c119265727ae23e9f13e859d
GIT binary patch
literal 58448
zcmZr%V{j%-w0?t)lQ*_)+qSc@ZQHh!jcwbu?QEQljg5KtyH)q!?W$*{&eKnyKGi?Y
zOwZ|YR}dEmfB^orLq!1a-woue_&@9aYvSTR<^Jhd|MlVihX@MGdkIl7aR4AA;Gg~T
zAFR;ju;e8a6;uF#avcDGX9WOA0q1SDW06o%75=AN1^^Jk005lh7d8%S1r-M7e-r-W
z01E$L;M&!pYHVxh002Ot0sx@B007u^LvIwbxuNqvF2m!W2I@b2e*?hGZ9L5Y03<>H
zz}*-Cf`I$vg{o>{YH0G$pT<88`+va5OmMgONB-jy{@FzTK#D*K^JHP`^3NA2DFE=}
z1^|F-&7Ev{u(3D(r-S(y3xxDP7zoJ+zt|dj{F@h~`QJE#e;@?&04UoT+L{6Y5h4G^
zKm7}Z-YHEo+ri%11ptWl0RZ4%0RX(WUCO@3|6NlM5Q#`SO*wra?;`;Azt91IH7S+<
z;s9)N-Hd+_eThS_L4*G@p#Y%&u>gR_hOvQ(fkD=XJ@Wx=ZMHzvx4fgCIUtG<%nT1u
zhYa~|!TwLPYJO{OU=9EwCo}^WhJb)%%mV=cHvqH$+izH5^fA~!+23z9?UWiAXowzS
z#yyoViVAVaQqR;6!UKj3^@#9n>|n5`$~VXUx@$73iQ5^|0Rsn@EaVi!kp2Xq`T>Ce
zh+0En`RA`j-PQtC0@KgU?dcj@RG71n$yNnk4&F%;N&*VG*b2?ss0Lq%6dZa#!N6A9
z#;QhZ5J^<@FoBpI4Voy5EFwZOL9sGh$OaWB#OcKviE1%bx!hvG<h{G8i;4cOXg)aH
z+{}%CXD;Iz8D%_3Kun_#QDMHQmiVHXxJDs=wj(&(8OC<i&_tEtO||a1=@W7DfzgS2
zw8LDH!TJ*6*SWTZYWzzLPK_6A<)h~EIeK^5T9H&*zO;XUs!iDpR_y6QS3<Hc91U$;
z+wE3-gJ3-IO}m9N0%6@ng!0q3;Wz31D%jx)e=%;{ncJc9^Gi+qRv^1?WOvpFWS^3o
zFB<%;ou5CBYNWfz7iwj8=;2?BpJ3y^wYx3ZBG|kL8avTiDr0r}#fc?4I5OQB*ef^7
zS+9r=t3kPIg4CDUN^_za;9W3bRvmLY$9m?BO5Tz$4z*!a8{m04q_me4(A3tcmbFl_
z(+X`NI@Arkov-oE7bLq>)G`ieg?9A9Y;1rVlroc+18y7a6s>7YSx%d=FVioFI6sx!
ziWfqCPjFYUy%{7{x#M6|885tzE|idD@`?50i$6pBO1DphyQ5|GF$JcB9{anb^l|Tg
zcOB#2)`s{+2-^$FufG_v`{*B@we|L2^UL#Dk4a3>^%9r`^J5l$KEGe;x)|XCKM}<^
z`Eg&ULtYB^nj(2}tvwhGa_wF0J)v2|bMOXt&`6rr%br(g?`)p@kyutA6F9`>6BL$4
z3#~CcLq6~{AMgu+(&PTgeF+}Rnxm6MF4A!hAB__zqm`+gwq<=VVSgLNHLeW$f_Y=+
zT)Uos9mJ2sle-a5XHSON9;k1yPTqlT{5e=o=A!i6YjS5ey-Dl-wdFzd9RZQdp`K?R
zi9s@D-FVn_m>t=WY3|9;fnnN(XYD$`Q*-S-q8LB_jr>gxA@U09IgQ77uG@&g{0uHP
z&9GB4Jx6Ps+2T+5G`aouK**>nXJh2elt|QT))$>kAs)%#n|n=yJZ$z<cZQ?U@14JN
zO2*}gAtBOJSiMb{S6;OOe+puP@@vcqo*VqhYO;C5A)Ige>Bl5@h;284H?p>;yR2cC
zZV>*(+F72F=S~j11?BIF>|;EXlTX^l;ne$CN)cv!5$0A5(@XQ5rtU1z)np_%gQ_jl
z!sw>B&&WMJ?qHj2`{;F@YXKDfZI*Q+bC_)CVsnzzV-IkvY;%kX*o-^>kZfW#WJ@fP
z%!sy+*5|Gq(Mfs|&{73~9N5v;zvbUF`2<X@1V$`>EmpD(rbUT*qlD)~#2E_m^uwq0
z`55$~;C$JJ*su67Jigia+`q$458C)|SRFImC;h&!SjH!|3(*(+qvyvsB-*8PQJ?U~
z@FzQ=o~D_ekca0S>n5;VDy43Ex8s2)LBqYB(Up`h;`q$(pZe45UL-=h+r(OmSPIB}
zcX8bNL%&0%d15`!5tx6`%y4iG@#IS%B`9^R^N%(t72;u8Q!xdhHJgU2l9Io&e~);U
zaWaU1(-GH^*g1az#wb47FREf9kA%8q@GxL6pI>c>E*LOMrgxPsVcSc!&VRI^Os(M4
z8h2&%N;hoVj7mVozLxq1x6j_eTC`3sSfqH@rkLcoBz!U(;blHr6^1znc`D0b_FmnX
z$2#wjZdy)0kzZ6WPq&RKnjPzpk+ZbLH^!>87F8>)tLavw%r)6{$8%pRmFY~LLrgR2
zn(%3grWvlLm4-RqLBEe_sIO#rmE}F{eUk6JP~P$Fc3G)F=8)=CBRTL&;1XGRs)g#x
zyHx~FakofrM^)R$XT4&z0xOKSh91XW)mE$Q&Y5NXJAbX2`k*=fdXYY=sOvjs^ySkx
zF@N90${AnvoZQA6$DTx+bV#^5;HBBpe@gZ})?)DbT6?5-c21QO!4bE`Xi;AlFR&Hv
z61_s&AZiY*;CbyW2cJ9ghkSbfU7PlexQg->ZvapGShxH~G<mtHdSJS1q|K<S8zaU%
z_(kHFZbrdXE?#!~1H;z|h5KK=?^%?PMH9m8zx!^BE{}j$LVWD#f_d6$$TzylXH^;&
zz8poyVOOj-P#=!#Q8&6d$`=RXKNygj?G{ozlHM8J?{M?;rhD8Tld}AkV`k}_@FB=o
zS->H-66LQ*0bdom7oPchDnqtWS<YQWCgC=Lac2qNY^T-zr1PE+d8YcH$R}%Gi*bJP
z@cwpy?U?JM&``EjZWV%~6gE~LgIUyrWDUnij}4rLT4FYyhI4M~*j=nPs47J>6Vt@6
zN?{wGY;G*n7wdD|)@M-49Z)**O(oDW>Gv+}Sy~8wL_8Y#*gvAY$+|zsOcWOxg7V<=
z^L_-?>+Be!Rn{XrTKO8vLjPpmi+tbyJmq(l*}A)iYAeFo4|{|C`2F?MKT6<t-rmiD
z_ks7noWrBP(|KwAd$Uo;JdR~$!iAg*OTIJt6wdk0bJttn3&YFYqp2IOd)=>VoW8kj
zJ?n~>XHCzh?DywY`{LrFJKFaLfp$GQ@;I&kM9>q@;+Lbuh*S=tD(t||QW#a}0_*~6
z8MbmMIx7@$yuk_?KH`XCekE2#)nzDPtSY%2ZRY$h3nL40XRY}pwdT^;${M%kGBCU|
zQS;9E6g!tP$cBevjGi-*?8*VZ2Z~$}I1|KRtU<*vrfj4b6VxIs5zxpOQ(P(1WMOK{
z7>XQdS<+xB6K2gGe?FY`VKcX;cq4g-?BwiYRCCU;AG6MhGCKRr>QwvK9PP7mcFxi7
zTL&OOQ#RNPadZ~4a8q1k{-%-uvgxD)o5KKH!^G@_WLfiy%VZHJMY9Q|<9cN?ht#e^
z<`hq5+B+v%1jaPgq$$+Zq&Yq4w9ZD+)JYgkswopL_d`?#Qnh>2p6cXoLFCnXokr7?
zYO;8`YGaxoHs+8#zen+-mQ#fh9Mgk;O<_uz1(Hr`U`ezVOD2(ifd8mp`S&K~u|YTJ
zPgw*H0PJOCtWCoW*|lYDPurPo&Bp(pdCJVm!eg19ov|}BTQ}2VFiX$Ku+1Yq>m;Sq
z)mkK@J4;G+kxC<#N+Q)w`tRzyOQ1g&EJ*nCyNPHN(SOjvFslBwUf&SX@L7^LDHuGF
zxfmasq%1LdyvV_ZlST*Qi_eGu14c|lN(tf@9$Q?dSdQp8$utrRk~h-aDJd(Ctsu{Q
z(J8<Sz9+ahMpPVzxFwM#F$(}l0O2QuBs7470uUKsaDvzs%`4q0EmN8<OD;Pt{|F00
zGWJRLoF+f!#*UDWN|hK|_&U|4g^?MuYOu2bbVczE@(=GBVKSX#8pjBjeF{);@*)g#
zIk2>sXc*B7xKX>YJTX5Z0r%iv<Uia$n7>F7P$Z(j4q_Z<+bi1h95)@;?tXc=e&~E?
zf|3U&=gX(du2bwm^9Cg|ifB+|Qz)Z2qsXAZN1{aHM8->fkV-6iIVm&$;!W(%`;aU~
zmWd=8_WUyo`vAKDI~J{qoLWpq*9)#Et=Fkvo1AEZ_FLF3`SVcj{x8i!)vUU6ojj#%
zR0^{Um$XocNlCd%sj^>*j&jAaQDriA@lKCfcWd|8>syF;!OHwc^UzP_N@7=5SCYrT
zTp9I6ic@P#K`l2e`wg89c#_BhqZcf7dk=f^9bD$$varsfp4o9yC&g7ud}jj>u@6HJ
z66hH}8Vy!Cr{-W9@L9^SlHi<2HAZztwMP9$(Qqcn8@Mi|*@y8>=NkQ+>X;icwq#w8
zCCw=Rvcjg3%{iIc_Koza_Tu?>4G=j=b1_o%|Mm*^81pjsZA*%sO*Nl0(LaK6;^q3w
zHO8gTF}?Lihtr+unVf+NI34^K)U!;7T7A~lF|7mRHN;1$muNSwFV%lg<Tjo6$J}Vv
zan1epF3Zl)P4JEVYl#ofcYA=Q0Tm{d2K_p^0l0}!svz||R5obeUfZ6c1ARLN=pb!@
zpaw}u$ZjajAl`n(t@sNZvJgEbq{M)jLmy*rf_^yzF3pFk2g9EzgN1gBeSxY%fkKuA
zXQT#6EfX3T)X*sT=<8v)28Bxu$Pw#5;lOF&F|bcS2c0DfZ)A<&kwBgQ+PC{xKB<XR
zttdH@WOfqSgf4Vzh!!y=72QG_xC}xGDHWxvjP2shqVCxh#W}@FnSEZXv{FgbV#Q(+
zvpAN985Ir<Y>Mu9eWOu^7&~=V+EohJQL&R=N8^_2SB+1DPt}KPT<X#ha);h^GQ(6n
zH5A$wjBaR`QH4?0pY9}`irNSDubM*jv#Lwgugb?t*dO`%G{rC)4f*Reqd)&vUabtA
zQ;?Or(1_$g=RM|`3lOX#TAQ#fu}SDR_Hp*%mY-H4tytDnyA`<gJheYb$*DxI%$x_j
zozi?-|6S`}lYti&S%_e*H`8g#*Ic=@o@G^V4e6}zoaA=SDYL7z>#*y)7LQ(vYG%`_
zaue^W?Wv<!{(E74zW*lpsri|MP-T2J%w1)t<i65x8GwC^Esul$Aj+p+x5D$t@|ORl
z=)WduiOR~29UL1T>kiQ=WEB8M5|*%E9l2U;^S0W&;Ibxbd(5Go6-*a_E*{-&T1+<Y
zWFU%ZDITe0-HdG>+cgS%&;xC5l<7fU0?o`Nhf?}uDeP=9hP`PTyJ@W3uGdzqO<C(l
zO-ha8g~dhus^UfO4^t32v#I8zwxfLb?NM8m7P(g2R^4Vk5fe$cGeRC#P&uK^c~72s
zc{|9>l&jpQqo?gBm2_~7p_zKH+(40AgbxhAYeF{@S4nPMJOS=j?pmJ6oFsW`^ZdW%
zu)Abu%C17Xy|pQ|D%Ra8_AMG^)!N&q*OzwX-ktCL5bLDu!84|RH`Q$3TQ~82fa}I*
z8CSorJkMvYQ=Y86c)WdkzIF_5%m_*qb~qL=&b#W9uZX*T^P7HK?p{bdI=|6<l6(e!
z@j(s>In?7O7;iKjw__3_)I<CN>kTXkw5AxHaG=hyJ1KS^@Z^)k28Yaolz>cw{O<gt
z>w>Xs>LAy-s>302NVJk*ZsB|CaXKL9;Jl^vobJ8id+moi3S0_*9h%*b-Nym`CV497
zW1lNFpJ+PhcrxnFZ(F+fEAM^M)7Oo&14}R}@p$wG+Do^qzdIw)?hj%DW=s9wvt+mj
zR!3EJ72x;Gf3ibk0s*R*=tBrfT)6)j6hEviEc#QFSeVo&M3OoQnTVJOn~*rkEXXX(
zOcJ3mgA1atFsXt|e(5D6*rL>;f-Clba;xd}W{m43*SiBF{pS1I_xqLq-t}q6x~Au8
z!@904N}a6m#V#__Bc-RZtg_E4Liz`84@5=2p^3r@;of=`(4iiZ8uB#O|KyJiuo<K?
zAUE!1u6t5w1kyiTdfva)AMSDhd7h;5Wu@GRk&?!9>a)fA#$GB;z~d}@$be8Ipw;?M
zUU@TGKN9WtHLWC=e2PNu3cr(>Kv9Ea=x9M!J*r)Re24Ko8|$|kix<i^W{Z~_6Jo!?
z9+={U)Bf%t#D1jo57nuS>p)P~j3^bpGG(qaE?`-pURWZ?8&Z!>Hk<aGdsU>r<J$wq
z*mQ(qTFV#COfI{di|Ftv(~_!*4Y4}#;8cSSf9mY)yM$-;@uhyO0bhr&J;rAo__>;|
zT)}tSW|vM)y*Jy@)qc=29v?oiJ%RUq%5`4^`NFV*U(SX7AQql>tr>M)d@bb**eYCr
zN6dZutvrwO{hE8L&Dx1FLssD2r9@3hF3=TNfNLWvGXd0F518E#{rP44<da^i!3O$0
zpjqU4#|Q_u6WH@W0t!bGZpl;{2+<=WkiUcgh7ZO@Kl|D2vp%yiWPL4Rmh(pD9&f0o
zB`?20JG9%!*?sv@-W*wm(fujZU){X5V%Y@RU#j4@wB519!e_OVm@O(2($^NQplD*Z
zERm+0m4&wpN0s$)UZ@Ucj_t!64n8xjRH52o`@D>FPAyww{*%74m)c$y3vxz~Av^%d
zitko2_F<5a(|EnO&lUS@<ck=9|9R->`D%>WH-p&;-;MJqE$QMS$=GIba-hKPxg}ur
zDf_{pWCbsrk^)fUY$LrdnmnGS%A#0q>U+ceoLlyEZkx+08|*?0-<KS8fK!B5ja9DL
z;_LP{IJ|3xbPOB{XnyC*M+ZM|?(ayWhx#T9oUs%YBs?AH;Cxo5C^#28X$11)V=_92
z5qZ2F<!t)hMrV97D}FXUsT*o)tS>$sS|(P&QecSY%a9C_BhZ#h$#-ksIZej=%ZC#L
zBle$P?==NC!{Cn+z^q2Uqg<RGkBta2rF|)>(Og5MJyZ3*{EDH2rrR$=e6ZYKW0c`Y
z*GAc)-?nrxq28=amF9OlUP2P#xq;m9uVRq!yE`4530j_S(|V|#C<q`z2DlHKiJ$!E
zp_vPG%kRFjp7U<dp+N%c9u)R&aA4kv`sBsee7;`xRzur7l&s!#?XX(`G%t^zmw1D~
za{L_BuUg^*)dmM1_0n4P$9{iB9?LwC9r@fqD~X;W3iYsAo$U7^pff{(N^gGx80gEZ
z<Eb0?X|P`;XwLz&u@E9RR`kiZD&Fz-E9Ep2_1>TB)BT8e(A+?Gi61NFfMWF_@h2Kd
zi61ZTrncPmQjcH-;S~J51}b`%4?s%H<vstUU=l`nawRRSE!_tPP)=pt8H17&ozQV^
z?h$B-KWA?2cs8`TjLD+RnHPchoc#~Zps7f7kuSIsZQBPdj5q5Q*|(AS5q)HrmG}4y
zOCVlwAqaeiiz>@-I(hscwdJo9tqi^EPmZP~EE+(X0k3A+2%)Js5cN~pFV*(&Oe(yF
z0LkGC%ZTO&2XF`g%3OXCVR-cfMqzkgFvht10yWxA*qYmj16IKBFRdHGa)JS4lji<y
zGITcu?fSDTHl7z0!e^}=tb96p!My$tqZiqYSuCm7yy_->_s4DjFRsrvoreh~aU5Vw
zGDO|3dPyT#`WJnGlEe{{LxQ`3fMDnkJ!x2Cq0)pY-;iUjhuwErt6nG6Z9E4>cMH$i
z+<UY-*<<nG?Uq+cICgb?;rQ=P9!k&8TGbmn2Qq5|s`0XYHmJIGk9ptG;u7tgr$?*-
z0yk?>`MrCr+f$rnfOy@}mto8g2_-13i`e~(urUNeRE}UMk{eM{OT;f6G_J<_8zq;f
z?}FZNW%bKfTIb1)fs})Nl7i7GbGchh0|B<@`oHmcaZ<Cp$qiO@e&rM&M>xOO9T?x{
z^pAhm!4x@p;p$Cc)vc(k02%lS2^<I;1EoU+;T!<JECNl*;3ht_pEC2-dcXf>Vn_sE
z;+)Q*9AiHdO!_>>QrgW$<ta|JtW?z<h556<(xb*8fL2dVbDq8rB*_M$^!{kCvYBIo
z=m%?mKCud^Q8#SftE1OmyHBwHiKNL^a0N0~MbprIS$99eipv-wD7=p)h^es&+%2Mp
zK}Fh?M>edF3HlayJk8TqBQ@a#pCT|LtzIX_ef68kym9u|ZLwafOzCoa^tpYkbPndD
zIk|qj(08AydBF9tTUm3MX<T31^?QHR9PLA6@T)36tr;0v{2Y1i?_NvZ(I*+NOyA`V
z2tynrf&wgcz9G$jer8#Z#)@+Z1kXbkv3x)6WM&caTYXnnTgl?Oq7g~Y&>GSl;;DUr
zv#zPZ+Kp6XsD|)Tzses|-)gQHo>R)xn0q>niBDsa<QlY-89l69E{(;+Ar>AD$sXSW
z;B}Ds>og`SPCb}EumViScy2_eHSoERWe_ptApQ_ppu&JiROomOZy&fPf8Gs4E{-5L
zJjW0?!iFNJ-$S9=Lx@q#*PYUba$u1Dz1|Wkg*2ctBNXWFa-?qq`^K%V6^8}lGUl<&
z8%UVVgi;AZx7{vC<%9Le=bL7o`CQvRw>zh=K7@fk&?{T>i9=M7<pLESe=!ASLLA~S
z|6NHa*gH*#L}DRF(EN5=Eju^1+Pe+-Q&!vQ>|y0L{R<k!7+$5RNO-Idm4*5BGw1g#
zJfIak(u2hpV#PXhCAX4Ll~XhAW(7szhDos#4|?AfqNk{-oX-It%rk-wg4P!l)b+4P
zNKbSD7E|<Es?IT>5_Ptxoo_cNjh?ph!$ILRV$S$^r>pI`#FeDA51HM!)o)^!CzN90
zjA6I>OF`-5ER9$M-&C!T;H#JUoq%w!X?{y+8=^M1{UcdnCD-y~p>OIszs%B@91G~b
zf_;h;`N6!)t%u)McX<`5U&u1H_Fcy9xLp6xuden56?R<%`sQ!ahL8Up<ZrQb>v;6c
zU-V*VryM_GE+5^Zf~GdW3Cte@jXyXsht|~e4>YFP_b|rxxW$YSc@%vLLQfBKZvp@7
z2a%%7O&t6XcTI9wR5n17mc!*PamGOgQ01G<>3dU9=o?l~<LBq#z}E%X(~awM0Bx1i
zbaxsL^Jnjbg~V7e+O5=FysT9QvYgipzX-i1$h&V2Lo7WKvxRWcuVwizgyOh~TD_Ju
zns213-kuvq>Z*5koBt4}Bmb=FYjK?FS?T{gUfm539_~ro;fK#|HPsV%@H@S0JSDSc
z=cV!uz-Bi!;{FxUY2i+*dgwa}+sPAzL_H*l_5~la3D-Q9N?ReOz)k|z{^9Q9%%Lw1
zTrrLA75aWHNByj_gI<k*1pVIR2z0kb>5;hY-qPf~Vfk5V=vAMGZ1+nM?lXkIpz==J
zRg+G-PAJ$4b$64h_c@`@jxdjW8S1x-op2UzBD;nN;>1Bfh8%f(lrqJ9B5c^#-+7=!
zYoUazOzJYod(Cd}Ee4Y>`$tSINZrLVe~4qt;L&uDX;~l(6aaxjnK2Z-PRzgKo9E4q
z?iS5a$=$piz}HcG3*7PGKb*!YFc$&y5qtqdhdMiFHY9k*ACB<C{^~6*2&9HXb?a+|
z+++IUpbbnvjx`^QVVU=C?xBCJ2}{m93s)b^7QS}7*zWAJ{)tTYs_g-H{F4B1`sd|s
z-@X8y=<~r6(-^hUAea%mE*GpR6w@<?2EVA35pj~EA7#{if`29AyEd};J<b!;nSTH_
z+F>loTk%I%o~@uC8I(SMji(R9DvwtyKcT>%Zs4%*A{ZPr{(;gxS+}cK-(J+6P0ytF
z9}xVS!^y#q&}ZqQ0(UV1mM`R^AwoaE(IH6sS~ur+e18)~Dm`D|*W-fPL*0Rl^}tWW
z@LE1vfZrbFEsdgj$!bw}$>#j$lD$eQKaf}+m&We56GjfWkb7HmJsWgP1**6NH47dq
zzR6U71i*`RKT)fw0sYD*U<WoT3OzsPywSe^_(ivD07{PTmln?}+J42fHA}7VMm;}{
zctM4Xp!En<4bF+Q4~dC@{jSZV&n6R)SJaM+Zg)8?{BFmuqkNk{l+>&omQ5U7tG6r>
zTyr{ltfH+Ig+)({5Kfi%KzspKNzSActRE`dej9#Mv?T#Qe^<4-9V!VF8V`Ej)dv)X
z_?|b0uWZ)B1!YL~J<W7*JzTlR1BYJVin{VW{vgRkp1YO0?-63=LSE`I`R~h3^0&G`
zzFEfoRzY-8SEut^`Eca;yV7_xP_@JA59xQu5=Ml0@Ke&XwtCk%)=d^{uv&Zbe1Ky$
z>LbpuC$i5w?v#_-OV|tEfFiU?FnInftjS>5-8_3h=KHWQrX%m;u`Vw`|C{Pu*(#-6
z2B{SS|I;6k7m$Myn&r*tZg;RtN|ETd7Hj^Yz~2b=^Xno|%^$i$Js-G`3bXu;ED37w
zi-4}U&C{A9dy6Gd<%)qb$HA1bCojI)u<x^^`z!q3E^s&D3qJyaX@(L20b{-|_tIhz
z(#@!fYdvh-CB<l@{P2xe{e#OrgyIDbI&VO8l$;1?#^@C@y33GiN}_hzptheoPIRDZ
zr*YvqtKARt-nr__7V1O#$&Z0H=rO}8ZLV|+>c1DxJD~bFP&PH${J43R1vJdyiJPFB
zHuSRFY7+)Y959J9;Fa)qM7W0=5+FqT1l0b~{8)x;6Tlkll?f^6M)-^7S;t6y-|6Lo
z@#zI>vjr})kGu~?pLqIVYwX$A>TN=Y=coE|`(arLNp|n3fnRIdxHv&_=XNCYsO6H8
zdtjLkBzF+^iF#+6ur_#W15>*5KD+fqa6x@-$4Jvw=O`wp`u1{rdi)4&30^pApMwB?
z(bg9C=329NrpiV!yx_sT^z6k-37=k05c1RZ`Yl10rFV~yTp0YOt0g&m)!U&LU4#J9
zj9=w`0;}ize)GQ3ozqujDXz71xDw^{m)f%HMmDE?!PEWjA@oy*-WkZ0vd5fK@U5@l
z=ZE{ZTby}H+&c)nI}3i4JR?H_hj>U*AmIuM?A~1WPqFh!$omGswVP7j=;L<nzudt)
zPj=(it*3t*7*7lZwNNqDP4nt1>k#1C6(+-de`m^vOz{+>XpTV+?Pl$C&uY29y%?SE
zFc#(}q@``-(2u;|j^fPu^<2l|=@G>GDR$m!2o1mYq2>IT9o1$Wt>yj9;j91tOTzAl
z#N=paq$O(jnFRj~JKytBg0sq`)v^BrSb0&Q#;mx6q38=RE-mXUlfU?JP?Q#ce;U7$
zOo1~gadEZ1!k0!m&2Q4(e+KO7_|{J2dxA4VmGF${O=q;t_H1mr{LP6l-UXO}J`y|)
z!T;d9nRp3V(5I;TGLpOkm2ITr(kE*@sKJ<P8<m^F!qCZ0!9q@Z-V8h3_&95#fL>@K
z4@{qLH$HuH^di5*avSqL9B%fAi5^#u88g^fxM(3etFyhUs={f=WBGdTcEIjDx%O~v
zL=69ZrlCPQYRZq#qb|22Jj^e}09G2~>@jD5J~Z?gR!=2zalVYU|CXpS2>0f|x^pW1
zTzds0jXd_$iW!|DuO?h9)!X7eS-aWL&=zi4Sypk&xLfu(w!NYSoo6RY;rPsWM!KDw
zkL2pQcWN&QiH)R28%q#4G@f%sg%3u~=qX~0gK`cvg%l|AE!Tu|lN|oT*XwLr@<`Z2
z76W{FJ6@;G*I(0P!T?@9D~<W@r;w|wA3+)GNIZW(mOnrFWtg$LZ6@1-YrZ1*n3=7g
zYZBrB8-|=DCuT0?P|&WWc-8y9@Y6w7=|=2LuiCMSI>%*xGl4a~pKT-J99x=f23bzH
z1@BnXFwD@-0QIN>b&gAR!(l+;$TQ4*yn58lNidNB+R6%Ajes9+baXhG`a=ARC<kbt
zNEtt6tw2j4J}YhfJUPE^syDeO>7bb^KG_NaIixpeqC^v;`!)Ln3HXW<nJ9i@GOsr?
zH82`ujzw=wwwK$DmIGpq-o(jR=nAokWxf$eSh5!Z1iMQ%N4cX|z{VfF;FYK!kWiO8
znNKvnl!){07;mecHD@t*d0Diag(E^r=A?~_!;9bTyYr|9HavSwJmnf%-N94SFk4M8
z<WpUloi;O#85Lsm{rXpTcv;kcrmrR-;A*tGZ}#}Wph3}p>-yWA7A-E9$vt`_=FV}m
zU@teZ%VVh1;CETf#2<L=>l>UZuiKM0k=H75OZc5WPN0CtM=koosjQmW>tz`F3Cobm
zk<F_qQ#_w!kypt#TlO669NzDInisA(k0XP5oi4G}VK!-=Wb$N04WF2^^|7&G2#JB)
zqo<yY-!v1hr>bH^;0rnS?_5e-g{W@mP2kBqVixX-$_qZF8%9(Qz-PdKIm{Y8H#q65
z=x%7&sAm-Kbo(~LrSKe{Qc%0sH^L3z@r}FDjfSNJi;oKEBM}f+_Yj(2YUyr}zXimr
zpR$=CUk@-qeb+(LigfvXU9I6*03}6D6CgPjIQGz%(*`16D{%0uOu17~usE)>9o}@<
zv_9}|3w#<AdB{$0+l=qHRbRAu3SmJN3Kz@axA+ZdWZBUN+CmO=oC^>~2&hBeUqnRx
z``^Yzs^j5EpFzm?BpK{Q5c~Tez|VD|AUPW)I7=@y_CS-V=NK~BtS4S4KR()zno2{)
z=;Nm(Lemfe6DH1Tv3gj9g-+A_12=cIBW8zF;PLd(qi0D^ekpBRqChQ?O53+>g+a*u
zK@lKl87&ZwGiT&9G$V}XF!LXpe*q27rK@v36Y^Wh9+fGZfq~6RK~Kxm_v9yIXFuYl
zrk#?4FgBVB18%d!#%R9mxV@w?EF?-{)ne7&gcTNr(<|kI%=q}v3f}PBJE2EviglXt
zL4eLVjXlDCEE%0V#k>E$(s?+(ZPU-XVZveULcPW;B)Cs>-j~<!VQ6sQzk8W6IqL1N
zL`S|d73o}YQBh}Fnp?M|OYHW&Fs`gx|7+(%vQJT6g{Y6@`>Ej(AsQ;??_@f&=6>U+
z+w?K7STdOObI;{uM{xe+Me6pWe>XFy@ef)Qthdau(aRf?E$ClBZ&C@39{0w>dH6Pe
z&^YI9|Mhm8b=&8;*5Bz3ODQZuup=Q4s+v;rqwuYJ0tNmJ_1ix&9h+)+)D+A=+0=YB
za$+8#6u7(OWpW*Mv~C1t0|ajb@fL_dT{rY9K}0hV`p3&j&~+&J+x8X0*>xy;+iz=R
zMlfJ|Yw3B_eGCH`kz-!ZC|r)sJ<h6vZrHPBWIOBk2j4#RYr8Zae4Q3w3t@i@mQ?Y$
zvFm+%WJwq7KIgN$Oe71<)gZ}%^-^|$m;(G!rCt4edgkabp8D6F+Jh~=sLv1(UlU$X
z8oKvxfybi|c32;2w>B4m7s*rkok+?=hSBFvA+#uW7<(rByL6J5%uC`w$qHcE?{~{7
zY%zbYK|JQbA5NXHI{%8lO?*%MwvEut&FSGCO~?l9ovjG68J@kM{6$a0)YJdk@t)G>
z0EfXJ5>Y`v>|yj6j-E>O$7kfsr}EmI<)PWpP*1_PPCw_k^?VMX!a^dOsPq5n$3<mi
z>|k>5GPG=n@4F2)xRdiW$$dK*QBQ<F3x*3Zy%LY`e8tk=pa9pzb)Ry}Iy!>T*Qv@!
z)rahcXYywYOiE=ac}S4#YlP7$N{S1~-}H-D&VO2EZeJHu&KG=M$ft=%5dMab7>YYV
zWcZ-(IB+r3a&&2Ho#j-)d^RA%yfq*f-qI5Obl`X^Too4mC*SmTyx&$I!OI{&n1VT8
zxBJar*BE3eZ=aR8ny07Q@&^B`2jY4JWND5NaA9|Th{bY2O7$Cj`%5#Wnc&-#;%(E<
z!N+d2Xh{EyhHVv*^+=&=?J^Hb_V)Q;bg<>z7t@X7$aY_F;DGTb#6lEIm#Jw-4_3XC
zy{Va?QJ&NHn)h?=6-I|iq*s~ZUDViZUEuM2y%43@dOrEwK9zTBQU}UU!w=ORkX@p=
zNi`d`_aAm>y`e?(JO+5#5&D&Zlz6Bif%mxe7jT`CcD~kcbUjnfo(k)tm(l?}dwf6`
zgnFrUcxs^Asr$jq(`dX0Qb02uzD-v4v^%7X3k6WV*~g>vwBj=ihvsG+-7&tmWgY9*
z_p*6Y+fOI!#{(=!fn~V3TrRJ)uIUEQGF*#Y+YS!nMa#^-8Ar5`+HM*mjUpMnM|ih<
zYR}l!9lA`kJyBmxYjW0-8Y9|M?EwUIv>t@S%Q6GDbS(nl(fQ`6x^i$nT^tLGVJG|N
zuNr>&tkzvwWlX~GaKZ#EhO%a;tI`*J@)CLAemSHEn}$?(&h!nH&c2(&lqkU|PYt)U
zMs-!cqeHD1aOG02R}im<dwjtYVeSn6LwIb~$5NU%4VGHn)CU>G6_qyVU<dG3(4^<#
zXeC_y>78dQ4ZqFU!yi3+2KQ+QlWI73zy%gbpjed_SCD~IzYjwl)(dp~yQX?fU*#g}
z?Y2KWac3X$@^U*wU^a>wjb6Le;`OzcM)F4_*!+Rf{|4+&lkb*Geh}|VV0?L+F<#KA
z%xn~jAez2&C>2fbtgJD1Y-)LtNec85E4Zz4GDi>C%1lLC;uitWwS3$kCG|4(mE4M%
zDqECZD%#vWzfT`m&8wxM%=)tM4zTanw|NpGBv$lk-=>JySRppbfXC%h+iTw_Odd0F
z@bZ7P2-#t{Kh}dDg^CU!yL=iz*x@K4;HgZlQ(a}!M4bz8Hgfs3MiTBJTk%&eqDH=7
zQ_TJ~%vhj7@{PJR6!-vFiol-4GX2Y=Hi3x@vk^@yl3Qi2hPZOuE`XRMQiITY5u8=u
zFH#F3sD%fEZlgn^=Vpq{y6?^EDwB{FGIch_grAKT_FAe@@PVJJK|j@yj|mUL4bqbm
z3?jrAc1v)V#lj6CD$T#Duvb!6HI^ZCCHG?|Qcxx@06%?QL)(B}x0-HQ{R<JypMgc5
zT2-B*PCSt?ADj*JeI?x-T7Fb_>oQd|e~#0O6-FL~o`iI+&JOA1IpJ(S?!>)q<j*gS
zEz(9*4kS2OJRGv7_u2Gv`+lR01iHw(n47YhNT1`L?O!(vW(%%NyOudryFf+wGK!6)
zgPx-UgCE-IMvibkM}I-8WMT_U&npjcWx<CaL!h{HLhe087f`_Ln#&+bpR{We&z<-S
z2^5!)C@CN+;Jr@pt?up@Eq~UZL?a;$qiy7QfEN`ChxrZR2H~Wd`A_1l<?pVjCYd*r
zmA~FDKkLEfF*3Q4I>JCi$_u4IEQ?zT`<iYDesr0a&+P;cMfF@V9eAm6umz*Twq=w;
z-g-K%PJ5cU(i4NBI_p|h*8*;|rc2;S?+2XNEo%*?N%+c3^e=>DBcjd>Ox47$^v|gW
zIAu;})j$YbUeD-Y0VQ0Y7P5-~D{mv&QOtm|J;4r9{kZ~zE;m)#{>GORzR#)4MtxWx
z9{e{AdJRtbU5XNe%XC8jwULX_7|%?zR%V9U3+d0T&l2@GQo}8@K6mCCc&+jda$ek*
zKi`OKr;SIAzGX4X;h~0Cl5dY^a#MVXtA0VIQ4uW@9M=`~#+1(!j~k^k-|Im!#-$24
z?|+sp2{)bg?ZoWK?AH@MR);Ez#KXhm@!1}wN=F8lA*+rma{4Gbo^eeDi1bGBzBQ*{
zvkKxJu;`ifQ1NxU2SywSTHgZrd`O&ef?h3n_?Kw*;${6t3lz^;@T7f9U)a`zjg{BD
zF`Li$;065!jxL7aGqQR0PTDAcTMF+BJd>~A7u)gNlq<wV;~wbs&-pMs{GLu0{0u~E
zqyD?=U`~~n%!AidqP73y$M((7pG9-OdKCu`rpVkay6z5tQI0w{yrk!!(2W_V=29pS
z%?Zs)C@h^RCSaZ4-L<<^Be!QB?~6y8gXNg8DER{@W=CF@hIBwXX#DCzkgL%;kZq7(
zf3_P!G*%@sP1I42f%n3+>h@m@D`D?(AchC78N~|j>L^{U&I_K2<y0z$+MAz}`$|}d
zT^7t=X}ONrJrNPXy{Aq`_Q97vN5OD<dRX&qMc|+euc(@ts^TxPTKwBJQ_BWXuuWS}
zwd9N%^NMb5PwOvUqpPhO1Tl+4*q|%*h7r{VQA840^ANr5;7K;8WZdMWXPSI>;af8t
z^S|_cj2wpcxyqb{hm&>O`c9z5gE}YHoyzzKnqoI>wk4Y#601)2>LAj?HYt-&zZpV@
zQ<GuTjpid7A3SiGA?QDSEzb9xij`pQu<4HJ>91eOj|A4s^Y)owpm4i?wd&_Nh>TtQ
zU3g*aSq^<<!YFYc_@*5Y^q5Z3B3PeTYJ~(t28cd0ab)9-PCq3)6cZ|bH+XPjXa^RJ
zNDAfCn~>p}>_S$FZ>+z5<-8mYAlb5;h5s(?imewhrZ{XgNE*pWjKwfCHMfbN?r#k_
zGvP3dE193Tw6WeKm;>?zr*#K^?_qm@dBUy>)!^N>f(5z-Sod4yNT7E{VK)U*SWOSY
z6hgXVvK9WTB=Zdf&@O}X9)1=yAlb>5whqB@UIbut)Rr^eLiALJO6H@SIXqX&kpW@z
zeMxS;)gv@5*%XCu+uI{qLNMB(P?*q9IB@2~qVAA~cjf0|P<R@sVe5F-LvCmc7&~?*
z96h{hdcL=v=zZS2w7hQZ_YPX;e5h${`m8uISTVRHdJa{Jhg2vBf+d>x>juV-pRioC
za)0uYad}|7=rA#LjGbo3%yrv3XOJz=?<CcvQYl)D2`Tv1i{n^f6gQvI@0=w)=kbI@
z5{r&A0TxS2V!#<kS|S>>P_@vb!HerzxLQj<NbC2|`48fUB=8HW&Y71<Nzr<_1c07K
zUo!<vKUAq1kiS`!$spjDDA|oBOaD&v6+AuS&L6y`1;|9}c3cTS;&10<^^sk8s0u&5
z6+cgQpl}3!=FIJG>{RV?{D#b1MxB<LerF~jr$34IIq)ElPLI#*e%CG6A-9i+<6F5M
zNQ-S7@W{Hh;zv-isI^`5duQYFS#ke^x4ZUJSJQY?L>?XukFK)+-}iJ|`_PXy*P!c-
z>PhbDajsT%cfS7iji>X)jf|IJxKea>>l=*nYfRMn5U~$-P55=Q?;UFq^H?BAXzTMI
z>4Y6Q<k!&L-dz86h$IX{6Q&;ipR1c57cRog4!;q@6mWdM>v!9Kht(?^Qk3?sx(iAm
ziOWIxqIf26F1=-qS0u~%lJ+E*%9G8eCb#ZKxB4&=oydD=ccruZ`Fl}@y+n^av^fGw
z9lS%<kD6@=QRLRJ`ZCp|U*mmnFrY2YN6DvZ0b#WL)_Ehya6pG?+eX7@o93V&hWSLm
z>2iRv6nsKnS|h6r&|(${hHyfKGM_<!P?QrXVEr12e*J}d!-BnKMTShicXJbM-=yak
zjnN5C;JeOGO<q#o^*|o2#n&O=(c7yV5#vGcQ%>uVyVD~2<$^l^N>i1US>SGNNGl45
zHDwdh*sPx>`5Cu%loSDPqF{c(tZCwGiZUyT@JPX$rWcTt&qq`hq*K0EBTqDcMR70r
zvf+jcWEkw9eS;+honO_w0jE_IR=8SrU|dGzPr>)<U8<=l77pB=cu*&rVXv-_1s@Gt
z>QV0I2yQq(&{CHe&k+x{%%_I^9@Bf1_p{$Pu%$64Xyg&jMDf>N{OZB-i|OXL=v=9g
znLqkWPokq5PL<oq3cpRyT?QTtU*;=5g?MUpvx7K?voWQZoD2La5J28w1&<i5EKhE@
z&R6MR@oo=BtrQ*xbL22U3Oou5#(?)%rhq+NWFYu-hIe?s{tUy|xIL>uySC@|baHkS
z3ErSne?cU2DZKp}r7`euqt?jF%GvIG;>7g%fO|WNk-oQ__qK>Qq9e7fpOxN%=M=8R
z@2Xydw;vO@c?6|x@n-O{*m{u)Pub_Zw`H*d@3O_N2|0=QAdq{WNbIdY(_U8lg>Y+I
zTaM6zTp_~FIv`#z|IT*my|jw0BX0b7!mz0k%*$Xq0xord$nR9@CUquPp)*OlPmzRs
z&^bwaWaUxv=plBuN3UFKwk4NJEK`Ykk@$%dGFLNMk%KU>6W7YvQ&=B?d__b#oWhG!
zlRtYx{TN30NvzGfCzhU0d-wCz{Xx9BW9mpFg!1vFf~Lk+B2?L8UIo<@mR7b(N0sPi
z_C)zRC#xewCAlAcB+u|ydL=kJ)=7}eAA<JfGh>b3#nE8MQ0b5GGo%;bJ^wsrlU}I}
zjq|jNydJLDY}SP-mNh%osh<1V<Cw_k?!^xMA6ly-0v{9@rZRR>3?N3U2p9Tl&VAFY
zu#CDW-9YB~s5apst&5b}7~Npn#!SWXU`nE5hK8S?J$s4pc5C|2%;fBF{wp>cvhs!B
zXf6IVHIz(wzl)`_W~L+msu@^b3G>l4u2Sf!9eY(65!FLF!=#S&hku@1%i~p@m;K&I
z7*SB-<|A2<iYbJ#=2dhYL6ssy5|1lGHioDfX5`U|Hl&~IOH6hq5vVYel#5l^9!^4`
zkT$!*V)lq<HdCxJ6`D2^3eJWy4m~Y96?8UF2sT}8xuVhA3?ygW*qwW17H?z_cb0uZ
zOKd8>75X5eJ<2cH-Pf=@TY^UC(!T6j(Dj2%^5ph}RD-EtoOqI{_O_(z+}P)ddHp%M
z?|dQrai1=~<K^~#)=*n?OK>$+2SK(MJ1=#YUyS{t{RgZn%8Xuvv8nu=%Ic)b$OFyj
z{DG{k+6tf->=pQ!IlDu4GEV?GaLmCzY!9KeX`d8PSvX!?JWlM=2BuRW9j?4gnpVz{
z=x3Ef83Klt)E|s_a*G5~oH}rebu+8S?|g5BHPwH+4iE^wqa1y-OT8&9pI>1Bn1`9)
ztSjYyB@oLPsj9pf(LzG*J;!y6=O2-bDQU?>(h}SdJkDMojhM}4Ry2Wzoh8P{(CGSn
zezc+8U>4DY^<Vs`A7VzxJw8_Ai7Q@kPcIk?KUjkO<d#}8CeKUVZ%uwY#Ps2yCV7jr
z-kc@fTf7@I$cc3UB^T9FsWC;9<V4=g`g4?j0i`gTWr=IBH4vwSR^wblAh9v1I+{H0
z?(XSJkt3&L2(I026z^&ymNNFD;^FOr<}H00)O+kDHs(F$8IAPU!QA|v1_xTZ)Xg$J
zGUl-{VB|?}2udj_h+X@gP5@r3%(eL>@K3I@mu`OUg~!}}Lnh396i8yul_zjdMKzS8
zXYr{t>yBg79YO^19U&77(8B(dV?k!3dth6IGuQB;9@T^r>58nrtfdgLQa{36b222&
zN#B-%K4sF{DLOReo`^psDIsHMOW!-26ZwX@&6>PeE{CpzjX#o)j#)jujT9hk^11tp
z1#QRgfu}wxKzcl>!@u-^_cqCDSoUPe-sR`_(SY#CuG`!>z4{uJ1sP&;eB64R0d2A{
zC-)d~q7jn2xb_7=hn<yfYZ`58ZdTkkPrW4FHpPKo0CG1hph8t_-O9t0YRDmG+i2pF
zet&Q>`G<&MM@k1}wE5o!@+Kz{;<EoOh^MJE!UoN>{ZS^&%h7<Sbxx3z9uV{w7yul1
zY8rrCA?#J*kqId-hQi>fuS*|6OrK)=(K=5k{j+(&t3H5PGC*ZL4MJ&R+cg>)4*gNy
zq25^}skp`6KoGPN`IoTsdY-yyljMd%esN>MZE6#;-sX@oydQ|)Pa3TVa>t_LdGt!p
z(+K(dSz2C_C4V0a{ApOTRlV3%ClI-@2@jDWb1|dX=-LiiE76-c=D($JaJtN+OL>D#
z47nU*>!?M8e`JQ=g^~0@XIGFtJ>gz5M$#gTkabyqq^DmNC)Frt*Ps;bq0uI&|M}6_
zWmG~I4fDwU9KY?)h0YCJ(+mXBM>k^cYvKkgTs8LTNOl4XK_@j&Jm=SIR!f`s`+MY2
zOoMV1*@Y@xX*t%Z^Vft$Axfh56{$?w;k^pKd#al>>+7p=%oX)E43$n19nrh}xVj1G
zeJ~i6E^(n01pNoY8Rc2_<OX|=veJV)-!2u%5kT<ZMIMTBp)P39*s$WeuABl?eV1)e
zHItTDml(Gc2L|cF-@`=@1f27%W40;)>Mo_N&RrR%_)*<S)`S~v>3P@Tcj3y>^q=!6
z?O5F@dNqt9f<T6-Gh{;zy0mNbJ3nr*0c4Lvz9g;UPBP`}kV3D>)26-cGv49WV>xoC
zrqpIQ<Op`33Ucj?u4>1s!ekhZ>SIRtA;#2sUW*)^PZ`<E)h<K=aEXVuq!#eOpnoR>
z15^4mcHY215}#=V!oR~8v$9n4ZR68bu2%WO(`F=CwVJlhM@KtL$)`T7{CL!;S(M-d
z>tbKr(3jGw{zNt4Gl|`U&RlM?Qcq%ZZ_d+m$Bn*{9dH>}(DlTpa22qgpi?bcfkStb
z;JX{K%vDx3u_m5G^pbUJC;T3Z=!=?%pF{o;t;7luV$oqVec<*%pkKL5-J2EP7d_z2
zKtrcncdPoU)NxIDRs<nMK!#7>citTYM!?t+@wGjl|C3+S7Y9N`pE|f^ZsA!_PpE}b
zdm*`#+qK}|KT1=okksD4K;4>jJby2W>_{d#QpBFic)}-GDJ(lB3#jLa(246YFmmGW
zb48WYUfAc>?3<X5W)%*@;PI3AKHdM}rvGpa&Kn@j&_YjUi#l=GjbbMv?^|N3xBTg+
zJu&7_+dKA|0+P>qyYvfoj%~`b)YWzReaxsF-$9O5VJGpks!X#aKPe327Ni2*<#^cq
zZGTwVOpPf*eMUkmyUn7=RRz8kuR+>P)Ryj>fETTY<g}xcd(Zghm+z%<lNCVv#)yRm
z?)JAI-GqXrHy~d5_rA^XP4UuFv#0^DT1HD_VOZHnTqlZXal&S52{C%OA;Qj35hb)+
zkyk?UV}~H7bK%C4)(<|$^hTCPu?96EGE~bb;bClPS{fC4&44<{Q~3CamR~U#H1=!*
z0wMUNNzw{WN&^hXreP7TK(Bfmn-N}skaxvbAY+p3lAMPTb@2Q>L{8WWe*KuYy%NvB
z7t?)B1&Aw9QNRJn<vepL<`A_apU*VgP1C})F~M(suVth(Til$C$|4TKtMpCdqB<<v
zmFSvJ^Yzxjts=>~(1T^GQA5d7q4f*9uAf?bKxwUt#`1SLa%zd#tQypKiq-a_G{P_f
zR{=tK=6IROMw!53yykQ!O^bVkL)F<ww)TS?0c&opAT4G-@o#XBruc|nSK=8%sb%)O
zAnrR6h*nmP&`qTv?TV&|TL$zT8&t@263>lW4<ikJ(v|lT4@&*`H{-qebtp$l%_R@n
zO6rIk{GxXDiBi&dyqZ8LR0RyS2%;*#7eg>gQA;*9MotrbsA^OM{0d%St83nY?+(gU
zWaov(YVyW!jOcS1*o*ctRg@UIdMlq556qCa@rF$kg1i(Xh!RQLYy2I#0!5hV)MtI9
z1iIIKxn2$m2^u|jyStX)6lq}BVSMYcy7|YTRRCXAHK@3!co2kMakl|zsp#aLv3bPG
z&b($_??9r6rCQ*DkC=0E3N<dB;iXCDX)zI+_Q+aM+`v&eBx9WBbqp#yT#-$caX*rv
z*t)0})3!t-gckHq3zNdJO<G2fKIW>bg(L6_7nx$`T@_afd>;y+T1PTt)+#m<zF7<b
zE341MPX)#upV%2by(cU*eEvy$pRCDA;bDYCvFGrWY(}XaL2<a9lZT19kMPaSi(`gG
zm?3B`#^64g5j!ggpqNvF&NOMKg71hf^NfjwKAov|!=^)yZ(W)mdtGKya9ybM%LWdk
zG;}fMyR0h)dYM9Vl05Ok&0y@Qq<P?p#+B5KbS@y8?38}YcK*>;9DY5yq~-^M!4=iE
znGIC5CB@q^AT!y-Z2y9mhNwcz@L_XF6Rt}NMdk|58`{0FsAB=s0=C^oybhIGsInfN
z!%+EH*>+Us*2xb%@y@rrW@IARc}!qK-ss(boTmf9>m+A7YaR>d0RVE!bHC)d2*ohA
zis+$JzAF?qMlQiSTg8x{53X$-&8h5^^RW(vJ7}+wB8?M2SI%<8|J#B*O>V^)ek6y_
zsO4@Df>Wm*eAkZelgLZC$wZ27-fujfEmsW4P-HfLa`Qg`NI<v0lGPADZJ#au2-tnI
zsC6Lkj;`FijV_9@qM7&{4^;Wkm|%0GnwPiqR;8g}X7)8jK#?|6W4uYiM0X7SaMb<M
zoNJRpnsOP|tFMK=@{gSFfQu<i^s*Rg1pE7Vib0}jiHT{94rz&Rz9WW+o-;)qvD&JK
z>0Rf@q%XehtXDpy32^XbYvm86WKnBn5JmmsEWHoK`~><QQxqkFXB!VxeCo4#vZ*i;
zyA6d)VR3SKB9;?}?=h)HV;`msnt+vIr3MHOu*n3hP$9yKt`a42uCT)dEPRc@`gEbJ
zbGF??A!7X`r}h<;r~+$NMM?Oj^7eJ}QcBW2X`-=4<UG;NYMsoI0T=eMMrr8|efozD
zT4#y0X89X*rDuQzxxr|5Gk<GFamieZTFhw<SshEtBO~POu1>01(x|&hU9M;;rFwO>
zt|eqOswhwSi7<6qsS37Wm<*j(t74qjvY;zPt0dZxx0gW??Q5MVriw69q##P}gfJIO
zB$RSQojaorBce<r@|Jr+37guYq=W^vz>mzjj?Rh%*=w6*q8S@4GZDoG*=R2e6iVx%
z2C`ll{Xh>O#iH9H6l;iatM28#<JQObqPfBwQ0r^9sibI*abGueds<ilADJE6eZ?&6
zR8sWE0IKmQl-4~n$nZfaml<@#TU`8S>kLC&Lsv{;5H2_|zC#uRy0<@Z4$4lB8cu<;
zn=-mXPBS}$jL#%bO=o-t?y{4Pvi}P{lYB+)U1T&0pL8A>ugUOr#%qgf8GP3|=QUYA
zWp~bOKoy92tx(j*-5nKMNz1&D61WqB9D=i_yZWUr>^CI>jpn2H#?~<}7c84?C{V}b
zO`(->fyxbq(wcNwL-Evjn6ZRqb-F-!@P!uxe0zti2zT{+cY!`GD$Z1?&b^y+(0j?R
z4lhv8`)K;m&)-oNt+U7Uqjh@e{`QdRrxvTyfXAXw?-VWEE5D;m^?M<yoyWP+H@qv0
zaBz~mlV(uD$g|Pg4L<H>b@a#^-xz9dwU(Y#KbTg@q2UVD&?^z~%9V!3Mn*j3pDU!n
z!d!5%R#D5ZuT;yARvA;kf~RsZtLY~~^Yp1y*DoRsXhdjQ(k4-Or2Ir;EB3UNSKTg?
za)Pz&YH-e!OlV<2etM~WCZQ?*)oOfwyDTqdl<HZ}VW3Q#^}4BC1x@W*Z62+9W&DVC
z>0IsOdMW<&4d=vOP|y<FWw1C_ff|0CJ#FX+pYEcaLK&f{j1k@;EQQhLk-SnUrbR3*
z+Z*!s>TCJJX_t}uNEu%dqR8u%UsX@`Tg&HUFq3zbyA^Mvl~81*?=#Pw;tKa+yO?B#
z3#-rJsmw%EEA(?qPUC#U)Uy|-O&>dE`ZRnunl?hJ&e4Y?nvsOFr>X4_6l0;q95|m&
zGvJh%o+fK=MriD`sV|P5HvMnrxLQwjv*a&M&i^IvGM(vF)-E?UKJO%Xi`C~<Do~VP
zfY$PqIXQlu%<aasHg_A5=`b5ljpsj=_lh~|m9RG+{)x5%KaB^Wb<77aI}B_Ws^Miq
zgztOi_Mbx{<!F%D^GKkWsLPDG=g^P7t}zBNB4p8P{ek|9oYPF~G%0)v){h5TM#d_T
zaFk<X`SCB11&S4z)DlOE;VyG>ysbn8ekyxBAgh^9^-CvNhf^zyr}GXh85{1gQbgV4
zY&0vS&gRwl$##B=d1>`gPcgLgxKz7zi&V{24ql)oOr?^hOs24WYXa|c%6W;QR}Wy3
zSp3AEBwP!B{0`R2Y2!5GnsXOE6X&5Z?<L}WDEU=cfIIYHpna@}&KE&4i)Cl9u-K;o
z9n?>eGQvSqyW%M>ru1+t6-CU7#zR&!-pp<+SB~d$NV55N%KGt=?O3WJ4_SRy`lgY_
zo2s<-y;fZT(oCgVrE~;E!BzzfuC&9OWxSUN3yc$eNhrt+9)-uu5~d0(^U1J*CsG*f
z9R+-fdf3)sG`Aazu#x6hkC?WDR!ER>NK$!Pst8!MVr-j4FSB_>pzu6JqV9WJE^iwm
zho(vZ6%v^dg6CPLK1@9LSRR?mozoP2C*TPi8^1r09_u}MPV{2P=?~4SJ{7o6pOS!#
zMpD|QZI4*Xp2c9A)v{Om#BJ8JZ)~i)al=Mkdwxf5Mn*0=Vx_iy+ZB`z8`SkQC&Y#7
zcM4;%IUb;j=vEE8#t*gcHzYc>zP{npG5Tlv`9!Dk!Ltt;PwYR_v%)?lo(mM~xTmm~
zj}%%fgVu3Zw-cm;l&v}NeXo)$t8j-VOqQi;%-5*#Gr2hjI%rH04TiM1N;kKeBSjgz
zFvjIOFyXkVOUJJ$8SayehQ-A#QceQN&3c~rSlhz{sYQ)g5FwCmOY+{n^vVo505A6I
z-@uTcitiQtB&?&R!}UdVn%=J&#Gtz+uMGrlCmIQWB}dS&;Cp62Jf*1HWyjO_QFqlK
z>4pw;(sN|h4r+J3KDqqXrD+Kr*YAp0=?40uJFf4|v{}VV%#+tfQW$(m2SwX3bRk#X
z1Xow`<r#p@uGurYmNhTt<!E`aK1nZ0GPbKcrkOxQT!r=WV@y~Id#n^8<5^xlm77CU
z{;D9W8Q=d+c0NxGMcnKEyA*vXNBqA^)W`386e9MD8PSz)F$BZ^qGJq5AfCalwvBmu
z!sIs`k3z&+@P1vX2CpTQ@4O;>KMYP6s={ZP;0VO@VSQ{AohpePR_ta2##G*Fl3Lox
z!2|eRj6%ek@~#yAc8_+QhFWt&lx5}d^B66~^L00W5aSmzie&Kww70_qR`Xq}jOK8a
zx;Qi|S~sY;w!>l%vz{zE1hC#^EE_;HUQ7MU#CyV>TpKQ27-?fBoVt3IHZ>C}SU<B-
zteIJ3BYaFdnSuCR?$4yL8Tbo|H8nN+Dz}{0e@%i1sV=1g9yGKJ9;8>}B|eD+Zm3v|
zSp3ae`i-9Rz?UhsXUWWrp{wXO5%4m7ds)2h;N|K(DSSO1m^@N+OeIL9dG(6w(yXMk
ztRk>F)X^}RzI&XMuO43RtOzc%_f7pf(<7Eom)|qeYl`B4QXq*ia8#*0T`aDK0C13I
zz4`e^rO#geby^zDbfW@)G3$vlpNpYyd+T}Ex<tDyu_o7cu{H~d6tfHy6KiV6an9m9
zM|$Nrf$F^aeCUN=G^RI<3K;YFIHVJ6^I{d(n0Ke*0Bgnr1P}ag)uy_=s2b~#in_f*
z1fSHlo$V{f_H}+`7Ez6*ZN>Au^Yhyr7+%0CUPgB_>xgZ{09Klop_{BGGqF4vqhPUE
zlo=Sa`3UsT8gHP7%9p;9H-u!q5+zaH(oC}y?urpnCLNN>0AmUM{-GFSWmB6M9HeY>
zu^iyk_L=2)t|NRjIwroeZM^Q0#cRGD&yjgD__#K2%e@v(=btKop7;rCw%14Sx4hXI
zL$Q^|E@5?7K4U0{pRaPDajq&NmrzdQ!AgEU@I|cNChCSl+P1c(kbft#HtBkGr5NP_
zP$UX%`!lP5?5U`KjLK{lKnBLsC<VpaIy1<Dof#UUci7_0&|0GWA@z2ZnnalDNjY8@
zPj{dLPb)!-Q4ftMw}!voI-sV8YoKrQ`B(y9!grhp`PyA_x{Cg1){CGgNsA&b*q%C<
zLg2Y;?^}NcaMK3>e4YIHF5oFVICK=<|DP-9{x^qeW&_WoTLAY}QqosQ*iwrc<7#RW
zfct@F;D&BB??;UZHUD4Mx2C<M)UkLT7itStP|8uP9<Gi>ca}uE`E?&Ts{Ab4QHy?6
zcjxpdFo9eh4g0|*+9`?ZM8$3Pe7dHXRv3`)17ASu8atAZT)p)>!I29k_rZii=qLy|
z#8*99BkW~2LSW)LawCZM!Q;p3vVZ^E{6bhaf9^WuCEFB;KIS0SNmUvY;N3eF)m*X_
zYl2w+)7+FG+Y;8A=Cd>Ff`XbWs?ol7txyQhx9)D=5XBNc7Jp@ePaR17&^K(r$71X(
zKx_vip3@9Qnl6kytilL<S~9J)ZDIBW3@5WM4C=_fpn#rBkZcT_^;4$k4O6GW&#j?i
zwSK7?TAYYeMQ7`qVG7eNi*X3I+oRo1Vz$JzBLV9rQjX8_K&G)$9_~Skg}aNT?7UeE
zS{i1hg6SuKixav~c`mJCIl2!V=jN?m@|hh7?X@vJXHNs@p|1LA7rJ5BQ1<qzw`=m|
zBxmJ-F_Y(fU2KfzDsO@QO8b&j&yVu1WW7nmd@EcrtLL{|VWF|C^+4-vhz6xXQcgQo
zQ6(YHtkeV155N)LVCu1u$Xzt`Hpl<gEm84r-BgNy<J~~E#+63xjg($!WMdLN^A^wb
z^bH>s7|!n)5RR2h^z1~R!0;(-+*~{^6OZdoV@EK*ih*k&6OWtc={sfV6!jL5aP$bk
z9xwThnlgoqX>$dc_#IB<zl|{^1IQ57b1(Uf4hv(tiW|H^;d8Tu(9;&YV4i2*q681$
zu+gKYn8QAb{h`MM9|94XAL9WVDITsCn5b=l2<Cd`B_|9T+781}J|XlEFAt%ULxL}C
zr1-j8VBCed%_uS2D}?^x<-wFJbWgW)<JI-b*9K{p&W%UE?n8HhZXcgkP8xTVgK3EL
z;2=ed$@fljWDy&}hPjh*pqDt7jt9NCA5h<NbP!A{2RGR?v==>sVekRbmpuo1(H9oJ
z1|pI+Y)DGpxDjpm+XnpqnA4WE-Q)ME5Vgy>t5iBg6zu?6o6%YJ>n0r&w!bqYa`_CS
z*g(Nqsz@(3qMk1I?$#1}z4k|ZzO{4}(hb|6*xXbO8qy$T1i3X-R9PZdq4c-BgGTJ1
z0*K;HZ`eWdSMca3kITy+!}x8+&~4z|s8>RwqeFtZMLQpV^2uY)^V5|pcJ5rUQrH%?
zekT66D7Y5?>#MH@hp06ORaOF6;f$r=sIob8%H|36?5{rmT>KUN&(wKyx6hjg%I3{0
zn~!OkpbCRqY0m|z`--XUilW&H>obABtz_Q@nzdyJQ#N4$=p5eXzG6Q;j6ck!T^2~u
zYM=^iyQi7J^c$AKKMZ=~_njh#U!`@|y7wYk{OXbeqtO>vHlm~NJbUFzkMyHL@kKD@
z#VY-+jTg}!S6HMMcDsT;7q{L6M<VxLn!o?wMdgJL|1^(g@1wT!<$pm>H;I1ZKQKHl
zET)Z^0kIB2jDO<G(t#-vB!Fu)+6lIx<c6#Gf6%+uXcTUOgaw6#3t;exoi}gpJW&H_
zS6*0(JPN<dr?!=nHe`2My${O=QpL<uVc(>d6jKdr6=~;0yecD|up+0uwDEpH!TpWJ
z_w)1bKb<beOP@z~j@5x)CvuH*)%)^)N7sJO&;K0^{5`*?DQcO2QeRSh`gHL|;|d~0
z&3o7kQRzGFVMF=d7qIrj^dfTXN(UQm@rNHSKAr`}1%`(Q9`9Q78s4{f;Uc_=+!(Yr
zJTPDiGLruBaGa)!+A5qC(nIR@Ok<;K(4$=RNTVvsLB>S%G-m_wQ<r55{ov1;#$9j}
z;ZosSZnm)W|3AH1t#WO9tl`CeAOpnwL{m=?o*5nGfNECB7RV3wUdqyFiasIyFUHSf
z!bVLh`MKnkF_T_LjX~3<ub4J%Txdc<XjnpmqG5>r#wDQ{eROY*nVX$GcT{3;UFPJ)
zAa~+OeQ>beFaf<8IsT=YGhZ4XNnmxxus#rbhy<u!R!!}shYt8}F|gfq$B*<F;cr@C
zHE!Cp71O5$iC9_d%meCdp*X=nP#if?iK2*OlZa!WYJzGS^MIQ*N+FHhwQ@QI@K9UD
zQN!8TrjXPsr*y}KFTXM;A!N?*rQYB?s_3ncUGYg3H>HfAnY3#6t0pveXPa*x^v|6)
zX#BK<;Lrp^Z%~`|tKYV)cR%MAT1T`5I$POHwW=q^Gg{_5TA7^mVD<p+o;G()B_pK7
zrzZ?X#l>JQishGC<Gr02_7&3#W!9TzMDTi6$J=5+9x|-t-q)$?`9H`D7@u^YwZrdU
zb$#niurE_6hN+n-3SO>23Gc#1FeL-UGu;AjX0SSrj?q#v<qg!qnP_e$iqC*4@LmR}
zXuZYvMbQ~>sredq(3?2P02c{Uh2jiQ*?N<F$NwRe2%FypJ1aoBsCTyA#cPDnHFS0$
zkcesiEQU$glmT{D!XG8@_`a>T=o-`b#e5X`-cw7w{%RLbesQ|ctJ^-vEfe;gPKS4e
zeOxNsrx9@o=aczJ+E1IBw=fDV+)}vDPUxKu?+E+&NMkXK6kIc4vxdTh@b^5k6sR#g
zoZhJg|9c<&vE4pAD{M{|Tp=DL!H+OJ^m{wt>7pj7Ui@OAH;ghC!F}D}uUwJfs==_O
z3j5$)_BM#YI?-@yW9=`#s1+7}9eykj&hQplgF7-hXwA*{*Z!RQ-VZCr2BD)j>{jiM
zdgI;9uMTkjao<eIy|#MwHK6%vg#+YUdsJnn7W*E|$gE<u2U(v(a_7=o)0hh(FLyd5
zJ~3UOSVmr>CMD#_g@5?lxgVmo(!(3h6`y;v6lhBKm7FzZ`a@0B{EvTzk4Ff`m<1pI
z#5bPZyuTEoEs5tge7CWx6!2T;o)H$zc@6UZ*FT;cCm8itRsRdy3(14IA@+e6^MCmp
zB;l9#S#NCh5POi5?-He4DR<gRk*Gs_qQHaIXcaUXRbU62t1hb*9tywLY=s6GxP{nQ
zpq_V!H>qBPhw$#kx$vggD~<5`1lD_<wr|1=?8K*CK&r3M9rRDI;|_s=^Q{qrrP_{F
z+6cZ-)tlA91S3>5A)sB*JaO7cJ{x!0x!Mn?v%Uvv)cSq?O`pL(=dJsnWk7Q=|Hk0K
zw}j+(4)yGL=$-6OKH)a6+PQO8YE{pkReT}(`QjyX<I}em?aaTC_dV#pW6@im7%%^+
zc*7Uf1%(F=t}CeCS5*8bGoP!@6cuifq!R#RMc1nAZp8zLPxXz~YSGWiN9_Mr%6%ez
zM=xVu^0nT1l<Rso(%x487UsFE|HOBsZyD=RDx4vI)ccl-ElX)`B&nsD&fYFWpU`>)
zdKlcv^%<zw;(bLY7lBFy-`e-qBM`C#g~M)f```GLgUtt;nhrEqpFS<;adG$5ZmjOE
z9n-jp=EL1HmU0V(5}0EAh+7Dfz{X8rV?Fv6-4%YUwj3umMI#30T$m!1a0|&`kW`PE
zyH`D0x=HB9-qui+B=*CmIh1zL!wxvvL6}YS_G0ud$bSX&#rpkib(x?S^zZ__(CQ7l
z!|uW(RtG}XDLd&{*@eE8U1>lm_Dho2ZMP~n?pska-2xoXq~WQ4w=yq=-GIlY(q5-J
z^V(9%%2KfML!hl)vZMw*`A{gVG!>OnmXxaDubCCgGR@6osQKJxm8-}T*J6!I#VVjX
zdStc|xR?vq%hjv=bCEwi>vTOD3*y46>$R%|W`<N`&F3UGB*O-7R?2}{qWFJ~BinOc
zw3E4b*s7dCw75WgTr_;KdEJZ)U}{kHh&vl@pvE;1Hvjq&O5<E>(dPC=Grs@rW8=m=
z;9D#5H&0_RUuai}s&g1P2W2%DxVTe@Ohsb^T^C^^F=9TZyo9a*B}TfXcMgtN_y=_k
z3knhU#;Z%aL(*isutUbmTTPwqQmY^JB}?=mcI06e)wH@o6wy!%PjpTdQ(+ooCxnsi
zZJ3}gXVy|jS}Fve4|kYQT8cDfWuS$>)q1(6hW8O;&=cDps3Yn3&ofJ?G{_VGPQD{u
zfrlXxw9q;D$So&68#%1Iz3%c({#I?RI9;uA2hEHH-8p}0Zv3A@v)*iq`C%dH&30Uz
zRE!m=lBT0}jH0#&{BAKf{~2nFl>D;oYCX{WelDf4`g7wCW@R5l|AOdaP!@oIBsv1_
zkn75spEm`hAW?a+P=*My>G+_V3b0B11;yK<pPX{S`7ERot(B${(}ccL&O_xa+@$Cw
zCpiZ&qXjB`70Rg%;jAuPxLE8y9OX%iY!{sqvU@@}Zw@!*9l7qDvKEMflfZnDdCy85
zWnxkEtr>5Pqw`7oPi*(fc`j66zXQg~lk_0lFAaypGKL=*QmsNP)Rny17ZwYtlO?>4
zm7?PcL|9D0e9Y6&E8=`vA_JR3DrW)>)Jd9jfdEiiI@uK6n;uJ98qdjsdSTl=49``X
z!$4x#6`2dBd;&A>o}fruFtvz5gCu_U7V{QZvrFN369RzpWiHi`R8$vpgt<Z`C?*;=
zz8Tp$C-p-LF^-N*0-8q=!om;bnjt&X<<il_yzgJj_?`=){Mp^CUWmH^=GD<?1Wkib
zDHR)e`YwMBFD%w9IZVg(w%}{!@}QVC=po!~0z-4IT4HepGz^y%>XV}H98XF;{Om1n
zy|C|8O%2rq2rtoZ#D1t^oGtqj?jTize<h!@v;K6gG)f*vb8IZeaeqW__W>DX9IKbn
za>Q&L3K3y?#sn3)NEl>3mA9{p&<X3L+6pE-oJ?bs=ZK93xwO*8^aybRa9XalVWP!z
z(hPFJ3`UcEczil4z(c_Y5&I(GPNAl%vKkRV!+e{x@0AL_0!{u)tsd<_rJth@(bh}g
zMg>R)Z&iRMKCX4Q$vG+3C8D*lGUS*o%oBv6iFbrQbh@>70~cr9O@3IV55Col@2isf
zUaB+1{gI@-a<m_5*0M-Bg|fPeG?6q9s|VgrDytPUa^1`df_o~bHi_mW)9ljRtPgHa
zMWnl8QjHkw?r%1o9FZ^0#fx)EeP}&8?9C)k52Gy`(VjC~r(GA?Tm^C3)0lM~Fae)(
zpy>{{Z#H8Oa|-h^4alAPoRr%wW^|Zl^3OZH4i3sYnRnA{$AVd=2>yv~M^~}_{Bv?v
zl4<X!2}|K)Gq8?fA@~mkZlqR93gx@rv??eb-M${Ei4oDuQjE*HvzH|aN6pyM6!4NB
zD=on4o+L`|PC#ub<iC8P@s9@XIpK)9%vdSZ;AdwdDMZB1LL~kVp4M^<+>P{W!9{eK
zixVEBDlQJCkr+fK%O>D;y5n^`nP*3e2GvL&b`l910PVqUG<LyCL+tYR-di4PXt{-a
zLDQ;Y!|Ti5dvDq62BW&HE@9fp>3Sbe-<Mup6LqlZSnAw>8T!G4d`JZ7o{J1Vx)$XJ
zA2FZW=8D{6tRH1{vZ{0<|C_Znv|@)cx7B=MJp~CgigEohHWxq7fRi&>`*#3BT>_&u
z3Z=Aup<QlrdkAP!a`)@sPaZtFXbwvt#TS$n03-lGNyXLf(|4_yMl(y>0brt#k{-b#
zJ(M}+*60C4#9Z=;ozg>84SRM*4gv$dvl#FBbi{o$4!JW91j)26QOB$xR<Ba#oIWGw
zoVuIxO^LoE=IS=#8_&H~`Mr~zqbg~Jiab>^Qjw~P{A<W_og}NW_+pxWJKD0YuDrOI
zgIfbmnt;Ik<y&2ienb}~U`%<(Itdm9Cn?}5-80E+Oi_ZX-`flOg*BK^nqIzQDrFLe
zn<?)3894)20!WN~mVhz-QhBRUCW{yW4OqMkyT^f~*OQAJJ|Y8Y3Kqz@12fP>OpjWU
ze+6fm;%r9C@ltMPqq*Hp2Py-UGiXPXvYD+fS)a!|e~mK(gxGU1`)ZW4i5i|U-h4WX
zoUTE(Z%VTzz(pzMO6pm-$ZWB8SEVt|VqHsm;M>y442fF@8*lyDY|<X>sX^w6YP-kn
z;Ebw1%$H6|P>E@SBL|HIq6yoZCJO}`M5o+?Q}@2||I>mGjsUy6AFoTDSMHTE_Ui1I
zZNqn!?M~v1KYb~b6i<6~_sX3IUEW+3{_>*Nml%s*o4WJ0*Cx)C>igq0n=#}h#>d@!
zMR#CYX$#k_cEsU%9~=hi>s7*6V9=~>{)dZpE+7GsMbl>PT~k_pNr;;<<>2g@lU`%G
z8pTKKjUn9}UBZeZT3l>Lm`MfEI>C=)_)!>~sYQ?Fm)OLT8%L}KK{jwTH9)e0v4k&)
za~@115zKT-ZI>{IX4>Yb9g)D4q8!yB6(kxExcZ4bl$5$|2h@PB%EI%)V;f{KOP21C
z7g+z0p^AQKT_TF8Eq~Y{M~F?-x<m^ChmP7@Mcuq3BIsBHM#S0T3OVJ~D(<|9JB!5}
zL97+FBvaSO0b(qBI0+r{NxQC2RUOiZ7iq-cl{ncdyd9AZBw03y6v}MzPy4=8U6v_C
z+IEeJ$g9P>Lqf%MDpA_Y6Y@6dh<VZvUb~IX+Cn_erZh3(+Fiz#ymz`LKO=-x;Md`4
zA!QmKb_v(+GOl=it|Qr*Af%49bw<Rp5LPh7_Z~{&?&5pTmqc@z`c+xQQcnq>Xe%Wx
zew5nSA||&(dCjSNv@o#&jS-QjrngBsFC>|xPN=~;OchErpgE=7J$yfs=(2X%{EP$!
zcSkGm)Ud361u?j83=Ch9CSrd75|Jc_r2baLCMFPNhnS1<0*$b=TJ@o8_bN=6+8?tt
z(^oq@4(-cA_;hmJO0q+Q*NgOov<`2FXC_~QGGaP`5>2pme5@9(uy&ZFvqW;8{8ca7
zs?ehAn*7c5916Be#lhv8EgkAB?HY17cIoU9c)e-Xr}c_59m(H9GmEg4zlHv=MNBI#
z!)?nYO+7xZFVIM{_Vo*0JZv3Z%W8%au$p?c@mlvk7tZ<EYwPH{>)$*YT8T-y>1)9K
zTq&R9oZ1l~-=&`gF2);h{<C}2-uWEa#gq-J-uUqX*#Ft_8~-#tc<!6;OnW68lXdT!
zt)KN3$vWrtn7OR0i15wSNmY@N5+0bwn$s8$Z`Cl`pQ2+tNWh>i@K5u`%l37}3FV%f
zM&Bz<she2le_$adF}#UM3~#-BG?jpT1unAtB7IdQ(#DkeA1>aJ4Ey`KZaq9>Vr*hs
zLUj4o$k@GWKHBl*)6&Q|bnk<=w{8@^o-w6r=FG5IS&zVkB6rV7ED1ua<}&|psLxlC
zteKYocUvIkHviWf!Lxm!c=?L@QRjGdup4|eMdyFFB@!B4-xyjyoFnxRwWi9UT%zy~
zI7sLVZ*qDe=MX8D$YNO%75zlzOn0t7X$<Hm4H4^GR<VJmamxB%(oj$$c6{Td(Q+-c
z%WOQ!aSNsJ$@vFL3zh^cdKBt4nphpE$?C8OxNLKv<sM~N%xkYWNwBmqYt!#CB4P=^
zb0P`k@kF<DvaCTOO(Inc*(pyW=Xwf-5>3Xn-Q|<ikHvbmrg%OY0H!0+cHVlOr%uf&
zB0jn1poztgOEx|(2JV+5qRRg{;^*`mA)x0-)cRp@!|CFZv)^)=b1uHR;g2Hx|Nj79
zK4?Ko!Z#^tDN9<5PHiY@XecQ<O?84z`~+GryFarbQw_js9R-J8TyCB5B+NE1n=ot7
z;K6%lmDJaB%`N-B1dD=W*W3cVZmo$67T$b47(|A|euHiwWj$avi9K}O+jsZTb+R^G
ztbD%}cLa#y4c8$=CYV^SUoT~6la7H&UP<3k+dBfoRI8^;Fci}Y5*~8TmyUO$xZ+mh
z?haQM`@3{$PgpMt-*$a@Iw8WyurTt-x!vJrn0oMZOb8}>5Uu?#dJd@fU;?bEL8>}+
z8JcQz<@Cl2_zT*fg?{Tu<$$u$i{8aU8bb@t5Tgzm#^>g4-b_mLwh_b;wbeODJ!qUb
zX%an%rzHc9ondp-$xUa@%gbXqpM{dR3sn9QuU~KTI197B7to%F@Cqyyh3#eyFV@OY
z7IV!ST~OIFAI~_K?u(WFA5trgp>=i2`a@EWkNIem^|4MEKWXeHv!N~f)KO9nvD7<+
zMr=>0BgdNjBk7s3N$i<nZ{`CxwaSoMWIToHsFr<FFOnw7FR#Z!5lr|P?N*ny_Blj*
z&yixjJrunsnElle_P|7Q#b3&u2gL8m24*XJK4{tZ*-<-u-@)HY4Ik*|M7E+3aeX@(
z0@_gme+$Bz4pfSu$YyzQje4o~gE&SRtwhHj8Z#qe?28c*;&{@)fhCo(nnh1Uyh#rS
zz1+hg&_zo}7Z%nP6&*SxH)EJkHG$DelZt8oj45`R{Af~Zk!KC^tdlsI{xqvhXGXlQ
z#qT*p_*VYdbXOs>`&#-sk$9Q2=n<1mg3>_KIJ0`4B4prlO~uNr%45P+E|ZLxZtFBQ
zfz*#dginDLwVqNaVQf=%r3uv5d@;wL^ar5h9H=E%j+LM(=($|x#~VpU0M{IfC_Ofn
z@GzH1>r{y_llMkHTU9R9A|z+67KMOi6rJ|1k<960q~}D-q?MXzA4X#yJR$rC+EJ96
z4+>c<&KrL#%L<YD;s{FpOlg>62k2mFEbDpX(yo)IG)so5F3D_wRY7Swvj#{O@UU~-
zLuog3_av^`@$7CB8n_HLk;y1{5ayC8u>50lNPU3%caRk>!U#$%>s5dt8P0Zj*hHAw
zq#prRn$@+rq8jo_`lZPuO>mZoW1llsdX#6lm0!;)Jc4v-8dHsZ#bgp3-6c3kbd^q=
zr3Dr*lSi|*R~F+auV`o&9I~prqt4wimr^(vjSc|cu|pu-SpZ#4Yq7pw9dI#Q-3AwL
zl_9di$r69H-P0yeKMl)zZ1++js;t@L1h1)0;Kh3V_<eQ1H2uOPV;+|gaz2f;#k?cd
zib6N=()|stVoS_9ukGfVKaEZh%R)6$PNMP)?!mNCOBsZ44=SzUnh()~#$++84I|wT
z>?`svx{AM)PK@;ZeDQl0rzO5Gs5TDV#d>=s&;yl9vWiq@IK*UC+bt9#!eS>u<3#+v
z1foCiRUtwp3Mrrp!+R9WYGe8U(OIlkde1fFw)g>k1X5xWURtm+Yr#v$D?t?65*pz@
zW5J>YGyG3(4<0{c^Z+-n(F?=YY~D7;FespJ|K7vK%zHU+JG+bM8WevoDPF<WO3;XD
z65P>hQkjPiS2eB4;+Ew&1&z`vZdzTgZGBc+X1_HU<fmMYVl@ju2u<<(j?i?t7@q*;
zLe$pr=7K*d@gE7j!WO;U+b=Z<Klg6f#2<L9*!rv?Wlr?CLG&GHiwq_jl$ghE+opk3
zOa4%YG?3nqX&sP&tc~wX##de)aTcvX)**Kd!k23YdtD;(@&A2&3+k_j)%+ute<c;$
z*5h-1xAwQ+YS~HC9|F&!f0=*pB*th<L)b7q9=C5_ZFMysdzaIr+i55c>`nuH#2Khx
zoAaT1<8!Pkk+)(Jm~b|F`_j8)X`(-FMgNkPXKSNSy0(t4Eg2jjsj+Tae#Q#{8|JzC
zy_)DcG;jg~!`KwDH#@R`H$)|mG9g3ibA*VvQ&>#s(?)L%6pmU{O+xEsIy{Y(dUgO+
z0jV#H`Gabw&iE#PaxGqiBk$h!{rR?UKFnOXnQInqUq+9C{bkrkxSjvm?gH4yLTB`$
zF|3Yy4jI?hK@}FfWIYTcU4zOU7fEBWT#3sfx=Vq)oFs~^ENV^vbQf1nVj2)#-c*N%
zf+53EVirG38pfK#0;AdaE2%X>BPuPH(8#b@*4%ZZ0Ko~&L%ODhLS#~X2uK)u1`pGF
zd^*y4EU5+dSQexK(@nryQGl%N9VfzGA;wVJUSFiH6|L(fk<?ae-!QVb+kdKYwZmxm
zozJVGliB{KS%bR$`-2Em2|_of!O8niZe$S*nchr)b5&|@?VxFqx+Z}<86<@wXB99`
z<^>kpDbFuMXVMNc7>K4R6@W_g>MqiYYKH9~>;X6Cnm{93qBWo-T6O<_y89BaD35II
zuKF58BnDJOqJ*VcB#I5RfQT$EAR;0HB0`I(ZHp*k3oSz1VhgB<N?hi;fI<K_bkHao
zgG&+}hcVG)jLG$Sjd7ez#w6D$H}O$4bormE@9Txl82_7j?*BheY#?2A&Z$$U-a56N
z!|RHAnf(Q6z{^w%Qb6rrHW~Q?CXgJ439dq3wl49&X&wiID4gQl$z5iVV&(h(Yy^xG
zPrXbrVsnz6$ywvb;n!t33_D;nc6Z&B4I<oTmU=#u<WPZ4N$2xZd9m9ajr}W}v3d0r
zp{{Fh^HJbw%(^LYG-e&PG%2=A_`1KpsP6^dPCfB9SOR|c>+OFqa*Usfo(MhBh21Yf
zUF?GT@k4v|;ybtQwpj*XGk^Y|sg0c-`WImC3)jJh*7dXw7TVZ}-7h<Z7XIGvjo6(>
zt5Ozox=S`kL+GI}WO={&&e<*bGj<L8K>u|M3AV`TkcWAqK9#=w7T-RnKR<b<@gAf_
zOqV#TAhQr>Rh)e8oFU9A-FRd<QlO61;v&myXD@_461E*p5vZMztJz=T0mzO#JWIx%
zyKteUHdtou>Ycm<oxfZg7-&OZOtYxe{DzK92uZypW0o;CHREZy1$cUQ+!J~cPbXx#
z<tz#BnQ4c8q^@jd&*Jl%^g053gAI920$!50t69+5UbWV~qiZjd(4J*M7tc}+sgmDn
z<nyw;KrA6ZZF&x`7x4h$eR>aai3fCIYihTNxX6r_BOf=h`$lppXS=k5z31;tkm=6C
zkGB(Z-E|>prYQ?AnjPjciRi%K!w={s`JCzAX1XzPGD~2C>1vnr=Hy_{H1n^QYcC-p
zH{Kgf)=k`il5<ds{#)cwe)=@M^p9j{K7amA{c5P!m(Lww2-3?(PmypkLtr7pFG(mM
zb9-!XtYT-@&Eq2{O&WRp+UO~<!PhzYnzV>TuUTsn=9i<kNPp>g{q8BF0w+`b20VYn
zrEr-ndYN$S*&(7bADl8p^68r9$4}nZUY6v(S&~$I@L+Muh9NhVMdbV$y?WNDh`6Tp
z8%|D-7VejV@M4a}7O^vWL9V7aNGRph<Hx5+JF?N#*K5i-`A?^t+c>%2`FLr`Vf}7n
zj6E(IXjq$_-|^3|4OFTam0~4w*3{Q(_j2+_r<&XH@weu55Yr*0(}#a`XD>wjPC}2!
zCKN+Xp<H6a$u)O0<dMd_d~dVz5@Tr{6#Y>{*73Bd?&3gqLzq{=4r6dv^#ui>J7_`3
z$(;9}K6!uT$C6y9_0!fII=p6mGLq+$4`ZeyeD=&QdFi9ZPI<R%!+TTq8fjMn?KjyP
z%)Uq3k^rtMxjW+Dq3a88bRm9ZA>T&Mn%9YOOi}l`Os6BUIvr|aR;VLIoSf9~<wajY
zc_C+cl1@hp@Hz;Ak6eM4`D@5rxt#nZEe(yr_w;Y)=i{YnzP#h}M+oThcfrlJj!{v*
zA~I@7kTVd|Z^~UWBzM3joFUYKk5Vq#@O}`Cpp<O?FD|*gL4$8GnvW5f5&u@l@OY0B
zl?0&jRJCH*0Ly{e3{I{lICr%;e+RmwHrmEF%*M>~apP|qNEK)yx}K~;>qr48p&rS}
zLX$n^vo|zC=IJCIs@317=4-T{+ia|}{LpOW7y>2ol{$?EtWzk%weUKCuX`gyRIa2=
z$OjGEtRw9U$TOg*!WHfHIEaS$(_a>7KNf7$81pPC7ab|iB?-_c`YNn|3JSfjKSgb5
zFW68Y7I|4%flw=Qphtwl1*v<@`t~#LKYez(sIjp~cl0QiNt(|+7d=1MOje#EkKRAM
zuH>{ZrlVQAI=!RAEjqo84*f*>(trHJC9HSF3L_TahK}R7V^lP>H+R?0yvsTiL*xHl
z=H%O#EZt5T=qseDql4o}Ysa>{&PK$5UJep3OXHA0{`3@H@C$7Ra(h+E3ZJaCv{lV|
zGASkM!;0dFqdI6II`-}=UC?;eJVJW;JgiU5D>+TuDad*gWTkmM!6G>G)R+RuO7P99
z+vPeGzPK`D8oA#*I&Mk`${;nIeA~jb)q~J7(jFTe6wTX;)AL7oC&<jNMNL8KEe)~2
z2AZPSqq&K-Z^e4&heRX}9h)>GY3$Gg>oQJ`N>Hzj$&gK(H*cEk9WHa+$OS$lLq|wP
zIyu>U`6rE9v?D$=W9g*PUIA}9O8nCz7HnXAJ)j*bXorODV3;<gjGJSYlwrs-(mE>A
zYiz>wgt1;bHjr<T>%?iopOu_k8~B#1-;f~#y=5~-Eh^JS1i8tE3?As^J3VOrD&PTY
z8GM$vbAyK_1e61a)H~z58eRv6a-H7M(6XtdMNS^!N_<!UZ-5Y4{P+Q#J%I9dLZp17
z-%TA=RI~ZKPMw^h)}6QcXbT>w{~@igFpd7)@)6Xv1sk=5x|Vc&&~UR7+l~MI(>ii4
z+c21TnV_8bnt(x2sE?3tsJ}S6jND!P<+W^bd*i2H$!}6)#!CHes9l70`XZq&ZsXhI
zN^1&%ZE@SqTeo+Dll((v%jF&9KAT$3q8f4?`H;DSY*uVe6l4RH+3vmj_Z_4KMlOxO
zNu)%L%0*^nJG_^nbv~0*m0h6KkcPD9=OZz?OKOmSlM-R9@pG4u%_bBtB!1|+z5rb#
zlQ_9tuhFF9b!<%efUZknpxUyyjiwA({4~SSn@+|g9adWHU`5c%Lk*fh<r}qrGY1Wt
z<&&^v2~N=;jB+2*R9td$ggZ$Y;r=%d>C)+me4ach(DwoQ-Mlji?CH*<SAK>1&CI}*
zoINFTr$q$M9u)0S8sa;1uyX3m5UeO(Ov+=2`6%ONW!j9IG{Q4FFhbup%qu%kI@F)X
ziD8UqO4oNLl%uN|r(|?P>$He5Qjq;keyM{qDGO*%ypNPJ(2*rR-V1rkSQHC%YeBcR
zhOFrP5T@_EI;7RHe-oi1&B243piD=NW>BVU;!;(DR6cNE8djVvDQ@z02NWeu?!Lq_
zCm~5C^SI&Z$G?Ang!4gHgL?MEf^fe3HhIa36D7$TU9$R33rJX;JuT1w-(pful$172
zoe=CLU7DDXGn?5$EC_7{dN8{yhONj4Xe^lkcKHq2o)2FNHc)JA*RR-sM$?@Q&|%V}
z4N&?;&@OxCC3QN%;<*TiSGY(_B$heT0@DZ1ajQsQed4%o#Q`EgbH$2r(Q{E$FZbky
zzMdgrktf%#Ik9g>L`ouUC*37%<8tWEj;xCpPq%Nk6c?{v92w{^e3W0<_3P*xwJfX9
zRpk^O=%Xf1B0<K!Y#gwBK8Rlt8$ud}qVDy<hQvE-;{M<vbIBb&dEOh!Kuv>DF%Gjv
zBrI79#&o)Hy=-`rhxd%A2v_NG-RgHiL$gP?-*)$1vOkCC?alOKN9Q#MS6=%(Gm4b`
z%{!nBX$H>^EIe?aaAf8i8e)A@nr=UEmNa2Wj_R7k=WzQUrIM)jYbe3b<GREzJ0X6F
zU<X?0-~L_K=+to0I#sDQagua~gGS@D7%R4}Bn>DgHr2<2%%yhfAr)vS1za(%G+R)0
zA-Sc2Lhbs4U{Fp_fX}NXy3Kr!ZtljEyZZULy7~GJ)@Zm)sf(+hubYdMwCWG?`L6Fi
z?<=~)>N<3(Yjo;1L47wDUtcHf9J$2i^#QSz68nFYQt|91I46_J@!Y$6Ip{r`GB<yj
zE6_;qAV&K&PHG%G`Ca|O{95vpMuYG!@*VOZw_sMdk8H^uk)e*LMevcSb~L;vWcUtw
z2|gn`*=Wh2wFl>$zUSJ&23$M!kwb^xqN#!$!Vo68>ohUrhbIiO)Jz^XA{II>Mp-i2
zeGyiaE+gAe@}xvx7cWwZlB1KnT}cW~iAqTE^f&Oi1;f769Tx7EhWw<3wo&k-JxTvj
zbTH^b3$k@*KlTRundnFD$GMaJ;FM@D9|H}gfJFjNtfa<cV91#9Sg{j$vHUCjTv!}w
zdJUHuKSkI($nwMq^6WS&ty#RdCL1f7N+}$chEd$hNucj247@tr{eVgQpJ)&H8o5!9
zI7Lmi^O)|9YuWra#2nLwruAn5i@QL&;{pSu+s`sUI=Rb9qbZ~>*}u%+lOeicY!{o)
zuXZoQR>%3H4A?4_J1_`=`4tAVlL=g<lE1n8^5DwbRU^>vE5rdb>z33Yr@n8+#!z{u
zVTG-#&u~woz$vjqeT8~epd$;z`5vnJQbV=EIs*Jh=m%?{>)9#WlDKoWCC&kSGLNXu
z!+j)qBS(#u#idROD4HK$=;`O0IZQexTDc$^71{WDCc6$BG1SS!*740q|Ab6WNvgB^
zTW+pH2Rg@$mMeKXqdsyQuy>+kE`aC9v02+*;6=~t2mkGF(%^|o<;3L+KD#;8H%g(1
zQm1=McvA85bcwU;PzRUTG1F$VxdGOikMT1)HyE-m1@VdS3|W^n*+n`X)?B|H<`+E7
z!QW=ZB4nektyPm}f#FUn7ob63V%Orf4rieOGJ8~8rrEfWR~pPIH8re1KyO-TL;nsl
zsVSeRNUMEwVUVg+lNd8%h68h*Jcn=5pcHaya@sHn(?dMa&@lf+SRV@n*v7hMw^p;O
zAbG<KoI=j(s%uuG3FK^SdU`CH&}p0*&JL;NYmi4FRB%Rj1^-nAWh~ilFb8fWRy|wP
zzV=81KQu@*I&JVlPeD1lMx~|be;hcx_in6Rw`I|`b;UQbwr+#s8Q|l>KN!dG6Y32B
z`UN<Z_mAoIP5z1kF4(<m=k9;rs@%M3_g7c10-SoBJx^{5{y2;im4FZ6j|XwG{^))w
zX(-P6Y3(680$Oo6i8J}QK~R?60x=|P!~x_WN(DHUlk+$wl@#QsqB1obgUQeDi&G6T
z7LmWm&fZou47rKy0&GPxdloozU<hJCGhuNz_h%Zk^XGm)lc%WH&pWEBcJ$k^XU`6c
zM)DE4`Iy`!AE9vM^%!}f@bjNvA&;+o{`nPTedTkJkRSaKZh|?d7hS!GT|8|AT?X1(
z_y^kZ%TnB%d~xIm8h)63Fmv9#nW^&?HmzuT>qC;YtQG~<)sc^DO>}D5T9?NqCqx&`
zG<`LUKYKU$6C{Q|JS8Y+YjyS3*XT!M<7<sGW^kDw9o>HPBlvjtgOv|ze*Cc}uh8g=
zAAf;ty*kP^Z{EL5HqI!U6;O3BA-^pHeh0*^)_b9Ww?MqZASSA6n)ezCDW5sdhWy!|
z%Y1O<lZPN|UR`BOI8y6x)K<oZkg_TK8Kt35tiYlNnbG#(X?q%o-!>2*mkPv#eP_;`
zV?+MZg?fU%w6-!Pgp^MKU8I6-X~DLbzhVJ41t#TBKfV0>+1b)_BRkuEIQuF2Cwk-Q
z6XZzlJ*jMHsN8e-@Bp+KC8FtQBgrF&$sftG578jh8(Gsb<S}_dzN5O%0{S^GZM*rr
z?e;g>Zh!jmQnt$(+tM5CgPrGqz9eq&;Q+1R71hs<Dhg-wE#X_W=kOpbNvf`F7^J^5
z9a!b*lSv-hh#Szx^Ry|O7vu7+zceF9WL>^czkCzpMaN{pOYzEVqrBRll6%PU>7EA4
ztETcW@Ja$+x%w26+dZ!hi}lMZjC+R@_&ot3Zag0bo-XZAWo=<=`fN~k74PqP$P7WM
zWayJ?`fIlvuBT|VS{%%ZLfLIFdg3j1v$7O@T<uT*lHprO-98?bkY=Q#v8=eyNb5sH
z=UQ+DIGVqL?%xND4C8{rj0@jD5xcY87t!-zSNnjWk(vz)D%2bE*6gu3XNd;<_2QH1
zJ6Hd8BfgxISgumd2~RGS&o8Sl8}unc4^GXg*2U_!XPxKINbp>+@j)yKvw?ww@QQVS
z<?DcxUdSgpBr#%gI#-MuRuxpIBsHVO>BuiS&38=PoP<8B>$j<*H&4Al@BPsLMhQSq
zkQoJqVKcD>s=Uz7i)YOfTUv2r0umx(QaKHGK&#&M$)f3~JSlElWH=|bz@C%*N5##K
ze|=Te_N+bm<I_U4EVm5XtH`7yr#qRc$k|959g<`ed2hoxl+#J#(wYleAy*C1+ijAW
z3kvK>WTmGpEiAy!Dl{-5Vsa{{6(7{9cYcxugE7fBb?oF}Zhb7wMXpx!A_EVY7}*%<
zsAs+-#cY?*)&4AgO|9IztfX4}5r-T<{WI!Sv4Pa!brraex@KDD`sxkV*O1MBzn}kK
z#iaT?FM9*fje98@X*9DD%d=o3pIvMN8~Lmpw{)_Rs|Y_lNo{2HADE3up5N#0SsXFu
z`J$OYkl@<Ueb@kFo?4^2Eb*7BcgVxfSN{9j-e>#QEYsA~t(8wzRTL}d==wJIetP7@
z)8)9`_1wyHPqWVD*=Lt-{g3rWR?MEUJadNQuB(5qTS)Uh4gmbkfw48@S8^frnA!mS
z^pMY5P^(jT3t0;Ux^hm*GVq#{>b@oXkr@igZLlopoiR2leQunWBs?iLWYiR^TyE#W
z#XA$iqZaS8*_zciJRv1KZ0drEfg>Zwj<iiLDw`Xou<4_nvTQ5ss|mn&?@P4j@lCP`
z`#c=T+H=S*kG5xaM$he`bjyBgdz5(xEtmJ16A?LUZk&f}#B|mkHMc8^wkL8?oz0d-
zJ+x<KoY0<Pz+wg9nG=nc%~D_UFDXPU+pBr9iepHlFXna<g{|{}e~|BPHk0q2k2#*m
z(;eJbIBpUbjOx#=Co%Zmv1{b<=k;NW%Ij9G*v!u6Hvx}$us!c?Y_G>Aa^b2P(t2I<
z<+|SI`qwPgR)g`yt13#Ab5;w+myg?B&aXV*#v9+3?8Tj}kItdMQ(VvR`Za?GPhCbk
zaFE6Ir&{zFUqo2%gKf}_X7UYk+QhZChvIJya|ie(a<;y*m(a9$G{!y6jE9eB=|}<!
z*g|fjJ~|{TXl_ozarzsC2t4s!HlMs^UX${~ccG#T)i;nBm^WPW8yb>uy8aUHbJIAP
zE#32LSw#5-Z6&vm_5T<@STBfovKZbb&B)l%w|M7!f6_~_)g5AoZvF{9Izw7ea4Gjy
zdn#%sQIuyXO$O4vUM)mD>h%YvS2Ikocwz349x{Ox)HK?xndcE6H=E}8a71U--}_W)
zSL`>OA2hxU?NZerfp#X)o)bB5xXKG%B?ldS(+jG5rcLh)WkH)=Slo1T+Lk<Vj$Ov5
zxsq`awA#7JsYz9E@Ic|@(xD9z_-Fk&FYnR`e?it$icbyr@}1FH>kq70bx1bc1OMX;
zQM(N#k3XApY*h@iTM4v7&1_*X^RtERE1C^0(6HOwi|v?aEUP<OsA-J+=e7&ZeWMl1
zsK`-LC6?8<9zMKcZZ29%KC>cY=1q)TFnH>?$oYo7>>5V<jV`pY8EsUyY#)$5Hj$DJ
z-aJ1uDk4NGP*$h}%8IDS1&)akk<=zYKk4k<-<xJRnnX0l2G~_cv2lgwi`6B6FTFy(
z4hu#_cgg8-PTFHlniaE?j&l18|C2oXdJP)VGWm=gFVwf&>T#o{;P|i#70rG@`$7h`
zD;4rNu(T{-f?}zD(`}rgKht8dd&9F<@$Fk^r-)y}<7DAnh|l3-%E`jXQ`-r?W4y;)
z#-Gs`Xa?XiQeJ@;kS3i_Qb$<XWae*}(@X}MTecttNqJS8;IWEhS<V>)%?j?Ee#b4G
ztoIXI#L8))Tz==a=j%7!f}!29?VQNt`D2_7^NKxK?Qp2~n-V7d<%Ri#enZ|M8q47^
z_}EI(K#`k}#k=Q}KoT=actP@4?MvnbnGh14y1ytqc1jos6ADsnPwyy5@ccmzg)ua*
z&$xh*<2&W`;Y8)nAAqmb0QzudbAyY!Q4W`G68ZEq{>Ft9IceQdm{HJ*@>&Zs3U}yF
ze_B%VY4MV_+S<0deh=PT^~q3=hPid0e_l7Y!DHwrtKP#Uxv4XkEt{Fv{zXOhqD`9?
zW%Fz3zBG<B_^O8SXU{SPTC$D*{(=Hnn3n*}vB>XW?71wf`{NU-B05T$88wc}Y*;aW
zDXBnyR{L5TNc7amiSra2m2{0y<1D`n-4xBSsC(NkjIe|8P3caYFVIz}OmwG*8|czU
zwJPvW?buCB7uHj~!0s&B<KsYN>?fH&79s0zO2z~e-CZ<sT(IVgq7odDbhP~_jBJa>
zHD4FUo=wy5#A4gR|3sEw>-TJ3qG_Zf`+xcV$?v(3@N#R^{}<PI|G)YE|7w5m_w$o}
zPygcgm~a2L%wxYFcKeBL;3jb_=mNCoPLI8Q4g7Aje*8&mYe(<{I@k}_SnPil?6cT^
z;pn!b7vSUF548`fd|hL7Cc(CiGs(oZZQGgHwr$(CZQHhObK+!@e6fv}d*8qJ<JIcb
z)$X&p)~c#L=j^@ldCJ-J@#n2UG6#Kg0LE9O19OJ_{(4sD<oGA6s<Oueb+G|?^<~%L
z!V6$K*NkKQMrfj;ejF%s;Fu5#YUq97-}a~0c)(@GB<giLV+c1yeR<62=2sAzbjin3
zmzXh7XWeGO!ZKZ=j9-EdfLkB6%Z*%>p=+!G42bQ2_1+P6;Q`wwHYlf)v}YT>v){ZB
zbekg#2=BK97>H~t6G8bpo(mAn2*y4kaoB6oyha3NR2)vDIG_bPI!%v;av0+!%rMzC
z-^iYN%L;y{Dh&1Yoel57@hG4Ht(h(V*!A<Z@IfIWRo=m1loRB<lHy}AeqG93?M-Dd
z$dBKS8xbT^pPx=;8RW|^`Uj|oHWkcWzPd9M?#2f<GJ~nsWrHBJFT^t`jSA-{hV_H<
zfTczoMKFpXu?zGUlSKOa`w`edL}#Rp@J}Exe?--<`=yWZfA8dT_xC}Gehd-%Gmg1`
zAb(PQl~5iZA_$<Ieg0c5=-<@5bI1|TR|0j`MSp81|2T*7i$NQkp1&T;gxvdRz#!Ep
z#%H`Zw@Zxb5n}pfI``W(INPkhT3(-@3M0QsqJ9@Ltwi9SBdI1xj%q9Yr6s6ow!FGe
zVGM5e;7cSOA2wZY7ySObVvbthmo0WwMSe81Z*22b8$z!Tl6mx%5bHS1sE&dEPnlZ=
zKHbfI5c0e<Q0DB%hrd@Yes;;mA$gyz2=CDZ^9-*sPR@@yO>A9Vi-GIu9?@?ndC=31
z0<Vsk6M_gM?(87vj1@^MI?HbtC(U2_j-OSt7P&(s1EriZUKcFsOGvutr;`Rbuh34L
z+2OVyZ^H~+Wg*rxp1eE@9AzTEVn4l?ve|&GeXgaZ7kIA#&kSNpohE%Qu~!aqd9w=%
zQD^n|f0y$xcU)qR*&jF7LA1-iA2~IbEOI~|uNrTksGXdT9es0%rWg41ooKh;!Em2y
zzE8YCiara}w9)tD4b?GZSl_^`4V?Y2@P+eOd!zdn?!OqHw2Dp#c7Fl;6@96!Y9oUV
z?9Qiuz^Xej`W?np9ER}fu=QPh$#`|z8U^#}3?wO(gUN}mFU1)8O0&E$K5oA=KAKu_
zQi$2~K=PON4((75m2Yof<@9#ip+NU46dCwXc>a4dP_`_4+{eLz-cc|J12=?uj0G`1
z99se}BudaDAZzrIy;(TT7J253$w1LF%u~86jG3i0*ef0ZT6K0_4on$TE&Eh&UEy@Z
zQIJ=ob)eCyR@9@#@@_GzYgLV@=NR>FX)~zN++9#h3apLN8-+x9!W4UA{A*^+NATAS
z#k(J10MXe|Xs3|TuG;=>l`AoH;f937?;<V!(Y(L{SU<J6)fVB&v$4+P^RN`R_x8Rq
z&vc_f)fr>l5^=Sg<KffxK0cnw(6H-^sqM>=$O|>=op~o#ObqI6ZZ`a9Q94BcDfQwk
zP!_k_?yv=zPoWMIpXJ)Zr@rxb!P9qoJN#1|<XKXTcHPD^iZGqD^IU+s9|j6%{`&Qf
zrZQB*k5AO%b!(sgSNOKRXS2naphN6S5NDOCx1<;xzG?b&8vU8|)&?+)uF%UA|057=
z=U3P7+qzC2Zh8~*6K~6Bns8V%K#UzC${FMO^zljg62p4G!m1GA|6xw{boCZ_VU%?_
zT`Uo%LdaaUlv;o{|K<3Hzk20r#ZQ7R@%J13%eu#!`0vS+fB&Xmo|tcKZs!Y5a$Y05
zCsej9H3kmW=ec8ud3!+m85@EI2%FH6)R!6tYFRc8r$ecw78g^sE!6vTbJN&795r*j
zc{@C7phRRQ-Lx6SnW02-U#M%fb3`<cQigmg6(yur&m{MdC2JiGbLITL@G5Zmzb;-p
z(#-btchkO_yTOiDEEpLrEMMOqVbvTJm7Hqhs|3)%%T!}AbScFK1kk=8!8n;K<{d%~
zBh44LoMYlz>`}`=@xM|Ty_6$A;^A)v+Gy~2XR0n-3KAdb$AfS~93NT{W-^mRYK!W^
zx$hd>l{e5v(~=BORh?=tzBt?on0Vl-UT~Jgi-)Nk`~lpg&l+LZ)yby28WAwN1`46&
zI^JsDQ1+=$Y%{?&aXc^hm>kjO@)chox&Fcuz?x-!<D3OSX%@9-Q>Xd?0JT)HCLQlN
zfw0mMv__*Lc<s7A|J5W^!<;W7DpIW33jbnlQ9FXNDUy*F>Z!xc*mF$SAQH)-U&qlQ
z*>$6hIqT^OW*xdlG}pk#VU5@`LU1M#S!u7d=RiZCjS0jYnNoj!fn<#6cmy2AfQjG1
zJD0@J>JgZ?MvkA_@*C99**^^Yn!f{<g`cnSBiHxzP%)WZQmrLs-2S1DmmuHA@ouo3
zFRlld=c)M$v+2;kL4#jqHIHsj+RVIkpSjUsrTsW8+<fG6E5+nK{vCPsfx$`<@NI_R
zuTI2WXtI5QS-jdC{q6cK8mEqj<AVFocPZSsEz*}0^^88IgBZ3-tD!sgQSTj1ihR*>
zT5nUB1+%Ayupu5ngse5K$ODh?O)2<6Lx4JI;Y>i{y9dE>Svg_@m6Irye|O4|DdLSl
z;=5x)WykwPHs2-Xh(|5*X^DGK7cK`RW~*Lx@d%WUMr!cug~O#BK{Ghi;hn$Wtg`4w
zV>f3}#tK;h1!B*QEQ!s6$)a;f-N3aLEAc_8iK?b$H%e$JCd-1EM5#21E92o1J^c4T
zOJFTnmm?(wbYNiZ2b7|McH|qRH#6tR&LCKG!34A0j@R*)7i2%|CoIduSLt39Yq~r2
zDr5NE0a)awOariz7lG^|{RrrTngcZDHbPm#N?io=_&KXXYA^Y$k$@i=UV+~8G(B?%
z?P|P~Y?Mo^*)UZ4XhnJtCjIcmOgp`)-Ds?Pd@r5s!q5TB^f3D^!lz`mARXcPN+V8P
zluT^i%7nuf3OsR{-+ruxXl9yjvzLw)#a@NF7+l^k)Yhn5LqH%}WYUAtEzqZ6wvg^S
zsqA5*Z6tI31C&FmZTS7xch~{qo5!4WYrHms6^4v1+BpRV|1i{B&yNXl|1p91#{}A2
z0w0xZA;Md&>{9;7uv78_7*XT^j;ULbhKXD1103i#xmyO&N4Iq0-|rlvK<Kb6KS&Cw
zZ)%EyXba0MH`?j}Q!?Yp6BXF9b%RaO1hpm}%(y;3tL7r=us%3H_!k{BB&V9s!+skU
z;q&3j<jJMgr-l|&$bxgNmM(6YF7qEs3sllg)xGANz5WWP9irdP`1=PGmL*!VB9LC)
z|J-xEjB7Yu+J>fJcdTw=(OaTm%NA~}3(;5?kE>Oi{h9t!-ZDo}u{8wJRx6$4F_-d`
zVcCqskd=d1H|`)ID^)Yk9>~tSYW1qnrAj8P%%s#jTGrJx=xVJ<T9_7g)@y2EWl_Zy
z%^BrSpRcvP&HGhw&fT)jb$5xd&7s}fdhIsZP+i44e8u|@ohn7=+V?6ZEaAwG?e7~j
z(5XcAB7dzV-%}^D3h1nFM0Q^5VO%E96>dLu*=3KU)ZUe%a|PIiT?H7;)QyIvt(-b~
zHx9W8PPkHi4qCGCVd~OgtG<P-n0qtHCkc8tmXK-PO6bYF-A?ebZk*KfumVQN-Byll
zy&EgYvTi3rGp&coFQdkc)qd`w_yQ{DvON2bwzQ49ivYF;a$ZIWwh%~h^U0O0(K@Fp
zmZVf&|Kq6yeXz_gSST5``KPVzpQ6+2ly<$ft<!h^{?y~KvHcLs|MKa3Z*96WYm7_~
z@8<_fvak2mj0q%O9sq&L0PcJvK<Zuq?Y$RfSstMIt6#Z1$btb^H88~hRUK6|XvF|l
z9eFjd#Q;|weKqLO0CXLJEilFac^&0P(A(hmI#O$(XF9OpFF_0<gb)bxU^4Siv~{f3
z02>4Dbv##p+yd0id|a-4WK$vZ=|BwUA?wosP6l*O1H22c`vUkk10dH>@cQ2{gF@@5
zu0h)d@akY*LO9fWL>w4!0bYB`FDP&!M*9>TkhT6%4$wGIbwh&oNZ#=L{fzclIB@WT
ze{XQJf2$43-4K7kpbxU%aPUB{4I+^2xnqQq5BQJmLAj&l2A1t(zD!T<{Jz2AfnO7Z
z?hQfSGjT`M4Mcna=?x^-M_H`L7$pGJ;J|mR2OA}X2d4;xLS}%<8&Xt<o6@IWhkOeC
zyAQGs{}l4M5A=f68_KO92CrI(UNsNtH>h_yW?4O;`GO4$%7J>0;D9ANmSS(B4X!q_
zV(+32x;ETmZ}S4~JK|!mqYb|{25T?8^^1K}#$IU)hJB3IUTO>ae4pJN+)N)q7-RI8
zXag{I15{Q6c-yFrz19}oP8hts*A}qbu&VWt>qWn8c7zVcZ=>~?rUuB`Fl>9c*%96B
zFz$Wtiy>c}{+xZc7f`<7ID6o$z@T;nw~=jow=Gcg{alPOLg_sQtn^`&SbZ!p2S(F9
zI-DVzV|yIFP&s?3<9$V%{0NckehKZ!DSOVV5U)eNcKUAU>jCT=vTpF}L8}|MtFZS&
zV0Qw%V8}y4cM81FutP+561)(xLq>NRys)t$4R?^-aPs{Lcj(;6vVDL%L~exHK#V=|
z7IfQyus!n@2>ejM4G<25+5oXV{T8I#;f6b0Zq&)Zf<0bt9Qt0!PisV{?UlG=>c*<=
zrMRPJ$EpqevnR_2-t8x~C(4FR8*s2^$_8=UV|Az14O=^ebtm2lW<R8Lr``!|JH&M-
z-wAO$<aMXt341>Dy8-q_z#j~|Mf66&9~!%b@<z%VEW5?@M#~#Ix$XOD^Kkn8sTb(p
z$h!UW7l_^nyZyf}aJ^C2`|mHnzTx=$V=qv?k$L+kKg|}Qx4-rR>l?MZ|NH{_70$mu
z_rm`H?l-b~<MsjiJ=}L=2O>nYSRPm!FN;@6La&k#_n|Dqsw@SZDe1xzQ@SMC%v2*(
zwP@MQRU>G%sM*X`BYeK-+017nfW0W%%xEK&wP@PRX(On$sH!Z-Od(DPtrT3iED*E&
zXJ%PQYl*Fy)<*2_5?nKJiz0T55~Rzb^jV<>N|9}*3@=OAu_FCUQI0v^qXIwd{A{zo
zYk{Dqc-NwBGxw!nl(}5w5(24tI*g=Zq$q_<QFA&Ii>Sur0tamY59)X^k<2^^ZHY;9
z_ETYx1^#B*Q_&{?h$|khFbIIal?+!j1o*|37+Y8bz~D-aEjk2%aV5wWCIL{mQe=yk
z05DuhvxQ9n9Imw4qI&>HXFOeD6ac9+nXYIG0Nt5bTUZ6a>P)RIItM^>CfF8c0nj>A
zY>WN^;5w783cCP!ooQD^?*RS>AYO6!g`tOEypnMXhYw)95^@Vm4=}t^a|`zmdY%Eg
z#c8J|o*}v=Yo{)rK|955r#7BpJ0;JjK5zbf#c`)bZ^3*ebEi&kfxX3br&e#Fy(QPD
zUT*=u#d)V@Z}^`=eT#Mv+@FGei+T@jD;EmHGD}p=l$o_2&5|=!Di(Y*bt+b@nl)(i
zrIs$6EH<63iLH&Tk*$@jnXR3zp{=E@sjaQ8v8}bOxvjme0j&kC39SvS5v>)i8Lb_y
zA+05?DXMid^MRV;zpW5yD_PwmH=O6+l-38<E1j*v%3Nr2$4?q_If{0;lG({%c<Ba~
zGNt}CjvC5kFRbN?pCFGk7>}x-NpW;n`f!+yB0~{dsghb(x{(|hP8o#4S$($y0LqY9
zXZ^B#^YQX40Bhs*<)w-(28>79orU<lb4Sh;N6wXh7b%WUNC9Dk1|f*MG;3#wVQSM@
zktfum)qbAS(q^HsytD(zN?&Cu-;7WEGf+{Ka!b)D=cNzjb$?w(%ZE0?#ES;Wz7Ldl
z7#cI&T{&{*y$&J@Elx3^EG4rBU3<cm;uGW-j7HQF8omHd9xAIXRGazr3VBmijoIzi
z-aKjy_Z_#0!spYaEW}xN50y8LRNcC8^yCGdZpG;`BljZ>(%t5&CtH$vbL1yF<m~11
zSeXy6-j2BiaVKabFK^la%Ifr7Q5TgM#GX9#EF7o4b^;~sc_2Oc1bTA;5qV`q7<(`g
zC?$wHfVHKh!bcq*MO5jjYA&!#g2>|$z5Oc(;2?;ucY6O#DAv9sxH4q6sC}zciK)Z7
zvVVcqNhFj@3$<0H1sdPOW2-cGRd%#}|A-UEQgywsUWJgKeqS8fh~+@$-yyOn5j)#a
z$~kzFQG)(D_5DfXA5=CfYsK~IdKH784>5|ct4>nA8nW^;pI9q=vN5qEw&|{AW>V54
z*j^&XM6-t@Cv``MDU0b=5J&G$B2`^6IU1|B#GFKV5Y%mw4qRuzfoWBCM!V-kvuS(|
z5iSr7k<mRjjB&Oc#<|TRR^4Z$e@O&HAU&g3n~_xov**s8kTHucK8UO0)Fq~)5Akqh
zXHDNd(VnHJl7-AbYf;5CI5=qxFu2^d8^idhmUj>$EpVUx4v{x9DdQxHi0d+oh}xm-
z*io@7or$iK;+so^)W(&^`ly_cP&j`OtawZZldH~RkDPe86P-qPSu=BtnBK~T)0=jY
zv<^|SG%WGjb29vQ@?v;LwNALgnzEOv*mR^cmg^GrE6drHX74O>QP(v7-Yd9gV=PC}
zf!k`vcZIp+CK9cp6ou(FP};GHudJ6r17pD}1p3YMGKOY0;WyqX?K6q`l9kx36Q|<9
z`J`qs%d3i<Nx`IcWTZzFv{d)OZlMrY*7AEW>&1D~c6~dpTze<I0ZbWlM!RxAuVdMu
z?+SGB{JTkLccFx1Zih!FUfgIsadqqfChpNhP_3Dv{fHGlO_%F#zSB&nJuQaeulIuz
zjzB{6wu_Qw=&qTe0C$Z)1M`ug_A*;3&-&%sn(wK#f)`fz_58Lc{1}eEt^V*v@En9Q
zv}s$omAioOS9=G}45_lGvW2Ynh9Y-r1wRJ+htnHm(dc#_=ofSCYdbI?i6@WYSGefb
z7C&LAHier>>vW<&mRuN%f>zwMD6Ak8DK3{zvPGU1;snC#-u4n}=ofo?2oucI7&I?q
zbz4lloy@I@jYNy~y_F{rF=+PoD$M;dzcab$r?rLLDYk6Ufwl{^Qjw-yv8<QCN1m^q
z0;lb*?nxNAjM4l!QxS`l45FXSG|*Fx$~`Qc>4@Y#p?cbi9Lc4of6A^}clxIJPvyhe
zwW8jABt&sK!FWcrQ4mjW*JbF?wI}TRct;YVyB8{N&l(n3#fdVRDZIa%B^-s)Liujg
z#7BsS5AYney1NV`;#=C`D9jWYMqITDLGe@&NG=#qYsQW0bQjK7zmjwo5f+wN$k{XR
zXntzB$91_4?L3I=i2GVv-lOMQLv;CVHB9Hmcn}_1SxEIFzg&=t|MV%M(_NtIn`C)i
z-zy_h<CZUP@?DtlwlB?PlE9L5R9H-+f~LSGF|%0DORq%4Q>=xLeBe|NQ;<_292mEY
zzP>!ckW>&weiH?fWy(10PI8O|Fh%xzNNs24$tnmFdjU<8al$xedbk7oA~0Fv3C+Pc
zUQ6RU80PsZHd5lxdebR`W4M^E_cuTf70l&3&FqaQoF)DsJUTebCJY6OvnMuPVh9?D
zHx|F;v8}PjGC*dRVk^NE^$6xB%tB+XkK(iSv&J$)er9Gf*%;=Hmf>Xkr;Ukgral;d
z0<|T)uB@*#kfM_$qkUja)fh3gHliYq43h*Sg_+7oleVX0d*t*O5uqQRQNOY-4fld;
zK~~lof<xE#YhRN-T}+Z0pmw^cOuJV!(kEHvtkQ_maC{n=H6iJg$eZ#fS|u%<<=xtZ
z43v%z(#3!wi4Ve}GEon8OPQwJr5vZy)q+_=tu&c^(U`1CDtIYl8ugxz4B3i#Hh{Qh
z_~$K8;!!E=A(l-KV0qVV8LY<%gOdt`J<7Zsk>MXuwq#XD>;$oC713)2ICWB5x2cq|
zX)>zaq=Q<Mb*kR9aHw<*Oi4S|n%fn-rh_h&(vKu?W==+|Y)RZ*#h>B7Nbx}hE2V~g
z(uCHymR*qVer0tuQnCNYHT@|Ow>^}1D;IK-PL!Z;8fo~UMAp8Nvgo2ucGinOLvrY%
zcdilExtb!jK6cczrP@OO(?Q?nQ2y7aV8dsk-8AYj1S_kfN>9A&O{2X^PeXf!o<>$g
zYf_D<L=mmK^N%*_v=RuNu4++g#cF{G>txgUWO5gnh%=j0T7Kj#?beE^#s^zdmGrPO
zk*tTrH;$&N(qtc#iYgBgF`SE~D2G0?gDO#4gRU>qZQwpOl_^swrN@jrkDc+z9M+vl
zy);_1#Xm*aiWIijrm?=$Kt><%nRhr1yN=Kvg_`8qZ2$Qv*ri+s70hhP-UO3F;!x=R
zakwaUt79X!Iw{neG<_qQ6Kj?kHcqVx3sq_flgToPB<2}pvxlnPo~=TONXyqk8eeAQ
z-68KYH@bmucpUctPoV!6;2-24i2Ek)bO)@z0D`c>U*X_#FoX~wOs{yS$6er62tCh5
zbx;tDtyr)ENRRg6V13ip-%&ammq~S>7T_%a|71fjqe$?8S-)jydirqoJlC88`~{$2
zL++b3>K@*zBd9&k(yEEx61~9)1i?T7!NH_ZRnX%|2twir#PZF`Fhs3{^CT2REy@Fm
zHVc(CtC}JrB3J@wz=%ZLz+fBfh$PS;REqij)x37S=mcz=dVfD|bKZB}k7u$voKA9`
z#FYP7W~2bmfMxVF-VZA=R2R!%A@cL3RjH9Ch=Low4C{fA$P<lFD<ILfkcys(S1+e{
zD(W2+2MkLIOT>#Vs%PHxlrBmql4xW~MHWl`?$8MAf(2v{D0w<hNyYQ>LL8BrJMP*F
zBROmLZC!xT&P~ip3taeb`9fi2jIpU_v*+6dLRIcNaP+W#w(SI3w)gE}UDc_=KBFrR
zeTJ;s`x&rq<&}lL|E}BrBKkV{(uBiSl`lkmg&RGFr|y+IyuTNVSs4C+Tv}u`aUTGE
zw%WWITqtV(3@<cz-K@}e_6;!AoNip-JrIa;)9PM35RVE$lY$<LrCPi#iV8(P8)ocY
zpd6V+FGJ&(=Mc)Hr#o)bB{Dr_F~s(XT33~dxOB#A=%FKXoo6x>(~;?xU`IhOD+8bz
zc3%Bl<!gvtL(sZXH0<gbxUM^gz<mmCC?<Zp&N%Xm!nfR=L`Vvcg3^Bt6UT6DSB(Yh
z_JO9_Ek{o$>XGJ6>jy}?Te#Nh;(fIRt_jnCVY*<%xNhh_qKPI<3!xF+?89MoytUG>
zK5y90BVpW2%$7(V9%F*k5WyoVog6V-dPmWaAx%Iw;cnO%P0W@C89}KiB%MNJ_#0E$
z+QQ4$q@iU?O4_JaRiaIfh-OsFTAy*%F#8-XaFN~e#mHC9^A9Kezv+}Fg;`}z;S%^{
zCuL7T04Qu48&l#g{urZNg4vNv2IakRM0GN3qwv!d`dv$Qg49`A0andzX0^qpLe~#4
zs4nOc54Xa|Ia_v_@-4cze91Yqw<^k|BlR3@yC@H!z*})b_{%YjGw6K~^FVpmM?S-Z
z7<ThuC$GK=5g{8V8<&Ug<zM_oypUd0yqFKIkgn?`%bKj)rO-Q>(4?^JOK$M0W6fBL
z?%L>!^}4_sTE{A`(z>*soWVbX6Cbr)7P(wIlw2V_?72T{x;Gm)NNaomEA-wWH~fX^
zmL8q1;TDrxUO-bqj84{x<}-IflJ<MG)wYIM=jWfF<?2?$F4Bs-WA$?E{OK^~a&z)&
zOvh`~tcV8$m;5@5jB8_=k+K~b^cERrKrSKf{nqIN1pI|GQa&1q#~3SP)EzADU@bkI
z@`X5eWa<T5cc3cW6e|;KX_^G^&ZUQb_Cr*}4)EMA#|H=&pQA2|5wi#gvxtBoJtpdg
zoSNiR_^oZ+ENL*j(SbO_eIP~?=u}EQ@P7I?u)^c%MEuf$t*aF(t`$tK6=6Bcr!!QL
z;}=EV_IKZo9yJd&F3#_`68OCxv)YMF`-t4U;$uFv&T}+OvG}0WES7Fzpx81<l37Z&
zN|JSq5#4pm3^UZz4iEpMU3Wu(H|qWY9cO^#8%2KWB|XCIj^{TT<qqXHLX&I=t6GF}
zUJt9)fXrQw4|#*x+s}t11P*ujt9Gvuij{EGXi&O2f{C%TwZiTI@crB|)w8uNLj%u2
z6F=!4Ka0Dmx+x*n5y5qi2B6U?u6juRTio`LfYTv-cSO$g=PI%5Ul~&}r&02UWa;G6
z){I}W**}lhqz?_<WVL^+9%9E+ARcmPlSCTRI=#AVlVq>zt>bK*1d#ujIZ1xf3EVyn
zUKL{}9In2Q9TjC6A+wXB93|+cP&+EO#-UlKp@FfATUexdv1$>qDs*EOH#No1Eq!8^
zAf_>3TN>CFOf!eK7P31IO=p&AE~_>+<(!<$zoerc1%1Lfj~sG_KNKpRTryGt(8uwY
zooeu>CvPFTU<l#z!9{F_2w^0}M4-6y34D1Xd36PIQ-FJ$hgVteKsg75yhDz!0Hqi5
z*#oB9`ICZ&jxfHs(m5k(-X(u1Cm5$o2v7YqoCKxpjIcx5)z;0NBJ4pJcA?+N7vs0Q
z1xk+|J`ozWWKCN_qeqsa_iV=<pq_vFR+ZS6!0nTDMGbk7Fug3IYh+Jjcg6+ro)>sA
z25^;w9ky7LFofqAJaq@O(qlZ-qnT!nL0k3F4@*JBZdqj=FVbM7{VPm}FfwkV=TVYO
zxiV`@(<azJr}deaVc7Z6(>qJ?7-O5DhtD~}=o_$^Z```cz9>?}PGfx1X@T9<ByUiC
znt`0$+r&o?u;0!>8G;I@!49<EY)MmZYgWjD@pyt;x^>M4Pp6=(&8lL^aLNF?aSZ|A
z)~pH-w+7ZHk8ny^#C(A|AV(6w4hT+s_Gzg|VQ1Qm?O|12(QZR7;10pHhfN7gf>R#s
zl5NOh=eh^#yac`&<lwSkH@<<oBwrBDOn71h8AmKa3>wkLdr-l5fNKO<Z|If?KLMC`
zA5AiXShqRCh!^r=Bs~F$@7E|~-f7p}Q>X(2ULl^+v54>Grs28=s{;#OVV<&+*PWVE
z$!;iiWII7}Lo?_7zTGp4gEY_N7kFL~?OJ@KrexjuUUQM@)z4vH^Rc!!f~a-SjN!lY
zS?jRd5xfggHoZ%~akibiW7y60%&fa7GDHo?b_`;}O}jkOrM%I9rqqGY?w@>U$mj-8
zdBggYb!kDw`aBCmijY;+y7nx{^-%bT?1s$W6=AfhjGSB`>ruQP2evq%W`i$#ADHM5
zaEAfRN1l26-@E%rvrD-{BHH@hoQa_AmhDudyDx|_Py`T7ok376E#V1dff%+p5WQ(!
zG|if<+2T3L2j#kFc`TJ3Pm6+F5~UA`Gyv+2MLEZkzAZY?BUwy}AY+D=No*#ao#eJ7
zn+!tyaYYR>lSawaaW2j(s7-O?mWonKDVa64P1*PPyrm9u(<QHajPIz^4*1iOf6>n^
z>T-3~hSGK?Uw>z|?@^ye+M@5BvhP)9m+XB5venunim!Q$-<C2DZy?P6okY(~n5vcp
zSTb#TRLYXx)b|8>fzQ&Vt*7&%$m8CP)m=I{r4AwGaq}3kS`zidQ-X4&9$uMed%JE<
znDgP1>Q)7;P^pyJjY)MEvQ3>Z>tT;ZKiaOSHhbp9mM?CurLC>3x7Kiz=Y2MyomW-q
zjqA;7mOA0_p*gKKCr)rBOyMPsRb-PZ34f=)-McNX$W2`TgJ-{4bD|6C5rFfO_@bhp
z+C4&T++eX>UTIXF%;@1yoJk#WWr(^?@L8^b95+EdY9MxOhK|Zfxj$-{I$|L8d$r1k
zP}=Yu*U>%uixIQpIsJVboiiGpV`?0TVpC94?#Vn3bxG^DU6}Kz+Ne%nxkqz#q1}_m
zmA-<~Vvo3C4bc)?WF+a5I=jkw6xqQQM`aF-8eJ5Ori4sekg8$naV2aTh1#j8mLRiw
zs;MKkqg97pt+-j7^=MR?Tw|PZ*(vOHMz<}~yHWN2fV^Fom*@efNvfofqg%lzN~N}3
zoQcNYd)z9r;;!3&o;r)#Ugx7QoRWE{tT63SEH}Cuv*wzoeAjb1u${tx7uRqY-g+om
zU&PefrX2;}wZ^|(`3PI1`MuX(8S4-LV`xfYqN;Ap>5j`hph^|NQ{@M#0q%{?dLw`5
z#uu72bUh}n1{*hcY8Cn8S0-xW4}_{E`-1hbWa_Yh&Hzv{EiS1N^NZnm27EZlP=7yb
zCK<(=z?nRb>8xc`B}rVSjgMy$>R`B)I8qzg);hZoNp~{#s*1yps&p7L#kNYxiO%Z+
zMou{i+B94*YtScZHIy~tdzCc<5@m3asXoiw2T8{ibP}buQOVnsyhs_zOA@87^C70<
zd8UfpxV(v4>}Bx`Y`9Z1n1U^-vn{JyW2kn`eQYYH%y#XRbY#05WU_-#kMJ)hth6w-
zEcD`R`O9YOo^Zt$Wxwh+)%XXmugUqlgpoS<*2Z*wvS98uAK5IQwc@wYm0m`HBAOR6
zXW6z%@#>TilO<VNv~8%`F_IJhsY%hSO6XXV)Y+xHtrI#N)lMDJDVeag7t#KnZLD+(
z_$LqsycQ2SH4z90X-oY3iUi(QY`glXN8i#6-h*dzxSiNu-5GfIfw3%(wXA8*<HC;f
zoyv^=@4A=^F9&A{7P$N+UbQWRe#wzs4|bSyo9kU$<pFCIV1V_3ORY3QX^2b3(uDP_
z{_zoOt~TIaH@K!VyK}N8&sxW^hRUD|m<>{qRKh=aaDUyJBcq7^nm1d7XN^N-aX!y#
zC2WIaZPR+oR%=HP2dpek`v~EE9rap3s)tN)`P5G7WTqH|Thzs6q1HRON(F~h2z)SA
zhh$b&%%qj7vbqXB(a1Ex=y*$=I7#U#iSS9o_C=Yxdhp(YT1-i$@w{ryy%AchbNa+y
zrxoq5_G@6wPhuO9u22SJZ-F77Xvqia&!A4VM_=LH*K}f*v8!aF8e@_7{XRqg09I9f
z=$V?bipg6?{zySx`}=!ehJMUfp-wrMetoG|w>pTLS`S-7+S8Yzdh5f6#H0a<Z<y+Z
z;TyJO5!P9j;3aO~l3S(W+Jb`TSCicP1n11HPt*oe;EEneA%aa`%2ehnv<~EBldrbX
z=lZS<ZqQrG30mvpB~xExp=p#VQq8~cDjxO<M_t8#$--^x-uuS!la}A1RNr1L9;I;}
z86%e);VZRnZ&a>pOMeqQFNclMH8+%!c*UD?ZRtVlRvzK>xLbHRw;qikHVbGs^Lri-
z2`sNkRzE(sqIUvNmu0H6xtc)E1}4j}JvW`!22v(JI~Cn_Ps3~5GajALjR})jo!wN$
z`9DI7$h9v`%u@b5o|X%lodin16l}ZD{;S4Dk3G|r*%nzb>DhhjWHOElNQ_=%fc)JM
z&H9S%j(n|8cYMgZ3XrNt!8Zg~-6ys-?fF3;amFb%<>{~ZLYz1ZFi1z^TRjtK^>|%k
z+9zi{o^`yx?@}Y`p*Fs6QjW%{5`BY+=G^oCC%$><1MjXhmU2Bk{>?C&Xd;nntdVLg
zkxHtOYNC;9b4#Awx$N+Zln3v;p2mcdG#LRT8A+%S3S5YyAo#|eCrxp(ktem0uAn+?
ztcVU+h=Nc6YT&>BvV?RF73VG*2k*Yji8WS6r1FWq&%Jinue0XPS6i=6$I}_iX2+7j
zNJ5Jf){Eln;hCKcBu=AkC$l7Ii9Z1kl3NKVVrWGX2p;(#Fnq<^Ca94IM`##4pM2a+
zLD)G$Y$bMp{qzt#ArAFoH10jYAm1y==Vvu%Q~0gdZbpnF-uKhRHR9UguGfbb35DWK
zR75h^9098Y8D(Iq<l;C`AtBiTDX};%rW9E`rE#cxXrhF8f@q=zId@@wNvLAi@H23w
z4rkZnZ%Nm*tzYS&MMO@-<b|vr?aXf-j%pA0V^M{9^m!zd<MFjSa*7z=)?zz=a$zpX
z(xbLvS)5@B=!hs3ME2W!x{+Zq$?&}7L?B9G>Pd-k<#>jQKT}l_!7yY|<M0!Pjlnde
z^zkM~B8}M};nf862`NWo#;6_{)ns+?tVewbEk}JxF2w;K!PVsaq}bzg<D!juI3!`?
z29Civq+{byj)~bsWfPc=soBIQ6C#cg+GJ|utd2R&iP*(4TO^moLC%Ta1bK)Nik0~j
zEGcmLm@G+^<(Zb{v1b(tndKRp<uRKTNuA|co)vL#7XA<x;h`3&HKjoSd9J|K`Gn@w
z)rByC8-Rr?XtucN)S4^Iw&Yv7Z9#AT@<Gp>_>**7P_Kj=P<KlAR@WTyg}p8FjlM1X
z&Au(=CzA!RJLP-B+LnW#4|;I%4DS_$U)o(Hevt7@?iKlO$mtW8S7Nuc?3DX~#iqdQ
zwCjP!CoZ?>dg=ZFq(_`TKW*KeE#MRPoh@s+cL>jjd4q-xzNb;?L5J*ZB#Q4>5+R6K
zye3noG(_22U|cfHrDvx6TVyl#{uGa~UvRMes%$mU(Jj9uElGq<3fD&2BA%;KC&6NI
zlhvZEtB$2{%_4*==sMX-1HZV8IAdyy6zzrFBHJpV4KWO7b7VJ$9}*+zS!qftLXNXP
zMhW@`B0yxBGzq4Jeo1Xz#WlufP;&$JiSnJ;%FNFY7n=KOEfS4X_;@<LnyNVj5!R7G
z0^43}`w{eLLkMQ^yx2c2(u=lzjml68!Aby+4Ur6)RDn3H;K*YTrOKmMx<nokPS0Ip
z2r+uVOlA-zW>NNbBc5Rd*y8#m`KIxqb(x?ge|4cb#;v&hhEt9o^L-1X`ub$!ZT-Yy
z59y|^kLSr-_6yc+poqP9?<<&_qZPDg<@0izep>HF_h~qQ?-Bz>^Y4lMH|p&MRnY}R
zA8<Sx`w=mwCM)=+3KVWy?2CW^SubUQPe(Egkp_@Bu74tr0N2q(^vNBgV={(?usFn#
z6-DtEuz#X7m~p%vXz~#^5a}zOfJ81343QI%<Q<AW!Ye)yDX!c%+Wzjs#IHW#GNNoT
zKLLn^S7$zn61j!4{3P^w@Zg0qz*0U<@ozzoP(IMfrU(4^FON4KiNYy><FKDBJ8}t;
zSc32(Duf8#gz^#fokZwhWjzs%Jjb{|h}ga$F2=&>CcX`)ol~Vb?a262_`7EFA4m*w
zN1%y6@Z-bfZ22vs{vh`WP%Zl|GUzTcNG;OnEz(G=MHguoX~qavhBAL9d}n-R5;qWC
z!RCahBVK7y|H1zOP$kzugw5>LfN$XjzQg-x?LwgU{l1z3Ui*T4d1P<S<&z|z?*m^W
zTd-T2$hOP2Dj!qYTUny`#GJ)W?!&>^`174$3n2#sT||WbQ6C8J8Um`FUo3H&soXpT
z)1wi^>4LZc+=+yg2;#E{x4k0t5Kj>Fz}6t}7AmxpX#ed$&#eidVu6#ZO6D?3-VL}T
z%YdDEk>W{{gb1ii$h1EJx#|wW!E$I**b?a~ArHqw(M?4&IW0C#b_4#`g5cE5lbgsI
zAS!4hk{<cBP^kBC3WeN0wTO1(TFjwTn`Avbqc;$e@D~X`OiwX3E8!$-QMC4?_+&tW
zs708(m0Wzc%9*e9eHD_kb@cj6#N{l@NNP8m;c&}E_roIB=$eOsT+eo{*_)BHA+(#B
z%+AHxw6jh1IgR*wgAZW4ZB_*@012flDY3v*qCoT^e|BM+l$%H<sagU1_62JJXWB=5
z$0eb}px29-r)T(FE16~ZtaaV$*B9upN@uw!W*FSu=i=A#B;51%?}$wF>Te_GlRJA|
z!R4PfZczp5S8Bbj#6X&@vGpWdP<Bb&yX&nt|545NlC`HGdfu_W=?k>!xQ&cK_piJ%
z)7KJ)zG~sGIT5cpk*~Zu)7L-b^SP2~^Q0tEpV^<;#*H=?=)YdGa|d-z>JniXo5-Uw
zaw0N(fVih-kyyL>>GVhYhUpBRzT<Vss9ZyI2thYoTyfp_P4W(|WL%Nm3_cHSA<z4_
z=hn^s)QGF}lPSR{!Vm?aNZK|~H$&5!N{RkRW{rful|`&j)>%OGLqu63SY|_wqpwCQ
zs@uC7EEWIyElay;hk7Z8b}5&#!44?=4Epol^yeLO0^A%7Ju&#v3~4nFI=guf%Jx~r
zhBv4F`ofrvKxW;|)nFGe@eOkHJ8PEif+VCq31(1M@fY_5!9_7jXE@#9S@eBp{8#c9
zikwYhR=DcTHr0zQs`ng<_uS9gv}?oG5N%5ieakHH9nSun$e{s#9delRjX=)6Pr!Z`
z_^NIH3V$8nKJ8gBch5IquNQPnj+kylU_{^m#V4M;U60^GGHI}w0XeIll=c*un**yw
z)G?UXkc6xg%th2k;jfy)lwOgpnL?hFHne=yImKtG`mIa-oP6r3!I7L&VQ!j+`kY&$
z?Bt(x(X>XGV^iXFYqmVrMjdN5oyqGzg7erLc&-@x7Zm-ozmsG~Kz#KYu-8Snl9a{&
zjwdu0J4l>&=80W(bt)t?5dqS}@9l6zNm|HU9F8iIR3@(6Hyrb&x&lrRLym5%89j*P
zNA9I^2_>-;c(4y^5UJDDB7#2f@?dshj*!LKM%+M)*?+T@Vli@aXu)_?j%vCw@Zp1h
z(Q-8ZdcX%k2TcP?1JMJb8=Cax=Yt1<2j&*Xs(}#wRp6crH$bjR<&42+6&@O!?7DPP
z!nWz}&a8Tol+c&*k)$ltDo#A%>izYb4_l!Lcl=Ui*??R%r!O&={F^L>tSew<@p&Rk
z5u7u&`C439;`hNdD1X5AmsfO}+ys2NB2|G#zqJL$OiJs5w*?O^kY_RJLncdvFA=Y-
zR>_G4_JLnwuN-&0AC+WIxT2JTC9PymERLQ)tWv_^62?r7I_yRN*@sH;W(3!t>s2F?
z@1*3oO)vR-q^$}88}fKE%8+h4%k&7W<07_@av8oCUL>n|!K17_3u4SLoH8OY-Br({
zC$LW}Yddw+3xhrLy*7U%|2TU)%+b3)M>iZ7F5q!S%4_;8L3Kyx#fdz>aKODWUYz#$
z_Eh%9`uRcaK)b|ANd+Kq;G&Ne;b!6kvL`jDDvTwIr9>#{1IFwYo%Ji4!{Czmit{AU
z({(-ZH&4&Y=Yj-mfeUIW;_`WfUOv7UZ{`H}v|L1TqLaCwLyK@L1W12LDgw*Sq*0aQ
zo`QAyyY!>`UxUnn?1JWk7W!dcHH7&$-Dsmt)``77QuN{dTPynTOxBCLJ6G`K{XYli
z{zh>v7r{NM(~4^GW7f14J|>Cu`Z3~rxt9t3U6d$M4u2{s@&TCZac!o>{y&@XaSn?!
zy1en^+rcQi#j5(i^<83kf4aNW6dP2#5(FxD!g8&KxojXuU#j-_)3jwPz(=4Xfs-`}
z|9$fF#@nXslzhW%L41JUu<rTp0q+U!o%ahbc3JZrWl6m<&vTBIw8_68mIK5P0<QZZ
z?tg#;7KG+m(CdvV^@o)PT_Oy*c%U|X=KR9l!dwC}^GK%o;O3y4C1^Jd;NI2+ga9g7
zqb*fnD~CF`X3YE<uvw5<;7{=R%!IkF7oOkV@a}Lgd^bE-dO*5hzChh1KKLJYz~5Ty
z#sHwV{E!)8UJBDtqJ5Vqs5qy1<zM_xU`AJ2#jzi1)jKE8FDZJ2{KVfda9>C`-YFlc
z-4Z)QV-wfoe~+M#!rr6#32XLFVa=1-?jRkLJ`g>%3ya6htjpleibIw~q6~fa>lV1)
z5OVPD!R8RZAg*lGMg4@>7cU>yK7Ul@Z~rZR>w{{bd4<Oq7Htq;{JmhjnnB%}E9Yki
zzqbH0(x=;OLR6grdPqX*RHI!%ZhC+(7O)i1+a`;X)H=mh3>u7i=UJf^SHYKZ0jek^
z8iBaG#z3|cmm^w98utNu+ko(D7vKV#gs{byh`}J_PB`R`FL1w7c$maqv`bip@t^?W
zgYNJMdJnrbMqcEbSlS0vl{-LtmLw`C#XPV-zz#Ogbc16Sr7-K$hD_*v4r)sj>-b!n
z19)H0xJ|ouFK$Wft*`!Rwc|$FrT*I*^=bV~UVi}Jh1Zu2_w$nT)Y?RR-X&^fzdBE`
zEM|ZAbQqy}w&}=I%ADk#`>g)RLpR`j4tbbA7C&zQ(M5VZf%Y4A-iu9N^8j15qMfJw
z<B9A^sJTkrPhPB6yyh_z?AM`UnH(8^XLB^<bMUCcE|7(!b*&>(<3T6W<ISjW)r+f@
z(}(?xAtkf0DKWz2Np0y|gui1eevx%`c^{jywBZ#(DRc3u+&r5{%|CYi;9C@O?c#*(
zZ5dHFa(j81zLp=hdHH(kAtYX~ytP{?Yq{4GP|=}t%xNL?%}3dTFx4XQhOp1v;1wsu
zpZp}AbDDL?``Xw4uK0#~{e0RBa?NV-k>drD&yT!~e%eWU|JnRIf`j1s%bB0L2aQ(j
zrP1&)&Gl}j_d#|JZYZl8KjDt<lTfG-%lIoWr(8v8Yi;0fB-=HhLdQ(|X@u&--~4`-
zeseDToxT0LQVdiZ{Cg*gA6TLX3)heLb8UL;1lOaFVmC&jK5r2zjX>jlC&+qi%*e`m
z?esJF<2A=!a#RLTa>aLAWijl&xip0ww3FZ)b>g>t>4jMIV!13F<j{lO9ph!C8_pMp
zWh?OGPOVWOwh4A-)0f?Qm$KaQNsJS{Id^$b10|6*m+cSR!+oyDy!jMscM|9vC@B@=
z(?cV!6Ru2s9A&d`T&|*!^s6eXGwP+!@E-Qo>(0j06eGe<^xnGD{qmk8os7EG7+3br
zJ8gr?`w`1k?Dp*`&sz9~-R=b6I()mcjH5$yP<Yd7z-k-q_IHSQRgaEx)vfBiOlXy4
zKjf=ef8m{HR(!ru*WUD+eCH?yczAS~@VcnTW1BR3=hwyY@}ZAvZQpxqle11rM$gh*
z_IQvl6?jj(Z2>IM?k}|%W1XMOAK*>y?qr|&1n9qAyqaAbFi(LT{i?d${=B=6(*%t2
z`A)OHUs1Svafig(#9c9p+X#t0x<@r!9pAkw_}w8ToxmHuD0sV`G5bAko~KBbz*rQr
z;M~_h?<)TQd1|d*e5pkDe|VXq$2A|cusvU(qPOs$PJK}oV}{hdzz@#_tiPKv&JFnZ
z%Z)5V?>$hNWo`UezqW4ga^owG+iOTRx3|Vt&zR>#Xt8Qn40rDt-p5<AL`$is+FOY&
z7<XjB`Dj1YJ)ynb9$v=EhDn>wX`6GN1vHbbbZ8l|jg+2uoTpc-fZ5vutoY+4J<jNM
zk{-7pf_go%D$a|ZFf19s-LSw<BHL+6V-PEJ(Yf_m3*p5iQVfsj6F1ENmY@JCx;8Ko
zTY4z3pLe*S;v0T9diF)!`V`Ii_J^GMYG3@NZoqu>$FSl*W~S6sz<-=@0>o+X(JsE>
zZEq~_E)KiLTP&T%`bhJ2w0`@}MStF^8P@ncfO88)(~!Xdk;?jKqJTsXgEQp&MT%oN
zfHIAAGNhq^CadGW)$h;(ABCqF(m?@57sJE{yBHzdItoBQ0XH_pRugzus|?mJQHS)E
zKr{TZ%CxMmBm$zvguxKU12r~E;lu}|0mB@L_-E3P(-Zb0ozH4FV(AaR-J=e##k+*F
z14SQ^8-w0`V+usEZi^;L!+|az{SuIg1zA4A5Rj4mIZe_pF8lKYmVSX*AebYXpXz6i
zhagEC%s5fXU)m8fK7#cMHQ5(7J}TwPCAh|4yV})`z#8<(aKsDQfLlO<3rKQ*LOK2$
z=qP>;i1QD~k-YXVROJAM46-Ec>UbqYxO`9q(|MTS43;E}ipWZY$_lK4P|NUuv+`JD
zdEOOb7DZwqWYz?0%Tm)|G)qDABDP>xOLEWZ_MBfj%OUeJ+kb?6mZRpye13f{$1w<i
z{t99VXZRy*jG%h*Bj5nt18u-sLk0@akNheW2FM2!1H_k$Q&$U<xirJF@k`;6BV9gV
zc({{h|8PXj3UP$Z3c}IPql+V{K2FWo=FR63Yzg@VpxKmhhe!5G>XCB&q|1^39@HMx
z!mkUv!V|x!x1(O_0Cfqy8El7FcBgjY(A#3qgkM;^U{^f@YU2+KpSX+NGIQ}XZ*)HR
zkDPM!h_qhuyyD5$JlJy4)*x8|$`{w`yNI}1(_~4`PZ(a&M~xcRXbVjq)_BF?iN+zw
zig-PWf2v^){qu5}B9>@Abv~q}9wrCOZv>x$48ffJOvidf`kb1LP*4S#?&J%jPH&7q
zK9oBI;NTGHV;5e>sp1Mmq6M&zSWyeu0;uZ8_+G8`ks7==bo`=48SV<z3NTK#;}@19
zcMvXsL1-cae!F#^`;eoI5omtf?g2m<>-V!@OH!Zp>)}<s;dU}ohwIz*rpZhl_(`NF
z$K>Hfhq%uX<qX8^#uwsx0i#6C)%?@Ka~<=#JI|YVM``IvBJa%p+pi+$F7#=2<fBt(
zqC}FUhTky>&WzZ_&&4<d#v=Ghv5{R=CEv>0Y}t(AQ&JWFk?UqSi}UyvjpQ80Pt;z)
zDC7ORptX26;ZDtNafFzEOcZ@lpQHI+Q9B{)C;h{D=PAJKuWjpvn&$|EuBIptS%J}C
zQ?{#ed)`Osqq@fZbf4W|&mvNwPJ0Wx4!ltul!f}(zroGA@q>7s>&&{o?Q+hp_!hP=
zhx`QHuiNIoUsJNrz6zPQ;YVd%<YKRY{J;73zs=?;yxF^v&ZjK_Vn+%U5xoJA6(Gs9
z`8*+ZTEkf&;Lz*_)1<kXvS<V$cfW0}JORAbn6De<5CGw7%*~O`T8Q%^tcKn#q8^IX
zWo%=s7xI!8bx1~CB7dr_ZtZ^d{Kw09`dTj(p)eSfPI{&FdCEc0S%g(5Buq$gVn0xE
z@<)GVCg6D4FSK|X>+^uY)>)FM@!#V~TI<T{)0ju>{?2z;1UVOwFhr2RLF0t|1O|On
z!6uyX#|u1(WEWEbEP2Povo+HJ%aS$WPFhX`wE5uSPEpKC>e(5C2h%94#?UqSs3a=u
z$(YTFs)L?~sV1TIG)d;cqT-G9^hu`(Two5iCS(f+&KL)*%@D36Z5$Q#9FU8UfFO#f
zzd0EXT@PLDQ<t+SZyY{2CHCjcA$)>)0@;7_DIX->NDj7;X7Zf?+R3w*3(}hc*f%t7
z^cnT^sDr)6gMrg1XvPrt4E&h=XY0ZwO6oDt2K1)TUW`Rzx-XiMQ@{BSt%I`}$JLlg
zDxut{8bD4AV1>Y)gE+YP_ekXOShzNz{w_FvpOshfUY$qOsAzS{xe0T%J#vY))3&yr
z_|||bRy-gb7!6aoY59>)79ZPwTUn!escf!XE(8?!EhgD^UI6?Mc5+C!b_t&OB`R}W
z94geE*hlVYQ2*H>Jlddq1&Agu5%sD?zFq&(fMv{E&&InRRk6r6x0p$*cci=gN4q(=
ziywX^>+;?M-d3pF-B!%=N|)ZgHhfOX11fjV+r;;dYZvtlQOJkbyY}E**x2`CY?yu0
z^ARXl2xr5H$C}?=)zed@pmB;T<vhu{;31S@_R)W`S}!fGCj9KNc4gLS>yoy(f_OjL
z{J4CINf_6qa4)m>1-=hu74fd$<#u3br@u_X=eJR1o7881LUQYclqVMV^sS?v)OWDG
z;mX*B8m{DRaqyv=>3(h)&ry>`D0@z6e63Yp4^mG^>$fCCDZ2%Wa;;7=ap>>ta78Da
z-9AMy2LvIY#|!T>F1KY8j_gMBWD4RJMI@5lWz_JD{Gx?IdGL@E_!a+!km<^(BMlMm
zWJ`7XHB4~(6|C6JUw~q^qlhY7-LZrk5-C*dOsO4?P8lgrN9y_%t{!*vSZ(r>LZ<ci
zj{<DXl$_3;n_M$*)$OUt=-UL10Pz9Dbm#|)Azwd^@K<Xj6wVZo9)tnf<+q}rtr*68
z0l?%3!-$l$2Fg<#1*`|_y7n&1F#O*R%Z7DCA0ZcyWM}Ynw3|<y5As?%3tL#Y592aH
zdtYz-GHBa3jN*U$tOB&_^pOP$*<w+Ju++ye(0M1#=Co%YOL!!QRhCV7)0^F`B)wx^
zc;+7Kl<b9|Kgpi#X}5S%ggs~JX#%=$eTnn<5^nB3?|uK%*n0*w;RWx)Dxv}kBE1Mo
zm5%g|(mT>Sh!9%nBtQrlQF^Zl9jQT@5RlLUKan1!_bQ#xOX%(Lf8YDz-f#EQ&g?UD
zX3p-+nLX$1J`W1;+$P%v`sK2Am+%PiJXQTg@-r5;j=?P`8Z;|gV*f5AC;BVV>F%Qf
zpl9sUUB>tWo*jOuA<vb6&l09=7hUPP6jT;D_(S&{LQXNKbvd?lJ&%><^oS{yqH&P$
znOaYYwTF9Hw;+D`L*I}1D8J%IAAK1^OoK<CZi*Gc44cD|A<sTO`n39>lzu$KY?^0&
zo>7yDT^>nR#oLA&KWHAQbI?-A^>1xEcaE%0eEJBkshh{4o6PcdnXJnt?eyQr54pc(
zT4muET&5ZV#4`sKIqQgazi~U~oIN9jR_Wm?Z#1^W4e|N$7V0)7vr!(puztCFjaKum
zv+J?xPL*w=cW)j!jM{w@%=z+5*}uMu4JJO`0`$c;H0VZJDXv|5H*)^TmK!d@S1duq
z_Q<q=C)la7YV7f!2XQ$N!6%o}0S{O&_+7|PRer}US73w3n4}qA@4(OLDG)OOUATCH
zQ-=Z$sUFZe6G&jI{ui{h+%fcU-s%Kkv0%WvF_n2RqKnbx1Mk64Z%GrYT=(9Xn0Ec_
zI%&7~bm9pekh=(NN3!QP#eOAlt`>e#k=u13Al1cwMgMqLhKy3ydDw43??UawcOW(=
z+4&{L|5xGqN@vopu*@81Vmtl!<^Dzu8vC*l>3RCCJ-&_IY8H#19l(>|H`m`b&XThE
z>mf8Ubo8mBiQi1I?S9~b^``&6kqIu%$7h=*Yo0!nY^@3aMrQY?O$Yg+)}Kx=u%P>;
zRXPc$h;qa)TcAwR?1HK*(B{kL(1iYHSF(F))Qy8aanZRCq1w%alf%)i9;fi5N$2?S
zT{`_y>~m?{&~#9xTH_dSwq-8dD;+iBQE}i?oa+OaElQbh;Z)A@6VQzyM{w9<FtAuw
z9Sg<{Iq5-Fo1j<y3@_t*t}RfX1ApxKr)HD5JeR46?lHTYcJ4e&m-vycDRXh4z28iA
z)ZQ+=u!x8%#Mu0We5-HALRc_qrm*?HP5!d;L@Htl<o9IHrQWHw-0yzPdSB{L`yKHQ
z!l!Ix4{8fZzrQBCkFu5a+QQksBc;=B32n0LX7zXBeU*Mi^A5hB%iRMbEa~z8VtuA>
z)P|6HQa4;sb=aV4#4c0nyt4kq5EXf)TR2NRnCP}*^IR;+L~mE;OWYOlclM@k_x96!
zpcPoML*e?3G1b)LL86~!QH=LCdbaM#LI}Qdwi0|e6nhokOfZ;8{UxrMWH9mcp46e-
zJ5~OBkCMpl5%h$A30`6S@0~Z*_q%@R+7*f~VOLDwU*|p$Ss}ak?40$T;?2Wuj$6`h
z4j2dezp|MxjKhT8c~=-}4LU#s@<Vy4GrhKAn$BYuMy1DzuBC+<`2}VoB(`FIRZ5D?
zLnKAQ?ez#mU-cC=JZX4r#ueD5u_$Ux469C(D*0C8{F+b6R4Mv-KuP$9^IJX@Qw7cE
z@L~NBK<!goNj}wXOwYFvW;6D{F6TC27RObGGyeX?J-xw2kc7*j=!IyKLhJk1U}TGW
zD-rSOOCFpe{&k9IRGb12Q$hUVE=cMn<?eeXw!p4$Art-y`!r=6Dvjjlfok$fGXf)6
zwOjJ*CkH%u;w2L<f1RI0zzCSuZ%9D4wdxhc1<_JPE!Dxo$O7x<Z}ac23@*5q+-jd4
z$PdcD`kXwkbQO<}k3f(3w%;n<hWJ0;c@+S9z%CU=a$pFL63i}2b`?u_#ztRk&z;nd
zs@EPE<z?{V^LeHHnuo=#4VK;WfLBS#l#t|`bw}xQ71M`{&&3`t3W_~l^b})Ud=ggO
z>0<h<Ezy;&&D=P;DsJqqERK*3d7)aZ#}+Ev$Mh&jgh6|B+hW@yBHq?cjT9Op`*o<j
z8=^*e`<Hzep?~1V^(uQ-y6A-mVl5i2InrTKWLUn2N1bXE?1dFaQL#mzm!0uP>TieC
zx2NYnsHV`Ty{t&Ey;n0x>7K{MSolb)jSm!}yzvA*NKTr^x~-u4^lO+uEoJ3PnL$#!
zgf;Rns{bFxl*j5F8YlSV<szk2+Z&Kl+#tDB9&3z(s(qsRJ>hevR~G)GqiqGnoHh4`
zl6CkOtwu=p%<)(Hd4s&9PH@=!2FuLJP%$qOw<YrFcD~w33nvQXAcsnxHlso&?f(x@
zPq%A-Ch<%bzwdrX{hxLm!K?zdJ{0an!naIq-pN-RW8p-H9u#uV1N>jW>)IGaC%UFV
zA>TZ}wgPrC6kbijXGdeMLc(|CSlCLq|HNwfd1L2mP^{txefXd#Qyx%8VZuM8^571j
zZtUa(#ae98qyG<}hpLgoWFx~e59r-7pQ7o+_-!)Jn+Rj~{L+v=FX*Ylr9|k77xB(A
z<#ZQUZM=aKGjdQ)@{>T>hmBVSlUQ!bj5nH2`Mi0Ns1Q=`Cok>ErMme*2^JgQEvsRx
zFEYAci448U3-S&*=^>xyRNXeCB7SL4d1HdvZ*?HP1%~Z6I}i^5RkmXtcYTaVa?4b<
zy&5*6@yZ(#Hx^<9A=C^?=wX`#7O#L(wQW#CyN*fTpu2J2oU1~US*ULf(T)Sv^k0tJ
zOd}`p%>O4~2!_0k^XJ@Ax*EU_L@eZ^pWl+HL_Ajb9bw@l`?z-g-KX#t1C~mF`|RrU
zU$$BcrdkU^FO(oCE8GaB(cVyeROMnED>326B$oC}B?~A^#?Twa(htrJm;I0FT^8^i
z*(LoY{gNeFz~sc#1o`^(;GdS?Ex{5KkxVDrC)(`|4)_Sndm7LZYdech236!s1;pfQ
zUdo(3!-FutmbUz==QVA`Rj*g%Jbvv%fFT1Ge1Y><CW|@>;XyBr`_-H(&7I1noVY8I
z_3$O?K2xV6uR)dNywJ_OAdYuiSBx3U%smc=y7nZl$rV)^pWoLhB0#MbXbN+Cnsbw3
zrJH9N_GF-L>qFp1;=(&v^CQfG6rK>4y!Ub=*zfJlLRbCmz2$4xn4|c&EwFg6>&G$c
z^0tgV7K>uIwh{nsg>zuarx@<HVl{6$d<3a|-aa~ddqONYP5c%mL#xM5s}7)jPem(C
z9+PC7h77lJ4J5EzdLQ#uFD)Et_fY13PcDi2_4A}Na<+YLwmmzwS<Zyo1;sYGQ2Xoq
zFOF5!hI5@BBgoVqvwa)*mPYuNKtT+|8;(m4rZeLgu9k6dIPYsK@9QK^%a#-rZ!{$@
z?7~K+$EOO)+n2-;nn{uzn7v4%2u^wWO)2KxUc8d-M5G@Hb7$ZwU!PC>bE*0u!ENko
zbz0~|s$I%*Dn+V6%1CNW%25hQYIF)Vg(nr5@~k(um#=rHH@kPP*J1VjD)}nWs`IMk
zD*LMF>N7yoAvH$}rBRqh#c)|4cT;c0VVOc{giF0kp-al$^NY2px~P$;qNs(a&2hO)
zwM&Cbs!N3n#wFRMtW@up<}cM>x?nZ1Ay^-*0=5R5fvv`hH$TjUy-k3prWf@HIhdic
z%;wAA8A|}w#v{t7%ZJRCQ5^Q#{TXmkVW9i?*Yb4~+@Ye|Jo^_vP;q>%+|5iHm4P~O
z*cp(@0`qPej!%_Go0*`@Yup{c4&Wcjqd1pTbEMc0Hy6Y-lC-pJ^v|g1rqnU*vH!8s
zG37DnnD<!bSmBuESm;>em>&EspdcVRpeX<ufNnGk`VgcPq!45tq!Xkbq#dL<SJ{%P
zjWR$9qsFE#Tz%?fXR>z&NCB#-aL4#Qrc_z6CON47$>=uSRLxy`mwYBh0tz`P-TpPT
zRU;1Y`0L&`pNipx+MMKV7ff;14A|!mv}6UzHt{vFUe<*<4~wMJ$99%fJ7v-T<oHem
z40BR(auF>MT^6AbNfsRu5f=3kB@r2LhPxO$83V_i#hk=|b<Vy{86r^8OVM!AE)k@N
zqbQB2iAY68W`>S7LQFlg(#RQXo;+R*b`F~UUZ)|gWl>Nu5;qh#p);y8q%+Y!+&{uU
z%0J9MQaM^VB<3XUf_6d!{apNjnZF+YQmy5zMgCf^Mbu8zir4N;TS(<=4goesr7->F
zHCwY%t^9@28(~{2n<}`@%}sDJ$N(%1ssSH?2R5m<;9JJH@y+oqF<jjyJ-7)31zm!|
zL8@R*up?+4i~vu7-5N>fb{OHw@Z|J@p0{hJ-dU3KS?_ct9M$?GvZk|!B$vH8R<!!l
z;ljd>?)_i0*1h3t<=tl4HT;f>{cBlnlG5H8-Y09@1Cm)jyz7SjQ(4iHCf?>g?rWem
z5H1;y3rMv_isRe@5c5c~%`*HSyy%70HSM+kwbC`^HRzi6TIO2en&n#PTH~5N=-cOl
z&(WWoJ|jP)Tg+e|U`j9rm^n-brVi7FDK1pD0$zntDu%sx8)WEv)b!hJpFtrk!nWSF
z&^E=kJa1mGM{rE=kKl~p!hX4JwQYlKs%?cW#x~iuEN`d^Th&!H;Mwgt>N)J$={fH?
z<vBO#upu}rCS;fKcBJGjeJpjs$cE^n7~AFN%TMeoylTC&yuNy&yfVCsy(+ze4<wiM
zmx-5MmZg>-EjuiWEtA?Md;RokJ*YZJJ4iVwKgd6bsvh$nnRMRn)$Y|EFk0R8)^4b?
zG^rF2PqRps=0$U$`TewZ&t@YV47`!+{k_?UCP6>FUDRx91JIknTYSyEXD+8n*w4@1
zXZQCkYXhgZ?s^UulLF;CRdI*z@=?4kN)ys@NuIYCo6>eqUZ50G*7~E(s$HOEr+r+k
zUi_n&u6U~Wyx37YS{tifrUlgA($3Vf(2gy(DwZt{Dpo8mC{8XGE<P&87H<`c8HD3}
znXsBzY~RQV-4QROU&h1wl|Ys_n}?FM+x$ke7lh)FV_h4YnJzBw@qXS1qIqPzQqgaL
zK)?3znXhYp3DJf?k51NK_DjS4xc6M=JDPbQ6%g?wsiVd|^)*mUZ@F-0pI8lcTln!`
zqcz(#dY^1a45S&74~c>_Lg0`bNFF2+k`5_?#5VkH9%}Aro@*Xyo|$!(bCUZcXDjC>
z=h$N3;?QF2F22yqdCqlCcfJ`I)FQc%v)6yR9)8Gr{^v+_ExwmI(<WD*3+8eYb2W8-
zbd>woCaYBjW^^-sb$p(Ah+Y$4+v{COZWXxwa5Hq3eC~KSux``enzI03NQBL_@jAn`
zqB}p;n)2#IPrRrK;Zx*g<^2U8f_uS<;5~3>csjfRP6%J;)#T&hE#uqbW#G-^o8gt_
z)rMo>*YH1ZJGcS71|ALPfg5ABHIYTw40~;KX}MHck4sHsxg2oN9PrYiqI%q{%e2da
z(;R7vv~V<AH={A9F+-S{m`|88lrvOVl+RR1S2R`p9sf4YI!-Y@GVU|3ISw3W828=c
z<8{+_r4#$5?IAp~Er$N<K{CZsw|GZ7pm#?*5HQ0wV>t74#&srerlzi=Zne&+?pNJ$
zopxPCoyrXV4AYF<%<CDw8TA>u8ORJ^25ga8@yEC63soVNJ+%{Y3?cl1V=(&Fhv)pX
z`hxl2>pp^h*MB7<OeQI&>f=qbsQUh|;cF2g53w;#InU&Ws)mmizvI3Zf5i}9l919c
z@<<qY*8ThRJN4`M*G$4}LaP%tXrmQiVoKXLa}IDi_3ecFQV<%8>6h!2o0l7s>yVq1
z!^#b`47PN)^tBAPbk2WVa9a4Z;J)Cz;O0@sT+3XYTa{avTVsQ=sadE(&v-~iuy!Oa
zevrnK*8JJ^l)j2fU|pi!rNz;%(r%UbX?bgb`#k%6`}}0KOt(#UO;=5~O!p>NCO4fx
z+uqw=+dhomjNmL#mRFXarLU!5t)vtqI=e4@tq<gqrGo6hfdaDcJ5EZrOSYM9G43#~
zXZ_3C&RWaDWo=ph3+!(gZD?zlZs>3L-7wYgx8cv!%Iwzcp3Jt)n#`)qR`W*lk}G%~
z<SW%goJ>4IViek_V31caXtcrel-ihzl&74loQKgsqQ&#mVoDn>w1#+{il5igu)f?A
zu$a*n))pITL5!x^<(;*fwDhmqS6hs18wp(}7ND}`Y1W&x_9xMdr!)!E+!=1_<8Q+5
z95yLbMgZ%9g}@YGdD*;ZkLZ}_AJG}nh2wHyHLwAg3akKPfXTqJvY}ttUtPZjz}?_c
z@G!U&JP)1%&yB5Zdd~5GQg&8OSI%IuODyU(Tj3+742=$>jD1teM{gxk>7+2UlJBf6
z9KA-(_a>my$$2O$e@uBRN{rHtf7-aVpyY>>$IwmwALXtn9ja`;MpIBx$*&CuJx4u+
zbWirKaRb;rd1lSGA!zq|Q?G2F)pDgzyjQ%RE=bo)*Kfdkz()Wi;4R?u6ZF$dyk5K!
zU5{?^YxHZ%oO?W{I?FkWoLiqo%udXT&+hD6FfFMc*!zz%VY;@fp|eb_oGX$3VNjI-
z6-Z~mrfjl|fvm7hjqH)^Kmauq4mF012aH3-Aaw!svQ08jnM;{)8C6+MSx1?5S%mC_
ztXmH0fSbTkF+><ce|G^rkiuAd{Ks%YdJwli9-s>Cf^hnCLXi;1fb{?x|J)<~<I4TY
zlgcCaefJahquqTm$Z5bSlnfdV(S-~I2n77}x4?*Fvga$83YYMojH)Xq4PAMBsV<|w
zbr^O0i2_D#MRife3*?x}6!`}nlrKR#NTO+|1bD2?nv3i6{Ou0lOFJFX(S9`IG|aqm
z`hodppo4}b&5nsE4@xx6!Xdd$l)xY42`<PXcW(XUJm|GeyZEg5uz0WdxXE7TO6F$f
zV&+cf9&{W%iJnFGqQ}wm=pOXg#Kz3(%<{~-^osPRG){V1dar4_X%~0|K66J@CDkVN
zz0i)VZr5MpJ%B&dXk%B@P)Qku$MDmc2<6#(_tYk(v-@aLr^uL~YVDPKHj}{Y3>snx
zSDq=oVM7$!9@ev%WXF!w=!aYKHs}wVd)E}!6k>B%bWh8!q*{ACj+(CIV1qWR%xgan
zCeFLgx-K~{kY~tC$Mf}bnhTn9#JS1E#2LdC!?nfL%(e7&)Aiq*Z#S$r6gML`J~x^-
zz#E1e-`fg({S5sKy;A-$M26;0(XKwsxkN-?su^UDD;cgXIxb4qK*IaqWLrtK;-`i?
zim;j{sg-<IHjO_(KxaE#<*;ngr3P6l6Q8VEKrucLw>rF4<W}^u2u;A7K;8Y+=|Sfg
z_J2?P34-5A2(k&Xd|;q`NlTOVJdH7pI*l%k!SuO5f0alTOO;d=f7R<M$tu1o)<Mci
zhDl}#MhR*ODhY;q+WHqx&t{*39yRtR^$ueV0gXZi=@orOYdj=yqts_SWn5)E3<hE?
zuAin-P&hzM?^-G!uZ3Y<xhr5Q0~Lmf1z7afA$E9Yti~;UtG3mqB2goN^<Mr|E1o93
zacf^5wXa;iW4=q3kY#*3`7XUd7WOsq-~PYue<lAWKL&Z~d%Ae4c=~#N-k<un^zZOr
z>A&fJ7ypX?jeoq8)AH2ywAuf#@3e2fueBez@4X*T#WzXbY}Ma8(kqM=UG4PNY^a&F
zV0uG_Amd`zv$l|H475BCIqX>->2-t;@Y#I0u{<9>q*^V*PNaIgS+H!W2vmg_oX;O#
zuZCmeQ**iHEk0LYX+VJIe-HZ)D?jMJku4;~DHJF0tq`9p{Qh9%WLm;uNJCFS&zwP<
z@gjpMgDQi@oYtJuoZ6huoWkXa+fz3pH?p0lJEWrFq+xL(Iw3mY{h|D!m7(q-(~4+r
zHDwKDmAjMF#<sG`q@L%BDlbZ{V?Qd9s(`!Gc+^CzG|l{N!6)C2t0+evY7^d{rbg9u
zpMpAyqbjIkshYS?47bZ9483ZX>^st<j;Ihkl_s@DAj4VG8c%Zx;c_B|KkzEy?PrhM
z7WWhOBRt*wLH?Hh3jW^y=Kc=;?*8x38uo|xqxLuV8~4NZ*Y@l7<2()gJ^W#3o@b_K
zhG(v4pUyPuMFWLr>2^!CO0|9(8f|iGr`Bnh6#W!3H8YYdM-`(gy|q@)dLvT}xRECP
zrCBBoRo;55sNU36AU6ZIxUu_>S?h)xZ$E&~>hE6GR8DSPlbo_<LrBG;3IMuV!L6UF
zpQ=|>Ic$<%wr#Y_4Kpg?))#Gdu^%fbtu{I~N-ixn{#%k^CEB_<RoY>MEi*Q%s4%zr
z^}_{dWH(k>Dr<CEN^117LdpzWycsyQTDoQAX7tkNP8knu4tDreJZ<Xs!`B#EMliOx
z`Ff1C6j-Ke5?_|{;f}V`{J3wkVT`0Sx6IlclOF&+Zrpq}mRW{2+ADD{owvZWUU(i?
zZyJm_mI)Z!{BW;mF`F}+qf(D5R5c{GYNxDLwj{S<yZb<Y8a=AC$zsQ{%u2y(z%s&G
z!*awz!WzwjW#M53vOG(SP2@}5Nz6{1OI+*z-JRGS*^TRN=q~F1)g7w_T^au<Bmwa3
z{CNEF`s2~Z%a5lYZ~qPcdy*HkH!*B~z;oeh;jOrhzT#R&azU588d=zNOelMYJ(7xG
z3MvDXT$)PFmMVKms|LIU%!pMV86D0*n)G3>q52qp5*!aIu(;<iouh!JC&y(SRHx%b
z1x{lv8(=iMn6N)@=3iF)@g!b#IV(q>YnP+MF&~fFG~?j$D~XZgc;Bj(GISb|)unO=
z%b)}hNrouAJbd+N`R2TbqkCAM)ti=oNmHV$aqHDjGNfh`q`O?`!pZF7htP$gj&A>R
zoN|AZ!+T~XU0uKK3%_vQ+%CZRm6{4O7qBC|`DW-MflQh7PXXTl!a6UeUuZWWImTbk
zZ`drhTuu}f*cq|^%_(anIp+l~H?as->~XPMU!JqqeO6`GnJW&!``LrcWN{4nO0U;l
zlQvX`q+Km<13yrn=sFT!k2r^y&q_umDkz*;u<5dUZ0G4x;~#W5G2I2gDkKs=5Hw^E
zO)>b4Xv3fp*K>t<;@k@&A7~=MAk<ptXk@JiLIU<aZ|n%ku?^2)@YY$mk@(5$A;``3
z@H1Jw1|3GfSe{}7b6IfYZHHG15W1gBvDtH(C!ceL$E+qd*C1QxBS)x&t`k1Ua}NHo
zp_a{Kx=Gr=99PA7MB?{@?iPaC$`RrB5z4xSXfsTVHH-#FftykCKBGeFmV{GuA6*wA
za6JNCw;>l|iB_PWIEudlk)61pBE-QuPtPs?FsHdH8;Wd?2UCtA8u1$ow7n5vzdef{
zmyX<?glaE9bD{>8F~^RUx*WC4FK~;kE&=IQAQ~Li06tfKCgZ=jprTvwho9$wfMyu9
zf=74>Z@u*(w!BpY75JiDk)l#S#&3mL`&RF(_nzpj5)QCMdPt5Osce07`*3R`WUH7!
zNbXcapa46v0KAFZI8D=S0|$MD+?MGK%+qWk2w1MeQOD662ty+ge$CwfZWEj^Q<^u>
zqWBSLQR~Fjos`In3ZL8DJ4im+Q})IPxreFb+BKWNqZk0b-ucIRESE<u!$)Ah*5j%3
ziBesbTT>QTx7GPy;ul4Tzcxfyu8~*Vn32<5-6gzDmb_ca5JT%^t17x@*y9p=Y|A3A
zm4EE2aI2LM-t|mxNs3=U+h3GUoTV1+iMH6|0Lwgi*$Mx_KL0%Qh2bs{$(Kp%6+#zG
zQ&(r~LsXB<WQ|T05mSpqBVFf;$cPNu`~&}sLe3?8;eilDEkQ2vI{nS{+z`dsQS=rp
zclD&4X2abKwu&9bUL^LetXq%N@CjWQv@M^8&;m|17PuLZr-{0IUTn#9)$=CXv7H@P
zhn4=DJb5JYD<)~MtQ=3Plc+v(sPfx9Wb1Ii`4s()3m(*PEl{AFRFsQ18U(EBY;GYp
z<6vf>^xW5a0l%3$!kjF{;{Cce9;MIU4B%~SFY1Gar7?YRJJ~n5hdAg)6b(RXT{%4{
z^$Y)ieB`Y11zY-j)A8-Sphm5JBupLO1^8U2gbP1C&b$WTLvD%&=347V+?!3L_~1i7
ziEsPdY;zZQFiLBxjq!n$y4aN+++Eri@1lZRe2AlBpjc%+@=3Zmy1h-3W6`_thbHH~
z*edRKvV^xNi!s?WGdK3_CJi~)@CIRR5^plf<-8mFn3X$Ph$!Zf0`dGjCMu3?s87rt
z*JSGPT?3e^e*e3k?AyDIfCJ<NP7&6FmzuYruXtTQByuZsjl4$CiW!}ZqY#pWD5xNI
zmIuk~@dyB3Qs^Qcdak*QcOkEf@{Eb$Kh^!rR?s?u^^nZ7X&oe`lMTLR9wVrRRItC7
zQv|H`ZbQsjHDMMfhMBzASq}pf_I!xs?#74C=C?G4dvN6mBL$a{bI@p1pcwXWfu+DE
zcCUD-EUR*oPlNorBY#$oe`6p5`C^~bR!+CKH}!NP#QK6w=V!_gCr3hA7XaTyF(@9j
z?<>N;X)uYNm`q0D1$qu4>MFN+F*ko6Uf=jHe74uj-vZ9Fz6}WdPNTXk)TpY$%w2%p
z7MIE%DO%2700QtmDEz8WKy{3qQNN>(iD05`(5(6QV-(&SDypu6*t$gm@LWO{lyp|(
ztt+Zo^7^XHv$WUBbn*qhmmUSU<#Rl-P12A}$~r5LUsX(tj@rkT`pbfc__@O?LX3F~
z>n|;x(Q9<HPQp+ZcEkI#PFw(`x`rRDdMC(Yf(PSzt(i-HJ^#Ku0y~ZaC3UaM>TnKV
zHy2)G%{b62VZ5+(kMOmg4b)Mth{;_d2=r4oej_@p(yiI|h%O$H>%{+DDmdgoC^m}g
zIA$&%J|COcldB4oK0C6#J$$$IVo)<+;|jL8vt&aw%chmoA&?Dh7gz-LJUJkdTVPpM
z4?g?*EZ%>4PSfFDp#AIbA-pVG5xF7>u`3Mr6p+*yme#NBT5QmbKla4g_ji1w`}2Ib
z@+{~Uf5r&Q2c;02@>%tFqFVEAyO{vG!!QS*Nyf(bpj<xW&uXe|&f3uHl<O30fV>?8
zPV{DL8zi%y&{nWd&ffC~K<UX_#Uw#=SLqT+Fi0?;m7le;tH23N@`8ZDTSJ@LXDvoo
z1`S`g+fHReK69Li%TXZ38e0UGp0QX8v~iGs=53o$s#k@t1sxccJvrZt1c!eL3_PK?
zAqqOFmVmu(Ce#PDprRK{h@xzSbQf($2Plu&^e!-_nXRnT+wAE|gi^OQ$B9c5TwK#S
zL%vaU2DRDwJCtrnCu!xGJJ+D^*Z*-H*1@QDQqT$f1;Xkttin&8D$Q}^Sqpzao|g=K
z8U;{oRx@2K5YO*DSU62oTzMm@-k)%^BGjNWt@gM<n@+nR-B2-hR3Up*lR!i(e2}~r
z9#_1)5#pZByU>)FTa}bsk+kPD+knV>lLvSb&LKM=E@wDc5rnH|ikgO_Ml7Q^$m{Tb
zb3e<|6Zdx;tN-vAc`2i1t?D#w&&2E*-8F@6wmcCMK)hT(NJBS#d$zXt68e@vojYBj
zMYBcKA(45oXz7;Ox)6Iw$LGiMmJ*``C<x6%zqyF+R3W)+6)|sak30;I9!d(U{*Lpz
zg%5XqP*m2P!6^OM4<;~MGD&m^E(l)!Lh&Uzc;t(4uum|_mx23;2PUs3#@^wHr@riE
z249(2`(2HhGrNi3#sVsN{F{>fU8><LVuZ@fy1~BxTKW56GX`VQsvE93-t;+I9kMr?
zf3M`@0|zCl4DU2Rtp5R12k4dR!wnkE8~nCwChC8l*Tjc*Nv6hiq(M`3@?ipJbgAX7
z@m}JJSNh&mp0o)A-@xx$eqz@HT6Fu|nB^A!KIkXlf5`uU|MKvRc76KxK2$cu^nY_#
Y{f`+d<b3i!c0%3Y_5auz$-VCX1q|0HEC2ui

literal 0
HcmV?d00001

diff --git a/studio/frontend/public/fonts/Hellix-Regular.woff b/studio/frontend/public/fonts/Hellix-Regular.woff
new file mode 100644
index 0000000000000000000000000000000000000000..683aa71fa8a105192ac1c638c52b96f2b121ed65
GIT binary patch
literal 58584
zcmZsBV{|1=7w(B|+qP{^Y)x$2=EQa;wv&l%J2SEE<iy6!``sV+&#kqd?ylX>R;{Y8
z>aO-sl#~R30KUt?3IOT*1oDafU+w=_Bqi14zj<BX_wfHi1eMjXl(>W>0FZ_JU03^t
zHKsg{f|Qb?DgaP^0RZsL0syJtyb+~?QmX2r-#k+Q0Pz<9fOquF!DXYU%EbEZaN%1|
z;Twi->TRqhc1Df>0Q9##&>jE)?CMWX6sv`i%eO2G`kV7Ve0>4HENs2Z0RWWmI6R2~
zAPCQm0`aJpW=7xkLA1Vkoc{r9E<wxkTlg(Y_^y+DgB(!>4%gDo)$`jvIRF503IM>j
zM5Rwl**chf^I+!y0LUNTFodH0v1e!G`Rx~^>06KJ8^mCq0A+h4J2L<vlK)%(J2$8?
zG_H4LM+X;I03Ze)06=&F00>^Use2p#*QOvKQV#CRI0i`k4*<CTLI(hrf2jT!2Vk8C
z*!?d0DFbr>{oPLAd4c{X{my31#L(2x(B^A-hC90IEB;^adA0*~An913HWZ*)94het
zah5EuO$;pnAe6-B;G&R_P|SHC0Hl7v^#5!a(U@lU^-c8kSx-5q1qB&lhMMzECXAqi
z9)uJd2IhdkgnEa+v2ZhASLRy$UuR;8e^kDStkQ8eIR6NVOZxgi0EXoV75~LllWr%y
zsEI!s@bo_WwZ(@+iW*iS(5_BON{T|t%Bp>C#LXEA4TI7y&6Ko~rc(QTnHVkobB>g<
zk@}aoG?RF+`d!!xm<E$C%04NI&%F1ac99B?yzTZUhgLox?2CbP;7i`<_8kuwF9pl4
zh42ZDCpp1U#Lzo#MT=j<VqQ|)HdRf5Mr5Y>DHp?dxW4G)Y9p$Je>e|?ddEmrqAwo_
z)Rc);OHIS}Rf_y$Q_9+<ihb*k?kCZufxJL#56MK=m3`D@zm)E>TO(7b<0Y8_f@G4f
zil*01R@2r%ejHe-7eOkqY64cM)jO6C`TJQ`4`e~|nhwSd2r6tbcOO0Lcft{u67mia
znG5&~ut61*4xUE(myLTFg14ayTpEtU>rnRKHao`+%g*BYm>+*^3`CEWdGK-NXV&U%
zM#-9|O19&Qmsw^vjDRjG4(3{ACRHWmU8*u&JO}#YFQjhzS~6lLQ*g311dc7PEgr`T
zg`<#5BvmsvT={p^Lj%wK{0TB&V79)>d)7XecM027DNKuLr3On`{~9Jv2Uh)~FPt7{
zlC`KA!_5EIeDQYa!G2h%U{|E5Z#g{UbTWlYv&ArF6R{$Dr^wX<ah^4q&poIdApOUF
z)ugmF+(|%YGfD0$pDM0*v)CP%xSQhIOTUnyJ?}6+3PEf`KHWI&ClNI`ac&0CTvh?j
zq9f5e+x9m!GwyP(5F8ARQ*^D=K60k4^J-fgM;d1W2yCj?WyjaAVV^-cucT&8t|2^;
zKXC~Ap)fxQu9iHT_)N!ak_WTMtnt#RhAO8)<ezQSUfCUX>KkZWKE#d^?gzZ8Gc|=e
zR>qAlu;JEb>Q>l)ld=89_4<{2kbfi+l^nq6mz<u3S9j>-ta^9!T6M7}Vdf>Z<hz{h
z!qReUwq<i=#`e-UszA<!x5qt4cHKMD8{IG5)5Kp38ACBTg?2`I&6#8{*i;v!_;LlM
zuJiMFT>`Tu5UuH4Sb%$k>*J1WhmsrYDvJISy0VYUSE$LOm{~WLzfzNlRX5cnNPpF~
z?B7dhZ2=*QwKctr9a1b`6ZI5m0|U4H&2>rHt^0)j{Y8X*k^9sSwuf>u_D*gz!hClP
zX?cb0?9vhO@&Z~}PmV|E$;#_(mB5<Y!L*E%V$PD)&y+C#)njX%j%PDpviPH-02O1D
zE6nEhrQRJvN6Q3G0oC2@q%%2goCd>@!}@@bjPS@WHb;tQ{ENjN*G}R^1x9a;iTOr;
z!;7^)_}O;Rb6*VCVJrUqIm@J-X%H-TOj9K$tY}^f57Qepz=88EK_(TF>BP$P3xVAp
zZOBL5_5|$!uzTjE<Sux7RaJ$OD?zJe-lFDV@f!m*Z%-_0Agfer@*PSIhk!PBgWNVA
za{G)vnk=^wWzuZ7y_GH^?%RfqO`}oWSpGWj%h>q5d-`CQL5X~L_XkZ}c9~>ziSRV_
zBd08*hb!8P!R6wBl&QdzE80VIWOmi#NI!@bY|Zz;!s;M_DDW{%dX{4b^50dnuf|?u
z8UA`1ycd#go@QHHUcG~&DMC!8?%IqwbXFuc4SNyOsNV|m{#giwnAtM!*n#SjR~Qti
ze>4EmxYd0_%qafC^*DbUOHKBL4>0Z7<AYi-KJLzs;!&;edQD?4vS(@~;wWV)HaTpF
ze9{XWKNf7KeX;*UZ!fpuO|%1D<(*G&vD(k^O!Y7kZOQT>zM;LgqqFpwdi&dLgnIDO
z<m(VUC%erjB_fJQd-4ULOCrSC?lGsGkPwe?{}kjf-m79<*%v(kQvq7yzBW7`h~wzL
zqUbX>v`R*jH-B|V@H4jh<)U4GwBI|Y$cI(&X;q5PK5qXjRIM#zYL3PgrTyo>21sWJ
zR{Y~BAUEBL(ABGRx?Y|4=hMSOCMSYzhFzzvkJ&nPy}V(%tq-_6wC&DT&IZ%MFW9GT
zk~5`}FA_P#FEUTC^Eb`yy#wi*9*fKIh3><;L~~@54U&fS%@Q)W%jhSp&sV4+c?#M8
z%`M%3rQ|R~u={SlTBFM&5R{Q0+B;z%xBlcC1M}IHheR(%Q1LjG>J2rA;<`1Bfrt5$
zNQC!><mTIjR1f4g#<yF%g8Uht*GJ@Rck--kz2m+_`Kt4Hq}Jks6{+A$BDbQ`cPFwG
zYnA1^RTNV0<5)LV2u=1nU5~nN`B10o_euhC4z*Zk$M<iq`?yZIzKVa!Hp;C-@sxj!
z*2iEMb)(q8Gc)2MO+o)-HJO5UX>H&BTWwfXieWCHO<0x6F*MQCP-r00>%M8gq@3HY
zeCU@(q+{CWQ`-G+KI8%UaQJ=qkQ(^=_6$2wQfv^~Q^4Qn0aU-EeUM&7pW<-wbMQCj
z2kTDc+vdlKpquQ*%@uTO5#H{vSGf1pPqlz3q1C({;J(kk&q3z_pTSngh0W@EgRVs!
z+v2z@B@d25NAd~0%d6M6kAXLqw}oe87eUvGf9DuuQ|rp_OMbp(ecQ6tkIS~dfB$-5
ze7zI>tj9zhqYoGle&qZ6=_ED$BZpWWu3xPbRvo4Q_xm})QT~I$8cmX5ph8xFG@@8g
znO#YJ0U9Y*ol>4YYc9po*izC(XD&&jsWi5-#=WTwjNnw<qGK-A-n9(s&wVjg_o-NR
zWxxMDRW1m;DRKzTfYL9PY?K&N^dcNF(8w4w{2%1WqO?{qR5>tm<RL#yS+%<Z1n^dd
z%-tK~jTIcTle3S|Ex1P^W?T|wb$3}cX?AhB+GgbKU83JM_Cb)$IN&lRG1(}>&G1bG
z8%qKyrjqt;4+8Ox60;MM<t!>LlEs{r%*U0F>Q&4g(>f1WQ@vE^Z=B^2nbXyiCefFY
zX7ypxI~v5(CSbK`CQW%f4$u|JHEzwitCPEgQJ3m<8_ZIxDdHKbP3RzOEui>TM+l=9
z(nJxRGD1=&v45BckxyvhNVOD8Cy~E{L)0&Rufjan7zT`_vtR%KrAG7ba-_@Va+ICP
zqiJqmp-soeMz^e!sgs$Rm6_?6nVDH39upCPf+8kLPD)BHN=iC{g2HM%6agI>fr83n
z8iO3!Ni^z1j9mT^061Yg;fBdwNJE$7m5U-52!=7_)B@N;ocC*7<9cHGrpCyalUF5c
zCbcF7lqIN$(`2FhOo&Da9iBzrN1jH0jYEKg_KRg8lMTUX2iuCoksK!oRa~oyR0?ur
zc!La!S}DP(LAZ(d%^}~TdxOQ6i2QZ_H)b`(F2ClyCS(<0q?Q5R0)QJ(FhV*lcNO*a
z>0t0<_<%<YO5WLY<l+ZxhfD25!mB}Yh-Dg$U7O*59b<t;BNKN;xDEMJvZ!X@jJ;>#
zNDU-tkq5exfsK(J_+bi0=Tb~JSTLB|nG=}m6{#r6rJ%S4%9x@xhOHi5Twd_8!!SrP
z6^5L!x$>8hOC^xaf;ju~SYTsI#$=9W3zP^w2<ipk4j2C6uLkLiL@+U#EuUSO?Gfyp
zDmmtL8Jwk=U7YosCFACw33b`maO7)~CgfC1<NA#?5xs}w2>=e}?GoM5ywtqNzVyH2
zfu{A5X~CR_c4CEMWnq0{9S`XirnJfN)AgZC!)`!8L9n34l2cA6xl7-q7pK3bPcqew
zD>X!P3CDF6_BeO(mL8}aqyP>>Z<VB}(J?^7Y{ELjsKS(KV`PeGE8NfU-9m5~L^1zF
ztwq~P$V&W{*p6b1^8c65&L1n5JP%mow~KaH@S&!}go|PrJ|EuQ<IWLV{wH?7Vl!iN
zcmA*p1DP{Fkzr=bV!uVa)zqWy@1iHl@Fx{Ny->!iB<^p{UpC5!NE6shmnr5u2b>-`
zog;=fjmI(LdbMh^B@d0RY+XiLu9Uje0J>hI!BzT}m`lUA6)XoM2Q!EC?cCm??&My+
z>Oi#!T6T<R3_Ofh3@5~Fu_H%rnYf(B5g4a9ZXuplU2v>XGbajNHjl*4na%xN$Jurk
z!X&BP3QzqlNMH-9_Dijb9*+Ue(clY*FUIyup4TUBK>AlmfM!??s+xACpr)NBf#z`4
z;$q9$%h}@D-BbNj($nYDF0OQJHKp!k-G@r2o{z4N9v)UcxSB|_Fr^l9Hmp5FB(E(&
zdz>bhCRTMc+>K~ly&g1l$UvjFl0em#7O*ZpqMS6Dn%dj8k9~o?94`$Q5w~_2n?;cx
zYkdgBN&LMNWnS9O)@FgT1J5tkCpJCS2}WTcZH&St$!^>ZtvYgXVT*meX{l)&-Nu~#
z_iwT38%%dum+1CG*+~)EVhYA#SqPPMcH&syzyRa`M?dJ+fuFyB9_!QzstBD4y$b3#
z7p$w<eYW)GdFJ)_c35FG1zZNYo$%ok&=L3$XtE~6kgHB=m@elG&0Nh0ANwCmScobl
zM<3|$5E~@C9gw_K@K*-XjluR5svD`$Q*&ZaMTMf4m{PB&avr9!R>!!JbWrQv)-Kjg
z*9LUdcJ6I9*~_=~zG)&=NNqJPG_ImuMOG%z{;Je2QJJm)tyEp~W2(_yJ->Q*@CU|L
z{5eBymK6Ao_3M=GRP~ABf#6ZkaxTj}R?(%n9awh1d=K`~^EvqW`!f`?<j^Kor_TLH
zXNADu&xH|JM*;<@<w#fB&1}YezIlPO@mKqgLcx5YIfD}*)DBn*$X1|jAAdckqg5d8
z3S>#Ytrpum^E-|yVkJ~}NL7eWNMe5|iU~PRrew{$**%`vh^&ci1UESWD*{jW<1c$T
zWVD$fwsCpM<jMO*bXGsE58M%iyYSe-=)v&8C-W|CQ}-;7DP<R9a1K!d<+wH}cPT~&
zXt;IbCR;5=$@wCVqPY3Xc|HK7QXsA{1b3jiAyP!#Vcd_A3llsOO_P8cX1e}~rl@L}
zs^JbBKkVaJyb;Zj-uO{l@SlzHX1|*+R}0aKBhO$Chu()ihfar>_=dO4TUVDaj~`FS
ztyQ?KayByGS-gOKd_!^eP&+Fw8ww<)+^UroVHF`2K2&3oPPf_S$5D5#xa`xJoN{%F
zMDhlm!i?Fn7Y&`9y2rJhIM?vzzc@{@VP<sOQS3{9o@jJ#+UNOf`<?n<&+%^(Mq4c1
z4evjJ3kV4y>4&DD=-vM6D%*<Sx?x&O_Hy{&T4<;nPFHR5kz3Fwk1ZshMsa*<G$JD6
z*--qpM!GyA>Q|_=bVLL-X;vW#DG4cPQgCvNX{d~;wYhl@k#<sMYnQ2iY}Xbk-DJ6q
zIkXFB?+5qsRHl*nj(@<{EHI!G4Ep$@duh$4?)z|Fu&4fmx>hxbQsJ=JS^Ys7L8T&a
zGVqfYn|nZgk*Liz0G68%B$e&HU&yPC4<tM>mrPIZy4NeUF$(Dg-cBk&vkNi1&v}lc
z!PMRq<|Hq*@QZ;7$mJ)qF@NqhDk4;mt!}xOrH07S2hnxEicyc-MqueZuxo^LFMu1O
zW{Gbdc<6}dI!O@w8HNh{?u-S%T(840RN}y(KCJvpCYS)2^-#)G*h*F%U<i4$$Z$nR
zg_=F@%w)}WU>^#Z8_1@>1_y=@2j8G82}uVE35m6sESSviYB<lwRluexh8I_4RmjY}
z9@INyS<zR`RZG?6Ol_Up?wK$Hu7u#d_PaX5O``ydx|;B$l|Yjx<LwRXLWk9j)M#Qn
z;*LQjiFE{o;K@qXG)_5T01K!cd<TrV`GFY#81o)O32vg~2cI6NaFB1QBqqpXXpQ&=
zBT+=mjFHoaRuQa|^DHbMX^s6E-Z|1gNaqNB4RclthmFH+904}MPOWr2kLSq?52F9U
zPI%S5=)1#&CByk~`#wPT)RG79ZDhA~)XW6FNW|!MdZ+BPZ<B84YqQlN2Gu$Z7N-8B
z8}R(cX6<58N>(#I5)pSHV5g0zJ@GZ5W3zkxVc3GLl)sh4=&0e(pycdUg-Vr-4zOyp
z0;vjLU9p^BUB9v*gZ2uy&L0fAW|4QN1eXN>wi8DJcK>9Hf`8&5FjgluQ1pw##sjF|
z30Z26Q)V7CwNih*vU1IA@9s6C6Jm6UD+c>Eeg9%Dwg1xr=c$v6>PCb^HhrlWI=S%)
zxZY;;^@)%un=Uz{*P#}s$$&D59%pn`gj<MPrDE0X+X2dZ=ZVb@F0DI<!5c`d2(`1d
z>B3JWy$d3ePd6K=5+2}e_HfMO@Kx*Mbd!=fZ_x9(mbwe44-fI_ksN@bLP`co8UFU5
zk4MmolI<!X^-D3Avu(N7tntC^JnhQ|c(L(pc<Iu)Kgnpo_`Ah?ON*i7UT=UY+g(MY
zxNJEBMb)>6l{F@_*0A6>FyF!)USEY34-KJ?=7<lY&{wKG&Jll~^S#v=9R`l0zmoB}
zA7pNLm19g5EJshE;Hrq|N<YTGv*pnay&cXwMA?uXzurcf&>2%ua_0B-ehrrrdHEnv
zFlh9_Xq~m3@8yW-<FTw6s*q{x14M$-9yB0uD<A7wajQ^ofa`>{65|X<cfS7Dq^g06
z+whlCCyy(np#7GJHC9eES9<tN0MGeIEDzTl4MRo~y4;@XagjrC5K{*VDCCr445ZPC
zCp%%0xLcyf-e@`RJ_~;nqdudj%?^quRf<6btgs}>doi>fh)r}KZWD%3-uxLDt6Fm?
z=9VX0P)u;li4Hmv?aj994sA9V8BGpoGa>H^lr96cVf|mfx(<n>8?}pPNvGD%-<-UC
zNpiz4tlsB=h;0KgG6lkRmI?!JytNO4d2%mn2*7K9t*BGbYpgtgrp3u3k}4tPXQV%v
zeSjdKw*00)sL=wc%jLC+Zv_@9q{cFq8LDO#26P7+Ij{#=P|P6VW$#ZP^zfZk#QnW9
zND$zcM*hAbM`rRD_@)39=$H!&Ee0_3_;JV)BKRhdr7UE2>=PY_W6pa>IIk(C8!sx}
z(*36IWdZ_(qd?2k9^?K;fXB?49+bXYrNon9TjZRtKvA^fx!)B^ZfCm8AvE))0H7#l
zqRJp3FMlf#7Lh*|ECS*!RGN_}%o|bgn(g&+gyr6+H@Hk_OUe1Rdzy{XWw1ZCL9O@@
z*z?9?{b@UUoysRj?O#&JmpO<vNDzHY?7O!sv5cLZ%II<-1=)|qOC(9VG6=9wDRd$B
z#PQhXhTM=481MYV?xB3QcM!E>i(uZf7q5&P?Nv4kEI58U(cj+d!CH3U@p6_OdKB?O
z+m+v4nd@I4U;>l6Z+vUs{5+)~vS0INbS%f*S_^UogDfho!kW($*AL%uy9_;cD!Jv&
z=Mt>x2<}m%1vfI3>iFSGJOC4?<3V<u8c=Yw&a*EWwDC08U*0z$n`gh8g6b1l<@vHi
z$Mdkd`oP0BGaX(rpePWFKx8*wN1&XW91!^zf3{E0>^znQpf9{0&PR<-7&<<HW-X?3
zhrsa(chpl+6RX?!8aNS>Q|-I`362Y^YB11+2rh#G5i=&Y+(T-_zp)FXM)^>0cbcc`
z4Nfs|!LsvJ<$j%<$?3hAg4~QJVxs50^8O0RX6kw)(ARw##!8Jwdb;~0JqByLO&$L6
zo0Ecpv*%~(%WW>#l39aGjHXep3k6;NnXLH&w_(DcvC=?*lY#~57=rL!ohD_Bq_&i>
zsTy_xN;bxma{p4i&}M_OAGRv_`vsgR&Bc6jlJuC5$zWPMV*yn2-}b50osY=&pI($M
z8t&)NH+<Du`1wAM+IoGaBywAQrt|?y(-8j}rAi8c$;*6T^h~HpqM}Knzl<wD$ZqyM
zvjOOVsC|T_(ZW=<iQ5uO!zd1-NK$<qxZ4Mq&@hfjC*UpJ;4L^x1RV;`CEy3Q3c7;-
zAnFfb$%(ecGe*t!6#b1z(hS0mOgbR`D$DK=k_7z(@Z$6-?KlY28IbOKnFtC+oo>h;
ziap0?i&;a}q3kb~%3sl%fX;=5hm=jlX6+$Oyo-|M!97&1)E}0cmqEIA<e=kbp%)5n
zII{Z=-m!@eqK?k2J^Ok(T5goXf9y^(Vr6BGMyIH-<Us2E^(|Yf9H}x-@Os%oc8S*%
zIyEH^eS%+TZ;QqSKVFPq-jUMvdIu6K-l_!@Uc_F+b4kwy3S%kPoXlLVL+LWv=RO~`
zeDs4z86$n%yzSC-N}7gEdkY$Q>MRcC)D`s3G73SC_%~hO?Z6bop7%;od`*2~#(3Je
zXDU(w3D}&sgi4oYX!%{}_6b?d<MJW}`HixfZ9%-P|J}}R1?+D6F)9A^hvc|?ht_*F
zoT|qPp)`Ox-6jI<c)du=`In0$h|!lgOE|mk8oU<)5xE8j<btX$sjR0R@d6>hbPt@R
zM>mdo9B>MAEMx&eloATzR+9*2^aKcL%nN9Ry##Qi>g87UXkB$!f*sV9_4B1iLexax
zi(xz~%9|)Ve~$NVNzf3pme?Mnx*|8m8F}#P6mTx2*6?0!JxuO>2s!sqMDO4Ij6yi2
zgU?WmurDk-uLH#;RCNQ8RWSxybl{>mbp(EjSbu1_o}4@1FRz+S&IHdKYW6opLKJ1e
zBPlarKR7b=Ny+Ly9McIw+G9Y@U8=l}FH`n5VGyvj6nc2{4lLpC7hi=qomD%R*=!Ny
z>3J&DzzLdzN|j$DG1sD(&lZ>FiZGN=gZsXd%@r{DOL8szDk+=aK9RjYeqom@8#P?p
z4hXto7`~wI-*-+wX%nwC%BokyD8pD9Fx<u8FzzbiOexr5-BqaA^O73A)4%<9crd?P
zB1%v2dyWe0Um4LS)Dp5BIjeHSwV6XX9JKSUvEi(^>-pgFoWXH=9^}7TeGpYDbpE@i
zTCSh6q_Vxb(bk6E6?65R9*0zY-Fvdk=TkHSr7AeyYK^;(=PaGVO`%Bf*n+}RaSZSM
z@O@NjYS<2`=`Mny>V_rAb)3+L7hcSg3;(X~)DI8Utd#z^RHGQ)qC^fgIq{>;`}l$8
z@N3jmKRY@kSRnd7a9kNpcFTSpn0en?T=Rw6bOUnmi6q`L1EAR^CXJ3TNC^E&;75~t
z!J|W||M3k;Dyw2cKTPf->?1a4mF#VE^gjdjpqcjdgs%toHcHO87espnFd__M*;gJj
zM*<?;;zQE}4#?u(gmvi%`Vx=>D_dOjG3tzUcCehEZQn&TWEuBay7X}it6a&-aw?gQ
zSm-scVI<qLfXk^92|XWrf8;>wQQ<-K7^}I%;|dN>R@yPJc9MH_mxT-<xv$}OCUQsd
z$UabCi{6P&c1mQ4Z;1~3UR1s^He2<#!avK)0W)v<J<Y#wIPw%u4=E><VzIQ@i0xE7
z4?~2$-T^iA;evDszfvtr)~?q(fghA=#<(LXWpyhQ?6gK6nh4|?=&slIxASAcaXX-(
z<tl+e;iR;<WyNh-NXCj^-eIJG4(1f_FQk^D%A&sZ9j^~J6e>@>_$<_3tyHFhB!vVt
z{^b776U#-t7}=Q6-(3bTv=e-rG*n6)nUN&KFoZp4>@s2=lLbc~y@K>oCVfd9zi#0(
z$gvlG`Gi-V<pm!Oh<(V?xhvt_2ogoy;E7ihGeZb$cvU`TP7QxL4w&zomJmdI-o=7^
zG>lE@fLlN89zd5sWcVIDrSDB441a7Rf8vEWjh8fNEj_*H9@q$L?K0ZfYuHSP3yv^v
z4bLhw;5oQrox+m}xk6kg6|}d#&#ZHJxMOj3I%qcc%p77jZg+bC^=|!;))Ypr2V6GM
z9_R<o`X;Mzv%!v^V7HG=;&KLwC_9q{mwyaDf_sgj%-xAV;Tt~{2)<YfOiW1$REZFd
z(E(P~sOqdDN5I7Qfy#UQqZ!$ANv#B{=kI^^EpcQ7ApOP26aroA_potc_CsksvHXcQ
z0{sWzZ#H?~?WtfS<xuO`@?q8Rv9jox4UYQd_611bj#xX8Uxz1I(QMwe?&eAj1rC|9
z{;AAL;5Zl|G*Gm&=WPGTV-i_6DwmWxWq>uELwhc;+7WxW8CA|iRM$ReUoxkU&U6TI
z%!qUG7o+Z54~|2LX$hfpf;Z0HMUobC+u?=<`#js%mUqgWQV2s?N#mBg#Fmt}N(9l9
za)a~*3_w>t4WantbL9Fgc;`(WMlv88T<Mr8qj=@TLo}b!QwnW%B`GY~s#+BAxgv=h
z`$O+#)mJ>QCyxY)KM{{&%qv|ZA{kePyLi-EA1m}u%JzzGhin-zR6R&opvECRz7hYq
zJ(92OU#T*FCEf9O`^oG$?`q`vS1v)w;zr)}2y^lD8o!EphVdS~loP-i0{QHTJkCdG
zh%lOYX3W?PGR0u)3Bj?zP(x6?{I)nV5~4{-cy}i>!g2WZcxLBnGVj{p8D7T3db9yb
z!i6;}a&As(gbzcQiydE9|3Pak43GVfkz9^&$S2NfVOx}|<z1uU@-6V9$ASE+qEk8X
zwq^T+<7be;M^)@kXH_c2_?Ns1t{$&l9yCAZi!aT4k}kV&K~qEn_)e-n5P_W`X}BhG
zaQL97YP8L13WIOABXDt=9!z`(t>$X`HsfV*q<w*PwE)e1s%!btWUqD$qr5AyCxb{t
zB!eza<MWcH-+fpgJ%0P2Tr9E74+SEAkm~}pbBh1ql)3>X#YL~>WrG0jyRh{B<dfru
z$@Ezq$!E8#nx*b8^5!|MK?$tVWnIS^H)D2@`3#FJ770j|EaX#8VRn*H5WahnfUm}#
z-L{CnJG$09_QbPWV#&&VI2gK3j<`=bEoL8O$7iTvFQZ7q9j?|uMdJ+QP^&bwKhJco
zS2i7A*Z!B|2iQD`%!G1z9!C5SE(-o-byIqtV+M6Z{%JPejuLI)eg*<fhAqgF&lZoM
zF;cw#9+f)rhj2IA4KA^?$N2ik-n~19z5V;6PmAY^uaW*CRPy&vzaz}CedZgBnnBh%
z3+^|m7`BFEDCtOAheT#NY;Ny=ay__kAUtsY`pSN?X;)DnjDB`RuyaMN-DN#1UJe1y
zz{dA0mWQty(nDG#a3yO$6YNZz!aX>6p2HNO?23Ms^2B&eOa=>X3QS8Mif~3Ad*)v`
z?L8{^DJ*RD$8F3kb51S4ojr(E>A%|bub}Cw4od72<7;*aS#|XGDv9{(FpFW2a$Law
z;3Z0=92|HWy3_uNgw#M-_GxD$(DrN7lV`{eIb_t4IG5BGgF88%YqM(s97=nK8uJg=
zoSZA#z@~v8grvV;h|NGRv=~<)5%(wWYTJX)R@?B*6KcwA-N1gNvy_A`QmCp##?+w#
z!O6n}($Dn8*VBj3=EjJPAI|^>1OE5vbuGpny|*gf7JmBq3>UP}4ITWTB?<ER?gR%$
zA-#)nsO>egnA3FqM==Xjp=&%}{-BB3BfmK?xCoK$Ey$jX)o@CQ!1svZ*162zF)fZr
zxkY&Wpbn}OdSFs#ya+@DDxBhVM3y^Q9T=Sqk(fu02z65Ab8ayGj}F}a%jvmK4S5z3
zI!z*}GEtBlS~t@v<3)-ROgcne<a|Cqt}T3$XX9WAJ~cdTb*__Q@zTK!Jo0#G{9Ltc
zPa-V3^BxgpGCwDNCy$mvDP;4Rt$^3*gN@c^WOk>|tXDmtuwTbVRNp)n{|te~#j500
zjRi|{eZ({3_IPo|N#s7?>w<O8nOy{!kT{C<Vvv@lxW&{lE8ZXAQXBmE>k(Mp4DRHW
zp~zI+FzxRMEf8Ve{cOR2PqpFuFc;mZUPw5=vCC5a6?8`Isb1uwtcH;RDLT&g>=J#4
zlDFu&Z}DfQu;lC)nW~;J4DEVI^X%1XKGaDARH}I)(Y)UKmRK#^$NgRhE>Q$0OfC?X
za8T4?w-#Rlp?nuwOP*a5VJAM{kC><?1lc#-J;-z+d@BC79qL&Rs+6rg@##nn7K;|C
zP(po2dw?cCq0{~yYo6$upW|6us_+iPp|GJGtlgRf2R8<?I{j(~Hw|vn$ceWLDspxR
z>zrP*zDLAXKy(NN%olVN?A^~aVJ5^e2#&3f#cmXUgstPx%RotanL@(l@hR)HyX#bC
z;ynKd=Od!K@=AE4Xb%te$7go18bNk$7dvD&UvK`z@ODJ567LX9NIuE5E>%s<GAXMm
zhI5}8z7+@FK7%A;Z`{`b8&?Rq+~l8+w=2aq2NsuIl$TZi{VANr^mfy@qw{c=7o9@E
zW#+1^@WI_-h_CxBn2Sn7!_G^$rGb`v{4?SCpwi+NT#AD*8xoJh-UFzh_!tVDhM?;B
zW{8VI>qc9F%@NZXmusPFt*m|a*M!&j)*w}|`EfPDC)tO^wBu_1pwL_ZWGV%YXEu3t
zNV5Q4z04x@>&&a`T7@zZ;ygrXDENtJ*70?>4sN;q-jL7j!;tY{2TVI(A!f^PY+`vn
z_>%egEdT3ywg07+P)Trur%NkQz!idbz&p&z-F<gn?pT~AHSJWK+~9R2F8C3IFS@K$
zXVdg)A`;=n?nZ1;{Nlp<`A1T~1HI4QV*BZD5gHkY*zGzx2EBid#K%HqSDHt!v+`@5
z6@uYo%IcvVY6JymKm07O5OWcSZ10?|VFtjVZ((l9*TfRa#5|CRePlYT!kNCDUmgmd
z1ULKzYV2R4@T}Z^&id9L@_v?F@Au^SZy0Q@K-c9=_4ZsuH{G94p0yD2sVk-Hr02Q|
zJiKb`4(3(aTy1XmK51Hgsp$34J(S3{;+U%5xLJZ?$2ZU_L1u(Xs(ug>0TX#dY2>DZ
zpd!<l@*qeRllUa*g}C4Qv$gxp`h9{eqS4+ZP84iKQ-&<6Y}Clq|6N<VIoPzLIAn*N
zxDprs>*hNCQH%KNT2K#U+l<BfMI}CK--Kz4hM$wwe44JVIueJ+=pgCH1@8-2C}666
zIMX36ThX5{;DCE~JmDg{sOVkloRn1q9SWWxz-8-d`6~xA+1+rmki`Y;RcO;=s41PS
z6dio3#`A*+uB<FvygHfyt>k>Ws?Po_OIuZ@FEl|*&w(iZb21MHho!Cer@?Y=bOln9
z5L9D%+@94;qt5WGn}0H;QK~7bmZ)25BwA(^t5yBB_-Uu-XxUAHe|S&j{QZm}z}4KJ
zUDIOh3CKsto9s7?W`P2b;VQ!kaUn4~c~V@eegSVw(E~|LXDSLg!FKe1-b$hX3QX}u
z_3(Eb2s{^QUcQ>d#*QxO-$yNB3iKYH%rpF1_i3;D1aAu#>3c|COrIxd0ww6+XE;0|
zn>(DJr;wT#+S-0Uq0ac)TkBOM3~2Ms-7a3WXi3eEM;vs;o{0_NISYA?xiWK)V!q~v
zSI-A0y-3YiVY6HK<*~KO%zIh!1cvqCN&AB<;KSRk#L@65un)<ZtGn(_WH+b>=^b>r
z7y0)OFFJjWRH5KzZiE|AFE*ssEWZ$n0XJ^+;XvzhjGn;3@#b4KrDS;P$NcD%eJ(kn
zzb^W-Pl1{d@i5GAJ)>2@koqXKL9J02TZ1>|=WOB_;vey*MqR~W&dpjR^k<u0+)PKe
z*Y0Qz0d2@m%C%#Y2gdEZzlzI+mOiw4u))b4q>6USl8<I`iUsKzBCQg$?ie(n{J%ug
z;JgvY^96LJcZKgivSIc~%l)!Ih;ER`yOMUQd?icU95kJNT!*?cMFe5`Js@6Km-_Tc
zd+zlS0lV%-gN86Lsx;KAccKjhT~7{Uz@S*oOG~D+Q*#v#)1y50sxmJ)VUV#1`WpK$
z4Rc(nXo(q<@pK|2hC4(P=22C(#;#I#xUp+$OFPG0g-%4`F}e~vc1isDGAu?%!e;cd
z;UCRvIo152qITOdi9KcZ>UU0x+jCgyWK>*e{q@90??MNSNIUT^D!T9%JZae3ovGX}
zg)A`^S}L>olVP{l+I)L|>bBljxTG=WvRE<$p*rZZqx{3d#xqmTu{^`3<l`~~d-3kD
zGA{gtubgepTh!4)h$1YQ=c6MP4r^Bs)Mz+0E8%Ip*=lHE1*+mw7ir~|9aHKa3tF6z
zP9c*%M_bZ<N4K3X3NG^5JRFV^ems0$aisybx^?ZSeX@(j{j&!*UuQd^@WQZ9baFci
zgEy%{$cSOC3VQ76xYTx}p*ziBNIVOf9kog*I&_dV@H<3>oEMPF$1^b3yi}H|XSQsT
zbLSOkb-E1?oJUh+e)%YJBc8~43?K%H2&elV5Utd&!8tG*<kX-ZhJj4NNga&>ujDXC
zyo$@nhw3SkGTuJl+IghnkHqVKVhI2b(-)z(rQ}EBf)s$GX2Zr_PQGA-7944CdD6}#
z5t}Mb7wgujdZ6KaAs?24_`DAY%$Ra3A%)4xP48r#RHr}gm0|193L5Z&GO~W@XY6Pt
zC)D}Y5xz)amyNS4OsfHZrpA~Dvs5V)U}wd6rSF(+Om@#@gEzs==KO{&3RR5w%Me#*
zk30&b`X?Sey1ZuSCd0%SJT-&>WkRz~MHTXqSv3i>KAB=W0{KQh@fFSd1>E}e^5suD
z3<D&CvY7bNVaXjyy7AjMurq>C&#?T1(}n&9u0O%9;HCX>G(o|Nkmk5MA%IGIiGOYK
zGb-T$X@FCeVe!G&EP(4!Vj^*WF^Ko6*7}D3ZsEGVYAm}J>l2rsg4Vh+pu*t9C9%NY
zrvY|L6?hh!h0eR{=^TVD&xkxawT3O==luOWjpJ49cMcQ#_<7Bv^W1J8?>`?uCPX~I
z?KG}$Tg?^weQTHgzcx_2e;m=WEiEP}+J?B%RTXsAkYQ;iJSP$P_1N;bbl7)4G;I~D
z0z<HzC}>5hrnY?>j>HKgY{51!*@UC7_<Kwc^V!Xj)IF-3tt%8Kk}J+qP{h^sc*3e?
znG+7LRYL70bb&!+tOOd<k2zfawAriXBYMe_ZB+KyeCF@a%R}@I0l?AmBbPf{dxQa>
z4X`^SgG0X-);I~CPJ@L|80G%2A=<w?B3cH{c5GEZ&r<eSs%<KS+uzsErNdbfcf82I
zry+pHWfc^XR4&WHs4Zb7(atlKDTRe86vM+IocvE4zrOdoe)Du5`L%f6S$GkfJg#Xs
zE^+$V3zBpfEwb-*FU3Bq1^RlNh?dUn1=!$**>X!hgCAkzg{~t=wV7?~VR~GL=?vDQ
zjqgUZy24Uz<xmtdd*_|W0#ym*RF*BEo4guvLu(CS?|oRw$)?7AR#9hZ)<1c>=8h<;
zcK_AxH>6GBbaege3cj9U3IOL#)$Vo-yU0zx<>Fap;j6@GE$Vwxb7QI0*h)lQ(x=l)
zZt=>yKpcMC;O-JoVw!wnp<+r4PZ%YB8(3_jLpS{u^x{QjtzZv{Sn9-G(?&6;$BEjF
zw6mi1J^dDCITIU6I)uTi3_zzz%4eZ-px|HUusbDu4gDA}p=iS;zbU*YP|pz7>|X*h
z5A~x5*rGXMJ3>_>-1F#lPd9gE|Bcdb5M5+^-yl2X5vAHAf`~tZeFB;zp%&I#`~>b$
zWzh&^Un|<{3nJ<Zj5kw43nr%bQhd3VIfom%9>Yuap5k{7AL`&;C(Dy=J;%5-M{#BT
zlNHBu=oRtA%VlEX+LeC98%!KZZq`Qy?sd)!l(t1v=~KT5>W8Od-toQgdC0Zi2D}$^
zrHA{d>>!?66ZWI;GMyscfTS8}{R-s#3rCcjYyF~+Rm@x6fUGa@M_P=k<V{~bLB(1g
zS2@4!(~L6%O%ICeY<ZZj)an>A!!~&3(LQJ#m9ixOrCLQh2*d@f2?<P*^Y3O{U*IZr
zqaQOlHgvq3CPIP-fdj~|sA(pBa+f_paUm^HP<(OZ6@i8}4$WqeVhZt4;SLFnxgw|~
zE8p@(iq5PQ{JBlRus#IeQiN<SlKb(XAerxX9>RI(PD4SKo5?f8YSKoGoS~1S8rUWZ
ztF>rGY4r<9dfz#N%iwkQMnWqh!@bG<cA2f574W#Zew1hKU}C90^+NBRZe+jcGb|7X
zS09okv^N(iC4r8dh%dePUO*lAJ=2_lK9x1m6R~$Kg?)A>RbP~aCNwCtyI{W0`c9;#
zN8h@Hl!i%4ek~|jlir+tE2=O_%Ev2b`U|O~81@gxw$|B<^u8Hb=O=}yII8t-dky=Y
zekb}sDLVVz@fpwe*u|=6P7~BsTQj*U+O)++`jH{j=i+AN5|;5#Z<#GUYu<$P)t&3V
zY5bM971s%!CA-52DT-363Vk!&tSw_cJuTd`xQO}Vy3chg`Tx@XWZ~AJ!m}EXmS|$#
zOr;lswdvBAg<A4R3iGa{B!ZS|WiZ}jE~3K~vBm%La>TEpXM$SS;)c(AUGJ0v7TH*X
zRVITO2UcRorv`dC10r#7`-5Xkpt)_o=58VsbssA`(7_(Q@G&?k?ER19J?8QP71_E}
zRWv)B?x&j}M%(zfvVa;jYrv%s<IjSxlq>y~*XDu3p<%vwOq^yp_r^}!JuJtvGK@bA
z7n_!Y9eW+BhRxG9c7mAysK#d&dsBIMsu5SZC1yvflq}vTb2?Zq6lkZlBVLx1bCqL4
z1TG7+a`HSXU)B7>E;IdR*;-mcYWY1tbG`DdL;sd4=fYU~<OlLCOlWiX>~uw{dc(ty
zIqP!RC`~<A<c~ax`f{PH^J#B|oGJ^)ft%vV>4zr1gh^1#W5#RTH0mV<0Z-E45FrVN
z<YUz5j_tra<8-~~J|-UAPpJpWZ_3<Ngcgx)+>?H9LQIpSsX8m~<|DWq*=F2T+RYWp
zW2%V{=`@w=`MZxDWnEhdg+BK)6mm$*MdL^ZUHsC6-BuDkNwCA5NOTrlfFH_HoncfS
zaU)KJZy+7oZ?6Z1XzIKxoiUmx9KbVo1XCx6ExKXZx<G_ow-at!AJa=&RLZnCg9~|!
z*+M-($jRC;S*w7u3o-xO!u<<RL;mNMmH+OPe*RMl=M9^!(DUDq?c~?=h^aq#yPya|
z!ze5I8E}aLBeZx0j2cH2Nn2N4y4-?fK|S4GnScg?Mk{;L0B<yJS&ag}Cb8j0_7**8
z-x*X9a$iF*aQ$op#aXR@hF-g8&I?fRS}FN#Y>o;AXAl0xtEg<%Eu398@1*NO2HJj&
zq<prN+W3J-{J0tnuI}79PXdIdpi2IC-H`iiTOZCHvBF@1F$FMdAN@hackxquBOk=!
z77PE+3I*@)|CwAclk2=VZmiJO8{>M49JU`_V_^~FvijugV)mp|v=UIQa0$itwHC-*
zK7ObNV8)&za<CtU*|9l41s(>knw?rNsNM6u=6xO5S&wahE;28G^PYx%U3|Qrcl+1}
zRJkjnXsDB^DXnNV)Hun`dppva-5sS#a6Mgqysj^AKWk8IIo%f@$K}<ij{cL;MdW64
z<Q`k7Mn}T4O{>j4<Z~EU@JQL^wTn;p(b)6A-X5K6K9eK0&h%U3jG}_Cb1q4hCTz%x
zTnh18i;LyymkCA2lFH`fvV&K+kbT?7Qu#7R3X>l%RiaS|EQ}=aB1O7nnFhs7rlEBY
zptzb&E!7SUF@@z%<(C3|@G#z7o~~@o_xV{(Xk;dwPU($6N`5{!4x8qyV7!#U@|RW~
z|IeDFEjrUcA=XD8nB97y-Ey7Q222R$-<-g$M#-xdAF_T(di<kJvMw|tvZE&2;?y-U
zJj++CDH0yZ0>cItB>sFM=L+MQ>CHmY-KztDo@Kdx+YiV_Emm(nirwAQtsGo$+?Kcc
zoHBS?y4V~T-ly~4H@bwDxAPU~dk-NqSM85NTmt^e7oy#=SAl42(9ld=mX>C!Ls><*
zSRIrSa3sv79Kfo1flR}#ElTZzxm?Leb&Rh}IbCg!9cQs9PYbphZsXik>0J|~H9TfY
zM)y-@yse7=wB*r{Qzql-GKjni3kJ8?3}&Hlhcad+JIz+Ms5w+rF?-fDl6F@LHI}$e
zq_V1xZ|;|IRCZDX*{EXVO^~xKle~M%02sqyl1mS0_xzJM)4%f%##t65q<2g|`2jG3
zjZiYoo1kpyyrx;(?4exJsr<EzDLwn(p37YwCNkrW^ZA7_4*_y)Ij7~j<b)Zo=63$_
z@hM;YO-`=pv#oo<+*Mms`cBTDu?dK9wl*tBtZefEYkz*PbdPbvcjF}jD|^6vUAxCA
z7p*qXJ@2JPz&3E}O=tJu9m^2}6`){`O3G&GJn^@V7-LrXUp#$ro9W|;LPuL|q<nu_
zfHMW9L{S)WU>oC&OY=WXq~5{K>2?0O7ErGAMqfmb#;q-*HV?s$>Z|(R>gJMEzdU0z
zv68di*r3Om+)ZSPgY_I~%iD>{Li59#%}wNj2WOGX)8TiTlb(sKA&HcgKa9-ztI46o
z6CdWZM><{Bml1!PlgR6!t=5*wfttxoX~>RCvp;#1_d1ba8}?#!vB(A^1_FP?{STl2
z%w;|*t8G(`Ul5sSuU^XM)N@Mc--QJeP5dyrkIKv3&+K(~cL0{Wy*Qg~F*GyC-z0Me
zRpEie=c(7<?EZ>Ap3Y*Kq#&fxYa;6R<;|w@O<`3sL3D{XIyjJAioc%+pdtHS@XxWM
zCt_~i+ghlr4%`;=Tb>Rcm(%aCiuQJA9+y@QR&3&cW4y^OmVprTEz5Y4b1SVdXnlDt
zOS&2+I)xWCoeH`h3dcCb`xkY5)l^H;TQCXRR~?<8mzivyyU`a6U6G@!9c<Va)qKpX
z0Kye5kHX9ZtJvjgM*h)TU7DF%wWvQh1l5#X-Z>j`M>5U1PqapDOOmg2sq1Iru`+_u
zuGv_T8|-K+=6|amA5OKW&5dz4Q%lk$!;I(sxTP}VOy1|;>0bjpJft1(xK<<eF-MW7
zrwZH)V{+EM5?4W7-yRhXOq`WFD4WG;X9yEAn#^P;jUl}E{5TfTI{sMl9p-qE(W7)t
z*Cu?hvB_e1j1)SAoIy_2iMoQl`KKCJdp|dy^Y`%v?>PLcAl+<?S9K$K;q+8&y!*YP
zV4nUMxdWI-X6;MLY)lF>F>ig^hec%m0CW=<kiS?*=<i6}a!%hPz9Ifr%%n|HNz4Wj
zw8awoA%NV(=|_6|Ar!j|!Fly=lus?7-2O1>e@G!j^%gAtE*$<L#EcFnS(p-nM)35$
z^7uT5?Uo{Q_cu>NyigifF`>K=JaH~X7FAz8@n#l01{PbGy#OA4%BrLuz4G#&oObLL
z!c)8yn1J6AR7{!#$`%`P2#(wN!!4wgvO6%f8c^v=0Z(YdG-y9k&vQ-{X-Lj=W$Gs&
z!9$aW5VeAQO@K1=TQPXv{0~W?oH&wvKM7?(>47S`R3mV}$C1?mj>k%dcTb+e1vcHb
zj)?Evxgu_#?i@?^O#;kSZaTuD&ONaF2%Bt;fkgO3#2LAwR?NmQlYT9Zf@}a47coI|
zB{5rE8D-*J+PpAQL0khNl_DxGDg`k`>v@j6cL{LcrdRlbQ*m<DeT;IRo@nq;*5M;H
z!HGa&-&GlOrHJgFZB<gpEijbyqZ}XKnDGzqfbCuh`ZnY4fu+Fyf;)~vAwgo1o&D4&
zTq#=6Tf3NAGH}Y?E#ihacG{$wAJVlr3(E1I)1~G;GHI3KLbK*l(IMtZ^ru}RTBL??
z8r7Y?TSBwUHm|nkd;vkGYz_|$0SSC(qZB5Kx3$DB%5#`yq+23vf;0<!6f60Ke-QAL
zS>(RpB_YULM^+H>#%m$cT<M6|<~5r~EP{NaAuIYK&TYmwYxaUh^A0t+(T(t+BXICb
z;}{DW&}uhGjH>jq&%>Yz6Y6U?gY+Cg4JRCGxfx{5BGh5A?yJj5scbNPTMXPB{A!(=
zK14FcV1-df#{L=2f@1zxiz`ORV%ldj{5eviIdYET>P+v@u4sKIqhIdb4E`Z{oq$Wx
zIRvVw+A*ZJbor-ucWcy>NCqTx!db^bT;xfIX9~7funG&kcq7I!Re!#|TA4b3I;?@j
z<n;<pW~SMmC)Q26{({t^d_f2*2D%WE56a>lasCXADLs?s*>6+Q6>sbSr^1fX=BJ#M
zH?0N@a<U@*c&s--@^9)i0Du{-ELfN2Snu(qEhC%zkD8U2y|<SszlPCwk9t&cYdIkx
zWoF-FQK`T`H_0{HH8&gbR@Rrl&-_DgzPjNkpgt-WT;WBZX{<AYeAVzLog3olP#4|+
zK6P8;_6$$rcV``|XELN+mh)OWB#(z$oduU0{Zd6DpUmB+-ikr@H0`HaYI+EU0%4WL
zC8c^mgyK~7EmEC~6imJLz4{ZmzVJ^TCsTny)_4P2NWdq9BvC=Xw?O2-)h?HF${NzD
zJWunI-@^Cdq=R{V!0k2Y%=od*3;!eq5moqpJ9EanCP!r|l>uqkbUOuf7Mh?sRP9jI
zp-u&^bCXO)fghVeR2v*#w})lD(2E)<wY2DzOJZjeti>g_V9$X7j~%;q#w3@Aao*?f
z9b4M?1=mnm|2<hGv~w&ioPGHF#B=Q8g#F^=Ia6ATxvq@3erw{5VH*c!u-&WCh*2PC
zwiN9(!0ixs&eAb+Q8@q4co*9P1MkGsAUSK61XVJsymR2~zRJ5ViT_!yHa%NKwj(f4
zVo~qDRLQ*1g{`ALOEFGmk&44VMBP7lt)tLd=krJWC&FveF$Xm~xs01F%)9;2_DM~z
zvZmHFEgS!*ig(&la-ly0iaiGb(N<34X|(Yv{E#-tBLz#_4}?JJ*Dh$T!~<q$^HPUM
zeeZ-|hFJOCE6RSm4iDslSyC*^S(Rra{|L2q#7fxfbo`1qrW)FUChWHp5l(q`uSe})
z2{$GZf@zK1Oyj}ika1$fi4<Fe*37@A+D)W*#=U77!i-2>NUoz2{CIlPvehNu4^y;|
zrUrw|h!dNInpF*SDvz%ma%|rbx3YAZaRfs;CRmnD5%Lm@pNpYw;^uV5lJ_b~Fw`~X
zlddKj(OHr;>_^%o&W9IsCu|%~VOOm$VF&1{<Cf7TWb{)E$SrqJCZH(|#b=CO#*K_W
z;1$m?*txqBr>ZQ~H1aANReltCA^sbdbN!o;iozht5i%(OWmtL=zM#CroY24(Cn80x
zPIEeZPdL7AL6NaDZ)IzLly_o=$xob>7O62N>6~)9vRp6!H(d?fP@Nm5kw=ScU^<jV
zXL~1fN09ihZybT1BocH6JB{HE^n$G!>7I{t+4=qi?UJ<0-`M%!_yJI^dMPTZSqJ=B
z4;Th!AEu%KOxHtIGo6Huaqtd>?jd#7@vFWT3`JNj0hFDyEgsVbuft?^S|2nO!wytw
z7Rk&rI{Fqu8&Y4s#Qgw6JIQtX4UQ$Z20rX<Mc{M=<Io;GHN2e(AbYoJz|IB)T1#)D
z1;eD@_%KrMoxK?$MZe2Ec-o6?l#qALo;T>niwqEOq+%;6*d1_TSDT|9%{HFg-#?*~
z)eL}6eoV8^A?(T?6*ldls;SvI8n(JA>#mN&3p-pF_EVPFt#MD^!hjbuNKy<(<@e-4
z{MEM*d*y*bX`?%*W=UP@m=CPJ5v9kDs!fhxEGM+Bt5icASIblP{yzXiK)k;MwQ^|z
zEpXAYTN8(5VIB35<+q$IMJ>eR4%DUmWaGck&7aX1K=-5M7)(14o&dvu<LRZ%kNH*T
zqrd(}A4+kX6?orjvEHYnS>R@2o75zArWUDK_ZZ-<tQ`Urkv|T*j2}5&-L>?PU&g4$
z*9v|N3AtB%3*7|1w+ermJo%@BE5H@?&OQ4Sta3crk|{<6f$zG!afv$AHmzE<iBJIe
zZCbHn6Z-wr?qKDco3g8oNQw}W>Y`TBv(1G`TYS1%uT02Iz)B13X{|h=l*4JQOrfYt
z?4#$On|S`wDp01vLU12GsA>lqW}1o%vBOWe5|JR67-I0^b4#kB_!QIpOu)*pQdfft
zC~GMSRSZ`2U_VwWQ-x9!urQm!`Z$E0WW;WHh{D7gNN#PbAd%tCD<PpLu$pOJ4oNx}
zO*AHYcrTML)jFBMU6Qz#cM^J7qqKBnK5n!_>l%7?c9&-A3@peErZPY6RWpi92PgW)
zoaT_#nWQ{2MAl#ISDWM_@Pf|kP{m0p5vxOWCn2jZMS00z&Od3V$`rhgfeC51(iG#k
z=CqC!oRVllY^&jlh+C_~FaZRovYcf!!5&3wWx7den@`%K3L~6}HpE67W&x8~!n*~P
zz*P~beIg;MZIep$OrzB%V%Q)T?O}oUDydI8lJ&4q<n9izMIqM6qT$TQ!Jf~u?A`96
zC8#ydCW)k`7H|CIF)lm#Ig-VjKhq&S+B%iw#ultr8MYxwt+K#2@R42{wt0#RuPHPE
z=s6|n>;${41-hnJxR{{pG<~#P4l~<xjK8GYmv;CIJY?oPTG>JVlCFMyI>2D2y~p@V
z36ov^$`FCN**<^CGAOrw1_P=wn7<_dl-5PDe6-vODMuSRxodFNtj_&bEbcdZJo4qI
zsx>Vx6AM<w1}^WkV0#3uQVUeQu&gJw<NA@9Ff%T&EFwpma*XLZYN-3OYo|MTXCwu_
zI49OrQLcAqKd12<Q|1m`?wI`gT;F{!DXYsl<vNpfrFC=Zx^^|$NMeN;*h#dI>S!Td
z**r4U(}lFo-Cb&z_Yl3^(oLfosMN0SL?BYsykmBIbGB2P%6F!EzO-TujWwW-TZxc&
z?;7siV+0h8DMB*bl>+%PMcut#Qr$gT1xx`8{&*<r?|uBS2n}#TkEAMmixNPf2+h@m
z1sEPFCy@Al2ieMVwwleDVAUD?gED!~!YlmvimhBn7HeuZ#$9bK&h(lb<354KxW7bW
z+`%66a8s*EROW%SA|0rW=E<xhU=8a`Y?hr=;)pG<$LH<hMQtumi)c$_yl<7T0<JQL
zSPHDPh>2xuHIDgsm6nXv+|UQil%am>6UrP>*@S9(*Lt2PTaSD;k20c?vldwVEeWJt
zjSVIeND0glF$<owh+P(QM#Y*L-pWj{0i+gBKaIOJ&V!%O(iufXH6#J-xmT5|Ar{S8
zFl+7X*=rvUv%W53Y_phy=503v%`AX=Nnuekp4*SkTW*f1m1Hx^_-{_mza=(UPuT+N
zY>m@kLJ={+3M&dHD-o2fJ{yU{oE#4>&B2;zS&cM@dBui8@NLoV!dUa5@Iq$Y;ZL*`
zJZQWJtu?M?J{Q;~6k~wOgbyFK``Zs@z3@!H=;18(;s!QX2|j9rG5b?gKji5}<HcBu
z^yuy&@ZXBGV(OPk<rA=e?A0dPRe_1=W`iF?Cy5GI;G-wuwWCadtxAO8G5LZ6vbyF}
z*K||dCn^e@^R6tq0Dfep(7EYu(X8A#8&~5k`}D1h%Xy0zmGx2q(sib)q@pgE&nl5h
zl}SxG&jj9X%5jNvR|jC0IA#x$^ej9$1FGZ<ZkqMXc?x;#tYfk4h2mK$MO?7Zb=Z^T
zuN_fD+=~Dy0@XPNi<tl2+h&auDJL5oRb*X}0w)1-dbYLdCT2z5+1Va%<|mdbyL0ix
zPyCUxzPscrma4o%R$rCcHq!fYl~rgYp~4BKQmayWfui852?lx>=FujmON0f+ND*CQ
zmGCHU$1GveTNzJ=4Lp*<TEA7mr#B_3S%uNUPb|Vl3gl^+7KEZD$hbyH6>82QV3nn@
zaU^<~jUxhuv8@vI&X48tw4>xu)1e|j%gJ62$K#}{kC9z~Ws<3e=>S^7)C~l}pMJW$
zKk%hNUSUab8IX%ET=Mq$zCH7U)}=HWYnZ>_fLL?B2bJ7rHRqN0-?nPc-@U8<^;aA9
z=lNqxGc%W>GRqqDtqo`uJf%KEPhAhH_bHsiiZbg&2{#a(>wLw?0gnCq#IF0~lM99s
z13UHhdf{b0BDK(%Q**9sZl`rA+&Z!D`pYc7BbC;$=tJU))NO>RqK}4$Gq#{4e<7)W
z!kw8mheXoRi`;}W=pTnoy)@7<W2nhvNUNZ9diuq&vas#?DVX-L+)tw%{fbiHI|{!c
zsfQxx9K_HX{NC2!Bf)8XRUYmTi2g%jf!~}lPY$<ByeI%);z;}-VbNEj|BQb+ahwE6
zOb^iaN<Hy2)}KK4!RO48_&=t)TAl7=$s|`3qep<7j-v$?Y!ugu3Ey*~c6O4z`dW)-
zuXz%9px@orM=r8eXG_dicL=2O`%n*~_4(UPS|<=)j!5iC-a3WM9sMCg)`9$WS`SRD
z-_u=^ifyZc`4jQB&U$GuCMbnED+R)MmLE@Lrk^T6HQs9G^?#Ek&)=^rs^-G~x2f`N
z(zQF5Fb`#4$sp=uYtO$%%OG{EqYIcu?u|Rr*5&aTdOZkbzlbmKv9({t-GNZN{YvhE
z7>*89bAQEjM=9C&4$7{Q#P=$8ngKJbPt-~6X2i$<Y9ovGJk79Hw=D&aMFjQWj!~9_
z$EGscj>mH{H;u8aj3QaQ0PVpb)@nEFy5KBzTQG~j;95_KJ;z*G^al_ttuVXV++i)Y
zCDf8*!h;JWZO4S`KmSadkO_NO|FO}m`Pk1t!$-997>HZST_=smfDw6_nvmUDdX?5>
zO@TR7cTxc_8af7Z^7HbyoUZ0mNKl4Yw-rB=MV}#g4mwZJoPVnN^p<?a#%Jg=k?<n5
zxs;Y&5WEQ9k?gA!V7hz~>%pD_6R6H$7V#ogem1EFv}SZPjG#7;o6_R(A(j^HD+?t1
zCS}6hxf90CnIqXWaS_CtqV;@HD)>niPoi|ZJuHrakzmhr2}I)i&Pa_(eJ4JiW{j~A
z>bDr3I3F4dW+Oj3$zf&0h;-&qAi3(}r}o6=y=4d(yIaQeye8T>l7-_2ii-_J)M7H(
zFDhV6<70&{#G1KStu>~dsxD?ta)1!X$D&eGJzZ4#&mv_#T_KxKY2C%vmSbz1Of{Vv
zX=(BJ!T1@-@UrebF$ODY!@VSjx3xtK7(@6BRLANi6Ay$j1eS;elYudlk3=`y@JArJ
zskpE-V-E~}8h;M4UM6N5r66@LF@-l#Ibb}DpT8vLwA#?%jg98jnoK+<MM&XsF7eUm
zuxMXfvFe2rDTRfk_+<?}oXz=+z_AOj3$k2a;auVFVsOiHeZRoZ{$#edPy)4kOnVFM
z;}cXG8oR0@@(As#D>=X?5F3Wl3F%N+>-N?Z@^>O*)9@K7MsNU>h5z<3t3~Xss6~v<
ziN%PWyw@g;GAl+5j5km!inO&)P!}4-$26n2<Ks;k4|t!@x}N+ob%jbzI$Y^UWn2$$
zE$9h56~InLCvFakE>$}=|6N+j2e7)GSn}4XNwsABwAPfGMy{Dzy8u^FJ#+Q+*ad=)
zeYo@#0Hq&;vB!_U4@M%VOP`{DqQU6;r<Y2TK;Xx_faZAed+7XUm(k~oN=ugj-Py-r
z1p55(DRecZ^#98`)1-gGAL<Z1j-r={Si)O8K-8nCFB4^yq6Yp~^g~8muiRYc3pG$h
zTNf1~P0BjFYMHH0{7g*=-h+Ki^4|}T(!wqz#8zkh3~=GP68S%~1RVx*N?hK)FP;Ep
zDdGtr%Kt}?)P}$QG93UfZe|C7@T=@IAbOXxSgX{`u;;P3;qt6_3&c8{=B@(SSCC~I
z&r#W}1oj}sPxo4DSwh&oWuKMCC6+L<xGEEj>K<5OfJPfIvbZ@5Fx$Y1$25bJCweA&
z$`+hJ+nn15QjWnX`Vjir_8}<gJJnnq!LO}3W~|l{HA9IL@zL9#+k7oz%`3+U53Qox
zreH>MO4~(ZeLjlWJ_Z52r&NJXt`_G5m?}sjr&JKLVrO@B^yD{xAHdY?ood*=VYee#
z@Z{d)tw(nQs6PB<1^V>L{<r@uH7tF3TM`J5T6M0>7{p!q8g!G_Pf;(A@~z~XNyj`U
z+%c<{w^reSv8>HNYh#GcqO@&wr$K?Dyg{s4X$GPPfUA5DQ?G+0?tN3was0PVNs7O9
zQYrq%4+XJU$YW^TMqulUmDB4tAgNDwbWFd2!3n{9RgcL4K41Z3^qmC!P85FU37R>^
zdC5)dK~%4tsF;4f!NZ0JtBZS0K#06HJUE!mmC))7qEyW^-qGfJ`O4RbPBIi$Q22US
zK<I59ToBwRJ1SZ`aFPY&0e#1suz&;yxQpWOlqD8w)4$9?1Hts?L3kE3?1r0R3JIQY
z7sc}_OH4a5cD~J_=~^D(0@ar+`_yxR=yn0R4ZI5Y)Ewa)49)>j=m2@1MCjw^-Q<WN
z_6mDVOWp%s;=A;H%y;*NUvtm_5S0Tx*f4mUD9{CZ!+JWeGQxFCLT+xt!u)*DWpg6?
zk@@5^uzepv``pvo%*yH=eXZ8?kUJJ^B%PxtKM=!a-KBoomfx`sQ-W>hqqNt>6SQ*S
zP}LaK6g)c>D`Z@9W~f9xTo(W&aYNTZ@v~G_3ZJpwil5<8!<x0bcduOoQ?rayvp{BS
zSa@t~co;XkJY%iFuy&0Bt~RXAFzjBl)*xuvAE(cUtMMbT^QVW$svV7Qy=C0*j%8v{
z&C;bcDZ=FJ{rfg<*uQ_n#(n#<$@j>mOC#qlT{>6%j_Dk)nu@hk(gQ*2jACkuqG*J|
zdOzT2tJ$++w}cE(jbfDj5vU$Uiad4^``%%S!0oi_0VyU8RAXB2YetFvj9D7~5a5Fi
zypf6&D3ycOLED|=;;WAoj73Mj-qET$^-|~S*Snnq<1T&%rc5e$rm6e~bger)<}a*s
z2P5uQ0RZ+)tp0rQ`z^2LoCCuSvn2f6Kji#`ZtfO+s3iWqJB=RHdYF1ET3DAPa%&t~
zh1Kz>id*=9(D&Gb2f}$!oNh3r!}#yATQL^d-<82?(n1$j3yp?<?_xEZs#=lbef2?+
z2SvZ8vlzOjs4^`rI0@0inw`VFZTAZb?r)>9YL9owvC&gruluUPaPubcGUlldZfQjS
zZrs8G)Lcy=%IZG1Sis&lEh)4eibb!PL8^fEd(9B7-Vd;j!Q_x~=|Gzq@b%K|+n3fa
z2Eh~OMop;iP=gunpG9A8L)BaeS{t=~E+R38UT~b|DD_#mD<q86otef#A4m6>pkFjk
zuS!Psk*Gd-1(>QfEEF`bd7*JHOczSXhkHSo^8b$>tycAHeWZRM_vjcz<6}&HI@HY3
zC`n|wge;UUM=2vvo=~*z;OAuSdtt_y@RHQ!+op}5zF_agt<OKd_4$g~OOp4cq$rwr
zIF+PMTmNMD#p4!i$ci7k*t7fk>8T~KQ}~+j@HME*3sa`X#!j8`0)^AF^^xYb*w<r}
zY7+6|+wX4!{yXsPwZjP_T|f9)nR^LR+su%P+C4jE2}a6B%N+1Q$B|U4dSD!{(H!=^
zs%KQesu}eBuxwu9q2;6l#SV&`>{L;%QSh?2-Q(TTpPjjE+03Mo(LUe|s_GTAFyr+a
zE|RWXRA$wD#mZG>Jv(pSvmw#ZA#+kv_9ZWY!?WUh1wWS(HY;_MJ2<%jJse)~@|wNe
z5_@RQcxeWTR^sW3@eDQRIno!|3uwLnyG3(_qfx<XkcKAnne8#&(xqxcnDIB*wCKO!
zJE9W|Rqx?%5la~FaNytwJb`*o%VltSp>P03Ekw`4+ASy|9!A2bcr;UdN2PXE2hn#N
zsA%#5YS#o5k%z)xgi-LjcyOYnmLG#c<6)Hf9FEY3a5Xk6V3cs+MR2m^GCvwk;cp2C
zgoE+m$QE!^)J0qG;W@N)4o^oAL<BT{BtfVV4!#JE=E16X;Vn2X9#dv)%Vj#p96ppE
zPM-I6ptt@W-hAi+L9N*extYR_BXQ6uyvBv+YsB$Q&okE+M&S_`0k=8``~v)F89&^3
z6fPC`IEXY1XFim>!#oCR3=gMsdjyvY5I^j&73OO=Rh+=XCB|dY`113P<Q|Wta`B;p
z3NFXPZq>jRE=N#l^0{!~HE6`LHX7?ly)e(PN-cyTy5ciHDtzgQtp~yuJ_A^?Z}q70
z;jg^EF!y#&&R1*0qR`<Z-3u?prRL;+dWrLoJv-yYcQ;PYuFrYp*2_I%kLJ4{ZOM8g
zpcXy2B-IRLJqbxwHCkh}JL~Q$b|s_}1_4&pU85!i-SPfZm;Pf}NmJ>G9bcqfUsh@?
zdw<(CV_XRIOU$Tl<objNH<qs3_Y>dnMahZM$J^7sSoXz^BV{c+mR=LG*2P2LN%!|>
zB?>n~e^&i3{4a!-pNRe0#oS*0lM3Ul3YtfyX>JyKf|KW>q+BR1ZFfjiAwJ4@6YN0=
zoUhRt97c20hBBd6xLleABj6m`a}!SBo#0if6X92QCu1JGT2v(SdQa91opw{g%<INy
z?Lkhv(D&$Xkl9RN;37(mk0jcja2UsD(YiDwvM|6^>YPiyk|U4B$rH!3R98qVRWrqc
zK)tS3Sn=hif*-xTf7nusQ0<l<ya)U!EZKOzTetHY3l1LS_H8+IXiL_aZr#rCS?HIm
zo4<YKVEW2;fakTVp!<RJm2VrboPSM!?)*;ug$w$!3+ME&omVODS3OM;;1&ruBVn9)
zjVfQQUQnMQ_oqu*3dB2Aes#?SG3QEU<RNzuk2l0tmr$<FZ-MRJp~EIQS^j=h+14S$
zCczx3rKt0U(jg;<%tZe3FV>e%ojOy>eLcWC@e`PBLzD>zuOPfu^rLgo@lNOp(Hq^C
z6ec4CLRX<Fz!*~$^CvVm7T&s5*jRYtgp{wv{lVX7b#NV-*3IETxIc_<atnk6xCfFS
zKrATC0NV_x4qZo1g{rpix-l)BIGIamm-z+8z2rRu7z+FiXk~`bqfmS=ndELtxss01
z%MrLaLKqM7Q07Z$J=ipgcw$kV+92oz9lQpwH6s`Z1BI)s-h-@9w%1v*BmE;g&?u7m
zMCEzg%*uu8(eIa6uf7ai^q+#~@HB_q@jslwVBmaN9u>nr3cnjK=08w980&$gR+O?*
z6s-IN1fNVjg_=u+@&lmuE>`gV@e{hCHXKK%uI?fQqDs{cin3RVab2iAVK%N=6)Iw9
zn2%pmW14~V!BEZu)hTe7G!Y3LZ6W)e&ebL#!odbIiL7YNGV3_WL9(_``+tfXTVo!t
zu@|eVV67sCN-Bt<4b<=@7S%AOO)Lm%-t`cD!m44QXpS#MYZRq0upzjr5S>6`C5%#$
zpLrU8In=zOR0km7b|d3E<eG+5xCH0h1xK#`VvOh@TqF)`kSQOb10eNap<OVseQ3lI
zJ`L@|g2Kdg@#)aYko4CMtj_QIPu8`!LM@?U&N`(Bjm;Wr;RWgdJf32CEcj9TR56vB
zsc(lcR4%3CO%krIWL{GjTK)rX3@r?=tVG^#yaDRfU0U+7X0c9;K~HaORQuECpJpCW
z=_POclYBn*0A7LVpq`Gw_u6xOUto{UPVPVMQ+FvV6Nf7^X0V)BA1+B68-GbM>yf4y
zA0`p^{UvcwF)rkmG#vF6>}qY~)5LuH&SH!_Q&!v5Z=gB*TtHn>weh}LDF^X>LA+y7
z7IlCm8v_<gb>S>)Ey;j_@0CT1gu5cd^5#L&CZM5e4*rs2+0#uqTAWw{yhs{H%nSNe
zfjY*R1a+pE`nkDUufnew_&zKD>pyYQ;6W27;^SQ$-mI}8`HmWt@KSz}>&7W-dMLOD
z#EHc`E=gQu8WuH}@e{#6TwZhHw^xpHlr)8zC=cRFmRx!rk`;zEa(zK8!Ik{j1C|Ks
z(GpfQndpv!V3_YwDvZW_{In^CH;5%Kupy*cagap_XWStGR92297o}p{?)n9kkMW>V
zIT96YMfl4(+M|m`2b=EPP0F1aSBEI;%Ve2)b0L3;)7_<H1$Wkf3k^0Q0!Wk&?e9zb
zI1|S-OcbsFs_EmyknM3&3nU%m>cXU-d88mrH`V{xtFDx`CFc7iLSugtXF&P0i&^at
zZ6HiJNs|>dyM-%K;gH9(@d2DvD%Ky_PU{yvX~EaZ=>aKTRDCL+aJLBzO+Bq%gnlbi
zLZcG>wkm!*vu(U71+kgCz2(OPyGu*y;ZJyqJ|p%b6=P}HlW-NeR=i0bXZ!qdy!4j*
z9?h?@c#pd{Gt(EWAn%EJG^J+mp)e7q&L*gO0v*R7>rs88PJCYYP^y<;!o$t<j&eV-
zcR|sJXP+IRn><+@06bLGz>vo1CXFB$!C*8wfZyMSR^nH|rHK3pcvdJpm4AxV0|o3}
zllGQUVav~Sd+891vQbU}%0!vf;BG$11Do=}cl?2tDAVO6tDi$_KBzP07%j{bgsI6*
z34f=e#iyLxXN)BuR=tC77Wg!kly|Jo6xT<R^vbtA@hX8F<q*nhDbhgFI7A-1w~@$h
ziUGMDN{PHzIj~78|7dM&)`Qlpl9y@Jt}O5#Ofwr$j<}ab<Jnkp|5vT8%IwO)p~1W(
zii6o4ZJNK9D1dKEU)CwW1b70Ls?f`9L|1d_@-qG`H{Z?dEH#ZZs<2`kjeh~qY`KD2
zMhN~yTU&dvc6>KE1Ie`ZQ9?TRq#0;6nVIN;ViZ#UB$e{rlT>=+x7*hBG%+FyrKuL8
z7#FI{*wVbTC!2uPGKrV$OF;c6<SQR<tS{%hgf(h|@v=~f&C3|*Dm1da5QD!|a<P{0
zz)Iwk3hL1gE*5Ff9{e>G_@g=|%fj$H5YOYyZgFT(rGw&%orHUigc?wVre^L4ie0-Y
zZ*6SQ-~UEK!Ob0eN4}7;X;a1vBmYtx^l7nUVxI9IFlfS@SDrh1Q=hYPY|Jx$1ANEf
ze^!3N;G=U<p70fOr)|#2g<;*Ou$@_@9r@qPtzlLAq}<Uq6YD5Qppj*$qZu|QuVuiQ
zQdsSPpig78#xk1FdYGK=EA^ym3jrYt5AR$H_j_jR0w-^hK#JB`13&^06ukYF)6Bgu
z&!Cy5Z2&M)NJ)=iksflcwhel~U}h4br`n~5lrr8PIS33QJ-jJD*yIHoY}_6Pf=Cji
z*bW;?<){?-rk%xnQ>`i26mOR|#9ZBCtm3RHQ$Ft|=ch^<q9R9?ys1o6Z7HQzI$vhW
zJIQTZA6K3nI-SZV^@PB66A-5AxDqg?ykk!Z76m6M;Hj*fK8P_-vUc{zCX{AYtz%5W
za5Kd@cb2nnMS!GOX9*aAuRd{ICW{#H3|PzzJI8@!GLVBDUaLg?D`|=qs5lODku%l;
z{vi2PFw;EN##ziIR=&*<MafM$0^9E`5!T0<V3tS5nE{fLpLcry>3nbHXqwyS2>ixG
z^WhXZTg1NTZ@b5f*%aU!v0_O1XjjqhNH&|SjY(sSv6$DA9{9v`GehEBaoy=-W`nx6
zr3M)%zV$w*r+F#;rP^9zX5fhaV}R(k_Myo@L4jU<$tB!_3qNcdaXPi^1nB&qgXLLE
zs(oUEzK%{f7<d3}Tf;Sd{hm;==lS`S>njiTS+zYfG%_aDxH9hfk3O3b&+b7u1J{6|
z?n!JQ^Vzh3&O)8$m9Sa=%!QP)<G}ImIiUhPz39-$PxhV2pn;H4G1Ymy4jBc{n5epC
zFFv=Boz)Z<v7g02W%aUlFygvmQ$5ihR#JI0d$@57H?f~*A31--CWf>N4|>K1o{EM=
zZLyVb&8IybhFN&hI)XFYDu+C00CdhC@}>;r=8oXc`pEcLz+{IG@O4GCcwwFmdYI=y
zTVVi+CalMYpS-&rIz$c)b6f1T%LQTwGka7}$#h^nR{NQBaeE{XGeiYsiyPzwR;#?v
zi@3Ac%VDJZrgfdYL&N~3Q4=Oa1=o>S%Rv(-{@X5dv}r_j**<wNtc*Sk)&ayzWy8L(
zLokbOP{J8oT+_BUlr<8zl^sN)c8GCeuyHcpE`?&-lN>55vV+(__82A+Pm8{{g>|e=
z$sz9(nj^t*#pi~)LvnJaumX&*i|d2T%;t91U&K}Hp~dn+137zgGC@e!*}_yR^UxMk
ze72!f?lwN#2GOeImYMJ7SQ3^~c6<^sK2FWBn20tQ=Uw=LcAP?!MAVs&ZF=soOmox;
zHK^GJ3QaV?=OBjd4n7C9R(eoLGzi2*F$0N_;{;33pQ)e+Y=#V)ks@6fG+`TuWr>6-
z6m#^!s}g}knIYzw45pQj+Ej7q(D?~F(Q*>S#WXMTxf;71Vo~McWxL!W$_lyHvA;DK
z?e3GFVx<d~#40?r#Ya|(PuMwQNUqXwqNAHCt$;S&MB9_H>%?Bx=5@Q{>{5@_VKF5=
z-RqcKNuZgx0k+_^D`^WQFwOF>P_r#cT7rn|$stWW-k-xT?@hAZFWSmh^^{X$`)U-h
zYU+988HX<oOWz%wG3wWn6}40IF)4R`2s-_ILf*LfmxfIKI_p#5X1oXwIGsaH=g<?0
z2LjL49M6D#p6tKxbD9hL!gCdvtkrq?H@!u&rX{?%0yVyxUHzu8GiuJ~3l`9G1JhS?
zTkLr~sy!TNT8?EluLS#T-grt=SCM+eclD^>UR(S5<jj$0R$%g~?~x6DXX0m>p8^--
zS$M?Z;^wXAQRju2-fw-d4)*Eq`p)&}8B-RoU7m1eUrZc^c{g}E`}l%osBu@$;kSex
z^P@hCkDr?)=?R*8dd`f%TEesTbC5eudA<_svhd;m9r>k%<o|LJB%MNYTh6T>*Acwz
ztp>yA(f@8qBs@C4Fs#0JDn1V!Ow~TQRN(}iD)fW@;-ZDXVp8Cc#h|1rdV0zk>YRq8
zrT39u5o=P`vlmV8DQkI2uY!YOEw3KZ+j4C)BYPjv8s-E(Ik!ORk&<9VcO@?ZaX}pu
z0hfIUv_zu}i+QaSHxU-7Vv6{WG9qFbzebS+^80#PkJh4Yh%|^)lfGOYM$Xd|K9Xn>
z+<K2sQ2&8#-vLrg9$+zfDqVml=in3Mqx1CkRC9`M)ZCp3A=@6lTJoS6^uL*&c;<os
zFYAs(fG+;XSX_Dg+t(}a+~%?t{xhohkD{VKioXQ|Cjy@pG0V<oXU$*Rl5lI!8~?gp
z`T93fAIf|_h-+54(pV4bMRBp&|D$TB@dniU)ut$Weth81dVX*X&HD;L<&=fM>*m*>
z-@<9a<ir_Z{mezz(Z_XSkCuF~zfB)v8@0K6uroEhZ@P}i0G;i@p)k_y0wlphc%X-(
z54>kPkR-D!jHC8oFqPJ^4S+O1$!G0OZ8TQj$?giI)ts!mXm@7Tc_qGMYeD$8w`Jck
zb$bL59Y&y=-DQxwP}JLpg~^>JTAN+G1)w(p-mp$DT+*uzXqvGX7j3M;Pk`0~;<GMP
z1}K}b=sdinckbTR|MnYk(va8q?MwIUSt|C2;hLxq*;(oj<HkvY=s-d~?F#b|8$++r
zo}sQ`;Ujr^EaUS~f@q_W8@wmY<0Z^`Qh)~1CUt3DQMGW?tQkd?94(P#*@S||3V^9}
zU99xxklI}Im73cV64R#k^xZU+4??ZI>xS`@lEv4Wz1qA_9V_JmOMN|Pw6(XM>}T>v
z(#s-W>}BC(cAsr(VIj4yXfmo}oA*h5LTaTub{!Uoz`F{4sx~zLSxo!E<>TMqHR}t(
zp1O;P)(Jl;-(HH(lRCX)l>HbS=<{&Qj7-0E_<5<v0)3qM>c#mTWe8}u2mCDP!5Ti0
zXJr?=I7cyk$3=ce83{t)-6CEti|;EMEU2Vj3AJL61)bbuA;`mlzFk_X-?_86SZ<aO
zvUM_}m1Z5LeP>hbG5KiLl?lJ-h5LGmlj)~HW!f|1SuH-#H4(m*A2yv+O=f4c^mI7Z
zF2EwX#bh_1^rGoKv$~riWZ(~W+a&zWEg|nqC-piufYdKRginDLU1tg<98Jo`$e_05
zi+TE_em{NBnR;gBxCfe1UcxOY%Iy~5>i$GNMO18(@;O&YCY4}LU^!mat5c{&NE+H0
z6C{=d?xN_lZC7MYmq&D;I526YCYr}1V>!GboQ0DqN)3gAj>T!?XJuI-QeS+7Qh%ZJ
znqmd$%hJ277m=HJtq3=1j8#%)$droYW_c*HwnP@Nusz&W=?~O;lPlK{(f9W;x0nrN
za`qL%Yh(y4``8%FXV!yuOShicEy4#f&6B#pI2pz^)h37#vq?7ss5EBK>FRCZzLIY1
z<Tp()7K-n7Gu6kJM`(~LGJ-T|7*i#D#b6>FV&AFGEOO6N9K{GrmdbCl7E~7RQ6ACI
z&NO6Jt&6ADrG!$qN_aRFc4f>j6hy)d6V9y1mL2Y8v)kZay$q2R4wm?;^}a(gy=!3^
zifybDqP{8(64KYqA(=N`a{R1(87Zy*+iM$ch8oMc#i3tk)dE*zEv$4rxjFaLk9g)j
zgWSbBPAd}5aSdl=L+L%nL{x4;ymjfhK(qaz9+lu<6w(dGWH38l9Yu|ZJw<A|oW@T|
zJ4Jdv8$Z8`%Be0Un}TZKAczG*qa)A(l}Iv*lvXHYGOD#1g^92@O3;vwKop`Y_=pgp
z5`_~`MdCFYg`dZ|nlBJ9d^O4&a8lg~Tt(F&E_wBc$Ym>*MUMDzF9=8bqFxM6&dyE_
zzL*&qHh8?YclSXf6DF-M$_$+_Y?4<mpHWN0@^*>(IGKauKR}8(ur(8`$1q0Qn@y_G
zV0Tv2oUCuCd7bLEvx6M5-rzZLmP1xSkj=Mz(`1%#5Srq1g3#38W{iP3-rT6l((u7S
zvVr$6o7jV1>vldN2N>8QfeHUYv%aZz@V_BOmloQ^U$Pn3*+HM$UtuCo4Y0>mv}TxH
z&KiLaR6Fae7kT*qZG8$reCu!z|B&TNNky`CE^yDIvcLW+V>>Lm5V)Y<&7XG@W2>cC
z*lRlc-tpsQ$B*N8A92y>{6Z8Cjx7X}#1SBU{{ht#yRk|_K8iu$oXVM*IgiNHL|@#6
zewU_abCVtxc<%0WZe@^yV*03{5yc54Mxo#A5ey3B`G|Km&SBmK5fvzp#m*oT2&~~+
z4TYf$-635tWEwn<lKN=?RVm5U!+b%ze7VEEVrAjJBLiLdUd<?_x$G_A{ub9D{CpEN
zflfE!lfuu26DJJt$u|0-FO5BQ(Q!@2*Yt24(Gc-`7)CmWvd@U+OtcCsSS?Q{316#|
zTGKVP;>=9fWYa4O#}F`NI10rc9HvgeN?2eF+jTW%CTK(z#S$6?ES9x!T_`}3w)2{v
zLZm3ni$TKBGkBPu<852ZF{c$M5zEL7nn=sYfLSI$+tiLhGpw&0C&K-Iy89BasE%ym
zuDT5_LBmcER1g8NB?g)X5s(BBfv|~)ZPT=E(?El=Nr(dtA_C&B$Pxz`HHwOaAfQo0
z6qBei#`yd*d44{hnK5R@aZHT*)Hpsvm#?bsy}hs)lS%%6-v51(hIG~0>eM->wsV@e
zdym86=Hxmk!V2X?JSe`R$!e&$2aP+8t_JWD-@9)k=r<YnlA$ObIOfD~^0d)N?T%s0
zX;LGW1ig&cbG4O>qO<{&JmE_zfeJl2M&a5JL^1lYXYXO^Ve2U~M1X}Xc?N(+L~?SX
z@t+Nw$<1LPb=Gi;)<71C1BV`0PEK6iE*eSNaMU0YNwQSn{y(nO<Ht4t<x(5q!qr3@
zF+|E%o)%_pLq|hGsv%eF!9VIPME9d9PHKH}Zkas*r;X+c&CN}11txW<YoT<~!29%Y
zj|ILQDG0&FH$hpl$xZ|}w%mD{>rk2F2F@KQ;|}lGOmZ;UAXAg1a_CRPk2i<UaY%d|
z7ckN!=RAzy1aE@m$rFQ~@^gj6<sQRF<)+*`=FiheNd-f6-ezEGV0ivad-}}vEYI0+
z>sgM$0SJGr=UDu$-~Z5ih$FYrJkNuwCi1C)W&v~5`*p+aDpmXRCf$x^<YksE{=iT<
zQIV*NqA#$`8+2WX3jN(gGs&P_r<i6g*Qqe0v5`-8N;VK_-4Y!8^qU>0FWN|^YA!nw
zw}WU%{*1rj{}Ohkl>$d}vk{3KKKZ1t*-!KWdA$rSeAp}&jh<j8F3rwe&+W`e`CQQ1
zx$G|-PV*p))5~;kXIO;vFY_RbZy1!5(@IVDg9c&6zp!;iMkp}ezyUe=eiFB<QR3GT
zDI=%&HGhbFhNlHwY$Wn2QA}RTR!MmVS;qBKh@rT`y%isZF6PPrv<AKnzy(II9N9dE
zLo>-lG!Ei|zN;SykjjqN6Jyy80%;^E;X_VRCJ2U-&H5q4f%O?jDU1^%$NEgF>w3D5
zV3S}oussnKCFTQZ8j$L)fY4p%&olD=R^Sa9s2?R8w4J!$SS(5GC-_uGY`PCOM$Cov
zR{Ss6(`D4n4Nh3nnEl=9;N;}s(>KDT(pPUX@|6q2@+?M`&r4qq?j#2K$LTcES7*_>
zRDn+)GBQR(<m=8IU7BPyW|KwA+&^F0`{+@x_N<%Cb*psKdq*}ZnkHV#C_{b6`wLUO
z!;?O$Z#XTFW#=!Zn}L|_k#sMf(u}DDB;7mtl)d-h;b8=X)rWglZ)fE9kM%xg<T~?@
zfpf1;L;7RC!i3=dqm9nr*muwz!=UN#g9nGv`b=O~)6-Sk!N?yR?R~6-{W|lHHgBR(
z2yaAw5g5)3>}xnXF4MLg{2kF;WS|1Q#K=`hjXqNuw;#o}XEU7XTW6D|kJ3?*fR|f5
zmYXw;9;~?J3<E_MAOZJ4%FLd;-v9o``z4=Fmbo-6*l_gyjr9tUvk0AznhU^*4}Qb~
zjF@|}zVVE-R}Xs!hCP9<tL%n_?IvM%NQVOc7@SM~`yk+Fl35<<*t}pgpGer;tkED*
zb~d;~HE<&;V&tfSZ7(>3+jBL+<=NS!0W04y{3DZtW&BDpQ>jE>DHLEbL^_pH33n*i
z_Kr^<f@0jB-pw}J{sqN_s_0i=3UVvQda$v0rSl@k-B8W-!9R+zH{;&(VI?R=tq;h=
zR>4OU<|ClRFkdu&%SV8^Nq|;fknUz_Zd;>b<O<z)#Veq)4IqVHH@=`cW?b(Yez6vn
zV=1U=Q9fuwIYbCxDatbFQxSdB!c}&bXM;}NEuy@7{kcNtIjzA=SA-(mqAE+HQeo}n
z+A!x?8Q`x+0X`d+yb#O*5k=V~4c1o}ViY{q-^wSz;$T1KbW@*gOs)i9EKwqFJSMs<
z_<P);S;xIaBRaGT>kwN>mdqi@YLSCc4bj`j`k(Y3?e9Oj_QY{ud*TG6LA}SH3IO<^
z7pXtlRCo4l-R3ifIUPfJ8}PObUR$EG&W7W3e<S~haxikJ8wnfq{b0N4T;{WpadWvX
zx3MwzFWDgW)dkYfzcBL6OO|dy&E#c}^tI51=%Oh<FVY)*wj&(eJk9mf0{D`XbfZJ`
zT_wt5pWHQswPXD#Hhtc?qPG%HtRrl89jhn_iKZ+f-JJmRp|Wc88N#c9-V2ysnp48n
zjd3$BOUb2~f)~nf@5~0lOB-?)qI>@qAzc^_a#1}a-=tBlvH~^e^8B#Sd92R3K79z+
zW4fmd&ScM~<gzzplLzv|7Oz)5md}b_BuUQ5OqT4aUerGGwbC{7a{S2Oe%L64`Xo*B
zo#XF4Ym&2@Pikn+*65k>xyfOoU>B!JzNxROYcNg`dHalcm_&Cc*o`K0;|(v0Ebs|>
z7(RDOc=E#J@F@*z&`mIYUYg&$4D~#}Mn<j<_4SxBaiaYc-}y6gs?{;Cih?IOO~x2i
z?0aE1jN=Ha>wJX9C3C}OlH_kw2~64zc-unS%F$IN`WDv1Cpv5~U^6?1u)OZE^I7P^
zl@H?LR47J&JCVKe<CWPKA8sMxbVhA$2Kn5A``KbVPjEkTGCyecSKE9b1j2mOE5oxR
z0v_+DXP3C0@FHh}aHAK<J5MLPwiPP#(KXriZ)E6C`PaXVzd;lkw&>KjZzAl{C31b?
zHI^4w(sLB;3g9*Eci(D(C%RpJ`H~iW50q#ZI1YB9li)R^;`-gA{t(mV#Aa}{Eq{1V
zi$2w0J#Pk?$OjdHjRK>#Chu2BpJR*ARRuatx`9e`O{oN~;3`@ViVbpZ)%z~4f=nn}
zjI6<ZT?M#{ycv0{E=d&wyQrS~U^!Pp7532MjWR^@-k%{im3UVJ$E(%j@@KAi=baTZ
z^T!>MWH^q?6sMMzkqlh%p5nt*)rZ8MC|2TiUnD6@le6zfBpALIhChzYm1ull(SPuL
zMfN73SMhp<OF_`WydAZ<vZ!FyOS2|d&X|!tAwMlE9I7gF(ApFqpNxdmtQTKP@|`j>
zAXN8v-<P!^{x8oI=y~#jDTCjB5R9%ynB?Gs+GQC=v~x`^daHA0`h>zXO$6abCKeL}
zO(;g}vz&zt1<4rhZY*0%zVbLbb;-Lcwu}wOj@<+RwB_yZlm!(!s#K2oQ`5@J(<GS=
z4w+DOsJiN~|CA&=+-+0*b@#RMw6Z|aw^L`3u_JUBLz<tSjlnF9d_8sbu-I*@tg5G{
zDrN1&6dPHPtfX{dwYV;7{vj~(P(qmZRKKz#W%Qm`GA7sp%fRTK7qlP=SWy%>1|A~~
zNvXz<zn-UGH2!3Y3_Jb=eg)>qjz&lWPNKM6sq7&XtPmngp;i`><rqI{Q`*Yz?iIh=
zk1W9kVZrMO`5<zH^Wxu$Mbkqg-&<YT(<+Njr#0nCWciTEC3(rkgQ9n`CEL7iT}fuN
zug$bz|Jm2BfyauVtYSCWgxSFX3N(_}m(g?%IMWpzyAtNp35I}Y$zw-Fe50Fa;_}es
zB~JN)Mv`&N;%3H?lC_eT!#%whC&YUAb+4-I!XirapK{+TfDAp4-xZc){4^JjgDX1%
zEJ9)>ZH$rDfi+IKf#vP(<$-zQ)X4UdBn#`6e`;yz`tLpbI#H+Jl*In?AT(g=Ew6wQ
zd1@)A3nh4rZ3eH;iQ<B)g4J263I3{aD%JQRsM=DB+CY4=N-RP$qNiRG4u*=eEBaOD
z5zwJV$MF9~Pju~AsYWt#t_GFh0;3rEeoL+MRFTv8DO2p#YDVETZoFvf1g9zJv96uX
z6LI6|Xxf+7HL%w2j5>%7A>EQ0Q>V`GvQ^KYKha{0Z6XUrG=AJQUAbZjloeB%hCX#2
zng()Drp9^NH0SYNUhyivGa;e>JR=eCb4NKiCRA3TD=HOmtU#B57rKl+!neQ@T^I0u
z1V^~|m`Fy&MnL>`z=4Yw=}E<r#iwv@q0;cZ5aRK79WXu)CL={pY#scW(M2_?`LFw>
zI+?5H$BF0TL7Siq*%-JGs#d>&wt<v+X#t)Rw2}Ps$`q6YQ{vqvYBh(4xrV3xfIBG6
zH5Iy;fi9|u;Ul9+{=pPIU$6wzHk9tzvAUCy7B_?F&cp$_15=VjVjhNyvqc<Es3MBP
zBZMh>ThWYhjiPyCeeEsHsSR}C*Pl32RaYH4R8`s2Qz<PekuEGP#Re5UK18gaQY*(Z
z?noJX)<^M~+~0sI9jUT;4A6{`FhzkhtW9@=yQ%+kHn!;;dp2BlP=iSiE+8hoz$>kM
zg`Al5EZ(G}FBoy6rv#hyW92iZQH##<<Yww$t)6LLopOVy70Gj=_BF$gT9N(>K*SkX
zcNBvS>{Nf&z_Plv3&VeLh>GlquegkvAo7}v$Hp-Qs>aGk<6)7*lw^@!&p{oQlNo<*
z2kNe>h-B!10D&i7ry&g6V?f0tp~S<q;t@m6#3#;PBl2=9l+22jB%~!sV3}%G@nq2i
zjVK^ICObPuYw?Q5Y}cv&FS|~)a&(Rm$E!UjN4dEBPVo@gkB<son8E5B$pUpBqaRP^
zo`l86P|s~=gv<8Qe*}g{XJ<zjs?LA;N?^1mE4n!Tm8l-y_7kFlVx{q3ZXVta6JtUW
z(x{KXd_06th*t5Hj{?}@b2z3Xg^!kJmaJI=RoAY~Ug~S-Yw<=F@KUt4D$t)oA}7k+
zC|%@r^fvg|<cw1ywgcO?s<v)rg@%30?%iyyjE8v6q_8J1L)7xL&B&OTA6&3nm6kAL
zu_IAOJW;lk0x9ZAQj0|tUkV4ql3ybm#lWy+o3^w`rEI}I9Q9T1+*t|2P~U>=>;)if
zz<lG+1&Q{nK$IGH7!4kA@g@F?)?y*0SeyV<KaoGlJ^leZJ&-53!o{IX4{k@+siYP8
zkHJCjtxHRqx65~J%egGuu@g5B1ZT-VJfHC=yIzM%C}m@0at)KcasZX(ZF|daT&?(h
zOZgS-j&bOH@e6d79eZ2s2i2UNz+VI~VjJ*yDQe5nJj(6Gv*7&qFba9IZ9Or3F>Xt5
zb?}7}p$}RMN^}oF3EB-)Vo<U&26QUuJXAcrElj3)2Fbe7S72plDG-540xOB7Kx{{o
zSlR~xn!pk-?9_JwKu_=5-a`n$du^SaZMOU0eRscECpv?!ApGAM5DYvJ@BqP$H@`-Y
zzP@?$YcTrjn*y#*`9r+c)v)!O%(yVllGJaqWb>|q!utLD>$mRTzg4w#sVY4$&*Bc)
zcIPh2xvTC5G2PwhNH-6sg|2mJjz0VGfGnCxIcaP?>_~t^X|^I^C&=Href!oiTedgX
zElN*cH0EQKlF{^aZR_egd2n0z33W~7uF4u<zN@0vl*SR{9wc62S^B%!Tbk)9ghKS6
z`}YHnedsJDap_V@;vJ+}XQ1%6rIFFoNcIn3!qI>W&GheK=}ruR1Mr~t<voG@pD~^L
zFuh=A`ci8kn~|YfYK_|b8O`?>KEH#>Ue~=>83|gHy5#P?ib$kZ7{MNmXZAQF%g?zR
zBj})z3=|nK0$nP}T)e~@wNvQl?-HS?jloBtOof2WYbVU5bp&jPd<9mP0`f!XVr^v!
zU7(dElZwWnyTA(k4Or{ug5(n?P{(mwupXp<1z;U|6CFV9kAN>22b_Qp`akHe=mx$2
zi{YxDTaSzH0>M+kUAwzPkM9yawtH?pX7oIe6IKUumtR1bz~YBitbnYosDO$I59kiz
z2MXJdyTO#SBkd>LbRBYx=uKS`dK0XJoj;(9=*EGS@Lb1_8p2mwxh|lM;vsWWLuJLD
z?fGInG@ri108CJQpCG#b057g>9=!8`9#gvsCa|uN!*#yh6c?a%mc{J;0GGd=O~_7`
z`IqyHEBM@h^=$m-F6I+Ri9z2b-JfrCU5r+%)v$;*#cf;g=!Tdz)3(G4xmp1U$%_Cg
z`WJz<=m;pzt0xJF(TzUL7a%#jnNxi3dcRpPV3vLnOcI~c49X+MGIp+Zfnx1}wGCx0
z6&t%oHjDzJKRbCpWvjNNR_O}Ak!Sp2>&kfx>(ZBP@7!W@_J>>D%AMMnxWyV(Sd-z7
zjTGy*1I>)gBeCONpxVBOO{OOZnjNf&oSUTqt=i(|-Pgg!c<|aB^($f0jQJzA+c)J_
zw%k^8up%)m7oKK+BdR%o^xzr`;N_gCm>0}v&yq#WQ!{Geo|0va>6s-8uq!H9qx73T
zLMWOS5GqwCk1lQ6tVya0j@Lxe*ffXkTRfBAB0{e`YHn6Gv*(%g5?FQ5M{lly-lTN1
z3}I`@vIZ5Qcjr&gdj^kt5gtcVdi8^7hUq;6Cl}=GF5ld3{+2lyd+wwTJ;u@3$s*`k
zcrdAWTl<#LpXzSC_n$waRBp_=s4jJq3>k!OW(eJYBgP6b*@q`blr-elYg;Nd930s+
z3RryBcW;_fo4R<pa{5c~x_s{HqGj1((ydrO4sN#j<k79}=+fM@fRN<m*knl)8{6Gx
zE}M1eYjCmUH2DnrA|Mg#w=?u2gKcGp(0_e?7~Qw2b2wDG?$G<&lVU1c8;dksthZWy
zxBK7^YZy50lhyD4<Ey>a%QoyQDQm7-lcQQwu+V-}@6{doZ5Xcuw!3OPx5oceU4kJB
z4J?~O3JrLBfjhuL6@_lY4I1rEQbZty6cHFpWk}#BF_|PjFvP6GF~pvr0dd~mQA&%P
z$aqceLU*?)<$~~_`J)z%s8eS*O2eazS}kfb$AqOU-jte@5Ii$3-0#H&Mb(SKG^0nY
zNXjE;Bas+q{qyjI?xAJS4IvA=taW=U^7h**tcvE$&CZGwnewDnn(!1!_h^0|PpZ7u
z(aN#ev*WY#Fi+8o=7)!{JmoYHp7uP0r~G=1b3L}nQd&nsFv^%Gfr*}w`Vk+2J-TG3
z7DYOU+HayeUmiwxCarZovtskH?$xp3j1Rcfkc(V}Hx7P|o_w~$XW7>GHm+(RhY^A%
zj3eS_>LBA#igoZp4|-rzXWvt@{=MGjq?na^n}`n9S>4)m@IejM!Lv0~2MaguEj8%i
z=J%-%LbA&vtf4l03?I$Ydj&XH7S?>+qJ!{@b2=9o0eb7v72sLSpnex-y8avr?xSb1
z^i0l8LVD6;Vz{qJgW^F@G5Qir*8tb_zCN<lJVR=!vau4z3olZ?GRC+!!3!@IYLFOo
ztVIzZ8CWX2x*}nw?lAa<;1rYGnM333Tb6}!2ET#606tm2CQh;^QHy0X)>uLmr?B?p
zW}O$b>_h_S*bFY!qdpL}l)2p>1a2TFg0l|yy*-S4Z=Y%=@;$3_gW)Ff_0^vZ^7g^Z
zboQ7JDC%Kcqhs29C*vX=7Nq({a!yP_-UMema7=+7`uw5vO}aLdVfg!=3x3tob{mCF
z;hfl|OX9p!9A%#11bW*!K=r$w2c7|++$EA@-Y3$0=0m}!7w?IihvkRSw=Nr8J}#^6
z?X6ARGOgw{cv;uuJ#AC$88EWqE87d(y$d%VtzOqBneGa&Y(-*^Y3LuHRs6da%NVcC
zVLm;dVxw1fn~apY_UAXEN6^i5X~F(?@4XXwbNM;Z_=Lh`@iC$Ph0uHN?Y6cz7rz0D
z(HRTmRS+MyY`iQgzJS%4M6wS<_05Tc?KTYXDh&L0B^^;1`ZVYGf@QH7dI5#Lf4Qp2
z82bD;0v!t0k{VvL{_h07Cj407LW`K-NDTOM#o#G;SXpwZa@C*FlWVJH1%j#`)RE%9
z?$oDiR^%#9F)gK6R#jfDsJL1gcR~(xbvK>*|8fe9J?p!(y!I5UOMzGqt+|>Dv}_7i
zMWMB>vI7?Bv@K@M`A;sHnO_C@-1;8Ia_2^DjTJ-D6h<D?e+RGq={X*^mudu40!vU$
zF(^WtbGVk3H)vZKjYEuK%3uK6@K3?4$2^77nlL=fdZt6Sp#$dVMspm|cI$CF^+4X$
zx~dLXIDmIX;Pvzo%*6i10c-74c8}zFiy1{33dzKfIyfJ#OG(iel6mg(F*LuyBiMv#
zuxH<@3WugaU_t_ST=d}ap47OYm`!=Pn`1&=UxZp%8WMAI66fU@h)|ZVSQMn8M1T^G
z21TSX4V$0d5r&zy)0eum%J*bF4K8{TgvPJ5p^N@^sQEYa!`mw}R{kC2e77oN<=g0o
zi^WY%#YK$`4#58Ckqzq)Ti<@K@}m2s&a7P*FYe0foaBD7^2m3LVpUdVb#-ReDxJ@6
zZOP6ZrCRQ;5RKpYIcP_U{MoQY=zxI$u`lo`eK+dX@^k(Z3iiJPz>H-HafL;3;f$v3
zO?4q!4}vW4Y(&9nbK(mM;$w>lZ0yS5^5=q`RMHV?fGx>Sh%bC2E`rhQU!Ij`fGv$R
zz$R}G$Xd>t-fVWEiZcL!K!3ji>z_IM;GKo<Vs1%NNJC)$=T#wrsJt5_r9_jt4il88
z(e2ZpuBkNZBs04&RaXxko#<58)%;;)?Wt4f{@MQ>dn~{89>epkW&gLW1OI<qAN;?F
z8GqY2>bHzPehXjxui)c;TbmvrP-0ti3G0ciK;X{VL(I@%gV<Kb8ARn>Vi3=Nem6sJ
z5HTF+16$1Y{Ze~qw(n%uTU{qVJor|3pL%uWPGS#tR;=OdA%@+KVLzWeY^&4TL;noc
z9-8g@{=(;X@9OR0_Qlj5hNCRj9>NeHW@95`d#Ep@NbMmEfl^F^19_qG-1ZQL#Dg$u
z5i_wxY}t%0;&&NT2XipIIt<T^z@vLMaQMw8FRvxj!Wp4C9Adu-hD7-&g6B^VF;ECZ
zNdch?Gm>8@ZQ7zqT1AXw2C-dOerk42{<CePzDVQH1sGY|h~X!hwX%2Te<j-peb!`J
z=#681xZl_XZ&WV|=Zzz~=EE}Fr(bIhZyy<j>f}jcA5W7j*PMOCeJ>iVHHWj0OuVwY
zdp69%_VGH4la;0OPt}ebz7*szMIKqYJTd%ynj_-5?c+3OVjp8e{0oISt+x*xVC~~+
zY#$5b;|iUzeJtecBj!2QWDQ`jk6iIa);{8PC8lctBm2m;to$2mAE6j#;;|<;%td^z
zm|@Cg4y=|r&_I?&EWieG!EDw-LNRM1HLQu$@Fp^jm`Ia~m()gb^Ya(l$Vjt&2Ur^!
zY=LEng0PJ&ER2s`Mr|Yv!uF?&wUOYtY$OaCY$OFdtQHLGm)glL5;tNyIn>!)1lFA+
z)J_gdXL<d;eJq{1xs*;HJ)J!us>hhl-`<}8%=5WGHG4j%x901NjJbRc%251Y@fA_n
zSF{fD6|KhSKrjl{<2Iw&`?CYjfnbOJAc**tD6F?#Ki98B@sW1s{88HNn{!u^NSlWj
ziQ&2Oe&rdlHgzkb&4viJVbmPfsf=K#OQG>L?^4vf|7v}<KOExR<zK}g24NfC`cL>n
zZ!e()jN#V`yULW4SbxZ^6)^lHyq5h{{b3OChw~S2k|ib5#RBz*WU)}8_lJ>K_v@d>
z&kuW^;3YEg^W;3i$j=k_l|R$Z|I~THv;2I1J@4nEu%B=JX+N(~vFjicKOcqteCtp9
zc`0588TolJ*2lh~V+)hL^Itc%VD<8!i7kRwD4n_3g08$C=wk~Ykm2#$_dKzS@fLQr
z*mwu~Ct{Z%9J{psoY-Z$Bfqv27VkC0E@KAoEo0i7>3ICiz2(9<92phR$f%9oTV^Ap
z*k#W1<49yg;-V-V7qt!@7qybr;6J#yC<@0#t;5AdY)cv!5m^&^);#1M?k|tELU63r
z$*)DhueBD1Avn@1H(rdElSpfnVKJJT$_84DK#JqNCM*McE>GjoGyv0ZZ<r0&>s-9Y
z?YJ5OKDy<a&272z@t*OXD~b15zxhko#AaPI-rIoJ#3bGePKz|ed(*L;mp@l5IpjVk
zjwRdJeNHZxl*SrhFBd!knZg#f-k*(e1rDR%?}#HJ#C{PGe%vS1Ns`EsWOBOLa9{E4
z*<xRk8!iRhg6_;Qa+giO_qGLpI9*kxO0NaCYSL9THK)nlNR$sQnUAD*5J{!n00(pY
z(*YdoxqP{2{DIFuKY;RUp$$|++gkTZB+&IBK_%|D5p}`!<~yhr^MF=tNBrRc#UKCm
zuOCa>KRJ5<b!~(0AP59P_iY(9$O`=hT}LlgQ{CQ(&O@=edLUltaOm>oL(UZM`C4d;
z@!Al)fX6Gw^j=}<wej?~wuGJm_V{NaC3QR|8s$%}gaQ~31(oi#y4$)I-S;HF-X+up
zzGUeo{jAuI_`_Rz;M241QcN`JQdfij0u#Y_aJMF7o9-$)i_YlSdxW_Ed7x$KWsLBS
zHQ+t(PVs`L1n>74?`^`T7rFw^nQQ5LDRSS^&%;xKV;3ttF~oLUp=&ZN=oKqF7}?yL
zFO%i>Hdj@pzM+VXRlJc}g|2+q3esZA9F;Fs#DKItC!w|?2X)GtJ*WO&iaK*DNS*Ky
zjOR}*9vk+Ks}(!<tXL<^qHwSqOFqp%s;*8gTJ-wsi;7aKtDE01Sh%p@eRKtb1xX8<
zr+RK*2$C>ZcybTg6SKlm=~xzn_OyP;=6tvUBQYNK7Bqdx3Y>rlMT4W}g8xp|j@YfG
z<BRD554m>a1{0~WlY>3+QQXqoT%didVB3jd6-;6*=pA<)m5AX)ShaR7SEn9U2^|gJ
zWip3Mz8jfo^xa4S{~Z)40Tdn+9>?~Tj5eu~Z3o=h4<(#I@;&N4#dregZ)cgYp!02X
z8lBn&Iu(XrF`P^cr;mq2%4Ks^v+)PqrTYZ5BDqkiJ4Zg5u?m#53xt5+w^5S*S3)O%
za2JKaQ<91O2SNOaF=Cb{LtA4s`;g2CJfZ7py!myr4nc?@mS$s`0O<=%Powue67FI!
zx#`T7*tWJ8>L1BWCgpGaCzD4YSS9(p;!JBRu>NTEt8Z`G6Xx3^fA`+Kcjs*LE00N#
zF80aa{jl*II@X(W6bS0l&xeOb-qn7CHvh5WE8w?(ZiDAY_pNc2hYp){B(6Jt<;`zV
zzjjuf4&>=(y`n(ls`AbqPCK}#G`bgkaJCLTJ+>t1Wko>5A3@<a?*iYe6+bpb&+wT8
zp#SV6HPuJy81{FXd!bWY9lv3++4SEr&0=GYjLY<qPXW;7V>pqt#*s<3WgVZ^)X<6Q
z75Y*rf7(LxKnY#|ZZ_(8trwdj6vKLia-S9pTFC8IPlo{sPy9a`m8+WJ8|dp9w^X)t
ze@o<o6=7VG6ZO&m0j)3|qj=h#l0j?JU>L{$Z<2QH)^^jA)E;v5U}M{D-Jo8^#&BD=
z!(bgK2$>6QMwX5=UEM*@gLs<<KZA#%;K7rMU%;av;!#BX3LgD8eZ*yuf;Keo?|t%n
zpZCf81VCLog~I4phcB#<#%o~>6Zjyk<A$mU8z`uU!YSm{3*j{5pM^8Xs9(Za<g}~8
zQ<%}#g>&QwBjzoix-P6xNGV|rdE=9?PL{otL4G968jM}xG`TE1LDq#c$m?H(v*cI8
zljO(3Qz+=Sg>#Hs!g<F3<S8Jp$vdiP6ttH-8D(TS^`cW6vf`UP7uHa~TVWk{@ln`7
zN$m-zP*hKa(~Q3n&R|CU5YFO?ri7<pYS)Bw<U1o~QN`4ZutLfBCaj@oNKT$%(U84(
zhDE~>W(`I|I8DAJJb|MAQ#eC@FPtSm6P_fy!c!>eH-&SIUlz`zq<t4Ips2kQo@V?M
z%jn_=CkWu8jSfPXIOD&D3aVHj?eVFJP1tloT0Unje8oN5HZ0DT*a45@-yFSY=5Ta4
z9`n9W6U*JBlfZ3vLi5b5RjLc--if(s2chemtta<7wq-7Po*5ZEGjMuNaO~`J-7O{<
zZsFoSCywX29Jd*4qtEcLGHg5TzGo9!(VjiZ-Kf?ocdqc7M*}<9z%I5(=Tnw1<Z483
zzlb@~Dw@o=MyJk9vrO*<J-6$dRjX2&Z|rRBZZ$>~m8F74t+F`RSZ%h}%Ju3j#_ojp
z(s9MdH~z=8E7W@I^YF3FeqzP#9eK7*N@^_5HwG9)m&HEDGg(~ba5)BpfweDLE#C?M
zYsv6S3Zo<NsSb-8emf@XZ}OoRI)QJ8?y+O8G?V4^xVoNf_Sxu|+pmq2<7%=QUQ8R-
z#>sIt*$m=lkZgA2W;faN<EEc%9>&eXWYdeAUb1-*HxH7{PTcGyo9(#SPB!g}ZL(|Q
z<hY98GzD6n0C?JMSam=fSrea`r5i#5q_pSmPGi_PRvJo6ufX-_NiT#=2n`8MHqiFm
z-QC^YUA}Xd@7&#8zw`U{y?NQ)^o~FH{eJUoX5P+Z6CjX(e*+}pU;eZi07$5T1{E*`
zbnrkWc%ce>;D>6cfm#T_RH%b#FdbHcm0=ZF6;^}QVGURl)`GQR9atCEgY{tp*bp{?
zjbRhm6gGp+VGGz2wt{+?0YL~s7$Ojb255vPXoeQp8n%ILVLR9!c7PqB722R3I-nC~
zLKn<}7<9vI=z%#f7kXhH>;&_n5BgyMc7_G83k<?S*cEnz-C++{1badpE8rqH7!HAZ
z;8ge&4uxakC^!eMfQxYo90{+(L2wd$4?n;$a2gyA55n8<9h?hS!cXuc{0tYuHSjn*
z3fICAB;a@$h9|**$KVNg7M_Bq;TibM%H9j`96S%#K?=Tw6X0cd5nh53_yWFyBVZKL
zuo%W516ep9a<Bx(ArA#8f(a8a37^AKSPsizFW3k6hTGu+*bnxF{ow%k6268zQAZC}
zq8F>s2Y<p}=*McT!CLqm0RuP{>kx4oPREsSWn2YU#no_iTm#p{wQz0t1%AbKa9vyv
z*T)TTL)-{A#!YZj+zdCzEpSWR3hUuFcm-!*5JMQo2u8618?gzSu?4rrZE#!M4&H|k
zaC_VVcf?j~!*=YzPMnEdI16LgjkB=_u7?|N4$j41oQFH%eC)%19KfC7clZN7gpY6m
z?t+835O>AhaCh7T7vY{5$01DMFdCS|6pmmTM{zM`;4U1)Eaq?=K88<l3FhHaETD-+
zoWMz3ipy|0?uC2fKDaNu3h%)i@Fu(k@4{>F4(^Bh;{kXe9)t(uA$TYrhKJ)3cqAT$
zN8>SgEFOo)!&!I&o`@&m$#@E$il^b}cm|${XTgnd6Wj{7!9#E}+yW27f$#tv23Nr&
za4*~kcjMW34xWqW!O?IYo{tycg?JHOjF;f0co|-fSKyU+6`YP&!{u-$oB`j!C2%5~
z4VS@ja59_%ci=U6EnbJ$;|+Kt-h?;fEqE*5hPUG#cqiV4cjG;HFW!gu<A3l0d=MYP
zhw%}76d%LK@d<nqpTei{8GII>!{_n8_yWF&FX7Ah3ciZ3;p_MYzKL()+xQN?i|^t4
z_yK;1AK}ON34V&7;pg}Teu-b<*Z2*7i{Ih*_yhikKjF{#Kl}xM#ozFE`~&~QzwmDY
zLL$<rf~Js89;zfSRgsVUR82KhO97fnbu^8p(@L~5twO8PYP347L2J@lv^K3n>(Y9(
zK5al7(nho~Z9<#UX0$nNL0i&RR8KP~NFfSSgrd|yjnqWV)IwX+Hnc5mN88g5v?H}r
z8?{phb<#}gqFEH9ZkkO!G>7I=FU_N!Xg>8(KMl~%w19S@L0U+=(r&an?LmuZPm0qJ
zC1{uoN>YkOC{3fZm@+g*S<2BkEulOW$fP1o&?GITWwf03qP=M!+L!jD{pkQYkPf1Q
z=@2@U4x_{A2s)CEqNC{;I+l*3<LLxCkxrtM=@dGZPNUQ53_6p}qO<87I+xC)^XUS*
zkS?N&=@Pn>E~Cro3c8Z6qO0i|x|Xh^>*)r%k#3@!=@z<`Zll}j4!V=>qPyuHx|i;w
z`{_UQ06j<#(ZlozJxY(!<MaeQNl($!^b9>q&(ZVrUwVOFq?hPrdWBx4*XVV6gWjaK
z=xut3-lg~Gefoetq>t!h`h-5E&**ddg1)4$=xh3hzNPQzd-{QXq@U<#`XBv5ztV5?
zJN-d_(qHtq)$CCtP17p0DVnZ%v`WpZRcSuWuT^U`TCEn)rfPNCG;O-JlD4w8ingk@
znzp*OhPI}*mbSLGj<&A0p0>WWfwrNxk+!k6iMFY>nYOvMg|?-(l~%9K(1Kb>3u_TA
zsx^3u*>uqQQ`o`Ss&s3Ir*$lz$mg=2IHTS=ls6^}Jubo1noH%f#$r#L(bt|x=M%-T
zWX4$POE|o$eK=>v6A2@0RwZnzcO>Fs%VCBNYyY_E=~M?A>cCDu(2!8o>1<@!)YGZl
zFc|et-qDcY>vB1jayixI?3Hr%5`l-pVTBRjEZ6E0hgZxRisvgvtbcmUOlO7-JuSf#
zQ~sxw|1tJI%?XPsPt%MP>!$RmuiJITsKeQZNTb3g|Ke2MFtV9=b~v5Td*X?rY3LaV
zYD$~JcD;vFkdaW)W5rjIvHt0Ec~#c6D%zm1QJ>4JvYfbVd^}e$^SSX6L+i|@G$Whx
z^s4Z3D!g6}FDIe8cchq2#q-6nOuT4T=iEx4&nM?yCpW0j8k+R^d~%+lkLMO7`1)Mw
zDXf^DMs-z<5xt)mn;dh$xFAzp&;VahkuPXKMNm``3~&TR2~!60>1=9BQT(qSD2t}(
zR-OTMkwx|7FL3!d>EeSfzRcnJLXL2`gsO$kqqy9rQ!=@1s=(1Tb47+aTNSodwF)13
zdEBO+)=tK_;p90SYF5~yuvH;HFX2cNV>IGT<>M1ZA~!bVmFoqQW07FkEA!{FL=`qF
zY*N^)uvKB3!gj_G-&iE9)`iu&uv*um_HAWvBW=8{Gcr?OAl=g^85VF3Cnb0n45y8}
zQAiiOlO<NwW#q@KY8lED^g&*;OoG@j91ONpbQbeD7HzWkg+rl6g-u@TVvTfa#2oRP
zBUa%tFZhz_i4yl0tY;>xgg2hg=O!~o()37H9Iuj?7sEU_oSV$@_ATm|mL`SG3R^0T
zv2k-*!7z34xMU1{R;all645*A3x!Z)Wv4kJiI2NyiHqJTx!zT8rp=76OFqw%2V5Oc
z8vLrsNm1+cL8<ka1W&A+H;T#h@<z-Tb=PNdiOk%d4qwc5N^FHwYGS2(;B#W-=lEjs
zD%c^_HI?9+QnTb!$DVbI2%B_vEa2*z(q3ke^LvE7*~gtyaXV|xe$_Lj@QN0xrW7*q
zf=qI=KBvAgSs2lKg$nw5?TswXkqaK^E%zi?(`!f0b2(DM>&jZBH2GCeWueMkii&xQ
z3QN4=-=ab~o7DS+HSFuN-LSYFPq?}-VGO4;nYdJezH+C+e$@)3(EDYEyrM~Rp1FCt
z9*@SbBdm_W86D)wgRW;{aK&pMEW38!pmWW-a4i^O2>artID#P;x9p|V3cJn2fp}R&
z!O)6^a82Bm_h6`8TKkr+LF{Yuus?3+KNxZgYeYSXJmQEjw_Is=sC5EAtbBsa!&Wzz
zHig!{h7-D5g!P1gh%aHAEMB_eV5B<XTyZcGNR-DPsdZkir1ltAvxsL*UAh<brR-!z
z1F7=0QPmS7Qq-o(Hj3(^9D)sA@xCKjV5EF`Lm<tqVYJj5ylKh$=g-zh#f(P3`lyj2
zFj~I8QO^iy@@4G6Ep7+iq-wQY3U7vMRZJ<>YOqNk6VR;71ued;y^+Q3javNbRY!_S
zgD=jSVP95@p5vm;xr#mKD)!up#hxoK_MEHObN?*%oU7P<IlI+ZyyQYC=*cnaBH5vk
zFK@eIal7C{e)aMsg<jx>P;fPef~z?c%9=yr-<v~=6tyY)W3xUX9v|s?#B1{Hhax`H
z-Z-L~MZ^)-YlRegQ9zWT!B?~$i@JD2pjhrobVWl$P0^j;Xt}iZweuZy@rJtMin)*e
z)70QEI@b|(3v0wKcQN9KhN_|}_)yfLmLK;0MP0n1R(_Le@36j2MXh(BP-Xw3k^13G
zI%SbGljNeLBvlqsHc8FWusttRB0Qv99~0tpp^c%Pp+ka?-xt`_$uN_lizo1l0-I(r
z#CRm;8tG;efAp5PGd!E2he!AWmrZjR<}&p11pXjp(>#Wq80OcPKH|hY{-k45Kb!hp
zrw%ae%&>rA7luKeKIodhkRi?^{2}Dx>YO2l1dnhdX45c(!H{G~@jPy=Y#Lz`H%m69
z8Af@8J0Y7EGq8(3ZhveVV-xo?Hf7nwJ&a8`HsxFy9cNg=kmm{9L)cVcQ^A#8lTBRW
zY$~#e%h$zKND~Z`3`-f7F)WwhU2f!a^@)jKh?%g}*h~)i9cd<~x`iArw+P`{FPKM4
zGwE08+Egvzqs6~Ut8`Ve6tU*$f}6=obCNxkM2uCJh&20bWhMhIE(djCFq1w8X{wT6
z72H%TsCB3us&;8JSy!f%H_~J#t4oB_namaQE+!LeGtPk{t;D*89QH_yl{bZIv-AX?
z@3)yX=8?k6Q6arlJvfSp{0VL*tK>Hzj+d#7*P$G*OIz`lDUO#-R=g#`TgtCid@T+Y
zF0nF`-fUXkso!S8btz2sBo1eDV>LFDJG-R)hzKPg13{`w#C?}S*E-a76Y>a~xw0xZ
z6}b{brATh$P<N<8b4-#+FSEt5e2vXq?n`OAFO+m&km?d~yDxODLtXC6BW&h!Uv4Vg
z7eu))w{fW3eWBTX$+%{V<AoZVxvO7DyZTwUt6y1M)h{egMO(^OBBfFmyt+hO70got
zE?!onmZ{dEWi`sP>&le71VwZ<bC;ls&Mn;4W?B3;v#U(RR3c7!N&g&LR#VEqGF4VQ
zA|noUYDPF<GjUcTm$6=c67hoJ8!BcphADZ?csy@p#m^e@Ww1;ZjA`XkPN|hP`LoBr
z=z=kQ1<m^oS#P<9y-}@{&YXbIiw)DBFjX0xDZY_q<0FO&z{n1ZH*4l&Hm^$+t(Rb9
zEaz-A%{Dn($!S(ANoC5XNs}|p6KC|R-@V+TNs4OuqAY$h6W_yp>US=|{|_8^fqDP{
z000000ssF10(jaT%()GKKoCPw+e3t?1Q3M)8GDd(_1^*3248y8SOWqW+>QjH2YI-}
zgkWa23yWP@?US$lX$7_EfRthcV~PcV&VSSjFE5BUpLU(0AJG5^!+6@QSb2O_#TC8x
zOjeS7Nxp;-AcO#cgoGsts|v^x5JE^0BT%$d3?jB6&?G1#wN$}UZPBW&)>=hHMH)aB
z0TCiQ0R-7M5hMx%DpH_;tUYhO7a%s;U)%igo%`PV=FOYAbLXCOM^r>=nUw*}1ev|E
z^TbmytfWxx5kEect5+qFMg?N{#9sm=P$I=EQ4%e+B}VE<tkjiyQeWbvfy7HgNsva;
zSeo#?rqWE3Bw3nEilj;lxmkj{4e6O6Q+jmEOOQQ12IM7Za^G%45_DX@-0TD`&&keA
zK<t3rehC<v$2M4;H=u8V398|7IK4Ifx88zkxWC{U?ypu2_ZL#b{e{+We_=J;pQncV
z3$Nk+BL3(8S`HglP@)m}4#wsWE51+T^YcfI(Zu`*)6z9%c)<hrYufO_(WA8eh@xTn
z+Nof4et~9M%(B?u;y{acTfEoeNQ;FQOA5!1D$?<zi-s5K#4)_lDaFngrxrg@I6|K<
zF6OshB4+3`VjF#n*j7u4?Q|Zoy)Gtp&@y62T}`}6Hxw7Ar|A}Ay6z%o=zd}wtsu75
zqr`T4g4kZq5Ig7vVn;~v*fGTjD1NATBtlCbC@Fx~860BA-&Zsm@$@XLcb#8di}ft}
zTg<Rn-=f#Hb$ssJ1zHTT7-rF9G1y{&#ab3aEe2VPw0$^h=G@h>*wA9QJV|7}BQ}yJ
zh>c|uv57oJOq7Ylrt&DUnM@!i$#`P2j3YLe2Z<?CLQIt+Vhb5VY-xLLWwEu~OH5;~
zR!O%{A}q#RY+y0gVw~#^g3>o?_+x%odPF%d<;<0{QO;C3L!}Rtb5wdk=>rXyHZqdm
zAIC9Ik!R#}nJM#RiL8_jvP1UE7xJ~7kxS~Qp&G66nyhKsQM+ny?XQFN9v!J;bevAq
zDf*1QsMB<YzN?FMxt8k|-J_rBQ9Y^WAqaAZIfMJc8RAsmeNdTYBMsW(-k-3m5fGdl
zoE<zq_~nr7kV0RJkVkzjLSA<5G}lgd?Lzw|Ldru<_@0E;a;?X;Uf0I>Y}i287P;0P
zQ`k&jPhsU@=Y3b6G*6yy6rMs~8_z`7UhvuQ64yTIYZt!5cOTKrwY^=t#I<FvJ?XQN
zja{4PTK7AVlU)0(Yu#B#y8T4D{Y38ZS#OGKJGs{F$Lsdvo$cDn_Qzd&#%H5E0v#vs
zn<hHh)G;Zz69e%R48c#4kNfd+`~n5|C5ljt5<FxYnZ|eokK)&uh$;9Zp2f@f8(zVy
zm}VN91k=nU>H9ig7wRH?M>pwu-KZbwL8R*y)JG#UKtnV|0uuF-UWMpoovDA<+3?d@
zF#0Y6bPoIxj38aAOY}YcfUz|X9$kqDT@A0ULZq%il&(cI>Yy%SQBOZc6WxG#-HgU)
zhGe9mx$Zy<-Hq0|6D{=<wAFoR!`M9J#3<7AQ?y5zX<<^e!X%oeCe|dI<`|5-@iPp=
zy|@n#pb%s5ARfke{0fg@5`K#(@q0|gGk6|<#*26ff5j}6VhvWK9P6+RTd@;6upbBT
z8NR?i96|*S;|Q+cvI#XI#$&=vw23k?rnaeX;_3hQxy!xW-AybHI4qFLe2~r^+E&_0
zN4beQ>lW!OU8Jk@l-s1Y^pPC7L-OP&GDPl@yXB`clsRLR{7R<E)7;H}lxO96`IG!v
zUX)kmHF-zo%K~{%7Rn-7%w4`-Hp(XXShmPEIUon+GdUy`@;P_?Vfj*y$#L%f6LL~c
zF$bKJi*iXW%N4n*+{g+w%nAV-q_s3mJsPeN8mV55(%M=_>uP<C(+12K4Yjd0(Po;&
z+|iu5<7Um&EbXh=+D~(IsQyfg^+A1zIplGj#QagJb9I@P=?dnM?YdVF=;!){UIY+`
zP=up4Vi1QUq#^?y(2+i$MW3&s&&%oaZS?t0e1cDL5TD~X{()2YC(hzBuF&V9^tp#V
zkEYLKOg+<p33Ci{?FM>ammHB(oSBE?PS&>CoueDV3^zgxl<dgyJ)u+ed3}jlZYHzb
zLdN83M)Nl2xdVDczt&SW(*<ypUPfX(5*d4GjIB=Sj^4;Y9{15b7{MJ>gmK(8lNdwK
zFnV6WTbPY`Sd22xWCONf7vrV^NAWdIa_pB3On|9nA{d$RjL;O5W;&X#rkBY!1I%DE
z)C@O27nr)H9x;xy3dHL;hcB@kGw^r3CjKauN9c`xm@NT#S4!{|eY5~`c(+K#;v4$w
z12BW4%mm_F`feqb@vYS+gdRMEwY*+u!syEitY@o@CLE{f-;c3@|7<l;^zmVA=k*R#
zn<GAgT>|7u0dWv>K5&<o5{K$sF}R1TyNq|kbp>&RuH&<jGC|9EH%hk?N9$hVSUn(q
z7$*fz1n_=>UL^hobTS@CAerPeLNHm3=HLm0^6BpoPJ9ZriGM&0`+Zs_AdYv>A&K}W
zq!M312LJhs6Ftm!-{=YAVLd`TrpJl@)U(8|^lRd2JwrUF=lR~P=t{f^Hxq9`Ct_!G
z;lJPNN#d7!l=u%lMO>~Q60?+E(7tFz%+~$HerU(H`XGzg1DV9z(3{u|-HAQXi*3Hs
z3mn~ayp1<#UuSH*$(crBCf;F=h-SNam`i*Y^Br}9_jvt2784g?37^DZDX*7<BVaty
z!%_#UtX^<56R-&%5jSEp|EY&9Rxvp`0@#Z^{NipL182hc;QSdE^bX^JGvs=o;s4*^
z63)|}!$n+Ryg8BsxN5j6#^3l6jR_zIv$}%`G_{D4CW07l90tCZIxMtWHJnJDl;Wsf
z>Yp^Kx%SdQI!UJVk}T;f+0svPrN7)R1LRH_D1&4$)yf+(UEY+p<ZYQDAIMT!CS|fp
z*2p?Js|ngj6Sb+PYD-PmHrh`6(4W(}CuY-o^BDh&=(!K*rR9wGwOCKDY+_7r=l<Kx
z-SZXq#kY*}bDZ;4&eL#yft**kiKOp5Wct&Z<M0M&<>!pg$!1oifod|^d8$q&@UXpt
zdj-xem)Mv+6i^-V$FJ>kbC$yrpF2EZHEJTgkSPN=mgDYdoP2iL$!I<uC0?t6)hbn|
zZ^ui#qBhItGhZ9VjP-8SR<wGbj;PYw*O%AT*tV@ZC!hVX7VE34U&vq1TBdUzJ{EJ<
zw35j>k;@*F$ElnrQ$zoj`tx_x)lbRq<qtRT*K;m|y+96oN&YG?%irV`ve`6h)YY<9
z%9*`ClFgRGb})<Yk$u#y2gqd=<gp`iv<i!zmGk7W%j7U6fB9>m22<I3$X#CQ+Zc`2
zdd&avjGe}eon~Z$6m6lcsCqNV3GFR^-J+fKR<c+xs^Bc{>|DK{I+)cIwQv#HYzDQC
zqhm^Smd@5W<chgEk1A*Z6^x^V7E==~p_W-{dG14Iixs+3SCLKD=vrMz-dnF5tWw&f
zoAqO=l&xf$?YcvE>Mq?)zS*n$^b_5$pOSeFS|w9aonEOvMk-g-q?#5#zUH<?)p%aO
z>z2+HxUYRt#Fd)kR8h=%q)*1yjn{UcRPw#;!QGWV@{h6Oh_>TOu%j}zw_J-4yVpO|
zs`U-6TL0j<s*Z7i3}R*|WmZ^ZId~m2!B#4XJga{PSp|F-HSkdKcRq%b!|%sPR}GKC
z=xd62EFL1WKY|IaCjJe1YBITL3YqCCGSbslBR@wTdV&0NO)LNRT9adrWe#Qa$6eQ%
ztCLCXm&nMQAzc`ErE)7{Zmx87B22n5zE(&N##OmwGLE)OPe#yQ>BZ<dAh%g|?X4Hd
zu31!BkiJw}fs#!f6Ds|vxWXmJQC*Tt-4!RdQ+p*bC#8~0@1O?DkUYzzcRG9`wQx=z
zVQo8>nKFcGE1YVpwv8&^Xnk#}%V*zjQ(aEHu}z$W=;R;l$!l8~TWMd9Za=5aul5~B
z(oX)b+E+iq@9!r+yF(JWKG(-yeVYvO+VwVS_gj$N7gPAg0Ex8Gn$9-eB_6L+5tK?a
zwOnl~xEPd?<yT8A)=6D#B(HCk`q&|H<gUx)E>mSSxT7r#&a$d=f#pDl0oRZN%gKTc
z2W}$^?j#R>N8POCy<lo*4|zP=#JD4HSabp3^f9N8HK()BdF*dF`zmKY+t^2SJvcd~
z8U;}89;|K#V5Q_wy&H#gw|4Tks^T@P>A^MhI#;HNQ^Pt{YGpO*u=VvC&#r;<UWBc>
zoVv=bpKM)nttw%=0X7FxB}ssU!r|C#IGLj_?>(GFJl8ly($Jr4)rZ+6*V=5{Zd=|#
z90aG2T;ve@*cI<UOT8vR)aY+h=gp)7J4SVN(XM=yt5qB&LIuK>T_ny{M`gAus;sLI
zl4E9(W9HmgBa>71lOgt5ZunFytitnYD@RjRvg@nYc(Mvdy3FTrUA7k6T={~HzBG;^
zUP75MzM*11My31}74kVM;4i6szq99Oncq`oR*h+&|NG4^^aQJ{<9e2r))_soC-o>R
zt_%O+O)tlOQLj{<FWA$*UaW&FD_<K)OnQDWtla!<BxSJ@KEx{cF#ipu(hYP@Ao!2r
zYCBc2?`bGJ2(wT7$vMsOeT&&Lg4JInq7ZGrmCKs>3S9^FvwmD-{n+1a9fSIn=N7j0
z?J~czqGpR$60PNkt2{TUa$Zv9ltZ`JeBoq}%4+|HD*t=6d8UtfnZ@ndh}Eul`YLwU
z)7z1H?ZjjkTG>;RJxH^jZZ89`)%6ZvJGDJI-1TDy+MzAlpuP3uAZqU<*8Kh4jNXA2
z)J?3MI<o@m!OEu_dfIc>!IjGWx>9%F=!mYkne}rQWFeC^baz%rKm3NHCtU4iXO%?t
z=%_+x=YmZ5q1|;Y<*XYQj_O)X-9^=9z2Ew}GGvQ^l8o0d2Me)`HTfHO6YpRy=Horq
z=}WK_A7TYo;UjEjy}sw#IgeAbpJ3H~2Iq0XFieaeYxf{qy?gB$aFm<%>bd*--zyPc
z6|(!<DaMahBL4zLo%1~a0C?K%eG7b3$C>ZU$S;hsWLa;?vaFY7$(ChVmR|xBN?qy@
zLJ5~;9hRjmrGyef2&IG&nh+p_P)Z4x%XKNGEX!r7*Cmu?SuS-65FmsQLI_Y^A%qY@
zT*o1<*JX*%{hyI+kB!BnB)#|keyi{IjYi*@Gc#xAoBw?C%^aIhLP#o^Ng4<nT0ML;
zNuT!JS<{J)BtT9iiGNeVh=Np*eh<`s(?%-3@xW*~93_VjK2VGBp@)Yf95H+-!ch-D
zgm5%ZGkVm62z%v|@DDSdnn6rI_})yE^Z0jX{(yKM|H0!wAR(SY{xF`B;2xwVi6n{3
zl1U2bM>NDj^2ty#oQx*p$Rsk2-$Tg+q(~u&w4EJR*ptJ_9sGAc)pq?(!`#B1h0Q^0
za71KQ<XE&O54T6ZM_!G7KT;O`9-TyzkRe^Umn1RyyA^q-q7JDjnUXh%lembR<Ps0b
zBVOV|T?&Yw6p{cbB0*A&rzjy|QcB85Ii9?dM94t0fV@B!l0{@Od66t3FOj9>WwMO?
zge)gNB`e6!$V&1GSw;RkSxx>2Swrf`&&gWy3$l*<lB_3t$$oNx93qFw5pt9qBge@J
za*~`TXUI8nfm|e)$YpYcTqV~?6KN(b<T`03ZKR$2C;5^{lu`vvq{*}&O{3{lMb%V8
zb<{vj)IzP)M(xx|-PA+9G@trufCg!ZhG`kCpp|qWt)fF|H62E4=y3Wl{XU&Sr_t$j
z2AxTtqMPYI(=GJZw4VM8ZJ@uQTj{HG8~rWaPXCqep#Mg9(toGB=>Mj>>F;PG{XN}7
zU!!~JKDwX2P7lz7^bmc69;R>7BlInLl)g=m(Rb)^`Yt^|-=in#DSDdzfu5mf={b6y
zUZC&Oi}V9}iH&9B*x$47vWe__>@hZ-&0w?9E8BSArK4Z+(FbK@5a*$!hKwW=(bB8Y
zj+9h&JVPjbMS2V2&*YGOf%_<HB`O}u@h_xPB!PV_-DIChjqJMA!qcR(tI{TRO=@6G
z($82kxJ9aBpG(`2Vkm2sPO(2p8(5pPfiUFI#PeeEort9&u9X%MowSdbaOYv^XQ)d$
zsiHrludt8VRrV=sV$JL`_62KYe`0M4C$87<CrX6QS)o}Kcos%9$iafU6NtLwY2<l|
z7^M%1LwcECJB4daJhlk2I^<#JVIolyHFBvU&l9FtDd)>n3Kfy*jvr|ff>HRcj+>%E
z+Qq-?;-=Uqy?>h$o|oQ~>ZDVsbu?|mEh$=VbN7?6?=9fkKJKlJeZLD{g!bzzk1e;!
zzwR#DXxAOCkGrMCwxYK>f4zUl{CV38ckJu+90MV1zfJ2u!@rvZNdwv&1a0v`KN5ke
zNkCU6FfJAG{-AXHQUZNtEg3*EK+rl6C0U?U=#v%qWs@AF8U#v%epLZ)9{?%Icu*So
zE=Wnf2U3#ngOub6&;arjNJ*Xs4Iuv$q@-i%-y!^aIuYUb=wyVm>5s?&I)~07$@FRZ
zG%$HCol6*fhCTzyJUS1OXX&$$%%?v_`seBMB!MoaOA-GS-AI%Qy+Th?6-I@bq${!&
z*(6PoqsT$%P`F5cg-_u_9T?HZo<I8J@-rv%r==Kq8h%}iC((iW6Fn%A7(nU72udL)
zkOK3O1+|gun#${%z-yVxYni}nM%i!JZ^!_)gY6(n_B-}FlE@CT!?+iy#OJqUl;0nv
zt5CKYEo?>G+R(a#aQy+a=6BJG-$&~`g;x8YXs^Gewe(Rsj{Y55Z36u+ohY}P+-j)D
z(`c(_Q2MiS+o1fVXps%{AL&N4iBVxvn9&|NXa}dlg;t<s2gjOZJeL)?S%W!s0oe+<
z5*m^OJZS-TogxNkkWqS**r0t?l017V5tDI8D}4gQk<%)rn=#4TkQ^j4fsDVweaU11
zq#9`(uAIe{_mD#at~T(icy8$lqJ~5+ON%me(hih&7Ui8H)@XThO11+s3TYF<Wb}^~
z{bE2*SkSsGWv=`jN$#X;BolM<*YJsso3A*>b;$Fev=;HR(odvT8SkXk@i4Dv>GGKF
zT_5Dw)P40Ys#EVAPxa0T^W0tMCEqRQ*fS64$yeJmr|v9?kC3;<#HWUC-t$xUmE)~5
zS$x#K>zXG-+p^<kv`ndnhp=BdK9+tcO_B1XcHsRk>1oV<tHnA=6EI)C-95j=uC+oR
z_uakreAVNjvljdrKItKAlYawK?km?mq#u7ZxykuR?+A13yPa(fwBvn$%P49hf8~#K
z9{jr0b{{!(fH!uJUimk|>(VRID_@5ee8oB5#Efpk?2w5WT%O-k`23#0=XV94-x;6Z
zDWBg{`TX9G&+o}}5nV(w*ss{HNIKiXevMF`)05b%>{XJ+XZA!svok)k_hV<+8KP$I
zv-dIMUt*Uq`+vwT6BWC{u8{uh8oP$8pRvzy^*Xzbt6#|L5Z1=paP>>}B~dGq6|l(^
z0~7<u0EJSv*cDoZ7PBQIL%7vqLj6<t>Qt_28psZdY5;1S4$9`%jFMY$PHxjAaSKk(
zpFf8`e<FYWWWl~k<<CEWKYs>)en!k74R3`6Zv9!f^`}Bh+Hi+#_eE_z9k=hCWF{zu
z+jmZG-x<iigR;3z=j3*hg4;<(ZYL?Yon+*8QW9N27hrWJw{<T41^or?+CVoDEpPW6
z-tIZP-P3uyn|ZruvYl)vvGUeW<*m<n>$}kU7m0^`z&;>u-UnH{^&PzRU1<FlT>YGV
zj;nGXWbxK_@YZ*s^;3vP(O=P@xOpFB@z!^6yV9btD6GWJZAlw?z=;-(S|fUHkC=#s
zE<&AAU)1(h)be+z*BR93lB@-&#Wj?Eov--X*q11a_b~FONm0+s46NtdNjqgcOreWe
z8uPNDY$lt-8rXI=lP;F8M2X8A*e-Sm;mjz(&MPDBin;Q2#ZGVsX-`lgZG+5|C5%Z}
z(|JE~-`>f*l%WYnd0hS_)6j&bgbtbI)QPG@Z+F=u(Z!u4=MkmEv8*9+JXj^?mpF}=
zHZ*Z2##;l)C9guC{)lgeUuoo1Xf4I>G<;`Zc2pzggiaQqO)9a@8i?;l!GDJqS^)VA
z;Dz8t;Kkq<!Amet@5T3i@BwfeZ175Ogb<4R=^%Vp;d?N?t6`-NlQz)jrFL3}cq3Nq
z6On%o`6nrtY?6vdJ-+`1-wmXMY{d-pD(bKeGv;rlr^&xbC1fYKQJO;DlzilEq<v2+
zB&QHR5B`)Ckv{_EKO;|2rIbrEz?o7ZHA^qkY-uEQfPK;v^r|$PekzTiP0~c#g7}}L
zQM65(h;i9KM!*u+qtp>7bqaF)5n5+5a(o7(xq(c?)lsplPsOf|!ASlmNrT#(QFAk{
zWTUp@$POumypG@RNolB|2DQ<nHnLA^7}lTz$hXm^4Y2;VBL7#h;`o%jjNGQuZ%7}q
z5rnXZ!6U)n0gnPdB74g)mY+rrkHcP9$Z?G7AsEd=FdDvzr+*MU0{k%eSKvM1kHMdS
z{|vqf?jS{!fEjoQIGRg2eMY*8oXS};DPl^bOGYW%(Z*+KD!52`j}DeTre8`QF@y9G
z(CC|}&x2sp73C3>M^GLi^87Aa5`PBE^HxXH$w+1A-Vi)TGFS<Ll9Zy)=14y31iQf=
z@Nn>p;6H%RfX{-@fziJV*Vseg8MrS4<yWEnDwJP^@~cpO70Rzd`Bf;t3guU!{3?`R
zh4QOVeih2ElAjlKs6riNt&BeX6Tn3kp8g4*?_b=GU2Uf#r5&G;a`gP1j!&o)>;`+l
z!@)0t{{TJ%J_|kvKHu>P!!`C0cn0nr0KHQ8=z%PvrrF>eumjxLBM&0p*(=FdtF^|_
zs`oI?8j$B!jPX~|{*ZD>IV3Jfav&MOwY=}U`_s+RDf-XgE#O~+Ujy$2?*s1#zYact
z73f&lsIrzn0nB+DHB6O0K}}ns0^h^v^pOIoo`iumrBXAgz*wl18c9TYmOKX89LRo&
zc$-vDEkFq?*ao(Pol-OOq#1hBOvBRqv`l)IR)A}8Z8&7pAe#=J0e%Yd7a?mzK7YXX
z8Sq)~Iq-S#1zdk0d=dNs_>%N2^s5>A)yyVJ_0X|qHXHSzm>Kf8*V)fK&LL9D{mc<k
zjS)PY{0&y`bGrP>uaIAoe?|-b1}*q1d4s%3-Xd?4cgVZs3iL<z8GlNe(4x&ilos+i
zxlX<Sj!Tqu`HY>^Mcp))dT3tEFZ>NUoIXUS#yr7KaessCZCFBoPS?|V`nx_oyM6d_
z>3_2REQ3A7MzE3W5%yj7nC#8v-p{B%Hwp7;1Q<69&$Av@O+Q%EO6f_WmTHMr`VnzR
zb?B=O@~E_h9+c`7UR;+`Z9uBpSgL(UH5Dl~k^jJ$I4C`dR8J!Je&~NY)?*H7A0(S(
z+em`$VUDSkenP5n-=ol;gVLKwYlS{Iz>iArLUIZcJJ##v$fruG$KCP{K@xhUoK%wa
z^!M~2VcJId3DQO?r@>k(BY*Xe^aglGEdD{4^h0s=-D$gbc@kuPt%q-+BNngkxn_;}
zYI~Kq_71=IVNU4&^`Z10u%yd>E3Zqsy`fTd&vl;c#g*$Uy)UkqrDtT%>#eIO^q(B&
z$-a;-*|XAmtbi9q{khy9@wEK3F|YGTtk+J-WlL{iyhqDg%S%`%y(!m>`<I)f2B`ok
z-$XuR<XUyKK{lz=$BkOcx$NlSi;1@Gd$HAd%-1WgyQOW>+QGIU(Y0c%gyufl_Gj^Z
z(>;COv2t$qTw%g;=y8`c;g0&#c&qnYYL(vZT<uF32hv+zSKh%B?u+>Y-s<W3=(>Yz
zAzJ){zMe$x4NT3gg6CwLG*x;s`sB!UoAfcl=VRA?Ed5m4DczLb1$E~CGr2FNH>9fA
z@6~dzNfV@<eU$f1bYGyOUGV+1z~T~W&Go;bQ#X<Csi?HG{?b-p1A0;#-Ps~?KT5mr
z)*r$(s!s41?2?}E8nHJ~&hva--58VHfULd8DqQv~yxiB`L+3rc_ZN7HZ+A+i?VWMn
zY7McG_ik7J?YMO}g}40KzC34r=wM^7U5F1+M)!DZitQSG6t{ie=64CLDQ%4U5pVIM
z9F%kEE{}VUcHs%3lhV3a>oxIF+!}kzMd;Cwqkfsr@g?6cHA<V1vQ<EhYG_^^=qG%C
zNsX&dVz!V+b|d<$4jRUhQ))wgAMu@}H2hwNy6!{ootHl7($Cd=<h4na-IyxXL&xho
zTdS+(Z=kK3dX7I{%3Hjo^W7e?DBph9O75@QdoEF4tev?|b?+0sc^f}U{M<h5Q@Rl&
zrLJ=?4y{*@{&~L(i+P-v{#LB#o1)%QL1RAR<D!L+x?VeWyw}=#`6cVSMrxxR>(=(J
z@>*g@FPFEo+mnf_H#<x2ZkL`u_Skn9R?9fwb$>S|H3%ghlv@V%@6mPwhTiOchFj*_
zgRvI5*@fuW1${Z&xew68f87$ri`Wdb54pe9>FpGr?`H2ggy(gzx35~d5&K@m^@Z;-
zNbT~@Vz;+gMp~5eVPE<Z)8~((^FXxSVzP$b9!lZYFiC^)HWK4C2fuPU&SP9xW2Fnb
z0i$m%M&xMpTRVFA8R)=IyRbh;8i#9b(7LI3^636)Iaca2(&cpT%1RqLI*>#2J=#`L
zec91trf-V=5^SyWy-}+3-q`Of%swKz83n8?M)(p*h_&rh@DHVzzzxybbIqO5t(hHf
zNiXmoNQqh$9q&YUrGZ1cdiZIpr75U)El^yxI#Mv-HsIHjd<LEh3#m40kD&$=WVGfj
z2dho0lE&~<Z}1(kYOcvOygd&}1%i#XL1NMm?`CI&*Y>)IQayJPpuJ~>dcOfIZSJ*$
zbT+yl*KKX``E4IZ`JYPL_*0CD&VRD@$1I;yF{>$toiX&i*%M7;Z5!238SN+X*yecX
zcHh=|^BsKNj4Z^O*oAe58>A+=AQSO`j97Q%krd(urSo;6nG}GOSReYa-Yo>_`AX4D
zf*=zq1|?wK7(%=Rlt#j!OspeIi2-YsGW2*kC>g8D3atApLH)1}ix4dt2<lHB17%{B
zIS1*NfQ;lNkeVzdKS6vs)}3au5@f>q^A*Hbku`|dflTD*WF7MVB}j`k-FlKp4uRC<
zFi1^~fYjtDNKKA`OyoGoL{5N=<Rr*QPJ@i(49G~%fsEt=D4kpcrISmbbaEM#jy3KT
zlztUt#!B}Z!X}UztKDXVEg&;iyw?%7f|R5Uq~xAV1^G{qo_q<?6A7fJ_#jkJ1;Rv{
zNX%FvCnM}f`youDX$aG4Ix$fdRUuSUH9`&55F^!59Z91GYCveBCWIDhA(_-ltq5(@
zhR{y!2%XeP3|Miy5qhWxp_h7zhUU|JgnsHr7@z@!K^jCDq9KG~8YTm<Ixizytk5eE
zR?<q+pAMwxFIq*b5DufmP?H*3gHndm;mG-6`Y@qb$4|q()9G}CGw2M2Ptm84!;ADq
zl7LnH5<J(>>CX{vqQAxy)zhs+MPH?_5*^(}w~<u(Tl!nXx6|!JN&l7ZM*Mg5cSzqz
z8*%mb^!Fr*?xA}Se~lhM>m8I=$n+3>2Wj4=r%49=13i!WT%ebTmOaQGBuX}xjU^^F
zj*TNK_V?`XiH=QV6G<w2j6FuoYzCV_3~V-=4K1Q1QQ?M`M?XnCR*Z2VpPsVt(=2^`
z<TEYNj~=7J58V&(5zzUa8T%FeEwdA&C=Z&Eg%UkPhdaB^PnB?%$;IO_Oync0B*r#A
z0`1V(WQ5u{P~ePx%WGu|k7Wy212GxF$cuh*`8e;Sw63S<Nue$x;Jg@laHX@}w^Bij
zqV?}2IejojcKouV)RHa^_2Sao@U3Y{j0^hYOk(BN!$L}t!(AK=Z>>jn3MYC)<N*4n
zlk@OoeQ_3g&V%Qb`(E3{^2fzPKgZINuKc3)x|ff?D<+rK$%L0m`2HIsseCp_?Mm5?
z>xv<!9i?|wO5dqu-EhaKTVIMemBzliDFywJBHSTUdY`&Sh(n-27kRqK$IsE;J%<H~
z>ZJH7=o)?ECwvc-BM{+xpiv6q+TCy|e?FH$ogcT*)&#96e=>X2vuED~gt&+N*XRB|
z1G-|xoH{?zU2%C7$*csvH~s}@ydQ%-bgyT-|8Je@?mW@m>BE1Kd+#^V|Mj(;>b}#N
z?gwM1x+g|=0MPz^?*C^%6nQtiTzZ~VO6$m@5-bPU^9o3|K=KlPZQ*YUI22yOn1|mr
zmCQv7<sEJATIp@98TZM43F%|38Cv+NW*pW>)qH)qMfw?EQ*RRdpD$vC^CZe+9LuP}
zE0+uU7Es>_{gSaq!7-#i$G&uKLnm?TIgMM-31kc?h5RijmDGZg$v99d`8!Y|x2^kg
z+d7>rATJ_M*}|sWVooE!2c?qNK&fOeXaLy{8bA(!25|dT!7a=FG>c}D6mDPkr`a?c
zaoNi3PjhGv;<ByTpE{@maoOTb=N4x=w>Z<eHJL<9X(>tOHf0(ur{%ETWXm#*Tb2nl
zLL(%R+m{J+5FLbgRdiRK4o1A1R>O{zEl$cU&i?d4`XI`Z?ap-iefoV|mF><X`Z#?Y
zrO4K25`BU`fw*jgCebJ9lZeZfXd0bGXQ6c29!;lpv<~;m_9f%?rGnd+j9ZZD+=5JF
zJ{BPT*l_j`N#|Chg4>Jf++IxU(ieFxCF_fP#@PftW6-bHsC*KWLcfZ6SdCBtY~W|-
z`yH3lCyY#V!*;aYKOA-NwVw2HY{gyM>BHe`*jiaGc*9?gt?f_A*QFxqCB(<@RqxAv
zohXyu?D8Bp#r#iaJN?Jd^e%s3)bHN8Z^FGhi)4R;=zo%Oy1Y(Be9gTrmS<6Hk7OU0
zKFP!P<r;GDz{}hV+Y(!EH%iY&_kU3PR%sG{((12Z<y<7jF=p#ght)myTIAL`6+0D~
z)yoI756?0=<{6mU(__usjJIOUC$cN*iRruSJD(#IeXxRG@JZCTJ=R8bDDVBA-c{};
zuI=*8*52-vYrOM>f@gf+ZEIM}y{2_NclXdnth1DkwwS-Yy{ksg<BomOPSkuITK_Ef
zgq`a0!AAX8Rb6jj^wHzp{Uiu$Z(GmJw|h_GcJgTd-^Hi)Yu=r`&Sy6VKgE~A&+n!1
zw=M?o(|T$Aw4OqE-$I2IfE{V&H!YO>#9e>>j)jsu4oc<c?$Y>)yEO7Fs2}j^dElXZ
z;x3J!xMQ(56XdrjlK8nh7JENIev2ZBpT1Mj7w8K>+J$r>>BrBxsrdU71Ni$BO8)+Y
zlAn1?R=Ba|PF3V8a*@)b@F1S2$V0i5IN2k7J(AV6mXSXx!rCUvYM`!s>Qa84ePvKv
zUH|P<p%f_4mO^l7Deh3L#VPJkpg@4)?yhN}K#}4uNhvNZR@{nvkl=11K+qr|NOF1J
zd*_`y|1Wpu&a7F#z0Zd;bN0-cGyCk{S_}Ptv5q=Eq~fELp_skj(zkgn$?y!~n!b-A
z#=W5ws&Xw){&Fp#6@wS!_|9Jn-bwLSsgtMWJ%9ivf1x?UBjZ8Je7oCs$|3ryv5(dc
z@7-|6YC!jTw}gWnjXYz&2YdWc99io4bl96463M36ZH%1Al^tE>N7|4{B~n|{fh%|q
zBf+o@e|Wp*p+(nNmcn&s^hJXYJUwKrdxWtf3D%r(taG$&+1#(A6#v$ycOt!AieM-H
zt8r_4Fh~Gz&otMeMd`wRnaGyt#kBHs7;%?S@7$(uvCoo3>BCpFh@V!SbuH^V)Ig^3
zUz^p}$sXwvW$=*J_t+IZgt37Swtd>~(;nI1N!!8GhF0x!aXt@oh9m8!FX8lzjLb{{
ztxZ%cVH5Fs8M+LqS(v7L7J+|Gi{%2yCh3%LY#>ri0-C>s`o|5wr>a;gGN|AUB8th1
zgOxiExWcMpg*;Z+vtcYhG*KOW1P+Y;3JV%rzKW)b>O=gS3_U81EI!<-3vq_+k~VtI
zi0gwE|Di*w!}B?R@N)I)J2*`8?p0{7py~{`{0iaP6%3w@%G5QfZeCmz&2<G`c>C<e
zAwDex9^o$B_jdeCzr$E{BC8ggc$-x@Kcddvqnm#@ze3!Q&(XDMYUJ!;`e4pZ$_7pV
zJrXhtL`n3gbLUN0NIUsE?t}<h5%0yu0e;@S@0+$#u{xf!nW0m2SkLYQG(j(2XFHT$
znshPnSJ27<Qv<~W35DkU9v=YuqzySoV+JHLwm|2!-8Mx?4mlioV@pkzcK5|;828Cs
zkAh{(Jz3>M<Hx;<!Li2KX>$Y(@)G)CEf$)yK8|rgu1(702`jD-z(<?0vX=YmccK9C
z6~Gwr`^Cx@al5V-3CE}MgmzO?JI!qRSN+xG{fj0oeIkJeuN%HO@)qoSuW(*zq<xq}
z8#e9i60T!~LxfKrjH@4GR-C#2s41fNZf8#KEjM{XLXyYs-|2DBM4$Oegml~kLB7|8
zRkS=PROc6eyJ$#QWg~(^QceWBRIje=*VWPnU%Y9b4X)y}maWi}I;2%K2?#{)LOy`E
z!%PpxDHYFk4hb^f9=>$%bVd(d|9#RF{#MZoVbl;9>yx{^E8fYrHGF+QYzLQIZyhZK
zxA8y(cB13sz%*Ms2CVhsR@Yx)l9eZdq=-8!%%wri#1)i}-94$^<GLKtoe_~DGgq`j
zrLiu^4Inn-Gye-Ypq20?h^&BV_5SW4UqW0EP0AD1pO5rs!x*`#NG1~^r8BE?t&qFs
z3E<e*tDV!@ThEq+s8f`%G;CqlnyFi^(hdR?R%`83UWcmcVZYAU&OTg}6nVxw5Ms#L
zqMR6UMC>AAURFdHz*piey_6Q>0;Uc;`$3-?vJ(@=yR@f*XS>!Q$G7tjCfzIV3cf(X
ztjr^ir&f$dpV|BY1*}J3ytZB_DPsSbOsr+36C<5WM9bU`;{U*I$!UR5&q|~)lm*O?
z&wsXb6C!S&6}*Ied9&r25D|25GC}H#z#!<ez#wFh_|LR@0@&;G*@uwIa2*Az4doZ*
z4q39gPm3n8kMYOvy-wyRP1>=0G%B5;Ecnc!lp+tjr~F7rhe=jYh$%&4K;`hAzDH$s
z(sfMgl)TA%bx(VqF7d3$amC;sX(GFOU9v{ss=iUY<SM#LJg_5>ozJE{XDaqwbtgnq
z7rTMn34=dNO26=WOc$CJ6c!Tp8s<9Ty8hc0<|_B<Ixh9><<-+6Y8k~fcKZDI^}P>o
z$)C@M<@5#j+b=oNo&$<u=Ux&hOhU*v+YU8erd!1`rk=`i&Tk;KiS09vw&+XN7u=ea
z6OP|NDpK;TQNk~=idTv~+4h1PkcqLUu6f+wKpIktH*eeGE&*pmJ;wH&8=#*c+1Bs%
z&mO$}@|><4|HLXSmM^t1No@vl&PrWm#W5LgJs>QhTdUSnHxhm)osPG1!_`IO8D>2B
zR9Nu@MX3LUuj|!g{A!uU__dvnBdcCMjueO$(jP5*Mq=6hbf${>Yxlg{PixZZNUe>>
z!lAEgBj>vl;;opXvVtRn@lUj!UQE;~&M{G{#{xeqFdyqh-WXV3eR<?|l#<k8k~O-@
zqrT;u%5pNDV|x6d5OwY>U?y|1i7k84uKIkL{Nh+pUMz_9b$VbGrM^N~qP_wlgCZfI
z^pV1)sKT7+zuYLQAS$MChA%7rBr2A~^hY;af&arTIOX~#<vN_we4EmIjdJ~tHz>`w
zDDyV|9VE)Ie6HBSXB0x0SbM~0fk*x+#kj;q2F{)>wNOZOqzota;#pw;DV+IH$@x+7
z`BCZr;k!Y0epFO`RLp<)Zu&1;@1YV!WF%1(1Bxq;Frq`3|1UI(4&6ai?0>@UivURB
ziiVN5pBnIg;coh0F!WRZ7yrE|*iZd`u$ql>>xlwKq1=|Dz_BQ|Gk&)-chCO`epFip
zFiwW0Jw0_eku8i#`;p?$Cvh@7aWcYjGLrv+O*;s_u4J{QGzH0zPyF9FGK70P5pc(q
zgRjl>mT*~p5|tc>{_12wzcZH*W$S@J`m5)`1dS{>?(Bzg7Pq@LSB^>jnws#FTprTZ
z-68mn4Ps||?f;8^;AC(QPJ9;PE+M;+XB{%Gj95tQQ{I%&XAFuG0If<X_xndF-yAam
z2G(A~8#^}lnp2;Pa6ahsHx%Pk1S-7cEw@VjHtdJ6elRcHV?A|R!=WGE7#nPQ@lW7C
z?o{{M_G$KMz731+Sg%^Gz6UDE@m^D0Q)~#Z1k5+gH#|#EoESzwlK9)B`0YF7&G6>C
zuk`J58N(8*gEz!SPT}h2O#0y_=8A0Bt%6Ns#@)dyck993ArnYTd7=JMtot)iNV;NO
znuDs?DkX_KV?w%+Upl{9I?w3GnCK5^<!4%)l5i5}RraTBrwBciqYf$8ZjiAnjh59r
zj6qG<4NZa-fX_+<L+7@I{u@90nWpjq@0JK=<pIgfy-l2W-z`4n5rcUSDD>o3F%~(-
z55|3U=UOBbQN~x`?zU4;RAlcfd2T64W~mcjZP!OR(l_V%=?~n(3d8r4kUuG!KU9U^
z<`pG#7{4ZI=iDVlUr!h%9ups_X76La`&m*%SywOLoG{s~^^dn|{&0uL!my8tub*U{
zi@n(+vYUk1Wmhnalh{RwlUh*pJbAo4&MzUz$0W!vSX;$qV?968=aG>smXX}VClc_H
z{tqjJoKK3=>fVkPKT-8_iZokgJB^1%0+iJJFWRJ@NJ!*#7+>;jw;X)}2R^(_A$_5r
z@G=M>8jd4=JhWBY_D1HySx3M8@~Ht^1yMwzSegdQ`*qzdvcj&)V^4;cayWrAPC1ak
z850v(r|w@^XYF6l&ap3mu=JO+scS+5-gF_Z4~JPP>Yrc11w9PS+PNa>#BV$-8@7f=
z#l?rab7e0HNymos-utmZ@M6O0YJ6$f$&g!t@C4ieze@V}?%yOKzeDVJ2Z5D9rh<)y
zezqn>(p?^>q56z0jffn&UJiDV@L=n)*VUW-FT-2Z_Ai~cSl+OP7wHlUY134RLLPl7
z4%^W(l2e@I_o6kbOldIx4&*Vz?~$c&UQBFtmdDgiRQU$TvsdMCI6MdMGgq#oLRV0B
zqzczO3-Q%wi+-0bM>o4Yc5K7;at-1bvSt$!vI%2$&Eb;|ZM3$P1MnJyEsuE_F?w}~
z8`;3uDqpPKrj<Zc5W`H*6BJW(2IRLtfgSWQa*|!jsYLCNyGR;8+B6Xt$}B&og#~ci
zh9W*A!T_P^p$3e;{;q-C9t)pWvpb2$5eJuT$jTz^UUA}C&W{B8DRdi~?8+pD%ttXQ
zD=aBk#*2~FnD6;s7+!9R8*N7TvgEtYS~Br>y|udH*R!ac4KflZx^H?rKOT;Z&!*(3
zSf*~KCa3783Z~Sio~96`sHP03mZx|Roe!xG1;B;j8rJz99jGc)2C4v+gGxe`)fmo~
zwca}wRZcC^DSyIIK2(Y-yI)qt@ss)|brJQQx}i)@T}xeGU3c`usP?D<_XlphUs}KP
zf9W}EIqM6R)R+5|UX+lO#FfOAYn1bp@Ra{9HTvLOmi<wnz^cI3PC$*S{)w+bhyq_e
zm!($Ucx+Z|B9)~;pS9XZMiRe-rGDR1EJA?7mTUN9W;KhYNnc+qH<huVkJUbOF58`L
z)2=TnR*(vCcP9778aQS*wxh*bqC3q|r{S!<I89$?MyF#?f?r-f-94Q@WjVb$ojj#G
zEjU#>eL6)nr8+e{T|VW#cfO~-7jQ0ouHl^@&~a6Dm2p*Ym2;JJRn}~{T-KhbERZY!
z6PV77P;)F2gk@6=Jh4>}h_oUZrcD%kTPy4-yS+7Hf7gSBu!4>OGn(Fb%5U$D)Me5L
zh}a5Sxr{BPHnE;cZzqmyW)=(bTA4yy)8>=r)BJV2->_$SeOIu&a~^e3Z)YiChdr5m
zGCgId;LvB+XV2nLWLM;9W54YHhH=es!78RICMsY~Q%)03)BEH5GyBuDleE)0<2e(+
zDd5B#``?ZWLUaocoy+tKr|r3oT4wUk9px4X4?a29`84{Z_*D3``Xu?3`9%2C`+$AE
z_(b_M`J~osw3xOi8&@tDEtfh^9nR9v(8It}V3_Ho>9nw;ah-_Q`}6`;SgH4;Ts_2;
zzxO?pkAMl{31Zmy@$WO!dn`LF$f8}Fz3rWCB*k9Zp8k$La%N9>M;J-%wdmdA73hWU
zrRde~UF;3;O81`i;%*jRR+oSk@4(Kf+mu~lE$5i4AMI*VN>U%p@)=6+tfz}1C6J$x
z-;hE`IY^`9V~1LYPKQE=Mu&EXe1~3#T!(Ro0Z0k-0VE4j0BM8dR|`6RyUzMjK7UgB
zC{@O`igN!^e5dPUDvl>%#M2S$5!w;e5hoGK5v37S5%UrL5qJ^I5%37Th_~GV-7Io#
z=Go@Oc@tHuZgXxUZj)|Mw?Vh*aVI;WCBEmDN&KTF!ql<)`LG?KN3rIXJ)%9LBcffR
z3m+YGKIJ&&*yPyfSmw9_?SKwIbD%TOap}{NlM|nd$f3X?KRq8kKfMT8*p$yy*i^8d
zaxqp!eI&&~MZ{i>C3$p97@i$kBmiTCY3Y8qFR+ib&$0h$A8p@kpJxAiR}5wW6M(6}
zs9|qm3NQ+oK8ytB0yBlZglWSFU{o-97`blpZgc7Kl(`09ah9G<?z}~PwuDWGnALAS
zQelJa$ZW;T{;d9N*39B8+YH-mL?$33JtIBaAQL%)9L*j%-*(>4-BvgaIW<4!KUF>z
zJ7qrA^Q>)}4(gWWQcimlvtUv6ky|5X_IcG;?q|bKhxvvd58J8!vJSPbvCgtCvW~NE
zx6ZRJwGOwgv;JirW?fs@S+iU-T+{FI*Q3W{$z#Ez(_`Labkf4Uc4@@Iklk?CFpEn!
zje9sZ6+%A3F-)Bz^roD}(P}26!rhRpw#Z6-EGtz=1ld&4YdBn^Z}V%=D7}u~QFUgf
zg3hp|_QHm4=xvtwE2P>?Zw05Jy5Vf?LhaiMf|(EHLuIu_6+Y$tdkqm4<TIT!4oEj7
z8}cL45BVDT0r>>!gcL{KM?Ob-AUPa!z1zH#>sFdan_&xf%dQem5;hX{61HttZ4S=$
z^+v8%$@>NSf1JGQUoK`Hj9l2@*L2S(?!TPxu2Hm#hyES~CrF3|8eNR+i_M?b0&I9k
zO%vPR20C9v?2pb9&o9>cS|1HcCIr8}_Pf~KCz?;Mt+Z~ix(gZn!*3kTy<AkCdG|wl
zG0l}|VaO@WNzPg5&V7Nu-u=#hfxUjW-tf+W0pS$mJnY2j?C6AaqH#8MGIoCF)Z!fI
z9O<O!%<5$81aLNRQm;qW`xq-~$Z9ymaVVx+4pt-lJZ{ADa-wrO#iqr&Jr&@wqf`(<
zc3TH%Y|gycsRzXt*PtP|mZO-*HW*rzbCu&P_Rf>q^8!H&6-)CJKmehIIpt!bV!y;h
z_ItoqnziKzk6NY={{$Y~3!I~`Caor|jjxKYpIu;`V_htsw_QNYAr=7(=?eyP28*)`
zB6A{(H2aGOEq5J-`1^|c{RhPdfc^A?*?sN{;cIn#&jQg&I_k(lbI(RPOx%zB5&i-G
zDeX<lUCK>LM8;0WM#f<VJOk+mt)8nMsh+IvtA<ujRQFa-S9e!0R?oT*xKFuvx%atG
zxc5$MH6GgedWn+dlxn2vBL0XTXvD*go``Yggyl{{*P+_bYUl}68CnXZg3d$zp?FYc
zC>*K>eTxV{u-LYFW_uRbPBg9h&-st|Px?ds2mPn#_no{~+@2XEagG-HPyzMRVQ_Ce
zpt(VhP>;}vP?ykx9x`h!Yd>o{Yd32%>&R%wXwPWf=)efMu(q)8;O6Xm;Bnwi=SJsF
z=bP(e;%4Gw;#t?Y02Ems$=%TKbzEX;7(MVow1gcBpc&CxzTX220%HSn0)GZZ2Q~+$
z1^&JkLtCH)(09i^=(lJEGzD58O@ekoo1$N$wb2A<DzrSB+&B5Uxn+4dF*_kUt9)o0
zZsc*$1O|d!*cx8>y195C9xrJ(S$g&(G=_4s-ZA<59JVguHK}@}BYqG48a9Enus2Bh
z8XrQJ`j&2+B0b~b;1P-RKz6i)i>Hg@Au;L6YyU}chxks*b5RLX&A((9@5SZByTzHr
z^CjxQ1mHoiJvb9A2F8HH!G&OQ@HDs%Obi|cKN|BH;~AqLyEpc1%zey$3~$VOOh_}P
zc+^uc9$vUrh_FYg`PYm4nueJ2RdTs%t&F#3wI=#oqWa+KBiTtJ60Z6yORWeL1%i9n
zIJ1Gp)nuiwmD}I=)Mx7ex|r?Dc4W5_)hg%@IQ_jP1UDIT977(n`02KBR6-q37TV}5
zS7}iCe)xgqZQX78t(I+@ZON^=ZGx?}ZKth7t*WiVZRM@rE9WcfD*;EsM;fmAz8&Z)
zbOyQror6w7mo>~U?zOEmwZAsY(mW8ntdJ}oLlRwfwSRN{=9=O9)%8#1j7X=*yvV2s
zOr&3AwPCzrs-df4p<$+>vtfRAZSimszPR~r?cK(^!*>TQYc2aN+m0u77?<tZo0OZg
zAoX_k)wCn%*cj<{I-GQfA<HaTN&uvWZM=!X=%R`dV?&T^91y#4V~ycFo!zP&Ud`xW
z203EmZk92Dr=)O(QK>`_3(f+YdQ)_R#2jGg@Cfiq@yhO-d@1~Ey2gGC_)5q@_z3U;
z@PwCAkFWkyP-j7CDV`vq4Pb=xqgGYHC;!f}(B{x{yllcfqI&XuE<3I{Gq=XIkDY^|
zo_OhmzbW1T{!lC(lYx??Mk_C9ZuC1~N9XuA=Aa&_9;p$jE~y1GBxDV;581xk^=(3q
z%y!K7%+}2g%#f>VtNX6UuIMAw(Fwyb!wCZ#a$$CCc42nfaJhIRLS$&JVjhrJH2%Sk
z(+jW-ctkWzh)+(!UBOOkC|cwOT<_^(4~G%il0&(Je!ID^*L0<ZI}*;4yKpV(G}gH(
zt{--_g+~&`Q-ImTb^WT3Rm^qQ&$}+V;-qC6g?<q3+)w)~u|{zDAWvHUliQECeByi@
ze8PN^eC&Ked`x`r_;{-rt6o=$S8-?oJ~GU`C?Rzs+JExq(FHyUejJ|0BPbpZ-tR|Q
z4}N{2kuP8?W&5RMV#QWNAr{jr|K+Efc7dJJMAgq{^7%0iF)T0Cd9w81)Y@rG-24oe
zSB=qs@r%dE_)nglr?%Y0#ZMFYz-W>ebnI_Uy_$DqC*VIF<+<gZ<we4{GM=lEI)oPe
z;NBtG#t-YPS^Kdz{z=`<%1zA;=w{)j;uheh@8)vcw6?KUvo^Grwbs1Wx0bdxvlh2@
zxVEyEw>GgBwU&9;F_-#D_c)+RVop0pE61pUV}{%~`k)tV0dg;C{ORZB?tOB+q1|g4
z(+|+(02_(t`}v%-Zs7H*MyCTRIev4RnzfcTw)+{MKsWj}ZhIqR;>jUgk_JJgmk#cp
z?v5wlpAs;KPs;8kytE<~lUvbZd_f{FD<|76%OsmGSC>bSH<)LimzgJ)hsg`iE6gL$
zo6f7tBhDMmix|Hc|2+;K|1zF9elos09zG5quhUX3;r6Qinp%)jkZz}<#$5m0S2;wv
zs-VJ3Yj8X%D=CrLQl&3djVn_}pv_8ua489)LXpnJX=qqeWMwkgm&DC%tm>1p&ogiA
zQL<$>7?mW*3{d@@@+K9?YtOs;nDwh}cUcvW6TxCPxJrta(BLb6WO;XYcYc>;_hxr;
zmu|OUS8exc7g3jL*Kl`vm-qVly88Onas4sW&C(Ba>2~??(&`d;sdMSjD7Hl1uC1by
ztO8S+E{xD|EUld~re-5fS5S#eA>pLce%t(~-b;3OYtp_@P!*8^<u&=({KrdvcW<(;
zkVZu$T{y*scUiBw`0Or)d2+L`Sd}-$l&4L9L1#ffpf)zOFr_fven`!_UclEd#IUNO
z!bxj>ye6wA(b{sTFIAl@TSvIfNq>H+1~EjD&dq6TSYPC1GT&FjZEZa4lXAebXzWw6
zZ#N%RBWMj6{+%L}YQpQti)7Pe)8sZtMGkSpFFZ7xI>bMiYD+e@HV64(N>J7C&0*G5
zO>T&`WJ{1KCKSaCA09SHolULeUHxEc*8VHV*549?MRmjbhcwe5oE;xR>ROvx>uwbA
zN|nl#%C!tAIV#AZgs$>GUICw#QtI$mSiJAuiC%~%AW!Bj(i^GB(B<#2Fz=m=CMPfF
zoY8kGnFCH}Rkv9f^n#-wkSlV;8RYzK)N*gyFp^i5H|`x((a+S$G%YQCqqo4ZoQg^9
z9I6?TNtNSt<WSU)E2T9MtM*WVpA8wNdU8U_fxxB;4{LbVkasE#XTt}VUw7ZM77tf=
z-cWujE63Oe(~4G|`49eWHElIaNl%r(m%M8J#%vMq_X?}Trt^%7wlwZZd~9^=_t<!~
zXtnPn-$vp_Vz|C>MRCPfL{G`jC<@E(F)hA&$Mik%HnBU=GBG*PGf_3MJaJpeajQ>S
zfY1KZg7mp`w=_~3A$|1s@89FUkMo=)(7As+q&lx$R+Xn>U{ptGoRfI*8hHsJwFy|o
zP~fQU_OV}7f@WxfW_1GA`6Z+Q&rE{|{ajCBv*FV3C%7Q}YY+<oM(z0)yvt;?CP#uZ
zkl}?U!MVSR63%AzT-)s^YvaQkp5f#&%(UovY&uw-9K{P5;wIk;C5=~qy0!TbBMCme
zv;(HG#_cYv5IDTNt^J9pJi4%|lfAO_BhiLzoG!P^2^%U&WnR0lB+9Y)Vc*~7;P4Lg
zdoq~QXqt^8JF<t%TVSwXvN#uVl)vl%qBu0@>+uV1()sM0JwtF9C&$sj^}ZVK7Sj$Z
zCgGYzY8Q&o7b|~utvX=KMJg5DhdFKk3zK3h04XS$an^Tejfs5q`N;Wa%tZ!U#0b(%
z=;%}33RsB*-KP0yHD09&daimbu%uke$rH~${d%_Bb8dnf1ez97GWT1HE%tT$U3yBP
zhzI-HVXWRg%2Lg`VA1y^$&D1u+bej7y`2r$$-c9KJcrBit%i1yevQiy(9Mg2^l{4&
zAhd(G2{@J!^n;OsZOxW;sYR*@=%PkX%DVZ-6PVM=fFX+_JK~1)c$?gwYc~K-5)&lp
zCHp(qD^=(LksYu6HEAgpxamn&(}7SN?+7ai2}lp+*BFev35+~DITi-fhm7uSZn)6~
zw^yFw-PZ__$Mp9u`uR+>6~&8m^w1QeQ%&t4mCU$z#qlk1-z*tc%mUW^H;ax+nmn;l
zyi4|NWMgq{vu6^B(nJ4N{1EU_-(cf~oH#Aku?VoA)}maQvE+-j0QDM6^_c+F4i8u9
zkt%tsffYnaU$H9BQBk!d-&@*CUZ4OlNWAW0+cpDlTUFd=@W8IrO$;i&HZHEkB|z5>
zLBEUVAAg3<04;gkJ#8cMO7s*vn+ra<(0n^!JH|kSO!o;XD&nICO!l8h^%%R>^oL&n
zU_`gmOt*LU>_2+untP1{UWsBe8<8iK@7fxF`<wLP=9-Hlm)Yh9Pwrx`_Q$T3am}rV
zhqaPOtkHQCf*}S$+QiwBDP|gAj*Z^5jX@afX0{&%Z4`HSVmAigEx6g@OvO8vBsTSE
zrPNx_oq#yCR&SRkQM1RJh1!41={F6Oc;hN{WdqEEdh5tGF2|1aRu6fDR|7oEu7b1@
zj$Ul{S1SD;Cpv$#61l`Sw|+8-Y*~!-09u#XwZ04Px^(gHdMd-PQ4ntn>Qe>zqD0SW
zL5QnWNNZr5G%_4JNwh2SShPwhVbwCAUOqtzDiK1<d}wIO8w<N0C`O1r_ll@QlTR43
z<c{$!UEKz3Es<h`wpY31TX?S7NzLN>$Typ>IQ4eS%z;4%s5kc@kv*r8L3AE@Ky7Se
z0`{a*h3ArLqY!~q>D^mFsw{NEy&KjpK!v!g$zrcXRtXr9kJH#0*=E5}66eh0j9P`N
z%DspFCadSlTFdeQot3=<xCb8YxuWOgNkk#_e!Jy?*P~bge;{ts8O`fbM{}g7QHzJU
zURp)*G~6u>nCUX{kcpx9mCPyy!GZXDJ6s-`3E(EG5g;#@6I6yL#k%_rhvS9oxo>ry
zvY<kg_rCRopM^?gwg)kwd>^)`CcsQ&uh7_N)Us^w`zF2)-FDY5)2*wU)WL8MJpVHw
z{`E&ck;BqMPu~3ibh|>=B~AtBU7U`9w@-Tdo9(jpDN9ugN_>az2L(+`FZ&3PUj*-v
zh2EIRU=tb5xCcP52YS6>k7S@@^5dlbqDBcZn`dM)JR?kJq{}G!ki1`^;&HbO%Ybib
z8Y?l$V_JDYmVwB}yRMIYqt1hLlyC)%4D1CU`81fklbR;zg?%cV%Q#K&`{zjNmkjR9
z2D}5=enaaMja&w;eX9N<(SIh|H&N-AChiNYE|Ohh-SoFY`aw+k2}IXiUO9|_89c8t
zMhLZHd(>>C{AR0Ucw6B2sDqo(gfXJf8+#e7S563E_||+X44}@7R#Kxs--pX@hPG-Y
zkUraz=u6~Ehn|L1QY@+~nY|i_dD!H}lcKRfJqbOfVm&(&ecAS@a`K@JPdzRk5A#As
zQa*twB(O%wg`qUR1*pA3g;>|1Z=cK6jMqxxeSW&T4CTjYHa;pkDUaD0=ymTA%6aR@
zNrflDPl!Qh;Nvxwj9CO~#MOC5(;~!Nk+)tFlSEkME3P9n2?m7K>g1&xM!dk8Nq04=
zc$f@e`Zq!f6F|`k@umqxSbPr2$`aSs-W{TwvPR7-`gGjaNTB8%tYN1+<<O*{B**!1
z20x&xg>J6dcmdJfe*Q(jKM~+FNX1Xaid$ti)lqnk!O4vu<I?k(<uI$5D-tSI+8f@}
z%gfc?0aloFhydvaC5i;0*V20f0VwX8FSz3CRawRUWRo*<fhdg!4y!*jCk#`@N7W)|
z`tF};Q&<UkvD1uV)38i|DH6)KOo?9b&%im-d7Ue2<Gv6L&1$nMv+?+O-uZxlx_FUO
zDisT>H;okH_r$5hiN(pp@x>uH17U?N^i|91NO1`>Q+DTs$muSwP~;TeQO(r>;qAzT
zFs^q0dQbJ84eb5=mL7-f^3W&)V}6Qk;R5*Zd|5WLZql>hIR7oymrJ7ulmS}xvh7O)
z-MKaIU!j`O!?`tp-Qn7$^b|Ek82_Ex><jh*=u~Q<3-|Q;-9Pp>yvzol@x!c%(-FV4
zKl#5HW&$dx``cvRyW860oBm^pny$0)UA$#@Q5SmouW8l2e`&ga_3^)TR$2GuzjaaU
HsO^6N3zOZ@

literal 0
HcmV?d00001

diff --git a/studio/frontend/public/sidebar-logo-black.png b/studio/frontend/public/sidebar-logo-black.png
new file mode 100644
index 0000000000000000000000000000000000000000..3db8fea46a8abf271ce37dc8996e53a816cc5c79
GIT binary patch
literal 9170
zcmV;@BQ4yCP)<h;3K|Lk000e1NJLTq00682001=z1^@s6+*klp000U}X+uL$b5ch_
zAW20-HZeIiHZ3wPF#rHaiJen-Sd;e_KHv9c4^~3h@UfR{fdC>StO&>uS)ve<0AYj>
z5<ta;hzeE|6%m|BRm6hT3P_!Z;-DzFXF-aHT8mX&IPM?Zr?q{0@9pjT*ZaKZoag+`
zdCw1k5fUbm=Avo<IZ}l@I@lkNPe{bcy?_E0NZ<q@5a!FmX9dLo0Pq8Ocz$%SKl}l&
zZvX-SaMdm(635~HgSt>AR{$W90N^4L=L-RlQUJ&<RLB$n$QS^yQ?q0W0F(d#Y<YY_
zA^@rl0POThJ_P{Ssgv9n0NA1&sR#g#0RWjOM<fD3y8vKSfl@32z`_8atUw|z1i-EV
zfLXROM*@Jo1ps!ASdb4uU;u!bLM+SxAUFa5Lmm^&10YNTpfJ+E;Hh75g}6uo0Km(Y
z&6i8kGZeU$&>DC0@ZjPh;=*jPLSYvv5M~MFBAl0-BNIsH15C~g000{K(ZT*WKal6<
z<l*LK=WOrv#iIUe@NXhMA@Se-yo>?_01!^k@7iDG<<<jm#{sZ)ziX*m0NPdnXy5r=
zGusKES_;s5`HOit{6HR_p-{*?9UKY^3+*Lhp}jEUi$4D|;cwuJe)jx89{$^U@Bnd|
zK$)$;ql5i%VQzM=QjX`#1VS-x_va)2$Ay3Wjw#XNG_hPP6^ij#iMUWAO~-k;QjtU<
z$(7;~>3=fuAC~<IlXndOtcFj(PPPZN%WS}Y?g0Xe4p`E506>28EsPoqkpK{9G%|Vj
z005J}`Hw&=0RYXHq~ibpyyzHQsFW8>#s~laM4*8xut5h5!4#~(4xGUqyucR%VFpA%
z3?#rj5JCpzfE)^;7?wd9RKPme1hudO8lVxH;SjXJF*pt9;1XPc>u?taU>Kgl7`%oF
z1VP9M6Ja4bh!J9r*dopd7nzO(B4J20l7OTj>4+3jBE`sZqynizYLQ(?Bl0bB6giDt
zK>Co|$RIL`{EECsF_eL_Q3KQhbwIhO9~z3rpmWi5G!I>XmZEFX8nhlgfVQHi(M#xc
zbO3#dj$?q)F%D*o*<oCahecqCSQ;kB7GvdDHCB%`V;xvGb`=}Io?-6@Bm#?o6YL0H
zgb>1Pf{>6$SWH+$s3q(pv=X`qR|$iJF~TPzlc-O$C3+J1#CT#lv5;6stS0Uu9wDA3
zUMCI{Uz12A4#|?_P6{C<kTOXNNoz>kNG+sOq(0IRX`D<Z>yT~9-sA|ffUF>wk++Z!
zkWZ5P$;0Hg6gtI-;!FvmBvPc55=u2?Kjj3apE5$3psG>Lsh-pbs)#zDT1jo7c2F-<
zhp7`Zb($s3n-)XMq%EV>(3)vyY4>O^>2$gY-Gd%Qm(Z8eYv>2*=jns=cMJ`N4THx>
zVkjAF8G9M07`GWOnM|ey)0dgZR4~^v8<}UA514ONSSt1^d=-((5|uiYR+WC0=c-gy
zb5%dpd8!Lkt5pxHURHgkMpd&=<EaVM7OT~%9aX!jHm=T6cT$f~&rx5azF+-<`eO~U
zhJ{9;hDf7SW4Fc`jUg7sGG*~tLe_Ft1M4hnm`!3^via;xb_M$zb}xHOQ$usAW~^qB
zW}W77%>fR^vEcAI*_=wwA<hlXdo2SkKP|D=YOQ9ieyum!dfL9)V(oJ67VYcW?{#pU
zK%Fd|^*U`j_jL)nHo8%|g}U2y&+0zYW9xb83H8eL4(Z*|NA+#=qxBc+@7C|pA2%>G
z2sV%zY%w@v@XU~7=xdm1xY6*0;iwVIXu6TaXrs|dqbIl~?uTdNHFy_3W~^@<VyraY
zW!!5#VPa`A+oZ&##pJ#z&6I1JX1dX|({#+t$SmBf*sRIyjyctwYo1}g*}U8QjfJH}
zoW)9uHjBrW+LnCF1(r>g_pF#!K2~{F^;XxcN!DEJEbDF7S8PxlSDOr*I-AS3sI8l=
z#CDr)-xT5$k15hA^;2%zG3@;83hbKf2JJcaVfH2VZT8O{%p4LO);n}Nd~$Sk%yw*W
zyz8XlG{dRHsl(}4XB%gsbDi@w7p6;)%MzD%mlsoQr;4X;pL)xc<r?m~(zVm|qno>%
z+^yMd)ZNTI#eJ*$O)i@o$z8)e??LqN_gLa_%;TM>o2SC_kmoO6c3xRt`@J4dvz#WL
z)-Y|z+r(Soy~}%GIzByR`p)SCKE^%*pL(B%zNWq+-#xw~e%5}Oeh2)X`#bu}{g3#+
z;d$~F@lFL`0l@*~0lk45fwKc^10MvL1f>Tx1&sx}1}_Xg6+#RN4Ot&@lW)Km@*DYM
zGu&q^n$Z=?2%QyL8~QNJCQKgI5srq>2;UHXZ>IT7>CCnWh~P(Th`1kV8JQRPeH1Aw
zGO8}>QM6NZadh`A)~w`N`)9q5@sFvDxjWlxwsLl7tZHmhY-8-3xPZ8-xPf?w_(k!T
z5_A(J3GIpG#Ms0=iQ{tu=WLoYoaCBRmULsT<=mpV7v|~C%bs^USv6UZd^m-e5|^?+
z<%1wXP%juy<)>~<9TW0|n}ttBzM_qyQL(qUN<5P0omQ3hINdvaL;7fjPeygdGYL;p
zD|wL_lDQ-EO;$wK-mK5raoH_7l$?~Dqf!lNmb5F^Ft;eTPi8AClMUo~=55LwlZVRp
z<RkfW^V=0{g-p?-v{J564i-!=s4I9|7+ZL_NWDl})HB~^{_6Q73xXDWy^yd_xUjR>
zxOiFd;3B_8yA~shQx|tGF<nx+WO!-t(tXQl%d(dJP~uopRq|?i-13f6gVLp?Lo0$;
zG?l58$;)o8^jf)N6=9WRRqtw-)mv77Di@VstZ=NTuK2J<xaQ(lPG8l0^?7aD+TKdH
z%I)h&>!j;$toK>JuS&gYLDkTP@C~gS@r~shUu{a>bfJ1`^^VQ7&C1OKHDNXFTg<ks
z+w!3{qxO28e_hK~{jICFzS$<)*1z3%d-D#19Thv?@09GkwJUg6YrRE%^={Jcg58fA
zVjH@@=6>C{M|V%fo{xK_dk6MK@9S!GZ*1JJzrV5xZBjOk9!NTH<(q(S+MDf~ceQX@
zDh|Ry<-sT4rhI$jQ0Sq~!`#Eo-<f~6^$6=oc`MR7zjdrFqivu)p}p_ujH8_$UL8li
zxBq_cG2>%($2E^vo}is5J@NVEf|KK?WT&2;PCq@=ncR8zO#GQ^T~S@VXG71PKNocF
zOt)Y6$@AXlk6rM*aP%VgV%sIRORYVwJx6|U{ozQjTW{-S_si{9Jg#)~P3t?+@6&(!
zYQWWV*Z9{iU7vZq@5byKw{9lg9JnRA_4s!7?H6|n?o8ZWdXIRo{Jz@#>IeD{>VLHU
zv1Pz*;P_y`V9&!@5AO~Mho1hF|I>%z(nrik)gwkDjgOrl9~%uCz4Bzvli{bbrxVZ0
zepdf^>vOB;-~HnIOV3#R*zgPai_gEVd8zYq@2jb=I>#f&AH2?aJ@Kaet<Kwh@3`-}
z-^aZlnJD}~{!sJL`eVnZ&`<Y1=Y0PBFZYD=`z=Q}&j0`b32;bRa{vGf6951U69E94
zoEQKA881mhK~#90?VWd=6y^QL-_Ok6?G^5TtMpz@fp`jvkSOYpK!}1|P=XYd5D*iH
zVviBU7LCM+iCsWIr6@r}x^xf)QItbbL^>STd)qS4?~iA9Z+3R~ZqYNm683Yiy`87b
zGtcaP=KFlx%nB(bje*Oz&;D5q`v2K=TOYfwXV`>jA%qq$!6v1Y5)4Bc)rK_kT!wMb
zFpSS#PUn{WkH5YvZ_{pfqf8=#|9OZXL)czhUg{F7Mg1{Ww~dO?Y%QWS9h=b5gaAMy
zF{FBTONk*RZfW55yt|}<+i;gU4P&{Zs^+DA*WS5DN@+CK$wlzLhaW`Po||9n6Qf5x
z8>`u_jL~h_H65GK(Ss0GkyK(x12<BUR7zZiQAb+OyUQ@Xc2-s2cX-(SYa+yrAQ;Yy
zuorH6Aw5=)S`ee#2E}SN?3#@z&7Y{!EKxo0R;j8atAWcbX=Pp6FpMvqC1vAsM?Lmc
zgt!ry5jJ5nH@`eN))w`0tR973(}~t}bfF;t2}miCfPvwYs{TZkKC*fV>NaHEwzscG
zDoZJGJDoENhyCS2z%3g~*^c198etK(@AlW1#%fU`VssnPnubki2$WPcd_)bVqe>qc
zorXb`+d;MA#A!&PH4R<Ue5!$0$MufmgWK3sv+46+R{Z1L(g<lIsF$#pZF?;?MvvMN
zquDO8*JDr_Qf)h>n{t<f8fOhImjjos;c`n{?i$*pwqa=73+dJ*nRq>#*AITo*5e1U
z+oN$B1_sh6r2`;kp6ke&D@y)2>rg~RIKL5+uvhGu+ayM_eP-A7ZdQXzO0z_jAOs~f
z71X$DXwfX4q3wH+*}N5%E(evk>2FV`^5{vnZ{N=0!-pv<Dq{5L(L7>1%x48B!9!G0
z-%O>fbREm>S91Gv5xw?2hAmV>g3Z1;TGP7eLIccts>z@NN@^=`RF^RHqO18!%geAC
z2HDx!>|eKrf`S5cUFQ#5x3YBUQor#tGc)<sh|5@aa2Kh~Tce>af{+j*-qt>SQ`09W
zTq1-x6(Q~UiJ*kNe)poqF}mJIAS?!za!*MFrH*Q<D~p+W>2+i!cjMrvpEKdUajaaq
zl9G~=dJmeKno3$)8n-6)pvR35a>Lg7ILa$&8ruYyG(t;SAyT4#)_FPbb0LHyLfZ2W
z0V=`pPZr&&>)LZrr>7b|Jylupg=GcACpKg0<u_At<QQYej%DZ0oo9XU`1p9HPoK`T
z`|stYg8fX}vY7OQB-{<~P*yu$EBwv$setpRG{jDxJV}QR9dNt<s{(N|fOe+xzkt2x
zW-5MY4LoitR|59{wP!PkA7Mv)x+DWQ5X$lIl9Iymd~VLTf@$q9W7e!$Oq(|CN1sVr
zS{iHCtl@Y@3R!DjBsDGtDeC8;kykvT=$5Bf@DsYc{!0m44m2^9ZwB5xM{_Zw5Adm_
zE)g}~u%Vd?QW}3Z`%qnmfm0edr5hKLyz)Hm?lp*M?Jr~e`0<TGR)D;`JbL%;P4SYA
zY`y+o@*U-99(5$-#FLvo2}IL?t#%&bkfmI9PG%8^8cVqtIBS*8r^9CLUOY%5t~IFh
ziTeDbKwf1L{aSWoY^z?39zB{xix&OJ`3K^6zxy4pI9}m{TP87d)yt$cY3A>fHAB}j
zk`_(Bz38UL|0RS7TQ%6SWs6@qJw2TZF1X;wPW^mE67Wyz1fK+^0!2U70eysZOXGfT
zDAh|=uj9Ym<-n=wy!f+WJo3mRVUaZ<A%VoiL{6PLbygFaJb5x7cK(oO2HeOKAFm=h
zDk@}OEirB?@G9W;yc)#SS6>}aw{+=JMvNE{P~WCan{cOnz9In_W2t+PpOCN`z;(M8
zUh0za8mFNmQVzpSt>MODxNu5?(#k^K{KZXtv}qHMJ@#18fzr~__~kEu$(S)?c;=aB
zC@n4J*T4RCK)<eCyONoiS?{E7zWHXR#CIn(sW~>Gg;otTt$os)4~_zKA;b?>4Q{yM
z2Ht)5-GKVJbLV0h5l2KlaN#c~q^IsDg9hMqxyCxB8;9qm$m+&n7@V#yp;hx%^tGo3
z)mak~5?Hio5xKd!tY5#Lx88b-d+)s$Ap~pJt_5KD@ZrAx&p!Jsd-v|8d-v`kCzX?v
z!}RIX`R66WI8#x8O=uzeYKie9fhc~YE<1AMNR}*F;wvv$uz(2@CJ>1R*9&Kub&Uo^
zSd3u~4@qkc7Y@Tkt!KN|QO%qEM>B8UJofF|=eP6V!GkF)D`V8CQ9%bEKYlz%jvQg?
z)Tw~-zj~$aH{X22Ew|hfa$>V)%_6f{qff^kRJt4iQdW3gT6EMkiT90a1L!~K6EI=a
zs8Ot4yOxzJSJpeXvqU0r5zrfdchlLRt!Mw<Kqu4%+L%ZJdZ8@&Zh$?^{eoHep{)L3
z0>vtcx_IVr0T2^x5bdgu-stO)Mu$47>ZaBsWpk=axuRtTov`!#^UnwDJZ{`Lwr<_Z
z9e3Qp-FM#|V9-4J=%aM#(1G~)cmM_s8icOb@oL_@d2HLZjpXFyfU(_fH}mJu=U<uE
zQBhOwcecXwUkCvLJ*sD?0fes>Oh>s_ystl*8;1-T!qA~Zt^3?-+2=k}+dY=H$)5UO
z0Pg{(fG>bOz%k$(l&>^ol+lA^;BUaUz;Vz1dw}mfW3C1!g)vSj=)j%8PT(}K&olpT
zJ>~a-tRVaSiMn(B1b@;APo4Sm@ESbTmT{s{<QBkFz+vD!&puy!=I{>C*SwRse9LRo
z+(Q1tWiql{hJg!7enlZm2i-=(7su$|zrX+4>N-LQTrPi=Wx|9B%$_|v;4HJUvY0q=
zVo<#nVBo-kY}~lfudZ+3zI^b}7SdKdL7KfOhLi{dW?Bjava;q4$Crm5D|u}G2`Od0
z#+MLcIdHwHYBn&jzI}kVfzhV&B2)P-{0)x>fUkfL8W{5sFc3vzE3QNlrdb1H0Gt5s
zr;ZqVflO2TaL-s~Tm{Sn+J`azI$$hN;3>Zj{638RyeI@xfQ~3qdhxku++{#U17l1B
zW&vmp!<|`cGO}t7k1lJtaY%z+u`O7=dUe1EZnxV<){Kk{3JMCCJ9jQ~=FDO7;>7{|
z-+S-9kYoi|v0?@3>FIuTyLRouai)+<+hyQ%8-8RJzCRH5s2))rTJ!|8Fr;jurCi)V
ze+aaCi|T2cNW)|Wwfucb;L_I-*h+Y01*I;G4YM!6NMJMJkriMt>H=)*DO>%i4Z-{Q
zSihH4lr(OLtN_mezXi~0rK`Kcqz*U@H%`OAE#0^T#A2{x$BueOn>lkP$;rt+DZF{}
z=7wkR+;jBr-=C>dr}FK$-x3!W#}iLH5zsF?JDUrWS`npdo>Mjv6%Yb6o7P^7wRb};
zbLTpc(AiSXMJ?gMkc#qSnqk$wih#R-UjsJ)j{@5)ea445w@ZK}A=<tGK0>J+R-^&D
zJlki&JeC)stk$+D{oade;3O(!8w?A9CYr-=w=xse>A54oX&7`)X-(1L<78)N*E^$a
z+qN-!^k_0N)Wy=LPaj{~v17+5FZb8Ov$C?d>#n=__~Vax;e{8-$jIQXyY6D};K3v&
zCi?X~a^whIVw)oY%|ukdBW(o;EmcI>yoT4gCaf2)q6pF&XajU5ker|-a&ymiEHKE@
zeicf>tO5Rw(tig4JyFDc703&7PQfJqzfjWlVwB$91-Km8V(Hh?Gmj@deYyZYC(y0+
zb5na4l-lDxw_x}ZxDjZDB5^C=CIVMJ4Va`kq)hZK|4!5GDphVLSEhI3(4j*V7W%KE
znx+wDa_G;TInx(R7&ves6DKN_aK?-obm-85K7IQ5)Q458R<V5fa^8RceWa8;^w2|q
zlDK{Q_VqK$0P^$m=^EFZN~gn5stTX91p%6-Ysmy!o#z4uN)k>+>66|s^bl|d@Yf*w
zcuAT<poN|r;y7QSh&wq%yBFRI_$Cw?%=nY%b=*?F3%D9sVX40p<x%+F^V%Kcc)LyQ
z-=lPF2>#u?kJ7i70wtchGGMW%-kKOQRC7wV?!B+dkcRJ;7l+}dTay%Wa&iKWVi*Q5
zyzl~5RaLBBz1o)q^Y+_svwr=0U%x$j_5g73;6V-?H~=Vj7QnD!!}#i}ulVGXPkjAz
za&jmx_P?;4PA5o-E`)`sCbdBQkMb4IEoP%6VJQ5Cz-6_k=e+=MSlT{8>2L8ra6aA(
z=2lDnT9kCH57z_vmbyu%?Q{YWtPTQIgGz#XA-p&a{KHb$$>x%V40daZCIm_LIC68<
z1+HnDU({#!?%hn9G>IWYhA?Ev5MTZ3)vL+Q&gTC6@8``o-(<>^DRk}Hm0rDi0dVcL
z*K*H2_t3d>XY%s$sHmu@XU-}G@kmuM&F!%`jk;V}Vf`-}Nbma3A(TWkV<Y~a1+Pj_
zjas(NapI0BiBpQwF}DHh2|s1ftB++`>YogA|5t&BOy$0zI-Vu0ZmwWh1^n4mPS6Y~
z9l^foD4`QAG>VIhNlQzkU%!5S?Q`bLVfXI96`{7amIoeqfU2r0a&vRZ%F1HKj2YzS
z=CWnW7BVt2c=_d*IehqVgJcDjoKobj^-F>gCb_B!fi48PJZtOqr$Ss9eXO1e^H=ak
zg2&>Uq71L6Q7)m~z_>7u<t0X{sm_5?DZ+s@B`}l(JzMBB{8<GML_^9FYo!3}n$AOe
zR`Kw|4|DwZaX+2ai(7BKH6V1Em6gS9x7|j3d^|mS_GHSGDJ)*Rn2sGg^7PYB11iDg
z%$YO1^wLWV9z3|-8F%W`iDT8JzIh1~QGH}Z)aq`-x^#ttRb@CA$*$I@ocTk{MMZi7
zeS`kE6j*@rlh&sOdB-;QaS?t9)v8A}c9g9?Sc4)`LrSZ{4S+5*PFIxD^V2!l?ZG2;
z2M!!yz<>b(^{>403hUObW5b3GOrAWMF=NIsY}hbgZUHFCdHe0Rvv%!Tq?BBE;e`R?
zL`O%{qel<EC_L%YS~X#kszO5-8aBikt;TsAW$-kl8d$&Fa6-C`odMI3o}Gs>l6s*c
zb888d%3cft=F~Hmmtf{4I1c5i3J2DdwTAL*8}qXY_?$H0K(L+?Nm^V2hjWjThcjsS
z6kz-I?X+*-o<oQH^M(fv8pKssUB!$UGjO}zL4yfi%%4A>ufP5}puTtS-c%$+bG$GQ
zq(`cv)e%()Y(k*9<YA+z;yc14tW|aB6SS8<s>0~=Q(!;v1TX|eoNIu;6KI%?M}=~O
z!F%_?VN3mQ!<>UPC~?rVtq~ZSwa%?kp-GqVS+E8PxC{fdC~{;qZQHgD)(wE;$B)ys
zYgZ;uo=j0u_^FTu1qBQqJeXT=y)|h2(4j;5mP$0I8%?03s-~o?BCF@29<`2rB_C|x
zk6PXvBCK^qxf%!*?{$bnxv9=IYETiXdr@8mt9}>i!t+Mb)>`VP)384JTa+KzjQ332
z6$Gjl*1$=lLc{G28paZMZoAf___n6lFCDOJ*RH;V7BgOb^;MFSllj%Je#PwBv)Q+A
zALZrcK0VjzbaM3QQI;)R#)uIkNKH-kbAv%$jsn1S*Imclqo2{-9_Nv$8a6Z^QFYH#
zXpH=lkBdLtT}gQNcWnt$w}IfK?T<<d>C(WMjfp#fcd6fv^&Np$b|^K&TW<|o#G;aE
zGDGY)hCn^~7|KK8MKN`LRxe&^U>-jTw!+^${9Vebr#?l~F10pFph<KL%T9j9y_c%y
zgoFeR9z4k7k3Y`C4?pbd+p1M7bY17fi4&|_w~lq|)&XF*+euAL#b&cnQc}|3E#Mbl
zd@;JNQ(0L_diTz(TeFlFvB|jI209W=)6_bo2{a9KN&czl7Jq~?B*M|3K1MBT-Ar{Y
zQC9Q6foxPJQ9qRQ{ZC^C$|G<!N;h4O@(Os3pLF0E{H=a~perSHE6#CLR+QOI)CT1q
zTZD3#m7^ka*P;?|tgzl)0Ll$_r>Wc<_4&SrG9V=?L8ws@yKPcRnY!|+WoQHt79duS
zV#zn3@WyqI(56ir3JO%x&bV>oe1wgTjwU-ho5aLK5)u-qsi{*FYinycdGch~6T0S_
zYq;aiX&gNCJ@a;MVuY(5Eu0C&+3lP-ag3U1L9ynh+~&fq3o4v7TPhA5apHN_N0^71
zL~uzv0)HRqe+u4t{~bkEP&(;-o>wt-s!R_8t>+~KI{1U(K@_2^;oqA)uljI5$`5SC
zQ}`1WR2X`uXS)VP*hX<XY5+!A)k1<?0Roh<BQ+MB&SuS;HLP8`mN#<t5!Jp0g9fSd
z-nnxp%WMU#cjxoK0}lio<e`Ti;-CKv7Dlf_Qc5bnxOyO$^dG=SpY5Zp#>M5at(ewx
z0Fy7ilD4^(^sJ5Lf;`E^N1R-iT}i8Niq2HMwc#~2ON_I)-Fqwk&hYxM6nN26r-#^v
z@WR8^eZ1vuz>y&Ro1>DzFCvhv`v7ycxwYN86c45@6){{)LrL{u$V2&uL!rd;Zsk`v
zdA!!S+k})7A%sZz;IFS^(<cPmD@xNTt*qdRZaqnElE}hc8_4RE$=f}K<8V0m(8%Yv
zAHK%5J^M2&=4W*7T&H(0S+azYBS!+zy?b{K9#lqVY-}vUhYu&GpoGlcm(lB@Oqw=r
zLQ>PFq$D*XC8;?vG124~6ti;GO3Ev%sj8}?rq)4`tGepw(Zln0EL*-1C0X);aw%nl
z%Z?@ca2zV)6Uw6W=E^>fKf&Lk&S^g%b??7*pZZ|UuU$YL=>#h0;1)vbmtMS#3N)B6
z)KmE942&XphSnIA)a+Hy_^W_N2(50IaXBiK+Am0(K&c;AeK&z1!2uKcp%TpN2J#A}
z<aZwFrX|LI8){DhR2werLMK+Y;dU9sCN!baX5c6;BOxl5)0O!gy?#1VesdGcmoI1C
zx^-MLXb??S{Dr;49^{St9%bfCwNlJkun;FU>@jwdn>QmRsTt`hNu(wxQCV3<USSb=
z1;uRIw2`W+YN~2$D0Ea+e}CwkMPGinX(eiz%tyuF%YfQ&9$ge9T~+|sp_a;Yl+}9x
z<&gKff_mWZw>*h5d_XOKW?$tA0-p_FMK_eL7;HQ<dZI$18J@Z6sPo<LnO`s;dLI-K
zyvU|5`8pW-d5+Nmm3UKz(m}U{m{UXW-pZ4P(!0L^KJ<)jz6-#KvM1^b+$4C$^1|gK
ztPnyZuXu8%h_c-sYHx4kK1x%oo_flbKn)?Obk#DVT`ykjJP^0rO=Ut1U03}LXL$kl
z^uLjtYf_nd=iLm+x{<_$M3R%5(K0QC7O5$;NK58Sej&N}h2#|!QB+#SuAMums;;5f
zSyl7>;qR7aZ``~RwZs&n2wR3qqHqv-=<#_0zeg3tTzOG+!k<SUK!{F$ojm^oZcU+1
z$<PGonoezP4bB*yXnP!`d1pw7Ng&P^O@2i&&&Tw|nO{L_auO}mQ)%5ImDVlN$jL3>
zbY32L1;rGXlu}k+!QRh4rN~uXd-TYmW&1X4dLKp9BG0QFf1*b2#Xe8qM_3_*NLupn
z04*_oGnj~4SMWV~+dqlAu0??FF+)o5N}>$NI@yF?i7jZEo<h6UEotAnC50uWoX*W7
zH@}FY;!;Y>%c-oYW_R|U+M`Dgy|-_}rsYJ48i5G`PcIZgXerAcyIaKCpAiT@JIzd1
zGg0gQ`%7K{#5yE5=cUj&IgR$MThXyyD>R{TA}5!eyaEb~N+>BSqq4G!oQk5_9osg(
z|K<7(@1d6bB7!gd5u%=_2>JwWA%rb;#gosA=%_zH!&0@b%_C)nKuC$*DVbEzf-Wsu
z(y?6|I<;%V@l!dR%*iD;zmSsBGRiBf$SEsy?Em`954J8^{BP=rT114X5m*q4umZ59
zt$6As5pDZ@_(WA(8h~4Jdr?bzwrES24sEg9qd9gmhn(De@{5Wouc#oWtjM`<|K1O_
zE?T^l2vH*l10jFoz)RZn_n!Ex7H7Z5QyY}1pfagc`kz1;l946JT+^~Eo!hsetfG>W
zr*k=zUqDfDDW^*donP+%eASkPi{2qZ)CkT3_5O;CkF@X17^^jHGS{r9s&}>gYu<Mu
z&?H=38OOw=UbJbMPJUq#r*jL)%P-(qVV?8LulBCpxZur2)VZg;M6E*k?IJ{tAS4?6
zoh}b)_2j2->tyfMd5N&;nE^zlj;ygx$zScgY2GZ6lF~BH<Q4LL?n&pTU+h`G{*8HW
z<4;tx)$fWBHG+l^)(^{hNUH(StrkBr*48|6rlx64eMB{>0%IL*xFjx}%Bm_3ojK*&
zx##0`D_(xpOVnUXe}t$JgpITMtyv+2hDut9?(ojz6Qi0X+^y?+Yp--|R~yZE*^&IJ
za@UT}KlxzcQ_n0!5i>&62pR=v{cGMPX=?+bI=}ws{xL06Mz_%uM)av_BAmHJo1Xc@
zlxIDCs!&E%DG{PZ&<OZJKR*wC#MLjDH$AgULW_Hvmbv6x<L~hPv`sZC6DLB{2pSDP
zn6Q3>2;sfQJQgK+T`1`qA!-DTipC<WK&8HU9~b1tbIVF3gDQf?M57QEe5CcdYGs6|
c5&RVRe<TwsN$H7SY5)KL07*qoM6N<$f<^SuO8@`>

literal 0
HcmV?d00001

diff --git a/studio/frontend/public/sidebar-logo-white.png b/studio/frontend/public/sidebar-logo-white.png
new file mode 100644
index 0000000000000000000000000000000000000000..f76b2ea396b8fde62827fb2920b3566e966eb28f
GIT binary patch
literal 9057
zcmV-nBc9xeP)<h;3K|Lk000e1NJLTq00682001=z1^@s6+*klp000U}X+uL$b5ch_
zAW20-HZeIiHZ3wPF#rHaiJen-Sd;e_KHv9c4^~3h@UfR{fdC>StO&>uS)ve<0AYj>
z5<ta;hzeE|6%m|BRm6hT3P_!Z;-DzFXF-aHT8mX&IPM?Zr?q{0@9pjT*ZaKZoag+`
zdCw1k5fUbm=Avo<IZ}l@I@lkNPe{bcy?_E0NZ<q@5a!FmX9dLo0Pq8Ocz$%SKl}l&
zZvX-SaMdm(635~HgSt>AR{$W90N^4L=L-RlQUJ&<RLB$n$QS^yQ?q0W0F(d#Y<YY_
zA^@rl0POThJ_P{Ssgv9n0NA1&sR#g#0RWjOM<fD3y8vKSfl@32z`_8atUw|z1i-EV
zfLXROM*@Jo1ps!ASdb4uU;u!bLM+SxAUFa5Lmm^&10YNTpfJ+E;Hh75g}6uo0Km(Y
z&6i8kGZeU$&>DC0@ZjPh;=*jPLSYvv5M~MFBAl0-BNIsH15C~g000{K(ZT*WKal6<
z<l*LK=WOrv#iIUe@NXhMA@Se-yo>?_01!^k@7iDG<<<jm#{sZ)ziX*m0NPdnXy5r=
zGusKES_;s5`HOit{6HR_p-{*?9UKY^3+*Lhp}jEUi$4D|;cwuJe)jx89{$^U@Bnd|
zK$)$;ql5i%VQzM=QjX`#1VS-x_va)2$Ay3Wjw#XNG_hPP6^ij#iMUWAO~-k;QjtU<
z$(7;~>3=fuAC~<IlXndOtcFj(PPPZN%WS}Y?g0Xe4p`E506>28EsPoqkpK{9G%|Vj
z005J}`Hw&=0RYXHq~ibpyyzHQsFW8>#s~laM4*8xut5h5!4#~(4xGUqyucR%VFpA%
z3?#rj5JCpzfE)^;7?wd9RKPme1hudO8lVxH;SjXJF*pt9;1XPc>u?taU>Kgl7`%oF
z1VP9M6Ja4bh!J9r*dopd7nzO(B4J20l7OTj>4+3jBE`sZqynizYLQ(?Bl0bB6giDt
zK>Co|$RIL`{EECsF_eL_Q3KQhbwIhO9~z3rpmWi5G!I>XmZEFX8nhlgfVQHi(M#xc
zbO3#dj$?q)F%D*o*<oCahecqCSQ;kB7GvdDHCB%`V;xvGb`=}Io?-6@Bm#?o6YL0H
zgb>1Pf{>6$SWH+$s3q(pv=X`qR|$iJF~TPzlc-O$C3+J1#CT#lv5;6stS0Uu9wDA3
zUMCI{Uz12A4#|?_P6{C<kTOXNNoz>kNG+sOq(0IRX`D<Z>yT~9-sA|ffUF>wk++Z!
zkWZ5P$;0Hg6gtI-;!FvmBvPc55=u2?Kjj3apE5$3psG>Lsh-pbs)#zDT1jo7c2F-<
zhp7`Zb($s3n-)XMq%EV>(3)vyY4>O^>2$gY-Gd%Qm(Z8eYv>2*=jns=cMJ`N4THx>
zVkjAF8G9M07`GWOnM|ey)0dgZR4~^v8<}UA514ONSSt1^d=-((5|uiYR+WC0=c-gy
zb5%dpd8!Lkt5pxHURHgkMpd&=<EaVM7OT~%9aX!jHm=T6cT$f~&rx5azF+-<`eO~U
zhJ{9;hDf7SW4Fc`jUg7sGG*~tLe_Ft1M4hnm`!3^via;xb_M$zb}xHOQ$usAW~^qB
zW}W77%>fR^vEcAI*_=wwA<hlXdo2SkKP|D=YOQ9ieyum!dfL9)V(oJ67VYcW?{#pU
zK%Fd|^*U`j_jL)nHo8%|g}U2y&+0zYW9xb83H8eL4(Z*|NA+#=qxBc+@7C|pA2%>G
z2sV%zY%w@v@XU~7=xdm1xY6*0;iwVIXu6TaXrs|dqbIl~?uTdNHFy_3W~^@<VyraY
zW!!5#VPa`A+oZ&##pJ#z&6I1JX1dX|({#+t$SmBf*sRIyjyctwYo1}g*}U8QjfJH}
zoW)9uHjBrW+LnCF1(r>g_pF#!K2~{F^;XxcN!DEJEbDF7S8PxlSDOr*I-AS3sI8l=
z#CDr)-xT5$k15hA^;2%zG3@;83hbKf2JJcaVfH2VZT8O{%p4LO);n}Nd~$Sk%yw*W
zyz8XlG{dRHsl(}4XB%gsbDi@w7p6;)%MzD%mlsoQr;4X;pL)xc<r?m~(zVm|qno>%
z+^yMd)ZNTI#eJ*$O)i@o$z8)e??LqN_gLa_%;TM>o2SC_kmoO6c3xRt`@J4dvz#WL
z)-Y|z+r(Soy~}%GIzByR`p)SCKE^%*pL(B%zNWq+-#xw~e%5}Oeh2)X`#bu}{g3#+
z;d$~F@lFL`0l@*~0lk45fwKc^10MvL1f>Tx1&sx}1}_Xg6+#RN4Ot&@lW)Km@*DYM
zGu&q^n$Z=?2%QyL8~QNJCQKgI5srq>2;UHXZ>IT7>CCnWh~P(Th`1kV8JQRPeH1Aw
zGO8}>QM6NZadh`A)~w`N`)9q5@sFvDxjWlxwsLl7tZHmhY-8-3xPZ8-xPf?w_(k!T
z5_A(J3GIpG#Ms0=iQ{tu=WLoYoaCBRmULsT<=mpV7v|~C%bs^USv6UZd^m-e5|^?+
z<%1wXP%juy<)>~<9TW0|n}ttBzM_qyQL(qUN<5P0omQ3hINdvaL;7fjPeygdGYL;p
zD|wL_lDQ-EO;$wK-mK5raoH_7l$?~Dqf!lNmb5F^Ft;eTPi8AClMUo~=55LwlZVRp
z<RkfW^V=0{g-p?-v{J564i-!=s4I9|7+ZL_NWDl})HB~^{_6Q73xXDWy^yd_xUjR>
zxOiFd;3B_8yA~shQx|tGF<nx+WO!-t(tXQl%d(dJP~uopRq|?i-13f6gVLp?Lo0$;
zG?l58$;)o8^jf)N6=9WRRqtw-)mv77Di@VstZ=NTuK2J<xaQ(lPG8l0^?7aD+TKdH
z%I)h&>!j;$toK>JuS&gYLDkTP@C~gS@r~shUu{a>bfJ1`^^VQ7&C1OKHDNXFTg<ks
z+w!3{qxO28e_hK~{jICFzS$<)*1z3%d-D#19Thv?@09GkwJUg6YrRE%^={Jcg58fA
zVjH@@=6>C{M|V%fo{xK_dk6MK@9S!GZ*1JJzrV5xZBjOk9!NTH<(q(S+MDf~ceQX@
zDh|Ry<-sT4rhI$jQ0Sq~!`#Eo-<f~6^$6=oc`MR7zjdrFqivu)p}p_ujH8_$UL8li
zxBq_cG2>%($2E^vo}is5J@NVEf|KK?WT&2;PCq@=ncR8zO#GQ^T~S@VXG71PKNocF
zOt)Y6$@AXlk6rM*aP%VgV%sIRORYVwJx6|U{ozQjTW{-S_si{9Jg#)~P3t?+@6&(!
zYQWWV*Z9{iU7vZq@5byKw{9lg9JnRA_4s!7?H6|n?o8ZWdXIRo{Jz@#>IeD{>VLHU
zv1Pz*;P_y`V9&!@5AO~Mho1hF|I>%z(nrik)gwkDjgOrl9~%uCz4Bzvli{bbrxVZ0
zepdf^>vOB;-~HnIOV3#R*zgPai_gEVd8zYq@2jb=I>#f&AH2?aJ@Kaet<Kwh@3`-}
z-^aZlnJD}~{!sJL`eVnZ&`<Y1=Y0PBFZYD=`z=Q}&j0`b32;bRa{vGf6951U69E94
zoEQKA7`{nFK~#90?VWjiRMpwXzvtdbCJPCWoqdN?phiJaT19;+)T&6ztF<gDwIH@o
z#l2P(w`!$UY+b5?f(whKh#*Ul9YjGCB`Au>E+P9mlVp~A-apQr$;_RZWYAFF_jP{x
zOy-{To;&lq&v~BbJm-#(Qqmf@VC$S8C!qhAUAGOk>w2zDh#VobWC=DYrIcV8(x^A2
zQQ|R-{f1$D;&Hn-?S16s9VHufdRt`>G5o)W_&$Ukxas-ciCWz66Ls5l37Rb@UemD&
z4NV9DBoafacdwKfQsR{cUh~}}4ZMc8+HDxiT+W8)_grz?E-9taS{E0?|37>u!VcK@
z+~5Q~?x{r0c5#Ak!>;MrgpMADsEVW#LmGIIilkEFF^neCn(rRN_`>b1zx&{*dsfGY
z8$&pp5@FBU@N9OX9`{;;ZX1!P*|2Li;<R9*%0P)~zFVcLlB@=vKuIf`(uQGt=B}!l
zP<-9PpT~$BLm;9iY~IEfrX<?pUP#pAuxmQ;nvO0s1Rwz^B@!?&{8BZTsM1eXA3?o_
zY&!PUHKnqY60h4mt9;a-?gPBCwUq4`{!b$+!VcN`@{&X?ZcKu1BVN<62@Qdgs)nDa
z;dE5#C!^aiaC%+T8*bc&Bwo|dHO;RY_;g%fKYn<PT@4#P`B~lH{#6|#Z4Av4_WUg`
zB_`-`+Y&U}x%OrZDnqJcxAao$anayzz~gb@(KS3?iO1VO=giJr)#WVuwn-;hkLTt6
zAF%o8e(d&m+=hXH^h@awNLk`Jbn>FA-_Je}QxSgHh)CFrw!P9OL9=~q*Y&<ugG$Ok
zi7G(|sv7EO@HCLqE}N^m^(U`=M;sm(4!jJtXX7}0oUL28a`50mDk>@%J9aD&+79w@
z*>Nz5Dw-Rqln&34;tQ*8c{--o{(xbNRFGn`Z;aQpzPiwWKs`0UpaQBI>u}Xqan;$E
z@}~~xV>1j23k%s>u$r>6GIU+%cbhk}WXY1C{`2zk`1zRgDcHY*%=VqoP&YwD2$5{-
zmc60vV>g{Eg!nc_+8-vu687qy3l}Bm`e1>u7*xtVB@tA+>Zx~BGVQ!8$xrXg{*ON4
zrn@Kb?z`_&RaMpONi#Du$;!&&#?=1wzve#vd-Hr;wGP@Qw!tHf$dXowjJO~7S_b??
z2;qv6_J@ZMm0<LT3vbhP?P+MzQw_hKs;v0(nlh48+p*+=>!~|*gz@9Yvwi#aQ$Bfe
zaxyb!%wYOGckz7LUZ!taM0QFV-WGT$>s>FE|8mAOz#TKhLNF8P8j$`6u=~smB?z5>
zM*`9g;7*|NR5}SF?3j-h=K}j8IsQFTQeIoi^|==@z1#WBo;{oC)2Dy$E6K{rV)g3P
z9L>!jfAw=@CS@Q+^E@<4D#ujZ@WgUrE-x!;fi?l@>w(wL&{zU782HGNmx>y1)X)MG
zQW}2^^r3nT1GhAAOD`TICAB5oF>nOayPeO(i4$9etN<k?B@7xgh|0z5*?jd~l)7rs
zOm!sUz|-442E@~Xt@Z=N0ZY2(jEo`_4VH8zaLOv3pRoBm7mY9wS9oc1?z^QIx8cPj
zB}H}RoZq1@<2w#y?AWo)n>X)!jz1K?`OR<GH}~JH7&(~|N0r|*5f-|Zo3?PqEr3mk
z<|@W;8j%LP0W2_IZ<=|h4SvFUrE!l(dT|@3WHr3F3<H<prP5W8Thn>&$D?@g!3SBm
zaAAubr=+CNsZ*y@9?+C2Q%KnMB~K2!h6b0b`N5|qO+!gpt=Wn(oGzpQ<A7hAud!y{
zX@drE<&FjCd8E9|ZKxF~m*J(+@ZvH&xTQh0qny`&`fJ|buz`mkemHDLSy@^9>}Nk?
z+_-T(`Q(#SS6B0kU;HAZUY|aF$ji%Xc2L(}e?3!^`;nQ}9-Gi2s|K3ZE$#LDt^@R#
zlszNx0D+;Q(#$(;&;Z;X&v>`=;xb=~tV+V3sIQ`9`;H8;XNJ{TQ&Lh`xNsrG#l@^$
zyOuZJc!Rs{x(gu$Yu2m*VD#wG{`yZn^%T2z@1|eBeh~*%R8+)_88i6%xuZB)SB6bQ
z7NA;c^0h!5F<tgdz#U~?t3eSKW0cDzX`|u6Wq4>bj~iX}ygpRvtb6wC32HoY<Vb32
zYPjyY>%w+EapFV{9XiCcY107ZfAvY-FTebf8*aEE;=pFlo=sk*#^4_Pad=!IQdXER
zEk5qD)Vr_i4CvqK6FB`y1<nQr5$tX{6?8G%4+46kZqU|58ZZ!L$@c~9QML<b;YYIi
z!wD3tB<gLB;YUD1xK6Z7-hXw7OB&sqq^g%jQ_2?AS94K^TzX>XnP;8}X*^-V1U7Hp
z%&oWH${lyy!KO`{g32Fy=pnjy?@n@ZG5{k+j6m0$cr|a{Jhp7vLV9|7NZ($sm-+MO
z^N+kMscWbWx>{lW7eatQj~mc)7}2W*Gf?go-|G)iw7JXD=I(&9J1u2X%>17MZv)=~
zp8~spBfys^UuncuMic44Ux2TGqh|YEz&B=}tANQ-^b-j>a2v24I05W2$N!a?eh0`8
z)9w#6UF(MglTMg<fuD!Z;HkIt6OSUd03HVp0$-bLzA(q|7BD2Rk+@*fOVhnV{>>8*
zUF9(hJV;9G%2_hvW>P*q!qA~ZgRfTC5klbc1gk7J-E<Rk=FADX%KZF%CQX_YR<8vZ
zK72Uq*RK!C8!}`FE8gEk*2+i8vbV*M5`iF)mI8rvG`#Bi?7+iS56?d)rEF=t%`)KX
zfb<+-QVVU~L>Ux5EDT7$figUNxEJ^wSkXeC2Y}%y5?gUGiZJb3=mX#wa1Twy*bU?b
zl#e$13dAMAJfK??{TBe^fig4wGVt3d+WAlpWB@%-r1aqvv)`paZVP=(2WA6kF2kGG
z7+_>I8m2C5cyURCfr&Y+TD2;q2d~%bCu?qQE@fq9yz<H`%$+-zMT-`N)PMW!w<D4j
zVEOXpWM^jw<?Yz91J}uN&g+_s+iL`oRrvov*yH-gb<Y_9Xi-Sna!b0hh57&$&7`em
zlW3W&pr*f%2%Y+R0Go-9tf16|@ln<VxE9z*bYuk>iMj#1m}#p&wI%pIAM5v$iIT=G
zkrm)+;8y@zqxAH11*ijVL%FuR(u+qxA_m*GZELo-S+i!5o}TWP!W%bkY<UDvKh4mg
zLzy;h8ee_&6-h}+Jo@OPA@vFi3pp#TBXPQBUUC3Y0U<!MY2CC$dtcNvcczW6UY2w*
zY6=gBOyKX9Jge?i0o)G!0{Cy>Az+K8&crCkb}q0uLfNOl`zV#eiY#D<d3-X;bNLX-
z$<qa;-}_Jx97iqLhQk7&jpj1E9RrE#_AMG12E8&mQE~7ng@uL9j%dr4EsPyImfT!*
zvkV?Q*k5+!$PsF5gZ1$I{CsY|{dPY1-~*n0_E~asbGiNY+Zj1>B&n&XL3IxuIz;cp
z_DDbrASz%=TLD7L6md46;dQ17>%)sEf^-5p1APc3Cn$;B-aJkOMp()(K}nd^z`s%Y
z?=WBhinuQVB~gwkoaFxpO4^=-(z|;D7XX_q^*Wg2c+9NR8~6#KZmpjLl=nudJ-%xT
zhtGg(fQ~2<cSQMgtuvnnOx9ddruwFTcffWPr`OHJ**!UM-~i?2!Lz8QX~e}f8LG2p
z&GIiM3?Dw6Nt2XHICJJqx_9r+;K76a>ch&FD_OQ|8SlLF4pK_)zyE$fN!%erh6EX9
z0Hvj+^hs)u!|e)^s=_aAL4c;|S~~wb77z@SB%FfMCw*V&0pM2P&tcm5NSZ;Yg<c%t
zJfEY8J0(K75561t*C;Xs;t%HQs3m^~a4E3dl7Ab@qwtOS+8O42I|ItULFw2K{=0bx
zrEf0*s?59^V3C<`4UD-;b4#!8+gD{s!@uRlWq9e^CWE4)qL8x~hQYJXK8w@oWYwxw
z{ven)-+Ys`YuEbg?b@{qfc^XTvv1!%K)JI3MvWT9=bwMhhaZ0EuU}MDL}g|0jpcT`
zK}vKXEJO`Z3)KHOfAY*?4oVV6!k-A8RtK2xWq@^g??aUSmi!&Z<GW#QwB)ZrN!R9Z
zHBf5Fn;dYQO(??UL8xlrAiNjChoivXEO|X`9%;yMTT?V4NV6wVT&!+zP17)prdWlY
zJ9jd9@?<W*{Bkb8{BnQ(s#U8fEG*=nd+y=&*I#Gq)T#98(}#fr2Lf=#6<2WQop;i!
zS1(FRN~o)=Yi7(U1o4owlJ@pQ+(uKZtg!wUErfUd=MYIE24X$Io&}#uP>-6n#{`u1
zKuMfxl#aOtSWEOFgFbyM&yxRGl<i*x?hi;0iB$0vVRdtb!%E<f0qGRYkkS?In~oDY
z@j|1rvXZQ<EH1q8!l3fGbLX;iXXuR3*x1Ou_uh-s>7=;0nEd>FX3m^Rad9!5Hf<s|
zH<uS)c!7fl54K2Ffb8sSD!h$B8xuu<T-Ag?7Xn?Lvh{kpkQ7B7tEVFHEBHO3o(F)o
zD8uUsluKwQFd>R_`G}Djkmo|F6w$yM5*SH>o+@-&{;Wa>q9J9KHB$g~P3QhyD|z67
z2RM54Xpqk8!;LrIh&2>7KR=(FZ@!u2<YWd67{Ju2Q(3fV5j}eJ;E5-m091g>$&)8}
z{`uz_IdWvPBktL=Cr9e5{o@h=MD>#u(WrY3>(mtqR+ZsQJSNGyDi`=67Nb^rLVbh5
zI1hLY<tJ@U4f6FJSjR*3OQ=>ovb9&)nuB#wL~2NBRk#7rg~o}xY6g5X7rQ<DO5MJF
z`xrKCSV;aq|M^b}3JO@aZXHvmOkv!(af})@${$+*N^;(E%Pp)~vj!<8XPtFcNI&uM
z@$~QCpHIt=`?Xe01V~k(p$iQg;-uE#K8iAUT2c+HUv4x@x~;txrzJhR1Z5-*M6JxN
zAyg{+Fans{OkX~N1x~?9C|6Z9=xj;16kpq#pH;}`qyhWF)s#rGl2SNWe3%m4VV6$<
zwr<@@w{G1ya3DBtc*KYiTyn`J%$zwBuh$!PF~NuV^XK!$7hi<r4;nOxx|Dd1mY0Au
zr7BtzQH8)J1e!-4G%D)8COX1eRfoZ0YX!p@MV-@uy}+Zu<tXA@2K<#!!)zjIDK{K^
zyAKXp@_!ZO7_5sD`vZ<OLYHQnj|u&r(4@!sI9!1QJca>U97VF8E?v5Ws|LW)qetn}
zrw>!6OrfG8oca|WWo2cI966F3Z@e*V`Bhh4#aB4c++H++lB$}Lu8OSYLp>T@d#YA!
z-HV#uTOzD=M!6IS74J<*Lb<8VG#XGVRCl4g3Re9t(v9a^Nn2ydpFzv|=&w+I;6S_`
za9l^IYGGYCX;o-=y<wNJ1m?DDJ&vy$DucoS`}FDK4`>O*i!Z)NdU`rP|M}0EGiMHa
z_Uxgyw$`uby4`LLA3n^|rArwzW(=8`nL%zaXo^t)xbn&?dFAlOw6`aj5>>;7<|nFd
zK8408t@@zyy`2uC+uyY*Ox`-elXfU7D5Q4_eYPfU1O7$xZmh2fwX!3r8NPh$qD3Mq
zm?kemyK#i-(MM1o3Lh$I^0WHzd<*0FUa*z_@`0~2RzCg_ns%PGPy%h@6IgoubM87%
z6{n=6uz&x49(m*u9(dpZf8CB9JEH44$BrGNprC+)f&u{Sb~~Awnb>SLs;a75+yZ{i
zIp?73Iu3_}?0&r{SiOXt#B{t~109K`X=)zQ1eylAr1aaT7rl=%B%;xuK0r-teFO43
zpseP91BIwaq6<;d_dku9D38FUDBW}!$}8YAezJil3AXwH!p@X5%{WI<QBi?zqRuGy
z*g}-MtQNI0cLgc{$BJ=b@>Za3G9Lz^KHry62Bbs<2(?OJw@peZGv9rDDH<V!1xVE6
zSp4OOyn5w>bne`lvN9F4GhxC6KVjqJ<0&jGBsDdal#~=28k*F^#>PgDA3q-TfG)f2
zGH$(XI{OcN!@TVq7~|<mjyr`UyPadlj?fS<sMNgF+B|r5L7lr{Q{BEpZp>$WM0uLY
zgqO5q2=;;gr{Ej+Ur}TQrIX%mzA9-_W%?6pJufEI!5<Fyp$KJN{`GBfX%6?G{J>T`
zPB39XEkn;Tj~h^gZ56kp24K`U7ZBzS5TJ}5sqxy0LRPO{&6+i9c(rH`aoutlF+yGU
z_U+qQYAa){x0HMDy*K0}_uqd%fB$>9W%MSbWn}WxONVpr&|$p)@g8a#JY102k?8}5
zG3A_#=~C=qKw~05Dv_LX$j$kM4my5Wankw5x|h@_F-~E-_hy2f;mu(Q@SG)2k8ljp
zEe~7U_|jW|Lt*N-M+JeOO(<FS0D+P4)^h7qJd&nR#Aq=cCDp^B1mz!&gevph%CBhR
zc&%f%2`ME)2$8Yk&o5)sZwl8cPSdG&)NxVY0i?G{Wx<Yh<oC?u&Hkftxm>(wl=7?h
zUgC-YLz$iMV|w*!(z_QgUd*-EUJF3Ke*M_LUm2N+iHVFJJ({AjD)I)M&%m?uXxp|8
zX>Hq*k=BlkwDu$<#8X;U$-67xrPfi8)9Iw4(M5%)-g)@&!Fk)3E?a<-ETurLl=4)k
zoe8KFpGX#^FIM&uf(iZxO-}pysJ;KzHqF5rU;7$Oq!Xx^gBysfU;6L@YM~+UM*WuH
zn1SmEAE7l4B{c^&)Bj4~K_aUgfw%y*lzL&9GJ#S*tom*WVHO7ha3Lzdys0CfP)dL6
z!M<8*@>h{s3ZUNbU>7=xx(%<#ATgy44x53ivWAqnL{2zLIehgDrvCEREL*mWf`S4r
z8!>`5%m2jgQTOrc-48KqmYON%zP122HtY#@(%ZKqBds0T8EIsur{QooDJid@q^yz+
z8`k4=*5hn!pxouG|K`A#3qN~r!@H<yvJ|!cUIR3K$43_l;7Zg~nT@h~_n{o}K37nG
zg8i1qQHBqw=`YY%d5q9!16a`)r7MQ(FAxJzOQE^u*mTtO?ls33&WApj;GL_5GzDLW
z!-eKKx}ySbYEU}pmIz~N3BIj7Stz~xQ{X+b@4#IEZj?RI+`>(Y*_RI<KVgLsB7ON|
zvqYTjj!3P2EBA4ln)TFEwgjpRiNn*#n63kPuGet9UN4T61p2J}3+~!7?i_jz*EeJ`
z?Y29(JpUR}Q&LG!Ye$Ey4019v$jM6QWNA6YrR9{ARZvk~!;bCSaMm|a>2@}JbMWgW
zh3hxIi<)A}QG~5Q1yQ(&JptwifuKhfC0u-Ve99lk?n8*4L6yw^AzM>uQZh6Fx~9|E
z*nm4hC*Gb!b;(Im5>iOA#Zy{W$ukK<aF^DRnVv?6>`Xf4WYQ@oi=yH(PLz~TQdUW2
zRW&uWb?pB5BPu-gjfW2%Sh{E3hIddztuSA;1QRv(P`Domf(R>w5NV4a7^bBrZ-fA%
zHYNN|-VP3;ZYmMrf6R~)yplKr@{hM+U}_E>vNPz~sRP|Qb)dYeniIt(6qi;|QCUrO
zZ7mLGJv$3`H6A{E;O#x@HY_7X)EEK~^7KL>gqE@N;X6d4{YilcveN>|8c5Wp|G|=1
z0EsTi^(7hfO3$KOr;haK+7V4?94jiOsHBYYiYlsVYH&E56xCHUZrif{ozK>;dmA<7
zR}g;cj}i3;im*@M7DCuEmp}H5h>!anv@BJd%1kLM1VT#eZpq}b9D3(;phwrv^z7Q1
zqu&;Byr`Js(sHV*Yp8WNDXJ-V?fv4j6`L0>`ZrBPtsq9!7%YfHSOM6wmOuWyh`0SV
zdZMZ$4Zth8rJ@4^a=OsFdl&5Xc#a$|qNuo((uzuI>*^?~sc`SvyL-jvg^QLDBWes$
zAmVQv_(+@m_M?B+lI(YyxnYS4Dw0Zt{|SU4xwa~u%R2O-SGUg8)HyhQqL`DVWmHsF
zbE3N3{n_46R&H9b@GW9Qjo}o~?61i9N&D8!@mkw9uLSC;>RnC$n(ti*GzsT8l9-e>
zkj@>lDJ`$yL~$7<rDYr`FL8hN`R+CAUweHaP3|clQJpBiU5uzPL_~|f(`AxYPk-X(
zp7w#g77Lr67eZ8;$eQSu{MkN;_U%%ss;=Q=NjcvXA9sKB>8`bFU!C_R!9)$T`aLnC
z#?TU?`e8Yfv>FiKanXb0ZS7NMX`0s7Pt*WaV7#j{=O$(2a5_0~@>|dLT^|%If8j+R
zQNvCBF`~v0HBRZbW`z(MDrh0T`&*Apiffm0hpy|LeA2aRV>}aO4@#Z2o^79exMIQM
zPcA?aGe*=HS_P;4Yu*9U)&|7&dijq-6FOv!?X0Ja8SHE$+{G0ep8Va^r_4G|lu=bp
zjHofR0>0DF&qEM#^&icfk=HvV=gzh@9{I+^TYW!mQ;&+oi4ir1R>OBDY)~gc`1Y75
zq9m^eC0%1gjiFW1T7(s-&^O=Xg1nepR$>uUF|;OHg|Of!t<O~}V?>SNbl`sgc60KU
TmTz6*00000NkvXXu0mjfdrp<w

literal 0
HcmV?d00001

diff --git a/studio/frontend/public/unsloth-beta-black.png b/studio/frontend/public/unsloth-beta-black.png
new file mode 100644
index 0000000000000000000000000000000000000000..beb3f6e82fe03a763f0600f4fa388abc131a633a
GIT binary patch
literal 160502
zcmZ5{2T+sg_cbes6)Y$%2)m-NsDRW62)GsyrKl(h5l};sme5-gSgflgAfWVOp@=k*
z5+Ec26%c7s1Ec|=hlHL4Liye(>hJ%(Gme8JGf%nq+;h*l&y!1*refP=w+jggiJ6^0
zdqqfS`?8SGCY`OKz`q>0L!1Tv+U9-U)=x-ihlJqIA418g(n6bj-<qBM=W1~36m<(y
zZX|7ec3SIbW2`C{1_gmUpK7%Do5Qy~5Z`v_P~|h48kZP9+;L|4#ZBJ?U$eg>zQyi1
z^w*)?n+9wAx8nQ!J-v%2JJ|E#EuJY=c5c@@E-Y6|*yi>1xw>5M_^PR}>Xcu$>5@Tf
z)%2u(qwbe?G&=~~X%eHcsOjqnKL0ME>*dr)ajepXY}+d(8ZPLdG|#w9lj*udC0ize
zLo4qi4;LvY_C3$M3bSu8+~VPUX!`JLuIo-F8;ZhZA6~qXYkCEnsn!^HBTv_i6|3)D
z!0rF)f1WlH*cfP+7nnlLIoHE}&-TbY(-YX(FYn=%P*x7HG3<HU<f3V>3&p)?^fwS&
zObgdvMkO5~B7o!i@%@|ZY(5n}BrGH(eMS4}Dd(LTXHT(C%%?k|<66wM>(L25$ry$<
z$gLVQJL;hH<?cPFWVF;mXf3P+*aAXNYB{v1;k*kP9n>d=oG0nTUkHp#-}`kZ5>c79
zsPGBnZ7{D6D`lRmh7B<BHVORx8zXN9Y>t*yk;XV4B|YUg^3r1Y{pEi{gL=(fa$0(O
z%&kHZr1%5m>khm&_8MN35We{Sq{gP7%e^IiaAC8MP}zU`w!Gc>O8xkMtrgrOV1n0s
zs$0+`Y&Ee_f@v)luxN=|t%$-vvcD!KX}VO!DB<El{fctVyL><gg(QC<Q*MEc1F*wE
z%=We!!e&AvcB_L&(=Mb0Cla@>!UA^bJ+?7z?^Jyaq@dfS5f)bl$>$&Ris{yM)`T=C
zE{g@UK*0;8<#48($Mly>EyJ%)p`P`D*#@|(DlFDN2-e%NG;X|NVD;eVs)eLETTy=q
zol<>$S>(u>@f|zwX|efrPR>*2jm^(AwB#C+2D`FcWVv<yGP-f;ceMiZ^TD}bgCx>E
zrFq>^hn!{K=GzB&qj9uvs+<1s9ymf~So~35Ze2f|vnl98xqZEVa#fYCUaN~Oj%J5z
zbeR&h^*XLna`^+CQyQ30WF&H@j|C0EInPdp{5r^&<-3<R3H@t#)<yX0qi<4gW3g23
z4)C`F>K+Y8yc+bNlPT5#0TLaov`+IQI-M1rp;;3qu$esD`b1kzUj{vt%jsT;$EAM3
zu1CNFqbpD!W8f1tSf0)CevGa11V=Yie$BHtdBDV@`-lhlJn)-hHf-ktt$+aL6(|Te
zy&z9ZY*49F6~gXCx#~n~VUIt=v`JVhtUgVi;vr(bvg7C3{rl?dIPin}9y75^GN&D1
zp3W(#_Dx0G-bG8wLIZW`o54{fHD=wa&$R2ZJK4`2z@r6mUAKUEOzZKWtnUgOFhFMW
z;-3azJV$uNnq_7m4#27Fa_yofbOS6?e6GMmRz9>iImEUzzdH3KFIQbJH(ng_*hS>e
zvyS)yXSOfo9uRLozM6r*egxkGQ9<xs%Er?Cx^E~NoT)c|(S5bSC;Nno>jkB>^uZIY
za*fTJ9p+H&GnyrbF+T9Ib(<<4N+$yDI{-2dP;7bu;+eE&1*7~=dxs6f2U!O`aKt=V
zqDM<E_Y_*`M59VzAA2#nX?^ZmfqAS6+H>|4jY4qw^k_@lGC#_9$nm4&&ov3{JaiiP
z;Zo#zmw&$kSE@ntrC(e8P;My$De}+mJJIUNNIX~XBPm<iiD)VGZQi6l^+NWro{8@v
zh9xV2qBBQ7y@i-i2pdS*N48O)X<=Uu8DNst{hf1F2qTs=7&gfKDWeCy3#mxT^y>_#
zCZN%F1>O3#{HHG*;uqd^UgF-V%HwBo66n2M#XcU0Y<!6UwV#Gx1|?L+-Md&%d_erQ
z)VE3-OLh9bRHK~y>ev6&Pr2ty$}*aDOiD0}3hU4unY`TCq$AKA5HwpRshI8cfqrBQ
z@phC6!3QC>vXAb*h0sp!`wTtwpyZ*2^A^IB7&yI0tZ*8UccKpb-qD|WO27Jy4XB_j
zrC>@)`5LxShFsC!mD3~R=2hPhMnfIQE8*d&)!b-Qo~Z^}!}29!v{Cxk68CSw%dzh_
zY*G-%UJbe5_|~@LhOL2RkEO1qJAsjCh|xeuLnUS2=$W{D$kd>9IlOtkPG-y?X|;(%
z5X%9G?J!8<8^_}*ygBRqf8=!)Lc4TDstUm;0zQ-cWfmEv`60aAFED47bH3dELrQ*k
zp6;83f|lw5ZH#S}`6z~+=4H^;*%=R;_H;umVdBTuClK~oN~@z$O>tVkmb(1lo0O1H
z*FjC`mzoa<(gBv$Ow9&qEwZ%hlT0cGItpSmAFnqD+jOFf`<nV{uqWb3`|eX`kPhqE
zCSjerpv=vL+HIACM*-lwKNh=>w@f>tJcBWx!Sop;?+=Bg_4F6m^dED3W2TpaiR-`X
zO!_(*kdr%@#b&AIDd=-PG50CTtw>HQTlM@}(%+aLOZp)7eMz&HH$v)eC|bl%opsi(
zi$or9y$$o7-m9qP*l`xxbvJXl220F75lz~8`B(JaZaM)I;7Gjj48|V7k;9wl<Ag@0
zC@Fdua+Uu=Vu|CGHuTxij2hct5RkLlPs(q}k7t~OfL|>+r<lRt8%}r`794=sWrX^W
z=fnB%w2=qIv%e$oKLC6MSdzO#fF<Y~%9MbRlFSgzeyWB`jE2i~Xlh9P-E0k)r<rah
zF3B;ep5Gwo&{et!$#Nj&eb&bsFpTKCi9Qg9+TrttAFsa)Vmkn;u%HV&I+l`|tJ|!^
zqik~;#KNd%M4SGDZuj#YihBxu>$`||J10niM$0S9Jvol<j(lTq_L<_MC*kA4d$;^N
zyi<~|)P;oJo*Py@<-8T8aYrt^ZNFH3zuSjoHMgpmQJIE5s&vU?XfsUHB_gvC)||4=
z=~*=bXN*0{;Yo6~W#QJy%UJP1LGVFnBJ@n?Rn?i611~#jNJQA5Jz^M1_Voc!$j%E7
z>dF)5pAd;4FQxhFGR|fh{J3^hi0$Gcyq}$pp{8KHJW+QR;}h0KOzqb|($Lgi+(!xh
zpR25SYqEjH-%oHSqM@LgqG1CYz~wL!SDl)TqM(6I^#qp>8bD^atbeAhaIu5qB!wx_
zZ2oc-m?BKCKrMrKqH5<(Q)oH7`29SS9?R`W6#OA4nw?<$uosao&8shfN+|K_>pqz^
z#DOQS8|0Usv4J=*WNRjMUr5S%e!;M>GLX@v8jcEsG&I2%JXc0Cv;rjRlNMTa*B{H*
z0>`-VeJ{Ee2c7=_if-iXiC%=7RinF<Y5!q<|FLI`V?J`Kam@}UvKi}2y1RL!YcpF2
zG$qC^WMsIKDPa&VsyxREAAqdwqjyBW_h<7)#c%QDc%t0gLo*i1Ln(Ag)XocXVC_0=
zY7LlU#81$q@o(SP<RqEb8?m&z@J`9H4hdx*onC20(-r(#u;c2-XvB7$t8p*%ljG}O
z0RC_L{U1p)xBq+61V-WcAD)=lc{)XTUuoSAO~oHJ=spj{G@mfby;b3XjDX+0Qlu_w
z!_%j0Bn^>?HCVbevT5gIVaN`wicT$<E&n2Y2xJ61am-R*0W!MEa(4H|I+63cl-%(7
z@>rXrTK4u_tYP`&3(lA$bG)xytFy%Uu~*1&h_f9z?Q?mO@-K4u@1E$5eEzXC$te4f
zC8Sv-ntLZONKk?MtW7Rru4c4_%+s-=zyWP?41bwIE#j&WL<rjwM32m<DwbkA#Vlr7
zZ6b7|$93Bxic}eEl55AH%iiAm8EaXtIW3tPF$<pVD-{XmsL2t3SMJ56o~VR$y+H})
zN`qGY35}qL0RZNHe78gte-z-Gn*{~fy+`AydJ!1-{|0!HhN*tN9_5Q1zzc${5vW-&
zZ%ud@<nC)YY|A=%&>tN36n;WjejC76E{;UNVQEK12xhf~;Vm=TRNCpjDV5JzEoL!r
z^;)mLXD~Y=8R^!-e|g(pE`~$yUMUVz;n()vF0g6rf1F>&?TczPwNnUL>u%Kn_2WqJ
ziT-6WPV?13!nWT)qbT<yo1KmbviY)Klx;}gS!lq&XuoF~{hiV!-(r$HGAlLR1GyR+
z5%Amh;OyvA1E4<a`>~gs2_q(i)m_LE#Ag<izmF$Le`;B$YCIzbFIl~Tjjh3ARS1KP
zwLZ*!<)T}*nzHd0JCSF&Ir-4~b9LLeojY<_F3t;&CR59Dx3#FhYdz^~fK%`aXLq|g
zgmSM3jRxuTSqUkJS4C(I?)c;9!PfkD>PK3;Z5IGz*Mm3Hm7e*<$;Uj#WCc7SO1EBa
z@XN8SmW;*Cc%K;f-Ha98_4t|qg5TD#Hc<Mafv+%tiroN4W5+!PKzep}tGrf3=oN_P
z+O^=NwO&nOx^K25S7%05MpG(qUNxD2tL#d28Aru=>VEdCwTOkNo*V<mn<cu`{$ei{
zait>bJ%SV0a29u)@aNAp3hi`#KoAmoDH)R~^y06Dx@QK<?502;bX0uv<@&^|`ruAP
zj!e?F?C1#ipWnnR7R7RK*(RIlUk}nV-|M<$#9AYvYUl3t(iG{md)L?ZmpewlC;YY=
zXDmum;!nb6uB~JaK49!$JLa+cetK;|Qj?s?S2*EfXF~?iJi9my(|KalC##uypp6jP
z)-=5`7rMSA3r$UY{O|w7XX4N9<a>NR`rLk0oqWz_2d1<}x+KA;xwby>J|&m%@Lw?3
z@UBsh#de^k4Oj9dh3P{XwHMdEfU8Mg(uYz~bLux2)<Vy{p#o&5Hg4STxy5!ED(*p3
z@>`JJ%ODNSTJWH{LJUgc_0SN(hpRPuBcf$C2JWpu59A8>yQf*W%Ihu$>`KgjH0I^X
zO=e>tjv*t}ZN#MAtD@0gq_+J$(SL9MgdU0S(4%cyP~H5c|4F_3E@<-}S|pTnT(c9S
zZFc32)YKzk$7#%f6KyDyes>@R=YD<@JtAYkz&`>$T(vg>?oXflc$+*{1Fop!K+`e-
zK$L6xFPX2eQvrY@vdcl3KCV}8*6)6x#HPnc5o+`9C@P$+*j6wRx`ZibgTdTV>UwEM
z3ih$_{;grZ!t&9cAB6FXrGPNrsVS)7f6!NC-Oj9w1pufI$fFDj=}O6y&$zxFS=qd0
z?q%}@P;+?XU+@`z|BS;K&{Dhvuby??asWgrg8v2cr(YSneRKy>7;7euEa9JJ+z4N>
zYh9h)g7qQu_wwSU_BR}8<AmIU2NQuk&Ft|8-|%{`%cv(<CHi3UoEBnPQx4Qn611pb
zeJIPGO;F-L8MEZI&^{8y9kqX+s*wB7O0M+1lHan)ug-6Gn%bwx@{sdGYmDB4HXm%P
zAS><VaRMWO_S3B4zqZ?vQbVmoV{4=$^T9_F?^_Sorq3Gbm7z3QUl6yCSO4&fS;Ns4
zTbpJ%DUpQaA&A2Oldd(vc(a=)N}4S0&^1drWT@^jjdkOn;(9QzbMwj71&rcu78%YX
z`m6C4!>f4xu7u!UYwVKUI9c=WCwodRfSVPGY4$ba>qptTply%15n{~tH~nM1)!RZn
zq9{D$WDxz%0AxigxKUy05m)wi<9G13Tce=ytLzNSlloB|Ti8-k4gE$Ugm#JMsC<l(
zolhHP5cR169i7B@J(M^<|Dp!K5bYUEp@hBG((K$ssSNIdO<Pj-BiCyQp_=%B21l&@
zY(qlO%A`82wO+b`n=28*-M#v&dF>M2Xe2HQR{z=^IAMQu<K5(#zB5r#;CeY%w5?}m
zqe*rvHRpI8cK7sh&a$ZTe7(hT3riB2QiB~fdLc>}^fHTZ3yJf(Kj@(s0W^9SMDm>T
z1!tWOmAUM9<?7AluzKp2a3Q(%^o!e(j5du>$5m4d-eNY|@~GG4dLu6k`*>8z2iH&W
z2{?xvxB!Nfx|i5y;%5M(Ha?yIHp=H-L3+J?x<AOi>YKFR`xp*5s*fyPLX43d)}q|{
z1_>JgB@TL^DPP~uxcp5OdsJhF5JPUX5lVHq!-{mm7Q$-!5X9w#T(9g?N5&@llu?#)
zhmu7XfD96(iIFu)u8Q&z_NUcmRXfd0@zkIz!7PKo%FFL9ZmO$#I6iB)zMyAyVKio8
zloefIxE|>-$Crj)c{TDs$>uJ!K|l`)!nPf5T3B7$e#1iB1}6iJ(v8w<5koq6Plc@~
zMhC`a&yNxt7q;M)O0B{kf0Wgj4>2_vNTI&5i+~SYar)pfb0cQts~FPn`XLK6#ynBg
zZ!av6IBuxEAK-SKWG4u@X?Bz8F+~?Ofd93sE6rf!uz)6`F@r%xZwB!CRxfwWt_Y2~
z)Zx|BzXj(QHebe;B6)h?R*TXOL>IK~Uilpgk{f}1{(C4m+E7pv=86YkPEnfmQ$QU2
z?e@wb6`axQk~19&noEckrYF9<9)sF}c%-(ZNS|>iljoTzzf(GZ^&bG0#9Q0|dX8)2
zNMX`*jcs7Wb|3CrIP<xo0B`pU_Lg%`>8ovx(u<#t9%oG9x=(E;yqY7&&lit|9kdOw
z^>B>WP)~eR^>`!^hf5DzDq<QhD?sq}S8|_gCQYSpYufuef_I8<M6l`iWqSgL-L+co
z9iJZv?)W2$A*VhOtcYm~`;?e507@Hzr1p~ODui12c@O8O@G@nd37iyNgKgecS_giO
z#eSy+V4UMkIda3L3iM2Oyo`>rg_xy2amJw?fM2Edw<jiUzUC@(UH^8mtj0_WmG4sU
zcnG!_TrRud3d7c~)CC3_zMxfjd5uBy!0`2H5Du*YBTpatKjAF=lMU{9sTr`JCjaI?
zXc^JT5^SOjf(B}JI5nbuh_(w`2y4Mu>6NJb8tjKJV5OH9PlGO+5Hu1(w;_)v4S~cR
z8TZ<2dLHBwM??r&_sFpgWdPTS<S)o;HPG80Jk#ZxHa@F5`?nnaCT$=60?=$3MKH&X
zjHHf<F9s$z7)_YVaTB`GX5q;*tr8g}JuYYsc6#o|W4Gv6>hZC}vR}1Bkb|c(j<-u4
zK2SN=VFE4t{6)?a9i?3_=aosVF71&Td8Evf%;qbJ5S9#qu=rJhXLcg<H0wx8`EmXZ
zsIACPgB~lx_VkRwSg`g`S+F1^D#G-b&$ar>rFS+EA2V`9xrdL48_Cyl&D5j)4)t0@
z`8J6mQIDhYrLe9;`{-A&E#V!`n$b;@FpOL9pCvwI=z-{y>qF&m$4)lu4%(Q9L2msY
zaJ#A?@m(uCDDk~*IeU~%@V^(M*D9R`N9*;RFF6$VH4HADIs`!n&BOVMlp(?(5Yf^V
z!FjdYkxdvRUxDYsm3~%(-4*ffYueqE_pp^=YUZW#vxkxaFjQwg5O!QI2l!eI;gtm1
zV0wJ%ut=N6AKt@>hg)%!8t|;{en4R<W5i8;d{s4kq9|7_AO#JZs$dlDf~`)rVH-XZ
z+9i|KPi<hFkkGYrKd9rAtpe(JHot=3(HASOtNB<4Dz0O)Mb_;KC*`BVCdX;9btl?D
zCjHb<3RzukW|64V6#;jSfrmuEJ2YbNkyb!iL(SuB&X%(*rW3iwYgwMR%piS;CO0`j
zfIh||Sym1Q8@Sd?DDMj4qVB<Y1dj5Im}8+0pe&Dscb?aUl8!<ZG#CvuxIq<rBJ6fc
zTndnYAbcLi=a3%Z8+hSWqtwBl%l_B>&mXOXwuqpW*bbDCQZpN%1a7NO?8LPDDq8$)
zNojdpSsRSqy)LRWKdSbabGiAdV<$jwvDaB!2tND;@Qm2ary|98k>n9kLQGj4soBxG
z^$=w$tlX;Bi=sbU>oqG2sh`~*9=~V*mXEjB?<+`Yj!9yST5F;8FsrH}{O4w$v7Q8g
zG+E`IuRc2f;u3Uy88qB*B7_y6#tq6x&b>H5UterZ+(wKD0u}#iY)_5+kdqJY5QNB%
zwx>9CYSR}vC>r4DQ?8lR!=2B>0`6NQ=j#e44$mx3$BnTeJAt-VR9%d+&k>MRyk2`K
zTXZbq_xBrDIj-#+n!bDm*p?+c0)C6WeG|QxuaJCIPLBwb@EE@*sjf>A@Z)l9)ONUU
zD!FYNVOz!)Lbu)X;=HSl+{8}p(bjYlINy^`Uq=)yhEJQ1-Ej=6f-99LjYjVMdEBRn
zdw$5x;+p`uIp{gWhnqEzX(q%*`O5jlW}?vMC=TD4(H#unmDJ4gCi>7-`ifEeHbNa)
z60UP20$v#dUl*oN$<NM0?`FhCz-J4*g~?=*en5^P<b#rqlONZ7M?lD${=<Tp8J$r=
z!k(WXm~>8<o*Cdv4%M3;PtJNXKCo(WqUc+C7K1W(kWMppfVOHHmt1$h)^Bv@z`Uz<
zLX|N-kfD`A<F}O2)aom&)Gc9cE0NO`NS|M2XZ+|!**PV^@2D#Perr4Da{`<!<1|b`
zZw{N-Jxi+qC4WotK+Z!BIL1K!3-g2X-q2UX*T%so=ve7F@qq7AP3Q%WwSvq=#(eA0
z!z(^6hpmzPeFV|gtm%j%<1kfaT|j97+2T^)R_4F0Xs#h)-?By$&6sD^x$;_DNUC^;
zj)r6w6v%*Un){+4s#PT~LY-I_W7Cg|1%z7`QT<KHMbn{`uHDK0T)(2x@c2_cwd(&9
zNAwL7VAJ<o*=GmbN+h~@t$sWh+U#c1Uv?sm^c)?uYK`3SkGFaR{ALwM40#5t@Q)y6
zmnC^FU?od9>&yUv>K8@dYW(B9)^_)V9DXd~I2IicyNO=Gm*-Nhof!MFiY69iogV$N
zR3}34Y1m2slo=h(s{w~b4pe0wq*4~2_`q_{BTsFj&lDJV>CZoyyn@Yr#g@n67c0v%
z=*xCwOc0ls%^!uIn)zi6eA~6r4u%QZ!2{=j4E979>wl9@b4TmhbvT%~gz9#+gndd-
zWtsm~qk|VAtfA&USVzp0_OI`BkgZT5oY~J4#V!I$W1;;)_A%GjUJ9G&H3eT2AEX#Q
z>@Dq+8dgfPH>dPR*MNP)o04Ux$nm>*%cP4@PLXj>B}mh*10bWY;p&g3$M*0Z!6kzD
z9UfQr@g8A|UNkc_N7tuX%gOp{3p0N9TtjO6YFCp_CrTK6nD@(0_^Fo0?h??$H#JXR
zB<R;C8=@&+NU2U}%V(vLQsbkqDId?$PFJVw2>nDFQ|BvjMmIwTR<edd>sIeUwjiGV
z$cijsI>qQP&}0Lu8X&RFm!WQgE5dHh*d9E-*n^3LdS$S!kUZHExQulqUtxd4lt<In
zhk&*)x^`@j)|P4bLiN|B8%O!jw*UxHvD_N)nkTf!*Lbj?<7@AjQ?tE`In_}U0+DA{
zOwdSng%w#P*a<@4{K$@9(#$Ej4MJo1Jt4eJsi;0_vn`-EC1{NXnIt^s@*CrRRS}%h
z$zqam681zLNn!Pf(z4hnvF;g|XXqSU6tN3=G!hy2u|E-Y!6Z&2!E^eu^NBf@YVaZ}
zxx)PrQY`hN??8&Z)|!0ZnjH2l&+ewWUR*g!4B50T>CB#z@V0LjfIP<uO$iev)S_bB
zlt09>Tmdgb&&O-=-L($rhpN87fc46G^=KnfK$~@eeKcX<?=)SFSMb@vvB=*W5!p~~
z<OMX?^zHtm>KA3)a%057bJ;w3wIsd1OJM1xQDOS&6;NU`(1?M+II!DOAUkGH=f2ha
zKzS)bXym(+FzJrSQ;>%>*w3q~?6{M)<qNX_%izt6%WJu%jgM5+fd4vJSX=`RD|@$@
zV7v4j49$p{y6<BC$-caRvBq(vFWza|kM);<PWpDDv^3iPH*L56SmL1{OMLmp#$oi4
z&tpAJDcLt#XpMO)tL|^`iCVDQ%(dp{bnk%_Iw?HwN(7u&H7L#F7*?Zgn*{8J?|bh$
zeSge1Z5NC16<M0%USd76xNP-C&igWK4R|mTal)PLaaAs?^xkD3%8^%4lY3N$Y#wLt
zIj-bVJu69Wj3d95baem}kGk{a{HU!NxVI;Y9za2^YpEvCD5H@m{Gk)&ZKFG6eing$
zAOGlwI>ZG5uh<NzBN<2gI?)=g(n+7t?On)tX1p(Lgu)H19mgiJC5P<kpJ~+trSzZi
z?&0<nKCLhtcxw}#0zz#=GFnL-AL{wLpqm7siu$L_EIU&&yazxhZM<1)B7|DTgC4Q!
zr<EDyUHn>X3+JyH)?pWc)#cCL24CE}q$ElZmRVY^wAv?A@B(U+?1KRZ881*W%GZbr
z`P@RyY{GlfS$oT=Uyw7fVISCT0jLgr&(K?h8eG!`2b%DzA3q3frU4p+<OkZR3P2l9
zo`h7`&$4c<L3W1Rr-Ol(xxWUyBw6bbcd)g_2<Ye%FUI_-D_i-lS%nb8Dz6?d*}XMk
z|5wo<w3wen&l9TgU&u0N3jO2KIfcw&#)vMk);fc}oEmVj!ksAU#pWpk&#BN6Wd5S;
zF0SGX#?foHup@Cw;+jbh4JtcX$h0jREpL5cRmkFXeo9Nikmrm`Zk?%xb8hnP2}UNq
ze<pexIjjFYXN64ZeoOll(?-q_G#`IW)LVLjd-yN*cpX@34Is&Ds38GKzW>8HI;{L2
zI8L#`14NgB^cTh9oPBquLws5^KMv6V3?<;@)R(0>@?#wE%%Xi-byyuxFlJ3d20%fg
zJ%zeZ*|j^#vTI#S2@Pv-GI7PfM5A5|#Y7gv*Ze>1(s1nhxxkvgH{1rj?+1Q+3OMi}
zbZOU(v{C6a6{vUsF~IMRFnxu*w2jzTFsMST^(H9@kE!n`9c4>V^m~C~Dj>uEMJt@)
zKL-d9{Gn{G=BDmB<$k4t6m6Ma1@a^Af1isUuD<qIbWZ!&v-g!Jz=Lpbp2+Gs*P*34
z>SNkL^4*(c!~m$e2>+;}JfXojoE+-|n~w12g?*lebVk5q_)d4q?0X8Y9$9Qy@L<B>
zxH+pl19s?1FLdtFsXfTXbY(}%?+tC3*)l&iG&zn?Q}D=-(lMFVtxx?k$<1;tfreY(
zv6;|!bR{n2&3DqX^7KB4XCt{3FU3<jpr{huLcDyWb&9)>uK8)OriYeImSj8L?6C!W
zis2ftlDwLtJK(55Nvg_D<&6>WaW+?PhP9}BUz!&R?*p*0Omaz7+!G*~WiFl!)j$MO
zhXtmsl3H=;Q{y4GF?>I<D0mfo!M;sO@8`0QKi%L7gg*rAO<I36y@`9g!^dRQU*-VZ
zN%sCfaPK?t;}i^CnRq|^9sJ5Lq$vW()E14~u+4xZ&;fMH&JW553+h($bZW6^Mn8{L
zuFl2g=hpf(FcAYxoW2i$iC=;GF961hw!Z--`9lc;`!&%xrT&qzQ;%0=&PW3hQOr<}
z20FgQN<5_(`z>+|e&Kg=n(Ww!X0RZd#T9{Q4#5}O!7{ZlP}&`*co~&zzz~YP8c?jc
zi}v;vUIn7S<j31UpzpUi^dIZ+x2f+$hH(Z$hm4qGF-t=-Mq!T1ll3qzJTm0ccRRr6
z=59HuBMkuiZ$MM!^<Y4>7@qc~UWtLRmp`AtUOhwukTb08_CB6Tz{SuTQA-6>TBrvf
zb2I~vA8ljRCqB=74*vlElM}G#mxS?eOw<opj1?2W;-}AmsgVKAgbH908l~M)efoI{
z-=XER2%&Bq=oHd%aJU2@UTy}=zG3n@z<N!QE7JWrMG^462=_Hi=0qV10@wuK*{}zH
zH_ycCxXK)67vK>n4<^<&sHh!@(j&&kc)=jU)p2k7XHb9)eZN6{mNy2s*KAdV3m~@W
z@Fquk>+BQ17862<oAXr3yNgr+R@S-}2THi*Yg4<J2%93GjO0n*oHqYoR@U*!jW8+v
z5hf=BM@DcLL-Tx^D@{g^D`K`;&8iZpfPGQ-1*@CG>v10d;VtN!9M@HsWqFUHZB8Z(
z*J$Icd<P(~_Y^=Sa<oX=-Np%b_ilVv`5v-@C<NrCA+I}vC`sOmgv#Hi^}O>TJFyJk
zSy7+aVn^2di-_}!96Fs87m0eIO-<<iRn^f~H^`#9fGigHG|>E(YE4GH>JzcjZj~`y
z=7i#Gq1jgCJYqp-=D-(V_96S>YV`lGk%RwuUb$%(HZw`q@##MQudH4I@6QuBs};iw
z+hG?U)gAtLo!kTz`p%1`AOiQ{s>ErRl!f5U(K0y02|_$*GgoUyj2$N@XKM3dbte*K
zpj=pRJMWAPf_SA>0nq3($ZGsg)?P&0K4-Mse@{>-g?GqVU?%Iwr?_i_lE?}uiAPs4
z_h!3O<{3=;Lx=IBEwisFrw3AY1fn8}VzXBy%a@=7Lj@N$6S~)Q-=~(oSIXg;9Bdpf
z2IROme5JEx&Z30&)fHxa!G0+{<;~c3iJgy&>TaD-8w5b(6IWYlPHZzFX8y=j{7QPs
zh$!w!1bk7Ep2@YRcmb*gqs{m87Q#O^ZH?nfiwi5+tMxCU%j|<pab8S+WqvY++Psgm
z5bi#IZuif%qyOAEc-QX-|4#!IQMc1J%>~^WG%Tm$b~;!et~)MfIVxsK2&r`=E?rzA
zi>-(_CaAhL?FI%ERzr9BAa($S9;~aWS`53w14TLVC3x|E;Q*N$@1F(mZSBz6x?<pb
zkBo-(UzvVPDK`Rsk1rc!4E!EcVhlj*C(*BFAlFicAg!*0$6Z+<&Cx*Xgyn#-uQ}|#
zwUH*U43yFf9*IQFD}?@5Z*P6x81~T?1_T1Qm>!3A^h&+^lNd|QdcdZUla`tm{X|_U
z|8|xBk#$ewz2j}fEx!Q3(oT`&$Pd^U&sd}jLnv2|wC<eB09co?oVZ9Cp+cv^DwtYP
zSy0C-n_|T8;l(H0Q6NCcfx+XRdUZf2$#*PP0++A}gp4QCV~K!vSE(yZr$uQG`8Q1l
z%B*VQ!<b*AK+_JLYh;A};cf~3&yMsLihS7lQj?u|rfaL>fypm7@1jFeVge7eM8M}%
z?~#-qVb=h14XAMfa{b<sWf;8Z7_;xWsBo_)gY;45`Bo%<4@If4QxM8=auxLtKh*)D
z+)q(h((GKl5r76wyD%KMUJ#!7mx9!kU@tFD+~*tE8MJ{*^U`NI15WkF=u>x6AqV$P
zLjjgmxRz$>Hjr{%{f+gZ7K@g&a_^~g%G6sy<%WGe4R7pYWd+Q(QYsF?G9rd5oD%tM
zp9-SKeVYWzqsE83hP)RO1-2Q}t4}@lW`fN@=BEI|D%>XmpI%h`yNLUuL|FF!)cGZ0
zdT+ox`CoQKi~f!h&!+9jH_C}(?$91g$R$PoqbL~#t;cVMD)_s@@xVxekpn76K6Nl<
zw-#UAYks}yT#e5(Ha0putQ+ux;8@k0lZ%WpVA^dZeoreZwj?RxkY8Lr%$;B7wX4(t
z1&P#fF^5IZIMIi!|8SK4DMQvAuR9CiBQTz&@x3Y)Gp1LUu0jK{HAdm+_C4&<M&vgp
zYR71#Y$W+!J3J73>piFf2AFFPRLi7!qqMgs1LTANXij51AVUB2)^tx>Zbb_7<M;1>
z+*VXw16E*X`zqAbqVA3DaVYZkdLpeGt^)PqWK~CxI%H4B^Q2`nYZ6?8+AOUdgC4WY
z9BE-J78ZNcClbWh<=M@6&oKYu{OU?S?b!eu=~K>=N2(2vcz?tjXE4sg6Evix-<)Ym
z0wy1lPu78X7>FR^gNwdCH!vt&|NnHnFaYYc<itJ2>WWsA2FPQKN2lAr^pBivaXwe)
z(-7H*O|+GK*EBVDDlit#e!iFVpF`BHBq$tuSeaYtij)Z3a%5(ao?x_vKuCLnMCi~0
zgEBl*uYT&~;sYq3!z;&aJXkJJb38e!!Dy;4T9~`#cJQw={|-GnRg{*+V<e_4eP5wz
zuuNi1dpaIqRbD88N39SQ0jC(PiXnv)09;p(aH))dJFzl1TsJ7dWa-{}ZU%aT>~azW
zmhjOd9IuOsLrlk}*)4>JZawT5+4&8$_?^f&g@l#+3E>9DlREq%OCB(h<Z{RORve8~
z%RG(X?B&%LNm7z(ux<HhTenGO;biDQpD|EAxo}SPLU!Y7O(a)MyqY`@9*R8l6ELMs
z5`XBaOPd8v#$Vn;`0H$s%%H|`q+-c&jhU9!JzJ5Ad`ZqR8Q!BJ5PD+*lgHC2wQ8LC
zz_Y1{P(^I=_*+OYGQhSr18f=f<bYA^3n>8;P$)JLL_c4C{}^ZST<+}jW$XXmVo0|`
z!$T8<P2U4LW-Y%Dpu2drYL~VD*m@!k{KAV>_QS_QSA&6J7!Yx@M*49JXdyFXsRd-W
z#<-q`%Rn}w-2HR-dw*lm?G2}(V~2pz+@H!s6rDWVu~`YU#E!*;+>Z;1jDUNK78%c6
z6HKw(qvON?@){%0<2X~jkM#qs9AZ1bTr)lnT>^|-13HdlR91nl!R}T-SgF-ADrXBp
zvo7jVq#f&!Ggtmj-8Wq2?9%V@yx1lH_P1x&0mHPfU*73;zwGf(n3OZM$Xy<CJu3rT
zIP^LtCpdnB!ssE7nZT`)Kz)d7ORT|KRr|DX^z?F>nvg(B7iS)l!KxUkZM`{_mpDK=
z^P6Bk65i1K)V`<k>o_)UPFisl6B7~{<Dvq43RjyE6RCQYOBnPgb)6Tedu0CA?vFp3
zJ&gjFMwNVLks^Y<#C1#>PxdJ??lg3L%o&hlGgt33AAjuH%f-&Po=%-Ij#pto)fI$H
zHw`5+Miy^?`^sp{DF5i77ofTFEZh!$#XoSoPpKC5aDy~U1NQDp)SCnd)_tw_Nk5I6
z3)T;Gc|$R`3$RR6($RKDTjz)>zwk2t6n*zjB51X@kr?diu7v_9U-a2#;hv`AcB>x3
z+p4M*Pe{Xh`RXut7j}JcP0dW~RWFGg_j|Nqr_@FrhkRGtM3kvP&LxMybNDY`s=KAN
zG}LBVU;-vjL}^IKn?=Blb9nXQP)tTSnN)*Ky#?M~^yuEeN=Cc)38Uds)D7j6PjuNQ
zcN9V8yUsrVNoXNz3)dN4Y0b=*sTMJFF}(V{Sr|p+Ceva#0K4d<`S2N5m6|nLr9qR2
zS}!JRxTIcpWNU?0MI~xlk*f@}jdV=-7OA;DxuW=ff<5EPqCRp??VaD1(7-VhRk!qa
zs#KO;_Xk^pDuaa!{6H#xNyiZ9cfZ^}=&r~0lptjgQ?@#n$o7#%OU(Q1UA?jBsQZxr
zSJsH7w0M~OUS7|i9*zmnwLkMf{p;LyHhihC#Q^Zfrc&U<)i~4>z^(c5ood+_PwLtP
zP=;wCg9*aX98ex0$w*1NJw&@LSWW3@{8NNgInDxys|4BKA`Whpt)shuhRv)Br6=Kq
zx_(xP_^4+m<;d;5uwd#GsuUP<jk!f4QsW5<sAxnaY2=Ipm!Y*@oJ0r9aV6Cr7jl`+
z3wN!rrv!5R8)Ig(nYtvMyI7w%oh*V`lX6S@Q<Zr<i@|KT`%mZDEcBagl|1ZTkUnJ5
zWz5pB7<zxLdiF5Bx9VI9@J|=WfdPS7Go`HTWfKdxyI1jMA!8XljbViG$d1cFq3;a_
z!Qv5cvEu$?rs46;z;l8Se`lr@oEF>nLGcBDM*K11H-UPYy&?I#eD4N?(vb8QxUb2b
z7>#L7lgG=-^GcfubDN+0Occrt)L<!Jg5C-yEAfvj0G#$214TI?AYhdhRhp6>FqPbi
z7)TNGygv`cqbR!`+_Ff)Kj;5D!zqRpj1y_1JzpAd51Yy>sNt}+3NRApGu=7lw!es3
zF5|^^@8qP52e@h`>tx$*u}m$fYq2i#T-n*8U|`lMn_SxVK*r&wL2r3r(0JnkCjJRg
zz1*wa+MtKo&Vs+o$vbl)OPg>zRfz<aR9Nrbjr)S4Ao5j>T(XP|PDi_b?3R`;_`nhG
z&Vr4Hu*(b}grHB!=~ARBGSjzMJ?>Xv)V$vquX^$w)Bo_D!(V5?GX2o)n>9!O_53p_
z(idMt!d>0zf2O@?LT3g;&eQzSfLwf7_6l`J@@a~%-7)r18}aHe#AX0wYY@`%QD+<9
zG|CmR3%NrETO7l!z!inRh7E9DF&07{f%XG@Rn>fUk=_GLz+5oYWm!y7d#EZ6=Ovi3
zW$T}9`~6n5Sw^sTQE64Dnv|Na7JR*t_X6V3X`kZQy`uo)&y6Q7bVP~C8oT6Y>uQHs
zBQB!ytD<e~Y}}ZEueT)*87zFrLgeJ*tyIRU(#IT#StCUFO>#*n27!U3{tqj?9r2^L
zo!c&`1i}N)`Gs~$>T^0{GHwdtfNh~B!;=)K@pxc5>&!ryIl182jKzO<S0Se(;Cj}`
zPk^NWtyOrsOc%H6Q<d4DP*GHex?mjG7p*~`>i=kuL4?Z&WFopcI&;$u)7;w@?w+Pf
zt4|raOcvU8oV<q6w(9;v!qua~H)oF{esmjZu+pIP4gOq7a*jKi5nKy@qHQF_msI92
z2Pw@I{N<43b^XZ5%8~ieQAiMStV#-zj?yvi1Ct3-j(nqT-f6+}qwk-)@DGxdxnIzc
z<&6W|@+Ero(*=aPHLmpi?%7db97*{DFeRa>Bf+yl+?y5xEW-02)+La9khro1V9H6Q
z=L0%nn1ZCS+EDgNJ;eY-!~+fPuuVh*uA+#qF_pf=^fD;L<Ia?T8aznF5;{3P)#4eN
zR*x;xbIvHYb)TQVK-OKI?+ouKazNd{U?jBje5$LoCTfh<xTk$S6t(PfwcLqhU~he^
zn9fI|%%l9gLRt0nvK5k;dZ01qHD4{4Yr$#RLR7o+tA)R$wc#x02?5R`<Zuop=mj)W
z(W%)0?VlZ!bbTjsH3VKxX=%U%d7B%vjX<*zXb>=BZ!p$rn2g;G_T@h?%PR(|t1Z^w
z8yFZ0(|oVK+>Bpeja0JQp=PLaVz2OH`R1p>^auQ2ug=NyIc3Eq7OK<(`hJ{reV=-m
ze&A#wt&0|_oYg*%Q)Kko%E+fgA5m3$YsG)jiINIlEBPdbM0jQ&->w^}LeP=p8+lDK
zVdsnJ%gp$2;@U)^spB(sP5}KHtk2?i?RMY3A({WUumM%>`p1~=nEG+N$>>hAPU>{V
zTu332v}j7qCIQ<idL@R8gKHcYjI$iN(gH)Lq^B64vi;5yYrt&E>J!+kA(Ox&jBSsD
zUo;=K*#6>E1l$L)(2nMx6sMEM)%ut~!);yP4uDSq^KSctDie!$@U9QrwEVxKsR6Sq
z!rFF*C4O&OLuTv@i2((pR>`b7!X91YZW0F78Pn#L{?OwRCywK+*BX#yK@RVp!#f&x
z0_w}eHz)c^&ZR-?C%?sworgDwiHd-j*o+G{w`wzS*LD@7F}f3~T}Opo0OVui`w5}|
zOYssR^r?1>BNZEDYHCOdM9EIt0RISii^;Tw|7+MB>hGHp!TU?XWw<(OGZv7eL7aR=
z#RmS%5?ZNk5a~Hq2X_ybVel^0`Z|twm!4~jX<6n|#Yx*bCEY9;!P8z9(N-zqg!zWh
zEU(B>u+=pyB`_AugDrHxxjt=&%#%&TK5(3^9l0qB@oA*86~4LxD|E(|OEvkQ%i6n<
zUHb&tb#$hOAAVUmZ@<Z?VSRg2Y_`ZI$Am_L*WM0$t(xD)un)f%X#DH2_u662`es=5
zJ$QgkW)74Gqd2jkFQL=aFaC7E8yokz!?r)YFLfWUigJ~-nt^aQFd<)-gV0vHRs=A|
z{(fpY+0@<pio5@0SNepFX%(4Ftc0PtKG>3+qu8g{?pu$Y%jH`pPT-NjwZ1H{k4<w-
zF(L2T4%lkYv_`6Kw%)$^{!=1slWsOYBHTqr^uu&1>A5fLJ#31<3X-3$=H{l~obr&k
z=eVo(m!B`~xk0Wj3CI;I2TNOzd|o{Ix7oB#6wM{mt*;WVXHr@D7I=`;;-J+VF<oOd
z+rqM*-AB*-4;%e`{F`x)?0G6>ee?vEFjp9fkQ-8(RH8rT<<{0hhLyK+ojN9)Ge@Kw
z_y;SpaQ1HMa%GdPPSrwvC_|i2bv78?&18|wyr&MOoj`#D`bwo<s%2=K0=P5L5+!L!
zfZ~<9bwdAuf0f2ls;)(QhFf=n>fP3#XyO0l&8W<4>y89s{vIL0jq>K%%SkTgsRjCs
z*6DzA^v4sFQm3^fXX#>208k`L$##WqOEj_Q0NT7$6}C4)P#k`IJ9RhI+rI@d%=whd
zb+lK{`-Ex7q1$C*lSY8H%^KAdi3H}%$iRF#ApA_}n~edAZVK)o%nrW58642X)uF=m
ze37I&YTVFTEnhx7eXxcUzQ%4iYpTSu7zsE-16qRq&?=q6!hftm_3yMg-#qszrGHMM
zJ#gT@8Rxm}=pOFMBWHgvkXb(W9Nfw84!2n2Z%GMH+ecGgdN@>B;mlK6cwpFdIy~r|
z`Xd<+$L9>nv%d0qQi1UbRU$3Okj9Ecb-wxrMCi}0-TJmcg!BbO2x}ZuRQgdG|LIGr
zTc+(GJo&@C!+e4->97c4ka=W3*a90Wcv1INnYYZD$ID3l-C#jxss4C~71x6X^4JjF
z*y5_DIHMYHa@}jCp^4}#OjWgK_4jp-*r)4d+m_U$=TFmQ#9I3CAy1={Kb(rJx3ED}
zK`^=cs<H-p{vtaM2ldZYL)NYtgzD>fnU{d^V013wQ-70ha}h9jLUdnhoX(hBX?g>j
zV%jYHW^}i{Y-~gq*vOs?l=tN7byAY&Fd`x*BxNq9sL^s5vje#b7XwUDVjbBXm<Qzm
zgF1SD!B&cqWjBX8<rz$qdk{J!jy@`_2%ZhVK5jdrHmXWn&zx^&vmU1FzG?OE-*KAy
zp0vtKyg_;^tkl}UZn4%Oy1kA{J!yr*q`U^|f`;Xhg|{yyE2nd&*@$;)Rot;jH_v!8
zUaGfop{6r(^wsFGtNMP~Nhb<%OK)7IJ1?#KHx%p&99H6I`xX~va4mJ~D~xY^GLZXu
zL4v;pqheU|97FFih6y1y1RG<Spxs%ZL;Z~1J;#v+hKO}ae`shnMs?uPE6sq}l32d5
zkN)LVQ{oUviZ_a_npSK$K`0xNe+*1EOw3MQHc>USr(N{0!q(@=YUjRcPx*rQGF<8K
z4Hrzgs+w|K@%HA0YX}4`IXQVdoV&2_wh-nWMz}rU820v>L62J~<3Vrqw(MnbzaFEb
z&@c}ycCB!+BPC<Jf)&HQIo{;-t7ac}+E5HCMJ~xGEyoX*ZYc~T9?(DCQ{5?@ruHVO
z7MRa<o$08tm_6VY0l$I<S~763nps5AbV1^uy{NWqbN{Ds&l5Kf{vDW82AgJgm9xbI
zoI(YV_z<e>mN6=$f44*KfUa};DA?j#b@r1OQ{l#PFY@SST#~s--u0p$k0Sgio@Ijf
zAQdha20<Ny$WaFjy4u@cX+Z1W78hR@6BA25;q$s40~*6mm606!q0xrg0h3oT2Mo&J
zjV(=3qq`8(>kmB{58f3<$_^&tpBJlW-M;Bgd2HF%TbQnFus%J@rZ_oXSK`Jt{NDNZ
z3q+ZWLqrw*o)uS}k80J>>?|QaQkqc(3*PReP!(YrU$n{J=Y=jIQcmp00zSPX@LVuM
zj#rgRLAzi-MH&}TINw<IR3P}w`*$E9om2q(ODjU2TBaALjOs>_{*Z8C)|i#(<PtRe
zi+1N^W!3En0-pLVEOc5iXB!l_AE`PXJ$Jj`U7MVmx*8OO<pw%-lSqhZD0nRm3{RIl
zcH`wIaRYNWf=n(&F3r75U3b%x^7zVg>rZ0OWrLv)X2Xuu9E&L%OB~H!sNiPwr}fd6
z&ACwZg$IOkJI9%ezbm6)!!NBUy>H=IX=h8~dh+VwRg6#b=2(M9Y`HM~LJkjbT*cTV
z?*G0D+LrEm!P1FdA$=lzoyzqxkpaqGX>p!~MQ*4?A-rqI$2wqq*uK>B<IV#oe9DuV
z|D8lIEU5wasbV1AaW-z6bDFluBr)4f6}o#0ATKO)@B=|FRqaa8=VXLc8O=Viy33^o
z6$RrzO%twi;UN|!vJJ}YPUz|xM~!2)xC>F0YIv8<qM$H_R2vtNy>L1sDU(dX4SFOf
zdA{yf0l1o`vbM!TNa)|>f0VxC<b%#>qKkrzy%RblAtulVTiNNbItSlOblg5I$VEK~
zB!d%~)(Y``3V$ep+;U<@H61Z?lFE&jsI+gc>a$IqEH@MgA_Y{|1(}8&A#{zPT+58I
z9B^e&PvYnd{`7e(tLSpizSm?pk($!QABt3;Oh>TWvn^7pRq$uLT2u9^Mw31Kn>8A)
zmbRV)@ITKn>{3aCRpjbP{|x2O{A(uV@7iRfo%$x%&9a)+L(@YRA02R9uR975Hfk<`
zZGBN+Y2om1WmWjzk6vi+HbF16%`$u-8ISfSfR--UY3{Q|Zh_u|>i}b!f+wRFhaoUQ
zo2N4KeMWCRhd=b-_0)h}9T_*>UGASi8}H4X4G;qW{#`se<sCGe^$w-n-1V_I?jTq4
zAJ}Ka`YM;1mEh2!ilw}E%|O`ziCOUEJ-D?wNh^R=Mw+Z(L@~Lt*9mJCH2y$Ya^FcE
z-F4-K18&7dsPi5B?W6OnDuIp;sZ|n|2U^yz{6-QB`>k`-{JZhY<)lF8fTOkKs{>>5
zX>MqmTP9xA;Qu7wbISMR`}UqPmjk?@+l?o{J0s#YQ8R0<149r~+g0@3F9vydH2^Jo
zM{>uv%=xj3I0X0GnuKd^>+qP6Vk5JNSol~z^gNgr{3bngQpJ$d3zfADBzu9WA!BL%
z&QkD~K^cbh54KxY=+Qd$nO%v#TdM0H6WsbfUUl1Pt+XxzosMDL>{}f0!%nSmbm09x
zQB{<(rElQ%{&T-G>)q2Im(#KJrDpq_=W_!_t;d*jUZ;~kA*Qb>+<p!6dVp)<<EPJu
zXIjjPjeTBC8v>n7$W$Q^3Lg4B5hWB}Wv@z7V0btZrzOK0XLl^ON&@NC!5>q2Y1m*z
ziimSTSR*gn>8JH5tn*B%zz(VgAAw3;T4cqps}p#y3kl2asO8QKHi;E<z~J(ZWTnZ6
zJ*vg8;A=~tpKKdmQck>`JtHB~8uPWc@R4vJ9KP07iN&ro6zP(q3ZHXJ<1omyZ`QyO
zR2chzqTl@~N`1%3v%8m<Q<nzHTIwec1;rLQaa!QYGc1n*$env|K!DUFEy989Sy|Po
zG$RzE;fI4RV$1A0tL()rzwXeEPS&@YP6>Z0o~*5m-1Y1@p1aH}A{B+NdQijHS7lWw
zPd*<waKPQub86npK+7$_spG<7^YHS@Xw!zon@hGUa;25ntznNrOoDZEZ0^NqJ>QI|
z1tw^X64>pf3Yb7KMe7qeGyDGMG6ZY>Q~VXYHEEXvGd|E%=FR*ezr@oK_8M~s043Ky
zg$2B2#aVV}Tmasfpzg!B+@U3smz8;<Ij>#lkU?&e)TQL98$5u|+mYJeX;@x(Y|(d4
zO1;<7)QWlyBc35YQnA<r3Fx!{UY8P=U+8-FzKj{o%}KxF>gq~7O?M51b$NI}bSiIq
z7V(!A{h4^p3kZeY_BK%0d(Sj`<u?+;CsLAIPxyMcbVhB_U*nz1$qc2Z%h3HD)mQ!}
z;RV~+ZPJ(s9k#rsJ^jM}&K2MVjo-{c`q~ykpWzX}z=AKj_QN=sOt+7KXMc5o$-b6Y
zoyvFQ`1*@YqudK<Q+F>ztfB+ml3#|?R?n+uXhTdb?02J%@BY@^oi&yb?f|AEOJF#b
zF4d>ufNW!ON{UX{3NMVxiSdH}-PxV}NTbEMu4)iPvFi%$3@gy=Dfp)ZaeM?@A7Z85
z^Cc0Rw2h(PK;UOaEi9|<3R({c2r{2aQJ_rW{eEY=U>8EXJ<AvLnYB#uc%5ktlN>Ym
zouUL9M1*i{JJRbRQ&fo`Dpdl<#a*&kOS8HA@0o}qTt-Y!Q0ZM$s|1dFdLc21NVH5Y
zJ}+n+GTq(WF5BCufnx$Yvn}jOTrk`lIyyQbqyk5x1C>gpE#_Xm;N|5-<2-sKiUEOJ
z5mXwDP3=kRWj?|OQ3DdN^t;}iKG!B6N^Kqr>)YfJYqp3u2wwwKq3J26=R(SAuM)4|
zc3;q?63wK}G&hML?eL-0!Zdo-@{{yy9v!h)2>8^Mv-A(?wEvSN=p%2877T$+a$@{n
zY3P4CK$A%yGU)w2q~0-15sW1Q&gsFwVA{&!ri9tCJraLy>4~zO5u+aoXS4bBVm9;B
zSx7FZ7c6BKNR}^{_Us0T6rj*28xdwMp`1wfheaaUw>=p3n9#-U>-YvJv!@_*c_IgY
zoFo=Z4`zs|vn_HuX^U(Q$pO)VI~g+P6yWb4TwIh#AUMP+S<Ibw1hWGw|Co8OO``fs
zP-#dLg!Xz@IPcSHl~=#rDvWw07l+^diW_vvLxUF}GtKBpV>1iKBws5s-UN&I(&;AF
z&s$Tiu}uqUE#U}QQ!9MsPQWsz<!c>*jniFGD>YyIWdjN>{n=RkS=oZt(3vW{GyV&A
zlNpXL7y;|PeDFyDONH=oqN1_k2pDKhAO}Y=eeyez-G~7Xbd{DHo>pv+u}l+BK5_kM
zb}I=n%E~Fu4HZ*%_?FZdT+5cQI{7NiAjq9s6%e{I+kp2biTPbGup6<LHwerRoa)u+
z$~ig3!gFa=ta7rpYinv4)yPY0dG_coW;i|AB&ZN|LCit|;A59n7>AJM9E-K3M&Qx?
zTVPCHry14YG!++H`GKFC5JsMxCL&ewVR>XPochGgsqd_VA7UIzJoR(9;@>cp70Igr
z?T|ga>5r=}U5^N^hQZX+&_#I}`&(#^9PoTGEE3x-?g{wWXm2~nk<S<d8;z}&$x0M%
zfY+S?rln>JA$<Gcn^RJo7}P*PA#(}!Uv4%OXgsCJ4o@+uw6wI)xklwhieFoZM5sa-
z?>77b+`}W!0XcgPennlhwYWh(FRzGDMUR2^UeCZ;uX8?LU`@diIt<-m$#0)5VjOnf
z%pDpNuAM&qA4=hw#`VDUfJIdU0*WPzd!u7~%~)aif6TY18$&p2LRVd`0^2bTrfS=&
z&!C8CP8+PH?eT%?8f?cl;Qhi2Ex=NCY|hd=L-Eb#+3e+wa>11o)iM8%tm}+wGV8i1
zD&r_JiV9N185KmOHvu7zj4()*s<fymNDaM*peT&e6hwhgL=dEg9v~1v6qMdW3!#?;
z5?T^S2>EWD_P+1OTCkRD`Qy3IbMLw5?6dbito5uAIH7_D+AfX$lr@D-I!3e@-7#!i
zNd_s!`u*?OU@?Z=M;t7u4BT*lJ&`$p`yv1>q#)Q#Ar*HUa~!<fKKJwpEyvxi;U-8j
zl18o2U!77uO7a@y({pw#B({9*q5f^KP+{kraq_eyx(&;9%;%6w&Sf@SOg1B{`In)X
zv1k9yomhVn0hhRqlWz$y(pi(f2Ju}=+G4ML%qklO^L(BFf&*Zri;HH#9>Zi_!M=BN
zXO)^wL-38bgTFQQHd@x`0*JQ!G3A?rU;jusZGA7nilq$#NyZ>2$55x#nets-4VwVx
z5Edj85i3mFWj3%enmbc`d|RcbK{(p=rQ(CU7PK`m_VusHL)kgHLY6q=wbxLX$K`&w
z%N<Oi``b>NL0S}qFyXhdqJB<kR^1J{jN}W|__v6DCsN5E=<MNYT-2s<nTfMqnss>X
zeV%4-oc9LT@|4mS755DAIa(qbGY$M2fvj4y3D?_*+sb)8KQ^&?)TE_1r`Uqi<&x01
z{%9nnW}{<}hO(M51m+x@6JG67c0Ht)dVNxnXsZ~aZg&I1Hf?_+CH#6uRM9E67_ucw
zY|HlT2IzhyRc4W3LHOV%s;=m}d@jb*q-}FU<_ft;L?%>)_|1Smm*MUF$E}3u*mz%M
zm2}U~5$S_2r-?yZfhw!M5=^z4Mx5{iavVHHj?6c|inikfjK6>GzCLBn5AP0KtmQrj
zzb#Ku><fsYHLCb$g6b>4`b-~3u|Ob7r^OxJcm*?=Bt4662}m%qM7#IawHypukZ#{Z
zM;$51d$$Hk#SL+qCSx(Hc6%OKk5*S#k7QI#&RmJbG6qpZ?+oU+amEYs_~wgU@;3-^
z&CQ+5UD1F1_`^cv63uAW`(y5>gH3{56T=Ug`O!7ftkDQk6=K)1`=PSCf3qF^{0edr
z$p=sVHw|q2^e0{8t7uqTbV?6BHL@6aKG{fUNh>~w*W$k<nSj9ilS&Fd9oW1;03wN`
zsfjB#p*`R}FDS?~^~#z8gWRs$?@=qhb@PfAMPtQ(s8`Z=nG_ta%J@4_IgJJjS1vu}
zH|#n8*5-XPAH=3|_6@awy5#8I&rh;+(%(=taC@}ATV~H3OK1y9fNx8r2_?YGf~{g}
z3TNnaTiiNS#XWxDW$>r=ttW&X)BCg$kBX4qf5^S8>wnS+YB(q6C?kENhnjFc%`Lm(
zZAAr*qohd$<O)^@tR<~sMF+6{Rg#Xk7M-egb4TcCOY)COIJ^LGUu7<adYyDq$~`?F
zF~0Icnnth;Z>0fJdFhR{dv9CfB`T7zqfNr0#E7QXWn!;P|A6<!+1~tMl79n~({Y-9
z3GN{qcif!D-XFA=a0$T+!e00yvg^BNUa@T;%hl|jH>HRw4GoAhtCvDv+?wHDsTs!o
zS1>F43vnWHg4t8-f~F5EFn;$|UuO8xZ433)n29c$#_Gz$Q6TMJT(OtfqLsPlHR5Kc
z@r}m`#*;|4gQ9&AfbrhH7Lg&dcH*;?K4j`@J!H~Wg0`k2eOXi-3Ckq+n;K~-Jy>{x
zwJv+`ybHn(St;+<90iz0h0l`7;NW21P+f1SDIs_%6uz*{PLOx%bWJD1yQ(wz=DZ*c
zePwpb?GVy{AVu`Pn)zD>++n4K;>O%GBDWWgnaJcxA7{+lLs6UJ9K9Krwz7T4op>5<
zbuD9P)J+S1j4N#P6Hb9kQ0*@WSdA<3)LJH`hR+s}l#TG>VOo@RJvstND;NCnlMLlD
zgyy>Moa1W-<mAR%Vsw@i-VNlq?`LwP1|$Sd=Qrmpzt3F%%x|<fkAw}ys7!nWwbP@o
zUEG8H0|T@B8zM#hk}L{uXL9phHI6zd6+d$&|AfD5&>i;bfK-rc|0QQv5mX+peS11t
zZ&Cg7eZt8Euh3Ce$ECNF>e+mAP}q`wq|`oIZa3qgPj{axwAlQ-R4=4?(#i#TikJZG
zh)!bD_cZqWMV;%|^ReTPBk*qcBBS0rAw<R_82k9g|4@1SS2}f%DSAX;@6@>4xKH1=
z2)AeOa{8r!U2(d>VtbLbnWG6~bNU3M@d>xmaX<2S9)z`IR7)Fm0sGD031n1SUZ({`
zycfEt_X+l~s@giI1G80D{CBq2^7n)5RD+S5>;4K51wI6(q?l0a4>SOx?KQu`Z6QaS
z{y|Yh9XnSY8`#+%-!flpz<fdoS^N`piN5>v7FXD-Fq9*gtj}MaF;A|G3iq`L-eR>A
zqD+QLyr!g+_R4-N_mwZ`E7w$Le=DCq7YXyivmT?E8XtfL&)mHare~CE=ZeqK67L>F
z(=~5_w(G!M>-wNd_u3_KTi-$Rpykrm1FS8<APl*-;EfEl7=UC$!JZQABbBZGh5dZc
zAY(}2nkJ>(;eP))C7jZ^0}W(Hn>f0vrhEoTCwduDt8vx^(Gxm@SE<^arCXmHTu2xC
zo1hSOG*<RUF4<$N)Wcj6NB-=u%<^7)sQq9un5=~}iXVI?&v=*3%?M&1Q-DSKP72o)
zbHqR@NwYNNCR{^<dK2!|TB1|ai5)b`JKD|(pHr-_EoS5KPP3s^E?o23CI=rnVQG^I
zWixPA7{8S>VxpDKZLcM*O7A~2)S#NlFz{eO5t|{Lm;iSfv<(Vr4f6H#cXr<}d*~{@
zx;-08uC^ZUd5h>fbyD6d6?2YNF&EshtJNw-h9wM4Muo6SVVTR~-;)CElb<edHU8fC
zpC?BdoE*>ZCnqC*P*<Ljl#QOVLjRgPZCwNaS-~kqW!29y5DN-048v7^00d{1_-Cx^
z{vBf6{*LlN4uBIBM8VOn+Hq3WyR|d1>YL>mEd2&1o-!~xs`jz9WmVp>OC-&!O)`j+
zcqbTX-T_IuP?>j)f4d7~ai#&sFKpLS@w{BIMOSi#dDUBj6<g@IeY7LER?#-TUb^0$
zWMtJqC$(Zu+opZfNCU0}3<+k?9~_!M*sJjW_jw+0KF|Hd;IP68nJoX`Q<Me&M4(46
z?#Dg`mk*lFv5PpTzZO$0szHA(XIE~qCAy)!?dd_cC5*i16?J1Ex0xip;L`=3l5~|J
zzF2Em{otVxHXXtC?}>f>>fZ8jsoLyIEgz5sk)jp4dO%XAc5Y$HB10^A*`JYN?w^<G
z^du+FiME%p<LE9vzj0dWqI$%(lIpyGMgh(aJ#z?Roi;L*(q@|=$UFuFo?KjKdF~zO
zx_fStbc*Y+=wEF~A#j&|?4gfG7H7z~$h?*l(6ch>a+_Xgx45(_g?RDZIPh=7uNv-Q
z!1`-~yI!9p*?BcQ)PE>pa<Z7pgPodL$lWS-$G@+M;osaouR~fM33e{Iy4uLi*rhk1
zT#9QPIBU|FqZgKU_3an1d6yNd=KOotlmzN(As-|(VE|uzYR2(nmmo#vX7crA(t)kt
z8kD+OJXc)A%+)q#Vj<&f%p0*?zk+l0Trt-RG^w4-&w>@cx5n>2I1{~>D|%mu26tHI
zUm74?aIYTaXQVTGs51|I(}8M_zUZvKR`jh=g&tA#X&*<HYSVE0)B+kqkWs?|PMfWX
zwp*fn?2vE8BX3KLVOq^?=ol%W(@6x?w2EukHHUyq(|0b>G9QBpHjaJXYfm1pQQMwP
zsGp=~Cm0HCBJ=~nCWX&0L6&lGr_Htcf9rhYYGEY40^faTPk6PK3VHt!@}RG6QG9I3
z^Ul}_^mzX+lR0{B@aT7D-dVB#|7A{rFZ0LuFXR8f_kFbH#igS77PDg<wMomq3f(pL
zZ(l*PgM*m^(Bf(HZqbM&R$1i>Zg&2uLqYlVC1GzP5&MBxhA!sq4xO-+k47?@CX^&h
z+=1pxsJdx`Rn|A7(Dy?_L%*tuYokFzU_}X~V|;hF7hI-4KMxODrql;y?5^cm5A@?R
z@66-|X9S-@uKn2iYPAkBU1L;i0m%I_3P^vOK~5o8M=BKv8qXacH+9pbb?RUz`cYBe
zo8W>IKV*TOw+GtIb$9Wv>Zgs%tflws<eT@7H`T-?*QKJ>FPDD!sYLuL>NDY&0JngQ
z>X{bNEKd0|zWX6(<v{Gr0c_@16@=fb!1Twy;W#F)6V{o5VO9wrsZ9vP|I#TewX`DW
zZ&U`WVHUd)@|0H8H#7IQQzM%RN*7*HXe(pXN)+7Sl0{*Ue*@F@7wF*>5XSYy7(J1j
zvHo>h4Lfr(mEeB*Rh^dWr*4DLAls`2Fn%c-D>bynjG8+V%_Qg>Ln?6L%5U(6D-QA5
z{Hc=sh9XQXTG}Va=%UR`upIUuto(bPeYT^7@B7s{Ql@rCyHz*k>egda+Yce$d4&R#
zZ>-co)5>U!x25_bkcR;aiQ5=7(1h4)u6Al;v<(3dfL0pC7qVfBz!zkUmfkWIO7U6R
zbzBq<&UkmnK1HdgxNiNXXz11y<C6gPU8M`|$ktkzEZuu%$Qnik0`E^006b__qS1V9
zrEd+A@_6F5OR$yfWa8q|KuVhMte_5=O^nuBwI+>A7~1<K`h?)!y<<h_<OIj0TC-jJ
z3E&({F>a0{{#}Uwxu%wq@99F2jz#O&MKfK!Pq^0EI_{sgBKD4I8m&rUPo%a9<0Yh9
zsD$SxK&+1cdYrJBvTMB^e%Jp6O87@1mG#d3*xj>{-8ekzQo>OFP^X&x&@t9piT_ge
zDU8!QjrFf*)n=j<HYey<%Wh5Pa+nz3g;sDc<rsyoC#bPD`+Y$Q%RBm%0Jic?y4c21
zskEz>F6bh2?d0X(++nq-8C$=jJ41qW(P=DF#5*in6?^ZCCUZ!vLZ?1E{g~;0&;dCs
z3OoAXb|kiJ)LI*s=u@3-hRE3xm$79uzB*X*+%!b5NHq0{u-No`!pKwa_WjsiQr1{_
z|Kq4B575mJ=>!KY%Fo+0#5Vbp4`FiOUG|UUfu5msX~5b`)2zN*6I#nYm10K7ep~@(
z`oYQG8U06fL^6@?wNBUK_HXKlC-knM3B`wG>Wm%J19RN5+w;Gwl7CKEAg;|oiut5f
ztw?B;WpJ$oT~%uETL6FPhH33B?^0NKwCcEGjB4#t$Ju#R{?@s(b`s+Ms+Kvcl6`V`
z?9X_9KFj-j+L^pk)MFSM6s`JSrH*ci8?D9E99MTj3*sZfZDdvrK!fyPIJb+A1UC=m
zj^e6spef{G*<)19!Pl&*65j$B9l%E1bY?2U9hbf_i0Q3P?^yFH4ZS}{ioAz~%m@>S
zM4{E0J&BBV=<fTCE_EYHV0UnZ*Ie5kNV<9xU4ogBN0+k8EE*&9%y%inuKQP=PbAho
z$A88uNiY-my%S3%hQvy?o|Ymu-XzD@dgCG_7$=`VbH3kKoaKD~X(p8L!q1DIUlP&v
z<^iXKd=zc8RsDe*`UP;JbunM;znsnj2~o1Vi0a$BlcbDkB;@(spO&B9pU5_%$LL>J
z<<-W1iE%$(m+zZU>t+QRW9!3xaN$ZXIp}g+=w_B0bGf|Ic`$#E#z|v00GQVj^9$<Y
z2&hnM!Gsy{vERrZ$EAiNHO3mr6+`6lkFOIhh8Q~LG98%4PY6N@^31j?xiNQ2T^nla
z=hZ=(1nE=}qn(ugmuvUuD#l6<e-8@W=hLK}ob<b_sf{YfdRm?p+l2`uNULgJ^6EUB
zTtlLf4g71Fo*lKN{g1Q2L3#muQ~)wO_&2Z?jL`&pZyai{UmfQ%@-&&znJGYwz7zcD
zjIDR4lG<`f^X7~gJK$Oz_u&doDX0f!#2?YvG-fU8K)`o`-|xK_v{fDVh@$IVH;hYL
zCbc2Rn8yw3cmbtfm?ESL395~XDrr4Cz2$F)I+3?51K#cr|EHJzeGz9rmmK=O2-SG$
ziauTP>M#IA#3cdJXxPoWk^rKCBSAc}dL*XwF^p{ZIYUSGuymp5ZG~$0tw-Cx<HQbQ
zL}aMmX9?B652L00t<jg;zY?FP3HvmSKV;0@6L6x;F9}^hHSE?-Fem6lTT0MdyU^{q
zJ?ckfhtm<ef$(K}u?<Z?w%zRAqcOD0@31sAqnR^eFR>I;<WlG6K8{43_Q++{^4Hqx
z(~I_u_<vHo6YR<=dnOSZ*xqD1p4n5Adl#`RoV5C9ne^xW0)Ovr3HfEB5|04*E#Bco
zm6`7g7y*dc<l^^2t>wYu#ySblT_0IHObRO`;t(Nhw+&__rF`t!+%dxa!y>9<lQT$p
z?V=@V4k5=7S@RnKO0(e2ZkK7Wi4+U11_8TmkH!(%*W+=N6PGL&U7*aAS`3C~`HF$z
zO+FFk(W0bs^pG1gw<?`{Z5$+5U<n^Cl)_>>QtRC;yI=7}lIycWze!Uky6+sHu|v%M
zPv7~lck{(fjfG57+)E#9HoVL_{}e}aGbV9t9w;GB00oW2%A52mYoOc!T^84?B9K$n
z9^KaggHs57%zI;EhvzrNtFe4fNzuJ2v7%PEk$joUs@e@#Rz-MUgEmG#ax!@Bi-60@
zXeGCw$UBRQ%F3lf(5)x2>f|f<kd01mztq4OejoZRu^s`MhWVJ4vXi}IlmNYUogafo
z<3DB$qfZXM?n@3g@y^W{-C=h@iTbBt_V)>m-)q%>J_1*Bf5|}(HpOrvU((Iht<scS
zH|PTztb>pA^w$jR)E$IMV)z`IkD}OypCeo3NWeo7m|6^{{Iune>!%Goqkf#aHk`)T
z4k0;kD19<~YEX6sX6zncG=aTKSDXIhtWT#B?(nc&n3Zq027R$^qpKgJcfM`_?rx0T
z7}SWvu|<M1YlGA(b9`NZ6!gP`Kj1FE(sYA!q76~j@|0@1Rx;soUcbPW{NR6G`GP-3
z+*l0m(!#|hR!aO%`|VrpucnD0>OQRjHo3FFA4+d7qPodLCGZ98%wg$Z{PfUHO=S4g
z@>ok0J#_*UJdf_vja1g*HEu{hr_^^O@Z>uB$+?ev@c1H`(6uPe^+#>HggX3HE^O*u
zR+d%<duPEH6m}%Q0yteV6Y{9Z3m-bbqga}G%f(7XaxVKvTwn$hyKv}m4c^V4v6bT;
ztNC%N@w`EN&~;Pz#=ZZbKy%)I*gfalPqFbD#@dd*k7m8$h6)Q$JE*)#odllANGK-S
z<BDt(4*EkR&A$L;M)nOYgG~ark!}<;138PiyL^td{RjO~@9;OFAB-TZimt00`nF!*
z&U2S(%8F=E2TX&QV}xQDf;r~w;v)RgUIgNn34BChl_i0-L+-ZxG4L35?$lxb-%q4x
z_w#&kFi>?{#u(YTniv}0n}8|OsIvd_w(XojI|%IeoNuF9=R_s)O3OH6+i}fTvgiks
z)^#^LF2Zakg=0AS#O8Y*z?yn7<yr8_lu6<AX_9t*v1`6%2*v|@5bZyBN~d6~$~iir
z?*vckhc#b&#h;2Bc7r&TBm;!fD`!aD%_FtGq|M~+5N%F|OkHo6;rl8#uNaF<yCgm2
zo-pscQl(O7H^9nhpM8S3PxyJB^*KUeFh@ZG2l}>;X7VCq5(ef#g)H+d?>{K`oR{9X
zdaFN^qRMBBJL3}fQqFx=eGLER_*M`u(MC%Aos`@cmB5urz-Iah{Z<lNN{akt)@}%t
zTaP8Nk6A3{T&=8ibVyWd=ZcGc)to!Ua4iJv9`T{LUS6ICtwXJbeb=^@9U%pWq^qw1
zj)CY)>f=|_5yI|0nUSaS0fOdtZ774Fc*`ht3$5lgblM8-9qua4M~!5I%eiqB|Jp0o
zNsW@kBGSi-<mem1fW5zS{a-Ye^ZbYTl|&hB&Up{I;(Y?xxgg4n(g4o?J;ylcSuWgs
z&kYd8&))O6J}Dd`IdSpoGZjDfxS?S&WnBy%K$r=C1a|xfc!9WzB9Si=tH-0iSxd_2
zv>grAR}|qSYeGI_wDI>!&lzpokhL#%XByy4T0ROgl-J*urTNa(|4xptwcXp92Mw+P
z5XEcD>YFd16|tTUM)0xWLva}r_Cdh6AmF{s<A2zkoTrwJ9&r7BR69$CPCxv>Uuf<P
zT%ye-SIu1~g;P_JSKZJc9X~dBpzFNT6BNZFlhvH*xputl(uZitj&*8ln{J$?=DIDu
z_71OMn){Wly|6aNN<5s7R|8l$H4?IuNp#^pYpbIZDex8Q(|aRmd+sgItROr1>;6&4
z7IKjN21&k#y=_m};rr?C57kD~vB{E1)er7#(maZ3!4*s;1vc69I?DEcI(`w)Q$YFv
zaQk`vmeb<NS?4;Jpum9crtm>;{+xX}*C!jzCLZk*Np2CoeVKgq*iZ}yd^lB|2z!EJ
zr!3ZvphXhk{*F7Q;Mnp_c+-`&yQ^G###;{7F1<X&p_iJEaS$R?L)C|kdvDyW*kVx-
zC=NHOfwclF!1^Va=JLdD|6!Pd&X^c4HFO#QWqN+XTIfqIt<P|s2g}#{jT_!CZ8pD+
zQ4MOy88?F_$PWI8$eZ)x1rIpOZ&sedeEz+)ZP6GdMgpS#<d?z<X*i@i@+V-VS^n*3
zcaI=<&uNsD>KF95Z@A2ttU2z7*4mTS)rsG*<%iInF&n%YMpI>;+_FockN#-;b4~4B
zNmPZ-hNAhHqvpn`dJ3Fk8XO!vh&!O21xQi>32>R3+4y{vhO&7sM08svXsnJ^d5<LI
zJEngx9z_q|Q!|XFFF2do(sbd4M3YukQ*ja1`KR2u$vJY{E@@Bym059?a4n{wriZ@F
zgL0|p+m&~CmL0gg0)%;<rcT_<*GISVAN*k*?Q7pHB|vZC!Dfz8j~xMN^_tC2@Nave
zFk9&H)Uo;7{iM%#5wpmy2b<$v20>#T8k+^-6UCCeWRI^zCpWjjF~Y7NoD(mJwOrY6
z-kgN5ML`EzHED9-mZ~I@2tLD@P4X8Vr+%8sz$>z({(5IC*0&{%s<*7ZSoa!t=W?rR
zg*#gHknkHBr>_Q7--e;wPyHdu#DA<(I13rTrW}GlTsf_+_FEr$0{QSa&~)5K=mOT8
zUsZ(mGIQ7GKaGAZ{j}sURmTrJ4Ra3bbe&f_b;YB5Z>c9>BzoX~A5FbfTV|@<{A($@
zw2kb7f8G`py5PWRL)iNF1PWPNqMVIH;A3MGIZzhS(%e?yS`SFI&!!=Ew?|9ubp~-*
zuk9MfIN%(;d1K(*ot<b93_!#cc;KnEIte&Q6HmD(9<7nRHlVejp3sT}9pa2dQU>vV
zGM|Bxl#8WZp{unc@v#YVkycJ#HKrk1)={v)HMdX4p^3nv%TD6wmQMFSAaTDxb!|_Z
zCXIn~LAwi>gM*YhV5)fSy*jBr5vXPnF=utUm&ZSDERc&bmi=q<-2iV;ZSJ+H&qA8N
z1>BA67Mc)>F}%9=8;9p)4rRIor@iy1_&u$1g$e|^dluXR!GH1@)&c98YEPt1P$$8i
zn?e*LlcRz+-u@SzVW4C*zoZb~FXa6?k#@P%bQxSwJ`=zq#8ck=)X(m~#2nORe>k_>
z;2M-5kB1>guk3&+0P~je%FH8%N5C+b2S&w)v39Ow?*V<OpO2<tFz82s)0buwj84Ct
z{OMX5hm}PMXjS)V74<iRuUz80YAw+@W<;vOk`13pz0Ml>hBdjUa-nLI&;4C~bjyXg
zWk;fEyYP`w`(lL}Rd$7#O6S|WKUvaujZfK&pR!+;4Ng`Zh!f^)9?&X5rGP{Yun(S}
zTDq&t)N?LWTm?1(Q;jN!Zw-<G36DfT-T(?R8{hzlF`pa{UpRFoT?Qy??)>gwpBdZR
z*acA<MjRv0cD@~}YLnxnnyL9l(>sHx&<~*FV~zrZD|6ib9<<XaXd>7S;&$=(1eSV6
zf|okoVAX)5FpLZ6mLOZ)C;S>Ons@a^%N=q@dMzQzo-gTw<O*_DIQAcb*Z1cg_?Ys;
z_vbxuCpcAZK0p_vP5<KncBCj1&XHeO@2i@-K3>N0kvDdJ0$T$4no{$UmnVA*7-IM~
zJC(znQHk<3UscpQRoo#^vt@&EpG_~Og@OB{KB-9CY<!EsEhSsOk%!;&9JjG5b6;C7
zcz=I?mOpkQjnU<gH68R7PJ2y;%r~pKH2oUn%~U%@;QR3b(ll=ZZYcj58{XDyVhZ)Q
z>95hDST?Z9F><A_b6XGd|Lf*5ws!tyGqK{Dyu?kl3ju~ri&fQ1$9sbK{q!hVW!J~S
z{(&^%-T>;rJ^bWKl>O7F$6mme3!fVh@yY||KOXYs#E}=xrs4HYNbS+#LE@JtOBslz
zUTe3p5}XoD=m#!=yZOZj-?7IRUKt@6{fwOj2$Wh0ySjN@DpgSjR^j@i{{C)A+@RzZ
z_YZ>CCc<4)W|P8Go%oM9UG?1dK~rfr>ZiWGUHXSz&w2INc!kz`=!HBe?*}V;<UzC<
zSRP6WQ<j{Fk!*Rt;m(<Gf$qKmvKyV;!}OC!bgEkPcMQ@eIro7GE-mHM1g_!dhIrxa
zs%+5i96QN+)PQ!%xTu5Wv@)pv2)e(yE&nyDCWO76!5%@bTN^ZJEZa^b^tE)IgjogH
z?W(rnTOJUarZw6yUy1J+EWfex-LOw)P{gco`d_`W?>mI^=Uv3%z{N`i+~>Sgv(J|%
z_#eXTCtNga;w9TBNbmo}lAxA3)&dBI#bOE%ulPt$9mE1*1A@G(5G?{x3G7E?Pb%JJ
zpPO7BipJmtQ3GvHa}EeVAF>zxY*v;|*N;{iK<sGBX`0^8^OvHaJJVtMhxZ$6uo)<J
zejKQ4&27C$woWm=H7yjBKTJ^eugdSzJ+DtYPZRYRIZC0HS<UO8MtrXJPp|lztk%mq
zb`LlWi~RCm)j4>6r^Vo*a9m7+Pjqx_;__$U_6=O-4teNN_(vS<hn^=_kx-6V3rE2M
z?x~d^Iz4Ka`MeX}&jT;iN!F#`T3&rs#l%W}_QslX(=`x&+D=FyxKv)8!JICn%Um%Y
z+)Sf{=)AS@(ly8Jla&nxlt|)PvF+c@aXXVc+|+h^&G)hT=qjJ8lR44M4}wDK0V16V
zGa`2Ag*q1TEtzOYT{&W~No}D0Pd@*VS4+JYH~LJT_QNHNBLcKELU*ly+e=9LN#*Jn
zoWEvoUM+TP^C5m`=$Z#fSV}}?O+{>rVLx^M9a7`2;;K~foFLSBSzSlCoY1eXOw__i
z<8U;!Uk5j-D1>ds?q+U5t$5gDDulfOVNCKsKGJBk{G!0!)Ou<HtJ}o7-N839y{;UR
z^fCsXOjK_y<10Dax}3!rc+9IKh5WfyGDLNCf5X&&6?|M-svJU^i-h%vePKn#=J)WO
z(k%WAU>{9kIJUO*x!6K61y6K`M|FvkhjbHlZ|epSL6JfpyFZ_C*m-C{=6ZZ>80~OV
zTc*mj7m#X~8s4PkJ2Qxu4f)-O-3_p>8z;9$x<kLipKTAu9bXf-CZBNo(%d7RW@x@Q
zRr4Bsw->UO`7Yd@9Qj84w%qSA8PDC4?%$yZg{@@F3;zHX^Iw;K^mGsICQo}(|9Po(
zw?oTQUy3sKUKZNRBQbqIYaIB;0jwZ=ta;62JY{zbn93s_iKHD7F+{>&c}~jxP(0K2
zb^fB%402M~J9H?BvfOFwW+xc3w8t?&1q*US5&)Y_INPFHx_;9HzSS&NOibZG2%V|2
zwl3Bxmma={dYNcly&Y=y6(Q%*Vdr2o<7po|hA2ISTMU)_y7dbbu}u19{l92#&W$gR
zTR3V4xZNC*e(RH+9pO<&`hu=MfK^srtvms;m6ytDqSe*TfMet}NkV)}yAI$olM5pL
zhz*(T8l?Lyl5oi@w1!Cp5-X|~kYu)M=05sXotOUx5Pe%29gdo&l~k_7D-Q@pA=CWk
z)-cz6y9Tw{>v;vX>OpF=TM?h<b=rnn_ceZ$r#Ja^99Eq-b{u;BB`;86JQ6EGr-uzc
z$4dVXTB+sIL&{eWs~Zy4z+^l)Qexr_XiqdD_J{kQQyPr@CP{8TJga+rP5G)5db9Rb
z(L-P!br1_&A}2jhAZ%P77OuUx%sxK-DT;!vEgAybd9gpra`C+_&AieEP90<bD9?)X
z%=KbBlROMi;8|BvV^idz&=I!&it9*WdQU=#CM^a@i`w(D6qD2v@zA~-;*hD}nVMh}
zW}m}>SHCyT{)u{b7Z2aKqlCK&E7vCfMz}u@(xV8+1qLYR!iSez07V?+QHVxl!!OdS
zG1%EIN%=1-K*aGB-COhdvF|T9?YLyVC-pzJZ#iyh=Wf*k-ot{G^3v9+s7(N;m{yJl
z%)|r{m=l)ZnFnHFpr~V@5lnP`5&a8j6n5D14MnWLwBJvJI1vq*G213#&Sie7^2y;s
z7MInl*0RILd&&43mt(pAjR*_g$V658Wk##&mPM^UhH<<oS9v3<zxDeuPi!%s;I|H9
zu{+8h<X%~ve+KuIo9pCQyV`D}3cF3TZ1K(Q<fn}|bIKJfP{WlABYpAf8)j1Uaql?I
z)yI4vG3tar_6R!dE@kZMa;OXiRTE0>-GiHJ{gqX|<-}Aoz{AZjj@|Q0G$Yl-))|uP
z+3_mxS$WRy#KT|e|EFlvc3BK=3CAhl9{CWDFYuW}A}&=F4cDAh{w-(ZWO3W+Ga}J*
z{G+pt=xWP=PLD|!J^cx(V&NkpaQ=X^albzQ)`Wcq6jzDB?draRzuG(i=bUd2k|M%o
zdY;zG<wq=}2OEOlVjz_;Nd)h=I1aHh!=YXjlno7=)$f)~v*JBB9mbSlqtzMA)Oq);
z3!>LXWRAyFVS=4lea|pBsfiD_Ow=G|$$w>gf9;CK3%u~|fYs60mC0J?MV*SoBT;c(
z7+#+8Ov#$e+bUoSTs6n}BWHaRfR?%n*bPeQFLUg0y~ndlt~3~mql;EAk?rjr&|PKD
z2L3KuI>=(RRgsmEs|JIsfW5QSq0l|Rpppv8PKS-o8?Nkr)zN&<X~UV{fh}z<F`AE3
z%xf{6BUlGgv4!GgK1ta@f36y+MQ!lBTE9fmqutMUf^+9PJ-i?)@fnMH|60#-77$(4
zgB$J8ckwNIV7Ms;96(X*w!5n*1bR*$Ii^0MoYAYDvjuD%jMk{#yPZa7Xc_lrw<bcH
z&VqJArn(?+6tHKGYW-R~^UU*E(v@ATkZO=$t2W$k?MG1!-^clxA3od=fQFP>6j_;r
zmwuQ}-EX-^ov>|KWLIB%kyP*7LS~GHIzDRcin$qi%!D3zc}_F@DD#L2J?f=g!fP+X
zk7M?`N2z-LzPSuPiGA2-Jvc$F|5?Crk4n}JV9cglUWe8*Sfk#ktOrs<%SqB9KXC2u
zC~dbXOLxTZhD51rU!R&7I=HW?hW{{Nu6ss}bX}Ad?_%G9nFIwbXcr~bUV{$L2%5J?
zYHt3F4`r^HTk|!pmb&BB(t^n~cB@d{4C}h<k=-LYBy@e=>x36EaJ-}XeZu4W<GtVX
z3~Y%nBe+i*y_Iuf8_yYUA~AdZbtvY%Y1v`oftfS81O)Z!H|e)>n26kslkCgxFF4HW
zU)b6mD=gE1cHno5mFY0J(!E$arI9eSG*^w~o$_6WMIt)klZzHNPwAkl+;Gb+p0&ir
za&sx@cAKRS5W-lvX>c`n11d)z2f=D!vnT~+6YSGFep2-;rqx*O4NNq$Fg=<%bNCKf
zA&nU_WFIwVd}CYe!V>w86`J7AjDas4v>rtSI!3LK*NoPDL)K9K1wU&7A=95f#w@$L
z0=iM>I^QZ<_n8seE5=>uXzti<7CQX77QBxu?3Np5w)5t=WS&;v{CK-KVNT_Iz4+33
zU6*sz!&F`FYn_qUT!@E3c%|C!vEHeP^A2e3=3!v6ISkM`T&pX(KC*c?Rk^O=zQk4w
zA@15HvczzCx5={idHqm+*8>Fta9p1*9p8Sd9;{YbpR3iNz6l@w;PfPailcnMg5zNM
zP>QR0tW$+h*uZ8s;1D~l2(!AO)RwiN#p*aG)Yz`6u)39^Z^xB~A&|zYUA<dvf**>g
ztUtK(#hJG$M18^FPf7e>oC;}bfjzphLs}Zw$`Jhp8uE>&VJMC0!aBSBcUgL^2r2CZ
z(GAwKy6E(rN%+C*73JMw9&WC8H`gH^@--c-V6RfjrFaY7Agexw_|tDIJh{60x$>PG
z*?sOsX%AzL$oEafyIpjp+x{_h{R6OC6W)r*U8gCn8mL8j5RMDrX;p^6kuiM>nXyS~
zaWB??ZeN2bbjD}H$iDG!jJy!TWnT1&V3g+it2i^T>a)$9Rz27ScASLm_L>zYF0Js=
z5jCg)9-dcA6gG#YHhSN^*%I8+ZkhY>7a6<~5Pq~~=NJW5grfF$)2^}^G@C^v$b!;|
z$cFQOfAWt8WJChw_r9a+yTbMUswyuJfv-H$vXT<4s_cWe(v7PQG$YU5(_PW>N#~NC
z&7>5!$OYf@i=g{Ww)FuqH?TLbR=FywQl@cyACn$cY<>R)>=c;xG<q!4V@lc)t=jT6
zZ4SAHI#nAN`^wIVkORnneGoD^wA-Sls!AEYo{%x{M#?%L#JyKDT4IcsZ=`MW&&Gb-
z*`BxcruTM&n||~y8JGIJ{Gl67N9e~=lrWo7l{vaxEbpM({Bv;EV%<XCn~oN@-9}tQ
z4kIY(p9Jrkn`}#T=QF+5g-;W6@hyj`fqS@Q4|jCdr@jitC<yFMNGu)E^{?F?_$lKo
zSHSp7$Huyq-2z8lKdA1Zjmnqnubdx@x&sWCR^=Yh+1UFdQ(%7KfZYqyNtkJ!G+K&Q
z4aUNFv`%|oe}xJzw1GVQu|Z1%?a)y={+8wwSLIhpLI!7iboPitH0>8y<Oq<=hqz_e
ztGkZ^Aw@x(m8EDvtRM2KB~sl_Tb!N}vTU7r_UEt6C-)z$Ob@tw)11_1{>a11SZ~Yx
z8Y@HVZ20?wh5dX#8cf&QIB4Hd6tApTd#4XFo3nLndoNcH#NG?4$NrgKImi+sg%+`1
zGY>^YGG@7=-MLRPuq%%XNQIr36Fp-~`tHN*IZhEEKz?ad7OvAq#YHqtIn_=D42ccS
zzoDh+$xiZ$bPE9Ix(--OlzVfj$2|(Lx4JFCK33a6RwR6KJ9b*#UPNQ83>dU++?L9?
zVm;{b?fl!t9He`Gn-!OK4i1X26|nWWWj=KCR92kSRkJiH;kheKG6vggW#o;YOJRv~
zI-he>A9hveWC;P3<yZWtJLC#BXiWo;k98xeXxYZ!?BKh{J=q2mU=)qqymd>GdAx-l
zLq=CZ&sl`H!3}7g&K>X~R`)p0j_$IGJ}24{cA4v}_-5shVPXd;ugxsLv?*srFjt9l
zRp@~QFJYR@`9XgPcAig+@TS!8%xWd@@9JAy_ca&yD4bj&fqc!+z2{)R0MR=t&O*rs
z$aoliZ?R^Hm0Z!T9K=hLUej$3W*T=wu5EzbJo=_{!W)^v40#(r1#9DOx|EkTlCLK^
zub)=^(S7NY99Bll^FQIn#xt@lmz7#VZ2O|(Pru%Dnu~PZ!xh$4b%kEJ)ijQCpu24L
zWok-tWi@j}U{j>BG#)DD3IQ(qI?PASV<!dJ1iVSAv$jzY*_!XJ4H#b>`SSD&_F4EE
z3TAy_m6}nGtN`-|#_iuql|(E1Rp#dv|N54U$jQG)B@Qwo5grmf+}8rJYca4z4k_jf
zNgv2+c_VEuJnRDk)T7^m@Mx2kd-;^c<lG3xGb;0P>U&1*PB<lF9@${?AMMHfc0q-2
z-?w3Ne0uz*O#I0soD&un_N)fgOU60VU1$>L41|aYdu6?Ne_sq9Wi@<EsEjJ=xPD3L
z0a`nE?SSx*@mJxbP~mhyTME%ya}99O|4Aeni#!yvfj($NF!(VF=sqELWslQ}Y`vUI
zFKt~SvL3@i6D#C{D&d&W>Se{1zAL{b63OJDX99K~IhxDqO5<P-85IpM`V=7NHqQ7_
z`yC&AQOl|!6#j)IqUcnGEn7ANRl)J(QzYJWU#R}-_EGcuZdtm?coIFe;p4sSV|2@x
z+59;S2I21w$Yp*vTJ^J_ciuTVWN7y-GlL$%NCiO7oi__uKMb@PkDBA*qaNHN7<k~^
zX6J)hZ7exq`2IIFPyP_#Kq&k}_(o<COslW<8JJ_daKLm7<l;@h>tP3Eh9Uaw;{@RT
zP0mOjKR~w}V0Ql!65$)m*@1n@<+Oo7Z(4C1bLA3r`^D+N*;o06&qigv=k;wpn)Wj|
z{Rnk(wU(ug=QG*_Momm6j4)d1LhX$Ut+?m{_x#p7YCn3tex&wBUee=I-<msM+>`li
zXr_X*N!uL<aU}|5ev8{7c|K={a1>yE|0e415Yck6a`Phg9UAecU;310=VWRE*sVv#
zpawO^chjq(Dnr58?Q~dgF40IlcLGtVzTX2fvv}qk*~IG!?D?cpH4Kb01cSD&z;TFy
z)mi+q<UYs4JmOo7>r<;E&u|+&0NvU;OE^YXAqYF>nrp|+HOL^<)fIttLTJgdqIokc
zbVC@<B0*7YR)Rr#Z>0x;?qK>$Jy4nzaT;L;NwU8Jl|GQLixB&pfAF1&Q6U1XF4TO8
zL4g}31l10{S_|FDJi`vT?YM<m4f9C-k|SLSZH^B-M5L!eT391#Pru?I|Kb;$kL4T+
z<yz9-t=DBk_0;Zv>;`k^%m$Cvpw`H^AiB%$aHi(ZT+P;0JxQ0=wqcDzFc@!y>%!G0
zTvSN2%AuUv?U>SjpJtZ>?&iagYCF(g{e;9WVZO|0cNbS9?iEn>oh8lEJ8yJ0od_q@
zS}u4i(|`4ri3B2JdO<?gTM%x(96Hjo+~Q#FywH`#VX+ynsR>${=IHlI=BB<cD6YMU
z#`_^P+1uSZIdo0HBMl<v<wh)v$gS6i`4?Ce<p-qk=gRaQ#9Z#5e`9d-*?T)=aC+fJ
z=!L!J39R?{yCROSlCQYiIS;))pu1!|ON#qvZrwF8LHm|d_L@>lCv+jNE11R^U?jbu
zF=hcfkAX7r4cyVfME<?9SIspxJ8!G&^ts39v>LAQ8VypPl{PtHzz7t-i@+R_3J^;$
z?ehf+6g}3{)n7JZ4q$=mn3N&3oV2<RO0Z8r(wIT5ep1P@S`?&zSpHPA*sko_f~;Sk
zuEY``OO=u9HVpAV6I8qSK1wl|wmb|6fM!F`Jr!Bw@i-J;$Rsy7>BRv>h%x3l|4)Th
z{kE@2;|sq50mywq{twafN-4m#(hoL_pfBZMrP<{YpmDQ0Ej0SigwSg%@-2#BrRSG!
z=n~r4XdjTRJL`9|L>pO{MuM)h5NDk)v<T=4|G|B_Jne%1DX;{<_%&H2@;Pl-ZZ%sX
z_PO%kw+2>VdI-Qna-Uii@@#SgY_c`;&8r)96?vsAlE7K8)565P;v{cKIlY(T+LESW
z%K&sK=a!^4#Mc>O0cr^3aMs}SUR7}D=((jM`T-;dBo2k!XQNYv0~*DHglLH5M!`_9
zzpCm)yZaO1XzY3-%B-%7gNm^*qM2^$D%v@Wi}anO_%-ayqfCpRSE&JUvu3)%I7+#P
z(~LD=+Uz0t%)e;p%SJ&{$LQ@pvIpaLP5U}BH6_66v&S0_g{aW|x#^T2e6`@$xgsL-
zHSgSZf}Ij|FXx9J;t$bcKHa@+cT_zOJ+O~3Yz|hvxdpe*Eg4j4QVDtU_@nBSrT~o>
zzXpt94N1JIT9W}ohw$l6BOetUSMsuQE*pPH4`n#P=M^&Y0z?lNwzOz>rmC9erh2`g
zjlm$Io}Z77g4NAkfs;1TZeuy10&{Ha;$LMBp|KzOjUcuhLC6hKNHB`um+BJd?Cltp
z6MbsIFeYw!Mo{Mzsp{he%lnxm%{p!S*`)f8>K0b3NKH4f3v-%~JTeqtFTVRFIJepq
z?zR;r{dWSD^;&M;RTt_V>2jD_19%=O3YL9w=W6)b>>GVCoR^?&xro8YIUw$kSoyCe
zz^tbhj~Bw}JHjJENnrFdn8Nbg^tgPi@NJFb+Ywr`)m3?D5X?o;Yf1+Jsucu8E-NW!
zxNO*p@B92IK*}#1D!gNwSZS=`>+5TU4XFWOgpAbfl4}wfA!R3}Z9a~aJ6cwdIr=29
zH!N1JJ~ruFs?gkBiz5MA!g_iG!^K%EA=Gbjr1sI=38Isr@F^_{>*Z7R5uD%$w@`Vt
zo023suMQE(N=s+j>zb~L=LPa(R(%s(bF;&^8?%4Ior2V8x!{Eydb;Ak0-(M-Po&4Y
zzua7~*<^MYXZ#d)^fUuIH&GxiH1aHdp3iL3`tZX9pu4KA>Np7IBj9LWnsbIy<cH2B
znCoo@gTU2%8jiLa!<qu-;NcZz_ZJtb9QlC?-pBL{=^Yykkpt@PO7<?(1uWj~E3g{V
zaoOG3PEq$SMi-^*ZnZ+36@mCN)ympBlA);m0VrU8BkatpP-WQDyQ_ED^)6(2mIv7c
zk0A$krpcNv^=#VFYvj*o>F@HKy?B-KDu224JhT72ue@J+vaI{B&6<Z=sdXiOE@aAK
zcgwVI^<9YM`~S3x?hYulL?>>Z*JT@-u#7YKKzy?p&x?zj3%$MA9zPnvc*+&s!vRPp
z<Q+=04?c6Nft>wfbRi^$&kYVnZ?v84{yBB^P5MV``O8NPR(T6e=hxbvzrf3O@(zrv
z$m_<v1$kDL^VG0zP<I{6EYG~aVo8j>CU(h=ryi;36fYz!tmM!kxKgu_Zufbtr{^88
zQ!wBW3n>xTPnz6c52Aoz&&F_-8xjzgG-H%GUCPaxZnz;wKF)D0IZR!Oz6k{>sLy<|
zqw`YVhZ9BKb9)?J;>NHAE4FSzP{MJhezLsbxxLzmW=G0enp5h)Y5Q1pk6XBY)Kz?W
zkbxT7MsvLr>WBWDzbxwvY{Zx9A#W-kee4GK@?mQ--;b<>Z13~%-}->i3mk&mi@|O6
zmR%L{k;3-lp3^aNz^EL4@nA?W63oR5nHiGgj7pBNRdKj|4+ciA$l<@@q_Rblzn1>U
zyhB&{s0RZWkRu1b_mFUPCt&L@y2Mx{fcfjgrS|oY{CnTum>46se%V@^ssF=&z9ZR7
zM5jO3h`v|HKNt?LM-z+#I9l_@V8!i$T+I*c9V0qVTK2>CH-RV!bEKPdOKOEUWa$A{
z6ir5NTu`a+c|T#5JA@Is8LL^V!E9r;W=ckU0%-?#j_w<hA5_UD#Dy!7V{+T%07}+9
zoV!-dJ>hz%-T_X&r+iRmLaM-@_IH37P_~a|7TjiQ^#yNz2nS8Do8w|4wS#HT#g{1O
z-A-|C2`_O**Bdz=z0u|%LgUl;L3!~p?W&XRrl3{op|`xr89fuxJe|Gq?es+<naJ+a
z)JAkk8S`%PGVE%cEyx*SfVS%W2~&YZn1E>hO%7F+vE7ZZTT{!55hhFDDf^8_)aRvt
zAf=u@1Zdw`W*6t``M{D%&+03|y`OFZ1lb_sA?a&z2%C-h&VCL<!q?C5$Vg3$Hs`kZ
z-iHbo+qrv#w8MJe?9)k#Bdry=DAhY%U*5=JhVzs8Zq2rK##-d4p?zL^okW^7+vO!E
zh{bO@=r1B0ZqcId(8QEtVwleb|9&glKXJ5xyE=W1yF^29cL>$6d3WpMCoV2E$A>Bm
zNbI}2?<v+I<x<u|3%~2<>WJqnR*WXW(6c{BX~U+J0jxi#FLopYiha4xe!qPmlZr=c
z;IrQNyv5d|nPYYrXCl(s>oV+P>aPTEaU^jx0s>KVyYy4AXYRUT>Xlo~y&UMPciLWz
zdHdXYr#fK9f4?y+`VG(0L?1x6J_-1MO~VN}CyF*yHu!oZa2gT{e({A=NU%dqcY$xq
zEu#+KtX~sB_2-~qE9t1>IGQc#pbu~H^IP!mY6u8_{cnFrP%2u^9`M=mi=Yg+d!`0r
z?30z9@jC<YXTUfd^FrrkdI%+vcWCrWeA<cUo>3oM$%tnKA;2D`vQQGcoA@HQMF;!%
z?7r9~w@$Y|T<K!!%Ii|aq%`lvq%XQnAG*4O`h&O~t*Y<&!CJ3a4c6YtfO=JMOo9^`
zEh^YdODMc@+E=@bMH@59U@bfPGFQ+fj+7MvG*P4{hopO?G0XC<X9=LYzO<B^G4#k|
zF*7oVQf^{+WojR5TA;z$-hSp?nm{(k5Q)=^`a()ob>U0S;aD$A^YiD<NA<t<)HZh`
zqFf97+L!+bZOv9FCxW__P!ba78Q+I(#6j?X_xQq&b_<r<^g*A806F*>rZJgvhD^A@
z6@C50Gpq!`KQW;+nfqY{Dlw%l=HQe|$fS$?Uc&tgBtD07<}?0@``2CC!K|i1w8yyF
z)N+ur{I;@3-HzN@&@8bO?6JO5VGb%%z&lVH@~sMZOgSHX8g&zN8W4RPy1)zEspsRm
z_21<<&Acj$k9kh7K3SZJgV=R;b_yM|L$G}7S3eyRT5{-0&1;v&bg+2d*wHm8bsL-g
z8M|gwDmC28{VRhb@8Lk|CyL}yPVcdGt)9+q_>0K#+>;jvdVgYCz+b@LO;nnQE8MUr
z(oY9>^+uVN=TtWwJOhy)IagPm5&8LV5GL$sGs_=l-vPN<YuR^N3$6)vh?y7rh*w57
z;{F){S%%qhhSkFsA^Rh5<#@j3OA2irXvXafU5^FCEhKM1c!FQEX<)d9`}l8oFgf>X
z6ENlh?XJ6KB6Z3^5Y~c#<{M^@#C|MUFHO#hP<yA<!1nex%muoj6@0}7YHDO;^u;CS
z+{QvTZUMp$<Phy@*V``n0%4p>N0OY76Cj8`tOpWwp(Rafo{<xXBGSR^DCyeYAt0;Z
z*-ZUu&ak8F!Ft*wc8F)3eZ>RfUN>jLt#ueMGuV9PdJV)b)SC7=q?35!XhZ1QIGprW
zN{ECc3JY(8c^iMlgS+SCTJ+<6!m|;H1F{1P;9}8M@_&!__d=+6vSat&ef_cUNtyGT
zLF|yT6HO)LQhYfhVdT@0@dfLkCotJnYUkIOWRX<GTPon)46W&40V#Mjur&iinUHwe
zK}I3{f5bc$D_E`|blR>spH{G2=$3@o=&wWBRH$9Yi$Xv+(jqX1&^{dRsCYf;UI66h
zzH$Zpq?@^TY1pjVWEl766d#?Xa*jikWHj-G%AQ$x8bBU!^-uBopjoqMOs&p@UGG3<
z5Rvi@pAD#@e>AGt5F?633W6Ms`o}9U{o{%`Z+p-s`%M2~&NQdVIV5LCo=;Qyo<vQ0
zTaYOi$OLn)=)K?+BuCF%+55Gl_{BzBjL_6p)HeFZI{3~(6e4Z*dF`(6kJfN%3FI?x
zDipArj7vnGzi4h<zkzwiJjH);YzLJP7tyi}Q8-n=n^wfwEi<hoe7=+a8PHO5WDwi!
zVAGoDXK3P)t|8!i62ZQ%aOh4i>YW6Z5TdVVQb9e)_$;BBnXKUGTmlC=hrZX6rVQUP
z4w2%!<}fEGCqF}Ax+aqq!7F^xYK8^kV_n`xWnR`Mn2gDDN&BXo+6DfQ1b1h}W;F*p
z=cbN><hOZWEA_uC!8u<&TDB!=;~f6dHx^|I{Ik*lJx(wmZz$&yoMMPF(gV7?g#mvq
z-r;H1VBPgaPSX_Bx8Y-7v++G@7fTJnG=LQ`g=Eg`;*>JsWK9u>(v<)Rs3<gzyC9RE
zrePbcfP<6sVa9P0zCpMA5V|uX<^r%%q0aHlc=wJ#4xI~lDA+WF_e_1nV60^tcGkmA
z+po`l0|`KAHWG#%L7uJ;+4kbiIU9TBI#3Zg+;9m2`10m`0waj7B69a#Jvb&4t@`JI
zs+xfSJ|V?4QL;qfvgXJ^;@K><6*l~W=vImsszRQ)oIX)PdB05!mJvio!S!xn&raRr
z`D^P(o3tc#-(-Bqp3l?-(J|~;iZ{a1lTF@xcOS`r(WMA)pxs`$Jo~BYlMA%)`-y*$
z#UFC$o7<V=u^AtQhd~)3oTgX^`p8hghv`NI3~x2!Sk0{c>+U{6U11vup|-v)y)861
zD>>ivi&T!<@+~y9M!OtnL~nBN$I#Vqu}5gUHi2&ps4h)8ROS*YpXSzGd=ROK0Xl~!
z!^-w8hk0|rng4i^1S)5-9x4t>fydsTtjVOLD7k>|fA-LkzJY@Zm8JH%sZZ%hv=Y5e
zU4)l$oKTHC(U4@|pQArszBJJ6kPOg23&bMnDa>u_>E8of4xepFQnT+cIu8%eJPCqd
zl59-_rWz5MCrnJ(W2}`)c~0LQFD(oHCKvg8%CR(eWSsv7`UgWMuRbd#8O#q)6$bNr
znu-j5aOckS2yiZ8Z(*-p5gR`sJPEL|$IMeV!ITUIx{kQOHbX4t{=QT^p`O;+m1fm*
z)gZ$*q~anK(BjO&-spd%=sEoxYpj8V++Ba$un3&u`X5!gjVZg0S9@L?!fs7R`Hs1v
z1505Opt)KJlxMtefSW+cv=cTIIL%eL*UX*lOjgKscKIcpSvnmWYw=#vJQqcQzYkIy
zdwqnp8j+!Jm@Vplh8x<C;<lngRTh$RBIbNCT`7(&aWRMB8<%Fo2~+gOzY}Y1Df^B9
zT@}p6y$FU4*e{GJ?wW@7YDDfH*-x7V)7=l`squGQW9~@1)R%_t%GEZ>s&M360u$xc
zQX*q5MW(~VF_omfJluqmuUI(%OK<Ooo(CiuRtz0NU`5?&W@#)6V@5Z}w4?`?TiQYo
zb+G6J6k(T!7K$;{+<w3!HS8A9@cz&kKVfB8V?8}Rb6jCTK|iN=1fS;5_`?P_G^Bj@
zo@Xk(Zb8F5&I=!x0R^v)0bZds?4t64E)FP6nc=u3Qs_crb|p3h8#o%6Cl@>M&is-P
z%sNl>Fdhk{0sp@m9_s1e6-EX=(7baDQyp6d2n)NuXXx}F*jlM96COYfHkyB>*h!cF
znI3-8AiibGZ{=%!HFT_M+pfyW#sBb>E!Ta(q^yBNr!kZK*gvCH>2mm_;Dcl7%q>5*
z9r`SNt^O<;9$8rx`~&$wn9%)br<aO@^l^D*`pz`g4|PZn({Q(qCGmbdikVt=qtC&`
zjrr5~elEepXt|ggimwB6*L^WXkb!I5A@Hbg5J<BPl+vME!l6A`FXSBFWJ7eVl)uMu
z<Hrf-@n@Hg`%laoXB-au`3Ssd@WITZSA}VU>8@jd1w8qtI;^VDmov<(g20h-63X6B
zuX0B2Jp%J8U#}Z&fH_oT7do!10g&35MyW*#4l9{Aef<20)&#~IZ6q)U<ETFv3y9r?
zOwGiUQs6`cY&Q!q7&SKBZ=x=+cC$~<7a(C@><AMRnvXIm19_&tpmB8g>nJA`281oh
z5=Wbp=0Bl&rZucCuAQSDVJsQp38z6qShjTmzLiP(5aIcpSq*WXGO(~*kq0w*w1>C5
zvgfybqvoS^8)|CIP2e41RB;sjVV9u2>$reD7uSK8axJW=rqK9tSn+tqMXUH=4H<#U
z=CxR+*eBdgtFueIx-xn}c75Jt7ndTYSnFfwR-xcoJHJFlXdm#{`aLE2kSehtM4vtr
zU`rMY$1P|MK@&Gk$a6Qqf3E?NLB2(Q172kFIyf{L8SI{H+PNgtl0+Nd_%;eYTgdwR
zY=Lp0qaqEeo#cu>&?2ZZsv%3iioHU&0ly-4Ya3`e2R$+cL8f0tEu@#@DS_#}|Hsl*
zM@6}Q-JwysL}`bMfJlijbjSb-(h5kbpwdHkioi&tqo8z)q%;EzC7~$YH6S1iIW!FQ
zJ^a@9&vjjwx)$&IKF`@_pMCbe5b351f1WGeS%<Zn`E()=Zl&v+)7<6+rHA}W=QQLk
zw$y>~R=#S#er;}^U)@>(3_DoImDGnZ9ucHEOFx+1ajvkQ>${u%w*nX!U)$#%qv@Hp
zV&7uX<r{r46*GyR+`n{@(wqTGogC*cTfk@&NA^{N!}W(=hNkvsDW@;2KhPh%ggm=%
zMy^!!hzt$pX)ztTs$yYSj#O?+S+L$C>7IXrt)z%JshjZ_(k6fn(Zb8^P_P#^unFWI
zBL#~4f|tfEXO;_9pS4hIL&c+y;yYyrL>(+WP21gHMQY5Yyp*WXS8tF?_vmj-XB}-!
zwrrOCKz~e#f1DF94_phtk01O{L48o=J&2Z6>(<6`3Rt_T3}?UHb-f5X;=LR#m7h$S
zuAN7^Wg3GATVa;&FQYuDSbyYP9_;L7u-S1I5cM+V=?ALmjNeW2e7GCnjj29|*uOm?
zNv@~=9cU>deaP)Uz5MA$m=BwFQ^3YF&Y#Qo-l62)!~1t3_;lX1X}Y^yDJSX6b1F!N
zf<?=}=kqmP*B8i}=@v0tuFcE7zfGqkT9!$vDXoA1HD6roywt(+cD%4!Z}9u7OWs2N
z7IWj}bO!k>Hf=TG-W3?*cm(<Jge8b#zRY6=Ym^}LJ|coh<L@;7m`{6KnB@q)GsHX7
z592r{9C}L3B~8}gnyBao8@wlsKVCw-%z?$z-M+Dl)4mH^fD2RuZ%e-3{@bysqz9<e
z%O3f1?Gpf_&E~a_={$|rT`d>a79QWubMKXT-4Z`EJzacB#a^;hhlE(zv?a+1Fw<C7
zcwNtJbtYNs+NhI`Njb0k=Ng*<D5>Q6@0+{jC#mv-B-r<3Muf+6ilxq@DZRfD;|}K)
z`L$f=lBFoKnxns#hf2~m0ZB7f6IUTEA!40%&j-!;(gbjCr^|9n_}tw@3zlbVI+OYT
zJ9|24_20}x%6IbCzq#sd7rstM%awT7jq@G;4HEhHGDh6_)IV5c*qtPyfKn->)jGz`
zCHR%uVAS+(&wRbrjAhR>?!%@o{v<f6dFc8=r?t>bS=we<TGc38MGO+-)5%_Ac$D51
z=vq)dHW+MX{>lGX+1ac}{9o_%p<Jz&Un+Jd{P|v3d+o?vvk-VwkHb#F9(gZe3Vi<e
zws;~iAP=vew<!E>uiF}C^JNqQ(dq;w41rGm89)Lle-ZSsD<<6=q)x)#jdOGINB%CH
zrZ22(fi*V)8a)4T8m9cn_KEeEtzREEn*-wmoLdM>qQOsr@Rv1`@YsIz(kSr3J|M@}
z#x^*+2F5T4l7R=68Nv=1U~R=#j05cV&-Oe!9`91jzjOowdoBK#E6OTzo3^eaPmF!z
zw<}(5Hd7t;DQyMig4cAV`t_hap4^;y<ZtsWx&lEA(I)EvYGNb?F<I+;{`>Xs<OuDW
zN+A}BKV!j<VUBlC+k?P_AT^1~BO+SqsY3?v2)gbRBOY8Zzhv*Qp*|vJ{rl(pgHgQe
z=W4IH;OwHqy0J66vR(IqL1mrw5~&X<ElZfwuv`(q9&s}wP#zzuy<E<rv0t9W*!|Mh
z_i@;KnX4V^=GKSU?CGd?0%?v7?pf(4J7V*qdKR76y>|`^-_QO%mnc2i?O{8gzCzIW
zXQ`N$!sDBfYsPnW6Q=NfOS0!#jBD_bEo_AL&|%(XSKs<kLqzjM<tW;)Qoq4<tA2Am
zcG?~L>_R<KGsZpm*mSd_y7T%j`@`m+fz#XGO2Zu1F>>Dtv=4_O|A$V<`-`XP7Sj?=
zoHQZosw)l>_T@$m*SpT1D0pSp3I|&}>pG&ZEwBCo2M)L}9}8P2cJBTIh1+o{Y#fzk
z+Uca`M?<gpAD}Pjx;p3gO@VVvnd_T>lD>3y%}!X1#pQ;nBwHQP>dPdHf8j|FjEs$!
zwdFBeiigdidw;Ukq8lF;A$+4xu#vwmu5(adDj_AoY?CJca>?qF#Zo-}sk@p4TJ}sX
z|9g-2&wvd=!6g#{F3oYfhSieYgCXuQ#gd7Ie0WeX)GYH=0fBa@IgW0s1y-J48Exls
zApRkR|1;rnr<dQ&$cH}PF=JmXGP=*uC-Z*;H_J%5!3z$@6wP<UcH7JDecaP|*p3se
z+zd|T+ye?ULs1;{L+)Q^KaO3Z{I%@xXBsZvghILP)vK`6>e0BjA*9=`TZQ#{XHNC>
zFYetC!l9<4i}+yM8}aXh5G5N}-2&nqb7$f-_P1$iCp3;Z^Vo{rp}yu@I8Q71Kb;=$
z;2!F~O;&@Sfvr~iH0<8pQ>BM*G1ai}vpWyDd(Se#(7|78g6|Nav)Oaa`gb)bfTO0~
z26AGHG2#skBxJ}-9!5Z1;V^7ylk^vB1i=Qfr<-xkTlmMOlWpOgpx+=SS1naL(;Uw_
zbFF>!)MM(@2o^#U(F|lOo1ZXqHP<%@qh9eWQ326I9)bF<z$AVydh3p(2hD}nc|6{i
z;-I!3YnBe?oADnng4i6Y3oErLw7=w#cX39I2Lq=6!)D{q*_I9ORF&P3$684(<MAgP
z5W^RE&5Ura8ejn8oI$UZHV~MP(RV+`7!>|ejPsWO<eXxLaW$YR+N1Sn1P3>(Y##Rk
zaku-WR4`i-Q`6;oS4Y!AiqpU8Mq4nHE*>a;g|+wTOnL=n068z=`Iq`N;lD$H&wA5q
zbivUu_pWjhh~6#h-}kL!*nutw7tEI?{|1T`<A=@c+do{35m)NfW@)1Y1N`y83{5|Y
z!|$hY`qz&;?RvrDM#A1V@qa)_1B(@p+&)y#yl4Geq%&EU-I^<D50zUM^S&F_EB`R#
z_gR#Wx7%Xyb?Z-b`wxkfxiQPfpH5EsaDETF;xoHpj#0VsT7KQvcXP%68FnUbs|Ffe
zZvxTgZo$FH3jafxV#rs^K`Egp-Zvq<ihI>wHIkDZn%bd9BjDW}2seu8)CpU1at~aq
zA;@AM2&~5kv@b7fh4#FQ;tfSSXLOZ*EMkc-({Ie^0P94v#}{`#-UHiNN`{^K`3>5U
zx`pe$!-4War`CtfoN<o+8`Vjyndv^+a=)pTT==nQZuX_Og}_&F2=stu!0N^TIJ<1@
zexJ6c-V6SqaA~PD7fNCOHHKTs@=|Ho*k$+gJ)f!<7q2#8*I_DXMlJr|{d@UWKxV^@
z`rg1<?fUR<;$gC;UZ5)i`5tBQe{Axl|7h>B#mhbWV)ympVts}pnfb{m`P|mt;&Vos
z*F^dZ9U(7>C89*Gi)EF{l*lxo|5cS9!<J*d7DrWF7YmgMdrHc;%yH}BEW>O_V#qbW
zL3aAv>ejy9XI!bvEl*?Pz{BYXg^0)iiN`by48HG>j>1q`{5K7VMJe6uM2la<X^ndw
zTi)<R<p{2si6&#X&fxqLh!ts2{!i`sg}^X~Ir-d>GXBsjVjp>5ZaTXACiL!&kj;>i
zu;*qer9GjnMIHr%WyE0=z1J8x@9;w1{`#jclWh5l{AqK}_S0LiP$9X;xMpx4^&f*B
zW!;{b1r-AWLn6@(N!NW2JbZB}bBk+;OX$r5t22$V)q;mqSMd?#0jh2F!lVJJn>3pL
z)HmkX`Shl%9`{!~KZ&8dLP3_mM0FGqoThDM-LASn)Fw2uDV80LP(7ouD!^EtjOCAv
z=LaG<W?xeMw*YjWqKfB7);<$pWEu}tF6~!O3|?H0PvwC^+h_&$1;`;eYG?g~fEj$m
zck8anSzQ&CP?qsW1q{4<>o5P|O4~95cCMUn(<$s;&uh~XZF);ab#2_Txg{wq2kjib
zx-XQ#l%tBKJ$ZFo&^lO2iWu{ALrk2|@w?2eBWbAeS=S$b)^#yu-S%lQ1$&>Cq);m;
zG)gNodN~E+;oX)~lH@d)!;Xbywrl3xIPz3vu;YiqNbmc^leZPwY}xFRg_h59L)}V{
zd-p7u6R$uARp6yp<;xn0pZ$lm>QGa<m?e1@ij=<v4bx_Y!<i4#mjy7(-^pcGiP$RP
ziw_wletW1a20{(Ek}JM6Bj9j2>W>r^2h(>MvP9!BOi-Sza=-4o6^6LueWD55`Eg|)
zss9|!n&Ccmtm4F&)-+%Cq*yjQ4R{6Uz65DR_jy|bxdzpdRG{59SpfTH?c|LpQO#tT
zJ$8eP4?EVXe{N0Ra}llT6?$XA4s-qijD4v@rvR@~BIcf<82BjZuW62YYz6wJ6&TN5
zmv}|39>?(xxYp1EceOU)Y=P7gsV7}p`q>n>hVAgUefxGBV5=z$Y^Q2U_G=HjdE>=G
z(ut{9MSg(4lQ1kLGU<5zs#$gCpf(6r$F&wauZ_BU_%#snbD|N^MA#UMe<gI9RJqD-
zflUD=J$}nH>rdILF6N#gT)eRaiW_=RXto&it}@*6ipREmi-!X3CJ4!eV3FOe`O#U)
zL{>8vySk7I(*9l#ZLVxx;!~=YROO^KUU+Vchdvy43z`z$y#V<~R7b1KKO|ZI_+|(b
zQ2;-bJF1%~z)BPTx${t7crbK%f#?(prHiI4V92kcXPEh652dhS1V`X0IayrnjyCIk
zxE&>dc#sH6l2!hB<O%G~DcfLRtzhkxqsB~Ui_-%Esb4#z=U+$<$($*d)HfwM&xzS7
zH-tGRI$%{AE{lp`RQ7dNwGxif#vF`y;uvj4FRV=JY5l!cA`>|YH*X-C_OB@(&5`&#
zxi~-D2IqkexKl`2MDb&ij4eD_Zt0{Ve}3R!02^=dERfCqEY)8Ji61^-)JYpQYS;<;
zpnc@yyd+1~w(tq6wBi;~YYKb)s$YwXtD1sp+wRI&o|y~}KhFNe-PGWL_9Nu;&Q0NM
zqr7dS&oNIMih>z_t2p;U?4FGGjg>ZjT>SS<Km}_HgF>OvNDt;j3dkvC2!;Ok1bK@C
zbf*USAB>MW?=7;5F|mk*$Vy?uO=w8cH<=V|9#8jZHy{itcJv^f7djQy-2(|Cb|Dt6
z$sOK=%u>G%if6vU&iM$AdJ)j~K`ik3JTs(u)!LEaJC-w2Kz3#$Sa(Q`9z1gyq@onM
z$AbaR++3$)yLm8RE%U2PRv@$a&EOCYPHG!QQv1$pICea)`dOf|R9!TpI0W%wWqvv#
z1i|xzU6^{he4yXN6+H8>wpK|BG>4SJHQu=*ge-kymLEyY9c{&D#~>@Dy~u;YYR0t&
zXs`0MvZPnUye`?cnlpt%k4>56&6$RUkH(hY&$%XiCoxYeU(ExrWlekbNKY>0!pVBf
z(wZ^+KdmQo;ksB96*1;dM1TYvKDI(NV-IyY{U?AWkIx{X;EP}y5x7X~a~oFPb^k=0
zF@+4tSd<8#K$34ly`z!)=czHLTp77^Rz&9QA&AauDi57>A*k-!vs~O8<oi4d7gDrB
z5Xqb#CSJ3sglGl#Hi$m4ePTWdX%M^R``(CIy7T-g{CYw@iaxOilKLD?3*H!%VzBE0
z!Q>sXk#X-yP%uNLEb$8cjY+4g{ZiBHO*zhUv-v+9Z20Hq5)eK#u{~1%Tywl=XlHd?
zT51b@UGNtvH6!?N_!<WX`jL$=&7q{{ZY(F>>8Mc(Pqsc;L-BrlK9bus4c_!1(mw;3
z^Au4D*Q=IUHt87~Cw+3N1-@j!W=$w{bKE+<0}%TWt_z(vVa9!=YVZ91zT<yC+eyR;
z8xpk?Z*Rmf-`jjxs^-Lac?Z0SVZe{9Hh(9yipV(iOf7mir?8`PXg7hPwJeoY{~fZz
zIFo8a1Hwh5_Z-cYH1s^wo%m*6pf>9VZ8%)l?m#?5gQz^G_^B^JIGI*Gs+*Z%;-`lS
z{2gOrDRc)Fr1-kFD~6r&+5I^gxHmOFRFe(6oBLQ7Qikl-f)37lh%Z_~vnljvZ2tcd
z4s~e~lHVaYL@>D@4D$Z&dKxIZK1iDGq^*yoQh#d&DjrpHM5H@=!x8zGY_Yp@v7YOq
zS39_Lukb{Y2Jp%e!JOrCH~Rz0gVSW|nP<V~%v{kDlCE-0xGUK?OuVM&(}PYh+=~N#
zO|8*ffqVMYoU+3nH&8qoq6!k+&Oit^ZeLU~M~?0V96vdo!XEvvo*wO_1zGUOWiaQ2
zwcMKTW+_}>1)L)HJx3VgB|V?jiq?me)j{Xj$2L#DJ^mFX_s?mh;mLT<bD!=cB(lOJ
zGiO8%f+3qbO*Gpa<ZLm6?&KozEfTTtTxGa3?J3umBbqC!oI6DK#t~mFeWDT6jWYyG
z#L;!At?ooJ$I0;d4)kZ*SHA4xjzf7n&*aUxJ5Y_!B5$WP;8J5r2EN5=plS`@2`t`&
zKhwL%P>;zG%`IZw)6*+qYuJh)3&`=_j=fe_Vr8lJ);WhFAgsMcI6O-AFZwwVHvc7C
zWfKDtcCJL-Yde;Dog6WwDM=O>`^S^c($Mf-4Ot0(ZchI}C+L?VjV%W%3BG**4#Xt#
zYJ3q1V|=Wqmrz6d8{C5_YxgVp(|w?qNTR4gv;zTbXnk;bN$)4sm)i{XM4iL-C7(Rt
zJMo?RH|C+<QaizgNnEv>*8ESYOry2tO!jyZwe)uBYdd1Ufd`sh427Z>aLkZeqHdT)
zd?Y9L%7eNz-2KJ}a1YA$o)+d;seTijp;%+Jw?ia}9~yq)aMNX1AL;7!X6nL&M4{fu
z2cK~-klhML>HNHt5FbUc7JV28kG=lfegG7LDP;opTy-(3O9#$HN!Z*O$jQTJWxQD}
zAyK%kt})Z!?-W2}wU(~Tip2{{iLgq}6@8&Q`gV;NWAs(pXn77Fg6KbE=%L`E8payn
z>*qn6P|}esLNbuWQ@_g~+K49xE@zOk8**NEL>eAhDe=|f268C;_E$gdY}6pIT<wu`
zk?c<OOP`Fqu|QwjPB_;KxU|17Hm|>1=;R@Ko8lXcM*~`!XVmaCF=2WC-+xK9PKt9=
zj@@IQ8L-b--mk3z)MOa{#)*KV8b0r&Q9tiV5M|K+#E=*W`3)&A!<kmVxgi<IZW(l1
zGVX=^@aK*D3OLdbmHf4=x2H5MbWpbfWR;W!b>b~(ljoQnM3iLZ75D%(>HbM`x0FHM
z-d;YuW5DHfL%KxAL(@zdeRPZEjw^&X!GqI3m1ZYSoC?Ee7REuONOF@5T5}R}NT0HU
zbNNeZnw!m!hUQ30n4S4=?0{q)t`hi3Mn>h31*Eh!d?&`-R&Rk#?x>7>&41XN<9S9y
z%L?=4Kd>%{TI^jGHLIa!;$LlLo~-~a`W&D|#VyDua{JP%>^~BS_QS|&+7fQ6&b<Oz
zABu`+dxohlZCCQOT*H+vPMHkzfmg*r+OWGBK%KWo5)^1ywg&9nh^)S|;{S|ixwi`?
z(-1th^&+NI{{--bE4ahf3>YisS2zRLxNVM*$nIVE;<U1H(Luh$&$|SLlbKLw<8{W*
zqzr|W9%4sya+UQ)Tp_!V)0@x&{B_hL28Ime3buor*rMR)v4?l>HgpG#6hmpl^|uw`
z#mz|DFxHP)@5gyDKop{h!9!!<``6#{otdFte1)pnzTn507LmmZ?_rv?zw8_5<spcw
za;Jq0vlTXu?=zi_uIoDJ`K%cv52B3-OGsPo3mnbM?hZi&*1)Q62|0iI5SiXV_>vX4
z&=ucR{IqgT(tFmIU)F1Q!p*HH&1+I{X=)mbtMwl*o~PbV60yJoRA|B(R6C*Gcdqqk
z09#-av))f<?Z1IsVmDHg&(DbWY<$v8`3<$~G((Tq&W=Q^7?CPkc`q(`{<Jf;To)3&
zx;4kXrVfX5@j;KFJ7OUXfBl*t;@tiVIV2gPG<-wM*@ht(oAeMyF`{$If-DJoa1Z0F
z*)a@>e;@(Hy1qmLQWkW4>pT#pcij*3scu4-=lr=e`pXKL)E2MO{>yfOR&a)}&#Cx*
z8jJYs*5p_8?OO##CvZr+Y|jbIE`T{XHy&h*Er3LY4A0n80C-!T7+Zt`pLh;nPoyT*
z6M>~Fz-~5QoWBir&Ug4He4H}$>Z@h>v(W9WhuO8!qL}f$+SnIET?T~YC?v<N+Iot}
zz@x7MnX$^A<a&%;0VNl(Do=z9#cY3RO;+nf{>w1cmU+&AcG-o9!-d$vR#<Mtb@0E3
zH}=o*PGsIfw0^}wYud{UUutgD3p;sg^PyQjzKD!ccAg|xNNY+ToplgtK6P6M)O~(W
z=6_D0qICcKsbXPa@vBDW#ScoK@blRib{MVy`ge-U)$K+gF3Hz&D+W?CN{QFKRkE^E
zo_bwD$zms!U*n&?GYsUCE^Y{L;|y?8{4yVc(5&UR<(0u4$guF0z$pU95EY#*@GA#9
zxpEANUGE<FW|FL?=R4f&=~pf&BKc?_q0PE*4O*cDIg4oc4c?Jk1JRKs+?mwngTc2^
z(D^t)-Pl5=h;oBgd4J^^K;omiv5?ekl<5mJ5A<&2mM@0}m|yf)!UXcZX|t;SdxxYA
zRqA>AbXrJB35&Mt=+ItRM0+{}y$=U6AZwtzGwR<s#0Gc8+>pS}`ftqxcYjU)KX7<H
zr{}P~QM-7p%+Xw!D;S!7775Glq2n-gD(gZei$oy*QWI_mTPX7QO=fu|hEBCym|Z<0
zSN~7VQ7&L2K@D2Vz&K7rb)-H#V)7xXsjs}nd2w)lDyBwsB-zx!^=*LrkH`^r<N4(C
zh8+1adaouR4juifCVJzp4P|N5!DMbtR&bw(Edn6>WWc|(+#9J4r(%5p*!x)CFt8IA
z1yq%-%e#_Z?`G;;bM{RD938*}`mc@VYqPKIf_%U1>}=u^({1!4c@K?$L1%<{Ff)Jn
z^I-46^o%!_({iZE-&_9TbayoL5V$%$k5$jtBW&%A7+Y!IUYRPWlT5IBYr3RS`c`WE
zt8UKK(@@|sO%Wi$;jiu4QX7;q`i%9=$#KEKaP(Gzg_`rs-;Xr>BRQViP5~3cGg6PM
zNJybjH$H~7z?}3NFC@-X0Dh><I+Unc*au$jd%bk$_O@QfS;dDXx{4e;&@O?zb1@5g
z-+1JT$Y6Y>P#u56NiTDvt7c&>c!TiIU5D~<o*W@H(7Q&0)dL>lHmT`qp_M*1S?}L7
zu1)4*Xt)&{w}Uv(N1kK*lrA_f%^~+i84}09fQrE-CrdTbltkuVUj_or9Hql;I76hI
zA;AwdCJp>}JBb6A8&Y8cV8zM(oQhwsT54Tq68v*KY2pSuU;<Z)4OsB8jIW$c+ONJN
zdt?3Wa|oh+Il5>x3*q~ETHZzUKULj_e<guWfKozI66@F=uixW*TKp;x4j`|0zkX>7
z{Vr0nYVZ<14QiHrVW{*eCB-53bl4(+DKD&vQ`SRytkj}J@7309)AyxMW-n5lYE(c=
zt!HAZRJ`#rP0qU$*e51eWR`92aLC(_Inc_k@Hl?&xK_7?Il+UJfM;NV2^>QdfI4Ge
zYT7rkQ?w-wFQO}Sjy%uqR`tbmjUK*hLC7I)$8F7}#32#g2gQiAs!xM|dwog#QI9Xg
z{MGtj3^>!2;BV~kL3c*UwhkeVFVWYcG%0%Xqq<u)1<%~G1@@|%L*t&I6gbP8nG%Bq
z^3v~GQ2RzA*Qu3tXNHEr`MgF*UKVDkXKDGkPOXCi5T<LJmQ?Jix0SJGiMR#z=Pc|l
zvGj-s=VGJ7CqfNaS$V4M{qIjoq^{!ekE>^^ujJuy=K={Szgy31M}s<Vyq*Kbg#|bR
z0jWSo$r~*W;Qq~)euD||c=LdtPv9MAos(a#;A?gvdG@`F{$iIsot9RAO6O&EhEAtR
z+r=PNIU)+gO(UM#?SX4;zl~n}G%Y>`Z|(J(R7n?2%2xQ`H1T*|dl4wVDhsn&R*kyS
z^KTPQK&rxX&-sy1ONEZ8O|Nea{;RNOerOiXgOb8R00MB?svYkCet&mvowu5K?p0J$
zQ-P-60&MQ+1Kz878o%?SXbFC4=V_Nq3&p)1tmZB%-CO20Co4Jecz`~ijh8>K8SP`@
zmJ+p#&iM34T}S20Y8!DA!B?I?K}P$I=a)2#0N$EW`U#e*PmQS}$G}jxvar#nI`|JE
z0`a;kQxuSVh*eWdyqBWFUtBQsuBPCA9PUM`I^3Dar4T18u&`HO%C82G?^M(ltOmWz
zqBGR(R!GC1nf_3<-Xv};CQWUTGH>|BPzSxfAE@{aD{p(IpcSI+m<bIy0GvO->j^-A
z|F9WQhyca8V8rns$Te#B-*Pnv@O%iUDV_i5Y#32+xIn@%y@z5Ub`Zv`(g*O*RD{jF
z*7g#V$xwioz8>EnvVgEHs;$554Bu#e_H|Njs9Puol+1}>2IP(+M3~S^;}!vCv`13Y
zqM;Yn*RPZ7Sr|vYf5XVpFJ2GHP4BdfG;g{eKRDr7`KK6Pnl7srpFKR|opk5v8zAE{
z<@5vTC8@4W+L6JYC4_tpzg3YO*^sYv%0TGH?bx(Gny=;YRzF7_1oJs7KY8-RV9IVN
zpT=wUDXje`ZH`8KCs>uT2@gVm#I>O%=}y4>zK~lfiruK-vrJPWc@m@qba1cPQl+W~
z3qngV+j)9`jzR6q8i%H)ipW?@dUbR62dXTwDjAb&`*L#ayo7;nJ)WLHVWZ8gac+a{
z*l)23$?+BChxHKGX-yC%lD^jR3F?M4SQQ=WNH{?vfiOWSgh_ks{*7w6kTQ}!&Qy{w
z-8taAczTiI(a0DOoOaKG`Xwse2pxulc3bQfmuhea-iwo66<9-Wf$fi2&@K0Tk@<w|
z7VeC>K>}l!_Ytd^@iGelN?5>3jLVY7Ri8Vf>7n*l=izvvC*&$HvJslAd5JSFDd;0p
z*i1)lctTl_v=QsvoU)U37q-*6UVDBz|Ez;yhh{P~3jFti?S2wnp@0V`CuQ_|zpLkd
zY^v>7O`aIO3(9jEDEOEzIg$?YQhiEO5IQ=vz?Ayc8ku=>A{s1aZ-a!ar@R85<Hi&L
zX^cCu`4ts|Y)ifaYOK!}yA!2>fAaQHk`;6cm$96w683(!7s<dY<}#;j^Fb#~b{@>e
zV4d;Y8Sl5USRbWIix)13@&YQ>LUylxN82I01~sBPgt->>{SPeHspg3j`N_EELU7^q
zD~_d;+}K3|`b0BIZqWl_b$!q6y71zZ&ya!1Z3c#y3CKmrb-IutvpTIrE~VSh)B2GM
zt~D<P>f`rPq%FqKzh{Ck1uP8fB^FcC&?`TY@Wtyx;q>~e;D{7ayjUqO<63#}R2hCi
z<dTY$?QYt8c{R{t7f{Vz)AjDIz-#>JSlK(|wAi!O4eg;lcl<;CG#fOv{b$9|!7g+7
z;&7&D<}V@~?E~W`b3-1f;%G!n2trl4_9kko>H$8dwYKFGI|(M%A8Aq(^gX}a6>n5>
z@#cWlEf`J@lTKitcr7jecZyjo&Jlhcy^$c=R#E83t6sJ7GW_TJ&uu}3Jsh5q6J>!3
zkZF82-xhuZ8~~KofgbJ%to}KB^7k4MLe>Gl_s~xctP%0Mxm!zd9o|!L>c3=VnRFMc
z+b;VMu!5jt&)z<DKx$v~LXl+c#2x;k>a!Cjtc-~mL&Ja_b{6m5DM5T;=HV|WVqUD>
ztriiYoO;xj@*fmx+;5>rch`b+!hrl)1#(a08dIVnWcnGJ)(zza*~^s7t^*(a=VH0~
z$&5D<jmi{1(^(SNq?12(yMvFu>ENn7Y}Z#t4(+Vez1xerZPDP>aS4-~&k0nQ?in~Y
zB=U<VC<rDJdJQum>Bj+0YX6xuuHF<d2?A3cw1nFz)!ZWD$+x7g?XpMeUR^X-M{4Nt
zY9=kU^BG^C&t=lRIb`Fy9k)KmUedV|I*li&j$XN)jT3)&YT8DO`8Ge3_pM`ei8C?g
z;fGt#+aqWmI)x1)fMnJO>=cOE4+~7Yw9}Z>I&)Se(#rF@Kbt7CQd$R>e)p>qv)azf
zI25qd%{k>*NmDU#vc(T;*qhz(C#fg<8hckS2UjmEe|suGWaI`Bc3308DM{2e`gFPm
zv-7tt^V8u?&VUayWPrR3Lvqm0L5Do`3?@MjCOe+&3Aa&#&?T3KaV)Fm4L=@IOOPW)
zD+OQ8ti_pji<c7PbCHYS=SRK=peQ{vI0E)r<jvqF8zk@%tI90(>FNXwv=#nqe*6rH
zOkigM)|rbl?*_Qm>3`qlKlCJEN+(2K^&1PK1XVQ1MsEK%H$}fgV=1LME_H8<Z9*HC
z3v;pVt+gxcFKyNCiQIzKIk*h{YpcE17+}K#Gp+rbxux2Ft_s*vD}QSfnZ<KLyws6H
zg0bEkrye=p&v@GkPGP~yy}n2BgyA8+MxfU3mQszNZngBsh3+1#*GxHIzh|qb>EO1j
zLICd9V5y*!$E@jOv%#d^EpKC=;!~A|+H*mv;U`_@>ioKau@-?&9@~<Pc<mY552994
zd}Cvy#-rapE-sNaKUSwYk(w?OO%>bIf-gd$6?`H5kH%C42UuH*pp4V8*j=~{B;<<U
zeTt##gaz~GXowpHgA9F(9OM_ch2v6x?Nlp!t3cfZG0WSi9{8fokeZ$!@_`a$`EzMl
zqZ63ncCsTHpGz+N>{;*;fx5+&)J*;|jBSgP(`in-#oFIx`BcG>Cy_S{Gu3<&qWsv5
zYy|!MGpKg0Z%8U)Eds*0XngO;>#+}Xz#Qw#6Py+mJrxoy0&*f<!`go6lVJ2QrJ2yM
zur8zJE@ctHI(Aq50u47V?~}7h_y<47=#n1UTO)tmTHlb>jPbh%oo;JzfR<#myAL2s
z8>TbQDM`P*mCj}fV#2{j^gAa0zrAu>a47<qYJz7<J+)RVY8&o4L;hU#<vQ&Y;pr5G
z)o;9GrOJ5B`L}#@B;wbvzOV;Vsa`BB7^#=C(hKq6(!O@Y2Zbhqnr6}7lbZh1(CcS@
zZw)j%*xQOo?-lx*TWT%JAR}4iO{y+DZa_r`q6Yf5F19Prz9GdZbr$aj^esH3Lw1XS
z=V&_pJnh3R&C2O%^HiU|?*Mi5y#2`IKD^fjq$%?9cgY1(j}MGV?hto9gD-xK!(`u{
z`w*umU?(%jj;tm}A4Mom?%imU#g0<ENqJE}K#YlfK3hkE`4%1m)P@7-s}yBSkKsNK
zCI|=jZlQQp+gq)^p9hzmH}^&IHryKf>;bZiKzH#9r<0px=_lALyLRWXao486s$>_3
zdyM3lJNjmh-^pmD=R&H67rlUeC)cd+GN&U*S?HDmlPgX8GmFVR2&y+--7rIeW@D9|
zU@K8nUlobC+Eb<k>8Cg-aW}2CH!gpI2pll$cnT<!`mR_=#TL(x6CL_QrbJGUuTQ;g
zd9Qlc<A}Q@Z(<837*hY1GlkTu<+gd;grmg3Czoyx5q<Z7Vklg171@n}sDqELqNgR4
zq-1TsAl>g9&H#xNbIawjUgZS6fX=sf8~_^NG`XW05)I#AG;^chco^xG6el3?)+*a$
z5fQcEAtiDKq%eLLmphW0<z#D#t?B!E7<mVacfCBhbXd^mWX?!B4fvG{rZ6`R`Uw>Y
ziI+D@csb9|c>T#x;sy(6wM~jLU6*8Y7_BU2b#-<3H84nZXj@ASBH+^5LO;bZI00L9
zj>`mRn-8G>7bx8KXEBD<hkw0t<wRZGf-jxNF~WZ^9}|gK?P1lIA%4uGROWu3aPAby
zrzS(Ajh`4vLUuFlg?1h1M`u>mTfu9>Bj0t;L*gVYC}akUGz$^htSY^?pbNw|u`Ak4
zNazOTvoFXhaSKX@4lu`PETRh8hPFl1a`8ZA8`IJZx!=3y<>l3Z?%5S=3#W*LBP+%M
z5%=!=?4;eHGmaA!jVP=|D3K=|^_+pfp!Q;}RbJWCP|t!{kMY#x!Qo8*spD@K#|JI2
z{ybnZ&CZN}GC)$-dJ;pu9zpJ^il)%O)v-^s!|pxfXpBw1iG>8i&1KwXVn^7&8?HHb
zA5JI*i+Qd-c9`_ri2${i-MTZT6-cWSX*^!f3YFGDvq+@lBAg}w4_Mrbk(Evt1`m=y
z08m6VW7w>0s+v1nZ%0Ph&ALBF?f)HjWVISQz-zm<8ST9rA(&-OkG@Zt;;Yq&Jn#){
zSfsGL9p>E=0y<m&nf7HDT_K2=3jBUAkxw|%gBYI)07~LsVM_29a6Gl)apZVrRg6to
z`oGm+sKF&DjH1Hs`b<%B9r|_(`n`An&gj8lY}Dn>l>8Bot^|GJ4PXV!(P(>wpYQG&
zV(NZRMNQ4|fk8h)Ej80flSSoej@kYnbT=zUf)n*Iv{wY2MG=fjNkEP#Z7JRQB^y7P
z2&?<@l?aO;cit1h+~kZP(K1bo&C$8f#kKE57{97`?9gA?eO^C}M#~DVtQ~fJaZaCJ
z0XcUqV5$Gj>Z-j>7`Sf-u5Q%~=N(?ogaBg}^M*K0BJ<PA))1h@Yy+_st(~##+sDfp
zQ<h2^0oNy<OAH$!qUvLUHsfyW{IP|P%oAZ-9L#?lz)g9_#(siQkyd6=AH;uc|D(iL
z<~6HC<;TfLWNM6|u0Qq_#5oK$LlMHsBKU<BWuOE~X{r{SmK^_|Dh6fq8v7G!;CP9q
z8Nh?XK7m#=0aT7@S;IZaU*j3P^j(++t48@fJlS`uKpIU;G?;FW)Aw5J@t~n#Hu*B8
zJG0D+MI~2=2c3h^cllbA6^(fI6X|m21&~3=5=D{)w5v7<lLZXC0F+8WtAf`8;GyLs
zUIE<2tvNo%Y_8S*y82v}rxwy~>K$p(e0IPN6b-NZ1Kmou7rM-N6)QVb=@i}qcA4uk
zbdE)Dq#o1OcAuG*)oe%$fM^C_w14kc&&1+?3~Fj>M$&}xc3M@bWW_*P;IZ#i*8I|r
zqGc;mRI~J6?o#ovQctA(+#nKRN{kthvW<07`Mh`}`_NT|-O9J<w9`MwdynP<%!ZJ^
z<t*UpZK`m1#<AjM+xWl}QYvLZDwBj<&(_64p2^+N92J9zQslGoL!s__yZB0IoAMj~
zG?L6p92Cix3bdtl1ba-$f7!offF1fkUi9X~iniY?N4MBgzv*c3XDYOR{8W3gEDE>O
zVMVUm267!)!68zMmF51~9(aIMOewH){$D=XJ(p`;Q@gUD4zvq1EXR^>9mO^2+B$Wb
z7_w&Zv|al~BagF!?I@@bW9(kIge^eH*nvk^1EOM?0A`DQU0_TJ^IA8Fxh_r&R*uVq
z%D{4fIA}=LKBe?HaC+fd_JGfD`0?)Y0?;LgCE=ET@M(R@bTAH_POh8Z@XNAzN>N-`
zc;C)BXjgD3rc>8&rWv-s)HUXATo4lIlqc3)UG`P1F$YlvW*q?O_2ob-!VT<c%yr;<
zMA#<q33d4y2ugy<_2MIFb^%Z7kKX@K_x*VO0J%-}U>YBw_)$cDjQ1g|$s%tVd7*g_
z5h)7_{W{RLGFHC&)8^i%@EXkw{q?>vJPGJQslnD`M-+?Dqon)W18FqeMCGlZ&P+@V
z7$5KE@syPi{{r(zDFwE?6K6deDX@<0BMrM8hG)90+->Fb&55)JEu#7;liVUZo0bva
zCsmu-My~5#-Fh`$_c-%AK7Z1FPvEv{df)ka5a6FmH#~nT<R}Cm%~RM!gZwP(vZjN!
z%l3y8Ny~N*c@xXrpO)xpY|}twQft!W=V@Tk3+lvNNZeUD#?7KK&Dm1}`<qW6w9&C6
zt+0a~zV|6+p>G<e6V#8gmcqRc0StjIuQ1l(H^}%YL>8ICi-=s(!14PGhr^VTi8HGK
z8g@VL)%gpD$GOBI9j(!nn&V0kP$x+Br2}IPVP31)=4MbryBBtLcEkdXTzmqh79@J5
z+dXbDD+SVS+_^lVACRQ>I(Du^Ztgv(`#Oeml3p#f;<oIRN95j1PL1ITTwKG|=TtzM
zB+6d@NBK*Q-Jyyr2t9feMLD59{@yR8^(Gcik1#Y9wADPEiz4A05hcm2A)i8dzOjgA
z73~C$1~0%T7YRCM6AGcAW~_(cqT3~}2P8)}o=5u?U`L7PvVeXhIiFnFHYGw0H%qH4
zjum{FRkc^e6zcS@h9F){Rbs5eySG0cM_t{q_Skz_2To!4aW8%1#{7#=oKM+ZP&h}!
zGz;tVqj{DY5{V%fWc}?P!i!Ihqr3IN??HdL$q;<>;OpMFtQf|`7hcbA6~JUJBUH7M
zRN!0%5`0TbOL#OOPCko-GUmKCpNlbp-(cLnHELxzhH*B0G8ZrAcQ^&+IYR~vDwCbG
zn5;kqJM<E649J~J1#qD90nhzSni!+K_Jj74=22MWFmSbf_{lU-Lx=Xj9!udB2=<Uo
zz<gt4|0f9R%IJlKpcuhHl(^jR_nC-SI!Ds5ulJpotE#c`@N;|K7BJ`L3;LqRIz?I!
zopT5d^TnTNxua`Eld=N>2hD?X@;*CbMw_#z1i)l8@k1|5-C5hA9b`Sekx#UU)CF8L
z-QV{-!sJp!-pFscGEB6BrU(AiSs=s^bV;sqSCr+z+n0PDuW-=fyPWnKO%2^CzwATU
z9WRIJfPH$Dq_REKW!nY*0RgkO=2cVkC(BW5-|W%tR>m9(9PG@#G*^zK!UNbZXx=*V
zlGLd<gnRbUU|+9Dt&!AcWzE?t1f<AP!fKo55}doy!fzvaKSd30>SLdjoC^FBy6-(B
zGnfyC4%h?qqT;2iOF77@9eSBSaS4^BPx0;ZPci}zZUmj}8&9pp+hQpcw5F`Jot|`_
z^jspVs2*LBxhWo2&-04u(hXT=$01#BQ(Jo^w`aUKH!nCjl{F+T9Oe(;Zbas`pc^mt
zM;SzepS+hB`kF*~)PobnviJBAlUi;Q?mn3IPwztw?9r4_YiBcR;9836x-=I$ehPcn
z9RsMDr!>tKD$dR#I!b3ALofAS?A80gU^F+~ViT0AR+oz8G)q0gA3Oy0UfcUEMF2bE
z`OPWv)+yv@2h3h$c21-j^f~iQzU*&mzsoYfgm27-!~Sl+8XY;+M}zC=s$+4H%1Hp{
zo^Tjv<?FthHb;~%s8!=2GRPGKs$PbXlsn751^mVGUqEHF8ksqT16R$wfM%fIY%i&u
z{j#%R;DrY0Zy-;L)kojSNjT{I_%-N#e*IL%8nKK&Hs*A{ab})A`M}wa7$dm#H)UOY
z)rxc%g3{7g={hYr`P9Drp}ZX7qXEb~sM|p}0;bd8DG=h^?VJL+LxDF&P1!@9bnC1U
zO}|1@x^XewAfNLB?YhxGnW1PDGW|*<sEGX}A|l}8sa~OnD1gVz9N(KJw~(LWeAj}U
z!8d^xbB&pq$7~(+<@s8fKUS{&rT{=gfPF<jA(r{E8H3epGE%ZYRe;%p7)!zs@J3jb
zAp);@^HAdscU5F&2x3oxbW5qK6d!NxiKp>x{M^N+WRQr<#eb1{f@mq%In)2-u+?z@
z=2mTB1@wi3MMZ)4)~qy`T8aDZ*;3!Q`{U~Fzs0xd4wsV%K)k;`bIjLCivkeB8gLzg
zMe~<T?@#`Wm{1ame8&`28t}}D1pRnlrJ+wF_*ME}vbAGbv-Mi355je~?>o)IVF7Dd
z#)MTL#Oly@!Ie37Oou$O%-qdgZqduhIMHy{{^`s&&;1FUM_|f3BvK&lX0AJ=&Iz7f
zI<YniIpS)ehBQ-z`y@e3@=)I#?MR^68s#-a{o?q_O@=xgQ>EEz?N3PKHRT2i!90?F
zcr4KDh)drTvrQEEpbn{{ZBe=u*r7T^=o3$57Q%AU5NJp?{Bk}CjK9ck3z-F&>A;z5
z!)Sk%QZHRIa|{Iopuf9)N-&G79lCUx{(rz_V97cuYnD~QNcsk&!s*VwTce%krYbul
z<%Oq-C5GL(!A45kpk#9#5l^z=rg|SEjy@-^rlg%w_o2Z4yN?;D<@D4FrQ=)8QQ1Fn
z_-;0pUcLrW=sQ6QkYkzdwvdQO=OvK^*yni=%G!pXZ`U_6y2h|GvUg_YT+erzw^n5R
zZ4(_ZAI`cOeP6Kf<)1H-KBwDVh84n45U!P0*t7zAiOTe`joW)m#6iIvSx+B+3oDau
z3PWhBVq#ecJYSWabttUX2JhNExvr~W<#kyw-Aa4Zaydk9LeGd+dQ&YciCTIfy$~@y
z!6#X$NmW758iQ4?ykQ5;!YvQwo9PK<`KZZE)RzJj0ZbF@pUJTb=BW$i5pcCVkQySF
z_v!aYTb4z(NQ*UavoFD?f2~j{D@XT-holA7OW`xwh9rR0kQK>_|C#`$#GHn7q<h9c
z<Vnw$4MyB`V9)e_#pSMTKMysR{PhWVMyU+$zd+6R=J1RUV_nX0RAFXWb4^V(vFz=y
zjB9b7jU`bt9WcC)%>$`&BRxi`#dG8m$<!5q?chGVZB;6Q3*!LTFX$6v=3AFc;>or|
zedN9aRv_`Ym540Ma4<Vlwa@m8xJN$|fTG|E=w!P<xO8Nx1yghWP;+oF2r9T}1&}E+
z=xooke^CL1Wr2F6qFJMs8?5B%ZB;g4(We9?8dGcDx9<=I$k^HOrhwvmCS}<0{T>)<
zD1N;9^QWC?Nt7l4Q}eC(U|u6ya~ptgp|1t$-fGSK-NMrb))B}<+V`sO=MB$m7E*jP
zxbgX1L-0unieM?YUuUp-Cl7^bv1C|L_GJkvgiH@-uV6I`M?i({I;JSN_Fwzn$=qx3
zt#j+b;jcJEp=qEI)=ztZhCt;!+6AE5j|U{~Yd9Dzdjcvw9>*x)`sg*&OE9dxC=6a&
ze2l6ZIp!BwR09uemiU5dD{PAS7tLFTb^z)&x~E>Wz~pPD=<8@`aO!kmhb#c-TdX}H
zHB5LXy`{q(o-)dmq;aaTGoV^H0=<<}iPp9p2;01r@4Wz{Z*4y_4NgzyTe|5nn81@;
zO(l6+NiPO-A3j7xK3MC|BDoB~_W83qQnU?@!K7tyd`**{UjNmP=jh~~;pHd18M_iU
zwzaibvCPh*tD}3cM|#8=Rl6*#sGHcU1VX#fgZ(6+)_~(Y6OqiVoN*B+xUk{K{X(|9
zVHM3nW0XP490YAvZm>#{H%y~h*hExCRIU`#Ks2tTtP7l5&-(y|z5}h037L8{Hc#vT
z`Av$i;mV+$GlGn*7WR@D+O0)PFOw-^$U_vbYzu^35FK4@xjSSPLbc*Dq540B^dcN0
za?gS^oRMKh8z%rh<B9M4qh{788xz9`NCRsy!01+Q^4o9ws-FX(!8;l9kfyj~ICIi;
z1l?$myy`Cv$1(IyJQ?R3|3h`qqXb+l3Pa#Y{{IS3h->~+%v2R-%Nd%oHpjl>BM~nd
z#LPFra>F2w>%#O&HA6!cE>u`^YAXGG5FqkvM)Qt1fY!ojdHV4r03&$7>zVLIdzi@G
z02lmSP>A<XBkB_yC_|m^L0Hd2d(qD_?zc{Zlmajb|MpyQPalUmhvy|5*Ix_A*hmP`
z0x=BbQNB;K@YX=&==%)gTfX)4N|_Nr(m8G%oEmql=Tbs0uOM(Vteiab9S-k%KE||}
zALg<A>YD?if|zKrmbf-zZqat8rTe*I^$d_&Ai#Y<%P_DE&mhu!hQ1bg%+%5a!9ei5
zAyz1Zhq`inP(QtvYY=kcaoQFUNX;UopWYh22(J6{rvISXmmTDu_<|cCzH+`pG{I+X
z1Hu*XU|8B|Y=v_LiuQv`TVKJHr<1Xu*^h+H_A{j?YN5#E=DA=Nl!?n>7?~wL#k|M$
zZ*C7|XuN&?WIY72a#w#d-h{tA52@#3<0A!9-}f&7T{9lKfE1QwXO0Wf#jAeHzZ&*S
zu8npk{0P(o4NuSNlKsM$K;G5``gkBnJ=~mO2b|VdNucK$Jg3eFZD9)-^bCO6A!<|u
z6fZMGhe*T9n(b^>SdXNnq=3MKQ?#SLj>v)S6jOZPF>!Y#_nx)}g_So&fa)h<6-#HK
zpWBk8KGr}OQI~NOAPmjU+o1NLy+p2h(3x`$ztbaYq8Q~*F{O}AT&4A!f$ULDYkW`R
zhaTR}W0T#N8^F$4rVv~F;CeiY(%59C6u@R9po6^qp=;yHnSChVb<{(B^VM=Ov>-v!
z?QRH+sId&CskKa$H>`OkrRd@h$MQp2AQ42;^erk7bb&Qi#JvppNPDa=6lt!@J@lKb
zTnvwSEgnJxu_wNe5}*av!IX{516?*2o@I=Iro+$9bjCFMVmPHg@xAU3(|%{(<1n^}
z?2Z8hA=iZC4<v)pkm$^5wD|%vdEmRd^n1dy6}omCzp9rA@(zKulOdgW=V244vx^*$
zZ5q=Nqn6tG1Vj~I3e{0fmGdc(ipf5yvNc|4^GiRJ-4htZ{`vuY;2_w5-PjV(!k2*7
zGom-XNa^L-2A&?VQ_hu2CkquXdoJm=OC_CPNo>Vf`SbQ-ozXt|g|tpFpv|s6H$r<3
zO{D&CdNNA%b!fTOdZ_P!acP8SUm?JdW9Oqsm71ncG}-$kvvnci(Jr=m><^PH@LsZg
zR&WxEMEDya+>p~^l%}7l3xi$3Yh!8;w&^P^n~=T0$7QS`BoOKP@HUJKKc}YB1HruD
z0tHpUyfA^<)fA*7$((dxAe3JEXKo5{TI9Ps(7TY9C*78^)KWw(yAZ|4nvxft(%LA4
zB9fcdfpS84d2uPcc@~M?#DfgIb0tV5X!4OU2B9F*tDUeMdMMv~J8$$>rT)MM-<oQ3
z7-ao=-3A+Fok5cXd{HXzG0>J)q#;*6Ln6MEhM9T+N__!vWsfEfu2NAA6dLlcq2o87
z)dR~LYalB5&L%vf6kFeQ37#7p<%^!vFeF;|Z#BUxo<`Dgx=g!v%>(#%?-CwanGJNg
z0Uqpa_mB$qbaO*)wAjQjg2T$hdv*nAZeK#d8B-K+W*^+I{p@MLfsZv;@&fuuOm;AL
zTf=pd>TlP&g<5n<8YXB7=i8AOXKkFz+5T7#0*kT3_JOwI%@w9wb3mU{(L{63r6(IC
zY6ZNCil64NAJk#n!l}faChu{0qMR8&)j_SecreNF8i<@v1Y}t)`iQ(NZ-+<V=h{sD
zNPaxV3bYI~vzZOOzd-}|)49ozc!nISuPLnEr9*63g<-9eG^ZTP%F*|eb%IRgTQ`gB
z(iM;gs^v=48@Z&g$Ue(L5m>nLQ3To+Z>Sqt2=Cr-cjyNUE+1>6^eOZ#L5nnfihGoV
zDKQX3?NOVai87^FR@7agKTw3|Tx~r6$I}4Bf~qJ3Cd0;=6e4NdudlS%6sf##u#QTO
zl)U;P5|VAo?9)8d|6E~---V=7{gkEdY1ba~JvXHx7sqv*eu{|_jm)0YA|P&AQy{4i
z!OY$Rn}8_i-AL}eqX@kzM6*J#^a3@9TldYcU;Al#(#q=Uu7kQycT0D4VG@aC^ac)s
zIqt6&Zd!QBfl%M;d!89NMCNvAN@yQ>U@c)K)SPVO1}63W9H|39R4W%ZiN^k^R^R>^
zf;j5>LlIy-!nII&4Pb+{dod9*ojC#8z(H-GA(zw3-dc1;4vdeA%?n>w=6HBaDlDhE
z->oR~+L)Zjx6^fj+!^kvsUt802N412DlhK*^=u6~V7(vQ1$1RJqWw9U%YTL>MkgdP
zbA=O9;$`y^n0PtazdX=(x;4^3goTiCub$@llAvpMKh~Q5wW?GJ_uPp+C>vy*osqYY
zWMIgPq2*~}rVI(N$<8sD+h3V#+3;hS;Ap`TrBY@pR19ID6`u`mR~v5|sOny%Q+OAS
zW?e0d3<wj<yG<mYhX5=sSNzw$u>x1fI88qK<9q8g`jzo??b*S!>W>Cn?PSDd$9^Gw
z%nUf1xwaMugp@WcxD46&;g@oafu=W^CpCks#(n<blu(^vcwzQyKHr=8^kpHY<#v<-
z!N<dN&g(mJo!O^PLL^%5Ed6I@8Sje71w0l*SQSe37<?!V6m_m(n)CwLt#tQ|S+k@%
zt@LM66BrYL++t@*ZweHefp1TQ9LMLvU6Y=wdfN|?)cgL)5a3k7qGO)`7Ho=EkB=~@
zxmVga0A3HdTbL3Z?UDL4H%6zDXrWxw(Q{k-&K!GDZ$$W$7$@Tqqbv=57&Cz_g^&Ng
zk3%VX0%!T6ABNB?`gyW?cHWxTIQJ}?{3OFLC<BlJB9iX&_e!xv%J3sZ-1f;<b53a~
zt&Ge&*wMsyPh&@;LE7EhW7$rY#cScj+yQu?8y+qq!uCs+Y#Emte3LJ*PVoZDq}#=(
zdtKi?-N&e0Ff!;L5J+3jAh&MkNwun=44NTzkKm<OVjheUsVSWKOAVBbfP}eU{xF?2
zzcd5Qi{96KdBetpFX6QDBy|Gxi&`&DPF_QF3tapc-M~KEiv1Vi+0imHO+kfc9UoB!
z?jEq?7Kg38(B7SJJVpd^UKkv7DF>@M;trwM3n4F{d0?O!NIoaQFwj@ePZ2jDd$VC(
z7&?P8*wtozUjnyrfcRw!>yrjLX;B32uYbH;jga+Pr{wSXai>z}-pgOicdSbJdw<bp
z1m;^QGVQKc_NGq%1?pqCb7@nmR?V3rjywp5N=kYyS;1@K`x)@TBWx=wrCG19PAG~s
zFf-)oZMJxCvMd>QKy8cQEHE4A7obJQz$O45@d9OzoC|SlF6A37W?+@t%7m<1C`QjL
zyrgHnsnDc3NFnG<(t+*R<}pwKe#Zu#y!6Ou0`PIF#_7Ay_TSw6M%-+Zzvf!@dVa7d
zP^ppjdF+oM(uIVCO7sKX*?;^HuP}MgW?iX7l1(1y*<GvH`&U<gZ@B0t_PWC{slYZ1
z%i~24#~jfgMja-6?!01vG%Q>tp7GyuGp((ECHr1c1gc9RYk|Y->)OP-LU%+dyym#6
zy$4#i&9qtPNkeG|3ZlEULz@d0Apa@87af$;6b!yvqdM!V5JHQ`e3nhlLk+4}>FQ|;
z(|LSpH1=MlZ9MloYg#0O?$|ub1oKdAXA9)NGoS`!4MiC^K&C~Ye)RwBLPLX`5;*tP
zdc@_8V5^ivWC^=K?dp)M;4d{Bu&b?M1#%w?DhDpr6+o~$;c+9*3lKqV?>u-V3djO0
z_YZL~61~d4^;~6-CRYY>ZtQt-D@3!0Vc>%Q3_U#+m01K;*ac8;DSS(`ylIHWTAKDR
zB<)Oc=ePTp*lSpX7N_cr#$~m&R@Wk1ljuLjCX`t==OrK(jjy3y@MFBs&jR+fbuVbG
zL=Nl=UVqe@K0V(3@FgPs#T4ir2Y~D&X4BA{ZgZ?GV`JabXFJKtM+A$zcE=<{E&AF#
zsB3LNu=N%g6|!7X<Rgiol?txWxwoPpF*4$J?cQr36g>}ZyN|fn^0d}D&GY|w`tEot
z8$NC$ltPMR=ag|Wvqx4bA<CASglyS6vWW=Eig2i`Yz_{QRoQzVtL&96dVkmbyzl$`
z^Zaq&J)fuRT-R@Wf8SA>r;0m~2{JiKY<TfCol{Z6heU(SN0V`@T-Wrh&HSzGQwP=a
zxRMzn^iKzS*CLY)c2wqq*5l8k9SItW)ogQ&_tDyU@m?1i<Op;_DXce3j0xO=|JV|{
zS8M;HOyX-)f=?(x1Dm&g37hBsCgq4vz~8?Og-RjInb{ST4xK+ZJm^+oo8s!a*h+$}
z%B}Q$<B4rLcmWWWf>K}LDc9Cu8Rbtx<3JC|v3#ViJV7GY)sR7E@bt%(>&N=He=*qu
zp&}HRc{4D&yKSF)i)Evsj%9`m=nbQnSs+ppp7mr^9%aipdu1!rP}_j}O-h~teT6!x
zqa1ab$G(bf{PU;*U}ejWFPpJIHl`SEhVDVCYAg>6mSRjEz9Ph|K5m6@@j$q{4C^N*
zi+MQyN|9Qm)?t<$cAzP>9=M+OLr4Nq6y#HOu=?$@frvm%OI!O9;jx7CFFoQ5E>#Bq
z#zpN$gZ}>Y(bCe=t0y+Xy2yk6EPu?OkoXIXM+RxGkrA1$v8weQqGLZL4>tE_UpK8H
zY*1-;7RLrcU~N*f;`a>HC%Y~8yn5DD6yCaS+!>C{qCBj4Q#(#>qNE;Br2U~^TSqL&
zyUBqfrF@Wv`uXETIT-Z|#EARC{}Cki(bW?ga-uG}#GN5%!uw*I286|f!sf^t$?S_o
z&pYLELER)2*HQp%i>yIMRCM-RVc*li0`wBis@6JzxHfu+v&3IxNrifZPXJTPVmlkv
z^D~yRzbjbz(;k9^Rao$P@(46FpsG?etZDt773Mq%yuQp5c#K^FnOf(<SF?C|R4=Rl
zF{Na@D}e9ODfZYimE%rlQ^O|2!`q?(<28krC2ACsMMFb_hpqDKfbOy;!Q6!ul1Z^D
z>n~Sh`{wmW%9zj^5A6uF^SuXJn1?A~k$wJx{5I}`{Mbg!^H=PeEE{$g?1r~mu=-_P
z{C5)@HHJSlS;w<Tdt3N%7){*J4wZ>hH#AHw@oMv{atC(C;8gBfe~s;kVBVOMiVN8k
zKnnVmtbWhF7cFcQQT~N1(rIK2avQbcTmc)iwlYw1&j1niu3bS*O4y>nU)apxi0Qn%
zCTB16SNZGE{fLSi0m=SaV&kg5KgNWEeFEud<L}w41=G;ZsmT9~#`WB)KWBcP+H=U(
zH$V>tcij8T+`&t+6U2R^ODR;7J7Uxln&^4=hl!?Q9p?p<&<VWE8FqbNMX5(byMuQ)
z2nuElcbqZ(*}{p6d?-{6q4?#=>~s}A4dr9jlKUClhrFn*>;9Wwsj{_;@Mi)7>B0=}
zP=!eupm(Hi-bL>`!tdXwYLTR$%o$s_v5?|<l)~pf{u|jTV-;D#?JAy6D#+G$swlW2
zTnn8q3AdscL?Yx(_C{XIn`HKH$>TM@o8<dfkZUk7Ici7!6H)*Adht&r+YhmVi$Afl
z8}%WI@B*qV*^#DK*m~8PF8AEReK08b9)XR*AEB1d8H@YWci830yYAdw>CKV)?zCFB
z08qJ(@PvTq61s0`X-UAlqVvXZR!!5dIgjd*SI&#+6L5|X0BO^?t9M^DAWIl1=7_*w
zHxlY->Oh|EEI6_^33c5Jgz88KBve+!%NxlD*$2t2(52c5@RVcn>}`sEO0;uSwjTJN
zo%u_Ss9R^F@~eBkty3PSa<w((f4|mD(J$6DZM1pBeGl#ALyOCN2^H_bP$gWO$K_e)
z49Om=@Q3&?a$%O@HHH?N&iTG0T3@1hC(2Aoeu!;v%{l9+8S?-=&^@5*&J)!t+h=&G
z`nAxQH@m|n|Fn`DbBr?GcVDm)jFUO^rbSW-8J(hTHka^!PO!W2q)p7@?nO*Lnjm^8
zu^DfarAilrxeMd__#z-*qDDm2>ff1My5)0J99MWIu}KWo5p`Bc;anW)y<)NvAG&qK
z*zAC?G}<gCetyibL9Eu@{wW#IY)lj=_FI<9#x&@r-T~((7|^z+N?MT*TDp-FW4mR{
zT2QE@rPwk0#>U3ViudiutEgcS5&?U@vwz{qK@pM_u<T^|3sf$>T94v^cI~M(AV6j}
zj$tk8n7`-47WEitIAxR>CZ6-OKhtGp3y%7}P5}$Ii@~>|mBst<_CgQBV*K)?IihTB
z==Yy#pi>0X`5eqA>WGg;XQtlV=HqK^8U6Kd&?kooH`D0v(Dq#;ClIF*qwD|jW_VT#
zKXHSC!sPZ<f93=YU2e=9Mv?goP;j<~&f(oXhNkdLN$HL35}?(=iz<JTC1$>&M)c)F
z(j8eXWz~+yR&$Q1%$i1uz`cdSf!pF34{2W?*tkO9H=J(Z@aQT>&CDkeh1KFkq~SB$
zF|2UG>nKZ8JXW}ZU%6ut88${rgG~;jS<I+KNVoD+3EaAk;X@z8!!CRX%$Mv|j&+Hv
zjZ^E*C2ha6eJ<?q^6M~-Yq~Ce+S%uI6sOuw+dkVH&w&ZdOr5AW%DtRsBZ}wxO%Q|?
zE5OCF5*qzayH;cNZD$ioN!qs!l*~+UW^?0tKqg)Ab!)yo-6r{_ryx*>B=S_&0dDy6
z_$v`N&j`qB%0LUr&JNzm%;Mu^Ru{p3DKvXI=jQ?ZTnP7QwnMB#<Xs*J26km(w1lmU
z5;^4+;2l8J(OXgvTn(BbLuXYD=I3d0p)0Orwl!|~3lUV8TK!H2y4If@(;bUk8))zE
z5AiPGwQwv5b1?j{??c<x#k(;-PwJo}a&cpTm76-&wp!qp!rXIr!m67F;O7vpQL6Bi
zj^N)}m)SoI6Z$iP;(;txD;$f=stf7?W=*Yko=W;}5^R_COcx3~WROT%D_Ik9C$4(v
z9xG-P*2jf0z7Q)BaK+_U?w)dbK*0G@zcsiL_7KN(!cP*wlO2Dkpx{RJv$mUdl4<;-
zi91>-MCkM_w?&&$1RdL$J{)S3*mjjgWh$ilY_+HERMTHojQ;k&t7?Ra`eR4@Q(#nk
z26uhulcz<i)AmU255CAJ#sLqqsb==B+lF?B%XHMagAjKzN)s{tT%)}FcWM49ORpKW
zD0rBh$8Ny6rSU<^lC?Sj$_9z9LCipYc(OMrJo3eq*ze#@*No-swPEj#uVk|g#8lq^
z*Lw|IoS)btpr0?T<w@aSD5!h~7Im&ikbRxs73cdW=k_<4<bd_uFUb5I2dcL(@Fy-)
z9E7CDbk|$b8GEp~O?p}`Kh{pqxL00iRD)a|1I7-HPqG{T21`_XUF9V;$dZ1)6jJm4
z%mJxG8z%<H3s*cPdfccBemE$rTtT7C2A8dJ5;T5WnzdC1Xjs$x*Ae&-u8-VOz<#A4
zzqD;+gC!8tR9;++af`ly^-u{_J(qBY@*2vU_%iwW2C=vz+Lh%tt{9EM5q%-{pSz2f
z>XO~@*Mp<A`f1<$g2D#3aTLH;A|pyZi<dssdnpJGy(BR;i;wiuJSg6TarIou(-be(
z^a6VbX0eZf`rpiMX3}h-V~I%0G~~0c1rp<~D>qcKNix<&I%qx>bxF%%Hv@q>QKRWt
z)K*fszk~CjPC;v-omYlQUPXnJhKAR|)bxK-*k0CanMQ<PrxfRMHRk7aKEAtvnhN2j
zY%_^>VWpI{uiMdS#f=zaUC>an?DI<3@6Z_qB?)Ot{aBy$uK8z(@jC$~tMt_>>G^6B
z$wyNwe{czM2|4d^)4b2(F#H3UUF)<ua?NAcM`K!CyR>t(`}&E$LfDf6Z`al8DJjhH
zdc{vf!&Tru-OMTFPa*`n0~f0syi=XtblS<AbWS8)!)PT?N)d@)l52YW@&`rYozB_#
z!r4#c2Q0bq*dZ4-d$@VAzB?RKA-1a^a`cBQ*2f+juZIB|-%pg3UUyhrIDIr)bAHya
zCNfM!mEJtaZ<V9O4s2x|8AK1@Enfj-dSR%DHFXBK!9f?9U;N!fuDTnSUvmZfeCIUp
z77HH7cUJEh`J`NWHE1Q~1)^-qT39$5%vT)k`>;JMXQ>WC>?Jxul--C>C@N@A-aW{_
z)E2w+x!5EW_r*Kr-_DS6?XP7f@HJ)`EHFsymW0!I3*-bu)!kS@l-YZ0c2B29@PSk^
z4IrY(7pP?XnIuF87egt?cy~ASoKJwR@m#qAy7Q6#)v0mB23oJ)Sd60VHxO_hYUo9L
z$BtU0|6?~IG}#2rn`zBPo)4Ojr+u-vFk#>4i5U;L3RVAbA!s5pO5+k<@xJ0OJI$+h
z<`8QqQ$yZI;t{N|^oY5Ez%gB8$)Im1#yPBQHz0b+nEXJ-5_?`uUAf@gxw1`-;Y4DF
zu+(W5f>17uxXSIP!+TdI8ZJz#lghDbK1>e3cgCL$=mzcv08pg1gii>Bw^gUoJT|Iw
zWLvA6rQ{Zx752)pa7wTvZXHmFCz9FsF(>Cz54fr3+fXSp$H)YZgy2h}_k_@d)*X;#
zJ{&Kp&sFSx%cPF8{Kt<!!l|Sqxoyezp+<<+Wx5=!ly)IntZ#L-Q|qlHbcT;>E;?Rc
z*ubnUseH<n5?;3A|J5}_VLa}hdF%E2K242rh~7~m^8+`pK;RiLjQ9}cwwMgfob~rx
z+hc2yqmnAQf@c)}=&x~Ni2lz1W3^kKiidn{hS8X)=I{b+zY4zZL5ua>g-tCxzSn_o
zd)&sD{FJ!4SC~`PSw)we7?|iW{jxw;k9Ov(ojn!e6UXo@%Onr?Qfvi<8bcTYKa4)d
zOX{LAWTp?}X9Wpn@w~?L;&U!iSVsa*^~n3imK><P=e_r9DLQ&x{s`wTj+S#%O05V0
zN#m~*QL7g`Gs#LV(pU`=cF7pV(aj3Bn~c|ER-^LHI;^bK^fWcO0b?iO8{<Y)Ij8sM
z9f`E@Tia-RJDX#e#GMpONXJ1}_?-XfJ9iPjd}w>c;1`^tMDOmcYlDwGWwiLrX)PVm
zpODxMdJS$wKFkDpTso~wPZ0@<V@<l>Vf!vN&-!}theC#G8I%52%`};E?0$}0B)wtY
zyxm>fp&}nRMlEr356Xjif$&N)Fe$uQ8sJJ)LsimyAB|D;U?E#yxQ?2*Pd9x_>K&hd
z+Yg28tNvH<6p3iedr_mH6f|a=#)OQZJB_0b&(ue9-n>{SLE|6%lRo$-1P}Rr(?gQ@
zuBfa$g+;X<sFC7`%ijG&IR(Dw|KX><I~^lJ@_+v6TJAgP%o4gVcO9TQOc>QebLCGu
z52HHv7>J9fr$<^yaZ$d0fooI-B?pf7@oNoi=_E4bV|N#_R1G2+w(QvOjCKZ_B=1mJ
z>Up0<AGZ!peDS?QHw?$r7mq<!g52}!z|Z#GO;4xDAB%iNFa=QD2d75bim_-4$=(}V
z_d|We?6x{rOks=tFp#Fl`qOVSxNd}NXI<aVPiFl71J)<H^*VOl1s~5bE?Xy46zCiI
z-m#^kq^5G;v@=bS0N7BO{H${FBOZ$;A0Pq*|0dRq7)s*%23_KiXVGqh(l|yn^}X`@
zu-@M;1fGud-U%9&$$X7x<28P4i(5`;QLMRVhq(v<dG3;=3VH{sqTw7lL3$(OW#zV4
zA7FvsTOKLpO#KDYApw|*tG_`tnx?_3PeWcw!RuVLH0zFsYr?Q4$f;3s5<60}EK_5`
z9-wK62RrECN^p8c^tkATAfq%sPyh6O3;Ah9w@r!(rv=;-QXj@k(bO!_6$Ks(^LH{I
zj1|HgV|0ge*jy}CF<c4^y$T+?@?d}D$%HBANIkm_be2=V+z;+ysQm@+SyMRY`!Mai
zTCq46S630llmc<ZtOA?8FFprbL4)g4^|B#UtdwGZVB!w9i_;Bylj_rdsv>sjN`4|S
zj$+Ktv`A)dKZ2dVl4~RFaRzYH4LAhmY<_4H;6la3KS&jf-R)evjsou(`MEnl)3lbL
zU)$sJ2YhjwUHQ`pK?}iY=J{nAuk77TI3#};{HLWn0oPHnGqtieJA+#nd&(}x%H%}0
z5#7E|C%2Yc<SNktTdf`T6v`<8ZZV2RdFi5d`fb10>;K850=rjKd_G<iVV{{L-p-`i
zl+V^9Y$jKpL5RVFh7EK|r+m!Ff>SCkX`eOnSv}=r@Z`>m1XkvPyTRdo0XO?T^x6vA
z4Dui{5p~6fevkzK0te!a#oS+H$v{kth#>DX_qeMKp~TJ(d~>TLV0_mM7!~l~hO?f>
z8`UyrSB_e9SB=axM;1S6CO23fye@nFxO^vITr0$!`ND>sp3p#YQO>mbvbLC_lg}Nv
zqcQ0P$io3`{T(G<FU03!Zv1DW4R-?L{YeQ^VI1Ra13K7p+p=7S2JIayJqe;`mTi{M
z7@3`W{-bA2-B6Vt_t36!i>UE?Xc3jr>}iUXrIfp<-cI6HhdF{kGOCa`o<}98z}P-=
zQ828H1i6U-9NUD3h>M+=uJjki4;FskG;3KXuARGJMCj;-%B0yqf4o4Vo!z)48DQ&h
z?`g$mag3@0)xPKuv-qgxCaF(Wh!4dnAIAO#=QYne!v%H^#-_^@Y8*72aDTT{5P4D+
z<gIPM7B-t?`Ti?db%ucD*%)<J>}Z8ur0?GR{lZU3_V5h&>&sVk2a+M-BHZIl4vxO~
zhJeENlegS$Cqa>4;A$}c%WSVgy7-fqrZuWGz~8~TseG3h_v9L$5k-efebU}xMVop8
zyq%#V=7wh&t@6cN<G{pW_OHSFjlJV*z@#;*U5b^5*J|rZCE1mo<t0;P^n0HBu&??O
zybr0_p^JseC?8LfL$n_XAFP_}@1;Fy``<x(H;sSgyKq$Zf(^Qu%kr=M+I@f|&5?I5
z{5GY;DYXxbHqvlRV%t+qlcWV*b3h*=YkM|<EngJm^1J!#sxo2v1vKfy{+ORuwxGvp
z6%BeC=uiyzh?^{BS$aKV;3(6S^&Dt1m!dCiL{;$CEM=+0_T&tRXb<N@6veZ#AA@eX
zg@)naQGm&~rTh(;A<tn9H!y7jfixv3O7hHzh7eIAL@!ui&vij%6hy?NPg6kJl$Teu
zrr`^$6VtvNRbH0UiErOF764;?`p+W0{_04zs#r#0#fbU#vQi}Ti@pmRw3SdqjDtvI
z(08LBusWQljup!%am>*8ZY<?X?iOXgMdZ#+7t0c2^D*dXLEY^^_`MH6d#t^su-$(z
z3vUk%7AIh&a7p}ZgKgEAICK%u@zjQf$X^dGYm^dA)jQ+;m@*QL4G>gd9a71I`VdW9
z4ovxU3%!`L2tFm=yuH{ry?1nV&HYwWtz&a@vpDdJfB>~Q*62d0V4R_;j%_k7URyAj
zY8Dg5GXjlE1uO^j?49de&D@EiD=@#{*;4K2WSKwS7qwab+*A66Nka6u<&dUZzCk4)
zd%5~dCUf??-M#vM(w`chTwkR81mS6edrD+?Ts`{jrn`VvMl!-;T|f6mC4XZdjDw%>
z>FVf&fImW2!rI+h$LU*R%SDBS_pj62koYvG5Ke`xj(U}yd`Qj6;Fp4N#j{CC89mXH
z_$E7Ep9*hXmA)9~Vo#ysz<Bxy>uybfvXr2alNnce(C%Ebh$>mw*;_yCzcRk;NZk%_
zO043d<5CYoWjeb_gy|P1h{M~T%k(-rzE(XB5PdO7{*(BweuLjSpWZw?Oy6nLD)&Q1
z27%fQD~;`@3SDDQ5;bA#__0=ROInsc)EGIqHeLR{w7^(?)?veP`s;YNIqGWl2LMuz
z41jIuYxCi3>$qsIVy3ULK~%wa$>GUgD<h=_D`iR}W^FNS!w&VwuLd_@8JPurVVR$p
z?7`f{yjhn!cSuIdtjvz{6cZ?z#JfM2>+*Z2<qOUCD6%$}SsmVdVm<%s*Atj=O6?EL
zn7NFBEeYOP{5*PJ8V<*;0Dh4|*tvsTU^_2G<$JgtetEM3v2B1HH^(eU;(9c7#OyZw
zb?}fxqS}A>^l)x%pu^)VdIRnxyO!eCL%Mk;YV?KlBCanH-|;YlD135<@!7oZj>k_k
z0zT4ix-}Ukv1a_qgy(q!2OfV(>Vq1M&!a9k6!_eQ$5qmqZW^PtEwM4S<nP1Vm<jk^
zN-Z8J?d{2VM7M>XVB!ydx=^!Zk%_UK+FL@S*aHGk@m|D8%Nh7^n`|7`aUG${obX@n
z;#vIE)qEd6GloeY%*EZXRDUn#z=CyOr{o-Y-pbrl>0kmpL{?CfzugftB8a7%(`Pf0
zH+3u%6%4pYzePbcDoS_W&AUO@VT_LbMgNzsujXmCcDn9L1JiPxAw=juE_5ano&)_U
zJ@BhjfE&k6kkm2$Ocu!-F?YpRn$-IZ119L$udZkcE($OEcua6iz_~V1uB?WpJ%e1B
zf$e1L+AtvhJA~Rz<t@)wz<D>JXHLU3RhCn<4|6<|*voy8v^FF1QTS7oz_WL1>nhXa
z^zJfO%cxQ1?EBz7R-)F{&MQ~2b+t0Mp-2uN`jUU1*+-}@XTpx7R*}_5<8*qFK=<A|
zW?v9+FDnwz!8jDzHxkiSPk8gxp5SrF;~V3vALTsoT$p}CuuZ6r4J#uSU|Mr!u9;`t
z<R=IvblknSJM6NGOpJq^1ok3EK|@>^9Bphn0j0k;#i<R<BI@S%f}*CUdd9{qC!ooe
zEJ63c{nzrp=KnR2%s?!+TNS@;+f}^qw+S9T8E22`h?>MajiK9!Ir0lgtdpd)1^f74
z*OCiRn-CtqMU4BvHq~vuWLEu^aJu`7{a-edI(E(wpOi?^f&Jr{zfaEb3=CbkQ0aH<
zm3Njs)@|4{bUN&6blbk)KPVlhK-d}V7rn@F|JjYXF@a<i_v5<7oRlwpYEq9hl^5(k
zt8E=F7a16ZYnno_u52vEN^MB|PBn9lVitHT*Cx!sEag_VCJS`n{Pjqg?ty75<`B1J
zxO_b#t0iOcxB{%=QIGny(G`pKbIDvk+k?ZpygF<aUF8(D)6Q>|TPvB<^V;|AviP1y
zpHC0!l4_#FDJZOED)y&l#D5kv3R7IvIioUdob#ZUH|U6hc%v>X^~zyGeAop{%_;1T
zgJODEPntOz4)NHyEl<ayK!tCQR+q!ZRnrY`YNO5TQwT&9sq9@!Cv(njps3qPx;sFf
zBqQy;Tb8_fjxlFSS?UM8Laz|F6c>7hIU>$9mp56GZ|dO;)FE>+6UpBg4&J3!pci~C
z^Q*7bPa*zPdI|PmkjxjnX+MMZw7Bx3VuNlGYp<C2;QQVmOY)a3;_5Ha9^6?l%C}Tj
zNu@mCGh7H>j*ZWUx@+inFi$^`yXN=#b{=Az0e+o<+1PI~fC!2&WCU(Nwq@fPYinz;
zQsTc>$L#yqcR*Wr8bMbe<ubV&8yp;rP`C^HH|kGz_~W4mT5Z!XLdL}ugxyl<v5oLn
z3^y+Etm?D}b-XaCzffMT-4->w{J=<Zz}Qig`M(P9TNW(he5x&U-=#b8fA2XZhNgxN
z6Rz=@p&`(bzmKl&-zj|X;u<%>O-lS(>gUdu8mNmc4RxKAwUjZ0%AbY^YDwICIWg{H
zP~}6%bK-kT-#ujD!pNNad0|#32V`-y#H>%bFYm!!<_VNuM}~<rf0Atq!3GU>gp;Fn
zs@!g&SXp3_@>-k)>N#01z3{X1ZsX4tVM+b5eZzzbAg$`O2l5dvic|OnTi8ouiw%J#
z5>g-`jUS`=9C1#<E`KEGD~7qdI3uB8b*b<plmsRD;5j!~4X7$cEK**rIzR;EFxW>r
z0$>6Jb~vT;QcBF7zj;!H<Fgw?#_hWH2X`SZuz<R9$jF^JA|k@pAvZt&9F!<t4%Otj
zwl77L*6E(Vl5}(}CrCz=vPf55u2*zKq<{X{wevO8z25yrgBS(qxrb;rmDqd+8!|m~
zKxScyopK?3B$uw_YP0ucEkN!5T`u6D{3Tr<R_hAk1PS2N?0aP&S{3@`6t=yi>Za8k
zxG^~(o_36mBzv`BKd2DT(80;bMa%Ni8}30Ye{mY)$S~=1a`4*&*3$wT54G$fl9CjD
z>Si<etKF^lJI2!-9%zec+JwI!7`EE}q<PFBU-;O|LJl?5M>pfrR{7-sILnq_nSIQ7
zNB)X7R?;{N$rDcfs!Twu8=N21GM%FIwR3Qu|DNzp_Jv;wCiVxj^UoXf(i`2=TX0^}
zH?r1Lbtf9SY9$6o)6!1Kvm6U*&4*EClwU>%5Ob-hsx`-y>p)E<Lp)|-4C?yaV+AKg
z`{U(b=i~+1=Ha@bkgHQrjH_T^uRh@AOV@fa_NpM^r>N#3E%}w;kM;Y7#n6Car8W<w
zI)J8e=jnBSU%bqrT)vdQ1gXutGmo#-n$u125H~ykRe+UDR(uI~_qWx!kvx78RQM<{
ziO)ZO2tTDT_lTgi$#+)-I#lR}^n=5wq0;86vOw<JXBNL^@Hf{^mX6uiAXm*6aK9}u
zfCUx*)a%j|<Rv1E0>`;n#V2q-2{J$eu--{(r)OSOPM>}4J~-9|_hzBO`TsUa_wnbB
z$uRL7m?h$Gh^W7DBGmq<bz5=j7cs7f#Rh~t=$IhM{kz*Mc76sqi*i?;oVs(XR5d0z
z!4FPDpRW1`L{;|9(479{)+zBjn|+RP1wslN;PRXw`z#3oOJ@PW`%>qpq;%#zynOL+
z;#yd51(TF$IJ;4+nM+qzYuOhb|7lv?itLdrI{#uI->O=rK;cD)vr0qd!74{(>pL3*
zHWTgzM%VFkpfMEj8pOz(6qHV9fRj^8W6E|)92vnl)9Fm<{P^Cl7O2m^Q>*zd{z*nq
zyE_lF7ztjTJ!kwpj`vP>o`!V*q6N|iaxVwU0SKdOS!ALbORfS&1njbAi@^FykGCs=
zp$+QNply;&I+eY;u5Kt|^vtc3FJKgLmC0-2gJWs6^KZcg6)Q`WB6XyFMotZ;6$vC8
zYzW~q!Zs*@Wl139GqGysIQCu(U491K_R1gH-;<T4TJ0Oe-t*yZ5aU34sG1wFed3($
zC%A{#xBhTN;$T5<`1?=cKXYEN61KpP7@nE=QzXq3>^cQZ8-vQ$bFtXbrG+D?LYb<J
zN2VT9$S(I6npwrrKOp{FxlX7if^M?+kbXf<5HkqzGzIqUZfYk!#nyy=M)gF@<WFRC
zm%*DQM!T6ge8-P$>eA34)#1bQ@6-^<Yw{j;$E`1ox$u5h$^mU{5mAHewyFZ~orCPN
zKg8>sV+Bx6sm`uEH-=zINan&=gopVVyP7eIO5a;hf(l37yr`W>>N}Rg*N`;|ukf3`
z>e59`!cIpJTtEKnF;SKp#)YF#PwU4R{^q`r#Crf-&TkNeO_ED5X3qqMORZ+Inb#JD
zwd9|6i|VxL2K(jQg3J5smY!x^lA?<hc4R?5j5T6+!vzi!UUnr0mW;<(w4Heb81S2c
zSOeGfN#={3YCjk5aWcQGg^^aMq8@4_1=Y=r6gBg~%BfL@5FyaLrot+7k5l@>qtFbr
zbp5QG_TWIHiLMRBN*K6>W2JERxSb*vW2%Fef7;3{K5eCSPy#b~;O|e-aI#G^G7)Yi
z>2I4XXxS>4IMs~WiU<RW<j*J+yVGfEG}#xC@!!mDe_XFgr`0m9TEGvVdFsAX+kCx3
z`9<nVS6}zQ3P72Z2SI8Z36A04kPeM9Yn8OoC_%o+5A#AjdY!9AZ_qhE-OgOwx{}^&
z<lFS5db%>C^}^~5qsSa%S<f@Vsb^z?7@S#_p4GxevdiWCPLt5`^4+4Vok?a+I+}v!
z(4q#pIHi5%IO^azyps?rEv=b^fdHw7ExWlj)f|6Bj9q%}z$z>vc*9)vHH&mFF5T9w
zGNq`(ImQ06Ru6sNXYQ1>bBM}f;z`MXUxU~H8`Re%!^;wDBFiRnB&fsCi<4K<;jKYc
zm#_NwSb<{(BcadBl{hN=d4h|C3!h?IBilZlJJz@~!mB#V-nYuqNAmC*roW3|GUR9l
zJW7C!jK;hLozz;^Yq8-0v7O7??7H6IhwKPMC@`@NGrY#nPlcGWYTAyLodNJh@78kd
z;ZmL+Am^z$X06dWx+YG-EJxJrN}={njb_!c#Rjy!UlE8LzAt%3`z3lNn+KgewizmD
zUya5=Hz1%zva4(#ocqoZ%ItP|(rcM_du>8$_Q-EOL01shg^YM<kwWHh1$if8^R~bh
z*dr>4xn5fka6pl<l72*2)N0^!!m8_!Kjb$uvYAB(bwq)x=)t}FvUKtEzNqccTGl7S
z(hpZls+`H4KsIwRr!nmnsjtyVoeHh5T>eyl?ayw0^H?Kj2H%<yBbNK_O}>W_VHb@<
z!JXP~yfedLY13P|F#8W7{9s~gH#ZgfFyLgJ3YC+Kn>$d`2Z+Xc<<{Db%Zn{;YI9}w
z?KA-}#3^l|Ki1;?9iEYy4A_!QJp=m{avmlcdi&=FYLyi^<aH{ePmUrAD+W!#N9j6u
zprqoPTw0*PAiNB490&v6tGUlVtoJss(au~n_FBZeT3v>{rtG9*1=%#;kVkM>BP8ny
z-?y20d3k{MGOiTl(VPrdk&{3RbYIc#q}o?5ef_j>_KR@gQ;aYDHefyxUFQhG(~U%m
zAabxHiLc@7hp9wRNKm)`Wz@*r9;IIU0{{j9iPjaRC+Uu?Vkf-k)pwmN1hdU)VZ6Pf
zHg&%+E5&m$16)f$ze<C0`T+XX2Bn<)#bVXtu>a!~pi74ih$JyZ!sPcRThk)A$9tJn
z#`;5`Ln%B1QXjKB;S(6F#l?!Eg_pgC(kd1+kP~cH{*`lm3_!09#0nE~L~mSL`MX*<
z1v~=st4AlmOUPHNNkY`B+$&bCK)#Prf=0{h<m4oXeNfa_KMS^N6@zKjZa4hKzA({x
z2J`YWy>TmNO;#=c6<F2ox=}vXJfyEFQlZ%S+F8>+)eXukmpRdN5kFprJ?HF0WnVJ9
zZl9~T@V575|6D>I>e$y{xk^1U>SC(Ue!h6p)t1V=r)!fRrYL<MW<Qp-(!fO`phM$Z
z_N3J4#}`DNr^awPMFX130?I{x7M~4y<SjYty4Xu1wOV#D^_Ge53+Vo~1xGM|KLN1~
zjI6u|Kq(Y3B5l@rj5MS_pAPY=au2U15}I&Z%>^nH4VXl}zcXJ^Sx}>9p`thijyBsX
zBNqYxlaO>-@bB%_Se^n^R3?ZnxzJWvmi)liAYnaxLNQ(lNt9Kq6=QImW_@~7fx#ZE
zd|Gx3q*g$q`kQ`a2+<3Y__}gV@m7vsNXk}xGtl4v)Fs_GH(ht<Pb71&gl8Kq>(Y#$
ztCJet)LUX)d$7+cFiD0gSlXz@FaA|^`xhq-{>ENan~jy4SOf3_APDWtda%RT&$uI^
zOnsoDutpb4tdQO2@?7BzLq^ebgVCg>vc$Ub<zVkiYfWG)SRJBaea8Q%3z-rxsjp_2
zU8_@kvyqQO?dDyZiq88+f3Bg{C05J&@{=;sW?+ymXZSQC0vvtu*gh<W=IJ9gDOAa+
zoF2;M-;8gaBLE6SZ#;aKFSc8esyvi*^$B}SqfNz{Hn~yc_c!F(NlRant-u<Q5YhC^
z%sk=V&TD#im_vK%lKuzdiD?fwQVue*GyC<Ms!nWrfVbt~*I?Y@Ygci7SsX01|G7<5
z7J?OHM<|oi=}ws=Tkf`%Qz~y4{#aNnqz?gREmXpPd}NxNm3NZlx;#jcH200#-aa?|
z!HQaQv8R&MN*|8j*jV7=M5KXOX|~_VgTDhjM$A5a<*h-)I5zG(CN;%Hh*d-LZTB2k
z#$3Z~KcWVa06FgRd@_F_qDocN7N5U6K`nqS9q>STFQ3|*ete4Vq4F^gYHI0B5}swH
zS53kGk1t-YUo~~>eXEK`=Q}1<6L*Al^Snc3NKfe91!{J){*UU8Z_t`K*UE8gOWxn8
z6&tKvhMrD%2qc{s0&Ac75CtcL)B8Y=yUDx})!Kq(h`*%!czxH`*5NjpfvcVfBf>1X
zEX;La!w-QRWsP-`1o=%_haCx&|Ca2_0nSAblN&Eie7NNb!2AA5T{NY3ZVHPhhU#+k
z&8>Sj-`~pkuKcsl4ousyx&CQ`Gf{Z+XlEtj3+*<=8c*68_;&p6?j=%eYcUNB477Do
zS!<>Hi+bMiujj-3X6$o)D`>=KB4u+IRAtx>7Z*Ve6l$xAby}v`kpR>0Lw&&Se_M5E
z<`y<<T|OX%B|fhvzmB=8E)N-#<T3>ez8Y;SMJj>!PgUJgNrnR*wN=M&Ym`%LIzzOS
z(X~=K>Js9nzPWpc`hT*O;v4~onEmhOL`m)0fd3j;kfOOz+X3X@xBa4E7x)@3++*Ye
zk_8gO^p>dsumcXdplQOP;%RjE=w_O4XuT!#S-!2*W2OWQlJ4@V#w7x)M9PmAk-&!U
zTyk9g!#?y-$^Fz|<*)w1LH!9E*L_|0Sa8&<9zc5_*1+dIz|VQm?9jQ+RuEi$EGM>w
zdC6FP1S;T6Qf56suGSB-^Q@yTt&bD-8P%YC9b6Aw6Hg<@{wO<P9I7w!s*!XoZ0huq
z#Rl#mUM+xB39=t$K9s~yfkki7kb#Lo)N9?&JJ0>f>ZiX9y?j<MX<*E~3?kRpT&B9x
zFkqef+56joa5$x~!SG%`+TN^NjNK))JdRkbKFo*y^zl-3efo9G8*fe3cwc7U`iIKq
zt=q<v0`v{#e1B8++YsM_=Z7OMgmb5FTHcTTP;>)eiA0ET6hmfDJvMQWl}apT$Hm0O
z5$Ep1!d}4P)qV&l+;bo|HCa*K+>irhS6iG@0=eqKHHHnlLjfm=$`BvweqVvnp@(mE
z{i~M~rwcHpj*nYUDk|!M&iOPFp`novLgN%!J?fQLtxMnVU5vr_1?CVZO77=1HDJ$f
z0A5=_@a4pFp|c*#_>t0T>9y;Sd7W`<G;>d(k(A9?nxt!qB#rypI6b1e1Q9o(*uq3j
zWzVtYQpSLNxBKd7`C!+Wnw;=3Snx%`gm@9LM1i(!>zZEp@#7bw`olfNP@kRkDUih{
z_xo{49t)NIpfY#08Irxq>9(n=+c4BQI=J=RWizVwfs<Ku1{*x<^U9yFa$*Cx9GmA?
zG5x<$TbKNg?+?uspWV1kT$#5$#zo!p=ctYp|6=8eQph>~n!jBxgz+&B+h{{+8K$+v
zl*iqPVdGCI$6J*G8N<1h4jH~sR(HI+L5!%8p>=jzO@?#X?M}UTA>MA_u&9~ao+AD{
z6Wrqvd%RyM4$&rl2hA>M1qF0Mietl2695j&Lyxlc<fnX0D3{={eE8k7A`k~*ISLis
zY~R5@f)zpwgmeNa*B37xA0p1r+pA-AFoB$XWC1(A&hJNc`IHt*l-~}WrJrb$g`OoR
z$LoOE2NxRMu4g&%SQ4N7DZ25{re9X9-lQvpzJBTj+!vzzB)h%gEt&n4`~Lf9`w6wI
zz(aF$^NK;<AG{4yl6KKGe(52ip1i3=QJbj5!c#Ur$4;8n6a8?7Sxd|@qpW1*fLs>y
ztwymx-0q^l^+4QW7FBI4seL(0koWwduMv8_rJJ<+RsZW=moevT%GNj)E{qGOv+sQ0
z_?eDxcK=%rsy5&C)X{~9wPiBOE<Qx!UkezCA1lP40pVJuh-R7QLW}zY7y^|mMEw4G
zAVy3ev>{Ga5X+Q;xY6ej75fgnz_>6z*L%RJYBN1Y>K5y6&@ZBguHUKVE<JX9q&PN)
zjC>{SP-pNik_6*fic~J^lM_oTqhJrdJS*dSSS2}^XIDO;GYsHt)QW-o?auyyyot_c
z?Cdtux(jN68n{yZ`~4sx&uyFqWG|9?dhS|QPU_3%9xY$+3wX9bkZ#@KFm%GQ+HYfY
zD{<JSb<|?xIHK<hbQ$*R2i1Lod8a5d;}0)12<0rdUGS0{I!)kC_>LC^DnC-{5w`O!
z*OWu;`Oc1M5$zAUP{)6HfagQRL@7LiMg-L6>4@<TM8{RZMPd^aSq7S&vVabX1F+vy
z3Y4&U>R3gDpyc=lA?bt5%Ds`@ZRbxHB?=-1snxIVH-ZJ>ZQLW84ZH5~q0+(%24?;8
zuQIDg!=n{zpY<Po>RPva&5zqBf^SRE=<Z-IbA&tizFj?9gR45}xwfCS$)M3Z<Nn8P
z;g_wY7di7Vg7hI`P4-$o%zMh;%b(t^9QrmXHzdY&OM<P{n;ysn5WWo>rYHv`*a0F;
z&lpOY#RM7>`DbkkI-K`^v7|{DfM4fDWv*XUX7}ct|9Nm1t&Mp5yc<*7uUcJy3629A
zpLN8BgpJ6k%{eE8?}~Yk!f*c^{AFf=M*S5jER^+KH~*XiC@m<P;m3LEVE1j;H%Q_D
zW(>YJRO&Oh(+-0<0Af%Of0%)s;zFph?s(O~`~s9%>uq$Q+}PhI2FuGaMQH+=DaUXj
ze<!NiHK&yG7i4ag*kz04HMczCcS#T7#0FL$WBPlDZ@NbsFroXBXF^jyeEit9dxiX1
ze|NV;HYQznc*aRmB{iB4*Le;$AvWBq^0(uFE`=*Azg0ka{c$Cklmmzwk%XxZ*l3=i
zvke#mm&B=bg}Lk0i#AxUtsp+f#0RQ$auq-Ob6-xg*dNtgLCtuxpq$R7cUg+?F#txu
z^5XAv9wRIbNKhofG+cc71_~t=E%EEJ%ToUtwGSy4i@&S1k_AkOK-xga*-I=I7FPQW
zf<hv@{T(#h#I{ZOLw!=x40}x?)kt<N**n_X+p*EO(Yf|G)x?+qhypRMqwmOuPXeDE
zr}lmLqykI<ZJxem*37y86(lb-ds|S4rC~c_H#o}N?(WKSlREUP_9DcY)SR-@zufS_
zgu%z^+5NG<9INXB5DCz>Y_c8Zn2&etXHRj$LBKO`CY4_&z-j{r#5e+DaJqW5ol?F`
z@bf|}|DCy8)Wu>Nai<pu-49w^2%pxB&AB_irbKBx2EAkEDaTz3lv6&C8xX4Yf8MR`
zYlfg|Wrm6nxL6X_W)41pXE_gm(w)CH|0z3#iQ`Ra+b{d?3JUEJOn#T{A?bA5gQt9E
zHFP7H4cyD?uq->3KAyW2T7&Y+6C;T~tAQli4GsHMbj%Vk@&GEx?I|8B*JJfJhIGN_
zyhqJ+CU}#7hlvHp4-a9I$gCPR&fQ5L<hyOB%D$CsC6xl`@v?42og0KtEgVxUPG+<&
zQ8D$Nhd>k5R+%gOy<;>9|C(Y?lNU?-|08(EiHcKr`YO!PSW7u*?ADAdkSJ`AXxC;c
z>M2m)?R~m#D-?E@+vujgI6>IQviW$2M84E}0bw&g_|CD)FsW^Uu~LqCjhn@T&L9?P
zFN6*b$qpQTaNVvp(##AeoO;;dS4`e4v0EeyE1S7_xBKzoPPA`8*QM35%2w&U@wixD
z#8nIug*XIxlLF_VsZ&9o7;cbh48F1F-_g0llVq3&c{&BQMPub3&q-7teAGT!fxrWo
z%{$g`&gSbRc88w&Enm#YUzZD`L)dNUGL&2xi~>hylE*0j1C&!_*tVz<P4&w$eXC=r
z`rnf?ipEYv!?<efpuEwyI=9SSXJ=<0T=^t>vY+KDtsV-wCr^4Fg_o2FiU+@^-vfe%
z(vJ#oN_v#9S4^bl0-$*cUqG^B2CF*xOWbF+eg=kRyEoc|Qy#cqsXn#k{eD7Mj)^M0
zT-`eF0UwJ$iH~i0kvIvlX)HMEHVOpJniYh1k|tIB;H$YJCN{Kt2u8#=SOD4l_}Uln
z1>_<Fd-khq*Rx%U$2uO3W!XJA$Fu4$B9g!xYLV~<ocnAz%WBKUn=7lqF6)@0-T2d#
z#fq!7?OSE|(hEF#q=JFBOK>?b$LxZWh4`w*gzqDT$@8(XDAn7&VU17iv6*uny2^II
z#4=ay9#Vb5sGtD9_BrnuZfz|sJlu-D0oAh@c%8|7p<ztcKQH@U@0CD&q7V;otC5QL
z(g$!udU<gbX-oOb3JS*fz?bRZE-&JZMsw0dUSF->Gpw+sFBt%7+WU${0xmyDP`nCo
z%C|Dx8qjHb!2eFc3)O|G`WL+ppTR-U1>yA}6Jx&Eq9i`AsRx`mSTHJ`*~Gux>^|q+
zDX*uLVb-xjhg%C-cmK6+C-GSg@<}m}!udNyGW#$N)VpewKl0I&2a^fPjCe1qhT@#%
zj7y8Re-IPiCl$D@5KT*6p@b3(rT-@kShG3hc=LaMYi)o7XFYK54DA3he<y6E(PE*=
zj@tYGlnW)Sv!q(ugn{L0<YofrB<ap4_u{Xm6Is)5L}gYx$k|bN(N$K|Ey5;<kPSvY
zg4ngFb}M30kjWB|j?1!sHxb*CoA8tBT#=k$@m{&{_fPu2`-`HAuUVpy*B!>P05ss;
zXnN*7I1kQz%!+PA__l!?u=~?55B;fXYDaEKMMo2(&HK>j;^%WtswIqonMDd)2d;Q1
zS_;p2Z%<TLvCoW%Yg*VAp25>pjio!%&hNpr-`;@GL$Z5*?<sY$tnXP*)nB>iOB{+9
ziMY@zcs~2XF{p|9Kork>I{*9kES$Y7Bp<mgWN?0FLlGh7lda}2-^2G`X8#X5Zn=e+
zdnPD#e!85|D(VE%iT&@hF2kh)__6Gf_27C^sfxAak(W0WSmk*1%i7=uk`b5%Z-5M(
zh{>3Iw|jIpb^!!w0K1|H<nEJ@kYF|Ok3wwR9~vha1UxEROF@BX0*QANq!`x&t%K~o
z6odx$<UUA`|Le2;^_iXBY~MA{{+Z7^56cI@@I7bs;UvYrGK$b7TnNv&4HFMIQgdNg
z9uE(>Vu?LYHQcABzuLQU^L{*Y#5Bu@RC%HpkMjuHS?t&j#z)K^>asd1U?=uE!$bYA
zauw!y$e3Nur_Lc!Eyzqi=(~UHcRyW*S<N@v9*{@ihY5L|XYUw_dFS3Yq8w*^?03(Y
z&COw`Z&x&4Lq~VYw%l!1|0Xz2wz(a;%egnyAtXVs9qZ-2%mB)_lP~@q54L533aV_d
z7=HZOU^?S3ISD<4vkH`s$iWinCm<tWfKv{W_#H@Fc@8{@6uv+~9<d=m=OU9KezJI3
zoWSRIXZBl)mlwQk4ACcG-GxaR-QxiKoTuwH1!6FEF2vW!#u;$nT22%(gqz&0M*nQ<
zS>wxz|LpuoZ|uKUUx6E%eY`R;>A40Nd+B%CzdOO*qrzS-)6%>%2)xR_hP#md`YGgj
zeV8zHWGkudjt}{Ukmp5N7>5py^&Zu=?~01rkH6d9$mx?l*`Ebel(r;^QfgVuyTcQN
zDg|iD-h%AxZ^=4$$=tvL-Z^i13-M+krW7YIuq&sn7=CgVp}FsmO`TY{{y?De|8o3x
zmZQG`1t6$zp|cHK3`LDd&BgI{L%H0)<(>Y#Dg1D7#UApjr2gx4I5%g)sW3iVz;0c~
z05$b8g@F!QyXoV9PM88K-opM25KPe+LX0H)&`^hrmT02KuX-yUdAE91DJCJ99)Nl|
zUxgK(F9PutF&={Un+y~EISC+0;{P<?h_oe*VwIJ-dgKZ27X=*dYAyh?LHYNehsxnp
zGaeF+3;Oc-O$3AG_W@JR7rgINWQnsRcPq<5bv<}x?7yiF1|sB3dU{pb?yG@A9qmvF
zmF!;u7@(@Mu?ZYce(+Y7*@N*>{LTTOx(Ei!JvUib^PYErd9>2OAvm$9YKFWI^Dx9^
zvA1~_YhF#lW>lU=vjL-EOu`eO<{sc0q&vV5$cW#oq4NDsT|xB03^W3Trw|H3LjCA8
zUW8yC$v9hx^*W@OGDQihp?5Hl=J;e7PGiapO{4JdE`BYvpHn(M+G|CEYN0dt!HW2_
z0c0&?oG{i?BE*7MD|t}ekTnwwitxg>Ln-f6+D`}o2GoanDg+z+o9xior|vc*m$jVt
zBB{XZ!PY@)gqjDI>svK$t9ZB9H@8k(TbGnRf3s$h-_*JJ4h)5WKI;B1wIu_}VIlc}
zsQ7Yz=@+=SLkyB&2Dni`e$AJ<6GG;r#Z!Kf6qY-)ei(+h&mHcg<$a+;C&@Z}6X1Lo
z4x_%_=TET?XGQKFajo3T@{<}l2m*sEg=p{{JKxYr;3s5sb|h2L^aaETgx`Zu6Ru<6
zb=W=EZ*eYK`<-)OHNLwCy<;Rw-efDE`s!W$TLXK9USK{9jI8A<pGLT&ruQ7z!TrNA
z6BxBNI?n{HG4KM7tEKOoZE&ccNUjVNR;~^JR5sO?Ra?vCfIHwoM&qgp1UrBRPd#$)
zA0>}L1<#s!?){`UyUDc>g?bK&K><1>@ixsp3})BT-Y&)D+`dnT+s*)?&hfLR3&>TY
ze!^mxgV%i{3SxpzABfKT#GT#Omp2XANiSh3u&cNSUpy)h8b0wbOM?P<(DQncr!Et~
zr_T(ZzLhSi`zNI5b*T20R<EuHf}I*;4Yqn5IO-ele+~1u0(g4}L27xh+WSc`qOp`(
z&u^u@s9qtyJ8U*BjBWDZKv-N#VzT!W;p=U1IWrOsHxG2x`I!f9S4Xm2z}~n1G`LHd
zdNWE$NGL@Sj6R9v?XIaj&M1x>ho*iRGXJ*z{E1cl{Q2`kEiIViv3dL>S-HXDo8K-X
zVrVeb?!D*!XCP|>Ho0T+jKWK}oNu=MF%D^4tv$^{EK)Wt7dN8TPvAX2<&*L@V&xyB
zuO{2;g^v!*SApFFp8Cq4Tv1N%aF@vBDksnaMi<#_-H|+Aw*RuAz(yGU)1hw-uRB!M
z)IhkhNUW;*BCJYVFgzZ%uUXB1v2#D${-K^;7@Qg?5T|3Xb>+IfJL&Bp-d^kZM{rWB
z@PBWAr4xgikjKVjTgZ1r97<FZ-xRz>bOUGQrzHFCh{uXZOVK!8VRJ6kKCn@bHF>2L
zjA?m^OP*Sf^4SGzt(QPaC8INPp)OGZ#z1~x7{o$<EB_0C-(%jfa+{&Dp<j1EO%L~{
zP<V-2T+s0xh>&cldMAtag}l=fvf#kmIC~(tdtHO^FaLF;7_G=@LyG$P7VWq+$Ol%}
z(b&<vInE9Tb`R19w;Y?yAXh>GRRuL|Woawy<y#f*x&!1;`<3$iy<yo!UWl=DsAj(n
zfQvZkcVxFbhzIgdS&qrjP)D2<QQVoDm6Edvs0&%;EF&`@aBBgy^t71!XO0pxuprT)
z{H6rlzR8tW6%eL(*SlAVbDH%Zu!S{BXkQ<PX`J6KAP@3g<;*U*2V-0sxk|TaQQL{=
z=RCWDApj7TJNzpEdG|mE9RPid89=w&6@VQTQ+}Qi0ePy=W-}cmKHt}0-nuS5ig18h
zU@o`;1oXc7hYUbj(!ZLus%M}XygtZx*UE3@w=45j6kqkF@+X!DT5g*O9z#|f__3-_
zm(%=qycIE@GxBFCj_Hqpj!GRni^lj}icKRhJcqwdu={{0=FFk;`bAW)5!$s5YQ#=}
zkFROVoe~=cZow#BM}Ub`Ef6{NA}D;y?4kZOGrNxmK&Hpnc9#sEl$SVg*O}hY`>yQ8
zVdI1g`4#c>6j(+mZ4nNi&JY}-FpmU;O=n2U)VzaZY;3qn{l$-qxw-5kU_)B)I+5Rm
zvIx1Y0te#Xm9d3rTzFfKnE&DzAb-xl3L>A#hpoQosp{!j?OJNK+oG*!Sj|%4<F%qP
z7x{F>=4m_JLB+v$LSwt!%uZ;_FYS8))j>tATG{JNyX(TInTk_Qv~Qu0_*0Qjfs_kl
zouDB`-0kqK4{>Dt;fmUd33GImO4720Z$$rpt?{O{jK4lagLwPGAxS4LALzf$$=4@8
zA-YA_|Hk2Z9)B0hHULAdDfV2%ZD0<4fuSaa?<(V>8O{?Z<xWy*rk}~jx$c)y?%t8`
z)WccvS`->R@vf0E@pa)*hYpxEpo}w3zQBKg#LE1IQqabM@YwWWnAqDwa0XDR=f)3h
z0%$x7!GTRcb}@53hr_=-EAV_q!FKoyp>bBz7JOADt8!ctH8!Oa(P|F-N5lytDKWo)
zgbRh)aOT*&Q02at^1+|q7UDM7ce(v9z|)ODsA^$ti2qB1MksN&<+7{@633`T@IYu~
zS;AiOMoZlJHC>`TJu^7hfh3Pqm$VUn%<8{eKMIyp%F8={MlIVme9B#_%u$%rhG7xw
zT9-UN4JHCYMz=(JTYW_SdjPEQy%Rops&q4#{L)T#4j#U@bcpaMfNpoDs1E_5%S?z9
zRn>=4%-0c?P^_xg?n*dVm2j`kg9i&jXV10$2h&q2t>`3_=I4+WG6Umrb6ycH0`V9R
zW%2qb(t)TwD2jSs0ONFn5E1TSX_P+zY&T}?e%sD<wk`VE2c6@4D~cTuz0P9gX%gT)
z1AXO=x0wU$4I*5zburGDTzfDI#-X<6cnwyNHTh&*b)CJ^xAX8;w8ij{^NMPs`CUa&
zfBJ$t<zMlOUba8u+m+BUv!9b&Du*K#4H?*{wN|#3=IpQ^7^=yn4jI0h$z;-W7ntV}
z8|`vU;uE&ATGlJ;qF~12sO=OTvkipFyYo;Fo$!!Tud<u14|YW2k-3c~SIv+6FoHM_
z3JbR<OM6$){yK)_COYRNdLauuyd$m1=~?7(eGveWZ&14};IC`yI9X`iT)=qyOV0@C
zF!IwDwp7g|o{;&ZoEZ$6?2G?nXKJwNvN=FKwNHeTS2?(&R4&YH<6)Cc4l9E>Gp3&g
z)2~M0MzV2%;CCmLa!(^}H|G&5cjVNbqvpN%@qd4u`=&^tWbCFu)P~f_LD!2Jpa=zA
zL~g#|+y`SSHAWs8hg=pO!k$PZ>Av<VD_Ziq3;jq)tB(RYgWvv3mIDn1I{T0>TGgqU
zOxqwZJkqbEZgpm6CTIwjE8m#(bQ6;lN`24#EQdgV3y8F}N8+Spfydch=`dB5I5d?4
z5NsE@H|K44(m`flgA(u}MI33TeD|0Ir$MMxIvKRQ=!&rBR8xB0?VbS7q5;sg=WvV8
zv15&|M!$Y@=FqRkLLEOZP`=`}Szv-b9Ni?bxd?%)CJmL49I%h=7@`g%Ir^YgzGcwL
z&@SzXtW**(Nt$8$^$5JMF@Tnd7+q|p-L2~btoQ;qp2Y|l;7W@G?$7`46Uu@oWPl`i
z_76xn-$X10ARYsy90<lnv?L@X01=&vV$OT8tU)?LHA1a^HO;nH3A>`C!y(rt!jiyG
z<}6xRQD-!fRQclH$#ec9aCR;@87eZN;!lQo(H^eGPb0UsoDfRJnSwv?;HDuHth}xZ
zfFwK==2zYa$;k|cfSBE@gjU{>T=5G`k{i?9C#_tW2IUvPv?bdv9S|Q6@WstHK)r~Y
z<AI)2YVYP-X>r5Jd6`T@Ay@FiMuHZJ&$4DU#Xn(9GZe5JYxeKvUFEvGqN}WZ>wr7(
zlEYfc4*`Ml?4Pg(D=2_K{!hGyE*Vk%5rId5X0Ftz1j;FC5yh3)_^R%|=R^Geo=;0R
z8Kxhk9Ec5_cn>f@5S^BC;QyS_Z7<9Ch=Z5>{lr*RkDrS1jT!v>a_p#fdn}g12k;{>
zj(oPgJj6eT8^~KFi%MIAi%ttnmjSj2SoRq9t_RQ3;CMj?BTjX}@nXgXSUx0lOM;!o
zZC6(>h8N5DV#*Q|r*?K7vgyZ~xoYK&xX4S3(RqtmHHSDb))0q7d^g-3%GM$^is00!
z^K7zk$5oqaWB!_vq9j01ud_=69W=|tc6X^Q8|a{XnEn_z)6cPgoI=EA>`;-Ndw<4d
z3`Vkh;IGxNXoQJhV;G}u5GT+i=ua+;LXGpFT7g&brli+}@XP`Vg@GzLc0tO@@_h<`
za$#X*>oA&3W)&Ss^Z<tuQBk5rVc1%}yQZnGZt7~5^J@AAt9r!C3s{jAQn&6%6JujS
zKmAIk0l%X@KLn{9bos;E1L+|4pCMT0kO)<){KEH;#C5i1r7ayqJ~JghE;w+wz|y5x
zuLujaiugjxN2o~ppguzept}IefR0Jzmmt7_S5E+nzdGSEd<fi%GHVC`oCt~>&-C6~
z8(ja$>?~IBwekB7SErSInemV6e&oJyb{9&H0p%d+n$`D%iZ2BU*!>E_P+`^$yW#h&
ze*f<FABcUkPk0a?0|I9NAj%cc<zU+jmO4+bY<r{p#l*ATWZk^}pEB3hfZTmr_}1Ed
zg>VQ0ScY-^0LB*KI|nS~)?4t^Zd%`Kvf5sWuk&7R_lMAU)rpz*;~1gfu6#ACQvcuF
zVH%9v1L4Xc4w>Ca*&BF`<=R_6A63$<^QN15*C2cvhz9m^S?Ff_C5WVp8GB5V0DyaS
z?&S05>)cFB=E#|@5LVBpk%*iMRG0TW>R>KD3q6SySqH0ziu-sUxqsl};ZeB0SlF+~
z^!(H$TuCk^`T~V)16c4^E+R^C?iElaoY29@=u!G2YwjFm&EEF&!xgwr86T3pwbdKB
z->s`L+FW@9KX#vPYX4<9+oy5y!EjDU3>&bssOH)e1OMRDLPMV<%O}YX2U)E*CivKY
zjFCb2nUhfdl%V{HGhTyiaoNjDgPMS!l+-9r;!t(%6g67N==2Cq4$ol#!?`WvY4=A6
z6BLDF#N99=@Qi#f@A8$%5@+5`Pl(#0OdGhe#~^C^ZZ+gdd`uC%f^)REhG0y{Vi`K@
zBt(I6IWeo=t~_DiV}3ll;sr>HOlQ7@BO1Q^|HIRF$5Z{j{~twG2qA>*6+(9QKK8M<
zLiR|>9@!&%g-AGN#~#__O-5FB#vyyl-hJ;^pWoy6kH3<~d7b-p-Pd(Jm((JC_`@K;
zv3&p>Q67U`nU`v&Adq}gOq&z{gywD0fEr{`p!(qn#<UdnHY7SXe*O!rJjX8PM%sif
zPVIJQhtd5Ymj-Z8em4FEeA4gWX14_1-GH^00_D3di&;lychy-pI{X1#urO72Ztgl!
zf-SV9pBS)BeBzojQ?_MAbbV>(k-r?bb<|$jgF!?8N-xE?ouZiV?9NYUIpPA+U&Io8
zA_`+8B3!gjggw)ZiVt0M1Zur4h~os<$bNx`1g`%wsX{bG3PN<dbcliO?!qgIF!GjM
zqQFa_Os)Hh+L|FOz|Z+FGC~4lt3bpcjROil2H%5K=NcFwT*CaaiokLgsMd1tPlNfT
z@_&PTVh;Zkn~$%;Hu$Qff6w7WvXz*y)e<vouTw|g(lLnb&G4E{b5J|a{$l!ex~p;H
zQ68aBYI1VT*I2@ujT|dM$n6kS@mjEXV}#~odBZ9dBwWEQZvxI-&w##g0;OMWHU(t#
zBLJ!>Jv_`1aQK&o3Mdr@b*e*w$zOHe#+Q|rRu@P?TjwP+Rv{q@e2NS^r7G}+83<ip
zga2d*vU~Lfn|;WaI%Z7vGbq7{8LPI|$1Gdt&yKL3@HJC42ciO9@Eru{CI~JcGbzo?
zijGL&nLf!5bg2=2;lkr9vX!Y{c%?V&p^e`}AEnRw?*&H&AxiMt^V`)vkxo&yE5R<e
zMzJnjC*Q;v{(e9_vgok2CRMgrm$Rn{dMmCO7~8**EGQzPsKW+4(Y=w$`7BWtW@4{k
z=l>LTh5lb!U_t-BC|I{3m?#2wE57Hr+&`Y?CKE_XPBVW$q=zh9RL_;}1HMU&)a$N=
zy9J_lKbdA|#O9X4cDWnupF;mdLI~bm9cpZ)!l{Ke3nP}DKqC!YmtM?f`35T@YMtk(
z3Yxk%Gk052!Q)VO0eJ>N1fvhUmUqCekLc`r#5+)16au3Gl2c#13kXEMO)rVz2T{NK
zKK)WIth;{@!(%}GwUB*{+9%+-%2T=?r{~~ZR*X4QAUF%#FBSdEqHEg3yBf`*^+NRD
z*<+(gM)1SdXHX2e8YJ6fI5(RjFrSRFp`~5Ygkm^R7;F6p0*`4uIlL6aG_M<4>}^of
zKdd%_P54U7My&+C{A&7%1CLhS6uAG?(xljPsEIZQfBb`$qx3f{pAF7SB{PVMG?2qz
ziS9f&5|Up_{HCppI*k4^Q*!QUw-`rRUR+SXQhF;C#8q|!ur<apr}Kw~_;K=h%%n1l
z>W4sJ5yVm(1K8&;K&70hBS8{<Il@TQULXDgHukQ7DE=Gp>gyUww*mkL0C)H=ZMQ@<
z{c}@JT1v{a(P^~qA@A!slA1xz_vzxbW)Iv+J*a^_aeVoIQ9BBAmdn@g2$cJ`#=bjb
zGBgwS3p#TgjihcB?fsx8MmuqEIW|OeT@|5bayUvSx6bf@+2$?ZM8NSJ(uSk7;Vr<#
zL5&@0ekq9U$`Ys#dn{!73ELeT#P%|<(R=@*6AFFl9=;0}LejW1WV;99|8}3-n#YsH
zZM>$~UMlLb{I8e(`~cQ-gF?=;!f!VJ!)0rc8sdf`RWi9d`&+BsBIS}vB+kaE*w!03
zt5`)cU|9+!+Q(ct!7E0%taQ5>S?BMt8=LICyhN%Zom2h-<|703dRZkH;=I3^W{++_
zE`<fz!=?^9o}f*s1UJwkm}F{eZ-6v1CR;pfrX7y$0|L^6fx9~`gsnr=D`>J%r1fHg
zXi<&NG)6z40TSddqtkLVhj+h>&hkSal;?9VL<7&IcW+#_x#Z)vKu{BdLc5S2SI4I=
z!n@c!q_vYS*Al*n(h~cL3DjK0VAMI^`j%X+YnmX#3!X)=C0s_pi_KFIL~KYx{5SP<
zm<x_zbN+?T^pm0q8?-}{{-f3B?Xg&=$Jr<=)DQv;k@rtRo34N_A(#u9e1Yx5K`Q7V
zzD45U;r(k^o`AroQD#lNoqu5*`&C4s0bH)lDJKMfc_wJTtks{bhP|wvRmEm8%etCh
zLf22Q6&I4^?*rKaBQ@kc0D_LFz#vD!>aIsC;$|iD&OyAT1FQk;+#AMe{C3ty3*%}D
z>Whks+JW@s6^JHY0*jG4nHiur`FMOOX3-8CUGa*&$$q<SyJvJYw3XV`;;6#4hmvww
zik1UL=KUSN+QrpG(BySwiWKOI2hP58A(kN?jP~{{3&sJ@@5g^M`dd)qF3_)69*RaI
z6wL>@Ey8v`EDD~G+<<nM7FU=9=?_SWKwPQ`+krXaFZjaBD`kV6P%sk^VFK^raZLlr
zgB0#GR^W^%*Vi2vP}JC_e4uLr2&wQt_B1-n$d&1VeH*X^0C7plGLYdy%AcGXPoQ^_
zGyARW|JQq+UF$#hxmi9Aql2QJwBU~)pFRB3g>Rs3oH8=;hg#54s?HE%(doVgl-Hg+
zkbNAGiUb1DV<$VbSAGa^r+<>IQ8l4ONCk7NIaAO{Moled{Ld-S0CWPiMv!H!$Q?jt
z>by<#EPm`r018WrijH1X@bdD40%B0H|3TY&2?)R_{R2Ym-<f*Mky%?iD%l;#4X|>q
z(esp7tYGQlznkd}aXi9t^6mIFtk=L*4ypJc@9Mm07f)U9i1*7E_E(u4LOMgHRnp^z
zmQhrCmPL~yvz~X0njSz~JHP$O^5hG3cK0-dea)^vk%b~G?e+&ei=+)I)8+fbc1|EW
zXNJf%4Y(|VY?1r*w-+RwFEasphez~ozXcx|(kd}s5aaK#{S3fr&=t&-1$HORgL}Le
zjTirm4CeXy{q4ga^2ss{`<JCKw1S@rZeY2B)_-t@arRJk`4xkTm-OBKA{Z_^zK-4Y
zM)m4K&j%&=Z3b)1EZpODJOocf#qDi@x>*=>9s6Jn_zf2zSrCWYP{^-6uc^4s1wFrM
zReF0JYMp^PiPX0Zg1?~DI?lfSepGx-E6S{e3IO2j5c|Iw1nPhbC2ZpWxIhnO*G(Af
z)nWuVnuJz0PtA>F{wx|X$NIgoK#jgu!~M0|{YnVSHq)@~l{PW)@2{H`z}eHt?|9LI
zletApS5JSU%t#v>DXcS>K07tyq4MwY6-hyq5Vohin*mfG#rhpg3m%xyW_<?CO?7fZ
z^Fpq(7}C??bXl{^t~~*cnQw^4o)x=P7l;)!^*%$6Q9+s-;%Q{^FD3@M&FN|o=DG_-
zxe4n6OaBd$2sY(^9;COB@6z1Z2@i;|-H4H?L-0kmX-uqaoSK9al(zpHwylu*HNUtm
z|Is0>bg$cB#E)}C1}A{#_*s>_ogF*mmD~(gN>hMgzsLk<e{9w~g=euPbpY6gNmcQC
zPW5xE8(P|;55(P;l{kM~{c<Qblr$H=7;<(Mpx#JG1O^6KDE9z##ZSCT@(Z^-iFwBb
z!C??F6E}+~dUhwVrn3J2Jac{e#h0Cs)3q4ZWoo3DiySsOY~^<mkFDuewrH){_=Ij=
zYo^QVsl~~U<F0f*0axG+H&kNREl$3WVs}qLz%}H}F%}-dh@jy^0$Wr~&KFE=zydsu
z-xStm{UET_l|tx796YVi%THkVr&fre!YTwiycFm$;ev+0I-@q>^k{Ruf25p_TL3Uc
z<;=7L7G>D8ek^Gr)fyOVO@9&U2OOGTdHgA-oG#qW%@_b$O~7;_on9W6CF+6<*TDK~
z5--ww8cJM==5EB9eluJY&`R`d9|o3!{1E0kq165lzZ)1iIz@qnblt?Y=ZeYLn4^r@
z)bS<x?RtoaS?}YlIanX+B!9`~s7*pO?o>u-$Gt$ULuF*ZZLj-$izAeS?XS_;nBHZn
zgL^zsv}X0jjACy0zzD(iM;u;!3@;W+Mig=)8*fkDX-|PkU=9CX|MEBoZ3VWVIQHp?
zO%ICfp;<jSVX@G|2HyB<kj)C_TEqeuXz?yhT|2PWXC@mg>rx*ZzJLb*Z-4+96%;hZ
z-<{YLLoGD;2G{Jc4*W`UH+CCBvxLv}MRWHXY4vfFOzCT`#~^UZ?3GCD<tFo&FC*x|
zqmemZ6*xS8=MCEdI2!VGSk&&=QxKaDvcC}Ulj8fl2c>J}4Srk(P{nsBFvt-Zz|xrX
z2NJje-p!rDthZE+Z#G^uK34qM)&!7-y4u~eQi@xxv)yYawjBo7UQ0_c_;PW>-4tI4
zoIGX@ME_v=Cj1I?QPLFZ2y`*%5!ye$b9U`;JuWa_(kKYGUcTpi_agO{g%nirv62yJ
zPNcIAe1kCcf}kK<W{r8>vSHgj84wVh%{%<mf%LD7g+boC&gN?z^<VN?mhaF$#jFsl
zO_Pr-xEW?9W1#R*;c37J0!^Kqvc-3O7iwM*ZYgm>2~RMD`GZWXgvqm`CKlwF0$VMK
z=9U`8vfS}=o@?_%phOL|HZZTo?UeAv;h!k*cxtAb;`~h&vPQbu`+Gw)T>!iSzCwN$
zgl!iEoo!%DxdWb)q+`qR1jIgotGM66>lqlRou49nD0~kB%fUJ;8Z76&8Lfd1u?wtO
zBB7<+QCbMtNr54+gUU`t%FvJ!h`b{Kq%(X$DGtQg6GLT1O6D#MYJb(&mZzG!tG>F4
zAGjT6?ktoB-TJ}>@JmjJ!%;UTAbg^@n28s+*R3Bu7s&I^lBcBYU0<8}_=WzZkgk3)
zM=Y~j1(|Chv{eGVivYZM1zubTMz-Y%+V<x#vmf7v_vk?wW}4X7%IFPchM#?)t~m;s
z3#lWILpN?!NnQ8q0UDf;_<B6s^9%10$ll;&=h)bA4Y<q44l%fu{xc0YC?lj1UHG~*
zE8H8$cs&WdvB~&t=F;bacVuWUzQwehM1+X0DAp&wJ#*h7x$E`Ue^z82_HoX}*62jn
z)HEw~r*Q=A=3;;!7G&AjBM5eEoxsx1ZVp7qq;eUDt@Ni^E-75I5<r}Ix~+=x>;i>w
zJ9(yD#+u#S(k85}1yC<a`ET4Ik^sa#<+UBodA~P901AV)x(F6XSiTEn)ghpiK6Ggt
zFeC(1A%*eQK)FdYTIh?Myfgo3Nl3m4P}M86us*&^?bEUv+@ywqPA!QBOc86;YFk-O
z+&A*~RoS%k)r11>rB~D(rJEd_nV018hS7`jIeVs{$N}OU2Xc%BR=%DYasT!Mf&~NE
zrP1y0EhsLSpUfI52$zYL2)kUrw}Hl(1+H{>%oq8Tt`f6r`r?}4J~0f;+l!3gv%D3^
z2|RRm6VuYd!B|Wo8(y&1=vx9&DS^XK8e}@e;@6TrXV$U@L>JXB%?t8d01m^z)2Hk1
zFX=X>fRQTu+^<@h$zBW9Mrf)s^W1C2`{UuU%9+)QiOGC`hofnMO`B>~ozt-Zt9?$>
z^rq$(Xa{2*kwhF~paY!S{sQ{wwOp`Iok$i0j+#;|kOwpHT&r;&s{$e!NM$$Cu>wHu
zIH)8neUBUJTWBL}*xDU`+gyD_^mx94q(x;$@vXhfkCh2*v+H-xyNfgq#4bIPfaF*3
z^C#JUEjdY!-gvQlNx@s)IS?{%<1Wf7Ls)xvx2B$u#*kkdTn<`BSY|GtCWVMMFc;)t
z-Zf4xo@s<n`Pa96M~|m`=GyUh-O3`K<s=_ua)-#g$9w|YJ$wfatQ<dMcVq;cE~q5Y
zJ>}efYAm<keAO4~nFb*9!W&>(s30f|7{MqO-RV^@V{`)}C5Ohl@I25<qu<~rV*vGr
z7`Tf<38<@_f~u2$@_3qq@Zz3mH!b5eMfE{YGj1_JHN#4D8~J@r4zu&y`u3ZdAibgi
z&TnT8=!NmFwin+gOnEEvY3ew30>x)eJIIl}YScfAlUmCZ2;TIZeyJX;Y8=l7aB@Na
zD^vob;tL>`te6kELjS#{TOIoJsm>3$EaUx)U<B-`d?z}yY=2uIwy$%(uz3qw_60EV
z43@OJ043-b_3v-5>ip%8SPydIU2eBkfz_dNR>Ak*5aY)r`7%$zBMIkB0iR0<O>D3U
zt2NWrFE2B0DeE)Woty2q@yyQP>qpV};bUxnFE$7F4xb7;GLFWXB*FMrc^AbxhVXv&
zBG}=ivLh#=dpuxu!A(cET`>bGR4HWWZ?HTvUN#0kPWYuxD<?Phk;3$<QI#^coY>Uw
zczojimiMZ@5QKOq5J2H_R(SyI1Fq3T0DQXsLYdrq30F~h1?+&0_z_7?JpU1+7#XO2
zIb&OT<9;>9>>H2zyi$_6%P~C7Yvzw!AeCNAd<*P5x)l3Y7e1p4m$u8zI|S*T1#+R=
z=mMOKESqT6aqso+of^~l0V1OFDDrEo!!|O_EUGt&b*ZGZtZGg`fNIXkz$4EO&<3A?
zB}bUA_EU}%aIbZM8zaauSMenU*>ooFE3I+wtW;-jg|vd$h7Mj`vJ2Oco=a+>`>(Tx
zk27%h?u;jU0&?<}t$~!sZ-O1whc-Pzn~hwdz5>W5dsOL+W<!<t;z7xWl|<6Vp*}Id
z4FDYFx}RA}DTu$?@SWQhPXZ?wxMb!3b25EVJ#Oq;xFydt`G#0WLymAF;MBFy12pO%
z0hbqQ1F9actK6?89t5)M&`46&-++Ko8)V>&doGA69odr=3skmXaJ&;kEB@;Zu5+PH
z7-R0!3U!v!W9rGzrssTP6J>E9k}@;cq?pfkwBVnI)1uYot@Wyur7K#?s3f_raQ_+~
zh^g<@wxAH3n{WK1az%iRVukaZC}h`Jpz^9#I{1!ZJ=8rxoN7q!P*w?Px%yH#mwqWo
zOTQ!Q9vh`z->nA1-}*Y`;$u+44OyR7^#T#5BT9dzytBTP;`7sTHhXNe>j3ui&fUnL
ze*`bQxiEe6#_k8SPnuz)i*&m{*lr&=>VAd0c^1yZ*kiK>-zA;^y*n$xorMKXJvy&-
zJ^^9|oY)XEhkw?lCL);=xgD1cR<pVJ?8ujxW@#R`U0oiXZM_YBw-FS;2U>Kpqn2B{
zdBzx7I1A55L#@P!M`*~L1I7GGlilVP)PNUjcKU#|H?&4#GiF2jlOzn8KEq*k0O<ht
zL1MAm%n4+$2>{VNu(}7cGtr&94=mIQcNH#vl@PL#CWvi(xothUWeo7UpQ5+x!Z$31
zvo)CEHp3n(#r2bKKltyF)Xuj_n2NbIUBo-P@kdXxez+gy*I;0-6Q<YTRadVIIQoDV
zG&VLiH6NC?p<V+tnO$vvuU){g&Au_E7JK_ez<vX0Bif-9TA;weiYDi(YW)s~&EFvR
zXTX1On9dgW$<58Z1*9_*7{icK$N&1mu&q4V8;oF(%}JbqLlx)J1jrdQ-tw#VCiXXL
zdq{(CBtY#Gv%N*_A22Mg)-Xk0H!fo{58`tNc6feW^R`9_%eHEjM*Zv%`jt@`+y^&j
zL+Q4Dtj+Y#YRYxqxpVzO_dfGjziavHtBs#vL$UA>cY%I1)H6L|T^&SX%5fp!YNW*C
z?)@<^ZV;H)&ZnU&hA#9LBo=RCm|e3c_L62tZn3_O9krZaRCnI7MQu0`GrPu_fjOU{
zk1y^CDDI7h_qr5^28(S90kc*F?7zRHN3;8^hrKQQ=!RF;dSMTq5ixcRDN<|c$W?&<
zCC>o*ZMFMNE%k$@Na+CME6tut2Gx^s4oRlAdow8>l`S#crrYJ2TmbkLEIs@K+{@#@
z6*ju^)eN{;-XVNwJ6H4TmywRXzVYbty0Nu2Ghiv&#^k$i*&5ZKgFylko!OZ7)B{_>
z^^s09(iBO-GAT7vW;efM(}_;y;Wm)Fx-}gpSQB~2H<LD+b%FZaws|guIR*nA(s&l;
zqU~F&>AuBjGlULY4>E^)Eib;$*&B2%40y7hDZ!9SlzVm!5;VR75dBfPirlgyEbjJj
z7T=sbHSC|D_$G@qfgMH|ez{KSI|=(b{DyuSE@yPx^>tyOvTJ!QH~xk9XJc{Vgv@4~
z1@$z-RfZ^F-s9__ysX#Ao-w+f-UYh*7p0V$nOPtS2=Hp>_}jg}yw#QC^7}~Y1Q+xy
z2URA3+By^waCvzVWLj?&TuYtkV64K`M@FaPP6kfRUE}1s`r4S>T!y!YF`W>H#p`f&
z5p0(#73MTImseMf>hD05_!vG}4=oAc#Iz6YCx#hlF9kH{>0sUjfSV>4mbvU+G*w#+
zNkmL5>ZMF;OeS250$oHumnusB{l@$ECU8vH{*qRiT6#1&=KRID>9b(pYLM7P_r(W3
z2f|qgPU2~8_T%ggnP;1#u$F&7!fsx&H(c&$s-bYv41#TbpO^zni`v)n%ecrfq5}M0
zKw7Er2%2c_f=LAXY$R+sZ@#PyjYwug&iXNDMf5Nq9ip<&VeQ2Q4)pK|$kXfUa0u{c
zBOHeb2Hyf%k?;V7pnw~T#u@k#h<(4~hfPHL9;%A`kF{ryqG{0l;2>)2r)*{zosDO-
zD|tuB++ws$RHIvofRzxIKfani5qu-0dq;oi5ODjS1E-%ZFsfTN-p@@osCT9z6kUJh
zknzd!DWCz&z4|`U_z9H}ch&qY{xa8NJBEB8rO~-*Dn!wp@_{atd+I@WOkV5)Qf3m*
zk<%VLI==|n(x@uuFNEFn#*Rkf_D&OiaFqVI0Ch5&P9l)a=!4L2OK}|bzWU&wYaU-`
zKt%0=JVO;{;2|*h1twR>WN(&xxvxu|!3F<SZGtqlBg^gRE*?YfU*uhqge+59T5lkW
zTRs0Y*|=}icLKGy;2i~znT_nh0kc~<1keKwo@)kJg2Bx8K}1nbNeP~MzD?}}O*lJ!
z9Pk}s^9gDdk?_NMDR4iP8r8F?Q&LASp))bvb~6lwfk4?Njx}x-Ufz0v2QgE~E|}e_
zp#%q)fJ<s<g=WRsH0K*&+;$6HX88W?dDp-l2kNI?s+lAxcHA>~6kui`KZma5>XGSF
zpw~No4_j~5Q>HGD=e^J??b~z*<E8LAHSNJT1ye8&oNv}4Hrr@s;N*V7^MdGo5@I{j
z2VVW{xIpx_We$7Uo;5pW$5+a3Ex_Z6zl)5s<`FcN6CBAh#CW!JkiIi3K~u1+v@Ph1
zR+dW!`Jd$`^DWspz-6X=K&Z*<ic{+?HUG=IAJ>7xA(#T{)Irf^{%8*}IzdcBl7{=!
zexQ0aBt6pgU(tu7ZjJH5++X&)31*qWFYZ>f5zgqpscrEF{ak!nTACx7;R;ZVLUy*V
zPUk!T+p#AVVk>;GjO~DWZpZOTOU1hnp0rOi@=tPgZI&DZxc%wnLt5I3m~9j4247!*
zXA3(%17Lv<K$<+$LGX9y?S>ICm}04jeY+4iIu%YE*7pT-%AApHRrW1y?9gK(*!$fd
zUw#hcE&+GyBj?_t?=$TC`96dxrho?KXELG|#vLl=elNin_+ZS=`loCe5l?r6p=DVe
z=C^1T4f}eHnE*16n)_7btX2gwduk!|!WYx6y2!f`N&q^~6d3(73VDJ92hCu8)pR~A
z^>J`H{f`>v{Hhc*JXNK6E?H=zw($`U+`wanQn!j^<^*4nGpA0xnbI|2gb04MxQJ=%
zBjCXbZx-c!8D>f_2Dhkz^l@*}v19(U14LoTkq!n{3g5$0^fKpDn+^oun8n@wPr+P*
z0Zv`1>G@F5yFxH_w;30d>l04!6PX@Z{&Ys<Yj!})!KMD0kREu2#ye>Ig3aN1b5yj2
zceUY8W0o%#oExiuLMJDpAL9y&YnEH4fwTK{-(Y_wZMtau&LpltzKfJy8H!Zm1UDiG
zLQGX<?VRi8L<ZUPyvhr8{q&sS=l?R_Uor$XfVZh?34k5nQuNL2K-PQJM$M?Ytbgla
zcYqIyeTp&eSr7o$g~x&Kp-UsC&<8SM9z{93RqYagw*!KqW>j!%DeUn(gtG*D_FJ{l
zh8fm|^E3Yek{P=K01dxcZ+{*27BJCi!FiFCxYqHAV|A+o*f1nUGo(8h54%hGqABwn
zu;l4kS>5JrF=oPqT@PtImezk$lZVHQQwsk9ZmMov;}j+;HP#HL)UBFUat&SaRY7zP
zx+o*FChu#}zuhGZIWDdrgq|(ToOyOu;#$wDenn?7AizRjh%})SV;01`gMm2TW>K9C
zl$f)ki^FuD&bhi_wKia$ZOxF|_Fbd%x_MvV@mss4<jQ~b<r}4pvydsaR_>8?EadN`
z$4JNERtU&09g$`rRm8Iw*d+Y{sM)t-0AkRMGJox$X(ass{PMxkDNoI;#`EurZ5=53
z<YQv<;f06S`|yKHt<+2xA8Il(G9cFWAR!wDTpAQ%u}jNwDxmDX_&ItGNlmWAu{1R`
z1EIjBiN-C5(m#h6D{MoYcN-+12~*FW3l;~uSW=)3F7ePW`RBZVQ%Ro3XiO1@0tLFq
zRPIi+Y#PB1(Zd$NIELHbuh1JG<1RI;a&y|rARv%N+b+)SFYu%vW6hRx24;efuE$Wt
z7+CnP2xlKt&GybK$?|pK(@!(XEz1Dw&4h^cg_}|se8JhA6WJPkIS?k+&_ex`IDuLm
z&!Pc*R+2vuLjuBXpj|)#So)Gfw#1`9GKGec2098<N0Q!&XoCyu)ztHknl+Rc4|X8O
zCm^VsZ1X*4VDHHs%5gc|992?YO_0+}`%m!A0;;Hpt$_;XaBZkw^fKFh!cZ^Gih<Rz
z#+Dc`T%5A*KFR?rcZm4xh!QJz<4V<YlG4h_a~TV?A0w=bi4#238m5W)r?X(Bz#-V^
z#<`_f$^me-gRQ9YXqQ8DQbo`^DlG#PM=80>)$<8^U1RF`5h3AOmBHV$i&c>#Dsm@6
zrXe#CJ&#~6w-zv1h`mxHmZN$)84%8msgI8{RfdFWf&Q*bgI$22z1+1e;P8WBEl35f
zzql3XzppUWi+>X+MnDdgc5q9vz>8yG&ivpUHLw0i4^GlpoqEI|G#b!Jg5poUoiXh-
z^8ep2pw9*FimaDJfokAa`}%VBs{DA&n60}4q~+S4=e}H5?n+W-OIf2>Y!Zmxzowd#
znx_54Q$x=`?E##ujo+b*wF`j+q~`$GV_5*~yX$W?2<b|^?|vCz|MKmr^Ql{FFe7WP
zArotFP#{iF1OrY`Yt#)v5l<^XJTF@BX#R*2$qS5%f+OU5)5W;MTUuHMybv1Kl5LwM
z+1b<S15KnS=d=TF^!4>8Lv}XO&rh_$4qK|Y2_g7T12<l3CQ#sHzB&Ca){1VEVdOV^
zD?O#4CC}Oettc0zeO<=;zTwFp`0&OqEc8dS7|s2z2rTsH#9;hmnSK)^d0m^TVt1d(
zHnVlgr`~~av>ngVy-VmvCjW>d$Epw>x!;)3ml%&Z+|6<5tr8Q=r^~SRAE8+bXi>Fy
zV~-d(;u;Yif$uZQ^9~Y*wA!@bG}@c7I=B1&y6gSSE|tzs9hx2|A*eUxCReaN_zeE(
zJ(&Xf`r8jK<4QtQiNQzpoSB&qi;ay%c!b{LBOH8osG7{!cD2Hz6|+56d0Ru`oT)Db
z+TM-W){M=uDK5`$mm+sD=|>}_57Y{f4&f!G8=Pt$9KpW0j*rJma4vr0)-YXmlM!pM
z5E65MJJgr)`X%4NB_lm@5iDXm9n3-W#AV`RAfFlgQ2x<3$2zz{hVimM-{if!To|b@
zZ%2^JFm0HwFaAkM`!yTOS;=!u^Y@7Tf!+S6sd7($(V(ZsHja;cum%N8N?iOKd6>~8
zd9}M=5|8+wiv*f0xmgyAA?Op!-$%;zrrkq~1J^a!S%m4c`|Xk`h|83(#-qd{=%S||
z-kN*OYJJ`q2{rR+(}&0GMCp^Ps=(&oZ!Qp`YI|DA;wFq*`KJQ=mVJ?93D5@)QDv=}
zI0)XfQJEu>31i4oUxG-Az`R3ev+SNsH~Tr**JZOQQl0><oQ!}N$x2@k7QX9FZ6clY
z!P!IDMQBK}&e6{9mf~X$bMih4Co3F2miI3m-{k+EaydbKtg=1jt=EcWrASc^2#|z0
zlOMe|i#@k1?ePSeQoaebovStPx|er?kyvFer3!Kv7GC{OiA{;xaZq%snHb6>NJq)n
zK2Q23@qWR98QuI_`aX$ObUj6x_OmXJ_yw_``)HZo4mO4}!Z|&9F5yDqvS;P}Te}*|
zPDV<rG&%*fR;x)Q2sN}+c2B8NxB|O-;`<{^3;E(p860};_G6e{Sn5jWnI*g!UuCLr
zLy3Zeh)`Vrxh<**!<TNnLk1S01>b>=9tWJAIA}h2Q%Zl)=Hr@HgDZ}g3{JXtD{fBZ
zSk|Nbdk3T40oRiOi(=?=_2MQO97<MT^vv9sVBZ$8(1gC`Mid{=9IbrAKNoPLbyGiy
zL3jkkZz#N5tkESgs)xn%bSV=jU%%AVL02xT3`YDnoVYG;>AOwz+LT5y?&FUnCa=6y
zA}cjDHRLW~`kfx-Eg$rynF(0hDzPLXch#psJq}NoLsu@?#iUwEQsvgllQ>rQm#M}n
z?bC>MdF)Hz^)!JOw)&d7bb*?K=IFtDoqcyC)<pQsSR-qas=H^TT*<yycg|CDgxPG|
z=5JBtSh&huq?;VOd!)%D0t@wj4=gDLVXv1!UuSdsEC#VT3TKaP5fsq*O+o2~L0$aJ
z8I^nQHs#MPVlk52l=OXTTk^4gQ~Jb>wgif}fuodL6=L%Md~EQhk{UTQ;Ovbecd_Y1
znZS*+_qV%Hav+VBe!Lx87rQNk{k*h5wfmOdGobfj(~p(B@g!AOxFyjUh$&N$-Em9(
zSG24pbxx+r@TaJ6^~v3<>;tjmpHbxm9s($;!yB7{3~LrX5yzQ2f=_4vvAP1;z|v5_
z&FkVO*^&={Ax{NO{LU@9{^3la5Qcm&CU&6ZOcAib$L4|>YWbt^8Wx#7ba>d#W2N@F
z##SG>*XQJm=>+j8aGep_SLQ5#)b)!#mGmHz$1>MX9YLkPvuc_g{q*B?VP`!dylq9X
zHnR2GUpC~t;ih^rK0_5d@&>z|2V*R?Q8MHnpL{cZdlmgefZY7;6hxES^G^kwDQ8Z?
zJR#*$H4_dPS5)Jj3_c%48^>shF+p=IIBxiJwlvcIVOevcy!oOYa_r~-o+2gs&c0i*
zn6#2`&srYcHJfw8O6tFIhta&k<<jvJx#AIpo|p*xbaPZ#^cMV({A^NOUM}?6>Ay#7
zFVGzBtT^ehcS-!*MY+2D82L!5oyy)b$=~x~&--3-abfI4_29B71tM7jgphA_r?&_A
zfwR81@|?C#gJyJgcMqi$de&vg<cAiBrV$=U$8<ny{#y3Yx5kq5_Xft8oE!S!PJboG
z*8^|YP3N+wV#y0)CZgq}mf_mI_$9T$bzIzMtLX6AMn<(iT=`f;n+S4LlRL}k$BpQE
zA5*{L=JDs#X>Km*xz}1WW>Uw01o~Q64OoIAd{w?BTgvp*NV5wNwPP!29?QXo*pX!!
zLC$J6Cvqzt^!glW_DKvYCcDzeF=~~$1UCit+TmD}Kjb<TME0n^p_@x~IydN*Q)UV9
zl<7m_H-aWa2^LzQ11Y1SK?G`@=+)3)rLtS}`Ba`u)cuEy^<Od5-@g|!oAW(wLJLr6
zCUXM2RkPwIP3~DH-rRpzM?y#Lo)iP2g;-AJ@>R{#xsr$fA^IL^;88EWjhL|Lms<O9
z%2m1F%M{>}fw!E$LD6=GHGfnfv<O~h+KbCel{sd_;?eQ(;%&M!7{qsd<H7&Po=si+
zSmz<g=Cqt|<sBiAd!ex4em~7nfG_x`J^n^}1E_}hc&+xB46eK#9+>Q{Sl}}LZ!Q6^
zwc+%QW|Yfkr<$Ojwu++F;&XG837hNc6t;qoPs!n@%qtx&Me1@V*AEKV;l-~Vjr6=?
zMdWellftiH%@!f<HUNinMvnC<?@rdpve!OnS1)?De94P=XNy{~-)Yh^{eiiUj~r8E
zCnp49j~uW#3}M{y@2r}kwy63Z98{6x;H`EXNl+qXY9`#YhUnmPM0_nE29{<}{Pdq&
z1#%aU-YUq}*0vp&^<&__SH9zo1(@bMfxuO(3~HUHUJt>4@P5!-e)wKCmlVOBTV7wr
z{`uvEfb`bR@H&@TJqPx=g6VQ)<&-jqQLsSE4hxy*W6$hxEPRCvALeu%J(N*secSbp
zFxd6&gEypTY>u9+5AbU&mdF2Iw*BRF*YC?#VhREb9O_B~N7Wf0(O_1w%qN9!6%u@Z
zeYxRqiP*K>R;gvw_(Q-$mguC&c=pogsZobM0sYyF<e<Gr<6o6BMSLcQ8yS<}E^#i?
z^MVzvykw$7IIL&nnXD_awY~SAO8^74I+{KYBM`*Jd^+vFZ>F_*b5BMIH}MBrGE8kK
zZSgx$x@dyCS}yti@25JnuRxXtc&oeRb^8)vH`MK>aCU!QTP?V9(B2a41iS_D6Bp7t
zJmXJ(j0dnI=XEyK@5%H$UGjtV2_3}g2tam`v7(!7ka-J_yFLy*X^+_aA3|t6UI_~c
z^ZR>={{CiBBOvh|dq3$b^0Jl_BUMX;<yC9eJwC%+ms1M=&n-?UHG73Drqm~OZ!ncE
z@@Q2`Ttwe7uA-!@-eWvWT!-T4;_Qt1B!G`ECF51=B7sZ0`ON_ZqbVt0V$yz`-?pXj
z(tG~>DNB+kvSF6Mv4V0kB}wD0z+!sYj)4JxwFS3JYYj(taexwS`hwWU=1iUqa@7m2
z{E1l5MdDzgWiGR%Iws3BX>%I+dY4b;3-#-L#IM~f9-rJv<2lRlqKP$SLCS>%&%N7R
zp@Icy!*yGZN|GBf+f_~^NM(B7EZ>H`jCnP42lgqAzAFWh!d8^Zc+je2=g94j-E%f<
zB*}&QwMczAH^e3x@!gUtFF6C1e1RKCH)ZJ|aPVA5=&zLf{%vIPl_?20z_)59%3)`2
z#h20r6etwRbFyhZbP~9fudlDOxqE=;`vZvQBixoh{k>YNInZjooZDka;XpfhG37f_
zF|L`Fs?6ca6!44WmQx>a-K4Sp8oM%N@H|TSVnf#b;YYse9el^WWrL#PV%{nqb^<py
zC{Ufs?hrKb-Ai#Qwm9^mP*#8|pn5Nke!6-3bK^~MlQwVj#dvu)*u_h(?o1xnT0Y8>
zQk9RgUR^b*Sxloy4WtRyeudxDq&GHDuB%i>Q&TVP=K88bNK)j1JSO^c=wgSF>({TI
zL9RD63RHMBW{NHjd({pXQJ!O=B);sYsgy<z=daIbN~ja)v$f$_;T%97(1zG_f%)SG
z$_|mr6o53K71)jCtqjq7?SQfvfPmBRVeF`>BqN$&o-`bBa60wuv)NlYMv@Wz4fDZ3
z{mOytzG1cu%}o~79^Aq(iQQZQt-i^7!+~cZn>lyd%h@kUWO|b0<||r3KQ$hQ^Y$Qf
zUmBYx1guC~z<V2_dJB-ztItgvA2tBEk2h<o(Va8f->7VcTVuIp0Yox)Keh5Bi~A_A
zw8C<M*HVv8QBbItE_TKf?qnoS3b_}1@(v;3erDsd^z}sCVZxv@vPbTqlkVo8O_n=(
zS>q<h)s9O*UXSvviE{I}`j#E1RlIh<M9v+XD;8D<z99UV-3DHbD&IE3l}g<moIZ=i
z#y#s0DqqcEn%BZ>WmNjZ#dfxe(_l&A`<e-zd&!+4!-L&AsH0+|UxFOP&$0|zF%y4(
zq|%2?ELWNp%J7sE$>grQ%1MhH+l3eJ8BZ^cAx^{Z$-sXMz*rF8Vxt9@H=69VlbCg4
zre{TDIUF%*Dsvco&+S~$QpY4%`}D9OPNaJGQz`aZ1NLJ}$`R6b=OVM^>SJuJ6<gFN
z>?f@}z5Uo;<(DytsJs$TA@1kYrnX0Jos09QtJ3-%X+g53m&ucWHJ#Qd780-g8)V|h
zIyrIm#k96c`0n|jJKxAEjC1QVNorrk*@e}uncT%^&VMLx9-OujCKHW$rEV9;aX+$^
zbiBUkMkmnAoxrokk2HIS>M|S>Axt$V#Qz+fpLY&HLrQLl=<>fNAu{<a|FYlwh7U4s
z^8I`H+azJ63K;1n1wEssb)I|QAaLA8>K$WJs*nuQ<v3D7eOSnxnJ~aaYua`EH`fug
zw{SzW+qKW$DvQ_ZJc^Ou68px(C4gn8p>L7L;LzF}F|R}ga!yLK>S3Pp?8oZJv8nRM
zY%L3y1Y#zGx_T#%z;&NZ>vL`OlwbiaG7spZWyzuzBqv(#Y{*y)&Ejv`=iTA)BMgy@
zrbW)n3?bNZfx7UWDDVnE>EoZo_gZd^4++&wj9M$HCe)Q;oj*HqLXJ^z9P_i#hv<+P
zIq<&oH`w%#-;}x43N0#|f&JHimEUO{WU_JS#|OODQr2MckVrAc<H;OQX^QoUnrJ9o
zXTR)e&^!M%8KSB`qTg>=nyFFP^tqD2$Rj!5hD?d&az7>_J>dagq{QrGmGh5FlLj~B
zd4+UexE#Yfx~m0jmdK;y@&IxF<5Kg_)z#Hwa}3(TzC%O)C;W^>^KVAgnj1YH_N1})
zr}anlTP}s}(in`-L}3Z=hu?dBEx=-=N7QPN=Njc=Mk8>G*t6(9IKlmME}!A&jl09w
zJinspG0}@;Xp2nM0`Vl(dSuhZogDYC#aD9ZtKQ`BA!((J5hE41F%4cn1K5H?Vn}In
zBM581dhwM{hTS~~u_*+b=)+ng%ScuN9Y)7@*7FZ+d!%VVD2CYN7NOPn-O53Q9CL)Z
ztys!8*OLe4x}@aPI%B!w7hp5Amm%POdZ5wM4gC0t6S?dkd31pVJJh<aQOlcj1dTRf
z!cRIKxbExCr}QDoqr4u9;PT}7<oPTX3>K&V7^0i{ZNIL%Rd9;^p}@e0ZawsDiZ>vf
z{)K22K%}1?{@`cR2Mz;7{J&41N)J}fn?qNsLgIR<`_fbRYif!Y{C1h!r3#F96mvZ*
zjf0ZDD?Fv@+MYgWqGbP06bxOskjT&ka1$VUPz<JONg81OPu@kp21<p@(yg8|;8FJ{
zM)+bf^>`hA@|in>!jTfEwJUDo+yTF@Opko99I3!@Mjr2ry~j{4uKRVM<C;dIt6np+
z6P?MxWuG-+BxIe~JE5cKQKfYJf`JDMeIOt9Ez7wXuYJhz+g=muqmCXM5>P3e5pZ$N
zjuja97YnbbdonB!kz;0pbG{{CrDDIkndre=OxXoq<DHsLup#B@ZUh4Tw9@lo%?6w5
z@wi<kQ=uCjcVTq5Dzw;-CEW-k`_%URny!tLhge6IKQtRU%*AdnU+69M1bv^zhg)M(
z`a3g%y78^mrcZu9<ULjIBgmnCzhTIa5PW(ev(v^&&}Ni$JH+GAcg2rn^uN0A^hXW4
z9$z#b0GoQsl^>bttCy;zW%U${q%wVhx#rv(Wv$$DjK;px`WkJI&v+LUNn?NNBZ!&Q
zZ0DyyHz@vUSQ<OV%ZE>Z?jDp4a;LM&_hR%@Hvm)3H=Pr3mx1S*;Kl%OJLL}FOMe7N
z!LDFri7kx*%q028Mx^;D>Po$pzS6!{;=W2_3gvbhrY6F~H_@DxjAXvyDUR68yRmxv
zG-^?MJ(y@mv6scNHb;8pO#zC&hi}iHml@nXS8mOJKVsC6tkSUXu3`<GfjR2KuQl1-
z!%|Y<K?r+R<XA1d*ayQAYhNCQomK+IJ;bW-Pr`KZ@t#-bVs0Px{xQ{Is+MwtxAz$-
zdkIte+iE77jq0O@jru>t3BB;<rDzYwa0J%SD(vpU9Kd824|LgbI(6SCuAj7c-fgeB
z5YTs+!(TWpQhzyvPjG`urb4HghTM~h%B^k1Z?6Z9HiELpb)Tns|5ht#5`IJA4hHta
zK%#)N&Z$P-R&VAIRUxLH(uZE)*$j{H`e9qNfhzRlPoqf!acd!loAdDuG+^@G8WHkG
zFPSDxA5(v>#G(@K+Dv%-LLgF8ki&<4>3LfELDuCINM7Y#9Vm^`IxERmWeG0K_Qs>T
zocjJFo(J;T0=8@t+UgucUh~TwRglj;(-7R7KV2A_=O215yj!xj?3WUH`Q2W&x9MEe
zfjiV{@@jd;S9!9xPPVj`k-`^Sxem*E_M6v#q|9f8$TPxU-SvgmWi%T{ssE*}?)dS1
z!H;fzUMXaLfZR8xoQ>U36A$EE`SzSL1I#4a$gy#F@f_>xLU?f~2zQx2sxAC7%Sl%-
z#GTNrfJ09Pl8OUhX(?Z+@wYW!@&K<UZMbQ9a)ut=ubu0ZXl-tJ>(}PRVp1%$sAEZ(
zFX<?e@>vUcdlDcy?yQrkXeuZ%G$J-hkYjYqjN*YBw+&)9!mpydVkW#r{hyz`))7*?
z1>*jW0}-3gSgt^ClfN$kgL*+kcH0KZylozn-)}#d?msYV+6UFlU+;P>ReoQl;sot$
zsxPxi`9t<oxtZRVfV1qmO4-ZAp^?gMYP3__T0ZrfQ$~rBn6oZ-jk9eWKj6=w?)Bdd
zF`@}1BlQg7o|*O^YnNH&VgqgOQ(t`dARDZ*_O<6AXka)3JIag(ybaJQ2`X2Bt|ztO
z#Z!_+kLDFseSHdl0EsN6wOug-G3E`z10+8JUs2+i*g0#T&pz0Cr?oGf2i{9>+PC&>
zCi>3(3d-GL!$eE_P}Vr{cGSCuUI{9)oRJ&!iGJ_gX?YZkiQbO6GMX<V-58kDD0f!#
zI8|2l+j-&=yy%|n*Z((s;_vN>iZL4Y+j2(UNm+7ow&Ys3+N)2nLj?4crpjtlKes9-
zzF~9g!h`utC-W%>B$qxsNl)~V0~hFbo3IClIrkr07xEat+R;zcv(?5g1>Zs?%>R~!
zkdDOY<z({aYB0>qW~j4(G5psE=knAM56ol6a#Tfoju@F7(Y*F+ZRsv2lZJ;xX>E3y
z?7Loh22;Vv#te~Z_0kjsI`!~2>UgzxzL>^o;i+=(gwd(S$qf#D{(W22C6HFb&I+$%
zGgM`*iY~c~>ns3pXaFL8vwaVg8T$dp2Q_BO7?Mn{Z^mxOv!ck6@(p;2_^Em<pZP5(
zxD8l&G30sw1$U1ESt6#ZNMI|%18jtDjeuCbn)f!R+l^&Y=idc4<VvbF#Sssbk7F=T
zMu*2H#xfUInnrPdM2^SU1srsrwCwr0sLmw*<a_cY5a8QAs-Mn(*gzvfdTBQk{ae83
z<90t)iU6eG>7EwJ%juyX6t}Go3;^g7E3S7HSb6Ar$x_4OT~Xwm?%QC~Hx_TV9bEN>
z)np%NA--Kc_URC6){87Iok7sd&0Lrn@vvSlmXdu~H(s<wErD*;h%#uo>VCV(+dgS_
zW=Zw_6caaU1@)7cw`}7(=f>Pq?#YzUi9Ni=V$)bY3rNUXZaFsNyuML(FeO9RvmsYu
zXCKtOio%Z01%0H6iot?l4qm(kyCNK=KHq-(B}<|(*&gFj<o7t3r%Bo8aie;Tu|W{~
zzTM@d2ooMu_fgfk)D7yH_#0yvHQ_C#poOuou1a)LOEGctiKcgjmb2hm#X-D`m^=k>
zkbZa%M3)IGD5MZzHT3uks{b$r?caf{0~h0i9-E}0v_>AQE2bo6e#fRYFC``t1>ddR
zT?HE(n<uq(b-73+7DS%{_fGiQr%z>*s>Cwg-!=;Si`~xY8Rb<xXWu9W9ZDR1CZZ?j
z%y@2yP#4``0+{iK(?X6V$@4ij`jy^z#ZFaolURiVW|yro_u?wvZx5dGX)?l(Q89AE
zTgc&*s5rwT)2jS03cnaF<s{xi0S8c?>kn66FnlmTW`B&&$_#i-WnaRheGQ?^r{8e_
zS9z3!g@raI=*lEwH^3rV8kS|Yu-*Q$<AAL*rqn8n+bW-?$aeO!p4utM<sJF=$7kff
zN(?gPPbK7{%W@0?k_-78xoY!T#B1|p!ya(mS20sPIo21Y&P?Vszq3!LI!Q(Lx=<RA
z3^Y`h87|CE7@4j?ygS4AwONQPz-Pksw5CEX>M_njT5t-C;@@<sf|iSg2z9)-gXQGH
z<8Iy;(FNnKNdkuhjoXG}K^n*c8#obk=CMy2a+`y_oQaEOxSkAED2tylMJ+T{)p(R|
zCk&0cOc=2~X*r*GO5)e~(3zDUP;jORlQ?8{y)Wxo13V5^@=dHvUH(pN?x5$?Ph6ak
zG`tEVOk?-*#=$xT+!lyts`joT1}ds$Q9>EF#jPM!rqMsg|BY1lX98D5%7j!^_BM?D
zHfUF8bvL)1R_TEkcDvvubE2=EibCKZ_1)G7{EWOc|M^su)Dv*p4ALC-e*YGVKQh6h
zL00@`xnvAblDkzwd>=LB*Yuh>KL45)<Vmq?q_OfRDVm#qYLM9liva;BhT-)wPr4Db
zx&z4R3Ho97?yuNA7ccUri6LqhKcVLQXn13u@}B{s*T6!+MViv%h`m*t>+Fv0+?}DB
z1pBRBmD4+nl7wjs{ZG+&G7fY3$rl(fWpPGu&G!)^$Ns<u<*fZ#m%{c^pGp#M&|>bx
z0^96R{epF}?96RJ&1CI9GPv)?|BHizk&$`UD!dYn0@c>i>V8xb93<C#1DKD7Wkz1D
zKe#6!1m^MjCEynkRg79mjR0`zSg9fev1#w~I0ZizhaUXNp8YRXp9!yMKE=cD`}wpT
zi?uB8)z`|Z*ay$wL0Sp3^W9~TRTlm{DAbB~PnY;nN;_wZljqLw${0NlD0&KTkaO0J
zkxvB$e-}z#7pKcz_<<=U1{k^0LQ;LmA^fq?e+g~hqpw;dKpzen@r6nLJs<J}M_XkM
zY1a=V7cuK%dPYgto`*>uxsqA!gYMPwc%iPv9}l(6Jq6geSyU&?3xhbGH|P>2El|S=
z1~0y0xh6V&O-yb$(bb~#022)Yw{`W;=)t}<R9@M~Cn$%Q5M6?}k`Y%Hn(4me?E<=d
z_l3YNcr(u^OeBq|wRD?_55;tLP2FyJazaTt(X5lD4Oiw{n}%^?xhjx-qUc$=v4svk
z9m+ipvFQd2DE~dlUwn62;7k&^D@XZ<879f;&RX&ysx>6u?%hPAF}r)h_2aHboB(<X
ziT|9iKNP*)qZSem2j^YZcB`yaaO^+9j>9^s;yH|-iuYCz<hV$^I5P@U$(TsG7<V#V
zmv$9n?94$F>Jq$-DTVp^W>u)I1M;W@dUY%)C4x}$%JsDm=#ig;Q<_9)t}-_em6Mlu
z8`wd002F|AHsJ5ICzwe<_!$639e%xDc!!-ad88RKpX(Msd81J;5uWx-a4{#ei6P}r
z)s~w>SSoYC9K2OT0I$Y8Uy?Dw4co5nU3}^5nn~YEPqq&I&El&Q)ep@Lv-!3h`xU3p
zF@wr86X?3#f6aW@rM7)IRnrpxCh^lYwfYeOI+0J&p#_W<vOZ+2;8Y<5*^p+&U3N)6
zQyXe;MNr9!&zsF}?K1?fVSKq$p*+~n;`SZhtJ^z;-A?T0SLyejsPkyya&fOq$sL?q
z9rj~hcxrx#l3VAB(bjg_gXPH+mqfdnq|T`@GNzDtJUNH!Dz%gA6`VE>R&ay?($VAb
z*=_I$044yOv&=0=z<Y_PZ+l40Y6TomJGc)8-ld;b9PwEGfDTo7WI8c6zLapnwSI`O
zjhJ{ZGJeVvaNhr+jyWZK&1<USd2sq};}-YRr%x>ajY=peAEyn<UxA1tjJJnAih%K(
z3tv=BY~W}inCLg4F8`z8@>}yw|EJF<2*P~_94gAoiF&QvOq?M0WQ4-?Bjm_c-7e;7
zK%JBH-1}+iM3+|`ldM82^`4t3j(h215*;4v{$$QBnml;*l+T<QZrLCg9IqL6LMCEV
z?FD6CrK#Hp-gUJ7_49hH#9ypApFgIP&qb>@jYMAlRWw=5%j8C=cqm0^JooksuWrkl
zzi0M2dGp>PMLRi6?2GV(vkR$U0>+6?mhJkri4|Rgn<sBzDC{jp1#94j_h*lX?USH&
z<y-NkZD$uO&WJBw<v<-z02u!xm|H@*%rRY!i!Tw#C%mAWK@A)IHDW&|Hk$%?0{Ehn
z<EEmWap_YyE=0Ze&MaS!n0}fsq1vQcw3hpw4$Q7f9S^t)Kifu(<%o=gqi)Yld8A&I
z33>yoM~mbr!|IpM0Vu45M7*?jNJ10$>M#VY0CBAa3q47s1+e{BI1>k<-#?fRfh7oM
zDvx=rf#a0qbrcJ29Akpg0+e<O)H{`4hc6j{JFDC4*8l|6cb^3|4F<p$7)oKTsi}Ei
ziUJb9YKwDM?xKj+tX_;~Gc7du4qNy*#i99n#xFBzjp_&n$clu&EE~JlxweIo_~K8{
z9Ss*<X`)TGEe)S^)%Rd=V0D>mYaDM_*}onl<#E!eb@!sD%-4M3^4ZE9A>9)<1TQwm
z@GTnigGwqUatsS*X1eL0jC;8A29y!R;_T;IXD)Zvp0V88K{n)aP+cKx+i3G9-ma^#
zW+H&iowYA{4|et>kPNx2L?#0YHQ&y#QGgyha~{&fc(WCpad2HPuI&w_Hj;+KC-5@N
z@QnT%C(Ug6qU<vdQxv2WdBN*79C038O@ZX5l3kNx&E9aPs$Opbwu``3zY`}`y9=zT
zq5&zNEneBd-hOy~J{AH?CC_^`K0GmzVM4wI&F;%apn_0^RB$m@O(0To5m@y7S88_5
zu?>=f4Js_&5!bHhAq+Vq--tSe%+Jr8Jp&~jENe@ZMQ*_DXdUv}aCdi?+g+%XN;&3c
zT~X4WY15-+%XG<JC9xNB72=o7$hY?Jotr{OrpnQ+PX6dID*KIErF$n!UoVK5AYYm2
zA9p=|#QRl~?wfac-AssJeRY|&=_6g?dXz;pgK|@Fn(nKCIwg_{oF<VK`eLJ%ADGmb
zmk8n`mPVgUA9EwL&8Y9;soCR!G#&0wwdN5#Z0?HS*{Ql%-DiOP)dj1&5}85WHUO-|
z{{(9&eYh#x?JieLd_MWv{?p3eJyKwut#Rv20UY>Hr3^{%n~L++iaa8<5R?eM26ob2
zY*)b7kj6=WG@|b>oJ3#iZlm!qfsAXogn)QkUd_hpgOoh#Gk%w3*prvZ6JP7SxVTMU
zEquUg02z}+kmEo>kcRJXQqLiu`QV1vuiGf@ztZ^`S`L=fzPqnKb%skm?!eD~xW2K`
z0ZdY@3V^2Sd-4DPc%b~?C^Akc;x_#p#F#z*^Ses{FBNkEsgg!TEKHR$VyY6%bU=9b
z?uT<=3u^(SaO-<}LWc~g{h|rp$oHyTF1}0D13RP|&v<w72_UF$Y;uywu+XBa{Jwl3
zs7(q!Rdg!$o9|Th-V@Zh=9bNL7uAVw*RJO85dIpB?C2tKe$DS2oJEwjVEk3p`qb&E
zVH^F-jx?fTraB?|b?V{qY{zAfGA8#!nU8sE8Hx(a7G43Pw(7O+9xQeBDj6<kj}my!
z5f><u<u97-%r3hCr3M6{5LF*;Yj3V}3;*0&Gi~U5`CRL7#O^K-_$jrKCx6qBU;$4#
ziQ8SV@8jOBv%r+9p8q=J%mF|w=a8iXH^p4ogAsDL2SC*bEZ7fW%d>2To3UU&CUr#C
zHp#NwJ~NU~5~*O6b8KYGOy?gaSg90fph{^g7p`<Hup;2@iS*dclZ}dK0DEdxtiQkn
z&WYq%_>wH$#XX424f()m|K#EGX4=tXA+|Xz2DxZo46EI&EQlMJ#f>JqUhpQt9wREg
zq%|gf<G^jIQr}dLs#x;^2$B>>3ZGMC&-CPEIG7#N<;<qJ(^WH8QprVr*fPE^GjF2>
zkW+!0ax_`7{wlHsIDaNSGh?Fg@<;E734J*KV`f$1d+r8h8On#w2L+Wg7x=avecCa;
z$6sbAI6e_s*`cPyUz<^G2O{xYQ5czw-^hty-bTIMU;Qi^yeI2yd>+~0Rp9*9ak%yM
z$mgskmXEi#|LV($4z}P=w}jEj>I)><joG>QRn=AxZBa*rkoe)(h$`*ZKJSx#R+&E=
z;7Eq?`vWw?%Jdw`fp*{8w|VUPk#49-?O7#QA|->}$xt%l@`+uFDua_7s@g;jJoscW
zqpcdmW-x5`vAcQ6-GBX2^h6_Ac##K&!?txHd=x5XIK&}?cclJ=#oNP=+~52kQ(qYs
z1@peWl!%Cwh=6p1fV6;s(hV*MNG+0*t8}B3gmjCbA|VPa-Me&3NIWdvA>FmqJL~WN
z;r)Q;hzC9H%-nO&bzc#9wYtVmIy-@1Wbn+AHEAMnzg}P3szS*sO>Y)6cDrzG<56qB
zK(WL9;Q)n%b1*6k>0SZ}r1#^OHfn(VeRRoLe5{`p^jwjHf!xCyT*BDwy09XFwSM49
zmdH3AUBhd!^N~jQ;}ox56JU@j?Yak*bTt6o;-e$0mt<>cdUpW4#e<!#3Jb6tdtrSt
zf~pi&7<#U6BZOTA*$wo}o#0X{oi?Xy)rz`)!3z(H#9+QJyglHhtX^W`)rO|%{Wn{o
zFQ!=JO8VgvoCq(Sk)5ypKblFE0@8NOooaN}5fS(LUk?~>9FY`iic}t3B^2z+f71R2
zSUJrq>q*dcABgq}y96bcy98qGHx0m1__?58bd-nvtVzD$GnMTu=WgFDkfea{j~F3%
z-IaL=ijLq((7+pSY(TwIT)Hy|YTfYdq9d`~ku}H_GhXsPQVRBO?A3?3@Q)b&I~dOn
zll5S#QwO8sTj84(uSKCVN?ZrlC*f@KYcaI1e(r1<@_q6(-g`Q!)bO?)$8rKquKeJA
z_evO@^Zg?+Y=k9)*BlO@+Xbd4$;9tJg?e;;s!(sP7LovSVX}x*=Hd*H!){=lYT^<>
z-m18eyktY(5!H>Fcd<>GDV-;JGG0pJJ75%NYQRJx{yE%%)mq-cUrJ0?v^owcsS@rU
zQJn{W8Cp{!CRJ1U%=L`BZJgnQQD2~+Nyyt8`dWE!qULrV^9xf0=6X$XU-5X=Tr_)C
zHDb%(P9FZlKENJs#__hMJSSn$NcL9}1_oi||6i36z;e-@x!q+|UJT;Sm4lZeH{&Lk
zEgY}pev^axG$T~h3S#oDGG)G{`>&VyJ8DRm1cXLb;}9UKzDq2!xg*?=C4X=GFm?8S
z_ki>ydoT$EkMw4NP9zDqE}1JSuf0&l|BV*n9o~Z!aIa0f`LYNxWUVDEu8$eky0BwS
zel_j^J6}2=z#9{E@&&F61O-JQaXrd0;+&t~1f{1|tc;G13Rb&=&V8Y<+^XVXT8Qm4
zE0{O&Ep2ED%uH7|_yv96X_tCnd-f-q@Tc-?x05acQsPw!6I~AWM>&EJQ#KbKe)H&*
z(7X%lMcj;E;&RqE(V|s_MdBL9`;=eO%dA%dx@D^$U{;0QB>kQViq{~v1-q@z?-s)`
zMM39tLy?3>&!*;IvlZ#i<ZW}BP5ilRQ^ynguQKlaCTv6QClyWqY3)CPQdh;IJ)}Oh
zjxy(G&2xuoz=ew2mvhL(e^W}|9uyZLIXVTtrN;vJVPbX}8n^HI6EU|12F5f08YrQ@
zBF9z$Jt@B9Ddqn60QTd;gsolnGrHL*uZp#0Q!Mg9NoklZWULjSbT%fg3JxWxM$6vw
zagd3|A^P7*lCD-WsK6Z+TfjK{13=&|DQ7t4RP3-!kT>9clzs*r4rs?cS%&;F{}`+6
z^XYNeRW{C6-I_d~D6Oetpkdv{m3;HEe8Fk-_fe?nBI<@<^|(i?kh4diR?rb;sOu71
z$9Mdgu$tQWx3k6axOsDDcbisxay>VFCA8#@cIff^r(2w7E8lK(Hx`oyc>jIXonBUk
zSXc2pc67e{-DoV_out0Smol`mt$J#PVQ!(A*5J7ZpcN@!c1S!E_^-#QDszicHU*~C
zF@ldz-Qn^V_K0&DZ7@J=C)u9Mb+q;wZNQW2D5?5axn{KQyMpO3Xg#O_WfJf4e|eMt
z`}}w(w<Kaep94J3^w6C}%jZ%|^(c97y!tjJkIH`6Ez|b&=f<&()lN#bJTk>rz1<6O
z@^@&^$3P7#4a{{{i8e@?bKFdTL1nX_vNB=zb0Rl-X>1<08ITYY+M&*b;DX)goU@-d
z=j@*R_g@HR!Y8Lvyi4>2fv%^2{HSW}dp){)j;=jPe4cc*Ri^$K@91()Z6<j}%$SDN
z_DeSCF78@1KBP?R2|)fq9q0zRCPW?_TP6mz#=d+-jlR`9KQgbg4nLL2sOy0D7e>Gx
z>|)?E{JfR~Eg^)eNcl`lozo(diqV9>=Y!CbU=7J@E_cG4%xy08*d`gjfOZ0>p&YOY
zmZL}O+ONo3Dkz2YufRt+4&E(8gGTvVp?NON+%$~X0~|*NIEg<t6d?QO00h7SO3bT|
zIrro|<nmr-Bj%i(R0|B<h=nkAR!y%+lvQJRU%ZnLbsrm@3=4EudA&dO7}&)2=^Q6E
zZ`a4d-9~_^z@jGHXE@~O0~wSSix(yCZh`I?2i!ZIBcoVPFYJ)Kt08NZg!`<=&nr^V
zVQ^~ry>0!clCOFEN$=<=%!!FgIgIstkY6A2)`qbg^R9<Y#Ao6MIWrxF;lHeIoHCp@
zc{`SOXcKB;kYX&(OSDRastR%uZ287zZaEotPF*)}<zHIR^v!BMJLF_7A<$Y%0NG9#
zE}J?vIP*<7Jxw~5i)}q@$4-#w@w+m(;sJqbE5Iwdas~<K_^=>Z2v@F^S9)ynza6OW
z(6_`d1RXb!!3RVTR9xhuu+fG#*nj^8jbf}JJACf4Ki~b3MzDmCSqvH>O3M&afII_^
zFtNjy#GS@r7OT2&)I=;qB*$zfY1GVyQYfAC+fj$=jw#iAtX*;Ro?fq2WEZVKMD(Sl
z=Z_QclHLS^h%``UT-zh}2suq%`$W@3qh?ttpxq`QvltAX-j+N(*gO53*p<p}YOer$
zRwcl85cYeYz+8c9^r-T?#7uBHPYUOzP3WZKPXzm;*YQ*~x&Ex>VTXFW-59&g(FD<>
zsX85csk7e!#wg*!;K?_>E3hA5v>HfuW-HT$LzhegC6+KZa9tn!k2MA+-22SRFzJua
z;>vU#HXL}Mv;QR=HG^4-iCMid>Xa<b8{xvB^~b#o;bg6YfqPIV`SViv7N6=vHq@r6
zg}5(di;lwnl|{%Vp}7Zm9sCRvuqz70cC$Ut9QX{D83zt2g>E^I8o3@O5-B-cl0ant
zeW<m$+YrzU(HU)MifzTn8)0vMz)GA^L%GDr-%NLoK(WlWU1M^zq>(nieinH%{LyUv
zDa}R*tEJ@_6*mfQtSV$YkXhf@cnd7&&9h$uYj0qCA-3(Ipn$cn)`@s%RB;n)8CeXh
z;tuYs%Ly_MlSte#0oJTGGhJ2hR1Zj^8d;ZRuNGx{XP--E^79<r<4g4CoqL}`@>>1&
zS0%sUlALvXqt$BKPSFe(-UP>v-%B^k=SQZ$@+E$97<O#sZtzyEN9k;OBpD09WEj|(
zR&yY$oH(mQ;C;5wh3Re`Kc}MzYK=g77qoD3U9XH1F?hOi_C`R+1oKIZeX{(mUrgd@
z-)6_GKSS49xi7srW8ym3=P=sfue}3#8VIP#97(=7e%PZPLiWtz693&Wq#_wS3>Lic
zn?9(4ky~juDjzp)k(^n6Pp0SyaNQ$#RwN+_+v*+t4gbUgJ>*61iO8A_6!at%+k9F*
z=&J1{0uO~SG9h|J80V~8!^tmIM6PS*r{`!9zi>vHwntfTQ&YUF859kknSF2lW8nKE
zNRth3qi*Qv15P1PYH|r58&BH>7J#o#3$USA_FI*jZ3Y5F<0;@AR0Cp0)TgE8@{+LT
zLLi;;bGmjeR0$s>jqqzps`9%rN&qKjv6-V{&-5+A8#iwCG=y@z{Br!Hw;(-20nQju
zzGeL|=0ADkS1(k5!sD`{o~VjBUfM*4S9@eeH@u)pFIz(foNHYi<!`zK_`s`TW@Q6=
zre!A4Q-@}O8RqqGE=!giQe1ZuYZ1?+GcX(5`dowP+?%$PQ3vdcn+MB;RrWu^*U@<0
zeMKjizQ%RG26*2v#0wmREhZ8_8^?~TWeMjry%9Lj+OA|<_?_78XE&L1!L}tm>-rtS
zE_7$<_m+C)sO9xgsD?w~rmxU4DHLC#%A~mR3JTDUzSl**IG>51UNLl!ef_I`lUebP
zktg0`(k4&!X=-fUh!fm3_2p+}2KXqCrNWBMU%~*^nVc-xY{B}57;c>{)YrheD>#*g
zW7t!_;zrRP!;{JBCo7lOmyj1wu7pM?eEwBc6+xs$p}PAYB0!V}C!U%y;PllL;KE>A
zEuh}998$NY_=eelT}xseXdi&wtK4lKX5ALLncW+#mz;^JtTR1D4;VpuMjClF*681F
ziIq_>P#ZiB2&`w`B*3pQX(AK_&3yt>A67x(&hf}gev_HAelb(nVBhwVchF|p2==)z
zviVzuN&RP%k^cxNgIlfVeIzzaKisG!FNQI-8(>mqQ|Fv(%L*TCVi;9go<U2K(_JyI
zC?i`F-p-wbwRi>xs0fwX{Sq9XDCI+W`$Ahj%WFGa8E5>K<*$$dA+<9+at{8a?@mCY
zU^S&8bP#H^0kB<ggY5&JLg>v-+*7RBfoK16>q(J2KqcgO=!)DK-=T3%pM&rB++rmF
zRY-q5MuV_oln&Cs@HBhGzd&atr4A23cl}+V?~OJbAbxJAH#z?;OWD&&S89n5uexxU
zU-Z0YHM%4fwjj$zlKhEV5>q{7{q9E=^-{BvCsIT5{p9GiZ|~MXmNZ=_qX6q~jR%v}
zaAnVGKbUlYMQK_*!S75oa@)NIur*4+*PWL;*scu>w-zJgQ;BGJfoahrE$J6Z+yp~w
z74aI<1H#o<bNe&n`$WKDyk0`7eR1z==C=gD){TabA0*t%_s|{GyZ;?scYxu5fA@bk
zi~Cba6vtL_=4=vu@dVM`WdtE6FMBm_Rt-61aXGdI%11PQ6LYL{cBUK01aqJVcEz@h
z=yaVa-yh_d%SMkKbKm`w;Uu0U@?b_xWv|j5PhuRoX#C8OfnH!ZQ#D7QebUFsO_H*5
zm2AGc{6gSKs~w8x@K~IsG_?Sg95ssJg*fxN%XZb&jB<!j%BhTVoY@r#;O*Xl>;a;m
z9{&FFal6q*zS=Z7lpb)Z7dn4I{0ad$Xx+?FdsES&FK3k-VA$Fa`1}tc%b*Apc1&C`
z`X!$H8ARpNlYcErEz2s;zJKlKdeGmc4Mw7FWUA$hmfXoYx&5v`Vn_eT!o7fb)Zxdc
zbgwdXrNjqP2Bz$dq8bmjwzdG<G`-{)8!G+Kq=sy{2P<E?z@Qt&BEGZEzr(xJgsoRt
zNoV4WxSU$M@hnCk3j`&&f=YzV+40*?{L%)3Tf!P{&V!$&o!L*RW=dB3hCm>`&hhTQ
zDP68u#fiVqU_TwA#}g5z(#tl6IV)w!wfS11QQO_{>KQr*?ZeV5Irv|9W%?5ubRG`j
zjndgF_1Qk{^FGxR_bo4e%`xq<zA8E`+&j4!H`b8es?BH1-a&7eqzIkHI#~q-AAg2y
zqAQ0=9OsmeSJWDz12Tb<iFR5ihmDU)-Wn&O89lS%GOY3b#Lo(<F>!jZJ8*+^KRt)b
z;eVH~66jhcAu;&7Q0yU~m5@TIIb7z0bE8Ksg!^d}GYpw|P5Hh1aETnE>`(-I=&qc{
zYP12v^#U;W^Y=Q~hLJqG@+$eZ{;j7<ssDrh#78Gxsm}Z~ZWg#GJ;d~gqn%N!QH^Tr
z+gFw0O0bYo=@D0@PMvJTqec&o(W9z`wYOa;U>?;Ql85R4EHTWsr*)U5LyhrwJf+rS
zoPSz1V>*W9051i%3&2x-=MTmULl`BMg1EMIfYEmKa#BQmt}rDB?h5C9g38d6fyaxt
z6}~ULx4~wZf4qI&?}-{7*;-#0T9#)&ThV4Y>yPSxTGmJ$TjAW+GI7JpOn?7K*WWe&
zX$y&hF)T=Q!HKCI6NGT}YPG8`E0nW-xhmdJlqn}vLLL;nWIFrpMQ}oxD<@3k4xE}I
zjxgf)*}3~ZLf65SVD6%At}yMRiT7Cd(0j<%gsxB}=O-hB>gnrK555*rCT*8*!jRM)
z%7Aj>->CV?N5MU5>&CTX2OS#_o%=-(QTB;yprx@Wu}JPh!4~6w=5*Dr8CQ&g>6x<1
zf4><K-!26CYU!H_j5g?Vx4YcHB_j_JPpV_2>X`a>dGioS?f*v(+2;QDiYJqkyXepQ
z{w7IFE|2KD*5M1abSI<!Go7eNMVQ>fiGC(jyye2}BB>W2Y2vvW6a=;>Q1-p&<lWnm
zyv0olf*?%u@bJ)_rS^0o>bdg+Ko&d&84EGq2Oan9OI()K%8^gZLi8w8O+9Elrriw^
zXLyO~&w`4(8q2mt5#A)I%PsgKr|cP#%xN>}0^s1wZZ{lgCC2%KL^Yb`>gsCoF&SB;
z=MOiSgjav~=-luxh8B4#F$+7*;AGwMvW+=>SjNzXf-!5OWH%^T;~0sut*_i{WUJk^
z0YBe(^iV~}pZ8_#>w238A;w3K&?+_aPA(_$U$%`^AOM1;{+ijHac~*}jU=dj{x0%I
zRhBJme5Bg``|V3|UZ$<E*$E{fM-FX<pAU*E@6QXhH6g%#E5s52nUe*55}1q;8*She
z9YsU5a7*$=E}5<6d(=>}-0sR<74fo5*}<|&fH9#Khre2FUo@Lrsg+VE+;%|{K;Lx4
zdLa|^2uvm9l%hwClxqB?-;s_F&|+~A=3Ex*0abno(<&Djgti#67*(=uszQx@4q}Q*
z{XNVO=Qr+r2|>LOwUCbxG*vn63pb(4cn`p4nrfBvHSIuJV29K072nnmh{+mkU<l7~
zkKZJ`s!A*gk45kb5NSIB=H2mM|Emx5Vc1wRxc`PRXCXW!DWN@zdJ#Z?H;4RMYf3z8
zlPwjVFo*}rr<(b2^Vcp>=)x$PmpwoVx5SbBvptFBGwbb|n&IAY!~e!ZCegBCxd|)3
zgvUeNJPAU3ru$*k`#Q<9DUF$Yq(3*9Z|lAAQ)L_5U9lOzHmKn<SV2EJ5UbJ0b!68>
z3(#3KI-8!NCbZI6?6m989er7NbDmW66noTHPpZ-dd{aH95~|}()NtCE{WD^j=tqnU
zm2mai#gjbUj<u#Qc)H~FoZFq|9URxox8Phpa-Rgk!o9l_O9mZ$U33lD%J-W%aT*4)
zs^iwWA@9WPivnTmF>!0bqjmD!N0Po5XN(Zh+-=JAyJm44-<-0aaI;p7&g!XkpWK^W
zot-WDSG@$y8>P-0u$2C#pca4i>3_GohLk>D+(;*<`ZT$mN;)TI<@)rC-Oe_Y{76K|
z2l#dtM(%0;_$D1`s7Wt&+bzbpKZ9#rkK(c7XoiZyjVHh!{rMeifL9~nc$NSrthS1;
zfvG#y41lUNRE$HCK2-x@+8A)=k1pK6pO#yz-_Cv)1I*Q{sy-dgSP4a>q}<che^^kh
z^AnLOypvurelz=YZ)NpQ;zvuZLJn~!bh%wsmio$nI$26>!K%}swUyKzC`5-_S@nz=
z2kvh$i=0_YY*d<7Ih|SWo^VU#JzLnB%b91AHng(ll>Pt>D_^TWQ*Gv-uGUFFWCnSZ
z5j8%;ASS<THIjR4a=CE5AA8x9S3LDsfg{~0({aNn!+t|;g7NqMu;28>YpdCdEn<Va
z=jnR#WjmD+lp%No2Tk#`PED*vkje2d6uFx?Tt0&+<D6pGOihmXoi5i|0+eYs)4r$?
zGWQsAfvUj75#Nu^TQWwyE()|CzGp4M0)22t833}2PaXIyqC?F6fBkPf*j2GGdHf(G
zL4bKUsgw<K!>X#<lbO~Z-)sfamR-NEb{21`@QZmW7ZA8hihmI`zV^dO-FBNxJ2_yy
z-eWukyZ}3e`aQC!_7baZ3JnpgwI<f@KPdqWf)=pV3E0DYC2~3efs?HOhj<;LEXR{<
zfLwo31EPO>1%EX*%98mUP<pizA-DB&X5-mTKH-mV%8~W+3a5Wm63T4ZpLlt?RcE_f
zUa39bHKCZ6SRk**tMCF}ROiQY@-{NTYSHkC$?%tq&!1|X9X$7(df`DwCAy(`Rs1qV
zl{uS*!_<KPLp;8DEgJOT{ISnf<a>kyI1oAb8KumZ=gg|DLSKexa^ju|EuI}r$WP4V
z2{9zt%<$R9b>#@gsj_CNrXO8pwQ!@Z5{M5-hCD8RT<T=M5o;dn{p4mAIGzj1C^Tc=
zur?K&9+Td_D&CaTuJ*;-TlM(@JFvK(;Ct)4Jy&;+fnxaRtIm)>`pQ^T`wSu}Vg45C
z(L_{VR}Rxaw{qoZf*4qs7avA3b`VaUTz-OG37WRt5q}ol7FS3=+6c}AisB)-i)x_j
zWRPpZxy;;cQX26|h{^ZLD{P+~rtIL4fUbK%8u3cXMlQK^)3EQHMFodx(m5cnW!!3i
zN_ox7gY6x{2kuDhK;9T&xd{b2g9a4G6{7`bFz*3fQNo?QLI#!3tpedn9V*#Gieka5
z^Zi@DI@mF4KX=Y&PGqKlH}ritX1gbu@*4RU=rSWfVi6_K(aCd%D}DcWcL66lZzwcK
z_G&M7ae61<B&1riOOLsG1nXyBS6>eXA4^g7?I}^J$dwL0HG6p0!v{Gx9HJ{$UvXp`
zq6dTEMQ7qE1?gkbl7!?WBR{#;_3l=NXY;dElaCbKw^6VNs7;Efp^KnhGjL|ju2E80
zs!S+J;)Z2KE8NYtkqA<0YZ#FDKH3=Xqo3_=f#@n3Vkn&3c$T4x4~utSl#H(7+;^S`
zZy?%ApY-aOo;Z?>4iJc#aic&QG4u0Ti3sWc5f!YZfD{(q-CBCh$z9)Rh@pHkb4xNI
zqugQ2w7e(PSM!DMbziVOuU&kwlPJk_mgAPmw9AgHc#bb-Nl}N+=T^D?URR=D>!jrG
z$0o9Y>3ZZw8v?w!qmWaYcjLL+10s3aVMl{Tv!>YDN0${?z)<Kq1LT^t=>tS0sZQH!
z==7UrLF$QwT#w>a%`;$)t?9c+K$zSei_-we*x>aP7Y|4&`rPg!?`ClygNu_}{#5D?
zm|Kz=ZPX`0Nhb$;XM|e5yeGo0wo(r%q=EFQ@4wZ43RZi3^U;J-+nU!&k+eEP-4PX4
zl-OEDCs^^WH#ZV*Iv?2bwe3{N#O9w!-ucjHDzoB04YCYR!Oa*6Y$gLt3bL@yc?#vP
z2U!|cL=oKSf6t>}7Zy-9SSMEb_fN|+tgW7QmpQrp;_M*$FK^r8-YZJKgH=2l(H(v0
zd*nw#0lsP)FAxbP?2(B~wi>_v2QSpOs;K2-P76vnm6?YKMK0@afImZPP3gC+pJNv9
z$6hK<AsT@!4Oc%tQZ!AUYT&eMNN{w(H)mp((xXwR;A-v~??CQ#`;Anrsh6R+9uhU9
zE_hvvRq7H(q~s}lEfZ!om@~WaaZA$OI+8fuB~=bJy->S!+W8+LpGy*Ra5;tTWG@YD
zHH1oXZsjL$4n#I-?N<tw38fUnRxnOkH&TfmZR8dz33>U{>7W!x26od=O1`~S-}z04
zadHVX{L_~ow`r{Lmwotd(Gg=36j5rBxeX?Z8uDV}Ir6Gc07nRLvV0&-`AJY%HPBFS
zUFcJ>9MAlyOc6j5aZ6Q<o^a<b@myWsU@{1Hb&KjTFV6wOs%t&4)}`MnXTU8?-gJ%Y
zrCo5YYtzez#hr6L{V|I2IsgJ~ZO6KuZEyo)SDPn`M6CTp08HhWb8Z2qSD98GvR+WA
zdY)bC?r34en$gksE%EdA!hFnY5=jc==|Y|I-uG4k+JgX_fvjk(K^V4rLkmXICLV8=
zv_5*-u%-!bi4A->e_!`EAXK=UfMCMJx>#T}=eHvKr5Uh|qzlfOt%Pb8B6+cv=V>A^
za`*@J{Yv2#If8UIrS<R%-ruM$UMO5>nK@mHIrI3UTb*^wTEho15-$|*PY~~2&FMp?
zY4rx$%CA&HMIdg!Y$Tg*z|69&Uze;f^CRcDh1yRuvS>;v=&`Y&(W^r<)osx=WPD-o
zM%zsIl$sM-NuMI630UBM@*e+l*uFU|Ksj&d93A4oW_y<JM>F=sOfutk;j&n8QaG~U
z?m{=)LN~q-CEj1HJ6{b}go#*jNNd>4y;QiO722o+{MXQVSD%d6a(2*awF$`j7{X7*
ziNQ_@2)C|j^y?cF^1slGv!ADCW}Nz9me(=z6C|9F*vJ2&V_2T5)&Q8Q@kf&66i0@s
zK3IEJw!!@^qYc1Y0t=(|Ui!-4klX#2=1H=Enr%`Y%*x~m?<$4S#tZCX$$zWdPS!Kb
zP4dI>kb<bvvM;OlaP0d>63r*aepGpVeSCFiuKunG-u*Gq;U>Nt#RweZBCc`Y3_lM1
zfMuM4dAQQprO>-ryf`L)xtm1y(S#qLHmLeeej90w@7@ekQOF^=qb`Kg^=P-;mahdt
z3u{Wf^Mp9cexq}06f21y@ZWaznmuP5^mhY9vpK-7X|-aGKw-x(Sczi60x|3qx47{N
z!#$3EX`A^kFcnBV81t+GLp1OKCxjCV=(7GeJ~^QqM0XM)od>A{6iqbqhgI^1RC0&)
z=Z;)fqT4nkYSG#?eH^FuQgDJ-iFq5GE2GR|{<i!=a)@}H<uA$3c-${?OboV%9HIw!
zt(l*rN4lL^hu-L~S8fDXUHC^dIVAWNojQ!bED5E~(2*_7s$)e5mY+-lof3+Ftq!C)
zwU-<*Hr}N$!zH3p#3jG|`YRgGXZR^x?Yls(;2ciG$BX^@ickeuvS16mM{6y-J~_l9
zT2GIa3a24l;DUB~>~WLWB$^H9!)%Qr(ed|X;qJo@EA>@gJmV_QN!KvBAOxI`8g@))
z{Yju(-eH;*ea@A8bz@EpBCI~0&!c29K8vWlGQt*H=H+*zb=_*RU&@IBSuUVn1rI48
zGThwJ`U@;MECA4arCjYL^=_cc+yI>2?q^V!t6*wUJw~GqDd4b*yXyqeu$uEG1y!Ba
z$bsBN=0_3+b=iMV0X+c?AkdSvUV4ad5lw)~5<%{-xJ}iK=Igw^@Y20}VS7bKONciX
zD{p=$7XFNi?(E6$!q<GCFp|5C8&?qn$o0dwTL)O*m5yk-x06o$wJuA5hcnysZ>Wy-
zM5zHl{cf&+Kap-TpwO-f5)CZQFEU;~R)j&HW7#KI@Of}*>b5}U@vxZ_4<hHlrZA2-
zp*ZJI?UoZux*W@V{-ByqjY?yL6>YbhBw45_*DuxPBhlf{ItJdh5ek=XW}JO$k%$6n
zwf{&sH*cuc2ZT1b^B18TqnA{W`m|!IjRB^zIAgFyVpK^GqG%Cwzw{P7#_)5pcP80q
z*26j<rHZ3-wzQ;0E8GYxif#pph??F`@>tF|4YoKfHZs+`=c$KZp|rw6<W@SC3GbHo
z4!_Qix6O%&EX3c&C|*UszO5j#@I2b<_KFzyTI2K67Q|_Lo-VB<h2T~f@>W8m^z_P6
zr@;}jBEbDb_?KxmqR@TRQwFHE4s`vyMU)*b>2pWqAt=JRQ|L+lEpLjZ7YW%W*xOts
z{Z?bb?)faFxGF-uJ{cTxL$>5SqoSjAPz2&izb?0m?gIIy0M3-7z+>||FS~fxWnd8j
z=2^;JOMsE5r<O@PM+^D#Wu<K6cCKF0sN_H9kV<8C<-KRCZn-0&kR_I3<o`XOlK32!
zeTg8A;HSZEi-z+=<B<pWHXeL%TVJabD0nFh!(kF4OMfI{)B8CXQRsgcxwaR$<ADv+
z1VPU<&!<5;APtspw+mW$n#8^c&aCi)IV_4GvmYO8{!;>2)9sE{quhH=)VmC^dJcZS
z%MDE<M7$pkTO#-2;qiB5IbB~7{a786V53^>)mC1SVR->LU6sfyT}$CAvFfxVsfqf*
z(f^`~XJigL7sELfFErEd!s4NddwLVpwu(QZm9oRzts=)Kxw4-J3RyU_S)?wH_MPjp
zQHV||ThR%Ri}hmeTMT*VO;k*(8W$sJ12aDTkd0{K-RiVw32$S`qu0F`!Nig1)SBFV
zxAZdZ==m~&IPb@PyS4BRN3Ax;BXrl`CEn%V4aP^ha2TV{`T2#h@N6vV6oOgAQ^N1}
zqEQf{Exk>4@Xm-dE&c}WlhClFmZ_YU7B77D#pe$h!_8N(xIEUPEl##?9&dg$M`(p2
zC#jDA;(M&)-aE1HrGeOV2>;j<d15^r`2DHmqEmFWN5j<dpG4y!Q~x!cY=KCVT%nO<
z-u}P*Rn9eAs>$Cg&G6>fAUn8C^KUuB%y<itpTN5-<Te-@8-Tf7+cv>mw3x*EBt~PL
z2E~3F@TjwY>M)&Kc~wH9dHUXj&}dyoQ@#q~Mfgsh`%H_JtqH)kPUoH{tQzpV^1m8@
zSQ9pC6q0#~GX*gR=%GAzxKU=2A*~$eN}vX);oYi-ed<S4FUQoc4nq4UEJ??52q|2c
zjATYVRQV}9Yf3ygXzMff`d6>sG1Bu>2gJl;xq)NAK6JMa9zcmT^Nv&L3Qv8Bg`=>u
zdZkFJaZets0|WN@pG!7qqx$k-=D4bQ9_zHD!!vV`buwqrh@rw@{tCqJE9Q)E0Mnwv
zF>hPPu?m_?JUotcPbLQC>!-MQ5o`+)IWJSr+$OH+hqNo)?W)9Z3|P#+2go`1TT#MW
zttC_&_*cb0k8VJ{<c~}&oJ;yuTIaV|#E{}6&N{RI9SjGVE+EuNV(QK*+4gL9KGZj!
zb+-g~SHV~IM8pH?T@+H;tRzZ&X$HavIGV+oiBJJPh@4XninqEPF788ltiEOwIVYI+
zT@xxT9P}@*gzZrL>@xyU)*{(DKZG~fqF=86F1kh~=+j;beicY*ej45KlqBxQbCtFv
zEoI@gilL&2_ouv<zp9-Fjg8wci-_7!f8uhZ&g@Y0KUU&}5=6|Qm)$niyt&&6uWE1_
z3P${Yns1lM--)Vu3TA*I0GD4nHW1>@Jucf%H7TvF8S&3pIh?>R6nI@ks5VU_$3a_k
z#32FHv@#@fy2n9yW;Z2j9SL`7AhTB4Ylq;e_mDc4D;LGSBLIo%IJD${2LMd0=N6c?
zP~_~n=Z>7<nAj@28$#Cuu!M~qt1jEpfgZ>P<kX*&*=k>LmJH=_SKH?6h>ewy08_RH
z&fUP9bT?QXqvsra$?I0cA2!8`OTW)Jesbf(0DYtKWu7>a2>2P@0j6CHg=WZ!98I-^
zmHp-4ZvmMjeX<}2g<=3x#p9d^)Q(rD)Ei*#^F0YW_fq-_-iCo6*IK$czkCF#zr@7U
z{35!*u?s6;zW{%9$gT+W_zd!$9-lW5m%MVIG%}tPG*=iF8OkCY_}%o|DlT5nOu#Zm
z6ZM#vqA>B5|G2n8FG)9Yf1nz+<?-^FVS>NwL^pNFf%RPYBnR)_C+>t~C)Q{5oa<q^
zvnKBiQZ!A`onN<FED$@#6F$%AuTg2D`d7*5I&xRxi6;R{`<fh=-QlRWZHmXyQuE?=
z&br-6@qWTI1JrIse|U+DH3J0`%V&t10;F|Y16xb2Z_$d<rX}B4Px`X3p&E&(Wi-mA
zYVhJfB7Y$*qs)dUQqzPRze#7R9+Mt&D(zXiq7LZ-I)8r|Kp2qK)~N6c(av9~hLYyr
znsdobYIr+)Yc3pmQV`_zpy_(pNcMjhOYZkdvDRGfT$vklNBXD!TloI5a!O$)?C125
zYs=7wa(S&>)Tuy{UD_(D7UZs3aQ?!&u$-DI0OpXH;rBF;-Do2aYdFBPM*F2owQ-n$
z0|P{&cP#}>F3O;4GOJTuvy0tKQMudH*jGf}M}0_wmRCKAIMA;>7c6G{0XPlmJB=HR
z>Fp;0+f*+oedY3pfP2xPCzoGFx3<xPJEfo2Sp>0S9DgTD7R)W+GTjawK_gt|0ZFe3
zI6I2X<NmnW20NcXlSL~%w^w*v>@jG-_LjUG5Rf&%SVAyO*jB3_la<Br?*++MJrx9q
z+X+Xmb-D)Pc<#bBU#p7m2-}QUZa37XHNGb5e{$k?$SLhOT}5??_cmuJfz5$^G+dxX
zNv}<%Zcv3S&h=o}FH?gz+>!V0Jx?$-cqZh;1hF0V!oxu)6Ihf8Y~Ls=c2J(E91_@P
zgXvmXo2Qa(CIl3<X0Y;SoWI&QJCXRz|2b29zsTgrtM=EdD>lFjrQ*v{M$(cQ?+J%V
zBtqa+EdkEMZP(3mhM_>rLH+#W#UElYI~+EHNf|?q3kcIy1#>EjFiQp<%Ow~SC5}+(
z%?hZgC=xG&u*%(8UKtLx$a*pIB;t&vI77hg&mA<qs*u32Z#PrLC^*J3xNqLyR)NgT
zfvwORN}!5+CK~DmlqH6aLyH2|SUpJzJdPL&c`(q{r}~|^2@Uak?!X5Npu{R7e!tj7
zgT9owsUBiNZ2k=FhWQJ;CMQUM<?tjCQOTEY0pFAAI#C0GDnfKRm;WGIif@%--%~*O
zzy+c3J9tdrd$4V0j%2(}gaGd8`@inF+qao>T;kmt_C(E#IHab8QZBZl5Q7DWgpkjj
zBN6ivpI$90EBQ7M5)K%@@6=9sZu&DZ1CFBka#_eWqJ3ML(O(3sRqOpMLAl)WZv+S|
zBor%|yOBG9tMC^1e7$Oqq-q@v1ddE^J7{dbB9Z=T7-c_&P16c1?i9XRBJ~tL@!_D*
zsOPhP-gzIaW-M99`?9N!BBiZ#-#V!kGy2m}*c8p#vl;OYtaKx^vW_mtSyXG49#nDH
zcc~w{Q<U(DFYxUrcx@e*QTkf4nRA{fS-rdxe?5B^R5r_SgB`MZ>-Y9mZJk!rk-~oE
z8j|HCvXSRtKTS;y-okdq;TVayi9&5=XsFiaoJKBXC@Ib8U-mIilqCB32a5AvU;+>@
z@rM_6Zm`t#jKt%Ob(&@)w%MB$lE}*Z>eW(u<7%0ihJeM1s+5%N;_c%1UVdPR$uS&s
z+5}lr9(4OV+E>4<@p{NR(Y&8}zvmWTec2{1L5E$G&@qyx#W11o4zGJ)Px$KQADPB7
zd59J*#n0%Tg-1n4!d#Y{Z=&}twQ@`9M|}VoD440EmU(*)NAHfQ(S{xCN3LA`iIco2
zNJH4rPDshj9fK36B6rR_LBCk0S4q(Q+)*mX4AI;!=w*ZA82bZo=6J!VSbdrXDls-I
z)sMf+2<iX1aB`|%G1{2bOTT;hQE%R13G1py_93qObX!*`!CUs7C19+A4HKeq{~CXn
z0OA=M!1`r$cQrt+qf>!IqF_31qToMcG4z9qbDwGIR7UHxbtzmWTFaU@D<)3{FK!ik
zAA6qmGb7|h_rV45AP^Ww?Exq5kM%_WF}@F6FHS<#P^1;`W(MIEb2O&b(plbW%$~x3
z14OaCC-sBu`0w~evq?f@9xvSdM}1MG?%r9j5(;m(|JWeq+ixP$n-v`vHl8L7XIK$1
z3@3=XXW5*sXjt>6^d%>5;b%RPHp(umC$gpDY-1~V(~{$&yoWZrp4J1cY$q%1hc?!x
zso8WCQKoVT^-&V_#xp15^f6EG%!d!pcSB#D!Nj)E?BtUn7E+8_xC{^R&rkYx%4K)*
zPwX=qGKiMUSTVF``mF~-V(n;ZKgB~)B2>bPL!y6TM2-M%K2&L6)M)#wh*_g4bAuxi
zl$SeN`|itsj;A}=@s;0s>2VHsIjSEXgLt@+i?O;?n9q9dYS#28wS1vt3;MI5!4S$1
zW=8G!rqVst(kJpXVDp!Z5%&b}CqmcRAmTWW8zTeX(8qj$067&ScN1U-*!?krm%@eu
zo^cCj4|<D*UfqT)yXM-RvF_|O67IS~PHLmNh?s-Jyi(>f8^0$>_Hta9-7pU*IM>g+
zAa0`HAAjc1_bsREN9mW+E4D<aJ*n#0ja@OwZyi&O4B`q<HKt0x1P54BuEgG8iqM8H
z*s5bo2}}IW+khK$84w0ucUwTRRdi%Qot7EY(h1Q^Bf-o%Y;<GpdQ35L&iMH&v|SB}
z$i4oTK|95DjQRW%^r<^;Oh2P(gl*=OWijMftq3+h&eGWy>s5;gIO2V1EcSM`@ycaV
zz2@*6e2$vaOq_r41Z3QHVC^VCRycWbKNh$c0Js%^wn{^JLG*K|*X0H*wthQ-IIYVo
zLh1X}-Cxax<K?AR<F`PE%AUe+{9aK#)uMl@_Q$V`2*ZbAhMLjj?;Y$dy1W>kP>%lW
z8|1ndO_lVTAl@W%)4p=LVPpn(-$`YgB&sfe?wgZloueUbm1=|jDAlq^ocDpk^zIdl
zM0N>PxN8o~7bbU})IB1o+VoO&3_cv$B1Sd#pgy{lH@m0{nWZ{Q)oc&~gJF))`ACrj
zRoFgBg720S(-{Z*ns|hsB$Z=EvWF=Tym;4PbTy($!s3m}-712PZ-3O9qj+?A<5l6?
z>wkS3yeF$TGe(PG!)4wMz788PI3yzk_guxA(d74h4hZJtLT2X<@y*WV@&t_>;!PN#
zYFjtvip0(oSAJ#Qfym<T@)aGaV+(&s+(V;{d%VfhuM)+z3EtFY{@e*ZZ`;keY&lQW
z|HKjU5t5<d8w_1X?o&~|`@I!<<0x&W5O-HMy~2-EvkM=v*elM!rFKQx)O`%!&c2mu
zmBGG7+XSW<*au4U>mtRQJ{sId*PB1?>|d^u|6d=Xbm*V|(QuFQJc9vn5}mT)PcdgC
z(!yGjOLm!qv8+Em2us|uPoDGL&*%<)7y}`%0zF9Yk1w&}vU}>A;tu3@z8E__!)*e_
zATe0glFC<LQ_}LI8tW0InXrS+&wJ(X^VNxx|8n&tI>5v8z`o&aK3>OE3fpZMNRBz(
z{UG~M%^E~1*DR+PzZuq~`0-c<Yhf3OqKy0bB(Rw!aY!m~o~1WkYCr|)x+>7rKtTM+
zbH=atljvj@a(qLm@TUF87o1B<Y){2AY%&s)=<a#467;lMzj{8J>Kvas<Ir<?3eGzQ
z*Ew!xV}hs~Illyh-M_<*PpDciuPs6GmMi_fw?(P?JrC<E*1Y)mL$d)jq|$b!1;V%r
zT3ncFWtO-=m*D5BIGaQ6$G#=*#5SvAgh+N3ekQkbQ(^Xx4t0R^wPpu*FGsV1^#We{
z(#m_()qm2Od;@U*<z(v7LLYAjSm2kmnT~@hfjA8_C?ca}+3fw1ylQ=tS9gfThv5r>
z+!0R5XRF536t;Hzx<3!O*TTg5PPK~M8%F2>{Z=S@sZVxHwVyLJEC=S@n6RUo$Lhjx
zdiPyD_X4?4K*gx)`H5RKvgzqin?S|TPS*$HpC(J?YnI1?r$wVa?gdf`Kk14p{RsEP
z)03cRyvH+F$V@zQS-p<gUKr%!=@3ihXav+V93Y&3z$PKX{M1k@z}sr)fPdn6PY7C7
zmE4lSFq*FR&hM!i;1QMdBjWy+c+y|qE$bEkZF-W*&IfNww_{<1xiR!7cYfaYKK!o6
z%+uN!`zFJXEwbp6UP;DxVaYQkbF-d#CV1Mmm%&piGc)saC>0fgo$i~OnzBx;?Ee<%
z1Rg@(o50&XjcWpmH+C)nQ&Q%MX%}A2&P9-retyHJWhQbwF!aj+(ebvE{sZ6(=+Sa!
zx>5Wy#c9Ih<VgFOpHGN&g3a=S^1oqW{gzjC58KujJKwPB%L-PEN;TgbS~jfmBe*AE
z95e2^lIG5Gti+BKrxk7aDKW)+t+Kf9)#vx@?8UNL30bXJy;c{MgX{8)qYonS9U*b;
zyc=w`=qMCc-uPov5FwpWk-@s5<U+8#D}&qo*e!$isK>u6@3L;o(GR|RdOsb_aqN@D
zwjDO)8$Ggt5D!9-zZV_i&RB5Zi1%F#?^MJbCHf^2#rx^1<R<JsQN{1SRcA93SvM{+
zJ5>h$Nde3Rd*F=K=7c0U6aDkuyF&$T_Qsx`=;1!l<<weF*Hvz5PD)J}8?!YVRP8c`
zo4iNop80$`tH{)z<ZkHitLJ`5_&xmYXIVy+;#C>FOk!L%m&6>T6UPgQKY%yJnEc_U
zMIga0FXZHPk+KJcNH<q54`1oe&a3;dDz02MZ^XT1`D%K#1PyXE?plcLJsv_+{?0;|
zj>twt>F`^zbh{fr6UC$sYfUQlIg5?}Mezr8&KI9%^=x<k*u;?k(PO1zQYCk$?w_RN
zi#@Q)(XHHVGG<_@WDDUOBld|Y{~|wq@-R1NAfW03sbRFSr8|v-`V!CM7v1JfF2>D%
zd~#|q>2?-{v-x~o2K$yCjyv)QNt%V6iU$`f5e9B^=G+Fy?*lO_re{ClywR&d%Uk^E
zhWf8l<zH{VZ&e38m#TQ8#Cm>QW#Qo@E%td&=3!y{Zp_eu0$S|z$(;DxS#I&UISES2
z6+ht4zGy3ZPJ`{83jsIf&wJT>QS+4ue?F{ltjU}A_-C3A31xRhpUjq0)%U)%$SUdc
zcNz=ac4_eHAz8FPA%T10Z8wycDXE;#5z2-GleY42dSemuSey?97{gkxGyYQ%p#|eL
z^%VZj7IyVgb5PPU#YxLodPTWW2~(4Q9udd?K<v>L;@>i`AOK=~YKd$IE^c(nhhewn
zugk=>`gJ7#MK47m2@>2668$f?SS1!pMh2*S<hRtW6)}Bwyy7H}$-;vD9~Yoi+qY3x
z=kO47Spjh}FXAX9b1k~}JfTPc*H`6bw?WQ~*W+e`6?tw{fm_drCoV#DSG1+%OPhnU
z<K%db8RxZ{yKMA0`$=60j5_02vYeV7VnE$-eQV3V_E*7K=*khD*?5hOrmlf@>c!Zi
zEYo<-bV&}7+>mIm;cm1}F6TpHUmW&H%;_Goe^=zL?7j0P;3w73W8OqYlh&@;-b67|
z)qs%8e9hNDzb-_gqxPR6^IvnwhY;t~uVrxu6)*N>Z1R<@(^moC=N=f=aQd>nqr`bY
zGRFns)wv2+IJ*6=b4zd;fcQnZ+jp72lTj~UUg$swamG5ec*|j9nvc29WiA>t>r)yt
z@+(L8FG;}d$D;_OjP6-Z0{aOE(pL=H*e1H)^DgNmaqRVLS>JfU7DN;gx4<BpNp|TW
zX1UW7dSC;4udyO&_WZHP{_Qpe7rNzK1{Ol`<IN^!VvlCw#Uw4xAld4|*7n$|`~LLC
zZ(3hb&mRX{FO@~eQv|^L_7wkR@7BC!UD&HpO_n~L3kK;>6#7DYCVNYnr)!u%%jE_t
zr6ysn(2jl%zk|QOcVkRDQM0hibpzl}JR7d|_1m}RoKw_M5j->Lm@)9;?SsIJJ*6PF
zS``<#e5)h2*Y-Z3HecJ%EE(=GEwLJR=}i}>uB>!d@`zQk<UpdSW!$y)iz!VBQnNRR
z5uv(%cx#?Ib#TK|lKdr`_8F5sF>Q<})QKw|-Cfrlz$W^5BusTV+G<C+Ef-^RK8H9#
zr{6Z`Wc;ubo9q`oZt(Q+9rX&cjsy0Y-nkJo&I=!w*`CFREG68jPQog_SyX=h>F93c
zt8R3mdTyLnzGKGK0R`ymyO6cG2!ZQ~<u?fkWO9j35nD60(ypejnL12+fLMwalrp=c
zPpO_F&Q4p~>l)+uu#<$YyYYUeF7SW^xZyqtF{*vY@585y7_gso!=Y*kf~3`D!3x7W
zw)tneBLv>xjV(X$2e#Zo4S>rFg&ox4CP7&++r;JqQd^dAPr^slA*P83rA`pD8@t(=
zR)%%dm7~+_o74af1g|zjqCF;Gt%seTkT?&yo_Q`^cok(%?I|d<r*d;8*aP>KwTXIM
z00J0{smet4m+x%37Jgl3<MVt~#S}=2gh+0(MT|y6svu)*!JB*VFaYd`AIl{s%D*Rc
zd7o-x2zZboaf2|Gvl<ei8pftz2u6Ln!2Z4L2_=*x8)4T=`+{Ym_j2h!;SH_Fr&iKr
zjrZciO;imT;>)cg%gT0Me<AVi%8;y{Uzm>VK*T2OE)BiuT*-fXl-MZ7?ComX-|@7J
zaiiah(Ibm_R$HIJF0(oG`idr?@#y8Ii0#Y%>Yv#(&TBI<;yZ6|2e_SvTFxi~*%vlS
zN<^&3vuyb{HX{XiPm!*E7A{@zxKGXyE3tB)^@4&i-eXn3Fl1S3>timjM@x&M`Xj)j
z1UroJfSSJz^zY;m3jR|6Ere*x){gd+vs@Z|6XW7#*}73s!+$iIOxWV&;ro43(p&t8
z?&D^0<CW@H&Zhq4Z=!B?I%fjmf|P-{VSjvz(<ivf4TcfIgFmO2aVN5{s!F>CrR^gZ
ze3(IJQZ44ynZa!X@_^yBr%Re_7b3Hc2?$ER%%u#mNmkiX^etpwFUN71wHYVgB*ph(
z3~tG_*Zpn-yF+W}BnESCwl;%W%v~YL(04LA?@2k1@rb_}ZR-RFelrJ@-oxKqIM42>
z?YTSMRX@o6tImU)JHi6-#U1O`5;|O`IREZ&`oZ9g&V^HaemGd>yjd3KE7eosclV?k
z+k22K0*>>KzJi_B5e2o+^KUdF|8#Hi<fpfHwM%@I4mHM%<AQ->#}v;4g^rYGA*&2^
zUPd~WXI9x4ElUH=pEWRpXz&#G<=^a&T7KgFEz`;$nb`t64UOgAp7Ls9(5>4l&1q-Q
zUV2hHNUyyq82aS^>>?m;Uv>Q+LjS-apLO%>pD}@pT}AT~=bSQ5x!KeNqA(a9xsZ%S
zn%BpjUzH4ohxG#uITnM;Uysam)SxUX<R72qidyWCMcc?1l{#*_Ox2!oKNv;F!0CN(
zVm^sivhE_!$7dIR{bE#@?v*eUvaFJL<jIcNEikx>fiw8=^TeZNk6kZb4O5QUU8%9&
z#z<tyWFL$6evS&PaFT;dM!YPY5ZtoF&Ne~n@h(sb(=7LV`@j&%c(waMNhM1|p|=4r
zaPyIWr7Tx8M$qD4UyVNO&1@w5t-twk`gn|3YOCxQl`j>`AXhfaX^+DoWxf#GY*?;=
zfiwH8d0BT&a0tv;_TeQayU4Ee-J^};RoOreZSmLQNxO7v{YULHQ%SJYM-NhhFLTT}
zkJQeu$^vK0CJE1lPr7sQNVRxkLKJK{f9DPJSBmp)t@NuSr}SXtZ`jAB;)0!~-FTuy
zR2qmf7BV!Jx=`I#JaK-~1HU!@y5qu`4(dY8hbNz&9<iJIF3`0>&aKzN!>DndWZKdF
zT{A`c|4!i?MZCORh~YCf_NlK=9CGztMxF18%DMv^L-#W001*GNj*xm3HrbOO_ebjP
zlX6aWpbwKP*1-1G<E*pQAry~@n62dg7}e+e0ie)7xeS?*^OEc0?AB5ou>xu7cgcpH
zm<Qz_sNP*N0@GRGY1CFOx1$94wfWjw=;xpE;{_{#TcRRRRnxtjo~DMv2icUd=!hj>
zvggz~voy5cUfF8Q_H9G)iwV5dM-^(fxvqAa%6G4jF|Ag^_jj5nvl=J^?fe)HYz36$
zEA$nFhrcKFf33`?YLwlh{aml7bp7kW%--hC?w06v)4qgyqzv_Qs^*n`r3-ZP>ipt)
zmI&Z50h3T5n%qzNhUKke#Rnb0QzYLi5O{!aX-RE~+`g^R81-FY#<b)Ox&CGYG5DfT
zPbq4e@+ST`B?WKB9T68v^U>OO>)i4?8%?6p*y>Z?rh0bIn*z>^BW}3Exr-SpnlAoP
zNT;@u95`6wx1nCoLHbyQsH&_a@k1`_N~5E3=(e^#&mGRIFrax_WFk#Xbk(#+qzasb
zh0Hu?Mn@u9X2Z}?J`SmxfGlI3GZUhEplR4FZWl7E$8?q#EK{>Oq=FQS7>9{DIMPek
zOy!+^Iy5`@syN3S{9doDpgezLd30q<pc?LMGFkD~QWre<J<WpPhPx!%X9)=PZ7ggB
zrcsmB2_J;zfigC-t1;Bd96Jd3cY^=#m+wsHiI?GVU*nQJ+&-8>9+F0{mpl^b3C0ia
zCe=@Pm$RuN1hz((-mqPuyw=3?&>@ez-^A$F7gsVi!{z9<!mS_V7T%s4xC}{(|LZ_|
z2t2?GR17c9ugCF9|4Km`rUS^B<e3J-axjrgViUGXOT))jvNteNZ#()47c)L&#ZIqt
zt_qKmRzHb4kq#7?z)Y0-^+SheHOydDq;9X2*upezk&MLlQ#ZdY_t~1ANy%VFG7TjI
zZAX-{m2Nx?X=svyzU>7G2(4?^@5cpc2;2`<T2`^t2EblpSZ<@6Jyp5@J}`1lTVq3!
zopO?y9oP^MU~71D_<da+O`33^%t~+O2W<N?n2<YWw_H5tz3iOU_=XwCh{!cGvmBHN
zaRpjbk<K!4zU8>Zs_i7}bw~O^l@?A+{bA5R*JT)YwR_Rr#*hcGYW1r@F_toC6^O^V
z=Q-%G#i$<FH#R7kabDF~nGJh9o)`aY^PoGjS1U?X2qDH@cV>rtRg1paq<?k9K3n+r
z*R)4ffEf8>rz?<*4Dg@?U;l1&s=qEYMU9zHS8_`>mA_d$>EY%~a9=5hGOh<_Z9rB{
zFK_zuTUodX@fsYt!bK4?s3ZDl%>mSY%YnAjL=B2NQ%$ufJhLUp36+!~uzfGV$JG1r
zxH#;XK6*(V%c%eVHUnrn|9!(s4ZENpNo)6=g&s*JOy!{+oqkEUA@UEOF1!d=FeCmd
z^l^ZKne-DUp;-L568E{<r|pZ0^fX;zyc1EG58vxQ?e1M_HFHOKv<8&lJExd>?Degl
zCf41#b?1}*d~wmhvv&JZmzi5R!U^0GH3GlaS1lY?Jxngf1a_O=!WC;>hlvAQQVnE3
zwjbz!)u&4JBX}u{u(?6&YxKr**uKGgx!H2UR#P)N-uSobp+7c@3Cqe?63JAv3Is+5
zEfr*dAv5irI$hE^b8dl_K5n@HQv_q<;5r)NUbetrW*qblV@rl}Aul=0>Ka!q){x^l
z#=L7`8$AulzOD+B51Om~NM@{T;G1%8_qhB~o^&=_FV0}Ae9{0XcRM&w;-y~A2eP1!
zW)yoZntjWN8?Eju7B8OBl)-xXsV<-gF4iv2h8y*UF6Tn>?c?Ai6%@;U7#oc96hqXg
zzpt4;N7F6NyZcFJ=D{~~L<0-16x!f;XBGn|gepH%zU!vPoTXM-RbBC<*6wm5-^0;q
zYG`t1W_WOB#$i@&iqRRhltGibl+kFS@|Q#%Iq!du#l-f<5jds)-%?`lXz<&(ZjD2d
ze2jgOu&|blh|oD+>K4&oG%Kw?=UtEde&O0~l$?g*pU;04d8m}-ourQ)4c}4T+lmZc
zec1ZThOcFh*i#U9R_DohEs4NU#oSu)R!yV3wAq?d`_pV~$0rYqzQm4^MN|jbnQ}XB
zIJVTAx!*#${E^@?n7O=Q-20Dur!}bdw;->+gTnH+UCUooRfkn`M^cO*XBqs*(@ri=
z7h@NJVXTtG8<1_i@C$Zn-+g@DD?Bi_+e(V~@yp9&*5C7i`^a+;+GP3X&!2aI!tW}3
zb&PJz{;sPOy;}fJ3uXWP|JeExXsFxv{}zdgLU<x&8KhBKFj-33ccntaBwHFxk$r}c
zJ!Rj<5?N9)WM5`v-w81oX2@>rV;RfvzdgU-`~KeNJ^#+Ba5|kc^PT&?uIqDsmRO6H
zADl)NPqUKqQ4O6&MV8FZrKFatigZ4%1Mv4=zg|zVBbEcLbi~16k;NXt;XYYBH+TGb
z?zfA&ODycpCz*<U;xU}aLi@=+CtI8@D=Txg&tK<_rxycl^`iUv7hpFo<qj;|gv##x
z`OZ`(5(9&*@%X^tq$~WpjEY#+lRwLYt3}ogA>>CVax6~-XXq@jFykE}0qe!*W(qgE
zJKbbLcq_mP4qxl&r=}v{uXMkhVyv=sf+LqIJ~U=c!?0eH{>2Cb`vO4+aR`k|bLdKR
zy@~c>2Ru8p8LyN{<I-AdxJDf_nUJ{I@z3*m?!y1v<{9V5^!DCGV`resMg63C%+;La
z*NfM)YpD*0e<PB9DvKCyJnk<E@w9!DJ^rhujQmoEk2SgSNMNUFz{uS6ShG9y)x2-a
zK*UVlZK9#YV<glJ<g$s%C5CIR@AQ~2oQirTC&W(uoERAKZp1j^caN;!)Q4PJ{TTP3
zShSvE`O3=U&G9ISy>p>FT;|r2(p6RGpL?cr2ez9Ai0Y!6;(@)L-+{1ZZ#Ay^Q}pnw
zI+ikEMWr<q_(nDiu*c7RlwW1Kd~X>y_x0N=;2&r)_`^zdm6Q4LnHB16c1oima{q*v
zCt)P0{$>8Y&v6SZdb7zvq{fvky*obQi4>~f_omFrOsa7=3_`OOCVkJRi2kwWa7G6k
zO6QAJ>6Ud6)%X_~S9804{TRlUJxX(jyONa2;a@0(LiPf!36{B?O~G*Ih3(x8(hD<%
z4v&Wf?zm%7c?c`gdp1<z4Wz+xOiP>28ViJ!{|lo-?>JGtCoB3V>-OoKN9;MuCt4mk
zcwiM$Z%~9AU@@8)X&nW6r?HRWY2Ec0uK^o^w?b^pUR3gfYXIVSPauwrY&{<jVELHm
z30^7DRnk=?O2bu3G2<@yjM~>dM_!L8z3*<#(E9>d;sdZg?Xvz<ic^aGKhJjEfs-aU
z-jht;tD+f4l;H_acBl8zoJC%uZ=r|VG0VvYVyp8(Gl+o7)K3K+DpmGc8%o*icSTr!
zxBpx$<VwHEsiubHJrjJSKvw*Fu9^-cQMB}f_A-l*AiO$G+0$EeZ|76tN;kSyZ3De?
z({<eQ+r`H8a;*m!bA4bZS`341dQ>yK#-e!(5cJ{eo8!PzQ<}8$p`fjHd@>h_hW@Pv
zISSwX!@U9oR2TopihXz5dr$Ol6T>&aB8>mSN8TJsp0M;1RsA$hGh_ORMLHF2RO%(f
zC2ssav(4aHXO+9EafjP1KaP;O&dQFmve_t!AEj85_^%Lt8~Q1ae3FNgoLef=Ht0oW
zqGTQ^Le}ySk-v|?T$$!9AU2?PrgKD3nq}ZBX3=Kp;qMZQqUle!qvYnb%5@hx*$C8C
zb`R4NNa-Ay$s(4cpubIPnWaBryIc(aBS|+?Clbr%nLI(!M?KVlZ~&m{-7{O0Yr0si
zC8_<vH}Ju@i)2JEh%?unj%^ajl-%wK?q4LO6r0K!8x%VX{jjE16q{6fR(O0L<}D7)
zR1~Z9k3^l+!1?!}>lXZ@P&l>|_^lt<@N>vS?7rLT_tbwp-*c?Mv<m!0OpO+$ATVCX
zvVJ=)=)Hv%yqt3o7B$#EbbiEr#ZNYY#Mj+iUdQtp65UD(eSRI<5Vo0b_*u}%0Cw^V
z#%hRTrBsX*I;4IUIwgM7^V;IrCp4#bYk<d%%zc&uSxa#6E%fMIxh-C(*`I|EUUqIz
zpI(`^mv}Z?H93XnjzR@iAP*dCH%i7kv*MQVx8XOZ@JU|Vvqlwf#1b)>>wQl>7m8`E
zhIn7cu>*190i!(S!pga}ld}Y%0@8@`dTL@4+S#cKiDFhLAEI76156XO(@L=ybCXAA
zE~-Dn(pl?$GofR_40gxh61jz|xmsA3ljR#H<G+_dZE<NCX3#B0-{mtJICBAlgqp#O
z6!D$q$lcKn;?*;Lg^I7FV_@QovNAAnAsdj98Hu%R(###(c^=zAU!j~7c>o(vX!<3*
zvi$?8-oI{l1RGR!Pan(aQRnRENK{Q(LjYpOA5DxPEvpy&4p)bCP?O)3=A6~P+G058
zF6dwr<%zBagT3yWvj)CI`$jU|<Y(xmjF^a6Ac@_EmDmwCCK~&`rH+(XJ*3`+!&^__
zoi`?pE;W>fkMiPmxQ(V;YZv_e$;<X`)k+<KVu`<iK<EEcq{SWp?B$kc&ZqKRxs_)l
z%NtIyDzR4qRruo}50RMEi1B(8g+YP{`a@5meqC%J;swlOH~tuTH%4$pW7{49=e;_x
zbp|z^f3o0<i%3@Dl=4N=@<7+}$E?Rj*OC4aeCT4CxQO@}{p{OPZ)FP38M$tIUb`vl
z;=1?sTV5@&?N`zk24RAe(MwK{B-Cn)+e}*EecgGG%cvRzHfDu7=?FucYzh0z{Nq-?
z`Um~9`0QUS&E13VA7W%@hOmnXB{#l=oHMzn<%^uroU@CxcL+h^2QU(}fJu|j^;6Ad
zSH()-u2C3>=VuLo=ZUxVw|vR->%yjDT@lW?XOm*Q3N?Pp%>ERRF<N}&;+?i*waHzu
zbK~RNfEY*gR)3*b`H%S?pbELr1@3`n+|@A{W)-T4l%9pbXPnyviQ6~wQ%g>|nKd$1
zMGjDUQ0N+hCSGI&X2n2`&bX;{alyP^Z`s)vx8;b%oW_+!`)0o_i<_UDTY5q${|uZz
z`t1>}c7s3c)tnw?bK2K2<AHIqvu)MbBl2!Ec~`8OuGoW2TO;kRMr3SNOfK7dm3!{G
z^#lIie>}-kM>&AI6fiXSuq^7W*m|uSCedeje<d)5b1>VsMXB`8H;$aXU)#T0hFqoo
zTxc#jM=}+GZ&Po2g@)E_5vK`0TBk~!5`HbO3jXX<ejXHN@b1bF#UTt^nEBScaG7UY
zuUnM`6z?^&b<2CuEB^B0@CLgBYVq`oMPyU!mK<-n#6CW3>ND1^Jip`uMh>$a7WQ#G
z>vhJq*@1K5<q_T33As-<jmX51qt+7VoOprH+{-7y_ydB@0irtwT$%0}8gc_twanp&
z3DN;|cWomrw~#lN&va&X0XB;tD>99E6p`AG5nWCE;p>aMGze3lUEg0V^u=m*bm+8n
zuz4#{Rm-Ftk<w_`nYZ@o4w8>xXI!Wo5tu{b3(9B0w$W{ajU<U`Af(SLrN?@BE$2$N
z3<>In7E$y)>1(E<^FSJI!+|PH$4YE$*CHKWBEttLtyVky3j~qFmA_4s@5wHl3nnn*
zb1d0|Esux+3jCh47~WQ6XI*$A|0geGlPkeB?q$$0MxQ5GyF!RWEG@6hoSd}BNawyB
zE_#h{Rr5?`*$x}poI}+}E4UMQC+!8&K&cgoC=D;dz!&>jjtj&8KMgccazxCS@b?+Y
z^WkO|^_DmjZc`V4oh4N8<++FGrT`9i`h?CYv9A?O%}EO*;j8yNb=Nysg(QZ|wOb(R
z$d!#i`1N?Odh9?iSnq|&@de$CoA>YKUh+4!?yRu{zG6q=#N}J~{5dav>7$j>?P{_@
zc$2FZ<@m=bepzm?jT9V{M3SS@gDsC+%9l_b2(!OaFrqrv3Gk0tZ#sp8v{->>P1^ni
zd$pl^e2h0R2{z@D;<asi&^L>;6#qU*Vxld&R);3L3^Ci=+pE`)t9zbXTU$5x)T&HS
z-q3-l>_#gOYV-dJsh)!nfy-l+!Ebd2a{hZ><J8hHiRAz*V>Z`haH6MVePG6tnS|n{
zxXL?5=LX)4q<*YG$IFS+Sv6{0WH7hRYjTt{aYYS5d+hm_Dq2Knnh6o!O8Qt+<ised
z8y4<Ix7YrPMG(Yo{K<CdScS|U+o&F(?-1@FWYQ|!g$}hjB?nArnkyw94Tjw+8QN$s
z=#AY$oGr?@dsN+0-}E;6Ho_Q#fjba~%JOipO`J?jw6Jk48OfWbcNecfrnxjJ9dRDG
z2by>xjTJsk2Cn$2e_pxT*AL9N0MI2<n@VTSIp1xVrA_C|R^uhPY+RR3*yb0Ub3Ti1
zQcDZ$x;@a)A4p+&e3=+h#8a#{9)6LzY1#eySHZuQ^<OM3G2F0iPn6-_cO6YAb07#^
zXOk55cugpYTM3LUcKhNiRKY<vy5%GdqEeRusc>U)nqMqKht5aa(OCMqa?<6Tlbj%#
z1RCLZ?K>L<AIBc1ZvTE)x>MN5{fd5b^HZhGaaOa6(-&k`c-lkyKe&0iibvK31jQV9
z_8geWL`Q~7;Qu1;ZIoUIb{R!Q^VeE%rnhl3Gm$yG3#KD8bF(nVFKa|5Y=mGiPS;J!
z7wfI?_+AYmYk`Xm<#!ey6;0819Fg{m@_hW^Cnm<Zj?$j{qRC*<S^w29^4($QQv3wP
z@>ywEHFDppf?@pWlUq>v(5<;A=M3LDxbG$Zq-1Y~BvJe);9^vjxmPcd*DiAT2&ciS
zIx{yJgYkJO5{KIq0B=_-`+GA&h5QTRh`gL|gZ<TzR>Jn+S-9wGg*ztH;bF+JOdo9q
zbw#1H-sBY@pDQC*u6!%WHQE0Y%R(t#+uVTcUh!3O!)u$UFJ2MZ-~z75#Sw7n3+1qz
z?}mBrjB)|DP~77Oe?QRwyN(+V0UrkNZM4wtU~2dcE_4U$k5q`ZY+AqU?H!O5MyOti
z-n8;Ke&<v!uSoIH-uDIGcF;U}NVpsu$o!H{J8R4l88Y;-(BS-Aa7uj&W9P%rHdiEb
zYK89ASM&kr`7h2KDN7EGsV4pA_C%3SB^Ri3JipH5UT6L~1{n|Dj)yKB>^MI1xry8<
zCg)f<WPkg15}-|&q#cHXg0o6WxDJy1R#G;rY-)*k3u<rKcK5|}uYN`GR2go{ko364
z@#Ncongmhj9RuGVc)>OtSD@XC4uM{+J`dk-gk5u*(zZjq1UaG?jb&xKhlk7|<UO||
zIY3_V=V*6dD)+Fxk&82Seb|yz#Q=x?X4NIJ5Y~Hr3k4mFYOh4LBYLVcQ`wZ$%hw&*
zJu@9HZ4ok8D(XqEynnRd%*O`)YPyFu)LX1#v$JB)=?^N`#jp3GaQAy0!g~-L9@Ho=
zbU6$<Kb5i0*z?M(oarZ1znqw%^7x~nI5(9w9TDZZm^IQn1qn)|*uHeKH1Wh>y%`MY
zPY$1fTAm#h^M51K*V{{|#OGQyX~~t?Js}q0+|`I13#R8@Xe@^XFpw2{a-*EpFvD_5
z!zSu~_%!GFE`y}1uS;dUs0E9Tn_mC8p)DB>n5_Rg&SFm8g!%e?Nr8vT%5X80)sNVu
z%$+|=f&HR~za*^bK9b?MLYoT3r;ATj>5Gg}(}H}ip5rz<QE1cIu((bwFrmvQ;a{9S
zU--;szAyh5<YOKM*)+}<X7q^g)>0&lES+0Bgm<x<SyIl1FdY4odi~@{1n>J{!x=9f
zNzg;j>noi8riUh!4VAC=o2-yF@!T9OtufFQU=PI!pvXHSiEyMRbKIR3t@aOn&&};|
z5toP$kH+(2e7JRSY);ReRZ6eC*?kSaa}Y{=Ag?$u$j<qjigKW)_zi2XZTn_+Pt6sp
zv1(75xj_~5lP!nw0LimNRmD!Y1b_*aJF)<_Jb2F+YZs-%%&Jv2c9xuPM+vPSWFUWE
zGr3g^JLBQEx8)t>@MJOD`{M8&!JtIRQAK@67qa67(>s%>e%@GT9Dt^|U(>hjZg+*0
z3%J9=y{PP-N|g6MF%K6{*6-z}PC%W*CZOy{&3aNa;6(dk%Q;tIwvk>ca~tvzs6sJ(
z`40fM6l*L=sb~shO8$}UyM65pd5GA*q1iSCPWX3|%7Vhe&qSQWkeK+WtUyV8zT6Oa
z@{UWdf^q)uND=Dz27Pqs$W+&9<E)8)Uf}P~|9dj(RS<{~+D?43mD(>`xEu(>2IW27
z78$u%&DGtEk0J$MR?+F{cL=MJO&8osfDAJa&IWn*EoF9NVkJ8m!dLDa65ht#oK9Rf
z86Wv8X!CG)d{}QgKNrSFrogyevIXbg<a^P6Q$?PO?cv0qu|0iDw1plN?>x~^=0Xs<
zX5iuB@e1TeSOAafcqU-s_1E7V)C1;W@RX7S{aC8^7BXJW=_*%WzmR?ATt?@o+vl3b
zMa}&Dro^w5*{4~2ggV)lHC^<U2f2l8+QJ8+Vqj|C3F`kn+7M@Gp@ukaJS5khs?MLh
z@%}AOu|~1qLwBX9=KR%{h1iR1&qvU0fGOdeYe_*A8I-ZK{e1EO1{Y-2cej_5S;guM
zv*@EIkurLhI(25lXo^^k{SFb%)0(R+?C+h9O@1Fm^m^WuE{B~#o@~ZDkkm`@@f4XS
zurpqTIruE`qC3#tifA)Ka!8$wZME0!2**jJoV`A(R~lzHVFsVq5oX7)VKfuhBmY1n
zv7RlhcNj&tK>yZ^iyZyNIk6-#sPL;g2OM##4lfKp-2b}>m;L#(gqYmgX_duS(Go*k
zm|D<~voEV3BS!?`KF6R&(03z#>^hd7$9G6{aV^{Z*EoFrz`XVE_Bz{Ju`$pn@?n@;
zwcuY!lQoPU{q~LQy<5o}m$qXyLXb65vU(aA%iX#PMRZZ4_Sd|cXyKa|N*X-1!vbEg
zaJ8FRTHaQ0zGYZ{Wv%~O;-c0_5hNlcJ4@yHc`x3GG`8T;K$BG7xEVRm4Z8yqf44Rh
zlJAORhZJG-AXAcuhv!=N(|xdc>Ug|qF+0rpBB+lc5-aZk9Lbbl9@<}gA%1C_6@Yy*
zOI6gYM{Bp~YrQL(Ce^-td?RjF3$kz4*U}0rxzR_DI!-8x^Qh>E%Nor|DPlQaTm4A?
zJ{)rBr05tZbIBQUzylk&<k&^MHRTnY7=ny`ALZ5U@8ASz)GG%M<SjG|!(_KMVOJjI
zZZ13)-cEV<TnuJ)Up0?YV2HU)a8B0UP?=PdQ{|TlFAU@Gw|~Mv&#JjmY>SKB#oQ~&
z(^!hp$UB;!Q!U+LI2SYly@4ER#d!gR!)BJ<55Rd>UG1RleRmH|lx~mY^roALZe{8C
zA}b^=d_+?0myCM52i7#l1J-!~2r%@W##77@(p;+Be+9D2rMc9R+7;(|66C1~Z|Fu(
zU_b`%;u)R)E6@M&mDuBSq|YOA5Gc%>7KT?7S1JfkQ)g$;yl-!3We7d~84Xa>WcG-D
zJ)vaO`x`PfS3cnJ#}!t3g^xBVC!P06?oZV?JXH5t)GQFq!}Z}a;U}lzQ=fx0a*H7s
zAS|HG+bD!Wu~qCGI)<`7(9|l9Y6R-AT$gcBd<f6%^<DM0;YN~ll{=yCKoEbbHIxr|
zA|~|K_OMge*a0Rj81d86X+viEPxYNGu`h|T@+)gp)n%!cq7n%&b_e_@;?z@kBTFmY
zut2LiW_`>#qMHd{<09}HR#bg=l|^_jVA8DgG<={#XK{T~tTvxB&?zz?na0cITS#YO
zFej$)|H!q;FoO0Ovl&93i0)cL3Jj;Sw~Yk`vevC3hX+nn`9*pv%~s{5R9}VCE0p0|
z(}b0_31RCww{uZFm{ZI+^KRIcNkLi&!@RpHd#5H&x&~G`N_Nd0RcU`Ml<?PI1H<^w
zTU++v2>vc|ACKKjNpTyZ%5VO-D){RIq{j79OX+=IWZKsAKZ`FW4=!E*DL!rL-`3Lq
zo^3^#|BWwq?Z11Zo{nvn6&NJ!PG%X{FQ&Ziynec_4QS0R51*aVdF*{TqU8)K@F~J@
z=EP%EQ%_<PN`>_lW>)Zg$b``wITLq>@2tFlJ<^Mne}D0OEa%gvTA&jOSX^}2Jpc(;
zq@|>!%C_gey{2}+EC{Bq4VJI^KP%K%m*C!;a(|ah&xT}8DGcdLA{)i!2TItIb8b!b
zwg2@x1j`|&0vYk%rvv?%%{U;*)mpA@hf35)5k}tjFCPB>Wao&-&NzOQFI3Pv<4H~o
zCo1eGB9`tfGT09Wr+D5x$~%*s^Gp@enS{(kp_+@?y8u!m>he#?G81CT*B`)J!S(7B
zZ*r{$M#MAWK}ciwm()&Pk_eQYxE+4h9`k|Xj(^0fIma16(4%&UtXD{y32jY5wksUj
z-W8@J^Y4h$0_Nlr2J~h-7ENk1P0LdkDddnKy9qP7bLFtfNEBTXE<_8N0hg&xjx;m=
zo(DFVDkTAQA?FXXC#>Do(*uiw*x1JlCcK0a@Fpz`6yQb}{B~w>BGEDu1u*%Gv97=#
z&|e=#jXRln_;k#uoI#H3r9sy(Tlc*GD1ASOX8DhE_rFg!`f_*a-LKerB8EEFX49Yi
zdO}4#i+pmlS&X86?7`(4*n7zK*N~QFzSIdyB#O!#rN;kqqVd<)BXT^;Ju^C@`Ey0C
z7KDUxCxV=;>@skJ;{aLc$gpUm_uJbZ2llQ99;C0^uQxsWQ8t%D9u4HZQ0L$j%Yv8<
zK2?!c$V`@BWpQhAP-Bhs2sDn}GHS18{oCx%p=w1{`_^UOa_ZDr61B6*%J8ZG#-f?s
z;LwYAhJ&RV0E{==ntx|ln8WF)^W+5@th+^512Rn%7GUCAJPk0z<(SC3q;=xvv?+Uj
zl(W@2)%FgJ^<v6o10@uPpF!+>P@?!bt1frw%&sr_Ro^kP+5hWE0)d*HA5RgbRZND{
zD8X+r%=#XF+xiZc!}8^@=pIU>C;IikiRyca*l$rfY%UAHk}W2}UK7iP0#*@GDB*5`
z-m*@eR;Iw=Dlg6AC{!-s??&yP4gUb}LftI(m8`|s{%L9K6Y5%{GeTPA&(YxUfo~_6
ziVvO{2qXty*OABq<!84WK_5iv=rz6j2W)3~_mm{0;V%+w{|1BqeJGp%dX^3p#2~AS
z?ed{_kHTjVF7b3r0l`PA$|()iJMC#-y&837WAIErzl$yWl8i<L+7&iyvN4RrQzyUE
zK0IDlVQYcyJLPxC6x*iwvpsNpNN^cO{X9o&_sTXGJTtgX@?2_2ltTO(7<i%lNMmUP
zoYGf7@(B##I};z(z!Fwo)l_?jC`jG;XgYnh)HY8^=wny^-qrKCXEySVwM3l+Inr5R
zhcoH;1y}^RO~x1=#H6N|m%G<4>V3~*^jxS^(7iowcYbS*@Qgo2SxqfF;+e&D(aNC=
zQa+6*)@^|TgVL>4h_6QOb~-z&v`XDTO54KJn52K?#vn|$!TYUrBcJ1yUo;6of*Zoo
z!xRQH0(aCOgs^$ssre02e{<$rFWc(fVyaJfWo<Ir;)X=ljBQ=>^D)rHMA<DtWcaY0
z(V5t-yNkuns9BhaL!_q_3395^>Vb3dEaGyqqHK(`A~py=Yk32*ll6T9s)5I6%!_Wy
zVhPBuT^w;5nsYN@5uai>WSp2v1mQo>1)T%(v$1DU>L_n4imEb<5_A1hWQIKJi!7)o
zJy%>ZVn;oAHuzD#(-ijGF>On+%}bzb^>4eLe;v&52TK3i>2MxWyll7zI=E@{)R4Qg
z=tW1btk0V~%0i;rwt=nDj_NDU%}`~_>Rn0VOEwhxVW%Dgi&>HiPxQ`3UZ1t$gb*RF
z;Ac$<7bVw4eSLfmOG-(NZB3`bsK+n3@gJzT+kt0lRL5)Yed)cyD#pqVQ}YWlaW1kD
zW1-xeFL-7eKZ}}%ex6>k6V^Nh-Qs!h5b?XeKXDRd!JIv?jch_I@K|>=);4J6Q!aRp
zeIc9=yuIng`6n2{xV^&`s=^j*Ttg5UtEh|7qOA8i+aGaI-(DAnn^>2gWEj57$QoQe
zwq_bt#jDxD(o(BC+MzSXa<ZmGKAPvedZq4KahtwEV%!j~dh^&FwOl4Uc<EI(LdGV~
zN`G6#>o6Lx=1fOx)tFTd)c2LqB1^FlRm7j*U_r4uMh4Sv90}@OK6Vx^v~S#Pm@)C*
zMuk+Qv7{pcj%S?bWWOE_9MSCEowUjI97hfm-SALVkm`{HWx<4D5}~`mqp4A+%K^~y
zhVt?{9XxLZhh3|2{Sr+|n=p%`nP@NbX@p3q3Tz7^F%cQGfQv<At+m(wYag4yf9Y>_
zO~&~&#=GkN!T4kP=!WSQ@v?EH%6fyV=s?|UEMxxp-e^aSWxl66TP-P8S+ozuQa0ln
ziaF}|JiD^WFUj3DR$Xa(9tW%P5+#ZHN>>G*n2?}~?0xEvn>vI%*rq+GE({l~%UiOm
z^}gW7cxUTtuWxkU7y~@;{AEZvm`W}-YKSIFPVERpQUF&nf75Hnf_#*<(ZO@xf9oU{
z&&thY-E-wq>V<sKqX=Q62$aqV?68Ez?sg<L!Z%qQ+VH?cjRmx!6BthSLR;Kk71@QK
z7SQATx!djtQnyvY@u7QW3de^Le7cP7&#{TID)fNYUtN!t?S<yInenO{EW#94-6>~a
zxP1hh%stz$M|1|1?pn&!(Fg#H6PocZPfEZig81>-g?kBAiXD4ZlKQifJuT6xBm@H)
zqQ9CX*9}`pn1*l9R^kkni?KED(K?pVjd$honZ?ChIRMke4V2`*Hi6sNc<YOs@30q^
zHn6*%)WGWwt*=f|DU(`tO2Ra*C^xml(NOc&Itq3A<dtcpArmk}_|H<<%y1BN1#06Y
z)8j*6uz;M2Ww*B?YwvD;l<wU7X{AT$Bt1S<e<{fz^=_7Yjcc^r`lR;b*S^RyGulFJ
zAv3-PcS9`iGuykHO09TDZyQ2K&2c!51Eeh-iv2D1spo|r>u8cUooSy`z*^ZwY>9gn
z2=1<&W?%WnVi0Km$@%jWuEF2Jd90#a6_CLfk|7`&i49nI6IVu-yj*{8T5Keh4_GXr
z3&26Mm^O8-(}lV!p_PuZ2fvHZ|9l;~YAoQEKZAJaRmYy6E86vje}$2`^2ruVw83ys
z<s=i{dQk5~XGocQPL*^5Onu`@cq4DH>ihSpLUhI){NUbXXp`zG>6W^^o1(Los;w-_
zBrq5{-0u2)y(c%uJ@ctZR5~vg0P~U8Qbif3qlwc*`K5av$V)YZOnRa-(;RJ`zE}2$
zT*L%m_!Smmnuuv6(22BqMEDd&V$t-M`{kXN_*o^oQ-hGaGYf+QBSjX@y#eb8o>B*r
zxytYt^nUG`S)7svzJcAVE(fQ?j2DVQ1)A9GJ!%Rc#$4<&6(59odwsd_Ur+5{H)gcb
z-}A%!=G&o5^kICE>Qyxnr(@AW9KG)xbyiq}u<UFmIt)%4oaD4hEaN$wCsKXD%)QOr
z-d&u#m^FEx0=q<!UYq~fqP1Cbw8_k0Q&IoJN3R(ZH^(N~`SNGNJ?CDZzJ9rdAdu5y
z!YI2k!%P^B_pbz?W{01d_G(8ew{>Sy@B*=P983%i5A|+KrSIG~F$@GJ%zTM{`sd-}
z7e83Fxz@G^Fbc7T)?Ze&@d`oa+f6=En3OT>bpC_9-;}u);6M7}H(;bc%FBm#|I{50
zlV!do;RAaLKeq0u@=S;>n?P__>e``LV(fbIIYR5`!Ti?S<DgAZ1}14LpFkdf2F|56
zS7W~T5B3SnXka~rsY)gt`N9-zPkKB$?z~B%VRE^HewM%&*9G^r_j?7YpwDzf;t&34
zh0t3m_lLUhE-H(Umjy*8iCQ929m#wP64J=Q-Q5q3jaP2-E#Lq&8_oH&E-*1|f|Orn
zqCRuQfXtWmG&w>YSEqsh%PM{86_q3L4`QDrw)IbH;rP1)W9frE_7{dT`djJP*Rw_a
zW_1|d(7tMV(s;NwJTS<E@9mQ>&$Hwk48v}EJQ1t^1k8;;7_MA_EQBm`b0O%mep<Uj
zepH*f+dNA)!!<vUK8!K7u^&`-m&Ulq_7}Ez9v$g&fsM~)FV#Ne3mZW93kp4E4dnTA
zOAw1$$K{rxCIx6M5pfLcwgqR&%M+~=uO7hRAGyukY9KoYaiB^YHhaGQ0|+v`fRf^A
zG3ns>tfgo2OHQuTk1zcxsiSt8Bel%<y>Bpel9XWX<x6yMOLF?=+F|UQ^j5+ciNv+1
zm>(E>1x!D0SOAB&{mxZDLzxYT@^Wgms~oe0S0i<=cAjWF$D$BM05^spk!O$)_cE**
z6ZeaPr8@Sw)6bG?G;m(6So^jURkH9yE{fQ}JJC~!(wplYE?cuWeA0Zi0VH^UI%yW+
z(d+dRnDIn`8kbP@X@9C1ioG6@%o(mMGLjmCoSmJ2+S~WLD$5x$J{7w_H0Eeg9ySPj
z$9k(ym=<6~+9jqjke^czAT|-&bA}&SlR)!UH9;Bx+8(Rc{F7OfaQEMpj0aS*5PX9r
zc?<4ENFS*2tYQoP-FJVE>5)JTgvP47{`gGpV?Ccy3a<lx81^XCyumep(c}|dgJ#G3
zu=_|0{`s-XYZ)9bK5mW{JaqW^08sjpwt9dMT}|AcZ&6(11%p+-?Q+)i$iOd$-emrX
z=ge)`?j`vdiCZgc+ul+rFAbldh?ZKn$FUdMbg4Qy-Mn+)Gzn1g!+&0{NEEhzLH_)~
zDQDh=+pN2@b1}#Y>eAT0?E!zep21=dHi|7(UXc!YqjE-5#p!nZJJzMIZ7W1nU@u%>
zT}B$TI@|sPluF2gY#N~?CW<FXh&Hqp{u|MQ_RAm-pM{rpMr%3d>Qu*MOubjftG2Hr
z!<xV4##8!b4&OTCUZ{wb*kg_PWAbS@q;d~sFds8L)Ldh23LqMTRcDG_c|?%Ps?N@V
zP~f3J4M80rxe67-{}}>Ra`m#7+FQvtHxjfZH(gwE2Mcmn8Wu3H+@+WH3z{Mvvq!LV
zC^o~>QT-<39fltjm-%cxkX|y438NTzEc}D{RNej6e~#)|hyI0(emM?U{pm>hR+9=Z
z$!UcxNEE9q>3IF^<QH}kRc}74N{(HbtKvIgncdk~C|8<mGhwC>0egCfO?!9R_%Z(=
zjGIlobj|ho15+SW^Wv~pwcr>De%%`pGM<>d)QFT;@OKArIaoE%3Phs(0t*IKl>Akm
zd3z}#dGU_R?rsFLCjMmYbA!K;&+`X(uR7)AIzrRuKwi=>5JPv6_y|HxQ_Jn`?35W#
z$SqZfZVbCBh#M6#uoO=58*tqG@iq0td^U|yA9*Pa7H+FV@ntX-I%;EO;=UdTJidE!
zb(p1t%XRrXvz8No@SDSUfc?h{15!){{yQ&wF!dPxAs0(<O$iHhtHRQ}Iwh4v*%t7%
z8=`8b|A?@&H%9VrGFl0&0Cmt<#nzv^h`02$5&mPh!qK^(Fr_-WFOgnfvF$?8H4q?F
z)$dF2mAOkpQ<X<`GM#`vQgAzD03%6Jej4YY`qyc86sP`j54}a%!c~s|jT|MC2Fl!<
zN*cK^fDG)(SC_f@0n*@F<h!H}+%S7RZq{Yco}Z-(f!1$M+1K~)5$<2L?92`Sf2}eU
zb>4=b;6j7AGU^@gk=-GojoT`xAy{Q`$CO*qUtT@(KZ!S=;Iw$7#>Tr)U6Ag0>INI_
zvf45WM=biJ!0J@vqIFV<9ej6v;M;v&`QZ{~Bx*PA#Ui5Vs#C1sWyP5@P5u#CG{#75
z&e!~c>N*I|{!TRIke<+(u-_wcnk9J|(vqvFq;wP%H`^&p3Cx1t*fEfA1s#(048O-<
zEC6wt%A&Grt`x)F7zC00ja<4L9aq)BhQG&?pODCGP|*?RWXk@kWZ3RvbM+G$f<~p!
zD6bvIdV?wS81I9D4Vjthpz0Ed)u%j^hhX9F{Go_fcY;rR)x$lwszx@#-k$eK0!*UD
zFc8pX_X6hr@`0%x>AcgGm!~zq-asBE1e7b-K5#zhG&WmvQ%9jafNN_w6d_7HpyqUD
zRK3xH+Yq2&x8!c!7Ep}vl#}(e2heoElvRW%;e#^RB+{m|@LLOEkYZO!F}x5PDo|_1
z`Tp`hPt#?lf1jq4XKvily!tC-jUy60>8A!M+3Za29*I@Bknpe?D#?0}q4IPNN6hs;
z2ra->U?|zmDRPnW{#Mf~y~nrK`@0s_RFp;A3?fH)ug2_Yea(><+4|_C9$E>U4yvW_
zZ=pcypdh1I`&Fx}RQh(XtAeXUWP;>IQsr<4=wDH4k$*6`QSiqWaS2Y#MmyHKFX|qa
zId0bZL`yiI7T&60AneN0(YTrUd1V{Di!MdZpE5t6WY!b-oDS0Ft-bMp*ZQWzTtvJD
zY}USip;V`!SER+K103p(<VB!_f|Qk<7jh+tnkEt7>k)I-$Kaxzyid5tO6nqCy&ws%
zO*P{Uo^j0*ey?vH+wsuW8p?~qj6$($p!0akX)c`v6vun5C0l~n7?(_W%qXJasVFOP
z%dS#){)+f!qREai6B0u;W)hZfqqF%Cn1H=XR+PGMVK82IbzL65A)WP=3EynVs1MMQ
zyw1J_BF4}aw)A4Sz`1S1hjurWMguJBKYK}3<ssB-*GmPZMkC!P@C_5B-Sw$7X8L*<
z06eS3tIyZ92d0hwb4~QWO@1?j{;D$hNDf}`iTlmYTPF%ht}WjVqlF}|9=ZR~_9~6F
z+$BV!{S>%tg1@9mHPpl7hQBGb;GzU*Q|j3qeWOg_JAz_gA6QzEmqSgotFGX3Z&|gF
zXtNP<&z_hb)x4)^vs{wx6`E@2cHRk4*fmDPGj4bpFNSYc{aR7f*LQ9->bt&0WqCLJ
zYr`1RWMxuy+{wb($aCBCK3wi7`Y0U_xvdP~ORr28_hgMi9Gb4WY(I<#&XWswZS=6I
z4!r9sgu97SSS|P;7JVep$Uf2NZ>fshKx7)n{pMx2rx!h~X1SHqA6zOuNoePdAPBEi
zI6`TR%s8N2EWPNBb}^M?XW8c9LZJl)Gxw~m%oF08OzQ{1pb#>F5H>6PJ~sPw1s3jH
z=ZlT-q3U~pkR<;eKzl^glf%!?N4f)?rOHxy57>7Cn$`EyZMgCuNJg`vn=QE6BFJWi
zBpklXIcCW2l_@{&ZjE`vpsw;pcG7)eB_(A&S%(N8&iCk}=62}hIwot7#?-hYUY`uE
zY^XOw`m&?cE^i0y+%~R>WB%t%rXEuRvQ@#o7%lYsI8)9M&`s?yMu8o{bX{Cm!t2Hd
z(=%?X$w6N^pu&BnJP%QIa?qm68vSn)I*HB#Ik2aMzEnWdc^tB&Zb`1WvSo#-D79$#
z?oXRa;~uTb+<A6+YrQ0HnT(ZxXJZ0;J}uLBpCkCJ_6pzcon6@yUoF?zhWIP(aVS3a
zgo=_v=G0rG+QW{&W`j0Q|CROzOg;*#3EUtg#dDi6-xk3kEiHY(lUh-3ebFbl#gj&0
zE>Ugl$eIZ7UsJRN+C!MpbjIj?(+K<ITXvD|e)eE4YOc6<QE|3PRj~v}t*bC2aas%;
z#5RX<5u;$3C;6#2{SIxjPuL>+Hhj6yV3{+=Al4Kt-{&wbS=)?}d2uTblLOD7l6QF_
z8&kp6UXEC>Jac+62e{85*+6Sqin7ZgNA)<BdlOrz-Izf;XjVOSJg`AH4a^iqIvvXv
z+#q9kV&zeK=C^Ou%&A<Hl%-)j8lk7A=IUx_^O62A=A_p7ts@Sgm~y$F%%n|8a7THn
z39hk<(pX_(QZ=Q&D(wnZm$z&DCmkI@Cewe+)7m%x-7>1)RR`n0bw5jTUhxLk-tv^G
zpcZ%4QDgEU|E)>!9fQ(Q6YZ5N+Izi^UbMXg#gVO>(42MbOR#c50hTW%u6AeR&f)#S
zeZ&J^HWrsvc2srgtk^K&$AxNJ)ZxqQ4Hwp_=$4YZbg;Cznl=!XVQQl0_^RU~@^Fix
z;L{k!?OriE&}^oPKJ_DrlaJmW8JTJU!gF(g0@(vs!sEYs`hDy>t_<}2_ATkWD)gqm
z@wr#6z3vU9i%hsDRoM*Q-T*|!?j$??yc6|LO`5J?CZ35NKdrm5PVr}e!%~H4#yPMv
z0j8I+2qNGuO_-I2nTkFyL}yLNEQ1NU-n3lG3YeP)<E}m+X#CWNVd_ofYYZ*N@&_mx
zBrw60h*#Z#B@?*a?0c{}J5rO)rY=!yZQLGC!ayUSQ*-kn4=}B+>gR}fAGfA|WhI5g
zVk+nI-C-!-ystluQu>nL9`%MS1H-l+IkI@a1-F%sO!H7Gh8fqmmerG5*ilph_kXt+
z`T_rLFOuJeUuHv}%-B{1F7ShdR0XFg&g|4IPsThI<>X_ZxO(p{^Euj|gU~);8|KcU
z`p&fA{r6gxx94?4yL*TsLUHrKr3;RKvBu;6YLXae_>Q@>S7veL&T0>NS-Z{nleLIJ
ze-}&)FBk}r=9qm?gVy$}g^;u*tpdN2HX1%X2!A~@Kdt(xTwiC6R(3V-O@vK4F1t=I
zlSOo^_C)=mZE&)jcUZrV{0_>Bl@@Xf+2tvG{=JH+7Elda*Ok!6v)8RDlOi<0=Z`_v
z-M~uxxhynpJqaR}2u$UUOFEw2cekbf<ms%gbNCka<Z+;!dOa6hN*4I9+=8%S=xP(!
z10;URmJ>u)6DNrB4+zXvKFz#AKT{-TJhpOhY0>0-8kn4wCPiup@<{xtB%S#Vk)7?s
zI3V)4S`Jg2zZyO_p9#A9b7G{7EyhS>PjfZ!T7f(ptUeMs`^eMPW`|0IA-%F**HskL
zO-IndmbDsXFXI85pMjj#dXo8IO`|5btNYj&NptzKmGoJ$rq)q50YuCF$BriJ-y84y
zQ2vZ<WpdIZ@=S=o$3};tc*sH_l`=YeS}=X%9E2@4XpRd71q;WFUS@|kM@mCj00^g1
z$R^oX8KKR>&;G5>+$nB+R7R`7hK@R4`i%ShZEw-vY6#WgxXwrCPnX))l8MseOZJ%M
zr4PBP=SMQV@+6*eEw7!}P}p^~hB7e!SkBb`d<_vtqEw4jLM#f4k#0OS1nHN1p@P0r
zBOPVW8urR4W#7UzMSjK1+*}{NuJO)*P1d!j*rKjq#EwV`*|_SS)DzSI>6U)k>s7ew
zV9?5vNn;eNzRJcuSVTIX02ANawyjqE!IDW=RU)bY@5#((LR;Hnl~+w~O)-$~dtfz=
zmgr+8=C_?Qa#3K%39m#Fy{WXoV`TnZ!)Xni!5+Y#0;14pbAs@W^)wp%kIY)N4$=iM
z%r6XvgqRrN&xB$65e`R#2e`fHaVljg40m?g7Nst?*>>H0j@4L^(!kp)G{7eHQMY@m
z%XHHer1Ejul{+yX8to;DrKBBRI1CRZt@gaekf2eQ`2q0^1;PvfRT$LDn?mXozR-!#
z>#g{)t><F+v#5Wf`X>VZwX2ys_U=){3`TTEH7Pz5+={6EI?pB-Q;$O3EbP1f8*3hY
z8hE9Wmb2uL1AHGIlR?z<Sa`8lE1!JN#PQI|qw~#fUviRz=|Ghga<1_8&m>`y1=o)O
zTLU<+J5%*@?b6t_l0Pwc>D^qbqbUmxjU_%8zAxGz>(Th#*JttC-cMthGtIGnv!)NI
zlR6`#CmxkQF3~mk<3FhOs@e$xuZ}!<Xamz{MD7kcz2qvpHyuPstNeD9S;D^FAVPlU
z_#f3((`t8KX=#V1yYgdg#e%eo3=@SPFufTT;SI)?O5KT=If1SIR@^;4qN+3+*v$+n
zn_2aGgvBFT?c}=cz^F`B^_zebK198xqg54ctH&O}iLOvixf()nZ%Pat^dJ&*njy$W
zreZHmZ0Zm+r(;SQvN#ojPrI>DaxsZSc?aUl)t{EM;2d^dfhEoWqBYZWf%x5LmqOT_
zB<*S(`J`g0Tj%r~%Z3IU)tm0Nkee5J&Yczmm@{z%_p&Lr<0I0zaJ!*|Pi{zwJ2d{J
zYJmE{x?(Aa@oEyJeXw+w^u=<#OZcx|GW_1Zug;FQ;d7kmn5bfKb3P%Y7?PJu3e7OH
zGdR}n=0P9lTYYrQyDMTegO8(>@VG4X)JTVKO#@l-ykZM!axr(Yaqo0ReO{b$dTIT9
z`kd1%TGiwL*HUAlX@@FnK%AnaLEw`DW1+*L{Kl7Xt^+$~B2gpI+W25W^C2#kRls$w
z(U~z46!EIxL@gM@NgazYKXt~L&7OgNvJd(Ey2#Tq0-Nea7|?jRI9t@k+7DFao?GBE
z4a(LJI`-px>K95jDZroft18MB!0?bYSs6mVQ=~nf#v4@rgVlwUMcbaOaN*V9W@R9~
z2$wvsyDC+i7vYziah~11&=+a2Cbg44x%$@>gcf3i5GKDd>R*S9NJCe6Uc(?;)>h<s
z>Abw=vJ{R$=Sw35FgrlDjeQ6i+X0bizX?!FapB7p(X~VmfdWF%wBY*O``xZm{bK3W
zJev$s5kPA-@Uum;7^bn?!FznBx04aF7>r>l()KC21nsE>_eR3^`~_72dL@QCp!Qy7
zal4h^`W?_00H(M43jBTaFvowLyvy(YeexLJJxZ7{xw|1%8{3R`QdlY3a&Ef#CPz1O
z|Kiz=B+YKu8w_Ltg6_&RHrN0qTGw<X@WO{T-$3Nu%XiL@w8s-<FA4v}i9;{bi{s%p
zIh+OKKU}+ARyK;*=a%p9Ns)Ng>NQewwFMTAQle<5^_{TTcx?4Y4ISEhhv&KLHnC;A
zH*!X*XPUq&b2?6fbn#m$%zVB7%D_<3QSee#6?238-qk+=_m4P@F&(S5%_2;d6}Tl%
zWfZIOc?zXo)2Om0GNBFwC5q}YOC&mXg13Xy(J$Ie*!~>@83;(>F7)d;)%Q{htrpAQ
zXy61_Ky{0AFJ!I+Iwg&bbt|~gc*V9fcqsckIk2Q<DCK)TMu!H-K+=;i=+?GvWlw;_
zM#WiprH=a4dfTLX$t+Obz};EO+&8zcM`A@Ck+PyVLw&T4FE&oGI&KYU-VD86yBas_
zb-L_{WqX-Zjg5S^Iq^En&e$@=oJtFw<trj50w4N_3IUc$4p76${kHFf%O3kDy*@)W
z?I5w{UAW`BA&~Y661Xkln_`f;$^h9yT1wY>-+*&4{sjD~6HX$yi%~^Zz=Km46|ltM
z>Ro#bPMm=trS!X9f0)5PlWH~bd(&R<<#i5}=>@EPT~<CLW>O2UYO_}Q@p8dc;RpjI
znP4N67Q+oIVu<S7g`+t2#;7veu`xGi>pRj*FYayrd^^|(nJ|7$$lM5dk8o(1m$Wh%
zV3~Nz=3F>ib#;Uw!Tp+$wOCO<lJo{{2*xNjUzItj9K3ljfg4>%R90CnKJx09>Tf{$
z83IDJLaXaUZ+C6wi5N>_{wPJ+RA|o_v<Oc%{;WS01OlX^%mbQupIwX)bt=xk$H!h4
zyH>|sA>)g@Zw>aO3Q*dVu6Yj3oBe&2cyRr!q`pWwFp+$iq7XboOr2S>Te9;mw^yUK
z*@qFOU8L=ivx|$-3_vgMi@YvY8ecE>^vBZuoS$3lOUftk8~`w^#?L%>cHg`_1@fOQ
z2JmO({}-+9z5>4`s{TIdR?{z!C{053Y>^>P-tDDQo&n~c<x%C-mhfwBD$B{I8eD6M
znD<7bVneM}FH|hW9Yq^m2_i<TQk=W6hZe%1FNi}W+apW5JnN*6AYE!+`aZYbX~B6#
zDAo^ym-cW@4Oh#Nc8g`+UnU;Y?e?0^s4SEm{~6Ivy%<*io&k{vO4Gf|*s2dgfMmlI
zs}q7<s}VP0yA>Xg3ltv50+pZI4uYA+Stk;yzRLNU&|Kw$BN{?PT3XH1V3pNL8_Ch<
z9|Rq9{IRNH#%30l&#SuvWc5J##c2`Nj37K2gTZ15_DlY;3k&=;Fo0OF`{h;v%@F;!
z*}nT}z?R;eI*IqIH+v$D4gO^bS8`P7wS}+700n&=RbO>^K}+-vGQGOEy0o;wQ>`s=
zN(W;`!n;4TfWwjAGnssRssLTu_5jz^vgD)4rC0h(MiBU>P0;ldAeL~fl}Z2-P3kGk
zKcB+?^nKwsKG^|@at@f8gr#5RJ}e~f1Oy~-C)2n_E_zFx-8gsShO_ui@<DacJq@Ti
zRpJ(c=&6J^R)F>JD!58hurPu-mRjQ46nytwd+@rH7(UOp<r-Hu>~q)~UO?*9U5(jR
zU5>7}kbmu5!_}k3MhRaU6tRpE?^7#y!!#~9O8WzR###oty$$HZUE6JgXUR!`6z-cR
zMlB5s(0{9=h>h~MGL?iZ57FCx1eJFiEqbXO>C~ZhWUU9Z?N$JTN%dO6>rkOtO?_;0
zCBB_P&Z1hGVg{g89~y7$TL2`u->GxVFEYR2DYs-bUD$%xUG8Z8iJyn+VX8@SPRFx1
ztskOu#V<tL*o735KXl{m`QrXyTq3b?s)EUiJz4@Of6Xk;%)42sjrVgK^YFsnl$Mli
zzA4)DNYK7KpUSN<d<85cVus|RM@__QO5>a53~EY^&IMe8zQOD?>HPP4!{F~Q!Gq1O
zGlSw+uy20P{MTbmmEwMM=;H3g(FlJh7sAnCXxi@|!2>mx%<(FqJlK5y-|`T5L04HJ
zgGIxLv!-a59fiI}C>7@HQLSqiS(~y`$e4ge!A;8f^&@WG^fD0)&?NfILw5=ykzvQh
zKQ!g<+`ZzW``bL_gFpb&8|`*8=&EcE0YXjxGD-=pJa|*82G<QX{4>^6brOQAG*!ap
zDWZE*nt1bBLT=nUc-$(b)SfaeOar%kpELdO1Lz52hPgZXua-U#vcmQzw~XT7>zQ1|
z%Y+|l{V9w2Ly1J|v?4bQT9c9}@16+_4!IohfvGQno_Pwg7NAU)k03xc60HWD2Q%P?
zi=x}x*o#k6b!|p7ANtH=G_$Xd3jd0lS(=$$92x+-r3!aymSLi_hA&bNuD<9DR<>+F
z*NO+~I+LsYBy|o?QR8Jkuu)0c!%(>~t|oS92Z63H|LBST_4yamYBgTtTxrhQ4hWP4
zUZHOm9_-|%(9_zya;*tV@7@Cf)BA!&n;#JEEShfhSu$$EA1Y}5TVyAoP4t_Z5PGEL
zo8q=51P-N7J!;`b)mf!EMvWr4m1%1>bQ8{I+#cfohLHX5Y|hw5ivg!ktkB5fFWe1=
zVlRRZu>!70It592X9`x~wmAGSMRaQ_n-%~ef094J8RHie^;*f<{&p0=rYBj7MJc|P
zqrXbP&J^U*tU03{1s%obv>JZkf{Mhtlw{RsV@v?(=9_0S9j5^Z60%7Cb>l%p;4Lvf
zwEiRw5YhlQgM62EF!uOjzhS_?bq29A9R5ekxet6=JT`U`OPKL3Z+Ms`g53m3#J<6T
zyz<u+st!i_`|OLC4zml3^Ib0%CIO{Ea}e~aTSu1e+v3f00KZTFBwlG`imsynU-<R^
z0y7zj_hUwBMuUcDB)S<qU*<*?=H2nMt|NA#m9oAqy;T0aV!;+CDnyO(UAi>%O~0n}
zK1SDwMM1)U<D00-G7E$0bv_k#=N9~!?tmFX;BidjoZ8_XcI5po4g8>0VX{daTa94~
zBFu;}9`pXWyL^D@pJJnwuMOJu;Stq#&@I^-m^u}>Pw=$#dy6M4Q{uLWTR`Lcs~Ero
z>zZhx*=?eV(1YsayHhb%3X=q+R_tf#bpp40SUrN>Lr5fn&`iANE;$nw;jSc@uW^{G
z#%SwfL^IA-V|-mXef>SC3E+{L?i=(6zI)2{*3=jKM4|;RIAE<}4(~4pu(Yoqa9*KS
z{5B&v_^g?U(zx`NoUMR)2Kbo3C8iAKDoxnJCh=OJuW7t18=T_iZKH=IOyy1SAzbVR
z%3k(XNMUw7JUvDV?(C$uu_34&uA#ykTv}AV%eaHChJDPn%2y)bQrn)uf4~3DXDq#q
z5A5HV#4dgjJEPwKs6b4S=otNinO5Q8ie6T4$Pc#z|C3O*2glzRrGX({)NCblRBq2j
z<qHeeNvxPPLO@8wiqO@}*M&dv_!>6&{>Ar)W!PH?L6~w+9`P;NErQ-W%i`PY@CaqN
z<$cT%E|Jeh=dcY9!Ls7ogugWU)$s;1&Nr~~;;PXxX2eIPaTC1s{RO&h`1#pnvish}
z&1Q~fhWlSmI;Wp-j%s+$`9!`#BSg&*mUCwyy<(6jC$swv!a6g?Wf5i}Z6S@2la`no
zX^(H2YQP&2i_rzvH|)Ebiz;QaPUrNESaOFx_TN$Ge+|j2k7&MmLvR?qP(GZx{N`g&
zyZR^g=B2>o_e`!I@A991Yqh&ds5s8b<wh?@S;3gRJJuc22*alhJe<E1e;v-N>{rE0
zPpo&X5T>TJ`R565>!>x@T)*e8Vemu!`{tP=E8TfvDD*Jmqq7>`x&zktz%GA-a)g{m
zaehH8UZyCK1ee#@kcI=4IB!2^l04%E1@As#l0<-;6t<rKzUy)vzo8I6i!H}D?gdKH
z5re|;!cj~Sc%2!2FQvG-7gMS|UT=H`R#AE++-v&A-0G@FViq4;u;0(p%Rh6CXRcsr
z1}fEl;xtXMKe-ifymciv>f3XYJu_<I<JJAybBF%pmz$CN*-(7uV>nrOPX5|CdMoeX
znm5%IFDSIKq$N@<EO^`l-`P*D>F@d)=w9cFA>)Up2vmvfpD9x92$$5{otB%F!bNC;
zz2PXrt45H}Ctxijzd;xXA}F+1;9s?>Roz*YTCh2@m8jig`q|<>$CmXC5}Uh=y)r9y
zmXmyaFf}Y*Wf;=aowv7D5?V0@w;mt_7H`}%k8{^T;6#?!PgX4&<MJs>$0M+C;`;Uc
z)2|nPAU6^9ukAV$2#(*X#t;L9--NPa^0gD4T|`dvhZ!OsHKj}_o#nabCWQ_}PMS+o
zmHTf>zx%?L;M6*S@)j1ccnnTW;wLQ_Z|G?GU_-rhnC$x9=+4_be6w8L4tN?OuN!Xd
zI*j^OM)+lgUBa2Z|2{rxZDYzsFjyp+mN|`WtRrleo~Bg3THkHYtUT$uSb=O9pbwg{
zBEF@QDvpsWb{>cR^kc*m%m47+VcoxUsj13YK;k6+UvK;;u2AsxGf_d4LoOOKCTu&2
ze80rohI;q6K4v$0uXQZhghIXb(q}5KauI2}x%;iBYeJSbMDgEg4g`e5qG9_ktn8#0
z9m>$M@DthWZ8+d;R$4YENM^wn3anni#a@wt3c@$5LYfWtca%+9Zg#M_+dDr^L58)(
z*GDGe40z|ZRZKb75}QkdUhA}Dj^n*D32LSHPT6%1Hke&JvJ+!kIf00CqfRqS)5qmk
z!z#JGI<VNa75=FXL-e|o3#U=)rdIV-uQ`|9Hv`m7-kFMK*eO(p+J*ne*1Ld1owxtv
zZ3i7h$s@aqY)HE~bwVMwO{t_d#Sk--6k8LMG=v#b*-k7LAtYN;G5C-ejF};6oCYaA
zPBY_tm@x)pj?B#O-9FFm^X&Kk@4CuWS8{c`-}n1<zh1BVelJw_J2T23Mm*gREu19>
z*TD^y$2Ju&c&?=7t>PHXT3kSuje;+yJg2uTi}Px6%bIk8P-pVQi|nwA7MGhMUr{UG
zRNj<{Q_UD{`ijv&-M}{F)LS1p#Z*7l(xu$v^voZASY(Hu<CeXXqOG&vm11R*T}KD?
z*lAG87T6=(O5x??Ol(^5pE%!orX^p#>{j*z{8wL}JhTNh{tEqrb@HX&VVX)TP&Q54
za-xs-Wyd`K9OXCfJV@z942;9Zv#d@`=B)cwXN>B0jr1I;*-(vG{2+$eR;=krc+895
z+1zgHO0`ac;%}0S(Rv%r!fD{~ZS}IKzT>EKda1DI_R;*d*S{5bxwGo3t$htA`x?n!
zGuEZWPP&%j%*pQi^x$~-K0TsG@7N)Z?dcNDYyMG|#Nxy-g85Zlyy9C3J;zLs$vyh*
zOzB!YvLlPmSh{)D5#cysjiDZ6xkGXIvh!WGYw*lI?8e!!`?YqTyPpL~Gi{{ST4Zmj
zh~ZmF**aac7OyCPC>kTIC47naYC1tdpv!&mX1F(J(_pwem{lDyKYuqP*VYpG1X;76
z7^oZ<ts?ZPHrOnY@rgQ|>y`axTv1|Fk)*q!rgt*}uUJgCu$0})v00A(D~bR7xFK+o
zRKJbP)l)km|M}6S-uUfGvPzlwLd!7u=+n(ffio_H{le)LN<)|6@L|_MN;X6xa0nII
zU3J*GAZ+ydfkOoH&4lKW^Jmsqi@K|=&r=5Vj?l!ih(TVUciA8QpW#<^5usI%0meTY
z=<v|tX#3=P%lQMJzu6<$|BxSLt5qBqNx0?nD_v~UE!estIOY!+QjG5^p$I_w-bE9f
z(q0b{HS`ufXpsM<f7g{Rj)*@}`SsNi*7x)_OSK8Dl86M2yQ2Myh~TVMoSAF38wi9^
zjHB-)365lnY}kC-%eS8~Pd^zf3%6R>T!#%G)H9cILuEU#;j>wgDC|HO$H{UNe1MFc
z7BN`+qi)M0x!8vReaT&(p}2_4Q_V$Yh=lg14s%Tzxj9b+FPf?sqqc{UJ^ZkqvAESp
zzJ?*Zs@~%(!FO@AthIRFbN1>@(mun}M*j%%;OV)a@6xPGp@hN<25i2<g%}jJv>L5A
zbSydWtiI~ygX#Vys5VwmQNuRYmpQ(!CBysmlviwc<sXX_rH$%z2TitAk?hoF%o*;Q
z7h19lWARL#irE`U&}b6G-oS<v#>!K7MG%7Ax40f*24jXd-*D@|93jMXzztQ0FY*$q
z7|}A=rE1*xW}(sA=SW={w~LK$f{UdeEY-x%f)CyQQelUY3QTnudl>HjT9(;0`2FrA
zjLT|T_WM;-dG~#a-4Y))Ci;klgDzLu`iiaL$G^b$Yi(xu>mpljL+G(0<g7(^l<fzq
z<(i5zrp*ivy;x{8Yu(Z&r~J6V-o=R79XTP+_6f8^?~@3H+Y=1wI1#rfwIRwvvfF>A
zpJX}9ClCAJ3iZ__J_0v#NCO!TnF=vit-V;M`@b1mYpf&pd`~M5ZhhXRTBQ4a6EW6%
zN)I|9vD(H&C3_J_(ZXw4oo=F?6V(1CF(vI1dG4@bPuaZ)=3==a(xN%%UmkSSmgV8}
zNH}-ryfv$s*!H0jh9m*`l%LR;r&wdOL44TR9U;k0XNvCCytOlR$sD=9^P42|&aYm_
zUB|s8aa>9UFuD;z3qB5(38^d4q4AZetjf2oBuK-Ubv}b^Kkw1`V4>7Vv%>7%Oy8E<
zkmen|BxQ|RS!>pfh22m>R9pknKyvRRgc{+W6Nx5}B|QH@qj}L#G4+EZ3<)=bb!#|T
z0e0ieI3wmRdncDw!X}~HmqM=42lLlfGPQK&o0A~qY!@f4^sR?cTv2&fLqk-$z?Tvx
zm$%ANbWi3yp@v39*Ae=y?_j&D0d&gJ#YYv1H7UdPlm7XY7CGr(YOTQW=%D)lM%JP8
zbA8{Mst$MnXJ}{S>VhIVW+}1CHrJ0Si>iA!l8NX?k>%}XK||}+X?m!{Rh-4yK5Ll7
zZf}EcVu%(_9}DiBb*yxJfcKz{g32nf3vPn-Pb{eNlJidt?@igBsS$z0`D)rSF4^go
zI;A?!b{!~P=Se^g@^tY*lOTMF_mhg^ZsC3-2^ezg*o|=Uvi>pl^-(=Fio8-YX5`)h
zw$4IzW90*aqOW8F-d~8}O1lfCrjFydgT<ORnWaX3)&O5bAO|sUblG>BC$IjBNmz9o
z>%Ou^Ub!}ev>7k_1vU`YVGj?IutD#A>?UYd_(@(mfzQ!m-G<U$x0>$@?|e(Gso59K
z?`%j6JgKi*XqJ_Qwx~-o<pj0ngGGr~NW<*a-$;Y4XZSiIPgc4k<6ZxoKKhG?_^<vG
z${)Sw(7E6Oo7OkcthLl2M7^gqNv#M!Je9i;`4-<8%KDR=h$hEh+Prl4X|*++b<zzA
zAC7X09LS#46u{}1wiihYo~=MLe6{t9e#48Zv7S$Cm@iRtUAB?qUCLF3xj!p*r9E&c
zvoUh<WQ_H8zmCJFdJC=1s!iuL5Rt6i(1y#b(xn<r3}KY_)8Mr!#Qn;#Mq9LqIl?;n
z?c4`{`F=+=8?xmR6A>dkd|uBas}?~|a32o^UsC~o8g%wOwoQGJn~S8=F#ag&0_M4|
z_Ok6Zvu-hs2%@H<8e1~UGe&8zp$t~6!MnE5l=zL(&1SR)p(x7AN;>`#O1#M;mdv}>
z1~kCgwk_>1`#n3G+`WTs5i=H6C|c-<yw|5Jy$zYMKm77c=iPq!855(B*@^4>-viit
zkD(|n7#R|c{ZhG9?H+>cx`UNf^VflQujeDCz;9)piGo8*={A3*ZKYrh;j3k}C!J@O
zt#?t;&k%o^8f<Ey$t?4g2dq`1_P$N|{%2HESOv6+VRbZ9dGEn*y@6T)gsIOw3YYFI
zlCsIQ;T<E<O=Oo*%mtLh_>wVNS5dueABcQRQ$_M2rdHYLqHok$(>M2y7M{?(l8)!w
zK}cTtT%iWG!9pY-X{auaj~vlxp`D(dS-gbSo|f<2_oaK-_ykg@oyHokf6y{9@CH(5
z%WG|?D$9|E(l84tSA;T{<&FeDgr9_x=%$@r5|$AbG0H<k##x^?WNB6p0@w>ayONDb
zkL1JcmJ8g8$*lQENYQ~l3Z&}Y-lw!Lm)l`5B(qR5gtiZvo1W56Z_B1$7j!Qjr?9%U
zifx92Ny`TgdPXi<NV`|jVl2g)Z-yx&%h8L*%|!01zXJH>V_4Sx$d6=30i-09zf>x*
z)7eXNWO{nJo$BXD`12<Neb(&ks(&I4R)H04Dy_*j?>WNb<s6N`J2_)Bfl`;~H@Fc|
zK%MQWO>w+8#f=CnS>L|BXKa>d|Kc(~eeMv-TQ5DZ-+%Yk(u+w&wS+-Z9LkU<_aCPv
zgry>z%Nf4w^*_+$#-5w9`oDYE1uu>viI$0BB20rRw~JI~w=B5wn(c4+IBPV0L67o#
zk0VU0*j*R&YUC%R?s(rHWc$z7i~(!M&&>8J|LAgC{1X>?T{4!&CXM(QtY6IIxDWEu
z!{e@nm(-FUM7f>@QO3~2Z<G|Z^pv;B<<zk#U2J=M4wW!AUxnGTCmni{Q^h)IXQ7%1
zo^?a$O2y}9+y+F)ra~Z?t25iZiaIu;R0M|)`h*wa=`dJJC;MG>SmUo~Zuxriztd5^
zY<l=GC({FMaJAuh{zvncv{$M!j&kIop>jB-H*zsVOBC$kTE52<zZNeLfXj(B6n(fn
zm1j#C<|%g%+{CFG5ml*djc+dOpT|Mz`JbQke@YmwD%aAMIez*k<@7_7uF~Qud|Zhw
zde)lU5DV>}S<W#U6v1i?-$D|8R?nGYH|=3M-!(BUFLpAdpg$GPXHLR7aU#Qw<2CV7
zlo7P*+sJL?uBpf0!_b_(0j0+cV42az@!XkoBeX#`cOs*wVhustVB9qSz<52|e9r`>
z8!rk8<&B_-HESZ61YI#VRIWa8=u0Oy+lAjbU?5c^wI6Sa>`n@J#q!14U<WAH#PQ~;
zGDcgDLx|r{H#0S$tT%+nmeXD2hpTWc)Uj0@R1#L!<GGx3@EyA#6zO!ZzjDwgDxvnD
zH~t~RFE_s9aO@*@WA8Lp)HHD^QPt3B51j}kF^@g$;wcNVwWRFC_u*MY^{_v1QRvy*
zPQs5~b43mX@J(LvXS6-ME?|W-&2HM-4YG2qZ?yeHfZ`9o^0ApGR81Lkypq4)BS5=U
z+Z9;*75VVWlZrI3qUPcak)p(+he51NNb?%AD1wIUw@xbFvD{Zr*|!Q+wyrP+O>e}c
z-3b~c<(@C=Bh{mHrJ0XmNLu|GUR|r9yL81sTe|rm<Sla3*5<u@s2-`fgQYF-ifx}2
zs}q?c=c(j{o|?3%tIN>_98dE_aPc=>fD-Y9mA+NOB!O!s+l&<g#oblFeJ@!V`B#08
zQT%oB*r!kF_^APAT>AL9<r8{&w{Orr-juc0jHfZ$P}*H23AUb1%@n?QC}Kye)3P-1
z-r(i5|M^j3{}bmHxDM`ILK`fi?4<N1U1G;gQSYlaVe?VPo9g=BbH6%5qgc)2zK{f>
z$(1x*ApSjNXosqt^bju$&1-}mXA<zt4QF4fx1Kng)~pnHp?vj{>UkscX9G_hajt5F
zi-mow(0a1W*S2)S)kD}yz?2Z;eutea!CUW7^sE<z$k%#2@b8>*p`9-2<Bgmb;zQ%(
zB8Bi-)(~i)Jet+0rsL&tHAr>pvCCbyzlFD`)0Cw?)7FOGA@S%HZT{@{E?=@|<`Sp6
zmTWeII|w1?aJM0uyF65U2z`;Z5L}1gyxo>s{+5M8W7%FcxW(3i)++x!jKsEq8XI!Q
z)QzR^I8{`xK1*|xOdLz#-5EmN#v%v$&E`dP9#f6ByLM!0&$$r#Oy=-+x2^sse#>sT
zST7WLi8Rs)xk*;~83$&amK4O7k0RYW7w$j*tZWZ%2w&9Z1QDr0D!haIw!rf#OA*0-
zA1k=mhzP35e*nBw>oy&UI+6a&W?gv_ME5+F$vyMX`LgMvPbt8gSJLnsOmjPbKK|zD
zbaHS>XO;`y^0ReZ&3GlX!i=kNhuvtgAU_IQDyu!v)M`?q?Lv9TMj2Naq9q$|FWm|y
zM!1>i|N35BiCpZW3p^QWH8MLa;w=V_6i&i-$*jbnj{Xn{?gl0&xBi?W(y>PajwCGb
z;7CQ)3`2Gf+i(Y~C9lS?;B$eX99%2fKLg`glDo>|uXdz_Wxi-JYe59FYr?M2EVVXe
zmFYsQXi9a-n>R0X@p%lO;g`1h=nA_nq~djWLZSW9Q6E~^?~MqX5IWaWt`AkeT>Sq;
z^S}bXY}&q4dE+Hju9Bu3x+!N?uXbk8j0_)hR4X*_WT&}pvxFN6JO&^X>{uh2B4llp
zZ9ZLc-IW%ZZA-5-n=gk1ZIhE{-5-8BvBf8GY__kmW<WnBrs=`jfIaj!((~e4_yaYX
zNnUfy&6m_J8}wAQxRA#qLP{!mY0=0M45)a^85(-B>WPer%yM)pr0Dy_i~bTRXO8&M
zH%~7|yID!Y4%ZyJ;12*kY%Z<#K+*d!j=K%i`fu41Y3jlC6WI0mm{b%U_(URO8aUVL
zY&S%d#p<AixHogbTFZEgpt#EGmET-cQS)^1DWdMaNTIx$MVY|1wLRSzwJ@3CS~d|Z
z^SB||df?7LK@S<x({DCw!GYFAf~V=za@jkVSgo<pRsTUBy}3=&<^P+IlP(9YWGDHE
ztixkAoLRW!-EB8Zlqp_)pb2LZ%Nn}q+e#xjIh1xmEI{n?Lgc@sSNz`kXyQpm=&}x8
zai=;hqc~C+ci3s~5M)Ud8l0tnhcV`?DDpy~4APA;{@1_z=A(UYa>h>hFwJs+zm9-=
z&x4wbQM!VAt~As4<r6`_@iyOdo8=)QaJ3|;{C4o!b&+~?>q^$fh`^SCJE9PYz9ds4
z6qRwS;!}tg_KGnp0X4(Fh%}?+drLGfpvt5^D{ID<^{lM%zUkyU4>%_6HCeT=p?W>X
zZ#ics@j?R}y<w7fCeOA1tS7?Ll2c<RX==^TW=sSG6q~cEDAg4eonp*nUATwcU@Vqi
zOAan2gMltJng?TPF?H(*yYN)2PdJ_Ts&sxZSFHQ@Y3|t-tE$!qJl@22smtyO8!hDA
zTg?r$p&U)qb7zjnzrU~9=8`B_oG^x*Y*Wka77ypc;kZ$sx6k_*pKqA<PCN-<z{?%O
zB6MXnqBC8`?^;OLy+)XdQfiNYF7o#sda^CP*hW`UZa|adM?0e(p2?X%11!Vb@OXdV
zauj)%goxxvP(FibzqV&v-wrk}ReC>+@dxBY2>oNZRBvKJPiAe&#g&$9xG&mtjW(S1
z@U5NQaCrz8aQP`FEiq^VA$O7-)djoI55DL5qjhKdCerKK^U0911fC-5Bw;sxg;+^D
zJnFNO7Vuluju{ceEB;7aHwC#EWfbE2_zhioE$Za)_O^DE13}iBp<brj*2W9*mxTW`
z>WPo~6FTCOAg45qlEHo}sfA&B?8unE2Y7L5Zq`e7^JXEQb<ePlo8ziHWA?u(c?(IK
z><{ckFfiHXEy>$e92HA9m+>;%$;U^<d;a+uQdjCYY(|rq&9a)yn4<b(>6L>cK8yM)
zq3>}mX8GmtQH=7Gl}ygCYj`{M+B+)#PNO7symA^&f_&yZO#iS5b{hqJg`T|68cAT^
zBRj6ENGe{p97XTI<gzL&vG05(k?}pnV(x@|V#;3$+xkK3L}Uku^l<qNYC(Dg{|*!%
zW*?o6xKGLZ9d@CJBX#H0Ze!u+5UtBE*j+P<FyJ$gdCT@eSlozwzZ`P`5iMGwH+MXm
z?Mzd8(FS}SIDiGM><_iMjeYLi4QC0}g|>+~_8|-PXL6KF1Jr&pLKN!C3C3*=W+h`;
zu~?ZGa^C>v8QF4PBZiIQQ31WL`K?rJhf@Avq408iPW<@Sp}u8s|1jW>O0<(sE~}3i
z-Apihlupo-8wP)%iMvauR7+%4&PV@GBi`F(L0Y}!Guj*%>sf=DjKmMIf*)^Vr=0ow
z6}r!5!QsN1p#m%bkhTT9h?l!loqHQ4bE13t-Ug}lrt=0md5mkww<O}d`9}DC<=lrs
z7@7_%l2&g+(`R{U;fnp*6PY^!kO|I%i-egNmME;rw!*AdE*c1(Uv-|&)oU1cGJm^p
z2ht4LLmkgL&ajY=KPyhar|euQu7l`;;&wyIEqE1myfQ*zv|!yr8G3`QZ1r&rNP9#<
z;NuqUMsYmq!57E`q5?OLA;EE}$zgp-cT=O%C$PRwAwgl{)*{z*OLW!FP*N59YPD+a
z(A&F!8oPtNS%?_O3f?tqaFH7RxRyNQC5wjqXIzq1kJkNf5}x0eIQ<-ZxeIo<aWJcj
z`Q|EN9;P|kD9}r7<Tsv+66ewTM!V;}Urjq;{IQ5OxFV`MSOXuI)P)ips`7%-m$o1d
z({`xOAJr4@2uG{D|42vYnxw|?N<jeA)7?h<cYb{BJ#Rf%^!S0N>L!j+<JW;XH0aRX
zMJ}}$BvO8ai<kOzP}=+t!TGiuB3W;+CEZ!~YYw*Z0A!ZWoKx({U%__ZTJOSsJ=@2y
z>#%kO5Jx>+c#~W+$h(6gnxpv^MN;mz!4p<rh2NzpYfgp{dU$uxy5dYL?=7=^H(?W6
z01MtCMDL-jeDbh4R!dm%qO;k~o5ZT7;KPSnQ$qut?B;}YhY)`A7-hx?fWuW}4@}Lm
z!-GD$%V5wGuB47hj8`0?d3<5s4b2SG9VLg%|Gryp8QlF(8oGlte4$IgUDW|7eEpt~
zEcYuXocrcdRx9!IWcmTiq|ARM3E~<j19PN~)n+sbp^iLrB505_GQEZPQ75iu$1f4C
zNf6N>G{3W~qyY!$2)D60fEv#kq}22L-&mCa9uq^A2akgPoiz9XUNKjogC>i4;5ktp
z6oGykC5pE#l-kU9kfXGE7<AKhVesV63f9!o<gob?_z10Gx8R<LeT{`+RE~!j9oxK;
z7JL_hPMKOYSPZ(pn*^zzrI72d3@Nb1$1tLk;QKY=Qox^~cN=`gv+^n(<>llinl0v?
zt$yovsj4cHH@+kqfdj0bRE7;L<@JXm1m@-g-l3e`rK*|N53r=}2+&if8A{I0@pTr0
z!%D+R>a@<0n_aRUEB_XV{>;NI4_qQY!s-{$L_L`*T;!crE>^+BiMcreRT5KJrKD^+
z%-P}~-Zo36w|_fseScA>%d?o%ri13lx|i_d?IEuemyj^83pVdrY18|ne+f77?r_id
zWv(>4`7T%5>P{`~cJ4N`o{(D$uc|Q>cg+<eXFzqqmx5zABrlE+9Vxa3-FBKW=bpEQ
zm{;StE+Gc<4lW6SqMs@WQ#;8iH%z*=LUAF&THCd01W{k`{&|gDwz==3hCll|uBe4J
zE~~qM-6*?v8(JIA^e0jm?J93o)(m3Wdv|)>sfCqUO4RA(iy4J4c}-k-Ur|>>Uax&9
z;#B)nyAbZuR0<3#W6^ATjpNqf-4_SI9Sp}}yGG_0AAPlgQ!vJ!#Y<DpD2Fcmm*I6-
z@LB-vioMq5RXd-6=#ek<!l@w*)sXv@y>|Nu+wDpu8(i4Y5hprt_&$;c2)a{hB}zx$
zkoO!Fdt2~b`&uIozSq`i;too}pQ-;js`Uu1*wp%9ezW`jk}WMmUArPKBVA~@rcc~h
zcOZ=$$j?6IfNNQ?nsMS52FAO6eKifwCs5lk#c})y!{u3r+4w4$7LfZ4iVt!wgxcGA
z6aC+N<$U@oh~ZrGP~!$+RBua*q3-SC_?t2Qb<I#b5qA}nJLE9WMBwZF8?Ix4s_HoK
zEktj-L20m-xqK}hAHqs@DYpN-qq$HL)_FFwXnZNVU-)VwATSYqz-dp^BDaeonPH<`
zMJaKsI6Gfh+^6D_1Uq?D#pWoUCw?_@KGjju!f0mXwJTzCDmo7;&wT&CMeo+)RfM^`
z+k8>J!Jsa0ZqMozf@(rFcx~kFtLX2K&67=HX=i2M_CG0RCvFLURa1Zeasxs`=M3!R
zqw%e@I&0>4@?zkbR6*6xFgsioP>nK}QDfYB4KJiv?zHBW$j@gs)6$UhguMM*6Sqe;
zAu_h16~_LaUpUt-;j3u!LyTI>KBhl19PMExwt1I_*5%8k<SSFKTD{$}?$afN1kG3`
zZ}OVg!y5~cP#jM92dpmy(74iP7WXk2F2HV)a2Ylm+Yki8Mqe#SK*XA>D^Qe9UfV=m
z)6~6Ma#RC(2Y5mNt&5>RR82gb92OWsz(q&+4K2Ov!UPsSv9}Km8J`x7<uwG%#&7Cc
zN%NQshC0al%i+7!m|~GQ7KCudV6WK1P$2(-R9V5rOX^q)y4(MX;my}2lYkWuGjdly
zzmKwGq39t&p>kckILg$$#372vsJN$&v&P3d=|3E&e&dj$^954h-!)53G^5qEkKF@D
zVsLZ1)Lxm_Sr|CVGko>jxbv*Ih%rCh_qwLqX6q@F$R^l@7Ipy81l*1Ad$MImm(y{(
zfpT@4&6<jd=~(O*W%Y>NXKNB&UWyWSL7Wl{jpa(U`8Ig(bZ9FtOPI8G0Ln$+ndOlU
zwq+r~j8(L88;$222exPqyK(t_SlVNI^Xfa+i<&_aG`AI^eDIxk=c}qIZiEknX_e60
zR@Zgi#%js>pzPhYLDa#lZQh~Oh($KbK5)XaZ3|_0sCX%xgb=W+2pDEV<>?Y%AQR|0
zD$O_J7mFmvbNt;y3-P}40fBsNp+jZA^4-e6BDcZdr+=ds-(~f^xM$eI)4be$+KUMU
z5%U)=eMt}NvS_M0aIDktp}Q4-L5JhAf0lJD393S|CTbC(PARFkAt%G%WSOhbD!1Rc
zRbJ?b0M<vS3Y&9j_>*}GwdklOl+a9KhC~$XF^&H|dG97&xZ2AWJ=aCHt0}(?5%h7{
zdoVPaI~0ElNPuvJDAZqfQT}}o8-50%Cdx~p_-I@Q&?U+ScrkutIf<BJ+Qv6YzOhhH
zoJgSLqRPPf=_UCC^1q~Voi7cX>KczGzd4+XTy)Tz@4LiyzE2s`K^scs9pD>ScU<tH
zmMrZcilv*v4g|Fv$MCy5JBRPy#lY>T1K~5%txf(2@RSUE+-%G{@jJ5Ru!RflSjlj|
z+2)z@0nisydbe>376Fz{tjDxSc*Sd@O~QxJkNyhg;4|lcr(q{Z!(VB7>4biN`mF53
z0#Rl!IC4hYDea`(XH)(JrmTdyR9#PZmhb=rb{{T^Xp32cx7H31#)m~i|C05s-}?YG
z9OYFVHdMQr)K+uQii&@acau~YN1S${VeIA&Qj@+<Gfz8T2U|{st)}`(dGC<g)792!
z@~qyLJzqf^2%_B(jni=&(jYLeL+X*tWYAS)F@klxh<g`sc1KQ%u49ezHSvVXthCJp
zWTTJylG-zDv^lT3Tdl_X(qz`EL`5>R_NHP#eVPZ~WL?058@GOt9Rexd^+H!&+%E$k
z$;qgYDfT-VN!JpOpLzfBQ#f{@l2Ums{J^nfvFT>et8(UpWJ`aj(<FOl_<)wJPIY9o
zOt?BFP-d-;{uQ*@pN;>0d8}32`2_imTyxW60#V{>{7E%wOMfjsv1HNI$=Q_mbc?B&
z)lGb|<J?C+;p1v?q09E!BKtGnfbu6tbRL|!l>|i_=>GccWo~zLlrwF-%EmPN7qf`y
zAZNB$*6)Y1PnW$hE8VRc#QLLlkt3!tMo7bnYDQfo6d%cd4pF-yX&?OQb}(%QlNP~}
zLvfKKueHp-;=vsxYH2)P<WjFzo1Bl2w^cRzz6SY7X7lP%lv$tmwU>wAp~#~c2q*c%
zcO7C>1y8|+bD5O@t0_Y&hk47<!pq?#dRPPa13CJ#+o&Es=ItBjv~C2JLm^)qni25=
zvB;VmL+_fj(}Ot~^a<sYSlg<>er$uS?|e~Q<fp~2rE<aG!4;hOoW{#z8$2XIEZIy9
zMPzm4?-+LcCx$<O7~Y$9w;R1)iuKfrjo+)94Gd~}lXFV7op``A#osvjg%gNn&oopY
zD+r^GU!y<oYZh&|(f|v{@;hWMZU7yYLChXJr?(XR8=jBh_U(Ob00u_wjoSCRj2;d3
zeR}dkYcjUr4(K<=^$G!-T)H-zU**uugrohnr5_Hi*XQM-6b|GMn0ILH)jUkl%ByG>
z4m<(MV5r;o(<{*Q;jClHH?9s~E}+Scm|$cp2@&l!OEN~0=XpTuTby~45Us7v-Rw#a
z$hQR>E{mk{-vI?wwsW4Di&8|0Cq%o*&&^pRs<KK6p6NU0vRhkPk#aF2DB`kM-Vm4b
zRJJgZo8F#7`RI#0Hqd2!aL|X$&hl`k;xM$aTG%Q4cd+)VNb1S@6FYyEklr3{IO^o_
zR~z?dM0@<gf+qWo#iFgfM}>)wS6uZKS2?OF7ysS+ZMGhhlZ>uk`yDo&WfjBt&nwW|
z*N3<0%7)k{Uj1=+A&b-Qe_B@-U^kCDwJ=p3cgkf`()Mn>>o4}3w&OlrbZ$sKA}01?
zx^FV?K#AeFx4gtEMnDsMm<}F&@!@K~1Y8Npw+3?uWgrCnew43Ez1@7!q&2#1n1^P#
z&~3JK({&AI!|ZHEffpt*R5)p?o)8YnK?U@hD<^?Vhh}X?bDU(R<5{uL1{}){a!mIq
zlyANbDVDNoZ8yXahA=xYXS$0EAr>WcqC1168<@y>LcKQBJ<!!}8|>Sq4l3TD*@5L8
zKtzK|mxN_yVxh<3X14_aK#*pIDNMBf?l%66*3Gli+zzs-SakW~#S0vw!q3ccpC~;_
zHM8Nvz~neu=k)4upF#V9J!E<21~Ii@_1C~BN9%$E#tE_;WL_1Zru8L^y$3$=((ar&
zw(aDehzkT~)a8VhGJeQ6>G!JM7fQ!>feHq=gM=zz%B#`4S8Li+E(xl{7$db+JE-q5
z=$0b@j8Jl{Bp(`vv4^U3(0VHwr3pk)EOv*1fh4R*EU!fnEAOHWh1IjRzEef#L`*~!
z*b2!Vge~W|n4m}fRBqPlni>41(JTg4{CET2tsk?6*B)i^@M33cU+2~GnPLk|bag17
z+mtcFNXT{+&*g<tSYaWMtnBr_y2Nq#)obyjJM3#8VZ5$rZKUCiK_2JR2g674|9_hO
z`TPB%XV^q9bKGs#WV4Ew=xNfV<g+;{Qt5+lQ_L5Cd}LSRW5i)?k#=Y2ug9~PKi-CD
zy2V0I-`6MPUpv%?Kh#Nr@{cNx7G8GcTR-%cGDk~2521n+np}(4_U0O+siS&;&?inl
zfbs7m7W6Dd@@sJMV&B(2@2gA)EIbK%rqXh)64W$kBjSlcYxEZ}3s;~;>l2gxC8tY*
z1@;C8(y$%N(S({xCLC%1@Zl9<4_vg=_XqHvbX!T~d1!+X?!>=_ioL-$ElzALY^Cf2
zJsaL_{q%Fr-rq_@cc4v|D|}kWSZ2=S15w=E-ZR_VELz(+J~iZy>qO;bC8ENw4v;GR
za(0x{a!x^`E#BN3z~e*zw->woD~&~>jr(0@A*`+>6fdPXq2hC*mfE+|R9$PEb|a0H
z$q%MK>3$K1d72Wlf+AnlG6bHooP&Y{=w=H=z*r;vgs1r{?m>aoLOW<#m1^Cq&+PB`
zZLBd9Si0_AwDYq~g5tsy$=DJuV1a{{={wUc2jh|J^?@I%CN6?`4yz^ZD@W_Cvjx{C
zw~N6lSC+nHT;<0?>erbCK=<exDozLyl@8XR1vrUNhx4MQ2Qz}!7O4$E8qtcPgeYVU
zuu6bB=d>{4^u-+Tidy;Me(kvz^naZ;=Xo2oeCfZ&8eG*8P<G|@@z#fDa-#8<aY2{5
z6cHm%%?E=ZK%ZR4%0eKI!P$1&>QZnx6UI3F{}ZY(v!6Kf9=Ii)psRXCtp4M}5fxu5
zVe2KC1bwf!IMZ7!`EiiKf++PXIM+jeK{X>s+NQEJ;LN)y74pi6`z22z>q4P2p*+we
z_3TTZq?)%jmM-1fFB1EHKPzyjtU?+HxE^Rj8MmX`v3nwRO^;fQrC9XE#u@hjroAKv
z>B)9zi~1Hz-$z>XP}4TV1Wotv1S40z2Ha{d_-^&#O5i`xRjSQlx2&MNCVMTqomj&+
z<o$?1=X3#yZ=ifav<^a!f)WNyk(LVi19&AJ2>h5avX@skl+5LzyN_)aShV^+J)jz@
z^3TzyhlPm8=cx@9(JvqTCWB^|ktiM*st1XBEC=qr+Cj<aKvYzMHPf%O|5>ei_-hCx
ze@?+;^W?oj=86%gN*y=1rMXE8gmXS-lq9D;IH4(0*Dabf{zg`?xL{WB>;gAv4W8u&
zN&MPz=+dnNOmxA*amRU$9L(^?Li>4~ALRYJTb!f23oWWYsixA#)Cx*$xqZXgMDq}e
zJV<)<y{;_N-*hq_<i^fckvcu{E~wQOo>8KpE#h%4=l}=A2@%CFO1R}8ilTE)+9xHb
zhcSy1aY9Bz0krlO^1aW1guw?QW6Y-)z|*KzvLQO=Ji==A#1IxxN|Ux-pwGV<cNZ;u
z=ER8jOoDzGMys~pSL&N2H@RbiBK$)9!u*uutxMB+pL9=U*WBRurBY_%bSSD_2P3M;
zpt*nZkU>SAQTq)~2Ha+({uV33#Q%xft(TL3*sMz<c+6_azP8HDoIat-rHiFfGFz7R
zU`gXNcZF6)UmsLX6%cG!(tIMqBoh&j8^ib9#xf7+PHl$H$aQ@8Wrl93Hb&<hs%8WP
zcNa@D2W?$173PlKEk7OMAN14Hsoz`=9Zqhr91yis+l_b4LGBSXnB~Y&2QMLxN3$91
zGNT=A>@R_G+o#~!5J)qEDIE6t*6-`9#btF^eKYO$%sjU9YVt8!UzXLpoMFnv!D?Ht
zff#?;i{!JC8Ub1xl<rKX7hH2hYepW*GNZDdK|uw`jZdnrVr7xvA3JMWRG7PBY`m|J
zTP~|I%XwlSJ^d+e;$*u83;D}Sxve|i^-)Y8%Fx4=rr3<%6YfAs1a~Coe%kiQJmK!z
zAo!H}e^tlY^RcP|y+3t|v1Rw_p9{{n(EQ$R&T}Qr4Ld1g>~|!nm^}y%_o(qX-Mr}U
zj*H>(U$(8J^((3mJT1#Mqh0BkILqQa<+3(PGtcB<|1eiCexsHDW<PeLV)KE#gxmIZ
z^KOqEKIx~XX8{d&x2f@I^Zd2gJV3<Vfi_%stJT|L<%#F7gB%43i`Q(|MC#=YA~@K+
zS2q?qEp%c(KLG6Z#b{Ob$sGpTvdv~SHa#L9^%&?yTc`)8dFgnO@WgL`6uDLo>Uz+l
zEEn2LWqcIR7#QbP@~GpF`RAGGt>#PB$<~{^`2b5)g-e$BdT{$d^$pn)I6uX$$eL^9
z6jKZGAWwzQkHr!mWnJAmYeDzC$yBVt^NiT2iFg0?2dS&_H~s}?zkICCy?dBbWVIOk
ztseh`f{JF@$}sd<_2;7uqh9EN`>BOkg5g+g`1Zpz`#r$NQ0Y^?fT4#hCBE>U6}d|s
z$ZCbF@yxqBw8t)-RqXrdcP*{y+4zH+D!Z*#6{Z&xuN)mWLg%GNB0ux|*~1|l#awH&
z&g#?Hl3`$!C}aB+0!M&taL-8O)=IzhC`Bj}7u<xPBg@$RYr7-eI)E89tbs1~3W!0K
zfk~+0<?I*QE9r7;LM?AO3s7HbRAq#zw9C-_J^gSGMi<X~)x{#`%wPBjN4N1|E02u5
zpArfyd|Y{eKi@o#eOh&WX5gXvtCPOf1mOqw=S2U8oUI-pu}a8@K_8vv5_NpbT(I3}
zB(}2qi=4yZhyh_SFq{70k%IsD+IfHG*~S)HZzQQyny4B?Rdg<WYBD60<sLb5F1g9d
z)ucVB@nfv#Leg*e@Z*EPV*7U!g%bGx(f9?tJiuo+H`<!H(8N#NZDvG7EpCa2li)|Q
zG~LYH`KwVU#FK_02GS$VM`(6~T}j0#VCY}K0a3RcA-jY!RH!B5K!1eNp6&BjM%M%6
z&%{(=7WQi^Y)srQ81*NtjkLF9i~O*Xb6EHdcD#*p9==~^v!szY<wLI3vy^9+bIupR
zRx)n852$_{h`94X#ZAD#g;QdZDib|lQNzOL*mtww@sk6U*M<g`##(HHQItri*B7(u
zHuU*D)jQ`kFv?lDP(369m(hM+<-)!+qD$F7U6adVwk{a>?mzp<&)3}B=FI|eQDtcw
zo~~ql7AXG6{qP|(`K8mwJXD{C84Xw|B-K{ay3)Sb&pATQm1Z@1#OBNXpO(JWeo^hp
z+#O{4i4Rzg_@+4i+Ke<UZE2r1PzEPG@))MAU9L5j%QTHR7se1^BxA-RLB?Z(eVp?m
zNj(@6czv<6TMLqX?|D-QAe3aBM~&vgIqJaCwghaxPGMUiVgh>HVl_Ri9B!cIH6olU
zzA9J8ivrmOLbSbsq}bIZGh58<<2I*8*#<wQ;l<MrGtMxqSXB}JKUge3Tc$*6eb%fV
z@X@?o&j4?V7u+OsMwlyU)Ge+JL9&{K%3|))2W{tn(%)fU{luw55SWL;v((J8?MC&G
z3bg8mTm0VWY6*jLIe)4t;la{9_v<&KILur_Y4BL@xZDvm2^h~KUrN^cP15Q;2TgVy
zFh(yTyG0JVvOeKK-HhJS%LXfWvp4pho^O@Et{M0?o$=|k_HcmXtcML(grNeVJ-8IH
zDiOJ$_XD+=6pJyUk7E!~BRj>+`tyrDkb=uFrST_Mq3J;Uk97M@Zy<aZmNuSc7}_pO
zDnUs_y8z$48AG!acY)Crudf$?3g|X)9a0cC3_{j^B?6M-!W*=kjbVP_E54{94;aW$
z^k^=3TDFI5iKIm;=JW23NAB^R^*C^0prC`iLtNxDX90bM1+y@W7vnXmqI+b;bJX&G
zIK-D#ZdgMgZ&{30=f7{D8Jnp_6A_;%am4+i4bhL^e~PkC-bUYgoa&xacmRJ}P+u`j
z$*+Z3S5FP<J>hG~(&+hTH|txQy%{=ua3O1l-pas?8#ku20_LYdfT|rWPrV%3{rm!e
zB#rpj4)41-AN)oEH4y`PxUf{z5H6%75s!O|Ny7(CXRQV$n{>f&86%(<;dIJsA#4-*
zeSi6Y-Uq79H2;Q$qHp3~7+M2?(nbOvoFeIFbsESh3yFLiT2}?;Br0=O!H*l^B2ymj
zB4SzA<T+K88L#nN=Q*?w7KN_HEl!t*?r}LqkPAa?i<uT`=(c^*WK-+g*qgz7z<1tg
zHo!Bg{+8`;z)~&<*i0brnJqft^3U^j-qvWrz)>tpdZM{>q+2OTwCg>`%bik9!rjAx
zIbNxI6{X)t1o27*Z4dE=FM&GP3hZUTS}KmXA4Y}m<&As$F&?LyMh6kDY`?$^e{$gE
z3J*e5iB1cNxi;e2o~_2GO)hVdeh6bCdK+PfgcjZx`P^K{)#3E`exi0`prFSdXjPAe
z9E=IwlUW-HE9|qK?JhGc7Wy{A&NRPSy$?0=(uX&Z9sN41!=1!tG`;2MP+vgzf|083
z98*J8wGXk?xmM1XcOcOIPJ)*&JTf9qwUAoUp?0KFI`Gh+F%Gq|D#J0+z%QNm&3omW
zrd6DI#r$<_*9Czm5KbD&SL(=1`kVju9RGRW4eoKCw8u~D-Mell#_Uy1d)Xt;9jFjE
zDS1Un^{0w1T^9F0AC1e3Bdo=n<$MyElnYmIVvp>iH$MsOrz107uoKHy-?({2+fBl)
zIBI>Obw`H&FAeh_`al=*^ifFH?<Y;3-}wdp6<9tjfD`ZncFThNi*g=N_a#v8DGd>n
zPQ72Lc3h1a-ZCJO2#UP;;4cOh@)9UJ;eqN}m{Akm3*#T5>YcwNS>jj=7!R{3ZmQl4
zreyT=%OuDFo)et$cv*GsChuVDED#ZdfK^3MwvAPl?O`N5rQIAFh#zckznKD@T0K4q
z_?Cl$o$53kAowfE=+BhiTJoQsO#dJMnK8>^)2;~b^Z*&yFlljClJY@#>gA#`kuF>6
zf!T6VVak$M=K_H$@1N)*i=S336yD!HZw*Vl=lE4md>o&v5!yDY)ADkPib*yvQX0yS
zOVR^^y6@HQ_@TP>0@K#oZ~Ef)WEio1G<*Efbh_3t<893iHDO>dd>Zg7^C88}x;F`H
z^j9Z}tW9T`>Cqz>0020WyDwP4tF_&%a}2(MrM<&kKpG79HNp(avFGVG0iL4GW{pXN
zkKw{4?uh)cy&wifZzkokUM%!jNHZ6aeo!0{giEFWInf?iZ(Xy%{XY)15hBMYUm0fG
zhk*Som#ey_-YRtKe=|#WY*w#PfJRoG<`IJZn%~Vb+&gbzGQ&y$?I3z*>IB&Q@fTvM
zC|;yDt1VmB^G)NyTSFV=!Nh4ER@51|$xU$VWwc=RsT0a(gfe#TFelYq@#C9<9!Z_$
zh~pD8ny8(Y1Ae^?md&tsVLess6TaX2PM}Yv*8~T4U8LpvWLO-g*?AfWoaZluo^a$m
zc{Gt6lC*F4Gwk$FXoFF}Drtr&fJP3MSjsXds_(Lom}`*Lt`4J#j=_(0&M*4vt;hmL
zP_uvYZS2NTxqm=5*?yT#9hNq#hXQk1+*@!1?ThhuI?GDDd25Mm@7+3JeE>2X;dPP8
z86!-#sEq|SXB8u>ZcO*dgcHF`BsF^MU2FE)H#h(+i62_#g6NfG%V`a8z~CuzC$O=J
zW1}x=gP@fPcHXuM`P(*BOZbbM-}r@_mpF;Lat6nr8WJ$=3q&b~{^E!paXZh1qmwi>
zeA#D7bYErR@cFjj<!1b9P5h#jv@Z(`g`{0d71V#I2|jG&#GUE2^EeU0wd!~VWk0dr
zMQU#CD}0~dxQAyvFYenCQMzYq?rHN3pXvAux(i4AMtL(?UBEwHJ;uK0q>C$(1J8Zj
zo%(H@rI6_hU}|UX>XunKaP@w9h%W3I8S@c-Uu`p6J~91m33^P#c*lqs%(}M~LQBfG
zLdbd0ct<^Pq0ydK<ji<>CxxY%@OVtU@ZbRl$Eabci;L4zQ$}81!zEqGhnb1}{Xl}#
zUrdW(7u1pwllGIbSR~*O{s97(w*9RLl?46Aoxf%IdKZ_gJw^tvSCvAlPe<>E4LKc{
z`tjO@u^eSq`PEB&w3|W+Ssqy}W~ZEp{@!RYO6gQVT)CW6<PZhKb-hGvP#sw;>RNUi
zdude9!3>O;R3(0{r)9K7*8^Q^yUxSiN$-7)=e@S4P4wr^pS3d`DcE`p)T>H(05Ma)
zl4hsu!cdPJDXg^Tt3#1V*lUF~OQaE$?y~;Gqi0`SEdd5LVIGS7Ws)*t0U&qs0tu4D
z(Udgw0`qth6u}=XRmZzEVv0ZI?AO!eiTvq(S&Ul87hycAykeVyw-?PeO2JLdO>cX$
zKSZ=7GYb#)MPDxW3DWu|6#cS#4H+@uQ{LBa1-h2v^_7EqC?8z>Szm276dSqrzlZP3
zAw0kOy9KIKDfa7bSx@*?5qfDXbMJ7@2~`Y7|8%bL!;8I8Z1O9$c!B^{iRY7$a~?5{
zfxp*=kMtH`fIpHMPEm^Pw4Iji4LxM;Yu@oP`Iu4uw3~56vGoqCx7#maZ7!taDULw7
zFOv58N6(QTg!6HZv*lkD8tM^_WLHUi3@}xB5nbbM*frg&gstS6mw>YZ_6>w1-IhK1
z^@|uu6Z`fV$khCx#Jh37Bf>vM?&`6WjVD3gVz;=aARSkqP2N*)lLIh8(XuvHC?c?p
z7%86I{>qBkHW21S4k`J7)C?)x@-f)sbJk~vS66gyQatgOYCSm762N@^sfE$X|8G=&
z2ItIHtYPe6pqT(Eh4f!Y-^<IWL}AR><lw@a2TKX@gff?a21)c6yYh||9R9gECi;r@
zZLAV!?&NBCokSJC8!i~cV7u{WqVgjiafR~eD?f2cPc%IfX+@M;Z$0%g*R;mX?Ez>f
zivhbbiNn-bb~0CR?jga#BC0ZX%`N_>lKsUuM)k6Qak?YSw-WQ*hyF!-(3=H%p@UL1
zA;h6|!Rvqrp#z9tsbV`~yn2*S=r~#4z6+ErdFIQr#N(-!L3@t9Nhq{IjAPj~(Mz4J
zy}6-*?QLvY7@s{qpNs;oluRv2-XxjrqazcaTj1*eOZNgGg%zsD-#J+I>r1;0%s2vE
ztKyf9|74?T1GMC9P7h-BO}W?T9VbG+mn@K2k!Is#cuIAj-u+N~%EC2M@zR-VA$;q;
z&Sj8Nwt`bU^3uN$OroA%yvy8k{s51cGjidqfh@BYmi#W0{Inzg__mXWp8^kKQunV8
z*$J-3vD@Oeq5C|BVxdHw!sRN|l*l^|yHICjVO_RiUtO!5<vD(<aXKmaWOAhlX*et|
zN9l^Y)oOvo&QNIRw`{teQ`|w8issAN=f`Ya8}8EW9(L#WQ)3HF9y+JlF~#%TdC^#u
z#FCXkiIjKdrv$d;?8o^NsJ-i=kK|_FWuZc_kTs~!`RYK%pLJX9#V?ABu4fE|dqtF{
zulgsX?%BGAVA%Qw+7L}mDkqBk?5Sl7O`s|3YDhxlP>e5&eKs}einxOM&40HZoDv%c
zaNnOythFWPyQ$G+{vN_e%Ta5NR`bf&s|>>z8)3ulx&lIRljo-s;&Q#~mDZN++q8G*
ze;d$*xKFnGz%=IIi)`0KQPK!O9lRB2!+H5(w9XdTtX4zyq{q4U*!H9ZAD&2^;y7(B
zIRiM*LD*su*mKNNUyLB4D2Il2j*H^x)!J_FwxJ2=z>fh`vX{6#r_3l5)$R`bPK8Cq
zOrYyUaKY^FNa#eMC5sjoGz7$ZOkHQibx8`)3Uf&6_>@c@7tB-u$H-5dgH9K34cgXN
zpsiNuZ)nFl{GXw}WrxnZ^^&-3T$1;{eyd*~*dDK&)BI9ZhjnxkvoV}B69w&ik%msw
zrvqixMKw8F+Z-o#IQ;eaxX_rDv=zPxp=0$3tMg6C+MijZ`*2>q#R=pj=F>^~IIq6z
zY_fUE0fTK|fO0+n1L+@UxINC20lBXka6rmZ<b<^r1gh$qY28R;`ZUSuhKateR*Cy2
z+Y6W9YlKn>l*p#=k#7ZHquhBg*i-0q<8ju?I_x=3d`qI+8uR3;^ZU?q8b)Ygey46m
z`}2fbse4mA-;`AEC10Bmw`Nz7_P_xFmcIK!xpFS|_#Ub3jE-NBM=PkqyxhS8p8PN^
z3prO5d+So187;;^%qW<Fk!SQrt#tmvLbuva%%RWxIm$O`QhRL;K|q|uZ1Kc92fOWk
zC>FTAQi`pdZ01;2gl(DnizK6Y<mq{C*n3*y)j?h`n49(q58{IjoO{kjpBjI%&b`@a
z)cPZK*CYw!p{tZn+NPdNJ8rBK@oc{-wp`p8*Yx7IIHVAWNg{xPTJ^wa?s4=)3@Y5W
z%7=IYoxZx(ZbqYx)L1xi4;acNeP7$osnZYkU_7G;qa>$+>)2@I*Wj_MruNkmw^kC2
z=NCY<llR*l=W!NqHJ)!v@k~3qhY#*{o=_6Zaf}*Us`A%8nf;`Spx{#T8X7zs-h@RU
zVV4@fMn*N%dVHJ>Si#vjDw=r5kcFR^iUy}{zH~lhDwNKnRL_4$%=jxv%kbm>x#T^U
zuYa{`RCv1H1#2l4-9~;bj|{0UDV$NXs}{m4YQ%vlhBd_Pb~|y$9xdJbUUW%!g;VEk
zEGuJZIj1^`r8{04oD=RvyHbzS5j3rS-@rNJ9WrpXpm}A1T=hs`i(DeLPY!xG?du3`
z91=g)KmOs8?}aAcQJxd9uNKM1TYx7HN1{cnoK7Ps)|JS1LI~g7cIyj@pjR(O(r+Lq
zN%W!=_d+mFr+w_kBc2h*VZAsB`y9=H7Ub~J*fI8ZTPjXNNR>rf|A^@V7|5Gwt}$%7
zc;NCYNp~6`B|~}Bd{;!9Fa6rka9&24ZF<^fgsW>`?`EKGh6p^DqYbBDX8BGVm~i>J
zi_NAp_x}e}e~H!LSN<O*Wj>Pl<6k|e<s{I=_=*WKo$$-By&xLx9Bpz;e(CfZ;vOC?
ztK~9z1Q+E?|2J{#P+N;oExpd<op^B0Hyu5)Ig0-S@KJOVSe=mOIox-YSK3V&<}n!_
zb9_tt2j|1b-xmt1^H7GeamI}v(!dq$bXp{#k!QF#G0BkkReCK;03L;7i9%^NB<b@4
zvkc>)pLd7obU&Fbc);Z&cJ7?bC_mvA5IkvIUOk#>_M}G#t%&F}tLK{oL6}$6ot;Ms
zZ$mxZFPZx?6JHe~>&y-aW!|SKCyMm@JsYaWD#<&z8i{wIPiDZP+YsKdA%E@dyz#XE
z*5_NuU;kq*%~)2ybwP2}pR;p;>5sPVoIIi6qoXgh18)f2_4Fa<Urml1x1AEj5e#4I
zaNJhWR)9u+1D@}uY9w<BXVK{%=#l8+rNs1y5h8CxQD2#*>;@}<8^bx#m)v;t`DkNb
z<wJaq+xCkl2_`KoI7VDiQt(-nsEZ+F*o>F^mApwr&SC`p{_#hyvd<)&Bv00Mzc07z
zu}BNYJ<c*Y_#0SdFq8@Spv)1+uf-&K#E;X)Xv){VOLmb}u!Sts+7Iz$$F`!k#;zDM
zGB8{?g)2Jq&e&SDqAIUPkAKLvC#N0XssJbHpb=V?>1bMfPQ;q24KE{6$8M3irxFWh
z$-sYhP}%wSY3eEc&uI!?zW#nk_=Nc)H?ikHcV>zqJ_65iTKZ^JK-G47baKLeQ@(9M
zC*}SRz`9bF3Pzzq4;XQl@+y9q6+!8$>9baf>QZ!gHpR&GaK`yA50dN-gp_ryR=nXq
z>wZx4X}}CJiMr%#Z4!U@?C1zmcd6Q%<0WJ0A@D(KrU7lJZy=IyJ5T7pHUvc*dYI8{
zx^V*LuQd;;o*a0nSeZ`iN{m_7W4<hK#ya^u7-&7ZFQkQLY0{k&1I`h(OySVeo|2qq
z^5$tpUg}#x5!>epHDa#wqqdW(_oFcI6cR%|cln1uHq`eq3&C^?=lWnjhDk=WF8%@9
zRC4FUO&Na5CzEdecU;2k{$no7URHm3y$pZ#j^-F>=~YuYYO%^n&1eQf)pbAnYLLQ8
ztYmHoj@OSEfAc`@c6xqWl6g0R-&#vn>ZoY~{ksD|LOaC@PCO$D++WT#8LxNkxy5~<
z0OMs0RYO&!b>JO&hwk7nkO$mYm&{|CV49-)V_Z!$@Q;YtOWnYW4Yo&Sp(lPiuku2V
zfY~86p4A_~|1?U{y+MiL3|DyZz%YLt;ZuAPq#2n4;H1%(3fH>eD!2%=rq{S?iF=yP
zn&dZMEUF5Vd=QZsmQM~4rnxequ7F(`A!y^vbaaQgIp0|I4g<GrX>{o|`X4ayqphC2
z`BtgmyZ?lzKSTB6AA^0%a&&&;WNR%(`mHTUC%AMg7k()V<%{PCs{4kPiP1x2KF>en
zv}Bp?++&EJqF(~%9j!f9aNdu9-i($7@Kj~+@3rK`B2i7}d&)SkaQ-Ij0JI)lO#uR|
z)X2)Sk#;c_-UJzYTz|dC=zhw9H|P&n4;L+b3+)ETf7n``R#yPYI&MRb&<b=o+nFBC
zLL4)hkp+n0=t<khlfT%m2jjQJz+lffyz>8*b>;C;Z~wd7=9U|`x}_|+l~UG9Qr2%p
zLUI)*+f0R2gGn0OY^AG05<&>2DC@@(W1ATzTQ$i#wlT7sF~&B=%$S+qQGL68@6R7z
zdS%9(^Eu~z-tXuAJkOa}&M+J*;Z?@ps@wtS%a_EDr7=3nTPjgw=P0UijaOSfIV2&8
z@spY#KOrixGj9vsR8MacEm*2KI!mx+X|Z)T0qkQ%LKmW3HG0*V^!!bTt^20dNmlML
zk%53{sxtHNztfzwBW54CUB6yL%ofaDU|Ch3$wn1l4z?*8rIE=2oG3CEUxnhJ++%Qv
z14k6KKFDzKrE{3h$Xk$FIz5}#vOeblmVLWFka@sI&a+Y%F?%M9{Am^BxZ9>wVdj(I
zQB_}JjPE9^fpk|=RZg!w6o#%j59<P5431#dn5p}L6tpZC@;<rq!9>V4@(YE0-kmp)
z@G~#=qZ#QaAPNAb-NKV_TALO53pAR5rH&r+h#6}Ny?4-=Wkp2w9rhOwg+-ml&djz3
z`1GA~Lgn-0=&$~$nwQ2FhnMsLlzFn`eKSY&0P~aBvpN$kp?wBB$Y8~hwfhvo+K>}k
z{w_H2=C1fyi7G|d_pS{D47~x*Z4$FhS%`$g0`-yZ8xfofR-8&nar@I-P~|w+cy&0q
z@%4;zu<})%B@ire#|T?~T2tRrjys}iEyE<A-z{BI$f~gPK&ZxRAL2DQvvEZ>CsZQL
z9z_MChX7+a!X8fQ%s~zAm}}pDO>tD6BEWAE3;YC;cIKo*`lS~j=W%XCRWP4MC*Fbx
zk?%~&{u!`(ioXb$jI@8KHM5%tK@cxZz#?d^r5d>!pBNRog!z$-4=?l6DAz8`f6QV<
zSn=ZNO-Hv;c#d>-Nn2$i@NcRnBBm<a%L2fM78|^}|Bs;fe8P@z96`r#b3&SQmcesH
zFMP+&AK-FCErS~wLA;);*tz-`56YfIL-IHxpp{cBDVB0=qL+fYK&d2yXS;Pd%#QNq
zB8pwv@+O}!l&{00;PVbeVcFnNMtCoKlC-I34|OPmyN2)NM^TJf(|B=Lq6|7N665_j
z2dEeie**fj1wjep+BnUWBselzMe11nW_O>aCNs&N-XFMais@$&@S98S!sZ;nXlOrr
z?a~eeWqIy##yc5#q5b(;)*sV;>Q?SXRxjvRv^6y~T*|P(zwd!$q6R<^X~{M*4<sw)
zz!e&UeMf1S90%+NNU!1ISFqF;=;MB_l<dAwN6^}RPwE~v)gSu&%=LHS9YS`7))aUD
zGq_Hq0nH}|19S`idO*DOJPbwl+X<35%SVgh@0tzL6<Ul5`}%<6(3`hT0jBE&PmV5U
zM0&Pw1`+m)FBQ#ag?h5@L+?WXHS1G__PFJK_AjnU*#~1^c0a#frn6nx{wfS@VEIuW
zg3q<DN0Iwk4w%^~y)*Mw;iluWT1;KqyTVr|2_o!o8ZF~0hblHH^p5NYB=HKjpiWTS
zoS%l(SJrZknI%?ae3ssQV3Q<DDD_TbtEN9c<LH7Q=@BsEkBhx=pceSl<YGh+*F~~m
z2_YPUWNVI7cb+qcjW+oW%NpoA8Um+eRwj1pEt39KST75{*v$KKZe9}-GW%hy`vg8W
zP{0v0kvk#7StDB}bguyeByCuu(z%dif{m=!l;$UI<0<+_;j1kZN?zl75=YyN(e9=a
zlN5wpE<f8)s>J&hzuXT=&eMn^S*WC;PB+Tww-zBzpuZrO(@)j{IlnAM^cIko+aoCY
zD?3;upGLj4@k`9nch!Y_r|tq;JrS_>BFH270^m<}dc($5ta||E3w{HHfP!<|mfgJo
z8TiEG*=GtQuWuCDb;QKymN;7>KUl$h;N{n`Bk#fnM{QEdDnOycTw>S8V}z^7!=rbh
z!o>Ez3j=To{m)IK9ygt%{@FLnT)zu5%E7+bj0+GKL1X|X`dy+E`CrCWkmyfrQ65S=
zl05%*`Z6PxW2s{D$1dn`4$i%Tx|HR=2mPu!SuATZ4v{Sh;nWR+@{y}RR@lLqD$nnt
z7xW`fkHKy<mps6ty^<5VJZ8&cJr2aE1*Pk#X}{PwbMDek#zDk{J!tKj_$zCT=#JWf
zqf8eO5LwRI#?nQIR_iM%d|@IO5bKUrB|cPde^kUShRVHL+#?E8@hcbuy3|?vex=PU
zLPlc#m9yh7vT{p1;!zBepIA_${?g}3T1vUTjvreP_aP$X;P@N>qAsAI;UzD)>wxwN
z6jO`u0WX;1^LQ0wO83~m7*hpS1!KagJm4Bi7VkW}(QSMI1L@9-=K~(xC-VoiE#h&1
zTH<W?a!uPSP;4k!;!I14q<OV|mFtPm^!%~lPbo>IZ34LYMt;08@l6smqfWt-Xoy!i
zv{O>A6B`ewdFc&NJK#Y&i7|V(M%zKWBr`@!2qe5$gGLZEs@k^~tO;?u^3IlAkTd8?
zmrwo=jG<TP^ZP^v*OfAqtO7RDn|xP6JEzN(Y`g$qtpT#@8h`bA{e7Pf0Mf;W>qz$G
z`<a>Ew<QXmo;E3lzLP`K>H#P!lqs;N<pmtcrB8wu)Pbp|5u{|tsk|lS6W1ZM(!N>y
zAC}6}J^{WB6Q)DzOU1JCr%t`!@Z2JZqIE@4e^0AKEStJb@?HTEw!04RSldiI5S>8f
z8j;&pl8yS(BwMqbcF7zE8LJGCe`&bPwx3#tgrU&~D^+c0>IB>P$2^_3CHNMPsPn^7
zDuvEmdtZ&hBa<QU;-MUs2ZM$yJ9yOnqN+PaJ~j07uAmR6wb*_kKLU>iN&r0lI)w4M
zeZG<j$gqB=m2D=tyC<!^hLGD{j_`6g#A4HKM>Qminqd%tYFZ+jEHXp8x_bQlW?|Z&
zv|h)aHdru@4}V%(_0Cz2Kkv)|gn;qg$NGRac}lzj6eC0xLiK+g>y}OI{`;FV;DulA
z90!Y{mfYgj1tYx3ULTMtnTIAAvu_d@;oBGz`^F-aWjxMy2fD$qa&8x_XNi9fmn^&j
zX{uf-TWNT+c^inL**wpVAS=1#@h&WhhJOX5UB;?9O^v3*zK}F-SFDP)hsvg)Od~bU
z>-k3Yu15#cPqyswQzu9B?_MK)Mke6U9~COw`>uT<mYrbK36jp4&`N7}x1oQ@Ul&AY
z`usiuB#Ls%!g_oPEtk8Q2d2yIOQbY;gtnj3eyXFTR3jhK($dy;C1covM5~uJX@>l1
z^Wp$bFMOSGxvjOGLS2RZdK5GUR5_X^pD30+zw-q&;0YcK3a9Dce_9-OPJEZuY+mY_
z$RehkH-SC^LJ5EAIY%+9p1i=EOXc9kNef{jK$jrz+5v^TuiKb8zXpp*u?1>%)+C*>
zi4iH$FsKek;UVGZ^TUI0UHzQKl~sh*dHinARe-hsn5leNu01nRapK^&-I5X`y86+S
z99&D&p1ZRjk|E3r3)kJ6MKk&zJ96#SqR?C@MVmV!UF9E5TUiAhhATPg2<7>aCf(6R
z{c8bxyjCE9M8?8^-4NUYuOt@i!we6EJ_pUnl&YZN_jv|-sN7s)hw~d=j%a23-E#Zw
zDEl$zO)Fr46<%8jRLyrNcaNs`;lHNrfFQAWOXj~7O+n6gYr_|-ww!ir7#;kj`Ioc^
zT#X60CF$CM_bqmLl0`31(Woc%k~k@<%wq5-k1f6v7soW}nHKTvb5nI~a`-woJ`eXQ
zDJ(dUUjjkhRQcrWS$rxxn9lN0=sC6V*M#HeXSV}&-l-Rssg@^(`Ci8|-P`gc%%`x#
zPMG#w_DM!S6gLl_I|!eyNB=(I!=<D*)K-S&*;C)=M_y;Ep*CwQZbnuG4J9LTYT$BQ
z09wh0(BDIAun|r5!AT7yZd?W?LRyXkrvyz3_|l`N34-ZRBn;9hYdSVz{9<=qo^$}P
z-2w?n@h`_2N4KF4O|iLwgL?IKvl*O2|Dr(LtWD?CR8FwsU^u;8;nn9T@*X!%)qx{i
z7G=R_@bepG%Xqwj)v!mjm&!*eDQP5qz3!Ob>NQJ+OZ6&7WHhH=noI{wcK^!pw5_w@
zLD^X^gV{Kxx$c)9TFROZ_+3`kL+V%5Y)Z|CX4zMQbW-9;S55V8H*4_gHke)99Bkd0
zBZm<IrQrrv&*4u{6gLUNhlF7wLR~kqnFKy)nM@_^f-BDpSlBt%ivIImMD?|YXx*U$
z4_*L+t8o>JqSsY7f+hz-qihxzpWAraq@_BO1EBV+5bIYOVJ)~6xIG*NI@Qc`bor?z
z0(^=eqCdX}Z~Jo$HQuQ4&GGWxsx30lI5o<`I}zmJF&bfy+}=`Evbel)oWbu~N<Da~
z^4?hIM}qu4LA~bL3iv!>`GE8K1qVrzGk`VM+!BIvRR?XN;F^_*JD-Z|xF+(mMzi$K
zSB@a@9YuEY{yUDeq&_0XTsfi=3-l2{>Ek?s&LP=)(Xif~NPnQ8F#E&H<fB|?XeGXo
z_$xkQFsMwylzd1Az-DFL&W3u_4WbS<3byx(XO>1DluLIk!%WvLWm*8VFUL;y#?9(@
z`XsOJ%nvo!-S2G=xI@7PMxS`$p-Te18?}0i;Xg5{ZYo87{(5-4(E56Kkot*=f_RQ<
z^TK7!OeG@eiI|QMw=~S!i1#5&D0?kt()W~5coiIfXu;=fKeDHFUDFxobyzONd!f7l
z@&rB|&mGPLkgUy%^T%lLoHIR-N?YwtoQam}9-Q&_bdq_Aao4zdQeItdl8LAR_FCF`
zt~2c)NM}}(KoX@P?9M>o^1=dVxp0L=|2zTnJZz%ypsy{876V0LaB~W|0w&|^B|?-%
z(tD?lXor~B9S-rp2_@CQccwLsaDDE%Y?HQDud6yP%^r9tgp!Ac``iI*!Cx7M#h8)p
zc|ZKjCT}lZ_+cM))A&31Grt7>@R}ds?5!tqC}WMCJ=#$d{C?3()jq-?J5bWbbAK_t
z!geP8y7cQFaz92Hfc-G6I>Uav+_gy>FZnV(HCn4BW;VqY(UQ+jcI5E|O&Ky*3>?*3
zEsh)0l46=vTW8ATW81Lyy&=#Vj5i4T2ug3xOf?LW)TtdB6~S8WkmgC_*JxGhls+Wh
z8z|}OFWyGX8g5kSY0w)F$FlvrvT{$6t6F?YQN0zuBQ5;I=c7@|Cb-#_VLppC5+8f;
zO4vz7Y*XA|PsL@T;y0AXS#BM0cVweaS=I;ihWxznSDSr1A7<p1&W9-Ygu3oC)E40g
zi@c+ySIDC5E8)~LL|709Q1=@2R4f?rdxDK0=PCUOm<&_&z|w$6Eae<cDY88W0{SRF
zla0f$wuthPz|?CM^}9EsJ&QlCtoEnU?8-)h^q;gN$xX!iVd+NWBTv}T&3g`Ds{ONQ
zkf6IAzEPy+aS`}6E4(8gS`C)^@4Td*9TO!K?C%@_(s~X#IJj>@H-|iFlpN^jQ4|mE
z3~Beu&(C+E;Mju3+4={#nQ0rzVlROH5?p!}DmkiIYp&J&MN}Ke3&%^X!%Br=M7|Pn
z_CTM~)`f1As`w@#TYUSf?*IPum#<iQ`s<5o(Fj}L7j=F3Bl5UmexVU4Oyimp()gRU
zZsnshhWuhYA}`WQ=UCjTfHViJ`c66%)-#naSVd0VR+v|IQj+_)L61_G$4e@pZJPsU
z`m@TkvL03ab$l03r@n?v52q8Stki+Qv(}uh71?YKX*?qo90Jc?A0i;e?|=pr!<_U6
zXo)x_w90-MB<#N0>5?#pj5}$-b`O~BBVC(*TOMVjY!G~LvbP)=Fgi+0Vp094jO;gc
z#owl9n=#;{7&DTX>;(*5aJ7dtcy}P3lWiYDeh?{^`5*j_>eVd4z_>X#1c3#E0ei0d
z1g_p&!9oID*<^}`7qvIbr3S-lnMUimW?|2XW?7_>7_aJTJrq(o9oQFMT$ugKkhJjO
z1g(6;@8N3xD_1t4woCiW4m$2m(TFRw8)}~i8RP-YruW;z?A<RqDR7tt0KWyE0OTWb
z7B64~Hw96W>zCX>6QI033LSwL4keeB!P4|TkmjEp;l$%`k8=rW?1OQY)ZvoQ42+-j
z?CdC%W<Z$v*lhDD%>W~a7rMsDE!7NxxmPpvTv&i=iMB=ApX^ni%5$Cv;uM3EKL|>g
z5R->L2sisyexHJOOH)91Ws%Qp<ojzbKB$_qJ0$M-(rl!R<##pjY{D5ODN4wuOx&)n
zjWLZ4D_p0j2uOmC<6H$~TM`^T=5J0Od6k6}q0NSJLUa+%o!tt~gWX)0xQrQ7Pdj5q
zVmz&AZ(@vXW3Q*dA!)Mya2zNg;QSI6xzKT%%c%>k)Y2}AhDhzU(AJT5N6^kk`$Az@
zy4$Qv<)peGc>VZEHI{|B{*{>A{QTa>mDl-WU2f(kM*<c*hn;B<bTAp2(tj2vbM=F*
z4+J0>Kj2{~vqA8FS^bN8UkICv=<Q)0{wX6YUG)sCgwC*&Zc9X?$<$7oq8je+XQ&{$
z-c}qxF;K!0zt<T=Q5G*9)SZ5?f$I_WJiyGXsavk}!j%Q)d?08694ggxJ^0tJPbhr=
zV{cJsriGsAPgb6MZvyN@Po2I>!$x=@FsI62IY9hLYqy~1Bv_wi=CH+9JtSFw4vbU0
zVY~pw$bNM%*n5igiM8c0f<p}frhRis=D@pjb=B?>bkwZdB`WZk7%}pEZ>Sw@7w=5{
zlv0ZUip!+5a`bphVpr_TOHC<;c()*ra>wv$*0Emv1<uo3C_a}tK*!&L>=d<?bDWt<
zm);O*%9;N|wkk;ZzT<Z+?f3v*zA#=6pVNUWUZ{Q~Phed$<VR{78c%2Fk+*PqgqoYO
zv?Cg%cU=F^qx*8&Ro3d{P0^Fp>f{xx$?Z^ry+p>#nl<^dR)8uSVQpyrXhs(Nr{1(Q
z%#fD&kk=y|&hU!5pLOSjz^Kl0NAq^*@i4r&ADNB-MG;`=7oi?WRr3ETh(ZGaC1A5p
zfqUaGptkY57if}Ks*Y;Z7dw$D4j6AVw=NG*^Ih}<Cfm9O>pvkX1E(hk29EWJWJElR
zAXmNLd~5Cz)h4X30x8zXz=!R9@Mvgm4Mu22vIl)8J*l6bzyIOyEC2o1l}{=}Q9=;}
zgjL(g<MWbFym<~kl!)*$6fPjMn**&ba#Q8Q7-ub&9-iu2i+)^9t7?)`ToAkr5Gnp4
ztUi>z4)T?q6Sd_JbPGqGRzgvkAn(1eYT4(-iQ~nrnECs@mApFW6f1!99hO+2caCcx
z4fxMymXF|l>(b%E%{SgH&4W_q02117Mb=-`T~U+eZn1t<^7}(p9vLNl<jNbU$q_0L
z*t$G6@*AH^jF9`3N-il2MA31}-$FS1znwR6;!!oMY+!=CaiOD@gdU~KxwQiFz1#3$
z%P9WdPgshq;`gP0>Fd%zpvxkQ@emrJsSq^cyA{KlT4<BvnCOY?A`tiIPhss%ai>)<
z@V(so_YTgCclL+Yn8!&jmv3)V+BcPloxB-Q9kg@8ZRu8mkH=7{!4<397-c}`+0WV>
z;C(CEa!2C3E4!OY>WA6xP7hwR9@uOK$gOcS47+g8x%qZ=%82gS=u_jor?=tbP9wTJ
z>#-b0C<!Hp_Xf6o_oyML?&y*ibVq0B%JOT2BjbLmQH0=$=V{d5Dt+SXilTHmY%MO2
zjk-F>^sxsx3Ex3;!3U@f;4CifaTC5)sH`&O@sGjBvZ((CyiZ>rd&boe2A)N}G~`g9
z{QPd3&xB|`VLpl=&hR{e(4RYFctrdld=T;IaImv!CN;ump)dP^x76BOIR|=wfnoKk
zF799MmB=ffrY*_g>`QZX_vi7Ak}?mm!tu0X(|MI>Y!(S&oHm(oW$XH|IUTRw94QQd
z*=SfU8Kt5YW4Ne!5iYD4NoWL>vQ@71dCVyeFUc<pE?5WA;=I+6DgM_3rYGZaU{Gq6
zKIXzm9lz_t=mX_SJZc<%)^wns<@~_&gYD9O2O)1p1V4(RaQooCfG~)n7lMD+FVdn+
z&i{z^Q`NtJOYi?)y_DW-Ijk32Ms*Uee$fld++EyAjJnSjMxv_ri9=yzdFuxhZHt+t
z7vmo`WF61EM0HnK&gk#+2V(GR^r)K=fwRJqLxz(uKkswMbVge~fA<x&CY8aNh`ig^
zReF%d$i$RM;IIVL6qMtjV;qmACcO9+!^U}ZV5mNji4WddbdJ%ji(qU=QdOekPU|=#
zOMEl|;dXK$?5v5Tks^v6glWOc5>zbg3j3X9P~a_cCSjT)htRjcj6_&V620xCnv%K>
ztt6M!|3uiDa{Tif#J+dkQrp2b{|{l(@r*G&dG{~G=}vH+k^zUEw$N)86-5!hwTiqg
zByZmt$x+)x+Us~I=5CRHgMW>-qY~4hH^(OcLxJ;ty+$L=Fbho5BrfiFxLz@(>mcAP
zr9;LeNzH9{iHWCF3#&T1)Hy1`Vc)HBrZx*n4{73>lJ`KkjFS^3fhNrQN?!C`(i=Bm
z{UC>T9m-iTXcwk(Wp{*ngdQ-_Bm@t%+8CGX`{5@?;?yIEs3xe4anZ#IoZP*6PE%|&
z;qxlS(GY~h;*~TZ=hKJXlUCMKJ)FQm{NPdXe}44fsy*+0D`8Z0mr9uZV-<jas>xuY
zUOc$>T9twEUhF4Ku~4V0&<m{W^_19>ZYZL`YJ<5=m)v|r*=;Fu4{=H2FZqm<Bp^Q2
zl}ujF1j>ihh-IBhR-gzKMUQPbOdUGu8C6B=Qf;~v+wxbT@VAPmv^Fx}^ohbD;cviJ
zSd=cN6;dI*S1Hc=@`M{gq@64k0U+Sb2Gkot-dsToJyd!&K`YHz5Sykb>OzS59z17x
zH0Zo^!PUv=Xw;FZbu>JtmgMX>t#nYJdACaBfSuL$HOuWm0Scz9)lWtH#LDkbSKboT
z6@h<)Dw}eCfXUg6FL9siQXL4>R@|Dlu=qiWhh@)l4sA;or-na=tl~+Jca1+T|7@Dq
z9#L`OS*6h})buOX<2C+z*Kd<96%sF7vl0654clt-KtJU`0ldT8YKNPJ3gRj!ffrlv
ze&(Wu3mj}<w^><Qv@C&r6lf6$nmX$zDlUj1!-WzEb)3;w<w@#8UUiCVbi&}ijI<-u
zwu!p#ni$PD-yNO&)C6hvl3M-K`T?^0z)LC&U`sU8>2}DL2dK`+1-;*`uqw&B)MRI*
z{&Cmw|NZ-c)wvY8Y8_}e3{WyVI;~eq+AW1=8A$PPQ(|PK>b!CoZ!*aKK+xIR=AI2%
z35zj#G9JVY3qB{Y$mNXphXZSPWVMTZhEkg7^sJm~BDKOWJo+XQ7C2}(6y|=4+}5Ry
z$~?P!z#2H@h(Dej+JV?*sV!u7)2beQ3KVINzU|9>J;d64J1%Hl&q>JN>3V>N?_|d!
zwbgI*(rFjNAi?5G6XAy(Bn!#fabU-e$o{jo#23I#a$QXyaDD)+Fq0g>L6H96O<S<>
zJ08`x6q&T|tieD3WfFO1uyFh!-b+7O+)*ht4~AzDsVaj#9{COtr-89=nYg(j@#75=
zb8U|}8inrf!4WUM(Zjkth^QIlM}<DX<FJ7Izv(<F9|>9Sq8^y_EognHr;izqyy3Jh
zc7JgN(j>w%z2QWKp!KTB+*!au6_t23Icx9*2pZn$(QB1xQ;(uPA8n1^3iRSc;d2ww
ztcm#f^WxD@eb&Wx+B|`*Fr;1C__ej=Gaq^|%E{Lo5~7cCZm$LA0+wlm!?PKvb>iwD
zbmZz!-)qbmg{2y^(-N#!d`*EKb^=v<H>0`pUblqBc(RQ!SX*C@2xNY4I4gxX^L%F-
zZ09u0Zl^eX&y}+7b}f+uDQ(8Kq_LVtQIczlZ@O!;5jd|OXV@V;(^xQnt~v0*Bex_z
zg8a9|dKRsTZg9B*jCpw{-UtII_vy1}(Dy}uyvyf$@5xO+=X7O9K*wf?0W9*~4eNMD
z!Hv-)i~QD?KJ=;}92J*8*pGb3Stnai?JUQWDFPC{>Aohwyl(iPB1-rBvh4e|EH5Jb
zUhNR%N#j36oei_|912^})d%7595F)7j4n)&!R<;_6o&hF?4oXiS*C8t*v+;Qjs+Mj
zXZ)vhx{8XV7hK!I7r14y%&89{=^JN(wcmSp`Nq-ivorXa$EUK^Ur{G5A{iz8xxFN2
zPRL_D>`pwDcIVxQqoswawfiolCyPCw0<I+R^V1n6%r_$zyIz``2&M$V#aU^czXC%~
zx!x_^GV(momB}8RkH9Dog`n)K|7S%1brS0TIwJp7lnw8yzD&P*J;iDYi#p1~PPZ3k
zp{j}r6SXg89~kq7{35d&iRshLkd-23r)BEHQzuR>TqWB_Br+!T4qP5$(QcqmIoE27
zHa~;JCG23xAd~Y#to_uk<zQ^HU1M?MV=P%mQ~TxFmcQQH<(G0wDBjBa3a?JXCMohr
zF%xcbc1M_ytp!2gCPs|7Gh=qtgcLY4RBrY1G^6Uq75{e83Hdjpb|Iiw<fTwoj=q38
zBFJ0&_XnPo<E+x9aw4L?hXvekwbMAp4a)cbTkkIm{Pp{m`$KAJ%a!h2u-`;P11I^~
z{Trj|&c=52@fc4!B_9wt|HZh`zHr{Ds%P@kemj_s8Ew)V5~J1-!&n=rfTk+89T$>M
zUW@n)Z2q}~u9fPP+Cjj*j1x5ircYwLo<VC7?1V~tPL<Xhds}z7$>J{WUKYZ<>(6t_
zupL$w>aJJw=;xf_sU$zHU|PpSvM^?D!k_&>MA5w}o17<%gKG<i59{rvYXb))VB>Mk
zzM+TX%Xfiz3aL322xABC;i^(Ba7ylHy;Z*V&+r1j4^QRm@V2>6Il~v$-DVheD`KX{
z1DVJ)hQTy*w)MG<vf;j&xi>Y<sfEPn&CvlW;c#0^nebO@8r|D8MDa;fP(v`)yy=A+
z_pnqYVRlR=ozV68?iJWn@};bDp_iu%xx@HRZPl~jo{0^jbW4?)5|p@SWAkzH2$$<r
za7KSkeCZ99=u=J>Mb|(w7gpZYm4-MRH1jbISng6&jf&g6@>GFgv=sT-zDIe^Gn!JB
zx3v~BlkW!+i$;<O^q)8t%fLbR)g7!w@oNa#D^2-=1I>=X=Z=H@#=4@&fO)o1okQLy
zm~+#v#i+{WL`RSAqNdtf!ejOXJDS$r&_cT7z#;!1^HQ<zl$}EV=ItnaSiT?CZz2Yq
z+Tu8m1BrDrmY!$&yg3P~#rwy{GT^hvEI-CWEjAWr9`mlu(XL2DWq)Nn+P%EBs=TP`
zfFGxhx-sE&it+R`VzIW?O72p;!IgoUv9h_IJy1s`@OQk>>&$(UEfihsqXJH|XVA^|
zzU*s2t^YIr=CAoZii~~TbE4ny)MwVKDPP_~m&iv|;RIjz8EVbO=XYgJKiC$GWxR{_
zL|is3<w!SoHYCKs`J0YBw8(_v>&>znJX1iPYB;j4<~&7OQBnFBVpb;|7D#oI6C~YY
z*vV(bcy_zZICw&e7V{0@HY+t!8>qa%%`>IvK6;w(9Y48S$$}fgNy*5-s6z1a#aW#+
zeDFX^<5?O6C%-PODw+WIFgDq~^06iQ!~t9pIJ$x6H9(mT0B5}V8R%rx;{fO72p)P^
z*5t=pG(7XY(#~9}wB=;4f1xpyZ+xMd#P2&`A9=Wk$JS;CO$ZSf{6Ir7bSQJa>gfsj
zyDk2sLAK0Yo6NNDor&VhHLL)aks$HQ!iksiKNL5gZ>EKkFDxKsGk~?~%n78WF)cM_
z#Q|mqCW3Ht9?Bvp(Zx(e{@$=$l|1JhYDsYf%?FPif7j@yEE;OkPBHMi+NjAp)SD+E
z$)O5`e%o=Np?I#vjjvawt!v9E0L+yVAc(-w9dKIN1^-CeZn}aIEcIDc43_fDX~&fm
zv@`|%48@n<c<+Ic(9}@eqVs&ZFDwuwVX8dbEFD%SqqKPOp#IE}_Irqp(JzY+raaV8
zfbAFzI-Km`Q$sY6(!5K~em$0KvYn~|3=(XJ=EEr+udCuuSsZ`#B3sD4SUC9BJI-PQ
zt6SIkN`Ag#S~6*YV=Nd)U=1;g?;-)->!hBeo*01{Oakq3x`P>!bdiyb?#mwKMg-k;
z<*-Y-F^Co5nGDGq{PRp&zCV-F6~CEl-(AkL-U3@K`RmFFlohY$9`H#nu4rd)%lYm&
zAAD_Id*78$AL%_jsy2KpW<E?-p%s&QkH-$umsbh`o%NV^%@G^SwD)9WdcR{y1?!1-
zoq>xs@yN4Nv`x^Uql&O~eOE$cQUeU)3b^NmyW@g<%)@|d<*e0*992$lneXCAm$bQw
zxY#852!1hk6;J2zfY0o`@;?OgYdvF=#)!;7`%dRgzo3A}Ipk(fd6|z=_?P~2+4*KU
z1LSbP!%#-Wpp$%L<csOKZMKmo^rvTv-J&E@&f2HR;<v2GaU-a9aM0Ye@rxG;x6dk;
z6q7|tZ+ir92F}^L)ws~3ky}t0p6ZLSPU^N?n07aXqu9|j{j)^~`IrT8bAc~vH~XAQ
z#Phh^*eB&idOTGVz8+!brHQE6B&E{8ByG2a`CU@v`8RrtE`KxMc4eQ_bD7emM1;>K
zhXrv0z)@7n<RL<TLL;tyXLjtB1?3uI;zcLe8*|UqmQSv=uTKq1FC1m>#MGr>uA<z6
zn}#}zZ43|L7e3EUr-miiE0T6cE|k1lPf<yHoPn^Q??xqUj;#qAOYWAl)=4}Wr|4tT
z05XT`d$z{j3GYousW<{kD0)&eFMn~48D|B+!->u;^_NX927>0h(pfmfpw?8RI$q*|
zw)KYM`i_1=0pZl91<N$ym+vz3CA#c?estTbwZ6BNW43*b#};o;__BfH!MVctLBP|%
z4kg_?sVc8v--;k$1ce-uWMT-ZU3H*{z&vSqWIkjM_f*HEVQNMfKH7`2;mA|d@jqc_
z<d#{1NM_c3Ww%t!IdXyVE)?;lyG_SK|3?;QJzBzxFo_NP5qfA^s%R|Mgf9`jF*&k7
zfcw7U-l=;HmkR*!G0ly>zD51RvS{xX)s^|u0hm>!+jYPp))x|P`w9Oli2T0(yT8UH
z1KhA)6@B6@iXSv^ydzpw=3T6Ls*-Y-iLSg2d#e&oP-qkjQ35YdjNgXdZFGaocSgc8
zngbnoaG$t$-D8ev_StO!)>GkzH2GMblW^$TfxW(pjln7@?GCf5G{}WdOT#a;rlpwR
zv3MW<((>1@UlB-mgnK9nQ@l~W^2y@uRkd4)K3TBRYxywHa9UUw#Ym{8J?WR-|6|Km
ze=qEVzOEF*wQrzD#~2k{##EWC%SPpj(}4W5a0^8ez8lKXHZ+*&oO^cp;$l`HcBt{P
z*Id@+*V8xT&26$zADVubWZ8F_twh+~<9FW9tR;SK;Qc@grP9}bq)9R$E8)c|;}_(J
z^k4|Sa@$@KG~*5Rm(a{zf1PlnPq|RaXbhj-lS=MD1TA#?30f+P`aYw~Z&BV4l-pqP
zNe|42Y4Rr6J*il3=yai~?EU)K!N|bt*sAHyxxvU4t=jKm2el`a6R{KPN(hpOqS3-N
zR}OY`v*ff~Q$<7q|1M!+I>ROk0ye<F*#ymciw+JgHm+M2<y~(M8oF94TFN=WO<;v$
z+c`wqsN3Kg@#lYB;MBgSLA~G7pwfOl6UE+mw{E<$EwyluWYta)Qf-&zoamh0A4cM#
zvu-HL?4+1SUTja`DIZ?Akwd=<2g>KS6oF(BC8yQ&@gO~%b9tOw!*jw*?0vmlFFOYi
zK6B}KUliptu)Vkc#|$M>b)n`TN9?h%Z^wqaN%gh4nLP#1LvMNhF1mD7gT=8LFZF4r
zA(ZidK(QWbRz-_xH+7;PkR2<HDaLxinY*YCp4T~TE;%KsE|vPkK=#{QW0L`VD)yAK
zoS<V>p}15nbzM^&#6o-N04xkbpeX=d<O=ik8Mud(;UDPN*L766C4wAi=YY>i@6Ikk
zP)gM^hFvLR&S(nOTV4?ax89u^T9l@WWry;P^r>enxqs19nhNCW29xmZn7s%DSNQ7f
zPaUKJ>+iv9>+egnwCTUp%%8q7DY(NTf4L5oYhW#}F`&hGSV#>QA$c;jGyB{Wn}_7C
z4XngRtmwNFJ#`6&W2LsLGeHwEuM)w-Oc*yXCVs5<uRm=0n;fXJaJWrV)RFgI;Ah;h
zt_H5mzSE47&lTQjl7`(9v1BVooMk*}3d)#9G`jMQcHU8N_W%c4`_pd!`c^CcoU7fr
z>%JMl-TfTR9y!ASm5R3xit9fY)J$YZOrA`Y2vc%`UC2r*z%@e5=9V#O>fdf_V+4l$
z6tb7z?WSdTolNDe2t1YE57O_dPG>TgmoZT4L)E)~M^8a?hxRv7EPk2;Zb6U=ip)k7
zfAdM7UrAW_`eD9)>&ae#%)kTkmpQQgYXpr1vgx|Gq?<4<Ft<u)RNOl(85XW2p@VBn
zZ`W1TtOlZ~e|a+WDu0K8<d)xQ-7>?kuGzDgTqB9~0app3a8H0Q&cU)Yd&Qgx><0>b
z0Uj)B^`}&qYB85a_vE)JFhmXyAF**U@<$*1>nE$u{9SO7sBTWV&>$ao%!SaG#(K2;
zy!}Q&JPd~+-Nkc?X>12zy$7&%65=OR>g$V3w?5PM-=vh1Js?3?2U!!<e7L@lY|yU5
znW)(QK)Z@d4?O^nm;t}w4ebkf+)13IW$VLv@CZOk{AX^Ke)=u+v)rqM)e$7)iAc*F
z@`hen7_K<HAEtKf3w7n|+xdd|N(7z_ug~r?r|-x7zenVKbWHk>_tF&c<2DVU1Dyxi
z!q;j+zubk48iX;{C&;4wEPVzGPZ#c$`PZc$%L?*-#oda0yN{?SiLU_`naSKBcso*d
z#yERhf?IDM@LqBFw|010=P%$jZyxQ0@6F?%*fjdy*PHSysLhyodO3Xi&)NRhLCWq0
zr@g;@1Sxy>M<80tZfW5>-OmAlq8XXwBUEB>;s4SwEnD~v#IoC4!Q#ZY;3JO`NEP+B
zS5ia|>_zfMQUCJ!S~hilZ@1CJ3CpZ~WqIK!dun<1VD?A_98wVXXIYz)zYJzS3{bCU
z;gv8CO;mo&`pvt?c!YHzY8B`_RQ{#rLbKc}csD>{^9)i~MSTDUD*R`!n=m;T7qpKK
zjc0iJpS<|%TU2{Se3l!7U=s!nmMF^kQW~F{FO&(+pZ~e@`1LmpRy@=GOlSuJoGtu#
zD5nDda7wS|?f!o!b%CaPyNxFtEce=@%bE(aXTSKC#1V_a+8IX&=s=G&vXbzV+mtMc
zDOvjm5PKDyC9V7o+~V+*0xoT%Akfxbe@ay7+AT^4dw=eQEParh=D$XeJKGNxGMxj!
zkmhw7WCx%6qS!}%TGRc_@6x~aN*m&CegXc}UN0n<vV^MrGwG-W;^(L86z3q5<RI6d
zUh#E^+kOEkSNn}cr(z}`->^R2r9C>{O<ipA&?9{`7}NdtS~q<1E3S~my@#WY{}Mv4
zs0K!z>`Q6K|2@My!#~Sy(U_-yJjM~pcwxxK|BiA|tB)bh7j)8GHuNeV6wqZ>gjW8)
zKjj1#YWTTQn<=DL-r+-Di<RSlC6xWotzTeb<eq8obD8Slz83NxgdqO@_y4=z0v}{w
zHr;&=gXsH_5`G=|vQ8uL2fTmvxf^zKP0;@yKsW6cBrD#()VcE)@lET}Z(Eq1HqAeF
HKIH!Zd}ctF

literal 0
HcmV?d00001

diff --git a/studio/frontend/public/unsloth-beta-white.png b/studio/frontend/public/unsloth-beta-white.png
new file mode 100644
index 0000000000000000000000000000000000000000..be689ff874c4f2daa56ca32f4ec7d0a5a3483c98
GIT binary patch
literal 156421
zcmZU*c{r5q|32QkMM>I}HA@MpB-_}Rk}ZZxh{@IrViwuP79K5H7;DzCBqDo;nX#+X
z7|W0_%#0byz6{B3EWg|H{(L{5$Mc)x=-}`N*L~fu>wKNB^E|I!-7+`gJuG~9&z?QJ
zCdN1J?Add;b<du?Du=j$pZLXTw*X(b9~s;F?AgP^xAX5md(zSc_w0Q@G`Vr@Zb14B
z^I-aYm-_AbIUiRzZc_U!H~0DM=lmMks_$(M7}P)T?At~r!38*S4)uNB2FI_ZIb1V1
zb>q0nghaiR?@8k_Mw`}lpSq9hSPzZHq-_OGySjCz2X>^&DT(}{4GP0c_8bv)8}tV!
z!r(9~WJw7~v6qzMAyi!kKW}APczgDbpb6jkZi#}=f|k?vG7$NKj$nm62Wm}OJrOwz
zTAPMTVdLJ+_xTO@e6%+F<M5y+?CZx_9neR3a<2I$znfFBJy&Sf@ZQS>26wQ<OAe4!
z>z3Mqb2JdV_kj;bxs@(T4UxmQ($TgI_Yx$LiY6C!3*#9G0@D@w0!KTVpI5`SFltnV
z3>`ZL?9~5#oCFn~;XZx$qG~q^UczQLMG46aD+@NeU;Uw``eezG)b_9rHiZ%R8uS_d
zG|E3Cn0Cyiz7HQ?kT`#(HXN~4Q*);UPS?Ml8jesUR(yt5oq$8OzF{2d4^eqx^?e@x
zjv_^m?=oQTrB{N&8eQSY96VjNx(u0vL%*Q7`qPV%t!OmTa}WpRv_|TT&JM~!&CVVn
zXkFd=e@oo6=hI8Q3p_{OD$Ui?F*b07QL^R|z6Sl{*_0XB3U6tzP4F8|l2vq#h)Q+M
zHJQ!5x4i^wF^iwSVqt~CiqwteH%Oa@QX<2iuUw7=HyW(kK|&?4IL9CFzauxU;aSrD
zH%mmiQFv>tR4tf_DxZz4G_Yp8p`aA-y&{FeB*X;5gGQwd{77TBqAz0Em9`9OFO|hA
zuP@I*HYE{^wkLH1B=-Mv!GE22d5d3gj~i<x(xjCx6rXwpnx`49#4KoSzd9R<??w{f
zPbhQ+94ohf0CvwZ<p^ZFv14&0rAZ@}B2yj*dA67S-HOnSy$(lWqgpK9BNGsjQSM4s
zRuu1a3JzV5M<-C|yg0g(qH7Lpa0cOFA2zNq$FQCKN*g#E_<Q$SvZHVDe~9g#<2)Cs
zicY{odt90|@V&gLg@yC>a!5<Cv{F-5_k;KDhN!DWsic%tMekvnho34x;YI5!Sg<&G
zY}uGDBixd76c*|EJg4uO-w%}M@c4;#R0+3dr8O9bemcTcL3bK;plVoJ;jbv=q4eEM
zr^iZ@s3e4uTS5&2Y6A(jwDExt?-K|iWjHX9H%zs7{<&Ry^m(!`e>ci6sq-j~vUQ5$
z7lfi!NbO3|N`ozTDapz-v|qL51EwJ1<{gXz%f;WW*lSz&d-%-n)N_!FPvlOw7!qX4
z7*YjIA>~_e^}MFL>J(B-)d_YR2peKzSau0(4=*fopl$U;mT7O4b`n3dzpALV9QhCb
zu0?S1WnaFPM`Zad2})|(lg#3JPc?f<YYMq}6N=z9!)g^$O5MgvQC+?*?eTY~Sy%2^
z@LS&8OIy?frPq*>E~&<1GEC>!SD(>0$h0pEO&D8*O6Rq9%_+9eau-~1?Ojz6LE4~;
zDD)CS1=P|8(ma-=A_CG1Sk@&Z@pYc;*j@6U&b>KsvTn4YsK^DGh$Gn9)3|^oL_zt5
zn{Ay&%wtq)TNM`SuUZ!-j0m>GMQN%~z52)k@k4iUxNwABv0b$uh-fspkM=1v?vg1u
zPR{=c;_=|4cx<Jsw}eP-UR##VbOX$#2UbFDq1zNM;&N>=(h3D?>6#86fom9Wi3pJZ
zgU~f<((E*>!-fz0-$AwnBt6=_W9tIhmybVcZM_7CwU?4S!WBBsK$Y9nqBW(^h2?4I
zQe_p<X$r1|aYDl^X&dc`g}#9E*h-sii?|YB36aJOu(sjN0K5LXXbG%v%QAxbdp(@^
z+7hPE`PRq#o~Q}KciGj_T2bDLR@ig}KM}B>YepQ^wIrlTZ46Hd+hj~zge*cbZaPJI
zU1wfBM%wi<Z+Sh#3F6IbR>oDNi?~dnF!nuL%@(!Pz-Z|5Rnlbz(q(1C3QYTX6FjTu
zUQSf9T3-7L(U-Eg(fzoPNEqMV9tUt7xat&-ZLqEZNUS?a$p9o9%j?Le>G0E>wA|_U
zWww%L-a;@&C*Xs6U6*ipnn3zG-Ky1faH5}9r=O8Q0nQeNBH!4xy3zQ!tiC7LcDqVT
zZQnl^sV9D-^+AvN{OqMO(DuvF^l$NK)#y~WC?%$#66GFPg*0|$d@t>n**(i64mk_X
zPoG1>z;UAP;fR3i@Ws|Y`)I$ba}VIciLKl)C?AY(IfnJN$`Ol47bSvB=?@h<T!&#G
zI~pDr#4IQaEPw_NPrZoV{H`JOlB?nyr{$^2Ki9Hn@l^I@!#X}|2X&Hh!g<YWw&3M+
zw#~LbEu!raurZS~r>Lo@yQ1lW)7GVHTGS(VB(VHN)>mRFZ09yL)TAC^)%t2I_`VE#
zS!yqBwPA8PqR-+e>?tRa4|uZ~zrNa<h{oF$`{h&UeZ`d0q+u<HO2d!~mQ8oktR?t?
z@F@47sc*>6?afAFB45Wz%C5tp7kJ1;;X%ful}_RFl*F#tXaydOLap#{yNaPwltO|O
zSVhqiGcxj||G}y0)|FTKGO=LVk7hlPd!zSW+V4<p9@}9~lEyIxQ4i#c%{+(;pTILX
zd>Qw#AI<Yc+G`SAODYY1aM{*BD6t>yv2J`=TjRi3n^>5}W`-8Hu(w$~r(d~%|I|hw
z^xCy$#`+f`PYkaHmli8f%@cc-O}e9$`_wvSNhwm%3bidMYw^<63Wx?Z--V{fEa8lE
zkbar?1};yL0gdnBh{1ETKhoqp(VPe2#bJLnaN$1t>6gP{mB_Naw0_Rp8!7Nw^{f$2
zZI^3BCj{;o+ZAbU&u%L$4edtzqGtN*kkD?vB4XWQnXbn%)+fQ?^Q_Zi-~aQ^^+`Ot
z#oxs8x)e4h2!Io_$*Vn7(N(-zp?36vtx}7ilF1$YM-LD>-WbD*&p%a^?Lm*~hQ!Fy
z>~03IO|twrw5|c6SPw)%+ewH7T*8GtZrqSP&syjK_Ei+Ccnu`m0M^W`GpV3EzZb>P
zFXpHwQcA5!TJO!qS_pU%N7M#mfnnzun&UGpznxwptV{J}BaVhnooL_vZuefdJ!0C-
zTBwLBMtkMjdzVojt9LwT?z+=5<DTnPJ!qX3d7me|c-!b|>=G!|NdmH=q#ci$Rk<36
znA=DDlTQX3)1|8hRNcSWUJ@d#eEDz${S<A3u*!M+)l1S6*2WFvoAY@jI~;NbbOD~#
z;>zwe2ltxhHH`V$))I=U*-N6A#KIPO<u|G-J_gSwXe*>kr5*U^MqW(=&S}3o`rB!~
zs7f;V%n&Snl%y7Hf_@!G5=<?RR~-7?yMHf@oZGKwRy5%k4plTD_=h9PfTansA_)8_
zOOGdrAKDkS4M0Hy4^yyc9YQ;YLFu6%Z2>_FY<~pVYNJZ5T^Woy&rn3~CArQ<D!m%%
zM(}An%2*{Yw1V(z1<0$%mjAh?J%(q2edB2^-Jofi@L*ORp#66`Cr#flSfoHBt7+^h
z_4_GX-B@0(((28ctaGLkSnq<yy)?z^z5P`3DH_WxmhF{-JxQi3&Uy}YD&`GF*eIZ8
zYPwx(GvgK!t^@taW`1L%rG)5Y9JoYY+viJj-+>snD6f;Vdp}?P=Su#%{rnccK2_mT
zm^HFkRK`E37#i&x<x>@>{#}r=l1JkD)%zRPzd<11unl|hlhL)P68IXSoFmE|jzF0X
z?z0WcIJ_U{{u%y=GNIfzd%b%=_Wb+L?r{j!&mQMv!H0bk#4PpCB5iM@oFK3}rT(%j
zEi{i3>lXTpV*Cn?N-#<j3*71@sEF8UP(9rT*BP}HaZX9x75`k~9z#)Jv(%f5>mEEQ
z*U%%WlLgt%QMNzH?Fv862&q(YQl9A%u3n>VOP(T=mgQ#Dox%~-f0x&Qm<gv}>;Uc?
z@t37-+c8)U10hdkn@TX4r)U!+^jyEgkfwbI=n5;33m3<+q$(u58+gN+d`?DJ;DYq6
zI_T}I5du~&nK-0q3_)vUN=d$cx=3@OHf=!QuLND@_tf7vb}iKD*||60VjnGP+e9f?
zC&1WepaTWv9bKdp$up^IL+&5LB|x(An3;q@{(+M<R}^A_bASW-KxpDI*i+(!yM&0Q
z9jwOV&wg5N@Eg!I(B@XZ;y#*p-q^BotentZ0W{C$sW6;fLAM>}4Q~0GrRAnxUvyqR
z!{L4fp{T%*K@Z#Xf;z7Z2*kYdCu*JDwKPFn0E8!^9lX9>iSFu(S8o-pAS$pH`wI+R
z>n5EgL=NAVv~<atED}2RH@H<^RdLRWw}kOs`S#?;tR!vrN!QQQ1e<Wgq|Zy8Oh19*
zBrVjc)t_9(o8i-2m*MNkpIg2Hdl{7+aShkxVnS_#E7*QXIKOn+%ccD_)y=VfeANGO
zV94*s7zOh$wSPJ7Vj`bjUb_%^&1<ps5<0GTuOvm4G~<58=cB@ygnuF8>|aZfy^m+o
z_{m%=Ej}x9T)1z#P03H<5WZz`z>(CRGdp~vsrj{0S>jPxlT260!!-V4E@@hlN|y>+
zZ|(u&3F6EA$rbP`??{d`v-27BIY`C`Hx#m$!L)c-<VU!~`B+dI^r@#$%afr&^+gP<
zw_X|#+oHUA2zq=Y^Zc$oJOTw+$yKA0x-VopH_79b!q}m?ZwjVS?aGqiH+RPmly&J*
zgC(_u2x{&iG4i?%?@bBDHXrPud@RP^=3Fdh4&o314?aVVAoe>HC-9S{GoeDaA6ev6
zllaJnF?6Jah;Fm138Vx{ZKb=x+Hd0hnS&c-A2=O=;p$-uC4;h}TT)74<TxJQzx#wl
zs9d*QtR1~n`a$u>NlCf*%35f<7%63Qq#;Sz>~bi@{(_1Ds4C{<JK7^1$PwqraKsOX
zry?!D-<4+Dm;<<5=5`kTc<|1+Gl!oXfl3eq#)enT?>`4rrgQj?SSJux&d18hw}Le?
z+4K8p6QKZ0_#OO->4g7GFBWOjaVAkq;=4=<B~ZU^W=uU{D>9i_%K=VFrlvE<Kk}!B
zmiw-a6f`{K(&d>{R`;$0#)e6<;b(lJ6vmvO?blnX7}CedX#dE}1333);|`NZf%uL_
zRZAFj`o@>8ky|Rit!b7pWQ=r;bji*RKCg3czO-R$eI)%<Pg2@Bh=YX4fD(Hl=Quf{
zZ>sVK$OL;{TOK|~v|kR7?xU8?-X{h^rlCa;!r*G(72Oox(CYnk=S;0#t2+|%3^yTQ
zR8sXgF3wik-LYB06r&Jqn&t*m{Pw5g)L!duFz{DoI+;X4GXM4w$HRuWu6i!%cy@>c
zmLch$<()^)8_%f6o@3PNf$A(A_tFMJO}q-&*MRr?d4Sxr8hmN8T%pEuF{MS}kxL$#
zw2!u`F|X1+==dKb2yW9drEclCP+X{JfkratiAbBR`BBlv^+;lc0*U+a<X0W78~=Lw
z7q8piZU2*)+?Q6IC!8wf?OUfXrUq?4Cka+mw3^j<E{l7<s-k8E(h>r>|0r&RtRnZ)
z;yH^du^5yok8&HfSoo+B&}V(}Z=GAc49%$21I?dld-<gM-N26(rAeLr`{*Mv_w6LG
zkE7=~!df1`kgvlkZQUbr%kFUBe6RGx3=u-zz}6RfiAZBWpyl}#sifTv`10}AcjuA@
zL)P{}+dmu92i08Vq3v6SX<423nyw+%!VwQ6iWR&s)qlapEvlUPk;m@c{u_jz?NN~n
zl@n@N){12>ouVl=F}_ocwbD)j#VM<OLIwE_Zuc>*#$0uUSY@Gp8+B|pzwu$qadlso
z`;hm~{!v3b$glp2E$2y8<{eVt9mN=5Bx|dm&eKpO-58@F&1W0^FBpowaUs${(|dH(
zkL2;BoK6w+sq|`hM@d{N;!?Gy7?rtqJ7XLxxt`wRqscVtJ7b=$eUy|0IP~p;9m)l=
zReoID?>)+e+dA#eXA0#rZ%hB0#*h!~G(CVbnf6t+?M3_4eLvCmG936p`GV;o7_?bV
zH2};D<I{bRZ48pD;(c64y#yD3<6Dbg{v4e=__$>Fn+hXsZWXdqv=#q8^cujSZ~b#R
zx%29eA;Yl?(DqPL%7J+PswU@!S01SKBI4OFS?yTxp$DfK#E)(xcdqUbuO6V<Zgf@-
z1H6lmij$qrWj1^XM@*zCN|%qBpPdw2LS*1%$kJ!z(OIVZaj#Z%rDEmk%z0t^h*M>H
zpgZOaZ=JpY`SV3+IJ-@8?0}`^(#JCu?=~lsw+UO4+ts}csf3mbSPDB8ce!gf)A)S-
zLL^SK`j0o!C_}c0iShkwSgqMqgJnGIeGMK847jfmUJnqS`Q(W1zHESm{osQQ6Q>O%
zAY=P`eu>O87!JjZWMD3KymO!4^qmm(s3%G2Kt8wxtl2o>Fs}y+lAgby%D(?=rFmQF
z<i32C^rJnDr=9pO*t|!Zc785%A4wbE2zcFIycdH`XtwrbZRx7wMW6#9mI;sRE0<mD
z`$$CLHT8=TQq3M=E$Plt<{;7xAWp>>_zm|tNr*%_Mt+_lABJ)GTcr1YaG!^_7=woH
zsD>jP^T>uxUAaFMD^*;k5bEKG0w?^47ZPB)zjT<*pITOwq;4#2Ul2N%d<Z$5+y+1d
z@IZAI6YJ`&YU8!~?=vD4EHwB?TT)6Pzp_Y;!(YdX@wU*YTDQEbk)V%nyvol1>u7yS
zxeys2eLpZTqudSwI|?#15{5>DexMxEgkRd_@WFVxT`c#~HpC{w5%ZnD-#eWbVZsBr
zaHmtx!7ZsKnq^}9ab1s;1OLE(c0S}55mK<zdY&76_kLr#Uu$Zy-4RrR?w1xzc<{;5
zh(Yn87!%dG@q541#PCTXWUiy45PA2`9$DOb6n6Vvo2z(o@8MpmRxGoMT~0O14D(_T
z%aHSE1!=){lFosbuDkbl|8?73yM|wX<_!U;@(ouQyC+F`V<s(77?DrrAMV>=e13_q
z@>G6lbvaLt<?(he&0CVjI*6;*UYCkxU;A0$EmN3RjMZ9GkyDmCVVh?b$8OQ+(gRhM
z4CsL>JxBH@bZCZ#%6@)yT~9>Q_0_@^)mU&nI18wKdW(tA&@Hxidf(l|YI{RfJZp`3
zkQll&&ft4M3ovcxJO0mR@K>neg~*iAOINa_CY>aGQWVB+L4DuGNsm3DqW0R(VSd8;
zf9KL}otvIZK1Jpl6rRx|c*Y?DfgwiW$CIuNk@vqyt6MJSaqLN0?{{R77@2Fh3m3zn
zz^ou+n_#RRwhA?2^`(K(6S(sD31YEfSvnTuP<mHly8iUzZetL<4}b%OVbXvngfTZ6
z;7m!t>i)?RsV~BKomUUP1^QFa!gpfVdba>7^<?ytd|7huIZ2-<3S&p0mF>pEk$jE|
z&TzF*xvv%3Eg!c=$vQVvF2ZYgVF}B^>i`?jspp01CqKC6t{M)l4o4u^iz*9!Z<JoC
z5FKAl7P^!soGIonD|b0l_S;S9HtIqhZ^J^ub*HedRgU`O_OERB6+Bi@ZE_&!BspS5
z_OxyCAy|SA)ev<in0Uqq%fu~gQ4=j1aKoHNR_aw}(4w>C?mGT1e#m9W7AOa8OuGVU
z;voQKhID&HmFn**S_UA;;xyD=+E1Op*@ka`RN=OIXB!+R6^j{^Za~w|_kDuIZ4Prr
zEzYiqz5g3&1;`Q0Y<&rAz_w(9H*~!3sU8S@rF`KsFi*W8GA+WvF_2*OYoO6-iH~-G
zl$}^uB<3>(pPPCd($qS(&6$RSxBYFB#oKDST`FEt6fAE3bBFg_e0dA~1_G=#(9qS<
z#7I3_-Q<UkVV;ISp>UaN<Uw5M9+5+^NK2Om&Q}>SzbB-wSMLCB)DTyH4CWJjzWHff
z$W4I565sK<`}5rb+HNooh4|ur?7u9*eQo*_f%K*88OOo!R5}O`VsE^2<39%Lrp^=n
zd|2};rbgd}I9P9oIBnDBl2HR3Gps>6-6hnuxF`1t=cQkJ7T9WRs}wJiMO)1DEL}n?
zkU|2yHnbPy!-xNj>{6EB&K+_wDv-2s$W*ZAXej2lq~u__9!OF;!H+R!FPB7U8`d(n
zD*4~pZeyI3v4dv?DG}ps{QCW$kM7`JQs&9DuuI6z5iMLm_FZ7!l*MTuHqsyev7iSM
zj7j2y=YiueCHCg0?pwcsxJroJ>p}apSNi=*u8j$Anoe?%XZui=0Zo8)QHU;G!aqp6
znSooDGO<WWE7bKbVfwl)l+A@!`}C%jMvpsCh^sdb{~MwgPiN0usC@iwln2wkMVbhJ
zw!0hB@oJ*fH&KpHEnz&@kN_axSr<MAlO&2?+c}Xz{NyoEp82sICdO-I%9YN24Y}1B
z5FpYxp8XLY<@aMREtiwI*l(pa-oI(qQ1<@%&;B*3{Wz26C{|NtXUvEmC^YhS&HHmU
z+v`15_`5JBpjIqWk-!k~tWc3yls6?-TR&MnmNmF+w*bDRv9H@@K~*AU7a6*B;u$W7
z{3p?-N4RJormKLIk|V`-hvki4@4Npl?NYDVStk|de%$AfKi-z^3w@^nemT`Bi+g$+
zm>QOJVlid*=b|sWz4HL-<$FnM%GWf_=U<i8j#uVo*k3<Gj|3_`tS!w>q(A8-nXAt<
z#w;?y+p=%YMM6Y7bf)f|(d83lL(c?X&JXyaEI|JpFQqoSv^%mp#u$+8#os>c6>TPc
zzNgKk+eKehYWK>&TLLTp1kg-PQks+@>O9<B(NM;i{*CAU<_G-M6NKZ9>(>yA>l>q=
zzV_EWGQ&3By<b-47%;785%)JU&heF)p0iYpm6L(7bz;F9j*KS1rfHc22~vB(p3K~;
zp5=+H^DlC^IY|mC#Fg7902QOY6N~`u$nBM&+T*a{ajTATJ6e|h-x4MBAg;v?`!R`g
zz_a&#afmr-0O?V*70;xv-kYYZG*AOZOOQ-(G9^Ra;{LAV+p2#-^NT3{*K0eIgCaQ<
z7=>n`)cUB~_E&@C_LoU(D1g*}`ON?%NV#}6jd#4_+i`**FeLpCBmM7K+~xd5iV^Zg
z5Cf~_G2B<1SDL2QqO54IqVNH8>Yd4Fx0q<ixifO5$(cfoOfKB$vrln{)T-P@^0be_
zYU{!Bz{IaJInD=ZSyqhYok_1D)byU8RIp_}U}$a)EGHXf!5ha5@|KMpANBOQoV-r^
zRhuOvzv~%XB;JT-1LM;<Xl^~mw<=E6VEIu?d&l$AA!Cp@ng18<JFNc?P|UP)cS{&I
zEK=a6!K?C7E<lhO)Fq-eHsbg70IH4DtfyDjB6s2Ym}EdW%<^ANai9JwFLMAVdt&xD
z4=ll^=$-^N!NR#>S*S(lp$B}HKq*-vZmiD=h?LdgnSz$17ReOp8x>!1yB1eta6xmg
zDE`l3)ytagan)qO4ELziag@Aa$q+ZrB)s@!Uc2Ew+D5}?NaA5wP9JZx9$|dw!52G_
zOGYIlnJ<c~^yTw3ZaB~=C&UQY9eT8biD-cxP8z`ZrK26`oikVR=nb~==15BUPdru@
z$Gq{4X@D~c&TB4V9wg{rjK#ESUxi666z7T{EvM(~Jgc`;No}6)N5A46?qK^>7jEqM
z&j0(7IFLPa$Y-%NixM~EPD)Y2v^$d?-qeBnnJUfU9E&1yXb;xyUOMjMgD!mOc3nt*
zJ$-tG$5ZsCC9HNkinz%~UiVxC_V(|JhGu&NtO4DMjC?np>=pJd9I<+D2p`Nxj)>v2
zRb9L|6iUA%{^Ape@l=n16|4f>9(r|tT`%}NR+zl(c3xye=N1fa9c}AQqB4KXfA<Na
zuOHLBH)ro+=I0#q>qYv#UCN1`$upef*V`)_#jJa5RBkt>>l-Ph{w81;!OF9Lt2D|f
zS`i-A{$dP0CJ27v-$%=>=K_cjV4~714=Gk!{7qXgU%(i>yXf;8fvA;}PrJOI;PO0g
zNR+RM`o{0%iUqe3QfP(vD0JTc&iecYrU2lP=F1mic?VxSgZlIBJ9rdZ<)EGgXOf8Z
zK+m@k`J@NBu%efkh+hGad|K0{q`DqCng9pT<q-ruOnAKDDft;Odq3d%E-&TyHK1nB
zgmiXFi%cTq;}g~IlG-~g9?R|j2n?^GYJkwmwPFG|m|;hIPSN7{$w}#Z=s;JmL;p$3
z4z?z3Ra+X@jek>VJLFT#12|B?oVCjN$f?2;!x2BH7MD0@0I5Fg%|V=rn-kl^60#rw
zOxXE^My=&O<aQ9v6&fYdYJK1VZDVqT^-Zf_7gIe1oHbH(&A3~1+%m{IU)|8HB}Jck
zzt`jlY{7F}rhSZHnBN}J)aUg7M?6V&v8b6H#A`V1ZrpZAt1at88UaOz(VdEgnUef5
z{L7JLy$OH_Y5FV=@P&Rgbn!(+9|vnS=5oH<=L*MI)*gd3t>784z37Fl&OriffF9x`
zQZL@ZISN=d<dCc7^eb?iiwSGcKKS(9!-gq~BFJ25U7Mh#Rlnp$yrxCw!2j0x&Ns(f
zk2)F+E&w{h)II3p6=*wmeCpIln0qy+(O3dodrnOVB-XO56H6&m`}kjrV=t`|ntLE;
zP4(tCOnv4#h--idozC>_{jzD=daf<&92NkMuSAy{pLdpD10WqpI~FsrDRvOYWV!Cy
zXFDgyXwyS~Fu-N+W1!VtON1=_i<m55WE=R>78SjcP?8~E>=kz96SrA1CTt=wepgW#
zs$8czPoxBg622M(Agu)Rt&1z*IaogQs(k3TN?_K;(a!}q_qdr`0IY5!%nw00fa-H)
z$wmTV_(7OyH_&*857wqu3TB~#y?2@-*0*lQJu0&7d*wI_kr1g*KLxN=q0w-}Efjjj
zT4=$|(q`esOpV-%^46-qJYj3Jt5_kAKG=qF;i>oA%G-@|pC$mDi`uE#LQJosWQNO`
z1%Anj9^+JR;IX}5>49oIe@=ZtzVML1hDP2zh~tI>N-An^^M5BU^}D=vRvo^*_tozG
zS=G*%NUPGWk{ikpKJ4E1+B@M1J@>n+I)Esyzn1&Gx&ce>#Fy!$*Y`K4Er+Rh;`}=I
z(nfkCZcB*tot<=6A0wiJyODCLC65~>>AFNgsA9ZgcndE39@XILE~7q5=NYc)D<JiQ
z7{<+l&<<tN8TZ`k=O23It-;qNAW^POoRMk{TJ;(*V*`EL=pU3dNSK@?8Zb6!jJ-my
z@Z+2WBpB1zrbvK=ta9fW;99~FBU`y=lr}<dzV_zpD1+rrc}8aV2fY!lF9S+rbybSB
zy#`5+g6xdEo8xd`dp`PjpVNX#((10buqz5%2cRB>TRmGpD^@_8i^l-qDDx6eE8OUI
ztJx(Y?72=G`q`hT%dr0e!9Y8d6kiAxfHVvUtLV-=GQT7(TipQtfeSa!SiBs|%X*g&
z@Uw^Cd`ylh`*3m(;zqd|nP6+)o|m%o?sI_bdL*2uRw``d;%RaoOZqG@d5N72M||PD
zt;I)FA#uxMR&J>FQpTUYB$3&m`q7mYzxDFsA9-yj+je<iIR~?6j{DDO>uUCs(@{}@
zk#R{vfsGDY{EcaE^gxY97W-&h>E|G*v_1|X@|c19^I^Uzk7W7Z<b6vWJ(v@2Z)t-$
zkY4U%a@50zbAB+V(g_bs0T=|HIWP#w1~<yOf*t#HrS=ot=gNL7@DcNwecm~A2QEOx
z{_&LAf0K|Lj&K_9!K)zf7vXa@q_rI6^s3Lo>9BypHPuTDjDiYxIpU{fqKxi8(ZER=
z0JeA0_2aEUlA<0_SN>eLbteem`S;rD2Jr63-NY(F`uU^WBZ<K7SbIM~RJ*fc)6a=S
z9vN=F2!ti(uqWRq36V)FJf+x5>pSPu`HM*wSB0bbpE7dyYskf7Sl!QzK_|(ij-ro{
zb)$|XkvhO+NsNJY>!3}DazGc-X*(fKj<D61$-(y0*04!fYaQz2%?vq0H6`ORe4{p1
zZboS1$C$~kd0epTkjsm=!?X3?HO1yih5~Bd(QlM>G(<rHUn69i41sd`T}j(`zrkJ)
zds5t%RN)f4V<9tN6f=Rza2zY>o^(%DE*(S4spp2ZRqIZMBObE>TJIfp=iR_pj!;nI
zBVA-c%+tmaBcN&<0K;H*FYW7%D%gPCFP{tR*5C5TNefnL?Y5kVsOqM>E}=}qpMqdQ
z*qR$QN&$5l%l#F%W}*GScGpIq0yg@3bp1+e6rj=MCPXRBDJUM4{#FKn*#9)>M`j{-
z`^D(@X+VM@FOYe#7%+so$7A_b6JLwybMm=>gXpP$%|qfS?;M}KsyphTf|wkq%4I+9
zecsp(F)C#+bMIr9b$Aiw(t)*47^Z`rJuyDl0f^I5-bb5_R{rHno;oVDiGy!a6H8P}
z?sflR1k=x3@%^iiV>I25=|77*>Ib``NJ>ZL+8g=opx1AF!H~zJtoS&@e|hh$ZtXF{
zr#86<Qp>sXxjD@%ZCTe}=@A4zdrWEkU9*$n5AOs(26T7CE^;zvcDxWeu&N@^T;s|3
zV)LXCr|U(d4+y*}N1aXeZGX04qK5tzXfU|1zzhD0T=Qaw5^?#9y$?q?;G@jrlx39M
zrk*GWWclVWX&bQVj2MDx;e39Z>!!ePBU;?C-a_KZ?hS`IzneM!XrNwK+%tC=)4LX}
zAk+3RtyWvSHtzs1S;YftT-K23XJb%}%^)q5ch^|{1ZZR$r-8o9S(E~#U4hPXJQa7_
z4#S3JQh4t0QNaQEq-D$*C3UW7!)lY2vui=BvAiJ-T;|$Z`XDK;qi5)2+v7%+CgQ~S
zx$TWYU{sk!L8^M#<8VL&9Fdbu=Sa9!$eWD0+A;?=A{pWjYsQuhp=KqGM!q=2eD+^5
zhPz7VD^NPH=0VPaENSX8w0%%0M{>aZ#hL^)@<01-J9$CW(kN{HqYnCgfYpS>xm*CW
zXOYB1JCedL<!(U6{a>fauuc|`*#~j&1}4@L1;}GfeNo%oXIL_i#k!Ub_a(hFSxClF
z#~fNYh*{Tl_^ZzB{~U$|`mW3<_geYWJ4&<g_x+niF*d9z5>6#NNkh0j=;bcK^sl&V
zkB=5k%9=y0tWI<;u_c8E)0I7H7Dw>}XPpx<cjRH{V{?L+z||F(g2h{s1OP%NDJMv2
zynPWh^MBet9m|c=)<K=GjQStx0O;AzLLyTiWZSie@Qz5gtmnU}G2HO>K9n9Ha!Gxw
zFO%CWg$?M^0ggdu$XsZ^grTZ2y1Z~+iYibRhg#0Jwe*WojK^XWdQrg{tEkQpu_|5T
zFSWZwGr_Tx3!28}b>oEx6oasFTjqwrzQ7(4mb8E#5~{|9GvO?PVtMUfA9&B)9}Nin
z<@oOFy`La17I9#Jc825_AZedHU>T@ro7IZq3#uRGTwWs{$;klrP@Zh9O&X51LV!tN
zw?bPQ7yfvT8)3#A2DmGL2omC#41(UQ@z@^L35*OQy<ar>$jd2-3YWlsy}j&XTAb#*
ztgS6dM>0!!>sVBP_^<CTw8JDb&;OI4{^g9hg+87hFlWUPp3G3J`9!{K%nRGpy$+~&
zgI_YG_R%K%_MP~f*0JNx95Y;&6}rMd_!^Y!VzCFa|IO$@ppC5W^z%r!xHaVH0o*)a
zyV2WnKR#B0c8vFC$iCZde~fxgor4YgMX|D2qu);;0D*8d6fI6Bl|`1O>b?s;n%x&Z
z{R1VS8Hv($1R{shp4Xnpn1!&HGUOWrp~GQY9)8UU^3}&Mv%a?f!p8mQ-iSKMFoS!~
z#$QpcHby7kKO&NO1abt{q~!E8y!c(86c=Ewtw-w-+@=koZJGOV0M$PR3qCbw2mm{4
z&7qQ0Pz_-F^MJnKN8I9_7)RzIkbiZR!|H;Hb|sj{j)^O8b4|osTcIAp--~7-TU!gy
zTIS@HqOH`Ur<8iXmAzC12IUodudH9T=(RcGYy+qU;3)P0?veK6w0~iE2%Rl0_p_L|
z;jtZI2Qgi8jIdT-eDaCyvGQHi<f_(*)}wC|W^1PeB{wtT02<(W(fxHX(Bt;@k?+zT
zu6slnOE8GW1WrNb^N^c{IPT<=cFpmYdTodPM?~;aKG~A@yykP+&U3nRUi8I0%K(P-
zq0To|ssWcRN#hPWZ!rg|xSQLiD^Y(u>ibIUGgwxrLyUpm+x^YF4~j538TM@NsIK^O
z%Qs$U>IW*V7D#5_m!m=TT{TJqc-Pvyqj%jOfHOQYF7~RW-q#cgiOGv7neWWli1g0`
zF>>@k8=1xdFZ_HEi}85*!on$6#=v?QMR#qpNiM9a(f$tjeA3eL`Q5rug#2~de;)k^
zu~9JwH2O$PI|QS^7Z1pp1b}USW(X)T;!cs5y<a1KUHv}dUZ%|pWO^L*=iOS*Qv(99
zXRg!5iup30rTJz)@8h29x%V~g_`cxJ@ZSqP8AqlmQA-EWhC2L1t19k$i<YFn)G#zE
zXxDe1xHO{B&ND_K)dO56I^7$|iNfYdN_X}WqOCD)JsyqRO3EP192*D$j>xlnVL>;r
zGB@m;9r*B#*Rt|`jzN*ID%{AZ$ibafdK&B4U7@M^(A21iKu`_#$CVb6uo}~1`6M;&
zESdSv-k&a(RpNJ&2e^qtW%_$*XUX)w(G*?s6e$VNiZ?O3^)4?INEnv|w(ld{<$u-0
zkrS=U6+lgRq?K!Q3;JZoCAFp0{QP<^;rQ1e?DZWnVPpC_hqrGh?&+=7zoG~7+EGE2
zZmT>pK>T?KqkJ#9BVBITM`0R4&=Mj%4o?}#Nlw>-8X>@5=b4|b?m`xJEhNVaG2}h6
z{+58q2jNxWZf><_ggk5D`c@dUh^9$G(KH_M3n7=K)ydE)t+8I1mbm~ugIO7-HS~X{
z=BhH#Q?Eot7RNF2li%{>=snb2n>1okNsg_eP_Ip&yEiza5KPkC?&-|_)GX_GucjN$
z<gPF|1bcD3h~mBI)-v!N5Up%L_UN{xnIuu6?QTus)_+mtrX$%1U5&MdB6OmPn;?|s
z9(9!0JOa!EJALSQ!VjD7&Yi^0jMD+!&}PjsTVRGy=uj2Lpou6uHAo#e2=LYOIeH+0
z4-w<ZL7b|rLqozU{w>?jelM4!*aCH|2r9X`?0FlZN;?4cSq&6~G7<0o<$k(s`lBrp
z<w;%cu=ayhq}ejw?vrW%@h(EvP6JglhDrTUpt3}F+whe)7kvoPv|Ad9USfTv31Ep6
z1)|eS6U(Z-G}c8IMPVBy>x9DT*K^xnW2?y_jsA9!$61N%Q!nduElUlzAI9VT3orUW
z$9>9}y{q*ieZCD<RHnmdGtO+c$q^xYofhi-r`~1`mFt7)dagM3*L|l6qY?*+Y5T+5
z>+7Ewc?=+(SC*|~K)9PCOp|fW&XvV}dmy*A5ex)aol`s}Ub@5iOHF9HXe~yF1cdw+
zg9-V{40+0DuTP9Qlh5F!o!ncS?dJR%@tZ(M*R!6gv}b+oM~Wi7qm;7?%{dk+1==KV
z?<_QbqF;6a@4In?*UA=-_ilsN1TPc?WxGdOBLczl?Ebf2c&PIGJQh|MGUd}7lI7P@
z*jAI_ayXBCwBf!iOBz0BNvdvGsP6MbZe(Gp=u(Gahhcukzb53X1{WB|%&X_7-v2pg
z>(dao%*};kO~?P&itN%f04e0Mzg6}%>?n>Nk4%Axi5rkD`#=f1vR}Uq$ePKPN#zWj
zT)|gaOs?3^`dq%MiGO61pp$6`;GV|}qPwF5gokOWG!#;KpfRbqB7Rk~uJ{O94W}=c
zd2MEu2jEn#Y<VsJd~~I<hilWRUX5sh_9WxB&3gnjTTqo!_{{NNi4^pE#f~zZC#nJ2
z>|YvIzCg2CNO19C2h98ZG{wjOH&y=c<MiR&o07Mdv$b?e@v^F-fKbkYl-z|IAGc1?
ze%`pXVngHD@W$!w_|_aC+hq_@pC!Ry{01SPARaYN!1^`_6U`r21LKrkKtn9r2pf<;
zp~S=Wjgg$1tftx2Mpm6ZL(@*B2$hZ&Rk~I&O`81O;^Wt!Tdm>y;|xZ*Oxd$LKXWS~
zbZM%out=mV)o-{bj(wQTw{hW=OdEaU(N7@T8OXyG#rfPyZyr(W^DAMl){?dx=D$TZ
zJNW9J?*&AGH7BaU<z1q{5$9*P)84HiHnokmTZe=e(14Ty8w0dLCo$|L#|1CILh*<o
z@3-|V+EMW<9UpAiDaNyH$-=cEt+}0F0r9mMqH#}dD^)8MbR4`czWwvvWVeT@MG59x
zWt_h&V7b2G^gGw~(0kuwkQ9%ZhrPBH%imvh`*)XHqL=bW)|#`PMdE8$f|pK&{CWlH
zq0&m%hVu%PrpG?EIEtbq815C%Z|AC;#fg3?pXtu4nIe@gLBg1;;;LHAtqGpPUvaZ9
zTIF(f`}cpHjlM<L%ktiqKVvKE9>s#fI^HN#F!vV%;UE2Z4J-b8&_46Rx;?vef%L<j
z%YWLDW@)T6T|L_!_Y;V~Ob$GK1)rCkZeWLw98Rz?HG2l;3CRgZcm;3Nr!M-PsV|$i
z_wGv+A-hK9uro^R`sg3#t^CsN+2_9>&YylWSvwscc<^Cmkn6Z%2Ir?=Q0TNB1*_H;
z=>kQB76PFu^tXn*z*AyRyd5j$+q_CUQS_}M(%j~H>hi%X_G_~wN*MR)hByCW+1bx;
z5n^^8YoK)Lq3vdd!k8IUnZ`uzIEp|jZr$_s%=beuAIL>fGAX({CLD29OfIi{D8{O^
zEOHRLdZTZVbE`^v@uKs#;w=1*1jD6dfR-N}b!IN1BdFV)X|Keh4ji5ho&7KZX3GWT
zxa7N1RVKv!HSf!q*?*w<4c8He8p6{2e%tF(FHs=MWzMfPDz*mtf4}H5&1bFzgIs<n
zL`c9LLA_RulXTp|I%#`#G>3gmXF9b1Ur*xv<QCx=D`;E35QZNh*D=&IOn;2rl9*Pk
zCW3FSd#tO+%BlXpM3p$eOn`*CS`hf&PD1@%S7tGmAMm3)?Y;L|&gF<;uK;K$qHF%8
z-t?76_g>p%Y)#F!PiGOn$WqQeLl|2?jqU&;*$?Me;ylu$rqVA0r~%aD5PR^J)pQ7p
zJ%d=WDKGRYtoz_IIImQfe2n6lw<+Cjrlv?Gs=99>w}<MF<x@d79^4ao{egKw9Wqqr
zvzRb&kgej9E>a@W7P?JpX3T~ZcE`9NZ(f}~yPF?ziDx6^>zB41VNohZCP6V?zBO@q
zV`HTCsEz9MQS(@ey%G>82d-5ttxPVQqE#LuyZ~a)M?4dr^xpq#Dzx2f9y)5DyclYn
zHjx%89-mK*@QllHVx+R~KnSMMUIWK1AR2Z(qtBDAMQOzXziZ0edNJLVX$nF0F17qo
zsUP_A*?DNj=z;V$=?-dfO?glS{&vjVy|PzUehz!E3WoBk$X-dtU{7UQIX&#l5%~|=
zkN@1qv{czk++172c1~^7h80foxfZlmEkNz(W1N$9OOkJ)1<RNHtXxnyhHx2jgN2E~
zGZ$bs{^ggmcXQ&^LoUHXJ)E+Fq~0OYL<VWXtfl=3kd`mN0p2wbo?@?^q5=8~KzbGc
zRcWcrIE`PVF+}+u1?i>h5S{+X&txUHDr~huCd*AZmN&DJqu!Oq{*(8m<}AX}rZ!De
zQ<+d`Xz4-vAjc?(@~RwsCd7JnZ>3^lwSedD+})xkwEH%<65f;<P~TH$6)BPz@cGjE
zpr&~B?Y;)zEOhGCHcr*3f$rAqUE3o{)P@Vq)e5!9@iJ`aYOM@?HndLicpKW_x~=qX
zokV|6Hp26qt$VdXP;H!goAL8O{Vc3<TBLIuFJNdbS^ZG%Ydx?NRnfTy`snzRlKV;3
zoeuX0w5FXaJOo#*Sh-o}C|&WYC_O^Y{1YO#nEZAt?|ot=@gYFn`~wY<PhF|eOV|4+
zRe0(3&ba6_Ek~Ya^E;+CFQyc&?9^QEj2sd)o2?wICT?NTz65s<<Wp5`(CY$R6TQv5
zy^Nid(de<!GMDASY0E)fp0=L&S7Sc@^<|5g+R_%h(CkWl{daPD>ErWoZ$9S)AY}pB
z2M9`@^8`BZ2-;Z^5OZ?c<6=D|e{DT`w5Hri<@Z*Lr%zNImR!TgaZ5ba#6K5C4;))w
zzHWlGg4iSr+^umg=Bz0IbmO2|N&!<O@1euyeo=MzQh<uu_L>{w&87c1e7-djn%+=5
zS5ULc$||ZTy(T>h|FT)vnlzvI;;dVE4bCl<8>%U+<DI9V(rZCvEjtzI+Jc6gao1P}
zX@bLBPKU)G{a;_+v&ZQ#h1&$PeC<b=VMAPeU~*hXta4L{DDwv4_!qPY;NPlLg(KGG
zzVt1JyheO^|4bG)so*gwi!0YpR@k{3MDg^~HH`M!3zw=F2jrp36}NmGz%W)E(lj)c
zuUO{1Qt@Nm=+~E~%s=TjfaX7(T%DA?^fuql&hD<GqhmmAMTMo6)xL#={{(N3zOchX
zGxB;d78~hKg)>*(*cFO@EA(?|R~koOQ5oWszxDrr6?KTurR@=YUtd+p^a<RUC~4w|
z5arOp{{?kp9$5DSDqtoL`Izv)_IXO}$Ilvppud2Cb`#^|wNgzv-(Vvw!LPDVg;7wR
z>LDknbep4W)_0B1ZTe%2k4jl+-5Ay4WxZ-Fia|g+{7$2|;)9kv>X+$o76Wyv7R|Ps
z3sQ~nBeMI;^-D@h3QJ0?>;aPH;i2sOWj{t#lNwC(>D0!<T26vf6%C-IsDg>W(4ZB`
zrA$uPS%_-GyXglnAGK$;(xZfQK}yR{-C)Sz@dRX8NL^Fb<>}US`M`gXVXpsVBZQ*b
zuaTx6sB83*&DxM6>3W0#W5Nn>u^=Wl<U#jc)18>iIkv~tT`DjSYul>Ug$7AXdv6GR
zMl~lrcGy}bv+YGeMFd2$014#h_o1Tz;Px#Vf$OL=OI7oOQCaWz$#neqI#olizaCE_
z8OlFKdV2PCTj-WFi-!&HIQA9W=2#V37nGEMMMXtf3~k_%*&$KP0QBrYb((Hi!%@wd
zs~Mr|lXe7ztwG0jrBI}D;J{t>jK@meLFlvj!k~(K)J(YWR4u7tt|xkCInPZ)EI3=~
z&w$6i&;K3-KMmY-dmE!(obI)a=a{Z*4yW))`Nz+E2f{RfeF6BC0gb@8{Qo>QyDIRi
zmXh}aj`X+6HO^cpqN26%eUYUI18bF4K+T=fkQ3H4cwtUAFm1Nz3kSeC`?WDnbGFAu
zHGuii`KfjBy`kqHr~AtKa$<A}OCZF;qM|k)R}SOLa~?mEklVzt_;wX~Fl0ri%zyQD
zo8NX)pK=B2{*+DzNX_?DQ#`(CBxSmmidbLj@Z9*d2Ix<|MAyBRE=&QD055%H=Hp!g
z)PM9YK)wjAIM&8QT>+8mx{YhA7z6G(*Q9}b_TL9joezkanWi6w03v|640+i{0;09@
zJFrGvP6l23%+IGnqSsjyU!Ma8hOjpZ(E?u@LONI)&@nOANPjAxw=4Mh%Z?6nBn@Za
zKj8I3Rd)HCfMbvL%0Lbn3ktM+aF8)WtMFsv`=(N$(UaX80=nKgrGabLD8i~L^V8o$
z(kCjdZE3G{T%G<JEnJgHRA`)dJk~4X|9DfN45o_o&u^_cCTGu`{qI+%cgV#uef|?Z
zSa@=;YJ6g4z*S)U`MRG5kgaUf@>o39m;i(g|M!B`mc`1qS2rj__}4^#-9m~zJ1EC5
zpJdhha0M7Rwa~AlQ)`NrMRoOgNM=z?$g}DbGn*@eWyuafCPY?HQD(@pZHav?+W1+y
ziEDdG_H=2jd9$eh{j$EQsio$n%!6g!Q4zAQ*0i!`j3Zsx`thZRFj%h~X4|J5OIw04
zZ&By<ZB_prz2*|kMm$jxuR~=Qvz((?&bitLHU2@YvijdpcgC?2@<%Fg6>I(AOUJ>A
z$x8QdxSd21Aoru5N1aje_j`z(1}uSE<?1{Kfp)Y^^OdeGKG(w5k@4^6DAQ$)r7y&S
zj8&Pw37J7tvf9i2P+@a41F6dN>$9i32zb8rjnN*gvOT!5)Lf_{)ax?vpPZI#L^8s;
zP$5D#A?&a66|z^UN7(Obh<_-Du8OpTCx#()-cTCRCw8~p&C3objJKOk022ZAON+X4
zRWX>WOsO1@PW;~PqYWt=fYOGS(XSD!_kZM(xd{9nQexy9#jR$!I%cb2PBg+=Q-h9g
zD7g*H@=FZ>H;?h^_Wk+QmUYVZ4YjxTz3$e0OwHyp!*3X)>Oo0U5iWIjJa1s{Ls3~@
z%-3ZWG3GQ}pN#cceCGi}PU_P(hF>KdPf>D4r^udK?TBloFQ@H8d%CZV1o|s)jhBHK
z)a9s^j|q{m46=TV3aLuB^7`(<YjNLo+Fuo7b)}#0WsI1|C$=}Nu>qLlg_#1EyG(%8
zbIs?HZ$1_cjijvem$v=0$5;G_zW0h$zznNlSly@j53)W>tiXd)Z(mj9tn88PhI&>F
zac!>mHBxu#D(*|hO9q`@SW{CYHb{_Tyr&|P6=l0$lx(j}2o5JFd@H*)-Dq%CbKu^}
z)1WtIRyOQ7P9?g&VW^X0GswD7K#GAqJ*NBo#f9C1?c1>9t?L~@zcq+=HCqKU<WrkY
zareBAn;{>94KomBfLZ4EBl`BvT^Es1%~MU%10c0)Agx5Dm9E#J?R9&)0vi9^*VJHC
zmf}tD=7ul8EQ93sK}RDCnw8&CN3qc=Hi6;0KvA8xX<SbPA^P9OjTPi)-hn_MW!GtL
zXjBiJ0#PY&H6wDkBvI4#ic^1d%7-0Q1;2Rv`6Nb}#||S3br;AbG_a=F!_DD+$oRs2
z!LC3Iul`Ze?%>P^t($sTahm`6UT5^<72r_McZNs)_qO~*umt3Bk!rVtk7q8pf=#P=
zzt`5k?8}Jbc35@Wyh>e60e*);Q!+B>??dZ!@SHe1kXKT4nxgJjhn{h8aBXJTRv^>6
zP44en8qe_eI1hjL@OzclaF8Fvh)|!62(Iw(5NUmFp92I`-Y?}*>|9D9mSf9Ktux2R
zwZGq!gqo`kHxn>{eI_OL(?qR-d19|9;NXTn`ndA>Uso$GM&7diK|AIU_OUpSsGtG9
zW3L88D$m81+>(HEs!$Jc#r_H_1kG37Bf1ZtM*wzg4M)g51fBzsMtEn0`+994XF7M1
zN2SEI*su_2kych#fqb`Bk@Q>@I5l9dXnj3su4ZYvuP7&{sa7YU<wJEfjp1NvnNtpJ
z4l0A#+0`*>_@K<I%d@p}%gZ9X9qQv})T1?5E5@x+fndc{2a9Kv!7D3k8NnHmu0C0e
z{T2H6X7}~HMJyk2n|db7E3Qin87Gzoz5{}uS4>hrIhcpOL)adh`zN>C69nsdfpgOh
z*wl%?qG;$=(<GT=2?KD94LOemRqxzW1B}V{_sGJXUO91yKY{0O)Mev``CvK3)~s4i
zx!2Rq4I|k>O2qP#WYbYb!4eWor*;)eLG!1eJV$$9_-h-RSG^P?fKUS;Y#c*+gToNa
z8Y>Ik%}N16&3|S@JnM$OmD}6fvzBI;`y7HPwOZ~!&f^J$pz`9Pni@Mk8S};KwyM#7
z)&F6y&JWicqbmd3Agp)CLMHU<YmmwKb>O$%F)}^-YnJ`?`-7{~imdSxO?#?+r`%;D
zlH2^Dk<dDfbkU_?RPU8;G-zWp5Tn^RAK7c~iAVw>$l~;g7vB}FcBQ^PoKYopJ@i|C
z+BYXPZGaQFN!Q-FiBueQ(q!^GkkU!%NDD3Y2D0uAQV;r{v4`}mfwUy)gCyg9AB0#0
z(4=d8V-MHf1O}}PY${9hZLr7nvjlFqRyh1#|H^0aaJ(bKj)=+-3)}FLf9%&p1vpY*
za&weWj;JR3$g=FQ;9xdB8P&Lnretc@+7Rb(wwtSoFP6eQwQ_2yI`7lgxp!2MD6cyW
zcEQrytH6IRfqK*SjCQtt^An7n5MNz5T#y{fSQT%4Tms*;CI+J&Yc}ptL(0PiR;EK2
zb*}9S0i9O#(5F?1*M)|c2N9#LC>sVJF_Xa7ePeiVfDD1Ap)y$RG)`KyNU$dlIIU|n
zAeC4@8TT<Ao#*kFy>7bA_H~ybnc09a-jP!N;!3LjUrFHf_e1Xh&S#;smYW$otL}My
zrs<gO4EPn&!{bM!e8U9zwZXPJ(8l~IDt(I@@tE}~AOFcg$9n}CqH=d^@`a~l9$xp$
z5a?dEnx77LLJNMy9V4d{X>1YJsiK1oC=9dfP1aQAF_(WyQinwnvuD(gIp|tZd0>Mw
zIh#I^g$0v?S7O=Iz=hAv)>Pmcp8U|A)U$APOW01rW<_<&x4%Dh?KVzlrH|q7TJc;D
zUFOxj6jLi&N;Dg^*Jql_*N=Sro98k82TOX3Y~TxH6xM#*BgLjys&5Q#Z>^}HXZl~4
za$Kj)wQ+1P6Bn9T8nmgPqKb^O)V-DfpR!_uQ%lUQPLpu6Z)h8!BeEH-bWN#E9P%4q
zn8!w^Bl05~Q1mtYamAeQe`WSBKJ0mp8`la5sn1=A>N<E{G_mc{i;n%YMF^0+0p1J{
zQbL`h0cF@vpM)FclWW3@`CdO94lmv?ses#ghtISC8lbw;%Hm#HYpCy~ahh9J?+%-&
zovkrFruF+TR=_Z?XV7J?7!lG|bJ)fUoyF{9OPq2s)^fxaa_O*kwNFkav)o~7>igIA
zSHcjh?s3@)*3Ub6ZAV??a5eg%V7~?Vt;a?Se*QDDcN6|Uvfex%>b3tLSBg54bB?lx
z%2s6G$#Qa1lBI+YqwL0#7=$sTsO-ClschL9#@H2Q-^MbGHTyD_WEp16%<r1c=bZ0-
z&hMXdpU1iXalPldUf1jO+^)^HjODHxsk^`ayQgSKprQM+!emsh4IGbGktHsEPrI02
zX$XD+cI~ltk?~E?m?OUv{LVwKHJNHOnfjpt%8=inTUzBha?+S{Ie~?3<c{&8^G8q7
zueHXHJQ{XQrC!43?&p3z%?GB8dMcrbs;8GC$rn4vC+}J1uDrkQ(&w|1DN(w@Mu7A;
zl{Q+ATlhM4roJkG$#(9cBD_5o{I;s^<Y`(Z-}gBoQ;>qagX&$Zm5Ukeukr8EUT4$X
zWI}D&nYXdoHT*I5`96+54YT~4b?Vp7a`~`U!)ApD&SukH<kR8F!va4lCva!zU|w7t
z;RgUr@y5VGs+2$nP()ataEerV>IEC*=qctc^<6{VLo|IayKwecBYiylGI~P*eh)M8
zB!cK;X*q6#<T!e+4GZpP6ZnP~l6TD*Yn5(|r3L(m@efSrHRA2zRx>vzIFx`d%d@&R
z_uOKK(POO?ZqoW-dUq_0fI(36`s|h5eOvAmbHt_ZC<=7Sn*Xc1E$7@}@606k@5%{p
zpA>`}4Ge|+*J*XdE^4W|A?@OLgcfcgw><jQ?O2Hi0?3PGhj9Zs;gdheNq`3Mm*Qb(
z37O<C89V_r3X#yt>oEkTzx9Tw_bboY4f=CZtP8-&f1uq_OfVx1_Dz(C;!5wjwcOLs
z&-a~cw?1BB9dntWpUijbH<e$6w+EkepDk``+616xqjN$w5ABE)?+J%(wpTSbTChK|
zS^xLWy4U{kgnb`;IZQXcS7t!-)>y#uPwx`&|Ln2UzwxG4asD3~ROzr>RkaVDl@mc3
zHlM*q0$5^|-eFdLeE6FB6eH-}TcoZqst9gmOOFN<akG*kkDzVdVn|a^nc>Mv<=K~l
z-o!~?@dQuAIHiB^VFjt(R*8_Sgd9#@l;F{HhzA3Rzj0zfRu@^7{~%_!C0b^~4zD@h
z+bf_zS@!j<qHcw7XeP6Uq5cgKX$eHWHrlHXcUOiCXzHnYpTCF)8pf@0JQz|ZZwG+(
zt_ePS4DXJUINQ#n8N>?A9>-9Uwvi~@YR;lWXZzOdXrGjZ1*k9z_p3anOTq<$?ki0^
z4h4OJhWhT1aj1{va?RHWPQo)5$$ky;S{4}Ri~}|sFa-q#@<a95TI1<}l_2=iLDMa~
zWRFwoEWD>0!nW|Fsj=3s;eK*HN;{HxM8>V4^H8JCXO36}Cy5g6N8Iw0@0~qqzg=Er
zc8k4}F47C#aMdM4_xGiA$C=2$&E3%Kb^7NIrFA@V|E~q!9%!l#wI`!`oj~FV=8`W6
zvQex;XvD)79POAiIk!^^BC{m;dguWET&`rlD95>i7Lm;lpMQY6ayi-39jmW2nGHrD
zr~Y2XBx49<^LRx#pHFJC6r_Za?Wz$Ar)1)sGLP8C#14wOl5e>?nrn8?OvwO%mWk?C
z$wof+zo4Q-<{sd-y2UleJy(VcQR!7?XEnUH=*t9u>!)vZcLxIG1FyvHblM<~F}ySL
z>-}r~{uullUuPz<pEckh!A{CMpql#gpya;DKS@rEq<i<T;rya`qwEs;TJ}`yC5}~5
z-6nSWJ6=3|0rw(`2U5XgL4GpGVe%$Yqrs_WN|UC=8kV8vyK0CaC4f4tIteFEj)Bhg
z-CJg+*0(BN_4oI{h>eZA!M~!P&Zp%5HtZP!F~d*kGREBED^8n`Hrai%>K<wibi3!3
zOar3is;Fr=%_#&v+QXWR%9f#`twL<TiT{fo4vla+$sld?Kda&WxtfhpY*WI{N!`J4
z>+|C=#&!{m57=YQT9!;`;hHWwPx@WrWwV<+g33+TGNoqYTYD87Ml1EhZ>ME>Y{EVy
zM<d^dQ~_*jGH|=6mb6;9J8B!ly1yl}?=uNAld!D1-NqkKM4UCq<Z}fmc%DGx&by!=
zv+yaip68@jE*}eYSsf-zs|_}xfp!|W7_OR=<9+R80+kb$v$?s*!=Q+47utInW%Ix0
z39i6D^%#Of{%2FQ|J+ommkJ~7c8=&W@9gXeJCvLz0AG$35WO~fC_Asa13nPp`>aTi
z{Zd`C_%RR20$0O8rT#j;NJE;Y1rO?wVbnOjjlCf-oMdL5522ApEC4{qAi$?RJl~Ok
z4m^00;lG-1EnjEX5&wreGJPu0blL)j;h!$I22^j_p!9X~bXK@}4Mc5&Ln+ZdZLKUe
zX>LMd1x2Tq*ouLW!~CxP`FPe=)C<|PpJNf%<Nq$B`sXs*+gF9S;r9ksg2WZi8{3Hk
zoh+YVT5@i9QAgXVMm%V@s(@9IRrd;DG58c!w0&&Hz?j{)HnFT9LmA_5gMucL=cez}
z)4U+kAad2WiJXy^md3O*%4O})msxCVpmN;<G|S#UIz*wfUtWQtKm=ohz|%vb|DzU`
zG}VW>E*X)vM<I?I+4#8KOPbG~vZdr&p+<a*$4LpfU^QENH_RL<;e+4q;lJ7apCCv2
z-OFSK2F4dl=YAga;9t02y<a<m@<&EaNjk4;9Yd{PPC%gIK>JNUA5(_Frj{UkU9r*U
za6+!tug7f0_xbn^F|=-zRHftQ<c6E>8HYf|=I%%JR5y7B+Drg!!}>fKK_em<eQrmd
z0$3>aLio84UcD+e95bKhf^9bPhYA}e?Qq$)X^>V@YpJk1r%+t=nZ<4!QiFe98gS}P
zjm)Rg4gf<vXxJn1LH1Wx*Q#<O5&AAy^e)l4*;AK32s3=rno(i?xq^2`b?#jXn9^42
zPC#>Zr0NMtNG>UM0E5}t_GA$9;|=Cjao|xZN$?4vF;umIjzs_%CM=p}JLg3%a+I(F
z1=ZYAUVDlA*asEZ-0s2G8-%y!0Z#9(e*eTsq7s`*p`{HzpQ2yIsoU?ZRrKLra9V%N
zidJ<o>YWiSy;{Vk9#k=e{jf5ZTiD6v@!G5E*RI+lYS^{dC4<?jy*n^_!~`jvj2PK(
znRZW_r|twteL=lIUVrSN^Aqnl#r5CC-2S<k$fV%|Ss)n75|_E|bCWJytUAh0S1$Td
z9_2t8ty0(lkvWoM-)_o1=VS%VQp8}Ehqxnl(CRTGYqg#{ClKWE>75n<s*#x$EJ;;W
zm1#beOMS8XO;&=T2A;*iP<2I!n;2ospkCpXrLnz(^xlTX6yydy%N<epC^x<Urz||e
z$8Du*ept}3uT328sjAt1y(sR{kOzd`3VH<>e~LH4j{Wzq|MTZxuZ&W?qazTvW2%dW
zpZx#(zzr|rr=DXz7uQ@y;gA=hhbl=mE`>&&A`|RwTxcJdv&o!8Pa*Q1p_J-f4`f9Q
zQ?qz!{j7d0rNc1jLvC*F5fu>~41jI{C_hr?{rNbjQ}gqajQ(qz4Sq=KLB8cM9$WBc
zF83)Uz&w5Tm#(<))ta+4Y$c}ScI#3V>AnvDJKhvnPbueAoeY0Kd#@<awJ2!5IWzJX
z^}nO`pEaF}I_2*CG&$RL;3B@cy5V=rUyeB4dXyLlguhsLd~~{Z<ZRH%ltYT1egG?9
z8>^jQoGW$h(YBB_A*__GBz-zf)0gO`RPQC)QhPl`X{>)-W+iie>Iu<m9Ayl(@L3Ms
zPFZ0KZ}QT*h+iMa5TFop8vS<M3?kP32?AAor{Nfl{m*Qx>mG=@po1=m3z$F|ebx*&
ze5$EF6E|SrD_>Y(Y%{8=FdsYb<=@&EakC-Gg^3VD=TcwA8$1??(e&&Nn6CZwMzQ>{
z{<p8ktbZPxzyA?g`BzC6E=;iZew(e!D9~OIsf9DtZwN*Yje*>Nr(;lM2NP}}{5^wj
z9s^>+8ouoiS4n?`Cx@3&HAJf?=hKm@1Z&DrbX3{{<I*mort$PB;aE?Rn-QXJwV=>@
ze<qy30@;6tE3q2iY-9W|Q5rSwUe^<04|GzRhcj%$0|T<>+pJZC!uYQ)YLjqO$9!~!
zKGerO*7jTttx)##=QX1xjmJC{h1qk3KI!oJQQwTKllSa@2$W<S;mE-}a|oONMcf%q
zZT?)+V>?)6q(M@pvOoOV>2bCYS0OEWJj<-*On%uGJpK{xTgD~e8^uL|?yQu%-bDH%
z42)s#Gdu<9(Ng>>{*4<!>v+jxm$4!_M@gIjTR@&cT4OqCTrlS92f4nh#+y(30(L-2
z*duQRKe-A<V$v5;jrVfB%Gx6k1SMEMm9q08%bO-6fSXEKbXu=McZ#K_pIFyVfz@|O
zKEq#Vr36<Ch_szMZyQ}|-q_bT%q)?gJY@c;;J;A&-D~FN>%v)>TdEK4=7O}B!4ygO
zx6$V~db$AawBbYj)8^Y9&Yd*}A+Cp{{k`wrV~_lJS?ySEu-uN@{a}W8T_kRqD#@Fv
z#rxoMh6R2IUn1&LG|0RtHE2&p=D!-=T{RrXc#JG*5LNg*zx-Bvu$WHXy|3kAS7AQs
z*~VA9{2{^{XYRNH+BxpnIeuB^XnswqE$P#27CMORkM@R<AcJ?4B9$I0Y6|mcOD4O;
z%7mwr#<B=~s*g(K9=d8gg8@P4-#hN-{|a76RNs8h>(IG;Raf+80m1qFc->{xNp=g6
zs<8sbOG#k^e+ivJ?7LwWJ7W+6(>&Hcw$)KN69C-aR#QW|Zyub7vLBtc@q<pf;IUy<
zIt^ne{b@Ye4i1Z%8=4N<PM=<m)jo|?YTRN)V=zSU6E%{$Ak+`reVP#Z(Lz62=nT@#
zRQqF3l#I+FN?Mc@k5Bh6RylZF^Dg<CSWy28`vs+Kup!E}cnNA)q+x?SX-nOZkQ0LK
zkLZYYay&E7{n<kQv#J*5#3b*;Y%k%S6&hWyIG};0r-&qRwTFo|aKt3}MFRJXYp2`V
z0}v?k__ZrlsQHg#PK&1rB9ls}I4#9<9UMxvch(DX^;gTRNGnvw6~Q|GB5LKs7AE+-
zpa527=`Ed)+a;6---5ci4G)_^*5TBk-9pH2?nNR-{Is$q4c>lzwcE#ey*H{;`AzeP
zw|lK?NvTjO-y>NE2=`Lu*n%FhaLg+cFvXvpQ3ujhkKg<o_syCtJ7wP&upGuHoN>W<
z^ZB`|*$~$&TKd0SZwOX6z+qvpjUOgF8=Oy7K#0a`Pd<mS=M92<p8hlVc2m%@5er8t
zD=1xJ2N|+Vf$r|~?plCT9ptrqUDa#x_uV{jlbX_-|DeM;0V2UOm|EpLC<0&6gABY6
z=PwYTEZ?96)yGI(z0D>-e7LUQS;g5^n_zEsPzqOy_08*g^s-&ORkhQBO`3IAR6cNi
zt)Xzypj0fdz4<?(Z8JPSJLL*tE4!S8iJ2TAwAVLwKhKMt0<uAnjL>VX&t~q$%g$&{
zGY~@FP%e@v;<J#eci(Y3XjkBxs68i@<RysDZdWYpX)3mKu(#j+(#u-=xc4q)t1JR-
z)*`=~0#3Nq&-35w+h{D+boiu+kD1G<Pm@-!LBhuPBd23DNKQ@sP}-K<?srdt;UXFK
z-Bz5!l%i5}kLKy>K~Vg#k^qc3Qp47K&pg0&ci^}vu5sa8$A-}(+685<t}v1A|KOTg
z_%aszV(;oCgIM{BvYZdbb^#n97z_A4#{f$eI0mH}%wHRpWyC+vFZ(>l?2$gL>5rT&
z=`^&9ka1gcZP}a#<MXP^$ebFE0La>1N{)vKf`so2%B7F_piMAo;%5fJi6!EXg|1z6
zDV)yxQheU}i)JxLQ_7P2rt}C^rN&;4H!elPq-wwt2IBAgynOt&!8KqhP9b~)UG2Xm
zIXu;O`OMs?jga2I_?+N79`^wrrs3W4TfAg!rJ=hXklxb+=)N=qj;-i-;QG!U85pz%
z**~B2OK(MbB!~`jY?upu?19Xgn!59+)1JQ!uj>eLOOmk|tkourrR}`<+P2KeoLNZy
zv6$}R&u%AGGygl_2&<-}8Bd+<>FFUgb2h%4i&irL=!{+ypQHPvU$OEkh1A@b=VCc0
z;^qq7nZ0io^r(_t#ru`L>-2HOtdg}q-lz;3xb26t?D>eqQ~+MVso>v|9ShcUqrHi6
z_te$Kc<(g3!oiDpd2lI~GoZhm!SA5ibFpLd-{g5dGj0C)<YRi7`<2+JaLoH+;4g7^
zvAY@{Rq`vV>tudHpOPEy%c?5n$2*oCsk%BoFWNq36HzJ{t$Yoj=x$AXj9dj9HB0;%
z2<ia|uvR62rrsO(Y^-hDB28;c-3@R|$H}{~enpi;o+@86iVf(`ypSbVz|p5(vNuZM
z6+$0B`u{6XmlM^A?|EC>$Jm7eKG;p3|9A})+GZRmOQbXS?>SGjagX=YG_EV2!6i4i
zdqSxbS)biD+~&S|82ADUoqP#7kGG?7qF%plC~Ab2+3Fn-l$#(PBNh%hoA$3~c(!u;
zFC>|f7Pvk=?g!LbAA-8AN?OiBKl=JQNUfK7NotVqMgxT1WwkQbIYHJS_i+XhE#O!G
z>D~CGk2fC#v4LeCK383Beb>>hz*QsC9`&yeuT%W+x6WI8THG0zojZW1)$LJx+W39o
zyi*Br4O*9N9`8Sfw|U5mpDxaor4kP9@cjom?np?}7*a?L(fudo;H58gCRKNpjZR(k
zsuHeP`LE<qP#V(b=zxp=`1n+=2k4ny+}$@)Y|m-JLNZlmHqq}Mrs3SJ*kkOY)I!tl
z`xu?&&#pV&+5I+fEY&xozTNI^Z*x7~?+W4JzjujF`UQcXAf2)O`knabgnJ{HzpjMx
zGPK=!%lhknVy<`3^TyD%+*uFZew@MYr@aW^02p`1MS@Ow$%9qXz`DncmodIIO|Rjw
zZPQ^@ZmXxC@H}QFh1AC_tV+v?#s*jc+J0M*xd2x$uGliKFK|%~w8K+5DUhHV2#9EP
zf4gB3^vWk)`34lzFteMyL3_~3w)o1P`3;est@q4$l6s8`_N_L6zFmv5_q2p~125lB
z`Zt00$w@X_8}B7h+o*%%zhFCY(<ix3LLzpzM7+l0TCl~V*RG_o86XzA00m76`$){6
z^^--sr|h)pF~`ir?K&~bNV|x(T_XDA_~Ce+u@J<4D);0tx_0Xu2KqVm)hZj3vgQbz
zdBN$jK8B|M7I!E7=Au-0`n6$TOx*1vZLf+hzhb9Z-{^ht0j8$ZPI_uD{Z=%Y?`YAa
zKKD{8X*&jA>fbcEdWb}vXf4Nb{+o@B<k415Od6hwSFs$JX#hT}%#UNhxdL|h%-t(?
zJ#LX<`QT=VdVvOqb3ezgaQV#W7YZfgTvN<1{20IKgwRJy@$ek`LXqJi(4aROGvEGR
zyYSkg!Xe5)ZCxHtu$(T~ANzC@?K!7o^`t}C+NKds-g{lD*_uDlnKe~v>}y^Cv$Rf`
zo6l6NU(USm^ZhKcGTPHYUj^QMy{#`vCL`p&qUxPwt$*RWS)H-0bZnHgf*Z>!@_vSP
zS$=xH@R#GLTfp^`U^$rA04$BIABOsCvh`zGC&W+)`5}qSYwkIb9kd_bh^Eo4M6M1n
z%og$F-BRKYe6M7W8UTWq_w;!QvEL#f)~F~wK<DQ|4uDRi=&#+fg$RG}5uUqQqNLa)
zIY*P?>RMzBJav@#$ApZWT}9qKiLr%X(sr~DQf^y-U-ZY=*L(l!=D>%RrPme$FUmY1
z)h;?-z7AB3KrxpquKI88^0GK<8$X=;&CQj^)u}vt`=+^)A^;Q|mQ6!WqCP`s35!!t
zZf~??uo}cgO=tx6O^$P^k5~WUwc1Kp)Rr80Cc|#xY-as5$3SHvF^J_b7&}XBqBYRf
zN0dEGQ}O=Tu0ayN@?wjs^G1-rSDPxVgX598SyCkDtZi(yz{?^l8<&EV5v1q-z*rWs
zkeJt*mYDw^JeTWinB664g->~0bQ1QxHDVZ~Cn&yGc&J*WWThYsf?5p+wQ=8fa3JcL
zuF|O?gh5{4=wb#DELv5M#S_RjobH5}gs4oz8I&iZ*h-w6hZFv}O6@M_7P21}h^@;Z
z4<e_?Q^bOtoSa!GV-_}ymAAeF1=KChUq|>8@&}O7hiVxbz6iI`S{*(`%Tyu4i`}3r
z;}ly3F-hV(4IPs=b8kE5Vv5vfKIP{&Dk6##e*bsUu;5fIi+$hI?<7_6bt&6qy|nac
zYxnc%@@Ew)4Y@;#ZiN+n9aY$xI8;_t0TBcx>E9It`(7{oJ|W1_?oVL)V1Y=t)yJ?y
zL*rb{6bqMp+y_?Vl9$bx8}t*P6En|~28V{SF%9c(B=E(pu&^UHqoSfnvz$#IV}f?F
z#%f{EuOm$dWxlIc5z!!^(BmVzutHiF4A!RHzPKypYr)%h?>okwTk0>EbddUJURB+S
z6eDPiIDQzkdn5M0di!9kP5TQJ%I8(w>A1=|q62pv2;Z8)c!jvi$47-%e$FooO65^A
zm2Z%E>$`dkKbkl`TJuG8Fl@v85-!Rxj<u9mFO+goj3z$7{%rEo)<lCVZuKA|9R_Fu
z80C*8sH+KDOnvLe&Y<B)DQKNvm#~~)yP-*nldS%B6o~t>W}y)D+Euvw^dLINGr<C7
z)tOWN`M%GxdPAh&RA?hiU?U}U%l34>+HU{T-fYj-82eCm`%vTmkeSjSSo7mo-ZGbh
zSTtpBua8M0Nx;t0h~7`;ECpF?Ci$09SlHc1xQVd&?U&c<t0pd^Lg*dBS|EsYo=hp?
zY~+!$sF<<@gV6Sf@{^tJ-l$reUTacEkmJVz(_)JpIyC4D0Nbx4Ouqu>sqZoX+lOpp
z)J?b0O;t-L{YTtyICbs8%G)?FwL-~!$E+_w7rEP1a$QQT*h(~sA1{BNnGI1^L~#Ez
zy?}q=!iJfCwOi-MrDhdX-guSqdtQ5MUP~@BDnG%z>x8*&h->E=yiuYBU3qhGvikF@
zN^A;n0yM>!*RVYXCZr}2-zO$db{=hWtjo9;1eI+%7gUlr>I_AZflr!_1uzV#gTn3)
zV5t(d&H4)PrD5Mx=x5dfvkUVTbADgnt)HAEOMN#mpt&FXTDHKX72kKyF?sl0rCqdK
z{smeYQeOWr%%6cF@-{m=HQ|R5!n5!H*1%Qseho-HG%twrU2#U}!r(w`9Oo(@+XMN<
z{g-JW{uFwOtVbT|DE?FNwIzOyzW%MFU=HiMMp|V(8_DeQ75%zx`=a^W^1@;<@CZts
zzKWSmIFA71)yj~u;c#JTqXpRaJFaQlyCV?D78>C@`c>5%m+9{{&bz^AD3q))HTSA6
z>Ap|!ybj4A*)v(CB6rO!VC@LI5bU@}hmP&PDyGdBp5K4|y%0fYj9lwEDd&-WegL*$
z6i|j8T2+1#{fe4)642KmJsre!=+SpJw>~0*ENU1hQNpVq%4lp`oVY89pO6RK_|ZEh
zN?TZLmjn?1<XU6+eah)#;w0~{XU&$~JOa|I+{R7mfBS7m$GoYnXy0YE=>vo7BgI~_
zwX{^Wm9+!rm7aBi-O=~(5`1F8WSfxo6}5Ng*4+vUqYE`Rmnj9r=?2LQ$x+8EZYTY3
zTGywWazfY4b+?7RK4gbBFP(`tjj$9vQlK=0-q8>fx6@~{+h(_O1M)^=9P4zWa%&aW
zKG;n)@Yjx9ED*-F9y25Cc_mdeE`84_J8E6u`awVsYP*tHy;6|93EvFxpC(V4VHJ+-
zWRjJCtQaCJkhH@k@y0#<aO>>jK)D;$s{cNJ=%B4!>;+@cj^Jj7>*_*k=-d1`wtO{D
zygVl9^q=L}o5?Hx%3tD&V^U94Cx0gJnFj<%o{p0j#wc9S(vN_!zb@RbAN~R6kafyU
zMNi6p;`m1M_$coAtBMt{f}{B7bG632S!5=qc{6~^ww9c3=&!%7IkxG&Zai~yj{u?7
zQm1KT#91qLt%XJn&E0&%r^XKAE12+x!2qMNQQWeAVQ-hFPR)4g6*Z%J;txrb>POJK
z?0wXDCvBq0)uAKW^`C)Hu=8Cw3Z<$g+oo&rR`~#3X1QNHdj|JiLvj7?rQcbL*lXMd
zK;uq2Y3wez^m_(KZ8P0o;=kH5x<%=T`7HP%56dKq1&GGvdaGZKv7#)xYF~_~i3tT9
z>QG_N_|9BT9wCxD!bBHxILk@4e<U2-d4NHXmEb#B0>#T|dTsjPnS(|xx2W{WnqplO
z0};@Ym%BD~MRAoT58rhhs1|H(oXdXEn|N}=___WscJM#{{i*LTDg6A@ZMRf)blPtD
zTmxDAX|=1zi*kss+$v5Igl-40;_|k(Z>By;m(5VDc95sAh7~*CdF%;Piqx|gC+`cr
zwQM^xTk)yjJvmcTElV<e*jP<**%sg;#2!tfq@9e-z#H^S^!muLeoe``HCZ_5a7vcb
z1X6w91wHK!A5L5EPp-;sP9(fLrLD&tomKPXZA;-^Y8_;Es;<`9UhN{N-lgjQ-PSVI
z?DoZ0=WS!kPyst4u|1od_1A%D<Aga-RxvKIBUgE<^$l@79<0X;L$Fo4?2Z5sm9CNi
zY4nT$4VkrwKABw`;#GN)tlnUorBO|gg}e2!-&=t!QNcE@{Qy0>455*mXoYNn-xyt-
zo%2W2IJ>9xH1RZ71toXW8}pylIQNzwf%$E(VnS^+&KA%Ji=Wc6k_}Sk&SaiG`q_2o
ze~%#p*KbS`KR>jHIo`IY(syO0Y$e3?s;E8-06AKYxPE8V{1R1+l-@#jf3GT2@Rpp}
zz1GHZ4E2?8Gz9u;&LPwu=uC}lc<;*zW^wXjoPZ}}cUsTTL>LaL-|&(q?cjCENgJC(
zpz|CS)z~^|U$G&0%RbPC8lvimX{q!itJ=T4y<B@ZcSrL34O&IQ(3{D;#ZO<EBUNar
z`61RS%KD0DI{&-3Kfil{GF83b`u2kJLmNvcHmTE5XYOsqDqf9KIEV6)yAC1&c&GTI
zJQ8rHrWASFnY1VOy~mt*&b{>3-VFV&T2fR5QF-&}82<To1&(av;qI~5Jl4ggte3O{
zaB@osX&^<cRwOtCcCe5_&=w#WT^~988nWcly6<c;UTMzFj_B+-^}}ZS&IOv?w^a3>
z(b_qem~liLx8;(?Sv$r!SLtNfZyo;|WMlIuSK_smBIeSr_A&eT=)`;MQIG~d^tcvI
zgb7sz%w+O^Mj7Twl<*+WgWM-3$f5BG@X88{Tf&A1RaldVjk4Pg_}bgoR9XTg`PN_(
zqD?HFk>FRmp7aCg&F>=*x*f%;_=7)ZTp@48AnG3xqB`yo%)%cCp)pQ391GumiXAw-
zIk@^fnB(kXV+Eg_RkX^`8<nSb95%d(R?{m8fmr1`xc@Zfr+gW$MKgB4!8Vo$DMDVs
zNj$YIfHpcNbyl$d@CAy|+(TRFIqUIUtZr$}NR*p|V622etOOrOV=J7Wvk<q9GLt`^
z_l?8)C~?>=BXg?!52SVF1I|rUsn|jqv;=!Y<44@<VrHA~lyA(M{f?Ss{Oe-hq4}qz
zPO77`<D3Ibhkt&%+;yPRd?&PM^6Sv>6+W-eGAwUT+GYe^OP4mO`nCJtPG}ot-bjEh
zCmCRLHvYsv5=jG+97hZj$UF`mm}mtk7wAB27}A?Af^^2F{iQSM7t?=j=V%_@e=yOn
zS`M9S>t;q<f3<riXE$eTXJyig^@>XOKHPahTIFxczw2JRBGN!?*eo9PX9$b-00SI}
z2#Ctp#l)|G^eeJr!-cJTw0;;3%wI7`*Y^|C5;j*e#jpLL(3Qe(={xYphSSz*TH4T0
z*}*@J)2GLZ-`*FYeL^BtbWtpz5Xi@`OM)cTqLo4GfV1sP&qn4?qCx@{yf?PW6dohB
zXIvhgAMa89tfffn<omIdd>fyrg)>=KwwSCmuUO^rrl7V>E)yrbVy1vTtuRY)`u-6v
zXX*lkdOgL;$NVd<&IkWAP0>9Kr|4#yl%V_4ew!VIm3lH4AXm`2wlp-%|HfR}3zc1U
z^IX?|hO|qJ02V#?22(Kg@#xqv2d{B8oyD83{mH@$+$1b!3#Wf&E%|ZM#s9qJgm|rk
zd@Nd6q={oBiU}2RvaFo=$Eyx>$css796hAyk&UnFTpEtApzj``HB@S7zrezz88~3_
zwtLkKEvsDf+rbwuiGi=#)^5}i#89}@+h{845FYQj<FKvhF$Rkzr!Kl@wXfZ%5q+&X
zjxG1yonrARwW!p~$SyfFg8Um=x5?%%GurD0C+~i``n`cToXg!E61c3L9`n2z;CW?b
zYiyTgY{iJtg($hbcC~?m*sbO-qm_V8Xwzr$+9udh++INLWO6aq!4gc2&B7Hc6$z`1
zeZ%~$#wI6A8}`!Za1hW8HC`kf;ToIB#|luthwr8|9p+pltgE}HQ6GN^7#hZY;L9kU
z0{5-0ndA@00$0r^NK;kSCFMa8Nss+~e1QF;f$t%*uH7*4-^rk#Udq^j4T9&!t=e_G
z<+|z)9&q`VG-D%nSxe%akSEGTLz~V(6>(==<bA82jtu6d-{1|D<{Mz*7xb0pU!-R~
z$HZB-v2^_+<NgD#@MLE#I*PRRVL{cuGtlIkt!<Vh9s0CD7tj~L4>XPMee37=EE@b_
z*Kq-aoylD^rFSSggLn9>uJmMEc_jIS&XsY*gy;oZ5wW?qMrgaK--ZJ=R}GTZZ=SSc
zd0!+G_V4KTmflY<Q05Zyvf;XED-BjIuU*gIT6Z6V0ztdhcY5G_X@^;X?nCG!uytfO
z#(Xdamp<--xU2TXR?32EPBxbk%@N(Pypq9-4(tqsKkSexNMsB89$E{~S$fTzv7M}i
zsczP9*V{Dyu7Po?IgM<H!FdE2bBiW$b=AiBd_hsZR%Zne6;0k~yPUo+uixr@Je;Hh
z!grbeEzIHyDqxyD?>K?zdC(!5AXBH4*TO-&;{MtYByL@#=`^p08ryU|09L8<uKFr{
z`?Xm0LEOq>bqFgK_QT@_LjURVvS}yRjZ#IA5#R-_C4KU`cNucfY32>^C=*w8hL%1+
zpXakk)f+X@D4XJ_DH`9Q;yS#Ih^}&u@Zna|%`hjn9eFZwLzTEyG{}K@|LlimEEZei
zI~-KEuC&M}>gs&3U=>xBC8vBHck)ZcR58;*(ck1H*K68TV?_3gc<>^Xu=A~N_64*t
z%DVg9A#G3NVVNNhgLHvgO1(K;O&EP^Y^}-lmj^+{1JHQu_)49-OEyS(hEoSddy&i~
z#pR~V9iGt@18*mAe)s8+7f5W~B0XQPB7av*&C(&70d3km`M1{&Amn^P^H5Af24`XN
z=79)rro83VWHp4RqhorNZ`G=Ac%xa}P>h-I;i_5F05}i}uZxb`_|+{WTkgcr_G3tW
zw8K7)QU!oD@~duSjcxDI23m1C`uTi?uHTr?Ptfir%IH|(n1cRfIGYdv*<6hj@Vd5s
zlDQF2{l1ML1X6b;X;w(yN7jKI%oVr{F=(jZ&_5sJAXj(HaKJi#FS&A*;;|Maq;|d~
zpMl|H>>-=$Z80yJfo3xN`hb|Op`O3az+AlI-rUopTGw~UO6$0>H@M~Vf_##AQuE%Y
z7k+NW{s6NHarG8+1z~fQUN86l(8ap=62SXn*jE(?x)ttM7CV!blk2(w`9@mI6m`b`
zNen+r-k1v591aH*HGS1~`y0MZg>)UB!2Ty@Qax_NR%E3!RI>l8xTIMCETbTb-Tzc~
zyI+jofeZWD`J7UR`g2RE-IcFmOgB|%_}(SR^hP-9z!`<jM;%!H-B?K9t)b<@W3vV4
zDZCh8yugX$44*!_cKPf=qN5&QOa}zrt>>GrUdgwQ?PX^Glgs3=vIAR(0deZImHA(b
z@+IixRkT*!csBMeXbYxyJ5ItjquJE*Vx`2y!h~WQS51Xo#b*UUB35QVTw>gNV(0~F
zX>ziuV%N&ct3qSzm7!Btvcy?)womO_pV@+p=pw^t<qw-NN}F#hV%*IuAB;43H#r!l
z3PZg#e>Zw*Kk27f#+u~dn<NtfJ&m{ZA6GY2a8Pn*VL~a&n&4XgxpO|9kZol{L=#&^
z?c-gT$=Lv3U|QUlawK_R4rbQi@f(TWDd_^NKU9tex3NAX^{fyfa>gj~EsNMvx5SvM
zw&3;z>YX}R7Bno{bhJsjNnz*i1B_q#ER0(9WlrQeYKAYO?bV;Cjq*ck`<Wo-swRi#
z@`;q>5(76k53CYV^-j_^ep)k&)X+3_5A(W+*=F8TKoB71W^w2ejqeZE6`i9MF&mNk
z(FzCmt1jNSmP&m7$Zq{Zw|fOKUvvBA<&eDML*rO>aKZkQE_}+dPLEWbJuc*y>>ZJf
z0G_blL5tY0hNgP0Xdd(xr&E^33`u<r4AMWY;Rgk(b34b@reu46Gqsp^i@cs$6NS~P
z;`+Gw47<O(Ri>!gqrih?%DRVG`Bd|qarFQ2h(lbqlLF5;$6oX?)O*+usW>vhAVFe6
zWC89NndQi<w9*xDuFxKM8h@8;*ttuW@trki`P9BcB9szlBcw{&Ex2uWu*bs-iaoQ=
zi5?Cm3V5t)@c-UrdOzsTr>*9zk;R$?i|>TD3pu*I^LiWwLxAkt<Z;2;y9}kjdm8mQ
zs?vI>FkBCuu0ml7CObiRn02q?2!r&s;e$RjZ9$V*@&g$$xnxIe;QrKp!oC=744TWx
z(@7@UPv*H}%>U|p20$7F3F<eTsHOPbGAFAIV_(Gni=7)lihSw=GZI<(!~@A<bIUau
zyxz@~$4L_(vP@0_TD75ZCr8S}PJ%&joX>aSeIL+KpEV~mY@hC*$Y()Uc~%fdIp#H{
z?CtgqYBc?XQF3}FZ(817X6G<nZ+NHpT|V*up+pP}!84CMbRj7_AF`Q+;T8uM41pQ?
zodo#f1$L0>O&gkKjGrE#wpY;8321lm0daK8^LQ|3%qwJt6iLV&psU>hfP{o?Rc}@3
zo$z9037^A%s=9fe_o}1PNjgJJv?{TY9UGm>CKd2lHPXxro&%Qw0ZHx12tLiQaL}sM
z_@1tCfm=J?g=M&3oxgZ_JZGJ+xcmwZE^d1c%y-C|f_!VEhQ4PV1tpzL+o6MY+bJxm
z|3IQ#&qQv44Hqt~F1!K}mf%`T0EOcM%i+>;-r>?(47qQHK{^m@wGG<!q#`Rjdyf10
zA{Ye?R>g^6ool20JHyMBBTnaeUm0duvOgOe-7(XQ?U@!)r6@>r>#Mqd51#2)Ec5Be
zuPOOn%EiWnIxqoBuR@9O-Ikm@=aky5)xzlL(^lTt>;Ay>N2=!t%e3!GUZF!uxrRuM
z(Y$BBEr$vmzQdJ5rk7x+)0Fif=Kcjfitev_uIw@)Yo$aE_h0@}Ue#;Yy)hinZF#@K
zwVW(2U3)jTlWF?zh-sUT%6|?l#7`jB4n2s-HYW0O1_mwj<b%FRkNv&{X12YPN4TDn
zgcExlz#{=0p`#Z&RCPgc?VF3dzd*O&yVz)t)xJ27{uL8uJPaD`8IW~%SShXK-~A0d
z19uFyfiy>ilEqs?9e-1A5o^PGsVS`#y$M)8#0Qwn&+R3Z240W3`s3zxTywv)ZR;s3
zznK>#Iu8pFN@F_bus2pHi<HvpA45ZT<6;7{FGAs-TUZNOSjd#?*oWKAp9aJb0N{2#
zUSsI3v)iex*YhQe0*%qM$#wqNhlnwfjkW)W8MOJehPF<6pi8{zuQebXzxuN%YLyO=
z`xe+ped|kmOCEa6(zW*vKhJ|m2+zF!b;omjLU^lNU?lr3GK?5?J5hqWog=S_DTXKG
z5G31wkED~HT28#`b<6%>pzzGF5we`6E(T0aRDRhne@3+TvMgL2Htc6h&dj~F-{TRK
zRf?bxK)nd{*lZx;fWm+EYx#(kZw8&uiW1PYR!~rIfkJuL23t<R2KALz$kVi->o^h;
zZkV8|Hk7s-6!`8u&F{YbzI8Q3x%Z-R628=7YvKU>xP>spp0IE{tHD3%9yafy;^qy$
z`&kc^3iHW~ih|b_#8ywxRH#pBD%0={=3wi>KxeykF7CQ7of%^3EKVQT?HX<>`d2(&
zO`A?t?op@(4ABcp3qcrQp7q?RREw+bhik9NoK;K)(p3gQ5Y1msmvUK;G^ip&Yc}PH
zPrP5d1~rMP_&YB#S;^8EdJ3RMkoAcgW5|9pC#n84(+H3U!#XeLEw2`q;?^<KC3J?m
ze5IW3rUBmE{6ExgL<VjwO%|9GlInY34RWkLw@P;LniH|Vc&BJrK(w%4KuqiFd}bvy
za_)*eduwk`_uB`8`QpEKFlGG%?lO(t*F`4pyv`Q5H3ly;Ji)-QQ#eUkIy5`nJTkF)
zoWbV1Hogd5A-pc)Yykvy%lAgK?(<~sfoB-l^O~6$@mfOZT<to!?PFkk5stRJI_(ES
zk}d;H1UmYaK`d*%$DwmYDaL(dc585Vs)qUF?{weKZ-%FBMKiqP8xI$xfHQhypcpe%
z0K8iFD;#=mO%MUEozAJlNcl=R^rw-hOQWNsHA;6k1Awnq8O%cgYpO&UY}}nNa1CC|
zdKl%9x>_&^-J6D|<(G8Q)^z0KSsSXEi2Q$F4?3SqA5pWjIfdckG!HB&s6(lX2)1uQ
zQr~HZo2(2rkFVh|Q@*=?y3Ac%zG<#-pmnm4<~em|+w52i*5*&B@uOkzNYG73;B{(x
z>>ZVD127bTOa^5hka@wAFW?o5szotDCDLnv6_cHYlh=oOdwnQ7!4>iiRhybZp1^mu
zTuUe+Y^!o959S$;+_}M*L*Li6VEBY}Nnn=sAf)(YNvpASx6QiE#v>{>BHhJJi~7wP
ze<W#?R?wcm@t+3#$ars5c|PigfPkFr6gxLK!bi<xh?#FMI+w1(Io7O*IBij;YxwM&
zf3vf})O*4!E%By$Bq?3FpJnRPyC+SY`z5r7&*7BKvkdPn&%(>f2CQXOCl*CIIYioZ
z_Ff&D*I)F!>zaQO^{Nm+gZo<nRR#~CHsCFrE82G+cfuvtcx&cBU}**iuWESzRuSiG
z^}*^pjy)6FH}o@08~sTudAo1=0(1Ka?-0Yo!_KXd+^P$S#+B1S<dv6#(fy6&4e_=W
z^QA_&am7h6fKz-jwfZrQyApIc2AxPJSN39vSMQ9}?|lU1g3FWj`@SwuT?N{dJ1sB^
zf7aY(F(Xbw?}@j8WPjW`Bw${naC{>u(9y@zb|Y<`b2D_-APJ2Xwfg(F_uo9H3-Jn9
zJGJ);T&@=!qq7AL*q3M~1Nx!xYYaMOm+I<d91!LNN<jp86Y;Rssgoz;D(^T?8hB9`
z#I1gr#mPqVTB!ftp`$&Ot@yDHl<RYtI0e9cZTJHliA36clTwfMrys;-Dv*>ST#>nP
z9NI_)_-;YS#GAbdFT)<BCIk>lD2^-l-<el9>^e^-iJPqgEnM;F)1x`wFxZN%06CT4
zfBC}ZRv{*LXo!%<oS78?oC3XRGUYPMQ*7MT3HDKUQhE#*buQaC#GI3}>|A^EM&W6+
z9IUxut*TYhBN+JV2p8l3##rw)TT^#N6PU<<0=G#lCOn&0n1YXSKyt6KAC+#_#$zS|
zc8>|xX${l^gGt1DXVBnSjwGx6QI$S7=H8n(0TMWG7sJfjtMko3jClQ~Db&(uH4g|c
zCtD3nArv~Q+M_f*u>|W6cyn8am6df2L8B;mE#8)Tt`77w3LaBed)~;kjhpHsfNoBK
z2jq>I>g$6K&XYAjm!bm4Q#xq`0Ymr|5~oxg^MjNfJtpfmp7wBSYFovvlyBJD>*(MQ
z**JE#>TeXFrlo3-y&vPty?XY`x1oFH(pB3p(3X@<?g>d1LQ}7%zRW*%TBDP;LU7#y
z&6@@MYYAX5Yk92(`z}KZcWb<o{}dFrPbXOIi+eVNZ8>y#Aft;kZgGc;Gfq!F^7s{|
z_gDIz#vW6w5Ayy1?T}UiPwG}0ylE2`?S!>RDt&)4-JaXfWLH^UtP27khU%6!@)|AB
zE=kk_B&`N)_5k1;zDJK(fkM&1k>dB8tG<2|JvM<a@cU=rsH*4uVt~|Wfb`jdz?Rz0
zMzZ`^J!Snb2`k-}L)vfC1PzLu*iUt#xno97197U(knjNZNyXv`*%#f=U_MtsU>CS7
z*ZS&u<+7XdS_8t7OSB?;ubPpwN|Vw#m{B<5ebBWah(EOVRr`DY!)FX)DCOBNLg5!d
z;TwH|=wE~gWr%{h;4biUJ6yvXZ#1}sC8<ns%|%y~h)cz?ZVsksRn-B?7o-<vG}GUw
z{Q_eTTYzyci0}<Y*-bXL$Gze!EJMN&j8!`hgwt?`mHdKZzRh!c>*L}moRb1DR7Y{k
zwMJNTN<LU_SP!y}<W=g8uyO28z0l?<?=~hc^{Y3=%*-r+%o>)53>ss}zL8)U^do^(
zN_&DZx@l4Colhju&l5;)QE7Knto8QsSrrsGc*!YF{OWEix}rwTEI{}#r?{lbi<0bS
z8&-zXTR(PL<aUz$DM4ke4O?ea97?<ESI9$rzMs;g3P*~F(>)>!U+e4_djGx_q<^Vy
zy0xt-+>j380mclmFO<*e;}dTl#8dAKB!AYpz#!=OVeJaHS6Q|&j?8jf=-`Si%Ec8r
zqnINC3WJaVV%-6x8+PlfAt&C?Zh7pS18+aEh}UgiNgziJzasd@6=5TQ?+@Td6iTx;
zNZzy#C=Q}-BZ*Hlu&t~RPATi6(9^2zb@kx?a7M!S?uaC8X@q$#c5gM0zF(!sBp4ef
z=-3LS6U$@Zx4o8nhsUS(6wzoj>sU>c4*h<xPqE6SGcs_CB>_(UTqp$bAjZS0(0O-`
zT|jA4;Ob+<i3^$=-j%XRH91+4743E|d)?isJF(eH?AP#zQBMpVHjdjy>XTmm{`YQa
z?oHlliYF1xq>eHKW7{c8YoPZl1g29`j}7wI;NO1jv~Z4&Ot!#pl|Rl;H$p3gxcUN<
z$GOGaW2hJ7iNCcc%TW~lgAST6?MF06tb4dmgYj0u2Td3csA+P`_LwmG0io?e*hObA
zZnO)=3{lCOr3+mt0a6?HMo+SfI4S!B>yUy)AmH<&uA`pP{WtD43mh3fW6obtjB9jJ
zey{S}sh+ng6^4Yv$$G|}SFpa0$Tydup)r{%8*?s6o<naQPdXXB2M-e9ZyUGbd)}ye
zqVkJ-$Eu<adk;80Vx^4#3A$F7Zbbt1nmAZJrGpakHeHDFC#W4R{PcsNOuBk>dp@AF
zpd2R|zt(-f+v18og?6^&m-=I~SQH2(F%#Q(bFL#S`d8Il%<%ND<NQ~B9tl}K7*Shl
zmx)Ml(aS*Y=6((9t$f@vF%|7uS>Rlu(X{WxC-by~w8}fZA7%~T9o8t7iwBKe5NXX(
zvruiLTALmnLtR2pZ*W@sz4yV}r#$5C01;36bPb?Rs5YDW5IHsJN%Kk3WguaEvAVj-
zv{ru}Ai>sj{s=UI=5i_RLxvzCnB&d!YB`9jL!?}ROOH*CMY-jo&KI>@akUDUy~L*q
zUCEJ2LuXIeg%1AnH7>Qu>LLW`mk8rpD3y4AgLsia=e@=Yl<2fCIsuJmw682pT5X8B
z3<@S)%8q~J#B<8Pi+r7Z^wBeyuKRgq3T2N6?u}3QD=I!!@UB0Uu^nlRG2{BxQn}Kk
zF)j+EUn$_w3jQH$IRWQN8-an0n!Opgqy}b9p(mnW8wpmM15rv-ADQ|6KC!%x(WLq*
z$jdtk`B8R9b^$8ll$|bP8&*<1`wMZT(YJ2%U<p~;C0TbHG#QLf+}jW6RvUo#Nx>~_
z-6_Z5PdhwaT*k!Gr+hiYda=KDnS$3SKL(-iDyo^IhXbNih|AaCyh)hU-8t}V0EK~~
zDwO&6^?9s@Ho1=!o)mZPutA;zFi_IO<{ujgJ5r~fAJsvE4`Pa!X-#dh@BRZ-5Ch`2
zAImqPV=Iz>YhOGH9uFK@zBl@~FHIh+0A6k~lm__z3XsY#gq223_kBCK9<$I$0Fj2o
zoKCJ{kFlb;+$+mn;q3?k3-~$|V(jSXh#cY3UF=EEqxbq2<&R#J)UJwU05ckh@EX3k
z{i92Y0PfzQn?$hQyCAp$gp@)8x4x1912<8+^jz8Zk(Xln=9Pe@2)$_REC_|ox4+S`
z`Cu}CLeRVkXa}w5qLvkm^+hTWyB}1@rczeQB7n@7iC>U9^M6Gk_~om3_rx8Kco!zJ
z9cUmvE*jh;N7EW!k=1_*zVpH3Y@A7G(H`0~U@1tu;*pAzw2Fjl)brdNkWSBt{}hRP
z-hp;zx0(a-^z`WtM8{<M{-|NzfZsZ?My-9ESEjcEPfAy;t7*8-ik|cSy!NA|NuEZT
zrmcT{*KUrod)2sG?6Cq0!wB8a4f1;o1@U0O85&;jaT6Rl!hG%>8~o!^)BB}(JxMGb
zlMhl674phctQ!m@ud=hSndI?y=6)$V$VYM8>P+0jQ%~4s<v`~<WeR-ufDFG*jdd42
z`5g5^`2T?xv6c4*TH<@bu?H=~3FUDb|5u9j@lTif^NbYJhtEN~-49^#z^u7a@YzqE
zW*UMqln!AnNzlM5RD<65T_bB)(fe&Z@R-KQAlp0cJNHAD(T_FRuR+pjuy0#NfJ#t=
zqTd(mPDA_)_R^(|;ah-+SOM0KCrBLE<y`&rMSUE&>;rb`k`XWf4LE%V9-Y(eo(Af>
z_u!y|8j0kVlkcaaOwK>%I~r=&fdK3d91KQga5(&HA4|uC5xgag6>c#Hr-M$Hw=46}
zJNL|q%<MY1idD|D+YP$A$WN*|xI0iIw>Advtyzft;{=)QP4O-jB2Lr0rLeTg!%dTk
zO>>6G#|@|L4eOVdU_jZHy@FkqEB*B~sPWj#AaPbfaYX?YeXGd$lpTe_F46)#BZ9`(
zwu2rCAj@*o@oH=jbK`-p-@uu%!*azcCFmx|xR`oME-amZNHoC+rcn+XM%qeY^`Hsb
zsp*54jQp|icnFJq|5Mo+BLkWT;H%IakibmL>PuOJHkxVRw;Si)SM;hD8bSkS*^%EL
z;a6RJn>9$|bekQ#cjr1h*Me3yKI7tnp9zW+-AG9PrY>91Gc_GP5K?rRcuPF1D%x3r
zEcw=~VdPLI=!&ISJm1Mppa0wNTAOgw0V!dHpCbZ8V*fWp-)zw4QJZ4~*d=K3q`r<`
z@mh9sRWm_r%|hpVI?HX4MZ?nJa;Z{3bD3uhObWb6??OQJ8s_3S`bzyq=e2N@7RvE&
z?v>m<PTL6%bXvDH+08=39#nq%XdARwfpnHvTm6{jO_#zkYwm}DPH~4)w%VJahW1$>
z3$mnr3uVK(VJ36gfCTo18+~fjs@<QB5J&aWA&yBPsh%3SL_+d*mY7%Mg3sbd{Uc2_
zC`_?xBQfEJ=XMn-rmh7W3?4M7HQ(xRO_O`t%)gt*uQ+(mYyrchZ7%8$ZQ=a;#78Qc
z#7idZu!f0&5$R#`{&d}gMT8Wi&9POJO98VDn!ZJAnNPYe#EZ1$6)u9eTqVG(Y1+n)
z?4S=4!GPR;r!p7gXTwHLycrg}SA))%^)MICgeg9K=DHlDGcegt*`c+Q;#|Ks+=+DZ
zBG-*H;;W3ed2?4QBBD(pKNTCbFoHafeya2zH0ma;gkHr}H>&64E!@D($E)PNmV%~R
zB_@N&>(*D&F+mi6x9LVJ#)VIeY-`){_;+-xF8$&J9%QR>g}83Y1J5QM@&^s$oKoN*
zR~=6qK6@||e%_Ib>RU}cWv5?y+G%nXR+V3BZi*&!*Eme1W*aFsUDHVA3)=Mids9Fn
z6TBnTkalH;00(-<Ikkz9nhP>yw0XW_ava&?P$*o{&Qoz4=Wo04%=SUcLnv?+j!V@#
zdRoex>RZt7LJ#E}R%X<layYmuE4NviZn4-!ka;%|A>@>@FmgbVQk4<`6aFCE90n&#
zfD?#l3Ib9?7kSl#`Gj2$+MdT4Fj6Xvxs8_n0ko@Utttu%ZcS4*e0M>qzv=_X5U-{e
zHM&#aznEQhfoUS>PRrFz@C6(4yO%ZTC~iYv7RZrl7=Z>lQDxy?yT3GGQC7y#)>e@=
zCcWO7cOFalh?A%V{-J;yX}Y;$6-YVz<_E@>a3py>m<IpzyVcu(0!9Fj)btq|GfK@-
z1m-`c+~XZB9Y%N}>#B}hSKS(udASsWb^AbO?4T|clLc&f5BtVdX~HgbG?llznt(Ye
z#gI$Ym}WE>J{Ixl8Xt{ewLNSLm!ZP$SZzpgKcUG*(PlL`T;nRu?{=MktI2T63`U(6
zbvWLr<fZ@ekQ04Kv~>$<GMuNm&rsovT>1{TND0|FKZh9x_+wOdfKbM4ui7~<>`Au+
z&N|KRoJ1m;{X>FD9TKQ<;?KNGMjL>4b21by#bR04Elm@7)xm3QzXG9Ox-4*%#cAQ-
z>(sIDX6YCZEL-8d*e$#oQB|c#m&y+~8G~Rf5Xo~7NgB#yBHs&(>`7OmM?~@QB^_X}
zHC{`r;5r`B70pQWgX`&a1;0QU=ibHwyY;9PRN(1++|L$=Zr`ITUz^S`{?L$W0y7^0
z)ip6~-5)@!UA_cge$ShoR-R9;)-3p@zlZC$4|{Z;*LF#Top;eh;;q&A?&6||?(j)t
zK<Z2lzLjYm%@(y~U?8=^XQvJ0$tj}5Ja-Oq=@Loi4y1<G-e5Ywqw8_g-+W*@K64vV
zM_Xpy5vjUA+q5R;oSA(P?-M8A!X`7ZSt`~+d0@3fW1iflE^QaaTmOJq?`TdRIhC0o
zvQ+VcI@fZJR{qfEg=lK!ehf2|l43?_3*kIW84>rp!oYQUVtH4K=IIwlyI<dqI92jx
zZj*yXE>)n~u0?=fFLSqvb1pnhI=&i#shOnl00w|-V?yBYpfEe(kQQ{*Mm`MHR$4D>
zYwzoeIPL`+vrW<V<g{fTkc7(1KC;EBmv)7VR1tXzvj&7H*F)Q2Emw#mK>=TlUMbYN
zn)a+qklH3sn?WYM2lcbu)7JO{)@F{sy>ZU-iw&|fRu;B;PwEm(v|(?UZsQj(slEm%
zLf=&l_e<o(bjEE)g}hAv0*CI$G>IBS^(H3WqtSf&a8YSD$|EpqI$Cbb3jA86c~Q-L
z_#QrQ1o~Fdwp<p?>eA`s63LFR*e>X~aCOPs?-)u&5lXPDgih)*_pAp#?1=!`ZVtv5
z@!#UYmIL@1q`Rx9pYsc@oYxJnMI8U0_hm1|LeumKB)*z9N_h#$9rd*)AKGu{+PwrZ
z_SCK;@q1fyN^&RIT|`L`IcMDdZ~F<<F5~#jxOuA0ncz;DpqC38lIPZFp%LU!X*fgQ
z`^NMoOh!ms$PLPp64C9NB!ghc9Ib4!oH{Ljnm_oR2Z|QIBcQ29vDrI9+vM1o7U$o_
zFt+qu)QpGGmA%1nJL(1+$yVy4qKr1bUc(omHVg+gHVx!&rhnmKxhEJp?+S*y0afP^
zR}h{AULF^0akBM<?i_+F#^hh@tPWOQxL*QMB%$P5?sYw&&CEQ`ykb?iR@g`atn54N
zJUs}8Ud~+xvUNk@V%ohQu18EbN5id`=yVJa;R`tbup>6DBxTyv+1{YhlE-d~WC1`B
zGaWP7-~Wz|kEY2vMiWU<N)!A?lZJ+d=KnvIt~;LU_kABbJ1Me-LbCVXB3UU#wqx(T
zS3<)kKK3SiuVWnv$I5o>z2_0xoZsX7`uVFrJJ0)h?r~k$ech54O_@fmf`txg>KXN^
zw~ZUlku(4?!tYd_?=d;!^#&W^o>uJd$_HNsxYtaT@3o{zZgQ3d?q&m1>lSxa#wd4m
z)f1XVP=*3l&8I&yY~lH~f(q+-`+*x<@_u{Ta{GQj?$PHwY1PZ`EN93E_o<?|RIj&I
z*w9_+6yo}OEA5R&j^4V`8Lf+?>-4Bj(O<+_eu4)>bgFs>(o(bs=$(V?f%J^6+lM|V
z){)t3FWY?dwMR6=I40bQM;vwbjr@jvL4JF2Gx3u9@T@`bBD(Hik^La%L`3>?l*!+a
zvIf$Q_6A^=_V&|Nh>Yh|V5XN(r3Nk>0^ZB~75_^eLTH-#r2qWj!|8)h6F|v+JL4fO
zE41v@aWfi4C4HEeC1m};foDdIG?@Q3CN~dA-CKc=_T}pzZIxSA?@Uwn`hFY)Hqc+p
zK~6yVS>1Nt#Z7WrM5Mdg6Me=sbO%tpZ}6mjX>>J%JON{t^`$q$z_T&f2@J&ha$~LM
zGTi(mkwD3`Zb}}GfSMe4uQ9jlZmPF6I*H32o$7SxX9FyRdjvIpKAp{{Ca32Teu!56
z>y*Movm*7?VxZhY|EBHTv4w=0jjhmh{pssp0@F)cW&aCm)<lmsubc;*j#vZx=x`^|
zV<?x%yvPtg5?0k2)y|$rahGV;3|O?;d}bI7PWtWA)zmJZDp2gl)nARcnVUei7M0P#
zgND8k|4&;<MG>qQ%UvH|F0HZj1QNL#aUC@@wHHG}864-<!(Bhu$owZRdihqx(HX5>
ze8t~LeLKg#`|^@I^R)foFy1XSPx^d}^y11e>IzGs+1g6-J+Ux@bD;H)u=e&#?POjr
z=RNO(Ydmz7Z5He0S=Wa%N~;U*i(Yj5k<r``N88yi%~f^PHto^c4k@seDkwl;3~f?R
ze!bX&Wi0TyTBwBDw(h(0T>JSH{zCn=dp2FX{3IA@BsUlX3y|Dn>+D^EW{%l}fd}}n
zHX;89_RAWnsm5Ek($u?3NZTDd<Q(iM<>O(is;~Xk>Hh!hK1SfgA{C%7=GIq$6&F8u
zTsiPRYeEN%@B^!&lV|>I>zoJe`w3`Kk^2jmlMOZik*8L^8*08AveE$c^yNBtplDVP
z5SN*^1F_+Dq4bS)1yaOgr`%_E?&hSYPQ!L`s<(<k74ayosi->S!l*mko{>gxY`SY)
z<mx{aVsbV2XSTB~Uw-AaOu=o53>eM=(=VKk4CSU}j|_;pJ&A)gPwC>q#)M1Ssq!@I
z^+uE9q_)rsr%vC@np2tnL*pHkN)zwAl1`(;KkBi1cZlcYFyIek;EQ->2GV<bBlS(s
ztn{nP*1UzA&(C=H3Vt@pPKw?-E>jbuB}JtruYjV6R6*bKV#qgl;)*3*aEg_8z+axZ
zQyMHt?K@kvRK1KrhY)nKM{$}{qT6R1xepvYV1LlQp;uU{aO2r|bg%~#2#Gw@ua5(*
zrqYAXe}TR~XkmLqT5iV9BZBupV|50nsi^4obV<svR*y*jc<kNr$N3-u+CM<nam$I-
zVZ-Z3T)IyWm^F&RCfuAkU`W241lOH9r|w9)W*9jp*>}8I6yJYYJ#zLA6##BPQJF5D
zT3Q)(4_zP9S9sB{<^a}zb2!30J6@#iK5O4V)|<FZcjy)FF+V|P^`_0Ub=Lp}MpJwr
z_QI;COZk%9{OLU{fjNq#RK=@gDSsHHD@;3L-n#W?+TF~-osceb@_Z(46`B??WB-2T
zpT&Ljv6tJpO}>Nm^FX+ZdT%W6cOXG15FdL+jb@Myakk@0<6Ib^!y<vmj9^NCc3MqA
zAG4yO5|oAdj##5%@li;X%PMJiyw0AE+n#+j^zpy_T%I_!(Ivc2>6nw^JXRL}u^Z7&
zvWJC`{9pMRSM3k3*j0?Ep2ph6HN{CSqcy*XU!bI)a?$`r08jC);SneLXcJeZCwl2n
z{+a6Ke|h28>^hkV0uGfL1i6#?=%&C)<4izC!XYW^8e^XAl$bm#^18N`y153WZyCP5
z#2T?NS^TxT@_AxqqL<7DwBhlQq4i(~pie+>LVX&WFa?6**|L)S;!5Zeb=Q?H*zUWT
zJ`@bH*T$h`w@9Cdz%HuS-<Q^XkwN;l>gv#gc@~&QZsVk}W^9q9%8AbPmK<y<iJEMV
z=Bv8j&94@hl#c?-EFZrYXOC~@f6=)ho`?a%9l;OXku-%Be(M|ySJp=>T_GOy%xYk4
zMYyAp`;=bVU~H!c>~H!A7)xj2Rf>#mqAja2(1Q%h;8kM?rKax*j1QHG9|RQ@6o>`>
zwqg;NvBZKrwmAJ4qIJd!1$#D{i$}~`q}Tgy!uMxyzSc+>k<4#b9UiX^;oU@kTnr!{
zKs3A^wY)a!JqW8o%f1LF>rVd^P!2*Ec|!ZTbj6cvQYG|~hM3<kQmgq&Ldf$$|LfLC
zv~O48%Vh=CGrYrF4ipxFWNHG&+h_Nag{N3SvBq<6<8qPk2umKWOky6b>+IbcF+F&b
z)6is8_2`~)XRsZc!~U}vVte9*vajS&@=tRshQa>KJrRoa!9fj{U_)yF#i7}4@m$5f
z`h&*SsfW|o7n_HLG!NT=;l-84$1FmN!~C|!?SfRdA;~lFlee-$PC!$;@ZEfbzc9_^
zTY1w4HFt_yHK0+pW*2Yk@8Z<mC||1-P)e0iN}Lw1M!5#MPA<6?&H%w%E1*v17kxrs
zoljH*<k#6EPcKh_7cu4F^vnIo)E_Fz!~5OUSnqO9ZbiUK>njg<hI8gXbv^K#xTp)?
ztI9fEdTNH*uE$S}xSLS?JBJ_Na2|iQuq?^Bs(?mLNH3#PK4fr*hZWscJn&?0%)C5e
z^*LjRK#xSP8=e%}|6u?(+Q%A}$9*16*q(TzLtoRsA{7u|6n8~=)BkiIoGG^P@0->Y
zrdz}18^#6c^Y+VsuT+|WPQ>JQ>JqxoL<$M6xmA5LT9Pc}QBW*(pG8?ZSFYmZ<1<!l
zTbE$#H~Kqx`Pm-?8(s=Ty?ez-lScG$N|-l>SR;m*fj^-$EQ&H+FZCgU;!w{UOqMqb
zs}G6b^E#c=#qCB^+w5442%k#G#uv|<2)2&R&6VEOWRuGE0DSS`MFIbKJkZ#b0UO2n
z)b1;AfbTZbT%+#t#7^ezHt<*i-p!yCE0F@zLDs)Mw@F4fXMVpaYHKqrOy3}xr=~7=
zoxI||Zvdcb?72rMV9EGDrhAMU-Eu`%43?GA=+*m@H;1%XFD{t5p@CpG8f+NCt;Ral
zsBIdaLX4+8w|s9odX6*U`-g#2B!DM^SUqXbV0w|c5CbAV@<95z3+w9)u_;#m&xNw{
zEZdH8e?YSbc08Ag)BSy<l6kzSWZoDUi_XbQx>4zK^bGuC4+02VCw*Lb%n2;7;#G?1
z6sy4N36)Fq-aMYy!}S^5B}<{!`#z3ua5}<%ej%>R$uaF%_&3!|e<gbQcNYBP!~H!R
z3|Tnt2X>ueCN;4q_(kIAf9lVA6Q>6)>izg@hW6{?_WlgCSvPBE9NzrF<)MVXb`1>b
z2~!w<D^dPM9kbICa!v=YNh3&%Q%GTJ?eF-k0RF=02K{sGCPMFPJ@`}|G)a;&$sp!K
zV_+cvD_3^jh{mNtOB^<drxPNVPti{^gk-Gbwm0}GcrD5-k=Ycyv{1in01fdS50kbu
zmUTZ<6>o(RX=?HyneM$n8@+3L#60t_1L5I+Ke{)b6*!ntp`Ir%+=8#Fn8ao9HfiPS
z;W~iRJ^?_167ZD2fsdgvw$*H1{07_C*x1+>*=vwnxGzt>9v(iW`XhO!zj}!J&kTyI
z^sNWj28I0n>oU9R*~|HBs>4k4-Mc7)c6ZNJ=)VLtC7Jqz5p3yf&cFWr8Re&HJXpz8
zv$i-*{(kl;lvc^fFh$MD9VG>YZTSK|F9JCChQP#lqXEPZVoVTpUO<~&zy?7X@iY3S
zUx+>pxA1sFZ>diHBHfotf`xVF5quUBfa65L?Zq-4V8N=4KbSF9>pG^8P~fFMBH-e3
zD**}o@2c@sy!L)*y!sLITS{4Q%4>|(jwfGsn7*)i@TN!1oqea|b8BBv$GhAaHh-Q}
z39&IYt}CK5=&Uz;@qvbic#nyOopBZ}Hb?z1Mo_*|$7BW;iQD7SnaR&%Mf0I%+Pq@9
z>K=w}lS)>7ZS_N}lr9EMzkd_&m*0)b>WF1{?2N0pvp>Yb=Wno^=Cht<;QiIKX$<q&
z{{}4bdd_{Jn`T!O+=dZ=+4@}<xhi(rLnV3k>oShmr1$l_-)em7!KWa*@ff>TSk)97
zTVpWW>Eu>_&Xh+sqPji_qf{|ZnO<@Kg$0&R_ikaThuZYZLN>6x#?xN?h&yZci4s-Y
zFC#(}=+o4=R=hSx7gN1vd1H@P^D*1)Tq+T8Id$Za<To3Zke*{-H3bSUeX>jmSB5iT
z+z58ray+MOa7N&dc1zCH=<Ao32MH55w-Pc*vCaFx<SU3V#$mo?;;@u6Ar_@h@e$lG
zsJ!h@qk5?!o9Q|sC-Gp6>9CEuV6@|u&OD{*M3azLGD>gfz;^2yjVR;CGw-?!oV_s^
z%?dBj1mwo&Q>uuZ0%nH?mK7&%*D&w#v#3a=uYR`HIs~-Xc<P=r)`iO<q*8Fed43ys
zo&5Dn$_~K(zxmRH*Bx>?8j^KZN^#y62p-ibM>2!8xc%W|nAJI_G#sQJeOB`=@05Bb
zlQBfs;3U;;!+muyVO1ukzc%nKLdXz?RW|H#QamHv+#+bpc%sY8O|P^9Lt-eZpwGP5
zB!TL68n*G5olk6>?zpIL7b*QbS~maicozJKW1_|s6h8Gl`Z_jGi_MgdKU2&Vrmu>7
zf4&TMj|u$YHP6mjGBJgJ=VZ^@9Tdo{2q9U+bGI%#9M)?${@^9BwGe}{vTCer`wkX$
zhr2mMKI|i2bg)9|eTO(|$_TZP+uziO?2rWffmdk9NoN~C<*}-n<d03ZC;vU3H@A$~
z7JwXi9d%Jw0hNr`2!N2H?{wVmRO1R#MlKeEpg^4GaaxjT3?3Kt*!v)HJiu?Y(9*Qy
zqHp;pO%!x%3@bC;ZDp`*#Tp^Dq9P*S(rm0o|05iz;CpPSnU7fe!fTE`$4tS>m$Hai
z6D#G!FJD^J40kIqP~?!!XgimratIW<)(>ekqiUyxoc-5c;#<TO_KOzNm;$RNLC%HN
zRXj)$qP|)6{Y^9EoCvM}IZA)=lu3BiS@AEerN^eEiu;o_v75CwENAg|w=4&^QX~yv
zbSSJ{MpX}s?NY<GhY5F@+rzM-Wz7gZ939&vWT#xoB*rwn(#9-%F00X2S*G43`O?nr
zgN%?<35b22SQWvu0&iq7jLd2e>OL0tjffROG!Ff+S9m=S^v#MrxnY|bzC12#$`j_A
z<m+o|EM-_IAU4{C4Q~S{>d1=r+nFsc)hZ4Sj?T~cqigWCsHVYT;r2(bil4_QpB@*F
zAl;N?gltHso$4GNLiId!8qy4|+)#DT`|kBV316_u`~(iEAcV@MnL(z0$5djEsAhJ-
zp_tIw``)g&EA<QY`Q0e^-W5coi|A>8m4)#Zg{zN^3$7@Y2q|11^6f39yaIdI>bEsp
zoF!t{`x!(OPi!JuDBJtbpx{2`{3o#5MjngzsoFR2$dCVg_+|KlJSMNt=cw|I?1?GN
z>TXF3%IRUwNp4H3n2B6G=szXQY9%(|H=8rVL3Z#Plm67`<bT2xwB=WeDfV5D=ZmCg
z)_;3)tpKi2!P=U=1*{<9ZM^kq@=$RYEBh#X_X_bB?{)|A+D-8^js#&~MI2O<N}j|V
z-DFR-{G0qZ)#z-dKf9HJr2Y9qF^nQ(P+%tAMsIl6N8o7v9rzFAhC$P*Vu{D)`cMA#
zZY^Yg{KMtHwo=_55VDNsV^LLDlf=XLk95g+A~f%72BZzciE8nEszTfQ_x-<dARccG
zhy;7JAOW~2o}k3y?@xxdOmM3faEBG$d7&z}{QCGyq!8AhTw!0{fJa!w@cE2W+)JCN
zM{86i(O~56Mw#|FTq`(`_U*tH=w%#bX4R2cFaPRc%kCZ!<EKo{Gtg2KlOTj%(A@tN
zIb2u2TobmB8|G4mSxvl*RL0Pil#`jXxd^g32~#O9gZ;q`bESc;Uh?KC`faWt%RYvC
zF8=7aw@?j|;%Ze7y|w}Fbe=Ue0l@?5$Nqq@`JE*jvSb0owofevQ$hhPrRvUJj@rx&
zG#zDW@~~bMo%h7?qxw0#Tb<frIxil{%~&!`h$yf_HXYCn;CdU4X?`AvSZMn-caYpF
z{s<R1Xyn@Mhx|1K`OCqxiAunx1QsK}+_M&)lAbk%QRU1zpmj}J$)9X%j?st-e*oD$
zV&U)dCOIMC_v1A9-b*+Cv<KrnCfkBZW7Rp7TKqo~;3i4<qzBf$2iLf(Jk(rEo46Gf
zND`*E9t5;TX2d?>WD=iPFrlL*R_n&avBXL&+uxvm@nVj;R>rA8r+<HEB&lq$Mu~JC
z7fDyt9b|d$gOq&DZ5#G=#Vs#$`H$GpgA13|KW)yd-6o*pL8SmQ=*RoBszRxAHi(!f
zUbh#U<gmex5Ikn?p&QX4^>nf2rp=7bFJ#Opz-pO-WlY7YE&|)XeSy|v)-Vv-EcLi%
zp8hc!NxG%U$p2k11_}Kdewjs6(XaD<Hhw~Oav|iaB|kk4>=Ch483&jzH?OtMIOG@@
zyM8)rBg}B|%+!8ERYBnovq>j9Iau(`ra}$J$_QlgCs%hi?tjbAw|}c1u+o`l;y6(U
z1!9~VxXZ3RZ(YByt*-l%>$4pdj`4rf=BGFlUH7$qp&I2{{aiJt)gTAU7h(jJG(qCa
z&aJ*@-EmL(=!cqts5%^cb1F3lF^@f~Ib^lD>Z)0CZxwL4vl__>b<TXjan>(5bM5Et
zt<1fRx>|BOY)4B)iM@K3Xie)|)wl>28&!cYRm_FO{4y${*;%{T|7yQM=~&c7Vmj%P
z7ETI2T+TCW^YA}@hGW4x)Na$hTttMQaJ%K@F{`uZjm+>7xN90_1PDF%7~<k5a(L6j
zo3$;{ZXbi^U9jj&=3$doGWCBpie21#ZNuh2{?}wi0~K>gPMouK0Lx_d0Lxt5pCtm8
zvN#|R2pOQs`&BRMckbjB>gRYzHt)5gZyUk<BMSJF6?64gwVkgJzW7iH-(s^v{yg>P
z3D~yl@2Q1_Y{wjI^xi`VUH;st2jS;bpmvnc&z@iJA%#__H}$LfZHtD2VW$^E`uEA2
zj9ic9cJ57t#+Yc-5RNZHD8OeiA;Dq98&oVN%>Fz}373F)Vl>jo)a`si9R0yVF$VqC
z2$U}xCkl$_lj<XyxXFbvGJPYA=15aKEvXkaGU9kgDc`3e_gC-_Vwu<$SuW_1!#<eg
zF=c{hme>Q-H1s!DXToxMtv>(cOiY&YToe=zo-KPdghIQ3ga1HQsW>$C>TL){0?6rK
zOo{q-Zd2@J;@KT%KI7>OqJ=491o^6$v9#pH8^Szdi3wZ7+yaRsvvI>+Iw$W1{)vIx
z<{Ch;2#ARGaqGKnU$CTRUsN{Pk*jYvGo)Cq|0>@eD_jL@55kfV7)X$&DUd>IXRqZP
zKQ=gY0hhYStb#e<hY||bu{@6MS0f<+yIHU^hu$yd*OrzJc)eCq>tpMK#U}jq-kmIq
zB>RDb(^Bw&kjj3hP=n$6*4LE!8qskEcu>5G?U_1`yOj!QZlfxVN8L0>$3uKq%s6rF
zs#LPn8k+T%0jiARq!TcP<P@kd&3J*R;Q8SD-2^<S!L+Q9TWYs4*@aO3-@4~EKSl}^
z!KIFLXB{sS>z!=3w@E43))Bu4is`p5`fQdYctu>FKe^%XXj3yn{>z0u#tU7^5WqW#
z@zfCL$N$_3Nn=6^y`@x$+Lyb|nf6X?nK0174NoI5{<Z`B4=L#1i;QhU^spW?7n*|)
z_ezo|D|3fQxeOSNPBey!t;NXmfE1KVxb{c<&`$s5Qzn-ZcI`QD<qO~t^O$(c#dUu(
zlFb~Nx%1@MPmrvMiLGRJu#TuKgyf+j0YYewl2&lS9v8ZcB!(f2ZOarXr-N9^GJXb?
zDn4#>AZCV|<5rxoVNAn@kZ<i2D>mqyHPBf1qe1BEu)Ui>g_@fS!t_CgeD&&81f&$3
z7kz!Kk&u)`WAfpmrLpmGKsy8Q=og^x9~e;Z@TeaNZ75=S%%L6Wb}%=6Pw;eW1Pcq7
z2o8)KK4nPyx@<CICa0K3OP(r!(A0qu`Ff`$Y2_+W<+_>|o`Q<@7tr;*OIl;oE>!QT
zLmJv%9onBIrdP>p;azK0li1gkbsk8mg)Dj)YP<U3O+yu+A@!DiyS0gWHP=zPH_eZ9
zZd=^s;iftbFr7W3-Unn&yQcN8$ptjspFHFx5+a7jZVa)f&s%x?&N<mGmUyqu4)FpB
zb&bszBWC*}tc2@xT)hz=72F@<U}0&^^nTenw5Ip<Bl5w6`H!GHd4=FzZwFn}A3^WG
z&{dsH;2F*5FuJFdvH)Ij=$eY5*D_Ayoeva`?c@uMLBy^zn(wSiTGQYjS=appoLZ9M
zS`t{;FGTP}*h^}c;@Dq>7--TzGkTsm#C~CLk7^T>$*{;UX-6m4891i`LG7gUjPm0t
z<?F}a6-g13{`Kx3@elSqAt~k?a4!5nSQU<+01&}F0i({PmBL=zm>j`Vir*=y=7sW`
z*A@fOkr&6CG$zvk@CH4>iwAuMKq%r&NHo;fu>cRoisWj%+N+7?b&30=@^V@sHmelb
zdpHjCbI`QcjlkpgQ^4lvJf)FeFhk%qnS3}&U0p!phJ5?4$#eJlH_)|*InP+`b5yKS
zSy<Fq7Y&|(KC!+IhA0V3)YDNa+@q_vtt<0hwm>Jw!5j<UX@dHBPMGsKw}p6S#J$Uh
za59LLybp{L?q2*oD<S^)9XFN;eD6=Cn!9Y?oFariwo`T@RCCn8GAu_U<`x^&KRHTA
zjlbwlg~dG>1}g1lVkfhPinu<_t%I(J$HDaX7c!C62!5xn>~Lx|Q2eCXb$r6PtMZq0
z_?S6Ds>}br2;#;NXl-U7%jaSfhj!QCz${*`PN{?&KIUYk@iMZaoUYWg(YzIq+nR8>
z7_P$vMYm@RvEehX{@7-O5PeKpvk^ZUX|+TlW3EO_;7Z{9;Y!N6eOpC&oFnhM-QhxO
zB&y`~>Jg|WAI}@1FE$y>@D#;<f3u%Wb1qZ;Rr97~?k*JN_UB`t-`Skex(&-tSr?|i
zP1?E#cv<X&mSgp|Z{Pl|qf4$guV`IMdFdkpgyOtV#C>LgyIme7Ts^!C-kWr@vbwwl
zW0011Q>o|I0Z^Md(Ls1^LF|G+b6<Dj3Y2o%xcf?<5DInc+Zu-%oSquIPYk^J=nZ%4
z$$p~i5f6d~sr4>*nVc8ztmN28ps~lNl&F(mF)z|_jUhp1kn4Qd<2Q0j`O^4H0Ey9j
zwU4)ickCD#1p*Zt6aD+fetL)$HqQf@<Q-EGa`cP~XnN~`L+V`Pg`bo@m?{$ijaFos
zyvu&*7lgIc#Tt2I^1xePkim1kU8%k+aSi{OLyBp;KT#OJZ;HQ){KsUql_P(MNjLbb
zxk|__R?uNou|77YRD!mM(B7D_u5H!MzhN@W$+q*j>wX-u7pA0~ve1d`JwvIVIZN&=
zn&wrG4CnvG!yVj>wcm+e3fM+w;P7!2NCLltBYolNhzO!NK$uQ>Z>Bqss<18}^)U2+
z%;WI7fy+b}A0hUX#WugCWljXuypdvi*BR*p1Icr+u_NYSZL<kXvUj)AQQ=Rt2L|TF
zlhBDR-64ZK4s(s4{YYDvfN&&o*L~^pAFL|RW~mSD_TSA_wY`&11*u;|F5DNiDH(rw
zsx`5LR$UV^V_Ph&8|Cvi@DIJ%=Y!P`5_^f)67)_kDqyNUWy<;82jTgXHj_5TtkM0x
zV0vi(Ss1ie;2!*pkN4}2obvSB1)|=9LUq0;#6j9`DdiLrz4VtCg6}b;gd8ERNHCpX
zJ4^`iv1IgNAi8s@BZSvxoNXXYl>ddaj6&WUbMvQQ5{eyshrSmT6>HB6Q?K+D*2c45
zWC`QdmIqDhCfeRh2y=^#ydP6W*f<%gM?-<{CxEhWXp&<5h$b@WpX8dnKS?}Sd}IVa
z#+hfJfD#kchMhGLM<Rr<-8wzDLK^5!6d%{WOEmG?-H+hWk<g9_1PBo70e(ukQ9He1
zrOm|i)ult=+{~-=>Xql5+tcX=v#3}L`lHw)JwASBp|;ana&?i8sSe32JHyQ@(982W
zul#PT=Ph;j%z`nka8%A^W?PfZq$wRT?tTukGf)t!`WR8=x0+h8&d8iwlNdkWjjU#4
zh4QN{u#l077RYb1O(@22Dng)}Mcgfipn59i`b3*g454aN+|F&ERvEt)U|GX?u;xPy
z-<LRGaiIk_8;P-D%VG4sb@AxJ`ZkSSmyRZ!n8pIzVj2p<i75aLK2(|F%sBvC`eN0r
zf{_9La!deK<~Ji`b?FAPt<&AJ-f6}N=7rdKc}@mt+Aw42e#VCG-c}Kf;riV_FU`)L
zLy=VL2<qYK$?uZtv1WY<>#0RmJY``lLDCuQOiUGd`~F@9j*fEH*4Fy7S!)3NAvg41
ze*{DuWQ|}9stV`=^O*68HaBM4rKUl#0~C-_%9dci$8X?W5JW0{eZP6Hns9XbWcz54
z-p_cJUdDo?m^ng!k~6np#5MiTJ;XC9CfjB|;<DKE=^3Piv~bqCk0#!<*InQKX2T1D
z{K@rcWwei||CQBn2?@s1!6|71(Wtsv$_^8od7dxo1w<0uz9?tHpd<_^sCd2*NtC0y
zhfod^ikffc39<^oHM|O*2Kz9vlJs5hbt(+YV6;rHTo2p1<DHupfm3uCDC&N4g`_B*
zHr<xM?&Y{q@`7T8ZB(!pnN5j}1KpLY4(`|c{4V4;XL(*f#Aj@B6H1dvaG*C^>9=)^
z>rrMurN1?anM6ujtLNTNd@hA85$H55gtDz5y~3S0y3#6YC(soQovQPG`J22gq9K<l
zKk<6tP9MJqs)_&j%n3#ZL2K#+$KczV%K$Salt58#CTHuyK?S(tSx2*ztr6w-cf|*w
ztn4cm+bcQs)`aA()y#jJvovj&96q<r{azXuWVe#RbNjc>k8=z;^Zu=7mbp20+dI8b
z^WB3al5wKJ2V|oC5C}KOPY-gE(ynmsPwmfh)C1TotuKjfII@^9D?yYNZV5>N1qZEY
z`ZAtHxCGOi@SaBt&aI;HSl${8koVwQjP{UgNCabG;Tm6`AEB;piu<4B2(S?U$?*HD
z;?-4aP0Mj0_dc)pwUfdDK_5xRWdg36W0}08Zo_M1%4wNkjIm3uYK&-HNeXIZ;$0@1
zYJFv8fR^WHpt!0hO(uhQ%FSgaS9S^5_;?y^-RK2eC4=Xkuox<?loRzn!tgiUW*><^
zs*5g3UFrM5VPB!Ka}BgyBHMZKsJl%dY~0*0#bPUF%fXd}-5y2+8MgW?gOJz-jHe^q
zPCy@Fe0zOSk917c4$^Mk`!f0$L6@=qZWhnzH_CdP(fU>3JMwvXA(M4W*)cuKTI-Qt
z$YNzVj)3nmrSFVEwU8!XfaR4shW09-%|XJ-`wln7?;s+%D1Iuq@LkaWP$anxa0_W|
z^QH|_WNnxPK~$8CW1}jD-Fwv-J{6k&R)9)u`7=2Rurq`-LUx)4BxL!?+->K-w7nz2
zHpgPF;1sU@IXq^jm+lD@Cln~_{p{Szqo7hMLcwMPuJjCwN=~d|64!>tv`hzM`jW?P
z07JJmHmLMsPRGC=3yIIhB(P@d=etARwwyf+ulR6l#ECw;a5CLI7c)~&5nwO>r^31Y
z*&Eo5EQ2Y2C3Ew&s6<0oFs*951P@j-bOZ_mxJfu&VnWK?!8eX_xMik)SisHYlX<_B
zk(bV8?VcqLLG;!MF7_V$jU{%dTrC|}Ytm`#M5AUbM)6Zlem#oXUaBd1t`EDsi87$O
zccu(Fne9cj)j-mS554;KG?gCi;7;IZ_Q_RORObqG8Q+5>uM~yu$bA{TAI;IiN)E?{
z?1Ll4cjXDXBL}v3fiMqVY0`-TIkK25^vfvXGk&%BDqTFQ259e9-zLePTByYN^D?=-
z($qoLs*BY|iJf3Fv(TJ>knyW6UoL(JfZGr|`5XaLyy?s45Uf(2A#3lz;lJ(-;dpvw
zJ_scO1g?ZZCq*P4O||6YkI9)2x>2`-;-(de2FA#5k2p|3Dl=`EnYIdSY=8gM%GES5
zh;(xXTbl*F(+Rm6AyiZ?8;^Q}@#wz*Qvtd=MrPDL!APRM-w^}4k@KAG1r7b92~+&I
zf<{)G6<@@&Hb1p7et{-Pa{2_Xujb-p+pm@g*t5O8=Ky^ps5W!X2a&CaF3ThIijH(i
z0OSnjR?;O_5Y9%^ZNi{sc(>%CvsmyXh#t6SY>}wOU(I3|k9#_vXW&Z}ZjQUJu}Pz^
z`R_*9)ald}*sH)!@s_`R94Kn*O2zV%FHIUs3hWdfrrxNC!Wu~Hcaqk4l83lb6yrC7
zB7xMr`ge><^Kaaj=2)23F!vxqlqi63J7g!b4(cfAj5>W0xOi?@4P?A5wtvIB{5z4H
zB`-Q-5)ESZYXT2sNbW;tod1TgYZn-UpvA_{E-M!Er&4&q%WR-O8cde`=MtJjBVZL%
zwaU@4>-L2CJHWlIeQ)MJy8Q~;psQ@EE{#nquxJrg+GpCGp|8eQ|7r6ermQu2y?&gJ
zU)<co7AaKnzR>Qtj1Ogf$FsPU_ncGjTZ=-vnb#gKnu|fa^xVnTlwB^@pZ)K=h-AvG
zcv)1{o@&$i6J#~rT{3KN$?QeWnkAMclBYD>$rR)YWRf;`0IyoU^X3a;wS~GbiDW(?
zFpuh8Z^vco6;vU?)Pyz3-C!uhA%bZCttS~LkFg#@oMMClKyzUT+}gOh9sQ3qUc`q?
zw$o%nmDK$~c2;{6u<v%7;U8U9a%;-hSw^$9ro*Bmmtxup=8yL>!^?__co%9fHa0jx
zAHxeeXo<HvI!ttPU{$lyd0iD(RbjdZ_1!A!wKashpG|~c=7P^I=jv4=Nnc+dkL4fQ
z*<DL3sR%nH&BTvcCNAcRc{2HG_DC-=2}awto=|&@F+HX+WXJor!`;XNZ=ETx_$8Me
z^n?Jv4FKGpJL{Pq?h-ErG$P6O?kSVCu&?fbho<7r5|l(0moFmTuZ|s40lS|g0AD3I
zY*zF+0Ab&*UpO#KA<AX1s%Hf_6KNUXl6g;+RP<pBE*yhB$24@X-XBniUIw72M%BDD
zCo48($HcHw+!6)}iSj4iV+E6l2$PuPWQ2Z_+bhcm37pJ)kfDggh1zBcRQA}Gf=<8Z
z{!_p~WmEhW#F>0^rd!vJix}t_PBbt<0=D~TJ&8~LvG*WYg60GUi<k}WxD#OEEnGPy
zb@doSqql2)V}nUrnt@p<f=B6lABoNJh8=3;p|`iUQsa5!YLvX#Ihe)=PuJe}24ipr
zBS6}ZqG=#2u~2Wn#BUgjD-!YSDWh$jtl8&#-)f1B@HnLle^hwOdfc>5(wWPnp8~9Z
ztr-GWZXA+@L}?jdi@qTaLW2S>&c@L-#ggjF%ZJg~kcEx$Fa-tYc;pDuOKY)?3u{;l
zTriy|k(GOf<g7V)AF84F&(u<|D^f(Xx<lg#-Ep1T{{O)0yE98{xAJU#t}a2xCqqGo
znS&8Leh-|BIgJ=1tm_TUr%FJd>12<?0diM7X<#WYHl;++1OUaz7C*0WgQ}upO#jrn
z&PpLk<0RVqK)aDyk7Tv<1^uR^xKD52f4pW9GNlm&{=n=X)yGj`e7ZONpM*D`QZ#)T
zOad7Pukzy#C!)fvh)4bgs{ndVFFR-K!i5y~EAxwUju?wRkyL+;Koq=Sq`53jsi;d#
zogW2o%Z%3KbHExXC={Kw^cQ3%b_hq8fT~s%7Qv-sfc-S`HU2eOh)9AA--k6(`uGQA
zUG~1o%IfZ9<Iq8f{#~(U6(zRA_QG&LUV`b>i(V$LWPu5=dhiW^?wQm75rAu%upp~D
zpINpJt(y$q_-6p6p|9Jzr5;`t%u=<JmqsF<OoWvQeb$q2Mug$&b=K=9^${2K6o0qt
zME`0JgFj`)gAPUyXcuMAyX_|Nxi-%hbBu6S+|!l1;_p;T7n?YRt&!aM-RaeJI?mQ}
zl87up(!gATl9LBu3=}M$rF#EX6+3=)ugTSt%4^LaRoF4gR^iCwg3}4K+HCBL)Pl;-
z>CCk~Cte&4{TSZwdKg2j*QVJ&U>BTLI*_Hb*RdYzk#u)^#<$n1zt3!@P_BHH>wOe7
z`wqOgF?x6*M6a0Zv&80WkDO;ruRd6g8i3!d=7JxjQ~aIro5fgxnh5ad)6&x`hCO8Z
zd#`;ldNVVr4P&7WFxP%y<79C^<JD&S-C=8q;j21X2=ouhT3E!EGv_|xp$4Qb!`q3F
zBm$z?ohQ*B-E)YE!L~J;w!tv2>R?%&k|4YN`jF;PJv?ZLe6@*z1s0PU#n2mum@)%g
z8TUsX>^qQVFc}}>*mlu<dbvgd`qvJSjIH+a$hvaPdZPa0G5IuNGT9GEDl+8d?7AS{
z^~B`STjJrm8Q~rj@=`(7jdYsDJ=T7dy4&*wB~=@fk)gWB=csC;o(E)PMV?QBI-I1o
znY_P)61^$|w^HAoEa9OjWC3ef4e*e<4-x<$Nv5E{m1J@dJ$?94z9@lo>G_1>{Vr3W
zpz{K#3dsEm^t^$$e%HWf8wHD&io^Ft<matQTWT}Cimk-@5cp1QHpTwY21B;{*#s)c
zuJetF?D!=~b%+1^U8WKp4Bg$R8g~Y~JS_mz`w#(k+?@%WMN$<BKHfW8{kUws0|Ww&
zVek73OzT*ftoQu@PwMP0kVGV&E8LBPwW0vr3GWX8=AVs9<cAA6<T0_4)RK&84rGgV
zZzw08xhTp^-s09QoeB4E&D1v~CULS3tkaY!Za=fVE>WQ|0wm(A^(Kh0{<suF(9KBU
zn)y290K}K4gRU-K;DP*Yp<I6$d`J50J7_>QLG^CxLJ}l&cR41&q#?ug`D!${15Bjn
zYw;Q}BzIMK!jB(CTp@e0{<Zvf?}zIX*SCGC!++N)--^po$hD8*s#r|9vf#(Y*xj^S
zC<{u!sVjAq&nAaAi54vEvjj!n+D^#VXNTC=?&E~%hfLbD_|eoSC^pMBmApA)=H?(Z
z$al?F1EWn1tz1pVS&m?(0{9=0iU8ow`TD4*^KPCeQu_A16NHf2Q-w1N`>K2x)B24X
z`J%i~$9fNCOj-O_>Od`vO`&}kk8S330&42&k5Xp8djLkp*5rZZB$mcO4xa$OvB!r9
zZgbbj=@Z;lLYjIp+g=Me!#V(}QQ-2#3$g$;w^DRjIXEya?Rk(7fJEUgH8p)OH_s<y
zRKx4)>N*Dv!aGnT8ti{6pisyV6)Aq-6yWO1sl7|2#zC~-M*TR3_;T(!BTc=6(|sXS
zaMo34BI%bUT<O*>>JB=)&IYm;=o+{(9~%NFh;Q#5J0s@}bhJb3rynG6exd_Ag#~{Z
zC-{Iwv$fDF?y>0aY1e*p`cGWz?zP<EouP!3rqk8t&hE9f)eiGwb3n}Fh3dc~hZ(%m
zs2)QDlW(($;&-T&ogJHN(Cs}2<x$x`LcZ$Zk}RCt=d)iDyWM#eyz-OKOf6Hwx;d*=
zUo{e+<z?FLH*gTjxN>qA`k$(|rX&qrlwsapda9K~RfBMxcbaqh5mMjb9fdQDv#l1u
z<+5b$j$+SsIUZo&8B@a#y1I)$ybmE7-!z}Q^Dg3%V!|KdlEq5$EiElV9JCDXDY_8_
zIe<@KQ+y{)>PujMEb5YXS0Uwt$G6&RyB^yDW*<cW4{%q3pWyNk!0`YJ_NpYSl%%z;
z;6iLiMn+SpF~shM738gKj7bGND08_PSq$+_BRhs}Nzab-Xy=$@wxE5<i>_f3f2{=T
z!;}d?_kRP2Ha`i2YDCjZI<3XNT-nNv7Q*zx!NEZwVstx0UalyA&<PL%k+Bn01C`ce
z1)nESV9<C8fNF~9+Z*zFwO1PUGqoLeed<W=%g*lw>gnHrk$u<crEl$C#1K*;FtSY5
zrJM4HT6~ZlilQd#BE>d2d7n?b!8&%N#n@o5psa3s73!b*72L;+pkZ^!LCCSPfnfs<
z;Jl78USOh<ysz+rXsr@^DkCu=Wsq9oz>N8U)HHoM#&8(q6RF+2#;1hcyAw{faA*-l
zMX~(|!)m01tAPVza;YlG^TZOr6`%Z3OvqTRN=<ZUHDQ9Z{rbg}II18X@QftjjtJ}3
z#c7AG$?=^>$m-^>5IW44=GK*W5cPqzbqfslWy@NHQ95Vs!(27)s<?R8@&}}PZ3kxF
zzlLM%G!=<=1+KZFG*6cGB=wXheP$Cs@=1@sU?BdvXcX!H(KFmY`Wl5*?=WxU>;KIl
zAVBI4jL3a-157?XnRJF_|NM#go27cdz+()Fl05y8YgXuX(+YODW<O=Ker8~MH@N)S
z?z^(wA`lY}QgP(h)U=e({@CruzP^1dSLlJRx4jy9#ut*ZMN#^<`o~m^oqwp5l~wRS
zghyNA;J4aOb8tqxyr}9uE>Jn-{UNsSe8akJ>ydauy-jo_6hw%D*g{Z1PK66o0tY0>
z#E?Xf6NF7t1F_EO;a3|TP`!y<jFVV-Lr613q=$%m07DbwEDmzS3}V6O_*D#1s!?6{
zCkc~>p?IKf{etL%W*)bLi>w-gK?LGNtO38}pEaUV1?YDH8DHSn5Tkw_APsEB%3(y+
zjz60Yl@y8y`YuaKKJLi)()naNzw|eJ7d@vq=_erpwJCoT!$J&`w)1use*$W=dv+fl
z3o_;d^^~oY1Gw76w@N-QP#2gD<m@2zHvKctkyqbus44<k)T}nRaE*o!davzrhhq*<
zfm?G;+eov^)BW#YavlOMzS!kStZmGZFMuhlz|Oe46AqBhZ!SeDBlki2nKo>fcX+bY
zul$N;47?@VhUBTpr|gV0sG>=^euP!Yqcy=Xm+BpQA~5`6&Z(740;1_p_LS5PHC9pb
zg(uBv3|Jz6(h^U-AR1W86A$@*SRnQApBUO1vN)P}XeMks7YaT}*u_TU8P|XYbeW5{
zFp2q%Vw7S}N(B&ALBaY1F;MURN-Kmxa=d!?GJrtgsoH2?|GZlt2k&{xw$JEZ(3dx4
z3Hw)lfqZ;=59<@J@4G>D0e<GVftI_f1V!4QlZ`T^ZhurIDE#HkD%^u7mSRfWWrRdn
zY#+r$jt5Vw%haofE=H%7N}O=iQmTIh^)fgq;g7p-_S7Yg6wFvwfaw(VS-m<~D*8H2
zgbZ-S-+Vs;F1ZUVM}K(c@kIe_=#5~fXt6ln7zCM7+ODEKaCbi@UOUJoQ0=$5&2h$Z
z#YIsB;#vN8JIn^lCc%p^YqrlCwyXU^>~i&F3ID=NW45>>!O^WS(<vV>(Xw{lKwlo~
z)`*v5Ll!YAiT^OZ$G3nqlEZ@`P^o%(3<(1Gt-8nG8?2?kyZg6`w-hj;*a)Y95JFzZ
z@=o#*at6%~6qQ#9aOdfp1o&}4M9EMRK^lWmW4mtyF}N6)%Pm2YR#-D_f|P?I4ky+@
ziWMYzHbcJ7LVa>3CYx_Zt3Awgtze)YK)s<me26yf0e=yuS00-6=erkr9)<Os*|`SB
zhf|=u*>7DAelL!p>#F($L|FRjU<`wj7+X6G-QV|PCOEOGvxmsp5gbDA5=_oE(kta_
z>gx+18UZt$GIw?%KDrw)QulRlZ|}2xjpyOQIY4qBJV^XO_c%vE6gcMIHCk|6;D%)U
z`iaTQvlfc2xLKm7f}_S9hIJUagWDL>P><L;AEnI6H~oK}k5=>flj)Gm-wUY$*=Prm
zH67SN<&iiq*qDkvjf*M505=uM#P(uMeIky}310e>B&YQFIl?RjPy%WuTTl+sPK1gK
zWI_z_oSLXC4oy8iE#SkSSoE~*4$ACxLg<0vuCz_S3yhi60TP#Rj46Zek(7#mS1A=|
z5ANMe=P<~?XgCXrK?DXTO9??OoLC$P;>z&f6_2XrSY7;`jAZ$lcWuspUBU(`7NPnj
z$xWm$6;vdi*dweAx5nICn@Cy_&mU5Q2X=xEg4?3d@<@CVMA$sTaueq+LD@6YMnQ>D
zn+aK=OChF3^79zOcfoY24dHYT{-*O=l04@iCtG0R0Mq@QAPvPP*q;H)8BkIZN>Q{_
z?>W$b1U0+(yS@c%LgjXIPuu~TWCEP>!jkS}sy~a1<<!;3DvCDDp{D4``J<I;$KV%N
zgJW?cvO?Ljs$M6x>@?#9u|IUP%+G5trD`T=el}-PT7CV(mZAosNTz%GJ~jPl63=aV
z3@gfc9r{MKM%CgazwvlaOEgn_%-SkO`qDd5f`hujy_|ow{b%+8!tzBfyA~eVaZ6K>
zP{Er)_2>*Yx}c8Y?@kbJhSYmY<d8887BEDpB$^=PIFx**0`X)E>Kr^k8pNN>IWQfD
zj>Vn3GBVXcGfkT=gt>YsY#|AlKl#vQ@s9NGYk8tMpqW2wPWqjuUK6QnLzeVirQ^n3
z`;3)y8heSdUG{%`VWR-}wxU>k3@X_&1E2GV@*BJ0AAURqaxK1)>I*C%p^}9UbNauY
z%NpqHc2*=-jnAONaSWNm^~{J7N`$?7oK&2Qvs17tJm@YAp{J0$Lz(CevmNHtX#1ho
z;IjVeu1v0<%<l$J|G>TKzV=?C6+^(PHKGHq`0>g=Hjvzv6TrRc$M6iBKQW!y_*R~T
zN2$Q>Q5YCZpbq3PfGMBmZ{t$bVcD$FdO_>bP5<Tne+(AQ`bEph!M})nVRR~Z!Xi8f
zx;;l-%q1E~&7gy4`%|mFEg=ixws{B<MkPZt9nSQhJ&E(G6)}=9@iJj~i@(0RPx}Yi
z7uM?ugT@9O5+6KP$!{KH<n!)1g*eDw#_=Q$U|odip9y2>J+~8*hq*LQU5hiA-Il2`
zzM$(>nMY%YD1mbLA*CqgKf<62i~-(gj@y{tN2_z?7GV~i*p=blvpqy#-jso~+7(L;
zbbU>9ki8<<J)0CD<#;sT%<AyzM@kBHLSmvRFzD@mUBH*S4U8)ywh{Y(CJVm}{PqTI
zX6GhhIy~wNTUt_i`0SI4e}7Xzk?XP(s&DnBiTkVs0BaI-34;eS_5KrKoQeoks9Sg)
z2?Cdk+#_)6FNX6(R(Xxqq>7V_klzgggr?2vTdB_;BRIwXY#2H4<n=tN=n@R770L;<
z`R2eL*ezggI-0q;h`w!b&bk1o>huC`ASoMFdlEyydA9SL--K(w{{Y*Gukq?}xbM}H
zHv2Ja5v=b@KL$OOV%jYO&8_XW?y%;66S-y`_1qiwB<G2^4lyRWOiabQ6ym*md925B
zWoo7|N3O>@XHqP|S<C}PiJoes$FSa~-@tvdn1b62{fB{%Z{#(tf9$~=L_SdDzW&?3
zy0_zqkll;^6`l8uzYZ?)$qg{Ddzl!srz3K9s4q8whauAA7d;^3LuHR^zG%PVa|ZFE
z`tzit6}iM!80PlRtpul-P+7ZV(CizGWWHAf|M_OaEuW+N)!?)R<B{@lT`ueBM`2gU
z^i?sTifeVdI#+yBQuwu>=nHjPYufQ_c|7J9f<X6`1|pXiC|!8W`yS>l4*?GXQsEfe
z0^7+_*+O-|-peU)d9~hwJEQY?D)EmK+#BLC=1PGCdI;AXFW~5H0ZA*fA5~~D{xB-W
z8jcRZ%o&j{@=3#(9*zbQQRWHy<r;u&>bb)d=V?bMjI^EIpZYwk?}*Vk{ZxItIn>h~
z$wyB8G<R$lI9VmeZQ6_#5Jv&upSwlo^$ka0e8B{$pj7q)RJY9msn9Jz2zHdRwPY0+
z69l3;Yx1XS{7q~b`Dp}UQ4#mHN0$^VP5ILoh!H7n=bEdIiFN)+p|%Yi^m#+V&XaU#
z+3kd&+L+tC%{@XgGVh4bA<zC>j^uvmP7`;QnmI5_2f4%$A%{;xW|$=D?9^GCKa3pr
zpDgM+CB5Q*xco21F3SZ{@n1s+-!oFUvkEP?d*LsK)C5|~WP{eRUT6(}4R*s^RyXZ$
zrUrUQxzHlRm^JLEXAna|4PQVN`HD4`eRW<5Gm392kxhs1_;>1tb7)?xo8nc7QA&gD
zV6cK^N(WfXBj8IGQr)K#?!~0asaFdJSew%TZ_gbWFKE%5Ky{tw00fsreGjy9AIFtF
zt$AStw)i0s$_>sH8P0K{Uy_jnU*pIU%GgWT0M0!N@yXVotUMsFolq2#7^Xh=HkJ|E
zyB&7V!F7{=UNxauZ;dgn>^{j^`f|!c(H#f2?k-zi(WdJdKaj>=Q>Mk220TWGA7rw8
zS0>?k8pCEH)hs&=ED-qFRPpb&l&k_tlbo}2OHG7mCm8OI*bdp-TRR-7=J%nj?^s;#
zvPyrnt(S?*;D=&$Vi+&Mm|hZ$JbYhJzB|Z2(_j<d+Ue1>UK3Dc^@4>4#SK0QQW%mU
zAYLnKPyKHZL!Wd>$MtvpXY{E3RBH=mX10@<o3J{uPxFOudU=~L_b*lWB{jct5~A|h
z5n`yebc&!Fn^8p$y76}TImXITIn)j8iEtf8Ba<06e_gSJI>%oeGc#>oJcMgoWGz}_
z*;u3sdk;z==U(mvowzVUwx89=xj(VCVRdC>(|VhnoUCN*0X8lC0<k<4w1yqME6H7B
zk#{zYwf3n}tBsyhvKCa)gHu&z%L=2FwqwtdMfye4E`7&AoX@jE9-h%)XkiB2URVyK
z3X_K562CDZm>wIm0|HlaGA&H=Q}mv*+S%?!O4uZ<U=wMs!vqAKEy4?+kMzjJfe!ZM
zVnfM+^tJfUaOl4v4{wuVj!JDU-zlv(Z$3%iDi=NzC>jG$A}yFgWcU9=FX@xzzB}3d
zmeIq0y85ZL__u_NwX-+sUS9bndq8ulWoQ`n=~IEQB^v_KeOuPhNvH}`zp}dcy_4OF
z*ZAKYd<0o&OgJ-fA71`-(Sz1HKBd&+U@DD+kr9E$s+uF+E!zKgaQY;Zd%xcZ(Pp|0
zZU~IoAA!7PPz-tR5!QU|;PEAIhU+&aq;d8-Ra8<|yRN_f8h?sJG~5m4fOq)O7&n)N
zZTHdcp!(P7)btDuwUE`O5=e_(+X8v4kNJiajP`xwvRoepgfs;UXGj<?w@%I{<}6&h
zv7&SUA9S~#vAZ#&x~Bl0>%n92gA~4M$lBW)$*r3Am@VS!tQxFJ+nW67=&r9RG#?NR
zJ+-h_QmT`>JvYNOpI5U$nQA7ob@=(HMUoC5yAhXV<?;yJv&B~3#mw1gtwRg`{Frxe
zw6wlk|A?LCW4jJQj99ggXdoEl{k)R0&ZHx_p0_`6xM|P-Sv|Rdu0I>|>`eNKHG_zN
zsx^R3v5f^}g$SjH$H80%P>Ab}^4W}(|K*gS(XYUE^q*st`d!sfYS=m>B7+7zDGC@{
zneB`hb*rB{w%>&1t0tM=T>kr6nn2Z<CR?D382qPnoCG%KQVTchnQ!Xpdm;zWn9|YF
ziAyEn4`S5plqU^CBTIkat8plJShyN#1w<vL(%HLL80mvyq4zLRVjF4KVn*T5P^@>U
z%m3X%ZfW?oI{Vl5E~LI?_)<3Z`)?fL3x06$_%YwHM@tvoWP!+N$;}jfRFm`rvvJf<
z^7su<ARS)2)exThd8N7`0%QPk3MGR2j_(z_Wvc=OLw)k{!or4eh#M6)%i*KE?j!Jd
z&e)$n=Y#PawA8rvRP@z}Fy&a?siP*Iiu8f`F*B$-q7HB!kXTCdor#H*#ANoe6TNMj
z08P}}=(>>73-pKPKY;#<q;Mf7u7AGIMC0}nu{K)psug!y^kS+_muBh+v|uZs&aSHC
zQY>77mwO*?SRwxe#6P5Rer6EXD+;Ep?A-2Uh`K(b;L`u{?Hd8}QX2ranTNTTtuC5?
zjS%tl^z@xX`$&AaNSoEHHe;@SF~Ij^7df=Gmy%>T(Is%<b9N937%3|5HnzyoIeSoY
zciV0P2HV2+hG>xW*y*v>)YAJiJ)JW_#T54#MMP>YwmD&K=>P@bU1^PE%XPF3bt$8W
zTPN%%KWL8$s3l21XJT@x(HJQD78rKb8@fM>5rK!o49Z4HZQq`lf6L^bB_eH9tfInZ
zwI#%$B{UWZ>VXi|RuR8Tk7RFJ{;S|d?NK}gxh8~XDwB>*K!s~u%Zb>Iy1&dhg{M}S
zkG-RmODN#iZZY=w?@UhYmEM2>E|@HlSvW=2(+vbO^PPWvQ^*auqSI$9tFs=%1#Q=h
z(>d*Kc~K1sey<0yXDVaGteFW)w{I;0Y<Lb3!z$p>Tf`N*w{1_)-)Nu37@KQOJzY-3
zsS&<FFnpZ<-ew>e8ADQSa>mNTLPPtOn0Q<11K%>nG_H<(K)+4#h>$o#x=U&Y3BIqE
zi`!wwW%8SRGs~7Sj<wBA)>8UvhNyG_>$syc_H<J)V?jBF;FzW|YW65|zPi{9^ZWAI
z@+nU<5T%)-mO35-c;j@+j5KcsWIz;v$#Y;c^Td@buiN~Z1Zt`*0izx~V0Z3L*T%*^
zeu-^K&H`JuUHMr4jGcw1@n@JECx32eVp4qBy22it?V}WPMgZTmCZ<pZwu<-}75GjK
zrSZLUkn%!mSs;7fdNl7GRC66O<ZC4@dns4iy+1zNc7WjUmJ9b>)5m-65iTivX@^o#
z+%Swrr@d{YgkM3rdq22)jom;zHmja%z3R;>Q^gnvIU5i<i-z)yG@<H8i24fVvhU(n
z{O&uz3Q^JfzBbHBQ!F1MBU9u9v@7%DR~=w<SrWvj{rkA)Umx`25lb1IZSKn0(&^Q}
zgMJ#Kjj$w9yec`koQsv}jH?01f$q^^3p<Gk_PDtEH1rj-uy@KrUaCsEiL>`wxLd%!
z8~IG(nfjf#I*BHSItE$^$p9L><~N{L&oc}sp?h<i2CULQz}~uGo+^8eMi6le>p62>
zO#1_%T?BY-@qlXZP9EsBou7mPbzqOAWCu_y-_2G<(1}xk8d;P?bnQMZ5EnH92!R)r
zhzU6FNyWAp_o~0oTGZE3XfWPn)k}`F{vS<W85U*2{XK*rQi3SbNF#{ST}mq`Qqn4=
zbT^0+f^<tODUCx9ND2;J(w)*EQc}KW_W8db_S*g6?zMBzJtuy3%IK;z6?;W)z9ufN
zTD0cI9!d{k{5$=fO<_}OY^gHrdWAW9w{mxRdFfT{c*n^NGO}oX%@ji&e0KpBnz3OO
zF+Dm{*MPm-qJAnagh}-FiKSYUbI8&6s1oF$1OE+S)><|dNdgVeftQE}gYd72eUxh;
zecY6i?X>Lx;GD{}4J?qJzljVo#jz*`>lKybYl0b<igoP4ex>0{QKxBMC8PXja!bQ9
z9D!aCR;_Npea$5L;dqL6VldO*n&Tg41~I*d*Sgzrkj<d;$H^H0(D|Ewt2jp9(Za__
zlRso-Vra^YXNq&9>Iyo@wu^6GLnHiL&kclvG5J%RzmZ5J`9xA>Q4tY<qy7WA7qEA<
zfxx5;3ep*%^z9W4gUXt=Fv`9%n77E+t$L6z=zHcKc9R!7OCh`kg8cojb7`|2{ACe=
zUq;3v-0iu0IQ!-et<U~|JKQS=2}@A6RB7L@R#4dD-Ngtx!|coHJg9V6qi1Q&h6nzQ
zco$$JlL>K<vRqd3;OulnI!49jwDqedVCgY2xt+)-M|>RO`5|0@mXs21T#?^()fv}R
zITv93t@{aupgbAcd}mMMH^|Y|gM05zpo_YtSPF6Ak<#d*>}H(K<3lW-Xl3he#K+O8
zmEX2|a%#J8MaxzW1lH%yZu`xf%Z!D(Vwa_@GYu-GQ}MBX=mIr1cwX}9#c3vq{N?1d
zSuUTzeNt%K(b)uJiN>9##7j?4lTO{`;uw>XQ|<dJ3~thywK8WnPwUfQBby_2aaA^U
zEP*dJ(tIT&G-#qqfY!6S<31le8BwpY`>TP-8GR>!j=cVCk<_(YEhsgWw#b;_pGR&8
zq|CzO7Dj#V+^5Hg_*;VD;?#Zcys1BXc3Pg;Gy`WwcR`Q&!%#S<7rvx$ukBj=4DZYm
zEIlo)_}^!-Y^p8ra3i4i%v|mTnwy`q&kdtnfKem7kh$=hl9Hj=y~#?&(B6+<+4eR{
z>l>`v-;C_s3Q==q6QZ@U<_)w8^jUNi%qYEt<ahGcO?sO$40GMGi6OKHVZG67WYMjX
z*mFSw@E*C6P0=*vndga?QeB!mp^(ZTnsJF;ium$}XS~sitVw4@V2f|d{tM0ZpYk27
zX*J(g)DZyAv~!N(BD%U69oF;QRnjlaFF!M?3+DK0is;Ef7~S_gX|G4{q#cR3rr_)@
z`NYQ9-Snr3yX^Q)KDh%eu%h<VjV>XWBMw(~G41+D7u%zcejsl%aG^4NoGw~sn+eN!
zeT#+Zn@_*il4n{9{H?cskV7~$6U{`LV7-44eGZ|uyOL0axXmg`JIyE@$>f@8^JSKg
z#fpd4VUp}VQhWZhjo>i4gR@CZe(S}Lg|^UfNdaY*hYx?B?(pHHN)GZKLI$vaD?LU#
z-klC7v>&@xW!Dus)1I&8WrWe)tM!5Sf%aeIo?-6;?12WO8WsBDKyKM^akAmiFN@Df
z+<_NUR(Y0g2uZqQz@gg1{aB1HkKj!@1Su|lG>LdS`S<*yre(L5DhU~_;Qv`Wk{vZ)
zXaQpk&ED&titcVeTw2#Ti%{Kc6}Z5lCX+i|S>PHHV9~dr^TR=S6Yh*-$t1$JG5HRe
zTFx!Sm(y!><8w%d`LlrfMHxh|m7-Y&Lz5Yz<hhbi!4B3Y31fr$Qm@msNS%-v<l{06
z>i4D_XXZh)IxQ)u*m2aLLz9_Vzr7(aSObjEVAxl+Tcd|*3g+d8&<6pMmD*k;V~u(|
zT0u+4(T}4*g@<SA5(<*w{hy!$daxwmo)*N){9b<pyXw|t<?$nV8KLBb59FIe1vDoV
z6;;z@(J>Ev-(%$Y`WwFB9=^%@5?AU(DfxM}HgGfGOHbbWrOHnJL>glDF}gFVx4P^x
zX(Ir6|6Jwvf-42jo8Nu!-SwP%yDj~gvoxdejEP@ra!7(-F=kEO_$VAVZ=_#CsN#}k
zJqu&X7?;iA?MM5gX4V>t4R_hOevBe@&6UVS?oo6RhJ)pbDXaewu^X{|pY|o8pTAk=
zRrklYAGg%r;Yi6rAxOX6YYob~gu9*dK+xmiZE=k}CKfL-zaLtZ3z@x>r(V|gLRb_?
zG=OD!@BKVzeJf+^&rH~L@`Oq^-qzY|{Sw|`3f8px9dcH=9}#r(0*=#h@E$M8K%}yk
zNZVqh_@B6D@MD&saFLNvSO}JeeDuhca+&$doUCjp$Ac~vi+$5`izePeOqONi-hf9+
zGg9fv4eg|U`6dn$i%|cPX>rx{laG)2^>%{_^S|HP@@ytqV>xXlHLA~DFA~0h5goG8
zG2=TJYo9fEi)sP{OXIn}`gy5#s2N}{OhaV`j~GB<t>EGahNejhIY0cs;KW94(F0Sr
zI$3>p${O%O4$kOY*F)No2GT1|DWB_e#nv|(6N>zbR(kxbr(Neq%PY^OTEf(9t_Wm2
zF6X@w2=Y<++=mR7t`eDZWkub_2kiB%@v`He4*&4<-*>^H-%F<Y<c{6+xFGa%0JXL2
z^x;#bLzm&enr;QBdm)J}NTU5~tnDTU(*iEnHB!>9h?tQ6IXd|jDJ<5+Ze?jD5Wdj?
zNot?WWB62bjhL*q4Kj)5DF#1TfoZrkggyOop2wD;=(z$<Wnt}ORXKt=|2~~;oIq5P
zMLscl1RC#y@~_~?qPNnM(T)^lq$v)y&cYfm<3UDOU%#VdTARA^;OOi|6^?!5KOWU2
z4qa2~Q&6vcRo19Ab{W>H5+7iD7eFa0#wVKnU>6<D3~YsfutOpM4~szBv_Si_2ZXzw
z?qhtuCFFv$*Q9wHZx-y03Y_3EmYz$=-Q;+oNR4aSlXhhlC9Tob#nWj$`V?;k7tDNj
z-Wr;P1gGv8BJevzt^+d=M=KGwKh}1ksV3b|yWOH59&$kE_YjB_0yok6MeZQeMe0XP
zJM6QqG>3R)DWIM(*cSUEc^~gq)Pe6n{_c%ZqErW@q=CqB)E8wDLUuA;Dsflt)P;a+
zYWFVaBs~RES74)P&UoWl{Nup{`rgq&Fb5XPkp`}+RqlgPmVW9fhe{rA>hA9L_Mex8
z9FKH1c8k;etRJw#^JA8bm5fI*pu6S!&JwEfv}bG7skxu<N&XHH$P4MgObMh?QArs;
zAtT%{PO4D{q7*FJ)hs<crn}R(F$(3mAR60y3fXPqS}hQSh+TaO&CsT_F28^_M=&0}
z|CgKLv9ssUDpShXom~6k_G+ps!^Z6)WO?%64|}W}R#UM7d2Aq7^TKKR=(je|F|lL*
zL$-94TNS)w^H2Cx?~%!m7?qGU)0E)PrFzg#E9ZKMb#KP4UP{d8@k6)qVDAEvW#7Rl
zf7d6&RGVC-<QB}scG2Tc(sBW!FL=nv^d8cl3D_Q|6XiuEy=a5Qn3O%4&+NlGIa&%c
z5`V{El}D9O%9Pq&XftpS6K9MFM-}hz)7Za3#|*R<-<W>wFU2o?Wn5u5i&!5?)FDJ3
zRXL$;jrD2w)j;ltr<(&vLcHo#(^KO5ii)2jD2h6W;ds9AJtUlm-B7{q+ky6%^g+3a
zJ&l(>P%0IaTqlKgx(zT_EV6*3GjBW}nRyhnSC^2MMl7rk+}4)MDMiKo#CoQ<VNH2r
zA8JE0!JfmpD?P0sB^x&nN=VZEw7f#0pv(QT)fM)>M@rf)8G;ZGKtSc-CSC7OX;9pu
zxgbX|ehGn`^Nqx*0m&%+SGD%_WV*3wwUq(uJb+axQ8$}3;HgE^V0FI}-fi*S+sa2F
z_HM-U+bRqe>f9Y>y&Byr!pS23^84le&GDh*8lTg{oBRr%Y9d{5J?B+7o?6LwAIVJI
z-)mx~;3fZyje~q4I}-G7SW&GQ2kFTScpugzDCC><NqkF!fp`vN4zPntN2qHwRFpS8
zQrP)bKa`}b%eE#wc^mXD!tXkwl@is*Ux(z0n=uhl3tSf8Eb||vt~Z}=3B(9>zP=t}
zdmORpvr$}a539stv?6`;$-zcSh3{;G&wtqhNwap1%C8HHBq+>HvpS~~GX1lSNYIdN
z9BpxMt2MFI%xooTmVo{(8Ch+YdyuJ3Er=uB9~6k?p=|b7r5w{h<R?eMI^>c{c$9Kw
zPGwl$z58iWAi2!0lGai5K)VHS^XS0Oee7D4^L&CUUq4PB#uis#JxQVcY)X}ld3Z^~
z{jXuSFGk@qIgVA%Ltmg*yQNx;P3N`w)jnC&X=M(!8xZ@9xAJ(&5iI!UE^o**kQjyv
zf{OH6g&+6AWwZjxgphU5Q)WZcEnHI?6FCbP=?R+fy<|0Pn}P>PYo-9o{%0M4Ve-AC
z5S%}u1vm0X!H4IJHUxL7UJcxLp-NX*SATIfl?ABpb0F+$KuI0GMhixa(QJ}-9DI_1
zlw#d-JP`7iRoYf=E7mrmg!+@wW%(c|bFpI6+qYb&7o2h72T5sqLVS{4Nz*58vS=^w
zOLGG7%<>WTxrz;L`P9YR?fpM1;fLnPu|_>{Y5P#vFIHktY+ImQf?VeC{>@Ylt#mB=
z$aN+IOJ2W}IpWRRDN~Hd$)j)R7rAQjye7R4O<4G!%KB$2-I?}sfm~dK^BnIq0)%p6
zHV)JTAewuaHA*r~w47M|<+1Z>pxC~IB+JfAa}$52yo1O~HZ5XM?!}9TEuetid4RkE
z|MLa=)QnuDr&Zgw;#0#vw<9SsKq5_oQMY4qzUn~@auaH1N#lhlt^3^DkC0`zzSoa(
zv7+_9XYu<{_M~kcF=)V;2Ll6k`rzs<GcB#vZrz)I7hqubOA*IQjE^7Js|FI_v~oMb
zCVrCqny4TSGK4s9f)s73f{m2ro9{x8^6_869s&k=$vqT7d#48^H487Duq3dLCRyeC
zG}Z^Ben!CZ=*U=&e|h=ajtJFk3F<1T7j{#jeRTukPz&Y|!bdpvA?r*MyUCnCZrCOk
zEPimVC|}1$>t9@)+IhpPZyy6z-+Z38!L%!28hNKdL?<RtwBZ6Fjp0dQX)8cS%DT8z
z!jMP5{fCUz4~9H%gojH<!U-PIyuO0Rl-fP?r@YYSZ$I`A6o#O$9Fe5fkP&+RqR_8s
zxN5T?{31ZnUr0ymlbgE!=B%f<{^WFxLl%49J2w=jGdANh2($#imn8-pi_{O?1jVqM
zFE$^kzr?5;j;3OLi3vNqy24?JX8jA5_D7Y?z=Kcl`z1{lw8E>7q4UgmY<^T>QL1g|
zm%NI9Y0r1?)sOLPTvNen=;Sj2g2xr!rpKml{K4q@`DW8RVzbgJPo^6C7=L?&R;HRl
z;<4zIqC)aFY(^d65J{MZ7HP5E>Mk`^7lbO|P~tqh;wG-;f+a^%-A(J>#P$1k^v<lA
zy)?<ahv`_DuHXUXV@^PAt6NEK-8|L@*7A7@NP@c}5c*mnQgr2}hy&gVFm|%An)H#t
zzp#g`d|ml{wf7o_{-nH+2&2`sY}Hrjn?-_pcBAODSH)En#i2zY2%);OalcyEP$Oi!
zwkP_9lkXO2!sx(?uc~D8(;2+oYEJOQv|m4d$g<Y2tD#}yox$>obN*p<AUlj0M%i8*
zXBh5~tbtzI*ZceOhYueNByGZ-TMaS&;6PX>;YU2TAgtQLI7pYt%7a&152&zYbzS4Y
zQ1ITVI|513NRIdyc$%}olB$o>R+s!j_eiZzChpk}Y;=%UM+G-&0S{C740B)HkduAs
zBF1+14&6N88@pVQvVLD;;o<94JU5e%{2n)xKLzN*kZD^h-ySZih$*d*{xfWLHQK*B
ziJ!e>qa{QL=FP5Cvjh`%N4-(|4^&5ggUJ~+Mi0DK6%=*5c*@>p1mfxYClW+f;NhF%
zY};k~7afeeql~DrVTm+9@Ni)TPP+VC4@|T|6Km0fKQKlz^sV<C#V-aj<^RJNW_vLD
z-BtIgOnJfz%i+%BYUtyuvyF}F8?KY%8`-J1F}j3?v!}hc*tE^fb9sH<O)Z%aEZv}_
zTm)583;3+)L=Bs#YMmTlR9P%6ZD_f!WAGMy^LUc!85xlnD_Gly>-Jf|WKhkX0_Qb7
zaw1BV<Y&(ODIMH}D+(G_+7gpt^Hf%|ar^?RT|?dt5>D481C6oC$i`Df{@7r810ka1
zQa|Cg6oD{jN8fxe9h^Rl)C;FNj^UgP8B7YuoKjKiPkV78sbPs$PI-?oz;UxAn0rDR
zFd7^)FO4uS<Y_T6MWm<vqTeC;)KAMH8&dfVe&sZ`xwsraQ6XOPx;I|O=(Ry8GgHAI
zK=7J5@kr^zz}KWJ#a8<vD?;oF&|O%mOE4xk5erW}U&29aofN=8M93E&Rm3)gi|6)M
zp83{z|0l}|zv4+8B+@2B-ZyVs)QxBTxmDd@jz)`%8qu)&x-zBi+8JjWuT~UY1^8ZY
zpay0QI7(9!bRVt_;p5XeU&Uyb6eE@l8hs_V8~<Gd*gC-xp=K}vM}9k;=vr;MEjtfb
z$l_OwI_2i;I{W$i7>gSeSTXZH*OA(}!NI2z7sre4U|)8CzZOCENPLpyy83*lvYTM(
zyQ&0TrrVu$zp!pv%?)2uA@YyIpPEVO8Uoy+R1cC}tH+lt_M7vTTYhy3o^B(vpc^@z
zW6y!e);)@L_ng7cV$LXkY~C2hO}wN2`0hU*9YdmMCL08izt?nc#Cm6oGoN5xxuS_8
zTP5&`6q_`NqkNtge_!l-WKA#ZOD_-&8ioP*wFi=Mv~ne-Q{ZuKKVZ%i1=qM%q6}1+
zzsXSSsMQ?6a3&MVx3vmDF)d|Wyu$0jds}c9K-2JwATj1Q(1gEQD;C9^ACJ-p)qn_W
zc#mv~1n)hkp*0PcWZ}|xe5oY#WOY(MewOr(IKz2Lhl^h>7bhn#ijh17&}b6&RQvm#
z8ArfH3+CcP;q-utN|-w4I_UR>P`xu?vczE3z0&`@vr_V6y}=}*pfe>GjvE_`GnjUW
z?Bd}^CD>wgyti}*8hpGKeaER1c0zH@%<^<vzN?jc9$2PAR-DM$ZYwWHO;1j9tjbc_
z_*WT)nXe_U@X!3A_F{-Kp<@p)4c8;rRmhsGI1iJ0r5CX><j5_g1g^MS%xb;ZPd=lr
zMtV{;>DB1mR%;QpXb?CJ!v1J+acBHvi68nvq4wDA&DZhqxS!5@-)OflzUlC<r=gRU
zaHku)7YS7)sQiAJgZFh#cF?>WBcY$pChHI!HSh=bD2|#F+$XF-br6!V&Oltk{FmNQ
zd|G>%pFiFcJFU1c08(u1Lh#cW+4tkdhMdFRnu757#3$<^P{DU&jiV<+ZZUFg;2#u+
z??37=6TEDi(nx&=`KO=R*nHeljG&`LZ#mHFEk$Wd%l)ixJ_U;JRaXC$wUV^L{=BgM
z^aPY+FMobfj~M?9R*n`>l~EnJX4jR{26T}eK`zr|whJ6VZ~$p9HP$4g-b%k3DWXa>
zhY`Qsne?&bK^z(1H1IcJz3sNwA3yU_?@-Xx(P88dDwTFvQ}b#bG6e^7mDSOJ!bmrS
zXuXmm==|SntSl0o<`a`g?&qXOC-UdrVFrEDcKszkN-Xb<z20Ow*RUWjL(2y3SZW6B
zk3?=U*II!!d54zW7Pct(wDfUpZ|AN*GUP`SZrd_kcH*aAlYnp6KKb1{Y)pgzMq~f}
zzkWvn&zBjG=;l?IB#20&8eL3NK<%km=sI2>|Fw||+d%c7wf?>AEbHg{0**(cyUz-z
z0S_2SrH`_?%o_MUptGRr#V_-%eyO~s5!W29yho#I`(q}NFMw}Q9{*EpST}NWdLQIU
z?A9sL=lj<pppVucXBu47wgM6MM?ouXi1!-%%8U^;Puexof$+SlT0>Ja05fqh`0PhB
zw<5`%6Rr^tYCS{ks)&pgXKR$$D0pjBdl!?MPWw!sV~w*oM-v-p40-GEc(*sDKGSAq
zViHW(G?bxs?`=z~MCHyc6jr=@c6n!vxXBq|jCHf)oEEYU8+S%TPje9;vr(5@%y{VU
z-GZh3%ZrS}m<_5LO3b6%;T+#P-DzBe3(>;VUmm4zcK=p^o2Q7n4s!J<P+rY}-PvCw
z6_$Tq!OztE=0vHvxw&}!7M(Y_9g{S>&lHwkJ7T!X+5sG7WyU4fsPGEYeVYJV&INO9
zY~(<7{OyE0y9KxH?dy!Hs%sq<6^fpHWttg8FRO_zQ>3~ibay9FllLa1@vg-rzZwiV
z0V7W{V8!;ZCqB5e7RU8jlfZ9cxXMPKnT3n}GPxN9Sp0yyouLm%M8AI-c#RR}@(--e
zY#jX#@IJDdrlY%}mzj+s4@?>l=V8!!GaOfZy-(=c$#qM=1VYt9!vDQ?{J#%G|1Qnh
zQTzGNNl}?UbJdN>u)y-W*iHCm#)pq+#Ey#oU}9FEyq*~x^PB;3sF+mXf@w{R{;W$M
zeo@(H_-j$hEHO%yC^a~&gfZG~GIbK3hWz-%-R8WdW-bcxC{(^@zZDnt;tg<|_>zb<
zN`!OyUtjqNRjFn;>ed4TLut|2A)`7HiF;MQTY&v^vUW3CyRBBza{f9~#0=ZBq-4}3
zE}TB9(zED-Idq6Yb)AuBgkR{SD%JkU+c2X2ZoBI`N3Ad;1&qjVI|4onsb5dU(HBqz
zAF3z`^=s)E20}WM)Jj^boL<E%GI0C6@H&T|J`hYopKpze?VZ=m2hrHWdEe&zuu-AJ
z&hCmgwl%7*QhMgg3#A2j!rhdIi)Q~!kw=9$OdqH7E=(vWja=^3ix9sB&@FRmtpEp?
zSS}-n_~T^mrj#YbA}1!g5Wg?E`8iXm2*vjP+ea~3sLKUz?0y<OSk$&mPByN1Wlv#X
zjk_^|6?GW-zSYNcmyT6i9Bjo42KLRD+64>bz;Y4FTenv4@du($FI97JaNINh-rd=G
zUrZdbg67T3E&9x#B#^6>%!|vzVD@!LO}wqb8st-9Scq%vozFaf-4wQ^1;9@>OUjYB
zp^UtZQ)4tde}Hcj&(vg%iH4Vv=7Ey7OR_?`^n=2*F!8bk%hi59<bYJQOpn0)k|!WN
z{6@Zi&6bvPO6EHQ2(I1pJv+{pehkKx0d-e^6JHH?`hJ{>z-(1-&VIk&I<P~nv1M&o
zA*bNfdV8M?_P4lan3OEkK|5cY{1-dn;_nQsH2pm6dsZl@fbTwec1J(maf`BXw(;nH
zPh0#MdO4Z8*lE8=s<XurJ3%Mg<k4=!4FnIb-7L1dGPMX|EE{z%QzXYX+xq2HE(&o4
zIp%FBLAlM0^TG(E$7DI-M~2jKHiZE5T4tdTDybc@tT>Nhpr-dc1i>H?ljFy>P)eG#
zgtP?^bl44A=kYC;4!uLZoTzv$vQ;qzK9H}A4Gik3EUIzW=rzMBVut;thrn6Z9NFG4
zuT5&eZaPVFxnsnQv?3Vs*M&~X81Dm^1bz#3z%sK?6fso*AGiG{TRv?X^Q{(4l{M?D
zhVe(@AWq<sN^yoSa`tjAY>>(fpOaEjtO9~LMH?4g<3}}HjWukD!pexycvE!x(8%0K
zOW&KC{JUMreH=B@+6XQVOB=lBZ%%M_+-{#-kr7j^L%tt3(^eyaX#a}Pypp_qSca@^
zG*g%PU1MEXlGLWcVRuB3<-|wQIui8s2KrRH5WAs9V1rR{u2xD4j}1nz`4f;x9nZ1d
zk%j?YWB?9YWukB`lYhGAGw(5G@EDpscv<|p8ucT2LVkpgg{DmXzkN_Mf9vKKs@Dts
z$t;hw7E5IvD$lLlyAoKvMpPU#W&QD9fVYN|ff%kHF}!Sk_p^5slct54t~tG=*4U`~
zM%@AJ{{*+um8wi4C2vflTpphM$)>5BHf@7gi|0RX(7~nQ`uW}{KkM`9aJC0q_@>wB
zV!|7Gj$7(^OLb523S@_t-d~&elm<Koo$5RBoR@lee27`bb5k4<yr#MpP1NF2+1t;$
zj=D1Us7x+mws?Vj&6!1r<f%T{Jk<H`p}awK`=6pOu<JV6ZuldYsQ4}&Ljr68<^@?5
zoWs~x)WCK_x(TO~RrM5+SnugQIa0^e6Zq#*SB}nu7g#>WhxI?Ra<|qZg;E44^LKEx
z2GMwqD0s5{3Z@PobXomcd4ov9z##0+MYI!DoLYqVwunhZ?50G{g!qn<f96jJ%C8&{
z71m0)Yi#?d?)eYM!L*0tij3D$&T9)}>L#|@Po6yYnqJYZ9cw5f?kwn{L7{~_xDMat
zLHyl_&FFSkh(UP?bLmPiXD+-RF)yg6IIqj<3cq<nF7Il&+_Yp9WZofv7<7v6!1>kV
zq2xjzc)G7L@5M?MopO{y6{pJWE8THmy1obJ4n@pHSA)?$z`9#Zr7lzqsXZ_^AG+(6
z?!FLW&y?q~k@KZyN0EX(gG7?+9wW_^tajuC=)*#bSe@qm&^B3<j;Mu#EBhSQ;mvo%
zCiUKctcc?->8gQJdO~chKj1B-fO+1SsE8g9eEJYY`3A(ePhLRY3#V5J`22^CCy|Q~
zViUskYFrbNRzkP7^Gu7OtLEh79F7dTaC1%uq;_w%i?^@$0<faapltX<zu1?R`A!!A
zQDo}He-T2>KFK+&9#YB!e`<IIxbSuIk~BuhMLX#2GxBmc!@&z;d2}E>FJ*hC&&YI?
zOyFku8HI?P->+M;#MHQz-n7vF=toYurRU_iZsc`oFVZHIHh=;bMjgg*m1S{5MaYS_
znw}g!`w~rpREH2NfXU?@rgM;%GywGmm?!Z)csT?LxJy9tK6E!+l7Mu0VX1#-PG^uW
zuYDuTpvL}|5L`4_s!in9ed+9}8i-M4zfcwrBBA@B8MAiNG>)4!;w(Z(Q*%tuP4+j0
zVj4!?(5(%>OB;8il0i(hVpVqJSPVU}#bi~KKU)2(C&EF)HWe9<amXwcEmqrSv`|R)
z86e(duhvf8NR}jNDqPX5GsdYQYf>v9%&K5N-SMAq7@ZdeM?v^S*o@!lW(f+xq{3w@
zk*l<fsCOo7C9q+#K<y7c4ByBDSP=FYGSQC)r#>gIsW#S$(!^X^Z0oknt)^|KQR5`z
zY_ZN8Y0}1?;!%W4Utr7J^7gc`Z=<;S8p2ORhIahT4GSBor~f2R&Qv)|K1z&04!SmY
z9c{0UmNNYREX#j~>%)YSp8GN{UbG%)6OgPdFUwk0<YfJE*`A_;`Cc0OboAa;^Dl6=
z`nH$t+O8O6EyPKTg5$vEl37mdCj4fiRCIN9!EK<j=H!`HU(&e|N>@vzi~i~TCFJjP
zhbSQAr090am`chhV#7(e4`0u|o3eR9>BYp9xnKJwt+V%WY2I$_l#;^pPT3VhM;@9c
zV?;Wx{Y^iK;WqV&ae5iDsbwcbA@DNcp2XX&@W$oD5T&cAzW?z3<rAT=Q&kKK0d#It
zWZ)(J0rf0MyLgNuDL6AWwYfm6S2RExJpR#XHl<1FH!;KFYV2c?j>Ly!r1@mg&8c{z
z*%{VKYUtaHG|_*b*mDc*Ri&z|clQf0H(%2$cPK5%axzjkc*~<jC#A6kM<7}-Db;6Z
z-pm4AFBED+rlt&-DY)odA;&RbVc{jS1h7wQHFX|;%(EVT!Qda1U5l)&tpCO7(#gfj
z-*5gw>lDUI{`6gw+HRk;WA7v4za9Ynva$BDHZr)1edz2f0xzf%EBjWzS73d*5@I~U
zH_*oZ$)u6H*D|(qj~$hnvd6jINu4fI49&Jdl|l<n@ea?elLm}mQ*#vJ5TAO;p0XmU
zlixnEG!RBmCJQH_%y0YkAs!_VC}>hDrdKmhy?^2Dc1+y#n2(cm%HOadgkyLm>GYj#
z%O~rXzf-wtgA79JAeAZH>WgE~M=(`jns+M8Ohg%s<<hB2s6!(-J?aqX7ZN<h$j>o2
z0cp*LQqx0N<IjK{`K;?pNIO=Qw`HSrPqQU`?KzYhVFmp5LP)s0{6oc0yCsqQL)>nW
z^Fz!~3Z*v!^e#8COm)>-YLF2DF|l{p+3_C)5j6vd5)3TYO-Zkbf742MbwdN}ALvgO
zVpx0|ySpuR<|N9>9d4sv>8ccEW->wW#l*~8yZCtgV=Hpc23}nY&e-cQmF2lE_~Uan
zY+129AT{bb$39YC)+)O0NKf1JmD5|TG%5m_)lJjQW1`WWg8NFCvm~<qhs~et&Jh1d
zc-?rtX3b>B)exrpcZ{t<IJA6;$99}Tf2Ic?N#R9y{5Bm?ea+(xOaEm0^{C*%SOMKw
zM4m>4?eqk?;rjze76_4O`jydTD%^WZlxA%Pc7#(%p2;*()Xl_ojvFrGv-V5|_s0Xt
zk+11Cf?BJovn{JKuM%-3hT1SeZeb+*e;2IqSh5IdAdeCvC!8P0!(-z$Z1nv$knei1
zKCBRR2nn-7^s#+r<L>)wLqBS^VB8kjLo;IlLJA6srVufgm_PgL*|#Zdg|AM*&a!wU
zD@M=$&RH~}(c!O8{Fcgd{g#$Ji9y-F08ux0-EB;b`jz>dJSx#YF86kR$@fv3L}uy9
zbvu!zZ%qZm?u%AAhB89?WQHb+BJlG?gqrPLLL5%a7ZN_t9Xu!MWD9V<DLz`SEVC=l
zFTh1-VbSpd#8({$QzPnNBwpifixqx^Yfm$EB>X!N_*9N!l%0m>qNtAJrb`Ee*^OTQ
zK|(JfvTgNUE-oDtKbz%|tdZ)mlGIL=_ZxE=k&ZI%IzYj<{E>1?9!V(rRQuPKGF9`f
zyUSD5G4e7;;)SK3Ik@n<9*(@&n`jfM_c7L=G_JO#8w{-PczCZ8&S)#Y;UN&~{1YIR
z7(l@)$N5&CpjYLiHmaY;5cTtNe`ntVu|5@H6#~)DAA<)vBUZBwk}#K=Sl2;HWIM;_
z2s-hXbz4=f5OatS1_|=}!C!r7w7v5aZH26kzDczvzJ3TR^Ls4y7dMphZaav4&7Gxf
zL3v8{F@BbTprdnVwn*1eGMQY}H^}wxXPW7jWs9ZnJjjkK=n37)XHt-;Fn{zqiA?d+
zRWh)aQI`){+y$a~j201#MVkq%L^M~dS8}!f(v!<gEC>d~aU9*%j@+$zp1nG??cZ6c
z3h}(X_M@}7@O<KKcg;!V)B7pPaWcw|VS-2v$sq&cF=n7jZAb}TI0PY6vofTL1P54K
z9PiHCXToW`EC0@S-e^c&yGy2i+e}E5abq5AFb5A4*q=TTnoN9~{Nbs0<!`%5*>MOx
zzGbL%Sv@yN`rxNV{4nFZFx513HKf~q)+se=*JOArWIOfBSt{O)b&LbwWzPddRb=%&
z9UX7=1?^_0#y%@+S&Yh%Uscso)c;Gp{A=atVOI7S*d91a28zgozriku$rs7is(zIh
z{mDyRJ-XN>l+Vdt^S!UT=|JME_+fPI#AQ3j4i&%lky63vNmXk&N8wxiBBOPyjsR+5
zNd1?dPu9-lE2d9`JWPLign7jFB8`(jkCPfwqdxr{GB_4vo40o)Z4`ZVe_FXMccb`!
zt$519X(#Xo(-q{ui_=Azn*~a$J2e-Gw8_+&IyRfZ6hNJN-~7f5G#YZ`9gm*4x0lp5
zZgcf4HJt4|45Jl&>cS}}r@T0nG9j;B`Y&%jV(z`<Ej_<43P(6eqGPpRzHM!lLUQ2R
z5{4!kQueC?K4NvEvT#7R0tWK-AIPkQm8hbdBKFT)T&I#%P`?52{V|Z{l(YC7({qdK
zKu~U?csQWQV5T&Jh0md@duVZYM%ifSd~mAxwcS-0I`p+WC$3x4S!tf(3Uzu*mCQE$
zyS`~deq>~1fqGbO>JDui;s}%L5_V!SvZ{l<3uFk;uj1RY`%TA+u!T*PhIX~^@EI?H
zfaDX~Dc*Jqrr&}^@zPcG6W~kPc<OiWFH$E7JVb%0<u`!f^B|sc`Kh@kx1<J?Kw#{D
z!6)UFa)4*V&dY)4Gh&?_hT~33>ZnS{dQ=B<Zbhkul?&^{^@<i*8?T9pD(8bo)i?q0
zYZ>_%k^VC(tqPD;oerm?tx3vByNu+dn<;?%EZHJmrUTd?C*ZNFlGb;<8XqsW5cV)s
zkszqvK%Ki~$>-2u87st&s9%d3eo7rBcps@g5-}n#RK*Ij`QV7O{aTm+=b;^fHafvm
z9J@xD<X6w!=C>6exwo5XH#L4KBD64{*lALkvijWcYII#k5n=;ls$Ii~!H*%kJZuex
zp`(ay-fQ}JzEf1*%ZHK${%@zMWSUIflSX_**W_0u=I3N<cWh5px-D@R9@T{4Idi4}
zQK!(2sN&Q9+vFRSe%9Qb5Ff+ecuCJ<l7^S**0{4c<1vuO;X+&4y-<c<JG-e0QYqiY
zYS)ay*@2QXp?3t~NOpT-l!ne^7<@^;PcbAhG10>7>bv?xRo!abahOth*mb8&ouggM
zWvXlmzuM>!*3F(h8v67#12ds>q&tN5P$P7Ja_X_cpA#$Z+AKxbV)Z|CQD3^8Hpf#A
z+O#U}sG+a+r2l>T)894kf@$->YjwBrj89E`OJILzcJ|}J-kX1?3^(rRQRuW@V^Sd<
z^s@mtv7qZqVs!LdARf5p`IjL~ltOJE2KPdsLpff1Ct&H~;+xXkM>dmsc?ehCVl-_w
z4kr-}bodsXtFNdFd6;s;tl5<iOHKlyzNbnCU@GbD0=W4|08^;&r(7Q;3lKn9fP8n?
zEV%kGaWTceGMA!8wd&ZPHB8o%|M~Dwwon022CeUZmmpj{%Wm-E8k8=Y{R<1aui+H+
zl?KRL>>FEYc#x9fB~-xEByWVbTLn-zBfdd(mmuMIIG(*A!Yk`~2O+B2ht8}cZJPX7
zi{u0r8jSvAIoEp|WaJ`_U^|%`e0W#T{4=efWt@X*2@?6#MOSX=iL3cL)m*q&Y0SZ0
zkwJ>ge4Whuu_<K6`;RIJHpp02-QNiFW6{t?4G*y~(!?fPf;6RkYuHE1Sg(pJ(Vu>e
z`{BC5oAZNLu!eW0KhD)!gPUWW>36xfx?D#M*7m&N+(%o*CV<JI7?isz{e4Ux3Pt_*
z@1dB_KfQXAVXbBKhrvSuIuv@Ogu-J>37l4^8-ri5B8Rs{SZ=+11iZJUKFFx~{$ZDa
zF`B~ko*zEqjChV2k+od*$VENwsV$-MKR|MjTckCilNUn4=xKKYS-BDj=+Ah*gpS5U
z%tFMt-KWgV`!33GWXG}R0wVqWW0jhusW4GTBGKL7>%9Lu4x2?)?kA&CIS8o9WdTPC
zJ!xA<_m#cbQgl5Y;^okP`t+rhr<Y19ps5Rh;6Lne>d?f(ywgwN%<&5cRc3JhIi?9C
zfrebD|I=H3P8%lyTC}RstyG<Wux@M3jf;IkbdXE5{7V!%=x#5a8_XcjSqlj5lT;BA
zcu$H@SFglk&0R+9U_Zl}GfI+GgHMO-O|nHkalwiclGUv~w}BtFjfNqDv(2ewNe_rK
zaSZB<iN;qrQX=?4Mh0|xUyW#>Ky&QW1HS$3Y4O*S5u=b^m<wvcc3DR`g3F!jE}#47
zo>}`>TtT<*30jz%?CG6`Tf=7Q*OM~jBt)yGRXPKYE{w*mo;Sbm79?GOFcCP@EBS2r
zKfzm8HQIPMVfDM4zfUfedIdzV-_QR3S^N?e(`GCD;_I7@?bX>LkkQ8r(~AwKLClSQ
zopP5|xLb7ijVPO%MjoodVO#z^F)bm8*JawfZPagz%IGP{aXG!d)NMcJVtzCmi50eV
z0c*EyXY#RfaGfT&;o-vzV?I3mu9?2^a?6&mkIs-t?mxj4>u2TW;p>jUN?#+T6NZ#~
z9e@<i1dEOGcvbFyh}@hU1WPhGOVY)@lv{F<THP%VwXO-;_PoKz_Z2mD@%0xR@Qx}0
z%*j7rBd>smc?(h@4KebdSmYKYXTE?6(cXA_ABs7+x9kljYv$e&+AoB#Q;RPU-r&AR
zGRSJcJ#G00JcR+nDj}qBnW{LKI{t$Cm}PT(EZ#dM@o!e?4P%)RH6@{miLaMTX|@<?
z1{;h@)fT6(V1kcQ`>;g*5b<4#nQm>xdW3s#_AF^UBl&%5r|OF)X@JEYbn)CAhGrqp
zhv66ZV>^jimGs}hg_`_Ib8|INiU@wHXm0L{Q5~=-GTgalz}km}Pm5qbfh!1My|YUq
zsW}lGYEpDx%w<OISJ*+KmFH4*#kV!kfHJ|Rvi1dFlK+|<AFTx_fiQ6IcEhQy$*;%Q
zZ5@Z0M48+aTb1L81&|&^05~Tb4@+Yh{T<cZg+DNX&1c7|wd~HfV$!fNdCnZ1^S<w6
zmzS5vjofSjg19{$oZ0=oQHZ@Tv<Mg=8zrH=30_DebGEJ!qoyB17%tW;!?Y8}VTV-%
z;!TS3w+JJ_y6$wsrV9$jC4u}5g?6c`T*vA+b3C>ulZ;>h;L<iks51dWbOU2JD6XFQ
z(AlXnynI#tgR;U)Faz0>C-n#=Vf;{^AG~h~fuy-?6z&xrz;c0;N}`(~B0>uVpM2ZZ
zc@deArlBA8z8;FhGl0_8`DkU|C1Nh4VDgJnz4tm4><{1pe3v@i^ap|3&E;j{p(gg$
zG`E=$@FNjHHn^O@BwS1&rT5ua$K(KERJ5An;QbkZ9uRQmjrmQjub(6;?|l)x&e~td
z75#bmP%2#I$W}(%N<~IH6yWn+m}|gODcM$5wBdYdm)7qR!ql1BjcocZ!*+DTkhn*y
zSYle0E9UzrtJQbB%S6WM2qks?>`$Mfu5J=!5pe_plWQo_fm!Xfi(|w|o(>75I(Bl}
zGritr3*SYni=^NJ_(Wz}kE~sLl%+E4D_v1!usNDrw*}&NTjzLH9SDfR96pKOUUHKW
zE0cl^Y^>g+{I{37;=3eR23P=ktjUi2h_N>RhW83n{rIs_P}_J*>)_Dg+~@c7EucdS
z>Efv)QX7E7BB>7TRB)_=qN}@N_TzeF%z@u_<d({}TxcP!8v4@u=PD3VP@HJ{ied7n
zUs0`un;<6jYT#yqOP;Ue7EaSI#7QAf&>$NTnrw1gdl6Yyo&tp(Dm(N)KMq|CrH~5m
zJW&^Uj(U*X9+_H9u&%YqE&eI70ZdLJm>vpv{NLF;0o&fB**YTczU@LV$@I>T_D-9s
z>RaObYf^^a;VTV($W#0MvOq2u0pa5mp*!n*8QyyYfDQ<zf*@77IH#=aYNC;6zQgHq
zqVYXQSnEy`r^Mun!rD&)A#^a?P<JWcbO6$_84_CF!_R~2NHp4cC0<=^(<oJH!9y|$
z){R$5x(=gPCDF+!Kkk=A=yumP5BP-i=(K%IaHJt)<AXX_@|jegK6B;w(?NW1e4v;w
zKm-{}M>sVP)e^t(1Sc0!O)>i@7I`TW__^A@XM$OlOnt(BNQOL<%?qkP*hUoGqNGJ4
z-O-Pi%=pc_Wwf-er^+RV<=k)dJ-_F^``jByKKo>VX~#a&CDEDy^A<d%HnH<n$2EB3
zd892BL1)}jbR%_X>BmUsiKEKObBL~HoIZyk=%%!9J+F;|Sj^noQb(+b9*AEu1Qi~&
z58Wx}1p4>+6h_Fg_y^<|)?;POmqXmPZu6U=9sq<HKDPsB)4rf6RHA_)9yhAy;Aj6*
zST2kWO6sdl__}{Gz@-!|Rbb+6$Hm-G3t4z%ZNF3zg>YaGGD=^5QC?EXqphH0=bykj
zR)ecIx<oJQ7+QOwSj>-Ug-Psyl0_CME01rH=tUOs1r|6QC|l>6PgTe%O4Q*m!pm_A
z+Yd%YEMsd6p{8E|r3a2bsrmYq&x&=%e?HfsS`Mp(?Gns|58={@>npJDPpK*V6SdKN
zQeZt%(Q%pZ2e1#9o&njX$;rVXQ*AqKa6P8j33${jNQS|@-RQo)$c^XAWrsd2%`>lB
zH9V$DHx85^-?3chKgk*x)78L6R^A#%RTgKJhQBxM$?4vb-;gMjKjSccoaHLM_V*p;
z!cjiowI(Azu8-^Y(Ko0<&WQc=5vFiF>LruciEYcDjO|~!qAF8|M{r+hyX_NqPGE{H
ziU1ZBEeLpH#c)QuKs2{1LqgS9oYXAj?_Oq)39>9V?}7A4!Lcuxsuz1Bo~T41z7Y?<
zV(o?30g~@jl{*<C9O0M!eU%SHDUmHbzQ^<Un9Err2b!;`V~v*Fp2IxHdsiGplQ20;
zueo@rs;iUH`yRgtqZU+g;fynd2BJ`IerR!Sot8#=?5&`qOI2whzkPsTA&22vOFOW}
zKgGY?VYKpr*Ihfb(BFDGw;NFm0Q$6&D12_Y-zUUP&k=B4LILh(ZO{&HIoiljyCIGh
z;mgj?U+%)RKBplHL)+6+sQX7Rb$CfdA!zN%vu7W7w>rC{9b?&~Fh9nrb!a?o@Nx&k
z_6`=%VDjz|*<<(tv*+}UmKsYF(Mya)8=8T5&WDIW3LHY3p{Krb1T~j$)iQig2kDHl
zEB&GEaL=I-AK#|1qA2W2UjZTcq)2)yYsg-^QBoIHbpQ@rCHCvrucN!VLOndzZa+N4
zgui2^gx_F{23#yg3Dse4)=^Y^$bEYp@l&&vl9h!EYn*yIqavy(O|Vi9&$Q=F=I}tj
zuux~!y72Y=U^z3Zo6(q1ObXQ=A6maItSLf9`0PJTPqB^%h{XHU4T&(lA=<wfXgJnh
z5b@Y_-%frsJUsjiAT-{MWngtN*%$RdkJ$S+UYrm{O6*;>Z8U7xA&*YPq#fJGU^HC-
zy92bX!kh>Hh3RP7arSig7Wb|AT;FV38bXMh<<P&jqfn3jXhb8%_vsSIOXo`!LN~pg
z(Kfl<99(C<c1tO5{++u?KwqnYTMSWZV$Q36&;+T#mKOH<FBQM}%`VsNDbck`k7(!k
zG;y~Q-od!~e1HQSzE$~toxl5vD%!v5?JZqlT<`^u8-oo_@pB52Xi&Rlr_^y+E(zX+
zC3EEH@5_E2h<5BTI5hIp)KPHZA>tpOIBx8pn*6#vG(3W~;!9JlHg6c<PaC~~HZaA?
z&p%Y7z*&9kLC_quL*WwX27rAPF015@xgw!s;aH@5umNjJP>OL641Du)tI<*;`Ef_X
z8aBYCng}GQ-DJO0DGjNssuG*`riMWB8{i+`eWWvM{<55Qlv!`yJ{NoyA0IlR>v34E
zuKU_ba-0ftv^PBRp^C~}&eg0iqetv~E-IrXh9*Y7K(=F?$gz5{iSVY(LtwT%93Rn2
zystc1q|3jwGEuj!YoPVEw#}-aTYPT=6<}H{zjJr;V{S=)+nBr%;3l>96}{zp-y`n}
zYn>X8p~y7`4z8@WUj%u#Gs9g-CYAVE$j`B7_Ok(Z3|nlp^<AtQd6!$J9X73VgI2eh
zpZ5IU<9P_boo#kiPdM$v$8At^wR&pdthF@JSS|8v93TH{U*85RyWaxUpZEDZkYP^j
zc`2}zu9N>)B4I&Dz~YE>d$7g?fLV{$^|5DW|ETqKlf1i9)D#X6z;u2*_ON*n0JM3i
zHZrK_2&5`69Nx$tjv;}L;0N*Tx!%dRD#S_a&D+kT54(h3sCp1w@g6jlJX7V{J19QJ
zMyir@jFg&wbFJ{Reio@S{_#ea7%!M!qO5<4oq38nr_EEqr+xaFx=9kT^!&tTxGm}q
zU+D@zpKK&l$>BXrW#D=WG%Q&(PVv1+kKl!o0xhnr!Xjr^ZVnDVjcNFbp8aP^3+9kb
zQvqe6F5Fcl!Zh|G-EsvNsx`I4zZbyvN9%}nc81?yy)pi!K8(9)Ry*uc+rP1yl0F@p
zZ<Ym6QnMP`yc`^O>z6RN0py!fX2(KlC{o5x#$@pD_9$ekW=PQif+R@-ZsIU{NopNK
zm+P2$43dOetXQls;DwkZ(Tpjt!<$UQHt9sMa+FR4aT_ay$!ed1A|KDRlH!u-3ob?)
z+L+2`nxl+#W-b^BE)!FH0*(lM(*?pDj#nmc8Y7I%o{cz6IXiw}$DS&^y8TriB*}7d
z+HVO6y2x3(VdnQMZ8+(y*<O&-($YExZ34u7$`RZHJwF^llX0P<$bux0e4bm=9a6^e
z0A5e^;u|K~A|Za!fv_>#!beJ=4=LPMjS1t59fdzi3`<_=N#7P;ORCEKB#irOQ2A(k
zI^4hKS$tw5Mzj5B?9V^=cnZ(l;Kol4&;wy)nV6Gtm@YvJ{)C%i&Kz)I8M`#PF$=21
zC&ftmf`^}REhnprw_F|;a&0IoJPW+P9F2}FAOywu;IUKE5oEO41h*^9NHIaN7=Gz}
z>a2NF7@h{xK?C9*MUxW3364FfI`Xw5BTVc%Sp+8vdHfcbG@8bwcOOe)Rz$MsQ->=?
zDm8G_xvj>0RWmg(7EOam(->xbat*!tEt>)(td!A9(M~p;;TV~V*;b67wezBZc2?yg
zg9QB;6~hK<dzJI<W(-b=^CXqnh5PGQ6m{?&(>bkte{;tT01kzlLKDGV7>7X^e{a+A
z@5On+y`z0thhBg&QOocO>pgedQHV$0c#u7^329s-w(EPuCbu9~Hw#z8f9Ey2LkL;?
z424#a6T+OJd+Mt?db={nhuAKBA6gw#buvq=8e$B;2{&mAQw_7rioWH0atu^8OHNTB
zS&sglP3>1>27+n;29{J?-;=so$7U(Vvc350m?RPuJf)HxAPKY~%qVP_0Hg$N-Aq)N
zbGwgJCT<iVA;b!N=4sQIlrNLbUGX<E40O}^XX!Qh_wQd-qhFu=N$$J4y4u)~YwkT$
z92W!mF7xf%gY4CkaJpuHeV7X<Rxc;;s@{CD=B{L+x5hCEa=ei_a7DL)Iy(9aAtldu
zRpIG?!V=UKcqe#(%${)x1VJ<%=6Ex(VjnAhefqik*W2Ne*_#QZ80YBtmTPpeA+UgB
z@A6gQ8iu}1RY|DzDr#Wko=K(es%Or&@S)&>*Kf>gpi*Co6^6yAp*BV0r!z4{mf-Wu
zh0JseakL6{nGX-eYeN6GD4c^S2H`OZbNlw#iYiD5V*$nGAQ_~!h7}R>Trm1f&BUnQ
z&G$cpzofJHvxFo_b$n-ws?EIfm#?hIx_oO(#bT_%=d``P9@i9JT{>rDWo3f;S1=RS
zz)TRugBhVhc~D=JH!${Jk$aVe6%`b)Kqz;2>1b)q!<oVd$Z6=DWAK*)u!xJGXAu&c
zQHgSgMxpH03*xc*zbh&2#hh<qm$*Sd%WLb}GmXScN~**mD_Y|uKwE}fL679>ZCr#r
zHGx@j05k)`WLJw0Hc&<+)5i2=?iF@Ctrb5%{<jfCv!4^sHJU#^WJK6Gu3IbC5j7ov
zEkq{}s~P~KwjB1UHq|#Y&^&#KRG$o@D-$<L-Bv!20^joq%taqsp=Ra*NHh=G)_&3v
z<JFuzQ@@y1o<B*Te3Q)7Yg<Am?)p95Yn$rN9(#s2-^u>kHRsuto<o0>nj35%bD+YM
zM~MKE&=33oTk#*zZZ5*`U3TD2UU3e5U~)w-CI`5)b4egQVXg$X8N<}j&{^ZCrU><1
z?kX|5Lnj@pi0br(RA3R%rok53l`TcdYJ^!O`fCkY1+Sxjn|(J_<O5qyBnuL5SqKR2
z9g>+~QsR-fx`to%S2yp>ng>1qHcp=QG(_VLL>2LMB+cZTj}9Ip)Z%sNurv#K!dj+q
z<|u~cvunfcbri=TI+<}Y*>&O3|F0cA@T@#0L3oSDp3G|!1TPIb7npGgAOk@BO4IP~
z!s`rXje;&E9n9l-(l4VIn}GD!-$WuEcQ-~3zU+^V&9q}w_KzrMZW4es{!dX!Nl=N`
z(GOPR*~W{5yK(V3ioSIaWm=_Rf|(`ScLt#3rqIUO6u{qU-j0}V;(~&8z#VorW}ZL5
zY5sTHD$1H%8-wzM>yd~lQ7c93J8w-4AJbKu{H?n4Ov~uqSy&OrL&5Y7YeyNOSKabU
ze%6c4nt#%RQmvUMEmbJT+c)|%p%PX7dOvF%ME&OVlJ=II>L567p@|4Bd3S6e6o6bR
z$MWvh^`Kq$NUBkEA$x6oXDU#mqh|Eb9!LdDZLKTDDZ39ITSHKH$Z2a=SaDc-1MoFg
zrF!CYHF{F1-_X30)h$`XA#<I4a6l#9ic!V-i8gr%QR%a&M~`3a=9y{pB-j{+_Ki+v
zko$sIQc$6z!fKRffe$0w;brxPE$iKO>6b6}TH=i`UbHEQvsR+SSc8H{8D^Fa=slLm
zFu75x)Pk?`^h=#(dwY81nUr;w;7k>kMxIJ}ZoB^RU0oz@ei;P9P`%Ji+Mb73G4e!M
zrMExlA*HE>#iX8F=LG9<p0i2Zr=o;@9^?VmIESIB2_fi(^h}<yCxB=Dr(mb!jUz<@
z^hn@at9|Nk-!FHLY+IJmaI|$zibLgHTH1!AbvIedA9sGIK+N#r)5nTAeiEaXiC(`y
zc+B!GWfQU}kj2tdX7R9V2Xkzl!6vrHQ+u)Svjn*@qO6u0JY|07E*k&9_5cbI-)gF-
zG)4l$#n%%m<}!3k12LWI#49imd+SIuad9KO%i&esOjY&Wq53Wk34&!f{$LCt-@K^&
zPKpIM#ww7TdY7A<NdNBLyM3-4-KrQs3M^a{uoJ%>RhHSp;4?7&1n%wau@#HM1w>k>
zZr=ZN`=0x+U_TP!Ffi2B4T*^%UpxfXat_W{l@EW_@ZBMoZ%`oUP)p)t&0Ws1+6HW8
z0GyoJ_hEt@SktzJ5aI8j3$XHrXT4P&Cm}6(e#D<ZGOg`9ejBIM5?>CtkpNFl`rud`
z&D2hU{5ownb1lao0_dC5UNGC;=6*+`L6?9vzH3&XzIb=2u)yp&$&nxj8EkQWz<C*k
zP{qXSllb{#(Gt&HMWks_CI)$8NBup4<yVQ?QU4cm8Y>L!ZGuUuxYJbIx7ev~W3o==
zUR#j(F~dL2u=0Nr8-Nw@+KBy!;8dHQWI^|Hm!zhUDND?#4qme`Zql4bLv{0H(K(M+
zJ*fo3>CBi1uq0UsOuK2fx&(4JBSzsTYBR-hnd60W;d5&af&f9mM{C;ufrNm<#a&|2
ziXmyKp7G=Y58w$~X&pF9>tS$ex53ctATH{MoU^Pyy9syncL9`IfbpnG63G`$7sh&(
z=k3h12S_G<{%eKW`LEl*0%1!d{P&_KU6KZq!wH6k=Ki7wXlH5AH<f+60O+cH!{d9D
z^>iNUJj&VxYy4H@t11kHa>eLgf?tsqU34`iD)o0y;lR1V6IXp&QIkv=I^f0R<S0I2
zHz}^DZsI;?@q4fwW$dbZeKe6rjn}`+MH*j)xcZXWk)|PH92qLg(R^p>&<&p9%sqV=
zEo_SQM9bZpCn{3|lhHOWlWY=FcYyc>*;3`rI98SCin?n7EuV@4C#>Ualh!&qdw*=I
zhNQeG$(NN&s*<8p64}F;NM>GbxrRB%LdL}n%()^^df3=4g565MHbhgK?Jn7Gt~+<y
zFlWwTf)RWzxn@2I2;!G*z1!9B$Yw6H^#SSzFIX|Jo+adCV9tk}`L4;T=4(EA;s;#J
zgkE(~?iFbp-SC#8`j2}lop2O#$&^{z7)?HaQyk81;m~L%Hv$<HjEzQ%v$;CjiSphW
ztd_~Hpk!rY`tY*BqwX=u0T%6ZJd2!3Yq#MxkEv^j+@xPM9D)fghOU>QxUNmZt>t9X
z%TsMOZz$o%pSCL(&dbTBP93c)7=ku5#WQehgr-PdJ>?k6t8tgrFkI9N0X9h%($LQv
zsU)J<e{+>e*ae1%_{13qg^Qa4v>LPPbHNr;ab-yrERL8~7)`=L#4Y}Qka$G7J(3|Z
zoJP34GJLr!aS;>zUNx)<cbtX92*m>fT;G#2M8E(-0(^7T0G-aoHy9!Si%`62=))<P
zP3ZQY)}V2^0-Lbv!aqT`maIIbohD0wtePyUn7G3-8csU(!D-mqTgAU!EEn^M$;Qge
zHiWOAy;^kn=^8{F+9(uv18Fe!(pPgdZhbkP)vxyxH!HTWX-F{oP8GI0L(|`tQypmp
zC6+sWlP5f_Cpw7HeBfn^9CcSU#Uyq<L#is$XO0Jw2;kj&tz7cw;tQN`D1m5yk!XR{
zTTEa~enlqU=_v=B6-IV2Y?d8bt56adOx}Xu*l-Eu@SDviCQ+5~bvpI8s!K+n9x>lq
zYGr<?!7!D$oel(-qF62L`M)nt9eoB)v#r+Pg7?@GPix0M%pmh1J>^H`PW;A@e=v+!
zep|#`>dx|V`z2%?^Bj}A2PCZ9^UblZ|0&zOHhEJ)qq;hj7rDMCp?ZHA!nBc<+s2b0
zmN!Yf^_Y2|-EqLu6<}qw$~9rO>k!+vydEVFhYqv#u%$z3Ylscnk)zgIPRA9N-!cno
za%-Fm$zsi(jD)0mB=GYk{pvEhpB@~nR4k2<xHKnH{@^46!E<Ss$4|8&^iOahR)vnC
z9CfthS(`oO+U}$6Wa6#oHC_LQ@18(!UfQCPN&AA~EeYg);_{XMKbpQWD$4I`dkE=}
zmTqaJkr3$)>F%MsJCqjbZfR*4q+6v+Nof!SX^@tB&iwxG`;E1frE|}@_u2c}S8)0b
z{@133dvZ&Go7)6)L{UCnVnH_CJMBVBuIi5X+WOEUu-Vyeq?~MWcvzwx?pX~V0ZYOF
zHoXDpXyPDGV7+9b8(bQW8fJcA<J0=5DE0$q)f8YG05v1ol4~qLN6bJ=_YF8?zSlJk
znnimBYTN$6@i7?o>e@volaB3fFE4Au9-F<4$3_L_+*i=T=Xw;RL1sObGXMXZZk~Y^
z=QlmeeQR>sJ2D;3)X?hFS4l)XmG?`m%{nzpMp!*P4Z!@~xvq9IeCems5CKSsj5@fw
z!&(o4fVhT%T;c~s2z9+usM8ZjEqF}XH<U;2t$T={6fRNLo1GDyZ*#!gdwgvKo{}+K
zB9tpH*a3Qwv)P%Mbx=$IKFatVETl-Ts=pDY?80@o_ytE-8RGLwAKPkk^c3|Et#@vC
z5|ff52)cs^3@bGSEjW02c{$3H;qcmzSx*pjB997Yt3kLUI2`X%&BoaPAxdB2vo^T(
z$N|#Mg;)>zQz<|W`r+mcOqP5B9PbK`k^A$#(07l;E)%L|ZcdvY{Lc>f$bF3i?pmNd
zeqBsodvHCO#XI<?G61AESpa}SLLHBsZlDSPyUp|W)|+cXBO^|M5?x8X_Ca4}gx`@F
za`CrnJ8bU5J<Rbb65|SE9c{IFzDbXsYu~h_W$17|mTT2<4{XiU4DjkrYzzd6@Ak@p
zQpFhTK~^INp>Xnklfm-K1;T~hD=_<{eaH`<xI-5wzrN&jn*>kW;TFL<+HCGY2Rm7v
z=y!wIsjfiI&sFOpr0(pDHSZfegjYhl`0fcN&^MrOUEOjA3G?PEJiqm(z>-N7&pNqf
zgP4+rx~p0`aA+i#%g+CxtrOC5){o#`iyPePQdQDdF>FyJSU;y1QQOJ&h{yYuI?-;z
zl&8=G-hU(S1uI|BMQ)`2Fab7+ZEh(v`JQKv04DgMppCc;|0`e0g#Cb2&J_rA`q;qW
z<sAOucQc@=j%D+YW?D#k)9g$ZfTO9qx7yWFb1|o#YN4?3oX|?<XksOC`*Sc!H0U}U
zuLK$16gL?-p?#FqZw}Tp$6L<at6u^%^zKF1nj{|1=~1@PDnX};x3zM&_zIMQ-l>$J
zivs{HLlweFo?nM<?IXf7#B~USe#;e4JaUhYQc;x{I2;B6pJD>%&JnL%Qbi`4>9P1F
z=f8rX9ALDdEBcJ1F^25=3BXo{B@*$PRyH<pT8YIuP`ZL2lFgP|-CkbmwNSQqjF3>E
z$)xIu(f1cQEPXs<$j|%i7<VMsQG}EE_eylR`;AF)avcoDQ{!{TsOLLCYC~hp=P-x$
zqU#)gHTc#Eo}%fxnED)e+i@d~;s3b@{{jJVau?_-;@*7LRjBs`R0!E;*PsS)1$UDe
zxxPJIw-HW`?*xrXPK_5ynls-QSX;O9=wLC5&iGMfQov_{!m#Qr>A23)@0>O2d%VpA
zi_`+?iPbVbKkIQ|UZeeQOnci0Uh(_Km;jCXqBHDA{nu*YoDH$dhigDtT2?|4oP@Z>
z3K98XLW&`xNM;dxzbO6$ZswAdz9SEVl>_9)f`?H2MQbeiBNf=~8Hxdr*KZQcia1Py
zpJ?fJ*j~IC<eeO1VgE}u{IXiFPcub<PaR(NGy{H!&oeGGD|m4`r&ZY5E>vR!vRi<!
z%Q!WQpX%<ftcjB10~r$Ia8EjVH@H8S5rFPwho(k>nAO|JW1L@}1T=4Xr)NRHKKesa
z+kcky2~=@l{;J{}KV549mb5<&u0Y$ZuHA&KZ*bCdCW}<X4T5-^0vyCuo0FXMg9u?P
z9546CYH&CV#05U`@xgMfv}yPEgjqk{#;R)@@77hP)3A$-2&a%Q&5UZzVD*5VUAs?s
zh7JV@<PU}!2}BSnlaaTef^!04OWZsJOr$X~oTBBeY&OiUPXtl2^MshV0f#bW7*sA7
zxD@6DEMC7<7&O_J!pHT#Tp<s5dN05=lLNt0u2Sg)#sV*LWYl-<IL0po)$DaEIEN~V
zn%Yua%uQs8VSjTr&yShREYku-z??lyOiWC-qVVYi4pX#0K`KSsg;R9kN(&&*eq&U)
z3wTf;0hik~8L%DzfJ~>7Je6{%T#|AGCejZ8<oyNTK>ug7{agNT2fSqizFjSTMC+r~
zAx<qA=_5NA3G*+^C?(=JGRc;mn@+tf(1++6z*gv^0|63@mRA4lOyl~#|6!z3u?x0X
z#N&=8h+>t)y$o+yp;VzAV>ln&6Pc&$l|v5_O6yXzND^hcqi%D70oQw0+}>xc3FME&
zxzFMEpfu-Ilu%*}Gc6?zz(DO^>Vy0XEg-b!_65{$@PooRaN#85#nb%4fpXFTu+Prt
znkEgN#xJHlgXsBo<%<@SDE;)D^b{Yaiz2O-eiNHp?V@1i-CzwIyU}^GDw@pCORKmX
zdxJi|>W9}^vJ$6+FnsU-0Q$B~Fun}{0!)KhA0jZ)pw?^mWC8KIaiCAJ1a17!b}u#_
zUf#G9E3lGVzq*BQA>a$Cm%bD{(CE{)aA(HQDu4e}xL^p_9(7I$=N5hXga)7KoF8u{
zLry>+^GuWEe?zuY&fZ^&LLmMc8mjlXhXPm~#?x5bEXeP@BlWIW$KnXYH+9+W%lOUx
zQ?YB_i>?WC{QM7*s_EMQ^GBDDAyp>DE|)G<)!*4(5)bwX^ycBL&y<1XOZwcXkp0<n
z^Tw}`p|^8*5R62zcm!A4P%>bXkFZ<h_jLI;1>{lXjH_+BW4PT-xYPyEu@J_M0zW89
zZxQmC9m#@t&JKljg>0hnTp<G0^dsJS{5A>D-NA#nzzM}4&*Vme<AVP+5QKH+UjWC7
zXX9%Bdyv&OI<t__*vD!W{&s3(jHUI2Um;w!|5AK?2fw=7O}<L`Ym}Cjrm#>;ZVU`Q
zhsMYI0d*_Z6FgaR;5qrFRRPT#EGof`{R$vUYGj{kYX!v8>mBP%x;}7KD*vEbOwwy_
zRctXpr&1acn1!>w;DY9#pa;}^Sat?z`y_~+gtH~CL7#mBR(Td9gx(8lpv!WqX({S_
z7ON;&Bx9|*74dGW00+PUF6F18j@S1M0G(zhspw98!ZDYE_uNojyK&gmz(ifsgzC^x
z+`i^*Kp4zk&JVDlg$h2v0tP#OBfTW+wS(XYSd;#3o`Y<aj0u8U1IhGj5JR}%^H3&`
z^=iKL^Hj05F{}Vf>*Mth1Q*OElfEccEL84Yryv~j8z6aMGJ#7u(=8}7I+wbry3&G|
zvMz_u4j;&lubKhQ(8;V>cY2#tm|^EmrzWQ`!+^_Qxo8?TF}`RP<cH3AYmV5x&`KeA
zzVc#Y#i}%xmA@54M#a7P^Mu4G)bsb*&4;8EZ&|gGc0lD2f=BXzQoQ;{BL_oBNT?MI
z8FA$NoFH5hm<k}mwP?y<^8xbj`di&Lr+8kzpgqm=UL2llFQ_qo$6reRqga1wv8)`g
zy=f#Wn0}v>0uj^!`_)Y#bCa?Id~q^NJMkuPAZM#81!o;Os!m+2n4qYJs6N8qvF>K#
zaeykeXkaK{n|!+lDYOw`D0-UegcBW^mpLcUG%GJ)MZoIcZ%955&i#;rETZe?Y!hiE
z=1DYjaW=^RkUbkipl}X+&Dak}+|lq^30~oIok3G}!ef`25>M$-+As~CX>~i6;z#gS
zD|w-(BL3~BZ4_V1!%qLg{^SH$fP%+KFv}kdg{Df(3yM!GYLRR~+ZUV}+4YDdVS_au
z7;o~@qTEJ9zTG-J^ipw%6o;6TV>!;@&@Wzo!x>*3&{m4VaaW1mlb62)=SrZxG(Y6w
zrgGG+CW&6o0<O9S(``I_^aGJLGE@2c(e&}X@G!fB?%N}U0GzsH$}uEd(eJIplfbr<
z3h+1Y5_R4#0O=HIp7)`i9<1+QgeVPz;b7eo?vSiu`9^~tB8>aLY4@Bz9^e6g1pVa<
zYEZr_osW!+^XIhaT(hwPqk_ooB$lS-MRlFUClp^Q<N)u+vFC9Od)?&h{+!k0>HB#s
zO=I!w#IqO=gaZ_xU8Hqnhzkg3e0PU>bY8lLC`xxfMI$Xee3~A+rm652xK#&&veycn
zkWRshqz@bJeC(^=ZUUhHqYl#4i$suO0eW0$r-0uI_!d3@W}f-?j}nzir|$(Vqs#D;
zu)DF#R}SU8R7mgIm9e$$ZRmH!1w}9iM(XxY?8t<Pn<boOpLgrEU`?|B!^n6z!ST*I
zB}7pU!4Gj-vgn}Ng(z-lzLl3-07j5nUM#i82CT(0OaNaqbZ+bSo>yh#`K#VuIMR0s
z{v3cOi12Ce8eFCIfUiZLcK{3<hDI9&$=CMY<13I0b{08cQyL;!=KG&~H3#eIwKw0q
z-p$A22V;M#*ri=quPb&&*hMB^CKVh=tp42|7Xgjnq%3WoOHFsFHfrjovuN8wcGxhn
z{BQK#a-UK-(}bWRB(kYf_ZnV$nrCRXWMlcXf3;;0S9L47@!6Lx=<{2C*`(wH=k)j?
zFI}k@dm0Y{zugnCz$5Q_-WI$AIxe%!<e?j2bx_%ktVlZ?`sNFlxoRrIK#W;ei6h%0
zpJ~+IUUI!qTE$zOKL*pdJnlhRe-d?}3}K$O%oP>i>Z*eK%OBalJXykS)>c%B+e#Q3
zfqU10jn^;mAaso1R~>?}X5cW}p|%esWpFty@)bMm0{s0$g|0<XN{VGT>D7|U>*Cv8
zU|;ybNy7pTQH9*m^0JiTc4+QtT|1miZ_nDMr`aq>2xy(t*bTiRrqr#OcVM7#P}e*_
z&<~)5a#X#q2&?WVjw7sbGmR@VSDMX^+aC>nY!sz2B^R!-Gk%^`{A(tuVA38!xgjQp
zWQ)`L@CP_D&{guT1-h(|^ln<a&K<gsv)F_H@Gl$L$|d~v9MM%E)CM?-4@jH%e8~%T
z{{#8fD^cbALj~RwP|@D@co>O77&$n=2E4R%c;8MNFnEpD4BOlp;j7$Y^LWv6@<Y%!
z|5?r3bcoJ90<gWhF=FX=VEY~h7QSmk<7v}wgs2Nqn{CQPP!ub{hyA`iZ_`%~2`0ZX
zDyD`cord_kvkDyu#D)I`Wib!4NlI9!A3A~MKP?*@CRoEZw`^EVp*OX*J_E7rkLB%z
z^C@dh+?E6Da8O4!eWdRU0D{iSX*6q_U=3mpZn6P{W_C8zG9V-o2jN}7dN`oSv?mx2
zPn_N=#<YMYgTnt$;ZGQ27mo$7&ueO0O~}#RY)QVbEzo?*FF>}cD{%f08bIb=8s}J*
zG*N#+2`zqAFnf>J{r;I0EL~gQGqK5l*ZsS0RGHmu`-@`VCF)VZ8N!T_9PNA==JXA{
zo$KjR34NhOG4=Q!U^M17Wt9k^*eGRc&gqclC+h~S1<1jE?m{=#!yoV=*?mC|4ytxc
z7b(K!qpeP@DJdeXqc;zdt%Lu8Ykw1WK<r8eaF&qvLa%k3QeYb9M+3%B>VyDe;30EU
z|EJevPt<*XX;>UX)*lhp7L)ub*uJ6iMrq#qcrl==r!}=94Q*@bS5!O}JtEY!=_o2a
zFLou{2i6s*ns-o_!wI~Df{V|NfPC^TP&Ll}ZZ+}t7Wj8{?6NHF<;4$tkCKv-%$I?A
zn7WQmBrq}AlpVeTMAlzmC7{-k=m`8*f7Il3q0lRh2z;z8qsZ!B<XkKK92(;Ak-SE2
zfAOOLKh4sK9y@@j<tz`as~EXymP$Bbg-bJmv)>&QM>1-o`H;m<YElEJCeTKXp7^wf
zODvQm=P?I|#6(XeV0r!%3^j#x+IstYRvC{yWng^tvBnSG+z@k5KGcEeR&g@KH6gEW
z|2U%VYkO}@oO&TC=s`zU?u*&`m6LKvns62@q5yN7&i)(NfMArL-UMdy+Qf6y*XDEq
zJ0##Zhq^8eS2u>iPLm{-Hv#6oWtEj(X4hW3ul<^Xb$G#MotQ9VU|t}NEUn_^qpr?+
zOS0F`LVZKSr9?p&?-ZwHDco1#fJROZE^--F)%ZPk(f?RAvn=fFe9oz4Q~g(|nyP)k
z;q<L1ukCjlz$jmuYwvM3#{CobL@w;O@k5;fvSLjO5hDnns6h5gYQkz0nBx}wJ5Qhq
zGVbc@*)7?1kLoHpt5`T|ntvLZ$t}oJYua;!1z;m{A@=T6VxE3lLO6c|gbH{p^d4Ko
z*A2QGCL*6WY2rF}rao4;QuHc8f_1<f-;NsEl~S@-gTO^T!B|`+(ssaasLJKOEkajx
zotB9-@K6e}x}}A;GcCWvy(Zy@Y-;QPFc_nz;Y`tIqP|?6C&GVbJ{|xXR2*3Ik7veY
zI{JkTew7}c{f<`xt0W5!iWM5kWtSh3o4;tC79r&!3O5P!(Ydz{AfFmf>6`qioXLd%
zx1EDqk<6ApW&8#PMsBbc2UZn=@^Zj{yZn&F%i6`@?E!mMMb}%eK|K&I?3AsQFKm#h
z9eb^Q^6tWI>e(Ke5vwnUhhe%gl;d_CMc1mqxdf+O!q8D~1tzd5X8hs@Ta}_$&udRe
zy!JDJ1546xmZygwxAc*HIuw9GZWC@O8(h~TJrc(+<>)*Mcp!Grq6PiAIG!=ie}7+;
z&Y5uHo&?(BFuPV-keJlDKBhU)(e3!Ef}$AX{X>QXZu3XR{Xbp^M3sJizRL{oA3YAR
zW_U19LF)%hE(ASHzT^%yN-SU%lxPS@4f>Z73FYR0a<Uamo&VY6{T;Av!1|FNVYVa+
z6fi140j|c&W^``^L(k@?;&`622zg!aj$uyrh0ucLYKfeKyH9;H!r0v7k<zx3h%ke9
zh%kiG!ld8q;8p-ucU*$L3yV&X4CYo^!`5A$?o}rGl+RMJv4{UjaWeBzRK+V-9avAP
zdw&8JhGl6&?rG+DbaeQS2wc=hTVrgfs|-jdOkca8A%6|P7@Z}_`{8(dVSJ$D<7Um$
z$xo2i<<{-vyW>&xoD>jJ;AZsj(?Kux?+^XUA7F2V`eJBeVgQU4nB^AF2b=Y`yUFoN
z>6oi-y?Sc*b09%^o>FStDeZYBA-}`|_UroAiGdn>wZ>2oYGu`LC}VH~v`^gDW6wkZ
zld@8Y7FG&q0C4|5kfs^$ISU+W;8D2EbH?>v90IzjG=mv<7yP1pAIVmQ_ycc#0)u|H
zDbTq!T}iK+tn=u+f=VIP%u(9b*f(Ph835bgg>?4fVnTj?^84RJj1Q3?9c_Vac`UmW
zY%dJv{7AZd{1@{;m2Pd21i@7c1QN_vptw>lSL7K<y#d&=ALWrZc__g5l4Pk+*0=pU
zbdg;DDUg9dTMjVjqKCqcTcq=8ru}TJtpQ;=nrxT(A80(j+A%OqfeFjEA1n?r%KhQK
z{oaQLvsSG?0M<pLSQiiD4}N_oE4L5=NoB4X>BCRoI<+!0-$hxf8kOh;I~ja+g~hgC
z@7Y{I%_l9CqH6*AdLHj}Ce_p|_UU=Bd5Z@n`^Q|nlA<7)lLh%PhT;5;i0B*pu0TCP
ze#gS^FDtyb_#=<%RRCFEv~KQOXf-h-O_j`{k>O~M%+;;n^(d<%Xlr}pVQ*bZb%4VN
z%f$#(8GTu$9fuNl(I&Ja^s2-0H@d+r<aL`?f{u~0_|I`}XXqJ@ya$FK+Y|Y?$U;2~
z*S}dC4Y9MRe#STwZBCKE0lV;LKzvgRBH4Y1#|r6XhrR*L2o$jS2eSDm$m@JSo>i}s
z_mY#K=ib)xv@sm5+{0hCZtzo$i%uRZ-)~A$%81XSlUPuJJiM*#xC!^3;#VGYDoD?Y
z^D}&C$U~{Dt_JCxqNcX<U~KxJF3@PLaOF<7lh8@dZ?Cl&1?>mcg^oVc`G7VKg@)HT
z5WT61ot8bmCX*srNPKBmGVxgjcW91dUN7ivlO6}bLuVmogj^sw_RZc&v0mNWysiu9
z!rqH2G^`PnNOLJ97{Fi-bKZ<_gbE5n34ur5?|4+#3^3b74>g#Ci8Q?dc2`Lr?_IvG
zctcL!f!9m5d$8`y{qhO_%VP{L$MIJK|E0%THcE+t9Zr$&b)eWXoU_m)z2*3ytlk#v
zSRyjORvkF-<%wLLH_b2mW2k3E8F4+$7$SIg^_-Vips%j6*paZwLrw{Wxw|dMtA|=9
zYoWOXSTkat|M@{5kMRfCDe?$=s^*L81NkW)5%63DIo{jHETO;y@PAhh<80b>4fIvE
z&2N+yYBEvmgHAGC@#gMW$C4e473<WadLYx=cdj+pmia?R{y@0Vx4do@b~|k_@3LC#
zpe_8ST35TS>dwZ}?IL}u{bRNEH9y}=CHo-Et9`6>P(xdPQt`0G|My&eh=qzQkUPnF
z`_V%E5ksqhHLp6pQ3$m-?eZZi#A72+X+w3Q>z5MHi%ufx)d8!s4mixQG#?es?A7eJ
z@DA{65mh~A)a8WQZ!3+bo=sH1<Am;R%9og6%A^Z&0s6oc*ba<Fb!(jlqr(P$K<3+I
zm;driG&?f;Wi##DhR)8;YkL~t>-FX5+t%w}v2mPT(yB0Ew+Bzrh^O?+uwUu$5iCop
z-#q?nQC-6<lfR%o$Fpv?Akp;jgPHotvO0BQ)3l60;M?^?hj5x2ZEZl2*)Y|0G=>(3
zs7Yd=RPs^gT>O2{Z}s}cYPGIzly@pN)?jjDAgW8N-URL^&My((sU9E?D@;LJH3NNJ
zA2O~@l3YgIex9rT7KD+EWk4Q6Z||wE*a(2zMGoX&FZo+~K|MsDMbh3kAYByCPoDI@
zY;EWuK*SP&MoZ3%RGKghP67mSBUN6&I`B<067+G0_`>y<O0Kb}5JzB`uWo>1HBT9M
z_`HVs@r7+R1(!M@4~wTLdcpRBUuK6D|F%;SO82F!f6TXd>D1nX&vw9WNfLSuZh1FA
z8?pqXq)H{X4}D>%#Q-F0p1b$jH*l<Emx+Ht?@qE1%2lyOfrF<BegfshQe43PdX{I4
z3@9zNv_F7!Sg=)@Yn!XYFczzv<M}+ERyAL@z;nA(W4^NQu*)5&T(J`o5A!lX{x0E#
z$B8){jPo%LgiJ(R9B}#M=LYOJbm8;#O}ua|pqPD9Qdqto?9P71847&EPtnE?1P&rj
z8YuR=aeVYRkqkCe!NL?&U(1sIg8SHHX$<%nD(f>crcBDKB&{k40##oA0HUf589+Oa
zr;>-`Qj!^FL^AZZ){pSVB#e$(cRlZ<U-cWBR>?9*DJiAp4RsJp)zLLGF1>g0+L_?#
z;9`7_j1L&$^7MTBSEEg@>AZ@o;3QRgb|q}Mz{$3WQYpvrU%S-+T2!Q9*HGV-aQ6Tr
zOx{(zIXJ+UkIq3qt$!Ymg{F!hKT!q%w%C(aj@9_2r2W4`YFCXKo2myaOp9x)9BqB{
zsaPsn7U=VOq~#zkS;nXpr-=s)k<geSMt6i|+u%;jP#4I#rjcg*5pOHDNLl%+66c|q
z<3-0AH&Cq#uzddtQH}5k?qOu#M{zgm#yavW`H%y*;!27ddNLdUnp(mtV37fDzCach
zU{~Vxztr3V>O_QcQjie=)O#p($;xjv$Fq1*d+vS>oL;hc6|^PDN{_sQ^_N-=jU^HU
zP3*@UaRsjgXS!=@3;pAFNZ4@l|G6p)U_!?_rMZ!j5x6}bxrp*w@P6rl{}%D2=oyx5
z99aZlHh$vlS2t*OjG>h}5Iog4YR6k3U1+*jMM^Hjvuu%_KaoX*S<4G-9dYR(Ob;z#
z8a)R`tv=Kd%NSZCJWnSn4m{?nEt({JG938%4ej5LvP$YSVSJ5TDfh4~wCB|(Nff_z
zBeX8QL_Hyb>`aVHNfRC*LX45bGLY<rT+g=e>?WF?9DUs#)u?;cJ9-j12N6U14t$C^
zpBLtiZeX`#Ch!vXcD+=2iKlMy3dyV#)G)G?q@=fiOF>F$6*EK$a03h`xcS7!g{)?c
z;iw&^P4c|^d-==ZeDILf+Z!@*^sB;a>v+{{HIcLZbhL_K-v7wA|M4``79q2Z9){$2
zxWDu;gN*@bw^%M=04VP4KLpb$Hdj`%8oBio!R3%|^z|#+DFzpSa59cENW|Wga-Vhd
zO=1#dVVM2b9j~ih?kCV<5Lpp|6MN$D0bwEXAMlSuI>NrC`bL&m(=q?+wAX3xpH7I?
zY+3gE_pJN(AL^Z9TkHj24JS}tyY}cP_I&LuILAo=N-9lQ`~^*uOOUr*wnwS9l;sl}
zW!ot<9P3xnOu&p3(y5g}dEL149ifPjav<xM{T9O@G&8(qE2?FOOrxd$ohnD(guG}9
z%H%u&Co;g*c>!8^b%Ou&5=HlyDfz6cX*oHuXOkI9<`Yi%I>yE|?O5gQIbpXk8ZSh9
zHH<ZJD61g%)K4y26`e)~O(x3qt#JfWoqKPqLR#+cA|fIN;*u=^%d{0p!d=1A(&F#d
zaOEg1?k&iLfd^}d{`+Jk#lRLt>7D83{w=;v;4Z>6cD~Th+xeTg!@7-E-CjEF!ZSpe
z&JtE^w?U*)G01C60ao~ipRv#drLE}asqe7xfM@g7&BeuoQyRHLbGmjUGy&<c=Ii(B
zGhYs7wj%`BLh#TSzTV}3KuxS^e@0jIycJnIiaBA;wmZQo<4g9Gwo)rDYFJgnv)-kX
zc!BNxF)F|Vk8y0Qs`o^?!WUH4KivLUok(DP2wp$wE!s(hYZY*7>6Ebj0w`bH{Z&Rx
zJ)naCrT`RqInNpZtAAEH1UrwuOW{Una+q{JgQrLB@#M;1_U8kJmbx$o<_Jj7OV%?*
z@A>REmt~S5IrTbt#aXM_yQZgkp`f82pd7aIwRwjZECPTh+J!eIa{7$`p5Qh6J8=*$
z(R>C+zSq^}0Ef*L48DsdA#-I!%yp}g?4~D%CcdqoFL{mUIoi8XqEw54$aWvckc8cd
zoz8(&1=~;%kRs0=&`bp8IAJ;si}lH5fBYMct{yjEaIEPymVRd4+UTuk|JpFY*obk5
zb?ldHzVOdAYRuy+XNUBb00+c~{HyyEorPE^uf`iq;r%C98=oonTDi4;{(P9hf&jGd
zmut=V0;XzI%g|)7!14}L>*Mi}RSrQ1kaoVb+xLc>#YkdPynG6z)I^uS#pJ&u2e%4$
zEdm)E3#|kDUOa47SbVOPK*yx#g|6yZ+NK67SQu=Hbp)Xr!86Mx*r5{Bom}DO8V|qn
zz1u4&bmQ15$LX~X$Y>Fg33z{fAN-$wYqul?eq9t*S1WrdY`Fsmv{rD&;ZDOJTQkSw
zir^&^F5C;aPb2CNeKC#JI~PNNGP{A>Cc^h^(HB@DQH_G@Q7TbDLtFs(xd3|J$M`zn
zh#%H7RS%xt{vR?vN^<o>Q%z%|_6mDVx2Aq;q^}J+D^|m#&%fj_um|u?&6`^@vE0(*
zMn}&i#kSU17&T{+mfb+a=NabH;uo_uzjwstZqo5@XNM%1B%`i#Km@s4Z7N=ex*+&A
zVE=S-E$>074((8gEqVN}`;>x;k1~O7zLI|4v|O#gnJ)&&fGa90`oQ3YTW(^hirUuJ
z2ImZKXsL;wBA@a1L8(7z8W0FEG8;BRy+D{=BN;ds!iqDgzp2~>jK1i4PrWjbB_>XD
z<ZGW=T-mjqWu0?~NJRrch{%M{4#fdjwE$oL>OJ?N(k(Mn?=foMb%fc{m-!&J3}m__
zWo6kp+>@;&Pz%(_s(U+!KYYSIRriys4&blFE*)5v5jf$ks<#J>ZEPZbelWJ64;96L
z44F^P;Z7*8I&h5Fw(2TPG#{?Av6s?ot1*5~q8-qX?^?R!)IA|ZTrpRD!8F}l7Atw?
zkdAb$S{X}A2V0OivM1lCv{N-_%(Fc_R8CGqS{LJ?2@C^CWx0l5BjVZ!dT&+G{+1){
z(%e#p=AYzeS<C6S{QaUO+oBjR3tnAmDt97?P?S=MsQ3julTd|R=#Gi5B}>qS3dGq%
z$`^o3!Zd4O$e%)<=I>bozz_@=rUR*rs=C#gksux;8?)ivx}pDg<waG~GR1ClpI!)7
z$?Zqy@+wEKC*BS}(8Y2NS4)czfY%1p!T9|o&;rEu9)LYt6wpeKp>YuK%L)$ySBdeL
zKwo!y*fM1lK{nmuVtN3WG|lsm=A-swfcsw4_?P|mOJt-NV}t%}G6hfqG@Llt4$Z=*
ztF`&~2maH~8oY7=$spTE2G}2+PIYx^S#EX{;`j|mw<5Tw$c=2p=dSf|9dDIbhqa0v
zHjH-=Gt-|J{@6sgvkxAZmUi~4A^q1<4C#H$(p2r>Uf-R$B5u(KHg-G@D5XeMB$Rf5
zgO9qa((c!Tvrcr@G#^_V)e-c>#kc>igC0Ol9LOipzzi^GxQQD5r!lZZZq6xvn&b}%
z+jzNufEXmynm1}w7JMcVgBEA(a{tF5QXA#%?s;l(eidybB=pAv%b6b!Zm|tUCDG+m
zS*ymZ92o=df<N<gP%I{;rCEhrDa3)B75GPp$o@HnFPDAR6ux2=4T*#R?`p_V`#z_Z
z?&ybe!q`_=%LMR1lf#RQROfHadl*prZuwV+VJNR=R$VBIY6w=-X6f$iQH-TBU>H@A
zuMG8C1wbAA?7;!FHN~jTUoD&7`p6xbk^g&AOhzU!>&U8a%HyJL7fu1wM%9nB%#(Yd
z$n~?PcVighqVB~zDWe?YI`LH%QI+PL$LqC1|2ZKJ(L_*Aj>Jd*Y0|!jVn(<;csLyw
z5TTrPG#j>2YhC^$z-mCMq(EvcXg_7cm^vE?V`UX@9uo%X9It1`Q<oHZd;u<p(hLC1
ziUM3MQ`n1@ToSL%$j?s*PMHF}97oxz?!YDcsBPjfi>`-hN?%}{<Km=RN&4uJnN2a@
zw+DcBNPIP!2<Za~81mWc{1@b<R=?MC;@CF*kG@ED#Cb6%K^33c*}R}nDO1TApkbIb
z1Pl5Jki`KPYr$`tJ$IuW;JN<6f#Fomutdl|)w#keA=_uJD}HcQLI65kuL?&YfOqR0
z$WJ7+ZK6)Z#u%0jJc=%U@jJrBW43+VuyW3+7X9ZS8rI-mT)6eH5QEvSSIxM1y(feF
z5!X`LF2cGM@25l)o43(JRnGFRe-55DHvX=H_%pe7?EZSr`*BO42h0O|BI!bBd-@ho
zNGne0n4l2WZXrkL{FnsVQeVf`_2ruD>zkDi$I_oYh5~+c<O+pFGt-uA^gR((Q>t*E
zRuMEGPG*GbFB0?eYT7C2-cN(*<Ob_;vX^cHQ*V*1mIM8hI^Sy!V~z9L_*>nTakRHh
z_FJb(t746i+28eluUAERSMr&Tq{SrVlcQ#KOUn>TZ*^RV`6}_P2rkbQXP&`;$O1@w
z0WUQmqd0$<y8*gm?F?y*wDj~}@ZcOkl#a_VO&`PO*N(!W{J`0HdTnv<&Nn%duuMCL
zR&LMjm4toAu7SD$hK#-?7(pU->pX1J!B<?M^WlV*3?%b`%$k-x0j+t*N={CeGu`<a
z4hr9Q{Cw<66VF%V_mj6xeA-EuTc3)T#UjDfa;BSd%~61DK<D=^Emf-L*$3sh*4@_e
zNL)1NlN&NpMilbNu6#map5s3eT~eMtc=C!bQm!(@L{^Q>|Il0y^159xp((VkOSH`2
zfr$S3wFwO2%CE;Ll;u-FVjeD#CglCl+Xg#gX%K<}KBqMLU+@Y0VkSh-l%dd)HLGMG
z4@8oYk#VgA3uH{Y18t9j*yvos)Y(!F`<21j+5Ybh814C&_jQ)PU9K<t874E(Lu6RW
zstqk}g#bUr-m{T$=hLT8y}&fr!t)CpJmfb2g6z23sqdwq`sK$PP=3qsPT*}RrO^8c
z4!Fr@Jf&5$aX>JqN$i@^T(P!}NvfH{A-$}F=BvQ^FfRZQS$%%??(>iKA=$w0^_?aI
zX1*h`u7_xCX>`EMOsX**JYM@#SH1eCtJIVH{L$uEbRAl;_xWDON3GSvV`7Uz|0(jY
z#Ji7Z(KPQfWoGz;91L`IeY|Fk3AXFot#xJacv;`owkK~!<ZLdtrJO03@BIsYB&bxK
z_f*T@&C0)OBh>bq03IqGRbM?{U+AoR6RD-fsZ}oDV@X;o+fQ9_`RJjnV?K|$h%{Gw
zo}-^@nVi~;;Cs)mazh`U=pJbH-gJJe>B0ZzC)zsBS+=5L<9CMi$@HtimxNX}I8Kn&
zgtoV6C2C05pK{p}_GrwV=dJ0PIAt!TO{@kT`D0+jNzKO<f3++n(Xwjir_&X2-)WHm
zx<oHf_eTQu^si#-duE(hS`!yc;aZq0D=XK)jwm53+Lz2DDKZiz?dL1an?Bs&(EHmf
zOB}j_hE?@liKmjXDa_=^aURSnj*`*Xgqcu%7j_FLx)`PahE|`zp)4*;k`WgjXjzhK
z`DtUpJr-OEYI0Gjy8)Oky7(LwCczg~rW?R_sULVDM1wHY<20&?>YckYChw+^wU#4s
zpJr;WQ(hD)iRQa1ug>NYc6;cOFPqEfcaG=^-hVhP?)^${DTG&(G@(itLVs&mrY)at
zh0)luLK~ghCh{G#wtq|OVZeAv6#B9qM3H*0RK6VK#69Di)<ar%f>5QcJ3@3PE!<+B
zN=7O{Q^OB-f3corv)0wrrJos;yRGq3CZ&a7t=o>Nx<m4-&uQfGF1JeB#9xK_ApCr?
zzBazox{ytCYky|+Z-RmVe~q3bNzy9@HtT!Q+BmlV-iR?ggd9(D==u*AlQ{v;#<^bG
z-u7|um5Ad?JCRH9Sxy5WD=6Z|KIM4=w}xRAw|}nqZ!&P3ms__D9Bb~Bo3GSc@6tWu
z4aZ_y!xKrx7hc4_J9=<tAP8t;CqGv^uD50qCUg1Fosc0e@az#a8gL*e0Q=-iYYUTR
zLO8>S6{^YIFOsV8G2D>rufBkW!1jurC+2QfEkt9tPZtcQM)$6QyEC)18{ps1dqW*^
zyd0$kd#-$|V^OKg9^$_p_r|0nj0e413i^XK*}<>W^Ygod=;Xb`Z}SZpFiV9b$~+qK
zi{d%qFv<GKV^Uj%V<DQ??P>co`3y?d0tpabh)2p-gk_ZV&^C$@Mj}-QV_b1nqz9tV
zAcz?H`ZtXDsHc);*|=FstZIUP7yt2}8Qc~?;=eA-Zne7PT=gxCJPt>gvc5vokf!P-
z=Wfx%T3>;9BVJ>KYI2D@E#ByoJx`k-u;glK#qA9#Z2yCFf~m!X85jM+rdI6$lM$Tr
zj%9PPs;4~9w|Lfo_5ky!^Gx+~Q`1J3>3yZ5M|EAs*#H{t><D-|Qe)B#z3gd<=C$r%
z=%||io%P-*%P*d`peEShzhh22Qg+<lBhpZBPJ(zK6T0etC?BpTK?jW+eqgzNX`3Ht
zDki2=YZSw%RQVgjc~+(HO)q?Pc-3_r<hy#7A0^m*8V8p}Y6)cp4Q*a7tvwLKOQ@n)
z+PN8zAKXsmC5UTbni*W}qheQ{6&|j0qx_ty0kakL<L7KM(D(A?_d0)f@Km9xWXwPq
z*9V5YutB&$yJUN-moA(*c|3mLB=-IBo40PNSO3pe1|@TEZ!X@t9Nr*4NGp1;%Xq1J
zH`7`pBy(>+oCEbFFh%q-5wCeGVRw{`As!r2Pj~iEa#qnut8I5ZwP)&z6aNYvLVRCv
zGuav?EQ}kcQ9)RS9}%y4wy7?*x5`lwL>+6*>M?jR6=u3|dSCnotORP5dSVaFKf!Tr
zzIsikNLqrEh^{X}Un^$f6Ii3uLdQBOd$&Q!8FRF^w|D)DEw~<QdPfHq9b%t9fBy7)
z=k?<aU<TTiM#-UNp=4)gpL87$Pcri2w*A{TDl05Rs0*qIyx`(Z`p&IDQ9$p=T~LEY
z%Nq?V*EkQpEbsgKcq4A<{`=PeP0D^=e+FufPW44G0fw9ePfQ+C9>L;HHh-Y4H95Gc
zzLIpGJPF(+UwdP7A&<Wc;uRx75R9sEdia=8dq>A<81<ca9-|MCqw4y<)E)(5&uFOo
z9p_JiR~$c=*CPA*FQjHBThZ)OpiWK->YiBd4idYt7MgQu4BI6-Cx4*B80YA0d`h}9
zpwTqs;|#(b^g~-$=c1i$9Bb3ZIp+-<t0*hNBQV}axtT}4U)x4<;e{EQ3bq2tUyHX<
z!)s)pzck}K6UA?m{h(*V$sPPbfitxcIX?T>k|g-NSqWGvs;20wywl}zFE8m+2AC=W
zXGc6Bfh~%hb(~BciCM!U<1rE*7sAy{Ic(1Y7a6D@$2yxGD|_{%!@ItSQYF)j|9P=q
z45_?uf8&lX-J&2|Yds;Da#wSyefH30Lw3m)GHndi$V4qO4+($DvoJ&?Yl^X6x81?U
zGYI$na0AodC$khA(M{5qd!LG3{}e`5Rw3d`Zgav*H#Y%1#!6Mls@dq9)gC+q?=(g&
z5RsP%on8c3CaM310=QGm5Kq#&oq69cp9EPVF~l>#XYa|R9R^QZ?Jo7=fsBR~LE$St
z(I$TEy}#?PZm@_;2d?D$8}l#N1Azq}^?Y*)3=0S67icxKr{%={;=xl__Sl&enDdaR
zajc2p!7#~|@c)eEnRWh~S7~5LfH>(Fe<$!d_+h4j+$qbX`jcT8&3CFYK}WLdMD}?F
z5~$K~&$*aQC(ZtxzePLBA-Fg0?{mC|qJrixZHf*Kn4CBTS?Q2itocKCvRLBpPL}Pq
zIlZ2jI&^m~t4zNHhg*Pl4Qd*Z71J4MYFYLfFHXAs20PULW<t7l^inLv^9?ro_ThF3
zF%R!9!ihiR_29DXL$H5*mKe&%S%#yBC%yO27uqfHuUdL@KYFo-Hja7zWVkfNj;_9Q
z`DB7Jp2o8p+Sd1Ddu#HMPj|m{ddQUSliE3}|6iXK1xP;nIu?XP#GD>hV4T7gEQwZu
zjGNtqDQ2?!HgQ<yUB>ogK?0zZRcz@hP#OWthwLpMQWk<?<!y~-``Jp;pc`jN`QO*u
zf;}-=L;EII!ZY({%WCNMDf7qqR5yE8+ow%+LD-;yb7(>xy65f5V1TAGj$lMQcuoXL
zEC~cjWv?)AU?5<e$fmIx=E#4XoGcW{plV@e3;C3|pr)%!0%w#jgDLmrMmT}%QQJ;s
zBcIGOqZa2{lxi*Fo0Dr$iZOudx9)Vh;DNBKi{Snt_XCR1^zl{-qmj*><Nh(Evm_ZB
zMF1VGP0&b;YbJMi=NLmZ$T5_@SZP#XbtR^p(EM@PSM)VA@?l$ge7uoTxlulB#+3gz
zRNvUn$(lzqDYYuZ3^tRgCyI)njYbQtoPydQUBvU{2x=_7_;A9MF_Y~iuQ9`h5coY8
zO@rg^lj4gU6P3F!xv3X>invge2K?SxYuL~gObziThi)sMu7EAiRCjJBV~E;tohjyc
zfujZQjF9e&#G5|Z#1*~1B?0op`ST2fp^xo1BNX^nn@n3jeOdfMd`CK6V!yY46>Zt_
zV>E?~#Pnisx6PR4>+X+x*cG?2Mwd|del<!8e=2Fs>;A5<5y2>y)e4Ab(x?!#Bu;#3
zQu>U9<E`zB*nbSgJatbq(%=q8Tz?I2N6Dd=(hFu>#akqh+i?=<zY%PtPb3a3R)1$^
z$QSaj#FpPKHn^?+aw|Y!wz+O<8I_W;7U@2Qhk2?`!+SqqS+WVnX-&IEn_Ra@#q1gY
z9i4Mkr9V6bi-Iia>(hjg5Ag8xgjbh(rBQr4a8hgoboe?%&A0Uy@(^%b{Ib=27;U%i
z$l#5RJEhWlM^WWKEfyG~)m%gQkZKDJIX?2?$$h#}fLciRi_Xe*d32a^N}lmh!pRz6
zj4AlQLt1VEp-S8$J-Q7UJ`Za2uoa~B1jqx#jM|zr+5ZXy;^KyXQW@i->kolyb&eaA
z-6g`m8G2p1mvvKNne;>GNVvuS{E&Nt@cF{qVsp0wa<j@1-^0Qa7vckbZduQSS(uid
zf(ja*nB=#JjDO<eDrsHJymXaTfLiqsXo!k<Q^@oacQQqZC<*Urmi(8~cdgvo?@BTt
z?xc>p4uc-L@yCiblm2s!u0PGA)KhCT$jHf0!PTy&LN_jUyD&KbS7&Ofy&{4$J=!TS
zvY!17>;eqWlc?8s2tD_QE~n+M?<=OHvKxWh%$pe4W+{hAXvQmF5tP5Aj-L!ejh;S{
z_me4*YvQWZ=+(}ws6G4gyq1FOLn6(7$((Gr6MA*Yo<K6sQje~VH6^D^v&OsX_D96F
zTa1n{<yw~c)ih41`)>iaW9bKTE|DAN(0Q$4t4ws-@G<4e-#qild_Ij?X83y#&fz6J
z<U$IZ0mdv0`hVRa^GK~F@7I=o2R|B?qDsg{<YtD@hrZT5J(agcb;o)T*~TeAJh@lI
zEqY(btb|s9;y<R%w{`psg#{1(EI-^)&hu~c5`y{{13O|&3_E)8o+<JEdUf^X42M5s
zi=LjwK7B}SI5S&2SsFiN_h+0SmH%SvrM9lFVBxt8Pe;i@!QB?q^}=LuWEhSx8dIKU
za=UWZnA*z?WVnR=M3&Z;;A3MW0kBZ+2kkQXGO!Ky1&WmKz)fw}eB3Yo^@FdMGH%ov
zv35Ac?b_1SGl8b#0rXK^+1hnEuYQg^ui;05WJmQ_{QC^jN!yY++H8B~&nPmBwnk%=
zs)=ddCY`T|tZP;0nR66HemOmhpm891NAm~!??Y|iquJ*A5mIS*f-kH(hk1WucubB{
z=7z(An1L34F^aalpT@`KomR>n^&hp$7tB(sc#UJ$s8E#lLX$~^<CqyzJ|fh-Ef)NI
z8|86)GP@BYDSJ(DkU@s0(N^$__qmMZ#hK+<u4VSw3|F_(An3rxHd#{GZ@;Ps&nkmw
z`;;eR`ucYx0eH60NBr}5Vz@+nZwrLUO}uWJKl;20{=8exEaIO!#*bmwP*#fo9&T*#
z>!_RIY`5}}ha;aJ^7&@y<;JZT@SUnCgIhhF?Mw~=vPvbg6rOd0*gsneuNig4vOGZ-
zlhP;=`agSn8q)wT3j-uzLO8Q?R*P9WrWhJHGdEXhS>Mw`z2N9A83aUre)7*#zG5_#
z*1x+pUAii&4K~kYFf>V=CA4WvS2Mm)lx|j4PNWPrD!nD=c|R2td6g6WIZ2DW>BV${
zrl2s39(vn~M0dT1tIhOkepEgc7f&R=moaH`HN$>>u9UUtv0$=VP$?=nUL||ix??rR
z=JEW~u@+u=vY9KGAN!*XDXsI}qz!JIpYL`u#eSf-nXE;wr|6Y{;GlBps(Gs-U0Z;Y
zJTNwhU}TmYkDles@Fb9Wo7nx=029Kp&hA?+7Ud?IaIK|bqT}BRDxmbB(OUV{-Hb`2
zWo@JqS#bodU3;1LQ&O+)jMq7Yn?6Oi-aW^1@+N2rkF*RfLK)8rJ|I8n`+>P7SkvGE
zJWh)=fO!xE$be)xGc+{=I_@In$_VfX3w(c}^P|T|Ba#a`Ykh&EyyegJK@d<_1E2|&
zXwZvY*zoUql4U8QgxoHz)dQg(0Hfs4VaY>Fz&&(#L!0s#d<$@Ns&nVnv4n%ee)vCl
zY=}~N!j{H+=A|^sZ^f@r>pPnZ!@uh*V(?p)?Qd9V+f+z@#baDZjx8PJ2MsCgKCxpU
zHfn^}@@O=f1!kps!pt&J>1`5{IDWVrWH$ZQD4^vy`dKNyr4r+4@<PO&>m4cNxbj>a
zoIfohT?Ak6QdDj!J^Qj0wpWSz6M~S>+Wt?WV}xT0`nU)$Tc^TkJ&vGprWVeB`60(B
z^|ez-6_U&&_>z@z3E6B`oy}87;2eV?>}`nQs12j!5WAUHvr%D9hhUSdist%lSSM`Y
z4<6=_Txe={tGE7MGK(j+_eN5FnOiJuG~<O_P*4{>a(V2I=#CKFR!x$$5l|y2DJgwH
zBE#v*ac~BD3z4*j2Cy2eDeE|DS4smk<~PHQz;!gdz4M+o{}c;2E+T;Kk7iNh7$cFV
zK+@z++S8@^`6#%E3ycDqEFJ%`j>Dbwk9T+BS00%H$3uC@KPE=no?R7}>^G*#AK*Kv
zs^1L^_lxuM#z@At<L8!kDnZbg_gADLPrM~{R7xsXSe<pN8&IUaEoxZ0Ui|EJ+;ns%
zcAAuk(OUY|{e@$Mp$o_OYQ&+^D%n=XmxI-(1X^cHyG*m3ks)qcR$kZUG94eB@|=7v
zpp8$TC*bDOjbxzy|6T#sL|o!Wja`YvN=Y%7;L52mnoU{!)VlQ8h!a*6X!i9iMF=$+
z@(C@(6ntxyPZa`~!V~Q`RklZnNOb&Z{yW{g;T;H_4d6G!m6PF3H&cc*PP_lKe6@qj
zdK3Ly`cG{iaEShpAfqYv;|eo3GNYYTZ9BCFubgfYkHJF)MM2}92{NquaDQ>Q_+b*i
z;Cz@+`=+GO0$0_aHY5aL9S66B<0oc`ySTgnTal775ci{*2Kd=%RXR0a0QsOk_};&%
z)1n`v+Ne!Kp($Cs0}yo>r=F*{TTPD8#MFm{{atxg@3pT31g+ReNJ#ej-yAClSae2)
z&j0=*>Cqbp%cZyIvHt*G@q~<wEYeAnM_=bXelI9LAL&IxwfV8dY4Mf5a#a7ZZ&9zb
zU-vJRv8$76ZnSV}3!nXkwcxr#&p4FMc~%s}Hm>CZjeI3t*%YAT?)<~w$c3^!p-6gh
zindk8zpC9vx;E{q1v1`tnO{5h7Dpk)%aq|)GJEO5Ug0PMPEhH8=UIcuX!2Z(XE;#z
z!4UvLL*c&9dVwWp9r#1-No&Xpe!BsWtVhI?1ld5k6jTB$blO>Y;$D8p1J-)DB5owS
z;FNq3brvHZwqcMQawM_pPasm)Nxx5@FjC?9@t)>qWN*sr;7GwN`Di`yC&=<RDXdi^
z4tw>bE_b>pgh}e0L1L75OFj2Fiis34K3)69FM;JFe4bVwZ5{fV^jSMgS~N}*P?nxx
z5)%%8rju0iZ<s07U`Z2a)2NwIGE_iF3)ihk@lLkbzZDY}5lI^Ttz}9ld3kkZ`S@_p
zil@^&YTanx)ue4=jmT(p^s@oU$;qiwO@WPwMJK(q=jiC@{jzDvGrvI7^CFUp-!Y2+
zKpa~?(2{EDM02nVk!j>#@;)v$kg)u=m#wJOVaK+KFSFy*pKZ}qwB$#zOOiK`eTm&P
zQ<x*-6uBDttNT|>4fpgc2RZE60?}rt8MEs9-s6aTN53HLVpRb3Vp)WwW#W|&bgBK{
z{X<`P3bSF+Jng$`^fSusw~V~^QCdkP6QBO8SY!W=>BZjs_pVGV2bh9e;CaL`(o6ez
zrl91CQsaqvekQ~TufLF0^z}<c+>()x@Qy}jYcOjRvW&SN{r|rk$!If=wi`!jw2E|T
zeWl64kVMmq@<~2gw0zl6<C9Ly@?hIpDO$9|XSCYQwAu+Z_j!A;4YvL51c%pjS@&R$
z>eF(c)z*zuOm2d^zAO3$9tW(6ly1_cB%sytG_|x??o8y-H)nWmFe;^gQu-Eq&1&HY
zN?o!gV_b7xof<k!r=EzkxwKTKDaA{o#F!@l-fp`s+Lc#W;bk#>0aCQ5myA5;t`8-K
z<>?yvBOwcO^KVNl{PVgE1Y`u*Ij1x>7eA3ST8?(TlWVK_y||b!w>PP+KZ@2Q(YaI9
zxlF6xD0R7k-NpOM946JYq*2kxy4d{m`az&4#MQ8@qNXKryv@fhSHlhl^`?(eYJvE2
zr8B@=k`ECX2rtV(CDHD;fgIF=vtM#bOFz1*2PVx*dZcxG$P&0GNWZB;?udI!)2QT$
z)B{Ncz(C4*iPZY6_i-fRIoqpe@WwIG&br`+eFpp@fd9l()3o#z0?&N-ni%HletyX9
z!OZ_x<gb-{Xf#M}ReA75izEAX@nrDQ*!Q%3bD*gX-ezl#1D1{Owi7_suGSMm?&yQB
zI@VN%Gl#~+HB1I`&J`uRz1@cyGiZ{&(-s9>9%Q{Sg`;iwRWk3=w#@KqM~7j~ZT!4F
z4QB|VKijC~IQ58U)1skX1p1(!YgQdj`Pw_M+<HX6u_w@-{Ym$v<+Hc&=N4L}Q^IB2
zZ>m5;I4Jw4{eB4hJ>yI|`P;vyPGPn=JeLk-iK8x)6n}oswb<DdO!MWFk&k}!LRGRB
zN<h5Eg&f!Z@1QE$BXst_pQ8+0Xo3#7QZ<r#NKMeq7(;h-{JBahRsRK4v_bh;w#Om)
zT*p74zqr)XpK2Tqq1%&WUd`;bEi=OJRr%jSW@1RE&>%@`mfxHyAFdgbcpq<l6Lbw5
zDbX$whg3%YY%e0IpPb6Kbk@eh{Zf9xt?0GfK<@Xt(4^l)YjYPM;Cz{RBRO;+XW|0i
z>o>96^?}!Y^JpLFKL)KbhG(-s;zvG7vii+2lvl!TVxp-TRxXlrx~kGM_&(d8;b|`>
z#>rMo-8*H8mww**tBcZ!X?_w{#S!|N+sdb2zo#l2sdIT+ozk4_7vsKX+=-T#eNtZG
zys??j3(4M?9#(o^v*FZqr9exTvSbpsgy|BdGD^vV;^vYP6PS4Bb0u1DXETYEgQ8Mt
z%Mzwrz$RHq)JAw@nMQ>S4w3$F6@&tSvx$|z;yT0qwUeK?*0d!f(574%xCOvFmj(Oe
z+~F;8@bGwaIfVjagzVz}E5D0&`~^W<yiaLM$Yk_0y6f(Fm_K_S%CWL=Pto-!s?%-1
zkzlC9t5u4FVtZqG#s|CxEycIgvXPO}A0+I0T(wT{l#tF%%Hzy)ZahiWulhE55x`wb
zg5!9xNog3ClJwjm;-^NkV+D_b5|D3~yM63IVsjvW^X3=Z({N5(hFV#ZDq@o~5wd}5
z+52`LhtIFslJ=yR==^4OU+jh0hB%vA6VMk#vXQZE8uV*9mtoDwH@y`}TZ}p!QAv*5
zz)rF`eCFBa(L$(_#J=yf-Qj=r`TpqGDZ+80Y#?1Jrm5vySDNrQouaT|#~$deimVyD
z6ObJhg*+Pl_q1>zuPedZoB9KD-95!qVMy8+H}nN)?Raw&)RhbPh{2Q%<_{H7oOQ{R
zT3ProqESP^Rp7VPh)Y|tD5K8?+wwWc(9&<~zFftVJe2<+AI+*3<b?FF+>L77b{?Kg
zTK+yoRoG5)Y0Z6<10+9uEs2YIsv%ShqD@@ez&tlxg8V){5BP*)y$IaE{K~h`ie;#p
ztZfG~K|DbcommNwTtwMI3W=`TO;Zk<rK?OckYbr7(uy4s)pQnw1ub1Zf6K3&6!(vH
z)|9ze@ox<!xhF;~CtG+~o%7CtNWO+)YAX3sf<mR&apM$c6`uXdj3JZPT-SJI&EGIg
z!|7YYoaq1@d8<n4%)_TF-mR0*2(4}SlbLVTz`@FLe`uLP0wV%Z9V%#P!CNYY8Y-v-
z-ckd6k*z_r=iZ4iM%bIhX;k0XX<SqzQ&Z@MznK-W(?U6EnfUTB#e-izhc{jo0$aQj
zexflsi9+ejY}dC29ww}9BoXW8guV(L4H{cyX=dswl$Tdcn~%bbpM3)1Ou;>kEzQ2d
zDAj|JO!%d$Wb!vzfgC(@o@&z`6Gml2V3HNNxVY$c0TyUo7=T7ylx!ZfMy->aX62*E
z=<s8CDMqd%r@p#uLcz=nc}hdFVniU#{WW_mgSxH;e^J$R!T|%Jkrc}up7P<tdrPO-
ze!~=ea`c+!>Q*O)g#l#!>E1@}W>vx(nD)x4P>cQ=af;$T!%}+xO1VmY+Tq=uNeI`y
zUUO+NFEgRWcB3@QkJ+x$-wcaSM_`TFYMBgz8hPhPzJB29MXMm&!e>0L0vj{!Q-d<y
zaTa}@ho>tV;HuCcl!FA;1Jp(aIR4b6QArcg*rX;LPdS1!C$eA<)z2b;-w|4vOJg<^
z!PGKCMtD|8loI7bubM=tuOn8N38|!?9%TRhKepa7D(W_T7ab4;6hvA<Kw#(;1O)`7
znW1aw9=c091ra0%2BZ}UfuXxgx|9y-lJ2han|JU3iM7uMShM)Rhk4?@pX<Kjk0#Mo
z%2*KZr@rL(c=?AFQ~W&sIG2F9rX3Iif9aJOp`R>D_53$%Yj{k_q20QVi<dlzwzkl&
zb1|>w>fbz>iZ2C{tcs*`#H+a(JL=7ygUBI-v)97E)Al;wag&i4GMMl~k^CVK+31%v
zuC#c6{_vpe)yNB?>er~Ma95(@+jO5x`rz=<Y40rWPH?4ixu79n*7DD4Pl@-Ov$-V3
zlw!+xjq-@imtO`IF)CP3)^y(Z34CI{4`<ol@2PeZ6{z(yO1fOK1h<X%mFFhWZC3o@
zV(A}(tseSe2ks(iaFdDeXam#HmtD3E5WCur9@r89fJ{8_L4n=H;&NKy;HBB3NF!!f
z?FDh*K5&FJTbH4Pfk0JFo@26)N&=&|q8$H0Z5(`!?+G2en6WRS5pj3>QG9Qy=(=$x
zr90@GwG{s+((lL|B2)8t{tLhDXy^SOz`SMHXa1OHn!~911z?zx>?B5iWavJw`8Dc)
z;ps!JP5`4K|5}^a1uxmQXS)?3w&o=wQICGU*3n9%=Da^3y#JOVd+FPZ!q=--INQYt
zu$v1%8+d8+YtoHTnf;rtn1*1MYF~j|E_RCwIITXv#KefLYCJo?p0%_rTZ?mbhMKJA
zOMinT)_fM}j3=?Z0`PK5@=;}8kg1sjPPEjtb^#}}ucFhw{n}s-_f#At_cMmT$A@<A
z$nBbz_L))w!KXBK;>|{f%csDXs$1Xr1BXWi*nW^TEl@LX4?dMBh`<5qV|{A^I`@i<
z7LcbFuy5HCIefl*iO|8{&OcfF?~JdXG8rUBW-<a3Fkn*#GP#<L0@E<`r6|#Z6+6tk
z@pV`rB({#FPuDBwxG1y#s>oor>8d_API9&9smuX<dux4VnA(Tu-$qx{7eHA@txng+
zRQZzM#PC|Tj(2%S9N{K**{i&ab|5SXxV81&wKeKc4!BZBgXXnsiF#}z?<8`O^4v|N
zmg(+^x<hL&&}{ui&sRgj%1WfPB4{gu`IFo?`MWqcIP!g^IZXM(*P`U(1CqHVv|`iu
zq_AAdr_Yu~<29yvn#f-Dj#XJaHvgrmZb`4z6s(}?joMp?sPK-MSlKL7r4kVpA#<dS
zs(LE|b0l%tsP<sgsWS9U9kVBxO&-(nsYsL3Nj0TnT<>C>`*%2kSAo$O^5!YN#d+o2
z>G|#Do`1`)jnB3C+g<00!R2=DhtC3&enn&O`o!_o7z!NaX-VKd=euVQsOB~QtN^+U
zzVg0|&j!AIGNk!%UZ8m2%MyY&DPz{W{#PJ<lMF|=uSx=A$bhCdNFQ&pg-rxqyF+4~
z4~*6;B15o<Bnl!`lVil)2rL+jS7{dhtnrWSVg}E%QX3%YJPb`j6sECzDJye)4y}}r
zYtLN)UEk7kz%GJ`#5r+hmYMlP7TjPdc0K)-;PcHn0M6n$`pnLj0%`Fax|MP{mVID}
zBZ`&_X*A^c7T|`ni)cvXhEg>BlZK`YTU0>xk#&ut=n#(|duD*)ttZgr7iK8?`S~%L
zYrlaA7-k}ri1|H-Nj#Pg_n1%x>6Ty^yO*w*074AuX#*aUQyO2k22TiC7g>lrawC8;
zP1Kt^G-P=Tf1~rO-IdSo>6=<M!*fjJ{IaSme5X=Q5bqL&umAc<eC#%~{yXh<NXw^{
zSh)VUmRWU?Yk>D~5Gn*un9%m7XmEW$u^^@<hg8qtFCO}O#Y*Nu;QR6i4nYZlIzG;O
z=L)(}2G9kIHJ=1u)j4$Y9x+O|cnNAY06sovUH5GPa2(<fn4WZj3OP8u@;dsP7{DbG
zs;uut5E!rwnL<~$WJav@iZ8yOmhi4ZG+51|aK{4%J_p60d(f-&njlzYyF8f_D}rsB
zz>hBR9cB;DozGi<i0(AMYW->W(OaiQkhk}0!+Ll;G$+CC&8XMiab^qv8;}4-br+?(
zT(ygsz8k^|$^1v$bSrB>O2w~Gias)~AYaiEu;Rk`_zl3{g$;-PIRN}$44Y(fMmI|{
zmk{tdp_WB6)j5fq=0iMRhwg>4?+)J+rI0Ue{M4Bxs#rq#q0t1LF!t2Mr@H1<Z@*8-
z{do0fB4k)7yh4eVX|Ad^=I3o}Z}jL~S4Lr0pCY}zLczFot%ZWpd%Lj`Ac@z3Dz_)4
z`uktog~KRg_VZ|?f2?p9Wy@I~y^p?h)}oTNj+Teq@8(|X40~NuA)obYOzR9<qCrRT
zSUkGJ`y;_j;9eHW-+Kruc@SHCV%&pz-~6qayj2o?)xYY&^L_Kc6L_{tSYR~_LY^s5
zX6WBNT3l;%>8VAx7a<_pS6UxqTkik6dMvIy#*d~$pQ?36Y9?r6)%l8osyohUF@E!3
zb`%&!I=oi!1C?0N(Cj3#<3(qtIwmgQF~vkjKN%YvQ_u}?nLY)K6U}mu{%(j;h}|8=
zXfCf=z1^!_IoLM3r$iG-5@Moap@$J#jhF-k1d^*0ZEfNl2KA)0SHHtbO*<n@81K-X
zo9AV}QM5i+*xlV7A{f<Evo>+uObAx1t+9XOQpd5*ti~p;xsenxrV~ifqmhQc{iv8t
zWSnAYEFuf0#Q6&ji<hCItgUxUy`xl4<%A|x-ChU`!k8O@rWjJG)MxjVhDyF5DybW5
zR)ze@HP5*$N7;npw{iO@H+HIjJyWLEu$km;ZWhv%<E977_<J)8tAyD_EL4`S%k5ZM
zI$^X3w*LOYL8Vo@X>@4}i52YB!g%+%H5WJu16Hdf#lC+IQ_-K_t~y2%KJYAqykK&T
zZvnlfZjA?CugD`wPziwWSkRFIAdb{AL)}z%UHSjI_!Z2;E>ZLW*8tX;ruwBkh}Y(I
zQ|67Tz`y5HpWsIjEOGAB#IjduEbJUb>0!Hm_MV=Y<tY7!e!S}D8-uBN2xI2QMDpD(
z`DdG}KtNX;ntw7jQLjZt`<w_n0Qgdufi@<kyo^YV75wTFaGg-?Pgclf+130t9TbFd
zn!><3kfiT|hkzWiFmToXqi2k7St4*$v2SQ-(7bellpjLC(~P~|aGd(Xq-q0gLt)xx
zT1dH&X0_)d#t9XTapHVpEy;Y1$K?}2?f^M#&b5!;a$;^m0gw@g%<x=l3jdss&7NI`
ze5v325Zx*)brZEe*;fHm<RrIR39pKq`eJ%H(bx6;z1{UOe5>bmmAymD_1%LY-2T9K
zJvpKnQwh({2Zxa6b9Mc(yt#ctHmk1!L_+&@x;H$lIq0$_-dV$^xev<IvIjQgA$+Sj
zW4&^iUCE?8>)=eBmq{(w`2d!g=7>>S1vH7~9K|OfU}*c0PK!&3tQF{n{#$|19m9i`
zq=x2kv4ZViNz6rt)n_Y3_YI1G9vd$oj`ML*CMIH(Z%<Wliv)6!Ec(}ZanpHU7F?Co
ze3J8(gPhwia(>=Wfoal^T}vTNC(UgH0sx%;fsFE99_{x4iF5qSC4fjx@h<?Zq3PI~
z-Y9Tm7q(WMWzHTD>-I~l<x0~RKpZ-S7cQ01FTvK`nuTKGlwp82%h+>vdKx<0J!%#v
zpM<0szJe2o-kdbm)|gYkCI{ZQSc1PiQH<qhErV%j9{ohZi|i<?xDlFR3eka7@p|qB
z_OHpl)DHQZslOweVCpHb-}OZP+Hz2DjPb0uV+|765$1ZTM^gRo&RNX`JJ~BR>0kzE
zddeH_Hw|1i!!__W$Ng-n1{t4UE`7f`R6LrEVHbJI%hrsyNc-QUIL+c|D-}i>j<IbH
zj-&;?AVaS|@Z6!R2=0z_yYB~1gTiCP%Nt<eiWysUV!V!Pgyq!Fa=*^30F>g`w=SV0
z6que#mubrARZwGfS?BB3Y>W`1bNzBC74i^qz=T(EDZ+WJCTl-MbV8WHi)~hD^NWdi
zr8a>_@2McGW=u=<etPNWt0lsKXJ|5xRNODQI)y8^%VtiJ>E+Rym7`;~!h3M55kOnh
z2CS|_uDsUyGQ-iFxGmn(5RC^wq{=T%Ed#RL-ukIh=j47{DWP=O@OQw7i3Sbu*l>rB
z@j6;tv)b%qAt{jE{X5&sS6+_~-LyzVTUJC-iLa->J!N#8<$Kz=yz7>};_O||-IUHs
zn2OW!EY0weKTd9LaJ>FbtDajgSMR4eNg5kQ0b;$N|I>6yk*j0j^AI+~9i~s}pj@^@
zUpkw*y@z9Uyr|oEXd>AoaSyuGS+UZJ{7G<cqW9y9_1)jXZvO^pm8p}{S39e44$MI%
zz^yN*(+tOYAPQXLd_iIl-V(Kjfg0z0&tAG}W%$1eG}7>X35Wf`TfAI9U0J{w)9`)|
zhkCZ(q|=43!oA~Kc2^sM?$?iPrO<Ph9+lOGyqE_U9|fQ9{eu#IMmI>V(Ssnc`#~GJ
z+Vgghm`g{yPd=s=nKNwnc@}Csn;nCS_`~$)fJUMq_lYo00~1G$Vj_waFP6SRk%@k^
zX%U;dRl#_7rs@JHt<eNj0t(Oobh3pZ0QM?EjY2e1l3H%o-U6%u<c<yYq}pZ)tYCj<
zM@mmG6&f>WQDJ3ouW7!9FqO~gyOQp2h6*db4By}J!f%>>O`c;DF>2@+pSv{+#1UlB
zzFd;bY2IQjP~X+?3zQPi4Nl&M)O_UIEYG@XMWr^oJwFj;l;u-MqfT+7%AJhZvCT1I
zUK4-@L@|%gxJ|1;$}JvUhZ9?Loai)&@%b~h7pNFAV@&E+78DHxD*#K8X1_n5XuCg@
zLfgJ<UH-v~qukE-dW4{phN*(uOQw;$l^Hl0YKY;l@<01lq=4lJeI4k5`F8Mg1$^(%
zqeV<;IXtarfi4r|L=N(|)Je};-%NnB_nj-yf75sVln|Em0wCocef~;Mr(gu!Sm*=6
zn|D!gi7x?3j1}zjBsQ0Pgd<cRr*Mbuzw|a6fpg4p^yi0Ej>+FlbN<t*`MHpLYhTRC
z5;6b3*Oy@d7dXDa#ShDxk&??%L%Y5(r2^RrxC<9uZ&V<FUtSRxNkBUuvw}?v;6*_N
z5NT)_@ril$yt?=@f~Xv7#hWY$@bN<=?A~EaD1NSTjkA!Ae^E%R@B_*70!IEQ6g2Jo
z#MYK#F>jW(?Yu@L#B#j2HA{jT5fql4^q%0|m?vI#tjaMZc<|l#EfSj8qZ5~r<~?r<
zD(?I(&AUR~q+ZE(`JZlDg6UGkR%@_+77=&v@UVB;sBe5Gj5#m^50!-{kb2~rzRLS2
z*zm{AhEIG~2j6KO#A)__XbM!Ii)oXf*_2Qk-xjH3E%zS$)ceW0L}+UcQW-_;Hl76y
z8{ZZPg_(DHt8u%l))zM|oE9sj)4E-J-0Z%hS-4tlt!}bDo^E?G*z`AxT0B2nM`HD@
zC*XG?qx)>MR~Oz#6Cc0}u>}@D-RoIv9jF-R;8opIxf~p++|X`Nk4{ws`V!o(qiuzN
z?nqjx*hHT3nw&lGBnULp@_s3G$PTi6a3L5Ng`^q44-GE|{?EGyAVbIb*)^%Jz4t^2
z^udxEH5K*vtBQuKE%I5|@f^jUt|t1QaT1QFc~1RkGhX+)IkI$%WTI4tb6$u#h#j1S
zNa;Y3e!zG)`WzVIezgJ|f}`T^(>HPe*I+bJcech+mJyhy|5tecrB(n`%|`4Xh;VTb
z;5uZVEB3h2_4c4SC%;EWfBe#-<7!US<7VzGNH#!@O?|ba%AI(Kanu~+C1>hm$=+Ok
zy05)5*zvtrS`O=dQIb@8k`yQ6r+m-5AR+G1D=BD_eDdd#@N9&IdB#)e*L3u#IJ=yI
z(Iad6CA$0x5;^eLbsE3gslv>2yzz)K`gmvnPcSOxy&`>2#7@eSBe`q(mUSd<DxNRD
zGR4l{F&r=a3!Rg-ihqyaW#@!R1O^l7Ybq3C{q~k&t9B_v`tJI1tl;fC5{hTN-O#Kn
zDS-Bu73A05ZYl9cVaC}iR)EB1tBY)1w@dEo`4(omam0&`pUS}Y{>8)R?2B@?t7A&y
zx0Xji)tDEefuX>%yvSY9hpk)wSN$Uq_ZUE$^*zp_0cc$dxFGv35+OoK=WG&8@~ySw
zH&5xq%}MtxnHMk)U?4P(!n&*)zjv(;vefYA;d^=q16K2gq-fr4uYT0l+K-#(Nc8%`
zD}6~?XEEG>3mAZwS^6G|KmFSOl07cxcrbI*z<uAc+A|-n9(#VdHq7$87ep*>w;8;4
zaoOB`ctB*il>B8f+Q-Nd;FY3TL4<%t@ghWxk%Ch<q}QS|g6u7jHN(!-cQ<^0Gn^n-
zo;caG=YDZD*%BQ_oA?qeVL<^92sE=$u}XQe3b9e#VZ4@hT=|v$ZDE;7UKq{$sa|bk
z$#vjV@<2ew>3To<bJ+W>_dF$-QC#5rsF8GF(V;>xYCt-I3<4uIBY_~Fr9!9RI-K==
z@FYLTM<iU6@i(U6^mncI^iOlhrDvz!i0UX6h*rT1r1c;_$B$4wMf7!0HM=HoS1wp&
zLbh227h9Jf8(TAj;K(G)u4}U0PAF3y)4odq0Zrvxb-%JxQR%Hj*rcAr17xzR5U4mW
z5%an&USWq*o+*9Zs>RR!qLC-;esT~sEo|}Q#i6byKjqyQJiUj-i<Ef@9C}Kh8C{c%
zFyD^);`vME;PBy0+pArHi(-JIMe-4JN7hOa#dFQ{s6sRJbD867BYhHHvcK5xhotV`
zfLvtwN}PCvr_6|r{MCU6YxxVF&Ha0fMNahDHFM6v0WRVSTp}$)jKu~fgR|fNyDU~@
zq^}EyE@+|`mrX9g-8@M`R@uz-@J4Z+SN<`RrPa8yaWj(Jl7Js<`#X8FgSqCOU_o>(
z#d+UysXg>RAq4t1uG4qW_j`AoJp4=J5Rue~wQw1YC8@xT?pB-upERXSQ5##^!?+k<
zEC80dOxO#H&0oyn_G66}+G0U6u`h4Q0cA@({EinJ+nQmd5M@gQ%bv0uk`>656y4wM
z9cI|4(C|i8#_`6PW5<~g0u_IN)M@_aW5PGKN46t9`9jYpBqjQyD%rwIA&{t0a=2zh
z!2pR#TEVmLDnN_cU2rlb1@6#&B|ztDZGZh4tcR!1KcBCgNHDHFChDGy+rX<uT!;u+
z%gi&^bq~N+<+Xo#jOjw(weBKkSs`c1_pcv6i&^-ys16rAK}y&v-fD?q1-l1a$y1o4
z6E9S_&MUW@2d!iFsd+%iuf;x@0@isR+f>wxiAigJ_ugCi?7eKvD@&w^FvjM%;x}sd
z0u-by53XnM4sVe(e#>1FSn9CY*Iil<tP0Tu6OP$qJJo>U+m+qlmy~{rq#;dqqW|to
z<5`a$rYmPc7?IrHMI8}5y#ruNwY60&JSK(V=x+H8y?}=YMx;`b&)9&2YqsIBzM30=
zaZrtS7p(}C;n+IJy#CVBgLSB6$Wvn8?_xX>&W!uCfg88fq3v{XU@7lK^H)q5b&==o
z*z5;Cx|k30SGYSOW<EKey8}b8MKjv#XY+R-Af&WZXwnqmc^959L2OPayu$;S;y)Pe
zWEMTd`wL7qM$1dkd9G+)g8!vF-fsr@2Pc4=vI67*s|9bgwgwWL^TR$@73zK^JLJn8
zW=QR4NE)WXnBdlh5u+T}w9`u|8G9aDOygFw4Qeu1Ptb+_*eFtwKq20)WyZlI^VPFq
zdJd!NNJ4h==zj$>QnR1ks32vzqC4{Y@>tr)7q-82a)#I%Ci*I5D&4_)dU*^HJD!S&
zLi;?5v1-$wTi4uSk(uIM^7J3E9zOUsjB|DL-iYgc09WfbLinR2b`4@3BJ-wTVhdM`
zKIR>LVwVT22n-jzR%I&_*k*m-cmC;ew(tqR74jUIY$`?hrNQ(>G2lx^c=B}^KL<<4
z{x;EQGM!g35kE(3C<)^L2ULUcRcjMN^A+*qyC)!{|EBVAVryQtp>BoXbleJ3H4BR8
zAX;A5kQL#=q}ap5z`FLJwD%`Ob&IM%ss4n^d;i|cALb$)c>mx)6Qn?}xIqjchTQa(
zjU=rBsr&fc1t)?3wOO{cv-%LZU>bmV?INWUXC{rP+TL84Ul?;b&IynWB*A#m(N7y4
zS^c@spianCyqsL|HvJ-IH=wgS!}&1e?6SJ1i2zWTwcFS9E+Btw$QxY|D~X(HD6Fl`
zmDF#s`EIFaw!0kwZxm1c_5-PlRvb}ah&aIJP`mUHZIx!_;P5GPc)L<H{t^3TDK<l8
z#3~zW86r5PcRjZE-)jv>3AwdE{2)W>a9CCm0k$S9N_Ujw5gF1WQA#bZLw8d{4X!E7
zJU*`INf2-7{!u$!@yqzJR2->owjT0z;p`TcuE7s%ooC_`m1SlM&#<fI2BP5F>|^A3
zW!34#nx|i}kmad^FXIR4Q+w&-6W3mkY~q(E;~ypC6HmX}Di>|k54@7V#2)(z)~yrs
z!#YqjwWq-t3CDaWm7$3BOaV)d`9Ze2$}uW?nCW|{&Y|o0H%zSXkS$pU&94dDIJ<Z6
zuQu-#g$2&vl}|JEL5r$#S0Z7tQF(pD0L|+h^Fk=Ff?eb*+p0L|O)9AoFzAAroXxGf
zY?bilONME9tH7mz_eT{-0O+jXLux#{3xpVr-MHrpdBq&cl&tGoXte8T6sQ<<KnbGF
zTq*zAm^+5q^B-M_SDvU%+W%S8OxjB?IItzqhxvKGz;SxJGSI*Q6iJy>QIhu|)98_H
zrjO?=lVe*WCrB!nE?y?^%InL)w0(f2k^k~Sz8~;tY`b2#zf-N6D>Zr0d_A|nY|icT
zJU0a$cre-~Jb(D45s>C|y14?xiY2lV&(6wyU`qnn)+7Dfc8Y>ZZN}cdH&@dyyssbr
z{xkk@y;^BU_;&>LbxNuIxv1UT+~;(k)Y68s=hrEEntt-Kgfk7-zvm`ng4`%KQJ*UW
zk@W0pc`OLSZ=LEyYVqM<HP&P`iwr5=d}hXKCSnVN^$$AA;Zkb$nX&d`@GF&5*6i&4
z>=Wzi_-4BRe(%(j><p<u<efS{9#{=;Jcdmk9E<2ZG3kdxzdFnLjf}c4C7e`V6ejM1
zYarOwY;~T-PN3TdA8od__hlE}w`3HqVVvaQyo#mKpto=1eyNHZ%LZbl{dpR6^^RxZ
zoJVZK#Cf9Q$PKe_6oR5wFz$FyjC0E9Y^97D=E3MBl@!Jx{O-jBHfsMN><nA`HST~s
zg~|%ffk?b)D%KHpfKh$XasHVP?cdb>y<y?-au(ObZYN8BZH9XNIl0;RElc*mV^Cvw
z5-fd8{_z)%Gdn0i#1=fMXSNKb!C35M5(yl5cZT6v=e3Xr$iy(`WU7-$?ps(lcwye}
zU1OU-hB%9GXt}1Ov6P#ap&ylNY8}TT5{Vu-U3-5S$N{?);7<HEeu*?~xr}_7NygGk
zERoBkDv-#=US5=WyG>$HH7w}h!SgkaBCBd;Hwph1=Ipim?)BO06@r*s;ssG@rL?qX
zv^R@5&VXInV!zQH@7@0W9bs~${w~k*;zz<RAepo%TP@B!?<0_ZjU`L%PFih8H0f>Q
zxyEi34)k#MQaPUedh@}wxBq1S?dRP~T5!>)&mVnyV?6hXV5-iC@Zm!oP2L@2p{SrX
z{D*ux*?c-xT>Y1gcLb`;1v9TT<6GH?vLWfebzdj{(yT$T;7zO#)LWMcU{6?ElIV2^
zG0xS`GAekZMjfZ-t4FP?1e_;wCzy73@#2kt%$hhlzMF<6$(pI3=cR1=uaOxbJl8_!
zxG$hDV0h)vA+h6|WeDJI$kt5m>D7rJ+P?Ne(iA@MdcZhc;AVKKoYOm#;9`)-h9%c7
znO+e8+Sr=g^|l#tj)|Wgc&CB!k)P(FZJ<I_m_py<qh1NoeQf*Wwv)u5!Rxb!*E7G;
zi=0j2O~kC_hp%<suz6NSGDRh7ym9&JN7G8ZPCUS!e-dc%%HwPQc!V2J;#~<Y1FsnP
zs7+Sc1#O+pb8rN@H~CmqGC@4rn{7|kg6Gz&mODUq<GAPVk{xXXL5mEa=7C>>1v%4c
zFAm9$uL6YDX}(}o0^@f8Zo-y}?V!N8c@sF9%8rRDIQfPy25PTX_fau@*g$<^po8m5
z>EIOYW>;W)&eM=SQhvEajpcbmWP_D(@H(R8uk>eKp@8Sj71tPzy9azueaqoKJ=|Ws
z@&~~Cj1ZtUGXsHYsF5d(z0Lpvf|f6h`wCfGS+xQoiK6CNJ7{@~@7?k8yhb$ONuq<-
zvv4`&!J~yX)<DM#83Xv04dt_Et-#1A%;^4VQul~~FIt|f<wcT_7uknBTM0cqoQKGt
z$+Gk3OV`*Jp_q>lkG3hsY~#t5rRl49^ApX#bi!n#kpVL_{Ik-*Mc{t&QxXch6<+&x
z2NFoB4OFV!hThIhT1zX1lib+cpsLCY|C)=l%jV5Q<>T3ZvspHg@7TNRiS-@_)!!$8
zQwO&o;pNuZ(us<CCIW@)b#_XS^3dQCY%PXgeG$9rCFBuXMehq|^U1FWx)Z!`RY*XO
zkKU)}sl)vsahf#@2S+qFZgc_P8h*GsmR1#sRakA_MElYS^Sq6H-a**uY(3%E{e`=G
z7Ldpf#l^eH)CKTfbB}xdh_euTus8-Kv?6yFL%-4hRt2z71Eul~O@ytFqp;5ZLaw+%
zBA*DVpV)%#=5W<R20pP<{{=1HuAfSRTT-Be1$|QvN|7XgaxQlK)QWDn1NjDWD#Yv%
z78{TL!y$X6rM@bsw!E6RYR1t6RLE#G0({2m!9nKw8!5q8z?`M)SBU=$-u!YKuf^zl
zPSnkKceN*fK2EUy9Tu3S?@3s7klV<D+-EY4Bsq*a{OXe$<p=<*2W)xX%(;<RJOcDi
z><&Qg2ZVw}x)rDg&EL4946UW6iaw~qz_vtXK&FTqdiV$AG^wik0mRp4K*6*EIF*Ze
zdwa794Am(a7>I0bZL8{Og<V}e<S;CMCa;BR&(UfGJ<y@OKHKkTHKwd<1SeDCz5Xse
z5|X5Xu;q=zU7PSsK_OMK<1|A@mTk-43roomV9&4+2#{*q>#$7N9P%d^*o%#n#90u<
z!!_gLtnDR=6jS)0*u`BGI)qv0IaZ`PP;s(*X)3w#n3e(i0ORPgQ3){!bwS0M@d-%7
z9+xA)sr^vh5;?$Pl*}#yMQ&oC3TsLIsi|2E{qajj-BMb2n8-n0ROQF2&bL_5qkJsP
z;qBs0s=s`-Ew*Ap*X6`~JB;HWpHb-Sy5v*X*X&v$Op<Bjp@$TpJd9c><{?I!_9n)l
z)AW19YyW`_V!Dt7Da4}yE2c}LQ|a~W$cp4=f;mYQ_n3joDcq%a$@V?Te@ie9xIyz+
zi&daRO>Y&=m-6!u7nzxEQwKOPkEa%_=ff5l>o9-%|A%;v8I~<qfE*qkd~lWiPlKZ2
z_%hgc_(Ji|C;;jJL^kP>X{L#wvS7@z;8h`tfQ!{+ajyFqP%(VDy$3vQ-Xbe2TTj8v
zr4x2N9A6lr!MyJ^>~ruZ1}bs2>I5*oSAYP|Hgs&by5}BXR{(Sn=RXC&f@cJdo)6IN
z9}1A7{5m5Py?Du0w@U2h=BChp@mc2A6j>q6Y;fjOyUL;{C6Uu8{1cFO0pC<Z-mtr%
zz{xW;VJyh;?8>cwnzC>CgMT7aqI@aiWs7_ZZbx|`MbJim){@VR{kq{33GB{fWnk@!
zrk!5{2~xIeQz5AS`%eJof3bDP0}iUz2J~CQ)m4~<+z1pThD;FV{}^Fjr_9JG@B;zo
zW0Hnp>2UU2xT<lnmecpHtH#!7HVH~K08jKYBL)-Lby<B8`@oEwPAZ#Tf`IKTtGEt$
z8jDxl-^kHu-;eY9Y;K{^Y_j@nVj^QNW3SE4n<uH?;yC!Cavf)|k*@}T@)i}u?TE)V
zLbY}B-Sx<ecvL@d#q%V!AT>%X0qP>(=k9?VbANb7fO@Z;&t0$ckz8o^!NPiX%5?j8
z&<9`$=Nfs8aYIxg?~4+cHv8cfQ1`GkIl+QClzJUuX@vbdbSt#v(3SAgbTbDu4_Zus
z5?YYci5cF;#A(x@C#I1S^RC?32ZSK5!sYYD?E$<flAwLT?3w5|!W0r!+~l48Cu?l<
z`=Eog|4rwO?=r&h#cZV>avT6qJqM<3mybbBU>Aquc@ci*7xfJPl=}(k6_C`}KcfEV
zi8{Wbv6>H_e0SyQyPLi7gYpvS`~v}D^6*8!5m?H>^gnBSK&>Q@kJzCVK3gumk_MD-
zhqc#FlPgoCrt9m4ZdSzZ80N1gYS?;RD$~5T@-Jp4n?KnW$1P;{YQ#dvSuWj|vO2%E
zIMJ;S8-13RPc+^vU(>EOWlSv64qEHiDi9))>lA;XV5!1#A@pj3NaM>WtcZgWuS|8T
zaK|fTbxK>p4%{$9swF#6g;yppJ&v?k@uPaK)ldqnM*3MT!R{z>24}$f@v*jvbKsfp
z5=G7S-UHq`dEF10E~}pQl|d6Uk8D5-++Rs`IC~;k3W?#2iWA^|3qBprk#!d+LQ{7;
z_O<q0$zm}T_e4*bBp`DZ#5{H|Be4NI+@HgbS$2j`PDQ6j$9xyg@&rUGH_K$t*N*Cf
z1*fH(jWw2PhrVknIT3VSRBsHbT}Ep>6nq$9oC|b=1g(z?F1p9!YAIT~F>X?Z=>QJF
zQ=dVKqD-CzBYkVc0QW1k7te36Pq;z8M6CulBPkk(=Kd`fxt1v*(^MdiqLWq&7ND3=
ztse_>GfP3~0=jRy6vC3RHncFcKW=5sDDl#V#3n=Az)AgXLLkTsZ@l|r27szyvFpLJ
zfF8b_YkKp9#BQ0kXQ6t^m^5Lk|8{mIc`r$WjfLOEWbN=+752D?%FMiMJ(4awFj8KK
zdaU&#f@+dB#^y(JGtI)?Sq;F@(^|~W%b@fS17uRK{ZD?-nhd7&UOSwIUB6kn@@WaV
z7d$^DAvWi3xp8^n@z>cC2-bTwTenl*e}Kk>fGO5R`z^qNv}rsD<^DzO(S!FMo&}DF
zh~(Im8ZSwpbSs{03)ubI;@>VcJ+<M`ZG7hLejE$Ui7m)irMFj9e1yFaR?;r<;M67o
znhS;g43^7+k@L^ka`0D&G;Z~j75^)^cB<Cg%q@1aMHolzu=QE6baY8+c?5*h|Czw8
zl^#LWaSn-^tJ!N11BAaV(I_ihHS2NP8W&?}4Kl$At<4eu??r<>2ASu_Jr-49DyHrf
zFP2^lMr(mD`l1vLRFDz@vTE(Gny0Vgs^w7bQhMs&n%VK+5&Dm)7x9MPWb6&?U8kP!
z2b5Z0M;8_OoVf#b3-}M;8(3ig>Lzln&ry#nbDS0ygCXxcLI&*j_($nLh9+w2@Vc2z
ztOHJv5k_e~y1S2M%O00ASl|)MHN;&s<<nfCU5M}axGuj<OS$<qwk{a_WB$=%(Gfpk
zAXSvrLSWtCD5%*o7XW+llijUX_puL#nVgL>)F%gAdMl^FtP6k|8hAgxSQX{F6?2sa
zUEUo44N3$W_x*8ItXiCSU~ruJ3}{lOq^4tUR)l@FhlMwnzCHaLlg*qots^ph{gXU_
z04{ka6NIbm9!|PNmAGt9c=NQILpw;FTx%9`)(Y^qBGAegNCIH7_Umg6;P2O=iP%w<
zrRKLbyE40W(7&I&hh0q!r@UF|+>W90FhV%_035m{K*{vRf?Yn51N~t2Z``#f{`cmU
z;q}a<T0(k=#zIa%>}SGMPQRA%pJhg9gh8X*^F$84Po46KrftEY(CDZkTORrxUeTi8
zDY2kQ7R4%-{BbfP62`tcUai;TBW$LdWd-`(-IRi`$0!qT)NXg2+1#oQ=U*YbDwC1z
z1NFF&ZSV>`)A|ftPcL`B41^7dIgKJMT-9dinsH~Sf;NFW3P^Qi5uD`a70_5@<#Zzg
zo?fZVP($7xGNt;t++wzru*q0ve-lfmB$kE)evW@eXt!r?<LA4O$f<1Sca`sV`{33T
z%C0=~;$^6|8Y=y-ON-R4(+E4GHt+NfZ_?KFH%PKyJ_G}mOHx(Nl)R6#jx*u*-|OF%
zrD`$BBKXp<16pyE!LZ;C#5Py_l?3_A5JkrxRLl$yydKV0HL}m1{$u1~d8Wog9P#%q
z6LSh+EXC=>_zoZ7@XcW^+8z6P(*&CF2xpMAJUCE{xB3daQGkXqnjv^_W5BIZkZ6U`
zbv48X-9!r9PMrH48>~#Ny#$~qh76<)Q*b;B`z=|b?~ePi?eN>Z8RM6?-k$TWOpPDe
zWcK{(8okuG7DW2&)xyDN1a53z+d;<*JS*_&^i9JTx-Cxaw{Gt4OX`_dq@n<9{H>wE
zM`IQp*?e))Efx(3ug%J>h6$E?7W^pF>t4f#^dcwjKHMCpO)lS%t=#K*0}&T*w^0?p
z=jWrRy8r{auwlPFSHLKw7f*9*cefLrON2w4KoZhp^)tUn+Q1KJI=9M<m*`R2b&uG!
zN;*@jwo6?Gbvg7a<;KIGr6!zA!PW3Gx7OxIOI`~Q!Z?!>WSdOGN$K-s+$GAs>T-0;
zY?i^YsM#v(H_K-f$J<deMrvozwuiw@<h{k=rfrjRvMm3Ej!ju%4w66zQNwpwwsiXg
zt{wku`x<evmEFz83Y%mXhlN*S90|tL7c#|}jmqUor;J0&rz7Rc1y@~kE}Fq3GQowq
z0$H{ROnUm{Io&p@6R1sd<e|+BIMBEK<33$x|Hj$*iJREpM#qJZJ~fUD6Mx*kRs1N@
zB!-(m_LC)jTq>YA;gNW3A?V0dGsHF9i?mghDTv~3aQ!vO*y-<*shIsJ&g_x{#GSC{
z3hFlf(5KUabHD)dz<f3r_ATGCI!Exdi@1zED7$am^z3Dl^!nOJ9)+N|Ifeyc8@{(<
z!E5@V4C~wq1tC?Dg7^=vIy(&mjez_sh3<eqBl0Bqg!I@;1M8Y^^@Wy(^3g`q`ssVo
ziVx^n{ZJ`JtKfPFvSMJt`>Y`nI^vk|xP@mlh;<5Q!Bwt`n!$ybM;VI=DbaiqPz~W~
zA#XK7(mX|vj&Y_<+D%rmI$HRw!hu<(l+JjZKwbh1`*z{p-Rh?JPz#y3lONf=@oN*z
z<0!)seaaVRapu=ErKQyv!rWfMiIZ|p(raGde8h?yH~MS0gY}%3vwHDJ&>=&Iz{7<0
z{I4rvwc`Q}6O-;L?-6<%9RdR=8Z84xahLf1@PHdk8o>PLBYBY;;Z(-mhZhZ=oEeSp
zi_72mqi6qsU6bYHc+EWU;6>X+{fvhS3I4}>VFt*nklhMO2va=Us}JK_Rocxq{6GgB
z1YA(YZCRac(lCVC#;{SKl$venBL>y>#Ao8Kxw^9CNZk>-igl^$Y}~rKR{SFcqXI+}
zX7b>=u6Xm)vvB%D*bK$vGNl$?|2T^;S8Pf@2tsF_XkigmYQ;^Qe=d+Jg~z&*7}b@}
zwtLEKbmR+kvrpsr5Gc7a>o?-C#B|bWebKxbJq#FQ58D!qxve01Qy3aib%c67Y27Ly
zIKeb9Gd0nA9D9t}vM{`}x4rFC(2#!_zD_M%volj)mETaK=il?kLoVTD3tF55*74{|
z`sq>gCGU5%RsyGYTONyiR~x14=Fa1&0k;AC%U9#l8GM8(t8|w^rzK43kwC54`QC*q
z%#8WAH@T#Jhn^=%>Ao=e&SK*7d1P!j-NF~q$H1Ig3CD>t(EaP>=U*qdf!Bk!?e~~d
zTEKI5`E$^bzYlL93Fw7+Zq%P7(|Dz!Ms87uk%Tp`!_v!WH1~K%s+qm<Z`vX_bNs?b
z0KJ3w7r1YF)7-9Yk3Gl$s&Xg`RRjmS|3%R0JR+1^`ul-*M%<fsN4VKi#8fLK%Jh>0
zXVr_@<~U1nc+k}NHb=dhm6;YloZECC=WV`{whiTkLQ_Q=1kW|#G`|dVJ?+H?4Zwg>
zvtTovIpedsQrx%b-i;i>`RM)K3t<5+ojfdo(2eh9g?L7yw+|cl77zE&WyP+D&<0dy
z0z3hCzDW9OfKX@(y#icBubqKNM)vSnQik5L;|=a#H=%%|#3nDwuz_a#4NZuq?S$zc
zMCZ?<BK2}l-ek4SGTpEE5eTCq7Pcu;EzZb9r8t6|X}ZnwVxH~XN*PnTvRcY`3uM9;
zxLtReAE??BLViCrrlj12i3KzTC_ffYy0d+p=xf;9eT6{ru0U}cibJ+2d^y|5Kjn7m
znoLl5Y3^APS!>qW>Ajynk9t{Xt_=Vw2vkA;TI}$KLgtk-UcA|t)hSa9Sbv4ZoFanX
z74PwrcaYdnV)$|M!py?#wUqoyGmeMj{FMgqp*mx)I1Lh86(;pP5^Ni9wgM?H25&zq
zUmJmkrJl61_8~T<5md^rvf2;V`<;>#l}7K@q8tkA>`ouZ#FAaUUv&qH(RkSBa@hxO
z+ZAM!FTaDxwVd>il5NBS^FZ&cbD0P9)fRE-)Q5!_F4K)gllp8s<ACS0F~`w!_tfN_
z?Y5P5{$R|P$@X&I$8GD`z3vveHpQ`Y`%PaWvCO4ee2#bq4(Bbo#^<bw<Eb!quuJv0
zvb#30fNgUNdA1?U{5R7=DFTe%XH&|&jSFcZ0dE3*K==T>4{WsjEY#M(GYEY-$KCJ3
z_3?Dq<;ynXt13D0szPHjk(ZF<VI+s}gskM=QEu^V?b+|q!<A8RV-2sb*>H#F_(w4q
zo&A>$&R@nKK(oNjB{0_S5mOQ4tI|3ia4+g#wy%mZs?7aIrT*vS?CeEr{}bT{z$A7H
zP$dA$A#}7JCDq#XL~{Suc;Vq`NdPq<R9KA|=Ay^X000^2S4PS>!z3=fpsMXR#;5hC
z+Rdh9Xu2Qp)H2;XTGK9dq9YE^t`;VMeI4t~mk}kip1{pF(RsefuV7x4Us(N)6!v2m
zS!}a5S*o~+HT_fJ{QG)!6E4a;#6T~vJFlj~zT=dBb27VEh<7*k)z1c*7|HqjN|eor
zG4jPW1zve|6kElvffo#4mB`}wQYB;%?v`Sg0Jg=}VeefZKl{wwMM(!kCjZjb`c%bR
z#HuP(KyGz<`yH&q{XT%MrX}E8fc$V!qT-;ihgg%(Y+@62IyL<gtMc@~N^i-b)KsIo
z1LL$BYk75yi*-4vd9R%bc3=kv>dojwhBus1w?Em2>8_W$ZasfH-W?2stRDQv|2EFC
z&+oD$P%$8{0*upwE~DX142M0HBaI6j;e%^ZgzWbN?1!S@eX)-gb;F-U=r1aTUA+ir
z%yRwK{QzgW<JjShu&={E&MNYKVFnJlIWo(!y|*KSeu|};$~c3z5;0z!R{SM6M><u1
zyhLKd18o1pWk~sk)9mhF@TJ>`rlS9l{Z)np->gLKH1OB}kS`9fsHy+XJ2L<H@|phO
zag5!OT5ros+~~j!hy*6<N8Z?A5Nq9jSGmbk<H6;sTEAgQPm2JYW|t$)xAwg!hE2PJ
zF?X1mv%Z8|>zwTcSE10=t)Y_@qnexNMxuw024oC97YGmf=Ln5&N1~v$d=xTTpDEu2
z%$wdgUOND+746sp>PgW=E`tbk^kL@RN!}}UL4f3cSmfv@kJqFg{yA#W&}TP|!hCyU
zP*SqPcfhC=^d*U_JH?&KsmDat_*NW;>eYlKeX^7qmX08~pEEJF*EmwfzAg^l(F?A4
zLS&)nqm?qARj^eS;$(B4M=OThpgh|K4+!-Z=|*xYC0F6q6zdnp+Z2(q-)_76&1UU(
z+~&{1N3Fw{)|g(XPW9zpHY!F%Wlt`fTWh7Uv11m3<-=TxtW$^A;)w7B%+%l&$@Dq`
zse_>i)c$oRUKuQqz+5|3rBH_IQH5^Sxc!f3x(12LSO+A5w4^gQKGy=YQw`@q)Ov5k
zrprByY>kwaVKNRwzBOWLUFt?F2o%y_CBOV|Oy|1&-x*n6!c*ZFH_nL~5dwVZ#T)JX
z7QZNmOGlmxXH9ZtMhgJwkhiJ3!+<SPZxnMfJLbC%eg0J*hTmJX4(OrX)~`1bJ6s^;
z)WY9AWY#qN%<BMK@vmY6|1CLr^X68H*JPT{huGd`Y`lH8E$$8m59N{LSPj$le`USl
z*FLl0PsI0LH|Cz*M3yE6)(Yp#NTAFt@QX(0Jk~PxFNaBS@@D}|Gd)Ktf9E}0r%szX
zm%lqp?lE^SjOMOdv>U86r@QGl8B6mDiSAvNkr%ChHllotTCT=RoI1())R~%B5%LEs
zwD-?|$knvOTfm+tpaY$T+}2j-;S$$V(w|t;T2FvuGWRJ8K=i1|xpi_^rx^Paznk3o
zjt@Z*XYRsWB?aY({IarDZ6XDtcw>1ylu46P;P^<vbTMjYE4#C=C7Zp(W>a(R&nMO@
znd+*eB7CV^ic`(4Gby#=)VF|&<In9&k0hj?rVWZV$&8w=S}vB~6b0DdHvKHLFktMo
zaD_$;{(L<hNd&Keeg)RqPV~h`KMo|REh6x=zTlO8@b@(U@5|{f5az_7#l~MRS2Spv
zQ?BBt^(-(`Ua8JCXh@k6n(;QpD=&Vzt>g@uVCt?Pv3n;i(Co*0S<RymVfQ`qogOsT
z94#(Dujv2!6`LH1rD^wC>)~76R^5#VP=7i*mRun6W>MX?xhARbKa%OFV$vmnfNil{
z{2Q%r)^EP^ey9VuR2E5fx@09P(zwZ|3I!~MLSqy02$v9K%f)#|EEI4FEF7f8gg;yT
z6RV*U_p9rw=X2w99ChpBdkU^;C;DdJqsLBi9b-39r$tRM%eyi6k9KdVfu*dXbfgnV
zeDHDZr*d}bb-9PJPE?c7>cBdK-d$vX_rDJ)!+MOWaJZ+}$y=8|kM)tC$`i&Lo%eyE
ztaX^{Q*j}^Xt_AR$9ii)%<X^p<q%kFs;a6L`F?6b7s0D9+|AE`oJ;O?On3KF?MgRR
za!$iwM)@R@&Vq66mP2E^DNo!)E~97-wj3-)B(pBA1O~Z-8p`0w+LWBJW<JV9PtOso
z#*q7IIX_-Vk7F2GXKjn6lX|8L)peI159uk6es6p3cwR_srpRVr*`iI4BD1vTMQzP0
zesv#pBZ<Sct5Rd^c*xetpFzG2pmX2&k5Gwl233L5+2k_-^Dz_&kWL^I2YTo#aH08h
zTEd!3Pk+MjcJ~NVk5F&*nTU4N1SzCaCQ`s_qI{<}c%$t+&@M}*2RpP{!RS7Es4G{t
z=H20iWP``9Ro{lo2%8C$mP(X-k@I0y0;z5P9DKT#Lzo<r`2XDl|FavCKdcT^<-gy2
z$2;seH)-g4L_~=FG1>BEoF`Y{w21-?w{Fz;Fi`3t)ejDK>tcD^4HH#Gkv$t<{5+30
zOAosB10Ta)1e=kg#)zk}JhSqipS$mN52t2yQD4)1Oe*N?V%(19mdjoGU52vOj`TCY
zaoY8}|0tGm`e?vy{_L0I!RBB<dmrPr$~H0yj-9WgrXJHZ54%BG^BGnj{&L<0{9t5y
zzw<#7{Lo|B65lm;=vbN(JYWORRHQuxsLyjwzv<rY^!#u>0PgI~BP7sIiW~SfrL5L%
zQ5-F`F_mD$-d557^b<{zkBLISqE@%6k>(8D<lzO<!8Z9q@l3YZTI|hz-fv|_9XU?N
zl@KYrdwN)t7hfaR{(9Qjm1R@y=^@PQM+)O~ZyOznO@3fa(Tl^X##t*=BTx(@5v7#5
z1G<ORfoXE(^zA|j6dByT9Tf*Z7QBEWqOi2d9_fL9=@P-DbhrEZ?oRbvsO6{##oHxg
zPhyAGW`uP0X#X~u%}R1!Q=Xr>yPbUa;C7N+^Yr>xlze&UD`@%QshR|ey|ezBBXriH
z*8BfH!T+;*=mTx>Aq$ntMujE*{q827<njICr;eb-P&uXr5Mh{FSn>>9hi;AWciN^T
zS5nu_%)=L=0ndkYoOzV=D<2MPmttCSi^<_RbsQQ``rB>UcXrI-IC;!I0UXgfjB}0b
z1W*pz5o0{zg~3~M$E7WWc#hs(*cUr^!yO81gE9A%nLbBc9D0r9kJYN6C)7`<Q+G@p
zjo1xqm!xh^PN=$?_OJI_Xo1{DAq&j^n3HwGe#NHO0YD;%6OhIb!kStU1X!q&=B0n2
z9D2KgKaeR#TU1oE_2FjeyhYYBNx$m-C{F+etVrhz@Ln)E)lHr-+pQ<iF0&z1<L5uP
zZG>=-x5M>$#v#Qr(N$R}hEf?9R7DXa1Zg&2P1vb1JFgf+Kw<#6fP9~#3@=nQD*GV7
zInh@Y%|jJYrIg-8T_C0^(NQ`5E4EU<VOWlIVmTid#YQ$kWQqh)B^k-%R(yX-Rf}ow
zyo`dTlR*TMB)L*0nQYa}sKqCg5cKT4_HF0XM6h8Zd4#*U4riat?ul6u_EMx)>$gn%
zz5VUk`RS>Xi;O|NjH&9=%_bSRqF}IHe)gmqq;w5%3^Y)YNkm({>1T1sXn)xFT*9ZA
z`2Swo|GBUSUaP|7u*4U+eP`XPPPNjU>3r{R13hxd0(-{#$}`H4BmOiZm#>5EbkB9w
z!{RAQI(olMF2;Uh^qfGrdk&RHS@oW*3_VCabTgt7IIUi68)8~@d7wK}Sfqtla2lRa
zhq^pK6cv&6veT0O{#!>#`)<vq@&4a2&*(+#sK<(=@*O$*dP!>|+y;I*x%57lkeSo)
zBhHZDT<tmca2r~;+5*sTR3Hvbr!Zi8o>}tm*(&iVkTEKHca+p(EPz{%78p1*9?GC}
zDr6w#^v;jyl|*lTcU=745c`rjf5-<`<9~a0LE`-DPxtau<jV%kUO2S(8%E{Ve7_#G
z%~F{JcdTi-;HVI-^d3sadK)~L01s;^Ogn0qfz(LSOVhf_B)CY=u?9J-lfbh5xwt+(
z&dHwMZpv^WLgfSh>0$O1+%B!auxfm{<lBfr6uzEHupPKVCK#pQ@0?}F(pE2Fmmoxi
zS5ZF{f%00A<w}y~ikB47N~0-EaxE}agS*x>ZmCD96Hefkw*A)7c^>BFc6wc=v4zvH
zD4x@^{>UF_wkG4Fggx%;uYqDT{%7Pba?qk^Fq0+jg{`!vRwO5Ldm<Zmq$%Z>m%<1h
zs(DTGz8$k6LO{Xsf1YW$MMQad=~2!7AwE_c73eQmt0?m%9maa}?=wpwmzF4GGMC)B
zN<=vDWfI2f(H{TlTemhBx3!UfJGMB)ke8ZaYgGh4Az(9?fB&#_D2SFG-f###tUqvM
z){hPlvRUsbFY#ZsytrByGwEC&x00J^rTqwI&wlL1NH6BxM&<0(vFseP0(f?^8J^c4
z_E(dv0o*BUz}nc+nCYNb*`Ijdvs;)a-fOzw@KeCuMnD&Uf9Hhs0P?}?QENtXrg+HM
zgdAI0m||+pqE<`O<5`cL_Hr`zGVehl4&8DoNG7%zZEp8W<rLuYvokQz54hocA{f=x
z^l|+Bk20})O)GV0>!f7Vw5YLQ3>J@RyKYi|A~Q@Ek}Vj=tM0rGAF&#&r}TGAMdZ@v
z5QN?4|JCEH(iMbB!4_f|f6b6iv9JoTO}f+PUX@`@mnv0pE5#vCeYr{yaO5*vMx9jp
z)PDNZaIg-MO^&0uV3~jj?ACI`+}ub=bRv6j>jNw-@tSD^``@tTjQ5pyjeAX{=v=kW
z{)EVE&F=Q*?jFC3-AE%~BJIVA1nWT~IFVf2XB$ktS3gMME9EGauU09jW-(Skx${3~
z_y1Yo*fKNj@pt*32TTD2gfAR=EIW?2N)Fmy%CUpA5%A>*qA2<0>~@<!Z06bJ;VBaQ
zhM2Zn<RJIQ8Uge}0y)<zsdA=sS^PGHeREXhB}atQPK3FU0saL~LV<I>ASouT_W_hC
zbmyXx(tmfwdGq0(M7s%IRKL3ho?dd8ZIpUs??+wRWKA;ff04XG)89#96bZzV|EQ~b
zK00V?bgk=qR*~7Mu!PpWwTC|bGs`ZzI3Uq=`kl~EE@yvYVj`=s@bR>r3^Vh!=)+NK
zc)8%oDdD%GLYxO5pTBS(IMlB8X7G)(ih8-d@lnpOYBRt1x6Kg$=U-ZoFYXA_C@|Hd
z>JD&cy+k3GEcjP}{flXSQe(5(P*Z#Gm`r|Y@rBJAOlv)A#L@r)J2lHzIj{cd?(XB}
zBhFY)?(0rF1J>s37b-7wBsP`~A&57Pdb(Xr(631Q${N;NCId%`O2-av06%o`y#?LE
zwi|UXTm&GnnM-<6+1|357=8&9nYldz#Tu_*B7$=h64scPy$}?17i3U$eJ)mOd%P}t
zzCSnjZfYTuFW4?!@P&L%Pa-jV&Z}7}ozwN&HB`rAZ_lIFJgFS7?Emk`h78Q0-3m}C
zlj%ODJon?ndhyy#JL91F$5_Wd2O??Mi?`r9@vjACwtJ|`{@T+~gw!yukib`v^lgwU
z8@&MG;u$C6=CgsqvaAm@xvA?4Vuo%D=Z7td%4Ty2cPROojnk_iuj+CXWgVV=Q9GUT
zKh)#l7W6;A0kX@xYMM_u6_^25%V_1k>+7_DyUSXj-+dMT78p+<ReUekk_x5%>=koy
zig>?o<NnrENTtZR`|sbL^nL5ghkVcByn>AOxi;YBX~~y(<yWVwgFnFI%cRq_&!{Ni
z240%1Inum^iZ+~dgSz5Xa=K)fn)}N3YtR}9XC2flZpS8jMlK}^VfAIinz|0=piwEd
z@gez^aiDoj<Y}v1k*?vXBD4@S`&uWt|Mq!qHyaw5m0_)*|0M&daHg9m{(>qn*G)fr
zbm1?KSBiLMI|{FQX$#gta;2&V+bN9%nMbahOELEV_eJds+obf5{l|+w?`MWTco3dA
z?AuuT)%z{n*>n1w?>SmKZ1H9lq`nkRB-KfMxSiYArEamJ-gvJzVwZn1<#}l}0^h9{
zj!Bxmj>)9_KaVE>gbm0gMCJ57GQ66Xl}@zQ1r#AP1iW>FxWVyxHVfmQn;4F_j{fpm
zTPs-(BvRO}%y!SVGEJo91iN}U5%t&*vu1a(ArkXr@3!JuQ3kb-#AdH~_5j44JL3kk
z(RlFHS1yNzmj8>cZ;xie|NrlzqM|I9A`+Y1LMf~0@>XtJD3^^536sk<B6GjIlM-^-
z+>_)Uo7>pjDa71!8zV|c?zvWU`MrF;=XX2bzW>c}jz7HidOe@d$MbRFbM>E3bd>z^
zx;a}s?zm_@$u3MiK`C(%ap$Jlk-D<Ez#7u7JjQt-OD#juOkk}7d-u$LAvC%jFpI2?
zKA=)_PbZAA-B8u7#20<ciZ#6E<#k>)xeOu+yU408I#C<&%W-K|)576AptYJ-(Ni`Q
zsdAs3O3nyk;^T$zsd(;b80PA#j3y_CC6Zev5?0#9fMDW}i)YtA-c^9~Fi<C5A6H{V
zf8xQ-$UVRtM;ADX->+Wt&Q8rcKat?6?yYvKwM$wuV$Ipaj=V_|zRi@HO98bxZorl<
zayP=nuCdA7LU|<AsZ=(MlT){M&*W+y^-h6QO8Z?(>)n_mMwrR%H#o5jtF&CTi<orW
z;x6~atdu*k>p+acq()!BFF#+LMcQz^-?zR<Fw)G;%zrtrS&Ce}Eg@<O0J7glr?W$b
zjHj}-hKzd&HSal*JTUBE9$R2EbHx5eR$KTqg#Miy8m)Kg`L|iASie1xF|nMss6+6E
z*-71laNl9xhKz*;*qg@RbVYHiy_RLQadNRw7*ks(DsAQSv><+)8vQE~fFf|iOyWU=
zndF0q0FE*(wtIQRddTPMmF2b3fAS^wbzA2vPf0qQq7us@LvuV1MOjwtx@mcjfc5c#
zYOytdG-w&5!rA0k0<({=Z&d=_dbdW$K7<`u`UO$84%$`c(HlRpI1Msonw@e}bEsPI
z&TR$RpUt8_4V&w8JzzDcuO3-r{f}$VRe>_kS4ccpAJc?Ho>L0hmH#~rV+>sR+Dbk}
z>co)kF}YV*ks^1jj-i{(;LXjz4B@qGdrs5vZ*rx?LyFVDyLI>8#p|PYj+M&{JX@Dv
z<{(Seij)-9Fj)h)0-3SMn0QrKx4JPs&&-`<8OJ|`tBpj(Gk@i;>m@}(?APPkr_m@n
zDk}aQI7UJuupEpgOV*`@^6?44hQ2Zf)^cxa7*rN6AJ*G{28aR&M@K$XGYcwz<ZT2*
zj%bK|J6~R2zVYMHwK3m5*yND0J0TBTY@xYmBdUZ!`20QTg4gH%ew?{l*yR2EDzz6b
z{)taa6U5(N(rWO#qRkC6nq4gQ@~reNAHA2&D9}qouAUa41w}RT8~E}a+`L%E@3HgI
zC6Nw`r++T}Hm>>o(ZNR5v1@HS>etQj#DbA~t|Q88_`SvcW7j%vavI7#N4|$zhh*JB
z@Oymfo~{*eP|<LF|LaMO?_5E-t)kC|t+CLcUxPuZYk~D2zVo>FOd77Ou|75p=#HyQ
z*X=z(WoC|z)rqvWKYkLnZQHU~x8-6=B^1?)SpS+eEAf~yE@QD?O%5%NUGea2?|&Gz
zDCzM{K)%wrxgexR3t-0Zr;3`G4~9iqQiR^y?Ys<|a5~DFsZijVlHOW!o3cZ^wa_K?
zSp<tPX$~KFJi-6gOk)s0drk`o-w&If!#tz5Tm<Ct+vDqz9g5e=F8#**3#h`<<(pBL
z;PyBzPjEl&&9dS{%*(rSS{{cnSg8q)Q)x@0AE)o@gs5fE&xRdOY5yFjmGbR=en6Zw
zLEF5V=w4}nD*U+TpOZWBf>K8&t9u_l^z#FEbgVw0CDzYtaY*CcU`mfgh|0LzF=pIx
z3&W>Mz>ITN<~Bm`BJ1>1kofvB0b+ln7HtFs=Kr@H`ahq}TE|7UG883Lc!Pg7lv|Fx
z`}*m?*zGUk{^x`5{Eid26LltVytD=Zzcz=@VhfCb4V02h-t*lR#8I~j?<t!K;sE7>
z?;H1cXALne{gL-Tr8nX^vHarhPZn8&#a!C!EMA=Da^<re!8ABGB};>^@%%6Ep^^JL
zmNb4F|C~D`8daUPxajiSf2DPZNf4Dy86L*@|C~6{VllRofLz)V_mga<{oYVIcgj49
zxhi7kw<TG#6w<C6U|mskeFFyAU=vrrc6rquiT{)h@@-TCXKo~cK)H?cYU$<WCkj%n
z99`y2sq_D+KeSXXdbDy&XrMhfwVLtjGAAO|#*UAnqbKp~=PYZ$iAiD4qUiQi8}w$`
z)3Xn>7!_7Vw)8SG)qx~NH+V{JIRa~`YA96oY}{a_zW0W`YJ(c*Lo7g_+=ma41KW)-
zHNKMNEUZ$CP6NRRN0c9JZr6UPZ4Sc01<Ch)l-`8Pr*u9Mg3pv3t35^LK7TR2>TGcm
z^_1DJlSJ0SrhnX7H+|BYy}q&WZNbr%#O-HzEwQ?GcPcIH6Yvl{tQSn|6~x~_(lDM9
z1owG8ccKEH=Y=z#6+&F&JjM~jSha6z+JC>hXSK1+Bc!_4xWh)6Hq5Uj2bBYxeM|ow
ze!lPJ-L`#3Q~E3^aZ!8iA8>VJ+4j@q!zEp^r48LS<27$Dl-Vc<IGG1Wm>qw7Hi7h7
za5a%O*}kw_qN2W35IzeaLiR*1=#ngg_IV0;q~Cxz?>yJ9@_J?X2Pz6qu9i(<&F>I!
zHV^I;t#i8e3s?-Ky|MVPiP4t@rp-;$SH4}W*_OlOS0{d@-21irY7=mxOc+GS^OeY)
zfATiZ_mbPVc!@mP%yY!J=qe^I00b<%0n5pknVGx$I=9n|z>>YxF{ysX&C9iWQZEN6
zB3ZSi7D^HNPE$kp-tZFKQD*r`Wp=`in8;o`iE)o7Qt%Ni1!f`7Tn@X6kuTzq2}<WP
zrBpHT{Uzhy*Y5WgwlDOTzgd)dkcG-rg01UpC74CsGDka=6wc3r)e{as&~Bvh;YmT%
zPcc+!Nt1c5Az!if6z=uF{ZuHN6c2?%h=xr=&qeSsO-!fu^UnM%EF~kr8p@T|W|=22
zGl(yH?a-9zlhb?j_Eo-k^JcBCZev3UkhreYdTp_+tgKQmZv^z{7ru6W)a%|bXMO~K
zqqRf^)|>#b)s<;Y00zwGqg4)JfvDqe4b<HTd78*nkmg~bgrK?0b5wD&dgxKTdXd(q
z9+N9i6yK&~6p$j1OCTwerjWU;d5``~v9GIN+HX9w+bjS56)hz?a#e)3889PxFlXxl
zzOCt^2;fuN8Rd(tnz{o-$|suqU3xK`E$_al*DaLiI!|gedqGyZ?e{hBs7h^Z?S-j0
zlVZ{cLVqs^IRF{i%%bs%iKs^WYN0`WV@ulRZjo+_r@~Brwz$GYaiMn<gNfhwtsfo{
zDNwE?<qb+&-0d?5IgP}hQz}7kJlwaLfZ`v%yFv-hUnfmcdU`owoG9gkcvbdv0PLQu
zf${7+Q*#GMvS}3J{!ic{>S|#WOm`+g(FN&ZHUsfaGCgX_<B#)-bJC=JIMBMsd5jgy
zD89<K3MG6>P5EnFc?B;$DlzG$#i{QAh(JawB~1?9vQ#?RK5P#wa4s1rXAtB&bn+Pa
z9eL3q*c1q1M0}YgPr{TR$wR*w@rjpCIS*d;7UwgF#8yWqL?U_}08&)?WxV|fK<Q*k
zZPSw3dnG4-#7G=HJ~TQuyyD$gXmyi$UQOc5!u))9o51=FO2EbNi2Nzp^GXQio+wO=
zMC4Ho43aapH%=iUafIw8sS>?^o0R?E4?BiWDq-A1W{l2TwPp5m()!PH^Zt97m~RSg
zT`@vjs+y+$%ro0E!Kw6<*i6|Xmpa9E5&!v{-@sAEs+qIyJ)(|!%#upjxVC(;_OIEb
z+tkx<SNCh}I_@J*38UzJwAhkNiIJ}chEg>k=cNVccii>k25Ua6eP7GH8u?>&+Ky2<
z)sr^S4oI8>5rZ~?`Lc2NsWI9*l|8w4`CC^rHDmrBmEI1Nofiu-=C_C4n*pDJykY)L
z0%@dYbR@d)#g(-~0&Q?n<5sA%acjCU9by~FR*V%iuvgxtKySjKVwlvq!$6^D2KwR*
zX26qo&QwRxHyd~=iQr|DSr;yI41yvdPVr;n@fOEKbQoly;G0BMDn87eG!+alc&(U0
zi?`Mek$ARW!}J+?#Jq<S_o!hm1`L8C@Yd2jS|ngOqyMQVr5_tB3nyt&g^wXzC(To(
zX;VYbkqKZUfMo1hNXx%&x}MH?dVw~Lp?1=Exc))vsbe1odn%cwB~CXcx+FAGQc{K^
zFFZSUq>T;UW+hld0xCLErDbZ+n2*sIRuT(qZMLTB`iA|-v-1Dl18rAe{Dy2>f?m7<
z$my~ES(k$8zxt=cVt;Bt5zoD$vV&RslMtF5j*oQ4?75CSUF#Qt?taIC1E0EEuWVnA
z)}`Z*M%8YycRq-&Pu=o8t`_uzM~A;=EnDU7x4Lh&Zlr3LQFVVnp?zy`el(;tM<eZx
z0od}Toxk#c|3HGw6fFMSm$5!Z8Ldb{MdQ1W|5hBWamBspe0H05jsJW}%QjnICbsoX
ze&tUevyawS){4G5&whQ~7!BBGu*_#AU!a@{DcV7dQt9KVO?lP~&3K6eYnEm|9=Z65
zR$70HN`^RCUF0lYKT!0xwZu0~`n?#qK{K52(!yKfkupL=gP>q`ejODR8Rx8pq39Q!
zwuosOl6s*?2*JgQSj6z)WL^s(qsYBBBn>N3WGTW3^LqwSr7v)1pjiblHt8}=RT;$F
zZQ3!yVVrOAc%a;s)kNTW%12i}%Jt9bAkJQ>$j!|DvHYvr6S(7+5|)60y86;q-DP?^
z@2gB?jD#UoDPGFLx$^mUb<9Q2bHGp5{y-ldgnIm+=IDQa(pv|<j`N?`43vNC<=Wib
z-#!jM`2xoeLm}-SZCX7d9T2L*!Slt;9O<NNAQAR0rPA5tLatK73%y|b^0SemxwFNk
zgsWF-rfz=!d{#!&<I;Q25z?2LnK#=jGb2|MVEb!UJB#)PT#JVf4_!HG^2R=BOXRof
zACn8-Tpbt+W$W;BEFlB7e&F8CEoA^Vnj4#3ofX2urNf{2fW!K<(yr1X;r0GO-sGdf
z+o6LszcFR&2W{tAa_`Q^aDqJ)!{Vrh*=Q(R5y1I#>hW~vsNHZ@6{(<DG4lGGfh+*s
z;fzYiZS;7H%)o0U11LJHgCoc6pKU8kAHu0>B4(|5xZYHvYF?ksC`npyuq-s|LW&3G
z!=Q0~tPot9ZWtR$wFR9L6(0eyCsp#YDDvW*>R@``SX+fU3^80DB0l1ts!W@fUI8G4
zJ}Rn{3&p0&(s+PGTAQk-(5k43(2qvDF=k`T_^SA%OxTAuHa190L{dPNo%IB3BPS;(
zAz`<KViGR5CI8k(b6a}#d|Q#rGh&_)_q%W2x#4jLz6JhPK(1Y&{L-Yt6@7`noFlL<
zIB&$#%BRndSg$RLPY+nF&<Qb}^}?@9b4aI4pP={{%2as`F0!a+QMP_MOpk#;BF2*6
z??%G*siR?@=e|t$BzAwCi5PqOaN3~b*W&zqlla)zy+6Ubh&Hyq2;B4Ujgkk)6D(Zb
z9CLX7An()X(7<5Vg_=X5r*LO8resy8x(a3ee_&M_<^+5=+upcjtYLOp%b!Oo{yVpK
zG-qbK`GA^<r@MM(PsMol`OM7RLw?`X6m3j53II8?x5z5}bUs4u8cp&3tQB|;L&(|a
zRKy&LfuewiF&i3ZP5$G{mF>ZC(VX-qPKGJgL}g!*h#ga&_M*iwz&w|S`zg*mK-?q1
z+~LK+^n-JAt-==MV7ir86V{rCb1exX&PsHLl{QbB%fY}#+j*J8TWa%+PCAl3H2k~l
zqci??ECifdEEK@G$hZMMhR^rPOVAydT&M*S=owb7{^;Jl!gt6=h5cD|Cg%l*)v`ex
z-TYd(VBuxo`|z0V({&VYUbZoQ0<2<yFEP|YnSwa#0DkaYzn2(ur$xH8>gJkzdp0Lo
z_U~Ozn!I?dy*5<w32*7j^ofgH!mA*Qn0D*kcixl<I0~Q`jT_h;V;&8^&acUIc9dC+
z_Aw+vx#8J!gIHOtm>|EG*NNADQPn&rE`FBQ-zy#Qd)BPJfm2}7Gc@!{%MclTdE4!O
zyZd0YE2kK1A}qeT4Sc8%{c!F|)I-rCHK&Y5nQz^74_uzt5qJM~@1miT18NoITC1j3
zj4ArECHCzi&q0(EP_Vx!8;Q-&dn2fOBI1Bb<?(LcjpU>H4n16MHni8Ytt~@Vfg`MI
z#m6~bPdO(ZmhUoh%Y{kOc1Eib<JNG{ensp6I}Z`SK{2BA341uFVc>eG*||~e3U&}$
zGK}$&ctWhx!tbni@s6=M`cI8~0>8^%M(9WI>tCZ&ZintKdi#myd_>fyadId<xZqT%
zJKZTWoBPTV*3!m6RWRo=s`fh5EByt2Sl;F<k#753iq%c%Jch(j@ytnDtX+X9ZOf$F
zYt`HS;kESKdaoOPgjG8wzYXGyqqF)u?ifcLtT)W4kr8yaJaR=Dh~f?=RyGZ}Uz_)a
z;EimpbQeVi`O9xRSN2yaXUYHlBrD~`bJ~=DunCF%sX@I5w(Fjzs#_O0#>c8hjPXkD
zAx}<;-ZUu?(Zr<ORi~(4y3p<+Uw#!N@>IYRbrdQ8v}Z4wF33Fu%a3zvYU9f-C(PPg
zP6S;7VH$Tp3e;hfPHfO`#oxm<!l9>b=Pq7MVO>0>`LF-2fBNnm`t3PJaNF$V*SXg8
zgi41<Jlw%#GM#^cMB$e0C{Ml*Q9lld*#2lkU-(`ub8k4=Vf?F1+5->Sh&K;*W*H-)
zRGk<#Bz<G5E$8j5hgt0FS-nymRn7ix^{qSn_BdT4xz8HCLW>=qWwS8<Dj7Z?Kf;+T
zW^o?((uD0htIBRR8{1tE8<e=;`=so11v8AUco`E$3#4@9H(}*YBB(Z_WG8rV0lY>1
z7brRp)JBp#DN^D$Vwf`}D;;DLeEcN>@MZ8lP$y`!s(2ZZ60ZdY&o)cBj?d%M_cuLx
z-2Hvu0yvrX-!$GBO#sW^J+$WK>Egm=Zh`KLMVwn)oIRr=muDo)7D$9>JU>o2;s91b
znMlgpqgvP|)-i){x?gUj>F0*O<3YxkGl*fP*kzmj?RCqA4~@a7q0FAP-g8Ikv;0|@
zrl;fCN@Df-2_f$C#$6Hu$reI#5+Xg0>?SlF!@5^EbdS|x-g~YLc(11C+OecF=H3SV
za(xe{&Nry#6UzXgg7orG8&8X_=Q-iqtZmaatrtw!zO4wo_zn2>9s!BSHEtgSC|Zd{
z=8t5eFN)}%@RAm|IL~{^ie(uV`@`R`HQnafYzz-ph7t~?8Q+=mRItO`V4(7uj65MZ
z198_UQUn-RFC?<pcAc@P0dQH5LA)yJUnNwswwofothAwU)Qlk?$uX#lTxfiH-6um@
z9X^hhV=H1gNKc7p2MI^)?RL?zB2MVsTHH7$_x149tq_1+gWuEJS>Ssx!J8<U&?=a}
zF+ELbpQdzw_VA@c@TKflI~Q$kO_5~Kv0P8PNGU?po8XJG`)~gXlz^@l7OU5{yp{qP
z0YMYLe-{=$pqsNkA3nS49##bS(eT>Mk}2=%3J(p-VB$qS&oVy7(e(EP(JG0!XjilS
zzyRRCH7%gML{WHZP>`coR?P3WAH+m|Pw*ewXBBXB_hA`V#&unXbA2zQG@4VLD|=Ev
z2^i^8EAo$(5+75HmsS^f!#Yhr1uyrq8I`VC4g2tMliXs19`)})s52gVkNpV{4|3&3
zT~DMBLgw_S^w>5yJHXQ7{D~mqNH#~CQx=vcgJd9D&n4@B0?D^;!9vsd=ODoDl3ZIF
zv?JVxD7q6L8>5;YKRoniGcxog;C<m$kHN^YCIAS#ws6aclhsiLDR7dD_K<(`n6HR(
z$s$e~{)Y<V!7>YDon<I$3Hu3Dn?^hr;#3@lzt#$Lsl+5j@kkZrMa2>f?8B_CVXt+3
zzgT!NMT+*hu4q`^X@Fg9tQ?DSBk#e1NM+@grr`|tRc~6$W&LO}1Wy{K=U*QJ-q#Z)
zc?=&x{3=y?4$8g@)>b0+*|+h-dQ`O**Ow>eg7;;OJM2}R`SmcMw%Yd@4sm|reF@X;
z!wV4^hHU4S=woE`c4L_cc+<@Sa_lG!RU*&uGP;FwTs%5dIq{UZwPfK_>i@?G``;&*
z7p298<T3V8>@dG)?Q~_5D+C`k+!*%>wTn7!VuH7tlovc>s=dR^lp+!MeeAFu+*o5h
zy>QTK!RwM{WusjCtU$JlgX@7V>j0tO-=C+PNr{OY?OXb_Uy{g;YD^?))k!T%HbYx+
zP^Q<?9ns&OU4(nhD-S<Tug)5($h9^nJ_0e8<F6m=uJ{xV{|XTE$XP@`XTWBLA$(zF
zXyV@cu-rR9f}3<cs1D+)y-Qd|^@X%}q4*Ee+=HU+1(@gjBSUfKX0Hrn$$mekg%1=#
z3<z%rv*I1?_=;WpRETD$dKH~GZN+!}YUy+Ml<i5Na?u;2m$wWmNeV9*o_Swk63}_=
z7;|)3iOzF0`T8TX`lmZApl#}88(X?T1bU0t_7kyRS;(ogVt!frPyhE-=|&(ccPCoL
z3139yQtwV(i2I#sXEtemDDYag{uY<Ws-%@pw0^GjlXP_=z8@Lk_9RLk?>MH}ACPy<
zFp3POv}?r5jSc%Y*0D$-4JtI^jBHy}*`+)M#^)b~$iLD-Ota-_mSs}BKYyrH$`5><
zolbp3{JpN{-Mz1P2fCvym+gDx>F@PRSsx>DajCDQEl|F~HeJFd<dk$+)%v@iCSF<-
zg80xx`x_o#e1=B2PySq(68l--*xf|@*z@9U&Fr|w*IN!U0#?P>kYO_8{HyBHEuB$y
zi1!uCQ2%e(!AFq0l|ooAhW(0r&%1a3f{cP|u=X*p4*`u)4nXR9;nNp8x3A;6wi72q
zZGc(?sFu(S`@`SWHHY+yvX9``d&Fhy6$J<23Z2eK4lh#&6S>)=z-Yba7dz%`d^_`Z
zW*NZF44$j~8qvb)x|3t?MzrbIyD;dIDe?DZG|LbFllx{<Rucm(&)S_8&|$?TA?8-o
zqS=kJLfmv@iMl;!&7NmSO8Fh`B!;w<)%t9~)Q@R(=`yChcGH{!wX|+8k$(G`><k8l
zQkQVR0J#uIo^LxN%P8K5H)0=}?Dg}ju3~wfS92we3|8c>zy8hAN5mowj&vR90;!gT
z2gTD4$*~hGlwnO#2u+d97<-b0j{J#8ssG%o;J@E1LnGGnQ`KvIt~iSNa<QmPte^0u
zB~#OrrC<nrrU*}rIdMB7^^_(?+!hxlUZ`tEk;rdP_!7jBaT)kUJw-VM$E1di`R|pO
zx&7YjO#xs?t9Rh9Fn_svPv?9%xisyTM8Ss!(Hmcl*@*2Jtvn|JC6r8LluN~<6evdO
zQF1K_R6N<;Q8O$R&|$Ui-#>e0+k`OX!%x{>r)5!@S@F+yomJ}>ezNWE^_R4O&-_V)
z?_Or#vfkv?4<^PPq{UiB?BS-X(gh8!V-ds6IhFZBgT)`x^NK0`&V7!gV7g2#?j&ao
z9fz}kvY*-OhNEW~HoZi*aC(8s6!p1tuNh3CLJG7w7VDN_#Sa86NU6e(*98#}Tjm7~
z%Wz_tyG-5el-|?$7F;Vczwp9k1Qkz8RWO4T>M?wjT5+0)X<i138Xt#6QiRPwA#5{B
zxVgiJhf75Cy&oK?W!eW##O~XeAL|+Aa8f((uE3&ZE7%}hQ^Z2q0CDu8l*TYv0bi6&
zk`h;mNVE&F7yWy^%98H5V8j~V;{cnkpBgrybd81JKjC%|j9~gvSJayW<*@q!(W`oK
zktkhVI<kq18VD<P`GG(EgGx7(#<GqI3Vjpsu+g^K?RonrqT@!+>Ao)=*~`yDgC3{N
z`|#cnOR)7J?Othwwm$Z<s~kMkE$p<u(pBf_ztIF=&XMW*=SqqRBWKa})zP^GZvU<1
zh%c3Q-t1cY_HAN&H)!QGn*BL4GE)9^BdA&}<GLpv+QTt6zk{{)2D0Mj^Guw`;Y9hj
z>U<^XPiOld@fD(SBJt>po1-CIJ)D{XQ?X=Ve}&P0y5fN&w`0>MpXuwY(8>lFG?`D$
zd$8yPk0xASC<RYq6u9K}*)3eqI^DTdhgw5Z^N|t!Q$xiaf(`}8=U9;uMW~E{nIfD0
zB>8r+H)<>cBWL0zG$uiiuY~FkZ^3z=dL_zx8DY&cWLb9siGXL&pOj$Ha-Yh}Od7~Z
zQgbKwUA-F9d2VCP)0M)*jcRTNZo=Xg2ElZLXy(&6NsVVhT#!7UX#fnV-iqkxfrc<a
zH&#mH{XV|^fA29C!x>X}F6>0muijyk=sKyR9b+M;pHOax&$KV!t(2OR)_vN$NBiKp
zKg|(l_&Z@c&iesglv#q(BX%^XpKd+RGMBfe-#<-mN*nf?c>C?o;m%(Y7aV(>q1vyG
ze<s_MDKOHZv}e9UYVYzO<UfIL=5}8Fr@0#dx{a-jMQCe!!_Ob9XI#fFG2eT<Jf`8f
z@TRMF>)qHu#4ZT$r3<!rH}&}#AuMQX!i&?*aq3tE?xs|7f3LTEZa$w5F0#D<5r$AS
z8wp(@Ld$V8YC`x%5*6pKW?TRvbtW|~M_eE@OGjrjWFN|FB7oo2sMb)rH2uVl5dJwW
zj9P$}D)E#i;vVSFy3|Q=pjP$hM3<5XN9>pvEc}VDMq)(=24q{EFJeBcYhtp%9ssjR
z**9|2o9(B0a(dcgtGcSZY<_f$J(fDcS|@w1C-Yz_HwJ73V@-5rS`q+9g-Wcc@5yS8
zf@yo63!c|gV2Jz=$AB?$+lkIPrkenGgcNo6>t-u4%x`icYR8}MkURO{dgy@L-O~e5
zc2)IZl0nT`WFk{1**+_@&A@7)l7gEoWy}x(<Hq<{eU@`u0Hpqq)~1GckK)htWxkrV
zHzUiVCa_bLEB^N19(u3Z)n8-9w2Vg#<_mG%sq>D;_)Nrd&Q5`6<-Fg>^KZ(%u7sor
zT;9WWQt-V1NbygaM-HxiO9%WVtJ+_VPD@$@hx4j+p}~n-vav>6eYON>MSkKjGu3U5
zp~pZ=sfd|X<v?=vE*cdNTvJn7cMI*&GjKs<aXtyk#wnP!7T(+PgjqOnWt&7=89F;e
zoB$zsEBK=1Ds^L46ueE#@}QN-l73!=EYki&CyI;OM6M}Lf%Sl(MS(XVAv3Vxs0A(x
zegFQciV+cs1BMO7?<EEA#2*HI#FYq31S$@=c`OmIxa1eo?2KDSJF-)JtEv%R#~sJs
z_r32OVz@3nc(Zn={`<VFHo_*Y>x`h9a-D!&a{@sFqse-HT)fE4D{^&yJ4U<a)a@w2
z1S@&=Uore++He1hB#|{aXzea0QuC|IVfPzx!H;4)_>(5K;lw^Ew#bt5uuT$sra$h8
z1Ie9U<zAJSykC>0H~gF~p!?vx2y$6pnrc7#&_bV88$hbKGeG%1R$AEjtm?-rp30a1
znq65l@%NhkW!_(Wlr}3OW%kagpmA))xJZ=RdrT-U{y8<*RlZDpWZ#1->y2E$k_+fi
zl4$)Ne&+g@6UX0CmrpA&FiG)OCNIRgY;Ep~lN#}UxySEY`Ab^)&0Lx+Q(g>}J6q8Y
zAtqJLX0dgsdGR(d5=So&0TPhVWE^dAHwMN0l#fhCx1&Os$YR!euu}(YPbtE;%(=`e
zy!5fC7GXT|9l^vwhKos<3KU&FcLG%`-73b1DcJY*p{*Mt&dN#?(^^&%XAY1%;N2OI
zkH0X-W+IMpC{9x`f<6dT%z$b^f3-WY<dns~Y2VP`6jfSF5C*`DN4f*mK#vpkw<g8j
z+)=3&)BPh*x2oJr*R1b>oR>@B<}&MZ4&vfXnT?eraTM&0u%f9Z^B^h&uM2OhnvkSP
zwe$RSqiVd~!E3~!*f%PGoq2}Whl*FfseLa*vr1k6>DBOJWwl}K!s*qe#?|7AkDO<2
z4O}#)k8oIN^W!(o!si|F-1N&cj8PCS>+;6k|HmZl-qdX7>C~Uze`J4-B!D!P5oMxz
z*<Q6UtyUrK$;{~TGyKFV5ILo9V@YM|;ip*uK>S*9WBK<HWgXu;xA>A9!Jw~(lF&kB
z%`cp2S`0848W;-MD3r`Pdoddp0i}hQ!(S;TtD&68j~0!c=m;``UTR#>2U<AzSU!H}
z-iRJmn5%aDVXOcRdOKDc-dfaCN0csDR2gnIJPjdwjLdtlA6+9cTRVKL4=^4G;<Zun
z<dii)<Ppg*<=%2%KcJ5G24;$Fr8c%i{pSOK5=do6r#ln0xTkFp2~xD#vU3Gdy19lD
z4v&ISlmRMT&<-qLKv{~FO)<-$L)_kSH!fw}Pv$Hx78(E-v@2uekFV>-y1HHfmFMkM
z)A%E2Pq{r%Q?9O({x7umB0%e}&Fg{pu+JOpNNd{@;9hf~B3*HZME^Plcb)Ci!ZICI
z0wj1I)c>?tx+q`b%bh#SQ{K?nn|!JrQ!Zs;NKrEQX`{%rXYw)JwBpc=yL`?N<oxW?
zzf03!H)Ois)SuY0?3dM&Nv-1H7A*zxdNCJY-xN#xH&O3X@#Acqu~JXqIV#<Ny$+ih
zD5D(5_>FE>K;eQJw|5+zyni8P@vU{Y4}X`X-%(#b-~NZw;U$IgG<)<RU@GWHT59_J
z88c6E*7h?ihe#7l6jF<EaR<c<Lfwh_Ap^E>prLJ>xJ$*=-Xf}z=BrSJO!tL%c<gAG
zAb!D}B1Om=x>do<D`vR@Qc$ej0wTw->CUQNtKCO&U=&cKYGw~;_u(aoo&i)@F9H>;
zlqLmy{NT+v1<zNf#N#X1-_8`{{#8Sk%91egncky@s>(`DA~zVgxXw3bCX$c*H6`Yf
zwcd1ZR&qRCKXff7By<{RcwMP=9gGUt8Dn;(+><q6guQl0Lir*>xh=u@yk(G`FkTn+
zG!=mQx2^a8PZjI5-GAoYFezU51Fk;66-uS{d&WzzuS`Af#GS9+9zTo^ElAzT$2c-q
zqP_>~nuO;XcdJrlw^p!a8_aPmO`rvS$X7PXq5=YVLV5!RBK0oZ$KPBwcS|~r?(Oer
zY<`>#)bW#T%ljECK_B!YjfQDQ7jcRb3@gp8K<T!eONTV}sh8Aw3L>n2-tC~JhDMDU
zriRbTwl?rq?AYAeSF-fa0oH<xmI<L60@ao@I96gRu*)oli7Zez^>oA$`wK5<BI3*<
zp>u7;IQ)8Chgd69ogc$j;68^AYQ6Og{fUO84)8H4YtDPPpVrJvQ)e{vX=JD+^H0e8
zmUp}I;_VEi>8RV0VTC|ls6BKmYn}OVw#CFQH4bEaF`Y^bD^W4%=#&Jf%%&lEEM{Mm
zh+Nn3<dF24Dy{tDG*KiXH_JU<Rvg8^!n*T8=iq4nM-nh`)qShfC9f7cbjw6)l)H+5
zS)A#{pT*d&>*&QufQhHDCQiyi5%d!c^BypE)#`Xc#BE`eNLSXOBY(R+AszROSn_B@
zrHfDT*UKta6m{pu)$x6R0H)#mB*g4wzC}#x9x73uWayq7TY+}w0Z3V@?7O5zO~m;p
zTKNLqmNg=B%DBizVCg~!YWTJ6xt<vMv32t*cc-6u0q3_v4TD@hMAmyPntYnl<d0w$
zn>XBt1AeW~U9k|poOi~V#{ce4k)+`(Pb1Ap1^h@D`<dZD5E5F<%spmSpdsS*oP5@u
zd$8*4ul&>i1IS6!5O%_b!2J8ckYD?T-}NYcGm*@TgXYT`3=^UCI>xkEY4IFD8OCTi
zm1wKd;a#(3SXiFAQz_D+Itzl2G^K&f)QWQ-$pjNa280c<0*Wp3YyOCywVNhlxQJVR
ztiI}x6<AkA@Y@>GDGsST*z;J-fg%LxEji%taZwF~xTE5*CXw70#UFqe1aQWIOL6y@
zr)h@{l&A6t%#r<`w)u{Ii$6*OMzvJp?OgU}U=>%?oG9Ral+7>sWnqTf%dH49@w1!g
z`rI7QdGTr*Vx~mZ2gbuO2I!xHCsY%ZLZXe3Q^^Dcf5s5Yu~JTyzM=Vdyh8XaYMbK4
z2S@AyB9^TsuM^4Y-s^tB#NV<!4=mNIXQ5o1EqwSqU5~<Km!f*nu*>9A-KVe0zr3vR
z#6y{1$iKh8tMSYM%pI#=*N&Gv#-ma$6Mq{YA1)S(FeW>vC<Ty)dOt*6En5VU>77GY
zhax%yNdH`ksH)mar>7P{@Y;IE*Ga`BpxVJbroLCW-(404HGdTlQeLe`9f&0EC|E?i
zj_T5>*-DC7Udp|4;__%J`@0Y9aX9qi=_CwwKI-yT32k}D-o9f>x#3cTSPU6Y)PF*T
z*fma?!5dKbb#DV*z)<s=V0QhpCi7U{xn$rj%gR^nyQzwLGN4LS)qlbKWgdoEIy;`#
zS(V4QJFnBm45kZ52gpQM1=B^WVA`ygA*LX_*2E5`X5J~W%ruW9x9zI_^={1^-qhx4
z^Qj0K#T=;)D+V}}Vq)NK(!<9qbeE#3q3_V0P13uj$1Toa!WilX7=u7bZg8~!8{Mw%
z^>6EImqr<0-M^~KT>=@k+1ZU3xz|$fozZWgEp_P1v`6X2NOW$R^I+jas((Ma7_4%e
zFUE6!uvGa+rWv9mXv)4@$z!8dvbAX{oq2wAA|)7bPchL_1j==jEF{v5znPnRNv!g<
z?oH)~SFT1#rR||AoXSrQ<zJ7u*G6Ryr7gb69PLCEK1#c;iVoeAhlBNm-@z(g_G}vY
zsK?lNa_erI-qRdiFouQ`FWkR<qX~c0@g}F=`G?xD>&~!Ur$0`Y1id_R_2L4oV)Kbf
z{Hc_U)i1M)+Y?ubL(o}u^=eAjRLp*RlhiNyCT3PAIat;TLB2u#8dxhEXu$KD-*Rse
z<JFVwfpJZrRVRY#BEG-4_4EwKfQoLQQ3;Wgb58(XD;~yhK9OU&*q8yf!zPi?6fb)-
z>t3`BA1G!~Q1Mnq4}w*|rq>C(?v}ISfH45pCbkok-KO5kuyLnfTC9`lm<=%>=n}Mx
zuTU~3-irC-L_5nH_j4AmfN(X&OCS+xW_Cx?tInT&*E>4GWYQ{q?zfCJO8}-_*~X!@
z^_H7u@-l!n7WN-ftW%jxYw?6I9*7dE=4AL7*|&L<r2e|l`pk3IFb+_N6vqKY`0WY*
zsotv6n-y2sQ-4leI({WJW8RdP(a15RNoo8DpEvzZKWh_`IqFQ$pr3C;P!B7v%>SEq
z{if$(wN~7zg5*$j!E?IFw=HhRsQo#xM3}jG;3>npy1;<S55wjf3X3VJ*heV}2jwa0
z-FJN=kyjqOm8;g5o7JlF+ymxf>2nHk!m;6VfL}>b-k#LaTGS^xa^S+JJVOe35BF9a
z6Ip7gN*8vvI)>1Hz&~Y*eT)i5kI25<gOO?4Uc@J;QHcz5M}8CA%QKCF_=phmT9=kg
zV~55wy7n6-WyligH$e2$UTg$jI%Rk(1DAkSrIB2F?MW+is_ke9Ht#I>&1piftw9?R
z+B#ciHgi3q1fk6#POyusEF~RcCFh`>ZD;NzE-Yt4vDTAA{vaqZs|^&ftD!ru&N;Xz
zFFtrNW+kzeA~{k|Xvz89+?-Hz?kNPnAZVtzPOjl&Oleu1kM~rB;6r0=Zvd)IXT*yc
zp}&$h*p@ar;f$zLgw(n$&-LJjHjS(9kk5l3!im54dYhk{nn=n}azKU8Pdcy-SH8+w
z#i*qkE@Lu<?Q>_pi>-?2PzS=RMDSoi()my=AW7)mz0wLO&EucifZihnBZ$#%Z%hOK
zQz(2(H2`+x^@BH8@O{`C(hmi1PaB1eva$?^d$&lL5M1PUz}eTni@twG2AwQ;`+6l>
zYfxT-lhZ!yE@*8qOEd7_4rsI2VF{;WvdrTZ=TM*rB$p$@p#+}c;^-`c=j2U+Blt{4
z+_C!ia?%YbMn0U4>p_Q=0z;E9H1$qE3NImK=vD?eI{9vq*5YMVxl~S8W7`G#M;w&>
z*j|UKwjiU)@^;S#ua^7|Gl<)&&!nk4T?Bu>?uAyI1pA-|*@Jv(V66A2so%tJD=UM2
zI6vPiP-+g}h*6SbCdA&+MFGKNJVN=6VUCnWG9P1klTozm`hJJmP@cb#)Ib0zXrv>e
zc;a2XTea6$t^KBaG}X6cy(?@{EvNTbbO~vCDhhFtORkExPPrs_Z7MXz@NU|~sPCkb
zetxb%AU_BzkxS2i^2~WB;{E&mWfYpcJMr3Yg>e8+UbjK;e!Nk=tN2`E!>0#34)`6o
zaHd2kB3Nv)CK%-{9A;WIph`F;jES1ix`Y%NuR>8CiQHY-pm-2?hEC^k1hLr;@zS)B
zd?;{U6JAEQ7UzsD!YNd|HE__|Fzc_6_dOzMoc?~!Ne~la1goi1VBn~ckUNp)`C8^C
z4k6r6rsmbPOZ1c<hc&?jiVkDy1Tvi%>S~y)i|JB?$)O4<4Xvl~<fgWcGtlz0X@DiJ
z|HBec^?|G==*p2b`@O?^H#f!naJX{B(%7&52bZhGx3;$SKEHlencX}!<eovd26}>A
z5W1*KwQZ4U-9L*}HPqy!Mw?0&sVG0sJ0bAjS%F}Wlqa$cbptuF%eaO0xu7w%>AFjR
z11c;TC5B+!zqFOQT`;;pC6*4gM`g7pg>5@s2qOeJvFy7-4eLW2XOIZX^7aQL1t62Z
ztDPX^q-Jky4^&}%)LzcM1c%rQj>RJ4XPM@UgHlHC@`_sHCYN;-icwF{a};^nTXUtf
z*>ym6l|REVxQDj)9m$HlMuXTE5aDE`?!RhViDW^G*t}1_!nG`(pBB%4TV-EWfQ&&?
z3-n&UpjyHdFq39Vjk7*il&e7@#|pf~nq`pjxOML;_ky+bMZeN{-F71eP<hKFW7-)j
z#WTK2<pKuLk;eAS9pdq#`r*Zkx=uqmkp`4bKt}17iS0f)HyzV;aKpos?Be9~{+kv7
zbW|@`tf;ZBs4Mli*cXFOHA{gNxR}OsDGOzQ-tcPN)?`^8d#nIgQwel8iTNvK{{C~J
zyN47%_O9085A*zjyZm`~;z+xwgxJBmF2U0U$xkKp<6nR2<kj?C5$5K7#X7X{%Gn+7
z4TbuLzq~jB!K3%)8ivGANEe-JmOsgI+_HgTlRXICdu_sbu<ruG%oPGc8T9!UJ%-wq
zV%5}!^oPu$n7l@+)FUAXud(74OhgeT9DHHr;cVySf&C21LGb}`tOym7Vrnnl;CtJ+
zm7(s&GN+QWu@)+Pr3ib>H_~tm1YdBRFfzRCi<B9DmSu7AN*vS~g5ND@8Toy-8d%Z=
z^6%lkW|wIK)j_O1x{Y`)Q59Z}OUEGdbJA}8%im|0%EQeFG1rZY3@cVVgDGh2*G87I
zEalo}Cv<05f9(3{uN-H8|Hq}hzYhAI-JM!{d9=DJ<=Of3OWfLQr*bHJV?)E8=zb8C
zPbe2i$p+ik&kr{ZwTgf-F^1gd5b>g<#HVue)3v|Pidm@S9sp!($f7y5P;$d+&C884
zJA{h=Apg}35d>ZLWX0ZpetCa8i-yRU;HPhg6OUiKr;XQthR2rs=j)sso-mBse=L@Y
zk34754^?JY*>6u)YH(@&S?$Z;u3f~Oc*r<0lX-H<hOGXmao=3AvPHyRBmCt<547Gr
z2FSSEzzA*}jV!gn(G%>*RN(Q!0X{S<(3-58C+5imPK4pSEVJ{%oSB=m4S1Y`g;HC%
z2WVnE^)ZtMFlHQlB;v`l)qu&yIP+w2O@xpoDOEX^XDGvC=vKHIgU*|a@m5=>9JV;c
z-s=)fl<!ankhc_JMC1U&MqD?BnL(0okydUU&a%18sX@kceqLpN2Zr}Eb&G{v3!RZ$
zlCHzG_X!Rr_krom;6EpMf0_d}Rh-#}<d!`{<Gpw*Kr)*`9RL;5u3PW#|K5-a@6tDv
zGYq=a?X>*~2ZN=mKcW-6Q~u1tZ<~Hu3=)QnaVG?Wa;5Lb1<xPdq#1+hBMtN2<3ytI
z7Jm*dhJTOQnLXN!GJwZRb;vg)?u*@<pXtct5oeVGNme1Nl)Og}KUE?CG7=(6k#vLH
zd$+1Y2=|#up+fstIuP5o4UNF-9u2;1n{{Q7eC56oC!Yb?if2Mtifwi+)pFglRm=F>
zk+xz+p~+pcHRHv(yi!~U8bU`sF?fYHhg}VwYQ3kAdQ8^B#b1L57v?U>Q0!Kzd*>oT
zX5l*#`Y~jv+2urET}R;u+li09+x%WN2#GoGtYt1U@Hng(^O%`e!ukLK7a>L|O_I=F
z2DJbnWt6B|;)bVkM&Wx;WX$^N(GG{66v;2M3v*+Aqwg6^`{}}5FV_!Mx%EZm-E|z@
zeO8ntv77!%iATZO%d=u_$T)c3F@zhopK&zjl&`e$-(Qm~Um(vpEZgq;A`#_c8J}1E
zC9WQd0S4OsHZaJTTjs*6sy*CRGrR)Zu*O=@cvL-w>MWRiI^Fm-S!Ygb1D*dsUH17I
z`G@xa`~*1>@$H-`71;0@KIM1)cPU*8gh#wY>^s#hTXM!K;<_r?{6&!wH4of`!vuTt
zN{cv0#Y{wAAz(QwTq%g%L6~cZGt+&JUN8V&xt_+h+MWhsubqdTk(h{LPw^21-6Xic
zAd@_o8W0gf7#`xmxk2vxz-z%Kxviv%Zcarz0rmnawG1o67Ga7SG39MI+XM^(3}6-<
z*>3fj;&gpNpZ#?tB0sSOlf6qrq4aGq9ddfWPMK5|p|I79WGHHKx=&iLWl}nG{HoNw
z=0DWVrCR}4r2(%0%Yc_fDkkzmD?6xBQWmH?O7VxqZ&H+Col&xyB4_wm{q8lLc(5F%
z+Q@0g4BWIr{SOPP({PBrT^@rocciwf=<=_mJ3UwHe|q#ots{4+>?-yc3T(a_ygcYz
zdhxm^HFU|qJqRR3j5tdNhzut9ea>4mS=HMc3uEU?=v2|3Jauy$Wc5JkR{Y_T7rIf!
z7%F-(OQuz^^Ie%>vN4jeb9VLh<{Po&s`3g9vN@#xl@{je$zIx}dU6JBGdI}2ASr)|
zFZH1KjR-V&X#Q0r!Nv!aH2{$42i(hqv1A6(?L;fycp$G%Uy(i%x@a<-RE9On=bs~o
z7pSTSAY&qVu&}3MSadc`gpcu76>K1%d6-+BXe|b&YQ<N=kKhLoQ9p4Y6YL2PJ3T{7
z7oZ+87qTMt3Mp%1vR(OzmD4Hi<X1^onw~Q2$|yeCzYV6(emOP|GACz;xPdR72w*Zw
z-SQ%S=;IFbI0?MX@*zb><Yi|Kxq}`&;RdW12p3X203%^EK;cPy@syUi>|fU%hm9jM
zh<sSg)UCamF&1kZqAL0IpP#n}zxX{UMX7kl##}Ey@#lquCg{WA41G!}{Ynd~<+bET
zoh0Vv3vIKn_Ti)VrwR}tS0kYsS!pM!M7j#;T0#uI(-Q{8{`gb#Jd`{mc3kaNbR^fe
z|72Kj+kt@$nt=xzq&<Ol0n|(Iq?*_kl+Cis@(tyhaTHEY%`5<Vc*I>M`bXApQ4K`o
zI`V({885Pnf#*S4PhOd^YwX>D(R7F@82$F*?x|7faCW|ekZtFC(Tf9SBUl@RkV89+
zEtWj+Z8UBO61=?%gy3bkXjm(#8YJ2t99g6D%tzVw#|lk5W8jl#zBo`?r}u98`EIb5
zM!q#Aw44aI_icl;wNgnxuIBQ}%@U;qV@6?sUqw^ZT&A#Xfu4pK77?hSBLiSMZ;gX}
zg{A@XPM0IbCjUlK=S|*aoK!3Sqh+<`nG~<D$e(ujV!`F<kOB5oV&kfzTl~rW47&w4
z>Q<(dM&XtC3SOmjXWdJ_^jhl8y%wUaOvPz*2olbECa2@?7cXQ1<{k*q35hvS@gh$p
z0S_DZ2<&PE&Bz7kJCc}bg3%nz=x}s@RtfG6`Ub<uoldnPrKTLEWuf6+&(4{{yB^yT
zcPU^o#s$1|q?M=(A=?ZNl9zxsoM3`)oE+BLrBGT3D0)qg?GCob&ekzL(NGcV?Xb6E
zNW?v3dkIZUcBlF-Vu>C+qx4C>OByZ-4Jkt<feFc9Lb^FY5KqddCs(f!%$KtG=kVBS
zqKQN1bp_y{l6dSmkwEZqUg}zi-fll=x!KH2=F~rlrXxor=TBY*zFe065{BW-xyJ04
zgjk!0ZaM7XWxMA{#^R{X|A50(Dz=4xw*!svPQ7r5=HYWiHLEf~T8fn-eDs~gE{Z%{
z8$vERCe>d%A$T6Yq2Z&O=6O>)DN?|1{oH99<0xt3k$ZvaqQ0+V<vCoc)EtW8yLBsu
zS@QD1#7@*!uu0Lg8_5as+HLW_l)GT}o>1_EpXd6Meq~8kQ_UlV0Q*<(`M_${)009N
ziIbe1&c~_+SQfI_rh$G+j-tm6a6E}fRz))T&O2dT6H%36OkAKr$byG0NxmnaROSdL
zyI-9)4JMX+7Yk>A_bFD;9;Sq04z3W~Qh;6baE|#)CBr9?AzWKc4iF$$o!<5z=kK$P
z;B+qAK?%$0WLvt@d2-bQ-%C?5dZLV@?3{g@zJA}B3nSkq3xP*{)zq@C*W`h~=9(pF
zXAOWpoA84@In7mbQz0Y9yXGAs_AS!vgjn0$?N{08@2_%52g-=`^c0+GY}~}$<67lw
z_G_xUPQ0$Z>xQf~@15nLZ?D0qI8zi3<IaGYpVs$@{*q~~R4|*X?Z|!v)*%xx%R?~k
ztMcM8b~NBi4qOwb>QJ@iTTy`vpDZD|>x1K8-t0ZKN<h23a)nsNAEsTimoVb63n*aW
z-BD%mf@w^9_MEJ;Q`Z;V^Q`8_NQSyL!U>OqSLJAh@MoLL)VEzRa3+@3r&H;u0t09q
z^qHt8VoCfY<}=1$*o}2Tn1d?SUNv{8r<|e<%g+pim*Jw6Ahr|WS%0LIA|;;PA`ovE
zd5dQEka?USx%Dzcibj1jB_DHVoF}I1?#hA$$zo&i{l?cQgPT{lte_x^pqpMC4-*H0
zBKQ9+{<GUY*)HhKlsrP}QzQzHH^x2c{^x&qYE_vX>PD=PQ-t}~QfIyvRBsZZ%>S6r
zzT)>vKYTnq%gg0QM=jF$L@k%z0uYY~s)1a|e^od`bfB9PWLzh#_tzt7lG~Ctp%6-|
zF{|#ph2-S4?A_CMZ8(2{35ECCkHMv8_JLrv@pli@kH08vaj)O2d5s4p!wg?joVZ7|
zBTIw%F?0RZDP?v(a0@DGHCF{<T9}#KF)*5}$Ds7E(7{<Di$2eznwXGb(`$=fH(`yN
zKFG&_$cOR=R{8OFVf<k2z~P9G544q0poF((qdKmHKAX%e^Kz>s#knU%HkGl5=TIfr
zd+oBXuDj=-_!<m(i4GIAmqB6x4WJ}Rt829<R^m92*^G_!^q2j3y&-XP;r-A4!l(_7
z^8KOg=MR~dv$U(N06ofXHaw^8SvDtsCVJ2=$uoO4;()^6%b4n^F(R4x6i6*Qp%6^^
z=P!|OGJ*nsq=%tzxcoC#f9!?VB=6<O%*T9;>zj7L^FnJ@Iko41a%{wi+yL&ATir&T
zsT)zyxktc<sr6Z)K9#PxHdViCm&!kk!*}e}AIa>Sb<?Yo1VF|wdRZ|2d&P@bMFMy=
zrD)n%r@4e`YCgqL`)>SH)|y}#UHuV8uB91<HS>;l7bL_gSlxy;i6PHZ?McH0!uVL{
zH0dH>E&rjdSjSO#CnjFImqv9e;01p!6(N4Xy`Tt&L4W3D1VUe<Dk0b#Fg9SPE0i`6
z`7|+~alw%T$rBCX3<MTa4-CN>!ApA-H7{7pPi}7diluDK2jaJK<W7u~+Y#)n&Af2U
zGlxZu3*vC7s8mBu&l>mnd&WRc_4L-~u8BAhN&R=ls-AEJqmj)B{Pi|hRutG)+8J5j
zKeDSvKvF~i9od|B_5XuPP1cJ$vV^d%4Z(CxrW@)8ia}Kt*BWT^i4(#mBRgJ+C}`)$
z<#p}j@N+(Xec(&)=7=FVCQb~+w`e581w}_bW=iKP9GKJAXMzp1(`Xh}5DVxvfHd4z
z#Fs_A-i{ZC_vKd#IyCkVpQd0A)B-d9z)LqrMWxwo&|vzLSGdPYue*WuMxQnFARJkW
zaz^heDB8mfmT7GZaO|^B4Z*_SD8AwF(zcoU3C#5<aOc@NBke;Ie2nyS1em#!^<!k2
zqBf_WHfb|;*^j+%&)QV)uhD?Zej3jX5lJIVhmpF4k&SbUT&}X0;J;a`RmVlK6ABEM
zN+mdYkrjQTd^yl4!zvFbv8r3Q{c&m}8_VG66MmarpU#v1iohxUYboHzk#2|qYz0gG
zN`sjXSCg(c_}Kz<XzeUCc3=7}@16gu*9wCPWA<Lo(|d?)wqY`$>$#^uHJ(Cgc=Oe`
zUJyTfH}W4)v>3fHtHlVbOe+5Tr1KdSy`9LLKJn^EOm5?B89;uV-GqUrB|&$l5EpJB
zfJ`ueb}|eS0I)9%AO#o}-v8LCG)gGNeb0OERAfTDh4UWlRoC9z1yW$t0<w5%@GV3J
zqsg*l46=|xgxEwtf$mm3y9eh)25fh2z;b+BHB+LlfVmo|eiFl?U;(T2=E^_$LRKpO
z<a`7J!yy5IwezeKj>BbsbDs_TSftUxQO<B$`;?ZguyaRE+-vn`aYGacep~5a@N^s8
ze+me(CUJjfW1qaQXyln>imkNbJ}R%dquChIT`$l6{DJpSH@}LSb$*(Sm+GF~C+<4L
z)>dz0nDh~+k$hhs?n$+-`<31Ys_)8N4`rEwqwf`E9tekG!7A9@{-aZv`)%b1>kk!<
zpD~YKuT<~24$-`3(T|A{mf^G&vh<ZA)@En2FjjZ*IJ1O!g_)CipQD--7f36#f&<xR
zGw{4Db7X<=1B88(8BpK!;z_;4-X|5l7Nh04Ap`znPYQL21J{TpU-H}Z#o<+t$vGX{
zg*g4(93VY{|8Y_HJwohVJ5ZAXBbx?$_!xke9%~E!(*g5OzMi@}bLUFz%;MY_&?{Fl
z13rwA3`WYAM$7H|{JOdw7Ul}?zl%SD54{n!Zs2#7b;Z)Uk*|0vWXQWxVhMkMG1c~|
zy)iyHcluON=3mvWS-~B%T3ppUhV&)a<!%~(@JrX8k-}<+F|<R$e$`6hf4_AYEBAKj
zQL?2pm=*?EAHo)T#BXku<gmq}T^#2qFl={LlR(e+PgMGh%Gdq*E`iDq_#dkMFS_16
zp6NdPAMfsXmr5seCx@JJ=hVrRvpZ+e(QJl=jo3zVD1?qGa+>o*M~ZD`Z*5NDE;)>r
zJ8ZKp2}R6VBV+b`-QVx;kNduVpN~KC=#RwZdR>R-^LbsDGtoPjcEVtX?!ECcbH3w)
zEuiNycg}GCW)GSpgpm9D5r`ak2k8)_aEN=3Na-2<;emQ|bm@F)g)@l}=@fnAv>0US
zoIrvav5UkW$?gg6zCZx`bi8j3?n1hw4|W`Kg1<Y@cVm@h=!%ki$CM)S75r~mbCil%
z&NU>|ap3YA6i=8RrH8d95bRn=(QMlUuLje2g+0EVCv@603p<OCj7EfCL5XtWgiK5N
zP^qg~oxoGlJsR=(^=H9CM#gjQWfpyAEq_3Zf0ciAE&t;^dhu9NkxgcvI*OV4KdYMe
z?F{55L)sXmo{hG7Cwb$|$Fcmtm4Pbb{VH_b-Hs;eqoAT-Z~lK6d&5Rk^@z)659bZY
zns!V~>GL-^xZl$X!zON?0jo1x4(#Xd$t>4WzNbdUcGw{eSgJWpB7R4_#_1s?mizPb
z+9f3L!b^k0+|T+@x9OciX_HAuk=qd*F9NOGC;RJ|-D<L8VB~6wbFqRiSLbcnb58?S
z(s}Y(f4c|9Hz`D4LJ}pzk7`|T_$?*f18o0$S9YF%lZN5l484s2wUG{+H18Z?JV^1p
z(&g_jDl_`KHOTOA(dw{wqJuRY*%+!V>aZO=XP~6@DI(PvqQ+4edsK8?j71}}rZ!A$
zxahNm&@*ZZi2Q4w5e{0UM|IbAI8uhdKLo(8Mbzt?`t3dbn@7Pmg|<}pbnO3P(r#<_
zd(c*}(f$$<YnRhkx$H7ktr_`M!m;^~kUioXjkF8eqs_0`4+ID=2;{Ntm<|qNW0a%b
zQg1c@1Z(rO$h>52r1<uxcNaE)S+`D~r4=@PVarxlkAGytR^w?9QM-Ksk(7hU-lc4~
zUnPv|`hdZLGJYj%!G&oVsGYy_!3%cEz=)$XSUL}T5PH0_33p}9=ES&|=0S>UhS=QT
z-GSfW``;>Rzmw3<BR%3%z;dq0%-!S?^<e+*izj~LR-MJ0+c1wInE=t&YTdIB{-SM<
z4csD&>2|8i(c%Y??gt#-br8|^?QniX_$ky<PRoR_sCVV-kPt=MQwPr3S$K{w*?W1a
zfIj;nxFlY!;s)5C)~)+l0+a<DDDDyRQy+J9d=uE-`Jd^wU-Ki%?6V9q4v2I*bb{T?
zukpIl`e(yC=ipCHS8pn+dVcwLD3jLa7L$+}pK!iW-K>b3JT-C%ESJf^)IX`*LB#jw
zK3g?*_>dmse5p24$UX5y-Q9A>omDQ7(_YilKo*+2#BhvOy}34mNY31$>|ALd0p<|O
zys8X{ek%K5zJPD|TDCLLwT$w-!8zJlprhghAIr?pQVw}$htO*}z~XKXi>z@g$QBl|
zaR42>Ye4jg1Pfc(d%_r3z2)f9mHJ&}pY7kBAMM;k3-P362M|0dR6QU0Za){THv6ZP
zAe<`hBgVbH8=eg|tA%^DX?jnu{<<mt39T^{?c&mXF2#&##er}O4cro`w^UkBR8RKZ
zfw{GVRq+k;H*NwHdti+V*vc1bw#P})A7N<R`s=r1!7io$1tP?>zua>s>NbW?`s6PR
zU(??pdXkuHapia0oOMOy5ZS9ue;f2{I`#7|K6I0^Y-Gq%jD=2U@g2*=pul|Il$Zh{
zd|zYf_Mn~XwB&xt!4i*QDpc<liz+_R&b$q}(Cu4q3&y0rAqsSqH@<Rq%Qn@cv|H&>
z%O;3!K*K~b2IM@(?zwBpD=i%JoUrT_(&%IXr;GZdG&A7DBDaJCJS=@JUt=S8fiXS?
z?iqpk;7z!5{+wO+5ae~emtr+3wL*CXOkozHz#P}@SNINT0&sBOTj0v+SXkFAPV2As
z%V1jJq{P@!W~H|giq|pAb5i?0HU8-%gyWS^)2te>o#g}qEL(VTxBW!}?G7Ng1=Ce6
zxh@;&O~q|3#R!jwDlSR-I%m=i^xph$W*nO-V|O1ag@QlkXKhznYZEov*bo!hx32SA
z*k4Dc&uZ0o<NJ1{C2fSx)pGXkz^wC#Ns$AD1u}aZ&eQ28ULB6_-3@^W%|n4rd)*2y
zlxg#e`C40*U$ze1ibRa@9Nmtr)@FjL^B&FaDaEmZ>8kr?&msqi8#Kt}yu)Ot>+Wws
zGYCp?LGT;n?K}O>xox8Ez-e2+ofv0Bn1|v@&ey+K+jH%+W5qgynqlqZFtuI9ZHi!z
zoMuzmJHF#}aNEoI`j$^v6DJtaJNp88EfO`HRJjun^e=$2Ty0HP!<YM|o_|^=$Uf&t
zQ_@2Ep-~oi-V5TWh^?Eq&X3^tRE3{cu+P^pBZ8=S2=;m5;yV6WY}D(#$9t6@s5gPJ
zniyD)K80!G^FLOT0SDc4b@f&gGqB#<ot)4`BOEQBVW=wwvD`s=W}?3tTGXC?{cPlw
zmU2x_rf~EX!QBE(1(VM1pu~M?7i+OK=&)d8TcWY_PUmhtR?_v+Ag8EkojWHs1cO6f
zU4LwO{rv9D9o~R#w0Hy5I2T;sQl7sXiz(bi*5=#Nm&?}jw1~4_PNrDj1Nc`D4>&(`
zO%#__q-6M|;Bn0;r>9%aVb-V@C~bsy(AQ<Ei<WG3-zRO>19LEC;0>#6<hDch*Ez!0
z=H9D-jjb(Q29Ltwslh&=TgGd(l^y(Y_A+IH_bS6<U$ptPwY%SsT(($K9}5z76?0Nh
zA3c?0V0LBVli<KW?Xbl@-?N|&0)Y|W8W;ldVA5KC#%=!9<IY*F9nOps_b*rJ`%LCE
z3%5jX%8B>+-~6$*wAPlt9i*C-_6;^)Xp+&!!iNavH0RBQVb80d1;_Wd!J~mvWFs`h
z_hdY}m_+$<48m5?J4f=`UkDVFK&E-5#S%oG`Hc}>+GW2pRV&uER@DU<*a<4yar#*e
zfcclp?2v`S2gIIUjI-1nFAIn3jE!{97+3oaXS}hrm<SkaCPV0y-`BXqKjj$iHo|~q
zA9r8+JF5Z=EV$633%2SF;!fUXbdf_swGa;J2F_X8HTS6TUJn?T!4mt`VD!lT8<b+M
zV~T~tU#(59@Nt5e`X{%hGOKO^)F}F`bGq0<J=3WwiorCmA#@MPn@f9}gtbE928^(G
z0Pcn{$OSt9>|j1{x7R9$`<M^G85i)RTLGpY_3GyMrEfEVLE}Nl;?P(Bvp8{enc*4c
z+8GQ#>)5Nvz`aZFo>rk3`rbiyVw=y6TmKx)FKdGX#SQ)M!Y2x#V3eo$!s@GI7fog=
zLdx)7MB^j0i)PrMw+pYd>NGOn4y<7uvvq<P(^?NWj<ngmo3xSIl{$AfRl>e8bnUpn
z+_#aL%fX_oPi61mdVi6!5I{C9AkCXP!Aj3vEVin2&9lax!|rs_P6+^8plrF;T^PBx
z0$cv+2n8qZ6+2J8_X2FUrI*QS%632ZnlY;#STHAEpfA#CYUwyn3=K=dztYzL3yC7N
zMykzm!qQfFSGa|I%^#}X(!N+!fjmA~z`(f(i!5RnCO&=6`S84Z`0I^#$3X1}IznjR
zYwz1*>-Zm^x4PM;8+a&!Vbk~K+W!lbaW<$7_V-b5GCXBl$*|2f%${=d%8-v{<yIzb
z!#?Z299T71hD=w1LwiH-b~8C%3Ycf<odYcczwtTDvmy`|jV}H8!=NQ50O=#A6@7F-
zslq(--M||0v6D{F@7msFT1ai$A5Jf4e;mm1_1*_he)VMhz})lFs&Dp}zy^2Ib#Nz|
zQ|2zPd`kz+RVQz;LpUUYE#hPSud+v;|61{67id2{u5&~h)!-}Dg}mdP{=3@QvyYo|
z2i%3&g#-^?>ps)*8bkc_wdWzQro#S#yF60&&)`M(-=PEE@LYcD$^9SKPp>4zUXf_E
z#*HjWmpUj7@~hTb?^wI>`dX)qUBCN0P5sU3ng7d-0DDcO*)x7&q4!i!s{psevJAWZ
z=qI{Db3Q}uXQ?I())MtIiC^d+XurQ%Zh^Dq=wH@rw)}{Ly9b!YdvIyc5ptRGmTj#|
z#iONV?<r$raxt^1X%P2b)QPjDfo#g>mz}itMh_dKX5QX?VI|>=D)Y)e)Foe2M=4Np
zsBd(%Z{XdnQ3`GIJKibwr~TN?%JKGyYPd?d>Q(zmo~|yp>K`BGUsOl!kScoX=oESK
z$-O06d$wiRFMn-#v_<yOWP|qECXHE(?W3~soqUHYcJpEAbgJpO-`d8iYfgmo0BdR^
zhq{N6it=2ti?j*e9uF|x=eOcpW|sxm_X_u>TjZx;w9(0&E1F90YptVPv_xsS)V5j%
zvT9oa?a#3#FwKE~jkROu+=V8RUs)CzpP|;--z`oFI7fVyf0qs;5)7I`kZD_Z*vR+6
zs{4r;8+8%<g4j6U=P^MJgW67!4{X9B;@(@ii4*bDy+&^(+qjlYIY{DAqUkAJgEL@B
zM!pZVujpnQx}QVx5p{96sNtTJj?5RKC*4l=<u_PXRUE9EjK})PglBfpa;QyFsB{C%
z$yXY}-|zKJ`i;9OVTzk(pgHWXm|y;3wCi`@Y1ZqCfXy0HmSvx==3vG3vL-9bgFg$e
zG&~~US4wR^n#87IH)5tHDkq1HrzVe-6?!!m=s^vc)SpZ59}T%JbqLCN<Y&kVCmeL5
z>bU5`o#f|d0&+2tTI5{JLUEjqCC3GVC30?1<Sw}ClZQO&_A0NZFZxmB{F=lb%2h1{
zT!m$nfjwkg=_%s*#+_H*z`ITJ9YRRWrIDi*jQpe8oPCsJ<eW8ZBlz|NxI@G(&wBN}
z3by$7swas%z)12NKm0aU;tk41G}}L+e(8Ais;X*rTqnN*e$$$}VkND6s>M2tbTP?f
zu*E{0OGoZV3GGqcwfP_X`{98h_>hpEri8iL+Z2{}3*CLejM|oCVc2kL^iaA1Ds@Ze
z-Gt-KR)&iIwwuFWz-}l=U(^n9Ybj^TtOQn;N)yV`<OUO;-RVbFb`>{#uF2L)udnKi
zxeCWjA=QN);Wr&&=p1%ujXGCayHwGoTA4v1A977Eh0f0tS}x-W)<yk28TiJHr_NQW
z<JNd6HRl+#F`Hk#IkdAvV2s?!C21pL(X=L#$S?Nu-B$tIVdzu_4M`rw`6~~MiRR`c
zi?A}zNW^u)(d~G@H6CiL#MaqX>Yp6@iZ}ENSZ+(DDV*_<pvoH7g9hl_%XpMJ&hpb9
zO5=wJ8WlI2b}*gY#Gm(<pdX*lLZ%T)O7T4=0V9*|iwYX9E%u1-yW~!JHCi?vO9-^!
zoC+~$OWYTH=m9keGZkRL6#C&gNqCwI)b%ZX1~F3|QG=f4!H+rq=Z#LQCfo6out?bx
zP3UNRKms(9u>7Jm*r7=BM2lsWr5jWA!#L`=^tF<O=+^93wGu$}dK|heu5_o}pwtD)
z`wC?+_n9I&?}@kt*-0gG6YmRRTQh0puXUt=!t-h;&+*Syxj;rCn^`w(jaUaNwa7Be
z`vVj?410)a4Nv%<$F-wHkB?1~XPbLHtB;DFdZ=~g7_bFj;2U0<mN7a%+LTADsC+(<
zQR0&pNyn+aBl*XJ3i65^WEew0WwU`(_O53xmQmz`T?Tvg<%0;Ta2|eri`VETeLLEi
zw6>gekgt}=FL0rfcVj%J>tZ>+oL#Q3e^TG8F>FQ9ESFpEReP`!di`qq@>5%d8+hch
z;P)(p#8>XVcp8pgsKNKi^^Z-#CbtS#+MUY{%=+V}!z4y-|9G*hzU*5-+<_G66xk!K
zQxYGG1F<Vh2|q9Y_|c%|LR`78D-NDM2^?;;Qnkf;UbVb!eL^m3HFD{$u=ZX5j1j2#
zG?B7n3M#%OP~efAXCmormn41f$sa^Y8fGGfjLzWfBj5M%{!%8(iuf6<gdv;;4@rS<
z{xGsN@_R5mQ0!l?Lq<0S@G!z$KSBhG%pb|)nT%+b)^tAez_1xHcCgAYT=zO?7`Ejp
z-#lJCgWuMb$oq+aUB$oUJqNqdXfyjLbuC^aZ{f1nOh>Lo)OKpahYgHxH}HByhLssM
zWKWC@%mqqer<f(rg)5cNyey+9y6)Vs5%ESwCO`RHIfUD#AGoliEiugs41kA-=w09P
z&hOPz%{5#QuB>*Ew+xZp+4xVBnXgu@JI&O}5ueGwkKNF)9DH;>bY(^PW6?l@x|78J
z%-G_FIJ@~XTQO?S9l2h#$uM<tO(Uzzn)^t;1;f5ck!d@My-I0XmGc+5pQ<_!&jjSK
zE?N-%alm*jkIgxI(>-@N?YW<P&U>6hYa2tUlnbo6mZ^dd`NsTfzjN(hSrWlLR7njz
z85?&VHzhg^r!@%>T>DDXdz=gkGFAMBiYnxwpWjD;BLJ@TV`KjOd0no4%dlLf<xPXN
zP`n+?E1hlisg`0{^z%E%vNN$F>NnWO0bH0Jx0P<0CC<QC?5HzNC&YbEh`r`in-9Y<
z8scN4qk}FFw7kC~QZn+Xju`&R`1Y+>{MI>&yYC`3=|XY_eg~y3Dv4lb&^o_^6>`Oi
zIS!sKe`s1TC@sU?v|_bRH|qW?SvahHGS0>FTmB4v?V~I+Nm!0L3j47$s3bu*Bc}W?
zS@2<`eku22g2ifn>~yU{XN}qoO2A+nwe!Q2f(EkS%=!$<;xm|%KnLDRC+4s7jO$m0
z=))H^iO)=l-aHWae%R45^tEzO6pLYTZ%Y$Wse$x`LuJ+8;ZNsnybz~L&h-^bNvVLr
z#pU>WZ^PwaZr$UWN0DDio|VwL4~3TH`tNj#R2MbzURCId{^;et#cyjLZiSdQ!m5Jo
zL<k8oD~q&>`)IBMir>ODXkq4H;t;~)sSK-QOPK`c1CjL_LcG3#k>N?NM-P{O4vy{|
z84msY5bc@Qf0Fo|p0bV)IWn)J%=<+*$Lk;X*R6vJLy9|Q1{dEx8uI^%PXDLV{i!)j
z-;9j~4%a9R>Hpcd@(~q``L5=iv+_h&`~#HYb$c?Io*jL0AD8AS4*UNuNA|SU7L0jx
z+2GPLDDhym_@1CKZh)CBG+YB~*&D1I9a4YF2dPFp!H%z-JwQL)@d}ox5MIy<>0B0v
zkz@B#S_QgX!*&5qnna1SP^lgkK>@#UshzdDu+-o;JE-wn>h~xM9pMQ*cx%1<cHV`#
zwD)()O^J~ry5$3gTqQLR=IRB8{4o=Z$@nlEZo>C!jscrwjM&*)Yt6OV>VffA$2ClZ
zL~_g9YOUe2{nUGyTj=zdoVM9vL1A<A9oAxgzFZaUq^n!eaf`^nkf@JRGlkG_8>TP`
z4_2BfG?y$KOcyu4sb3E@CuDSYS(VKT%jPCFsPSZeYyW9GRU6~zlcU?(n<1prMNZX>
zQe%i?PTbLzerY{oi~J8K;tC?mw7RdeDQP^Uez-5ro{=e<;w4e~<;Cmpd{K3X(Hw23
z7w#6Hnn8(k^~SAdT*zTvIQKO_#qxN=A>C8^*917h+}{u>=*%r}e{oIk+RAw^RmL!`
zlD-i$q+XIDK1+!&LftYRpxnoa8L+1e+OsK9`26wV8l6|-8x(W~Tg!<AOEsobP`8b%
z3==tel{$Hx?G#jgq#U#ucg4-2^8Q9|luKp6@32NgJxZHo8&l!00AH;F2fAcwyi?>K
z((EDd@Rx>5<sI%5OE5a=1T9v(q<!{VzMhArg>1BSf4Q62>l!+fR83yGY0V6_fa%@$
zBXcpyHz)}qM!w_qo*4#pXmhx^p~YUKqEM^<rrOJ@!qdzX`l1t*hLpzTihW@}L>4l3
z+6K(k3y#De3$~a9Bbuvv9g4R2as%YHyFt-twuzqa_B-k$KHe(Lz+0<e6a?pwFZvBX
zkGt4z4dH(M*>c?+)WVhuE_?FtVC&Z76ftqjWzRek-dlipZAfD0V*AeGs}0GwJiZNj
z?0h%2n`_h}7&5KpWZ+$CERq6JW@<@d%urW6BUTCVf^A$|$TEcejIV#}bHf3VW-m91
z;+;)E6RC&{u?@aKO9aQneIElSmTD*%ccwy{V5)H=X0C#(1eKhHVaLXXC4-+nP0aB?
zhKLl+e*HQzD!LN#`E|JxCi8*~+;lC}r-(^zZP(zJ8S~u^jk~SpPZ+jE)e@FmG}71o
z)8D>WaG&7q^hMo&NUaE;t7g<JU$!o_#v4mcHCTPIS4;4)ok;k!nqL<??1r9OY4C_t
zTl04$SQfm0ZS<&tX;HO}MH*VkD+%=LJ4-lF`YuRkgUQm@<`kRbpLS>2l;*p<Pr|D<
z)vALZcrp(~H144JI^u46#P%oBV3&<&C|xcU^9;NeIjwGK@iP9C=R@l6RT{Wkf@9nn
zicf}Xap_PmlY>xe8<Vd(DDp=reMr=!W698B@OX|kQsuos8xcoM@f3Ps=T>nGjZ!dk
z$GEC<Hz+~=Y=3E7npVb17?g<b5zUEQ^v;Ewls~G882ic!-)~hEe00wirB+fx@}@Jv
zRgTPDYo`9aElOB+=$Tyg`S=aciUmP>f*<3}U!1?S7W${j{?weJ%VW*n!As?OBp$M=
zMz3tGrGk$MD|PgHd(!n|(OH6pYFj8RRvi^tax^{C#9|#J@u}X5y8Ptu=!NOc{o-s}
zsO0Sr&EgXFE1p(>FD!Ham~phPzjYB!o^>x3BPQa%YW!U&BIdS_e8hu&{7oKphB$xE
zomHf12IBzg$Qw$^MX?`ct1rPEcS6vMe&*rdH*)X^o^8*l^$vwastffuV#G6CLW834
z<jE{jD-?p|uBS)YF=K4FM(nXXcdHi)4I~BrL88G%jC7E)1%t|<+z;|TT9CNC(kI<r
zLpep#UMQ@5VT3-mH1G|4HhllV(9poJr?1h<G&S2Bkq*jVGivMSWMDYr7B!b<`rypA
z8~Xn=)b-R~vEVnuZhBiw@rf8mORpp_6s=;v1fJ+$noY;#8!pD9Z0Elx|BAUHiu`~5
zRf3uaM!1^qMw>#n8?XbK9k~y2^7zQMQllT8pl4WF$K7y_U$wnsCQ^wNV9bhG9io9$
zs+VZ<FSv7Q&4TN&dja07l<V?f4=@waN=+n2FMKzS7eKg(>gJ8=f_j!GX(g$iW9UJK
z7mdlFynx@OnFerLZ_dSAaQiwvFqwmXl$1c%UCOsBVR_;z)6~crGCDo-5M|}Be%Y&|
zL{9%a*VytoM??DTo*)1H4rD3Kz<{RjY5>YyiC}T=MJL!I&?AtHk(L%(5_a9{al3(k
z@vV1o$c_AJpfPU2AH2@J9SOpX|7xmMjqiGbkJJ}^>wsKj-HqteYorNBKqD&Aa>BK@
zK$>-8RbZs0NVREeX76}@?qyzexFn#15wMRd%Mlzqm(M{~A;wnjj=EO-UiyHi866*Y
zIyH1F>+>s}($98La^<G+95Q4{<UA>hqjWMfJeB{fMQ}HO$VITLm^YT<KyJ*}<$g}n
z=HeKWjB5y-Kc5s$P-c$(#c3`#j3C^hS>B8sz;EjuK15CSL*YP-zc2MysU8zmCOHBd
z&cF7#!2ptA@LPB%C`PKiAzk1av6$5NOvQLH`ME^YL1J|czAYx_?zgPII-7aUq};24
z0af?>)wwI<7un89Uoiu3z{DHFMrTchTl}hawzqH1#mx0btVaIRrB+=@KFu`GF-WFs
zICTqV8>N-PkG+4P#f{R-=d!eRluq4moA!<ygyR++?V&N%ID61U%^I$SCQG`PDn<mf
z-W?6YM($U&L4(+7JcH$`vT6@<_47)o4W}KTXDN=U<??NE1-BrhiJkYGws8&G4rsUA
zLBn|){f4gD!8W6$Zz(=E2vo4c(O*nZZ5iH9u|OZ85j^}SsHqqI!+je212i*A<b2rF
zH1D}qsylM$U$wt7Bm32=lIfP^u9@7+l=~Os4l1vu#niy$13}QTuPoh)aa|?!%rjP#
z#6AQ`k_0yw$-e#7f6MkTvYbZK*EjI;$*YeHnxE~d(F^Q6*;gG#_$;mPQyNmr^#?7x
zj_~TIQRCuP$EX*;*IEZ=-iBKK=OOkz`F@YNDF^F5?Yq)C78$FMvc;rTCQgXDFs&RK
zf5LX9L08;x@4>`)_{3yv!rs;VLi0N5(lwKe@u=1Oq*%ktgnGAuBh}M~M^s-qK$iO>
z(%7F$y)%s8;B(ro&#^MZ@zK~g$D*LE&aQUEX2WaZj&jpP&J5*j^m6cOk8pEDW~2y$
zUFYaQHou}OY0o-GdDCGae6i;S5y8F)TzTai{9e7Ck4VM6wj0P!4SNaz4%DiX6v<n^
z?{71M!3QnjCcilZ*MV+$&6rM(PJ#a%>txX0G4Dc5rzji3SiJ7=$P=wI`8|~-Td$dL
zdAyMKZs?`2;7aHtpPH*7G12k9njEyF$m8km=TQ<}6mf#7UQjc{xA00gh*$9}x}@~M
zMZSuV0Gf_}Ib2kd9shF<w&}Z`b1rYNIAJL;{6MO#`_Ei7&e=c!zz!qldHAr`y0uW9
zv{t345&%SHypgOv+@mDW7?aN3K|j`)D~Gn1K*#ah*vuGPm2Oz1R)KDGz%#T@@9wM@
z#9u~}{hVB@4^SPsMs1SEoV67LGk@0uH#j4GVQLNQZSd4>?*Y(8+7c|j>qANjrlT@a
z-t+qLC0$1|@IVZ@o)&Kb+xUmz(s9^5tdt1qxBEy!gmQTcQv<2Y7|x)Wg$(tMf&NbO
zO$_CDfZ5qO%_DMvP*1Vtk_trO$#XlegnW3|Xb~Cd7jSu2@FSbG7(nXj1Kr<3vA&P^
z3=X<w&HU*(V7#CKqW%T5Lb=f7k>aLE?1)?fcw7H6mw>KYH-|-@StDiB6J}y#6BA@6
z@h7sDZMBwE&Wv4nf%yNL&G|Q*F~({Qe?t)Hy|Vp5ikSKyi%#mtv0sqkPugv3^3GYg
zMIZ{WmoGj$<>RePd?w!a+Zkh)mlpZQSxT~B+-70S-rMvTS9r6{E@e<;9^+D4nQ)Bj
zW@X-ZSp8{KWe0twX4te!i`-uh5S{fny3V~ViWuP_GETqJhJws-)eidw-?jNBQcAMC
z9(e<YLxKdk5)EouwXUQ<lrII0efGnz^v8FglA-z7kc9Zsc8n&W-Sb()>zalrf8TlT
zB2LMusK`7vX6ysCpyU3^TTfYHmO-0hQ!fbmNJpVTx<LzV)VsVj5l8R4$sd$@%VQ?`
zx5%*T|7o%ddWzU7KO{^@;pMDMJTeF%D<_yO57A8A(P;0(i{-Yn<2et>98XJP#`rDZ
zQ6mu>G2{<f>;G>|`~Ex8xY15U1Nq`8k)4!5v3QY{;vc8>vgx$+b(|QHivQ6u&~p93
zcdIa!Hv4va$n-iVgX|MeKxIX`KrzPfMJz&%y*@4SdzmI;=br-qm3VePZ-kAj!TmVC
zlg-sS>!-mf+2~hb7%%q4FL~mH5PtYW-51VX2JGRMv{6*mDrLiV5`%xi0EGLoDHIC;
zM1hfy@#U;s%tYi$U`BOXOeNj2@zt?tQOl3fg3L^%K$=fI_=~tXB}E*wt<BkPZ$Ts|
z@89zMcA=tVKDJO{s>@yf#n7X2|Msj&r9b$e6;2qb<TMkpaUC*|tTOApk*u%RYAOEQ
zmd$uaQam~<W+e=fp`A{x<;D9}S##leD-UJv=U{p4%0HBG!@Ru5-QK}DXO8?9YI_+U
zXPpX5#CS<gs^fT=XY=ZC1=WHIccVA8po+kt^eRnnio@`tyvEP^9{~KAVD$hae&npP
zMdb}Cjk^i2+9vtdd(DDNV@twqHj9cL8L=d-hPTC?GaVJp7r5muwa1v-SYxd)>!=%)
zl@E3PD}gVDdR>Fs?JMcb24)OlW|$i8VdQlzZgjk~v<798+YSFSi2K6jW)DEZlJ^mQ
zR-e;CDK<vMzAm(8cKiQcV^Ea4m6u)EA9K@D@t+Q4E&nSv7M*P@-$(dq^{o6v^$J5K
z4duC`Cvu-zEyBH1<W20p%4ZS%>y|=PrPfjVJ7K&l_c3Ij%A-X0ddnBN*4)C48ZM)|
zS%Q9uIa1@uEt0Q&`#x1gln1|CAiYlvge`r$BYC{namlrfv}g_I>w;PfKz_B&2nF+~
zkND?54p4%<cP<(u>-H$$rQ=n(g4TEf+zK`md=5mkW{;{}I6v-}hGgG4WBzhpY*89s
z{FajDd&`l@&=L6|>sN_V@7Pc_Ug&;j&`N5YTlU<^m00qP<|Q1Txs3JzL9dsm&+E&$
zfZ=g*$Ig+^Q{MHK;gP|<{^6Ql>Fz3iT<A1YKZgps!t#$H`_sj(&j-Poxh7%A8U6p{
z!3{f}jDPS)hCLxwBbcG$eksS(wDAX-o*-`0QdJ{3=PcjQ75hcbil^5>^1`HX=4#}X
zO0ygIOUx5msO93b$V)wxyhNpSAMs`Q(*@YB9y*@H8eqpq2Zf(5I=P|vR5Q_5BV+M&
zbX!?MN|Pg3ziIsqyz1)%^{YjLo@%VI?;L%W00EQtyM=%^TRu)<Lm*iD+#~3OY-AFS
z!yfn(b_?++5~DMS%3@9mBM~Z1Iw5vUaDopM_?IbYV&n`Z1%q0s{IX{!tCr(a|3S3L
zbEknwhVCRmMLi}w6O(|P(EgfBLxaUT1FfXr2ieNqE^pF~Y>j{JWT&oNBy`h4F&yg(
zIlTw311`Z&y{^!X-?tf~*SWoI^*^1ecjF)Yx-@JYresY`gIM;{_XvuXaBKy=yb^Qm
z=NHoR(Nngp4zo0@+C#^A-%TJOjccX%%_QM-uUzE$6Y_$J6r!#pef_xHf~coG%Po9=
z)9Ie7B@zAgt3z52q!V3AcEsTucsvi2odO{0>S{w_5(d>#0Wee;?-i)zakv0?u2obr
z6k@{+uynwDj;hReJVL0wgHVP@j<C%EO}l}EZ3k3}a8NMc8gCcn0><-13P_bVl2(cE
z>k#saC@udUywO{&eGE4a=5bygI5HK4C43uu|AUieRAd=_d0y~?-7Kus!{NxvHu^?F
zPziiCGHlH;Xqh~kZjg-W06lc4ug$+6+n*;=PR~#J^A8^Frh5^FUhG#MofWVwWE1`5
z1(n!f3ztf(9onG;qa5vYgE%)~X3s6Pww$B!i|)e@g++vp&Eoh{i^ol_1;=VVs_b6)
z`kiKB&=xr!+`Hc{`Xko6yOJ{QZF=glvz?UAVCayQ{O8oUcPlJojO8E@y>;RVg;jcV
zo89DRy<Zx(UKAgfe)c5Osm4g<y4oQ_Jiuep#fK>99AmXRyn|=(+d9Vl7cWybVo*E`
z0U-JJwxsfx3{%DZqLU}BKb^l%=`n%9x$ttM4}JN0b*6CxU96oW<tKe!|F*pJTj-H=
z>L!1Hl^(iRU$Bl}_LFgLuh1Z=FGL;nx_>lp>4^Hjk%UVNy9~PK0iPEA-BM#=U}`^O
zr0C?rvWHbn$QM1DR2`>dm2oug6zxM)w79VTg^Gn};68>e_J(834678=Uiv%=SCWJ7
zG*X#l=JUH$T=t|ATTr?s$h~W94+;mBZgja@kZ~=9;Q0B|?#%kG{Xx%dYj$%D+n9>}
zu3F@4pUP&}A=J8B-uTC*0H~yxUk&cJqn~k2L#9O5FhvC?VKBa7HO|jN1rj)NTWL;D
zb1(EI;TsMXd|Q~?&855_8<OOiP2EB?ZX!S9J)iD1Xd^8Vd@0F+-g`wdm7cVM0D#Sd
z!ZPH&&!d$e%lA)Ny}l}W_l@x_JEoF9ZMFHxB<iQG;h#}*7&C*moMyW356{R(9zYh3
zMVWw(R<`b3Ose(2eJMmEt51sH_Ky{_wDzbuwXE<L<hRN`<i@)M&pdL@X)r<}^<xuq
zk!l|l_+x?n9pF#vC=q{tH;fz~s`FuU`5f(q3!-W+=rF&B`cP2=sS>*trg^;fkUA}N
zb4MlXvE$zS)3B%~l&zkq8<ed`DTJ?3S2kpkMr!Z$dg>cj+2aP036|yg3Z6?|%H04D
z=-o<YuI8&>6t})N#m8FXZt|}=!Co7BJ>dL8NygsN=3hg?&<_ZAUJVO(t*>kbtX_ur
zNhR^*eTs=m!CdlOZirWbX|%uJ;2=Pxj_`Dh*u_Lb?K-5)#OHxWT{9i9J-gd->fNfc
zd&_n+&mGkhBEU<-PU)m;r~V6<x`f`|{&@)jPq<}bAHSTEF_W;k)JT=!WXEti{fZeg
z-0rBKi&HBuS6{B-6Aaqs_qIi$(-|fztC0pB&i!{&eEX7u;(PikBVXkCb)z4jK$=Ad
z`<>3tXPmo#E(0XAuPNf~#|3+dE4n%F4O;B!a(>Sa5+fJb!IZ5R`Yz+ykCFv&?A%4J
z9b<rqUpW?ah_EaF=TZ3!aXn)mm^nLS-8Lap7kHC$n;<+Yg-d%Y@gEp6Pj>}PBdr7A
zdxk8DHP_;UCKA0*7<$D5=d9^MWV2HW!c9U8Yp+Rq?(Vnz*}Mp72?eFrmdUB(!aLT?
z-5rXzdLe7(IbESB3Gc41r=se@BFwyws4gF0=I&h1-S*E2eLdFE;X>~|m8;;%rPW8g
zRD=;tE0jH@3$h>!l(%e-p0E<@^0DEg9qa2|4XA2Jc5qR1TifJUzL|G=U(B^YIVhV(
zi}@AfqRIb)w9z6b#XV8YVMoA}?V-uzrH)zmwTAa(>?hI9dEawibp7(%d1EJy?F#Sc
z!wr|!bkad{=acNdV^o&6ja!II#@zC6{)SX(t-a2Tw|%OCz2#B0s=x@WbLcfsXFo%#
zwbm9R<Lc2_;sxnbJp3MpjoXk0AoV{G;e{(7_|tIOl!Vi3ULxKHq0%^2kwNWm+Xuhp
zi+FA)3#}3)l6UVIH3dh=V*C1i8fzw7&nwWQlKea1f71}nylhn^OcSo7JcEdPyM+D!
z<$Q<LPsZ!-W19}LTcI^LSwD?itSIwaMoV9w5-(^Iv5otki^hy|>XmMcYu0O!W8&P5
zZj5s)u{`*Ch&1IR6|F0LJR3BW)5~OBCpqife}0%`KZ~zNryfnj9u-x(Km@hFB@LK~
zLygCTN$814%XCH>=)&sl`xg#TQ*bzm03mZuK3c%aImR_=Ct-Upd2_u2d48_e1-Eg~
zgbOt9B~WFp_obu}Q2Qv8Nn$^$H<7K0tkf@H>O6AS-{OQ}AWy*n)S?$Yr$j@Oe*SoS
z|I`Op?}mt0X<y5mJrRd6Y*Skr_6+{)%tOds(C3+vt?GfT>ze{e=rgTW=O;(!Ke1Z2
zt^KnQ0ejFJan+=HK0*5W$<yyIo+Zq#IKN8plAWrP(NnOM4gq%aK5HTOK<71l@HN@!
zVC5V5m2iDa7q!b|`utn2`<u?s6jsUn**RU#h*;Wl8(Nc3JW&vL`IV*CsS>G<m#b^r
zm_{!X;EeYoaTSJ_*=zVD??f@^&7LXzCD8T^_ZoGC!8#yQYeMZb$Cey_ocj*o1Y$iX
z+FaB*B7zLFTm`ofvE>Ra0@1Ru;RAN(k~-J2buDD0izA(s8o1Y<XKS%Y5R+~iE8@N{
zEp<_HJBhBowD=)KIayDyBgO_UyD0gT_EZ>yM`{U!&M?oZqZm7c21N0p_^Uoq#{u~S
z|G(+1`@iW-1_F=}0pEa^Aimx_9*06ntrBXBLVB!JWx0+059U*;jwGq0hL{xXxlda&
zJ+5L6pE*>r1^}<@GfTJFM?5X`M3092+*DX3Ywp7B#i>6Ilsap>*%f`Fux2Fb=7n3G
z4U)7|w9zr{14)!LZ>a;G#!3@?;W*GI7o*%^QNIwX4~i^<jn8Bd1BlgP4>d@_Kq$4q
z*T1^6jjJ}N=4&l#;xkV(#f!6y-V0cmOY@yd)8D!M{`Ku-t^2k)1<EPejO_gE=H}9x
zQPD`rhoa+FVJqWo5Qbu=P4>9!zXp%UjLHJx*j-&Hx``Jw1>@b1DqC(%Xi7?^>;JnT
zyov!X^9YMndmj@gN?5VcEJUqf#ve&1&-S}?esCg5pV-dk_!P$V{tX@pz52C@6MIS$
z9nRA_N4K>N8!_0{ums8hpUyiK!zO>^U$1O#{b)%P-;JCOdW3^Ik7in=koJ=0+bdb=
zo5>_3kuy-1ikt)aC3(x+UM3c){0_v^yA|<g;CEWZn+R;H{jf9PT_>N9a{Z)l$@2be
zX^#VBnitstfu!W39R$M3{9mvKz5v^v)X%)+1n=h|=>P~Y=nzaH^-Mp%KO5vsq|g9E
z%l>Q{8#&umI(pyb+?9||2*krjDEXYC?7Bzi-_cjR+|>)tdwBpM7PK3s&^|BOXR)?&
zfG`s?I&k%ByvXK1@TB}IyN-y^wQqo^Xj4bFB>t?K<UDMJ7XRbZpJql0=ZHUIpZ4(1
zYn{6CzAI`resXYJ>XXsOwq{0_0|Mf%=)h-(*kw-y)A*McvEu34ql`Nxa#+HiQ647l
zXTCELd${ZT<!k<@eW+>ePa^)hP0dFW#pH@;Es+}N6?VP>T;=Uo@%v!29gyUN12atT
zsk`%_r#_=Ppu8&7!-y5)eX9@Jrc7x`OcUSTipo^ab4|X1pP_FIvOBEw320+%v~UgX
z5{)_Up6yg(%JU+9ioq$P_h5s)Nu0DR#U)o~f9In@Bns8{>&nV`r#w%^rW|UE<Rx>w
zHUvP09+(adPyRp>-Y88<hnC!VueQ#j5BxJ{tx{FL4J}w^8Oys^YE%W}+r~O{(-VI5
z?N4lwPUb!xxY4hW5t>ZQjnd4C!>mDG2{&5JFT?49_}=(rF!u1s`E(8DABdf7=6Y=8
z#OPb38oQ+H{n^rD4F}lzuMEk(y{?Ja)8iz3!ya-j_7;e=b}+caRP9~uP<x-mUs7xS
zm=lJ-<YO05Z~Op_3ZtYTcQzhWzD*A><~l%<CWF>t-W@<d7cO7B)c>v`#+6&%J|>Dd
zHm-@}?GQdj94-yKkAKd_golSm&3q`7<LMX{Jv5IU?kda|C3m>Eh>6&d!CG?Kg*pQu
zF9_;*Dx6pk`Ti>EypuN9pZ8YkhJ%{@n}>i<NEjM9YO~VuFY5QtH81CpjW-;!(^|X^
ziq<d=W{C&E@_K`nPT7hZ6I$sWSm-ZEo=<VTPqf-r-0<B`;9v^ESNN41cuxJi+*~iC
z!%W|Y@vWA)@O8~6xMdzAwV?f6{-5V$k7lLMk^{xRsXQ3lu(ZecICppY3D)Dy&ieVu
z7^wrmz=2o;33q@(++k@K`ydMXiLu(S*pv4?oPm%Deh-ppdnmP!p+(QP)d>JXyF-Oh
z>j;y_#c36M=gsKRSB!~3QnjrgaPZ}w%n!mLoh+~bVWIssSX6N&=y|e8&t!Nw1YLb3
za$@=Y*RR<iNQW308BrB-9#*OVs&cTma4lqc)wQZ-x6oym&>$z9@JaqZ{pi*26}n4&
zYKQ%u>!<opEurHLl^36ds_v0+h5b7-fc_``OxAechuZP^%RPNp`?m6?hLb2`;^Pa4
z(m4rMWrL<WFEh(KXkp3p`7Cq}K%+x~;2WNp9#jdkpPg-~c<rN>xrTrJrQ`iPF@P&X
zngFNR>Yt5DN$ERFk@uDQ%l1kUXyq!!i~e=GxUN2PAF+rIivn%gA_<QJLAM_=FEDDA
zTx5zpVYdls7sa5pI3Q|5*5e9#6_C=nF1@N0ePcbSKrHz*yZq@6IoB}$+RD5ABcVQp
zBw;UNVSp@))oUrk(s74U4U+wu{0Br4n=rA+j0y3H1AlmyR%8Fuh=S4a`N|Y*(`i^5
z)*VefKA+YQd+F;!!VnpVEgUrXOHaR+hS0~GUTM@(t(hYV{K*@X`cYL7?$c&P&@Ap#
zf3|Gy@F9w&=ouoGtDRXMdou1|fzyc{qb^x{(P~FC#s?xCC4-KTc{;(4Zh1K+i3~}u
zuJpE4z$otXbh&_uxL682nH}p)1Rn8{bj*DNjrz?)g{kATtuiqKU#+7I^c&UkA<{9t
z>KBP4*O1G!O!G+vz+(WLt5D$VWI<@3m5fc8g;y#CmXv(>OnPk|O&IKA*VG6*?re*z
zyMZ^#pd`4e4dqbVC!^Hs`Z&FQt5EYa2jnqSRD)DqR{T$m4v>fap}X`&zhf+|*MbX6
z?3WFGDUwNA8%yo)#_O_&tF+G2vWju?pa6&r*e?f7jF{R&i!Kvq)kSC{necZsDbl}k
z()5a5PSW8wFUXd2>yUMa$?_5`=Y-fYI^6%f&q;giX^vKk-TALW9zYCh%ttJ?qC7AJ
zN0A?5VXsQFMB6hC6+S?TSHtQ-u&Br>_U(%M^&s;MkXO<gIBB=8bM^lUm^3E7GiV!Y
zgN1cKhVMB2x6~UL^%YXf-l<GttrnCIdtP)@k*Anz*k5}%a`1;kqvfHnf$<dk`Xa^E
zKMtPk`;de`L^4<p6>>3V%@qrJ=@Sup-l<BRO7X>uM*q?UuabWx2$Zf@-EVqS&W}Hx
zVBgmow^F?|U;}Su`Bnh4VR7!GRkSRse<rXz>809z;uSjxffip_4{>OJg{zA~CQZ}X
z=ohi-b>gdcL3VP4C0Ll<2v!RWInK2@;+1P`0?TvN&fE_IiVu+Hj6U9MO9zR&(KKc|
zyhZ`ps9MkuiaK<lV&I&}%Tc5z!U6QY321tIZUDud_$Ds)pb#9KY?e%OnxRpyBYXqh
z3CLyi8W=7N@Ly0{n12X*woC>POC8T<vm`B&Wp=)3|9+!~X3=qegYC;~RC4&001xlB
zGFiZD%w0VQ+gjVq?7NAOo#+qI)w5`U>OVQ>4mHo+K?nXfO&k5e-=LWzeoH)S)G<$$
zP25izPBD8iznp93H7a{6S7lfF+xEqzvtv#kJK{?+C^ZcoN%&CT;7~nCD2Iuy{ySSs
zsgbX{-qD6fD6>B~<|e=TMJ|_mU_neQhUIZue?Y+o?w4Qk_qP(a^Yzv+|A5DAf#(B6
z@C|9*-(>Z*@7C}~yc-m=(5J{oL$cP6g=>|jIIADZCnJ9BNu=L-$ypCkTdDWB>!9dg
zaZ^<W-Ap^6C1x)lQfdjX{428YFM9l`1DTBH8PpI;zmJbswYgcO`2AN_=8MjUKYw~2
zK%S7rRX1jOrpZC>E_8P~-W@OeJ*f|@5CZDnEwXjGmRo6mHwPt=>Hk3`G_61W5eK`8
z;F{}S>y7zXjv>i@TfBHp>$gQIEj4xJr}9*Dti4P<b3B&FqF>_wDF57<+K||KDBYd<
zoiWacSX-uvTv9()RJ&Qw%notozCK?@3x9Jlhn1N^dR=o|=MgTI^*Bj8>?0r|+(Wi-
z-<A7LmyI8$q@ktJYR#l?y&y~TxX%?GqTayJm$&fzn=WwUzArfqKkZe2yhQ^UZV3*$
z;l{N2({qd3$hrE{`C5|LeHzMQ_BXStkfMjj!X-`JkFG_8Omf-Z*eQ{XvwIY7ehj&R
zheL`e#ddf-{1B<07pDIoexX9{SFGA=1pMY{*d}m`xa`pK#FfOBm9JIk`;U%K7aTZi
zm3~ZO5KgFVTBvxo&?n`W@vOOh5ix2Ob|8$Ds?^If3fJ-9_Mg;CBq(PPBJUX{u}_g6
zc1iy{2@mrB=$hSL6rdkUc+iGRv49*TTFG`<X?~G^D}7MpikL(2{ub(K_}y1w4{3nx
z^F1c&95p$s+AO%pOQeTrbB*4PeM4U9Uk|Mq@7T!_WFu5M1!DdsB;06(`PA$0Xyn~U
zssr*zUqBSkPw?fcO3I50^Yb3QzP2*{zH4?cY00c@XvUCCE6nk$ssv*OZBcRoSN61q
z1ln+A>lOZ!Zm*xvUAjO6cmy)es#Hbr=z;WBzvaOn)T+kP^-BpF$oT>y>joj=Goi6_
zvU4)BK9aYZzZ}_<`q?4K5p|ZGm`u-}a^65&$*B84>69>f%!v$_O4GfJUkJhQ^UscE
zj+os2s;3<LEq${SxV{|my$rt>cMGA?Hr5)Z${Mo<tV*IwrA`v3Gi+pn0efjYgY*5!
zq13K#om&Ne<Rb7FaFg&j_rJv(FWD!xL>?C-bqjiJyc?ZymZ$Vo%R!kIaQ`SN<=eMA
zcUWX4qh(1C&Fa`8<FGK86KO^H6KQl}Z|i)mDvGIrx>oO_=sy);!#t2Qq1XofTeiK*
zqD&mKZ&F32Y#klHn_B!gzFOqd{h+{P@@2E^o9tmH*F<*tyD0SYgZY#~1?@)7upcGg
z*H^kML>a*`v2y47wsS<`F&i;eMeb&sF;$6m3cPPLr6R@==ztt><zu+U=LXnU8zM?e
zyG|Ab|LC9(r;oc!b)pCd4v8va&cGci7H?8eDL|g?$e%l~qqwt9m+Q5&^{Iw*Ztf7(
zu`<8+qC38V)LId}SCsnAmm-I~)r?eWWR@?Sp?DL;E_7dg(UeH1o7pt4{PBFLKB3&;
z!{>YpE3>$X<rJ5lx#g!$etLa8Jnr~U=|VppD(01md)*eAteFQ=C$#U0apgh)ozU+5
zJ4CFSJfw(KOU3dBYbjd;7{gzbtd!DZnf<XfQ_X=5n$FUkd#^Q}RoZzjaz@qCg$UVd
zKDTCK10;-A)4oNoH((DZ=#@=-E!EGzm=oPJMrMy>F{wuru7tQUi;54EY-c4cdDosC
zH;gStD?vJ!QqF(!%n<uh(0LqSlQ8f|Vl22|EDfj_S6=4l!C+P{$}N(soWzQNGu%(|
za7<IHliyQMK-BfR1ODkET&-h`;NXMER(2P~yG-1l6Ohv$b2e9O9~T{mT@+ApJU4IC
zI7!btem75Pc#=Z!w_)OsGg0cOYw<t#2$3oE1;@BDm?JBCbbzb*ZzScqAtEboIu|RZ
zS}7JC5kbmo4*{_{H~3vGKD<WuP4mT53(fu<$9P?5pQ9d2j^Vj$_+>|mfuw&_wGG~<
z*KmGe9iQ473n*O^OqDG(5bix!i?!iC{EPPeDOQCwhA=)-**F|_q~p1f`{fG0kwe$M
zXpPDYJjtl#K-F%tTAe&lIfa)gut>(pD=9Tosv={gbUn#cnLgGz#q(||_GnyPkk?Ap
z=4;&x5+7IoQ*jNSizAKH`J2Ay$efONKP%}vc@9aY(LUurD#)t|n;2)`@5n7j*GBjc
z3pm9Aq@#6fq2`0k?#p=2P5hyBswbcY0)X0ti@sta2MyqE|9h@x)nf<!cFSubXaJ(f
z+Is%7HvRdbgYnTy!kt<1NLyiIKw;X0axT4v6+zjLOk@d!D@>nC7d;55<o<6pCS2i~
z`tXu`fv|`<UB88k$wudcVteHe@00*jYxc?Yi<oCRG_J!p4(4Rb^HXQ*RdI?~IXP%j
zb1M%krB4I*(0Q3stmv@#mYj-?B;!_l#Pz=`x0xQ}qNPMUv09hAmKGHRZw%bYl3akv
z13GQOnTq0bZiM8pJj2hy!8{3BPp@0fB*drs+Wh;`<Ulv?f<}wFoU#3k_{L=4u@Iv+
z#oVA+?+1M!ZsPe#_!qiN{c8rv0sql-T_^wv<pTn6Rbq(JXb8fn(U5F$;gRZcmmWGP
zaeV`M@Vjz;o7}Kut5xu7{?@5ZSxOdlF!2ec&JE}fuPT<af_2mx<Z@F}6{oD%&~Xj)
z#ng^pemjbTh_p8Q9d+(Vh1MMfLx%T^!7bClV;@ccUHRVbPBMQ8c5MsS+f_^S7OBFl
zy$CwHC<51@wYC-Zl0k-a?ir8#n0e<DI@tTp9vTh{Zt#g~%+T&?JB#f9geWR#3=fjD
ze=QYd<{I_q7ZugVf1CY7GYEFG-~acGXOmqj+%xc1;$FX91tFTZ>kF)5uM_&$AeZ!5
zR@47(^+Rt%@@RcM6)?KwP!=2(zN{ww@Hu0l@uz3Rg{fTajs`>Ie2(o{dUVu)LwVE@
zAe40P4D#l<fO-=@@{$p0J8>}GAT}XDg?WjdZ^`EQPuy)+@}CGlJooVAdDYh1)9}Mi
zUFRidGXnQWX9MEWUgxDa86;_?9Inz(9(`1gyvG2*Vq!{PFuocTn)dxI5)#6S$E?wb
zi2Q}T#kX$^w~Z-KKqyT~TH{r?lxF6n$dTK^L@8m|BZttzI-;NxIWS%h9G*k_(4TkQ
zDsu8u{t>l=tI_e3A3UeoDF>eR9E>tZzF=loi3KQ1*vdJ<9Y<~>e7fK-FsPMw_doCA
zi>i87+|QeQ2*KR_e(cg=QNr9+^J%WjkY<1MLgCLZpQn<~_HWC`wS`C<&A^*0{P0p`
z1LVFi30SH_mnVYkV2!+$g^#8~Z#}|2N3FD${-m-5f$IQVkvC>_OsTjee>{}?l_Poi
znv}}&F}|RItk)1J`2n8vZsnr1H}TbCE&kGbO8bHA)^bVshzosjHO2e3@h|v>y`ssw
zi`={P#c=lVMmM3VKwu3wY76LcEP(Qla+Q=64bgRnqX)ZKnSmkRby2aPuHvvI0Tw^w
z82yjODj52=4dk#Lega5=5B(H}cwm}nPjYpcLTbcuAM-}>TG`_@|MU#&0?(xr<}V{w
zjrj2o-geP@+obDqj<NXh&hW5<Qb3n5Il*MUwcqES{A9g){xw&v*i+48715q{aGvP0
zXpM4PPQwe;HT>zC$)$7GiYEN}L|By2erK=Z3D3EGL`?qXV&@`f$+sH3W6Se?YUh|b
zJRbo!dROMye<e*!ji(GDW?Capi$cT+PurSl_o)`_c5;g6!~wjb{(B}DoXLST!tc@z
zlf)nKsy(&Y$enG>2DO49jpPDyN3x5Y$1pI$!~DG0S0}sE_J?{kGQTv)(U;Pb@)!rM
z8GzA4R0cZ+Og}`w?gO%oA!_4Z6ll$a|8a&3RRGSIHtdLLcIO+2{!dxo9?x|9|L?xL
zyWO8V(&<JyCaIjuDdo_CgpeF(SlARBlJjA6mrBmenNU)~SSEAcLWnt(vYCaD(`Mwb
zVNSoR@9+2cJ-&SY-Q%(CvG;Xduj}=Co<?xkE;T}Y?-q$_$NbFkq3vgHzkttZRa!eA
z=-@SPlvk_vOIZGoQUhRwp^iyI$2zzqX@+D6Z@MaXE9WdevnU}{o;pmsj6*w2E>ftS
zICyb1{G{1L$F=;1VbKh`FTO`@(0y@)+s2ydrw>&gKsMVqin3^kG_3tdn=b~9qro|c
zZ-_G-f;B#aVO0n|^fwOj5Zt-s9-~UV=Yr`7MuCH0nbSvL73o$?bH~b36exB@_#2pT
z-b*3<#_+js%OH-fad_bC!P0GE;g+vpGno6L=oKXzgTJKb7p5@py{d-_>~~FK!Iutn
z>TOh@YEDDr+&81{C&S#POY@~9x5wujzO9T^&cE3j)G{mvpx_?8#czwXa}zNrV`=D?
zc2J!R^k82gG?{a!WJ`1KmeKlq9U88JH<j(#AMh^FjJi>VD2@Yc=&)!y4eP5d+X4&;
zYQN#2aRH?jmOyymC5jA$FRhoop<*(n-tL2Kx}p>^y(D^DXwuj!-?Bt_OaB1X?~CGR
zpCL@GiSY?jM2KGi+S_HTvwH?Pbw}Ck898)etSZ`cYCiJ`wcN(&pR=A2aj1Z6jzmvT
zj8~Xb0z1G^b0fNPS4Syy^a|ZbB~$4Mp>^F@ePBd%i(B6^9>E(5TF(nxjDj>S1Q=sq
z7JMD<VG-&ZW$H~I;Jmb|bxhgnN&VYCl#$11U9tK<#fq_YEU(b4D#J>yL)xHRW%BF3
zj2Lkbflje*U+wgSkC&`d)NPf+`}13rKQN^Lqts8S==<B1^AKp?Krha7INUS|AgoMa
z?(1(V9dPp)D;OZ21WX<8uBLrjKg`OV+$L*=0L-PXF_TrQKy$~N=j$!a809}({TUhk
z>hT(<c^*Q4h_X2ibxbfL9{uRPa?g+mG)sZ{4gXK7<fH#lpZdR#KHpofM})*c#Zq}!
z$mUgxKZk=~`25&h$&(D7PtzJ}bQ?(~4Xzk=SWXDv3_@S#O8!A;EC=VkwAqi5ClbCn
zy6|(pRk?>bpGKhNSj|z(I(YN1_FLva<j0N8^k7mcv=m?M=YPs-=>G$3q|NqXKx{)*
zX)n=9>St9}#B_&jr@V8)p?xldYs`1z+p~Mwu8?helqxds&O>Lh4mmI2nMqJWr*%Zw
ziHXJMRQu%RXV>_<j(sim1LIUr&$mox<Gw;=8Q-55?c(f83uvy=VAlc&-Nv8crAz#B
z_KbMMcVYPMK41#;0gd*y@>l4p>D@GhA>3`5pL~`$b!r6D7<hM^X7B@vE%dp5L|3Hi
zHTOUV_s~agU52(NC$T>Pm@QW9YGqip<mfR4W~;5Aj)nA)TMDGE65{^5L%keQ-gZex
z0`Qmgsbi6gkKu{plxv0hF<p{nPBENPLZAlKpwcAEo(hu0D{%21Xjw5p;1U6wW+C+f
zzvXbadoWl_z^%Ik7uGAcWs`+FB?u|@LE*~EdcM`3^>ZKl(P5#tHPNfw(>(+U^hM2L
z@(X2d3tpuW{vs-7vo(^~ClfgBo9ha(Zc-nC;rPOWkou1G{w4@E`$y228MY=L3m$*`
zX1jIj&ub(OH+EA+XX#=eZDr4^A4^z76}8TDbSa=ADA5)W(zHc8#$tKhUP7LvrBdzn
z?XpZzl-tVMvsI~*%linEFY5Da7?IuC;UsNkJbe6|GWxaQKNBbHMLM%j(rwj|dW-8k
zP0gvz3%Cf_4DgAdM~o_(2smhB6|MfZE#T*wRL6{IWZLx;UhaNNODT48Dwy!OAJJdF
zh^P!_=I1(%41Bx$2iXdd?0Gg=37X3GCs`d~%~LdO;X@WE%6-)8g+_zj1ljwmPi}HQ
zCn7e>V4DIb=&E`B<iBzv=P};*M#!s8rkh52q0m;Q#6eE_htFaKU*FWqvd$)TGrmTh
zPU^>tZt1?u%QGr|?X58qss2%g?Fn?^PH(fNde3>d7Yh{1sAydKv9F^>JpEv_4cygQ
zjLcL8BM{SVi=-`LTxfU4O{_cHiGkx_{HHx1+WR%YjWvCJae6-KZftG8VI>OhtzATJ
zyn5PNd>ZHRMC;6BIz2io_VAsa{8jka*kixyAv+tOe|l0eT0uw~xTgeq6kEdz#BuK;
zV8DV9zl5SxEN27v01>Ns`|N+VF7J@A{_A9TdOf&|Vqz%EAgVHgeiWmAR=I7Sn2%Vv
zx4m$Od`wUXV<E%I!_Gd3_XWX3;7c?X#%ZRWc?KJR&k!kRC*!m|nbr|b3Rs(_Y@Cuc
zJH8A8EV-Sxv<GH0XgTAH_ci6*`c$&kz1!U{+ZMgOWuF9mXMs+dnhJ!ip_-rl6Iq(r
zs_F|d8ZB=x#A*OwH9;d87lz71f}ICbkuftysCk&EnPID{`g6X}%JXXOEq*v9Vv*Yr
z8s_dE;O5qNG@KQ4+4HO?`yGIXy{mtxJ70m!jAr_upZg~<AlB`o#Hn3#U)ZCY+=XaF
z_lssD(N19+B4tY?FM4HbW?n~awQV2%n%h+$c}rO8`B12}{{ZQ&pjwMvb$h)LA+5r5
zv>@-w;0A{%_EFf`a$D&LfKKi~^*pXZ5chA?%P}0%tJ+gzZ*j6`?;fKl#9M|a9nCm}
z_=M|V-n0ew0HeYddvs}m?Ga=oB*?k_-C8J07}v>^rs~`WX#@XioW+GblTt$xG_e}<
z+E$Yx(S?LtBX@h0l$^<aVcb4;aq&_{@3Y46g|6Zpg;E8v$4~3z;cVeH?y--;FuXhl
zfFLFv3u6@}uLkNNN$llBj~VS<mv&7xsAwYr^s#0Xy$&TjFJ={qmxzrbb06=yI!{`(
zZMd^6CVvZdB>994!6H?F=p38@j<pOe&ki0Y0H?rI6t@3H#WUxM$^e9AMm?Dgcqq<;
z*U{F8Lx|e4^BUaFCHhl=H`#?e-8K`M&&P3@6ZVzT$O1^cJTe2$A)r$E@oP7M#llRr
zG{e<Nm4#n&K}mKD22;0-PeJ6xjIF3SoHpLzU^heSPLG9CLom0wIOQ<#oMgcq9PZZO
zKW*n70mE^UM%-;m^FW~t&>*40Ao1)Y9VQv7MCcJHQI5fz*rf>gH){Wj*>N#SAT*P+
zQf0vDhtEqz+?oIA)8P9qaP>W<#5T;#m6O1X9kDcFr`({&EP4M9S<;&aP<s9_H|Wz~
zQI(}<aD3SF`J%wI{$VZtAv9V&O_l;zwOsUcvmQ3)c3eK1s%|?8eWms5o<8S-dMjTl
zBl3|Mn|iu&<CsAOS84h$pe$lu3^b%VyQRYe&LjWyw(7vzS|jEA71R28?rxY6UmmqA
zYoFFBmHhk*L0g?az}e177uFimlHQ<a7~&8M$OX-Ya+Gc$(6!o#Lj#64Rf&z#&|h{+
zr@8+cT7J89dOfq)K7#|oAt;`Z5jB4J_bRtYw|Vl)>~{sDk(g^wNQwn|Dp;UA54~pK
zx#sWjvw8Go-@gH-(}Ql(P~JC55s#%dYWv%=OH?(Gr6xaf%p_u*`paAn4X@gZl#hlV
zD$3Gfh^A_RE-f#Ca_I<N17hQtSk+!|8)A#^%BI~gJojoOa=+^<ucOp&S}$-I5iZFo
zvkCx0&47UOu%f11^FyOIsm<NcDdjwE&P<0;1bT+bnVG2Sw`r7Cn7iXeep6P82UD7j
zER+bGpAS&JN>rE|Y5n9pIpb3L-klz!V1t95daval0)sul9d6T}xY?dp9*<cXYK+9|
zZ{OHnipooj;vcd0-N-p)<n#1nO~_{@GIXNC)8e=n1lT@fX4CCZ@t+(f(cYU4uaa@3
z&h%orYZ|bZ9=9rqcLUW-;j+;`hT5D%bcYni`%WF2xifIyR;Z16g<9UhMANTZ%Z30Y
zkqeuZZIPlvVG7%cyyIpLCu&UHYE&0<#w+w&Q*+|s7>}(%4vW_7NY)9Nnfx$O3d-3t
zh@<nj1Ig!C*M!HvxE^63r5PUJAD+lo2Y&S@R^nk=fGe)}Rl_%W6C-Z#q|ek0POpn9
zL?d2ZRfG~%a}3;C=s(5U`MDcSv;+anXse1*OXR#Uwn5Q@_YjQj#h2CSk1uJ~R0C<D
zz9Z+<cbS0=QmtU(dcUpM>+HLlRY-zycBfj$^9^@2S-kgJcaT!$mkMuPTYMWcXc`!1
zvADJ5cg^-;0N$~$6k{dk^cG98p!IUBi_Tsrfaa+-8AoGPXH3mGHr(gU8j)9X$a(?n
z&L6H=S<4OpMLCu^V1W;qs16k#41~t9h1&!7xm6!-b+140(N=PIz2@KZV%Dwv044kC
z<GY$ngbR`TAy}PP7R23;-dfuenSXIH+-o}zZwzBIObqcR>70od#S*uM7SB+w(y(C@
zY~&OTDfPmu-p;X+GqpsH?dq*}=f;{vUZN^?GTq8+Z=h?kJAw~UDBYGJYIFtuLf9eQ
z<3QlHq^p$-kl26~Cv{}H3;Gdj*hsSN&%|z^=FSXA#_jjFI*BdJQY=#*dX`7x&CFQn
zul?Y?ik^Ye%9Vy{)#{m~4>fOrW@*)dD9xC`;P(+4|ES!`x%p2|dTxf-w~UzEM8?9;
z4Qk{@<lj45FQFq>Aj@^i&^UN(HBY4Yj7<D!khA2sZHtJX%umR6Qi<n64j4AH4~968
zO`lnB#w1s*Tnb#V1y*;KR2^VXk@WNZ65al?$Pqt<=K<|~)aKHy>;A?$?)MwCW#-ja
zK8~X7J>>&OE{lA~cCeiVmh$9)k;{#erLeCNl8i29yE~Nns7^1vf4d^2?7EsMqPRA>
z9Ea~#>{YqZWA3CIG$ukmzrccR&X%lWdEi@N#4kYhC<_&Uw^V`Qan>9l=<)8c$5GY)
z&^L@657mQYpR$d1>@F}`^2Q^=v~bsxeU+`}i6|FaQ=X7iTombE_`PY42sbf&f>VvO
zh>I|4{cLxVlHhK9;kL56<=wRcq<zhtxilm6Yhc$OcxY{teWfVFQ^TvUr0c2^?iXME
zg^8d&kv*4Wu;}fEZ^+O&-g3xvY|*-2PAo@Nb8!@>dza?InL2m4s=U5`yCV_;n$A?S
z?l3B=;`KGqOKWM0csr37W};ycKPf3vn0vmMz(BMMrhJst)C$y|#UOjab_gGuOX#)a
ztOfCoJI#qC{*!2$vF=4a^ooCt|HzK@l)U?hFfdbxDoTS~jFm6A+6^C(!XUTvRsifC
zcyE3o*#F1LjvqGv82MZ;%v&)nxu8B|EM0#xxeQ1=#|M?j>m~oROll)uRNioEW!m|g
zsh94tt^ZxDNQOV<w~*>uJ7=@@8-L_XbL=;BlRp{~Oj4cG`>SvvXL6f(7E{dwxVN$_
zVPsl}AVU<d!8%8-Q%8q!skJk)8Rjr;Kku<Gux%!fH_3bbbUFIE(yFbsQf=Vh@D^1f
z?GiDgb)Z*1X=v^tYQlF{D-nGDK<AkEL;9Pjbq9wqSDpBL(cU)l0+*x<ZWwp=cM`ss
zv`WDuRE)FPRx&MF<@!!926eGYLU6`>x*`ZdiS_jk*4eK7^ck0EH{@G=rIhBxt|~5s
zxa)DXkc%dM*3QQ56aP7?T8ih^zmbhW{U{-qaj)w4t8@(ZBEkl!Slo64mr+ot94#xK
z(tXX%Wt7Xd5;4zPo#@?Ss#rLZf?f2wZ#9au^k?^_N37n|IP<97hE?IwQI$Jqx3}Vd
z^Y0p~{D*fgnH*R$V^uICwy0``&-%1RWqrPsKe!%=PZT2EUw;2RxoAH}w`x12e`KYO
z8A~qoT!0o7PI_9_t3$b%7^8m8J&^4{$8ZN7sVy@w0|_9k(7EVxSV+H=S@r34S3i5w
zx9}@@9Zw8o{9eVw(c)rz=SR@T>QAvALRLb$qSiJv-`Q9ya@3P+Ij^#Es@=_Dx)Gxv
z`YWeS$@SzmqJd(vZr!=PsaJZWp6&A6FU|*XnW*@%F#{M$5A41cFNs}&%*v)~M41%O
zOKuZa&!=72Sl>TKxkg^(O%Kmw)>hvyMUjpO356S1zs9=rr(%l?ie1+G^s<NIFKH%u
z>G~{LYoUO=c*~y2GgXsi{Z6yScf3CL?Aqd5TjsCx<yK;G%tWPkFvyF9Ho<q6NIwxA
zc?wIZI$j=<eyi`sVMzBey8Fws)x>xHzC6Ia>u;=C=<6<w@~h>DiSeSVq?5cNn#x{y
z;(Y;8p%ms_eV&U^2FXNkcWD{<{sJ<aCfpxqQNcR$id{UvF}Kd-YUg93>w`IA{0VxE
z<Zb$?{IYmKvaldG!e(sM%<}%tb2DnAZ`N$Qy1R(vWakqQ@??qI*K-&T;y<0tUN4&4
z&{^~;4yCfV&_iys7~9hEZ^N*7hDhJp;xh<&OGf*k3d4w?T(8j!(h1he>u9~H*T+cT
z(z>fIR)d_WXW5yn%bi(jB1JuU^%@!hA>HWXRo}zpo4`G1D>zOqQ6-9Yr|%`Ut~cU3
zhP&^hssXt8KlthhCMiO6=S7};>2&PSU+y*?Ui}1=_5H3784CmVG<n0NSHsBb6D^S{
zdg~54=ayqiT6<&Ow3rm1IFyxmyt;#VsND<j%a)@nq_nrhEA((1s?z((7dS{HC(+6n
zBz4K57v23FoxKt$Z<dWq>|C6hOA0<VU{=p~iL+ov^yjuxb}|%WU8i#Vh$9SAs}qEt
z{JNH?r?)Vwf(hjGu%Uu0=-45nqMCQ2)pC#5TYpvs;}i)>`&{aEmGfB3m=-_83J9%`
zjG_t8+c_QHa9rg-xbr-nd_<_pZ$kOH08(XX+V}hPmxTd4XDNx$nc9V|qLq18Sm<`e
zUHJF-H{^M<=mL>6?52$8^>Z^?rj?H;E39No2Tg2(6n_z~;QlKl4f(&og=7-<R(hmo
z*GJ_<E~d5Vg&aScasz>#7O{~@#v_(gEDVG?vS~DAl?g~;Kh~FOaIW}NNApJC#>@;Q
zLqA)`Z*rq*pNJ2PbPp5ql9W}m7b5ZTN%N<7p2kGc25y~nj<?#Ap^j{zo4&JCz{U5L
z@fTv@xnkFo6|>Vv1-W;sHp9NI(uTiPPE`-<^=&U7?rneB(ryhK@x!|Jo<7-YdZbZW
zX3qsJ4T3TVLV13|X8U|_iJ(37I0_S84(;{>H!udZm4Z0;tZfmLhGFm7v{RM8c`#qu
zFI_R|O2_qC7#kwEPRuEKJfvQ_K2HPjs@4}7fZ+6vbwi85#=E$=&>Mk4crNyVQ90aR
zT!a)i=IJ@;CM(uff#^L`Kce&mxPNjYGQ#ZRmG;4a_59FIjWFMRus#`0IPG`cw>2Fz
z`@2-8KTRcY1<9ePH`|}KMz^l=a|I;y!wb0|@>C4=pkdab2wONzF3UL!#Ky_w8+rrh
zj;N*pa#Pvsp5kAK?vba?4>S^;P9y(xw<-p>zK6g^WVfYuUfjojKVan>kAP`#6IrKe
z*A--R#=}I;%$W0V_S@p(7;p}hWBj+iSXfx(YG3r*WPas!GGky#f@Gr?C5E&T8{m9n
zoWX)om$y#nzHVCwA{k^PTnC)~10U*5i`dw)>%Jc}Ivu-StIC>RsZ(W1nTmRpI6Pyo
zP1c{V2%lHz6_n85xaEcV@swFAVItN&Wl%h2b423ctGu5oN&Ig?g2l#f?}oE$m3x&V
z4Iv_(Pe9+0+G&5NLxFR4E!^z}*O&}V)PO}Wkj&nf49Mn;rfhP5AtubrZL_&A9x7##
ztLL}a0^2|_$JQz--rB@T?@iipXe@hV-wX%>Ysc=VRF0yHpSsQmUZ#=*RY2ut5atb!
z?o93D<9kF{z44G<@|a$7lNcQTA4UHCQgQtG5wG<nC1Tcv^tBtNSU%+Q_wH^(Eh9#z
zY?Lq%gS?(T|LX+Q&1iwiQIGUmK|1#tLA0}^xd65GNIm6i8AWesIzx*E(3x{`8kX1+
zE-kh-fuU(FXN*IBMg_QealS53C`7D`szhupOjs5900i|7zP!v`IU!y_7~T>D6)p>m
zK~ShJlsuZ!lST%R9+XzBfXEbmqpgY^oYBASan9v`%w?&X3uHso>KtmMuDu13LN;YR
z-sZh>IbSMEy_EvI$NQGY^6=yUukSybggnPj@Q>@BurS(08>QXqH5HdgJAI)b^;q>m
zoYd2yL33fdJFSiLAl<EX`nE7g%CO+VH|4PX$O=p&m1K;_V_3@|XC`MCI|s9BbF?gm
z>QqC<#)0#zuh=SIv4-5F7nEc4{_#zsRdBYoK_#Lc011Srfm*`wixLX!*@GA_`<;jp
z-|U~lhz04gE-pb88sKR9@&QR*QtF^h>8q<M;*#fow)lukY<)}a?Qb)3PT=QHjK#Xq
z-Q6P6<z8ZoEiFB7;A;loYaCAF*_kX|0;uj}O{1Kd7a~l0CZ)IE=*Prl=4@KFtRA-8
zUp&Y7gmA~$T@9q%)cg&LZH~$oXwKLw5xnXzOO=%#SoHu==sxiL=MCS;>nO!Kwd{o~
z*`z)jF&7BK@1dJlBVZ9nceQDkPF{iNTnAa1BP=6`0Yf-+zmDHq&BeBb<fc3M$I9Xp
z1%$-)RzG~NxZ2x#;;PD<^Pv#W1&XB-#`)Y?(@Jce*?_v2+3@5GlqXER8-Cj6KJKh+
zzE51o3-~2Z`TMbiWO(vNcdDbKGjwx;gR~-pjF2C6<W$edd2qY$D@Ma2m6PI4Nm~Ua
zr>nl0pj4)U6cUv7!Cz4}e-eY8FF1m1RG<fazT+O``~49iIO!n(*E*v7?K70xDo2s4
z*VPXuyQ_b~w~={mtA|HI=70XA7&*@}E(YpIN}iW7(M{3F<|n8#6tHX?nf27u67Niy
zzdLj+cjB&}=!U}%QA9M`f5f+C#=$i_46;yfZZIv0nb9{;T`C)k%~AOqobB~H+jGmy
z9Au8nIa!%aaDjuUIwO+Htb(pViE%~}Q9hh~<OBB@#U!ddcP?yKe8RtBF*k*Uc%7)=
z`k}}4v84}D&(ZuVtu5_BLR_{4E3y3=Z#^I53Sadz5)#yD+-?oN``ynH#y(lx@0@g9
zR{=sT$`0M9nP!0Rd@a*U<^Pe(lhjTZFcmi<95wbUBQ3;cA!%DJMN%7!w8KT_-R3Y-
zLq$VFp+2k|ePqORl@qch)Q1|FQ*|$Mh)I)p#yZjlWMVDjE3!!%r&f&wb`-{&4{i!4
zUQty5hrko9sWZ3d^M|N{78R+Z-&W$Qh4z#Ef7}~6JRVi>qN!-CiFE4agRP>1t-&uI
z;=0Az^iNFzX$`leBy_Oxvn;!0b=gKj-Mft-Efq%|y?1*|KX2`OWd4XRt0FX+i2q57
zsm^`)*<8xzI6(bc{D#KtBkFU{6EpVG$7t;jccc#SeJ9DJwzi0O@fhZpQAOjH3srw6
zO2kCLE}T36^9QLZIQXrwkkEc=o~Zs!h89LvQc|a`+GjP|?(1-)5#mSm@#21}SWb$S
zoyvAMerrrcFORqUmD$|f081=$s4nX}qr*9i3`S|`T>y5*<K7@QFl;}_!=R9c%|{Z;
zz(d-p!TodsoE%gW3Uj_jsqM{>UO|kT<<@^rY-*<+I?2N*%E#>2S^N54%!kytlB<GQ
zd)&Cr)pzpws#Wfs^=xBldwn~zA#|qi2{f6l0$MHS*m<&|a@pTDrE@EFTZ2joY~fB}
zSpBRch~`s~C|9`ni*C2yc?^igDzM*$-l--D6wno)nr))be6v;R<gr>3sVsO_Csr{L
zs+Rnk1j)}DdN!vI%ZLy43-(9VIgi}%cFcFKI~xMYpZANf(7RUlH=^|6tq7=a{*$Q~
zTEGQj?r9>8rcysL(-_FBqx4iu$?i09@_pM2O5Ph>SGr*_0$GV++x$(zV=RBAEe@)m
ze?2}>b2Sqho^SJ;hl6i}Rq7z688tY*FoyIG-6X8sY9P(BPPgn{+&mXcVWC&921orX
zN9$1FpM#b6<sbUCqEOkIHh2$B)jzfI)I=Cwk1jX=4aLY9821vTdk5JNncBGp1J1xr
zvTg%&Ti<Qld9Qptwe>_(dz!Zg(X;<+{?cWnXd7=@BhLNHZ70oe=Nc0hj7-y@yWTza
zq!uw>%f#SI2`g&9<SDyih57ha_P$V*JmZ3b9PG}fm=sXQfJUee6<qd5drT|=%>B%z
zpim!3RC_DK*_B>0K{rpN=)wNK$@9PJfx+iDqrTq@Abwgkzuf6<!aiUDYs(x`1Ya`P
z*D6OfM$14^E<w>3oSk3J`q@e-4A1*_<$afu+HqFj>s~t@+kQ4vsSS7Q;QNyOFHjJ5
z=s%;A`6SZ=&{tNtL8L%f^@G_!k^hbY;G4@BM6cg?A9+GR1N||(ZoL&MRm~IC6r8(r
zvb{uxJ~p*)B$m0*=zS`cy<9W<Z24!+>{3t|x$N-nAG00@3P!!Fy4aKFmdD|osm)sV
z<IipWd4ck^A$|#)=IUJ>n)BEA1x)v^pWJnO>fAdbC?=_8Qe1McN|bo*v0LjpqZUJ9
zOy4VqW)7Y~HRjk8nSN;!_)vUV<i_}pFay5TxPzTz6_xQfZ6CJUN9cq*2I>PILedVM
zcW$Vl-<<Nc@kzx{LRw>Di5vHi?KeA~9G~QYs$QNTpHLD`)|vCU=>E!=@7aX)Ee-2d
z77y7>o0~v)Xr7>Z(^7Ni{mIPDcl&pD#geLz2=8rO1{~t<#EzmSsqjkWUWG8W^RN+p
z+_kGN1&negyOJvSzJnod-|}nGCKDGkO?hp*IIDWYccWE^mC~Mozg>1Q%zTMThO8&;
z@wZAui*0pnb;TKT<Lqe-EZd*Kwx>_c1V7sanf-iw+fsNs9`=C!P7%tfSv}T>A(gZP
zU=ch7bV5S=?#~wz1LxRY9@m-PT8aRI^7dFtg-v_&)Z>U?PqAdwYK`Kfp^k-;UN@n|
z#2ua{V9GEA`|j59RE?-k<<j%CncyN?ap*I4GXWRTkjuKb8>i*dnf1))i1A7O4ms}4
zNvxE#bFJT7bWsq|(vHA4?JfotZ3T*wrYB<wllL;*7?51{M6Drxf5NVObiPY_K@BDb
zo=q1_Tsib1UALOnPj?=pd4Z_jKel(6s`>aNFaO`wet?s`tMq4u&5N1940xDRn$pFd
zUEj|4+pmYhXe&YW*RmZMW-1YJjcpZaQ@f--e2M#n5f91;`;(ErmYQ+qVNO{?9qaTS
z*6wZczCiJ2t$a@Y+{DSbvD%ByPuQGeyV|h#<Dfkl$hGHsmkW8V2<=8f|NZ+B;gU}U
zuGT<OKw|su?p1vI7vcfCm9X~e{`T^+k;yJy%B71}Am8U6(uqYw6GcN@sT00CuT{Q&
z@b)1g_xUqp_B#(q=+)EvbiRL9blLTWi{SO}wu4H?cfEe!Vq<%c(ZZ*U(X(B@KFQe7
L+@So*?T7ydUZWsG

literal 0
HcmV?d00001

diff --git a/studio/frontend/src/components/app-sidebar.tsx b/studio/frontend/src/components/app-sidebar.tsx
index 032244ac39..959c2a6b31 100644
--- a/studio/frontend/src/components/app-sidebar.tsx
+++ b/studio/frontend/src/components/app-sidebar.tsx
@@ -36,11 +36,14 @@ import {
   ColumnInsertIcon,
   CursorInfo02Icon,
   Delete02Icon,
+  Download03Icon,
+  GemIcon,
   MessageSearch01Icon,
   Search01Icon,
   NewReleasesIcon,
-  PackageIcon,
+  PowerIcon,
   PencilEdit02Icon,
+  LayoutAlignLeftIcon,
   Settings02Icon,
   ZapIcon,
 } from "@hugeicons/core-free-icons";
@@ -50,9 +53,8 @@ import {
 } from "@/components/ui/tooltip";
 import { Tooltip as TooltipPrimitive } from "radix-ui";
 import { HugeiconsIcon } from "@hugeicons/react";
-import { ChevronDown, ChevronsUpDown, Moon, PanelLeft, Sun } from "lucide-react";
+import { ChevronDown, ChevronsUpDown, Moon, Sun } from "lucide-react";
 import { Link, useNavigate, useRouterState } from "@tanstack/react-router";
-import { motion } from "motion/react";
 import { useTrainingRuntimeStore } from "@/features/training";
 import { useSettingsDialogStore } from "@/features/settings";
 import { useEffectiveProfile, UserAvatar } from "@/features/profile";
@@ -67,7 +69,9 @@ import { useChatSearchStore } from "@/features/chat/stores/chat-search-store";
 import { ChatSearchDialog } from "@/features/chat/components/chat-search-dialog";
 import { useTrainingHistorySidebarItems, deleteTrainingRun } from "@/features/training";
 import type { TrainingRunSummary } from "@/features/training";
-import { useState } from "react";
+import { useEffect, useState } from "react";
+import { ShutdownDialog } from "@/components/shutdown-dialog";
+import { removeTrainingUnloadGuard } from "@/features/training/hooks/use-training-unload-guard";
 
 function getTourId(pathname: string): string | null {
   if (pathname.startsWith("/studio")) return "studio";
@@ -76,8 +80,6 @@ function getTourId(pathname: string): string | null {
   return null;
 }
 
-const NAV_SPRING = { type: "spring", stiffness: 500, damping: 35, mass: 0.5 } as const;
-
 function runStatusDotClass(status: TrainingRunSummary["status"]): string {
   switch (status) {
     case "running":
@@ -107,6 +109,13 @@ function formatRelativeShort(iso: string): string {
   return `${d}d`;
 }
 
+function createNavigationNonce(): string {
+  if (typeof globalThis.crypto?.randomUUID === "function") {
+    return globalThis.crypto.randomUUID();
+  }
+  return `${Date.now()}-${Math.random().toString(36).slice(2, 10)}`;
+}
+
 function NavItem({
   icon,
   label,
@@ -130,13 +139,6 @@ function NavItem({
   return (
     <SidebarMenuItem>
       <div className="relative">
-        {isNav && active && (
-          <motion.div
-            layoutId="sidebar-active-indicator"
-            className="absolute left-0 top-0 bottom-0 w-[3px] rounded-full bg-primary"
-            transition={NAV_SPRING}
-          />
-        )}
         <SidebarMenuButton
           tooltip={label}
           disabled={disabled}
@@ -145,12 +147,12 @@ function NavItem({
           data-tour={dataTour}
           className={
             isNav
-              ? "rounded-none pr-0 pl-4 text-[#475569] dark:text-[#94a3b8] data-active:text-foreground!"
-              : "rounded-none pr-0 pl-4 text-[#475569] dark:text-[#94a3b8] hover:bg-muted! hover:text-foreground! data-active:bg-[oklch(0.94_0_0)]! data-active:text-foreground! dark:data-active:bg-[oklch(0.3_0_0)]!"
+              ? "h-[30px] rounded-[8px] gap-2.5 px-2.5 font-medium text-[#383835] dark:text-[#c7c7c4] hover:bg-[#ececec]! dark:hover:bg-[#2e3035]! hover:text-black! dark:hover:text-white! data-active:bg-[#ececec]! dark:data-active:bg-[#2e3035]! data-active:text-black! dark:data-active:text-white! group-data-[collapsible=icon]:!w-[30px] group-data-[collapsible=icon]:!rounded-[9px] group-data-[collapsible=icon]:mx-auto"
+              : "h-[30px] rounded-[8px] gap-2.5 px-2.5 font-medium text-[#383835] dark:text-[#c7c7c4] hover:bg-[#ececec]! dark:hover:bg-[#2e3035]! hover:text-black! dark:hover:text-white! data-active:bg-[#ececec]! dark:data-active:bg-[#2e3035]! data-active:text-black! dark:data-active:text-white! group-data-[collapsible=icon]:!w-[30px] group-data-[collapsible=icon]:!rounded-[9px] group-data-[collapsible=icon]:mx-auto"
           }
         >
-          <HugeiconsIcon icon={icon} strokeWidth={2} className="size-[18px]" />
-          <span className="text-[13px] font-medium">{label}</span>
+          <HugeiconsIcon icon={icon} strokeWidth={1.5} className="size-[18px]!" />
+          <span className="text-sm">{label}</span>
         </SidebarMenuButton>
       </div>
       {children}
@@ -176,14 +178,16 @@ export function AppSidebar() {
 
   const isTrainingRunning = useTrainingRuntimeStore((s) => s.isTrainingRunning);
   const chatOnly = usePlatformStore((s) => s.isChatOnly());
+  const [shutdownOpen, setShutdownOpen] = useState(false);
 
-  // Chat collapsible state — open by default, syncs with route
+  // Chat collapsible state — open by default, auto-expand on route entry
   const isChatRoute = pathname.startsWith("/chat");
   const isStudioRoute = pathname === "/studio" || pathname.startsWith("/studio/");
   const [chatOpen, setChatOpen] = useState(true);
   const [runsOpen, setRunsOpen] = useState(true);
-  const effectiveChatOpen = isChatRoute || chatOpen;
-  const effectiveRunsOpen = isStudioRoute || runsOpen;
+
+  useEffect(() => { if (isChatRoute) setChatOpen(true); }, [isChatRoute]);
+  useEffect(() => { if (isStudioRoute) setRunsOpen(true); }, [isStudioRoute]);
 
   const isRecipesRoute = pathname.startsWith("/data-recipes");
   const { displayTitle, avatarDataUrl } = useEffectiveProfile();
@@ -219,26 +223,43 @@ export function AppSidebar() {
 
   return (
     <>
-    <Sidebar collapsible="icon" variant="sidebar">
-      <SidebarHeader className="group-data-[collapsible=icon]:px-0">
+    <Sidebar
+      collapsible="icon"
+      variant="sidebar"
+      className="font-heading group-data-[collapsible=icon]:[&_[data-sidebar=sidebar]]:bg-white dark:group-data-[collapsible=icon]:[&_[data-sidebar=sidebar]]:bg-background"
+    >
+      <SidebarHeader className="pl-[17px] pr-3 pt-[12px] pb-[12px] group-data-[collapsible=icon]:px-0">
         {/* Expanded: compact logo + close toggle */}
-        <div className="flex items-center justify-between gap-2 px-1 py-1 group-data-[collapsible=icon]:hidden">
+        <div className="flex items-center justify-between gap-2 group-data-[collapsible=icon]:hidden">
           <Link
-            to={chatOnly ? "/chat" : "/studio"}
-            onClick={closeMobileIfOpen}
-            className="flex items-center select-none"
+            to="/chat"
+            onClick={(event) => {
+              event.preventDefault();
+              if (chatDisabled) return;
+              setActiveThreadId(null);
+              closeMobileIfOpen();
+              void navigate({
+                to: "/chat",
+                search: { new: createNavigationNonce() },
+              });
+            }}
+            className="flex items-center gap-[6px] select-none"
             aria-label="Unsloth home"
           >
             <img
-              src="/blacklogo-c.png"
+              src="/circle-logo-small.png"
               alt="Unsloth"
-              className="h-7 w-auto dark:hidden"
-            />
-            <img
-              src="/whitelogo-c.png"
-              alt="Unsloth"
-              className="hidden h-7 w-auto dark:block"
+              className="h-[34px] w-[34px] rounded-full object-cover"
             />
+            <span className="font-heading text-[21px] font-semibold tracking-[-0.01em] dark:tracking-[0.02em] leading-none text-black dark:text-white">
+              unsloth
+            </span>
+            <span
+              style={{ fontFamily: '"Inter Variable", ui-sans-serif, system-ui, sans-serif' }}
+              className="ml-0.5 inline-flex items-center justify-center rounded-full border border-[#e0ded6] px-[5px] py-[2px] text-[8px] font-medium leading-none tracking-[0.04em] text-[#62605a] antialiased subpixel-antialiased shadow-[0_1px_2px_rgba(0,0,0,0.06)] dark:border-[#3a3c3f] dark:text-[#9d9fa5] dark:shadow-[0_1px_2px_rgba(0,0,0,0.35)]"
+            >
+              BETA
+            </span>
           </Link>
           {!isMobile && (
             <Tooltip>
@@ -246,10 +267,10 @@ export function AppSidebar() {
                 <button
                   type="button"
                   onClick={togglePinned}
-                  className="inline-flex h-7 w-7 items-center justify-center rounded-md text-sidebar-foreground/70 transition-colors hover:bg-sidebar-accent hover:text-sidebar-accent-foreground focus-visible:outline-none focus-visible:ring-2 focus-visible:ring-ring"
+                  className="inline-flex h-7 w-7 items-center justify-center rounded-[8px] text-[#8f8f8f] dark:text-[#5c5c5c] transition-colors hover:bg-[#ececec] dark:hover:bg-[#2e3035] hover:text-black dark:hover:text-white focus-visible:outline-none focus-visible:ring-2 focus-visible:ring-ring"
                   aria-label="Close sidebar"
                 >
-                  <PanelLeft strokeWidth={1.5} className="size-4" />
+                  <HugeiconsIcon icon={LayoutAlignLeftIcon} strokeWidth={1.75} className="size-[18px]" />
                 </button>
               </TooltipPrimitive.Trigger>
               <TooltipContent side="bottom" sideOffset={6}>
@@ -259,18 +280,18 @@ export function AppSidebar() {
           )}
         </div>
 
-        {/* Collapsed: sticker with hover-swap to open toggle */}
+        {/* Collapsed: panel icon doubles as expand trigger */}
         {!isMobile && (
-          <div className="hidden group-data-[collapsible=icon]:flex items-center justify-center h-9 w-full">
+          <div className="hidden group-data-[collapsible=icon]:flex h-[34px] items-center justify-center w-full">
             <Tooltip>
               <TooltipPrimitive.Trigger asChild>
                 <button
                   type="button"
                   onClick={togglePinned}
-                  className="inline-flex h-7 w-7 items-center justify-center rounded-md text-sidebar-foreground/70 transition-colors hover:bg-sidebar-accent hover:text-sidebar-accent-foreground focus-visible:outline-none focus-visible:ring-2 focus-visible:ring-ring"
+                  className="inline-flex h-7 w-7 items-center justify-center rounded-[8px] text-[#383835] dark:text-[#c7c7c4] transition-colors hover:bg-[#ececec] dark:hover:bg-[#2e3035] hover:text-black dark:hover:text-white focus-visible:outline-none focus-visible:ring-2 focus-visible:ring-ring"
                   aria-label="Open sidebar"
                 >
-                  <PanelLeft strokeWidth={1.5} className="size-4" />
+                  <HugeiconsIcon icon={LayoutAlignLeftIcon} strokeWidth={1.75} className="size-[18px]" />
                 </button>
               </TooltipPrimitive.Trigger>
               <TooltipContent side="right" sideOffset={8}>
@@ -281,7 +302,7 @@ export function AppSidebar() {
         )}
       </SidebarHeader>
 
-      <SidebarGroup className="group-data-[collapsible=icon]:p-0 p-0 pt-1 shrink-0">
+      <SidebarGroup className="group-data-[collapsible=icon]:px-0 px-2 pt-[8px] pb-[12px] shrink-0">
         <SidebarGroupContent>
           <SidebarMenu>
             <NavItem
@@ -292,20 +313,20 @@ export function AppSidebar() {
               onClick={() => {
                 if (chatDisabled) return;
                 setActiveThreadId(null);
-                navigate({ to: "/chat", search: { new: crypto.randomUUID() } });
+                navigate({ to: "/chat", search: { new: createNavigationNonce() } });
                 closeMobileIfOpen();
               }}
             />
             <NavItem
               icon={ColumnInsertIcon}
               label="Compare"
-              active={!!search.compare}
+              active={!!search.compare && !chatItems.some((i) => i.id === search.compare)}
               disabled={chatDisabled}
               dataTour="chat-compare"
               onClick={() => {
                 if (chatDisabled) return;
                 setActiveThreadId(null);
-                navigate({ to: "/chat", search: { compare: crypto.randomUUID() } });
+                navigate({ to: "/chat", search: { compare: createNavigationNonce() } });
                 closeMobileIfOpen();
               }}
             />
@@ -322,16 +343,15 @@ export function AppSidebar() {
             />
           </SidebarMenu>
         </SidebarGroupContent>
-        <div className="my-2" />
       </SidebarGroup>
 
       <SidebarContent className="gap-0 overflow-y-auto overscroll-contain min-h-0">
         {/* Navigate (no header) */}
-        <SidebarGroup data-tour="navbar" className="group-data-[collapsible=icon]:p-0 p-0">
+        <SidebarGroup data-tour="navbar" className="group-data-[collapsible=icon]:px-0 px-2 pt-[8px] pb-[12px]">
           <SidebarGroupContent>
             <SidebarMenu>
               <NavItem
-                icon={ZapIcon}
+                icon={GemIcon}
                 label="Train"
                 active={pathname === "/studio" || pathname.startsWith("/studio/")}
                 disabled={chatOnly}
@@ -353,7 +373,7 @@ export function AppSidebar() {
               />
 
               <NavItem
-                icon={PackageIcon}
+                icon={Download03Icon}
                 label="Export"
                 active={pathname === "/export" || pathname.startsWith("/export/")}
                 disabled={chatOnly}
@@ -365,16 +385,15 @@ export function AppSidebar() {
               />
             </SidebarMenu>
           </SidebarGroupContent>
-          <div className="my-2" />
         </SidebarGroup>
 
-        {/* Recent Chats */}
-        {chatItems.length > 0 && (
-          <Collapsible open={effectiveChatOpen} onOpenChange={setChatOpen} asChild>
-          <SidebarGroup className="group-data-[collapsible=icon]:hidden overflow-hidden p-0">
-            <SidebarGroupLabel asChild>
+        {/* Recent Chats — hide on Studio only (Eyera fac13); chatOpen = ec695 clickability */}
+        {!isStudioRoute && chatItems.length > 0 && (
+          <Collapsible open={chatOpen} onOpenChange={setChatOpen} asChild>
+          <SidebarGroup className="group-data-[collapsible=icon]:hidden overflow-hidden px-2 py-0">
+            <SidebarGroupLabel className="pt-2 pb-1.5 pl-2.5 pr-2 text-[12.5px]! font-normal normal-case tracking-normal text-[#62605a] dark:text-[#9d9fa5] focus-visible:ring-0! focus-visible:outline-none" asChild>
               <CollapsibleTrigger className="cursor-pointer flex w-full items-center justify-between">
-                Recent Chats
+                Recents
                 <ChevronDown className="size-3.5 transition-transform duration-200 data-[state=open]:rotate-0 [[data-state=closed]_&]:rotate-[-90deg]" />
               </CollapsibleTrigger>
             </SidebarGroupLabel>
@@ -385,7 +404,7 @@ export function AppSidebar() {
                   <SidebarMenuItem key={item.id} className="group/recent-item relative">
                     <SidebarMenuButton
                       isActive={activeThreadId === item.id}
-                      className="rounded-none pl-4 pr-7 text-[13px] font-medium text-[#475569] dark:text-[#94a3b8] hover:bg-muted! hover:text-foreground! data-active:bg-[oklch(0.94_0_0)]! data-active:text-foreground! dark:data-active:bg-[oklch(0.3_0_0)]!"
+                      className="h-[30px] rounded-[8px] pl-2.5 pr-7 text-sm font-medium text-[#383835] dark:text-[#c7c7c4] hover:bg-[#ececec]! dark:hover:bg-[#2e3035]! hover:text-black! dark:hover:text-white! data-active:bg-[#ececec]! dark:data-active:bg-[#2e3035]! data-active:text-black! dark:data-active:text-white!"
                       onClick={() => {
                         navigate({
                           to: "/chat",
@@ -406,7 +425,7 @@ export function AppSidebar() {
                         handleDeleteThread(item);
                       }}
                       title="Delete"
-                      className="absolute right-1 top-1/2 -translate-y-1/2 flex size-5 scale-90 items-center justify-center rounded-md text-sidebar-foreground/55 opacity-0 transition-all duration-150 hover:bg-destructive/12 hover:text-destructive group-hover/recent-item:scale-100 group-hover/recent-item:opacity-100"
+                      className="absolute right-1 top-1/2 -translate-y-1/2 flex size-5 scale-90 items-center justify-center rounded-[8px] text-sidebar-foreground/55 opacity-0 transition-all duration-150 hover:bg-destructive/12 hover:text-destructive group-hover/recent-item:scale-100 group-hover/recent-item:opacity-100"
                     >
                       <HugeiconsIcon icon={Delete02Icon} strokeWidth={2} className="size-3.5" />
                     </button>
@@ -421,11 +440,11 @@ export function AppSidebar() {
 
         {/* Recent Runs */}
         {isStudioRoute && runItems.length > 0 && !chatOnly && (
-          <Collapsible open={effectiveRunsOpen} onOpenChange={setRunsOpen} asChild>
-          <SidebarGroup className="group-data-[collapsible=icon]:hidden overflow-hidden p-0">
-            <SidebarGroupLabel asChild>
+          <Collapsible open={runsOpen} onOpenChange={setRunsOpen} asChild>
+          <SidebarGroup className="group-data-[collapsible=icon]:hidden overflow-hidden px-2 py-0">
+            <SidebarGroupLabel className="pt-2 pb-1.5 pl-2.5 pr-2 text-[12.5px]! font-normal normal-case tracking-normal text-[#62605a] dark:text-[#9d9fa5] focus-visible:ring-0! focus-visible:outline-none" asChild>
               <CollapsibleTrigger className="cursor-pointer flex w-full items-center justify-between">
-                Recent Runs
+                Recents
                 <ChevronDown className="size-3.5 transition-transform duration-200 data-[state=open]:rotate-0 [[data-state=closed]_&]:rotate-[-90deg]" />
               </CollapsibleTrigger>
             </SidebarGroupLabel>
@@ -442,7 +461,7 @@ export function AppSidebar() {
                       >
                         <SidebarMenuButton
                           isActive={isActiveRun}
-                          className="h-auto flex-col items-start gap-0.5 py-2 rounded-none pl-4 pr-7 text-[13px] font-medium text-[#475569] dark:text-[#94a3b8] hover:bg-muted! hover:text-foreground! data-active:bg-[oklch(0.94_0_0)]! data-active:text-foreground! dark:data-active:bg-[oklch(0.3_0_0)]!"
+                          className="h-auto flex-col items-start gap-0.5 py-1.5 rounded-[8px] pl-2.5 pr-7 text-sm font-medium text-[#383835] dark:text-[#c7c7c4] hover:bg-[#ececec]! dark:hover:bg-[#2e3035]! hover:text-black! dark:hover:text-white! data-active:bg-[#ececec]! dark:data-active:bg-[#2e3035]! data-active:text-black! dark:data-active:text-white!"
                           onClick={() => {
                             setSelectedHistoryRunId(run.id);
                             closeMobileIfOpen();
@@ -456,7 +475,7 @@ export function AppSidebar() {
                               )}
                               aria-hidden
                             />
-                            <span className="truncate text-sm font-medium">
+                            <span className="truncate text-sm">
                               {run.model_name}
                             </span>
                             <span className="ml-auto shrink-0 text-[10px] text-muted-foreground">
@@ -482,7 +501,7 @@ export function AppSidebar() {
                             }
                           }}
                           title="Delete"
-                          className="absolute right-1 top-1/2 -translate-y-1/2 flex size-5 scale-90 items-center justify-center rounded-md text-sidebar-foreground/55 opacity-0 transition-all duration-150 hover:bg-destructive/12 hover:text-destructive group-hover/run-item:scale-100 group-hover/run-item:opacity-100"
+                          className="absolute right-1 top-1/2 -translate-y-1/2 flex size-5 scale-90 items-center justify-center rounded-[8px] text-sidebar-foreground/55 opacity-0 transition-all duration-150 hover:bg-destructive/12 hover:text-destructive group-hover/run-item:scale-100 group-hover/run-item:opacity-100"
                         >
                           <HugeiconsIcon icon={Delete02Icon} strokeWidth={2} className="size-3.5" />
                         </button>
@@ -497,7 +516,7 @@ export function AppSidebar() {
         )}
       </SidebarContent>
 
-      <SidebarFooter className="border-t border-sidebar-border group-data-[collapsible=icon]:border-t-0">
+      <SidebarFooter className="border-t border-sidebar-border group-data-[collapsible=icon]:border-t-0 group-data-[collapsible=icon]:px-0">
         <SidebarMenu>
           <SidebarMenuItem>
             <DropdownMenu>
@@ -505,7 +524,7 @@ export function AppSidebar() {
                 <SidebarMenuButton
                   size="lg"
                   aria-label={`${displayTitle} account menu`}
-                  className="data-[state=open]:bg-sidebar-accent data-[state=open]:text-sidebar-accent-foreground group-data-[collapsible=icon]:overflow-visible group-data-[collapsible=icon]:hover:bg-transparent group-data-[collapsible=icon]:data-[state=open]:bg-transparent"
+                  className="!h-[50px] gap-[8px] rounded-[8px] text-[#383835] dark:text-[#c7c7c4] hover:bg-[#ececec]! dark:hover:bg-[#2e3035]! hover:text-black! dark:hover:text-white! data-[state=open]:bg-[#ececec]! dark:data-[state=open]:bg-[#2e3035]! data-[state=open]:text-black! dark:data-[state=open]:text-white!"
                 >
                   <div className="shrink-0">
                     <UserAvatar
@@ -516,8 +535,8 @@ export function AppSidebar() {
                     />
                   </div>
                   <div className="flex flex-col gap-0.5 leading-none group-data-[collapsible=icon]:hidden">
-                    <span className="truncate text-sm font-semibold">{displayTitle}</span>
-                    <span className="truncate text-[11px] text-muted-foreground">Train</span>
+                    <span className="truncate font-heading text-[13px] font-semibold text-[#383835] dark:text-[#c7c7c4]">{displayTitle}</span>
+                    <span className="truncate text-[11px] text-muted-foreground">Studio</span>
                   </div>
                   <ChevronsUpDown strokeWidth={1.25} className="ml-auto size-4 text-muted-foreground group-data-[collapsible=icon]:hidden" />
                 </SidebarMenuButton>
@@ -525,7 +544,7 @@ export function AppSidebar() {
               <DropdownMenuContent
                 side="top"
                 align="start"
-                className="w-56"
+                className="w-[15rem] font-heading [&_[data-slot=dropdown-menu-item]]:rounded-[8px] [&_[data-slot=dropdown-menu-item]]:font-medium [&_[data-slot=dropdown-menu-item]]:text-[#383835] dark:[&_[data-slot=dropdown-menu-item]]:text-[#c7c7c4] [&_[data-slot=dropdown-menu-item]:focus]:bg-[#ececec] dark:[&_[data-slot=dropdown-menu-item]:focus]:bg-[#2e3035] [&_[data-slot=dropdown-menu-item]:focus]:text-black dark:[&_[data-slot=dropdown-menu-item]:focus]:text-white [&_[data-slot=dropdown-menu-item]:focus_*]:text-black! dark:[&_[data-slot=dropdown-menu-item]:focus_*]:text-white!"
               >
                 <DropdownMenuGroup>
                   <DropdownMenuItem
@@ -586,20 +605,24 @@ export function AppSidebar() {
                       <span>What's New</span>
                     </a>
                   </DropdownMenuItem>
+                  <DropdownMenuItem asChild>
+                    <a
+                      href="https://github.com/unslothai/unsloth/issues"
+                      target="_blank"
+                      rel="noopener noreferrer"
+                    >
+                      <HugeiconsIcon
+                        icon={MessageSearch01Icon}
+                        className="size-4"
+                      />
+                      <span>Feedback</span>
+                    </a>
+                  </DropdownMenuItem>
                 </DropdownMenuGroup>
                 <DropdownMenuSeparator />
-                <DropdownMenuItem asChild>
-                  <a
-                    href="https://github.com/unslothai/unsloth/issues"
-                    target="_blank"
-                    rel="noopener noreferrer"
-                  >
-                    <HugeiconsIcon
-                      icon={MessageSearch01Icon}
-                      className="size-4"
-                    />
-                    <span>Feedback</span>
-                  </a>
+                <DropdownMenuItem onSelect={() => setShutdownOpen(true)}>
+                  <HugeiconsIcon icon={PowerIcon} className="size-4" />
+                  <span>Shutdown</span>
                 </DropdownMenuItem>
               </DropdownMenuContent>
             </DropdownMenu>
@@ -608,6 +631,11 @@ export function AppSidebar() {
       </SidebarFooter>
     </Sidebar>
     <ChatSearchDialog />
+    <ShutdownDialog
+      open={shutdownOpen}
+      onOpenChange={setShutdownOpen}
+      onAfterShutdown={removeTrainingUnloadGuard}
+    />
     </>
   );
 }
diff --git a/studio/frontend/src/components/assistant-ui/code-plugin.ts b/studio/frontend/src/components/assistant-ui/code-plugin.ts
new file mode 100644
index 0000000000..5df7ac4f95
--- /dev/null
+++ b/studio/frontend/src/components/assistant-ui/code-plugin.ts
@@ -0,0 +1,66 @@
+// SPDX-License-Identifier: AGPL-3.0-only
+// Copyright 2026-present the Unsloth AI Inc. team. All rights reserved. See /studio/LICENSE.AGPL-3.0
+
+import {
+  createCodePlugin as createShikiCodePlugin,
+  type CodeHighlighterPlugin,
+  type CodePluginOptions,
+  type HighlightOptions,
+  type HighlightResult,
+} from "@streamdown/code";
+import type { BundledLanguage } from "shiki";
+
+// Fence tags LLMs/users commonly write that shiki doesn't expose as aliases.
+// Keys are lower-cased input; values are canonical shiki language ids.
+const LANGUAGE_ALIAS_OVERRIDES: Record<string, BundledLanguage> = {
+  objectivec: "objective-c",
+  "obj-c": "objective-c",
+  objectivecpp: "objective-cpp",
+  "objective-cplusplus": "objective-cpp",
+  objcpp: "objective-cpp",
+  "c++": "cpp",
+  cplusplus: "cpp",
+  "c#": "csharp",
+  cs: "csharp",
+  "f#": "fsharp",
+  "c-sharp": "csharp",
+  "f-sharp": "fsharp",
+  golang: "go",
+  rs: "rust",
+  rb: "ruby",
+  py: "python",
+  sh: "shellscript",
+  bash: "shellscript",
+  zsh: "shellscript",
+  shell: "shellscript",
+  yml: "yaml",
+  ts: "typescript",
+  js: "javascript",
+  kt: "kotlin",
+  rsx: "rust",
+  "vue-html": "vue",
+};
+
+const normalizeLanguage = (language: string): BundledLanguage => {
+  const key = language.trim().toLowerCase();
+  const override = LANGUAGE_ALIAS_OVERRIDES[key];
+  return (override ?? (key as BundledLanguage));
+};
+
+export function createCodePlugin(
+  options: CodePluginOptions = {},
+): CodeHighlighterPlugin {
+  const inner = createShikiCodePlugin(options);
+  return {
+    ...inner,
+    supportsLanguage: (language) => inner.supportsLanguage(normalizeLanguage(language)),
+    highlight: (
+      opts: HighlightOptions,
+      callback?: (result: HighlightResult) => void,
+    ) =>
+      inner.highlight(
+        { ...opts, language: normalizeLanguage(opts.language) },
+        callback,
+      ),
+  };
+}
diff --git a/studio/frontend/src/components/assistant-ui/code-themes.ts b/studio/frontend/src/components/assistant-ui/code-themes.ts
new file mode 100644
index 0000000000..2557b45ef0
--- /dev/null
+++ b/studio/frontend/src/components/assistant-ui/code-themes.ts
@@ -0,0 +1,30 @@
+// SPDX-License-Identifier: AGPL-3.0-only
+// Copyright 2026-present the Unsloth AI Inc. team. All rights reserved. See /studio/LICENSE.AGPL-3.0
+
+import oneDarkPro from "@shikijs/themes/one-dark-pro";
+import oneLight from "@shikijs/themes/one-light";
+import type { ThemeRegistrationAny } from "shiki";
+
+// Canonical Atom One Dark / One Light themes, shipped by `@shikijs/themes`.
+// We only override the background so the code block blends into the app's
+// `--code-block` surface instead of painting its own. Every token color and
+// scope mapping is left intact — that's what gives consistent multi-language
+// highlighting (including Objective-C, Go, Rust, etc.) out of the box.
+const withTransparentBg = (theme: ThemeRegistrationAny): ThemeRegistrationAny => ({
+  ...theme,
+  bg: "transparent",
+  colors: {
+    ...theme.colors,
+    "editor.background": "transparent",
+  },
+});
+
+export const unslothLightTheme: ThemeRegistrationAny = {
+  ...withTransparentBg(oneLight),
+  name: "unsloth-light",
+};
+
+export const unslothDarkTheme: ThemeRegistrationAny = {
+  ...withTransparentBg(oneDarkPro),
+  name: "unsloth-dark",
+};
diff --git a/studio/frontend/src/components/assistant-ui/code-toggle-icon.tsx b/studio/frontend/src/components/assistant-ui/code-toggle-icon.tsx
new file mode 100644
index 0000000000..6d7abefad5
--- /dev/null
+++ b/studio/frontend/src/components/assistant-ui/code-toggle-icon.tsx
@@ -0,0 +1,22 @@
+// SPDX-License-Identifier: AGPL-3.0-only
+// Copyright 2026-present the Unsloth AI Inc. team. All rights reserved. See /studio/LICENSE.AGPL-3.0
+
+import type { FC } from "react";
+
+export const CodeToggleIcon: FC<{ className?: string }> = ({ className }) => {
+  return (
+    <svg
+      viewBox="0 0 24 24"
+      fill="none"
+      stroke="currentColor"
+      strokeWidth="2"
+      strokeLinecap="round"
+      strokeLinejoin="round"
+      className={className}
+      aria-hidden="true"
+    >
+      <polyline points="16 18 22 12 16 6" />
+      <polyline points="8 6 2 12 8 18" />
+    </svg>
+  );
+};
diff --git a/studio/frontend/src/components/assistant-ui/markdown-text.tsx b/studio/frontend/src/components/assistant-ui/markdown-text.tsx
index ea409591f0..2a0517a44a 100644
--- a/studio/frontend/src/components/assistant-ui/markdown-text.tsx
+++ b/studio/frontend/src/components/assistant-ui/markdown-text.tsx
@@ -8,7 +8,7 @@ import { preprocessLaTeX } from "@/lib/latex";
 import { INTERNAL, useMessagePartText } from "@assistant-ui/react";
 import { Copy02Icon, Tick02Icon } from "@hugeicons/core-free-icons";
 import { HugeiconsIcon } from "@hugeicons/react";
-import { code } from "@streamdown/code";
+import { createCodePlugin } from "./code-plugin";
 import { createMathPlugin } from "@streamdown/math";
 import { mermaid } from "@streamdown/mermaid";
 import { DownloadIcon, Maximize2Icon, Minimize2Icon } from "lucide-react";
@@ -16,8 +16,12 @@ import { useEffect, useMemo, useRef, useState } from "react";
 import { Block, type BlockProps, Streamdown } from "streamdown";
 import "katex/dist/katex.min.css";
 import { AudioPlayer } from "./audio-player";
+import { unslothDarkTheme, unslothLightTheme } from "./code-themes";
 
 const math = createMathPlugin({ singleDollarTextMath: true });
+const code = createCodePlugin({
+  themes: [unslothLightTheme, unslothDarkTheme],
+});
 const { withSmoothContextProvider } = INTERNAL;
 
 const STREAMDOWN_COMPONENTS = {
@@ -425,7 +429,7 @@ const MarkdownTextImpl = () => {
             panZoom: true,
           },
         }}
-        shikiTheme={["github-light", "github-dark"]}
+        shikiTheme={[unslothLightTheme, unslothDarkTheme]}
         BlockComponent={StreamdownBlock}
       >
         {processedText}
diff --git a/studio/frontend/src/components/assistant-ui/model-selector.tsx b/studio/frontend/src/components/assistant-ui/model-selector.tsx
index 08e69bbf93..3628252177 100644
--- a/studio/frontend/src/components/assistant-ui/model-selector.tsx
+++ b/studio/frontend/src/components/assistant-ui/model-selector.tsx
@@ -68,9 +68,9 @@ function ModelSelectorTrigger({
         className={cn(
           "flex items-center gap-2 transition-colors",
           variant === "outline" &&
-          "rounded-full border border-border/60 hover:bg-accent",
-          variant === "ghost" && "rounded-md hover:bg-accent",
-          variant === "muted" && "rounded-md bg-muted hover:bg-muted/80",
+          "rounded-[8px] border border-border/60 hover:bg-[#ececec] dark:hover:bg-[#2e3035]",
+          variant === "ghost" && "rounded-[8px] hover:bg-[#ececec] dark:hover:bg-[#2e3035]",
+          variant === "muted" && "rounded-[8px] bg-muted hover:bg-muted/80",
           size === "sm" && "h-8 px-3 text-xs",
           size === "default" && "h-9 px-3.5 text-sm",
           size === "lg" && "h-10 px-4 text-sm",
@@ -80,15 +80,16 @@ function ModelSelectorTrigger({
         {isLoaded && (
           <span className="size-2 shrink-0 rounded-full bg-emerald-500" />
         )}
-        <span className={isLoaded ? "text-foreground" : "text-muted-foreground"}>
-          {currentModel?.name ?? "Select model..."}
+        <span className="font-heading font-medium text-[16px] text-black dark:text-white">
+          {currentModel?.name ?? "Select model"}
         </span>
         {currentModel?.description && (
           <span className="text-muted-foreground text-xs">{currentModel.description}</span>
         )}
         <HugeiconsIcon
           icon={ArrowDown01Icon}
-          className="size-3 shrink-0 text-muted-foreground"
+          strokeWidth={1.75}
+          className="size-3.5 shrink-0 text-muted-foreground"
         />
       </button>
     </PopoverTrigger>
diff --git a/studio/frontend/src/components/assistant-ui/model-selector/pickers.tsx b/studio/frontend/src/components/assistant-ui/model-selector/pickers.tsx
index 8f318a8292..30b8fdecb8 100644
--- a/studio/frontend/src/components/assistant-ui/model-selector/pickers.tsx
+++ b/studio/frontend/src/components/assistant-ui/model-selector/pickers.tsx
@@ -152,8 +152,8 @@ function ModelRow({
       type="button"
       onClick={onClick}
       className={cn(
-        "flex w-full items-center gap-2 rounded-md px-2.5 py-1.5 text-left text-sm transition-colors hover:bg-accent",
-        selected && "bg-accent/60",
+        "flex w-full items-center gap-2 rounded-[6px] px-2.5 py-1.5 text-left text-sm transition-colors hover:bg-[#ececec] dark:hover:bg-[#2e3035]",
+        selected && "bg-[#ececec] dark:bg-[#2e3035]",
       )}
     >
       <span
@@ -376,7 +376,7 @@ function GgufVariantExpander({
                 handleVariantClick(v.quant, v.downloaded, v.size_bytes)
               }
               className={cn(
-                "flex min-w-0 flex-1 items-center justify-between gap-2 rounded-md px-2.5 py-1 text-left text-sm transition-colors hover:bg-accent",
+                "flex min-w-0 flex-1 items-center justify-between gap-2 rounded-[6px] px-2.5 py-1 text-left text-sm transition-colors hover:bg-[#ececec] dark:hover:bg-[#2e3035]",
               )}
             >
               <span className="min-w-0 flex-1 truncate font-mono text-xs">
diff --git a/studio/frontend/src/components/assistant-ui/reasoning.tsx b/studio/frontend/src/components/assistant-ui/reasoning.tsx
index 4d46fc8d59..387f8cd458 100644
--- a/studio/frontend/src/components/assistant-ui/reasoning.tsx
+++ b/studio/frontend/src/components/assistant-ui/reasoning.tsx
@@ -11,6 +11,7 @@ import {
   CollapsibleContent,
   CollapsibleTrigger,
 } from "@/components/ui/collapsible";
+import { useCollapseScrollLock } from "@/hooks/use-collapse-scroll-lock";
 import { cn } from "@/lib/utils";
 import {
   type ReasoningGroupComponent,
@@ -67,49 +68,8 @@ function ReasoningRoot({
   ...props
 }: ReasoningRootProps) {
   const collapsibleRef = useRef<HTMLDivElement>(null);
-  const lockCleanupRef = useRef<(() => void) | null>(null);
   const [uncontrolledOpen, setUncontrolledOpen] = useState(defaultOpen);
-
-  useEffect(() => {
-    return () => {
-      lockCleanupRef.current?.();
-    };
-  }, []);
-
-  const lockScroll = useCallback(() => {
-    lockCleanupRef.current?.();
-
-    const animatedElement = collapsibleRef.current;
-    if (!animatedElement) return;
-
-    let scrollContainer: HTMLElement | null = animatedElement;
-    while (scrollContainer) {
-      const { overflowY } = getComputedStyle(scrollContainer);
-      if (overflowY === "scroll" || overflowY === "auto") {
-        break;
-      }
-      scrollContainer = scrollContainer.parentElement;
-    }
-    if (!scrollContainer) return;
-
-    const scrollPosition = scrollContainer.scrollTop;
-    const resetPosition = () => {
-      scrollContainer.scrollTop = scrollPosition;
-    };
-
-    scrollContainer.addEventListener("scroll", resetPosition);
-    let timeoutId: ReturnType<typeof setTimeout> | null = null;
-    const cleanup = () => {
-      if (timeoutId !== null) {
-        clearTimeout(timeoutId);
-        timeoutId = null;
-      }
-      scrollContainer.removeEventListener("scroll", resetPosition);
-      lockCleanupRef.current = null;
-    };
-    timeoutId = setTimeout(cleanup, ANIMATION_DURATION);
-    lockCleanupRef.current = cleanup;
-  }, []);
+  const lockScroll = useCollapseScrollLock(collapsibleRef, ANIMATION_DURATION);
 
   const isControlled = controlledOpen !== undefined;
   const isOpen = isControlled ? controlledOpen : uncontrolledOpen;
diff --git a/studio/frontend/src/components/assistant-ui/thread.tsx b/studio/frontend/src/components/assistant-ui/thread.tsx
index 36713d665f..54f64220e8 100644
--- a/studio/frontend/src/components/assistant-ui/thread.tsx
+++ b/studio/frontend/src/components/assistant-ui/thread.tsx
@@ -16,6 +16,7 @@ import { WebSearchToolUI } from "@/components/assistant-ui/tool-ui-web-search";
 import { PythonToolUI } from "@/components/assistant-ui/tool-ui-python";
 import { TerminalToolUI } from "@/components/assistant-ui/tool-ui-terminal";
 import { TooltipIconButton } from "@/components/assistant-ui/tooltip-icon-button";
+import { CodeToggleIcon } from "@/components/assistant-ui/code-toggle-icon";
 import { Button } from "@/components/ui/button";
 import { sentAudioNames } from "@/features/chat/api/chat-adapter";
 import { AUDIO_ACCEPT, MAX_AUDIO_SIZE, fileToBase64 } from "@/lib/audio-utils";
@@ -34,6 +35,7 @@ import {
   useAui,
   useAuiEvent,
   useAuiState,
+  useThreadViewport,
 } from "@assistant-ui/react";
 import { motion } from "motion/react";
 import {
@@ -69,12 +71,7 @@ export const Thread: FC<{ hideComposer?: boolean; hideWelcome?: boolean }> = ({
 }) => {
   return (
     <ThreadPrimitive.Root
-      className={cn(
-        "aui-root aui-thread-root @container flex flex-col",
-        hideComposer
-          ? "h-full"
-          : "relative min-h-0 min-w-0 flex-1 basis-0 overflow-hidden",
-      )}
+      className="aui-root aui-thread-root @container relative flex min-h-0 min-w-0 flex-1 basis-0 flex-col overflow-hidden"
       style={{
         ["--thread-max-width" as string]: "44rem",
         ["--thread-content-max-width" as string]:
@@ -83,10 +80,8 @@ export const Thread: FC<{ hideComposer?: boolean; hideWelcome?: boolean }> = ({
     >
       <ThreadPrimitive.Viewport
         className={cn(
-          "aui-thread-viewport relative flex min-w-0 flex-1 flex-col overflow-x-auto overflow-y-scroll scroll-smooth px-5",
-          hideComposer
-            ? "pt-4"
-            : "h-0 min-h-0 basis-0 pt-[56px]",
+          "aui-thread-viewport relative flex min-h-0 min-w-0 flex-1 basis-0 flex-col overflow-x-auto overflow-y-auto scroll-smooth px-5",
+          hideComposer ? "pt-4" : "pt-[48px]",
         )}
       >
         {!hideWelcome && (
@@ -103,31 +98,29 @@ export const Thread: FC<{ hideComposer?: boolean; hideWelcome?: boolean }> = ({
           }}
         />
 
-        {/* Small overlap and extra slack so the last lines can scroll under the composer cleanly */}
-        {!hideComposer && <div className="h-40 shrink-0" aria-hidden />}
-
-        <ThreadPrimitive.ViewportFooter
-          className={cn(
-            "aui-thread-viewport-footer sticky z-20 mt-auto flex w-full flex-col overflow-visible bg-transparent",
-            hideComposer
-              ? "bottom-0 gap-2"
-              : "bottom-[140px] shrink-0 gap-3",
-            // Compare: pointer-events pass-through so messages behind footer stay clickable
-            hideComposer
-              ? "pointer-events-none pb-3"
-              : "pb-2",
-          )}
-        >
+        {/* Bottom slack so the last message has breathing room above the
+            sticky scroll-to-bottom button (and the floating composer in
+            single mode). Without this, content would butt against the
+            sticky footer and feel cramped. */}
+        <AuiIf condition={({ thread }) => !thread.isEmpty}>
           <div
+            className={cn("shrink-0", hideComposer ? "h-16" : "h-40")}
+            aria-hidden
+          />
+        </AuiIf>
+
+        <AuiIf condition={({ thread }) => !thread.isEmpty}>
+          <ThreadPrimitive.ViewportFooter
             className={cn(
-              "flex justify-center",
-              hideComposer && "pointer-events-auto",
+              "aui-thread-viewport-footer pointer-events-none sticky z-20 flex w-full justify-center bg-transparent",
+              hideComposer ? "bottom-3" : "bottom-[140px]",
             )}
           >
             <ThreadScrollToBottom />
-          </div>
-        </ThreadPrimitive.ViewportFooter>
+          </ThreadPrimitive.ViewportFooter>
+        </AuiIf>
       </ThreadPrimitive.Viewport>
+
       {!hideComposer && (
         <AuiIf condition={({ thread }) => !thread.isEmpty}>
           <div className="aui-thread-composer-dock pointer-events-none absolute bottom-0 left-0 right-0 md:right-2 z-20">
@@ -140,7 +133,7 @@ export const Thread: FC<{ hideComposer?: boolean; hideWelcome?: boolean }> = ({
                 <ComposerAnimated />
               </div>
               <p className="mt-1.5 text-center text-[11px] text-muted-foreground">
-                LLM's can make mistakes. Double-check all responses.
+                LLMs can make mistakes. Double-check all responses.
               </p>
             </div>
           </div>
@@ -151,12 +144,24 @@ export const Thread: FC<{ hideComposer?: boolean; hideWelcome?: boolean }> = ({
 };
 
 const ThreadScrollToBottom: FC = () => {
+  // Scoped to the nearest ThreadPrimitive.Root via context, so in compare
+  // mode each pane reads its own viewport state.
+  //
+  // The button stays mounted and toggles visibility via CSS. Conditionally
+  // rendering (return null) unmounts a DOM node inside the viewport, which
+  // the assistant-ui autoscroll hook's MutationObserver sees as a content
+  // change — during streaming that triggered spurious scroll-to-bottom
+  // calls, especially in the narrower mobile stacked layout.
+  const isAtBottom = useThreadViewport((vp) => vp.isAtBottom);
   return (
     <ThreadPrimitive.ScrollToBottom asChild={true}>
       <TooltipIconButton
         tooltip="Scroll to bottom"
         variant="outline"
-        className="aui-thread-scroll-to-bottom absolute -top-6 z-10 self-center rounded-full p-4 disabled:invisible dark:bg-background dark:hover:bg-accent"
+        className={cn(
+          "aui-thread-scroll-to-bottom pointer-events-auto rounded-full p-4 bg-background hover:bg-accent dark:bg-background dark:hover:bg-accent",
+          isAtBottom && "invisible pointer-events-none",
+        )}
       >
         <ArrowDownIcon />
       </TooltipIconButton>
@@ -227,7 +232,7 @@ const SuggestionItem: FC = () => {
 const ThreadWelcome: FC<{ hideComposer?: boolean }> = ({ hideComposer }) => {
   return (
     <div className="aui-thread-welcome-root mx-auto my-auto flex w-full max-w-(--thread-max-width) grow flex-col">
-      <div className="aui-thread-welcome-center flex w-full grow flex-col items-center justify-center">
+      <div className="aui-thread-welcome-center flex w-full grow flex-col items-center justify-center pb-[48px]">
         <div className="aui-thread-welcome-message flex w-full flex-col justify-center gap-6 px-4">
           <div className="flex flex-col items-center gap-2 text-center">
             <img
@@ -235,10 +240,10 @@ const ThreadWelcome: FC<{ hideComposer?: boolean }> = ({ hideComposer }) => {
               alt="Sloth mascot"
               className="size-20"
             />
-            <h1 className="aui-thread-welcome-message-inner fade-in slide-in-from-bottom-1 animate-in font-bold text-2xl tracking-[-0.02em] duration-200">
+            <h1 className="aui-thread-welcome-message-inner fade-in slide-in-from-bottom-1 animate-in font-heading font-semibold text-2xl tracking-[-0.02em] duration-200">
               Chat with your model
             </h1>
-            <p className="aui-thread-welcome-message-inner fade-in slide-in-from-bottom-1 animate-in text-muted-foreground text-sm delay-75 duration-200">
+            <p className="aui-thread-welcome-message-inner fade-in slide-in-from-bottom-1 -mt-1 animate-in font-heading font-normal text-muted-foreground text-sm delay-75 duration-200">
               Run GGUFs, safetensors, vision and audio models
             </p>
           </div>
@@ -303,13 +308,13 @@ const PendingAudioChip: FC = () => {
 const Composer: FC = () => {
   return (
     <ComposerPrimitive.Root className="aui-composer-root relative flex w-full flex-col">
-      <ComposerPrimitive.AttachmentDropzone className="aui-composer-attachment-dropzone chat-composer-surface flex w-full flex-col rounded-3xl bg-background px-1 pt-2 outline-none transition-shadow data-[dragging=true]:border-ring data-[dragging=true]:bg-accent/50">
+      <ComposerPrimitive.AttachmentDropzone className="aui-composer-attachment-dropzone chat-composer-surface flex w-full flex-col rounded-3xl bg-background dark:bg-card px-1 pt-2 outline-none transition-shadow data-[dragging=true]:border-ring data-[dragging=true]:bg-accent/50">
         <ComposerAttachments />
         <PendingAudioChip />
         <ToolStatusDisplay />
         <ComposerPrimitive.Input
           placeholder="Send a message..."
-          className="aui-composer-input mb-1 min-h-12 w-full resize-none overflow-y-auto bg-transparent pl-5 pr-4 pt-2 pb-3 text-sm outline-none placeholder:text-muted-foreground focus-visible:ring-0"
+          className="aui-composer-input mb-1 min-h-12 w-full resize-none overflow-y-auto bg-transparent pl-5 pr-4 pt-2 pb-3 text-sm font-[450] outline-none placeholder:text-muted-foreground focus-visible:ring-0"
           minRows={1}
           maxRows={6}
           autoFocus={true}
@@ -483,24 +488,6 @@ const CodeToolsToggle: FC = () => {
   );
 };
 
-const CodeToggleIcon: FC<{ className?: string }> = ({ className }) => {
-  return (
-    <svg
-      viewBox="0 0 24 24"
-      fill="none"
-      stroke="currentColor"
-      strokeWidth="2"
-      strokeLinecap="round"
-      strokeLinejoin="round"
-      className={className}
-      aria-hidden="true"
-    >
-      <polyline points="16 18 22 12 16 6" />
-      <polyline points="8 6 2 12 8 18" />
-    </svg>
-  );
-};
-
 const ToolStatusDisplay: FC = () => {
   const toolStatus = useChatRuntimeStore((s) => s.toolStatus);
   const isThreadRunning = useAuiState(({ thread }) => thread.isRunning);
@@ -638,7 +625,7 @@ const GeneratingIndicator: FC = () => {
 const AssistantMessage: FC = () => {
   return (
     <MessagePrimitive.Root
-      className="aui-assistant-message-root fade-in slide-in-from-bottom-1 relative mx-auto min-w-0 w-full max-w-(--thread-content-max-width) animate-in py-0.5 text-[15.5px] duration-150"
+      className="aui-assistant-message-root fade-in slide-in-from-bottom-1 relative mx-auto min-w-0 w-full max-w-(--thread-content-max-width) animate-in py-0.5 text-[15.5px] font-[450] duration-150"
       data-role="assistant"
     >
       <div className="aui-assistant-message-content wrap-break-word min-w-0 text-foreground leading-relaxed">
@@ -791,14 +778,14 @@ const UserMessageAudio: FC = () => {
 const UserMessage: FC = () => {
   return (
     <MessagePrimitive.Root
-      className="aui-user-message-root fade-in slide-in-from-bottom-1 mx-auto flex w-full max-w-(--thread-content-max-width) animate-in flex-col items-end gap-y-2 pt-6 pb-0.5 text-[15.5px] duration-150"
+      className="aui-user-message-root fade-in slide-in-from-bottom-1 mx-auto flex w-full max-w-(--thread-content-max-width) animate-in flex-col items-end gap-y-2 pt-6 pb-0.5 text-[15.5px] font-[450] duration-150"
       data-role="user"
     >
       <UserMessageAttachments />
       <UserMessageAudio />
 
       <div className="aui-user-message-content-wrapper flex max-w-[80%] min-w-0 flex-col items-end">
-        <div className="aui-user-message-content wrap-break-word w-fit rounded-2xl bg-muted px-4 py-2.5 text-foreground">
+        <div className="aui-user-message-content wrap-break-word w-fit rounded-[16px] rounded-tr-[4px] bg-[#f5f5f5] px-4 py-2.5 text-foreground dark:bg-card">
           <MessagePrimitive.Parts />
         </div>
         <div className="mt-1 flex min-h-6">
@@ -844,7 +831,7 @@ const EditComposer: FC = () => {
     <MessagePrimitive.Root className="aui-edit-composer-wrapper mx-auto flex w-full max-w-(--thread-content-max-width) flex-col py-3">
       <ComposerPrimitive.Root className="aui-edit-composer-root ml-auto flex w-full max-w-[85%] flex-col rounded-2xl bg-muted">
         <ComposerPrimitive.Input
-          className="aui-edit-composer-input min-h-14 w-full resize-none bg-transparent p-4 text-foreground text-sm outline-none"
+          className="aui-edit-composer-input min-h-14 w-full resize-none bg-transparent p-4 text-foreground text-sm font-[450] outline-none"
           autoFocus={true}
         />
         <div className="aui-edit-composer-footer mx-3 mb-3 flex items-center gap-2 self-end">
diff --git a/studio/frontend/src/components/assistant-ui/tool-fallback.tsx b/studio/frontend/src/components/assistant-ui/tool-fallback.tsx
index 82a5b17e04..e407163045 100644
--- a/studio/frontend/src/components/assistant-ui/tool-fallback.tsx
+++ b/studio/frontend/src/components/assistant-ui/tool-fallback.tsx
@@ -8,11 +8,11 @@ import {
   CollapsibleContent,
   CollapsibleTrigger,
 } from "@/components/ui/collapsible";
+import { useCollapseScrollLock } from "@/hooks/use-collapse-scroll-lock";
 import { cn } from "@/lib/utils";
 import {
   type ToolCallMessagePartComponent,
   type ToolCallMessagePartStatus,
-  useScrollLock,
 } from "@assistant-ui/react";
 import {
   AlertCircleIcon,
@@ -52,7 +52,7 @@ function ToolFallbackRoot({
 }: ToolFallbackRootProps) {
   const collapsibleRef = useRef<HTMLDivElement>(null);
   const [uncontrolledOpen, setUncontrolledOpen] = useState(defaultOpen);
-  const lockScroll = useScrollLock(collapsibleRef, ANIMATION_DURATION);
+  const lockScroll = useCollapseScrollLock(collapsibleRef, ANIMATION_DURATION);
 
   const isControlled = controlledOpen !== undefined;
   const isOpen = isControlled ? controlledOpen : uncontrolledOpen;
diff --git a/studio/frontend/src/components/assistant-ui/tool-group.tsx b/studio/frontend/src/components/assistant-ui/tool-group.tsx
index bf7a6a9a25..e91da4cee2 100644
--- a/studio/frontend/src/components/assistant-ui/tool-group.tsx
+++ b/studio/frontend/src/components/assistant-ui/tool-group.tsx
@@ -12,12 +12,12 @@ import { ChevronDownIcon, LoaderIcon } from "lucide-react";
 import { Wrench01Icon } from "@hugeicons/core-free-icons";
 import { HugeiconsIcon } from "@hugeicons/react";
 import { cva, type VariantProps } from "class-variance-authority";
-import { useScrollLock } from "@assistant-ui/react";
 import {
   Collapsible,
   CollapsibleContent,
   CollapsibleTrigger,
 } from "@/components/ui/collapsible";
+import { useCollapseScrollLock } from "@/hooks/use-collapse-scroll-lock";
 import { cn } from "@/lib/utils";
 
 const ANIMATION_DURATION = 200;
@@ -54,7 +54,7 @@ function ToolGroupRoot({
 }: ToolGroupRootProps) {
   const collapsibleRef = useRef<HTMLDivElement>(null);
   const [uncontrolledOpen, setUncontrolledOpen] = useState(defaultOpen);
-  const lockScroll = useScrollLock(collapsibleRef, ANIMATION_DURATION);
+  const lockScroll = useCollapseScrollLock(collapsibleRef, ANIMATION_DURATION);
 
   const isControlled = controlledOpen !== undefined;
   const isOpen = isControlled ? controlledOpen : uncontrolledOpen;
diff --git a/studio/frontend/src/components/navbar.tsx b/studio/frontend/src/components/navbar.tsx
index 56b1e0346d..716c9d791f 100644
--- a/studio/frontend/src/components/navbar.tsx
+++ b/studio/frontend/src/components/navbar.tsx
@@ -7,13 +7,13 @@ export function Navbar() {
   const { isMobile } = useSidebar();
   if (!isMobile) {
     return (
-      <header className="absolute top-0 inset-x-0 z-40 h-11 pointer-events-none" />
+      <header className="absolute top-0 inset-x-0 z-40 h-[48px] pointer-events-none" />
     );
   }
   return (
-    <header className="absolute top-0 inset-x-0 z-40 h-11 pointer-events-none">
-      <div className="flex h-full items-center pl-2">
-        <SidebarTrigger className="pointer-events-auto" />
+    <header className="absolute top-0 inset-x-0 z-40 h-[48px] pointer-events-none">
+      <div className="flex h-full items-start pt-[11px] pl-2">
+        <SidebarTrigger className="pointer-events-auto !size-[34px]" />
       </div>
     </header>
   );
diff --git a/studio/frontend/src/components/ui/sidebar.tsx b/studio/frontend/src/components/ui/sidebar.tsx
index 6d9016f8e2..967b0dcaac 100644
--- a/studio/frontend/src/components/ui/sidebar.tsx
+++ b/studio/frontend/src/components/ui/sidebar.tsx
@@ -26,7 +26,7 @@ import {
 } from "@/components/ui/tooltip"
 import { useIsMobile } from "@/hooks/use-mobile"
 import { HugeiconsIcon } from "@hugeicons/react"
-import { SidebarLeftIcon } from "@hugeicons/core-free-icons"
+import { LayoutAlignLeftIcon } from "@hugeicons/core-free-icons"
 
 const noop = () => {}
 
@@ -337,7 +337,7 @@ function SidebarTrigger({
       }}
       {...props}
     >
-      <HugeiconsIcon icon={SidebarLeftIcon} strokeWidth={2} />
+      <HugeiconsIcon icon={LayoutAlignLeftIcon} strokeWidth={1.75} className="size-[18px]" />
       <span className="sr-only">Toggle Sidebar</span>
     </Button>
   )
@@ -471,7 +471,7 @@ function SidebarGroupLabel({
       data-slot="sidebar-group-label"
       data-sidebar="group-label"
       className={cn(
-        "text-[#94a3b8] dark:text-[#64748b] ring-sidebar-ring h-auto pt-3 pb-2 px-4 rounded-md text-[10px] font-semibold uppercase tracking-[0.08em] group-data-[collapsible=icon]:-mt-8 group-data-[collapsible=icon]:opacity-0 focus-visible:ring-2 [&>svg]:size-3 flex shrink-0 items-center outline-hidden [&>svg]:shrink-0",
+        "text-[#94a3b8] dark:text-[#666] ring-sidebar-ring h-auto pt-3 pb-2 px-4 rounded-md text-[10px] font-semibold uppercase tracking-[0.08em] group-data-[collapsible=icon]:-mt-8 group-data-[collapsible=icon]:opacity-0 focus-visible:ring-2 [&>svg]:size-3 flex shrink-0 items-center outline-hidden [&>svg]:shrink-0",
         className
       )}
       {...props}
@@ -536,7 +536,7 @@ function SidebarMenuItem({ className, ...props }: React.ComponentProps<"li">) {
 }
 
 const sidebarMenuButtonVariants = cva(
-  "ring-sidebar-ring hover:bg-sidebar-accent hover:text-sidebar-accent-foreground active:bg-sidebar-accent active:text-sidebar-accent-foreground data-active:bg-sidebar-accent data-active:text-sidebar-accent-foreground data-open:hover:bg-sidebar-accent data-open:hover:text-sidebar-accent-foreground gap-2 rounded-md p-2 text-left text-sm cursor-pointer transition-[width,height,padding] group-has-data-[sidebar=menu-action]/menu-item:pr-8 group-data-[collapsible=icon]:w-full! group-data-[collapsible=icon]:justify-center group-data-[collapsible=icon]:p-2! focus-visible:ring-2 data-active:font-medium peer/menu-button flex w-full items-center overflow-hidden outline-hidden group/menu-button disabled:pointer-events-none disabled:opacity-50 aria-disabled:pointer-events-none aria-disabled:opacity-50 [&>span:last-child]:truncate group-data-[collapsible=icon]:[&>span]:hidden [&_svg]:size-4 [&_svg]:shrink-0 group-data-[collapsible=icon]:[&_svg]:size-5",
+  "ring-sidebar-ring hover:bg-sidebar-accent hover:text-sidebar-accent-foreground active:bg-sidebar-accent active:text-sidebar-accent-foreground data-active:bg-sidebar-accent data-active:text-sidebar-accent-foreground data-open:hover:bg-sidebar-accent data-open:hover:text-sidebar-accent-foreground gap-2 rounded-md p-2 text-left text-sm cursor-pointer transition-[width,height,padding] group-has-data-[sidebar=menu-action]/menu-item:pr-8 group-data-[collapsible=icon]:w-full! group-data-[collapsible=icon]:justify-center group-data-[collapsible=icon]:p-2! data-active:font-medium peer/menu-button flex w-full items-center overflow-hidden outline-hidden group/menu-button disabled:pointer-events-none disabled:opacity-50 aria-disabled:pointer-events-none aria-disabled:opacity-50 [&>span:last-child]:truncate group-data-[collapsible=icon]:[&>span]:hidden [&_svg]:size-4 [&_svg]:shrink-0 group-data-[collapsible=icon]:[&_svg]:size-[18px]",
   {
     variants: {
       variant: {
diff --git a/studio/frontend/src/components/ui/slider.tsx b/studio/frontend/src/components/ui/slider.tsx
index 573dea835b..705182e5ae 100644
--- a/studio/frontend/src/components/ui/slider.tsx
+++ b/studio/frontend/src/components/ui/slider.tsx
@@ -79,7 +79,7 @@ function Slider({
     >
       <SliderPrimitive.Track
         data-slot="slider-track"
-        className="bg-muted rounded-4xl data-horizontal:h-3 data-horizontal:w-full data-vertical:h-full data-vertical:w-3 bg-muted relative grow overflow-hidden data-horizontal:w-full data-vertical:h-full cursor-pointer"
+        className="bg-black/10 dark:bg-black/12 rounded-4xl data-horizontal:h-2 data-horizontal:w-full data-vertical:h-full data-vertical:w-2 relative grow overflow-hidden cursor-pointer"
       >
         <SliderPrimitive.Range
           data-slot="slider-range"
@@ -103,7 +103,7 @@ function Slider({
         <SliderPrimitive.Thumb
           data-slot="slider-thumb"
           key={index}
-          className="border-primary ring-ring/50 relative z-10 size-4 rounded-4xl border bg-white shadow-sm block shrink-0 select-none cursor-pointer disabled:pointer-events-none disabled:opacity-50 transition-transform duration-100 ease-out hover:scale-110 hover:ring-4 active:scale-95 focus-visible:ring-4 focus-visible:outline-hidden"
+          className="ring-ring/50 relative z-10 size-4 rounded-4xl bg-white shadow-sm block shrink-0 select-none cursor-pointer disabled:pointer-events-none disabled:opacity-50 transition-transform duration-100 ease-out hover:scale-110 hover:ring-4 active:scale-95 focus-visible:ring-4 focus-visible:outline-hidden"
         />
       ))}
     </SliderPrimitive.Root>
diff --git a/studio/frontend/src/components/ui/textarea.tsx b/studio/frontend/src/components/ui/textarea.tsx
index b71e593958..36c372bdf5 100644
--- a/studio/frontend/src/components/ui/textarea.tsx
+++ b/studio/frontend/src/components/ui/textarea.tsx
@@ -1,19 +1,28 @@
 // SPDX-License-Identifier: AGPL-3.0-only
 // Copyright 2026-present the Unsloth AI Inc. team. All rights reserved. See /studio/LICENSE.AGPL-3.0
 
-import type * as React from "react";
-
-import { cn } from "@/lib/utils";
-
-function Textarea({ className, ...props }: React.ComponentProps<"textarea">) {
-  return (
-    <textarea
-      data-slot="textarea"
-      className={cn(
-        "border-input bg-input/30 focus-visible:border-ring focus-visible:ring-ring/50 aria-invalid:ring-destructive/20 dark:aria-invalid:ring-destructive/40 aria-invalid:border-destructive dark:aria-invalid:border-destructive/50 resize-none rounded-xl border px-3 py-3 text-base transition-colors focus-visible:ring-[3px] aria-invalid:ring-[3px] md:text-sm placeholder:text-muted-foreground flex field-sizing-content min-h-16 w-full outline-none disabled:cursor-not-allowed disabled:opacity-50",
-        className,
-      )}
-      {...props}
+import type * as React from "react";
+
+import { cn } from "@/lib/utils";
+
+type TextareaProps = React.ComponentProps<"textarea"> & {
+  fieldSizing?: "content" | "fixed";
+};
+
+function Textarea({
+  className,
+  fieldSizing = "content",
+  ...props
+}: TextareaProps) {
+  return (
+    <textarea
+      data-slot="textarea"
+      className={cn(
+        "border-input bg-input/30 focus-visible:border-ring focus-visible:ring-ring/50 aria-invalid:ring-destructive/20 dark:aria-invalid:ring-destructive/40 aria-invalid:border-destructive dark:aria-invalid:border-destructive/50 resize-none rounded-xl border px-3 py-3 text-base transition-colors focus-visible:ring-[3px] aria-invalid:ring-[3px] md:text-sm placeholder:text-muted-foreground flex min-h-16 min-w-0 max-w-full w-full whitespace-pre-wrap break-words [overflow-wrap:anywhere] outline-none disabled:cursor-not-allowed disabled:opacity-50",
+        fieldSizing === "content" ? "field-sizing-content" : "[field-sizing:fixed]",
+        className,
+      )}
+      {...props}
     />
   );
 }
diff --git a/studio/frontend/src/features/chat/chat-page.tsx b/studio/frontend/src/features/chat/chat-page.tsx
index 32c3b6cd16..8510aece50 100644
--- a/studio/frontend/src/features/chat/chat-page.tsx
+++ b/studio/frontend/src/features/chat/chat-page.tsx
@@ -11,9 +11,14 @@ import { cn } from "@/lib/utils";
 import { GuidedTour, useGuidedTourController } from "@/features/tour";
 import { useSidebar } from "@/components/ui/sidebar";
 import {
-  Settings04Icon,
+  Settings05Icon,
 } from "@hugeicons/core-free-icons";
 import { HugeiconsIcon } from "@hugeicons/react";
+import {
+  Tooltip,
+  TooltipContent,
+} from "@/components/ui/tooltip";
+import { Tooltip as TooltipPrimitive } from "radix-ui";
 import { useNavigate, useSearch } from "@tanstack/react-router";
 import {
   type ReactElement,
@@ -39,6 +44,7 @@ import {
 import { ChatRuntimeProvider } from "./runtime-provider";
 import {
   type CompareHandle,
+  type CompareHandles,
   CompareHandlesProvider,
   RegisterCompareHandle,
   SharedComposer,
@@ -166,6 +172,96 @@ const CompareContent = memo(function CompareContent({
   );
 });
 
+/**
+ * A single column in the compare layout. Hosts one ChatRuntimeProvider
+ * and one Thread rendered with hideComposer — the composer is shared
+ * across both panes and rendered outside the pane flex.
+ *
+ * Each pane is a flex item with `flex-1 basis-0 min-h-0 min-w-0` so on
+ * mobile (flex-col) they share height equally, and on desktop (flex-row)
+ * they share width equally. The `min-*` constraints are required for
+ * the inner viewport to scroll internally instead of spilling into the
+ * page.
+ */
+function ComparePane({
+  modelType,
+  pairId,
+  initialThreadId,
+  handleName,
+  header,
+  borderClassName,
+}: {
+  modelType: "base" | "lora" | "model1" | "model2";
+  pairId: string;
+  initialThreadId: string | undefined;
+  handleName: string;
+  header: ReactElement;
+  borderClassName?: string;
+}): ReactElement {
+  return (
+    <div
+      className={cn(
+        "flex min-h-0 min-w-0 flex-1 basis-0 flex-col overflow-hidden",
+        borderClassName,
+      )}
+    >
+      {header}
+      <div className="flex min-h-0 min-w-0 flex-1 basis-0 flex-col overflow-hidden [&_.aui-thread-viewport]:px-6 lg:[&_.aui-thread-viewport]:px-10">
+        <ChatRuntimeProvider
+          modelType={modelType}
+          pairId={pairId}
+          initialThreadId={initialThreadId}
+          syncActiveThreadId={false}
+        >
+          <RegisterCompareHandle name={handleName} />
+          <Thread hideComposer={true} hideWelcome={true} />
+        </ChatRuntimeProvider>
+      </div>
+    </div>
+  );
+}
+
+/**
+ * Shared shell for both compare variants. A vertical flex column with
+ * the two panes as siblings and the shared composer docked at the
+ * bottom. On mobile the panes stack (flex-col); on desktop they sit
+ * side by side (md:flex-row).
+ *
+ * Flex is used rather than CSS grid for the pane container so that
+ * viewport sizing stays stable across viewport-size transitions. Grid
+ * rows with 1fr were triggering resize thrash in assistant-ui's
+ * autoscroll hook on breakpoint crossings, leaving it stuck in a
+ * scroll-to-bottom loop.
+ */
+function CompareShell({
+  handlesRef,
+  children,
+  composer,
+}: {
+  handlesRef: CompareHandles;
+  children: ReactElement;
+  composer: ReactElement;
+}): ReactElement {
+  return (
+    <CompareHandlesProvider handlesRef={handlesRef}>
+      <div className="flex min-h-0 min-w-0 flex-1 basis-0 flex-col">
+        <div
+          data-tour="chat-compare-view"
+          className="flex min-h-0 min-w-0 flex-1 basis-0 flex-col md:flex-row"
+        >
+          {children}
+        </div>
+        <div className="shrink-0 bg-background px-5 pb-2 pt-1">
+          <div className="mx-auto w-full max-w-[44rem]">{composer}</div>
+          <p className="mt-1.5 text-center text-[11px] text-muted-foreground">
+            LLMs can make mistakes. Double-check all responses.
+          </p>
+        </div>
+      </div>
+    </CompareHandlesProvider>
+  );
+}
+
 /** Fast path: same model, adapter on/off, simultaneous generation. */
 const LoraCompareContent = memo(function LoraCompareContent({
   pairId,
@@ -191,61 +287,87 @@ const LoraCompareContent = memo(function LoraCompareContent({
   }, [pairId]);
 
   return (
-    <CompareHandlesProvider handlesRef={handlesRef}>
-      <div className="flex min-h-0 flex-1 flex-col">
-        <div
-          data-tour="chat-compare-view"
-          className="grid min-h-0 flex-1 grid-cols-1 px-0 md:grid-cols-2"
-        >
-          <div className="flex min-h-0 flex-col">
-            <div className="px-3 py-1.5">
+    <CompareShell
+      handlesRef={handlesRef}
+      composer={<SharedComposer handlesRef={handlesRef} />}
+    >
+      <>
+        <ComparePane
+          modelType="base"
+          pairId={pairId}
+          initialThreadId={baseThreadId}
+          handleName="base"
+          header={
+            <div className="shrink-0 px-3 py-1.5">
               <span className="text-[10px] font-semibold uppercase tracking-wider text-muted-foreground">
                 Base Model
               </span>
             </div>
-            <div className="flex min-h-0 flex-1 flex-col">
-              <ChatRuntimeProvider
-                modelType="base"
-                pairId={pairId}
-                initialThreadId={baseThreadId}
-                syncActiveThreadId={false}
-              >
-                <RegisterCompareHandle name="base" />
-                <div className="flex min-h-0 min-w-0 flex-1 basis-0 flex-col overflow-hidden">
-                  <Thread hideComposer={true} hideWelcome={true} />
-                </div>
-              </ChatRuntimeProvider>
-            </div>
-          </div>
-          <div className="flex min-h-0 flex-col border-t border-border/60 md:border-t-0 md:border-l">
-            <div className="px-3 py-1.5 text-start md:text-end">
+          }
+        />
+        <ComparePane
+          modelType="lora"
+          pairId={pairId}
+          initialThreadId={loraThreadId}
+          handleName="lora"
+          borderClassName="border-t border-border/60 md:border-t-0 md:border-l"
+          header={
+            <div className="shrink-0 px-3 py-1.5 text-start md:text-end">
               <span className="text-[10px] font-semibold uppercase tracking-wider text-primary">
                 Fine-tuned
               </span>
             </div>
-            <div className="flex min-h-0 flex-1 flex-col">
-              <ChatRuntimeProvider
-                modelType="lora"
-                pairId={pairId}
-                initialThreadId={loraThreadId}
-                syncActiveThreadId={false}
-              >
-                <RegisterCompareHandle name="lora" />
-                <div className="flex min-h-0 min-w-0 flex-1 basis-0 flex-col overflow-hidden">
-                  <Thread hideComposer={true} hideWelcome={true} />
-                </div>
-              </ChatRuntimeProvider>
-            </div>
-          </div>
-        </div>
-        <div className="z-20 mx-auto w-full max-w-4xl shrink-0 bg-background px-4 pt-2 pb-4">
-          <SharedComposer handlesRef={handlesRef} />
-        </div>
-      </div>
-    </CompareHandlesProvider>
+          }
+        />
+      </>
+    </CompareShell>
   );
 });
 
+/**
+ * Per-pane header rendered inside GeneralCompareContent. Contains the
+ * model selector aligned with the global topbar height. The left pane
+ * reserves room for the mobile sidebar trigger; the right pane reserves
+ * room for the global settings button.
+ */
+function GeneralCompareHeader({
+  models,
+  loraModels,
+  value,
+  onValueChange,
+  onFoldersChange,
+  side,
+}: {
+  models: ModelOption[];
+  loraModels: LoraModelOption[];
+  value: string;
+  onValueChange: (
+    id: string,
+    meta: { isLora: boolean; ggufVariant?: string },
+  ) => void;
+  onFoldersChange?: () => void;
+  side: "left" | "right";
+}): ReactElement {
+  return (
+    <div
+      className={cn(
+        "flex h-[48px] shrink-0 items-center gap-2 bg-background",
+        side === "left" ? "pl-12 pr-3 md:pl-2" : "pl-3 pr-12",
+      )}
+    >
+      <ModelSelector
+        models={models}
+        loraModels={loraModels}
+        value={value}
+        onValueChange={onValueChange}
+        onFoldersChange={onFoldersChange}
+        variant="ghost"
+        className="max-w-[80%] !h-[34px]"
+      />
+    </div>
+  );
+}
+
 /** General path: any two models, sequential load → generate. */
 const GeneralCompareContent = memo(function GeneralCompareContent({
   pairId,
@@ -299,88 +421,64 @@ const GeneralCompareContent = memo(function GeneralCompareContent({
   }, [pairId]);
 
   return (
-    <CompareHandlesProvider handlesRef={handlesRef}>
-      <div className="flex min-h-0 flex-1 flex-col">
-        <div
-          data-tour="chat-compare-view"
-          className="grid min-h-0 flex-1 grid-cols-1 px-0 md:grid-cols-2"
-        >
-          <div className="flex min-h-0 flex-col">
-            <div className="flex h-11 shrink-0 items-center gap-2 px-3">
-              <ModelSelector
-                models={models}
-                loraModels={loraModels}
-                value={model1.id}
-                onValueChange={(id, meta) =>
-                  setModel1({
-                    id,
-                    isLora: meta.isLora,
-                    ggufVariant: meta.ggufVariant,
-                  })
-                }
-                onFoldersChange={onFoldersChange}
-                variant="ghost"
-                size="sm"
-                className="max-w-[80%]"
-              />
-            </div>
-            <div className="flex min-h-0 flex-1 flex-col">
-              <ChatRuntimeProvider
-                modelType="model1"
-                pairId={pairId}
-                initialThreadId={model1ThreadId}
-                syncActiveThreadId={false}
-              >
-                <RegisterCompareHandle name="model1" />
-                <div className="flex min-h-0 min-w-0 flex-1 basis-0 flex-col overflow-hidden">
-                  <Thread hideComposer={true} hideWelcome={true} />
-                </div>
-              </ChatRuntimeProvider>
-            </div>
-          </div>
-          <div className="flex min-h-0 flex-col border-t border-sidebar-border md:border-t-0 md:border-l">
-            <div className="flex h-11 shrink-0 items-center gap-2 px-3">
-              <ModelSelector
-                models={models}
-                loraModels={loraModels}
-                value={model2.id}
-                onValueChange={(id, meta) =>
-                  setModel2({
-                    id,
-                    isLora: meta.isLora,
-                    ggufVariant: meta.ggufVariant,
-                  })
-                }
-                onFoldersChange={onFoldersChange}
-                variant="ghost"
-                size="sm"
-                className="max-w-[80%]"
-              />
-            </div>
-            <div className="flex min-h-0 flex-1 flex-col">
-              <ChatRuntimeProvider
-                modelType="model2"
-                pairId={pairId}
-                initialThreadId={model2ThreadId}
-                syncActiveThreadId={false}
-              >
-                <RegisterCompareHandle name="model2" />
-                <div className="flex min-h-0 min-w-0 flex-1 basis-0 flex-col overflow-hidden">
-                  <Thread hideComposer={true} hideWelcome={true} />
-                </div>
-              </ChatRuntimeProvider>
-            </div>
-          </div>
-        </div>
-        <div className="z-20 mx-auto w-full max-w-4xl shrink-0 bg-background px-4 pt-2 pb-4">
-          <SharedComposer
-            handlesRef={handlesRef}
-            model1={model1}
-            model2={model2}
-          />
-        </div>
-      </div>
-    </CompareHandlesProvider>
+    <CompareShell
+      handlesRef={handlesRef}
+      composer={
+        <SharedComposer
+          handlesRef={handlesRef}
+          model1={model1}
+          model2={model2}
+        />
+      }
+    >
+      <>
+        <ComparePane
+          modelType="model1"
+          pairId={pairId}
+          initialThreadId={model1ThreadId}
+          handleName="model1"
+          header={
+            <GeneralCompareHeader
+              side="left"
+              models={models}
+              loraModels={loraModels}
+              value={model1.id}
+              onValueChange={(id, meta) =>
+                setModel1({
+                  id,
+                  isLora: meta.isLora,
+                  ggufVariant: meta.ggufVariant,
+                })
+              }
+              onFoldersChange={onFoldersChange}
+            />
+          }
+        />
+        <ComparePane
+          modelType="model2"
+          pairId={pairId}
+          initialThreadId={model2ThreadId}
+          handleName="model2"
+          borderClassName="border-t border-sidebar-border md:border-t-0 md:border-l"
+          header={
+            <GeneralCompareHeader
+              side="right"
+              models={models}
+              loraModels={loraModels}
+              value={model2.id}
+              onValueChange={(id, meta) =>
+                setModel2({
+                  id,
+                  isLora: meta.isLora,
+                  ggufVariant: meta.ggufVariant,
+                })
+              }
+              onFoldersChange={onFoldersChange}
+            />
+          }
+        />
+      </>
+    </CompareShell>
   );
 });
 
@@ -734,9 +832,9 @@ export function ChatPage(): ReactElement {
       <div className="relative flex min-h-0 min-w-0 flex-1 basis-0 flex-col overflow-hidden">
         <div
           className={cn(
-            "absolute top-0 left-0 right-2 z-30 flex h-11 shrink-0 items-center pr-2 bg-background",
+            "absolute top-0 left-0 right-[10px] z-30 flex h-[48px] shrink-0 items-start pt-[11px] pr-2 bg-background",
             isMobile ? "pl-12 pr-1.5" : "pl-2",
-            view.mode === "compare" && "right-2 left-auto w-auto bg-transparent pl-0 pr-2",
+            view.mode === "compare" && "right-[10px] left-auto w-auto bg-transparent pl-0 pr-2",
           )}
         >
           <div className="flex items-center gap-1">
@@ -754,7 +852,7 @@ export function ChatPage(): ReactElement {
                 onOpenChange={handleModelSelectorOpenChange}
                 triggerDataTour="chat-model-selector"
                 contentDataTour="chat-model-selector-popover"
-                className="max-w-[62vw] sm:max-w-none"
+                className="max-w-[62vw] sm:max-w-none !h-[34px]"
               />
             )}
             {loadingModel && loadToastDismissed ? (
@@ -784,25 +882,36 @@ export function ChatPage(): ReactElement {
               {modelsError}
             </div>
           )}
-          <div className="flex-1" />
-          {view.mode === "single" && ggufContextLength && contextUsage ? (
-            <ContextUsageBar
-              used={contextUsage.totalTokens}
-              total={ggufContextLength}
-              cached={contextUsage.cachedTokens}
-              promptTokens={contextUsage.promptTokens}
-              completionTokens={contextUsage.completionTokens}
-            />
-          ) : null}
-          <button
-            type="button"
-            onClick={() => setSettingsOpen(!settingsOpen)}
-            className="flex h-9 w-9 items-center justify-center rounded-md text-muted-foreground transition-colors hover:bg-accent hover:text-foreground"
-            title="Inference settings"
-            data-tour="chat-settings"
-          >
-            <HugeiconsIcon icon={Settings04Icon} className="size-5" />
-          </button>
+          <div className="ml-auto flex items-center gap-2">
+            {view.mode === "single" && ggufContextLength && contextUsage ? (
+              <ContextUsageBar
+                used={contextUsage.totalTokens}
+                total={ggufContextLength}
+                cached={contextUsage.cachedTokens}
+                promptTokens={contextUsage.promptTokens}
+                completionTokens={contextUsage.completionTokens}
+                className="h-[34px]"
+              />
+            ) : null}
+            {!settingsOpen && (
+              <Tooltip>
+                <TooltipPrimitive.Trigger asChild>
+                  <button
+                    type="button"
+                    onClick={() => setSettingsOpen(true)}
+                    className="flex h-[34px] w-[34px] items-center justify-center rounded-[8px] text-[#383835] dark:text-[#c7c7c4] transition-colors hover:bg-[#ececec] dark:hover:bg-[#2e3035] hover:text-black dark:hover:text-white focus-visible:outline-none focus-visible:ring-2 focus-visible:ring-ring"
+                    aria-label="Open configuration"
+                    data-tour="chat-settings"
+                  >
+                    <HugeiconsIcon icon={Settings05Icon} className="size-5" />
+                  </button>
+                </TooltipPrimitive.Trigger>
+                <TooltipContent side="bottom" sideOffset={6}>
+                  Open configuration
+                </TooltipContent>
+              </Tooltip>
+            )}
+          </div>
         </div>
 
         {view.mode === "single" ? (
diff --git a/studio/frontend/src/features/chat/chat-settings-sheet.tsx b/studio/frontend/src/features/chat/chat-settings-sheet.tsx
index 2e03591a8e..24a9ac4077 100644
--- a/studio/frontend/src/features/chat/chat-settings-sheet.tsx
+++ b/studio/frontend/src/features/chat/chat-settings-sheet.tsx
@@ -52,12 +52,17 @@ import {
   CodeIcon,
   Delete02Icon,
   FloppyDiskIcon,
-  PencilEdit01Icon,
   Settings02Icon,
+  Settings05Icon,
   SlidersHorizontalIcon,
   Wrench01Icon,
 } from "@hugeicons/core-free-icons";
 import { HugeiconsIcon } from "@hugeicons/react";
+import {
+  Tooltip,
+  TooltipContent,
+} from "@/components/ui/tooltip";
+import { Tooltip as TooltipPrimitive } from "radix-ui";
 import { AnimatePresence, motion } from "motion/react";
 import type { ReactNode } from "react";
 import { useEffect, useLayoutEffect, useMemo, useRef, useState } from "react";
@@ -738,15 +743,34 @@ export function ChatSettingsPanel({
 
   const settingsContent = (
     <>
-      <div className="aui-thread-viewport relative h-full overflow-y-auto bg-muted/70">
-      <div className="sticky top-0 z-10 flex items-center gap-2 bg-muted/70 px-4 py-3 backdrop-blur">
-        <HugeiconsIcon
-          icon={PencilEdit01Icon}
-          className="size-4 text-muted-foreground/70"
-        />
-        <span className="flex-1 text-base font-semibold tracking-tight">
-          Configuration
-        </span>
+      <div className="aui-thread-viewport relative h-full overflow-y-auto">
+      <div className="sticky top-0 z-10 flex h-[48px] items-start gap-2 pl-2 pr-2 pt-[11px] backdrop-blur">
+        {isMobile ? (
+          <span className="flex h-[34px] flex-1 items-center pl-1 text-base font-semibold tracking-tight">
+            Configuration
+          </span>
+        ) : (
+          <>
+            <Tooltip>
+              <TooltipPrimitive.Trigger asChild>
+                <button
+                  type="button"
+                  onClick={() => onOpenChange?.(false)}
+                  className="flex h-[34px] w-[34px] items-center justify-center rounded-[8px] text-[#383835] dark:text-[#c7c7c4] transition-colors hover:bg-[#ececec] dark:hover:bg-[#2e3035] hover:text-black dark:hover:text-white focus-visible:outline-none focus-visible:ring-2 focus-visible:ring-ring"
+                  aria-label="Close configuration"
+                >
+                  <HugeiconsIcon icon={Settings05Icon} className="size-5" />
+                </button>
+              </TooltipPrimitive.Trigger>
+              <TooltipContent side="bottom" sideOffset={6}>
+                Close configuration
+              </TooltipContent>
+            </Tooltip>
+            <span className="flex h-[34px] flex-1 items-center text-base font-semibold tracking-tight">
+              Configuration
+            </span>
+          </>
+        )}
       </div>
 
       <div className="px-1.5">
@@ -889,7 +913,7 @@ export function ChatSettingsPanel({
             value={params.systemPrompt}
             onChange={(e) => set("systemPrompt")(e.target.value)}
             placeholder="You are a helpful assistant..."
-            className="min-h-20 max-h-48 overflow-y-auto text-xs corner-squircle"
+            className="min-h-20 max-h-48 overflow-y-auto text-xs corner-squircle focus-visible:ring-[1px]"
             rows={3}
           />
         </div>
@@ -1202,6 +1226,7 @@ export function ChatSettingsPanel({
               value={systemPromptDraft}
               onChange={(event) => setSystemPromptDraft(event.target.value)}
               placeholder="You are a helpful assistant..."
+              fieldSizing="fixed"
               className="min-h-[24rem] max-h-[50vh] overflow-y-auto text-sm leading-6 corner-squircle"
               rows={14}
             />
diff --git a/studio/frontend/src/features/chat/shared-composer.tsx b/studio/frontend/src/features/chat/shared-composer.tsx
index 1c69ade6bd..e133669a1c 100644
--- a/studio/frontend/src/features/chat/shared-composer.tsx
+++ b/studio/frontend/src/features/chat/shared-composer.tsx
@@ -2,11 +2,12 @@
 // Copyright 2026-present the Unsloth AI Inc. team. All rights reserved. See /studio/LICENSE.AGPL-3.0
 
 import { TooltipIconButton } from "@/components/assistant-ui/tooltip-icon-button";
+import { CodeToggleIcon } from "@/components/assistant-ui/code-toggle-icon";
 import { Button } from "@/components/ui/button";
 import { AUDIO_ACCEPT, MAX_AUDIO_SIZE, fileToBase64 } from "@/lib/audio-utils";
 import { useAui } from "@assistant-ui/react";
 import { cn } from "@/lib/utils";
-import { ArrowUpIcon, GlobeIcon, HeadphonesIcon, LightbulbIcon, LightbulbOffIcon, MicIcon, PlusIcon, SquareIcon, TerminalIcon, XIcon } from "lucide-react";
+import { ArrowUpIcon, GlobeIcon, HeadphonesIcon, LightbulbIcon, LightbulbOffIcon, MicIcon, PlusIcon, SquareIcon, XIcon } from "lucide-react";
 import { toast } from "sonner";
 import { loadModel, validateModel } from "./api/chat-api";
 import { useChatRuntimeStore } from "./stores/chat-runtime-store";
@@ -267,6 +268,21 @@ export function SharedComposer({
     return () => clearInterval(id);
   }, [handlesRef]);
 
+  // Auto-expand textarea up to 6 rows, then scroll (matches regular chat composer).
+  useEffect(() => {
+    const ta = textareaRef.current;
+    if (!ta) return;
+    ta.style.height = "auto";
+    const styles = window.getComputedStyle(ta);
+    const lineHeight = parseFloat(styles.lineHeight) || 20;
+    const paddingY = parseFloat(styles.paddingTop) + parseFloat(styles.paddingBottom);
+    const borderY = parseFloat(styles.borderTopWidth) + parseFloat(styles.borderBottomWidth);
+    const maxHeight = lineHeight * 6 + paddingY + borderY;
+    const next = Math.min(ta.scrollHeight, maxHeight);
+    ta.style.height = `${next}px`;
+    ta.style.overflowY = ta.scrollHeight > maxHeight ? "auto" : "hidden";
+  }, [text]);
+
   const addFiles = useCallback((files: FileList | null) => {
     if (!files?.length) return;
     const next: PendingImage[] = [];
@@ -455,7 +471,7 @@ export function SharedComposer({
 
   return (
     <div
-      className={`chat-composer-surface relative flex w-full flex-col rounded-3xl bg-background px-1 pt-2 transition-shadow outline-none ${dragging ? "border-ring bg-accent/50" : ""}`}
+      className={`chat-composer-surface relative flex w-full flex-col rounded-3xl bg-background dark:bg-card px-1 pt-2 transition-shadow outline-none ${dragging ? "border-ring bg-accent/50" : ""}`}
       onDragOver={(e) => {
         e.preventDefault();
         setDragging(true);
@@ -498,7 +514,7 @@ export function SharedComposer({
         onChange={(e) => setText(e.target.value)}
         onKeyDown={onKeyDown}
         placeholder="Send to both models..."
-        className="mb-1 max-h-32 min-h-14 w-full resize-none bg-transparent pl-5 pr-4 pt-2 pb-3 text-sm outline-none placeholder:text-muted-foreground"
+        className="mb-1 min-h-12 w-full resize-none overflow-y-hidden bg-transparent pl-5 pr-4 pt-2 pb-3 text-sm font-[450] outline-none placeholder:text-muted-foreground focus-visible:ring-0"
         rows={1}
       />
       <div className="relative mx-2 mb-2 flex items-center justify-between">
@@ -515,13 +531,13 @@ export function SharedComposer({
             }}
           />
           <TooltipIconButton
-            tooltip="Add attachment"
+            tooltip="Add Attachment"
             side="bottom"
             variant="ghost"
             size="icon"
-            className="size-8 rounded-full text-muted-foreground hover:bg-muted-foreground/15"
+            className="size-8.5 rounded-full p-1 font-semibold text-xs hover:bg-muted-foreground/15 dark:border-muted-foreground/15 dark:hover:bg-muted-foreground/30"
             onClick={() => fileInputRef.current?.click()}
-            aria-label="Add attachment"
+            aria-label="Add Attachment"
           >
             <PlusIcon className="size-5 stroke-[1.5px]" />
           </TooltipIconButton>
@@ -542,11 +558,11 @@ export function SharedComposer({
                 side="bottom"
                 variant="ghost"
                 size="icon"
-                className="size-8 rounded-full text-muted-foreground hover:bg-muted-foreground/15"
+                className="size-8.5 rounded-full p-1 text-muted-foreground hover:bg-muted-foreground/15"
                 onClick={() => audioInputRef.current?.click()}
                 aria-label="Upload audio"
               >
-                <HeadphonesIcon className="size-4 stroke-[1.5px]" />
+                <HeadphonesIcon className="size-4.5 stroke-[1.5px]" />
               </TooltipIconButton>
             </>
           )}
@@ -569,7 +585,7 @@ export function SharedComposer({
               }
             }}
             className={cn(
-              "flex items-center gap-0.5 rounded-full px-2 py-0.5 text-xs font-medium transition-colors",
+              "flex items-center gap-1.5 rounded-full px-2.5 py-1 text-xs font-medium transition-colors",
               reasoningDisabled
                 ? "cursor-not-allowed opacity-40"
                 : (reasoningEnabled || reasoningAlwaysOn)
@@ -579,9 +595,9 @@ export function SharedComposer({
             aria-label={reasoningEnabled ? "Disable thinking" : "Enable thinking"}
           >
             {(reasoningEnabled || reasoningAlwaysOn) && !reasoningDisabled ? (
-              <LightbulbIcon className="size-3" />
+              <LightbulbIcon className="size-3.5" />
             ) : (
-              <LightbulbOffIcon className="size-3" />
+              <LightbulbOffIcon className="size-3.5" />
             )}
             <span>Think</span>
           </button>
@@ -616,7 +632,7 @@ export function SharedComposer({
             )}
             aria-label={codeToolsEnabled ? "Disable code execution" : "Enable code execution"}
           >
-            <TerminalIcon className="size-3.5" />
+            <CodeToggleIcon className="size-3.5" />
             <span>Code</span>
           </button>
         </div>
@@ -629,7 +645,7 @@ export function SharedComposer({
                   side="bottom"
                   variant="ghost"
                   size="icon"
-                  className="size-8 rounded-full text-muted-foreground hover:bg-muted-foreground/15"
+                  className="size-8 rounded-full text-muted-foreground"
                   onClick={startDictation}
                   aria-label="Dictate"
                 >
diff --git a/studio/frontend/src/features/recipe-studio/components/recipe-graph-aux-node.tsx b/studio/frontend/src/features/recipe-studio/components/recipe-graph-aux-node.tsx
index 904c5e9df3..003eb6837f 100644
--- a/studio/frontend/src/features/recipe-studio/components/recipe-graph-aux-node.tsx
+++ b/studio/frontend/src/features/recipe-studio/components/recipe-graph-aux-node.tsx
@@ -132,7 +132,7 @@ function AuxNodeBase({
     const hasInvalidRefs =
       findInvalidJinjaReferences(value, availableRefs).length > 0;
     return (
-      <BaseNode className="corner-squircle w-full min-w-0 rounded-lg border-border/60 bg-card shadow-sm">
+      <BaseNode className="corner-squircle w-full min-w-0 rounded-4xl border-border/60 bg-card shadow-sm">
         <BaseNodeHeader className="border-b border-border/50 px-3 py-2">
           <BaseNodeHeaderTitle className="text-xs">{data.title}</BaseNodeHeaderTitle>
         </BaseNodeHeader>
@@ -195,7 +195,7 @@ function AuxNodeBase({
   };
 
   return (
-    <BaseNode className="corner-squircle w-full min-w-0 rounded-lg border-border/60 bg-card shadow-sm">
+    <BaseNode className="corner-squircle w-full min-w-0 rounded-4xl border-border/60 bg-card shadow-sm">
       <BaseNodeHeader className="border-b border-border/50 px-3 py-2">
         <BaseNodeHeaderTitle className="text-xs">
           {score.name.trim() || `Scorer ${data.scoreIndex + 1}`}
diff --git a/studio/frontend/src/features/recipe-studio/components/recipe-graph-node.tsx b/studio/frontend/src/features/recipe-studio/components/recipe-graph-node.tsx
index 8afc4e26bf..0a7cdb0f71 100644
--- a/studio/frontend/src/features/recipe-studio/components/recipe-graph-node.tsx
+++ b/studio/frontend/src/features/recipe-studio/components/recipe-graph-node.tsx
@@ -399,7 +399,7 @@ function RecipeGraphNodeBase({
 
     return (
       <BaseNode
-        className="corner-squircle relative w-full min-w-0 overflow-visible rounded-lg border-border/60 shadow-sm"
+        className="corner-squircle relative w-full min-w-0 overflow-visible rounded-4xl border-border/60 shadow-sm"
         style={noteStyle}
       >
         <NodeResizer
@@ -465,7 +465,7 @@ function RecipeGraphNodeBase({
   return (
     <BaseNode
       className={cn(
-        "corner-squircle relative w-full min-w-0 overflow-visible rounded-lg border-border/60 shadow-sm",
+        "corner-squircle relative w-full min-w-0 overflow-visible rounded-4xl border-border/60 shadow-sm",
         runtimeNodeTone,
         hasConnectionIssue &&
           runtimeState === "idle" &&
diff --git a/studio/frontend/src/features/settings/settings-dialog.tsx b/studio/frontend/src/features/settings/settings-dialog.tsx
index cdccf2fc72..bb22fcb89c 100644
--- a/studio/frontend/src/features/settings/settings-dialog.tsx
+++ b/studio/frontend/src/features/settings/settings-dialog.tsx
@@ -83,7 +83,7 @@ export function SettingsDialog() {
           Manage your Unsloth Studio preferences.
         </DialogDescription>
         <div className="flex h-full min-h-0">
-          <aside className="flex w-[200px] shrink-0 flex-col border-r border-border bg-muted/20 p-2">
+          <aside className="font-heading flex w-[200px] shrink-0 flex-col border-r border-border bg-muted/20 p-2">
             <nav className="flex flex-col gap-0.5">
               {TABS.map((tab) => {
                 const active = activeTab === tab.id;
@@ -93,17 +93,17 @@ export function SettingsDialog() {
                     type="button"
                     onClick={() => setActiveTab(tab.id)}
                     className={cn(
-                      "relative flex h-9 items-center gap-2 rounded-md px-2.5 text-sm font-medium transition-colors",
+                      "relative flex h-[30px] items-center gap-2.5 rounded-[8px] px-2.5 text-sm font-medium transition-colors",
                       "focus-visible:outline-none focus-visible:ring-2 focus-visible:ring-ring focus-visible:ring-offset-1 focus-visible:ring-offset-background",
                       active
-                        ? "text-foreground"
-                        : "text-muted-foreground hover:text-foreground",
+                        ? "text-black dark:text-white"
+                        : "text-[#383835] dark:text-[#c7c7c4] hover:bg-[#ececec] dark:hover:bg-[#2e3035] hover:text-black dark:hover:text-white",
                     )}
                   >
                     {active && (
                       <motion.span
                         layoutId="settings-active-pill"
-                        className="absolute inset-0 rounded-md bg-accent"
+                        className="absolute inset-0 rounded-[8px] bg-[#ececec] dark:bg-[#2e3035]"
                         transition={
                           reduced
                             ? { duration: 0 }
@@ -118,7 +118,8 @@ export function SettingsDialog() {
                     )}
                     <HugeiconsIcon
                       icon={tab.icon}
-                      className="relative z-10 size-4"
+                      strokeWidth={1.5}
+                      className="relative z-10 size-[18px]"
                     />
                     <span className="relative z-10">{tab.label}</span>
                   </button>
@@ -131,7 +132,7 @@ export function SettingsDialog() {
             <button
               type="button"
               onClick={closeDialog}
-              className="absolute top-3 right-3 z-10 flex size-7 items-center justify-center rounded-md text-muted-foreground transition-colors hover:bg-accent hover:text-foreground focus-visible:outline-none focus-visible:ring-2 focus-visible:ring-ring"
+              className="absolute top-3 right-3 z-10 flex size-7 items-center justify-center rounded-[8px] text-[#383835] dark:text-[#c7c7c4] transition-colors hover:bg-[#ececec] dark:hover:bg-[#2e3035] hover:text-black dark:hover:text-white focus-visible:outline-none focus-visible:ring-2 focus-visible:ring-ring"
               aria-label="Close settings"
             >
               <HugeiconsIcon icon={Cancel01Icon} className="size-4" />
diff --git a/studio/frontend/src/hooks/index.ts b/studio/frontend/src/hooks/index.ts
index 2f64d3249c..f889dc7445 100644
--- a/studio/frontend/src/hooks/index.ts
+++ b/studio/frontend/src/hooks/index.ts
@@ -11,3 +11,4 @@ export { useHfDatasetSearch } from "./use-hf-dataset-search";
 export { useHfDatasetSplits } from "./use-hf-dataset-splits";
 export { useHfTokenValidation } from "./use-hf-token-validation";
 export { useInfiniteScroll } from "./use-infinite-scroll";
+export { useCollapseScrollLock } from "./use-collapse-scroll-lock";
diff --git a/studio/frontend/src/hooks/use-collapse-scroll-lock.ts b/studio/frontend/src/hooks/use-collapse-scroll-lock.ts
new file mode 100644
index 0000000000..fb7baa062c
--- /dev/null
+++ b/studio/frontend/src/hooks/use-collapse-scroll-lock.ts
@@ -0,0 +1,63 @@
+// SPDX-License-Identifier: AGPL-3.0-only
+// Copyright 2026-present the Unsloth AI Inc. team. All rights reserved. See /studio/LICENSE.AGPL-3.0
+
+import { type RefObject, useCallback, useEffect, useRef } from "react";
+
+/**
+ * Locks the nearest scrollable ancestor's scrollTop for the duration of a
+ * collapsible animation so the page doesn't jump when content height changes.
+ *
+ * Unlike @assistant-ui/react's `useScrollLock`, this hook does NOT toggle
+ * `scrollbar-width: none` on the container. Hiding the scrollbar mid-animation
+ * caused a visible disappear/reappear flicker on tool-call collapsibles.
+ */
+export function useCollapseScrollLock(
+  animatedElementRef: RefObject<HTMLElement | null>,
+  animationDurationMs: number,
+): () => void {
+  const cleanupRef = useRef<(() => void) | null>(null);
+
+  useEffect(() => {
+    return () => {
+      cleanupRef.current?.();
+    };
+  }, []);
+
+  return useCallback(() => {
+    cleanupRef.current?.();
+
+    const animatedElement = animatedElementRef.current;
+    if (!animatedElement) return;
+
+    let scrollContainer: HTMLElement | null = animatedElement;
+    while (scrollContainer) {
+      const { overflowY } = getComputedStyle(scrollContainer);
+      if (overflowY === "scroll" || overflowY === "auto") {
+        break;
+      }
+      scrollContainer = scrollContainer.parentElement;
+    }
+    if (!scrollContainer) return;
+
+    const container = scrollContainer;
+    const scrollPosition = container.scrollTop;
+    const resetPosition = () => {
+      container.scrollTop = scrollPosition;
+    };
+
+    container.addEventListener("scroll", resetPosition);
+    let timeoutId: ReturnType<typeof setTimeout> | null = null;
+    const cleanup = () => {
+      if (timeoutId !== null) {
+        clearTimeout(timeoutId);
+        timeoutId = null;
+      }
+      container.removeEventListener("scroll", resetPosition);
+      if (cleanupRef.current === cleanup) {
+        cleanupRef.current = null;
+      }
+    };
+    timeoutId = setTimeout(cleanup, animationDurationMs);
+    cleanupRef.current = cleanup;
+  }, [animatedElementRef, animationDurationMs]);
+}
diff --git a/studio/frontend/src/index.css b/studio/frontend/src/index.css
index dd559a3765..3943c7f882 100644
--- a/studio/frontend/src/index.css
+++ b/studio/frontend/src/index.css
@@ -14,6 +14,22 @@
 
 @custom-variant dark (&:is(.dark *));
 
+@font-face {
+	font-family: "Hellix";
+	src: url("/fonts/Hellix-Regular.woff") format("woff");
+	font-weight: 400;
+	font-style: normal;
+	font-display: swap;
+}
+
+@font-face {
+	font-family: "Hellix";
+	src: url("/fonts/Hellix-Medium.woff") format("woff");
+	font-weight: 500;
+	font-style: normal;
+	font-display: swap;
+}
+
 @font-face {
 	font-family: "Hellix";
 	src: url("/fonts/Hellix-SemiBold.woff2") format("woff2"),
@@ -23,6 +39,14 @@
 	font-display: swap;
 }
 
+@font-face {
+	font-family: "Fira Code";
+	src: url("/fonts/FiraCode-VariableFont_wght.ttf") format("truetype-variations");
+	font-weight: 300 700;
+	font-style: normal;
+	font-display: swap;
+}
+
 :root {
 	/* Animation timing */
 	--duration-micro: 100ms;
@@ -100,39 +124,44 @@
 }
 
 .dark {
-	--background: oklch(0.24 0 0);
-	--foreground: oklch(0.98 0 0);
-	--card: oklch(0.28 0 0);
-	--card-foreground: oklch(0.98 0 0);
-	--popover: oklch(0.28 0 0);
-	--popover-foreground: oklch(0.98 0 0);
+	/* Exact palette from apps/studio/index_chat.html mockup. Using hex so the
+	   rendered surface matches the mockup pixel-for-pixel — OKLCH conversion
+	   drifted ~3% darker and shifted the neutral hue. */
+	--background: #1a1b1e;
+	--foreground: #d4d4d4;
+	--card: #222427;
+	--card-foreground: #d4d4d4;
+	--popover: #222427;
+	--popover-foreground: #d4d4d4;
 	--primary: oklch(0.6929 0.1396 166.5513);
 	--primary-foreground: oklch(1 0 0);
-	--secondary: oklch(0.33 0 0);
-	--secondary-foreground: oklch(0.98 0 0);
-	--muted: oklch(0.33 0 0);
-	--muted-foreground: oklch(0.70 0 0);
-	--accent: oklch(0.33 0 0);
-	--accent-foreground: oklch(0.98 0 0);
+	--secondary: #2e3035;
+	--secondary-foreground: #d4d4d4;
+	--muted: #2e3035;
+	--muted-foreground: #999999;
+	--accent: #2e3035;
+	--accent-foreground: #d4d4d4;
 	--destructive: oklch(0.6368 0.2078 25.3313);
-	--border: oklch(0.38 0 0);
-	--input: oklch(0.38 0 0);
+	--border: #2e3035;
+	/* --input one step lighter than --muted so form borders stay visible on
+	   muted surfaces (right config panel) and keep subtle contrast on card. */
+	--input: #3a3d42;
 	--ring: oklch(0.6929 0.1396 166.5513);
 	--chart-1: oklch(0.7511 0.1407 166.2284);
 	--chart-2: oklch(0.75 0.14 136.5572);
 	--chart-3: oklch(0.7554 0.1285 197.339);
 	--chart-4: oklch(0.7503 0.1199 346.7805);
 	--chart-5: oklch(0.799 0.1196 84.6633);
-	--sidebar: oklch(0.24 0 0);
-	--sidebar-foreground: oklch(0.98 0 0);
+	--sidebar: #222427;
+	--sidebar-foreground: #d4d4d4;
 	--sidebar-primary: oklch(0.6929 0.1396 166.5513);
 	--sidebar-primary-foreground: oklch(1 0 0);
-	--sidebar-accent: oklch(0.33 0 0);
-	--sidebar-accent-foreground: oklch(0.98 0 0);
-	--sidebar-border: oklch(0.38 0 0);
+	--sidebar-accent: #2e3035;
+	--sidebar-accent-foreground: #d4d4d4;
+	--sidebar-border: #2e3035;
 	--sidebar-ring: oklch(0.6929 0.1396 166.5513);
 	--destructive-foreground: oklch(1 0 0);
-	--radius: 1.1rem;
+	--radius: 0.625rem;
 	--font-sans: Geist, ui-sans-serif, sans-serif, system-ui;
 	--font-serif: Source Serif 4, serif;
 	--font-mono: JetBrains Mono, monospace;
@@ -305,9 +334,16 @@
 
 @layer utilities {
 
-	/* Heading font utility */
+	/* Heading font utility — matches the sidebar logo's letter-spacing so
+	   all Hellix text (select model, menus, nav items) stays visually
+	   consistent, with a small bump in dark mode to offset optical bloom. */
 	.font-heading {
 		font-family: var(--font-heading);
+		letter-spacing: -0.01em;
+	}
+
+	.dark .font-heading {
+		letter-spacing: 0.02em;
 	}
 
 	/* Elevated surface shadow (use ring-* for borders) */
@@ -332,11 +368,8 @@
 	}
 
 	.dark .chat-composer-surface {
-		border-color: oklch(0.38 0 0 / 1);
-		box-shadow:
-			0 1px 2px oklch(0 0 0 / 0.2),
-			0 8px 20px oklch(0 0 0 / 0.25),
-			0 22px 48px oklch(0 0 0 / 0.22);
+		border-color: #2e3035;
+		box-shadow: 0 2px 12px rgba(0, 0, 0, 0.2);
 	}
 
 	/* Fine-tuning Studio: equal default height, expandable when needed (md+) */
@@ -375,8 +408,8 @@
 	}
 
 	[data-streamdown="code-block"] {
-		gap: 0;
-		padding: 0.5rem;
+		gap: 0.25rem;
+		padding: 0.75rem 1rem;
 		/* Wide lines must scroll inside the thread column, not widen past the composer (flex min-width:auto). */
 		max-width: 100%;
 		min-width: 0;
@@ -384,7 +417,14 @@
 	}
 
 	[data-streamdown="code-block-header"] {
-		padding-left: 0.75rem;
+		padding-left: 0;
+	}
+
+	.aui-thread-root [data-streamdown="code-block"] code > span::before {
+		content: none !important;
+		display: none !important;
+		margin: 0 !important;
+		width: 0 !important;
 	}
 
 	/* Chat thread: code slightly smaller by default; step up when the thread column is wide. */
@@ -416,11 +456,11 @@
 	/* Keep monospace for code fences and inline code (not KaTeX). */
 	.aui-thread-root [data-streamdown="code-block"] pre,
 	.aui-thread-root [data-streamdown="code-block"] code {
-		font-family: var(--font-mono), ui-monospace, monospace;
+		font-family: "Fira Code", ui-monospace, monospace;
 	}
 
 	.aui-thread-root :where(p, li, td, th, blockquote, h1, h2, h3, h4, h5, h6) code {
-		font-family: var(--font-mono), ui-monospace, monospace;
+		font-family: "Fira Code", ui-monospace, monospace;
 	}
 
 	/* Align fenced code blocks with the main chat column even when nested in lists. */
@@ -478,6 +518,9 @@
    full-height rail flush to the right edge, without a separate decorative strip. */
 .aui-thread-viewport {
 	scrollbar-color: oklch(0.5 0 0 / 0.54) var(--sidebar);
+	/* Reserve scrollbar space always so absolute-positioned overlays (topbar)
+	   can stop flush with the gutter edge without covering the scrollbar. */
+	scrollbar-gutter: stable;
 }
 
 .dark .aui-thread-viewport {
@@ -489,11 +532,19 @@
 }
 
 [data-sidebar="content"] {
-	scrollbar-color: oklch(0.5 0 0 / 0.7) transparent;
+	scrollbar-color: oklch(0.5 0 0 / 0.22) transparent;
 }
 
 .dark [data-sidebar="content"] {
-	scrollbar-color: oklch(0.72 0 0 / 0.65) transparent;
+	scrollbar-color: oklch(0.72 0 0 / 0.25) transparent;
+}
+
+[data-sidebar="content"]::-webkit-scrollbar-thumb {
+	background: oklch(0.5 0 0 / 0.22);
+}
+
+.dark [data-sidebar="content"]::-webkit-scrollbar-thumb {
+	background: oklch(0.72 0 0 / 0.25);
 }
 
 /*---break---*/

From 21e9a91a5746e1146201c8493ae3245fef6c762b Mon Sep 17 00:00:00 2001
From: Roland Tannous <115670425+rolandtannous@users.noreply.github.com>
Date: Tue, 21 Apr 2026 13:17:20 +0400
Subject: [PATCH 143/216] Studio: forward standard OpenAI tools / tool_choice
 on /v1/responses (Codex compat) (#5122)
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

* Studio: forward standard OpenAI tools / tool_choice on /v1/responses

Mirrors the /v1/chat/completions client-side tool pass-through from #5099
so clients (OpenAI Codex CLI, OpenAI Python SDK, ...) that target the
Responses API receive structured function_call output items instead of
plain text with tool-call tokens leaking into content.

- ResponsesRequest: type tools/tool_choice properly, add parallel_tool_calls;
  accept function_call and function_call_output input items for multi-turn
- Translate flat Responses tool / tool_choice shape to the nested Chat
  Completions shape before forwarding to llama-server
- _normalise_responses_input: map function_call_output -> role="tool",
  function_call -> assistant tool_calls (preserving call_id)
- Non-streaming: map returned tool_calls -> top-level function_call
  output items keyed by call_id
- Streaming: emit response.output_item.added (function_call),
  response.function_call_arguments.delta/.done, and response.output_item.done
  per tool call while keeping the text message at output_index 0
- Pytest coverage: tools/tool_choice translation, multi-turn input mapping,
  non-streaming tool_calls mapping, response round-trip

* Studio: merge system messages and close inner stream on /v1/responses

Fixes two issues surfacing when OpenAI Codex CLI drives /v1/responses
against a GGUF with a strict chat template (gpt-oss harmony, Qwen3, ...).

1. "System message must be at the beginning" upstream errors
   Codex sends `instructions` AND a `role:"developer"` message in `input`,
   producing two separate system-role messages. Strict templates raise
   when a second system message exists or when one appears after a user
   turn. _normalise_responses_input now hoists all instructions / system /
   developer content into a single merged system message at the top of
   the Chat Completions message list.

2. "async generator ignored GeneratorExit" / "Attempted to exit cancel
   scope in a different task"
   _responses_stream consumed the inner chat-completions body_iterator
   without an explicit aclose() in a finally block. On client disconnect
   (Codex frequently cancels mid-stream), Python 3.13 finalized the inner
   async generator on a different task, tripping anyio's cancel-scope
   check. Mirrored the same try/finally + aclose pattern used by the
   /v1/messages, /v1/chat/completions, and /v1/completions passthroughs.

Tests: hoisting of instructions + developer, developer mid-conversation,
multiple system messages in input, no-system passthrough.

* Studio: accept Codex multi-turn shapes and fix cross-task stream close on /v1/responses

Two issues observed driving /v1/responses from OpenAI Codex CLI against a
GGUF backend.

1. 422 on every turn after the first
   Codex replays prior assistant turns with
   `content:[{"type":"output_text","text":...,"annotations":[],"logprobs":[]}]`
   and carries forward `reasoning` items (o-series / gpt-5) between turns.
   Our `ResponsesContentPart` union only accepted input_text / input_image,
   and `ResponsesInputItem` only message / function_call / function_call_output,
   so Pydantic failed the whole list and FastAPI returned
   `"Input should be a valid string"` against the `str` branch of the
   outer union.

   - Add `ResponsesOutputTextPart` for assistant-replay content.
   - Add `ResponsesUnknownContentPart` and `ResponsesUnknownInputItem`
     as permissive catch-alls (drop during normalisation).
   - Wire an explicit `Discriminator` so dispatch is deterministic and
     the fallthrough reaches the catch-all instead of misreporting via
     the outer `Union[str, list[...]]`.
   - `_normalise_responses_input` now accepts output_text parts, flattens
     single-part assistant text to a plain string (keeps legacy chat
     templates happy), and silently drops reasoning / unknown items.

2. "async generator ignored GeneratorExit" / cross-task cancel scope
   `_responses_stream` awaited `openai_chat_completions` in the parent
   route-handler task, which opens the httpx client for the inner
   passthrough on *that* task. The outer `StreamingResponse` then iterates
   in a child task, so the asyncgen GC finalises the inner httpcore byte
   stream on the child task, tripping anyio's "Attempted to exit cancel
   scope in a different task". Move the `await` inside `event_generator`
   so the httpx lifecycle stays within the single streaming child task,
   and surface any HTTPException as a `response.failed` SSE frame.

Tests: assistant output_text replay, reasoning-item tolerance, unknown
content-part tolerance, end-to-end Codex-shape payload (developer + user +
reasoning + function_call + function_call_output + assistant output_text +
user), and single-part assistant flattening to plain string.

* Studio: call llama-server directly from streaming /v1/responses

The previous fix (running the inner await inside event_generator) was not
enough. Wrapping the existing `openai_chat_completions` pass-through still
stacks two async generators: when the outer generator is closed, the
innermost `HTTP11ConnectionByteStream.__aiter__` in httpcore doesn't
receive GeneratorExit before Python's asyncgen GC finalises it in a
sibling task, tripping "Attempted to exit cancel scope in a different
task" and "async generator ignored GeneratorExit" — the same Python 3.13
+ httpcore 1.0.x interaction already seen in PRs #4956, #4981, #5099.

Cure both pass-throughs had: a single same-task httpx lifecycle with
explicit `aiter_lines().aclose()` BEFORE `resp.aclose()` / `client.aclose()`
in the generator's finally block.

Apply it at the Responses layer by dropping the wrapper entirely for GGUF:
open httpx, consume `resp.aiter_lines()`, parse `chat.completion.chunk`,
emit Responses SSE events, close everything in finally — all in the
single StreamingResponse child task. Non-GGUF streaming is rejected with
a 400 (wrapping the transformers backend would re-introduce the
double-layer pattern and isn't a Codex-compatible path today anyway).

Also surfaces upstream httpx.RequestError / non-200 as a
`response.failed` SSE frame rather than a dropped stream now that the
request is dispatched after SSE headers have gone out.

* Studio: silence benign httpcore asyncgen GC warnings on Python 3.13

The streaming pass-throughs (/v1/chat/completions, /v1/messages,
/v1/responses, /v1/completions) all use the proven #4981 / #5099 pattern
— single-task httpx lifecycle with explicit aiter_lines().aclose() ahead
of resp.aclose() / client.aclose() in the generator's finally block.
That handles our own iterators correctly.

The residual noise ("async generator ignored GeneratorExit" /
"Attempted to exit cancel scope in a different task") comes from an
innermost HTTP11ConnectionByteStream.__aiter__ that httpcore creates
internally inside its pool. We hold no reference to it, so we cannot
aclose it ourselves. Python 3.13's asyncgen GC hook finalises it on the
finaliser task, its aclose path enters an anyio CancelScope shield, and
Python flags the cross-task exit. The response has already been
delivered with a 200 by then — it is purely log noise, not a functional
failure. Same interaction seen in modelcontextprotocol/python-sdk #831,
agno #3556, chainlit #2361, langchain-mcp-adapters #254.

Install a targeted sys.unraisablehook that swallows this specific tuple
— RuntimeError mentioning "cancel scope" or "GeneratorExit" plus an
object repr referencing HTTP11ConnectionByteStream — and defers to the
default hook for every other unraisable. Idempotent; guarded by a
sentinel attribute so repeated imports don't stack filters.
---
 studio/backend/models/inference.py            | 207 +++++-
 studio/backend/routes/inference.py            | 703 +++++++++++++++---
 .../tests/test_responses_tool_passthrough.py  | 667 +++++++++++++++++
 3 files changed, 1469 insertions(+), 108 deletions(-)
 create mode 100644 studio/backend/tests/test_responses_tool_passthrough.py

diff --git a/studio/backend/models/inference.py b/studio/backend/models/inference.py
index 956b99f75c..324002ddbf 100644
--- a/studio/backend/models/inference.py
+++ b/studio/backend/models/inference.py
@@ -599,23 +599,172 @@ class ResponsesInputImagePart(BaseModel):
     detail: Optional[Literal["auto", "low", "high"]] = "auto"
 
 
-ResponsesContentPart = Union[ResponsesInputTextPart, ResponsesInputImagePart]
+class ResponsesOutputTextPart(BaseModel):
+    """Assistant ``output_text`` content part replayed on subsequent turns.
+
+    When a client (OpenAI Codex CLI, OpenAI Python SDK agents) loops on a
+    stateless Responses endpoint, prior assistant messages are round-tripped
+    as ``{"role":"assistant","content":[{"type":"output_text","text":...,
+    "annotations":[],"logprobs":[]}]}``. We preserve the text and ignore
+    the annotations/logprobs metadata when flattening into Chat Completions.
+    """
+
+    type: Literal["output_text"]
+    text: str
+    annotations: Optional[list] = None
+    logprobs: Optional[list] = None
+
+    model_config = {"extra": "allow"}
+
+
+class ResponsesUnknownContentPart(BaseModel):
+    """Catch-all for content-part types we don't model explicitly.
+
+    Keeps validation green when a client sends newer part types (e.g.
+    ``input_audio``, ``input_file``) we haven't mapped; these are silently
+    skipped during normalisation rather than rejected with a 422.
+    """
+
+    type: str
+
+    model_config = {"extra": "allow"}
+
+
+ResponsesContentPart = Union[
+    ResponsesInputTextPart,
+    ResponsesInputImagePart,
+    ResponsesOutputTextPart,
+    ResponsesUnknownContentPart,
+]
 
 
 class ResponsesInputMessage(BaseModel):
     """A single message in the Responses API input array."""
 
+    type: Optional[Literal["message"]] = None
     role: Literal["system", "user", "assistant", "developer"]
     content: Union[str, list[ResponsesContentPart]]
 
+    # Codex (gpt-5.3-codex+) attaches a `phase` field ("commentary" |
+    # "final_answer") to assistant messages and requires clients to preserve
+    # it on subsequent turns. We accept and round-trip it; llama-server does
+    # not care about it.
+    model_config = {"extra": "allow"}
+
+
+class ResponsesFunctionCallInputItem(BaseModel):
+    """A prior assistant function_call being replayed in a multi-turn Responses input.
+
+    The Responses API represents tool calls as top-level input items (not
+    nested inside assistant messages), correlated across turns by ``call_id``.
+    """
+
+    type: Literal["function_call"]
+    id: Optional[str] = Field(
+        None, description = "Item id assigned by the server (e.g. fc_...)"
+    )
+    call_id: str = Field(
+        ...,
+        description = "Correlation id matching a function_call_output on the next turn.",
+    )
+    name: str
+    arguments: str = Field(
+        ..., description = "JSON string of the arguments the model produced."
+    )
+    status: Optional[Literal["in_progress", "completed", "incomplete"]] = None
+
+
+class ResponsesFunctionCallOutputInputItem(BaseModel):
+    """A tool result supplied by the client for a prior function_call.
+
+    Replaces Chat Completions' ``role="tool"`` message. Correlated to the
+    originating call by ``call_id``.
+    """
+
+    type: Literal["function_call_output"]
+    id: Optional[str] = None
+    call_id: str
+    output: Union[str, list] = Field(
+        ..., description = "String or content-array result of the tool call."
+    )
+    status: Optional[Literal["in_progress", "completed", "incomplete"]] = None
+
+
+class ResponsesUnknownInputItem(BaseModel):
+    """Catch-all for Responses input item types we don't model explicitly.
+
+    Covers ``reasoning`` items (replayed from prior o-series / gpt-5 turns)
+    and any future item types the client may send. These items are dropped
+    during normalisation — llama-server-backed GGUFs cannot consume them —
+    but keeping them in the request-model union stops unrelated turns from
+    failing validation with a 422.
+    """
+
+    type: str
+
+    model_config = {"extra": "allow"}
+
+
+def _responses_input_item_discriminator(v: Any) -> str:
+    """Route a Responses input item to the correct tagged variant.
+
+    Pydantic's default smart-union matching fails when one variant in the
+    union is tagged with a strict ``Literal`` (``function_call`` /
+    ``function_call_output``) and the incoming dict uses a different
+    ``type`` — the other variants' validation errors are hidden and the
+    outer ``Union[str, list[...]]`` reports a misleading "Input should be a
+    valid string" error. An explicit discriminator makes the routing
+    deterministic and lets us fall through to the catch-all.
+    """
+    if isinstance(v, dict):
+        t = v.get("type")
+        r = v.get("role")
+    else:
+        t = getattr(v, "type", None)
+        r = getattr(v, "role", None)
+    if t == "function_call":
+        return "function_call"
+    if t == "function_call_output":
+        return "function_call_output"
+    if r is not None or t == "message":
+        return "message"
+    return "unknown"
+
+
+ResponsesInputItem = Annotated[
+    Union[
+        Annotated[ResponsesInputMessage, Tag("message")],
+        Annotated[ResponsesFunctionCallInputItem, Tag("function_call")],
+        Annotated[ResponsesFunctionCallOutputInputItem, Tag("function_call_output")],
+        Annotated[ResponsesUnknownInputItem, Tag("unknown")],
+    ],
+    Discriminator(_responses_input_item_discriminator),
+]
+
+
+class ResponsesFunctionTool(BaseModel):
+    """Flat function-tool definition used by the Responses API request.
+
+    Unlike Chat Completions (which nests ``{"name": ..., "parameters": ...}``
+    inside a ``"function"`` key), the Responses API uses a flat shape with
+    ``type``, ``name``, ``description``, ``parameters``, and ``strict`` at the
+    top level of each tool entry.
+    """
+
+    type: Literal["function"]
+    name: str
+    description: Optional[str] = None
+    parameters: Optional[dict] = None
+    strict: Optional[bool] = None
+
 
 class ResponsesRequest(BaseModel):
     """OpenAI Responses API request."""
 
     model: str = Field("default", description = "Model identifier")
-    input: Union[str, list[ResponsesInputMessage]] = Field(
+    input: Union[str, list[ResponsesInputItem]] = Field(
         default = [],
-        description = "Input text or message list",
+        description = "Input text or list of messages / function_call / function_call_output items",
     )
     instructions: Optional[str] = Field(
         None, description = "System / developer instructions"
@@ -625,9 +774,31 @@ class ResponsesRequest(BaseModel):
     max_output_tokens: Optional[int] = Field(None, ge = 1)
     stream: bool = Field(False, description = "Whether to stream the response via SSE")
 
-    # Accepted but ignored -- keeps SDK clients from failing on unsupported fields
-    tools: Optional[list] = None
-    tool_choice: Optional[Any] = None
+    # OpenAI function-calling fields — forwarded to llama-server via the
+    # Chat Completions pass-through (see routes/inference.py). Typed as a
+    # plain list so built-in tool shapes (``web_search``, ``file_search``,
+    # ``mcp``, ...) round-trip without validation errors — the translator
+    # picks out only ``type=="function"`` entries for forwarding.
+    tools: Optional[list[dict]] = Field(
+        None,
+        description = (
+            "Responses-shape function tool definitions. Entries with "
+            '`type="function"` are translated to the Chat Completions nested '
+            "shape before being forwarded to llama-server; other tool types "
+            "(built-in web_search, file_search, mcp, ...) are accepted for SDK "
+            "compatibility but ignored on the llama-server passthrough."
+        ),
+    )
+    tool_choice: Optional[Any] = Field(
+        None,
+        description = (
+            "'auto' | 'required' | 'none' | {'type': 'function', 'name': ...} — "
+            "the Responses-shape forcing object is translated to the Chat "
+            "Completions nested shape internally."
+        ),
+    )
+    parallel_tool_calls: Optional[bool] = None
+
     previous_response_id: Optional[str] = None
     store: Optional[bool] = None
     metadata: Optional[dict] = None
@@ -660,6 +831,28 @@ class ResponsesOutputMessage(BaseModel):
     content: list[ResponsesOutputTextContent] = Field(default_factory = list)
 
 
+class ResponsesOutputFunctionCall(BaseModel):
+    """A function-call output item in the Responses API response.
+
+    Unlike Chat Completions (which nests tool calls inside the assistant
+    message), the Responses API emits each tool call as its own top-level
+    ``output`` item so clients can correlate results via ``call_id`` on the
+    next turn.
+    """
+
+    type: Literal["function_call"] = "function_call"
+    id: str = Field(default_factory = lambda: f"fc_{uuid.uuid4().hex[:12]}")
+    call_id: str
+    name: str
+    arguments: str = Field(
+        ..., description = "JSON string of the arguments the model produced."
+    )
+    status: Literal["completed", "in_progress", "incomplete"] = "completed"
+
+
+ResponsesOutputItem = Union[ResponsesOutputMessage, ResponsesOutputFunctionCall]
+
+
 class ResponsesUsage(BaseModel):
     """Token usage for a Responses API response (input_tokens, not prompt_tokens)."""
 
@@ -676,7 +869,7 @@ class ResponsesResponse(BaseModel):
     created_at: int = Field(default_factory = lambda: int(time.time()))
     status: Literal["completed", "in_progress", "failed"] = "completed"
     model: str = "default"
-    output: list[ResponsesOutputMessage] = Field(default_factory = list)
+    output: list[ResponsesOutputItem] = Field(default_factory = list)
     usage: ResponsesUsage = Field(default_factory = ResponsesUsage)
     error: Optional[Any] = None
     incomplete_details: Optional[Any] = None
diff --git a/studio/backend/routes/inference.py b/studio/backend/routes/inference.py
index 6103c90915..9c84eabd8b 100644
--- a/studio/backend/routes/inference.py
+++ b/studio/backend/routes/inference.py
@@ -12,7 +12,7 @@ import uuid
 from pathlib import Path
 from fastapi import APIRouter, Depends, HTTPException, Request, status
 from fastapi.responses import StreamingResponse, JSONResponse, Response
-from typing import Optional
+from typing import Any, Optional, Union
 import json
 import httpx
 import structlog
@@ -27,6 +27,58 @@ import re as _re
 from utils.models import extract_model_size_b as _extract_model_size_b
 
 
+def _install_httpcore_asyncgen_silencer() -> None:
+    """Silence benign httpx/httpcore asyncgen GC noise on Python 3.13.
+
+    When Studio proxies a streaming response from llama-server via httpx,
+    the innermost ``HTTP11ConnectionByteStream.__aiter__`` async generator
+    is finalised by Python's asyncgen GC hook on a task different from the
+    one that opened it. Its ``aclose`` path then calls
+    ``anyio.Lock.acquire`` → ``cancel_shielded_checkpoint`` which enters a
+    ``CancelScope`` on the finaliser task — Python 3.13 flags the
+    cross-task exit as ``"Attempted to exit cancel scope in a different
+    task"`` and prints ``"async generator ignored GeneratorExit"`` as an
+    unraisable warning.
+
+    This is a known httpx + httpcore + anyio interaction (see MCP SDK
+    python-sdk#831, agno #3556, chainlit #2361, langchain-mcp-adapters
+    #254). It is benign: the response has already been delivered with a
+    200. The streaming pass-throughs (``/v1/chat/completions``,
+    ``/v1/messages``, ``/v1/responses``, ``/v1/completions``) already
+    manage their httpx lifecycle inside a single task with explicit
+    ``aclose()`` of the lines iterator, response, and client; the errant
+    generator is not one we hold a reference to and therefore cannot
+    close ourselves.
+
+    We install a single process-wide unraisable hook that swallows just
+    this specific interaction — identified by the tuple of (RuntimeError
+    mentioning cancel scope / GeneratorExit) + (object repr referencing
+    HTTP11ConnectionByteStream) — and defers to the default hook for
+    everything else. The filter is idempotent.
+    """
+    prior_hook = sys.unraisablehook
+    if getattr(prior_hook, "_unsloth_httpcore_silencer", False):
+        return
+
+    def _hook(unraisable):
+        exc_value = getattr(unraisable, "exc_value", None)
+        obj = getattr(unraisable, "object", None)
+        obj_repr = repr(obj) if obj is not None else ""
+        if (
+            isinstance(exc_value, RuntimeError)
+            and "HTTP11ConnectionByteStream" in obj_repr
+            and ("cancel scope" in str(exc_value) or "GeneratorExit" in str(exc_value))
+        ):
+            return
+        prior_hook(unraisable)
+
+    _hook._unsloth_httpcore_silencer = True  # type: ignore[attr-defined]
+    sys.unraisablehook = _hook
+
+
+_install_httpcore_asyncgen_silencer()
+
+
 def _friendly_error(exc: Exception) -> str:
     """Extract a user-friendly message from known llama-server errors."""
     # httpx transport-layer failures reaching the managed llama-server —
@@ -101,8 +153,14 @@ from models.inference import (
     ResponsesInputMessage,
     ResponsesInputTextPart,
     ResponsesInputImagePart,
+    ResponsesOutputTextPart,
+    ResponsesUnknownContentPart,
+    ResponsesUnknownInputItem,
+    ResponsesFunctionCallInputItem,
+    ResponsesFunctionCallOutputInputItem,
     ResponsesOutputTextContent,
     ResponsesOutputMessage,
+    ResponsesOutputFunctionCall,
     ResponsesUsage,
     ResponsesResponse,
     AnthropicMessagesRequest,
@@ -2083,49 +2141,229 @@ async def openai_embeddings(
 # =====================================================================
 
 
+def _translate_responses_tools_to_chat(
+    tools: Optional[list[dict]],
+) -> Optional[list[dict]]:
+    """Translate Responses-shape function tools to the Chat Completions nested shape.
+
+    Responses uses a flat shape per tool entry::
+
+        {"type": "function", "name": "...", "description": "...",
+         "parameters": {...}, "strict": true}
+
+    The Chat Completions / llama-server passthrough expects the nested shape::
+
+        {"type": "function",
+         "function": {"name": "...", "description": "...",
+                      "parameters": {...}, "strict": true}}
+
+    Only ``type=="function"`` entries are forwarded. Built-in Responses tools
+    (``web_search``, ``file_search``, ``mcp``, ...) are dropped because
+    llama-server does not implement them server-side; keeping them in the
+    request would produce an opaque upstream 400.
+    """
+    if not tools:
+        return None
+    out: list[dict] = []
+    for tool in tools:
+        if not isinstance(tool, dict):
+            continue
+        if tool.get("type") != "function":
+            continue
+        fn: dict = {}
+        if "name" in tool:
+            fn["name"] = tool["name"]
+        if tool.get("description") is not None:
+            fn["description"] = tool["description"]
+        if tool.get("parameters") is not None:
+            fn["parameters"] = tool["parameters"]
+        if tool.get("strict") is not None:
+            fn["strict"] = tool["strict"]
+        out.append({"type": "function", "function": fn})
+    return out or None
+
+
+def _translate_responses_tool_choice_to_chat(tool_choice: Any) -> Any:
+    """Translate a Responses-shape ``tool_choice`` to the Chat Completions shape.
+
+    String values (``"auto"``/``"none"``/``"required"``) pass through unchanged.
+    The Responses forcing object ``{"type": "function", "name": "X"}`` is
+    converted to Chat Completions' ``{"type": "function", "function": {"name": "X"}}``.
+    Unknown / built-in tool choices are forwarded as-is; llama-server ignores
+    what it doesn't recognise.
+    """
+    if tool_choice is None:
+        return None
+    if isinstance(tool_choice, str):
+        return tool_choice
+    if (
+        isinstance(tool_choice, dict)
+        and tool_choice.get("type") == "function"
+        and "name" in tool_choice
+        and "function" not in tool_choice
+    ):
+        return {"type": "function", "function": {"name": tool_choice["name"]}}
+    return tool_choice
+
+
+def _responses_message_text(content: Union[str, list]) -> str:
+    """Flatten a ResponsesInputMessage ``content`` into a plain text string.
+
+    Used for system/developer message hoisting and for assistant-replay
+    (``output_text``) messages when images/unknown parts are irrelevant.
+    Returns an empty string for empty input.
+    """
+    if isinstance(content, str):
+        return content
+    parts: list[str] = []
+    for part in content or []:
+        if isinstance(part, (ResponsesInputTextPart, ResponsesOutputTextPart)):
+            parts.append(part.text)
+    return "\n".join(parts)
+
+
 def _normalise_responses_input(payload: ResponsesRequest) -> list[ChatMessage]:
-    """Convert a ResponsesRequest into a list of ChatMessage for the completions backend."""
+    """Convert a ResponsesRequest's ``input`` into Chat-format ``ChatMessage`` list.
+
+    Handles the three input item shapes allowed by the Responses API:
+
+    - ``ResponsesInputMessage`` — regular chat messages (text or multimodal).
+    - ``ResponsesFunctionCallInputItem`` — a prior assistant tool call replayed
+      on a follow-up turn. Converted into an assistant message carrying a
+      Chat Completions ``tool_calls`` entry keyed by ``call_id``.
+    - ``ResponsesFunctionCallOutputInputItem`` — a tool result the client is
+      returning. Converted into a ``role="tool"`` message with ``tool_call_id``
+      set to the originating ``call_id`` so llama-server can reconcile the
+      call with its result.
+
+    System / developer content is collected from ``instructions`` *and* from
+    any ``role="system"`` / ``role="developer"`` entries in ``input``, then
+    merged into a single ``role="system"`` message placed at the top of the
+    returned list. This satisfies strict chat templates (harmony / gpt-oss,
+    Qwen3, ...) whose Jinja raises ``"System message must be at the
+    beginning."`` when more than one system message is present or when a
+    system message appears after a user turn — the exact pattern the OpenAI
+    Codex CLI hits, since Codex sets ``instructions`` *and* also sends a
+    developer message in ``input``.
+    """
+    system_parts: list[str] = []
     messages: list[ChatMessage] = []
 
-    # System / developer instructions
     if payload.instructions:
-        messages.append(ChatMessage(role = "system", content = payload.instructions))
+        system_parts.append(payload.instructions)
 
     # Simple string input
     if isinstance(payload.input, str):
         if payload.input:
             messages.append(ChatMessage(role = "user", content = payload.input))
+        if system_parts:
+            merged = "\n\n".join(p for p in system_parts if p)
+            return [ChatMessage(role = "system", content = merged), *messages]
         return messages
 
-    # List of ResponsesInputMessage
-    for msg in payload.input:
-        role = "system" if msg.role == "developer" else msg.role
+    for item in payload.input:
+        if isinstance(item, ResponsesFunctionCallInputItem):
+            messages.append(
+                ChatMessage(
+                    role = "assistant",
+                    content = None,
+                    tool_calls = [
+                        {
+                            "id": item.call_id,
+                            "type": "function",
+                            "function": {
+                                "name": item.name,
+                                "arguments": item.arguments,
+                            },
+                        }
+                    ],
+                )
+            )
+            continue
 
-        if isinstance(msg.content, str):
-            messages.append(ChatMessage(role = role, content = msg.content))
-        else:
-            # Convert Responses content parts -> Chat content parts
-            parts = []
-            for part in msg.content:
-                if isinstance(part, ResponsesInputTextPart):
-                    parts.append(TextContentPart(type = "text", text = part.text))
-                elif isinstance(part, ResponsesInputImagePart):
-                    parts.append(
-                        ImageContentPart(
-                            type = "image_url",
-                            image_url = ImageUrl(url = part.image_url, detail = part.detail),
-                        )
+        if isinstance(item, ResponsesFunctionCallOutputInputItem):
+            # Chat Completions `role="tool"` requires a string content; if a
+            # Responses client sends a content-array output, serialize it.
+            output = item.output
+            if not isinstance(output, str):
+                output = json.dumps(output)
+            messages.append(
+                ChatMessage(
+                    role = "tool",
+                    tool_call_id = item.call_id,
+                    content = output,
+                )
+            )
+            continue
+
+        if isinstance(item, ResponsesUnknownInputItem):
+            # Reasoning items and any other unmodelled top-level Responses
+            # item types are silently dropped — llama-server-backed GGUFs
+            # cannot consume them and our lenient validation let them in so
+            # unrelated turns don't 422.
+            continue
+
+        # ResponsesInputMessage — hoist system/developer to the top, merge.
+        if item.role in ("system", "developer"):
+            hoisted = _responses_message_text(item.content)
+            if hoisted:
+                system_parts.append(hoisted)
+            continue
+
+        if isinstance(item.content, str):
+            messages.append(ChatMessage(role = item.role, content = item.content))
+            continue
+
+        # Assistant-replay turns come back as content = [output_text, ...].
+        # Chat Completions' assistant role expects a plain string, not a
+        # multimodal content array, so flatten output_text (and any stray
+        # input_text / unknown text) to a single string.
+        if item.role == "assistant":
+            text = _responses_message_text(item.content)
+            if text:
+                messages.append(ChatMessage(role = "assistant", content = text))
+            continue
+
+        # User (and any other remaining roles) — keep multimodal when
+        # present, drop unknown content parts silently.
+        parts: list = []
+        for part in item.content:
+            if isinstance(part, (ResponsesInputTextPart, ResponsesOutputTextPart)):
+                parts.append(TextContentPart(type = "text", text = part.text))
+            elif isinstance(part, ResponsesInputImagePart):
+                parts.append(
+                    ImageContentPart(
+                        type = "image_url",
+                        image_url = ImageUrl(url = part.image_url, detail = part.detail),
                     )
-            messages.append(ChatMessage(role = role, content = parts if parts else ""))
+                )
+            # ResponsesUnknownContentPart and anything else: drop.
+        if parts:
+            # Collapse single-text-part content to a plain string so roles
+            # that reject multimodal arrays (e.g. legacy templates) still
+            # accept the message.
+            if len(parts) == 1 and isinstance(parts[0], TextContentPart):
+                messages.append(ChatMessage(role = item.role, content = parts[0].text))
+            else:
+                messages.append(ChatMessage(role = item.role, content = parts))
 
+    if system_parts:
+        merged = "\n\n".join(p for p in system_parts if p)
+        return [ChatMessage(role = "system", content = merged), *messages]
     return messages
 
 
 def _build_chat_request(
     payload: ResponsesRequest, messages: list[ChatMessage], stream: bool
 ) -> ChatCompletionRequest:
-    """Build a ChatCompletionRequest from a ResponsesRequest."""
-    chat_kwargs = dict(
+    """Build a ChatCompletionRequest from a ResponsesRequest.
+
+    Tools and ``tool_choice`` are translated from the flat Responses shape to
+    the nested Chat Completions shape here so the existing #5099
+    ``/v1/chat/completions`` client-side pass-through picks them up without
+    further modification.
+    """
+    chat_kwargs: dict = dict(
         model = payload.model,
         messages = messages,
         stream = stream,
@@ -2136,7 +2374,46 @@ def _build_chat_request(
         chat_kwargs["top_p"] = payload.top_p
     if payload.max_output_tokens is not None:
         chat_kwargs["max_tokens"] = payload.max_output_tokens
-    return ChatCompletionRequest(**chat_kwargs)
+
+    chat_tools = _translate_responses_tools_to_chat(payload.tools)
+    if chat_tools is not None:
+        chat_kwargs["tools"] = chat_tools
+
+    chat_tool_choice = _translate_responses_tool_choice_to_chat(payload.tool_choice)
+    if chat_tool_choice is not None:
+        chat_kwargs["tool_choice"] = chat_tool_choice
+
+    req = ChatCompletionRequest(**chat_kwargs)
+    # `parallel_tool_calls` is not a first-class field on ChatCompletionRequest,
+    # but the model allows extras and _build_openai_passthrough_body forwards
+    # only explicitly-known fields. Llama-server does not currently implement
+    # parallel_tool_calls semantics, so we accept-and-ignore it on the
+    # Responses side to avoid breaking SDK clients that always send it.
+    return req
+
+
+def _chat_tool_calls_to_responses_output(tool_calls: list[dict]) -> list[dict]:
+    """Map Chat Completions ``tool_calls`` into Responses ``function_call`` output items.
+
+    The Chat Completions id (``call_xxx``) is the shared correlation key across
+    turns in the OpenAI Responses API — it is stored as ``call_id`` on the
+    output item and must be echoed back by the client as
+    ``function_call_output.call_id`` on the next turn.
+    """
+    items: list[dict] = []
+    for tc in tool_calls:
+        if tc.get("type") != "function":
+            continue
+        fn = tc.get("function") or {}
+        items.append(
+            ResponsesOutputFunctionCall(
+                call_id = tc.get("id", ""),
+                name = fn.get("name", ""),
+                arguments = fn.get("arguments", "") or "",
+                status = "completed",
+            ).model_dump()
+        )
+    return items
 
 
 async def _responses_non_streaming(
@@ -2156,35 +2433,44 @@ async def _responses_non_streaming(
     else:
         body = result
 
-    # Extract content and usage from the Chat Completions response
     choices = body.get("choices", [])
     text = ""
+    tool_calls: list[dict] = []
     if choices:
-        msg = choices[0].get("message", {})
+        msg = choices[0].get("message", {}) or {}
         text = msg.get("content", "") or ""
+        tool_calls = msg.get("tool_calls") or []
 
     usage_data = body.get("usage", {})
     input_tokens = usage_data.get("prompt_tokens", 0)
     output_tokens = usage_data.get("completion_tokens", 0)
 
     resp_id = f"resp_{uuid.uuid4().hex[:12]}"
-    msg_id = f"msg_{uuid.uuid4().hex[:12]}"
+
+    # Responses API emits each tool call as its own top-level output item,
+    # alongside an optional assistant text message. Emit the text message
+    # only when the model actually produced content, so clients that expect
+    # a pure tool-call turn (finish_reason="tool_calls") don't see a spurious
+    # empty message item.
+    output_items: list[dict] = []
+    if text:
+        msg_id = f"msg_{uuid.uuid4().hex[:12]}"
+        output_items.append(
+            ResponsesOutputMessage(
+                id = msg_id,
+                status = "completed",
+                role = "assistant",
+                content = [ResponsesOutputTextContent(text = text)],
+            ).model_dump()
+        )
+    output_items.extend(_chat_tool_calls_to_responses_output(tool_calls))
 
     response = ResponsesResponse(
         id = resp_id,
         created_at = int(time.time()),
         status = "completed",
         model = body.get("model", payload.model),
-        output = [
-            ResponsesOutputMessage(
-                id = msg_id,
-                status = "completed",
-                role = "assistant",
-                content = [
-                    ResponsesOutputTextContent(text = text),
-                ],
-            ),
-        ],
+        output = output_items,
         usage = ResponsesUsage(
             input_tokens = input_tokens,
             output_tokens = output_tokens,
@@ -2203,24 +2489,100 @@ async def _responses_stream(
     messages: list[ChatMessage],
     request: Request,
 ):
-    """Handle a streaming Responses API call, emitting named SSE events."""
+    """Handle a streaming Responses API call, emitting named SSE events.
+
+    For GGUF models the request goes directly to llama-server's
+    ``/v1/chat/completions`` endpoint from inside the StreamingResponse
+    child task — a single httpx lifecycle, a single async generator.
+    Wrapping the existing ``openai_chat_completions`` pass-through (which
+    already does its own httpx lifecycle) stacks two generators: Python
+    3.13 + httpcore 1.0.x then loses the close-propagation chain on the
+    innermost ``HTTP11ConnectionByteStream`` at asyncgen finalisation,
+    tripping "Attempted to exit cancel scope in a different task" /
+    "async generator ignored GeneratorExit". The direct path avoids that
+    altogether. Non-GGUF falls back to the wrapper (which doesn't use
+    httpx, so the issue doesn't apply).
+
+    Text deltas arrive as ``response.output_text.delta`` on a single
+    ``message`` output item at ``output_index=0``. Each tool call from
+    ``delta.tool_calls[]`` is promoted to its own top-level ``function_call``
+    output item (one per distinct ``tool_calls[].index``), and relayed as
+    ``response.function_call_arguments.delta`` / ``.done`` events so clients
+    (Codex, OpenAI Python SDK) can reconstruct the call incrementally and
+    reply with a ``function_call_output`` item on the next turn.
+    """
     resp_id = f"resp_{uuid.uuid4().hex[:12]}"
     msg_id = f"msg_{uuid.uuid4().hex[:12]}"
-    item_id = f"item_{uuid.uuid4().hex[:12]}"
     created_at = int(time.time())
 
     chat_req = _build_chat_request(payload, messages, stream = True)
-    result = await openai_chat_completions(chat_req, request)
+
+    llama_backend = get_llama_cpp_backend()
+    if not llama_backend.is_loaded:
+        # The direct pass-through is GGUF-only. Non-GGUF /v1/responses
+        # streaming isn't a Codex-compatible path today and wrapping the
+        # transformers backend's streaming generator here would re-
+        # introduce the double-layer asyncgen close pattern that produces
+        # "Attempted to exit cancel scope in a different task" on Python
+        # 3.13. Surface a typed 400 so the client sees a useful error
+        # instead of a dangling stream.
+        raise HTTPException(
+            status_code = 400,
+            detail = (
+                "Streaming /v1/responses requires a GGUF model loaded via "
+                "llama-server. Use non-streaming /v1/responses, "
+                "/v1/chat/completions, or load a GGUF model."
+            ),
+        )
+
+    body = _build_openai_passthrough_body(chat_req)
+    target_url = f"{llama_backend.base_url}/v1/chat/completions"
 
     async def event_generator():
         full_text = ""
         input_tokens = 0
         output_tokens = 0
+        # Per-tool-call state keyed by the Chat Completions `tool_calls[].index`
+        # which stays stable across chunks for the same call. Values are:
+        #   {output_index, item_id, call_id, name, arguments, opened}
+        tool_call_state: dict[int, dict] = {}
+        # Text message lives at output_index 0; tool calls claim 1, 2, ...
+        next_output_index = 1
+
+        def _snapshot_output() -> list[dict]:
+            """Snapshot of all completed output items for response.completed."""
+            items: list[dict] = [
+                {
+                    "type": "message",
+                    "id": msg_id,
+                    "status": "completed",
+                    "role": "assistant",
+                    "content": [
+                        {
+                            "type": "output_text",
+                            "text": full_text,
+                            "annotations": [],
+                        }
+                    ],
+                }
+            ]
+            for st in sorted(tool_call_state.values(), key = lambda s: s["output_index"]):
+                items.append(
+                    {
+                        "type": "function_call",
+                        "id": st["item_id"],
+                        "status": "completed",
+                        "call_id": st["call_id"],
+                        "name": st["name"],
+                        "arguments": st["arguments"],
+                    }
+                )
+            return items
 
         # ── Preamble events ──
         yield f"event: response.created\ndata: {json.dumps({'type': 'response.created', 'response': {'id': resp_id, 'object': 'response', 'created_at': created_at, 'status': 'in_progress', 'model': payload.model, 'output': [], 'usage': {'input_tokens': 0, 'output_tokens': 0, 'total_tokens': 0}}})}\n\n"
 
-        # output_item.added
+        # output_item.added (text message at output_index 0)
         output_item = {
             "type": "message",
             "id": msg_id,
@@ -2234,62 +2596,215 @@ async def _responses_stream(
         content_part = {"type": "output_text", "text": "", "annotations": []}
         yield f"event: response.content_part.added\ndata: {json.dumps({'type': 'response.content_part.added', 'item_id': msg_id, 'output_index': 0, 'content_index': 0, 'part': content_part})}\n\n"
 
-        # ── Stream delta events from the inner chat completions stream ──
-        if isinstance(result, StreamingResponse):
-            async for raw_chunk in result.body_iterator:
-                if isinstance(raw_chunk, bytes):
-                    raw_chunk = raw_chunk.decode("utf-8", errors = "replace")
+        # ── Direct httpx lifecycle to llama-server ──
+        # Full same-task open + close, identical pattern to
+        # _openai_passthrough_stream and _anthropic_passthrough_stream:
+        # no `async with`, explicit aclose of lines_iter BEFORE resp /
+        # client so the innermost httpcore byte stream is finalised in
+        # this task (not via Python's asyncgen GC in a sibling task).
+        client = httpx.AsyncClient(timeout = 600)
+        resp = None
+        lines_iter = None
+        try:
+            req = client.build_request("POST", target_url, json = body)
+            try:
+                resp = await client.send(req, stream = True)
+            except httpx.RequestError as e:
+                logger.error("responses stream: upstream unreachable: %s", e)
+                yield f"event: response.failed\ndata: {json.dumps({'type': 'response.failed', 'response': {'id': resp_id, 'object': 'response', 'created_at': created_at, 'status': 'failed', 'model': payload.model, 'output': [], 'error': {'code': 502, 'message': _friendly_error(e)}}})}\n\n"
+                return
 
-                for line in raw_chunk.split("\n"):
-                    line = line.strip()
-                    if not line.startswith("data: "):
-                        continue
-                    data_str = line[6:]
-                    if data_str == "[DONE]":
-                        continue
-                    try:
-                        chunk_data = json.loads(data_str)
-                    except json.JSONDecodeError:
-                        continue
+            if resp.status_code != 200:
+                err_bytes = await resp.aread()
+                err_text = err_bytes.decode("utf-8", errors = "replace")
+                logger.error(
+                    "responses stream upstream error: status=%s body=%s",
+                    resp.status_code,
+                    err_text[:500],
+                )
+                yield f"event: response.failed\ndata: {json.dumps({'type': 'response.failed', 'response': {'id': resp_id, 'object': 'response', 'created_at': created_at, 'status': 'failed', 'model': payload.model, 'output': [], 'error': {'code': resp.status_code, 'message': f'llama-server error: {err_text[:500]}'}}})}\n\n"
+                return
 
-                    choices = chunk_data.get("choices", [])
-                    if not choices:
-                        # Check for usage in final chunk
-                        usage = chunk_data.get("usage")
-                        if usage:
-                            input_tokens = usage.get("prompt_tokens", input_tokens)
-                            output_tokens = usage.get(
-                                "completion_tokens", output_tokens
-                            )
-                        continue
+            lines_iter = resp.aiter_lines()
+            async for raw_line in lines_iter:
+                if await request.is_disconnected():
+                    break
+                if not raw_line:
+                    continue
+                if not raw_line.startswith("data: "):
+                    continue
+                data_str = raw_line[6:]
+                if data_str.strip() == "[DONE]":
+                    break
+                try:
+                    chunk_data = json.loads(data_str)
+                except json.JSONDecodeError:
+                    continue
 
-                    delta = choices[0].get("delta", {})
-                    content = delta.get("content")
-                    if content:
-                        full_text += content
-                        delta_event = {
-                            "type": "response.output_text.delta",
-                            "item_id": msg_id,
-                            "output_index": 0,
-                            "content_index": 0,
-                            "delta": content,
-                        }
-                        yield f"event: response.output_text.delta\ndata: {json.dumps(delta_event)}\n\n"
-
-                    # Check for usage in chunk
+                choices = chunk_data.get("choices", [])
+                if not choices:
                     usage = chunk_data.get("usage")
                     if usage:
                         input_tokens = usage.get("prompt_tokens", input_tokens)
                         output_tokens = usage.get("completion_tokens", output_tokens)
+                    continue
 
-        # ── Closing events ──
-        # output_text.done
+                delta = choices[0].get("delta", {}) or {}
+                content = delta.get("content")
+                if content:
+                    full_text += content
+                    delta_event = {
+                        "type": "response.output_text.delta",
+                        "item_id": msg_id,
+                        "output_index": 0,
+                        "content_index": 0,
+                        "delta": content,
+                    }
+                    yield f"event: response.output_text.delta\ndata: {json.dumps(delta_event)}\n\n"
+
+                for tc in delta.get("tool_calls") or []:
+                    idx = tc.get("index", 0)
+                    st = tool_call_state.get(idx)
+                    fn = tc.get("function") or {}
+                    if st is None:
+                        # First chunk for this tool call — allocate an
+                        # output_index and emit output_item.added.
+                        st = {
+                            "output_index": next_output_index,
+                            "item_id": f"fc_{uuid.uuid4().hex[:12]}",
+                            "call_id": tc.get("id") or "",
+                            "name": fn.get("name") or "",
+                            "arguments": "",
+                            "opened": False,
+                        }
+                        next_output_index += 1
+                        tool_call_state[idx] = st
+                    else:
+                        # Later chunks sometimes carry the id/name only
+                        # once; merge when present.
+                        if tc.get("id") and not st["call_id"]:
+                            st["call_id"] = tc["id"]
+                        if fn.get("name") and not st["name"]:
+                            st["name"] = fn["name"]
+
+                    if not st["opened"] and st["call_id"] and st["name"]:
+                        item_added = {
+                            "type": "response.output_item.added",
+                            "output_index": st["output_index"],
+                            "item": {
+                                "type": "function_call",
+                                "id": st["item_id"],
+                                "status": "in_progress",
+                                "call_id": st["call_id"],
+                                "name": st["name"],
+                                "arguments": "",
+                            },
+                        }
+                        yield f"event: response.output_item.added\ndata: {json.dumps(item_added)}\n\n"
+                        st["opened"] = True
+
+                    arg_delta = fn.get("arguments") or ""
+                    if arg_delta and st["opened"]:
+                        st["arguments"] += arg_delta
+                        args_delta_event = {
+                            "type": "response.function_call_arguments.delta",
+                            "item_id": st["item_id"],
+                            "output_index": st["output_index"],
+                            "delta": arg_delta,
+                        }
+                        yield f"event: response.function_call_arguments.delta\ndata: {json.dumps(args_delta_event)}\n\n"
+                    elif arg_delta:
+                        # Buffer the args until we can open the item
+                        # (id/name arrive in the same chunk as the first
+                        # arg delta for some models — but if not, stash).
+                        st["arguments"] += arg_delta
+
+                usage = chunk_data.get("usage")
+                if usage:
+                    input_tokens = usage.get("prompt_tokens", input_tokens)
+                    output_tokens = usage.get("completion_tokens", output_tokens)
+        except Exception as e:
+            logger.error("responses stream error: %s", e)
+        finally:
+            if lines_iter is not None:
+                try:
+                    await lines_iter.aclose()
+                except Exception:
+                    pass
+            if resp is not None:
+                try:
+                    await resp.aclose()
+                except Exception:
+                    pass
+            try:
+                await client.aclose()
+            except Exception:
+                pass
+
+        # ── Closing events for tool calls ──
+        for st in sorted(tool_call_state.values(), key = lambda s: s["output_index"]):
+            # If id/name never arrived (malformed upstream), synthesise so
+            # the client still sees a coherent frame sequence.
+            if not st["opened"]:
+                if not st["call_id"]:
+                    st["call_id"] = f"call_{uuid.uuid4().hex[:12]}"
+                item_added = {
+                    "type": "response.output_item.added",
+                    "output_index": st["output_index"],
+                    "item": {
+                        "type": "function_call",
+                        "id": st["item_id"],
+                        "status": "in_progress",
+                        "call_id": st["call_id"],
+                        "name": st["name"],
+                        "arguments": "",
+                    },
+                }
+                yield f"event: response.output_item.added\ndata: {json.dumps(item_added)}\n\n"
+                if st["arguments"]:
+                    yield (
+                        "event: response.function_call_arguments.delta\n"
+                        "data: "
+                        + json.dumps(
+                            {
+                                "type": "response.function_call_arguments.delta",
+                                "item_id": st["item_id"],
+                                "output_index": st["output_index"],
+                                "delta": st["arguments"],
+                            }
+                        )
+                        + "\n\n"
+                    )
+                st["opened"] = True
+
+            args_done = {
+                "type": "response.function_call_arguments.done",
+                "item_id": st["item_id"],
+                "output_index": st["output_index"],
+                "name": st["name"],
+                "arguments": st["arguments"],
+            }
+            yield f"event: response.function_call_arguments.done\ndata: {json.dumps(args_done)}\n\n"
+
+            item_done = {
+                "type": "response.output_item.done",
+                "output_index": st["output_index"],
+                "item": {
+                    "type": "function_call",
+                    "id": st["item_id"],
+                    "status": "completed",
+                    "call_id": st["call_id"],
+                    "name": st["name"],
+                    "arguments": st["arguments"],
+                },
+            }
+            yield f"event: response.output_item.done\ndata: {json.dumps(item_done)}\n\n"
+
+        # ── Closing events for text message ──
         yield f"event: response.output_text.done\ndata: {json.dumps({'type': 'response.output_text.done', 'item_id': msg_id, 'output_index': 0, 'content_index': 0, 'text': full_text})}\n\n"
 
-        # content_part.done
         yield f"event: response.content_part.done\ndata: {json.dumps({'type': 'response.content_part.done', 'item_id': msg_id, 'output_index': 0, 'content_index': 0, 'part': {'type': 'output_text', 'text': full_text, 'annotations': []}})}\n\n"
 
-        # output_item.done
         yield f"event: response.output_item.done\ndata: {json.dumps({'type': 'response.output_item.done', 'output_index': 0, 'item': {'type': 'message', 'id': msg_id, 'status': 'completed', 'role': 'assistant', 'content': [{'type': 'output_text', 'text': full_text, 'annotations': []}]}})}\n\n"
 
         # response.completed
@@ -2302,21 +2817,7 @@ async def _responses_stream(
                 "created_at": created_at,
                 "status": "completed",
                 "model": payload.model,
-                "output": [
-                    {
-                        "type": "message",
-                        "id": msg_id,
-                        "status": "completed",
-                        "role": "assistant",
-                        "content": [
-                            {
-                                "type": "output_text",
-                                "text": full_text,
-                                "annotations": [],
-                            }
-                        ],
-                    }
-                ],
+                "output": _snapshot_output(),
                 "usage": {
                     "input_tokens": input_tokens,
                     "output_tokens": output_tokens,
diff --git a/studio/backend/tests/test_responses_tool_passthrough.py b/studio/backend/tests/test_responses_tool_passthrough.py
new file mode 100644
index 0000000000..2f1161c329
--- /dev/null
+++ b/studio/backend/tests/test_responses_tool_passthrough.py
@@ -0,0 +1,667 @@
+# SPDX-License-Identifier: AGPL-3.0-only
+# Copyright 2026-present the Unsloth AI Inc. team. All rights reserved.
+
+"""
+Tests for the OpenAI /v1/responses client-side function-calling pass-through.
+
+Covers:
+- ResponsesRequest accepts Responses-shape `tools`, `tool_choice`,
+  `parallel_tool_calls`, and the `function_call` / `function_call_output`
+  input items used for multi-turn tool loops.
+- _translate_responses_tools_to_chat() converts the flat Responses tool
+  shape to the nested Chat Completions shape, drops non-function built-in
+  tools, and returns None for empty lists.
+- _translate_responses_tool_choice_to_chat() passes string choices through
+  and converts {type:function,name:X} to Chat Completions' nested shape.
+- _normalise_responses_input() maps function_call_output items to
+  role="tool" ChatMessages with tool_call_id, and function_call items to
+  assistant messages with tool_calls.
+- _chat_tool_calls_to_responses_output() preserves call_id and drops
+  non-function tool calls.
+- ResponsesOutputFunctionCall and ResponsesResponse round-trip tool-call
+  outputs without losing fields.
+
+No running server or GPU required.
+"""
+
+import os
+import sys
+
+_backend = os.path.join(os.path.dirname(__file__), "..")
+sys.path.insert(0, _backend)
+
+import json
+
+import pytest
+from pydantic import ValidationError
+
+from models.inference import (
+    ChatMessage,
+    ResponsesFunctionCallInputItem,
+    ResponsesFunctionCallOutputInputItem,
+    ResponsesFunctionTool,
+    ResponsesInputMessage,
+    ResponsesOutputFunctionCall,
+    ResponsesOutputMessage,
+    ResponsesOutputTextContent,
+    ResponsesOutputTextPart,
+    ResponsesRequest,
+    ResponsesResponse,
+    ResponsesUnknownContentPart,
+    ResponsesUnknownInputItem,
+    ResponsesUsage,
+)
+from routes.inference import (
+    _chat_tool_calls_to_responses_output,
+    _normalise_responses_input,
+    _translate_responses_tool_choice_to_chat,
+    _translate_responses_tools_to_chat,
+)
+
+
+# =====================================================================
+# Request model — tools / tool_choice / parallel_tool_calls
+# =====================================================================
+
+
+class TestResponsesRequestTools:
+    def test_flat_function_tool_accepted(self):
+        req = ResponsesRequest(
+            input = "hi",
+            tools = [
+                {
+                    "type": "function",
+                    "name": "get_weather",
+                    "description": "Get the weather for a city.",
+                    "parameters": {
+                        "type": "object",
+                        "properties": {"city": {"type": "string"}},
+                        "required": ["city"],
+                    },
+                    "strict": True,
+                }
+            ],
+        )
+        assert req.tools is not None
+        assert req.tools[0]["name"] == "get_weather"
+        assert req.tools[0]["type"] == "function"
+        assert req.tools[0]["strict"] is True
+
+    def test_tool_choice_string_values(self):
+        for choice in ("auto", "required", "none"):
+            req = ResponsesRequest(input = "hi", tool_choice = choice)
+            assert req.tool_choice == choice
+
+    def test_tool_choice_forcing_object(self):
+        req = ResponsesRequest(
+            input = "hi",
+            tool_choice = {"type": "function", "name": "get_weather"},
+        )
+        assert req.tool_choice == {"type": "function", "name": "get_weather"}
+
+    def test_parallel_tool_calls(self):
+        req = ResponsesRequest(input = "hi", parallel_tool_calls = True)
+        assert req.parallel_tool_calls is True
+
+    def test_builtin_tool_type_passes_validation(self):
+        """Non-function built-in tools (web_search, file_search, mcp, ...) must
+        not raise at request validation so SDKs that default to them don't
+        fail on Studio; they are filtered out during translation."""
+        req = ResponsesRequest(
+            input = "hi",
+            tools = [{"type": "web_search_preview"}],
+        )
+        assert req.tools == [{"type": "web_search_preview"}]
+
+    def test_function_tool_model_direct(self):
+        tool = ResponsesFunctionTool(
+            type = "function",
+            name = "send_email",
+            parameters = {"type": "object", "properties": {}},
+        )
+        assert tool.name == "send_email"
+        assert tool.description is None
+
+    def test_function_tool_rejects_other_type(self):
+        with pytest.raises(ValidationError):
+            ResponsesFunctionTool(type = "web_search", name = "x")
+
+
+# =====================================================================
+# Request model — function_call / function_call_output input items
+# =====================================================================
+
+
+class TestResponsesMultiTurnInput:
+    def test_function_call_input_item(self):
+        req = ResponsesRequest(
+            input = [
+                {"role": "user", "content": "Weather in Paris?"},
+                {
+                    "type": "function_call",
+                    "id": "fc_abc",
+                    "call_id": "call_abc",
+                    "name": "get_weather",
+                    "arguments": '{"city": "Paris"}',
+                },
+                {
+                    "type": "function_call_output",
+                    "call_id": "call_abc",
+                    "output": '{"temp": 12}',
+                },
+            ],
+        )
+        assert len(req.input) == 3
+        assert isinstance(req.input[1], ResponsesFunctionCallInputItem)
+        assert req.input[1].call_id == "call_abc"
+        assert isinstance(req.input[2], ResponsesFunctionCallOutputInputItem)
+        assert req.input[2].call_id == "call_abc"
+        assert req.input[2].output == '{"temp": 12}'
+
+    def test_function_call_output_missing_call_id_rejected(self):
+        with pytest.raises(ValidationError):
+            ResponsesFunctionCallOutputInputItem(
+                type = "function_call_output", output = "x"
+            )
+
+    def test_function_call_output_accepts_content_array(self):
+        item = ResponsesFunctionCallOutputInputItem(
+            type = "function_call_output",
+            call_id = "call_1",
+            output = [{"type": "output_text", "text": "done"}],
+        )
+        assert isinstance(item.output, list)
+
+
+# =====================================================================
+# Translators — tools, tool_choice
+# =====================================================================
+
+
+class TestToolsTranslation:
+    def test_flat_to_nested(self):
+        tools = [
+            {
+                "type": "function",
+                "name": "get_weather",
+                "description": "Returns weather.",
+                "parameters": {"type": "object"},
+                "strict": True,
+            }
+        ]
+        out = _translate_responses_tools_to_chat(tools)
+        assert out == [
+            {
+                "type": "function",
+                "function": {
+                    "name": "get_weather",
+                    "description": "Returns weather.",
+                    "parameters": {"type": "object"},
+                    "strict": True,
+                },
+            }
+        ]
+
+    def test_builtin_tools_dropped(self):
+        out = _translate_responses_tools_to_chat(
+            [
+                {"type": "web_search_preview"},
+                {"type": "file_search"},
+                {
+                    "type": "function",
+                    "name": "search",
+                    "parameters": {"type": "object"},
+                },
+            ]
+        )
+        assert len(out) == 1
+        assert out[0]["function"]["name"] == "search"
+
+    def test_empty_returns_none(self):
+        assert _translate_responses_tools_to_chat(None) is None
+        assert _translate_responses_tools_to_chat([]) is None
+
+    def test_only_builtin_tools_returns_none(self):
+        assert (
+            _translate_responses_tools_to_chat([{"type": "web_search_preview"}]) is None
+        )
+
+    def test_description_optional(self):
+        out = _translate_responses_tools_to_chat(
+            [
+                {
+                    "type": "function",
+                    "name": "noop",
+                    "parameters": {"type": "object"},
+                }
+            ]
+        )
+        assert "description" not in out[0]["function"]
+
+
+class TestToolChoiceTranslation:
+    def test_string_passthrough(self):
+        for v in ("auto", "required", "none"):
+            assert _translate_responses_tool_choice_to_chat(v) == v
+
+    def test_none_passthrough(self):
+        assert _translate_responses_tool_choice_to_chat(None) is None
+
+    def test_forcing_object_converted(self):
+        assert _translate_responses_tool_choice_to_chat(
+            {"type": "function", "name": "get_weather"}
+        ) == {"type": "function", "function": {"name": "get_weather"}}
+
+    def test_already_chat_nested_shape_passes_through(self):
+        """If a client happens to send the Chat Completions nested shape,
+        we don't double-wrap it."""
+        already_nested = {"type": "function", "function": {"name": "get_weather"}}
+        assert (
+            _translate_responses_tool_choice_to_chat(already_nested) == already_nested
+        )
+
+    def test_unknown_shape_passes_through(self):
+        obj = {"type": "allowed_tools", "tools": [{"type": "function", "name": "x"}]}
+        assert _translate_responses_tool_choice_to_chat(obj) == obj
+
+
+# =====================================================================
+# _normalise_responses_input — multi-turn tool mapping
+# =====================================================================
+
+
+class TestNormaliseResponsesInputWithTools:
+    def test_function_call_output_maps_to_tool_role(self):
+        payload = ResponsesRequest(
+            input = [
+                {"role": "user", "content": "Weather?"},
+                {
+                    "type": "function_call",
+                    "call_id": "call_1",
+                    "name": "get_weather",
+                    "arguments": "{}",
+                },
+                {
+                    "type": "function_call_output",
+                    "call_id": "call_1",
+                    "output": '{"temp": 20}',
+                },
+            ],
+        )
+        msgs = _normalise_responses_input(payload)
+        assert len(msgs) == 3
+        assert msgs[0].role == "user"
+
+        assert msgs[1].role == "assistant"
+        assert msgs[1].tool_calls is not None
+        assert msgs[1].tool_calls[0]["id"] == "call_1"
+        assert msgs[1].tool_calls[0]["function"]["name"] == "get_weather"
+
+        assert msgs[2].role == "tool"
+        assert msgs[2].tool_call_id == "call_1"
+        assert msgs[2].content == '{"temp": 20}'
+
+    def test_instructions_plus_developer_message_are_merged(self):
+        """Codex CLI sends `instructions` (system prompt) AND a developer
+        message in `input`. Strict chat templates (harmony / gpt-oss, Qwen3,
+        ...) raise "System message must be at the beginning" when two
+        separate system-role messages appear, so we must emit exactly one
+        merged system message at the top.
+        """
+        payload = ResponsesRequest(
+            instructions = "Base instructions.",
+            input = [
+                {"role": "developer", "content": "Developer override."},
+                {"role": "user", "content": "Hi"},
+            ],
+        )
+        msgs = _normalise_responses_input(payload)
+        system_roles = [m for m in msgs if m.role == "system"]
+        assert len(system_roles) == 1
+        assert "Base instructions." in system_roles[0].content
+        assert "Developer override." in system_roles[0].content
+        # System must be the very first message for strict templates.
+        assert msgs[0].role == "system"
+        assert msgs[1].role == "user"
+
+    def test_developer_message_after_user_is_still_hoisted(self):
+        """Multi-turn conversations where a developer message appears after
+        user turns must still produce a single leading system message, not
+        a mid-conversation system that strict templates reject."""
+        payload = ResponsesRequest(
+            input = [
+                {"role": "user", "content": "Hello"},
+                {"role": "assistant", "content": "Hi!"},
+                {"role": "developer", "content": "Updated rules."},
+                {"role": "user", "content": "Continue"},
+            ],
+        )
+        msgs = _normalise_responses_input(payload)
+        assert msgs[0].role == "system"
+        assert "Updated rules." in msgs[0].content
+        for m in msgs[1:]:
+            assert m.role != "system", "no trailing system message permitted"
+
+    def test_no_system_output_when_no_system_input(self):
+        payload = ResponsesRequest(input = "Hi")
+        msgs = _normalise_responses_input(payload)
+        assert all(m.role != "system" for m in msgs)
+
+    def test_multiple_system_messages_in_input_are_merged(self):
+        payload = ResponsesRequest(
+            input = [
+                {"role": "system", "content": "A"},
+                {"role": "system", "content": "B"},
+                {"role": "user", "content": "Hi"},
+            ],
+        )
+        msgs = _normalise_responses_input(payload)
+        assert sum(1 for m in msgs if m.role == "system") == 1
+        assert "A" in msgs[0].content and "B" in msgs[0].content
+
+    def test_content_array_output_serialised_to_json_string(self):
+        payload = ResponsesRequest(
+            input = [
+                {
+                    "type": "function_call_output",
+                    "call_id": "call_1",
+                    "output": [{"type": "output_text", "text": "ok"}],
+                }
+            ],
+        )
+        msgs = _normalise_responses_input(payload)
+        assert msgs[0].role == "tool"
+        # Content is serialised so llama-server sees a string.
+        assert json.loads(msgs[0].content) == [{"type": "output_text", "text": "ok"}]
+
+
+# =====================================================================
+# Response mapping — tool_calls → function_call output items
+# =====================================================================
+
+
+class TestChatToolCallsToResponsesOutput:
+    def test_basic_mapping(self):
+        items = _chat_tool_calls_to_responses_output(
+            [
+                {
+                    "id": "call_abc",
+                    "type": "function",
+                    "function": {
+                        "name": "get_weather",
+                        "arguments": '{"city":"Paris"}',
+                    },
+                }
+            ]
+        )
+        assert len(items) == 1
+        assert items[0]["type"] == "function_call"
+        assert items[0]["call_id"] == "call_abc"
+        assert items[0]["name"] == "get_weather"
+        assert items[0]["arguments"] == '{"city":"Paris"}'
+        assert items[0]["status"] == "completed"
+        assert items[0]["id"].startswith("fc_")
+
+    def test_multiple_tool_calls_preserved(self):
+        items = _chat_tool_calls_to_responses_output(
+            [
+                {
+                    "id": "call_1",
+                    "type": "function",
+                    "function": {"name": "a", "arguments": "{}"},
+                },
+                {
+                    "id": "call_2",
+                    "type": "function",
+                    "function": {"name": "b", "arguments": "{}"},
+                },
+            ]
+        )
+        assert [it["call_id"] for it in items] == ["call_1", "call_2"]
+
+    def test_non_function_tool_call_dropped(self):
+        items = _chat_tool_calls_to_responses_output([{"id": "x", "type": "retrieval"}])
+        assert items == []
+
+    def test_missing_arguments_coerced_to_empty_string(self):
+        items = _chat_tool_calls_to_responses_output(
+            [{"id": "call_1", "type": "function", "function": {"name": "x"}}]
+        )
+        assert items[0]["arguments"] == ""
+
+
+# =====================================================================
+# Response model — ResponsesOutputFunctionCall / mixed output
+# =====================================================================
+
+
+class TestResponsesOutputFunctionCall:
+    def test_direct_construction(self):
+        fc = ResponsesOutputFunctionCall(
+            call_id = "call_1",
+            name = "get_weather",
+            arguments = '{"city":"Paris"}',
+        )
+        d = fc.model_dump()
+        assert d["type"] == "function_call"
+        assert d["call_id"] == "call_1"
+        assert d["status"] == "completed"
+        assert d["id"].startswith("fc_")
+
+    def test_response_with_tool_call_output(self):
+        resp = ResponsesResponse(
+            model = "test",
+            output = [
+                ResponsesOutputFunctionCall(
+                    call_id = "call_1",
+                    name = "get_weather",
+                    arguments = "{}",
+                )
+            ],
+            usage = ResponsesUsage(input_tokens = 1, output_tokens = 1, total_tokens = 2),
+        )
+        d = json.loads(resp.model_dump_json())
+        assert d["output"][0]["type"] == "function_call"
+        assert d["output"][0]["call_id"] == "call_1"
+
+    def test_response_with_mixed_output(self):
+        resp = ResponsesResponse(
+            model = "test",
+            output = [
+                ResponsesOutputMessage(
+                    content = [ResponsesOutputTextContent(text = "Calling...")],
+                ),
+                ResponsesOutputFunctionCall(
+                    call_id = "call_1",
+                    name = "get_weather",
+                    arguments = '{"city":"Paris"}',
+                ),
+            ],
+        )
+        d = resp.model_dump()
+        assert d["output"][0]["type"] == "message"
+        assert d["output"][1]["type"] == "function_call"
+
+
+# =====================================================================
+# Regression: ChatMessage validator still accepts mapped tool messages
+# =====================================================================
+
+
+class TestCodexStyleRequestShapes:
+    """Regression tests for the request shapes OpenAI Codex CLI sends."""
+
+    def test_assistant_replay_output_text_accepted(self):
+        """Codex replays prior assistant turns with `output_text` content.
+        Before, this triggered a 422 on every turn after the first."""
+        req = ResponsesRequest(
+            input = [
+                {"role": "user", "content": "Hi"},
+                {
+                    "type": "message",
+                    "role": "assistant",
+                    "content": [
+                        {
+                            "type": "output_text",
+                            "text": "Hello!",
+                            "annotations": [],
+                            "logprobs": [],
+                        }
+                    ],
+                },
+                {"role": "user", "content": "Continue"},
+            ],
+        )
+        assert len(req.input) == 3
+        parts = req.input[1].content
+        assert isinstance(parts, list)
+        assert isinstance(parts[0], ResponsesOutputTextPart)
+        assert parts[0].text == "Hello!"
+
+    def test_reasoning_item_accepted_as_unknown(self):
+        """`reasoning` items replayed from prior o-series turns must not
+        fail validation — Codex preserves them in multi-turn."""
+        req = ResponsesRequest(
+            input = [
+                {"role": "user", "content": "Hi"},
+                {
+                    "type": "reasoning",
+                    "id": "rs_1",
+                    "summary": [],
+                    "encrypted_content": "opaque",
+                },
+                {"role": "assistant", "content": "Hello!"},
+            ],
+        )
+        assert len(req.input) == 3
+        assert isinstance(req.input[1], ResponsesUnknownInputItem)
+
+    def test_unknown_content_part_type_accepted(self):
+        """Unknown content-part types (e.g. future input_audio) validate as
+        ResponsesUnknownContentPart so the whole request doesn't 422."""
+        req = ResponsesRequest(
+            input = [
+                {
+                    "role": "user",
+                    "content": [
+                        {"type": "input_text", "text": "See:"},
+                        {"type": "input_audio", "audio": {"data": "..."}},
+                    ],
+                }
+            ],
+        )
+        parts = req.input[0].content
+        assert isinstance(parts[1], ResponsesUnknownContentPart)
+        assert parts[1].type == "input_audio"
+
+    def test_codex_full_shape_roundtrip(self):
+        """End-to-end: developer + user + assistant(output_text) +
+        function_call + function_call_output + reasoning in one request."""
+        payload = ResponsesRequest(
+            instructions = "Base instructions.",
+            input = [
+                {
+                    "type": "message",
+                    "role": "developer",
+                    "content": [{"type": "input_text", "text": "Dev override."}],
+                },
+                {
+                    "type": "message",
+                    "role": "user",
+                    "content": [{"type": "input_text", "text": "Weather?"}],
+                },
+                {
+                    "type": "reasoning",
+                    "id": "rs_1",
+                    "summary": [],
+                },
+                {
+                    "type": "function_call",
+                    "call_id": "call_1",
+                    "name": "get_weather",
+                    "arguments": "{}",
+                },
+                {
+                    "type": "function_call_output",
+                    "call_id": "call_1",
+                    "output": '{"temp":20}',
+                },
+                {
+                    "type": "message",
+                    "role": "assistant",
+                    "content": [
+                        {
+                            "type": "output_text",
+                            "text": "It's 20°C.",
+                            "annotations": [],
+                            "logprobs": [],
+                        }
+                    ],
+                },
+                {"role": "user", "content": "And tomorrow?"},
+            ],
+        )
+        msgs = _normalise_responses_input(payload)
+        # Single leading merged system; no mid-conversation system.
+        assert msgs[0].role == "system"
+        assert sum(1 for m in msgs if m.role == "system") == 1
+        assert "Base instructions." in msgs[0].content
+        assert "Dev override." in msgs[0].content
+
+        roles = [m.role for m in msgs[1:]]
+        # Reasoning item is dropped. Order: user, assistant(tool_calls),
+        # tool, assistant(text), user.
+        assert roles == ["user", "assistant", "tool", "assistant", "user"]
+        assert msgs[2].tool_calls is not None
+        assert msgs[3].role == "tool"
+        assert msgs[3].tool_call_id == "call_1"
+        assert msgs[4].content == "It's 20°C."
+
+    def test_single_output_text_part_flattens_to_string(self):
+        """ChatMessage assistant role prefers plain string content — tests
+        confirm we don't forward a single-part array that would otherwise
+        force legacy chat templates into multimodal handling."""
+        payload = ResponsesRequest(
+            input = [
+                {
+                    "role": "assistant",
+                    "content": [
+                        {"type": "output_text", "text": "ok", "annotations": []}
+                    ],
+                },
+                {"role": "user", "content": "next"},
+            ],
+        )
+        msgs = _normalise_responses_input(payload)
+        assert msgs[0].role == "assistant"
+        assert msgs[0].content == "ok"
+
+
+class TestTranslatedMessagesValidate:
+    """Verify that the messages produced by _normalise_responses_input
+    satisfy ChatMessage's role-shape validator so the downstream /v1/chat/
+    completions pass-through does not reject them."""
+
+    def test_round_trip_multi_turn(self):
+        payload = ResponsesRequest(
+            input = [
+                {"role": "user", "content": "Weather in Paris?"},
+                {
+                    "type": "function_call",
+                    "call_id": "call_1",
+                    "name": "get_weather",
+                    "arguments": '{"city": "Paris"}',
+                },
+                {
+                    "type": "function_call_output",
+                    "call_id": "call_1",
+                    "output": '{"temp": 20}',
+                },
+                {"role": "user", "content": "Thanks!"},
+            ],
+        )
+        msgs = _normalise_responses_input(payload)
+        for m in msgs:
+            # Constructing a fresh ChatMessage from the dump round-trips the
+            # role-shape validator — the key invariant for the passthrough.
+            ChatMessage(**m.model_dump(exclude_none = True))

From b13ce6556a4d17c817e92ca554baa3ed527bd6ea Mon Sep 17 00:00:00 2001
From: Roland Tannous <115670425+rolandtannous@users.noreply.github.com>
Date: Wed, 22 Apr 2026 00:13:59 +0400
Subject: [PATCH 144/216] Update model_mappings.py

add Qwen3.5-9B to studio qwen3.5 mappings
---
 studio/backend/utils/datasets/model_mappings.py | 1 +
 1 file changed, 1 insertion(+)

diff --git a/studio/backend/utils/datasets/model_mappings.py b/studio/backend/utils/datasets/model_mappings.py
index 7fcac637c6..36f6886ef6 100644
--- a/studio/backend/utils/datasets/model_mappings.py
+++ b/studio/backend/utils/datasets/model_mappings.py
@@ -369,6 +369,7 @@ TEMPLATE_TO_MODEL_MAPPER = {
         "unsloth/Qwen3.5-0.8B",
         "unsloth/Qwen3.5-2B",
         "unsloth/Qwen3.5-4B",
+        "unsloth/Qwen3.5-9B",
         "unsloth/Qwen3.5-27B",
         "unsloth/Qwen3.5-35B-A3B",
     ),

From 92cee0ff3e4a88cadd5b9ae01b7627bd1c92394a Mon Sep 17 00:00:00 2001
From: Roland Tannous <115670425+rolandtannous@users.noreply.github.com>
Date: Wed, 22 Apr 2026 03:25:07 +0400
Subject: [PATCH 145/216] Studio: support images on /v1/messages
 (Anthropic-compat) (#5128)

* Studio: support images on /v1/messages (Anthropic-compat)

Translate Anthropic `image` content blocks (base64 and url sources) to
OpenAI `image_url` multimodal parts so the Anthropic endpoint reaches
llama-server's native vision path. Mirrors the `/v1/chat/completions`
vision behavior: 400 when the active GGUF isn't a vision model, and
embedded images are re-encoded to PNG (stb_image format coverage).
Server-side agentic loop is disabled when images are present, matching
the existing `not image_b64` gate on /v1/chat/completions.

Adds translator + normalizer unit tests.

* Studio: address gemini-code-assist review on /v1/messages image support

- Preserve interleaving of Anthropic text + image content blocks in the
  translator (previously flattened all text first, then all images).
- Let _normalize_anthropic_openai_images return has_image so the route
  skips the second scan it was doing.
- Use module-level base64/io in the helper instead of re-importing.
---
 .../core/inference/anthropic_compat.py        | 139 +++++++---
 studio/backend/routes/inference.py            |  70 ++++-
 .../backend/tests/test_anthropic_messages.py  | 239 ++++++++++++++++++
 3 files changed, 405 insertions(+), 43 deletions(-)

diff --git a/studio/backend/core/inference/anthropic_compat.py b/studio/backend/core/inference/anthropic_compat.py
index 400e6053e0..263718c540 100644
--- a/studio/backend/core/inference/anthropic_compat.py
+++ b/studio/backend/core/inference/anthropic_compat.py
@@ -13,11 +13,44 @@ import json
 from typing import Any, Optional, Union
 
 
+def _anthropic_image_block_to_openai_part(block: dict) -> Optional[dict]:
+    """Translate one Anthropic ``image`` block to an OpenAI ``image_url`` part.
+
+    Accepts both source shapes:
+      - ``{"type": "base64", "media_type": "image/jpeg", "data": "..."}``
+      - ``{"type": "url", "url": "https://..."}``
+
+    Returns ``None`` when the source is malformed so the caller can skip it.
+    """
+    source = block.get("source") or {}
+    stype = source.get("type")
+    if stype == "base64":
+        data = source.get("data")
+        if not data:
+            return None
+        media_type = source.get("media_type") or "image/jpeg"
+        return {
+            "type": "image_url",
+            "image_url": {"url": f"data:{media_type};base64,{data}"},
+        }
+    if stype == "url":
+        url = source.get("url")
+        if not url:
+            return None
+        return {"type": "image_url", "image_url": {"url": url}}
+    return None
+
+
 def anthropic_messages_to_openai(
     messages: list[dict],
     system: Optional[Union[str, list]] = None,
 ) -> list[dict]:
-    """Convert Anthropic messages + system to OpenAI-format message dicts."""
+    """Convert Anthropic messages + system to OpenAI-format message dicts.
+
+    User messages that carry ``image`` blocks are emitted as OpenAI
+    multimodal content arrays (``[{type: "text", ...}, {type: "image_url", ...}]``)
+    so they flow through llama-server's native vision pathway.
+    """
     result: list[dict] = []
 
     # System prompt
@@ -42,54 +75,76 @@ def anthropic_messages_to_openai(
             result.append({"role": role, "content": content})
             continue
 
-        # Content is a list of blocks
-        text_parts: list[str] = []
-        tool_calls: list[dict] = []
-        tool_results: list[dict] = []
-
-        for block in content:
-            b = block if isinstance(block, dict) else block.model_dump()
-            btype = b.get("type", "")
-
-            if btype == "text":
-                text_parts.append(b["text"])
-            elif btype == "tool_use":
-                tool_calls.append(
-                    {
-                        "id": b["id"],
-                        "type": "function",
-                        "function": {
-                            "name": b["name"],
-                            "arguments": json.dumps(b["input"]),
-                        },
-                    }
-                )
-            elif btype == "tool_result":
-                tc = b.get("content", "")
-                if isinstance(tc, list):
-                    tc = " ".join(
-                        p["text"]
-                        for p in tc
-                        if isinstance(p, dict) and p.get("type") == "text"
-                    )
-                tool_results.append(
-                    {
-                        "role": "tool",
-                        "tool_call_id": b["tool_use_id"],
-                        "content": str(tc),
-                    }
-                )
-
         if role == "assistant":
+            # Assistant content carries text + tool_use; images aren't
+            # part of Anthropic's assistant content model.
+            text_parts: list[str] = []
+            tool_calls: list[dict] = []
+            for block in content:
+                b = block if isinstance(block, dict) else block.model_dump()
+                btype = b.get("type", "")
+                if btype == "text":
+                    text_parts.append(b["text"])
+                elif btype == "tool_use":
+                    tool_calls.append(
+                        {
+                            "id": b["id"],
+                            "type": "function",
+                            "function": {
+                                "name": b["name"],
+                                "arguments": json.dumps(b["input"]),
+                            },
+                        }
+                    )
             msg_dict: dict[str, Any] = {"role": "assistant"}
             if text_parts:
                 msg_dict["content"] = "\n".join(text_parts)
             if tool_calls:
                 msg_dict["tool_calls"] = tool_calls
             result.append(msg_dict)
-        elif role == "user":
-            if text_parts:
-                result.append({"role": "user", "content": "\n".join(text_parts)})
+            continue
+
+        if role == "user":
+            # Build an ordered part list so text/image interleaving is
+            # preserved (e.g. [text, image, text, image]). tool_result
+            # blocks become their own OpenAI "tool" role messages.
+            user_parts: list[dict] = []
+            has_image = False
+            tool_results: list[dict] = []
+            for block in content:
+                b = block if isinstance(block, dict) else block.model_dump()
+                btype = b.get("type", "")
+                if btype == "text":
+                    user_parts.append({"type": "text", "text": b["text"]})
+                elif btype == "image":
+                    part = _anthropic_image_block_to_openai_part(b)
+                    if part is not None:
+                        user_parts.append(part)
+                        has_image = True
+                elif btype == "tool_result":
+                    tc = b.get("content", "")
+                    if isinstance(tc, list):
+                        tc = " ".join(
+                            p["text"]
+                            for p in tc
+                            if isinstance(p, dict) and p.get("type") == "text"
+                        )
+                    tool_results.append(
+                        {
+                            "role": "tool",
+                            "tool_call_id": b["tool_use_id"],
+                            "content": str(tc),
+                        }
+                    )
+
+            if has_image:
+                result.append({"role": "user", "content": user_parts})
+            else:
+                # No images — collapse text parts to a plain string so
+                # existing text-only callers keep their simple shape.
+                text = "\n".join(p["text"] for p in user_parts)
+                if text:
+                    result.append({"role": "user", "content": text})
             for tr in tool_results:
                 result.append(tr)
 
diff --git a/studio/backend/routes/inference.py b/studio/backend/routes/inference.py
index 9c84eabd8b..e17f8f5882 100644
--- a/studio/backend/routes/inference.py
+++ b/studio/backend/routes/inference.py
@@ -2866,6 +2866,64 @@ async def openai_responses(
 # =====================================================================
 
 
+def _normalize_anthropic_openai_images(
+    openai_messages: list[dict], is_vision: bool
+) -> bool:
+    """Enforce the vision guard on translated Anthropic messages and
+    normalize any ``image_url`` parts with base64 data URLs to PNG.
+
+    llama-server's stb_image only handles a few formats (JPEG/PNG/BMP/…);
+    Anthropic clients commonly send JPEG or WebP, and Claude Code sends
+    WebP. Re-encoding everything to PNG mirrors the behavior of
+    `_openai_messages_for_passthrough` / the GGUF branch of
+    `/v1/chat/completions` so the two endpoints agree.
+
+    Mutates ``openai_messages`` in place. Returns ``True`` when any
+    image part was seen (so the caller can skip a second scan). Raises
+    HTTPException(400) when images are present but the active model is
+    not a vision model, or when an image cannot be decoded.
+    """
+    from PIL import Image
+
+    has_image = False
+    for msg in openai_messages:
+        content = msg.get("content")
+        if not isinstance(content, list):
+            continue
+        for part in content:
+            if part.get("type") != "image_url":
+                continue
+
+            has_image = True
+            if not is_vision:
+                raise HTTPException(
+                    status_code = 400,
+                    detail = "Image provided but current GGUF model does not support vision.",
+                )
+
+            url = (part.get("image_url") or {}).get("url", "")
+            if not url.startswith("data:"):
+                # Remote URLs are forwarded as-is; llama-server will
+                # fetch (or fail) per its own support matrix.
+                continue
+
+            try:
+                _, b64data = url.split(",", 1)
+                raw = base64.b64decode(b64data)
+                img = Image.open(io.BytesIO(raw)).convert("RGB")
+                buf = io.BytesIO()
+                img.save(buf, format = "PNG")
+                png_b64 = base64.b64encode(buf.getvalue()).decode("ascii")
+            except Exception as e:
+                raise HTTPException(
+                    status_code = 400,
+                    detail = f"Failed to process image: {e}",
+                )
+            part["image_url"] = {"url": f"data:image/png;base64,{png_b64}"}
+
+    return has_image
+
+
 @router.post("/messages")
 async def anthropic_messages(
     payload: AnthropicMessagesRequest,
@@ -2896,6 +2954,12 @@ async def anthropic_messages(
         payload.system,
     )
 
+    # Enforce vision guard + re-encode embedded images to PNG so the
+    # Anthropic endpoint matches the behavior of /v1/chat/completions.
+    _has_image = _normalize_anthropic_openai_images(
+        openai_messages, llama_backend.is_vision
+    )
+
     temperature = payload.temperature if payload.temperature is not None else 0.6
     top_p = payload.top_p if payload.top_p is not None else 0.95
     top_k = payload.top_k if payload.top_k is not None else 20
@@ -2922,7 +2986,11 @@ async def anthropic_messages(
     # 1. enable_tools=true → server-side execution of built-in tools (Unsloth shorthand)
     # 2. tools=[...] only  → client-side pass-through (standard Anthropic behavior)
     # 3. neither           → plain chat
-    server_tools = payload.enable_tools and llama_backend.supports_tools
+    # Server-side agentic loop doesn't support multimodal input — matches
+    # the `not image_b64` gate in /v1/chat/completions.
+    server_tools = (
+        payload.enable_tools and llama_backend.supports_tools and not _has_image
+    )
     client_tools = (
         not server_tools
         and payload.tools
diff --git a/studio/backend/tests/test_anthropic_messages.py b/studio/backend/tests/test_anthropic_messages.py
index ec432df9e6..0825ef9337 100644
--- a/studio/backend/tests/test_anthropic_messages.py
+++ b/studio/backend/tests/test_anthropic_messages.py
@@ -10,6 +10,8 @@ import sys
 import os
 import json
 
+import pytest
+
 _backend = os.path.join(os.path.dirname(__file__), "..")
 sys.path.insert(0, _backend)
 
@@ -32,6 +34,10 @@ from core.inference.anthropic_compat import (
     AnthropicStreamEmitter,
     AnthropicPassthroughEmitter,
 )
+from routes.inference import _normalize_anthropic_openai_images
+from fastapi import HTTPException
+import base64 as _b64
+from io import BytesIO as _BytesIO
 
 
 # =====================================================================
@@ -246,6 +252,141 @@ class TestAnthropicMessagesToOpenAI:
         result = anthropic_messages_to_openai(msgs)
         assert result[0]["content"] == "Line 1 Line 2"
 
+    def test_image_base64_block_becomes_multimodal_part(self):
+        msgs = [
+            {
+                "role": "user",
+                "content": [
+                    {"type": "text", "text": "What is this?"},
+                    {
+                        "type": "image",
+                        "source": {
+                            "type": "base64",
+                            "media_type": "image/jpeg",
+                            "data": "AAAA",
+                        },
+                    },
+                ],
+            }
+        ]
+        result = anthropic_messages_to_openai(msgs)
+        assert len(result) == 1
+        assert result[0]["role"] == "user"
+        parts = result[0]["content"]
+        assert isinstance(parts, list)
+        assert parts[0] == {"type": "text", "text": "What is this?"}
+        assert parts[1]["type"] == "image_url"
+        assert parts[1]["image_url"]["url"] == "data:image/jpeg;base64,AAAA"
+
+    def test_image_url_block_forwarded_as_url(self):
+        msgs = [
+            {
+                "role": "user",
+                "content": [
+                    {"type": "text", "text": "Describe it"},
+                    {
+                        "type": "image",
+                        "source": {"type": "url", "url": "https://x/y.png"},
+                    },
+                ],
+            }
+        ]
+        result = anthropic_messages_to_openai(msgs)
+        parts = result[0]["content"]
+        assert parts[1] == {
+            "type": "image_url",
+            "image_url": {"url": "https://x/y.png"},
+        }
+
+    def test_image_only_user_message_emits_no_text_part(self):
+        msgs = [
+            {
+                "role": "user",
+                "content": [
+                    {
+                        "type": "image",
+                        "source": {
+                            "type": "base64",
+                            "media_type": "image/png",
+                            "data": "ZZ",
+                        },
+                    },
+                ],
+            }
+        ]
+        result = anthropic_messages_to_openai(msgs)
+        parts = result[0]["content"]
+        assert len(parts) == 1
+        assert parts[0]["type"] == "image_url"
+
+    def test_image_default_media_type_when_missing(self):
+        msgs = [
+            {
+                "role": "user",
+                "content": [
+                    {
+                        "type": "image",
+                        "source": {"type": "base64", "data": "BB"},
+                    },
+                ],
+            }
+        ]
+        result = anthropic_messages_to_openai(msgs)
+        parts = result[0]["content"]
+        assert parts[0]["image_url"]["url"].startswith("data:image/jpeg;base64,")
+
+    def test_image_text_order_preserved(self):
+        # [text1, image1, text2, image2] must not collapse to
+        # [text1+text2, image1, image2].
+        msgs = [
+            {
+                "role": "user",
+                "content": [
+                    {"type": "text", "text": "before"},
+                    {
+                        "type": "image",
+                        "source": {
+                            "type": "base64",
+                            "media_type": "image/png",
+                            "data": "AA",
+                        },
+                    },
+                    {"type": "text", "text": "after"},
+                    {
+                        "type": "image",
+                        "source": {"type": "url", "url": "https://x/y.png"},
+                    },
+                ],
+            }
+        ]
+        result = anthropic_messages_to_openai(msgs)
+        parts = result[0]["content"]
+        assert [p["type"] for p in parts] == [
+            "text",
+            "image_url",
+            "text",
+            "image_url",
+        ]
+        assert parts[0]["text"] == "before"
+        assert parts[2]["text"] == "after"
+        assert parts[1]["image_url"]["url"] == "data:image/png;base64,AA"
+        assert parts[3]["image_url"]["url"] == "https://x/y.png"
+
+    def test_malformed_image_block_is_skipped(self):
+        msgs = [
+            {
+                "role": "user",
+                "content": [
+                    {"type": "text", "text": "Hi"},
+                    {"type": "image", "source": {"type": "base64"}},
+                    {"type": "image", "source": {"type": "url"}},
+                ],
+            }
+        ]
+        result = anthropic_messages_to_openai(msgs)
+        # No image parts emitted; message falls back to plain text.
+        assert result[0] == {"role": "user", "content": "Hi"}
+
 
 # =====================================================================
 # Tool translation tests
@@ -772,3 +913,101 @@ class TestAnthropicPassthroughEmitter:
         parsed = self._parse(events[1])
         assert parsed["content_block"]["name"] == "Read"
         assert parsed["content_block"]["id"] == "c2"
+
+
+# =====================================================================
+# Vision guard + PNG normalization (/v1/messages)
+# =====================================================================
+
+
+def _jpeg_data_url() -> str:
+    from PIL import Image
+
+    img = Image.new("RGB", (2, 2), (255, 0, 0))
+    buf = _BytesIO()
+    img.save(buf, format = "JPEG")
+    b64 = _b64.b64encode(buf.getvalue()).decode("ascii")
+    return f"data:image/jpeg;base64,{b64}"
+
+
+class TestNormalizeAnthropicOpenAIImages:
+    def test_noop_when_no_images(self):
+        msgs = [{"role": "user", "content": "hi"}]
+        has_image = _normalize_anthropic_openai_images(msgs, is_vision = False)
+        assert has_image is False
+        assert msgs == [{"role": "user", "content": "hi"}]
+
+    def test_returns_true_when_image_present(self):
+        msgs = [
+            {
+                "role": "user",
+                "content": [
+                    {"type": "image_url", "image_url": {"url": _jpeg_data_url()}},
+                ],
+            }
+        ]
+        assert _normalize_anthropic_openai_images(msgs, is_vision = True) is True
+
+    def test_rejects_image_when_model_not_vision(self):
+        msgs = [
+            {
+                "role": "user",
+                "content": [
+                    {"type": "text", "text": "?"},
+                    {
+                        "type": "image_url",
+                        "image_url": {"url": _jpeg_data_url()},
+                    },
+                ],
+            }
+        ]
+        with pytest.raises(HTTPException) as exc:
+            _normalize_anthropic_openai_images(msgs, is_vision = False)
+        assert exc.value.status_code == 400
+
+    def test_reencodes_jpeg_data_url_to_png(self):
+        original_url = _jpeg_data_url()
+        msgs = [
+            {
+                "role": "user",
+                "content": [
+                    {"type": "text", "text": "?"},
+                    {"type": "image_url", "image_url": {"url": original_url}},
+                ],
+            }
+        ]
+        _normalize_anthropic_openai_images(msgs, is_vision = True)
+        new_url = msgs[0]["content"][1]["image_url"]["url"]
+        assert new_url.startswith("data:image/png;base64,")
+        assert new_url != original_url
+
+    def test_remote_url_left_unchanged(self):
+        msgs = [
+            {
+                "role": "user",
+                "content": [
+                    {
+                        "type": "image_url",
+                        "image_url": {"url": "https://x.example/y.png"},
+                    },
+                ],
+            }
+        ]
+        _normalize_anthropic_openai_images(msgs, is_vision = True)
+        assert msgs[0]["content"][0]["image_url"]["url"] == "https://x.example/y.png"
+
+    def test_bad_base64_raises_400(self):
+        msgs = [
+            {
+                "role": "user",
+                "content": [
+                    {
+                        "type": "image_url",
+                        "image_url": {"url": "data:image/jpeg;base64,!!!not-b64!!!"},
+                    },
+                ],
+            }
+        ]
+        with pytest.raises(HTTPException) as exc:
+            _normalize_anthropic_openai_images(msgs, is_vision = True)
+        assert exc.value.status_code == 400

From 7ef8cde3c22a3bc2240353353caa21051a42ea90 Mon Sep 17 00:00:00 2001
From: Daniel Han <danielhanchen@gmail.com>
Date: Tue, 21 Apr 2026 22:39:35 -0700
Subject: [PATCH 146/216] Coerce TRL's tuple-cached _*_available flags to bool
 (#5129)

transformers >= 4.48's `_is_package_available(name)` returns a tuple
`(bool, version_or_None)`. TRL's `trl.import_utils` caches that tuple
directly in `_vllm_ascend_available`, `_llm_blender_available`,
`_deepspeed_available`, `_joblib_available`, etc. and the matching
`is_*_available()` accessors return the tuple unchanged. A non-empty
tuple is always truthy, so `if is_vllm_ascend_available():` (in
`trl/extras/vllm_client.py`) fires unconditionally and triggers
`from vllm_ascend.distributed.device_communicators.pyhccl import ...`,
which fails outside Huawei Ascend hosts and blocks
`from trl import GRPOConfig, GRPOTrainer`. The same shape blocks
`is_llm_blender_available()` -> `import llm_blender` in
`trl/trainer/judges.py`.

Add `fix_trl_vllm_ascend()` to `import_fixes.py` and call it from
`unsloth/__init__.py` before any `from .trainer import *` that would
eagerly `import trl`. The fix walks `trl.import_utils` once and coerces
every `_*_available` tuple to a bool; the existing accessors that just
return the cached value then naturally yield a bool and the `if`
checks behave.
---
 unsloth/__init__.py     |  3 +++
 unsloth/import_fixes.py | 26 ++++++++++++++++++++++++++
 2 files changed, 29 insertions(+)

diff --git a/unsloth/__init__.py b/unsloth/__init__.py
index 73a9ce8fef..52114bb544 100644
--- a/unsloth/__init__.py
+++ b/unsloth/__init__.py
@@ -137,6 +137,7 @@ from .import_fixes import (
     fix_vllm_aimv2_issue,
     check_vllm_torch_sm100_compatibility,
     fix_vllm_guided_decoding_params,
+    fix_trl_vllm_ascend,
     fix_vllm_pdl_blackwell,
     fix_triton_compiled_kernel_missing_attrs,
     patch_trunc_normal_precision_issue,
@@ -159,6 +160,7 @@ fix_vllm_aimv2_issue()
 # Check vLLM + torch < 2.9.0 + SM100 compatibility BEFORE importing vLLM
 check_vllm_torch_sm100_compatibility()
 fix_vllm_guided_decoding_params()
+fix_trl_vllm_ascend()
 fix_vllm_pdl_blackwell()
 fix_triton_compiled_kernel_missing_attrs()
 patch_trunc_normal_precision_issue()
@@ -179,6 +181,7 @@ del fix_xformers_performance_issue
 del fix_vllm_aimv2_issue
 del check_vllm_torch_sm100_compatibility
 del fix_vllm_guided_decoding_params
+del fix_trl_vllm_ascend
 del fix_vllm_pdl_blackwell
 del fix_triton_compiled_kernel_missing_attrs
 del patch_trunc_normal_precision_issue
diff --git a/unsloth/import_fixes.py b/unsloth/import_fixes.py
index ca44a0ce7e..6cebad7939 100644
--- a/unsloth/import_fixes.py
+++ b/unsloth/import_fixes.py
@@ -489,6 +489,32 @@ def fix_vllm_guided_decoding_params():
         )
 
 
+def fix_trl_vllm_ascend():
+    # transformers >= 4.48's `_is_package_available(name)` returns a tuple
+    # (bool, version_or_None). TRL caches that tuple in module-level
+    # `_*_available` flags and the matching `is_*_available()` accessors
+    # return the tuple directly. A non-empty tuple is always truthy, so
+    # `if is_X_available():` fires even when X is absent, triggering an
+    # unconditional `import X` that fails. The surfaced case is
+    # `vllm_ascend` (blocks `from trl import GRPOConfig, GRPOTrainer`
+    # outside Huawei Ascend hosts); `llm_blender`, `deepspeed`, `joblib`
+    # share the same shape. Coerce every tuple-cached flag in
+    # trl.import_utils to bool; the existing accessors that just return
+    # the cached value then naturally yield a bool.
+    if importlib.util.find_spec("trl") is None:
+        return
+    try:
+        import trl.import_utils as tiu
+    except Exception:
+        return
+    for attr in list(vars(tiu)):
+        if not (attr.startswith("_") and attr.endswith("_available")):
+            continue
+        cached = getattr(tiu, attr)
+        if isinstance(cached, tuple):
+            setattr(tiu, attr, bool(cached and cached[0]))
+
+
 def ignore_logger_messages():
     # Ignore Environment variable `HF_TOKEN` is set
     try:

From 4c53191de902a8f2e415d0a722633f45b4466d2e Mon Sep 17 00:00:00 2001
From: Lee Jackson <130007945+Imagineer99@users.noreply.github.com>
Date: Wed, 22 Apr 2026 15:34:31 +0100
Subject: [PATCH 147/216] Studio: Smoother thread switching in chat (#5126)

* fix: keep single chat runtime mounted across thread selection

* fix: stabilize thread switching and remove transition flicker

* chore: keep hidden-welcome thread footer chrome mounted

* fix: guard composer during thread attach

* fix: recover from stale routed chat threads

* fix: notify when routed chat is missing

* fix: ensure stale chat recovery opens new chat
---
 .../src/components/assistant-ui/thread.tsx    | 59 ++++++++++++++-----
 .../frontend/src/features/chat/chat-page.tsx  | 33 ++++++++++-
 .../src/features/chat/runtime-provider.tsx    | 11 +++-
 3 files changed, 86 insertions(+), 17 deletions(-)

diff --git a/studio/frontend/src/components/assistant-ui/thread.tsx b/studio/frontend/src/components/assistant-ui/thread.tsx
index 54f64220e8..4c6569c2ee 100644
--- a/studio/frontend/src/components/assistant-ui/thread.tsx
+++ b/studio/frontend/src/components/assistant-ui/thread.tsx
@@ -60,15 +60,31 @@ import {
   Trash2Icon,
   XIcon,
 } from "lucide-react";
-import { type FC, useCallback, useEffect, useRef, useState } from "react";
+import {
+  type FC,
+  type FormEvent,
+  useCallback,
+  useEffect,
+  useRef,
+  useState,
+} from "react";
 import { toast } from "sonner";
 import { deleteThreadMessage } from "@/features/chat/utils/delete-thread-message";
 import { useChatRuntimeStore } from "@/features/chat/stores/chat-runtime-store";
 
-export const Thread: FC<{ hideComposer?: boolean; hideWelcome?: boolean }> = ({
+export const Thread: FC<{
+  hideComposer?: boolean;
+  hideWelcome?: boolean;
+  targetThreadId?: string;
+}> = ({
   hideComposer,
   hideWelcome,
+  targetThreadId,
 }) => {
+  const isComposerAttachPending = useAuiState(({ threads }) =>
+    targetThreadId ? threads.mainThreadId !== targetThreadId : false,
+  );
+
   return (
     <ThreadPrimitive.Root
       className="aui-root aui-thread-root @container relative flex min-h-0 min-w-0 flex-1 basis-0 flex-col overflow-hidden"
@@ -85,7 +101,7 @@ export const Thread: FC<{ hideComposer?: boolean; hideWelcome?: boolean }> = ({
         )}
       >
         {!hideWelcome && (
-          <AuiIf condition={({ thread }) => thread.isEmpty}>
+          <AuiIf condition={({ thread }) => thread.isEmpty && !thread.isLoading}>
             <ThreadWelcome hideComposer={hideComposer} />
           </AuiIf>
         )}
@@ -102,14 +118,14 @@ export const Thread: FC<{ hideComposer?: boolean; hideWelcome?: boolean }> = ({
             sticky scroll-to-bottom button (and the floating composer in
             single mode). Without this, content would butt against the
             sticky footer and feel cramped. */}
-        <AuiIf condition={({ thread }) => !thread.isEmpty}>
+        <AuiIf condition={({ thread }) => hideWelcome || !thread.isEmpty}>
           <div
             className={cn("shrink-0", hideComposer ? "h-16" : "h-40")}
             aria-hidden
           />
         </AuiIf>
 
-        <AuiIf condition={({ thread }) => !thread.isEmpty}>
+        <AuiIf condition={({ thread }) => hideWelcome || !thread.isEmpty}>
           <ThreadPrimitive.ViewportFooter
             className={cn(
               "aui-thread-viewport-footer pointer-events-none sticky z-20 flex w-full justify-center bg-transparent",
@@ -122,7 +138,7 @@ export const Thread: FC<{ hideComposer?: boolean; hideWelcome?: boolean }> = ({
       </ThreadPrimitive.Viewport>
 
       {!hideComposer && (
-        <AuiIf condition={({ thread }) => !thread.isEmpty}>
+        <AuiIf condition={({ thread }) => hideWelcome || !thread.isEmpty}>
           <div className="aui-thread-composer-dock pointer-events-none absolute bottom-0 left-0 right-0 md:right-2 z-20">
             <div
               aria-hidden
@@ -130,7 +146,7 @@ export const Thread: FC<{ hideComposer?: boolean; hideWelcome?: boolean }> = ({
             />
             <div className="relative px-5 pb-2">
               <div className="pointer-events-auto mx-auto w-full max-w-(--thread-max-width)">
-                <ComposerAnimated />
+                <ComposerAnimated disabled={isComposerAttachPending} />
               </div>
               <p className="mt-1.5 text-center text-[11px] text-muted-foreground">
                 LLMs can make mistakes. Double-check all responses.
@@ -268,7 +284,7 @@ const GeneratingSpinner: FC = () => {
   );
 };
 
-const ComposerAnimated: FC = () => {
+const ComposerAnimated: FC<{ disabled?: boolean }> = ({ disabled }) => {
   return (
     <div className="relative mx-auto min-w-0 w-full max-w-(--thread-max-width)">
       <motion.div
@@ -277,7 +293,7 @@ const ComposerAnimated: FC = () => {
         transition={{ type: "spring", bounce: 0.15, duration: 0.5 }}
         className="relative z-10 w-full"
       >
-        <Composer />
+        <Composer disabled={disabled} />
       </motion.div>
     </div>
   );
@@ -305,9 +321,22 @@ const PendingAudioChip: FC = () => {
   );
 };
 
-const Composer: FC = () => {
+const Composer: FC<{ disabled?: boolean }> = ({ disabled }) => {
+  const handleSubmit = useCallback(
+    (event: FormEvent<HTMLFormElement>) => {
+      if (disabled) {
+        event.preventDefault();
+      }
+    },
+    [disabled],
+  );
+
   return (
-    <ComposerPrimitive.Root className="aui-composer-root relative flex w-full flex-col">
+    <ComposerPrimitive.Root
+      className="aui-composer-root relative flex w-full flex-col"
+      aria-disabled={disabled}
+      onSubmit={handleSubmit}
+    >
       <ComposerPrimitive.AttachmentDropzone className="aui-composer-attachment-dropzone chat-composer-surface flex w-full flex-col rounded-3xl bg-background dark:bg-card px-1 pt-2 outline-none transition-shadow data-[dragging=true]:border-ring data-[dragging=true]:bg-accent/50">
         <ComposerAttachments />
         <PendingAudioChip />
@@ -317,10 +346,11 @@ const Composer: FC = () => {
           className="aui-composer-input mb-1 min-h-12 w-full resize-none overflow-y-auto bg-transparent pl-5 pr-4 pt-2 pb-3 text-sm font-[450] outline-none placeholder:text-muted-foreground focus-visible:ring-0"
           minRows={1}
           maxRows={6}
-          autoFocus={true}
+          autoFocus={!disabled}
+          disabled={disabled}
           aria-label="Message input"
         />
-        <ComposerAction />
+        <ComposerAction disabled={disabled} />
       </ComposerPrimitive.AttachmentDropzone>
     </ComposerPrimitive.Root>
   );
@@ -537,7 +567,7 @@ const ToolStatusDisplay: FC = () => {
   );
 };
 
-const ComposerAction: FC = () => {
+const ComposerAction: FC<{ disabled?: boolean }> = ({ disabled }) => {
   return (
     <div className="aui-composer-action-wrapper relative mx-2 mb-2 flex items-center justify-between">
       <div className="flex items-center gap-1">
@@ -578,6 +608,7 @@ const ComposerAction: FC = () => {
               type="submit"
               variant="default"
               size="icon"
+              disabled={disabled}
               className="aui-composer-send size-8 rounded-full"
               aria-label="Send message"
             >
diff --git a/studio/frontend/src/features/chat/chat-page.tsx b/studio/frontend/src/features/chat/chat-page.tsx
index 8510aece50..32e34b3c72 100644
--- a/studio/frontend/src/features/chat/chat-page.tsx
+++ b/studio/frontend/src/features/chat/chat-page.tsx
@@ -122,7 +122,7 @@ const SingleContent = memo(function SingleContent({
       newThreadNonce={newThreadNonce}
     >
       <div className="flex min-h-0 min-w-0 flex-1 basis-0 flex-col overflow-hidden">
-        <Thread />
+        <Thread hideWelcome={Boolean(threadId)} targetThreadId={threadId} />
       </div>
     </ChatRuntimeProvider>
   );
@@ -492,6 +492,37 @@ export function ChatPage(): ReactElement {
   useEffect(() => {
     return () => setSettingsOpen(false);
   }, [setSettingsOpen]);
+
+  useEffect(() => {
+    const threadId = search.thread;
+    if (!threadId) return;
+
+    let canceled = false;
+    void db.threads
+      .get(threadId)
+      .then((thread) => {
+        if (canceled || thread) return;
+        useChatRuntimeStore.getState().setActiveThreadId(null);
+        toast.info("Chat not found", {
+          description: "That thread no longer exists, so we opened a new chat.",
+        });
+        navigate({
+          to: "/chat",
+          search: { new: crypto.randomUUID() },
+          replace: true,
+        });
+      })
+      .catch(() => {
+        if (useChatRuntimeStore.getState().activeThreadId === threadId) {
+          useChatRuntimeStore.getState().setActiveThreadId(null);
+        }
+      });
+
+    return () => {
+      canceled = true;
+    };
+  }, [navigate, search.thread]);
+
   const [modelSelectorOpen, setModelSelectorOpen] = useState(false);
   const [modelSelectorLocked, setModelSelectorLocked] = useState(false);
   const viewBeforeCompareRef = useRef<ChatSearch | null>(null);
diff --git a/studio/frontend/src/features/chat/runtime-provider.tsx b/studio/frontend/src/features/chat/runtime-provider.tsx
index 5747e17970..105f2fb4b1 100644
--- a/studio/frontend/src/features/chat/runtime-provider.tsx
+++ b/studio/frontend/src/features/chat/runtime-provider.tsx
@@ -681,9 +681,16 @@ function ThreadAutoSwitch({
 
   useEffect(() => {
     if (!isLoading && mainThreadId !== threadId) {
-      aui.threads().switchToThread(threadId);
+      const switchResult = aui.threads().switchToThread(threadId) as unknown;
+      if (switchResult && typeof (switchResult as Promise<void>).catch === "function") {
+        void (switchResult as Promise<void>).catch(() => {
+          if (syncActiveThreadId) {
+            useChatRuntimeStore.getState().setActiveThreadId(null);
+          }
+        });
+      }
     }
-  }, [aui, isLoading, mainThreadId, threadId]);
+  }, [aui, isLoading, mainThreadId, syncActiveThreadId, threadId]);
 
   useEffect(() => {
     if (!syncActiveThreadId || isLoading || mainThreadId !== threadId) {

From 92d43a4f686c0f3a809b157b907e42d6272accf6 Mon Sep 17 00:00:00 2001
From: Lee Jackson <130007945+Imagineer99@users.noreply.github.com>
Date: Wed, 22 Apr 2026 15:57:19 +0100
Subject: [PATCH 148/216] Studio: Replace assistant UI shared autoscroll with
 per-panel scrolling (#5127)

* Replace assistant-ui's shared-store autoscroll with independent per-pane state

- assistant-ui's useThreadViewportAutoScroll uses shared state that causes random scrolling issues on the compare page
- Replaced with independent per-pane autoscroll implementation

* Fix scroll-click scrolling for slow movements; polish sidebar animations and UI

- Fix middle-click scroll not responding to very slow movements
- Add animations to sidebar navigation icons; font adjustments
- Align app menu styling for consistency
- Fix rare lag on sidebar expand/collapse
- Smooth right sidebar (parameter tuning) animations and expand/collapse lag

* Address PR feedback: clean up unused code and polish sidebar

- Remove unused _SuggestionItem component and its dependencies (SuggestionPrimitive import, SUGGESTION_TOOLS, toolIconMap)
- Log delete-message errors to console before showing toast for better observability
- Further sidebar menu design adjustments

* Address PR feedback: auto-follow guard and NavItem cleanup

- Prevent auto-follow from breaking on short threads: gate wheel/touch detach on scrollTop > 0 so no-op upward gestures on non-scrollable viewports don't flip userDetachedRef
- Simplify NavItem: collapse identical-branch ternary and drop unused variant prop

* Address PR feedback: model selector consistency and nested scroll fix

- Adjust model selector to be consistent with the sidebar
- Fix auto-follow breaking when user scrolls inside nested scrollable regions (reasoning/tool panels): gate wheel/touch detach on innerScrollWillConsumeUpward(target) so bubbled events from inner scrollers don't flip viewport intent (Codex review)

* Filter layout-induced upward scroll deltas from auto-follow detach

Accumulator now counts distanceFromBottom growth instead of raw -scrollTop delta, so browser scroll-anchoring on auto-collapsing panels (reasoning, tool outputs) no longer falsely detaches auto-follow mid-stream.

---------

Co-authored-by: sneakr <hauzin@hotmail.com>
---
 .../frontend/src/components/app-sidebar.tsx   |  68 ++-
 .../src/components/assistant-ui/thread.tsx    | 307 ++++++------
 .../use-intent-aware-autoscroll.tsx           | 451 ++++++++++++++++++
 studio/frontend/src/components/ui/sidebar.tsx |   4 +-
 .../frontend/src/features/chat/chat-page.tsx  |  31 +-
 .../src/features/chat/chat-settings-sheet.tsx |   2 +-
 studio/frontend/src/index.css                 |  30 +-
 7 files changed, 670 insertions(+), 223 deletions(-)
 create mode 100644 studio/frontend/src/components/assistant-ui/use-intent-aware-autoscroll.tsx

diff --git a/studio/frontend/src/components/app-sidebar.tsx b/studio/frontend/src/components/app-sidebar.tsx
index 959c2a6b31..6d029a59bd 100644
--- a/studio/frontend/src/components/app-sidebar.tsx
+++ b/studio/frontend/src/components/app-sidebar.tsx
@@ -123,7 +123,6 @@ function NavItem({
   disabled,
   onClick,
   children,
-  variant = "nav",
   dataTour,
 }: {
   icon: typeof ZapIcon;
@@ -132,10 +131,8 @@ function NavItem({
   disabled?: boolean;
   onClick: () => void;
   children?: React.ReactNode;
-  variant?: "nav" | "menu";
   dataTour?: string;
 }) {
-  const isNav = variant === "nav";
   return (
     <SidebarMenuItem>
       <div className="relative">
@@ -145,14 +142,10 @@ function NavItem({
           onClick={onClick}
           isActive={active}
           data-tour={dataTour}
-          className={
-            isNav
-              ? "h-[30px] rounded-[8px] gap-2.5 px-2.5 font-medium text-[#383835] dark:text-[#c7c7c4] hover:bg-[#ececec]! dark:hover:bg-[#2e3035]! hover:text-black! dark:hover:text-white! data-active:bg-[#ececec]! dark:data-active:bg-[#2e3035]! data-active:text-black! dark:data-active:text-white! group-data-[collapsible=icon]:!w-[30px] group-data-[collapsible=icon]:!rounded-[9px] group-data-[collapsible=icon]:mx-auto"
-              : "h-[30px] rounded-[8px] gap-2.5 px-2.5 font-medium text-[#383835] dark:text-[#c7c7c4] hover:bg-[#ececec]! dark:hover:bg-[#2e3035]! hover:text-black! dark:hover:text-white! data-active:bg-[#ececec]! dark:data-active:bg-[#2e3035]! data-active:text-black! dark:data-active:text-white! group-data-[collapsible=icon]:!w-[30px] group-data-[collapsible=icon]:!rounded-[9px] group-data-[collapsible=icon]:mx-auto"
-          }
+          className="h-[32px] rounded-[10px] gap-[8.5px] px-2.5 font-medium text-[#383835] dark:text-[#c7c7c4] hover:bg-[#f0f0f0]! dark:hover:bg-[#2a2c2f]! hover:text-black! dark:hover:text-white! data-active:bg-[#f0f0f0]! dark:data-active:bg-[#2a2c2f]! data-active:text-black! dark:data-active:text-white! group-data-[collapsible=icon]:!w-[32px] group-data-[collapsible=icon]:!rounded-[11px] group-data-[collapsible=icon]:mx-auto"
         >
-          <HugeiconsIcon icon={icon} strokeWidth={1.5} className="size-[18px]!" />
-          <span className="text-sm">{label}</span>
+          <HugeiconsIcon icon={icon} strokeWidth={1.75} className="size-[18px]! shrink-0 group-hover/menu-button:animate-icon-pop" />
+          <span className="text-[14px] leading-[18px] tracking-[0.01em]">{label}</span>
         </SidebarMenuButton>
       </div>
       {children}
@@ -230,7 +223,7 @@ export function AppSidebar() {
     >
       <SidebarHeader className="pl-[17px] pr-3 pt-[12px] pb-[12px] group-data-[collapsible=icon]:px-0">
         {/* Expanded: compact logo + close toggle */}
-        <div className="flex items-center justify-between gap-2 group-data-[collapsible=icon]:hidden">
+        <div className="flex items-center justify-between gap-[8.5px] group-data-[collapsible=icon]:hidden">
           <Link
             to="/chat"
             onClick={(event) => {
@@ -267,7 +260,7 @@ export function AppSidebar() {
                 <button
                   type="button"
                   onClick={togglePinned}
-                  className="inline-flex h-7 w-7 items-center justify-center rounded-[8px] text-[#8f8f8f] dark:text-[#5c5c5c] transition-colors hover:bg-[#ececec] dark:hover:bg-[#2e3035] hover:text-black dark:hover:text-white focus-visible:outline-none focus-visible:ring-2 focus-visible:ring-ring"
+                  className="inline-flex h-7 w-7 items-center justify-center rounded-[10px] text-[#8f8f8f] dark:text-[#5c5c5c] transition-colors hover:bg-[#f0f0f0] dark:hover:bg-[#2a2c2f] hover:text-black dark:hover:text-white focus-visible:outline-none focus-visible:ring-2 focus-visible:ring-ring"
                   aria-label="Close sidebar"
                 >
                   <HugeiconsIcon icon={LayoutAlignLeftIcon} strokeWidth={1.75} className="size-[18px]" />
@@ -288,7 +281,7 @@ export function AppSidebar() {
                 <button
                   type="button"
                   onClick={togglePinned}
-                  className="inline-flex h-7 w-7 items-center justify-center rounded-[8px] text-[#383835] dark:text-[#c7c7c4] transition-colors hover:bg-[#ececec] dark:hover:bg-[#2e3035] hover:text-black dark:hover:text-white focus-visible:outline-none focus-visible:ring-2 focus-visible:ring-ring"
+                  className="inline-flex h-7 w-7 items-center justify-center rounded-[10px] text-[#383835] dark:text-[#c7c7c4] transition-colors hover:bg-[#f0f0f0] dark:hover:bg-[#2a2c2f] hover:text-black dark:hover:text-white focus-visible:outline-none focus-visible:ring-2 focus-visible:ring-ring"
                   aria-label="Open sidebar"
                 >
                   <HugeiconsIcon icon={LayoutAlignLeftIcon} strokeWidth={1.75} className="size-[18px]" />
@@ -302,7 +295,7 @@ export function AppSidebar() {
         )}
       </SidebarHeader>
 
-      <SidebarGroup className="group-data-[collapsible=icon]:px-0 px-2 pt-[8px] pb-[12px] shrink-0">
+      <SidebarGroup className="group-data-[collapsible=icon]:px-0 px-2 pt-[10px] pb-[14px] shrink-0">
         <SidebarGroupContent>
           <SidebarMenu>
             <NavItem
@@ -347,7 +340,7 @@ export function AppSidebar() {
 
       <SidebarContent className="gap-0 overflow-y-auto overscroll-contain min-h-0">
         {/* Navigate (no header) */}
-        <SidebarGroup data-tour="navbar" className="group-data-[collapsible=icon]:px-0 px-2 pt-[8px] pb-[12px]">
+        <SidebarGroup data-tour="navbar" className="group-data-[collapsible=icon]:px-0 px-2 pt-[10px] pb-[14px]">
           <SidebarGroupContent>
             <SidebarMenu>
               <NavItem
@@ -404,7 +397,7 @@ export function AppSidebar() {
                   <SidebarMenuItem key={item.id} className="group/recent-item relative">
                     <SidebarMenuButton
                       isActive={activeThreadId === item.id}
-                      className="h-[30px] rounded-[8px] pl-2.5 pr-7 text-sm font-medium text-[#383835] dark:text-[#c7c7c4] hover:bg-[#ececec]! dark:hover:bg-[#2e3035]! hover:text-black! dark:hover:text-white! data-active:bg-[#ececec]! dark:data-active:bg-[#2e3035]! data-active:text-black! dark:data-active:text-white!"
+                      className="h-[32px] rounded-[10px] pl-2.5 pr-7 text-[14px] leading-[18px] tracking-[0.01em] font-medium text-[#383835] dark:text-[#c7c7c4] hover:bg-[#f0f0f0]! dark:hover:bg-[#2a2c2f]! hover:text-black! dark:hover:text-white! data-active:bg-[#f0f0f0]! dark:data-active:bg-[#2a2c2f]! data-active:text-black! dark:data-active:text-white!"
                       onClick={() => {
                         navigate({
                           to: "/chat",
@@ -425,7 +418,7 @@ export function AppSidebar() {
                         handleDeleteThread(item);
                       }}
                       title="Delete"
-                      className="absolute right-1 top-1/2 -translate-y-1/2 flex size-5 scale-90 items-center justify-center rounded-[8px] text-sidebar-foreground/55 opacity-0 transition-all duration-150 hover:bg-destructive/12 hover:text-destructive group-hover/recent-item:scale-100 group-hover/recent-item:opacity-100"
+                      className="absolute right-1 top-1/2 -translate-y-1/2 flex size-5 scale-90 items-center justify-center rounded-[10px] text-sidebar-foreground/55 opacity-0 transition-all duration-150 hover:bg-destructive/12 hover:text-destructive group-hover/recent-item:scale-100 group-hover/recent-item:opacity-100"
                     >
                       <HugeiconsIcon icon={Delete02Icon} strokeWidth={2} className="size-3.5" />
                     </button>
@@ -461,13 +454,13 @@ export function AppSidebar() {
                       >
                         <SidebarMenuButton
                           isActive={isActiveRun}
-                          className="h-auto flex-col items-start gap-0.5 py-1.5 rounded-[8px] pl-2.5 pr-7 text-sm font-medium text-[#383835] dark:text-[#c7c7c4] hover:bg-[#ececec]! dark:hover:bg-[#2e3035]! hover:text-black! dark:hover:text-white! data-active:bg-[#ececec]! dark:data-active:bg-[#2e3035]! data-active:text-black! dark:data-active:text-white!"
+                          className="h-auto flex-col items-start gap-0.5 py-1.5 rounded-[10px] pl-2.5 pr-7 text-[14px] tracking-[0.01em] font-medium text-[#383835] dark:text-[#c7c7c4] hover:bg-[#f0f0f0]! dark:hover:bg-[#2a2c2f]! hover:text-black! dark:hover:text-white! data-active:bg-[#f0f0f0]! dark:data-active:bg-[#2a2c2f]! data-active:text-black! dark:data-active:text-white!"
                           onClick={() => {
                             setSelectedHistoryRunId(run.id);
                             closeMobileIfOpen();
                           }}
                         >
-                          <div className="flex w-full items-center gap-2">
+                          <div className="flex w-full items-center gap-[8.5px]">
                             <span
                               className={cn(
                                 "size-1.5 shrink-0 rounded-full",
@@ -475,7 +468,7 @@ export function AppSidebar() {
                               )}
                               aria-hidden
                             />
-                            <span className="truncate text-sm">
+                            <span className="truncate">
                               {run.model_name}
                             </span>
                             <span className="ml-auto shrink-0 text-[10px] text-muted-foreground">
@@ -501,7 +494,7 @@ export function AppSidebar() {
                             }
                           }}
                           title="Delete"
-                          className="absolute right-1 top-1/2 -translate-y-1/2 flex size-5 scale-90 items-center justify-center rounded-[8px] text-sidebar-foreground/55 opacity-0 transition-all duration-150 hover:bg-destructive/12 hover:text-destructive group-hover/run-item:scale-100 group-hover/run-item:opacity-100"
+                          className="absolute right-1 top-1/2 -translate-y-1/2 flex size-5 scale-90 items-center justify-center rounded-[10px] text-sidebar-foreground/55 opacity-0 transition-all duration-150 hover:bg-destructive/12 hover:text-destructive group-hover/run-item:scale-100 group-hover/run-item:opacity-100"
                         >
                           <HugeiconsIcon icon={Delete02Icon} strokeWidth={2} className="size-3.5" />
                         </button>
@@ -516,7 +509,7 @@ export function AppSidebar() {
         )}
       </SidebarContent>
 
-      <SidebarFooter className="border-t border-sidebar-border group-data-[collapsible=icon]:border-t-0 group-data-[collapsible=icon]:px-0">
+      <SidebarFooter className="border-t border-sidebar-border group-data-[collapsible=icon]:border-transparent group-data-[collapsible=icon]:px-0">
         <SidebarMenu>
           <SidebarMenuItem>
             <DropdownMenu>
@@ -524,7 +517,7 @@ export function AppSidebar() {
                 <SidebarMenuButton
                   size="lg"
                   aria-label={`${displayTitle} account menu`}
-                  className="!h-[50px] gap-[8px] rounded-[8px] text-[#383835] dark:text-[#c7c7c4] hover:bg-[#ececec]! dark:hover:bg-[#2e3035]! hover:text-black! dark:hover:text-white! data-[state=open]:bg-[#ececec]! dark:data-[state=open]:bg-[#2e3035]! data-[state=open]:text-black! dark:data-[state=open]:text-white!"
+                  className="!h-[50px] gap-[8px] rounded-[10px] text-[#383835] dark:text-[#c7c7c4] hover:bg-[#f0f0f0]! dark:hover:bg-[#2a2c2f]! hover:text-black! dark:hover:text-white! data-[state=open]:bg-[#f0f0f0]! dark:data-[state=open]:bg-[#2a2c2f]! data-[state=open]:text-black! dark:data-[state=open]:text-white!"
                 >
                   <div className="shrink-0">
                     <UserAvatar
@@ -535,8 +528,8 @@ export function AppSidebar() {
                     />
                   </div>
                   <div className="flex flex-col gap-0.5 leading-none group-data-[collapsible=icon]:hidden">
-                    <span className="truncate font-heading text-[13px] font-semibold text-[#383835] dark:text-[#c7c7c4]">{displayTitle}</span>
-                    <span className="truncate text-[11px] text-muted-foreground">Studio</span>
+                    <span className="truncate font-heading text-[13px] tracking-[0.02em] font-semibold text-[#383835] dark:text-[#c7c7c4]">{displayTitle}</span>
+                    <span className="truncate text-[11px] tracking-[0.01em] text-muted-foreground">Studio</span>
                   </div>
                   <ChevronsUpDown strokeWidth={1.25} className="ml-auto size-4 text-muted-foreground group-data-[collapsible=icon]:hidden" />
                 </SidebarMenuButton>
@@ -544,24 +537,21 @@ export function AppSidebar() {
               <DropdownMenuContent
                 side="top"
                 align="start"
-                className="w-[15rem] font-heading [&_[data-slot=dropdown-menu-item]]:rounded-[8px] [&_[data-slot=dropdown-menu-item]]:font-medium [&_[data-slot=dropdown-menu-item]]:text-[#383835] dark:[&_[data-slot=dropdown-menu-item]]:text-[#c7c7c4] [&_[data-slot=dropdown-menu-item]:focus]:bg-[#ececec] dark:[&_[data-slot=dropdown-menu-item]:focus]:bg-[#2e3035] [&_[data-slot=dropdown-menu-item]:focus]:text-black dark:[&_[data-slot=dropdown-menu-item]:focus]:text-white [&_[data-slot=dropdown-menu-item]:focus_*]:text-black! dark:[&_[data-slot=dropdown-menu-item]:focus_*]:text-white!"
+                className="w-[15rem] py-2.5 font-heading [&_[data-slot=dropdown-menu-group]]:flex [&_[data-slot=dropdown-menu-group]]:flex-col [&_[data-slot=dropdown-menu-group]]:gap-px [&_[data-slot=dropdown-menu-item]]:h-[32px] [&_[data-slot=dropdown-menu-item]]:px-2.5! [&_[data-slot=dropdown-menu-item]]:py-0! [&_[data-slot=dropdown-menu-item]]:gap-[8.5px]! [&_[data-slot=dropdown-menu-item]]:rounded-[10px] [&_[data-slot=dropdown-menu-item]]:font-medium [&_[data-slot=dropdown-menu-item]]:text-[14px] [&_[data-slot=dropdown-menu-item]]:leading-[18px] [&_[data-slot=dropdown-menu-item]]:tracking-[0.01em] [&_[data-slot=dropdown-menu-item]]:text-[#383835] dark:[&_[data-slot=dropdown-menu-item]]:text-[#c7c7c4] [&_[data-slot=dropdown-menu-item]_svg]:!size-[18px] [&_[data-slot=dropdown-menu-item]_svg]:shrink-0 [&_[data-slot=dropdown-menu-item]:focus]:bg-[#f0f0f0] dark:[&_[data-slot=dropdown-menu-item]:focus]:bg-[#2a2c2f] [&_[data-slot=dropdown-menu-item]:focus]:text-black dark:[&_[data-slot=dropdown-menu-item]:focus]:text-white [&_[data-slot=dropdown-menu-item]:focus_*]:text-black! dark:[&_[data-slot=dropdown-menu-item]:focus_*]:text-white!"
               >
                 <DropdownMenuGroup>
                   <DropdownMenuItem
                     onSelect={() => useSettingsDialogStore.getState().openDialog()}
                   >
-                    <HugeiconsIcon icon={Settings02Icon} className="size-4" />
+                    <HugeiconsIcon icon={Settings02Icon} strokeWidth={1.75} className="size-[18px]" />
                     <span>Settings</span>
                     <DropdownMenuShortcut>⌘,</DropdownMenuShortcut>
                   </DropdownMenuItem>
-                </DropdownMenuGroup>
-                <DropdownMenuSeparator />
-                <DropdownMenuGroup>
                   <DropdownMenuItem
                     ref={anchorRef as React.Ref<HTMLDivElement>}
                     onSelect={(e) => { e.preventDefault(); toggleTheme(); }}
                   >
-                    {isDark ? <Sun className="size-4" /> : <Moon className="size-4" />}
+                    {isDark ? <Sun strokeWidth={1.75} className="size-[18px]" /> : <Moon strokeWidth={1.75} className="size-[18px]" />}
                     <span>{isDark ? "Light Mode" : "Dark Mode"}</span>
                   </DropdownMenuItem>
                   <DropdownMenuItem
@@ -576,11 +566,11 @@ export function AppSidebar() {
                       );
                     }}
                   >
-                    <HugeiconsIcon icon={CursorInfo02Icon} className="size-4" />
+                    <HugeiconsIcon icon={CursorInfo02Icon} strokeWidth={1.75} className="size-[18px]" />
                     <span>Guided Tour</span>
                   </DropdownMenuItem>
                 </DropdownMenuGroup>
-                <DropdownMenuSeparator />
+                <DropdownMenuSeparator className="mx-2.5! my-2.5! h-0! border-t border-border/70 bg-transparent!" />
                 <DropdownMenuGroup>
                   <DropdownMenuItem asChild>
                     <a
@@ -588,7 +578,7 @@ export function AppSidebar() {
                       target="_blank"
                       rel="noopener noreferrer"
                     >
-                      <HugeiconsIcon icon={Book03Icon} className="size-4" />
+                      <HugeiconsIcon icon={Book03Icon} strokeWidth={1.75} className="size-[18px]" />
                       <span>Learn More</span>
                     </a>
                   </DropdownMenuItem>
@@ -600,7 +590,8 @@ export function AppSidebar() {
                     >
                       <HugeiconsIcon
                         icon={NewReleasesIcon}
-                        className="size-4"
+                        strokeWidth={1.75}
+                        className="size-[18px]"
                       />
                       <span>What's New</span>
                     </a>
@@ -613,15 +604,16 @@ export function AppSidebar() {
                     >
                       <HugeiconsIcon
                         icon={MessageSearch01Icon}
-                        className="size-4"
+                        strokeWidth={1.75}
+                        className="size-[18px]"
                       />
                       <span>Feedback</span>
                     </a>
                   </DropdownMenuItem>
                 </DropdownMenuGroup>
-                <DropdownMenuSeparator />
+                <DropdownMenuSeparator className="mx-2.5! my-2.5! h-0! border-t border-border/70 bg-transparent!" />
                 <DropdownMenuItem onSelect={() => setShutdownOpen(true)}>
-                  <HugeiconsIcon icon={PowerIcon} className="size-4" />
+                  <HugeiconsIcon icon={PowerIcon} strokeWidth={1.75} className="size-[18px]" />
                   <span>Shutdown</span>
                 </DropdownMenuItem>
               </DropdownMenuContent>
diff --git a/studio/frontend/src/components/assistant-ui/thread.tsx b/studio/frontend/src/components/assistant-ui/thread.tsx
index 4c6569c2ee..0b97d98dfd 100644
--- a/studio/frontend/src/components/assistant-ui/thread.tsx
+++ b/studio/frontend/src/components/assistant-ui/thread.tsx
@@ -6,19 +6,27 @@ import {
   ComposerAttachments,
   UserMessageAttachments,
 } from "@/components/assistant-ui/attachment";
-import { MessageTiming } from "@/components/assistant-ui/message-timing";
+import { CodeToggleIcon } from "@/components/assistant-ui/code-toggle-icon";
 import { MarkdownText } from "@/components/assistant-ui/markdown-text";
+import { MessageTiming } from "@/components/assistant-ui/message-timing";
 import { Reasoning, ReasoningGroup } from "@/components/assistant-ui/reasoning";
 import { Sources, SourcesGroup } from "@/components/assistant-ui/sources";
 import { ToolFallback } from "@/components/assistant-ui/tool-fallback";
 import { ToolGroup } from "@/components/assistant-ui/tool-group";
-import { WebSearchToolUI } from "@/components/assistant-ui/tool-ui-web-search";
 import { PythonToolUI } from "@/components/assistant-ui/tool-ui-python";
 import { TerminalToolUI } from "@/components/assistant-ui/tool-ui-terminal";
+import { WebSearchToolUI } from "@/components/assistant-ui/tool-ui-web-search";
 import { TooltipIconButton } from "@/components/assistant-ui/tooltip-icon-button";
-import { CodeToggleIcon } from "@/components/assistant-ui/code-toggle-icon";
+import {
+  IntentAwareScrollProvider,
+  useIntentAwareAutoScroll,
+  useIsThreadAtBottom,
+  useScrollThreadToBottom,
+} from "@/components/assistant-ui/use-intent-aware-autoscroll";
 import { Button } from "@/components/ui/button";
 import { sentAudioNames } from "@/features/chat/api/chat-adapter";
+import { useChatRuntimeStore } from "@/features/chat/stores/chat-runtime-store";
+import { deleteThreadMessage } from "@/features/chat/utils/delete-thread-message";
 import { AUDIO_ACCEPT, MAX_AUDIO_SIZE, fileToBase64 } from "@/lib/audio-utils";
 import { copyToClipboard } from "@/lib/copy-to-clipboard";
 import { cn } from "@/lib/utils";
@@ -30,14 +38,11 @@ import {
   ComposerPrimitive,
   ErrorPrimitive,
   MessagePrimitive,
-  SuggestionPrimitive,
   ThreadPrimitive,
   useAui,
   useAuiEvent,
   useAuiState,
-  useThreadViewport,
 } from "@assistant-ui/react";
-import { motion } from "motion/react";
 import {
   ArrowDownIcon,
   ArrowUpIcon,
@@ -50,9 +55,9 @@ import {
   HeadphonesIcon,
   LightbulbIcon,
   LightbulbOffIcon,
+  LoaderIcon,
   MicIcon,
   MoreHorizontalIcon,
-  LoaderIcon,
   PencilIcon,
   RefreshCwIcon,
   SquareIcon,
@@ -60,6 +65,7 @@ import {
   Trash2Icon,
   XIcon,
 } from "lucide-react";
+import { motion } from "motion/react";
 import {
   type FC,
   type FormEvent,
@@ -69,8 +75,6 @@ import {
   useState,
 } from "react";
 import { toast } from "sonner";
-import { deleteThreadMessage } from "@/features/chat/utils/delete-thread-message";
-import { useChatRuntimeStore } from "@/features/chat/stores/chat-runtime-store";
 
 export const Thread: FC<{
   hideComposer?: boolean;
@@ -81,6 +85,13 @@ export const Thread: FC<{
   hideWelcome,
   targetThreadId,
 }) => {
+  // Intent-aware autoscroll: replaces assistant-ui's built-in autoscroll
+  // to prevent the streaming-mutation race that makes the viewport snap
+  // back to the bottom while the user is scrolling up (see the hook for
+  // the full explanation).
+  const { ref: viewportRef, context: autoScrollContext } =
+    useIntentAwareAutoScroll();
+
   const isComposerAttachPending = useAuiState(({ threads }) =>
     targetThreadId ? threads.mainThreadId !== targetThreadId : false,
   );
@@ -94,154 +105,99 @@ export const Thread: FC<{
           "calc(var(--thread-max-width) - 2.5rem)",
       }}
     >
-      <ThreadPrimitive.Viewport
-        className={cn(
-          "aui-thread-viewport relative flex min-h-0 min-w-0 flex-1 basis-0 flex-col overflow-x-auto overflow-y-auto scroll-smooth px-5",
-          hideComposer ? "pt-4" : "pt-[48px]",
-        )}
-      >
-        {!hideWelcome && (
-          <AuiIf condition={({ thread }) => thread.isEmpty && !thread.isLoading}>
-            <ThreadWelcome hideComposer={hideComposer} />
-          </AuiIf>
-        )}
+      <IntentAwareScrollProvider value={autoScrollContext}>
+        <ThreadPrimitive.Viewport
+          ref={viewportRef}
+          autoScroll={false}
+          scrollToBottomOnRunStart={false}
+          scrollToBottomOnInitialize={false}
+          scrollToBottomOnThreadSwitch={false}
+          className={cn(
+            "aui-thread-viewport relative flex min-h-0 min-w-0 flex-1 basis-0 flex-col overflow-x-auto overflow-y-auto scroll-smooth px-5",
+            hideComposer ? "pt-4" : "pt-[48px]",
+          )}
+        >
+          {!hideWelcome && (
+            <AuiIf condition={({ thread }) => thread.isEmpty && !thread.isLoading}>
+              <ThreadWelcome hideComposer={hideComposer} />
+            </AuiIf>
+          )}
 
-        <ThreadPrimitive.Messages
-          components={{
-            UserMessage,
-            EditComposer,
-            AssistantMessage,
-          }}
-        />
+          <ThreadPrimitive.Messages
+            components={{
+              UserMessage,
+              EditComposer,
+              AssistantMessage,
+            }}
+          />
 
-        {/* Bottom slack so the last message has breathing room above the
+          {/* Bottom slack so the last message has breathing room above the
             sticky scroll-to-bottom button (and the floating composer in
             single mode). Without this, content would butt against the
             sticky footer and feel cramped. */}
-        <AuiIf condition={({ thread }) => hideWelcome || !thread.isEmpty}>
-          <div
-            className={cn("shrink-0", hideComposer ? "h-16" : "h-40")}
-            aria-hidden
-          />
-        </AuiIf>
-
-        <AuiIf condition={({ thread }) => hideWelcome || !thread.isEmpty}>
-          <ThreadPrimitive.ViewportFooter
-            className={cn(
-              "aui-thread-viewport-footer pointer-events-none sticky z-20 flex w-full justify-center bg-transparent",
-              hideComposer ? "bottom-3" : "bottom-[140px]",
-            )}
-          >
-            <ThreadScrollToBottom />
-          </ThreadPrimitive.ViewportFooter>
-        </AuiIf>
-      </ThreadPrimitive.Viewport>
-
-      {!hideComposer && (
-        <AuiIf condition={({ thread }) => hideWelcome || !thread.isEmpty}>
-          <div className="aui-thread-composer-dock pointer-events-none absolute bottom-0 left-0 right-0 md:right-2 z-20">
+          <AuiIf condition={({ thread }) => hideWelcome || !thread.isEmpty}>
             <div
-              aria-hidden
-              className="absolute inset-x-0 bottom-0 top-[10px] bg-background"
+              className={cn("shrink-0", hideComposer ? "h-16" : "h-40")}
+              aria-hidden={true}
             />
-            <div className="relative px-5 pb-2">
-              <div className="pointer-events-auto mx-auto w-full max-w-(--thread-max-width)">
-                <ComposerAnimated disabled={isComposerAttachPending} />
+          </AuiIf>
+
+          <AuiIf condition={({ thread }) => hideWelcome || !thread.isEmpty}>
+            <ThreadPrimitive.ViewportFooter
+              className={cn(
+                "aui-thread-viewport-footer pointer-events-none sticky z-20 flex w-full justify-center bg-transparent",
+                hideComposer ? "bottom-3" : "bottom-[140px]",
+              )}
+            >
+              <ThreadScrollToBottom />
+            </ThreadPrimitive.ViewportFooter>
+          </AuiIf>
+        </ThreadPrimitive.Viewport>
+
+        {!hideComposer && (
+          <AuiIf condition={({ thread }) => hideWelcome || !thread.isEmpty}>
+            <div className="aui-thread-composer-dock pointer-events-none absolute bottom-0 left-0 right-0 md:right-2 z-20">
+              <div
+                aria-hidden={true}
+                className="absolute inset-x-0 bottom-0 top-[10px] bg-background"
+              />
+              <div className="relative px-5 pb-2">
+                <div className="pointer-events-auto mx-auto w-full max-w-(--thread-max-width)">
+                  <ComposerAnimated disabled={isComposerAttachPending} />
+                </div>
+                <p className="mt-1.5 text-center text-[11px] text-muted-foreground">
+                  LLMs can make mistakes. Double-check all responses.
+                </p>
               </div>
-              <p className="mt-1.5 text-center text-[11px] text-muted-foreground">
-                LLMs can make mistakes. Double-check all responses.
-              </p>
             </div>
-          </div>
-        </AuiIf>
-      )}
+          </AuiIf>
+        )}
+      </IntentAwareScrollProvider>
     </ThreadPrimitive.Root>
   );
 };
 
 const ThreadScrollToBottom: FC = () => {
-  // Scoped to the nearest ThreadPrimitive.Root via context, so in compare
-  // mode each pane reads its own viewport state.
-  //
-  // The button stays mounted and toggles visibility via CSS. Conditionally
-  // rendering (return null) unmounts a DOM node inside the viewport, which
-  // the assistant-ui autoscroll hook's MutationObserver sees as a content
-  // change — during streaming that triggered spurious scroll-to-bottom
-  // calls, especially in the narrower mobile stacked layout.
-  const isAtBottom = useThreadViewport((vp) => vp.isAtBottom);
+  // State and action both come from our IntentAwareScrollProvider (scoped
+  // per Thread, so compare panes are independent). We deliberately
+  // avoid `ThreadPrimitive.ScrollToBottom` + `useThreadViewport` to
+  // stay off assistant-ui's internal autoscroll path — see the hook
+  // for why. The button stays mounted and toggles via CSS; unmounting
+  // would trip the hook's MutationObserver as a content change.
+  const isAtBottom = useIsThreadAtBottom();
+  const scrollToBottom = useScrollThreadToBottom();
   return (
-    <ThreadPrimitive.ScrollToBottom asChild={true}>
-      <TooltipIconButton
-        tooltip="Scroll to bottom"
-        variant="outline"
-        className={cn(
-          "aui-thread-scroll-to-bottom pointer-events-auto rounded-full p-4 bg-background hover:bg-accent dark:bg-background dark:hover:bg-accent",
-          isAtBottom && "invisible pointer-events-none",
-        )}
-      >
-        <ArrowDownIcon />
-      </TooltipIconButton>
-    </ThreadPrimitive.ScrollToBottom>
-  );
-};
-
-const SUGGESTION_TOOLS: Record<string, Array<"thinking" | "search" | "code">> = {
-  "How do you fine-tune an audio model with Unsloth?": ["thinking", "search"],
-  "Create a live weather dashboard in HTML using no API key. Show me the code": ["thinking", "code", "search"],
-  "Solve the integral of x·sin(x), and verify it step by step": ["thinking", "code"],
-  "Draw an SVG of a cute sloth & show the code": ["thinking", "code", "search"],
-};
-
-const toolIconMap = {
-  thinking: { icon: LightbulbIcon, label: "Thinking" },
-  search: { icon: GlobeIcon, label: "Web search" },
-  code: { icon: TerminalIcon, label: "Code" },
-} as const;
-
-const SuggestionItem: FC = () => {
-  const aui = useAui();
-  const prompt = useAuiState(({ suggestion }) => suggestion.prompt);
-  const isDisabled = useAuiState(({ thread }) => thread.isDisabled);
-  const isRunning = useAuiState(({ thread }) => thread.isRunning);
-  const tools = SUGGESTION_TOOLS[prompt] ?? [];
-
-  return (
-    <button
-      type="button"
-      onClick={() => {
-        if (!isDisabled && !isRunning) {
-          const store = useChatRuntimeStore.getState();
-          if (store.supportsReasoning) {
-            store.setReasoningEnabled(tools.includes("thinking"));
-          }
-          if (store.supportsTools) {
-            store.setToolsEnabled(tools.includes("search"));
-            store.setCodeToolsEnabled(tools.includes("code"));
-          }
-          aui.thread().append(prompt);
-          aui.composer().setText("");
-          return;
-        }
-        aui.composer().setText(prompt);
-      }}
-      className="fade-in slide-in-from-bottom-1 animate-in relative cursor-pointer corner-squircle rounded-xl border bg-background px-4 py-2.5 pr-12 text-left text-sm text-foreground shadow-sm transition-colors duration-150 hover:bg-accent"
-    >
-      <SuggestionPrimitive.Title />
-      {tools.length > 0 && (
-        <div className="absolute bottom-2.5 right-3 flex items-center gap-1">
-          {tools.map((tool) => {
-            const { icon: Icon, label } = toolIconMap[tool];
-            return (
-              <Icon
-                key={tool}
-                className="size-3 text-muted-foreground/60"
-                aria-label={label}
-              />
-            );
-          })}
-        </div>
+    <TooltipIconButton
+      tooltip="Scroll to bottom"
+      variant="outline"
+      onClick={() => scrollToBottom("auto")}
+      className={cn(
+        "aui-thread-scroll-to-bottom pointer-events-auto rounded-full p-4 bg-background hover:bg-accent dark:bg-background dark:hover:bg-accent",
+        isAtBottom && "invisible pointer-events-none",
       )}
-    </button>
+    >
+      <ArrowDownIcon />
+    </TooltipIconButton>
   );
 };
 
@@ -273,7 +229,9 @@ const ThreadWelcome: FC<{ hideComposer?: boolean }> = ({ hideComposer }) => {
 
 const GeneratingSpinner: FC = () => {
   const status = useChatRuntimeStore((s) => s.generatingStatus);
-  if (!status) return null;
+  if (!status) {
+    return null;
+  }
   return (
     <div className="mx-auto flex w-full max-w-(--thread-max-width) items-center justify-center py-2">
       <div className="flex items-center gap-2 text-xs text-muted-foreground">
@@ -302,7 +260,9 @@ const ComposerAnimated: FC<{ disabled?: boolean }> = ({ disabled }) => {
 const PendingAudioChip: FC = () => {
   const audioName = useChatRuntimeStore((s) => s.pendingAudioName);
   const clearPendingAudio = useChatRuntimeStore((s) => s.clearPendingAudio);
-  if (!audioName) return null;
+  if (!audioName) {
+    return null;
+  }
   return (
     <div className="mb-2 flex w-full flex-row items-center gap-2 px-1.5 pt-0.5 pb-1">
       <div className="flex items-center gap-2 rounded-lg border border-foreground/20 bg-muted px-3 py-1.5 text-xs">
@@ -366,7 +326,9 @@ const ComposerAudioUpload: FC = () => {
 
   const handleAudioFile = useCallback(
     async (file: File) => {
-      if (file.size > MAX_AUDIO_SIZE) return;
+      if (file.size > MAX_AUDIO_SIZE) {
+        return;
+      }
       try {
         const base64 = await fileToBase64(file);
         setPendingAudio(base64, file.name);
@@ -377,7 +339,9 @@ const ComposerAudioUpload: FC = () => {
     [setPendingAudio],
   );
 
-  if (!activeModel?.hasAudioInput) return null;
+  if (!activeModel?.hasAudioInput) {
+    return null;
+  }
 
   return (
     <>
@@ -388,7 +352,9 @@ const ComposerAudioUpload: FC = () => {
         className="hidden"
         onChange={(e) => {
           const file = e.target.files?.[0];
-          if (file) handleAudioFile(file);
+          if (file) {
+            handleAudioFile(file);
+          }
           e.target.value = "";
         }}
       />
@@ -411,7 +377,9 @@ const ComposerAudioUpload: FC = () => {
 function applyQwenThinkingParams(thinkingOn: boolean): void {
   const store = useChatRuntimeStore.getState();
   const checkpoint = store.params.checkpoint?.toLowerCase() ?? "";
-  if (!checkpoint.includes("qwen3")) return;
+  if (!checkpoint.includes("qwen3")) {
+    return;
+  }
   const params = thinkingOn
     ? { temperature: 0.6, topP: 0.95, topK: 20, minP: 0.0 }
     : { temperature: 0.7, topP: 0.8, topK: 20, minP: 0.0 };
@@ -425,7 +393,7 @@ const ReasoningToggle: FC = () => {
   const supportsReasoning = useChatRuntimeStore((s) => s.supportsReasoning);
   const reasoningEnabled = useChatRuntimeStore((s) => s.reasoningEnabled);
   const setReasoningEnabled = useChatRuntimeStore((s) => s.setReasoningEnabled);
-  const disabled = !modelLoaded || !supportsReasoning;
+  const disabled = !(modelLoaded && supportsReasoning);
 
   return (
     <button
@@ -463,7 +431,7 @@ const WebSearchToggle: FC = () => {
   const supportsTools = useChatRuntimeStore((s) => s.supportsTools);
   const toolsEnabled = useChatRuntimeStore((s) => s.toolsEnabled);
   const setToolsEnabled = useChatRuntimeStore((s) => s.setToolsEnabled);
-  const disabled = !modelLoaded || !supportsTools;
+  const disabled = !(modelLoaded && supportsTools);
 
   return (
     <button
@@ -492,10 +460,8 @@ const CodeToolsToggle: FC = () => {
   );
   const supportsTools = useChatRuntimeStore((s) => s.supportsTools);
   const codeToolsEnabled = useChatRuntimeStore((s) => s.codeToolsEnabled);
-  const setCodeToolsEnabled = useChatRuntimeStore(
-    (s) => s.setCodeToolsEnabled,
-  );
-  const disabled = !modelLoaded || !supportsTools;
+  const setCodeToolsEnabled = useChatRuntimeStore((s) => s.setCodeToolsEnabled);
+  const disabled = !(modelLoaded && supportsTools);
 
   return (
     <button
@@ -510,7 +476,9 @@ const CodeToolsToggle: FC = () => {
             ? "bg-primary/10 text-primary hover:bg-primary/20"
             : "bg-muted text-muted-foreground hover:bg-muted-foreground/15",
       )}
-      aria-label={codeToolsEnabled ? "Disable code execution" : "Enable code execution"}
+      aria-label={
+        codeToolsEnabled ? "Disable code execution" : "Enable code execution"
+      }
     >
       <CodeToggleIcon className="size-3.5" />
       <span>Code</span>
@@ -523,6 +491,11 @@ const ToolStatusDisplay: FC = () => {
   const isThreadRunning = useAuiState(({ thread }) => thread.isRunning);
   const [elapsed, setElapsed] = useState(0);
   const [visible, setVisible] = useState(false);
+  const visibleRef = useRef(false);
+
+  useEffect(() => {
+    visibleRef.current = visible;
+  }, [visible]);
 
   useEffect(() => {
     if (!toolStatus) {
@@ -540,7 +513,7 @@ const ToolStatusDisplay: FC = () => {
     // tools show immediately so the badge does not flicker. Fast
     // tool calls that all complete under 300ms never show the badge.
     let showTimer: ReturnType<typeof setTimeout> | undefined;
-    if (!visible) {
+    if (!visibleRef.current) {
       showTimer = setTimeout(() => setVisible(true), 300);
     }
 
@@ -549,11 +522,15 @@ const ToolStatusDisplay: FC = () => {
     }, 1000);
     return () => {
       clearInterval(interval);
-      if (showTimer) clearTimeout(showTimer);
+      if (showTimer) {
+        clearTimeout(showTimer);
+      }
     };
   }, [toolStatus, isThreadRunning]);
 
-  if (!toolStatus || !visible) return null;
+  if (!(toolStatus && visible)) {
+    return null;
+  }
   const isRunning = toolStatus.startsWith("Running");
   const StatusIcon = isRunning ? TerminalIcon : GlobeIcon;
   return (
@@ -649,7 +626,9 @@ const GeneratingIndicator: FC = () => {
     ({ message }) =>
       message.content.length === 0 && message.status?.type === "running",
   );
-  if (!show) return null;
+  if (!show) {
+    return null;
+  }
   return <span className="text-sm text-muted-foreground">Generating...</span>;
 };
 
@@ -736,7 +715,9 @@ const CopyButton: FC = () => {
     const text = aui.message().getCopyText();
     if (await copyToClipboard(text)) {
       setCopied(true);
-      if (resetTimeoutRef.current) clearTimeout(resetTimeoutRef.current);
+      if (resetTimeoutRef.current) {
+        clearTimeout(resetTimeoutRef.current);
+      }
       resetTimeoutRef.current = setTimeout(() => {
         setCopied(false);
         resetTimeoutRef.current = null;
@@ -794,8 +775,12 @@ const AssistantActionBar: FC = () => {
 };
 
 const UserMessageAudio: FC = () => {
-  const audioName = useAuiState(({ message }) => sentAudioNames.get(message.id));
-  if (!audioName) return null;
+  const audioName = useAuiState(({ message }) =>
+    sentAudioNames.get(message.id),
+  );
+  if (!audioName) {
+    return null;
+  }
   return (
     <div className="col-start-2 flex justify-end">
       <div className="flex items-center gap-2 rounded-lg border border-foreground/20 bg-muted px-3 py-1.5 text-xs">
diff --git a/studio/frontend/src/components/assistant-ui/use-intent-aware-autoscroll.tsx b/studio/frontend/src/components/assistant-ui/use-intent-aware-autoscroll.tsx
new file mode 100644
index 0000000000..ad01afcdaf
--- /dev/null
+++ b/studio/frontend/src/components/assistant-ui/use-intent-aware-autoscroll.tsx
@@ -0,0 +1,451 @@
+// SPDX-License-Identifier: AGPL-3.0-only
+// Copyright 2026-present the Unsloth AI Inc. team. All rights reserved. See /studio/LICENSE.AGPL-3.0
+
+"use client";
+
+import { useAuiEvent } from "@assistant-ui/react";
+import {
+  type ReactNode,
+  type RefCallback,
+  createContext,
+  useCallback,
+  useContext,
+  useMemo,
+  useRef,
+  useSyncExternalStore,
+} from "react";
+
+/**
+ * Intent-aware autoscroll for a Thread viewport.
+ *
+ * Why we don't reuse assistant-ui's built-in autoscroll:
+ *   `useThreadViewportAutoScroll` runs unconditionally whenever
+ *   `ThreadPrimitive.Viewport` is mounted. Even with every
+ *   `scrollToBottomOn*` prop disabled, it still installs observers
+ *   that write `isAtBottom` to the shared viewport store on every
+ *   layout change. On sidebar toggles, browser resizes, or
+ *   mobile↔desktop breakpoint crossings, that write races our scroll
+ *   correction — whoever writes last wins, and the scroll-to-bottom
+ *   button flickers or sticks depending on observer ordering.
+ *
+ * Strategy:
+ *   - Own `isAtBottom` as local state (exposed via
+ *     `useIsThreadAtBottom`). Upstream can still write to its own
+ *     store; nobody reads from it.
+ *   - Drive the viewport with a single rAF loop governed by a follow
+ *     deadline (`followUntilRef`). Any signal that can invalidate
+ *     bottom alignment (resize, mutation, AUI event, programmatic
+ *     scroll) extends the deadline; the loop pins to the bottom and
+ *     reports `isAtBottom=true` until it expires, then settles on
+ *     pure DOM observation.
+ *   - Detect user intent (wheel up, touch swipe up, scroll direction)
+ *     to detach. While detached, resize/mutation signals don't extend
+ *     the deadline, so the button appears and stays. Re-attach when
+ *     the user scrolls *down* within 24px of the bottom.
+ */
+
+// 2px, not 1: subpixel rounding on HiDPI displays can leave a fractional
+// gap at the "true" bottom that a 1px threshold reports as not-at-bottom.
+const AT_BOTTOM_THRESHOLD_PX = 2;
+const RE_ATTACH_THRESHOLD_PX = 24;
+const TOUCH_MOVE_THRESHOLD_PX = 4;
+// Cumulative upward movement (summed across scroll events) that
+// counts as a deliberate detach. Summed rather than per-event so that
+// slow 1-px-per-event sources — middle-click autoscroll, scrollbar
+// drags, some trackpads — accumulate instead of slipping under a
+// per-event threshold forever.
+const UPWARD_DETACH_THRESHOLD_PX = 2;
+// Window during which the viewport pins to the bottom through
+// layout/content races. Extends on every resize/mutation, so streaming
+// keeps the viewport pinned as long as content keeps arriving; settles
+// this long after the last change.
+const FOLLOW_SETTLE_MS = 600;
+
+export type ScrollToBottom = (behavior?: ScrollBehavior) => void;
+
+type AutoScrollContextValue = {
+  scrollToBottom: ScrollToBottom;
+  getIsAtBottom: () => boolean;
+  subscribe: (listener: () => void) => () => void;
+};
+
+const noopContext: AutoScrollContextValue = {
+  scrollToBottom: () => {
+    /* no viewport mounted */
+  },
+  getIsAtBottom: () => true,
+  subscribe: () => () => {
+    /* no-op */
+  },
+};
+
+const AutoScrollContext = createContext<AutoScrollContextValue>(noopContext);
+
+export function IntentAwareScrollProvider({
+  value,
+  children,
+}: {
+  value: AutoScrollContextValue;
+  children: ReactNode;
+}) {
+  return (
+    <AutoScrollContext.Provider value={value}>
+      {children}
+    </AutoScrollContext.Provider>
+  );
+}
+
+export function useScrollThreadToBottom(): ScrollToBottom {
+  return useContext(AutoScrollContext).scrollToBottom;
+}
+
+export function useIsThreadAtBottom(): boolean {
+  const ctx = useContext(AutoScrollContext);
+  return useSyncExternalStore(ctx.subscribe, ctx.getIsAtBottom, () => true);
+}
+
+export function useIntentAwareAutoScroll(): {
+  ref: RefCallback<HTMLElement>;
+  context: AutoScrollContextValue;
+} {
+  const cleanupRef = useRef<(() => void) | null>(null);
+
+  const userDetachedRef = useRef(false);
+  const followUntilRef = useRef(0);
+
+  const isAtBottomRef = useRef(true);
+  const listenersRef = useRef<Set<() => void>>(new Set());
+
+  const scrollImplRef = useRef<ScrollToBottom>(() => {
+    /* no viewport mounted */
+  });
+
+  const getIsAtBottom = useCallback(() => isAtBottomRef.current, []);
+
+  const subscribe = useCallback((listener: () => void) => {
+    listenersRef.current.add(listener);
+    return () => {
+      listenersRef.current.delete(listener);
+    };
+  }, []);
+
+  const setIsAtBottom = useCallback((value: boolean) => {
+    if (isAtBottomRef.current === value) {
+      return;
+    }
+    isAtBottomRef.current = value;
+    for (const listener of listenersRef.current) {
+      listener();
+    }
+  }, []);
+
+  const scrollToBottom = useCallback<ScrollToBottom>((behavior) => {
+    scrollImplRef.current(behavior);
+  }, []);
+
+  const attach = useCallback(
+    (el: HTMLElement) => {
+      let rafId: number | null = null;
+      let lastScrollTop = el.scrollTop;
+      let lastClientWidth = el.clientWidth;
+      let lastClientHeight = el.clientHeight;
+      let upwardAccumulator = 0;
+      let touchStartY = 0;
+
+      const distanceFromBottom = (): number => {
+        if (el.scrollHeight <= el.clientHeight) {
+          return 0;
+        }
+        return el.scrollHeight - el.scrollTop - el.clientHeight;
+      };
+      let lastDistanceFromBottom = distanceFromBottom();
+
+      const atBottomStrict = (): boolean =>
+        distanceFromBottom() <= AT_BOTTOM_THRESHOLD_PX;
+
+      // True only when there is room to scroll upward. Guards the
+      // wheel/touch detach paths: a wheel-up or swipe-down gesture on
+      // a viewport with nothing above (short thread, or already at
+      // the top) can't express intent to leave the bottom and must
+      // not flip userDetachedRef — otherwise later streaming updates
+      // skip extendFollow and auto-follow stays dead for the session.
+      const canScrollUp = (): boolean => el.scrollTop > 0;
+
+      // True when a nested scrollable ancestor of the event target
+      // (e.g. the reasoning panel's own overflow-y-auto region, or
+      // any tool output with internal scroll) has room above and
+      // will therefore consume the upward delta before it reaches
+      // the viewport. Walking stops at the viewport element itself,
+      // so only intermediate inner scrollers count.
+      //
+      // Wheel and touchmove events bubble, so without this check a
+      // user reading back through a long reasoning pane mid-stream
+      // would falsely detach the outer viewport.
+      const innerScrollWillConsumeUpward = (
+        target: EventTarget | null,
+      ): boolean => {
+        let node =
+          target instanceof Element ? (target as Element | null) : null;
+        while (node && node !== el) {
+          if (node.scrollTop > 0) {
+            const { overflowY } = window.getComputedStyle(node);
+            if (
+              overflowY === "auto" ||
+              overflowY === "scroll" ||
+              overflowY === "overlay"
+            ) {
+              return true;
+            }
+          }
+          node = node.parentElement;
+        }
+        return false;
+      };
+
+      const extendFollow = (): void => {
+        if (userDetachedRef.current) {
+          return;
+        }
+        followUntilRef.current = performance.now() + FOLLOW_SETTLE_MS;
+      };
+
+      const detach = (): void => {
+        userDetachedRef.current = true;
+        followUntilRef.current = 0;
+      };
+
+      const requestTick = (): void => {
+        if (rafId === null) {
+          rafId = requestAnimationFrame(tick);
+        }
+      };
+
+      // Single rAF loop. While within the follow window and not
+      // detached, pin the viewport to the bottom and report
+      // isAtBottom=true every frame. Otherwise settle on whatever the
+      // DOM says. Scheduling is edge-triggered: scroll/resize/mutation
+      // events call requestTick(), and the loop self-perpetuates only
+      // as long as pinning is still active.
+      const tick = (): void => {
+        rafId = null;
+        const following =
+          !userDetachedRef.current &&
+          performance.now() < followUntilRef.current;
+
+        if (following) {
+          if (!atBottomStrict() && el.scrollHeight > el.clientHeight) {
+            el.scrollTo({ top: el.scrollHeight, behavior: "instant" });
+          }
+          setIsAtBottom(true);
+          requestTick();
+          return;
+        }
+
+        setIsAtBottom(atBottomStrict());
+      };
+
+      scrollImplRef.current = (behavior = "auto") => {
+        userDetachedRef.current = false;
+        followUntilRef.current = performance.now() + FOLLOW_SETTLE_MS;
+        if (el.scrollHeight > el.clientHeight) {
+          el.scrollTo({ top: el.scrollHeight, behavior });
+        }
+        setIsAtBottom(true);
+        requestTick();
+      };
+
+      const onWheel = (e: WheelEvent) => {
+        if (
+          e.deltaY < 0 &&
+          canScrollUp() &&
+          !innerScrollWillConsumeUpward(e.target)
+        ) {
+          detach();
+        }
+      };
+
+      const onTouchStart = (e: TouchEvent) => {
+        touchStartY = e.touches[0]?.clientY ?? 0;
+      };
+
+      const onTouchMove = (e: TouchEvent) => {
+        const y = e.touches[0]?.clientY ?? 0;
+        // Finger moves DOWN on the screen = content scrolls UP.
+        if (
+          y - touchStartY > TOUCH_MOVE_THRESHOLD_PX &&
+          canScrollUp() &&
+          !innerScrollWillConsumeUpward(e.target)
+        ) {
+          detach();
+        }
+      };
+
+      const onScroll = () => {
+        const scrollTop = el.scrollTop;
+        const clientWidth = el.clientWidth;
+        const clientHeight = el.clientHeight;
+        const sizeChanged =
+          clientWidth !== lastClientWidth || clientHeight !== lastClientHeight;
+
+        const delta = scrollTop - lastScrollTop;
+        const distanceNow = distanceFromBottom();
+
+        // Viewport resizes can clamp scrollTop and produce spurious
+        // direction signals. Only flip intent on deliberate scrolls.
+        if (sizeChanged) {
+          upwardAccumulator = 0;
+        } else if (delta > 0) {
+          // Downward: reset the upward accumulator, and re-attach when
+          // the user has scrolled back within range of the bottom.
+          upwardAccumulator = 0;
+          if (
+            userDetachedRef.current &&
+            distanceNow <= RE_ATTACH_THRESHOLD_PX
+          ) {
+            userDetachedRef.current = false;
+            extendFollow();
+          }
+        } else if (delta < 0 && !userDetachedRef.current) {
+          // Upward: sum across events. Middle-click autoscroll and
+          // some trackpads deliver 1px-per-event scrolls that each
+          // slip under a per-event threshold; summing catches them.
+          //
+          // Count distance-from-bottom growth, not raw scrollTop
+          // delta. When content above collapses (reasoning panels
+          // auto-closing after streaming, tool outputs auto-hiding),
+          // browsers scroll-anchor to keep visible content stable:
+          // scrollTop decreases but scrollHeight decreases by the
+          // same amount, so distance is unchanged. Those layout-
+          // induced deltas must not flip user intent.
+          const distanceDelta = distanceNow - lastDistanceFromBottom;
+          if (distanceDelta > 0) {
+            upwardAccumulator += distanceDelta;
+            if (upwardAccumulator >= UPWARD_DETACH_THRESHOLD_PX) {
+              detach();
+              upwardAccumulator = 0;
+            }
+          }
+        }
+
+        lastScrollTop = scrollTop;
+        lastClientWidth = clientWidth;
+        lastClientHeight = clientHeight;
+        lastDistanceFromBottom = distanceNow;
+        requestTick();
+      };
+
+      const resizeObserver = new ResizeObserver(() => {
+        extendFollow();
+        requestTick();
+      });
+
+      const mutationObserver = new MutationObserver(() => {
+        extendFollow();
+        requestTick();
+      });
+
+      const onViewportResize = () => {
+        extendFollow();
+        requestTick();
+      };
+
+      // Fresh attach always starts pinned. `userDetachedRef` survives
+      // ref rebinds (it's hook-scoped), so if the viewport element is
+      // ever unmounted and remounted without an AUI lifecycle event
+      // (e.g. a parent layout refactor that remounts the viewport),
+      // a prior detach would silently disable auto-follow for the
+      // rest of the session.
+      userDetachedRef.current = false;
+
+      // Pin to bottom when the ref first attaches. Covers the case
+      // where `thread.initialize` fires before the ref is bound.
+      extendFollow();
+      if (el.scrollHeight > el.clientHeight) {
+        el.scrollTo({ top: el.scrollHeight, behavior: "instant" });
+      }
+      setIsAtBottom(true);
+      requestTick();
+
+      resizeObserver.observe(el);
+      mutationObserver.observe(el, {
+        childList: true,
+        subtree: true,
+        characterData: true,
+        // Layout-affecting attributes only. Excludes `style`, which
+        // elements may mutate in response to viewport state (feedback
+        // loop). `class` catches Tailwind show/hide; `hidden` /
+        // `aria-hidden` / `aria-expanded` / `data-state` catch Radix
+        // and native collapsible toggles that change scrollHeight
+        // without triggering the viewport ResizeObserver.
+        attributes: true,
+        attributeFilter: [
+          "class",
+          "hidden",
+          "aria-hidden",
+          "aria-expanded",
+          "data-state",
+        ],
+      });
+      el.addEventListener("wheel", onWheel, { passive: true });
+      el.addEventListener("touchstart", onTouchStart, { passive: true });
+      el.addEventListener("touchmove", onTouchMove, { passive: true });
+      el.addEventListener("scroll", onScroll, { passive: true });
+      // ResizeObserver above covers browser-window resizes (they resize
+      // the viewport element). visualViewport.resize is the only signal
+      // for iOS software-keyboard changes, where the visual viewport
+      // shrinks without the viewport element's clientHeight changing.
+      window.visualViewport?.addEventListener("resize", onViewportResize);
+
+      return () => {
+        if (rafId !== null) {
+          cancelAnimationFrame(rafId);
+          rafId = null;
+        }
+        resizeObserver.disconnect();
+        mutationObserver.disconnect();
+        el.removeEventListener("wheel", onWheel);
+        el.removeEventListener("touchstart", onTouchStart);
+        el.removeEventListener("touchmove", onTouchMove);
+        el.removeEventListener("scroll", onScroll);
+        window.visualViewport?.removeEventListener("resize", onViewportResize);
+        scrollImplRef.current = () => {
+          /* no viewport mounted */
+        };
+      };
+    },
+    [setIsAtBottom],
+  );
+
+  // Thread lifecycle moments that always pin to the bottom, regardless
+  // of prior detach state. "auto" respects CSS smooth scrolling for
+  // runStart so new turns glide in; "instant" snaps for load/switch
+  // where any animation would just be wasted motion.
+  const pinToBottom = useCallback((behavior: ScrollBehavior) => {
+    userDetachedRef.current = false;
+    scrollImplRef.current(behavior);
+  }, []);
+
+  useAuiEvent("thread.runStart", () => pinToBottom("auto"));
+  useAuiEvent("thread.initialize", () => pinToBottom("instant"));
+  useAuiEvent("threadListItem.switchedTo", () => pinToBottom("instant"));
+
+  const ref = useCallback<RefCallback<HTMLElement>>(
+    (el) => {
+      if (cleanupRef.current) {
+        cleanupRef.current();
+        cleanupRef.current = null;
+      }
+      if (el) {
+        cleanupRef.current = attach(el);
+      }
+    },
+    [attach],
+  );
+
+  const context = useMemo<AutoScrollContextValue>(
+    () => ({ scrollToBottom, getIsAtBottom, subscribe }),
+    [scrollToBottom, getIsAtBottom, subscribe],
+  );
+
+  return { ref, context };
+}
diff --git a/studio/frontend/src/components/ui/sidebar.tsx b/studio/frontend/src/components/ui/sidebar.tsx
index 967b0dcaac..8eb8c51491 100644
--- a/studio/frontend/src/components/ui/sidebar.tsx
+++ b/studio/frontend/src/components/ui/sidebar.tsx
@@ -518,7 +518,7 @@ function SidebarMenu({ className, ...props }: React.ComponentProps<"ul">) {
     <ul
       data-slot="sidebar-menu"
       data-sidebar="menu"
-      className={cn("gap-0.5 flex w-full min-w-0 flex-col", className)}
+      className={cn("gap-px flex w-full min-w-0 flex-col", className)}
       {...props}
     />
   )
@@ -536,7 +536,7 @@ function SidebarMenuItem({ className, ...props }: React.ComponentProps<"li">) {
 }
 
 const sidebarMenuButtonVariants = cva(
-  "ring-sidebar-ring hover:bg-sidebar-accent hover:text-sidebar-accent-foreground active:bg-sidebar-accent active:text-sidebar-accent-foreground data-active:bg-sidebar-accent data-active:text-sidebar-accent-foreground data-open:hover:bg-sidebar-accent data-open:hover:text-sidebar-accent-foreground gap-2 rounded-md p-2 text-left text-sm cursor-pointer transition-[width,height,padding] group-has-data-[sidebar=menu-action]/menu-item:pr-8 group-data-[collapsible=icon]:w-full! group-data-[collapsible=icon]:justify-center group-data-[collapsible=icon]:p-2! data-active:font-medium peer/menu-button flex w-full items-center overflow-hidden outline-hidden group/menu-button disabled:pointer-events-none disabled:opacity-50 aria-disabled:pointer-events-none aria-disabled:opacity-50 [&>span:last-child]:truncate group-data-[collapsible=icon]:[&>span]:hidden [&_svg]:size-4 [&_svg]:shrink-0 group-data-[collapsible=icon]:[&_svg]:size-[18px]",
+  "ring-sidebar-ring hover:bg-sidebar-accent hover:text-sidebar-accent-foreground active:bg-sidebar-accent active:text-sidebar-accent-foreground data-active:bg-sidebar-accent data-active:text-sidebar-accent-foreground data-open:hover:bg-sidebar-accent data-open:hover:text-sidebar-accent-foreground gap-2 rounded-md p-2 text-left text-sm cursor-pointer group-has-data-[sidebar=menu-action]/menu-item:pr-8 group-data-[collapsible=icon]:w-full! group-data-[collapsible=icon]:justify-center group-data-[collapsible=icon]:p-2! data-active:font-medium peer/menu-button flex w-full items-center overflow-hidden outline-hidden group/menu-button disabled:pointer-events-none disabled:opacity-50 aria-disabled:pointer-events-none aria-disabled:opacity-50 [&>span:last-child]:truncate group-data-[collapsible=icon]:[&>span]:hidden [&_svg]:size-4 [&_svg]:shrink-0 group-data-[collapsible=icon]:[&_svg]:size-[18px]",
   {
     variants: {
       variant: {
diff --git a/studio/frontend/src/features/chat/chat-page.tsx b/studio/frontend/src/features/chat/chat-page.tsx
index 32e34b3c72..9987af230b 100644
--- a/studio/frontend/src/features/chat/chat-page.tsx
+++ b/studio/frontend/src/features/chat/chat-page.tsx
@@ -10,14 +10,9 @@ import { Thread } from "@/components/assistant-ui/thread";
 import { cn } from "@/lib/utils";
 import { GuidedTour, useGuidedTourController } from "@/features/tour";
 import { useSidebar } from "@/components/ui/sidebar";
-import {
-  Settings05Icon,
-} from "@hugeicons/core-free-icons";
+import { Settings05Icon } from "@hugeicons/core-free-icons";
 import { HugeiconsIcon } from "@hugeicons/react";
-import {
-  Tooltip,
-  TooltipContent,
-} from "@/components/ui/tooltip";
+import { Tooltip, TooltipContent } from "@/components/ui/tooltip";
 import { Tooltip as TooltipPrimitive } from "radix-ui";
 import { useNavigate, useSearch } from "@tanstack/react-router";
 import {
@@ -351,7 +346,7 @@ function GeneralCompareHeader({
   return (
     <div
       className={cn(
-        "flex h-[48px] shrink-0 items-center gap-2 bg-background",
+        "flex h-[48px] shrink-0 items-start pt-[11px] gap-2 bg-background",
         side === "left" ? "pl-12 pr-3 md:pl-2" : "pl-3 pr-12",
       )}
     >
@@ -564,8 +559,7 @@ export function ChatPage(): ReactElement {
     if (search.compare) {
       return {
         mode: "compare",
-        pairId:
-          search.compare,
+        pairId: search.compare,
       };
     }
     if (search.thread) {
@@ -658,8 +652,14 @@ export function ChatPage(): ReactElement {
     },
     [modelSelectorLocked],
   );
-  const openSettings = useCallback(() => setSettingsOpen(true), [setSettingsOpen]);
-  const closeSettings = useCallback(() => setSettingsOpen(false), [setSettingsOpen]);
+  const openSettings = useCallback(
+    () => setSettingsOpen(true),
+    [setSettingsOpen],
+  );
+  const closeSettings = useCallback(
+    () => setSettingsOpen(false),
+    [setSettingsOpen],
+  );
   const { setPinned, isMobile } = useSidebar();
   const openSidebar = useCallback(() => setPinned(true), [setPinned]);
 
@@ -686,7 +686,9 @@ export function ChatPage(): ReactElement {
         .first()
         .then((msg) => {
           const metadata = msg?.metadata as Record<string, unknown> | undefined;
-          const usage = metadata?.contextUsage as ReturnType<typeof useChatRuntimeStore.getState>["contextUsage"];
+          const usage = metadata?.contextUsage as ReturnType<
+            typeof useChatRuntimeStore.getState
+          >["contextUsage"];
           if (usage) useChatRuntimeStore.getState().setContextUsage(usage);
         });
     }
@@ -865,7 +867,8 @@ export function ChatPage(): ReactElement {
           className={cn(
             "absolute top-0 left-0 right-[10px] z-30 flex h-[48px] shrink-0 items-start pt-[11px] pr-2 bg-background",
             isMobile ? "pl-12 pr-1.5" : "pl-2",
-            view.mode === "compare" && "right-[10px] left-auto w-auto bg-transparent pl-0 pr-2",
+            view.mode === "compare" &&
+              "right-[10px] left-auto w-auto bg-transparent pl-0 pr-2",
           )}
         >
           <div className="flex items-center gap-1">
diff --git a/studio/frontend/src/features/chat/chat-settings-sheet.tsx b/studio/frontend/src/features/chat/chat-settings-sheet.tsx
index 24a9ac4077..e4574dab74 100644
--- a/studio/frontend/src/features/chat/chat-settings-sheet.tsx
+++ b/studio/frontend/src/features/chat/chat-settings-sheet.tsx
@@ -1271,7 +1271,7 @@ export function ChatSettingsPanel({
 
   return (
     <aside
-      className={`relative z-50 shrink-0 h-full overflow-hidden bg-muted/70 transition-[width] duration-200 ease-linear ${open ? "w-[17rem]" : "w-0"}`}
+      className={`relative z-50 shrink-0 h-full overflow-hidden bg-muted/70 ${open ? "w-[17rem]" : "w-0"}`}
     >
       <div className="h-full w-[17rem]">{settingsContent}</div>
     </aside>
diff --git a/studio/frontend/src/index.css b/studio/frontend/src/index.css
index 3943c7f882..880bc023e1 100644
--- a/studio/frontend/src/index.css
+++ b/studio/frontend/src/index.css
@@ -281,6 +281,22 @@
 		}
 	}
 
+	--animate-icon-pop: icon-pop 0.3s ease-out;
+
+	@keyframes icon-pop {
+		0% {
+			transform: scale(1);
+		}
+
+		45% {
+			transform: scale(1.08);
+		}
+
+		100% {
+			transform: scale(1);
+		}
+	}
+
 	--animate-shine: shine var(--duration) infinite linear;
 
 	@keyframes shine {
@@ -306,6 +322,9 @@
 	body {
 		@apply font-sans bg-background text-foreground;
 		letter-spacing: var(--tracking-normal);
+		-webkit-font-smoothing: antialiased;
+		-moz-osx-font-smoothing: grayscale;
+		text-rendering: optimizeLegibility;
 	}
 
 	html {
@@ -334,18 +353,15 @@
 
 @layer utilities {
 
-	/* Heading font utility — matches the sidebar logo's letter-spacing so
-	   all Hellix text (select model, menus, nav items) stays visually
-	   consistent, with a small bump in dark mode to offset optical bloom. */
+	/* Heading font utility — the logo applies its own dark-mode tracking
+	   to offset optical bloom at large sizes. For small UI text (menus,
+	   nav items) we keep a single tight tracking in both themes, relying
+	   on global antialiased font-smoothing to neutralize the bloom. */
 	.font-heading {
 		font-family: var(--font-heading);
 		letter-spacing: -0.01em;
 	}
 
-	.dark .font-heading {
-		letter-spacing: 0.02em;
-	}
-
 	/* Elevated surface shadow (use ring-* for borders) */
 	.shadow-border {
 		--tw-shadow: 0 4px 16px rgba(0, 0, 0, 0.1);

From 301153587129cf008719e208df45100f696470d2 Mon Sep 17 00:00:00 2001
From: "pre-commit-ci[bot]"
 <66853113+pre-commit-ci[bot]@users.noreply.github.com>
Date: Wed, 22 Apr 2026 09:02:48 -0700
Subject: [PATCH 149/216] [pre-commit.ci] pre-commit autoupdate (#5117)
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

updates:
- [github.com/astral-sh/ruff-pre-commit: v0.15.10 → v0.15.11](https://github.com/astral-sh/ruff-pre-commit/compare/v0.15.10...v0.15.11)

Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com>
Co-authored-by: Roland Tannous <115670425+rolandtannous@users.noreply.github.com>
---
 .pre-commit-config.yaml | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/.pre-commit-config.yaml b/.pre-commit-config.yaml
index 309c7d9aca..1d3137b373 100644
--- a/.pre-commit-config.yaml
+++ b/.pre-commit-config.yaml
@@ -1,6 +1,6 @@
 repos:
   - repo: https://github.com/astral-sh/ruff-pre-commit
-    rev: v0.15.10
+    rev: v0.15.11
     hooks:
       - id: ruff
         args:

From 77756faa4612590582861d9bbd1007883b9b324e Mon Sep 17 00:00:00 2001
From: Datta Nimmaturi <venkatadattasainimmaturi@gmail.com>
Date: Wed, 22 Apr 2026 21:33:20 +0530
Subject: [PATCH 150/216] Fix tokenizer save gemma (#5115)

* [WIP] Fast inference for qwen3.5

* fix tokenizer not saving properly

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

* extend to VLM and clenaup

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

* gate tokenizer.model saving

* fix for gated/private models

* Fix tokenizer save review findings

- save.py:261 restore dict-based _TOKENIZER_MODEL_CACHE so negative
  results are cached; the set() in 0129fb5e regressed non-SentencePiece
  tokenizer saves to a fresh HfApi.model_info call on every checkpoint.
  Don't cache on exception so gated/private repos can retry later with a
  valid token.
- save.py:282 guard `repo_info.siblings` with `or []`; huggingface_hub
  types this Optional and returns None for empty or new repos, which
  made any() raise TypeError out of save_pretrained.
- save.py:3487 split push_to_hub into local save + _preserve + push so
  uploaded tokenizer_config.json/tokenizer.model include the fix rather
  than the unfixed copies written before the upload.
- save.py:3352 call patch_saving_functions on tokenizers passed to
  unsloth_save_pretrained_torchao to match the other three save
  entrypoints; previously torchao saves skipped the preservation patch.

* Fix push_to_hub repo_id conflict and torchao token forwarding

- save.py:3493-3496 pop `repo_id` from kwargs (defaulting to
  `save_directory`) before calling `self.push_to_hub(repo_id, **kwargs)`.
  The previous `self.push_to_hub(save_directory, **kwargs)` passed
  `save_directory` as the first positional `repo_id` while also
  forwarding a user-supplied `repo_id` through kwargs, raising
  `TypeError: got multiple values for argument 'repo_id'` on the
  standard `save_pretrained(local_path, push_to_hub=True, repo_id=...)`
  call shape. This regression was introduced by the earlier iteration
  that split push_to_hub into an explicit second step.
- save.py:3314 forward `token=token` on the torchao non-PEFT
  `tokenizer.save_pretrained(torchao_save_directory)` call so the
  patched wrapper can reach gated repos when HF_TOKEN is not in the
  environment. Left the sibling `unsloth_generic_save` call at 3063
  untouched (blame points at an earlier full-finetuned
  save_pretrained_merged fix and the token gap there is lower risk).

* Fix torchao tokenizer reload and push_to_hub repo_id default

- save.py:3283 after `auto_processor.from_pretrained(save_directory)`
  re-runs `patch_saving_functions(tokenizer)` on the freshly loaded
  tokenizer. The rebind at 3283 was overwriting the patched tokenizer
  passed into `unsloth_save_pretrained_torchao`, so the subsequent
  `tokenizer.push_to_hub` (3309) and `tokenizer.save_pretrained`
  (3314) bypassed `_preserve_sentencepiece_tokenizer_assets` and left
  `{save_directory}-torchao` without `tokenizer.model` / restored
  `added_tokens_decoder`.
- save.py:3497 fall back to `os.path.basename(save_directory)` for
  `repo_id` instead of the raw `save_directory`. The round-2 fallback
  diverged from `transformers.PreTrainedTokenizerBase.save_pretrained`,
  which defaults `repo_id = save_directory.split(os.path.sep)[-1]`;
  nested local paths like `./out/my-repo` now resolve to `my-repo`
  (the Hub id) instead of the full filesystem path.

* Revert tokenizer save_pretrained repo_id basename fallback

- save.py:3497 default `repo_id` back to `save_directory` as-is rather
  than `os.path.basename(save_directory)`. The basename fallback (added
  last iteration to match upstream transformers) stripped the user
  namespace from the Unsloth convention `tokenizer.save_pretrained(
  "user/repo", push_to_hub=True)`, redirecting the upload to
  `{current_user}/repo`. save.py itself treats `save_directory` as the
  repo id at 572, 593, 1723, 1779, 1836, 1844, 1858, and 3025, so the
  wrapper should follow the same convention. Users who pass a nested
  filesystem path with `push_to_hub=True` can supply explicit
  `repo_id=...`.

* Guard processor.tokenizer recursion against None

save.py:3511 change `elif hasattr(model, "tokenizer")` to
`elif getattr(model, "tokenizer", None) is not None`. The previous
guard only checked attribute existence; a ProcessorMixin that sets
`tokenizer = None` (audio-only or manually constructed) would enter
the branch and crash inside the recursive patch_saving_functions on
`model.push_to_hub.__name__`.

* Add review tests for tokenizer save

* Consolidate review tests

Drop redundant assertion in test_patch_saving_functions_still_patches_non_none_tokenizer.
The hasattr check already proves the patch applied; the or-chained
repeat assertion added no signal.

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

---------

Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com>
Co-authored-by: Daniel Han <danielhanchen@gmail.com>
---
 .../test_patch_saving_none_tokenizer.py       |  47 ++++
 unsloth/save.py                               | 204 ++++++++++++++++--
 2 files changed, 233 insertions(+), 18 deletions(-)
 create mode 100644 tests/saving/test_patch_saving_none_tokenizer.py

diff --git a/tests/saving/test_patch_saving_none_tokenizer.py b/tests/saving/test_patch_saving_none_tokenizer.py
new file mode 100644
index 0000000000..e5aa7f9b21
--- /dev/null
+++ b/tests/saving/test_patch_saving_none_tokenizer.py
@@ -0,0 +1,47 @@
+from unittest.mock import MagicMock
+
+from transformers import PreTrainedTokenizerBase
+
+from unsloth.save import patch_saving_functions
+
+
+class _ProcessorWithNoneTokenizer:
+    tokenizer = None
+
+    def push_to_hub(self, *args, **kwargs):
+        return None
+
+    push_to_hub.__doc__ = "stub"
+
+    def save_pretrained(self, *args, **kwargs):
+        return None
+
+
+def test_patch_saving_functions_no_crash_on_none_tokenizer():
+    proc = _ProcessorWithNoneTokenizer()
+    patch_saving_functions(proc)
+
+
+def test_patch_saving_functions_still_patches_non_none_tokenizer():
+    inner = MagicMock(spec = PreTrainedTokenizerBase)
+    inner.save_pretrained = MagicMock()
+    inner.save_pretrained.__name__ = "save_pretrained"
+    inner.push_to_hub = MagicMock()
+    inner.push_to_hub.__name__ = "push_to_hub"
+    inner.push_to_hub.__doc__ = "tokenizer doc"
+
+    class _Proc:
+        def __init__(self, tok):
+            self.tokenizer = tok
+
+        def push_to_hub(self, *args, **kwargs):
+            return None
+
+        push_to_hub.__doc__ = "proc doc"
+
+        def save_pretrained(self, *args, **kwargs):
+            return None
+
+    proc = _Proc(inner)
+    patch_saving_functions(proc)
+    assert hasattr(inner, "original_save_pretrained")
diff --git a/unsloth/save.py b/unsloth/save.py
index 22ae5487d7..4e9e1bfe11 100644
--- a/unsloth/save.py
+++ b/unsloth/save.py
@@ -41,6 +41,7 @@ import sys
 import requests
 import torch
 import os
+import json
 import shutil
 import pickle
 import gc
@@ -50,11 +51,10 @@ import subprocess
 import psutil
 import re
 from transformers.models.llama.modeling_llama import logger
-from .tokenizer_utils import fix_sentencepiece_gguf
 from .models.loader_utils import get_model_name
 from .models._utils import _convert_torchao_model
 from .ollama_template_mappers import OLLAMA_TEMPLATES, MODEL_TO_OLLAMA_TEMPLATE_MAPPER
-from transformers import ProcessorMixin
+from transformers import ProcessorMixin, PreTrainedTokenizerBase
 from huggingface_hub import HfApi
 
 try:
@@ -258,6 +258,105 @@ def check_if_sentencepiece_model(
     return sentencepiece_model
 
 
+_TOKENIZER_MODEL_CACHE = {}
+
+
+def _has_tokenizer_model(tokenizer, token = None):
+    tokenizer = tokenizer.tokenizer if hasattr(tokenizer, "tokenizer") else tokenizer
+    if tokenizer is None:
+        return False
+
+    source = getattr(tokenizer, "name_or_path", None)
+    if not isinstance(source, str) or not source:
+        return False
+    if os.path.isdir(source):
+        return os.path.isfile(os.path.join(source, "tokenizer.model"))
+    if source in _TOKENIZER_MODEL_CACHE:
+        return _TOKENIZER_MODEL_CACHE[source]
+
+    try:
+        repo_info = HfApi(token = token).model_info(source, files_metadata = False)
+    except Exception:
+        return False
+
+    has_tokenizer_model = any(
+        sibling.rfilename == "tokenizer.model" for sibling in (repo_info.siblings or [])
+    )
+    _TOKENIZER_MODEL_CACHE[source] = has_tokenizer_model
+    return has_tokenizer_model
+
+
+def _preserve_sentencepiece_tokenizer_assets(
+    tokenizer,
+    save_directory,
+    token = None,
+):
+    tokenizer = tokenizer.tokenizer if hasattr(tokenizer, "tokenizer") else tokenizer
+    if tokenizer is None or not os.path.isdir(save_directory):
+        return
+
+    tokenizer_config_path = os.path.join(save_directory, "tokenizer_config.json")
+    if os.path.isfile(tokenizer_config_path):
+        desired_added_tokens_decoder = {}
+        for token_id, added_token in getattr(
+            tokenizer, "added_tokens_decoder", {}
+        ).items():
+            desired_added_tokens_decoder[str(token_id)] = {
+                "content": getattr(added_token, "content", str(added_token)),
+                "single_word": getattr(added_token, "single_word", False),
+                "lstrip": getattr(added_token, "lstrip", False),
+                "rstrip": getattr(added_token, "rstrip", False),
+                "normalized": getattr(added_token, "normalized", True),
+                "special": getattr(added_token, "special", False),
+            }
+        if desired_added_tokens_decoder:
+            with open(tokenizer_config_path, "r", encoding = "utf-8") as file:
+                tokenizer_config = json.load(file)
+            if (
+                tokenizer_config.get("added_tokens_decoder")
+                != desired_added_tokens_decoder
+            ):
+                tokenizer_config["added_tokens_decoder"] = desired_added_tokens_decoder
+                with open(tokenizer_config_path, "w", encoding = "utf-8") as file:
+                    json.dump(tokenizer_config, file, indent = 2, ensure_ascii = False)
+                    file.write("\n")
+                logger.warning_once(
+                    f"Unsloth: Restored added_tokens_decoder metadata in "
+                    f"{tokenizer_config_path}."
+                )
+
+    tokenizer_model = os.path.join(save_directory, "tokenizer.model")
+    downloaded_path = None
+    if not os.path.isfile(tokenizer_model) and _has_tokenizer_model(
+        tokenizer,
+        token = token,
+    ):
+        source = getattr(tokenizer, "name_or_path", None)
+        if isinstance(source, str) and source:
+            if os.path.isdir(source):
+                local_path = os.path.join(source, "tokenizer.model")
+                if os.path.isfile(local_path):
+                    downloaded_path = local_path
+            else:
+                from huggingface_hub import hf_hub_download
+
+                try:
+                    downloaded_path = hf_hub_download(
+                        repo_id = source,
+                        filename = "tokenizer.model",
+                        token = token,
+                    )
+                except Exception:
+                    downloaded_path = None
+
+    if not os.path.isfile(tokenizer_model) and downloaded_path is not None:
+        shutil.copy2(downloaded_path, tokenizer_model)
+        logger.warning_once(
+            f"Unsloth: Preserved sentencepiece asset `tokenizer.model` in "
+            f"{save_directory}."
+        )
+
+
 def _free_cached_model(model):
     from huggingface_hub import scan_cache_dir
 
@@ -353,6 +452,9 @@ def unsloth_save_model(
     maximum_memory_usage: float = 0.9,
     datasets: Optional[List[str]] = None,
 ):
+    if isinstance(tokenizer, (PreTrainedTokenizerBase, ProcessorMixin)):
+        tokenizer = patch_saving_functions(tokenizer)
+
     if token is None:
         token = get_token()
 
@@ -480,8 +582,11 @@ def unsloth_save_model(
         )
         if tokenizer is not None:
             # Set padding side to left for inference
-            old_padding_side = tokenizer.padding_side
-            tokenizer.padding_side = "left"
+            _tokenizer = (
+                tokenizer.tokenizer if hasattr(tokenizer, "tokenizer") else tokenizer
+            )
+            old_padding_side = _tokenizer.padding_side
+            _tokenizer.padding_side = "left"
 
             getattr(tokenizer, "original_push_to_hub", tokenizer.push_to_hub)(
                 repo_id = save_directory,
@@ -498,7 +603,7 @@ def unsloth_save_model(
             )
 
             # Revert back padding side
-            tokenizer.padding_side = old_padding_side
+            _tokenizer.padding_side = old_padding_side
 
         if hasattr(model, "config"):
             print(
@@ -579,13 +684,16 @@ def unsloth_save_model(
             print("Unsloth: Saving tokenizer...", end = "")
 
             # Set padding side to left for inference
-            old_padding_side = tokenizer.padding_side
-            tokenizer.padding_side = "left"
+            _tokenizer = (
+                tokenizer.tokenizer if hasattr(tokenizer, "tokenizer") else tokenizer
+            )
+            old_padding_side = _tokenizer.padding_side
+            _tokenizer.padding_side = "left"
 
             tokenizer.save_pretrained(**tokenizer_save_settings)
 
             # Revert back padding side
-            tokenizer.padding_side = old_padding_side
+            _tokenizer.padding_side = old_padding_side
 
             print(" Done.")
         else:
@@ -865,13 +973,16 @@ def unsloth_save_model(
         print("Unsloth: Saving tokenizer...", end = "")
 
         # Set padding side to left for inference
-        old_padding_side = tokenizer.padding_side
-        tokenizer.padding_side = "left"
+        _tokenizer = (
+            tokenizer.tokenizer if hasattr(tokenizer, "tokenizer") else tokenizer
+        )
+        old_padding_side = _tokenizer.padding_side
+        _tokenizer.padding_side = "left"
 
         tokenizer.save_pretrained(**tokenizer_save_settings)
 
         # Revert back padding side
-        tokenizer.padding_side = old_padding_side
+        _tokenizer.padding_side = old_padding_side
 
         print(" Done.")
     else:
@@ -1993,6 +2104,8 @@ def unsloth_save_pretrained_gguf(
     """
     if tokenizer is None:
         raise ValueError("Unsloth: Saving to GGUF must have a tokenizer.")
+    if isinstance(tokenizer, (PreTrainedTokenizerBase, ProcessorMixin)):
+        tokenizer = patch_saving_functions(tokenizer)
 
     try:
         base_model_name = get_model_name(self.config._name_or_path, load_in_4bit = False)
@@ -2865,6 +2978,9 @@ def unsloth_generic_save(
     maximum_memory_usage: float = 0.9,
     datasets: Optional[List[str]] = None,
 ):
+    if isinstance(tokenizer, (PreTrainedTokenizerBase, ProcessorMixin)):
+        tokenizer = patch_saving_functions(tokenizer)
+
     if token is None and push_to_hub:
         token = get_token()
 
@@ -2916,8 +3032,13 @@ def unsloth_generic_save(
                 **_save_kwargs,
             )
             if tokenizer is not None:
-                old_padding_side = tokenizer.padding_side
-                tokenizer.padding_side = "left"
+                _tokenizer = (
+                    tokenizer.tokenizer
+                    if hasattr(tokenizer, "tokenizer")
+                    else tokenizer
+                )
+                old_padding_side = _tokenizer.padding_side
+                _tokenizer.padding_side = "left"
                 tokenizer.push_to_hub(
                     save_directory,
                     token = token,
@@ -2926,15 +3047,20 @@ def unsloth_generic_save(
                     create_pr = create_pr,
                     revision = revision,
                 )
-                tokenizer.padding_side = old_padding_side
+                _tokenizer.padding_side = old_padding_side
         else:
             print(f"Unsloth: Saving full fine-tuned model to '{save_directory}' ...")
             model.save_pretrained(save_directory, **_save_kwargs)
             if tokenizer is not None:
-                old_padding_side = tokenizer.padding_side
-                tokenizer.padding_side = "left"
+                _tokenizer = (
+                    tokenizer.tokenizer
+                    if hasattr(tokenizer, "tokenizer")
+                    else tokenizer
+                )
+                old_padding_side = _tokenizer.padding_side
+                _tokenizer.padding_side = "left"
                 tokenizer.save_pretrained(save_directory)
-                tokenizer.padding_side = old_padding_side
+                _tokenizer.padding_side = old_padding_side
 
         print(f"Unsloth: Model saved successfully to '{save_directory}'")
     else:
@@ -3154,6 +3280,8 @@ def _unsloth_save_torchao_with_given_config(
     auto_processor = AutoProcessor if is_vlm else AutoTokenizer
 
     tokenizer = auto_processor.from_pretrained(save_directory)
+    if isinstance(tokenizer, (PreTrainedTokenizerBase, ProcessorMixin)):
+        tokenizer = patch_saving_functions(tokenizer)
 
     # TorchAO must only use bfloat16 for loading (float16 fails)
     if HAS_TORCH_DTYPE:
@@ -3184,7 +3312,7 @@ def _unsloth_save_torchao_with_given_config(
         quantized_model.save_pretrained(
             torchao_save_directory, safe_serialization = safe_serialization
         )
-        tokenizer.save_pretrained(torchao_save_directory)
+        tokenizer.save_pretrained(torchao_save_directory, token = token)
 
     # Clean up the intermediate unquantized model
     if os.path.exists(save_directory):
@@ -3223,6 +3351,9 @@ def unsloth_save_pretrained_torchao(
       `push_to_hub` (bool): whether to push to huggingface hub or save locally
       `token`: HuggingFace token for pushing to hub
     """
+    if isinstance(tokenizer, (PreTrainedTokenizerBase, ProcessorMixin)):
+        tokenizer = patch_saving_functions(tokenizer)
+
     if token is None and push_to_hub:
         token = get_token()
 
@@ -3342,6 +3473,43 @@ def patch_saving_functions(model, vision = False):
     '''
     exec(push_to_hub_text, globals())
 
+    def unsloth_tokenizer_save_pretrained(
+        self,
+        save_directory,
+        legacy_format = None,
+        filename_prefix = None,
+        push_to_hub = False,
+        **kwargs,
+    ):
+        result = self.original_save_pretrained(
+            save_directory,
+            legacy_format = legacy_format,
+            filename_prefix = filename_prefix,
+            push_to_hub = False,
+            **kwargs,
+        )
+        _preserve_sentencepiece_tokenizer_assets(
+            self,
+            save_directory,
+            token = kwargs.get("token", None),
+        )
+        if push_to_hub:
+            push_kwargs = dict(kwargs)
+            repo_id = push_kwargs.pop("repo_id", save_directory)
+            self.push_to_hub(repo_id, **push_kwargs)
+        return result
+
+    if (
+        isinstance(model, PreTrainedTokenizerBase)
+        and model.save_pretrained.__name__ != "unsloth_tokenizer_save_pretrained"
+    ):
+        model.original_save_pretrained = model.save_pretrained
+        model.save_pretrained = types.MethodType(
+            unsloth_tokenizer_save_pretrained, model
+        )
+    elif getattr(model, "tokenizer", None) is not None:
+        patch_saving_functions(model.tokenizer)
+
     original_model = model
     while True:
         # Check if push_to_hub exists before accessing its __name__

From f9682e656c2f0cb9c6c301e6e523d6da8613ed01 Mon Sep 17 00:00:00 2001
From: Datta Nimmaturi <venkatadattasainimmaturi@gmail.com>
Date: Wed, 22 Apr 2026 21:34:08 +0530
Subject: [PATCH 151/216] update gema4 chat templates (#5116)

* update gema4 chat templates

* udpate template

* update template for gemma4

* Add gemma4 chat template tests

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

---------

Co-authored-by: Daniel Han <danielhanchen@gmail.com>
Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com>
---
 tests/test_gemma4_chat_template.py | 183 +++++++++++++++++++++++++++++
 unsloth/chat_templates.py          |  95 +++++++++++----
 2 files changed, 255 insertions(+), 23 deletions(-)
 create mode 100644 tests/test_gemma4_chat_template.py

diff --git a/tests/test_gemma4_chat_template.py b/tests/test_gemma4_chat_template.py
new file mode 100644
index 0000000000..6211b4d433
--- /dev/null
+++ b/tests/test_gemma4_chat_template.py
@@ -0,0 +1,183 @@
+import os
+import re
+
+import pytest
+from jinja2 import Environment, StrictUndefined
+from jinja2.exceptions import TemplateError
+
+
+CHAT_TEMPLATES_PATH = os.path.join(
+    os.path.dirname(os.path.dirname(os.path.abspath(__file__))),
+    "unsloth",
+    "chat_templates.py",
+)
+
+
+def _extract_template(name):
+    src = open(CHAT_TEMPLATES_PATH).read()
+    pattern = rf'{re.escape(name)}\s*=\s*\\\n"""(.*?)"""'
+    m = re.search(pattern, src, flags = re.DOTALL)
+    assert m, f"Could not extract {name} from chat_templates.py"
+    return m.group(1)
+
+
+def _env():
+    env = Environment(undefined = StrictUndefined, trim_blocks = False, lstrip_blocks = False)
+    env.globals["raise_exception"] = lambda msg: (_ for _ in ()).throw(
+        TemplateError(msg)
+    )
+    return env
+
+
+def _render(template_name, messages, **kwargs):
+    src = _extract_template(template_name)
+    tmpl = _env().from_string(src)
+    ctx = {"messages": messages, "add_generation_prompt": False}
+    ctx.update(kwargs)
+    return tmpl.render(**ctx)
+
+
+# ---------- system turn and <|think|> placement ----------
+
+
+def test_system_message_emits_dedicated_system_turn():
+    msgs = [
+        {"role": "system", "content": "You are helpful"},
+        {"role": "user", "content": "Hi"},
+    ]
+    out = _render("gemma4_template", msgs)
+    assert "<|turn>system\nYou are helpful<turn|>" in out
+    assert "<|turn>user\nHi<turn|>" in out
+    assert "You are helpful\n\nHi" not in out
+
+
+def test_developer_role_treated_as_system():
+    msgs = [
+        {"role": "developer", "content": "Internal instructions"},
+        {"role": "user", "content": "Hi"},
+    ]
+    out = _render("gemma4_template", msgs)
+    assert "<|turn>system\nInternal instructions<turn|>" in out
+
+
+def test_no_system_no_thinking_unchanged():
+    msgs = [{"role": "user", "content": "Hi"}]
+    out = _render("gemma4_template", msgs)
+    assert "<|turn>user\nHi<turn|>" in out
+    assert "<|turn>system" not in out
+
+
+def test_assistant_role_renders_as_model_turn():
+    msgs = [{"role": "user", "content": "Q"}, {"role": "assistant", "content": "A"}]
+    out = _render("gemma4_template", msgs)
+    assert "<|turn>model\nA<turn|>" in out
+    assert "<|turn>assistant" not in out
+
+
+def test_thinking_template_defaults_to_thinking_off_when_unset():
+    msgs = [{"role": "user", "content": "Hi"}]
+    out = _render("gemma4_thinking_template", msgs)
+    assert "<|think|>" not in out
+    assert "<|turn>system" not in out
+
+
+def test_thinking_template_emits_think_with_newline_when_enabled():
+    msgs = [{"role": "system", "content": "Sys"}, {"role": "user", "content": "Hi"}]
+    out = _render("gemma4_thinking_template", msgs, enable_thinking = True)
+    assert "<|turn>system\n<|think|>\nSys<turn|>" in out
+
+
+def test_alternation_violation_raises_template_error():
+    msgs = [{"role": "user", "content": "A"}, {"role": "user", "content": "B"}]
+    with pytest.raises(TemplateError):
+        _render("gemma4_template", msgs)
+
+
+# ---------- strip_thinking macro semantics ----------
+
+
+def test_strip_thinking_strips_matched_pair():
+    msgs = [
+        {"role": "user", "content": "Q"},
+        {
+            "role": "assistant",
+            "content": "<|channel>thought\n2+2=4<channel|>The answer is 4.",
+        },
+    ]
+    out = _render("gemma4_template", msgs)
+    assert "thought" not in out
+    assert "2+2=4" not in out
+    assert "The answer is 4." in out
+
+
+def test_strip_thinking_applied_unconditionally_to_model_turn():
+    msgs = [
+        {"role": "user", "content": "Q"},
+        {"role": "assistant", "content": "<|channel>reasoning<channel|>final"},
+    ]
+    for agp in (True, False):
+        out = _render("gemma4_template", msgs, add_generation_prompt = agp)
+        assert "reasoning" not in out
+        assert "final" in out
+
+
+def test_strip_thinking_applies_to_iterable_text():
+    msgs = [
+        {"role": "user", "content": [{"type": "text", "text": "Q"}]},
+        {
+            "role": "assistant",
+            "content": [{"type": "text", "text": "<|channel>r<channel|>final"}],
+        },
+    ]
+    out = _render("gemma4_thinking_template", msgs)
+    assert "final" in out
+    assert "<|channel>" not in out
+
+
+def test_strip_thinking_preserves_plain_text():
+    msgs = [
+        {"role": "user", "content": "Q"},
+        {"role": "assistant", "content": "plain answer with no markup"},
+    ]
+    out = _render("gemma4_template", msgs, add_generation_prompt = True)
+    assert "plain answer with no markup" in out
+
+
+def test_multi_turn_strips_all_historical_model_turns():
+    msgs = [
+        {"role": "user", "content": "Q1"},
+        {"role": "assistant", "content": "<|channel>r1<channel|>A1"},
+        {"role": "user", "content": "Q2"},
+        {"role": "assistant", "content": "<|channel>r2<channel|>A2"},
+    ]
+    out = _render("gemma4_thinking_template", msgs, add_generation_prompt = True)
+    assert "r1" not in out and "r2" not in out
+    assert "A1" in out and "A2" in out
+
+
+# ---------- thinking-template gen-prompt injection ----------
+
+
+def test_thinking_template_injects_empty_thought_channel_by_default():
+    # Author defaults enable_thinking=False, so the gen-prompt injection fires.
+    msgs = [{"role": "user", "content": "Hi"}]
+    out = _render("gemma4_thinking_template", msgs, add_generation_prompt = True)
+    assert out.endswith("<|turn>model\n<|channel>thought\n<channel|>")
+
+
+def test_thinking_template_no_injection_when_thinking_enabled():
+    msgs = [{"role": "user", "content": "Hi"}]
+    out = _render(
+        "gemma4_thinking_template",
+        msgs,
+        add_generation_prompt = True,
+        enable_thinking = True,
+    )
+    assert "<|channel>thought" not in out
+
+
+def test_base_template_has_no_channel_thought_injection():
+    msgs = [{"role": "user", "content": "Hi"}]
+    out = _render("gemma4_template", msgs, add_generation_prompt = True)
+    assert out.endswith("<|turn>model\n")
+    assert "<|channel>thought" not in out
diff --git a/unsloth/chat_templates.py b/unsloth/chat_templates.py
index 326fd59289..8376dc7e39 100644
--- a/unsloth/chat_templates.py
+++ b/unsloth/chat_templates.py
@@ -866,12 +866,29 @@ DEFAULT_SYSTEM_MESSAGE["gemma3n"] = None # No system message in Gemma-3n
 # =========================================== Gemma-4
 # Gemma-4 uses <|turn>role\n...<turn|>\n format
 gemma4_template = \
-"""{%- if messages[0]['role'] == 'system' -%}
-    {%- set first_user_prefix = messages[0]['content'] + '\n\n' -%}
-    {%- set loop_messages = messages[1:] -%}
-{%- else -%}
-    {%- set first_user_prefix = "" -%}
-    {%- set loop_messages = messages -%}
+"""{%- macro strip_thinking(text) -%}
+    {%- set ns = namespace(result='') -%}
+    {%- for part in text.split('<channel|>') -%}
+        {%- if '<|channel>' in part -%}
+            {%- set ns.result = ns.result + part.split('<|channel>')[0] -%}
+        {%- else -%}
+            {%- set ns.result = ns.result + part -%}
+        {%- endif -%}
+    {%- endfor -%}
+    {{- ns.result | trim -}}
+{%- endmacro -%}
+{%- set thinking = enable_thinking is defined and enable_thinking -%}
+{%- set loop_messages = messages -%}
+{%- if messages[0]['role'] in ['system', 'developer'] or thinking -%}
+    {{ '<|turn>system\n' }}
+    {%- if thinking -%}
+        {{ '<|think|>\n' }}
+    {%- endif -%}
+    {%- if messages[0]['role'] in ['system', 'developer'] -%}
+        {{ messages[0]['content'] | trim }}
+        {%- set loop_messages = messages[1:] -%}
+    {%- endif -%}
+    {{ '<turn|>\n' }}
 {%- endif -%}
 {%- for message in loop_messages -%}
     {%- if (message['role'] == 'user') != (loop.index0 % 2 == 0) -%}
@@ -882,9 +899,13 @@ gemma4_template = \
     {%- else -%}
         {%- set role = message['role'] -%}
     {%- endif -%}
-    {{ '<|turn>' + role + '\n' + (first_user_prefix if loop.first else "") }}
+    {{ '<|turn>' + role + '\n' }}
     {%- if message['content'] is string -%}
-        {{ message['content'] | trim }}
+        {%- if role == "model" -%}
+            {{ strip_thinking(message['content']) }}
+        {%- else -%}
+            {{ message['content'] | trim }}
+        {%- endif -%}
     {%- elif message['content'] is iterable -%}
         {%- for item in message['content'] -%}
             {%- if item['type'] == 'audio' -%}
@@ -894,7 +915,11 @@ gemma4_template = \
             {%- elif item['type'] == 'video' -%}
                 {{ '<|video|>' }}
             {%- elif item['type'] == 'text' -%}
-                {{ item['text'] | trim }}
+                {%- if role == "model" -%}
+                    {{ strip_thinking(item['text']) }}
+                {%- else -%}
+                    {{ item['text'] | trim }}
+                {%- endif -%}
             {%- endif -%}
         {%- endfor -%}
     {%- else -%}
@@ -918,15 +943,31 @@ DEFAULT_SYSTEM_MESSAGE["gemma-4"] = None
 CHAT_TEMPLATES["gemma4"] = (gemma4_template, gemma4_template_eos_token, False, gemma4_ollama,)
 DEFAULT_SYSTEM_MESSAGE["gemma4"] = None
 
-# Gemma-4 with empty thought channel (required for larger models like 31B, 26B-A4B)
-# Injects <|channel>thought\n<channel|> at the start of each model response during training
+# Gemma-4 thinking template
 gemma4_thinking_template = \
-"""{%- if messages[0]['role'] == 'system' -%}
-    {%- set first_user_prefix = messages[0]['content'] + '\n\n' -%}
-    {%- set loop_messages = messages[1:] -%}
-{%- else -%}
-    {%- set first_user_prefix = "" -%}
-    {%- set loop_messages = messages -%}
+"""{%- macro strip_thinking(text) -%}
+    {%- set ns = namespace(result='') -%}
+    {%- for part in text.split('<channel|>') -%}
+        {%- if '<|channel>' in part -%}
+            {%- set ns.result = ns.result + part.split('<|channel>')[0] -%}
+        {%- else -%}
+            {%- set ns.result = ns.result + part -%}
+        {%- endif -%}
+    {%- endfor -%}
+    {{- ns.result | trim -}}
+{%- endmacro -%}
+{%- set thinking = enable_thinking is defined and enable_thinking -%}
+{%- set loop_messages = messages -%}
+{%- if messages[0]['role'] in ['system', 'developer'] or thinking -%}
+    {{ '<|turn>system\n' }}
+    {%- if thinking -%}
+        {{ '<|think|>\n' }}
+    {%- endif -%}
+    {%- if messages[0]['role'] in ['system', 'developer'] -%}
+        {{ messages[0]['content'] | trim }}
+        {%- set loop_messages = messages[1:] -%}
+    {%- endif -%}
+    {{ '<turn|>\n' }}
 {%- endif -%}
 {%- for message in loop_messages -%}
     {%- if (message['role'] == 'user') != (loop.index0 % 2 == 0) -%}
@@ -937,12 +978,13 @@ gemma4_thinking_template = \
     {%- else -%}
         {%- set role = message['role'] -%}
     {%- endif -%}
-    {{ '<|turn>' + role + '\n' + (first_user_prefix if loop.first else "") }}
-    {%- if role == "model" -%}
-        {{ '<|channel>thought\n<channel|>' }}
-    {%- endif -%}
+    {{ '<|turn>' + role + '\n' }}
     {%- if message['content'] is string -%}
-        {{ message['content'] | trim }}
+        {%- if role == "model" -%}
+            {{ strip_thinking(message['content']) }}
+        {%- else -%}
+            {{ message['content'] | trim }}
+        {%- endif -%}
     {%- elif message['content'] is iterable -%}
         {%- for item in message['content'] -%}
             {%- if item['type'] == 'audio' -%}
@@ -952,7 +994,11 @@ gemma4_thinking_template = \
             {%- elif item['type'] == 'video' -%}
                 {{ '<|video|>' }}
             {%- elif item['type'] == 'text' -%}
-                {{ item['text'] | trim }}
+                {%- if role == "model" -%}
+                    {{ strip_thinking(item['text']) }}
+                {%- else -%}
+                    {{ item['text'] | trim }}
+                {%- endif -%}
             {%- endif -%}
         {%- endfor -%}
     {%- else -%}
@@ -962,6 +1008,9 @@ gemma4_thinking_template = \
 {%- endfor -%}
 {%- if add_generation_prompt -%}
     {{'<|turn>model\n'}}
+    {%- if not thinking -%}
+        {{ '<|channel>thought\n<channel|>' }}
+    {%- endif -%}
 {%- endif -%}
 """
 

From 71014f4f4ef7bc597f2e10a411c224f23747df71 Mon Sep 17 00:00:00 2001
From: Daniel Han <danielhanchen@gmail.com>
Date: Wed, 22 Apr 2026 09:17:50 -0700
Subject: [PATCH 152/216] Update _utils.py

---
 unsloth/models/_utils.py | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/unsloth/models/_utils.py b/unsloth/models/_utils.py
index 34450549c5..2e7e17bbb9 100644
--- a/unsloth/models/_utils.py
+++ b/unsloth/models/_utils.py
@@ -12,7 +12,7 @@
 # See the License for the specific language governing permissions and
 # limitations under the License.
 
-__version__ = "2026.4.6"
+__version__ = "2026.4.7"
 
 __all__ = [
     "SUPPORTS_BFLOAT16",

From 2bd6d544ff458577a0202bcd64ad664d4f1cec08 Mon Sep 17 00:00:00 2001
From: Daniel Han <danielhanchen@gmail.com>
Date: Wed, 22 Apr 2026 09:28:47 -0700
Subject: [PATCH 153/216] Bump installer floor to 2026.4.7 (#5134)

---
 install.ps1 | 10 +++++-----
 install.sh  | 10 +++++-----
 2 files changed, 10 insertions(+), 10 deletions(-)

diff --git a/install.ps1 b/install.ps1
index 3fc9ac4690..7f98fc5cb8 100644
--- a/install.ps1
+++ b/install.ps1
@@ -912,7 +912,7 @@ shell.Run cmd, 0, False
         if ($SkipTorch) {
             # No-torch: install unsloth + unsloth-zoo with --no-deps, then
             # runtime deps (typer, safetensors, transformers, etc.) with --no-deps.
-            $baseInstallExit = Invoke-InstallCommand { uv pip install --python $VenvPython --no-deps --reinstall-package unsloth --reinstall-package unsloth-zoo "unsloth>=2026.4.5" unsloth-zoo }
+            $baseInstallExit = Invoke-InstallCommand { uv pip install --python $VenvPython --no-deps --reinstall-package unsloth --reinstall-package unsloth-zoo "unsloth>=2026.4.7" unsloth-zoo }
             if ($baseInstallExit -eq 0) {
                 $NoTorchReq = Find-NoTorchRuntimeFile
                 if ($NoTorchReq) {
@@ -920,7 +920,7 @@ shell.Run cmd, 0, False
                 }
             }
         } else {
-            $baseInstallExit = Invoke-InstallCommand { uv pip install --python $VenvPython --reinstall-package unsloth --reinstall-package unsloth-zoo "unsloth>=2026.4.5" unsloth-zoo }
+            $baseInstallExit = Invoke-InstallCommand { uv pip install --python $VenvPython --reinstall-package unsloth --reinstall-package unsloth-zoo "unsloth>=2026.4.7" unsloth-zoo }
         }
         if ($baseInstallExit -ne 0) {
             Write-Host "[ERROR] Failed to install unsloth (exit code $baseInstallExit)" -ForegroundColor Red
@@ -950,7 +950,7 @@ shell.Run cmd, 0, False
         if ($SkipTorch) {
             # No-torch: install unsloth + unsloth-zoo with --no-deps, then
             # runtime deps (typer, safetensors, transformers, etc.) with --no-deps.
-            $baseInstallExit = Invoke-InstallCommand { uv pip install --python $VenvPython --no-deps --upgrade-package unsloth --upgrade-package unsloth-zoo "unsloth>=2026.4.5" unsloth-zoo }
+            $baseInstallExit = Invoke-InstallCommand { uv pip install --python $VenvPython --no-deps --upgrade-package unsloth --upgrade-package unsloth-zoo "unsloth>=2026.4.7" unsloth-zoo }
             if ($baseInstallExit -eq 0) {
                 $NoTorchReq = Find-NoTorchRuntimeFile
                 if ($NoTorchReq) {
@@ -958,7 +958,7 @@ shell.Run cmd, 0, False
                 }
             }
         } elseif ($StudioLocalInstall) {
-            $baseInstallExit = Invoke-InstallCommand { uv pip install --python $VenvPython --upgrade-package unsloth "unsloth>=2026.4.5" unsloth-zoo }
+            $baseInstallExit = Invoke-InstallCommand { uv pip install --python $VenvPython --upgrade-package unsloth "unsloth>=2026.4.7" unsloth-zoo }
         } else {
             $baseInstallExit = Invoke-InstallCommand { uv pip install --python $VenvPython --upgrade-package unsloth "$PackageName" }
         }
@@ -979,7 +979,7 @@ shell.Run cmd, 0, False
         # Fallback: GPU detection failed to produce a URL -- let uv resolve torch
         substep "installing unsloth (this may take a few minutes)..."
         if ($StudioLocalInstall) {
-            $baseInstallExit = Invoke-InstallCommand { uv pip install --python $VenvPython unsloth-zoo "unsloth>=2026.4.5" --torch-backend=auto }
+            $baseInstallExit = Invoke-InstallCommand { uv pip install --python $VenvPython unsloth-zoo "unsloth>=2026.4.7" --torch-backend=auto }
             if ($baseInstallExit -ne 0) {
                 Write-Host "[ERROR] Failed to install unsloth (exit code $baseInstallExit)" -ForegroundColor Red
                 return
diff --git a/install.sh b/install.sh
index 6915893ddf..0352887de1 100755
--- a/install.sh
+++ b/install.sh
@@ -1316,7 +1316,7 @@ if [ "$_MIGRATED" = true ]; then
         # to prevent transitive torch resolution.
         run_install_cmd "install unsloth (migrated no-torch)" uv pip install --python "$_VENV_PY" --no-deps \
             --reinstall-package unsloth --reinstall-package unsloth-zoo \
-            "unsloth>=2026.4.5" unsloth-zoo
+            "unsloth>=2026.4.7" unsloth-zoo
         _NO_TORCH_RT="$(_find_no_torch_runtime)"
         if [ -n "$_NO_TORCH_RT" ]; then
             run_install_cmd "install no-torch runtime deps" uv pip install --python "$_VENV_PY" --no-deps -r "$_NO_TORCH_RT"
@@ -1324,7 +1324,7 @@ if [ "$_MIGRATED" = true ]; then
     else
         run_install_cmd "install unsloth (migrated)" uv pip install --python "$_VENV_PY" \
             --reinstall-package unsloth --reinstall-package unsloth-zoo \
-            "unsloth>=2026.4.5" unsloth-zoo
+            "unsloth>=2026.4.7" unsloth-zoo
     fi
     if [ "$STUDIO_LOCAL_INSTALL" = true ]; then
         substep "overlaying local repo (editable)..."
@@ -1487,7 +1487,7 @@ elif [ -n "$TORCH_INDEX_URL" ]; then
         # runtime deps (typer, safetensors, transformers, etc.) with --no-deps.
         run_install_cmd "install unsloth (no-torch)" uv pip install --python "$_VENV_PY" --no-deps \
             --upgrade-package unsloth --upgrade-package unsloth-zoo \
-            "unsloth>=2026.4.5" unsloth-zoo
+            "unsloth>=2026.4.7" unsloth-zoo
         _NO_TORCH_RT="$(_find_no_torch_runtime)"
         if [ -n "$_NO_TORCH_RT" ]; then
             run_install_cmd "install no-torch runtime deps" uv pip install --python "$_VENV_PY" --no-deps -r "$_NO_TORCH_RT"
@@ -1498,7 +1498,7 @@ elif [ -n "$TORCH_INDEX_URL" ]; then
         fi
     elif [ "$STUDIO_LOCAL_INSTALL" = true ]; then
         run_install_cmd "install unsloth (local)" uv pip install --python "$_VENV_PY" \
-            --upgrade-package unsloth "unsloth>=2026.4.5" unsloth-zoo
+            --upgrade-package unsloth "unsloth>=2026.4.7" unsloth-zoo
         substep "overlaying local repo (editable)..."
         run_install_cmd "overlay local repo" uv pip install --python "$_VENV_PY" -e "$_REPO_ROOT" --no-deps
     else
@@ -1525,7 +1525,7 @@ else
     # Fallback: GPU detection failed to produce a URL -- let uv resolve torch
     substep "installing unsloth (this may take a few minutes)..."
     if [ "$STUDIO_LOCAL_INSTALL" = true ]; then
-        run_install_cmd "install unsloth (auto torch backend)" uv pip install --python "$_VENV_PY" unsloth-zoo "unsloth>=2026.4.5" --torch-backend=auto
+        run_install_cmd "install unsloth (auto torch backend)" uv pip install --python "$_VENV_PY" unsloth-zoo "unsloth>=2026.4.7" --torch-backend=auto
         substep "overlaying local repo (editable)..."
         run_install_cmd "overlay local repo" uv pip install --python "$_VENV_PY" -e "$_REPO_ROOT" --no-deps
     else

From a1fb7c1297bd534df7bd7a58672f19cbe05edc6d Mon Sep 17 00:00:00 2001
From: DoubleMathew <mmathew23@gmail.com>
Date: Wed, 22 Apr 2026 15:15:42 -0500
Subject: [PATCH 154/216] fix/llamacpp_prebuilt_install (#5135)

* fix/llamacpp_prebuilt_installinclude libllama-common.so in bundle as its needed by llama.cpp now

* include support for rocm linux
---
 studio/install_llama_prebuilt.py | 4 ++++
 1 file changed, 4 insertions(+)

diff --git a/studio/install_llama_prebuilt.py b/studio/install_llama_prebuilt.py
index fd2b78d05f..1ad1ccc851 100755
--- a/studio/install_llama_prebuilt.py
+++ b/studio/install_llama_prebuilt.py
@@ -3640,6 +3640,7 @@ def runtime_patterns_for_choice(choice: AssetChoice) -> list[str]:
         return [
             "llama-server",
             "llama-quantize",
+            "libllama-common.so*",
             "libllama.so*",
             "libggml.so*",
             "libggml-base.so*",
@@ -4929,6 +4930,7 @@ def load_prebuilt_metadata(install_dir: Path) -> dict[str, Any] | None:
 def runtime_payload_health_groups(choice: AssetChoice) -> list[list[str]]:
     if choice.install_kind == "linux-cpu":
         return [
+            ["libllama-common.so*"],
             ["libllama.so*"],
             ["libggml.so*"],
             ["libggml-base.so*"],
@@ -4937,6 +4939,7 @@ def runtime_payload_health_groups(choice: AssetChoice) -> list[list[str]]:
         ]
     if choice.install_kind == "linux-cuda":
         return [
+            ["libllama-common.so*"],
             ["libllama.so*"],
             ["libggml.so*"],
             ["libggml-base.so*"],
@@ -4952,6 +4955,7 @@ def runtime_payload_health_groups(choice: AssetChoice) -> list[list[str]]:
         ]
     if choice.install_kind == "linux-rocm":
         return [
+            ["libllama-common.so*"],
             ["libllama.so*"],
             ["libggml.so*"],
             ["libggml-base.so*"],

From 41a6cc8692bd47ba0e4590b45bdf88d24b15ef33 Mon Sep 17 00:00:00 2001
From: Daniel Han <danielhanchen@gmail.com>
Date: Thu, 23 Apr 2026 00:22:40 -0700
Subject: [PATCH 155/216] Studio: split vision-cache exception test to match
 transient vs permanent (#5145)

`TestVisionCacheOnException::test_exception_result_cached` currently
patches `load_model_config` with `side_effect=OSError("network down")`
and asserts `assert_called_once()`. That assertion is impossible by
design: `_is_vision_model_uncached` in
`studio/backend/utils/models/model_config.py` intentionally returns
`None` for `OSError` so `is_vision_model` does not cache the fallback
and retries on the next call. The module docstring on
`_vision_detection_cache` itself spells this out:

    Only definitive results (True/False from successful detection) are
    cached; transient failures (network errors, timeouts) are NOT
    cached so they can be retried.

The test has been failing identically on every downstream review run
against `unslothai/unsloth` main (e.g. `unsloth#5115`, `unsloth#5080`),
but the failure is not introduced by any of those PRs and does not
gate correctness.

Fix the collision by splitting the class into the two contracts the
code actually implements:

1. `test_permanent_exception_result_cached` keeps the original
   intent ("exception falls back to False and that False is cached")
   but uses `ValueError`, which is one of the exception types
   `_is_vision_model_uncached` treats as permanent and caches. No
   `huggingface_hub` import needed.

2. `test_transient_exception_not_cached` pins the opposite contract
   with the original `OSError("network down")`: the call returns
   False but the second invocation re-runs detection
   (`call_count == 2`). This guards against a future regression
   where somebody caches transient failures and then users with a
   flaky network permanently see wrong detection for a model.

Both tests use `assert ... is False` on the public API and mock-count
assertions on `load_model_config`; no private helpers are touched.
---
 studio/backend/tests/test_vision_cache.py | 45 ++++++++++++++++++-----
 1 file changed, 36 insertions(+), 9 deletions(-)

diff --git a/studio/backend/tests/test_vision_cache.py b/studio/backend/tests/test_vision_cache.py
index fae1e95311..9e7bbdd1fb 100644
--- a/studio/backend/tests/test_vision_cache.py
+++ b/studio/backend/tests/test_vision_cache.py
@@ -124,23 +124,50 @@ class TestVisionCacheSubprocessPath:
 
 
 class TestVisionCacheOnException:
-    """When detection raises an exception, _is_vision_model_uncached catches
-    it and returns False.  That False must be cached so subsequent calls don't
-    retry and fail again."""
+    """When detection raises an exception, _is_vision_model_uncached
+    distinguishes permanent failures (cached as False) from transient
+    failures (returned as None, not cached so the next call can retry).
+    Verify both contracts."""
+
+    @patch(
+        "utils.models.model_config.load_model_config",
+        side_effect = ValueError("bad config"),
+    )
+    @patch("utils.transformers_version.needs_transformers_5", return_value = False)
+    def test_permanent_exception_result_cached(self, mock_needs_t5, mock_load_config):
+        """A permanent failure (ValueError / RepositoryNotFoundError /
+        GatedRepoError / JSONDecodeError) should be caught, return False,
+        and that False should be cached so subsequent calls don't retry.
+
+        ValueError is used here because it's the simplest of the
+        code-path's cacheable exception types and does not require an
+        import of huggingface_hub errors (whose module path varies
+        across versions)."""
+        # First call: load_model_config raises -> except branch -> False.
+        assert is_vision_model("broken/model") is False
+        # Second call: cache hit, load_model_config not called again.
+        assert is_vision_model("broken/model") is False
+        mock_load_config.assert_called_once()
 
     @patch(
         "utils.models.model_config.load_model_config",
         side_effect = OSError("network down"),
     )
     @patch("utils.transformers_version.needs_transformers_5", return_value = False)
-    def test_exception_result_cached(self, mock_needs_t5, mock_load_config):
-        """A real exception inside _is_vision_model_uncached should be caught,
-        return False, and that False should be cached for subsequent calls."""
-        # First call: load_model_config raises → except branch → False
+    def test_transient_exception_not_cached(self, mock_needs_t5, mock_load_config):
+        """A transient failure (OSError, timeouts) should return None from
+        _is_vision_model_uncached, surface as False to the caller, and
+        NOT be cached, so the next call retries detection.  This matches
+        the documented behaviour on _vision_detection_cache:
+        'transient failures (network errors, timeouts) are NOT cached so
+        they can be retried.'"""
+        # First call: load_model_config raises OSError -> uncached None
+        # -> caller returns False without caching.
         assert is_vision_model("broken/model") is False
-        # Second call: cache hit, load_model_config not called again
+        # Second call: cache miss again, load_model_config called a
+        # second time.
         assert is_vision_model("broken/model") is False
-        mock_load_config.assert_called_once()
+        assert mock_load_config.call_count == 2
 
 
 # ---------------------------------------------------------------------------

From 72c1c3b254cdd6034c01f4d0ec0e4866ad7e6ee6 Mon Sep 17 00:00:00 2001
From: Octopus <liyuan851277048@icloud.com>
Date: Thu, 23 Apr 2026 18:11:35 +0800
Subject: [PATCH 156/216] fix: patch CONTROL type for special tokens in
 sentencepiece GGUF export (#5080)

* fix: patch CONTROL type for special tokens in sentencepiece GGUF export (fixes #5070)

When converting a Gemma 3 fine-tune to GGUF via save_pretrained_gguf,
tokens like <start_of_turn> (id=105) and <end_of_turn> (id=106) are
already present in the sentencepiece model but are typed as NORMAL (1)
instead of CONTROL (3). llama.cpp only recognises CONTROL tokens when
parse_special=True is active, so these tokens get BPE-split during
chat inference and the model produces garbage output.

fix_sentencepiece_gguf now reads tokenizer.json's added_tokens list and,
for any token with "special": true whose ID falls within the existing
sentencepiece vocabulary, updates its type from NORMAL to CONTROL before
writing the patched tokenizer.model to disk. The same CONTROL type is
also applied when new tokens are appended for the out-of-range case, so
both code paths are consistent.

* Wire fix_sentencepiece_gguf into tokenizer save path and guard np.diff

- save.py: call fix_sentencepiece_gguf inside unsloth_tokenizer_save_pretrained
  after _preserve_sentencepiece_tokenizer_assets. The helper was previously
  unreferenced in the repo, so the PR's CONTROL-type patch never actually ran
  during save_pretrained_gguf.
- tokenizer_utils.py: add an early-return guard for len(added_tokens_ids) < 2
  before the existing np.diff contiguity check. np.diff on a single-element
  array returns [] and .min() raises ValueError, which would discard the new
  in-vocab CONTROL patch; the guard flushes tokenizer.model first. Guard is
  inserted before the existing lines (diff = np.diff(...) and the min/max
  check) so their blame is unchanged.

Dropped the separate refactor to fold the four duplicated "if patched > 0:
write tokenizer.model" blocks into a helper because doing so re-indents
lines whose blame is "Formatting & bug fixes"; the duplication
remains the author's pattern.

* Fix review findings: negative token_id guard and np.diff single-element

- tokenizer_utils.py:481: add 0 <= lower bound to the special_token_ids
  bounds check. Previously a negative token_id from tokenizer.json passed
  'token_id < sentence_piece_size' and Python's negative indexing wrapped
  tokenizer_file.pieces[-1] to silently corrupt the last piece to CONTROL.
- tokenizer_utils.py:513: replace the loop-1 'if len < 2: return' guard
  (which was too broad: it silently skipped vocab extension for single-entry
  added_tokens.json) with a pre-pass that substitutes a trivially-contiguous
  2-element sentinel for the contiguity check, then restores the original
  array before the append loop. Lines 519 ('diff = np.diff(added_tokens_ids)')
  and 520-529 (min/max/boundary checks and early-return write blocks) are
  left literally unchanged so blame remains intact.

* Restore real added_tokens_ids before min boundary check

Move the '_real_added_tokens_ids' restore above the
'added_tokens_ids.min() != sentence_piece_size' check. With the previous
order the sentinel [sentence_piece_size, sentence_piece_size + 1] was
still in scope when the min check ran, so any single-entry added_tokens
.json with an out-of-range start id (e.g. 99 when sentence_piece_size=2)
bypassed the boundary check and fell through to the append loop.

* Scope fix_sentencepiece_gguf to GGUF export path only

Previously wired fix_sentencepiece_gguf into unsloth_tokenizer_save_pretrained,
which is the generic monkey-patch replacement for every tokenizer.save_pretrained
call. That caused the GGUF-specific mutation (and the unconditional protobuf
import in fix_sentencepiece_gguf) to run on every LoRA / merged 16-bit /
push_to_hub / torchao save, where it has no purpose and can abort the entire
save if the protobuf runtime is unavailable.

- save.py: remove fix_sentencepiece_gguf call from unsloth_tokenizer_save_pretrained.
- save.py: add the call inside unsloth_save_pretrained_gguf immediately before
  save_to_gguf, wrapped in try/except so a protobuf import failure logs a
  warning and lets GGUF conversion proceed rather than aborting the save.

* Broaden special-token retag to USER_DEFINED and narrow save.py except

- tokenizer_utils.py:483: the in-vocab retag previously only promoted NORMAL
  pieces to CONTROL, but the real Gemma tokenizer (e.g. unsloth/functiongemma
  -270m-it) stores <start_of_turn>/<end_of_turn> as USER_DEFINED (type 4).
  Extend the predicate to cover both NORMAL and USER_DEFINED so tokens marked
  "special": true in tokenizer.json are promoted regardless of their current
  sentencepiece type. Only tokens explicitly flagged special are touched, so
  non-special USER_DEFINED pieces are unchanged; already-CONTROL pieces stay
  unchanged. The warning message is generalised accordingly.
- save.py:2294: narrow the except clause from Exception to ImportError. The
  loop-3 try/except was added to tolerate a missing protobuf runtime; leaving
  it broad also swallows OSError/PermissionError mid-write, which would ship
  a corrupted tokenizer.model to save_to_gguf. ImportError still covers the
  protobuf case while letting I/O errors propagate to the outer save handler.

* Harden fix_sentencepiece_gguf: widen except, protobuf fallback, revert USER_DEFINED widen, guard entry id

- save.py:2294: widen except from ImportError back to Exception. The loop-4
  narrowing let JSONDecodeError / KeyError / OSError / PermissionError from
  fix_sentencepiece_gguf abort the entire GGUF export, a regression vs
  pre-PR behavior. The outer save_to_gguf try/except still covers GGUF-side
  failures; any fix-side failure now logs a typed warning and lets
  conversion proceed.
- tokenizer_utils.py:445: the direct 'from transformers.utils import
  sentencepiece_model_pb2' raises TypeError ("Descriptors cannot be created
  directly") on modern protobuf runtimes. Prepend a sys.modules.setdefault
  pre-population using transformers.convert_slow_tokenizer.import_protobuf()
  so the subsequent from-import finds a compatible module via the module
  cache. The original import line is left verbatim at its place as the
  final resolver.
- tokenizer_utils.py:483: revert loop-4 widening; retag only NORMAL pieces
  to CONTROL. Retagging USER_DEFINED pieces caused a concrete tokenization
  regression where an intentionally-USER_DEFINED in-vocab special token had
  its sentencepiece encoding broken ('<user> hello' changed from [11, 3, 8]
  to [11, 0, 12, 21, 0, 8]). The PR's stated scope is the NORMAL->CONTROL
  Gemma case; USER_DEFINED handling is deferred.
- tokenizer_utils.py:475: defensive guard around entry["id"]. A malformed
  added_tokens entry missing the "id" field or with a non-int id is now
  skipped rather than raising KeyError / inserting garbage.

* Add review tests for sentencepiece GGUF fix

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

---------

Co-authored-by: octo-patch <octo-patch@github.com>
Co-authored-by: Lee Jackson <130007945+Imagineer99@users.noreply.github.com>
Co-authored-by: Daniel Han <danielhanchen@gmail.com>
Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com>
---
 .../test_fix_sentencepiece_gguf_robustness.py | 132 ++++++++++++++++++
 unsloth/save.py                               |   9 ++
 unsloth/tokenizer_utils.py                    |  72 +++++++++-
 3 files changed, 210 insertions(+), 3 deletions(-)
 create mode 100644 tests/saving/test_fix_sentencepiece_gguf_robustness.py

diff --git a/tests/saving/test_fix_sentencepiece_gguf_robustness.py b/tests/saving/test_fix_sentencepiece_gguf_robustness.py
new file mode 100644
index 0000000000..49bd70fa2f
--- /dev/null
+++ b/tests/saving/test_fix_sentencepiece_gguf_robustness.py
@@ -0,0 +1,132 @@
+import ast
+import json
+import os
+
+os.environ.setdefault("PROTOCOL_BUFFERS_PYTHON_IMPLEMENTATION", "python")
+
+from transformers.utils import sentencepiece_model_pb2
+
+from unsloth.tokenizer_utils import fix_sentencepiece_gguf
+
+
+NORMAL, CONTROL, USER_DEFINED = 1, 3, 4
+
+_SAVE_PY = os.path.abspath(
+    os.path.join(os.path.dirname(__file__), "..", "..", "unsloth", "save.py")
+)
+_TOK_PY = os.path.abspath(
+    os.path.join(os.path.dirname(__file__), "..", "..", "unsloth", "tokenizer_utils.py")
+)
+
+
+def _build(pieces):
+    m = sentencepiece_model_pb2.ModelProto()
+    for piece, score, typ in pieces:
+        p = m.pieces.add()
+        p.piece = piece
+        p.score = score
+        p.type = typ
+    return m.SerializeToString()
+
+
+def _read(path):
+    m = sentencepiece_model_pb2.ModelProto()
+    with open(path, "rb") as f:
+        m.ParseFromString(f.read())
+    return [(p.piece, p.type) for p in m.pieces]
+
+
+def test_user_defined_special_piece_is_not_retyped(tmp_path):
+    pieces = [
+        ("<s>", 0.0, CONTROL),
+        ("a", -1.0, NORMAL),
+        ("<ud_special>", -1.0, USER_DEFINED),
+    ]
+    (tmp_path / "tokenizer.model").write_bytes(_build(pieces))
+    (tmp_path / "tokenizer.json").write_text(
+        json.dumps(
+            {"added_tokens": [{"id": 2, "content": "<ud_special>", "special": True}]}
+        )
+    )
+    fix_sentencepiece_gguf(str(tmp_path))
+    got = dict(_read(str(tmp_path / "tokenizer.model")))
+    assert got["<ud_special>"] == USER_DEFINED
+
+
+def test_malformed_entry_missing_id_does_not_raise(tmp_path):
+    pieces = [("<s>", 0.0, CONTROL), ("a", -1.0, NORMAL), ("<sot>", -1.0, NORMAL)]
+    (tmp_path / "tokenizer.model").write_bytes(_build(pieces))
+    (tmp_path / "tokenizer.json").write_text(
+        json.dumps(
+            {
+                "added_tokens": [
+                    {"content": "no_id_entry", "special": True},
+                    {"id": 2, "content": "<sot>", "special": True},
+                ]
+            }
+        )
+    )
+    fix_sentencepiece_gguf(str(tmp_path))
+    got = dict(_read(str(tmp_path / "tokenizer.model")))
+    assert got["<sot>"] == CONTROL
+
+
+def test_entry_with_non_int_id_is_skipped(tmp_path):
+    pieces = [("<s>", 0.0, CONTROL), ("a", -1.0, NORMAL)]
+    (tmp_path / "tokenizer.model").write_bytes(_build(pieces))
+    (tmp_path / "tokenizer.json").write_text(
+        json.dumps({"added_tokens": [{"id": "oops", "content": "x", "special": True}]})
+    )
+    before = (tmp_path / "tokenizer.model").read_bytes()
+    fix_sentencepiece_gguf(str(tmp_path))
+    after = (tmp_path / "tokenizer.model").read_bytes()
+    assert before == after
+
+
+def test_save_py_except_clause_is_broad_exception():
+    with open(_SAVE_PY) as f:
+        tree = ast.parse(f.read())
+    for node in ast.walk(tree):
+        if (
+            isinstance(node, ast.FunctionDef)
+            and node.name == "unsloth_save_pretrained_gguf"
+        ):
+            for subnode in ast.walk(node):
+                if isinstance(subnode, ast.Try):
+                    body_src = "\n".join(ast.unparse(s) for s in subnode.body)
+                    if "fix_sentencepiece_gguf(" not in body_src:
+                        continue
+                    handler = subnode.handlers[0]
+                    assert handler.type is not None
+                    assert isinstance(handler.type, ast.Name)
+                    assert handler.type.id == "Exception"
+                    return
+    raise AssertionError(
+        "fix_sentencepiece_gguf try block not found in unsloth_save_pretrained_gguf"
+    )
+
+
+def test_tokenizer_utils_uses_import_protobuf_fallback_pattern():
+    with open(_TOK_PY) as f:
+        src = f.read()
+    tree = ast.parse(src)
+    for node in ast.walk(tree):
+        if isinstance(node, ast.FunctionDef) and node.name == "fix_sentencepiece_gguf":
+            fn_src = ast.unparse(node)
+            assert "import_protobuf" in fn_src
+            return
+    raise AssertionError("fix_sentencepiece_gguf not found in tokenizer_utils.py")
+
+
+def test_all_special_tokens_are_gated_by_tokenizer_json_not_by_type(tmp_path):
+    pieces = [
+        ("<s>", 0.0, CONTROL),
+        ("a", -1.0, NORMAL),
+        ("<n1>", -1.0, NORMAL),
+        ("<u1>", -1.0, USER_DEFINED),
+    ]
+    (tmp_path / "tokenizer.model").write_bytes(_build(pieces))
+    fix_sentencepiece_gguf(str(tmp_path))
+    got = dict(_read(str(tmp_path / "tokenizer.model")))
+    assert got["<n1>"] == NORMAL
+    assert got["<u1>"] == USER_DEFINED
diff --git a/unsloth/save.py b/unsloth/save.py
index 4e9e1bfe11..df3df13a04 100644
--- a/unsloth/save.py
+++ b/unsloth/save.py
@@ -2288,6 +2288,15 @@ def unsloth_save_pretrained_gguf(
                 quant_method = "q8_0"
             quantization_methods.append(quant_method.lower())
 
+    try:
+        from .tokenizer_utils import fix_sentencepiece_gguf
+
+        fix_sentencepiece_gguf(save_directory)
+    except Exception as e:
+        logger.warning(
+            f"Unsloth: fix_sentencepiece_gguf skipped ({type(e).__name__}): {e}"
+        )
+
     try:
         all_file_locations, want_full_precision, is_vlm_update = save_to_gguf(
             model_name = model_name,
diff --git a/unsloth/tokenizer_utils.py b/unsloth/tokenizer_utils.py
index 85b3c5308b..130894e385 100644
--- a/unsloth/tokenizer_utils.py
+++ b/unsloth/tokenizer_utils.py
@@ -434,8 +434,25 @@ def fix_sentencepiece_gguf(saved_location):
     Fixes sentencepiece tokenizers which did not extend the vocabulary with
     user defined tokens.
     Inspiration from https://github.com/ggerganov/llama.cpp/blob/master/convert_hf_to_gguf.py
+
+    Also fixes special tokens (e.g. Gemma 3's <start_of_turn>/<end_of_turn>) that are
+    already present in the sentencepiece model but are incorrectly typed as NORMAL instead
+    of CONTROL. This causes them to be written to GGUF with token_type=1 (NORMAL) instead
+    of token_type=3 (CONTROL), which breaks chat inference in llama.cpp since parse_special
+    only matches CONTROL tokens.
     """
     from copy import deepcopy
+    import sys
+
+    try:
+        from transformers.convert_slow_tokenizer import import_protobuf
+
+        sys.modules.setdefault(
+            "transformers.utils.sentencepiece_model_pb2",
+            import_protobuf(),
+        )
+    except Exception:
+        pass
     from transformers.utils import sentencepiece_model_pb2
     import json
     from enum import IntEnum
@@ -457,12 +474,45 @@ def fix_sentencepiece_gguf(saved_location):
     )
     sentence_piece_size = len(tokenizer_file.pieces)
 
+    # Build a set of token IDs that are marked as special in tokenizer.json.
+    # These tokens should use CONTROL type in the sentencepiece model so that
+    # llama.cpp writes them as CONTROL (type=3) in the GGUF token_type array.
+    special_token_ids = set()
+    if os.path.isfile(f"{saved_location}/tokenizer.json"):
+        with open(f"{saved_location}/tokenizer.json", "r", encoding = "utf-8") as f:
+            tokenizer_json = json.load(f)
+        for entry in tokenizer_json.get("added_tokens", []):
+            token_id = entry.get("id")
+            if entry.get("special", False) and isinstance(token_id, int):
+                special_token_ids.add(token_id)
+
+    # Fix existing sentencepiece tokens that are marked as special in tokenizer.json
+    # but have the wrong type (NORMAL instead of CONTROL) in the sentencepiece model.
+    patched = 0
+    for token_id in special_token_ids:
+        if 0 <= token_id < sentence_piece_size:
+            piece = tokenizer_file.pieces[token_id]
+            if piece.type == SentencePieceTokenTypes.NORMAL:
+                piece.type = SentencePieceTokenTypes.CONTROL
+                patched += 1
+    if patched > 0:
+        logger.warning(
+            f"Unsloth: Patched {patched} special token(s) in {saved_location}/tokenizer.model "
+            f"from NORMAL to CONTROL type so llama.cpp / GGUF chat inference works correctly."
+        )
+
     # Load added_tokens_json
     if not os.path.isfile(f"{saved_location}/added_tokens.json"):
+        if patched > 0:
+            with open(f"{saved_location}/tokenizer.model", "wb") as file:
+                file.write(tokenizer_file.SerializeToString())
         return
     with open(f"{saved_location}/added_tokens.json", "r", encoding = "utf-8") as file:
         added_tokens_json = json.load(file)
     if len(added_tokens_json) == 0:
+        if patched > 0:
+            with open(f"{saved_location}/tokenizer.model", "wb") as file:
+                file.write(tokenizer_file.SerializeToString())
         return
 
     added_tokens_json = dict(
@@ -472,10 +522,20 @@ def fix_sentencepiece_gguf(saved_location):
 
     # Confirm added_tokens_json is correct
     added_tokens_ids = np.array(list(added_tokens_json.values()))
+    _real_added_tokens_ids = added_tokens_ids
+    if len(added_tokens_ids) < 2:
+        added_tokens_ids = np.array([sentence_piece_size, sentence_piece_size + 1])
     diff = np.diff(added_tokens_ids)
     if diff.min() != 1 or diff.max() != 1:
+        if patched > 0:
+            with open(f"{saved_location}/tokenizer.model", "wb") as file:
+                file.write(tokenizer_file.SerializeToString())
         return
+    added_tokens_ids = _real_added_tokens_ids
     if added_tokens_ids.min() != sentence_piece_size:
+        if patched > 0:
+            with open(f"{saved_location}/tokenizer.model", "wb") as file:
+                file.write(tokenizer_file.SerializeToString())
         return
 
     # Edit sentence piece tokens with added_tokens_json
@@ -485,10 +545,16 @@ def fix_sentencepiece_gguf(saved_location):
         f"But we need to extend to sentencepiece vocab size ({new_size})."
     )
     new_tokens = deepcopy(tokenizer_file.pieces[-len(added_tokens_ids) :])
-    for new_token, added_token in zip(new_tokens, added_tokens_json.keys()):
-        new_token.piece = added_token.encode("utf-8")
+    for new_token, added_token_str in zip(new_tokens, added_tokens_json.keys()):
+        added_token_id = added_tokens_json[added_token_str]
+        new_token.piece = added_token_str.encode("utf-8")
         new_token.score = -1000.0
-        new_token.type = SentencePieceTokenTypes.USER_DEFINED
+        # Use CONTROL type for tokens marked as special in tokenizer.json,
+        # otherwise fall back to USER_DEFINED.
+        if added_token_id in special_token_ids:
+            new_token.type = SentencePieceTokenTypes.CONTROL
+        else:
+            new_token.type = SentencePieceTokenTypes.USER_DEFINED
 
     tokenizer_file.pieces.extend(new_tokens)
 

From 114908cd9fbc6921dd1dd3839b5bfcc6c707b909 Mon Sep 17 00:00:00 2001
From: Daniel Han <danielhanchen@gmail.com>
Date: Thu, 23 Apr 2026 04:46:03 -0700
Subject: [PATCH 157/216] fix(install): clear STUDIO_LOCAL_* env on POSIX
 normal install (#5146)

install.sh's normal-install branch passed the inherited parent-shell
environment to setup.sh without resetting STUDIO_LOCAL_INSTALL or
STUDIO_LOCAL_REPO. Consumers treat either as truthy:

  studio/setup.sh:491          checks STUDIO_LOCAL_INSTALL != "1"
  studio/install_python_stack.py:868  reads STUDIO_LOCAL_REPO, falsy
                                      if empty

Net effect: if a user or developer shell has stale STUDIO_LOCAL_*
exports from a previous --local run, a subsequent 'normal' install or
desktop-managed install silently takes the local-dev path: version
checks are skipped and an editable overlay points at the stale repo.

Fix mirrors install.ps1:1082-1087 on Windows: set STUDIO_LOCAL_INSTALL=0
and STUDIO_LOCAL_REPO= explicitly in the env prefix for the non-local
branch so setup.sh sees a clean state regardless of parent exports.

Co-authored-by: Daniel Han <unslothai@gmail.com>
---
 install.sh | 7 +++++++
 1 file changed, 7 insertions(+)

diff --git a/install.sh b/install.sh
index 0352887de1..b1663ba0ff 100755
--- a/install.sh
+++ b/install.sh
@@ -1587,8 +1587,15 @@ if [ "$STUDIO_LOCAL_INSTALL" = true ]; then
     UNSLOTH_NO_TORCH="$SKIP_TORCH" \
     bash "$SETUP_SH" </dev/null || _SETUP_EXIT=$?
 else
+    # Explicitly reset STUDIO_LOCAL_INSTALL / STUDIO_LOCAL_REPO so a stale
+    # value inherited from the parent shell (e.g. a previous --local run in
+    # the same session) does not silently flip a normal install onto the
+    # local-dev path in setup.sh and install_python_stack.py. Mirrors the
+    # reset already done in install.ps1 for PowerShell.
     SKIP_STUDIO_BASE="$_SKIP_BASE" \
     STUDIO_PACKAGE_NAME="$PACKAGE_NAME" \
+    STUDIO_LOCAL_INSTALL=0 \
+    STUDIO_LOCAL_REPO= \
     UNSLOTH_NO_TORCH="$SKIP_TORCH" \
     bash "$SETUP_SH" </dev/null || _SETUP_EXIT=$?
 fi

From a5eb2e3d505b6cc04dcda3b8c4a9a97189636b9d Mon Sep 17 00:00:00 2001
From: Wasim Yousef Said <wasimysdev@gmail.com>
Date: Thu, 23 Apr 2026 13:50:10 +0200
Subject: [PATCH 158/216] Add tauri (#5144)

* add unsloth studio desktop app

* Fix review findings

- studio/src-tauri/tauri.conf.json: retarget updater to staging repo
  (danielhanchen/unsloth-staging-2); switch to unslothai/unsloth on upstream merge.
- studio/src-tauri/linux/postremove.sh: drop the interactive read loop and the
  /home/* iteration. Package maintainer scripts must stay non-interactive and
  must not touch other users' data.
- studio/frontend/src/app/auth-guards.ts: honor tauriAutoAuth() boolean. Failed
  auto-auth now redirects to /login; requireGuest/requirePasswordChangeFlow
  only redirect to /chat when auth succeeds. The new early-return on failed
  auth is intentional so the login / change-password flows remain reachable
  when desktop auth is not yet established.
- studio/frontend/src/config/env.ts: keep fetched=false on health failure so
  later calls retry instead of caching the client-side platform guess.
- studio/src-tauri/src/install.rs: pick the available system package manager
  (apt-get, dnf, zypper, pacman); AppImage bundles run on non-Debian distros.
- studio/frontend/src/lib/open-link.ts + markdown-text/sources callers: return
  boolean from openLink so callers only preventDefault on handled URLs; relative
  hrefs now navigate natively.
- studio/frontend/src/features/settings/tabs/about-tab.tsx: fetch(apiUrl(...))
  so the version request targets the backend port in desktop mode. The bare
  /api/health predates the Tauri webview (blame: the earlier onboarding commit,
  which ran with same-origin frontend/backend); in desktop mode the webview
  origin is tauri://localhost so the bare path fails.
- install.ps1: gate the install_python_stack.py hotfix on a sentinel comment
  instead of a content regex; append the sentinel after applying so reruns
  are unambiguous.
- unsloth_cli/commands/studio.py _write_auth_secret: use the atomic mkstemp +
  os.replace path on Windows too; chmod calls are wrapped in try/except OSError.
- studio/src-tauri/src/preflight.rs probe_existing_backends: fan out the health
  probes concurrently; desktop-auth status still runs sequentially per candidate.
  reqwest::Client is internally Arc-wrapped so the in-loop .clone() is a
  refcount bump, not a deep clone; annotated inline.
- studio/src-tauri/src/preflight.rs run_cli_probe: wait() after kill() to reap
  the child, matching probe_cli_capability.
- studio/src-tauri/src/process.rs + main.rs: add stop_backend_detached and use
  it from the tray quit handler so the 5s graceful-wait does not block the
  Tauri main loop. RunEvent::Exit keeps the synchronous safety-net call.
- studio/backend/main.py: drop the permissive localhost CORS regex in
  api-only mode; the explicit allow_origins list is sufficient.
- .github/workflows/release-desktop.yml: drop max-parallel: 1 so platform
  builds run in parallel, and lift releaseBody to an env var so the three
  tauri-action invocations share one source of truth.

* Fix review findings (loop 2)

- studio/backend/auth/storage.py update_password: clear_desktop_secret()
  alongside clear_bootstrap_password() so rotating the admin password
  also revokes any previously provisioned .desktop_secret. Without this,
  an old local desktop credential keeps minting fresh admin tokens via
  /api/auth/desktop-login after a password rotation.
- studio/src-tauri/src/desktop_auth.rs provision_desktop_auth: wrap
  cmd.output().await in tokio::time::timeout(30s). DESKTOP_AUTH_LOCK is
  held across the whole desktop_auth flow, and previously a hanging
  `unsloth studio provision-desktop-auth` subprocess would pin the lock
  indefinitely and freeze every subsequent desktop_auth call.

* Add review tests

* Consolidate review tests

Merge review-added tests into the existing studio/backend/tests/test_desktop_auth.py
(the PR's authoritative desktop-auth test file). Drops three scaffolding files under
tests/python/ in favor of five focused tests next to the tests they extend:
- test_update_password_clears_desktop_secret (runtime)
- test_update_password_on_unknown_user_leaves_desktop_secret_intact (runtime)
- test_cli_provisioning_delegates_to_storage_create_desktop_secret (source-level)
- test_cli_connect_auth_db_reads_storage_db_path (source-level)
- test_desktop_auth_provision_has_bounded_timeout (Rust source-level)

* Revert auth-guards.ts Tauri branches to unconditional form

The review loop on PR 5144 introduced a regression: the isTauri branch of
requireAuth redirected to /login when tauriAutoAuth() returned false, and
requireGuest / requirePasswordChangeFlow silently fell through on the same
condition. The Tauri desktop app authenticates via a local auto-generated
secret; it must never surface /login or /change-password to the user. A
failed auto-auth should let the startup layer retry, not expose a password
form.

Restore the three Tauri branches to the author's original unconditional
form (requireAuth: return; requireGuest / requirePasswordChangeFlow: throw
redirect({to: '/chat'})). Keep the rest of the review fixes -- the
apiUrl() fetch wrapping, authRedirect helper, and fetchAuthStatus refactor
are all legitimate improvements and are preserved.

* Revert release-desktop.yml to author's version

The review loop's workflow-file tweaks (drop max-parallel: 1, lift releaseBody
to an env var) are cosmetic. OAuth tokens cannot push workflow-file changes,
and fine-grained PATs cannot honor maintainerCanModify on a third-party fork.
Reverting the workflow file to wasimysaid's version lets the push go through
without needing a classic PAT with both repo and workflow scopes.

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

---------

Co-authored-by: Lee Jackson <130007945+Imagineer99@users.noreply.github.com>
Co-authored-by: Daniel Han <danielhanchen@gmail.com>
Co-authored-by: Daniel Han <unslothai@gmail.com>
Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com>
---
 .github/workflows/release-desktop.yml         |  187 +
 .gitignore                                    |   12 +
 install.ps1                                   |   78 +-
 install.sh                                    |   69 +-
 studio/backend/auth/authentication.py         |   50 +-
 studio/backend/auth/storage.py                |  121 +-
 .../data_recipe/local_callable_validators.py  |    6 +
 studio/backend/core/inference/audio_codecs.py |    9 +-
 studio/backend/core/inference/llama_cpp.py    |    8 +
 studio/backend/core/training/trainer.py       |    9 +-
 studio/backend/loggers/config.py              |    8 +
 studio/backend/main.py                        |   17 +-
 studio/backend/models/auth.py                 |    6 +
 studio/backend/routes/auth.py                 |   25 +-
 studio/backend/routes/data_recipe/jobs.py     |   15 +
 studio/backend/run.py                         |   23 +-
 studio/backend/tests/test_desktop_auth.py     |  597 ++
 studio/backend/utils/hardware/nvidia.py       |    8 +
 studio/backend/utils/models/model_config.py   |    5 +
 studio/backend/utils/subprocess_compat.py     |   34 +
 studio/backend/utils/transformers_version.py  |    6 +
 studio/frontend/package.json                  |    4 +
 studio/frontend/public/studio.png             |  Bin 0 -> 27420 bytes
 studio/frontend/src/app/auth-guards.ts        |   65 +-
 studio/frontend/src/app/provider.tsx          |  167 +-
 studio/frontend/src/app/routes/__root.tsx     |    7 +-
 .../components/assistant-ui/markdown-text.tsx |    9 +-
 .../src/components/assistant-ui/sources.tsx   |   15 +-
 .../src/components/tauri/startup-screen.tsx   |  401 ++
 .../src/components/tauri/update-banner.tsx    |   84 +
 .../src/components/tauri/update-screen.tsx    |  173 +
 .../src/components/ui/shimmer-button.tsx      |   96 +
 studio/frontend/src/config/env.ts             |   31 +-
 studio/frontend/src/features/auth/api.ts      |   49 +-
 .../features/auth/components/auth-form.tsx    |   16 +-
 studio/frontend/src/features/auth/index.ts    |    5 +
 studio/frontend/src/features/auth/session.ts  |    2 +
 .../src/features/auth/tauri-auto-auth.ts      |   95 +
 .../src/features/settings/tabs/about-tab.tsx  |    3 +-
 studio/frontend/src/hooks/index.ts            |    1 +
 studio/frontend/src/hooks/use-gpu-info.ts     |    3 +-
 .../frontend/src/hooks/use-tauri-backend.ts   |  487 ++
 studio/frontend/src/hooks/use-tauri-update.ts |  219 +
 studio/frontend/src/lib/api-base.ts           |   24 +
 studio/frontend/src/lib/open-link.ts          |   31 +
 studio/install_llama_prebuilt.py              |   25 +
 studio/install_python_stack.py                |   46 +-
 studio/setup.ps1                              |   10 +-
 studio/setup.sh                               |   15 +-
 studio/src-tauri/Cargo.lock                   | 6320 +++++++++++++++++
 studio/src-tauri/Cargo.toml                   |   42 +
 studio/src-tauri/Entitlements.plist           |   15 +
 studio/src-tauri/build.rs                     |    3 +
 studio/src-tauri/capabilities/default.json    |   24 +
 studio/src-tauri/icons/128x128.png            |  Bin 0 -> 9194 bytes
 studio/src-tauri/icons/32x32.png              |  Bin 0 -> 1930 bytes
 studio/src-tauri/icons/icon.icns              |  Bin 0 -> 263568 bytes
 studio/src-tauri/icons/icon.ico               |  Bin 0 -> 34589 bytes
 studio/src-tauri/icons/icon.png               |  Bin 0 -> 46007 bytes
 studio/src-tauri/linux/postremove.sh          |    9 +
 studio/src-tauri/src/commands.rs              |  476 ++
 studio/src-tauri/src/desktop_auth.rs          |  388 +
 studio/src-tauri/src/install.rs               |  555 ++
 studio/src-tauri/src/main.rs                  |  189 +
 studio/src-tauri/src/preflight.rs             |  831 +++
 studio/src-tauri/src/process.rs               |  537 ++
 studio/src-tauri/src/update.rs                |  293 +
 studio/src-tauri/src/windows_job.rs           |   72 +
 studio/src-tauri/tauri.conf.json              |   66 +
 studio/src-tauri/tauri.macos.conf.json        |   11 +
 studio/src-tauri/tauri.windows.conf.json      |   16 +
 .../windows/branding/nsis-header.bmp          |  Bin 0 -> 25818 bytes
 .../windows/branding/nsis-sidebar.bmp         |  Bin 0 -> 154542 bytes
 studio/src-tauri/windows/hooks.nsh            |    8 +
 tests/sh/test_tauri_install_exit_order.sh     |   95 +
 unsloth_cli/commands/studio.py                |  341 +-
 76 files changed, 13533 insertions(+), 134 deletions(-)
 create mode 100644 .github/workflows/release-desktop.yml
 create mode 100644 studio/backend/tests/test_desktop_auth.py
 create mode 100644 studio/backend/utils/subprocess_compat.py
 create mode 100644 studio/frontend/public/studio.png
 create mode 100644 studio/frontend/src/components/tauri/startup-screen.tsx
 create mode 100644 studio/frontend/src/components/tauri/update-banner.tsx
 create mode 100644 studio/frontend/src/components/tauri/update-screen.tsx
 create mode 100644 studio/frontend/src/components/ui/shimmer-button.tsx
 create mode 100644 studio/frontend/src/features/auth/tauri-auto-auth.ts
 create mode 100644 studio/frontend/src/hooks/use-tauri-backend.ts
 create mode 100644 studio/frontend/src/hooks/use-tauri-update.ts
 create mode 100644 studio/frontend/src/lib/api-base.ts
 create mode 100644 studio/frontend/src/lib/open-link.ts
 create mode 100644 studio/src-tauri/Cargo.lock
 create mode 100644 studio/src-tauri/Cargo.toml
 create mode 100644 studio/src-tauri/Entitlements.plist
 create mode 100644 studio/src-tauri/build.rs
 create mode 100644 studio/src-tauri/capabilities/default.json
 create mode 100644 studio/src-tauri/icons/128x128.png
 create mode 100644 studio/src-tauri/icons/32x32.png
 create mode 100644 studio/src-tauri/icons/icon.icns
 create mode 100644 studio/src-tauri/icons/icon.ico
 create mode 100644 studio/src-tauri/icons/icon.png
 create mode 100755 studio/src-tauri/linux/postremove.sh
 create mode 100644 studio/src-tauri/src/commands.rs
 create mode 100644 studio/src-tauri/src/desktop_auth.rs
 create mode 100644 studio/src-tauri/src/install.rs
 create mode 100644 studio/src-tauri/src/main.rs
 create mode 100644 studio/src-tauri/src/preflight.rs
 create mode 100644 studio/src-tauri/src/process.rs
 create mode 100644 studio/src-tauri/src/update.rs
 create mode 100644 studio/src-tauri/src/windows_job.rs
 create mode 100644 studio/src-tauri/tauri.conf.json
 create mode 100644 studio/src-tauri/tauri.macos.conf.json
 create mode 100644 studio/src-tauri/tauri.windows.conf.json
 create mode 100644 studio/src-tauri/windows/branding/nsis-header.bmp
 create mode 100644 studio/src-tauri/windows/branding/nsis-sidebar.bmp
 create mode 100644 studio/src-tauri/windows/hooks.nsh
 create mode 100644 tests/sh/test_tauri_install_exit_order.sh

diff --git a/.github/workflows/release-desktop.yml b/.github/workflows/release-desktop.yml
new file mode 100644
index 0000000000..2d1c6f51f2
--- /dev/null
+++ b/.github/workflows/release-desktop.yml
@@ -0,0 +1,187 @@
+name: Release Desktop App
+
+on:
+  workflow_dispatch:
+    inputs:
+      draft:
+        description: 'Create as draft release'
+        type: boolean
+        default: true
+
+permissions:
+  contents: write
+
+jobs:
+  build:
+    strategy:
+      fail-fast: false
+      max-parallel: 1
+      matrix:
+        include:
+          - platform: macos-latest
+            args: '--target aarch64-apple-darwin'
+            label: macOS (Apple Silicon)
+          # - platform: macos-latest
+          #   args: '--target x86_64-apple-darwin'
+          #   label: macOS (Intel)
+          - platform: ubuntu-22.04
+            args: ''
+            label: Linux (x64)
+          - platform: windows-latest
+            args: ''
+            label: Windows (x64)
+
+    name: Build ${{ matrix.label }}
+    runs-on: ${{ matrix.platform }}
+
+    env:
+      FORCE_JAVASCRIPT_ACTIONS_TO_NODE24: true
+
+
+    steps:
+      - uses: actions/checkout@34e114876b0b11c390a56381ad16ebd13914f8d5
+
+      # ── Linux dependencies ──
+      - name: Install Linux dependencies
+        if: matrix.platform == 'ubuntu-22.04'
+        run: |
+          sudo apt-get update
+          sudo apt-get install -y libwebkit2gtk-4.1-dev libayatana-appindicator3-dev librsvg2-dev libxdo-dev libssl-dev patchelf
+
+      # ── Node.js ──
+      - name: Setup Node.js
+        uses: actions/setup-node@49933ea5288caeca8642d1e84afbd3f7d6820020
+        with:
+          node-version: 24
+
+      - name: Install frontend dependencies
+        working-directory: studio/frontend
+        run: npm install
+
+      # ── Rust ──
+      - name: Install Rust stable
+        uses: dtolnay/rust-toolchain@stable
+        with:
+          targets: ${{ matrix.platform == 'macos-latest' && 'aarch64-apple-darwin,x86_64-apple-darwin' || '' }}
+
+      - name: Rust cache
+        uses: swatinem/rust-cache@42dc69e1aa15d09112580998cf2ef0119e2e91ae
+        with:
+          workspaces: 'studio/src-tauri -> target'
+
+      # ── macOS: import signing certificate ──
+      - name: Import Apple certificate
+        if: matrix.platform == 'macos-latest'
+        env:
+          APPLE_CERTIFICATE: ${{ secrets.APPLE_CERTIFICATE }}
+          APPLE_CERTIFICATE_PASSWORD: ${{ secrets.APPLE_CERTIFICATE_PASSWORD }}
+          KEYCHAIN_PASSWORD: ${{ secrets.KEYCHAIN_PASSWORD }}
+        run: |
+          echo $APPLE_CERTIFICATE | base64 --decode > certificate.p12
+          security create-keychain -p "$KEYCHAIN_PASSWORD" build.keychain
+          security default-keychain -s build.keychain
+          security unlock-keychain -p "$KEYCHAIN_PASSWORD" build.keychain
+          security set-keychain-settings -t 3600 -u build.keychain
+          security import certificate.p12 -k build.keychain -P "$APPLE_CERTIFICATE_PASSWORD" -T /usr/bin/codesign
+          security set-key-partition-list -S apple-tool:,apple:,codesign: -s -k "$KEYCHAIN_PASSWORD" build.keychain
+          security find-identity -v -p codesigning build.keychain
+          rm -f certificate.p12
+
+      # ── Windows: install Azure Trusted Signing CLI ──
+      - name: Install trusted-signing-cli
+        if: matrix.platform == 'windows-latest'
+        run: |
+          cargo install trusted-signing-cli --version 0.9.0 --locked
+          echo "$env:USERPROFILE\.cargo\bin" | Out-File -FilePath $env:GITHUB_PATH -Encoding utf8 -Append
+
+      # ── Windows: verify signing CLI is accessible ──
+      - name: Verify trusted-signing-cli
+        if: matrix.platform == 'windows-latest'
+        run: |
+          Write-Output "PATH: $env:PATH"
+          Get-Command trusted-signing-cli -ErrorAction SilentlyContinue || Write-Output "trusted-signing-cli NOT in PATH"
+          trusted-signing-cli --version || Write-Output "trusted-signing-cli failed to run"
+
+      # ── Linux: build + sign + upload ──
+      - name: Build Linux app
+        if: matrix.platform == 'ubuntu-22.04'
+        uses: tauri-apps/tauri-action@fce9c6108b31ea247710505d3aaaa893ee6768d4
+        env:
+          GITHUB_TOKEN: ${{ secrets.GITHUB_TOKEN }}
+          TAURI_SIGNING_PRIVATE_KEY: ${{ secrets.TAURI_SIGNING_PRIVATE_KEY }}
+          TAURI_SIGNING_PRIVATE_KEY_PASSWORD: ${{ secrets.TAURI_SIGNING_PRIVATE_KEY_PASSWORD }}
+        with:
+          projectPath: studio
+          tagName: desktop-v__VERSION__
+          releaseName: 'Unsloth Studio (Desktop) v__VERSION__'
+          releaseBody: |
+            Desktop app for Unsloth Studio.
+
+            **macOS**: Download the Apple Silicon `.dmg`.
+            **Windows**: Download the `-setup.exe` installer.
+            **Linux**: Download `.deb` (Ubuntu/Debian) or `.AppImage` (universal).
+
+            > Linux in-app updates are AppImage-oriented. Package installs should update by downloading a new package.
+            > Linux AppImage on Ubuntu 24.04+ may require: `sudo apt install libfuse2t64`
+          releaseDraft: ${{ inputs.draft }}
+          prerelease: false
+          args: -v ${{ matrix.args }}
+
+      # ── macOS: build + sign + notarize + upload ──
+      - name: Build macOS app
+        if: matrix.platform == 'macos-latest'
+        uses: tauri-apps/tauri-action@fce9c6108b31ea247710505d3aaaa893ee6768d4
+        env:
+          GITHUB_TOKEN: ${{ secrets.GITHUB_TOKEN }}
+          TAURI_SIGNING_PRIVATE_KEY: ${{ secrets.TAURI_SIGNING_PRIVATE_KEY }}
+          TAURI_SIGNING_PRIVATE_KEY_PASSWORD: ${{ secrets.TAURI_SIGNING_PRIVATE_KEY_PASSWORD }}
+          APPLE_SIGNING_IDENTITY: ${{ secrets.APPLE_SIGNING_IDENTITY }}
+          APPLE_ID: ${{ secrets.APPLE_ID }}
+          APPLE_PASSWORD: ${{ secrets.APPLE_PASSWORD }}
+          APPLE_TEAM_ID: ${{ secrets.APPLE_TEAM_ID }}
+        with:
+          projectPath: studio
+          tagName: desktop-v__VERSION__
+          releaseName: 'Unsloth Studio (Desktop) v__VERSION__'
+          releaseBody: |
+            Desktop app for Unsloth Studio.
+
+            **macOS**: Download the Apple Silicon `.dmg`.
+            **Windows**: Download the `-setup.exe` installer.
+            **Linux**: Download `.deb` (Ubuntu/Debian) or `.AppImage` (universal).
+
+            > Linux in-app updates are AppImage-oriented. Package installs should update by downloading a new package.
+            > Linux AppImage on Ubuntu 24.04+ may require: `sudo apt install libfuse2t64`
+          releaseDraft: ${{ inputs.draft }}
+          prerelease: false
+          args: -v ${{ matrix.args }}
+
+      # ── Windows: build + sign + upload ──
+      - name: Build Windows app
+        if: matrix.platform == 'windows-latest'
+        uses: tauri-apps/tauri-action@fce9c6108b31ea247710505d3aaaa893ee6768d4
+        env:
+          GITHUB_TOKEN: ${{ secrets.GITHUB_TOKEN }}
+          TAURI_SIGNING_PRIVATE_KEY: ${{ secrets.TAURI_SIGNING_PRIVATE_KEY }}
+          TAURI_SIGNING_PRIVATE_KEY_PASSWORD: ${{ secrets.TAURI_SIGNING_PRIVATE_KEY_PASSWORD }}
+          AZURE_CLIENT_ID: ${{ secrets.AZURE_CLIENT_ID }}
+          AZURE_CLIENT_SECRET: ${{ secrets.AZURE_CLIENT_SECRET }}
+          AZURE_TENANT_ID: ${{ secrets.AZURE_TENANT_ID }}
+          AZURE_TRUSTED_SIGNING_ACCOUNT_NAME: ${{ secrets.AZURE_TRUSTED_SIGNING_ACCOUNT_NAME }}
+          AZURE_CERTIFICATE_PROFILE_NAME: ${{ secrets.AZURE_CERTIFICATE_PROFILE_NAME }}
+        with:
+          projectPath: studio
+          tagName: desktop-v__VERSION__
+          releaseName: 'Unsloth Studio (Desktop) v__VERSION__'
+          releaseBody: |
+            Desktop app for Unsloth Studio.
+
+            **macOS**: Download the Apple Silicon `.dmg`.
+            **Windows**: Download the `-setup.exe` installer.
+            **Linux**: Download `.deb` (Ubuntu/Debian) or `.AppImage` (universal).
+
+            > Linux in-app updates are AppImage-oriented. Package installs should update by downloading a new package.
+            > Linux AppImage on Ubuntu 24.04+ may require: `sudo apt install libfuse2t64`
+          releaseDraft: ${{ inputs.draft }}
+          prerelease: false
+          args: -v ${{ matrix.args }}
diff --git a/.gitignore b/.gitignore
index 7a24d07c6f..b6786ee655 100644
--- a/.gitignore
+++ b/.gitignore
@@ -204,6 +204,18 @@ tmp/
 **/node_modules/
 auth.db
 
+# Tauri local build/generated output
+studio/src-tauri/target/
+studio/src-tauri/gen/
+studio/src-tauri/artifacts/
+studio/src-tauri/icons/android/
+studio/src-tauri/icons/ios/
+studio/src-tauri/icons/128x128@2x.png
+studio/src-tauri/icons/64x64.png
+studio/src-tauri/icons/Square*Logo.png
+studio/src-tauri/icons/StoreLogo.png
+studio/src-tauri/icons/squarehq.png
+
 # Local working docs
 **/CLAUDE.md
 **/claude.md
diff --git a/install.ps1 b/install.ps1
index 7f98fc5cb8..44464101f3 100644
--- a/install.ps1
+++ b/install.ps1
@@ -12,11 +12,13 @@ function Install-UnslothStudio {
     $StudioLocalInstall = $false
     $PackageName = "unsloth"
     $RepoRoot = ""
+    $TauriMode = $false
     $SkipTorch = $false
     $argList = $args
     for ($i = 0; $i -lt $argList.Count; $i++) {
         switch ($argList[$i]) {
             "--local"    { $StudioLocalInstall = $true }
+            "--tauri"    { $TauriMode = $true }
             "--no-torch" { $SkipTorch = $true }
             "--verbose"  { $script:UnslothVerbose = $true }
             "-v"         { $script:UnslothVerbose = $true }
@@ -44,6 +46,20 @@ function Install-UnslothStudio {
         }
     }
 
+    # Validate --package to prevent injection into shell/Python commands
+    if ($PackageName -notmatch '^[a-zA-Z0-9][a-zA-Z0-9._-]*$') {
+        Write-Host "[ERROR] --package name contains invalid characters (allowed: a-z A-Z 0-9 . _ -)" -ForegroundColor Red
+        return
+    }
+
+    # ── Tauri structured output ──
+    function Write-TauriLog {
+        param([string]$Tag, [string]$Message)
+        if ($TauriMode) {
+            Write-Host "[TAURI:$Tag] $Message"
+        }
+    }
+
     $PythonVersion = "3.13"
     $StudioHome = Join-Path $env:USERPROFILE ".unsloth\studio"
     $VenvDir = Join-Path $StudioHome "unsloth_studio"
@@ -609,6 +625,7 @@ shell.Run cmd, 0, False
     }
 
     # ── Check winget ──
+    Write-TauriLog "STEP" "Checking system dependencies"
     if (-not (Get-Command winget -ErrorAction SilentlyContinue)) {
         step "winget" "not available" "Red"
         substep "Install it from https://aka.ms/getwinget" "Yellow"
@@ -688,6 +705,7 @@ shell.Run cmd, 0, False
 
     # ── Install Python if no compatible version (3.11-3.13) found ──
     # Find-CompatiblePython returns @{ Version = "3.13"; Path = "C:\...\python.exe" } or $null.
+    Write-TauriLog "STEP" "Installing Python"
     $DetectedPython = Find-CompatiblePython
     if ($DetectedPython) {
         step "python" "Python $($DetectedPython.Version) already installed"
@@ -736,6 +754,7 @@ shell.Run cmd, 0, False
     }
 
     # ── Install uv if not present ──
+    Write-TauriLog "STEP" "Installing uv package manager"
     if (-not (Get-Command uv -ErrorAction SilentlyContinue)) {
         substep "installing uv package manager..."
         $prevEAP = $ErrorActionPreference
@@ -746,7 +765,7 @@ shell.Run cmd, 0, False
         # Fallback: if winget didn't put uv on PATH, try the PowerShell installer
         if (-not (Get-Command uv -ErrorAction SilentlyContinue)) {
             substep "trying alternative uv installer..." "Yellow"
-            powershell -ExecutionPolicy ByPass -c "irm https://astral.sh/uv/install.ps1 | iex"
+            Invoke-Expression (Invoke-RestMethod -Uri "https://astral.sh/uv/install.ps1")
             Refresh-SessionPath
         }
     }
@@ -760,6 +779,7 @@ shell.Run cmd, 0, False
     # ── Create venv (migrate old layout if possible, otherwise fresh) ──
     # Pass the resolved executable path to uv so it does not re-resolve
     # a version string back to a conda interpreter.
+    Write-TauriLog "STEP" "Creating virtual environment"
     if (-not (Test-Path $StudioHome)) {
         New-Item -ItemType Directory -Path $StudioHome -Force | Out-Null
     }
@@ -806,6 +826,7 @@ shell.Run cmd, 0, False
         substep "$VenvDir"
         $venvExit = Invoke-InstallCommand { uv venv $VenvDir --python "$($DetectedPython.Path)" }
         if ($venvExit -ne 0) {
+            Write-TauriLog "ERROR" "Failed to create virtual environment (exit code $venvExit)"
             Write-Host "[ERROR] Failed to create virtual environment (exit code $venvExit)" -ForegroundColor Red
             return
         }
@@ -908,6 +929,7 @@ shell.Run cmd, 0, False
     if ($_Migrated) {
         # Migrated env: force-reinstall unsloth+unsloth-zoo to ensure clean state
         # in the new venv location, while preserving existing torch/CUDA
+        Write-TauriLog "STEP" "Installing unsloth"
         substep "upgrading unsloth in migrated environment..."
         if ($SkipTorch) {
             # No-torch: install unsloth + unsloth-zoo with --no-deps, then
@@ -938,14 +960,17 @@ shell.Run cmd, 0, False
         if ($SkipTorch) {
             substep "skipping PyTorch (--no-torch flag set)." "Yellow"
         } else {
+            Write-TauriLog "STEP" "Installing PyTorch"
             substep "installing PyTorch ($TorchIndexUrl)..."
             $torchInstallExit = Invoke-InstallCommand { uv pip install --python $VenvPython "torch>=2.4,<2.11.0" torchvision torchaudio --index-url $TorchIndexUrl }
             if ($torchInstallExit -ne 0) {
+                Write-TauriLog "ERROR" "Failed to install PyTorch (exit code $torchInstallExit)"
                 Write-Host "[ERROR] Failed to install PyTorch (exit code $torchInstallExit)" -ForegroundColor Red
                 return
             }
         }
 
+        Write-TauriLog "STEP" "Installing unsloth"
         substep "installing unsloth (this may take a few minutes)..."
         if ($SkipTorch) {
             # No-torch: install unsloth + unsloth-zoo with --no-deps, then
@@ -960,9 +985,10 @@ shell.Run cmd, 0, False
         } elseif ($StudioLocalInstall) {
             $baseInstallExit = Invoke-InstallCommand { uv pip install --python $VenvPython --upgrade-package unsloth "unsloth>=2026.4.7" unsloth-zoo }
         } else {
-            $baseInstallExit = Invoke-InstallCommand { uv pip install --python $VenvPython --upgrade-package unsloth "$PackageName" }
+            $baseInstallExit = Invoke-InstallCommand { uv pip install --python $VenvPython --upgrade-package unsloth -- "$PackageName" }
         }
         if ($baseInstallExit -ne 0) {
+            Write-TauriLog "ERROR" "Failed to install unsloth (exit code $baseInstallExit)"
             Write-Host "[ERROR] Failed to install unsloth (exit code $baseInstallExit)" -ForegroundColor Red
             return
         }
@@ -977,6 +1003,7 @@ shell.Run cmd, 0, False
         }
     } else {
         # Fallback: GPU detection failed to produce a URL -- let uv resolve torch
+        Write-TauriLog "STEP" "Installing unsloth"
         substep "installing unsloth (this may take a few minutes)..."
         if ($StudioLocalInstall) {
             $baseInstallExit = Invoke-InstallCommand { uv pip install --python $VenvPython unsloth-zoo "unsloth>=2026.4.7" --torch-backend=auto }
@@ -991,20 +1018,52 @@ shell.Run cmd, 0, False
                 return
             }
         } else {
-            $baseInstallExit = Invoke-InstallCommand { uv pip install --python $VenvPython "$PackageName" --torch-backend=auto }
+            $baseInstallExit = Invoke-InstallCommand { uv pip install --python $VenvPython --torch-backend=auto -- "$PackageName" }
             if ($baseInstallExit -ne 0) {
+                Write-TauriLog "ERROR" "Failed to install unsloth (exit code $baseInstallExit)"
                 Write-Host "[ERROR] Failed to install unsloth (exit code $baseInstallExit)" -ForegroundColor Red
                 return
             }
         }
     }
 
+    # Hotfix: patch install_python_stack.py for Windows GUI stdout
+    # The PyPI version crashes with OSError when stdout is piped from a GUI app.
+    # Copy our fixed version (bundled by Tauri) over the installed one.
+    # Remove this block once PyPI ships the fix from commit 18c5aae7.
+    if ($TauriMode) {
+        $rawPath = if ($PSCommandPath) { $PSCommandPath } else { $MyInvocation.ScriptName }
+        $scriptDir = Split-Path -Parent ($rawPath -replace '^\\\\\?\\', '')
+        $fixedPy = Join-Path $scriptDir "install_python_stack.py"
+        $target = Join-Path $VenvDir "Lib\site-packages\studio\install_python_stack.py"
+        $sentinel = "# UNSLOTH_DESKTOP_HOTFIX_APPLIED_v1"
+        $sentinelPattern = [regex]::Escape($sentinel)
+        if ((Test-Path $fixedPy) -and (Test-Path $target)) {
+            $installed = Get-Content $target -Raw
+            if ($installed -notmatch $sentinelPattern) {
+                Copy-Item $fixedPy $target -Force
+                Add-Content -Path $target -Value "`n$sentinel"
+                substep "patched install_python_stack.py (stdout fix)"
+            } else {
+                substep "install_python_stack.py already has stdout fix"
+            }
+        } elseif ((Test-Path $fixedPy) -and (Test-Path (Split-Path $target))) {
+            Copy-Item $fixedPy $target -Force
+            Add-Content -Path $target -Value "`n$sentinel"
+            substep "patched install_python_stack.py (stdout fix)"
+        } else {
+            Write-Host "[WARN] Could not patch install_python_stack.py (bundled file or target dir missing)" -ForegroundColor Yellow
+        }
+    }
+
     # ── Run studio setup ──
     # setup.ps1 will handle installing Git, CMake, Visual Studio Build Tools,
     # CUDA Toolkit, Node.js, and other dependencies automatically via winget.
+    Write-TauriLog "STEP" "Running studio setup"
     step "setup" "running unsloth studio setup..."
     $UnslothExe = Join-Path $VenvDir "Scripts\unsloth.exe"
     if (-not (Test-Path $UnslothExe)) {
+        Write-TauriLog "ERROR" "unsloth CLI was not installed correctly"
         Write-Host "[ERROR] unsloth CLI was not installed correctly." -ForegroundColor Red
         Write-Host "        Expected: $UnslothExe" -ForegroundColor Yellow
         Write-Host "        This usually means an older unsloth version was installed that does not include the Studio CLI." -ForegroundColor Yellow
@@ -1015,6 +1074,8 @@ shell.Run cmd, 0, False
     $env:SKIP_STUDIO_BASE = "1"
     $env:STUDIO_PACKAGE_NAME = $PackageName
     $env:UNSLOTH_NO_TORCH = if ($SkipTorch) { "true" } else { "false" }
+    # Tauri desktop app bundles its own frontend — skip Node/npm/frontend build
+    $env:SKIP_STUDIO_FRONTEND = if ($TauriMode) { "1" } else { "0" }
     # Always set STUDIO_LOCAL_INSTALL explicitly to avoid stale values from
     # a previous --local run in the same PowerShell session.
     if ($StudioLocalInstall) {
@@ -1032,12 +1093,11 @@ shell.Run cmd, 0, False
     & $UnslothExe @studioArgs
     $setupExit = $LASTEXITCODE
     if ($setupExit -ne 0) {
+        Write-TauriLog "ERROR" "unsloth studio setup failed (exit code $setupExit)"
         Write-Host "[ERROR] unsloth studio setup failed (exit code $setupExit)" -ForegroundColor Red
         return
     }
 
-    New-StudioShortcuts -UnslothExePath $UnslothExe
-
     # ── Expose `unsloth` via a shim dir containing only unsloth.exe ──
     # We do NOT add the venv Scripts dir to PATH (it also holds python.exe
     # and pip.exe, which would hijack the user's system interpreter).
@@ -1109,6 +1169,14 @@ shell.Run cmd, 0, False
     }
     Refresh-SessionPath  # sync current session with registry
 
+    # ── Tauri mode: done, skip shortcuts and auto-launch ──
+    if ($TauriMode) {
+        Write-TauriLog "DONE" ""
+        return
+    }
+
+    New-StudioShortcuts -UnslothExePath $UnslothExe
+
     # Launch studio automatically in interactive terminals;
     # in non-interactive environments (CI, Docker) just print instructions.
     $IsInteractive = [Environment]::UserInteractive -and (-not [Console]::IsInputRedirected)
diff --git a/install.sh b/install.sh
index b1663ba0ff..07473e441d 100755
--- a/install.sh
+++ b/install.sh
@@ -35,6 +35,7 @@ substep() { printf "  ${C_DIM}%-15s${2:-$C_DIM}%s${C_RST}\n" "" "$1"; }
 # ── Parse flags ──
 STUDIO_LOCAL_INSTALL=false
 PACKAGE_NAME="unsloth"
+TAURI_MODE=false
 _USER_PYTHON=""
 _NO_TORCH_FLAG=false
 _VERBOSE=false
@@ -54,6 +55,7 @@ for arg in "$@"; do
     case "$arg" in
         --local) STUDIO_LOCAL_INSTALL=true ;;
         --package) _next_is_package=true ;;
+        --tauri) TAURI_MODE=true ;;
         --python) _next_is_python=true ;;
         --no-torch) _NO_TORCH_FLAG=true ;;
         --verbose|-v) _VERBOSE=true ;;
@@ -142,6 +144,24 @@ if [ "$_next_is_python" = true ]; then
     exit 1
 fi
 
+# Validate --package to prevent injection into shell/Python commands.
+# Must start with a letter/digit (rejects leading dashes that uv would parse as flags).
+case "$PACKAGE_NAME" in
+    [!a-zA-Z0-9]*)
+        echo "❌ ERROR: --package name must start with a letter or digit." >&2
+        exit 1 ;;
+    *[!a-zA-Z0-9._-]*)
+        echo "❌ ERROR: --package name contains invalid characters (allowed: a-z A-Z 0-9 . _ -)" >&2
+        exit 1 ;;
+esac
+
+# ── Tauri structured output ──
+tauri_log() {
+    if [ "$TAURI_MODE" = true ]; then
+        echo "[TAURI:$1] $2"
+    fi
+}
+
 PYTHON_VERSION=""  # resolved after platform detection
 STUDIO_HOME="$HOME/.unsloth/studio"
 VENV_DIR="$STUDIO_HOME/unsloth_studio"
@@ -192,6 +212,12 @@ _smart_apt_install() {
         return 0
     fi
 
+    # In Tauri mode, report needed packages and exit — Rust handles elevation
+    if [ "$TAURI_MODE" = true ]; then
+        tauri_log "NEED_SUDO" "$_STILL_MISSING"
+        exit 2
+    fi
+
     # Step 3: Escalate -- need elevated permissions for remaining packages
     if command -v sudo >/dev/null 2>&1; then
         echo ""
@@ -752,6 +778,7 @@ printf "  ${C_DIM}%s${C_RST}\n" "$RULE"
 echo ""
 
 # ── Detect platform ──
+tauri_log "STEP" "Detecting platform"
 OS="linux"
 if [ "$(uname)" = "Darwin" ]; then
     OS="macos"
@@ -804,6 +831,7 @@ fi
 # ── Check system dependencies ──
 # cmake and git are needed by unsloth studio setup to build the GGUF inference
 # engine (llama.cpp). build-essential and libcurl-dev are also needed on Linux.
+tauri_log "STEP" "Checking system dependencies"
 MISSING=""
 
 command -v cmake >/dev/null 2>&1 || MISSING="$MISSING cmake"
@@ -868,6 +896,7 @@ else
 fi
 
 # ── Install uv ──
+tauri_log "STEP" "Installing uv package manager"
 UV_MIN_VERSION="0.7.14"
 
 version_ge() {
@@ -922,6 +951,7 @@ if ! command -v uv >/dev/null 2>&1 || ! _uv_version_ok uv; then
 fi
 
 # ── Create venv (migrate old layout if possible, otherwise fresh) ──
+tauri_log "STEP" "Creating virtual environment"
 mkdir -p "$STUDIO_HOME"
 
 _MIGRATED=false
@@ -1304,6 +1334,7 @@ case "$TORCH_INDEX_URL" in
 esac
 
 # ── Install unsloth directly into the venv (no activation needed) ──
+tauri_log "STEP" "Installing PyTorch"
 _VENV_PY="$VENV_DIR/bin/python"
 if [ "$_MIGRATED" = true ]; then
     # Migrated env: force-reinstall unsloth+unsloth-zoo to ensure clean state
@@ -1481,6 +1512,7 @@ elif [ -n "$TORCH_INDEX_URL" ]; then
         esac
     fi
     # Fresh: Step 2 - install unsloth, preserving pre-installed torch
+    tauri_log "STEP" "Installing Unsloth"
     substep "installing unsloth (this may take a few minutes)..."
     if [ "$SKIP_TORCH" = true ]; then
         # No-torch: install unsloth + unsloth-zoo with --no-deps, then
@@ -1503,7 +1535,7 @@ elif [ -n "$TORCH_INDEX_URL" ]; then
         run_install_cmd "overlay local repo" uv pip install --python "$_VENV_PY" -e "$_REPO_ROOT" --no-deps
     else
         run_install_cmd "install unsloth" uv pip install --python "$_VENV_PY" \
-            --upgrade-package unsloth "$PACKAGE_NAME"
+            --upgrade-package unsloth -- "$PACKAGE_NAME"
     fi
     # AMD ROCm: repair torch if the unsloth/unsloth-zoo install pulled in
     # CUDA torch from PyPI, overwriting the ROCm wheels installed in Step 1.
@@ -1523,17 +1555,19 @@ elif [ -n "$TORCH_INDEX_URL" ]; then
     fi
 else
     # Fallback: GPU detection failed to produce a URL -- let uv resolve torch
+    tauri_log "STEP" "Installing Unsloth"
     substep "installing unsloth (this may take a few minutes)..."
     if [ "$STUDIO_LOCAL_INSTALL" = true ]; then
         run_install_cmd "install unsloth (auto torch backend)" uv pip install --python "$_VENV_PY" unsloth-zoo "unsloth>=2026.4.7" --torch-backend=auto
         substep "overlaying local repo (editable)..."
         run_install_cmd "overlay local repo" uv pip install --python "$_VENV_PY" -e "$_REPO_ROOT" --no-deps
     else
-        run_install_cmd "install unsloth (auto torch backend)" uv pip install --python "$_VENV_PY" "$PACKAGE_NAME" --torch-backend=auto
+        run_install_cmd "install unsloth (auto torch backend)" uv pip install --python "$_VENV_PY" --torch-backend=auto -- "$PACKAGE_NAME"
     fi
 fi
 
 # ── Run studio setup ──
+tauri_log "STEP" "Running Studio setup"
 # When --local, use the repo's own setup.sh directly.
 # Otherwise, find it inside the installed package.
 SETUP_SH=""
@@ -1554,6 +1588,7 @@ if [ -z "$SETUP_SH" ] || [ ! -f "$SETUP_SH" ]; then
 fi
 
 if [ -z "$SETUP_SH" ] || [ ! -f "$SETUP_SH" ]; then
+    tauri_log "ERROR" "Could not find studio/setup.sh in the installed package"
     echo "❌ ERROR: Could not find studio/setup.sh in the installed package."
     exit 1
 fi
@@ -1571,16 +1606,16 @@ if ! command -v bash >/dev/null 2>&1; then
 fi
 
 step "setup" "running unsloth studio update..."
-# install.sh already installs base packages (unsloth + unsloth-zoo) and
-# no-torch-runtime.txt above, so tell install_python_stack.py to skip
-# the base step to avoid redundant reinstallation.
 _SKIP_BASE=1
-# Run setup.sh outside set -e so that a llama.cpp build failure (exit 1)
-# does not skip PATH setup, shortcuts, and launch below.  We capture the
-# exit code and propagate it after post-install steps finish.
 _SETUP_EXIT=0
+# Tauri desktop app bundles its own frontend — skip Node/npm/frontend build
+_SKIP_FRONTEND=0
+if [ "$TAURI_MODE" = true ]; then
+    _SKIP_FRONTEND=1
+fi
 if [ "$STUDIO_LOCAL_INSTALL" = true ]; then
     SKIP_STUDIO_BASE="$_SKIP_BASE" \
+    SKIP_STUDIO_FRONTEND="$_SKIP_FRONTEND" \
     STUDIO_PACKAGE_NAME="$PACKAGE_NAME" \
     STUDIO_LOCAL_INSTALL=1 \
     STUDIO_LOCAL_REPO="$_REPO_ROOT" \
@@ -1593,6 +1628,7 @@ else
     # local-dev path in setup.sh and install_python_stack.py. Mirrors the
     # reset already done in install.ps1 for PowerShell.
     SKIP_STUDIO_BASE="$_SKIP_BASE" \
+    SKIP_STUDIO_FRONTEND="$_SKIP_FRONTEND" \
     STUDIO_PACKAGE_NAME="$PACKAGE_NAME" \
     STUDIO_LOCAL_INSTALL=0 \
     STUDIO_LOCAL_REPO= \
@@ -1629,23 +1665,26 @@ case ":$PATH:" in
         ;;
 esac
 
-create_studio_shortcuts "$VENV_ABS_BIN/unsloth" "$OS"
+# Non-Tauri installs keep shortcuts even if setup reports failure.
+if [ "$TAURI_MODE" != true ]; then
+    create_studio_shortcuts "$VENV_ABS_BIN/unsloth" "$OS"
+fi
 
 # If setup.sh failed, report and exit now.
 # PATH and shortcuts are already set up so the user can fix and retry.
 if [ "$_SETUP_EXIT" -ne 0 ]; then
     echo ""
     step "error" "studio setup failed (exit code $_SETUP_EXIT)" "$C_ERR"
-    substep "Check the output above for details, then re-run:"
-    if [ "$STUDIO_LOCAL_INSTALL" = true ]; then
-        substep "  unsloth studio update --local"
-    else
-        substep "  unsloth studio update"
-    fi
     echo ""
     exit "$_SETUP_EXIT"
 fi
 
+# ── Tauri mode: done, skip shortcuts and auto-launch ──
+if [ "$TAURI_MODE" = true ]; then
+    tauri_log "DONE" ""
+    exit 0
+fi
+
 echo ""
 printf "  ${C_TITLE}%s${C_RST}\n" "Unsloth Studio installed!"
 printf "  ${C_DIM}%s${C_RST}\n" "$RULE"
diff --git a/studio/backend/auth/authentication.py b/studio/backend/auth/authentication.py
index da59ba9a1a..6ddcbc8e0b 100644
--- a/studio/backend/auth/authentication.py
+++ b/studio/backend/auth/authentication.py
@@ -52,6 +52,8 @@ def _decode_subject_without_verification(token: str) -> Optional[str]:
 def create_access_token(
     subject: str,
     expires_delta: Optional[timedelta] = None,
+    *,
+    desktop: bool = False,
 ) -> str:
     """
     Create a signed JWT for the given subject (e.g. username).
@@ -59,6 +61,8 @@ def create_access_token(
     Tokens are valid across restarts because the signing secret is stored in SQLite.
     """
     to_encode = {"sub": subject}
+    if desktop:
+        to_encode["desktop"] = True
     expire = datetime.now(timezone.utc) + (
         expires_delta or timedelta(minutes = ACCESS_TOKEN_EXPIRE_MINUTES)
     )
@@ -70,7 +74,29 @@ def create_access_token(
     )
 
 
-def create_refresh_token(subject: str) -> str:
+def is_desktop_access_token(token: str) -> bool:
+    """Return true only for a valid desktop-issued JWT access token."""
+    if token.startswith(API_KEY_PREFIX):
+        return False
+
+    subject = _decode_subject_without_verification(token)
+    if subject is None:
+        return False
+
+    record = get_user_and_secret(subject)
+    if record is None:
+        return False
+
+    _salt, _pwd_hash, jwt_secret, _must_change_password = record
+    try:
+        payload = jwt.decode(token, jwt_secret, algorithms = [ALGORITHM])
+    except jwt.InvalidTokenError:
+        return False
+
+    return payload.get("sub") == subject and payload.get("desktop") is True
+
+
+def create_refresh_token(subject: str, *, desktop: bool = False) -> str:
     """
     Create a random refresh token, store its hash in SQLite, and return it.
 
@@ -78,21 +104,28 @@ def create_refresh_token(subject: str) -> str:
     """
     token = secrets.token_urlsafe(48)
     expires_at = datetime.now(timezone.utc) + timedelta(days = REFRESH_TOKEN_EXPIRE_DAYS)
-    save_refresh_token(token, subject, expires_at.isoformat())
+    save_refresh_token(token, subject, expires_at.isoformat(), is_desktop = desktop)
     return token
 
 
-def refresh_access_token(refresh_token: str) -> Tuple[Optional[str], Optional[str]]:
+def refresh_access_token(
+    refresh_token: str,
+) -> Tuple[Optional[str], Optional[str], bool]:
     """
     Validate a refresh token and issue a new access token.
 
     The refresh token itself is NOT consumed — it stays valid until expiry.
     Returns a new access_token or None if the refresh token is invalid/expired.
     """
-    username = verify_refresh_token(refresh_token)
-    if username is None:
-        return None, None
-    return create_access_token(subject = username), username
+    verified = verify_refresh_token(refresh_token)
+    if verified is None:
+        return None, None, False
+    username, is_desktop = verified
+    return (
+        create_access_token(subject = username, desktop = is_desktop),
+        username,
+        is_desktop,
+    )
 
 
 def reload_secret() -> None:
@@ -173,7 +206,8 @@ async def _get_current_subject(
                 status_code = status.HTTP_401_UNAUTHORIZED,
                 detail = "Invalid token payload",
             )
-        if must_change_password and not allow_password_change:
+        is_desktop = payload.get("desktop") is True
+        if must_change_password and not allow_password_change and not is_desktop:
             raise HTTPException(
                 status_code = status.HTTP_403_FORBIDDEN,
                 detail = "Password change required",
diff --git a/studio/backend/auth/storage.py b/studio/backend/auth/storage.py
index 7d55a2dc59..2b0e359d39 100644
--- a/studio/backend/auth/storage.py
+++ b/studio/backend/auth/storage.py
@@ -6,6 +6,7 @@ SQLite storage for authentication data (user credentials + JWT secret).
 """
 
 import hashlib
+import os
 import secrets
 import sqlite3
 from datetime import datetime, timezone
@@ -54,6 +55,10 @@ def generate_bootstrap_password() -> str:
     # before the user changes the password.
     ensure_dir(_BOOTSTRAP_PW_PATH.parent)
     _BOOTSTRAP_PW_PATH.write_text(_bootstrap_password)
+    try:
+        os.chmod(_BOOTSTRAP_PW_PATH, 0o600)
+    except OSError:
+        pass
 
     return _bootstrap_password
 
@@ -63,6 +68,17 @@ def get_bootstrap_password() -> Optional[str]:
     return _bootstrap_password
 
 
+def _load_bootstrap_password() -> Optional[str]:
+    """Load an existing bootstrap password without creating one."""
+    global _bootstrap_password
+    _bootstrap_password = None
+    if _BOOTSTRAP_PW_PATH.is_file():
+        bootstrap_password = _BOOTSTRAP_PW_PATH.read_text().strip()
+        if bootstrap_password:
+            _bootstrap_password = bootstrap_password
+    return _bootstrap_password
+
+
 def clear_bootstrap_password() -> None:
     """Delete the persisted bootstrap password file (called after password change)."""
     global _bootstrap_password
@@ -114,7 +130,8 @@ def get_connection() -> sqlite3.Connection:
             id INTEGER PRIMARY KEY,
             token_hash TEXT NOT NULL,
             username TEXT NOT NULL,
-            expires_at TEXT NOT NULL
+            expires_at TEXT NOT NULL,
+            is_desktop INTEGER NOT NULL DEFAULT 0
         );
         """
     )
@@ -146,6 +163,13 @@ def get_connection() -> sqlite3.Connection:
         conn.execute(
             "ALTER TABLE auth_user ADD COLUMN must_change_password INTEGER NOT NULL DEFAULT 0"
         )
+    refresh_columns = {
+        row["name"] for row in conn.execute("PRAGMA table_info(refresh_tokens)")
+    }
+    if "is_desktop" not in refresh_columns:
+        conn.execute(
+            "ALTER TABLE refresh_tokens ADD COLUMN is_desktop INTEGER NOT NULL DEFAULT 0"
+        )
     conn.commit()
     return conn
 
@@ -201,6 +225,9 @@ def _get_or_create_api_key_pbkdf2_salt() -> bytes:
 
 
 _API_KEY_PBKDF2_ITERATIONS = 100_000
+DESKTOP_SECRET_PREFIX = "desktop-"
+_DESKTOP_SECRET_HASH_KEY = "desktop_secret_hash"
+_DESKTOP_SECRET_CREATED_AT_KEY = "desktop_secret_created_at"
 
 
 def _pbkdf2_api_key(raw_key: str) -> str:
@@ -233,6 +260,10 @@ def _pbkdf2_api_key(raw_key: str) -> str:
     return dk.hex()
 
 
+def _pbkdf2_desktop_secret(raw_secret: str) -> str:
+    return _pbkdf2_api_key(raw_secret)
+
+
 def is_initialized() -> bool:
     """Check if auth is ready for login (at least one user exists in DB)."""
     conn = get_connection()
@@ -374,6 +405,10 @@ def ensure_default_admin() -> bool:
     Uses a randomly generated diceware passphrase as the bootstrap password.
     Returns True when the default admin was created in this call.
     """
+    if get_user_and_secret(DEFAULT_ADMIN_USERNAME) is not None:
+        _load_bootstrap_password()
+        return False
+
     bootstrap_pw = generate_bootstrap_password()
     try:
         create_initial_user(
@@ -406,12 +441,19 @@ def update_password(username: str, new_password: str) -> bool:
         conn.commit()
         if cursor.rowcount > 0:
             clear_bootstrap_password()
+            clear_desktop_secret()
         return cursor.rowcount > 0
     finally:
         conn.close()
 
 
-def save_refresh_token(token: str, username: str, expires_at: str) -> None:
+def save_refresh_token(
+    token: str,
+    username: str,
+    expires_at: str,
+    *,
+    is_desktop: bool = False,
+) -> None:
     """
     Store a hashed refresh token with its associated username and expiry.
     """
@@ -420,21 +462,21 @@ def save_refresh_token(token: str, username: str, expires_at: str) -> None:
     try:
         conn.execute(
             """
-            INSERT INTO refresh_tokens (token_hash, username, expires_at)
-            VALUES (?, ?, ?)
+            INSERT INTO refresh_tokens (token_hash, username, expires_at, is_desktop)
+            VALUES (?, ?, ?, ?)
             """,
-            (token_hash, username, expires_at),
+            (token_hash, username, expires_at, int(is_desktop)),
         )
         conn.commit()
     finally:
         conn.close()
 
 
-def verify_refresh_token(token: str) -> Optional[str]:
+def verify_refresh_token(token: str) -> Optional[Tuple[str, bool]]:
     """
-    Verify a refresh token and return the username.
+    Verify a refresh token and return the username plus desktop marker.
 
-    Returns the username if valid and not expired, None otherwise.
+    Returns the username and desktop marker if valid and not expired, None otherwise.
     The token is NOT consumed — it stays valid until it expires.
     """
     token_hash = _hash_token(token)
@@ -449,7 +491,7 @@ def verify_refresh_token(token: str) -> Optional[str]:
 
         cur = conn.execute(
             """
-            SELECT id, username, expires_at FROM refresh_tokens
+            SELECT id, username, expires_at, is_desktop FROM refresh_tokens
             WHERE token_hash = ?
             """,
             (token_hash,),
@@ -465,7 +507,7 @@ def verify_refresh_token(token: str) -> Optional[str]:
             conn.commit()
             return None
 
-        return row["username"]
+        return row["username"], bool(row["is_desktop"])
     finally:
         conn.close()
 
@@ -480,6 +522,65 @@ def revoke_user_refresh_tokens(username: str) -> None:
         conn.close()
 
 
+def create_desktop_secret() -> str:
+    """Create/rotate the local desktop credential and return it once."""
+    ensure_default_admin()
+    raw_secret = DESKTOP_SECRET_PREFIX + secrets.token_urlsafe(48)
+    secret_hash = _pbkdf2_desktop_secret(raw_secret)
+    now = datetime.now(timezone.utc).isoformat()
+    conn = get_connection()
+    try:
+        conn.execute(
+            "INSERT OR REPLACE INTO app_secrets (key, value) VALUES (?, ?)",
+            (_DESKTOP_SECRET_HASH_KEY, secret_hash),
+        )
+        conn.execute(
+            "INSERT OR REPLACE INTO app_secrets (key, value) VALUES (?, ?)",
+            (_DESKTOP_SECRET_CREATED_AT_KEY, now),
+        )
+        conn.commit()
+        return raw_secret
+    finally:
+        conn.close()
+
+
+def validate_desktop_secret(raw_secret: str) -> Optional[str]:
+    """Return the real admin username when the desktop secret matches."""
+    if not raw_secret.startswith(DESKTOP_SECRET_PREFIX):
+        return None
+    if get_user_and_secret(DEFAULT_ADMIN_USERNAME) is None:
+        return None
+
+    secret_hash = _pbkdf2_desktop_secret(raw_secret)
+    conn = get_connection()
+    try:
+        cur = conn.execute(
+            "SELECT value FROM app_secrets WHERE key = ?",
+            (_DESKTOP_SECRET_HASH_KEY,),
+        )
+        row = cur.fetchone()
+        if row is None:
+            return None
+        if not secrets.compare_digest(row["value"], secret_hash):
+            return None
+        return DEFAULT_ADMIN_USERNAME
+    finally:
+        conn.close()
+
+
+def clear_desktop_secret() -> None:
+    """Remove backend-side desktop auth state."""
+    conn = get_connection()
+    try:
+        conn.execute(
+            "DELETE FROM app_secrets WHERE key IN (?, ?)",
+            (_DESKTOP_SECRET_HASH_KEY, _DESKTOP_SECRET_CREATED_AT_KEY),
+        )
+        conn.commit()
+    finally:
+        conn.close()
+
+
 # ---------------------------------------------------------------------------
 # API key management
 # ---------------------------------------------------------------------------
diff --git a/studio/backend/core/data_recipe/local_callable_validators.py b/studio/backend/core/data_recipe/local_callable_validators.py
index c32b2fccaf..afd10b02d1 100644
--- a/studio/backend/core/data_recipe/local_callable_validators.py
+++ b/studio/backend/core/data_recipe/local_callable_validators.py
@@ -33,6 +33,11 @@ _OXC_TOOL_DIR = Path(__file__).resolve().parent / "oxc-validator"
 _OXC_RUNNER_PATH = _OXC_TOOL_DIR / "validate.mjs"
 
 
+from utils.subprocess_compat import (
+    windows_hidden_subprocess_kwargs as _windows_hidden_subprocess_kwargs,
+)
+
+
 @dataclass(frozen = True)
 class OxcLocalCallableValidatorSpec:
     name: str
@@ -256,6 +261,7 @@ def _run_oxc_batch(
             capture_output = True,
             check = False,
             env = env,
+            **_windows_hidden_subprocess_kwargs(),
         )
     except (OSError, ValueError) as exc:
         logger.warning("OXC subprocess launch failed: %s", exc)
diff --git a/studio/backend/core/inference/audio_codecs.py b/studio/backend/core/inference/audio_codecs.py
index bcf3ec2937..895b112e85 100644
--- a/studio/backend/core/inference/audio_codecs.py
+++ b/studio/backend/core/inference/audio_codecs.py
@@ -8,6 +8,7 @@ Supports: SNAC (Orpheus), CSM (Sesame), BiCodec (Spark), DAC (OuteTTS)
 
 import io
 import re
+import subprocess
 import wave
 import structlog
 from loggers import get_logger
@@ -16,6 +17,10 @@ from typing import Optional, Tuple
 import numpy as np
 import torch
 
+from utils.subprocess_compat import (
+    windows_hidden_subprocess_kwargs as _windows_hidden_subprocess_kwargs,
+)
+
 logger = get_logger(__name__)
 
 
@@ -81,7 +86,6 @@ class AudioCodecManager:
             return
         import os
         import sys
-        import subprocess
 
         # Clone SparkAudio/Spark-TTS GitHub repo for the sparktts Python package
         # (same approach as training — the HF model repos don't contain the package)
@@ -101,6 +105,7 @@ class AudioCodecManager:
                     spark_code_dir,
                 ],
                 check = True,
+                **_windows_hidden_subprocess_kwargs(),
             )
 
         if spark_code_dir not in sys.path:
@@ -119,7 +124,6 @@ class AudioCodecManager:
             return
         import os
         import sys
-        import subprocess
 
         # Clone OuteTTS repo (same pattern as Spark-TTS / BiCodec)
         # The pip package has problematic dependencies; the notebook clones and
@@ -139,6 +143,7 @@ class AudioCodecManager:
                     outetts_code_dir,
                 ],
                 check = True,
+                **_windows_hidden_subprocess_kwargs(),
             )
             # Remove files that pull in heavy / incompatible dependencies
             # (matches notebook: gguf_model.py is under models/, others under outetts/)
diff --git a/studio/backend/core/inference/llama_cpp.py b/studio/backend/core/inference/llama_cpp.py
index 2e26995309..6ad05ac52d 100644
--- a/studio/backend/core/inference/llama_cpp.py
+++ b/studio/backend/core/inference/llama_cpp.py
@@ -18,6 +18,7 @@ from loggers import get_logger
 import shutil
 import socket
 import subprocess
+import sys
 import threading
 import time
 from pathlib import Path
@@ -26,8 +27,13 @@ from urllib.parse import urlparse
 
 import httpx
 
+from utils.subprocess_compat import (
+    windows_hidden_subprocess_kwargs as _windows_hidden_subprocess_kwargs,
+)
+
 logger = get_logger(__name__)
 
+
 # ── Pre-compiled patterns for plan-without-action re-prompt ──
 # Forward-looking intent signals that indicate the model is
 # describing what it *will* do rather than giving a final answer.
@@ -440,6 +446,7 @@ class LlamaCppBackend:
                 capture_output = True,
                 text = True,
                 timeout = 10,
+                **_windows_hidden_subprocess_kwargs(),
             )
             if result.returncode != 0:
                 return []
@@ -1658,6 +1665,7 @@ class LlamaCppBackend:
                 stderr = subprocess.STDOUT,
                 text = True,
                 env = env,
+                **_windows_hidden_subprocess_kwargs(),
             )
 
             # Start background thread to drain stdout and prevent pipe deadlock
diff --git a/studio/backend/core/training/trainer.py b/studio/backend/core/training/trainer.py
index 77cbda6b45..c1e2ac4a85 100644
--- a/studio/backend/core/training/trainer.py
+++ b/studio/backend/core/training/trainer.py
@@ -49,6 +49,7 @@ from unsloth.chat_templates import get_chat_template
 import json
 import threading
 import math
+import subprocess
 import structlog
 from loggers import get_logger
 import time
@@ -69,6 +70,10 @@ from utils.paths import (
 )
 from trl import SFTTrainer, SFTConfig
 
+from utils.subprocess_compat import (
+    windows_hidden_subprocess_kwargs as _windows_hidden_subprocess_kwargs,
+)
+
 logger = get_logger(__name__)
 
 
@@ -1765,6 +1770,7 @@ class UnslothTrainer:
                     spark_code_dir,
                 ],
                 check = True,
+                **_windows_hidden_subprocess_kwargs(),
             )
 
         if spark_code_dir not in sys.path:
@@ -1982,8 +1988,6 @@ class UnslothTrainer:
         device = "cuda" if torch.cuda.is_available() else "cpu"
 
         # Clone OuteTTS repo (same as audio_codecs._load_dac)
-        import subprocess
-
         base_dir = os.path.dirname(os.path.abspath(__file__))
         outetts_code_dir = os.path.join(base_dir, "inference", "OuteTTS")
         outetts_pkg = os.path.join(outetts_code_dir, "outetts")
@@ -2000,6 +2004,7 @@ class UnslothTrainer:
                     outetts_code_dir,
                 ],
                 check = True,
+                **_windows_hidden_subprocess_kwargs(),
             )
             for fpath in [
                 os.path.join(outetts_pkg, "models", "gguf_model.py"),
diff --git a/studio/backend/loggers/config.py b/studio/backend/loggers/config.py
index 0d32a64657..4c0d8ade28 100644
--- a/studio/backend/loggers/config.py
+++ b/studio/backend/loggers/config.py
@@ -44,6 +44,14 @@ class LogConfig:
         # Fallback to INFO if an invalid level is provided
         log_level = getattr(logging, log_level_name, logging.INFO)
 
+        if sys.platform == "win32":
+            for stream in (sys.stdout, sys.stderr):
+                if hasattr(stream, "reconfigure"):
+                    try:
+                        stream.reconfigure(encoding = "utf-8", errors = "replace")
+                    except Exception:
+                        pass
+
         structlog.configure(
             processors = [
                 # Reorder processors to control field order
diff --git a/studio/backend/main.py b/studio/backend/main.py
index d146a8ef12..05adcaa2ea 100644
--- a/studio/backend/main.py
+++ b/studio/backend/main.py
@@ -179,9 +179,22 @@ logger = LogConfig.setup_logging(
 app.add_middleware(LoggingMiddleware)
 
 # CORS middleware
+_api_only = os.environ.get("UNSLOTH_API_ONLY") == "1"
+_cors_origins = ["*"]
+if _api_only:
+    _cors_origins = [
+        "tauri://localhost",  # Linux/macOS Tauri webview
+        "http://tauri.localhost",  # Windows Tauri webview
+        "http://localhost",  # dev fallback
+    ]
+    _cors_origin_regex = None
+else:
+    _cors_origin_regex = None
+
 app.add_middleware(
     CORSMiddleware,
-    allow_origins = ["*"],  # In production, specify allowed origins
+    allow_origins = _cors_origins,
+    allow_origin_regex = _cors_origin_regex,
     allow_credentials = True,
     allow_methods = ["*"],
     allow_headers = ["*"],
@@ -223,6 +236,8 @@ async def health_check():
         "version": UNSLOTH_VERSION,
         "device_type": device_type,
         "chat_only": _hw_module.CHAT_ONLY,
+        "desktop_protocol_version": 1,
+        "supports_desktop_auth": True,
     }
 
 
diff --git a/studio/backend/models/auth.py b/studio/backend/models/auth.py
index c55e646508..23eb0ac4c0 100644
--- a/studio/backend/models/auth.py
+++ b/studio/backend/models/auth.py
@@ -17,6 +17,12 @@ class AuthLoginRequest(BaseModel):
     password: str = Field(..., description = "Password")
 
 
+class DesktopLoginRequest(BaseModel):
+    """Desktop-only local secret exchange payload."""
+
+    secret: str = Field(..., description = "Desktop local auth secret")
+
+
 class RefreshTokenRequest(BaseModel):
     """Refresh token payload to obtain new access + refresh tokens."""
 
diff --git a/studio/backend/routes/auth.py b/studio/backend/routes/auth.py
index 5cd23bd450..3deeb6793b 100644
--- a/studio/backend/routes/auth.py
+++ b/studio/backend/routes/auth.py
@@ -17,6 +17,7 @@ from models.auth import (
     ChangePasswordRequest,
     CreateApiKeyRequest,
     CreateApiKeyResponse,
+    DesktopLoginRequest,
     RefreshTokenRequest,
 )
 from models.users import Token
@@ -80,6 +81,24 @@ async def login(payload: AuthLoginRequest) -> Token:
     )
 
 
+@router.post("/desktop-login", response_model = Token)
+async def desktop_login(payload: DesktopLoginRequest) -> Token:
+    """Exchange a local desktop secret for normal admin-subject tokens."""
+    username = storage.validate_desktop_secret(payload.secret)
+    if username is None:
+        raise HTTPException(
+            status_code = status.HTTP_401_UNAUTHORIZED,
+            detail = "Desktop authentication failed",
+        )
+
+    return Token(
+        access_token = create_access_token(subject = username, desktop = True),
+        refresh_token = create_refresh_token(subject = username, desktop = True),
+        token_type = "bearer",
+        must_change_password = False,
+    )
+
+
 @router.post("/refresh", response_model = Token)
 async def refresh(payload: RefreshTokenRequest) -> Token:
     """
@@ -87,7 +106,7 @@ async def refresh(payload: RefreshTokenRequest) -> Token:
 
     The refresh token itself is reusable until it expires (7 days).
     """
-    new_access_token, username = refresh_access_token(payload.refresh_token)
+    new_access_token, username, is_desktop = refresh_access_token(payload.refresh_token)
     if new_access_token is None or username is None:
         raise HTTPException(
             status_code = status.HTTP_401_UNAUTHORIZED,
@@ -98,7 +117,9 @@ async def refresh(payload: RefreshTokenRequest) -> Token:
         access_token = new_access_token,
         refresh_token = payload.refresh_token,
         token_type = "bearer",
-        must_change_password = storage.requires_password_change(username),
+        must_change_password = False
+        if is_desktop
+        else storage.requires_password_change(username),
     )
 
 
diff --git a/studio/backend/routes/data_recipe/jobs.py b/studio/backend/routes/data_recipe/jobs.py
index 00546b47a4..606ef1832c 100644
--- a/studio/backend/routes/data_recipe/jobs.py
+++ b/studio/backend/routes/data_recipe/jobs.py
@@ -57,6 +57,20 @@ def _resolve_local_v1_endpoint(request: Request) -> str:
     return f"http://127.0.0.1:{int(port)}/v1"
 
 
+def _request_has_desktop_access_token(request: Request) -> bool:
+    auth_header = request.headers.get("authorization")
+    if not auth_header:
+        return False
+
+    parts = auth_header.split(None, 1)
+    if len(parts) != 2 or parts[0].lower() != "bearer":
+        return False
+
+    from auth.authentication import is_desktop_access_token
+
+    return is_desktop_access_token(parts[1])
+
+
 def _used_llm_model_aliases(recipe: dict[str, Any]) -> set[str]:
     """Return the set of model_aliases that are actually referenced by an
     LLM column. Used to narrow the "Chat model loaded" gate so that orphan
@@ -154,6 +168,7 @@ def _inject_local_providers(recipe: dict[str, Any], request: Request) -> None:
         token = create_access_token(
             subject = "unsloth",
             expires_delta = timedelta(hours = 24),
+            desktop = _request_has_desktop_access_token(request),
         )
 
     # Defensively strip any stale "external"-only fields the frontend may
diff --git a/studio/backend/run.py b/studio/backend/run.py
index 9675b9ea4c..7590ef1067 100644
--- a/studio/backend/run.py
+++ b/studio/backend/run.py
@@ -248,6 +248,7 @@ def run_server(
     port: int = 8888,
     frontend_path: Path = Path(__file__).resolve().parent.parent / "frontend" / "dist",
     silent: bool = False,
+    api_only: bool = False,
     llama_parallel_slots: int = 1,
 ):
     """
@@ -258,6 +259,7 @@ def run_server(
         port: Port to bind to (auto-increments if in use)
         frontend_path: Path to frontend build directory (optional)
         silent: Suppress startup messages
+        api_only: Run API server only, no frontend serving (for Tauri desktop app)
         llama_parallel_slots: Number of parallel slots for llama-server
 
     Note:
@@ -275,6 +277,10 @@ def run_server(
         except Exception:
             pass
 
+    # Set env var BEFORE importing main so CORS middleware picks it up
+    if api_only:
+        os.environ["UNSLOTH_API_ONLY"] = "1"
+
     import nest_asyncio
 
     nest_asyncio.apply()
@@ -310,8 +316,12 @@ def run_server(
             print("=" * 50)
             print("")
 
-    # Setup frontend if path provided
-    if frontend_path:
+    # Output port for Tauri to parse when in api-only mode
+    if api_only:
+        print(f"TAURI_PORT={port}", flush = True)
+
+    # Setup frontend if path provided (skip in api-only mode)
+    if frontend_path and not api_only:
         if setup_frontend(app, frontend_path):
             if not silent:
                 print(f"[OK] Frontend loaded from {frontend_path}")
@@ -391,10 +401,17 @@ if __name__ == "__main__":
         help = "Path to frontend build",
     )
     parser.add_argument("--silent", action = "store_true", help = "Suppress output")
+    parser.add_argument(
+        "--api-only",
+        action = "store_true",
+        help = "API server only, no frontend (for Tauri)",
+    )
 
     args = parser.parse_args()
 
-    kwargs = dict(host = args.host, port = args.port, silent = args.silent)
+    kwargs = dict(
+        host = args.host, port = args.port, silent = args.silent, api_only = args.api_only
+    )
     if args.frontend is not None:
         kwargs["frontend_path"] = Path(args.frontend)
 
diff --git a/studio/backend/tests/test_desktop_auth.py b/studio/backend/tests/test_desktop_auth.py
new file mode 100644
index 0000000000..c8cf1c7081
--- /dev/null
+++ b/studio/backend/tests/test_desktop_auth.py
@@ -0,0 +1,597 @@
+import importlib.util
+import asyncio
+import hashlib
+import json
+import os
+import platform
+import secrets
+import sqlite3
+import subprocess
+import sys
+from pathlib import Path
+from types import SimpleNamespace
+
+import jwt
+import pytest
+from fastapi import APIRouter, FastAPI
+from fastapi.security import HTTPAuthorizationCredentials
+from fastapi.testclient import TestClient
+
+from auth import storage
+
+
+@pytest.fixture(autouse = True)
+def isolated_auth_db(tmp_path, monkeypatch):
+    monkeypatch.setattr(storage, "DB_PATH", tmp_path / "auth.db")
+    monkeypatch.setattr(storage, "_BOOTSTRAP_PW_PATH", tmp_path / ".bootstrap_password")
+    monkeypatch.setattr(storage, "_bootstrap_password", None)
+    monkeypatch.setattr(storage, "_api_key_pbkdf2_salt_cache", None)
+    yield
+
+
+def seed_user(*, must_change_password = False):
+    storage.create_initial_user(
+        username = storage.DEFAULT_ADMIN_USERNAME,
+        password = "human-password-123",
+        jwt_secret = secrets.token_urlsafe(64),
+        must_change_password = must_change_password,
+    )
+
+
+def auth_client():
+    route_path = Path(__file__).resolve().parents[1] / "routes" / "auth.py"
+    spec = importlib.util.spec_from_file_location("_desktop_auth_route", route_path)
+    auth_route = importlib.util.module_from_spec(spec)
+    assert spec.loader is not None
+    spec.loader.exec_module(auth_route)
+
+    app = FastAPI()
+    app.include_router(auth_route.router, prefix = "/api/auth")
+    return TestClient(app)
+
+
+def data_recipe_jobs_module():
+    route_path = (
+        Path(__file__).resolve().parents[1] / "routes" / "data_recipe" / "jobs.py"
+    )
+    spec = importlib.util.spec_from_file_location(
+        "_desktop_data_recipe_jobs", route_path
+    )
+    jobs_route = importlib.util.module_from_spec(spec)
+    assert spec.loader is not None
+    spec.loader.exec_module(jobs_route)
+    return jobs_route
+
+
+def local_recipe():
+    return {
+        "model_providers": [{"name": "local", "is_local": True}],
+        "model_configs": [{"alias": "local-model", "provider": "local"}],
+        "columns": [{"column_type": "llm-text", "model_alias": "local-model"}],
+    }
+
+
+def local_recipe_request(token):
+    return SimpleNamespace(
+        headers = {"authorization": f"Bearer {token}"},
+        app = SimpleNamespace(state = SimpleNamespace(server_port = 8888)),
+        scope = {},
+        base_url = "http://testserver/",
+    )
+
+
+@pytest.fixture
+def loaded_local_model(monkeypatch):
+    inference_module = SimpleNamespace(
+        get_llama_cpp_backend = lambda: SimpleNamespace(is_loaded = True),
+    )
+    monkeypatch.setitem(sys.modules, "routes.inference", inference_module)
+
+
+def test_desktop_secret_round_trip_uses_real_admin_subject():
+    seed_user()
+    raw = storage.create_desktop_secret()
+
+    assert raw.startswith("desktop-")
+    assert storage.validate_desktop_secret(raw) == storage.DEFAULT_ADMIN_USERNAME
+    assert storage.validate_desktop_secret(raw + "x") is None
+
+
+def test_create_desktop_secret_rotates_old_secret():
+    seed_user()
+    old = storage.create_desktop_secret()
+    new = storage.create_desktop_secret()
+
+    assert old != new
+    assert storage.validate_desktop_secret(old) is None
+    assert storage.validate_desktop_secret(new) == storage.DEFAULT_ADMIN_USERNAME
+
+
+def test_clear_desktop_secret_invalidates_secret():
+    seed_user()
+    raw = storage.create_desktop_secret()
+
+    storage.clear_desktop_secret()
+
+    assert storage.validate_desktop_secret(raw) is None
+
+
+def test_ensure_default_admin_does_not_recreate_bootstrap_for_existing_admin():
+    seed_user()
+
+    created = storage.ensure_default_admin()
+
+    assert created is False
+    assert not storage._BOOTSTRAP_PW_PATH.exists()
+
+
+def test_ensure_default_admin_loads_existing_bootstrap_after_restart(monkeypatch):
+    created = storage.ensure_default_admin()
+    bootstrap_pw = storage._BOOTSTRAP_PW_PATH.read_text().strip()
+
+    monkeypatch.setattr(storage, "_bootstrap_password", None)
+    created_again = storage.ensure_default_admin()
+
+    assert created is True
+    assert storage._BOOTSTRAP_PW_PATH.exists()
+    assert created_again is False
+    assert storage.get_bootstrap_password() == bootstrap_pw
+
+
+def test_ensure_default_admin_does_not_generate_for_empty_existing_bootstrap():
+    seed_user()
+    storage._BOOTSTRAP_PW_PATH.write_text(" \n")
+
+    created = storage.ensure_default_admin()
+
+    assert created is False
+    assert storage._BOOTSTRAP_PW_PATH.read_text() == " \n"
+    assert storage.get_bootstrap_password() is None
+
+
+def test_web_login_token_has_no_desktop_marker_and_keeps_password_gate():
+    seed_user(must_change_password = True)
+    client = auth_client()
+
+    response = client.post(
+        "/api/auth/login",
+        json = {
+            "username": storage.DEFAULT_ADMIN_USERNAME,
+            "password": "human-password-123",
+        },
+    )
+
+    assert response.status_code == 200
+    body = response.json()
+    assert body["must_change_password"] is True
+    payload = jwt.decode(
+        body["access_token"],
+        storage.get_jwt_secret(storage.DEFAULT_ADMIN_USERNAME),
+        algorithms = ["HS256"],
+    )
+    assert payload["sub"] == storage.DEFAULT_ADMIN_USERNAME
+    assert "desktop" not in payload
+
+    gated = client.post(
+        "/api/auth/api-keys",
+        headers = {"Authorization": f"Bearer {body['access_token']}"},
+        json = {"name": "web"},
+    )
+    assert gated.status_code == 403
+
+
+def test_desktop_login_mints_admin_token_without_clearing_web_password_change():
+    seed_user(must_change_password = True)
+    raw = storage.create_desktop_secret()
+    client = auth_client()
+
+    response = client.post("/api/auth/desktop-login", json = {"secret": raw})
+
+    assert response.status_code == 200
+    body = response.json()
+    assert body["access_token"]
+    assert body["refresh_token"]
+    assert body["token_type"] == "bearer"
+    assert body["must_change_password"] is False
+    assert storage.requires_password_change(storage.DEFAULT_ADMIN_USERNAME) is True
+
+    payload = jwt.decode(
+        body["access_token"],
+        storage.get_jwt_secret(storage.DEFAULT_ADMIN_USERNAME),
+        algorithms = ["HS256"],
+    )
+    assert payload["sub"] == storage.DEFAULT_ADMIN_USERNAME
+    assert payload["desktop"] is True
+
+
+def test_desktop_refresh_preserves_desktop_marker():
+    seed_user(must_change_password = True)
+    raw = storage.create_desktop_secret()
+    client = auth_client()
+    login_body = client.post("/api/auth/desktop-login", json = {"secret": raw}).json()
+
+    response = client.post(
+        "/api/auth/refresh",
+        json = {"refresh_token": login_body["refresh_token"]},
+    )
+
+    assert response.status_code == 200
+    body = response.json()
+    assert body["must_change_password"] is False
+    payload = jwt.decode(
+        body["access_token"],
+        storage.get_jwt_secret(storage.DEFAULT_ADMIN_USERNAME),
+        algorithms = ["HS256"],
+    )
+    assert payload["sub"] == storage.DEFAULT_ADMIN_USERNAME
+    assert payload["desktop"] is True
+
+
+def test_desktop_session_uses_real_admin_identity_for_api_keys():
+    seed_user(must_change_password = True)
+    raw = storage.create_desktop_secret()
+    client = auth_client()
+    token = client.post("/api/auth/desktop-login", json = {"secret": raw}).json()[
+        "access_token"
+    ]
+
+    response = client.post(
+        "/api/auth/api-keys",
+        headers = {"Authorization": f"Bearer {token}"},
+        json = {"name": "desktop"},
+    )
+
+    assert response.status_code == 200
+    rows = storage.list_api_keys(storage.DEFAULT_ADMIN_USERNAME)
+    assert [row["name"] for row in rows] == ["desktop"]
+
+
+def test_local_recipe_token_preserves_desktop_marker(loaded_local_model):
+    from auth.authentication import create_access_token, get_current_subject
+
+    seed_user(must_change_password = True)
+    jobs_route = data_recipe_jobs_module()
+    incoming_token = create_access_token(
+        subject = storage.DEFAULT_ADMIN_USERNAME,
+        desktop = True,
+    )
+    recipe = local_recipe()
+
+    jobs_route._inject_local_providers(recipe, local_recipe_request(incoming_token))
+
+    local_token = recipe["model_providers"][0]["api_key"]
+    payload = jwt.decode(
+        local_token,
+        storage.get_jwt_secret(storage.DEFAULT_ADMIN_USERNAME),
+        algorithms = ["HS256"],
+    )
+    assert payload["sub"] == storage.DEFAULT_ADMIN_USERNAME
+    assert payload["desktop"] is True
+    credentials = HTTPAuthorizationCredentials(
+        scheme = "Bearer",
+        credentials = local_token,
+    )
+    assert (
+        asyncio.run(get_current_subject(credentials)) == storage.DEFAULT_ADMIN_USERNAME
+    )
+
+
+def test_local_recipe_token_keeps_web_marker_absent(loaded_local_model):
+    from auth.authentication import create_access_token
+
+    seed_user(must_change_password = False)
+    jobs_route = data_recipe_jobs_module()
+    incoming_token = create_access_token(subject = storage.DEFAULT_ADMIN_USERNAME)
+    recipe = local_recipe()
+
+    jobs_route._inject_local_providers(recipe, local_recipe_request(incoming_token))
+
+    local_token = recipe["model_providers"][0]["api_key"]
+    payload = jwt.decode(
+        local_token,
+        storage.get_jwt_secret(storage.DEFAULT_ADMIN_USERNAME),
+        algorithms = ["HS256"],
+    )
+    assert payload["sub"] == storage.DEFAULT_ADMIN_USERNAME
+    assert "desktop" not in payload
+
+
+def test_desktop_login_rejects_invalid_secret():
+    seed_user(must_change_password = False)
+    client = auth_client()
+
+    response = client.post(
+        "/api/auth/desktop-login",
+        json = {"secret": "desktop-invalid"},
+    )
+
+    assert response.status_code == 401
+
+
+def test_write_desktop_secret_file_is_0600_on_unix(tmp_path):
+    from unsloth_cli.commands import studio as studio_cli
+
+    path = tmp_path / ".desktop_secret"
+    if platform.system() != "Windows":
+        path.write_text("old-secret")
+        os.chmod(path, 0o644)
+
+    studio_cli._write_auth_secret(path, "desktop-secret")
+
+    assert path.read_text() == "desktop-secret"
+    if platform.system() != "Windows":
+        assert oct(path.stat().st_mode & 0o777) == "0o600"
+
+
+def test_reset_password_removes_desktop_secret_files(tmp_path, monkeypatch):
+    from typer.testing import CliRunner
+    from unsloth_cli.commands import studio as studio_cli
+
+    auth_dir = tmp_path / "auth"
+    auth_dir.mkdir()
+    (auth_dir / "auth.db").write_text("db")
+    (auth_dir / ".bootstrap_password").write_text("boot")
+    (auth_dir / ".desktop_secret").write_text("new")
+    monkeypatch.setattr(studio_cli, "STUDIO_HOME", tmp_path)
+
+    result = CliRunner().invoke(studio_cli.studio_app, ["reset-password"])
+
+    assert result.exit_code == 0
+    assert not (auth_dir / "auth.db").exists()
+    assert not (auth_dir / ".bootstrap_password").exists()
+    assert not (auth_dir / ".desktop_secret").exists()
+
+
+def test_reset_password_removes_desktop_secret_files_without_db(tmp_path, monkeypatch):
+    from typer.testing import CliRunner
+    from unsloth_cli.commands import studio as studio_cli
+
+    auth_dir = tmp_path / "auth"
+    auth_dir.mkdir()
+    (auth_dir / ".desktop_secret").write_text("new")
+    monkeypatch.setattr(studio_cli, "STUDIO_HOME", tmp_path)
+
+    result = CliRunner().invoke(studio_cli.studio_app, ["reset-password"])
+
+    assert result.exit_code == 0
+    assert not (auth_dir / ".desktop_secret").exists()
+
+
+def test_desktop_capabilities_json_reports_rollout_safe_flags():
+    from typer.testing import CliRunner
+    import unsloth_cli.commands.studio as studio_cli
+
+    result = CliRunner().invoke(
+        studio_cli.studio_app,
+        ["desktop-capabilities", "--json"],
+    )
+
+    assert result.exit_code == 0
+    body = json.loads(result.output)
+    assert body["desktop_protocol_version"] == 1
+    assert body["supports_provision_desktop_auth"] is True
+    assert body["supports_api_only"] is True
+    assert isinstance(body["version"], str)
+
+
+def test_health_response_reports_desktop_capability_fields(monkeypatch):
+    router_stub = SimpleNamespace(
+        auth_router = APIRouter(),
+        data_recipe_router = APIRouter(),
+        datasets_router = APIRouter(),
+        export_router = APIRouter(),
+        inference_router = APIRouter(),
+        models_router = APIRouter(),
+        training_history_router = APIRouter(),
+        training_router = APIRouter(),
+    )
+    monkeypatch.setitem(sys.modules, "routes", router_stub)
+
+    import studio.backend.main as backend_main
+
+    monkeypatch.setattr(backend_main._hw_module, "CHAT_ONLY", False)
+
+    body = asyncio.run(backend_main.health_check())
+
+    assert body["desktop_protocol_version"] == 1
+    assert body["supports_desktop_auth"] is True
+
+
+def test_provision_desktop_auth_writes_secret_and_creates_db_without_backend_deps(
+    tmp_path,
+    monkeypatch,
+):
+    auth_dir = tmp_path / "auth"
+    auth_dir.mkdir()
+
+    code = """
+import builtins
+import sys
+from pathlib import Path
+from typer.testing import CliRunner
+
+studio_home = Path(sys.argv[1])
+real_import = builtins.__import__
+
+def guarded_import(name, *args, **kwargs):
+    blocked = ("auth", "fastapi", "structlog", "utils")
+    if name in blocked or name.startswith(("auth.", "utils.")):
+        raise ModuleNotFoundError(name)
+    return real_import(name, *args, **kwargs)
+
+builtins.__import__ = guarded_import
+from unsloth_cli.commands import studio as studio_cli
+
+studio_cli.STUDIO_HOME = studio_home
+result = CliRunner().invoke(studio_cli.studio_app, ["provision-desktop-auth"])
+if result.exit_code != 0:
+    print(result.output)
+    if result.exception is not None:
+        raise result.exception
+    raise SystemExit(result.exit_code)
+"""
+    result = subprocess.run(
+        [sys.executable, "-c", code, str(tmp_path)],
+        cwd = Path(__file__).resolve().parents[3],
+        env = {**os.environ, "PYTHONPATH": "."},
+        text = True,
+        capture_output = True,
+    )
+    assert result.returncode == 0, result.stderr + result.stdout
+    secret = (auth_dir / ".desktop_secret").read_text()
+    assert secret.startswith("desktop-")
+
+    conn = sqlite3.connect(auth_dir / "auth.db")
+    conn.row_factory = sqlite3.Row
+    try:
+        user = conn.execute(
+            """
+            SELECT username, password_salt, password_hash, must_change_password
+            FROM auth_user
+            """
+        ).fetchone()
+        app_secrets = {
+            row["key"]: row["value"]
+            for row in conn.execute("SELECT key, value FROM app_secrets")
+        }
+        refresh_columns = {
+            row["name"] for row in conn.execute("PRAGMA table_info(refresh_tokens)")
+        }
+    finally:
+        conn.close()
+
+    bootstrap_password = (auth_dir / ".bootstrap_password").read_text().strip()
+    bootstrap_hash = hashlib.pbkdf2_hmac(
+        "sha256",
+        bootstrap_password.encode("utf-8"),
+        user["password_salt"].encode("utf-8"),
+        100_000,
+    ).hex()
+
+    assert bootstrap_password
+    assert user["username"] == "unsloth"
+    assert user["must_change_password"] == 1
+    assert bootstrap_hash == user["password_hash"]
+    assert len(app_secrets["api_key_pbkdf2_salt"]) == 64
+    assert len(app_secrets["desktop_secret_hash"]) == 64
+    assert app_secrets["desktop_secret_created_at"]
+    assert "is_desktop" in refresh_columns
+
+    monkeypatch.setattr(storage, "DB_PATH", auth_dir / "auth.db")
+    monkeypatch.setattr(storage, "_api_key_pbkdf2_salt_cache", None)
+    assert storage.validate_desktop_secret(secret) == storage.DEFAULT_ADMIN_USERNAME
+    assert storage.requires_password_change(storage.DEFAULT_ADMIN_USERNAME) is True
+
+
+def test_provision_desktop_auth_keeps_existing_admin_password(tmp_path, monkeypatch):
+    from typer.testing import CliRunner
+    from unsloth_cli.commands import studio as studio_cli
+
+    auth_dir = tmp_path / "auth"
+    auth_dir.mkdir()
+    monkeypatch.setattr(studio_cli, "STUDIO_HOME", tmp_path)
+
+    conn = sqlite3.connect(auth_dir / "auth.db")
+    try:
+        conn.execute(
+            """
+            CREATE TABLE auth_user (
+                id INTEGER PRIMARY KEY,
+                username TEXT UNIQUE NOT NULL,
+                password_salt TEXT NOT NULL,
+                password_hash TEXT NOT NULL,
+                jwt_secret TEXT NOT NULL,
+                must_change_password INTEGER NOT NULL DEFAULT 0
+            )
+            """
+        )
+        conn.execute(
+            """
+            INSERT INTO auth_user (
+                username, password_salt, password_hash, jwt_secret, must_change_password
+            )
+            VALUES (?, ?, ?, ?, ?)
+            """,
+            ("unsloth", "existing-salt", "existing-hash", "existing-jwt", 0),
+        )
+        conn.commit()
+    finally:
+        conn.close()
+
+    result = CliRunner().invoke(studio_cli.studio_app, ["provision-desktop-auth"])
+
+    assert result.exit_code == 0
+    assert not (auth_dir / ".bootstrap_password").exists()
+    conn = sqlite3.connect(auth_dir / "auth.db")
+    conn.row_factory = sqlite3.Row
+    try:
+        user = conn.execute(
+            """
+            SELECT password_salt, password_hash, jwt_secret, must_change_password
+            FROM auth_user WHERE username = ?
+            """,
+            ("unsloth",),
+        ).fetchone()
+    finally:
+        conn.close()
+
+    assert dict(user) == {
+        "password_salt": "existing-salt",
+        "password_hash": "existing-hash",
+        "jwt_secret": "existing-jwt",
+        "must_change_password": 0,
+    }
+
+
+def test_update_password_clears_desktop_secret():
+    seed_user()
+    raw = storage.create_desktop_secret()
+    assert storage.validate_desktop_secret(raw) == storage.DEFAULT_ADMIN_USERNAME
+
+    changed = storage.update_password(
+        storage.DEFAULT_ADMIN_USERNAME, "new-admin-password"
+    )
+    assert changed is True
+    assert storage.validate_desktop_secret(raw) is None
+
+
+def test_update_password_on_unknown_user_leaves_desktop_secret_intact():
+    seed_user()
+    raw = storage.create_desktop_secret()
+
+    changed = storage.update_password("not-a-user", "irrelevant")
+    assert changed is False
+    assert storage.validate_desktop_secret(raw) == storage.DEFAULT_ADMIN_USERNAME
+
+
+def test_desktop_auth_provision_has_bounded_timeout():
+    rs_path = (
+        Path(__file__).resolve().parents[3]
+        / "studio"
+        / "src-tauri"
+        / "src"
+        / "desktop_auth.rs"
+    )
+    src = rs_path.read_text()
+    start = src.index("async fn provision_desktop_auth(")
+    depth = 0
+    body_start = src.index("{", start)
+    body_end = None
+    for i in range(body_start, len(src)):
+        c = src[i]
+        if c == "{":
+            depth += 1
+        elif c == "}":
+            depth -= 1
+            if depth == 0:
+                body_end = i + 1
+                break
+    assert body_end is not None
+    body = src[start:body_end]
+    assert "tokio::time::timeout" in body
+    import re
+
+    m = re.search(r"Duration::from_secs\(\s*(\d+)\s*\)", body)
+    assert m is not None
+    seconds = int(m.group(1))
+    assert 5 <= seconds <= 120
diff --git a/studio/backend/utils/hardware/nvidia.py b/studio/backend/utils/hardware/nvidia.py
index dc5295c302..274d9beb48 100644
--- a/studio/backend/utils/hardware/nvidia.py
+++ b/studio/backend/utils/hardware/nvidia.py
@@ -6,6 +6,10 @@ from typing import Any, Optional
 
 from loggers import get_logger
 
+from utils.subprocess_compat import (
+    windows_hidden_subprocess_kwargs as _windows_hidden_subprocess_kwargs,
+)
+
 logger = get_logger(__name__)
 
 
@@ -61,6 +65,7 @@ def get_physical_gpu_count() -> Optional[int]:
             capture_output = True,
             text = True,
             timeout = 5,
+            **_windows_hidden_subprocess_kwargs(),
         )
         if result.returncode == 0 and result.stdout.strip():
             return len(result.stdout.strip().splitlines())
@@ -85,6 +90,7 @@ def get_primary_gpu_utilization() -> dict[str, Any]:
             capture_output = True,
             text = True,
             timeout = 5,
+            **_windows_hidden_subprocess_kwargs(),
         )
     except (OSError, subprocess.TimeoutExpired) as e:
         logger.warning("nvidia-smi query failed in get_primary_gpu_utilization: %s", e)
@@ -135,6 +141,7 @@ def get_visible_gpu_utilization(
             capture_output = True,
             text = True,
             timeout = 5,
+            **_windows_hidden_subprocess_kwargs(),
         )
     except (OSError, subprocess.TimeoutExpired) as e:
         logger.warning("nvidia-smi query failed in get_visible_gpu_utilization: %s", e)
@@ -220,6 +227,7 @@ def get_backend_visible_gpu_info(
             capture_output = True,
             text = True,
             timeout = 10,
+            **_windows_hidden_subprocess_kwargs(),
         )
     except (OSError, subprocess.TimeoutExpired) as e:
         logger.warning("nvidia-smi query failed in get_backend_visible_gpu_info: %s", e)
diff --git a/studio/backend/utils/models/model_config.py b/studio/backend/utils/models/model_config.py
index a2d48cf009..a2b0c90e59 100644
--- a/studio/backend/utils/models/model_config.py
+++ b/studio/backend/utils/models/model_config.py
@@ -32,6 +32,10 @@ import threading
 import yaml
 
 
+from utils.subprocess_compat import (
+    windows_hidden_subprocess_kwargs as _windows_hidden_subprocess_kwargs,
+)
+
 logger = get_logger(__name__)
 
 # ── Model size extraction ────────────────────────────────────
@@ -579,6 +583,7 @@ def _is_vision_model_subprocess(
             capture_output = True,
             text = True,
             timeout = 60,
+            **_windows_hidden_subprocess_kwargs(),
         )
 
         if result.returncode != 0:
diff --git a/studio/backend/utils/subprocess_compat.py b/studio/backend/utils/subprocess_compat.py
new file mode 100644
index 0000000000..bedf8cf2e6
--- /dev/null
+++ b/studio/backend/utils/subprocess_compat.py
@@ -0,0 +1,34 @@
+# SPDX-License-Identifier: AGPL-3.0-only
+# Copyright 2026-present the Unsloth AI Inc. team. All rights reserved. See /studio/LICENSE.AGPL-3.0
+
+"""Cross-platform subprocess helpers for the Unsloth Studio backend."""
+
+import subprocess
+import sys
+
+
+def windows_hidden_subprocess_kwargs() -> dict[str, object]:
+    """Return Windows-only subprocess kwargs that suppress console windows.
+
+    On non-Windows platforms returns an empty dict so callers can always
+    unpack the result into ``subprocess.run`` / ``subprocess.Popen`` via
+    ``**windows_hidden_subprocess_kwargs()``.
+    """
+    if sys.platform != "win32":
+        return {}
+
+    kwargs: dict[str, object] = {}
+    create_no_window = getattr(subprocess, "CREATE_NO_WINDOW", 0)
+    if create_no_window:
+        kwargs["creationflags"] = create_no_window
+
+    startupinfo_factory = getattr(subprocess, "STARTUPINFO", None)
+    startf_use_showwindow = getattr(subprocess, "STARTF_USESHOWWINDOW", 0)
+    sw_hide = getattr(subprocess, "SW_HIDE", 0)
+    if startupinfo_factory is not None and startf_use_showwindow:
+        startupinfo = startupinfo_factory()
+        startupinfo.dwFlags |= startf_use_showwindow
+        startupinfo.wShowWindow = sw_hide
+        kwargs["startupinfo"] = startupinfo
+
+    return kwargs
diff --git a/studio/backend/utils/transformers_version.py b/studio/backend/utils/transformers_version.py
index 36c3a4c22d..f36bdcd6e8 100644
--- a/studio/backend/utils/transformers_version.py
+++ b/studio/backend/utils/transformers_version.py
@@ -36,6 +36,10 @@ import subprocess
 import sys
 from pathlib import Path
 
+from utils.subprocess_compat import (
+    windows_hidden_subprocess_kwargs as _windows_hidden_subprocess_kwargs,
+)
+
 logger = get_logger(__name__)
 
 
@@ -499,6 +503,7 @@ def _install_to_dir(pkg: str, target_dir: str) -> bool:
             stdout = subprocess.PIPE,
             stderr = subprocess.STDOUT,
             text = True,
+            **_windows_hidden_subprocess_kwargs(),
         )
         if result.returncode == 0:
             return True
@@ -520,6 +525,7 @@ def _install_to_dir(pkg: str, target_dir: str) -> bool:
         stdout = subprocess.PIPE,
         stderr = subprocess.STDOUT,
         text = True,
+        **_windows_hidden_subprocess_kwargs(),
     )
     if result.returncode != 0:
         logger.error("install failed:\n%s", result.stdout)
diff --git a/studio/frontend/package.json b/studio/frontend/package.json
index a2eebd5cb5..c5cb949ccd 100644
--- a/studio/frontend/package.json
+++ b/studio/frontend/package.json
@@ -41,6 +41,10 @@
     "@tailwindcss/vite": "^4.2.2",
     "@tanstack/react-router": "^1.159.10",
     "@tanstack/react-table": "^8.21.3",
+    "@tauri-apps/api": "^2.10.1",
+    "@tauri-apps/plugin-opener": "^2.5.3",
+    "@tauri-apps/plugin-process": "^2.3.1",
+    "@tauri-apps/plugin-updater": "^2.10.1",
     "@toolwind/corner-shape": "^0.0.8-3",
     "@types/canvas-confetti": "^1.9.0",
     "@xyflow/react": "^12.10.0",
diff --git a/studio/frontend/public/studio.png b/studio/frontend/public/studio.png
new file mode 100644
index 0000000000000000000000000000000000000000..4e531499b7cf115cfa016e1b3976f6741fa6f0a2
GIT binary patch
literal 27420
zcmeFYRaBfy*DeY~u;A{YvEc5O;7)L9+}#@w65OG2Pk`VM+}&M6<8Fb*-PygqZ~gn6
zG0x4o+!y;D0~l0wS50|pR=v-h74cO`8V!XQ1r81lO;$!i6%Gz50S*pc6X`9i#VcG>
z1NP^msf?;39Go{T99%#!9NYt}Dc~<09Ec4L?$8JhP9O~qj=(9aMMVhKfnXvpEdlo$
z7B|)j`w!VkM#mKn4i)F0A9%QoEJ9c#qMNLuB;w&)cw}n+R8;sbSQEUXi;_D0WS&O`
z9Gvx<ti%^}&!yv4ueOxgwZZdAo0-q<LDhNKGyu7Tp9%?^@ndQh>Dz=ora3+Q1LEn4
zUUGZkW+=!p3}W;VQ$Lc_q4!kc=BKM`x9yPY)=D4$b%RML|6IrXL#6HMH;hhOv<WJ}
zKTp3a1YSadnmxJvP58ghf?9;c|2#C<SpPoBGXK-|^?!H2{r}%sY(n-~q7S^B3sE8N
z=$~vL&y>;qd#eB&v!>e%`W65P)e~)zwr+~ZwG-w^sGB-XH_NO={|Ep?OpB5He`Qby
z{7>7L|3}3C+xU-B{$B?95eb6B!b*tj!((SuodrGFL8l{NPBu%oNcbOT7fbvH2C!dw
z*#GqX^B^(&@9rP}r!;=;7YO)AHkm#3pOlJ=0oTRS!iOrk-s+ZU4Aq4oF06#&y{aTb
z?IGo-zPP78W#2(w-+nH9p!8_29QA)xG_S@7&Va`kf(N+%JX&Fu{D8bY#J#+YN*^#2
zu_Vo9k4%TKyXF_U|6{`h@Z1k(vW3_NuBuo70BoXRrC=)Xd|u>upJx2Q^et+6P_iTD
ze@3mr)=E@5>T$0#=7ur>0O_*<@1IcbV&c4VDL?FLL1oN;{?~#-E_$3dW`^#0Eg)>H
zgfU<Yy4%J&(~iBt9u2JyB187SLI$|j{USlEs{ZFg-_^1w-1hK0jNm5O5FE*hJ^$+#
znUH;+Z71~n2oO3}0zxA@ffKypC1D)Lgus}nM$JD<1^_S_#A}{iMH+BPK!GWbY~3Rn
zZT_}M&5X=k^4~$;?JhMxrcE5zRVx7is_^xj?uD)t%i4(dng5kUb+bwC1n!oc!iAK(
zyf!>s-H4kncC&2|{-9W?01*5iSE%YKdZ6HVg`z`bB_!SkpbPa%DE%x1Vg5HN6!bXN
zue?*(Q_9exgIN@hYg>ud6~3NH{9nl7eJ|Ub*pm0IO@S$kMGw)v+LQNQVh&XQZ+Jn;
zDckZsd)HVM`=9PPft4|Ws~?e)`~NBs=&Da7NkOc=kBODQVf~q6V2b-bR^vZmApv1e
zw`HF2vpkd(rW*@0Vb?}MZu|fA?b@_JiH1)g{IIY`*UAni0SoO*{Wplv%Q$!pIEiRf
zxWa%oQR+f&3T6bA{da}D8=M0@dnOnD=X;KdUdeuc6!RGWRkf?>nMm0q3+53l<eus5
z^COB*s(-rznldU`Cu<u1(`RF!vs=0F2>#au?|SPuag-XAC;<SKtUkKBh{eEvt&f@I
zy3Y~4_n*l!_Sz=B&hb@C{}rSQk3E2-#3EZ<LL#qp1nt9*PfFYW8fC9pGg`)kN)!uA
zHt?C~OBo^BztR)!qni<u=oG<>)o0hohW=r_{J%4Stxv>q%s^0F7|e_3yuhg9yrus|
ztXWFg*2(6QC&_I@xQq>d<t}+?u$}6E4K@6!@ja{acMu8|7P~6~YqeC<-ggL4!o*dh
zSozzbTBh6|BW^hS8q7&M3(ia(ze><R&t_z?ygkGSq?s=;_xcdV@0r4px;SQz?)rQR
zk?qeYlM+5u&CDqOGQO*nsNKW*H1d0F*V69-wcIswr$?hm1getB9l-sXR>p=n={RPA
zUz>hRpley)w9?XsZ`mq>g*6;K(Kbl#OHjz3EJMAk$J|iB4_oXOTzP8@oiDr{=ESr2
z84(IbdAat&-^-omtd(4eYvD;|;z7mAa~vWrJe+s2r|A>89b1eYSDiX>Qd=(9)=Hci
zGimR+%${pBPF4{*wt*eyD25ynvVNu4d=ioS5Hs71TK4gUw9vAbdv{UJY-@#Gd0LK~
z-|Boyk<0Pjc6Qlf5)xzeTg>j3a^Z#QfP^HM%;|)Nx*|#b)SwC}wG5W|2{#c7J}SZ!
z{|0h3;^y~tKhc2I{4uTugZdI!%$impVjRL=!hwtQTt{Ik<X--a*-(3{{=M>SH5_iu
zhvTOZ_UK);P);cd`u?1<z7)4S=H;YEaS#2&ntMZk-J^j^N(#>J34Ybr#VF~^Y29{K
z|45{2Y`SLR(J|CXxnB_ms6lAt(O-zs*I{+)&6#eS$ow!T?G87erF1kDYlqpK|4U)x
zP3R@1>Fm8q_jSHcQfTJgFvJwKS0kTHa4O=sn5-gkov^$0!>;ga)V1fftK?YZAGhqf
ziA?#&({x}i%h7<tAB-$V5pl0XUc%jLcSgQOqM)EoEf^M=GP@n3M;jBe<fu<R6H5`z
z*uK<$I;QQqj{b8-<!7-`kZnywbMSnBg0KbIgG0|#+ffWAU{AC3>VCznPyOS<y^zh~
z#@=*#!+6S;^fOvUtAzTPVwI_R;HW?6C~M<OMwC~smREX$_0#8r8}s~+XmcA&GFYsD
z7e7|Vd*#vBA6G9q&DRG5-F6iBTuIg+%sx8)f-bTyuk3v=X8zBb;NV<LX20i707jU7
zBFq2qS9_gRt`#|)hs=^ul*4HTw?hS_aZf;34g5~W#Jr34aBe5LBr-E+#ENc0VMBzN
zKKRvOQ}Q+D8c&BdrW5OC#lkJ0v0@xS!TjSJqQg{O!-(DHz$43S%6AhtjY+Rxte=Ek
z=N57k7x}yMP5k%RAMUXtuC==KIM*N_{)v(U9!GrzzQj{MJ59#8o##1{iw?5>Dx1={
z?_*gsbvRr;k;scS#z03sgvKfN-eGHJe`^Y&cc^Qfu!n*lodD)r#4V^Rp9Wk)j04{i
zJHK~i8(-a7(l*4Rm#B<@j&tQ0rM^X_{+c>_Hx%P+*&HB^qbX5|<yia~wx{&##!+&O
z6u~FU^u;gt2=@@^4GBoW+qiO=t{63XZ2Eom;9fa*;Oga1=dP$NFTiW}%%wl+e89{?
zuPhbpY9HghvOtA0`D@435^YMga_Z?rKdJJq;Gd{;Y#`K15@))i&aNm{R$nI%F*K5k
z=bh%$004_~;gt8dGqBK2<p8|J&gaV-fygez5D*>Tu(okTl$}=Yk(~TW27y(%wn(L#
zSXc3sim(Y?Y+L2;LM{J-{|aB%SohoeLQh<Qx`+u!kK4)nY?oR&OIhR8^OM}Q9e~Ac
z$c~2Fs)R%V*SIJ5?Q$<HkiyTe2=_i7nP;2wXI?VpHu>ueA`Q-Yw5N~tcTZY|=hLTN
z)RP>cJD2H5mw4)-dWfnugFzLENqGW;>@)4VHw<m6h14{l-$Ka;6K_2lM{PNrB)3c1
zKF*o@!{Tz7=7%4;4j38{64-kX=Wm3A@DTpU>5TmqTb2ng(Fm9N6?28X@3=>Ts+9{z
z1KSF&TCH6E>a1)IjJI65&UTfol*r2wse}jUAn#qUi`+4zvbmC3UhCLg6*`!xu{sL$
z78VMF^_yrK5f5VusfYMI5ODq)i}W{}wQ?@49@+p0hq7HyJZYW121cm-%O?AMw;x(7
z%7bu4=4XU^>D=TuXEJ8&x=o`&h}Jw2Ng9WSUN%nr+xib9y!1QX&XJ(^|Abn#zsaM$
zuz1$jl+rdXjH|P+9Ws(NtUbRdAmyOPxVF8f&YJcqs+6>!&LMsoA|SKrg0>PAt+)wu
z$YK)Rec!PKnRv}ds^3B+7tuZ=I`$tku_{UCbuD2rL7_RtVb?!ijv9=#ztZeGf)3#y
zDaQKU;>Ti7*H!)<IIcffWM`R8l_u<IXbvXSWA>mibaET8h$}DW4q6>`;Di!_Er@(}
z`QwpK*r1pPySh&P1G)9E-QGGDaiOjaLltKF4Tw1=uZ3A-kSJ(Bk%>$Ss(Xpmt}2)y
zq~&wgLGqWUI#xwSkC)a7Ii&jX3xQN~W=$KLlF7&hqq06Z=vhxQRaR=(h7<0&GAdlg
zt3u^vK8^vB{_CuH)hn1uiNaxo&4awr$C1oTBWs(qa{Lcs;Kf2*`wE%ffu#=Ou*iI{
z$dPe2pm-!}+_<vHzegavK+(d`tEDzZa3QWXHi2cr+5i3&q3{bCPvLt>_2=zL(b{0g
zyCs2?aG3~(>7<rFKi-$UsS5C7GNCoB?Gw<g)O8X?REbq~V@QxOJ7IDdIkNj@vohj-
z!q?d)AY-kEg;CWBiYzXjNjy10I8pR(W@y)=nb;w=(cKR>(kJ(jb}G%aAd{G(n*A-r
zOmq@^HK{u&lkDoMBky30BT!8rqa9aYJa|a6(yBcow0HhyoiqFQeK4=K6`7s{Lpj=F
ztfIA?_~F;Er9Uj2yR(<5uIc@Dh3DMcL=*dv>CsA(eHYWkPCXjzh#`$_|17I)fO6PO
z-|?xX_TZ<8<>-BDn`{6)s3|jlMb_^-DBjLx+%sHkZQk5apt<(>8-$%24P_&_i1y7&
zzg|(^8Am0W7SQsOKG}^aVuKOYVafrNQ6+P!{B=H;;F1}Kup2W}LQeZmRE}WA2hWgQ
zbV2xgkf*>B-}@TFA1OJE^Q*N^T9$M8)oJ}qCRna%{iGJ>!ij{#h)jx1;`e6TmPKt4
zOTz|ZS00w_OME^Xwan1b=W#t+_7tqP9%|@j`O@5n&Fvb$YB_4`7+@m&kx4cn;O2ve
z&m%j8J#zbICjzBTz(-oD*|rJ7+JFw`J~4aATr~YLYJ+wHdUVcvT@wgjU`W%-v9&r;
zx{4|{U2x_0;%i5h6`YAVw!WSh%UDr27Uwa;VA-=2G5PeI?jX#TU{(5jjoSW#;{Phh
znbxB%ViNE722eV(=~gJL_~yu8KeJ*nBR{ajx6g>&q~rQbPBSLV8VCu{$qr>G*@u|O
zd}&jBXVnL+YzKw=yU<;#zGH}oQ{P5e1Ci(q4q<Bz3d^s5XN_m7Wi_qwHfju3`TAQ(
znzIXUYo_U<B^@7v6*(K*;oJXp!U!saN<CUOO0GAEE3)xgY$+^}AffRKOK>XaF>M6*
z{Y+wg1r=VJu3pzyze0+zT9u7AJ;cs%-|iAHnuOZd8}V1vA*`_2m*-^59$ZnO_EtTN
zS8rs^y*27DkkUVBnR5ozh@ODwASl9omTjY1hyS}}P3e+~vG}{s?qX(Yt^1R&)rn!2
zS+=oxGJl-%eL$i$c1ajR-9AI%g<OdD1OB}l4Udti!yZt|1@X9i8#kEBD}m^$aGL^H
zHK3Axqc?zSP$VsYu_#nh&bY|1g~${!Q8A3k)2M`{&1~S@3Jkl;gpPAVh{np5MCaG>
z2CC~-n$^6b&??lZPY~9ept@o4J9s;c%F+g9Lz389sXpbL<;9d9mt?$C0%v3kTYrp{
zbIeKGXYN(RsaqKAhD&_AtQ<b$gu7~lLGMGu76fga&AjFolSO<7v#otmy*(Bi3{)Z{
zzQZt3zp++`puWxDSPKd_*K=g4nlL*0v<9O0-BZM9%xH}hM3{J;0B7Iv#mH|{7k>Y?
zYdv_0=h-~2Xq{2L8vIB&8G)QjvkHc$^_Vjx%7pA%^?NB8A%*Em4`*p!S#s>p4*fW-
z=L<O(>P3eT7)#ow+E>TGFq~F4wvBrjmPq_Tp%;<W!#RXC9?otYsSyU1Iv1Fiw#(HB
zb#oh6u^^Px{O_JeYH=o8|2Sw7CoX5_S8zU+2sE;1Y-6{gw&uOpn=td%HGGRDuCb6l
zm^Ngt-WUtU8u!eK;~R0C?I9-Z`f5;6aow-IHl1Xw5dfQ8*YT!47E{IEz&&$8%89P<
z`L4I`&pg!Ze5KG=1`&!oP<md~I@6zv%6jm1;kQuhg;W2djyv~3W|vJ02Y(bRe{ZqS
zMN#O=7~PTmH+l&L`<`8eJ%RC3-H`g`C}sZOx$+>m<i$)6uV@R%mzSLpmvGn%7Iu<g
zxN8Ncgl1~0{<qeXRr0Iqj<+S}DgM0zV#LkIUZP7Z-xpnwpY{6nY}C8e1GI92njivV
zTP#H-ptk`jZNtorWJ)&XjPGYUzmtBYRRx3o{_3~<b=dEz#KCe{X~m7ctLC5}zxk6s
zwc=1-6|9aDv>c6C&Pi=V^<E&qrrdTS(OdCuR1yL|Gju#|#Bg>K%gCRx&u}J!jwkWX
zYxiJRsVX*}y3ttxlAJZlCZH3ht%MhALb_UR8ct~2wW}ZjwSKim40Y9Cbf;EaILsR5
zyQz+?R#4Wgru|HVjXa>Se}LlRYU-1jPFjgSnZ`n|kgD~}21Pp9j@{sQgG^quoJmQF
zU=^Hc{Lsn~sRS_RwhsH%9_R}7#?Y7ZISy@`FFSrM_+j5C-X?|$C8SqP7ipW|fw0R>
ze-9LrB7+is1!*Yf!L}N{dG#OrC8pW^nrYt_#=JM^U4DWBuAHT(3~QrbjDZHOk}%9<
zL|%@lg%s&UPIyENgBm*Ln(%U+0zl!|>YH|JQkI$WlKTsZaZM8fQgUu)T*9i=k22N_
zuAiu8&sJJ^v*jNxVXt#ZWc<qlVtFTSK$_*Bg5mk)h}nQQ&qWZdnE^6cf8ip6GP2Vq
zt_gwUCyjEgoNKQYjQ~J#S<g6x=w`ECW4xEAf$9`|wv3s8tD9{QJ#~ukI>=a8hpr+;
zau)oYD6}G$41QH{ETTzTo~gO?71P|7P2EtEanOIz#o(6|A^=Luka+N691^ulWL`$5
zEaMhL-i*|`hIpO`PCsP(VdV5V8QyR#h0h<+Er%C;CJEInwz(g%ITzA0w@C)!d0$js
z^DJb#m<+!b^Zvn>O8^D*ac>X2^=YL__F>iW8uPaxqtsvN{9Cktsan5nZoi@%5Ly0)
z`UHC(hC97V+SCPFz7Qbig>k+}TumnkDFL^5Tv5onGcX}S^TXs$$?fuG$F;=*LF$wG
z@5VoXX6R-M^=!hjFxu#is_Aek2DDUr<wnTh%Q{h#G-8gs#)YF)pE@47`{LbGP8p}p
zWX&V_iWoQ4_N#^Ml%XUx799H4HcBn`-1^NH2x{%YzZ|g758?RyMv3Jh7e=xU54OuH
zta;2et$P0`nVD3siH)aqP%{`BVpg`*ZyMAnHz!gJ&v1<a1$(X#XIU=3!mlL#L_qE$
zzRJ&s39oL|pLt|X22dHiz|~pn_dz>p5dQX4OO|e{L;?VgyT<V<^F|o2K+FJiy8Z$M
zb2&rVKB4UwnfI}01X91-P?%p55aTKWvM|_B3KNy{{vOW9J6mg(afa;5E_)$I2!wT*
zeS7DOSGnokdNRLRxAh8RL=&E}uzmEK%8c7t1kE2F9LuQ+!hkWBi(LYeg?fAFwXq+;
zMI%?BclR&jb33H1Yqpm%QoPT)xt4Vh)wcn<SG#&p%m}_X5)PjI3eVU}=E+9>-*+vZ
zUpo3Csf$&w8cXSFE3ZGQnNiN`NXsPqEKCT9?X8?PRR1&X&9CFWT;o9?IrSwHiB6oT
zC9_wbz;_ur^?Y6eov`o|p~`%xiGb5#MNR$rA|Ge}3hW4(jwL6zwgG*-7kl9@_2zMZ
zOEOB~6(UmeY|5Zg&aMh`GBIjpt!D0Ws^Gm+?X{=l<nuRf7&Dpg5=7Pa<dzj<@&rR{
zQHwt#OwBsrJ$FWxg~1pnaAz4Hhjl6@bRZwt6{m0y?Oc4aNHaIcLu4+>f_C{=aGI%$
z=QxF2uZc0_)bp(ih&drHFq7EHvO=f#kHKR(wv^bq=9Sw_P|Ru4K1JLm_;0aMe7%KG
zi~j6%J*#=GJ~8o$=!W^J^oTeM^nAi`0tI7~b<R;%y?azW=EE9vw8a24s{sox+47`k
z!&lR-d8%J}-*WzJU1lARCF|M2bYk6a5=6DrTwQe6U;lzII4jve`^~K(W{85p^^GY0
z4niGSQ|2Lr5>Pybrm#;GRXMreoct;Wb1x=~+qdVev+)+izwrx%dzUDVY=4xs=LjC|
z$1EABUgUIQ6s^TcmZmc10JK^C%n|xANv^Sv)<)TYSr#|(ClUyy#KqUd(Z9xI0Mhvs
z`-%sV6eUP6h6&xomL?PvK(7)bc%SkU!j1q;eZhq}tw7*gdSCI5;K}FZPSH80{16@h
zW%)3@%91c`!c`6de|-JZmyXj?RO)^q-vo-?)7zqZdwLT~mQ6JLl+I+1rtV3l9dEPl
zgMLNSBo`;40y0_Z?UvT-a;mpx4e<6fEOq2QZ2R~-tL6zSfA%Odw8M~&O1_g!UmRYh
zjk_+KGroqvDEwgO%GNnnq_xA%X;;=w5r?b99<DtCj-;P@$2L_s`U{DB@QHqtuX%<s
z(N4^&O?0EA9h-s{XJn}9L6$7G8`_CB<%(}0CgL-i@!D~%d|d*;GUTO~r(*o8GZnFm
z&MO2})z@c1)<fJ&4H`SFXkO9iO!cu?W5-8EWRx|EUTJR<Pj8SWOxG^_VI;Q<;WZ!q
z)sto;4T+S}GqL3^hohOs6N166h7<Ebu0tK#jFqD(>#Gz>U|Em23ISL0pxK%!-*(zY
zr@r|~OP!h8Qk8J#W_i$?sX4M;iW+oEa2R=|N8r(v8=D2)6=|+N4!sML7TY{bNSnHS
z2c;@>tfB6N!tx?{-U-<kqz!Jqw8zv}RM=)wVR8O${gRCH9xE*fv9#(vsCd+y%gppp
zTUt+6*`pN$SiHt+v-a*Ovtw=Ws7Lt+Z~N+d(g78_6fHHiX2!$0AuuWeUiJlhSgKb9
z@AFlSXy5BtYul*sM&#eFwIg{dsIgP4Pri_8{CD1D$FzytF(Z*qPKp2+cZmgB40z8)
zwRUgbT-=X*)qsxGC#y+DF)1~@bThUC^fTg#N|Qz_RZaEgGHMTpu0WlP;h9OXhu7KL
z6#b!?@a3b-@7$*xWm&crMHE|RaWnu4T#sO5-!AQe$MmuXRjpiRtKNnqiEq#aI4%oT
zgAYq4`{CC9W`?g{R!}eKUeV@V+il0E-HIGyWAsRjiPUJD_w-DNjYO0$J~`g6yE^tv
z?wK3xam{B0@h`b8FLK}K;xDUv<~X+vhDDp_8TKE0>@Ngp*1?FqucHEda79fd&)87R
zdlzPF$GKZ)72W4_55f^tU!9{)$mU1)pHxt(gGU(_mLDPS0QaS1F&~UGK+$d2zv6jv
z85HdwXkWoq%*^*kr=l?G`hYF!OIx4OtN0?tmcJR2co)BJEr$N$h&N^yd5y%D%?yj~
zOzuLrvcg6r(_oxgL2J19--O<hcZM#8HP*8u{3kqmW|vqV=LN4c>aKp&RI4;c6X$N+
zt=uznWE+S{=eED|4dX>@&z%aUN$&^S!qNsLPAU1<&#_t0Efy!KE_3J=HYJ_rtR_48
z0tyFA7XrQ!S!GfBhiQXWn0io;-=6I9$dv&mw;WCiu?JlQDuQm;<4=Sz+cXl6bTc0p
zdIcnMV)qRruItkdd?!iwE*YbWdk+5;@Yv1<*dWtb4dUsgEO@>;@x9*OQo!~|+s-pw
zj7#?ukM6d8HIZG0`R}HD6K?WD6O}r}G0K7H>aYNW^(xo$xhC3hC`jn<l8XppiY;$s
z*=2g|*RQeWbon`kyWX!BcNWjs@7agAH`hrm>w-7kd7JFaDPE1!(Dha7s@n?Ga!0Ia
z<IdTrTnugg5}c!-I^`^<nDWiINe!30`wHM+DBV^==)`(Se|51Z>U5{Sb*3n~@b5lF
zqkyIR((W;Sbz)V@GTT(Y2kE4Sa>@OD0#9U#Q9$Jr!w&Tlx@a9x?~*WxMr>(F6?D>9
z|4R5Y7+?R@@t)}n*72Ms9iVx-&VijctGS{Di7;*8LvlZ}dhmz2`c1j2kMvFpZvetQ
zUB@b~dcjSkH0M-&3*6g!v5clzR=!`D0BBtm!ysC+PAOIjkWTSy#_Z~@Uf03<qV|f(
z{wujQt?EmZZ-ls|M~9*tZ#>}XoE!=?aLO7oZU1r#hppIuK`rYLd}nX}A&j#jE3P>)
zP)Kg@2fup(dadSNW<1V3I@-;qG3MqCF*%u2FgOmwtA*51KmFT6I(l8KyVjX9!KF+n
zFGmOeMTXg=w9sgFS(BK2N_PKe4W+G*?%@K@jXbaBuf?V>VTE<O`MWH#_phc_N*KUv
z|G_>OEB2bly&~W_<F`jJx~(&2S{HZKUfjd3z(YJg-(<Ay$)P%Mu3r6e^UB_X)|A_C
ze%BL}Ul@K(GR-FjS^m(Wm=EJ@Cnfw-XhjAj>tt#KBlEcRldgU-UZ%~Sk8tLSaK&-;
z5MS-k<vZ&eeKWpnbD!4+=?}i{_1^c*Pw)?K!@GK(K9^wqUV1X3$56Ox&5l`l6ZDHT
zt)y&r%5ZlOarA53BgbRy>;P-Hr#}9pmTys(sxZ+Nk(;^vlV600_(t=~skdw!bG@Y8
zmRB_XOV!sz$m{fz$$Fe(bM*R$0SMnl;BEz#dD|MR(vEI9_;r9nT3)jrqfaKMMkxcm
z@?8<_%C7H3VZA?dGx{^{i`5dHpI{Zkt8A0B?3$RvuM$xj2td5EYh8@ffp?Ma(ke(q
zTwthlyBDFdJa_nu*Gj$!*BR<6%NHO{ggtR;6@3_+kO_tKOt*I$e+&E8+vT{$M3xWZ
z2MLI^-<0B*GUfDeDvq(48ozbINWr?)g&qoWr@>g0`*$yLx~A1xJK=2OO*ULnU`LiX
z8c%M`VgHlXYWWh*MAjG|Hs?^i+_RLzpv9GS?7NRpjx&nsP*izsB$ClJhy8*<dSu<6
zRjzB{=UeC2^V+9$JLL;O;d-|UoKsH1Mr0D;b-Q9$gTs*P>eJ8B@w18C=ixKJ?k)8(
ze~u%d-eBCS!lmN|x$Wbb<7HD)LtBnCD%C>z>^;3|$2cFO>L`uX&8;RH>_rLOVM5Uv
z$9gKZN_vlc)Q1<bKD|5{a$jLXYHe6Z(AIyHp?c4|7+EXF7qKn=`W@8a<|Ihll+OP&
zK-5v2YsabCxo~U2C`Ge$5X17KWM#5Vwse>H>@n=NCu+oaqUal~k8{eNC4Ah1-srrj
z?l4jt@{5$hsVEptTTmMakZsA`6Cykqe~djc6>cNd!#;S^P2jdy3U|En1yLlH>t+^E
z56tUC7lARO;F>hX#j(yLvsAx`&a73frCO#2&>Q>OB|g=TTkKZs-3@}r<%O*7{-@b~
z$f9b#$M<iFCcszG<;@3VyHPS^_l3|m_R>r$PeTcAR}m@8PYC{iRh^{2e7DMVja4lR
zpmvq2lr^NY1cwhEh@;y~jLHBnogsG8$s;7@)?H=Lcz6(iQ~{k0W=)FXib-p&N=5Hc
z2UC{t8Hb3Qhf9ChId+&(TF^gT+5J!m&JXGM4lvDfi#jtBnlKY1dm@$;mpevI+Bu^K
zZiJocnrFI?(>Jh}2o{#&>d4veLSEd^X{REz%$+2E5E+5VC##$N`%Fr5PA4beF`(M@
z%zE29(nR0Ou<qDWU5e$NjqKaW3X-GlD7Et--MZMa2BJ46Ho@-8%l6-r@?G=m9zL|b
z7hbodN0by$Z<em<kIXKvKJx#I_LF)Oe`!FNqmIiDds#XYUNh@kP@O8PY6~s}+Ch6~
zZiUVZ-7WyCP~z-mBD^Z61xuVcz3$|i&{<IbSu7;A6E;A$5*fUIHa!(l<q(a|zT@de
z@5S9uoIjZ4Ef`)F*vQ;qdz^gj7<nEYt%Mm=f~jqirEj6>j1(G}Ji4Z`7cbNsh?T;T
zJ#vgy!=vP0t%&xrkgfzHVC~5fA8MKR;-1DQ&E|_U&3+gEvL*bHtW5(>qChznwG?!C
z-p2Ut9fjthHgt?{bq^_J+ErMeWvwPZy?8=up#1q9)ujEHfEzt&eM(UkvoSbJ-whac
zXviN=-<)RLbO!`ZBGk{aYUhCvVwL)OsBfaHK(>!onYja)+aG(or%w>GhkLPf74L=`
zL=6&Q8H6U<sR>b2GK42m0Xn`!>ONoxOqS9J{SeaobzkLJ6H2F_pp-v<lVw=t`IiQp
z-o?|2#>65g^YOm=;FX#!x!P?H>V(0GMkrBp$C=Y<^?SGRvjZ*OfXW9AqA~6TzRKI`
zU`SFs8qUPzzG*z*oAuT<hfs^5c`@lz<(N$iPv!OQ*2^}4Uj}W#FluPF)$MG%LY!%P
zzlt{z=a8xI)TMk|KMK3$z1KNLEsn<m^MoA1e2PBKUkiHaN9poGU7=UbY#at~F_RXO
z!!}$l_};}2`;oRTSLv83=kid~(zbO?yz*?3T3i0-UVd7IAvUtr+oD=Nr%WGAlNt}~
ziC^9u`2KrU*y!;eOz9EXX`C`+XQ^mk!YYdY2^<P<7FOweG89UlZrx<9(q+SuCvBxh
zAk+zr@w~4I%2!0}b6|e_mxVff8Mr4>cT))jGW*jd?Kq7Kw?NTRx^D|U^iR^H>0h;E
zlT^iR369-A6a7J1&}?eW!g*7lj~#&<I{4l3jc#IX1)_%3Bcj8ktHD+BsHhWm=R1d5
zSFq~~1MBjOBn&sZD(DTXfjhZR!NK@-+(<PfvZSzTk@yQ|sFMuB>ea<_g|x1~E2@v=
zVza&ZVtOEBKmXji5t*^MAKb8++{vFZr-u3?p&j5xdr<uqP3QZ2R&&QL2^UD>RAOXp
zb>WQ>1Y>1OZ6){f$^2iErheV5q|Yavu*HrgJTiT-bmdI+O3F(+lW?%yp$ck1|G4{y
zD)8CV;dUN~;X3!Ea3f=67$&~D1ffdK%HG?~5=i?U<tZ9yU_$DP3FIJAN&UDpt>0R!
z|0Ct3Kr{Z8r>?00fW@?X)kV{Ze`F<G%z@Nia`?fUI_Zg2?6jszS!_5^)F2gxPQn}o
z%fPn*@`TBsS??$3!t2$Y5F1qxwo^ar(I|zl4MTO(T@O3+7(1m;3W!%L2I=`*7S^rF
zetP;LOH1p&v@3j@|KnY-Yte)wt6k<iU-Rw=fW5*-j<T}F4iVS=HR7KmLNZoQUH{mv
zI)TwfWA{jX8)y9n=J*e%F6uU@)OYKAp?3!tot*fmK^tDv(yW_(>l$_pVZUx~LFO;d
zZ%6pgqg1hb^<Q+MNHmv;O?-h496WJ-bB(Lq&IVHuUG}7ulwK@)qkRreE1y;%hE;Ox
zPlu@+5!zuYR710`Q#Yk9E-srSRShuguMr%b!U%~^F%-kh(}LF79;oLhd^sDsk`$f6
z*pJ%}0Q4(SU5px`C$hdZZgZT}5Td4<JYMkXg$SYt8;3VVMFpac5^%}=Re$5AVM~~$
ziVQ3hm_g!65F_PQLq5ZBL{!H;nT)*O$#mAfCNUYd4l5=>wrBBq%q<r$pV^)E;L@J8
zkGDiUch@Z&3u3vVH^B~JjO=*2z|I{rZlTN`&`R);Ied$aylabo^>Z@h!x*cQ5#vUA
zawl)?v7gNue$?Ob{VaPgg*cf&ly(I(Q7C+c*upGc=%sJZz4V#O)J=ojg;)JAn5s+X
zFQ`zYvb}w0ohzk_Y)6vX@LD_!f%c%$?+)rwOD~U4^WNdR`}@uc8<Lfbf<yi50IeRK
zu}!}QP`JkB>6zFLOSC-tu0<P*If95-8Hp<nr)T)89U`5o(cXmuY@$Z;uO&-w*PQPC
z-<vj3{sq?0;qb4WW5671<Z)M2L6tQ9&6lV-%y<1kF)HK4h}^|l)K?xUOxpe%{Az72
z>hQ&r;0kVikU*eilbl}a14cPjlFHH6Q>X(7N|_?>x_q*aZ98N8E!K=p%`0oWKB7$K
z<NPJC3!?k$!J`&91))8!>Lv`uy7}c5x0B}xJL?B?#rgxKKQVTukpMiZy>jAmvG~`G
z+tP1MK1Php#kHy5j4+MeGj#t_h+TRKguF@OAdz|>TW&DmPwcsOE^2QhJm6w&NA~f`
zftt}~h;n6ALCTe4P;Jt?3P8jgf$^q6Q+ST`%6L-Tf7%Myc>mchS?_EBrY@g6KO~Il
zbN~`Zk=L-!H#w*r=l4#4ZG;SZMH+ALDD^Eg2iJ>MO5G1$IN!bMcS@dlpNZ0ibE<v_
zUMt^vpLc-ye(6Gv<(o4_gm3Ekc!SQY>zoNi&1`4V6bX^kM}EZCA_z=7xUDoRrd*M&
zX(UzkF^S7D1(X_B=Q3i=!~SMuKZTwE4^qb;*Lh15ariQt!XJrF1SC(TU1qs$pq2%X
zJ0%+Dljk;ld(f)HYouQa4p-YSYlrzqHF<HW1AbC!lv713R{b8a6XUJiePeM>iQyzL
zULM8@X8#sMew%jccIuf?<<|uf+2qW39^20jsL&{qSM4sV`X|*ek_~|$InF5titxC2
zwclT&-U}s^O@;EVwpK>DMqJUF@0Cig9-nek+zff6T1qifo!ANs(c_e5nl`sG+CX&}
z5(+*-TWQ<zr(;c<vRTt6fKPGxI?E4Y;buL)V&pxRZA7Z{EtnUI_qL%_M1*zqZyOiv
z%VF-uHZ{F#3M@L8>f2+S{WOU`K<!T7ux~!lLSCt9B0N%h{5dAP;<kp|;N@vw<(2|!
zCoDprIM#k~#>XB<=U^w?<|6YYVh7KU``BODeT$3SOu+CQd%mg&(ixS+u;V~|y*T5|
zM`MLkS05Nk<pqw%NokZ}X_BK>{n@{|;nglgA<EE<3SU=%F}reHsgc(%H*Atqa*_4=
z$J_IZ9+@-wutjCy^;rSteY#3&_%1L*Srw&FiD}amiUg-#i0tEx^_wkK*(;n=dtLx6
z!w`}U*P4CpBsaxn1e_%_<^wl>N#*2GV!0E5K}_<-_fM<WeTod&Ux{(cO{x_`A47}o
zyu;j_J#|sfSh$JpiJClo;pqgjoru8AX0i-B5mb^M0ukXnM+Lg%Y^F1)nqx|$^C<?u
zt_sA<l(ykrm4_&nhpBy!_plOSCzl_JVYqxY(m=n4@i8c4lNMHedzbCvHx5&8HbyJS
zzNKyEIP2?=g^;T<G+OlC2M3{=p(A5ntLKnaAkw`%3<)0%I<)-DRH&Zy{qgG`L7L_=
zG5MBDxLU#qG%N3-p<FSCKl2OTY&4$=u!j-hmrHv?W&113up`v%*c;3B=sLq54T)0z
z)@vJZqZ`3-u5MRmqR`N~e`wK~k94Rd*W>f^?fW()pc|)zlK3jJ#UYpa(~G*~Q-bfr
zBLquMBio)YP9=I9-MtE(3XEX*-i|Sw5eKNsX=DqDLSFx{q-%o6JrAq+W@tzwY1q>N
zL}XPm**GI9mwhz>tTn7AdM<i9eOV>0*p1P9>FfpcN8GREWMj+9fkiCHetaNh=1=yx
z%K}<ilS%dc2@bFi5z^(&s(8~-pEe&O%^8vo5425CKdOgqrIGerwR!dJQftg@Njtwj
zP3MMyi&j)v@N5SegFQnoIz;x2=Z8u9cC75_+*`#M($YwbaAA%bJJdzgE>xasopH7w
z(}~`*SAq=lsv3u^6#pLbX6v^htg5hPsv}1?M6_e9*6-ksl3?KJ<ob&RWz-cub)!m8
z`@8!|kr}-}biMU<MUtgkt&xFt--@K<;8%K+xh}dNeVml7F9~#7HZbzG1t*U$H|fuj
zg!gYyfJ@0(?3Fo()~|hpNa{~JGff!<=WgHQ%>x`O5Z@@EE@+x0VrczixE4=GjVOe=
za~YHQpLPt>fwqLZg-wF12D-jDimo=&HeTVm4#G_vTV!i6W?y2nb3fB6#<EBP8>&;e
z;(#Oo5+#bc6YbcP8>dY`u@rHA(~vU2X)l&y?^j3u;jM5`@u!k0+rwrJhL(hBUG#IQ
zYxtSwPbP1aHfj7^CSi3{#D<17ENkcQtQlM0Z4BZ-gUQEJ!P;B)3mGsswK12BTGO|y
zKBxj)>YIXD`Thum%Jf(u+Lc9I8%1KdDEbh~QF9_RB(IcHB+C@!kq}hu!jY&xS_@@v
z)PMR-d^BpW;HI`BeDX1MU@=ms-mAbp_2T>WR30H=$?EmB8gqg#e#LHEswXY2WPjGa
z=Zt_@Y{muBWPwX{fDDu9KR#(^9x)ydu*X{x%j`4Y;ez?6P4<brWS4F9tAoOc>W<wT
zr43m>H;N|4)!52NP!<BFE&j06Ya(ota1SZvwtFGf>X=6Kc!UyfL~o{Cp!ARm4f=+f
z0B{gfcI@aM#2N4=&B0U%)uCR8P1t;@Rci1<0=%Cpp@%~bSPyHn$^IG5Fr6dsnVGwS
zbK<o6ZBqVH=D5Q|xJpI@*}HJ%!sHuWK^|&z|7&EOw7L?$i|-T}I=G&4(m4f|t_lv;
z!`E7bK@I7C(>LMTD8R7%&5i^t=KyniZ&RPZXa2c8Bl1l|-oEVJnWiDm>nziJ8AXg5
zav-Unb)A?AC$##%*t&Rc7IML2S-l`2F;VKu`C1L*xWf(KG$aosM^IYI75o%p7c~M7
zufBm@u%O-nH1EFiUc06^|7tQ1$YMAeWA@#ZYAsvkVc}lBFc<5Up>V;MT88{ttCBf}
z(W}nBmSmq_N+`6pD#Ulj&K)1L!!TvLu%U27V3=KMVYBFMiot}AkJ?Ega8QoD3=-+2
zF2f_>%=)e8Sfo!jQfgh`E_z;lvs!)orNG_v8*9P$%cy*2^h6=O{y8ER@MnyZs@RX%
zt33hszxIyKcZ0L&*(bf(Ft7Nzn?qD!Wpf3_tof?<L|OLPl*%#90n;a(9uZOX!3FW<
z6eX*-6^QuLX+Jm}U-6+nl;8)goX7)Jmhm$e626eA3aW4kO<^OT^-7@Cm#&7Hs`yVk
z5*Irbn6P{x_Qh3|jIU@;!vtnmq24g{ZNg(Ca8f|7)E={BRk<7yKhAJ$=o8xS<GR|Q
zKK5mS0Nhg>CHemB`4v8v?a#2(jmkl<?V?64Ivnmh7&dhL9_Vpodkv(@+_4RHrI8Al
zUyVzBx<Md~iu?lgas*Xl3Qhrb)mDamS<*$%-IDl|Litcbr!p7mI<ij5nID>64Pp*7
z#BxKj)wbFgp;ha@`EWv`s?}kZRajxUs9g|EzJzyGOaDgL#WfGKT*EazmO1uw5UwgV
zNtQAB#tVZPf2u&HSk4O3LaW^{jy1Y=X?^oNEFCC8DcV-(tfN0<nZq>(kQGlaT=9by
zhEP*Esy}PDsA@0USB}5)Juq<7ciS3AIh3>pWhiftgPmaGZu#;`a9oVwsFdzckL`br
z-GC+S`49VaAXpN4I-D@O`T|~<jeyq`h~f)^iV@=g*p-@uqcd^75dm?h*mke<>61^)
zt@rX-7`urTVR(!42}`gNoh_t$7QReXf48yix{Xq+u(f`W@f0(&<V=A>SA$;_xdOYG
z5|MIm&LtWQTnl}#UN8)ph5LV3gOyfP^FQCuWN-A4C1sgY8a}`B^m=P2^goc|?6*<^
zh_nS-lH{$&>J{qIRUM#C6`;2Xuyef~lo8sEOSKbm2HT%QtAH&u?mYw8uqt|v{#*Jf
z05aDW`9M>puABJyr;1SDamD=fgcI4pdx!8LF({mQj2AnJ6AA^j=ZYDUv@LlAZo%On
z^aWjFUh7P}stPWqI%J)~^lbIN>;L4h(-2^IMt#rLXKzE+s|{0oT?d1arIT6}O9tjh
zKI~R9LNUXJd^C^q7*IS%nY1z#34rB>v;g`3V>kS(SVua2T?mgTE#3p&KjIDEZk{15
zH-Wse33n@<@$Q4Hv&K8s+PO_Iol}ZK+w^Gw{-Y~n4Jubg;x~o<Quu46c4VKazuUt?
z)&K&s0>*+S)e-Lls+XI&iO=9uF)Wq#{(f*Fn;$}-S27~$dnnd{A_ra$0Mb7J;$<lB
zhXp=67!(SURPjFF19k-`8Z2C4;&s;}EocszNbpg>VmnH9jPxwq-v}NYaO6MPWS;xR
z1(sy_PvVQg#wXZXup-S1y>iQ4NeERG%R~`gyIf>n!Zh|U2kUIM;zHJ_Csp&@`R)#1
z4-A%0B=78?K$v4}?WTNm?W3h@ZKoh23Jf=s-j{S>6+a<+B3KeNk3nx3voRJF-Pf<X
zkdRP;b%|+dkc__+-j}eDci}0rE$5z(zIsP-`m0L4rWO>Sj(l6qxhK^$_q{%8;k+1m
znHq*tp~?3>4&;sX=h_Wu3lP#mYCK@Aj6}`oYfoU-xFduguoR)J=L1{P0Bk*zsF7`w
zE~2B)WbS*!2Ijgp_#5h1?UPz{eA@E$fI>ZHZTS#>kei^`8vkPt?4XL&yqo}vpx!CH
ze{1>d`x|DHP_I)*<FX&4__G$@)|>~m!(`OUg)LBCp9b?jRp6iGhyogDfrv<spI!%Z
zxJ;;&iYTbkY_p1ZT6GM;xH_2SznsjY{ZSg5EsPyLH#kl~LV*}XZ()a!GAtlbY=Pz!
zSpv3@id_xtmIcxNU@bGjJKtmF%u=BZmTL3_QVQseY13jWp1iIDViayh-Gnp7hV99Q
zc7w5^`<_eAc2oc-;fN(7R6YT)xkft|mBz8MG|{utEwt0OMWN;s#%@<x8A`>^W%b3`
z&7~Vw9z875xnL-b3Ja51)5#~5kxd}muOa@?HSTCw!(eL*P|#bnU6dP3oI_{o#!H8-
z$7HVB4mf&j2gg{uBpDQE6f@?fcXrI}mgFB$ygddk6f&!7{nq3eZcGYcdy6XO6YtAM
zfr}1Mtd&xpe+WIeFFmup@VY7wC@fbNa>t@Z{qD)6P{&ecLXUh?0hWbnb)^!=UM_PN
zhXxi6l^Ze7#iiAJ(w3)ed{T(Q3F?@Nt0h;8==kukYK!9*wR}-b>^Ay0qz$ZoX8lwN
z^G)}e<0HD8?1@BqFqwgIB(*a>Cc)j}=Smb?N$V!IRM)72ZZjwxzQMk)yE(If6L(Pl
zB!kYRcN0B_ptk&Zr43oKj9k6p6lCVf-OHq_p>(R(3b*K8_Sz#}Cmo(I2aHxy{rdE=
zlUeW#SH3(^!in!$U($0Tvi7$^7J`^}AF<))$pw`ObNl5A&Y4K+RQdzyuo@OBO^E2X
zs-qsVpSZszYRFi$fs~?VMdFD+5QkVy0<Ht(u-vH@+he4rnXPPdsR^E3VX*-v-hYxp
z0jE;I1-c(hIx!Bx!ah?=(wuK3*EF-0w&9)d5SN${niOW2_Y4lwiMzaGl6z!Gc(V7d
zO4(}9a=<VDWtfi5{43euOWYm_p3{Mbr|~Fnrc+us;Zm*s)Qy26%Y&sgV}ez!aKh&&
zb$=qoh3r8TyPsX@m8qcMzgxba9PNb5hccvWv|w>fdrSQthZW_kF{ii0-B)08C?rTe
zMz&kGv_Ce1!vU-OS9${WQVF@IQV$~J4a3}7mj}NxH(_scupXIc9340+rv+QKWhUG#
z%ka?fHmnw1nxi;C3B758%otV<kegrE(gE!CHd3h%|KpH#8XDg92?}@MLI9M!!^FJk
z@M(zou%KBaLj-(p3(FZJAM4@XQU8Py4Z0z5-y0f)AE~qaFT08sqvagvunFI3fT0y1
zU^P(Txa6X7(w=>D=A!6Ew?7b*fX(xYXV$`FBEe3mg<5sUA-;1NFhfE-#0z>ZIaj$d
zWzh*{G9<fvQ)-SDN@>R4_-$($k$1w^YJc!Weacw?hv<-ws866RS*!-v6s=SHQU1yP
zUXf2-J~q(3ltE1MGkYlux32x*Lp6-CYQuMfL~pAaC~&=k0!5uZz|7tF6yJa?YrMXV
z%f92k2T$%)R?Dlu7^VyHx=_ugTh$bqm~OH@voCN5;=4td&B`2y0y|g$jE{YV5;G!U
ziR}2bNZNfSM)2~1_icE_4-Z~;#BQ$r#N9X+^M8zLxf&wCr||(4!5!h?sfp{rxLu!d
z@6Tm!6!VDLf>mgAu39Dsw|VkYH-7tpeKo8;zq2ZTF*AUoBzE`j>7IW6**09EI1ci}
z&TuIL1&0?(!Vu=uqV6=ez#-QwdM61n#i2KnDsq^Z++^1%mQ{>|-PR(mrX-o+n9=*5
z!}MW6cc`J}1Zu?0{lhfamr_(cHX?pf2^fpKvL*fxT}jhM)YLDJy?j-RJ&(8=_EuzT
zG`Qw+&{^z&A)(xs>~jI&tn;4W>AZsyI1}|#mgNNXFA{-!f@ww{Lqc66xwb?F6*Fbj
zbvq^&+cG=*pFKUayn@WWs(=dX!Gq3_EFEXFU|&H*+u=)tU|;u#_iJkVmw#Am;XN6g
ze9-=8e!U=B5ORuQrAZ={F^`~z{EPvpeh^(UJoP9f*KZaKD(YR`w9%Cg-zk$hcGyMn
zSnsf>swaURHvw?HIwsTh#j_&BQ`B-D@g&@(5*Ll}^PtlWil4E_+AfbgL5_!VHxx^6
z9nSHfl*z-UV@NAolqzyQjLXTh2#rkvBFH;0y<qcNl35J^j0XE$uyx*f7x-efo`#<_
z!ue#CT{K4Q90&@QDU{A2(oc4Wh(}q3$sE)2)vCchdU@BExBeBVBY++aW?dOvhTV_?
zvt{O>dN!?lXOZ&h-v-z#e#jbE<(>GjvL)uwK(0=N_F=_<#v#`IX+8JQN%_pR?mgYT
z5D&OH-WI0$`A?9(YE*%Q8I51w={AO$y{_~7Hyr1chwCgyC;If8!ifNGGWXl^KTaBd
zNlOmImd}-_?bjOqa;@ax`bZ9^jvCPSk51^vQIBEVWrxiMSuDq7BQ)>`Sn<dMr&CJ<
zmwmZ8UF{rbuiUM!y@c+r^`>XL_aR;<;TfJZ;;GM_D{|DAdHqYDhHXCr7%$i2u3s-#
zizhX+#$5~USJHH^YGN^mgJIAYqNP-9upF0R9(+&=AL5_|BKUUAd}anAb?lLgk)IV~
z4@XM9w<|NBx{PK)7pQ){4zSl*3NI?AXi}U!hV(Mj68aR{4PRp6h<34bFF$cDUWUCa
z(D3Tl%KXp(&wrMX+e&>%c){-!1z8V<v~?puCHe)X=c;2d`xj)68(e4+?W73yrC9nN
z|C!xUx2@#~^N8!zMn%&%v=udTHy>)^NoQr7`jz#oe9<($QRQMAsauvN&(<!|?S%AM
z=ylf!<p4RU&xL0}9WSD&7t#;^g1Ua)GfNEy`&t<u`3r{dq0&2=7pnE;Qz?D-J)6))
zy9%v5-7GUQ%1F<m>wX<R#}WP(NgMCdTvK6zgq}TuvlPkBO-)X6!p*8G#i06v2q#8w
zhQIo2pk_K5LB%Gfzn`Sv+mFiu4NMb}j@N92t6aQ9M%)#{&dh=}9SbdkHR<RCq5H?8
zQ8&%WG8WRP_}-%?uq&smSK`k!em;>4sjaGRxQD|m`k6{$SidD)>aM2%e=+SNEZR<k
z!k>nBoFbfjxYhS>x-2QwQJQLR2EQbluSa(&L-m`fPkBl&-rBxXytbHv*pYUB%#gD3
z;EU77sL1k`w!Z#okgwTw!*Fhugn+z)XQyM9{=^N>*l2lenD@?BE1tqb0wpwe-|$M<
zmrD22zb*IZQnNZ^YFa?d^Cik^c^9@I9V-pmY9_-V;y5L62K9fhLIUjVdtbj>2vRMa
zpmuV8vZ@io^q!e9&&r(NF5^hE|4yq5ZU%IXl)7MGya)Ex2~DCwf*2c%2+4n~t`d>k
zS@cM?4H~PcDk(h(-pn`9!8XG3pTBqjIvdjWw!VmvpMuKoR(wSXoO6;ZV!|O8)ev#i
z(YcKctC+z@mOj-*fdCn=`v`nG>GlZ>c-Rp)q<g7!gs6Locv2Z&T8EIl7t#JVNBh0<
zM4Us9Pxt9%^zfbawKQ7}4xl})eUKB<JNo5B)bN+>aQw$2a?(k*hM1)D3A59Fr3W8-
z69*b_+GJBJtrv|1HN}hc`Z~p!F6j`CMF{wf9;F{sG_U`rCxMH@ul9WUNQiqDke59~
z5J{YxUs=cP!LO@<RLfcAN5ct9lgn3}%@=FFZG=YcbpK0ZrQ7Yvjr=Z7nJ%kc-H(r#
zEEv*2!PZY*HJ?zG->q5PIo3DF8hha#=$<p&R=F4W;xw_t$`CwnPG8euHyEY*oc?)9
zi#FWIJrAqlk!Bk*I#6s!f?cifVbvX#RN*1z-6FM=w)M@MT=PojhHbw)9|tA^kv(+N
z%MNgil);QaN>?O=ZqXrIzmV)3w(ZIDboj}gF)3Vt!Tr=xO`JV<Tfj_t!tq)=it0FN
zA{*f`((J(=D|^R5=m)b_Qu1bgjB~7P=L9hG?rUC5%Uk8wXHi!lo?IgK6^BZT8ZB`N
zv9rzx>2CzYPV9zsD<Df~we*=oUv1C?!w=!ymMJ0KW$_Bc)Sx+-Xi2#g{e3$=-?+{K
zn_|g*hf=)izNSBZU)z0Y`5fzqcFf`zHD#k1Yxd!{EB>6;d2&TlgtHKlliJ!Yn*6li
zgLy6D{3W?s6IfH_F}<4~B0jztrgEEF(&N`Q>XRi2yAqs-J;<4+%*Sz<8LI_0z`w?$
zVD{H=!E(iy1>I=-WkiLWMdWtZSMvHep|{dNtm&XCVf*{d3DZh<OKIYsW8T4mdc@Vo
z+MGaB-9Z@Elyd%8d*}Vs)c5s!KB6Fk^bP?HO_1IRB?_Sl!bd<lA|0ej38AARy@uXV
zs!H#@N(;RiK&q4g(m{kkAouWnp1IEtckVxM=kCc&CYjl%>~_vxYrS79{G!G5L+XcI
z=H(QNKbyBOn8VqtM(#r|f9}f_+s>I*RL%?I^3n3$!TXx4V%%cs>bOC~ll#A`=|^VY
z**f)Iyih%RGL0U}yW+$2Nk@Kp`uukdTa3M?4x+CjqMjm<di1l#bZ*u&&rgH~Vjn2H
z6Q!QDYRn+DJU^NB(2WLNaA6bB_gLsQ^uZ)e{Xd`}z)hf*f-P+Ae(XY8ZruTcI34g^
zdV2_#wlW`4{wZ0|1ysWXC!5!FaO95_djTO}3WZJ>(?_5RV3U~~!9j8*>J;QQ4;qBO
zvBp8-r{+WsC%U4xS-hP>@?8N_w&)Lgb&Y&v!1JK|9DK?$3ACz}>U`mA!N>;R1?<2E
z4w;B}#cvG}OY2n(v|*F@s{Z^<jo^Li$M#(y{~SZ}Aa74BbX^a8M7oG5We`}oc6kov
z`4t5QzsFAO9(b+)ZN-36pTjdw(oemE{Um1hgYdOf{ca7wC2#P#!Fw>!(+L-%^JUAj
z1)E^Q`%yjXP!D$Qj*{t_R>qAdrQq1lvQ|D(Gd`cE?>F;J#R%@Q7S_uYm?~#PZpf%i
z{)E72DU6BLzOu^zOW4vyHH9C~v;jP0a@w;t_qH5YX|#grQ6kKxM(wqG54Ad+e50JR
zLj8JSA58+A9Ozi9|HV8SwKS1(#}Mq3YYxuQc2D;k?w4(sX9C>9jjV{+3pxF`Ntd-R
zUzJyz`YqS?SZ0MLjeS)g8@oX|M3Z^_hvIO!8F9Svv>k|fGMu|)p#h9&P<3pg{39?R
zBf0F+2LGscY#M*g?IZH%HSCVjbJ;lEgagk)mawK&HTSuT`ckbi{fM3)!}_m~>*vcd
z%Q0)$onqkz=;qgd(dSY4#rvzD>_;lbHgNO4(b%`}#)of+g<#GBE(O%;I2jYLn_Slj
zirx51ByLLn4TXdQo16A_LR)_nQJZbQS(ZaRBX<rlIL9x+=Bfz@6sMdzm!bexRFzxh
zw%(Iw*VV@l?V19eB21#^7l6p|)4S`>G7MEI)rIh{|5(|Q9%PeK`WO_|EMU-t!$a^=
zx~fwqe@cmf(%u3Av#iM|HS55f=btbA1MAs*Z)-FzKooy3+Jhc6I+y}mjbsIk^v>Nb
zxo-KJ-Bebu(-+nBhE?-*V=xx%J|30W$ZXLF+H%N#%DISBGG(X>4n=cBui6C0&HCEx
zyQjA?{i-Jor0tE&Iny|CEud{;^?pI0Y|kRmpeDzCa252ZGj8xuv;vTPv<0@hf*w|y
zh=&z0PRS%S3T>hwF%)GSu)i{@CZ2u=gF#5g2y4^zPc4n)M-?h}%LU`~!)3I=*rucj
zDw8<B&>Si#&v76lh<cWb`(6Wtg37sH_gylNDUlS1@eq^G9<5P2JC)DAK8f#1{pmPi
zX73!DO7s_)-bnTCmhQD9=wS%PjX$we>2_TbVfq^oI4@U5?x%c#PU~2BGM?{@I%4by
zX#>xXx+IdmyV;nlg64%yjtOBtgpXA0eKU#^j#LFV1cf*{%lBk046D4D$Hz)>#65Tw
zQJ}HD4nA6Nn%d)?&~Kk>%n>T%u<rhstv;4=I4_B<&e4&|?l=P;zEb3tq)hdYDoQ#e
zP5>fkYoi^@MU8))$Pu>J`M#<`m4Kl-7qHzq9R@S9RAE14sdkeN$NdbHU~{W0B{q6h
z{+I{0zj`z<<n-FNHbu6vj>C>2!iHek(DRasLAB!`H4EYSO+7u^;e?RkX;7%4s#-Re
zX%8X#VV-vs+Z_qQ-`Y8^CW;T~9d+DT{%i&g_M~mSclD~vKSMGgC}#}qC{gigQcjVj
z#0<f?V^hc$dsF~ApPzju;d(+gVp*xxh%;GQ+SeEP2xnPRcMCB-=6m>lc;(T~T_(|7
zF(W4s_}!TF4`66{mkX{Gs>Zn~XaG%(X!`-{&WxNd=RN}_!jw1jTjxIt$16XNKEa8Q
znm&S-^DoPY8PsvN`W^}@h2fjQL@E%$GgtOY<OK2Y%bt3xu_~&F%~yf7+vguY%-(hr
zu@mlvz>4Hl*>~>xTqe`x@9*U(Ycw@?YQvk8HAwCESvs9f?Wd07dJP6VgpmG8PqwBO
za!CFp(9A@pezC|_<fAv}UED+0Y9YhM3>0)8e;O8G;AO#+)8r%Em3(s@4A#FLF|w#0
zV3u&1hJ|S+cbw9b_ePcVF+p{Z*S5TuJcrgSCI?$I6r|ePOEt0a3}c04HRMQ9jji57
zps8C@hVI#OqXe>Sew+pw{<~ma>_bn|dF8SQyWr)SiF`QSp!JY$lq73l-1Ay1Da4%l
zh`;nlsVc*BoQY{N<4S!U8|Ir#e?X5A21gx`X!_1IZ|P4|lSH+eDdMp0dB2$SgH`UQ
z>fZ4hr5x7dI^?5&VmT@34&4(L_*bSs$PHfRmaleyFnok*n%k!>KfszF;p06bdd9R0
z7TVdg8~I57=n9t81?xRP7Bu}fy<KaY=4g@4Wml75XWHn|%oZ+vhS8Sro&drV#$Pt4
zbI@#U{DarPY2m2bV`p(4rt{}klo4HH#SJf)bKO3@4B&p3D$@s1RC&I6H?W=@nJ*;+
z2&@Ki0~QO;Ah3v(yYG?i15s;|RJT~Ovzb&+e#QM7wUshh-hrMDygzbhpyiIQZ)kx@
z9F#oGrTgv74~S;$YIbSb=}b5^wkcwl!!WMPg=VKJrSDYBn!ZSTmyTyYWs`1=K7+QX
z<sAKO=WM%MZ|7)poeq#a##<b?`*SbDqOF5~;LsF-wK;*`^@wc`232F{71|zutqs1G
z{f<Icb$pU`1n1Y2LM^X0>fc|<7M`}M3~K*GDz59ByBy<~$==R14x-cDDP3iTU4eb)
z78;h4DY+s)GwA!J*JJcl-Tr9c<T~mjkp$nHo$Q5QBr!^y!X2H*wVm2?vcL78y0vyI
zTztt&eV>)o1V$M)0}>0^H@SXEq`C2wJ%L}sG)wAuk;kXKtXxNp>$)+;h}-2J#hhpZ
zUQ-(cNLyzF5~inCT(io?$sp1rQBVHT1)f)D+Ic{OEBNTOQZhz4GwNH5U;j1{$3_BA
zS)j$w9;^XTES!?`$jW_-R<lj3*@ss1<_bl()Cnl;O4USD7CCa*a^+7N<F7P~#a^r^
z6bV!qhYfUvdA<gmuSbPH?2aEH&(ouyJ5eHjGN_vd5Fzuh{5r@+@aSc~`F@Z6%p$}{
z#+kMZux~3x@qMM(KV|w&UeC50X)6C}ZR;b1<u_E#>2<D(ONYW$mWGo!_#bTVpW59Z
zBH81{Q{_kbT&82c_~LIL39E_?L10X6W95AD{C2e2dy*p0-Z$-{Uh>z*z984wQX<~r
z7#tIHZkYGCMRpF6kgsW-1(E#yOW`7<Qk7|Q>+86$PD5IUS#aRK%*+79)TpV>p0e!d
z0GK_A?CG<sKbhrz32Ke0IyUfr-%c|`7;s97LX#bkEweSZvV0%i8SLoONVyW)5caiw
z?Y=seRp5vCM}9;heKyMWf$VWGJbd~*8DA}NVA3y);Jvi9?HcinUDw;0+}9}7ZbMbT
z5_V&y4VwJ2P+*3J=9aM3TICW$w^J-40)H;l*K#t6*w?M{ia{SM0>0K2u*TtcSp%{0
zA0r*<xTbt;cu@rWEQLK|>u@P*>HEfhB}|o*dQ|2GMW0}>$((!_YO?7(DR#jXl-Eg?
zPk3M)_UH$=ywDNMl4SFp67shEG;A-)JA0&Y`69w2Ipvt&WOi)ovMotl|4BPn0tL08
zWlkrH;E`g8Z1|Y;!|hkfDIRPksmB#{jqkP1T91{aiIp9(EIXfjG0=^O+bx$29E;@j
z9A1u%s^0O&Bwp!%MrJve!k}B%$%kKH9}aUpqTr#woKQj79}tzQz%|LAMXetbpHIyq
z@yU|-rD;#{UzR(RUytth*Err;_*7@z;UQ<H6QJC_PD%dY4no4!!gAX@-XNk!Xd~n!
zaV=r~q5=xOY1}O3oiNOA9j~~KUK@THIDu*1Cu$NXt4V>x-UsKXGy0(vnN2gc{OVQ6
z;B98`lCEZX#^=>`)pEHj3Y<f&$xql{`WP!yqrN;oAB1%4jM}8qx-3B=<)t+cqxmwy
zS9?L{*Ig&9Cc@)@_|<<5F8+i6n`P@$M?t%Kc(f&MuuNQg3-|uz`I{deH7A+Mx3swx
zzWBDTuz8-YBGxk+1EXzIdsH7)=aKTx6aSDupFTy5l&+<0S$<P-oIm^%TyQ?7q0lBD
zIcVC+IKzuYHrFkDC(Coo%NG2FE9l?uKXksesAuzTN=^;+0Gi9@y+eEWaNff*Qb8Es
z*;}hs&~1!t>#<!({I)05@yq+K_V>jz+VcqfWLi}1!WWpqcqFl|1mQk{sPex2)RFwj
zAH&mMsz4adlJFeok-Y+<xMv}tOTsN6T@As<60Kz|;}Y!7iuf%Z8L{uykPiQCndi&`
zZ!+GUdOG>*9)@$lN^HrKqUC<#&iG4@O<^PlH5<>LQySnMZ6QHsuUz!35pQeV0Y+||
zcl~J?tsHK1>aOFRrSoSy8~@-~W~FzRR3<{J6#yQ9P8x$Ta<4%gbnO^OhP%tRK?4Wz
z7%Tg+cY3>;)k$5R1uVHn#8}-RqKb>CV}9g0m)slC#^wc-J>ttyLys<5Lq1C<r&g?Y
zze&Na;NwkgOD=MYQkE^RTJm<e2BNED!GV+_#UwRc95+uAYuD|WdAi<%ftT*8!brRh
z&ymQrwdY?AQ7-Ct9*n=4GoxQ{vTqh_ti*DeYH|o_TmLpV!ni>*-F0kZ3k7boPa+AK
zTrw>Fp{S85F5&E?UfM9saDPEGn(-{G$}^u(TX;<PyK5`V4pO|`^O;@0?!}HMHpt#*
z)+s#mm@&6SY=QYRv*%88&Elh_BFAY853mg;V?w@|c)x$%B}sPaM<v@o?@9)z#N%eW
z4x($=tQ7aHYoAv9XnKd*hcvjf3YtnF<kq~Ue{DzvHIL(+CJ#Q{UOsghvqwCOp1%h^
zI?t7vgM433d3%RtGBJ)O*GM>felBV4aZkO8gV?*jo>pH_x855q4AHap((XjPcDdD1
zO(#1!22sOAuly^Cp-+@~3v9bgDk)qI%|FteKAQKfYh4z^BBO;_F)9mvWHZZfFP{<Y
z+xrU<`!+>xG$xF(D%Lgzn5tPv(tVqZ%ed2O7K50uE!Mgy3Y=@bm1d;*u!D+Og}P9?
z7ow%?yVM>U!61_E=U^dT|7*c+sw%fXio~md2yy&*9%Wk!ZaN}tnumNpPI>#-c`k;B
z<|mz{$bkoQjI{0sl@cK!2l}J^_F1Bxc<N*tO}e$bx30!{^yO^C^KBI4j&Qm}aE-$N
zNo&UJK4+A7SdCFC&G~}w;p3J}yP>}!5A}I~07|)XZdEthn#IQeH|u)mSZ?}9J|So}
zxJ*;^fe!yng3^goo**8xyw9<}=?0y*CR3JszDpQkUc;DvA79=wv|IiLuC)%6O1Sjt
zCiW@bdQ(Z!Us=4}+SZ%3vPT~3okd8qS0Uiq^ck0$dcU7vM`3eZ<hNW8E1AXJWa(HJ
zr~VA&#bLA+WdL^4xAIo!+fdzfnmM7>Gyn7opTuiAf2CJri0D;kwkfiVCn@giXD`C0
z<`T=orF#9TtQa+yYXg@RHnXY&9m1(mYq+r6T4_chVDdHLx63CY`Dp_Tnl_>MenE+H
zX2%+Z5&JG%r%`TnmGUz;0YNp{`O5nZjj7&|3;}r>)MmtN>aVwZO4MUf3)tPEi_ohE
zjHHVj`ys2)Mb_Dp`%abA<z!QIu|)6myvb6AZ@LU}*%`0P0}+{6{H}YQd8{z)+*iJ^
zGgbS%omLLI4C)g820(@8boMu1@wwggo5+c|rSBW`izf8g{$jsF-&<o#!<^4Uyk#Ob
z=2n~X<<u<|vFLwV9pyE}G}2JCeMcn0dd$(5^agHpb<zoo_WeXp=X1JLTf4qHD8L=4
z!YVE}Gog2Y_lS`F_VqXo{t3u#@2jBet`Q5gx!sdemHTgAEJHr2%{7+I3BC3xaU&qE
z9B&)eC3|M65>YaOchkC6ta#Paf25ko7FaHu&JeAnmq!R{TpC$7TUz7jOK}#WXn+7>
z&A(3Cx@%OpZL#wD-6f&7aqf{%`$w=8c(>d`x$ozlc%2J|7Go!rzSEFPDXH9;b#RZ)
zdu-ey-Ypm&xhk+B!bXU;397deW8U;1drLBf&C>Qsd}^OykRkBvA2tHXb#o^rjf{V#
zD$wm48Y)MEW5ug~kk<T`A{vIBrkF#qnsiiZ^s2n`v&?YpLa*5qvA=l&nN=c9t~=Pr
z=EqDNjKhz1EXsazZD3m|F^rui(qebcZbjADuAXhe6eHZq&Ee!L1oSNkA0&f+t%+Va
zbHDqY9at+Q$nRU80MnsR^QH6><PL%3L~=R7zG|OH$iaEef#CqFoQDyis<Xb8GuqBs
zvfxfkDPe`Rh-$`xl}Cemgmrq^9JAr?z7v^5Y~+E_fY5L;cgc&ypW?5Tm?xrR3>uA$
zt^Iooz!wyI+qR%QaWSB)2{I;-&m9JuOO?#tq#pnER_G8;WEy@g%HdPD+aML=Oy%?!
z_KasORR(yb;wo%C3z%5O-!)I?O9C$yU$}7HW5XD~B9Ra`GA-1Z`~m|0vR!iSLJ}H+
z6@y01U!R)`IH;0ljK@Ptject%-ogln8EI}lZeLmBU(9FijM=~*KLB>V*?z*${bHl6
zbLZd7=SjQ@(ehk{y4<_L+WC*0tt>RNem8>2U0HKcaG`a5vqs8Q>^hPP6hLJh>nrzM
z15In$w?n5!#TB_E^W`!j4CzJ5_fUQa*35M0ZBxNS{hzOEAelO0%}?NOMB<j;v15@(
zOt?>t(J4<#oi+}9xH75*0f^spZ&f_DQwTSJ&}<+#xfNibp}S?UDW-2KWtK^QgN4!!
zk-&y1;|BfdMIN?<#cHq29)pzWZRtt=U_RdFo{rLERoBD=(pFF6GdgzBEDEM=hV`i|
zlT1NY#Z6c)Y5TJ&iyif&j>hjks$)$)zF|A190YI0;*3-urGj?|bf~aR`veNZRtk;j
zA!!DdToxMm7veySK@(X+EG1fqwR7QEp7Y;`9&N8H8MH4B9MZj)vn7u$65kMhU3Hl?
z#@NruEo$@fk_MMtN1ZtyTJI??!}Kjz$lzt;zvpL*_`-F<K>e$pUkV^8w_Hqf(g+D(
zc1lGg33Lh6pBPa_1OUPmS7c(3HUj?rV#N8DH>}VTN{!rY8Z4OMt1Z~b4U2C!(zM`%
z9@0QhXgZEag>wh>h|vY%0KSUI2}bihk%)e>ztq)Hs!}YwOv&2ttdexteUqUj^V*m)
z>khzVDabq4Zi&)krl&%>H-;AjjKJ4i)X<T|!8_=gCk2#^SN5*i|9nvb8JmO|h7jZ8
z&{U=8h5Nz^7D_v;Oz`9&6r|gIzc6+C>F5T)TYbVOxI8bh0s9Y2km_twHZ?3nwh!d1
zTV%zqTp&K`)&viYK3s4Db$Omcr(W``Z>(>4Ud3YXm^x&CX007p;m;nW1j2j32_#C=
z0u97O!0XD~pXzpU@ST<T>9Iw%C{-K-n#u{QEi(#HB6_|aJ`JJ`Fb|a+76MSPbm`_j
z|3p`50esP~QlW%~N%Kldqu}?K=babE;ntZY-8N_@QOk#TJ5X0n5Wm^UL(vtUzkh4W
zluCz%F#bDaHu^VrAPEhVhnoV#*Q7VpW@(8#!NDnNI2~W>zi9}hr*r~9hZ0@YhxS4g
zfgqrK;K#aD^M&5%%psxo4~P<z!{->rlc~9wB&P;|kIO3_vv@9f78aJR{DHkUrhZ0=
zbI4Reux>P-O3Y(E@av)XSQS~l7LrOsrn|!mln1k0hk(Q_fPR7DXuWJhZ-}Gng-pAe
zucUMcQ`O!LF*$q5z^eD90rTQO_|#;BVRxiCOiBSb1W7U04hZBQ4niWIn?wN)gLk<w
zJ6q!$1nP9U`lf%X6_F4U(AV&BAi7~XeLq&HcgUDT#-TQ#htF|Bo))F>$my2xv%1*I
zdoy-<(}f|D3HJiYTI{(2OmgXKqL%fJuZVLa&hrUVCpUS`Ho*OTuNzegSPW?bF|<2(
z9oFosqa=m-e{3k6YSiE?V^^hiXiM~f2DN5_revS@K|;dA3;Luy$=s%dflc+f5Mg2)
zb05Z|7JeJ0(9o=9&zn!vw$FMUoQoRucgY6Vm7f|Isy-rx=H#H6^uJb{EJ8jEh+dO=
z<$Ykk76gNX(t5NkmPl23k=|n>1;@kh)J9LuO;qD=jeh_*A=cd^f$!iSJ&&hmpAI=k
z!at7<FQ_a_5%SPFe2pmx%~kBj73%>oJ4}p4F;3Fq$|MW8H5t~<3X~a@u8>D6{5J3H
zz}qdyi9oCP#gPL1l+;|5V05Jx60%tk6fts7%eEE%@#Q06)E!IVL9&M#|AE7vGRq#D
zedz-yt98a<gj2XZlr;l-kxM3F@!c!Sh1vU|$-iaEk2*}k)Wo2PS{Id-S<5mD%TAy?
zf1KQ3Ev!1|rsq>OH>)YmZs4_4VrNwp9^DcQWw2Y<ShCWuqcatrr!}QYYQGN6C*Jdv
z>K33YcGo)U$dO)w3YW+6@W`aJA%$y^9m<W(4qLpQ>-;yMA4RVbLAMz4S@gkRy0}eC
ztp{5I)8SmyhYcPk3y&@LVbGtD547~qG4?@q{@PKbV#m}-kR=FAiUT)0?wG&62l7|3
zn?AQ-&0h6MyScz?l=;xZ;hLLy7Ln2a*_5Su;tLn`C%dQG0j;yeK_>F32SF6h*kOqB
zAb(*CyEO>>xb<x2?c0z!if`Tr6Qgl}K~5IW*>7z3=mio^fLuK<p3SK!)deWTmhF~)
zIb_RD>(mivSy}HFu^uN-_Et-->S%UVll|#}Umg4#Tl|7)m4JG*<sQ!B7rECF)G%xy
z?qKXIn(5gm;BIvS#htyEZiXA@>YbmIHp;5IK;LqCZf?)EF?z>~m2s#&JR!w7c&?M~
zuzo$1z0ox)S;w(IH+fa9WwBxqdsKX>j21cGyd3992FA{f@2zLdHc0Sp3%3~nlifrJ
zzJ}WNJIMJ)+{Rgw3p8#>2FSeHoK<tW2Gy^qq!?}r*S{mwdw`0`?QPF8=~Cd2F3b)Z
zhrk|P3&uYCbyhI|c*jt!XHu0Qnz9@6vN+9LEX(RhgZ3xbtw0hUcU@|Eu62E_Q@Sgw
z5>CL^y;*br063_4V)j)O`eF5uAKtv)E&(3fvyi+8QoBY$2d}@*f6yUy=G?&bJB=&?
zc5xVIq~B~toRwF+TEy+vGdyNjPm59T?BB+LvuQGF^;rD$LC<H+s!<C|@%eKK;K7iM
z4;R?B0<eTrb3<o`q@`9LG3K_!MaIR3Pj7nY<x$5);`XyX1SbIMPR2bk(25M;d<q0u
z69kdsS5wz7qAV56uHfj0K=g-pC2y+eGDNleVyeqvc=F%Z*9Jj+eze}{dlvPF6^3Eq
z-{bN??1sgLS~o|oJ|vt=BCb1X3Rmm$j(^U?*pYc#gLB2)vPDz-JV%ijp9R1BH4Tte
z)5SEkcPFW1JfaWw9}n8ote9JQxCn$d`Q!iGH^aTS=GkF9I%nGLDMRyK=K;fj>e?=d
z%<r%c4nV3=^D~|p#z3vS5*Tpo{wj$$o_Raru_?5wjJ{fd6E3cufyLu7X;~snDzZML
zX2aI_+^3yqkI5yQqjI?^CnYc*9zBaYXy+jo?S~M5GU80&4HVAw;E4K9|Bce0uuIy#
z)mh{+1^<JOf=1Jze!e*h$-<nSK#lt;kCw4H;u}DKx@lpjtLd;LfP9SZURSLI@w|{8
zvk9K-KKu5lCI%<wq{2*3LuW)x$xDGRf#wSMe20XEhp$`THF1Bnd!?F^>K$>rb9pGG
z)p5cjR-ve@V+!(dAQy*dXNZxGLZQ{c-eCe)bqWFfS@dKb)5}U?<wEb8)YfL_5#@%x
z;&=f+j|89hwaq2A0bH=b{vZruji9}96qv&sq7j6%$zut7(*?jXLke;C<@S1+)<0J(
zRI0>TNO|saEt%matW=tKTRbLzdyB`o|B-)S?13^V8z*39)861TzDzB%;vDk}q7d*N
zkXAvS!YO&7cz!@DvwcJKF3>FU;1lqX#Kr>ry_M7pS^H5d{j1c8C&m{h{$5ZOwWlW<
zINd-4NsNlEu^nn&PSAn|A5T856`8vlXQ7)ju;M@=CY@+|&1P<KMt`_`nLjt-X~5(>
z%a^{+aH$}^<^&?*2Pe2g-;$4P1*3-DYp6-y;nD1)<o-_SpJ~}ACJ?8(Wme%yquR%t
zQ}IgPTs89(Bb)je*pdMB-$#RT_z-9k{es=eablv-_EW2%7ISM*PexFx^LFSD(R}dE
zQPldN&|yO*Rogv#ZEh-L_c2G?Y<T9Lmym=2lHmMoHnhB8uOLlGChhOF^A01}W{m;M
zbM>mUpSLx!%IF`D#@N30Vt|9Grn~_?;~7udNtA&J2pzZ!k6f8uOYJ<^UTI^0@|fJJ
z;OV9^I-da*mb+PDr=(?*aD#$3>f+S0Q7X5~vAKIfP2zn(Tum2t^c9`}jmEq+J}(yk
zWnSHf-@3fWo{jflN{LWJ8Dwc&1tCWxt1-pcSK{k1g(@wb59U7a%C2Isc?JATk><PX
zc8<-5&$>H~jO{tnH-k<C?6ssqZ?N@L1fdf#m_O>ryBVv6z;XoiR{s^+b5&-I!ppTv
zs_e>&4iVF0otyg`ztdm`id}1(W~*pR+T8B$yH<hDLn8Ut7|#6ySn7)?YD#NLR|vIi
z<%6=cHK^p_t!KfwSK@0BTTZkC^7M<c;-SVBL5KWa6SS@(1$Z!Z-P$jfpBH;qh<@=v
zf-AuWXA&5c;U0o|QB>2IJmSxTueOa-dw3{zMHm?LyNaSg7!4V@r+_hKzP!Rc$c^NA
zYLtVY&K+#t9ZUR6=g<9?cLe8k?E;;FJiEYF;FQI!rS{_jIS<cp>|=&;2Ar(W1$J%1
zQMJ7#Z!kDy*J||BKmD05N*xJ&=zsJvJRVIgA7WX%0!x$CQlq?M`hEI@YS~)U>;wN1
zI-+ib&&DinsnHAx`*(ZD{qCr3=hK07gl~F}{`e#YL`V8Y5E~A@%~>-7;Z0u!lDHm*
z?sSpwet%Vacd1K<=;Q%mElV^I=oy7Sf;IsB7{-{gDInv(-9(>gRb8*g3<Rl5yfo$1
z^EFVvNl{}4Qt#|y@1-%4`1(gWeddZ4Z3v04^jzBurK2QoUmD6k`>xQVt-n%L{@u%|
z|8MPQ-w%t|Jr7=4#cOYfL|w!@iYe(q@1k6shTf2Pp3%9AgwF(AG{lY&%*2DX2maQQ
z37dN=RGfeTyG82p)|r#dMEQW?;vp)|qjr;^d14v@(D1u1BJCJE!yFc6MS8gsI>6;a
zf;TB3?$-0$ldy}W%`g=Oa}-*-zjQ#+JKcPKA|XebuNl94=aLw=r@+f0&T|o7PTw!O
zSj-316f7AJE*#Kn{IGIvlkp6Fr*3ZX?&P)49b)fA$m|$%--HQ%wKpYguH=#B@>O~Y
zE2tD$cP2WzM^hS{BuEIHKCgTfU2EfEJpKV9BVHLUbpDabe^?E=>?;m-_=PT#8L{u<
z7SC7D;d$v@B$neAZN<M<8w`8szEg9jY&Hwyq`8=tY1G5XEfQN5ZPBU_$kSWa7g{=W
zQK}7!G)(-dR}5J>?6ky)L>b|R8#Dm`{*EjMU3Q5&kC%mP{pW9RV5c;Eksap^WW0FZ
z^$KAV|Ct(fuWF};>c0IGC3Mj!yXz_6i%jS;pw31FuBo=$aYIFuh0=PaOSSsx)MHDY
z5~A$_&~^gof~`2b3CJ*p8HZ*0%R>6j2RF<gXh{zhNp?LCb+|_CP^r0_f|#e`iNrrC
z00}@qZ*&rW4qYF`p%PDwI^{5;QR%pzGyfqeO{)xB`Uo|zM&{Z=LXf=@;eS$d90{<C
zMO#D<pqnh11pX_-38zI9-5bX$uFM+V%<&{5y?QNfZ~~vkc?wGNwVV}GujY<)6ClF5
zDFrYtr&Zl#6>(cabUqt!_JH3nLX7pkzp&0Wf^lZddDDyJ_5z7Va5ou95dPn-i*yLY
z2LCJR1QqWKpT9CdJ2{cURibZlX8=Vl6p%Olcln>(E8&1(;J*tHrvBgM!~g&DrX~I#
zX^>kuXf&_Baj~Wu9!UGd<*97wX>H+Y1F>?q0lo-Cg+(O=gvAAfMf8MaAYzgbapA|p
z!VqEMa;@r{OjQ3<!P(W?!Pe*hUSXN$-ZoI-(f>Y!r-QSNho^<}+yAqUgw%hX<UJ`J
Skm8DfKwVi2R<8K^!~X(~ihWK1

literal 0
HcmV?d00001

diff --git a/studio/frontend/src/app/auth-guards.ts b/studio/frontend/src/app/auth-guards.ts
index 1dcdfcb143..509b8f61af 100644
--- a/studio/frontend/src/app/auth-guards.ts
+++ b/studio/frontend/src/app/auth-guards.ts
@@ -2,12 +2,14 @@
 // Copyright 2026-present the Unsloth AI Inc. team. All rights reserved. See /studio/LICENSE.AGPL-3.0
 
 import { redirect } from "@tanstack/react-router";
+import { apiUrl, isTauri } from "@/lib/api-base";
 import {
   getPostAuthRoute,
   hasAuthToken,
   hasRefreshToken,
   mustChangePassword,
   refreshSession,
+  tauriAutoAuth,
 } from "@/features/auth";
 
 async function hasActiveSession(): Promise<boolean> {
@@ -16,55 +18,64 @@ async function hasActiveSession(): Promise<boolean> {
   return refreshSession();
 }
 
-async function checkAuthInitialized(): Promise<boolean> {
+interface AuthStatus {
+  initialized: boolean;
+  requires_password_change: boolean;
+}
+
+async function fetchAuthStatus(): Promise<AuthStatus> {
   try {
-    const res = await fetch("/api/auth/status");
-    if (!res.ok) return true; // fallback to login on error
-    const data = (await res.json()) as { initialized: boolean };
-    return data.initialized;
+    const res = await fetch(apiUrl("/api/auth/status"));
+    if (!res.ok) return { initialized: true, requires_password_change: mustChangePassword() };
+    return (await res.json()) as AuthStatus;
   } catch {
-    return true; // fallback to login on error
+    return { initialized: true, requires_password_change: mustChangePassword() };
   }
 }
 
-async function checkPasswordChangeRequired(): Promise<boolean> {
-  try {
-    const res = await fetch("/api/auth/status");
-    if (!res.ok) return mustChangePassword();
-    const data = (await res.json()) as { requires_password_change: boolean };
-    return data.requires_password_change || mustChangePassword();
-  } catch {
-    return mustChangePassword();
-  }
+function authRedirect(to: "/login" | "/change-password"): never {
+  throw redirect({ to });
 }
 
 export async function requireAuth(): Promise<void> {
+  if (isTauri) {
+    await tauriAutoAuth();
+    return;
+  }
+
   if (await hasActiveSession()) {
-    if (await checkPasswordChangeRequired()) {
-      throw redirect({ to: "/change-password" });
+    const { requires_password_change } = await fetchAuthStatus();
+    if (requires_password_change || mustChangePassword()) {
+      authRedirect("/change-password");
     }
     return;
   }
-  const requiresPasswordChange = await checkPasswordChangeRequired();
-  if (requiresPasswordChange) throw redirect({ to: "/change-password" });
-  const initialized = await checkAuthInitialized();
-  throw redirect({ to: initialized ? "/login" : "/change-password" });
+  const status = await fetchAuthStatus();
+  if (status.requires_password_change || mustChangePassword()) {
+    authRedirect("/change-password");
+  }
+  authRedirect(status.initialized ? "/login" : "/change-password");
 }
 
 export async function requireGuest(): Promise<void> {
+  if (isTauri) {
+    await tauriAutoAuth();
+    throw redirect({ to: "/chat" });
+  }
   if (!(await hasActiveSession())) return;
   throw redirect({ to: getPostAuthRoute() });
 }
 
 export async function requirePasswordChangeFlow(): Promise<void> {
-  const requiresPasswordChange = await checkPasswordChangeRequired();
-
-  if (requiresPasswordChange) return;
+  if (isTauri) {
+    await tauriAutoAuth();
+    throw redirect({ to: "/chat" });
+  }
 
+  const status = await fetchAuthStatus();
+  if (status.requires_password_change || mustChangePassword()) return;
   if (await hasActiveSession()) {
     throw redirect({ to: getPostAuthRoute() });
   }
-
-  const initialized = await checkAuthInitialized();
-  throw redirect({ to: initialized ? "/login" : "/change-password" });
+  authRedirect(status.initialized ? "/login" : "/change-password");
 }
diff --git a/studio/frontend/src/app/provider.tsx b/studio/frontend/src/app/provider.tsx
index 68ce3061bd..b75998a169 100644
--- a/studio/frontend/src/app/provider.tsx
+++ b/studio/frontend/src/app/provider.tsx
@@ -1,18 +1,181 @@
 // SPDX-License-Identifier: AGPL-3.0-only
 // Copyright 2026-present the Unsloth AI Inc. team. All rights reserved. See /studio/LICENSE.AGPL-3.0
 
+import { StartupScreen } from "@/components/tauri/startup-screen";
+import { UpdateBanner } from "@/components/tauri/update-banner";
+import { UpdateScreen } from "@/components/tauri/update-screen";
 import { Toaster } from "@/components/ui/sonner";
+import { useTauriBackend } from "@/hooks/use-tauri-backend";
+import { useTauriUpdate } from "@/hooks/use-tauri-update";
+import { isTauri } from "@/lib/api-base";
 import { ThemeProvider } from "next-themes";
-import type { ReactNode } from "react";
+import { useEffect, useRef, type ReactNode } from "react";
 
 interface AppProviderProps {
   children: ReactNode;
 }
 
+// ---------------------------------------------------------------------------
+// Tauri window helpers (only imported in Tauri mode)
+// ---------------------------------------------------------------------------
+
+async function showWindow(): Promise<void> {
+  const { getCurrentWindow } = await import("@tauri-apps/api/window");
+  await getCurrentWindow().show();
+}
+
+function easeOutQuart(t: number): number {
+  return 1 - (1 - t) ** 4;
+}
+
+async function animateToGoldenRatio(abortRef: { current: boolean }): Promise<void> {
+  const { getCurrentWindow, currentMonitor, LogicalSize } = await import("@tauri-apps/api/window");
+  const win = getCurrentWindow();
+
+  // Ensure window is visible before resizing
+  await win.show();
+
+  const monitor = await currentMonitor();
+  if (!monitor) return;
+
+  // Convert physical pixels to logical using scale factor
+  const scale = monitor.scaleFactor;
+  const screenW = monitor.size.width / scale;
+  const screenH = monitor.size.height / scale;
+
+  // Target: 75% of screen width, golden ratio height, capped at min 900x600
+  const targetW = Math.max(900, Math.round(screenW * 0.75));
+  const targetH = Math.max(600, Math.round(targetW / 1.618));
+  // Don't exceed screen height
+  const finalH = Math.min(targetH, Math.round(screenH * 0.85));
+  const finalW = targetW;
+
+  // Check reduced motion preference
+  const prefersReducedMotion = window.matchMedia("(prefers-reduced-motion: reduce)").matches;
+
+  if (prefersReducedMotion) {
+    await win.setSize(new LogicalSize(finalW, finalH));
+  } else {
+    // Read current size instead of hardcoding — stays correct if tauri.conf.json changes
+    const inner = await win.innerSize();
+    const factor = await win.scaleFactor();
+    const startW = Math.round(inner.width / factor);
+    const startH = Math.round(inner.height / factor);
+    const steps = 15;
+    const stepDuration = 23; // ~350ms total
+
+    for (let i = 1; i <= steps; i++) {
+      if (abortRef.current) return;
+      const t = easeOutQuart(i / steps);
+      const w = Math.round(startW + (finalW - startW) * t);
+      const h = Math.round(startH + (finalH - startH) * t);
+      await win.setSize(new LogicalSize(w, h));
+      await new Promise((r) => setTimeout(r, stepDuration));
+    }
+  }
+
+  if (abortRef.current) return;
+
+  // Apply constraints and finalize
+  await win.setResizable(true);
+  await win.setSizeConstraints({ minWidth: 900, minHeight: 600 });
+  await win.center();
+}
+
+// ---------------------------------------------------------------------------
+// TauriWrapper
+// ---------------------------------------------------------------------------
+
+function TauriUpdateLayer({ isExternalServer }: { isExternalServer: boolean }) {
+  const update = useTauriUpdate(isExternalServer);
+  const isUpdating =
+    update.status === "updating-backend" ||
+    update.status === "downloading" ||
+    update.status === "installing" ||
+    (update.status === "error" && !update.dismissed);
+
+  if (isUpdating) {
+    return (
+      <UpdateScreen
+        status={update.status}
+        logs={update.logs}
+        progress={update.progress}
+        error={update.error}
+        onRetry={update.retryUpdate}
+        onSkipRestart={update.skipAndRestart}
+      />
+    );
+  }
+
+  return (
+    <UpdateBanner
+      status={update.status}
+      info={update.info}
+      dismissed={update.dismissed}
+      isExternalServer={isExternalServer}
+      onInstall={update.installUpdate}
+      onDismiss={update.dismiss}
+    />
+  );
+}
+
+function TauriWrapper({ children }: { children: ReactNode }) {
+  const {
+    status, logs, error, isExternalServer,
+    currentStepIndex, progressDetail, elevationPackages,
+    startInstall, retry, retryInstall, approveElevation,
+  } = useTauriBackend();
+
+  const hasResized = useRef(false);
+  const abortRef = useRef(false);
+
+  // Show the window once the frontend mounts (for pre-running states)
+  useEffect(() => {
+    if (isTauri) void showWindow();
+  }, []);
+
+  // Animate resize when backend becomes ready
+  useEffect(() => {
+    if (status === "running" && !hasResized.current) {
+      hasResized.current = true;
+      abortRef.current = false;
+      animateToGoldenRatio(abortRef).catch(async () => {
+        // On failure, at minimum make the window resizable so user can fix manually
+        try {
+          const { getCurrentWindow } = await import("@tauri-apps/api/window");
+          await getCurrentWindow().setResizable(true);
+        } catch { /* swallow — window may still be functional */ }
+      });
+    }
+    return () => { abortRef.current = true; };
+  }, [status]);
+
+  if (!isTauri) return <>{children}</>;
+  if (status === "running") return <><TauriUpdateLayer isExternalServer={isExternalServer} />{children}</>;
+
+  return (
+    <StartupScreen
+      status={status}
+      logs={logs}
+      error={error}
+      currentStepIndex={currentStepIndex}
+      progressDetail={progressDetail}
+      elevationPackages={elevationPackages}
+      onInstall={startInstall}
+      onRetry={retry}
+      onRetryInstall={retryInstall}
+      onApproveElevation={approveElevation}
+      onStartServer={retry}
+    />
+  );
+}
+
 export function AppProvider({ children }: AppProviderProps) {
   return (
     <ThemeProvider attribute="class" defaultTheme="light">
-      {children}
+      <TauriWrapper>
+        {children}
+      </TauriWrapper>
       <Toaster position="top-right" visibleToasts={2} expand={true} />
     </ThemeProvider>
   );
diff --git a/studio/frontend/src/app/routes/__root.tsx b/studio/frontend/src/app/routes/__root.tsx
index 69aeb11748..19b5763557 100644
--- a/studio/frontend/src/app/routes/__root.tsx
+++ b/studio/frontend/src/app/routes/__root.tsx
@@ -3,8 +3,8 @@
 
 import { AppSidebar } from "@/components/app-sidebar";
 import { Navbar } from "@/components/navbar";
+import { fetchDeviceType, usePlatformStore } from "@/config/env";
 import { SidebarInset, SidebarProvider } from "@/components/ui/sidebar";
-import { usePlatformStore } from "@/config/env";
 import { SettingsDialog, useSettingsDialogStore } from "@/features/settings";
 import { useTrainingUnloadGuard } from "@/features/training/hooks/use-training-unload-guard";
 import { useSidebarPin } from "@/hooks/use-sidebar-pin";
@@ -33,7 +33,10 @@ function isChatOnlyAllowed(pathname: string): boolean {
 }
 
 export const Route = createRootRoute({
-  beforeLoad: ({ location }) => {
+  beforeLoad: async ({ location }) => {
+    // Ensure platform info is fetched before checking chat-only guard.
+    // fetchDeviceType caches after first call, so subsequent navigations are instant.
+    await fetchDeviceType();
     const chatOnly = usePlatformStore.getState().isChatOnly();
     if (chatOnly && !isChatOnlyAllowed(location.pathname)) {
       throw redirect({ to: "/chat" });
diff --git a/studio/frontend/src/components/assistant-ui/markdown-text.tsx b/studio/frontend/src/components/assistant-ui/markdown-text.tsx
index 2a0517a44a..7eb4b21ba7 100644
--- a/studio/frontend/src/components/assistant-ui/markdown-text.tsx
+++ b/studio/frontend/src/components/assistant-ui/markdown-text.tsx
@@ -5,6 +5,7 @@
 
 import { copyToClipboard } from "@/lib/copy-to-clipboard";
 import { preprocessLaTeX } from "@/lib/latex";
+import { openLink } from "@/lib/open-link";
 import { INTERNAL, useMessagePartText } from "@assistant-ui/react";
 import { Copy02Icon, Tick02Icon } from "@hugeicons/core-free-icons";
 import { HugeiconsIcon } from "@hugeicons/react";
@@ -32,9 +33,13 @@ const STREAMDOWN_COMPONENTS = {
   }: React.ComponentProps<"a">) => (
     <a
       href={href}
-      target="_blank"
       rel="noopener noreferrer"
-      className="text-primary underline underline-offset-2 decoration-primary/40 hover:decoration-primary transition-colors"
+      className="text-primary underline underline-offset-2 decoration-primary/40 hover:decoration-primary transition-colors cursor-pointer"
+      onClick={(e) => {
+        if (href && openLink(href)) {
+          e.preventDefault();
+        }
+      }}
       {...props}
     >
       {children}
diff --git a/studio/frontend/src/components/assistant-ui/sources.tsx b/studio/frontend/src/components/assistant-ui/sources.tsx
index da97ff66b5..81c8b0c213 100644
--- a/studio/frontend/src/components/assistant-ui/sources.tsx
+++ b/studio/frontend/src/components/assistant-ui/sources.tsx
@@ -1,5 +1,6 @@
 "use client";
 
+import { openLink } from "@/lib/open-link";
 import {
   memo,
   useState,
@@ -93,8 +94,8 @@ function Source({
   variant,
   size,
   asChild = false,
-  target = "_blank",
-  rel = "noopener noreferrer",
+  href,
+  onClick,
   ...props
 }: SourceProps) {
   return (
@@ -109,8 +110,14 @@ function Source({
     >
       <a
         data-slot="source"
-        target={target}
-        rel={rel}
+        href={href}
+        rel="noopener noreferrer"
+        onClick={(e) => {
+          if (href && openLink(href)) {
+            e.preventDefault();
+          }
+          onClick?.(e);
+        }}
         {...(props as ComponentProps<"a">)}
       />
     </Badge>
diff --git a/studio/frontend/src/components/tauri/startup-screen.tsx b/studio/frontend/src/components/tauri/startup-screen.tsx
new file mode 100644
index 0000000000..af5f241416
--- /dev/null
+++ b/studio/frontend/src/components/tauri/startup-screen.tsx
@@ -0,0 +1,401 @@
+// SPDX-License-Identifier: AGPL-3.0-only
+// Copyright 2026-present the Unsloth AI Inc. team. All rights reserved. See /studio/LICENSE.AGPL-3.0
+
+import { ShimmerButton } from "@/components/ui/shimmer-button";
+import type { BackendStatus } from "@/hooks/use-tauri-backend";
+import { AnimatePresence, motion } from "motion/react";
+
+interface StartupScreenProps {
+  status: BackendStatus;
+  logs: string[];
+  error: string | null;
+  currentStepIndex: number;
+  progressDetail: string | null;
+  elevationPackages: string[];
+  onInstall: () => void;
+  onRetry: () => void;
+  onRetryInstall: () => void;
+  onApproveElevation: () => void;
+  onStartServer: () => void;
+}
+
+// ---------------------------------------------------------------------------
+// Constants
+// ---------------------------------------------------------------------------
+
+const INSTALL_STEPS = [
+  "Detecting your system",
+  "Checking dependencies",
+  "Setting up package manager",
+  "Creating Python environment",
+  "Installing ML framework",
+  "Installing Unsloth",
+  "Finalizing setup",
+] as const;
+
+const EASE_OUT_QUART: [number, number, number, number] = [0.165, 0.84, 0.44, 1];
+
+// ---------------------------------------------------------------------------
+// Sub-components
+// ---------------------------------------------------------------------------
+
+function TealSpinner({ size = 24 }: { size?: number }) {
+  return (
+    <span
+      className="inline-block animate-spin rounded-full border-2 border-primary border-t-transparent"
+      style={{ width: size, height: size, animationDuration: "0.8s" }}
+    />
+  );
+}
+
+function Logo() {
+  return (
+    <div className="flex flex-col items-center gap-4">
+      <img src="/sticker.png" alt="Unsloth" className="h-[72px] w-[72px] object-contain" />
+      <img src="/studio.png" alt="Unsloth Studio" className="h-auto w-[250px] object-contain dark:invert" />
+    </div>
+  );
+}
+
+function ActionButton({
+  onClick,
+  variant = "primary",
+  children,
+}: {
+  onClick: () => void;
+  variant?: "primary" | "secondary";
+  children: React.ReactNode;
+}) {
+  const base = "rounded-lg px-5 py-2.5 text-sm font-medium cursor-pointer transition-colors";
+  const styles =
+    variant === "primary"
+      ? `${base} bg-primary text-primary-foreground hover:bg-primary/80`
+      : `${base} bg-muted text-foreground hover:bg-muted/80`;
+  return (
+    <button type="button" className={styles} onClick={onClick}>
+      {children}
+    </button>
+  );
+}
+
+// ---------------------------------------------------------------------------
+// Per-status renderers
+// ---------------------------------------------------------------------------
+
+function CheckingContent() {
+  return (
+    <div className="flex h-full flex-col items-center">
+      <div className="flex flex-1 items-center">
+        <Logo />
+      </div>
+      <div className="mb-10 flex flex-col items-center gap-2">
+        <TealSpinner />
+        <p className="text-sm text-muted-foreground">Checking...</p>
+      </div>
+    </div>
+  );
+}
+
+function NotInstalledContent({ onInstall }: { onInstall: () => void }) {
+  return (
+    <div className="flex h-full flex-col items-center">
+      <div className="flex flex-1 flex-col items-center justify-center">
+        <Logo />
+        <p className="mt-4 text-xs font-bold text-muted-foreground">
+          To install Unsloth, click Get Started.
+        </p>
+      </div>
+      <div className="mb-10">
+        <ShimmerButton
+          onClick={onInstall}
+          shimmerColor="#a7f3d0"
+          background="oklch(0.696 0.17 162.48)"
+          className="text-sm font-medium"
+        >
+          Get Started
+        </ShimmerButton>
+      </div>
+    </div>
+  );
+}
+
+function InstallingContent({
+  currentStepIndex,
+  progressDetail,
+}: {
+  currentStepIndex: number;
+  progressDetail: string | null;
+}) {
+  const stepNum = Math.max(0, currentStepIndex) + 1;
+  const stepLabel = INSTALL_STEPS[Math.min(currentStepIndex, INSTALL_STEPS.length - 1)];
+
+  return (
+    <div className="flex h-full flex-col items-center">
+      <div className="flex flex-1 items-center">
+        <Logo />
+      </div>
+      <div className="mb-10 flex flex-col items-center gap-2">
+        <TealSpinner />
+        <p className="text-sm font-bold text-foreground">Installing...</p>
+        <p className="text-sm font-bold text-muted-foreground">
+          Please wait a few mins, then you can start training.
+        </p>
+        {currentStepIndex >= 0 && (
+          <p className="mt-1 text-xs font-bold text-muted-foreground">
+            Step {stepNum} of {INSTALL_STEPS.length}: {stepLabel}
+          </p>
+        )}
+        {progressDetail && (
+          <p className="text-xs text-muted-foreground/70">{progressDetail}</p>
+        )}
+      </div>
+    </div>
+  );
+}
+
+function RepairingContent({
+  logs,
+  progressDetail,
+}: {
+  logs: string[];
+  progressDetail: string | null;
+}) {
+  const latest = progressDetail ?? logs.at(-1);
+
+  return (
+    <div className="flex h-full flex-col items-center">
+      <div className="flex flex-1 items-center">
+        <Logo />
+      </div>
+      <div className="mb-10 flex flex-col items-center gap-2">
+        <TealSpinner />
+        <p className="text-sm font-bold text-foreground">Updating existing Studio install...</p>
+        {latest && (
+          <p className="max-w-xs text-center text-xs text-muted-foreground">{latest}</p>
+        )}
+      </div>
+    </div>
+  );
+}
+
+function InstallErrorContent({
+  error,
+  logs,
+  onRetryInstall,
+}: {
+  error: string | null;
+  logs: string[];
+  onRetryInstall: () => void;
+}) {
+  return (
+    <>
+      <Logo />
+      <div className="mt-8 flex flex-col items-center gap-2">
+        <p className="text-sm font-medium text-destructive">Setup ran into a problem</p>
+        {error && (
+          <p className="max-w-xs text-center text-xs text-muted-foreground">{error}</p>
+        )}
+        <div className="mt-4 flex gap-3">
+          <ActionButton
+            variant="secondary"
+            onClick={() => void navigator.clipboard.writeText(logs.join("\n"))}
+          >
+            Copy Logs
+          </ActionButton>
+          <ActionButton onClick={onRetryInstall}>Try Again</ActionButton>
+        </div>
+      </div>
+    </>
+  );
+}
+
+function RepairErrorContent({
+  error,
+  logs,
+  onRetry,
+}: {
+  error: string | null;
+  logs: string[];
+  onRetry: () => void;
+}) {
+  return (
+    <>
+      <Logo />
+      <div className="mt-8 flex flex-col items-center gap-2">
+        <p className="text-sm font-medium text-destructive">Update failed</p>
+        {error && (
+          <p className="max-w-md text-center text-xs text-muted-foreground">{error}</p>
+        )}
+        <div className="mt-4 flex gap-3">
+          <ActionButton
+            variant="secondary"
+            onClick={() => void navigator.clipboard.writeText(logs.join("\n"))}
+          >
+            Copy Logs
+          </ActionButton>
+          <ActionButton onClick={onRetry}>Retry</ActionButton>
+        </div>
+      </div>
+    </>
+  );
+}
+
+function NeedsElevationContent({
+  elevationPackages,
+  onApproveElevation,
+  onRetryInstall,
+}: {
+  elevationPackages: string[];
+  onApproveElevation: () => void;
+  onRetryInstall: () => void;
+}) {
+  return (
+    <>
+      <Logo />
+      <div className="mt-8 flex flex-col items-center gap-2">
+        <p className="text-sm font-medium text-foreground">Permission needed</p>
+        <p className="text-xs text-muted-foreground">
+          The following system packages need to be installed:
+        </p>
+        <div className="mt-2 w-full max-w-xs rounded-lg bg-muted p-3 font-mono text-xs">
+          {elevationPackages.map((pkg) => (
+            <div key={pkg}>{pkg}</div>
+          ))}
+        </div>
+        <div className="mt-4 flex gap-3">
+          <ActionButton variant="secondary" onClick={onRetryInstall}>Cancel</ActionButton>
+          <ActionButton onClick={onApproveElevation}>Allow</ActionButton>
+        </div>
+      </div>
+    </>
+  );
+}
+
+function StartingContent() {
+  return (
+    <div className="flex h-full flex-col items-center">
+      <div className="flex flex-1 items-center">
+        <Logo />
+      </div>
+      <div className="mb-10 flex flex-col items-center gap-2">
+        <TealSpinner />
+        <p className="text-sm text-muted-foreground">Starting server...</p>
+      </div>
+    </div>
+  );
+}
+
+function StoppedContent({ onStartServer }: { onStartServer: () => void }) {
+  return (
+    <>
+      <Logo />
+      <div className="mt-8 flex flex-col items-center gap-2">
+        <p className="text-sm font-medium text-foreground">Server stopped</p>
+        <div className="mt-4">
+          <ActionButton onClick={onStartServer}>Start Server</ActionButton>
+        </div>
+      </div>
+    </>
+  );
+}
+
+function ErrorContent({
+  error,
+  logs,
+  onRetry,
+}: {
+  error: string | null;
+  logs: string[];
+  onRetry: () => void;
+}) {
+  return (
+    <>
+      <Logo />
+      <div className="mt-8 flex flex-col items-center gap-2">
+        <p className="text-sm font-medium text-destructive">Something went wrong</p>
+        {error && (
+          <p className="max-w-md text-center text-xs text-muted-foreground">{error}</p>
+        )}
+        <div className="mt-4 flex gap-3">
+          <ActionButton
+            variant="secondary"
+            onClick={() => void navigator.clipboard.writeText(logs.join("\n"))}
+          >
+            Copy Logs
+          </ActionButton>
+          <ActionButton onClick={onRetry}>Retry</ActionButton>
+        </div>
+      </div>
+    </>
+  );
+}
+
+// ---------------------------------------------------------------------------
+// Main component
+// ---------------------------------------------------------------------------
+
+export function StartupScreen({
+  status,
+  logs,
+  error,
+  currentStepIndex,
+  progressDetail,
+  elevationPackages,
+  onInstall,
+  onRetry,
+  onRetryInstall,
+  onApproveElevation,
+  onStartServer,
+}: StartupScreenProps) {
+  function renderContent() {
+    switch (status) {
+      case "checking":
+        return <CheckingContent />;
+      case "not-installed":
+        return <NotInstalledContent onInstall={onInstall} />;
+      case "installing":
+        return <InstallingContent currentStepIndex={currentStepIndex} progressDetail={progressDetail} />;
+      case "install-error":
+        return <InstallErrorContent error={error} logs={logs} onRetryInstall={onRetryInstall} />;
+      case "repairing":
+        return <RepairingContent logs={logs} progressDetail={progressDetail} />;
+      case "repair-error":
+        return <RepairErrorContent error={error} logs={logs} onRetry={onRetry} />;
+      case "needs-elevation":
+        return (
+          <NeedsElevationContent
+            elevationPackages={elevationPackages}
+            onApproveElevation={onApproveElevation}
+            onRetryInstall={onRetryInstall}
+          />
+        );
+      case "starting":
+        return <StartingContent />;
+      case "running":
+        return null;
+      case "stopped":
+        return <StoppedContent onStartServer={onStartServer} />;
+      case "error":
+        return <ErrorContent error={error} logs={logs} onRetry={onRetry} />;
+    }
+  }
+
+  return (
+    <div className="flex h-screen w-full flex-col items-center bg-background">
+      <div className="flex flex-1 w-full max-w-md items-center justify-center px-6">
+        <AnimatePresence mode="wait">
+          <motion.div
+            key={status}
+            className="flex h-full w-full flex-col items-center text-center"
+            initial={{ opacity: 0 }}
+            animate={{ opacity: 1 }}
+            exit={{ opacity: 0 }}
+            transition={{ duration: 0.2, ease: EASE_OUT_QUART }}
+          >
+            {renderContent()}
+          </motion.div>
+        </AnimatePresence>
+      </div>
+    </div>
+  );
+}
diff --git a/studio/frontend/src/components/tauri/update-banner.tsx b/studio/frontend/src/components/tauri/update-banner.tsx
new file mode 100644
index 0000000000..98652c4db1
--- /dev/null
+++ b/studio/frontend/src/components/tauri/update-banner.tsx
@@ -0,0 +1,84 @@
+// SPDX-License-Identifier: AGPL-3.0-only
+// Copyright 2026-present the Unsloth AI Inc. team. All rights reserved. See /studio/LICENSE.AGPL-3.0
+
+import { Button } from "@/components/ui/button";
+import type { UpdateInfo, UpdateStatus } from "@/hooks/use-tauri-update";
+import { AnimatePresence, motion } from "motion/react";
+
+interface UpdateBannerProps {
+  status: UpdateStatus;
+  info: UpdateInfo | null;
+  dismissed: boolean;
+  isExternalServer?: boolean;
+  onInstall: () => void;
+  onDismiss: () => void;
+}
+
+const EASE_OUT_QUART: [number, number, number, number] = [0.165, 0.84, 0.44, 1];
+
+export function UpdateBanner({
+  status,
+  info,
+  dismissed,
+  isExternalServer = false,
+  onInstall,
+  onDismiss,
+}: UpdateBannerProps) {
+  const visible = status === "available";
+  const show = visible && !dismissed;
+
+  return (
+    <AnimatePresence>
+      {show && info && (
+        <motion.div
+          initial={{ opacity: 0, y: -12, scale: 0.96 }}
+          animate={{ opacity: 1, y: 0, scale: 1 }}
+          exit={{ opacity: 0, y: -8, scale: 0.97 }}
+          transition={{ duration: 0.35, ease: EASE_OUT_QUART }}
+          className="fixed top-4 right-4 z-[9999] w-[380px]"
+        >
+          <div className="corner-squircle relative overflow-hidden border border-border/60 bg-background/95 px-5 py-4 shadow-lg backdrop-blur-md">
+            {/* Close button */}
+            <button
+              type="button"
+              onClick={onDismiss}
+              className="absolute top-3 right-3 flex size-6 items-center justify-center rounded-md text-muted-foreground/60 transition-colors hover:bg-muted hover:text-foreground"
+            >
+              <svg width="14" height="14" viewBox="0 0 14 14" fill="none" xmlns="http://www.w3.org/2000/svg">
+                <path d="M11 3L3 11M3 3l8 8" stroke="currentColor" strokeWidth="1.5" strokeLinecap="round" />
+              </svg>
+            </button>
+
+            {/* Header */}
+            <div className="flex items-center gap-2">
+              <span className="text-lg">🦥</span>
+              <div>
+                <p className="text-sm font-semibold text-foreground">
+                  New version: v{info.version}
+                </p>
+                <p className="text-xs text-muted-foreground">
+                  {isExternalServer
+                    ? "Run `unsloth studio update` from your terminal"
+                    : "A new app update is available"}
+                </p>
+              </div>
+            </div>
+
+            {/* Actions */}
+            <div className="mt-3 flex items-center gap-2">
+              <Button size="sm" className="corner-squircle" onClick={onInstall} disabled={isExternalServer}>
+                Update Now
+              </Button>
+              <Button size="sm" variant="outline" className="corner-squircle" disabled>
+                Release Notes
+              </Button>
+              <Button size="sm" variant="ghost" className="corner-squircle" onClick={onDismiss}>
+                Later
+              </Button>
+            </div>
+          </div>
+        </motion.div>
+      )}
+    </AnimatePresence>
+  );
+}
diff --git a/studio/frontend/src/components/tauri/update-screen.tsx b/studio/frontend/src/components/tauri/update-screen.tsx
new file mode 100644
index 0000000000..a8b77ab1fa
--- /dev/null
+++ b/studio/frontend/src/components/tauri/update-screen.tsx
@@ -0,0 +1,173 @@
+// SPDX-License-Identifier: AGPL-3.0-only
+// Copyright 2026-present the Unsloth AI Inc. team. All rights reserved. See /studio/LICENSE.AGPL-3.0
+
+import type { UpdateStatus } from "@/hooks/use-tauri-update";
+import { AnimatePresence, motion } from "motion/react";
+import { useEffect, useRef } from "react";
+
+interface UpdateScreenProps {
+  status: UpdateStatus;
+  logs: string[];
+  progress: number;
+  error: string | null;
+  onRetry: () => void;
+  onSkipRestart: () => void;
+}
+
+const EASE_OUT_QUART: [number, number, number, number] = [0.165, 0.84, 0.44, 1];
+
+function Spinner({ size = 24 }: { size?: number }) {
+  return (
+    <span
+      className="inline-block animate-spin rounded-full border-2 border-primary border-t-transparent"
+      style={{ width: size, height: size, animationDuration: "0.8s" }}
+    />
+  );
+}
+
+function Logo() {
+  return (
+    <div className="flex flex-col items-center gap-4">
+      <img src="/sticker.png" alt="Unsloth" className="h-[72px] w-[72px] object-contain" />
+      <img src="/studio.png" alt="Unsloth Studio" className="h-auto w-[250px] object-contain dark:invert" />
+    </div>
+  );
+}
+
+function statusLabel(status: UpdateStatus): string {
+  switch (status) {
+    case "updating-backend":
+      return "Updating backend...";
+    case "downloading":
+      return "Downloading app update...";
+    case "installing":
+      return "Installing update...";
+    case "error":
+      return "Update failed";
+    default:
+      return "Updating...";
+  }
+}
+
+function statusSubtext(status: UpdateStatus, progress: number): string {
+  switch (status) {
+    case "updating-backend":
+      return "This may take a few minutes. Do not close the app.";
+    case "downloading":
+      return `${progress}% downloaded`;
+    case "installing":
+      return "The app will restart shortly.";
+    case "error":
+      return "Something went wrong during the update.";
+    default:
+      return "";
+  }
+}
+
+function LogViewer({ logs }: { logs: string[] }) {
+  const scrollRef = useRef<HTMLDivElement>(null);
+
+  useEffect(() => {
+    if (scrollRef.current) {
+      scrollRef.current.scrollTop = scrollRef.current.scrollHeight;
+    }
+  }, [logs]);
+
+  if (logs.length === 0) return null;
+
+  return (
+    <div
+      ref={scrollRef}
+      className="mt-4 h-[180px] w-full max-w-xl overflow-y-auto rounded-lg border border-border/40 bg-muted/30 p-3 font-mono text-[11px] leading-relaxed text-muted-foreground"
+    >
+      {logs.map((line, i) => (
+        <div key={i} className="whitespace-pre-wrap break-all">
+          {line}
+        </div>
+      ))}
+    </div>
+  );
+}
+
+export function UpdateScreen({
+  status,
+  logs,
+  progress,
+  error,
+  onRetry,
+  onSkipRestart,
+}: UpdateScreenProps) {
+  const isError = status === "error";
+
+  return (
+    <div className="flex h-screen w-full items-center justify-center bg-background">
+      <motion.div
+        initial={{ opacity: 0, y: 8 }}
+        animate={{ opacity: 1, y: 0 }}
+        transition={{ duration: 0.4, ease: EASE_OUT_QUART }}
+        className="flex w-full max-w-xl flex-col items-center px-6"
+      >
+        <Logo />
+
+        <div className="mt-8 flex flex-col items-center gap-2">
+          {!isError && <Spinner />}
+          <p className="text-sm font-semibold text-foreground">
+            {statusLabel(status)}
+          </p>
+          <p className="text-xs text-muted-foreground">
+            {statusSubtext(status, progress)}
+          </p>
+        </div>
+
+        {/* Download progress bar */}
+        {status === "downloading" && (
+          <div className="mt-4 h-1.5 w-full max-w-xs overflow-hidden rounded-full bg-muted">
+            <motion.div
+              className="h-full rounded-full bg-primary"
+              initial={{ width: 0 }}
+              animate={{ width: `${progress}%` }}
+              transition={{ duration: 0.3 }}
+            />
+          </div>
+        )}
+
+        {/* Error display */}
+        <AnimatePresence>
+          {isError && error && (
+            <motion.div
+              initial={{ opacity: 0, height: 0 }}
+              animate={{ opacity: 1, height: "auto" }}
+              exit={{ opacity: 0, height: 0 }}
+              className="mt-4 w-full max-w-xl rounded-lg border border-destructive/30 bg-destructive/5 px-4 py-3"
+            >
+              <p className="text-xs text-destructive">{error}</p>
+            </motion.div>
+          )}
+        </AnimatePresence>
+
+        {/* Error actions */}
+        {isError && (
+          <div className="mt-4 flex items-center gap-2">
+            <button
+              type="button"
+              className="rounded-lg bg-primary px-5 py-2.5 text-sm font-medium text-primary-foreground transition-colors hover:bg-primary/80"
+              onClick={onRetry}
+            >
+              Retry
+            </button>
+            <button
+              type="button"
+              className="rounded-lg bg-muted px-5 py-2.5 text-sm font-medium text-foreground transition-colors hover:bg-muted/80"
+              onClick={onSkipRestart}
+            >
+              Skip & Restart
+            </button>
+          </div>
+        )}
+
+        {/* Log viewer */}
+        <LogViewer logs={logs} />
+      </motion.div>
+    </div>
+  );
+}
diff --git a/studio/frontend/src/components/ui/shimmer-button.tsx b/studio/frontend/src/components/ui/shimmer-button.tsx
new file mode 100644
index 0000000000..d675cc0979
--- /dev/null
+++ b/studio/frontend/src/components/ui/shimmer-button.tsx
@@ -0,0 +1,96 @@
+import React, { type ComponentPropsWithoutRef, type CSSProperties } from "react"
+
+import { cn } from "@/lib/utils"
+
+export interface ShimmerButtonProps extends ComponentPropsWithoutRef<"button"> {
+  shimmerColor?: string
+  shimmerSize?: string
+  borderRadius?: string
+  shimmerDuration?: string
+  background?: string
+  className?: string
+  children?: React.ReactNode
+}
+
+export const ShimmerButton = React.forwardRef<
+  HTMLButtonElement,
+  ShimmerButtonProps
+>(
+  (
+    {
+      shimmerColor = "#ffffff",
+      shimmerSize = "0.05em",
+      shimmerDuration = "3s",
+      borderRadius = "100px",
+      background = "rgba(0, 0, 0, 1)",
+      className,
+      children,
+      ...props
+    },
+    ref
+  ) => {
+    return (
+      <button
+        style={
+          {
+            "--spread": "90deg",
+            "--shimmer-color": shimmerColor,
+            "--radius": borderRadius,
+            "--speed": shimmerDuration,
+            "--cut": shimmerSize,
+            "--bg": background,
+          } as CSSProperties
+        }
+        className={cn(
+          "group relative z-0 flex cursor-pointer items-center justify-center overflow-hidden [border-radius:var(--radius)] border border-white/10 px-6 py-3 whitespace-nowrap text-white [background:var(--bg)]",
+          "transform-gpu transition-transform duration-300 ease-in-out active:translate-y-px",
+          className
+        )}
+        ref={ref}
+        {...props}
+      >
+        {/* spark container */}
+        <div
+          className={cn(
+            "-z-30 blur-[2px]",
+            "@container-[size] absolute inset-0 overflow-visible"
+          )}
+        >
+          {/* spark */}
+          <div className="animate-shimmer-slide absolute inset-0 aspect-[1] h-[100cqh] rounded-none [mask:none]">
+            {/* spark before */}
+            <div className="animate-spin-around absolute -inset-full w-auto [translate:0_0] rotate-0 [background:conic-gradient(from_calc(270deg-(var(--spread)*0.5)),transparent_0,var(--shimmer-color)_var(--spread),transparent_var(--spread))]" />
+          </div>
+        </div>
+        {children}
+
+        {/* Highlight */}
+        <div
+          className={cn(
+            "absolute inset-0 size-full",
+
+            "rounded-2xl px-4 py-1.5 text-sm font-medium shadow-[inset_0_-8px_10px_#ffffff1f]",
+
+            // transition
+            "transform-gpu transition-all duration-300 ease-in-out",
+
+            // on hover
+            "group-hover:shadow-[inset_0_-6px_10px_#ffffff3f]",
+
+            // on click
+            "group-active:shadow-[inset_0_-10px_10px_#ffffff3f]"
+          )}
+        />
+
+        {/* backdrop */}
+        <div
+          className={cn(
+            "absolute inset-(--cut) -z-20 [border-radius:var(--radius)] [background:var(--bg)]"
+          )}
+        />
+      </button>
+    )
+  }
+)
+
+ShimmerButton.displayName = "ShimmerButton"
diff --git a/studio/frontend/src/config/env.ts b/studio/frontend/src/config/env.ts
index 91e17f6bb9..72bb3fa815 100644
--- a/studio/frontend/src/config/env.ts
+++ b/studio/frontend/src/config/env.ts
@@ -1,6 +1,7 @@
 // SPDX-License-Identifier: AGPL-3.0-only
 // Copyright 2026-present the Unsloth AI Inc. team. All rights reserved. See /studio/LICENSE.AGPL-3.0
 
+import { apiUrl } from "@/lib/api-base";
 import { create } from "zustand";
 
 export const env = {
@@ -21,9 +22,21 @@ interface PlatformState {
   isChatOnly: () => boolean;
 }
 
+// Client-side platform detection as fallback when backend isn't ready yet.
+function detectLocalPlatform(): DeviceType {
+  if (typeof navigator === "undefined") return "linux";
+  const platform = navigator.platform.toLowerCase();
+  const ua = navigator.userAgent.toLowerCase();
+  if (platform.includes("mac") || ua.includes("mac")) return "mac";
+  if (platform.includes("win") || ua.includes("win")) return "windows";
+  return "linux";
+}
+
+const localDeviceType = detectLocalPlatform();
+
 export const usePlatformStore = create<PlatformState>()((_, get) => ({
-  deviceType: "linux",
-  chatOnly: false,
+  deviceType: localDeviceType,
+  chatOnly: localDeviceType === "mac",
   fetched: false,
   isChatOnly: () => get().chatOnly,
 }));
@@ -33,16 +46,22 @@ export async function fetchDeviceType(): Promise<DeviceType> {
   if (fetched) return usePlatformStore.getState().deviceType;
 
   try {
-    const res = await fetch("/api/health");
+    const res = await fetch(apiUrl("/api/health"));
     if (res.ok) {
       const data = (await res.json()) as { device_type?: string; chat_only?: boolean };
-      const deviceType = data.device_type ?? "linux";
+      const deviceType = data.device_type ?? detectLocalPlatform();
       const chatOnly = data.chat_only ?? deviceType === "mac";
       usePlatformStore.setState({ deviceType, chatOnly, fetched: true });
       return deviceType;
     }
-  } catch (err) {
-    console.warn("[platform] Failed to fetch device type, will retry", err);
+  } catch {
+    // Backend not ready — use client-side detection so chat-only guard
+    // still works on initial load (important for macOS). Keep fetched=false
+    // so a later call retries against the backend.
+    const deviceType = detectLocalPlatform();
+    const chatOnly = deviceType === "mac";
+    usePlatformStore.setState({ deviceType, chatOnly, fetched: false });
+    return deviceType;
   }
 
   return usePlatformStore.getState().deviceType;
diff --git a/studio/frontend/src/features/auth/api.ts b/studio/frontend/src/features/auth/api.ts
index 3bb2c9139c..97d815950d 100644
--- a/studio/frontend/src/features/auth/api.ts
+++ b/studio/frontend/src/features/auth/api.ts
@@ -1,6 +1,7 @@
 // SPDX-License-Identifier: AGPL-3.0-only
 // Copyright 2026-present the Unsloth AI Inc. team. All rights reserved. See /studio/LICENSE.AGPL-3.0
 
+import { apiUrl, isTauri } from "@/lib/api-base";
 import {
   clearAuthTokens,
   getAuthToken,
@@ -34,7 +35,7 @@ async function redirectToAuth(): Promise<void> {
 
   let target = "/login";
   try {
-    const res = await fetch("/api/auth/status");
+    const res = await fetch(apiUrl("/api/auth/status"));
     if (res.ok) {
       const data = (await res.json()) as { requires_password_change: boolean };
       if (data.requires_password_change || mustChangePassword()) target = "/change-password";
@@ -46,12 +47,32 @@ async function redirectToAuth(): Promise<void> {
   window.location.href = target;
 }
 
+async function retryWithCurrentToken(
+  input: RequestInfo | URL,
+  init?: RequestInit,
+): Promise<Response> {
+  const retryHeaders = new Headers(init?.headers);
+  const token = getAuthToken();
+  if (token) retryHeaders.set("Authorization", `Bearer ${token}`);
+  return fetch(input, { ...init, headers: retryHeaders });
+}
+
+async function retryWithTauriAutoAuth(
+  input: RequestInfo | URL,
+  init?: RequestInit,
+): Promise<Response | null> {
+  clearAuthTokens();
+  const { tauriAutoAuth } = await import("./tauri-auto-auth");
+  if (await tauriAutoAuth()) return retryWithCurrentToken(input, init);
+  return null;
+}
+
 export async function refreshSession(): Promise<boolean> {
   const refreshToken = getRefreshToken();
   if (!refreshToken) return false;
 
   try {
-    const response = await fetch("/api/auth/refresh", {
+    const response = await fetch(apiUrl("/api/auth/refresh"), {
       method: "POST",
       headers: { "Content-Type": "application/json" },
       body: JSON.stringify({ refresh_token: refreshToken }),
@@ -78,6 +99,7 @@ export async function authFetch(
   input: RequestInfo | URL,
   init?: RequestInit,
 ): Promise<Response> {
+  const resolvedInput = typeof input === 'string' ? apiUrl(input) : input;
   const headers = new Headers(init?.headers);
   const accessToken = getAuthToken();
   if (accessToken) {
@@ -86,14 +108,18 @@ export async function authFetch(
 
   let response: Response;
   try {
-    response = await fetch(input, { ...init, headers });
+    response = await fetch(resolvedInput, { ...init, headers });
   } catch (err) {
     if (err instanceof TypeError) {
       throw new Error("Studio isn't running -- please relaunch it.");
     }
     throw err;
   }
+
   if (await isPasswordChangeRequiredResponse(response)) {
+    if (isTauri) {
+      return (await retryWithTauriAutoAuth(resolvedInput, init)) ?? response;
+    }
     void redirectToAuth();
     return response;
   }
@@ -101,25 +127,24 @@ export async function authFetch(
 
   const refreshed = await refreshSession();
   if (!refreshed) {
+    if (isTauri) {
+      return (await retryWithTauriAutoAuth(resolvedInput, init)) ?? response;
+    }
     clearAuthTokens();
     void redirectToAuth();
     return response;
   }
 
   if (mustChangePassword()) {
+    if (isTauri) {
+      return (await retryWithTauriAutoAuth(resolvedInput, init)) ?? response;
+    }
     void redirectToAuth();
     return response;
   }
 
-  const retryHeaders = new Headers(init?.headers);
-  const newToken = getAuthToken();
-  if (newToken) {
-    retryHeaders.set("Authorization", `Bearer ${newToken}`);
-  } else {
-    clearAuthTokens();
-  }
-
-  return fetch(input, { ...init, headers: retryHeaders });
+  if (!getAuthToken()) clearAuthTokens();
+  return retryWithCurrentToken(resolvedInput, init);
 }
 
 export function logout(): void {
diff --git a/studio/frontend/src/features/auth/components/auth-form.tsx b/studio/frontend/src/features/auth/components/auth-form.tsx
index d9190429bd..090a3081a4 100644
--- a/studio/frontend/src/features/auth/components/auth-form.tsx
+++ b/studio/frontend/src/features/auth/components/auth-form.tsx
@@ -1,6 +1,7 @@
 // SPDX-License-Identifier: AGPL-3.0-only
 // Copyright 2026-present the Unsloth AI Inc. team. All rights reserved. See /studio/LICENSE.AGPL-3.0
 
+import { apiUrl } from "@/lib/api-base";
 import { Button } from "@/components/ui/button";
 import { Input } from "@/components/ui/input";
 import { Label } from "@/components/ui/label";
@@ -49,7 +50,7 @@ async function loginWithPassword(
   username: string,
   password: string,
 ): Promise<TokenResponse> {
-  const response = await fetch("/api/auth/login", {
+  const response = await fetch(apiUrl("/api/auth/login"), {
     method: "POST",
     headers: {
       "Content-Type": "application/json",
@@ -96,7 +97,7 @@ export function AuthForm({ mode }: AuthFormProps): ReactElement | null {
       // (e.g. tokens from a previous install attempt).  The server's
       // /api/auth/status is the source of truth for requires_password_change.
       try {
-        const response = await fetch("/api/auth/status");
+        const response = await fetch(apiUrl("/api/auth/status"));
         if (!response.ok) throw new Error("Failed to load auth status.");
         const result = (await response.json()) as AuthStatusResponse;
         if (!canceled) {
@@ -147,14 +148,15 @@ export function AuthForm({ mode }: AuthFormProps): ReactElement | null {
     };
   }, [navigate]);
 
-  // Seed password from bootstrap credentials injected into HTML
+  // Seed password from bootstrap credentials injected into HTML by web CLI.
   useEffect(() => {
-    const bootstrap = window.__UNSLOTH_BOOTSTRAP__;
-    if (bootstrap) {
-      if (!isLoginMode && !password) {
+    function loadBootstrap() {
+      const bootstrap = window.__UNSLOTH_BOOTSTRAP__;
+      if (bootstrap && !isLoginMode && !password) {
         setPassword(bootstrap.password);
       }
     }
+    loadBootstrap();
   }, []);
 
   const blockedByState =
@@ -241,7 +243,7 @@ export function AuthForm({ mode }: AuthFormProps): ReactElement | null {
           accessToken = bootstrapToken.access_token;
         }
 
-        const response = await fetch("/api/auth/change-password", {
+        const response = await fetch(apiUrl("/api/auth/change-password"), {
           method: "POST",
           headers: {
             "Content-Type": "application/json",
diff --git a/studio/frontend/src/features/auth/index.ts b/studio/frontend/src/features/auth/index.ts
index af629cfb9a..9cc1599195 100644
--- a/studio/frontend/src/features/auth/index.ts
+++ b/studio/frontend/src/features/auth/index.ts
@@ -15,3 +15,8 @@ export {
   resetOnboardingDone,
   setMustChangePassword,
 } from "./session";
+export {
+  clearTauriAuthFailure,
+  getTauriAuthFailure,
+  tauriAutoAuth,
+} from "./tauri-auto-auth";
diff --git a/studio/frontend/src/features/auth/session.ts b/studio/frontend/src/features/auth/session.ts
index 6012174077..49a2722bdf 100644
--- a/studio/frontend/src/features/auth/session.ts
+++ b/studio/frontend/src/features/auth/session.ts
@@ -2,6 +2,7 @@
 // Copyright 2026-present the Unsloth AI Inc. team. All rights reserved. See /studio/LICENSE.AGPL-3.0
 
 import { usePlatformStore } from "@/config/env";
+import { isTauri } from "@/lib/api-base";
 
 export const AUTH_TOKEN_KEY = "unsloth_auth_token";
 export const AUTH_REFRESH_TOKEN_KEY = "unsloth_auth_refresh_token";
@@ -78,6 +79,7 @@ export function resetOnboardingDone(): void {
 }
 
 export function getPostAuthRoute(): PostAuthRoute {
+  if (isTauri) return "/chat";
   if (mustChangePassword()) return "/change-password";
   if (usePlatformStore.getState().isChatOnly()) return "/chat";
   return "/chat";
diff --git a/studio/frontend/src/features/auth/tauri-auto-auth.ts b/studio/frontend/src/features/auth/tauri-auto-auth.ts
new file mode 100644
index 0000000000..d67730d2d6
--- /dev/null
+++ b/studio/frontend/src/features/auth/tauri-auto-auth.ts
@@ -0,0 +1,95 @@
+// SPDX-License-Identifier: AGPL-3.0-only
+// Copyright 2026-present the Unsloth AI Inc. team. All rights reserved. See /studio/LICENSE.AGPL-3.0
+
+import { isTauri } from "@/lib/api-base";
+import {
+  hasAuthToken,
+  hasRefreshToken,
+  mustChangePassword,
+  storeAuthTokens,
+} from "./session";
+import { refreshSession } from "./api";
+
+type DesktopAuthResponse = {
+  access_token: string;
+  refresh_token: string;
+};
+
+// Concurrency guard: multiple route guards can call tauriAutoAuth simultaneously.
+// Without this, the first-launch password-change could race with itself.
+let pending: Promise<boolean> | null = null;
+let lastTauriAuthFailure: string | null = null;
+
+const TAURI_AUTH_FAILURE_FALLBACK =
+  "Desktop authentication failed. Update or repair the managed Studio install, then restart Studio.";
+const BACKEND_NOT_READY_MESSAGE = "Backend is not ready";
+
+function authFailureMessage(error: unknown): string {
+  if (typeof error === "string" && error) return error;
+  if (error instanceof Error && error.message) return error.message;
+  return TAURI_AUTH_FAILURE_FALLBACK;
+}
+
+export function getTauriAuthFailure(): string | null {
+  return lastTauriAuthFailure;
+}
+
+export function clearTauriAuthFailure(): void {
+  lastTauriAuthFailure = null;
+}
+
+function setTauriAuthFailure(error: unknown): void {
+  lastTauriAuthFailure = authFailureMessage(error);
+  window.dispatchEvent(
+    new CustomEvent("tauri-auth-failed", { detail: lastTauriAuthFailure }),
+  );
+}
+
+function isBackendNotReady(error: unknown): boolean {
+  return authFailureMessage(error).includes(BACKEND_NOT_READY_MESSAGE);
+}
+
+async function doTauriAutoAuth(): Promise<boolean> {
+  // Desktop must handle password-change state internally in Rust.
+  if (hasAuthToken() && !mustChangePassword()) {
+    clearTauriAuthFailure();
+    return true;
+  }
+
+  // Try refreshing existing session
+  if (hasRefreshToken()) {
+    const refreshed = await refreshSession();
+    if (refreshed && hasAuthToken() && !mustChangePassword()) {
+      clearTauriAuthFailure();
+      return true;
+    }
+  }
+
+  try {
+    const { invoke } = await import("@tauri-apps/api/core");
+    const tokens = await invoke<DesktopAuthResponse>("desktop_auth");
+    storeAuthTokens(tokens.access_token, tokens.refresh_token, false);
+    clearTauriAuthFailure();
+    return true;
+  } catch (error) {
+    if (isBackendNotReady(error)) return false;
+    setTauriAuthFailure(error);
+    return false;
+  }
+}
+
+/**
+ * Silently authenticate in Tauri desktop mode.
+ *
+ * Delegates bootstrap/password handling to Rust and only stores returned tokens.
+ *
+ * Returns true if authentication succeeded.
+ * Concurrent calls are coalesced into a single in-flight attempt.
+ */
+export function tauriAutoAuth(): Promise<boolean> {
+  if (!isTauri) return Promise.resolve(false);
+  if (!pending) {
+    pending = doTauriAutoAuth().finally(() => { pending = null; });
+  }
+  return pending;
+}
diff --git a/studio/frontend/src/features/settings/tabs/about-tab.tsx b/studio/frontend/src/features/settings/tabs/about-tab.tsx
index 417922139c..bb4454aa43 100644
--- a/studio/frontend/src/features/settings/tabs/about-tab.tsx
+++ b/studio/frontend/src/features/settings/tabs/about-tab.tsx
@@ -5,6 +5,7 @@ import { Button } from "@/components/ui/button";
 import { ShutdownDialog } from "@/components/shutdown-dialog";
 import { UpdateStudioInstructions } from "../components/update-studio-instructions";
 import { usePlatformStore } from "@/config/env";
+import { apiUrl } from "@/lib/api-base";
 import { removeTrainingUnloadGuard } from "@/features/training/hooks/use-training-unload-guard";
 import {
   ArrowUpRight01Icon,
@@ -28,7 +29,7 @@ export function AboutTab() {
 
     (async () => {
       try {
-        const res = await fetch("/api/health");
+        const res = await fetch(apiUrl("/api/health"));
         if (!res.ok) return;
         const data = (await res.json()) as { version?: string };
         if (!canceled && data.version) {
diff --git a/studio/frontend/src/hooks/index.ts b/studio/frontend/src/hooks/index.ts
index f889dc7445..cfbd8e5115 100644
--- a/studio/frontend/src/hooks/index.ts
+++ b/studio/frontend/src/hooks/index.ts
@@ -11,4 +11,5 @@ export { useHfDatasetSearch } from "./use-hf-dataset-search";
 export { useHfDatasetSplits } from "./use-hf-dataset-splits";
 export { useHfTokenValidation } from "./use-hf-token-validation";
 export { useInfiniteScroll } from "./use-infinite-scroll";
+export { useTauriBackend } from "./use-tauri-backend";
 export { useCollapseScrollLock } from "./use-collapse-scroll-lock";
diff --git a/studio/frontend/src/hooks/use-gpu-info.ts b/studio/frontend/src/hooks/use-gpu-info.ts
index 3eb81c35e9..64caf06b50 100644
--- a/studio/frontend/src/hooks/use-gpu-info.ts
+++ b/studio/frontend/src/hooks/use-gpu-info.ts
@@ -1,6 +1,7 @@
 // SPDX-License-Identifier: AGPL-3.0-only
 // Copyright 2026-present the Unsloth AI Inc. team. All rights reserved. See /studio/LICENSE.AGPL-3.0
 
+import { apiUrl } from "@/lib/api-base";
 import { useEffect, useState } from "react";
 
 export interface GpuInfo {
@@ -27,7 +28,7 @@ async function fetchGpuOnce(): Promise<GpuInfo> {
 
   fetchPromise = (async () => {
     try {
-      const res = await fetch("/api/system");
+      const res = await fetch(apiUrl("/api/system"));
       if (!res.ok) throw new Error(`HTTP ${res.status}`);
       const data = await res.json();
       const gpuData = data?.gpu;
diff --git a/studio/frontend/src/hooks/use-tauri-backend.ts b/studio/frontend/src/hooks/use-tauri-backend.ts
new file mode 100644
index 0000000000..ed697fe31e
--- /dev/null
+++ b/studio/frontend/src/hooks/use-tauri-backend.ts
@@ -0,0 +1,487 @@
+// SPDX-License-Identifier: AGPL-3.0-only
+// Copyright 2026-present the Unsloth AI Inc. team. All rights reserved. See /studio/LICENSE.AGPL-3.0
+
+import { useEffect, useState, useCallback, useRef } from "react";
+import { isTauri, setApiBase } from "@/lib/api-base";
+import {
+  clearTauriAuthFailure,
+  getTauriAuthFailure,
+} from "@/features/auth";
+
+export type BackendStatus =
+  | "checking"
+  | "not-installed"
+  | "installing"
+  | "install-error"
+  | "needs-elevation"
+  | "repairing"
+  | "repair-error"
+  | "starting"
+  | "running"
+  | "stopped"
+  | "error";
+
+type DesktopPreflightDisposition =
+  | "not_installed"
+  | "managed_ready"
+  | "managed_stale"
+  | "attached_ready";
+
+interface DesktopPreflightResult {
+  disposition: DesktopPreflightDisposition;
+  reason: string | null;
+  port: number | null;
+  can_auto_repair: boolean;
+  managed_bin: string | null;
+}
+
+export function useTauriBackend() {
+  const [status, setStatus] = useState<BackendStatus>("checking");
+  const statusRef = useRef<BackendStatus>(status);
+  const [logs, setLogs] = useState<string[]>([]);
+  const [error, setError] = useState<string | null>(null);
+  // Guard against double startServer calls
+  const startingRef = useRef(false);
+  // Guard against React Strict Mode double-mount
+  const mountedRef = useRef(false);
+  // Track the discovered port from server-port event
+  const portRef = useRef<number | null>(null);
+  const [currentStepIndex, setCurrentStepIndex] = useState(-1);
+  const [elevationPackages, setElevationPackages] = useState<string[]>([]);
+  const [progressDetail, setProgressDetail] = useState<string | null>(null);
+  // Track seen step names to deduplicate (Strict Mode, event replay, etc.)
+  const seenStepsRef = useRef(new Set<string>());
+  // True when we attached to a server we didn't spawn (can't stop it)
+  const [isExternalServer, setIsExternalServer] = useState(false);
+  const externalPollRef = useRef<ReturnType<typeof setInterval> | null>(null);
+  const externalPollAbortedRef = useRef(false);
+  const authFailureRef = useRef<string | null>(getTauriAuthFailure());
+  const elevationResumeRef = useRef<"install" | "repair" | null>(null);
+
+  function setBackendStatus(nextStatus: BackendStatus) {
+    if (authFailureRef.current) return;
+    statusRef.current = nextStatus;
+    setStatus(nextStatus);
+  }
+
+  function setBackendError(
+    nextError: string,
+    nextStatus: BackendStatus = "error",
+  ) {
+    if (authFailureRef.current) return;
+    statusRef.current = nextStatus;
+    setStatus(nextStatus);
+    setError(nextError);
+  }
+
+  function clearBackendError() {
+    if (authFailureRef.current) return;
+    setError(null);
+  }
+
+  function setRunningStatus() {
+    setBackendStatus("running");
+  }
+
+  function setAuthFailure(detail: string) {
+    authFailureRef.current = detail;
+    statusRef.current = "error";
+    setStatus("error");
+    setError(detail);
+  }
+
+  function clearAuthFailure() {
+    authFailureRef.current = null;
+    clearTauriAuthFailure();
+  }
+
+  function stopExternalServerPoll() {
+    externalPollAbortedRef.current = true;
+    if (externalPollRef.current) {
+      clearInterval(externalPollRef.current);
+      externalPollRef.current = null;
+    }
+  }
+
+  function startExternalServerPoll(port: number) {
+    stopExternalServerPoll();
+    externalPollAbortedRef.current = false;
+    let failures = 0;
+    externalPollRef.current = setInterval(async () => {
+      if (externalPollAbortedRef.current) return;
+      try {
+        const { invoke } = await import("@tauri-apps/api/core");
+        const healthy = await invoke<boolean>("check_health", { port });
+        if (externalPollAbortedRef.current) return;
+        if (healthy) {
+          failures = 0;
+        } else {
+          failures++;
+        }
+      } catch {
+        if (externalPollAbortedRef.current) return;
+        failures++;
+      }
+      if (failures >= 3) {
+        stopExternalServerPoll();
+        setIsExternalServer(false);
+        setBackendError("External server is no longer responding");
+      }
+    }, 15_000);
+  }
+
+  // Keep ref in sync for event listener closures
+  useEffect(() => {
+    statusRef.current = status;
+  }, [status]);
+
+  async function checkInstallAndStart() {
+    try {
+      const { invoke } = await import("@tauri-apps/api/core");
+
+      const preflight = await invoke<DesktopPreflightResult>("desktop_preflight");
+      switch (preflight.disposition) {
+        case "attached_ready": {
+          if (!preflight.port) {
+            setBackendError("Desktop preflight found a backend without a port.");
+            return;
+          }
+          setApiBase(preflight.port);
+          portRef.current = preflight.port;
+          setIsExternalServer(true);
+          setRunningStatus();
+          startExternalServerPoll(preflight.port);
+          return;
+        }
+        case "managed_ready":
+          setIsExternalServer(false);
+          stopExternalServerPoll();
+          setBackendStatus("starting");
+          await startManagedServer();
+          return;
+        case "managed_stale":
+          setIsExternalServer(false);
+          stopExternalServerPoll();
+          if (preflight.can_auto_repair) {
+            await startRepair();
+          } else {
+            setBackendError(
+              "Managed Studio install is too old. Run `unsloth studio update`.",
+            );
+          }
+          return;
+        case "not_installed":
+          setBackendStatus("not-installed");
+          return;
+      }
+    } catch (e) {
+      setBackendError(String(e));
+    }
+  }
+
+  async function startManagedServer() {
+    // Prevent double-start race condition
+    if (startingRef.current) return;
+    startingRef.current = true;
+
+    try {
+      const { invoke } = await import("@tauri-apps/api/core");
+      // backend/run.py keeps the existing 8888-8908 fallback via
+      // server-port/TAURI_PORT.
+      await invoke("start_managed_server", { port: 8888 });
+
+      // Wait for the owned backend's server-port event. Do not attach to an
+      // external backend if the managed start does not report a port.
+      for (let i = 0; i < 120; i++) {
+        if (portRef.current) {
+          const healthy = await invoke<boolean>("check_health", {
+            port: portRef.current,
+          });
+          if (healthy) {
+            setApiBase(portRef.current);
+            setRunningStatus();
+            startingRef.current = false;
+            return;
+          }
+        }
+        await new Promise((r) => setTimeout(r, 500));
+      }
+      const message = !portRef.current
+        ? "Managed server started without reporting a port. Check the logs for details."
+        : "Server started but is not responding. Check the logs for details.";
+      setBackendError(message);
+    } catch (e) {
+      const msg = String(e);
+      if (msg.includes("already running")) {
+        startingRef.current = false;
+        setBackendError(
+          "Managed server is already running but did not report a port. Restart Studio and try again.",
+        );
+        return;
+      }
+      setBackendError(msg);
+    }
+    startingRef.current = false;
+  }
+
+  async function startRepair() {
+    elevationResumeRef.current = null;
+    setCurrentStepIndex(-1);
+    setProgressDetail(null);
+    seenStepsRef.current.clear();
+    startingRef.current = false;
+    portRef.current = null;
+    setIsExternalServer(false);
+    stopExternalServerPoll();
+    setLogs([]);
+    clearBackendError();
+    setBackendStatus("repairing");
+
+    const { invoke } = await import("@tauri-apps/api/core");
+    try {
+      await invoke("start_managed_repair");
+
+      setBackendStatus("starting");
+      elevationResumeRef.current = null;
+      await startManagedServer();
+    } catch (e) {
+      const msg = String(e);
+      if (msg.includes("NEEDS_ELEVATION")) return;
+      setBackendError(msg, "repair-error");
+    }
+  }
+
+  async function startServer() {
+    setBackendStatus("starting");
+    await startManagedServer();
+  }
+
+  async function stopServer() {
+    if (isExternalServer) {
+      // We attached to a server we didn't spawn — can't kill it,
+      // just disconnect the UI.
+      startingRef.current = false;
+      setIsExternalServer(false);
+      stopExternalServerPoll();
+      setBackendStatus("stopped");
+      return;
+    }
+    const { invoke } = await import("@tauri-apps/api/core");
+    await invoke("stop_server");
+    startingRef.current = false;
+    setBackendStatus("stopped");
+  }
+
+  async function startInstall() {
+    elevationResumeRef.current = null;
+    setCurrentStepIndex(-1);
+    setProgressDetail(null);
+    seenStepsRef.current.clear();
+    setBackendStatus("installing");
+    setLogs([]);
+    clearBackendError();
+    const { invoke } = await import("@tauri-apps/api/core");
+    try {
+      await invoke("start_install");
+      // Install completed — this is the ONLY path that starts the server
+      // after install. The install-complete event listener does NOT call
+      // startServer() to avoid a double-start race condition.
+      setBackendStatus("starting");
+      elevationResumeRef.current = null;
+      await startServer();
+    } catch (e) {
+      const msg = String(e);
+      // NEEDS_ELEVATION is not a real error — the Rust side also emits
+      // install-needs-elevation which sets needs-elevation status.
+      // Don't race with it by setting install-error here.
+      if (msg.includes("NEEDS_ELEVATION")) return;
+      setBackendError(msg, "install-error");
+    }
+  }
+
+  const retry = useCallback(() => {
+    clearAuthFailure();
+    setError(null);
+    setLogs([]);
+    startingRef.current = false;
+    portRef.current = null;
+    setCurrentStepIndex(-1);
+    setProgressDetail(null);
+    setElevationPackages([]);
+    elevationResumeRef.current = null;
+    setIsExternalServer(false);
+    stopExternalServerPoll();
+    seenStepsRef.current.clear();
+    checkInstallAndStart();
+  }, []);
+
+  const retryInstall = useCallback(() => {
+    const resume = elevationResumeRef.current;
+    elevationResumeRef.current = null;
+    clearBackendError();
+    setLogs([]);
+    setElevationPackages([]);
+    if (resume === "repair") {
+      setBackendError("Repair canceled before system packages were installed.", "repair-error");
+      return;
+    }
+    setBackendStatus("not-installed");
+  }, []);
+
+  const approveElevation = useCallback(async () => {
+    const resume = elevationResumeRef.current ?? "install";
+    try {
+      const { invoke } = await import("@tauri-apps/api/core");
+      await invoke("install_system_packages", { packages: elevationPackages });
+      // Packages installed successfully, resume the flow that requested them.
+      setCurrentStepIndex(-1);
+      setProgressDetail(null);
+      elevationResumeRef.current = null;
+      if (resume === "repair") {
+        await startRepair();
+      } else {
+        await startInstall();
+      }
+    } catch (e) {
+      setBackendError(String(e), resume === "repair" ? "repair-error" : "install-error");
+    }
+  }, [elevationPackages]);
+
+  // Initial check on mount (guarded against Strict Mode double-mount)
+  useEffect(() => {
+    if (mountedRef.current) return;
+    mountedRef.current = true;
+
+    if (!isTauri) {
+      setRunningStatus();
+      return;
+    }
+    checkInstallAndStart();
+  }, []);
+
+  // Listen for Tauri events
+  useEffect(() => {
+    if (!isTauri) return;
+    const cleanup: (() => void)[] = [];
+    let disposed = false;
+
+    import("@tauri-apps/api/event").then(({ listen }) => {
+      function register<T>(
+        event: string,
+        handler: Parameters<typeof listen<T>>[1],
+      ) {
+        listen<T>(event, handler).then((unlisten) => {
+          if (disposed) {
+            unlisten();
+          } else {
+            cleanup.push(unlisten);
+          }
+        });
+      }
+
+      register<string>("install-progress", (e) => {
+        setLogs((prev) => [...prev.slice(-499), e.payload]);
+      });
+
+      // install-complete is informational only — does NOT trigger startServer.
+      // The invoke("start_install") success path handles that to avoid races.
+      register<void>("install-complete", () => {
+        setCurrentStepIndex(999); // all steps done
+      });
+
+      register<string>("install-step", (e) => {
+        const stepName = e.payload;
+        if (seenStepsRef.current.has(stepName)) return; // deduplicate
+        seenStepsRef.current.add(stepName);
+        setCurrentStepIndex((prev) => prev + 1);
+        setProgressDetail(null);
+      });
+
+      register<string[]>("install-needs-elevation", (e) => {
+        elevationResumeRef.current = "install";
+        setElevationPackages(e.payload);
+        setBackendStatus("needs-elevation");
+      });
+
+      register<string>("install-progress-detail", (e) => {
+        setProgressDetail(e.payload);
+      });
+
+      register<string>("install-failed", (e) => {
+        setBackendError(e.payload, "install-error");
+      });
+
+      register<string>("repair-progress", (e) => {
+        setLogs((prev) => [...prev.slice(-499), e.payload]);
+      });
+
+      register<string[]>("repair-needs-elevation", (e) => {
+        elevationResumeRef.current = "repair";
+        setElevationPackages(e.payload);
+        setBackendStatus("needs-elevation");
+      });
+
+      register<void>("repair-complete", () => {
+        if (statusRef.current !== "repairing") return;
+        setProgressDetail("Repair complete");
+      });
+
+      register<string>("repair-failed", (e) => {
+        if (statusRef.current !== "repairing") return;
+        setBackendError(e.payload, "repair-error");
+      });
+
+      register<number>("server-port", (e) => {
+        portRef.current = e.payload;
+        setApiBase(e.payload);
+      });
+
+      register<void>("server-crashed", () => {
+        startingRef.current = false;
+        setBackendError("Server stopped unexpectedly");
+      });
+
+      register<string>("server-log", (e) => {
+        setLogs((prev) => [...prev.slice(-499), e.payload]);
+      });
+
+      register<void>("tray-toggle-server", () => {
+        if (statusRef.current === "running") {
+          stopServer();
+        } else if (
+          statusRef.current === "stopped" ||
+          statusRef.current === "error"
+        ) {
+          retry();
+        }
+      });
+    });
+
+    const onAuthFailed = (event: Event) => {
+      const detail =
+        event instanceof CustomEvent && typeof event.detail === "string"
+          ? event.detail
+          : "Desktop authentication failed. Update or repair the managed Studio install, then restart Studio.";
+      setAuthFailure(detail);
+    };
+    window.addEventListener("tauri-auth-failed", onAuthFailed);
+    const authFailure = getTauriAuthFailure();
+    if (authFailure) setAuthFailure(authFailure);
+    cleanup.push(() =>
+      window.removeEventListener("tauri-auth-failed", onAuthFailed),
+    );
+
+    return () => {
+      disposed = true;
+      cleanup.forEach((fn) => fn());
+      stopExternalServerPoll();
+    };
+  }, []);
+
+  return {
+    status, logs, error, isExternalServer,
+    currentStepIndex, progressDetail, elevationPackages,
+    startServer, stopServer, startInstall,
+    retry, retryInstall, approveElevation,
+  };
+}
diff --git a/studio/frontend/src/hooks/use-tauri-update.ts b/studio/frontend/src/hooks/use-tauri-update.ts
new file mode 100644
index 0000000000..5863ef7aa5
--- /dev/null
+++ b/studio/frontend/src/hooks/use-tauri-update.ts
@@ -0,0 +1,219 @@
+// SPDX-License-Identifier: AGPL-3.0-only
+// Copyright 2026-present the Unsloth AI Inc. team. All rights reserved. See /studio/LICENSE.AGPL-3.0
+
+import { useEffect, useRef, useState } from "react";
+import { isTauri } from "@/lib/api-base";
+import { toast } from "sonner";
+
+export type UpdateStatus =
+  | "idle"
+  | "checking"
+  | "available"
+  | "updating-backend"
+  | "downloading"
+  | "installing"
+  | "error";
+
+export interface UpdateInfo {
+  version: string;
+  currentVersion: string;
+  body?: string;
+  date?: string;
+}
+
+export function useTauriUpdate(isExternalServer = false) {
+  const [status, setStatus] = useState<UpdateStatus>("idle");
+  const [info, setInfo] = useState<UpdateInfo | null>(null);
+  const [progress, setProgress] = useState(0);
+  const [logs, setLogs] = useState<string[]>([]);
+  const [dismissed, setDismissed] = useState(false);
+  const [error, setError] = useState<string | null>(null);
+  const updateRef = useRef<Awaited<
+    ReturnType<typeof import("@tauri-apps/plugin-updater").check>
+  > | null>(null);
+  const checkedRef = useRef(false);
+  const updatingRef = useRef(false);
+
+  useEffect(() => {
+    if (!isTauri || checkedRef.current) return;
+    checkedRef.current = true;
+
+    async function checkForUpdate() {
+      setStatus("checking");
+      try {
+        const { check } = await import("@tauri-apps/plugin-updater");
+        const update = await check();
+        if (update) {
+          updateRef.current = update;
+          setInfo({
+            version: update.version,
+            currentVersion: update.currentVersion,
+            body: update.body,
+            date: update.date,
+          });
+          setStatus("available");
+        } else {
+          setStatus("idle");
+        }
+      } catch (e) {
+        console.error("Update check failed:", e);
+        setStatus("idle");
+      }
+    }
+
+    const timer = setTimeout(checkForUpdate, 5000);
+    return () => clearTimeout(timer);
+  }, []);
+
+  async function installUpdate() {
+    const update = updateRef.current;
+    if (!update || updatingRef.current) return;
+    updatingRef.current = true;
+
+    const cleanups: (() => void)[] = [];
+    let phase: "backend" | "shell" = "backend";
+
+    try {
+      // ── Step 1: Backend update ──
+      setStatus("updating-backend");
+      setLogs([]);
+      setError(null);
+      setDismissed(false);
+
+      const { listen } = await import("@tauri-apps/api/event");
+      const { invoke } = await import("@tauri-apps/api/core");
+
+      // Listen for backend update progress
+      const unlistenProgress = await listen<string>(
+        "update-progress",
+        (e) => {
+          setLogs((prev) => [...prev.slice(-499), e.payload]);
+        },
+      );
+      cleanups.push(unlistenProgress);
+
+      // Wait for complete or failed
+      const backendResult = await new Promise<"complete" | string>(
+        (resolve) => {
+          listen<void>("update-complete", () => resolve("complete")).then(
+            (u) => cleanups.push(u),
+          );
+          listen<string>("update-failed", (e) =>
+            resolve(e.payload),
+          ).then((u) => cleanups.push(u));
+
+          invoke("start_backend_update").catch((e) => resolve(String(e)));
+        },
+      );
+
+      if (backendResult !== "complete") {
+        setError(backendResult);
+        setStatus("error");
+        updatingRef.current = false;
+        cleanup(cleanups);
+        return;
+      }
+
+      // ── Step 2: Shell update ──
+      phase = "shell";
+      setStatus("downloading");
+      setProgress(0);
+
+      let downloaded = 0;
+      let contentLength = 0;
+      await update.downloadAndInstall((event) => {
+        switch (event.event) {
+          case "Started":
+            contentLength = event.data.contentLength ?? 0;
+            break;
+          case "Progress":
+            downloaded += event.data.chunkLength;
+            if (contentLength > 0) {
+              setProgress(Math.round((downloaded / contentLength) * 100));
+            }
+            break;
+          case "Finished":
+            setStatus("installing");
+            break;
+        }
+      });
+
+      // ── Step 3: Relaunch ──
+      const { relaunch } = await import("@tauri-apps/plugin-process");
+      await relaunch();
+    } catch (e) {
+      console.error("Update failed:", e);
+      const msg = String(e);
+
+      // Shell update failed — restart backend on updated code
+      if (phase === "shell") {
+        try {
+          const { invoke } = await import("@tauri-apps/api/core");
+          await invoke("start_server", { port: 8888 });
+          toast.error("App update failed", {
+            description:
+              "Backend was updated. The app update will be retried on next launch.",
+          });
+          setStatus("idle");
+          setDismissed(true);
+        } catch {
+          setError(msg);
+          setStatus("error");
+        }
+      } else {
+        setError(msg);
+        setStatus("error");
+      }
+    } finally {
+      updatingRef.current = false;
+      cleanup(cleanups);
+    }
+  }
+
+  async function retryUpdate() {
+    updatingRef.current = false;
+    await installUpdate();
+  }
+
+  async function skipAndRestart() {
+    try {
+      const { invoke } = await import("@tauri-apps/api/core");
+      await invoke("start_server", { port: 8888 });
+      setStatus("idle");
+      setError(null);
+      setLogs([]);
+      setDismissed(true);
+    } catch (e) {
+      setError(String(e));
+      setStatus("error");
+    }
+  }
+
+  function dismiss() {
+    setDismissed(true);
+  }
+
+  return {
+    status,
+    info,
+    progress,
+    logs,
+    dismissed,
+    error,
+    isExternalServer,
+    installUpdate,
+    retryUpdate,
+    skipAndRestart,
+    dismiss,
+  };
+}
+
+function cleanup(fns: (() => void)[]) {
+  for (const fn of fns) {
+    try {
+      fn();
+    } catch {
+      // ignore
+    }
+  }
+}
diff --git a/studio/frontend/src/lib/api-base.ts b/studio/frontend/src/lib/api-base.ts
new file mode 100644
index 0000000000..664b9e58bd
--- /dev/null
+++ b/studio/frontend/src/lib/api-base.ts
@@ -0,0 +1,24 @@
+// Central API base URL for Tauri vs browser mode
+let apiBase = ''
+
+const isTauri = typeof window !== 'undefined' && '__TAURI__' in window
+const isViteDev = import.meta.env.DEV
+
+if (isTauri && !isViteDev) {
+  apiBase = 'http://127.0.0.1:8888'
+}
+
+export function setApiBase(port: number) {
+  apiBase = `http://127.0.0.1:${port}`
+}
+
+export function getApiBase(): string {
+  return apiBase
+}
+
+export function apiUrl(path: string): string {
+  if (path.startsWith('http')) return path
+  return `${apiBase}${path}`
+}
+
+export { isTauri }
diff --git a/studio/frontend/src/lib/open-link.ts b/studio/frontend/src/lib/open-link.ts
new file mode 100644
index 0000000000..b7e6a0fa6e
--- /dev/null
+++ b/studio/frontend/src/lib/open-link.ts
@@ -0,0 +1,31 @@
+import { isTauri } from "@/lib/api-base";
+
+/**
+ * Open a URL in the system browser (Tauri) or new tab (web).
+ * Handles anchor links and mailto: natively without the opener plugin.
+ * Returns true when the caller should preventDefault; false when the
+ * browser's native navigation should proceed (relative URLs, empty, etc.).
+ */
+export function openLink(url: string): boolean {
+  if (!url) return false;
+
+  // Anchor links — scroll within the page, don't open externally
+  if (url.startsWith("#")) {
+    window.location.hash = url;
+    return true;
+  }
+
+  // Relative URLs — let the browser / router handle them natively
+  if (!url.includes("://") && !url.startsWith("mailto:")) {
+    return false;
+  }
+
+  if (isTauri) {
+    import("@tauri-apps/plugin-opener").then(({ openUrl }) => {
+      openUrl(url).catch(console.error);
+    });
+  } else {
+    window.open(url, "_blank", "noopener,noreferrer");
+  }
+  return true;
+}
diff --git a/studio/install_llama_prebuilt.py b/studio/install_llama_prebuilt.py
index 1ad1ccc851..23e6b7c0e4 100755
--- a/studio/install_llama_prebuilt.py
+++ b/studio/install_llama_prebuilt.py
@@ -46,6 +46,28 @@ EXIT_ERROR = 1
 EXIT_BUSY = 3
 
 
+def windows_hidden_subprocess_kwargs() -> dict[str, object]:
+    """Return Windows-only subprocess kwargs that suppress console windows."""
+    if sys.platform != "win32":
+        return {}
+
+    kwargs: dict[str, object] = {}
+    create_no_window = getattr(subprocess, "CREATE_NO_WINDOW", 0)
+    if create_no_window:
+        kwargs["creationflags"] = create_no_window
+
+    startupinfo_factory = getattr(subprocess, "STARTUPINFO", None)
+    startf_use_showwindow = getattr(subprocess, "STARTF_USESHOWWINDOW", 0)
+    sw_hide = getattr(subprocess, "SW_HIDE", 0)
+    if startupinfo_factory is not None and startf_use_showwindow:
+        startupinfo = startupinfo_factory()
+        startupinfo.dwFlags |= startf_use_showwindow
+        startupinfo.wShowWindow = sw_hide
+        kwargs["startupinfo"] = startupinfo
+
+    return kwargs
+
+
 def env_int(name: str, default: int, *, minimum: int | None = None) -> int:
     raw = os.environ.get(name)
     if raw is None:
@@ -2469,6 +2491,7 @@ def run_capture(
         text = True,
         timeout = timeout,
         env = env,
+        **windows_hidden_subprocess_kwargs(),
     )
     if check and result.returncode != 0:
         raise subprocess.CalledProcessError(
@@ -4357,6 +4380,7 @@ def validate_quantize(
         text = True,
         timeout = 120,
         env = binary_env(quantize_path, install_dir, host, runtime_line = runtime_line),
+        **windows_hidden_subprocess_kwargs(),
     )
     if (
         result.returncode != 0
@@ -4444,6 +4468,7 @@ def validate_server(
                     env = binary_env(
                         server_path, install_dir, host, runtime_line = runtime_line
                     ),
+                    **windows_hidden_subprocess_kwargs(),
                 )
                 deadline = time.time() + 20
                 startup_started = time.time()
diff --git a/studio/install_python_stack.py b/studio/install_python_stack.py
index 2b9fd084d4..b56b737cfa 100644
--- a/studio/install_python_stack.py
+++ b/studio/install_python_stack.py
@@ -359,6 +359,28 @@ def _ensure_rocm_torch() -> None:
             )
 
 
+def _windows_hidden_subprocess_kwargs() -> dict[str, object]:
+    """Return Windows-only subprocess kwargs that suppress console windows."""
+    if not IS_WINDOWS:
+        return {}
+
+    kwargs: dict[str, object] = {}
+    create_no_window = getattr(subprocess, "CREATE_NO_WINDOW", 0)
+    if create_no_window:
+        kwargs["creationflags"] = create_no_window
+
+    startupinfo_factory = getattr(subprocess, "STARTUPINFO", None)
+    startf_use_showwindow = getattr(subprocess, "STARTF_USESHOWWINDOW", 0)
+    sw_hide = getattr(subprocess, "SW_HIDE", 0)
+    if startupinfo_factory is not None and startf_use_showwindow:
+        startupinfo = startupinfo_factory()
+        startupinfo.dwFlags |= startf_use_showwindow
+        startupinfo.wShowWindow = sw_hide
+        kwargs["startupinfo"] = startupinfo
+
+    return kwargs
+
+
 def _infer_no_torch() -> bool:
     """Determine whether to run in no-torch (GGUF-only) mode.
 
@@ -412,9 +434,11 @@ _UNICODE_TO_ASCII: dict[str, str] = {
 
 
 def _safe_print(*args: object, **kwargs: object) -> None:
-    """Drop-in print() replacement that survives non-UTF-8 consoles."""
+    """Drop-in print() replacement that survives non-UTF-8 consoles and detached stdout."""
     try:
         print(*args, **kwargs)
+    except OSError:
+        return
     except UnicodeEncodeError:
         # Stringify, then swap emoji for ASCII equivalents
         text = " ".join(str(a) for a in args)
@@ -495,7 +519,7 @@ def _step(label: str, value: str, color_fn = None) -> None:
     if color_fn is None:
         color_fn = _green
     padded = label[:_COL]
-    print(f"  {_dim(padded)}{' ' * (_COL - len(padded))}{color_fn(value)}")
+    _safe_print(f"  {_dim(padded)}{' ' * (_COL - len(padded))}{color_fn(value)}")
 
 
 def _progress(label: str) -> None:
@@ -509,10 +533,13 @@ def _progress(label: str) -> None:
     bar = "=" * filled + "-" * (width - filled)
     pad = " " * (_COL - len(_LABEL))
     end = "\n" if _STEP >= _TOTAL else ""
-    sys.stdout.write(
-        f"\r  {_dim(_LABEL)}{pad}[{bar}] {_STEP:2}/{_TOTAL}  {label:<20}{end}"
-    )
-    sys.stdout.flush()
+    try:
+        sys.stdout.write(
+            f"\r  {_dim(_LABEL)}{pad}[{bar}] {_STEP:2}/{_TOTAL}  {label:<20}{end}"
+        )
+        sys.stdout.flush()
+    except OSError:
+        pass
 
 
 def run(
@@ -525,6 +552,7 @@ def run(
         cmd,
         stdout = subprocess.PIPE if quiet else None,
         stderr = subprocess.STDOUT if quiet else None,
+        **_windows_hidden_subprocess_kwargs(),
     )
     if result.returncode != 0:
         _step("error", f"{label} failed (exit code {result.returncode})", _red)
@@ -627,6 +655,7 @@ def _bootstrap_uv() -> bool:
         ["uv", "pip", "install", "--dry-run", "--python", sys.executable, "pip"],
         stdout = subprocess.PIPE,
         stderr = subprocess.STDOUT,
+        **_windows_hidden_subprocess_kwargs(),
     )
     if probe.returncode != 0:
         # Retry with --system (some envs need it when uv can't find a venv)
@@ -634,6 +663,7 @@ def _bootstrap_uv() -> bool:
             ["uv", "pip", "install", "--dry-run", "--system", "pip"],
             stdout = subprocess.PIPE,
             stderr = subprocess.STDOUT,
+            **_windows_hidden_subprocess_kwargs(),
         )
         if probe_sys.returncode != 0:
             return False  # uv is broken, fall back to pip
@@ -773,6 +803,7 @@ def pip_install(
                 uv_cmd,
                 stdout = subprocess.PIPE,
                 stderr = subprocess.STDOUT,
+                **_windows_hidden_subprocess_kwargs(),
             )
             if result.returncode == 0:
                 return
@@ -798,6 +829,7 @@ def patch_package_file(package_name: str, relative_path: str, url: str) -> None:
         [sys.executable, "-m", "pip", "show", package_name],
         capture_output = True,
         text = True,
+        **_windows_hidden_subprocess_kwargs(),
     )
     if result.returncode != 0:
         _step(_LABEL, f"package {package_name} not found, skipping patch", _red)
@@ -868,6 +900,7 @@ def install_python_stack() -> int:
                 [sys.executable, "-m", "pip", "--version"],
                 stdout = subprocess.DEVNULL,
                 stderr = subprocess.DEVNULL,
+                **_windows_hidden_subprocess_kwargs(),
             ).returncode
             == 0
         )
@@ -1140,6 +1173,7 @@ def install_python_stack() -> int:
         [sys.executable, "-m", "pip", "check"],
         stdout = subprocess.DEVNULL,
         stderr = subprocess.DEVNULL,
+        **_windows_hidden_subprocess_kwargs(),
     )
 
     _step(_LABEL, "installed")
diff --git a/studio/setup.ps1 b/studio/setup.ps1
index 218be7e045..24ef3ef1eb 100644
--- a/studio/setup.ps1
+++ b/studio/setup.ps1
@@ -1077,10 +1077,13 @@ if (-not $CudaArch) {
 }
 
 # ============================================
-# 1f. Node.js / npm (skip if pip-installed -- only needed for frontend build)
+# 1f. Node.js / npm (skip if pip-installed or Tauri -- only needed for frontend build)
 # ============================================
+$SkipFrontend = ($env:SKIP_STUDIO_FRONTEND -eq "1")
 if ($IsPipInstall) {
     step "frontend" "bundled (pip install)"
+} elseif ($SkipFrontend) {
+    step "frontend" "bundled (Tauri)"
 } else {
     # setup.sh installs Node LTS (v22) via nvm. We enforce the same range here:
     # Vite 8 requires Node ^20.19.0 || >=22.12.0, npm >= 11.
@@ -1207,6 +1210,9 @@ $NeedFrontendBuild = $true
 if ($IsPipInstall) {
     $NeedFrontendBuild = $false
     step "frontend" "bundled (pip install)"
+} elseif ($SkipFrontend) {
+    $NeedFrontendBuild = $false
+    step "frontend" "bundled (Tauri)"
 } elseif (Test-Path $DistDir) {
     $DistTime = (Get-Item $DistDir).LastWriteTime
     $NewerFile = $null
@@ -1539,7 +1545,7 @@ if (Get-Command uv -ErrorAction SilentlyContinue) {
 } else {
     substep "installing uv package manager..."
     try {
-        Invoke-SetupCommand { powershell -ExecutionPolicy ByPass -c "irm https://astral.sh/uv/install.ps1 | iex" } | Out-Null
+        Invoke-SetupCommand { Invoke-Expression (Invoke-RestMethod -Uri "https://astral.sh/uv/install.ps1") } | Out-Null
         Refresh-Environment
         # Re-activate venv since Refresh-Environment rebuilds PATH from
         # registry and drops the venv's Scripts directory
diff --git a/studio/setup.sh b/studio/setup.sh
index c2044b5141..142d253554 100755
--- a/studio/setup.sh
+++ b/studio/setup.sh
@@ -183,7 +183,13 @@ if [[ "$keynames" == *$'\nCOLAB_'* ]]; then
 fi
 
 if [ "$_LLAMA_ONLY" != "1" ]; then
-# ── Frontend ──
+# ── Detect whether frontend needs building ──
+# Skip if SKIP_STUDIO_FRONTEND=1 (Tauri desktop app bundles its own frontend),
+# or if dist/ exists AND no tracked input is newer than dist/.
+if [ "${SKIP_STUDIO_FRONTEND:-0}" = "1" ]; then
+    _NEED_FRONTEND_BUILD=false
+    step "frontend" "bundled (Tauri)"
+else
 _NEED_FRONTEND_BUILD=true
 if [ -d "$SCRIPT_DIR/frontend/dist" ]; then
     _changed=$(find "$SCRIPT_DIR/frontend" -maxdepth 1 -type f \
@@ -195,6 +201,7 @@ if [ -d "$SCRIPT_DIR/frontend/dist" ]; then
     fi
     [ -z "$_changed" ] && _NEED_FRONTEND_BUILD=false
 fi
+fi  # end SKIP_STUDIO_FRONTEND guard
 
 if [ "$_NEED_FRONTEND_BUILD" = false ]; then
     step "frontend" "up to date"
@@ -484,9 +491,9 @@ _PKG_NAME="${STUDIO_PACKAGE_NAME:-unsloth}"
 if [ "$_SKIP_VERSION_CHECK" != true ] && [ "${SKIP_STUDIO_BASE:-0}" != "1" ] && [ "${STUDIO_LOCAL_INSTALL:-0}" != "1" ]; then
     # Only check when NOT called from install.sh (which just installed the package)
     INSTALLED_VER=$("$VENV_DIR/bin/python" -c "
-from importlib.metadata import version
-print(version('$_PKG_NAME'))
-" 2>/dev/null || echo "")
+import sys; from importlib.metadata import version
+print(version(sys.argv[1]))
+" "$_PKG_NAME" 2>/dev/null || echo "")
 
     LATEST_VER=$(curl -fsSL --max-time 5 "https://pypi.org/pypi/$_PKG_NAME/json" 2>/dev/null \
         | "$VENV_DIR/bin/python" -c "import sys,json; print(json.load(sys.stdin)['info']['version'])" 2>/dev/null \
diff --git a/studio/src-tauri/Cargo.lock b/studio/src-tauri/Cargo.lock
new file mode 100644
index 0000000000..62d520437c
--- /dev/null
+++ b/studio/src-tauri/Cargo.lock
@@ -0,0 +1,6320 @@
+# This file is automatically @generated by Cargo.
+# It is not intended for manual editing.
+version = 4
+
+[[package]]
+name = "adler2"
+version = "2.0.1"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "320119579fcad9c21884f5c4861d16174d0e06250625266f50fe6898340abefa"
+
+[[package]]
+name = "aho-corasick"
+version = "1.1.4"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "ddd31a130427c27518df266943a5308ed92d4b226cc639f5a8f1002816174301"
+dependencies = [
+ "memchr",
+]
+
+[[package]]
+name = "alloc-no-stdlib"
+version = "2.0.4"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "cc7bb162ec39d46ab1ca8c77bf72e890535becd1751bb45f64c597edb4c8c6b3"
+
+[[package]]
+name = "alloc-stdlib"
+version = "0.2.2"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "94fb8275041c72129eb51b7d0322c29b8387a0386127718b096429201a5d6ece"
+dependencies = [
+ "alloc-no-stdlib",
+]
+
+[[package]]
+name = "android_system_properties"
+version = "0.1.5"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "819e7219dbd41043ac279b19830f2efc897156490d7fd6ea916720117ee66311"
+dependencies = [
+ "libc",
+]
+
+[[package]]
+name = "anyhow"
+version = "1.0.102"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "7f202df86484c868dbad7eaa557ef785d5c66295e41b460ef922eca0723b842c"
+
+[[package]]
+name = "arbitrary"
+version = "1.4.2"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "c3d036a3c4ab069c7b410a2ce876bd74808d2d0888a82667669f8e783a898bf1"
+dependencies = [
+ "derive_arbitrary",
+]
+
+[[package]]
+name = "async-broadcast"
+version = "0.7.2"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "435a87a52755b8f27fcf321ac4f04b2802e337c8c4872923137471ec39c37532"
+dependencies = [
+ "event-listener",
+ "event-listener-strategy",
+ "futures-core",
+ "pin-project-lite",
+]
+
+[[package]]
+name = "async-channel"
+version = "2.5.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "924ed96dd52d1b75e9c1a3e6275715fd320f5f9439fb5a4a11fa51f4221158d2"
+dependencies = [
+ "concurrent-queue",
+ "event-listener-strategy",
+ "futures-core",
+ "pin-project-lite",
+]
+
+[[package]]
+name = "async-executor"
+version = "1.14.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "c96bf972d85afc50bf5ab8fe2d54d1586b4e0b46c97c50a0c9e71e2f7bcd812a"
+dependencies = [
+ "async-task",
+ "concurrent-queue",
+ "fastrand",
+ "futures-lite",
+ "pin-project-lite",
+ "slab",
+]
+
+[[package]]
+name = "async-io"
+version = "2.6.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "456b8a8feb6f42d237746d4b3e9a178494627745c3c56c6ea55d92ba50d026fc"
+dependencies = [
+ "autocfg",
+ "cfg-if",
+ "concurrent-queue",
+ "futures-io",
+ "futures-lite",
+ "parking",
+ "polling",
+ "rustix",
+ "slab",
+ "windows-sys 0.61.2",
+]
+
+[[package]]
+name = "async-lock"
+version = "3.4.2"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "290f7f2596bd5b78a9fec8088ccd89180d7f9f55b94b0576823bbbdc72ee8311"
+dependencies = [
+ "event-listener",
+ "event-listener-strategy",
+ "pin-project-lite",
+]
+
+[[package]]
+name = "async-process"
+version = "2.5.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "fc50921ec0055cdd8a16de48773bfeec5c972598674347252c0399676be7da75"
+dependencies = [
+ "async-channel",
+ "async-io",
+ "async-lock",
+ "async-signal",
+ "async-task",
+ "blocking",
+ "cfg-if",
+ "event-listener",
+ "futures-lite",
+ "rustix",
+]
+
+[[package]]
+name = "async-recursion"
+version = "1.1.1"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "3b43422f69d8ff38f95f1b2bb76517c91589a924d1559a0e935d7c8ce0274c11"
+dependencies = [
+ "proc-macro2",
+ "quote",
+ "syn 2.0.117",
+]
+
+[[package]]
+name = "async-signal"
+version = "0.2.13"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "43c070bbf59cd3570b6b2dd54cd772527c7c3620fce8be898406dd3ed6adc64c"
+dependencies = [
+ "async-io",
+ "async-lock",
+ "atomic-waker",
+ "cfg-if",
+ "futures-core",
+ "futures-io",
+ "rustix",
+ "signal-hook-registry",
+ "slab",
+ "windows-sys 0.61.2",
+]
+
+[[package]]
+name = "async-task"
+version = "4.7.1"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "8b75356056920673b02621b35afd0f7dda9306d03c79a30f5c56c44cf256e3de"
+
+[[package]]
+name = "async-trait"
+version = "0.1.89"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "9035ad2d096bed7955a320ee7e2230574d28fd3c3a0f186cbea1ff3c7eed5dbb"
+dependencies = [
+ "proc-macro2",
+ "quote",
+ "syn 2.0.117",
+]
+
+[[package]]
+name = "atk"
+version = "0.18.2"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "241b621213072e993be4f6f3a9e4b45f65b7e6faad43001be957184b7bb1824b"
+dependencies = [
+ "atk-sys",
+ "glib",
+ "libc",
+]
+
+[[package]]
+name = "atk-sys"
+version = "0.18.2"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "c5e48b684b0ca77d2bbadeef17424c2ea3c897d44d566a1617e7e8f30614d086"
+dependencies = [
+ "glib-sys",
+ "gobject-sys",
+ "libc",
+ "system-deps",
+]
+
+[[package]]
+name = "atomic-waker"
+version = "1.1.2"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "1505bd5d3d116872e7271a6d4e16d81d0c8570876c8de68093a09ac269d8aac0"
+
+[[package]]
+name = "autocfg"
+version = "1.5.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "c08606f8c3cbf4ce6ec8e28fb0014a2c086708fe954eaa885384a6165172e7e8"
+
+[[package]]
+name = "base64"
+version = "0.21.7"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "9d297deb1925b89f2ccc13d7635fa0714f12c87adce1c75356b39ca9b7178567"
+
+[[package]]
+name = "base64"
+version = "0.22.1"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "72b3254f16251a8381aa12e40e3c4d2f0199f8c6508fbecb9d91f575e0fbb8c6"
+
+[[package]]
+name = "bit-set"
+version = "0.8.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "08807e080ed7f9d5433fa9b275196cfc35414f66a0c79d864dc51a0d825231a3"
+dependencies = [
+ "bit-vec",
+]
+
+[[package]]
+name = "bit-vec"
+version = "0.8.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "5e764a1d40d510daf35e07be9eb06e75770908c27d411ee6c92109c9840eaaf7"
+
+[[package]]
+name = "bitflags"
+version = "1.3.2"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "bef38d45163c2f1dde094a7dfd33ccf595c92905c8f8f4fdc18d06fb1037718a"
+
+[[package]]
+name = "bitflags"
+version = "2.11.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "843867be96c8daad0d758b57df9392b6d8d271134fce549de6ce169ff98a92af"
+dependencies = [
+ "serde_core",
+]
+
+[[package]]
+name = "block-buffer"
+version = "0.10.4"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "3078c7629b62d3f0439517fa394996acacc5cbc91c5a20d8c658e77abd503a71"
+dependencies = [
+ "generic-array",
+]
+
+[[package]]
+name = "block2"
+version = "0.6.2"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "cdeb9d870516001442e364c5220d3574d2da8dc765554b4a617230d33fa58ef5"
+dependencies = [
+ "objc2",
+]
+
+[[package]]
+name = "blocking"
+version = "1.6.2"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "e83f8d02be6967315521be875afa792a316e28d57b5a2d401897e2a7921b7f21"
+dependencies = [
+ "async-channel",
+ "async-task",
+ "futures-io",
+ "futures-lite",
+ "piper",
+]
+
+[[package]]
+name = "brotli"
+version = "8.0.2"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "4bd8b9603c7aa97359dbd97ecf258968c95f3adddd6db2f7e7a5bef101c84560"
+dependencies = [
+ "alloc-no-stdlib",
+ "alloc-stdlib",
+ "brotli-decompressor",
+]
+
+[[package]]
+name = "brotli-decompressor"
+version = "5.0.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "874bb8112abecc98cbd6d81ea4fa7e94fb9449648c93cc89aa40c81c24d7de03"
+dependencies = [
+ "alloc-no-stdlib",
+ "alloc-stdlib",
+]
+
+[[package]]
+name = "bumpalo"
+version = "3.20.2"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "5d20789868f4b01b2f2caec9f5c4e0213b41e3e5702a50157d699ae31ced2fcb"
+
+[[package]]
+name = "bytemuck"
+version = "1.25.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "c8efb64bd706a16a1bdde310ae86b351e4d21550d98d056f22f8a7f7a2183fec"
+
+[[package]]
+name = "byteorder"
+version = "1.5.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "1fd0f2584146f6f2ef48085050886acf353beff7305ebd1ae69500e27c67f64b"
+
+[[package]]
+name = "bytes"
+version = "1.11.1"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "1e748733b7cbc798e1434b6ac524f0c1ff2ab456fe201501e6497c8417a4fc33"
+dependencies = [
+ "serde",
+]
+
+[[package]]
+name = "cairo-rs"
+version = "0.18.5"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "8ca26ef0159422fb77631dc9d17b102f253b876fe1586b03b803e63a309b4ee2"
+dependencies = [
+ "bitflags 2.11.0",
+ "cairo-sys-rs",
+ "glib",
+ "libc",
+ "once_cell",
+ "thiserror 1.0.69",
+]
+
+[[package]]
+name = "cairo-sys-rs"
+version = "0.18.2"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "685c9fa8e590b8b3d678873528d83411db17242a73fccaed827770ea0fedda51"
+dependencies = [
+ "glib-sys",
+ "libc",
+ "system-deps",
+]
+
+[[package]]
+name = "camino"
+version = "1.2.2"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "e629a66d692cb9ff1a1c664e41771b3dcaf961985a9774c0eb0bd1b51cf60a48"
+dependencies = [
+ "serde_core",
+]
+
+[[package]]
+name = "cargo-platform"
+version = "0.1.9"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "e35af189006b9c0f00a064685c727031e3ed2d8020f7ba284d78cc2671bd36ea"
+dependencies = [
+ "serde",
+]
+
+[[package]]
+name = "cargo_metadata"
+version = "0.19.2"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "dd5eb614ed4c27c5d706420e4320fbe3216ab31fa1c33cd8246ac36dae4479ba"
+dependencies = [
+ "camino",
+ "cargo-platform",
+ "semver",
+ "serde",
+ "serde_json",
+ "thiserror 2.0.18",
+]
+
+[[package]]
+name = "cargo_toml"
+version = "0.22.3"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "374b7c592d9c00c1f4972ea58390ac6b18cbb6ab79011f3bdc90a0b82ca06b77"
+dependencies = [
+ "serde",
+ "toml 0.9.12+spec-1.1.0",
+]
+
+[[package]]
+name = "cc"
+version = "1.2.57"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "7a0dd1ca384932ff3641c8718a02769f1698e7563dc6974ffd03346116310423"
+dependencies = [
+ "find-msvc-tools",
+ "shlex",
+]
+
+[[package]]
+name = "cesu8"
+version = "1.1.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "6d43a04d8753f35258c91f8ec639f792891f748a1edbd759cf1dcea3382ad83c"
+
+[[package]]
+name = "cfb"
+version = "0.7.3"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "d38f2da7a0a2c4ccf0065be06397cc26a81f4e528be095826eee9d4adbb8c60f"
+dependencies = [
+ "byteorder",
+ "fnv",
+ "uuid",
+]
+
+[[package]]
+name = "cfg-expr"
+version = "0.15.8"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "d067ad48b8650848b989a59a86c6c36a995d02d2bf778d45c3c5d57bc2718f02"
+dependencies = [
+ "smallvec",
+ "target-lexicon",
+]
+
+[[package]]
+name = "cfg-if"
+version = "1.0.4"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "9330f8b2ff13f34540b44e946ef35111825727b38d33286ef986142615121801"
+
+[[package]]
+name = "cfg_aliases"
+version = "0.2.1"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "613afe47fcd5fac7ccf1db93babcb082c5994d996f20b8b159f2ad1658eb5724"
+
+[[package]]
+name = "chacha20"
+version = "0.10.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "6f8d983286843e49675a4b7a2d174efe136dc93a18d69130dd18198a6c167601"
+dependencies = [
+ "cfg-if",
+ "cpufeatures 0.3.0",
+ "rand_core 0.10.0",
+]
+
+[[package]]
+name = "chrono"
+version = "0.4.44"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "c673075a2e0e5f4a1dde27ce9dee1ea4558c7ffe648f576438a20ca1d2acc4b0"
+dependencies = [
+ "iana-time-zone",
+ "num-traits",
+ "serde",
+ "windows-link 0.2.1",
+]
+
+[[package]]
+name = "combine"
+version = "4.6.7"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "ba5a308b75df32fe02788e748662718f03fde005016435c444eea572398219fd"
+dependencies = [
+ "bytes",
+ "memchr",
+]
+
+[[package]]
+name = "concurrent-queue"
+version = "2.5.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "4ca0197aee26d1ae37445ee532fefce43251d24cc7c166799f4d46817f1d3973"
+dependencies = [
+ "crossbeam-utils",
+]
+
+[[package]]
+name = "convert_case"
+version = "0.4.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "6245d59a3e82a7fc217c5828a6692dbc6dfb63a0c8c90495621f7b9d79704a0e"
+
+[[package]]
+name = "cookie"
+version = "0.18.1"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "4ddef33a339a91ea89fb53151bd0a4689cfce27055c291dfa69945475d22c747"
+dependencies = [
+ "time",
+ "version_check",
+]
+
+[[package]]
+name = "core-foundation"
+version = "0.9.4"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "91e195e091a93c46f7102ec7818a2aa394e1e1771c3ab4825963fa03e45afb8f"
+dependencies = [
+ "core-foundation-sys",
+ "libc",
+]
+
+[[package]]
+name = "core-foundation"
+version = "0.10.1"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "b2a6cd9ae233e7f62ba4e9353e81a88df7fc8a5987b8d445b4d90c879bd156f6"
+dependencies = [
+ "core-foundation-sys",
+ "libc",
+]
+
+[[package]]
+name = "core-foundation-sys"
+version = "0.8.7"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "773648b94d0e5d620f64f280777445740e61fe701025087ec8b57f45c791888b"
+
+[[package]]
+name = "core-graphics"
+version = "0.25.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "064badf302c3194842cf2c5d61f56cc88e54a759313879cdf03abdd27d0c3b97"
+dependencies = [
+ "bitflags 2.11.0",
+ "core-foundation 0.10.1",
+ "core-graphics-types",
+ "foreign-types 0.5.0",
+ "libc",
+]
+
+[[package]]
+name = "core-graphics-types"
+version = "0.2.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "3d44a101f213f6c4cdc1853d4b78aef6db6bdfa3468798cc1d9912f4735013eb"
+dependencies = [
+ "bitflags 2.11.0",
+ "core-foundation 0.10.1",
+ "libc",
+]
+
+[[package]]
+name = "cpufeatures"
+version = "0.2.17"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "59ed5838eebb26a2bb2e58f6d5b5316989ae9d08bab10e0e6d103e656d1b0280"
+dependencies = [
+ "libc",
+]
+
+[[package]]
+name = "cpufeatures"
+version = "0.3.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "8b2a41393f66f16b0823bb79094d54ac5fbd34ab292ddafb9a0456ac9f87d201"
+dependencies = [
+ "libc",
+]
+
+[[package]]
+name = "crc32fast"
+version = "1.5.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "9481c1c90cbf2ac953f07c8d4a58aa3945c425b7185c9154d67a65e4230da511"
+dependencies = [
+ "cfg-if",
+]
+
+[[package]]
+name = "crossbeam-channel"
+version = "0.5.15"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "82b8f8f868b36967f9606790d1903570de9ceaf870a7bf9fbbd3016d636a2cb2"
+dependencies = [
+ "crossbeam-utils",
+]
+
+[[package]]
+name = "crossbeam-utils"
+version = "0.8.21"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "d0a5c400df2834b80a4c3327b3aad3a4c4cd4de0629063962b03235697506a28"
+
+[[package]]
+name = "crypto-common"
+version = "0.1.7"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "78c8292055d1c1df0cce5d180393dc8cce0abec0a7102adb6c7b1eef6016d60a"
+dependencies = [
+ "generic-array",
+ "typenum",
+]
+
+[[package]]
+name = "cssparser"
+version = "0.29.6"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "f93d03419cb5950ccfd3daf3ff1c7a36ace64609a1a8746d493df1ca0afde0fa"
+dependencies = [
+ "cssparser-macros",
+ "dtoa-short",
+ "itoa",
+ "matches",
+ "phf 0.10.1",
+ "proc-macro2",
+ "quote",
+ "smallvec",
+ "syn 1.0.109",
+]
+
+[[package]]
+name = "cssparser"
+version = "0.36.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "dae61cf9c0abb83bd659dab65b7e4e38d8236824c85f0f804f173567bda257d2"
+dependencies = [
+ "cssparser-macros",
+ "dtoa-short",
+ "itoa",
+ "phf 0.13.1",
+ "smallvec",
+]
+
+[[package]]
+name = "cssparser-macros"
+version = "0.6.1"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "13b588ba4ac1a99f7f2964d24b3d896ddc6bf847ee3855dbd4366f058cfcd331"
+dependencies = [
+ "quote",
+ "syn 2.0.117",
+]
+
+[[package]]
+name = "ctor"
+version = "0.2.9"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "32a2785755761f3ddc1492979ce1e48d2c00d09311c39e4466429188f3dd6501"
+dependencies = [
+ "quote",
+ "syn 2.0.117",
+]
+
+[[package]]
+name = "darling"
+version = "0.23.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "25ae13da2f202d56bd7f91c25fba009e7717a1e4a1cc98a76d844b65ae912e9d"
+dependencies = [
+ "darling_core",
+ "darling_macro",
+]
+
+[[package]]
+name = "darling_core"
+version = "0.23.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "9865a50f7c335f53564bb694ef660825eb8610e0a53d3e11bf1b0d3df31e03b0"
+dependencies = [
+ "ident_case",
+ "proc-macro2",
+ "quote",
+ "strsim",
+ "syn 2.0.117",
+]
+
+[[package]]
+name = "darling_macro"
+version = "0.23.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "ac3984ec7bd6cfa798e62b4a642426a5be0e68f9401cfc2a01e3fa9ea2fcdb8d"
+dependencies = [
+ "darling_core",
+ "quote",
+ "syn 2.0.117",
+]
+
+[[package]]
+name = "deranged"
+version = "0.5.8"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "7cd812cc2bc1d69d4764bd80df88b4317eaef9e773c75226407d9bc0876b211c"
+dependencies = [
+ "powerfmt",
+ "serde_core",
+]
+
+[[package]]
+name = "derive_arbitrary"
+version = "1.4.2"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "1e567bd82dcff979e4b03460c307b3cdc9e96fde3d73bed1496d2bc75d9dd62a"
+dependencies = [
+ "proc-macro2",
+ "quote",
+ "syn 2.0.117",
+]
+
+[[package]]
+name = "derive_more"
+version = "0.99.20"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "6edb4b64a43d977b8e99788fe3a04d483834fba1215a7e02caa415b626497f7f"
+dependencies = [
+ "convert_case",
+ "proc-macro2",
+ "quote",
+ "rustc_version",
+ "syn 2.0.117",
+]
+
+[[package]]
+name = "derive_more"
+version = "2.1.1"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "d751e9e49156b02b44f9c1815bcb94b984cdcc4396ecc32521c739452808b134"
+dependencies = [
+ "derive_more-impl",
+]
+
+[[package]]
+name = "derive_more-impl"
+version = "2.1.1"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "799a97264921d8623a957f6c3b9011f3b5492f557bbb7a5a19b7fa6d06ba8dcb"
+dependencies = [
+ "proc-macro2",
+ "quote",
+ "rustc_version",
+ "syn 2.0.117",
+]
+
+[[package]]
+name = "digest"
+version = "0.10.7"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "9ed9a281f7bc9b7576e61468ba615a66a5c8cfdff42420a70aa82701a3b1e292"
+dependencies = [
+ "block-buffer",
+ "crypto-common",
+]
+
+[[package]]
+name = "dirs"
+version = "6.0.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "c3e8aa94d75141228480295a7d0e7feb620b1a5ad9f12bc40be62411e38cce4e"
+dependencies = [
+ "dirs-sys",
+]
+
+[[package]]
+name = "dirs-sys"
+version = "0.5.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "e01a3366d27ee9890022452ee61b2b63a67e6f13f58900b651ff5665f0bb1fab"
+dependencies = [
+ "libc",
+ "option-ext",
+ "redox_users",
+ "windows-sys 0.61.2",
+]
+
+[[package]]
+name = "dispatch2"
+version = "0.3.1"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "1e0e367e4e7da84520dedcac1901e4da967309406d1e51017ae1abfb97adbd38"
+dependencies = [
+ "bitflags 2.11.0",
+ "block2",
+ "libc",
+ "objc2",
+]
+
+[[package]]
+name = "displaydoc"
+version = "0.2.5"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "97369cbbc041bc366949bc74d34658d6cda5621039731c6310521892a3a20ae0"
+dependencies = [
+ "proc-macro2",
+ "quote",
+ "syn 2.0.117",
+]
+
+[[package]]
+name = "dlopen2"
+version = "0.8.2"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "5e2c5bd4158e66d1e215c49b837e11d62f3267b30c92f1d171c4d3105e3dc4d4"
+dependencies = [
+ "dlopen2_derive",
+ "libc",
+ "once_cell",
+ "winapi",
+]
+
+[[package]]
+name = "dlopen2_derive"
+version = "0.4.3"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "0fbbb781877580993a8707ec48672673ec7b81eeba04cfd2310bd28c08e47c8f"
+dependencies = [
+ "proc-macro2",
+ "quote",
+ "syn 2.0.117",
+]
+
+[[package]]
+name = "dom_query"
+version = "0.27.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "521e380c0c8afb8d9a1e83a1822ee03556fc3e3e7dbc1fd30be14e37f9cb3f89"
+dependencies = [
+ "bit-set",
+ "cssparser 0.36.0",
+ "foldhash 0.2.0",
+ "html5ever 0.38.0",
+ "precomputed-hash",
+ "selectors 0.36.1",
+ "tendril 0.5.0",
+]
+
+[[package]]
+name = "dpi"
+version = "0.1.2"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "d8b14ccef22fc6f5a8f4d7d768562a182c04ce9a3b3157b91390b52ddfdf1a76"
+dependencies = [
+ "serde",
+]
+
+[[package]]
+name = "dtoa"
+version = "1.0.11"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "4c3cf4824e2d5f025c7b531afcb2325364084a16806f6d47fbc1f5fbd9960590"
+
+[[package]]
+name = "dtoa-short"
+version = "0.3.5"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "cd1511a7b6a56299bd043a9c167a6d2bfb37bf84a6dfceaba651168adfb43c87"
+dependencies = [
+ "dtoa",
+]
+
+[[package]]
+name = "dunce"
+version = "1.0.5"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "92773504d58c093f6de2459af4af33faa518c13451eb8f2b5698ed3d36e7c813"
+
+[[package]]
+name = "dyn-clone"
+version = "1.0.20"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "d0881ea181b1df73ff77ffaaf9c7544ecc11e82fba9b5f27b262a3c73a332555"
+
+[[package]]
+name = "elevated-command"
+version = "1.1.2"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "54c410eccdcc5b759704fdb6a792afe6b01ab8a062e2c003ff2567e2697a94aa"
+dependencies = [
+ "anyhow",
+ "base64 0.21.7",
+ "libc",
+ "log",
+ "winapi",
+ "windows 0.52.0",
+]
+
+[[package]]
+name = "embed-resource"
+version = "3.0.7"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "47ec73ddcf6b7f23173d5c3c5a32b5507dc0a734de7730aa14abc5d5e296bb5f"
+dependencies = [
+ "cc",
+ "memchr",
+ "rustc_version",
+ "toml 0.9.12+spec-1.1.0",
+ "vswhom",
+ "winreg",
+]
+
+[[package]]
+name = "embed_plist"
+version = "1.2.2"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "4ef6b89e5b37196644d8796de5268852ff179b44e96276cf4290264843743bb7"
+
+[[package]]
+name = "encoding_rs"
+version = "0.8.35"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "75030f3c4f45dafd7586dd6780965a8c7e8e285a5ecb86713e63a79c5b2766f3"
+dependencies = [
+ "cfg-if",
+]
+
+[[package]]
+name = "endi"
+version = "1.1.1"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "66b7e2430c6dff6a955451e2cfc438f09cea1965a9d6f87f7e3b90decc014099"
+
+[[package]]
+name = "enumflags2"
+version = "0.7.12"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "1027f7680c853e056ebcec683615fb6fbbc07dbaa13b4d5d9442b146ded4ecef"
+dependencies = [
+ "enumflags2_derive",
+ "serde",
+]
+
+[[package]]
+name = "enumflags2_derive"
+version = "0.7.12"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "67c78a4d8fdf9953a5c9d458f9efe940fd97a0cab0941c075a813ac594733827"
+dependencies = [
+ "proc-macro2",
+ "quote",
+ "syn 2.0.117",
+]
+
+[[package]]
+name = "equivalent"
+version = "1.0.2"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "877a4ace8713b0bcf2a4e7eec82529c029f1d0619886d18145fea96c3ffe5c0f"
+
+[[package]]
+name = "erased-serde"
+version = "0.4.10"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "d2add8a07dd6a8d93ff627029c51de145e12686fbc36ecb298ac22e74cf02dec"
+dependencies = [
+ "serde",
+ "serde_core",
+ "typeid",
+]
+
+[[package]]
+name = "errno"
+version = "0.3.14"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "39cab71617ae0d63f51a36d69f866391735b51691dbda63cf6f96d042b63efeb"
+dependencies = [
+ "libc",
+ "windows-sys 0.61.2",
+]
+
+[[package]]
+name = "event-listener"
+version = "5.4.1"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "e13b66accf52311f30a0db42147dadea9850cb48cd070028831ae5f5d4b856ab"
+dependencies = [
+ "concurrent-queue",
+ "parking",
+ "pin-project-lite",
+]
+
+[[package]]
+name = "event-listener-strategy"
+version = "0.5.4"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "8be9f3dfaaffdae2972880079a491a1a8bb7cbed0b8dd7a347f668b4150a3b93"
+dependencies = [
+ "event-listener",
+ "pin-project-lite",
+]
+
+[[package]]
+name = "fastrand"
+version = "2.3.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "37909eebbb50d72f9059c3b6d82c0463f2ff062c9e95845c43a6c9c0355411be"
+
+[[package]]
+name = "fdeflate"
+version = "0.3.7"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "1e6853b52649d4ac5c0bd02320cddc5ba956bdb407c4b75a2c6b75bf51500f8c"
+dependencies = [
+ "simd-adler32",
+]
+
+[[package]]
+name = "field-offset"
+version = "0.3.6"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "38e2275cc4e4fc009b0669731a1e5ab7ebf11f469eaede2bab9309a5b4d6057f"
+dependencies = [
+ "memoffset",
+ "rustc_version",
+]
+
+[[package]]
+name = "filetime"
+version = "0.2.27"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "f98844151eee8917efc50bd9e8318cb963ae8b297431495d3f758616ea5c57db"
+dependencies = [
+ "cfg-if",
+ "libc",
+ "libredox",
+]
+
+[[package]]
+name = "find-msvc-tools"
+version = "0.1.9"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "5baebc0774151f905a1a2cc41989300b1e6fbb29aff0ceffa1064fdd3088d582"
+
+[[package]]
+name = "fix-path-env"
+version = "0.0.0"
+source = "git+https://github.com/tauri-apps/fix-path-env-rs#c4c45d503ea115a839aae718d02f79e7c7f0f673"
+dependencies = [
+ "home",
+ "strip-ansi-escapes",
+ "thiserror 1.0.69",
+]
+
+[[package]]
+name = "flate2"
+version = "1.1.9"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "843fba2746e448b37e26a819579957415c8cef339bf08564fe8b7ddbd959573c"
+dependencies = [
+ "crc32fast",
+ "miniz_oxide",
+]
+
+[[package]]
+name = "fnv"
+version = "1.0.7"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "3f9eec918d3f24069decb9af1554cad7c880e2da24a9afd88aca000531ab82c1"
+
+[[package]]
+name = "foldhash"
+version = "0.1.5"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "d9c4f5dac5e15c24eb999c26181a6ca40b39fe946cbe4c263c7209467bc83af2"
+
+[[package]]
+name = "foldhash"
+version = "0.2.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "77ce24cb58228fbb8aa041425bb1050850ac19177686ea6e0f41a70416f56fdb"
+
+[[package]]
+name = "foreign-types"
+version = "0.3.2"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "f6f339eb8adc052cd2ca78910fda869aefa38d22d5cb648e6485e4d3fc06f3b1"
+dependencies = [
+ "foreign-types-shared 0.1.1",
+]
+
+[[package]]
+name = "foreign-types"
+version = "0.5.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "d737d9aa519fb7b749cbc3b962edcf310a8dd1f4b67c91c4f83975dbdd17d965"
+dependencies = [
+ "foreign-types-macros",
+ "foreign-types-shared 0.3.1",
+]
+
+[[package]]
+name = "foreign-types-macros"
+version = "0.2.3"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "1a5c6c585bc94aaf2c7b51dd4c2ba22680844aba4c687be581871a6f518c5742"
+dependencies = [
+ "proc-macro2",
+ "quote",
+ "syn 2.0.117",
+]
+
+[[package]]
+name = "foreign-types-shared"
+version = "0.1.1"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "00b0228411908ca8685dba7fc2cdd70ec9990a6e753e89b6ac91a84c40fbaf4b"
+
+[[package]]
+name = "foreign-types-shared"
+version = "0.3.1"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "aa9a19cbb55df58761df49b23516a86d432839add4af60fc256da840f66ed35b"
+
+[[package]]
+name = "form_urlencoded"
+version = "1.2.2"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "cb4cb245038516f5f85277875cdaa4f7d2c9a0fa0468de06ed190163b1581fcf"
+dependencies = [
+ "percent-encoding",
+]
+
+[[package]]
+name = "futf"
+version = "0.1.5"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "df420e2e84819663797d1ec6544b13c5be84629e7bb00dc960d6917db2987843"
+dependencies = [
+ "mac",
+ "new_debug_unreachable",
+]
+
+[[package]]
+name = "futures-channel"
+version = "0.3.32"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "07bbe89c50d7a535e539b8c17bc0b49bdb77747034daa8087407d655f3f7cc1d"
+dependencies = [
+ "futures-core",
+]
+
+[[package]]
+name = "futures-core"
+version = "0.3.32"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "7e3450815272ef58cec6d564423f6e755e25379b217b0bc688e295ba24df6b1d"
+
+[[package]]
+name = "futures-executor"
+version = "0.3.32"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "baf29c38818342a3b26b5b923639e7b1f4a61fc5e76102d4b1981c6dc7a7579d"
+dependencies = [
+ "futures-core",
+ "futures-task",
+ "futures-util",
+]
+
+[[package]]
+name = "futures-io"
+version = "0.3.32"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "cecba35d7ad927e23624b22ad55235f2239cfa44fd10428eecbeba6d6a717718"
+
+[[package]]
+name = "futures-lite"
+version = "2.6.1"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "f78e10609fe0e0b3f4157ffab1876319b5b0db102a2c60dc4626306dc46b44ad"
+dependencies = [
+ "fastrand",
+ "futures-core",
+ "futures-io",
+ "parking",
+ "pin-project-lite",
+]
+
+[[package]]
+name = "futures-macro"
+version = "0.3.32"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "e835b70203e41293343137df5c0664546da5745f82ec9b84d40be8336958447b"
+dependencies = [
+ "proc-macro2",
+ "quote",
+ "syn 2.0.117",
+]
+
+[[package]]
+name = "futures-sink"
+version = "0.3.32"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "c39754e157331b013978ec91992bde1ac089843443c49cbc7f46150b0fad0893"
+
+[[package]]
+name = "futures-task"
+version = "0.3.32"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "037711b3d59c33004d3856fbdc83b99d4ff37a24768fa1be9ce3538a1cde4393"
+
+[[package]]
+name = "futures-util"
+version = "0.3.32"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "389ca41296e6190b48053de0321d02a77f32f8a5d2461dd38762c0593805c6d6"
+dependencies = [
+ "futures-core",
+ "futures-io",
+ "futures-macro",
+ "futures-sink",
+ "futures-task",
+ "memchr",
+ "pin-project-lite",
+ "slab",
+]
+
+[[package]]
+name = "fxhash"
+version = "0.2.1"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "c31b6d751ae2c7f11320402d34e41349dd1016f8d5d45e48c4312bc8625af50c"
+dependencies = [
+ "byteorder",
+]
+
+[[package]]
+name = "gdk"
+version = "0.18.2"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "d9f245958c627ac99d8e529166f9823fb3b838d1d41fd2b297af3075093c2691"
+dependencies = [
+ "cairo-rs",
+ "gdk-pixbuf",
+ "gdk-sys",
+ "gio",
+ "glib",
+ "libc",
+ "pango",
+]
+
+[[package]]
+name = "gdk-pixbuf"
+version = "0.18.5"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "50e1f5f1b0bfb830d6ccc8066d18db35c487b1b2b1e8589b5dfe9f07e8defaec"
+dependencies = [
+ "gdk-pixbuf-sys",
+ "gio",
+ "glib",
+ "libc",
+ "once_cell",
+]
+
+[[package]]
+name = "gdk-pixbuf-sys"
+version = "0.18.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "3f9839ea644ed9c97a34d129ad56d38a25e6756f99f3a88e15cd39c20629caf7"
+dependencies = [
+ "gio-sys",
+ "glib-sys",
+ "gobject-sys",
+ "libc",
+ "system-deps",
+]
+
+[[package]]
+name = "gdk-sys"
+version = "0.18.2"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "5c2d13f38594ac1e66619e188c6d5a1adb98d11b2fcf7894fc416ad76aa2f3f7"
+dependencies = [
+ "cairo-sys-rs",
+ "gdk-pixbuf-sys",
+ "gio-sys",
+ "glib-sys",
+ "gobject-sys",
+ "libc",
+ "pango-sys",
+ "pkg-config",
+ "system-deps",
+]
+
+[[package]]
+name = "gdkwayland-sys"
+version = "0.18.2"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "140071d506d223f7572b9f09b5e155afbd77428cd5cc7af8f2694c41d98dfe69"
+dependencies = [
+ "gdk-sys",
+ "glib-sys",
+ "gobject-sys",
+ "libc",
+ "pkg-config",
+ "system-deps",
+]
+
+[[package]]
+name = "gdkx11"
+version = "0.18.2"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "3caa00e14351bebbc8183b3c36690327eb77c49abc2268dd4bd36b856db3fbfe"
+dependencies = [
+ "gdk",
+ "gdkx11-sys",
+ "gio",
+ "glib",
+ "libc",
+ "x11",
+]
+
+[[package]]
+name = "gdkx11-sys"
+version = "0.18.2"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "6e2e7445fe01ac26f11601db260dd8608fe172514eb63b3b5e261ea6b0f4428d"
+dependencies = [
+ "gdk-sys",
+ "glib-sys",
+ "libc",
+ "system-deps",
+ "x11",
+]
+
+[[package]]
+name = "generic-array"
+version = "0.14.7"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "85649ca51fd72272d7821adaf274ad91c288277713d9c18820d8499a7ff69e9a"
+dependencies = [
+ "typenum",
+ "version_check",
+]
+
+[[package]]
+name = "getrandom"
+version = "0.1.16"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "8fc3cb4d91f53b50155bdcfd23f6a4c39ae1969c2ae85982b135750cccaf5fce"
+dependencies = [
+ "cfg-if",
+ "libc",
+ "wasi 0.9.0+wasi-snapshot-preview1",
+]
+
+[[package]]
+name = "getrandom"
+version = "0.2.17"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "ff2abc00be7fca6ebc474524697ae276ad847ad0a6b3faa4bcb027e9a4614ad0"
+dependencies = [
+ "cfg-if",
+ "libc",
+ "wasi 0.11.1+wasi-snapshot-preview1",
+]
+
+[[package]]
+name = "getrandom"
+version = "0.3.4"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "899def5c37c4fd7b2664648c28120ecec138e4d395b459e5ca34f9cce2dd77fd"
+dependencies = [
+ "cfg-if",
+ "libc",
+ "r-efi 5.3.0",
+ "wasip2",
+]
+
+[[package]]
+name = "getrandom"
+version = "0.4.2"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "0de51e6874e94e7bf76d726fc5d13ba782deca734ff60d5bb2fb2607c7406555"
+dependencies = [
+ "cfg-if",
+ "libc",
+ "r-efi 6.0.0",
+ "rand_core 0.10.0",
+ "wasip2",
+ "wasip3",
+]
+
+[[package]]
+name = "gio"
+version = "0.18.4"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "d4fc8f532f87b79cbc51a79748f16a6828fb784be93145a322fa14d06d354c73"
+dependencies = [
+ "futures-channel",
+ "futures-core",
+ "futures-io",
+ "futures-util",
+ "gio-sys",
+ "glib",
+ "libc",
+ "once_cell",
+ "pin-project-lite",
+ "smallvec",
+ "thiserror 1.0.69",
+]
+
+[[package]]
+name = "gio-sys"
+version = "0.18.1"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "37566df850baf5e4cb0dfb78af2e4b9898d817ed9263d1090a2df958c64737d2"
+dependencies = [
+ "glib-sys",
+ "gobject-sys",
+ "libc",
+ "system-deps",
+ "winapi",
+]
+
+[[package]]
+name = "glib"
+version = "0.18.5"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "233daaf6e83ae6a12a52055f568f9d7cf4671dabb78ff9560ab6da230ce00ee5"
+dependencies = [
+ "bitflags 2.11.0",
+ "futures-channel",
+ "futures-core",
+ "futures-executor",
+ "futures-task",
+ "futures-util",
+ "gio-sys",
+ "glib-macros",
+ "glib-sys",
+ "gobject-sys",
+ "libc",
+ "memchr",
+ "once_cell",
+ "smallvec",
+ "thiserror 1.0.69",
+]
+
+[[package]]
+name = "glib-macros"
+version = "0.18.5"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "0bb0228f477c0900c880fd78c8759b95c7636dbd7842707f49e132378aa2acdc"
+dependencies = [
+ "heck 0.4.1",
+ "proc-macro-crate 2.0.2",
+ "proc-macro-error",
+ "proc-macro2",
+ "quote",
+ "syn 2.0.117",
+]
+
+[[package]]
+name = "glib-sys"
+version = "0.18.1"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "063ce2eb6a8d0ea93d2bf8ba1957e78dbab6be1c2220dd3daca57d5a9d869898"
+dependencies = [
+ "libc",
+ "system-deps",
+]
+
+[[package]]
+name = "glob"
+version = "0.3.3"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "0cc23270f6e1808e30a928bdc84dea0b9b4136a8bc82338574f23baf47bbd280"
+
+[[package]]
+name = "gobject-sys"
+version = "0.18.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "0850127b514d1c4a4654ead6dedadb18198999985908e6ffe4436f53c785ce44"
+dependencies = [
+ "glib-sys",
+ "libc",
+ "system-deps",
+]
+
+[[package]]
+name = "gtk"
+version = "0.18.2"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "fd56fb197bfc42bd5d2751f4f017d44ff59fbb58140c6b49f9b3b2bdab08506a"
+dependencies = [
+ "atk",
+ "cairo-rs",
+ "field-offset",
+ "futures-channel",
+ "gdk",
+ "gdk-pixbuf",
+ "gio",
+ "glib",
+ "gtk-sys",
+ "gtk3-macros",
+ "libc",
+ "pango",
+ "pkg-config",
+]
+
+[[package]]
+name = "gtk-sys"
+version = "0.18.2"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "8f29a1c21c59553eb7dd40e918be54dccd60c52b049b75119d5d96ce6b624414"
+dependencies = [
+ "atk-sys",
+ "cairo-sys-rs",
+ "gdk-pixbuf-sys",
+ "gdk-sys",
+ "gio-sys",
+ "glib-sys",
+ "gobject-sys",
+ "libc",
+ "pango-sys",
+ "system-deps",
+]
+
+[[package]]
+name = "gtk3-macros"
+version = "0.18.2"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "52ff3c5b21f14f0736fed6dcfc0bfb4225ebf5725f3c0209edeec181e4d73e9d"
+dependencies = [
+ "proc-macro-crate 1.3.1",
+ "proc-macro-error",
+ "proc-macro2",
+ "quote",
+ "syn 2.0.117",
+]
+
+[[package]]
+name = "h2"
+version = "0.4.13"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "2f44da3a8150a6703ed5d34e164b875fd14c2cdab9af1252a9a1020bde2bdc54"
+dependencies = [
+ "atomic-waker",
+ "bytes",
+ "fnv",
+ "futures-core",
+ "futures-sink",
+ "http",
+ "indexmap 2.13.0",
+ "slab",
+ "tokio",
+ "tokio-util",
+ "tracing",
+]
+
+[[package]]
+name = "hashbrown"
+version = "0.12.3"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "8a9ee70c43aaf417c914396645a0fa852624801b24ebb7ae78fe8272889ac888"
+
+[[package]]
+name = "hashbrown"
+version = "0.15.5"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "9229cfe53dfd69f0609a49f65461bd93001ea1ef889cd5529dd176593f5338a1"
+dependencies = [
+ "foldhash 0.1.5",
+]
+
+[[package]]
+name = "hashbrown"
+version = "0.16.1"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "841d1cc9bed7f9236f321df977030373f4a4163ae1a7dbfe1a51a2c1a51d9100"
+
+[[package]]
+name = "heck"
+version = "0.4.1"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "95505c38b4572b2d910cecb0281560f54b440a19336cbbcb27bf6ce6adc6f5a8"
+
+[[package]]
+name = "heck"
+version = "0.5.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "2304e00983f87ffb38b55b444b5e3b60a884b5d30c0fca7d82fe33449bbe55ea"
+
+[[package]]
+name = "hermit-abi"
+version = "0.5.2"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "fc0fef456e4baa96da950455cd02c081ca953b141298e41db3fc7e36b1da849c"
+
+[[package]]
+name = "hex"
+version = "0.4.3"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "7f24254aa9a54b5c858eaee2f5bccdb46aaf0e486a595ed5fd8f86ba55232a70"
+
+[[package]]
+name = "home"
+version = "0.5.12"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "cc627f471c528ff0c4a49e1d5e60450c8f6461dd6d10ba9dcd3a61d3dff7728d"
+dependencies = [
+ "windows-sys 0.61.2",
+]
+
+[[package]]
+name = "html5ever"
+version = "0.29.1"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "3b7410cae13cbc75623c98ac4cbfd1f0bedddf3227afc24f370cf0f50a44a11c"
+dependencies = [
+ "log",
+ "mac",
+ "markup5ever 0.14.1",
+ "match_token",
+]
+
+[[package]]
+name = "html5ever"
+version = "0.38.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "1054432bae2f14e0061e33d23402fbaa67a921d319d56adc6bcf887ddad1cbc2"
+dependencies = [
+ "log",
+ "markup5ever 0.38.0",
+]
+
+[[package]]
+name = "http"
+version = "1.4.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "e3ba2a386d7f85a81f119ad7498ebe444d2e22c2af0b86b069416ace48b3311a"
+dependencies = [
+ "bytes",
+ "itoa",
+]
+
+[[package]]
+name = "http-body"
+version = "1.0.1"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "1efedce1fb8e6913f23e0c92de8e62cd5b772a67e7b3946df930a62566c93184"
+dependencies = [
+ "bytes",
+ "http",
+]
+
+[[package]]
+name = "http-body-util"
+version = "0.1.3"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "b021d93e26becf5dc7e1b75b1bed1fd93124b374ceb73f43d4d4eafec896a64a"
+dependencies = [
+ "bytes",
+ "futures-core",
+ "http",
+ "http-body",
+ "pin-project-lite",
+]
+
+[[package]]
+name = "httparse"
+version = "1.10.1"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "6dbf3de79e51f3d586ab4cb9d5c3e2c14aa28ed23d180cf89b4df0454a69cc87"
+
+[[package]]
+name = "hyper"
+version = "1.8.1"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "2ab2d4f250c3d7b1c9fcdff1cece94ea4e2dfbec68614f7b87cb205f24ca9d11"
+dependencies = [
+ "atomic-waker",
+ "bytes",
+ "futures-channel",
+ "futures-core",
+ "h2",
+ "http",
+ "http-body",
+ "httparse",
+ "itoa",
+ "pin-project-lite",
+ "pin-utils",
+ "smallvec",
+ "tokio",
+ "want",
+]
+
+[[package]]
+name = "hyper-rustls"
+version = "0.27.7"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "e3c93eb611681b207e1fe55d5a71ecf91572ec8a6705cdb6857f7d8d5242cf58"
+dependencies = [
+ "http",
+ "hyper",
+ "hyper-util",
+ "rustls",
+ "rustls-pki-types",
+ "tokio",
+ "tokio-rustls",
+ "tower-service",
+]
+
+[[package]]
+name = "hyper-tls"
+version = "0.6.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "70206fc6890eaca9fde8a0bf71caa2ddfc9fe045ac9e5c70df101a7dbde866e0"
+dependencies = [
+ "bytes",
+ "http-body-util",
+ "hyper",
+ "hyper-util",
+ "native-tls",
+ "tokio",
+ "tokio-native-tls",
+ "tower-service",
+]
+
+[[package]]
+name = "hyper-util"
+version = "0.1.20"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "96547c2556ec9d12fb1578c4eaf448b04993e7fb79cbaad930a656880a6bdfa0"
+dependencies = [
+ "base64 0.22.1",
+ "bytes",
+ "futures-channel",
+ "futures-util",
+ "http",
+ "http-body",
+ "hyper",
+ "ipnet",
+ "libc",
+ "percent-encoding",
+ "pin-project-lite",
+ "socket2",
+ "system-configuration",
+ "tokio",
+ "tower-service",
+ "tracing",
+ "windows-registry",
+]
+
+[[package]]
+name = "iana-time-zone"
+version = "0.1.65"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "e31bc9ad994ba00e440a8aa5c9ef0ec67d5cb5e5cb0cc7f8b744a35b389cc470"
+dependencies = [
+ "android_system_properties",
+ "core-foundation-sys",
+ "iana-time-zone-haiku",
+ "js-sys",
+ "log",
+ "wasm-bindgen",
+ "windows-core 0.62.2",
+]
+
+[[package]]
+name = "iana-time-zone-haiku"
+version = "0.1.2"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "f31827a206f56af32e590ba56d5d2d085f558508192593743f16b2306495269f"
+dependencies = [
+ "cc",
+]
+
+[[package]]
+name = "ico"
+version = "0.5.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "3e795dff5605e0f04bff85ca41b51a96b83e80b281e96231bcaaf1ac35103371"
+dependencies = [
+ "byteorder",
+ "png",
+]
+
+[[package]]
+name = "icu_collections"
+version = "2.1.1"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "4c6b649701667bbe825c3b7e6388cb521c23d88644678e83c0c4d0a621a34b43"
+dependencies = [
+ "displaydoc",
+ "potential_utf",
+ "yoke",
+ "zerofrom",
+ "zerovec",
+]
+
+[[package]]
+name = "icu_locale_core"
+version = "2.1.1"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "edba7861004dd3714265b4db54a3c390e880ab658fec5f7db895fae2046b5bb6"
+dependencies = [
+ "displaydoc",
+ "litemap",
+ "tinystr",
+ "writeable",
+ "zerovec",
+]
+
+[[package]]
+name = "icu_normalizer"
+version = "2.1.1"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "5f6c8828b67bf8908d82127b2054ea1b4427ff0230ee9141c54251934ab1b599"
+dependencies = [
+ "icu_collections",
+ "icu_normalizer_data",
+ "icu_properties",
+ "icu_provider",
+ "smallvec",
+ "zerovec",
+]
+
+[[package]]
+name = "icu_normalizer_data"
+version = "2.1.1"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "7aedcccd01fc5fe81e6b489c15b247b8b0690feb23304303a9e560f37efc560a"
+
+[[package]]
+name = "icu_properties"
+version = "2.1.2"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "020bfc02fe870ec3a66d93e677ccca0562506e5872c650f893269e08615d74ec"
+dependencies = [
+ "icu_collections",
+ "icu_locale_core",
+ "icu_properties_data",
+ "icu_provider",
+ "zerotrie",
+ "zerovec",
+]
+
+[[package]]
+name = "icu_properties_data"
+version = "2.1.2"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "616c294cf8d725c6afcd8f55abc17c56464ef6211f9ed59cccffe534129c77af"
+
+[[package]]
+name = "icu_provider"
+version = "2.1.1"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "85962cf0ce02e1e0a629cc34e7ca3e373ce20dda4c4d7294bbd0bf1fdb59e614"
+dependencies = [
+ "displaydoc",
+ "icu_locale_core",
+ "writeable",
+ "yoke",
+ "zerofrom",
+ "zerotrie",
+ "zerovec",
+]
+
+[[package]]
+name = "id-arena"
+version = "2.3.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "3d3067d79b975e8844ca9eb072e16b31c3c1c36928edf9c6789548c524d0d954"
+
+[[package]]
+name = "ident_case"
+version = "1.0.1"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "b9e0384b61958566e926dc50660321d12159025e767c18e043daf26b70104c39"
+
+[[package]]
+name = "idna"
+version = "1.1.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "3b0875f23caa03898994f6ddc501886a45c7d3d62d04d2d90788d47be1b1e4de"
+dependencies = [
+ "idna_adapter",
+ "smallvec",
+ "utf8_iter",
+]
+
+[[package]]
+name = "idna_adapter"
+version = "1.2.1"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "3acae9609540aa318d1bc588455225fb2085b9ed0c4f6bd0d9d5bcd86f1a0344"
+dependencies = [
+ "icu_normalizer",
+ "icu_properties",
+]
+
+[[package]]
+name = "indexmap"
+version = "1.9.3"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "bd070e393353796e801d209ad339e89596eb4c8d430d18ede6a1cced8fafbd99"
+dependencies = [
+ "autocfg",
+ "hashbrown 0.12.3",
+ "serde",
+]
+
+[[package]]
+name = "indexmap"
+version = "2.13.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "7714e70437a7dc3ac8eb7e6f8df75fd8eb422675fc7678aff7364301092b1017"
+dependencies = [
+ "equivalent",
+ "hashbrown 0.16.1",
+ "serde",
+ "serde_core",
+]
+
+[[package]]
+name = "infer"
+version = "0.19.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "a588916bfdfd92e71cacef98a63d9b1f0d74d6599980d11894290e7ddefffcf7"
+dependencies = [
+ "cfb",
+]
+
+[[package]]
+name = "ipnet"
+version = "2.12.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "d98f6fed1fde3f8c21bc40a1abb88dd75e67924f9cffc3ef95607bad8017f8e2"
+
+[[package]]
+name = "iri-string"
+version = "0.7.10"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "c91338f0783edbd6195decb37bae672fd3b165faffb89bf7b9e6942f8b1a731a"
+dependencies = [
+ "memchr",
+ "serde",
+]
+
+[[package]]
+name = "is-docker"
+version = "0.2.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "928bae27f42bc99b60d9ac7334e3a21d10ad8f1835a4e12ec3ec0464765ed1b3"
+dependencies = [
+ "once_cell",
+]
+
+[[package]]
+name = "is-wsl"
+version = "0.4.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "173609498df190136aa7dea1a91db051746d339e18476eed5ca40521f02d7aa5"
+dependencies = [
+ "is-docker",
+ "once_cell",
+]
+
+[[package]]
+name = "itoa"
+version = "1.0.18"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "8f42a60cbdf9a97f5d2305f08a87dc4e09308d1276d28c869c684d7777685682"
+
+[[package]]
+name = "javascriptcore-rs"
+version = "1.1.2"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "ca5671e9ffce8ffba57afc24070e906da7fc4b1ba66f2cabebf61bf2ea257fcc"
+dependencies = [
+ "bitflags 1.3.2",
+ "glib",
+ "javascriptcore-rs-sys",
+]
+
+[[package]]
+name = "javascriptcore-rs-sys"
+version = "1.1.1"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "af1be78d14ffa4b75b66df31840478fef72b51f8c2465d4ca7c194da9f7a5124"
+dependencies = [
+ "glib-sys",
+ "gobject-sys",
+ "libc",
+ "system-deps",
+]
+
+[[package]]
+name = "jni"
+version = "0.21.1"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "1a87aa2bb7d2af34197c04845522473242e1aa17c12f4935d5856491a7fb8c97"
+dependencies = [
+ "cesu8",
+ "cfg-if",
+ "combine",
+ "jni-sys 0.3.1",
+ "log",
+ "thiserror 1.0.69",
+ "walkdir",
+ "windows-sys 0.45.0",
+]
+
+[[package]]
+name = "jni-sys"
+version = "0.3.1"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "41a652e1f9b6e0275df1f15b32661cf0d4b78d4d87ddec5e0c3c20f097433258"
+dependencies = [
+ "jni-sys 0.4.1",
+]
+
+[[package]]
+name = "jni-sys"
+version = "0.4.1"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "c6377a88cb3910bee9b0fa88d4f42e1d2da8e79915598f65fb0c7ee14c878af2"
+dependencies = [
+ "jni-sys-macros",
+]
+
+[[package]]
+name = "jni-sys-macros"
+version = "0.4.1"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "38c0b942f458fe50cdac086d2f946512305e5631e720728f2a61aabcd47a6264"
+dependencies = [
+ "quote",
+ "syn 2.0.117",
+]
+
+[[package]]
+name = "js-sys"
+version = "0.3.91"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "b49715b7073f385ba4bc528e5747d02e66cb39c6146efb66b781f131f0fb399c"
+dependencies = [
+ "once_cell",
+ "wasm-bindgen",
+]
+
+[[package]]
+name = "json-patch"
+version = "3.0.1"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "863726d7afb6bc2590eeff7135d923545e5e964f004c2ccf8716c25e70a86f08"
+dependencies = [
+ "jsonptr",
+ "serde",
+ "serde_json",
+ "thiserror 1.0.69",
+]
+
+[[package]]
+name = "jsonptr"
+version = "0.6.3"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "5dea2b27dd239b2556ed7a25ba842fe47fd602e7fc7433c2a8d6106d4d9edd70"
+dependencies = [
+ "serde",
+ "serde_json",
+]
+
+[[package]]
+name = "keyboard-types"
+version = "0.7.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "b750dcadc39a09dbadd74e118f6dd6598df77fa01df0cfcdc52c28dece74528a"
+dependencies = [
+ "bitflags 2.11.0",
+ "serde",
+ "unicode-segmentation",
+]
+
+[[package]]
+name = "kuchikiki"
+version = "0.8.8-speedreader"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "02cb977175687f33fa4afa0c95c112b987ea1443e5a51c8f8ff27dc618270cc2"
+dependencies = [
+ "cssparser 0.29.6",
+ "html5ever 0.29.1",
+ "indexmap 2.13.0",
+ "selectors 0.24.0",
+]
+
+[[package]]
+name = "leb128fmt"
+version = "0.1.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "09edd9e8b54e49e587e4f6295a7d29c3ea94d469cb40ab8ca70b288248a81db2"
+
+[[package]]
+name = "libappindicator"
+version = "0.9.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "03589b9607c868cc7ae54c0b2a22c8dc03dd41692d48f2d7df73615c6a95dc0a"
+dependencies = [
+ "glib",
+ "gtk",
+ "gtk-sys",
+ "libappindicator-sys",
+ "log",
+]
+
+[[package]]
+name = "libappindicator-sys"
+version = "0.9.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "6e9ec52138abedcc58dc17a7c6c0c00a2bdb4f3427c7f63fa97fd0d859155caf"
+dependencies = [
+ "gtk-sys",
+ "libloading",
+ "once_cell",
+]
+
+[[package]]
+name = "libc"
+version = "0.2.183"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "b5b646652bf6661599e1da8901b3b9522896f01e736bad5f723fe7a3a27f899d"
+
+[[package]]
+name = "libloading"
+version = "0.7.4"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "b67380fd3b2fbe7527a606e18729d21c6f3951633d0500574c4dc22d2d638b9f"
+dependencies = [
+ "cfg-if",
+ "winapi",
+]
+
+[[package]]
+name = "libredox"
+version = "0.1.14"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "1744e39d1d6a9948f4f388969627434e31128196de472883b39f148769bfe30a"
+dependencies = [
+ "bitflags 2.11.0",
+ "libc",
+ "plain",
+ "redox_syscall 0.7.3",
+]
+
+[[package]]
+name = "linux-raw-sys"
+version = "0.12.1"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "32a66949e030da00e8c7d4434b251670a91556f4144941d37452769c25d58a53"
+
+[[package]]
+name = "litemap"
+version = "0.8.1"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "6373607a59f0be73a39b6fe456b8192fcc3585f602af20751600e974dd455e77"
+
+[[package]]
+name = "lock_api"
+version = "0.4.14"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "224399e74b87b5f3557511d98dff8b14089b3dadafcab6bb93eab67d3aace965"
+dependencies = [
+ "scopeguard",
+]
+
+[[package]]
+name = "log"
+version = "0.4.29"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "5e5032e24019045c762d3c0f28f5b6b8bbf38563a65908389bf7978758920897"
+
+[[package]]
+name = "mac"
+version = "0.1.1"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "c41e0c4fef86961ac6d6f8a82609f55f31b05e4fce149ac5710e439df7619ba4"
+
+[[package]]
+name = "markup5ever"
+version = "0.14.1"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "c7a7213d12e1864c0f002f52c2923d4556935a43dec5e71355c2760e0f6e7a18"
+dependencies = [
+ "log",
+ "phf 0.11.3",
+ "phf_codegen 0.11.3",
+ "string_cache 0.8.9",
+ "string_cache_codegen 0.5.4",
+ "tendril 0.4.3",
+]
+
+[[package]]
+name = "markup5ever"
+version = "0.38.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "8983d30f2915feeaaab2d6babdd6bc7e9ed1a00b66b5e6d74df19aa9c0e91862"
+dependencies = [
+ "log",
+ "tendril 0.5.0",
+ "web_atoms",
+]
+
+[[package]]
+name = "match_token"
+version = "0.1.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "88a9689d8d44bf9964484516275f5cd4c9b59457a6940c1d5d0ecbb94510a36b"
+dependencies = [
+ "proc-macro2",
+ "quote",
+ "syn 2.0.117",
+]
+
+[[package]]
+name = "matches"
+version = "0.1.10"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "2532096657941c2fea9c289d370a250971c689d4f143798ff67113ec042024a5"
+
+[[package]]
+name = "memchr"
+version = "2.8.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "f8ca58f447f06ed17d5fc4043ce1b10dd205e060fb3ce5b979b8ed8e59ff3f79"
+
+[[package]]
+name = "memoffset"
+version = "0.9.1"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "488016bfae457b036d996092f6cb448677611ce4449e970ceaf42695203f218a"
+dependencies = [
+ "autocfg",
+]
+
+[[package]]
+name = "mime"
+version = "0.3.17"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "6877bb514081ee2a7ff5ef9de3281f14a4dd4bceac4c09388074a6b5df8a139a"
+
+[[package]]
+name = "minisign-verify"
+version = "0.2.5"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "22f9645cb765ea72b8111f36c522475d2daa0d22c957a9826437e97534bc4e9e"
+
+[[package]]
+name = "miniz_oxide"
+version = "0.8.9"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "1fa76a2c86f704bdb222d66965fb3d63269ce38518b83cb0575fca855ebb6316"
+dependencies = [
+ "adler2",
+ "simd-adler32",
+]
+
+[[package]]
+name = "mio"
+version = "1.1.1"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "a69bcab0ad47271a0234d9422b131806bf3968021e5dc9328caf2d4cd58557fc"
+dependencies = [
+ "libc",
+ "wasi 0.11.1+wasi-snapshot-preview1",
+ "windows-sys 0.61.2",
+]
+
+[[package]]
+name = "muda"
+version = "0.17.1"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "01c1738382f66ed56b3b9c8119e794a2e23148ac8ea214eda86622d4cb9d415a"
+dependencies = [
+ "crossbeam-channel",
+ "dpi",
+ "gtk",
+ "keyboard-types",
+ "objc2",
+ "objc2-app-kit",
+ "objc2-core-foundation",
+ "objc2-foundation",
+ "once_cell",
+ "png",
+ "serde",
+ "thiserror 2.0.18",
+ "windows-sys 0.60.2",
+]
+
+[[package]]
+name = "native-tls"
+version = "0.2.18"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "465500e14ea162429d264d44189adc38b199b62b1c21eea9f69e4b73cb03bbf2"
+dependencies = [
+ "libc",
+ "log",
+ "openssl",
+ "openssl-probe",
+ "openssl-sys",
+ "schannel",
+ "security-framework",
+ "security-framework-sys",
+ "tempfile",
+]
+
+[[package]]
+name = "ndk"
+version = "0.9.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "c3f42e7bbe13d351b6bead8286a43aac9534b82bd3cc43e47037f012ebfd62d4"
+dependencies = [
+ "bitflags 2.11.0",
+ "jni-sys 0.3.1",
+ "log",
+ "ndk-sys",
+ "num_enum",
+ "raw-window-handle",
+ "thiserror 1.0.69",
+]
+
+[[package]]
+name = "ndk-context"
+version = "0.1.1"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "27b02d87554356db9e9a873add8782d4ea6e3e58ea071a9adb9a2e8ddb884a8b"
+
+[[package]]
+name = "ndk-sys"
+version = "0.6.0+11769913"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "ee6cda3051665f1fb8d9e08fc35c96d5a244fb1be711a03b71118828afc9a873"
+dependencies = [
+ "jni-sys 0.3.1",
+]
+
+[[package]]
+name = "new_debug_unreachable"
+version = "1.0.6"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "650eef8c711430f1a879fdd01d4745a7deea475becfb90269c06775983bbf086"
+
+[[package]]
+name = "nix"
+version = "0.31.2"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "5d6d0705320c1e6ba1d912b5e37cf18071b6c2e9b7fa8215a1e8a7651966f5d3"
+dependencies = [
+ "bitflags 2.11.0",
+ "cfg-if",
+ "cfg_aliases",
+ "libc",
+]
+
+[[package]]
+name = "nodrop"
+version = "0.1.14"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "72ef4a56884ca558e5ddb05a1d1e7e1bfd9a68d9ed024c21704cc98872dae1bb"
+
+[[package]]
+name = "num-conv"
+version = "0.2.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "cf97ec579c3c42f953ef76dbf8d55ac91fb219dde70e49aa4a6b7d74e9919050"
+
+[[package]]
+name = "num-traits"
+version = "0.2.19"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "071dfc062690e90b734c0b2273ce72ad0ffa95f0c74596bc250dcfd960262841"
+dependencies = [
+ "autocfg",
+]
+
+[[package]]
+name = "num_enum"
+version = "0.7.6"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "5d0bca838442ec211fa11de3a8b0e0e8f3a4522575b5c4c06ed722e005036f26"
+dependencies = [
+ "num_enum_derive",
+ "rustversion",
+]
+
+[[package]]
+name = "num_enum_derive"
+version = "0.7.6"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "680998035259dcfcafe653688bf2aa6d3e2dc05e98be6ab46afb089dc84f1df8"
+dependencies = [
+ "proc-macro-crate 3.5.0",
+ "proc-macro2",
+ "quote",
+ "syn 2.0.117",
+]
+
+[[package]]
+name = "num_threads"
+version = "0.1.7"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "5c7398b9c8b70908f6371f47ed36737907c87c52af34c268fed0bf0ceb92ead9"
+dependencies = [
+ "libc",
+]
+
+[[package]]
+name = "objc2"
+version = "0.6.4"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "3a12a8ed07aefc768292f076dc3ac8c48f3781c8f2d5851dd3d98950e8c5a89f"
+dependencies = [
+ "objc2-encode",
+ "objc2-exception-helper",
+]
+
+[[package]]
+name = "objc2-app-kit"
+version = "0.3.2"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "d49e936b501e5c5bf01fda3a9452ff86dc3ea98ad5f283e1455153142d97518c"
+dependencies = [
+ "bitflags 2.11.0",
+ "block2",
+ "objc2",
+ "objc2-core-foundation",
+ "objc2-foundation",
+]
+
+[[package]]
+name = "objc2-core-foundation"
+version = "0.3.2"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "2a180dd8642fa45cdb7dd721cd4c11b1cadd4929ce112ebd8b9f5803cc79d536"
+dependencies = [
+ "bitflags 2.11.0",
+ "dispatch2",
+ "objc2",
+]
+
+[[package]]
+name = "objc2-core-graphics"
+version = "0.3.2"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "e022c9d066895efa1345f8e33e584b9f958da2fd4cd116792e15e07e4720a807"
+dependencies = [
+ "bitflags 2.11.0",
+ "dispatch2",
+ "objc2",
+ "objc2-core-foundation",
+ "objc2-io-surface",
+]
+
+[[package]]
+name = "objc2-encode"
+version = "4.1.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "ef25abbcd74fb2609453eb695bd2f860d389e457f67dc17cafc8b8cbc89d0c33"
+
+[[package]]
+name = "objc2-exception-helper"
+version = "0.1.1"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "c7a1c5fbb72d7735b076bb47b578523aedc40f3c439bea6dfd595c089d79d98a"
+dependencies = [
+ "cc",
+]
+
+[[package]]
+name = "objc2-foundation"
+version = "0.3.2"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "e3e0adef53c21f888deb4fa59fc59f7eb17404926ee8a6f59f5df0fd7f9f3272"
+dependencies = [
+ "bitflags 2.11.0",
+ "block2",
+ "libc",
+ "objc2",
+ "objc2-core-foundation",
+]
+
+[[package]]
+name = "objc2-io-surface"
+version = "0.3.2"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "180788110936d59bab6bd83b6060ffdfffb3b922ba1396b312ae795e1de9d81d"
+dependencies = [
+ "bitflags 2.11.0",
+ "objc2",
+ "objc2-core-foundation",
+]
+
+[[package]]
+name = "objc2-osa-kit"
+version = "0.3.2"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "f112d1746737b0da274ef79a23aac283376f335f4095a083a267a082f21db0c0"
+dependencies = [
+ "bitflags 2.11.0",
+ "objc2",
+ "objc2-app-kit",
+ "objc2-foundation",
+]
+
+[[package]]
+name = "objc2-quartz-core"
+version = "0.3.2"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "96c1358452b371bf9f104e21ec536d37a650eb10f7ee379fff67d2e08d537f1f"
+dependencies = [
+ "bitflags 2.11.0",
+ "objc2",
+ "objc2-core-foundation",
+ "objc2-foundation",
+]
+
+[[package]]
+name = "objc2-ui-kit"
+version = "0.3.2"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "d87d638e33c06f577498cbcc50491496a3ed4246998a7fbba7ccb98b1e7eab22"
+dependencies = [
+ "bitflags 2.11.0",
+ "objc2",
+ "objc2-core-foundation",
+ "objc2-foundation",
+]
+
+[[package]]
+name = "objc2-web-kit"
+version = "0.3.2"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "b2e5aaab980c433cf470df9d7af96a7b46a9d892d521a2cbbb2f8a4c16751e7f"
+dependencies = [
+ "bitflags 2.11.0",
+ "block2",
+ "objc2",
+ "objc2-app-kit",
+ "objc2-core-foundation",
+ "objc2-foundation",
+]
+
+[[package]]
+name = "once_cell"
+version = "1.21.4"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "9f7c3e4beb33f85d45ae3e3a1792185706c8e16d043238c593331cc7cd313b50"
+
+[[package]]
+name = "open"
+version = "5.3.3"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "43bb73a7fa3799b198970490a51174027ba0d4ec504b03cd08caf513d40024bc"
+dependencies = [
+ "dunce",
+ "is-wsl",
+ "libc",
+ "pathdiff",
+]
+
+[[package]]
+name = "openssl"
+version = "0.10.76"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "951c002c75e16ea2c65b8c7e4d3d51d5530d8dfa7d060b4776828c88cfb18ecf"
+dependencies = [
+ "bitflags 2.11.0",
+ "cfg-if",
+ "foreign-types 0.3.2",
+ "libc",
+ "once_cell",
+ "openssl-macros",
+ "openssl-sys",
+]
+
+[[package]]
+name = "openssl-macros"
+version = "0.1.1"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "a948666b637a0f465e8564c73e89d4dde00d72d4d473cc972f390fc3dcee7d9c"
+dependencies = [
+ "proc-macro2",
+ "quote",
+ "syn 2.0.117",
+]
+
+[[package]]
+name = "openssl-probe"
+version = "0.2.1"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "7c87def4c32ab89d880effc9e097653c8da5d6ef28e6b539d313baaacfbafcbe"
+
+[[package]]
+name = "openssl-sys"
+version = "0.9.112"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "57d55af3b3e226502be1526dfdba67ab0e9c96fc293004e79576b2b9edb0dbdb"
+dependencies = [
+ "cc",
+ "libc",
+ "pkg-config",
+ "vcpkg",
+]
+
+[[package]]
+name = "option-ext"
+version = "0.2.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "04744f49eae99ab78e0d5c0b603ab218f515ea8cfe5a456d7629ad883a3b6e7d"
+
+[[package]]
+name = "ordered-stream"
+version = "0.2.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "9aa2b01e1d916879f73a53d01d1d6cee68adbb31d6d9177a8cfce093cced1d50"
+dependencies = [
+ "futures-core",
+ "pin-project-lite",
+]
+
+[[package]]
+name = "osakit"
+version = "0.3.1"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "732c71caeaa72c065bb69d7ea08717bd3f4863a4f451402fc9513e29dbd5261b"
+dependencies = [
+ "objc2",
+ "objc2-foundation",
+ "objc2-osa-kit",
+ "serde",
+ "serde_json",
+ "thiserror 2.0.18",
+]
+
+[[package]]
+name = "pango"
+version = "0.18.3"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "7ca27ec1eb0457ab26f3036ea52229edbdb74dee1edd29063f5b9b010e7ebee4"
+dependencies = [
+ "gio",
+ "glib",
+ "libc",
+ "once_cell",
+ "pango-sys",
+]
+
+[[package]]
+name = "pango-sys"
+version = "0.18.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "436737e391a843e5933d6d9aa102cb126d501e815b83601365a948a518555dc5"
+dependencies = [
+ "glib-sys",
+ "gobject-sys",
+ "libc",
+ "system-deps",
+]
+
+[[package]]
+name = "parking"
+version = "2.2.1"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "f38d5652c16fde515bb1ecef450ab0f6a219d619a7274976324d5e377f7dceba"
+
+[[package]]
+name = "parking_lot"
+version = "0.12.5"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "93857453250e3077bd71ff98b6a65ea6621a19bb0f559a85248955ac12c45a1a"
+dependencies = [
+ "lock_api",
+ "parking_lot_core",
+]
+
+[[package]]
+name = "parking_lot_core"
+version = "0.9.12"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "2621685985a2ebf1c516881c026032ac7deafcda1a2c9b7850dc81e3dfcb64c1"
+dependencies = [
+ "cfg-if",
+ "libc",
+ "redox_syscall 0.5.18",
+ "smallvec",
+ "windows-link 0.2.1",
+]
+
+[[package]]
+name = "pathdiff"
+version = "0.2.3"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "df94ce210e5bc13cb6651479fa48d14f601d9858cfe0467f43ae157023b938d3"
+
+[[package]]
+name = "percent-encoding"
+version = "2.3.2"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "9b4f627cb1b25917193a259e49bdad08f671f8d9708acfd5fe0a8c1455d87220"
+
+[[package]]
+name = "phf"
+version = "0.8.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "3dfb61232e34fcb633f43d12c58f83c1df82962dcdfa565a4e866ffc17dafe12"
+dependencies = [
+ "phf_shared 0.8.0",
+]
+
+[[package]]
+name = "phf"
+version = "0.10.1"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "fabbf1ead8a5bcbc20f5f8b939ee3f5b0f6f281b6ad3468b84656b658b455259"
+dependencies = [
+ "phf_macros 0.10.0",
+ "phf_shared 0.10.0",
+ "proc-macro-hack",
+]
+
+[[package]]
+name = "phf"
+version = "0.11.3"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "1fd6780a80ae0c52cc120a26a1a42c1ae51b247a253e4e06113d23d2c2edd078"
+dependencies = [
+ "phf_macros 0.11.3",
+ "phf_shared 0.11.3",
+]
+
+[[package]]
+name = "phf"
+version = "0.13.1"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "c1562dc717473dbaa4c1f85a36410e03c047b2e7df7f45ee938fbef64ae7fadf"
+dependencies = [
+ "phf_macros 0.13.1",
+ "phf_shared 0.13.1",
+ "serde",
+]
+
+[[package]]
+name = "phf_codegen"
+version = "0.8.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "cbffee61585b0411840d3ece935cce9cb6321f01c45477d30066498cd5e1a815"
+dependencies = [
+ "phf_generator 0.8.0",
+ "phf_shared 0.8.0",
+]
+
+[[package]]
+name = "phf_codegen"
+version = "0.11.3"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "aef8048c789fa5e851558d709946d6d79a8ff88c0440c587967f8e94bfb1216a"
+dependencies = [
+ "phf_generator 0.11.3",
+ "phf_shared 0.11.3",
+]
+
+[[package]]
+name = "phf_codegen"
+version = "0.13.1"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "49aa7f9d80421bca176ca8dbfebe668cc7a2684708594ec9f3c0db0805d5d6e1"
+dependencies = [
+ "phf_generator 0.13.1",
+ "phf_shared 0.13.1",
+]
+
+[[package]]
+name = "phf_generator"
+version = "0.8.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "17367f0cc86f2d25802b2c26ee58a7b23faeccf78a396094c13dced0d0182526"
+dependencies = [
+ "phf_shared 0.8.0",
+ "rand 0.7.3",
+]
+
+[[package]]
+name = "phf_generator"
+version = "0.10.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "5d5285893bb5eb82e6aaf5d59ee909a06a16737a8970984dd7746ba9283498d6"
+dependencies = [
+ "phf_shared 0.10.0",
+ "rand 0.8.5",
+]
+
+[[package]]
+name = "phf_generator"
+version = "0.11.3"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "3c80231409c20246a13fddb31776fb942c38553c51e871f8cbd687a4cfb5843d"
+dependencies = [
+ "phf_shared 0.11.3",
+ "rand 0.8.5",
+]
+
+[[package]]
+name = "phf_generator"
+version = "0.13.1"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "135ace3a761e564ec88c03a77317a7c6b80bb7f7135ef2544dbe054243b89737"
+dependencies = [
+ "fastrand",
+ "phf_shared 0.13.1",
+]
+
+[[package]]
+name = "phf_macros"
+version = "0.10.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "58fdf3184dd560f160dd73922bea2d5cd6e8f064bf4b13110abd81b03697b4e0"
+dependencies = [
+ "phf_generator 0.10.0",
+ "phf_shared 0.10.0",
+ "proc-macro-hack",
+ "proc-macro2",
+ "quote",
+ "syn 1.0.109",
+]
+
+[[package]]
+name = "phf_macros"
+version = "0.11.3"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "f84ac04429c13a7ff43785d75ad27569f2951ce0ffd30a3321230db2fc727216"
+dependencies = [
+ "phf_generator 0.11.3",
+ "phf_shared 0.11.3",
+ "proc-macro2",
+ "quote",
+ "syn 2.0.117",
+]
+
+[[package]]
+name = "phf_macros"
+version = "0.13.1"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "812f032b54b1e759ccd5f8b6677695d5268c588701effba24601f6932f8269ef"
+dependencies = [
+ "phf_generator 0.13.1",
+ "phf_shared 0.13.1",
+ "proc-macro2",
+ "quote",
+ "syn 2.0.117",
+]
+
+[[package]]
+name = "phf_shared"
+version = "0.8.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "c00cf8b9eafe68dde5e9eaa2cef8ee84a9336a47d566ec55ca16589633b65af7"
+dependencies = [
+ "siphasher 0.3.11",
+]
+
+[[package]]
+name = "phf_shared"
+version = "0.10.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "b6796ad771acdc0123d2a88dc428b5e38ef24456743ddb1744ed628f9815c096"
+dependencies = [
+ "siphasher 0.3.11",
+]
+
+[[package]]
+name = "phf_shared"
+version = "0.11.3"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "67eabc2ef2a60eb7faa00097bd1ffdb5bd28e62bf39990626a582201b7a754e5"
+dependencies = [
+ "siphasher 1.0.2",
+]
+
+[[package]]
+name = "phf_shared"
+version = "0.13.1"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "e57fef6bc5981e38c2ce2d63bfa546861309f875b8a75f092d1d54ae2d64f266"
+dependencies = [
+ "siphasher 1.0.2",
+]
+
+[[package]]
+name = "pin-project-lite"
+version = "0.2.17"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "a89322df9ebe1c1578d689c92318e070967d1042b512afbe49518723f4e6d5cd"
+
+[[package]]
+name = "pin-utils"
+version = "0.1.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "8b870d8c151b6f2fb93e84a13146138f05d02ed11c7e7c54f8826aaaf7c9f184"
+
+[[package]]
+name = "piper"
+version = "0.2.5"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "c835479a4443ded371d6c535cbfd8d31ad92c5d23ae9770a61bc155e4992a3c1"
+dependencies = [
+ "atomic-waker",
+ "fastrand",
+ "futures-io",
+]
+
+[[package]]
+name = "pkg-config"
+version = "0.3.32"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "7edddbd0b52d732b21ad9a5fab5c704c14cd949e5e9a1ec5929a24fded1b904c"
+
+[[package]]
+name = "plain"
+version = "0.2.3"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "b4596b6d070b27117e987119b4dac604f3c58cfb0b191112e24771b2faeac1a6"
+
+[[package]]
+name = "plist"
+version = "1.8.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "740ebea15c5d1428f910cd1a5f52cebf8d25006245ed8ade92702f4943d91e07"
+dependencies = [
+ "base64 0.22.1",
+ "indexmap 2.13.0",
+ "quick-xml",
+ "serde",
+ "time",
+]
+
+[[package]]
+name = "png"
+version = "0.17.16"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "82151a2fc869e011c153adc57cf2789ccb8d9906ce52c0b39a6b5697749d7526"
+dependencies = [
+ "bitflags 1.3.2",
+ "crc32fast",
+ "fdeflate",
+ "flate2",
+ "miniz_oxide",
+]
+
+[[package]]
+name = "polling"
+version = "3.11.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "5d0e4f59085d47d8241c88ead0f274e8a0cb551f3625263c05eb8dd897c34218"
+dependencies = [
+ "cfg-if",
+ "concurrent-queue",
+ "hermit-abi",
+ "pin-project-lite",
+ "rustix",
+ "windows-sys 0.61.2",
+]
+
+[[package]]
+name = "potential_utf"
+version = "0.1.4"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "b73949432f5e2a09657003c25bca5e19a0e9c84f8058ca374f49e0ebe605af77"
+dependencies = [
+ "zerovec",
+]
+
+[[package]]
+name = "powerfmt"
+version = "0.2.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "439ee305def115ba05938db6eb1644ff94165c5ab5e9420d1c1bcedbba909391"
+
+[[package]]
+name = "ppv-lite86"
+version = "0.2.21"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "85eae3c4ed2f50dcfe72643da4befc30deadb458a9b590d720cde2f2b1e97da9"
+dependencies = [
+ "zerocopy",
+]
+
+[[package]]
+name = "precomputed-hash"
+version = "0.1.1"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "925383efa346730478fb4838dbe9137d2a47675ad789c546d150a6e1dd4ab31c"
+
+[[package]]
+name = "prettyplease"
+version = "0.2.37"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "479ca8adacdd7ce8f1fb39ce9ecccbfe93a3f1344b3d0d97f20bc0196208f62b"
+dependencies = [
+ "proc-macro2",
+ "syn 2.0.117",
+]
+
+[[package]]
+name = "proc-macro-crate"
+version = "1.3.1"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "7f4c021e1093a56626774e81216a4ce732a735e5bad4868a03f3ed65ca0c3919"
+dependencies = [
+ "once_cell",
+ "toml_edit 0.19.15",
+]
+
+[[package]]
+name = "proc-macro-crate"
+version = "2.0.2"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "b00f26d3400549137f92511a46ac1cd8ce37cb5598a96d382381458b992a5d24"
+dependencies = [
+ "toml_datetime 0.6.3",
+ "toml_edit 0.20.2",
+]
+
+[[package]]
+name = "proc-macro-crate"
+version = "3.5.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "e67ba7e9b2b56446f1d419b1d807906278ffa1a658a8a5d8a39dcb1f5a78614f"
+dependencies = [
+ "toml_edit 0.25.5+spec-1.1.0",
+]
+
+[[package]]
+name = "proc-macro-error"
+version = "1.0.4"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "da25490ff9892aab3fcf7c36f08cfb902dd3e71ca0f9f9517bea02a73a5ce38c"
+dependencies = [
+ "proc-macro-error-attr",
+ "proc-macro2",
+ "quote",
+ "syn 1.0.109",
+ "version_check",
+]
+
+[[package]]
+name = "proc-macro-error-attr"
+version = "1.0.4"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "a1be40180e52ecc98ad80b184934baf3d0d29f979574e439af5a55274b35f869"
+dependencies = [
+ "proc-macro2",
+ "quote",
+ "version_check",
+]
+
+[[package]]
+name = "proc-macro-hack"
+version = "0.5.20+deprecated"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "dc375e1527247fe1a97d8b7156678dfe7c1af2fc075c9a4db3690ecd2a148068"
+
+[[package]]
+name = "proc-macro2"
+version = "1.0.106"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "8fd00f0bb2e90d81d1044c2b32617f68fcb9fa3bb7640c23e9c748e53fb30934"
+dependencies = [
+ "unicode-ident",
+]
+
+[[package]]
+name = "process-wrap"
+version = "9.1.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "2e842efad9119158434d193c6682e2ebee4b44d6ad801d7b349623b3f57cdf55"
+dependencies = [
+ "indexmap 2.13.0",
+ "nix",
+ "tracing",
+ "windows 0.62.2",
+]
+
+[[package]]
+name = "quick-xml"
+version = "0.38.4"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "b66c2058c55a409d601666cffe35f04333cf1013010882cec174a7467cd4e21c"
+dependencies = [
+ "memchr",
+]
+
+[[package]]
+name = "quote"
+version = "1.0.45"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "41f2619966050689382d2b44f664f4bc593e129785a36d6ee376ddf37259b924"
+dependencies = [
+ "proc-macro2",
+]
+
+[[package]]
+name = "r-efi"
+version = "5.3.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "69cdb34c158ceb288df11e18b4bd39de994f6657d83847bdffdbd7f346754b0f"
+
+[[package]]
+name = "r-efi"
+version = "6.0.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "f8dcc9c7d52a811697d2151c701e0d08956f92b0e24136cf4cf27b57a6a0d9bf"
+
+[[package]]
+name = "rand"
+version = "0.7.3"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "6a6b1679d49b24bbfe0c803429aa1874472f50d9b363131f0e89fc356b544d03"
+dependencies = [
+ "getrandom 0.1.16",
+ "libc",
+ "rand_chacha 0.2.2",
+ "rand_core 0.5.1",
+ "rand_hc",
+ "rand_pcg",
+]
+
+[[package]]
+name = "rand"
+version = "0.8.5"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "34af8d1a0e25924bc5b7c43c079c942339d8f0a8b57c39049bef581b46327404"
+dependencies = [
+ "libc",
+ "rand_chacha 0.3.1",
+ "rand_core 0.6.4",
+]
+
+[[package]]
+name = "rand"
+version = "0.10.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "bc266eb313df6c5c09c1c7b1fbe2510961e5bcd3add930c1e31f7ed9da0feff8"
+dependencies = [
+ "chacha20",
+ "getrandom 0.4.2",
+ "rand_core 0.10.0",
+]
+
+[[package]]
+name = "rand_chacha"
+version = "0.2.2"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "f4c8ed856279c9737206bf725bf36935d8666ead7aa69b52be55af369d193402"
+dependencies = [
+ "ppv-lite86",
+ "rand_core 0.5.1",
+]
+
+[[package]]
+name = "rand_chacha"
+version = "0.3.1"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "e6c10a63a0fa32252be49d21e7709d4d4baf8d231c2dbce1eaa8141b9b127d88"
+dependencies = [
+ "ppv-lite86",
+ "rand_core 0.6.4",
+]
+
+[[package]]
+name = "rand_core"
+version = "0.5.1"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "90bde5296fc891b0cef12a6d03ddccc162ce7b2aff54160af9338f8d40df6d19"
+dependencies = [
+ "getrandom 0.1.16",
+]
+
+[[package]]
+name = "rand_core"
+version = "0.6.4"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "ec0be4795e2f6a28069bec0b5ff3e2ac9bafc99e6a9a7dc3547996c5c816922c"
+dependencies = [
+ "getrandom 0.2.17",
+]
+
+[[package]]
+name = "rand_core"
+version = "0.10.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "0c8d0fd677905edcbeedbf2edb6494d676f0e98d54d5cf9bda0b061cb8fb8aba"
+
+[[package]]
+name = "rand_hc"
+version = "0.2.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "ca3129af7b92a17112d59ad498c6f81eaf463253766b90396d39ea7a39d6613c"
+dependencies = [
+ "rand_core 0.5.1",
+]
+
+[[package]]
+name = "rand_pcg"
+version = "0.2.1"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "16abd0c1b639e9eb4d7c50c0b8100b0d0f849be2349829c740fe8e6eb4816429"
+dependencies = [
+ "rand_core 0.5.1",
+]
+
+[[package]]
+name = "raw-window-handle"
+version = "0.6.2"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "20675572f6f24e9e76ef639bc5552774ed45f1c30e2951e1e99c59888861c539"
+
+[[package]]
+name = "redox_syscall"
+version = "0.5.18"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "ed2bf2547551a7053d6fdfafda3f938979645c44812fbfcda098faae3f1a362d"
+dependencies = [
+ "bitflags 2.11.0",
+]
+
+[[package]]
+name = "redox_syscall"
+version = "0.7.3"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "6ce70a74e890531977d37e532c34d45e9055d2409ed08ddba14529471ed0be16"
+dependencies = [
+ "bitflags 2.11.0",
+]
+
+[[package]]
+name = "redox_users"
+version = "0.5.2"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "a4e608c6638b9c18977b00b475ac1f28d14e84b27d8d42f70e0bf1e3dec127ac"
+dependencies = [
+ "getrandom 0.2.17",
+ "libredox",
+ "thiserror 2.0.18",
+]
+
+[[package]]
+name = "ref-cast"
+version = "1.0.25"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "f354300ae66f76f1c85c5f84693f0ce81d747e2c3f21a45fef496d89c960bf7d"
+dependencies = [
+ "ref-cast-impl",
+]
+
+[[package]]
+name = "ref-cast-impl"
+version = "1.0.25"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "b7186006dcb21920990093f30e3dea63b7d6e977bf1256be20c3563a5db070da"
+dependencies = [
+ "proc-macro2",
+ "quote",
+ "syn 2.0.117",
+]
+
+[[package]]
+name = "regex"
+version = "1.12.3"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "e10754a14b9137dd7b1e3e5b0493cc9171fdd105e0ab477f51b72e7f3ac0e276"
+dependencies = [
+ "aho-corasick",
+ "memchr",
+ "regex-automata",
+ "regex-syntax",
+]
+
+[[package]]
+name = "regex-automata"
+version = "0.4.14"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "6e1dd4122fc1595e8162618945476892eefca7b88c52820e74af6262213cae8f"
+dependencies = [
+ "aho-corasick",
+ "memchr",
+ "regex-syntax",
+]
+
+[[package]]
+name = "regex-syntax"
+version = "0.8.10"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "dc897dd8d9e8bd1ed8cdad82b5966c3e0ecae09fb1907d58efaa013543185d0a"
+
+[[package]]
+name = "reqwest"
+version = "0.12.28"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "eddd3ca559203180a307f12d114c268abf583f59b03cb906fd0b3ff8646c1147"
+dependencies = [
+ "base64 0.22.1",
+ "bytes",
+ "encoding_rs",
+ "futures-core",
+ "h2",
+ "http",
+ "http-body",
+ "http-body-util",
+ "hyper",
+ "hyper-rustls",
+ "hyper-tls",
+ "hyper-util",
+ "js-sys",
+ "log",
+ "mime",
+ "native-tls",
+ "percent-encoding",
+ "pin-project-lite",
+ "rustls-pki-types",
+ "serde",
+ "serde_json",
+ "serde_urlencoded",
+ "sync_wrapper",
+ "tokio",
+ "tokio-native-tls",
+ "tower",
+ "tower-http",
+ "tower-service",
+ "url",
+ "wasm-bindgen",
+ "wasm-bindgen-futures",
+ "web-sys",
+]
+
+[[package]]
+name = "reqwest"
+version = "0.13.2"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "ab3f43e3283ab1488b624b44b0e988d0acea0b3214e694730a055cb6b2efa801"
+dependencies = [
+ "base64 0.22.1",
+ "bytes",
+ "futures-core",
+ "futures-util",
+ "http",
+ "http-body",
+ "http-body-util",
+ "hyper",
+ "hyper-rustls",
+ "hyper-util",
+ "js-sys",
+ "log",
+ "percent-encoding",
+ "pin-project-lite",
+ "rustls",
+ "rustls-pki-types",
+ "rustls-platform-verifier",
+ "serde",
+ "serde_json",
+ "sync_wrapper",
+ "tokio",
+ "tokio-rustls",
+ "tokio-util",
+ "tower",
+ "tower-http",
+ "tower-service",
+ "url",
+ "wasm-bindgen",
+ "wasm-bindgen-futures",
+ "wasm-streams",
+ "web-sys",
+]
+
+[[package]]
+name = "ring"
+version = "0.17.14"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "a4689e6c2294d81e88dc6261c768b63bc4fcdb852be6d1352498b114f61383b7"
+dependencies = [
+ "cc",
+ "cfg-if",
+ "getrandom 0.2.17",
+ "libc",
+ "untrusted",
+ "windows-sys 0.52.0",
+]
+
+[[package]]
+name = "rustc-hash"
+version = "2.1.1"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "357703d41365b4b27c590e3ed91eabb1b663f07c4c084095e60cbed4362dff0d"
+
+[[package]]
+name = "rustc_version"
+version = "0.4.1"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "cfcb3a22ef46e85b45de6ee7e79d063319ebb6594faafcf1c225ea92ab6e9b92"
+dependencies = [
+ "semver",
+]
+
+[[package]]
+name = "rustix"
+version = "1.1.4"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "b6fe4565b9518b83ef4f91bb47ce29620ca828bd32cb7e408f0062e9930ba190"
+dependencies = [
+ "bitflags 2.11.0",
+ "errno",
+ "libc",
+ "linux-raw-sys",
+ "windows-sys 0.61.2",
+]
+
+[[package]]
+name = "rustls"
+version = "0.23.37"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "758025cb5fccfd3bc2fd74708fd4682be41d99e5dff73c377c0646c6012c73a4"
+dependencies = [
+ "once_cell",
+ "ring",
+ "rustls-pki-types",
+ "rustls-webpki",
+ "subtle",
+ "zeroize",
+]
+
+[[package]]
+name = "rustls-native-certs"
+version = "0.8.3"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "612460d5f7bea540c490b2b6395d8e34a953e52b491accd6c86c8164c5932a63"
+dependencies = [
+ "openssl-probe",
+ "rustls-pki-types",
+ "schannel",
+ "security-framework",
+]
+
+[[package]]
+name = "rustls-pki-types"
+version = "1.14.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "be040f8b0a225e40375822a563fa9524378b9d63112f53e19ffff34df5d33fdd"
+dependencies = [
+ "zeroize",
+]
+
+[[package]]
+name = "rustls-platform-verifier"
+version = "0.6.2"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "1d99feebc72bae7ab76ba994bb5e121b8d83d910ca40b36e0921f53becc41784"
+dependencies = [
+ "core-foundation 0.10.1",
+ "core-foundation-sys",
+ "jni",
+ "log",
+ "once_cell",
+ "rustls",
+ "rustls-native-certs",
+ "rustls-platform-verifier-android",
+ "rustls-webpki",
+ "security-framework",
+ "security-framework-sys",
+ "webpki-root-certs",
+ "windows-sys 0.61.2",
+]
+
+[[package]]
+name = "rustls-platform-verifier-android"
+version = "0.1.1"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "f87165f0995f63a9fbeea62b64d10b4d9d8e78ec6d7d51fb2125fda7bb36788f"
+
+[[package]]
+name = "rustls-webpki"
+version = "0.103.10"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "df33b2b81ac578cabaf06b89b0631153a3f416b0a886e8a7a1707fb51abbd1ef"
+dependencies = [
+ "ring",
+ "rustls-pki-types",
+ "untrusted",
+]
+
+[[package]]
+name = "rustversion"
+version = "1.0.22"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "b39cdef0fa800fc44525c84ccb54a029961a8215f9619753635a9c0d2538d46d"
+
+[[package]]
+name = "ryu"
+version = "1.0.23"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "9774ba4a74de5f7b1c1451ed6cd5285a32eddb5cccb8cc655a4e50009e06477f"
+
+[[package]]
+name = "same-file"
+version = "1.0.6"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "93fc1dc3aaa9bfed95e02e6eadabb4baf7e3078b0bd1b4d7b6b0b68378900502"
+dependencies = [
+ "winapi-util",
+]
+
+[[package]]
+name = "schannel"
+version = "0.1.29"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "91c1b7e4904c873ef0710c1f407dde2e6287de2bebc1bbbf7d430bb7cbffd939"
+dependencies = [
+ "windows-sys 0.61.2",
+]
+
+[[package]]
+name = "schemars"
+version = "0.8.22"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "3fbf2ae1b8bc8e02df939598064d22402220cd5bbcca1c76f7d6a310974d5615"
+dependencies = [
+ "dyn-clone",
+ "indexmap 1.9.3",
+ "schemars_derive",
+ "serde",
+ "serde_json",
+ "url",
+ "uuid",
+]
+
+[[package]]
+name = "schemars"
+version = "0.9.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "4cd191f9397d57d581cddd31014772520aa448f65ef991055d7f61582c65165f"
+dependencies = [
+ "dyn-clone",
+ "ref-cast",
+ "serde",
+ "serde_json",
+]
+
+[[package]]
+name = "schemars"
+version = "1.2.1"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "a2b42f36aa1cd011945615b92222f6bf73c599a102a300334cd7f8dbeec726cc"
+dependencies = [
+ "dyn-clone",
+ "ref-cast",
+ "serde",
+ "serde_json",
+]
+
+[[package]]
+name = "schemars_derive"
+version = "0.8.22"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "32e265784ad618884abaea0600a9adf15393368d840e0222d101a072f3f7534d"
+dependencies = [
+ "proc-macro2",
+ "quote",
+ "serde_derive_internals",
+ "syn 2.0.117",
+]
+
+[[package]]
+name = "scopeguard"
+version = "1.2.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "94143f37725109f92c262ed2cf5e59bce7498c01bcc1502d7b9afe439a4e9f49"
+
+[[package]]
+name = "security-framework"
+version = "3.7.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "b7f4bc775c73d9a02cde8bf7b2ec4c9d12743edf609006c7facc23998404cd1d"
+dependencies = [
+ "bitflags 2.11.0",
+ "core-foundation 0.10.1",
+ "core-foundation-sys",
+ "libc",
+ "security-framework-sys",
+]
+
+[[package]]
+name = "security-framework-sys"
+version = "2.17.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "6ce2691df843ecc5d231c0b14ece2acc3efb62c0a398c7e1d875f3983ce020e3"
+dependencies = [
+ "core-foundation-sys",
+ "libc",
+]
+
+[[package]]
+name = "selectors"
+version = "0.24.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "0c37578180969d00692904465fb7f6b3d50b9a2b952b87c23d0e2e5cb5013416"
+dependencies = [
+ "bitflags 1.3.2",
+ "cssparser 0.29.6",
+ "derive_more 0.99.20",
+ "fxhash",
+ "log",
+ "phf 0.8.0",
+ "phf_codegen 0.8.0",
+ "precomputed-hash",
+ "servo_arc 0.2.0",
+ "smallvec",
+]
+
+[[package]]
+name = "selectors"
+version = "0.36.1"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "c5d9c0c92a92d33f08817311cf3f2c29a3538a8240e94a6a3c622ce652d7e00c"
+dependencies = [
+ "bitflags 2.11.0",
+ "cssparser 0.36.0",
+ "derive_more 2.1.1",
+ "log",
+ "new_debug_unreachable",
+ "phf 0.13.1",
+ "phf_codegen 0.13.1",
+ "precomputed-hash",
+ "rustc-hash",
+ "servo_arc 0.4.3",
+ "smallvec",
+]
+
+[[package]]
+name = "semver"
+version = "1.0.27"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "d767eb0aabc880b29956c35734170f26ed551a859dbd361d140cdbeca61ab1e2"
+dependencies = [
+ "serde",
+ "serde_core",
+]
+
+[[package]]
+name = "serde"
+version = "1.0.228"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "9a8e94ea7f378bd32cbbd37198a4a91436180c5bb472411e48b5ec2e2124ae9e"
+dependencies = [
+ "serde_core",
+ "serde_derive",
+]
+
+[[package]]
+name = "serde-untagged"
+version = "0.1.9"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "f9faf48a4a2d2693be24c6289dbe26552776eb7737074e6722891fadbe6c5058"
+dependencies = [
+ "erased-serde",
+ "serde",
+ "serde_core",
+ "typeid",
+]
+
+[[package]]
+name = "serde_core"
+version = "1.0.228"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "41d385c7d4ca58e59fc732af25c3983b67ac852c1a25000afe1175de458b67ad"
+dependencies = [
+ "serde_derive",
+]
+
+[[package]]
+name = "serde_derive"
+version = "1.0.228"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "d540f220d3187173da220f885ab66608367b6574e925011a9353e4badda91d79"
+dependencies = [
+ "proc-macro2",
+ "quote",
+ "syn 2.0.117",
+]
+
+[[package]]
+name = "serde_derive_internals"
+version = "0.29.1"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "18d26a20a969b9e3fdf2fc2d9f21eda6c40e2de84c9408bb5d3b05d499aae711"
+dependencies = [
+ "proc-macro2",
+ "quote",
+ "syn 2.0.117",
+]
+
+[[package]]
+name = "serde_json"
+version = "1.0.149"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "83fc039473c5595ace860d8c4fafa220ff474b3fc6bfdb4293327f1a37e94d86"
+dependencies = [
+ "itoa",
+ "memchr",
+ "serde",
+ "serde_core",
+ "zmij",
+]
+
+[[package]]
+name = "serde_repr"
+version = "0.1.20"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "175ee3e80ae9982737ca543e96133087cbd9a485eecc3bc4de9c1a37b47ea59c"
+dependencies = [
+ "proc-macro2",
+ "quote",
+ "syn 2.0.117",
+]
+
+[[package]]
+name = "serde_spanned"
+version = "0.6.9"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "bf41e0cfaf7226dca15e8197172c295a782857fcb97fad1808a166870dee75a3"
+dependencies = [
+ "serde",
+]
+
+[[package]]
+name = "serde_spanned"
+version = "1.0.4"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "f8bbf91e5a4d6315eee45e704372590b30e260ee83af6639d64557f51b067776"
+dependencies = [
+ "serde_core",
+]
+
+[[package]]
+name = "serde_urlencoded"
+version = "0.7.1"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "d3491c14715ca2294c4d6a88f15e84739788c1d030eed8c110436aafdaa2f3fd"
+dependencies = [
+ "form_urlencoded",
+ "itoa",
+ "ryu",
+ "serde",
+]
+
+[[package]]
+name = "serde_with"
+version = "3.18.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "dd5414fad8e6907dbdd5bc441a50ae8d6e26151a03b1de04d89a5576de61d01f"
+dependencies = [
+ "base64 0.22.1",
+ "chrono",
+ "hex",
+ "indexmap 1.9.3",
+ "indexmap 2.13.0",
+ "schemars 0.9.0",
+ "schemars 1.2.1",
+ "serde_core",
+ "serde_json",
+ "serde_with_macros",
+ "time",
+]
+
+[[package]]
+name = "serde_with_macros"
+version = "3.18.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "d3db8978e608f1fe7357e211969fd9abdcae80bac1ba7a3369bb7eb6b404eb65"
+dependencies = [
+ "darling",
+ "proc-macro2",
+ "quote",
+ "syn 2.0.117",
+]
+
+[[package]]
+name = "serialize-to-javascript"
+version = "0.1.2"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "04f3666a07a197cdb77cdf306c32be9b7f598d7060d50cfd4d5aa04bfd92f6c5"
+dependencies = [
+ "serde",
+ "serde_json",
+ "serialize-to-javascript-impl",
+]
+
+[[package]]
+name = "serialize-to-javascript-impl"
+version = "0.1.2"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "772ee033c0916d670af7860b6e1ef7d658a4629a6d0b4c8c3e67f09b3765b75d"
+dependencies = [
+ "proc-macro2",
+ "quote",
+ "syn 2.0.117",
+]
+
+[[package]]
+name = "servo_arc"
+version = "0.2.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "d52aa42f8fdf0fed91e5ce7f23d8138441002fa31dca008acf47e6fd4721f741"
+dependencies = [
+ "nodrop",
+ "stable_deref_trait",
+]
+
+[[package]]
+name = "servo_arc"
+version = "0.4.3"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "170fb83ab34de17dc69aa7c67482b22218ddb85da56546f9bd6b929e32a05930"
+dependencies = [
+ "stable_deref_trait",
+]
+
+[[package]]
+name = "sha2"
+version = "0.10.9"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "a7507d819769d01a365ab707794a4084392c824f54a7a6a7862f8c3d0892b283"
+dependencies = [
+ "cfg-if",
+ "cpufeatures 0.2.17",
+ "digest",
+]
+
+[[package]]
+name = "shlex"
+version = "1.3.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "0fda2ff0d084019ba4d7c6f371c95d8fd75ce3524c3cb8fb653a3023f6323e64"
+
+[[package]]
+name = "signal-hook-registry"
+version = "1.4.8"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "c4db69cba1110affc0e9f7bcd48bbf87b3f4fc7c61fc9155afd4c469eb3d6c1b"
+dependencies = [
+ "errno",
+ "libc",
+]
+
+[[package]]
+name = "simd-adler32"
+version = "0.3.8"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "e320a6c5ad31d271ad523dcf3ad13e2767ad8b1cb8f047f75a8aeaf8da139da2"
+
+[[package]]
+name = "simplelog"
+version = "0.12.2"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "16257adbfaef1ee58b1363bdc0664c9b8e1e30aed86049635fb5f147d065a9c0"
+dependencies = [
+ "log",
+ "termcolor",
+ "time",
+]
+
+[[package]]
+name = "siphasher"
+version = "0.3.11"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "38b58827f4464d87d377d175e90bf58eb00fd8716ff0a62f80356b5e61555d0d"
+
+[[package]]
+name = "siphasher"
+version = "1.0.2"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "b2aa850e253778c88a04c3d7323b043aeda9d3e30d5971937c1855769763678e"
+
+[[package]]
+name = "slab"
+version = "0.4.12"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "0c790de23124f9ab44544d7ac05d60440adc586479ce501c1d6d7da3cd8c9cf5"
+
+[[package]]
+name = "smallvec"
+version = "1.15.1"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "67b1b7a3b5fe4f1376887184045fcf45c69e92af734b7aaddc05fb777b6fbd03"
+
+[[package]]
+name = "socket2"
+version = "0.6.3"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "3a766e1110788c36f4fa1c2b71b387a7815aa65f88ce0229841826633d93723e"
+dependencies = [
+ "libc",
+ "windows-sys 0.61.2",
+]
+
+[[package]]
+name = "softbuffer"
+version = "0.4.8"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "aac18da81ebbf05109ab275b157c22a653bb3c12cf884450179942f81bcbf6c3"
+dependencies = [
+ "bytemuck",
+ "js-sys",
+ "ndk",
+ "objc2",
+ "objc2-core-foundation",
+ "objc2-core-graphics",
+ "objc2-foundation",
+ "objc2-quartz-core",
+ "raw-window-handle",
+ "redox_syscall 0.5.18",
+ "tracing",
+ "wasm-bindgen",
+ "web-sys",
+ "windows-sys 0.61.2",
+]
+
+[[package]]
+name = "soup3"
+version = "0.5.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "471f924a40f31251afc77450e781cb26d55c0b650842efafc9c6cbd2f7cc4f9f"
+dependencies = [
+ "futures-channel",
+ "gio",
+ "glib",
+ "libc",
+ "soup3-sys",
+]
+
+[[package]]
+name = "soup3-sys"
+version = "0.5.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "7ebe8950a680a12f24f15ebe1bf70db7af98ad242d9db43596ad3108aab86c27"
+dependencies = [
+ "gio-sys",
+ "glib-sys",
+ "gobject-sys",
+ "libc",
+ "system-deps",
+]
+
+[[package]]
+name = "stable_deref_trait"
+version = "1.2.1"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "6ce2be8dc25455e1f91df71bfa12ad37d7af1092ae736f3a6cd0e37bc7810596"
+
+[[package]]
+name = "string_cache"
+version = "0.8.9"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "bf776ba3fa74f83bf4b63c3dcbbf82173db2632ed8452cb2d891d33f459de70f"
+dependencies = [
+ "new_debug_unreachable",
+ "parking_lot",
+ "phf_shared 0.11.3",
+ "precomputed-hash",
+ "serde",
+]
+
+[[package]]
+name = "string_cache"
+version = "0.9.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "a18596f8c785a729f2819c0f6a7eae6ebeebdfffbfe4214ae6b087f690e31901"
+dependencies = [
+ "new_debug_unreachable",
+ "parking_lot",
+ "phf_shared 0.13.1",
+ "precomputed-hash",
+]
+
+[[package]]
+name = "string_cache_codegen"
+version = "0.5.4"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "c711928715f1fe0fe509c53b43e993a9a557babc2d0a3567d0a3006f1ac931a0"
+dependencies = [
+ "phf_generator 0.11.3",
+ "phf_shared 0.11.3",
+ "proc-macro2",
+ "quote",
+]
+
+[[package]]
+name = "string_cache_codegen"
+version = "0.6.1"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "585635e46db231059f76c5849798146164652513eb9e8ab2685939dd90f29b69"
+dependencies = [
+ "phf_generator 0.13.1",
+ "phf_shared 0.13.1",
+ "proc-macro2",
+ "quote",
+]
+
+[[package]]
+name = "strip-ansi-escapes"
+version = "0.2.1"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "2a8f8038e7e7969abb3f1b7c2a811225e9296da208539e0f79c5251d6cac0025"
+dependencies = [
+ "vte",
+]
+
+[[package]]
+name = "strsim"
+version = "0.11.1"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "7da8b5736845d9f2fcb837ea5d9e2628564b3b043a70948a3f0b778838c5fb4f"
+
+[[package]]
+name = "subtle"
+version = "2.6.1"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "13c2bddecc57b384dee18652358fb23172facb8a2c51ccc10d74c157bdea3292"
+
+[[package]]
+name = "swift-rs"
+version = "1.0.7"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "4057c98e2e852d51fdcfca832aac7b571f6b351ad159f9eda5db1655f8d0c4d7"
+dependencies = [
+ "base64 0.21.7",
+ "serde",
+ "serde_json",
+]
+
+[[package]]
+name = "syn"
+version = "1.0.109"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "72b64191b275b66ffe2469e8af2c1cfe3bafa67b529ead792a6d0160888b4237"
+dependencies = [
+ "proc-macro2",
+ "quote",
+ "unicode-ident",
+]
+
+[[package]]
+name = "syn"
+version = "2.0.117"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "e665b8803e7b1d2a727f4023456bbbbe74da67099c585258af0ad9c5013b9b99"
+dependencies = [
+ "proc-macro2",
+ "quote",
+ "unicode-ident",
+]
+
+[[package]]
+name = "sync_wrapper"
+version = "1.0.2"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "0bf256ce5efdfa370213c1dabab5935a12e49f2c58d15e9eac2870d3b4f27263"
+dependencies = [
+ "futures-core",
+]
+
+[[package]]
+name = "synstructure"
+version = "0.13.2"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "728a70f3dbaf5bab7f0c4b1ac8d7ae5ea60a4b5549c8a5914361c99147a709d2"
+dependencies = [
+ "proc-macro2",
+ "quote",
+ "syn 2.0.117",
+]
+
+[[package]]
+name = "system-configuration"
+version = "0.7.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "a13f3d0daba03132c0aa9767f98351b3488edc2c100cda2d2ec2b04f3d8d3c8b"
+dependencies = [
+ "bitflags 2.11.0",
+ "core-foundation 0.9.4",
+ "system-configuration-sys",
+]
+
+[[package]]
+name = "system-configuration-sys"
+version = "0.6.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "8e1d1b10ced5ca923a1fcb8d03e96b8d3268065d724548c0211415ff6ac6bac4"
+dependencies = [
+ "core-foundation-sys",
+ "libc",
+]
+
+[[package]]
+name = "system-deps"
+version = "6.2.2"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "a3e535eb8dded36d55ec13eddacd30dec501792ff23a0b1682c38601b8cf2349"
+dependencies = [
+ "cfg-expr",
+ "heck 0.5.0",
+ "pkg-config",
+ "toml 0.8.2",
+ "version-compare",
+]
+
+[[package]]
+name = "tao"
+version = "0.34.8"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "9103edf55f2da3c82aea4c7fab7c4241032bfeea0e71fa557d98e00e7ce7cc20"
+dependencies = [
+ "bitflags 2.11.0",
+ "block2",
+ "core-foundation 0.10.1",
+ "core-graphics",
+ "crossbeam-channel",
+ "dispatch2",
+ "dlopen2",
+ "dpi",
+ "gdkwayland-sys",
+ "gdkx11-sys",
+ "gtk",
+ "jni",
+ "libc",
+ "log",
+ "ndk",
+ "ndk-context",
+ "ndk-sys",
+ "objc2",
+ "objc2-app-kit",
+ "objc2-foundation",
+ "once_cell",
+ "parking_lot",
+ "raw-window-handle",
+ "tao-macros",
+ "unicode-segmentation",
+ "url",
+ "windows 0.61.3",
+ "windows-core 0.61.2",
+ "windows-version",
+ "x11-dl",
+]
+
+[[package]]
+name = "tao-macros"
+version = "0.1.3"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "f4e16beb8b2ac17db28eab8bca40e62dbfbb34c0fcdc6d9826b11b7b5d047dfd"
+dependencies = [
+ "proc-macro2",
+ "quote",
+ "syn 2.0.117",
+]
+
+[[package]]
+name = "tar"
+version = "0.4.45"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "22692a6476a21fa75fdfc11d452fda482af402c008cdbaf3476414e122040973"
+dependencies = [
+ "filetime",
+ "libc",
+ "xattr",
+]
+
+[[package]]
+name = "target-lexicon"
+version = "0.12.16"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "61c41af27dd6d1e27b1b16b489db798443478cef1f06a660c96db617ba5de3b1"
+
+[[package]]
+name = "tauri"
+version = "2.10.3"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "da77cc00fb9028caf5b5d4650f75e31f1ef3693459dfca7f7e506d1ecef0ba2d"
+dependencies = [
+ "anyhow",
+ "bytes",
+ "cookie",
+ "dirs",
+ "dunce",
+ "embed_plist",
+ "getrandom 0.3.4",
+ "glob",
+ "gtk",
+ "heck 0.5.0",
+ "http",
+ "jni",
+ "libc",
+ "log",
+ "mime",
+ "muda",
+ "objc2",
+ "objc2-app-kit",
+ "objc2-foundation",
+ "objc2-ui-kit",
+ "objc2-web-kit",
+ "percent-encoding",
+ "plist",
+ "raw-window-handle",
+ "reqwest 0.13.2",
+ "serde",
+ "serde_json",
+ "serde_repr",
+ "serialize-to-javascript",
+ "swift-rs",
+ "tauri-build",
+ "tauri-macros",
+ "tauri-runtime",
+ "tauri-runtime-wry",
+ "tauri-utils",
+ "thiserror 2.0.18",
+ "tokio",
+ "tray-icon",
+ "url",
+ "webkit2gtk",
+ "webview2-com",
+ "window-vibrancy",
+ "windows 0.61.3",
+]
+
+[[package]]
+name = "tauri-build"
+version = "2.5.6"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "4bbc990d1dbf57a8e1c7fa2327f2a614d8b757805603c1b9ba5c81bade09fd4d"
+dependencies = [
+ "anyhow",
+ "cargo_toml",
+ "dirs",
+ "glob",
+ "heck 0.5.0",
+ "json-patch",
+ "schemars 0.8.22",
+ "semver",
+ "serde",
+ "serde_json",
+ "tauri-utils",
+ "tauri-winres",
+ "toml 0.9.12+spec-1.1.0",
+ "walkdir",
+]
+
+[[package]]
+name = "tauri-codegen"
+version = "2.5.5"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "d4a24476afd977c5d5d169f72425868613d82747916dd29e0a357c84c4bd6d29"
+dependencies = [
+ "base64 0.22.1",
+ "brotli",
+ "ico",
+ "json-patch",
+ "plist",
+ "png",
+ "proc-macro2",
+ "quote",
+ "semver",
+ "serde",
+ "serde_json",
+ "sha2",
+ "syn 2.0.117",
+ "tauri-utils",
+ "thiserror 2.0.18",
+ "time",
+ "url",
+ "uuid",
+ "walkdir",
+]
+
+[[package]]
+name = "tauri-macros"
+version = "2.5.5"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "d39b349a98dadaffebb73f0a40dcd1f23c999211e5a2e744403db384d0c33de7"
+dependencies = [
+ "heck 0.5.0",
+ "proc-macro2",
+ "quote",
+ "syn 2.0.117",
+ "tauri-codegen",
+ "tauri-utils",
+]
+
+[[package]]
+name = "tauri-plugin"
+version = "2.5.4"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "ddde7d51c907b940fb573006cdda9a642d6a7c8153657e88f8a5c3c9290cd4aa"
+dependencies = [
+ "anyhow",
+ "glob",
+ "plist",
+ "schemars 0.8.22",
+ "serde",
+ "serde_json",
+ "tauri-utils",
+ "toml 0.9.12+spec-1.1.0",
+ "walkdir",
+]
+
+[[package]]
+name = "tauri-plugin-opener"
+version = "2.5.3"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "fc624469b06f59f5a29f874bbc61a2ed737c0f9c23ef09855a292c389c42e83f"
+dependencies = [
+ "dunce",
+ "glob",
+ "objc2-app-kit",
+ "objc2-foundation",
+ "open",
+ "schemars 0.8.22",
+ "serde",
+ "serde_json",
+ "tauri",
+ "tauri-plugin",
+ "thiserror 2.0.18",
+ "url",
+ "windows 0.61.3",
+ "zbus",
+]
+
+[[package]]
+name = "tauri-plugin-process"
+version = "2.3.1"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "d55511a7bf6cd70c8767b02c97bf8134fa434daf3926cfc1be0a0f94132d165a"
+dependencies = [
+ "tauri",
+ "tauri-plugin",
+]
+
+[[package]]
+name = "tauri-plugin-single-instance"
+version = "2.4.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "dc61e4822b8f74d68278e09161d3e3fdd1b14b9eb781e24edccaabf10c420e8c"
+dependencies = [
+ "serde",
+ "serde_json",
+ "tauri",
+ "thiserror 2.0.18",
+ "tracing",
+ "windows-sys 0.60.2",
+ "zbus",
+]
+
+[[package]]
+name = "tauri-plugin-updater"
+version = "2.10.1"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "806d9dac662c2e4594ff03c647a552f2c9bd544e7d0f683ec58f872f952ce4af"
+dependencies = [
+ "base64 0.22.1",
+ "dirs",
+ "flate2",
+ "futures-util",
+ "http",
+ "infer",
+ "log",
+ "minisign-verify",
+ "osakit",
+ "percent-encoding",
+ "reqwest 0.13.2",
+ "rustls",
+ "semver",
+ "serde",
+ "serde_json",
+ "tar",
+ "tauri",
+ "tauri-plugin",
+ "tempfile",
+ "thiserror 2.0.18",
+ "time",
+ "tokio",
+ "url",
+ "windows-sys 0.60.2",
+ "zip",
+]
+
+[[package]]
+name = "tauri-runtime"
+version = "2.10.1"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "2826d79a3297ed08cd6ea7f412644ef58e32969504bc4fbd8d7dbeabc4445ea2"
+dependencies = [
+ "cookie",
+ "dpi",
+ "gtk",
+ "http",
+ "jni",
+ "objc2",
+ "objc2-ui-kit",
+ "objc2-web-kit",
+ "raw-window-handle",
+ "serde",
+ "serde_json",
+ "tauri-utils",
+ "thiserror 2.0.18",
+ "url",
+ "webkit2gtk",
+ "webview2-com",
+ "windows 0.61.3",
+]
+
+[[package]]
+name = "tauri-runtime-wry"
+version = "2.10.1"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "e11ea2e6f801d275fdd890d6c9603736012742a1c33b96d0db788c9cdebf7f9e"
+dependencies = [
+ "gtk",
+ "http",
+ "jni",
+ "log",
+ "objc2",
+ "objc2-app-kit",
+ "once_cell",
+ "percent-encoding",
+ "raw-window-handle",
+ "softbuffer",
+ "tao",
+ "tauri-runtime",
+ "tauri-utils",
+ "url",
+ "webkit2gtk",
+ "webview2-com",
+ "windows 0.61.3",
+ "wry",
+]
+
+[[package]]
+name = "tauri-utils"
+version = "2.8.3"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "219a1f983a2af3653f75b5747f76733b0da7ff03069c7a41901a5eb3ace4557d"
+dependencies = [
+ "anyhow",
+ "brotli",
+ "cargo_metadata",
+ "ctor",
+ "dunce",
+ "glob",
+ "html5ever 0.29.1",
+ "http",
+ "infer",
+ "json-patch",
+ "kuchikiki",
+ "log",
+ "memchr",
+ "phf 0.11.3",
+ "proc-macro2",
+ "quote",
+ "regex",
+ "schemars 0.8.22",
+ "semver",
+ "serde",
+ "serde-untagged",
+ "serde_json",
+ "serde_with",
+ "swift-rs",
+ "thiserror 2.0.18",
+ "toml 0.9.12+spec-1.1.0",
+ "url",
+ "urlpattern",
+ "uuid",
+ "walkdir",
+]
+
+[[package]]
+name = "tauri-winres"
+version = "0.3.5"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "1087b111fe2b005e42dbdc1990fc18593234238d47453b0c99b7de1c9ab2c1e0"
+dependencies = [
+ "dunce",
+ "embed-resource",
+ "toml 0.9.12+spec-1.1.0",
+]
+
+[[package]]
+name = "tempfile"
+version = "3.27.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "32497e9a4c7b38532efcdebeef879707aa9f794296a4f0244f6f69e9bc8574bd"
+dependencies = [
+ "fastrand",
+ "getrandom 0.4.2",
+ "once_cell",
+ "rustix",
+ "windows-sys 0.61.2",
+]
+
+[[package]]
+name = "tendril"
+version = "0.4.3"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "d24a120c5fc464a3458240ee02c299ebcb9d67b5249c8848b09d639dca8d7bb0"
+dependencies = [
+ "futf",
+ "mac",
+ "utf-8",
+]
+
+[[package]]
+name = "tendril"
+version = "0.5.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "c4790fc369d5a530f4b544b094e31388b9b3a37c0f4652ade4505945f5660d24"
+dependencies = [
+ "new_debug_unreachable",
+ "utf-8",
+]
+
+[[package]]
+name = "termcolor"
+version = "1.4.1"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "06794f8f6c5c898b3275aebefa6b8a1cb24cd2c6c79397ab15774837a0bc5755"
+dependencies = [
+ "winapi-util",
+]
+
+[[package]]
+name = "thiserror"
+version = "1.0.69"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "b6aaf5339b578ea85b50e080feb250a3e8ae8cfcdff9a461c9ec2904bc923f52"
+dependencies = [
+ "thiserror-impl 1.0.69",
+]
+
+[[package]]
+name = "thiserror"
+version = "2.0.18"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "4288b5bcbc7920c07a1149a35cf9590a2aa808e0bc1eafaade0b80947865fbc4"
+dependencies = [
+ "thiserror-impl 2.0.18",
+]
+
+[[package]]
+name = "thiserror-impl"
+version = "1.0.69"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "4fee6c4efc90059e10f81e6d42c60a18f76588c3d74cb83a0b242a2b6c7504c1"
+dependencies = [
+ "proc-macro2",
+ "quote",
+ "syn 2.0.117",
+]
+
+[[package]]
+name = "thiserror-impl"
+version = "2.0.18"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "ebc4ee7f67670e9b64d05fa4253e753e016c6c95ff35b89b7941d6b856dec1d5"
+dependencies = [
+ "proc-macro2",
+ "quote",
+ "syn 2.0.117",
+]
+
+[[package]]
+name = "time"
+version = "0.3.47"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "743bd48c283afc0388f9b8827b976905fb217ad9e647fae3a379a9283c4def2c"
+dependencies = [
+ "deranged",
+ "itoa",
+ "libc",
+ "num-conv",
+ "num_threads",
+ "powerfmt",
+ "serde_core",
+ "time-core",
+ "time-macros",
+]
+
+[[package]]
+name = "time-core"
+version = "0.1.8"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "7694e1cfe791f8d31026952abf09c69ca6f6fa4e1a1229e18988f06a04a12dca"
+
+[[package]]
+name = "time-macros"
+version = "0.2.27"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "2e70e4c5a0e0a8a4823ad65dfe1a6930e4f4d756dcd9dd7939022b5e8c501215"
+dependencies = [
+ "num-conv",
+ "time-core",
+]
+
+[[package]]
+name = "tinystr"
+version = "0.8.2"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "42d3e9c45c09de15d06dd8acf5f4e0e399e85927b7f00711024eb7ae10fa4869"
+dependencies = [
+ "displaydoc",
+ "zerovec",
+]
+
+[[package]]
+name = "tokio"
+version = "1.50.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "27ad5e34374e03cfffefc301becb44e9dc3c17584f414349ebe29ed26661822d"
+dependencies = [
+ "bytes",
+ "libc",
+ "mio",
+ "parking_lot",
+ "pin-project-lite",
+ "signal-hook-registry",
+ "socket2",
+ "tokio-macros",
+ "windows-sys 0.61.2",
+]
+
+[[package]]
+name = "tokio-macros"
+version = "2.6.1"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "5c55a2eff8b69ce66c84f85e1da1c233edc36ceb85a2058d11b0d6a3c7e7569c"
+dependencies = [
+ "proc-macro2",
+ "quote",
+ "syn 2.0.117",
+]
+
+[[package]]
+name = "tokio-native-tls"
+version = "0.3.1"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "bbae76ab933c85776efabc971569dd6119c580d8f5d448769dec1764bf796ef2"
+dependencies = [
+ "native-tls",
+ "tokio",
+]
+
+[[package]]
+name = "tokio-rustls"
+version = "0.26.4"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "1729aa945f29d91ba541258c8df89027d5792d85a8841fb65e8bf0f4ede4ef61"
+dependencies = [
+ "rustls",
+ "tokio",
+]
+
+[[package]]
+name = "tokio-util"
+version = "0.7.18"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "9ae9cec805b01e8fc3fd2fe289f89149a9b66dd16786abd8b19cfa7b48cb0098"
+dependencies = [
+ "bytes",
+ "futures-core",
+ "futures-sink",
+ "pin-project-lite",
+ "tokio",
+]
+
+[[package]]
+name = "toml"
+version = "0.8.2"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "185d8ab0dfbb35cf1399a6344d8484209c088f75f8f68230da55d48d95d43e3d"
+dependencies = [
+ "serde",
+ "serde_spanned 0.6.9",
+ "toml_datetime 0.6.3",
+ "toml_edit 0.20.2",
+]
+
+[[package]]
+name = "toml"
+version = "0.9.12+spec-1.1.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "cf92845e79fc2e2def6a5d828f0801e29a2f8acc037becc5ab08595c7d5e9863"
+dependencies = [
+ "indexmap 2.13.0",
+ "serde_core",
+ "serde_spanned 1.0.4",
+ "toml_datetime 0.7.5+spec-1.1.0",
+ "toml_parser",
+ "toml_writer",
+ "winnow 0.7.15",
+]
+
+[[package]]
+name = "toml_datetime"
+version = "0.6.3"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "7cda73e2f1397b1262d6dfdcef8aafae14d1de7748d66822d3bfeeb6d03e5e4b"
+dependencies = [
+ "serde",
+]
+
+[[package]]
+name = "toml_datetime"
+version = "0.7.5+spec-1.1.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "92e1cfed4a3038bc5a127e35a2d360f145e1f4b971b551a2ba5fd7aedf7e1347"
+dependencies = [
+ "serde_core",
+]
+
+[[package]]
+name = "toml_datetime"
+version = "1.0.1+spec-1.1.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "9b320e741db58cac564e26c607d3cc1fdc4a88fd36c879568c07856ed83ff3e9"
+dependencies = [
+ "serde_core",
+]
+
+[[package]]
+name = "toml_edit"
+version = "0.19.15"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "1b5bb770da30e5cbfde35a2d7b9b8a2c4b8ef89548a7a6aeab5c9a576e3e7421"
+dependencies = [
+ "indexmap 2.13.0",
+ "toml_datetime 0.6.3",
+ "winnow 0.5.40",
+]
+
+[[package]]
+name = "toml_edit"
+version = "0.20.2"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "396e4d48bbb2b7554c944bde63101b5ae446cff6ec4a24227428f15eb72ef338"
+dependencies = [
+ "indexmap 2.13.0",
+ "serde",
+ "serde_spanned 0.6.9",
+ "toml_datetime 0.6.3",
+ "winnow 0.5.40",
+]
+
+[[package]]
+name = "toml_edit"
+version = "0.25.5+spec-1.1.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "8ca1a40644a28bce036923f6a431df0b34236949d111cc07cb6dca830c9ef2e1"
+dependencies = [
+ "indexmap 2.13.0",
+ "toml_datetime 1.0.1+spec-1.1.0",
+ "toml_parser",
+ "winnow 1.0.0",
+]
+
+[[package]]
+name = "toml_parser"
+version = "1.0.10+spec-1.1.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "7df25b4befd31c4816df190124375d5a20c6b6921e2cad937316de3fccd63420"
+dependencies = [
+ "winnow 1.0.0",
+]
+
+[[package]]
+name = "toml_writer"
+version = "1.0.7+spec-1.1.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "f17aaa1c6e3dc22b1da4b6bba97d066e354c7945cac2f7852d4e4e7ca7a6b56d"
+
+[[package]]
+name = "tower"
+version = "0.5.3"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "ebe5ef63511595f1344e2d5cfa636d973292adc0eec1f0ad45fae9f0851ab1d4"
+dependencies = [
+ "futures-core",
+ "futures-util",
+ "pin-project-lite",
+ "sync_wrapper",
+ "tokio",
+ "tower-layer",
+ "tower-service",
+]
+
+[[package]]
+name = "tower-http"
+version = "0.6.8"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "d4e6559d53cc268e5031cd8429d05415bc4cb4aefc4aa5d6cc35fbf5b924a1f8"
+dependencies = [
+ "bitflags 2.11.0",
+ "bytes",
+ "futures-util",
+ "http",
+ "http-body",
+ "iri-string",
+ "pin-project-lite",
+ "tower",
+ "tower-layer",
+ "tower-service",
+]
+
+[[package]]
+name = "tower-layer"
+version = "0.3.3"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "121c2a6cda46980bb0fcd1647ffaf6cd3fc79a013de288782836f6df9c48780e"
+
+[[package]]
+name = "tower-service"
+version = "0.3.3"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "8df9b6e13f2d32c91b9bd719c00d1958837bc7dec474d94952798cc8e69eeec3"
+
+[[package]]
+name = "tracing"
+version = "0.1.44"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "63e71662fa4b2a2c3a26f570f037eb95bb1f85397f3cd8076caed2f026a6d100"
+dependencies = [
+ "pin-project-lite",
+ "tracing-attributes",
+ "tracing-core",
+]
+
+[[package]]
+name = "tracing-attributes"
+version = "0.1.31"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "7490cfa5ec963746568740651ac6781f701c9c5ea257c58e057f3ba8cf69e8da"
+dependencies = [
+ "proc-macro2",
+ "quote",
+ "syn 2.0.117",
+]
+
+[[package]]
+name = "tracing-core"
+version = "0.1.36"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "db97caf9d906fbde555dd62fa95ddba9eecfd14cb388e4f491a66d74cd5fb79a"
+dependencies = [
+ "once_cell",
+]
+
+[[package]]
+name = "tray-icon"
+version = "0.21.3"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "a5e85aa143ceb072062fc4d6356c1b520a51d636e7bc8e77ec94be3608e5e80c"
+dependencies = [
+ "crossbeam-channel",
+ "dirs",
+ "libappindicator",
+ "muda",
+ "objc2",
+ "objc2-app-kit",
+ "objc2-core-foundation",
+ "objc2-core-graphics",
+ "objc2-foundation",
+ "once_cell",
+ "png",
+ "serde",
+ "thiserror 2.0.18",
+ "windows-sys 0.60.2",
+]
+
+[[package]]
+name = "try-lock"
+version = "0.2.5"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "e421abadd41a4225275504ea4d6566923418b7f05506fbc9c0fe86ba7396114b"
+
+[[package]]
+name = "typeid"
+version = "1.0.3"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "bc7d623258602320d5c55d1bc22793b57daff0ec7efc270ea7d55ce1d5f5471c"
+
+[[package]]
+name = "typenum"
+version = "1.19.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "562d481066bde0658276a35467c4af00bdc6ee726305698a55b86e61d7ad82bb"
+
+[[package]]
+name = "uds_windows"
+version = "1.2.1"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "f2f6fb2847f6742cd76af783a2a2c49e9375d0a111c7bef6f71cd9e738c72d6e"
+dependencies = [
+ "memoffset",
+ "tempfile",
+ "windows-sys 0.61.2",
+]
+
+[[package]]
+name = "unic-char-property"
+version = "0.9.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "a8c57a407d9b6fa02b4795eb81c5b6652060a15a7903ea981f3d723e6c0be221"
+dependencies = [
+ "unic-char-range",
+]
+
+[[package]]
+name = "unic-char-range"
+version = "0.9.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "0398022d5f700414f6b899e10b8348231abf9173fa93144cbc1a43b9793c1fbc"
+
+[[package]]
+name = "unic-common"
+version = "0.9.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "80d7ff825a6a654ee85a63e80f92f054f904f21e7d12da4e22f9834a4aaa35bc"
+
+[[package]]
+name = "unic-ucd-ident"
+version = "0.9.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "e230a37c0381caa9219d67cf063aa3a375ffed5bf541a452db16e744bdab6987"
+dependencies = [
+ "unic-char-property",
+ "unic-char-range",
+ "unic-ucd-version",
+]
+
+[[package]]
+name = "unic-ucd-version"
+version = "0.9.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "96bd2f2237fe450fcd0a1d2f5f4e91711124f7857ba2e964247776ebeeb7b0c4"
+dependencies = [
+ "unic-common",
+]
+
+[[package]]
+name = "unicode-ident"
+version = "1.0.24"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "e6e4313cd5fcd3dad5cafa179702e2b244f760991f45397d14d4ebf38247da75"
+
+[[package]]
+name = "unicode-segmentation"
+version = "1.12.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "f6ccf251212114b54433ec949fd6a7841275f9ada20dddd2f29e9ceea4501493"
+
+[[package]]
+name = "unicode-xid"
+version = "0.2.6"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "ebc1c04c71510c7f702b52b7c350734c9ff1295c464a03335b00bb84fc54f853"
+
+[[package]]
+name = "unsloth-studio"
+version = "2026.4.7"
+dependencies = [
+ "dirs",
+ "elevated-command",
+ "fix-path-env",
+ "libc",
+ "log",
+ "open",
+ "process-wrap",
+ "rand 0.10.0",
+ "regex",
+ "reqwest 0.12.28",
+ "serde",
+ "serde_json",
+ "simplelog",
+ "tauri",
+ "tauri-build",
+ "tauri-plugin-opener",
+ "tauri-plugin-process",
+ "tauri-plugin-single-instance",
+ "tauri-plugin-updater",
+ "tokio",
+ "windows 0.62.2",
+ "windows-sys 0.59.0",
+]
+
+[[package]]
+name = "untrusted"
+version = "0.9.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "8ecb6da28b8a351d773b68d5825ac39017e680750f980f3a1a85cd8dd28a47c1"
+
+[[package]]
+name = "url"
+version = "2.5.8"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "ff67a8a4397373c3ef660812acab3268222035010ab8680ec4215f38ba3d0eed"
+dependencies = [
+ "form_urlencoded",
+ "idna",
+ "percent-encoding",
+ "serde",
+ "serde_derive",
+]
+
+[[package]]
+name = "urlpattern"
+version = "0.3.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "70acd30e3aa1450bc2eece896ce2ad0d178e9c079493819301573dae3c37ba6d"
+dependencies = [
+ "regex",
+ "serde",
+ "unic-ucd-ident",
+ "url",
+]
+
+[[package]]
+name = "utf-8"
+version = "0.7.6"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "09cc8ee72d2a9becf2f2febe0205bbed8fc6615b7cb429ad062dc7b7ddd036a9"
+
+[[package]]
+name = "utf8_iter"
+version = "1.0.4"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "b6c140620e7ffbb22c2dee59cafe6084a59b5ffc27a8859a5f0d494b5d52b6be"
+
+[[package]]
+name = "uuid"
+version = "1.22.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "a68d3c8f01c0cfa54a75291d83601161799e4a89a39e0929f4b0354d88757a37"
+dependencies = [
+ "getrandom 0.4.2",
+ "js-sys",
+ "serde_core",
+ "wasm-bindgen",
+]
+
+[[package]]
+name = "vcpkg"
+version = "0.2.15"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "accd4ea62f7bb7a82fe23066fb0957d48ef677f6eeb8215f372f52e48bb32426"
+
+[[package]]
+name = "version-compare"
+version = "0.2.1"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "03c2856837ef78f57382f06b2b8563a2f512f7185d732608fd9176cb3b8edf0e"
+
+[[package]]
+name = "version_check"
+version = "0.9.5"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "0b928f33d975fc6ad9f86c8f283853ad26bdd5b10b7f1542aa2fa15e2289105a"
+
+[[package]]
+name = "vswhom"
+version = "0.1.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "be979b7f07507105799e854203b470ff7c78a1639e330a58f183b5fea574608b"
+dependencies = [
+ "libc",
+ "vswhom-sys",
+]
+
+[[package]]
+name = "vswhom-sys"
+version = "0.1.3"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "fb067e4cbd1ff067d1df46c9194b5de0e98efd2810bbc95c5d5e5f25a3231150"
+dependencies = [
+ "cc",
+ "libc",
+]
+
+[[package]]
+name = "vte"
+version = "0.14.1"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "231fdcd7ef3037e8330d8e17e61011a2c244126acc0a982f4040ac3f9f0bc077"
+dependencies = [
+ "memchr",
+]
+
+[[package]]
+name = "walkdir"
+version = "2.5.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "29790946404f91d9c5d06f9874efddea1dc06c5efe94541a7d6863108e3a5e4b"
+dependencies = [
+ "same-file",
+ "winapi-util",
+]
+
+[[package]]
+name = "want"
+version = "0.3.1"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "bfa7760aed19e106de2c7c0b581b509f2f25d3dacaf737cb82ac61bc6d760b0e"
+dependencies = [
+ "try-lock",
+]
+
+[[package]]
+name = "wasi"
+version = "0.9.0+wasi-snapshot-preview1"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "cccddf32554fecc6acb585f82a32a72e28b48f8c4c1883ddfeeeaa96f7d8e519"
+
+[[package]]
+name = "wasi"
+version = "0.11.1+wasi-snapshot-preview1"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "ccf3ec651a847eb01de73ccad15eb7d99f80485de043efb2f370cd654f4ea44b"
+
+[[package]]
+name = "wasip2"
+version = "1.0.2+wasi-0.2.9"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "9517f9239f02c069db75e65f174b3da828fe5f5b945c4dd26bd25d89c03ebcf5"
+dependencies = [
+ "wit-bindgen",
+]
+
+[[package]]
+name = "wasip3"
+version = "0.4.0+wasi-0.3.0-rc-2026-01-06"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "5428f8bf88ea5ddc08faddef2ac4a67e390b88186c703ce6dbd955e1c145aca5"
+dependencies = [
+ "wit-bindgen",
+]
+
+[[package]]
+name = "wasm-bindgen"
+version = "0.2.114"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "6532f9a5c1ece3798cb1c2cfdba640b9b3ba884f5db45973a6f442510a87d38e"
+dependencies = [
+ "cfg-if",
+ "once_cell",
+ "rustversion",
+ "wasm-bindgen-macro",
+ "wasm-bindgen-shared",
+]
+
+[[package]]
+name = "wasm-bindgen-futures"
+version = "0.4.64"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "e9c5522b3a28661442748e09d40924dfb9ca614b21c00d3fd135720e48b67db8"
+dependencies = [
+ "cfg-if",
+ "futures-util",
+ "js-sys",
+ "once_cell",
+ "wasm-bindgen",
+ "web-sys",
+]
+
+[[package]]
+name = "wasm-bindgen-macro"
+version = "0.2.114"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "18a2d50fcf105fb33bb15f00e7a77b772945a2ee45dcf454961fd843e74c18e6"
+dependencies = [
+ "quote",
+ "wasm-bindgen-macro-support",
+]
+
+[[package]]
+name = "wasm-bindgen-macro-support"
+version = "0.2.114"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "03ce4caeaac547cdf713d280eda22a730824dd11e6b8c3ca9e42247b25c631e3"
+dependencies = [
+ "bumpalo",
+ "proc-macro2",
+ "quote",
+ "syn 2.0.117",
+ "wasm-bindgen-shared",
+]
+
+[[package]]
+name = "wasm-bindgen-shared"
+version = "0.2.114"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "75a326b8c223ee17883a4251907455a2431acc2791c98c26279376490c378c16"
+dependencies = [
+ "unicode-ident",
+]
+
+[[package]]
+name = "wasm-encoder"
+version = "0.244.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "990065f2fe63003fe337b932cfb5e3b80e0b4d0f5ff650e6985b1048f62c8319"
+dependencies = [
+ "leb128fmt",
+ "wasmparser",
+]
+
+[[package]]
+name = "wasm-metadata"
+version = "0.244.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "bb0e353e6a2fbdc176932bbaab493762eb1255a7900fe0fea1a2f96c296cc909"
+dependencies = [
+ "anyhow",
+ "indexmap 2.13.0",
+ "wasm-encoder",
+ "wasmparser",
+]
+
+[[package]]
+name = "wasm-streams"
+version = "0.5.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "9d1ec4f6517c9e11ae630e200b2b65d193279042e28edd4a2cda233e46670bbb"
+dependencies = [
+ "futures-util",
+ "js-sys",
+ "wasm-bindgen",
+ "wasm-bindgen-futures",
+ "web-sys",
+]
+
+[[package]]
+name = "wasmparser"
+version = "0.244.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "47b807c72e1bac69382b3a6fb3dbe8ea4c0ed87ff5629b8685ae6b9a611028fe"
+dependencies = [
+ "bitflags 2.11.0",
+ "hashbrown 0.15.5",
+ "indexmap 2.13.0",
+ "semver",
+]
+
+[[package]]
+name = "web-sys"
+version = "0.3.91"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "854ba17bb104abfb26ba36da9729addc7ce7f06f5c0f90f3c391f8461cca21f9"
+dependencies = [
+ "js-sys",
+ "wasm-bindgen",
+]
+
+[[package]]
+name = "web_atoms"
+version = "0.2.3"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "57a9779e9f04d2ac1ce317aee707aa2f6b773afba7b931222bff6983843b1576"
+dependencies = [
+ "phf 0.13.1",
+ "phf_codegen 0.13.1",
+ "string_cache 0.9.0",
+ "string_cache_codegen 0.6.1",
+]
+
+[[package]]
+name = "webkit2gtk"
+version = "2.0.2"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "a1027150013530fb2eaf806408df88461ae4815a45c541c8975e61d6f2fc4793"
+dependencies = [
+ "bitflags 1.3.2",
+ "cairo-rs",
+ "gdk",
+ "gdk-sys",
+ "gio",
+ "gio-sys",
+ "glib",
+ "glib-sys",
+ "gobject-sys",
+ "gtk",
+ "gtk-sys",
+ "javascriptcore-rs",
+ "libc",
+ "once_cell",
+ "soup3",
+ "webkit2gtk-sys",
+]
+
+[[package]]
+name = "webkit2gtk-sys"
+version = "2.0.2"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "916a5f65c2ef0dfe12fff695960a2ec3d4565359fdbb2e9943c974e06c734ea5"
+dependencies = [
+ "bitflags 1.3.2",
+ "cairo-sys-rs",
+ "gdk-sys",
+ "gio-sys",
+ "glib-sys",
+ "gobject-sys",
+ "gtk-sys",
+ "javascriptcore-rs-sys",
+ "libc",
+ "pkg-config",
+ "soup3-sys",
+ "system-deps",
+]
+
+[[package]]
+name = "webpki-root-certs"
+version = "1.0.6"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "804f18a4ac2676ffb4e8b5b5fa9ae38af06df08162314f96a68d2a363e21a8ca"
+dependencies = [
+ "rustls-pki-types",
+]
+
+[[package]]
+name = "webview2-com"
+version = "0.38.2"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "7130243a7a5b33c54a444e54842e6a9e133de08b5ad7b5861cd8ed9a6a5bc96a"
+dependencies = [
+ "webview2-com-macros",
+ "webview2-com-sys",
+ "windows 0.61.3",
+ "windows-core 0.61.2",
+ "windows-implement",
+ "windows-interface",
+]
+
+[[package]]
+name = "webview2-com-macros"
+version = "0.8.1"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "67a921c1b6914c367b2b823cd4cde6f96beec77d30a939c8199bb377cf9b9b54"
+dependencies = [
+ "proc-macro2",
+ "quote",
+ "syn 2.0.117",
+]
+
+[[package]]
+name = "webview2-com-sys"
+version = "0.38.2"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "381336cfffd772377d291702245447a5251a2ffa5bad679c99e61bc48bacbf9c"
+dependencies = [
+ "thiserror 2.0.18",
+ "windows 0.61.3",
+ "windows-core 0.61.2",
+]
+
+[[package]]
+name = "winapi"
+version = "0.3.9"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "5c839a674fcd7a98952e593242ea400abe93992746761e38641405d28b00f419"
+dependencies = [
+ "winapi-i686-pc-windows-gnu",
+ "winapi-x86_64-pc-windows-gnu",
+]
+
+[[package]]
+name = "winapi-i686-pc-windows-gnu"
+version = "0.4.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "ac3b87c63620426dd9b991e5ce0329eff545bccbbb34f3be09ff6fb6ab51b7b6"
+
+[[package]]
+name = "winapi-util"
+version = "0.1.11"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "c2a7b1c03c876122aa43f3020e6c3c3ee5c05081c9a00739faf7503aeba10d22"
+dependencies = [
+ "windows-sys 0.61.2",
+]
+
+[[package]]
+name = "winapi-x86_64-pc-windows-gnu"
+version = "0.4.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "712e227841d057c1ee1cd2fb22fa7e5a5461ae8e48fa2ca79ec42cfc1931183f"
+
+[[package]]
+name = "window-vibrancy"
+version = "0.6.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "d9bec5a31f3f9362f2258fd0e9c9dd61a9ca432e7306cc78c444258f0dce9a9c"
+dependencies = [
+ "objc2",
+ "objc2-app-kit",
+ "objc2-core-foundation",
+ "objc2-foundation",
+ "raw-window-handle",
+ "windows-sys 0.59.0",
+ "windows-version",
+]
+
+[[package]]
+name = "windows"
+version = "0.52.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "e48a53791691ab099e5e2ad123536d0fff50652600abaf43bbf952894110d0be"
+dependencies = [
+ "windows-core 0.52.0",
+ "windows-targets 0.52.6",
+]
+
+[[package]]
+name = "windows"
+version = "0.61.3"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "9babd3a767a4c1aef6900409f85f5d53ce2544ccdfaa86dad48c91782c6d6893"
+dependencies = [
+ "windows-collections 0.2.0",
+ "windows-core 0.61.2",
+ "windows-future 0.2.1",
+ "windows-link 0.1.3",
+ "windows-numerics 0.2.0",
+]
+
+[[package]]
+name = "windows"
+version = "0.62.2"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "527fadee13e0c05939a6a05d5bd6eec6cd2e3dbd648b9f8e447c6518133d8580"
+dependencies = [
+ "windows-collections 0.3.2",
+ "windows-core 0.62.2",
+ "windows-future 0.3.2",
+ "windows-numerics 0.3.1",
+]
+
+[[package]]
+name = "windows-collections"
+version = "0.2.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "3beeceb5e5cfd9eb1d76b381630e82c4241ccd0d27f1a39ed41b2760b255c5e8"
+dependencies = [
+ "windows-core 0.61.2",
+]
+
+[[package]]
+name = "windows-collections"
+version = "0.3.2"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "23b2d95af1a8a14a3c7367e1ed4fc9c20e0a26e79551b1454d72583c97cc6610"
+dependencies = [
+ "windows-core 0.62.2",
+]
+
+[[package]]
+name = "windows-core"
+version = "0.52.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "33ab640c8d7e35bf8ba19b884ba838ceb4fba93a4e8c65a9059d08afcfc683d9"
+dependencies = [
+ "windows-targets 0.52.6",
+]
+
+[[package]]
+name = "windows-core"
+version = "0.61.2"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "c0fdd3ddb90610c7638aa2b3a3ab2904fb9e5cdbecc643ddb3647212781c4ae3"
+dependencies = [
+ "windows-implement",
+ "windows-interface",
+ "windows-link 0.1.3",
+ "windows-result 0.3.4",
+ "windows-strings 0.4.2",
+]
+
+[[package]]
+name = "windows-core"
+version = "0.62.2"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "b8e83a14d34d0623b51dce9581199302a221863196a1dde71a7663a4c2be9deb"
+dependencies = [
+ "windows-implement",
+ "windows-interface",
+ "windows-link 0.2.1",
+ "windows-result 0.4.1",
+ "windows-strings 0.5.1",
+]
+
+[[package]]
+name = "windows-future"
+version = "0.2.1"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "fc6a41e98427b19fe4b73c550f060b59fa592d7d686537eebf9385621bfbad8e"
+dependencies = [
+ "windows-core 0.61.2",
+ "windows-link 0.1.3",
+ "windows-threading 0.1.0",
+]
+
+[[package]]
+name = "windows-future"
+version = "0.3.2"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "e1d6f90251fe18a279739e78025bd6ddc52a7e22f921070ccdc67dde84c605cb"
+dependencies = [
+ "windows-core 0.62.2",
+ "windows-link 0.2.1",
+ "windows-threading 0.2.1",
+]
+
+[[package]]
+name = "windows-implement"
+version = "0.60.2"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "053e2e040ab57b9dc951b72c264860db7eb3b0200ba345b4e4c3b14f67855ddf"
+dependencies = [
+ "proc-macro2",
+ "quote",
+ "syn 2.0.117",
+]
+
+[[package]]
+name = "windows-interface"
+version = "0.59.3"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "3f316c4a2570ba26bbec722032c4099d8c8bc095efccdc15688708623367e358"
+dependencies = [
+ "proc-macro2",
+ "quote",
+ "syn 2.0.117",
+]
+
+[[package]]
+name = "windows-link"
+version = "0.1.3"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "5e6ad25900d524eaabdbbb96d20b4311e1e7ae1699af4fb28c17ae66c80d798a"
+
+[[package]]
+name = "windows-link"
+version = "0.2.1"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "f0805222e57f7521d6a62e36fa9163bc891acd422f971defe97d64e70d0a4fe5"
+
+[[package]]
+name = "windows-numerics"
+version = "0.2.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "9150af68066c4c5c07ddc0ce30421554771e528bde427614c61038bc2c92c2b1"
+dependencies = [
+ "windows-core 0.61.2",
+ "windows-link 0.1.3",
+]
+
+[[package]]
+name = "windows-numerics"
+version = "0.3.1"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "6e2e40844ac143cdb44aead537bbf727de9b044e107a0f1220392177d15b0f26"
+dependencies = [
+ "windows-core 0.62.2",
+ "windows-link 0.2.1",
+]
+
+[[package]]
+name = "windows-registry"
+version = "0.6.1"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "02752bf7fbdcce7f2a27a742f798510f3e5ad88dbe84871e5168e2120c3d5720"
+dependencies = [
+ "windows-link 0.2.1",
+ "windows-result 0.4.1",
+ "windows-strings 0.5.1",
+]
+
+[[package]]
+name = "windows-result"
+version = "0.3.4"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "56f42bd332cc6c8eac5af113fc0c1fd6a8fd2aa08a0119358686e5160d0586c6"
+dependencies = [
+ "windows-link 0.1.3",
+]
+
+[[package]]
+name = "windows-result"
+version = "0.4.1"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "7781fa89eaf60850ac3d2da7af8e5242a5ea78d1a11c49bf2910bb5a73853eb5"
+dependencies = [
+ "windows-link 0.2.1",
+]
+
+[[package]]
+name = "windows-strings"
+version = "0.4.2"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "56e6c93f3a0c3b36176cb1327a4958a0353d5d166c2a35cb268ace15e91d3b57"
+dependencies = [
+ "windows-link 0.1.3",
+]
+
+[[package]]
+name = "windows-strings"
+version = "0.5.1"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "7837d08f69c77cf6b07689544538e017c1bfcf57e34b4c0ff58e6c2cd3b37091"
+dependencies = [
+ "windows-link 0.2.1",
+]
+
+[[package]]
+name = "windows-sys"
+version = "0.45.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "75283be5efb2831d37ea142365f009c02ec203cd29a3ebecbc093d52315b66d0"
+dependencies = [
+ "windows-targets 0.42.2",
+]
+
+[[package]]
+name = "windows-sys"
+version = "0.52.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "282be5f36a8ce781fad8c8ae18fa3f9beff57ec1b52cb3de0789201425d9a33d"
+dependencies = [
+ "windows-targets 0.52.6",
+]
+
+[[package]]
+name = "windows-sys"
+version = "0.59.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "1e38bc4d79ed67fd075bcc251a1c39b32a1776bbe92e5bef1f0bf1f8c531853b"
+dependencies = [
+ "windows-targets 0.52.6",
+]
+
+[[package]]
+name = "windows-sys"
+version = "0.60.2"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "f2f500e4d28234f72040990ec9d39e3a6b950f9f22d3dba18416c35882612bcb"
+dependencies = [
+ "windows-targets 0.53.5",
+]
+
+[[package]]
+name = "windows-sys"
+version = "0.61.2"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "ae137229bcbd6cdf0f7b80a31df61766145077ddf49416a728b02cb3921ff3fc"
+dependencies = [
+ "windows-link 0.2.1",
+]
+
+[[package]]
+name = "windows-targets"
+version = "0.42.2"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "8e5180c00cd44c9b1c88adb3693291f1cd93605ded80c250a75d472756b4d071"
+dependencies = [
+ "windows_aarch64_gnullvm 0.42.2",
+ "windows_aarch64_msvc 0.42.2",
+ "windows_i686_gnu 0.42.2",
+ "windows_i686_msvc 0.42.2",
+ "windows_x86_64_gnu 0.42.2",
+ "windows_x86_64_gnullvm 0.42.2",
+ "windows_x86_64_msvc 0.42.2",
+]
+
+[[package]]
+name = "windows-targets"
+version = "0.52.6"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "9b724f72796e036ab90c1021d4780d4d3d648aca59e491e6b98e725b84e99973"
+dependencies = [
+ "windows_aarch64_gnullvm 0.52.6",
+ "windows_aarch64_msvc 0.52.6",
+ "windows_i686_gnu 0.52.6",
+ "windows_i686_gnullvm 0.52.6",
+ "windows_i686_msvc 0.52.6",
+ "windows_x86_64_gnu 0.52.6",
+ "windows_x86_64_gnullvm 0.52.6",
+ "windows_x86_64_msvc 0.52.6",
+]
+
+[[package]]
+name = "windows-targets"
+version = "0.53.5"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "4945f9f551b88e0d65f3db0bc25c33b8acea4d9e41163edf90dcd0b19f9069f3"
+dependencies = [
+ "windows-link 0.2.1",
+ "windows_aarch64_gnullvm 0.53.1",
+ "windows_aarch64_msvc 0.53.1",
+ "windows_i686_gnu 0.53.1",
+ "windows_i686_gnullvm 0.53.1",
+ "windows_i686_msvc 0.53.1",
+ "windows_x86_64_gnu 0.53.1",
+ "windows_x86_64_gnullvm 0.53.1",
+ "windows_x86_64_msvc 0.53.1",
+]
+
+[[package]]
+name = "windows-threading"
+version = "0.1.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "b66463ad2e0ea3bbf808b7f1d371311c80e115c0b71d60efc142cafbcfb057a6"
+dependencies = [
+ "windows-link 0.1.3",
+]
+
+[[package]]
+name = "windows-threading"
+version = "0.2.1"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "3949bd5b99cafdf1c7ca86b43ca564028dfe27d66958f2470940f73d86d75b37"
+dependencies = [
+ "windows-link 0.2.1",
+]
+
+[[package]]
+name = "windows-version"
+version = "0.1.7"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "e4060a1da109b9d0326b7262c8e12c84df67cc0dbc9e33cf49e01ccc2eb63631"
+dependencies = [
+ "windows-link 0.2.1",
+]
+
+[[package]]
+name = "windows_aarch64_gnullvm"
+version = "0.42.2"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "597a5118570b68bc08d8d59125332c54f1ba9d9adeedeef5b99b02ba2b0698f8"
+
+[[package]]
+name = "windows_aarch64_gnullvm"
+version = "0.52.6"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "32a4622180e7a0ec044bb555404c800bc9fd9ec262ec147edd5989ccd0c02cd3"
+
+[[package]]
+name = "windows_aarch64_gnullvm"
+version = "0.53.1"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "a9d8416fa8b42f5c947f8482c43e7d89e73a173cead56d044f6a56104a6d1b53"
+
+[[package]]
+name = "windows_aarch64_msvc"
+version = "0.42.2"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "e08e8864a60f06ef0d0ff4ba04124db8b0fb3be5776a5cd47641e942e58c4d43"
+
+[[package]]
+name = "windows_aarch64_msvc"
+version = "0.52.6"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "09ec2a7bb152e2252b53fa7803150007879548bc709c039df7627cabbd05d469"
+
+[[package]]
+name = "windows_aarch64_msvc"
+version = "0.53.1"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "b9d782e804c2f632e395708e99a94275910eb9100b2114651e04744e9b125006"
+
+[[package]]
+name = "windows_i686_gnu"
+version = "0.42.2"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "c61d927d8da41da96a81f029489353e68739737d3beca43145c8afec9a31a84f"
+
+[[package]]
+name = "windows_i686_gnu"
+version = "0.52.6"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "8e9b5ad5ab802e97eb8e295ac6720e509ee4c243f69d781394014ebfe8bbfa0b"
+
+[[package]]
+name = "windows_i686_gnu"
+version = "0.53.1"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "960e6da069d81e09becb0ca57a65220ddff016ff2d6af6a223cf372a506593a3"
+
+[[package]]
+name = "windows_i686_gnullvm"
+version = "0.52.6"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "0eee52d38c090b3caa76c563b86c3a4bd71ef1a819287c19d586d7334ae8ed66"
+
+[[package]]
+name = "windows_i686_gnullvm"
+version = "0.53.1"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "fa7359d10048f68ab8b09fa71c3daccfb0e9b559aed648a8f95469c27057180c"
+
+[[package]]
+name = "windows_i686_msvc"
+version = "0.42.2"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "44d840b6ec649f480a41c8d80f9c65108b92d89345dd94027bfe06ac444d1060"
+
+[[package]]
+name = "windows_i686_msvc"
+version = "0.52.6"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "240948bc05c5e7c6dabba28bf89d89ffce3e303022809e73deaefe4f6ec56c66"
+
+[[package]]
+name = "windows_i686_msvc"
+version = "0.53.1"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "1e7ac75179f18232fe9c285163565a57ef8d3c89254a30685b57d83a38d326c2"
+
+[[package]]
+name = "windows_x86_64_gnu"
+version = "0.42.2"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "8de912b8b8feb55c064867cf047dda097f92d51efad5b491dfb98f6bbb70cb36"
+
+[[package]]
+name = "windows_x86_64_gnu"
+version = "0.52.6"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "147a5c80aabfbf0c7d901cb5895d1de30ef2907eb21fbbab29ca94c5b08b1a78"
+
+[[package]]
+name = "windows_x86_64_gnu"
+version = "0.53.1"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "9c3842cdd74a865a8066ab39c8a7a473c0778a3f29370b5fd6b4b9aa7df4a499"
+
+[[package]]
+name = "windows_x86_64_gnullvm"
+version = "0.42.2"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "26d41b46a36d453748aedef1486d5c7a85db22e56aff34643984ea85514e94a3"
+
+[[package]]
+name = "windows_x86_64_gnullvm"
+version = "0.52.6"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "24d5b23dc417412679681396f2b49f3de8c1473deb516bd34410872eff51ed0d"
+
+[[package]]
+name = "windows_x86_64_gnullvm"
+version = "0.53.1"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "0ffa179e2d07eee8ad8f57493436566c7cc30ac536a3379fdf008f47f6bb7ae1"
+
+[[package]]
+name = "windows_x86_64_msvc"
+version = "0.42.2"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "9aec5da331524158c6d1a4ac0ab1541149c0b9505fde06423b02f5ef0106b9f0"
+
+[[package]]
+name = "windows_x86_64_msvc"
+version = "0.52.6"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "589f6da84c646204747d1270a2a5661ea66ed1cced2631d546fdfb155959f9ec"
+
+[[package]]
+name = "windows_x86_64_msvc"
+version = "0.53.1"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "d6bbff5f0aada427a1e5a6da5f1f98158182f26556f345ac9e04d36d0ebed650"
+
+[[package]]
+name = "winnow"
+version = "0.5.40"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "f593a95398737aeed53e489c785df13f3618e41dbcd6718c6addbf1395aa6876"
+dependencies = [
+ "memchr",
+]
+
+[[package]]
+name = "winnow"
+version = "0.7.15"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "df79d97927682d2fd8adb29682d1140b343be4ac0f08fd68b7765d9c059d3945"
+dependencies = [
+ "memchr",
+]
+
+[[package]]
+name = "winnow"
+version = "1.0.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "a90e88e4667264a994d34e6d1ab2d26d398dcdca8b7f52bec8668957517fc7d8"
+dependencies = [
+ "memchr",
+]
+
+[[package]]
+name = "winreg"
+version = "0.55.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "cb5a765337c50e9ec252c2069be9bf91c7df47afb103b642ba3a53bf8101be97"
+dependencies = [
+ "cfg-if",
+ "windows-sys 0.59.0",
+]
+
+[[package]]
+name = "wit-bindgen"
+version = "0.51.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "d7249219f66ced02969388cf2bb044a09756a083d0fab1e566056b04d9fbcaa5"
+dependencies = [
+ "wit-bindgen-rust-macro",
+]
+
+[[package]]
+name = "wit-bindgen-core"
+version = "0.51.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "ea61de684c3ea68cb082b7a88508a8b27fcc8b797d738bfc99a82facf1d752dc"
+dependencies = [
+ "anyhow",
+ "heck 0.5.0",
+ "wit-parser",
+]
+
+[[package]]
+name = "wit-bindgen-rust"
+version = "0.51.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "b7c566e0f4b284dd6561c786d9cb0142da491f46a9fbed79ea69cdad5db17f21"
+dependencies = [
+ "anyhow",
+ "heck 0.5.0",
+ "indexmap 2.13.0",
+ "prettyplease",
+ "syn 2.0.117",
+ "wasm-metadata",
+ "wit-bindgen-core",
+ "wit-component",
+]
+
+[[package]]
+name = "wit-bindgen-rust-macro"
+version = "0.51.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "0c0f9bfd77e6a48eccf51359e3ae77140a7f50b1e2ebfe62422d8afdaffab17a"
+dependencies = [
+ "anyhow",
+ "prettyplease",
+ "proc-macro2",
+ "quote",
+ "syn 2.0.117",
+ "wit-bindgen-core",
+ "wit-bindgen-rust",
+]
+
+[[package]]
+name = "wit-component"
+version = "0.244.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "9d66ea20e9553b30172b5e831994e35fbde2d165325bec84fc43dbf6f4eb9cb2"
+dependencies = [
+ "anyhow",
+ "bitflags 2.11.0",
+ "indexmap 2.13.0",
+ "log",
+ "serde",
+ "serde_derive",
+ "serde_json",
+ "wasm-encoder",
+ "wasm-metadata",
+ "wasmparser",
+ "wit-parser",
+]
+
+[[package]]
+name = "wit-parser"
+version = "0.244.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "ecc8ac4bc1dc3381b7f59c34f00b67e18f910c2c0f50015669dde7def656a736"
+dependencies = [
+ "anyhow",
+ "id-arena",
+ "indexmap 2.13.0",
+ "log",
+ "semver",
+ "serde",
+ "serde_derive",
+ "serde_json",
+ "unicode-xid",
+ "wasmparser",
+]
+
+[[package]]
+name = "writeable"
+version = "0.6.2"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "9edde0db4769d2dc68579893f2306b26c6ecfbe0ef499b013d731b7b9247e0b9"
+
+[[package]]
+name = "wry"
+version = "0.54.4"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "e5a8135d8676225e5744de000d4dff5a082501bf7db6a1c1495034f8c314edbc"
+dependencies = [
+ "base64 0.22.1",
+ "block2",
+ "cookie",
+ "crossbeam-channel",
+ "dirs",
+ "dom_query",
+ "dpi",
+ "dunce",
+ "gdkx11",
+ "gtk",
+ "http",
+ "javascriptcore-rs",
+ "jni",
+ "libc",
+ "ndk",
+ "objc2",
+ "objc2-app-kit",
+ "objc2-core-foundation",
+ "objc2-foundation",
+ "objc2-ui-kit",
+ "objc2-web-kit",
+ "once_cell",
+ "percent-encoding",
+ "raw-window-handle",
+ "sha2",
+ "soup3",
+ "tao-macros",
+ "thiserror 2.0.18",
+ "url",
+ "webkit2gtk",
+ "webkit2gtk-sys",
+ "webview2-com",
+ "windows 0.61.3",
+ "windows-core 0.61.2",
+ "windows-version",
+ "x11-dl",
+]
+
+[[package]]
+name = "x11"
+version = "2.21.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "502da5464ccd04011667b11c435cb992822c2c0dbde1770c988480d312a0db2e"
+dependencies = [
+ "libc",
+ "pkg-config",
+]
+
+[[package]]
+name = "x11-dl"
+version = "2.21.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "38735924fedd5314a6e548792904ed8c6de6636285cb9fec04d5b1db85c1516f"
+dependencies = [
+ "libc",
+ "once_cell",
+ "pkg-config",
+]
+
+[[package]]
+name = "xattr"
+version = "1.6.1"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "32e45ad4206f6d2479085147f02bc2ef834ac85886624a23575ae137c8aa8156"
+dependencies = [
+ "libc",
+ "rustix",
+]
+
+[[package]]
+name = "yoke"
+version = "0.8.1"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "72d6e5c6afb84d73944e5cedb052c4680d5657337201555f9f2a16b7406d4954"
+dependencies = [
+ "stable_deref_trait",
+ "yoke-derive",
+ "zerofrom",
+]
+
+[[package]]
+name = "yoke-derive"
+version = "0.8.1"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "b659052874eb698efe5b9e8cf382204678a0086ebf46982b79d6ca3182927e5d"
+dependencies = [
+ "proc-macro2",
+ "quote",
+ "syn 2.0.117",
+ "synstructure",
+]
+
+[[package]]
+name = "zbus"
+version = "5.14.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "ca82f95dbd3943a40a53cfded6c2d0a2ca26192011846a1810c4256ef92c60bc"
+dependencies = [
+ "async-broadcast",
+ "async-executor",
+ "async-io",
+ "async-lock",
+ "async-process",
+ "async-recursion",
+ "async-task",
+ "async-trait",
+ "blocking",
+ "enumflags2",
+ "event-listener",
+ "futures-core",
+ "futures-lite",
+ "hex",
+ "libc",
+ "ordered-stream",
+ "rustix",
+ "serde",
+ "serde_repr",
+ "tracing",
+ "uds_windows",
+ "uuid",
+ "windows-sys 0.61.2",
+ "winnow 0.7.15",
+ "zbus_macros",
+ "zbus_names",
+ "zvariant",
+]
+
+[[package]]
+name = "zbus_macros"
+version = "5.14.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "897e79616e84aac4b2c46e9132a4f63b93105d54fe8c0e8f6bffc21fa8d49222"
+dependencies = [
+ "proc-macro-crate 3.5.0",
+ "proc-macro2",
+ "quote",
+ "syn 2.0.117",
+ "zbus_names",
+ "zvariant",
+ "zvariant_utils",
+]
+
+[[package]]
+name = "zbus_names"
+version = "4.3.1"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "ffd8af6d5b78619bab301ff3c560a5bd22426150253db278f164d6cf3b72c50f"
+dependencies = [
+ "serde",
+ "winnow 0.7.15",
+ "zvariant",
+]
+
+[[package]]
+name = "zerocopy"
+version = "0.8.47"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "efbb2a062be311f2ba113ce66f697a4dc589f85e78a4aea276200804cea0ed87"
+dependencies = [
+ "zerocopy-derive",
+]
+
+[[package]]
+name = "zerocopy-derive"
+version = "0.8.47"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "0e8bc7269b54418e7aeeef514aa68f8690b8c0489a06b0136e5f57c4c5ccab89"
+dependencies = [
+ "proc-macro2",
+ "quote",
+ "syn 2.0.117",
+]
+
+[[package]]
+name = "zerofrom"
+version = "0.1.6"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "50cc42e0333e05660c3587f3bf9d0478688e15d870fab3346451ce7f8c9fbea5"
+dependencies = [
+ "zerofrom-derive",
+]
+
+[[package]]
+name = "zerofrom-derive"
+version = "0.1.6"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "d71e5d6e06ab090c67b5e44993ec16b72dcbaabc526db883a360057678b48502"
+dependencies = [
+ "proc-macro2",
+ "quote",
+ "syn 2.0.117",
+ "synstructure",
+]
+
+[[package]]
+name = "zeroize"
+version = "1.8.2"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "b97154e67e32c85465826e8bcc1c59429aaaf107c1e4a9e53c8d8ccd5eff88d0"
+
+[[package]]
+name = "zerotrie"
+version = "0.2.3"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "2a59c17a5562d507e4b54960e8569ebee33bee890c70aa3fe7b97e85a9fd7851"
+dependencies = [
+ "displaydoc",
+ "yoke",
+ "zerofrom",
+]
+
+[[package]]
+name = "zerovec"
+version = "0.11.5"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "6c28719294829477f525be0186d13efa9a3c602f7ec202ca9e353d310fb9a002"
+dependencies = [
+ "yoke",
+ "zerofrom",
+ "zerovec-derive",
+]
+
+[[package]]
+name = "zerovec-derive"
+version = "0.11.2"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "eadce39539ca5cb3985590102671f2567e659fca9666581ad3411d59207951f3"
+dependencies = [
+ "proc-macro2",
+ "quote",
+ "syn 2.0.117",
+]
+
+[[package]]
+name = "zip"
+version = "4.6.1"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "caa8cd6af31c3b31c6631b8f483848b91589021b28fffe50adada48d4f4d2ed1"
+dependencies = [
+ "arbitrary",
+ "crc32fast",
+ "indexmap 2.13.0",
+ "memchr",
+]
+
+[[package]]
+name = "zmij"
+version = "1.0.21"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "b8848ee67ecc8aedbaf3e4122217aff892639231befc6a1b58d29fff4c2cabaa"
+
+[[package]]
+name = "zvariant"
+version = "5.10.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "5708299b21903bbe348e94729f22c49c55d04720a004aa350f1f9c122fd2540b"
+dependencies = [
+ "endi",
+ "enumflags2",
+ "serde",
+ "winnow 0.7.15",
+ "zvariant_derive",
+ "zvariant_utils",
+]
+
+[[package]]
+name = "zvariant_derive"
+version = "5.10.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "5b59b012ebe9c46656f9cc08d8da8b4c726510aef12559da3e5f1bf72780752c"
+dependencies = [
+ "proc-macro-crate 3.5.0",
+ "proc-macro2",
+ "quote",
+ "syn 2.0.117",
+ "zvariant_utils",
+]
+
+[[package]]
+name = "zvariant_utils"
+version = "3.3.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "f75c23a64ef8f40f13a6989991e643554d9bef1d682a281160cf0c1bc389c5e9"
+dependencies = [
+ "proc-macro2",
+ "quote",
+ "serde",
+ "syn 2.0.117",
+ "winnow 0.7.15",
+]
diff --git a/studio/src-tauri/Cargo.toml b/studio/src-tauri/Cargo.toml
new file mode 100644
index 0000000000..ccb07cb9e2
--- /dev/null
+++ b/studio/src-tauri/Cargo.toml
@@ -0,0 +1,42 @@
+[package]
+name = "unsloth-studio"
+version = "2026.4.7"
+description = "Unsloth Studio Desktop App"
+authors = ["Unsloth AI"]
+edition = "2021"
+
+[dependencies]
+tauri = { version = "2", features = ["tray-icon"] }
+tauri-plugin-single-instance = "2"
+tauri-plugin-process = "2"
+serde = { version = "1", features = ["derive"] }
+serde_json = "1"
+reqwest = { version = "0.12", features = ["json"] }
+tokio = { version = "1", features = ["full"] }
+log = "0.4"
+simplelog = "0.12"
+dirs = "6"
+regex = "1"
+open = "5"
+process-wrap = { version = "9", features = ["std"] }
+fix-path-env = { git = "https://github.com/tauri-apps/fix-path-env-rs" }
+tauri-plugin-opener = "2.5.3"
+tauri-plugin-updater = "2"
+rand = "0.10.0"
+
+[target.'cfg(unix)'.dependencies]
+libc = "0.2"
+
+[target.'cfg(target_os = "linux")'.dependencies]
+elevated-command = "1.1.2"
+
+[target.'cfg(windows)'.dependencies]
+windows = { version = "0.62.2", features = ["Win32_System_Threading"] }
+windows-sys = { version = "0.59", features = ["Win32_Foundation", "Win32_Security", "Win32_System_Console", "Win32_System_JobObjects", "Win32_System_Threading"] }
+
+[build-dependencies]
+tauri-build = { version = "2", features = [] }
+
+[features]
+default = ["custom-protocol"]
+custom-protocol = ["tauri/custom-protocol"]
diff --git a/studio/src-tauri/Entitlements.plist b/studio/src-tauri/Entitlements.plist
new file mode 100644
index 0000000000..10e6df3427
--- /dev/null
+++ b/studio/src-tauri/Entitlements.plist
@@ -0,0 +1,15 @@
+<?xml version="1.0" encoding="UTF-8"?>
+<!DOCTYPE plist PUBLIC "-//Apple//DTD PLIST 1.0//EN" "http://www.apple.com/DTDs/PropertyList-1.0.dtd">
+<plist version="1.0">
+<dict>
+    <!-- Allow spawning child processes (Python backend, install scripts) -->
+    <key>com.apple.security.cs.allow-unsigned-executable-memory</key>
+    <true/>
+    <!-- Network access for backend health checks and HuggingFace API -->
+    <key>com.apple.security.network.client</key>
+    <true/>
+    <!-- Allow loading Python/venv libraries not signed by us -->
+    <key>com.apple.security.cs.disable-library-validation</key>
+    <true/>
+</dict>
+</plist>
diff --git a/studio/src-tauri/build.rs b/studio/src-tauri/build.rs
new file mode 100644
index 0000000000..d860e1e6a7
--- /dev/null
+++ b/studio/src-tauri/build.rs
@@ -0,0 +1,3 @@
+fn main() {
+    tauri_build::build()
+}
diff --git a/studio/src-tauri/capabilities/default.json b/studio/src-tauri/capabilities/default.json
new file mode 100644
index 0000000000..7b6033c721
--- /dev/null
+++ b/studio/src-tauri/capabilities/default.json
@@ -0,0 +1,24 @@
+{
+  "identifier": "default",
+  "description": "Default capabilities for Unsloth Studio",
+  "windows": ["main"],
+  "permissions": [
+    "core:default",
+    "core:window:default",
+    "core:window:allow-show",
+    "core:window:allow-hide",
+    "core:window:allow-set-focus",
+    "core:window:allow-set-size",
+    "core:window:allow-set-resizable",
+    "core:window:allow-set-size-constraints",
+    "core:window:allow-center",
+    "core:window:allow-current-monitor",
+    "core:tray:default",
+    "process:default",
+    {
+      "identifier": "opener:allow-open-url",
+      "allow": [{ "url": "https://*" }, { "url": "http://*" }, { "url": "mailto:*" }]
+    },
+    "updater:default"
+  ]
+}
diff --git a/studio/src-tauri/icons/128x128.png b/studio/src-tauri/icons/128x128.png
new file mode 100644
index 0000000000000000000000000000000000000000..c1dae04e6dfaf806cbbb4dc6ea25c1e38276f07e
GIT binary patch
literal 9194
zcma)iMOYk6wCvzckRMNA@W22e*x>FEAh^2(2oOBjFlcc1!GpWIyK8U@g9eu%6YOz!
zZ*#xzF1mZsz3Mtur~A|iS5uM0!=}Ur004Lj^3v+BEA)Q`Ci?5#FKRXp0MOSeNK0sb
zUp>vjNK0B+7@nRY^$0L3b7MhfRVPtH!tscvq@t`%U7qn3Z7K;5@1){|41GDNrSg%w
z#?sc5mLa5iFMrOKWwqve;Au1U-7}4S-Xqn$YPzLgh})_4U|r+6cs=uW#|dB2`q54f
z$c;vUJ`#nCVksMn#6{l^NRKyzPK<DhTU(L?xw&xTDE`0gp9-;=CJ(5VmnPbW=O5_b
zCrdojOmK>8FnTwscdBD`l-5Zqg2oEzN_O$*VbfRjt_6C)wRh96CuThrm!U-61$BlI
z<BE)65O4$5oRa;|4rTzHtasfgGtq97=rU^8B>id#l4+gPq(tS$tzmql&12!`mHxGY
zEzx*#ZDiKy-?&*;Q=-V)O;XB=m%0qf#YO~I>)!C2y&(!3MOeop(>SU?H~Q!9Hs8dE
zyWErIvc$}W8v@2>{5g(lujp=vyt&^;r?(02dxb93CoFS{j1n4}7@ty*Tr@kJup|G?
zp8AZ1bjpoM7~GJKmbLu?JFdT<?@s5)`GipJdU^I4otjGN#}`t*U3XQzLpVI$rG}5&
zj91(YPH}v(O8^$LRNg}A9cL<kk(3R$?~;vfpu9=hX~V0y+88q}RL<o5ob7k3=k;*f
zY;}9Gq9xkB6C<>nZj&8v5hE~^<vFaKwfE3CUTCF~t;zp}cG*%mIJ3y~mw>%7I|8q)
zx!!%xVj}wUyXm9T1QT+^b4H{6-mDyT{KI*-N!Rng<1&xHDqq=p>q*Q0P&_V&Ri0r0
z(<j9-r2^e*ovMj`#3jMWszIIR$m(2zWc}N16vw$Gog_gphgclF5f1iVn97sod8M~P
zb>QM$vz3tZlrK<;m6dh3s{Jfs?b!pyW>~{HARGYo(bCp#OZ|wnwP6i&*sY%yP2_V}
zJrrQIavY^hFEU12qVE1H=cQy4skkT*R3}~?ZTWjDj;$9LrC=!+G_PQ+p`HJaUnjKr
zjIO=$nB$?`kw$Y_0^9W({IHS4qOtpTUNyGjJ~K0uRXSEak(2vt`)SkFpW+<Fa_!2p
zPO>U7_Z#!{4!eLT6D|!=j_h+y$viGe742KnLOMzj6iWdx2arl@pt^jS;dH@!OBD+u
z8=F@r*G|3g>^Pe*a~13VC{@+{?0t8>zHZpu_7jU{5+Fz8*KpGP@_gnOOD$4hpUN>;
zXaKg2<O$<VS|<J_?`yf4%4N2~rdQP@7r@1Amv4>}hZ&uLG}Bqg3STOrT56UtIB3tQ
za9rnO=VKnqY<1oHsgSt)bhqURr;3DZw7F|Z`&1_85iyY$P75sNNn!k%FOJ2f;#2n6
z^7CA=S~4m$SgL|OK?2GmmbB>B5%a@bu1|`|n3I{6(2Y1VD1WrjGl*mDy@O_j$+}Ha
z)Um*NC96N1VT+nE1Yza%j5yd!R27ZCIbL$IUan0DawfO6rb2zk^EQg*BQ_3>m5x;)
zeGI=O<Y8?~ap}Ev51*I$^ODO()Sac-sJ>q)a&Z?WeauV=c@M6z&3fsGnDcQje$*dr
zH3pTGP^0@dVvlNP?dPI$kBxRK52!YSl(RuBX%`iojsT6c;EO`62>@}X-|b4>nq#{j
z+vvuzZ!%`{L}tMR=Vb4+3@afejOPmvt|`ZYKFOU>iG-)#*2XS52|{#x%lG<I>)xWl
zURF4<|7T*mUn`1ui}Tn)gi*9(<Ky(hcdPav5IZz4?=h&q=OW=x;UG-+3|oI6gZHRz
z81(n{!K_YspE;rRXUS0A`};GSvH&uN(U(7}{w}%Eggt+~YXS!CDtuNvxBT`(sT||B
z`+xkn2Tu&x_}!EVa~|M?<l#srlo6H$i}w?YS#GDIgLd<1e+a<W$|O5-e9l_yuBU{A
z;PppoUzsw_QK@=S|K0U<I>)Yq?tC}B4n{-WkKbK_k|oZld`s4a?g_LoNK^(dPrB`}
zpvkc#7?nP5#q^fau*$5M-Fqsh4qBURuOc?x8X%p^vH2)Cbek$?qunzV<3F!;RU@S+
z@q!Hu(Ef;mhDnkYY*9!xkM=ZBT#Cp)8JTs1q!Gq{^q-us5ae=@Y`hnP<|Hikh1%ge
z@2%P0Sm(6Xo!biv{x{Z%zPWY~<Z`n_BTH!f?JiF)!xsW9aXQ&r(fi!|n5#Jsi$Ygp
zC1xg_YGd{vG?12#hqHFaK_Q{CI`yXAZKa@B2!~rM@U4F7nN;>H=H&=)6AXGBv8M1s
z88CT*dz&Sn9v~s7R*zxCbOSlp*VA?LDnbJAdKdHFMg#aoVQ(-NvE@jN)b4OP_Xc+}
zHcuj0F7thR!`ab(0WE_>Zr9**M%^c_rekr0TP$~Br3!!;$quZ{zOPqr-5Smc0wosl
z=spP(Pv2gj<XDx^MdbKDx5A$G^#o$DLP5kl!Gf5-nA_x9n#uzOb^8P)I^8^a2>l1b
z2<v<iE(h~qUid?h%?`y<h2D-h{kp-oz<s^etE(xCtWN;Uly}Jub%WgEDUa}L^V0*+
zlIxy;>;4F}&k60Ts<;sE1hIZ{zlajHS5OF#q88CFQq4KrNB;Ud-Au0BPx=H%o?O%w
z!YvL!Q5{^{!k}iLuf6R@-`ey!XCF^te^?+8%fONsM&h9>>Lt&`e6Ne-s%0VxRHO0s
zm-!<rP`va_K!)B6J&@Y_Fpn^6i#$9h*&6#|LiizpQE3<4jre4VKlG*Vh-B@IcChfq
z;y?5`hfCtIjn+>(hzDrUTT1kOF~@a&#P6Utl$E_U5;Q2ebGF2s7$$x<$(e3*3bfS{
zpZh?~RO}C)hdWVyZa+f;++=UXTjW!{)o6`LA>w}UVQL@cXT-F>;0wReKy!&H*TD4j
zbS!E*!O|Z-+#K&CmeihvN(2fk)7}&kMjP%=mynVACTa8${2&qkv-Tffx9U1~FzcMK
zU>WvFG0EeW`ld|AOcy%O22P!fFHh(GMin%dUnO$f%+U^V{JVT#^IHt79}m*LdURNo
zcC$d1IxvnS)hVluTZ%W!JM#!a&$lcZ8oRewm~laxMq8XSvGskhnecvqSk-NF6!LI-
zfGMex*E^e;XkR!d`+I$;#U|5PcK2RJe;oSHuzZjufM<mvpds9!@uG3<d<U0;>G+iu
zZlCKk^&e-0<-VHuoHpWr8kn}0{E4NNXbdi5_z+iXJI}<z%9`-1CE`rBZRva1`hR2X
zi^pU+r+@#96-{VA?-Fu(Ts^(>XbsUbh(H~|$MC{GTGqiooLPHk(CSiX2tn+z^xl5@
z@4?KB>8xR1#kKq<0#x;F3&kR=X`g)ov?!v>^!YU&tVZH!&vcEAjV<?L-wMo?hTLuX
zsKy9ijbPGWQ{iYZgGiJ~9&f7slM9tn+wOdt&lxd127t=3tnvs7B;P9@o##5ycA1#?
zbo*yWKJIVgtZ?{3tE)}x??9ZlX@S9i7Rpl%8lNp?ZJrYPLcG7MH93%=*LJ@`(dc&c
zi^aMRqg#%bU<R=db9!4J&o<^<FO(3Eeft21n=_AoHC(juWLB6r1EZ;T{?oBXoBW*h
zAh0pmvozcmEtB*;fhYRVz@ahut^2RfQ+DXGT~D`bR<8*-l3;|*W6ShfXaDuCSi4^_
z-%yMl1Pq)<e}3B1tNXEmaf3pDCHs9ZS!Z2@*J)Fjj^yMaTrt4syi*5qzm0JRA<I&_
z2+-Oq^~6CbHGV3P$XT;L7^yXIY#pproL_JbkT)j$5u%=+bdc?%5+ib}Tbxr4yI8C=
z(20NRdr($h<&xfk>+)>z5ujek^kGnpW=ZxFz|L8<Bp2q`rqHwnd6*>UGG^&%6-bS~
zc2k+BMMEw{^&2^(s1&qM(>~#}whv1sNgX+ARybU#FYfqvR6YfjzEiKVD(6fYX}k-<
zc;mrne7}<U<5wN1x`QiFB(9DG=kj>2P$_F&^BUbZ8CAo``%j??<LB?I)h1$|r{7Oe
zVCBr&-zBYy1Q&ZVsmq>l0gLJsY}f^Jw;1~(rE9-O;xQJynGOl3kHA^u1JM-RL*dI(
z_yPqHHxA{Tiv<}Uf5<@xfW4^Bb}5Zd`?*2*5w=O{s0k&X?e0c=gt5qxmX<o=+qeQg
z@=YlIYmavI#<$Pp1H^;uF5Af(=ISaP)@yH$&V&99+9P1uorn36yJ9a0Rq5)X9e-Hw
z58fAe<cBdmn=F!2V$OJFV445RGa+b`{FkPjWa1wMw00ZJu%nN#{3_+&N&#Aw{1zvg
zzgFDAEVIFLnli5#Aww2Hy0z6r-C8Z;*OQk!MxvROQEd@^^v@2>8!7+&XRmAuU&-;6
zx^0OuS`s~8uC<u!1@<$a{41drMz?5cQ$*T#yo3*4l~JZgNuH|L#cGRHds<yYhar^&
zPipO)w^ViWnD;2EN*`z>It?5o<Ib2rd+`L@<GLS}77I34H8k+_{U>x%BQ=&E0KpQZ
zM!}@VCHx>I_Vjl?{_dCD{vX5QvId4Y?8{jBqsrx*8f{t>yUT<@j+e?Rc#c9Kphf%T
z`L3HD)WZ}zJyFJ2YC25R7;)OA#*Oi1kz-K|P6@850xf_6r<>oWJJUVk+;dWIL^KCE
zD~?5wMJKbsOPw@T7lN)?th@yBA@^G)|NLzxZ9yYgBu-rqQ|%n%!%$e8GQ*C^%fHm)
znSiw~!^+heFdMb1caa>^5jrxae?PmR8=LX_It+l{ol`+QoH`CReb0lBfmZzJ6^1(1
zBfkJL)x?|l>?O$21YAr1%@@###wB8}=jb;RI=s>!DSc5ZD&uh2jw!5vpa-?6+;-vW
za+qd3)dcd#?m2yglwF@c6{jO-0jDiBY>boXx9BBhi-!Yfz5wO!lQ!{0P(%Ob_T4&;
zdN8hyV!!Ss)0G!B(8~NpzU@1lZi0n1@?+iU;#J8l{fOOR8$j-{h8$t~%eF47$UC*{
znCr`Lag60<)FmtN!+?K6n9Ab2A!NqiwsocY%c#N}eVrsrMFWTe%<L_6U)I4}zV-G@
zGP0_lE_kP<hu&ljn_^a7&Z~+EV79hpqt>SV7n)PW;jCo!T)o5=FTyO6Nu<fif{v;8
zV#U>>C5QO7YnxCw>Yr;CgpL=XyXo%5uZ;!@BojouN^+!a14Vq4nza>onrYh5RQi!J
zwGi|6*SPNRZEUNpAWrL8E!n9acC?8;a_J3#<3ghsnneU(7dfBXm-kGUv-DSmO{KPZ
zcHnfn`Akc^9<E8<KN}sD^rAbFAG*TA=2odE(t@<`=Q&$zDf!dWc%biB#|}PpB163p
z_7J}^^V=@FfLWj@K@AP+l}_iz&yVG1R3#sc!JFSL7)U>pVcEJOLKca70GUopd>fW^
zw@j=_9R<RH;&uOWbZ_uO3}j+j1Ahn}`P0uukbvmI>ALpLRq?khh_JD-Gs|U*V$Kf|
z4dm29m!<LpidV%cq(#dY|KU+^DWDRLtajN?z<t<;q=o4xS;Au$nP`6o>9KK8J-JT>
z;hmzxUx;%L&p$%L&)eE;x^!ciyxq!3KX*`YTemPM;$SyAZFMP9P~eI`OVITeXPD}d
z|D6&#uju>F@8Q-R9r#%RXnz(-ElHXgQ7Aj%Qv$ktYIXZ#zFk|{)C-cAIGAIg0RD&U
z_)hjkQ4u^kO$mqqA5<54aO^84C@8<}hIyO>0>2{%o0_~zBBlg?3{PG|2OrCCrPz$K
z1@g#!4c&jCvlhMyE4T%?hLbqFLycIHhy-BvL1j3Syo+8v$1kEEXZv1yqbQOAmh)Xe
zNNk{*Y9~B-5Wh`XYpyu_$>VfYX0<atspdMCl6MR(^6@QnG?94;d-MK%o>6hpXpW$}
zU9$aeF8~$s$rYZco6v+lnJ}#}de4q-lbc=|(p5iRMWPbfxWXdfBzs9x@Nj%1mnf&L
zIZ39hZ)l#Q{RAG!Bg<b;e29w<4KmoP)^?rDKPAur6TC9a-))lq^5xIHOj;|`c`x}|
zRj(66CZBawZ|CnRBB-L3ZTkioK;gP>z=-R*spUiaVWUC9#kG#u_?y6}E_4|%sVief
zZCF8kF4a<Mi3(H0<qVUs;m|RpAVIeV#+1XP(+=PG-v>#(b67~AmVHFZEfi#eem1XE
zRKni&E$CQteu1dBl2=5rDQP)TRWd7UBQw~bph^{*pet0(U~g%rXsw{mwIxF&2y&cn
z7IcK**G|OSU28VL$Zpn?Czj$&IJU-w)?Uc0ygCc0xlNmg+*g#l>-8@~hy@+Uqaj%O
zo&9*>QugDGa4T+J=q+-IHD0jSH+OzVc9x0MAL5oHb#{0keIGA}>C30V@F+9R318U;
zt!S!2C%oo6Bl#iTB}0wDi@#VgmAyW3ibL*8Mr3Bl<2-fh&2BtRH8#B?yjaSbY7EOY
zy}%BD!TDk-cd`s|==%hH*@ie{-v(&3$4|Cr6<>hmY}&95Zcy@1fBPOhLEq;`gcd_+
zDL1Dd&?u(<$QdVGD(=NbPw3|J!Qq;2D<M|g!Z*iU;uo{P)a2B{%6{}<pcuaMJ9hJ@
z+z7<cEHEMGLT}?pa(c&Ddw8JPE>W>6<Uc4DkY71jD;DqFg<$@jq@A!klI>TpmHB!U
zsDA(H$|XT96#reDUdP4&|Ez2vH>vYqsb<_<OC`Z*j`!P{6&^jz=*m5fpZw%SZb>gh
zYMEMHOSTV0116Ta`_j%QSs5M`Taq-LQG$KlX8wx!*CO?hd@IGB3Q7BggnT3Wa1Ocs
zcpi?X=50+e-0`UH!p-*dz&Dg!)MTZcvV4}2XIwI>J^hbe$G$$_woz6YkdH#R^Do`%
zyxj%zhcqW>ar-5aal%Ael%h{NhbD``ZrYL$?La&MFrKh|y_={ih)MpTu*pY7CUOl5
z9_}5gv5D)?@ApigLYE<`90lLu_~rm`kZ9n)fsSu2_K-1YivoiEvK$A%J^@UL1ZcJI
zCQ%=!;s%f4Pfg~|U^7b1&}r1ihfLQg=MPo?^MbR<HOp0Bm8rq#K)xr%Oec%3d6YM!
zQ;wriNvWNzrZhyJK}8nMum$1%;NS7$gkfUhkk`MKvjpSudyg{`4-WRE#I>16vs;vR
zc!ko>HngRWlvuPmmO@5-;cK1>y9H|3p4|^d$pgyWvhjDEc1z?`^7lbLNPY|}Xr75l
zt<q^wyZ2O9EG3lq3|Vh-**~!+j+BC*MuXLM<XDplw~+Z#ENvm}TL3z3N%VMu>r@u%
z8xb^v08LP0Za^@wavWC#WMQ#H81`Y?jZM2zmGfm4VmjJwBGwlE7kO^TT^W_5`B)ZP
z=O*W*$ULIOEUJeM%c1S9wN%cxaPdwab54_0X|DzeEe)->N=IBl*n7!FbUFm7SfRnb
zb}!&oAV153Img#!x8oP1eg$yR-EU{Q+hpA7LTPRkz3!Vc9bs>Fr?+fOB@Q?SlP?(c
zHv|g(@=OkwW-mWlrX~6j+e{j>MR;5s!L+&&E|+y7#1K!S4KFggh>r-lM?0bACX-(v
zk7dhWGV$L)OU68?Q)j=8=~hKot*fG1LY4zW^Tg6+A8_p-&j}=Xn-rxl=iS!n;UyB2
zD`MhF^d@=;eQGysN-ng!|Em#?txH@Dx!w?}bJTT5uV#v%F!Wfl;_#ZMN%5ezANYs6
z54^DS2w+&o!cCS`|3ZhJ>qis2M`blZZAU-3R--5jxmWvMKQT}!4qUM$;Iqzd{#Y=t
z-{{z=XR9TQ7JqA`K2V!Y>_Bfr9Qt2R&BS#FaZXRC&s~+dri?#2z2u{Gnb)H+lns+X
z%`G#pDzfi6KlH0ue`7?OI$cTyw!9!q(zjhj5*bIdy?p_k?1q}}^Vwhb^nc(Aeo0uC
zDB&|!FkRY{dCK8U0+yoC?V!6r!Xj~oMSB_SiN<?8fScl0Tu~aS`T-f<&j2p7`)#J7
zH5wN0v$V*3ZQ9U{|JXc<bSGy*Eum`ahi{Ix!)W*h?fB?S$>Fg&%pT=1enTc?p<GAJ
zbA~=tYm>MdTak!-3<oPUl7$tfaDuI~<Xu%w3b>dtyx-T7Oz$LUP~sdJr_-<7vGuDO
zJ*NqHu0kI9jl3v%0hq^W=YyGPzIB&O46lf3541(SAPOsv;p?J&R;StXr8hZbqif*;
zVYYiX?F;5MhCeuL$Fq?qYVr<B`NHT8f9=GIb&VOqnbvZte{z-bo0ioD@-jXBJWH^3
z83}Vh5rgS;n3-E3OEW7&F~QBO#&2lbx@Kb8rTw)NVR~y%S5#rKr{X_|a`(UMbQz`U
z1{G#dY4oV^b>TT#Z`9asE4q0jPj0GJa43|N1ikx6@W6Cs+b5375ig7sdE{L6%G@cd
z(>OOc*ep)AhCYlU{81(gDg}^coFksy$2183&lSB4bty$Ogiq9@^F5;#TtNwHy-MWR
z)W&dLjd`;NVr6rIaL>H=^IFF3n=2*i?Rt`)c`Y+7my#=v;D&R{mQJ(p<~K*|ph2L2
z`%igB6L5p*kiJp<aYVkfbw$Y-xpzF}N-w$z_*w}2Y9u}flid!Qo4%b<SsSbGk3kF)
zsDFD8j(aycAc37rLyQ+F%?z2@#p;it7N5;Cut?J+-Z*))#n&TWNFI_Fw6(`DYtw)*
z^D&&8A*WL0Q@nl*9Pz||+;wigc^lnAm5Job#P1I?+ADf@TWQXS-VbBy1CYCqTH`(-
zd~wY&Du@yU4r$5Xy}z>mdo=fR_>Ig1>Ye`&m?8ybdQWcTZ2I_UQ>ok|irLzkc~_m7
z-W`>%I=IL&wcsY_%3%jN220Vhxk<$GYco<Cp1x}{##C-!%o)<mR@jB+4R({(7in|`
ziMxcFfCo(K&s6bGJt__p5>Ckf(4@Yv8`|$tPR5mv{4kv!STgpRB=nyxYCN+3pd;0;
zOPvnZDyq;w^q{YnpCxq_DE&&LkUTRyc<oldE1+J6xxH<KJ`yQJ14=2yK`BtwSqp}>
z*k}>`-bS;G5S9HwM_WTxqrj=c!ciH%OrAM#I@)bRf=m!kEPl8g3zy-}E#SN+1CBvc
z4E+DepQR!4c1v4^50w}%HO2srTuCj5jMy@KRtEw#=E;tJbNV>svVY1fuWG*})}ol_
z{UCk+@gr_NM#N!i*la5b!=y21xC!ZylPMjP%(zy4{pq~ef>|RmY6u|VSH&`jQ=oP*
z-IP`-f4QF^@aGHZDNvDd$Pg{8OTT=BZS^L}_-GzJt)Z44;|Kle-W6`my(v!brAK!8
z-GjQ;wZQd3zdl;tm5PV|*H_n}&}2u#ZCkwo_sErx3<QqU0=9b{BOgI<WzWEZt;%s#
z>W}W*!)H&aerNq7qaO5;VCKM+!ps_Pg#1Aj|0GPI=yW34EXovKw?ddiuru_ylaU6@
zj>%ZywT<#XaUhYiIluq#Ae^y0dPbz4dT?p5u?#R6RpDA)GR5s-nCtmX6~boY#1@kq
z2$13(L}LqLB4pcy9D5R|YOmhc21`@w=-)RNZLRvDq{q&bF%79=SI3$N&O5x-N+0@E
zI6n-a<IuCz`{V89xcwoEU4ExZ1@E4uk=(XRrMet{@Et>A6mq;YIQB}1{XqOfqPGgU
zKsLH2#DGeyyjbv3m0H@7eB<q_<C4{m=ZGI<aMTr05`Y{6GSU5;#g~XlcwO>QP5?@;
zkGIU)WA&3V?cGor&bIb2f3TYEisTp>t!{!uK1_HJmi6~Q>}3Vb{ZIuNUroc&v?5s*
z%y}#$t{QOYQ_I8GrP|T2J*^~YgD8^(%_}~;aRY3X#{e3QSt~x<x)KK}Sw?j-?(N=j
z#9=S^WuE<b;1-p)kks=G@x1QGy$>9fnFT5XOryY9uS%O^R;-CD8!%BBl$QK}!L#QI
z#a>+%9Yf<QvcCLVnj0Iu3iC?H34LUSic$Rtk|?`6^alLhHq8pJv)0Xece!<&lvR}O
zOK9rpo)lAbp?j@M*>SbKEMu*U(n;<JcW#QSmtJdTu3n3>uTdd1Ik{K3AZblxKLke#
znsw|VR**!?d!ol+IBB^Y^1+JyFIkQAOXhTcD<pv3|J*dpqyRWHk5e@MHky*lL(?sW
zhrJuC;1lMhKMiT}#j^mU^m|1>OS9{S$b*J}&8mGuVtGyd^L~O7hqHQd)4FI-DGTei
z@80%*ji7FSxwEtz=iOEL(A94*uZ9A%y2RHg$%L@#vx%Mt5o`(^rNuE&q)m%Jx%|&d
z9Vc9XKZ+U0MuTxx4nks={}aihG&al-O81a;JR-xajwZ{|bq{C7-Ko~53&gjWQ7|f6
zLcscp@237qSHUhJ1I4ocCF)zVtS%HL(ThMUHV`*H(;SboRWL-?I<I{ln>su$G>AFK
zu{j3F_kH>)ZdB@IGedQmmAUUA(oaZbn3&4x1B_9LuhP|(^egMyFp62zEPro3D_T*x
z!AK(EJ0AM!9EK=HFhY_4mS$pfDx4Enn;#UBY{g>E89?Flm*xAnGtJceMH$b<76l4N
z_!R+uz@>F3in#)z1)Cju_+a@}a{5t37WmYiY+1iTx^;K$C31p%i(bWv#k6U)_MU!{
z^`>vj?qoB&S1?+EAMKrW-iO)4*`FIXm_^Rh?7wcxONt*uy+Lo}m{UubzfHb5Zv|p3
zvddCS(9Orsf*uQHw#F03c!(UkWqGUqXeyBnzn9wL3ScOHK0g_CTq0tMvyah0u8U^R
ziG3%uNMSHr)-yL9PG|gmJA`(&ywd4I=CieBA``OFbh*$twv}nYv&e+n<6?Gil;{vh
z$+~5xGQ_nKM{q)ACOlt5Q^x8`;I2xtzF36`Et=gEMj?b8*p9$@Fw&wSg9l{%5D5%e
zHEc5(S3Ak>Q(|=q>xdUXjjueqqcxgre&QK*NDmcTFLuh2&lX+iO)Z}VtJKGQ`$$-R
z;4fk<YSh?+rUIlR7HGC?+&#u#ulG}}H5O}dt3`jacpv{2hLtfcmj%c>9X?P^T+=$A
zk0Zp|*OcZwbh%VHuhsaXN_WE_CV1r|OvW1bg#oXgHaZ4zqQ|nZ6@zzP(C7J$u~bn~
zW@b0xxR)kTdn(!4q`Rt3OzgDEME?p2eu+8bRDJH<aZ=E3V!EGK=)VZW3n$Qi7Vfr;
zj>`TU#{G7M=dv`<!jA2I>rQd(-WbiMb4yUGiec>O5>_50cR|VM3Oj4!C%u5dd|0ZJ
zD#~H+HV_wye6`jB)w*3&xL00>!w$Io3lR!=&^&l(Tkx$?q^>|TQI7h!4uo?Ryw@z(
z%{wU2%>11eEl^f>`*ir(rkh7e;$caa1&NN%Scg?6nphJorJJjQr9%qSmEDR-zls{?
zW4aDO9Z_xIzWuf#k?1c+<y4Xk>dN`84f*?P@93r^4@F0;^FSp={0ZZ)Cp?wmO@_$K
zZ3wz%6<nKQ|M#e>p#Ay{{ATJ)&tkjxx&2=X*``zKvtN533pS^1HB1uZ_Wc<Rc<r}Y
zi@W9?J8L)4W%Z95FUv|&-_O38`|kr?F}r#ke4&I9x}}FJ)*PfVvmz2Y+S@kkCU@!^
zUIj5=zVmL}G((OI;CBA1h{eTMX5=YQry=$yZA((rq)~oz_MDg(G=_)xQ%fDX)sM>T
zM#b<3Pk=3H>TN%P<uQz&No-FX<OcELbxgAobDt+|{o%2`c$yNLdYz>hbC>dRg1Ncc
z=!_@$V3y_0Whoc^Y7zT2S4JQhvE#eDEQ{8K4!CV@nn$L}$(pNQJ=0HgA008*6q!*!
zhg<4a-Gdbv3t$f^oW!Hu2>jrHaDsE`w>Zb9y#1&ZO)o3fWUa0}0Wd%O{&#nXYmSw=
zPQ~;q{jlB}EKfPbhJG)CWr2>zhiz9rul$rZFl1}?MQMs#z0gr3Z%@u;&ZjiW2qa}W
z8=;*tE@e%;J?jvrpMs#YbIVrmmfVN&f+hDW$+fkVv7Y=GF2iQ?)8&dndO8pEa_W5$
zwo+ExW_P!9qL%V-8%-D4x2I$&JrPcZ>Ilm0#!+tZzm|$3%4osjPc?Kw)E2<NYvh22
zhplbzEp$e2eT}{ZXX0av8bQKY_WJ9x+~j0OEA&FqxX-o>15TQ_YB~K^q})H_`64F%
kkG(JdTWid~@C%B_coT-7W7*1U4-Y^=Mn$?x(lqe@0mld9rT_o{

literal 0
HcmV?d00001

diff --git a/studio/src-tauri/icons/32x32.png b/studio/src-tauri/icons/32x32.png
new file mode 100644
index 0000000000000000000000000000000000000000..bd08d7bcbb6badbdb5abc9a230b5f45d357e1a46
GIT binary patch
literal 1930
zcmV;52X**~P)<h;3K|Lk000e1NJLTq001BW001Be1^@s6b9#F8000M3Nkl<Zc-qBR
zeN5E%8UNhx-tRz$oPb0fFP8WMTHY)Zq1DOEMvXI@y5bukZl=l>wL)U(wuW>WG;7>4
z|5&MwS&WIZ$;O<xRRV|{IB>NzYgwnG)}D!!*E_j8?(O&6^L&r{2@0mOYx-O+_x*i;
z-|zE$p7-Z;woCO2zCzM{Q3e|NFe!L4J)|c_uA!@n0=It>zEA*?M)TzZU?CB~gB}-T
zorE>X=}4X*kI6s)dP(~~1JQk6!E9ld&x=Vpi22resL9!c4WSr(+W8M8FI|RLldNbR
z=!3;zMU=tBzJaVLUmO5y>?Famq`}723>3$uV3psDp11yhijybd82$`(wLisgst=)g
zK^i(-4z%3<5SoyJ7^4|MMSgq?JuVBhBz(wwAJ2Y$8IC69z}eD^znps;9o=2HOM;p7
zdOh?;BjTc>QJ9yHk{#c}{NiWu+TC8X-R?(%W!_^!%n5*nv!APTzJ=#MQBZy8$GG<X
zdk-w5(;*ZJQ8+w${Tr*Y@XO!+8W)xTzq-+acylyDbAXWMGW%yFyzZIjk=u3~MMXtI
zI1A;kR;xuY7*vJxbL1U=yWhWoXX#E}>&JK{uLOf$S2%b-4*;Do`vl4|3h;E71KZ1X
ztEOgR!K}H40#lMC@ypM=FU#_bvpXvGBD3=oRAv+kSI!1Pdbp@~gMOsh5>OdP#E#NZ
zv2QRK5TG<;AxwFBc|v;@$iD{<9>mU_JH<1f;j@M*5HHYv+2jH&wJj99BZX^L09w+<
z8Svp~dLiC;`4{le+GH}p@Ao4*IvS+}sK37-$B!Qu+Lx4+h<gVP9Ke|~XVBi>jy-$!
zz-%_dPayd2Bo(ROz5FVUq!++N{In0%2{j0L+aC%dDb|M9HOukswMEeBbqEH6SigQf
znwy&iL#Ba&0l3|6csw5D=jY?rty@S-OH-Nh{+2CU(9+UEJoLmPfJ>cM@RIZqJ{=u`
zQL9r6xAdUcNVrc3i<8pv_dowN>~9jDm6e6o)>aG-4&vIiYhph&H5HkenaIh>!MSti
zkeHaL3g<H;BO|zc`7*MzvxRV`?xsuFu<$8NkPK;BEJzRhFHm5WJ{I;XSJeYns}%(W
z1xQX#7TK_Q^Jd{=4m^(d?c2ADGwl0~8#juxOO`Cbx^?Ri6&0oOv3GRfY2vMsq}gMD
z^VDiIA;uHJ&4K@@>DAfUiLS0LR8>`pSg_meICbijc;|FF#h0nEu@PRcR{&eHW{o)4
zK|=W)6Z4EE-Yg@MEi(Xc-8ShBkneji?i>$;B4&}4;!7136&M>ELqbA=*h@)C5oTtx
zSS-RKH8nNp>FE)f!@`*d)7bbpLQXgJ3M0vqXXyn>I1bon!Kq7}<AEt3o}{|ZR9RVx
zty{N>H4BfJdGFr6XlrXjLqmfQ%+`wV;8YEei~K&H$Q~u)^63CHqI=2E5pq3g0tyzs
zxM<NLoIigaadB~oiHSjcd^}E^IDy>UT-4Ro3IB73+`D%Vr%#_20LDq%(a}+C+qMl3
zheNF;2FfDYCC{h?(*RHrBGj5FGieo#F|S)_!VEa~%gf8vOPrRqwY6}$Tmq!cW<z~_
zy?Dm}`O9^I??&7ZML;#tW_X5uFlr<<4RzC-kB8c>+azIKb}rs|uLmQ8;f}|?A08gY
zv17-?^Xk>B(bUu=+{3ZKg=adkg$^G+tj5v&1vcboWFufP&;;d!*<hxiR~`yN!epWX
zapq1R>dwB21C{&Ly3T&9uC7L5VIeMDxF9s=JtMW@i13Mt30%B*QCNqC^ZQ5uSC*Bd
z;qBk!)SY)>q^U`ns~z*?!^21B;#Aa*#pAo*+@uB(M|h+c7)UVpg9I`))6mcmdU|`|
zo19czHJ{Z`T~TBOfBIW98WP9R`|-aKYqHFklx9ps{36z9L7!s)a&{J0KeY^fH~)nd
zD_7#f4?YkY@tL(bd1$BlFrA)TTL61+k65#7B6n7#EJbMfVqEXLjyRL`k>TW#{>a$m
zkc@ZygV?`6?1rmt?Ko2P6Tx`*4_?3@?N@P(ToR!pWb!6^C(3s37H3}k;g8UDHC#>i
z@7ss=&~SKum~HT}>?;2GOaKZjIz4WTI8eUfJ6Pt{LYt9{6~%=}@k-cIa}Y&+XYtba
zcj4~kZlo<*jD3x-p|JZc{H&xDOM@nCI(`Ta=Lk}LdaSD2g;zdmheguQ`rl#Jk&ONB
zqt0j}lO4@?4rgy(#~qq#UZla|{H=G<H*yy%<C0LnFdrfEbG5^co1=qR9-DyD^gJ~7
zcOgM<#m@9xy#CKCVi+;$^s`RMbIybef-B;f&y8rBZgu1>UeBWxvQWu(2mESGgvh*B
z+Vhb4U49R~X0#$id3B!xn%4%Z1#^w)kDZD&Q}uyUlrw?XoWK0xj0rat9?djD08cwS
zaPS$ftTeERCZ{~<{{p@^J!+?~g+^1;O+-5e7L2TwaBcb0{u}<k=qm^CUotRQeuC}4
Qc>n+a07*qoM6N<$f@%`GBLDyZ

literal 0
HcmV?d00001

diff --git a/studio/src-tauri/icons/icon.icns b/studio/src-tauri/icons/icon.icns
new file mode 100644
index 0000000000000000000000000000000000000000..e67a5e64744a7208847d39bca3943d1e99298549
GIT binary patch
literal 263568
zcmeFYWmFtZyRh4XI|P?7KyY{WKycTPAR#~q?gR@A&fp;s+}+*XJrLX>XmATI^Ud>~
zwbwd(@8|dVb3RtDrn{?J>aMG*ZW}8zdlvu+Io8USmj?jQD9Z^!L`2DiNTh@qIK&Br
z=)|OiAbcqf95Ui>gxJKyXaK%POM3fj0E3t~nh=*19aTBDBkUy*I1{ddhE5tu_<{sM
zx-Q><0sxRPDBtBbz6Ozo6C#lS?ma>f0E7ez0FZ>cT|p#aglMFoz!CsJLPCT903rkl
z832j{0MZab5>iCdTKN4)002S&0RX~3MV01=q(OuzM7Rw77yt_V4u1IKY~Div6chlh
zj}ezBkdTZNNHzh`0PJ`3l?ckfG_1=Q9e!$*3?dC6#3jOIZUm43=^Geqb!rG$fxytN
zWZ@SYS#gN`3DHRq?BNgPhHVT&U^oxOe_^n}O&DAvzIF&C{)AYh2x)HsKuiQcR=IzK
zz%U<V5CBmL0FX&Xgcnl~VV6<^hyY**dyxn`JA?fYfeFAH1%NC5c{l|DQ86C8L4b`Z
zPyu`K3YI0|4xqrRW0#T=_!A-%kmO?lAb;S6ISeMF4uhHe^C%F2mQRi!KuAth`~m=8
z4leD(VtNOD)u+Su7YAPg0M38zMJ2*2p@dh@fWVL^RV2hUmmo0IT=)g@vVY!;M2e7U
zKmc!8I0T09bbIX&17aWm624a9Kc9_Bj3Kk)4Im=`Vm0UfFwDz3DF6W(a9@$dAPyiT
zBSz(`f`10Q8&E{Gw8T;2eFZ<Q5<w;oB19)euHXb%0Axfde*j$y85w@eQjS6zOh`h4
z_&FXpb_PD2x&eUmxdmW-2E@Rn7(z%$gk243?u2v_HA~C16Sn%dwfNQJhY=DH;k-%^
zr3~dk<Y=s8LFNdf6-l7QAr2!XAbJ5{zW{JZvGMV+NpS!?YygiqobUw^DKQZ#5g{=l
ze1jhllM-zcjy57SHB{F%B{ZTpH8oVU{2HijZu-`U-PBlJ+46fWXJEXvy0Iy`5x1$j
zc6xWij}Y*=*qo`WZi;Mt(Ny{KPo63vyo<0%RrB_LRy2jfpH}g^hvUDP<>=X{Yzk{c
zYpU4E`tQtUY*#jgG?Fw`E|vaIW{Z|9n}QpW>w0IA{~NM!X(3Mc)&(_^HC66w{cmD5
z_A8qL8*%FT2jI;7zko%B1B(pw4b%m|+gQ2o@o&7ktyebrH)1taZwvn$uY%jvO#zMM
zO-m&IF!n!jC0c4~@NcAOXdg%WH(ZfNJL>%#k?T8#kpBf&P+u3k6n^th^nc?Ob*!U4
zun~EQ=-+6?U;O9ANKMt7@+AL@RwC)G8hD#9o9g<G|2M5}$NhEiCXqE&|7`z%Xhju+
z(+Y0@5`X|S|Eg*Vf=jDrn;u|=PbngBT1DXUX9Z|BYvG*(@3*yZ;J^+rgt-HN6YL}K
z=>!OZ%P_2wVDr<=j@4rOF*o0F`;yh~1&iO!4WW%hP4(^jeXV=FRlO@~JyqR%9liVQ
zb^p|Fs;{W8uc&LTYi@$?{yDC1ih%Rh)Y#P6*woPQ&%Uv#X{WIq3JMQTh9X6T1q6pD
zK+(e^!oG$Vr3Ocae}iI&hx>nx=$p>!E%*`;9v%(FjR*)V|2gkZ4EP=lmj?z!L_%Lg
z_!cbYX%PRzE172Aa>3V#aQM@{_V#f77q8qsJ-!iPP_&4z!@2*R*X-f15g|~L2*0NC
z|H*4fi*G~_6eT?P$0X{10~Q4tK%Php4iAKqMfgtZ{cmD*rhFp;p*X?GJ#eZ2FJik>
zf&-xF5x$*X|3<7wr*DKm6f44SQvBbD6`6!Y1VG6`$I1VVSd#Iu04PORR4>NA0gKub
z6Y38|4vXnP{TE;nTff0e;WzhT{u{99y)mHyP?XSd(*Hp$%0Kff;TZhIND+_;72^K|
zETP<#f4DysGdw8k&%XfsCo3o%-lA~-+~)rYEC^^QfP@FZh2=lP^dDfw;J}Lh2Uxo4
zfQaCKI<7qmICKDv{tH;k6Ce~W!cZtd=#1UilEbg)H36aZs9%or6SlMAVNjy*kZ-@z
zW5!bal7CGlLK4T~(tpJT!wU$)Lw$nbTSRa;G&mx}Cp6e6G&}+hTsRcI!7IRb5#iwx
zJ5XC8I~zDfXt4kL_kShue<kq$TM1a1aq|Lzh6hwbO&%MA96sj)*oq3WTK{OP|It`c
z;QtJri$1|C2`b7;>v$|3r=zCn%w7$ij-5<-pT0s<An0BTc!6nsBpZbnrBuf49@TCW
zNe&wz_C9UDA;#20A_P$=>F0fE7=!tqwp}(|den;69-Xu$w+~J<9i1GFr9PYp-^8!q
zrEj`d8m4Qgr3n=$;(t-z=!5bC@DnGLE+_%xPe90NvyTj5gkM2~7ZF$j@S;sD2m$;G
z$Z_-k%gW|~c?;~)Pd`Xw-m1{Ul$Cn!JH=RcpXjn-r3y`+HSkXzc9`e<p>r@5j#1yz
zrjtO=qSNe4x7w3*^N^=lG^sRkdg~eJQEhFgQx^XBbTpmD5v5D<8BJ-_Dw?N~>Ag*S
zhfyGdwPOU1buL5s4CK?X?)Odn*|w%^Up@`&k8gHoOtP{20_s_3>vlSe&y{yeibs95
zNB^!jvRVCU1HmODyV}ng-apRVQRf<^c~AV#DYWh=FZ#4OVJUxgphP)$m#h<f@Vt6~
ze9%HAp@ln^VvZHeiva&{XxirB7}TVM$oHE>JB8K!SIhDvZ@Bqzyw=fpuw0f-bxGTF
ziOzZw6U@{>`n_X(sbo3iDA&un@^>4vLuoX2DtYgJdg%r9SU({Vet)^sdCcNxOt%kv
z&(^XJ#>^A+gw!C?o5<jzuLx!v;_-=ee3cXtcYY-EAscBDqQz8tY+5(tX&=@G$^L1`
zw$8bkwWq<<_-UcKNBnb@eV}e;f!(-^<19Kd-SD6lSm)M5ETasgPwOk-tDj8}^0*d6
zc<-IzvbXzCLT^zzi+Nd4tVbX9B%AXOF*Dhvora7%b2{;*<?p{}`kmI7FEwgQZe5Q;
z%iH$)>1!(!gwuN_(na;hlX>OK-&XfLwtaoLww_iM<cAT{;@y-0ZM)^EPtvMydDNc@
zWTNoD3%Lsa^Pxpc69<E{OW{1}*(SLSR;#+{Q{JL(j*0%yr=x9C>0hTFFkZ$KGCjsK
zYtMvI`4%y#r4{);p3Cz)tclh8+&N#i3D_+ajSH?Drn^ol9^7=o+C9_XvRP^5#c{RS
zsk?RJATRy#YrXgYrldYLZeCg~7aKp~WHHL;y=9+)QaI$R%@z7q7ye<Q?@Z)1QhUhB
zc?IM5z29fN8lZ3}5?uGuj-n7$8&4DF&AstG>qedZ^DFnOOpo_@M<eVpSj1^cnS*3Q
z3}CGH>aD9uU>Lb7?tFbbU7hLPd83!!&Xd8EONtObvyIv|?K>Pte^5VftgSA774LOe
z-Z&G0jHz{rW-Ro95uis8J7;2N*CjFb$R-zd&a+-_u$v+{%c#}d`Y}N3c+v?U(mJaA
z)_r10_BB}YY@HgJ_vLOi#{=Eo#RRUDRLteC&kI#k3&5V+FD`GdmmTJ{Jv=*nu)~9J
zhRO_TRfechx++LE-ROFwg1(G*g(UD|{4?AJ+GA3`>sW=^C%-E<d!)nCAKb(yPu7Nw
zJkzSMuj!mz5hzbFJzHjeEqH!dhm{m&`o3c~Zu2lwREK5k6=>g1`UGABBcr3URi?c<
znG#<g=ka%pO9^6G>z(k$Kqk!<a^wOl8jQ5x=MPl;;Z1n0e&1B6^5HV^@LqhX(MptG
zhxS_I`ednpEn5}0z8YnH;Y4ECTR5A30QPt}F$;TIhZQqs`e>CK)DGx>@C1ja&0bjr
zcCTlhB^rAbTQ5{)-(}3lt02x_{}srcZF*BXQ)#0|8at`4?59a^Wi~KEUn1Lyq6@`0
zOZ!++JVX!DZ6DmZq;m6AyH6l@h<RA)FeR96Ritn;j~Sm$vX>Eq{1kbS4;`FKxZLi8
zx2L7$prmx>>&<H6T!qmgYX=>uv|hRmpQh4j)os3bT1}LFN8B#{LlPT@QMvu?x!Kk3
zBh~CgDY4B~)fAIougAA@tlKek+{}$WGKiOEOK-l(3ICSkEBlgE#|DRKqo~jG%Jb)~
zB^AU1CWZ$|`j;gA@>W8@5fR_jURFq;x7=Os&4+C+qd?|`F2{IhJD;yOFH^FmXw1@B
z9;FyeLD90iWyS>(bKsuZ%Yu}UoJ>8ADU^><=maUm3AfpS@R5swU)H;8T%zfs`u9_z
zeRPe(d`s)y$V9*hVWX0#X_#1i?8RzvVTLu&14F?y3aB@>2=oSjNAkRnU{+Odb-T0E
z1F2g{y2(_b<v!pI9*vW*+0MtSkZUaar+r?2i$x#ZGD+>)9F&Z2Doxwjyg)E$!YG%(
zGwoK@ZEpqV-S3Ix*6+|{yKV|R=V%W!21%7}u*U|-?RwENwXCgady($CB>cg%JpId>
zf%wge%RtpxEebYyUK6Y&F5I8)<<r?4MNjS@_)C-aj{}vGMXi>;#m_g(>bno0yvosD
zd^Eo5wbQZ$E=I=pj?8EgC-UWdfu_fbz}a1q;>3fc!AF#gEQG}_$`{r*CO+rE!rx3h
zeh0^6eYI0)bwAdLAZCwzC1^7@CI&(Ia=fJLb~vZ%wm;SX*lIbEQR01bQj(WPS1yfL
z`KIk9^eKY<^uy+q=fQ6Slj6cXHBDYUcJ4J9%yIQesXeThT|;D$wy@L0gOHbMP0u0?
zk`;SD4r0hNmP_)!?ckTj6KKSnw-ac?B3qu|b4sqn(_iiHtBk);bp(^LM)w(}<|ngd
zP-mN6EbE6EQj5wPmPVvPAFd7y1(rVN#_yqc<HZL%G1a^cBd##YUweP{{oe7n%GJVA
z<?IUos7`7egE7;N7=k76YWHAkG44Es+H4~F>&)lX;KtIER?g~;=5r~0PL=)c4fP~0
z>GtRQL%Ds;=Qk0=`Q@YY!Oo~#_iw|_F$mfML5L_tc8#lZfAO%uVhbL)31orcc$FIz
z=~PVztv9pEoj#V4Hp<5@V#VChWh{Z5U(mwCYuk0r`#j0!c>!yM2sP&F_OCU;zb~M_
z*FE<rucnkai*Y4xrN)!EX$f;E=1QrcZf6za0>q|7VQcN)rIcS*G>@D2GOTaUewSz$
zx*gI9A##VriK%#;Sq4S-K3J;hsQ*ka+dXvtXmhNa&}<Hej72`_an{(m((m&6-~iNG
zA};#(e9OC`D!m`q-tnE?y6=PC)fn$Yk-o{-SyIsZ?4vy6SL*5_cP~wi5!SlF-bdLu
zLL0yqW$b$eOX_|+g@wOQ;lT$<(23G7Zar1`0tiuVHIrm;X&xNS3`T|0)#^{r9Yn4z
zm-e>ea+BHgRuOFOE~k<iKAXje7i4&BqF-Ijn^@+tArr6|w4*j?Qt*9_CFi$}7rEUC
z)P+4iX}1M@@W{eUV<_txTH`lsYk3ZJvMym@?WdsD{L=$}k99rreEeTmcR!WJ@fm(Z
zcYQkwzH3FnfUh?Vo1Hgn-Uk(}Cp^jC-8=Ji9~e#A5k+I9>rL0c-sqYQM)T9W7LlQr
zr9-RQlK<qFUyYQRS^^3Igz7I*(m+MOc3-9#W%ZmaR3*1qccEQbVo&SpF=iYeL~Q0R
z^j*PGtle*K9Q6+NnbM!<#WU-=X+3U875N=RFeCABRI7GgUY{`Hzm2+U6JtAL8B+|i
z`k+biHO4S*k_<vVd{PSyyA<WOE|7BrXQ|C|_*p)ka;bEzLT`?0CrT*no21(xPFk;?
z>(=oq0qQB{!n7Rt%G5dO_5+1w#jnlZn=k3l8X1Agx^(HL+<=s*6N;2>p@C(+#Fggh
z<^8i7Wofb-SyFbHS{v7w%sw!-FxN5CvP1f&r6t-Ce=n^Ne8z~9E@EMWq@Kl2WNMr~
z3K&tG?3R?}tRN%(1k6$Y7+5jYr<O1_JmxOt=#uz-?=RC{cULS_OE0Nt+Qg7aPB`In
z(TehZah5U=NtVmew1F_&=&&AJ^&wrbO2LdL`tgD+kP-;$7k4=xq&ZwRAG0TEwa_Tu
z5l`P{olE`m+Y~jKLwpm~)E~v^<{_|FqFXvMt8Y>sEph)DamA^^cRai-ds~%3HN`xV
z?GSS_b`6QSdO)ni#Tk)LpRy!--p)1d`sz+HDO0o9iS+dzD354FS;O9Kdf{oW;rTmx
z(cwd1EIl#HxP+6!(XYtMokSCn{t4;m4-xDQ8uP~9c1mwI7(bN1#Ln(oK%Yo;H`k_j
z^z?W0<Y+`XT=D%l$o}3<x6<LpGWXHq$AvOXoWgd0D*xcr57xQ_U(%v3E_sJPitbbz
zIY_B_&!rU<8M!a^I}(){Hcy4hQ+QPGRVP@V9^ZJ;Eu}6~RqWBY^W6w7q#TI>`n%KP
zjHOpnQFb%+h+p)(Wi!Vih~_sAeh;bLf9B~K)}bYJR3HCLW*^MGchFt7tyLX~r`<`g
z=H4a^%w-TO1ZVOtdLR%*=T`Z81ByF)+a?xXf8X34-`AL@zKHXom5byWQTJ;zYIWDT
zd9?=IAgCXG>3m54{TeF-gWlLUOLor<OzNJMY!$|cSL?-Rh+hSM?PUhey9wy%o$Y^e
zZ{XT9Ml~@=UNr1{yok7;&U`li+_+LqCH{?v?IxlbRmU6+J4L|G+Ls}Bj75yx<u$^|
zd1<F_CAqZ-kEy9USV~;WS0_E=Qu+SL*eEYyJ=Uc!ul@_=Iym`>(tSa?{dO~gxw_-~
zXD`GRS}*oLB7Lk?%u}0s>Y#<_&<3HQAg_3wt?DFR;7akXcwe$;Razh*K4k2)tcHLB
zs#Z_sdz&rSe4SSGP8Hv|GRaLr;p}!Rrcei^?^5QR#@ZAY4;(+MO^a9gMQ7z6%sI&j
zy9$&mk!iuXXu1^(w~sg{#yFC}6`z^l*isT1#{F&VG(l?bHAE9y<oA4;Iot-Bhmu~(
zm+2O@i(c#EdS^&|<wH2X&wF(I%Rb9DlX$cKhBEuT$7nfCihBW!3U(VRvK7gz74lJN
zxHDGjWKVc6&5T8``$5sUsw{)T-%Ark_HKWBSxd&tfwZnreT*5A-FBxCG0Hh^ylacD
zvs<P|8dwKZMgqX`lTu0@oKF|!B1aL0aW2IBZ;DWFB#fy<?uG6&pBoz`dzd=wpePBj
zqTFbnk<%|J!aJN{ZIG^pVbn=oG@s>)ZNm2%kYRKWSM#fES<*_}dSAll7JqNjFI0wp
znnQ2PX`_W+uewZ#OYE*N`<XJJW}5`M(v6`3giD3ml2?_W*5nIEv0Z0?Wxe`|?(zE9
z4@z8!tz`N~mhJ6M*z*JB4-+C^XFOV|QUSRSR|m7qa#?*E;krd;u7@|lEqp5SrrEtA
zTfr-{@@5l1kG&3%CYaVKVhN<<GpqHTFW>{kL~bK15a)swMdBA)uhNwt&`Y`<z^I|s
zB}Z*s<Rm%mPFDPNeG>7}6yH7F#B=fay?t~43-=jq+OvyM&eC37$(2eRI<XGaMHE4-
zHSH^#xt~;jH?gEL`W_MX+b#b*mqdIDVy;IM&7V_G&y?_OK3BUtWYpXECJ)wxE5Xi6
z9n>e3q>*&q_J=Z8L)QhL`rgz0K*g*e0p1LLT<117`953!-mCvY^8Qcm!Gc+gQH*O3
zv?MVokIL_{+IlJv>W`~zKU9Z**gjn$O)y(%MR5fID@-j!a(w|q!kwp1=lMrxzljUz
zhwlF}bl1;hf;Bm9&A{KDym>9GdDS8tCUw+PhC*>=g&6$v*J{q+Xu@||T*eBFu+5s;
ze~i=v$-o;W)1w6Z$+2W!Y*N{p&gdV@&~&`#QIqG7C{;@9p+~F;N+|<s3m&X?NO}ch
z?5-%(udpR%Zk*;%W}Q0!`YBJTm)&Q=9xeH8=E~wJj{2Z=1k*R&;Jf3-^+&`WH4I}(
z$2{_a(cvs5OnPQgPQs#7e52Zar7w0fJ#}A}w7p{CI{CQDp7foOnff@^g$DARXxAD<
zc_YazS`q!lmhla>iNmC-`5WXuuCyD)u8)Y<2F0tj<=CaTf(1fg_2It(umT!c(y!^Z
zqz%S%Rg=z4o9JD)Gew%YS8)NP{@|9nNYt|e2678RpN?9rJvZK1vhi|<fv#Q38F|qt
z{&k9vxI~w!;(svVTFCLZm_+f7=pxttVr}95eqkm?9J&G4%ghKGPDhL(<GA+V(B@Ut
zFC|u9Ow50`?vAjx-OysijwARI*THgkqX<rtf+tN|(YzxjXo`}*TSkK)*wbCc4$Mlp
zqxca~u+9{<j^L}`#P@)jFVT|E9`)wK_^ed1D`sPqV)~aD6;}-M&3M$I8tb`USh|BZ
zD?IW2H*gO1Z&tsUNITx{;8)RQXz%?@pPg2hgj@3tl_)qbKjVMO&(ANYd-qvRs);%q
zvJUq}eP@!6(FsE0HQsBR9i5er_*7Ex=8giR?~>>^^vdF36s|yAN+=vOadc6V0^V8X
z9U1qwa(7r~L|uWLou#EVu{S5bazls9tA0jae&QNrt!&E?SaqEybf8}U9_0K1AhPVK
zRi%5Cc?>r?bp+F0jmzZ+_G}Iv#Y}9TN3V`Lc>m_DMWxnV^VZee2_+2OU+OM-Unblb
zdk(7*;x4e{4rDV_DkJs{IIOnRo5~T#NX;dzN4*rD@&J(*3`9|o1kqHTgQ{c{nLhZQ
z1S7Ct+4_I?wo9FFM>zMf{cBJ_;#0(ME#&f@(kJjk=HGb}zxKu0yMudpq~%q&E<_$t
zo`%Qo`BCH&kVSW2vSOTV{c;+giH^xRzgVKndR6K7r2O^vcj!Tk-(#*{_--vtYzdIx
z4R0imMa*!9S<=|hb!w7$a!fmd=RU|WBeSyk(IQNKti@ouK=u9Ur=8SImsO^T-BgDP
zV~^j+ZgAIM%B}Z!3jeNWtkIkW8f4+}rbmqE93^Q|CLk8;p6dB5H2#~_-Dfm&+^0QB
z7<cYa-$uU}3-`7TWk>}#x-?vyY(GLR-W$S6Et34~;~t!hPNl9Hg2~y&)o(N^-+#K+
z-ny^1&G!a+D~#(58p?SSae&lDJ2F~t*Rc30ft>OjC65j*U1R`L1|mNRH|yOVA7^H_
z&f7{b04htiX1lX+Y?ZggM*NRy@)99%8H(N+Y-qRlnY`j640yIz%k&{#qrspJN-%V2
zv|pW}Pc+PdYm&U*d$VLz7wqhiiE*iy*?b#NG}kKO+H1jfB6OZ+o_QpCuS>i;NxEaI
zSetP@3(E6mm2^m)dU3c6ZJD>|tPL_Mu-Z==6drU5=-4GZ5lf%u3opCdPWOsFX;(kP
zeF0WVk>s-?9mhPE3322rNK6u7EDU{vuj$*L1fa44tluMl4_5wM!t>R*9~JPoqCk-I
zTz8#T-Q`_#?8UZTwrjjf7?oe%0re%_#j@&^ot>I~+ZR<N9T9GuqPMtvLrXl7SVPsm
zSnjqMqbOm@O@8&PE^$ZaEcPs&pE#XiD|}_4Mu`8zGTrTmFr^KT@sEN+)83F`xZh;X
zk=k{Cir$Qt`Lq*m*(J@#z4LIx@=v?>5RTL<<P1F9fD=^8&2`~@ccjU?8E1BXP+B<y
zw~Ah*nB~x|3K<`HBmV?u%u6fnugRwEc=;6&xdJRYEk=JNd%4aeWMpPXHr+9aU!qv4
z+po2G>V;=Q@x%3loPX_c><Eu|E{2=K1nCN2F}8U~I9`5dY<jw09)Ij#wr?E|!lb9C
zXEv6O#}B=@OO{fh|CCM=%Z6#6-43B`aj+kfz09`|%HS;Yd;j&GYTc!ja5CoYM(at-
zXuNd}aD&YvKGHs`_rCb@r|3kQ2i4cV-<&Q(vB=`^QOs+8Ss+doX(iMsexZocX!p9}
z8OX9*ZYWVr6)@1p6RNif7gltps*&3yn)JW<DSApe0eM7~lb_>aG4}TWNx~j$C7G8P
z39hcU6<gTQq{?SX)r1i>2nbD<a==p-u}iYrya}jkuwYD*lm(vo8DsYgOqzjy*1L7s
zvt>M3XY=fL?~`qlBZ1ytQ0{I9^a}t+n2e1aYDf@HC5TCrwewVr%Zhh{%ivfA8L4uk
z8<mRBG|q$5y@Gn)7(g`N9ZTM#5}9Hj0$Wsaj1wrG51G)Q6Mt%=s<6Ccj#zom#>wwo
zaI<^x7xzc!^@@u+BITQNb;&1J?NJgY2i5dX`)cM7?~;NCWz^U`E8;Ipcz)nFUN+96
zQ<P|<)VCvuCH%^jPvwGQATY^H#$G%a2BTc8x$OPCasfg6TK1=gGWjSey#=W55jTHj
z?kASPb!xExnB)~)pEpc@mIlV)3XrAJH}B(meI+*5`cv%Hx{Jmj!luh>M*?9o;p5~m
z@2x1og0TUD#ZRerH3*9SH+$^|2p;C%$TbycHkczCEND*;HxpJQIY<1jCrtdF_~}Wy
zz^fH*g7xno))*X!+&HN-aO?f>S%>`?wFEe?b{smN1Q@?Q2cT`WPe{UiT$H%&7)a;m
z&8?Ir7@uyLJkMtKzIGo<{+*OzCj8?**J;w}yL(z%f;6=Lt+z;REvrC`_3oul0}iD~
z5(dI7xQF#MjrWkHkvGUuWy!|QO#A2jY{`!ly*9DUW{8I?XXJ0Ng|*bByY<a=`@E`M
z^?m(m$HPe){k?}DzOG7A^HI&{qfmW|)7J*$F8+B#vq<2SIo(r-1nhhJ+s&yse`9k>
z+j_+`l#)m9Pl0M`%6v3}mR2OI9P}*w8SR4QW7y%86qKY(r&z~Oiid=&%)+wXq)?S#
zca^->AstjCH$A~3zPk4zUBfiPBsus_e$8VGyB2gX_0*VZ0?aucXpfXAf6*bekwWKL
zhr}Yv487AZLi;I>wOr@Hk{kkpPu{$8Qg6`2hxe#wx>G&jo?aDXX$fT`GLV6pBWT;_
z=qjFdq1q02l+h7|nw9yIq4ucoo1c5~8vy~bkD@r0rNWm?qAtiuJj(8d9;k;);y)%-
zroJ80a5P+EC%9dkww)QXcvny&gqcPti5m;y_dPM<BB`06X&<C8`pdd1CFV8ihc+G;
zD+j3lHIC+R(GVM{eF|wfD7zPQaIpQn;$mvxTJxOd<2HNFI5Od#?sZVS+ec(ZjV;o=
zS@9iDN-b|hpwp1InwRWaCH@@?plMrHQ_9)ZRS9>7mUU_I&oZNRY>jeu{=ErXEcKed
zww)TGL(-2=kZaHt0?*!3{UEA=jo94P3)0l-xc1G??sE<c#>3#F^NZ3eCJYWI0gE@9
zk=15{DS0W|>{G#*+gLeb5~>n{H-V;=+h%>3MRv5*rc`!Iba}^a_&Jl}qSxomEa{-3
z_!H$;m2`3eZB+M@Qwk)5qk5b963YzFf}aBUeUI`~UtPm5_LN0!MB=I02RD#xceA+`
zJ+`8BB_<1?^0IG2kW7tMX%_?G2)FG|{WS;_+p4t;l_<$w8`Y%8i;|ysiYc`aaT)2k
znsPM}xWM&~_vg_tA_btl&{}x>?Pr_*^G31f2y<bQfXT2~o(<ASD5&JxN*x*tj%Z<e
zq4aiy4b)}w4E9ekt~9tAV4nd~dZij5XqKV$<-WZ`6{}8kPw!!4FZ9D()qBq<84g;c
z_wZ`vG$aCU+T*7`v{W;0o!uez1wE)!IK;m%>z#}kLEMM*lws0R+aThx&63jJF!sLr
z4wmjAgQzU{iV9j6xX+R4D58ZwsJACvnAl}1<WReoY3?DTlejDG)+&6Ksf1)P-$gRT
zo?&-s0!IrdggeS%GIw!DVkXl&!qy^K0KsdAaB(dxQ_e3|gyUS);PxfIWE4$TZc1G`
zE%V9HCKv&ON)Z7fBv!YeSY9rIBqfaG4JBo73*4GYjRlx^Eq=nSnzkt|(|Adeo5*P8
z$hUU-$bC}JaD$fF6!1<IA2a@>UfY<fhs_kd6mPP`<hRb@d9SSmeJ^kc4}~ZTw_9JD
zufRQyD(|gz6>F+IFMVCY+-mIt=@m<v(n>Z3PFoHucBTE^jQ(Ih<XhLIIAj=zGC<Mr
z{H3Fl`XDEU+Xi428|EIFUjK!Az`bT6|0oD4KRXM&=9cuzTMEo!7EW+S*{TRMYQK@>
z)9;ZL`PS3*%$Hl;dii{exOd5`RMb}wMNQ*AD4NeLLtloLEL!a^z9gK#T)G3qnwM~p
zw?(8F@JsGZAzvWXg^SBOFO&<2<@f)Zl-j_|RZ$y0cj6pvHtS1qc+N1~$`ks>t^as$
zPnIHupO2e5J!t3|@@a}?1hK3K?iy0jio!;#1jifJ)xVyisbnuS;V`lj=C+`n|J2dT
zAd;|xXy6JM^DS%yYn%j&`b!|4VbIaR{nXYYaovt8gra{ahoF+5C#k1pgETAIM}!On
zdm!gqgQ>z8yBcQ_TV2GIf7k4bKUKWTA-7(tl9g?PY%=x3AbtWizqo~M#+7q)-kwX&
znAFeSF*5QnZVWSKrfr6!BpAmdB3o|td_?S@zghQlzyb)TN-*M79gkX%Y9f~4Q@=`#
z@7QaRT^7Eg<D2?q$P2>YF=2aNMET@i{bZnh006w@x<5ypEH_pv&K2;YTqLWB?9;iy
zo^DDFYCp}6fNPvssC>@+9e_<RQ`od$*xK@X<=ZTB`5^+PXf5O{70!`pJ3%FA`gT0t
zO=p8ZaT=DC`NQA+txWt)<e`FeL5B=0gjW?o`=RGs`w-^+cquU?34#;m6<8LQX>av<
zZEWAOO4D#Wl_FmbSGV)K7B{oEjJV{PML#F~(e;556c;?$z%&Lqq8OG~SZ=h)h*XHp
z@}v`K-ej+*w`Z%s>a3XygM0}-p3Tk3ER>0<Gg>OjhR=R5cUF#ECcO-xOR(A*Nw9{`
z;(Fg1)hmb&*_n6@)zn-Rc#n`s8O1#uS~`Y)xP%A&@<;w|kdR(;vGnn?qmA^gG^N~)
z5e*sWjxE-*%NHxq#qrrx*9Fsj)pjPc9rt!|qdRas?KKw5n!h8Xk}t{$+Vu;XkW~%c
z=?662m^DD`M;=HaZrP{hbHkF`%2SDE&})bSDPZ`tL(W_uWuNBZ{Xjcs)BFezlc_GV
zYm6>6IpMQ9$0?A7o(lg4fkBsa64*TR<~e7{H5*|xAGW4;S!FZsjC~1rP}@BhOhc1N
zg%^fuv+*`Is7aH}Wj)>;@<OP&sZl9iWB2KcqrxFaswa%7$TY@PaW$8rHzgA!6MCu<
z6t}<a+A9(IR1vCHG<-w(VX==gcW?HsS^Zr${83Y6!AX9RAUV{Js`2X#_puyp*sEO^
zYnvrGN0Y|VF{<{`SwnXfk4%UV+W5=~b6l*&$$HFEx-)Ucvwjj2+adL)m=*0FR4gew
z`Z#*rE*oW1>P*dDH2dk6(qvlF7zJv)7#u(-`<K?Z_UD8?>+ftO<F_Vp4p{qu*lS*N
zK@EYbaZUaA03A)Ul=M*0)(FO;l>|A+Ks2tAi87yCf+DlJA<`p>I}1bR?Vxx+D`Y$|
z&(!WlAa+It-KZ_oq4E+4nV-@Hu_6E6gq6Q@J68_aRH`;b16fl%4ccJ##v0+5qVHfb
zG_qh)*wS)Y&oI?`G6&Dn|9YdE9uz^ssb@A2$vsP<`JD1LSH=HLH<$lQq?ndJIDG(G
z(MZP^8GhZj?^W?@-sK{E56Q~#2Dm47LZ1p9cug~nZoF3-_NNQy73EWA3T#H&8~s_Y
zb4$>hKO#ZIWkS<&S3OLXy0{2Y8gvRC-SN4a<oLVm2zSY5`Yi{&bEjw}2@*R)2M}z_
zS`EjA&e2H!ZZZ)9qkliTeqK?j=diN%1QZ+7u<Jf_2Z*F2n|z1930m`BvL04)VW|<`
zrc>J691XE>CElv{-`|z7kcu<*(4I0UyuLHP96f~xplCCFf=V!_zfdn#p&M9ZI2)3n
zKqdoA5LfdqcvpO{-Cgh@EabM`=PUX^JO}{AecrW@Gn;$u3wJa;o2U8+As?6VmDg4M
zjf;KtLXPZ3V#euj5+M5rj<XFd;_tpnJ6B`XQFs9Vh@z|t#E=uo2lk(~RDDYU-p1~F
zm-h8~vA{xjU5Ir19iwVWy!Fd=H`1c-tNBE~KA?xogAi0g>eQzjd?mVwW8E}g<&FI%
z1iVHVLJBk8KRQ|ZBO{`s*pD$Q4jQ#!a}<IVgj>bPuBNGQQS)&wOOALrWCXawAs!oG
zq|k0eF5H3dPP;ztbF7(>kRx3&NL71lImeg5S9mb_9|V&H3qM+%I^OzLJVo;U<RY+L
z7>m2Yk+>CxP)Tt&-I~4(tika?j=*!1`Fzt)a1i8acOm{86l>Gk@=-D<5!k@-@y>d_
zfF=So-&nyJ?Ktep<X*y{P{gw*pa6z~jFbg9=A1k$KP9H{%7||N{oeBB%#?AUiyRLY
z5{&4cNQUuZM?)e0J9DS$fD+crM?|r{)EB2izG%EJGLzqR)P33uB%!`R?fd+8mSzz(
z)=>yBh-7FH{%BK)ybv7$z!s8cinMLdiOv)Yj_VNqW{;|dxFIFTNeC9AL#vA_g16Ze
ze18!V&5d%7>vj2;9weiYM|D2#a6U6t6?ua$nv=m;Rb!#F-rf{#$^@%tddBPDAnxdn
zTLtA$d0A{5k9U`<*uo|sJK&b<FRE7E<cnpW=L-YBo7Q#Rk`?m-`j<<1h<Hc&jh&TH
zRrOk(q+dU-fERCPHSoQow)r7)W=xkfRCM!^sHq=;w!^hL15qOkTOxGX4`LWtQk3KZ
zn&sgX`H`?6+A#$5CRo6kDf40ea^Cf$6gc3DBwz7MO-|O?xE<5>TRi6g%_{gyw3tvR
z02Pr%xEHR}hYEUi%rORE-PYC7_LTV*+JJ{H+Trn{)@$rk21b<r18y7$uF1P@EXZ!j
zi_00rmq4xM#5EDah@epgHo~|{TxuFd9m03|T^-aLbuu=B{)N~hu~Tug0j%$NsRDK>
zP<6!$r&rVAk<J)+B`x&w7v!c{R0gyPnmp+AXLMPRTtq@PG+n%nKFxK^Zfn%T^bz)N
zbx1~k%OK8o454?B%@xl0;Ah;oGQvbQ5&iiM2c7!@0;xd3yMo&hs~=JRG|YNNDiU{u
zO{iI;C`M#IVyOLx&C@@Kx;E4jdZ2|ddc4xJ*zi!7v;WRmr9SZiotsJ=;zoO7R-!*G
z@Z7cR6@|3zOFS4p_p4}djvXuv@o8C1dzZjWQ`7;EU<x2Jyp#(kYuMK;foIa_DBmfL
za4Q+6l=P^5QM2n(@?iL~;uI>f8Ae-Tfdrhy{@JR-`R<w>ZOFQ#c8?k`Z$L14WEge?
zt%{wb0g0zOd|3SxRR7iu8C1dSFG>9eJ#2YGk$S6D^OItS#nKlkl$TAm1&UnVL|Kk1
z0<Z5Z&O+NNpJJ!zEChK@G~6t<Jcs$E188E`st)OXs5i=yV&9`4VJLDLq&U1@V_=J9
zj~<H8@*pJlrx*y=GDKJuy3+umsX>uf8ToXaYhG!D{r1AN{MxXFOy)Qw%gYa3ygb<R
z0D^FNZgXqg&4t~hc{cqmMEn&~GE(@gK00G`AI9GxPn^MhvnlJt06sTQ+Oz8~MUu}F
z5oKf$rk=_yz8R<#DzcuUkhx+TQ!pxE>JkPDYq@HCpEx1B_mT?eyI}%%TIPAUNklOH
zf;%2msk|9J3wh#fzk+A^=!DaZ(AjP*DH^}1G5_upKOx6Kb8BEXr4otRSNt(P5B~yg
z6bYVwSl30*mjX*$;c+^Hr7R-42heKMGzYs0WO$R}R61>K2lNQQI)$aG8MqTh1t_u^
zf%E<}<L=QpVkBw4Eg=~rkwobcL2S}y7Sn+sAB2TI^7AFW=h*#;#2^PZEJ%9@{Q%;S
zIGF2|VwXn|Xb8T(CcW603GPGsfM+3*9Xg_ShpY`Wq4_qeWLM?51`L(jE6C&{aG<<P
z=B1%~frwhlgK`LWk{GIIBS%Q}yhxMGjB@5QmA3ooGo*{w1;$w(9G00l2YF0Gj;%@#
z@B0Gt^L%&^((Dmd%kA%!o>4?txltBp_UH#HtZw)T{C|^%^H<Q^-dRL-(MYwVIM8?4
z28r9AZY8B7MK<GgSt6??#*YXJq*0$0w2X^tF=uI?K@f2i_G8Cp9CeUeij@HL^C4o)
zZnr>sA-JJ2{e#l3&ZNUpTt}V@@-ggqVkmoW1oE6rrL4RYUs*##>a9r4gB;qhc|b1B
zK;fXU`K`umB161ZZEV=35*`Jfp3jV4Tlu8Vq4SOB6DgMQ>=#75(KEVI^;aXfSttTD
zAu8sgMcNsy@JM#*lxYJ5?0EurXqxd{ExK<g?osuDFUT<5`n(uWrIY+4Mv~5&oO5yl
z7~pD_A}U@eL7PaEk7k$q@lyWZUzRz2S}C;O20tGkTyH4<3|RbHg$TmKr`1rf=wLd-
z+bstvY<s-5&>=J#v1J<S@<+tZ3VU}S!nYGP)zcyami8d(O>)m?)rf_7P;n7SPuJ5t
z-F!p}!KV1s^gMKrDWj#z^}JgP3KK@GvZv_@q1$#=vbSSx(+GeE`{^b_v&+uVa=0i)
z0Jb5qpyH;;of^~nDeFVrs+oy5!ot;XN08ez<y!)_zK=*rAy{&oD=rgPmQ+jg--<)y
zVs1o7GFidctjW*jV~G^DtPM=?^$E8=o{pxFZNVGzI@Wyid7kj8(R@3bwy-yvfogXf
z+38vgH<4-wsa9ikGM9x`>WzONwG?$qj)?1@Dq~66mU;qOnvho}IAN}rR4gkVXCETW
zy%1v$$(<Z|>{@W@RVrSTiLrHS1e{h6f(tZ5wFTdjms(E-AFG;wM7kAmKYH8yTBQ>K
zpz40+ZG>d!sZeVfL^3v3=9dtxT=Sv1K%N3(^WYuyN@wFQ*gWD1t-}A<S)S&)VOXyQ
z-S2SzGZ##*U(4Rs?S9>6${BS~Bh-$`o{Nz!Ty-w>hmo2axXTn;csFvB<`F+DHMdya
z-hi*0?<kG<O)<pe;7Bad7vaxfPym~eodm(o#Pz%Fy9-YPOfe8xs*vUx_!>s@w#BvB
z6Oxmw99uvxa`*>tXZ<m9{adX9D&<etzaDn_Zavr4SY2)?)WnyJg(wc-RQ`EN-&ah6
zl%B3FP9HBENFld^794~xX{FdE+$?is7b8pYgL9KZv<Yt6&OWq-szgH4ZR5{mjCmC>
zf9%(vB}^i8yk=mnWk@Ecd(VBB5so}*1>O&u_=(sb>K<wAO`~ifev9##EXRN)OX!8M
zfLRlBYyj*GXu}hQ)9&(VP=HbB)Xr;-k1I3XfvBmKedB3*O(#rV4Aqt*ak!$9Lnw=b
z>yA9pVDY7mKQAiwe#ujfd)X#2`7G>b?$=XOKbb$0K=R=!f=enJ<Ah)Ec3TSHbAoeJ
zjxaT%J+xSY#E2XaMtJWr{|s-~R5Z~Cd{(-1wISH~hKBUY6qwVCuhb|1Ew?2xU%x!D
zYmq^*a*a<An!aM1KeEwj!=gyP4TfhtlxzA5moLsRW_aXSJ&c<NR6Tj)J2~QioTTQ(
zR;w;QcxT_dL!A4lC1LK1Ncpwz65rVz6I^nz)Z3x%KYwr-ccx4p7cb8-8pOcD#?Nob
zrL88#0v4OQ{B+j!N&_cS9)xE2P7~as*^gLMYFlYr08gMM-&AoDJ%-kI<ENzGRhA#s
zwPmPsDIP^^X7`;v>hcSy%n{BfnNwRx;0I(q-3q#SwYt7dzL@^abR;)oEfTOoXi*x4
zWVo;q`=P%pgQYLfI~0#YiZ3prk2*BzrTKNo2XKM2MOPlOp|&Pi=`oCcVpgF>nX`SS
zFU8|k&2ML2ued8bC9XT;_yk4SzAmk`QnteGEKP3G*UAzWBZON&r|%LgxCfuf@)$j+
z;wHcI(=0L@Yjeyoe||}KCsz-mjb-P2tx9V5=U1O-t%a_q)#s~H<4r(`V2Ak)V8Dhi
z!-6_aVN2KsdGFB9kB=y9!7H6(Izd_G34@#Qc}jy(%5CglZ!W~Rv9?L3RKfW1uPQ;M
z+_m>e_O>Qe8A#c(gr9Dm+6@usF{+_o+ufL7ZxLfbb9i2|y_3r73fB-A4PItY0(}l%
z(0k|NBJN7K=KC)27%2zv{8L(px~^ycKr>HetC~<*#fk2OJYohbzwkmrBiI(VwMCl9
zCS*D8WI@0dTH(bH*vs_uRTINE(B}raIlA^)NyyY7vR_Y;vskU8NN|xI@UiKdKR~}}
zKw2)7Ei=f_x#HRcd66BNcVruQZ~sYj`X?o;I6stT|BNtzyCwej#vkwPZ}N{wws_qN
z&2T@*HLoy|=r@LtKk?dXnld3+&^}U!J6jSX*KhYHt*2U5sLkZWwCkh9neAT5@<HE=
zI(30LB<s9PeRfSuW&L6S`Y2UHx)4b7nXO!$uv<Cgdf00({Yy%p0mu7y!S8IkAfN*B
zE?1*vF6b*6`2of8mw+QeJcg!y)_{iVw=PVEhwb$?j_ir-P_Xp{b{m7+6`b-OKBmP4
z{8l%zKUG+7q5=*l-24-o&Ic%5)OZ;}-%cVj3Cc)+6utpI!c&_?zAgXGLgytunqH^O
zZ8<0|9#fm1l&bx;-~-P?$XDFh-bDH_{u8Q<j(%&e`f<C|yQYfRGR-NZ<)r2Caawhb
z!3gJrOyQwI0DI9Rsnw7eNMXmz&bEX*65nyJh2<f*oZCj)0h!BYE5F}NZr{(S_O;!^
zGo`iD{y|zONEjCseue#!rhqN=)En73TmzZ9IKK4zn>br3Evp3@xu3zJWDIF@gOpzk
zmGq4qZ;P5XO=sSB?;)k?2avfjS1z_8L!34)VgfFfr$f%8HJ%Ok?1hAE*%X5d_q7vY
z1-`%JfKMB;U)2$Kz{xMWNR)TW=1Gwt0R(b8NmI}$+1e2k*!_Hbo?!(GLQQe8)?~M(
zKBukB2g{6p<tm%qxR-UF2tDL!M;zkCXpm?+Hpm6*{D?#A-39@*_OLskY^P)Q=SG1&
zg5eSWgD281f?W8E&lU#zUL1tU79Qu$l<$@1UqZP+UE6W{{q8o!I<bXW>uV=BjkSNJ
zaU@JJMJ|SD|C8$Hh1e78=?(+xeI0?w+wj*4=Ic~S@mPXU80|uLiH>|$DsO`E>0Z&<
z9SR~rAK3HaSrGE+2?Ma@KBA&k&|$ckruAX|uKvfp8btNz4G%Kr6<>v_>LM7RFQTyz
z(V=r!=!o2kMxq(x-2KRM$`wSgShuNc$MVwPy4BS572~5GY6$(Ij7a?eLWGbUQkbK6
zzM8=^e2L5f!B0oK^IEKwFse~P%p2O4EjuRtUOCg#T)mC#Y=yKPze_G;HH1F-PMH=9
z7bQ+hIy>&QaQIK@O}Aq6s!vVDKHytemzV~9RwNn+(4UB-5T9W`XuBM8`1&sYfF}P}
zi^j9c8Y*J!l=%MBV`u`V>V;^*fhU;;KCKR@<CmgTD<M+=Ds@CUN1&1Zobjeb$EJh}
zyH7p3h)%QuW1s5YIv?-(P#*T;vncszj~UqqXCr1wA~Z2-v5XQH>iR_vFoQIf)QaeX
z<U6714U?Dnlb4{kr1-QpuUn41t`2te5vJ^X1t4i^ZWc(fa4(+{Ji|b>12xvBk$2&g
zQ<eMrCJqft53S~@)$r2;J-4zY9i<;$&`j{fXhvQ$>eQTT65b+eTqnl!(@Sgwr5`Xu
zPDn5co4)dhUZp-p9tJgmea}0G{rZlL0L({%N)CY{JEX1+Of&|JK=wc+D))Iqi)<w~
z7zatRZw0j4?62I+&M>Eq)Mh+S86O4mO@7*K4Wk)z?kSqf*0*e=2w5TsvO!KYkjZY`
zNvN>N5_#KwBGo%v@)IdFg$JR_Wf&FGG-`2I9NYi1{_!#s#`pP+!D$OmOM||WkBI?J
z&`P+g!xBMNR_A-YhEMH=`t1S$8!}o~F#-z4OsnaxV&Mv1CxG>>Xhhcn5``Y~9QJ}1
z%S#hr3KqL4;V|-LzWh}YnVpZ7j2WY>=8i*$10I>#k3`ZX=&o8Y?#69F|0DnOnUpgc
z$;yha1=Nl-oHM1%JSJ5m^%3YH#@ZPqq7ZTY{F6MC4vMmt;WQt=*%!=_XK>RjbcO6G
zWZHuyV;-J8?)DQIN8v(issc=$!h_TgE@d!2Ha9vBM?ez#h(c}3wi%S)c}wHpo-|XT
z>DoGB*r8O5EJZ4y)S#ooKTZqaq$`STqtHpXtw~+sf8G2zy%T1NTu8m9%b~8m=yFCC
zQZHwg_2FHOAEW+5FpdkW3i{*@p!K<%8QHhd$G<xEf)`q?*a8Ua8k@AFM}5Moe{0c=
zwasE`F<7jF`dPvAYF1yn&HrBLtWKm*;)awBF&s$^bR#)`W{ib9%TV=VXwCaHJx(ei
zFH5!cgrLjIMAC?F`7fX&ONUKL#>B<(>(w#2xY%K79!R6OPT#eYIVLwo6`>kA1C%es
z4fGi`hhQ%!TMJ!Ct@>-^;+mXY|HKPo?T9CD9-K@wT_Iy*k>gMN9e>1(|1%%~^4Ya<
z)ty&PQ3$PANxJpZUKoqmB<@=h{8iNmrB}RUMMc9&s~mco!$=4G2HEF;yaOK}J=O=a
zSdZ*71RbI`^5?LpQ@>OpEVHhGJ9vtZ>dTDLPw8CS8@K0M*egG5g7^@F(ry2|PWMkz
zeeT;el>O~bL2$@R&&0Bt=*%hXD)k=Y5Dz6oBL>y<4?GRuEGhMf<r_HES`5<q$*0)(
z;mUiqu1T7X3lQEzR7k|VbsVfNF{y1RFjK;kZdq76(F>%dDowMw&ZxR?Ni(M7`zIQX
zXq3R_9FOPjmM%new-JatR;=z_?V-^eebt-Qi5F??HmiH6>XKve(xb_+vGUOicWX6z
z`cJ2%&xPZKg#2f|w|VX_^_=ssfuHFY1({f7Rrm9DBwPzuA&xDZ0Xr-&-Wt4Cmf|=0
zdX>Oz@oi?a4)RginOxwYLL*@<`ur~A<*+m=WuE(*Cp`J}qO~Yk(V2bQEdZPmJAjnX
zIivlYoN|5;%bB6ElffA4IPE{(=kLY+o={MS^BJDLxL=z1odgZl_f_(Xs^<0MyNini
zZX*Qp@RaxLsV4XVLC8VewoC;GGKY4#Cm7wNND0P~)RbX$Cpc6RD@~IsgmGn^q$)(!
z=#z#d2}~=GA9mMI#`s`!Fz&+u2Eo^GWz<87{cWof|5QRnc-ZR1P%z+p!TGjTi*nJd
zsKxWln8AkSw{-LDThq-wfCpEAbl}50{p+$8j{I~!Iw<-+0}U3xIfePF%-(@hd4qjB
zT007<NbqC88S#|oXI72)sPrTHU1E61ytN}c?m;8tZQ?ZiM{Pc9l1&_gpLZC~=UehL
z|6r-?+yXj(wr3i!Q;NJT)ik!QHlK1Q)U4;P+jUsm!}_fRI&Hy1NZf#?q;Q~Fz3GvK
zxd1K}klSnJN@Mw9;zup^Ydh6fBXjIHzQoPonyvVYjh#XhD&sfV^o`gbIw1K#DwG><
z>qnI%>nmlY`6A0*Rn0wh&rzL*oRY_HSHRKK#>GYbK>EQ_Q84*JPT*3WqD^}e`c>*R
zUIkTM=m9)s{TvG){O!;7@b{?qs*>LEasJ8Cw2cY=Y`mV>A`^~@G>KR>V@nn8ll<k(
z4&4lB=kMQ7^PzS3OIl6Q@xK2yjqi+~EQn`o^poV<S0pQp^g`)Z@uI%ZJ!$usb8Pxl
zoGHb3?0&p;>#BSW$a9AmXs6u@6u>ZO=(cKh$6_FE^HVwV`_hb6_qwIZ>4E{X*jUKa
z3j%8<^xYJ)PcoJp$m()ENNO}AD)F!3{zI*v-HA|U2L>6v8ZyU`rDyms{B!~yRt<@V
z|5|O(dalfX8A3_XlIkFgt?8TMck5Gkn<7pWlPn&ZgX^{TX?kqHKUfebDyR9785FM4
z%vt#wLal_y3?(WKp?@<lRWnqF9ra#lloV+73?mo)`<4{qQ&zc*LBWR-gf3?6duQS2
zjE~J2$+ah)NRHy4c`ho#Z1YNX1Icn*082*f0~W$|&{Xcx<?i_ViMPn1Zo8!lZXs1;
zyM<*6*JTu+IYYHcRIMfyC;XUvJu^sKHrq6@z5^=H5%HLF*WcgNK5BGJteycmudxii
z4^S&0%B_QN$t~jf_0p%qKM}ptx)U4#x(M@GtY^0g1322K7mF+rt)+i^##US6G38PC
z)UV!U^(@aDepc0}DLk@9zWCik#vuFVcP{xJNe~E)1%*DrL7z>7*R`Nte(MKwJL1fO
zFgkPjjv1+jK@WRm^ZcjNH_(3VqH)7FXDJxfhniUI&7{<w4KzPw_FRrZd|`MEl9f)t
zVjXVnAC<SBr;VA9Yar-6>97hv^#OftN?9$YJ+;uJ%r3CNCYdaDv<SJbirsi1VsRW^
z6P<ZRuhv@eV_!emtygRbtXB-Gdb;WX7O+^U*p+ZvBaugJG~WIn>fSOesyF=iouN|(
z5GkcWQW^w-p`@j|MY=;eW(XA!1nCBql5UV1Lb|&{N*bhNVD{qoe|OG{{py@+?>Af*
zGhD-3&wA#$^ZU7r{UXcBJ0bZcsS+XaAMg#viuphr39I0<o+|$WasQYNvkHQ3AXmMK
zLA5i1OI|X#N2#DMOgM_=R<S7XT(xNZo^kCS8#|!&-yV@KqW)wWwCXO!(2?P3U5fd)
zllQe=2mEH-ay@rT4c5RSqJsP)nJ9b9HDhJFh+^&V-}?ZLl_fY+8W~AX>Y%4fGSW9N
z20Al$so+cAE8-ub`(sd^XjFyee=f9!K^vR|bd<e0m{yf#y+aG0GbV4i7V}9w)|jXZ
zg@%;a$cWVDRx(gOu<{eBa?<d@?uyJ$)Ktkr)Dvc-sF{`wES;r>$do8ziR4NkII5PY
zQ3Iv)6VXd(1T?kD?^mPUO+iuWp>e5#_DdjH<GyN4*b{I!L#r7ZYSB*gaurU6kM2Gd
zz*ssGw;ukTx4P#O_CoFH&{eNP+coxGy-#e&q5r>RJL?hf<<R9S-%-r#RL=JUYS|z7
z@YXwc4QypPK>h)``o#_=mOox!=P{8d+KCJKtLcPs)h-!2U2eGawj}ivM?>FM-{qV~
z_S=^Gxqx2&qc=b**;pb>)B0Hl^DRa3OKYbH(Kj~8_=r|(Lx{?qTqKv%_h7&mafnka
ze@#-J%k$dU4_%0plJ5zY*XRX)`AQ1!JN7*H;%K%OroC;<x!BrD-@Ai-D6aLRwAA9H
zv<$6@wYcWtd-5o7tQIWw#ekUTUk58a^xy6>RT`PgAPvc!oQJH$pVH9u`@72(y-&})
zQebP7-B{!?=pm%G)hU7?x=lb9BlNrAA*75M9Oz1MPRGQy{sB1ovL?QOJYu8wYqw_*
zP)7~$TAb$teMi<Pb&7!g;4Fm$4URj09!yn&peUO4*4tHD7jAK=OOf+yvpjJ)gAPt@
zUT;0RjU4Ns>lohz>G7DVeDRj3$h$>;mtI!>jcp?HP|FEwI;>ToH$kHZ=mz&gApU%$
zo&*FrRm)ceX%;zua^)+{vA?(ImYL*_cYG7}JG1Q!R8J;QxS*1@@jB^y&Or|%?xb^m
zDUt#-HP&^PtCfL2E(0k8Px05EK7#~Uhm=#7seNha8A}m9F+x4G4>e=^kJ)pg;Mq1`
zkOn(;NKaN(K&zj{u&@Lor=frq5*asS$3#sjte#+IUU`xj4Ff!~nJc#E{aq34{W0Sn
zZ}{%Qvw5E-Pn%1LO@DZ9&H>HMFA3gY-k@?~KjJ8B=t!_|Ao<Ex9z%{5N>a|Th!ndL
zp1TDk0u05YYFXHV>7~f;+XSVl4ew2JeeN>B6;UAPxnfbNch4O+@})c9#H;61Yh*E!
zMp#PbIl78)4U6t^k{(SbaJ(pkzF;k@30zy4=^kmw&-9w_%c%I>8dp$wu=BDPXhe6O
zjdto73486$^;{xc#dg%8=rKom)Ad$*x&x$>v%4pVD3vO4XNOzvtd}vy*u{nZayZ3Y
zzj(c}Ln`UX;Pc?HZ<W>-sTG!4hK0Ofx0oOUXU>CmR-HQPuPm_@q|06$xR`~ME`@0Q
zwVHkmUkG(yS@QRPe9YKphdodx9-S{#eXv2B;$;~lt?WZ+z`JL-c})T8Zkn3BG2d4Y
zY9Rd0^NG?WDElWkb`R~6dzRP^N#7%&{2T?SS865hs5b%+5a!}=Ia8Om&!?^DeXkD)
zdA;^$OO`ZFuoIoAgy<53_sS7Lp=9bp^ro2W1&$vg6!q8dp1&w$D&lcwwRQ^5+Sm{W
zq-|3`%|4@pkglI7@D;^%ir`u%qCMQyYtD)~swZ~DvmKM1B1jfDdK`BVC-tMv+T84)
ztW2c8HV0}^@Po!Jw*ord9PDb=d4?(4K~VW~lzCDEnWcH1DJbeyq7y@^E`I0{kYO}&
z!b})v5QD|(*AxuUB?sPS-JD((B)T%|>SJW$L@J^}8-5FxA~V(i2YSKpqbk7xwAEMZ
z8Fc6M2xGAVt{ug_lypTS1>P^LML&YvA!+R;e{zB*itWC@8fw_AOmWQuYII+*$Uq}B
zB;pQ#0<@jp`6eWPN0*2~o5poU{e^L7Q8P&7r#U6dKvrL~)S}nbRnhuxzr!h&x>w)1
z+wSw75$j^=Af&$dONThov(L}tV{ItK>}t#6h;86`mKzRnWZ0X}#VXB=<jg9Ug&ll2
zqYaq)UDK@&4o14)FFBHi2)+E2%fVEb*z(bsnt8!7uB;%-&mON^CW^v0uTS~!9;BMd
zB^9;{7=V4K*gop|AL;pLYXS=oU7tJfqJA?-I=D#Hey#>$_Rw^d*jRot&k8-X^CX2$
z-t_G+6yyq|8w;E}bw)0>xZ$5Sc*)05ph^m!?wO)@G0o8bQFxQUY!wYO;flq(*0g+f
zSG%^061hzm)*guaxMc(QYYR;mKK(asmI3p21y%>leTGkgdSLz*?Z&1W#vye-e*GUn
zBvNFIZQ$Q{k8%BXg!4G-w9GoQWA%x}D}#1Qkn>}%G-$bDN!j&%XTm*5gx8h++icl>
zZc?=DQ`}Bh9co3>PmKa)mianhktNnm#Wl5MY^zAh<354&;gIZ_5W(?7+tl4$N?Y0#
z!7+i{SR#1B_#=6_XeukdGc=2N^%>gwTcA?$<~oC~!4nzx%N~#^re+Zc5EFC{AOZ@U
zZ+ARA4oF&iS2BIKQa@U7XmKl^a8lGg^uupTL~71Q$Qv~M)GDCDTJJwh&%92asWX64
z9K2RtKUm36A2;&_*CmoOX<?<uvq=-#d0!VKevUVlWccc?smxH!rzfzjG(WN_4sa1Z
zHU3OC3-XQRNbSQN=<B`mFyh@fLjX!ow%OCF$c>HM0mNPQjxslI=W|V&XS+wNZ#-qd
z*J4PnaEaNEo2N|8gD)NYPV~xsWO(9#r?BEWz#@kAkQEl+5EmzYJfzqhPg4`#{VQQz
zzcFV_R0;3E?~{pwv%cz7R9Mx0(ijme0cbV})H%Cs{bBCA8+ua$Lw7R@)bKEOEIJpe
zuoT9&vmELISR5MB6DyF#=cY`>4##`s^|#}7q9^wmn%_y`cx6>#l!+~M85L0|yvFXn
z325IdC_P9*(SjzfL|`z>X#d~1B8Pxd$k(QgIAx*jGx__AdIV)mw^!4q=<v8R>gCYq
z9!Wju^5sjapa~`4_DOHe-R(7JEk%$c=|?vl7JXl{>G)f+q^sL{p)U@BGR>+|-+0Og
z+mw4{Ewl*YpyN?e{5LDyJM^`a_X`5H%~`=fZG@;9o|HunveK_qRASa_<766?4A9Y0
zttwcYjp+Jh9hjFj*(<X!RZAK|5NG~`MzMHRu7=oCZ03bpKkyiCF3ZnxsWlB%R5hK=
zUlNaGApIgOBIDE(bH1_~#m2E>mo=K4RGYxoEJ$EJ4DBPsm!-Ai%{dF!ZrPnAAwG&s
zuR5ofu|)s)8hbB!ve6__^vwK4<z!{uM(@rR{v^U9|4JG57E6+VyL{-WB1w8(JqukG
zAXz@!LE!7)T}&wAev*nV2)HZyt6831^J74jZ^2PCUDrq`jNn7;5C>NeMG~L_<K{A|
zHce?KKmZ!A@%Md|l$f;Kw$eNL=u9#veY)d#M|7{`lcyx6JJr$(+B4tL%?sNWWlbtc
za++FFiu}Zj6&ztXb>^4{<ocFr_L>Quw|-OJN4L&CJ}|j)Iuo%26Byi@Ar_Jw({@N7
zTT5KX&$A7HG-fA}cfn7m9ZYh?qkj3(tVAU$PcP9Hr;o_F_6L%cqe)%6-8mh}siORW
z6hPL*FLFi4Cq<71DXSN{J~&{+iKCf05L~ux9ZbQuo~d+15i(H|V`3hu^*l>mr~4~W
z_5iwQ`m)N9<S1c5himm~j_N5I^%$+_#krB(W^F|KPuTjG;S}CZfqbAI7Vp<vf-l*v
zC&3Q7;g|_P{w9wm*D^SXxMbv2;5bV;OwTi?ZNQ4lMBbLxFDs?m#;abx1_Ry5?4E$~
zrhpYY3&+9x^lW4jcD}RH$kgI6*rYC$UW9G6#_YJ%SPY%fSVkY$g(NT;P_nZ-gxY{I
zsh@oasEvMS@cKmXa0<`wZ`8l7n&rwNIs1vNgSbNR&qJS{f=D9{Mg|@JeEo)GQaYGP
zD(b!z^>&-RDi}n(E2{Xpf%l!d!kzs29NVmkH(pTTD@<n7L@i^Xm#2rrK#Tl49Chug
z6Gw=#x$*0tcnV9miDn3`Ti8vZeSz>L+7ivp?<AP^?fYFD0s=*m+uE;eTKR<!wI)V%
zzAw1ld=9FnuiiR!@Yfi@;)l&@*nf)NZ(MJ@GM(Mx5v=NDxff*NbTR)OPQ2T4xqND}
z_nuc6fsPLY`mG9#(@Q8LTAo207Je_i)`_)9!F&DP*mKmhzag_bO@H1;aoUpJG@aP8
zM*lQXE4#s%BJ;pm?An-O<=HY~faxu7tBq4tXybPj&N^|TezuXXI6j4IzU;RvAJQ;m
zA5W~o$1%6dk;XGGXqRldhXuJ`>`x#}DM0l2AGpnnbxo!|8!F}|YB@KQ7ETnX32!p0
zLMXeq($zIx9rpvFzWObB>qZ3q>`vzM4l->lJ#@Xmm(}C#Vg22!&|@_wwX+8W`<K+?
z5D_W<KC#nl9M4BPq6$l|-lH}Wn2z@2Mt}wtf%U_+XpCsMcv^;Q*ywALv~j{7&u^}b
zHSPgFI|)1DRQfGg_t8poKfR|cd=Lq*^wGI_RV)cLNVG4aAN2{2#tqDuPdkjHArPG!
zjbihI2;Bu7^Y^@M+H^cG+GkX+-dQK$l7j_m@%x+_AapUOVActNXVtJu%P&+*P8Bnp
z3EByC#VRp0;sHW$sh=2bW%^lWxr})_y9M|80b*UAc#De(QxG8|iyJvbG?yE!BjmU{
zK)_7%x_mURW#VjyXkCT#^9VMEA9C&2R>G1YwWtOLF6Opd!Td3x+01({G=GiLNJZ<n
zO;4*<p(+*kryV!mRC8qSx2e(HuK~W2h<8!?MAOg8-zIyt45lfUaJ|uKudR&E&p`H8
zlA7_%jLMnG{H(t?6@adkOes%EnP3&Va}!jXCC8$@XL6`$J>gh6&f?#ELpn`dl`w<8
zSuTtvDP)14IeM+>jqe`0Kkobvmf^9;a9bm9K8I(o%71r9>cLVx@f=~T+7_myuNA4J
z!Z^NubGcA=Xf6H7Ei+Hd2AWVWHMI5R`MWo9hRM4(B9!m*EDUrqyKrL3j`COhKLaG3
zl)y~lZJ<ZkEf^OwiPK<><LQvf21D7eEvgl|@Q?(*e(z6*<qewD&*Q$#Wkm%95gb!f
z@hZF&`X0z#y;XI8!6^fCYUW*N){79vc85N5*GP-6G9L7Iiefg&kumhrzDuta{@6|7
z#jyerXkOD%=)zR#+VneE5Oy?lNkA-zf!cE^W9L!S0{>;%1pUO`E^7Xp#MF+fU}+C&
ze&y0{v~^o?m7G1st=dE3#`Z#l^R-ubh~i_v^ob;Xgs%yDkc3c~&Fk-e;inDWRY}`~
z_kX^;(>cm7)lyZ{*iYrRp%Tu<?O;U(W7F71$sE6z{rd9lPU-CFIyPYk-xK(<HFYG{
zc*=(s(K8M53+l3ub^94(^Mhm0k+e|Q;+f3*W9gFBzox|=grHcOIg13|w)UB`BL`_;
zu2x=XyL@-HyTiIq#0NuHtQ;O_MSZYdLb(q1``vQYgO<$sDxG8{ly%)Q95tFle$&3N
z{U*Qp8hwX=pc}(gmHb)!;b{J+`GR*)PszXcAIl}f9#!}GWr6>ly}R*5)vwwFRyj57
zrEbU@|M+OP=@-}pO^?@O$8Qh##O39JIPmhQI=tOs+!7q3*e49@e3Cl8Z(seSWQa!8
zP9%T9TF&=%?+oX(FM9q_(D!$DO)wgl{=nbJoo1r=AFI_n<*+3HVr4wSr8B}(%HHxs
zQa8+FWH8;LBY}|D8Aw&65{LU25I;p*sNX4jG>IpwLLCx)7BLi2X-tiy<Z$#N#Tf*(
zyqU{Q%^2o&`QZ`rVxs&9ez^U+Kq?};F^vZawdMFu%8#I-TUWv%d!F^EwYX>U?FxP@
z(QS?WpMH0@pJe+;SeY`fi@bsRmb`dvQKCMZ*3m)k`<5(n*x|x<T$<2+S%c7ijQ*F<
z{I<Ddvh5F<Uig|YkLzFA#ATBYHUYS#h2g7C1a@<M+6)aIQKwYpWT@0%S~WKO6HLTL
z>9=X>=WcN076H(OU6^$FD0N1(++2Z`8zVXe!}=Qw?uZ5lipctr!O`E*hQwYKbCz+u
zN@|7b8=f4~@3Td8!mn>`ypKkWT55)D%fFaq4SE-*tSF`&*GT*kL99`sOHzp(N}bYE
zS>*&w31akfOjJoE5PWW_rRTOdrm{CFqni*^`_4K*sBPFN6Q4+%7u|0^l0GJ<N8R1<
zK~o!04yy6?F(j(YMtqJSNl4M+)jdcN(<hDbr~4j8o;`h$8h?hNOHF>8KXkzd@K;Cs
zu)iP{2cA!QksI&MGl?;6d@NNRl;>hOWf))JzD}R#X&eSUR`AK0cHruBA;6FWF>NRu
zJaWnFd}{GdDMf%H_^p0nXS&%;8BZYQk*y>8HBPurm~#7)7C`#i@GX>is<uX&=WOf#
zh;UG^1I>@pHMP&+n14symHNkV9>9kCgI^|g2qPsj2><+MFj}MSfQ5XFRoAhDZfD2)
zfmx@cFUk69D-k0j+PBcvrWGJc?S6Wx<NX~kmFRy1yBh{IFaxx^|1&^%|NG16)4rz_
z&7)bn|AN;9kOV=BuT^Esq|HMB?w|k&RE_w5a|h9Yzd{4yJa7lkd?s&zKMMLE?%=)&
zHe6e$sx)FS)?--VOWbcb7(%*rM!4O$n>Hi(cG^=%oi~)2@uQ9`M5+t4O`@#=2Y7|C
zOd|Z(D_Rrvz7uCEh{K)39oOOedzo0=@|7mfqTR`#XN}on#!MJe5DY0MGEHAxoDe#k
z5DYl*gARuT^zaLv0_eXN>A?TJ`2X#j|9}3k|Hu9P-*nUAvofOGU@1tef*TTcrxS^S
zuEa2$LPyFYmF~0t=fPPB%oe2JWbfkcW3q(&B_~sO>-Ka;-PHEJWc0MMSIxBt@kTJ4
zkudu?v(EV75x?fY0=Ao@$Vdg8{~muh1<p-Zdt*-FD8{XH9~*0IMn?eQvCcSuo~m4(
zwvzHp+m{u+8NDQJLE7q}Y@6KN9fZ942sdX!%Lm^oE|+X%Vzp6STFk4H%f6(=7A2Od
zL6{kv;2TslGckr#EC_<30S<<h<6bnKxaZj$=1=wy#(mM*Q_MDu&a1;ka<R`Ov<jvx
zWDfRIl}rd$t4$F=At&Nt7Mq8qC!KY*Clg=JPkvlCyq{s2OKCD<PcPRK;gX5KmV%H4
z&`jQKyd8H7Q81_T`4JoAC{|z5X8BiSywXtoSA(6mNP{)wj>eB~lOrW-yB=f_KM-a$
zh)=QKQ5?_kz3-W`QRl_Y)n4C_HBOoI&*mKQAsDzIDacoX8=NcFO+D3-SlvX2zMq5(
z10UWr?UZLq>Y8qhVLZ0EWeW-mcE$;@b?vdKo(JLA;5zHS)LAxa0ze&nXC5aY2AsF}
zaJul5h--)={u|k)wECF^1PZ}ovXLTMDzM;UmiOgiqf4tyZ4BgOg#Pnw8VxHOJI1II
zsJ$@tD|0zbK-yXHl8Up*S?Vr9WFhYVJY(s-J;uI|6fQPFWtLR!OV@82`-wf%lwBqU
zj!avEuP%0oGbXgY+n~o4O@Qr8!K-9ketjn<>6;N|h$O7UA_W>jFoqPGBqYp;ynL-K
z``yir2<G)TU3C%eRrIDisqu*Yq8;qwKx3I4bvZ_eybp&ngD<@)=<vN9V^L1`Ciay7
z-byHfq}0(RA+mG)x8dI?7D5ZlTl;c8J$v7fF_Bw6DQMC9Sgmi#j7UN^C+v5C*(gSa
z#O1SVzrE~vn$`E9|MtQktu6U-n!J-1mIe7#gs+$9uOcfmkpedx{;FiI+pd7uqPygK
zk`<p0%|DNEz%h^wftlK4#8Nbh|6M!*GL7wTnj>v6BB?JfrffmUSz?>VPAji&QlWf<
z@TN`_5;1!8c=ww0AkPO!UMcoBx)rR3>3+7!<YvV0_46lu(YkQ-5DfZB5Jc7;`yOr1
zDJ(F$x~Q*rSiW!7&nY*x51Bb9Vd4CeKZttuK%SY5jSU;)c3j=|cX%r*e-gbH0~QN^
zGAFq4M>TY#|5{5{^Xi`TVb?IdjiI%2!?|6Z{A3QTI<4f(mGajac!dtLPeDVtw5h#$
z!$rIe&=uGnX37!)y}Xu?<1AgeuuGD-|Am67Y1X~)+}xaIlb4g-*!K_b4;Kxngq`=9
z(JnkN8NgS*!~F61v?WID@zdPr+hBdn|8_*|2G3C+e=y8pnPy_8{97<87waJRZC=rG
zmT#ndpV@NT5y4F>7<PHMv}}$Dy7M!rwNETDZ83zk*iEp^)j6e;z>&y+J7EI@1I=pd
zH-yN8h5BRw@TOdD+R`kWqfq>q4NMRg$rb*nd}$E^w`!*R@3e@cKPGQFY;VTSyO&p4
zg?<Y%x~mmFE=X7@?3ruwzSzG!SSSF~f`b4BUX|+~X@h2;N=~CX6`-2R2=Ms`clUCr
zT-PF=_XC&-20#E@LlPy))Ru;I{&Pwe^UDX7dt7N&G^TJ2y^A9r@t~-#%(=oJiz{yM
znhX82Cq9`snjKVUV9S3(t0*9M%IAG(d>39qn;K}xfEW*hUmRt^z9hBver4PQ6hE;w
zH8pTD5V9oQ{j!%>j-Y|wCfS)Nmm-!<A50e>HgI#Q5)+WVbDBL<^EvJE>D&Gli_Dw;
z)PJuq1D-6@yCfs-Z<P6OfjFT*ijv<F;Bck=NhUk<v{QozHABDu1FAB;0Bj>y<A<F4
zZ({j`pSY68@)hL{LXeJMuH!eQnrw_iv2@EE|7~1crjrG`?xiB?ZwrCXzI=5sIzN9g
z&NhDD^amd)A|g^{HSols$}%G0{`&99oK1G$$>u=qJ`u5*fk8ka&0XDM%*+u2^#af|
ze|f(<sc3|5Rf~ovR{V+DTIc+bZMBFl1hClQ4WDY;Uza0{&Oe?6T+)wM%P|fG;7>I_
z7u*)P8Sg%SXlbW>iFOZE=WELm?|L3FW8C-OI9DOn6+k?myV}2wJMi-W5I`A5a2IcO
zHT-`ryfTkl$BTT!J{IwF=<UhTMUR4<ebEfg^U{o9Gyd%mG{O@=d6NXFTKDA7uBEV)
z@NdRIayY`yi>&RXd@TQMnYG11Yzqm?yQMeND9hyaO`d8zZ&=@}+Qnen7SE`i2N+A*
z!H>n5oX^nm!ki<F0JLNG&(cYExDWov8rwj9yO0j1Q`dms1*Mw_Gyk+#b>GdA6RhS-
zLt@YG9s}u~^1nmr1#|a6Huq<U#My1j$|oCSSz!pFXQTJg3xHwa=|twJM;0oU_c&1-
zocpo~<3GD3`#Spg4<lbd6QG^1@G&3?(c|DBp2;%|yw*z(GW~!sfoo*}d+RPk#A5@f
z4<-ZJ*gilt)e<6MGg{u;X4YI;q0oaG7CTk3n<>j|x<|BIzY9xv*A4QftV-B2Fw59-
z!yn{kA9eV0fh+OdFbB%vqWIr)vMpKvAP=p&6(fx;O8X$DKaSqkT%b0@GCmM~`%%F0
z=<Zay+y-Xf8%&G12iPJtlTaXuLWo>0nIpY&nuF@{0FCz|*BkLIE7=X(ci9eg)Ivi=
zFHXH3W4m6V_!iJEDUCJ)HRl}hN}S4UfOy@={gR!&S4mlzEa^Qn-iMk;EI3qPV%e+L
z0rnk6Pp{tC(dxD=IS3?#4&f^WHAC7Ux$0Zqc=J=v!oBVYm(@=>F8v8?H-<H~8b7yQ
z@w&B$t?)Yr_e4=L0#FL?%scYyZPZ(WFu`_ZZ|BL#e)`%p`wy!Nr4E+M7w0GO*G<e@
zM;DJbC>fIaivF_!VF~`0YM~KHm!DNeNBG+j>bsWK1A%hwFt_T%IH!;nC>`l*$arNd
zq_u8>2ixF2S22DA$eK-PwCX=X){z7s(ek5W#4l2e1O9cHO0RnTu56o7b<^T%uQjdn
zuU@QiEedFL%ey!DNN;YC&kj-ptC-F5Du>X8ae+cuowP@=@Jm)-*TqD&CgI-5&)C{k
zjj*^Iw}j??>br5OYg{SwM#yqaF=8Y@FoSTutDnQp#^m_`>B9BRUimwT<5!{3$0*^Y
z_wmNA?U+*7u$4A5WRQOG>YBr6T3QT_DSoT5N_&&k*-Mlqa=t6yszT#QYF#_pgoI?e
zxmvq~)>p-3_8rQu&k0X4q95FChQ8{6uw)CPW$`re?*)r932ABq_`ZJAj!Y)SGs%9b
z*kx;P{uNGjg454s20yu@sz)R^0Ez$T+`T>#Qq|4E&#%!KAd4u0`1trj9^K#nPO^LC
zu7j-2#c1!i4ZhqD$zN)Q)s59W3KC*k?Qz<(Jo#w#;8)3&YZSKACQ_?dzbqEZMaM7#
zVSJS}X#KL23i>RmKH@W)JzW@jkuo;oVpcm0X0|=w=KkxRRn>hnU262)rR6+H?0izI
zF>_>2EoNtLxA){Bx$6sjlk@DSiZ&nJOPR<I?SDr9#mzRro0yTBj*6wND^J{Ps$=E7
z@*H`h^{iJ|iyBjNDE#H;9)3lP2md_ww^3_+5%Zl%cPMn}>BSt&JF!9~r}!P}uArfA
z(fm&<xA<3VY(nIO9-}813IRw!0%i$hR8Ih*)3!yHPw!12Za=_ewF1yznm^>C`N&bS
zp#-)QZ4>Or%ot74=B`uA2PZAWBtbUp=(Htf3=S8E32D|Hl6u5i{E~p5Dr@1mBy+rM
zopD8_Tg3i6eM~HB=Uhk%@p_X=lbK5^pv%UPLRB#$g}Q~Gzx!jW+j{rTyfNoaL~_Xr
zx&HO%)s+UfcTPrE`1fN)SXYb~ud`ZVZpz;_E`u-Hl#7EB)Py*bp_d1f7a$hA-#^Id
zaK2#yr`tdXIrE=Pxxr5|$dtxXOUr`x@vgJeu9&svmprmU@6Mkcgci|mx?A1>5UVfg
zB`d<W^2fB|el=xxI^&5+TQ`QXp-*#PKBX$HsOV+n{Oui#2Qh^_QZ=rZ-EuT3<w8de
z6v=VJ>oneey(|CkJY^Dk9cUqUVLl#TFc*jt<=RL~3VgG*qUvY>?MmO2_xh(|C*=&F
ztmpS&fHCsSHw#1)!@FsqH^^FDsdy_x$zu+;w6f|;6LiSc%u`~&52b*}wy6rvzWFe0
zruUdv!L7zK<6Zhvk?<%vk5wo4$)JmeRhb>>e(#C{!6&{!+qY@*4w4L#6_}C`-go1n
zGQKWvTMhpS4V5`R{$mpT!<c+Wztj}B*2fToe@L>OLs(cZUoHCuck^*b(C@8QpMje|
zaHSd06qN_`5a0M?0OEDsGGl`x?VoFRIA(BB&)w}F;2p(KEW)$PD{nfYJRL%$NwR_t
zJ}0rOJ(fA&+2Q<$lZ$NVz;4T{s30btb4Fbidk)<HjGcz=8M#kCvLfu@ZQa3=`{-Nt
zeGZX2y+V|eQUbaJd{Yc?35q%OD<qu)Y=$$~079XE#II{hP>5JBoEanO>12t4Rxyr`
z6cnB(#bsDC8gumuW*DNNhr2HuTrC<_*>R*IXuc6S$LG@H=ShR10XL<KPXs?$IVSt+
zTi^UaL=&B!O=G#!l-@A!T+kkSt;!s<llrWaYpXiAC-N!)o#^I@L$54>j6-)@z8NSV
z6B!_aTCnbh(uHX&DxT9n`U+wWtcN~FlPE&-nJP8X!<8!5Ev@N*$2iT`Xna?npQ%`p
zX}}34BYNU%{0xk7Pu5i5n|nTBd6S=5Wk2F&@GdMg^=C!!@R_~Rf2O-5<-j3z<!RgY
z`i!8*ZL$DVf7rpkAE~o<ZF$S~mU`X;i)Y30@!TEYzzB@Ww4(sV%r;UYMB?U;dgra^
zCbQ-$Q4-VDZOar%f(|a2(ssHeRTtB+&XEnEOX*0DHhMZRV~#!xkNK%dXF(!H$x1_b
zHWh_zwMpk(s#TQZ%x0^uFG&)mobyACX~7?S9Cxgsrwuq)uLhAp)=Ftq=f*D6dtO2<
zchlA)Q3p#;L}cM#`%lKI<#Xf}X@g;a(?)gnuVdx9x|@)hRCOKzQDYq)HU2r0!l|&J
zu*{M9^vHcw3Y(3-jf0<m!o`#Bsb!l1nQL95dUT^l-+N!rDYZt<qDEeFrOTSat1?n1
z5C*aI>WcMr+jMUs6UuQ(?TfGkl48AhpY3#^jrP&oY{JK_1`<`imRC3@z$Tp_@4f`w
zuFZOaLjqlbtFQD}_eGet<nIA&b)(k-d*&<e3tilO+waw)kYU~6a#zR#E>z$kSOXEr
z@#@5V(uzk$5KVxb0WcHNXy;dM>I}!7sRvXOE6{2U`SqR^$quK!J?`z!&#<W?9jdd+
zh@*6u)~|ptMpyY}7=B1CMzPOPmbvPGz^egYg>3`EQTk6$zvs5$;b%@@B@u)b@TrRt
zn;$985+V@JmCcUqXt2%wq^JZ74|@Cmj$~Hmy-(kv3#>Q9S8Bq+h^FVq<g*e1A_K@g
z%sT+(V?Qci(G$Mxy%JdPA}|j#5DR%lhvAd`jedKgAm*vZAn$m+iy1R|A%VFi@1kYf
zazP#Y`9RrF)u`k1fIpW<(YD9JKdls;B0p`AJOoj}W=fIH>kC#lg(k&<u|16hNk4XL
zIKOQX$=4kj$t~B|#arJ}4Bc+Tt8D<~(9xWu$49)|J>2xife3TeFmcQl|I4KHFRxU;
zh@B{8^0Z9`Gl#WaMkM?YTy<1?Yrk2R*NwgwLwk04Op+t<u$h=r1A!a6e6mG({@zOD
zTf5YW{@?W)7pgCOK&2@F;4F$J0jzP(R59PGL10co&*{a%0yz&H^3UX%Yisr~Gn|c>
z9jze4J@P4lu`(np)Gakky;SoRM6<<TtT@)1vz<p9t+LF`_bbMFI9`~<O1!8jp*P6q
z(=HsBo-gI7Y;;bo?=~y9&&3-mO^8y({ceB}9|@HF*0yT2NzH?2idgDowg3gR4oCIX
zfq)3m!E-*M&Wl4%%=!fUB!2l^q-o;?Kksp$*?`WT=F}DTs@6gg_#w>3ZAudEiBde^
zZ<<AwE0R+#+r<j-8QnP6g@y7=zCavFMQciIofN2~GCZJdIZd1Ofp|1W9EKGb2VS|P
zNdxIW2B^xW#b|)2YV3*B%Lw?y7r#t;G;hv(>1BgzbyP{59e$gA$_)rOycF691Fi2$
zDWmNzns4O2M0H;WEhiSU-EyFB5+ZSqxBos6%n!VIn-HFRx>P3+D^W_(9{fK_I4<CA
zjE5esDMI}*bKeNK*Znh6bX$C}%LRaXdHAM^&AZ$#*vh>J^ssoU)7g63bg$kVBXaD7
zDd&7EM6a*Es3r;@cSc@*=58J3gK8sr_9{apF3uy=dqLe;8pLfc>5I^MX#7hKd2#TG
zE3jT?t^8ncjj<zaZuQIBH6Itl|MYOk#rkCI$m=2mUCJkgz^L{|=##B_V%5kU`itv8
zBGbZa0M5#&b0RZ&34pM97oK;RlEDdin=vKPMrqn~J6(yv&vAo>ARt=|bALiodOvIk
z6G(QEYk$se-``{CY69L^x<Nm&2W^o1;cj);ip_O%i-XLD-Tz9uk;2G$(%e^N7ZZMb
zw=cu+hr<={H3xE}LN{{bRW(Du0iS(~45<63C*pMY%J5lRjU;;uI^ZfX&=@0EKV-d=
z4U$hYxUWusA2*PGIgY>=+`AM!3Kl;C6rDxffN8t{F88|$nIs1&2n`5Vn*l<+KH<lH
z2X^g^)R1<Mlf|QzRIj#xYFFFhxes5JogNnY_qKlaHPjNRGr_J%U7h5u5bEVF-*j*K
z8liRf)PB{h#jp0_>f3g<sArM4SN2`tcZTSX710o;)w2n;WB^9L{r&q1X)r^em%5aZ
z!!#kZCLpQV!tuk~=Nava52h|0lTJ&jJ0@r|Z^*y@4fZpZXbzqDlA8A}<xnnWpFrp%
zHR>zW%^o+`znF(99{%p`?i5!MDU4`5<RL<<06Qo7{_fVHxb4z32_laL4o5#S6_xR#
z{7&cu0<u$Rk_xXyubl+tfYQ}Wg1jhUk``4Rx0}z0wFMWw`be9*7n4Yr=6@ffXrZP$
z_XJ4U64GD*^t8LZ`kLG%cRh64WYpIMyD3OtjH_<WNOyp3rM2`&m_fHNT{fE3!(PGL
zzC02N?4u`~ANXNiQvNgbfct%Tj%GjovCKyVfODmM<yF!QG!|C5t;=BrF&u_V-rbyU
zcHna-e3SXK8)<Qc6UP4dH=<uS5hDR*cC#)|S$E|jIN8x)PVbHPuybnQGPa+2Ij_?x
zcl=uyAZ#Q0*{71f%fIcz0BNG-U<z+zCXojcxbkKaaKimUjU6KZ&9BwP?0>~;zJRvk
zV#FcLb)((7)F{^#oIye-EoHcgA9E+njs8dUDxD6|HF)k_qN}#N<C$ccW@_mhf*@1;
zwn#Ma&>5<;3~=~s_iJGzZEC+ch}&A4fV9L^h56L`Qo~wprj*pwox59e1rVBeK8Ctm
zXaSaNh%J7O9LNOp90Nxe6drz#tfv%4luhjo%!aDzkwE&_Xjb-Z0LbMDP$&OcwIVVw
z@>POiY#t4QCC%#7Nhqu^Q)Z$JSu_uapZ-d3-4XufxcsTA{UxsMrQ<l$!$wz;xbI5E
z7D|@dLFp&2Yr{A8%%7C>5;^#o0sZs7cjq&;e_fRGfz@E}$60*`VB$Bb242J?0C9~b
zK(UOK4e6rB`ka8nIa?`pe?P%w&~VCK{kED!kATz>(DBo}GUK{EipP(RLE$f{O}oPi
zQxW2;=Bpk6UbGJo>9GLWx^tRX;N!n{_@ds`cn)WXRGGJ<*)29yrr}^Nkg%Fo=r~qP
z^_Gm&%$KAau++!{UboXCI)7rrsojIw#`Wx`6_@k5UK!m|r(8OhtxU>m$AH;HcmT-3
zlkFcd{eaGEvYs~m)^S=5(r3vXWtW_%oZJsEZA+b#3h}hdj77GS;8U2EX+?{Q-Ee$r
zVVjgXkO0`ur_SdWF0eqegh2}`t1h;Dp9|isBE}#f*k|S6dEba0RmVW@^$%MB=q}Ku
zt3zQmq(Om<uq?$|rEE*t1Ln)Tck=A~Na(o}TWtb~A{4&AZ9iW~iImhyefK-KDOyU8
zxn@t{;jEE1zPR;HrG!0Yayb{B@VTjHrQualbbE%(6lzsJi=h%!Iz*s?1TQLK!&c%L
z5t9F^W3#il0k+b#B3O+4ndtl2{e8zH{<V)dw)pMUj>X<X-F%~*e{L+-@l5!mk{-4W
z38*7+_9#s^aC0hDE#8oFPHN#=wE7;%yxC7ffxc!SSxNmm2!l9>iU$C=7}Psw*4R%e
zaiF!^qQ569xIQ`=<X@Q8K2jBuGvV-Z`O4fR<a_YzA=o*lM$%v0>g@&bjW^sz^!PBP
zP{OE@Kt`0~A%G>I7E)1!d%Q@a3m-2Ak9Hw}6YK;?^e7+9*X9BKX<Sm$6K!R>p&$5#
znW<cV1$WpNhVJ3zLZ}yW*0nqjW)>R@wHPS3KQnT5L%RpL&BI_m>jZ!FY=K+@J^q1)
z_XQo<bRp$fl02UelCP=(M{q|q*c>CIYpu;<Xt?rxzL(|)c1mjm(}Oeev>n=|b@6HR
z=&k&*_rj5P5)v3`!u6nn_PhE~-q)rE`9=!X9&bgSC^JuuVhR9Qxxys4*3*}TrVf2<
zy~%dedhEEY(*zG^h$c47@vNy21NBLi5KrXl@77k)ha+I(MR}fb5)zjoz^eloq@wA*
zRH(tY#061NmVAz<Qus;298M4RVL;5*XlDN-_<|b;I_6W*q(TB-nsT7C$hIrx@)#*H
zkCinK8Up<t$a;+eXvP}<NR;m}h=a(bY5+WPdj7~$$FS<Est^+FC%y`88lFZ<BUI6r
zTU$Dflf`kw*zVGBz+w2lN)$>Pb+a^fsBo?J349xhz%NBq6J6gPAsqTpz-*BTcV={`
zl<csWE%2~*DJy{V^#hR`i0nISU*x1<gqIOHo7I*P2u#6aP==}`l7NFSAx#?|ysKBl
zz<vHwr_6S4VUOcb=xQMfLpA6DL!go+_A4v~Dg`0=#|ECo@dk42rD03EXeFz{BTOK|
zXoJz5MmjLZr1Z<rOb^$%KM2bv1`C#GzhSw!gYLKx1o5!vg2lsEiBtf0#<}^Q95Z<T
zl2rD6SMXJX`=-*;^U)eRExVQ0K=n1rE5EjtA5zT9Xh{RsGHCj^!1E>Hy;A-+9^1b|
z$+r`Gnrn29m4=y1jW|xAdz2Lt0-&bzk~aX+dJ=H3eYgt>iH#h?)K7kM<tRxs+T!o6
zsHR3_FAcImD+=ynS|yi3n|<`~nbUuc==ryVe{&K9Yf&V9P=F|*fEiwWq-tXHzzCNd
z=i_&M*>vo!%oRYk*Mv3fZ?$|3jVb;gbHz4E88nFymbQIR`#mTQ1RNybq?=XV<urtv
zv&(`R8R~;n!|mKEQ635)<X8bR2UBG;ziU!i$5;@KYo)yT?Zv@(sgW=lHkLBl=pzvB
zl@Gci2p2^E&-6s?ALn@=H+yV4k*LbyW#Jt&bC=HC3Kj~;!}bAEMaG9&T4g*|$IruO
zjae~>9@PMz6i}QdrNS+ge|V6;FH^E{Z<<a|H~Mu0EmbbHHuk$STkK)Ne^#(ygXa&F
z_LH^G+({5RVhrMRpp*r>HvnD}&V<0-)GN|J%4ib7JfpF2&N5OKFx{H+j@iuFnVhG1
z`%C0A7EeA7f7%uE5NLBR?>Y*Y7pA_W1|fmwOF_0naNh6FY5;kSt$B_%MKmm*d2qVH
z!;@S4u^%%?3x@U6{~Z{!rhVCsj4V0LZs5M{^ndy;XuUi7$9Oi~S}aD1VXe}`oHJ%H
zmr)((zsoX3$rjy{ul3Jj<?v!%c|#Rge?hBI*rJbOzfA4^T%eVO+ezKr{0O)jf!%9$
z$Iu>$b5bC2rPuPGnGf9a5Htc@X{(#F9VzjF9hz+T@YL==n8aVD^%DEIcf+LhLVtje
ziGIxUYQfGcsB@0!7Ey^MFtM4E&41N7Pu`9KD4&+V8e>4PExtmYzOoriAr`=GYj928
zJ>N{EDcC`b4TVVmx8Bp8&ZensJMIRjSyCt?YNQ5V^H!RJC6MzaxX1Vks?Y){Cvko)
zE^B1BC5ip>6XU6AQ|h91claOXxHN`b5OCg<fK&qIO~1S^)YUW8E!#O*$0S7p)4<e*
zNF?sB-eJK)6xwiu0RZ;1(MPd#&9cr8{$Y`kl@ekfN<rUnV#6cGQO?zi*y4@C{#o=x
zRSveP!-U|9cG~nV2ofO3{a#gr;5NT;1y*{OY&1`qoD8e+AMSAmj^PchWjDGvmCIKH
zBVENhLoCt~5QI$V1pDoy$;Kn<Z;z07_#p)MJlDbVFkosTcF8R>jx<TpGq_FAzqKxR
zz<U_jpQ|=W_oKfe_Z$8oAC-A!5G|T2*l}z8Dw2c=6c#qKg3ym;Jdqoj_?&#3VrpUt
zX{W9GR=?=}83~N7+suGUF%jdYKw+)OJu_4s5KFZS;!0X*m@7*DeV6AnHD6WElN}ER
zR_L`QYP*Y&fBF8?w0*vGCIPwQA`J*2g~kDRZe>&d1Hp$yk|u%cwgFASxn$saCc2?@
z;?Ys<XS7EF8T-5kJ$FuM(uF|GiuWCHAa~}iQ@ng)-V<CF9h6~o_h5m20m+ZukuoH%
znK6Xo)&y|VpZaZo(=68I(n|wn2o_4zl^V)lP_2C3q1IEXGvFU44FjBU>Ej+&)$FwB
z&86kPwxitcW0&<rVl=KLXU}*o!d?))J${a(Xg?#Vg6+86Z14z*bp5GYn%1-g&I&_+
zg<hBX=0~-V)?)?0>&_jxXZ<C0cDZ6|qbR<9X_ga+(stv=!e~C}=WquQ`S$qm_G}>E
z4NAd$o5!OdsjI6iNs>dwG-BE2c1{$R=wH`gzc8&ZRipCa5d)a6gZoB#thTk5uliXe
zs;zzHI!9&)26$yVjq1Qk(E)0}{#9TW9}^eXn@6wqd4S{5TMA9HSxf%YjPr~58I%PJ
z`h}95Jv$rdHC)z*R&#70HKrPxXA><)(*-MXc+L}r1SAOJ^G7lq57&j`rj7HBFyw{p
z{%mF0hj!h^`ZJt#{_~EsU3`3zXa>NtBLh(!Mi*bH7oHZS!u%kSsLg@oy~QKI%>Ok!
zP=?5!P+|sSm<vFtF@W6Zo}WJ*JJ_jcQZqisE<GroeJ9nJRs~HmtaHaD*|*S8?<beb
zrD(NJyILfMaqrtu6p&LFyi&^)O|$}T4h`>v_x>#XL;9!>4va%+^1wd}NBi@*fz3N+
zP~dx8e+Gd$KAaUXVfIW5*gC5dxkYcM<~>NgBT<Ws#$?DAfJM=qf*}>YkVM}MUL}U}
zN(t-8aNuE#;x#l|ahGD{j<Ll8p-gPXJ)YQ>gbQ^S%!U6#Z?$7*gl&-$K&}Ag&X00X
z3@MCgZUj>EPFnr-`ybRwp0k+HiE~+zL5c(?;Zo%Xqahu_byvV1+B*00?k$;35F`=?
zii8c`{wR<iqXi6YE-S<bfUs@=20cC_<EfsaY$Z+3s1a_Vi(I0|8#4Y6Rg(4~oXd9F
z!Ry~xPD3*9Ek;#g6n}pKLN{&X|9r%n58Yr#(EYWI-9&ygP#2#|R8%jr;R6LR&)HF)
zwBO{-x8_2DS<_*kePKMdVmy~MuNq%g_$S>r;NI(BMzlg@#k^XR>2MgdHRy&D0DZOE
z)^O&Y=(n~6l8(DXF8_gt_Ku|H7Dr6LopTO@`RG?{6v#(^%o<1dkF5Z#Jr)3j3+845
z?q-;*V-y{5oVz`h$9BOL<rq_Nu{t!s$7icPLv&u*qgt-p-9OWVk`$XTr*9;H>3@&!
z01M^yrVSW`xd9X_bWJpPSfmb$rxT49Xy_MCJqCT{&;Wy%Fn<Xx(e<F5xJy{W9(g@7
z_~if?#v=cbh*mWQ#$f%%`dH~LC|nuM{2!oDF6QA$@gQs{dhJ4~UY}RSV*@lvz+>YA
zQDX1D8I)^U`Bn}5yRjJsHwfNdpN>E`fmtwLA6$y|>{xv+=5?zD9VWDYKe5}UC1){Z
zh*q~s*l28CER!EMh%E4s!82I`E)_vIl_9B*<M00V9yr}91NOos8aao}$IY!~bN6r@
z<179yNqcIv^+jZh;%L~{A4T%709(Jo^&87mI`sA$v_|dX^3oxv5;UA15bjZD<q!sn
zaj%YN4u#vb9>jK-%Xf7ISiABfSjmayh1&16qB*HhtiFf3R<em#D&W{k0!`ygMs<!F
zCuwWHRwAxCfQ%+fBD+o;AQkN3z<d9We37m=QboUhl2d3P@x3!46Ij8o9(sy1+1nc@
z{7Axt3Mh*w4J?Mc^JIS}gmK=HR#G6{pcdZ<2z~;!fjs{$|FdoShs2QfPtDg4S@gNq
zz^39=DFu}s5y@q9QUVu{4+e%IRl*uCPEq1lF0Lwv-iF%GxtjK(xFB3O(|Hm!+~Q+8
zZ)!YM%><u8<L7ZGO+L`c(F;U#mK|ZZ1ov)nxX+D_%MELb!~n*=JAfeHKKj?u>RSW8
z0Y(IJCwwJVFn`%;5C5G=1A5Z_&=djyyp(J0r_e}5n1m<@_q9v2N=D|GJ?fBbI#C)B
z!Vm>e9N(>cThI4w;!G?%vNorRHrOdp9qtDV{XW+ZlFDaS@dx74a0))laOdwB#I6^}
ze|u3;1a`~Kzs`<U%&3GM3q7`GrIVv(*t5W3K7}P<Yq0I?>;xB!WWC8p->r)y66TEe
zy!BGb6VL_rbsW#_iuINS?e<GaJ=-S@kMIBdxw*eR+5`ZI^?<7rDS_h6!bAt|Istfe
zz!(A_W_Nfs)M%;E)04B-sNUK1UkbuOtHsZAXQ`>)l=0;-Fk{vN<!3+=obuck6BF|U
z&@3te@qg_SX$9G;n`oEesgQNo^{2gL-YY%Jr)$iG6!nNDOyBV_3QJ_E0nt(RFWh^c
zKk%GgAeEIn0wchdY&HXi4kbK9`+V=W)$M@d!v5kftxJ9dl9X=A^A#W;tcf0S$G?au
z#3G>Jkp$`vKl!&FN=c%_W1~uI7M{nAo_p`vL6+r;@hsR73D`1-4y+LHrEo%8V*dEE
zJ^DTgeaJgFF43&T^3);gPe(s<XX|^e<$%*S2e-FW?;(|6iRVl&fw=|wvV^2mkk+Z6
zTJqIvvOz13>hdwD43FWlY+-Y=S&t!zfMLBe7Z7mA1HxcMlz4dX-6cSo7qT8oW9DwX
z`tD5lUN4a|2wKjAC3Qi-qyS<Pd5Uu1u!z6TUTTPsv{+2KAAA1V3u9A28C6Mk<tT_{
zfC6OV1xbTIRwyf};h%(5G5C))!)3$4$On6DoY(~snE#z?CXb;PAfX=k5hKCQDE~$K
zTMo0103<G@^h)wK?%D+q5;igXvJ)j!XX=@9VI&vabLJfN2&oc(9>xXt;k-K%%tH;!
zVczN|x2rNdJqDv!(*UJM<Z&m_;t9D7c%+rwC;O53nffo}$$+4LS4uk9*ZG_e7tu^U
ziWDI{7*5(^(#()6F*;%hM4IWP^WsavM#IMUJK)483|<N#7PKtCUACRkZ=rav(A6*>
z!=5g`%^6%oenM6`RBN}-o41w&Ebpx@p>_X{AOc=Sn9ev~^qtGM`&8-xU^bKeO{WZZ
zUH=`9ZZwr(Tyc4;p>j^Lv3mp-%&g#*r2<edRh`9yYazWSHRKh&yu5m$#r$nDPjPtv
z0HdKcB00*`GeKM*-jP3y+S+r<)rExTK%3X;&N0o8Z7}KHR--bnf7ze=%cRS+JZW=I
z5tJ3uKgCZJ_yDUU-ae+G>!p`zIU84gXf!(hB6W=)psxH*L+iVmINhoIs2p>S7#38O
zmpO<~SPJs_K`*`x%+@>*0eY)$;Had+S$2?a9MA`L`cNThAzH*kiz$c2W{l>PMwvC~
z+godoUjWI>Sal)=X8xO#95(MM0e)0)_kBi7Eii^*_k@tfz4F^(Pw+J04nI4P>7(6L
z{p-I)^pb)!6{JrmBmbI-Waim-$rQYHMF<&h38LEr%t#MIiEHpOnFAt1rSVk~oHz1x
z47Mvjf#%EJ20?uB5ZhK56DaOw2ZHOD|3*o7=B8|?d#D@IsKPuL4>bz4zA&o{riJ?s
z;XcpEVTKJ#yy}_k7u~2RZm$E8{G*z4YgNq@-T6kVXojh6`cSX%%((^5r;TXg<ye>2
zYq>yPbvt_ELZ7xmf~m@QH3<>mt9FW9>KgLzrG^bBGrtnn#Fs>ieaR{SMi}}rQ_k0<
z*I#aS+j{^GRS`=)<)cto7iN&<HLDbAofb7p=}o<{nDmPnse!3rIay|61mGnLOJfym
zp9QSM_a}02P<Gy;nMDDaG!vkJTt<Ti8H&Tv21YmFXG;!_;<cd_R%MYEV-~(0?uFGu
z<z8I<32ZfE0!6G})z4X?{tE4U7cVn^in1QbIv~3rgD^L?IQo|oPsa91!aok#9E*8l
zJ+7B}#0VF96`ha&RscaGM<WI5UqBIkz*%j1DAi0QKGH%5>N(nLj`G}!Sl@}l!EcY>
zZzk$B7NX_aTb(g^$Yoh||IonA5rN=JE_N0fk2s9$EDqky?bFgM06?Can;-mF^4tsc
z*gC)tw_I%t-rev9h6eid%!d#npY$sHkOJ+uSbWeME1_yf2e12+>EyMaVJ>c)EBgQI
zB6siyiGuZQexP+;NWL!y9k;@tfXJvIo$z<TS=Ij<rvo=b&y^{ViXf&#f0WVbMbTJS
zMA{v>@2yGjLpN9|?>j7Ri^u3VAuQdqqxHWzp(TzQt(<DB<UJJTxuKPW_qqje&oPg!
z!3e70uq&b$aI!lIX+{tUT(A~G@WP-<wr3(=C3Q{jxg;M%gakx_MdAv~b%;n;WXt%v
zR4kk@LG6`LS%MY|`t5&VHZ^!_wop_~JwkibOYn#j2eqFMrBVG#^w5;3_m;AIi0-gh
zuiR804RP8My>V-MM4D0|1}rjrtpDvnz15$N55(mRb0he)KCe);9EllExD2=PTebns
zuzKdMB!*<vEKT?$^EC`<dy{7arcbn^9E@lgNN2c#Gz2(&Wf{dXaf^~4FzvTpvP9{=
ziAKk>Qf5Gw1rFa!++6Fwi!_e_1-HfURe6jlict*BX5%yZ5Q(3~W&m?oZvI~RLXhgX
z;b$}5{bmr5e+&CbqX^Ka5hODZd`ZEL!Kg9=#|QxCCs)~yF(71PEa?K?1?xg=-()E@
zK`->cRK0eDuWZS3zWhYnh{uTqDP%a^hVG2q9bo%Bdg46?-AXg_-6V6C{wz6KKST1v
zep{L0>T-8#ps9P3@@na3wnyn$9`%glJ2CPA{Mk{=kl7apBn7g#qiHK&sC6jd-hcXX
zucUgT>8N!=^Biz2by^NbzQ(fBO4jgv3&u-tFeGK_<|)lq;nRT|SHC3eBVEBHfnU8$
z$n#lPff_wysO7IV8M-r?$c~Y?Q4k-YL;?*7UvV#YCgi<GWF|kEEjX@FK4l=gV*KZO
zFcNWn3!n>MD=!WY;v4+RGz`ZA+b%Y^4eF{Obue86D{64KUy6JeKL(f#MFpU4yn{n5
z*c2l}1AAK}8OH{oK%Iz4liZT*(t%4?^XIiLE*&cT3hWd3X&*M(e8B{AgCMmmvvKdE
ztBU$4tutPWSx5@cLA&pB3|T^sZbg$%W<$RNkwTF~x2Wmh;X4C0p!h=6d+|eKx5w7_
zr6N3@&H@E`@^{r?HoFgNQ-Kt88@RgH;3gY2<ps$nKMrYPRGKctS@q2TMp$r~wKj*D
zhRM<i1YB<>p8)8DUd88`8%2?L$QEBUU-wPnDrT2#%lwpon$utfki{!H%C341N;8`<
z7*8EMUB}))C?)GF7tTG4T2_fl>Kx+Marl36bRCXR{(t<rbM}^*b!H+F+3W0G6jBIT
zWfR%%Y>I3lTf@!@*+j-yR@s|`jI%fQ`}q9>H=pP8yx*_MzrN|I@+tPIWQi~J_D(xC
zb4+G<qFDi{9osRctN_*%?4dKeM_JSlKC9{{&~pikV9uvlu9aR(Dz(Nd@Ag`M5cqm)
z^!FlPR>Xz~2G18mbw<L^mHhTVTcSQT9Q<^WGVv$ocKn52LYcRdhZls})-~Z&-zU&-
zi7(0d4U+|)9KISkt8MLO@6#qaTKDP{^?N%kS{cxX$#10mdM6NjzwoJkyF80`_aKkX
zCc+dCaWneSf}Lmor3pOhfkSn$$ZW*(ugrOh1h;~eYF>|l(fm_@qwMF@63haX2?A`i
zIYq(z&@1UW%}-Fq$9(jA2eOevEnEI#02EbZTrkjekCdKjELW2~nrIR0Yc1#lXZ6|<
z_E{ro5J9kLbYm|1&8|FsY|iI(vPwGE{mj;*l(aOan)%pl9LPK{5DZZmX=<u<G?yQC
zPGEeG=+W%BqkrCPSZrjE7gGdH!o8nzTvze40Hc;8bzKR3{)uJB2_RZV>672{L;Vjj
z3Xg7OawI<hB5GKq{sQ~&q8ME)t|%oSxc9OxGTzK}zhEPT(b+$BOnrj+G1KoeAdh~E
z^+5+I7(s9ho)C++m!9(#@Lfo9h?4V4-9yUE0TqjisEqZ0DgK?`SFH<FAVWHXG-0_b
zQEw>`hRnj3(4irq3IAA@-TTY@6GciY9XiPFU#dc261h|ZcgPd4Nk@GfYB88d%;xI)
z?L=pANV%+_g<HK4Jrk1K&VPpn!5q*3V$z)L#5ZNFtvp|`D^Asuq&(X9ny`6SxH=Fs
zPm!lCiKmtTq2?fX-0D&O-c!7_j;~=>+}(UdUfa#J(plgfWk;zbW|B!V;9s@j*0;b0
z+;jM$V7dO^Y(MVfTLmSH4EH5LRusgTj+AwCF)rhZi58Qb=wmQm#}bBSqat#GZ8(X%
zg{fiG+wYP)m48GbiSKaA6RdBqo}c7AXj#efyL;_xUQ6XA`4gzTBdk*H0hSICNoNUT
z5io=qc;zSr)k3e-Ei$$NIWnRNciczx*s11l3)@G<Ya8sruSG0swf+Qzu({)tLN!GM
z_#AIn(62YZ;;8TWL@V|>`B%wh@w15ghv0`zkU#T!T|mWc`8!1L*(z41>%V6ejaXib
z9ND){o&|7=p@)1h!na|dmsroz{BheHF%e=&Jf5glBCdzWD!EsaVDN4G0fFZCi#LTZ
zgQHnAF5q#)tJ0&Bd$ybw1I<8AwXOo)y;YZucLJZcKPa0pK206NiQ{s6=%XNFl+OO;
z9*2%77n;(8k?}Xn-9ELP*EXXqIFw2`DBLRkJ=Fge+i5Z=v^f|;OcKaT1xYBjN*!Q^
z3c=2)Q{^MkHg_pQMcI(YpzX>@?XL&J^kW2zFStHEr<@PdB(Ekek6s0~`C+<FQ>@@q
zbI4dEzn5XPuNsX;N5I^tC0w4rN#;`k2CiP*HdeHzg|%oj`KcIlvTy}a(!~-AL-|dv
zBypEuzu;56T<&km+Vb_XhI@D)_fRo*mT4EW9No&cjNOzoCoYcLR{={SQ>{*av4}=$
z&TpT_Cg2)>b8MS<99_Ho!o=x$4TURqK-Fb0X!|JXKvHk5@4klV<1<P)XVhqBT+)3s
zg!OHIF@h0{Itfli-_b*chy^2%{B%(H5TMBE*vD^EqWb8jX3-*S@cs>YrhkB%$bI7|
z{M`j78&9F(S7+wC_E8JIVVu;GJs_n;5H#`dzTC$&>+-g_KxZm2%Bu6DZusHn+!6tg
zV3nNbc5Sr_pi|ojzwwtp-{a!IEBOkqV%e7NfVsk?PkA-KINeG147oMpAcKre7scl4
z!Ut0&=W-o}!nJMWO^W-MiFX{_D=$ESs@YY6_(fpk`Jxlk{c=LKzdh85972@ltR`Y(
zAld<w=3A|#7Ry#@a-r07zBz_~wbF+D;^WImCPzrDLXoX15OzI|L-`MgE*@Ie9jp!A
zo#iE&C*3|-$=1Gi?_R3|h95dCx;s#o>u2=p?+U?mxmFyc;;}yW^t)0++DP)mi?#HN
z&vmx~f=;$(8YC>%1gJw4C}fGy&P@kKazyb=kAV<bp2!ImPCtJ|iAe=|<BeytvV*a|
zfa~-EtaHDOG<@KpO|>}>Vt*H~HQO~9N4csako1k9fc{Da`%hq=mPm^x3bykypWLDt
z3VM=Tn02j2Jx1-pR26S3Es76b!nl*Ywqs-6BlhD+A(ff^BhLE2%S<^Z#$)91cf+5z
z%QaKTy4ids_p<{IA2TTgHkRdgX2`?uRt8u{8htZ=`oQ|j*;Q<#`Ab0I&_vw?<KMiA
z`7sgAI5-qb30jqrO!&c)qyUhwvb+C?&LHRESKkM<4{Azg8h(xAea2!b0bi+q4VOmq
zJNIjPkLB0*-`Fmc4=H!p`-Bsg6}>JqQdiW!u(#K^BI>-IgIeVoYyyq@+7&52daT<s
zmh}*xVJ!&|a<EsGfaDBco~^>M8AiK~v)ovcgE%p*M?f#K9?7RtPDOO%yK#9Hi7?1I
zWs-Ew;U%<T#-m#=5hn2V-u%1!<KIbwjZUy6O*cI%Nhss-xm5}JzVrAj*X7Hj^7cDq
zjuMZ11Nu=fr`@`Gz@}N$HdnZP|J59xfnxa_)TTg_NLkY!OoHN0!OMkHIR_|;e8u0s
zY%N~{ja^C1E{*y<29;}YJ-W^xsuG6!4tV}*An%qKWnw{tDR_t5>l@CNPD{4IR9t=y
zZbj*M0nSqd1chU@hExVFGuv>W$?oC7OCWAYD4Sb*B^2;M?fb?`NCc!|Tqkg!77ABq
z2odiE#I?{>?}t%HLJ(rykBt)*hXHl$-*)RDVPCTwq9(f}g?QkZx7*e<R)A|@-Z2Nr
zUUJZ<ddcdFTZlKxdcnFLJK5wS$5Q%>;pwc*8ymSlbaqgjRD}-D{wr}^54jm*Hf56a
z#cs~rK+qAFEnd5P1J->BdLErKY1=wbB8G7Ji~WZq-!-5+7DaXnsN&wQcVSZw{>Be{
zvcQF3bsqZ@t$clQ0_}y{0s^IBxy7hJ<H>o<u(<DZtn6+iLAEuU*K0az-XD}hBt&+H
zF2wlMY-W&ryYMP-=<B)L;D<y##Yg$k%D|`N9<JBA2PQg`U2;Q1w{Z;iC%9GV(nd7>
z>ffh85NnHjj>jzmbG5xp-9&G<5gN|)&$BaQNkeT5x)GC|Tl7SicVQ&Bv(Q+OK;7bD
zKxLKEllLx2Oj>6c{<U!9>vNY$=s=|SxsuvLBn^fy8Mjw`J~yIzGB?{%SD4#+?+@X`
z$skB@T%H`T-j0<+g+{%Ic3l<LfH=o0+tefBmR!;HPekG%M|ICCc6>u)W4zh|Q{}fM
zzm!ZW3!`!1GxOnAlJ#(ex7RI82%7{2F10;Oy%)kn{EuA^BJ99R1`~930s+eCRLaf!
zdTlF_9oDcZSmCVAy(<Xmy>tD|0#z%T&rTRtnOCdn77~F(huGVWqcIX>#3+wo99!Z=
z{)UHs)fSsXYwOElaL7ma17(>Og2wZ3?oi0`*&cv!s#e&STKmMFF5`D2Ql_ZlxnR#>
zIV6~(C!C3po1^9hILFMSD&Uz#qlvaZ*d!yIz`gxE35Of1d}koBj(+3BhxZQmPRM+V
zu9kaO5x>P}eC*k$v^IsBVu%Wx+vQk$JN)!1$MN@ptjLWj{GSBV@Cp^aAsGBh!||xW
zn52(s`ffx0MCb6MS1T{CV-U#TMTKccv6urac-*IKf4sfE(Amx2i{o?R$vYlk(<?7i
z-Np{);)S3yAP4&h3f7{0|J)jQxh-S9LyzM>+EP2Rj_ZlAK^N}DgBleNK*ZaV{P|Ha
z%xj@O=JcZfua>Xc-feDJWc0z8$LGHq-F06$)Zp-G^d{cF>U>RCzsXM><5FTw{22K+
zW%B_$0q&UE;pxktV)hiYeHSlq0l-o4LHph9#vFae3d=``j<3t5(;cUij@6K`_~zuF
zA}zuD%*bq&Bm`2V7(HZFU;aX(>O_bO`KwxDH-Sg{570K;A4ev{$8$(byM|=A{b2Gq
zdh@FS)t6!QIF2s!o190*NI7s{HKywS07Y}lAoy=c{9c|f!a{Kn7@O8AMNGJ_05HaK
z8BZ`uQ{Te9Z^l&|JRz*U^i07G4YW_rsnFpv{kFdG--A<IFi8Tl<|t);OSINKMMcHJ
zdolO0zPA|a!RYsur1ePdH2M91w;Ob(B%(%*stw6A(ljzGs9EaB=Q-5=I+n%vi-R#I
zv==9+bJAk7oL=@H4t)1ysh5AbHxv<%q>fXRiL+LnZKB?i-mXK5v5Yfdj)uR*3^spx
z<;OZYCJ*ZD-GRwZ<oMBG72!-L*!FH;)t~dSL>c2713YQ{HEq9b#?aX~0TM>kaF%^j
z(6h|QTw9k<>f5&_nu>E5N@QOkmjhtqtR9r@>vojv=DoQ+9Hu*nurPEpGD`8%q=r+d
z$><9*irPkB6^=Qc7;O6yz<4JmfK*M3Z!_kXEV&Z~h(aa^TA4CGJ3;N_!#>2O({a)-
z)A?5<Hm@g1Is+W6euiJa7k9YQYPw7J$TYV}gR^48E>`*<ACD)aNhcru`GlL01gKcT
zQ3xenBwDh`vAMPr-{`*bF5*rG&@?~nI(TU%Wq>WfA(N!_m5Rg>bN{NF1B08QFfr=5
z1HGN*oNj0)ZCJcAd4>c3V!Y4ZUmoim%r>b!PF&fVN+ZKY(OV%GzqL^+SdPM6`_$rO
zolMw&&MKtZEZ|_INbD5kFw<<Ughr6?77AZb6=;#pNB?--bjPdK`oo7pLm3W7Sg3zc
zUuM`)!l-Q@LLy~qq=l5>wujgc8W%2f_hk-BO|5%AyDkV{Xv3{F35pK1Sl=L%2Jf(v
z7D_-!C#B=g-7ES+lh>%gN&@OAw7D>qOmq`p0&fx{`o00|we9edtc}O{mQi9!LFE9!
zn1NFHmeCDo7uCP+?UBUtFZvN3s)o`VXQ;IL2qxa#?)rXVdjY#@x0eETMH**`*1tdB
z+W`BgG&$At#kQ=UlEtJa8;MY86;2_?*&+Mc!!-%PluHXZGG0dm)1N6*>=r#0Xj(^U
z&YT+T6Hl9zO~kGJ0~@>10baP}2RN0!IK5XXgpX&#dT1jtGT&+@Uh02|`H&|1{(ik4
z9@X1RbYhA9aIXCf<^oL`i4%s!Y1sS#OXh!%?J_@dBWcod<T^^A?zxq&Bbc^L<16fJ
zsXS>n+4f4oy`1|o3`8WOzZaJW#(((mRV8KqLCi&%O2aPj-~-^J9gNkD>AfbH_Y3kP
zmu8%HdH!xbP+&7~mj}Mkh%J%8bF5qVDGZcFq1$W~b22pitA)Y`^J;{cs(Rw9d)9Id
z$7~~5g066FcfzrYr=(iye#mkK_3oZJw@m-Jpwm+JET+pXL!9+Zt88Ox9&7{0S&Q}h
zXlbS00Gr&rD{^Idd2K~8rz<zwO^l`2u*g7vD1FmXxnVF*l{C|`>lvK{45RTj$7RlI
z|2>1y(~>1SU{Pr(F>5d)hPEW89grDB)xNIR+oSw`Cfq}ySGkcn)uaiaS1T`EvyLzT
zD|qJ9?m0Ix5L5^`{_zK_J<Z>Xc0Y&}dJ^khrY(am0ZJd{{*4j8TJ4X~cSSk7{jv~9
z6-I5^wXId$=j66+w}+g{s1tT3<wjB0_ZM8>+j4`ym&i_NzaSL4BqO{{@e~ojhu-`G
zki611_8@tKc?<yzgO(0`(4C~$(@Od{>_WaIjXDbTj4;qHjQEI2tFK$wQaZ%$2*9`_
zIp$es<cjP^^iPpl!p7};Aq;;H6(0l;2MG~VEI-EI0W#+^^%|!s`Z2pWIRV#{yrQ5Y
zb|MtAiN1f8?QdUJ%?lnEUmDEr0d7e9`f47|rNVxMUk*ic^Q<d<?q$E!jSm|w)j2J|
z9ZkNSD%u5$E+lT;_CvSL`bgGG4QAk2D<!DN8cSLvDQwS88DDg~`U>Xu;#%n_)FNRZ
zLQQ8CSzytB4n*RH36ImMM2zPHze{1>;qx1$X1FEXkJHyC$>5dGH6q+VzY&bRk?E1p
zcra>0yvMdUp&CXol3Y2Ev#6l($ywkETDo-!?vPnrM7{yn!+ZZ$pIw@>T%cD%*vCdF
z@4Ey5C^)^dj{LHjuuM+F&Fm6LZ3L^@Ij9{Hb{vGyQ5X83)sU%|xh(?`Ey^~qFTkm$
z3A832TdZ34)zX6lUlkNsDVA_-HAftiT6HV(Cx5P8c^sN?HZxqgw*8k|<^}|c-aDs2
z4|(7RC*B+>e%EYK9796ajS>Bi{FADUKpas5awCyfSBsIW5+Dof6C?_l$}G*#>?o_3
zuUz_PxYmBd0Fp^`fouANo72bV8Ik=3-7KRm_U>K^(2Nj0Qvl{cWs=kGKSS!Bilq58
z(R;`xcw+6dqG5_TjmW6;0J(Eko%i!jR}!sNK01o~EWN8cmDj1};3CtpR6891u>+ko
zfoi121rt*BU*Mis>{4@Os0H9&i#S$COW!{c%;yhFflV5X`uCK2j}UZa$@%?TNf)uv
zQfiDviWH1v3YOB7O7W=3FE@)ufhNXFgZv+)0V&<YEfF;2*<f={ajWLax2%)D?NCJ1
z<(jTxg9HRu^BZGxVl8mMq5Wd?D-bl%C@MUNa48spoimAafH1%Zy!SHruGEHy511vK
zb*82yf1Y+$#o{%PQk`$|rR~YZjy5qwh6aB?t%lnXq`wDOtNrpO!0v`>V6Q}vUq8AC
zxUD@s<P4>SWs3T)?3m)W9yv=CFG>eNN%)R<IufN*&Lhm-49r)fJW<7YY+uX$pFgtR
z3vah5`!MI_sjOY(t6#Y=-!aLS9uqyL(>lM(+t<mIHuNje+jwU*8=IWwKbK9|qHYwf
z>tOD`6E+RHmM0&2*3VN{V7X#3Fo)@d?-9jy11GbclL0IKx3_O`SSTkE>Gpw!qYsG!
zNIc-Idm!@mqYmSL2N$OO*_176GNd^Ahni84?7LKi()x!IyP-0Bg~nGeeSDW|uP$T_
zjb(ISqIiM9*ShQTN-&EV9F2{-L@zFa#XWLPnV^-AjZLFO_7cpbo<U@RSOdK?Tj>ga
zvkV0;`r@2NZ(jFjdd(Jc#SJRQAv&7vS?CG?0c@HD7Sc^tLN2x#I!=;#wvHo85pt1j
zKsec>)D5Afz}1L+e$37zbT;8N*cH`ZQ&*DsR2I(((s_|1$IwmB$T*ynnScQzf6bA7
zoqD_ksO}5Q;`~eTO%5q-t)Nq!w}!sjy`;;VNjs@duTl*3f2h~QX7LBh_ok$dcu9Mn
z5L|eZb3>nK%H8P%W3JTmUA24^(KAP6bagUYb;_NYhXL(f-u09h@v!RA?dS#kJ1n%9
zR9~Knvn?7{6YJ5HZ%Ur`Aqfju8ZS@RwV*{avPgSQelAAMaJ|~QB5xIXjxr|HM_+K5
zJ6&e32WdjVYd8ib!a{@ZZ10o#ZhlL{5n9C_1u{2#@g;C*RieZ3h!87)=N)e56tO}C
z&c%OJshIM7-96aL2&7yz*;m?^E__Y+Goj=B-!ZC&1=EtOyEF^(9`aVtUFIjg|M!|%
z5y}|@)(k+zKu-u#=iTf+GEZTJu5g%)idPRld&HhCiZ-~E@@gSc&k`Yo+jqJ%T<sm_
z2hLqMDsD%d0`(8)IH`R=447QMB2@_jWc{}#QEPKvOqzM88&jxQG7`8V6~}%u_!I(?
zIRogW7>VBM7r(Uje_b1C8_W`YUK<p>gda<)Q;A$<ck7^j!qjTaF^T_bT6}hZ_dJ7&
zc4@`^*V}dDuUWr;mfULCeQDUC%ZY-|7s2&F;LLC^A4{%*<A1oQ`IeYg9GIMdKdpy4
zaQZShFpss;CLZ!<0ra))BTy^al47a}=a)bJ&P`y0w%fk{=i~i*_sMQBVU*Yg_@5#-
zpNH(P4k>4wO;<Gsn?^k}v4{j9ie`@1r<3tM<GSxX2YW4nVoR809hrBu7zFBr4y<Q3
zSXM0Id8(n<JOZ;>n=AX4AHyAqHxGV~uhUrO{cCrs+_mA3hx_1EbwJZ85)ul81<^5_
zQ}3gGxPIYP-lkL`*bu`^aqz>vg9_>&e`cm^9f0X`@w?f``Q2khVXwz5|7tPX&RvkU
z@Z!Rp=_s%xa-R6!vAU&;H%}J1ip#&7Ohn1ljBe2FyUDM0=vn`&gY2IpBViLD=gBVV
z7|(Zh3O{E(h&PdabML#4*>=X0O8acHYGQvm$;ISV<I-d~2fWN>RLi8U;jhl=Y(%Re
z9@LcgRP4IUtsw3kX}S_x&SVu5f<x5JXA>KkSCCo)oA*3SOj>KSNJAAb@LItSB%h>!
zfyy)a265`3Z;7nfTk9)TH{9$J`$oj2G5%Csv)8$ypc>6w+}uDX0)szZa6o4#&d5xt
z%Dw3)NQ@yl14UX7E(r1TJgU&LE>u=|Etz~}??Yu!j+?ecZSuHs{H-TV%^QM!D<?Ye
z`F?nvk#7oONiUgi8Rt!LY!tmBlRSSbB77Yxw=vo`Huqk@|Ec6tr{cmwW`iF88@`uT
zbrhwbZ^ymG!Fd<uux$Frzz83bhPd5;S17@BX)C?=I@o$1OOW|qiiuW!NO(NA_O}(K
zz_^S~3XOVAeBK|0U>O!(56YH4tTZ{2{~XRS9U&PUZ&u%~b9XnuqF`EZjYNI9(}#V6
ztzJR7SYuW>^z4Woj?xJ<rrD~g6fl4heTLvV(1z$<Uc~hwjl`0FHYxj?9GLgiRfXGF
z-aHp$1ajQ!G07{2dBe~5SPh9nGJ;auvb?YM(_-k~bL+T;ZHMwaR5ZHz;&94-==B;z
zg2#2EHI(TWp!bS5-|f=eo-2eo{`t;71I&n5Cg>#8;@%QdS3sy4`*D3;#D<0$r+1(R
z%Rg5Xq~}>`Qd?HMQkpdceb6V;fKwF^U=sAl><<pSPGGk%rREa8twv%4-;vTm>_UjZ
zI=W{8LbvoQU_Wgy`lBh*IDd4e2%3D#lZ%diJY?%b>2`*lmyZw(W-#}^Gxvcg1$W1p
z6h5DU;1dc`p=uj0zQSBj4>yasp`*B}dW;{Dz3+ZD4(Mpt<Pa;0(~r2}LxlaN+R(Hv
z{H|ys^uj+@1L~Ex7J2w<Iqr$B87UNhGUXT=f!eRUe9J1+tAIPX?<w=WcebyuPXs3@
zvd62Nz6wM4@Vt9o9=P>wrJ7$a%i@{n!-52A+N+QbPhx92WMeC^hEb?%oi!#O$-^JZ
zu6;=S1D+Y9ozg_SHfhq9LIhhJxAK{!JqpUch0FW_bm@P(8j<s}GE{dyZx#LiiOEt3
zCn`^QvAK1JO~?cBfuI%rqGt&rS}8!$w;n2XwZ(2)sd%vY`^y4t$1(uTlYk@I)QcW@
zEOS<6(?pd4f@QwcR2;Kh)vC2BcR0~OQ&gMANb1X3p}#E^m?GJ>(My@6<2!9ZF+t@`
zYNOb7uLW#ai~2ILz>5LNB{I-D@aPXa?pXaBmV;BWwea`PZw`rdYRN^KcFOQWHyVJW
z#8DuBcG^@(3{8G@Xu3dG;<^dYAXtQ)nfJczPUt?6Ja>GO_^i}W@=mTp@%rfzRtR7l
z50otbYszJz=pVY<_esVKFKU|_O2S#{vc6S)v`twD9#P3e<9^+b*05L`>@@-U(7y<k
z#Yu-RJj8LUWdt?FV>3E%<T}WTc>9MD-l_?(q6g~G!vjSxuhbmwC~DM{HZLc``A}jL
z5O@B`^4l76{P}if0#9aKU&t@A)@jPfi!B4l<d!exgH+$p7owtwz8q6>p8?e5DBTHP
zyKJMSjVA4iKJnxdU!pu<YbQP8c!F0aBvdyK=Y{KYhIr4#`Qul;3HgYo!C+{SBdRNN
zv!RmFaR)r@2{Fd#d7-BL2j9S5(?se$Y#RNdUaDi~TO4)!eYPeuW>=*P`3ga5Im>jV
zFa|28X9rF`8r*QPz7V~7uY@<d&=JQLhfc1A1Y(P$`Dh46gwnedP`%lAp$TaT(JinA
z6SNkeQ@YmXz~{Lk>&u0KOW1}~SC&`wi!i^ltWpAxwPDpx3nH4fK8CXzm@thbh%Lf;
zU&lsmQ=uZo^v#IHj=A)fE>s(COn>zCqx1v`KhDi3wr_*g;vqy8Mlcym2#zH}d<5r}
zc=R;6s4gAeYRlh85JI$V8LP9_0<zWl#~pzGrsEM-sV&y6T~Ac<%q6aLaz~g`mYd4T
zJhg=~#b5$W`LjFu(bF}-GxdKyVFf^#NV&|+5a!qZ80Px^53vJSQ9l-{TGCCF`r?U2
z^VP)NVotqK|Iw9T{^xJAewiF-l;-q}$+X5Z|L=Nh=;;AaEq`Vbpw9Ax2AG}*w9Vp)
zkLSB1&;-XS<I!>tj;Q!G2I2Eu#XudTiLubI+}pUGRf;L;jfYqUFND}&8)n5(b*!*Z
z8nL93P}i>Qg;lR1RR&GqzT!&%DmB#L<s!J8^4T<IgS|h5JA47AAzs;}hs~#8l}=hc
zQ?ECJJ>>9yxP1|OtDLG^m$bF$l2XhTl{`EwSJMi}L%eZVTxJSKlbi!)rdw4wg2YRh
zQv@|p12A!YISrss6a=Cg{xQJopzx8iUgcfGiY@|9xuAFZ;&kngMs=hEu;_U<#zo#5
z3O?-hg(V%7OtwgaKoYB+r@HQPq`^cq7(5kDz~aF0F*VZ>%ET9F&hw83jE5NjBYJ`#
z$qPe$<%nx4sB?`_Va6!6ZbODO7QRjWXGHx0-QAnNHMp@{XY%=BhyfDLU2D?yoSAAb
z{e{ki+K`x@DfuM=4Mij|DvJ3-Yk51c#7z?L30zJ!vmr3tP&^(H#1Bd8d}l=k%*<pE
zl2AC`qT$EZKS7ijlbl}$+Df1qNKvZz+88BbD#n5Dq3U%}#*dGvy0hPVepNXMS|JOS
ziIY%iUcsg_v7(+@a_pfOZbRyST@V2$KD`v{sipcBZ=89d!FK>I)xn3->NfMkMP>GC
zf6#sOJp;CsFCNif8ew=x>bDRN=jQ(~x1O8D4qxi3MnP_Z%1&Q%SzJt{;~TS#NeC?V
zxMg|fpYWMdvbv%0Aelpc!=k!Wi0Tlh37tQJT1iZ-$M!2bqoD})O4+3BjgSEm7w@1f
z!6ImZjQenGm346t^O`a3x|-BbRoC$>P(}SQ+<xa}=z0r3aQbt1UWEpsf)K^MxJSmE
z`OB~fS*qJCAB=d|1youmx5Blc`5E@DGorC;$;77Gw>|}N@yQ3vE_N=XhKiva)j#5o
z3mj+~_YKPMd12ei)SKQKum2;XL|~M@bR%v@|4AO(97tPOya1M<lrncS5M)N6giz6O
zpAY*kJDt)`=XorHM+cWXyasZJ7^KrMb4x3o2jy>H6U8A-sLCt_*ix1(ScYlJz1i%P
zNpG@Pp%1U;Ov?tGdYGLetT;1ZbzZKzZcdphVQmW12*~<ra4Mp}v19i639z0#$f6Nq
z3%nf|MMjm{d(9(5kyOC?%JOP9N0NOx3~#E^(%T~ZZ{v34t3eOjjqcZ1>cW{U3G}7^
zn3czsx%|qccJ>!=@?~!a6dCm6xm!=SBA&<*jfq_V6}b(VZg+ujIH?}nHmQAW2o~JJ
z{7cYpy{7bmdSzelQDYwb9#Cc_ZBY9NV$2{pO^Lq5xVSL$T(%GQ?y!x1eddSkJ?mGV
zG|!<_pRcW{>dxX?dtyw#pjd+g)-3xJd#wJ7cn9i-l<UlQ()AccqOyi9X*z4Q1|Ex+
zdp-sS-uLv&x>ooG#6SriPT&ek_26-PG4|_rDkpc*HO@9-m9&C*C`kzx%mESUx$<#w
za=|<NUHG(9G9nH^@6R9bdc(@?r(e|{q}sH2A($oGRAr8rd>^-4#YAolY^p6_1(HH@
zHi4ihkw-hF8!V7F!Fui<R*gV?8l-Wry8K!VQEPA;(ZjBq?*7zFIX`rYG`qZ3eC=8f
zL1S&(y-4Hht>@WR7#`<vG6tsia<s0$+(<3{^o}DTB2=#J->U`3Ik2itAw5Fe_aD_A
zHHSl%Y<9U-YA&^_RJgmxQ*^b5D-?$>Cr9b!DQmXP&=&{0K;L5&Sx4T+{*WVgTHPDR
z@tgEOUWMlmn0t2#(9Z)IkF5wLL)_>aMxK3~RQVE#z#v+?5PYIyGoI$ui;1ppenckp
zU4mNOQNk&wn==GASWZ_*@~}>#HP7J_ykyTgXW-=uLch=dPp9!m{>}*f+?YLg+akD~
z^|A0I$q>OOS|!iXkVjpOK6}AMPb9l20XkrqNw%P2?#N5xec#n+{?;pHot?Q4ZY1EX
zYY8y<ZRrqx{Tfx0>#rSD<a9{P`WfIf3&<VI4WYi>5LtuYxiOyQ<+N|BX{HIPfi$&u
zDORB@!h&1nc?tq`{0d-*|Gs`5340E;u><3~d#D!8Cj}*@F|%g>Z$@$(o;H7l?yM0+
zR<%3H(9=P;5uC8?UvP~(h*mRv2$3VpDb{W#6mYmPk?vD3RhKy-{9`umLcPIY(<9pf
zK*|P@G?l+V6iA)QU-lk*yh+%gQO+{eTF&#i76ABa1Cy&#pu)DCa@TwcG`_)pMEyfm
zYWT(-1!BzET);;O!N?#N8VonF^qClR^R@^l3r3cflX*G^hB>`XHs@5X*u$_CI5f$&
ziaq-Q=PJpG@mOPjxpJi+FYAr*P}^!e{L*q+fA6}=UOFg{@6{6i=<g>d3=^^c&9})w
zSRcDFr|2r_wy2WHIiKp{h=+|g`P^MiYXA*ztL?Re%i9C*H}m)9Hu#SoWsb^;yS+6)
zCi|%`L#iq?!09peKakUBSVy;6@{@pgzxa*_Av;Ec7y|O{l2^kYH1&&BCgw?%!q>^z
zqcsTC{Vtw-wWk|Sw>`UPo0<n)Fs0<~3mgkq94Y;+`4GMx4g{GF^q`RrLl#SbIZ+@+
z5yPoKozGi(h(eJD(@n_y91=#8i!!fhymy)JLX&=#_%RyC^vD3PKmVl+x1e#6G*qU9
zEWU%%hz1@ZQ;a;`$Ajm&z$6Xy%X4v9F4NCf_`;M{KU3ftn-qR5GHyV2FWTrN^`c9D
z5LbI0gTfjTlS52kzuM8YGJ#%j-cE;*PyysR<KUqIHIBH{ukrru-td0E*V2}~aN!N;
z%4Goalni2-9{&Rh1KOPgN#9?tk{-o@6i0r+F<!16ZwM;dUxufU5JBQ%1nbHcIKRkK
z3wvt;t>!mzuIxeMSMq8t+Q{QVsE)Nq4IepE3<Cx<C0BSgsSO`Inhn^Mc^{d1CLh{t
z9)FY@=Qs|LX5}(@&TOD^80!aH#sM%z`U8^g_5EPakHvgfA(mncL-u$WQ%?Zdn|9#m
z^<2&NVz3(DXE+Kv?rHS_xg!?AW}fD|(3UEuFyY0MCV%s=dzf5l*rF>0o7bh4qh=Jg
z>t{I`Yq{iE>WUphVRT)n94z}tnM_oPYc|GI!hMG|Z>TOi1VcF^`thv<ve5+L()sMS
zV5`&Ls;<<gPke|v<GL@IsTiv9a(24{5Wjd=nFfbZ*tws9KAbiDP+K(2of%1tF=xl@
z8yq1Z2^Px8RZ_f#?!&_=mtrFFaG86`pI(z(*0IkU#tQlp@eY;u#^7`D$HhH=LFF1#
zQGZ;J|26^qH2<!z?AmM3(-T{9RWoEbyOUk+;L8#^XNUbxnt#<yWCA(Dj$dZCX9(-x
zNr(3B7rtXQ<_=p0wBA^@h{JyYgp~E`mx!}3c_B#;{pm`x&%iO8f;(%$35=a>HpUK$
zR66(z?8<X0eW6$+D_?O#-9YUrM8oNQ668S~fCIw9OYL)A-3~CxgycPA^DS;FSjC&N
z#Tgrg%v0`VYuMoq?ZTw)-K{16p>)vYQ(F4#S;5%kk^G{2_A!3{qD7JAi!3t;0R|`%
zls(nch0!NTk>mj=)hF?s$jrdgHT{c8FlH0K=ya0Q&k}1|{x%DID-Su?8|m8>>llxh
z=7bR3tzxh^+8<z@z>`oRd>LIgBd^N68nRR|6Sh8g#THKe%UhEtf>UjTH@e~p#By{*
z*NoG2EL|~`QExO{Ot_!1BOzy+O3p4mq#}HNj&j`QM{9zmyB+H*@rzLtmt<uMsg8AZ
zs?uAGZ3?t@jjr&^^IY%!h!DhG=V{|DAgs7Dy^Xa-Xxy8>puC%qxwyi_sR3!ta;x5A
zrrCm1jgo=md&Iq8YQ;cf3O2QalvNRDi7CWGuZ~yb2eQ*09K=d45`6<{Buyg>379{K
zqKY;H>xZF0jExy21=GGJ+j&q`zqEE)%cU>fM^zaRC&97J{%YK^9{HKGO!12HzMm~m
zcPfvZ$8EZxABOOy3K@35`IMe%F>Z{4jAIbU6&H*Hh5+RE2+A>}F|}e>U=v75`p>P4
z^$~|6w&U7;1Neg>ETKx|m1)-0y@!WK?L6|2&+Yh&1};eTZ6I__5=8$Ii5yd2zpIhu
zy-3We{Kjamr5gMBX_K!zFva`+jNrUz8vjBH<NU<7UBQW0b`fa*yWs!+_NxR)ZdeA9
z``2SZ?+FPIA;s;v9YHSWzZmb+r4<)+A>2F981=?ltlfw8s<(nk*4tHwSfnDt5;)qm
zA~%k-Xiz|Zr09fCRzs%E9k{KIp12*?_Nm2qRcs1N9*!D!m9S{VM9#eHA0TFE7WmRA
zl1Cb=#M+~x3VP_-;sMvw6e=DvV83XGYcqtvM&LfnBx^t;vn|v0Q!Sb?S4i@rGhuQ%
zFjjAI=}cLyH4wc}uaN2d<|<Z+N1E<qfU7cGnA&yB%&SLqKsNs&iMZspkIh;(66R_{
zyC9CDief(Ap;oE|kcz$pRo)!yn|#~DvrQ0%V1EtSR~1~a>;Gj`V2)t$iiT$k*w+`6
zFXGz_F`<de%q&$?^u}%n$f)y3@u#K^ICA&<fLA<(G~fw0R;L`{q;~ysN2Y>aCi(A;
z$$jqu;nr2whG2-5jmY%S`GtQ_Wx&*<+NHoNQ{Lf4#i7J#)eG$>9ZT0h#v<WGBzc0U
zUH|@NX5ute)D8<sMq4gAXCQdFFzmBN!{JaKM1i6wPsTCXW#n!J9%7e*HYRe&cZ0Pc
zQ}`TETc#<X1EY|Su1C2MC>QT5fG&2M)xli;!0HK4S?j_6Bl<<ubNcNn46QDpE}B8k
z-WM-_VMJ8qPkLxelkw#@xwSWQc9XDp=r}11IH_;({4h6~->{LP^m8bDbBj9PX*FNQ
z@6FD>{1kEqq!xbd2~%$ct!G!cVR7alMI8(e5C1ma6*&hdo&Bh+>pV(oIq6|zeqe<u
zjM#v@l@z-JvI_G5Y=uzD$6W=#0<*C-I3i|`CN&Aht=_UDj|pG8(2Hy*LdQ2?s*0?)
zRW_22{c*Dr`7bZ8+q)613=V{6Zh??steMa{83D~_?bjOvkVpsvT%mWafB?uY*{@h#
zmu0+&8@_+)*gzCq3J70zC=cybpe~DW(ZddZeA7Jo#a!tB(|oqP8K_|41RjVKkh^wD
zetenbDba!X!{K|>*g(>HX&p<B2M(%JTv!d_-=ba{tf%>*HN0VOVEXpc-J>HSN2*Xu
zZn8IiaSP9)ujp5Iz!3?+T+4pz(tfS0nhY}Ms#6|-%tB@|{yIA4WG=r~?ev?c^w8Dg
zQS@5^BD%5p*w?Kf2-kWjPgRTJ)z<XQEA#qHgBPKVD6@vZAFMo&KA!@NC@GF^#}{w|
z^@{N1@C?qIEa}`(8=<$IxX}}fqYF{y4A-BTLT7g3Z#s2kpAsE+?*ejZTlrONsT-fm
zqrwLtm4Y){vPFlFV)vd*p%9B86|u5;;Q7o{^BB+dv5pf@H=%0Bi3jFL4WiSQi^KO<
z59KLd5LEfOh8S>ZpRAELZm1uEe3x(EzRe1H9NP5ImAuGQ`A@fXXDra7oMucAl1MM1
ztpQA5uaY9wbxb9NKEdfQxpYQLZ4k>C4doqq&NwcOrLYoKIbY+ZXKUnvv8?;YO+Wu_
zqTkGMP1Oqd-@c*9Corfgy_#h7I8v>XL+|i$e4=25t69boY>R2y+*^@~BsXW~ZbWPB
zC@~z@9B`IIh9R0CJC0<ybOf;(r;`;{5eWp1HI5T`EnRaVAZ_Sz0_F>KB;T=v3-wui
zT3Vsd(TabruooAMuZ{eTtVH-!LSmIi#s;jVT4y%xD;JTK8!22W)P}EZ5kH^RdqTN1
z<`KXun$l#;Dt_+oU<H5GGWiQDwx3+63JgkRuDT9Rg|N@bXlPa6(0o?-%;pPu8F6ra
z<{V*sm2~)X|CW!`+;jHC8ob>Fi1FdozH_IzE_LnegHIriaS*w84oreYGN>>Zy99`c
z3x%nCPoVVQ9}6Z5AJ+`N4?}1}yQFU#?c|A}1t<6)LPdzGFF~JZ)y0|wZ5tvIEH2`a
zG^cOW$s)(HPqjJ}qK@gZe7zol8x17mVU#u@w!yD^&A3zrBV*Tpa0zt!|M#hcfQB91
z<%Q=*;WF(?sga3nhNeSMVrq01enX!0e8_?Z?-ya6+EDXtXHh5g(fdp?cjoAADOBJF
zDFPP6K7vPuSIw5Adkp4a8<0bO>OjS#%^9E><f6*;SO5yzVo1XM9LXn&L*!Vy6lkR7
z`oS@dJWOIx|Lje;T(C^2r#Hv^5|z4ItPcAG%TvidystVLAKFV7=Z>sy^)I)8QFr(E
z!}~qJ9nT4Xq=U-~a6WTn`|n0g`(c}gznHfa5XzXt83c`mf<RJdH>=k)$CXf-K<v$z
z=2I<vA4V&cu!6i=asm_0n?STslZgu4;op?bHN!9z5fqT+C_!fMvYExsH#Q)^@+wJ;
z2hqcLEZC;DEwm|p>PI<k<f{oKM8f#Egl@Tt@deLEKz0JqWrs`@Q4-4XZU_VIKBzEQ
zSsK`B7bE6rox-~lPed0u+LL(Wzbh)04w}<KP<W&c&$Gc5p`h0u<$&5&e?w+01hcNi
zT^_Y8gScfGi$018EU=J&K`?6v@cS;iN>1JSxW;;Z!8V2%NW_ZbKP&gyvn}Q3I2JN*
zHVp(G{nFb&IDfyGsE44MG~PN=JCmXRgmx6%Yg#Gc9n*{_i$<RRxS^i*nhG)k7<sQR
zn$q-zqC7R8f6lJBR;}@1$b9Ts=j%rQF9TPT|9aR$6?@NYEp`QX-G%0ij6Q@2mVF0`
zn6lBveyynERHn8r_N8>-e|@S{t8pH0iT=~7NhZ7y%#gy`wm?-1o3=w50c$Y&V3PNt
zYU^u^*$nz`{U0WHT<`IHxE>ohWNcSi4LOG41!ys!Z{}`-DzFBzn`N-xs_s7j&G3LK
zOOu<Dz{WrF9yuH=lRAeqQ7Ol`vnJ9q#;Y*Hog?EpISyZfS@QEPh<Ecw2hKgMIk<tE
zVCI#(k>MKrPc6YMKU?93HhPkQn+u$jV+>F5K<t#!1G5`4CrhLF!uPS?Xr*C^(8d^q
zVoYCu0i*p^0AbvCr-DMW4uHy-9|#EqGpX=cCTk>^`Ivv8W3#rv2$EM8FsKzOIzBOg
zFCt>w$GG|JJNe*|UpVPJ*vxMMdvPI6u!cY7s^yX&Cg-R5_7Q&XRJY@hJEzuItFc)p
zfd8*}37w*v4?Iu}=i^g3lj%`aQsg5OC>m!<xM78tVPZUA50BG9wAwm!_v$@WIqdea
zCN9u<rJS%K7wUKfGDf2`MlXP<kN*WTh>IXV7aMy?7YLfx>6y@91y&KvY9QhyX1$d?
zDV?#b27DXCY<U}dL=S0=FGDW~dT7sK29ewDcm_U4X+|#mF;^GJF+>}sj4(d5bfVjC
zJHCJ8^(lMcu4H%o4DXcUwLZ^w#t+Okm8Ec$D0ByT`fZ`EO#Il2vQ?Q<jvIG2hijZA
zK9>0$lKJZk^u|5Qe^-;dap7FPJjewI2>Bnp+!7%2C~w%@<P~tF!|$*MQ81b(C6@n)
zRx0Z5!r(9sOL7b%=MQ5}0|`cMaGpLz#}-vXg27`Ul=7WcS>vYOyUN$4{5O}U5PR!(
zM#R$ex0g1+JYaKO{Ny$6D)9$YD;SRRQ!(1$6{48ib=T(L^!<YwR5l8U$e4n@AeDNa
zM^U>Oxi$o7>gScy)ukmwabO2;g|X(HBmXLMWNC>@p!{5|=$V>KCEr(&%(E9j7OX;K
zZ{e&u{UPU9lV&!!&LB_!0?Y(OpUrgA%;sGFNQitu-6x8i)2`|e6Pagzav)K*C@r%z
zf$9=Als&cn{y7E`iBn#JsiAybL~u7?+e6SFIywzu1XLo*Bx;%z5$~2$t0Ra(;~|sE
zIEucrG?=*Cd?vJ@_+!M!z^B~a=v}^dkxGvmWqV74@F34uk2SjX!J_hE$wsIBqRONO
zBXIzsDV*aRuSYv&V~$wX-1!XG2@~o}9(<2L5^F<Yin6kLS@oMTJd163$!DOIVWdlz
ziIPpCjX{yj%CXHz(}LL_SV9Z!n`OCMZTH&RPe&tY{7ajmIf>wFef}u>X@k2NV{5=h
z&jBOaCR_AX8jlkDu~5X=$-Ln%%rG@eCvfCw*z(Rm9Ox7&T=A0T4Q4H}draZk;!f!x
z;W4YNB?%SY_ot}1Y}XzJ-@4_+vcdQMy9crMvs-QMewmps7jp|%<4c#;35Lm&4n21+
zl$2e8CZT8mU2e!3&7I{jthvmqkbtoIwlEy6!P1aeUkHuUwZ5z=NddtG32F~C;qxZ2
za*43_`6P?i>fCsio>(v!P?@r!&)uBxxg7{-h~#|t<3RFb(bEoHp{b`P(Vk6*lMe1b
z#l{Lci(i`iBkyD<YFGp~mFJ0A<Ah0-c#wl*9+#OHY<FJ20!&{o$i6qV+df>M_o?5i
zU*E5vz(z=JUt(-_fsbaviH=8rSe=Gmhv$`Y|LqaqA(B)?PADWiBdg+YNcT?rm+3SM
zI8g#7^;@k49!d+k+-VNSiz<b1Z?wwl7D=Q40c19J<8boXZyGBf^x4$h|5HjRy+4;$
ziXKY~hCBTw`^Ek~h;XVQqJT`0Ee<9lzlwE`ZefpPv|M<DWte!u7k+mf1ZBPWyhA`z
zZb@@SBRrY5ah)X+%I$JDU$vR{;w;?_>$-gZF4vf)vMVXP-(}LU@|TDKt?}f~XY5Mj
z&LB|nVO5Fs#>2^MpNUEv>BCpR;9PFEz9L^?a_!DrhG^se2#okn7IwqepjG%ruT323
zoE&NDS;Wk3B2eU_6QFd`9Dk8tSp}4-7h)<gl8%brndaAy<bBWiRIPqvfmdjBwk1)x
z?oMSJIwF&(+x{w^XM~!Huvwxm7D-EPWSkahHvl464eYM1PTx7B`Pwey^h+zP$>*zq
z=D>apeyCAv286<&|1v#OS&V#SZ)=@gPdPYiEg5<TogOU+p-UkHm|jh%I}O_`;WMDj
zbN^4Uc9q1GZ)0Y^HdD?s@&rlIu|Vm@Vf!an5JgO@0$q|(SwB;`Ru;I-v{Z!^kiT)4
zt^4kdn8{cBk4rJ#b=yrLeHho)G6vDhSJT3&vrQK+iqFBHfBPx{PoOE3J9mMyjCj^c
zQe4R@zRyYG%n^lH4Szmav{fI_uwP=g3hwlEy3D_8uRqkSCIgO6qHr4eub#g9bY7B>
z-6yq{ducbNvIsi%2DaAqwiq2IxxiYc$`Ok39I}_}js778EhW;wHwGWis{0Wdbvlds
z)GGE62z&EPyM8CHKM}glOuTf!^j!`w?*e#Yx6?$%5O!afK^UE^F>b4GAU?b1OX9}w
z{T$xr(CZj$M!1H4M1k1Zt0YEdVr#{<A6sHfmDgkegO?ZTV8ZzuFG;rC^~+Thqu-x<
z&#3ymGs$~vgs_d5qf>JPqpst}r)Sgs*#f`6tTifJ?!=;3Dc~%tWS1YUfU9^HL^$!x
zRAvuj_)9i%?>q$$V<G3nD|M=xeR={s7-4~qBAPH%Z1!*R^o**YCPt##-UMJTyaB9W
z&~)#Z(mRy}IIH~r1e4w-x;gdrag*ry>FHU7jBwK7H|2};{^kSl4NI~J$N&DAH<$hv
zo;wUCO7<Ev>L_Sh)ER*hA4glMLEnfX5jnrFqjH%j(`o|(P+tq=KFlK`*=T3B58vG!
zA`JIE2T$|G=N314T!`T8Ks0mJw9@q0#m9Gc#{)a@3SWM4MO}fp^ezEz`0R4<B^kT?
zv_nD1ywFoh46u<M#d_lGbujL3x>Cd__&k;>>=fPhmV8n`NfoJ=M#NJ9sQumCeW~*D
zIW}JLLHxnKPW0gvkOUuGsGl&G!u&S#=n$EScthvD$5M(J6<Nm>U2G5aIYQ8Dq1OfC
z6$@_M&+2=#h)1L@2y+!U<OPpo&)2=el+)RMUfnf~85n73#~VoF$k`_d_#Rhwj88h%
z&hUbK9u8u@kF1#ah&~EnyGT4P`x|48JUpO?^q8+Tr%Xif*>LT34fM^MA98*K(|bw)
z`euNchaMs)(gZy6P46Qj1teO*{{4W+m}J%uE4PUysKS$H+6-1qdJ42VYal~>_F*%x
z%mPV!KPr0mrWW><iDg|eyxDx<wtD+tX}-&uG8f&^9q}E<6Fhwq+h^QqWz=dGZsrwm
zBnuwdI|`_nlhmx_)>k@+3S(_T$`WnGG)lnrE39|wr~R}wH;m}iwdMP%iK&pISI7xv
z1{8pmswGfr&Qm;FA!<xrYgx^E^Wov=dquYh^+O7ytF)HRq?9hwkzHVzdk6l}Cs!i(
zpQlp)80z*ZMQ@~Dv|hy1-MwjuM@4o8G{ZMRg)yo~yTOR44!Oz+Bbxq>FlhmiAx+5;
z+a=-K?`_JkYw+^#VX>DTQP*%SO?!}IvGf$KB2%+Fm<nzwQ4$zUse!aE$HOP7?o$e?
z%-wjLGbo~U^To|QbFCN64<?l+^m=*wxUDHGk7obtcdaSKov^M&GFzIc*I55|xTiPr
zR#5xVQI_BALF2%BXh0-XEYe;$WV!Db*d@&~sqi<|DkDGW8P{6@3Z{0M{W&hFK0hVN
z;(r(XFIDnzf3GK6u7VRsY9-^P4gkDo?B2-VA8Xn5BfwWviLW~px%{L>f$@Trcxv#a
zAj7=ASMZYQ#@UU@N+-RfH`gtaqKZ^)o0UmI3Ac;QFM2aS1BeQ_#CH6H)8QYnJL7ZS
zMX-wa*a9H#-axjvuuABlaY66#>Dc$1!1d$a&G8AK6#au<Yr~!KM)_kAy0iys1}xvi
z0Ve6bWWz*o7vShCt*YdDs_RoSOp-p<48HrT3%2ZWmY5B`@9bG(%AN5EgIDRgs#e34
za86R@`ga}CjF3B21jXYBm@mj4OU<!Ne|nf_YoNFMDTV?f)h|e=jSQEwmUz%^m*w+m
zxYf{u=r~=f@Fm-a{*{4jEP+t&W-QCCn|{%vB7M!#gOxVg3~N@$@g%%5swZJzhPr3e
z(}#|%SiLqUN8}1dxOV;vIkLVhJQhoTA4^rfvp5(ZzAP8V^JDxB>t_96?ta0fzU?Oh
zC4EVM65gWzSMg2HHa||OPv_JIA5XrwnS$QaX&^KxP;~YfYe?<Mw{AU)!SVg1XIj~l
z+<CwfC^PfZ4=f)ps;{o`^A7CDW%9~i1@s+>ittkj^bo|A5vt5f!WiAZ=<27vwkS$B
z|94iA<ugwPa(AAcpZsH`m>m$P*a<957*B4>*7I4vbrhbR$Cp|yI9|Bk$RFkP1Nyb7
zTU=|6S%wUHX`G(^?ua&T`kge{gDD->!M_`I4;sK52bCvZ<&JbTo*o3`Tk%n3?}0cR
zJziwaeFd$eC6)xYBL4Lxt#i^l?S)u|rrRo~Ym=q^MUySzoR`~v!A}%^uv4fh3-Q=G
zWas<wKBquMAKkTGOkcL0_~iMD0`<@aik^f7nmIP@UpN!0_8Hd>-<mSM{sA9#SNq?u
z&$13QJQDk4!E@W6?q&X~hg)R7k)f`6aheq1QEG5lpwpo6aOFELp#Er&U{1lj-_-U6
znq~27Khd8iotRvie;U>e`DuD_&!!zbDEN4j5NUR!mEP~Ab5t^+PvtDETO}@!a|0Ze
z&J?~@T0WIKL%V-}b-Kevu<Kt|_2=dW?j|IRhZ0|#mP6I5SJVE|TZqd0y;a=ce0~zt
zm##_J+D`7Uk@eJ5{1F8vuH<^ys_Gdt6CngQ@-*(hJb7g5rkZebZg66FmKbvlT8o<X
zGMA@bfBbxPNN7F7`OTDb!%xj~@%`OwoPXRy5LvD^c3o?&n)i$)gHhU8gm?oX1zA3U
z#7~6;r)bY5yz@B@3W^v@BU^#qv}*&{K0e*~h_bBy#kvjjrVW~Q|NQ8$yT50T4Z<T#
z2MgRp)WuzoC3dl0k5AJczjU3mwRvJ}PklYoy=g3rNpmXSxI=J;u}{WnOh%PR?}fP6
z{{R71sHq?@gYdt~HrBOrpbi}g#9bV{p98@Zwg9I{Ag8}xQiB5sh}{+2FA9KjHOq5>
z*OL?LMgQsEiFyF1fJzH;+mvJvzR0OkpjwrJ@Pr$!b9+9}#o_m~m)zUe!isY%4L^G-
z`LV+3ZFQ>L;keabz>|1BL@-bN<-5Capxa;zn(X?-#!Rh#yk(s#7_A*RtWZC_sO0@u
z7>8SJ?Bzaa8KQv<aqsY`lKAayWnjEzk8Q=1oGPq>NPnHiO!vhd`|I*Q#EMIgrH{RE
zbpJM_n*if`^wj5LqOxHGW{~RN#dAJkSbVxd@V&xcjLaduJQKR?)8%6)#6=^pA9Djq
z8ePlJa8F&1b~IlpxL6(?xZA+~Kp!0N9WB2?pUTXf=^c&DKE7kq)#Rv3C%(xul7HY7
zN>pxe<au2yeIEp+iC#riF&l(zM`u8}&7)^_Khu{ZU6N{LyFT72v-x+B1U2I6WfB)%
zHnt~{6ynbIWaqY)9&QBoH{Yg4N<pw2A6tE5-dioVy-A>N{$~r#Qyt%i%bHX|F#$^%
zz{J5tc8B8YhYuh0RdtoVyIRNZcgA0IDO@h_%81(r3y>ts+jqgbKZFUL%JzQfD;)>X
zWx)~kvHXw!gwBwV8o=Z5;$Y%-u2o9#{e;#}*Ncp%k-QwPisHEyqb>Q9JJB7$?eRbA
z-ZCo6FMR)<VMu9Eq*F>nx|N|z8fhu%Zs`yZX^?J^kdTm;#-h8sySrh|=KDKu{^#8}
z>wnH#$2a(_VVEcOJkQ?ueO;fcoK9hgYmV46|I@Yo?H3DQfVSxy(=@Pc0m`5}sCDlg
zs|ml)sN(g4n}Vq)+s)B%CPd2=m3j#&;z!Va-!h~V%Kpoh#roEkv8_-Yv+nrWh~!Zb
zost47xql%{q%q=)YN|8RV3Qx1rS+O7>8;BlNvIBE;vLt_4;<zJy39Y>n;dr%b$grM
zxR=b%*M$^F)ULPN@#39P!TZsYp%<B!pCSKp{ZMe6AOW9|{`C$nqcC^nbwA}dAXy57
z;?f*~D&T!^8gwQ2^BdH*i!NTX{F+UX6r(@E_4AhNdE$9tij{;x-ded2aeK~hvpm*G
zEOn9(J_vej=?BoeD80g^E{ukA936uMLZztcvh9OzLauj-I<glSi4bBcq`Vz!H0mFO
z*+r(&itnmMYoqfHkPoE_LHeQMUbhMVSa+Q9?am=7PXdu$2t8CH&|7v9;IL%vnrtea
zkb|aqZ--Q$yh|Xn`}yl34p6uB&0OmOssMVXqvE0v#}ucFuN-%{10V|Wm#I%vvs6&X
zv*>Mg&XU4jpi$>a&iwX_Py=Dk;d{hggO4e9-<EGsk=7gUHp3PX!Ed#{MM)TkRztpw
zt840+Qu<yRaRIy9DjuorPchqF?3Pz34ie?f=Q;q5dhu)`hf_`0TmZvN_0uqD9=Dq4
z8gA-)fc05242PfE0)D>d`zLHe=w;HSpA&>L6=e|lQ$rtF?+zwo{u=c1lOnetu*(Ys
zyb%ltL+A};zK%O?y*GD2`YHM>NX>)SUxy`aAJmo)%nI!%f6$~jjR_DsZ2?u`1t{pu
z_Z=E0M~C=+>+bRq&OY)YhYQ{8R*l47-OhJsh6#tTReU4k+cU4ns=)E5M$0DkgtVVU
z3iEjdY1?hch%;fx(qw#?wp<6TY`G{Ig^cv|j&GHlM*o}$sble*F|=0sAv{Apb}yzN
z!9@=}#IgZTAgRzbg{`8aPwCj-`|1N2h^{mH?EzSJN!~HoXmkU~YW>(%UgiUfg<s0g
zrM!em1}X0zklj|ZZJ^EfG55hb=Y9xNE`=S2KJtP$;ERNho9$urMf;~?B;6PH0%|=X
z|9`+Tp450sl-Dvkwe`Z6vKnh1xNq#(rCe5@4}9g7ViuKXJFSuI?*s9VeW19t>RF~j
z@DAicl0#{}lQ51CT^_CUfp=_>_dn46Rs5(YcX9|eggvE6lV9U$v@WpZ{sL6FhgeOI
z#gAz33f`jsnntsNY4mbf?<w=~zZ3h%G`;>?_+p3q?J~vGC$()N<&OfTBQ17DPf+vD
zNH$?Ru3%l&#ZiZ6XxK^{><Auz4Uwy$qwU1}V>9P`fzI~?z(5*CYsn#m5(GNH$;*N^
z==+HdH7TG{zIsJ$L$&j7{VMb-USkvtm^n$reP%p^Wa=H)7ZRFI0Z>b2X+mT6Q>9Gi
zcgz7@wMeylZu|glzZ($g(-2`%_^N?y!!GM$wU;KUV=SkzbMv`+1cM;TRs%Hb9jamA
z9;(P2K_Y(L)Kj7`bx;v2dP7*5Y<aL^zKP;K*49l_!-%wuj=r>Qy)#kn7BvDE^|P~%
z!R5Q)w%c{J0=`>#e{e0<#1lzje{CL(C$6g{N7TH^>Hao!)4xz>p5dBk{nhtk&aDqn
zJ>(r7E9h`2L4?sd=zh781Szky{wuEIt@t?8_~u)~DWX;m(E}~i<`W_0`coU*YtTCF
zVp*f7_2~H*-k^%Gjcp98>ef6(==YL+l&v)|qFdewSygp;pL_#GQ#fL(qIyaD9YCMh
z(H)y6v~%0YcG~P2Zld8*?vCC%Kr~9UDiW3^?koa?Ab7W-;|LZ{9D^xFjEJ8~B#+vC
zx+!JWxli?3-w^wGBJ4>JU!75BA=G+o0+6u-RHo`V$WC8Jj~PFj|9PyK8eV*?+0yl>
zpX6^EhR9d!TO4x8GU3m+JH5FI%J>0g;si8Z8ipar3f(V9$<^Mx48Ve4kbDs~_8K~3
zbWZYwV={d2_8D@1lZYecVmj6A@)r)jI_S2a=fLDEsaS1Y?`Otag(Ppg!QzF8w(P%o
zTQ<}!{3D<KP+*eET+K@{XE9t;F5zXqM6f9Gg9Br1W>#R~kR1q?*VZu>(bL_RL(3!r
z0~21-(Q!wGPz55oS;~_i<|rTpsq@f$2zp4c^|iY&!4)a*`;}VdYroZ<^+5WOz7Ksm
z-IqPkPJJT7vBb!irKgtBS@nKN0?}MmvKlGl)%A|btzx(ah4pV)q9C!qjoTwQ9>OXs
zb4ID~iH|0jxW(y4m2#<?sghpmSv@R1PnJGYr2_F@@<ZP1p*e{Q7k+P0cdjOy@EKyJ
z2xhbA7q&H7qCWUcQxr#J5h*-{mp6kO%wDd#LMHlp^alDTG1$ipYjV;2(G*_mcJ}`u
zf|p`f;zt8$T4yNVmnWM5g340=){H&qdaipzlp%yy*u*RJk|v|;(7g3t>2P@jQUAQ=
z%4Sz6ydas;42O3wt=0ujxC$|u!eDY!Z1XGqEStKu_NV^?Pr;y*f6+RYZ3AkEVxc+a
z6bCB9aL(WQwWBP=ZHFl`p5yk|_X&;R{O(7i_MfArEiH@a&asKv+wgA+&u$XAFs@c#
zUJtz8=ZottiN0N^ia_+VH|@FTq~`Qwe}fA^>Yb2kt0)=f*RSXn*x4vPB<lL}Lrst>
zdF%zJF$Qa_H5u&!ky1kZ8Fseq?%#<Mtv+Chg)&feVu-aK!qn&s61zwW4Wy_*;=)t}
zxxg@quZ~lJQkI?S%_#k999f%D_dkmh&bkF6hMBvU6nC|BP~Mu~w(oDj<z!|*&r3NJ
z1U*-`lSwFJirt<dqo4F={~7<vQSqKh`37_cNYE_dZxH!sIq$xLeL;N_fY${I<Xh;0
zfK%IDA``2HBYIEl(=O@tCwhpnk7{p8$rb6-tiKnkXqinO!q^ut?yf8}t)G+&SLgk?
zT2(Nqoe_zj0Hh`rBm1ApXtCt_=~o1=o6DBv%b7jAkuuBhQg>K_)FsBal?;);unuJe
zxjJh0IXUzmTRR5r*|!HI@q^&|B^^wF@C(DhQ<*-dPLj?&o@1W7&3b)6MFs;Z>LC^A
z``(}JXeaZx3w$*?2<iTGED?*lh5S9^h5c~i%feI7FIoBT-3BVYCnYyBp7Bw!Wh|Y)
zyO(M=z!9O~hxU=roc4!iCHp!{rCMLSnkAnwQmWeE-Bakv*-h8U@k2Gk>tlLf<RcAH
z1{UH7m5)h|U~VVSA_7=46taY4?S<okm9QzkOBLe*&FzbxL{?|cW>C@pFtGtD8BS47
z@oaQ|?6;flsYctSwcaikl(oMTs9|!NDPfthF)L-^&3c(;N|||{pk3pUyESM-R)_n2
zoeeF!zX6twsKCR|{5e*OVQ1dD&}^!EayKrNOgXc+-O{e6;aE<jy|kuLwEZEC=Mc_T
zNDEuc9RP0bxh^CYOl3D)V`A4bsO1`Hr4Co5t5^r|Pv%PFK0RXZCP~?_D9WaVJu9H*
z3dCg!bi5Be{ftYsUj{{1oW62Ef9^EzQa*~k*dK%kxx44Hzk7ahXU96zkQG0vvb%X1
zt=&9{r7tkBm*za4l)S94qd-P>R#$^takmCp{iF!NtVl^nw^11#^?^I?aVCB9wEV^2
zBgZPQm`kfGQf%d1((6*!|BGYh;EVE)=Y=lf5(Fo}F5wy^(udNSRRVF!ua4$$hR~fR
zwch@-Omr^wJm1~AgvHx+V#VMnY~XS87G%?YW*S?}_4^fQ8C|{qYxZ;3LyRvUe{h4Z
zWK?Q|@0@`qxqr;HU-f5127O$gsO<-rB0s4g1Jwk(0MU?0yMxDFK~H7kgYg5;JgzgE
z+`oy8e3c7nMoT(7n@N9fW5XA2r4kcERI9)Hw8vZstqnQi>3bUno{gz0a~4>Pxi-|l
z4G$MRD2_qFiEyP8$DuD7Z3t9@H}LTAm+%&`B(7lCuho<jDxwA7zDiCLl^Ag0CubSH
zVZU&!?S!~<8c0eRTzu*ZzlK|r+ag@f{FFAu8xcKEs`|G3(t6vTI1|FzoKUl}yKyhq
zYD_M{Y4wp>oQY&gti;!ep7{Ip(Bcg>H`GK#1UXcP;x~^?teCGL%$N673MnN(+dFp$
zK4kE)U{dc&FhWgX-?BLJW0*oYqmx3*EUHR+f2@T&$LwFKMvstLA8-SCLyv$Wu}{nI
z<%<oxdr8wQZK;J-)|;$d&FND=+H;ydxR+ii-DyJOwV(IKi|BvcT}n9qmWEX*5ozrF
zmVHhgWRvxh)^&p{PRO@?e;$htcOOEiUo&IIwASpr{ewmm`Ir}%d;Njqv4K@>zc8QR
z3UIQUAwlx;x#p>8DKOsw^sUo0ix+IPt;KBuhSe`iytgkTPCNIHU$<v&`hxgHyKVjA
zz?~|Tq7ALifg1=T(H+*fgVxWthQUwRWSsK;t|!3;7vG@t*$iw4V1Msht?9{t;bA6Y
zLQcqQ>&gAE_wWw6A}FGnm`L0U+RQOO%zKz$Jh6kOAhnodA9@Y2bSAxk6nShP6FXwP
z#??8w>`q-<!hWp+j}9CEk+d6`DPHQTW<QLOKQ%6YUu@8D-KzP9M5VGVgn$Jt`)KwB
z#_=04In|pnM8}1|L=XZ#{bGL{(QgMhGGE=I>)^gccYI_4|9iN?q-^|8dr@=(3O{MR
z$P=eL#OkCyB(>?qlWYD`YVn+Y4U=+yQ#JZ?kaRI6Oi3`)&-iaq$4LmOOUs79{I4Z*
zbqp<alijgxoz`ETSX+n-ZKX-XiUymj?a12k$tr}K=*MM;NgK?N@H1N5Dkrb8DU>uy
zJSJ5tS`d}k$aBWyEvlTvA^L4#O~6tI^bnUEf@nx^#8%CK1T!4nXRdBiAVmloT)8!n
zJShr-lK@v6_nXdlpQ;fbnQO=FhuV4U`op&GoFMYXOHl4Qtck7se`Svc>m}2kdq8Xw
z^Snq@#cd;bnZ7Q?KKh<2kvzTKu1PKi4pN?C48dakOYPADcj_rTioxq~e(~TnVsQDY
z_-4YCx>!xQ?v;P2#^Q409a4hVL^ZOU%(tVhxf^M#RAExniE}zn@1hR+KMR*4Q=y5P
zoT@7Hz1^JIhJ~ewr(saALa8{0=%>S9UfLI;jJ&7gtyWHmQ`W^54G9>BH)zXGV_Cf0
zz93^s<n27atE-IN?ndf_TDW*Z<$ut^{QG&V{42OEHLU77k^VdyXb)l)#S60G&yHuZ
zR1dzR(m{NN)zsR=smUc=fibpeM;c#-*OqKV&o&}?cJCJPEuT!p6e5ZzpJ^cya}UA{
zbM>BzTrWOrNy275-}B={I6+8_P$2_#COxQJO(sT%wjyDDbyMg8)T@Coe_<4-2}>9W
zq4SM?1;sg+#o9aR;KYZ|v6zFP>Up<#Yp{<bazgg}PrVB0roo#=r7j*V!!z#;#xhr}
z@5IA41Zb)gHFgZnz(Y2Y7qcy%#V*L@l|K^`%$+lv8eL4U7C|H83iNP79L1;kU6%(^
zi7!;I4RHLM%-tc*fPIq779KctC!cJb9pMfEf8X^ws{xKI63;LaDs(kxmAr>w!@_b0
z<uL3tG{YUEtn%c9k}sdo?bzKi%bx>$v2xOuAA>HCtO+pPMGs|JulX^S=@FG9dSEfI
z!AZ%FI$8r)pF9?xbwi9;Ie<INl`r#Qx4t2wh4%S7WaSp_pDs*U_~`%ejdX>C76129
zk-x*+2{WyGAe--{Lpjx<?X2361Sqm-A<#04cNZdZVKFV<BnYyG%x)NJHU>DuX!wH(
z(DAgP&?S*AyH~^ar*q55gHN(K{i)Gip*peRqlla{#8NE+p1Jk*lnqTp=DgiE(E{xS
z`e}<LS`Z!+IF7U-xF}i)0-)o%EfEDh5{dNN2(YxPxJI1x@z&lcISo5Og7RKC7^xER
z$5XsPg`Agy1F<bYTud6{Is+WjHw8ooEFFpEzG{A!^Ilx7fxc-?8=M3ScZfV%dg>zd
zt*f{PZu->0A><aj@L|`!5sf?)^y)B3?qO1s^N1tf&MA#Qeyb4lg*9l&Z1p^W3DIBs
zRPOr<zH<|36LP`B2j(G2uED~oBN>2y3KW1oqq)+s>F>jvzQvM~WFAqOd+%MFIHGrt
zm_X%o(4G>NNg;1cFl<o;^qY#yBM9#RTbOYQM@#*>b?~nj;)X5FGT$$4(nn!K1;_tT
z?=?ghHM#af{?)-Uv!&$6cZqGD%XkhCrSFL+*@<npljKu)cXU>NgL<)B`%Qh3jw`a{
zNMr8{s_VT+1o@EkcELatqvyla_E%P+ZpEj?^!|%4z3O^89>cd+Sb^lgscgAwJxUX^
z)IA$2re6y**!!YA*2ervMHdQSxM$rjSQcdTg1T^86Q%k$1feWq;;ACL?t9ZNKg_F-
zV7YpgfMQ|xdNpMytS^Y?jY8drjPFgqZU-)=%&ATSsczdIsyba86Y2uguCQ`>(UPys
zIUj*$y0Ny<ZmuB(q{O8F(O5EowxrOA2Pc2Re9GWDOg1gh`LEt)uA#=-dF|ak2szx}
zh+|aZ1z5(E6k#9FjPjFr3qo7{N_zTK{L+PpPrK$W+<n@b?}fDl=W6Z~I=X8H*C6QX
zU!bxHk@_{00+6*(QXO?u$SX$%<PW=(Lof6ixY>9eTI`2%KBW=%53mOV?$FHk#hK5^
zdDh2r^K{4HD`3;I1BD>CbW8ww4N#@2y?bY~4jLm^6M(aA2#WC2)RCIJ{dBmVJ;41T
z?+glqLLG_Dp<HbTa&mIL4<)C0&mC4!1{nD1JfqxiY=w$pPT-*@d;(T13IMd5QBH<v
zi6@`39+pHLrlw9V<T^C`-nT6Md!GcOPOb)fy9OViv{O8&F5epU*4>cROcV?^gDUbb
zA)gGtVH?{*mFlhqb+&a7pX!7x$@}Ps_mvYDu;ZKnj7)MUA>+6g9r4^HyIkx2S$QA8
z2{)bfnesX5@Pry#PCT>iKJx&VM#0k3QuUuoJG=!o<as$zjVDPSRjuV3vL0pM%i~__
zxy`MS_gs5Z=BmIOuHm;!xN|;Dpjt10<_TsgX6wBvW^Mk^bzp<~kX83`pA3<ieiO4v
z^_W3;MmL;s>eSEUJ#HK<Msuh951Nmw<S~}xPz3h+S=mO)h~VCe9VI7B-D7<=8&u&^
z>jA8-4>WhJT;mQ8J`p|yR;2+`%Jf5dzs!ydrGp0uwjVM;Qh{eR1?YmuXZmLR)`2^`
zV;vYUT!TZIA_wBECNH=B6i2&VJwY|67i7_p)(|ejg^<Tt%6|TS2Mh#V91|)YJEN7c
z@Fc(66{<cUd|u&XJsV1ipyXn8_4ZBAdMiy3Bl-JdyN5S`ty{kiUoU^sarscQ!O~PG
zfxxe#YuDpY5j}gBoCpsK>)QGs19yVh9<>}g#;q0J;g%zL^#ZOPKEp~=a<d`C@7y2q
z#EN*9V-|ueFt6the9d3huM#|M&hJOv5Rmu_60yENM24i%2m^K7(<OIK6`z@(_@Y~&
z^6dP9{oZ?>$&w}E?^@&>C|<8Wl-+<Hi;TW%j_fgc0k~OB+%7+Db)Yl7OI^As+W3c<
z5klj~<X16?-=4V&Cw)*fF3br@jJM3<yHY}@p!*!1Ewc+?l3dKwF+|1E5e#d)KFjau
zR!%HLwAPuVdUy@x{k#OA9dgJo<O*o&bYh#V2G92aoCPkS8jHPG+b-{cWlSO&hKA#U
z$~=*lvp_7PE_~PqTL)#c>9Bmgp}@$%&5w~-zHLA_zmovYJy*w@QxD+Uc!sUPP-CZ4
z$d3jn`W(Q%ZT(i!pYI-^92&3wP~9&^JkleaFU%9dy?-@~6fPb*($i3GZ;igl{_qet
z10!TQ>{;sd;?*r>$5k5)*X808NJgzq`z|j;>{lCh%AAoWiUVS`Xv6F0Q90_O5S=B}
z1Z@hBFir2rP?=1`P~SD1=6cwyFaGUIY<c}z;S^j^#ql|7%se)ag;5tb)3jE;fOr-M
za%#`9y3k;6vk<EkO0Izm)ELigK<PCF84uM`BhNSCLO6KLx1hbdC{%PO521-fpEgh5
z{&RKGo^^K8Nf-SyLR)a7qns>m@g^O#e4`9h$*kL{*vN^}Yh4buvf%9mDU9WM`zHO8
z?g`*)-rMPq4vVay3m8Jq4j{_GXls|lZmG>79=<Nf^em=h{baX1)VBg^4t$`O=RBSI
zURcsjLWG=a$lNtjaTE_gNJrMC;nymkHVSA+{c`m9@jz*BTtkm$j6Ne))Zbl)sz0L@
z&J_3><e2D^Amkeq72$WMtHkr~eAf%dWk5nlz+-8jV!zpKdb3FE_o<JWuEE{Mqy%Gd
zw%bgxuG{~BEyZftmS;`0eG3Rbv8+bAP-PHal0W2|ZY`eQ0VV<u{BBEWM`85?pinem
zi74<bSUi6a{JZJkVDqRRXC*|T58o62*d|51T?wK?4UIW<7dD693Ty(&e?0)Il7G%?
zNWl#WsQI#WG7-Y7X@#)7l3`fKx*v|rUoUr-YWG5z-h;dOSUo?ASryN<I3X-ROuZUU
z=_^V(FD>nL?Ur^~XeBY(1<+}%TP<Iuq8xTeoXSx=Zjt1Th~#@zUGuAKe<)WvkOf7e
zoXdpo`qtX&Rt<Yp$@iFW!ZWp88{&tv$okM1EX&j<G>>*ND5HS*Q8EvGQ9peZ>C%i?
z<oSNfmt$fib}NnJPY$6mj%QqlMN}GdD+Fi}I{m2}QhiEK7PAwAUO`X^;o<SmBYwcd
zZpg2uK_g-*@NbU#)m>@piu<88bBvkpzbl%*laZ`jPa&^BEn|1z^1h)M0#CVzpbSCo
zeEp|5&uYA)s=jUb52N+F*<wNIbOWCoUvI+u^=K`w>N0L8bb22i>3unA>*rEghK*2$
zQoY(Shsfgl;67YYd@Q*#I$GWw8{4S7wjq4<A>#6Iuzr2?S5nxy^IKZiAb<~h;Hs|X
zxt|Qhs3+|yj)r$CRon8c-hPsZzW1Rnx+;tCexC1+7L7FlgxpJ)FQ>XG*RH3(RQTt~
z)zZvyc%>M$-wz|oH^H1W=98U20n4M)&gzerjnJIze>jZ(Wc*2Yze8?DPtS7qJ_0k1
zcC1ajQqc^+82sdvWU8Z0WSaU*n^&mpFt^`yqL`#tWIu=h1ezNG^G6li+&J1!I*m17
zmJxj}DbHxY$>e*7R~q-4b|eA2BWmC?M78Xr+NUs68ETbyJQgqCeG7a9m}fDt%s~<@
z%)9*mgg;Hjt}W-yBWtTV+3LCIwyK}T8oB_OQJ?XgXKY7ua~xTf5D;N6SJ_U!{L^vt
z6#n$tPjXWdiAwKJqf)hBK{#!`Z(^BVW2(XZagywptQgl#ZZ<%$<1OJ458r4!7!gpJ
zAN)PE*krnFGm6UOqny~IA%&9>4S@V>&;D$bQ{V_}5X4%6%jN51aYJU0mQwTI{I*k-
z&p!i?-|f!oq};=}POc^&#`jVR)0|pU3R%DaHPM7?OmWl^v;fyne{#w8tVqlOU0qo}
z`TMuG_H5)=-|B68{Rnb1Fw^s#E&hf;|2X7eX6?Axm-Ulr6Pi6^fBll1*6oYmGV(n4
zy^r(u;M2b4Nb3y3C(Fo}5rJ$dtAT{bUYQ>FG|w88FGbbrk?AqvX0M*Dpxkp?C--P+
z(CfYIgyM0^U1*P&&%N?{eZ|F<>yrPbMtx0Bnoc&popZ|ndlf#bPBjzDmk>T*Vja&$
zkhi)h$mOdt&ijRIgMF3DlMYdAD(YANzShI>UdwrQf=vsj<M^5Ab8a%R(Ock#6nrV|
zvjn?534XzIX2(s-$?Q>45_*|c8@&1_pUZ~ois9t6_p9h_l0`#CXM&BYp6QpIRq58m
zDi3VUvUgjQ^@4Wf%27hA<W72Fv!ktR>t~+`S61PXci<js!iG86`vDejGiA7+J_)BY
z#$-u-AyTW>bzdVRnbV=jKuz{X_SSKjC2~^H)`j!!6NTYKG&Y)`CJoB)3lXI{#pyjl
zVW&qH6sfkx(n4&UaEhy7?b@PI?wZyGf=%Q@UkO7Ks=<JATyKcq9ZBq8kCT4+PRAQ4
zr(fpl-?Oie2q&O_UPXXfWr|TPa(lgkrrgm6HYG$muYmzM7DPm|%@ld=3uHk%|83;D
z{1NDtxZKeDuKAnk<Qs})G3cw{Rboi^>pbD%%3{r0Tep6i7aEe&o@~*1F=@^fc6#tD
zs1*g1$mEWTPMAQ)ZQ#if0E<Q3#>fmKuA?xh(oD%Fv<pf*ZJx7n&G6V-W&}K0|0&9r
ztTIkw0e0Tq;cnA;z%5^>wU#M`efIOo(}wlV2t3K()UADjN~19f0mzNsCxR!_IT*h(
zo@BxT4b?0j2|ZQe-cGK*9hy$X-smg;1+r{Y04A?-sI8vgj+03OVUj(p2UmJ)=Jy-g
zwxzz41#XIrYDjy1_M!e4^4HuJYT;2^|B}~>zX^ZX)#5AT)*X^@xt;{kNLqCNs#7Hp
z)$mW_v3E@;h?kiUQ}(`U+iBXO*`M@FO$t>hA6eR)R&w%UEO=9+9{4%wv$idXNI6?>
zoJ2Imjw&08y`AZ2-+L)|k3XL*Pa$q`(6+NqS~$SkP?j_;hM*gy<8DZ#S*e)`>iGT%
zi52bsKoZ_hCIw4MOk6j4#=cVmOrAz_q@fpPw$ZxSSLLCdprMN#mwN??81wczGC3K(
zb$p45M=Gn@vHq?96Ae8<3XiphA}uSgGOd7BcMZq=C>du28mlSI>lxukId4{uQp5>Z
zl`9VT(f3N#*r&Dy(z%w$o^h}fb=paLqnj6&$<cllS&N3kzxEq8`M$F-D`(+&8tpmq
zc%~nN9AYo=9;3spbPZ|t)xCHDtHIk}SEaq7#xhUVLZ(rr;!k4wTz3W-FJ?eYWu<4U
zhxGe`j1Nva<kk3}YZ}|5Evuif=-N-X7}1egv!aQx*OX%WC*2{;uw9XhT<&UzT=KH0
z6Io%)eCYboSE~{XQ(-bW<$?aapoySu9V9XzIE4%{?XaAZy(O-3$@0Z+t!;P4pGb=G
zL-$gv9lq?AGx||V*o5016Kfhw^h=M!CAuD3ou)6uB0iU2{Yl}UC!m+s*_zbgJ<CcZ
zS@K99vEOw_X&FJYr57MmZKuKXH|lXEqE4-E0Q66ak1x|m$ozJ}QcjToQbln`r_xIo
zcZZHuyO$c@D7MxOq;+DhRS@2cUT^=_UXm{#WMX8CiU<`pkbp3e2h>q}x`MkQBv=Mz
zpPMf)=+u_IYQ_JhvhQzZ)yb>D^2eK-=~12bP)pNya;99g(+iaev@xU>*Tba(;Jh+n
z)eY+As3`QM9<_Zn3M1<Yr010qsB868bQu*G%qW+}+2p@GP98<o3){<oFu^wE5{n*3
zFxQJV^wq+o$HO<_qaTguoG4nfglC#;>dE4p|68Yx{*Hi{@Gj_cPsJyGGKEg&$x_kb
zJmwq`QUhp%kr(<+{N`CC1my2u)zi9otVCZ^=CI21V>$BP22V^p=kaZ6I^Jxhy&WbM
z9{%>O)#Wd<!NhsK|DZxc?IpoGvd*xtD$MPHn^~&kpp`Y$qK#QqANJPc1qY_F?UdBX
zn;8tXb(uh;olK4Xth-(Fb$=L6=hZdMNc&k=o^{}vm@E@=h}{=x_I*|)mOotjFmSPP
z@m(+~ZHcZN7hfcTT6LHJJU?q<QHWJv4ZPyAc2_29^STPzW%uPHR5-_b@$M%(KWmr$
z7731#i}dH_7!+!`E>j|iG>my(`G_zAFL)<6e`G`oQ8Zyc*zK{d|9x+<hLV$`_KID2
ziVRl^A%Tcb5*vT3P3x3yx^v(9BM-jOhU)EG3A44uFwO2uf5m<Fom!>DLYLX{%0|Jl
z#3nM<l7BNhFkoCsTY@Wqw<Bg4mBhc^dBbdv>vF=di6=jnDYT;4^}za0)AuBb4q~_C
z`9G7xnnaXr^+#fh4HAz@Ja0r5X%#FhUtlzW{7aD(%@yJWEP+Q)+>O)(_kXVR0y1(!
z)h0HHHs`ai5<&-?FQ$LC+XnP;8Hx}>?d_-bPJbfCghMjLcBZQnhKwJQUdD?CBW1Ay
z19X4?oKzkrRwPXM-cu?^TkIHfFLrj%k!ovOw5T+GMDhL&bW5T_zB>Y?yd*i?`Qs$+
z1j<ns`%${s_rv;b(`dn($qy_~x{AsDZDb)gXA)8|CI6y?_KADvAc#s@&2sG}s%p~i
z_=g$zDD-KY6plVKYjvW$LWV^rS&U6gyt^K6!6}{iN?J#ICF^>hWLzx)0R|Lxnk>zR
zypHd_iAco;Xji!y$nM9ba=()cHB%dlG3r41W*$Gk{rHnmnR3)x$xld3mZqH{$s0rK
zoVa=L;zF^XkSln4`shUgsb-?+Y9|+ds<vefR6m)I!Q$@H_>=f=5k>=~p7Cp=y_4x`
zD<}I;IN?Zjx_#(5TX(g4Kh#b1QmX1f_)it^$HlQ0d>VzP@H{5_!|<;C4F|u2(qz<A
znKTaihZ(#pC>1SNLef}ab(SJt@k*iZHwRNMMv$s#RG0Z;Vj_`$Py88oi2Bicpe**Z
z94riP1OwytKJfRW?Y;LE$1OEN%#qT3%c?1pdK7Q{&l*f2YrP@d2jV)jO&@8<#5D5C
zR1u?%Bs!0+PQACO3eKXrT56w%mKHw|IOf~4IbWstUeEqa79C6IV)SR1*5!FqYDw1S
zLaZLTE#HebA(TUBhryBAZt@Zcnw`%l>5;W!e(<lVo6-)SO`hCxaHg4PSRQGxrUj5g
zz9v>d(0i$QcIX4R>Yvgys$BG-Qu$*eCwjJxuiR4|4@&7ah@3@Q(EdI2_+g1(I&u0l
z_b<F379eY6>*t6xnyen)SI*$G9^yAP2dfW`cUqr%dw3Vu+FAH4K$T3}e=u{;=3685
z3mu97U`Ud+eEoJ(+;v){2G`dgZEsmVoT@Ob1(J0%+Z%j;!3yom;=+R4=KKD&Rj7pA
z^K`Yd|Kbcv-Wq^ST;4Mq@S{0|W_X6&i>f!^-SY_e=A7He&!2Ws2zqc4F{>!oI*iDU
zU4RNKyX$lO*>-~5%qm;c1J;yv-uTKkRY)HFb*&?^m6d!b5{Jy!8rhI{r6o&2w=R2=
z7Xj%>g`AW{vxVbc-^e`PnnTQgyAcOT><N$g?F8YjOmC2k{$O!tlF9e2w2|Tj&r@|3
zoR3TU>XgpSG&d)~a<oHm`jA1_M=_Q6!`sw<ciJmR1zPpV{N+W^@yDrz^4apjgj}~*
z&9vW|?@2Ur0+gclIq=fAU3=UTXVqM=gnj7^6G65#vvhR2byGLoi$nQthvpBr(_aCx
ztx7n|kJN2Wt4N8R#&rI+dPe8izT9<gRTO{S8gl+EC{%c;xL2=w=T})7`5SSjV?{`b
z1@gPirGjQALMQxzleFiohQze?c+b7CeIv9wsY<djk(oR{cj#aY!F1fXZ))~qW=)1E
z_*8`Ou}07t<&zgL(eB+J4c|7i1s=XG*#nkYwxWs5e&U~&uIC5)Rbs+yY{`qBBUPgQ
zR#pp7%>(4yw1xZ@^qPt#)N|A>I^SWDa(DI_cb*Rxgkdpd5I+!&1VB;cPrvs2NZQAN
zsS5-<tK7^ldga!BhWh9=81--6!uyMua8sw07qNIv^ML27A@drKk-Fi-ZQEC%h~iEL
zB3nNf_&eci_&)ie{Uu$&?$78c^O?Ok)34qAujz%zFYfPd8a3MF%fOw(ftu2DlcoTZ
zAQk0kEe*=Dp1RY~yfOa$%Vsnv1g(A&G{o#dNw(x8LRXcxZS~R-PnC=MjbT{A7Uru`
zCL*{p5G<ybU_q<LI?2*XbA6mgkN2hFjLvBZX|d^0U{?MTvhw*vY%g=gpuxq`>4l?9
zm7?pT+x5P6rRT{WXPI~Dc66s5N^@6IZ)UAi_{VMZ#c8JxYNq?iC(Syj%|lk&J(eDX
zYfsC_=#qT)PwTjH_z723GB3$5CY`#%G27{pjVqn!J6mFSv@O*(3*4spIvtO#<9VXR
zsA)AyvI@VJ^roCT)yJr4pQMCvCeTOREU=odvi8h4=F{C!6`j7Wv$T)e`(d6SA>w!I
zz?R5n0617AvG3}(_SNlghw2uuY0#z6q_rAZIw&@@)A^2hG!X?h`Cs9n1>r3!daI#l
zl-v2nP3T#Xbe3Z3h<%96x5|`K5S|@LdWP>BPrgbu#8e2p)9VjqkrnX~$-uSHKHHz_
z9Nu>QGK^+_zx|$^_(V27iT-o+c2G(zPgFePsXN+c*@#ypK6j`QGqIRogh~Ky+?~6y
zt<va#BPHito51-%=_;EZ@zgBx9S2tJHDqJohEB(tNxYm!8*=?eI*L5y3_SEFh_O)b
z*jL+6UW<kJNxpPSH@2(r#(kP^ew6-?VDCpRE7eMb;!R3apKd+RlV$!24cU)$coNM1
zJOkICEwdWzLYheAKFPoI{sLlHx)Pei@_k?1o20uQmXX|)IloJvF^@(OW5z%G)JPuR
z8<W4n;jr)pdI?l>vqu^(<(|^*emMIow0;qO+?a{;AHx@--E{X+*AhQycY9)Z81f2w
z4{F-I3uH!_bjhKYn)@+<zkacB5EsAQD`6E+xdI-&{kJDs2$!O>ZsHD!0GnWt^;9mf
z8Y7ORy?Ate^=AY%8SMqMxY^X{CNik?8G&d1j==hAlounm^rrt~#5}+b@d1!rK8rct
z*b@37Nnu8%k#osufLDvyCIc~RQMN98)C{V#mW{gIzM#Xn&v!#*zP~-p;?XZwHV0VE
zvzN;NdL>A5WZJ{L^cs)q_f(IGGFE#8)<9PvV(XbVm+cKE>qm~Kfmp*;Xm{)>MO`Ly
zYH|<Ez}~ixu#93;Sy@?C-xMs&CMdaduv$f8apMO?vg_EbRvR-Xoo5^7Y9+n978FSX
zp6}iZ)>&kXZ=D>3S3TJ6Q=aO(|I)kZ76}gLp~@I0_T2Tw72Vw4T`Fg&)CZ3s8hb+M
zj!pF4_ql$K0)BB)%zdS3yGI}aj(q@*XaeY$A8EqD$jnloriFMUbTX-NW94w8vI#qm
zbBz$ZR5J?fa<uRN%o|2GI*0R`IUKO0Prh;GJ!YC5ZRUll5Y~)zlcN!`muG%>{djZp
z#(t*SV;nmrRKDd#wBzM^q}>Ke77TF(&a7$DlvbsQY-EOi;ZesFP$C*g!R(p{-3rui
zylmPHrQK&@b2}^8hn^fB{ORX4->>$oNngkts>0{dewKA+I41$O|7P7m)not7*OD^*
zc+IUJ)}e1zeswd)=CalpV;oJ%71aYY_k2lr%r9Slu+BVFl!_=pA!x0R@XT+QZ|8Qt
zBOZ5&bzNkt;#Qjh3(;Szj=IvXmTAYX3j!563L4@)&#;)uFkH_-he@EwK#4z4MDi`*
zo~1UVc$=#IGRO6!*T-Skt^0U|)vjdwym|qVnK|TWUddtpJ26@kUD@R&_(nQ-XL;^U
zIWNj!B-IOF(2V2L=2JLP=ERY7b+^f1?Cx*JBj>bd$wv6y?HEw3yjEL51V8Z-J1W-}
zHBzMO_$~2lZpaGLdwAu2m2OWhiBPMhWC<tt?a;~L$H;6p(zD>k7L*0;3ly^&dZhbF
z!Ix~5o*Ftx_L4iqXg<I)`VtoWM==<ul7YuB{ak)uop7n~vIK?Vxyf!V(;0YlcRDjt
z4?Pi)y93t+9Pc-Lm2Ihworp~Sc<o2rnZrRS^Tr}@uBm=#;R~Pr8<1skKvKrAjU4qT
zT;A2ZKy$2v(rLDZw&!ov_m2CC7js&vCrnIsc)Tf9rJ|P|U(Fek_ZV28wi)1JL6%>d
zz=~HO5|Qu>q6gyHPTIie)E^_kKW1M1ktv(6l1by0UXdKulx_HS#EhBjz5T4Ubs&K3
zr$}*~ql-{r(H4>{J5{sF6uNYAPFXt)nlCT9c#a4jX!9X`y}nKAcljuyK5@1jA48e_
z>>CS26<=kjW84{@bNABh!gd<P6*aH+g8pgPoN|uMip$=IC$>yjzM|i=6%t+?BDw9}
zCG~M4M!w}vxy|6bjK$>ef8rB7`ngNQw1WL8D(|kms#IymHZ+G<HM>c$W_*c0yjQW7
z!CR{_V}%q01;tWVs96zdXaTs>`T#Y6Yuh<m-%aoZ`~6q)Xp8S;0q~7T_rBWOgmF^i
zSHkpFDI~3CMf88l_!y@+!+OF@3Zs?dT;r^6>o@vcd#}{@CyRB2N>~Nu6?ZRR_h3x&
z{><m_>{IRX8*A45HBfqlm`NYvB}+(iN}>iA@GJh;fn?8Pq1v_0Rf>z+tyS9*&ODYG
zbTvPsu)fg9!#R4j3>_nMOmdi5_33^rvl;FTrK`UyGP{rISjc)Y3+KwhXYI2Xq9qmw
zVe6{<sa`>-!K7VaKkH0#pp9;JSN3<<^3yYeZoW-e&?cS4$luWj!0(VvIv)txO>`K2
zNp%t>bk47t(?m4k(_5p4qb1N`8e6=BoJ<<({Eq}tc%vtHO`&;iMS&+8N1<g_1>whp
zZ+VL>dz*bI|8}Dt!p$R+zJn+^S}4<g^B^x2YaT=U!TO<RPpW#dJ2KN)w`cjnO|(?T
z-F=x{r7u5sK~duKmU*z(#wYAJm7;2Nv9LcJlLX2CI)jf2_O_xEo7x#dwo3QDGMZ33
zB<jCoAD0!rn6<0R9-zulh=vxF#u5F(6WNw9Jpxf_4kiPGseasNN@*qwTQG$iGuE9K
zw({Z3!FBx0$nk6Ld){eKX7((d*d6Ue9(ZqUQ}FY3$5P$LyfvPb|JD!BGy3#Pr#HNt
z3vd|UoB#aP&4L)y|Fic(Tg4I_NI=m7EO55Te=J;Y;rh$Wo>eXX-s$gjI|fty6~8hC
z6!Cp)yB_qfMTU$a?-3_M<hg-2yFZd{Uf6^+fgGw26e$)vOQkDtPNhE5?fCo2CS%vm
zjkeI-==(yP#<V%z+8&uz3`U3#+V=12Fwx!9HFCsaJrh3wgM}*21{=Nk+8x3`zZTiV
zdVTwjPpt#ncpgKm)-gli*SE2>XXKqW5yPc6eO0~=zP@&Jo30<)CBK#CWVUj3KHtwU
zuY%0*T57N;`Y#YBA6^_*mlwY-l#PK=uQD_G3-4ym9Bf}gY~<G3O?-@xEIlc+bmlHF
zrJJ`RTED|I=@L2(VuUHocN*jsXC}Jy_vfs<F0N6&;CEI8Y_IVUoM6UT>N;f1H6cI&
z*?OXQ=hWOKq(6_R#}?!x%8KbRtoZ|;nSj(!d&@vZ5nbZe9bm1j2Pl9Heu9M?S&tM_
z1WB#lt~}XUSzX68H;uI3#0C!IqBTp&=~q|zMb%v!fhQ;!M;Up#nlNe3#E|Ry$S{KZ
zvb1&|LyBx#%@MX9cd@(*!t)aLM0LJ&`zIHhSWNQd_M)Qa&TjXYsXy<GG9&NdKRV_l
z(!QxY^*1EdFo(FhMr;M%P+Gs7+VhKZg_qkK#Ff$~zLz-2xJ~i!TvtxQu}wu;em9G5
z=afe09sFm;l^0(1o6S43R4Rdiw<~}fHLD;ojo+vbo}>ypnL@<HSoRg5cXG|zPk;An
zr&XFyUcnk1ueX`vOu|^)FoEHni?*j6`0N;31nuhX@n00+dUavnSXzHpTT^Q!o^X$}
zt(uune8xntSL3BQGA0NsD!M7&|CoI6DAXrSdu3{fS3GB9LAxcSm6u2RL?1Wc_@6Pt
zId<)LbvXYdk2CW7t}=gngN*~{j${L+tJw_3vujXFCUx+lCc`Y1s&m^iYS5ayD$Tuc
zi1O$Aw#E@Ri(X_fZ%1#ip!cc|zwHrWJB@C~fOUneW67wjrQ@rTImPuDUZE7l(E4rR
z=dMvAzv{%j-!r2w80VWt_}S)W8ECD*|4{mQw<*0~lgv!cI%Jz6bx*Z*{T@0QS@cy{
zjQe<F=ZemtZo=D2b*X=KsqIA+;d^Rr|B(-~6E0OdXwv?oZ-+RQU;Jua^sA7#DyTf^
zZAP{{5M?<&Y3Ng?iPKG{Zwavvo%<JN{Q7-{a60r7dCYN^r9l%)3y)SrOyBnPGiJ{Y
zRwss*JuJaBB4uWoDjijF3;NwR=9|-z3{E6ko6F6QXMYET)%;S6v@DA3UohaDD{5h#
z{UOMFLe}#UKO%{BdD}9il^tT)XvW&hR^sw<cWZCgOH+N?P#;O$t)=?Knd1k7B?(=9
zLlOLxN85W>IyVt+Vt*s|TXBR&oM>B;z7XP31V6StDp71qD|cIAy8ByO{}rDq;X}V>
zg8YI}0AbfzJ8`CMS4Hb=V_hSf4v~81y_qWA2HE>1>hF$iw;wS++yM{M5;>RBdAO;7
zS4G#h$6!ErTQ=IcP`jQRd1P36;^-T#q2`96mU*PKy=@`eY!B04?$YD`F1;9mfQgzr
zzy^}0UQ{|EhVzWj;pC!WCZwT9!}83<kj`geT$3vK_ESo8%U_Z~vl<hLl~?}m)t?OU
zeoIrN*;=2$J~)g3uEys6k6NL|NZ)jNQ5;r%Lup(#j+MP{%?>N1b(U4Em0OMY-Difd
z(|q!%3;OS*ZcX%3X4}ZgFGeFd($z|`30HjZWvz6hQUCtU=;V+7y#OnHp$X~3&{4#-
zZH(J8>Z+_R;rwa!^T{zuk``Yt`Qib4r@VlT5dx<}AT6aVj2uqxI3MRMFT@r!Db#T4
z685<d=~LxTO_gWf@7wX7UX_2!vxtT#bv$&Xwm;Iftp{Aq{Ig?1n^rhn<jr^&b|;cX
zX&W2k=(@^m?aiKJ6Cp7qF(@JRsw48al#Wg(=2eSRRf~5Oi|b;9;p~+_mVMyd=G5gC
ziFTlSB6VU!>^4~z?9a)v(KyED56#2Fywvw^oL=6NGoGqdng;U3pLqd58kz%jiuA9I
znQyXc#x<w2@33jjR99Fr(-^qkNQS<MccZo9^6IGB;CCFou%7#6DFLuWafHB^H+a41
zp}YH=vd%1&WSxBBUzgX|vvhqDUs2F5okfPGg-3H6D%-1#x_oyLBtwmGp_IfhPE{Im
z7JbImR*^^_RW$fm0?YSK$>qI2F9RnO3wt|Z5WUlsp{-Teb&IYvkIOV}Je0R4*a_Et
z52aNqj9`Z$QF^Dn{D}43omVQIX!vr-M#S{ODwdt46AA{)+ju+fe*eS{Yzv#6ImQ-d
zB3@!+ovaS{pTO)=XNQ9xJ&9yYMruJkjE%#bA^TL^7#GE+tQacAH`<TtRvIsgt!~AN
zCzGk)4)#B#C&IVAmQG(H=R-H;ymt+&MD_^p4`*q;D({(h*YMS@9%x3r+e@P$8GSZ#
z-;8>n#dSZ!o88$e9B927R3UT2$SIz%@V!A7#T0`4^_S#F7uUi5^VN@xY5TQA>2lL2
zX@077H1lgj2um?S8Xf%*-lQe1SMSI623@IP{zpl2>C|96Apcs8p2W7i`hKx#<ypNj
zMn3sLx=QWD_WPM9-ZF#Em1*(;A{7y-=6br9P)vmh?i8YU$1ey)s-sUiR~6z98nmWX
z>FQ?w_2Fe(8*0w1RT!UWL}3Mwh}0nyZi3E?EJ3@}Bo40UY?79Mjzp|`nWogJ84Wt<
z5Q%mgSEO-jwz4^2qq|`86Ho=uS_%KvoqeSDU8@PjrAphj3N5XB#q*@^(X!xqbqfzO
zQ}LJvpD3OcZt67<7aC#v9<RQt#!2O=kb<iWTm=1a)#a5&H#*|wi(h*bI}py*z9Js`
zcIPP&cLL#OV^oyZqhICJwC@SOLK)1cn7aSx1@1{KQBwo~@-G=Q5j<p3y#Ib1C{Z8I
z4cG$Xj2dkHx5HoyjDUx&|4svV8C3fpw!ob8|LI#F&gcJ@KlK08efZygLjUiahlf%4
zzvZHeu`VKDS>08i6F9V7A>f~^q@qNbm|=jWF}nZ+Qq>x&tSF8Bi1ZP730qc1LKOl*
z0{=yVpre8xr|zTI;Fo-5B}Cu4Pw(oX`M*`KEDi0AaqgFiiv8;k#Uorl#olb#PMzQ}
z8h2G!<_N^63s#rF6a5F-BzP<D58fB^On~cZL2abYW8_r6b$@Gr%c1}NUOWb)e4){`
z7{qIzHe|he_w2vpef9*TzyCLIk^g6K1BULyD3<v@M>F(c13lpX)ffN2{4DhU_4xi@
zJ+A-DXHowz=kovlL5EC=3$j5)@mgdYZvl8V3;|z=ragfVl!wXP1EKlDfxYLREJ#Mr
z+Q!&LVF=iG`b5Uak$gf~-{QVx@T8(s$)TfFfhUU&JByKCeYp34OZBLL`T8I%Oa}eG
zD;z?Me*N!PLw4aH@{L#*GgD0ZZ*S~F^<l1jMXB1ia&IQuq82nKG!x(QP*wG1ndjwg
zwMr`wu)T<H{^U`~YM*tRP^DX&8ZC5u(UmycEXPpU>t_Jc)DV&M_~32{!9!Moe)KEH
zIBz_5%D2+_Hrm}A8>PM@o23((UyE_e`g|h3g(poWqqm#1WQ3<mX$%XFcl<i&{n~!%
z@t@k7<B_Pd<J_yCE)xt>$&I?t)5<jkSj9upMDd<@Q;y!P8Vx%J$QV+)=f*_aysG=r
zX0jnLT%q&2{->pzz)v%}EtTA_qXQ+&+s;oyb6X9nTR+F-AXq3c-R|kL5NF?;sy=+Z
zt+vgep_sD4#6t!EXuL02*MNPwrl~j(qmf|URfIj$^GTs`t2{$QLw|J$`LX#8vv1Hx
zJM;hxhYs_qX$WRDhMg9_`n+yqXEg!msq-;}7TpM-twWAOukdUz6(r_T>n3Je5v>e*
ztI2|;+%xtD`BCSq|J2Hqh9I`Oup*;WN+#y#$b)i_Uxg`O=*ua+Q%}Fo$=m6jrfg${
z6=M9)H5S|1q3a4Jwm0`JGa>yoccswKjqjTJ#(t#dK)?Cp<@pwV`iNSNIV`q#1Zt@d
zU3|jo*>$X!wias8nuwZEEJMiy>;*_7ctN^P%a_}-jIXZ+P_BlltBNr$BiEb=-wjyJ
zTKb*usmwn`Tnu5~-iP?pLoeKjsWCs;#2{?%^{hyaj6fo^s4`3hPhyJeCgdyeOkiPo
zOIP-13YVYhBY9P$Jnvf`D|O8o;0kMG2mSRn7(`ANzM#1B+{v1zTy%l_Hx`!E)}_lS
z^N*VuW~5VEJwCYB7n>Rg6gZl5RVK0C{PUI++<v+v^5gTqVfYY>Kl1HQpNzLd_@b1u
z|Lr{1Z7TCdstwi0*2J#Z=&~6(JK;@sOSSykQJL~p?CV-VywJge$J<wgd-?9@(sD6>
zVWxi7&+aE1^{xjzB^il0BQ^X1^^Il}f+yjGc8@e=8{`vNRovCtFWt51X`7eQb(=9I
z{N9eA%a?3%Pn!M-Gcy|U&9E|1h_xWTjlw#S{bKxyrg&DD6a!bguhb+|FYgH-M$I#$
zXev`j^qa-e&xY{IlM2o}Q7`o#A()?4GE%F3o05V8EB<Dk-Q08QiLwNcE1<?}GfADs
zXP*e9A~O2=nfH8CQ&XyqA8ajweBNb$R-2U1ZkHbE+!=)ydg;;66^BVxqWe5<ob7BA
zs)h33h=^HbKj`B6=x04oIkHgx^&=t=buZ5-zj!{=BTTx>V7~1D>$(N%2mG7!h5&lv
z2^f<JCHl=ee$61DcB<Akjllo**87eR$mUh6%oMP10huES5TxG#0c(>)w#@g(%uuYL
zFxC*9^0`?&f72$C|LzvCG>1<c_kT5^<=;!oFT%h2>fY7x9Tvnd6n0ECx}EP{?9CKF
zsi3|9n_KCyB&H2gZyq4afR{}Ng!S0lJK3ZTz)I5n5P|)`86b!ddDfPSdR8<hffDse
z{vJb&3F%n~vgY{#`)l8be;HGSxw8w7{?+GNr$p`<*Q&osPoXA9yo(3`&)@_(Sa%^M
zR4G1|w5`KI{^ti7eo={S02H|fFhVia)z$t_Ah#uHP8Xf{QdmD}%#-YJ^N3@pwE%WL
zsOS1bKH58N>m+NUIz07)+^Bouea3Zn%8?L?HxWo#O9E!TH(WOlc5WaPG68^^)TN{+
zr>>St{=QA{9ME_2dkAXDQ^}Qb0tS%yA`Fp3=^y3Rym-Is9fq$<RhjAfVyNd?{u{U$
z&yHs-+vkeOzRvhiM18T=Jv)0j%shP7xP*BtARtf)n*G`U)9~QzI0nl5F2HZ7*~P^-
z(AM@Yq`a%0jh;AYMZA3AiMW7pFI|{v4br1m-#?R?sh{mLFBVe=fTv8z&(Af#B+FZM
z&w`J=FKC9Vr0Dv*F~^!1c{T;EhuhD9F(jXaAL$;%g4C3^8h7B(Bj0yl+g08w%Rty&
zx8FWM<??~Dvkdv;KaQ*_|Nq(XihMTp4^LOEVp@y(jE)b^J7lD+iYL$+OVd9ZaBT)4
zVINnS4<sI~bnHl<UWua0VO|eGq|o{7W|@AKax(lkWR_=p(cTN2+|4PBBTSN3*4V2a
zx%qVo)y#gRY6c}FAAk~1-OHVg&JKr{7iJ$Idm|k>g^MNLVLZ$~mYI9%e(|cK*#7JB
zJSTC~qvsm`P}OyPVCz>M^)sR4Z-=&6NBIUY3QgUEeDtCK;nSPuh0o@<WrYFQt_^Mn
zFM-s~)%J;v=99oz`HvIcLi6|=k^g%xNk~Q>F41v%H+J*3{&3gE4bWua>YvEh@ww7W
z^VR><s^_nk3C7kPpq8%!Y+up?#r^i`q&Yy?e6YN;&7i5`hfD|JAsNteqAa8FzV(-x
zaZtQ*JH(BoGJaj#Abs5tvzP7ppmo$ahH#EfHk`#?_P^_7QL^$$8eVzxiZG@)_0ub@
zVOU#Jfl?pCa1Th?<F+}tI}s~4_p|E!NY#1|`d-ze558aQ+xe0y!b{sJxCXoXp`pOl
zYFzU|*3ZqmEbCe_-oE0OCvG+||AY{nGe{RCx~o2_Q#OxEY|G3+Y=N%R#dDfYxi@@I
z5?$yYeX7P`KqrObOI%8NTm9B0zf`cK(rBKu=8g~S!<6%F!fV6JQ(kv_G(Bd=*Xh)1
zzxX-ZzB``zTBq7VrD$D<!?F3*!v8_iS;sZ`y>EOQqidi@cZh_PgyayUrKFMWwkU~>
z@FOTlBOPMU(jbj=mmnozPy!B+W^BK+@9$rIy~eZioSk#Wbzg5O?}+|5W?q6LLhdEw
z*(TnBDq6ZrH_-R%%OURiEYIhwixpnZ%g09tDQ7MG>zL!KznOW`2TT8V2SlfaI2%O8
zrk}jhn;4hs>M;D{>^dB#(S>?c`#0GqvQ35S%3Z`%bvvTHVOpHj5(g|ye*-`1X`^-v
z8f*i71cp-zpP+P{LFixG-dPRksqUQCe{^58tKPODq;{$9VQFalqxU$=|G4cfKL|k#
zpn7(561tIG(V~z@K_8QG63X(`yA26(?djy_n%~?#YQ_Z~KWK}3_hRter=fYJGO$Bk
zbcPq6iXaXrduw=v0wftpAi6vQFm)&8J%uQht9aS(Pg5Q|>>^erMXhw&VZ$xTR)2cE
z;^c%2&q%pUR(slH&Yj?$u?xK)T&j$&Wj1udrj?P|_6A+b_mi~W2yHR<zDi|-#}mSB
zp26DC=r@km%Su^FZ$B=5OU*L4MH&2T)|=0!?Ax1XnWU9J@9h`QbVIX`)J9%B#se=B
z;VS}q6Z1F=TCDyf$7j!)AB61I>p&zWB_nBYxSn(mn#ZQt`gicIFBcIfJCPs0x1t&*
z>uADd_*VOUww(|Db0Nf48lQ}cyExNW^<3r36yFiSts(Fvb<6hGTbYnoNW<}<iTv5(
zsN;;uae!nTgYvr{ZHjF?&a3G=pRKUIPMEw_1O)Ep-0^vXgstsA0|%d){|FJl(eFKH
zO4n#`sy3RVJFkd7i`^KONXyC0#wT(%d`equX%Lh+^&h`>pM5~~J}a^DXpG#eekpA@
zA&gDzv2IT(<-fG)i&FW{c04cOqzGtvDO;?);iG->A6~3nP@dknA;U=hWa0p>6^h;Z
znD5*ZLwC?c%DL&77c_7lMm_{4WL&-$8yN$aKS@TsdK8uWbmz3t9zVP#-u~gt^6o(!
z9euc)5D{m&9go-X-_$JEFGzE``%*{{r>Z-8e^=XEcFj3uF)J0io;|4;w{;|=+HrS{
z#h738erT_o6_dVpYzAwalvLk;?V$U)weUb>EA~zK3Zuo@tJM|YW#?mkN{O2+B{}7V
z-_2`BJ<=`weG+lpsaqDFY9J%>M&)Gp>oG)tqNj+Fi>#0YoNmLBjQl@xHAY^jVKbU5
zoSi>*P4yoB>rMFCDkrWk<8;Kn8&%4=_SpHdB@pvEyL?6VLUWH3K!__ZO{db**Uzog
zquAcbv9VNCRSocp^aMswAZ!sd`VSh_*S&2jM2WD&r5a-B1|x^0%TND1Px)l-hB;~+
z+fSu@oDajxi~i0^54*R%qVH{~(wn`e8L+SCq3R1fZWp!z6=B2wgIyR*k>b3G+cIx;
z1%NP^#qH6~E-r(BY4^_fy^atriV30KsV_Zu@A;UW`Be$6M|J)=PTAi}WhWTLUAn~%
zM*N<*RC-+58Cdb6dLcRD?vSPFh2%k25hD>2PE%28!F~?yR{JuemA+Sh*u)n-VBE5(
zuqCe#vVu#EBD;iTWz9br<liJPv%-7U+k=MB!=Tl6K)3oH2ocXi;HWq#@ba)MH7WUd
zNoI$Z_y4om5B`Thu01?C4dkMIZ%U1|d81`YDMI(qlf+)~@bAL&gX4Ua;XiE(ap8QX
zN4#cwp2Fn+yX=h24jF@nuvO8!4h_5CA5S<4?Fip)FfYdY=%f-Uqt_HcNKhtXQHAsg
zbsNjM0jB6eVn3fbLn0Li(ERXpwy))u_shtFR8`RLRe`pDBH>g4WfiGqPQIfaQ7a!^
z{RN{ZZU6h_yrf_MGk*@a5!#k3B@Ifr<@f-oBKX5dM+=v^Tk|KM8S^oT`S5L_dcAp-
zZq~zY(e>Jh{+FkrM6~Cp!seB!V7_xxvlVyxi`KFu9Fz!wy*Ou8)pc%~BnW?4qsn!d
zaw#lmro!5sT(!)sqV5Yi!Dpe)=&j+xx2hGjCb9?y+H1iN*r9M7LrdepyQ5*}^ABk?
zp5p<QPSH`BB~=k)hn_nBySl$*?0RLcuyt<!`X>G4;>|4;i!m?HA*|{4ne)XB2iAor
zB;qUHSLZK50CQ_XtqTu;R-0JmNagb%hTRwPYy8HmwCQ}O7i}~2six#mX7}0hOfzDu
z2Jahy{AGG)qS@bzpLl{jCZWWb%aLA@S&)PJa3&7h?v^d`y<S^GB>zTjV|ltf^ZYaX
z<o*BAf*zBEvo(>ODvV&mU3Icpjvn~UZU@M;UCz4Rj@$iy?Y25PY3N|GR<l4;n==9h
zf;N`J{VywLwSCn5SM(Q777a)+nC2fCRkA8aCg%dR7c>ONnH$`l!ctPxe*Ro+&YhMF
z4;#`9<C~uhJ`MJtF=!SkZI)ovIjJi?tpt!7INhDq71!C$Ss*OLdw(~0Evt-W`c<}b
z<KJQ@Cyhfs_0@Ju<(gpUQ?diFNv9e6PC(eT)=2f2%8cqXiJJshO6<<~bbIz5Fk2S_
z%$8#_@*VfLwepBDvxrX*5sTz1w{|0pI>Lk%4jzAX5m%Fj-C_jHCbB!Qqti!*JbT@l
zyDZZyDz!$Mzxr2@Uwj6)#WuTNp@4q`%VBja1~@Y&`E^{H>7I|F|FSAF?Fd)qu7>PN
z7@=2Dn}siixL>fneb-4rnLACAP8I!8(om7k9-}=+-O=G&-Rdm_L#^$k$E7+x;r9IB
zku1o+{o<|4qTo5*sex=nN6Ym+#yREC+r!xR#Ft=P6yVXT`eT-Xo#V%wVeg5<NU*0|
z@Syxc?oD95%LYcc0o=`wpQxD1-dSSFxpVnrgV51%<!H@>_w~>pCzyEmJ=qc$EuWV!
ze$x|@s0cfqm%hIiUCxVb%A^zfo2k-^{?v&SuG4BZV5~Kk0ZYX7f@zfC^+mw&OcY?Q
zVk9>E#klvvI_&kMm5AFyPSSt9R?vT~c%YRl-uX3xKf3)SHnm84)!V?qbFK1yAJNYQ
z&cl;E`T}LbX5z~X17Gaq+B);mQ<vL?U8)Bb8^7xOSYAtZbaaPqjOEIw%X-a<WCB;j
zkz57y{(oS|no%5$*tcPS*q*<~kG?@C1Ou~AFWJIRe^@aT17ddOi9xV5%$;(nYVZEM
z`vPpD#`)ccXJsUW<AqPLJb;B+<s5r`)=&7vUrvL?&Get^<y!UTky2B&4Q-~<v-2z0
z^;3a!mAe6Fdh5ucHi;pF`_#MG1~M=zdF%G+YShuVp#&Kto!~X-wUp(f+bzFuN=fVu
z+6|j-8_%4QuHIiPg%Z+i@|JYj{y6Pjsg^nXCv8UNWsk&LlGf+m4bf5JUvG9`RO5}6
z*AG7GW%3ZPB0y4a0rJrTB@{_mGIZsHBMV}26=3;n%U}e~)HPMPSoHMkXSv+4KoQ)X
zWy@Mae0j15Wv6GxIRqSDTAkEku9p?e@t%&Y=b8cXW_QDv)5>mK2otSQW6Ab5Hw*wu
zbT?3<rNe*QjO75~BF`+}Px(WK%n!N=ACun}nW46rv=5F&2jJMR5V=z-o-Cdm4FA1X
zb0hGRM}*7ceh6Xl^k?(UO}G^9?R6A*$TH6dSIM89z1B~Y-RpjN@=C0ILQ=&9E52PF
zseF7CYd8SuBQc`ul;bXi0pxS}UkNfwx+;>pc`HRc4--77Tic)I{jr}?q<ld*<Pw8&
zwlvK;r2Rg9m4V{BXo$YHn=960x44hb!f0)a&%m%Oujzr>*Ap<NA+dP<i!B41TB4O0
z37g>PH0ypy2fa=nK8k?2!|g+;uW&yj1wb(MCDEOEkDZVwqo>p8=8E6k)7vUd8byy?
zF55{(&o9WZ1SV7Zxhg#8TI*nqM?%cr<&GK;(3*f)a~Rmp{RZ>r#!-dfvoElr4g2P|
zef}y~v3J%Xh1!TfsKmqZ0K57;&q+O8Gs_ZJoBcF-IQwL(gHn3?L>d#Jg!z~sUMdEz
z#$)ifZ*3Tm!jNznNLO2_I20|Y{~PiWGWneu+4baL3A2(J&>33$(7kN_d6KRVVWE3A
zE%MJ)oUx|ULfWjg>492N0gp9XaJF|l?q9NbuG+Odt3N(1+{~BvFAWUHzYKfJ6aQaT
zJc4f(*u=hp5|M1aeS2^vf+sA%P}SONmRhAQG`-c)`?<sQoGxv`RTo3dbzikFA$;xq
zXvok#Qg(rOVbhfIx~G1hR7ClM!c=GsSWofi+hU^o39PdEqalwUKV~|O&ESQJGe*j+
z0x!KcKxFGx)_G!^j?g55#t>m_<<$b1-%|TPKy?a+)VeEw<^ve*6`-!DDNmh#pQS<U
zQPA_Bi+<*fFWoM0e#HVW-2VXF!PXQ<g;g)V5>aYe_GfeTE~Cx6Us1C*6TyC@E#Vf5
zq7C!bX2UmD+AE5pE&F7-@?olky+W~hf{_Uu<fdL2E^;mZRFb(X_B5uzc!+yX?Y|D^
zL3)BwS1SPK5uiIH;XHq1keBC2YhcVb6$py|ymUMz0~FJqjv?7Jcq-oR{Ff$k!|4<0
zuV0$%xdSN(JEsn<=Ke}T<Sd(cE2Mo9abw&f{~xmFQ{fH@fNGpaG9;RFX`f)hlef}C
zQ*n!R9=zsuVfALjLkbf1i?9_xYhhW@bCdS(%^JNCIrLn2RIS!16E0=Nh<?bQW^;A)
zjvNJ+n;9%``KLQ)**XVDb}(ocF3}DzTl=+{E{9+E)bG@z#ybsqgy}llT0mK1rpkWi
zX@yn2313EL=GNtfJuoI$`T#e46!norJ<^@BKm%d}xeg~YgT%xfVH=rcJ1S?khv%XU
z%;^zBKRH%*+?HE|59XW_N@`Ya5057)^Gq(l5F~d5gSx3@wZ2u_=pvTvW6=LTXSZ+3
ze)e8|QPU+yZg%25#Yfoaddj%oOj4OQs#|!AH@{5KzqjqLl@HK*1rdaj!NAX;e#1}q
z16U275VBRLkNUa4fwM(Q2cY&Cw>>*bRFCXsC3%%fCNfv8io;FwSvLI>t97WQH>bMd
z4Mos%DA<W!U{WaK6Y!MZ4Q3ysiZeTuRsm3J(Wi0H>r<NmsMZ~EB!Z8B`z09fROb(*
zgEjVDFps6C>MSziMS4NoDpT*OnSt^tj)n4UOMyB~5Ow=35lN*r{qrDP+vNUzoJtx&
zXDbuyhRolO)T`6qcfnC^u}^^2!vQc(93mKuH*#iQc+VOj2A##?Jl?$5eKQ37%_@A;
ziz!ShAKczdN6!#DXO%8#_n|3SW!*9wAh)0%K@CBl$-x56QHL+;t@^o3zWR7?wPO+i
zLA}!Ld0$9NQ%f-T{M&Y78+<flZp!2`YQ%(%b*>;;W$w(_F1M_G>f_)2AGQBhVyz7#
zj#?IX(RH+#@e*m8>C_X^60d5`U$?DAIBV`aIKBwXWJa;^`5nnm2c0(xa!BZCKVrw`
z%Cs9ECa_5B{O!PpOCMLG#$1%(U^cThF+X4I6{Fa!HdK+3UH)m}&W<-m>gRuC?v!1%
z-erNKeUcL*KhB+hQP@bur4zOe;B}_WVY1A^)@Gs%QUI%5b@5ET76&SCp0fm`AShO{
z-VKM-6|jgq^-)<i`sUVo&gclky4;DL(tbR{9E?aU&Y58JWi)Jr1N@TsTV#TFKNFIj
zJytCD@rC(zxa3sRMQnc%Uj{RZ6$>h&!h{5tz>LsFKe?I36*E-^o#@4a6YKyKdUOFm
z{ym6Klhf0$ndowj7Eu=GW{Pe|ZwW1q;?SRD@HgjO>%|GzEGY?oDNJMM+xY1@=QiTv
zJ)Hlr8~UGr8{&*$aczbB_$8k!@)HRXnCZSbtp$PLmVSgiJhJy^r{m~Y_0hrrM-gcT
zn2;s}BWztNz4wn_Okf-|_W~C&f$3Pc@jOw2+xWB7qECaJ28hBdP?ISF!Hj=qg7}vE
z1Xpo7`aYXnv8~tOje$2FtmZ=Gg0obFGsKt_<$c&P2*<yWr>3}ly0N}q`WyqSazDM-
zO~;b+ga&*8WMkvU-;0fSzl$RDbdlF7^on07+oQRmK|CFEb;kMoRIkaCA$yWq#w=J6
zrRj$G-rju3ygYdY`(9A+y-K7?Pnh5t9vG1t?bGYx;B?`Ps&&ABBKyNQn|E|=O-&>{
z>9t_3P9uM7opF|U=k;~d=C5VRbfk~(ph3X!c9k}YGw%HR<X^2bgBQ??s1C}Cj#}EY
z3rvUCARZ7krXTYYsi>-tDY}D~^@q6%l&=Z!5|G(<*WWD2fX7tQ`r6f3Qr((CC*ZC0
z>5(8nn7+dq6Y<Bqlt=7H&a~2fesNp)ugvLU99%yfaIh+XndIl_1Qk^T{r(N;92sye
zwycQ${s&gRDm%^xp^mqlC}`$_3V*#L_ll43j5{N!d^)Ifxye0&<4ct-KdNwXp?6TF
zm{nQ=8QmB7sq;g3POhjw?Tt8Xdc3Cd{rW_m$9<2L_AtYr$kS(?D@CgOy0G+NS2dUg
zIe5SH5~sTRQ;9=c%#vG)|K3?2kz^6SW9`^W<)X=pjD!V)N%h|V@0|_~w&$1Ok%=!S
zi7npTJM~7=PPBytY8x2PdftJ!!Ac`~`BoX#U~@Et&z${%;g(vL?Gd4hxX+aSTnnLt
z54$)X1Z<!G;Z1$_oCfo`gxdM5K(*IKFnXg_Gl9dF@`t^4CsGZj94XMb1L?v+k|E$A
zfu^6YN-Sq}82I`v+A-jtt2SNCuhPE9@BQd@0gMG(UAt#zSJb@|A!HA$HSGb1a0*xf
zGLVw!!X{`S<flPI+7NOG_rC3c!M@1RpqNl%Hm$tg-;?4?e*Rw5`4tj8sKX8dXBb`s
zrK8kJahJX8F>?<D;j}b$Ad&*j=_@Sc#hQc{*>jSSPk!F=@89nsvoL49ll#AS`f}Sm
zQL;Z)NKhk3D_|J;05Iay5nOa|x@^$OLg6gIzscqz(2oqOvLL!J<p}YKM6^id6#*#M
zPu(rMZ-<8({z_ePSatz_N#PLA75hk))&YrMcwp*dxMhH#SNWid*o-85x-(}4>Kf|{
z!kyYMKqZ1^TN2*f&!U}#5l$#-h<jsr(v?%L5BswGUw)I=MRy3BQ~0mG_@b$NuAhnU
za;tisZ~*h-hl*M^i0q$KY9rgs4w4$#6E!FjA4)`N34T_o!IR3<B)*>cQ?kgZj`mPJ
zKQ9VB9Y^8r_a$&*lth@Y<ad64V7DK}iBmO$nFp8i!!1>%;Vq7Q^w`Xw;b`RzonPgi
z$xdTe8fAWfl!<%N|8&tKAiR5?_M)Sj<ks|BM!wW(_X1-VzMEFX8LTlJLF$-<U{i1#
z$)LMM-r4l<&7Y&SG>(s39h0MxcmB8Dv)#V7nK!l`TmH*a#apvtjU-#wv%H)^&6f%X
zPl8n4C)n*Il~pzixzi{9D1o!twq!23_Qm|*PtM|b2PnYm<q*9vP21N`i_Off%qq8b
ze-U570(Q2+D7`Z7)QJQYsntmy0XX376Euljt?IsBA<-{iRx2w$uYeSiCC0=~;eBhD
zNR^spL-M#sYrNbu$EcxIU7Xpy2<-KHSv{);2=dl@55Y?RVVHQY%g8{|yidNDLuPf(
z>D)&Y$l{k|X>F$6U`2AJ9D-nwIUseQ`Pz(OEu_I-QbtnY#LprYP=MJ+T7JP#c89){
z9qksr|Ff6ihBORh={DKVxKA0MjXl?l%RRM>m(P^`a`8anB|RS`I{Mp6heaaqfyVgs
zt2Y-Jwl-FXF3yHR0*M%V(t%&gQBmciBhOGlwPUyK@W~*T>XIf$IvUw)BR8Dh`^+rV
zd=k%3fr1sf>x|#*rIuR8y_og<a3`0F(HlT#2?SJNR-|hivMWtkq&xIzv(9_6-`_Dn
z8~M0Kf6+}$n6PtVLUVQ`cKa`VV0Vf^&PwsNV^`xa(6ns%z`j4C5(o~hF9|nr@WJ49
z4~#A%xs^AP>B0tt>1@w73ysUnM9s4xInu?-4HZ_J$1E#JTdd|f4VF@4SE4~MerK;=
zP(ME_e(n46hWmur<H?goT1AeB<%jGNj?p)19rmu1X?uP{>XCXcw^|Y`(n`!KvRb}F
z^P-6qh#E5Q71hezzp4eI?svmD!OyCPCo8sY+DgAp><Ys0CXb{@;H?Kk!jCa+C!Un(
zu6$7MhGY;sycgF(nwgm)=?hr+#+^GK9nmJIg)|H`F3xJr)al)%;em2}5j)qNtZ%QE
ztYv?R@9bJRD^S})fv9Y&*$}K0*Umh8XcZ83`9wwU(J1VQgX1w!6=vLeU-QK`-<v7l
z@QwmR$2uCGLN_3H(dt2*#*;hvN&TqzZt)r%y{J;J>*8@(z<yF%z^Dlmo{Q{}30Wa^
z<V|<L`>TB3Wp>r#n+R9Pf;VTcq~uGOC0KTB7+%==IEi(Utuzz$4Dk}bHvDFL2?MVA
z4Xa&Ug!%z9aRl7{7J?N{;DbF2;Sz%Slzuf%De%Y+PZ8S6e88y(Lq>OR`Mn7)wlXgf
z^*fQT4eBy@34bu)5QPUdb?H-sT=_H?@a13>&jWYnxC!N>L3Cd{Ys(YL!x-3)t4%io
z6Gm>m?HtOXvZq7~c1+u`-v?V~b^2YY!_2~yD}gWZOG^(Jux)tUMQa9JHD)oLyA`@h
zhn7&4HB}R)fKN~~wYrE^ki46`kqA<m#2g%j;<~aQYcI-A>ezhs%hvdfC1${51ntiM
zG*sZK@Oa>=YkYaf@b1$hRvrI2VwLG5_1hy%sXnq5x`a#PPNeANN1>?pdAUmm2Dfm;
zOBCcKYUHBmqvj+hxM<%AB7(Ln8h?WeJtZfH&0Jf(nxkOCn!MOgBkjpO2C3&YNKXja
zNf+nHSr3WNsM=G<34IjP#%JJAVS@etkN6qHHS&dk1|^WpL0^16P1~^4O%f8$KkvbO
z&~^TPq5Yi-L>LnNDi|g1t|)#|r_d}{6_So?BF9;r#I~!bD+aW`;UeQPG2$9a1w*6;
z>tngw@`atL^j|L1L_>xN@{aV@Hg94O&iO{8f-I_jf7FcsFK?<NWOBtRK}mo!bPp=V
z2foc1!<Pvn5IFbw>rVcGR(*ocpcNZnAjZE@d)U!^>WObV>w5f(8<MWwLOgq}3{2{-
zUV<Ac;9e)NO#2@g&&EhxKy|llHqCmGgM0DclaXKz7`X6K7fw{l&7Lq%U#9*f9ltw1
z@_Kh9Q%dtcT2B206p!mY*F7BvNQ^Gbp5T*RCN7@wgu1B|AWaP$-&cye0g(hcF*yvc
z=u~L=$=D^hW_bVHZ35aPeQ^ey%GQ7@MG_&T=*myD5as{nvS`W&+b<#QoJGzNtI@8m
zm*3!UySdD`*K~VPoB^FH5PG5wA*+teyqa>kF|g}%p$qN{n|SOzsic^gf!ihFIHuI@
zDbHfFUVnXioe47*9D=9c|43>9jH(6LxQM#yV9h3FpDIRu%3);v5OPd+^(Y+AyWJo@
zk0<Z_6@KIIJIPZFxa~fbqRJ1Pj}3Z$md>k2kqi?0S_D!!2yBruJ_~#R4`y_b_4D&e
z?CBR!(G*A%GEF9UWeZb0bz)rNDtoDC(fCzFW;pGsFEt-j>sh19KeDd_1Ie;9$~G*Z
zEuKES6!Ta-{}nY#<nl^26ZVmcV<8n)3Fr;H4_Oa6+~g({hD=I~?-Hu$^X*|RWvi-M
zdRw=V%l22O{2)PaxD`w4PYJTpv?Ni{8n1z-`d9CaJ>|(E<Y>O5bd?y#|FR`AQ&{Sz
zB`up@O{s8v07{Mlkj*)Ni6*DQJxV6OZoT)(s=ibaKoTDV*v2Mi|4Vyt9rXNY`=eND
z5}lU4T(>8>CyfJgt&1QE>53xNt@oUP(ck8iVWQqKsmLoIpXYzlfW2WFXM~`R)Pf{S
zUTAjqzs;x1CGlYB^r`AZ5v==$z(v0!T0~#XewwnYbO+5O=^W$x7Ebr@7`wk67f0o>
z-1_+tvtq|0<6Zn@eeTYixNkyvP^hHVcd#{XY;A2tl-<s|_krk7Lo%(bNQ(bOfX@3{
zW?)|@i(jm`t~+u9TXb{xYfV>iKYpBpIq<a=JgyOhIxm&+5_!Z#;9IAHS7&{(9K`<^
zU8`aZ80Y>X^<XmBc0Z%T>weoae}WgEEw9{Ig+;X^{+T6xM8@@CVq)S0K;qQ_`G5WQ
zD^(1ukK}u;{>ixZp1s(9BXMf(%=VMNn5nS?P%@_`nVhi|mb94s&*V7qALxQVpp^A)
z97RQnY_*g2noT8S`;t#L4Lv|}VQ1;{{SzrI`iws0(F&*s8xuudN-cF1lTa~<BSGJx
zL~5PTN|GO&oY48{=)c$Ozx`AQ;`~WFMSv8cj9P|pLA63(7f;_&Uf6r(Nwh=H9qEK7
zr=4?Lp7|?S;{D9t*A*wa9QyCx?!^VmQ$%$V-MsAyuo3*APJM+1(LPghU$ge7diaXB
zp=JUea7QQeW$o?eo{S=HSvC5Kf`mH-@O8EEN-+_aC*>qPGOnXp{9^5=Z+)qsnx~0`
zt9%kCQ9Y*O(}M8bX2U=J?U-_w|Gg>YrQ=dMZu0uw0QlN1U3@jeskbzVCEov=&Byc+
zP!-BcZ`xN@Erb4W<vD5E9e?ggN|v~I8x?Z-Fjw3v02s#&1Jg$#Ud`7gg$4Ykw-Cu0
z6{pA^@}I{bC2ZmO>><x!$k#vP$IB?a?JF|zB(g^7dbB7yNaPYD{T@H2L40BHrc0lP
zZ4XLR%K=)C!1j)I>3~rUI^ND0lux5{X!sd>Ff1L?n~^P=v{3N;IG!&^n<-Kp?qi~O
zhaYjO$VG=h+~Je+U6PY-wrYO51x>rg6QBh!VtLR0+}N5ZjG|B=G9bQ+Vq1177+J!;
zz}9%xnshDLw|^>!@U8dC{0jLGLM6eAGM$o)zw|4_WwM3>ZkgwKx0+bP&?T8!Jd1QP
z0jH;1Kwu-0pzJ;>IBS7^ss0=V`aU-BtC6O5KtRBNN|{ur8XK9!k7A>u-!FwL4Rax)
z&z%@KQP}EjpLb@6m;#m7UtCAT_Iqx`Tn@GP+_TS~M<Ht2YHim#eKVxhWi0+trrmmu
zs;1lA<KP<LR%-(+K7!C_@3ra~WvHR9lhOUR21;E2G$%Cp3lvFI^d<PisexMk6@a{`
zF9&b1@7@O5s&h0xy?9OtVg^7f*nfl$NgM4F1uWqcDluoGpd!w$#lq9oWa=0cGZPJI
zwfLpZ4+?Gs&fJp1NA%%x+Ul#F`u`kIb3Cpt{M#Qf3&LUcU9~~ZHT{i{by_*5E13`r
zE;Y@&eDvSg-y$<8y0j4t66GCzPoTWqD7=!91fM+)d!{qBN>6rAlZ(fFr38pC2b!cQ
zl_K5SQGAeOxi1}}pF@6^_vNmscR!ALgtZ2c77F}?itDjmbp$6mc$EBlP60n^L|LK#
z>yZ5Ks<N&I2>m|G{G)dLT=}g+PHj746I(*R!oT1GoKL@jM^&O<`<X_VgrNsFU2#xn
zG1W{p;1%5lKed0@@4cfT1FWdAH~b2+#*|1{qMYC@XouA^e&&U`?8cL|KF=p0pehw;
zWTuHi^%93Wp9!kse{td`m;+gVFQtE`!x|B5Ie)FRu?BwY#TAKK?(Cr}DMM+(!pz+l
zFuOQVNwWbB<Vp@0Vl<h|<N*VOG9NiILD2~-uF1PnMqKPLHh^kW(YSf~Bdpzy4-)(9
zv~k`Uzag{rR;kjS4evUhx66Q=MDRDad52WceN7xhqW>Fp+fxiAx!R~o!;6+ti2p$8
zaI1qugF_WEw1}sDPPW=cXw}T5(cC8j{T!2@-kM_CByVYBpbmRa|K~~wd*|tL$oE7e
z&!w-(h`iu7f*iTjUHV|$Ysz$K<Z^z8lVj09<+AR)Na|EmY_XBl6zp(kK!*MEI}p&5
zg8COCsj=4vw2D+AJ8h27%_qxQx`?2^LKsZncO4QhZQ5&x0A~lU5Qsd~!mS9_eT;ls
z2H9&zUxP5PAiB{`5qY)y&9h-^QAfH=SZxU3-w?d@zvI%$-q@^5#^4K^2!a|cQ^JWv
z%<(GlzZU45n`rD2Q7QK}Y8NqBW$b6Ne-~9vjWhd*_MZgCqp>Uz(WkUG(F~W$cX%Q6
zXz6+c)lJJ9_5SG(dYM1XuOlT9x9K7DB=ipft2s7X`$kUE_cGbkY1V+$>T*~jq$lLq
zZf4}dZn5+e>o_N7fQm+h48QY1g+o7ymLN(rcVO-t<@#F&c#amZj(=_Ozc1V|S2C&;
z!6FMKh5$!dmmgoA(|zKZAE)FDQowTxr{!D|wc4a?+XS}4hPi)`aAe#ZM+}YqPxu{A
z8}?z_YbJ4C)|@<7zKMY{1T<!4880w>f!7?i9diHfj5oU%PekFO%Y&^98^bA|pIJD)
zw2y^EbS6mFJeky1WE!2zr{oQyRXR+}dG@E|eB(42fi)zLIsDc)WQTx+v<V$F-v&C3
zaJAux6DBb@uiiH_JQP?N*8s$DhkAlDSEy5j8N&Tuo=%I(u{o4wz=K-BouS}$3G8<Y
zSt3L$$LAtyYy5JTG>GO};JnIumR;}~gYTVJ$ce^p^hKVVx;&>Re`bbT`o1!sem|e<
z*V)s=vkSka!|p;~c`I7+2a`cmA&!?Eou3x#y*`TVu*q$g*Zv0HKhwrXAh0xT`#YYL
zD9DMd6EBRQ$ZoQ_!q+EWk*_CdihfWlM?H9Dg_s_FDwp~`XmLfwnw_U!s@7)o(t0`{
zkZ81&#+k7|kdQ<UEVeZR$JM^Ruv_$AVP@lDIOW|B-W`uUy8wW-yShtbBb1h(bFE@X
zpzZ(&HDacRH6?x+R#itPCU^U-(jH*HmwtrwQSAODLCx^;GzoRSWDx#c|IKzfHVe6q
z>@`K-sg-))zV!XC*5|OnTP2>+Ut5p)Ade8(`_B2~xcHjVL1x#SyNZrT*?E;dT!EE9
zB<4c?&B0vMTaYP~D)-6Tj{J49Gyu&P+JWODqd)!b59GAbDO`?t$hD1Cuetm|k~OU>
zkc+U@?WRZSaWesk4~%<}ExbCj#bm3&IkJ9vB6jVq(YDd*T(?32bn5_sd<V3z=l(8z
zNr7z(HVA&aCR@Yrm+$-`W8Zidss(XW;9~A=G-9@Mi$;kzptDUqEoCzDl0?aH0tR{I
zdYSVb7v5t++q?ei9};Zk%Ae8g?SH}Kep8s5Z&%^cP3ZZirUKFv%&{-$)5_RBL8m$^
zh`X$PB#$kIcVox_q0tH`i5**G5#Z}>UDJ<wQk@Vb8o62qX)+UczZ3E=!|AmV=0C7>
zatcXjc5Xr?fnvEg)24oib!}_I5Z&{r*W@^AA=5NrtK)!~i^k3Y&JkU*lkLD>sgT!G
zQZ->C=%QBIdWmq1L5Z!=7iG4ffeC)SUHAhc_>DO8A8!6*IBQs}u~%KB<njymde*n9
zBq9-a8eV<_q4|d}Z^e%p<>(&>YdFYg3rZt}AOZ4C+8-b}?~72`UQ{#3I<7)xpioqu
zbIHgsFjH}^wN^)EKHnid(pfwL&g!ibjAJWh1WBY+Tx%if)q(Q0go2OTR5c9jhq;|6
z>6w|#4XX+Hc;L5fA{wPK)7IALZKpito6M9AAJp!VFv7M2w@^=__)^FMEXbCdMLonG
zgj(KoP35pvD~Fyl&}f;xmX<9988gl)IT6X_PBR7~YG{nn56-Wp@%J$J()6&%AzW8X
zlFh@z;+-fa-_VS2n)9rd%wI2nJlYl$jEYb(gTT;4Vr;sBhQ5TtcL~KYQYkRwAD6-k
zP_d{>DL4<OhxT@FIv1-$Oc_iv#gz!6-U>1_mHjhPua>Yj;!{P=piiq-s<a3_28i3A
z43+Ry>f0@RQC1+6j{Q8@VKSeR&)fgihr#5Sc1=YGzj-NjA*Qrf^oD?pb6-bi`nt*N
zDmpv)2XLw`HdCaV9|q35yeZink6)#Et0_mMkqkkuKwycRCq@5kiMD#`pF8ml2$lx!
zwKvM=fpe4-sh*NcCC5nI25iBbJD}r_3xpt%N@HI+`V5GjBGM$MMp7Y8H00=>^ldw7
zUW@Aa4(o!rQxIM!kVfYtUlc^T@Q?+G(?RL>-lX-a^~G|LOYkU@Z0~Ji&kBqQrf#>c
z)W7Yh!BJa5s(PN+C>diI;4usiP&Q#xs7YXfN<<?hK+`^F4>U(+v|-+d@Ikk`1$^S3
zu}Qin|6o^M{M6|j1r1?4Z>tgwRXBK#*BhvpJ0Nj1^0(4SV4<#;Ta&qXVQ>r<Hd*;1
z@MSTA&*5v7sN*I^p?}!1x)mc}U!eHf$G;e6KY3peO8WXaFmEw*u(Rx1At!@R$|R9B
z$|er-JEaY2lT5t+azvutefg>cYI3rS!iQM`V&BP`zAKOYcsr0&ZL33Wi8SRCoxxT;
zk1FRa&NIH@W$=ZA*J2^$NWDYcKW;rKUKH(9CZ?}!w?;GyUb@V7;F0Q8&3{VEzD|z4
z#`IcGi0w{9kyC^VARw};owtwBlO-VM)T<4a?OMSpAZT0OGU`6$p$j#HqQ}spWtY4k
z+-X;zYg5;eSH*2sG%imu^gh6d+S)-ZB#VX^H%IDFDAWt+<0V-?_g86xDqw(X2)~Ds
zYUp4uolUcqW=#{XCQH5ig+$YSRw_?9AUQ1lkfc=jmA0{JyRt=}=>0!ryqiPj<udnx
znk&-)?e{Z3?>!XA4rUfRebzDXRyv-~AHL1QwAKq;+xWd7_J=%|xh7yLUcC>zhlWOF
zSIT?1=>BdWyo4GNGs*hqG-&7G7~Y#gNA^CMK}lwPL>J>E47sADBjH>^43Mfx&?0Bx
zoV!Vj9AQ{oMoDtO#&)hT{{gLu!dKqn-F|q*BpR)dCL6%8(}C?+q;}+w-_{|Cm<Pt-
z?=zjNx>mw{5g?S+6hdAPAubS+1Xr+nfxBB*<BzpBc4ow_)1iWvW#I(*ihye6-+SYB
zDht6?b)d%inBt!)pH?EoBxl!8wY{#yc(EL-)MF~%*hSr@dWcJr@OoTxIc=a~(-+5F
zbQvCl{prJOP({kIK1c^IAVq2~Yr}WO<Gj$B&+6`IBydz%U)~wS?tX(qJFh;U78J}$
zqlU|FLSj1AA((9d`um;fmN&U}Z=x~!U|E1-m2&TFBVSivUmtMSg&@bJx5g_AL(Br!
zH%OMMbP^%zPq$ZWyYIAQ&ZNzIY-L@3Y!V5JIQ#p(Mb>^xm@Z0%Mv)BV+jeB8M3%&C
z353YXWIo9MuB{SUVz<MCi01OyIYHQ8%zJ(b(z&l@TEVdBw#EYEgs;UM?QSg=u@5!j
zT-REGr4vB~`$u^7Dwz&_EOh@#5w(38B*Ln&B=1UtX1s>+Vy!^NRiq%SoM}IQYv09r
zP`d9#C4-goJ>Gd3m#gH%bjo$wZ900hN;{oufa96mA>c$<GOL8QRu=W<D8p{mggM8U
zeYUeTb_Vtor0nnPo`4n(ee4Yo{;gWuos!YdfkUx^q*INHnK)985)|ZX-3-hbObYJT
zKQnGLZYclWGCfE15rd%x{YoR?piTd^us&<hp+5UF$EDga?LOy-c=DQRATB3kL-P~o
zmxhP%rXb&lPGv@yaEp;Isx<GPZhJCS4w7DA97y0w(104y<os`Kxfai18uxIH?fNez
z2oJjR1n5P!V+7T!5M<Z8EvjlM#2voCndKf92#~t45;1I-i<1Ne?XDm6{W(i^(TkL$
zAGn5070a>2I+ddgeCL*VaZgUFzDQ7c%UTYFjUk@^9_b*+G)ucyO7;u_R)|bgt5zUg
zD)cF|4PPQDkbLPxg)oF~n5tww@gA<TYHRR%ek!Y9tI?oI%@)j%p||g@tf`R~QK$xB
zu#h9I?dj3QZ}9&{rt;-8(_GNixcbOWq+THCc}fDXB1UIYeH@qDg$Jze`$tcJxFNZ6
zWh+1|?43sU&PLP=h?Yf@u)$Txf0~R@GDDzoEf%=+=!IESas+%Xgo6i}hz2m4&j(Eq
z@n<&IrK}GqN{GOnchJ@Ltr*|Jx^D+6dub6Ln&s-MMc}X03?uJ7-Os0%JiTr7^S>;1
z)}5WgKD~V;52D1&ak%{9cGQh{n?>uqPi`xACZgW>e3{0z>(Bul!IwSvYF7_Z(i8@q
z^Y71eYe6LJOWjnEWkdD1po?Ci7RJGOVB^<f6Gl;LEDQ6fK>S}IP?}Qu85?e~u!^3N
zdA3Aw=Of8?<_LVrKqt^gJ4r$2cI-z^OvhmZQT`&n2|Q^GAG;7o#o7`hL#`^omJ;u8
zH$Dze@ukA?K_+kF89mSNoAMQ{sB4?+wm=Z;igzdClY~BW#bxZG_PR(-zmKhQvSKKs
zU5f|c3%!4@k>PGZDexE33HbAG9ZyXVPU(Yh-$Y2u>rKJxOLpqr{T3icl4b69H12cJ
zqXpCO|LU+SGk4EcmV25?3On_Glg^(_*a8;o!npHZf)X-1)+6rWrnnY_=UYveW(>@M
zH_p>aG7;jf>0ir9Y-(Xa)L8sL?REL|9qXEsI6Qb}LHtIVA)fT+wnI5-o2W2O_vwS6
zAI~L2Paht=a04G1Nz&g71SqqM={F1vyEbBaoS}=*l4TctLJ0}@#@#5s>qP(22hFY!
zu$g5q8%}nNx#>L{FH1#^d>V=8NVzQ9d1_SqmqWI*^T`xA<X?mWWtk3~-v4;zSj@rk
zA1L9}ZgBqU9FewNBOZXKFCvn7q3+NE3UsLzPYUFbxArAC$82s_!*WY!Q(XHv<X-rI
z-?n=y9zR+0#zb}-^~y((=nejjnB8B7M!o}8;*Q|CZ;l@_yHpy=AP7uhzjx!!|ITf>
zPrJwSVs>hYM@g1o)#`$iP#8hOVcudvF~YobtEG9qcgj3q;|U8I&K3DnWyxDQ{s;rE
z`}Do{H@AQE4sZ_P1%3G6o{n=ER#o2J!%P$sMWJ#aUY;*#*h`B-3mafnu1rN<gWi9x
zR=831u%tjI{P=zzHK;!Yjd)Lr-+k3n0`{8Ud_2b1b)ISbyU7QQi97maiJfkJe9yzH
z0Z&YCIG-KR`;wt~SBNg&uiS#%l50JE*O-$8e@f?N`(#wwljiEkr3XF?OccD+eRH$5
zz{tDW!5rRGzgDr-bH3nR2N9RqUHDz9BWl3Pm9L%(=aMW#O*%DKdC1nDiScqx*U26v
z^XvWw+6IF;uH>X7ZrP=WQ8|zLn4g}!nyyBU<TzO-GUR?%dRjeG1*WfBGmQFHyQ0?8
zt$inD^xr;%+pA81!KSSmNozhr0gS0i!5@UubRzidc6{}b6=@yLKOH|g{$*iBodKT<
zD8oYfFgkl?IWVj_OIy?trvq%0R7>>Z^)b&x7@I-pS5MJ-qI8~?J$$o8Z&5aO)~wEy
zIww;r$DWR@iF%b=W8J$lX-o!$InhISQN6Pcm$j_QVLbTmX}6!O^X;g@&2OKktC8nz
z`ntrv2GkCuG}{~_`ef>J{6zbwfDrcCZ^{5K@)!*HNuHWd)RKPh1=-$BA`+@S0k&8R
zVhd1D+J1R;%r$57Vvd9hO4f3be?!#2(#%fxp5X1zpWEoGv3|5%BjLPWpd068TzPn}
zr~Cl0P^O;Wn}FM!J~A^)57efE(P$_bi84vK#^r<^&wHPl?1hj(1r}t1R82>4H-1`?
z+6N6pA@d}i%()+ZAYZ7bg2^8&CCWd^655d6Wl5Fu1$D6I??OgH_~VVv2L}x14+`6~
zc&c~Y66F5~^7}Jc_X=La&OZ{9^;s~%(}<;BrZ{jYaR5^1^#Q`{A|Aku==<Dy1g^>{
z5U_u_6rBO;vL~!#n~%I0AE}CyBTsu!dzrpjrZ)1XWg80@c<>gJg8!}aI~Sn$D3$Rt
zYR(9)92eCfl|tgq*4vRvG<Le@_GjBv;-M=}QFS%|tWzp|4t)OU|87LTpb#h#zeE)4
zP_D-HS++?8Hafq1S7NHb?G23%jTp&&KAAl0Is%tXU!3WnWW4!Qx{uzE7d3z@Kx$|V
ztmZdG5R2_R>P&^ih1&x*q&)Zw8=280qzt$99`g-Jd1wom6+6j7dP-a_AIPOVlKBPx
zNW4^b3&?A~z{>M>EQ=gsrBfrSKn3GFBtviw)pBuJH?9AVB0=b92-)%7=xY`^>Yc%&
zDL3}|yDwcYp|4eNLASxUBE7F{C!lz_fbmaxYJ~gGuDnsXGRm`^6i9R}UZud-E8p?{
zmaJ$x&K|~<q^E@*%T*|Q6t@`upo!LwH6t=O>1t{|8J}(+CSj)sd~in}IF&y6WUH6J
z=JKI~S7Xo$pBv_%7=4O=mnoHP&}>MAc#T7)lzUDUdVWV=q9|kV;?P7bmp+hW4g-_w
z_k3LRnFUHc<&c4ujs6p;u1jkGX1PL{asW`3i}ec(;u*;(X21Si8=vb77OYLp{S9Av
z@jxDWNdy}Qk9PE}>6_t8lGTui5Ndg*B|rB!tKq`C;RpP%AFY^jSt9SIAEV+xSrom;
zQN5x-FSJ=AezdAVimq)YC;Vm|w(zVjqUEUShhKc~oHFSd4)@;WJp}gB{(2%0{`jFX
z)Bo8oRXIfY8a)7L>jvXJFmcx53@}<XZsQzEs}H$0*4DN*R15kG;~q(~4Vjjj7)@sF
zI;gcwyuC}A>(KAWAPYrny)N)u2|UbZ6tgY=<pxGnTFPx&%*Y{sWtY6Ta>(kQJ!JV3
z8#0=#4@YhCaTU~2CTHF9c>AAr6KNPm^dk6+|H}M$L<z+EcPQxE)30BC34yRftP+AM
zbrn$MK<UFewlfpbsQW(dmK4uGNFJO^ok{oV*56J1$F#1!NB4QukY}9CYOPWavoCqO
zy9y(_%N6H!JxIlHRHQd)Y~f*osNGMXl2_5i8KG>liY9@gQS#Bo1F43Cos^c(FO|zP
z>0*(Nq~UJQ$xoPdMh3(k<fGh9KpA(Yz%K8CT9xzUnk`qJxW$)$QH<-ys>WgD5n|*t
zYnH_OK<0d*+2FHy?VDSolJLXyx1|xKoMcF@wrgXX9P1-_4Icb{&*;$yNBBuymd&?#
zziQ7JAtfaJjf?)Ql_z7jufN;rs4M6I)6vv>8B+bwxDxW#y%1E_a<E+U?TH*bdyOoD
zs<pgBj>Z-HfXSmLpdQ5SWsUN&$e*O)a1Fgpu44N!ED(vCCR=79BrR6MyKkf45DS@P
z<UC3?NHo%=$PrL3G$RG@6VZh0xlfZ@fq_5yKaQXCcb}8Yq}7ZU{8Rz#U136LNdyOe
zA<Os|MJBxW_5W--;_TQiQ5#X5-)3kH{K7yfIIFjb8rM!*sigJD1_z`zqP1VR>AaHn
zy~MG|OQVk(Tp75+j`0^AYA(=EU{XyRXic8}W!Lenlb;x`S5aZ7`Gx0bIN@H<Y1#k;
z`YVJPhjzTnciuwRG2CA}6B5;+{d1s)JPLu4@6MFHX}2$nr(hUBOAS+hxa%UENS4gC
z1CUobOuRLqvamf*p@Oc-)Bebbbb1otH};)(>njvgGRZFSZ6EL}M#TIxiqnuAl~>Dx
z27ZNUzmU360mFl86z2oKCpCLjDU0ai{&D?+r8GL0PSLDr#l&L6l(6i2*{ePTiPokd
z1C0UO(B^&(3t|N(xs-s|=N{$+I%|@<F<O_*TzCJ3{|ij`)m{^A4|=bq+?%r%*;b-O
zLeJBo3ud#SgB3wDB>j0xA%8ZqBwcjwv?g##7SExJe`Tds5Fw~>yQQ;06BDRK{Rg52
zxpM>mi=^eP3#Q;4ziFo?;+*z*pC*o>@WB=)LRfUOs5PM=!5$ALv^{3)fuM<ARmJ#)
zU-1kS%PiS*JwMb~KtD&2pf)@<W|j5TTU?YIJ@2heAZnpRV0Q&G|0yNRwlT&;N4`UD
z{^v%Ll?`^Q=h`l)-Ay(?1Efx0nqP+9)LokNg=9X@m3p?Z|A4sj#8<ZLr+fq?mDrm|
zPqt#wcZPL<k(Ch2lT~8@Rm@<tFE3ksuR|@E_3c~rt;-@!!W8BkX2ptAGD`*>tX<eq
znqIEFX-K)@&KtAkgtW}im3-0;O|$>*dD(^TKVJ$v`Mz-9zj>9e8pE4_hI%alE=gI_
z_}OyrLfA&=%{>urd$m-udn3AHZSPWqxro4Iow4NW_j*jjN0$%A@@YFX6e#i6?rX<F
z6yIdP?`*#-cblyARA~)(68vnfk+6}qwAN4~DKf<-^-V0RccAv)@$@d5vP1Y6Y>#7B
z)JQr7IXJY+6>%UY^^c$m$CwyixXM?D+T^J4UMnjwfAw-KH*mRxH*rEO5#H18$;MC&
z8o&?oKtj69PRh#>&%i^m%F%N|D@iT62O3VENKI47ukg(mMMCI(b5uU*6)2(A-_TT^
zVyj5x1JUy+P+}ao#>6yLkeiGKB7g0fL%n9AWJuE|s6F;m*$wX7x;hc(_#iDK4gFNy
zjnw@NpMZ1|qdv`sggl{0<)QS9nLv5}Gm^_7YCgyZ`l`naqVWXve1D@LP2BerDuyP7
zzfEerxyNB&`h%KjFO#5kvwLwrh$Yyr;t-!4WjKDC){z@BRPD-PN4Q9f*;;Yu+d7b+
zHdHa5JVLxo+W#3g-jaN+GI)|RuQ}_%ZRdl_-Hy<PL~h|3nMq4bx;e7b1aExK#FIKD
zoP@Ks2MQ*0>(rqBBO-%ufIjb3JCCFjJbWeTy?XVc|I2}iAtoT@qR%I2U;2r)VUA*_
zMPK9Zniem~u^-U?P=2cH<nFgR-#z@2RTWb34UA@hMhwJ?BxCiF+k4hUj2NK~lT#gF
z{PAhR#ZRdgzX|~zc*aEv1pcU-;R4?%;cM*Mk4Nx%<5lSTd=qaUg2sR~3!$k>5~dis
zCx_fx31rrOd%m-X1e6_^DuVkk4LpUgT){Z%SG?>Hu!Pn*oW3&CHIXOf-WU=0i#UN&
zuZC-r^HC3-6?3Nr_X2UilFahB05%74_1DJZ>DSv9^?6@M%SBoao|yLB<3Yk!OJRn<
zaAqn}kgd?fdl(<P+7aK02ayw)PaC8QU&2L(zh$p+NrHs3f%dhl`N;~ZBmIFkOi1~(
zn~%f=<+f+=`@Mm_0o8S8^c{|IAzQ9p_o&0oNws{NrP}t$2eJ39?PEX@MY}-f!`a-3
zMN{|S(Z7yx>0jtHJ%u+{(QxEDJ?P+aq@r}Pdxoj>DjdDrSg80~2+bEmv<lXvehI^R
zctxW2u`bN}_z;|`UML1tQc~gP5e$r2x<h2&!%qTgd$j5#JJRSyZXsAuL~*m_M^@U-
zai|fm&?6V65Pfzeh9!P&y2EU*@Q1v8ATRntPlXd+VD(JGN#q{UDpkxTzUWpO87*@=
zs^#9u4I!Ol|K@-msz2ULq-~%%PlaP(!tU>ve9SXWvQ~Vh-yLkTmt$4qnQv1^9;zhw
zGi}qNB2CGQD0dCnv2f3Hx_2ob-f2n%X?trc&7vR@!B-&9P=1vsO`U?|7<t2Seg_=@
z(MaY9T7^o>>&%vF-6aaYQryQSC?{;9?x@@%Ll^Nmg&iZZy-|Dpkz2~hjEp=w6v4a9
z!Uu_H&@RNUjQ7IP#B;?*3~n+^tfY7Oc0)wT(G(YjFF1@ZNr?=f)}j>s5KeD9as?~1
zqZN?eyRM~O%J?eczgBJSJE9{SXL_*JF<6t?vvl|`!!*G)d=Sm4S=@#~+N#Kl|F$7j
zE@p=oc0oUgNACnTsr#`}xuE9O|NEC&dlIW+<i_`xo97nNYwf{r6Ek9nJcRENm=#KX
z@au2Z5DUn9EM=*l;iRQ%-X&YEY^`@9RhZULsnM}7$+2UxaJDJ&?TCE&;~MJ|<&Xby
zFTIe9OtNYIqIc^c%)WR@bc;fBtv8r+o}*bstxRiKE&Aew6Nc0ax1j&qP$O&tCHn}0
z=|Og+{=I#{GO3kP{>LT#aF-jMO;=m8hY@(|X94G$bF!oeU|co*sLyUn7L^l`(UliO
z$WO~4f83oDf9!cxy+y{M+Aoh6JttpoK`4lLuXjc>PlNVe+3uSI`kN~yQ19Q}Lf^qK
zB4G$R75VeuFXsNRXdBMc=B5`rS~fgE;aY5e9wH%z{uS1Zm1P?hd6STLMr2wrL@^0E
z+31_+(NW+TR0LgNC;P0^OnSVVOFn{A3?AOZ;20(*VpI(sb-f&mqsipy>rEH6{_s{Q
zF3xh&HJJ9%1!h(Gg=i$BU8uy$JF;}VgfAs*H3u#z7NL&lnkpmITmf^Uj1O{>Slf&a
zA@h88n2!fK+ASschU$_zA8e9z?13(1=^pHsR0_l+G(ih8r0gtte|jy^>YfcHB<XC?
zJNgCkum<;<U13NCe`esX&^@#~GBP5G7nS@ca4%~UiW=m9<6adGz=?H2hI#gmQum9K
z>8=t?I{e9<8Ms<I!Dtwbrp`rc+KDpEQgQ2D%5QLI%=XHY3Akj+JBX3|<rXPomVa7Y
z`T0ME-=JMOd`~N8by)!+@$qlz*HLty%70{4=^nd(B{;;M!rzf}qC5tFLC7|Wk=Htp
zRr|XVcAeCpa)gE|0&T|{D4M5&iRcG`*SJm<E^1xc5IMkD=2JuUDO*2#;Db)J9v=qC
zB4XEqiw?Iew%$>VE>-NB#VMqoN_0iUM^v?G%wo0!e_)<>Xs(e9dyLEdq6!8Ag5Lnp
zs`(ki%_G-Y@^$pHS4xvcTB)|18tnKHJ*cB3(r}ITx>U<fF4TLqU7~9610*P5EJDpH
zV4%1keF!AaJs)I0-f1a+qtv6ya(;pl1GNp~I}XEbg={oqleb1bDA*9i?$JR~@ecRc
zU*9#~qiq6LR2tb_$bk742IGRcB77}+9j^Wppep&v6E`bK8p^(X*Mo6w1FwkJznNg2
z+MrfstO>b)q#78QQNR~Vhg{SZ;9<m0rV#LBzw<)=PI`~N{ytBnFlQtd64U6jXcowk
z17vb*IOT}DudbCKW63_9((+w^s>w-~53z3LPDdAgCZT=e&nxqb_K4#P<q5YHQInWh
z(<)2=X2cT}w2~M~T>C2OJ&GQUzRERocVlHaS}rc}i2qA+yaj4itnJYFGx*iCQ3g#t
zi1TQ^-LwBWk?xBDN1F}n!#kJC)uMDt4q11eGa`H(k9>l)_+Zi_QHITd$*&F|C)|Iz
z^$P7};eWZ?PbQ&cNJE_P-5Vp9+7KNvh94Y}B%Nhg+J}t=uvI>YGw#PY4!U!@KQA!O
z<9WzMUIoe1t*N`c_GI*3Ba9afP;q)WU|WO<V!_1jA&@cBMmFTqr@V&0F7LKnUwZ#6
zgw`K8{P?z?xxS9nNP@s?(V}S>QgAGh62o~2;?eW8(xxm}r>oG&|8aB{3{iD!6h6ZY
z-5}kKB7$^xNhyLz3P^V=%}664AtDVD3MeW4(JCbk(%mT_opblxA25AppS{<+*0Zoe
zaSklvHMd$-;%M(pI)ePo&*%FJ?eXp%PXxuTToX#C5BXW8*hwtSQd@~r^rnJ|`@26p
zf3d}TrMm4KUIdJZ#A|d^;r<<lFt?0VTt{$3Ei6{HpqMK4gA2!U*CyPjOTAI;>`F3Q
zU2oATlU|NuWA%&6vVxoarfmzmJOQrdFEl))+5XTVjS)}#JX~Zl-vbehb*?lSEB*A0
z<eu6P#v<ES;0}6(G*`1Y+NI4dMSkl|gy@HE_!wckreCA$8DL>#!pYyl+`0~yHeU}b
zQ#}IdE4K6+@!|hoyYjB3d@+gJWf}-&k641KiBvT|+0O^Pm&uz~s*R>FT^1kRgPZvK
z6(rqS_-!S(#KJamA4W!`Yumu^5IzY@$Vy>pmUcwWbu05n;=Sf)6_A^%1)8{yv>Nay
z@&Z%My$LYak;7!3e_hcvBJaxM{0XrNs*tx<)vBW$K}65HDIx0qaL8$|A1wLg+jOf0
z0Fqc9zR>cJ#t*?EL&B9<!4wCEiL9NCRK&drX1y-dqdv_98j%rZ6emosh9#l7sNOA7
zi4LjIb^txAHTP>CxWX9-?C#zhQ)Ay}yOPNdM-1Yz{<Fg0$(^fyt+VtisV+3GXGUfP
zOHG~`S1y{apsk_<MB=8gxOr}8S{M;Xb|?{!1KoqfG=DJ21<uW7V&f6H+$X!Y-)MyR
zW`K8n8)PFIOocC}jH!WCz$Kv`{1K*FA8lZGM$(;=;a#J2@n(Y{Ofo@Ksbyn7orXc~
zg$2t|@X`ZFb=?&ac;VYiw3Aw@V;<we2@QP&a;YXJlvcl&A0a4tRM$yS=y!##;_!+s
zZ-T+kRPQ5n7Zz9P+OAFGM{c#0qak-*MdzAN^sdGdiA~uCcv$8~>{1+yM%*UE4DNDp
zNb>aFvy%E$h~yOY2)cd_wc#0CO&nMKiiIMXs-%*0c0&gRTz%eT^OitM1nj31n+(fC
z$hXvKv?}7m-@DG|fh%hD*}=!x&)n`~2F-ry&Z|_zmgOVTPIzv>n!g5%lp=XRU!c#y
z#G}+U{Wn4#TAXA0dxbMW{sGrS<NoJ2Y}_&-Qp>;A<c7aOxyq{|hDDBKO~-m=n4GW!
zMbbSVwYR?sh!IGI@@~X~*wqgcdxL39%Qqkrlv3tl3P5I93J4`lkHzpGQnM)o6yE2O
za0(3RQ*v~Vjzk)AVP)gjNyWRjI0?+gBxM#nj43PT^dn>yK8&`C_;=|ogHLI5XQcu!
zy-Y6=maLhu`qyq+?#@|C;q9^#2*~P1NGhVp>F2!N1&E$J%_iex40;e0O+b>`OYW5^
zkI!R;j=Y*K;N{$oU^IVMfASam_wIwJS3_PlyWMZm?!vij(cs&F2}>_@xEvi)yIKb`
z`LcI|@>DvB>}{8SBaNhSCWLQ*i`*LQZvO$}aB|~*`?Lo6Ff62%Zj{%5r?&K|YE@tF
zSyLW{HgIMo?~?lRB26LbN1}c26B5GBJ~0+(KV%%QdFjvGdo`dqZI(-{x>#5Hy*rz2
z%gBIoNxl{ZqFD|p4*NPAB0uFi#ogw9;P1rA<COjPSAxPy<v$MdR&8Is6CZ7zvaSv8
zLE)SIU=$BpO7-G!x1Sh&kjlzlLeAQbtCUuh2qi1PLRcW2CvMzqtZW!R16;W^QZge?
zQy*Hcg6A7i?6|B^eVS_B`Wite)~+mhzT#)tVHp>-JGiH^w9k_qmb(WGMejK@Qo6wb
z@-8@@dxTdb<US9Pd3?XEsfI|^*!3x47vP+*T07&S>l|fzOa7Idya%hPu3bCIfVS;A
z#}df_YKy37GNkXgjk@ETul7zdLpXAuY}U~S9G9T)?Xu~SB7WN>4^<pb>2sK*H%Zx4
z&{Q~~^BIb|(+#3i&*^bW8RFUl)8Lzvf56|PA5~A-&ZNuoX;#$-#WIF}BBR8y3iiGK
zu!8M^sLw62#X{XF|NHZ@a9XK63V}qlbs@L~h3CA@NSEKc#dJm`_1$jd1dn6SINzPa
z;-bIY{FAruELdxYF~vz>$2x~mA<H**{r_>=y{NIwurEzH3lGdgIv5Q3r}2idKHpLB
z9uIxq)#Q5=Qeq_5MGVw|XITVGYGzKHcs`Ha^cU~HQq(+LC~(IEFI|U)%x_PJFdMf>
zlHEoR<)UUo<94pVOtXm4sp1*b_b*gRZ|qZ?cSXgWchzLGSk=Iq+6#K?%OcI$mF!Xw
za_6srhPYi*^C#R6YG(q$caJc2vd^*#G!v#R0S)@nyWZC8d=Iy<qP}-HOHxum4-l-d
zgHa5%hln;)ObC%H#VXuk$`^RLJC*L+C|;j6#osxfaHIO4-kw*EBaoE!qR6U70ScsE
zX`QLZ0X_{I($8HZ*~+zBXa$0=1_-$-1j!yqD|RiWK$E*nXQZ9V;v-xSWpR;L3xS2A
zyispl$&l>05?8{|-3J1!^hhaCV>6ozLtfGnEI3!l_fV|_4No&}?q79cuzkyo^V(wa
z+(7%s%X;IyRQ@)df+nLzZCa(H^t`-~UUmM$0DozIn1I6=_Z}5?WBl%dyqlQ&vQif7
zVydeXd_U3nOLz4hJ!s@W<){N(-d^xAx*u!p7@a+eELBrNy|ut5`x$gxeh&+DHr)RW
zIDOCRDfYhgW8ofueJFs<gjB<YfW7<MtC507I$xEh7V#A#b_kea)v#6lZ;WajC`Qt4
zu5Q|A7Qrl-SnQuYik>Y2pYr}<DEA=?f<O~`+C+gQh{r-+$l}W3Vvwl8ncq=DMDk?F
zZfrU`h(DN6kZwaQ!*#JM7{5leb3B3OxgMDQ{7xBZ4Ms(gk(gl9`wdAT{^Q^?LCU}j
zPwYN{kTmcwFC^@<nS4R(3p45iG(lI4;+XNw3FtB#{fpF_E}0=z9W7FB{~0bJL<7!O
zhgz1#kUi^zbch)$g52gDz0{zlKdy~ioc-OqJ{d=??Kw*~K44tA24tRZL;Ey_+rVLP
z=P*gkZ`4i9>uV6vnSV&!Yqy^<Sd|@~5h-{$kccqWj-olrKkCxl!HP${B__d*>CGhC
zUafv7>bw}LXXsJEWS;v90|7JzH;ih0Yi^8KMi5u#d~WLf;naHXyij_Q<vdh^fz8;C
zPEYAH-XFGx0%D3p0W-zh$06Q@U%Al`OJS;E2RPEi8%XvhKQZ%q(0hAfSgqen6cGjc
ztm>fj89i?cM~e{jR~cEHWN*|QU>2^8{3HQec7tGxT6d&LsrhXO=r1N(Z#h7Z)`T2V
z%az2@q7R?OSQ)o=cS0$`Z{!h|@|t4^l>0{~u?<Tq7Ary`pXokqbGG3-+HGp&i}Nd?
z`<9M`sv4gAuPYES%DG81G$OaZ&>!T>T6+&_kA->AG2<f5n2^VMX9!4?o;Yd~AHLLm
zdOG9!6^AfFQd{x!TfEzPrp0IRynZ;G!xg=8m~8hF671|r+~O)5&x`WkC4sMIn|9-$
zgT^8yt|ePFRi=vv!R;ZY6t+uF_*nAds}>w%$O(4tnbVQUuk$b+I(At4f!tj<ZR1go
zv1k>+7zGL`Lrr<)Re4@$vTguHY0ec`W>ZjCttg&}tG%Z9A%QB#0G@v`tP15s%LEl0
z?s8mmN3yXPl)gza00-cRFn_J`rM~_Ugk(bVUNZW9Z7$k`&)A?0^g|bkk8;#(QKz=y
z;@U!Wgq;c}UB0EIqc4jlrq5)SJ#x-52bRrCEbOyQAq3b!;mJ9WyjU7HN{%87lv6fJ
zWM$3@y4=#anFiZzT>D?nVmjHv&1((wAh)8+!qi0BAzx2@zOulFU~dzKCEV!*#|gNo
z61Hb-{hW+4`)24$<y`pA0(x3Fvz@mkgM&e`8`0zj=ZR<e8CyF^_HzY2sf><Mb2Vmv
z$;1q~kX5mEaWl(fEG`gFT6eZ3S$Nnoyb`$?H+KD?NF@GqN0p@XK6Sh79or^1jN9u^
zKF5)v2qBkQgTKI7feyV*v_-0EFWwLfC1ovd(6FjO>htW%_vy&~VvvjzfaUv-$AI{T
zp4tp-=8#!RNrXPG7#_SiS(zWiL~(KwFSd-+0B9u5f2fj>tB8`yc0DWIFkr?;4&j5{
z{v*cgH{S;&G|F0U{U{2RB>_%?<$&qcq(vk17uGU)^vD}DJ5yZBFfXF^T!VE(xl{R`
z{lwr_c&SdkI}RAf06Llwf&zg6<{y#76U+vr@?AmAfRgmvy^G;Fi~Rmi^2d4@Plxxh
zm7>t0Srd;Q4i1&;s8!zwi8ueTLCWufAaXH){=*SCC#Drr11&wc42m)O3$4}rUqIcB
z2MERc^+&SaG*6b}!&pBv9#pcz%Wi@k#)|%*Z$}d#*<qPD9yRB@K2xFqA;s=}5J|`z
zu$<`IrJnHShQD`_Iy%NmxWktLa4=~>hnt%wu0Um^1xU23N9~@elgR=9k-Re|K`ns>
zd(eSu@YI8Z_Rp;bo5C~ugb{L+ZldOG$f&uG1B19!Ej;B-0(toH3Jg6;%B$#L9hmhr
zhlzv^IxO3w+Mhw-MBp(`Bc(?sc_7)<uM$iBiBIh2SJL!sQ2dkSl`BQzwxHmpM%gTv
z7&KOiLxQ3(&`lA8pVV!_^mUKmpj5sto`~3iuk}_AGtABU&XR~6Ni^O0A*n(gkW_R;
zN%9sLV)AWHul4{6!GWB)?|Vqmzku>_o&~HS^cbEo@K{G!ri6Ph)R-(GE2~sR-e><H
zP*Rmc{9bD6pcDImFL=gNNDU*2Yjeg41K)N)`pks)wQ+ui0imBZFm9n)8@!>G)&jG`
z*Ea!gssd-8*R2GhLwTn+m8YWP-|aPwey)%M#v=Aj6k(E}?ZEME*85qK=tFuyMq4Yn
zro!@E8u8sCV{xo_LnKE@_yI*@oB8Q)(CPllo875g!ha#^kO>B;91Tp94GxNCeoh<h
zj-~kZSoU8xoAu_<r+Y!wQ{GZmLjz}&%W`&<2j7u*w7}@11#tGtUkC6bqNDmLp}&u)
zJsYIAVis(tVTsUrayUp*-{<Hw(_h@RmL&FfERMNPn(w@sFX<n1cq}u+JO`+SH9g^~
zZT0ReXl_`78K9_x&4ccGlaRnQSm_+brQ8<f#MhE_oAQGyO<=fsgl+iyhk#X(zxp?n
zSSA4teg$D;E3ib&pG`l)nzVe!#C(qJ=}IYZ@E$tnf+@>0JW$$AJ`X_6i{^XY(mJ@~
zYzz%XWc>w@;C)lR9Re(}FB)%m2O)t_Dh%1)g(56qyJSL7bzRrskz9VO^^5=omjd9+
zjum0OvZQ5^u21$y3S(Nv%gy-yznjmIF-^H|$O<wLDS*3nNqByn?JfEfxys^q*7P4<
z+pSeRAsj4Jm#FYs#P;%&c34k~t`)|AM}xBuUi>>d!*L=BvtTEP@lRNK8H*mjdVm=b
z56rD>OzTcf{rBlNW^DC}gOF+HT;{ryb8gmJhVn1}MPe_lCmizrE5JlIu^9ih4S;a1
zhVzuwiC+Dky^9X7&;9or+ggxLjc0XVhC_#27KW3YKyl~?W&@4#7#|`sS?|)PvqSCJ
zKDHwIM&@TXf^?Z~{aJihwjwdVG^JkPoOk~N!=(eoSMjCp+^)}ypB5^FWVPl9j-16G
z8O_KcmH`#9s%6k_Zl+~|gLdNQg||CjwbRs7GiEiM%hsFI4D?NTh7*7)ztrM_S=whS
zW-cz(Q^0p=XlR({^*Xigq4;K>rSw~??QkN<yn<|k7ZOeXcBd9h`g#@QNobK(L^@cP
zBZQ!xrtS^>1R3!|8P)_gwUzL14AOoE%`dkIgW?&E&zt+V_kv><*k<bZ0v>S5bMp)-
zOKc|V8%C-8VtI0EnE0Nz(#<sU4EC31*33tq1n*PsoKR$2{5UQKswMC$nE;8iIB_0D
zb?XGM8JE+QmXS$VO|?!_d97Uwp@24In1n1Rjp9CcbS1q?OiL@~JKG5O#Q&NN#@$XB
zBPAL!ll1<(SLQCPwOVsNt%eQ9(j6Z|Jj|NAeHk;K!DmXbH10XjDw@-z$}0Ph5BGUT
z>u3V_<vT8JlzE24vo_s^W<r@3B-PZbxgNbNera7!ScV&tpS6IU_+28xE}*rLlzqVg
zSB<l~2r)Ub*>~-n(50$DJ7fg0OakQIH3$ioNXmu7m_z|0E)1shBZ)ZRcp?NRV)9YQ
zV;DjM`iCFWWGjOUEjeSXV#`A$eMvevo32)P!8V}+AtC}^$qPFAzv!juk4ZL%LsgMo
z7H@YVQR6`b9MlrVxYn5UZ>d+l!w8skp56kV{{M3-p-i&DsM{OwLjE$1D)B#4IaJN3
zwmE_NX!s3bvfZ#b8GIDGUS+uDfs3GX@L5I{fd^gefw)}IE<OVGhUpKS1YXUUt91_c
z!PbC7e(6ZUp}`vX=#8r~n;{SM<}Xz;s{c<uP68s=%C$%>?UO$S(uspc7#f(riy<8%
z8RqT7vbaK`suHisG)4bHtPftJnOV?Ly1Z~^d4FK76>N3UeWE=e9nT7cq?6klus*YZ
z`oHK||NZ8Xa<kSVY)LZ=DqaJ=H-OaH&G34b1#MKu5{@Zvxm3sWrM6TGFUqSU#4>(#
z7nl}mv*dyf@9jx^GDT9AU=<PMDnO<fWpm5FVypqc5=|20K=e?b^R}xT@a;)l`V-IU
z`#pk^qF~%?eD~dixkDBsAzL2k_NQbtPO_YZP$<=%V^CqRu`+nrA&kq>HUoc{h(i%H
z-t(U8H+qyx0WGK_h`drq78x<*p*L^6D!^!4HHN@|4`x+|x;<-M1Gr^L^FE@feGnlp
zN6={m-t${?6PvkTxW#aN!#IH$e7`S<`LY7klb5oyob#Erm;`~0e(3{XoPYckrw6MV
zKhY{mBa3R_;?6kuuc@a*{G6q}Sl0L6W`}w@YAXq_V1&JX!Nex^@-n3GfZTr~>U9P~
zp^NcXzuq<lln<hJ{yX7I-<f*m>-IN5)}3!bU%w!fx9kUS#FULU4X8(-r_!`{F|DM7
z{Ob#aI<@OWi{RC^M+E#!Ayg?0?Mo!3uvuGXeGm-}KAGm!Rc?EWG@T1xZ(OCpNa#I(
zjPZn#5Hhf>s)n4x;eqOO*L$D#_=dRt6TVvp>;2x{7qAx*_}#+zt{8~%|H)t;36V@)
zU^Z5&KzcB|zeAm<M2F!LmB`9+>IwFeU;Y8Sn@@1i!i(AyF1aZ>PHC=8w~%d>B=`Ir
zSuTy>X)1O$uu@J?8NmT|O8==Tm*mCDI41w&{f0Xdu=mjJ1SChDzGDx&(@+E<-8pC8
zgk_%qm+?JdB;d^=frB0_Q95qA&fxg$zhDcJR~9&=9ws<BHHax7VADsv*YJaI=*&L?
ze-WI_?}K=8F<FRO0P&{9ia#>9-|XNFvv;Q3Y1o5ReWK04^cRr-(Q^sSlG*|esFv06
zQrdWSTp6GFxiJ)rGQr-pgl8HXEH+{!Xd>Ed9J_m;yihvr_O-$-(tM?uv@0FvbOsos
z(Q4y2z|?ooo(|w5u!6rDyp|~9HK~7MOo;|oVa=-{5=7^H6upUE_8IiJcSjiWc8_p$
z?-<;M-C*_HxrXUQ9eBX?e9zMKUGK%AFOYNSPP8Jz;MBsI;-LNfG1uEmrl5af-HCIY
zGxFqp-W}8hbk<d+7;=KpL*~ndrTQ|Fb4%hjMPg}o)YSs2X&(1n@=IveXgL()kqv4%
z37eL#Wh&md0t4Z_r=EX#aJ(x1+uP&haiYNd=>SkLk1oDhY~yTH)(gQfU}_eGNJQ=`
zb?$$7Na>+P$`nl-xmr9VoE}@;Z&qoGo%9IJ*CpOw$diq~{~$ARWwyc78iWU|X+<vH
zqR_-2P_6K6@*WAbHN04ku0v~U0fTa2D3iom_C0f4QQsS-Mvmvv|1v*WV+E7iRn1nH
ze#1!sXYe){L*6y>x*|)qx`?Qpzndi`O|ywuMkU@NQxRm&z(?>7gF$m3^cua>%mLRK
z<Q-6ioFeM8o=u+HTgd+tDic`$nJD+Nt2)$J@}<8tAj+1cWtAq8+`@))W_B{{kdQ!v
z;tEVf&d*f<#RWSUhE^#k)c6t5H$fUfljO)m_uM*7USu$wIk}9b<OltK#vayRu+529
z^|`6I6+4=IDvoYaDdol)-{B#=2=i6r^~np&t8~BZ{&HAWnpUI64a7FVV7Z3(Xr$~e
z;7XafT%mr!_<nsD%0MvVYCz%gQc_Q{8}}qRmfPVUu0SippDrmHVn)$+DtQ7+r*{2E
z=5zrd5?bugBE{ZjbJX5(IUY$CP}%}5h=!2&1;{zf>Ir2|Y=Ia(3k>IgVA)S$GFrHE
zsf79$-L3~e)yzBv&zaM+){lB3z^6#$22Wnpn>WwtF~P_YaZU$>$L#jjWVwjG)snAe
z{~TZ#fA7Dh-{sEu;f1U5@_xIAe^!>~@~5KB#L|@=tP#TGQ}4qY1w}XDNhlc%u6V{8
z%bx8u@@S1yHVI+burv~@M*rXY#$sraqRq4RTM7UZB&j@ogt2H0s}K#($j4hIuXpEI
zF*2tsA~9hMzIJ!UWOu|O!;$toP5|V`k{3U<_-0-h$9gw|&Q6bh;fbPOUp>tNm>=f6
zS2GWEuE-OxLh<7(a4-)|c->~*Fg|?y3QYQX0sG#>_TY49(YNt$;|{19+>aDHxJ6q3
z138){X9^A;TvakkO^#QJ0}uZA4dbOEa>F40IVmN_Q;Lroqb9TTU_}X>)@ifi(UstJ
zeb^EKk1mB!G3sUY%Xs2o0A#wrbvpfWjLgzE_-bb1|CAD9pD(3Vg69&v5zgxbqf8lZ
zuxI{@EF$1#On^zsZ0<Wsv@%6eTP(%wQ%%`(M+i*<P?r6dLoB=s3$iOR{^_(`TKXvH
zQ<sO?ruE`$*3y6BUAGw@KN-+hbtOj(xK2N-8WqsHV=&$Sl1X9G1ppOwzkjpZ)t%1q
zovN~yIDG{I&K0&h8#0x~<PYCb#Txv^(&xTd`WLYUZDQ>9S|{KyNaLqoMb16I0YxsF
zfeIHbi8uLG-+?prMp!9M%t_uS%Z&U?#_yV2*>Y?jWQE4(Ti^57KdfpGj?BX8c0l8K
z{*aPjw}{rqGv6W9H%N=J9R$c#JzMh4*@stTH64=9qv~nRzBPJ}29Iko!}QxSAvw&&
z^4YnnuguRKY^**s5)aK=iG@83PLJh<f=eMi*pu2{9%PL3*e`)I&*L|5-6oz1_wL+r
zU6!<W)CDuq&n04a7Ms;iymGjAl%QJz604U=<Yhr?G%Mftd1SbR7`ufIg^g<*3RmK~
z>kpbk`;cyJWmJN<uV(pE=bLX_<?X<a{{T(E<7p0K|FlG0hC6R5CZb@O*yk*I<s^sL
zjIf(7`P&%y-|;uwP4G|Op~za_dRtJx`2o!6M2lyk-;I>rmy2T9Oungg>?{9Ls!E`9
z9}sKZX^+#Skq)Y(sro}SnM>fw)D#d})cQ?gY<I|TUezC4|CfuPZ=HM(7QYY2tlJO5
z#tS}LI^2~Lnjg~eyc>{-JxCLnK-iYk0T`W>0qSqxU}8>f`TO0m<6O>`Fj}M)HHMl_
zWRdXIt7K|ATr2sl&cDJmRpe4&f|nQOXv{hWPnN21D@UUk{a3YLlJxmx5%$*d?Kj<y
z&n#e#yG@>7Ud;~V@QjsjHObx{#s_Z_VbE_9+!k7bRPjH6aN?M&${9i4`?iPr=*@c?
z54pTYyHiyh(vvVk5az+rI8)E$TLPNBy`w8hahc`XV*>YYe1NQ=*Q|X`?2}3ll2!hz
zA@~n)?#|R0HVaN(US35?@+Y4*DBh$Gw48uX_$?cIa=mlWOrjy;(+Jp6GS!-rMnkiL
zE(oN^WU!?Q)PNI($Q`4V`$R*WRu>p3S5qWiu!x9ayfb%j`tj~Cc7)$GxSKEAncwAb
z#lc_#rWth8O0%J>uiyNk*Z$NiOqt~kRav@HAuQC$)$Ncc0h7$EW6{q=z8Azu5F<N_
z_eMErB0b!-#Bt>?7WYZQFN53P5l-_cC^J7v!{I0bqy2xMj>Rh~78p52hA@Zveg&V-
z088-6jp_wmDa?N_j{=eP9zJ~ScP=iUS($xa*~O@<!x9?26-LXGC||_&IJ+-q8IGgM
z3v=T+<pj55@3+1D#Irg6ue)oTGLaI{p%1Xe5i(6-aT`|soSb&9o8tt09u{2gLI&hw
zWFHZTUA#XpTaUA1K0P6d@>;AjBYuzIwq`r(8thv%J7p~dyZ4m9;JbmQUQZCYQO4kw
zZ}J!s#Ut7V&fiaQ4Djat_oestvA)A)X!b(nlV1Su&K6*Z&+E2uN-p7b49LaK-&Nm#
zWo%LZ6{E#$@PTT_x&-&_T$!uZ_@T(5(*;}y&*mk2S{bQ|xw}~<1~WaJ;3E<Al@;Ig
z+x;2}hy-b6j87kHNi&YcHX7bL)9)~A#SX){bZh-_X>7vh^g8r{I1>uo7q7!oXvvem
z+Q4Z_-D=&;iz(3kqFr(yTPL(Q_PhGZmAJx9I&&A;%zXsEs1e%a-hU~+O;vwPEXb95
z({__cA$0c{oP+=^G-K?+NA%@NH2%}asb}6~h2hNpKp3|IWJvP|i18Nt-H&!f7&%78
zkMQ_gP>_ylZ9amWzDmrXDzhGShmc@ch<*!-C01j$s(@qSRre{psrs~Qm^&n(e%Jo)
zk(s)E%hPFvsVBXhee71mRcG_R2i$5)Q5OtbQFIo@s<l?XPmiAbdB>~q{4Cpl{-kMe
zCoC`u`YO;-JZy37AJiqmF)h2EYMGh;<|W%Z9wM3!$>Rk!@jibAv9H@V_dJ#IQ0uhs
zEjGc5BfgR8sR;z{IolYS#}lpp{IM`qloIPt1#Um96Cqzi3LLeV;?T3azE>FEW}D`B
zr>mTwB*)O2Cr6hk+q5X+g<&6jHM{A}`T`^><`u^CPFBZlTn}nH&Sh{E@wEX$oc3Uj
z2)|O;kU>%J`Q^lqyCC)B(arK1s1yT3ueU~iCF<u-L~7kRQPHDs5CNK`$BH!#mJra<
zH`<g5pD6FlNYaS;T2c8OZ!Xy|C0HPLxqmQa3oHIgOd3Mdb(Jkgh%s34=^8)&jHQM=
zpb{t%g++HmU?{#oKl@oX&qnXb+UGbT2w$ft{Z3Sbw3X=74%=+s&m(Qmyl~Fb#fv=|
z3kEg@bM~?LKJCTR-@of0D=5&{5<667twFVAd7g;JDXDxBUOwDCr<y)|X36k+Z~BjP
z(I2+M-=Sw#Li`i)l#lmGDh`*25+l~66F532ulC)oo-RBtn%1%TjHRF>7J$cDGVm&~
z`Q={WjOuJ|UC8-#hV=}TL7|3FBa)+VKw3d6Mh)Fh%!ekAlV7UmOta^KNTB3gzdty9
zxGKLQzsEUvD4oSAg$DGU2@2dJ;pxFjD8p7-6oXNFlxyjvy)`dMH`~6d%=TR*0NkCI
z*B9FiMDv3@m4`vaNs}L%bDsF_+&_!R$>UC~=AA6wX}TBvx)T~**7{m!g<OO5ddZw$
zjQxx?YaUCU?m?E0XkrRQYeR$ZNl<xGBYmc+_TnTk-;$ds=Lq0%o^UefK9*H4S)osI
zFS)mqtbUFEu%mdNs`-J^<<@j*K*@A#1ncdAe~6K6Clis1A|Hp1V@|$5ryY@8?3s|w
za{8Lh)MxKkL~^>;Q0z3s)55anVDEyh+-FcXa(~8vwg6L3NMn2Si<Bc7hv+du$il&A
z?X2xa4D*~ANzz*T%j7_>QoYk6&HrR|H-4Z38_$lg7G%u^Ol<6f>6g*x6P;Q7sp*Zy
z7vbHI|D)IJ^Jynfiwf^zGn;a?QTltjM1R2ct(u2*D<$Ny?t-P#g~-oR-M8v+_}}{I
zFAofX{hjhnorQ%#p`@hAFkJFkX}LPp>N}%PmZI~<{(k+><@(}HU-~2Lwhlta-Ru|M
zBF{lj@i*G=P30>(8f*wd-`il|_TssTyK>Uqg`ugDd0Zqpv=u%7+DwLY$Ixzbm~SW3
zC1%Fuzy3$pBFF!7Pyq?L09mdwL94!1&3Q$iNiAU@fV&GyYAuW)kxM?_nLF2_KKU#s
zMJ3dwQEk9)+O-Q}AD<r<%2`y8G8_QE>3@%GxBV$;Ju;Z$-@vic!2xb6`bNlWg-Q6I
z*XLQUQ7xAo4URaQOFyr4A2M?T{M@P-4+w*5;<G^-jeZrg&r-tMZ7_f;ek98?hX_Ek
zjkPQtNyGjGp>EDH766#S2Iv$?gp{=3YEeJ|u|-e&1%Ys`VsXv$c6w^3<hRzt=%*M&
zP-)Spb_J=Ea%p8Exi*D2aQ-H%Pdx<`379<{-?Wd__eIzhMqa)UE39;WSDh+-I%&BM
z<|N*_2)dbdZlT>1t^aI5lU<+igo$P6JBFE}@w&m&O4ZAoD$ecV1k`3zFZ)UBFd6g=
z^$|{ze9yr~62@6(Xd{yBTxt0RfAlhLwlCp4K<l3FzKFy``h>lcN5imI5{&!#3*W-`
ziq9gELnPZbcHI21#B|vZ?cz~ty3k&ZDJ`biiir!{l0UG{g~4RCuC<q_7p`ZAk6y{T
zTAZDDSi@GK0!K_IiyG*2nW+n<lYwdBM@FqjEZ@^{?{fUfKXDGjsn9#~rd3Zr2B0)S
zG@^=5FZ3Wb6Z+ILer4OAz82-0Tqo64_^`}+`y?4^f<MW^ExB##c#l_%x;m0t_`9OJ
z8#K`JfRtGrf^ikL`Nm~fuC>P`QMPQ`K#L^j4=|*RtDv~Rl}r%gU?X@)R8vq;prfp%
z@Wah2@%UHbO_%KL5~rkyO$ZO(2N{PhSa(4<-=$P<L0{=4K$nF?HpbsGTn(GU!`H(|
zfG5C2Y~8*qU>qm4eWoqZpJnD`ag!JMR5{+7KYbYc6Qn)L*|dfQ=cxS(!!KR06YYWk
z+NS@{b2H}_;0!8&Irh!*T8RaYX<sh9*0T2FyE+`nh77E+n3phOE5iQvFGJlh{&sI3
z$7^S<wxVbFb^q8$l@E*A9&2IIBa7g2|HURgP4~nYY6=Fsv_9*Prt9h$N>4}dN&eN$
z4;~Z%y3DUfS9v}(&s?4Q-aokOd0E7O!R!rc!WFyUMf{Cdjy}t_kA!~Z2V)X>Kp~MB
zzxsxjG4J*i^iqkL(JZ}%5wjcs74SA91GW+ose#pY)+1t5ShFdYX7MAUVcvedK(XMh
zb`^P8fI&XGZqHLM$9J8^-Y9kdm4xrsix3X4$5O=1Me(u^5C4G%!pHBb%eMD>$puYP
zj2@lgra;t87zI1b*vu8=xy9D;+9uz}YU2y`5eF(op#4yLuh&ZKPw#un>zxBy;SZE9
zQ5-O((11sa0EhL+rOBzv0}(bYay|T%!sG*;OMN@K4$!z%W3F`pOB^TLU3*dT;go<^
zkh)Ne8z2gbl|7qg=BZ@R<T2gqo~4D~$7aq~p6T$5)tx8<#Bu7Lp*xJbdgU8Tv<;TK
z&G1E3gr4E&cS@Gg)llO2v*)JPj6vrXf*`K8N<!;gui(5(*dm4Lrc~a1Y6Q@zXY3Ps
z0=mYw;<z?X!$-jIxYf$oeDj4bIG?4$iNu)gWKZ{ke!Sg~eDLu+*aH$s#~eb0H}=!@
z?cj42uEEa3wS|H~TwXjxLn7oY%5*SW`u)GwTU$4bdU^IRU0)HT5l_Znpe^s)6uC}T
zu%vnX5hwT90<OX{An45ZADE}cM+JTE?Fp30J@luSmAu;hKALoKJ>Q%CRwjzC@-v<2
zo^1m`CFthH&ZYH(I#1$cME%1IT{hGdxp5z{WWAcUUk9UX^>=DoS()oypQ|=4etD2H
zCy_Se8mx*z!o$B^?oUAx#poGg*`Ob=R2W;sS8;Hrj9hPnUH}+~u_yoaJ~(zM-*7tV
z_X5jm!=H<SY&45SOeNT>{2R9#M!_v0yRGIr!J2OqZX=9NgHhIkT05Nm^j|c<ABmJW
z*ViJ^u--k6Ez#!-vneI=7l@31{O_2upk-`o>wXYpHNiYc-?;Frc&*+W{3N1sS6->@
zxJG%PAJjkg195BBuS|!`1oT4E!&tu3aE%Y2AFlL+M{Lj!py~cP!OW98d1M<heyX&o
z(j@w=3q1MlfGUTs)%0C_gKhFf52t+^+X1fM_t0_giKv)~!Y}UW^$wY{9U;AChN*Df
zZOSKa#7jq8Tr4QC3QlM?;XB^oT-Ecij>O!&l{La0-2NKlR>4Hu<8H-f-su8c&=G)v
zG>+BML*z<iMj*+{gFO@!&W80dq)M}TMPcJ<_s@n^*hRAb7`R~Olu@_Y$(+jRH-d>Y
zEZyQjOXX?8z8n0MF;&bx4@}kKbZ>=7Lxh5_K&4M(tX+{*Bi)8e&e>`o%e$^Wc}3lu
z_nyUaN?>j^!rq!-nTPIS$!SPXiW#RLQ@(u$6tQ9r@~Tw({T16yOrJk(y_7Xv7|S>~
zOWTe+6Xmbpje<k{?5ul4`7T&?doEVMe~aV?vE7<t3N8Gn6W(~rx^8N0jZ|Lm=i#e?
zg*w|T?-a*RL1%NX`vKKM)7`z2jffFc7_Ecp*K3-vCo8SLO6o)^-^~1{@wxFBRjZEr
z4J$Gfl_MAYQJd6f);jHFUt?;3cP~*Stn%%~Hm*Z;Yk@ZG>$ks{TWjEoZqbl*_}&w6
z^ce(AWl>X=)k}sZ0Da=Z_Rso5r;yX5Zl^u-O>APu-LY#oNdKEb<%gw-8@mumh~zqY
z9L3{DWH!Zx8(aUF2Cp-)mr-q9==cS%26eC>%9RH7$rEc9Vm9RyhyLtiaaYelJN;e1
zZN#vVr%4LBB*{rOOP6E8${j46aZ>zS57iNBG55B+0|dX6i2=&Q5g58Ojz9>l-Unm!
zx*88c2xQM_5@js?hYz_tKl;hybAIg&9QM>mAyV+No@(~`DJy%i-|IU65MT6L<!bBt
zz+IA6D0TZaJXw-*%k^`BePi9iFZ$_MU#!yaR*Nvq*^ShcEBV_lkuAzqJjDGoGb_Gu
zzz+<|YwNg+IGH}n(Pc`Zp&uTwu?fA4VhTm|@|35d_b8wQsr$fo7<NFm^{Kb$gEvOO
z*9+Y$>EP9!^-zw{{#X4*z30DS-7hH3|D?n{C_T21&uIw$C>}5P{gHm!cy)vOlU4;{
zv!Vt)o_A1Ehh=B%Lv&eXRo)mADfQt5w~!*+*yDVrW~Pr1OdZgN&!eSCrgUK5OGWp+
z9@x@&36l1O_2xgvmN`Mq6vLhNg5PdUetR~={#>3&ZV@9UO++M12<%=idZH%!f13^t
zOyctYF|WzTMq+D88+Q*>P|B7PR*}X7Xj*r)Xd=ZXfS~d;=-KdxUC#Ax$a6-C$XNMD
zpR;839N4zrK0a6;MK#>3xp3N*jQNtvWkV#gmr?5_OTG$OP2qCC)@}<fjeL~8wf1A+
zmGGA#4`lH=fpa5Ply;FV{?tP(&XK&I4Qq!vsOv6k1d03hpRW`8BZa+qW3G|$s`mE9
zY^Q|O{B5LHMJHD&g18r}4=xAw{))cu`xbw_@I4mwyR&J}%P2kXckX9daj3xqqxL%^
z=luEw+X6oy!>bfyq!`Q!qe|FSVj6d-#_>6qODINK6q0ji+vV0ok#ZG{C6>&>GD@H}
zK$od;d`an{DKb-G0*wn(vGn32G(kolUmoYUJiQv@Sba~|X3_i0?nt0+fs%9P<^jV^
zEgMXvro;K=HTXI2@}Cx@9Y}zg>+7RQ7+0Fho&;C0Y9w+7Sv98kds4Fz#|;XbCGP;0
zfA;e}D}-mvS0N-lKp@}32?d<m-fwD221N3=)PbGK{y*Y}x%!#*mL7X!gwG=VvE<8~
z3Q(3oB=2v^qB90)1&NJbp6Y&&pmnjJ2^NQH(s6TtS&fw_uOELxiM;mOvVSnMM>1Mw
zA5-cB|DbY?JMN%HiIg#7jHQ2um3vALyCpPCz-Ir9MiQe9-w)W};^c{(gH&q$+}$+Y
zd&2*OZ#ElV04g%Lpx&Wdfxhnj*pBzG)msp)F+!;hWD>~P-7FODSuXsIp*}A<_Djqu
zeEE8?^6SUc|F}*>nfS7nPEBrAI?agW7{p-x^fSi;(K)F>o+{~%XHv8D6BduZZ;0$^
z{m$FXG|CIcvLWf`epwu-3OxZ4;#i$GAKk&;PP|19uw)n>k^ku|8V{|4Pl=xESPnkl
zKKq@*>nYF-6#Z8d8$ikMc;}JK$M(Z@yXlr`tW(uMZ}Cf6r-^tCx5vyko*5^bQXY|<
z2N~9k*{2^2Ykc#!hMef?h`+A$VdoAs!gEoTB&6B(e`;}EY+Dzat&NXv#wAl3XZE&R
zI(7Bk%P9?)*7S?FUu6g%$nq7j!WZ)gL7IE62ZIOy$!ljzYVR_b{Tdjhj#OsqIEIN$
z=BpA@9rE|mq;1$0=d!}tzc34i5_5;T-$oxt5;OfRgCVr1FWhkMdCYs2j}a~ogpoiu
zx1z2$_s(uycxM`Ok|%X`H_zh@n<oigh!5^%c#eNeUDn#sqN6*ht0AtuS%X%?wITe<
zv=5n1I%8vjvhI5VA3yuqw~PH&=hf8CXEl~9aq#@s=T$e*{%~eL@kzzKA}>WHvLg_e
z@D7vf$L!Ae26+@#$3OIh*p5H8UjMRB@htT_-Q7BeC%beLBoJwBkO+x<$z_e?{<D}L
z+#YHlU;VdzHnIm@<0~o_(HNxsE<HAA&g?loGT}0~Iuez|@xEW)`IT33uu8>XHQ6pe
zG^DWZkO)<BFj)oSR*=}f?~bQOHqp_q3S!I{sOIK!zu4Q@2$FTsNl2iqeX&dR+g1|0
z4G~NA(uq@eW9qvt4?Mxv5Qc0cA;t-N&#C1>zS2$P)}M+!4AkHa5)#rSl0^cg3pk;)
z?voFd@e+DcsTuN0gI;3vJR?{9XYRG#kdJ_wvWnSRcu&lwtRuZM%IhTf@uuQ`sNWRd
z`?vcu`r0Tw$z}OGuyS&HiO<(+tj@vL8mO{36UUuWMQRj3@ssMnPJ>wpW+f+wh&E#A
z5OzvZ2$FyYiAbf<G6J-{XKyIFgNFy7c~^-GW)1(G^RO_1J6a(AV|1C__iud_YcZY)
zf6t%BkJ35r3juq>Z*gtvz?P2Xvkj74W$PS6m4#K_tDN2E)5pQAr!4)l{-)7v#~=Q?
z3=Y`1AOB0J=P?n|-uydeYzDuWeZNN+-DHEZW8Dys2jaT_#~t~R-UEnP`YwJ#Yt4?{
z4{VmWJ0iqF>u8S020^v!!u$tsfRo*f3e(igw@t@Rg9i=bY#nFV-REO%Eol=suYT|?
zVEatzxcl!v>CWuUAW*;Pvi+hYbmu!}@rFV7;1$G0c|$PnW>BwZ9znsU=27r-{Udy6
zQ3JEzX>dD)@N55S&F?HY34SU*bVOg<K##oKBRLR^Wr*kIrt$gGW{dx7-q-ew!UdLw
z(PB$@;6KRI{qa6j?7RJs+MPg}*y!lIH+^l1P+CVe{_S|h$KAMW#Zqrw*AbNFv1R$o
z60^q3*5?{DI#q2^WIWiphqL!_|7n2Dsp*V44lx9mLx~3tDEx58xgLC&Ep?4!M68G7
zj%O$P^I(PhiRCZDMfnMs>{07kfg<ApK{x9Gty3R~dUIl_-93&qe8&0Br}2?tswIr@
z#}e7WmOqQTj-qJ2S~kSz+m~#g;Tk-%+WoU_)Y|Sxu!TA^d_0L-(dYAa9$ou)v<eAP
zR-AWPwZYBFBeC071w<^ZVYKfQ6Vg@UC9tS1{APUhu+)_faIQmZLYBH<2gLLcs_}z6
z;nOTAg7fe;dv%imBUZxf;&mepMRAxc4M?>KX>^-}SEJtCt^MOV+$rob@OJyg1Jbly
zf(g|TOl%eYD!Vh(pqz2f7jjA|@TYmI=o}}){b?x)?`yhJ>hyN!b9x2vAvGEPpm-eH
zb@3L2(vM}|ar)mWEE&2)4J}`kTuoRrm*_sJlR`%8FE0OggYm(Cq8g#D_W5vY?n>1m
zU53_r;*`xVpty_U$HKYXRCLO7fu}kgeZ2y?=0&9_s<$w3LaF=*y~vb(aPC@!Ir@@K
zr25H+_fL$8<)cEzWg87OrwQy#w$JEzQbf8>Z|bU&wtF$UVRl}AFwF{9IC4PP0a+<z
zukTRTjqw9-urrKTo+QjkEH|0k{#nEglMyNsUQ_GzURV9Y1-QmGU1*cbNZL{@IQT}D
zPi|kwelDbwvxekJG_&pG5^h16VZJGq+~p#B%SS@q)4gB;lm|p>fdviLS^dTmY_hUA
zaF%=9UpIvl!n_&^N6KJ&OxVL|$UU!KR5F|j+O3(WMx>x$$8HY7G7V@6(B~gb5rFoP
z$Npc~ron^8RL=00WwUS0{$#H@-YCA^5ND}Q(cdvU0XNw=5&X7f9+xmLDI_;H+(#gn
z8Arji7R4gx4g7GD4@-^<d(QV&Qtm&!G$TSb+4?}9fPIqB7ZW;lqnT=%8|wps=l8zM
zX_Q6CC3B8OOJ2-5r0${k2nc+D97dRdZN6iXQ=a<aTVgnl3;*lv@_PVZ{N&?SFsCuF
ztci2q#1H2<t_5?InNpUce!~;sLzBvHj0^@Z!hILmdm#(nhaes1Et-vfTc1(!l7Gdz
z9;q+fQY}o`1-_^Vit~ovmi+&v;yPkF$+I1QLz^#EqXnM9I(ZGD4={v$5%4l9_7<V?
z;R!7PG$^{p>|QulE-v_nu}OzMz{WF%qnG5iT%<;BkLQ*VLln6JNM;;wm{F4A7%J}s
zwN#6e&EC>G=EIg#JM9cgvBSQPbKGK&9Y(?pK1ar|tUUHN6hO!I+M~YwR!ZU6MuDSU
z<t6H<U!?ZtvB!u96jpHG&EhGgSTchK7Iaz)K8W)d)Y+sZu_wSWebz#C!80*<ZokjZ
z3fwB{HgYtr8G<jt&Ii)O&P-o~>3J*q5@$~BA3)cHMd-KonQ{~@=+)uU+`?5R=TV0u
z-BbEMf>$A&g*Di>+3I;RE6Sg)=|Z=aqNi4{Cd3!?3l^YgF2TX6D;0o$zGwk|Mst;U
z(@*q+J}1%AWFIow2JBs0x#RSXTER5)u&F4^R1g~zoLfvG1J;U~DDoS?7Ur5FGI(~`
zI`q>Yb;XxqUl^P*8K||PLxemqm4=M5QfproUhFU5wO4*)lG5h4OyY)?z9%ShQ`)X4
z>8D6;*c^U_^%1rXSO;PJtIScSO}hP3-4HM;A&OzT3ob+j4pDqV<P#m{R#H~{fS;tQ
ztKP@s38H&Nl^AXUPwZD6$5;}UdS{~*Uep2)_Fwtmj+SB{i_f&c<(_l9VE-kn59q=f
zO^nYzqsSkjCaARejbB>#1mj=40ms#g514i~(yM7ZZ~Mc9HMHtpWqoaGzaBiBvSm67
zWx8%dt2#X!6VJq%z2W7W^4~t)&3gkp)Bn{Lxy&`DftI*5U>ZvW(3Uh7#fa2!d@4@w
z5xQyd?q3Z~bB#5Qo@*w5LCN9YjrUxSMF5sDElnoSFRT2>WI=Lk;IZioCNb3_RQRr~
zmr%dq^OrIP5_2`TAG&&LhSngQXNj;!AL3rrOo~Hm(X>X-tRX3PPQ<I-$>IB^jY51P
zZY{3EdEpu41B3h#fIBp^eRdLfbei*~+&0ra;sV69T!0WHtC|o(Ujtlex+W%0>tHZK
zFabE*=0Jp>W{!I<GQdXc_Zy@?G(CYZDA|?b87<hhudc4%hnAe?{dRbP3@}*lIrC(|
z(g6$Cmdw{w<_Mfvv;b%~tDFwil1xA4IHHU?NKc<!$aib(_-kMK^Y$a0Ikg(R+cnui
zMi=e<>hi6z0OO5E&r>8~Y+%}AONgTp4Zc5HSSr1>KxbPA^{E~RWf8oJfC~>n5XU(J
z7@5>)a;|ZIHtM-^e)ZPdlk$Fm6K*=`w-)s<5{@>vpI~?HJ@ExUjh4N={nKAMow6-h
zQTNJ$Yy6|~;rF$CbKb+;TTS9i)7Nur^uI3y81r>x8!u(Amx$*An}Av`j_n6_DK_hU
zX*O+0*gA+oy~?S3@Rtsioq3h8$#jQPX2v*%Yw9@I_a*Urc*66Y@*miN-pc<7+=pWc
z*H6ke(nclrR$Leb;Lm*5XLErHm)-ziZT(=l>)`$VU_YE3U09U?LMhW|dB4ns4zo)b
z7`D+JAn72png)Eq<1_s;!RsI$-n9+_7~T=l+;M~N9VXAWgSE$cz5Rfi(+9d}Xluw9
zWhD{gJZ1GACLlo2^Ke4PcW10BN%mv#^$JrzFg~vc@SY5(#WD)=dItn$=IE)CB`9~i
z>GTZ%uysf2nDugsu5+|zL-70=4J3OJU%UPu3-x=?UI685=UCfNF?b_E=v({H$g;IE
zAjW>Qph4WbD{$oTl=^HGwTTb9PpnvYIbk8p4*zm)Kgjk$!zvk7bKwBihPcvC(1`UF
zT^W+WA_LrQR7*YrI)O9wr1D!pdG>t8e;csQZO`-J=UUtx5U<zKvK#PY(Q$mAqq{@@
z1uRw**UMC`Zfw>!=}TwD8^2I9lGtMSVmemI+cOt2v}i};!knbic*`uQHzRBc_V;0T
z%j^Q!B$o(x4Kwj{MZnuGPYSzwpQIF_TI;ORef@_E>dygahaPH2RKn7yQ`#Qs3x6FX
zS`d`1vD=f{_Ie2-V@j!TY$7kLyAv6C3)E`QWDeTk>p(V}ep_fd92yt8`6iAas0|qB
zcRqmcUa(8!*cYTW*a`JHYg~+q#IONH{~?HPJL+j8MQ;Jh;lGO?Ot*`%c&6m@MFo<?
zw^Ab*F^bWnzZ=V49dQ=<(Km4$aD`02Wlz6cytroUx@dzFd!3yF%c!Gu|M{7m>uTdp
znI~eRBqYgzHKt)6>tWqHRQI=PvNkO|eCyjkSlm_$SSHQZ`Myr;i$D8QTBIYjj=?Xg
zJwE4%U%=<PFy{5zI-^xIB$>yJp4snDU35f%O_W0#qu}5fR)XI)p!8aUj)!WM(R-V+
zl0+nTufceCQL^|(6Jm+OnYPW`{&8{CnR9Z~%@$uDYbdeNRZjPQ@hTIHeBYTprE~0L
z;-ja`to7R8%8~6POXDgx{cAO#?DGNq&HK8MIPkbiwvb^&ZU|)_Zd<21VM}cu^~mLy
zY`+pV-f)-Y;r<n%If#N;p67J>OBrPsB{_P*VO#Gw?J*JnAst;;mA%vn-}u5p8=Pk<
zMgpY0aeY&kKO9*}@<<;erh%+hS?*A2&|_jugVJo&)|S0FUZtKl30f~2R|5?lA$L>*
zOM+*&Ic(wxU#8#Tz67g})i>Os*<Ks%x?bcyUz)?Rv+$aH=N2%25;%<ZV5y<}l`Axx
zuI=vK049R4?9JBW9j(<@fI`toAg3j|V0RB~_<wCgL@509U(SbjI3u^z72C8Zy_GOF
ztmuSe9~oQNwfH8m{5JrgDt%-@V;XT(NKN9_(L|KUa|e|Dg&OBN!R<&~;d;5JN~b@{
z`WCF_e;ULvZN3X%Dw4xP6rNR+JpP1PF2c%sS-WLe7F|V6cLsbK>ki8o>6iyyO2_I9
zcUqK1V&g>ds%zTI{toA>hVo!4l?z&lUS2yoT<a2!Jr4RuKH-;MZU_YvE%Ku0f@Rq+
z$Zg|2%*q&`V8*+{iO-Iu;=Gzsi^5;8Mf0pI6s}c?kn|9XWir<~JhsYGNGrsE+~Y^p
z@KY)dx`dr5oJz7PNLUzIK#j)4ZfLG%z+#gaNUshDo_Vv@l?=dY=D4zbIx3sL($TCt
zP9auc_DQ>XA{q?C;4b&!d1H};P~i(IBK8{pci%s6AV;tVUON-O*o;H(Szd0+4w$mq
zUDRb=PZ;%~Z|VJc8SAGiIp+VtI7>}y|G34K+(z^h%aanQm$9*m<T*LND`*=g#Tlk9
zj{)abyr0T4p531_dWHae*q8X}YJtzuaKf{Xd)i|$-H)rCg;%e`mEvy$nTs#VVgv3K
z`e4TsOaLSI()oj_UdFY{>BLH8fqE^=+(Z8~v(DQQRQV>@vnGVQAjx>}j=QVh*#8I1
z%l$>fg`^Yvc+(MeHFkWGzxM`&X$+H`k{=h(0E|JsfHHR-YYO+&Pu7B>CvJ0pn~t<U
znijjxky60&W8p|Fh0Tq_?T^Pvw##br_mnlc%mlcDZb(YsOS6uCAnbZK7zsTsd!rlv
z)>@5OCz*uDKj^O=DGKbfIC<uv6g%4<<c}DtRKnVFkpjB5??+p|&w3pirU{160A@6B
zJdd64P<`%UP89?y?A5C}>6h!>$Bt#G*z4)7X_Tr0!pBr<KY?=E!l0Bg)Bn=VZvTCJ
zl$ev?y(z>82zDahgcV~p{@WiF*O?#c7+!3$UUnM8;tqV0@>^epC@UTS`PbNgY?Lz)
ziEof4Ie?!lDA4YT&J{bY=KuTK?x&*2EE2KTHx5VTzLs_BHHC1|2P*is%+FO2;$~PW
zR)T-D$K1gPaQ(Pm@X_xTr8(fMD;uDHsdw$lM}P6T!Kp8pEI$iBv%u5tXB6y*TONMS
zj+blM0G%PZ4LiT|10mMeiNVW=d7;}t&+Q?q{^dBwEOLrv#CdEeALeQ(Il@2Nmz0IQ
z@yWToZUce?U)Dy7eFgJY$T9V|fj)=ngKij!fclx?c=?=Eu=ItXV7^zOM$NM|Q&qM{
z$(;gI$gkf?d5x;Mc@m>UgQ$)C{v(z-Uz89G(is;?q}$+M74&06l~~IMH@vL%b-y+6
zoSon^AnN+ZE`LvmPGRgCq#-39s0J><&yOPR3!k_Mu?pPvt^5{!o>LpK`lC?LiRps#
z=(z8b{Pjn>#;ooSPESp(6Z2M8TT{OK5<V|8*<x&vaG`(lPI8sr!}RU!SnJyQNjUk+
zs%+d1SVOJ&@Q3<d!IPb)%>QOi%CcGF^Q7OGtJUqftx;3X>(b_Aru*^e+I@s4Zc^LX
zOF)l8Ya|7mk0q>0pE2f4?s1*=^d7m42c8{6y0fLKB%gpR!$pK)ZSj~;P3r>LCgLDS
z$=r%*D5RV?01CdLN&4w~G@#k-egzW<&X(?&l|Ce&fc=n)1zKgAMJ=MULF>8t;RYci
zWLVJ1iAVw!(R?$-et*StVBNnq^1XhD_bHukn0|fU!F}`@Q@I4pRY<CEm5@dh!h$)x
zIcw{VR0Yvdd7Y_t-Di`wf^U!a+rwHh@hPot1leTBj9w2?90IUd>~(_L2<q}3E>ngz
z-T%mY%b+@fHs15#8XSVVyL)g=2pZhoA-KD{yAucw!Ciy9ySqzp-{F1lt-D*bUv}$$
z+1gVFsHrnRcTdlB_tVe+|65Qmsu;AQWnx?5ur|-~$J}ocW=d6<$1wv6@A~ug&|$)*
zkgv0WI*oZX<oa{V{$M1Ecn@*s7_Z!H?00|2R?lnR>*aj7?kvn~hyXn$GkCtw3hc+J
zbuUxP>7Ne9s=5J|Z5m*c*E-eJ%x%L;E%t7dJFP7!_rt`e2dwKz`6LH;C{ihbpN-ju
z`QFPsv70GHL?88~?w9@+a6VPzs$|!kl5%`n1kgxoWPNq%V()c=LG$siZP)LN*&kE(
zu{!Hz(vsCCimkTqwP}8I{cSt>1+iwF>CqqlIhm{09m(MNM=q>bB&Bu=2g##@nYU%5
z1f2guuD6%oUyxvqbFMpBfVH7CZbmHHFhIvW5X-PoGU3(m{u>%EJnW1UF^(qzk&>Le
zZ$!>~QVxhb_1AF2?oF&?G(SC7g$)85x{wLK9zl_Fo*q|5*V8X{j0h;W(uxE7zsG|}
zDA3Y4ELCJl893!hc`SzOS>9&xSR-K=j7h$%2*Brm+qp^;MPrb!zTk#Et59NII_Ak_
z+nyt5Va6J?k@SQ$Ew21ZnkTdu116X^uGi+RWoA-E$MQMGeFkG?91b63D`o&U;8L*%
zzRLe5TEw9Hqx-R9G|W&6b1!rmS|af}cFgHyg6e(+z*N?_ce+U$6lFQvYv9+RhSazA
z$C%e)(rMZ*IO>z(Su(%~G1pgo@=bYtH~HiQS>pKGFy&Z~L!8V2vCRdm6Z5z$M!6I&
zm0uO$I|^(fs9OdKtp_YY0!%wZ`&3V{CuF=rkrzwrlX*;WVQ#Rq^tzvMr&Uxwgkn|^
zHvh0ybr;4Z=LN-vKDIiI=|v(#sviHQaj&CMNNOA{s&L-sq~mP4WzN{1{!Htbfw89G
z!BgxfLGaZdal#@_Z*BqHKMDL8GjQ;HPJyMIEE=em;FU~alq}&D7O~+hJ+W0}Zx~3e
z6LG2qab}Hr`gV@u#5qgFN|%(Ne?I^R5)(OqIO@ny@GuAmQMWRr{o$TWY1^Ygv@f0c
z{4l3cMhT*DzP+6S+Wri>BJ(7F$x%JCSRPFsPGWOELc(8=Q!2c!McEV@ilV};VXXf9
z%zhH?U5yy@Uh@)JRy7<A)PrFzez)5-d~~ybt&FqLCu25|nE7{_X5p5x1_;Udh(p0y
zKf^V9f(|u-l{Twpyu|kZtkY(DB0)x+8#?SuiOE66V2cIPL}W-9tTTK{fVII4y-pjq
zX$}q=z5#zTsiWIY%oAZggA6yK9p_8%!U8RacSqa5!%otdX<UKn-&&oHy)?QDcZI%_
z-&-0U(6sOd!}Anq`U4Jg6z74htf>xlgxcotA8zz42!_^664&2W;FR{I0`yO^RmO8(
zPfhoIA&>?ipGao<Z-)vj18zm6sUbsc;=pp>a-tA@1tl*7Hd{Bff}u&vHNUcPMZFU%
zPNM<iXMH>rwj%!oKbxhiJXV*-W9TWfHy8T%I~01YCT4DiA=@JyBz;H8koH(8;;%!-
zSRhFVnsxar^hlK8lV621GZOD53$~NPZu>fAXPfnetSk+W%mPbz$ZGFm?};fQ^FP!{
z?K6!}-a7vj2p)7ndwSPESX;vx=Z<Cav)`5xE0mjQ(wIM5eYY#Oii$VqKFkg9pO;e?
zWAo=6h@FPU@ojcEFgas;SkP<ZD2%5Lt1flAuzXeZzK$jX*_^olcXC*t43(=4FS6Mp
zhJoY$EG$d<-Moe#t_|Q{3Z<#;U~eGuz+<}VEAgH;?TiA>$Vs(_pK#P!$@9ecE)MCJ
zoBFN&$Jq3Q(7|kNm(FgQ-sc2Dvqesp>yoAnKS(e0C6iIohynhZJxz=9)7Y};X=Ni7
zUr9^tV&6n=-&oSEtxGoL=C240zJVQyRm%)VLRFQgMmYE_A}>H)<uG4mij-Y84;#nu
z)-O8KVGfnz`&vnZo^Qn@V$1uY`OdLN*Ff(zq^ecwTSRrZ!-*d=@L90WT@pw-Gz@jg
zGT*bz2JvF8ViR4pIg4({M0YX<`fC{W$HWsF(9qz(pqKG7tneGT-rBGftbp8AHcGs=
zd5K?T_<Zfeh9XoNU|cI0v@d>5=<<ZKmU2FPBGM%NlqsHYYIoS}llS+s<LGR`%ga~v
zMYyWT!n=cPsOjqF^<X-wT$E<74~9XaJwjBv;3M--`e)b6br$xvK}ZqcjhbVy`A4q}
zXMdE9w9{&v0r*cXppQ#nC<>YtOyj`7`%9_S|BZ$Fr`%%nXQ>PpijNt*6A&uet^s8*
zKs1^Qc_hk(85~Zg-_L;8k|=I-#l}WK_ALCJ{~7&f^g>?ba}}^KeC7>EICJJ6hdDFw
z7DcYme_z8T`JGc=DFL5o*=Gq%AsakF>=&XMt8IQHcp@qVm5T4PtvDJO7B`;9L`AnT
zY#j}>VHKsAJpZ`PtnPLR%9@$UrC|~I?q{2Z)E@5I(#vxWH{!Knt-0vGg%VEPUIs_y
zy2yyVlbnQHXGS%M_z31H9!mZUF~WRdVa+g7F~3q_$neJp<t5jGU`L5LPAL4@nm?1Y
zD%_7i6Zw9EOm^>@-+3eYHz}dnB6J&NM%sJn_QxExV&SIgSFhkWguk@DwT~V6Y^rj^
zSQVw$eyGpf8bot&qW%8TkIUDD&cR}Ge?>fX-^uJVtKY3)ab!5Ylc6b=GR?;+2~QbO
zDr|Xwx_;0(-zZS;1>g;|+v=9lGk}fduptUs7kc+vf3E?(aSZh{ce4hj9!)?jJiO89
za>M)#%W@BW6IO0PdE?;m&cCygS-%+|;B{leqEV1<_&FmzcMn8hnVo1+SNqX^W!G98
zUofQYb0*ewDS`?po*L{REi7chz*(g7>ZL=qD$2J4UmVXC@BK4VidhLuR*UC7e&f0Q
zFa??RxL^m0oe9kOoCG58EFa*^{-JZAmMZkFv65f~##3blB)_e5WkQE`lIQE-ucT9g
z6rqz&@Ub;-)5pZUC;ip9Je@jtzA{3vsPja8g-iwEd``!^Ch9*-&%|0;0V_qzJ3vd{
z_2l+~ol}2L7alhnE(B?AVs2;u;-ai~mH<^|1LG@bqq75GTjh{w;FYazYH_jc=QMgd
z-DC2t8BZNLWktVqhTi=S3=^0t9o4Qo>8`BA|0YWPPZm^e2B~$pRn$(6ZjU-~ok7c>
zhfQjWLhJF#J5p_ss5}<|lG;6FKm%?HLc@jqx&AzM)o7Z4OMwp+aR!!3CUx@x=FJs;
z`lX#I;POlP86cTuDp|-L$8Iury1O{96%k-!O5JpysTKCMu-N!)>Mzr!&gZkC-Bv25
zoUe30sD+5jJ~(DLcsE%Tj!2z_{o!cD0|Zt6j;k9)kzxd-Z=jv*veVFye%)&d^U`k7
zA3u5#9510pPG6GWMC3Fs05n%E*-t1`#4XM*U3ox86niR=Wb5N7s3q_uc#eN*`#@H7
z+7z>7x^k9aoHsoFg@O<N{_XX-Ri#U&5_od_Bqnq}Bq>5bONYAJ%K)R>Pd^!GKbS9z
zJB$GWLAyb`dI%#>spec*u!@q_ogNyZ>0hIJ;PlFw!o3wLg?M)+f<?5G%t*Bv7U?=k
zp8ge3pv1M@lG$&8Z?;|Xtjauqc0#U2&a!uOTO7^p>FpeAWu4$(_Q&?+Xj9#8vtKi9
z$ZiJY)*dCkty-pW&s*t;k}hA=FOTCdnhX$|hVJycZG9xJeXhhKOYu6tX=KaiM&D7$
ze!#z9v>%E<=%;`*tZ`T$?1<%1H&;3=a#`jYwEJh7$Ppt#OsZ0zQ=C^mns#H~9IK#y
zofgWPL=pMC!C<<}FtTD-NcK!ra`UCp+&22`k7<&akk89crer2vzy}K_UaRrwT-o+z
zs&Vs)1Xcn@QmvJ4fZ#wqlj|Rc>U)t@VV)qc1>q?y{Gy_*|7-A@3m<ujWR7h5jBTjY
z@0zp<0G=I%dyDFnh`&oTMO_SNr}sa?B0Hip;sJYMW7bW|ES@&wQdB0rzk3!Jg7D@S
zVVh=OCM84)ghjJHyTTk+&Ui$jvWMx@V2k)fD)^Hoyt)cl%gs*M5wfmz@!U;R>@sO%
zFRenpvLH4*fey~C$TTdeMXN~EK~H}rqw&*jfkA%(8xi`IdAI-ii%6)CIHP^0p-r_X
z^5;U+tIR&Mvp>HWh;|}npVOkpG@Cgvx4ElTr2WWH#AtjuCZ0mfb6RXd+i<=H$uN4x
z0T`B{q&AU4?=QASnNF9LIM0<1TA3@RF;F5jsJB6__=%&jg*z-iH{!q^fULRID;38o
zcgcPqq;myQpGY5egyeO<h>aK*&2#9zWM}nZcQ`jaPF~MRRU54U8mL9bd<u!RKMSb)
z_Zt@piN~X27KwyAz|cGXa-H+;SaLgzJs{?96%4SRs(5x|MZwiKubv+N&OoQa(1Vq>
z8{0oe1$L67xfh=B?C(Z<P<@g-^u>5z2W*G903*3VI#ZOnE!az(;;b5dhw>YLj}DPT
zN^FLbTus4Q6EKavT<GKe4GpSuu4f|C^W$kcw{elmHNdNR^<f(@z2e2WG9IDX`hr5#
zvovBPkJulHI58CPenjrcX8nx7;K%Yg0CBn&=9M|EWXNbu>FbAQ;Aod0L{_P>w6wIM
zcN!u>8yLQ1uu^q$X{)m=-c$Thr<IAl2Kj-hQu*kq89~Z~yOsfOqgmGc(e*_{?T6ex
z?X!++xAya}P;dkXQPwoJ`>8jw@Zs_4RuyH94lo3f*rJ1-SjD`St&Ou3af{+2oXf@7
zzykz0<_W>~Ho*GjPliA+B#lJSG9QPSMm8~W{8!}YT=ao|zx3ZZiB`Tde<d~eyRH}0
z>JY(c^7Dc&bMc!K=RfMj*>+Aa1@!uvVSE^L=BjMxFBpf1&$cUdZu6hg!elz0g$Ee-
zqihbKav<IxftEE+SyHPpl8(yqExsC90x}VGB_M3t&|Qj@pFOPl^(0+a;(rZRGf!b&
zUi=;BG(E5Lsn6Ufn5sqPP$$p1)msx2wEb;4Kr~|e+uNKl^WUD!IK+UCMb+c;8k6H*
zYph`mAzSnakaN$K@=C+V=xmvND=QIM4u#fP7wKNuFVoNN@QOX}81J-6UCXYt0xU$k
zckMJK`L{{uo{9owS&CW`-ER?T@Zg+ofgL8E5?wj&03q=ofbUs)OPZ&#Qnx9x4}}gA
zvu2~;BVxatHErF-dv^AeooNk=X-{&DIIO%QBXCFl^vrQTU2@ozf=g)@xF?xMB`qYd
zC(KX485-`Axj#KWPK3<w&ykMwc|Fl3*!iNg^B#=pA#zovF03z0HqawRzBXln;5ofx
zP^&r8fWy~mE?&;c{xWra=@*sDgnJv@+5xqpeh+0*PXYe6$jit?=&qsxZYzF*4dVq!
zMj0W3|H=j<)lhQyWZub~E2D3<K9oZd&>EdKP~QSW_n-q6@f4=e*H_@W0QLQis&FiE
za1xo#ov04ao;@81wQeXR$hI^NR!r}8{tU28E^sO-k0GPsBV?RRi&W<Z2<=xpNJn~W
z%Ld*S=-1TJuc@hRP&m_SD})~g@=PgH&nOu_cj+P{g0>lrAWC;YVo`!wSRamOCm9p7
zOMmrw|C)GkN2ML|r{X5c@#9?9R~~o|#LgL={RmMzy5NEIktKM_*L?p@r_Lu{d81^J
z&G+EwkhXUjxL#FqPx~G`(dC8vW&ar0=K(&lIeE1T6;7V{_BS0!5mjMoVBSG6|CQ0?
z-g+6z3A&)+p5k-(ntZ<1j^mj#rZqL9w{ThR_aypDaF^59lrh%#ncu&bTvm`6;}KYV
zF};FkLxzNmtC_E&3tp>gE96eB!}2*5bK7|9=eH;#Mr9i)J=I#Xc5vaKpv*P-+GW9~
zHUM?n7~mSfcI*(N<HAeN{Fa9wV^)UeFL)5;I@a)#G>>b@FF;Y7hSO<MLh-kfi)x8A
zd?ehcI7U9fDZ%2U`C!~>^ilbIu~dVvoPk$H_VnS)2;3rPQz47{nBtJnT)S%bM8(zn
zO6C+NUQ&jA3Ndg2^HB39Qr-XYHS8Vk65KZ&?OM;U7SOH0D*0fAkA+2Du2CpuY3Rcu
z;6sSi-JHkMm>{nZI{7+5GJ6@%h3=QqvF>bysNcq7ZPEP{u&jMs8s!DkU9<@pXIM!M
zu+q#O$}NM~{(P%D%ykG6ct|ET(>ogp_;*OB+)ad@CjXpeB)X2~yW>{PZ^K&f8m(7C
zQsWsgjxRldOvMdz=)-w0e%9uFB2c}vAV3jLAW$=@1#u$?9(f9Fdzv^`{eD&tW#<se
zJb5p_+Nf0j_K{vF(mn_ChrxN|OrmaaI4avvb7cG8MYuxB)pZ+RVXW}ihOAh~k!i5U
zLC~iJg_1g0k?_APi)g8RgTYrtXGbx~ZT*y?M-^vzR7S)<lXbM1=cNViS8W<|Cy27X
z$AA@8Bw%%;2px+VUjeAJpGFgt=|1eYav4S&M-bniO&Ct-tz{x;f*ZNFArtpp&pb20
zXqa;};!o6*IRr-=+Jb5K?J6`eIP2XB`}QyIvc_~OmJdAJi;$=cOq;rg>E7pb{+`jR
zE0_Zn@h{nc2v}|Nor}=kc#2y&vuNNxyXncap)^L_@u~a{C3<dYGXk4eqDK{K@P0jo
z|1046^iRq&y;XP{z@c&`NHaUxs@M@^l^er7PJCNzGjtj}=nDH4^Hz-1nz5$YFe0@J
zM+NeNIqs<o7e2k&!++mwrsf8W!NO!$gZ01V4Tn-v>_xRPJiTafDGhuwT!&L@u*>4<
z{ympLjz4G>IbC5jR_p!K+uMfh(8;-9{C7oub|>2)?Rl1IEog<)T!l{7cLP23^8T`}
zs`N{-bSwn%E)A8hz-ji%#qk5k>g!&=k(c3>xjSKw#@aoCWcyKMXBmVlS<;|xtN?-O
zNsEl^%EEBr`I?2t{S#Ci>R`41@g4`-H9~^9=FhCPws#<aY`IW+a%1WkI$pprVh!*S
zrA4$U_k00AGaiX1Tk`-aAx-Sg6F^$o3|IkVaieWKOS`4vzT-4#pUU9fRyGbSvy(__
zFB~9IZCcWm-{d*TY$~2w@nAy1U1b$$szOMzCWk&XM}?ymR%Y~j=@I0Ts?IQtxQY~1
zqu-S?Co6Mh+G5@xB2vrX+X@TcIk>z%q&L0lXGgsW{ux*k%lNJE*;fx&#T4Y^6nPZz
zOlZltbmo)bBv@sun@~ZKY#?@#^^)f1zAvAGWStJRt+fhkW1m6h8T@y}iBqt)hsiU$
zLL!NhbI6|^I;SW(gIj-0FhvpKdg(nO)_kl8c93n=cDc->pHyx=bq8^BzS(MtH3e?-
zzz726PPjkq!s|rOEbvtGjQhR_*<<L_v$<u6(w<T)_JV7ab?wS}GC4JccD;w{%p5O7
zNy&5RxnJr9e3(~;`p(i6r)d7chI&V6CntycwGOiXzdplvhxomox(Mz?4hKksp-Nv{
z-Gd8<fmB_&$JH#V+b1BDjQi7r7!RRBqS56@zeR2BvEtYL&uCw+-+L?xtFR@y>oyd+
z8`}J1s9o^r{Un+L6PDG|cIC6u=63w$YqI;XoP24rVa>+^v`*1N-HoE21~kwchK0tF
zKGwf-bk%kQ{}TFmcFEB*iD#$gTr#cTx~5w@m4z)vmE;MCu>U(ac_h<qT=29|+#28A
z>Y|TEHy~E`opD}WaI8Ink@OY*F~urR-`%<CQ!VybRC7Jr4rzWNO!x1)WlWwVK{J)2
zBh)r*tuNg0i$Rt^CfEbyoZTv2iz*nYV2teh^2^vbRqpC!U2<6E#}Zs6N@|t5#!lf^
z(YWiudV40g?zLEFdzI<G)gJ%w`fjBt^OC6X4PDl?k`9K|KfE;8cq4wOktqz@$L67(
z%pmht6NXWya!1C~qq9>FRpn(p9dJ>Xjyn2VJ7=^lF-;vkA=EVZ<FiMN=SUZk-l+3V
zB)%DY(vFlkbQFSM4C||M+189ImmTWY-iGEpRH7v3an&T54Sj#~q1%4!Z0n)w&ehh&
zRu~N|<?J^TMY01tgDv7RyRH{MIOkVD!?cCZCU+NM%;Ql#bnG_iKiriIbH~@O?SdZ_
zo|!!RO>L^ZWvXKxJmc(`&pOx5xZ727zVE?<3P@lgX7{%OWhj?aEQlbHqyN0VZ&?X#
z8BsC6b<`vC+L%`*N_`1RYwzgAnKY?465HYTb*&51L+Ozu$gsA&g>e2k1Ndqjp8sj!
zYmM^Gq!318(9x4bW@6bn``!L?2e;9@mZ9dT6?OPl58|c}Kl+}c4A-TNLc-)2GL?Qd
ziX~I2JQsb(3su@eGX}c1H*1hPre_19f?gFg2B#s5Y~7l0q(4+sSI*jG(S-RACq<2G
z6o2!Ac~FMOO8*^cfF~oZG8|tJ-)=p@L57bhaFMU&#xXo(3_Pf&NmYT|)1V*a^J7&|
zfmw`T%D~62)Uh8~*S`PbN}mli%<}ikP0lQ>@K>P>LhJZgJEvV5OHbx}t4Ogau}Lur
z{(-1}6=Y-@vHTtOwH=<dbWWR*daI8-Iko|7hf5ESIO+ke$;8Q#@yB>M5Pz??_2=OZ
z|EOMG7Ni?|vuAw4r@B$9F%IBJy!8MK(qMiPr%84nto)W%GOWK@)%rwgqPW9=kU`1z
zO+1V~(S_84&10bcfZJ~N-g2$mTnw-+N<as6d6Q3@Zkne(gpDR)IQ#e;eM25|<e54+
z-on6k=_({xMnnv|p1iHntYeuYFCKKHBcV8)VY=LugD^Q;S9LN)bjc)!7^3&9oTGs+
zCnYNw;-}-JN!USS%C1fUrz5h80yg7>`7qA<V0&cOGpJ69aI_Q3WXY5Msw;*&S5Apc
ztm%g-D<R{1i+E<bK`@}R{7AH6ANNfj_+(~vvPRWGgT;w$sF5=u_%|T8!r|w|h_+ZN
z0u?bY3f#eE{*-Mxa;&3lTTU#I>^F7z#+}ytQi~Un(#2HbACu#sDX>tjpCmK4@VQ`(
zS>K$(Yarbs#v|xDAFD>zT~)l*>n7TvU(YfKaAwJ8-rAwxa@gKhICBR(1p+L01FNN;
zsaQpmHp*Hwp^QP0dEMfEj!u)~ce{R68RrdHnO~Q$Gkg@uNY?kT-pxhONi=jqIa9XO
z_zmXHCY^{Oe6LczW)cJa0kV5HX7Q8xW7++og?rt`9R8xSWUbP*b=j>u$~L9OqjBoR
zd-{8nlGDvuQYjGv@T3q%`A3f~R2Or@y8AuxqD5_Km#lH6Zww{ZQcrbduiEfhB^oh!
zMyL@I{Serhkw)v6SVTg0UrkZt(GZJwtyGnmH6eioc8G)r4XZO)RXdp+o?smjxY3A$
zS1km3HCN%a%hcMS9Bb9BYhf~mcigYX;J112>N+@Rs7vQmxP(zGkkg-l<U;*V-v4&_
z>yXkpswD&!Chi0OIB9Z9!s=gfawYD;$4^AC_3wzr|2X*!BzFRdpRLi+YVh4v#H0r3
zd0><#MAXCo{eXQDQP}t$4YFGbMhFE`809}_1DU8FZ37$u{frnm`cHv@BcKO-9Q~&m
zfRBM_|Hlz9=lp;Bs~_$8zvU18f3yt$+Z*)%NjrS>!v8H7RfJ*l9U^DAHYABf%?Sj&
zq{U^$Dn<1C%?(*tKp>d-FnL+YPw=?#z(=2?rNk70pTr+uu+YGF2Y2@y;3NfUG2tJs
z%YCX|GPCzM@2eN<7ec|AzG4JS$q(Qf>-6Y(b3>yN09dnS8;{uLH=tM)F*s}pcEn+~
zTxES%D~VE%467aL&umf#55?#HS<h~%Ui51|J1lX5ubJ=s?0)6iGv*l%{&%hAlNRIs
ztC!El#kFVl{efe{w_ErQk!ddlXxt}e8dugY1ymB1d1c|?jXr1{@}PK$>pciP64dSh
z!p}xofgD3eQrSc?LqZ^T>1RQRDKDLFEB%y-2x*?s$9cEnp{2l2dDx!IkI3Xq+QF&`
z<z@tj0eyPniQT;&e&y7<_#n$u^6r{~?*3UsjDbSEq>aez#y8*a1gb6UH$<m;!9Lqk
z<UBWFTwX1ty45$7^jLrFUpT_z-kykG7!m5oJqJCACcM3HqAB;fmZl>*Gd=JK*0h_x
z`KR5uf4Pw$+s@BdNKJ`}C)oJ=vTb^~yrxsPb}s2}ylyBw3g;VCn^We{5p*D^An$cJ
z&~*3aflU8vh6Dp%ft%m`6|b!35{QoAeoh8;woLiEI<Sj4_{Q$DKAV!aou?hdtvKU}
zi;IhJBkWbFl!-HA{|LHv*d_33-RSf)J(ATq8E~bjf(+16<#rWB24yinLk_dTUpl5h
ztc*C9{WdY^E^iJ7SBfXU8kuNg8u=Gl%SU@hT6>ND!vq6fbsSp6y{XtWz`I}xO&&+8
zH+ovJ+Ljp`C70chM8?g{O~eChuoIAY_5vi)KDL~#+F%+Krb+LmU2yt(28=ecAWIdH
z#&kc<<(8Q{$-3*~=gY}uBi8PVdoo^L+(O9W&Q79%7^2V1DT#@p9xzDwU4XJ|a(z7-
z*skuf;`}Q|{BTZ=S+ddoLK|K1xl8<|4tWla1J-SY2%eN?K6rs+<m#+meDcT%7TRB}
z)B8($Mh2AVZ$Ce<rZpQ^e!|^@gPr*bO{s6+l+yma0br-e>1h%En_e>CZ&|9B@i(Xf
z1O)yHWl9K*H@S;aO>(di3DZ~ig6|xswrJT&NkQ}TDu_r(+W>ROK#+q2Gc7G`aA>F-
z*r|+7O)Z8$U?U?VYw~<#t*NaI`uY3(_qWAAmozf5h~^sfQphh%@_I=ESa)=)a}X+J
zm{3qqQnIpGOCSD{TZTyd(GP&@Km^<Aa}*v|crB1T+TG%Gf`W-z_^;Z<anpReBCcS`
z(D!=Qf}Nxvy=Tntz4&NuX(_C-lJSSKa+unml{0Rq<3!g3N8Y}alIm(yohnt3CZIm%
zUJw>}#7Do5q*$jXi)Ua1%;-6fKE+;mFrVWi2auY?C2oJi#KgqL!xI$|@n8HVEsbho
zQlm!SL(0Lyad>oun)bIR4Q6KXfm=kD>lU7DDBko4>MJ+WqG@-@x6B<s8lvoimWdzg
z>R}a{jo};!F{y6AW~7;=rKqTA0Ltim_~qu0Pq7g%0y%?(6x;ufK<xP*LzAfXU(CVb
zu{xRnOS4QtA3{4jyU?&OVE`#5o5}6Bx~erhHwX9b5WL#zf(*ziu|WOOD^pVnRx1sk
zkU~T&D4>T)mnbHRA!k7HsHC&$)NJOMQ7l!OW<7}9_ah@CbKf$P>+9=(Hfw<wjG2gf
zd9}L^fF4Jj{4OL`QN)Efrc7Ley_hY?IX*efWW<ClSFVujYdbTpyH&xE1#CqRbPy2|
z_5+}Y-jAg1so7cb)xXu<L8iRbK}bXb<^5AHP@KD_Mr3*{laPA>OM1U^RSDrmdW`bH
z?e=Afj>kK~04u<`mT~X0%1Q}+eG*2kCSvx3FJ$#wp_gG{a9VAy3Th!i*UwTQ)#RjK
zTW1?Pvso_s1trpHWdzr|fB#DEZ~vH$?%QeISS^4?m2@rmj4j1{9{|b9F*7-d;_mLT
zm67>FlOtzNUR%Kk%I8Uw4Hs=->iMb@)v{(ZHPJ*y?1gmROgP67WbcfnB3O^;#q}P9
zNvFE~7g+!W_bRQP$IYzhi8(7<QxoK28>3chN$R3fKre3gXrOh`8qJ+>8@II^_cpk=
z8yo5J60q1Qcp6-0ea5YLi{9;ekF3*0bf#(-sfCtDX7rEJgKF6M0?SnAWh-XVi9_rA
z{l!Yi-Q-VK?^Nr{>QmacM;Q=2J?|HrNN)L#Ek3pVd6A5`NX)(<sjIJxiVi5^UZz*m
z_byxKdo~3O8j)U6Re67&8D7c@Funy)TJd_rA|1dgdbyXEE(fU!ScHgAPdsht&&Nf+
zp8py*BTFC{b44<>)fpg2x6mo*-hQcYnc}klOt%v+-^olw^3S3EZ*<1505!r4-{~8E
z_B$66VcWj(8?Cmt2|5F8mZO_m*&&DVKtzAK?bL}VvveD$T?U<rUl=v=L(VWS7BYc@
zZ7R#xv5(r`$Q>qiq+)JPIpFCF(4|lgQHL>E;T%f0eoBdKehZnMFI$mtT$qGo>>?mJ
zKK`GVIF;q>f5Iq>_>bPfJGj4C6jcRDD9wBq<AcVEEKmAr{W$%G_ttqx2Xm>*%B<$?
zf7a;+jIsY~M0Q|f1%be;{4t;~KcgaIVFqKsU?D*hv%-JE3cx_X#DHVe(Xq9mgU7)1
z$ACl9)uji481!_I{4gM}5R3!@Kt5j#5wI|@;6>#@tky76sZ=02QFts&Bn!}wmG_D2
z_eE8ZIT9ucsWymQKJ@)PL6QWdO@@L*qX2?rj?Ky`CIf*K=#Y@9)j;a6XW@}&?^+-=
zS|kiqPim0+*2(F%Cy3G$4FeOAT^j@<U}ho&fwVXfF)^?{sl?m4fBWkG6Fpw_6BY&r
zCMpp>Kc2KS?iYSyR7_u>=x7KC|9zu=6de-_>wmm`F#;RGYHPz9p=#>-hw5vB8sX|2
zMn(dhr-vHq{TmVLTL-L3K_Ck2f!2E8Mu>){UKXI5%zv928XIfI<v{Eo)uaK*j@L9c
zv{!*t|65HBP-Q#7?g5JX-)fSAJO*oPhD||G|FfFL!!<P{)*#LQR#O{fIZ{*K+CT*I
z*giem^#T#pwbeIO&)9-MSd5IHK_IKy>ZZoV)}mPFKWeJgPT+CB+x&pSwGRw;4GeS*
z53~bi1#hUU^{WT3t*ei13<!e^4+;zq4+(<~3`ofi3=9tpgAWbLDR*^j$O;Y(_X|S^
z4KFpP0D-8?OT$C`!oWjHIDm?=7l(v}g$ESL|F>e)Ai1J|@KC@}N9Dg2QwL>4h6bfM
zfGGZF#T?Ru0<-l0SH*O*0t2&5K^p(FV#Zm4p%H~Z1N^sQ#08O|;r_oZKp<=;MjQ~x
ztjIqc_>Zx1PI*6+@*KhAVn50j5|^7FUr-SDD>p90KMXD;Ffb%2C?p^-BrpsnJUr|_
zZ@(}HZfi>r2n_c9|K~4YA`RTOsGR>d6KNsf?iTvbM7q@6ZXE<dx&7}%`b-m0lg=%B
z^Bi>PL#HX(i9y2jR^z4T?}vsL>TnQMl&#gE&6LYrYaClo3Nlfr{a&jZ9XR6E4f!hk
zogQZPCkO{0LQ;PJG4uTyt9(sh5OL}Hz1kDWbkWks!)xaGlVyp+Qs-i*;ML@&oB%H5
zB(bmH<K*iU0`rcK2}Mqn1X-jHG#GQI2HPjZUK<@d)bL2?tBk_=OmD7$ZnKwnIuHR%
zwk@@UhCV0#AcUqGPnWoDgaMj}h%BzXG;uO60bvRQKNgsoR2l9xi{+D!Fw&}%O6qJr
z%<l*Opy>VR;kMov$H>ShtVV>Ipj<WTcT=dYj<J0$nf**nt<ckak;QnE1J~Og&kb+p
zBA(7n?jW~T7t-UR96kM#f%Pgbg9bTU8Qu=CXwGP6t{GZ(Z>EUJSDoc?RGH60a_Ls4
zC*h!!{nTT0kgfyFi|LFBy~aN{gYj&(=6i1)PVg@)dIlrsO^J5M)#~@Law>YiYJD-c
zAQ198u7p)Lx{|fEeDS9nQj*Bt<{P;B>&_!p{$M#&_V<xp-kRExlh%l-;eK-DC4h&A
zzr545o$`YE(|(Jx;Ygs&O4PiViXC;UIzdcxPjaI{U?q)xihD}|UHeoFwJ5XHNrbW6
zy)pmEf%m4*jEe|ml<Z&Y#UJwl=NDIg`0CG0$x>OV`?0E8T3R)*EgNd3>Y99f?6ynb
z?GWw;GMBZ7d_QvsRqj^hdh@SgP_x`%qUg(E$-ZjsAb$o0C}Wwgp=pd-%%ER196^an
zAPK;0h9~J(N(_(beh(^~qm5KGWxuluFJXEmxEOOq!1b^U)aCJmvWpueJ6bkp2;6*U
z21^B|gvV>A{FfVDo!x1<4D{@X?`oF<HPdVu!RmjZ1M<kHN<=LAQ*=)WE$FJ9{tV<J
zW9b2g&`mmtQJZ8ZC>a^hR1RglaC!6KlS!;ZyVJO>+3CJ33$_M_sRuvrW{=ReQZpj@
z*mj)2mC+y(s{j0v_q=#ALO$JSA*js$&g`Gx_clx3q^7o@0h_+enxm5X&3j#i*=JWC
zl7Gb=Za>X#o|vhra`ajcS0=)o_3RLxYO1%DgA`a=ocqFIA56(o`!C!8EZMQQs`HM;
zn_acj^?C%qi5p1;Z$A*1PWhFKlyMs64A#Wph-4=i#2=>-tnv#1+=*s4vLp;eJQ6dD
z@H-rg(ibbO-T>r;{Q$}6AD->S#l<!(X{l@pNeWN_Io%W6R}0mKc6t{AR&MtAFV?GO
z){mqZ?~K@8BecH{%UH1OYzFa|Tis#CJlhmxG)mx?D2kOozpA4s!zsB@PRX`lBfxCT
zA2td85QAw5JT{3=Kep5ja16ix_8OcUmWKHh1&cO%Z9r!biNl0<J-7LLp7Z%0*oxFQ
zH>XKWkyaGP>b`m4MIG?n)W`61R6FH@#R+ojQ9PN0O4V~$AOz1!H&mRFO?^GL(&})P
zrf=c@UhNDx-#!6E(HQ9jN}AE(Kh&wYsMy${0CfwRN;b0#^v||q{C=2+KK9?LpSiZQ
zRPQL)MsF{r_u;{;NO^sWNO3YKkR|TuU9M5T2`49zA|izvfx1lD{Mhq*)q>$@Dhs#U
zKXO0@08y<*h-k9DzW%4NQI-Ce?MBD;+QbAhkgN42n(#|Fu<zczysR;4;P*#w1YV?)
zgq|H^!C<F4b>58qN2lZTkCB_9a{txk+}c8f^;EtU@+@V`tj|ML0$xvg015bG71e;E
z=yI{HJ-U^d6OxjWqBl7<_C0T^RGH2gX~q>Wg{%hDndIc;#EGdVvn2`fs!}M4Q^Fdk
zRTZ^`wQ47#Tkz>me?Y%t4>JhB?Xt+*3|AT?>zKfXy1T*noKL%e%=Rll#+2uFy}#{}
z*84{SV10zQd))V|tr1u+*D07?x=4T1K<&ikvZ1f(bT}B-{~V;Hs=Bc8qH7EliP`FM
zE;b#cHhOl0$S&~%0hU8KsX=T0X_<S61`j#V>#8~bwPutJ!gM^dTM9nn@$qqRWJIYe
z@_P;khvT7qrhu15s$dM@$tRw-b9h(`D$+oa2{5llk&MLlJ3KTa!EoYEymkw$d_mM6
zRP>_xalOo;tS7t&p1U~^6=|B`gpstCe72H!D#NJqbP3=$FxX5sOxaVY#~qCNsMZ1n
zVPIf(9uIOpt4vjD^Pz(cdZa?n#6N5u9nsKBeODbnxFV_%!oX#ou5G6t;b1QF67(&z
z$wrwi|C0YJI$}7wo)n5WA?ME@<`1gt1Jm<s;kT6D17N+J4;Yq)Ck^&UJu$|7!sCQ>
zuX%`zl#Z(512pB|K_&qDW2(i;^5OZJjGrGL5c?_3cipo_@ip3P9T*Ov%+`K+bE831
zc0zJYs;`4Dc3V=j4Y6@IrRTi~@V{x()1ekj^@``<wq579BJ&K5h+xb6TioIO>V8#e
zBC`fT9pmPl1W+hYKY#Op{g(}hi}YrT6|nhTnfdwoe?H$?2PJE*bOdbZ>u;=^8JLHJ
z1cOg$l))m?Vu5IAbFgom&FDLbo*$dS_IEv0RFXrtjM1|IX%;xJjQ?1C0?Ssrw--14
z4#gD<+<2a_OD@}W4E;EBHqOhJx;l22oTE!ifDF?!H>^1mBlu3f&c`F*2XP;t%SDC2
zK92&i2o7J;dsE|p8YZOW`9c=4AD59qmugCs<jw1_FQe6L4+S)yxV${d^vui^fH6qj
z48aJT?RZ+QF&=@uMl8O%S;gbBNhMdPnH?V*ax2i=MkUJWa^q_|(^e2uu~LCUGNT=N
zbrZJ6*NV${0=ZnQ(x`$gn2iedC-0Ana^N@Z?d*hwgnU!~QdUS<nq(UCvrz>RGwjTk
zC~bRPZTI!`Sy@>@CQJQE*t2uPB#UWvxgbr@sW{UvwaT|B9RzJ!7)LzcpOY|AHHv^0
z`y`|r@4IdRY{P809mNq_1@y$f`T%V366jVLVi*Vr2<T4$&!->2!&!lMVOEIhp@Ur#
zpa~h`!Q@F9R-lu#%=q`nLqFE=iYd=FZ9>(&+k5)l&Z!24)u6ZniH}mKkdk3Qx1vqY
zVKYTyWo5<qY~YKYwuWa5%bN)~BvU(FTJIPYiQ}HtOr3l@i*?{wQ_P;AMCRhg;hS0O
zervI(;q2T1t66J0p@Oo$w4}cKFdg#Pnv|3@%?8|FEmC>N_}qGEZnG0{k7DBBgTIVq
zil%0jc(lUyDcl{Tc2n35=P3`K0EGdjo1me;Bg8l^aJ}j6VWNpg%dHBBMp;Nn!AVF+
zNKX>>E4-r-`;zKnzs8kE<B^Hq5FaHN*E4baW$|A8qfdHeK3|TLnVBiJ-4oGRUytda
zq@)zI1dw}H%w~(@Qa3t%d{BoARM64UgY*_D7M*}R!)U-U3wCH|s3(509ETl87l1#K
z*$39B&<L`vc^cXgKoQ&fiEyK^S;WR6$KmQlc>^S)y?DJ|lPLS97r>D*E<WDh>1004
z5}-8!ZFaua+UFu0!K3j9WEA%o9Eg&fkvZ~w)&omCW0!8!J84ypWh!4NZ(!_dTpt??
zWD8=W)69_cl~*wqe7S@Gm`nM)P{*FS{ilFYb@=52U|1MIZu;_Y1&*km0H`*wxVyJv
z)`VFpzS&5}m|t(Qr=IxLRB)S1yZ4|YD-Ms9!_*Ie3~9Q1BcCVvc)GI4OOyX-!1817
zcKotthGtXGE~Vo$F#>2*BF~M~x-sG5Xy?p;DO)#U>k^QR$^z)S6Vi3J!Doz?BdHMM
z_bSA_U4A<HVkYemEgdOysyik13E42}jC@7k(2)0cI2%@eczPn`<W#YPazYX?8br|_
zKi#Myp!{aa^B(kB2c@dCr=(^6;DVkQTEaUR*Jn0BJ1Bz1h3l9>%`7fHfpK_CV-E~^
z9E@E<i-CizGunIGt0xc+lQA6!U%-<ai&3jMu?DSq&%hYKB78&GK6z1RrM!rH6@r$B
z{$9yCvJa^^vg*A<!Y2=B)IKU|92NI<Wt+Yo@feZ_KGXkv$^DcEE9@@tfuNW^$)r+3
zj|S4)?yet}F#)%J<VS;DzD>g=xBV{c{&-ddkf*3x?8Y;$*pn!Qg|d2pBc^#^xH(n7
zy{b?Fh%O8aQPYpls%L$u-FfQ+8OH9eZc%@$tQ)9NAAgT0{Nf%e5)9QHs+-j<j=?3-
z{EZ$b<?KFOge7D{3_eDqA8q+-j`v-sRN?ub?5|tH?%*w@7GMZ$1o78wEDr-tm(^B*
zi^6P^A;T--ki=fP;6yHgk9~+g7eXAM&!Oy_=^Y@Wue25%JP-b<bN-lphQYK>9OMAN
z>gfBuFS?{ETVfLMne^6XXUof6Vlsh3-n+tML3mJUO(I2{5}%O!k2x#}kKoHb(|T$C
zu4<DqiPoK4Laj6t6_*I9>AX%qM{l*`DfwAjL)XV*_*&2tFHdN!;!&mB)%6Ckd3aiI
zvui{_WtT@UR4Ob#e?3afM#iNu`x;ntUkpXo+V`)Se<GdU{9%)1F%3U}UNr6yLgk)b
zp}w(yG4UaHdf&1WG>_%V=vvT&Vj1q@&=f>b+gD?_^v^g|oPyajSW8P<qG0o(Z{c^4
z%wGVa`khs-zQmzn@H^J6ch}D+o2fy<ld4SAgHQ{5|KjD40=0Zk4LaT(ECYlYmXmrE
zPv9{YKIHIy0*`PXYhSI_J*tMgW4!+5CN_SzFA5hwliggIrhFAKOVCFEi(ZY`yHzHO
zc>PSlwmRz9X<SBerj(t>_jSROTa;c*Q_?4Wc&pJ{4qoZsB}Yqw5H!0xsiI~Jb3_3p
z6HH_5EHu(>RceIMl!<p=E0z^h?)o~D^+o@=jXoI@eejU^C6}^goG7j#XudLEzej$F
ztmhghb-Yq+YHCeVL)3Q3k9WH*RsK3fu3e~_vwryIDCg<CRlJB(In-|!1=Te$gYTHU
zm~M~C(p>W3Jcp~6JP$j>WkVVfK~&n)p>eAzNX+kuqh}aKVJCT2Rj?E4>I#w3p+)zi
z0x$hdrPgW9J(DW^nn}_?*or)-x9js39R{y%rIvU|!Kod>V^cz=qCW&LLx=XOEjgZj
zbr|HSnT`6+Od*E=)gbsh>d$t2dp_^uYnNWJtghfQO_sgB6cq1Q45an1U6r{9c5g=P
zibMr@6J4g4F^jNT<6a<1i>k5E^?=m3D#Nps_x?pYL#VMYD=uB@YfuSfs@=l2K`TK#
zuwhFfmn^?fG3~IE$!8hCe7YZ$G^aY^VD_iFLxbS1uzAd=(>*-$-mB{eNqif)X67Fb
zwJMOL?NUOyh+vxQdQ8_Odz37rDjE%060p2Fi9d^Kw_jH33N|TmYdCMgs%XaqXN`;G
z0?N(w)mb9C*N~lo3*_!3+V44xS<FW_oK`JfaN~~05nkLG4fXw8N4{Kq&^?hV*zSH>
zI}5s-`uYKrvZ5@*YcQSh`=YXEy&WE5W@7hq`&M=pa$(g~#?y8<CL`J7Cu<DDs0uy)
z;}Eq!BZK1;aK-D~o$=D5?cK*($CUW_b_$)r2J-@`PPb7{JMe6KAQ4VTF0S#u(khYh
z&}F!SO^xL_gxNT5@+Jf(%6Z!|$;RLkz4HjW#iqp+hvolrlvL5w@Vjz}N;PyX=$B#|
zeXsD65K<#Vw=ySPVx<Gy+@5|{?Yh6N#ME7wzv?y18iyZ&xgKJ1I;5%GtmQ4>Rc!L-
zq0K6S9WQ=C^>MQq;vt@WY_e*{Pv)E#c8$4C`}A&JDh&DrZiXTh0Q!hLfD*1c<%)QY
z6L&2`{x88!LuM`zi0t40P1M8}IM4S#QIo#{`Xj(e)&C=Ek_D5RI5#&iF^=Wzr(Nnu
z15T@mDG!F=9EFRATb;5v=_$}q92VA&$Kl;?c3O?+A$kp`@<U7<6^~l#f<DW5#q+?$
zbllY?m0`v?#i??l=`SCfZS%pZ(p%AL=G~4h(pQrg3rS)&C>-d)NJI!DiC{2RvcI6T
zIDOyIL1rNnBWz+eR>WAG|I^36q+>F5pCOH|bX5*7n8>J;gx`opnT3=n+#3|z72(=S
zYDB)$4HuFY?;_1)O#G{LC{QO}A(#Npm~#2L=!@1_P@@$-@|A*%j(9DjF**CmB14mS
ztkz+zR9B@%xI?@32bPr<op|%u4_rt##6J|sD(r?{ZfREK^a(m+D}z(oPb2zS4GCXX
zZWEK2+!VzjF4w|YoA(Bg_xf>&zGk$&P>w+IbV4h1ws<Cl-{+n#mL_D@-J(#uAx$$@
zxkYta=S^=OpWVef?G@TjoYG8xr4Uy70rNc>%>GB4Equh&)S1U{P`l)?u;wk+P-)97
zm(42mOlKNn&i6-|4|?U$_;~VPBtDtDRR_8IN9))7l&}%gk@DNVaYi$Xc;X_OiaTF2
zo5_k<%+i6@-Ood75Xi|pEuYH&tqtoH$^eYrY_B_YH^6+w7?=(z3v})P!tXTG>^Q?{
zp8hPC0hO%1=lYRCW7+H<+{h%0M*M-9zx8H$EOi(jKb1DtI_(*bMkx?X9G}JOVn4p6
z*IVvQ0Zb`iC9K==_H<I}>|g0AQEM`0^f&kuE8tN$0HXZ<b@*F>T9s<$=-<aHl+$I+
znjeG9)A1s;_*)P*(+jGJyj+Zev1Hl^@Oz=MuSORY?$TBM^V5ySe0JlW#NTLXX+IF{
z@hfl68T49zn0xvCd_9y^R9aG4z&6%QGOTw2Pu>Jh>*YfpT4S3b+_c|1U<-tuH<E7O
zbR)jb^90lgRYe(Xj>pn>BSI7`<kHPZ>nN$@KXI!DH=dKW)}Jswm)Q{gSQLhLxaN9Z
z159mp0h{}n@&{n9Nh=m3mB7qq)q2+O519LZEmNr|ZU0;;=yYq4)@I=ssmrP)z?gmU
zLnMz?L{{YvtB@4;3xp957b7vAa&J}HBKg^z`-U8xw?s@{jbtm~+?x%sviTR|^CJ3z
z$-Tb-cHkOYe!&rqfj)eeKfHL(y<!Nz6j-J(P8Vu&nMbgPawIOI&q{e3t*5Z+FVU-4
zHc0xhQd#61AjHB(rGrhj7t(Go6yq&4ifbOU=9Js4ax!pI^=CFa?ER8X*aZ+XF57q!
z-oX4-S<Is%ArFlbyKsVM9-yv%AW=S0InEniE=$G>+J%}6l^L(zex>0H%A~80GXty+
zuV0g3$1*E?*8mU(cNE_n9pehM^^bDhRnx@C6Q0!yS|32eDz5|R6h6F;_5kDYxZ9Hj
zTa(4=_y9X>a}zvB0(SgJ8Wwm21Y=cWf56^U#2X+(eO;h7>Ed)Vcw4Yvi@Z0|AJXs&
z1~2NsC5xUc#_mGoH(f0m6tp|(Mv6RAktdf;4%U7^7JQKhfcFBDFZC8n&yc2lxKjZ%
zsh8!<HXx<cz{^6oQ4o5j*WFSLV4kE-KeTq@nFQN7nprT)Jk~uSPK%10!EVNmsK+>`
zfqBnYEbOAbv9^ni`6#fp;d%YNd2e2GFDne)M*-bpcIoS<`5E{C)JT%ykrA?i`(;a}
z#~q?~Y8XP-Tri|@gh#zSt>(=Wu06b4a*e%xE?V0>1!mvc^CVxjhlg|1QoqmEL+?j&
zKK8j$s9iVifBgC^$~~3<%E4YRo=q$e4GDSR8tpapx-I2rK0pY{LlB8CMJ)~#`im}T
zw3Px)x0^>2!n3_nD$<q%xP;dLXb!&Z+T&C!%Jd6Jyl%*+`<`~Ym{q!aPheIu6zp_D
za7C9Se2(W?yvp~0q6~v6+jn)^X#wY(6f+1@;o(?BW+awTkrloBfM?rA@_{0uSNlUu
zYJXxd#1*`Sm-B(2fx&QFmDBzfDO9iqL$^ODt)!c1oH5TQmq&8uFC>cQlLSt?hFPD%
zeT*UL$>$9S#YVbeh)zRe-U+Q;IqvlDhubHV?;YPEkbcsR_Dr|Zd0(v;D`oMG;P0~M
zlAEEx36V-v3q0iJNB^5<G%R$``9?(vtk4+%d~Q=4pyjja<Kru>QEb@V`W66O!hz;;
zB;#3i-HNXHyd3^5ls@NU+GGx2a#;4j?naT<XYZgh<Cjo$(%ziw>xr5fSw5caT6=@;
zdd=<2!fpVeVKf*mx;v1@w#L>7&z`^~nMvJRcYeHIKte8@+tK%yUh|!`;Y8@sF@~+M
zLKcJ$X2B)Hun+i$H;2*E(Ixz5S9|40pSZg|%`q+}4gcWfX1wmJ^F+f1)1k8m^1{wi
zwMaHKl=<_j_3#L{J34ov`t*jP)&SmM2Qyq8+s^@}J2(sF>N`SYtC~Ok_tl&K{TnyT
z`VN9kCP=EQ>0=X0e%Zb@I6I(QaM<H<*dHYHI3)p|DuAnM0PT0D%Se7pY3Z;?!Y>-X
z<#NvV!L2sa^d-x@#7>E^lYVyuu?c}7<oZ@NU<k>{s_*_nZ>)P<FaQR;&vPh(>2Ol~
zVC<y7yRma&snx(7lyyaj<%!&V#E&F+iWYwIh?BWN`vXG6Jk+cW?69096Zp$f{^xj#
zZ@XNbkKc`u`ptTdF;~t>`U-CiF`%cduh56r0K1LHXWsyPT<Cp4n^o>dz-k1yqT5uM
z2qJgd9Gw{kV1_4U0%SXqDq)2lx<)*PXP3jB$R5XEL4J-BcS224Dem$lI#}L9UfTnN
z9x`u$>l^Pow{~x1u^wyh#Kc4lWE#rC5hr4f`!P*QSA4}I1TE#>I2cR|;xGH4!P*8f
z=s~1_FFxn3AyykT?K`kFws0BgmWk0xBSso}lsc4`s&AUM?GztmPVI7{D=Xm~M+2yX
z9G{L4bFpYzFLtnf#i-Cql}4T{MdDb-6x*y8Hc^f&_snB&GR{ST;F!I=!t_h5RFa~c
z(GAzWM$`v@GVZuO^x<*=D<YfMJ(ZbYSvW1Rx!T_Z*w#txK8S0K_&#Zs_0jmTFOmDz
zh51li)~{UbAmUJ-0C$C>Ld_42m#IKWD_xJXdZh2Y6DA_R;FJ?|xPFr}#a5foP}0!S
z#skk1AxiU>v^{u@n;6TYVF~7mO~8vS9?)d-*}p8G-8(l2scVKq4kE$0AssKOA{|by
z5NI~r7is~X|1{lq-!Y!`^(oKmW@H`8Zo}y+e{Mh+hBoXo%+bw%QKM7<`h)RcEXlc=
zj*gDeUJT%xT@nN+`{bhe{|&;DUE?7rQPE+_V7}Z|`6L06p_Y4(#tRA<=U!r&7+R@E
z95By+?5b}y0<GfFacP!s-uZ}|=viUgbIlH>&71xR_^JMZM{{K<n)Po+5~i>5Jwfhf
z00$2fy1Em%6!nhBvot0>FrAVdD3bsz-WI<$j(*s#mM=aI{_c5u+kk4wN^9QKg-Uux
zp9@N{{YljlYD^*PnP;u9YXRT`h$p(nvqv3jTGuDxJIyMLTR=;601+NzU;vIl8KigK
z(7RUM2il6Z6S(GUUt{_}y%$jmuN&$$A!{(V5GZgGu7KIo>K6_mwn>V4`W*Jv53q4l
z1p+%T_uikgzFqn$@0GY9K$Pgb76|98SRM>ke^5XwzRt|q`AO-Zh6E|5B?7`o+2}8K
zYDGC^8J9pFlWH8k=K%nIv`=e8w0|>X0Vx(zGW7`(El7L^S=h-H=K|_Y>4pvO=P_(n
z9h$CYo|LF-N7)$?DDV<UufcPi3SP@pl~ZOD%g_|el)>Xh=?@lG8=zwx_Z7QWtTZlT
zP9Chk4}d{-rqFp<$_$yU`Jlu5e~JA+csr-)$lfT*SL_bEqmIpv&5mt#$L!cnCl%Ya
zZKGqeW7{3un){!Jd6-$V);!H)QdO&}Zr$(R@9eYB{?&a(x;mLDl*wFEy@B;ggjUf3
zV*ceRl-yRT^#ond+|HoV%K>HeyD0|GLQe)^*&8xwK^d1Gv0&yN?ts5^b!#{lZr+3P
zANu4mXXaRcBrfY<=#t1M?t-v8>vE=rf;8q}32;Ag5462SQsdh}j{m1H^8{t+xDq9c
zhha}X1RRLP#rD`X7GGwLap~tjk&YgpEHgMjv2+$!f3O>7>ME_*YVVFO{GSIbpVG5B
z0LIy#;K!4KX!YQ(PkK)<`^QfBw^1Mv;&Uk`Q>-j;na{^NI^71&oT|k4_#-KpKW6YD
z$IR(@RkB+$zG`?}MyINC%g&rM(*ZN8V!&?%>AgUwSzC?KrPd;JGjX-6Bal%U(H8cb
z>>UUrOTxmkl-1u^&h`?yZ;8@b6gXL`HJa%m?xQ+=E+OQDHEL>;hB&YR1lj8{ywr&A
zXUcWa>Vnm-CYO;R5G4T<YP%OLRb6a`-O>u8hbr;5{fCLjQ-<&EYyp<Y&c~(2Jk3>r
z8r_S<ds-thn&<0{z(WWHPlk;CO$11In~i;#lQ=liE-q^zk3qbOmOQRpx~owqfwH)Y
z8({pXtlG(z@&mP~{|7)OqwA)Oo*XaZDAgarXbd~+RAhzwwZOO_xP!-8Q$;t=2|C+w
zo9sw+-C>;pI3213Ocf{mkOC8#oQoYqRhK-DnY8R^l7Ww#CC@%)6XtZo2%l{`U;h4K
z92)|sx5<+4>U}&XpG^6#{u+|4PD?jatRe_!oD9<t)4x@6fYmkN^0MydB)Gr_yV$lL
zZg^b;oDiFE!B%K%R1eRA#HukjJ~5O)MxwDSV$Bwi3dF=Cu4VsdMz;o*KSb(+T148v
zAq&cozW#3LqH^>3YhcEDELqR{$l^V9_zS}iu4JV4?0KZN<d9J&$}NE_ycQR0y4eD#
z2RU`jasP(CjqQguB;^2PGwD9fE6QtkLShrc56L#S9nx_c+HmH&vxV!Dd$M7Re{CR%
z*BYEB{W<fx%p#BEvJ;kHTS9P4OVEoZLjQc9dEsS+_5jt<Zd+5NuM9uL#>@75semtr
zuYskJ=EvH9BrZ!zF>wXCOZG|8LDbA4efX-YSp`9E_|~>8=-L#lpzJC}dl{4W>P7lk
zek%S93{@r?Sa_`u6P6Y=i6^@~^SHthpBw{!Sd=i$4QF>Qbr?E7Y@R3Jk{oW{fEyd3
zXllZmqMtJOH}zPSke6!PDyA!R8_{IbpUE^@P5f^+1I&0Yj_5kbX1>uK#wd)V6Nkg;
z7d7S8G}(1wQ>l5b1!pShY=$vPH_L?P(RzC&nZR!Nx6Y7|nH9qE6o0i(vrMhEcwDI|
zY+&k@(Zg>YkYM*GOYh%hhWDKozSG14Xf;I8*BTw``OM`8_$AD`020}V{EHGcg1O^U
z-~vWBD8qJ<W8Jv!o{~19y@1b8sO~vi^X^mNPqC<0zhItYAF`RS&vc}rq@DXfe#52_
z2GDEIC>Jk^x;TvgDWMp&B$DS_yds1vDp0=gjDpJ|1&uzu(rG!q<4Hd#%11Uq6B@NZ
zNs{lcMbC);=KR+m<qUS`12gB}1v5DGqOHxWQ!|><!>R0xQaeD~Xd#zILTm(jmD0Gl
z$U^VJq&>xH`dT=*KuT&wFV^puduLc;B`IRd^KioNUoyf9#m7BM=&s&cosJB*Yb%?2
z=p=;?XUK7hu^?>-#6P5^IY%b(Kw+GR)kQ9h2hwp;vR}K>T~7Uo-5>+>^?*wvyf9Z3
zTW&);2hEmDblUkmPWXX}=5x@rkyqR@asjgT1QPpTBf|HmIO43s%1{VN2d#R>IgH@*
z19z<vsPDeUvz^4=(SC{w9Xp8wpW0;AW{N}Kfb3kcm5$Venww}m_EDJd*RNoJdA^9a
z@kpJkQ(QEX&Ess5Xu0JM!pA(lMiFr09shyNN1_Ycy{p;eq?H13-G@>UFM~ZMwE#NJ
zTKvxQZ)`mWBfG6RL9DEIaF+4U>5d<pI9EO9KV(?2|4+a|>^PBkhNi-a2COh2rwOvF
zkE2Jilvc`%9-P&x9$WGZ4%3L9j;+5KU}+QcKkJ+zT*tMa6v&PnYMvzD)*FN!9P2Q3
z?YKuY!AqbCO)&#P?F!5bk(N?p=yXM7rVwE>Mh$J;&#<P%QRV5PDgRJ?9{oiKtmoqh
z#a|(E3VA5ON`{rvGKkw=1?{W$AKuEXI294}GHSN?l~l6okmP1i&?1F;uu@f1h?}ZO
zYRk|wZHeB(Xxa9FNXxr(V~e@nxoSon-bH)*MpK-zgQzmDwjVexsZ2xH+@{Wk<0Zh_
z3Dh#^2;lx)BDBS=tk+A2vS3ucR^;5Ed&m+~lmK@-XD%BCn(^ddA>-jX3lwHAAm);+
ze3Ek~!hmVqOT0lX5`VxJrTIZeau9&FsSI4+B1Bd8c*aN%Ixp&A8$gb+)hRbSu{G70
z^$fEk$f_!mFV*xAw}XCOER?b)ieUy($Ndm*h|%?G07tsy)4$7mfsCh9hRl%t6Z3s6
zyHR+&-d~^8$b(8*nSK-NMAaYLBZZ2@e3)tR-YFgaTh(kuM~IoXv&q5yFz`!GOfIbK
zgAE`S{A5qSVEC33_H;Z=9G88mwSN44a#vS<sK41FUb-p}3yeU_C7Y-gjY4qAllSo5
zg3lSk+=|o0a4mwa{^0G}Ax<&q6G59+`+7guw74JZ7yDb0X5<`W8J<W+>g|jQmu@Op
z+3rTU-;(_IUoQD;DO(*&whsk-$Cp@pQ!XZG$zP>g5>)M>0=%52Zuxy`A-X@?l_LKQ
zOqqj#yc2)9@V;lh2!&DfFvT5ee^qv1rGI<j7)&f`GLcPMI#0{hEg4atd}h$Ftj{xV
z6qg0%L80$Hr<z>+aqvF+lI_2+{Si+wt|uyh+pC^Ulu2$eY0Lq0C=>@8i`%{d5?*=W
zalREcdCH50uYx&;dIoFEV*2v>T;uRz#V{&II3JL_vO!1?L_1r=HeZeQA>m1id;@-G
z+VpdJ`clTD!PF8=K)?Qt893f~YcjOwG{930nuLCR0T|u(->UxG!I|Y4<S4I*)xfnw
zK7N5u#g43bl{6rg1qP^+YA34kv>~UUA^FnG>5zlDcHNoM;o*^R>H#^6C-zg%Nm~5j
z;l7BFI@L&4i|j5tZ|eEFy6CYCjXL9E;E31Gs;ksqf#Qv8*NaYKzigLy>;sd<A`ZUf
zqrWG_Z}Mdr*Z721(G;)+HNFW=3En62Ow=3(IokN)QqH##0N^A+n~1&%$&qAi?)_&I
zM5_KBHkSJ)nTGI=ALgg8DqVbzZvb)S7&1Sdk<lW0$hU1Ldi6#HrjHeG{gEy`!M4y_
z$eBTBS!l-Q6LCb1yX?~<!>|^Eh;DiW>$b0^BH4DKLLF>|OnNJ#?hV3fDrzy6Hpo2b
z)ZZInNuRz57XCa??*ZL&=VcmEWqUd7wa-!fs32Z&{=GZdr8j1KDKj$yTW4oair$me
z;UOMXiNqNN&k+E<<qZb^ddJxjWhg&hBEblbZvGOriGETXM!B*cDv^1~OCI}KeW%Fa
zGBzym5$2SH6<cy1Cz?L*Kf(%3BK%>UGQ)jTmjbLpT^0T!qy#9EEt)jzkmcZHhC9JS
zuPAjX_r6XGB_0z;8XiTsC*Flu?vHj;Vxh(3tx7DSCT6ww%{p(LjixhfHDwsCw#%{!
zqx&pTk_(|_|1<Id@ujhgFZmJza^iR8U!<@(zloyv@lD1FEyyNTYox`!9~Ir|$NLL~
zh?kAgI83venG0rrG}<(3nXB=^#NO*D_t$1&T9cV!24QvAjNi0lW_JT7lPW`1F&|j6
z@2{d|?ytIFdU$e0r;OaH@ZR^lAS=PX#;`VJ(xeJRNgkR6yFKa8Vm2`Q2i!<m4K;4F
zSyuZ$f{}TC#VrY!aOg_uFYb%IWwR#`mqL;5!a8_|gd+_J^pIO(jCH#ZZwQ&NM5rYH
z@J;i02eBADZc`4f647{^r-bLJlLW0}(Ys=3PD}+EgB6wkp`NIR5OEAxaFFWb>_lr&
zxs<1KX;VV-=GdrSkoQ8H>c!NU^M~cZS(_++o?oU6McX`2+*44+-4WE?>GLwi);jec
z5Wav!>iFGd(`uzi#-ztNQy~e7Dk*?dz{zI2`^`W<&$Lr4ik%-|X0|97L?FbFTP;Xt
zb{M>0dEj`jcg|lv8SMR=^x|aI-FdOzj-|(mS7jd%@k%Gto&$o$J4dy5C|7d`^I1x{
z^vmk}*eT!g&*RJ;hC{5O1k*Lz4GfJSMX6-L@SM%Gx~L>=om0^aqCV>J=~}CA*Z3jP
zXF|akIR|bUojS>y{)K7yD&2}4ohY`Z>ow-v(oP<b6B~*Zj8Y{f{shcuFO=8jy+Y88
zv3wBW$M#jg%Kd9{7UQ(TX%Hh`Ll%M?`YM(Q76FM;%wSF*!2b-ua)d2|UQ7}Q<Pgy7
zpr$a{k&*$Mu3*?SwUJ*`!{6=G(b7A3vrbX}c3(p7ohilW>3kEN0%WG8QXJ`F<WN@e
z(n*HByymc77;o@x-x=FT9I`(tgdP5o4dj)%BUIY(qs<vhYSEpqRbjf7j?hfH__qJd
z<o%TV>S%pm)YAZ(vK=*N48cghFk%i7CW@aZmG{&hLSGc2&~)xkqZC!l^;6VMj&8|9
zoWK<S&3*D|vxX-FPwkm09DHdG>6=%-Vb@Qud-lzDUn5)aGa$SuxqQ-f_KOJaD-9`N
z`_d_UK{(DMrpPZ(UdV=U6&P{c|448MsIM(=k7x3SP{m%LAAEw-rE&36yK};)Q^!V{
zN+n*Q3|7w#JL?3s9`L=CIg4zP3+}S7t#=`#5TuQpoA`~bnjzXyemFM6{mto(I)|9r
z47pUjLu^v}B??O}{18{;`vOm>gdh9XEpI(8Yzz4eBjUEEZMly(5mP!GtUuYmsOvd_
z?lWD~cx?JjL*$PpVJfFuQN@pcE@ah`(_b99ORX@Z5~qd+Zk+1(xRuM`x3_g*hr@Y^
z=#om2pbDflRs+&o%+xTpwqcCJ1jK_$Now$Gq?qJs7%M}Ua5DPOM!L*CL!!lE3jJG(
z-VtNXDPX$6CLRSR{q%X3JWqMb-79Sw`X@tir7{Y#VM%EDM}a8DVRFb_W0+|3d*%lc
zj`&fT@pbL5_*y8#+~6<N%*@DnaAE(FL#A7y$R~7}LiN53+Uk>nv2|;e*WNCQji^-O
zBL+djzpH2lkO~wJC!10#C9e+RxQ~8)IU|;)7}SOd>HJZ?PQP-OpnE*KGpVAO8uc5T
z@7x(`%DN#$=B|Z(<>o?I>sa9U@}oXd(h;AHYtG7XFeuRmecN2C-#L7lnH<fAklTE}
zeVCbUN7mJ^V6$>efsomGd+7WP$k*x{9&sTH=cMvGEzGF#c#=G<;+ja8DmoiaG>FjO
zsar-*K-(RB-Azj&&WcK#XE%><fwKNAVRix7!}!vcM$R#mlMk<ct}lUpj>xmDEb1e7
zlh1T-;Rn*2+0sYl_<=;&2Vm&^DbeXSyiZ)w6x3H9YXd~_G=4lb7j3TmhDwc|Dx(}!
zK&+0|<C(Sos+Kw^S2#NaB4yMv);jWVx7q%d$soB~C6Ds>xe>>_Q@*<V(*OZ^V+3TZ
zC};Gw2E(Dyw|EbE9PTVwRqvnpg5||LR|<rpHaP2FfsRY&9~7HU{y&d9eM@{H1L^cM
zZ!<aK;n8nOm?gNuWIs@rXuD1F@lpr|%aFF!hqwY1&6mHAVv}gb3FoEr4Wwt@{)aP!
zasDR{`AJd5M!zCafs^S(Oi01^pJy!_N2fx2pZcT>kJ(e%ce+{WcPCbmx$G!Nr7?5a
zb6ZpRP(D+qPR#j_M+|1j$EPyaJ`~Ot`LEu|7imJdU9bD!7%MXh<ohW{h@;&r%}!_$
z#;?sdG0MOsIEUn}-Pcf-$_lV>Dprs`B)<~fnW0n|R(hY3g{R5u)DJ@lFeroXINi)s
zOi(&%olN(ZTDQN@3XuK^N<P~cp^Pkau5~CoskRiSsCAGz%^7CRNpf`8YR$;eYLWHQ
zDFi1b_V5)XtnwcOB8h-APaFgb5=hujwaB$6jF$qxnc&=F*Vun#O!l>U`!e`k=!fVP
z5D(5G6^(t3#A9($b&6tR=t3xvgTL}2`jUA0&h1_5R^i*y?6}VVqQY&qVi^}-UQ_>m
z5GTWEuUy=;Cg5L6L%Z#@zm3&M*X1K|o^ofuw;~y|V)p?w6sVMits=h1g;bx9cQ^1O
z;v&f`jM9ahwQ!e9=3i;pB7=OO3>eoNbgQzTg!j1QKEH}0rfY*qUouaI#i*2F#2Gsu
zcW9A!tJO*UJ{e6(=@czKrS}#;{CyU!O26{%FBZp&|6xkAGGCZL#t*Jof4cLWWPBB`
zN`LBHW4ElMSKf&U@~85*X^tZ1aGQLK84)?%NK;s%rRqHl|II5ugo$taEuBJ!qtemw
zixuta5R^gFG*?eOElg4Q&*6A<0yeVAZ1M;S&L?TEd!q4?zoAUX>RfbTi6%6LOuo3D
zw=`~c=c>th3u3MdEmF8PJJ)ERe6LJ9pbVwZjp!|4LkG&Q6H|}FGC9wjv6p^Sh_>#n
zeuR(XY?8^_(&#s>)IO3;(BAcKTAXfV_3%VWalsIn=6;+0H=VzJ2VZ1A$uM_UUQ+xT
z<UxljL6uxWWjBF((Mk-rz#vX2OgbAoP4`+TwmB9*%7$UnCC*-Tq$-0wL@lz(;!9rq
zesMZtvxq?%V;Q9aSr^HW9ZkTyfcta0tb1lMlvLMkJCJ0$ywdht#=EI;JSC*gWI68`
zqKSUNJO8-i>q1sfguo!3jA_eMWuRjv63@8&RA`=xs+h?yzdiZHAA%KnBrq0laE0DD
z#DAU;UUbxmuy=gZg8BV?SG3#o#uQJpdSz%GLfT`wp<^qLA4qg2n%~$)tW$#o*NSbk
zC9?$Pdy>niIpyo4?3mHZ4}JJ`1#}v_VdRNPF}a(K8~08S*Xn;O)anX0IMu?UE<DCs
zr6Z(`No0a@&xQ^a;#bu!W@FG14pc>%{yAL9U(~AnQXswK3gNl-<in<o`9+RWPZAmR
zbgD%&zZr#cQPAsZM^P&MU2JME?xcq(Uj1*Py<S&Uo1oxXmEMnQh@C6=DckA`kM`4o
zKYIEHxrIIp#3-R?>hFAA#*q<Ow;`-wm)WjLbB!$Msato8qxVONHtbvcTjjN*&lVAK
zy>sSebgmII$Mea!f6j&^+bTf)>)9qohQL{=HG($%Bf!@qsljMLyfpU|<o%+0_|i7-
zRV7VW@zjJj;_28Q%2GhBTCSOUSfHA*l@iHaR(JpQPsyx{O-A@-QJe;XlvG!PRxA=z
zl~Y7BM;<}rOS-1`W>o5R#25$VO&~N7_|JN8w=Nv-BY|&Ql1SH?y`}DbJGXy)SCR{*
zA=q&!A0_k#ck8;7OpcnyKXo4nt6H_Aj(e~*qQGOhcDHl)_gDAAACC*mTU_y`Gs5$^
zeddCVNplswIEe!vil6M3+qA`<Ko(~0I;{AQ<HoD9l4R;>)ETUAq{R%%F+1}m6yQxQ
zWWi>C`Ke|8ppl-oX(x#@uh1&*pHvSXjT;7#;l8Z)R?-M89A!GL+;u8~M=6`%1@syv
zH>NKz+37}6o{nk>gEsr1shsf1QBk<*6aIefLo+@}C!-YH7ov0Wc4xOqF%fj0{n8rj
zvbJ!R6qJ0EDIN8Y^l=KmvC?Rd!t-K~>A_+w5ec-2e>Yc#Jy9UWc6AyTt??eRn%y-I
z|1HOEt_FIha-_^Q@T<~P&`P1knpKaSQWOR0FG)<8BVA9Q0(?W!E=0c~o#?aoK?5N8
zWmD`{$8O*B-=F#j9-eL(m+LyDQ?Gx7^xPr1N=P^KxuY#{x4*t@J94<^C84H6nldbi
z;@)cojp%r|GA%K^rQn4@$V!;;?v}C0s-o<h22$qp(4|}$w|cbXyo?nrI$wWZT}>M8
z&WmEvZZ<qyswgBQb%8A>JfK6A$!y!`>avellO1X!>cqBtOBB)KXCki-!^>(MVHLVH
zmgbj*2@ra#A@wITBKEt1^lf<A-1gXnrSSNn(i>-wd16$<gFekre^Zu|m}p}HTPP5t
zWKQ01tBS0c-FN+kH9wXkZ2bSx`|^LY#`F(;K=F??!Tq)=TmIi1KU^S?qTB!J_<;bP
z1_An?jvsA@5>wzwod3)5lc73)H*_(6KI3^o3NMM-v*Cx1WN|7Qi54kc!R!**VHtt<
zG57`Ot2}%`QisNZz?ar2aBUj@@CG2~EjMoUy!EH&?Wr9@lP#y`r{n1_=iCnon@^cL
zF4fwZiooD;X)=bZ%vOIWD+u_*9=02RDg6a;+3)n@f`HixE-2t3OfwMh&<-*f6Zj8=
z-Om5dCmZ`0EHMbbCE-MC%2IsDNcTQ>@zEW;k>`GtOSZVz0Uve5cJW)1>hVk%Vq;so
zY9a-#YO6Q-dT+|ZOMz6$l=S4qqr1Obt%bH~1wa@W%OtUb?Us6nmmV{ZVyUM7X_?Te
z<4<W}7yij2pR#feY<i{^zk@N~-jeIhrifxFb}*-xi_-7cNH^cG-&J}gb5K?~=B+Yz
zyV*={{;wSZI5MK^1K+|1CTKbvoFbK;zx=QdZa6K7x@b*YDO?{clgU58?IJvWU%!Sq
zZX*&<MjcNxMh;|!1U`6Z#?H`qs9xD;?^uBjeDlTMW|gO&E^DG^8=cL^s}-3<H)Jh0
z2z2Msfz+Ki@tqSZWvfA_`5qS4+wC;A<xwc<czyr#S3(ax(@0DPUSDZ{6}|im$yxvK
z-9q$5mu8Xzix>juAtJEkH>A-Pe?l@jTQz?8lMnV{&{l>3Zz+*Fz4F6EhW^G7?tfoo
zPxZ>k!d+`-;-Xl=E#bY!#$PSF$a=!jZXN-Kd}PR+Q1#J`FRKEvU->u4TO$`U=yf9i
z>)A8Q@$lfKjKZXR9_gm2RGlL7O*HSnGiENgyi1#Ee?c{&yz-~3lFvnB<w~=P;O_lc
zXl48107ZRuB6nu*WG1i1L@KL9rDkpKYdfGuTg=LFvVVLbLwhI#wI5WbzX{7}vM9V2
zi9}+=b2)MU_d!cmCXa{aSHf5_bM^9@%-7X2XFPeG?UH@pFUHzuGJns!AbyQ5rhWz7
zLvujISwbnFQR4l2CGnq%)d-jaZrTBZQpp78rgkRKE<b+g`si@a)TB38EQn)jw^nfO
z`UJD`&*#szJ|O||nQrUKdL`e)={H)PLe@uyxlnxDLivSa@7m&j)D&IGtUB^9d3mHC
z?B35uRCfT4tORI(r~r8e@)H@{toaY#mpyRvz@Fu|NU!HrXY<EvAdmfw3?udyABd{a
zqpzVZk#h8|wCnrXY;Cqn*MoXy2TK-pJ`QBU+#X!}toKM9#c|`JE<k0vOYk_UY@YLj
zK~lbf*X7cu0#P7@Tv0PHr~&b*xp>?T1r|W(cLwt^t6ph0X^_nBybD;;s&ecW{d_4l
zb;uI*U5eSymCki%gFT+5m`(&lG?m}4isjOa2;KM4Zyo{u+oFn_d#4vlSiq;@3axtC
zVPe?sD(oF+^1jFb*NN_+L{`NAmfK)Qbb7q1d5BHwk4mFga%7F69TdD&m5<SPGFgTV
z)$=<{nHlPLv+UnR?*JaXtT@~I2ZL_An~s#iN7iAH%Hx!m|2+us^UT*6^r>bG{C-`;
z*w-z`jG=3^$KZp|YpoK;<5*LqB8y)<mh}a`!gqz|mSS1`o8*&c{+VWTUUpTod&T?n
zm4S_1In>5l*v+MLfe}ycT#CUDfapH|@wWL<3Mg^P|Ai83=(`h!Wz651`}b_-Tqf&!
zlv*s+<UVCBCdfi9-rsWM&$o!x&sAHh<HSs9$N-+BJEOr-iZanIShY|LqYT5U(qRe+
zwT_|v8zN_S`R7DD+vu0IP6N#OKT`P4#?ce=DK;W};J-ZRg`q<Wi8p)w!1rluJ1#4q
z`~9$9yilccLf1(SQQj!rjzLmwzwW$P3V37~_W7+7^i${=bt-KhuZ->vUWw)>%fDFe
z*33}*^tr`eAwP~Ipk{CN<AObu+WHFh&e?bENNvi}?V4<7bs~Q)$}C>}DXaRzp{ISE
zqH%*gAYslG7#<!k|Fud0q3!ABa4}?O6&Ae61@QRh0U+MDo3vaZ5~B>-S0PFRh$zv6
z3f&@s1;XC?o1(O!yli#G8Cb(e1kAKAiI2Jdz{*9$sOUS;EmLxoyEQGgiK??*Z2R*x
zI_Wox)hz995W<%bbG=?#oMplCLRmBm3(*%-0wIR6FL>3DIWNb#zSq^|2Hhhq++rZv
z_UtEyMq<xxv~PG9bdPNFcErkVvh1Z+A*fQH2b&cu+p?F-3JQdn)XB$m&v=w`K3x0s
zV|()KPkd;qbq_w8LzEjlrJ!sNq1zUOGr$~7uV`=CTPD9R3wv>|%yiw*0)TdogEg1+
zuqb#1EgxlZVZP*F-!8?Z+?kUwR;FxT2df23{+M}}zCWxg9K4u%RKlYh>fZHPE1Q9?
zM<)(XjmV%T3&p)bEw5Gn^9MYo$;T^0udrD;kjvc!=oSxpURQzKv3fu`kbwN#S`rwR
zQw4^BB1kzc7smO(u&!q-YQR8>ob%Dl!0R8g$*eMfP*_$_Kwc?~RxQ^4HS{f<;X;3B
z#{GC(ORu!}P+o~uoq>5n1ZhHHO6U;zYxgiNxIN?|`8epSe9JpelVH_h(s49i)@oTn
z>^??$0;XbuaR;W#M?~8ja81bv_-88d>r~Tls+=h~V@H%TivV^G?ms6;;|OJXNoCW@
zs8I0B-AOUW%CG!{Ls(C=gg|@hI?a$TRXT+mKQH5-?Y3p_mQJhZ*VxBY)8i<0srUIH
z%|N6F$GgjMS3$%^lTp9teys;KmjjvZwOg%MLKxp<k9xKgQkaB006VGpk<z<Z_?N=U
zvBf|KxZP*XkSj#Yc7F&c*b=}3uyBipOvtz7hMI`$ABI-Fg`Y{(a{T9EUZ%^-EW%Re
z3_XU=89~Ghls6q(d~$EKsq|bR*t*DJAsH@DQ`_;oE^zxgbbAwU_um1)$5K>*N1=%n
zW-_ci{DpF&P#}bRg5!(9=a7vK&vF3PrF7PMm}T*Jxm~7G?0iDb1;rc`$0zG{X%-OG
z_hKfms_-|n;^4%=(DF<zvDFwj$!H}~Ugvcks(o(mkB=e#l=1W4UhR4|)npE!+SvWD
z2J9l%cL2SzL__cQJPyRq+|vSGQUwK`r>_ROkQ+UKFD&;H*A{4(Am*o}8Swy`<kIb1
zSm1F==w!L8b8k%{2V|sk-4rfxG>=c`h9ZN>>osN<jw3c!%lrPIGUHnI)nM)%tfu2?
zzZ*sK7XhKC2zS81ky!yf3?{8s2V9d9KHIMtJa&r&p2sbJwU75Vm3BXUw;ZGl%8K6M
z4R)jUw)bFri!w^O0eoVmf4#u_STw>cCfqu?_{cU-%&|kcXjtR3-Ai%${eEcL=>iN!
z$AC+hCDpTMf010DO0NToH(I#SVDtBbn$b`cJIQw*5n@qt_?le_Q=h_G=<M_|h#(MG
z;|**EM9K8Q*EF4+-t(oJ)HaK5ctC!cRa2+RIy(;E$zSTf`>(ry)X_ZV8R#{mvB-*M
z)cpYXfzo-l12E?VUQX-fuPPf8M}7C;4lJXs=FFl?Kg{)&@P9{Z$4%ja@kY+;Lql$O
z*)591oe6W~7a4ub-Y%GAJJ&-WPU|Pj@NHUzI{+Kv-Fw3(S~ZAx254XA0e7aZDd!|u
z+BKhcUr)A_e;c@%vYW!?JK}>vy!Nm{YQ<V+jRJQ{7dOxE@&x6n&bTSL74q#&uG#&B
zDBMhE(5tqYJ7#9^r|f-XT)-70Qn-Yc9-Mv|GnuV;@yelte|}I_k+%i|{TE~m_fN~5
zx-q?ss_8X<B~K08Cw`zpWz$8fSU$6?s$~aJBsKAz%~3hh6a6v`7!1z`*octx&9<8{
zHTs#HHIhawQLopy{sf?a0e;7`A(E3-<8d48KPHN$`}~=EbPMVKwhiD?8TofUS_UG&
zIlFOel&O`^&1>jYMhQIsf?Bh$@}3B*$la5pluI*?pg%#{iP?arsU74ib98`W(;z6z
zU9@(JyT5x9OvzR%wa58>2vP8PR7TOpd3NdTu<1P>ujJ&VKZfE9?Sz26<mq(8&3>{T
z&%mT`lzuovlOmuWTT5$hQR!EHMak`6LYPc<F;=0lbN6+2XKaQ(S@Sj=V)*H-R&D!W
zmj7y!bS?A2k8gD}Q+Ryg^YLc|xTQs2->{At^6ppb*b2#eE@Tvy=(sEo*nO_hZk-91
zz;~-XtWC7IID7CQUrApjsyZZbVSC_ON;~BPX&lT>P?g^aMOx1_Lb+=6h-OcKp^P7F
zeO}Ug{w-2aZibdM5CM>s+~b9xwrZ<Z^>U*LWc!I0%zHTg`ILOcgxRdiZjhg&@@u?3
zK~npNdwM1ww_;Cc&vnM>=y6{ErNNIX__PBfxylcu8=wbBg;SuNn7REOIS7$LS2suW
z(3ud&B`4K9gbJ<RgH0QwhVZ+G5n;iDUuWO^z?(}G)1fY$o>uCzcGv55`15S`yYa8)
zwNfJfSQh$+@K!igV|bJ_4r>c<%KUL!K0L?okmpzBUEbAr7CbBl1}=m`{L0>{nOQe7
z&u_Xq1&Nz6j{OCVt^}Kesc!@>OTrxh@s*~wGyazc)Eb!w!#|#Wx*D399d!kWrKsR0
zuHgWWgipYrI4kH*>Z$Zduw-4B!!IFd{Gy@`6F*eGk;wZoSG@H;qvVGihC_9Vv!vuD
zFcDd-3fq4paz$ccfQm+#Ft0*}R_&^~_DuL~it6LeU%X7D?bCJ3BVU+J_|+H0Q^3wN
zH_5mw%`<|!t!qDtW8*PQ5?tc*ev>`Y4qgn!xsj+)E9v09S3~v868g;sdG%cIYIn;p
z&o-C*uqj56`_pZ#k|fQg=mWse2J`Gjuqp=`a*cGw2%R5tA7&WQa`wDPIn-2S;rn_h
z!HPZ|?X4<{c-Z1J6f2C=KrvYD7ehrl#7%VXl6Uop^a_JEL9(NMgb7nZ(w*Plt}A&?
z!;9k_zZ{8`z&!})67f89Jt@66Hw*SscQu5<CVrIUN3o1vxTbOMe+y{`cQ=i|O{u|q
ztyb+}{mcT7Ah<ah-{p$pRHHU}W4*Wedg8beY5OP*YgYc}b6<Cy<QF(tqwz7Igv$k@
z2*}6bL0BurDuQ>_5Pwpy?f4EHK(w2+Z{)A{(@6xVU@O5)L)yJPU;yBSAW84Dw*wlP
zP&tRV{@w9Bjd)JKVwhTqk<-aTU>lpPgh6g!&~D(`yoAx@-!qS6=t=5L{1{B(gzQ=k
zhihP==!tIuPQGh8Sb=HyKIuCjP#?~oAC;nVw>YtGiM`;g3vTJ}%_*o?0}Pi;J@@7J
z{|u1mE^{hmmjDR0ybo1yt@;x=Unkr$EGP1Y%AMuHU!vO`WTCA7SIDCdvw!bp;idsJ
zjqtpM3ksRp0^Y4x@=qsJ>RYiAgi8Olqoq5E&k1lM$USXNMDB*~i%k1}k|e<)Rbhj~
zh732EjrHQ^8-IEXT>l52Jzg@3)`@oN4J}I!C?N8At+kjb2=zsku^Dc_IO&+J62_b_
zHpjmM6H3l3MKHM%g2NnUE`ap8`B;7qjo_ncZ5NGf>W>!t-8tahQ+MvUmG0U^LxfIy
zD`4^O%%KARPOs<PMq&L>X3~|U`q-(P|F1|r7W99RHaJbhm>N%IMZpoR>xxQR4b4P*
zAJcm`gsqX@3_hiUlujFzUvi_fhE^{cXK;cgCPfjLd$3<TpSSP2^^uuTsCdr)cr|0U
zT&PIEKkW}~z?^;PA$&S(-h74Xl}FSSv@5_X8XL)xMxvm>`G!?;fuU1BAnj^B*W2K_
zqT&$)?8zr&4~3s}j1(r2uN7eyc)K^iGFvG|QBnx#ySfj=dbU$?#$qu2Oc@VS-G)&2
zTBYmtl_=$?oJCxO3M03EA4MdhIKMNkaGG=%YNi~hcM!TCfhqm`yEs1_U&6MA2)N54
zN<0%Rug-epLub|)+=)ut!R`ZsISJlK_Dy_4)Xz8R{Qr=EUC7x4pFrsY?=sU7;7SGZ
z9<mYR5VVlLW`~n}vqK!#jq3;uZe53SEi?C^rn%jHIz`!YhL7Pp3ujB-Bou!dgR~dq
z1h`*OtfPAHQi9uUV}UOWnU3SfMrF*A>`<`Cmr}~7z};^$-mlIzO7PvS(Rc)djwoi$
zXq<pg;|8(nj#8o(jW}3wGv5PQZRlZ*C1KzJI1e=*)?M|4?H>0rYRI!x4*zD(18zQ`
znzyZn{q*%0hG}78VNt`6U*bY7#JS*2K#Pi}7LHa8fM(S_Y@Z*Smk2j4D-wIcN9@1(
zd=`9Xay$lXAZ`RCk6ReK32;E3ImVq?&sO43$d{kH{l9gUms`d>02tfgk;<CC5jUnx
zLx2M}k7M0w7R#1+Gd{pUAN1L*w_c8%H2VxFI#oF{opdWD1`jPy?D+I7-^cDwJ6UfF
zHX_p-?pYh^p12Z+pKsJwJZ};oblpc}u~3(2^9OS&t7V}22W<iFx`FtYXrYC~%}BtT
z?FNBUG#H7G9Y9iZ1yLg+MXm3B9tg>BXXP93X`Q~<0eR(Rb*oi`?Nvf~&*d0T;N`EM
zeG6DJI+kOej-SK*LP>knz!FdrBrHDfuRJ#(Q3RJ2bE<hDbs!-d0qNV~GLYl}IKAIw
zegnhp$H3@OzE9XeJxNR%sIccho3TmsNR|-*O1@7|;ZKe3fOJ2Cu*=G>ZhbWgkr;2&
z0>UMJUYPEu?>MegPXd1KDxg&X;|xfDX^)-&qx`@<MrXkU9=vq((98Gv3N~d*#4iT<
zndtp8I3d>j=@+~)>KoA1W6mG$-x}bfW!_UI2&!U65C(RW9j9=mhr^%JO9X#;xdf&n
z5Gg1H0ql3C0i7|40U#ve&yo5bn^@>0H>v}~aGqHBF?2iG=vbq5)3RwI9>EoC0Xcjo
z&mhoIlb8a^dgs^Ix%u6z_A*2ekr{of^<@}}tY)bW`)h`TKoD>Yd7relR66?gNSUzw
z-fiTwy>K>25Xk}(wQZYi*5@dawex_TBx`(Mj)>fnwKdEaRN_^7?<QEKjcQ=`TC|rO
zx=6A}Ga7Z+&EJzE+&NRKLbaI_$|AOoGt5uCJW>I_%35;S4j$uJAHV^U$8JEX)`@Rl
zgf9xYS6!@T`*`0}YTp4Zz+5^4Z=R>~jO8XV4sQ*bT7VE4QA6u@=AJ1>s3-?re8l!p
z_1_gVZ^h>^4qtP8NOAW~fSi25x?$IcVzFx7e3v*TvAPe@pYjw#r(SV+VGw)7D}^)4
z-9AHM@+3y~Wiq*rsAIX(c{y52z<?R=?p;CPiQIwjUAQnAaI|pWnaE>dBmoHPByIvr
z7Tu(xVuQY*Qb0FZup<T{NGOcRXfC>d$}VLw?uVN*vTw$-8+U|yF@W9L0?wdvW~OV;
zr&A@?oj9ZaAUtzGRfL{qlt;eKrF-fL^A|K{TwZN+k2`D4$}S1T<Y&@lGM1F;;WU?+
zm7N>W@<h#l18c5ev(fIZ9+n-75vCE~Fn!3l&pqnC9A^B%NnT7!)$S%>cN0(50$Bbh
zUI&0-u89C73JMAuUEu_b;OnPUAz2F3OzaqXB%9n0Fj<?e&9LZAp$S*kw_=~4zn_UV
z9m}z%qBXYw?E6@PMIPt@g_eJ`V_yAdDd0Vv%y1+6eH&|k6AZwoFkp@ArcIz`N|Y1p
zq+IbM6+1lcSO#;fSDVTJ9;=oH8dsxb7`K!IQJwhC=PBO@z`2Su34VnWmsnt;)%A6Q
z5d3(p7o=IC!o0iRlWL=f7pj~qm*<93#Kh8D$s?RGiCGa<VNHaS{|H3FPFrGGoYQr=
zMj{y;pnKZ<csENRRNcAU?t8OpvBOk949Gvo3Uvh$LQcg*3^yfmr(;HEh+4Z##pNV8
z1IOS@76!U{3z#lpGl+Bh=2Atxs0)HJJ{V8kCE}T37$!8SW}Lv3z8cmeLHJ_YLR4k;
zL=(RDll~jKL(#**@hxgn*MFm;Py}LE3W9G=Dr4BxwsM)KNAkumKT-mRMC2LVs}gR?
zSduWBZ<^;3@XJ(S8#^HR5&@`pI#VbjB(=y?%=L?QAnf&q<Kf>sM~Kk=wxgNhY&HS{
zPY$AIsGaHTqvQ%;Pfh4ME_es*=e0B6g+b%M2FOf07MOz3kfx_=yg*rRxNZ)>YPq?$
z!{o;0K1&Vp+>PWc8Xv@521YpRAf<dC4m*w^-Hbh9>Z;%^kwz71;on{!Ce5+)PT9Xt
z>iN8}Q($)!u2(s8HvW9sptSw${EavZwb2KIZp4R5nd95tzHQeV2i5O)Klt4afYRjU
zD9vn5iL<z9Y%U`}_4Y{ZemQseyXQpkc1nZ>>z~U)7eKB)xF|1!9^Q=YE0N#GDdMAh
z`s&pL<lUqoLe3NR(tRiK95&POgs_ucv9vZ)`MWq@mXxO6&bQMFc60g`u}x@VA#~$n
z@o?X<C}&;!+z22Q&od~V-Fz_AWK&vC>&9NW8r$rDH|cf*hE<~o&;?DVyDByzFpT1C
zUs?#veXCfk83YrssQ&{3LDCkZAeEIP=)^-W!`|VqX$?b8W`shqyH!h7wS`!)nJP@o
z8ug0h*wxnY>TNS2O2lU;Y5CU=^)nR>vh-2|pCmTic2Vjfj%VJQ)Ac|LcE>8CWiqa+
zIF>>PESumMD3ReGiaPLr`H@!}+-OsSAb^WEtGJLDJpafc@mx>3JD~K+!Uzk<=-@(V
z8QDSXd7a)R&@I(kqmEJ8!4lKa+)&n^7RUOyq>6EH;2QFNsxIfgq2_gjNnw$3(RPD7
zS>aEbl%0t^Az^H~K}mGJH)y}qrS+^LfDADRm*&^y!svgaLWP#sgI76Dqw*DXl1?sY
z)(CDsE0yt+yVZ?ibX4RUt$zz@I<9!;w6(SRwdQD`<y7~c;pIGkMKwC<ndxy{deHyb
zh!};Zb*CyGO-R0Al%q?VwU!n4UN+$eElA0#qOSZ~cXu@)hE{Ydv(MAOckTkQIkyiX
z%jF)6_f|8b<j@)kiQ-LaT!izFL`k3PK1MB{8aYWCRGs?g=a0UPa3*{JtJ77Qlp2xI
zp2I{;DFO(mO)E&-W0(m<+C$Fc6Oa?&eDF7@-ZSb)DzPRbHXyQIAul*{#>kuE=e@sb
zrOkvGPB@nVf++AfRN!3BFYuwYP8$J+%YT!Xe>n>KUx7Jar?Bfo8D2}C1Y(AvEoiHQ
zT&88W-AFZosiIH`QL!Ls1D$oUW&i&`V@ET$TK;^y^=82WW%(OpN)%|35|eMy<t99i
zqrG=CPI?^IsJ;olEQ&ffpitJ(dZ7RIw_W3XtJHmzrZ|N|Z^Wp;5_&WkqU_#WAvA_C
zyp0-NT62^hqFe8s&^JxDTI*quVUCc%BV7wpsRFh?U-Jo$uQu5wvzMNs*avN0{U?=R
z7{oG#o5!DT+MhuW9SO7gZM9Upmrv0B0WS*pwh2Eg`liB1p^m`46-Z>nmQZN)^Eea_
zRDELcgu=bJVA&;aUQUZ5mjzr^DfqA#g^t8)J?m`AJYuH`r9&75Y!~T+ddXiR)!-bO
zrwHnpOO$RU(CHE^*1k-L$Wz=YpWf^~w*?O}2=lvb7{4;I!8ccPtO+J!!j2W6R9Gb^
zW&$;GW#g&fmJduy*%C}BaEw|}sf2hqc3KFw7y&_F8&FN9$AG9=P5uH^P5X?pL4qLm
zLqwEJgr>CyOux`OP|(sCfIO4=$8}8=>YMY`QK+L<67*K6eKr~0pH-woe{$Ratq>XR
z4vQP(H9&Ec{b`{lRR?5Adpm|W*IWBjNR?A(l<ja_w4GEP2>XZuqjx84n@ad6l!K94
zu!`=}s;cQP;(Vy>Ammc*{8NMbTR;K#m<Rv2K}h@CRpha-qF&inWc*=b4+v%JJfV?C
z&Df^fqiYg>E;h5j?)In`vQeQvrLGeZ-@TNyOnAX6y|)d!4zHN}vAK3ly!yKG1mbI5
z`2@4aBSeW&_G|!l1=kiYukNGE6n&96y45SU{4kb-Yx>po>2#-6Lx|CRj`C3g%QtQ_
z>8A~Dnh-`IYWnPuwtJB24DBdXMK2%?$tp*pz*iF{Xg4%|pCPGcDAr@tvF2trAzL)<
z?4#sKT!Yeb0_n0XZ3QZx2lDy~Kwlz~lLI=n#b{iQoh%mbKZ4;<!CzCvv!Ve?)eNIt
zTAaP$tKFe=Zo~t{OM%_)F9f$6Hl?OjKl1P_R%%2=+rc~310P@?4$Em~Zp#UoJXOsr
z!8yIg`6ntW7OE{Ez$;@X3^q|W0Sd-!x7QGAVDVwo2QVsO&6FX=$=RL$Ijsv{0WSTr
z(ca_Vc@AjVO3oe{Q^Dw1f#cjZtWXlEeHy8tjzJK}Q@rO_`1xvcwbVipE9`ZuJkJri
z^T*pmIWYV?KT23<PfO%=>1zwxA*2qO^$FQn-K^Hk!IK!qq>fUC$&vjw8fDEX9h$k9
zfcDVU#8H}oEM)w0d$gO4u>&()l*wtEWe!PN6>t=MwR;4nIZ6=XgBHL%r&;^RK{n{C
z-K>x4f0u0;Ng$G9%VX+s_|fKUq)CN}mtFFA$`?TcGz#m8_Axk%hzG@o%oma$#RH%T
zEmx=PaSEpT+&$g@aIDW8IZ{fLVW3$)jLwHrQ+GuPMOw1i4CT*@<B4QefXEZg_eT>g
zfUCGV5G7Uh`GkR*MO$9pQIho(8i$JC&9<#`ILR?AVA>ZZUMn%P?s~=Wmz|~XfiQvi
zgANHU#1nmtS+_T`zoXr|ftn+d@w(Mqb_eR?@>Y-kM8<by+6`YjDrs-nbNG97z|yR!
zHv0mP8odGHODcenj9c|;`_;7Ky!KY66!0A^i38Gp+sC76jI_yc^L!y&uxxz=+GKL8
z42okbE%q364snvW!57@y5QGpn<N_!xi^h4EsdZZ+B>KBE2XvxYPrBY%`*0oZO9sKI
zINVFa^|@$UTf{i2#-eUwC#+y%W@0!3r<fy((#SCIsoXgg91MwWO<dhg@I%=o_N2O8
zIR4|dbw@R1zZ_)EnxZ!_hZyrJ@+5X-LFenZ<qMZ4N|@pk0U-`&D3`EF`5eR8j<Vi;
zy|Gi4cRHmj94%)foHP7X^-2u|CQHbkG)6^Uo@zuYC%@#Ux*WjPvl>=tiC$Cbg+msy
zrii0J?Y5M`A<kAhfVY`#D^DfEiIyZri~a=S%AHo8Q2CYEZxK&lHu0zzXN!CU;``2u
zz^TYlGohsM6GTqZDkMA{use#lY%YKYp~V~5OifV8EP$U~+Z5rJ!kmLBqB+DrKnI>k
zE-<iu;E0)%MbK%_wynMa1uf2~LD0k8?tE1KaK4i9+mWs{fQPWae;cwy>Wk6AC`Z^w
zqHLx`!nY!0vY4Z;cc%%Qr<fL#%M1v|{-$m;7{NS`uk@a#nJ?=r*2Coc6*{`@-=}^M
ze9361s|dTA=1)0{x*z$F@j*FRo*?GQ{ouD^TULW?od?ggCLlLrQA#3huE=t<quH14
zKEDj1)esuYFA|)Iy6$Em-OYpr+oW3b>Vm=4BF^5^fOUX7H()m8nLk4&h?(3KJP2u3
z@ke`t>k1y{c83}ZG<Iv~^lMGFk<r}B9VFG9fzqJg<HwT;qZc147O>&DVlg7^NL$Cf
zM=pK1GZtjx^kuiv_vk>xL?}+zO=ZRy>;B33X6yoJK#^s81(YGpx)QI{AZS@2I%s3V
zf~Wk;pw<hnS=YSpT^upMw0KrMS8E!e1PBmB+^+)@aifLr-hj~XZk+DL1#?!xR@qQ<
z8y7>0E{=<yoOQ8{4X~&`IarcGZTDAOJL$@e0S)|9{E8X~%Df0RLf=_4IZb@RJ(TVr
z|7A}UnQ(7%;q3jOl1odl_}bx&Qxd;k_<33%Axr`SQZ}eTVYbOzp!-XVvl3~+_$?O5
zW0W$eINQb0-pm&U3J%t0oJML$rwxU%m{5}Y4<GK`ED<VP;isFjQx--M4(2eh+ZG{o
za1Rs{>R`P6ftSkyU3LWcRE-Z@)A6U0(RGLv?L#67bE;@5sm;Dq)4S>|g7q&Grq$AT
z+}$UEM{Y2Y5Od3;!B_vfPhK$LXwD+P9tJRv1Kh2z`L`irEdR6_3I-&Dwmx}z=Dc5r
zCWA;~=?JqrJ~7DPd2mAnLtVasiauarpmRWWdFSskZ^>z_BD{OI@olb`22_LHcxWF%
zflw~VxQOWcijoQOG+hRR(#T(5q4@gK(JzL*;aSnMQ-5?em>&9L6F<Q9|I(Z%S%!<T
z;{s_#P_}U!T2{j>MTLVtmQv?RRIIK(Ujl~c2FTmoF}WaT=tOY=&SHen+LEfkJ$gy+
zX{^r+W8Ws!yM4=#Q{nL<|MT{~Q`eMuLR`0|5+c`(2a|bv;<w9Su3xI_ZpMOL5bSn~
zD&Gon=(k><Zsbt7^$a_K%5|FPk6P;Ws@MCqme0eVO|`N$<3WnAD`-$?r-jX3)uD0<
z^{Oe;Np~Q0&3Q!(&&WM?u(%QR4G9tXVgy{eA*lUiqd|*T2hr*?f~Y<pBC-$x9*0t8
z7=B^IN0Lf3CWRg{=+c1Zq;R$1{#A(3?+&|A%C#;p=W@cDdQTJ0!B43McoSu6SbUHy
z&oZFGm3t9|-kmXy19!Jg1^7c5cFA_oOE=lbM9H6flypie*nwl_PheENA3ez6gEI7+
zIjFB7Ae8a`Gvz3!P8AB|glt@T24VwbJjH=3+=BuxJ?6kt%qiamztJFad_g*gby}!`
zRFysHY*>T?B3fA+g~T<UK@O1?nWPd60>veH4mckQiymGLZL42t6RF1n?j&=RA+`Zp
z=e7du@I)E>1m0QumJs-kswu+#*)n{fu<5w#h=U^?RP>|hanw91(wBrrT}M{n39AJz
zXAD*cHz}Ifcf>eTpVz6W9?K0rgvyOn-DJy6LEI*uszzZ_AAy;OAM8waZd9f*%kkd5
z>JbUO=lx|UY+<@&Xn}!{8|>A#obe-(hNPqu=wNcdP<|B;!)-cJDg&G}sxnVfqs-FU
zX=T0guJYF1(r%QlYxcoBJ0WCcCeWbsn18zspW>ZzqqOPv<)7jF7PT;^PJtNg(3+Sj
z5(xfGTmAKa0gal@Fc4KVzJkR65JFZbrHFU`D49xinyk1A!G3MADw1OA`J7`X%kllm
z<TALu`YmRL+=P?mT+!KN*L{Rt*pDP;qvnJ>NugOB2jv;=6j6#vE6w)%1|@w2L)36W
zjvE%9FaBVdvNq&0*OMXyynHCkIu)C$LmjDN$hHTP+3%(e7+kwyQC4=)_1EJ=H&6g7
z%VU0>i!ryeFw2gw36HN-T2>mHxuFAM&q=}q%;Y7Y&8BURfH0U@GTxo0rLcbqgje8#
zse7w)*ycio;9%6HxQwORsRQ9a296;RA#HcfKa(f955E$D`X8tXyUYsQoCU(Ervc$X
zj>wbp7nl2|JyO8sqZ&psN^Z5aBBhI7XB_X9Fe%PR@~CA!BO8I#Uuu|8fN>2JMVuG<
zn`#J!LWIKRXx|(Ng-pVG2I1>7l!kgRMOahfWV@`a2GubMRg25zvQQ^=ieN>v{1<&m
zCS0QO_^>m)+k&!2BR*$_2ha-}nauh_ctI}p<6W(=y~i9)CI{F$BZE7FC<dX1`3adw
zrMlfpAcleaYYO_-Twp)6KAMR@Zt$r36O0O|1>U<=Hn%3v$#1yaMpC2@k`eYtDk}*&
zIuu+v3+#Up9NG%GFyTVI=oy09kq+MsgsuO24XeR-6MkA98WEYif^eG!pP84PJooz-
z7I?8hX4pWkSK2&Dzr*s-F~cs;9a0QdnLn^&`flTdu~(5iKAA*zlL)n?*-~^`1@K#4
z?51QwN3^1Ko59E@Cya7(WDs8#wN3CU)8wdJf}uW19>t8$*{Q;`l}dvUu7<xL^*H-e
zZ~=wJY!ZQWgI=c{zp4Zi*f8X5aya*J6#Sk_BqOt*P+doY<0(bV0u$V{bBrfUiEk^p
z6I*9AnZ;i(KR)7E4RnEL7qg=_)=V`RRbyF9agcTAU7^s%F3HOkNJmj~U^z&FWQ}=C
zRI>g6o$T}(gC;Pc`#B(J8nN6hyKG4v67>^a<03ludr-m&r}&1Crd+l-<mLHM0^2Mp
zIJ98QcAge5rEXy4x$t({EU#ZVjVyNP*V*y?mdszj<=-_>5NH@=ijpRs)R$-nl@OA9
zZki^lSbC#Y)WZOX8zm>?$8!+de#lI3n+Tz>+vmO%mqI$l7_b`=(`Vt?Mv}J&L+BtB
zeAAZq;b$ZfWjUtzgL;S%Zm1d?lHMTlJqKwUYr1wtKcLx9J{6o>aS5Ntgg*+RALa`v
zZHd^gGiaQ#I6<wMn-t^bt_6f3=UD<xO!|I9=#(I2@trlt$vZQmmBrZ7;JD}q-qCD2
zLKM2xcjNJ7d@H&pYT*8a+L%C2QcS<}KbU*#u&BT9UvwyGL6B}lI;C?!q*J;<x=TWG
z07XO^2@wgEuAx&H8j<b}DQS>~sq^;p`QAV7z0bM7bMATWbI<(q=6QgbJ$tXc*V=2n
zUau7r@if_f12y*$*<`=<<xz1@B0JO07LL!A3`r)_Hg=uv%J?_#NAgOMz0C5=pT0`@
z?N(cXpS3oSKVo^yYJLBm-#HWa@~cyuIQw9%<O5nC?<XFu_Zu{;9#qKicIif))C@2d
zJ&(~BH=!+e9gjZJvbV*)kqS68>3OON!vryPp@f?vz7%LO>zT%LG(ii?$d|7~@LonB
zqmm1N1!6T=zhJbFL*i9M9{H-UTs00F)#5`A=9~r5%6)pC?*4acZj-*a1G+IDoSuRl
zyzjMUvokEU{TVx7#1`NFy#D+uZAN}}p|ZV^%&^c~0V_rAt@Zw)OlCOd?!cQUUP})-
z^6jxJv#r~+AX7pa3`Y5478EG#9g9h;UujUp*KZohMYK`}yQJG|i1@XXI#pbTQNN2<
z9>)INS2aYP0AZi<oTK=j7h|^Q978xOr#M59U#pMpnHP1$jkp8<L;6SZykq`OU!N_+
zmy<<*%X+I%e#3ie^F2m0J|f3G?Nrf9Sd}niui-R(9JAvoJ5L>Z7A>2(&~5H}oN*V%
zy*Fc{SbZ@8@m3)$8V<5I1c)qUcA{qwg9+vdYm<&lLEHWM08sezZ6V7&&`|8;_ETNk
z<>{{1xY^L&k<UiW$DG0JHBM5gB+~Ie?kx<gc|QTf(HGZuA5{IGP+=MgcK^s3I=`2<
z*TBq-v`0`y6ao;zMa}h*hk>L!t;Oa7j9)bm33U@+>2bf2q42`|DSSt`JGkyv)%-Zh
zS^d_}jeI)=51aJ`Xx1nVYE1h}c~fq#VQFl~A(w9Xii|uaXZc0p&vlp^x7y<^MgZ|p
z`8-^*a$y>fX8Gy@AF+?BMuw32c$3+VGYb=I*HrGCdA4j*enaTV*@t5>g!P`2`PvgQ
zmhCV0bm&CR?H{C`YS5;psqhcIVdv%*6?GES*OupIl$kwuJndxFy&tcFfoEa%oU!$J
zA67}ZJJh`h08rCzX!=MW#Wr-2eag9oRvy-W2gLu?4&yey^qwLNMa48{AI^QWXLgVy
zi^@mdi2Db(`I%&$P5t6LRQ~BI6}9}(p*#WGVtzf@rmr)XyZ3cS3@N|7NNQX!bIeD2
z`>PHc#v)&b&H@|@{pXD8h<A_2W>jl61lp&2KfPkD{pD*IoO(&BE_iE|mag`!w^MJm
zoVU0u|G5zLQw=#lgLrdv>Nc}VXaGf1!10PHb^L=U%L12`KL0G2Gd<g_as$SrWIlnX
zTGSr96TQ-P4u(N4&X?s@8z6P^Z7wB{DKD8KH|_|X`@`=MfX0w08J47ju)^0DV+_?n
zXy6)OpgxeG@trU1`b<WMXp3r6i;*mi6^c<VRCkB%>2A%Gi~Z%<L&qDRb_=XIf|}T{
zc7Lv?o0LScU!TzPn#t#PzSk8Sie6$@$8e6GH!^$qQr7R`YPi|!BkZrBpxyFf+%+Rl
zILjQ9yH+~1S^(b%=cgUI@+=q|k9<oOQ2HMG@{s!o#*IlVu`Em$1&B|DYs-+C8ViBC
z_&fJ_sA#k!o`s#@<ny>DP?6wxkrA`C*u<vjMzo#-qKh>6f~nudg>k&Owx2Z$-~0tS
z{e@(e7RIpboj!~bYEApKz8kWYqJqG7C+$*g0rDJI!jjn1DeP}|)AY5UE50R)?WOj*
zb*HlQOSuCS^7N{4TWBdCtqoD;wFhUZy!lW9GX%|IyB6db^F1fjFfJ8)oS<dF_BNvB
z)Ll7M(!Vm|YB2bBPT;3rQ+{)^XfwCY2#g}yPCv^f!B|#Bm43AmdXP6}8o_hV{C-`(
zlukl+_}1EY{x4%+Vi;Y|h`+P@UjkO%EyB5wPS)m6vs+E1ln`|=<{z2f{HLGprM578
zjLA3_r?{emt>iVx7JxQOr7Zo*#}}qNoLXb}*1BI>I;=f4E?+k>9}3_Q3f0!PHn1~B
zMq@Sb@ozk}Y`4loYO5&$>8-b|$E^p5DXm#{%lkHp#RsZUd?g5Kmme}1s@uUH?qx#p
zWZu79x#7{3LT(CPID&4Qg?)C)dl1XIryg(=gR9Tp{^uABNfM0rm&CR#MZDQ3AvnJ8
zb#a(W)5<?6rMk=Oxy-XDk48(=uz#K%U<kBOH@5V?DQVt#F>T)U8#~)LipHM{y6_z*
z!e{*~De7!#>g}IIU6jRdPYDTkUbX1rJ^f5{QTT*1$Z`Eym=5L>#w_|yDvh0rIcn^P
zD01cPkCU+p&+2{xVm)(d1r}A@n6;&)dgEPA#scn&Hcq*H8r5z)sr2<hnBBCW5BX(F
z@gfr0cq7hu=6-s$-L8`}>kkHiG&~GEcCS_W;@A;nHbK}sn9A=TI6DBmP5y^j4zK?B
zGMKB+(Z>p6bZ({Y^#!<<8YCCzudN<mH`QGz+?RVnD0TMZ(LYeXV65)spa3+eA>11j
zfAjvSs{I<1dK!^<0ztdPZKk(~i>6XE85=8`$ALI@ES#?(jT^J@@xv%$Wm{alDmH?b
zFFyAY{;K(#UVX!a2my+W1*MlUT85aW_zQS^KiKevstlxVq~r0d?%!tQ`+O6}bXaqP
zwi5;G3f??_9xfxo{G%}@Lj#K$j)4_HbwFp24`ZDswV=o0k08szyA57|%30Q`V-@_^
zdRB2n{Q)}N-BPoK<LiRGm9|4G;qn82+)U#U5y`z&J%ul+PbJ@vDs1?d+E+U^mxeOl
zpgU!BALqy8@q_v@@2jTedJecRy*+q(Te#0sIMJ$$(p<&GN}iP6Ln30*3AN6oi}r(P
zbjcnWV028V$+tb^jKXD(%i(`*X*_GS;n1-m_mVHP0bj}>QI(@t>-U<7@Sh(A#0w~C
zTIW~O&)~k6Twn@38D^Q>GH&LE1%5_$1tR%nX*k$SVrt!*o^1Ra!-Sgbk=xVO!{E#P
zZDY(yk8rVw&)WVD*vUYaPaQxQXm#KwyEO^Vp9p9PU0tW*5gFmtBHJt+;g5whoY)v3
zq;IC9&xUdfT5u<4{XUW|;HJU|P)K@WO!W-3g+N5~JzistpGHt)f&v0&a?frvuf2v2
zK&K5s4&@gs@xbm}C&XhXc+K}3n<-$<!r_a$KboH^E4(VU#_mFSdV5g7O@1ToNrea<
z&UNAEn_3P#u5TxJFE>-ZV8=b9h`=%ZS_8`Cvw}TT-QbSD={lD0nJF8^&i?cSv-9O3
zZbb8t!)<ADAE3&8o`)83Rx&+lebU-!tR8A@djHMxdq1y5^6Gq;xlyBIyM=MP7>E}K
zud@{M9>H|miygJ%Wi}XyD5d15p+iIheiG{4gGWTc&q18gGG}G{mf>9I6IJnF3W>4^
zlQgsg?z7!z{5icBk8McaRXuOjMbe7DtAcc<7D&W)aS>_7XvZG>I%&u?EMF^c3+kpM
z+8&^wlk#&OrHx^Wy|<d{Gncl}8_i!}dfg&%i4!F8q8nS${{5E`|52R#s%LtWRg9FM
zo?!Pema|(O*;^jH$HbPfy~q55cjHYV?1m+>{o{1ibHBDRiw^ZV9C>On^+p2&(UC`>
z`#EYdTld)H{8!~K$-*{lr?%g{z$s>4HRRXPS$KKM^tM6SF5kwi7Q$f+kG}tsM-zX1
z8>Hvl#f1~z6dGBRd?p-Qqt*(NG_<mA%}H=1YA|u=BHH49;V@8Yfa|Op1gPuie~-MA
zIIWi|mb<3r#d?pejd2a0JGCMrnPIPf@B<Kl%t@6`FUZ$wJ0|Z8wgy{@EM0&Mp0N>A
z({Kv%Pp}>V9pMhj3ovv`>y7<jTuI-Ov@mOMaxn@egh0KPEpLgJvRoz3<X0kfzmZs<
zUX7Bz;ps@DZ5bH<{9>7gms?ph^H<s-7ujf3dW5rI(@KD_vYG^5sk%a&<L`IeGUFsB
zRAeh!aq6tXG$kd2A6NK|o)2R0i<*A<15)u45qV5xW0(BuO9iF@g_6pjzdDWC5=3^L
z{kH(bM~gmp$T3H7YyIZWCh>BH+Zz$AH#zRRPjezaYN2{}ES~*}q$58Ne$2_elIbfT
z=_hYaa6o!5S2qdw#V!EDxBHlV$eqF%<0=!;<``OP1-}fLsc%+b69h^A#!}5Bx$z#T
zDYLF?EV5H4QfQrDJvMs%h^hRu+f{D$UF&BnHj%&8a4gGoUf(p*0RJ2biofcX!=*YQ
zHLrAA5-)r5VWja^{xgOLS}(sk(7$T7Xo4bwNj6s!r*>iB&@=BdITdFSlY#(xBj3U+
z&}hzCQ65o6_1&BYmEip4TkqD5sBP{CCZ<m{<VF9=WO6yAOmEaj*lPIFitShFrmrUc
zxy_{?RKR5@2v`jQz^`X*CDCfWd|UoejJe7E*y*roeN@(`Kl{l7xw_lA9Le5OkyE{q
z!9wN_#SH{d0DN(;Jo5t;9&R{m)`RMnwWHg!vveU#OxpLK%=xmd$)euiyb*HeEW%Vg
z@c4F2&_#`%ZWYhW@UH$C=&iZjJg)gLwE{-1Nl}A8Zb6mKxs0&cSvwvULC>Ov4x-q_
zp8~<?D|J4THg(a-^s4voR>porgM7~fOxpAq7VJt|gHEm3-MD`#w9J^i*!T^4LLx@p
zZ!`D!X+<l4VU7q}EdCxl3z4Wjojq$_PydOE>7K_U4?6jHMnu#p<z$dEk8WB*&f()7
zN}y!k*6}43uA6I;IR%i~LRZ1f`~%M09Dn9o3!d*1X>Q+u40gAs8;L)agg9yIZeDI6
zzl~|T&Rw+|@U*|{a}k5NGfH3!fu4gUU)wdjICQWVBOwBP>v8d8ak3f9s3U&rp~d=h
zmhXNzWea2NX4={McCj^+mC~2TO~f`G5rv@aSRvBQjB0NlRt@#J5~m%l=f8D=fEE&8
z%MdA-pu@@av$KZ&oc+a;XxjO&uNMo{+}c0lUuIvCRx#Db>;sVXKSX4VDZ5*P9}>*9
zz#(a=ky(k4n$jbAg@cHt#=J9GGKp%37pnrsMJu_yy4dmln43?D;C1zZT_)Lt%~Phx
ze95xJN!=|+sZzqQT?lfD6<E`x!%^Lz@6Km=jhO^KmEQ6}gzML|L>h5s56<vTx>V^v
zgBU+<s%N$x`cpR?E4j?eb5{cD7ok%{{dUR85t9$dT{-c0KG8TTI<4dAD0gFPv;5Rd
zdkW+a^}2V)Vz|866^&|XynimDfNkh_%s8m^HVt5_4Y&f$L%1Rs=vuSAB#EDge}deE
z*57=RrAW$>jrmFv{M&JAxIZ#l98^-tVw3mgy=IF5^l1dMIw@ByMd{neO8t|yKkA7S
z%q51XL2W_rXr(Vqs3Srvpi2aFBKO|l%luV}hWL^ZbHoe&R>-8i7n%1*T8hh!_1QYM
z)0P;KqKpj@=GL3ZZ-?hQBj#f!I5GY^<*NQ~tA}@rD^#y4=$r*&^r^I5#?;0joZ&&a
z)U6xepA&rmX`<rN2xqT8DAgf!yX00*L0_K&4ORM?Rk4)UFi2<}(nO7NrrtYJnCab$
z_k%hg7IM4JY(0$P|IWNn;)La@@GB^}#)*`(fG)IQ<u<>2Y0kn~OSiW8&=u$GS2qp&
zGo@eOXn#|^!NB9b_c-0lm{(WWuc$#~GlQ!gYkFQ1pR01)j#}5Wn=ig)&hbPkwokug
z#6s!x6G6?vb0WSLYGzm?ONQd_mq!>P??@ZLP#@4jJxSfJrb*CAQyyX!BX*8@P*aq7
z|8d=?iaJ71W{Gjd9nd^4I7{YmA>yh!c^yu{?LD|EJ&nQ#G)y3TYx_G5YOSvu)ZSDh
zFS|kWL_AD<>i65?aelh#nv@5>hf?*B7FA@)MSX}SG@U4a0@_Mj$G8}%4JnolN!qfm
zCfWsZ)!W##dsFyS6~hNKswU#Z<M({3mqj?V%QjJuHc+^@p&fq?NP=KZXS1MnUj^3g
z&;E{;IKO*@l#IJ0?k4Rw3wCrM&Fqqzh(!7HXK=nb2j^upXLra!G)D3>q6_u0vFwa)
z7Um3d6H^NSnYm9Ju@bl@8=deo3GI(bTLS%cr9TSV;w56H`H_#~P*e3|WC_r4)8bn$
zB73>PMpq&>qRvG@qO-8}8U5{EQ95n@CqfKgwa1h?+7wZ`Y4h=PY%8Yro(dxr>eR_3
z$`zz|+V<#iQ}ygKscUd7IIAtVrPbxGxGd|~qEgjB0!Y^Qu3Hcfgd*pd^paoOc9BS|
zN6->t`ObtssvJw*eD_tTZXgDCso{L&=7(G7EiSU@RdUpE$e&CX$1#lSk?Zv*CrL_K
zynaJEdC^Y@HoJvPofW%5A)!Vl<!+Yt-;{0(*{IT8WW_>s4BzlIteAORZ+ZP_&-nQu
z0sNv4nSc5Ir!zFz3ncM}#T23LWDm3Y^pkVA$X<%DvSEf8Yio<*3CVgx6baPBOMW%`
z00L~p5nh?14S5wl|67Y-EHPf1CzSUB#;*vWYafN|xC?#D6S&&h4t8-Ck{fD*kVl7L
ze8>BrvdS|)eb%4aNc}#FqKd~z)`w*A56MXVc5^Vs>h)EueV(%ldM=-z|2He~tMYX=
z3VFR|9K#Us5wbDee~&5&D~jB??xQG(c^ja`i1il#y;3F(5A}WY%*MfW{00R1&n$xg
zdc@X`lz#tkppF_Eu)O#L^c7LB-XjV6g}3q!066XizhtWw1;sONcEHz}yaZ&yUS*z2
z)`hYWtcG|Eg+H3GoRm3-y(gYbk)KXztCa6ZN!l$7d;Q=P-r6Oxd2K&M$9!)cz$QEg
zL=X|dFxaoRbkhJKr*`$G_>*n^uL6}C8{Dt$Mpc$Yligp$ftE%+VY-<_s#mm6odV7V
zIe1t=4|<qAB_uxrpvK02s7@8e+f^WC;4iBb69*FE9MQ;Hr3+$Yc~l8|l_WO6^tWd0
z`ZjMt3S+)A2=tsAH)<faHnbzyc2ryrmfu{=N%}r@#D$HHMqD?|+NS0#J>hR}t+^X6
zj-P$8+&_~hUur!;zRMG0$eguNkeQ9<u;Bq06jelSJ>H`A;6CB0YVJ<6bt6mRf+kVj
zdrb!ASw_g*h%dkaiU4c*xK1vvXm%yy>n>4cR<oZ~!7F4AMs+;MbD>;H!H&aytLSNu
z)(71py63r%$Ybs03*EgX`9`G<c*##@(|Gu+!2Fz5^<f)JbNyq@ML7YBgP*H=J5q~F
zj`k#e06=ul`2>&du{TPAd>nG`G@ZxY${zAS+W<}GY&X&#-u{7bk}TTvy*&Ztiyx1Y
zEWEr}u17!Gn3QkUbSq>~4s%4l`BLL(n^kR}YgQtJ;gb|@>dAZ5#c9|`m%^T0O}-ky
zgO5`}<5hzBvq3MIHIdjz+FtnQ+lP-jU2uo0WD|<S>W;RUJ_gt)J=J(cY$|kMwtY(t
z>TjEwzOy;h4R0py6^x<r3ePLWNIt;47MOqFf?ydC(cnM>?3D&NB;8in5zIz5LfOiz
zGw!_Oa!~1rSSa9dzGCJ18E(1<tr&B9<N*{G{+dEpjKvCPv)Da4R?TD+$-!U3Rwn4l
z>F5!eyR{_?P}^pJn*HZ)V&<PDVbavMA4NBEU|sL50QEV#_<`gx|1KOV>hNdy36uEC
zcv)|qHWuc8<>nwl40zDXqS4Rc0;*Wd3y_<6&pEaPH&Knac$<u73VWMID^R?2x(92P
z5#j3-Aj4?sfs;1LDg#M1sV^R4&J2U+-kskRr+Ys(GQrNld#{Fm-P|i$iTJz$1kg*t
zZ?%bzz#T#Q4B$Pbu}9^q_%75ag{Rf6)C4W}mV=23MxGk1gisN+QST0h)zd-d70c_F
z(Bp?{Wv0(xN5vuur$D>eJz7ykdqyPG28`Zwy8ISBW#2)Pr8YF|L%D<B6_x{TZpt?I
zf4ZI1>IMv6`s{P;jX9Rng(FO4CEQY_F5);oBs<Z_xHMFyK5&XCwBK?|rNG_hkg2gY
zSGKNM6?c2Zn_$ZJ(>vSY=xD6pZ^fNFN=zcAfQPLlz5T5P-Qy+q)T-j#U{`{E#dzwV
z!a<GS2c&guURjWRzyTbBWe(A|f5<Q69EogS8ofedESr1>J&i2Wc3r3^wf+mbN^gap
zEpmd7UHr)*(|3c1OT`5u*%l&~9zE}u+kFTxn*&smsL>V0^aoa0eQa|qzf`r-9y=re
zAY7Rb-u3NqzB;!q@sjXtaf6}Mw>wTiytdSK_3Gzcr+w(6OR>Yz<3TfepdMJX!?d-n
zgMCc*ld$O*KoTjlz%>nNMPc9m9^*aDJ+E@i>0YO_l{W370eL>;`viuXRaD)gdeTrJ
zNg;2RFY{H01sMs?=<$2J4e8XZVp>J2?28OBBv&}M)i!iiaUI^%oDPayjz;CxM~O}z
zJ7?_|&^R-F6rB($NG6F$o4l{0oIvaF<O0)nQFo512@X^$v~E8?Y4%5?wzvW`#jIQ+
zkq1QmL$J_dPx#);mq%nBKh|=BcCy~u^5_ewo$*pP{>8&@%e~iIj8!phE7dQiy|?-E
zI6LPybFR@8LVcv9y?L}&ls#!3gwdEz#ioBR`vccg5|_Z+;`F!=R`RSVzIqz0<xh-7
zcGVZhwq*fbgqYS+K)pHWJx|sk{?Oo$o0sN7TNf~Z>B%*JUi*G)t6&I*%-hot5bEL3
zQ)Jjb;neb1#JH6~@V%3pf4hCo_I+5wE%?kB`btsoL+?jUd^bq!s4=42_6zLlELe~n
zmxn$p&F)VH=jNTwqjb%8-GqaR@e!QTJ!-1!MwG3wk#YcMlSr4h&p8C*yT4^TBQk=V
zQ>BY|<9m<Ui}uaON1Ye>bj6TVOcF{5kZoLBj#{@npvwB?bd!V<^;z#Lc?3aK?Hii(
z_QucE@o0k9{=q5mpfs2TB&TM}m|7t+oNIpgXHtzvL_I2{Z7WqnZ1+M1b=3r;N$L-1
z+m;%g^h8UH#0|(BgIBL5Xv3$}gSw^z_4ao+JoVK;?&NQM@Yqd)tY<&KDKc*0O=5{|
zVT#|i6}|{UhdVWXJhRm&N(E2GD+u4M3GA^nOrw@WoWJKr0<{s6?-3cfR3HbF8a4ID
zdY!y%!}6hqhT65o%k!~)F^*w{xzj%smu4Esqli*%C>hnt*OluZ_{+@k>-+?k;qDrG
ziBG3zrm3yxX(RDq>@y<xz3ux{-SqqvPV?kcPTZ<i%d<L5$c8N$<P~ez*ywd-!{qn;
zB}bpU9<r!cYHZRyv#a-{eg#<suK05?W>62z{YK~K=E4*9=SxaS2&htM-n_UcPb2^x
zp;sfzZfs&_t_3K|7ke;5LxQU*HT)QbgyK+S*>62)Uj4Tr?I%m_QrSl4VsD6|lSg>?
z2B<Rt7MK8^d7ahAE+QDf@tX7-tf9eS5^#RHXM)A1cs6J>NpQj}t&%w-H`}B0skkf0
zj(Jhsxh!{DU7pv<o<_2iAX#*b{oIph5uoeae{$7J>wyQ)1fIY>U%i4TPcmD|99cpl
z99fe|3z9mIS&|zbln4v%{*`W=Mj#{UXWc9dWaC?c8Q0>|HD*_s%CpCmy?=&LKrt0=
zeUKg}%G&4<AO(;+)k3A_PFeO)l%{T}FWL<|RTk6Qjp({_=V&Hkhe5MDo|uj90S?ZI
z&HzK!Ci8E(s=q=PZHeow8SZkn%=iv13$*l@bQ4TcSC{6>+YPZ@rI5|U(T_qsB1J$w
z>_hO63PSmQV>xcn9nV}SqL(UxO5gM>_WF^KD$m8o<Lp8krsjL8IS6=V)2e!w^9Su)
zB@6(K%<T^_Z;Cl_bMf2)&o9QNA(wk=t!(XX!)>}^*=4v68;{*rTFb$6&sVV~jY!kj
zOli2e-CjF^a_AVMLmLw8OqI?=k7o#if5-pXsb8%bQFfi$IZ7>&jT?!f2a(4fjSai~
zO8J6dQ$L!^DC@tL3iO%3DISK|l~tE+5qzbq_NC%j;F`A#BnU5&#(8X&u5TeGaeh1s
zK;-T4bPa0H+(|8LER;$KR9E2X)-ZiI<gUcESo|7uh4HR817ZO8+jk-&Qj>%?q;TmM
zmHe$WJ;HjDh~NLjP1{QPg-5q=y*cE=sC7MesT6AKc1zl7en(KWwuc=h%FgR$19V5(
zZ@*qWw><C@5{F?CLIAu~u|;+TO>8>@xOu5}MaeMP_9KDPR}24ftDns|{hv%0U#ZR7
zvsh(6u&*~cPuI_DwxG^A@|3x?pk8BGg$=R25bAL9sD0h~6^*y~Al)R-JV=(1+Pmo4
zmz!7QZ!BK<-y43I1Ydn`F~`re;?zGXD!_j@1+$_Cu@FWJ*pwSt&cvCi6{PEXHdmHR
z73+v^Khh?p>Ep}R)$?{g3<C$5v=?rg6aD1&uvv6d>|`Hc{sC;(moJW+kkYSDb=Y(+
z{xaATvreNV6@&+6&TsMj@9#;eu1Nc#x6;^74pYYffQrcG@kRo6LWJz6&)#pwmB>C#
z5)b%)@qSeA8(P{!+?}fV)0XoPv$7z@kEY}=NqB9L*(acSMXXt_YZ>cgP;@+XXc0Q=
zHueby>v`TPvp5VhTEesO6Y4Zz7Ub`m)4s8DOv9(b5NRMB^k^n!PCCcnoDyNEgVfk3
z(yn+^&+(=iq%oIk0==X{#a_@+n(gES+vj>s_<Q<94h92cUBM5wS5sCX;z#U0RMZK4
zK9KIH)BaEqYrR|Ocwzh0#U9C~CU4vrE_N_tqh%*;#f(nsIW|7du21peDZtq*^b=dW
z#cQQy3f^U5(y7)~Nd47?|G`Rc?4Z}m9N9ATq$1WX-h^b90s0~{pndq0Mg^aiVOK*<
zLeXc$j~a4o!MSl|YsJ#0t8)?XTE&dUjDqF85?=wLI_u12Og}ca>W(v>wbNYTZ7uTI
z2eoN)SldwXd-7toggN84dVz$#?++(E?I53DF0;bxRcsarT(yOfNb;BL<x?CmNA1ox
z>L&W{)iobYZr)ulH6A-Yz3-D#DB}c9Yf>25N#wB8N;S*ezmuf#E3`E=%<02RrZ_2D
z3yA}ia6X3RJlF*=!akAsI6%YJ4IcUt%`MidmL1wP=66wP!9N3Ij-kzZbR4OP3%T*3
z;Y6o&v_h&9Vqe1q>UL^TSG<Z4kMBZD-;HC%aecuIzRy2>sI?dl@rZwHnXhOTV1UeS
z5P#cG9l*0jD)N29P_++7vu`{2Xi41N%qtDH`UcchK$E<PrW5%m%O)Bu^I}==&oquf
zYBhUT)OTsGpXNL8>YL2G2?6Z^Y9B6sNnWJ@Xq4K+;OwakVOWqQR=Av4m6Ouf;CJWE
zfwdXC#He41c7`WKmHOH`&kwVNooL1L@VhzDk+_V`@rtK@&r&2_>{ZU6Z{iYnKcS3R
zb)<XGH~BHTUFzaF)fHXU+s4Ds6N|$W90(>bWcfnT@APTK`fsc9m&BlC#s%9nq0X+k
zixan}L3|xT;4YQ^JYToXSFzD%-uLna!PQNX#%WMPT)%y&_ECGzVy$7OI@+;uH3B`J
zPj%O|w5wJ0{T6GFh^QaiTbqjE!|`}g%wn+}*kArP>Md0M&HcK;;9QJ97j}34=%#h2
zuv(AigRCtTi*IjTwu8gkz}X*+xe2>MWB3BRU`GKjYmec3k9I^ys1J$X^iXC^9=g_1
zR*W!8xkwf*IVuM!{g~sO4Z<oq3IA$`Y=ba*{S50x?0qNs@NK<r4+>cUM64{v_zcI`
zE4kY#<&ECF92?HI?M@>W@&r;9Sr6WY6hENHTx!~@x<5@IrAZf+a1lEaTVp|or|x#b
z|Irf!w!d2_$of1g<n`@k6#o?T8{s=wyD(Z3mx<?pDb)6pdmq1rkHEc&M_dIr<2O<n
zRJv4y*%LZjf5!Cocb(<El5?<nyeX*_5mdpiWLu#-|Eas1D(D5p`%$+m=gFtUuB*?9
zT_;#t#1?mL<TIVWDgKDq5Eu0R{VaXeGTJE=pS<MVx(AWV!k__b^UL^iS||lNt3|)w
zNqCCwfq6FklkTNYghjgu=*s2IQ|LI|=LF@2Vh5i`Sk$bWTDAhQ&2H51o5qI6dlSqa
z1XM5BrwXa-l<02x^UV6?NgBSpy}JuM88>gQA902zTIUW2mV8`O`*>O}_e&DCL5n3%
zD|xJN&O+;uA37t7{gY>^RxXVwuDyXp!1k2Z)v}6tN>t}7=Mb@hS*zlQbf&_DpRQxs
z6UxSP{ms#O1^{zVM`(~WU2{G*E|x6qqrQ;Q(MK5*@}v;vuWzXGX0Nh7Twoi~QNfEw
zuAaP4j~ySxO@u8EaZCqLS?n*e$*^s`t<)Y?;bT8%om>*Q&0Z929R)p9eU(4!#y99i
zgslu>+fqHc?^W1CZ)>OiQG`13g-J<Iw)I?<U>MGcvpd!;-aErL8eJ>;fYR4iP>CYF
z&IWm*r?c;U;$h=%%wQVttU=RL;XPSzx^Jh2fHvGO!YY|#2sw%AyErYN8)D}XJJkfI
zk$X4u-d^Be%{oI9c~0q_bnMRwK_%W!9e|?LevE`+;8%hylK(>NZWPq~7*M<W-$MxR
zzaGY*0g6egkZ}W*NYM6w)*TcQ1YuOa`M>K9VgrAM{jcsIgn{QZ@JEXO>JA>R;D@d_
zh5d%FNlK}}`AugX%~u1ZS2qrQ$YS19>T))5=`Q}7(awC4pJOF;G`{qdOP8Q<$*%~y
zI=8&Kg!!0;SpvUC@pErYY-aDDj*~`v^NfnV#lmmfa%(DgJRFJSWE(FF9)+!5tbH4)
z;Tx1fuiZUU7Y`5n@ps&y&Nq}h`v3d{!vFUnkNNL2`QHbK?LW`|=U+4b)l~#us{cMf
zWdA-R{(f!X^<Vt@f9x~<J0tub{EYv7hwwi))Bm5$!vES3{^u6^zq1hkKMvvloyq@S
z7vg_u7XAl^z^3{}2iR$^#NTe9(c=kgsJkX&!uDg?FLUjD4=0CM<XS+We<pH};5f#@
z;DiZExT7dWUx3@y3X!D`KH6?(mb+ko0Xem9M_tp<({{X7UWFh}aQ@x65rNP9jB&tn
z!#Xb?1hUaLdULu@Br__aHYb;dXnwuhR(i6W?s=P{$@9+9<i)gXYtwd7p?P9RNPyY&
zla&Fb9v9S{@-lP^(v1s07JpAnY_hhV?|=WoOa(G*W-B{8Kj)Iro?|pO<v0>1J1a|F
zhw8$h^M={jIFEx9*>P0yK%jr7;d^^y7+q;qOqjPE{V7j@cAx_M?0ac|`5zH;eFK+3
zO08=zIzrsVs`;u}sgRc2uSb4K#vyI)-oX7xVv>WDl>#`{%n8yx5qG3{_dp<aoprF3
z6Zi~${&lahHGguJ(r~F#$uOWT#2^|yLXLY?M2w78Gx+}JT5@T{6<TjmpSjXd*D|Ud
zDB9*3r9fzT9SEFae~+k-1^5vx<)gz*=7i-Q9zN;ihsH}A%Um5ZU<xcJL-Zwt0W0k4
z+9_b+!{xxyca8hK%O(n^fxvnC&p*%WoV8kz$_UneLVu}UxN3+#sALr$>+aZw*p7_h
zN@LzFTSa>1ta_qDJ(K)fdLyn#i0=EeVS)a!Gj`2R@b|~y@sqN3vs!WC@*i@tcF|#u
zJ7(4tO=?eePBnYD;D}#&^QBdku#eSik8wdD!3+>4D8is%<nC@Qe=QNb`#QX#k7qp$
zaw`L`!#Sv6`{a=2lA`tiVZKb=qv-I`-P1ERZ0{y6TtprM@s}fdLJho!6!5j>-`8K_
zR>D`XgqUu|2Dt(V5oIfyu<QeYih9!pCDadDh90L6uNGc-OLd}++txaKH<a2o5~+rL
z2a_?Bfp4!23<g9UGpQUEjsvgh!Vq6tJ;7R7`&cs>?sesn!Jr*zh!PpXQwd%7VpCTh
zVBMN4;%h2_e-GX3rF8RFi}1Sw-sG=iiTJ~I`1?+Ac4F-S)2I%or4UupVW^beF$o)Q
zuL1yH8_lpmZ9ep!O!)WU(wrmL258>0<BMxayNCa=toYaYYqDxFlW&{m+Xm1tM{}+Y
zraWsjshS;CzIHT#OD;$(jW5Nb(Bh$TGm<Gga@Me487#-1?an&QIUtIxmf6i#t{V%C
zf4!WB<4`NfuH_b#62f=hHaX|1aW;gU2xIalL>}@vuDk%-y<t7n!T9(o#Vs$}KRf4T
z4v0fKn-BcyMHm~fhyU7yeC-Z=+qGh8(t@*90y=C&X*6%vG<|ZMx+f-3+zYc7KImbh
zeV^?!AgzYU`J%R}z8?j>y<TKu2$~hZ8TMIu7{)Kz`J3$DZ8JXty@}x^yAQv*Xhbm%
zCD|_;Q?7cQ@MRxhPWoL3w^9^4CEY6wPn3Q3?#{D~_w5Hx7PxBPQ!wQz$3Sk=UGDu<
zU_^fpvGpCMUsRu!*2b59{f3|L=6dGky^H}jt*(x5fJ5UgKaO7%>5?g_fn>ugd~28p
zZ@MRbS4hroYU{mA1~$q+$Ipo;eoh&ZK*mlHE1(j*Lp2`u;!jUN$RCmH4Q{09LNv+_
zHxxws-tCfuUh5f|O=lRNNbnyXLk+EBj56~47U&Q|A@!v-VEg~X6CuSj;f7fPm-V-^
zsK>VsVrvX5*IZ2#=*Y<|C|vZl94GPlVBOZCC0;5>icvR*(psjY&kVu%R!{{S^WS%X
zXU#ZxoEr<SNzb1o{)t<tufT*)-5$7S5u!9K8mx_@my`w6?{li<TMxveP1rErCgCEU
zT|RO1?Lsp#Ex7kh@CM4*fb*FD-HbNk6*oM|(<#2h$2T3{os0N>W^a@u(JvMWU_q-@
z*+)(5yWsfSKM{9Qp_7j>8pUT8?K1D1J2>X*U?!(x{5wOG)_;sigGM~g^Rg`MA&oXU
zBO_dmzQ)^gE$0O69tDd`nwL6huITacadSB<^jS83La5vZ(`%kz;)=%y#XcZI6a;eR
z5m5haVl=R!H4O9s9q{jL*?fRggBy|q+qi4EQMdfBLgyKNUBl}<o7yyrmM-g%@9j8j
z8_;*1D|!+skKE$QjtI|>YsTnr{`@cwdgt*U6PRbqp(8Ovn%lKk^}%}OL{F@;e>L8$
z>FIjbS>i?{7{Ox{f5hq@!GH!Vf`T?RCd0wO!HAial_O|X6+e1@Y3Tsk(IJQSxx4A=
zQh@Yvkg%q_i(bLUr=g+A$jWkASXju&%=`@n)O?2q2mhRC_=(H^$m<mR{_yk>1);}U
zLoj@&G_zv*$nBjEH?Z#kvjT~ga-#^g+V0&|v;)8jVc$dK_W9wp<Emp|pCO{bfd7yT
zyVT-{Bnoj80axa`Ln4uKonCHkbvrwbZNZmiu!}Q5j{q^Vx_bNpQ~loVrau?{x{m5Q
zLd5y`d1ZcnXOVV}D+L9Gk)`F)T5;3C<_6k5Rvnx3hU5w()Mj4>{#FBYvld4QjO1@5
z>^^CLXqbSl7}hq-pL}~swcyvZfKJ?S5gRf@P+XKN(>Ln=WZb-V6owPbzV5V4N8Dsm
z%=6sL723cG3;V49D8RVUU4`h`OIXq>BOBSY+W5r8FLJyI=jlM)-RtLM@`Di-6}%e2
zYxjr3EBhnsaJHVE?X>1UiP~tfL>%mYjHr`6pbz#9HYQ~!YqZ6UV&cv7`g_u>>|vti
zb%nIEnd=aL1b8I8JUkdG55+}AZ#6Jrt5r7ENd!02($Zldrg|i<G3ok&UMG_Q`t_R1
zOc>7P-rk<vx^X^K2-JYe%*Zfva3~zfm0famb#<R$&3u(5C(8xDr#@NduJj^W_Bmv@
zc4{)ca-9k1-$f!n9u%72KXF84Ze_M?hrg^~FCU-H9I2%_7TVcCj|e1cv>Txh#Lo^c
zndRpwhOgWbt-DZtaa5I6Ripftts=pqVq$mKT&E<M;mFn-S&S3=b4aK;<g}IyYadhh
z5^}uU(g^5ip+l;y+7~%+(wOmDG00NzxP0o+Lo3=b1N@A&%c`vhXX|VK$AI3NwNro^
zds~qQu-%NVuz*>!w)WkI_b|$CAYLK=Iw?@L%#&{%ZJkuMf3SUb7wqQje0p@FkZ$dP
zDnlJOJjNpv4Iqubw@Z*gv_DqxtTz%1_YkPZ)dR+478DfpGy}u+1kwtp6&T_<w?0XZ
zz8S#y=e84}l`ILtK}$^v>_m;?c#fhoxhWTli=ov&hh@NMrr~Q9{gMqPCIejI8wH!n
zvUbc6oEbpZw&DEu7<_<`ihq84C=&@68DQc)BCN`9MH_kCD4x_>qV9aKA9Gze>0d{Z
zff^bbP+{uWy9E*1EBJyE1t<7*`xT++OD1Z`iAeVUSSqjoZv=%~V~ZsEIPxR-?5L}>
z?uf*3?Re}p6~lGrgTw;A2J5a4Zg|;g`B$h>u;C4b>cMj_l2*_~E9P<D<r;(fx+Wq)
z3M;B0x|O8SYwcL){`it73>mP7?=5_diH$Brs1{Z7SJYEMbL)od-+|dl|8*lKWroNe
z{^r1?-?;ZYO0&<c4Ru}4UuqnpODx<es`i(J$4{0zr}Rr!z|J`&QI3Wu6vI3v#yCho
z1V7+xOHDfP_%js2!$nFzMg>c-?z%_-wl)pAM$_3@<OKkf3MW#CjQ})(Z~G#}UlPf`
zpAtqgM&yZnZmg*1lm))~Ur%PDZitM@yl%L@zFSAV-afbJL*KXFBX%Cl<L)SBFS$6F
z{N7P^M{r@dH}|b7zt8#zd17z6{)*ORTpNV61ZZK~I`=+;Asyhy>y8PS*urQ@PdM5&
zlX~BOjYR4})D6==o;Tv*7iMjbWy`8<>myWNiIpqbf7H^5bflhN9Fvr0VIRDjiNJ~U
z)_sr!y(sU79L@_#`|qdaNWSbJ8^K+=^DlL+;zKXiqK)Ei+&?FY?M~)KZ4M+Y{pyJc
z!<9ERUC`3N+=V}DV{xq|v|I-N05kr31nKQL!x-nzpvkwhU%BC*myXe2&TuG)Z=K+j
zUK&|O=)}bPOcy-OB-}{XbK`Ny-GvLfcD2X`@)e(q;*OpZ;}npxx|0R<_4SF@5OylJ
zDa8Lx1pEz4s?(frGu@D~;aPc6C@61er&plyI<!}wd*s&5$eIsDa6y5%_Yz5P)v&Ik
zhJn-HI!)spRtRy9;;kjTVzBuBY68#_vTs`UjRVxk=bt=za_AVCK#Icp#~0gdqF;eW
zyoy^$J%f!Rk-pnxaKvY6KJnD!HjF3wHf^vdi;;tJe`|&IMa1iM>q=B>{j|=BM!G2~
zTaAOZWK-eWeOjL%gM&5MVS3Lbo)P>fw4;L>LZaJ`JZG&dkq7FohjE0*b$vnl%S$m8
zYdHtG++8rSRkvJg3<$ykR)#dYDAvHoZ0O*b&vQ;k<taV7zlbwFJDZX06EprJ_Q4?<
z>)!*mf9)$QzX8T7JQrT*T8eCHxo);wX04*Lxp4-!d&$t!(_hdXzt)%uZ@4M%3tQ_5
z>sW-YTd#n=LDo_zA;`rqYQfs=$_##w_fN)vZSk)o0}htOoji)f#g)fXM?-`sO-e2L
z?%vML`*}i%<tPHHnbE54Fjp7j2J56xM{}t#9^<N;YuuMeh3fK7$mH7aGgit3Wv@ld
z&bctl3(0^yH@*LudwkqFcncGO0SUEM6IqfI^~gdFeh&$200tBsV*feXn_zjSmeGRX
zY-R_t-icv-WISUH4E>sCey%3)PPA*Y0R1V69;c)eR$o62^vMlnWoE$#0avJhtb#g3
z5mnL{KIF}Z3Z&3!g7O8A?`%JWmsO7XGC}E36!UYlvv)Y*zsK^fvk`_!z~p{)<5>S8
z!9)#;Orf=tFnPJSYJeuW^FXfvvQ&^DOf+gt_TMS7Q#M@aN&$hVJTB2AH06gtNb=_#
z>3Ti4(ReChzFQ1T@W`oK)>^J?2%-Uk4rjW+Rk`jnzQIQvX-R_~o@QodE&&{Oq@67p
zD3tg=M!aPVk14+^i!E*o3a1$G63BIgY?q;4TFh8SeoDpM^cc*;gm@gmLT_uKcYtU1
z1sfv$_dL{zI7=SShr$(F7SVwEqFd8ya5*3uH3SP6mez>EApd(I`nyL|pLRlDy^;XE
zq_C!9c4~8lI4I{<a#hr)M^jvtPi4R(tYahV$8Q$q7~Qvxm8`qaDUVbR(~YfaR@wsL
z9;i>Z#vJn2)_G`Z?vblRY0%mExyR^d$;MZIY}YVIKfm*?%TQMzKFSaTYd?|r6iN^+
zLG`bpN95H_(qi91ajeRY%VWWO^k*t?3T0srrSO(8|B2x2T*K41_(%1+&K8CUs8#!Z
z)0JBsECbGk;eEQUfS#Y6?509lPHun=QN~a!kC~YnanTzlT1joa6*dm3DzkJru)yEg
ze~m|kEP1oqc(a)bBRsCno-(803mYAolA5Id9C?CjSg7^F<h&`87J`nlGJ=g5W$s@D
zXI^||jSWhM$2C9su6SsWhKfF=QGdvc_QnoXNG-NuU%qr*9j_deBZH@X!qx9N_gfo6
zzDt{$nl`#GJP84ib?y^imBbq^!5w!{woZ=nOt^*uIx#~kq_AN$puwVEwlcG#66RH)
z(HFL&(jl#g3Rw&T$BdkO8Hcu^?5yCc|2q9n_Op~QG1_LGk>Q3V!s3O>zVra|2@&&D
z;~lJ`yVZd5pTm$iYjDR?j)v2wrY5)%#1JvFf&P0Ou3J+;JP-pTDJdzxmQW`xzZuP&
zJafPp%+H6=8(fn9nM>kdcy?tZmsgzy+Uz^|^QfD5$uQdk++L(>m>I$gW{!?S?sP+5
zA30ZDGBtfGoTUwxUf=|7zisg9$pq`)L3p6#@d%hT;gW@@pOt4U;f1BQfyI&2jaw~2
zSSOSAMAm-(hPAb|7yRnzoC4Bt>o%Cz3Yy;6Ifa~ajzcRd>MsxGmL4U^*#4#cZ*OhI
z2dTuJsK(O0XbD$|Se28g{kGw5T&_b|5@(aCLG*do{R)l~bqv_fE?~0-^f)5zudy<+
zGd%OMTNdTS55zP`3U6(!D<{4*Cu#I`T1SStxs|USLT*l9EH$`g!VHicyE^dz0B3Qr
z!7c0nMu|@FKlM`Rgrb>(50daZ#+)3f%o5t<^lHXAaDs8+M_!|c)eyPsGJoP;L`?=3
z@9Tdf)%j-60}@lxAs%Me1plS(Wv;7)K9a>dt~!se^s>boN2o4f5m5eVYd|a*#sLyW
z#V*|BWo7OCMy{RbUf*0n9%Jd<uEl}NCfiB<G0=IV=D(G3%Z^Br5WRd1?g4}X!pd#<
zZv$=TcXtT?y#}RPU($QLWL(Xt65FwALdrej8xc4RDRFN=s?U&R=b1~fU_t<Q7Mr(t
zr}J6jg>=ASb)|XJA_@r)<B&h~x%=+@d2J*;Gt&tOHO__60Kg>E(9rO9?EKkc>>uw1
z`s@@-o3(Oj`!SK+#t%7Kv$EqSDo)cv<bRUBOgxLwQT|hj0kQ!SAI#4q)g?qir$0p@
z!vp&);$W*n9BuQn&4o@REzCQk8+%np;C-{KN2uNUxf59(Edo@-vVh59aRyqySbL%6
z<?m2EQ0%u&q~G$I0HRx99v6J0kPU{QwASvtHrMUr*Vk0?;NpePK?y&n{hJm7nj}sQ
z12-Zl@pJcGub3*TB7dmbb|S(f6@ZH;=o@gOK0U{^dc=oR$gyVWZ!k%6Xj>0AMH#@h
zLmS$RwlbH}F5H>ox1N0s_~sw}<zdy`^~J@FBpSd@EdbU1Egp0xzh!MlN5{mKA2%u^
zxe>x}ZT$TCFLTryeD!*<JDQM>kI$n!H=bn?BUvt0!aPm9Phd2(z&BQ>RrJb#VYM-M
z>4*|<-yjmp!#K1yzcYbocWvd==XDR-=-;^ch9*+=haqK5Q%X4*%}cuDOKN|^c^XeU
zw6#F^55|vAhMp`xfdOHrL(73O26`M@V$%X6+{2@CZJ)^iMM5vJE{&1=89O2oKGhE~
zM02>FdU;JgAcgTT0|AD$z&ppq!_TZCkOd1c<G8qp&6sFz4H(?fm&c#Hnfg+_mn)K#
ztw<G!DgNdis<O99ZZkH+sSd@j{M#PeXEl~pcN?{RyEAHWD=okIt@5OTS9SohY>va5
z(cJ~ZEW&6SkdH$E;fgSdPTcbXKcc9pXrc^EMt4(wWUYZRxJAzo$T<#W3N*#X$E!6O
zocF7VBFzn{68UfKFQ2r1d-aKleE#K8E!fr@-4N`yfAFW+vmmA?oStJvfPr3{fM}6a
z8v7rQY4b!E#Keey<G*Z@C;hf!%<wMco*n<VKoc6CL2^rpb~j!b^HI6MnHiYFO!?B>
zjxJ?59l%8q$jqEUJFN}Tf1OYOa6K6r842BF6T%P##8M#Bd5$RK3KOByD0I_@ht%&w
z)bJz!CisG(2SmHnyXPwyImbRt8uvK4s@t|}kG$Fr(K+SP%sojo&k%YFM`zUbZ*cyZ
zICf<Xh=lz&3KR?E_$C4z@_lOLzFPpq|3y0V3dkY6a05~<k-c=~Ns8al?a9&X)4w_8
zZj6UuWhj7}R1e;O0*>MYB!BjhlJc(-zEj>!FNZsz@`a7|$|AyRuYxNNw%)v1VGShu
zEs!Ble}$53n_aV0k*Zqkm>p^yune)ZUDxx9p-`#|eQIN4Z}G3^Otzg=P9)OzhYsMt
zWwCk`V2*DncA=b{*L3i|r<elUl)!_bxbKwC#CQ1;c()$iI7o$c0U)>0bPcqcSroX_
z$Fu<=f`ILx6dft4yv}e40B`a*T5JF^8iO3$41xEE#t;W<(A4lny8QCRGY=oDELj<c
zMC+ZPO4yuh8+2h#+w*)AYoTgMB^HDnrp|x#ubt^VqSify?uWOv4MKkm*Ga!GS~`m=
zZCq}+cKi0~3*k-Nq;S~wiE=tB-`s|kFbX2;+JlZ0=y1*!_qsYlEuOB;&i0Gzr*8lL
zy#xS`+yKq^fuqJH&qp`tmJb*#gs9v5eb*;M#2naX7BLZ3uGd9<$7zw#$9J=(M_2QO
zJpmJsu?d#PqIokp>1oxlCxP@i`C-dfzs<SZ!TTU``Dq*!*0E!IPXm0_{7Db$yrQA1
z4?Li5nGI13o;~13P#W(s<+kO1G(?FLA{v0&0g{T};#2`$Hzwz{)Z}%|hgPjpWwQnD
z8I8f0uK$p_mmvxYkxWDKtlHGC_*ewLgQhDM_)s!SGaJg7i$^D;hqalX_)U8GNT$Ck
zAT9Rhd^P82K%ji!^b(u}{|IY?%zK=ihy>nbB2kA+M}baNHD$vH@z4sh5Jevh^fTzi
z5;5hl3t&(PpifgBpLyfWCUP_n>@`<qNr(K%VwV3wxeMlq0waErECG^8zXCzbH6aR5
zed@5(F)7=yzf>6>`}2X2zt2Msw|AtZmsfY&Fnuk~bnUS3=EKP_{t1DS5!pBI$^wt|
zF125xPO_!u!kZiuRxD+D&Q>%bf&+Beh+r;{R2;;gN>te$&ktyM`8gdbFKCA*WUnw&
zxg7N+WsIW>U<C=0B`ACR&1-}sgCSfuNoWKvA?@1H30BCEzQQ!Qpo9pb&Q{sz3v%Lf
z+o7*co9IQrUGD~L`nZAPK&P1j;{iaB+9Jn@2Yd~Xn3C@<N4CC5YT5i=b>)A{j0P^f
zN5;i`fx;+#4{hxlV&gm9Fyl^p3hYD$P{}$P5P}aM%uBqQ5`qNF%H43B>hO=;7#kSO
z*E$TjUmY$O85r2RxVVT(NJ!j?;>YyhYZ0X|S5LgLHQR0AXMu5a9uZEAlA7*^HzMxp
zDXzR4;774{n{p{8y8$@6stT<@0RqU~tvedSDezk8j>@YRndoS%)ti=2zNN<<DsB@V
z8U@C$Reb(BksZhY7dL&4zG1eiM6Il>%mRrzCnN;u4smA+89_AF*MH|9XS-p_UicfW
z;p8jD!>>=S1$a#D9<5;fO<Hl0THJ|#&3uimki6T1+Fox(T{4%R26XIv5K>4Mn|zS3
zqoe!F*!KAxpJ&RhfH7YU@#TQK+f-27Rx9?`FCtA&Y(q^nmu(PzML(b%rUAkLorGow
z+Oxeg_OBeoL8)Vg*4w6p<%i}jt>?LDET>pa)e<O!*<}VD@7BuRm#?GtXUYRMdhwY|
zTt|Tu77!!6TW7Kp;Ncl6(#*020&OEBqn!^J1cz`q962YV3fg_nRK5<DmR`a=?OO8R
zI|?>XQBYRyr{hO1u1^l^PC|?C&{w5Dq2qElSp}O|Bn{#wandeZT;jB+uZL7ZPoc7y
z6U}+%%Ql#Q7rt}My^r`d`-pvR1KP4&Rv`nM0m@900oH?><)gtKEo=4!2rq*gL(*a!
z>r8G`umO%~7y6503?d$ucq1K-X!iH_zg>SqO_IM4Zj5sMj(KzQGZZ0r&r+R$vvUr<
z`2jQ7AcBbv&)3(NS6qC~TKyq*l9ZLMv6a>8F8uZN-W-k^Hd5{n1D5zc_p7>-tie>x
zX1NE<{wp3=Pp3>nihk9f5H#;v8Q;?{PTF;v0=50eG!kaQk%C4c5THhw)5B0hqM74#
zoj!VD*Z|xkj8am5gXP2|1nHACJ3G-iqmDo{VxpoFV|`q#p#m?>vA{qBLbZoL&$j>|
zy>W1Kv`y!M$xgbVg0%r$3+LBLu;ih*mLPS^k^^Q{B2JckjT6FDtgfXri}6kABl=fO
z8bkvjtnv_8$r2XxWjjM-dEuzQ<CRIjQ9fcnwWi4Mp@e71kMi%TfH)2apFZS3uSke*
zDIqH?W5fKu1J~ONAwxThUUL-zag9$@z+y#x;pEMWOxBhO&~e<lyI@OWUdnTfc1(1C
zzcR$74h=-NN7JSHayN6suvI8_q!iX_SsZBdzp-@X@lgHW`_34<B0`Z&A(d<uikQ)-
zEGbL&tx<%?mNm=Fq&;h7OQIzEz7s=2lI;5)W#5;<Ecf@GzQ2EZy?WJa=FYwMea>^9
z=Q-z0T_GlR?t4H}BMPgs$Vz#nJi&HcOt}IEAsF4n#>MhFT~T+Fl9n!wii+xV4Njq;
z@dmf4T`dz7^4{Gyit5!!7ArH+umz3^frV0nGW^Qr%Y!bNDpiC-wMC#(v=6f*=B1kO
zo~i2js4EjV2?n!4WVc@9b`2h;P|h?MQ^seI^xmg2_|K#h)paHZjdmiIyqhUcN1H$2
zDB8uB&Px$Zhw%;$=>)D@t}ZTHG5FHmg{}qv9u0VC$fEkepC1ds`0Lc^`sy2tC>7S=
z*9ZwFFaCoEclx(7bleP&V_OgsEWfK+-e|EYGY2OzE3%VOvnXB>!IJ^`)@2*k{nTH?
z(d9p)xQmm@s48lP;WJ*j<*Bn)tdH<Q4|34|yI%gyVJ)6~MYGZt4eLMduDUOSegIt~
z*{RPgT*Y5XZrUwFvd;jo<g^lz+-#bL15`HMqQ@7^b{GAbokpsy9V6v95eyBZ38W~*
zTe4{pmLxOF)j^oCXMPqRZ_V6K!pxp6eBI3+90b+-sTZW9)wEMk#Jm)pl%k?8*>mrd
zy|V~LQP?Erz0$Q&_UOqAmn9Hi#k7B6vn>y0+<g=msEN(Cc7OTEbJ?Kfk($z;(b+B$
zx$h+Ne=iQ-?Mz&9;^>OUxkBbnu7C*3*>XAGSJ|6!!wt)K=RA7xqma}ziT-!+gjqHz
ze1<b?&#9X`e8ErxFtbPb7g_VOt3-_`{L_2`Vg@gR%PBSdXGDAah`~Y3G$3Vcp-<`G
zy0t7C!}bF65<MWBf@-+JFD}ojS;chrYd+1on&pm9QE`i|*ZJXuU-l|P9=*~BF8xvY
zxNd#iXiE1VQyO`v7cadi?4^0vn_^<>vLf61-9vL%x}^Q5h4b8plFmMJn%i+B%DB^9
zLRJpP*yH(H8Dp~dsk(k!o@puh6!2G?khTBCjAh*T)w)p3*bUx^y_LiT(sMVXN^blO
z_ah&+ktj_rmX=305kwgYwl_Q-w#G4*^X3MV&Fo{R56HcaWt*R|g0*atkK$?LrYb0U
z^I+T(e38^IDr+`fg}D5meGV6jLF&paXVa9ouSw?y<$c5*Y%2=5=5+s-Zed%ibYbOk
zRcXPG;9Wd3?H*swP=3nQ`l_tGtYs;`{??(cd__7M2r!L7Rc|2uiF_Q{MEz1OO)HJY
zBqZEXg!P6T&tq+o{#?9J)n*kh^N=XdJ1d@dMm!`<o68bR8nN)cJv3aJgur-ML*vTD
z4L;}lcavYvEicu#8*Kw<0)p@%C4Rzae0C$L&4tu=0RQK3rDCHog;44AGRU>hC7*NE
zn3+?rBMom+n~uGuUUPbnp)*j`pOSKOUA$1(4f@+Zpei2M^@cV0Ep5)Iu#V-}qH7J&
zW(N<Q+250v7R*8KJ)EM`0!c7$7DX`&Np|FO>IxOGf5FPl23IZmps1q>)_5H8*0PVS
z;jZibYt~HffQk?JDx2$4bU*`pUsKAwI+svZmN8^+&p-70%UR1em_OdVbo6X8TQ0r+
zs~Yq2TJ!on_2lpgm4NLWLXS6{`Y;*l2qg(@H_Fa&#^xIgHrve}1T=q#hBC_NkLg+L
z6a=!HV$hx!W7s00X92vDK|-h4n)dZb$aU%-J|tv*^w8nm-JPW>CZ7q5nxr|`Qfq1N
zzOMeM=<Ei~bRm)d0jemKseDCve|gF2*3!(ZQn<U@bil!WY_BRJZ1)EJSfbVqx_S%-
z{|qzGl^<97MQ;Tha~Bv8X;BkxLvieT@`HMk9V8HYg2h-lIXP*qRXxlpu0$Iv^z8nQ
z)_L^K!)=ey6+YB6a8%FdEliS}F?COCe2P+As;ti{C`pfsie0#J{2wI89+#qj=x~q8
z>SwNRbT#PhqGl>eRe#49-gz`?s@zS|n=yY5PS?$vp?n*hae%6m8qOHcyaTF(;&Tqa
z2EeFTTH0yS5Jh1F{Cf;7g_a-@RA(v4{JWX`4O<8Q7wZ`O5|2L82DF_k25jGlGjx)W
z^ZVJhnZJpe$Iq)Uy}4w39DTM+Ve4ye@%GLiTc6L_jr+^23qRM&z&>FtPiQj{cK?)i
zV&WmHb7ryy?VPpn{C!{@P;$sA8P7D|7}}RTjNtK5O65!I%eT2zk9X%vK1~IJQK}9l
zICXEc=rm1<;{w~xqr_r6{akpSrQ;NmV{fad3VMU;xLLo`&9a#hQO<r=sAC0RrZJ_v
ze<yk7Rw1FXpgnKXXQxKW`mubgSCW~<2`z9WL9tv{%|R_Kw5TT(4$opFn_bp@>*40?
zI6!fqqyENd>vRs%(T=8#w~Xl%8ES0{QUj*mFe7aPz=T_iNdD8`%qr+Y!~WJ}2@Z!u
zgfmW)!jJS+5P6|_ws;laJg~O5`EOg|Of%{{nD@Gt`Q^I)s4<fm#?)wQ92CsM``kOX
zE1Qe0bA3J*x4gABCF<_iXBBw)zqGw_VJD*UBEfijW+wJuvNnc|f&3ovtotT?C6kCV
z4wPQ)P8-BC!xj_i3G}KeWjG_E(x9N-1aEU+?^wspx`}x8Fb>CuHU(PxNyw@m1)ei}
z*S(eY2D3#F^IR8(@VzVCFP?R20h<a40)7EH^k*8bXT76yZ_sh(J4A*!cC?>8^6w0#
zp?2QwR8>9h(PR59Z&dPMGPkJOVm38vC;8IDuJUcx7<|yXzwqNQnnsN1A>US%@hG=0
ztG)v<3%<tj6ZGQ<oB>Ndkbm2g-4qlQ@aaS~T{y{TiK7D#c{|L_L@G0#e_Pt`%@jB}
zyHQr0)|Y%R<qCFn<H)~)#kFFTP|@qvaSdqauYX8vKlfZIp=v#IBkSo+z9I$h0-;=V
zAiFF7FkYlT9Yq|XY-Ri)ou;A4*m*<u{CP&L5Tr>w3Uu=nYefA^CjaYONr?J(hoqnD
zVTQv?)q>muy|1r63I%{4+=T*Rx1^(+xv<kN+!ah&thwtMC`lx}yk7r`gnVecT-u?R
zOUv41<tM1lyC@gsS62@cLXYNi)3sTk%o-)E3#@z!&A`SKAwK=%SQRD#8LL^*QcbJ~
zYDh3LE1fk%DrKC;WqN#Ku<6`(v!>GUWEso22#@v}pZUL<W8_m7X-IR6bi}u;Bz)V_
z*z9&P@b8!i812LoT;g;UIHpSzaJ}@{SOMIiX$+2&X0&BmO&S6Q0dAWI1{9JmN5|`6
zA<m9e4OEFrN=3y$1ougU$Sexd%R(oXbC1yFH&pxJPPWL3<N8Z0Ve^3>VY2_S6MN>(
zgYmD89f4;iJ?N_oxx6#C`@d64Gx<&g09?7(BSbOFvNbt6F?b24p`jrvKc-tRMl{}P
z6d}Z|pOSw7T0(VX5Wxn<11WsWJ%>|d5JA^ZFSB>FlT+s5q^ifRXC@?6#z&R+@XE-4
z$VmI<n&^=P7UR3PVj9EF0kmCLosD#h{A}K<fz|NpDJT{xO?MjYKZ~(`yMO_zs)&lu
z&+sD`e_PUkO0y~|I?K1pB+j&06@*~;9ZXH@em92DQlp@G0U)!ji6E+ON*t4$j&K(#
z;ZovtFIaj-+nN(uo~=`MURc>0wcg|6S^>o;BQ+nJd!`>ld%Jj<;^w(h7GW%N^dZGv
z|DBU3$UXgibiil=@T_`zf;*NvR|s}bZQ3W-iPj}bFl}zAKdNnlf>W8iOHwpY!5#={
zo*s%}rWCdsvFXc&uYy-}o-bufmJe3unUKaMzLo#)3p<g7(!ggqo@o089fv_B0aEkd
z>4Vjk`u!(__NQGh#Wk<pO9ZsiP+vZg0Mt33*^XnlC6fE(*V*Hw9ZqOW`!d>$B}0t4
z6zvYAhq4EMu2~k@y%gosHzme$%+r+B(fk6erx7D7zgT<Li0R6SOB>G#cM|j8dr)v6
z@pkAdz?}kK_INfI_*nTQq0Zi3a4_!gxC5|aD0Qu0d~;JN@GaOFj#ICdkzVIi-0ko0
zAC8eS%_%SMAAiBO$yAW|<-Gn&C<&b+1BJ0<NMR$++5aq{QUi6;bPH_0FMwityU|v5
z;9|KbdR~K>MFZ~>#qsYn-&||zE%m(>vNb@l?to20j&5ob>kr-o+W(=|UcAe^o2%<~
zJq&iZx`1h__(YODt@IDBmAYAoqCvc){3&+CZD{rz5x5K@ZgC2nDF9ZS@qH-~oy{ys
zI#Z=8Xg`??-Mx3ly+e@XvRh?AE?a#=<F@w}S^oFs&;Ke1j7FuK%S^!Q!IY&S!<o<D
zl2t0qO|oI|i9u-~9hUAnN0V-u9LM(JgSeo9L*AeaAyy%J5ONN`j*K5jiHkX+pN<6u
zZ-p*omCVI<4Q^KiWbDPgccOgX-COUl{?-je{Clf@l~FaVgdIF_akM*)mnp4J7hIlM
z?!<9CB4zDVoo~pmBY-WgWdUSY*bh{JgeS?$NUymm$mYzw11Y=+pzhQ9FVuCV>A3V}
z83S+>h44?VLU^Th{jLiE)oHT~d&<*CS39qh#E8Q746IbwWY~+Do>409Iu3&5auF`k
zC*;r7F2j$%^KJLex902Q(%!BB=`<p1N9q&=O(keB9o$?ffx=K3VptL6kcgc3yiu|^
zu%_JnAeG{o(oQe+A<_P{BwNH;qO(07scYo^JGq7Vd+F6#u!3*d#t=~OhZ|Q)fxB%=
zdA2`oFvvP<Q52k0Qwo@&WDFO@?#=C}NpE#G24aXt1AHllh&OEC+uBNtiq0l;nHAo7
znhOju$!L9jeL#SF?WXSDL?3YfG?hh*`!DgngJ<sD&sQ}Bng1~b1=FUq7-U7)@NJU3
zZCa2AT^R^sqUDcBjy^bkMx~0G$v;hs-DB3=1xe<>DpSErx#Rj47W2^%So2W~YFJ7h
zw8Ywxi3A&}@K4bE%E(_;S09l)Cz`jfwO%Dz#XkB6Inghh_tL_Q!iVnEloh3n)D}6`
zj9Qre5l{@f|Nf|E^9APmK)D&h+V86>eyb_@+oBiT>)OfAtprQUM9jRNH3E_|h=PE{
zhJLdyUljr5LOr4~6If=hse){>J&vYV+7X8N1K?jPJN>qJ$vSr>L8UI}rnbelq$u?`
zF#cv5t@XQ@hNyf#rTWG``EBh;ZRO4}%UDdW>D}MW$fo}~WK~sFYuiY`dc;b{(B|}9
z7W#6_?zRdrlrKVcFu{U`xPE>w;d8Rf-r@9a9j_?u|NeW7`D$beN@3jntZ(VzJY3=2
zM`T-w`c}Ltm&E$R5}z)0Rxt^9q(c7BV}EIAL<v$iS!}!*@QAeAbt5HFV2d+;XyhII
zd*mzyq{aKUZ)Ngpj^@mL|43T9xy3C&nKMu-Q?6b4{#p#*#PB{KYJqpnZOgP#6uKWJ
z&Dq07^g&%uRvV3AIfJVm!GSGx8<MQwByTnwnqtQ%h&3bror((Yihxo&>fMspRF%__
zbdbkG`(8K~3Jc%~em}>yac6%~+E_iF(48jz6ZU92O_{3nwQRefG9+93Am1rYV29b5
zATm|%(0+8d7POm<$qIyo$ske0mMS^@xGcKzMH?W9m-#GNB%}0iqA{59qT*uBXvCEC
zfsRRCw%Nr9b4@~W(!5g{Ls{8xRsCD<{T$gAK@DVOT*8<DGo544@{X#af@-ttsf+@k
znEUv)aFPB1HMsL6_B-rG)pyV5AjR%34VSl@QoVDM%k~ytU}{96RUrw4*L53j@6z!D
z1&*D^33nLhI_LqTK=LvjyfDblvl2*FpgdENSxs&yIWgx6quV44=KdG87TW*-qc>f1
zp*wR(Es=iJw0^!Tu#<SI)bwv->0U(oMS}Ep_dAlkx$)$9wQv5?K;hIrenKn-QJ3h8
zay*a{e!!?B-<8N%vSyNJHHT{l+3{uYTG1JIyVC5rd+1B+4)<Pq50hm@3iIC&*xy+G
zbEaI<CQ?_H5mY9^Jn5i%yE(WtM1?ngmTm_WyA>Zcf2ve5CnU7-mJrB)O}(aU58y8y
z9ePPGU=RVfl6(Y$z64pI%*@T_BDkyyY#yzkIt+sS&E2?tazfJ!i4&FD2>8P`3+1CQ
zyB~ete7sZOw+;t`bNynn)gNY9otSDA0NiG^4a+Cr+BeRb*M40ey74|Nqz;qv2Oks^
z1ay~1oG{IIdE}m?-uq2{*e2VX8g?3*aPUQ<v`tFs&bKE<Dm|CLrr=VrY_cCSKU=rv
zcLy67Tvc1%pyoMoZ+OVd)yrxhqX>xaF(w#mC|ft;IO+Iw`!L|P8|wUn&-(bJx$|rK
zP8O+*oc)9F_vc|1_X*<oYQBy!!Jdk!bKMJv!VstcM2Q$y`w`emV{Ty3Sux1Xu-)zI
zX!hM&yjM<Mp@v&Oi11gll>|0$v@|7}8~<95i8##y)7>$c1zb#m8}61LiMISx!Q_2{
z$|c#uMIL-qV6S=%@a%YcE*%u7Mf-#m2f%2F2|@a)Qz==5f<Yn3-4h_alf4XYs97)A
z*F)>f8mj!r2VfqN7u63}6<&8;*v%eiraTp+I3KO?8jvk((++u{3=!o8Ck-L=U_Hq-
z)QfNP!z<?P-HWa*hU%4s<gvUw;+hZB5HQ4A(!4(<2L}OfNltPW-G6*5HyeVgGi^<;
zA-%l3cCsrg2Q4fumwv=ZF=`>p3il7aJv}=#=o&}pb64WTp<!H>)ifWHK(LRi8S-yg
zPdyx~xPFO^VX7y-9NW(Cz$yK=lC0cL$Ik`0_~^dYW0_Txb$N$^)f;goqcC6_YrFDa
z0k^60m|jc<dMEuiH`Jp2edzoCSb%*oKHk8NM);PcTmt+)XJlpFMiyB*Jd^SoAKt~6
z-?;BshaU>8kEppzas5|I*<6;Ta+*gtKejU0zXvbGa+_=Wmyr&vTShfEEdJH%)l!P8
zs=P#?f63p(fc4*H4>#-yCCRQAhKGkWhiM3?MGrOWP|ySk2@3~F5;ahp#pW-~YxFo#
z=gytOl^eO|_#hp(NoT~GU4fH?`C%GfeWG$%Ammv$bz>Zz`7f<p7q2RPIUi4=$8+u#
zFUpSmgdg;8=i6a5E}YGCJ^dJ%E*au$31d2Mnal9s;p;I6*%T>EP*d|u1gLO5Ek`Y?
zHj6D%8OAsI!AFIc6y@-e*jwVGt{&|{<&{^k`8xO&A2CF&Fq-#nAQuDP@o83Z(n}$2
z<M45xVLBn(zLYd~cwquVf2Oawml^dbQPX(Oc#W_<=n!-~1!$KMAj~yM(n9-Op)iM2
zhg15uJgqS^4U?-2;rot+i04H?77wzo5Cz7@%rCXcvRxYMwA)wX`r*R@OGk6$zgR^~
zTmMPu8TGUF9Hy6Y&G}ld!k}C$0=hSj<w9S~=FYzEq0_YAA+kf!dTMi0Fb7rEszOd^
z$1n*&>^Zd0(b;(;#8U75r8&_p#U+-Ga7cowX}vw&T1h?*YWEn0d@IE3ih?uydiw7=
zR9R4E-o2n~tleh<frI#l&*&^j*(OObZUi~Xe#&CbaL)HIC<A%cgcA<N-ipZDKc`{P
ze9dLBs1Z+{5~VyO)a32!u78KA$3UM<c(E#ogVby>EM|tb5yhETDvX|GIh0)pslu7U
zHg&2xzkbLRd8aA()MoeYFHwpgelg|CQ~(Sr;69(?r)GcsFdVmKXkVECvo7-Www`Hq
zQ9ZhlN)OH=n3~3ksQUM?Qr~V6=({sZU_%o=;mu!a<{lH)NfNnPY!!lh+zDALYv>0c
zdiJH*)tPyJe5wI5($U%S<4PRv&IU2V(cKOK!}GbrVL$)u7$W`lBSA{Sa-iTUhfbGP
z*4CLp%$Wd^3i9479X77L*TljMDywkH)}fHboXpG?R!aM&jZoU^a||eKhj#LX8+Rg@
zp-=R}N7r_4)dmg%1J@Lp>7;y$I969Td<CWNP-m7TkRpF^>HDd0`$6l*ZrY2HxnK6q
zUIB=Yi_8otU02~o@m10f3rZ@hy<bL4<)o)qq|&Qc@5H`VEQ=8^umsSSLVJKIrGWjp
zy^N}D_`gzCVu~nq;sh<U{~p5sXvZDXFxJ_-%<vF*iXl_e{KeB}iv-cbOUxD3cF&}}
zv^O-$1t%_(6$npk-qidfYYzEhu4@d@M@8PK>MOU_mQ1K6MjPH&&8soUg%zI_)8Dch
zbIvZ0R=YGj+eaU&^h^i*p1GUb%B&ecGAHgU5oKsUO_5kmWbbZTUctzS72%$^O->ae
zxCW9JTx|zTEUPsy?TZd{<eu9RU0%=?!uX|r2ez82`*wuma9L({a0!)3v0GYk-oqu#
z$>B&VDAs8V_eS==I1Sd3?n)4x*n9UKH6}pump0Gg#31F#tL@ytUP3p}VY>Q~`19va
zd2J5XbX10OULE+c#ddG4L85r<VoIfSo9}01BthYBTl!_Zt%nXFA{nTL_gWct%VJ&-
zFMbe;fY!JhR2q==BlQ~TFO{&FDms{*UppU1&HUZ)4g*~ubA^oewk~~`z%|#+-@IcM
zfZ*WdlvY&@I?o<St%2_0WwChiZda+SvRJbS5uYSTL`Gh2#Gu(zkWsZnjjfB8)qX&9
z*HQ;BUfmv6a9-r}Tq(S4S~}}F3LUo9-{lV817Mppt*1%3*O_cj12&{dw`HT<W2>xC
z;${-0SR{;X-{(KpkrA(qD3QXpOIjSvXl-(oCBE0`Wp~HTU5CpIT`x{3#JQWUTesI=
zibepDsxGQ@QLmZ*?$gxm)t<p^ZXHmJ=9x`o3p8+sbVc3T;=X0lO91Y&Z7`0Ca+MIU
zy?9u4C?X33)wvchzR!XqqXtrz#x^tCJ2+T^M7&Uih)Qzo@2mtJfTL5fOfoG}#Z~VW
zvsaef-9+!m{;R`z$Jg^Eu4Y}!XoJh0Oh6ruu`@jPfcAx8XNUJV^UJ?4EHPGjPKy*K
ztTeRIj#<g3SH0|}AfSR|kDJ8z=e+DvG?+>xu{w6%Mk&gmvMG?)s&HN=wW~(7dAA}k
zMzwXI$z=%O98s$0<^O%7C*SX2f-%|H!^}YcHlfn5Cdhy9n&@hC;M3!6l5F?608Z^U
zFQ#L5>bVzRO1@t;hJm@CaIV>h01;~%49-f#Fun`vWt1bpBp@hAnL;)n-_3gsNep?8
z>?eX=8Oz5iu4Kn%;{_kLkY`;xUl})d;)E&PI!wKv;W{Chuidts{&lLakT5BqvB#=(
zuj*9y@6Ui-8TemJ)(k(g#(xA95zw|w6{3V2s|`scQ1<pTR1kn;ef>z0B>?fpy<J<U
zoHhMt5x$=vgpGEapTVyGj8GGQQSkETsnq2cBsb}Y-(61pTD67dv~%`wqSS@i6UH>9
z`vW#$iOm~eKxqZ76Ps1hz_R40)UI$1l>r{4EQ9N<260eMK%01;mN2rB=(99pX<^&C
zo<5{qex%@3-=FcLM;|}A#CTgzTwlcm4nzjS0BzhK_u8NXnTqu0WI_o2cP0m42^IKb
zKVn>{rfJW>{i|#HtO{LI##OlA_R7qOg6eNU-YjNOWF%K5oXKPpKFql)Pp0P7K+i{S
zLEQ(Q2evTE-PH{iIP2ue9XOVgzG1z(LWV*Oo~qZA05+lmSCR|EZ7ZC#l|Ii6X&EVD
zBiZzMu_4G(aXDGp_T+;+`=3XTB?&H0w0NbXW;3u!Jno-rJa+7uYVDx0E1)TsSk6H=
zHs4ujTqy({AcUYZe`)*^zCJ#CKn(&042FXzdGrX1hJt~(zq&rSyzHtHxKkj?+L1Q%
z+3buxD0O3_qoX%<OELm<`#g)^r^5Cv4Rj;oN}tGjDze`5B6LUu6j1gZs=QR^zc?j<
z$7(xz_#5iEu^ihtup68-NdQ(sFse@;_wd*(D`du74Iyi4YWhY-7RgM2+R-Kk2P4N1
z*w}g)r39$;+<Nbw$CpxDKDxg|;gGEd13&u(sFzp2-f{eM9CxGR^?#4?h59GwdIgww
z9e%wlOr}mtZi5jtgNy?tCF9h9VlW(MZcIV`?{D&=FPzq=AfnWG-K}{iuDLzeL0SRp
zHY1;xyVCmieaknJEhKrIg9l~6ll)M4!9q*PSe8UBQr-A@uIyg-3ApIAv=Mk!6~hHk
zPVBi=h|P>|rLo8?J`e~_sIaHZ+?PNs`)U*anw2kAnzXMG8P{9kv5Jx7r4npA1J1Ov
z!ip!FsS4-M5A@|)E>XteT2<O+x6kIR0mgwHME_5tkf7|X%~44Cv>d1zWm%1Q(Fw+W
zstya4metU*bmp2iu||d+EBWtyyYf(Fz3busJn9MSZYjDTG|teJTt^%&%kCF~7MbpK
zZs<^dHKHV61015dz6->tJ`0;{TV(IP$e*dm$1}?>K+}X55ieXK(&lYj^i7a~FQ$}t
zO#fiktAyEfexLcn=;YDF?7b%Wtc)$WeRAAog+4`<l_^%cPX@2B>EFR7K~+%+s6#cK
zzUuI3<GK6B>Q8I@mp+y@`R^hT;*1**t+d0vI{_PZ#}4D9{oT*+l#9z!3`?hbO>w1Q
zPT&vGwQ{#{H3HMsRq#?PaO^3txbLptg)?T(N;CJ!KLXBYZMWwB6(yDQ0kvDDBlG@4
zi&_B<@u6f!u(9FFrdx3IV9wi+6qf`jLm#(dFS68=&VxR#RNB#?O#{-Z-bVC1u;<s^
z3{+2fKLw(|v}kH)#*HN9UP(k#a4eJBg}brVv#!Mrpy8Bp%Ac=(4^DYskD$nUE*HrD
z!Z{>NPP4*qV|oHfu{2dMp^%<|Db;bZ2mdz_t;y{4?nc~uatR+A7(PS~XS^ob`7&J@
z(IQ4caRDRpkCse<;sHwA0hHUWX)nRw=Mk;Dh<m3G4i#sCogVkJ>7EfN7Sjx8#KU_G
zuayh%U-@j!8`oli0YS4p2;`cXN)VP8=I4vPe!VPp>eN~)qTbW?Y&67uJ>lXzlw<W*
zc0`>xA1Ji0A7ygoTWvlqIBaU&O3*&<eBCwJOLx5`S~92+W(4>U*vs`SXcK(YjF_LF
z*NBGZxjOQQpqktrg<AH1jiJJSf0K9e>fHK8R&5{c@fAE|Wpp>2{b;AOrf}9`Oe(+A
zU0VXAF;Vg->It^Csi6LEP`p1dF+&5>q~;?)`-)AGII?8NL=185h34@Cz>y&mT?5Ud
zD#pE$R&z@wG_fRT6Zx+)>%siiQ=1Ql-DL3JWhuhfy5tG<0))1$1$#|~9n1}2XVc4G
zVkoHY)5BXn>Sa}9J+#lCDgZT^XI1JH!w(YD)kpQs64Z-B&r}e#wL9u?l`>+J@@x(T
zh3*GPS=$(Xktj<A{kx{Sgn<Xg=a#jDPM*>s1Se=P!=R6BOC1KgwHLt*(9Kr3XO9gc
z-sl2(ec2?PsidrI?};BCUkkb&C0u0d+*{$utN-2e0f55UJqa3Ow^3uq$}%7uG0kEN
zcbLjtX}CC?t#)VdUgsxfSfGF2%+k`iU5%21+a`DE$PNNRdtj3rB6fFn-ey?;z!5);
zWhYKNfh3Sog+~S?5nmg$iF;$=uf^*kb&yZ~ogU~DQhH;~vzhwm%E#^UymbtwyeACF
zRaGuYZ?OWBX`?|RJvm_c;&6pW>e5JMw=Cu(IP=6XALOiG9gwP=-Q39O(~evzpmlSF
zSxZ+E``wpoza+wGL>H1k2nF)ir5!uO3Wb80-SlvEd?on4G<i4oaJEa0%1SxGac9eJ
z<a@@Q4;(O~<}bpyP%u7=@opXfG`fA7!09}QIccgyi~?k~ETFDhs(kf%(XHN`_@K_m
zpZV3sYy9dY5G>MdqnO#Ce-C8p+>WT%?w+0=?jAvrkVN>c+hmVphO2HHs7!%%O-o>f
zT^YleSq8dSU`3D({~FN%gidEV`8p4f4OeV@e6}40vB#?s)Ul&vtM`-2h`vm|6bC;^
z;$_09PUOEn=0kdL{AhGBaDb{4{P#F08_eYD>-cXDOAYL<qFfdS0n?yJ!+<2Ah5&j(
z_%0JPvBW%a?UNKDCFr96bv-?DGwSc3Kf7c}e{u9;H1wG6MGUa!MVoPDun<e+wEyJ`
z`Sd=zS6uGO(a_y~LeLNgW&JBfy18_1PWzH{eJ)&exHC;SrY3;AXA2C?n1}11wzjs`
z0JA(HqQ8YpXPUhUq!XmmK@dGH_Q!{c6E#@R#h^|m+}_9NdIy1}>0S7&fjc30T$=lV
z-Y@3Y*D2E_f3^kl*7K+6QSxNbktee!QyEtL)CjftIQqTRLRXt8k8pW#wCj}-pNfmS
zWm)gvySJ8vm`8P-LbS}IR<)siZNCVR@?Pp7uYj1;#EXTr({9w%Jf`(cg%1XFNbgc&
zveFbKRGVX`<0K?N!;3k~Ct+S`@^DQ+IV0{dQs8)*9}P5b5`AN>`aN@a=Z8|edwUfu
zXrSxj`VV$QRm6)*%1bL>+*;{V?wzj<q~0apbgq3>zZwe01);iG;IUx8l+_$qP4L8(
zr#<glEH3UFO0K#}@I9zkY<}@7UE4zMd&|I1e&T*<shf?sbMQP=5xD!{X80K-h%6m4
zBHoCTlX+K&O^6>qeng$!ANiL)Ebp6Bpvzh@?2SC%SaqLZSu-sA=?*FhB6b3hc;K1F
z`zNIx`7O*a=f<@qvi+!igpu@Gi2vS5bn(#Evu5OF_I&DsK|Kr>R@s4FhUBHEg11(c
zB-QT_e_$ZU$1rNiU$^%p@xABPNcJyCn)RJ(K-oBV<ZL&h$2)_<#ecZAXaYfyU3c?l
zo&8_nVgFgjs!nHwVVsPG8+T0GVV8D8Q{q_M?2iD)X>vw>etBD>5@mt_jJNb>;53OZ
z@0x~qmwIaI_?C)dC7G@#oNbh3;vHw_*(M(Yu`}DQ%6e&GT?GN+oRuW9=HE}7_*_bp
z8b0c+R}0!H?Qzx>W1=g7o;9YXYCDo6l6OrD@_pcypp}4Hp%1a&CO$uA{G68oM*$Ad
z9Wl_vl7^W+O%Pgm`}f7ae!Z$o+<dJ|1X8!<g-avst0MjEJi<qGk+ajs{>5rv_{p%)
zLMgZDnG%xuUI8~fU<8S+Q8h2#R_DB9ZJm@)g(p}stAOXjLco-Ma+O;d;U5qH+C*;O
zzI{`Gi6sfT_&Q_M9DH^GH~<y!)&IX&+pd4<{L>39-nW+zQWk$&O1HLA$XuHg+lNus
z-W@)Kwmq0af~y$-emqbgzz_6p5~S=z25Uj^2SK#}a#z?2TY<P2a<fVtd;5Gh{cF<x
zlq%R)_Grm5(|7ukb!ak*@g_{nyr-Bp!VxoLcSTPouM91+NH@69Zv2(M;-8b7E0Rlf
z8?PNT1(A>#_7JdcYnw&7tS3*N)EYYQs75RVv7#@DbxRkSb37tq%}@#IR1W&+<ykYy
zQ<mN;Y*UKe$S%GgP!3y5EG}?ZSQz81P?mj3&Y(kgvmVn-1xk(DN%j<5zW^`<;AUM`
zR@SE@2M#5pIu1N)R(^YbymYTq>?MLx0_=ZpWNE8fhW~Mf@;lhu<lXx}?UP|o8nh|r
z94C`c!%md`GKSZ;ynLe(RI*L&bC3|Lz92Dts&`?C9e8={ypfLQ<(ruS&RBP`>}Ia+
z8D7Sj@)92)7?(Z#@Ur(Ke0CxYetgk)h?aKF`kA#3993QPpC(OptH&@kh7e2u0m$um
zN@{8Y#y4Axq3eIvOMd@O7xHcY82$3|UH*#YtHpwO?}gQBUZfEjHsCKww(w<qAL(py
z-RI9UW(D5;QD(aUQO8PK-Us|hnQilafM0%dGC_40`uh7F()E(JBPuH^K{a6;A|Sm!
z3cVCKnz;o#Z9T|daxErYpcFl&t`ToHAsFpAb(S5`e-wNd$G=ifZEJ1a%>Mp!<M5_?
z)ZgXOC76*k7;m`|9gG~22jT7l5ikJ<^;HnZUu%6;;Q{_8!&hKvh=oqh@CuXuZ1Ue_
zXDtNdrG~2Of0HKLV5juz_ELI#+bP{04>f`EEgUC&D{$>9X^8-|F2=_e^awF`Z0v&J
zrh&(?%}7x54>J@MQ|w2w;M~#ATFp!{72~FK?@KAp(wk(3Biw9^D(IFdRVrX?<^4&F
z0zjr;zXaseslMD_Q7&PNV(_1C6vAewCe48G@4{~g%@`uFz=iu)P(Z-q>eZ`XRS;bo
zfCEMPnXYkF109leNcg%mG$|f_4hbL`aqTEos>8I5`Z_CZ`)=GtQx@xL=%e-kHGir8
z?G|`LydfSc=+AxoOyl&D|0R%t^iU^P`*j&ww&i)D8?!6P%dzm>^v7VwXF)EVj}>P>
z!f?>qi+w8>&;IYG7qrxF>gk1ZY%`pBw3vXRBm3#BRF=^z&Q4B3s`G9jcHv!lK);-t
z$bpPc4CgCue}*oU#1lCy4xWIQXmqqdX@HK?qLpv<Ql-ye(dfeCPyf2B;(Ipy8*=?c
zGp+1!xRNmrO)2;#H@#N?IBCN74F4oghyrBoE`n%|5G89vgpHLdgh`Is?MamHAq`Ds
z#JbqwZB3v!n_A2-2CE!Hw@M&-w%8cC<%F>G&$-!^V%$lbBcXiyw&`wr(SZJgA-aNK
z1{h{Q2YZ1$0krPzEK>(Q43AARnc{(wpqHN1w}z^^y1IDB*i%$Ox;F2dPJAx{&Ufky
zMwwMNO=+Ha4={lJm(%CQ=^lfWTxBW3g&n<)sCVd!em(*K@L;(?3w;X*_bR%LEtbQ4
z;MU6(p4!cy!f(C9nVa8BOH0$faL@63gvzg<GDUdudfDfV<qE6@0<uLx1={Dym@lYw
z-;{9|3TgQ7Ozd8{CSS!dCBI<PZ31v6l-+J%87{gTVqF`UAj<}F{F)}>Jc~Q18VTo9
zl%ccd&bc&&9%f%cfesmo=E`L|VtY?cA&@s0)7u0`@I){0J>VF{pAUhnSZ%il6P%ZT
z1+FO6{hI{p*}3W1iM!ZppQZJ4a26E@`qEdx{fDBZ4s?K`9M``_1aD*PY<D=z=t|;>
zE=z__QUtv6w8m8s^40>ivrY8WcM|;rLWxYOqmPJ?)6)q4H@PG$nDvGU+($1tMqua_
zs=L$CKFcE)lfnA}MTac23keD305=nmQ*yq18F}F8$)~cPM@Yu=V7kMu%=fsr&Z9Eo
z+XzrM5!uzq<5NR)jK119ozg|#m8`cz!CC7oK=nKtuqYq!H^E_Xaj+QA4L-R*dyqMW
z0(m1hK|kIBB9Q%C+uQY4ccvomR^>JKf^k-9<qi8wtk*)2j7WR`cp1%k@`ky89xb0=
zwCL5(kxwYtKME7HSCYS%y_cU#d;{_g6rC7M2tztz^tH8LGOEU+q9R_r?3+JP5c$o|
z0%hMb6M%DzVsP^|$4y``Y<t|9P`!D(I@|6oftO3-*bW8^GGK((?agkzVbDzgj4g4p
zpdvjq%Gv=MIB@+p>%i>f9f?OTuOk6$S-9rBXl0r$vgBBzb>SzNRR)NG6F)~t7BJlX
z({c$@l}4Du&n26G<R0sc!jw{BK(&`JuqGGC@7PPDqKw+5kO*o;%>GEwqdOr^+!e2*
z8QKq-$DuBCEk{Ae{&%%tEJe^`rtL`}%Jn|wHhVuzOq`v-9S#y%A%EqR9pKZubUXjc
z`YXa6a1#oI7ddVU5@nW{^m$W2<9k<kHy>WSC<^ke`t@D3Pqa9Xe&`>S5(%380{lZN
z!(c|c54~QR26_y2F8_f+=Le9Q6o(6LKd#yA(&_8znT`e$5ag!Bo>BvKI-dWa4Kz2M
z@~qH63R#C$`keH<=D)t!Go`Wg1f18#YR|*%2%T|^5LNQ+`wAe&0!c@<cIV>?3cUla
z5EBy4TI;j))*A?hb+~4x)9jJ6Nmp-18#)H%ITjVCg{Kj*$g4XCq1b{x7@Tp+LL;|?
z^iKcrWG2@A1Hd=Qa5mssAt2B|e$`dg)wC{~Bo^QySzlkz`S|hc@z>v;?Gs|&X&Ce!
zdJj72X5PLM43uV^E3KpnQZWgLi!Bt;sneG$;Wvg$*rn6)4DO6JtBF|;flFw(gg}5T
zd@r{`!@$ud|A<lkf!dE3{LnkC>#d7~cL9w>hyL{jW_}!J{zEb^{=v|P#~l;3IOs8`
zjNIH(1~ohtfpKN9rg~%G=3%%+gmaQE`x_yF1OMy?9RyR_+{L~HjzhziPSQU*iHw(W
z1WtW`fTef;;FmFyyrp($mJC)&0G?pTpqSXS1Cs!3RzBA89O2H0D?O+$h8Wt%<0Wbl
zTy?Tn_2HMPED=xBC*q7kx4`4e7Rg!8e%yS^AThz(vBi<r|M6%Wl`x|+UT9tIvn#n^
zcV+<r!qX`q<ascn=GWJI-Jf?|i76C01G4bk<oW7+ug^*I_mAo#pZp_5h#tV08?Z!(
zul<T^VWXFB>IuW5;82q`^)~$LqXAJ-*Oo%zv<5%{6l<=;g3}x9*O88_byaBgj)yVP
zj8cghmMR8YtzrfkIJ*$AzyoeYH@sH!!S*2$mBrz4df5up@$M+&scL=8JH{{KbtT2f
z*qEW7&r<jvkf!qPC!;b7Q}Z|{>>Q#+xf#FokZ0isbx|tZ&_xE)t;yHNgf<$(XH*MT
z)D1Iu>j;V%c;%?@V$DN$E7RD&z6EaIG<Zg^?zsA(@x~iHm$i+fvYh9#-p+4_4o@y0
z)^n3z^3i(|Fx2gHqoSGKAisZ@aN|kC?P}}p(l{T3H(IYk#G|6L3_fXj@cG`f`j!OE
zQmX5&N*bV<kt@f>TdB3LOcU7kp|j%w<o6p%D}kCkzmd^TPbzl!4|H@)3?oiUN!c1A
z4*Ky<ya&Wf4KJieOjVMh<nzbBfYf;qXf|^wk*ABW*F$f}aky`09^;V1dFMKPkjRM8
zq@70D#k9CvMGeon11@8O;9xXcFxo8i8l<s*N&)4_>UjTtXadUpDz~8dz6auUKb{qC
zZ=7RYd@EdhhAo&4Qvd#TlW@hnl!p7Y>2_ro?d&UTW3^sVQ1~89Z;7B#TJ-RD0?5l-
zxlk6JlkwSdJQbcZ;cm=CE+SONhyK+9eY0>`cIZDCS9RbKm<&<O8uUd%?|&E0^meEG
z;vi<nqgB~bv0~6o_o}?f<e=H>F^06-N*^b{L9z{o>$2dA*}h^kA$<{Ef}BVSIIz*s
z4&tDoV6(DNWQDUvo199}f)r&ZdqZtGN|K0v0r5ORs1JDI%kO=rHGbj`f9V<#_Huuy
zvZtRsO>B3($~#-MCG{JQ2c^#Q<bEU+kNn6&L=hpIvyeFU{xxE7_%Ha+W6j~%kmjY4
zC1>Ts!8Sn&RwvlLiaovtj|?5y9@n#6P)hw8A5GY!&uxtZhQ?*=%U~isbLD@Z`AFl%
zja6g{{~FjbPz(f-oy%9h%(sRkuX}qWgpm5owQs*3b#;s8!17#y*e-wRA}9(sby0^i
zopZ}Rjkswr%exj8EHEHaFJT&H1cV^psQ=yGNYGdU@M*1t2XH=tVMJ=<LUrWduk=pV
zjd*(u1XsP@RuQwv`NnVKCk;N`KjKahD!3z{q#EY0Zga?7*u~w%{8+BIk=`TU;o*|^
z^y$;umE<3oa5$9y|M&SauNFX;FAwcAGcmDewEnvwJW7!Vt1CiIT_>FL4il-scaD()
zZ%Y2hbDaZHe=%u4=l($iXY7V~zg+=aXy3k_lQ<99i$xI2${0Iu%YrCA33PVGyJ4ff
z&IC#7zFU3D@%doTTBBfLIZ}=}<yWR0-1er|<c|hw%+c9c>VFT=an!-Bv|KsU{57qm
zLP9aG=$A#(8cv<-V{u))7)^4Qv5%?di_B)<sa-+v98vYWJsd>S&dyZLC9nf&TcZ(>
zfMJ#VDIMi@ADgIuS2JtTsV0zm$3bw>XLj9OG^vF4+Trt~bf<OOTYEU($D0^Kndc+0
zBxd(zK79&-?#6&mRp&WVAB~V>na%*CCq92Z0Ut7@cF70Ly_e2&D@w}SHxXiKYAe!w
zhV3SJkniK^uc?MK-0+1la-w-?7mQwcevz^Epr&{kaA3XUe{XWI#3=U!a2Q*!wE_md
z>UZ3QbmaLrCPO5j#)PJ>)74`>rnyis8j0n}5&cXcQ{g_Y!w&GTuy=>&72__X!ABTf
z7QSN-n}y-YoroOW782t~v)w}@(Fg{UZ<|1)kOQt_t>mv48X`s>&|oQ+h_buCQWB;~
z)@xDdcnb*iAtKe@wdx1lqaL<R`1!E!GndeBO4~KtH}qm{(2CoQ2W_WAAaB+b%XD4*
zcN<VT>Qaz=8{jsUwwH!YEG(S<d%nLEaMy6ng&yc>62{kQi)R==_av!mD<)=F8xf4w
z0<kF~k}4)9JpJx2yj$^9u`;XPwW|NFd>EDK`?P!KXrb+MNrWzNX^>8UpRlR68)^Cq
zZPsUrh_`scG0~<pzIht@EYz50{{*T_yq6`_L+y-zYO-{vd?DBO4=bAOIQS)G)w{m(
zIUW(Rq4D+@r}s-rv9CTx4Jfu&`>yO2UfgV(d>=B0i~M(bjw4XVT&~o`;qdnunD~`O
zXqdspCWSUO?-QrnM62S%2EruJ-yQ`2yG8Z>^6PHK4Q3Q_cXGk}@3Z3@qMF#lpJ2H^
z;&26>K`4L`Y*~EA*n&GEI6k4-VvUdQ?_j}(8iI~t!1<RnH8oLcMO}=G#Y_@SmAcew
z{rmP?lkC#2Ix+J+Sw7jPiDvUj-h`bNM1)Pv9SCZ-Gs8~iZ-+j3*F`u@YYrFLDE2&t
zkZU*Cy2gW=Jb)x9C8SkV4T)|ZHN5xf>8mXT9^`QrR6Wa^7h<tsNJe-NSb}C^U~t>E
zDYWo_n1-3P^=Oc7M5-(EAp^zljqa_B-`b?YQ9E5(uXOHmon+j(NrZ5$${~v8mgo!!
zV^dM|=hN`&WCPx6e+%XT<DpM>%K5F@FdSKo?h{iXb5m{+<>?wD-DI`gX&g^cM$;qI
z#QRVDHa!^ziUfLKc5ZIiQ&bB7#Iul56Yj&##m<@onx^d!B^;`~pkNy(1kaQ4EV|&3
zf+-C*@yV9(=Wk26S1){-T>AA;8iTVh?Z**D@!Q48gSA1FFWTXZkWmwL$Z%Kkbvz*E
zf~JJW3&5K7F~B_7^8v-ySff@?hd+xL>8N#Ef{ls$gwTO09=L<0&fw-vUQ7As@yioi
zoZIz(y77UYg_fLG5dSh-h29>aq3w09&)GuQzl$TlfA)$b&8;*l{uBIRMbz^hj=mm;
zH(3Ux{=s;TVDu#<u-pdyfx*)DLWZPu;>O!)K0;IYYa2QL&}P8($Xq<D+7gWBI^)<<
z31%rtPF2WiMjREfl)|pMIOSy&GWE+0IV_(%6}eLRFEerV!#=(v4^Va7Zyg44x>m~r
zzXNV**vo7G9)N%Y#D-zsVsz}YLj_adF{$(T<7mLlUIrH1^_jw%d(U4TAJ5Lo*`zf^
zA~%ozhn(044gQ$szro4#<Poz=M5O!V*pah#FJfY2_b3YUTaU8fi71Xl3)OEv))#G?
zlSHwV&U0TRMgB!uR2JMnv^>D~pX9)}uC?=48o#yuSr2Xa@CLa_*lnOymUJIv_n1wP
z)+P=5T>5A9m_2}rd>hmdJ7a_%=vb!Y2XrBNPE_%Jysv2b`~}Ajw3|i~>5<pfeP{xt
ze&_<viazR5PD8a=BxE3kcrm%^A!KZ@z^R-ldvs$Ay3{Ws3$ir=ZHAAa+Wr@>uP{I#
zZpj{*jXyB9MfxPCmU(EA42E7U3X*o7GW00ZR7TDoV=9MxvrUH?Lm)h7ewDs``*!@=
zQRLw|^`d$L(>*yn)i2pkKkS=djNKI;l(3v_-9mpp<*@(}&^-R45f^$cG|CLRd0r!$
zD>N*|8$yc}fn9puQOla(N|4y^wgndhA5yAl1rN(;lLH_{l5z?znEXxjxWK5HYWC){
zedc>kB{8zj0<8iQwaX~+rWtJ2HjU05#p4hKeGk{B9z#6i7G2H`4^;ggbxxCoBiN>}
z45?3@sJ(j{T7o9=B`+*5t$3eSO|LW|Hn@sU(47T)I2f-8xE*8Lbl=Sv$5;bAzMZ$U
z!u)7|DgCPD<g>bUFdT3@ah%#D%F2CoGcg4mkq(teYV_(kjr)K{eb07MMrPSq&;7Fy
zIuhZ>wz{#im!=c<MU#kTQ$$~b?x2@(Z=etSb$9gr$2@-gI+hmGXQuew=a%xl8CM)1
zJNWjFX7u#Z0I6)a%<20DN~#AZ?Jp$FW4eMl+szOq#$nnxpofO5u4kas3-H2fuX_$$
z;=R6Z5i>G^3ZjNm1Hcx`8VW|cAvGs;a{NKfc_4^`6>JRq7=^XqgqY6X0|dzWKX5ZB
zh!oTWJyqd(z}fIj=vzP}6m3;KV;)B(c}#i=8(9rya`T`LfXK1iMTuB|<9{eq7EWac
zZNYiNM_$>^(j3m&PK(+T4#D#45w(UOQx_Syqaup!+loL>{NBXt!Ihf>(?{ht#CA3J
z+AEix8|5+(=ukNpJ0mnk*cif1?Y$Pl^89AD6J9$&mUlYBrE+YB!*=cV{K?hxDu&n-
z$VJNhJ7X>8f2WqJ{xX_|bFYEn#sTJSR@=YT)1*;?^nQJ9Wl#y$(DxR$fkk)_gr{Z7
z4Q6m>$d7H0KIIDBUH?;dr;!805uVirv|0u0h6j3c@}O5VV7YSUeoK_Z*d&RBk&T4J
z={y|ZPqdXvBy_U;)5!y<kX~-$W6I&D+~nh*T#q~z<xmZ~YfSfBSwB_nKJzSs=H<(|
z`wRfQ%W5WYe)*WkGr6#`KzC0E^r7*U$D?EAZpkh~B`wwPh6T?P`lD}z_30KoRXECB
z3p%aESWIm?5C~XGA0Ws}xLyoqG*TRK|J59NcnYVfea1aW?x>s}M+#d!X#8Fps~M)0
z+6-hlYxA6mJ<d)RWaj<+aH6Gt&sCnEu!^^svA_@<u7uFG(%q^|h!)kJdQS4{C?eV{
zk4(o^26WT?zTO0ZR$S@$3AHJVqT=b-;&ia*voQZ8dvFKIZeNRH%mibUrGdM+OT5<W
zI_nDe`s!=l!r^nDji5?I*5l|CH`Y<m6>Jah<aIX)&uxfAL#u4qnnnveSQ95Rbyl6{
zBhxkE6XJ!g=lb~03a09Yv;;=3LAIxzb3cFjDutDQY65^I096kZcV!ElbIV=UCOWt6
zrk;ei)zR|N&oT?t_uwJt{M}jtFP}Qk-un~-{*V2@tDR5=eD*6S3>XlUeB{~unc}g`
z<HwFkpFZ8i>QI@wGx0^%tf0<xW@_%hvZK@__8C{_x2=P3Sa~Wz5-0^@EVILOc<vlW
z=<D2Jc~QlAQ^}v-v3aPlq;}9mA+)r=U9_w!>OndeIP>7(KY4eA(!bnC?h3v`iJKA}
zJ}U`pa+n-3+!AFvW%u1ELJpYWDP_6<Dixm9IosFO^?Tn$#-3iHVw<ExLd{32CGWZL
zzCIJ}GEcUXm$befBVJfxl2>O|r~{p*LB>&ve0iAfBOtd7fCEBRXc|WhO0ivr>rd)_
z7FVzHNLemA#G(ETAI+*TAX;s<x*8_C)RUV})tdELI?SjA3yz$lOO<rHz6R*~_Kp!;
zdO5^lTJ%?0Jk15bA;z@O<`K5w)!CG{db}FzrC^kw+vf9O__2STc2fl$T=*supE7+&
z{>~FUmSCNVeLTCU`fmt?`sJ@;1SV<g=ley9LIh>~^VzMPgF6Uv&!Pi3qqx6v8|uDq
z?&P>uM9TB{LW}-8xIMlY)b`d&r_SXef`Ma>J(uw+AHdZHv;wRgXDhhZDJxO~GZ%)I
zejxh+tb7YWJe2^DsQq_=D1zPi=ANu&*-x^)7))1Yx)4<;*J`bW(I;RboBfLVEu^gu
zAtEpF9v{yEaoFgz`OLk2DGv_kS>D(OVtxOB-7-j57GF;2zMBoZm}>k)xu{}em$M;#
z5R{58Q%}WAEUo_b0_Y3gvoo|j*jFTb*yxrm;cqDXdA|^^<vO?44u-n=U%*D2Qq3l=
z_!IMFC$IErHE~OESlr6FFK4Zt9ZAyN69GjEGy0RT(Rkw9x6~JEr&k9zy=_b`q7JfV
z9SlVpFky8A_vX?WGe7cKQsT^cp43J|BRVU!lsqM-C#!$top_0d3fd$nhE`H_$>BJs
z=Xp3UTZEg%fu)aUQ3n}C=Uz5E<7!<@dRNlbTBEG>Xh%^-Bk~v}{j3*gPcRM!kQ2Ke
zQ7L+pU|(anML0O9gd$GEl}ubavlL0h3tPpCt<GrVl=d4a5>XNR-*(Y5VYQyM*8k-j
zb8w@IkD<xDWQ?ITdidiBumO}U$N7i@e19{L5>g{OobkiIx_2xq`rjLW{_$pYQAW>a
zqkU`nE6Ppf;w)1b-k>au5j-BSqPlWWn!N}e7<Bn+Z}MwB@x9w$M2!he1<a<~V@O-Y
z>yZ3t4=s-FEOuel(4#$;3g0$7<!Jcm9Wmqgfy=~LBJgPZTwy)=wc<o|Qv`@K3x4xp
z_c56+8|e}{9ole2`Q&BN#lTn2OgkfrxJ&krt)#Gt76?Ypi5#1U`THp(r|)KoH%Dze
z4#Vc@58Hqxm^OgX7<`hV$MfVldCzdv9H8TT>bX7k=Y%Y^pv;}f&F9)s@Si8_;x85K
zZ9-APY-CuoCcIL&@|k?*lMHO-*P}#Mu4cw8OyCkzP6%FoAr6fR9W+KG9hvJ`#luY?
zE<^&ysvuw7utOgI#E;QoH4~%%MXs(3+wIb0Vn4EiR^aI@7Si|TPD;DR(~;x@25ux;
z89_$ur26c|+Z@__>>@9_KVA*FPutrn^P0OUe9k^}Wn+2ao$n{u2wsDD&PZB7=*Mi-
zxq7{I@3!$5zRW0r=+hZy>q1xSz|`FuA@q)+-Lgn9dVs0<)Kq>UU=;H!J}8_CHzD!}
z<Q+cyi{4n#xSbw=D}4Q@*C*XN*Nzb$wD^Bm^LqT=lc1QMOD=0x34o_gA2pz9u4|_}
z-^BG2)cxFhlW)1&jUJ-N<B}fti~G|Q9zKS?8$mB#6dMU&zC+dAN`6ndfP^z@5Eb}~
z=psk+iMO{v(+{H$Du5~SXMyJTBPKkxM{}k?fy5wI{A76!-YE%bG?4pB6YZT{e0$r(
z-0U}}26K|jr+=sLCS?ER)`Kp%^gqd}+PqAWmyaw*Lhj50{s047a8@Xib<sVBJa`rP
z;|XGq#Qbb?y0eV48AwWPH~pGW7F~QHDCXhD_~qIUkUl4N@8$6(7&!d=c+o8_@Z;~C
zp9NFgCxy~gp<qAI8aQ?)vJ5(i;4!f(uFt=07lR%l3D~e{-EX8d7{ymrRhji*nUj`2
z{C7>t#lKCTLT!puv+q$o#R&XwwBVIWSfS!76{m;)dBS+#cK(zW9gzozX5U^ERZl{*
zktH99yB(t5e)BAWkNp7ez3X(X)iYmBj<8)oYTQ5+7?J6h+0?miXEocs__ReQm-R9q
zha?~lmNC{Rkn#05khp@862oB~ki@i`_T)dF<^N;px}&N7|Nr}57uViV#KjHK5LtzL
z?W@R+WK}4mtd@*>?Gf1(G800!jLfo23Xu^qlD)~e_x*d_&-eF7=lG*@?)$#)>-8Lu
z=i~8sy2S;Yy4BYXmSoRHsO`k%>#KcY)L04^Y6d-<g`aMSFZAC<h<rLiYXyqK3L=iI
zd;Oxl)ex4X_)$kv0p(8hUo$)2b%FM`P#W1AOSpCAf*7&TyaZIiov-KNj)9X4c5&S}
zEJ+N?;g`@NJbz0O?;<yqkiJcl@sHI=X%`=Tt)|TsxMhDzk{bH=7v%+vQRt7j;$>>`
zq;`W`K}#o%;yVVFIAcFVkHbXyC&hm5dx_@p1uS@wGtnoZukW@zPQMlU5ts95=YbZB
zWR$J%dPs)ZWz3`H5jnqq>f@KmL5D{_iDAFp>UdT~wcoAx-t?VK81m?<)lDH#oDhan
zi7PG<i}XLxYN|t*N4OysI;#~w)aGUTFIbuS4hAl<$6Ao8Ox~5Wa<DaWG=zz_{+ivL
zryNHyGZzu=zzt8xg~RK+F_#>|?&z$@7YKfQ$a(Nd1y}8O8{_wSxtQngd1{_&;T%06
zzS(}RGAKP5`hq3$?vhX{2jW=!d8$WTFRfY^Q72yGrYtB8R~LE-=W{5go^Kc0;|{%W
zjyv$@;C|1WheAAF;Uvlv4GVP@B*6~T7Aq^ec?jsRyO(4*1={J=r8Vw^wBmeOMUcV1
zClGgPm4UFd3HP$D<@->?IW}iu7JnlHwp9DiVNumFCFxa>**A9|{Cf1?-L+dK()(Qe
zx$S!H^8;6hG}>z~+OTAh-8b|D71TOAJp54(D7!6%{u2Ek8NnH?j-cvNt^Mue5ww8g
zF|mYda7oxV2P0D{Eo9sa@CF+U_LG%r9=+s)=37Mlk76$@$+CYj#mxt??xva|)K!tQ
zCoUy}k;fC`y|NC$&D|I+E5cOqF%u&fD#%E_{egJz(J#ej)<-A*gSfluUjEAV`~8*W
z;NsAMP&?(uIA<7=WJ8n}-dC8vY|>N~8}r9s_^7b%>nQ^wPPCV>uX`tdT;vgJ#HBG^
zfe0Lmw9>H^lqu-#3nYIFV$c@PnA8|3w5fVtU|90FYZI)UUy1?1+34Qp-0FwMPbw!w
z@TJN|2zOhsqxDBZ4}e>rzJmTrocOirOnLw!vjIr0w!RxFEDs8d!~ch8GpJ{YLW+Uw
zwpO=bf;1~|wn!lUQR0f~s=d~!kc=z-vvrRC5s{M^MZq8mheA6O9Xx#Mko_~$&$C{!
zpBMG>qv$zL+M*$<nGxGcp~rOl-te6AGeMP+)Z5uW)Y$A58Wh2+Qk%=9UR=ylw9ajF
zC+QlHK8t@reiwN2f;%1-**&;J`%j4N?!0lz0xK6t^noYdfZ=WN`|<ZYK{_AGPS5W4
z+^-6EXvdPHveouuuqIticHZZoMMUBvhJnFF!<UQ>YLjg4h)rgvp)p&p7M#hdM9eIp
zn|-g{u4zaAeM{+QWOs;hOZh6d$}a+`L$9073a{=xhX@cWH$u5j@LPhq4!h@i<K%uW
z`=<udZMsX)4rV1GKMzS-Pukv@<9ab{>$j0B2j&5fR&{TPf$)dAtPMALMV~oSL7Y7@
zZ^R>4whiq1)ZzgU5#!0v>)Fu9%r~Utt|g<b&N)LdJ&Zeb>*t_~?(x4fI4|F0I>loz
zl2w^r1@-kA8+h8Wz9`hBUyXB#Ts)mo0JB|KWful!?~Xkwv*NUNf$f?s@j_o{&gtAD
zPSv4@LPuzVt;7&(?U_8JBj*#<AK%ujU0q8U`L!x3b*g+W%%BRM|0aJ=PxpVpFmW7y
z`sm(B%=N+H1P$@=p(Y-=m;HKEy`8xr>@9ra^=$0ZJMnyBdZsJEuY3a3=Ug22FJci{
zunfv!SlZ>Y-hI!Ld%cHNM&@g^TubpwztueF``)}N+6g`9eFKRH3x=PFe1FWspeY4$
z$<+uxKWn>U<`fH2M*z!Cov(GDCH`ZoWgf&?o=d(U3ygCNAkBBGOt{;6T4@R|D4z{%
zs_F8Ocaj+Yi;WB~k6If$P8|C_W&7!;wl%r;mD<7YFrod)OFwvoc@MhZHX1)T@ja9A
z^6I%H|0jo_R9*}^#f`u^BzN6@Z$0v9B3=rGe~j}TtohvwtXHV^GZDP;Dg_Niy0w2F
zH9yeVV5?0m>4Jx<=%gu8t@K>yB`Gb^?z=0$DC^AuG=LOPTU_=sU%M`*V_^J5&YtRI
zirA4+AS|Y0mFbzGicuQ0h6|?9f5<Jx)vKd-=$$Av8M>OoZa-}euHF6AP1p83oL%ww
zh5MZ`<k5=TG1g7~)`-MduOAeuU&CjEkTRejO+C1w9*QKjW@uJ^z+Pn79G2C|p8fc0
zOK(qM?k13GH^Qm*Y%#P1sGy#z@ii;8vV8iHpc_x@Wc%GRq>a(REdQ(jH&$aHy!1Cl
z`@~1l(=^m6MKDRs9$owJEFc9A`XDlmgL1TwbJ6Rner$V5l)sTo?g^-VHDnrM3Q+@S
zXZj{F;D=IwlTcKO1CLf)nJ#?p;Psp*7@^oWCcDVo5{4xyYQ+Y9w#LR&T<CbB9vQgU
ztV(?~H6=(AS_y}H)xV@Dy2-uiE?u}e)HbbGF;!Gi(wr%CrLSW@h>eEMT>?>eu_bS>
z;k7cu57>7ZCMG_Ap=b9s!W_DL!IYvKl^>OjCe`FVVBv(?pd5KsM6xR9NI&Ybu(!T1
z9io}iHi9tfw&8ip*xdGf_DuU>OMdbWPRM`gI^zF9j#HVfy5(QT>i+n^VK2-aPW?Di
zTcMaF|3~^Y694ddVvY{`lT^cpY$=2VT|RV5zC?nn@p^{^vyOvgMDibhr={d8UcI;s
zxbDH?((vt!Z60f7=;JN{*oI<g@?d@>hPNI<kLVtH6K|vhe>+$_{t+81=Id`I`&irE
zErC1hSU@ucNIDNC)JEtqsC#2IS!^bR(>4CW)ivY7+QUuAYsZs$@wZ)-Dl5ASA<74u
zRI5d&S|Nqg)bx6z*$VEtS5-AKupqo+pwsJh=H0p}tV>~t&!_a8ye%WDrxcR2rVKpU
z<B`||lzxld8AXbyRBqp^%1gwrZrx`BM9Tm}rp$nn?4@s3c#slgB#2^}J8AxvWi`kp
zLDgnFqqw%Q5cyGb_k0fHjG!h5T|7G?iamHhTNpa&4vGzfd$$XmDu)=HAm`_RHhq9(
zDUd*$p!<NHY~?`#<UgQ1S_tvse&vp8P)iqiQXi}?<t4Q-p-=os_raz}FLY2u@IZ%D
z2}bM1WyC<clZ9dldN=sAMRYxNiMTEtwi&3I{W6Fn1?1X-0nDEDd)H(t?<pOXd2IZb
zOTgT#eyfD8O-dk=q-w(fkwwYKf2Wjc-WwjePrR1?-Poa$mcw97dX9)Xk^JyeBL4|k
zJ!R*D>FG+Ov?P~E9Qn+GM}KBZ+bj6TzqXxf`td<chzjB`1M2Md;KlzUcW73^&Ep3;
zZ9+rvqY1yEYD6bqm%#N`R2(fbnU|L#iHWHXMQ4r}WNOMBEnYrve%PN<8r*wqmNxiH
zzX5JJ8qOboC{)tvVk+=Tmi4R0nG+@00;>wE7by^XgRU6KzC{`w-uMW6p$fbyg7DVE
z>~<f@IRFqX|I%<9tKIKs#Ai89w@0UIYW`~9=W~#P9<fW3Z~_(ID63~x%Qvgt9-koP
z%l!A$q{9r_e=+T+BjxsVu4<QHH;4tIl{5s-lUSACY^)2)to-s{==ZiX!4dAsXl}}*
zELi2gcNFV(Dm+a1a;y2@7unv&(>Td0mAGSB(V~^kGx11LFwS%ZgVcmtUo)bI`04LB
zJ8w8ctb_D}Pr-csU=T7zbQTJOwL0v=7;F-1as+#o>L4L*tduQA__WezXnBJ;v=K~h
z32WLomp`RWea_aDL*AduJMsr}XZA`?o{84q7m6poN_ZPwT87%at40tZyv$!pm<e0=
z<J-Tn7Ph5eYNq6wH8Nom|4Aj~{Vpd`=j-uum^v20t*YK_sY{*>L7ll&?Q)sKU}p&B
z)7sfZscAIfA~;3Gop=IIY4VU@76zWv$SCd^yH}ZdG$GhPKbH7Nmm?>bP9rMcYc)I@
zTQtp9Rq%}@&%J0Wg||C>t{NN8MEtMBITCg_adfk4q5>z_)Y{E7{lPbAE2E59<9hqq
zCC*@1nrVf2oMf|UpzfOPF^z3j6C4@}^rq<V2FsdD71p1P?>ygt3p9}|Exh`!oT^Zd
zv0-Ky2|}93%W=+Ff#S)n{_M<$tP7`^UztNL4D|OFpkDqFNZMS(l7$0*Q|kj{gNWi{
zz9nIX;E6{npIW=8-q1DPxh>EjY}YX@6l3cQf1bOT#!=P9Zb-dj9>2jtk=`OfXP%IC
zn>v(&DI6DXL~G<OY7r<&RunLKCps_g-P$7b7I%xB|CoKvQ0v6st~Tc;l^nLjU`3YI
zbJ)FC71WIbVAAk60bU6u?vO4iF~|(5zeL6|xp1!r>}nn`wkB~y!%FnOEvH7*)mp3C
zFH``?ut`WA2&)a9-k6bwDjCq~oI6msRCq!4g#W0|uTuas<huU-8SN5OCUI__;xd=U
zOR{`UZMfJkui~aDls8_s?<697|1s9$5G+gQeh{@q-|pI$p6bkEXjaPq$Jrb^#uYBQ
zp4Vsak+8ug%pTve!a!`?GTKy?jX6)+)OGK!X*zNF`PN9)1;VV7=Uql<$Xz;@yL|CA
z$BuGwzx&;CoO*19g3K|6C{E_w0Icor%%m&qbSWr1Hf6wRJ3Cpb16E~Fj?#x~Md-rz
zt5R1G--^ZliwBy~^3|1;eDCLe)lS;3O+ofJKG<j&6K=HHWOq$^^Akdl3-@I*|7P&I
zHTJ+aMiTeq7WSF&`sxZ_hMO+at|Ws;ycrarhwVDCM_hQD=Q>{R-5~ghFBdXAG1vL#
zB&0UneEXB!X|Cp(N(TD^q(zMMu~`lo&$dV&|B2S~>>C1}q@jG;pC+_7GIb$eNz9u&
z)-9bf3-Y95xzA+1n|-^*ugc-azSTS_rkAOyximk`oZ{TqZJ`i-`Nf%Gs;(Qr2jh6L
zp4V-`GxS4P<*eIjt#VtF@&9g3cBs7(y456I$L4fj`U4Yu_ss4#E;7!>@G^F<V$y9^
zz$7_)F2jTPYE|dBSLXu=(cIh36T&(7Y?qcHa|V}ekN+dbe6p+RsSYc{5SYu>8$i0e
z^xaVWS!JXl>)I*?Rhh@=QBqbU^jOJ5CENJMYOrDqQ)@1~{_~y*g*pohMt}cvWb18E
z=+DgTJ_}+mXg7T~&LZ)A`I0zoKjqN5tuf<SHUv`)YS=Vfqdent!{K1e5z;rwIMZvs
zn7f#VeA^uX9-^&nuj{He1#89fpVKE9k3^90&THxS{qln9zii#h?CwtaF*fE3W+fYv
z&;TR=%AzG>TKthawmef}2t)TNvvsiJ>z@iY8&hYSL<58j*Uzir+wOPat(W{CX$2)}
z-p%-QiLK*I=@g&d)sK$&kU$QdPB>2Hk2?M|Rwq6E+Pt`h*!R4012)_<a!Y3nl!6J#
zkwy(4D%j33mn`Hwcq<VPcKGaxD`ii>M&NR-ChdkZrv@iyaMWMwi^9M!keIYvwBEt3
z!7ryq91R>^VH^XN3I+0!7^LB7TqbS5TGcRvQOAR#I0-|l(?!yK;I9_lb3OQHBH#8|
zRqtOb7tTV<ebgiH_x47w`u=j<=^cZV+t*(;z%SdEetj6)jTSbu-~9FuCmfj*Q|5?j
zH{JBs{Ay4h6MRGjBXD@}#J|B+!Ht{?o%H=pRO=ML+W64^Zd-0X5NduEz3T(;wE^+C
z8<oPrp`FF|<K2G?GNWz8iu~BqfkbmR656?DijY|58*S%C#6ju;Hc%yE35Xr8me?$p
zv;i9n^<CKrqlpi`eUyv@pAR?9|Lb-AteP*UZNd|J)-KCi586uIH!sFgWOrg1X<0k)
zE#coX@QQ>!2E}~kZ<{IrEO1NFnlg|fR9)D5_ef|icP7s*!;dfKMGm73k#{b|%V3rW
zI*jzS`9&qxu5+tAT!ex@iluqvF*SR4bnR?Ui_(d`uq{m<yqmlv<cr){ome$P=y3!3
z)_u0nTorFyw7Of%gH~0Dfy5Z{ofNz3!Li($-&z3A4~Rb;jV>A#JVQOSs`Md_uce%e
z8grE8=W(wi(c0sFyVeJjL2Px@8>o8Kn?XANOW&e?UzmWrvN+n`BBm~oZt=xPV2+~G
z%0Da&y6~+eVu4|lXVjFk#xaj~9(K*zo=Jnj5k7&MalMx|;IUMZSlt!r25f$2;Q3UW
zyOyhlQ(5RtEz%Amnv;8<f6jw_`a2`E*XaIR-GEc)olG`ftLVb}@~^BmP5ErL)O6sN
z0Dov3Cc8O&dTBv4;3J0}z+kHWC+N;Lg(>FP-}uVdE#R~M6L60Gsxs!XMhw=T+Ve*g
z6&$XMjkXY;VY5GY+sGHQm||RpimRw8SNx*Ui5hH=l~-j_X=vHQH+O$A3Lbr>(3xs=
z%tb=E9>>o~I)WJN2LIhV2)|>Oh{#_zq{X|cA%iuquouP9^-?7yLjpNmD7=-0sY18v
z2a1C^v($fZk)Zw+Gc(hhL?v&nqINA){=Xbl&M8}F(ChaYi?dI@>MrQ6y<HITY9l`Z
z3~|-H)wd%h<9`B#Awtyrxv!3W&6a@!Lc=&wxJE4dm0@!)qskzvl3mbA^sI(Vr}wmq
za>2=LySMVoGpe(Y#GAl#l94BO9MFi)(*sZ}O2C*F9aFG1Q<p_cQtyrAky1V<l)IqW
zAawTbo%itS!1z~rD2;|%zbG4T3z`4O3)Qa-jqm=s|LJe)8#q=~bD#5vn}qVEwv6t@
zS4v)By#6DjImte%(e;_M*kw*4zo9W561t|$<FJpbco4m&P;|ZeYvXRis`q9^e!*xm
z19uZmP?hwBv(by@vEdrIA-#lMU4P`qD3@f()9oY>1LlTb>rVbG@p#Ba2V7`F#^WI?
z1FOq?hS{x*ttJEiRyyZ*+V?)~Jnhf0{P4dP?XQaEuR!bTR?dK`_Sp4a0Ri?99S7x^
zADCFbef0eZotTlZ;)xdQ_Y7d;t8o9*aM~P-TiC<ME~UZ7vy)d+S$VwfflD)wiWQ9&
z(^iTqRXGmkyB2}^1%Ir<u=-XBPWmZ9k7#zED?_JN{gjabkyiY2Jd~p^^UDo3<t$Mc
zx}Ozp<Ms2ps&99K$Ds}Mr`v~*Yu;l7UtdL(ZgPESe7I6-?bLAhgorFHs06-!#x%b<
z?3pdDuBz-SL}cecA8*Et3Ej3pLy}Bb;Lga~c;@YE+oa|jpwn)=q=mm!uDflVM6t)R
zM-90E`@<N;r)>8<@iMIEG<$I|9%d=|&tIz_Rr1Vv(L2Y8z~rlB`tY+1(eIwOU!>|^
zTp)T?gg_<w&gzjy)=J$QJIan<_zcoL^{+7>yL!|7=FgJiw4t#7uKRfAvdxF2596@V
z`CevWwNp)DqyQ*zCWA5X2^^9hX!Imr_fo87!eKv-U`)?|Uv1L+9G|2Z{Qc<`r)YGC
zwLn!bKF$YQRvDyxuvD=#U}E4cgnp_!c&zhT=EF>K0N4e@k(`Wz<_u<%za4LpA{{&d
zMZ$mME;J@N2=(bKX@>QtkW7Y;4f~S~WM<k7b*!`OQ<Rq9o^6J>fB0%OCgle4VA%1)
zkB38%j5*Gn01Ep^^L@n!e!BFwJz^&e%n64=IJlrwTLk)f8qvMHvwHD&0UYC5U;vuL
z5?n)&DI0+jG82G`FJPTx5Xlr53RO&iK}q#^7*3IMIh^1ym*Mw6GsCTRN|)W<%uwc9
zH#bD^Fggdqp<piKq>?>{SU`SpO$T8Hvnh<g92%nRHTFcYLhLz}2s}~%aE@2!fBo<=
z%7I&piRbUdH9Xo&B>!2uJ@}{5SmNw|mDX1Mo$&|!JCo&%2ao!P9#tVVStL=X4oz95
zgKNPcSk?D;Bb0y*#&9{w#rFdQbIhC^#BHY1-K9!I-H$zAc{|p$eR-85o-?p(F5TxV
zRm@F<@h8e-jlNtBBjJbbPr4#VooK-0{M#e<3B{avE!MB7cWEx_300C}4(#emn_knQ
za{~fr8yHD02u>(Foa}e_Z#`~xUijgO=fC!6JuZ43)FtsAN$%GBFVVIrzUe&uwTDMe
zeqh_+J0trWSIJWp6v~u=nxPd)c!+?=Y>jgvX`&Q}GZcb;oY!jzlg!;mBRTc>Jf`LL
zh>jHlG{nN7Xhz_KfhpNO$rF=AZ`$khXi_C%E9dku10xj9zk)jnp%!Ob*;omp;i{IX
zI}riC8oxSP1li7!`q;EczW%7iEhGuMzxIiZ{NU5b(d{}7=8Hv@-R-6W!)+H8k`9@=
zEOAoA8p_1qTkP?qht(%v@E$q%PF&;{Z@rn*-h1o!M5R*#Rm|Ye2-%u6znu{>Y<ts@
zh+P2+(o*pZ-##wHmU-bq#u<^YR+4`&U?B|yW5@<Z%q3m-8b^Ya6WiWkb)y4ypO5%)
zx+l+73UWbp;_X!lR+~OHmeto~O(I&23y)$(FsApL7zV?3ga)0_W5ANk0Ofwy16O*)
zJwsxTS4KxoLqLsFBeVRr<~(h4%nj6JYeCD-=L-6z&`o4!o7sO1;?Tn;p3SBy&I6-O
z*VJpWUx?=x_THMdyHoxn$sXVW0+9MAC|Fn2(=9jA-MLAncw*osYtiSIBBbY@WwQ3I
zg7#)@*N3~s8p9GN^qNy`2X5suczkjLFAm?Ze*H0`C~5tUd9jsl5+e!AlFcF#VkdBd
zM7Wk!QUI-}iFw2hWT7WXe0b}R$4K*T1Ft87?=K%<;g-XreHo%fzv4Kc6(aY0ZS%qX
z1(@Fa%JA~e)sp)>C=(_U=00C;Vhtt#2g|MHN2!kcS{0?3{h5pfzIZ$V`chtGvR_hw
z_y5pta!zDkJmS`$^8uo|5wR-}n^%2@pwE|!p||$8fZ4>aJkGzp4>SVHz#cu@z1uP<
zM|+x+2i5nI492D2(G#w%E9zeF)AkHSiYt0DsT0*6@ZMxF>jERvdR((hNrNL;a<{FS
zTrB(f`dNr18Emlz{3j_&KK=u(>>j6@mV3ocRk4+%m%Vr;j9I@Iz>J`V=K`4aVa2;2
z+U5fzF?lNEC(<J8c7^tYzQQlzvcF;H)EJQ)CL&j|10n{fXs}2%SZ@RNmifENj)nx_
zdIj|L^{IBdCD5c?i;W?2YiGQd8LcZR9j00QY!o-dQc`hEKN!Nl<J1<~xS&Tv__TjS
z4|2)_{AJ+<9~xT7`M4>>{;24n6;VW+AEfRQdo$pWru*$A7l)~cQ3AAkfr|nba$gwr
zo2@+F=-*Wkd5Jtk3lg^7o(pBjXu0FiE%D;Y`)Y4F{6LNOX08xVrOVJ6TMp)@_40n7
zFMBI8P$!{>MGs$*;sn7t_t9jFF-0^1mBhg$Jcom873hPYr5<v|Q8cH(6)FRT{7*CU
zNW>@Qg0n-i46Lue^ZC7$7xz=e%~WUo(ngWkJLKAr#Td|2pnS;2f6rlpG2A0JHiIO=
z5+WQfOfZe4F)~2aOizvEZX9ztQ&J^5bXIDQ7ldu3C$RE+BVYYc^M$=6YZtHHD!6`A
zY&p%0GMm6zvbNbS%AXkTAE-uv<S8qW9Kw$XQ224HeI?lR;BhoULUdE+Aizh2hM-sL
zNGNsUxz^}gs}MFb7n1%RYb9H+pp8&>U8lVCjeY+;6|lHHCPc?Rvp+KRsU64AJ_3Ad
z9`(-VADnl*f6%cfW>tp89`<+hjKpT&qlGG5LNA^r^{2B-thoM1Ra3X=hOe(y4Pz8m
z+#W8k6_;FN7Q41}m<2V{XC8=QX8b30rgTcHAFofHNx<oi$|S&wVg~50^(UDdIvIO8
zx(U9sF$HWcNKO216=~@3`hM8k{ac1%zlH8?n{vwbP~ts%)^__gNJ3C3DXXW;_~uyX
zN0IBQjOzY9fpe+>kha_=XxkpUkyLzk{y7O7)BRxs-ii$(iE*S#WS^jRs`<=0!fmUI
zP5wFLy#9lwrRBvGG?j*Ns-y>wwBBf{@lW*gv45JH`sA!}n5n$Sk;mEr!Uqm|&!51X
zB}jpN{QM~8k3!RY4~4K4IACA~A3_W=Sny9*+0z?|8>D@;c}*%1hq|)mD-<8TNH-PE
zLG(TQc8HNjf1DRZ&jZ#3j+;!;aO@^;RC|HXr7MsNKj3#@yQ|D@Xyqe;GGYonc`TC4
z6MkYBO%=9^2N_-Q(APZ96)YszbT8OUsm*2aT}jcGJiEF^0G6l=UWMH%f0Z$qZST0{
zy<6AB`?RK>eyZq6jEC89hu?Phoiut&y3cme_ASe~3_h-f!Gb=&AL3q>8S(HT5+N8G
zgQ5UTU^_yF$E$#t<v?1!k<V+t9ZRRX`c6FHw_8Xh=4IRDDHarz;UbL~XpmqY7`ID1
zt;*LT)6FIKvpt&sF4gmBVc@O50W(J*jdJ!KrccbVi%^x_!1{`{!om|*tu(pBtDQ8|
ziHwEKC|t%8^De*oq?yK;?|{72({G1ow>Db-Q02ll<2v|LGX^9bcb)Q>??Tff)RC3K
z!xKm2A=zbw7k-9}?Rtzr;h;ohEEKDQ{UMN?^a;xavGY(Zvw=u(1n3Wa?NVp3xjt|c
zhWY$YaSM7w6#k!we%}%3WH=`ti(m0Ul+xp&`rUD#bO^nCGp=~=@XYS(c12D*^WXDd
zJxEm2M5k<JZCgosXOCS;QAN!cxX9}ZlWcR`&iBQ_p0SPW2_Xgrzfr~aC7@#eTr<0I
zHrVc2$2X)p@>^<SZw{A^)qw<W{+*lmb)nSf4A2718}GhaWKs5!%ZMFO(nP5<?~4_X
zyw`cxj;_;!@C9<({>U!9uw9X#sqx@n9@p~``WFVHw};L?p;>xpZ4PP*VK@2f)yd;`
z8e9?zjxDjZe_gc@xHYMMZQ#aw+=Dk=JlFsF$f#`?$mo28fqJR~6glo0PO1B3+Qlqn
zV)Eo8wypmA4Y8Ur0!n>ckYalOXv%#8<t*0Wc;^1+a05%HpR#=XPk`R`Bx0kPcM7!l
zjfR@_;m4wMIgHwwS{rKM-h4slTh_l90hv7avo81@XNDa|Z}_QiXSBs?m7Tx22cUTs
zxU?=7x-mUSmp+=Rn3SsZ^vJMmukv(9)pt0fHbAP2z41(~!J-^zpjqOXSQ`QC*)Mfh
z>X7(pI5?+>#?=2P4`Eg)>6=tZ{ghvDO^haT=H7L_nAXS}mfV`o16o0AUX}%6Ki^;e
znmEF9>O8AqCmrK92l_eBE>Yp>Xcw$wKJegJj^^R!uk6UNjF4B+sq&EUZjeHO0zr|u
zy6fM&dl)uji%5+%*hP3cO1^vysJ8@9Hx;7GU%2=DAMRjRvEW$aDDvU=7=L@J@G_%_
zV?~t1thNxk!HhAvAt4a5UTW0Nb|WO!{9+&73)7k|^W}4z&D7&UC^&2aY0zxoC+@gx
zhm6nbY<U<^8Zs0B6()96myP}w`Q_r_uT}q{%ZeWE20<RFnASK`**V{qKw5thynxrc
z(;uJHT0-`a1U8y5oianYasO1XOu9@Q|0jkBWI+so2L7#=0#>Z9=yCd=$xFRn)6jbJ
z$3x2(@!;c2>*L@LaO&TE`1TO<?gpIC0%_oQQ4k#|YM0Y<=nKo7*B6REyi1(;Off>{
zPuO<O`OqL;fmNlvLpQfuc=AB^3EiuAfH(nr>61q={F{w<hiVl<o?$OGUp`&y^ED2Z
zl~4J(4i>=16`&&?T(|x~^mS<1ZI<+~f5#n3J)yfaCX2D~v@bv#<URL@7K9sjZ+JK%
zNa$GCvkD5%@pSeSQ_9<Kr>>Q^^6yd~;gP#rW@J|2QFK|(FD?n5k0|KLy9Z5-x~L(s
z;<bO>^=luQY#5-oVw@O-mYO@<PyiG@a#1oLdPygp<ymU&_JMzt`xF%|!sg2G%)k?Z
z!#BZ#-ggx+419FQ4?#RK8E-KEL##dN)jksi>Zvaqv!CVu7_iHbwuAzsW<nw$-BG4k
zxF7E2(3W9T?=y82emH4!vz+;W#l#U)%x!LQ<r(UbI*dCHb&~RpW{L~1FTZ<V)7X{E
z>|FX`oF31vX95L}(dCL0GLHO+j%*_fweJ94ssH-bRpPzC=BEpPK$o<%H<wWj5KOVU
z-fQgTNzM59-?Yssgv+MJV#v~~8dKj*dK#FHV$ix`$Qo_o{&c^cYeb>%XB!`_+DzP(
zs2CnY&hc_i0zVLW?~R#a`GQUY=?YIXRtFm3d2zi^%8JM77|9MKZ6!(A&5PgO%;>G$
ziw=|M>=7bRq&J<^fX3>*O@w@(iOKQHyOvU|BrZ{Ce+%z2_CwFlST6umru}_Q@?BGM
z?IWWV;H6}D(jzj=H<t%8fZAz2k4x?KzVrGr4Fp5;ql9F}=KCbVD0E@yf^TVP;?;)*
zP$h*y1PT^l(RQn|$vekWwezvl=}6cCF6t0$!1-_Qz;DOzR}Gz%emw26sGuqq6dKoA
zuGc^In9Au1jK#)UCFI~CW`6loMm7-WcWofc!jQ_SVdhKFGpfE;{)ljOI5Y$2=gR2l
z>NwJY!++NEF0}n-_TDjk;U?nS9JL$wG)0hRn%?AozsfAQ+>ybp#l04yb(FSwxZ*ZB
zki~_68_g^H!LIUX5^-Nz_m1fvZ*U2{7zuk4d)nyowS&cJH72|v)LCOf;A($%d@%`X
zI{47EC{CLneWDr-G3Nvwoiw-g<Cz}`o}aMhr7>aWpA$pcF-x#S*lmChKey-hfXP`u
zV)0jT_&6E2HSxVV1c@QMx2L?|$CRpf9|Ft1y>ueBNB5FFN_Zas9=Id>mPB1}pTrOC
zzr0Hep`fsKi1E3)sAafZ@ib7s`~)wmVNd#!rZ71(WQ*zq-;;BkWFgbt#rAvd&}LKM
z;P1k<gTwt7S7*iStE3!xa=>GJCd0+io(-@8YudNjU}ui<Dl8B+itfGGv%Qy9R`xyp
z@f3sYUf!H7kjp9WtF-1YKl@Tf8Ww{J3Jd-CW0axb%654`*WNwt6qTK(=#gDKE#SOM
z8VEpTmU35RSN^PO79rbCD31RbV6I4A@eTm=xRc-l7(?8OFZ$(EuV4IlRUzu>7`O&x
zu!;R4ST^3<EZ#P6`4>lEeSYfN)SKsE=C#2P1enNlPZDo~+@AIhxBpjkLXt;2$1wq-
zYQVwfof4$s6gKBasv`(c;~jz6xSY6PqXmu24M#Dz-b9>E&Q`zS^!HoV_W}qM_#S54
zS8b<q5sI>S$%;BpR04>`WUW6(x&gnPT$A7y^J;us?&(i*#sV(%bZM1tY&9d9z6%jb
zG`$*2PB@gx=L~KP(DVWDA$5Bdzw%>u5l-;|WpJUK=!aH85NXiqF)>h?B|8U?VUT}e
z@#2}xAEFEY`1DEyor5k*E<L<nIH*Q#DwIsfO|v=rmlso)lW!!{@7a3jt-w+6>yK{)
z>$ifOy8)!SOciWX-6G`eJ7WpSkJLJKq#5i0bd`?oe4j#xB+X<rX@HngDIV$I;4uDb
z2W)x{pn6k9I8|}%)5lYKaqt-LOT#O)XGs!yh*J;yA+f7+7OBoKH3XXjCDZ82&jB@!
zrP|*Kr=N`z0TIReA{b>UxTqC}xL$6t-v@GCNzld4AG`qlf1p-2GV=|c1B<BKIhl%6
zSM`ndGPN<@oWWK#O>1wx0lk{+*S>uk5Er4Br?dhhx*($hcG%+>K8^g;<LQ_t>B8y@
zvwRC3X>SPq*70R5JW=*0D0MmYQ=aztfq^=qoCqiZwtHcy$*M89f4N8*WE<N(B31P<
z(-{ZCkg!PXooTBp88V2|N$>I<-ZfZ%I68gJ)^;^PtO`Our^)qLGOMi0L}5H~6w83S
z0&GPPUwhDOwGTvk*8NbnX2jD7qpo}IjKdijdjqOUI{_?`p8Ze`9Xq%WeHZsjemOt=
zd{@sqt?k5zTc)3Z_(TPPH!Qr?H_w{=@|_;riE>ysLCib3^lKhQ^D^>VNj+$uQ>AaH
zU;XpK%RrRO25!nsw_tnA&mMroJUM~CeEBK&nUnPGeX&qw%7*g7OKc@F3JBcluE^rG
zR04Gon4AZqt>1FGqA4zKoub+fV{C|DV0XCpe$~}&T-qIgNFunjFGXGDOcq5SNo^4`
zyOf)o2~B!n@{qCYzZY93;(Fv!YT-1o&asfx)uOdBh7iUnZheXJfhWpVb@#iT3E}Pm
zmeS_J8~l#uT{^|``jN&O?;qVyBY8eEf<U~L0!J6+zdtIcvC8kw$rC$NhLIg)q{YX=
zO!Fd<KWF~Zh*><Lnv~xRj51~q{>SMqnILgDwu*CXL(y<oxlu6R&k&*pR1e!ARAw5B
z7JmfJ*wZ(%T*cqyFaI66Hx3KmvNZgEMS8|x3Xi^Cz<>AgFy`WY^5sGa$?tePxCgOc
zk3df-9p;0h_H*|lbKXkO?E-Bj)z>)*5`j)7eYxP*3_Ov_sm;ivje-KnZ8DSk7Yr#)
z&gJ>)-8MGN)~ODb92hVhHpPP3{pxx{tZ0W6Lf>O!S|v?m=p*lNWz$Z-Q3_*+@-K#8
z&o(}Jq)zDj%FXnXU!_0PW)E79{y<`0f>?IH4xYBZG}YATp{|oXq|2!J3pSX|piO3c
z9^bx|1qAf;1XEo@NIF-qkEq7S$1J2e_ThPnWmR|euhwOHe3j5VPE+W|_sRI;1M5<O
zbY@b)-CuPN<iX+^eLBL)z^VJdmAi40`%OFzE&h7m7!2<GFG6!m3LUkX><bgAeres{
z^I1^nLQA6xaC(HFiWXCX04L@%88`b3eytt7;b<?6h?~#AFuRt>yn`yo-Pu@Mmpo_S
zSbwlMcE=f_UV-V;_pCPAA-(qj|GGohbM4|JAj4Oo9{onyKi>tlcDH_M{o49Xs}F(B
zJpu`vt-$xVujnR6zx!)f7bnl7Q}e8Bt<Qiro|p(-qV)6fottu7pK7CGF021ajO+rX
z*DwsX9hFQ>-Rl6ZU+6|3?yS-y5U0JJF+-L8$lq*5bNH)|N1^Sf+JRrn@4`;KGa?U-
zzTAAfi9q*5;TK=OvNQg6RBAK$XEI-kMIHVMl?Wz$I`!A-ISOv%wB43$oA9b;WC)5$
z<GJEX3t|6na|Rm*TJ`|Swx${}*suBrqna`9fQ0(+_ea{Nm&$n1IPKRrlnFALYehab
zo#H2;1|XxS)WlpLe>}}1)Pa63*(;(ffPv1~)c@`3mzW-1s|Zhv$ddby44u<WhTqG=
zx7JgX^M5xU%`-aPthgWuaOPb~KV4?qcVqF3?WidGNkF)*0IgzolNx(Dv*bq9ZvZph
zr~(nD1O)PBsjg!)WIo5Nt-Y_08mA-Z_OpK+oSorkAkFoFyhi@3@B38#?uh+RiS1BW
zf(Sb?ds5MJCfqGKvOM}E_)Wks5>{f{cb<x0+r2&ayk;3N=6z`KLALK1nX(=17_KB=
zk3Py&TqgqcB^?4n`u9}IY3sUyR0`1X2irq3&oqxG@wA*4y((;Zkdl-5;>AIMxOPy!
z-yOM3AxoF{u_cWPOWv_|5`0d#4rdY!=<N4H*;&GIPBhI-q^fs7#dZOinJM~1{GnK=
z*z-T)F77$h9!1%rUiy&#`n9<qP!#gi=~Izb2oj2k2V#U;jOCt0c@RN;xTcAZ)dnif
zh`(lzG7#!GU7B0N_Vb3rjn6wk+BrT~74O{}Iwy?&Gv}TN#I-&GUAXQ*Q0NtIoc$~4
z)yx#s1!3b`Wn&Q{h(5#^d{WsD#2Wt;hDCc*P~DxAJ$JsWGFz1TUb;#sk2+x;&@%n&
z-am6IA>RYA$V(vLn30^+Y@(a<=A$b!O)0fr1gl5LE@m)o3uj(m)+K`^&MukVd3qt=
z$rB|-4SY9*Z{%_gf-P!EKmoeBn(Sv|Vlqv0>rllVr?c~AxAF>8uBs>TtX=R?OT^k}
zW}X#+9^t;Tbj=_#dlK`!EIhgr0Ul~A=|lCqRu05+Hj;{|+@#UrMWui6?(vs)&h;yC
z3Z3{H$@Ka-tkWD3!1_lCmF)=|&=VMQucSHuXAattc*vw)t`@KULr?}{fPsF*?;<oK
z#rg+(#`J7$m@rFC-no5^BcDy#+S8(sUmU!gzIUzQkm)1&HNpsiFQRHY_}wqj#`;c%
z1#Todu}UB$)XR3XlvraIgZQ2-zUWxS`8IyiRusu>u#W!kie*j1BfpsYJ#Smc)E0a?
zTrKvowq#E`%$Ui@On|sKF<1(Swd*`uoHUZKA|~djtabs58bK{v-HypS%t_oHGIjte
zu2PGq8rLCCivkL;CWm4Vy?VVeq~>{p({-U0UV_Jpl&?I7Ca^%eRI?D0Q2+iBag=0x
z$#n6R8%@p7pIf(YI^A@>=@LytRZ~U$`UZNb>hothVNvEoJL|a5zZ)4hkIlYv79PjY
zJFS5R#xGNY2@>rvW#JCq8tV{1)#{<`ETIV8X(;QaZg%gkG&u;JIV3bi6&BHCFe}U4
zp|NXyWeHA_BwUHahsp!j<L$GOvd!%L9GAIFGHU&L4v-Rmc&MK)06!{x_oLnH$7rxx
z(5<6z*o;u_E1KF@vA(4zLuWG5J1}HkREzPP-4)M_laG4pUub-jylKGx#30to;Iw6B
zC$ECx&=Qb%Ru7_|Gx9(MES6~)?bm|zgj4!@ep~GhQc|a$)8WJK-%Q15o^NdBfbQ8d
zM9l^l>=(z09a%p#(=(TbgCIkZ<ilXZQR!m16P~9%f)ZBXMhJ{e|3=$#0V@Fl39R#(
zxEXX!?DTo|xxf?SOi&vgP;^3(`G2S8pXvPfIv3U7y#q0{CH?~3OP){^H_v_)VI)R-
zP68rOl=d&Yu;i;T(&-F1cX<87!*_t#`}s2La41f7zt~+(HjNYRHlkrZ4hW6spZn56
zsuKC~9~C<!c&hvbGM~PQc<V=c_v8gN7|i@@@_8p;mwo2_I<du@XV>XT@yzp7*KfN9
zwgM_}EYvALtq+0c47m%p&SHW9t)eps(jN@26sySAq9A{k7H+wPm*>3%FlBP*9%jv3
zQl~HxBh9abBd;@G%LhZIhgB#|h9l;Bo_rs&XC^pT_hZH~j!u5*JPwrks@zmZjmP26
z&ssC&A?-g?RarB*VSLi1>yjw^KqHHAsNgd7Y~6IVW~FA78)P`-VZ-+Op11rf&p6_^
zpslM$<lAwxMudm#CRSEU(J;_L`zw;8|7XGTF{lAjmudz;fJ%Vi&K2R>Ic6S+?wL_X
zu5f6&twP_Q+L{!;5BYefRJ#-;21|b`u<f2^S7L$i)fw!iN7k#xx;Kn+8sII9mdeoD
zesxa;%_rn@=heT4XBnw>@(CTu5nf%Ip9t+Oyr|TdI;X;M7Q88%+V+s`MK#-1r@IT3
zs*>x|8xDR<@c@<I`3jgY2Q!|l;|(_3d^=3S*LF2rTncgK-1)$s`Om13=r-bS)Vqkc
zqqTYGnW!Kswhd&l>pAUpf^RTzjB#G8$kS<DI}y0yRD<5|4ws%Zfg%n(7~a4=Jv~Z7
zm0RcOdSsrU7+4KUgXfgh(ZNvikgy?Z@j|7cz3Lo?8Lf`4ZkJoK#sk<_3yltB-#+%D
zwmA4fbVq4%NDwijw%x=2N~9f`anaIZTG3*ZkX~0}`)lZ@)zyUc@qf%eHG~m6OT*J0
zg0{7uTD&&{dyhJjxbuNpk7!@=%-Bq+u6)pdMv*vICywN%VFE>H5zX^>juTW3E#K-@
z=eT1@rQiI3RSi&~eqjV>&v7;b0?cNQ@z!v`cB`Wf-14I<F~6FQ#QO*`JgdEDROG&v
zq7&#)p!B;F3(3&jcow_H#V%tM&|3l4a3`Eno;yemcwq43e2*h9DywACYIpl|(cRdM
zA~(J1tli`iT6K)FBPB{tM7GGS$mO@UmR1Okj^)_Q(1k``O6p3mxb0#DsxwLM45;Ce
zhDM;fius^=tA$gERa#cN?E|9!*3(WqenUIac2$qx1xY+du75aBjkztHnNZD-L1!*H
zmiJ|8D10JVYY7Q}Z8Wk)$KV%Mv5B0e%D{KIF_0t2-HGXops7(47*Mmd$_ObDW^`AC
z^Ay$X1kj`}kdo~B+yh4+JO+}HA}h~%%~s6BVUBW=Ff?!PJnq>=>Nbn&pbIpbPbC4%
z0=9&$iZi9TWzW7p>A3Z{=ZW@-6!J^|+csgV6}b+QX^+!%1@D(^JrY;?^rX%5$e=<e
zY3)KR?Tx!kITEn`&)b5gJo^2XFZQZsct~-iFUfOWVwB&TJn=e^JI*~FgSg`p6$SmR
z7YKhRZp!k0;+4vs2PbA15Y)>IisRDt_rsdiie^WWEEXFcN<I6QAKaEiqH){?z3j~v
zE@pWZ^ndQ^O6^Vz^@7g@HdTt1kfmU!K;PxV$1|2Qaf14RH<6GaLyD420vQv*!Wf|E
z?P}1Z&<V>5$rj!aX4huGXIg=V4|f3Efez*f?0bjaGm}23k>sSxbo+)t%uI$;p+o1s
z?w@&wKbf=7%03i%3+Vk;TM#IqB=^B9<?vel!Rlw%2Mgmz*a!3R56c_N9){3-tjGeZ
zlrU<JADI^gHSva;e%4HCDfK|un>;Np6n9uLSLrSO*c*X8?N4H6+P$3E4;lXCG<m_x
zbHI0ClSlubGf4RvxFg5@v3K+fi3fMza}*n*jtZ|E!$eQDB*&@Ni1EX-SXg)I$I_Vz
zHpIyQGe<L=X%+gR?gp@hl5St5Lbf`^;QPnJhQq7rle_YEK8Zx$XPN96yci!V3ZfT%
zLp8@>3v`fl7C?Q@iWhSrVP;;`<BDX99to(`wpwZ<xj2*U>OWg{jPS-&(&G<wt~EWG
z4i+LH2NlqWco^s@i)S*b&agwj?UsYs9jt?}4%W;%Ye&$Lh2t*^#eTnv28e$lG!A6M
z6O2$<5ub$j+)0V^>Z#6&Wihy4D>~60=XMCZE?iz*J`)Md{|#scXJAK#SBFh+9tO)L
znd_)h&gwy!f0(Ym_7`6+(YE{uX5c#P#I?gGYImFRnqyki7*%<m8<Iu7=S{T@)c{4N
zFwM^~jmcNTjBBMX71cw_uMa=0xd{vWBaexr*0N|ZTm9-^*cESeqR$2CXqcEJ5V18_
zJCG}J1>_q&jw+%p;=_3j&su3;TXPb3EC*{#$3055wX6KMLvzrQNHqQ6Pj>n>g|Msn
zU#{liNp!|gJVqi5qrP{6XM3`8qJQ-YCkJ~cqoA<+`T&qR>X{cN-^{W{+!SkNvnn*Q
zxtzsfXR1}DUwp50_iRSFWZwI`{U@j`Te%LzJ!1auk!!IilxwD-{WZ5@y<*-Zz;Lzc
z#K<1C9U8BCIU_UL?xkY3r%^7zT%g_tc!*JJ3<wYDO@<ct+-tVG&5LTr`!zI`D@`$W
z=lED3nlrL*siW9YU=7!>I1CmSjoU~1Yhf#2efoA?N(-1nd{x{HbH=#{9Z1ZG>B68T
zSpqAqD3fxF82g(Bw=RFE)?&HugCv?@t_uyww2jctcU_>b?P~FFJ=Wm`h-!$zIz2dG
zvU&q!rtw~vAI&<r=)L)8kBrXFJy7B-XDjY8EE2K#d(UQgV$rVPX<0;-g0-TvT${r)
zC{`#K6at+xjz7S>8^)QQR9_|jX+LX`%?+8xF&O?e&zaiDl)Nlocxb90ct0Zuqo_QW
z`Q+<%yBbdD6y<g76&mX}?{7=X={8(ZC?YbE6?I5;l#0I=c*~T((rpcWk~DUeJ$UTz
z>RdVutU()ehe2ITwosbs&hm?R5GQw*k;mlK;%JaqX|UXxO4oIN;rrgbEpaLmDD~^y
z2i%&7UST>zW=**!l?5`-%_DAUz>er)varU@&rBHaNq>=ayCw%>bRb@GX$z6>z5>e@
zhCTduZ~pZW5W9xeZ`_9AmywqIW;VBgp}*^l=`G#Cj~2o$oU{Iyw}T~_k{kjH>CKGk
zy6a<_PQF~FWjHy+p~m&YFK`lMAC}Yt@7;0o7n_0fcKD6a{%0TNMp{%nyPWAKMIxWV
z*R+d6Qc7rf2Ajam_u34Mfp6M7&tpFoxm2EuLpQ;c1OMDZVBd7;e@RZo#5qIB=ImW+
zmST8A<07Dw%NSV|wf$@+<GDoVl>tM-V$dZFUv4N-DE;<lsS`^YOYcCrPz`j!BcR90
z3?u}0AhHE=?u0C`ZirnvZ#GclF<N!S+h-^;{E;R^#ok?A>#9drMzZU~hEh%Y9T@p|
z&^HNVguWR{hF&a&^L!5wuF)IPSgXx<wOvr$3L7U+W`4;Ph__`$-m=)&nXm2*F|z#?
z<i8>nc!V|Q%}afKV9%ZaWmyB>rW$*E2CBl%I6xUl#Y=WJ32hmE+;Ybke}5`ibh27}
zBBQvZOG|q)kn@zO)1FoSp~k~(RucNSoO7Z$`tk8Zq>-8D!E2xyA)(Nk7Z>FY(zv0n
zI#%Idh8gT`nWrKmA#|Bs+k2;=W6Np4J;KoAZ6!6+-k3xjJnq?s4Ft~rOp$WFs8D!^
z*Q6betLLZ;U%!%mKF*neXPTDj$iJR5Ul$0wMpu%j^U%TQCBi<#6Q#dy<fV$<`oQ2|
zTt&AKWpssy5~i!?b@$JsKT1=ZOz-4{{QqTv<zhXlSKfh<>_x-ZtQa&a&pGf_Jq*IO
zrC!d){WKQ#KUl8z`;pNqbWOR-jld@~(3`=u_-kl;J*+{r(pT+Cs1uE;tL{G|l%4ZM
zr#>t(lba-gN!IxNTq#~|g(D~#moU*gL94`1aOKgCr(IEtCr({0PRs9ny(XGsU3RxS
zQ+Q`#@6%NiLow0+K(mnO!Q*d2wB$#(Ve^Y(oF|QMwDM!Xxs^;B9j{?$qG3Xlz%0XO
zmxv}Q7Gjfe7hh^{LL6u{_PFNwC9j6qPEK?MNB=IRG~-2y<E39T=J7$_96#MYa>$M`
zRs|3{Zk)pyns@Dc;y~N7QM1tNcS(-lErx2VOY;`jb=Tm)vC%AAXy!$H?H@$J!jcr$
zEw<D5ndxXyGNxCygOxSW4X@bCUMf-k6Bzs`uChZLNW#|#?c+z=5!y~x`k(Le`(OTg
z=(dqb$6H;Qd~5@B(I1;_6~F#n{d`ejdGEK$a#wqVbud1W^hT#Z$5ZElIl*Qv;=Dku
zBAurqj}A|;{nfWJ2OA87Kkan4f_&uY$%ohjcI~V_9zh|~S@Y(tHC;N?r!Q%t&`&^D
zRd{SweJ@QRTLjJ0r@e<nt3O=1HmE%^W4P=(R`dP+y<ZDv21P*<=@zie!BEv76)Ogw
z9R^jCoudO>*k4h{bK);whP2-<d7C~O-o?<rR3;t6^t)l6eGfcj4M(FP%ZqCdzunKV
ze-F899<Qz=bdH%h-jZ5$b$3cqpv#e)Odt7A3d{2IU(+)R>=y01PKF=-Qa3ebf46y&
zpFfExe@ocgOuPUloKPZArvDvx&5gyIUZ`PJ$KmK6#zLkw1d0)2%XFTxr9@k7Tc3x$
zopEDtV~<{o^)4bGksq$*w?eD-?}@vC@pfJxn_7pO`cFT-^ufl)=qMKLazuXjzj_Uy
zqnGmS1jF4RXA+p15b;Ow$29eyd7epjh7c_+tHRox@j;Us!N@qW%q7nJ(uzO0?4l9J
z0p=<d^RE_KMrHqfhmX(+(hBiW+ToqV#RDN@<KexkI{`&3Cj~o#uw&{wK)D1Yi-GLO
z48!%u1JT8K-LkeCWHs6p&5Xz+^Z~pj46Q#coM<LV4&u_?MRgY1A$or^7)4Fn(C2k!
z6nVG!N^G0P2aB#vOMx9I$C(1x)@F!6ce2whRq{`#(=gmP9@Y~FD;*9D0xd&jNHr)V
zw+t0m9hW5q3xF<Mf*YZB6#qum<S2XF1P?`?ljtuL8BAa0jUN8sAAUL9q(=O8Jv~#M
z7{YWEJ49Ey{dxN8?1<awO4WQ2l{Y1i`2q76@8e4D<wW0SG8$Ncll@Gw$7SHljsW6p
z0=DJ0veUAKr(bQl+h4s-*ne9CVn2^(=0UIF8>vP+F<BBrBlIdmTFGv>+`)#0<b74T
z&G<SAqsi8J#eSaM-K}viS7BcRk))gBd_3xQ9X{9mlEncsc6yo@?*obXVdwbhvss&^
zNaXnSpjg`2Ilo_0G+jLxp6*9ZXfO!`2Tx%y4=Cbybd@7(ZY!-CwcAEtq2Nj$29q=1
z>j=FE`I(P^Y}Tz9wS4!c?z?uPdB+q*2y3b=nW-ztK>8Brp(&ueYVc8KQr#>Bn_(wL
zI3-LX{JYW5@E$`~w`-oVZ&{Iz&Q!S#b}msUnh&FnK}4v)hB|mjpAnM1Qb5@{sJQA0
zhj4hbNNMpYmL!6#jj#WxN^7KF&UQ{Pe#<?RIrA=Qb=M({8D^Qg_;1zyc<>-wGD7S#
z6NDta*BY70ewqGhxC0%_V8tJLrQAI9c)3C^@t)jQK6c!Ha=F%V<hPM=HE5tNgq|i|
zf(Js$VEt?IwNGbpW{X1V(m~7{MhfkiUvr0a9S%9OTNRw&w$AUILn+h=mY1L>t=N}5
z8`UR2`#n|5%tfICwd#c833i=(z@};#Y&4mfolCx@rxkzO_Gt6wN;fp9W;tU>i<-u>
z?&v2XmeMK>)(K{?5zAwv$+*MUj11&I>N5C4btb8UTWevZ;q7h-OSe$}UIRy^!v2GI
zjX7g}`Wdz0wTH{?Dd1L0MoueEOD&Rz?66OnvhO0$5yCvrR^r`+KWq?@Jo4vfqEW)a
z|Mu!4>C`0WWIOlciZ@<S%;Ya&y()dz?i{7__gpsEIrNC<sz+e%*WAu{vTzc2jC0o&
zcx%y*+@(UuO_c2tzb4!+pU=(weWtEg*&n?uM-8h(y5v?HcGD+h&#?q^+p&c(XFh<0
z0><CdyRQrf3G@6r8l(P^iDr%lZ}n+!P9O^%ap^d(|E1TPlgEr~7WHfXhlN{=kzobv
zZQ{co_n|P;quwylLV4clR#7UNG;`Orb68p^JCLK`hzj^y<eYvWgO1D<%+tyY8(m3@
zEq&u!JV&$A*P=7@AMpg8Z<jMGz*6**G0fjL8zKKk(s{>I{rz$LbFYhguf1nnTq9Ca
zR^eWkxX37EhDr$`B%5oLk#P|%8KvxzD9TPo%gD;k$liP2-?`u4-#s4g?Q=ipocDRZ
zU$5uu6zMTg<}2M5PF$$lU-}GSm4`o)A=d2ty&T6^_}k_TvUf!Ma^{4!(Ns}nNlN+e
ztn^oxe|&0#iF72(lQQD6)3|Al1pI%9Q;bFq^?!sXrFaRyQ8J9GihROS_RNjHeoNo(
zHa`_rR@y&B`FNE=*?6v9x6^bQMskqjjmtt{5;>D#VQfD}!oIAAE0)VKJV6LQxh{Pq
z7{~-r7orU(5E{4M)O~b@i}-1rnif=%?nAa`0AU5L#Oekd0H>1aTf3-e72BXqvjEEB
zwi=&M?*(!H*|V3J>@;P+1``9r<sp-nYhNDj263(9IpI=p;*tpZlMGFxH+~8ii+L=H
zwRF}+3pbT=mWgIK^zaF>g`Z(wDPqtm%Fk^Y0MkdX`*6R8kENPGKl1thy(I3-?=8-g
z?#0+git!F2QOesV>j8IqqX)f5IrGvLWg+6MjBzJ#Co^`5#d~ly6Vr^?gONgTu6TO0
zGS0EL@5M>$j~c`4Q!!8Ky}`ivG8`GUS?lw3GkKCCeDqhi`p+#ne9xz+ZWr+XtoW+q
z`%gJNc>Su~Uw6FyD5hH28&BTVtb8BeWkuWeE%=8Qk%p0OHHu_v-q6GjBq%~Uoiwz>
z@(Jkt_3?pm&lk38-EmWHi&4oHs#G7&qf0P*w|_I)`E(TiGfCrNjl|@r4u)U<aveI@
z%?@+;PrHw(j!qabKu9yjemBMQNkPm{-I=j>c{y*gtK_duE22ni;fRM27Zt|?EBy5t
zqpG^w!nhfyG*W7JIA{JLVB1ZcqGHI5kWXtmHpMypB2~KAxD{M*OkvnPRh`flv4GQk
zw3854M%J5k>6&~5`Ks|uW(rrujc&R)TE2JN5&8ZF(kC}ZYyz)3+JX*RQ-$m9RF3R8
zlLT?f;;inGrLwLdLjN(G<kyT4Hl_UO#<<*F90f21Du>}}&>`Zq+?ZfP(Vhkuo5(g1
z`lVtf>;EL29b(5flfpXAV18i?MScbEez)_tR8G+pr|f1iBAxZ308DbreCtZ@#a0vc
zBRPMcy^UG**%lg-TI%vO2_#!dqpqhF4zoc}{q7fcozvcjtDbfGtHO6b&&9g?BYF&v
z?u`HyHQ`5F=7rHS7rIE6g0aV6@Cv*j7aGJd#l@(*b*ITlA8g2Xm3$LBJWNxfEfR%Q
zW{8B31g`%qzMrV#^>3%v>^KbZ_XxrI1s8in)f9_UQb!Lbq~L^ov3w}%{?--F`>X|~
zdn(3sWYaY@B-RhYE{{m;=5Uc-A^#<?4MP2;Wg3b(_n$7-dJcSJcvN(ipE|c<Z9(E4
zbTo&;Oe}5qt6=s|eFXo-$gI3;(j8F7VJV1QEQQF0GG1{-x*egd9{e~R`V@aO&QrGn
zlXb;1aG%(bozT%}o%T-gz6ZwD{~h?3JoJv<FwUquJ(!PBJkPH87-n+{<c0+oVxq;n
zLWq8eP}{kTG!9cJP51}@>w)E!iR%l(Xp(|v^B>CnNi~@hMaC@u?Fqh&&@h6u>s`zu
z%L{1}VWl4Lj2k`0LVjzfo<M}r=_<8^?+tL4qo4J`x^(id5MQ7yP*<$2tg6e+`eMk-
zSiYXC#GcWEe@fKAhKw<%V_g1L`XaE#qwn5fjWL^#`PT8S$!#?Ib}R7u78tFKDHaN%
zT-d>zc7JNLNpb&~^8c@)%m@KGcU9GEH#pw%%>`K){@gEu?#*YDmx&TD;vO@}<6-$7
z*19r%Xj#!yH5*AKM_xakp4JqOx8EJvPM8%g3yGx|gJrBX-i<95>o${qZXwF0P7Y*+
zROPvqq~RFOG5$2P{fBj!FXh|uK00I0DjjO?UhX*L_Sr-1%<xd>UQ&z&C$w?&8(vSy
z`+33-LDFVK1r^zL!t0~(fb>_e10S<50y;>$WC|7RL)H0%4)Vr+I%Y%R{^e?iXfp^y
znLc`3s&J=Y?oqtW$CG09e}2*1gOO9wt5OSoNJPfgjS#<2xgihcOg9d!U#!zeMK#EY
z#~QF98a&;8@Fj~9y24j2?_T7N2o!DVpRQ#251b9?ZW|5Au&E{G`Z3p=iRZnZ^`t!B
z*M&G|&==k^NYIk(*~}?}Ry7A9M5m1&3R$rJahGF!RvNFXMyncrsdN3uZG5wJ5hx!6
zXl~1s$fpnx`s46tpMXEO36q=Y(#X``Xoykpwc^K<D-SA}g-OT5`Oh^&b7DMT?|oHc
zJVSBclo#8Jd$mMyyp;V>kOMyeS)Ao!L=oIXFD3*2)q7!)yk2$@yL$DHj_pDvq%2ze
zgd1&xXD>wpjMxvFP7g~*{0+ZqeNr9boXdt&_w&S)4y}*Kd`z=g=+za-n9m_A)>c9s
zN$4VB+56V1IC6E$fft`XI(0rJgn7UgEK`FJGUsi70Q5H8Kg0U_`bzW~JpnV&U!4ry
z{Whdh@zNEN0Gvb^M$0#=yb1+**)Tv0*LjuL&2ZR<MAIV?*lk*Xb@!}QN<z<+gdDn1
zJxP)jY&iFcmI6vf@Lrx|?bE4WwP5};b#}HTn{EJ#a(`Vn%>b?!V<?hhMHfnf1roVT
zgXew@w(ZywPOVgAJz11N0#9Hrq`k=MDg2}82sl*lG)1uimgaSDZ|`6|(3XlfgH`qN
z?E;PTvY*&RaY~rI-&Bxh;v<WX007^$Q8%2($ocPIZ+vNQfNMVGzu&V{FAjZ%QTvtk
z^_kG@y-2MXmG!X9fTFRF#L|_DAFp#LCtWd^SCal9hNPAY9}EbAne#Xa`|{6sS5!xP
zPl3&1U--s+rYDo#BfUlLU_BJ=rvUJ9z}p@{<ZGTAzbq47c}|CnuWxTe!tg0tU`Qyr
zDRQ^<!1no@bpR$?dwcsLO>TaE8F=;Pniy1Q@(+nev-=cae=>1A;OTw#LqX#GRmt^z
zwL@!-)@`-403jAAw%Ez*CfaaAZ#aIz&L^E(`E@|0KW~|`c&}^Ayb^-Oc$$B#hJ=wz
zbc21h<YPH&!3YAtu!|}yEA{!?>;CRHoK4uLXOGJaoBlg`lWt6xr8zIA11$%Z(~HwL
z9!f%Ixs6K&bB{dS%Rc_~X=ts>E*Sb$av>|>nfHV2H@UgF1&zYBJVU=!eCrAgo=KB(
zWzxgQ%@t6Ez`K)0#o&UH`0?}b{LY2FQmWtPf+Bn+-i!OCwNfKL9A?bAm}bfeA#|l6
z&JBj$$2*aja|jfoe*>v__Uz8<4<CBXziQUA9t9EiM1CTr!V>5boaRscJIsXkz|5mY
zci{1H6Vb(FPQ>@5PVoRB^r3GE;qQI@!{Md3V@IaO)p=tOjl_fMcp2p|e2pctBYbAe
zWjHyQF9@BQqmEhleGLG|Vn7Rr&C4nD(J`CYkrAh^qR$t>>%4JZs&Wbm32M*l%ZP46
zn8XGe(u32cI-D5fZikq@F`g8+dkm)G?Onli>|FPY@zp47dI{UTZ7qoP?u6XM4uiWI
z>_2_g{9{p#yZqbSqmq2HhRRr3{B<pU2`9k8Eu%)d0)SuFMW|z>45TZ73`ZYP>mhlR
zPx||jQX!A!)&BnenG}`63x%43&!Y8x;G|6{-@VeDNbjg@$;!+~?~4I7_I{TS+p*AX
zTKj}k$B##EY<E9??Du9Vr|+bNN3)s)sC**S-#-#OlP|FBB^W19ysnXbj-X)jD3yF_
zB9u4>Fu^xKlfOLABxeB7oex}+Lg07dh)Ig+&?j>S!r0Fn<Pk#VB@b+(y2p6LG4;TH
z{e;tf=Ry1u<)bcL^cT=YJaoj5pY|Bnx_W}(yzG%QObyBfxi0MUfQ)9{%e$n~`VojL
ze2OlVPXNzNOcQ_8D#kD5HVyt+(>Vkm3g!utrt7#@`2=ynL&*rRNOK_)uh>YO{4b`z
zzCpTGKMMq$gl*O44?e|Y|2NY))ICb1eEjoRL-P*NNwufaJ{}IstjD)*s)r=`=N}?^
zsPfN=<=p!m_Y#0kAtWKYuB1o~F=on-s-}|fY5d{t*j6bTIT2A-6jmz<sk9XGqbv6*
zwIeDiBJ!6tqI`JSw7QDEijvg5FT5nVBrAPX&d&tIRmo()k6Tw4EHuDU0Ol#7jM)Qc
zva0I8kt-b9I$aM_@uTXMmm@#aWhW<x*|YgVc&L=39Z=&l)b4xlr?=iPW%&4g7qniH
z!EPN`s53IT8>6>fDrn=)3T01%7Iwyk#C#ls7`@rWO(eRCAG+(Rq@eI8jFy2)*Qi}<
zpYK>sn6(Sh5KqWVpY&T9YXE|3hl~xA=m<pr2iF(N6_SwAb)%EIur_%91AxE_;45Q<
zJt!ZZ$JSv#*|g}+AF^zpcVX5V1pWv$HWaMwc|)J2=APK4TgTApu%R!7I@G8?6st2S
zL4B`PAVi)klo!8WYp+rawQkIYVjlTl=2_i~eV&#0aTZAR5Z{e<DT}|3DCZ<k#rR(X
z17~@{8|+2Mm=!Z=)sqpU-mpSN7PD60U$hdwmwdqC&~jSo#a5Vz=EEKPbY~$n4YNaS
z;3C!a+XQXfCWS2n&gMKkMN>Mdnn$CQbMxr8=X;QjlCdCh^OJ#vqul1}PeyALYI-=B
ziwCVG{J<T&I4af&E!R{d<)cj6FTK{@Rm&J<_)Gc-CDURgoG4jP=Am8d6gSk0-?Ay*
zRPL~_n&te=bbo2fyJV~;NAW*_Q;;dt0WU2rZqvJJ#P&#61A)LJnSC8DG5*7c*{fVZ
zAzEE@56p|Krlu-d*HHDZg=jW1Ny5gRn&bNQanhtF>Vc4s#OxTdV`!hzm4d@~RJwC$
z=maf^%wPJXbKB5cfL*58zV6gKTh9*A(GLJkeRU(9C+l^2XU(#uZ5NA27VE}*0NmG7
zqRYy%T)iG&v)OBHD=Q*SA{QX(o=HKsgW|S==Y0$B>kZSMw59x+J-l+swD7;L-+_ym
z3;Aq)F$Vpr=y61%#=r*c_jjwot#)}IOl}7JeZX@Ebbi#cex|`xS=Y8!&hOOl&a<s>
z>?AX;_)s|(<`PL=k@}fh0uTp`aLRMYAsVCQcbK&E+c}&aGZ-MSt)3LvD{7TcyZh69
zfGVL|F0hy9KWYxOdSon;NxUwLMhG{dM#LM&Oj%E~Jl43zPpAE?bKZLx3+X^(G%J+!
zA(xa3;_`5+#wLjVnvaX)bz|0d*qEk_bh|1FL+Xu`5a~tJKY4ppFVgm*nBU<iJ<Flj
zwa6_I2@G5)<iO+9&j(0fkiZgD#_rGY-ZEN@%{~dnJ_j_0d(OJuTzVQ&3=58x!xNuF
z8)XOQQpVKaY`fGbw<K}>z^R&jWPPa6dE_o662l%Efec&Q?6cc$jvh`43nMdN{yjh-
zPrI9LU)6BTPq^%JDJXfkrtkEY`Ttia6_a|MCd_}*Sn$&KEQH95%qOXc{k)v?G9J6Y
z&bJqzb=_Qfg3vsFL?=C04j<#FA8noo=_-=9Y%pe~Ng+{CTH|XuA}2zCWIHMsbLJ`K
zrc5YPN=Q@-NO>GwKKQk~6C6UUBU?Cn$)f8ru0TiXLViq6VeEKUT-el8l-#7^g%Y`F
z$#45|2WbzJ^xV<8pQ-T%8$lYU10l57?EXig{6H+_#6#+8Wc!1IDb;MXtFFi5?(Vgp
zeWy<_7C53iI(dM)ZIHRv9z?{QlnB(#{q#-|50`rfrNFr~HHe;T3Dr;LaAHdD)p}C{
zd_fsiJ1YE#Le0<os&?Gv(!b>#aeJVi7<tMGr!|6Fi5jh6JxC23^337Zc87gBvS9MK
zS7l{++N)Al=-DZlI{g|>_b{TI=GnQ!LXg<7-J>W?ak^WY!}IhC?PRp61V9ST(mcL1
zno!7K3SXtk4Dka}_<S@v)nOecO(FK{I=s)z*60lX6q%h-aT@@GM&fX7VhYb5hgnKz
zh($~ml+@0(0ajlHTUKs31}@hWyBB=Ic#n$Rc?G=<@jx(|3go4dZ<qcB*3+)qcK~2;
zLw_pcw7C<XozBgddLn9#2ub3S&q9CVW}BcX5d$#`sm@GCahzylz!Rl0g>V?2l<`w7
zE`YWw<n8Wle*9;6HZ20JhkpC`Hi1lbK|l{qjoS@)jQ9W6f+KqwLX<RoeHri<pXg#B
zWee5|KzL6l;5YOBtJ8Vc_w#wY<CO{iq0S;bU+&PrAwD@2SJGT|tO-X%Iy`{AhdN|u
zBmKcbL@5j=Bjm`>9gpTCzdhu7!*T<DTE@Yk(b$aD^XAKXh}_8I{JC=LthTSrE9!)>
zcXam^?P=O=Q1>&|k#E@>EBP$hSl9ShEZ_M0Lpn}~j?4{mWL86!Oc-W9w@YA9t77)*
z$_cuBCQ?Y2f!~<?TpNxtM@{F1=Z;<3y1|X~5e_$Q+_+w?Bebbjc=dBH_bJ;n{!M_X
ze+ZPNIxY;d(6u?kZW`E||5`(bNvsvaE+5~nIRjx?{7nhz9?C@yd>&B#!+zubS-SU^
zHgKP`nfY1RC75Qc*3?8sz-ctK>29qrH#{LJl7N98Bnw$|sX>^0xuPFetObfxjnsFu
z&DX{j6DQLtk)hcco16;l%bKj7BD>4QP*gay()qdC6O3gDqvBe{IqzzI0?Sg_r=+!$
z3i!vKy4-j$_JGRmQBH_8tJs3vD+W*HM0yN2uifF*MMy(dd}T$qkT-gWg}G@+!ev;N
z!ScD3nmeomg=)MBue9dWgRpV<8qzB?fju0r(VD^yZJq6XWI<d!*;9;m1|z+!1T<-I
zu?Se0D|<Mf+>Cm3qD*D<1PP%p592)XrkJvk`wM8<B!SK^`(<k*1H?sB1R+K=2x$i7
zo!3z7hj-60{hE^$o7b!mV&_k#JKwAdNF#UwOe#iCe_Mev4*MnGl{je={Muh>Nc~F8
zaMJnO{<!5U{{f};s&_wp;wJs65GyP%mFn8axWM`c$ADqCb1<eFJp^}kN>m?OBXX)9
zrxD;QN5BK*GBq`o>Us)F3d!#IlU{TZ*&D7$s~{lF#p(LzwwZ=}0!tf^0@p;_4(uHX
zBRGk<e7<(N7%^Ir{Nqqlt;f#5<IkzX|6=%g>PC}!s9g)+dvx!H#VKFD=eyp!;HpFB
zxaRrgRf}jVoY$cye*u*$4>)_2*a37G#YXg<FYY2d^Lc-F_v4;kx?YBEf=EM`l?WEa
z+E2j+EOQC=)fDZ@LO-&3o6_pjbWiJXUPz45`AamH+<APskni^_-|ywS!*OZz-3EpG
zere4LQ=U&ful7|etGt$pW*67Pe#OM7OsApbrY}Mxb+~t#Mcl;uGyy*She0`8vvf#U
z&1aBG-M0tm4G*$;l49m;-)+O(2OH!PKD-P*+%pT@9KEhJ2`v5L?cV_9)J%$1IvujN
zk=j0UDp=g9pUF0c^K;3juv3Cn`qPs<pqpfTcN4ckVc~VEnfrw8r7VYjKA-jWD+wLO
zl{c<kZ(tvTP}1+i^+~;E5`k`eNxeDfy_w7k5hZWZNm)E=qgXy+g}&Z#TzBC0EUv9p
z&rJlTWAjdRY%eogGzx>236`O5;3P|_IdfA@H_}U+S!Q_Vi=XqjMxhIngd$GrOgXFL
zh@!&=!&4~w$2TaS9WKAMCA^Vn1qohEDaZXyPY-^xvPWmZV0Zy&+Txocw8gN6W<S9R
z%_>VW5L#&bd&TG9bIM9zoWra}=1%xs$)^}*k2%%{KN8LBiesW}Sa#&BPZA7WG*^}V
z?6uGGz9line3n|LO8H(>b$bbQ6coEO&eqYQI2onP)Y|+W=;;GoDXKE>A{+UZ2h(Rh
z%wd!1$cm~|<e#?LI79Y`_#y*Nh!qkX8jJy4RH=kX;n|QY{*;O71~Dui%*gbYbVQNS
zQg>blg$3{>=r*&A<e&fG7$PuK^F%>X@H171x3Fy7+1z<OdA~YMkAnLl`NA$~AqT@r
zgUAHNutCC}5yh4JW?q^Jz}0GAhW%~uOJ!gE)8;I|Uid>5vUoxsXw&nTExCf-o^AE?
z{86O=f1DJ>iO?@~jZ#(~p?iT~8_OcLw<7OeGp*NbEBL@i{iRvwUQMm`(=0xBI94;U
zL;=g|%C+L|*zEB!zbo#=QV{Y307^eDE9vKg+|sb%@wC|2_9LJ^N%S!!WC~f%9lE#A
zkk{h^ilpAL6%q!md?U_YOz#NxH`qX_Ltny^%*~^zPu1axd~;Oq+q>a!=9UFAepE;3
zf%i~*;+o!to2msGx!)gwo6bmT*>);)d_)m^PN}>Jw4)_}Bwb%bspo|sMzQ%7ceRbD
zYmK+>3nuWHs<IY_1wqwJ7~QUpl7sY1^Cy!bp^s;sI5VlAX73df|77ly_0`7Eh;4oL
z6Kt*PHcB21(nRirHtrt9X17E^M}T#&Sr=cOPSd}7bt8x-Sz(cnP(`~w)NYaS6;q+s
z;JBQ1A+iRU5r0Ak@^xQ@j=yaIF_bZg^IzrMX^35w$y-<D4&O9zJKae+qW*e;JQobe
zOtP{4U(~&a*_EJsw50Fi;u9Rw$3^bnP=zm_pdU(;p>Mtg>|Dp4#tiN@*=nq{zXm$R
zZx><pn+<y>ZdnU(g$$4bao?o*+IHH3%6t%np6JQn)L`wxS07W<9q$mElk8N_K24%7
z5DZ76PU4;0M<ll1&-x#=bRc$KHmO8Y;M>pbMnR^7(mjWD<UYnrK4bNvcUf~Sq6-QU
zx&vpDC}|HIkkYNLy^imQu1Lteb9FVu()#tnc~IED2X6NPKmjlOpdn&{fLt}5Dk^$D
zXfJ)@KrUuI8iD`)xNyxtj$+*rqBSlCp<??t?ymo9F7~h`TyTAQ>1&4li+t-hS14*K
zB4bgKJ%WKhH`~Perh!yYbEO6l#gaH6EJO84p_C!+EtP%8`DMGM_C(@kwO?~u3^H()
zsXpV=%T9*SnEvHFm&cnkHHSwtE-9i7;I_I5>Yy^vKx+dwlH9Kn7OBmVgtE_YQr>x|
zruF{_0M_c94<0FgUM|fMFq;v{*S+^Ue06e#8+zuCx+3yAmw4n?Xj&!Pmw7&QTAe)>
z|I35*jxZ2|8m1{<10C<=jSp~~qxzYQH2fRXCRaeQd7dujQdw!bF4l7JP5O>;-m>?*
z<Peva@D6~@Hf4{z4t<T_p)sZ{yYeU87u@-tgN%j#qr-hyusE;bE~jA}P#!FhyQ_W@
z>1)$xpC`p&Wr3HZu_KL;BI_0ihhbDo?72*fUUg}QX^ma|m2z2fp!$`zXzQkS<|0f9
zrl_HmAar-#UC4KxD7RF0E6=;IFrZ^MZMSfz@Q>Z@i4xnIfvu`N1VBWO`si*X#=kF~
z70Ic1a0C3>8nXn?HmgXfF~|3Je4jlYfAzG>PhCJAm(g*lI3Q_sTi4Sfq|j(f$bVl&
zZu`c|GTWfxbX>>?zktA(cE3UlGBs=KL@`QoB>9-|-aX-&o^vSs?{Vvw&f{IB{-MCD
z*PyrfugGsC%&ebRI?OfSRo(6TH`Dl7rwsZIL_eZaRkB0kk7VZ7$d{q;<@Gb@%A)J}
zL~0bld&l+xa3h5Kc>%tkBD`gIcDKrRZQ12Hw#y8{9vIS%oxt6j>9{Mm`3f8xPk*MY
z51jH|TTZ$!bDaTSm7F-VoHz1bJIH8!ETCBZdV92RVEguQjn*|z$Q?$U2xfqlV%F9M
z?k*N?<`yXQ3{Z<F%RBx<GYWLZx!Fe{a5vny)2lgq`s|h9)_VTlj4^p7L)A~-v>OAj
z9I<icT}EwZ%Z)hN6`P<U?ASw)$JFgq`IUEmw5O^3zH(`8gtFr}II@ps8(TirZaGhe
zVUK=1rGnRff`5N&l;H;2L3dm@I;JlPxs(f*m@m6Oe0m*6Xij)0r3FoL0y}^BxJ&o8
zC>xoe@U_u~MtAYpk+r`zqP`#zKIWnSe)I9M|2D6jT1k~XbPB@g=)3G{R%3$Z7HE-%
zc?c(%4(8G&@8nF$+blGF<LGgwGc!q&s^;5CdpW5B%&<GvD5+#fmtPOw%T||xa0Al|
zMWA(hA%8Q*AYE033^UD4Ub<S;w3)Ysbg`y4UDn;wJ@}IVEQ&(Ck}jw&y>B=T-+@n4
z-_D<UC|;59$)~S6_)F1oQAn$YYE@KKjF;-vZ9CsBLq3V)hG9&fcbZq3waz}HW?}+q
zmZ<X009xgfyC)y0{<(a+A#={=XNn^b)aL^$<^WL61No)a%l1D^M@a{TmYQU!@N^Vi
zQjit=R(rZDH4LU^_umV};DenGb$c`0U7rMmWI<PTujlakp?gYF)N0+&nqvy+Uxy%a
z2({1l4Z97kOxAU6i)j$Ahln%(+q*lZ-__?H_wT=DMld!v`F`vi*A00kaWQ)^G_jz4
zw^{3^Z`rVh3GLGm_XT-yTfnC_01|D&(WGTZkCUq37!G)<xxVDLo`}I_NK2dbfDJgZ
zJ#A8WN+%%0B`L$ZT8}n&jDQ?F6Dm*j04q`pTHrm!!1FILjiI4!`s3X179;#sOuKi7
z@>brn6xr6CThJwIV^u7N5)fJM`xyF)_sg?yM#)a#s7G&qV-%9~{IQ~H$;Et%U9J|K
z^t#*S$$Zqe0l=wtq~<&X?QhZkS$rSn`ke$5)An|2v|vKLmF&U}f*2F#j7<<KNib3v
z6}x?6i+>}rh_~|H2u8#>Ba?1t=5+#`t$y0n6${OCv?GwLrtIyaf4lnWiI|wp?evd^
zA`f$SD4QiAOfF%)uM3xS!pNo?u~#gL|85X9dnnW?CrlP{Vuz-TFK|V~*Vo7(@93wD
zu}XDNYJeMY|KAID!sV#5s`Hq4g_F;wA4r%kBxCnY4?=eSMs|#Zj*I>AxOB_}i+{*g
z?upl~oW;Z-V&U{m{HERruo84m_Z$yoUP`n3wuvt@(nn)^9x@4p&E_(S(*FIr?O7N6
zTMROF-~hKiT4^Zh7PIco1R4U4vV!__HFhDT(v980uf@pUG)Eun4rO`Yi{IP>Z7XPL
z9cyW6aPr{Ht&FBIcMxJHvb$|dxhz51610Q&inQSnUjc9K4lu}F;pMZC)q@5u=N_FQ
zbQ-Wdx&W;lt&ad_(bwRnQSGcXBg%%+76>V!7bjtl!pNhW`WOEm@iuYl3Ji}!tXBS1
zC>5FIM!)PGLaH)s?3US^$0yJP$6mQE$FQaA33;NKU}8&okxkY~qBs*vtLymT5Eq_X
z4N}dz?)o<^zHtJ?R5HE|$%PgENIAX3K|Ww>N@A@8iF~1mT)H|sb*0{0qqRX_s(8+X
z+Ov!WO`nmvwND!oUYDSE6PQ{*IQQgtq-K8uW@lleCKMEZR@LoOcegG*RTphLB+F?9
zIVZoM8G}W0K&>y4w}Hdfbe!o$43Am|Vq_^e{Z_|(ABzS&r59jRL|cz)*6d_c6h~Yn
zp_pe5Tr#`UnI$2>{IUf=81zZg0FZGr<DjKIbUw_1m7V0(AYi|<@;A>;?2X}3-rjkb
z@egh;N$6>JeJvo*oqEuBIW~w)V-}OJKk*(oxX@R)_8v2i-%w&VxD4ws5YwZ+TD3E_
z0&!>cL~h?za(ep9IpdhDsRXj}*+;`s84X;tlhr&(_bg%j&0-k#VDP>|T<5rrUj{uR
zFRMeZ<-NT~-ZyLQ$~yjeiHAJaH<Ppob2mmI)alpo{muC&vzxkIKdS%{woVFxA9@rz
zhcokiE06!c84iv4YkAanvw@1-jRKOX3D@4{yPaI{@F)v;?VbPRKgWnj#@H)*G$|sP
zU+IJ=4IZ%O#cc&xEwE!bMj6bv9G%+nK2wY#2_%8FP?tRZtG#x?yHmzfKVAW9Uwkw(
zAou4=%kH+D!KtFghs?+L*(aWzaeO#gsrWiYs#gq4uQRZ(v82W&%5GvTGz^?-p2HlR
zu}3q9kz`e$r8M>sjhV5{Nmt0pfF}&P?d;U`g4%7eUEmuiI$6a{vY~|pUeZgU`4I*6
zy3h}Ymh8Ac)7Upo5{foradC0ix`n;^f`mSsS-y$*hKpd(y>bN!saq#0tB(R2c;sB2
z`H_rw$VMNNZnK55QuB*5j)s@4ZWU^wAtbXT;?d?G6&Q5w-rvRt-7QPIDv7yxi_RFt
z@mZ5g_i#KBjU~sozmA_<>+S!Q?5p5P8CBR*GrH$QJr)9Opie#bbORYP&E*qS*eazg
zSf$yBtFr+w&pIW^+Ey!9#Jr#whPj3X9KAKC6c{+qBkI8=RP$Ae_;1`^Zr-b4U^V)C
zrg)>h&rWGy)q$?<v>plM$UHzXTiSely4;!G2m>#fR|G4TPe){OGLv$cy^U2m`Vz}2
zpOnY)hF?PpY1vO1Ms^I|NZx9GMb^<@KKv)?Dr6T1O)`p<7eF&Y6;SsFS{8+9GnsJ!
zjx0sL9DX#*jfU7gOM>6f63>LWCTZ>k7OW1eqCtGfY4KjCWd6cqNs~Y>C=qM?y^E;x
zE1~vEAk%Gn`@`#|Pt4ZMtxPO3cEUKVvj~KEG(>rDb}+m@z~Xhe;N+6dryPbb?@GCU
z2z*zO-2&~fC>;MJ2V6>8y%_DU=A$86hA^|816weCx+Y!ewBfR^Ku97!VA1hqtjk{=
z5Sh*C2}8&*ta+5|WiIqrxR~EP+TNYDOm0b*TLNja!GQc)`tW_eeQ5UexGo^m(uI2S
z;hD0iWaGWn-)FHm8KtwinwXoym!`t(U!Ig^<>3&PG=#M1SE~VaB99jDyBepCzI9m+
zcJHGJSc}~F@pcP!D_HJMC*{Rs8lQu``a2AHvO(ic1x$sE0d1@1m{ky!Ymr-QrJb+7
z;pRU0L6$e~t4{9Co9?mu)tUAUqtFw8q}ltys#1m76_$4(%BeTWWLpHS7D^Hu!P)<?
zE`c8=FFM<cJz{oA2}&3N0r3C?8!>4LwS9y#XNs6+>+XD{1t1V#wokp-94Ja3@%!F6
zu88Y!l{2UzfTQH>hHORjyLQvDVI#JTJ^4j_y`%d1lnrunheN}`DYt_bWCaL%{a{9U
zKQKL4#Egx%ky*Nup)lrwhc7wUzauh_I8N?7UF+0KDo102St6ma_YOv)PHc#*bp%NU
z>RN$(^FgE!&_IK}-zBIU`->-^7T?Nl{2udJeC#cM^MIlB*z_J|S4Tc>AE)7SB_L;!
zcA|%lMQG@XLUF-&ZOj*;sC*hH)u6$^t4w{5w&N$V8a9(E_DLmuAl;p3xUg$5$-)GX
zlNlGqS!Hz>1p=?+rW<?V_0Eap1yo_M90EB6s9V)+V@{fCkM+CA?1XwNQgPe}mit!a
zA3-tH0rV5tN%_2XK37m<Wj^pbNg=^XJs+IU4vGRc!neu`R*>{n*cg|G(q*y4Z~k08
zeFfjUbqz&uW!lh6)vGH2q3iI}*Eq(BYsmVv<l)li<*-#zwY>##-RkzjKR|0j-;PG6
zo!B!&67sf)83=MUJ>EcWQ&9ni1jL-RXYh_92|pIn4oG=yMU&mnET<jWNq2*Se(CO;
zfi{|AH7my4e;%d<`#2PCEhLb;4A_$quNN@8IZ5BV3PdL^K0bjZ2c~1&Hu7^$6MpO6
zzJh{*%;TMovPMUnJ5$U7v^T_Sj)itcEa)8b%b3__%VyG!=VjU}M}u=;<XNg*_qnF#
zT03mTsowGr8*_SG7txT2@I4k2sOvN=HUVY|iI|?VXg3|Mr5puVS2&v?A0KMl&g}iG
z%eo%<TQQow=dwdh&wyh1i@hi|F#_R&ox-xg8xF|SD6zqWspPGYGw8Q>nI=}*T1;5F
zxMVHkYHjs9%f^c!(9~xMVCRr<19_o}Ue)BFrXC{pU_oc+Wk+gLkPT~uLr!CMCmf>f
zpatA%S88R@Zcyj)?FLlvF2k}_9YxyMqMr8c42iA&Z{i|C$N0S}LA|T!%V??@vh$ua
zPszRXnsC`{q0`xk*8igSm{*Y^he92<a)_01uBqdyVB!&&?J7!LNqUa%N(@j&a0GPX
z4!+<J>le1zl6C{Pu%ij8yVG?!Z%FqgOF(Koc_^rX3)1B{Y@o^1&P%7Jn6KIJc0Rn6
zOfYVg(0R}(!^SUyb)Ypp>v7K~a@}qC6RnEe<sqI#n9yC0Y>qpp>}pj_SH>s)=3hE8
zVpq8J<7v`{iM8H(x-Kd4;ThV@=SWC0$d5(Y2NzPDaAGq_bNtd_WF;ZkE<fVgHg^SW
z0J@JI!Eeh?c3v<zmBhR5mnGQ$-qFaIUG7>vsGvRqGJx4)_AU5r&`YM)c)n0~18;A4
zw)|Yg&yyq8U^~C`i&ajnN*pz|OO1pCte;S%@1-en)e4qU&PG1?tX0il!Zw1lp@7UI
zUR6lXOxGYBB1yesf|jG4O$mP2QB@ZRs5VW)^EcjxWj_@X*rCSAd1WWBlE6_Ht0Yi2
zbC1kZ1*P04ELfvRBQ*iI?Yb%N-^=a6<^G#<m_r{=5Hg0r@+kQ-K+SbdPob^bXJDMv
z4dM~EwyS|O8d6kQcD%B7%7RY3{^>Tbr^sUcHpb=8J*pQlkz=<GZtkm<xkg0qy62X6
zIzG}(Gmk7d-M`dR@Pv93<CO+2mg?~q%5E?qC$8zQ+5J4zeLDQ+#bbRspzNb*NKV-g
zHKM^+tpVYEL=XvE&50mY#M^CjKB{IOO~CYFV4j0;-<M=s(<9JL;$t*SJn|atRkLPI
z8a<_|_|M894a&WS#y|PM2fd<ZgElg7Yk!i3sJ;8wcWPx|)O<DxABm^Un}X(UVQx>x
z#^FY53D~qb+#GE0bwS&)k@UZB<4q1g%UqKSh?1a3$051KaBTj+ngl%0R|XvrNr!=z
zeF)Ag)3i@i^)j%IHsir=K8+#N%9U}xZUlQ?q%PxvZlMnGc&~&I*8$@V;)FW(lO03Y
zn=Uvx6*cyon>1io+aZ1eR5<B<@Qw%hcmgLrkojIUEy$t$QmEc0#HEy<MsxycNR=Qr
zTu^$bSNi=|#`y4Q#B`4h-~EntocEk5xZCQzP101iMrJruVpT$<JINyI+=}}7a{fz~
zihrc3oY@J>Tslra9hej_ba49k%{#{z6&7@C+>5osO0(6%+v!9v-w2>>xGXsn75Dr3
zR&w>5kJO|31TCwLrc2i|xUux{Yz(J0s@`9cv{k<c*cKuy_h})zYo`BZ4wI`=D>jY9
zGPGaCMq8z_Ya)fzU6$Vz_qu$M`Dw%Bq-<x+$4YuC@X8I3!Y3IJuFhe<nE5W8FGuVI
zPins@G_cJf5;dfX=^+n~K*E*F0iq97pFz}`s(O-Olmm5JcAbQlJPjdHT9UDA4CJ<e
z=tGRA&zxUhMinPfmH<I=`Rr>-S+{^&QHpzFW?iMN30Hd=n`pGcJRgo>;F#X<-rtbH
zlH=+$2U5OQS3MjCz%^94PfSQqwM_CYn}C;iUne}%#S+X|C(qMnN=$*qqOq)!-=k;t
zdYAX_bR>YX)P@@TtC`?=Wr(6SyZL%x8#hAZ?$ajCM_@N_KLoi?VY5PI{;KNVvOWoo
zCXbisps^aINfi*Q05;e?*KclC99h+o&pL47D%wm*66znDlsJJToa8-J0$6!G97Ach
zBJkoaHk;S+qh4IV+XID`at9F6(0kzvpt8%X2$@eBcaj#372fHn*4(>EZM0Tnp-#?v
zJ6DV@EsxL!3N&SZ&ER4wcaJ2UzQwGE4E|I^HxiL5W%u2ViEE7~JYy_<)h;ONLfuCC
z8+nac5m`0;{R|Ge`p}$~GWdgt@76GoJY*pKLQo*h3PNbytu%c;83K@3&kV?UkXJXz
zr}iFi-}PEialFdb5R}l%-E2ibF!G-zBFGqIojl`xvHG)tHFJ;WsQ$h{HOQI$BXsa}
z@vMIFNEqf`t)AY^jiQeuWq`!qUwZ9}yft6*WT{bx`m;5~nbsgud#Wz4V9(uCq>XF&
zU%@2W70jPqi}wMeKu93xHFR5chT3!!DS421*_eBIpA(^x$sc2Kt~um~fuzPSX`*i)
zop}HpIZMOWxZn9k|4F{FYUZ75ai?r5<-Cr#!545Nu(si}Nlq7GGhtOZBhJ+3LsV!~
zn*)(Gf-_ur@6baxM?OaN+@m$$5>Jy|$0dKT3l)4~B5lJ&n|fF+&0%+gmpuMHB(Y*F
zxAEZlm7*-$T3Y2_Qfn33v#{>$v$U9CWSH@Aru1F|y>NfPa*ZRH)cgz#VAaLxE`#IG
zW-EThN#YB(&M+bqo|~%mv|W&W%%^Vu>dI#zQe&XiSlQ~ohxP}LZ}b)DGiUj3QtXJ(
z>7+1wGxr0X9?BOM9Aknh@qO(nJY>AvJPomv-Wiz&=QEY#8HH4p1<A0UySOpIs%<7d
zd93eMKr8ri^6DoCU|#xt3xgz_``51!z3=zmk6q4`i9v?FQCxEeyt*v*Z@y>!UlnAr
z$BZ_BYf1PttT__cZwjp7_T9Oc{@Tl0aQO*m`e)LPfmRU}=l|JjI$l9h(b<GGLR#DE
z$U>n3QL61xlgSNbl_!Y+1{A@YrcFt#6NHHj+ec{=Y-HK*GA=8Rf-4Oz7il}x{L#j!
zuv0h*`dc%mmMeinECsP*Ox|8Z>5)MA&NohD*1}JcP&S`H0oVy)7|tgC-fV!Y2{W!)
zT7G&|bGS8_wbyQQ&kB}*(+&?K^})F8O`#qw$RZXIdH|Bjwam(c%5@@k&e?BF<zxi&
zA3&AN2#s7@c1#L4i98bxv3A43{~z;MYPDfXxDA(LloYb}x1fz9VjbJ*wHu%iM{_SJ
zQb7(J`(vTTi^;5W^YdRSNIdnRbGraIZW?r8{dx}z42xm8Xh~}!)R$?FNfX?wa(_XQ
zF_K;#H~`F(+e$!`@-u0uIPTR04u&w11V;a+23>u1faZUYNx~l)iB~@_L2+U@BrYP8
z<yPP%)=-4KP$dT0<7IT0g><&N`oE4j0)R(k9!yXGX7)N2<vpl<WH!NguVl^m@B&>s
zQ5gy4*I63v5#4r&X8m#~!#^qcexi)izohWROp|q>8UG&i>#a03!xfn{a^=}EYP9&c
z6A)FL30{j;c0uVwY!H^Rab&Vtd`p~l85OkBI(A9RrrP(ka-&^cu-flz`svu+zhL~}
z%pqx5`0XepfbKM0quG?P7uEv!s$U#P&S|V7$eue0DXRBEgEb<?a4&FZ)b41Fhw4=c
zhuK${2Nm@&vho}N4{U_WO2F1a8y3b2;Ay)Vvn(Lri=v4{)>VUa*79awmwJFbr9q7?
zXJC+j>Kxym`K_;20FESpUy*|Ix;`x5r7ned%0{(J;%<e2I#4E)D4<ib2#%(#@fU9*
z+Qo~Wsu2)b2x`fHJG8@U#}!_$5H1~YpVb_01Bf8*@zkIH|9^2*ny!ojLI4IqrVO|C
z%fI)Fzfs&rC8cYC+KAb31uTU3U>#HM2WPzFbB`&vP(hpT0V$Z4rpfti$dBnmI9c4i
zdR1FMr1bjNm@y6@;`;wzFTE$DG)t!fkYVrJUjWIG|1<Za)fVAMzhym}g=a<3Qy5sE
z-EJe*eDUD3Zp|GMmW;+J00oX{3Gg5B7WO@QEI4xJK;BpnkIX{Yn_|9^%5OaY#qA{y
zd`%7555apOQZy;7d%ovR+R-Hla7D+LcO7om1s~o4`{(`rE!XovsVPTW4a_PR%mWXH
zFd@VPOBP;8Z!ZI<5zjtW(Pzsr8jiJke=F{o@?FeDG?}0Jl);+A#;5j*F=<Kx*jRF@
zap#3^hVDw96x`8jf<|L(;2#<OPfK>H$3e8j8e&3;U5!Z6ySS#RA|Co?jS`7F5ZMvD
zy^!JEsuS+w3{;D~;F7dbD})PHB}2I|AfXA%r<@u|-V~y4M&NCUkWRr;G#@g)1HB?T
zdawK=<-z1isXsD|x#DvAO-OPtHFKJ7*GJqe;BZ~Lks1}W58AtpVw2=vLd!_IkeYvg
zEZ><%uK^dTL+$+i1!R$Ef{+@zFLMBn$SNBQ_RL_Q&I`5Bq*T)u!f{?sL_vSm)JV%>
zw9w&E<&!i267=2}`*y{u1_M9L-<le)X+t^M>;Y?3GAz(YlFfT|FXLz<XUtWOI-4;h
zw1hU&B{3!421t!8*NDx#fZU6R@9<xSiy%`}q!oTvS`1QOFY?VO4Dy)a<*RE(;z{p<
zV|T^cH0KP~jHaw+hHm58tEs@^-tKM(usoTQVZ12wVl3W<iO`Nivufk6UyqcG=+%-Z
zVHz1oE^*W4ad&_NJYG<RJy6W}(F$MoadB=?$_QbopH$kV3}t`6FO6r5_%=8DFVP`h
znkt*(j~ymzgF~M~`}6TmeTf~-H8tRSng^x?Wehbm%AG|>Rcf%}TG7Lhayh*1%HA$N
zL6gW0-gTBmMY2qFJ6Q|d{XV|dNf-c0iJNq&u%r;`+i~wAjQmY1NtE~!p+WpG-foKz
z14%*BIJ1z8N6I@^ePwS%chePJt%yfO6*)rU5M3ykiI@Su+<4NM*ltNl35!Dqhy!4^
zBKX!=@yymxs;333+skR}*(*W6ze7HZANm?qa?HeqgfUfM%PgSAKs^rp(!qFJBR4jt
zPLkPFa2EG=&d6Qqp@XT<>Al|-ck=XVR}rMZELr=Wd^74?gKqt!ucx_mi!**)w~+#F
z`w#Os00^Yfzw=Y+D+AQnp<|P1#1Oa#Er3X<58#o|mr41cBcnibtZb_<4{LOadXM#R
zyc`GTjF`()_I>aU)GG~tGbWQKGql1ld1HOUxWM1hRak_|>Brzy+=iS4xz;^)+jB5K
zslRSwzA*#`la@#*WU|t*U|FuC1h0hF&)VGzd1FEo%da5K>P41UT>>CAet!Oit;NA>
zMLSI-<tj-eVHaHTI>m*BKU$o@W{zB^evt%PdoOfa);6m?Ee3fd)j_yV5`{1e&Tnr$
z1Y9M~BnB&pEWq&f@J`?Woa?^eCK%`nXd^C9sR0VY)hbW%6o->fjPI_u95;E}UugNh
zao#@8IOw6&<u!M4$w!q}s<fxuGlW&C)TA2$+e;PT=ceyMvX5Ch3v00I=}f-i;`#XW
z|Ed(|>_;~NewY;U&!Q}LW(M%?W~Sbj>5E1VWgx;33FuSo_l%ZQEjtu!4(>B1xnoDA
z<cW=^?Izb71xSgZPZ9F+WNbnUVGuEH7954KVzuB=eWgkYe5&f_X(3P^Mm~6bk-Yml
zg(;dIMLI<vaZHHaumk8;Wu)L>2+pH@@j7>_zaL1maWB5f1`jhnn8c~ue!R>{-4I0z
zBp}%~BZTQV207K)t2zE};(*9|mn-9(w=<EK%$mG1_3hC6Fg3OE@0oB!CcK?jx%D|n
zhmV{)NuT7-Yzk9y8<J|}O5bc)&I$Xmtc_}Hm{IAanGc_2pT&8GM2Tbt4*K5a`|JPD
z>-x)5^_2wW`wLHBp7WM2d;tg;Pp0mnG;*8wJ*~F;Zq@vxexFexQaM&LiNZkeco|;P
zh!=l?XfHU#?rb~}TbI<xyiN?$;p!}48?h^P`A{8p!M`>!K(p6)Rr0UoJI%Q`jspNt
z2p3TO1|r^!5fON~mQ$FMYKx!>BOa;!b(AG<XtMrYV_B`|_k4h_zMvu1UeW0I%@Nt4
z1dyMfL2`+4Y$n~9vhRvL2DXC2Gm)Glbb23^aG7z^HpGm>$mwwWatYWyS~-5o)}4jQ
zu~LW?5-958#djBJN88b@RlP*+yh$;hoF{2b8R3(O&L`VoT*Ho=9x@Fwb_x)EL|jO7
zU~0K5d1koU?~EX^Rs=;{BF+u*jo@66BFxgw2SP+BDJWQY{(-l~izv>kDSy-(#hRGt
zGhPcQkFRU2g@9%|p7&~PpbA<0$;qqe2zHm2=-0m%48kzgl?sZ6g%Yy=T$%Zy#UAR@
zbY~$=DxHHHfXby~RZF^0U+iec3?Ut|;hNlZ>8JSV?CTfBpXjhdB{TkTULR!p;c!?=
zP2l6Emsp|ekq)CQCI8maGTP0&eaG7P?_O^BX?3{&m7ItED#GN1A|dTx8ZYM1f;{R}
z64lfIvbDlu5Osi~PF)X0Sv!c(ol7$Vtd}sz#bn57q2%YMA<$PK90_Ra=iib<i_&JT
zkosk6gUOR#nAvFz3|pDw`CTIcN8eua&1f@&a)oR74x7%&iBk$1;*W6+GazV?5eeI(
z-G({cxx<(bxcjP+h)VxWXXB&}L%2vi%wvq+mgB*5Lv)yN1lu2Afd`+Kh0@jk=9Uv?
z+}5+ZC2iNfZ>z54Q>Qr2Wg(Q;m^70-^!ei8&C!!R|LV(`Xq4dowe2AjDOZy;OLnH&
zne|r6bg>mWC)C_GtOIU&5<c`re+)0RE3E^E7Z$}HJfJI=y3q@D!&%`w`a_?*Wy;a^
z%BbUGN3devl`n%vwr6KWeO!UhBz^PDgN5H=#>IWF3W}0Oy{rp$6{&(85hIp*(`M+c
z>pfBA*TP@#A5L^R)%kr+Q$LxvHC@$KgG)%TePq5p(T@6cnAz2?vH2?+@^Q~;*|y{g
z7Qf#!O;^%x=bPx+Bpg&19qy4F`!WIv9_325uo=_>c4!(fo*{{nbR^k~=~q<M)8&^g
z{KaY<RDCSoY5(`3Y^P4;&)A}Dm~q|>NX?M=Nx9jqQ)}?af0sRSf|V3gv#}6FsCqU%
zwLjXN?7jSKJ6n)XR^v<pu#AP4>#y$$QJS~M6Kqqqm#eZzDL?Dl|Mu+Lz#F1zLN?Hr
z0pDoEg8|&DfN@3b%eEp?+vt<pYP{quc6WrG)tqJ}sFuf_CyMd{H(mF4WppCpXw(CL
zaz|`5;Ujhj8(ukw84bb#9oQzx{z4j{Ctu*-z75i_z)w2DPkQ@r(i#d(e)J*^MR*rx
z<@4oN(T{+&v;bd)be-k*+u7<Lx7E|@TapiU+I?Rw*Ck9J94e#6ZuFiTke-OH)80mf
zkt5{mx4A}n)-P0-wf_~PQ`1f0K(&j(#&tu|h*2Cck134<7u>jsoM#V>hv*l|hjEUq
z{=Tbe^+um^x3et~Bx(^6$oD0+=~iU3_(MZTN`)=v($8&d#^ivGy~r~kKDG8km!yrI
z*WVW=xDI`%)D5xak`vz-|89E=n{zY;olZ~r!P8YzG@ma=AhLS)J?y=3w7;~zB~drp
z(l9bdxtb;z^_}812=lyoROJd)O%haRb@;|)(T0qb*Xsv)KyUdNbMQkNI~t)0Te=6i
z^oXq2a~2>esAxW@vG(o9|2i}iu$4*`V<_6))7CXE8UD!wP3vk^bh&EiXy13aMshxV
zGI=LpZQ1TOZkgVWBW;a$xng_sxpj(>sP1hexBV027fY+9<AufozP589MgwfB8XOC@
zD?TL$_CAy(ofL6os#=pYL=bkO1*L|tYku4Jjqa`a88P@RPq?B44bi?~%Wb&mT1m=L
zTFcJ1%6K|m)e(({0!=$@)_Vm0a==A>mlMLAknI0;PO6E`4;z?9#rxBDWRN9GFKk5)
zpJB%lldDa`1}g_ESaqd$KF@0Dw{05|%r{?Zt*}{Extuw&nlQ_XFT7?ppKNq=S`v`Z
zO`l%-g*;O->32C&SM)V2X!qU*b~3bG!JKuC?g{PtV4_kEilzt8wv`9pb7P|V&+|S~
zy4Hwz0nR+RMj5T6ljU_v>dRIEoi$^Yl=Sm{Z`6`fUDQUL@nuW0w8er^FSz`>SS-hy
zp!u&U@YUt7!Xn%|hhfIZaoOnTkC4_A9gfxNB@gQA7N*6U=-q%O5Tr0S6Y-N0(e2(}
zeP-2RRp&=0?kBLN2-DCpA`_Ov+8f$6I9EoSnbya-HmCU})JlHh&34z|q30WvU$+7A
z{cY-&N<+`w56(9l5qzoPgZPSiD?VIs_0M#*LEC^u1*#K~iT|t=D}GP4z_I<B+SqM?
zCcF>W4a;|^**&1xDvehSb)e*ylzgO)m@66^d{wuyY%D)_>?Y>F{9`h>zdYdba$E>v
zTNq8i{Y&Z`xi!`k)Cx1&Q0ccGH(|ph9E|OIm^4<l-~*pGJZk1g_YA|yLuz^A|5wwM
z9suxDLeI4u>v{!nw+5L#(<R*j-4Wf%3C|#kP@8~V%ggNv_TkETb(cc&rC`F%F#As+
zvEaj<KkP}$>=dAPkC?-0oYPRoO^~o-qlOvk$6Db@pTJF%UX-MO6~}M2X|tbFr&?m$
zK(s9ApSv>HX>xvTh&tcf0a{oKP0<|pfmtqKB2MWJI3^W=KeHx_PmF&+jKOnRES#Rj
zZJo!yW(N^gSptnZ4&&i>OXHQRE9#?ZYFk1|bZM+|n3b3fzaUmTINRrcNq{c%+Mg6r
z7Sg?DC<e++7qC8vGT*LI%9u>gbOwq!D&PMutIc|a`OYq%&uD_c(k`3h_d^wp;DC93
zmy_10V{H_`^1lNn{x^v;#Wq!IFIMe5D(HJa{~<VD5yqTfOV$H^Iffn{k-<f6>;^S$
zHRPZrr2N?V`w?2Pn_VL#`0J7KnORvPK>4K9+X@C7@Z))E;)6x#MZfZ$4er+4b$v~7
zj5Xg9u9NOVTLL%>TyPpLM7{7rukiedzOjzwYohdp`OT1jS~Ad_tX8KcgKf+A5<~r$
zTN05Uu(qq4=%Vcz1sB!mn-!nh9Sv@dzS%?kL>Q_{{*mkiSR%eLigx3&L{Okz&_}{a
za>wNhujlFlcRj7^f`exM{>t&`PgaB9^RK%^E@ig#BGxKoHObGjg`RWx@{s8RxZE4>
z>J%SV-xoHTvvBv-FXef6r7{G#e@-p9I@9N>zyA_=bs&=Yn6sF6?CmpxT6|+Mkh>TO
z39OIXHhj?OdatkAJr|&z{;}*Fs5DaTPH@*PB(-A_!NUp}u{&m5UHhrcQ!*4Y^~-k6
zNHigu`=N`VY6~e2jf;R#PfEgfo>uP=)tJrJ9}L+)#5_A%1a}H`Q!m{G^f8vUcCm)x
zx{we|Hu2wx|9lZ^siR8XcUP}g|H1H#3=VComdfCT;seBq3<Zv5mPiM1={K-`;|3al
zt||5!`<DYoy}m-*np>%rJP<t^YmUfN4848Mf8(}sTBa%C*jO@M#GRIl3&)#U+GHTB
z1&!ayy{)!Xk9qrVh6@5;H63oJs(znU(n^<EZC`q-o-e%dqw9tmfTof+Zd7W&sUsuw
zqk*$9#U^aZd%pJp5ZZZV?Gq2C^BMu9UV#-Vw6wZxkzr{NJE7}d!TauVLr#0fW*D5L
zRZE`BFX1s215XwmN{vzY;s7BbKOJPdSYC~Hky`0!tdQEdi!Z=~W`eTX#OPi1q*CM*
zdY(4i5r95LUgDR9r2PL!(sc(y{lD?goionJ77>RtQbJ`#oDt66vk(eNcG-7UMz%!R
zqe98ZmQ`kDW@K+dX0o~8`@X;a^heI;KJWMQJg?_9DwT-1^ydfRHWv~rtUK>(uU2;N
zbw~G}FT8SX6mZQX(PxM*PKv@J5l^P*-?{3~wz=9LWpL9skVfLuibNmaLFo2TvVr%T
z5^PGDpNkmwiqY*v8_D~N{*wVWH{K+K6<xtHMh<?wTOa&~Yp3A)>Coq=s<9kp#JMs!
zsfu3VcY!{>1bc>pqdK_ZzvfNzSSb~?&rYWa0S;8w!Wv?Zubg47AUxdy+JfSPmBGNf
zfRu>N-?+Zr@A$~|8tliJg9KO8QrGRz>MxTKN`j3q(hx5nvND*#7tw(Naq-ZPP-1N4
zc)ObM5og<RNOY>@(=q;upme7s-292I9k<T2e5E&EYmXhe)f=@>Kiyu{VW<Q=Hjqy2
z%7bNUO8R!iDSnTCO5O)Xk+IFP^(*4*2>%E=Qr<5?_aXu##0d@W7{apSEr_4Y7Tz~F
zz#$QvFDad<u({7tj?@`YD@d*FX^GL&H_Q1H^eN3=FH-BK+1ve>g>ZBe7_mxB-z1te
z>ZxZx$~N!z=!AZB`b$X5cSt8%$M4oNvGKJM7i}7$e(r{=$1?%Podj;hMArKD1}a-g
zAoHQ(*U9A2)F9=?hCw5jouZNOM1uYC)KZz-M|orYqA3C4YcGh4RG&qH-W%bEFb8ZG
z2|h+U(z;5ajlN6xlwinQ_+qigh$U1p*oL#v(j&f^|FFJ6C*3W0_X%;eZ207IOvRU9
z@kQKrWGaz{P5XaF8C6Eb-kq2|u(_(VG=aSd;>Ex&@;`QFj{(*$qWV@mMnMt0-(u-H
zceeY@O3Ap-ds}g9gHM6w52f?^aCeV)k8gStP3T=Z?H;%NxeY}yL0USBcS@v*>JIL>
zOHg-NnwAF7x!gyOc%UL)Gv5B}Rkg*Kd^ous1q75;vBm^s@TuBOy0^<OL$?7mpVN^B
zalo5}G<`xuzN`+tUk7YH>cppm3<dlsvP}D<j(?PH7#of9=!e$dUR-MVI!|0Lq|WGF
z8WDC1c}M^{?O^zq<Fh37?1Q_Xvp$#f@*06QYpaksHKTo`#QIex9D>C^BX16QDd3w#
z95*a&p+9;z$V9s@L7dT1{Z&8eDj7qQ)KHf)Wu886<1-9SzMF*(Vi;P@e~)fFpMbc2
zA%5YNF|@Fqj4@A$;FR?iMH8%Wr(SA-35CnCCw~3n-%YP|(wckg(#J<^SvG;-!=P4;
zqk!rccL>~E31>;lIV$SD#aou!Pu0%6Xc-fCi?^PjnLNF<T%Dia=sjCNZ-^K0B^g`f
zKs6(~rUw5t(1DcrjvzN}drCZYQc>PR4U7_LqFvNCv{5VAY6G~$GYYey=6mYRpjskK
z3AMLz2V_ZmJQ)2x?bD5>oDMuL76#jXI#$CSDn3&W-0-+J>Jj#d9?6Wa)yFg8v!lYE
zBn+_M-JCg}fF4bk;a2o}6>qmj$&A}yw-%-vx(BH}YBlZ8#||&C2!OTM{y$=BF<U+q
z4A17vfTpmJk#_ksB}P01#o+-$3-m}L5Og=yh5DUh5<>&fG`jt7j<onL!mYtx?B0@J
z?MD+!k;H(HA|NiBaK|HK(&E%#Px*s+r4Sudlj!lnqe*6H007cq3t6g-+I3IIA|FkW
z7d_#r2+Kce#sc?<cV1!J$=Iro6pN!{pxAZE$6+@G(5JLKw2+XWxyqN4jr+b-q0^5w
zS-|x&swi<2&L8}gRBH$(X_A*hZ&04USfl6_luy;+o1f?yH1+s1lzNetO*i(|R3=^g
z%#L|AXmEZUZ^6{e6E!uUby7heZ<lYizP_L%!%Sf!YSwTUvTZsh_HGD0QT)fSl&YS1
zqO_2cX)tGXQ5iVr?^o53W0sN2FRKW+z7u6(;!&7wvuHP9Y<}BHfj>BTDj{`lA4XS>
z^T<2TCg2WdPyclPHJ_<~FJIx0>$z8H92+;WbVkQ}>r_ECC8O@t?bsflj5}7cI%6hn
zxRhh~4G9T;^=oM480W!f;}53vY6%r8@7%a}x9w?iBj{cUqZKM<n}QyOiT|j4mZ;i!
zX0LPQ&qJOl+paXIN;zukt#b5d;+xN00f|ahyy45R6j(^&0LPO&9zralg*v0|XYzS^
z(zc|#faMtXp*pW=x$z=pz``_+5D=eaaClYVr~b6e3jN!M6xdn!ry7o*Q3eb(PYz`k
zfz|BmEwBi0kxN?gYpFFTv!zjlyNWP|_Miy|hBK2bmLWD*g?H2i<>PlhW$CioieOJN
zAFxnBx5tV1g$M$Unq`3=dm66)TDZa3SeGAzUkHN2!EuXOb!NI*9*%)DfoU(lW-1;0
zT}AVS`_YRaPSaXIC-aKNB4s2zd9$MX3x)b2yC&((W->U4d$a;htO1>WJO?RXmQ(U+
zD<;=t^R4g8i-qqIoomT{YQ2CIz3@`0-#}L4F@>MT99ea-3Z4i!cREov8^D|Od7%O8
zn_)ZW+^<?Sa!^2;v7%D?fD%IPul$Uf&BkNt2hwBp=QR?~K2_~C$}T?bV|3bL`D34p
z==lPtfsd^2r;5;4b%F8AuYNN_25LyP+YExADF<F^&&!K}bjQ+j#5*9>yru5`4n>??
zi&CfDE3Ge^r-asaIZm0wVaqmi@MKzcOrU)6jaFtz4vCShUxgw`lIs8&?pYK$_IcV<
zq+(<C(M=!@0MbjFq6c`+{GV|emmws-`>}pC{9sxo0i&q>1rkeGlKVEj;QNM<)nsv%
z60)RSrOpSj0u`6?|BZR$3WvpdJpcSEyMGHT(bCa$*EqOnj|h(91GoB%)M3rHs2Nbg
z*Y$wy3{jRu+C0Cs;8#&Q>I1<3bL~l5I^bi$VD%<46y9n=zymjoAu@Vd35Ma5Sm%5c
z?4J02v+?+9feDz`!uOaN?y89&O*Y)(E*@FX_bXak?4xX1%W<y*v9_XTt{`Pd0cm|n
zgJAX)V2-r&nqqLx_F{`tq4iVyPyuHZxjmhNt2-SXk?#521lnU`tztna=d?1X8si{Y
z^9UO*iS`!Q4u|j*D17o|e=k_LSGy;@rUx;??aspiNp3#{#+K_X47Yv>!bH2jAVjp-
zWSw4?;iCidtN;ySn)-CQ%~;gOt8bc|LY(1m#SmFM&53ZVyO<#FJ@d+k$-)*@R|KCA
zG{(7Z7J<=>$S?w#BNv8)P@YWjj@s5`)pS(IA;}>!BYM>V1V-!q`gEhk`8;)(7&W3d
zL%jWrNs=YqV!J>*#5(mNslLkkwj(KOi~H4=PG>;6=1w)1iHH|LHp>Mw`GH`d9G*BP
z&bY`MjI-cjgg?A^eE!=Tu%)fi;t?FKiy>TOuuR;xIbhhTzT-SoEc2!)JyGhOVBM4T
zvA)+q{&lQ*Nbbwr=|uSj=-G7%O?{X2V^-ypaprRa6ZN4x4RLNF3fLE(RQR~Rfqf^f
z64O#Hcb(P-DTyreZz2Eez5-Hp?b4(%m+R_-KlHY+gr5TjX~wka@!c~>#wxHR3~NoV
zVmNC0si|2QoOi<%^xSD^PQGT~n605`xxJ5Ve=eq<G_AAK>qSMcD?|%T*(UNa5SmWc
zU2@!CKiS!QaF<TIA|qQ?Z|!y8E2qeYsLJoBI?rl`d<f;nw2;j`69^sSrxoXX{Euu+
zogM!&6`V~$aR-?7SB;+s%7Qq$16`Pcv4%*x$q;aa+Tml^hD=MX6--rA|K%XmWLDZe
z&kZD^Q!9^F-<OavG^?hu7|`fm`6?Pg*3P3?)psu8!FbCjp||Lx-9x2j|Fgs87S()K
z-k+xSS5LD*`W2v6>pbMkF-VfpZN4Y;&324eHf)Cv77?Zg0iL9WDT(W5m+iRc{G)uR
zppsvVc-iL+ZsdG~@CZDlONBRbz6^Q9xwW*>;?Y9RN|M|bEjRqSQpIc)83I8;sS-1P
zS$@bFDVhf}`)Zi42NwnT0v}Dgxc%O3X0;U<D?l^V&9)PsrKXWvKqsg;pwlZ~zEY8d
zYZ56F$kF!&yJu(c-^@BqmW1|17u%f<Xt!n>X<pMREl-2f#f2rQ7~JSyJIW&;ZgPq<
z*2fUX8J}z?t;v$3$ix>0b#N5TXp*jUcFK8sFU?9~51J?~?L-xZcG#V5+HPkDPj-h$
zk{P6#MZqYhUePsQQ1RK?8+RdRfaITh+hi`7@3kgH$+nWHjxF6g(Rv}Bem&s3=H^U7
zk<lH>tX&`9G}29BIa&~#Rtkg&Uw~5r+ZcV+8`<n303N7bismBy&a^^<?HZ({D7Nu7
zp;#<_cArbjh~|B<I_SfhcsRd^@BW&6FHvMay(Q~N3MrJQIkw#`4qGL`QZsCTptvwE
zF(C3H9aWq{LoJNSW<oy!fK&=H?ZYrK>$k$DV>mdOse)ufA$3IboM^T3XyWDm&Cb_7
zr<)Dosr{koua16PtyUyM8bT?o-ho>!<Y2*CAwkmHLkFuZ$(jzjaRgFwS{CB|<{vcy
zP-l>tbNiU7OFnDmv(Miz9+U6DWAAU>Ukc|6q;RTZMZeTN6NVl~WU#=pPyhQ1lp)>I
zj$6C`QdBMC#;!&dLjFHJ7%|e9s0^<oqglh1aqEB$L~{8m#mi;81d!+{UGr(49254s
z%(X3XnCEslV#|)14{~BP0`ee|F}Te1+L*?Ztn(kXHw~wyrQU(fDu~kIoCCK~XvxoW
zyXo;?kMVXK!C@hrw$UAy8$a1HpK|v4#%y~mNVig2QbkkUALQB;QC1aGr5i{)CR(<t
zas@E;l%=j;RNTua6qq`&+uX{_rbp~)Pz>I%y;pT--IQ3IH>U`X`yW)ytAtF(Xe#@9
zt|dC~Ldm!yTZC!Xg)E9>iWHW5Zek>=QgboJq7?2sD3;*Aobfamou!<BxVCN#J@4La
zA52`mk?1uD8FrX`7c}3h#^+o2{;d+wd6t#J^!Et+47;QE%{2pc@d>%xnReMyxL<Ns
z6H(i`#c{hCkRO!R&N1%<k&u%Ogy1Do2EAFfTlTs24(|FtK_3Z91HvUoghI~{Hr9r0
zspeF~tQ^e|EwuRWp$@~%a``d=>x1P192)`2a=lgiJ4yDdng$iXXaUp~Om@dapVkPk
z$+`fovvqZ5;qEOrAh9hSmY%Pv+5A<;E9IPYrQhZKC3?k1KbBS=A&9|{0h@T=uBt8G
z`FI-DCvdMV9RCm0Ty@Bt2}PnNWHxRXFtJsBixe6mv4eMsw)>&K?w(7cOXB#sfnm_2
zBoVU|5vKjj^3lv3Uh_bDYuh>><UbwDEFXMuVoiMJfjV5?cg>`u;`og7e|Ewy&nQ4O
zASaoKQ%eITNaTkE;B5u?LKmk}e#PmMxn*Q)C+=F68YlK^L3b6B2hWxk>mKWIOn<00
zCmy15AJ53DxaYK+WAiVUlw+WIp!({TV-{jWGw&YQ?A^W$^{k@lDL;q#GrI0EG4unp
zV1&4dq<CX<q5bgYUV>L=Aoj5%E$UF93@!aOJX2gd_y^~2>@}<yxiXv?muTG+vcP(q
zr0KmWGvPe;wNHlO5;w1R@n;nfa>pvjiNc(4NK<$>oPSMV2@cthAT$Dmj3b^sU0~X&
z_c)u?iGq~GuXhtiIL=ny84;$qSgUP10))y`J2)m8!70S0VQ)Jt&Ag+{#=Gw}Cn<U>
z{fJOUc`0q!KV?exr{9h7d{8i-Z(0I_flr>V5Jec;Rhu(G*#K+5z=E_d+k17#QKn1Y
zmIS%?>lg*RwA~qA1kmZfXocqRtp{M&INML#s57_uarVvB_4$S^+3?>p;(7YnuNMt0
zPdFwQqO$%iWc{6`4{q;`SvEZJz~RagadPB_)BxCY+;Ybh=QQGSKB>O~;fV-lOd0hR
z()Sdk_)P*<J9^t@^6L^mC=YiQmW?j|=PJnLxnI5{xaFwLy|p8T=KyqF_aM)4Rg%OX
zH0%<Zuo~UIRK0~g;52MmpzEP)cJ<{QIY~3VobL5QKDGy&^9lJslWa?3{~ml+zb8%i
zc6$WshmW=-Tz<Fwcc#DOp5%4UpH0j3f^Fznl*2C%(Nl{G2R&Ue!Nc=6eVyY^=6?d0
z5R3Rk9CxGKufGk<S7L50^(3h##F^MSEIZSa6080P6iLETsSkRLq)#34_EHmu6#d@f
zcOenb_k2hI{kEv#9D;i~G*(>(4ZX#u{uc9nt}YvCf2nA>7RHIZ9=_kQoX1~)_IYoz
zek%Jnz$Hp=-9_YZnZB)`>uihs7z6<*);~jeKDa9uy9+2BlnAI;7PwMbnhNH5YbHXd
z(Krr`KMH?s$2ktYXGu_D#l(v5@Pr~w?Q6)m#77g{do9<Hozcn141jiua<tu$Y#0Cp
zj&t)hYO9DGU~#Pp!<M_P2=u=tA7N~?A-mnne!9;ya;pCR5wm;d8(CGH^Tq+KnesFN
zP-=ixh|%f!oA4hpZ1+}LZOh&!P$ZMgu~FdjpIeI;C)YhAfBI+lG)<zwNyHD>UlBh_
zeq6*VVW}Y8L4iQ8W7dke3-f2d@5@K#Fu605WIff^x!QgQ(8!qXp_yJx<ZzgQZe)DI
zH@fC`hX^TB_c6oSL%H4qdtmzos8ZuZ)=eKAgZa2%^;@vUrT<G6;HHc2(C+BbeaJup
zuL)WH$`SAADW+FXH|vI2*&v`I1!vlH_ODX?(BNoKLHKXu8<eiP)s~Tauo_CdPzGFh
z)~wl#1If1*y1->t?av#k2qE#op_zAt+@##+uPeNd>{NaK@cd)`YoCE9*WR)`Vbl_L
z`meN?_?A!_s%Vx2k*-T{FpaIILMPvCZZ_=MB|P7I$ooRx^tn?4&G6mR|AqjPi%`yL
z6Q4db;p989*U-~`+imm>J1mzhURm3LR;m_cOz^$aS8a|BsRMV=vNgKP4$|4sg=63^
z%#o-U+R4o|U-qYQ^1CO!88VWZZ}oNnG}$Uj2Q)++(SHo@@XCHKV1YNk%R{G~SqP3f
z*9OY9ZgtmH5)Z#(jIskvDsaE<uD|!ls^8ux^gIs*#C86*d$_DgP)#4FD@(`iS9<Dz
z<xveRc|-T;6*^5;y92+DnU&jN4=$KpRX75R3Z?ju4e#oN2nCA7EHxbP5)Qr@%EGJg
zTXP2_b=RU!m+=(3J0h}9T#DUl)JoLp1-WI(uo^Dr_1LL~hrl>xEa~Ngh81Gdx229;
z!udVETXA^$EPNR&4y!ScOhXkzhQ3*nd+G%n?EC@AAZhubZFk`>zOHntAqBYZzuQRY
z*VEyTciDrzDoG4`e-|Ge`GOekGNJ?W9%xv{K(6cf+1(*r%S|<X+oO(!*Thvr@H5nB
z!9Bj0Sliy&t15FPlSF$*PjAMYaEIk?vg5y7juP2UajpK4KjW?=RMWWj_LA&n*LSaj
za|WuX1KEXY)#}8HG=6v1NUGt%4`kFDu@u}d3#P<)bUQbed+ocG>8S~<`{9WWpz?yP
z{Mw^w0w}s)<lx9=1Yi%HPqoeqY$L&hQ4@>j%Se;zv8z?$t9*5b2MQ=CaSImV5{#JT
zm?A?_^umqCz*#^rKuD1wSvx$6NfA!xTJ|_TQ0WSNDA7umShRzExcz20zuur9<od(z
zzZcDF1YQNQ_-6nE`9zej{^1rJ%11NoTnaoy;BQbs##SB_!xl|MF_&VC93)=(1Py-s
z_=^)k_GsO7;Hsn%yd|p`HMI_QPGSGxA=;wTC06nc(uytWfJv&>0z}G8u|yXPzz~~O
zfjh{5QyJn4wg=RpoK*tYm5+k*Iwq?ODr^Q!1t)ceJ6uc&N;hWDi+B0+y|Fc6*f8(i
zQ;~IN;`jcQz4K=RUigx8Fv14yx@}ef>gaM%zq&>BW15C~u*5nyU+=(2AF6l@rgkzJ
z@q*+&(tQd9REC!YDK5fo<;Oya)1wdpO(61v^P#M!L|o2w2L}gY&L;T^0XSLO9SJOW
zZDNNITPBM#ZT-Y<EV2wMyt1nR{PQ7eC8khdXD8VQ^x|aI_Ze`Hvw`O!rKEv^U@Lv*
zzt%c#tTIEa<_WPwwS^b{rXPGB|KUn0m#bBa;O^wpX&Rw8zyu*H562nIiFmTmg-!?T
zXPgpc<3%}u)$5c+w-akM2LN(Ez3)HECW+X{Z|`i5i(0}frI8?ndZnl3S;3TaPQ3I~
zO*IY-B(o~&I}!jG!3ZH0bCmJbJUN2*Zn*)^KE4lPN4`LGY_YxkjgwvA&We?amdE;7
zWuLjh3PyV^YZlP9g0KC~csf-EQsy14=&oQ%$CkR&LUyxMsiSWTB8?#hwk&+|OJw%{
ziCUlN%WpYq+h5++>a&2rw-uHMoxggg6l`BeYehmCTIAw_bsaSDj4-(iv6MT&phzhm
z2b|eNi+0Tjccq2c-IEAnw(X&<-he^KjeR!Q+)SN24=j{&qRbq#OinDX1)g}P$>SP{
z8?gAG=wo)kwoPk_PP~!x$(pu#B`XuxekyjR6GoNO*FdG1_)AiY_P=}uPfHlx9Vl^{
zViZw$mh_oETS$$Ok`MO#=Ue-<%GL|HS?oz7lZIij=pIAcE!d3p#RLq`MWCvTxG3$O
zUu+($XDuB|v#Z=%LJrs5Gflh~y`>g!<+GS9N*ZMM>?L`R7dRNi1@YFco|v_wJ2c%B
zO|ZicCZ~rVq^g1bAXK$S#~sIwb=CSaSucqu$lWvyASkMCh;zON(B#1Ji{FAoAY$gp
zK%zI@GAAUOgP5WmM66jGJyYk8gw`S}mA~*ZV18>dXwCij;7y^PsNRmqMMwc#P;mYH
z0PZZNFOiS}-_ldVR%|pN=h+eY-P0kR)<>`bgP!Mz3yC3eK0R?B^YjD~bFOb?<z4@1
zE`9asAMfQ=CZh}|Du+A_cSO)g6{eqSI@1KAGr%!;?^|xp;z(f-eZ29X@2r^hK;ro_
zaG+$zY87!pkDu{e%IMQGz_gRNE8V}_f2?s0#Q-+zoxitV&R`^Hkr1r;<~Y!xc?X`I
z9O{ax%JnMf>waj(S9Y5V;L+Khx0xeFAMbabN_%S$aJj_HP8@84U(9lD5LWYADL;qo
z7x1gql|8tD7(AEG71w4+wspa0lH=n;d?~&Uv)a41^oe_=_g?}iQ>Kc^xQ|}GXWS6c
zSK}i|u}DtdaTO!Pv67G+^@a$p!o4RZUZ`hxlc^0EpW9Hh$KnCaH2h3chhqvH5fmv)
zXb%JIgTOEYvqw=cI#UfVhE%>8cb+r~IyZ(G*mx&^4^2gt2^_Y4!jGsXXN4>S^oAl%
z&{l11KGN-@33O|>;K8rB!%dWol*l=?b3JQdhzz2extYiFTEbhnfNAUU%+}xSVd|cw
z&21m=xssx$c(mHC-ZAR1!<B`(8DtC@exSQK38%|~$M2fkPz|mUmrEHbt0RMBv1v3C
z)mDQhT@Qv<Y2uY#yYpmRQ{I=o)MZ<HfeuI92P<%A?F}L!T#YjyApP|wFc&kMN2p|+
z-45$-J>li-n`&P)%Qc-SmGJ`8meO&XF6--F+la}mPk{{=dZz8Ql6DQ&RabRxr>i!G
zQ$oRbLi>A-UUNci*R`gUd_N*V`JYE%JO^4KUEpue>4~dS2Q92kX+!#`wld{KM{lqB
zWgGwQ!`h-2MjV5Gj}rhbXK$$veceZ9pK@e*T@CJIb6qc=VneDgW}V@puQTUSWHe-9
z(0*%97rgb~o(U7*tKepdf#Ry?>z_v*Br|18Q|k8~C9@2n`NMe*-w7mV$7Y+;-ONMa
z5vKJ|oV~T(gT+`Q6Hw?fD<6HWp>(@6${Wm%{JN!kYOj4Bfq(4WYHjIuY|>(JnOCY!
zUI_?#n#ud+rcKg|+Wb;&%W6Wh&60(Ea7j6QY6#zy@10)U->N=5)xHtG3-UYp(iw#y
z+&R7GLp?Sx5(S}mfd|q%$A=AW*S_}l_9pBnd7!(N6PkK`wkg<}r8`Qz6x81D#L>1B
zv2U<@ssyqCgIIzhk8(MCYTb$arJY2h!wSrZ&3In$-IUiii;ci(@2fS-oy@a#wy(&U
z@ZNm^S?geh_H0OYXIL08;4TOaV&C}#2AB%$I_tWMk)Arl?Pvt8J*N~;^aNhfY!BpW
z1iEAe^qx^$$AutVy)ci1|D0}YxW)};Rj&O+9&i$tm|b<=HoShJEXw8r%yZ~J=m!Xh
z&7F9krc)_H%S+@~{S(>KWocCbNtCwK$%&MdJg|)k;Pla%?ETs+EZ3vc)nt9i_)M?%
z1o)iptyfPIZCk$HGy&Jd{lIlHeAkAp4wZ)_mpt{h@BY}$q;vlTBig4c*N=FX-L1Z`
ze#CerM~)t|VB!J>C=stL-u%d8`DTH{(1rUaEd<$xK7oqf1jnoB2!w}(vKT>;*Zf`!
zv>A3xYWBXLznb>6u<85EyzPk;N^M>x&{&yQPG<t=fs3$|S#CW{I8pLH@HvuJguK~<
zlQ{kyp2+od^(HKH1nUZw8aw)!0gFnoVw;d~nc|SW1qu;1C}5Jgq^tcF_{i0nvK8x+
z7d}{2*jIOgj{{n@Vcms$p0=`{%SGLrl3!LCc3`EcIn2ylS!AHNs;MLKk)u(}mlbor
zXgzHSCRI?m!iU$Tj$cqDYmGVZSI6&nbYS04eC>i#dQ$XPtb27`%j{($*M<Uhj^;BV
zH>(iu=_al^{xJRE%5P}u_7m>o=^D2@A%f!Eap8Bu==Gn?-5s>K!FNOc6|X&xv-rOY
zb%r5qH)27#o|g;X#C~)1=*dcE0WH*0tZ}!HOb^Qd9OiOC_C|Y#<sv|z$@uVgt%`b&
z%Zrhp@0&U&EqX8&Ez1QxrJg@a<yTsZ1y?zPv4Ky$69B-Son$?JZy422@1moA`Z+KW
zXQ<yF=uvWz9y>-Ey5r|5PX}R-pqfwO%;=O?tL>N3g$OCXy+`q$JbFv*R7hwxzbFr7
z`j8T)P!PJ1C6<XM09VmkXV6d_n8`Y+hXde!<7sjQ94f@+vwa%Z;k@m3^z#f?e3w||
z)EI^u_RhntAC;U%GCY42y5Z1Gc8TI;a2ccRk9eP4M7AWmErz4#=Y=+2qF}1Ty-oF$
z`}o~X+@I;oz+(SWG$C&b4@Y&0FOk-=C!FAxW7_FTB;TJ9%~=z-Sl(<BR`dM!N*5#F
za6ivI(kZ#}qX3ov29Z7RRnz`KD&q;!@hImaMjbQ;p{v+4u>xhbXS35LRsr|Nt6fYM
zt<)L_T%=H(V=IRw=QQCy96wx2d<haMN?#o1`u{L01WTXXu<kK?^UJKIn%z&|&V`)E
z$1(e@SB`v%cH6vX{7)5#Wa^bg_q{eqh+2mG-(Q9$;NriG9e*O>N?3iCSK;eQ5HXVN
zSj*nov^0qX#-UtX5|X2D!k;>u#>ov24&>h^&iz-!t~clJYfIQls@${H^t}<q+}U@I
zUE&95z|LECfapYU&Dttly=r$E`s{yLUQArr!lH#O>O2dFMn(t^hk>jsIKG&;N!$0@
zBLMaLFv#*4ejt0LTpqwLVtlvx5+BQ%w3oS0gHN-}Xz-a7MMH{t{z`>Foak;m&h9Sj
z=6iJsnusr~>v#vf;Mv}n9zq9k0lQ>xVRQ)5{^Xn}frV?F&-Av&_5nw4mONK5rU>pI
z_XH1ovPpT>6NrMZQJz4(f3u{3&hlDx)ddtQ?!ytwEMYnWFlkHx2QN<lz=H|yfP)X*
zkovuA-P6i<qo$3_5%@>{IU>=Ifep726z}?zO!phuMP5*A5eruWfw-lB?KQ*EtEJS1
zs|r?tYC~X5fzU3l5oq-!yjBOZA$lb7?~nN=zrPLSCFBE77r!OvEL$bxzN2e(u1G%k
z!D>-%7YUN)B_@`|AUVKF>9R~=gbEmR1@aeV^4}0u(utGJ=06E_ug8>`p+@OM*IGE9
z@(R+dti`FBE|z52duO35hsS?=?sg^2&O3Azkmw!(tDqcasPv42)0Fl#9ry^kwO#V$
zUBuYZN5E;gWu0jA<9yr@f6J9twHqr&bjEG@v)!4%>nta4JXeKaDk*GZ5x4n|Xz(SO
z@LKRt-gmi3_<|_-dO_SUuuDnf`3b3gj>1Ax$D3rr6X;***fT7m)=l@~)v!mJg0Tz`
z39&H@?L7skVUD{faPKPRw4)XTpT#VDPGi99^vG+x*26yjbpD{pzBdy?K`}}ywg6T{
z&R;^2&H#-xo}?0kaY(sZM|d<88<vHE6CmhEf_d<K<MWU;Zr?kK=Jyz{9y652y(1{0
z5&>;~zij*^#xA-k1|HN5+Y%43do&h&-%^4CSqdVt>H+IVl-fqC5ztwjPkrn3Kxn#Y
zTYUYWdBD~1-@c849M%kWvQCes5m3Y4_DlVg*_0hGdJXUQP8YVFHF9a{K!c#*;Kj0T
zInuT__1VFMbCL=H`I#ghUqV34CJZii`<W5vZEaPqk!{cLLa-825r`fddV`R1oxzy;
zMbpf!0p4f0*US(JJq)~(i}fjaDD<=JUIK9lyG3*~pm<iIxQB#n)EJyoMm;uFM$ITm
z@d8IvOW)SiFTadeG$MjL7bx+W32^I<ZnFBb-+}ol`)<7Ak$cVwp)wv+23FBD-AD*+
z6YaRI#dggaghTL&^D{+p$<FO##>sEx($LW4MC}5xU1%i+!#$b_q>zUUwVInPe_Vil
zI?!Qqne13y?2<3zz*c?#Q*1Pgh4<JN&g6aciB2=%|CR3q_JB7ii1iA+3DaeFHbxT!
zHm6HR$fBvO1dxfBC7d<&)f!**KfZk<=AP}+=8+c3HGF>DNUPK3mplrBw|=w9g7HZg
z%0{5PTXh|FfA`;u0mKj>ag={Y0DP%&VESTb74R<^pa61ofRfQ{t2S5V-77zYo0FdV
ztlw~z(t-~99NUV+y_eqa@*Bw#I#)roXHXn)hD4)}0o0-su3a_db@%K3Lxx=#(8*#c
z(<JjOK16&dV}!1<&%d2b6wcB_@iwo*10ngd1KU(44D(7}oE+s(?tz*}@K#V<tPfWj
z8)*Co4AjQ~7R7akkbGjo1)CAe_uKcrqDsAN07bEVJv@-NWJKeW=&;OmvSdxC3pvhB
zZrt8p;+Iv3>gbma>=;Al4)U2{zWS4`kN-A?<n_3q&B-qf$#$PpnT;jxKW5!#!Phua
z?TAgx8vQGlyVK}DD)D%N6|VULsvwExl9v-SlFja`15c+kSL19n>D7P-kQIV(gjt~8
zoj!lh>ONA-Q~n9_rLwf?l)Et9C>F?sk8VT}La!^$h#d^6Wz@^;{QmNO&C%~8@iGU`
z<2QZT^LiJFu{RMAD;3PT>kz8Z6f$Wv`oK=!a7G7n3t@SDfexC`{Ak%wbX)h0!V<vB
zh{l#Vd{E@dX5hXI0HWMj#q!<cz;kL#Ir&SW@O%LWEZ3^;eI+B(wk?{#$o5`{nPT5&
zvoh|yN5rP!oVUNKjD1-r_nbKe(kAQ%r9~i_WR`5{x}=D*m@f#L1qX}!kOlv^f64vk
z4rKi)(nSu-H|qv(j*`ks)_R>rA{i;Q4cSt|n#r#6=B(0OfijuYGz%rR(`W7hK+ZXd
zM>ds?o_Nt-y|l&;C=icnPR|Sp8TlrQO*d5!89LlV_+x+YL37)8#ZJ>BxDyC*9MacX
z6R{oSmBmdPKTa*4Y|ma~@6Rg{ue{-?N2I2PN}4}z?q7OE_O)}Vn+$1S{-@vX#QE{L
zS_CHjMRs-cf=<wFFXJ>0M8*NUltq;C4!Fg;UtL$7=OI;~i;Dl#A6pH?lonu25Is`L
zc(Y)&7AzGT=!f1HygA<rO)vY&iag*BXH7hC19Tc-);2{TWFKM&U;FhF?cDye!q%U3
zWSq8<?c(WrYw|5)|GC$7`=li@n9~F&47$|fqy+!y2o2q`iCPAlw@;wPkAo$U5?|D&
zR>n>74;Df+50T)-QF}tT&+~l$7zaoUvtU+taOmGLRp`R~3;q>@<|1{R*v_g{a1(gE
zbB7X+a-_vu9T7H1Gb5niW+QywUjMld2-D!biVL;6qvN^<{AyU&lBym>tJ;|xx@K_4
zx=fW;b$;fCqE<9eFC^7Y9Ff+iqI}jwlTX~4f7%U4{r(VY(BoV&d0IG@DdqMB>`|F~
z&wOM-p-b(_3&gb`A%*~(+M2AS$|BjwONaY6=D*$9+6OEe27tC(SOm0L`Fi(0nRK9L
zggh@2W!4(s$x)A_;x?-GxQ5VU_T9W)^0)TU6K$-dzipvfN7tH9e(8`k$x{<EEOpqy
z??N!uj)8mktAbZlc1YSS$-v-qg)kt$z4?|>*-bT!>#xz#Ns6$=0}zQ#NqI&3l&W^$
zF414;bOJZ;6k;078jGt341C=rrjYn8By<3^KG*JQn-_mlvXXGI;pxybuT9&y*Fywi
z=cmsD)u~*W?{Q#gey1cbu5q`n`X4vnZFs4I66A_1<A>BNp2WeSCoh{WKK0sXfw;5;
zL3Rdvww&*-pdb_oXa?5jI{yOt#3;zf2-%J!s!l^=FYN+*R%hVLkTU|XrL6V=OTU-E
z*(bs<9X&}0z0AAH_i}!E>$zGQwDpByJ&)t_$9%n!3l=w2mT=vAb0um@#G>Jw2X*@@
z2dI!gwu<FY-IU{l%#Y&}HQV)Ix?ZZJ_*tFT2>gC{!c0eu?OX{dyx9U`a-~>jLx4b*
zpJ0G_1lUy60fodID`M+<@1mc$B6z#Jw;NYmg(Bsz5&(^c+A6PZr@?UFhP!$?Qz}+j
zzM3ZPSI^~brkAFwj!i6Q!+_uLWjZD<MSB3+(l$Ju_W8v9x^Kv6o<O|5Un4LYH|#4k
zL^*0X=pjbGrbpfVDu!vO{Z6tjGC7;p2vpO%qSb&vW5lqOWC2D(g<-ip)cpd>*!2P%
zty1HJA|5loW|VctEx#(j_ZBw%Pj8Kq&C+b}-p&OGS}PD0(DQ%n`@6Z0uiK~Wh40|z
zZoEjZMbL|Jx}`Jjw1IAtK-9iB8e^xA8@5z98aIdUuW`1$Eq_N$At<bZJl_|8`n3P}
zS1oz9xiid(p``O``tZm|cubOYHtUBnfrA1XcLA8U20RK<@YaL;4tMTb0OpHmu`ggK
zuM;cGJRB)URDSRA5&mSbOFSwi@}F2wXN!aD)Qxbnw^3cHg>3t7<)+->xeU*Gw4r<~
zzk;GBF!<Lb-#H^JoYp+dMMC?|Ci&2B?*H$ddb;a|&rQ`bP^+R4iC|<Lj*=bHm(by4
zJG8yDPe?kEcZrpeWTC*{II0PK6#M>5epGYvH|B59k+Bt+7U)9zjCsU?aR1BCfTYuH
z@Y+>@m))Y){r>NL)uQ#*A_g3HR(_g|_MNlMeX~X#J){+9(O<_9Z{Q5HvA?3Za7~7b
zc}{z}=^n4|ZZa$zv}mP+<B)IF5p+X&iPzS5Ph*fA-<_tu0>E=PtdBFz3HhnL;zG@%
zADLpA@mV?IL2O&W@h-R5zJyT`mt5Io);7>QoJgLn1R=7hUj}VVLxBrt#5>)OZ`~mA
zgvz*?!m)_JQLC;sOiMU-2oBHht?Q1{$C2kF?>Um7+BMy23|zlLTW%U2OyO3M=B`d!
zCK1HxDGMrnCkGw#aN*m@tFsuOVt2Gg{H^;hF1QP48faBA6`h-4a=1R@ZE2MXESSEL
z`f0{h1K}H4bJuG;%Qd|9`TNa)-k~TSyo!IT0oqt2H|4>H9df_=`sjR5O#tVz=<H8I
z8*4wm>Dvi{o}0VKKdkmTj2S{h3o>O~DoKuXnkN~6uS3?RaG^*~q~>lkrR%Ya6=nMb
zPZ#kB=rN>i^gr`h=rO84`E8jdl>L!^lqMseg8?Ep%Ex2(2{|`rp22v3OuN2O@G~+C
zWEoCnCs&5B=@NHx`tT`UAT&4+lmAf@NM4&Sya1L4p$}n^@_7b@C!fyS9Fb$c$rTng
zS(8?|F{CV3IUd~_8cFI%He9;2G$WF{dvku1_q5|A#anE@5<N=5ZMWPV_xo!j*eR|-
zLhKy|B;o$%YpQ`2D@W4Dr+wIA@f{c*QK03Tls9k=Ch9iVDrz3X*@BjIdQ3KfaCHnT
z49JAzvlhveoZ$JeD8w&Y1wlg^%#2319}qFq;Y&YGd(X${zQU~Y<Y9r`LIk*BegWCQ
z*TxQ05RfT)Vym+-t(GV?z;S%pbjf^u4$=Jl=s01i7lcm`w?o4``_uUp0)ez`=+A+`
z%qu?XxT#}i=Aj{8hRMf`=Fimk%!2=xqD&-ldF;D0j~mtrvY&ohuE-3Y&C_4MLSuG5
zSx;hsNC`SyK|iQ&m-J&e0!u*zL$eFVy7Mn>s4o$M;mt7sCF{q+5H*N~=<3{C*5Dq#
zmqA=j(pk2#8lb^z@MD3C2M1(KIlVaGnvt{GD_39%mNol^S%b<uTBMB)tUXPt`r5&<
zvuI030BQsm$?v<*saqOcNUEt*=AY2&_i<@->*1`t5f+^Ok%PEBt8l|VQXJtS+u`b2
zuLP^Z@z5!eaIM0t-0ev;et-^mFZx0dd>*g%G7)~u7JUW`R;ZPm{|3|ULwE?Upgo?d
z+xet<P8nJ|1OZa_2HJ>wx)EaDQVeJ+K}zYnf?<F>2r<(w8k%Ek0Hk&v1Wcb?zeuW&
z8d<bL<8{@d9g2k^^mwb=gk+XiubuPurqo|XlU&7?!O<asjJtlHpRrwkfCz&rV=hd|
zbr*~6efuu(XF6l-fl}T6IZ|n&TY{-rH{>_s^@(x+W4I&OS!2^+iJm&IuJHk!ccS(>
zJ-pckexjW}Qb6s#Tw@_~>jC-<boTtOs*u=9Qab;cOA1DR0F%9i3*v)<T7lmbT#pYt
zZpRwXSjC%z`Bld7#Al)j*L1hRSA0S#gPIw!mQN0S(=0k>-MV!-_AffiSZ1U6U{lvr
z5!pBD)j55)!XY65_!PI0pf0Y+cj25P8Vb*>YH;6GcK@&&0+;{{m=EDUg;!7IIsT#U
zM!bbp@QI6emDL+!R(|3ZlT8%$7p|^mxFPoz2;e#|Co(SiOnWs}eS}k?I&wVWgWYpQ
z{fAI~gkigGJ?4op;2QbIRqEV_4Mh%u|Ng~srpoi5G3=6x<D75bMeh@h$aUeO3If%=
zR{OdwH>Kh6a2iDO-*)Mapt=`+R{~|@xa7~N<L`*8PemGfy70d}ndc%n5Pggz%nZ08
z^GQOlYNR~>L_?h?6(~C3p)*ZV)BS7;*V4&Nz^4S$LJ#nzFT{~W(65Pv)vq7sV5J}x
z1SL>$E}BL8WmxVE$Rw)<(UBt254%P@yYD}kaPZa&j`y7Tany(-G)^`K$Z+g-3|zop
zfLx$xc;z0L(^N4(Y`0WzCkH$a#BrQWx@v{i#*=NixYI(J@-8txHrj2Oxru(##dm72
z6~HL(X0;|)Bn0k_+E;n53}CL9Miq^Z30A$GsgjSXz0}-w`k%)ILWhS1XFV9Qi_4n4
zUqZ{u%nz51l<512EYEuM%W!J+4h+k1^`w)^s-pjRwiuDxbj9-XuO;Bzz)#KeS~G_Q
zKdEI~^*;LvWL5paP3S>1O};pO6`o8g#L`K%NCReE>2<4$YYCWslLi0)#h?5Y-(RU%
zDyD)vMdfGJL`#*#7v1W*ndhzNSstkOKKoj?N%EKJHjSBH+?j`<U<}*!rP|_rtRPfS
zEG}Ea_@er>ZZE>Wdhoxd@52o_X!7CepFbfMPiAM=Q?)ORHW`?hoG;qn{qs{GM&qWw
zqc8GstUL5WrM^zEa^g9eH4iIbE6mAo>6dxqer_XSuX%b=MM(ya%EywSH0%OAteL!l
z*fX0V`=OZQK)&m(Y?TDAFcifb_%VYKj{jfZe3Q@LC5$;wp~_N8t5xZX7^rZGG39VK
zUD4*IA&EDv)Hret_0fR}_0e;O@;;SQs-JKfpWB=B$qsyr%{3`f;#%|TAB*>oNS<CD
zKYiOFk!{giF!NTNc+IO<!Q{R8Z#~&Jr{jIoPi9?4f0@gFOF<1U%BW}_A%=2y^rZ^$
z;1`>-zvnG>NL+o!N0d4K-K=#lRW!9Vfo)Z*4IsF?6M|&tQnl{K{5u0)+T);V2G7z(
zu|!{zDSrv(qnn>3b6S<=dNW(+K{<r8VF<+aHY;jHpo;hE<r}blKYlbsX<(QWy`>b;
zVg7cADVXXS{-0(WS9y@Syt1tP^LhDVA4kUR?*99R=Nl$mL@rhm=67S&e8w-U6PY`p
zU*kh`;r7R_20>C@o6lm_hMJr|sr$E7eXDXnVkiSZ3DRY5y8@sR35TbD!V%3CB882P
zk3~ycPb?qb6m6q)4YkpvMEzRA7;9K7X2)>zN3l>`?Gzq<3yf0e2H;(ENV#L$fHTm;
zQ6QSZ>hf!o*NZ#nLlqSUv_7c9r(gMTHHGJ%YkICOmnjnuxG>|q&{k}243kTdO-0C&
zxAW(!v+((}4`Q$D{^_S|0H+Kva}HDku$^H(FX3__BZQ43dF1kyAK9@DvbbNdu-}*#
zz}n}%9oTFb!2Y|xeSGgVH#Z7W>qK!PTB6WT?TT&7!^{A9Mjj%K8lreh3R*b<gp%VP
zH?UF2z&Lhfetp3pbcV*pClqW=Oip7p=;sy3_7#f=6Pr?MtulVSyp6mcc?0u{;@sX;
zLv&X9_}n%o2E2s+*@LnDv8CY!!sDO3$5uMdWn5Res%LLmmaBGOI28b|%m->sQy$Xh
zZ8ezdO&%zbll(w-h2?{b-01<<x$lHCf|Q9uv>hQ2{y_Z$65WjK-j{d|d1_zRZAuLN
z%mOP+{dGVjFpFZJm{tq@W#OPOR(}6EVB22WJ)YN@%hOEWr80)^<7~tRGqFZ<x7z<2
z_j_Nq^aGZL%Jnh?!oM%DV_ffMbxXp^p6&m22zT0H!$X5(lLEBst*2|Ve*PIB$+B;t
zUIm}>wFE<*PeZd&`3hvNeP`t^(~T%2G^T@#gaNObA6;NH*YbQ2U~>y3dn~>=+vB6a
zg!z{hic{KwME8rx3kLKl9n@;;h}}N5G>h`%zjbyu*VK1?=;vNj`UFGc21`GT;lDtb
zau~Y-;c>$sf=p8{%O00w7Ins>xV%lT;;u}kT1II14lj~1R6pbi1otw8Iz6O+>$URY
ztDuFb4;B=ci_H}1#z^$-PgGcyqP7JYaFT~04MtV%2<Ve<pGRCyIGHwQhFWbvSS3T*
zdKY2!vcS5$%vQ+7Nb^U2!(K=b*b53-srz-ftRLZlwNS^u3)VlaE^uh7uU=Dz_wlNH
zhkdyIOicZ4Xn8GJ9_Q^N=7d%KTBXeACJ=A!7BcmwR}p~j!xK_#oniW<6v0Lr&vB1;
zbVTHb0;o?5_LLxuP$tOEIG7Y%IsXmmEeqi2m(?F8I^qwmYps26|GJl@+9m%MS1~HF
zcEkuy;vz-jY2aO7L{cJQrIr6!^16H!>-Ud^F^CF=y^MP|m5js-SKz_%XcRm(K1dZp
zKzX8UY!O0GC8P}9ibZI|!vNc;rKl)ay&U1#8T9=SlwCzDdGcGJU7(9Rh;aoBwQn@&
zQ#4SmwAz@N=C$^t2~uZ-@uaknvM19YCzDTzhFh2Y_#kvM&Oyq7iY&SuXDlySlbn05
zlOYu`)AMN?uTJ+3+oNF(%sDmyUfoiW*qph2n?p^2jr;dARX3kEhOl7kSA}~Wn>zk6
z=~?+%Me=ss!peqcg$I8lwzE*Cuf7p{(C0!RSKi<1O;$c-fUC%Oys7gjdV%onk#2Qn
z;wkZne`7`d5>M=!_SyM$R&6l;$UIA->R_>dPc$C#dZiD`hOF(PQ7>eEuib7RCyo`q
zY;HH4hOYn;kEJeIA16li_VecxpGmpXkc=bZelTZ6@^c)efX_tg(FAY~vL?Dgg_^TG
z6a(V2$DzC!tE#cN7?(<ZKwLpDVLEH-{&${3-1OCg3%h2-C-R;NY4H4CR6Zem<4zT>
zmf&~n^0vuEgFLp23E4~&4q~j>uH7j55UktddP(HdXOX*gOhP!|mqg1fPXVlXQd#lU
zq(;o0_=%z%;LRj9uXr)_C}C!vN^vNwEN6Frej|>DLJ>nJgzXPTUhk4`^<vr3qRlIk
z?;_}oXpa*DAB0c3_}T>2c^dsAb^(T&smM{nl*Rdwkd>y)?I2;FS~;~eU%Tv!c=mJU
z*U^I$p%g_0A1D;xAoS><co-V~VHU~AAG#eRNU{WMpUg>g>p*|_$_V#7X6I{LR%q5Y
z`Ft*6sDd=|GmDV3b3HAiD$9l*=lPQ7?Wd>jjXMaI?kB`jk^Jh+qVzPRe}eths~Nka
zX>CIDRKysp90S+A2xL)FmxlC(E`u*mofoDCNuf%AlRS8c=OTt1gX;B6{W%vaS$hL4
zpOP+gr)ZCNR-jg3``v19rT^B90+D%wI(9uwX|*a?_(j!wAxfmxGWl`qpY5$@PWHJ{
zKKJd;o{FW~|7X)ShlL)bY&81xpDhM$u3PC4?k*tdF2P=Be#;PIn+PpP9MI9iRca=M
zstb^)(E-8@Yf%`&w1U){CE^-hb-1423V``)IrC_a5mBT(Fv&ViJ#=$XoQtCujN+mn
zQhbvKoV;c&m8l*8K=*pWiNWFY&(|j-745zMN!uupeFqg6vyG6hB`a=Ql%<sSa~xr;
z;tvgVVTL<eYIJiAUsy>8I59iTYNfzV3*pradh#zGDwMfo7#In17#O}!Yf{Hw!_W;B
z?snWg4fsah0e+^qvv~~{eHwq{p>=&#hpycHq7)4C$EZ*k;v_t-c_M)(w001;<KO}R
zNr1%fd5jMgAr3Yl_&Yq$6$^@{`0&0kCia^KD+($&<5gmWq3nb^zlg>UvoIhHVN%cX
zD6@s}dyp(idmNX`zkmN)DE^uupLYU6V(a8d;culMKFL|j#1w~e+*yoM%7PVL`ICzn
zSw=rM`Jx-QFXWYHRon<)z|B{B?3M83{h8ET%sQws1{Tmj?8k}hV{m>9j%Jno4&q?B
zK4lbVE<bX1_{&6ThnBvt_zOQm_Ye(IGcE{Gm?6sVjad}%moITQI?!YbTLbo(0vm*E
z-i3<9PN)@L9h0gtqbM@o@A8hnbIA&E>i0X|UmrhDanqN`bUojGu>F^#=!nXU52_*e
zp~CRRK)X`aZ{M_^bugy=A>91*U3;pYbIxXNp5+aZXv_T_jTvzuk>zyuCHzb9sv?f%
zA_7g$KP@jX9p%(1KZIh{VZ093x!6Di_xmlo+)J+wG>z!H#PSdc?2MuG(K*z3ifrL$
zCEV>}N%m1pbfNZEw6$busfYIlu)%Rb)4*Ls@lBTTMXqIRoP5+pg-+ic+7?;YnPyj<
z0#X;bPgisnXZ~8TiO?cr@~7$M=XPT&xqJH-zyjBN-j+Cu*wtNxP`ED?ThE^<4Ij+J
zo<83$8O7c3uanj=XLnvr#vpW8&@dYg8fQpK)Ne?IuACKLR&|#iqYr+sau%LC4`T@P
zCarNqp<i8wOKtT=m`L7+QxxqI^jE^|JaS_i13zB>A-%HpZXj-gD^ecWMEJnyS+`GN
z*7QK_h|Piu+1!vnu`Ly4f64b>!HP|mBu53|uyslIKoN=Ipq9?QY)rg$XnyUc!khVP
z7TbM}MRK<VtRy(Abe?IJ6YLg#Nbo=yvaP0faf>N*_tZ#`m|UglH*nFM3q{G9{=m5^
znlq09o9qt8{vbLG4c-y;Kr7zvj)1@X1{>WzFv|z#%_9i{V(5R~&bmX(AoO}<0D$!4
zjZ4id^3w_UU~4v>tXgAT3wIIj4WSPB$j{KmH^OC`oNYA#r>(+y*yI)0x_`SWTjLCF
zNviEfpFWxp;elc_bWbkr-q)`(dxZ2!+cX~4LFSaWaeII{bA{_15A{mkZoi~eJd`-b
zDz=qN^D4Wbu<{f|^GV@|fOVBBh)$$NDptNq(hCcPFy;P0$kmtDRjKeL?jEW==p!}j
z^#_`$AG+Pylo8)E=NrEBz(cR_9iLTlS*|SCNdxCM5Rfi)M@E+CZ#Wai=-D-VR8!p6
z-!(;KldkYqm&TN(@wxWznZG5gEuS~BnvbN0zcoB1tak}l?X4E7ZA^=|*0<y-jeG<L
ztYGgYB%zeA8?ahuW~+=iIleCA?14@QrbzvDd_Qg2-gS$Db;Jo%Q9I_MX_Q$l{j|7{
zN?E~q?eUMd7HQV{3aBLnB*}kt4J6SHy`I@h4-F&9rp)%mzK5<ZroNngHTw;MOXsJ9
z#XBweJn5~{zi2GLHde4rCHE+75|4<~SRZ-veQKaPr!U$0dnY+i;ozTm>#blZDpJga
zC0uJb)9Z)shVNPvW&K|Nsb=u*-`6Wm1ZsuV;qk$;`jK2Y!uKazd(1NQK6i^_iq77F
zO=!?vzpyO5IpR7XG(9FIrqW&+u^StK|FrKi)+AFqMD}FUnb%MGgG*eaM7g~c`}F%J
zt&Qt_Z+F?4V*UBaOAiM;#H*)W(<c0^=g^=T!qSjA>^_|9yF#H)dSZi8bE5u<K*6?1
zvVmnA7X$Pc!vqJ?SSs?ojDlV3y<OQXh@M2Tmj7gXDzNCJW?$)9!ms%$9WHrhoaEmS
zCJYjjJUw~lF;bHl*iv){2Ej1t$<L1Ex0eaMQ+{|eUnBdv_wHWjRv%+8zOazSV_@{m
z7&VGeWl_%KsoLIvjUTd4m&Q2R$8AH(XviLwmWj1o=M<GVXU%RKtq_6E;izZ6xqN4+
z_VTjphVQ;~y=R?a1|eWI-C(-ouSb!_B=gxPLc@P?rOZ*c#l2MH{!oSGo3n4(FuI?O
z<*v%L2d350j+5!^`AklYER{LEtK2<aEO8DAka(Xn`d~<Mkr8+63vmxGCCy{qIcnOT
z|HyTDP~oX&#zvOKl5%soqsODdBzOV*ZEXrP@cx&9_$q;YH{bS`2oq^>BU{Oi^m`Y%
zHUHy5=IU@<3LjvHLYKS46k<?<+}|2jd)vN26tJT-Y%q@xOET<pyAZ82T-IRM&s}Tb
zBI+7OFojY+OBF^odYd%sy(uZ<`{U1vCqFm2vXDkcq%GP~UXgoZYI9dCAQdX%-1^U3
zyPuTCofYOi>nDufo!80Yl9T`ACfG$t$Z6D_6&Rf!@c8vyBO}MOvcCH3o_G4F%r&Mc
zeXWf@z0>j?&rS09Yl%gOPWu%__6XNCV9*!0vS|OXS$o=*jXmOoGTg{}bnLZ>rs=Oz
zTR({X8n7;$_%t?Ewq<i1_GTp3`M?Y!_cVvp?tl&_I7642E{UB1&A@GVy@Gt}ir%T}
zMNItp4oZgI+t(F}Fa;!A=lUWP?p;&;*G6D@f3mxy`d4mPocCDmz&5$5q(tP7;jvFl
z-m&4zd4x<39M2-wL(=o?3NuuOQsyJ@u5r3Au6Wj4A94JTrRxr-^8Lf_IgULuLe`OE
zM1>F%4$0n`kqD8Il|9a}N4AK_NH*CkTQ(80Wy{`so%4J3{dJxHx~|9jKJW8<=KZ;E
z-IRXvt-N4-Puo(?gXG6+ob$7TQ_Pe;(EjMn3w;?oEJ(W?nBZ&3N=5^gy#4I)m?2(*
znwI5O`(dbZyPktD8~N=Hin!)F>Xyr%)mE)2VNOD;tq{8fUYG6ilPDwlhg;)g@!EE`
z@Tu^rLlxf1(Zc4(LrDk7LZ3O%d>FVft+B;9ALRq#-(7T)N4k%~Nfz2?qq6-v3x<eV
z*K|%Z->_)=jNUE2OyZUtxlG)A0tM%v=9<7naM!gXjW9+&#E8}p&M4O5cNj<~%am{g
zL!TZxN0B_=K<Y|OqPonn48*=!AJZNu6*ClUk4;TP2)qo{AST2{3u3O%B^N2Haj{!*
zeOb(k5i9N{O~l8aHkM_TpYaX9kG$6X`FypoxS>aD_np*&lxFMGD1i|sDGF8bK_cG|
zy0ZL=Y~9jKXn{xTw<x{M(jQwyLn;W71iSo1jLNuR+|7cS*`xG{V4#kyzS&Vs>~(j1
zIB%D2g|igo%^X~bN2FEV<S>rjG(GO5DY&8%J>glS4TwX@vbtryu}QdiID+!HJn^hs
z#kmm1U9_aUJP(@GZB%O+V2eCm9MpEc4#5e`kwhtdH8(xVqTNjTp^)ekhobKAjG%aq
z8>~_{!2h>M4VMDaDUcFOl4^lzRun~0SO6DJF45CKFj;E@h#;4u@FjI>`#pYa!iG$&
zbqvl(!Q-S{qJc;LZefeZHF_w-erYQUmCw=@;!5j`HP_O7ey8(BLc2nv>@i_$Qgx)%
zru5~_i)J%lk=y5I#|eC8c0P+0bHecqEQafQ*SHEl?XNk=%<}JI{5ESl^84cQ_vPi~
zWnUd>N~OV5&MaFpp;gjNLZS2n^s(02cF6i0jQfg8WQb-zTw>{VOR%3gA1`^p-+~BW
zGHdNdoNJP3y#_&mGv&W>HbTX+y|s2(KZU)lmEQ%sCZngGU6?bGZ%vCyr-Xj?ptmvO
zb)dO$<#$iUwIVL|#7|juifd|rODX=PMCE$>wBJR;C|!R){O-@QU*z%8+k6?v_~{}*
zZ!W9pr=<l0D|`Y}GDa%Uj>kuOKiltySxMRgh>K`P!wMo9EbvVp4@tMXxo>w`y=RRi
zYa5FbdP60}S4(rX@Q^ZpZ!kTo;B<gOKDQ!zGj20(sr)7F%hE}fI4{hL%`sk#9-fb1
zG#RK*GuEbEuWE*|vY5jR!15$qcV<_`v2WRnr2m<dd{VZ<m@`2%hzl&4P(brgo&wRv
zb9#+nU6|rAkzce=oAC!DS{(g{ZGfh2tXIB#J1r6L0m_A=4%>A-HN32A$Kqij)DnwK
zbLFeTv_{@hk*+A!cRQyirQ~Ts$XGSX0UlI#*g%?TJc~Q(vhK3lZO>gE%6{a>=MJ9_
zE@c0<Af$AHHOwG>_pv(g!DR6SY4%b5bvl}{UJrF_nP!*l$r7Ld^v{7d2Q;JBAq!6z
zEiAc$-(QE%j;-`b6)7t%&t>U%H9mzSNlJ)e1zFp%@3?exy8x*aKu^(0AuB!SHnyf!
zlJn}w=Sy4DI4j7%JxKHCOkv5?kNoV3UzAtcNh&w0G$rjlO!9b}7IhE9sK7qjE%P-F
zPQWg&=@s>wMKuEm*_&myBYiRJ{Ogmr1Y<njHROpv$5{fPgS<VV|5kVY$cS+1feye|
z7H;O-tU(dbt#bdp`5cw6A|E#-1icrJqQoQ~%C!Rn4{6hWIHkLhnC$6EAN!vwQg}Yr
zH0F*Iq`IoM?4iv%m*0K=vNLMIe<@VkoBL`oFl|5%EbjothB-sk#bz%lIMRR+BD69g
zTKdXD4VBXwXj_&0CWthm=g+?@s~!BD{6VmsoQ?m9=OE>;xd3E>1b`qs1!T%^LG`34
zdW0bn;){rIM1oc%>rPm4?k#u*=iZ7R7n&d>+dqaeByDA&)EjU_^Ta|2?clKiowuZ1
z3)89@f8$WGcW|Ch%dpT=3G#QZG{8%vWIYjE^84<~SE0ok`7GMZm)^(I<C$tSk~)X)
zy@9Yl*!W#y>9gCBdf7Elh2#-tx#_XF3tk`fjql`>Av>h{?Y`HkCPm@27a(i<+B>dj
z)3Wm8qm#*vzjw)c(BH$!XkuaKq4aXHb81L#0>Aho#q=6ptec|mtLPf4M9OGcXFyl^
zFw+X|fggR?*~CzsOBX@}IgS*`=Zm}-o)(2I8c!)1FXXFYqel67`FV1OeSNidB~!jp
zD<{9wnY__1BZjFybzm&&<Uw)zFcw_$`>qIaU@0ImBq6kEoNH0qSGh1Q=9ZkEMz|LF
zLX9rL$X&vIx|5=<N5(u;22Tt{zKCDvHekF7h-7ZI?2o2a&Ef14L8AYhz-v~TrE!r>
zYq<z>KlK>4rA<03UrRySINj8P#1I2t0>-q&xbe76r}ii|eTq=DJms#t#NUH<Y$UO{
z1_J*y5z8z|8|oMOoK)3iDT5}$ASD++?J9g$jAPZyFlFwb*3E9Avt2<P6-k@+J{EI2
zdwR=o@8ZDuy*4yPE5!Sp)3h)bJTBbin<$4c;tRtBIJar_%<$N9_AfbU!^CIU|L#8(
zUA^G<!QQtgK>7;LNJb#^8m%)rG(^d*O@zeFcytL1>qft7Z!SeFFA@g4T9Z_8N5DER
z@TsYQ_E5~JnWWD*J$60`*87-L7fM^oMbhIADja@=C>k%y{XOj~H=2_zxmNVpsq1@{
z6PW}eKMGH*#<PXIfIHR++t@&6^e=2vIqwdk^V@^FJ6M*%i0cZBbfsx>0f6WI!>(@}
zAMGE8vr&K0Uw%EOw@%wq3huUe<gOX_W_#%U3uw2!@~_N#NmVW`j91~hoI7?;f>}+T
z`lLQBKfFrh4li@9NMmPXO`4?6-^}TzQz4&J>&pShGRb5DZK>S1NGnbG+BrXBze`D_
zXUpSQ$)$T42A+|tudgkTNj_{l+*1lt^0T<t;{NVKptq0|?N7|syxjY73RHZs1rxpR
zYkquz@i;DiVV=sL-iNk^#-79}%$46V5Y}*Hrn)+v?Lhrm%!79na7G>pfmjdq#~EJ`
zk*`RstFEwTZg#<)D1#IL&iq>fp)6#nnq>T}lr+8eo5U%Q4h=~wCwXCQ@{RZV>X*LG
zQLvWv`D@GR#BqD*@oC$I49!Jo<LKI@Vf&aHdvmu4hGoM-WH?QQ0#>m8<{ns=)&N22
zH1Od)RG?c_%r9SH*UcVGy!;fd{Su44Fby>zR*$v5YyUIEI2+nY$47R*&srV!3x|im
z&;ENDl)cvvu1p`FeUX;_RR*`&a^zneyT{ysok{@`hxvX=a$;|^d<D~*d<jFn(|$hk
zaFVH@Z%^i(NGm5EHMbfG)ZN5m5a0#X8NgqQs?{Y64B+nByucn<K^K~QqHyT(xje2B
ztOU;hw)+;JXGQ)r*4g=AJ?s6Xe#z25?)z!aXeHJ2?dnUa>lLIZHUnfqQ51Fc2b4le
z6e~A&yghdwUASy~r@$w(KEFnEP}mSf;1rgRV*2^;I*59J-tjiP<6Acwz!KEI6+~Q!
z*`HD^a7<&%nX$Bsk7cEHws!~|($B9!-QZ^gOBgf6K{rUgmFQPP?CqrhL&t9RDzom`
zIb0A-c>SxtRiS(z9C1(N8peyZ-B05Q_kK<Ct}Y^_Lr?3U9ne|h23$8!ei&QW1>0_m
zG7|U>8dg1ch_C@$*{^?^U870;;9A6qUtkdAbif5_bP?OEBaE4v4`f-8Oez2+pI|N7
zpXiN!ZUOhe9=U!L&eEJT1c-}BiAVrDP1mn6Kf+H%ga#?&2JY9)`SGdTQG&v`AJboT
zQgbbUwbXyFLL>JR3DiK1pPkJ2$=UK(yMS?y%U(Ne@J$ec{O=hET>5T;{E;18!_ep=
z=o}3wTZeRd(eMv^DU}$u2y9qc7h1T%OIldp4A?W0;OjEthd|-h!Uq4I0)&!0jRTS8
zct?n*ACwmxph?L3u`P#}FI0-wugg*366u_UwclDsU6@Khiynw$A}IVRuLro9?0`N6
zy6k2DI~GRTS2=ppg8_aD2_t7FQ5zvUm`djcZ`3SD^03HH6+j51c6~w~YI_6GzFYeb
z(WEbeBOioy6V+f%UxPX4W|jQ^w}nAKg<~(ufpQVJH~dOXqQqd<=~orli#Q(8_-|8;
zpJ_p&iB(JLyqGsTwx1}31B-?RhPEcvLS&#46!=!G<lLCUtCw!`fe%4pmeSMi2i%8(
zzn6xY9;d0}aGl)%4O1(%Tm;t6^4;$o$h!bZ;ef^x8wG$GlrV8@77`ZT$o}_z)9Lz5
zzh0B^E~Vuu=wGmBtQQ{blurD_9v?bC&!qK6V6JVPUpEN_4e~t;NwOCs9RF#8Ex-jt
zS7ZRqe{{gW@cOt(rvmBgYB2$Os8bH}2`mfoBZh{bPSvy?(byF;>~*URb_NKZ`~Di5
zHQ%TsV|F$+bFXnI8JrCojQF>JpE2tir>vp|BD4Wh;VuoR?+%_`>rKOzouTP-IA}qe
z7b`89WZ{L)BFUEaV8=Ne0K2g;DL#5&62xEGcE_@t%63m2VON&1#-1;eok@-i8ahBS
zM)3R2!RL^xblc>w(cSVv+1Y$V6l};!$iG`+{7mnh92`8QibeIsmfw$5?_+H=nDzek
z@EuDPN7^~4zy9y4nD%1@HW;=-f7EUIUM2e}^r?3Uyb>KZu3d8yJDGOE*|_)r)qY>G
zn44{0soj|J)n>Y|bu9+(>-AfNn}%Ba7JBr=s|!=>yfywKW|`lY-uMujK5)-E=f7hL
zD@dO~O=CmfC&G;BBe6VhR}fWGf7Gz3hX1cIULXs7Ko;%5?|h%3w}%+cY{_U3&Rvsp
zqp5AG|N9EHICft_xt{T?k4Co$(1oP22V0*5G1wWG4F-BLA3*WHhhY>VO`o{<K}LEw
z`KcdAAsZ5XD6n!9dR_JfUBvZD$_mC|${I_T2Ll7s455mWfho2PwpFg_JQXh7XSCPZ
z?{5ANKb%*^Q8<RfwX<cZcodC3rkVF%d>1VF<FwLGYI{8;0REm4wmli7CG<C=h9sn8
zQJ<q+XQpO9vtT)i{xLVT$_xXO-pYLmGi+~|Mbu8;;56?Yr#SV~3qXJCA^+W88zOq(
zRY+gR5_NW0DY|@==5vuZ7XacHt^u|4bb`nR&NZ?6mo`oV^XUX#llTDN#!a?nkz?85
z*ndy(5Vr#>Bq$hgn}7=Va`tkR?vAqK<lxv_xgd|C7UIWY6|n2?GhRS%N^JVd636Xt
zS8Ar2nLe|RGEOrwaAKf>^Y6l$<fCuOe+4C%ENn6Ocwor-{Dr^xR8^-=MfLZlXZ}mr
zf5G8qA`y^}xxn|Zu^sb;zE8Qz@P)v;CI1TNLViR4J&vCW<OJn`;jP;&n+pw~MVpTk
zi9X+TvSL^-AF%cA8FghU(?5z(c=Zp4M4-KZgSNIA=t{o;DwbJdNjf|1n;oQpW*;Zu
z0iN7z^uJBk5{t$TUtD8+n4bb>G(Xziw&IeR%hMRA6I`*n+y8A;fCQilWqe1|?6Si|
z>}^V)#;l^EB6S`FH3sQ~P5Y`p{AX&l1R!WA`qTpCX`3HN&}Uf5rFF2{jCmq<eeR1;
zW_sKIy{KxW)Z1O0#a-jHqV@5#K78ZRLVe$>Ai{xt091MdcdX=rGnR9&6BfOWN(hXy
zoV?u4*fRMtO!5&!PTO~VbwNwn$pNgg6yy;Tp+AZpM|k&B($Dje8iTXW3QbY}HmlXl
z&jIYeWS+*fUZK9K7fx*7OOcz#NmtER{I>vS0bopRoLt@+4u##JTgrukPKJ0hR_VTI
zKFxoAv?agmGHMA6YJ68-&ynkN*>qx-Td8xjKVR5BIbkgjXn*$a!>JyX?3<-L33vm9
z$0<lbWdxdezSXq$Kiatae;=eLbc|uUY63~u{P+1Ofuve2Z9W#Y-j8LdJLtz|YHAbx
zEA7QglqLDN*=&narkEuWsasE|;*|K&!|_>RK(<KNq8T1ILj+n&AzwEOaAu|Die-n5
z8j{p*!~EG~e0pe>o=6xclg-2rOy{mo_wU$uOQx!@IgX2_s$zoRZut&wfK2AAJ;<u%
zz8c!QkNj`1xoDwiKYWlgnYldN^$Tj~J(FjVryG|DHY)%0zpwwh0dO4E(zRpyfpgpz
z^CDq~pW){;KKdbG=D@e0f6@lZ5nkW=hkRF=zkDZ4YimUfPOz;T#Hs)z^CoF#ib^K-
z-!(}gbg3(}iDQ61tZh*q6*sSZE7AqJ6bBMV<zX-y=%6yEdf}%YY5j}>=BbU9wCV&Q
zG?Ixd27Xm4B~=^hT$8^q5S&jn{@d~k^`v<`v;naZ&@L;W4}F;OgI{qGjtib0ercO8
zN)$Z~wnTo$a<$k@$NJ5(Ug;Q|Vz$({zCP@wHS5{z#Ms-Z$2z?z+dr;SBFyW~#IX6(
z^b?e^+qlk!Wzq}p&n`0+6B**^96xRd$YV`9iW`ofgDuS{XJ_Y$xjADIIE^~dGxyhm
z@oe|yOZ7~cn3yU}P#kiyvaK7xenr1cFwQDG?{rphd^X+NH@Wl*`})OxDO}lHIZZ-5
z{E^R3Ya`PY&`^yVy`{eeyc)U;jfi>#`Sn_03G;8qTOQ=@3X0TCOFppS5qpjULDwq3
zP^lUQ-~e6J!NPmqNszn2c`@)#+T}2Bn6;pGCa9wK+|<<6?X`!eXHjP64-hA6bIle4
zsyJ?JZ;z@_Df)rHDcts(-6R{73Id)Y-99K(Q)P9vfs&FE-05UzmX}gDaR@mNh6J=h
zrySNNs4)J+i|S+7TF9Kl1`~bR1gQ#mrml@$is<^|HA|ND>Ui5Cms!z|!B09LXn`pI
z`+6nK3A`_vW0@&GH#SQAaWhf$?V+$v3m}pXchGoDb!xinr3?bpDS*$PKaY-%-i*a=
zO;t}_k2P<^u*pe7Wshi_(QrN4tKLGT(OJ+fKM^|cBfd|dpn}NBb&Op-IXnz6>IE_G
z6X!=8IP?@_SzSO&cC=>DQGa<$d275g7In9y(8liQ_D^-L(N%h?Yfocr(a-19{QKN8
zNy=c^7BCGGpg&2xVLZg?I{1_}5FJaDW6X(A&Ao3wM`f(^)?fkzqH~;2jx|0!H9-c^
zaOxpmp-*o4h3nAA`g{QDTz7o5huA5RU9VogR(q{wt$eAGRl>n<*FI72W=lpw0`pt{
zWc{u5TIm2}25<F{nWzL|cO|a_*F~Jh{Xf_+OP=9x&eOj1au}a2e`_7yo*2K4QH#ko
z`uU$>d?40$i8@T$THh&Yz<J^FcDS(58vr&%MpX0Dv|<Q2mT{tsN!SQigme}O>P=sD
ziAzJt-n8<ZM9l2E5iVmfn)=s63R^FBJU%w;Mc9#;(Un<2x5v4ue)7Akewn|eAkjv?
zBd&l2FBVI<Ac)*H0Xh6g%^=oD-|zO^-Nc^Cg@!XAI@v@;T9v;wyn*vuXAQ?9&j!gW
zw0Mehr%#1xXC(jH0Vr;Hmoi;?6tE!Mp)W>Uf$S!w;G_vYJAo(J&b=+>T!~#zW8MXA
zi+-NC@bUMo{gxl+lX`|O2g1VmdFw8ltj&IX^45%V@%U$llS=!B*Vl{MRRGEF7*>#j
zZ-#*^^?Ufl-&I`XQHa@VaLR{><L<1J5}-1BZv=KfpAekW65z?NXL|IG3*mD(>xiw1
zlD7-oC#_H@+pbxRv~R#16_eA9F2q=M64RNcE0o<FgCTAOeF6!Gh_uH*Za~;9*hrx-
z3CcY!CwH1R>JPk~<#sq5N|`?XuzAeC20{O7$78d!vsHp!PoH7TVW#BcZle=s7bAsf
zlSvQ{6u{5Ir}ZJ<+q}_j`p~o^{8wwcK-<;(ZLEHWdBpuQNx|_ID&Ua~;-fGQBMq2Q
zUFSLvl9Sek0a>|Ci(r@WriyRnC%lahqJc{KXrP@&l{-WCg|oK&b9D&F>HCxhS$G`x
z`$qUn*FVlJG6Z*6PePF4OHDQVvR~BZ+P&HKGH+eB67#}tqm5N!clU8dh78`;woB~#
zl8<i~oqVToqt(??dupYKCUi%}Csei`b+JVS6jz%32trcQ#NqiFFE0Jja%YX2!5mjl
zCC?mYM>j`zm5i0VRUK_EOoSV=)GdVZv>J7ceX+V+ic^a$l56P+3O6gBv&pa`rSDD+
zpe1Z0-C^1>_p%dJyA2Lc_`C0b?BqCmHDKjp?wQ|h<NPgQ2erQv(kZ3K8meaAEU#|-
z&D=m)dDe>v{4BtNu%>HEqDRDwmg0oiCkFyulO)k|IC+fXzYuqMbia}8x8}(L(Onwu
zq1qyTA(pqPc?hYrgTPu=r-sIE&lvm^&h>A^?0uX;eZ2Mn)?oHrrw+vP`u9CjEKt+F
z{k)JGnXgnPn+(M3Qa++a3_Lzih6nBvMd!<w5OOyjp0gbUM=!Jn`yIV|WL@TF+u=+o
zdQ?^2a1Rj_9bb|6>&tDu2`)+bOV&A2R}rxqb*Rk3irxmg@`pQ)N!zW3dXuJ>IGA~}
z79y=SEhfaATq178a$%`y6l1%CchTKaWRyZ*#X$m8hU3nERTsXZoyA;yC-t+BCH&4}
z<SjuKT`JI2vpIQrIROLf;5CVICX7NOeE=avwBd};#ny2PxzDeey-3#pbl5;lU%Q4q
z)r)o+qHY8Nkb2zl0yYPmq3ci}G&VQ+CL6f5g2Jq9jT+ef6hwx<^j`@nCU&4+o|1Pq
zFS!%`@_lqTIJ{W5g}K-vH=-`2Jr0kHxk&5zv}7&qC8e`Kny!Q12Uc*~#LSjJ@m5*_
z`0v-7ET!wHdV%1Epkm*10!qmKt*l$JzMG=YvT*Zi^jZW0bBI@ji};c#fk3y-<a-5M
z?r8vDY-u$bek~XWr%r}A{@QdAXl8rj{jfMg)RYv7L5C|4)3AAr=O{8D3ooD1aM;87
z5o>AkI-d1#u}4Brvh8_Idy$O%a^`rqk<V~!h9<?_(qi<eEU5%7Y{}ZkZrfX2*~h$l
z>bNy1$3jZ^!q2f1XZ%CLC7Vii&wFg!3)eq1&m;OPz{eKqzJ=iT=rx$3%x?oVv`IW;
z_}o2jTnvB{tt|2u_hVg;L*rOc9GWbpv%emoEr6G4;(sSB(mrb{0^3sICYGP)3dJX?
z?u+4`U};h>Pfi0kuG2s`ICAvW(~^iAt&@hgyIs!P3kI!T>QZT3+yJ+X){T5|{bko|
zO|_kS-IB5P(A?1cdpWloDy7NqJGPCpsLpb{R+-k-CRI#(r}4u3AbFNC_Wn}UU>AS8
z=Z!IlHE=uRm_i<&%9~mw|BdirmX@qeG}ja93~AOgxelhBu&;(xd)Bm3Wm-JVONEf<
z+ip7=$x=Nfk|~4`t$G(dC5m60!YQKB;v+MD9XqP6Q67HX^eKy{^=1mYLBD64v`%)T
zuhvF$8t8^uu7hVjP=<>24utQKSyT<Zapx}j=Qq1PLkYQ9nPM?KnZvuCse@KJ5DGzc
zZ+`$Tim>nb@=BdCT*IAhHA)}_9E4k=@FLE_X%SMu+E)L$`88wkkk^^AU!dJ{;MV)u
zIBpgUMFmWr(D+&Yx>DeAcJkY@llm2L7m)ij$R(q&;6vxsBCGK1Zjf+5rJOk(n`k9R
z{LSbV#8n~0Xo5vm$JV9H6lQuszoKGoZ7tGU@aK~M7-O29<h;{FlZ5qd<9xtGw|VCf
zV~xED179QqMg_rhVjG>DH9tx5ej^fHmcZM|4U1m~NveW-q$+i|3BQ{jf#ZXIdxZt|
zd)9;R<)8WijKQzBc{S~|eU)(Xck4T0f8=GJ=^Ai4p2;3Z5cwkGSHfW7>4w(L*)k)v
zjz^z&yex7rhu^~Jo4<1{KSGF_UF~`AL7US*T*+IP2wDTCO`hAeoiitmClPr+HH#qB
zz~0^9U%{Jse=pt-MV2#a23p+fy|`6?ExP)1T*tst_0X<e3xxhD6M=oxczXCSz&zg6
zjq#pkKTdVJ-`htF*9no=;-~szMf$!!F~dp)v$*<%8&Atz6Srl0Y#h~p_0&ZPd+#b~
zgz<wuy=akuCh>tJbhlf*V)x1tAGuB19aM$~T?O-aO`58nmJ8Lgs2x}k3e^1NR!cbd
zKML0Rv;tx80mG<eB~9oN?#df2RP@(uIt}4ed-xsMy^6Pgz9S#$B(HPL7=+xL<lfal
zn=KDaoYyV+(GG_l-(=Wr_&zF=NcgYBn#Dxih2^zMHL2#UvD`H7d4xo$!!6Stk&c?S
zG`GXg?APTeRR`cLPq^u~PceNW+GrR>&TDPo;l_b3X@?1p)u7cofrnT<Yxr?~EJYa^
z9I6iOqW$7g(1<VrsaLBTKm}xS1fZDSt>?c;s-zGLXJd62%D8RoM>N4G7gdIL>a;FI
zL1lU=yUQ#^h~z;E#fY~z-kR2(w1NW<gt5KLO=GiLMl|y1K?M(=e~w3a9_<$jj&)p0
zete;;4*WLrNOZK<yazv<1w;I!!$YkX#ScPHNrk5=pm@_Yd3_%jkSm_*XRH81o|Wqf
zDg0TDAbjwzILEH+i1dQM#7WZ1d$dshSqF7A!1MqaKfXSZqtNIU#hUtM!eCP+bu%qO
zC`2YDEk@iDjAbVMDcP|p_fFTn=5SaD-eNUUjq*=KV}Dk7&fzy-7^<8kUetM=?ITNm
z>9m}E5~_{m@oE^b@wu&RnqL$HcH@*m2?h0@81(BWwHi~S!>_M0f3$>fv}Y*#g*nhm
z03xz+=~{90@G`@EObzd|(0k@{N<(tm6vI-`2Yv$E0wXnfd@iLQcKCOAGpj7ESt;)4
zc=if(bSe07bUc+f9yj$R`E@Hd)yJ-j(+rE?sPEK~ea5eT``4`>64?D-UVJs6sBSY3
zYbg<Sc3VOqOCAIATBi9sMUJvv;<?Qh5gDB6{g_@!ytF8;e~0rjs@mnaa;ju@?C-6(
ztF-GmivUVVWWR)K6v^aHA!;16l<#Wx7l+SC)Ags&*f^-pD!rUOiTHx>*IJav5QZr-
zo$iYh_7IS)L2^iq!IY*YTB&P}ey@k4TaugnTsqI??}LXw`=At3;c4M)BD&uU=?SXo
znu+6UkL9aYPa3h93xC0+swhWgP1?J96tOvDjmMQzYk{{ExIfs$+V#`b--Msd99LBb
zT6_+u)VfgTkLePnrC}M7DY1`~u8`XQ0uABl(C5TDGdo>vpQk;)ux}DN?VL>dcaFx*
z5+P)@^m9*wt2l^_2=vxsI^jDYjcGL)25D;>%bN$Khm3U-QRL~U)5}1R%OBtGq4R+>
zcVqdI_ucd~dIachxV!32gb9KGY=b;^`<r9|ozHewPl`uPs>+|wT+E!{LBo1bwGC12
z1BCnU>gC5h`8eV!*ZK;EQsTMp!hI7ZT1B7d4`#ZACkPYlP!pD^Jcx;^y^pvVz<Xga
zdXjjv7iFctI5tV)DuoB;=wRo7<0r{3Sa_*Cm2YPMSsshXMRbKL|Ii9<_9fB~f(2cY
z!5X6(OK(4*&$ide?L$K0O-~`n-Je-=7l&i_kjG)QY$rE`ltgu~wkec=tG$nE`1eFI
zG$)#}wC8rJTdG1X|0PfAo>8PV?{P?Z({Q8=17%0?ZJ>WM71RZ}`{Z#MUao3qjMu}^
zE%i&15m{hYjKAmOmB7>Yx>~wE<}1XbLXXoytm>w6XA-fWeMg~SgXKUin{EEm2VKMA
zHL9ntX_w!RsAsUA<#1DW;;Oimd2}+l2GGyv-X|C}>u^4|racqNERGf$W_iBdjk>Jm
z$2W}>iZ|sB;@2B{p3BI&pKRACT92~x7*3XMucVw2`^~SVrqKt_e65BE9r-RME+`}v
z=<|+45?R=mM;>>@>@?#L-{TJ#n&{Y(evfOx(PogZNo=H)>`4~o&Ig7D*LylbV&C=?
zND~`+l}GOa%Cpl&L)zK43jKc06AW>P4c_N?ZlBwEm!z9-NroL!v`he%qD19mwF0?|
zJ~Fre1V3|Qw9IYWkw>}VOKC9T0qSW?jMv8}Ll!ywH)%MXv%i!`T5WxPRN5!L*svlY
z4Im5kl1?>uQ>D3e%5E4dWAux)Fd9=x=x$vWcZJuh#mrIR2R8dxw6IFB!<ZC~!&qrC
z<-$p7!wL}C&=ZER!oAn&MzqUhJ^&$0O@`d84n_jL!Zev6<1&oUO(^oE^5>9@Y@|!?
zQ(l)W?O*zFunI)Xy?ghhB7c4*81)^McK;?COP;N~n`*Ki`aFNg2mhFbk{je;-I-+>
zW$gIcn#Z1fDvm@tvglJ~30{c}4s_ObUKp+s0X!5siK|~~LaB^2!!eoX*a-207q+77
z`>5+W85m=I@P!)x#(96Q!S>-?@Nap)C3JJ(TtYsALrnfduSulGO$8kizn(jJ(nBF6
zw7PEcwA<c7T=y0AX<ijkQV*dF^0bvx`SVM;AbP})UGqDhPk5-N&t$A%ztwFM!L+g@
zo@jkjFw$VnIHpt;KkIY~u()QGRgL|Sx;CJH&y(QfrZaAlCjWf}RxI|1O4LPL7G!vo
z=Q@4MD1=5e#;{)WC2+=;4D;)SsMZ|@cQ{zZGgovLFhaka3Lq8LN7Bhtk_O{B*ivn9
zacPtds};J`-LxEslzDKPh`V9MhjG2+>Gj7(qm147`JWRHpAO||qBuD;=h7gA&7IGk
zIGa%AH*+rqAW#~4(LchYh(@yb?XBjW7SYh}@7-Xn_u{`~pv>AD8!xh?hrU?+tAR^T
zJeC)d%0AW=dwCdxo{GIS!e17dkD!f@R7|)LdD;^B!$TBr9YKL8fz8!R*J$~h2{`Yn
zw8zdR7d)q=O%>oaq>wJemxtVe^c4RI5R!YK$=$Ac--ISvS6*IOA@6l>)Y8%@bcL-0
zH7%ee@4)z?_Oqv9{NsMwp@=1xvIyr-R{4<i9$s5i|My`0%C7(NI(;(O=KhoQ;EQUG
zH#i{>vi*0^6fi?&&xLS{-eEt>X`<ovcf1W*7-((aXlAVq(R9I`>r{Py`;FtpR(0??
z`w8#Uq1L)5yq0%#NwlAmv`2q5?VvYrZ(ZW>qHwC%;QM~VF<Z-YMh}(gGmnj3cbB$p
zm0;nS(1b9e23{|*$S}!J3|NXrPA;x6GBcA|uoTCiPfHc52OwF4brtg;61$x3Y`(Co
zr0~ecT+bg!adkSNk9_2db{){Q0eOV+@K{ag6OSF^1)j>BXd(h%3SCj{Ct*2Iis@;m
z0r7dCj|X@69O*AQC)<Qs&Ep?e+&X1k)gnw}%3OGwMNA1-eHb#3$|pL&38PXVfHCcK
zgKU&>R9_q{EUrdqzG+>EE8Ilz`T!KxmyNfGMM^)0(?mG5!yLh?x%pXn?4c=Fk9Ic^
z&_y$ffB}lc$F0<KJrU!PGcIvk`TsoVzKZ45<hXaP`sY4&Vjl&WJbQ{u&7<7kM!G)R
zbR_W`GB{8sYT7<3IitR`WvFDS13g~6yj>>5AnYOl@xkg>_4Ba^cH=BfoVI*g8vRim
zLJOfp2b~3v_mS9oS_GC$P%$GY|4uY8`=_o2^P|UZig?R2PK#sFNdDDRKN@kN3T*wk
zM2x&#rOvx6-mx2=6hK2m!+c+X8S?nSy<mYDIz)WWU*f4$(!V`^@Wc*)NLtj?_+4A$
zrnogr<7tAKf|FqAZTMHpiHQk%zjQ?s68h3kT!^u`;T9U!Y*q-?FGq`CzO0(Tjbq1u
zdPgeiv0|v+)xaax#kX;nDBH33$F)Uj6VK#arix#9mETK#$Vq`iVV@@_lhG_?rAjgT
z=MPhdgNw5>Q+x&C8v~W`jp)v}Q;;oMLN6>K+|U=$8*^@!sNe3aDq0C)V&1G|X_&{-
zv1Q-YxF9qoB?rYygUMN{Uevh+S`@_cbm=Q$O$%;zqw=jP%zqzu%xL6gGU>ZU-^<OY
z87p1p2%O~U(A->`)W(&%4(2yv(2P1S-;JJ^{t$~}ehw3dHEJ2ggeF+GFUqWLY!HW1
z`)Y{#!&9u-5CF~F2wX3AfRo_sdDP=nLLhIuw*)_6?)_9+<>&jvR|PM4$FWyKu3KR5
zM0&ryT==;kjBy{Nh%0>9L*A_T=+g~KJn6`<r{^2R<Q;Z-zQx1^9>2!sI)x>yimpCa
zmAnZOU%B@Z)o{^{=tg-(zOoZ*;w7Rc5UzwS?T{1R@g!Sf^K`0Jv&q;WaM$e*sQ%o{
z_)&;}P;vT3cS7OVLbNUxI~KW8WNL{7XF{$6H^maD;07IY33>5V%uytX`SwjFa?twA
z@5+CVp1(n@^GY)e5lh6a9pm^dRda-6B1Ak2;qAEGe_|2`bFzE2pLSc6wbU-RsJv;m
zyeF>szIA?4^_ny3aXHu@svf=ijj@>;9$nEKoTQ8aM2w#hX!zc@=ys^|^^Qt)^AhF5
z@&vB2Y~0~iYFz(1Oipf7<xV=>A%V3pNTZLtn{#`@t>5y(<IjeX@TkXw_z?~bQmCk(
z6i>J0S9NYt4;?l;(T$o!sCuXsFUJ>|GFtL4TDHPJe1g1Y(hB}QB8>lMxU>jviEHTp
zshw>sb6VE$&eUh=()vP3K<LHotV|srT~$>Dw|$DM-Y+RRLp*H_g;PVGd2<6zlA7vf
zSS>CEP0b`k?0E6UBg?djOPSn`eG3wR)}6*h?q%Io3<{$=cKJ)8`Y|D2$~a^6Y>eV~
z!^sT82jH+x5tAkndi$&aL^^DNdbi9@9o22|SiL`39y{@~#d~Ua`)_j7Wqo1eD*?YH
zMfd~xjTL_#BbRALXcr$UG-j`3emqAXX5&VNvJE0iQ?d@Crz=PX%XHtskT4?RZD9*z
z!S<gXv!**itF|oLW2xN#9GW9iT|RL2z#y$ku0fh!0ZnZL7AduY+}OKvD{$hFA&uvt
z$Z&)^{Re?ukLcYS0aVdb)UGBuCkcvfjB$y60DN%gZmyL}-Qo8{Rkj<+8XrH-p=?)(
zk3Wj(>b-lzqFA3GkNiRuoL-~FwrUfmhy|Ib=~Bf!uo}r<w~v_b<X@uN$3|Id4x4Oq
zB`Kf|xPNm1J$Sl~=vuPU<9&)O(gB7x1Fg@>LX>b!V|0lY#?(NkR;}K5bq3TtwS?p%
zutB+Zzn%|#H!bFUTiNAtbtD?9v<D0PIP_o7JwATiV`ykwq1U47vB`@cI7!t|y%@wz
zoQGzm4ubkw5l|aDyK1ltq0V&2l^1vpS;fAd*18i*!E<0m0os2`3Dx+JXfto0DS_jf
z3oF6X&czpFi(Jj_5yP7US36fbe*4*w%!)Qv!~lya`RrN0YtV&yzjl`*o*9E;+>~9D
z_@--~!Tm6EENUwrt7hxbXHa);`$-quS)ItYXC6iMEE-5iwEwbmixM^@oZn~@mLFAv
z|JXHm&W#L|0>>&~L|4GK6a@J%iDFS9{dzWCkIn`nRG!@;y>6ym_dj()jeucsJ+FM*
zY6iK8NU?>o0>5bVkbgMNgOPDM0&PWY0&-vp3-&C?U`Td(Hhy2WRm2_}|DE>JQb$!x
zElXYp#+Ng4lZYmR{wShTWn6C$95+Y}5oZpcd~W!OR6jZv7Q$hcE1j$SapQ7NG3`Kg
zXh6KfABP>;b9GNO8JL}@n~|uRw_rW9JHGPzGD<m07Z%6q{dts{z*noxH6%(Nyl$WK
zjSyPcJv93Cejwpj^EpftsA^{4ukANvdztt&#BN&{!HD6!GispHa=DZ2IjYxmQ7eD-
zuM3uOn}5L<9@^b>+P6=~mAwmvugw|XCkx^Z1>k$L;9xKu%Iyol=Qw~YmXC-;%uXH>
zO06WNa5)QBPGN+P<GZ!2BbTnmUYtkWJ(T(D^CQu(*5-qYujD(*xv??K_+*UDg_H1y
zE-JWA=PphS*f_x<lq-2mqrj@fb(OPE$C+y16bBYLosa9qF0uVV!Tx^Mll-yRS2W19
z<VK9s#-#2YEzJs0tzy9|3bLuC$+~A4gWkWOjf-0L(g8X!8NfI=1!?JCKuwOW`CGH{
z7&YzMUssscG@W>f>KZXpyO1zS3zr8$U)RNOEBZTx-Qa0`J5X7PIDo>rz6gId6YLq{
z$!cvLxE7sKr#I<$W)GLUeahPH52SO8#y^L+H4y*`E^7^e1PbxtH!$8g&fa|30?}?F
z$}OORBxQLEz*gBcTy1TCY(-@{8zVA6A_E(0*23UdPh{r%nx8L<W^UodvY#t|g(^sl
zph7YwC-y(sZc3PyuP58=!mK%C>rP=d>k=u47GKw+)`@um`B_CFwD<_(gYFC}T<;C;
zyeN<E7%Dj(7f4Ia1#&-(k}ePsaE5RkSAQ5m*|*UcUf_LK!EyP|zPh!VwTP?>k6pDz
zk{m^7zb_+;%x&#p7N;Cibj*!vxuFP1e1lQAQBt!ibohRs_Eay=Z`|`NZw%|s)@r%i
zJ-%b-g-De8%)Qsq^m*+s8r8Tp@#}{zF0XG)fUj^vDT2e0f~_@a>kV4wql6u`>K$(m
zFsf0-pjPfL%}!N!_uGCl$GI;jF|iXy+iX{O;b2KGPpBobDTZh%tjtRnF-h}vGv6^0
zryj`Qb>)KqxS#|pe$UKG_Ngf1%;K|@V1}UbACCj+6(a_E?X8#B>S5CDo{%MJ=)vkd
z)3L1`E*)V*kB|bHH6;aOcg*^YySkd%1o{Y&w=y;L>(x?%!-#PI-wt$4kD-wNEN?7x
zyFm4`qS75tXjMV+Vnyt6eW*c^?gx*plltI`wCAJL^5{=yLz?m;`M({wfsH33S63wN
z!Xu~;9FJe#i8CCPQO00hPI_CE9c8`!Ik9XxkD<q%_*IkG>Ba@^{lA6g%OJDq+{8{M
zv5!zbu&~pM`Ek&`L(ETY5XBAAx>CR_p6L({uxtspQM)CVtSq|5IeH(}U80^E6pS3C
zyoX#XO8BX-(H+7a9S-ig#`-+&ntJ6Id*rK9e0RtO`qgnKBBcvKhzn%K?&9WtKqjOU
zI3Z#i6;oR6xoPw-+24L&%l>ttb|lHJ;U9i}@(^?P&~K7n>m5ItS?VF@LvgDp5kiZ=
zy|;M^l%qHseu_o*>x3+trj+E|4S8t=@&rsSXZ&F|MtfR(Q(Hrtrr(64ZL4@s$K1b(
zt4!&7F)zZx#CGZhwA#;{DyyKo7&*s?+1l8yO1G-Pb)7p6SdG0(ka{KhGE-2BH2PIj
z)XN)XTN72{z9Dg78SNUFrv07;`M&oF%F*5$Gq0o`y3{P29c&nd2#{D(ToXd@-A2Ar
zrgVzsSzgIj^dmxlP0Y6g{}2TlRE-fzIC^>{sa-a>6NE}TEVb6c4g_!&+=0bAttZ=h
zkFj1H6b+2~aLy6%1yh|t?A742HEQx>y!E@%wmvy$$_`+?bJH|0_W39ztPm%47u&Or
zzg1nYUGT%8YWJ<^A0fZ_CHuE+KF&YM{qywmZ<OfpS@<JZ6&6hk7WkttBg0>sv*aV%
zoJeVqSrovg={0I<z+|r?S#ZGQ+wOGucz)y+ZhXSO)?)?{8Zx0DF}yXL97YPUj)e^%
zm|#jG6}n+}naWuwfiJYkESMtfsk}p8GzC}8Wmc-jTlR05xJx$Nz&Yf3n$?iZ+q=HE
z2p?Fbj5*OKD)DPtIUFco#KhUX<V}SBEl_^5O^acmu+Md<d1-}P^6VjI(LHPgtL3Tp
zt%R&ua#P=ScBbxrqMUSpB<$53$uTbWJ8#8~bF1sr_#@>JB72Y`C)gtHDwW%|@(%EY
z`m+vCYL24j$Q&l>Fb#tW*;3?NF~5qwEYG%V$!I`i4O)l$Ge4^M;~^VNY5Dp;U-(ep
z2sp#eiv2Q2RV))0%N(1%C9r%tLg}|M^|tPwmeRs$l_Mvx8fB*A$esj|r9i|~PzYwx
z#R=;B!eRJL<SoLnn_AznwO>khVt$Af4dAp4lB7kjzI^HC;N^aCaNLYh2^@$(6r^TE
zr>i9;5X$0MaW{9zyxDxAb+u-j^Aq+|rEDEFe}#>SQvvmR4qa~KpXgUuNWuG#bkMU$
z{C*(P8exkY@(cn}l1S?eKW96$7%;)ki{9ohLdYCQep3_~)_Yc%12)zlhbJwGD%Yki
z2$WWBpuJ!HV3Tmj987D{unqcGYbp%W0z{Ly*s{Pq_{oWLNb?-$Zs-%i5jb}=EQoP6
z9-fyY?FXF$Wz%UEKiE8d27v?c7d0n&o$>Jx;0VBigyE|uw_PY5!P_0%@~9%z_@{>R
zv@A6|GwE`Mimi3qY8J+E#-%@WRav_@bE)kTF53)c=CKo2Ov4H6kY+GeS5bbfxn)ql
zLt2k*yw7}i)coCYU(%xX?<uobeZ(>w19MqwKoNlG=3=Ll3dItuo{hXB=+c+nS9=hK
zkIc#|2%`Zh43>S!4i5IF+fP|lsTFmpOMfLCi8Zc(coTNpk2n7LL*;zRbl34eZ>J36
zzQ;aO@Ndrvj#__Li?7NB!eHbvQ&Us^H;l6&mAFzOW!+qH#qZR817&WP)|@W9_djqM
zm9n|C`uqgvSc4qqdRWOx2;~Pw$BF>#<o6lO>oFzVyef8UT(sNbndx2l&~hFh_r5Z6
zL2+Tg%X2bO^N4YUXh-bS)AbU&GAd;9g5-4}fyv9X`TRJ$%EHQE$|`7^OcIf3M_ku(
zL7+_-Gk$IqU#&_~NcL?|%zbWQ(UAjKNX1QRd;Y4>o*PS=8u1Cre=+izcl_kojwGlk
zQxqIzzpfkRQ~XF>?YB<+YXn3^!~l67Ga_BH`w7(n0Jrw;2WMg+6G(#b2?}aCv@kd9
zANs_#sTbuR&e&;@_Mv4vRHbj?v=rjBo{ku8PFu{{=rhG;Q-9U75m;rWuV@<IOA+4`
z&pAwsL<qtV=pa~goWGEULT0aG!zf$<J^)W~bJ2{rVz@(x75`hr@pAe&#hH|xL#gkL
zfN0&D*K|*7*)@2Y-6+wOE;H^c`dJ0QuE5-h6o^H#<#saK{?*dLJoVejP5Isjb_i05
zLHFg$wD`Ko_hFK8H^&d2<u!y|VA39!i7C&B<lOOus{V5&#><&N7j9f0h_5=HxTiL0
z^s!+CK}Sr7lPgjc(N;074ag%DFd9&^H&D8`Zl?PO;+szM_Xe-3q}x*zMfVlBx#Gu(
zC?c?53KM79wRdd2Xzi6y;|8Yi&LFB|bjQP#kJGm2#4{%iyz^S<aiV4?`Ilm5j1U*0
zB$?;3nN+NwR!!t8sXHh1Z;zp7><qE)*q5u8M|W<hbjv>ZOGLQ*(9YJR(7baO`C#WG
z_(U+WA=;ftMX<jm@I;AbMfb2&AvWY@<L1eUZCR6(kUV0o{DEaaSI6{CZG+mXN*Ygv
z5;cBroCS6oefrsM5G1(8w&Alr&}q4A`QMq=t5ky}4*y#E`^;dPx<pJLX^9(b$t^8k
z^9EO5Te}w_;yQuS)zwvpn&kl@-I(=nSFEcgUz&Mw<@a}?>7Tj-(269@dCf!kiVrl1
zsKipvbt|@hD(zDhAn~7Qy9D)?RZr;}`dHx#pKA27+u_2SzaeOycw#mrFQ}{|%*g!^
z@tY#wG1n!OmdmJ~MdU&c+XeaG*R#6{I8n@#{0|`<A1w`y>ObbX!1o<#I12tYtQ59N
z;%J5l+->)uyT>J^b6yc-S5-*3WI8&>xcGRHd#wARLzG`h(B{(?Suxd5o~bY2i~V6(
zY^uK$H`rQn@K$*d+o1bj7?H5LZ}n@k-#_65nycahqg2*VnE#SfmZAWp#MYLPBv=vq
z_Sq8OCxPH!)!fON9TZAKhF!c+fB}*`gmm}V<jeB9z)qA>E@YL3o@`|N(yj5Z1rD)j
zki^%z!n_^GZ}m;=i1NP(O*#I)@r{9j<_@Fy0_m%Oql8$#_dx~N)$W<uO&|VxEE~Jx
z<J)V8dl6CN+1o!9BB&J-Yd_{Ty1DeaH%X3uN)PlsAff@R(A1$^7H^}falGGLlpiZk
zF&jLzb@{vKo_C&U!{hqQAVk>fX3cE0e4fj>+fB)>A*{uuix~gthfT(t|Hbp79*~p~
zkv|f3M}!6wr6*hy>Fled?TqVShEL}E@TzK=2+9s>zH{pTA%{N7_YzpYWow|8EdFxW
zNS^(IeU4*~bAIu5US371K_M-7s8`AddoI|NwHL|n8)@Mh&%V(ALV#Lki%1pC<0HSz
zF-oo1i*%}^u#@=rQSp~6W4A(?7X?^lbM<oIz-oJ^{n5*lZ?2tt!SzS^Voqb&xHrxj
z11bC^a(EX+jX1T4>m)5NJ?@u@{PQ~qgd}GG#Q*_9mb+59eg`}By&p3}*vznUuXI^p
z;FbX1PtB2zq=(NtC|&zGxv}zC4eqR1Os8ig(N5GuNfTlQ3xw|Mf6b;IqD)*-Zo)N|
z3zKmWpE|xr4Y}j{upxtSp90Ww?XBdN8B!~0vnG-bINLYyICyOs?%T^A>ut~y5R*^T
zt&&9trMxt0l_kcpW?vp)+PS>{(ch;}W#~2MW_y+CDMY!Tp!&%SIfjrEf@z`h{UrTf
zm1quM3LU|TAF40()4F>2-Bph6=StZrl2GgXHvbsZ9R7vyl+I3P+FI;o45iYd91-Xf
z8eRRF?1yVFeJr}C4<<Z!KQ|+bHR}PmuSiafW{lY>QvC{>RBh_jLb@8%aAjFq?Pn<A
z)S?kqTG(gH3hB|tu*k^|ed}lkf%^=w9k$sQtc{qH{`jiW3w5a@uC;&FGg$H`BY^xn
zEg*wb<OL|jJT6!B5UclB2GKEqJf-pX^#z91Rc{nF`ztq^E`SOx@Z*+gv!gJIZTH7!
z)G_WTp|45)SJp@S)ucxkGU1)JFUDHT*!a~xd=R?$4e0|MGCSfQ1>S+5l#YKTb$ZJE
z`5CE$c+BR3GN_#x6`SbxRnP86phb#<siZ{B$Y0Omlcl7{5xpt<CQf8<XY5bIFIUW_
zPq%rQnJL_tw?SFqjsYU61gj5Lr6gCKKHH7sr333$8(TIkI1`)2II4JNdD7e~Jt)CK
z^J2L`|682&gt~ejx%H#(nC%xTrM&<1ZZ=`6ur8%^F>%5$?X?d^fIK$lYtzPy9U=uI
z(SCMLV?_l(opAFYY=>5S!gJ7oSN>R2+pBi=Q0EH(r&Am3wW8PYlgIF5GS$RK#)}q~
z(turqBI`fVTY(CBUCdA0YR~7+j&EM<AFssrTSwtpZ~--AdD%RANlMtYXBOnre<Lhp
zxgAMv=d`5lbw+KL6>mPL&qfRWRyA6VsSKK!&H)n;u#3mZq^c<W6Z0cg?Tc>Yh4-FQ
z%Om3*lO|<zy<U%W$}78zhvU`9-^!n=3}1=UD|n=8T=B~7<E}~C-UiWmy6WCph}^}4
zj)UqR1=>kF?w=`||5A(SH^>LG1w(1G=izlXsyUc6I(><IQP-_XZsK4c?hdT-MOJw+
z-rtUG*9cC}DCYiNN&Z&>?-U#cYG<PX*zDmj8&-Asn+<*BtHl-N<hx5dLfTk0_al@~
zW)voQ$w)oF&57aD)*IpM^Drf?4az*AP~ClT;z{U}_VN@^dovw36y%$dV|<@z)n{Ov
z@#!L-iD#>}D&Z@*?5HPn$l(Q;qxAjd+)9s&5lT=!la9q7lokUzJeDW%ng<jX)RrYj
za+SC3|BPRry{H<|rc*IGs{ZvhK~+0mapILlE`=`|oKzI^SG<>X_WGOpI8ZTLQA-4m
z-+%g_wZ8#}f2H+S8n|&02Nvt!?gW2RTEe9ucjLZ?U`d1;d8}obuiyiBVy*_Xo_$DM
zA*E}ov&*b&F1}UJm-T&CYdB^>`g_};;WFO&U>@~twN7zHX?XkzE}@zd>ZAAXpW}~S
zkbPFt!4a<{q8VU4m}do2_jSM^XFXcP+}A9;IoZe}RW4b22$c#MrI)qK^N2i<Ens6E
zH2%#Nzycevr`KZAe^%abv|;DcWHxK_f(nyjbr-%w87g#-bMq*VWM-2jYgG#)8s)~U
znfFTiXm;#Jb;EuX{nFzQIf<14apuQ=BtYU~xY8RB?q|IMe>U8Y`L0ZDzdpoeX85u3
z7nL@(uf^p?(tGx=NZ_Y{J7)y^JO;;gwu-$aj?`K0RkUetLs;qR!MQ)bTWDCNp*-Lx
z_3bVWC-e~_xPk47YKd`GVlfVB-656!uA0|fco8JeFP3g@0XI6odZfSPmW*`R-Op)M
z@zO*f4Bbx}?|e-q>P)0#>kM6|{5Acuf4D)|*<Tk=fkM+=X4iYQTNb0*`lErjv+XXf
z-bJB6D$df{V}?7Fh*+!CHO~5>3vJGKWM`k0%+KAwKQrFFeF{=q*8eV*OxovBO@j|E
zHhxTo+?6EGuXt7b!EB?2P<*lx^XlTjgh(cHN8Bq_+%k#^07tW4OGE^(d<v*DynJ~o
z3ZIk5)Xv^9XfUF$)>}NO>y=o+jU;`G8PjWBV*t4(OnDPA@9nVOfAFNZf|q{yv7He!
zryB~d7}l9}Tt<8S=E!GOpMQERhy3;Dk1hCbbYw&*&)c2<2K$}c#y|5Z1)tYMg@4Li
z@k+x-kuo3h<6kx-?P^w5)Y7a!p3~*(u`6mEW!>uZ&f%<RXb_)1>E8BVN`2{bwwysO
z$W8r<A&C9f#mVrKsm&kOC^O<uJ;js(T9tpa2`%b)P6zJQde|7bjc)j!h*Q1*5$jbb
ztk1^kDwjdEeQ($NXT`(1*{f|N&^8(@*nx7-ag7yE!SXZl2qP7%L|D&`%XzDKEba-u
zeG<l>I6W)k@FgdwlZ@#R^=L}`|C;-*s3xLt-2kBq(nLU{M`_X&1nEVZARt|-3W!pq
zNiP8esY>sdfGAap(rYNvg;1pTqLCVUfRN-4dLGVNx1NW)?#;twk~Onu&(37#-~ay=
zqCL%7`uh@2Vcr==+g$GQ=hL+gTbzMws16mTEbL(2o7V{ciRs{D7^7zBEtp?$Jg>(4
zse*^u&VYDXa&YSkB~g2a0y3CWgSapS@0sQSCwQP@+|WqJjDn4S&zh%!`y;Ejzq7Zn
zd$*ADpzb0cJRqiXre|RQ)3GFlA08j;&Y(FT<^uDm2UXXUr3{Kx<otaLaoMKc<f<X*
zu2pTPezB&3C|{fX@~U))tLN-OOI_#A0JD-C#K*t2xtT2y%|9-cbX$iTjUeEM8M8_-
zish?i3mf}vPSBN6LTK0~^pxmX(O{2pqQru<>&vHa16{s(Gi&DI@xe|mE;Nu3<Kja&
z)nTN<Vcyam`ga6_GMwhz@X%h8c08;2<5%UY4PgQT$nV*|0j1lNPc!HSLXqh6p&ZtU
zsrT%%HwSY>{F}w>*W}TD@j(|Ts%?$pV;J}1Sh0*M7}xb4^=0t`e$&xfeD+r`re0=i
zW*Rk>jg+Zyv-<h`+s^djJyT#P*)+Xj2YoY}nKFN?&L)?7(|!g^bVhg{i<>T=$9pNV
zvvn2E{KtUi+`mS-FIb^zeGlubr;rw&E6GHpq_3nO#Sj9bX9(en$o>2?lzS0g3x2{N
zMy{Hbk$Njj3A`%ouZV!!BqLW|`KyJQCG%Oakkn04rUQlF!Urcn^eC+9LFt`ID=#5o
zM?ylfT$6&y`NUWy?Ee+rgl@IzSNBLTAoHI}-cX!7507t|bH1~Ajc+_|`y`piv8<EN
zb8CNN%eabphI9;N56fU;BUb-B0BAV6AYt;{;-%r4+kDu7{p<O+@<s=`dIXXQ%Ftk~
zlh%!~8?DpdZWfg~4(tYhJ3GAxOpviB4fIXz80=#O8s0NIJDh@2(YHyGpBf1XTc^q5
zj#u&@QU3w3M*?wFq}oby-!ON4@T&@bp<kgI+k{Zxiak_Nn-!|RN%pFhgtwKQ+BZ*h
z!=&6SKjP-9^Sf0IKm63DVO`kwbv9qNpzkeKcWvv<Oe;p9e<#7=wq#lh`K<$kAnQ${
zK0wDPsdqG%L^a7ENs-u#xC7u)q%<2S19!WV7Ictn!RYrXpP|XKGg$g|#BqzQvbQ^3
zaBuQhi!O2VL@IM^wgaClA_Tm&0^?<j@QseP?q3tF<rRA#3(CO;;f^NbZ;%z+<u=n=
z`;1lTxdu1ebl-h4v3hTvc3~a46xz;#J-sm*F8C~nhY8e2tECLkV>B81^CThKsq`-L
zMjY$+)a@A__81)F6#@DlpN`yf*??J>Cqb&yhxl1A9+J=+U$3=H<C!MCT>avUy}<Ya
z$@)OVOl<!lz#8)C5ZH0bxl`>kAAoLAK8{}z68b#3_laOVPsv(6lg0OfbQ$48Ln|{P
zHMygRi8)`3iklIyU!>#cB}h3>(e3yM%m#k!A^s$j3LP>tdLbZBsos6l@_M!Pfr%)!
zu^-tu=5O6ZSz32L6iDqgUQX9CWP0{WBGTcZ$_>!B*LDq%Pk@xOnG&&BPyLlXMJDbN
z#=3cL^8(&zb<a~Oq&mC-r)Nd;eLIt@Qkt6rGgr#!$5?dDm9+lnELc71MquLWJi*SO
zctfHSISiY<l1L{;mTVCm)T@xVkd#0lZriHHrX(Za4Quuv(y$2j9gJ5AGoj?oedfAL
z&-qw#?9F60LpAaDfFU6y2+6HA4QTcX!N5@nwoSG^#Z;@D9=}?{cWMJn6pHEr`2H-;
zGtC}({DNev%Rcny!FokgkZYqo-Vdm3cTagHg(WP@hl1=X_)a4a?UBp<{>W&jY;^QG
zFC~98zl%nOO~LKZ>lOGjb3Q2CTR>)QwYUDL4HLPvikKJki5A|+DrO}*hF1)tS)#2d
zK(Otf5YS>e44MQLBpjYsO?v+R1XE!&jwP(bG$EVz7Uf)B0H@Q7%Bixn*2_Q7(}5E}
zbeF1hlCpaBhlD<Wq;gv&e&z%O;xg}7*Gpw6CD^6JNQ`&fB1vJ1lXLeqWx;AxI0QW+
zpdNKQ<NaS0O%B2DPL^lY@V`V$Es7rh8r^AA^NItI0rVJXAXiWUCJV#VAdJZb<+RRz
z>O=erc**^(wHbF0yWU==fd6u@uea+;k<SfXpWU?=b(_tKEcv}(qU%p;H0jJPL2!VX
z9&IovLHQzI^6>IwSGif~_sQW2ai4S)e65P{W6xEZXv?nGNY*fooRJbW%sYG<g9oq9
z-LSoVbOjBa+bq`MCFN6gfb%KK8!{Q}A?#_Cq<7_k3Y3DzK9b@|FR0bE|LCC*#bv^x
z`(k><>FW<mxZBGMg#KK!A#-Zhs#h(%tSDF&0zF$NO1C9Isjwq^Uo>0>$=NFNxOV)*
zAJ*??nB?-d_1E^K(6qyUL6RGAFO~kol*fh=t&$p26fM4I={_7n6|zCs$3L#=f7S4?
zu!oAY{5y;CUr2DD%7pUMlo0%EE6pq78iBb<+eDFvUS3ZvrWQiW_4fu%!#uR!7zJAs
z+ns%Ez|19^(fr0mP2M+NUA2Tf0SEBYJR9d+7jEb3BfOJZU||?aL70E}=(7E?s-gjm
z$sJhNDj8_u36P-F-{kFR?8vAjR7xyTC2ZV_yXnFY5_|cEqGWB2GhbXV9>Se&;h|+V
znWp2Nla7Z@!VeUE>h><f*OZX%YGsmyQz>uuZ`ml(LA_9|EkZ|Y2rUKnGW318q0uGw
z(vQ$DvCz%{ovv)JACvY_#Ckz?Sly)nT257&0^R%5b=>}mdC-mkLMG_Ws`TwcF<fVm
zVj4CGx*2w<d+6Pl+17egPN~kbB}wfF%Q*}Xg`PJ7F7~RirV}t3zDCr^dZ6Rj>M=l{
z)jhduOuhUU%%;pAZ!<${$S^wTYklUoLn@Y0$s^yYm#h>Ki`TLHcvz$gwtM`pU=psN
zPpq0m_@cqU)FcNDLTW2B1oB9^B~ei4pEWaJc05H6V&*i9AO)=`2!76|H$7IQk)Mtj
z>|(S|QA9bMuY<X7M;$6{9ShY%N{&kryNQ&z25zUU!n;dFO&Ts?t#{1mcU(RTDAPL}
z@YeYWwdI7}mrj+FCEGQ6THM=h67nD!cM$V4V+~%*F~Nll7<F{bvqxQ&ZcF3K?&g;l
ze@r4woz)<N_+FX&y6-}@9-gWDDST%PS~mKnb$#0Nqn@gSpnRs+0PIb_E#TCpY#F|m
zQ=*v9IsNA3?DLdQo6=MHK<8Ct7_rTJG?tI&>;>lW-JtjLC<Rc%$BDQvXvtMkt!}?s
z?x`xrmL;d%1$T4<n!n{468fMMVLIcw?-+xWCBi8B$Z;NPYY|BE<Q`R+1NL^5aj)GT
zSdV{QY^~pZjNLLuS9%8CJ=I(>2{B{m3E|6FN^P}d_<6KB8GjUe)l@_`C82xiZN-3^
zO6)0PP&<UOxa3Vk7%TnTn1DG$BGZ*Q$}pB4;ph!yv;SU*5ZNwC)VJ=Rxq4}!#PH}P
zH^6_R_SL-1Y`wc}LIf{+MAWi87e6W{zq!5woN7TQWwcndBmOwFqKD#c3K@8#_)WX%
z4%EstHgSQ-)cBS=Lu}iep!03=Xx8y|#KKV$u4B|Gh<*I=vgv@F-%aH(mi#ll)PzKr
z=EbkytafZ%9<6^G)&2#ygJfMVNbx`5z|{;in|keBB72`oQ8ILgt#EqL4{;`3e%=`P
zJ1kfFrRJCOXK)R6Fl8*|FQB?{{n-Eaq^*vY69`LxUn~InwMh<36CkA}i~@s+jR^xH
zO}7H&hiHf1rx<ctrJgfSboyXk?bhbIERg9lQ#7qUWbb3T!V-XOnuiwhW}pl7q3#tg
zBcDRwTA=)<8@-%%=LI+he7`w326EQ>()x08`|tUNMlA=VguA(S<OJY5NzB8RdCHiF
z`I7!Y|LAm8Ct_@I03>cYLLp{rHGT0(k>hv@B!Q3hhCAdbUu6IwCZcLn571}DW~EI;
zbg$Tax6=Ov2S-9a9YSuukK($KD&bo#Mr8946-qbtAr2x<CmM`A#X@V3*6G5nx4sRh
z49%?yy-cKTV@b@-f^aC3m8e19f>&82Rv78tbQkeNB@r}<6XNo~OAF}?-0pjl)q~h9
zLN_Cd^=PoY=cC6X#wRQ+yvJm&v|8iWI<Uy9x2CF)U`@Ih#tM%#eATPs^U+a5xdq`5
zeRd+%N6mlad#sO3Y@6aedFlI%ZC2WYY1gqo66M?PtgfcdJ07qV-dGG+Y3=O7>x@P{
zpcqq6f}V|Yf1L7sCd+~%fx_h@4?@?HV=<f`>2DZ9u4_NcNd`<V#wJCbp%ce`*rB*H
z#`x>@ExYLBrof;YICCh7KZo-%ymW-^4R#3smzVf49q<Ci7*X3a5Vp`c`8kYqt~t{Q
z8;`mY11m)7M5_I94a`E9FCT_)5Gp_YOBR7;g&(ss=$~!iyETeh7WjypXN5+CUfD^N
zm2<lv-Icm4o8>4^#68~32M@e$0a$8Mj_18F=~;|7j(sLj$#lk<$enaD7OxrB)<Z`v
z&_f>q?}38Pl+5MnoMp5qWeArI%lU!=>DAStxEajv(m1cr)3;Y!_(D3bBnn=Oxg^36
zEa*Om*;}9n&w^LK+UHJb<eqF6kWp%@vXn{MK$oKdIHh+ZxGssL!-DJ1nZ{ceHFGB}
zkVkZ%lqM+fTCMXm!xa$SE54*>dfxk-g?_^msz-1;v7;>8hcst*_C}5XaZfbdwY>Dl
z8bUg>3$!aFYYK>eiPRFDrky&_Q)Y&yNkNe{%*mJ)`N%c6P~&6|47e<E!TbxJD?O{o
zaDxPX-}v+PZVlA|X5OoM>OmdW%N(kpC1y9w^gz#0{1wc-+P1vWmB66GM+853ungT+
zLd&YL_bQs(`c2v;fW$Miq@%V&OW%e^Vhv(&u%j56oe(};+<f3%&*LG3#eXv$?(7n}
zFK6jFmWY}8b<x*S`&N8J1$+>qZf$B?*QeL5HM}VDZ!<8KsIpv3Okt?Me@|wO$e24!
zi<rGc8!Tib_{@ITz%;K+q~3{he`9-kLB;*H)(dCSrH<Yx4udllZlY`2!&Lf#=$>p4
zOrTg_a9S?$;G#qfU61||*T`tP)DMp;{rbwMI=HpvSJ9&QcmT3Wby0#zS+G)1ZtuSD
zP=6T{7hVp+YjnN{@U2MUlu3l2(bjOr+hcOeJ#Jf>wex<Hgr#ZyJ||(lyL)WIqoC>6
znWYI2zGC8B#|6pI=*%7aVtV9vQNLdV4H31-@U;f{OohYqpo&YVt*h`>!@%HAIU$|X
zy^s5GU06~tc9b9l<S44iOkoWZA2RB_dQ8^qd484acSF&cXZKYhkq2tW9`T6g)qB{U
zwDG(1@2);>2D}h1V8wLbxQHyePrjH+1P`Lj3G3r%YKT&fyG2Qf1d~FNPswGY8|#nM
zNh<*kwvE|UlC&2p8g3+yl_VKSL3W^M=mV+UXODT>mkh1CpxtQmJE^&b6xtNdnilK0
z0;VNLPTHF0`(>aWqdPiu_up4kvTveO1?kYDhhYkf3y#)q9gfvYo%-t@`UV5Ak@l}^
zTt`(s*obOo5=nL9kKhmxe_b%aOtJvQHm%5b(|4iRs^Bn-i~Hs&AsXWg&9^r$I!od*
z_1Gu$FGxB%Na|cd4l`RSAFK<#LT0C@`J(SF!0xnOlYgv*U^>dJTJ%KX`=gUvJK%a<
z;Ud*UCs|LHtycy3QeI`&-6|$mZH~`LEdp`%`gL=U?>$>;h{s=gBuPiZcF&X>^n(6P
zim3fKRBm}c2-AOhR-WG3`|$h`1jih{uQ3=#uE_gQmrn5eyPP+vToIXTfbj*})<Tu&
zCC_)-TEua*=SAnq=Zr;4RQhEhpxg(cjv*Wx3<LvKB$KH_@$Yj)X)9-m@?j{fgjXCP
zkYBHPccJD*18)8Z6Io$F#niSAUQd?4qhMk;n<H`RTFOxg_4);t|B)b`N;v;D*!A1r
zb+b^88>=Qoa!bu4WD5$I2p_+54eaFbT?$pIRYSk)Ga~+uMtbVXA}{cq6o7P+CiSN$
z@ri-lr-bbz#iNB)-W#(d+6{|M8S@y}BMh-`Hg2QKm0w`lT*UFCVWvFcOQ~n@SfNt$
zmZ63&rs2Nw=xQ{ek8WuP_1kW8Q64+o-I_F854N7UwHvT(5kc3+C06~J>78j@{*@kh
z&*z)|;!`+1Q7Do_wwibA-Gya@1ercvcpRr0tz!{jTn{ZT>l*T5Y8$U&CcW937e0E(
z^5Xl5I{1bWPmzv(%^J@hpS}PG!MFqzVdb*Z!|AKHB4wGz!XB-Mf$>@cQW+dcvlsEW
zub4y%C8fSxWWYDjSC!j31j!i6uwAsDaTM10p>3|&r528p+c)<I9X@atcqrW+hd7c(
zjeAkEbB`5)RY2d1LCG}>JelMpk(tvMd7oiSeE7YgYkEfqUEb7jGkEL_xl0Q+VCl@-
zavjQHWLRwHx})GcdCG4VXM^onFfqTn$}vFXek=zWt|-GLmbN_EFDDe@bSA>@cqcD<
zMlH%RTi*7DnNg-nLa;_yt;6C&&w`b~Hu!0v4nCI2=Cdxo4pTb|z`nj>aOf|-Xqvw;
z%COXc8w}I)?d?{xsJCjamTq~%&N&>He;F7@b`K0{fmfvY{jcN;{PaC&FaYz9e4!0n
zt1^dIcLu{>z@_Ki?>DG$4NZ3aI#HK;7VevO&BDc4L-rv&;%wN-oK2k%e|#3X)k9wQ
z7Hes3Ztlg+t$mVEAz;zH7_ei{eX1u}Dz6|4A^BR!;U~uYAiGcV`!g@Vydjta0=T%o
zB+~qF{^lukM>C9YFhF=EG=Iu@=jW5sA($<UJ4KsiEU%;?Xj$#<EX@SjgZfF+4t!_O
z+2?WE8P#%%(AS?VQ92VOeZdtK;swETTO8C|7y!VY+OPWA&5CDTRKf5x1Ibk9m6~-f
zFL>(n5u-s~7q?CL4~-{^=o>q|!{&3-u#I6FSc-^q&LabcrTfN_E68*~rRecofZNS2
ztxP3Th5teG=VbCw%9|R4ot$Gn2MryFKT06>GL)u^@3yPQjiCvYJ`2xvIEW;S`yEU0
z%mGA|Jl_5wi&ZqPOt<Z`6}g-uWb*KAU%VjX#MP$pafySF!sbXEHTeN;@a`LLq&y}(
z(H6pAoSb<6qE80fIT$~ySXWN-{rdQfqpqAsgm`4!7+~wOTo5wQ5ha-j^83szk3E6d
zWU#=JsROyVt`zts&F)~YG58pP?rg9n6$k15vZ!G&#qvGC`<yIbvN!kp6rGaq_!#|1
zp)T)Aw5G!DXNHDBp|M0Tmu9&TZ}f95#!-t_K|Q9k#p5zsOx|5ycQ=>3v!=p|F%zHd
zy3*z&Vulc<9JFCS$B4%RylnIBQ!Tw-u?Nm<JC}lT8w8o({DxOtGe?(!Cb31eu@+f_
zxFW5}%Q2-v)id8m`~fc$7B=QI8CHJp$U=6a&TrVACgjdVYW2Vxp#MoE2mb7d!vT?z
zN`<>+Dj&f%K(D<Iw)Z-KyitXXp&-OC72L}d_*mb+Wl1n$aeB8hxr*dzoVB;CUT_w8
z+k0F0eBl>|SSM)@wm2&EBaC2;St9`C9O+X7<_T*_hL(SMxPA7ACfOKp2sk7aT84>M
zA&yJS)Jje*x^`dbL^NP>8da+x`CK#I-kx|g!m;NS9}z`&#SymUhZZQQHBl?;vRGyx
zw2*urVrxBrs&X=)X{{h(tdo0=u^;4CMLe765KGN3Q0gCtJm^Ryj1poQK<Pno9ra>V
zx4I7A7xn-$6gROAaO}MRl<Xp_Ki)Q5Pe3#3Xr6qVV2&f1on4gwd|WsD*60t`VKMJ$
zGwpPM3ext|Q7s003vm#O-@?sgaYjshxk7b)#}Bb+{gXMNWwL%hHd?O6aLf6|sjDDB
zY}w>^Z8)zEcqLtcZ^HX>^sLy{)}1VN-m@+>AscRSUYVHCEq0VJ&@VYa>=02^{NUB4
z&!IbWcf}m~yxm~TCl>wSXF_o`vQWuQI_;d6W(xi4JToW=qs`5lFLQt2`c6c%^hpcG
zPtDsNyTBzsRPm<q69&W(A}Ggfb4CUM4J|)@@Fo@;%Y9vpr|Dv*zeNDh0JT6Jt-_A?
zrq-FiI#~;v5D;8pp1io~F~7MAdr!1t<x24H%wWhp(+aT|IV3;Mb*(9{iVqkV(OyV0
zwl?=9D}eS8U`7p1Z^_>@tc#wk#fOTy+r)e*tEtE-Q<Sc}$1l9$1ULcxjYw`vi5=!f
zt;KTP$oiQ&T3vLVVe~@_dpqfwA1A4?xp}hOX+FWX&h$Qq$EvJOqX<k~v`Dn1LB+`8
zl<$<lmIMv6qos>1)Ze$J-XZ_C6*(UgH|g^s=%okg<>xj45}KH6z(G*S+}vD@?8mig
zcAI0LNlPEnX+<i~T(5TOdHhwoN%;K!CtN4YxP1R*WwAhqLKe8Uj->e`xahfJ(y&3w
z0Nm4jZkrGfp!VVO=XO)jLZ%5-k>yvM@5roIV?>Wni$|YJS?tbDu?rFu_xcIH#hxnv
z<EMc0oOOscRiCNz=rG)F2`^B)8+`z0yF}C<2~hn~iprNXEn;kP?ChqZ(*X5-YRrcv
zw_<HI4XWEGj!6Qv;Scmbu(NJeEza-W!OqHy>Yc7ZzdZPZi2yH0^ENfJ+dN_!%*^q1
zW8fMh0bJ8qDoaa+_=#c)U*&9FyM=EGly6l4lK?4I0tU!Zd#}Eedg~7F%I`5iJ0-l-
zI}7TXRMlNw)dZOm+kmle{ZPh7uS1r;EN?D3C^+E(y)~Ku+Gv43sK5_Y(TQj6E6@Gq
zSBmE1X=TS%sf+Qbck7It$NdB#o5u2!&I>S9`ukh<YbX84dkFQ<ZdyW&eTY2}44Elm
zC!|#9u6>SHJY0j@RVzR!8%9aBIIa&J-VLd-O0?Z*^QuC+GuU`jn3Z<LKCMhR=fx7r
zjudMUm<P99_zdMw6&)Xh6XL(ysA-hWw=VI>4rq85ct`dGiV70#lC?#1Dgx#o&9L%A
z!d4ZsR;Z$Rbm45G?yVI8z*!EKH*S;w!F_)s3}S|?-KOk^;Yj3X*7&%HrhJ}Q`t28-
zqR444ALm#wCC2CH;cR;q5ZjqzxMe?AqUGtB?x@dl=H_z&VIv`QTb;&%wZHq0&e51h
zi)ccz{X=Od<A2hDTq)erxy8gT`y(Mhq0e@*@#Ui$Emcpz0uBt3b%)$MEfoD)+bYJ&
zZMqNVi+krmOI+f9Lj1!u=W5RKDXVImku8(CaLhO8bu`qWQ}wtaKq`ExX=L(R%e18b
z*86Lm>TE;mQoqiZR!Wm<!@h%Q85L4M%oc9Sf5zg4iGqTHYzZG?kq53P8K(Fj5kSzE
zm7RSdAL-pmBNbjk075_~*T7XaRRmNZ2j63xkG3V)k7JyHYP)O0$Q(9W$A;*4RCcpW
zP-tDSus}k5>W)T|P*%vU!z`?J(SuSZfF44QFn`4`WOBuUEr`=%PHte9XsB$aHntuG
z9nf&W@1+$6HVze)&TjzLLI17;$rme292BAb^i8N7?M}11#LR-VlE(TpJY0}JD#wIU
zr(hO4!eDaBxgLbn)r;W*0~2|FSA~zL$RzQcnY=93;6rRE*X?1#$%@a`BF2^4F|u*y
z8tI!Jtt%6L#>WqU0>Ee?Ny=Kl)dG@R7WHEM#J<16STbaN$_KL~g%0-*t||UrL5!$W
zI#1tjT1(R2I;wyM6&)}7Z<cZIAugnZCWTYSpyFnp1g|5KycJn_odde}_b9swb^U|8
z9NvlyQAxk<4DAO@=udeH#)n+5k^9<<{DCDr^*MvD3t!&C@g`TJ*IK+cI{4eXbcyGJ
zj`x)g$w-YJdP$Jn_mD?Bjk+M3&mwD~m9jdMBf<aD^!x{^Inp1%uiuahUN~Q5uR@d_
za)*{1y?mjAKgEfiVWTi$#$6uOip;e&s2owJJbtge?5XqU=<jK7SgH8Qw9}>0^^>YJ
zrzz<tEYD3j{gtpeFHRr*>4X8i^1&%6^ZOdI4C?l4Ew+!&OhPJQRPFZwCFrvY%{s_Y
zHJN@xyddZa_|G<K#a0s(6_#<n=-$H)D7x$*UW|O)_4yM1$#{=Kt)$g@_DDsnp6$`T
zdq=?C&RM<7G{=qC9Al5_op2s-P5RIKsS*VeI|t|J&+mgm*X*~JHAPIXHeIf(9KeAl
zyx0$GV-JYit=1hjeLX&4aYCh>)V5chAbSkqCHF5TPx;BlbS1;IAt8O+ZOxpgD%jr7
zu3sk`-?+PQ(|v)EuUh0SW-hJelGhWazRloYZ;p8n2ZT}jd)hOwv;qoJj6?}y19Mk)
z(?uHl%Cj1VuSz_uBjreek4cPu9Pn@vs~6;yv4LP7xNbUD*bzp4^I42k3Y@?Gd}vxJ
z=r-Is4sSvD54PMzCWj}1_wr9=?u~&msD=U2?efk@+Q?dv0&<a+gM-6~JVPPKU!MxF
zH<eBx1N?9)#n&r@M|!JP)(EU<ECPKBrFOb}l;q~IbZQ4hbFl8NgHY2@1%g*?{mV2q
zLlFwQwhToj?nGfes&5%r)cg)B+3-RAGz61|3Mz<e!<K9!sEcGt_>xFvya+V5C3t7W
zt>C>0Mc?g_jKSRlEc)nUvIu<P&@AIsXXRic@wy1b-~X223*CF;xoR*seD{sIDWq_#
zOz%}&153l-qr=Y%e{HvHAG2~24#z(Tx#$LY7FE-t)CUx}dcFV=7}FiLYug)-9cJzC
zFi#OY@ke?75|5a=se&p#b9}W|ldsFb;Wk^nVCRagM$}as*AAT2^Ogi!Fd7!}d*tbh
z&k=roY}Pffi0EW!$$Lcg8F?PTT3UYfe$w5zgT}yTJ<~gsyX`kjr=f^Hzn8RcKico{
zHryclY7xle5n=MS16DVNIjz_Z9{G#XK!0j+`t;-Y)VKEdGHt}c;Uvx(H=G233AYZv
z20Sq*a&W3F6sQ&zoBsuWG8qmdkd%Y?O+75>z=oFqg<ihP-&j)bi)j+>j;b-}hoxnO
z{rGnMbF8>@+~|1f&KG?4E8-(dPBjSxH=)Wj?3VA({ZytWV%M6Xhrs0nk=DCD&Az}q
zyosi%Q~XgS;$f<H)VdZEwo^|QqtU%ww#%pBiNoBLdlrPR8r)ZMGb$#$Y{9Uk$&b~_
zRxjV$Gf4<6jY)+FbNLp|AiG|)ow%QS?x9q2tq<kKQ!+wgq8V@th^<lE=X3eFBU!Sz
z8jbP5a`b^}1APE6@ijSV`GA7%c&o7Fo5?*5^Lg-;Hfkm$i{Zw3jaq;nMDCPTv3B_>
zc7R<-Wm6-F6@73jUU@Wu#w5}EZR;LeKGA=YvH^p+XR86Ckuzh@0eXQu_LJV@CB!mk
zUM<^SEhXh5=Wx8I;zp)@U`rD-q1(gf_!Vq*|8Uw?zCS<W!LS|KE%AHrh`4#vTN`>x
z8O-6-#_KPfjr(HaX2Q6axcNwapSB=|_#`b*+pHBw82i(>QyBlyKILW4vU(X)PqOz!
zU7?u!{b3?{6n_p}52)z6^`PI51OjF#snln>Oc<7Rx+60!m$x~s8;^IyQf&jfW(y>g
zPao{ZAD$q}{JB{8bxW5}%@#~VL9in4lRK?Ne$QEq&$8}0M4D1OJ<g6`%^~RucqpFu
zST<D4N`yuJ_%YT7DxD6X%AMd|UG%n=sK#YZ7f=J0thrw+bPPFhsc`U8z72Cq-?YpK
ztJ~Bhu^qY`e;I3wK<iY5z7Ca3T83XRhQaM6+G#pRk#??6$EceKb3zud^P5dCx|=CA
z#Ku#6*9`YCvs?K;|0oj9cG9VI&&HGE7|8Bkc`j(E%rmtOXb!^zI-5T)%eUK~nFUky
zP~G%gRmkF>`)=7XR^d%+#`-d1<Z21n3Udm6bI<BVj!DA&hFkAV<v$oq+S(G-(kI^X
z*BH@#;<eJBHEO9)ot~8QLRHK5rFZfvcS0olsxxk4LQni$X|QF#Tx+%GrQK8vYF_|V
zN8Ju<&#-77_D<Y{q{IvINPia~wST@a);ek8we4fyp`@AN)QOaSlIS3W;B`?1a$|J?
z$d(`YNc8F|`lSGJ!ZxVF#cD6ohyX~oFMWP<8{8EIll_})P+{VO*gk(`g=_IYR>4kv
zcBRSBPY6IB3W6%SRx=EYL~fQm7><lemK?i;Z==bbz@DS;S<zg(R;Eku|5cEyU;X7+
zE@3}?UV3-IQ>1FWhUV@7TG@)hT=uK0Qr_b0kC-Sb#db-i5<o!XSe5gU<}mJKJPj|U
z&FiPh+K>Ipgusg>XAbb|!yE2zvLk|)vzmIkNpy`Fmj1w?#(3=()c2N~y2k4sUQE#&
zJt6_6mg|m8KNYKDhC*#Kzs-QV)3@#%cB9wP)N3~l^<M<}4;4OZ-a<=`iGM_>P6f#A
zeMd%An66_3<;<s(r+q{*X(90qL};dnfc`E8m;qCQ&^Mzvu(=#bqdFPDJt4<`f|r$k
z2w6sjvaAFdj{mCyi`f&m&$ylBQvljuph@Io=gCHiL&XLy_VBi^VX=1-BRMt~4c_5Z
z(aPtovO}vdMPvYOQ?UA)or78ND#wUn{|6o}D6YB78#m{<Em?)E=@MDK@rcJaKXS7B
z-ihi)f^%&^4e*O9r+e4V^w+j1KfBa^yQP&IJoj)%%O--}gL^A&+0Z}?-qNH*9M}gK
z#T0N+>Yq1#`wf$B9taM=>8VQkHYO3Y<~;YYx5hX9j>ekhmvWYzkuc91A(M<};w_b^
zpG=7fkdrMb--zmCpIPnJp5~>sl>?Wu+OH`6smNzBvwt=Ux@Y7@rp$JIdi8D;wYKoV
z%JyJGn1BbdladitM;@+Cy&KN*BZ%2KPY4U@_QTMsrQr}n&$~fgqo1d^^nUHXC{r6l
zV|snlq9@d-eS_B}%AAOElcohm-MIG@_yT5~&6^yHMxH-**UXlwyTifnjq*$ZNEenZ
z`1x$JihNbMKwtwsb$`JNUjsvmekaF$+u?*Xp2-`z-|Lodx-zy6EEo7`ZxQ<(|8Cjb
z+Zd`+P<UshtAh|TU}GJOu#vTrjCf>x@5oR0GUbvM9pxC?Z{aFM#P9O2_iY3LO-saT
zFed9{5tb78bhhfLa8g_x|0(>i_LILg9J0D4^(Oc#b>kzuve{{CdC$c8r2v}nws)TC
zfNhWPldU^jQd_czJ5>%{U_@Re^l_qr0B~#h@RG-9X_E?*5>1xllU|P;KwBT*<s(iH
zuVT9G(=7lm>rpV}w)DS*5)`ahJn&@l_<)X}cX1IjG3vUwv@UK)d7AHGaB8s)To#GK
zs|o?553U&qh*9_-TbL<JyVM{_Ok>teGEA8EtCCIh-q?WEySBYkKUyy9eXh@U0xN!t
z$!P5#n3Df?|77%#+jmug>{%{<LaBh}AX_Z)u0_gLNr$yMM9K2SLA^ay63C`%00DK`
zj^Ao0G2gn5;by%)9vAH)Pwq$s<q(|ld`o3)%h0F3XnTzcOXHyYiVB#z1lRr!V<s*G
z6%p%$TNe64oE$`{O6m;nncoxQX7Y8_4=js8mDD|6+8o^}WB0^if4;3!asZyte_<(<
zs6Q5eHm_5*WP+haZxTKkhQ6iCP3qHB#s|0zmI?(!@(vFbxATWhfIk###g#I&!k_#e
z4{o<lJP(d43X~XL+>!YJYWdC2{pZhAk?{P(M39MzZBH#W4rBcoM}l3?b7|+Np={7+
zMG<wK&zSj$R9}v^kPGsfeB@Ay@Sw8VaZ#4SPM^uRb~X9oN5wLUI)3r5c3D_yg~t*M
zgUfj~p%c-EUJE;2tY;gZl7&@(#`cBmPma|tGox-8by(6<T8p<XuXQ%ea&o$zpMT<7
z!Mq$(wxZ(JmomvT`6#I4q6gAx;2LTD0|Bj0l_UZb8O5Ok_|(aZ-BYubZZ{zF{h^Ez
zUe@(Is)n@V#&bfx+H2Rc@vBL=^Ut2f{|*KzSDLLqJNSyq?Qy%G%A+j(gmQTKZMu)>
zlUMMjGhJENnVhfT(<PtKipN5VOJjWa)k}+%l(9b*!rWG>#$T5MK2a1$jif-Y9rfXC
zn`5R^+!TeEW*ptj2*(RY3hzA=aka>2NZC&)2I_^Joba~<>-C^M_`<oFKJ~cyJ8v50
z;c39$=V5_X@n4PdWMb;(s|~CR>2->(@z)o`Cs5|6eLDDb-<crAV86jPPk(aS3fb=q
z0CWn9n}9$P@IG~=>}-#4*bDU^TLlz?b(*>hU5p%sln%xJ?7K%=?~6mN7ru6Ay?@F{
zYg+LLz}QiQ*%$z1S1yW<6f$FXb*OmbX1a;up9(D1L8B+raljh;K&CAIk`TfL{<;&9
zaagMT_(HV)f*=d*tItK0)l2|5LqtvJ252=PGcQ;~{AKBr^-vV-m(IyaPa2?Yjv`FM
zq`wV3pO7rs5WU!~;OalK3g?gBpsLmpZfpxp{L~mA*}`dkbh*jxtl5S|GwcJJD#IeS
zO`Lo`=iMNO^`zHuu<dsI_8zaS<>0VB3#PaP0d(M>swnV&{gTmncFT>WJk?oMNt`Tp
zA`;S6TkzcZxipk~yzK=GQJVIq&Z~y}R_lkMVN0lG)EQR3bW2qTVXr6iwEfDP6CdEL
zz{>oJKX3W&-ud|F!T}=hhx(s)S4*}Mh=9j<6@b|Nmi=4!E8DsM{9M=FvSTLt#TVRK
zoQh$hH&=QN7~=@eJc4=U1;6=Qlrg=zWxHR4k##jwcQlTtx?!r`c11`sKmi<6G@hY&
zaU?N7e(Ue>7yd4=W=z^xNaWu1HpNYZ?r?oA>>tu>xm1AsnvD{!nv`XAJ81swZL$i$
z{@EPg7tq9fMIumD3jMAC6<dC+wZ|LdWrpCxsuG|9K2@V)bfh+ix>=(t^zHBJcd&?r
zOU&g=;f}(%*gIq3&oNwA)a1##Zb^{(V6rnm&5k=<aX!X2GUh&fSC>wrG(`jg<v^&e
zo1AX?w2Gj%21G)%7P$3?=F=l7(wI&>g6?H@1$xsToS%o3u0PJH8#@okZ~UmJn=e!3
zdO(Mm`f`wfh|2z@YN5E$<HE5E-xnmfRWJF~y)))GG{qr3ek0cF2vQnNB%qK=5)deg
z@%klAuMpR<%5B9BeJgK@xvtx;ztCrvB#}&~%Q|>z>&eg->do~dn>M;DwSGm`=z$n~
zxq@@$_lQnYQ?z6W3DgjMuXdCcL<|DaEuPhe{6zRyJmi$G7v0deCRUzok)K^`iKY65
zjAX(wP>=G#U+$8FfO9-q=5jlyYi~5qwLT*B%Tu2MyipKn54#Ei?LsB-q`-%ZFF+UV
z{vWJTTx=C`u>j0ufLF-H;(xOGf7T+`MFSxhi~mV8|KsEQUuvA|Z%6(=X#9VC*8hud
z|0_NJUo;c)cL4s=2LFYz_;(Hcr+xpotN(?@|GSR;ziRxyGSGjw{nztzdeKA-{e%Pb
RT{H-IX{qVnFITY%`)_e+9*_V4

literal 0
HcmV?d00001

diff --git a/studio/src-tauri/icons/icon.ico b/studio/src-tauri/icons/icon.ico
new file mode 100644
index 0000000000000000000000000000000000000000..acc6eb82b9b52829178ffd2943fcfcceec6ec0ac
GIT binary patch
literal 34589
zcmcfpc{J32`2UYTEukb^MTt>F>Qz(OGmJ{nM!PmMm0n56mds?#h$LC2Qpq}%P>D$r
zG7Q-gQpCtM_Q^V%F*D5aeT>)peSZJ^&iC`j=bW#jbIkT!9?Nyz@3-rkAV>~Uf(#8I
z@M8#lR)L_a;4cl0{~jMvfS`3tAxKyEzsJo>Ajo?;1a05`-{bZb5M-$ZL5`09J<eVQ
zK`QHIf6IOhp@5?hly3z=u}58<)t0SY2A)(q;No<A@kQ_k6(#Uj@a@~P;A7zjoOYkY
z$W2!%yf8ldte!0mok{rCoEvn=v%s(C^Z)#M$7en}NY318zF|gqtK|CaoESTw4sxa>
z8@@vt{zhcCRPJIkRj!(P>{LX{;D^h7`}YjdBL>jxjo-=p&e(Sw`LG_S<BioFuxfdM
zujBT(h_)QXYTi}6SHXLH$~C>WcQ6SxlYr^F72iXU!ZEBGSc^{Yhem&gY5%Uj)_*1#
zhAwFYxu=BPrvkhyjQ3I`d(r9EHarbf&>2=PU6@5-TH_Q#)60pi3$v{wK1g<ru+Ch4
zIFvs#K=Th@cYUnOw|ZnG+A)_feyCaX=*o1D8SXjt+ZM3Unx?>UnlzpyeFS5KO8=^t
zaeKp|n;y>AX@S>$(rB1y_pz4#@o5{O61g_HPv|A;%qKFs6ZB0?;m|JeA@t=;;;VuF
zEPDs`X#YcN!aZTFy3|D?@Rz18be!|fzm&$9+YA@(#Ib1vHDstNt`05UMf>Now`-<)
zBqM8cDf1e!H7whjF-n*`jpACOlaP`gYEI|D{HMp4wOd@$i#Wcl`1JMs-{*u`RPh$n
zBzT9rpe2`UOJ-f|E$_|RjB!xOJaj7H7PcykD#&BE5irV3K|fQ@kro0SZg=ugJK&`F
z_*(vsBiboTWzSLC7b<n&?1k0=F<NA4Q{J1q8M94M-LN)jEv9SS2P<Xo>L2%@uZpN-
zu4#DWoeF^=z%21iO6eM_ahR&Xl!S&^?@aPe)oplGISFBs!NdNrkb5zt%FIy<!&(Im
zgyW-9ivnS(bEOh}ehvEDuhOdyB=hA8Jv}{?$@0oAM1p8o_}fi<azmtbSdq2DKon9C
zC6q9wc>F>MA&rk~<zFd0vpwF3D@;G7Cw$f_kaw7SM6T~I&2d5eH|CJs-n7!eTJOr^
z#K9vdp5<GQf7hC#7$K+_w5Se!#et;2`jl&*P*bxvF)<M{>W=?6lk+5(^&!SSenDSf
zpMBb&cUm-0mhLK^d|v+uJujvY4iCrn;s^gJI=&-W6LDfw7mP-!RwItM;pdX&p4Z1F
zV^fj`tqEM~GJb!Ua2P4^gbe0?#&ERYY%NM}Q^h(Y#|Zf~gyzv8mI#Q_=UhfB*M~ai
zLz!x-{HNVUp5=_>9-_?fhyxcdXfMl*9Y|5s7}{8_EvjaU4ER00y(1l{=?)Al3fAmX
zIp)Jz?ay8vKkVX@aBd4zkjVTtHfGXp!3LK+PwCF9SX<H?q)L`dkbiV`uI4X9(ZwF9
zFLrk}V4z5^>#kN_!HN2onW9g*@<$I2@(EJDum&mG>o8>G5P3;g0ESzka@eji>iDwc
zZ5~5y*@qxpPf565)9B;QwPe8q@;8{M8pbN8E|fnx4q-U$ncrix>c@OtuW}A6wmz}q
z7MSvhT@{~H9eolCCZTVK`}yJY<TxF8&~!0Pp7k&WaW^^qPx6wBo}N+Xwvc8IkZMVi
zaZ(%KK6~XkALv4FU$)1tD_NvRu9>ml#miq_T<e_E2|cW}NrZ4UU>ieBsf5|uLuu)O
zSRdgFAASA&<X9*s4<+jK7q3RoXQL0AMPJf|9^Un<_r83F9ILYYe$yL7lH&O5IeU-%
z?FEG`N38p5iQ+C|;KXmUM18w@OWK7RcLL9y%YjQdxPC^sL0VdxXb3LZ!0!!Kr;GdO
zEy?)t<d5a$P0^}4n3zzE@=WRQ3TVDEm(<!=;e=3q>h-ubNcU%FWPZNiwr%QW*#Qmu
zp)5fIGLw%b@dxo_?k;joE@3{msH7yu&z0oRl8kFih5}kFX|jJTF<mO|4Q$YVa{O9;
zVR2dPXPtE!=gm)C+?K^Tys!G#?Z!a_hL;?U#q-G0CtQBNa@9g5b0KJ^*VbWB+d(cO
z`S#o7Aus%x7ua^qV0*9}(c%ttO9QUC0s47vcqFBr!il5)GrYfb@Wa(#@M~63wfV-Y
zsLgaj51oZY3Z`+a1p)+l&@?r1L#Fr*WcFHeb*s3ywXmeb*=c)P%b9;xL+jS9W2sVE
zs+1p-loc@U4%l9&pCS5SKKhw8$r1sdqhBK4ShVT6*Y3qQY93vi&3l|&Dtw8Xuk*o7
z(!z|4FQuK^(rzL5v-K+F>lTVU4Hc{gd9`Max=E*6&yY_;zt5%QER7CGB}ty5M-GvD
zMq4u_#;(oQE@@t`IFNVn<<HghUbY42U-Z4?+qGH3G*jfAhRUuub}2!i{H4Ii55*fr
z-JwgwbjBUj!YOnDQc#3ka&eur!NO&D{jg*ZhMC%H;2PH4A`;R?wz*?zhg}UK2Fi55
zJrhj?z+o&o7*xd+clvV+VeDh^!{5Rv+;Y|l0xJyn6$}f1p^GTg#gD1uNqjt!502|F
ztAM<Z6jjg_1{C_4V!`ySMDNt?tx+G@GybeUh=?6}#lz$u^y&r5Od98J%^oaUncHQJ
zQ8W0syniCeA)Y2mgN+Q+C4FUFn&f^=-%WES-w$2yBl;D}IqN@HfWW14_mk!adjrQ2
z>K>t#5od}WiCas*YP2gJyY25RJUkY=?eUqC*mmv_hnYHb8VtuF1@{CI)x+7efrd;@
zvZzI9hb=r|&Duz#?%Y~y+Rs)-2)|Y!#%hZxlSb$}{$V0mkY7n-xjko8_?;@%|FMmq
zLQDMgzejg4-cej%$6Mc1wB+f)(EP&;eD!`>u+0jP1iL?ZUoM)oUF$q|&nU6g%G-Eh
za$|#6+HCeit5)U3?1zHn_U*pxdt#-9p68wo=5m_r+xFkz@XOa&#VPnQ%5XFM?w!U)
zv56`Fke)Hrrd0L3QK{-)tw>cx5~eBQ55eSiZa;&!9sD=}^ayldpQ}^Bo{N$He^iT1
z7ilc278qB034#<>{$Hw=)_snK_q?yv`u;Xhs7L5TGPJ60uPf{7TxS>Z*clb9plthL
zt+G7(0thTn?jQGIO&a~%p&~V^h>_k!!tg=~9(YfzwHuNE*ITjj&JO6QS5WFp_5Q%g
zm#hkE9a{92%;md<?S$a$6Uc@w6!s)1h&jJYG+UiNyLXQYlmV1epzaDIY;t}K=QU9f
zgy)kb@eGC~YJL<uLJLPAae|vx<}SMxhc2$gjQSWfm%$g-FTN=`T$s!-0@3_=%T)?n
zkGg{{YN0o08E~)RnPh2tFnVq;g=48cc5$tWQ`dpM3#>O@>p1O5ZaYP6S7C(Y>_yhn
z@Y6I-13}b)`xb*!fz9rQ)wYU$yK!Ba^V>wtt&-(L3NPYIS(WQV*V<fbw-3_uMCAdp
zbb>6n>n2b{37?`<+~!}q4P}F9ECfmEHeSLYWSZIw8=-ZgI&B;7YT}US*V*gc&98z2
z=H#Ig2l468R#w<)8s4vxYe!AGKhT_Fju<s_8yt5FTMH@kGdvB58~ye~EGeo~9?<V+
z>kHzj0@eOu`~E19N`oaLZ2y2-e?%_$3}qpVq6`&PHpoYO6wmkz{wxfqTI2u51=E>Q
zOR6}G+Vp!ZT=*E?xTSJv3uh;syK~*vWSlUWV2hT9pc{<*@!QGlU(6=|;^5zk8NyLG
zFQ{VZmwLDjQ8-6uEL4u^$(0IICv8Idigkn8{YAaEn0*jsA%?KsO*B9oV!MU!jQg{0
zaRQ;H&HW=6TJz{zGVwJ1e^9(SVH!+g>&tCp&OcixmeAQh({7BZ|9Fn3T)ns>Rq^J0
za%HW2&f!~QdcBmu<p)`_SstjMTUrs+vn0kPJ<~v+-oTr&hw_ZBD=IC%7m`m+3bESW
z{OW^=+;h$M!Sb=F=(Tss;=h}(*|f~+kX+UCBg>$A-}>c{XLAvY$Y8@~t*U=|{U0`y
z%v9(uZl-}ydjSM3N&3IsOwwg{w%1c!R8UU)HkjM$o_?KRj8asDJGrc^wK$S~UVkUO
zdt?<%<no|4=*$U?_3S+w6lmYj#fr<;%C02a8f9u%9`}GUjr#>DRIC&|BDWUP&<!`v
z2Y)~vTS<XVW)h(vWFqG2d4tBZ<jKF1xzm#Qyvih^@DXgo9U>w$Q9XP#`DMjy9$R%9
zB^g-bHgbh(h&A{xct$|CaxG!(=kAPp^n6$)FOC{%+ACZRX*@aZ-L6=b<&V>V%{Ias
zru{`jII*@g!ZW)8p&Y5aNlT}G6-m%Vdlk+mazM6tgc$tcb#x)|{tKs<N3@~7_l%{X
z4pBtmed3k*A=sF<fq3my;+ADYaw?f`T`ru_XRjdQY-jVS&F+K+>vo<!)V{qfL_gnh
zzLKFy#3&BmhF+xcaMPi<Ex0E5;+wa<CvIBQhV^d)TN+PLI?eZ2vJO?ML<H($vfXcQ
z#LsyVp3$%wv_Y_YE(Et&il*t3tw34OTk|O2^R^c#LzRD<D(})T^Hs&BZaZ9!TAiSk
zu@&BRdL9cZl7Sw=bER~fz(}O12022M3W(Ua5{zo4FpZi*5~h#_9}#iJKy-(Om4ptf
zOIhlRve7jYaX_BH(nmJS4l<Wxxli5@@_kdZhKSIhO6{nsRWhl@(jl;Q7NZeW+(T`N
zGMF_)jh;e9lSGXq#Q}Hxs5>Uf8mlb<GC#8fn^qz>z$Y+5u>ymeMQf~}lM&Anw9lWf
z)n}AVSC$_$!!bhfra+?M#596E3?%g*{*p2OprP8OGBsDh?{A{vT4*epNX88);|-zB
z)qY;gm*re}Q`Z|s-AKa?DiNQ~-n4JxmRlI^ftne?_*rMdlVHY=K$#d@l+hjMD^}4^
zVOzuLT$ussg|9RYz|6_(dAMDbKPvZO_rPJO+lKu!#^H%<#N`Hp9T>r-!-;l+0d>hm
zFo!?vm?KFJ<4J-aOpz;!vl0CmEg40Rjk^(cK`F7ukm5<`!1DH5U#~dj`8j{-tr@;8
ziEG=7Wn^@X9dNs?MdFx|zRIQzV%>K}L@sPi^ef6tR9UL9>c#%Sp<i-w=M8??RSc%F
zv#7!*#I(2DuxxsevmksoRrEHo&ysj&54S5i=C6vNf*BmlpbImw-%n&lULW*Bad}<1
zpS%mbQoXTb+h4R(XE(0Ce9cJRK2VJmJT{7zy5l@%dMZ>LT_}IfQ*6mnCI6~JKxIYM
z(#2vrcakcZ6n$-#BJwRES3lTrxpc}}*vuu=oq+c@y}zk@LN5O7?@a0!#%VlfB!N<&
zM_G$xA40au?2kRaw+1fhg|ot_lC&uUefa?2!s@}>N1E!3X&y}#{NAQKDkqQCUP+{W
z;g4S^2~2NX>c~2rl%!pBL?UR*ptRXlY$VTV!E2bJG5;M3m4DMJpTWYvl8X9T|7^BL
zN(4wo9F?c{_YQrYM;}<sm&KvCw@9urU0r8^+RI!j3(Y%PPa1Da@aZ4=;><p++#?z<
zu0F(6qOQ>q^xSBSaaf_kK0>_btZv@_J6D*e-PqLEE{L91z|Uuz2E6}q&8qdTtF$fo
zxw6LPr$%5MUg+)d1a8JE{*(7mA+6u32>UfW^r(;XwA~6~qD^5@P4(dVxHr!kj-zF|
zMfVfg0gY_DtDea_d7ilefu5n#HNohruc>9@e_%?l812&H0$BlzJ*L>Av{d2q+1~~c
zpX3~0{n->*i-4V!tcn-917Y+iZuQUjf|9ELLoAm8q3w&Y++AK613@cN|8HYC5N6E%
zEUI2LVdwfidqla00yo1Kn>8+X?4N6}OiFqvP~Xpjxi}vC={w^w6TJOBrfbd&HQKFQ
z)Yo$M^CR%G(<uFb<0eSADDaZa$dwGP8Jssy;qOHyw#OvYsFn+M^iafTi7tUkfHi!v
z^-gkarm9v`_nvxKCFHH=>rc0D!7yW}@@|-Rw}FaKA~yskD5Qwus4K1ML8_q=35k0E
z&c6qyV(OOzgBD|)vsn=Og5!VGxGjJ_5#bq9zhKXKifvG)u7sN-m+&u)B=s+dg`b(i
zZ8M=_HS6tut=Rm)m>&92l3;<Yg%r(FhfFi)XuZ}30YRU4yS#3pYAtG8$*OmehlZE6
z;d&2ItYEx{?xTNW-53q{5obXPTCz_v0+&W!s@Sh3v$Z!pV*EZ^m0%im!glm?68i6!
zgf)a-WE@i;_VE`RRf?E+)tIgVa}!hPvj5VOX2`M9(6UyaO#F5_?>ci2R^*5Knj5|h
z$vJ_nR_FjWDVES^SJ~3TUWFFyWBpa;&QbaHqCe4grMC*}o;SwbAJ}(tRd4vsSU&Ds
z6X!g6u3z+*fio*X%smo)a}z5<#itLNxm->)$xJggK6dbB=dvoI<hNS@*VCX;J$xpa
z)j(!TC=Gj2&%zSt-fB|rzoDqSiww?FaG3Z-u69Rq)Ih|(ZG;!P6ILB~a{PHttamEr
zlCJ`AoA~0>-M<C3t@F$ATt0z_p07g>)%sS`5E6dbEzdNwr2j#6w?eAykbkoU6V6mQ
zGG{aSD@@9Uwe#(VZFU~fRySVSgD5$^>~rJc8S5k3kN=iuq9w)X(NVN;m5_@R2LB)@
z3aZabADDBu0@$Wlt*WV|++G#OywJp_&i1gZ{Dr4QY?yRMI32*H&t+&Q4b85xvIDP*
z_g@8hR<l~{3!VqjeK9lrNuw^MFC6!;Yg$LWBlbSxLeWqJ-o!gq@Qa3pS4s|1BT14n
zQZ_T}-X+KpOzHLfD<&z8Z%^bx4;3SfeYuZd!bga?LgK&coOdRj+mgL<#(MGSaa2(g
zu0sd611UJiboEi=?s+K-w`0OL%Vy5&vkKvJ=Wy%P$L5_8E1?5jIZyWZBrIgu2HyiD
zhs^lgb7NQ=7<}mOLgppiQpx;e*<?Za88QvW)*<Ym&-u~oGeY|t#RttkbpqT0&0vO3
z2vSQE$buO9V5)F+QW7I-K&Y|(OW?=zEHib&R)?`92hE6cTRxj*|2(HK;MYIqC%6L>
z$SvZpL_m__hgGV)2y<TCXa2%x5MpcxQSPEu+F~+62piqToWbhG3|r$pp>sY@j(<1;
zvI|l8+HD}oK$4n1wL1RBeMa1XP-Te@?NC2Y6W*Flz<?c;giJKf@qX2L&?mv!f%8bz
zfflbq3X_nH<&~4=AB&0<s?sQpX{<(|25_BT;j784Wn`!-#oP;8TvmWRPymoJqRt+o
z*2=gB`!am)@~6%>#bBL078H^sX1Ansu@qVs>QUU}#+L&^Lrmpi-G1&%cvEONEmZF4
z!Iw|IS{y+Ls@2Efk|8)|JQ?pj<CWIA7A|gsx9Eib(vf2Uo>J+gNb?BQ=f|n|22*9k
zXUt7H>xTWi7xvxa$dlunvLTi}ou^M}ibAJukH9*-LFlYM?KgC(!Ik^LyPhbSa+A9V
z4ysx04+uhGaWR0o9q7Qhsk1WR(^NUtbo}6@R1@R%vF*8T+~2tlPp7B-AY)BUNc4v!
zUZVp?jeMCb7~Y8HY!v3x=eNoTV%xc_C-PN9@l-2|jTYD<rn<am&d#(u*F9G@#B5t{
zldgo~eTfN@n6(bXy5V<1pXD%R$zis5Q3D~}*?!{KB4G5N-S7XODv$xL-r!dI-{&4V
zIf1QOt&ZQG>zV;Jg;}&6lajQ^F;}74tFT#MbEC%oMcva0E7Tb_C2&l!GSE9?8k>ul
zew!ZNmQR`IsSgLNQ;B8lVO_ZU{M_)ru6?z`k~t>yrucEnfA>UUEhSYYWzHHyi(U;b
z`U)ReoB<g_gN%R4?hbIm`_O;~o#!D0%EHG%%0-ewvB{z9&9S}11`RTX#gds(=u@X`
z21=>o^C)3+)bpS{2=&eIs+!NPoKjPVScaEXS|N&zt)PY3SMtNCIuQgx3<1c$0p(I&
z=Xoq(r-#j=!)D523V?Z=SP9A49)OtCCTsk#HD^7Hvwl%VF$faIA38mEa#wxeNSVE4
zsoTIR`e2C0R|FvPmk%Ecct6{VXXqo4GLFT!ui4Ie-hj9V1N-O&Gi-)mgj^_ETxiFL
zw^st-M_A_Q1OZG&09v)B#$ayyiWMtlFcX0+=?k<o{T{z|iD#zjQ^sETHQ*J@?t^9`
zFiAVrvETNVg3S(L0DC=eAR16sV0j|vJm1rNIwviWTytcN%)bW7^y&@f+zninG7f4W
z!jv7g9z)mi1Cc$@g)cXkw%x7e?jDW)&=Q>F1R_fxkYF-}UlQI^GOAZOs8>ptjqLxf
z#eQHFNDJXIbq#dBOjzeEjzY9|{(1mwT4Ur?uNJM5MP@+5aPnl{IdUD?r8ZcOF<3-H
zwXxa|4hU%r81ERY?oh?=Lu`9G*B$`cKrxTTmBke>{~~#k_z$V2gs@P;9R-@WH<TVe
zH9Z~3tl)EcfIC>U;*USHh{w1~T%I%HzH>Fa%^#-Q1}i}aAScCq%hhp9{O7Lv*DI*i
zIMVrxlpqAn4-uA;7Hm=c5OgvawnVU*nZqG~`D0r7I5!B485kUffIZv`Ajmvf?!L-7
zAKAffRF3a|W^vA{mJ7C%{M5*gQ3v&KR+QZHoTU++Yaa`GTg4Y-kwHCzcl*d9&=v?u
zQY>nr7@bB7@1u=AMGK$W29GWGX91BM1AqxDfi6gZny;Fl^Lc-I3$*AgWMM@4MX^$q
z#Sy6Y=J1i;3WpcAnnPtrv3TRmTLGCg+~2(rrq;-yTDDh=1F5nqN39i8n8{pPWs~Ug
zs`I-nkg+2_b$&bir6B9cSEA%Ek>kpob%k8ytE|)sf*LF*iPYR;O>Z_otsJpf^|`#d
zm_eMu?t(Augf}fzHiU5sz)AsQ_&HVu>YuFWkCVZFMsHa#V0j>tU(yBe8AV=m?@s61
zn(YvXnv=V>d3>pj0UHUhA8@Y!`RL3vJ(GXCB$QDMN>ajUJ5ufYTe{o}KgZGYK)c>O
z!>IK_aEb+`G1ScjUNWIkW=PgA?SWkOXr(m%__1P<;FnXH%2~(ZyyG`L8o1_1w845H
zG!aGLh%De*r!SNS+__(6^%+dm&LrT!kOQaCyi>xjrcwj+>}$Gb)~(kq!T)h)?)m{y
zKyB7%S8f=27DX8XO3<yGH|`>DxZzs<F`s`^O=kG4802C&M2xIg9|Wranwpv6#Csv4
zOdZ&<i17n%LqN2E6!m1!Rr!l)dSvPyYS_>)Hi|lbNx~V*NYC^KmAa_!suu^k|EzZV
zgKV|e1NL&Z9@kQW(=I+;ANacX9!PO}v^;1#mMgdcqlE!p>#OSjsu|7QK57vKlYq`!
zheuhD^xhhT-C_=Q3$2opbRq_#w{N!<j@!W+Krkwo-=1hhyAZ7AFRJ&C_XF8sc}>+K
zs=5T@yp)w5X(7MQjel7Yp*|m>zE~F0y?5unBN$oPh7JNd%{<WvV)6=@Kiv9wnA5gw
zxFKeJYRfGoV}Ug5MXB|v*bL`?g#V8*KLoNWg%(z!12h4|vZpUp;KbH=1rq-csm3O;
zYrod84d*Rswh_lI!K>2wV!E4IG?*AB3sjxRqVHrD@V7r8iXg5+7tWn0H}W0WZhyvg
zEYfr$!F^rLx{0{Nbij{@k_7W`Nf0ENZIv#M@SnfuKae_E@UN~?r6`yhG51XmGXf{9
zB1z7Y8YihUlOVB!^S%SB5Lc~3(1eR8(XD^<LIq#olARzd3u3i#T<dUch%oO(P^zMZ
z|4QcQIft^<%lRLU0WiM)zoocGWryvb``_smznr<jXsMs?E%RIb>=5h(XM=x5d83=;
zYb(1N#yv(6q9rQS*?8*U3G`fx;E6D+t8(<Vm6dfFbE(_tmQJ^^BFPW~!<&TF`BOPz
zoN2U}M-l2#)>CmUwU&KK>|nyz44ogG0@%VM$}Bm2D<gMf#<F^f#-f%N>a2{7@*X&r
zmMiFX$Dcp#H)13i$;HeM?sJ$cBsWg>&u*%m@W&_7OPSv3Q#u#Uz}TK-_6$W7o<l#w
z@`(%Y=c?nM1HON&VHG)ubJanTjV4m<LQP3rYifOzRU+4tJlpxt^#*4t+l^fs73@pP
zyvl|DSImwxG=Anc6mVvJY(D<SKg$^IF`qiS{xW<+>_2<!mW*tLiL}-oQJ`XahM$Mp
z%kKd++2$Pi{&5@2o+|;v7tdTuYr)sQy7{^#ZV^fEfWLaEA#dmxZ?;U)*u)P~)a}xL
zcDJ?=FK0FO;Gv3Q)riymhtPHtKa>+Ns%=Wm3YMm=Ef)f>f0lDB!oK))pjjO6Z6#GY
zDbcn}lU`AjmjADrl->{-Es(y&``%2{|36q-NASAiqNRQ39oP>+t9<`&EiFM)B;C)F
z>Qzhq6{XZ;LVuIHwf}zJA<cdxver1d()enJiHWKFSaudLe+13V(z<t%Ay(VsV`dfA
zQR{|hedESP8l82{e2&GQj2FTkN2!i3ikE|eZrKcVe;Kfci5^hsS?K8Gzlp)r-1l&H
zCxIpIHD7mpm4&Ia@r}q-^OcK5+P`|X>H9yw;sBW*;=tWOKBO#aA&J(IBt^&;nCKVm
zS1U-Ke31re_AhP3UdmFosl=wcMuS>PywJ^Y6nVx)ap+CLI#&hrCB(j6HF=qM4VdV!
z`{<6o9e7a;;cn|f8+^<kFna8Z;Jm%4g)wKAfjV}HK}qPd23Y#P(^rb-{y7-4!<!4c
z|DB*Ax3Bt=o=xh0gG!qp3y29zY)nkWa`k@|B65*~2Pktf6>3y&8g+@I(@`I=haCoo
zk(pWE7bj;tD*2=RD<X1>Mq)BKIyic}^l|RC(3Gd&PfvThHWO49skO*%RcvEp-KDxi
z{vI{C6u~5Gj)qvPg`afCN5dp<Nh5p-OC2LZiWGoJ^=&_=;HdRkuB*F$Y_~dgJ6ZUS
zTyIC|=wa>h7u?fe1v4_gGQ@pkuFe8K=oR|t#?%(7_J96d|6OF1)%ILgAQ5BfAk-wg
z&giqFf%U+(mk_pu_TiP)Pab?(Tpu@aFU@}_iY`?Ud?HFPm%XCMVcjH_t(#!$Y2%)3
zuOJRHViETcvRteG4Bk(sS0<7O)BM4e`e^Yfbc+}6hgV>Lw>oIWc>L|KuV<Rheyzv%
zRb&@_ak6X0VkC7Y%sO-9#~+r=XE4XuO{%~v9d?x@{1>MDkFHABPs9|ps}FmjM*B(`
zYf88|Ox9;n2T5!TYO=r18JP?D_f-y>$?REEMcWq6A>>?AdYN#4KS!Mqf!-xO>B`pw
z2UJs$sJ)9+9e!Lw9Xlp!&ZFp2xq8&b7UFFi0?&)!N(*thEHkJCa`TmS&c?rhEM=P#
zz9$fIx>VjI^+#*vL^=Bh&Hh<YjpnIpS;Nlp+<~cpT07K&E?)4oKb9m09Jm-Q<vs6I
zkfps^E9oy^kZ-`>9jP>s+6Eec;1O23W5W99Qk{eF7mlwIZP65MkzwY6jn;15a58I7
zH0Lie@#l5=t2vrQ1CPyrQ@kNo=W43SK{E(Ay$WrrNR|4HSlMRBZA+tlu+BGICRz30
zSZ&bwx&6@?nLd*Bvz<Bm%o+LEDp6K6XGgkh0O_9nYS9$Jfzt0>>w@6H3|b2t5b?=^
zZ1UxO$5-icHyC0xS1n1t2RzPW|E~$>SYP~wS3d>5mU&Z#v)1Bz8@{ta{y;Bza@V3~
z5#oMW9C1=CK5fhv$vkIdiyoyTE%?FrWXgzMp>@F~E%=5=o!y|RB_A`Lq&{Ttf7=Xj
zUu3MQq-OZ<7f=qy+>r5tE^vR4!(@>|oJ`y)zL?y#Q@iK)IXTQ?)e9JwWSJ!%R0D;q
zZ-;ZdzgiWa-Vn*KI->1CV}CM5MiYfy#8LYjxC1EuNK~0cJ~*#@q|P0=Hr$^ks%;5!
z{4H)g8D|E8J`hLE-N4*=rkVh?&Sb4%R>>MHIKX56x#^({^dy4wh<L@zE2eWTBx_7i
z0C0EY5GRj1pC_7OO16Q7&TOu)S8BI-l=9^GeHDXi;L0mON#((LnAEs(WR$)xcL!W6
zlCb+&fcFLVRT?5PIfXQq<-kh<*6(uK!qem54)08&-Oi^C$utRONSpL+QLrup&K_D`
z169?)BmsDSC?VOWpiHR;Ld?EYUvg1b3492XKpTj`|Bb=O<g9ymhdW!Q5oNmfpjmG#
z7!=51;si1loN%b60IOI&{NU99+K~m{>B=pHOS%!mY`39Y{E(>(9e@h~R%ohg2Yw<K
z{{W~;`q*AT$gw}841gO7TqlPCR>uo9Gm1(?3!>2j-9F?|S?z^iKG*Fdw<%jg^HV1{
zK#iI;>laM?OT<G<|Fd4e^vI;;D8AidJ=hcDq_MXIDaL`9jUsK=<;wKd9uSNXSGg<6
zAF&`xY^AF>=0E-KazGahCz&8jrtr>FEhl-Y9oOq*>&t_7Da|^(-FmcDDr&`!7{QiS
zDTn`54x|n@5ETrUg5aWAZE*xR3Lt;B(m7j&vI6Lh(m^GeYW)fj5QOWn4!4KfAzM7H
z+=73D4!dY@9rA%T36z$taJi+(`6Q&A<5hXXSTOBWP(S^)7dC$s0!svVy>nYY4;YZ>
z0BXhh^-tvYKaly2(NQoqm?=+~YJ@QWJ~A7Ea{$H|x=|^-iXYU$%N+pVNzoB#TbT4M
zDRKvPG7TM{<!!71SA43oVaF;=(Roxn6{z=-f*z*fhHr<VgJy@@YZrTMWW75;+G>*o
zP6qta1|XYdo)^P&Cxpm&)~0O3rBDwfs!|t0DF>V`>VZ?q?7z$+_THY%pPHIt$uJON
z>XQ2KVxjp9d103WC|#v5cE+XMALuV<juC;w!4|oP?QoC=Qukd(sceQvXzaDR@!XHg
zDLR4c_ZO{&G<RR!0O+L+aB*;c5WGRwr^D4ovb7O|!2I7?EZ!{oNg0{(&OdfgPL_~B
zgf{|1nY~gXYL%w*@l5`h%rTdej8UM;EWISGykuBr1<Pi>OqMb-!vLV$5ZOGULlC<Q
zdTH1rH`4u!(iDJ91W+FZE$wWtH6r$b<`cj`<uB@iTbH%lbd$y1pE@xwCdt%Hu{c};
zDn-v^wXG4RWQJKP97}6Fyg?r7>2=deIbasu<4?_4tp5!RM1W<ZtqEL|T#u-aF4iZ2
z&DOvO?`P~tYk@wfMfgE@CQ+&qk-Q6UPsPo>ZT_;yyj8tg)v)&O$Y4)yHS4aaJs5w)
z&!Do`#rtU?vcR!OVq}eTxk&Li<xg708d-xFSch?PU}kqF-8bCuK@F@hmL-~NDZ6^m
zkkz?pu4Eunh63b3JS><5Tk7V`T7^GK1EA2g(*iT{V@-~)3R@mjedTv1Ij^&;;fVIl
zl=D%LFSi#kB2Uo1Gy6iBEu#gXRwsjLd6g4+@+=vufj|@5eU(hnh-q+-Ux}t>m&pbQ
z^!zNCLm;MA-Zl67@u94&Q5M?(ng&NQX7~Ck1kl6zg#>yy!k1y;%?0IVjr#PnzJW7S
zXHPpHlARUszh>hNi!lha){*ACNlK1pOCp*T2S9J%uxxuaM!`fED?xvSCRms<K&yN^
zB^kTMfu(?ac;C=m4j`AJ1mm1b1N{p#!}nS|Rkp)5s#QJz`uyupfI$Z+hG0<#EKHYh
zcZ2?l#XbvXXJ_9VH{@CS0j>jgg^|H8cU+yjj9Z^@7(LO|(~|}EUcVrLt{Xv|e?uJF
z>oBrcAY*0+&90b8wGKcs*L-%mYApH%2fk@pPM{)KsMs4I8$gD!GA22SkBNz)0E}uO
z>$?Ln@c1$SZZ3n`RaTY}<eIUo?s6e8-NJ7~k;@{ukku@eWE*!w<bBA{yzwxo7C@&b
zEiEGna@BErhL${Xx(d8nyTz|_3Lv8ul#Q)*0Nee!tY_|D2hxdya*IKA5gY$W4nMod
zO+YAxWl!z`))rWQS>=|a31e&a2GpoQT_)zBvf-Wj?XWRh{!fHJ=crKUaLR1BdW4_B
zCU6;xdr{VB$yl$BRgtwhn!07mQkmY3(mVP2TcIB$WUFk@0C!|9m7%W1H}gP$AZUki
zQiNpITmWFbCKynM=$4ivm2*DSXNz1?R<UKV9;)hW3jiyLJz|PXAuW`VhGn}NY~C3!
zy@)@)_1y`edZDxbL+1X(=_v?=r7k5$El5@=MzfZntGGZ6dC&?kY!kkM&u;)*e_JI$
zjCv8Wh|jpB9;OTMAX2qT2EIA6<%|2Ci(3YAwu}~M8Spk`_Q-~gEi13f@YGCl3M8xK
z_u9YTy@%omY7!d!t$f>%d5gIGBM<;ZookCvgKiysf{bJ@xmb4Lamp2w&*NVbJJ(wG
z-PD%N-w&DIVC@2#9uOCbDlx2gSRi9%i+!1kl)F_j)hYlP?%0M%+3`R#?-i4#QLPB_
z!WZ(e4hZy{K)GNTuPImMTR89qB+0p`IMAbof)q5XPt5(;VTX<PA(tsdG>_Z?9h|r}
z({L>+XAAY~<-&_H+$~F9pM|Bw1smB|q?I!((-w)2th9ll^23m_RC1@1DFk7BV?W5!
zna!`C$VV7&i1bbc?0->{fbt4#spD0SW6IL1!r_K9>+I??Vv--$s_q_Lvh!4RzOoJk
zx?>G$^WDOB5wXpl7|}3{Xbi+S*o#<MYr$f3glsW1ShZUm^I0O>8D#!#_#y|>F*c7?
zk(HS=chmWxt}HR}=a~4*J32i~LHt~t7D26P(6l9O54v>nuh#lYD8on{L$&agnfL(B
zh-6%Bvf@Br|8yVbhAdU3m0;pZ27;0N;73mRPn)l#n(Wk$@J{`p2Jjz<72gia&|^xY
z#gQu}GL0wesU485P{47HFH30@8_NOB###X_wn##5{dUf<C7vi~(K}I~^@Y~&1qn1f
z5;Yi5-zt{)gKCKI@gjf$YGi3oFMPv?1iwXezqqVrISBirbM0sB?(6x8pE7aBdP|yr
zD;k7J*2{7qM+6EoQcE}NkTy1tb~{{h*|SAYG84K3u3t}q{wqhy-PJ_|w)@7AG2_;w
z9j*>{x7K9AxYshluxM#sF;O>;l_gy{)&p=e6h>I3sf=DV%WyYpr5j<sT?k4(r({$<
zS=L;l_yWjnwI0M#FuuJ(O<NJ8jcz%!_SBm*X1Ecw7{SP0Wxh&_D}NNbK7RaES#U+G
zp0q!d`&YQsK)N5+;!^=SpleMX9?9m`um^Opn=Jt)d{73G@pjM}-vGc01w?;iB4Dbr
zve}>RZutSQw`KI-6e(<6Y>bg@1l@ab5rCZm%`B2lNnS;F-Y=d#S(HX^CQbF#H405C
zmc?S%zY(d%k7ot+;nI5Ug=f}3E(wb(89Gm$?g5zkzbYz_M}UsXMpFj17mqJ}eE&t~
z1HW@Vh@&8NbgosOz2(n7oo<Bv`U_id)>q-7l@tj0-QCdpUO}{%$lhOY+rNr^!W6*v
zr68vO0(}5+R2E%RWR3YUL<@TUWlmy|0YX<m$M>TYFor+RWdZnT3_AGr&Y!2$=~0&=
z!x{9ThuU{s(sV$S-n1jGjcrsSQS+UpOrNi?1db9&O$_jA6@c~?CKzY-&~Vc<Rv;+9
zf{s^<M49LXSvS=r#~Zjs2a%COy8nERKj6%dL8{%Ev?vW*`lnhJ>whJfBI_(<VGZAY
z8lLj>;AMb*P)^=OXTv0ynKulOsG(tpS);`+m>h5U0X-^P4=bpFxza*GS2<n}==mJ)
zAq6B$0VExP3seAVpR(S;SKzi3UUY)G8*a=CZvt9lND4qJ&KfWJ5KzGcX(AM#96GpI
z-&vR`F9Tx8J6{gm#hfPM<uq5PJhbpn3m5^-YHFm0ncx*wtY)wf1aHx!j;De06k_zP
z+mZpj%DKr>c23%2ow3USAMY>xP99XA5!pG+_~ReBiE7<O{D5%)s?R4|b~Thi&`~M7
zni>XDQX(wqkvt&JKPC^%7SjN?ZMO*I{5GrPrBT-`_9B8wB1_JZ8@GTa7&ea}Eq@Ps
z+^F)^)IGsL?|%pj%O!WH9+`{<s<7v|zn^>n!>=S5iZ5D27q-y9iz{dHDwV7lxN&WK
zG8s1k<D`|BuxscN4_{wDFxw!E%_qVZW|+gwXAEJpsbn>YE1*ZZ;+kmT3jBpR|M{zz
zy<*DG1~W54f1TS~44Ta6^?Oqe;RQda!x-yaTcP_44%vA3)@b$lcG3tDHK`!zK?=0-
z1Ez#{ei)O_n9O8*iRSy<#=QEcJcQ*)vFZ;rrOKybaFV-*C_R!{O8LgFTMZNIARFkF
z{V{A0x<sGSZb#LIko+WM-H+m2ES)}oHL+!NuekwRlduqn>$9WS6If}~7M#`azi8f5
zY%+3T2WqM|1UC<SKWSmal-=n+n}hiKNS(pMOUrG7Je{Mja>@TgV6tCb486eKp9}j=
zqkcUf9Q<dXVFrtUaZ9M*$}TmgWnyheqQ~T>;yAy`F(x6#A779D;1MPtc<BPC(z?Qe
zKHKJB%@j}g2aL^n8;C!`>%t^LH&zWFk$9pn@{_Hv5i@8w4j4&x_czulg0u%m(<jt~
zh)vME<Nn`!JQl9aeA@_Pdm&rknc>S!zbt8K{NH8}wFk8li=Wz4Pcj#Bj8<o75x>ur
zeoZ;KB&oS@SA%ZEL-3s!3^fgXk58D&3$j>_;~_;WkYV&Bn_pY6xkB=VcKMr@Dy<8k
zq?um{`mI#(R&}oDTk4hEIg2!c!hh6EZ&mD7bKCK(i%~oFeT`6C|0^AF)aAljnoNFP
z|5x@-<;}}{qa5##IXSv4|Fdb?1E(UP^(zkZr!TZ?iFYvdD){C0wvAz|qY}Nmty=SK
z<*Fd1D#xO?%a()FC~ua^HB(VVLjHE?dKC%5NFYrZkXlDd>Hoh65>2;U+ljjaB<%k?
zq?H3N%R!5f_CnAb@YS`W){Bs~dpbio{bUEPf!RFM7RKD5;-X#MAE~K&eab08BjMot
z9k&uHeB#$i>sN<?df;l+(~2wQ^bWee4*lFL#WN|~QZA;*u4rn8l2XyoS~@i|)tt64
zW6OU6n%jnMkuPK%bxpT<o2(UjXrMOM90Cslm8xR3O@BZ(fkU-xAXD&+0(eQ=3j!|<
zEsxL!&&UOU%H{ueZ@v=sx@+rh@iSfGnZp~UhYo(7ry{nDiH%=M^IS@?AHk(I`qSXg
zPENEvSXTUna`NfM%_qz7#(iI(^A}zpc=PjMD^q}t!W3S=aPs|wS!Pql`T(V>0}_>k
zP2Pz+$(zpmJgM}EGPxcQx8i!1N&fGMtJ9~FhO{~;r7!UoN7XLx8~crZsa6|VyrtvQ
zDD^G-(Ac}TP52W{v-FaGy#G?<z{oiD)xjUswU||Yx|wine)YM)m`_u$6PaaCyR-Z1
z-yO;tTSGONko0kuOniYk99M5ny|O&oTpnC_Z2C|`b1eGZD*Pjd3O#RHxL5vE*sWft
z>Hf-+iJtcvFs@N4cj=ZHV)Uj;-K5Hv?sq-+rm}8dCl6B0{PPmk($>~2&XQ*AwENR!
z@cH~eHu29j)sTw}$#tg%lv(S>m4<S<{P^fM;qsmX_D_<HEy#LtLV^zM-axv8-CNk{
ze~$B8(l7QmtdZTw$Vs;6b?n);7G|y+=85FPji^(k9KV*}fR3e0j2j!g4NeL$h|KrP
zYLACQc=wmu_eA}7SI&oJ1~bMM-fjFhuS1oa^Y-+{1hLaA*^2r7GLL$}bhPv2lf3-%
zp~nc!;{5KC6UgDYrr3N6qi$mn`KfKj*VYU>_m&iM=lnB;UqzJgg}LkPN31QStBo}H
z?;y%pewujqkuy8p#5sEtw32LY+R85UO?T@=Lr3?6of$9DS)aZ8PGz)(*@Xlo-;pqz
zC~X<xZ41lR5py?*o|v8Z9RtI^F>y_QgBP)#E&clu#W-OQmty7D{RR|&FJ#<oJ21hg
zN-MA#Xa4a%`ug#9il18ub;XkIzl3tmMFV}qX;fKvUp}H`3clGR+gz~qcdVX&wrgiD
zzA*Rirj68Oa}U>rSFbimE%Ec?ro4JR|2Nij+=&D|YuA=^Tk|YFo>i^f@%L9&_};H!
z?8*|UD0)|5+o5eb0|>~p_;$^wk55g$^WIYTOt%-3Zc+KCGb(mwZpzY?fAV`onbMAL
ze7td@n1(vxX3u*9Fiy$u$R(=Bxhg1|i>A=VrT5vJwr)M8gTlOAYa8_XI{1ERn>H)6
z=-BYHdZU0DD%hnbk62!>MLfUpak=w9)T5;N{E&~+_0?f-wQuUf%<}iJbC0Cu7+^;<
zxq`VKf7%HQwi2iI;O>fV@6Q$;{${9DRiHBzVqBAOH?*bd-c$2svh7x1K}<_J>3#o3
z%DH?`k@0f(h9R}JDJP`gB}Rw0_Me>LX&-9aB=I7J=SUXl(s#Ke{JE_t3dZArn>3S=
zb3*VF7c~ccZ))oJaH-}b$st@s(;7wPX+PXj9H@nmLzfloUu)HS)YK@6HgWh4xP?7#
z^QF0mFLIM7=k41{yzMMc8qFP@o9V9We|bc`xKN4SHRI42X8Uquy;Q_)?Ep;~(zm80
z-0}RgMfL6%u?7#)J9yqv)$|vvWYq1q*SkKv6lT(%99E#s&01x3l<q6~P4+ph`|zjx
zA;K|jo@f2{jqjYON~dDAJku{1y#2OO?o>s?C>Iujb)A2@)}Of0TX{*lqx^tg$Q5Es
z$MYL|5fMLjX=cYZbUx*d)Pno@@@3*(-tTaJUv6iC$K;kuW4XNI-4w0$<Ur6*^0wX8
zZtJMM-;;~a|Jmk|e?!3b<c*179j$q*eTF}@Z6efQ9<i4TnwP4R25Q$t+&=cDCL29t
zIUKOz#=EqD&;IQm3D;<cXzcQL1*@&l7bc#&b9L&Ry=|i7;*wm~6*w&YBIGh?_lJ6v
zB4{?;X7diJga_xQzI0sgnZ8-%XmUv|(P`{GD#xMI;A;^#C-vSd(&=q&N|zIsYNxJ#
zDtH+McJ4B*_cddvcgKQ{%wEmCk@(S{_N83-y*2Xt%94ZFOZO2^9&-BL=4M{sxnPph
zt|V9U@QvI)tx+VqR=WdyRd9q##VA%E++BLf<;#5JK8^K(ww|MxdH3d)-w=<RTmJh8
zcj`UzMA0jy%twbyM_!sk(Q>UGS=!ii!I2O~?}~GSt<&X6u_=DldKy8A7!?!b;c5fS
zfLzGjAFs6cl+uwu#&ho;EMW6NE8khr9vxi&fmNiWw)S<Y^xfkJcZ}DGS^ExPcRbVT
z{&_=GPe#5e_rkxG@O#|G7OvjTSJAwT;??*a_kuXzTNqQGMhdOj`|(hzs359iY}ec5
ziS7oGQsxW!KUMmg*ZF81JKFXU8Z#LEWO=xsOZlzolX0uJ#&59p>1;+sD1}aUp9-1m
zJQ6bARwpV4-^+U!#-Dlj`n7TXZVmE2%DPx_+*anrp*HNq;92zBT!!m0^V3^*^zT(|
zar^1WSiY|6+nNZ<eP;5+y>+gol3kyX1&n7C#I>0{?_NI|)yjLKee}td3hfip_%C7*
zQ?eYyvnP`JOn)h#w0{1gY22iZ_VeXl<&Lsoui{3Nv^V*myT3ceEbu0Ct-7yeJz*$?
zX*`Jz-1PCx{nZ5?+5P8PN%H}Nhk0F7<c?m;rju!pO;DRg5%ONphOvp^Kabh>3_V*D
z!+&4vi!RBVDTfyhl(8MP;D^Vn2i%@--(3N0;r{Vsl6`Tjv-6v1(Lu_?^Jng}muXX?
z<P?<N_?7f^&T1?-Ky+c$pRS2|pg|tc%Yc<ml=C|dQE^`JK8L0?A0k4Q?)8FR{ff<<
zob#n0n}3Zgqgh>diE}k8tO)-YJ;;e2q+=QSytYH|x9Sc8$Cl?ij8?wV>&%12hOi1+
ztX5xIb-%wNEKfhQ_t<n9BlEg|HTdpCZpfsujlzz5j}eD4EU&wXH49#@C*6KzydRqk
zy6iK3>S@^(u*fz$KNoeNg2)8!+{A=j`8#{NS@v+)=MNcm>Nf(;`O)zUe!OP$);I3x
z@D5$M^Dn1fqc*s??GmoLgp%*CHVB(~xx!`u8rFY!-o?wXPQ#^ZR{S7(;(2UqUTDbI
z57}1o2_YYUt^uPtG4;D4A;!4Ky}ffHzP~50rd)l;8lRdE+CyVKX=^S>Jc;%>nV6xa
zJRS}0_4=PB+MCZ-D1SbtXL0S}TFdKCb_oWePDv%=6O_n{m=~()Chxy~>$mize38Ti
zUVmq@rB2WA*x#?<JlBht(4NfRyhS)%((>C<;g-9fp2ge&>&S5a=OHTiR`o<d`OMCg
zu<B8o@kLW~g@PS%ckv~_aZh<RBwDWDvum%Rld;N&Vdtxa>_SCS+B>;>kWDdHDP8W(
zud#Kh9xuMmbbUzqa=l80=cU$u>a=O*^hDfHR#z<#EXAVnijt<VXk42+&0NE?ieEnc
zIc?YAUG3iv3sXg|?EK=V-|vhlkNFUNe|r(%O1&>0(+ekFc>Hrs#M;K0qS*UfJIm`i
z`$G&~xYD)}yu{4yhb#MH`BO!$?*NzFT><PVPx6VbL52+5%yQFTfy_kx8S+eNb1#8%
zC(Ppc-~Kh)hv~cXhW6Jx+65{(p2|JzRqVh!#^laRT=nx(Le@Ng`QDYXJ(N^ypk}*$
zTCvAJW5~-(Wy-R~$Oc4=gxzm8{}YkMde}-j$`o07=*^71d;h9;iQ*6Fit^vH-kXZk
z-kE+DWp%$gr9()ndw+s{>j2=s?+QwXmhDY>I%5%hJRwYzm0BOaCTm;ifP6=ZKmFl{
ziy77*Ts(Isia2Yc^r5?T_QBH)>nD4zG~duE|M%$IQTvRMEuCq92QMk7Y_lJdmewV}
zLoinT?@s0Y?r=xvCpyeuQ|JvWz_&bj|8nGr$&u76@&8P!4n6EwG%Kt}ybBIeuyEIZ
z_mbv!^YI){h<r{u_BK#=4@2(ts_%!6-Uw;$5;HzalGeVNT&R7xarNdFhd`I9U-8`0
zWc04O*4>F0<F<Z23b?D^!7~G<7xUMty{zh5+M0Ij%888u*gGNEZ6%76z4*%wThE7_
zBKz~bvPAzr;~bLO>-mniWlu1H8Se%uErG)|Z@g?bO7`$~bPelF&w1l<>rY+4s`nSl
z+G3pbFolf5r`MU&{M*LeX+5w4#`;?pe4DP+DFoy`*4|>8$8${Z`(3OM>R#<cYKc&|
z!uKaEq*ec=Z8V|BzWW5b{P*X}iO%!>r+R#gj(mS&H2U=VjuG9cEE7bo0m=LiMt)Ud
z)(3nTbYPS*g8ujR>^|Z2{KqS4nvZct`{TENcOy_d%5NR#Z|H~k@@`Y1)P;<}J<IPc
z+lWHFaAJfQ=-zsf;(gy#qv*E91+5PTdv1Fgyyiz%)^yZ~Z++g*Kq;foru?};6>;K#
z!<SsSR?_<xX8&mCKm52d<&#%b)KabduH#Lznl~fX2JexdVdqit<h9p#?Ywm9mVu-F
zas0`QOzxq1G0NlhQ~JZ;+SkRQ`gDU7vHq>D-4%kNIJ3gaq-(bodW~*x{kyAn%LlWz
zq0?@1U5PQDZNA;T{ba>(;d66{cR=`d9Qmeiw^ihmd(F)EAGP&jU5jCO!OQ(+bLnr+
z9nlIRKM!$nVF`wbxhIusxqI2`uU}HvFnH2&!bpQ0da`%k0RGcd%8T0nZts^BoKgYe
z!Hqcf>Sa@V)$M<`ZW}(ht5JOr71*kK<My}pF>eSGE~$|cL5tPpI=?^lrowLSlzLdE
zW4MJpd;Ya3V0LSV#qVT3eV_iz^O&am^{KaVq%f%<X4i1M`SE*~Z5pW$9cLJ}jC9Y<
z*3}CKf<C;@)Vp*0n39ule5B|2-rN4VpK{%r%@nr!j^-*P1U+u48aAd@@BO+PqC<zj
zM;bi&>3FabF6QL#nu^PP9K3pb-y3DV0}8fl-bQ##Qc{Bax{3NJR_UqqO;+O0@66Ct
z+fn#^N<`J?M&+NURB$~7BP-8mMl>$P+`Phj>7+|mFUGHwe7O^*8w$HXIQH#KzEPsh
zo<7jR?J(ABM!006{1P2?)3{j$TG^d@0?8xGm8Wn55Mx2mW_qF6STy(RnZA02FEZou
z=8+MqRI;G|47~~;q+#TkXSM$#Z=%C&|BKqA4^F-DygA8_{$g?1`O?dpd&AMa9nPMu
zKc;U_D7J2*>pj%o{e)EL9>f6~1-*aZMbrw;7A1#YDm4dr1gJ)LB+~Tg@v8lXs5gPg
z_FHS-hSPs4h%RZ}%0OfPNY39VtJA#o`-BmT(*)0}QAhTYR~REImH#MNFYiCW^XdEn
zn;ly2m{}{5AFuHGD|r`p^{!d5id}Z6TL#GiU&eM7PMV$`c;sww42a;MgO!Fe`nvJP
zVK*lC^1jXGT&+F7{+Y6Bfey5<;WB;46?9TZ@%h_z9OV4ptcfm9q6hKj*VuQ-cVEK@
zqQdKKuVe41AG-1Flh$NKd%<q)j$Cg&UW9>5TUY${P=kmEL2XQ$WmCr^dn@;t@n08i
zxsx_YO9O|0gR`d$&DkE~eQ~<)IQ8|tk{-`nA^OL!zE`t}E6*L@j&d=T4t?y9nbZkn
z6#w9*si#^$!M8Bgbe%p@6QA|OW@t#7(2~na9}d!Ernbl*Os#kA!ff$VJe||L^(LsW
zsX6@S2hwIb0;#ivLv3#F=SfnLm!(hV`zLx+wXvVH+xgW7!s!yaNa3sNG88i4_1c`K
z#uo=wH=60fSH59tc@)(h4E6i{^;2l~iQ5l>CT}^+*e&$%bZc48Il83NuBt!c(7<!g
z!~;t;hf#b(wExc|SN1Kb-JZ@rP<2^h?(EyXqI|VHb?Y1(1GmQ6NNLV`C*AN2U)|5B
z&JRC>HVrMU68wI1EQ|LzQumI*mrwD^tQ?cI|E|PU7A<Flm_J<8lJ8$%HKzaD*)GA7
zu6J2|6*tZPuPRU>(=7;xw|u)l+sFSD<<5`il9oMQdUp9bQrvoYz_M?s#}&~rWqrz_
z@4RnsGMgP1PN|KAs60eW$5|xP4fYG0<O7k`;Q!i4H2;oPIe;Ae(iFY0H6yrr!t>pZ
z1WN^_<*Wn8r$F{=O^W;&nuwHOimx<kZGk=P^=#HhxO0~k@JyV`Gz?>o(z9+aKh?nR
z-T9=B57wbOzb_QMJ3uh1^}#sv1mj0WN4pymxOsoF(u#H0w611nXXku6cWu98so~2A
zI$%*rn|2dV-c>Y5F(@5P9nNuA-{tHR>Mg6~u9}YF{X5YF1O(T?WukOjmA)<%_Toxq
z=KTND+*vh5`G;G3=o$nGDe3O+2I=nZ?nX)&kPbl@I;BgxBqXIvx<OJJ=^^*?-$(D+
zK7Bs{!^{)+z1FqX&si1*zQFJK%-&s9<&<=H_PZ!*<a^y@{FN=o5OZy3A>X?9ES?+1
zW=goHISAjbuTh1DD*qH%I<@(8y-mN%4IVfhyNj4Py^P(Q{N%na-bl!9y5(+afC;1y
zKV0do1YV`ynfi|?<6$i^6%Br)t5rf67<5~2Z?=$qmngB2x|#4^aK;aguyinigdm)<
z?jKy8v;u>9@L2>r>y2}0N}x;jVi)|sAm?j?Qi7mzx?2IS6w;v2=K0r)W+7e6NwA}b
zxUd+iULA-Wf*d7R$jg(^73c$+?}`Kce8<OgXg(SS`{Ht?4)aZI$g|R?`#(`fKzC13
zNbG(iWl|-mq!;)nxoi?g@~kq(hi|hpL^Bh+;QbHb-1G8le>~Yl%Z;}i%$BMco!jnZ
zZ+opXOzmg8)tLJILGl5lzpVSfi7eiI|0I(I8&t^B)m@(m{y*f5Y3Z;e@B{hFMHKX_
z1I!-P8teH${FSq4cwp<72ovY7HfdxHCz>QMO?I82R2&Sy$u1YS5Au!3L?c(zh{Ry;
z<oIPWCim-nqpRakcb7*d=3WrX6K=RrChiEalX`Tl*`;OKHkFw4A967bYOc^AhBQQO
z8d&Rno}T9Bcm8!&qJzloSbn%%#9}JzSD5fV<;aOe0^gx9%wS8aYru?(1262wRW(0|
zaDx(^Iy}|bt<`mXjyB!65SS#nQ~C>}RhC>_kltZYtg-~RAZu*Y0@G{hZhFij<s#$g
zmxCVB-VDi~GZk9&n*}jkGV6rHq7=&`)u^l7<rkfhF~0R*gyFH_S0HVd)aQ4S=Ou!x
zzRt7#!p~auFMt-{pcqd)FZA=2>nb&wcny(3j0}vfXYeU^*Oo8l10Qor{GXxP$rbEi
z^@lOOPzMrtS^rJ%S(Sb64ZwPL1U&DX)G@i$J-C4kSQ4{t^~Jfx*F#|i!coD_8Cn~d
z4E4LI^ak>-<?lYri3(yCoWwUTYGN=NPmveNlJwa;XTcjARXm(DyIh|%K~g6!)3nku
zi~h(8KsQ-%r|>?Up|zrBJnsfJyNtzTLq8vIXwHM5V7zV_`{#=-;0mhZ<hTrkooH}x
zCtE!nRn^V`ThZGrt3sM}0n?K}q)E`6bycnFEzztCH?KSbN0?2w%~)DyfcIQ#UjDcE
zHW-8G6|#ex>qe))Zft%GPOM(I=kEiyJ;71`<yh-iews2W`c6MF_p228w&&Z`iKkzy
zt{oHM7__vsjHZ%MoTy7!ri3!BZ7x9)3x?~rE(mqIo9pn0s}dW5Jod5>qfZayn_g9T
z(~0_99cS%hP{%^h9VV0LXxF^1QN`7y@KlZ;`KRlz9#>IdqGTLo>xSPp2s7oHsSOGr
zNfOk%0&chl3xKk;N|kIr13hekW~W#|1yAw@*=_vk&^rJ<Ae@3cp~%WDa4?yM`oW36
zJ~fInuF&J&-0mv0vjFtiT$QRIf;ujq*-9bVj7`$Y2QBVY6xCOtk0@)2YjMuh_Y#A0
z@E0>|^Yz6JN~XPi@u&aUxy>C{_aMAzKQHDZhzx!@DSo&mRWKVjG3SGezd~{W)Cc$u
zPL+`mYqxq)$ayT1{n&kLC>8-<-+F&MbBA1LhIN?Crj~62SMlGl870~~+cxqVI~Zf!
znh^^-uV?Aq{?Rp7TKDamml^`8%s(~pXK$@B0tPpgT-!rc>qo<kh#_fJR{t94RV7y%
zPU}_cJQ_)*7IJeJyh!TrBDrji7<70B>FK1)N8<?O%MBoZf8zxgGthoGGn~&uN*c&V
z{(!LkJO41f8knicLdV5#fO+0H_eByk4h%qcieCdyoAne~n48Zr*PAX|!|~d#ZryPO
zi3CqGqXTym_)EtJahGkgT^isOLhlZ`j^O>Q1CbhPP@OPF)tOMAAMd6d2ntVlrKZe6
zo_T2rddSvmeE6G<9yjRR@O{`R^01miaF|Cz=r#G+Z}!}}pZVxNy@a9ebWMrB26-uR
zy3i3WE?PS%iP1mbGx%T39enCN7QddBX2kpJyU=ac9lU>DRf#yf`L(}Xb)%qEgc<fe
zsO1f*PzE~uJXs&J6lLJBok<{^yYh;Yi<Q>o;(TRVmTsrW_78}kH+%daG8;#UD__UE
z+pa|wm->h1^Pi7rIkXRcAvik989z=M#-0S4+dV$DnDzkuRaWt!bH-eMZ3433`#(M&
z6(w8SGg^*o7Ez>Jx_?Vmva=Syz-wy8Gt0(Y#J-?jGFe6+&q%}&^k`RT8%uBza8%pa
zHJg>G@an7+H@fA*mCMdfF^R4pnCGfn<e6ngz~nakb}$>^j%J=)v&}#Y?nhdql}aDA
z37sU+xHch42-3rb>L#d@qTtmgKc>t`IN;&UEh`~|DmrpNG1r^z4=BCLNRnboCPZ)s
zR_<`SK_@p*=B0WUtTB3bWC~`+E4s#$vacb&nKFEQM3%yDYO4gV7=*o$GPsm{jr~xL
zS47jMlxMykQ?j*OVW#@rT6A8RG6mL<!be-gDT<m3;0!#|V<D=Vp=upv(T9HUR!lE$
z)r)F9txyV6xi<a6=A|w&+V~vVa#a1m@8;%gzvg9O;N9?&6XY}hkA8G2FgM_+Vt)YN
ziUL#U$97E$wuEZ&C||cRcRe@JtuoY*38dj%-B88e(^CtGq18Q_yz`8xKX=C1yRYw}
zm#YF6Z=GjGX%O|Gsj@9P0%Y^|<Z1W~ucJ<|W`4>hZSR5k`9t;*e&{Q3cYai)VnAo}
z;Iom@h_AOA$|}y<Wu1w@*aa7gh^dJ2-+i{I-L)FPD0iWzupoC?p(#G~!6}>;6~6uV
zgDDqo7<#7Ep`1(1r-kBsc20t5aMJ9wSZSB%UpmQGGVmlv{>eM`@<2)0NeD{8I<$r8
zy#I}3*>5L7M{K$@M(%@5B%+1MI`#5r;1cdUoVhmmEV9#R7bR9%v@xbZi=7}h^_*B`
zBjh#OcQfN{#&?Mo3Jv8_Hz5SYaK|(P{@Y}y-pf{n|0rWw2A|o8Rk0J|XcS!Kt%F)j
z5?Ne3!&^oDQ5Lu!vlp_^EYn(pyFu1DGSYx-19**U<bfi67>Y=Jx^HeD3u{>j_PVYS
zy?89#GOb@g2fHyo=&lPoYu;W@zjFaY90-3@BXNToRrgQFjUpUE`l>OgDVz|nS>_38
z@96twQphCxh#<;K!NUBGrM?S9+6t(#k7`}1mu4>c@`V)M)fxv#Xav5B`;GGU(zTER
zMp!&U(gkLZ2Iyo78E;Q1S{jyoB4ReXC+H{y2H{G%#foZzE!aOg;7xE)kaewuWFl*L
zbCT-NXj)H4wY}ofDVO6SK$3JyE97M32(qFHWJpQ-+ktH=I|;<VZ8HgMHJvk>7Ep15
zyZA3k@%k=$NCOftz(UJt0SZhP9XGXAYiupj#bo%h0ySG<2-#%wf6!8lG2*)dtPmx^
zF2^hDH9&DxHgeQaXsGkQ3igWjX>|Eat5C&|Q@PD|4#a}@B|~J!(pZmqHZ@3&=>`)F
zkk$QvYinme%8Fohg1{BVMJE=w*MI^ZFc0}}K**ZxE)7^%(f!_DYW-^C0SIO5LNN)4
ztpv7zM%Uyz&NuVl@AT@H@X%wyGS|r|ZXV0qCW4SPUfPF04zJjBxL$rt{r7$a1`_$P
z@&;*FNP-Th^1%Y~5mH~ayt?;Rsq~%P;k8+n)2p=#N{fB>o0IJydJ=5@b9DD|c%jLg
zX-2L@SwK!6D|>dx*gw*ChG`U`x(|?slr<ADQESPd#!byqGnBQgWoB$9E`ppk)Qh%1
z`{{&I*ANW6L8d%QTM_DK5yGKjh!^NIG=NU+IGWt+u8b%AmvlHr+<uy3_CvTvE$gU&
z0e>Il-_B6BAo{-gh1gEdJJRb7*9zMj!$M-mmAVffIw9K(zg{6Bd``bP1)V083bpnB
ziO-oe&%@~Hx#+h>=<{>7W06x$p$JHJJAIZ2zZUN{L)^e1yqQY$WEJ<5j+2JC72wgY
zoESVVCA81PQ+|5aoC$f07cnmA%8ekG*{_!g>KX)r0%d#c<IYxFYZVqsxsfk3RfP^|
zd|sdLstg)!=SRsJJebIXE<)Wv+hh#UvmwzNtDCj@1;ld0xC~!3kqVUAN58o6E5_vR
zLb30<Tlgw+z!KJv*M~d#INM0WrMdiWc@FSYHQ|R*|8@=`jE7JO5kxWEGsd;o0<cAY
z{bpm*z>9L*2$Wobr;wx9)3DvgN}nE!IKO;yIuuP0G>Ytn{W>^{PK+P|7KtwUA~Y%y
z>9jiSL0CN9?;q&j!M8qd<wYk~iG%HQH##4~z|j3gBEgBrb*N}wmRLHs8bp)ous53O
z2t37gQ|Q%d@Q+y;xQta*z2v!15DDo;{oLAr4ySp=hW`#lN-;>yZMa;qw0C!s+?OPk
zy)&UCf`hRn+4Tg2KYO{qSm=0RSg$+J<#%D-FK_jJp2(2`Gi`*r(<=repP}BqrHoux
z*O`4p#Y)@&iKG?7$tSM{wcEF>_^drwYJ^Kc<OxB>&wIp-%?YkKeu0nF3r;_t0Gmu@
zl~sLgrOg9}*)v(5DC%7B4;&fX6^$4s*P?adMP|cRG^P2@%s%snJeX-P-X2Pq|B^*i
zCZXWcaN{@Ztt|?|OzRJRGRNEy3Qh_XQtzZg+KPl&$cf4sJqi+~X<c%|Rn%SO6v33P
zN*u}kAD6CL_yHC8x;6FSDBjnkCux|>p(Ass*H$P>)(5hTka##*6iXH8Cf!34TPNmv
z&*jE;W#P%Rsbr#xt7O42tmcUU0YVMUpR&dRFHbjPRMA|B(qH@{FmoPOYfD_y>O%ob
zxX>okV)ZyF5mMx9?4!EQwpV3R6DG=2U?;u-34HslIiY2rI^dYXQaN#Nmh1*T1c^v-
zqw%Zr)lF#V8G&dhe@I9UhwqG{FFS}4!x;!Cw=$5HaEg)S*SEy`WpEatOY0Ab{$hqq
zq!(Mb-0>yNDWjQm=DXEifg%@Ybl_N!uD4&m8~XfH3foq!w?Kt+BzYck!st&j!Kp&q
z!=P(r!XR;`=5U;2X!K`{n5X?MqmmmAw0!DX4aRfMlW4qT=@%)7%Jg!CzDG=K|NCYD
zgjzoO^P@Dcj=qr!PJ>|)d`e_NULc5bYA@=!%#GV3-{dZEt>tjGY*9fjYp&F3w5v6g
z`L?JM?S~~IL{vH|7i-<mLa~Pf54lCV^vM^8qfM5#w+U~bXzrKYP+-vvwK#5ichn%f
zb9IODgup*k!s~4YJkZ#+rMLZ>ax<HQvp-0oH3zfFyf;iJ7s)IoMkahCaK&*%(Tk};
zaF<5$V0$dm#{1n)bLio|w2efvsh`%2HQp`E`fBVPFrcXOgTgB@W<OG_)S($TqI((>
zAVa1<S0bzzUvjSn-}-vtK$wV~d;hKJfuL|8xMcf%8(FIbsbD~8c(Kk75<ohw=K0=K
zcb%L>^;VYXZF=7M9|F*~v^Snk)ChkDYF)fdRmK1V|AeHv4vwxco`)=S)=ovAglreH
z$1vy9?Q*G&;HCiKt|7fj7S!>5mk(ii%6bX@Z*#O*IXHOb$R@SfmSC}-cS$}PRK?@h
zc%Xn$y2!G8UrP_WP$UEtWY=*<g&~u6OtvyIdBF}5qMKQAER>QrSCuDRY|?z3u@Jv4
zGQ_A}1P-jh6p#HN-v#FUc*u#42&As7<2&2OAu8-wxisAA(xtR^kDvO%HP7+flN`9t
zOXJBmZ^Z5eA><OAZTA-MKR3JyLW;xok+#44g?kk4?{X>n2QJB}quo+GJRP+4CMdAr
z<uWE6MERAOEU)Vgs}ga5AY2r}#WSe%6&VS!0OVeH_M-Hhp2aOKynCI}{_(<sez1oa
z`!zBG!8e@<{q3H*JT!%|+hR}={Qe0+WFY(P`EW2Q_uKqT!=FvI2cHQj?obBo_2(&<
zQIgyRKnC%2?Sht0wMa`};y|xUnRDe@&j0W)6iQE<;IF@psYLo9rpk)(m!ZYfm)At>
zvd9Pj#=~D2W1rCN39UNHL?uq>e%fPyVW=w)guDEaNd~SPkD?9?B<WPbUBA#Z-TVsi
zMRVUN{r+5Bz_RrOyHddvG_(8(Y_7k_J9IKHSA$+I4MOfZHgzi3tOse|uV5o!pOmzA
z*T$%*HEL)4PP+lU)t^_#2~61Kg~(blTv3wKEXJc`TY@@|H<}EDP0*e3(LR`qpo1ky
ziTO0X$C8x9zouy=;?kOdK^GQ`$0e)9w@(seVK)RN3LhH^3oa&H7<TosJ;OBWfIp(i
zfJ_2X7Fq^WxT+xXn443!ap3K?sfK!>#4FzkdhDSdnJDkL#muIoNB(uh2}H`x412+l
z{mQpja|rK2jdoMF_;jQECN-Gw6UxciIp|IBDYX0AD0gZ^EV#dxl1@a<MXd(GDaF}*
zE?F^Z3N;>7v$64>=-8F*v~rik76s%6)bbi!XtWnJ1&|^HJQh?P?5zQfO^jYgl;hk{
z)~`*7CfC&vPZ(Vk46?n(4vl!h(w7#*cZ&N}(sI;&$j6rrDm5&<A9YAe2&H7yHBlCW
z;kBU@j3Jv4r6p2?j#%fK3wyUT;`yOQ)B92B+HCo$$-1V{*QrgJqj8dpLVT#ttSa?p
z`CfWf0}>E-gWnCsF8nSZT3{s;gaoxO=NP6kQkMS&985kWy7#iNL@kFJl>kYjtprmT
z<y17zs_aw!sOr+A=tuW)%_B-^JDR%E1`%|Y^mnJ}O^WxoFUHJ!st+h(iw3yUC&tky
zF?C7PlyIWCZszNM!<+Sekl<<<L&YinqD8MxDNyWmXxJ+Jv|0Hmf&9MBxm1Cp7r($=
znNJF4a}m{9`<ygGW5dsNrtV|2<3GYH8Ah43QFly}rq(J;i1~nWg08?}kmV+|LB|r$
z`ehhe;D<*XN-`L$X$-$C08@uURgFPfr{~f3Y@kw){u6*<_o-zAiO7BU12-?|^8L|)
zA1EA)>%OSL*ILj=l50EEMkrJvD=&-3!O|1G_ZWJIG<5-Jvss%XARJDvoEPuk3Iz6I
zan(c+hQ8VYp1Bwa6eL{*0c(X$h6ogpg;z9Obo))KQTmkN!FzJhz#Riww_UNHk60YT
zZ$NlZArGXp7w~_xO9fCq+Od?QG|pQq3Z`!xtW$!ZQ?hK7_XaLA%JCQj6_(IqoJ(LS
z;y*Xv)IlqeAd_^!X7?nMu!-v(L~YE`80y89=FUh~?sj$_)Ws#!E~`?>!<sTFMgEZY
zc`=l7!uLy|2tiJ8dt}~dJbrFmIE$o}&Fp8mAo!&L;(seVFG+_}>EUiZU`SUa?I6Oi
zC>aNpLXTfL+%WKd&3L;t7cqcnj%_3MEoxL3hNJ~*Lk<3+{H?CgJ8ZbhRbILTo(<VB
zlbe#}Edojv7xFP6k{GLfLyD8=dz&MkpWw-EA?Y$1G^~T#L-uBMXheGIADrJT<kX?^
z_+j94NpTPte2y#p`gd2D;tR46Gbi%$+yU)ijl&%;Zs;GvSl$}Sdzei^52ZwVmK*I)
z=WtQy^PP-b#P}cBJ$6W{>CjPrz8s2+()I~qO~wMP3kbp+`NO2~Id^TO_6kK1+P~p<
z7`;B9X$62qV>XS{rOE84yQsDt2gEY^bZYq9!6@V<lUzw@4_ezmNf@X=!G#pnvVBA>
zNk`%)zx}nrYAR2(QFVO8s}^v9XBYFoY^>Spv1xzhvLytY&VNL}9=o8aQllEhDnRC=
zj8wK3F4xNI0G#aX8H*MOng1CeXj*aIEc<RL9FPx?T@s=D3<S`jNM?k_jb>c5c@`Fi
z(E($Y0t$8%Zl_RNkVcR1=}O7<Z@a<)%`ED#L-wafw_8e+VauQD5a6(JsMY0dellEO
z?|+Ar-}TeC(Z(|yb!Hgu2}Qsxh&FtP<k^d!>1&rJlk~&y&+si_R!@TXk#pcn&Nfp%
z-&rC?Vv^Xly$nBKNNcKayzDo^MGGR-xl;B;((HOFy1FoTs)qq)Kh1R1x9SVjLJpEq
z5X-Pgctu<MUV}yRjN>s@-Q1LnpkO^91o_O8>f^EuSR!Ubf@QaXvZNb3@|DG}6;a8F
zcfzCj%w(9%nJ?Dk=_Jm~Eeyc>39A`OLs`bM6bO+@vfh4K#CvYF-u*^h*8hc$e18|o
z<5mPKoqP|mQGI>7h>2Q4CUk&8f+DL>$op@dsrZNXW?X8@NGh{8(cV|&Oly7@=5f{m
z2ua7p9`0N&?QfcuYu;9iu=J{jo!1YMm1;z3@#_;;IZj8Ms#seh-V6Dj==V!0cf*6o
zdtbO)A>WFXDKrh@>04_{Dsk6tc~HF|&tXZ$zz;;Fy>(4y9ruh{6ME_?M|s;aqT7HL
za=d8I@hbaE)79DMVbf{G6J<y}%7wv|gPtW;Wg$DwMAe6^$0DW-HhP!i2c4H#SpMGC
zf}>O7E{X6}A=2#VL?k^J{_jwD7>kLE81CNGt>G^0(%%3>1ddFijPeEaDVkEh-Mhjc
zQdp#vR7xy#{1<y~^C^DwYok00>7@6ypNpPP-)%j(#|N3>-4%TqvKv`8?;>e16$3x1
zzjwQb<)s@T<euM#4evd*1k03<UEzo2_$r);qRdDw+<TS_^Uf&ccu1}@^g`N{TOK3r
zu=yf&8vdsg9djdHCNYf>CoC@(Y1)D8Fnnqf;a8MzylEh%l8xv+`ctMX9ry!Y0QwR}
zL*l6cXm3yp0EKg4C6vgZv6z`XDRs-Wx!%tx*|h@`Il66U3;}fYc0$Qm!tuk%%R`&)
zTwlmUSGNAXtvUEDM?MkYoWT&ZbeMnGOhNH%{X_sM0T96z<*f<4kJ$U|Wk&2|g~}%w
z>R<dcnZiZz-QY*LA29xoY&q4m(T3S8!n~bu_r9VcQdxi&bfL9+#D8RW#1@-Zr}k~~
zNY-v~aHDe9EJ{YVx}BI5Xm`l~;^DhSu;BOQIr==mLWjo*>#(}#KpYQu9Ls46ZcNqs
z??;BNKla`gJ!y(r2P2Sv8o0vow8kK-JX-1hsTR6;bew#lM4SwjV;c*nV`Aauwd2rI
zm0%(hS-7&j=%G@76E6pcYHX-M)~@jjp}fku*0~e_sEN0gJ%mqVntO4wa$&XKPntUO
zR5%n);<mpHTs-OU@+mLiEoN9#*ofhT6+GYb`vi1&>t|lh{$V(g9d#56Tf?)dN<cJT
z+DbD2)sx3G@HsFFn@xfzIc|U=D&xKN?N4*EQcs(nVkBcN4Kl^2XxgcH`35ESuDO9M
zKdOd5o;m@^H`t0CFjHu%!iRw#&5bISvfcs>PC_XqF`H4moyl2PdJX5$3sEt>A9?b0
z3NPg{qp23#0;BzV8klS|9CZ>ayOau{%irGv!i_dM{tosxRi@h@Mch3`8IS=Bjx-a>
z1c@_VC&b9Dix&q$(1u&G&|-?T&i@ry<BJuC5|lbwKiyr5aDsOUW>m;<pj5SRRh*3v
zh_23N<avnSKH%BjdvqBiETY%PeCqOHl-hX*jw$4N&tfQ1&=af9Hx{u<rwC^sv7~G0
z<t6Hkw-Idk`4q7b<o~y-3}sW-^^tOs+*u{HwvHXm18LOi_4{Q2A}a2#sIxQT)Hgh)
z({3hsOo26SoUnuZkYH6290NU0P%m50Au|DyD&)hbbEE=h#{>c_BsUyPnjhvdU)3S)
zSDAKsL}(nzo&4NL?u>gMTDb531D*auvMNfVC=W02!Z_QZf42Sx?tYbfLUhLNmHz?g
z9Jkzph{9j#BL70QR5hd{!7&4bZZKy86YsANXC3F7btpfG-%)Rlz02<k$dn6DDeu++
zEg(7;=j*X*U?}NT@X;oy7}G>Teq1=qCJXv}hun??9OQn?8ZclpGK?^E>Vd$O68CtU
zta8LqNy`l?OuPrV!$Z+ETnh%(y}$Ng&^_*McCvk&`W8jzc!}9b=W_#G<$XL1%c(dW
zK16@(z%mJ8$5TF`scruTNxW3K>7w+f5g7QTB`pQ-K$Za7EcA8tPXQYDyOY^X(xUdG
zii&a7*=dQ!-%CLNhft!hwY!a&HZd8agobwSs$#iY71&Vs&MwCzvK@$OKFz8u(3!k3
zmo7V&4`VHVB6JuQfs@}0aB;5WjK^_5XlHti_|EAh>4wDNv{UlSO7<|sq*2P{@rBgU
z<M1da3QiCUF7^h~lCqQ~`#ccIGgcjmq5@i#B9rVaq3N(hDLWY<OhlKn02~crtf*(=
zeqY|UZ82xmdw`g&7e?g6Si9Vb1o7CqObokRosInWMg7J2z*RuNnMEO@>`*KHwKVv*
zEXa22!zVR(F0#z`Jp@Yo)r*9PkT6`?y^NWdgb$6QX0H#6&|>2nCisSm3eD+WJ3V%1
zX>+^0UrKdO`^oQ{{NekEb57VqD^MX{XjmYZ%!{LL9k4B23at@ekdlkG%ilYB);PK+
z9CLrfeR$c}xAry$hXHJOnKpj^JV}id7vINl!1e4VAO7-ns?bg<E7;sPyK8N{mV6^-
zfgyA`O#Q#8UjRa1lE3dOs5#gj6n`HpC2ze+t_TJ5C!lu;z|!4$9F%1uaA>G#T#oq>
zV;)(Hp-k{4w0L2dvX&^QH8kj67C8eLf9n5}SHsDl0s)h;xZ);CMF-vhZ5fqym<Elr
zMqB7!I3CsX4J@s|qlh2gX49#*3ml-%aj&5fEW$%E+!B?hgh25KhX5fsCb35ArkcYx
zeviZk!O2C11uWN!nKURO6c6uys6Hi3sh#WlQNN4i>43Nk-6s}s7)F~mRH6oBAt!4}
zeoK}TjGdI+_NlP0vu&#gBD;U>5mBcth(~1u{Yrl$56yEOa$b!*mV%WWQI`B}SAS97
zKtV{F5j}i<ib};$xfCuv@+VTqq1J}``CCDv1CJpLg(5DO?X!vAg6X!+&uuX;)}UrI
zA?+{n^aCmfn>^hAhKn(mUxbP6{pLPAdYUkb<D-gDh~!l=Q8X{Jk<m$lCDw!=#SH~!
zx6Iz-Okct26XH-iNwuE@+#K!c!Oyq^^FeY{eQXeu0F_S>APiJ~q9i%Baxa~;t8m`l
zC8L6MQ5&8+jBOujIhE{aNJFs0=OQl0@``_;%q)0kU@xO2ccZ^NzsH1EOnWtEheRN=
z=qZir*6N`bqf_9z_8GbzHvilT!+652W#cP%LG0PWK&3<f%=#IT+;`E~=9{9=D;q&(
za7|3T)wS&0-Uz#s#5R<xnui4Gt|aHaf!>O-=p5B+=j%7bxDWUcB!j|wP(G_^H=g`9
zQ~Z7JnMD75<s@Qu78iVv*9Z!vZOjH%kpu+Ep04s=dF*8j&fB@#Tl5rz%naUyuf=-1
zt>D%bbQ|fm*mfD~b@72%kWhOn;E~biIxO}T%GPMQLEx|DqdGQ_1hmAL=(p7101XgB
zgvezjn@KR^)$f}4ZzbSNj6@|>-#0XG$VTT5;}Lamd+U}=d$HQl{>nYu6S99nba3El
zhwDNdDV)(^9G7U2umttJ1MdytlL&d+PZCGb#2|0vc`QP=2O`*t4eovj+#vZ2So9%E
zTgQHz@R>w<BY&wmQ$zMHiwp4=Srwh>skO;zEIgusB{GEt%XWB4_dR83SH@h8hIhx5
z@lVA@Bnd)3#TIRC-U({Zn_LBvU1S<DpACr{oKM@9vwP7NNM#foI&5lc%U&1ck<GGJ
z1?GkgA@q8W5pTShmC>g6K$`Zwj7Y()L80|Ym)tS+3hf|49aFRR+yq;2v%XC)c$dk-
zW~f3N#a`b3W?oOLGxR~=qDiPs?2eEHAr?^;?oNF1!W4`(PgnPLcq8yUH(4UJxIm@j
z47Vr1Ox%QL^%|u8fd-S1h=GIcH`OVzsK{|uF`Rlulb&}sV`5RF3Vc0M9$bk4CuqRr
zM<nKIrlY{6#CoWD5tiA-?Id<M^G_)8kD=)ti!~w^CRyI}KhP6KoXN0Mh`o2~x-Yk^
zf&glTqGX5dK{S)dG?qRA&bmsRA{94LdHG1jI-9P>2;vd1!MA@PIX502TCllQlHa##
zcx`+cxqq+E=ONhwV5^=%7=ZYwyw4l6&E?qL0_y%S*V3HAc@V;Lo&QSZhGwX|4D1_!
z_!CNkd(2JCz_gz3$u8(EVT68+jhv^Rh+^><!0@dyvQL=4l0`X+Kss!LDoh`50_U6B
zBxyK6f(Hol=~(yfL-mzrjV+~CieSn1rHwP)&(!2qIZn5Ebr0=1rZhbNx#0*VsVtsQ
zY+s*T0esljXRPrGwZM8m^&eku`U|?T<4t|$b&gfM3T@u|wHde8J_TUytbdvP+bte+
z>3%6DXV0Tw?E7BVv*Z>unR{8953a6zSac!aSh|UHZ{H5vV|uG^Af+V1Yxe0TmC@$w
z+;$VhQpuB;@2EySb>quFSl;^)Nfgpz-wl5Ne!c7{k5KSr-Sr70%S#$WOzoc2ddbZC
zca+4Qr@og*pX5FpIy(>=z-ffXug(4f;ERV<=_v%LD8W>jZ|i<+p299KQ#noGiDR>j
zShLM=!orclIh`3w;iZpViq6n`2@z9G<0(j^o6g>li><XyE8`_scM~e(*P~4v6QnY%
zJ(=%so{jUoF1&Ichtcs%0mG=DBCCFfBJWIUO>Fe~)NlmI^O9Y^Lz8sbs=VF*!j#U5
z>5t@(d3}rR0}vM$AK{?+BCS+)J6lOE4^0f(Asr={*P6tdD!+g5T+ZOoh1!KgBA)Ci
z?Bd;wzdf@$G$HqdcK;n<nRopBmi(xmr=LCxgxZ2O#M{`0?Dy#ZEw&eH`~@rT-Gj9M
z?#{JfW|arpsjBbXY(E!GsX8v+c4;$rMgMZ(>vkp+K;#5zhzowUYPLABvF5`9gNphc
zyeaL>r_vfRrCd~~Mi*G$1i$-1*02M;+}bNMBR7@#M%#*M{u5FH%8ud0-bt%-XQon8
zTr9WSSJ607^B>c0DXe_@dILI{*}A-J9?U&jDUTptD*U`stl-p@fp(L9i(NzB6m<k3
z>;J$wWMBX8j-(_QsfY(clS4DVP`9RrvT*xj3Qf7EQ>KF(##d^5r+L3Ky7khb{xdR~
z<w5QJCGIdo!)^3+7RM7OlONmJWRl?PCqxJI+%icjsBrL0U(UnT0*f9wdsYREHH5oq
zQ-!AmY2o-1^}JV}1T+FSd|x-eXET_*ZTp?ks48#Ww`rwzwsg=cDG4(37T1vhZ9j|1
zR@!b0Nlms7QI&F387c+nhnju+Q&Eg=bke#FMDC+2FTibRdqy^*5(x#;YD10-)dq|Z
zQj+#;H$hB|;H;4Qpr-pQQT)VA(WpYKfCJmv@xjmteo*;$O7r~iSmht=wNelYMQp|x
z{EA3gnZcQc;U>%kBY`nOP=|jsvGBD%Atb1}b`_n32RR%~#55RY6XJ=3NgvN~Ael~#
z9Gd@6ScuVy>cbYI-7X(kTErX-zZ*VNbaJ&nVKk+W6y>vDCGQhiH@;t1Eq_~0V$U9>
zMWEs^r7#u35bU2v*uLGV0Zsuazr;Z?7H(5y8lTjA#nvxCg;HRHltBu41ldhEEMgld
zuWs&atS!EwCX9a&)Pq;T<T$^J7sl2}v0QG4;3)aWKdIgho1vH_sCgY$(6_p1Y_Fo;
zP<G;obor-`i0*^TpCaM|f^axgCS=-FH$4`0Ztv1&xt%n|E`+%yK{Uqidsc+%27RpY
zKNfAzWnzA5l}{MUTx6lwA8UYFe-Ki1w@{`@A9$U@@kC>{h}U|6mYc8||0?VIpSR{e
zZNSAW5{@YIQXJ7XW>q(0xKapAOYei0Sj00$PL?6JbxB)~_)P928^UuhX!V+FA*`E6
z`^^gNpPLoJ>z;4=Kucg|a#qDR9q~w`PU`wq0kO3N{n(||nPOon5f}!)tGGbB&@H&@
zp@zUR(ZIxAvpVo!K&#$EBtD!)Bd8ufCX(0Z$DBs?s9WQqSFPN6q27K$K?(Z#?-<L!
zyY+GbwB;#D)tBRA`yJ_^pHoJ!5ArZ$y;Hoe7Scv1sQmOrBK5-~>!OYQ+ABj};Bf?8
z(g*Oh;?z`jS|24vnvuSNF+gS>5kuBLZi+@Ej3&PNz!R6({<+nf1nshbDTsy(kZkH}
zh9_2m-eVQkwNf0fZ5ontz6nJ}d@Qx)&79XF<kD2)f-DR;PX*d{YO45Q>d>Wl5}I`b
zYgfq$d_}?}EZJ&oR8{NOX#>UVOQAdQ81c-GfFJD+4`r2^XU5;<wFLmN#&gRUb13L}
zkz6zBvt=)qz!roU1Hp4Xi2D0f@=kOiXX8*5%4+k~#Qm^S4-DnG)h{XREbw2tgY6XD
z-NfA%*Lk8$CX4r&T7Co<`c5CGfxUDeC@@G@zskwPx=03ALPL=1Aj%i0rW4NEyl&`n
zx9c|Cl{We|LENJ8x!@*t)V?;r4J7c7(%?za&br$)Ym{NU)LIDRd-FUd!tO3U7uHsD
z7*?%|o5V);aR_iDPOylSZcE6q`oN3>5cpV#xZWZAPTykGZf0;=F_m~$B`|i;9REdH
zPHL|Ae?C1CN3(sCm06XRm7_Jgo!mM3LJ$v^q!p6MHYOr;;AEqRFyJZOppm-)(vT=9
z_)nGi^Zj%7%k$l)UKFWsM#%QuATmKBLKv=nV+JpXVh@PL@C}sxw^AmKkNwJU%|=48
zEdqx84<`OVJYskF>!43C;G+iluB~u^3ZI%3`vpONP}idY!0|A^i>84W6i>GE^KpyZ
zja^jSt<qJ-tV9$-rGwgBGTe$_C(AJI{+nwS_u{vzT-Cmipw$|;+b{$7?q4kP&(^af
z6v$fuo1ifS(BJ{x-G3rU=72y>!^VACmSw@8V!3*Al)^6E8k5qQzHj^i*Y^FN)YHN8
zx5QF*zLz8P%s^jCKZUD+U<Lp+whY#))xbq=0Fr@^s7zEGh=*Z9DRYCwpNx|Fd$(UA
z-w^qKp0Q(O{<09<Qjb4Kg9#;UD6b*tXMp7-zgTxcTNwj(Z1RKy4G9syI@HX({xUTo
z1enMc?-}Vwf5kA3CeC=Na6L!obCRYnm`jQ*M9`Zvix-zxhwjqu61y;+GuO2BK`otd
zzB1M(5+HxZsbyY!61nGJegq-{R8`Yzc__Tu-=7Nqg1=|Bd7BpdJ?FyJ#e-aztAr%2
z=$&^<rTSG;)JsV;@~ClRtR+gE-33`Eg^pQp&ljM~Y&GI+3^h%kwpSMir`k$$eOE?u
z>IQx$mzAF$3Je1vy6aMcOW#zu%tuywv1n!cX{Sm&=GqRRgd^JtyPsunC>Se~IQ8(5
zUG{pII?>q8jq+|X!(6{=r@l`z?cF$iX!y5!Tg%Kk>pa79PB@RmU;|f{(_RLh7Lu=Y
zNp-jzzRYMy<wS1f@T1L|0aRZwb%m|DM<em5dmK>4YD5!C`5I4m$uoSd6Qz{=Fbz15
z4fkM#puvv$xd-zT_24$l0gfmlx8VE=xTItFJNBiw4o{RrJW4=&bmnuj*fYuQr_*k8
zQHZRmTTk58&zljMQ%p|Zlcnl)jZ2hN7h*n&)KH*uE%-COIv=Ge(oUIkM2v#|&NDq*
zIZY*pD}${|Xx{FwC=hL%2R!?nK1_;HtWUy(ungYq-0t3JQ{QD9lK7#dDW6^>0>a=A
zXbCfDYp9u#cH7HKfAVsl0=1c6*LWi|9@*s)D3(L+ms}TV5}bII;^NKI+VHK-TTDUm
z!l^D)nYtLC&jE$e#03dDLnRWDtlv~NMv?yMG4J8(zAV+9R#zV}7d2Ml^>f<*@As#i
zZD4?24Tw|)pNjwV*CG`^l$oNgl1FnOe36u@YbQkifxH#~E}oFmUJoPycM$#^`P4)b
zvLs#AwE8-IqEKm|@i{E!8DkdoH@jaYtn^S9i%^@)eNmmwxWBR!Bv@fiL^qZ<(kZ#-
zdw*ZKb2RF7MXc^Sa^rDCe>i1ZMH2j^FDl@aEOZ@5pOR!pB;wFqll;~WQewU9l#GwE
zM=w%uW+ZD?zrpY1$C6+`GwPmgb9y>8=)LZY8^$LPRm@CNp4t^@OhUWroLp0u7vP9K
zC>>AeUoxU}aE#qZ<CYn61iS!F#6?bCVgGP%;%vc|Ub=qIaBKRVAgRzwRmbHfY_tEM
ztJL<|%M}LkGY22skhzDElhv|fo@`^D8<+mrwJr~gn>Js$M8enVviHZP2)}3+DgVf;
zKxu6f00>tk#l5L3&QlHM5HI+c&96O{9Jy}?%xkM1w|=7!J=Q@h4rMl{v?GS^0Y9*G
zpL};u4e^X*6r=SI5E7|0MllF%e?f#DOtD<#UDep;_HDhh6gKE30=eR`W{KAtR@cB@
z`k^l%K~8tdZ%Z{N#c2s2-lO%q>yRjzMz!<QSeNR+39d8jDQv54pxDF`U5xPDOort*
zh4Idu*=HUV6WNnz@c!m0PQrpfXWq!kCJ@_jUBg?hs4tSYJ_1gK3M_}K&F3lkha(VE
z%*+Gxyaf-AbqAHXJ|22`o#Ol)-puvi&y2KW)?!z<WFc(%kB8qAi`86-ku&H3&N3lJ
z-Wfg1&V{8fwitvEp3121oNkt8&zSkcwWJczXd!2(FiK%Le1w&y#-*4_P<{2(RHCZl
z(&PE7xH7dT3iKjGwO6E@1^dS`XO5tajSSy=8FB7jcLPaJ#-*!<*xlXYvF_*mLnU_3
zetJ!5(xY<*6(4Chn8*`rwAfPQ-W5&fxPVi@rCzO{G)Kxn1_PQ?NX(?(Q=R3v?%T@%
z4qO!G_nA<uf7J{-59ZXV%F%roBhv8@2H}1M`7Pad?J=QZKxPxH&hpDXFvRm{S8twY
z;(1Y?1QPCvOyNcxo<ZGnT|iP6k_?aa&IV)|*O9Bx=lt>n`*yK|_3kA{vk(`8Uegdx
zl-kvjQy2e=-aa@oD?TDDP6-lIup~>a<{8g3I{7!Fz|5l<_O)X-S&8rOwcN{XYd2Bi
zpl`>X(BzEktH4Cukc6K2#+?9h@T{VL@7zbtqrcm(S_&X%+(-{pI(>h$g_KA9wEM?a
zzAsLnq&rn5zj4%#_b3g2u+#!4i_gSMazAXcA5u2Yy{z)scjkox-UwDFBrT5sWTRiN
zph&CP!$LDI5u~G|+EBK(6!R;}_ESmT+_3cOd^2tsINAIinL^c;Y|~pGkwrGOQQ&8I
zxU0QEBhfTeR@HPh7kE3B^Ar$k8Jn!0TJV*@C@Gl%rKa8Fve6`D+Y%?lkE(ZS^6q={
zOlQHWtw(-8PMDtpjd1_M2C~o}f8&qqK8$2(ghPu5uLXH|j|oRh<rP@y1Ye25Es!O^
z?6ni`6>zd!TInbnfXMRoVK;^j`t7U&T9jl$S<rLkfo5%fQ_+|z*Q&EnwyqIhI5;9{
zf|+%QFbzn7v9lUAnr8HZyMc_?jQ2=A5fVAOz0{#T0*&<Li0%yfIfbxX`n=dezgkvV
zZ>|-^3cr12-khQYi>Wn{U<G;-?-bpYEAt8v*SF4c)P(jw2F!mve{}Wp3z40nFcCR5
z34z#BC1DpQ_MA~BHNP$A=J*Q~X)Mh>J%_$ua55<tjsFoqwi%zQw6IQIl|3cvKKcp2
z7G4tO@$7OgtNI!UNC0`WKL`|@-&NxA5;d;=if}?q7KJx+0^jxQpUy+}NtOHJF=<HN
zA|ait4UuN<P#lQW{8x0*@nwrD%~{Nn0?p>fGVyzOlHcS)w>L(zd(APu6(KucCNntu
zc}f94EG1yL8be}GPmBrlz`PjrG(eC*pk;6wb4Sf7&wQP6mR({_-iDl<`}Fv|bwe@J
zK1CHK0|(H^Og=#Jri=k457p`8)%Da|$j#wqJ586<c!xS)b|s3<Hm&n|dzJX2#s<QS
zE;f|LfQX67>9ZXumxMGTs5!yPK<1M7Y#u$}ApYNe(?<P-tmEwdX>z$}+(guS5N^!r
z)VR~1uiu_%6i*k^DhKa`K6)(OmyLH@Rn}GPq8F0Xc@oesGcK8YL=P?(Mxr%M)iUN2
zxH_8z5P2(9b?t^rXKZ70W0?wc`Sr(Cvu>@&kcV=|GX6XGb+U(nv=Hq_?;|@fSV8cy
z`75JVY59L%6Ey<Y7qmexCsk8n3zt6bCZmv)NwX%V3Za)>*sieY(mn@oLqFXMFCB}U
z`9SpBBjCWlE0beyPX2BLj1YjgDl^WmCW`4I6>nP|SeMaBvdlo2DKz$(HXUut9n8{S
z@l#l^rZmldYu%)Om8zBBW=xoS>M8;=Cfp?5=pLi7;QVRl((t*x@D+9EZK{60k-sPg
zp?m3vZ})z<;l_SG$m4j4j~lVZi)`fUc7v0=>})5q-KK;fN{k40^D14F`8Y#`;#4iy
zw(sS@IY0b+)T-D-zgV-?HQk+0K8gG5ca`iIfk&BK%vYSGd+3HJhJl;a%i^r{S@`pX
z8jafZe+l+4nHOP#lH4OAS1?qcH-|#<>%!izyHJ|*ljJD?P~q7*+fG1CfQV-0xQ9>6
z;AG8U4*7g@r*85Ls_4h;OI9AWWH^EUUL563RQ_K`cyolp!?$jouT89X4dHx*cRG1&
zrFOw-DyzG@U!z@QWxQK=71i9EvqzhPgROT__1?-Bibep((~L3V(uPfucm>BW%R|~2
zQg8i*c-^ILkp-#^rKnO)BohtdvmkkAxStzfo#*!3$JHZrBmfBOa->+^&YFTSsp&ij
z2oqR6Li)lk27`FaG-0*VC0(=EhgdtxEOApPhyhRAKlY*Ph9p87h-gTEJ<3)tfXrrz
zH{S{jwVjwe;4dY)TAiw7@}FL`6jRNq;{j8n=O1HS)iGA_`dAC3wHE2VUE^6w)vPKy
zz0LIrr8!T-^|)pni_@}Z(iN?@S3KgI)$>a8k|xOIp6uX8v-Bi*ZyKk%pO?&=7kS)!
zD!2=88=#8_d$s(?5`2~zxzjMsj~Je@Co`^AAvs=aRF6$^=F8}gs@%`dxL$Nsv-I7z
zsy*R~`dYDy%G5JE4|l6AXSPyrJaS7!?8Kq1k`w!1=&e+e4bzVv1c|&$EDdyWf1xJf
zpO<b1#sN(@8J}|B{skCek5DwEG!}zx=JylIyHx+RrfyP1hd~2IKSrI^wrP^kCx2Pa
ziw_D0Uyu-U$_wxneqwLjZ+N+7kq&X`1kT6Pi{VG{6eso6$VzE29uIVhr!^^%HuTkg
z&Ti(993=E*-o)nV+}4r*g{1szFW_{Q-`UU&+Px7D>MbTpT6rbm`CpZdH$cQ<P4nLz
zl6G<(U2j;Yu-mB7{$t&Jdj4;A)geL;Mm9kf8Q<D41-yXl*);BMe-ngYF}@l*nZkhR
ztG15~X@4<a{(Q00IWPULrK+ZJ0<>Zi^XH@WF}#MNklDveUwD7`DqwN=ed%fk1+$Oq
z9c06nB$jn1Bce;_T7%%0q$aZEB<J_a_;30ra`BM0YiaKbsp_o*(<<Ol@FcQjODJbg
z@8b2jlaxQ}Pfqb(xxsu-ryakT2t#+|0uJ%I)=)iYtqzL!QLXA}SMExKPP*c&ZOaDa
zb-IAyp?CGKN@*ty;SdZSM08grAWb=&E{$3#vl91_ICzPyl?Z>+I1-Qt_wU;3!RNJg
z%kEQyOWSeguAFgEq{Cjorw;M#6g?)4-k>N}U$^d4U$2I<$3yCU-U-4J%<%qqnKLJj
zjqj=_$b=jOOIK}W{bhz1Sr+^eO3#A}t)4qV$lOLh4LluoVx<&qH6GT6tOF%h##5|1
zQ*_^%y533XhI>toXIu6`F*#iUsUq`jbYR)r_wcK&hc$2J(1nyq!V<1yCSvN1Nl+D?
z&e<|tLE_dA%f*>FlbmivUSVvrwM7`wj#i(Du^fJD{1>6No;l1Y3Z8g$$DDBF*oogx
zCY9@z51>ovX&;Rm80@{w_Y<=*rQH!!f%sRm$yipaFJ<-h5%^o+$4)xk+RsQ~I&NrS
zI{v2o!MF0)Tq51RNO~Bu&ClU}@F8`>B*HET4Yxdc%LVMPJfcn2<`sWMT#Nsj`9rJ8
zj(e8otx@)4mimnc#JGzGbn6f<RXa_RlOVfXX5&GPKuEQtV#yxU=0q6VIyF8$kYM=M
zw{F=wnNv}%Tz%JvdBHnhP$wGp@bK|`+Ni5(!oK#4S>E`^@{CP|jEg3*KZ4!c#0V0^
zf@hLfl*Bd#LG!$bqs+4nVo-2gS2HEM<pr^$Ne#s;uUa9)7^b#iyL3t_c}c>k<5c!<
zSv``$wg^paAUUYUIYO1Hyc81`g9FXb;?zCO5Yfj?45TOwC&*v8%}lvQ)FmN!ES<RJ
zicO7~9zpriy*5Ta=SyIGv_d07vm5zcbzF{>?uu$=l^vG7!qGkn!jtzaSa4z;aRVdD
zf@pT-Pv5we^uM>XQq15X47JcN@6R?{tl{{CbZ+mA07H$|30LY}*8-ZpcKyqx->Yro
zmblt`zQG*V>w^y<a?czwxZpk%btfshC;__R{$SLIoQ2?G8AQjaj3;RHoYE2eX3%x+
zqc}YL_+Mw8j=ltg@O~;{PJ(~AyWLNqQR*m4K<8s2dM4KYQFb>8YNG|(?*9KM!u!8J
W#)$R{xljq&_RAOaj{moB@P7dEKw~lh

literal 0
HcmV?d00001

diff --git a/studio/src-tauri/icons/icon.png b/studio/src-tauri/icons/icon.png
new file mode 100644
index 0000000000000000000000000000000000000000..635900d8a7edc917157d5e1ac9552763950d128f
GIT binary patch
literal 46007
zcmdSAWmuHo7d1RH3=J}%2#C@kAyU#UNJ}?T0@5uVGlT*Pg3=8lN_R;OAl==i(&5lK
z%se;0|F`%1`&`%aMP<0W=bU}^z4zK{trMlGu1HGsfCvNvkt!?6YlA=#;GYl>0Uq$z
ziSOhU2vi-cEHA6;JHKa&kI>btDvcV9^Bh+C^raI4hTYvbA>C@+Nt+S0oc7Yw<O?Hb
zey=AFlkEa+QRu26fd3cAA}VmXqBBwNH*und{I~sY+ie(wk&7cNUup6x+L<goY0Q>*
z#R8RuL1kGewfzVQLKz4`p$Om)1_BDu?GFq}p#MH(0RQ*l|G&TdzxiGNpZ)#6wCTuM
zIWZovEUZ<@9Sy(H`+${RiDf#L9x4Bzg30>d!M%Pq`y+)Qdxvljmo0RYic-na{o#zJ
znH{EN^tiHD!>tGTQZSpDH2X2D-uNI&K>P3q#}(?s2PK04jz5x`;Hs;=F{fY@di}hQ
zgFP;zBark!Z(JZ>U7=1_MRlg_(~7~2L6WW@ef3bb?YD2+NJY&No+k+{@BFH`U9-`N
z)h2amvEoPPeMyTgDr{AQaB~j9m)ICqa;R(^2nN*xhrr7T&zg=r@*Rx7PxcRf`J}h2
zoNXMFUq^`M=A21r6--yk9PEEkG9g&4F-0N`JCY24z5cKC=vQ6s(Zr|IqrA(8w=-;W
zDNQDv>E#BZ+;UO)vM|a(y2+b0%W?NmB?|`Mytr5=iTa8*>rK(|N@K~N4fZ~w4R4sY
zweoT&M@s(ecv42?A<b)$5piF!T#tx-Fs#|w)8gi8?_6|^b0*`XIVWNmln^8f`%H30
zaKXNApgt0(pXk_ENV+ia?q$<<d8UlM+1eQNp6xY9aCnFdL8zTuk8Sllh`5H(#ZW|V
z*`%qrhJydZ^9aO5U`dQ%h&+nAggFtvlwV4#pIJa+k!+@GDPpD17F^BqKb@_0>6B@V
zft*cn3N6#<*f}_%qbi{Gg4EBf<#d5*C&f!@E~Y1`J0u?p2>*A+&v$p3``%N#+6I?d
zKWJaNeA(Df?v<wMIx&DUYYDkH+a}ML(D`bM^QCA4Y;Oi$rR4VSJ2FjQk1|Il;UyL+
z(FuZ~vJe?qxCvGHpSEnPt1D65%W;P4BEpN9b&tERMjRIH;b;3=%T(C&F;X-p62S^S
z_n~GW_I8THI%7;7Xbvr<ltHqZI5IH#If3iQT<V3eg7VhBoQOwn8!{%oRZj}OZoQ|`
zw`5KxrJobt8E8HV&5$~Obm_mFJx{m#7W7{({MOk}ET_vqYGGSYOhx*6d;Ki3F&F*e
zZYxlg%zfPz_(E)lYFDNr;-AIGF)joY-4K+iJ4P-`r~KdINziE=n`ut;A;_e@FR^6{
zDlSr6y!JZzb(2cvYou3oVz8)D)V-a{yZiaR1d1wgoj5k|8WzlKlj+rn|BJ`>`D64E
zIH6F+Nf1ol10MsKa}E!RsV?g49aijH^>_Z3+K0}ZlX~qUA`ncwy06Ge$-#jSy&l)J
z>x^v0exJnYg~H<y_vZxHeyfMA^<V19YhPgQ-ewJpWh}joJHhqpWQ2uu)o~^NH`xHa
z0dX9-Lka}x+NSaHB{%t2y~`4(<4jqiG#0HR<TT5WF6^2l8E~d#W|oB!o|~J~Zt`}v
zAN%_5?Y~8%2f{9UtdKKLTqf{^->^VDF@1>%XZ$qJ=@!@!_rHvYTjNFb3538Mm+2-}
z%5y`o-|+UoS>_ilXZd|l>@#0(Ly=syg5l@?mX<A$!8iU!wGN3TW-Z3>7W)adxjN@`
z3IrM*cq43NWTajF<|Qe5f1y5kxg|ig+^nTpK1Zqe9tW5t`~!F7o${qc7{aER=D%w3
zh4Fx@>0f&@em+J~ZB;rq*yN^G_~1vvN<q(Dlh4`S`ToKWFg-YUy+5(a?e}w|X5Y#u
zCUt7lB}U8v=l@7|c5@!M{fT<q&zk&-Ns&INh9X*qr7aEbv~Wru_tQHy4B>Nj2umc?
z;0(np864e}IaiRkxZ;keIWs)D@0)q0-SOZAY<(!SioN(7%kT5=)lFmxeQJ<B6LLHp
zafZr-e@bfW{mi@`EifEcQ&WSW1ffgPJ<fZ{6-XKwZIfNdzEQ_97=jrh!w0U8)nWtF
zw~w=DYCfi&KeX&$d7XLHpL!^c8+d=A-ZdGCxl$FlP8G22kERi@-X70S`<+aA;$^Ra
z2yT}Cde}n`@2#->^-U{J!K0I0F`<x9=0LHc+))VD@zZVms#KeUc_@xynd`sAC1g2T
zu<u?fqRm|hdi3eDqsi&%({YaR)283VXi-toDw~1(Mpf2Pftbt9qdD8`zN7Vl+C4II
zb0ed`0=k>J#n>4X68jX@eEjL{&ZM#lj!i9uNTN7`_Kn`@KaSNRhESkmM>a&%w!bJx
znw-8r3Or{VuU23l3M8Itek`~pdNtmC>b*aoPa^`sto6m$mLsit?yy2J{Z}qkXiX&$
zuh)+5&w~!)d~;-Q88oDeFS{D?zlB%k^XPd~tvSRY3x_O^P-i_#3Jygx1dmHILd*rW
zLLsC_)wUx^hktu^6;Cc@@l=Sf#y|=L!Y+&K?WO!||E0{I#X<bnQr0(1FQ>8A$$!^*
ztBHKzed4u?A@nU?(c3{6za!H2^A=-sK1!Dt<e;E|kOPm8&y#KlZ}G<;j)D4iAw68@
zt^xlu8h2AxfobpRzAKb7yyjCwVozs}(es}2&7t%kbC_iF;KER;lk1k12wQYnK`5zL
zqYvuo%XA?xXG$jn$}oxidx^T>Zz4_5|N4^ri<pDo%>02({X)nJUn8<m11^E#nSA4*
zOM~=avv){SgiaQaTQ?b^o@+PfE6GOHZ~D-e$1R~!wxi{}ZRX9D6-qtWVTofk`<b%L
zCJeIuja7JpRX50orYd2>$UI}iop_LkbJX$E8KKly;~Z%&SLOfC$*$z@J4NZLYl*vY
zMQQIO499WWntx~vv5gNzUcZ0lgt|F?UTzC_=nbJqVs?tlYbK?g=bw?HmrLgEUO3N5
z>+|{^ScqP(#kZ_vH*DQxJJ!((4HZ2-_Hl~q636l{K+b7Q)`GO>oQO)C%WTVjZkTwS
zb29d-s0vdiy=5hOS96Dr-~pIi{^CWTLx;)3i<kEF`YlV2&l19hh*g4{VQsK)nj1bu
z^HVOuy&g!{)rcI|{sfLI;~G1y!VPgg_ZEp20jH3jXd335>vNyX8>-7KtR+dfV7sc1
z%jAcC#@aN8cdHAfj@HX(r$_OZO{^QJvwLeaOi6u3|Mh_IgaB)eu&AW-k7}bM0_{l6
z9qTs(K??0~_v(LNoI_ir8ScJ-jaRn9TI(iw@r^Lwl;hW0&X*^&S`F`@>nK7{^a9ve
z$+MJKfrnjY&sV)WE88a2-F3LzYt8BcsuyeAiUM2R^D##687+*ybAmL$YUZ<is-X<w
z+@Mf)XWbDz;*!-Db+OT{NrYF5Gj{e>BWxb8TEf0X^j$gEHLjHTAmzEIn4t+U++c#Q
zny2uSF-3mwJ;TeZ-Eu3b1Mx8FdsyM6xACvs+Hqy^;VW(C=wQR*)jy6O>FJ?dQvx<)
zl@6w<v*%cA^nBNMn+mP_sdep;2`QO$3ypRuozKe2oZB>A9}^xzV{YL#Q(tvpc=BtL
zWyv(juRj)N64ErD5&QX1JF%D+&m{Y&;+Oq-c~~&@3!Hu`HyGi8tsar$y33|EI)^bN
zL#w;L_V;gm6|i5c0^;ZA550rIbSBx~@z6upeuK7mT!)<Rg??XZhS!bN+zA$9S?zJ&
zwLW@pbBn7K%C!pG=@V(y-iQ=Se8Yj7fS{k{jaol#r-DApXpZ=fW=|J{ou!P8xSH1v
zgIVoPw|F)^vZ}hTrb|s8ySAJrNt{mVG-i&>X~b^t?(`m&H|}uy0dI1?!&K4w9gM80
z;?UkloS!@#1AK`Ysp;4_`nvMO^`<&@z6-CB`#O($g>`6gwTB`_KK2MGLvQhUT&GEE
zd=cx7X?K|P(!;YkHY<q&73cVE+OFWCZn5tXE7!yq92`Paq@JTkP^Cch_K$4q&Pej3
zc6|D+*IB;3S3!jRqq%Z6lLcDZc?vQ7R9N{i61$1E3C;snXj6=Z+tl*@Q42Xmuq`JJ
zeTg}f<JrH2v^O0x2IM*dG5}7Mf9<p+cd+~><KlsS5$EIdF^TByQz2R8i}eTEtlT<*
zUAD&5>dH|mv@HSx-S1o7|8{TBzk0GAm0YqyWqA2<b)~`G%Gu<C7&BIccfkyOk<|)!
zSIu2J4>@a7Ee=l55PFg<eZD_=24W-X%%fr;$i)My+dwE4>+ej3!3a5YN@J<D^^f-P
zu9M@g*gwr8yz)X;r;ql-is;uptZ$lpP@mFER)nt=59lTRYszl)#uJk^u8ifw9)1&f
z_@K0+qL=wer%wnG%nWu%{Z+mEhLdS2Hx7EBNP!1Yr)By1ru@Hp$|Ce4=(WO`#d!RW
zxge|<_gY#~(94Y#btfa~uJm<9??W|vS(k-6=lNYY(2Tskn+HK8h^`tKjj~o(DlFw_
zcr6gtHa2}}f{x#`^Hn%8Vbn1BHg&<-m+ywn4es$Nxz~7QSfwu&36E0o+Wg`<8g%uv
zDYL)3*Sq3K62U)cXPKtxD8nREfhz;!vl<VR^K-RqH9iy?Ds#E_+cYNc71g$3sTpCd
zuQ60$NT!`jSlHmZM)p&l=7Z4S&W%>zfvX^JrTOv3P`>}^&Q$;u9u3wndu3FlTlnXO
zz#JjwwX@X&{6;wpkM!jH!iRw@Uyl@RnyjQp{Dj=i0ndW}<lp?eqqA)3ft?od=wKGT
zQ)YcN2QI?@nw^&Z36*a@x*~kvvTlFLW7Lv!k4v=9paAQvl7J(HSeF1GLGcsA3K{1>
z+u;ljBo-48_4Cpi6e`h+V1*_<oGdZYDJJlhl}6;tavRr-#$JfSjYE|T2>0Yes>Q-9
zJ5XwZ7Hc2o_+5Maz38A?2s4KGMDT5wV{xdy_WM4FY+^95Z7g@5G8pEY3*P0dRhyIk
zMSJp#d!ss}=fg!H4%yWOmqA$qC71q|Vl(FAFPRZCxCQTKC|#JoqT(^*ozEcFpnB=Y
z5UC=F?^LOY0ikTMerZhyBG!4nM(eBQ{7l7)TmwM}CE0zySC7C@3}sV&@3+$d>#OgH
zRSqNGMpofrsf86G!zT_Z|F!Oplzqq4m4|Ixe`f?eualoi8xA`<^rQ85FRiaRENSOG
z@pxC9?#<l*0Or}KTswAv2f2lo3YEJ0t@-O(Y@Jnml`M(n;<{yuB0-N3Ok+1)lB$nu
zT<65$yilk2e6-Qakrj9JQDkhPHp6QQ2^w}f(vzuZbgOOplcic^g(uk@)%7JwVl;F9
z*fAZ%EynS{3x3!@a3Ma34t}GO_TcoD>-4U-P|MBq8`0?frTe1th|m2;W7Uc|ipumM
zZ~(MDI63TCxvcIcWxcCDf4-oBhe9>}M#&OXyr#C!k&C$FF)EAC!Pv$nATZ(T#qiL&
z&4|*iE>Sb4(X;QZpVyQ|<CCIBJ}Q;-nu3e6yDT6m`SaD4H`8s?y#*{-rzMR~!cu7J
zzr}kT#|v%rcPz6>@3k69Rry(85F7zXIziQa4#2MUdXj%6`Xm>h8SyYhxOP-;cc))A
zdhc^)iu;`D6Ykl4trmj~>xY!P!4?RmpY4ZeA%nQYk31%Ac;y5k&#0JyHt`&Cdf~3g
zbnq*6|G~tHbhVb^-<}nj4(GmIo~>UW;ZsF=4^AqhQ0cC%pIwnR`l?sMh<~&a)O%cI
znX3W&d|HTA_*QO2KV!thuix5;h%+bfl1RdT@M}ttTcDI@Ns&mG%4R1{2z-4nDLUb`
zC!@oEC7GRdH{z@G0{a#Dg@$klvgz>w)vQ#Y=m0t&_omiy#$i;kq9=0MXC>&z)1Z9Z
zAUyO11JpM=mvL+2N9;q-LB8>NS94aJ0ul>rzD4V{<sWsNrvqg}RijRi1Am{RV(bot
z3vHB~KSZoi+zwG8<|-dt{w~;D6_^$a#`QFkB<1bYJjvZ4Q>;UoC@k05$G^Fz9=cwO
z*I0|74<5}y-9zzh_3$tr1R*Wd!zFQB0?w2EeiB#zBypsa$=fy=!W!Os9+i+Mxay=~
z>9Agw-;MJpmj2}YfFeigHknvbgMf{l-`}7)eQP6{+b(-#xcRrn^}#28B=T3_=5VH1
zlCb0SlT-=6s=-WggP!BF{RJvs1nkiCky~r_0V{%ooD=dR!{ful;EUhJlm+^w#;NBT
zeu5A?;>C&sojJRC$Y_=Ix9|SNc(>(+MWV#}0gdz(`th_Y*SVL7!UJ2q<2N_!m0PEh
z4V9*3sgnLzcOR=K&<MP-)1Xgk9z0RT)1<T;{}xq8pf0}uO!Q(k>~utvkARAt{XY19
z{PL-2)7n!3zJosV0li)AsSEs7orNOs?U+s2lqB2}t-LSLG>a`)rlMK4k9)>%a^+MP
z9>zQQ6p50J(U#gc`k|J}L?3*S`^2o-5ST}EB;j~LU%)HpbZH>Nd-bj>X2lS>_atNY
zWksS86QBGu86iGTFwd8bsx`4CU+jt798#`8K=D#)BMo|UQ%V!#@Vfa*(OXRaMeuTB
zF~>C*&N?ZY;9zT0BZXIX--oXW>9wo>;0R_TMkCfk{9A><6}$x<7TbL?MQS)^;TMJQ
z#yq-<V@E7;zHojt@b6_6hfleEh>b@N=r(xj)7b{P4B{^jkU93kG;@9xVwabn)Dwjd
zetkIq$kRH?FRhE_-K`9jIy;Th>;-k>X_2>yFcv|2=mJXq@evS{SKz(KTFGN_i?t_h
zZuQUFu^5*ij<_vyalRSbiuy=F*YZgrFt+^;&Sa~BL^XPw@$52)%&g#2`f7_=??`U)
ze8CmTxA3^bj1oc0*NiIz8KrB}|K&yweoPoV1OwSYEdogIGWz4gSU~cN+<SBOdjXz9
z7ZZra(ly43UFimeJP(_jR($TOYXWpO9P=~jN*0=Nzqzl<J~r~;W>1dkH<ufLH3xEJ
z!`8mVt80hl0?)oi2i6@Lh&ulhH-6MsBg5H(13)Dvx>xAccUe~Q!HQ`{nCkSmUk1|8
z$C1Q>yXS(a5J}XJ?BF6Epf#QW%l&FfDZ>Q{h5+MgGYOr9A?f>mM^4?f)X;X%qeaw8
zs&`vpwVPe>+`G@J&bNW?+%n9*l(t6eO>ipHRwwx=g?W1@Het+OAa!mYI;@(v_}89Y
z<Zfk)c@_D1XWs;UWr}%U5d&jcJ(<u*2IJsuef@fLH-st3TT|A=ahg=RCNQb_wbMJx
z#~JO)x2-M`mqACiJ2rSde<+~;CH^C}7%si|lA5=!<<i`&zCqG=G-xldSGzpihq1K6
z>O%n@9v;*eQ7Oz2UaC-`RrK*lGUn#mvAFHrED5HF2ad!+nTg4H(|jd$1_A68L`La_
z*rl_e0$VA-D=LbSCh0t=<8k+W_h-S?p#FpH&C^M=YxCjzXnJWgJq!t2zT|F5kzVo6
z*6Ir?({F#nrcFovT=AQN4JEki=1lYlI96Iq^TLg~g&DFTvbVWHv~`XW3hHAdogc`1
zQ&L`-y3g}AGDo|g@j&i9(z=g=M%=rk`Enz@%KfhbUNF<YNST|f)AbHwo`hVvh@B6w
zF9^ap?{y;kg%hC(So5pDiZpc>o`RDd4Hk?(M7Nn!gOIcP$oJ$jon||rbpggPVwim_
z(^;Oo2j9%}u^vp}Ys@6`L<1{trT`~k7HaI74a|dT^>O>f`79P7E3PJ7!rWK7txJsx
zT_G7149{ha*NJ0qgn4j&i(RBMAiD-neM<B-mbbl<tkcY_{X&tH%DIbV1Is3^Eqcqr
z*Uq(jweXQPjZQA|ww5M<mYAxrn0i}kT&v5HlA5}Gb8Vpnf=GUcYPyI0z>^QPBhFC(
znSvfe3G_jck*DZ-8ewGF)b7A+n1%rbtp5+)%AW0Vv)|FIRcv9^is-<|XDOz!c?b;e
zIlJ#KQemZ;GE-I9qD3U)_-A_Sw(w7<<%p_w5kmcQr*W3sL|2o-bfw}8CChGO4U_-s
zBGz^-?w9nEIr^FhU)uQHoX*s4x~hH$qQMkEuxj-~J#(#U;AuP(dxO$$@js1|5ACAG
z`<OuRWVTWkgPCA4YB=VpwydTwAi3)VhEdRon{!{I;)!EovBXOn(;o1GRHWpp#j59m
zbA6v<c?%oB)}7MD0gwOM;fJ-V@fyw$t+Hr`*e^CzrV-#SP_Ua-=s8tP^_Gm&&6lJb
zvDGL7sM~oFM<B7`*#5R{6MA*iODgzYu8eMI(kz|IS0?4RL*cejo<nJZM_YNZ{gLF%
zllAoJ*G|(Kus&;^X#3=R)#Uyeb9kvsQUQ@}*(=elB*YZ1bz0G)ayNpQR@gSB4)hFc
z?_1~llMo2dENSq9+N!G^|HmIMSCL~d5d5QRXMQf(o$6R=jN!kQb2GowDFZ!fn;|V~
zbd+@|-YQL7%5I5K^;>7JU++o1w&SWzVbP?*nCte_g_I96dZ|{OAx$x|2COx^O1HJ<
z`Kz;QpHv$7Ll)Olu?gR+dUiTK6=nBF=uDwj&6C&%f-3)z*kHl4O8BsiBs5CqusSX~
zo981H&$KdFg6fgj+qk_wCzQaS_XKvt?bS}jK11F7qfdTcS^p(66^Kr{O&tnQN8&6h
zO+RRTDoi8Zm}X9P;Zm#`12At6)6)JYb^uvP`yv=hp7Vg$s++{9-X*ifVM>Jy(r$<I
zmh8tR>S$14VOAHVE~H?}<?Z^JwMoct|L5(nb3lt`Ji9j74dx$jxQ^=aWl5od)1m=J
zl<RhYC7~5kQ$~0`O=E~0F9wfxp@9l^1Q0!{`}4K=06zValyqNLm0>84xF9o?dsA?m
zb72UBC>O#$oqJQud)sF5@vw_Q3VSmn7gzMVu<Lv%>&Y+hd#@JQ<r092j4?ktvKd0l
z@nm>E?k8VV13++FJ;VYU+V!XH_0VwT>3lC;9)3z|6w56bdDt%9rE?ZBin3HZ@L53l
zB%y6bvbcp;0;~Nm^2+<#G+?=R!MfwE=p$9usZrc#cB2dhNeG>XA_ZoSeH^{X_Ou3^
zgzVEKw>3l?AMSM0)Ca{zh>;SBUTkh`6um>8&X||yt0tidnF75#KHlfn_E;*=VqW5g
zsj135CQ>Vikg`B9f_<5gvo+e;ha{f}zkm+-m9!tA0hFd1<RZG|MzcJ27oE?}o-Z9L
z-5JDwiQSaOY8_IjVxZ)~RI)V_-(RMGA9?5$UR_lcN`ZghPpM7I%S2`5L5%grhF;@j
z@fUJ@kLL&gFnnDl3!{&|S{nPObg2;mz79hYmm;glF0WBY$3E=O&8!IzRvc+r`C$n=
z;IMXRDggR=d#?HiFxmgAeVUU3jVvQ`F|REnc{YWJ#Tu(q$N&Ig;yHa}$c{k~6VIuL
zUYXt8!Y<c8p^JrRsCw{>SwN{;mc-BSSZP@p#UTfHmcR$#*h|Bgb|59I!Xqpo(io%B
zoJIyP*W`1Nk1V%q+#7^vlY<3IbYHTa-AHe{k_7W|egjKJu9B$%aK@$iP=OV^cYas?
zZCA)egU7nc(&Nz@dma0g)*#J4G8g`BD|xc4s*t3CH*yd|Lg0KU_%2ir$K(39Y52F}
zj=!0l;-%q!qeUJhFx<%s4Tbn^=O?fIYeFRfh3(x<aA@3zF<issmlsYlWTPzsKFS&z
zWDd_kwveKbE|yg)Imqmt+h<PyMllL(2zNdq3DKcWdZz?a!3JHQ_1)EwU?h{e$Z<Ka
z>dR)}Y-Ozg*j`ii@Xcz)SUNM}-xkVkGI9{9P`0)`P&)?n1q2ilaMIN(-*OsK!^L&M
zoD%y^w&8kil`J27@cWm|TJL8wRdfH#yYfzPAOg2aMT_gR{qa&0VM=^FRmkWa5aER{
zjxvZ4#CT|Sq;dG<w2y}~E}cwF?caI94J&Jx-rNcv7T{s~B+nO_H6%|H%6M%K9!Ji;
zVuzC5sR57_AWq+XKv<x7dyt*yDcOWqO~=P;{rW-HYUes@d)=8W4shY$D|qn1)0N=s
z{i8o0d6HlZ<WTZ-K+1w+jDXLCGhqmK&5AUTDnu%TcQg+1r0gymnBkA=w)xD-iGr78
zy9oLbn-@P<0R4(ZsC08L-(PGut*hp?28=@byDV%gl;G{&tQNp)Y|L}DDMRpl=fUYl
zw<q@}Z!dO`9t`hi+#DEtLod>e{!nt9-N1AGE8wA3@Zav3yzy*?KXK3!<64zq^r0dv
znA@c8$>DjKvP_Hq(dYU{aSB9nZhT=%>_4Teu=rwk;yz976fV%qBkX0buJQseM&KBo
z?pS)1<P&N%;qyP=A6X1wcu5+sc8YDTPPSzw2e#?55yMkE1L0DeDt}8HzE}<4trz+Y
zj7*GUUKb1Y-od}-$gYu<c+V!*Q?dmve$7+0V}FrJTLUqM!th^zhCLLw9ZVsAM%dQi
zmb`Pio=Ep&8#y)<`ux9mPycc;OXb-1FgnhX#hTEfwfLJi(j2V;&X)uO{S2zm0jVZE
z`CMGq$Z1Cr_q!1K(5xwS@lAK+Z`LnqOy592y(s~y1u2?+dRw4xV60!Zz5f^YE*gtI
z))=CY!dzJ4!9$hW2t#ZJk{*rTiDPJ%cX12||L~zwO5$B9D3>5EGHM*_QoV>T*(e;4
z#W+;uXqP%n3a)6UPw#@EAHPiNT-AUPHotTOBE3U7ny*SliPv~Yc#uJ0d_`~FjpOsc
z^|O(QzH*&0-rW)qj8f<b-}266BZ@Zn4*G^Tlmx?j88Q$5(;0<da?MKcoTBIv!Z!Hu
zPnQRP9tQU2s!h}V884{(hu<kiXI>b^h@}d4T)z_kK*0hE51&~<8pbgnDU3{fOukMr
zGc|^_)7RzJFM51LpChsB8Pc*W<b)~G@K*G$Ira-MOSKCU%DmRHP?p)W%6FccuPW!w
zjt2t~dSQ*->LL|b#zagze1D!vLglnb2Ws-(|I+lYQPnJ9U+^|a&nXc4Z7+qFzEOhf
zSr~@?l8=t+KB7kjX6*6p_uM!`o)-W!E74cvzQT!*Uh(pgMNddsOmK$D&20zvJL-FE
zk5YwwX=V<kzBUD5`a}P%T<v0gZi6&XhG2nIU8%9+*@KnO+q4EMbw&cicf$cN{`{ba
zT|GN3W_@XS({7Z<W9+=1OoGm><m3_G>+q*!mIsdslpSVd)bO2_n~m;BqumPiOVgT`
zz**rq;y87wFY~H}bnYnusQcRhhW)4P$@z+zt+M3bbMu@atggEN9<=$WpUVT)a_&Hk
zXwL@tZcqxYWj?QxjK03U3`NcZmJ#bV_fxVji2-%}^$XKVQ#EQ&?=XQGI(V*B$7)+^
z`Kup&z_zuoT;|Aa!vR#b-KYsfieYaK(Z3psOU%N}{qm0Z9xqTHePkip%{qz^GcHf#
zXRxo?aL!Z|95^{ZFA(y+^x9*4*fI66eA^fWx-NK;<73`v^rGj8<UC4_>vmlP?z#lI
zq@hpk_GT-~-nHxBGn{$C5HRmV-^I`W0b&G%9UX+_GCBK9yYR3m748rFfL$L*-d#ii
zZGO{uUlk^QM1vawwRi@jg#z5^u73anC)l}YQae7!K0P>|b364Fy&6O&{MWW?vR{F*
zK_R#6xmdMtyT%9Tt6s}6EWoJ?UT9>BCE5U+gAl#**_&m&rH}d$-7&OHp2R1Skl*(j
zIDBFUpM7oX&mgfNMzA9%%pd6hsk1uqt;lj}-t(@{2kheFD@t?=7IWR40+o$iNMdXT
zuaYD9WQFzQxQL*mL=DY0Jf(Qw#yH}DQ6??}LnN^w<x1NHcNI7@Snb#z;aH>*bVvZ?
z&i4w^P+4dU4-&0?^IY@A+dNtouUTB_iBoyeLFxo&;ZoJxrtv(K`{oB{SnHg~jU}aR
zFzf>y^Z`D2o%cg=j2>vT-`HWkyQTGOKtqqu$arX=EMG~NGipLu;Hr@5`I1uLU6qUj
zh~T`Pe(<sr&v{7h?dwr>IQ8bw-|uO4(f@nIA76&Sj$q8Ct^LIJ7(f@FOH|e@vgHQ_
zv(DMm9JOEN&$oV)262JHKl;IW?Id{5Ys4ExDgu%)4TKoO^Qcy7c?s{<WCj8zT`h*;
zgzJAz8XLozyJER*2^1YSiQEALxAu;t<`ySh0M5CD!+i}a)_y3)yw4g(28^v(#Y(c#
z2OfZVSb*IOQ+ABv0KmE1OLc4qTu~04LP*p>0F2L3eS-XT;ec(qZ1?!c2uf0J!kxa7
zIvy0bcLQ`N@0V=}96D=&SfOvK#mgqUUp$>?vOveUaO^qgrvL#OUc&sjbcwzv&BRT@
zAN-LQBZHsz2U7(U-;>d+$HJN3ynJ(@VhM^=g;*RO4e%86@}_u_HWa;ZrO~X<FXOcZ
zOcIE=FF{xdt6ZaUZ5zL;fx|1?QE-Fc_2u!1^!jH-y3e<eqCGp#kel_d&4L~a<ggIG
zZCYj)SB`9TqlAOb_USU!L4)W5FC`+AE%00!L{J%;dN2NFvv=S5S{3LQrZMO_e10At
z4cnXBa!jn;S(5h9WaE?Q1~qEfF91t%_ygat!7Z2VAp=f(4Wv=GxV&`8xdcMc10qEI
zsvLr1nRn}W=dgrbe}g&xedE7C0o|^=2wrkzeWua*r)W+#46pB2*Ge|^P6dFijF9s@
zi%Ff6)=}D@pDR%p9RQ=rmdL61h53_62N%&>E2>3?;ty(u^^;G81`^-8kg|Z4{OhHU
z2_}1czX;!vGJOEZ;z<LGksiF+A4%a)Zthl6qur%n=aP^V0@^@+z(&By7UL~3q+6){
z;+939YYl2DUX@i++ZL5swzx~;3i5?QjUOcZ;Ug$Y{LIZ=<=ER$`|+E$gBT%*5W#Yq
zBpvzseL7!i{DYbaext^Z<I;5B0VhW<FwI$agcFir+`kY$HaRFat}T){8I|+s=lHRO
zI_zlms{vn~?*8CO_)M;3A@a)s(V0jGy5D|l3TY3+SFLrJf>4OE2vL*n>6T`djLflm
z)}cA{qP1Y8p-P}H{MU+YJzulQGx6*x+ng)f;KzVE+z&MRJ?=b;%10OR`;yNQ)cn?w
zE?=SKZfEGj-RNi%`{m}JC#V(k2SQE-o*T2zlcQ%iv%p|}r6nLWIJUR9LyASSUVg{f
zsry1E{3PD%+FK?6nLd!$UwE%q-fX<4-x85Au)E)I5A*x?73LbXzJkTn15oFK6qYXw
z7YEpN0&sLD*UP@F9*Amblch#4uP3!8^)6<IDM&}17JskprKWl_W|84i!^+oKe<O;J
zl*fL!xVT5V<wzA^{;yrSTS2+%F4kpyEcB-9GGaHG@4~?P;UCrl>U!i7uHX0=wKclb
zhzynelMuuE8<DdEP+8wb;3W7m&E|rR(+RiJKL6V-O?yCG*jxOmb1tAnk<u-5x&rXQ
z+Bl&%0*lB3JQ8YN89;X^6xg_xlEj9`Mpgd2_Bv?v+I`CjvMyJSXTyg{!Iwb{V5PuM
z1ryJu<_|tP;OtQ_hFT#A$!1?KPyJ&rbn>@wd4u6z4m^Ije|`PnEv)i0`JCCgWyGWJ
z@}zekz*?sYbrh@r$Oo@DX)4BIGdzdKvV|?oXFZ2t&y4F`xPgH?KH1D)87mnXa&umS
z*D3U7D2<h;_2R1w>05)uC&AL?ym+!_BrHlG7SV@TkG0qFm)T1V@gH6<CSk@Nzwm~x
zKU2k4QeHR-;u&GRW=wx14FarCR#L;Elx#8h_Zz13hW(Lu4)_Fd3!?CV8@EheV{i0f
z&p=+R6eqLdC*9l}R=sDiFDa!LGM$8f&VZ4yiRq`k7^Nml&y*`OmEf++lTpvmD#^#;
z+z8(%Hz>h;?63mvwPA9*I@7}gFitfcAU&cFevvI6QOSWvTB&@q??|3#{zM-Q2nKYe
zq;r3s&v|zi!{Vz<9m)%J)|Gh93cHYCAcw)8vxvAXiU>9uH@@8lC*EiBRsyj>vix^A
zwnuZrh{SO;aPPq%E?eddE}|pQRgTrV?ei9`<s~4NjV__T0q;R1e9UmYasHSa*IZ00
zZQ#}R4~MH?ay)hYHw5}I4+Osym$w?L<|M!Jh{A)L{}8uU0`yY#St5iE+GkQrQQ6zu
zyH~ncpiS-}0pITet-Q4lTxFV>AntcoRP=Ck<*xHLeOP3Ubn{<^Q(TJ!TU-XqYHa4^
zPlwY0xpcXf`)w{Mg7QL!$Ha-x-oY!$w+`qSdKu+fPR3PljYbEbWG{&WHC3&&biQgx
zG8`+8DzN5A;7P0Vu?CY0%fdctNL|V=`)xkj7X`dkR|srU!7L|8{|n#)JHAy&TF4fO
zAhG4}xQx-9(rEK0Lx(rI<7WVw8K+66#42!gl*8dO^-KU8(v8Vzsjal?**PMm^Qg@I
z*Ap@gz~M*xa((pc>YD)@WFpjeQ$dCda*8ilC}tjgl}RCLSB6pYl_0x4!OV<stfUqn
ziv<RKsWQGwLGV(Mfyr*A5HMf%HV6_+hT65lSwLSzI*{By1J+8qGuP#Rd4##6O)4xx
zh_Iv5Z_dmsL+BBHLxhhra#-PmQsO<6{bFks#qD(<io*wU?yc&XV%xd&%I3JbX198U
zSI#w1pVnf8m*ZSpFBO9LH0>G53w+xONTw>|HKas=SM3<R)HM{)OA8-PW)&CKCYFK3
ziLgHdBaQu8Y36Iv>(AG_9XtU*Rm4_Lb0-Ymg&S;r$u5iiOOG9;@u6K?O!`TV*1}b?
zo-8vp5%b<FD2-FHdla}5-=E0EMf2+#Vjc}JX{LaHTt)|h4SgZdeMJc(&XyS*C2E5d
zRAt>Q#x1ZM?uFM&D?GjU9n@;h0*d;3Q9oym-4xpXDp_Xn5c_5%Yo8J`24ih(aSA9U
zpN#91LA)QbJ&^FhyH_uJhZ!Lx9`l{p@)?p&fle0GzknrsN3hy*E7eRT-qFDU^c>wk
zPKrFrcwfmP!IlSa*AopI3n1SdY))7_6|!u)^Rx(aL}7$7i@%CqjW~|$Ee_ty?a|XM
zXh`4GT;&N|DDo`S<Ldz#ZoS$Tva{yX%NXI?GapKdzTc~qCkxtZdHv2{tmHvE4)|{X
zrCxsf3GU*Sg>t}Q7nNfGNDORfn+N%ICiAu!bkK^p52Abk`-QLy$*MkVoDNzKJ5{Ad
zD}z}61z=5%&x*#nqS9`t{H{$yZrxz1d{%foukYaygtB$dj@EZRkuGu4YJH-yO4UPc
z@hz;96r=wG;Wg&DF&OnAB>aNxDT4Aw>Nzur0wGunBYA36CEqjgT`l#G!DAVIm?#B^
z0*}INCFndVU6~`}^HQ;J!UU~%LS+dg4%8WNW<E7|ZN5-cPCG)6>Lt1JgaEtuU7Ak)
zGuf>v(ZG_Xdx+s*u|c_+Aq08c60>$~hq{|mApr!LGcMqIzux9|$2;<JrnwPfdS7uY
zJy&AJeQx6|;+8F9o@UL=9T})h^ekQE9g9EE=MJWi2F&j3Mmw6&Gu@rx0cZ$t<jOLZ
zZQ>fMIAGRqw`7gge;I>AWTVQ2E(;pQNL^hTT79sH0)@22@>h9|DN9ff&1Mrb`;tkX
z#AW#J6kcs!_`%ScgpnsR-Tme;P(TakQKKl}(+HLu2sx+bfikPjAfSO~bGB7>V@yc-
zSZjtrs}Oyd-ODVMCh0Q+@Pl4^QgJ)VoKJ<2wRnO!kWz;8b=daE%|5>Go%=p>(i>^!
ze(RJj&p*nH*3VGnIc%viU7YVs4K#I6(p)TE&Gx7qC}JOReI-ZlgFiY+7_<5jfMh|h
zZ|K^}7it}U@a#Pl*)6GFYeKb7XrBVWQm^IT$mckAdYKyD+z_Jl2IIRd-MppQYW#YL
zSJfh<eRo%IDG(PT3HiPYE7B&9m}&*8O^0qwCbDCtu9PH4XwZNm;WOd+_JpF(h}>j^
z`GV65%|j;23+6+={gJ54YZI5BFH{$Y2Z@b-W*SH0f$bI>+z0j5(0aITK@~LwJR+iB
zB@g!NSBidsx{3Dx;lZbvnHo6TK2UP4)y|krM5W1W$aLu;o>vRxw=ORIQ~DXy_pHz%
ze6snB1>_Dx>sV(KVq&U_`e@!{ypVV;BRnVFjma@)3q@UvB_GX(eFY|kBB^dMv%!B>
zMjC+lLe_hhr?t~#_ew+=5zp`%3%b9#>NuO-hqtbD7jzx8y4&C`A3fy_`%ZNb+Qh6f
zT|ltvmm%z$^~AijIovE<p5a;G<$Cf_7nOOh^5e|4q7Ovq7C#L?k9FZHR@ZFn?<t4c
z(_kggYjFmeu6ivRbK7t@Zyh3C&%sD2CF?Uc0ftQ@t3)kz4teb~%(=VoCI3CvOr*r0
z{NS(yl{q6mKG!PEpc>mbBPR`vC#W+Q`nP3K-RxF1#vnUki%=#r6!T`kB}}0ib5@Vs
z9u5Jvw|!qd(z7x)j5~C-7}TK8Yxg|hIK}Fd7V13^IuY^Dm*%#dg+4r)PZP&~^R(|O
zLtww=;GYOFq?~$*T*hbKQ`hxvz4U{s1n0XxT>=50#s#W^1`&D9WVL+3D2+lhtq!S&
zzP+QYYWvVvI8csggWtAZ!x$4tl(u6{DF1SVT`lcbS-dA9&+9(S0MPt<kdwrZloG^m
zm<be^Xmg4}IYHiH4a(m^8TonebVrh@GnF@-#Q;%MnQ=o!**jHwrL<l{Vlamk9BeBX
z1Zws6IqG#YaR^?hKy-61{Ns_-{n(rzyChYVk4`e%&XZD8Y3f#Ev$4bCC|&L_>8X~M
zdM9(KahEUD>CisqPChMkt6s6b15Qj4XddEg#`r)!zyg3;PUH<GkX2*L&PzaOnZBQx
z&IuaU&L}*8lF68;4Vb9Gky^j$CyQcUqOe6tL81Ma_Q=nsZYKqMVbm^xDKm<5v_>?O
z*ML3R4CMz8k=6%65I8&!RedyFZi(+gqB9vupOj+;@fE<uA}=6rJ&+XG^=r$zKpvz=
zshi3xd28w|B|wl^_~Sb^1-0N_NYMM1uNupe!PF>0wmT`(!3m_IO)O!?z$6{@a}23F
z7oW}CGwDpJdq%b{t%BXU5xC}&JjhFix8qXTZ6H!zY=!nEtgXDf>17uhiINOXeCBMD
z3%5pMR_}aO6v0vW0+Lw)LE^T~^N!7MwmWOzS+Vr87x^5t){A8Ub(CI4K0cE~gbKIi
z;?KuMBzVLlhZ7VkIXp?%t$}M5lE^<km;kc6Lx|{1+BFwsHmsaOn&d=Z0Ps2%KRjC|
zA}7>_iNKea987+YoY*DT9mPP%$0UWfd$5JR%+W?}W(7RETl*DRg&{QtRdl|qlGH{~
zLL(_H!4J9gz`8y;(joOAZ$*oY13(;^QieF4K>KW8<goBMM15A(J%-$k_@mM`1PH?B
zPR4}_vQS_hpElti_JF}r%gb0L_5o?F$hy#VgvJ>V*e_DoJ|7BTES8gD+^@G#;ynYe
zE1OYl7C921oV^Ml7GpZ>VEj+-m=&TmEzOPER|pB9V?v(^>V@O_SgjKKmGMSDb)4cU
z|GN2D2-ZDchGT<_n!HQTFV){LS&Xy-cB)-@(6c8E**KSwRfp5EIm4@z8LSXCx9@%w
zh)_oD1ar)&Ccq3QdrM6{`S97GLe2-9X(U!gzM^rbsCaU0_!FwjWRz!rG>niam<<LJ
zmTeO~MT`{!GpAa$pK$vMMjS@g^p#50jfp%^6O0%}a2MY&f43uBeWy%XLs${LRa(D1
zPTBPe#cgH|GUU(er`j5<fy3buV2?#%SG$ji?9#vmu72zRN}vw;sAxLTOprE_w~`>?
z=C2Hb?5AW&{1M(s!S~OSxszn|6}x3k>NxqwGBLK6sW;1vy>f4;d&!nAU7Zf#z;rOR
z(B`~>f;N*g{ro;N2T|I{dDFt`<kl1LPUt?H9&hF0M!IYNknl=LCj<Gv-J=g+EnI40
zmy9~)3=Czy9(eGeTpx@#?aaLp&8Nu_CKw83;G_gqi~$lkCH?GEGMPcjg=IKD9b|a#
zKFtmwP2_%b;{D}{mH2!|DWJjBe>-0VvS*lp%pMU{!3&ww(U8nfwXSGi33h=2P*#ys
z=5YXS&J%v11<U8y*|yjJUQagH$8DYnWH%}fzO}Eg$(H?lIbtq7?^jU+aGV}Q$9gPE
zaUk7{eOKAmnnLY`60~He9&dd+X^ZR$CZ5mHqv~ctL)x@EnkMfiI1>HGnMR`mpKha%
z9GZhKqr9vP-5ZH^M5OxHJXeUND>b=!-iO|wfr8uay_;ZX&qySN3U7fT+vGv0-8e?M
zwN%@zvGtdu_2Ea$Y(%TX2bY`Ks_N?MZH@>|&{@&5k+R$XeeaD;yu}KYIFS6?-4(N6
z&zn-G66fr<({6q=JP8W9{I}F3Y_ZKn9wvQ9f&lK)a;h&$@R`O4Fd|D4ILo}hzj`Z)
zi3a=P%w|8L2Vj2z^VJP7&Q0no`9Z>4>T|SXCkyCWZJP|E+!Ub<_nXg`18!ZgJ6x;x
z2vjJdz=xiBq!z`X5aZm!th;rJF$&rX)odyEWY{4k)Q8#Ihc?!Ig5BrRDYW$YSnB~y
zrld3V1;fQJy`iTS%1I=>bp9eI*+FMUG}6J%WqDm0Qjlj=LDrG_Kh4dwt;?@v@gIFN
z_XH#yiYN{M{;isrUl35t0!6VDuT74D1~*iM7$D@UZ8Z!ib#ruT{k7}0>q?fICT4Md
zpipFhUa55#n@Ta6TbtHrS)2Zo?nds6?2vwt_shDh4<;jJQ?Z=Bqs|T5;Oi37CPifv
zY&h5<dnf<xt^;*hAO1DU5)Uc~_O1d%&iCuf)mS>cXt&vikN--7m=JB}fL~-cl3l(6
zM)3HTVMR3&uVn|6M#L?L4c~?qhjO=s7thysf8(fo=kkk<TBry`?|qpBo(v<jRS8_f
zWp<ev<9S}u!AIgGY}6k#EaSQ8-qQz6nnmp^g$EkmD>%Bc6)T{2X^MEVx{gpH8I~lR
zTnNl1NS41A_W;vYvEBDLJAu}<S*uUCY8#?S+133*SWn9lCr}~plNh0li3J6sAQ^i+
zIG!z*nrMTl#MFki$*B1NnkT<a|2>q-nEVJPvmM)fJgjr-378wcl&x%g^8|fU__eng
z76DQ+Y~a$k2YRna6(-aV2-iG5i|!HnVP8U^vjKEWGI0pgChwm|4Io~B)5iiPM?{4<
zK+ijBZ<#5;HqjoM1I%7x$oEE(n#w28k8+x!FW(+!lk#7PYW;bi_K0?GFSlFmP=*Os
z==ge|#BDc>BgS;WB&*zZ#ax%$37aiczy289i$S$syJX&L@08)!gGk++=uG`m0`XZC
z*-Fb4_t!oHFE|DoYWrmYiC>M59fHd}n4g3DVE+LIrE$qWQNf1utB7$S|HW8|qe#4L
zYdW6~l;mvPWMf1Gwr8${xa4%EAgK=Ctq^=VO%EH~uqZQJnSguZki|Hi-FlDUcoz~3
z3ux>qmdfE0yCqiI48Om%VFnmt-(c-<SopziZ!jtQ@Pl@Iy`|w*dRi25_?v=W=zQ0|
z`vjP0U?S{wcr5m62KiPOW|h?UH9163Ol=%eTew$i=Q<BM<uA5-uAsv}fnZO>9@n5(
zre0jGEO#~(=C-M?;m=);ngw63&yQFi#7fG9N7+ZaZSg9Bm}aWl6(b>*%+U_U{Ba;B
zMXzdlTs=cWoce-Sa-T{jo|{w^Mq`0Bvtu_CHL>_lcP&fsTe!I}s&8NU{(i?Fc;R*$
zVGBGm6tAZXFrZ8?BynhHwr@suT7wtBh08YTw-i#kOCv|Y3oD8r&WK0i-dkxF!odV*
zsHaZTF~TH-GH*k%bn!QNdvCR>|IrDzwRw&MMLr@BaF(e+DZI{B&Uh?e9|IIl^(Ot_
zwn0I&b=+QP(gN%=GuRHCLxd<YW{L;7J1O4)b<9+>5|UXo9sj1AP9(w^c(?5mu-LJx
zWL@E1_(x}UoMdb=kNH2!dX^(uT%_OZ%<J!|?b3C{AQ&pQ$Ep75`ztfXi(eyIk$cs+
zLwJjjN_qA%Fyz+5VbEkqG)S}ftf_IXYuvzl)AIoW$`JZTdeKQR<`f0A`=o>Xr@Ozq
zdg=SI?9QxTFGlDzE6QFRphk0X!r&PoM~8?zkBagFbL$`#Z>aMe`<!;}mD-YYK8Od8
zy0ZK^t&@KX2=NX?Yu%OOY!-?$&i2C_D*g(`Pg%f`(Wm7`=!s^Jm-dczSX>ItxpeQY
z50s7joa8aCC5D7X3>!)N+VptX3vx#@&mlpFJNE`}?6E<>je>8g$xoYew45p}4WOO1
z>!pjGSMyFaAYP&U`L!YyZVg(7Z21Hz1Aj4m%&M`%Uby;_hnZobM)>FptLhryZP1uy
z`11KPqwu0zScZEy&D-;j6O}T98CFJdl$k#z-&Rgl0JpE2Q?$BQ+r!qAOkBo<)W7;e
zEo4W5i%r{A{3a~71Q>ONxEBDW$)8};Td|d=#`rZDuO#f)NXPt&JS8?W;NMS9bqKiy
zjR<gAbDAs<8Lgr&D=S;59;1%(e?rv=K)+g|wsXm=#PoruO==6mQPcW0dZZbtN*Na9
z4;x5V85K61%03SZ0Wc@LAIq(FiL_ZyD;vN9druVg++f+0g&K%nB*_tGZMoP)eIkC^
zAS3v2mI`q`{xfE@wcI=4(e#Ye)mEMdaLJGV!eF!-f2j+Y_MYP4ksPo+j55SE1o))g
z5BG-OWQ<+U;xT{;ny#}sxV_5s%~fBri~juCLQ#o!C1V&2W_ASBI4kY4lb4+(M_)fa
z9e?+76l$U8uCJfuqf8FDqad!uO)c;yIwR<8&goM3AOH`{HZKfV)l}H`V<sd>oe_X3
zWDc*5Ci90gsDpIekMPxEoS0`C=ce%fg9H&5fCp<_;?(NLo^7_hI-)dqmD{4kRJmsx
zE4IVV>P2nR#eN?>=guSSJg<wr!;^3mZ%HRf=hi{^xcBz%BKC3wPSI`K$#zdf9Jm(X
zm`T*;EiZgdJG|xYNaZfeODJ>E2|q}6Nz*eG(<|Pbzs3Tm_}TAxgVj0*aX>7E6_T@t
zDP`Em`by{Gwl#}}O5QP7y|TF6CE*QRu?nj(4LY4F61-XsBcu4Y86H8zR>*q;D^MX`
zjqWyT;q$4t{`Rd<Pn^*S93B`lnE7t(%k-N;sBqH46q1<g>07~W3Rh-$FD6GuL7{gw
zyCDRZXV1MY0Tdl<vAahi27JS2YB(M~rKp+*OTU1<GMj#Zl`yEY(B|fqNW8nyU*Jn(
z1k#&;xwZpRlC@`)XBj1!5K;j!7)ziO_H}sE^-ax$`Y}-~=br$Avlrp_4Kw81{J7&e
z4tjpwytx5?lE(Oo0@oEOT!h;;hSF_->z`tzFuOnPSwkYl#Fu;Vpzvy}bdHN-_G_JO
zVeTZ11%%<Vni66-Q@q$cdLj5#1DQE(N~quGdkNVDEUMk8*u75R!N_i)DwR8@%NIgs
zv%!7$A`#+0>*hSQ%45Ez3Z!c^YU02?Vc_v44&%8FONbjdaU_-(9H(T{4GftBMz)#x
zEDRK>Ig*_vpx%|uo^!COO|v&@xm1d{m*ilxKs`qzhKhh_dh*Zu$ZWSCdv!wQ8gwP%
zl^FO22QmUI?R2JLM)L#SYCuQ;sTlR5t6lPHFxP(Y5i8_(Gpa-w$Eo4>5HH{?3O}H$
zToI?>+$!WfT~)wGR5uddzO$CwSXvwI5_qLshclL5C@IDA<+m(1)N+fL+LKp1uJF{P
zV~fQK%P<$MSJiJ~>a-7l8)r!C-RaUQ+Yvg+RX2vs_4Vyd*_@u-Xm`Pf{dz^ZT4QPZ
zmU2y_UtbVsTK2rA6b2)dKIOQs_?)Cu@tBqTwFR!IG?kb(=@Wwf2`@S_WDwN&yFKWL
z3K;sL4ux;AFyzz_e@T00|Mk6c1AY*S``WL=Yh`XEq!8q^76_>J6dN}k0iZ`9<5=G^
zRdIL;;Pjy#-kS=jSIv)pCcxAikOgIsr&hhU{cj8VBeDI!U580d=8~R9u35k>{f7Bh
zdv55j5{Wrgdwd=Y3I5YNX3!vZ_<lLS<dwG5he+wJBJjWnxLCM$Z-QoD8?n*58>y01
z@+g_t_`$aC2+wI$277re#lmdQ0UCEI$2{wrRF?kyz8OOnuVKe=7}ds^tacD#2oK?%
zbtBwEz~+3dSm(TOf5tXWlFKdWYf(rMJ%J2E%l+Xkx{blCI(t@Ee+tCWDRxZt!PZx-
zYo)^!r=$!8$8}HIis!KC<8OP&nj9o>JDT($MW6>9T}aq`5CCsq_7iCo9nHW#suG5g
zG?ySn?z}<2qPFk!t_5&=alKfS%pd$<sDj!SLxIIG8Ze3LeKAUd@f)rN{}M$c<8tbz
zWVnCPh|^Lf%3zbq)yLn1f8>sRocZ=k^Xaq+;W6Exxfk#7rV^`0a{fq5e|O=!rIwyx
zfH!0r8<D5WtfRiO^%`UT@CLpaMn5w}rr{a{P{C<k4WyV>{4z-;cT)^twc)PrU?g|^
za_Gp5mbuaTp}>%W$^AbZorfcpZyU#-Gwe<Fo{2>EcI;h5elkM#9+Azl2^k^D3YF|l
zGAkL`W$%?ev)6fV?;n73j^{kjb$`d_Y8{AbRkVhcgQl7~$eMWWFlso|NDdCv%E>Vj
zFJYMKPFN>38dhb`mMLyJ4lNk#8TR)!{ntBY`q=V5`{y7JdE^hr+a4)?(_&s6k5Ac!
z7Wzf>QQ4X|5jUA->ou3Q1`T@^P!={P@a51|+3KH|QI^kNxc1Mmum6DoB@_1w)BF)L
zr;EioBK;e>Q+BJ^r)w!tJxb_84wMHK@z1-Khg3V{3G&I~{;@2<U)Mh?8YZ4oi;lSr
zl(}TodN=QU6VY1ZrX+sI(6hE*MMpXZ7o1KY-Dmy9406`k%F$|9v@FUSLI1oHmRhR9
z&4Kq?#JV<G`p$?ipC=*>Hfb;#&|T^?g58-d<G-<*A!x0k(3HR;SU8R@T*^o(MUWC*
zZx@Y%OpLc0(FUXjDcr#<VSmlD-pxJ7teGj^wMzZ6PaH@2-ei3@gqLqEzbPR%!5jk`
z+V-QhAkaiEFZU?QwQvM>Nh|o{c22+}t_PXiH)+G;NAzMYT2oWvedit335c%*q?g;=
zZ~w_8j5bq8hlduR)*@}OGv0x#)nR2DXm>+(uopt7FaNs=yr(%m<O02o$P)5f-8aE%
zJ8=;!UX%=hQm~v5T4JSBE+h0^)bux_JZ|M>LT~HCzMt9e1om1Led+V^l-95ERc~6D
zZ)l}UPjR16YFwHH9;Iq!Nj~;3(b;-qFq@G2HefCXr&ZM;QrpoiU_W9y@MLCE&#Q5s
zta6t<0S$AUUicnW+%Rx9+c6oq8gOs#E~~j>3a)mqCVz9mYhD%vXx2RveDy(#=GW1c
zNq-JWtBMo>hDuL829kc0j#StzD6t(XbC7F#@!Z#MrT*qZR#RIE?tLv<s9)PaH?IIQ
zpTUq@tBUntq8L1)=M=HqxS5#LN~EvBOzIUv7__Ueb73P{8DN?z$4*t8`(NzK{w(j=
z&+LhViiyY{Ee;Hng}?wd$p#0~ZAKh+rg%y={CTDyCnSPIf_uPlazLp*hL(cXqVjpr
z`~L;!;84JBtp2)&lGjhA5zdg7eXb037ZnZ7aBfyI8if4SM-H_b5y?<PIn4aBy!Z~Q
zgr-KwImSm#SLH#<^_`Ucbmtdo`no?=>k_hgLS=i>(nq`{z0R<&e26%pkK`2|lzj0w
z>G{rjZsNF^6GF-cshtMJj;!OrZ=F7kBv(nWn$f+u1+4oFx2{RcpNTRpKCZ#jp{&>z
zzwBkfDP(B6KHt=U7EKAG9oTr-Y1AWi>V*Yg$#tJ(PN<IBvzj?yXKjY4L!s*!YFeDn
z`rnz}rE=f-@)m<*nQ#(B-{Q@k%&Jip9Ere1t^%KTxP?v75*aj?^g*d|%Ijs<U=Iz5
za*^lUWM4Y5)Zx$gPxJr8D?cur7H2#lUy${bwX}1cpZNakCA~b9I|iy5z=(m2u+!(=
zZ9mXY?egE~FqxGv9)0pmxL6eW>RQUBflR-64I$mXQ(j=|@4Gy5>BNw7IAP?-e!3(|
z90Fs&nC`|@#SWAX*b_&s&w10T=bdj&p%Mu3;qs)chpFII@QUXS1TV#l_0-t^);Ro4
zG15MmEo4_861RkvK%iB{vc~NGgUpDw&5(5x>&3L_>;TtgCh4uERgd4VHVtdD|MZpI
z{d(~H@ege_6nwr2t^<HG!=c;^@AaL2VPfW6<J&M`asvIdZnB{1>(HP)#wzP1D1ZUj
z*Eav1%mufmnW)2gWKX|yU|R><?mb-o@bJMy!rQcH1*QR>r!3ocVTWr&iaDm!)h(eW
zF?z=4(Lh8|&(-*NHr{L4@V)!!Uu%%a5;|2&>dh@Q5>=oD>z)mj7D={Ce=IVOL~qx>
zmwv^A=7>hjgU_ROdYAr?;(oP<Cfw=x5ZqNAgDK^4aCjp^D5)>W4pBed%DEKxNR+U*
zM9@>LJaC_o!p0|`=t<fJV7lx)?$$E?4;WFqbnz>{TMhQ!FG!kuv!gGx<d~7UMt=7#
z?`k9F38U9A`NFBVB(yETU$uMh@Ms)+HNN;kxZy;D(+tdc!fQ&J%l-Y&pRyk%8B51L
z`0i`EmuXbxkYied7a$|Pn7U?InkwUn$XY?QPHI2?-7%emY<rA=n)9BD&`I45;mDPw
zEV;#&s)UbyjJoq|VhjBOQb}g=nTLr;YK#`CDI<a?3V*V2i}UI$J(K+^N*40vHRJBx
z&DHAL?zXRcM?@vj0i^7+bR1Afo%(ys+&~8sjWu3)L}@EZLyx1(vF*=?hsM8nAI0?O
z3J0O%Ssg6xN@|(cn#EoAE?f%bv~5$=E{mza+A(Tw+2ZS6J=22E_rn_u{L+w1I;q?%
z7$4$OgSb_x)cLznk(*G3wZWmGna=`NpMr0Lk}CtDb@2G#$bWCEPh#YBZ8>&W*@RJ!
zD<;eO23U|J#NiIQLdhmeI~hH6;MDU(jL`2|M5wAD`N`b+-!_yS%}Q`ec+5+@%l;T7
z!?3_+NRH%jmGOz}r%2Z6DDluF)5dRF!Uuuoh0}cN_^K-%zRVL$jdF^`YO{*r7bnbc
zlva=-`A%IGuRaX-69m_SwuJuWMbQnZy)N0XPCMLYMZY7f{=B!#mFH@RWEr=7g8zbg
z{_&>=jE`}{GDFhavwd#Pr^V2QomJw(o?}HGDlWL?>Uhdw=;b<uk6^#u7Eb#c*n7p>
zZw|=s&3%SBEq~{k0cFIS5_Ae`@!t|{XJEJ~^J!y4)Rvkln@^A$!-g9Q((x)at}iQI
zEzKT+3UqPR;G~7v=w#h7hod9!GuT~pshOBxn}Nu{cNWPIW`1N)1Ld<o{=2%Bus)ls
z{y5^dY(G2F_>4d1$;8Dy8M5&ualhD|myO~Jr8Wz=KUaX8hPm%T0H4o9a`T5Mk+u&P
z-)OGq$J@mm&`Dx-BibL=!S65!19G(MGRRf=>Hj$3Lpc2=n$Wa1Tv+HeWFL^A2KC5V
z3F`e`Ni@<nC4iF7rkui~P={65uNb9z<S=Isy`;YP%=Y&73S#&K|8Z$&tigi2Ip5e-
z1nqoTt>MwhHh(6hSC~w8>ju){g=a;{($og3VZ^E$7fq=rvhXL;>jkfu!82pDLlT$E
z`mLk|KlTpm-F#X}&%&}Vky6XRF8!sg7Ck>JMSB0!PSKw}bhcb1Zbh2?_Rf7Keotfp
zc3ZH0_Y#D=T8N@*JC^TkPuR9p@?;7KkOtX~6(E|YfJU^5Hx<jN)J3&*GifFO%aqqu
zo-%YYdYiN<{&1#*rbxF9S;($u`Tw?7qKl;4N3W$)PVcve#D`QgtBme$dN1row5qP)
z@!AiFFA@5J0Kqad=2Z2|E-RaO+vh)hUmRaIsH7ID+bY73-N}KDl1R+b*I`{LHZ)o5
z*nAaSh3Uc$24E2)daj4k`{9Qm^8DkY*e8XrC2wSY$kUyl?D7L`<B@{Jujcm*#Qj6U
zy&t7a5ixsYPzuIEoAH(Me|sbi;1QLIJMQ20-}>&Z^)3Z3Rrp_|(&D6JIVWD?S{Zg-
z@z{(OoMjWRB3>=iz}wV;R`f^}(mRs(_D;{`h#^BwX>zeqUk<%)2I0=-Z2z4hr@mJ+
z69}nsU4H-QdgmzvZ>CHTlUuo#4N;Dz`izRfEk7mUxB#ljNrp3)X4zJ2JNeri`^1Y~
zbcy7M=^Mcbs}Z7spTA)q&IQ+H3-g&v48W?64f_yGjz-^N8Bt!Hn++F_OFZKImK<*w
zJkQ^J_~;Ax)-)6N44cH+H%k22|B^`d?IBaMDZQJ*m24#+nT$n-LIgFb^RpvoUo{T6
zNN<=<qjz%b0d&H;!>aY(TnxFx+Hx`k!$s*_^D5uzy;6rX_$e2dLb2P5FG<|$bK&zG
zkk$3Vz%^`3qBGk&&OXBbBD)mZbA4F3!yK2qy_fo;4kkb@4zNYLK6JY=d!(pn5nWR}
zkyCb^r7PvHx2He&`IC47!jDVKna!(El_UsRjpj?;#R0b^94sW;O?dP?wWuKj-e$wo
ziyel$XOW<_-l`H$$20B-`ZquR+f`^!aQ~(wB!1zVSUR~cz$VQ>YH60<N|L5O5lqt8
zmHOY)b-oML4c~}D5KJUlp{I`U|Mmpt_HG%^5gbuJd{(xgoGA4}h{SQ%CEle^zf%3#
znQXTFw?(&1W-*$b%`ZO73Sst(t}Xoh2&9%j(ejdI`$Ge?M!fB_2+{F;4`eX*snU3y
z%%c-h9<@REJbN)n2dSgY)hzyP(PfvSOL~$ZhQTX-CfJr~aZCdvESy{<r6k;~b8lhI
zdq|mD9kj35GioJ<zIwazt)zW2iQi)G595ejfT@XAHS6r=6EjMttX-%!n!@y0ef0LO
z67E)zc4-r|6<w2v*q~B}hh^&8fO&`*hb3mEu{O&%qG!641)}i11=s{p6Lmln*OgHN
z`9wYts^J*}z77hDW!Af*b6DP$*ZDniIq2$qeOaw0+7VRryqXfD?+%3?_xQn5j!Gt5
zB>|Aca{sBehYUd|E;$;Z#0Fd(7#6y2I!Y1mDwyr^gFel1CeVnCu%fwOs9M&<=E4TI
zC?$HdLfannxX#?Kx&H#UKd`H7`;QvOF8hUSegtv=pY6bkU^927##?uxBe^~-zI#e`
z30qB`1rHTNU(i<Z4OHSLvAKD#r&^ehXbva|feZ0K;#%JsQGqivnK<~wE_ca!b{mb5
zC5HHy*FiR7!PEpOWh_m!0v;*N!1r*~h8V*qC!}3D@4RZ2&O%lR!=)0%lv-AIGiVu6
zPc2yg1uxu#RR6jn1J8VWh&R(qb<Ja4xS+u|KrYq7g5EZ4=SK=j{j2{;`N8i3)Kbbl
z<I0<0h!fSjNWHoFW%{;D(}dw`ZRHrqolnuZ_C15Ev1C$Hwjn;Y`9BV6&Uqtl6B0&u
z6atbu=6PJykPeX^W7MI`|DZN}W2>>ls*X4)in&TUC1)#aK+x4EB%7}YS|H>&9$RBv
z97MmQc}u4vF;v}oItx-!%a8Zoct3W#ixo8esVlEi4M&cj_(9@-hHUvOuqbKLdkh5z
zoXosRZIe5Z8qoX<^UekC7)2_ciRRspA?)0;q0);TE2yDjC|7w|^hu#3dDEeO85S38
zPmyffNA2Y=LJ}lep}Y%uFK#(?Y<uAC!r~RE1f`XEm;#U)wgOT~%VR#`yYzHgKc&~H
z6oL{ib4&p^MAVXR(Q`|y9Y+<fUg9RQ7?YM+@G_+>nKKNNSNJg5DiYjbunaz?%bk`E
zJohv`M_RIF!Wz8YwB4Pvlp@;YB$1HS)6jHeq0^69yE9Ndd6Z4g&lGeoD29+Uy@$dx
zQ=Wj=>ZbB)I)|TgJq&NI*3j9(`L%T~`o*B9%~sdTn{?q!wpj3Wz?h}yO}YH0q;~Na
zc=Bbh2IQ%AlQ`PWccP4BaK}WhK#JTNOt%LhIGobB+dipDF$4>3rT@+6zgbuMNVTfB
z=cFkQ{s1JiQntu^`Ov12jJjCw>%_zev-eB|5AHLK);{xR>AC1voHWZNQJt@^tM1BX
zUpF$OT9B{9fNGXQn!~Q{s^|~YPYJh~?*yCi^0;MRcO)sTRK9YWw?6RIKk|8?Th_VC
zJs@(m8;s$-ky1T5-R;MI-%DrXD579%$5VP+m;|LLz(QFeTpc%Vb~bkSj{sM0&9uy@
z<MjKM%i#A7D}Fn#Reh9h-Rg~`7jIXVI$iR6^35_ndTU@?Wnq^$B|LW<1d3jBYNmC8
z1LPfWJokvGL83kml6zEN*VaH}Y8(bsurqK@SgxCL(Q}SAy{0Ippy<YKs&9V~ZAjO4
znPZ9ObcrOSrhO-K%k8&2f%$UJI17Y}dT;ah=78f8RNXF@5hd!kL3&@s@t7fpS!Rul
zUFC)fCww|ZS%17reEfWJluDMQZqGFM>gWLEdkmr*h}xO;Sl>^p`e0c95FE)WaV~?o
z_W(QCE{NvT5=T7Do$Bkzvk#L><<UqqvaJ)zEhIAIWk$C6+Aa2HbaL<YYEJMd&Xn_=
z8SL8(=W8Q*yUs#&cJK)<LOZq@c!eDQpUeN-X*|(?GQ&SL<;>kP5B<jYL|_ts2>atL
z1+USt|2ms||AiJAiFc9!b>MLp;ewi(6BoYELpOu@yDt>A_U8)R@xh<2$42M3XFynu
z>!c}ezxPqm(_!(O7r@gjByy^F4D}v_N$dZ4AMaIBaqCqL`80M7h^F=|CCHUUnR6)F
zr6EzLFMx))QCm9_VF$G{gW|hKxCZ%0IR)A=)0TiQ1~OY-)_?i$uVY77e{+_iqJ-`t
z*<gFW;cEAhZKhZdGFO^S<eMpf;PKW(hHs-pL)L`A&)LK))vx;7o;i*{Qr3?qulfy8
zAPq`?nY$ejldwU9+!fOGT)VkeAoyy6lB+_H+@6eL=X@G8zQue(_ETA6`1XA{JoLp}
z;0H0j=nz+OGzXsKg$Q)#o*)|oTKX0n{d6u2eNIO>=UgG*O}!K}G|9BKd+`&_UXmN{
zxz7B2^(H@F))VikveR^2-g@2mfKKUOMqXZMkA}d90Dl<)n4rTS?rmzE#)Pdoc{g$Q
zMWrmZ`E*w&#BP%Dr>>e?`q1#6%D->m^7cghq5r<p4*%J$$XYca+*1c)vLAahC91;%
zouBOf0-V0b4V2p@eb{)1#rFkqn9*u@5Quk|ycjM}*DY3>m?uz(+$3a<Q^Qg9zcQ+I
zpd8MyxwvYdng?DmiTHs%hJihifa>mi823Ial28kJ)I^CUOu$B;$>E{!;G}8@mRD2|
zu{=4t3y0ng5(p*|qF+^e=Q`gROi(NKb2O3mKYif-{7M^c4aP*1lbYZ#_zg-TzjE@M
zpk)ysj_lrpk~GLK&n51%n|!*_7p64&X@f49B(M@#5(lLJ#TlHXUv<h3V(RJ8sNKhS
zL=Y`FU+rsK8bkJM_c9<Bs1S0Sar9J!nnqk3*SY$-wtU|GYi-Y2xbgvZ<qD8_N(Ogn
zpKO4H!L9vdalhYg;-1Ap#3%lt@!oDfVzDc~J&#Po$Av^ius0RWG5*o#<_=c88ZEJj
zZp<O$H}PtXThXVVp$5io6)cvSVi+i(DY(IF2&}o`aZI4D%=MqCSL(6#_UQ+ian{o?
zNk(>KJ9>Sk;{<=$3I>QNk_9Z3FCT__eJJL>fmn)A4>=&vCSE|YH~E2;*L^eFi@@sq
zo?(b7Ii^(yWKJ0PS~y#Tp&e!P=VW`M<^Z#Z2k7^buthfrHm`k4hKxqQwx8i_to536
zsWV{=h1PZ@b+qUupfy&;tJ@k=iu4;+zpcFD7z*W%=q9ycOUGeHO6D`)g{@70slG{@
z8u{XOBz9fXlTz0pat}HKk-xdtXa|Q;yK{X(zHD_oP<tHAgPsKsZN`i~)IUK&VhkkF
zYXpdeuH)k=*J50vNT~;kA7A2MH!#mXPT=#y<r=E!iN|8+Nldh}Cv}UjY&<Q@f0YbA
znhm<u0|(7{Dm+W}8tP0J55ntxENL8<oQOXu^DkO(jUgx4>GPa#nF6}^GoV9<h41LC
zx#KoojaZ9TQTT75kTTYmM_rWXg{9~PP?qLgfZJ>u=AsqDJ9e?%lrSh*<ru(wAj_su
zPP|B1vFeVxjru1S2dDB)mIXKfN2Iy8%BO~geNd7K%X`M;SKM5<hM2O!7#f7lll;q3
zv&9_SMo2smt|$7baMbBrTKfB0;n?Jf?4n1`DOUfYd69*Ewkd=J0~9{Wf%NIZs8LEZ
zQ6Ng$D2a_FE9iV(_i7T1*?9IH&f>b+BF!sbW<hU7kCnNJ>YIE6&FRt{Ka!(O1eSQ~
zCpb<Z#FTKJ$2H8zDs!xbEmh7$Y|h=dg;N`O>$13T(yhoQHw142>yNm)aq=HaH%?_t
zteUGa$1`RY$c4O$t&^KY9zH)uGH(5|E!o1umhpw?)u^#+sv@z(k4;t5(z`V6a<^=o
z+~C)j?|lxV!jQr)(}p`BSaDN&8*7VFdoX`RBAlGHxJt{W25HQ4DBq<g-+_~k5`x?J
zh)2J~s=nG3Y-*oHT1k{4{xc$YZM-r+h?(-}C_#J?_Y0trG>=dxqnD9ImF@agdf_09
zjUFTb)4n>>Wk_|uq-I&`wIAgNWhsD@VBKSWF>cYw@`<fX{>Jitm?KMhF3U2H*>(-q
z3*%1bfBXZ^t?*2PW@{8MjscJ>F%$y|0W9C6NXA$U$>ckOngJ#0mwPAUf2{JmKPVpR
z!ygUp;wVMmlx9slx;Z&jE~A%y?<HM*Wrvhs1wj<z0R4w6cuGPitd{Mwh{vcHYcSVZ
zv-|03v!4ej#ryR|v0XKfmlMF)J~Hi9vLVW@f*k%7{=eUTLxALfW#W3&p7QxjhyjEY
zhu2;d5nsS!l5eL*V#t+1&pb^`td+<&Uq-;eq@$Jganr&Rtc<b%jdqRbtrHD$6v!XR
zJ7W>n5o&S-?WqP&+)Hf#*lM^YGPO$-i5hnkGjBsj&%Eg$z@u*AEpHOcBS=tS>{e1<
zzA4rLuctX&G;F|O(H7JG7y>5(k6Bu2eR8Qism?x?IGXqT;#VEXlhZ*7I*UseiXv@6
z!3&LYSuU|RSS3zL$`65Vif{okw=q-iZlM9`d_8<o@jYMb^&A$MoAs>)Q50zm{pmiL
zLOqaFbj3*X<``r1ZH_Ot0Sdu^f~B`QwD2IHe3W+%d+^4FX9_&j6_G9C-VQS+Pt3|H
zRgw4E-3ydb<&@w_PakmN==TM`@fcErC*NM1a)J}s_RE}@@Oc~OzcVEAdjNu4H>?f5
zFiUH}>7mQ3fRL)dssHMif^JHA$5)ldVx!gennpjCC;(#-=PH^gS;)5k@H*@DG-=E}
z10bWV6kSqdKVKO3T_<OCtOy}SQ4ytLXl*m!?*tw1E`@AO<Psf(YCtCNIgqw|D;FFT
z!}1?pj61es@uA#77rXV^;Cr5+nh7sytHJ&gszsC?)m}CFmNu|1S^#IS+&e%3857+{
z1?{NQJpUrI9y@0{2}^=bQzAf<`Yz{BGlTgpYbg?c$Ir2M$?~1o@}>M^_YY;KSY`mV
zu(mrwwXMN@>4qDYXa*?iV0h5`+axS_3GQ?bqtb5kD2bI6y{7!2N)s5KK2aON?ml1@
z<S*}pk;o?AfL}q`*b3YtW=|&7vBxc6F|(ZFJa?rM+<Oh3-i9g5Gu~6$N;wU{%!=hd
zzov6=$6Xy9h|Jmnkl<ZY{!Kz`@=uyCw+0}=FlxA5&s-rkh+Q(@xVo+@h^X6s%MDBb
z1(yck%Z?S{J#u7aQLZ|>!yjT>M$66k|3A&=$eO0zeZmGh5NUwBc20D9o$V#|1HH`Z
zchd9~zwO#8fd~O^ROgt8I^@QpPCKl-Mb8TU_20nsy{88!C%8_e;T9Z(vHpn*&*E<E
zR}bJ3@xt87{%GH-ZK$3MF=KB~9Dq#2W-|XeIp=1ryi@M*pC|Fu*5Q=*UjiYzvH66T
zZ2*L8HI%2ULHuH8`p!*xeden-j<pcI8t?M1ET=BF91J%lk#gS;cms{{@YKjmwmS?N
z98f!skFBVIk@?A$5PhawUl#v`t!QkAmh@BH)2;(xF6}A4NGNsZcKz@3qYnz9S*<xj
z!zT&<jHXb?ML<QYY8kMbnQ9s1q#OHj=H<>`<23Qej71Iiy!Gn%-OWRJiVJ`$Kh@y@
zFYTih%kA5$$AIth<;#~@KF?$8Zpsq-ETv!CZTn+E<`v{)e2`d1$*nrz^z|snlhUE9
zh;^~ghlwOJTkAs@#>h$T%d#c1t1U&8Fv|EDHa}Y@3QAx+JZ<jV*ba`JW1p($54d+*
zo||`2S#mAK;7PPf2dmEUlcd*tm2Re)C$JsbX)_;rQvCP1Gs01A38Q#$OiSQJ3LzSI
ze(W@w`q~L#GtMU~Eu)gLo9dh<@>)CR!T@dPNiw>eESme&(Ut5X>FwLk{3ojc?*+Ws
zVchLRvC?9ZQ^~KZJu|mptu<P+Z)@3cE!_#=65-a|?Tc9Xj6M^JrSbm(t)ls@bXjHJ
z;r=e)?|Rw*0r_udSIWGD5?O0*gHvJ5b5d#=HMiBDeST(LPE>{$nx8d?lT<AkX&2D?
zfsA9$0Z)yqs}MOpyw-c^oY<+VNjGQ&v5o`e-X$mr7D=HZV9a6w5f=_q`kqV@a5xr<
z8#%5X`VfZHgboN|n`~wApap06GLAe%+MBG4yXI<zA8ZpQ7%D30nKGwq(7_<Xa7el~
z6sC&qw0OB0g&7SZ<fM@_#<Rw1cuBKV4I^aMeRK_S`v32#gfYtnW3I2fJ_wX)R!NLZ
z<WM&s+vWrs+`w;$QtXDz$q~PC8dQc_?zsp#2cNvlBJ`k-+mk>AZ4n@0A<QEPQbY|?
zuJ$RIgRKFF{M?b0Q<E)FJ;YU+{RuA=vO}GM=^M$%O+@BexfZIueeVxPJ8{yAK>f3K
z;4-07;a)zh^Gl?vDhXQ56AVwqdl9u-nFZfU7w1ka@Aj{>f>HP2kKV&>(2i#VLekOo
z6}UgM<^&wXO#AOP50{&@7UD>m!KwKS`9lDyvy0Jtn)N2Aj4cvd-g2&i<x69!6j7L0
zPlRo(eg}jW>atKl`#jr{?@iIvMc9RexeAaeyliH%Bi0%KEN@6+oXBpPQ@(bUJ^pRU
zbAOU)13z^rIU2^z&VScUggbOT3bN&eu760y;HID~gu|$B9fB1Gt4jm>-$d{@+ollr
zlW-}6M!R3%{&i!eQbKbYNMg_Q;dv&wJQVWMvjSM#s<DKI{4lF}%=Jm@3cxK(nfDS;
z?1Bn;Ig(y8kjHPuO?>L^hjqrwE2c5z!0TNhtY;Oz|7=P*SWo%QT1<jKN5AwQ2+lt&
z#_h(gAxN@{*36>rKf5&wzH1t3kw2zs&K3>4HaMW}Z*`S~*f63VzhDv*dwE$hL_qF=
zs7AfvVAy=ZMaRpgfbxNx$$v9qp_;jSwtjaNblv&q3=9gw_{zS6L`>OeQ@=*cX*z9t
zC-YJU=)XQys8_p8vIt&oQzsNy2&GPAY+oQPg-zSC7=UVU@X;ifo^snuwCPOn-^OKH
zcw*1#L%0qT5oBmvRRcLiAObb$FSp-s^AFzsDsra`)>GZp8?YT2SZ!f^M;z4nN8Ygv
zhf1Z-u^1~=pgkC0-=axUqKCUgC$X^}KL@kqrvrd@^9c@`ds=sN8#O`CC38E|Ep$UA
z**!l;?zU#|BsB*+xKoZ%8zBI8%HWadZK<=RQ7nOnyI*ce!d^pLV-SiaW78gX>q{Yo
zcITQ33C}(PDPtZGB;d;;MKDa(iP3Y@{|rvZ-T@;>URmIvM!3-U!~m9{piM8$_LuKO
zgD3ux1oPlzeizh>Ka+>51(2*+Ecv5z`^@%EuzIGtoQ6EuG{)KtO*?@6f8$GN71b4R
zLUn9U&Si|JN0kX!{xgQ+FeW%#mWWJa!}&&dq875v#<8nM=c&?hm#-CGq1Fq<<Sm(S
zrxU;!jZquD0--)0dwPJ2zz!}p^p-5-GilH<rn&)EVb7`|6UAnI6un4Xb{X}#w}zSW
zw*KMj-7>rmzryambqUjt-t$1{`<}ctaOH`=xj;^#TQQ1A!($6)%Dwi}hqqszGY1`r
zcO}hmP03UAdVQlQptr6ng`<R^`z+^Q78=S#Pc2E>6iH+_Fc))}rdhmGsZU{9zssRm
zkL-X0b)u$)OWBGLR}diNdGvgT7uU1m>-IJmuM;KK4+nsPQ9moO*uY(_Y!HUQVQLmc
zXk_j(P3~8Gw9Mc<RhpI!stzBGV8D^^n^s!qAp3X2*CpAQ%acpEdoMF;Y5L1^Yfv7r
zrV~ATiMb*EfYl0*$9YI;tP!74^xw4C=ipTRgPEk(a<5t93wuM98ae-qImmo(jU7y8
zS2bNzT7sJh&fsk@#=J|GzlyBc8lqw-e>Y1i+GZ2+ca`|_%!QCSBR}CQIHOj7*yYWn
znFFpf$Sa@_Jwe=SJ)JVMJ(oWcCL7rBkvR9fvnI?~>Y2X`Aj%fK%_>bMy@n0tOl`ii
zLqmdzic2sRl%K00<~D3^2wJA3R1-i#B|@}9CMi)#?z#0^eCS{VOG+7Q(RYTg#vaz6
zaLkF94Y;Yf6~8t4RQ$V2r$UV~y~0O&66LEV7*G_LSLu~(bvP_4O{&r01>%^%SuYXY
znrT~ec+zGr7nlwhe@E)zJ0uI9CKMqrEv=K?xGlxG*p5iO09zSGI;Ckzn8ezt<q0jF
z+6~mr=>tF|^s_^YG)J4wzxHqEqfz7mr7h5$SSUqr0Lo!nUpRAY9n|PqVYqvQi+&2@
zF(N+~ifB6Mw>$)>r)DX6Pn;gNzR?#2IYnYOM9RGWta(ni2|P#CIRg+Lv)fx!P?5dM
zMa5+Y4lwx6U2ldh?swlk@id>^ZTIlc%6h)|zHlw6bZHZNm?-7gYyV0?(G6r0iUxu!
z9<#-9WP1*)uW-pFBQ3uy49BT4e0|;c85*Z-dtO(P2EYW#Dv#9R^Tx0Wv50s1_=^+`
z?wm_T=JbW6CQQMX?#@^oj@ab5GJc1Nfc#kW^oKV8)KlX)ujb=PM~^;{vBHkx=Vk#c
z_j6vWnFl&o<Oy0~1PBy3Sq8^Eud}Y0?!SBioW35wzBjSmJKmi4ZQN<xJZzlUjS}Cx
zMq3|%9?gO?B_}VQDmj%F=L^ODdn0~B`02>pa7bWATFLR4@{Q(ilW7KUM+uzNZL{Ll
zljL)~-x7(4DTOex8f6WO_!7VXGM&49Jo)SoxutLL#njyYDJ3L6pGvEQP9^ywo&OU4
zW_}lfGxas9kdTil5hf+Ow(BU_${bB&u@JjUJz>usDLf89S@xgyvGFS`$S=qRCf{z+
zF+@Y}J3Y+StmnPiN)IABuiw3SZ^%&9nG)IWI{CQjx1jzl!^ys9%nIW!0H~-}U1GJR
zH<{x*QDrT8`~no5D{MDcWh;#-?!TgrGyH{Zz<suG5V;Pm!MA#>6A9*I2+}X2X71sF
zMJ`%_3TG`zSNT=dAenk4q7*OgB=3`DMsXtRcgd}6`DYh&g+^yvUkfzcuWApD%EIk(
zxWV&`kdflFh&3dz+#)hCd>d^$0FbNtwiIjA_b<q6zezd&)_B|OTdS`=a9D>GZqSwq
zp|Ix5r)R2)S^jgdu}W<u8Jx8e55FIr5yuAwmqPk5ow^PWa;90FXCRs9@r$p14c~-&
zYv!;%OU5hujD`5e0*O1T&GLIb6y7Z*=$eq!>X{NnS<njYQuQvc>}_GDF5!I<<64Ig
zOYvO|d(C0JXt%a9YN6{F(*o(U%~!7ScHp0X?}mWK+Z@jEeu1P6Z`M*=RKYT-*IDes
z358sXw3{s2X$<^&SYo>dzUiBkS%24G7BsA-0*_AY^EC9Uk*e!_UL1$nH@%)?=^(AD
z2s-rvwbsq{cr99)pnBS>5#sS&!spCQ0bzx$C6a%(2A|BT`r{aMxCr^y%Xec7_;60U
zeJ5%><ENv?TRNisE`!Ls0-e~sw}N9x+j4pUqmwqo?DP&K<<ynG-uiQx%heK2hqj`D
ztLa7+id?)%p`pjKl3)M1BSKq6Aq^b7yl_Wjwm*mz=?b^<8x*7e@`Gojy*^n)J$3xM
zP1mDSbJ(M9<EQ5r)BQQTf6CXJ<gWJ<g4c-Q3~PkfA1pzu_y8cBIA^MIhS5AF+n6_A
ze8&lp%j=tTs)|EKGCUM%9vp)^@mRhkpxMhSrjiVg1=SuKxO?RTWCi`^2d5-H=?tJ*
z<-Z(Ca1Zy+RPB>yq4D$cizq39l;baoR~h{+N8lZnWaEtg{W)(Y`6crGFqkNr>rBaF
zplKl&BwBPl*ir@hf*Xy@{X>U(PfPN)J}?kfTPRa7kBnxzHM4j8=FSjKq~9fYnlIX!
z-{Ew{g)@WD%+02i<|nScezW_YyAv<4WEWRe<>*U=u`$CJ*MrXqnPsOP3xCY>KP5qf
z8reyL7sf#g?cuI1fro<6?~+EG2e-c>n&eebX3=?z%UKAl{e$<15)~D5OkARaScAPC
z!N*e|5`1)}dPZLg^WV;+L}tB43|;!2O2}tcW}jAeGU@5Eh6S&O)A1(B7v6rD-5a}z
zz*Xggx$z!zfyc4e%N_xe=^TIWuDYg7v?R3e1EO(6%oEt$PpW>5PdeAnZ~;CKD<1a;
zM)Z7CFEOZHygn`a8*jyOd_)}WIbUx^@*2r)&Hk@*pm*NvnC%0Y-qQku?*y89>L7EY
zjlm<|<RLPeSF8=3zaQZm;?Mf;%53jqS0iL;w?pMqo`USoI$(&;>a}o5E#QCaN5#$F
z(b#=qY|&5*Z!sIVr~2)$B=_}9nXC5bzUaQw8A2D|<{8J^GBOo&ce6@33j>1i4KY;A
zMqpZUx0VtjMOzsYFvMBXj$*U_j_8@{bC|Z`fZ?9IwSGT0HsN>j4m%^sgaUUZ>ai7C
z^5id8ahuZDTi5bp3-mrcD7uTI8}>P_T4U)#Lg6Zdr4tNuZ@_=l=qBXecP_C(-Ec@E
zbUXd3?J9{<_|9VlDdCOK4Bti!8=#6bzZ&2+u&lAcaHqc`javaSq&XF0y2g3+y<HJT
z0k8NTk#PMZh62;t{10*}mYl*=W~q0DlEN*-N`m4@)L5)45Loy%y$T^!@3)@h4hm}A
zvA^@rOvAqA(WJtJP7hZvhZRZH$?UIwx4Kfy8RL31y@j!Aoz<`7e>x+t_%#1J$@ZT;
zY8u!K4~&M21;2eBvN-e)>XhW1l>3`*nVBE*jQtfaG3__0!#Q?|UVjDg;*Bew=Sq2)
zzjUuH*1#P{Vm0%*77)B=Y-42~j<p{6V`Hf(B{duiUVqdeMtegFoOM_d(Br(`7x0ql
zrrE8@DrcRPSUU5Rm?CAH7DfDUoV{YRtDdY+K%!z<VmkfF=D309L1V|Y2#zAYHb97b
zFpwiEpcFo6SlDxVKKA_%X#IF}v3>+9MgO4p`fx{*LH<~j_N^loeTFZhK$G-XvZlor
z20Hp`n=+A(^5&Ejt+=lhwcp{|f(>(`1$vA7J9D;(Vn<T);0;|@*>ac!&PG7r_~u6(
z4dek8K}i^F`YXaG5_1gGANBHV^mSG~#uGyXx`i3Hq9bLj#2$UK&G!8`-1gWL_cTM|
z^K+(x{?&n;U2OjM+X)PJ@A$_F3HG+c4OUreQm<Q{CgF2QDW64@4|UC`W(=KJGJ0=M
zj>r^_u<!p0JFyZL7)zjfxJz2Gzc`o_xgwLu`E&eY*WK#T+{402U7L^C3cBI}_*_N(
zFOr&{ZGV_jozATfJ)L}KJq5j^R70u}qbMEFR*;I(mo6Ri!STbCXBs(^9C@G;C^ggP
z4-Ox$$}cE*xCZuRvbdyg0DUJyf;^<W-Pnm`I7;*4FdC0?ZQZvo&5JV3HZCf&edh@Q
zcjwvV*#;x=>;P}&e$eOS@zmxV9pBBnCy_aM-03xZ<DWO1cw)SNLcbTai|eh>E0BH<
zx%1OMKjO@q|D;TIqf3Xiu!LhCKm&+zu=1o<=0r>F=}}<5B{y--KY+v0;bO^sD5p`h
z#E|S>#Iu>AaY=Ch+vi>C=6g!#>yxDcMU$<OY}b4Kp+<5)nTb^t`8jPIbMpPU?1)iu
zC&D(187npuAH7}>qx7tyxJih&g>~D(-UUaw*RX#0?vx>20TxPFbL00XX-9HSu|vYp
zxxJ4MvNjsw<~gxaWOerEDS@7)`p1P@U*+^xzheR$PyS)g$(i+=*w_a%EZ&??bf*a>
zCRgX5Msz{`>fRi)Z;u`oez=3fVtTuc%Ky1bOe&6V)hw(_DKU?23*0JQi2W=ze5>|{
z4*ve`@PG^MZkDg<&dm)7Cnt}G<55h@pz2j?ZvEC-h{^l2Q~cHC@+_n`Lmj8>8<FEy
z_ERs>|A^6vC3F#M$`|ysI1t>x%dr3Y>^~EC<>WhagA>EEcxVb}J!aP1OqOi(iQU=|
z|7NC3?3BybKJ`n{!-E`5K%yQ%maB}>X{^_9T`*+QNE!;_Z6PHf3nNJMoS$#%)}@$F
zKI>6o5lv}y8_1h>Zh_jz$NL{p7B#;a_dwqCtGexmKNX$FJLZHC1WpDxz)i$l341Ov
ziyU}<oc8>!?UJL(8E<p$=b7O{Zf;1BTNUd8!KueS8os49sABP1NPM{g45-iQa=bIh
zfE%{4wxuIk_(%}u>g3%V08`iiog$ftims#%0~8S38@FEw2<Ivmm%J}0CpL?IY2S}|
z1Sf_{3*WaZNFSBUC=;XF6haUJO;+!_3n&w@y1$h?IMmn`<xm)Y_Eh{srSq$rbeZFE
z%fG-U@zO)mPyOW<-a69$Y6CXe^@@y{SpIy)I8``WKX6>BdVW>KwedL-v)0tZanw3Q
z4n4-aL6D~KIM_(RxXPZ`h^9DKT80q(K98U7O*{?I=F!^~l^n|$vv>0NGNhdh<Noid
z?}yilkE76oq#IXu+ybzq47tz;pMTTPhxKqyXfsb&jGf^Xjlh1+4Wy`bt~|p$bv@Zv
ze<9~;adPBg4O@l^9I>1%YN3y1rY=-YhNd6hFlno^R%hVd;T*|7at_C>&_D5_)5tgk
zpfsTyL>0Y$*j`*F^uA^E!nQACCE7KmUb^$c{W9x~qZFtKp_7GIblvprHU4MJ#Xsq}
zoh7}kp#GM7WGoU8?DmH?-}rZyEA6q#R4p4e&^+nsJ-D=S6%-%1lnF{4?1cA;YYPer
zbd|LgzPni^9d;yLb;?~Ya7l^Ug!1C2$~ttyx(Xur&!u|`dP~Ovx-2xRF@fjFa`+5B
zfj&GDkq8sDb*ol@A11ebq$@I*X5nIWlNWtoIog^(xgYlfv^~n%wTFaeX#74zoV#8o
z*#!Z#P2b3WX3i}j8B_ps?49AW5)b_I;C#+o+uD!+Vt+UrGPK5KS%Ak?gzxk%LR~O{
zb}wGXOXu5dg^vm94>*1+?|){0sEy5lE<~Wd#w9(<cn%+I3I<bJuk{<#zZ!6*N5cf9
z2eq>Udj)_l)1iKm=R^D0)v5ROy*tm(3mM@oUY8pwie0ZFcM_CiPO|NzppN`tEMgBR
zGz#9)JGh8-r@NqsTHK6w;U$cObq}b3*OBjFOOa8U*lj1>qBb9D*HCXQzDItUwf|e7
zSn%>e6-9V}VgAj!z32KlzJF=$jna2tNcyfn4dwKFC{4mrm>_q5{{U<teE6!SY-6{F
zLdYb|Nc{va4WeO&7i_ZNuvAdwezs0{U{d|3E}>u-xu;SHwjVy&?y(Xd>3MB=xw%It
zl1k+g%?VQq3s9d2I4pISCZ{S7WcZ}$<<KKalT><_hW49vfX1boGp%#j61ds!59X!r
zPY8MjX$Z%<0ivLI+2ctT-bzL-UeooiX*$GR9F~0Lsc(L9dShjPI8M_&c#COEzkHRM
z?u+GCGh!YSssH5TD<#XA8YpS>@qeb)OhKm>LZGg;OiJh6py0ek)FO@Lrc~a1Yy{A#
zCmdsWf_lcb5_mR`B8Gu^+-hZPzV_4?oX^q`#NsS=a>v_2-(Rju-FtHy>;Va8U=1Q8
z8vE#bHwkWkSb?2JJP-~Bb$N+UP07%g7}J4lnb!xc*S2o(26>KfJzr6@5%0Slpe^s(
z6uORAu)g&ek)ZHc2dTmnAn45Y?U|=1L<fEB=?;|5-S=mZle*Zd{+)bwIop%{QZ|~u
z@*}<2w(S?fN@6q%PA;7v^z$fQR?I*AiOZ^*A~&8oYu1ZN`@g`-)_A3+ot62w^JCST
zMTZ9kOEOtAp5d}MBr^2N<?aL&`T52}EF16xkqTpL#4;}KgpuoY&{F^dF@7#^xeJb6
z%2!-Y20b9M`eo#-Ap3^JBCZnbSzf}e1~0e<WVhv9Cs^}!;&r6aaWKYONPCm3kKvOh
z_#%;!<oa49J!x={XHW9^#9~T??f{kXH(w8_3R?b5tlte{sv(>O?Hd;X70+d!f%l>+
zcjT4Y4r`VB`@s6gJ|J!_`<3aEn}EHL3^3MW+S{W;r~6BN;3rlY`fup|y1^{tn|b7`
zvVN*`=`y4St#iEj?SLwGvsTl0J_N_)lRj?yB#r~Zp!dGxwvL#%i9!eW<lk?yC!50h
zi;NQydK*+aArhs(TU;zCu?vo9*ASat;9S*xzn;|Gyp=7|9X$RT<ClS>?Qy4KE$?`a
zJ?I~Rfi(W9V}K}>$c;dgmltO+D1sgPO=y)?&62|EqplxcmSJZp27kZ>Goy^T&Q9S{
z&bSgvqGj!p09q<<6V9FBM@;GB?s>pfi`TmrCJPk~z5pwI8sqE=r5ov2U2;yAds$y~
zj^q_~t?@jL<C4T$Z-l)x!8Q-u#ztvMQi&U998$e}3>2}?niN&(_Pa~AYgj%bZ9P=A
zx8aMpxC<MOn`7nPuYQ9={q(eZWce1j?RK9nf$tXScM`i5#WXs^4=4Q5w7+`kakbKU
zJs*cI`seCxv%JzA-v^z{c=rLShnBm0B|9+_SYh-RxL@A1;W|sL9Yyt`l_67KH9s~U
zV(K(7zhH$=#84DM-|Lcl%~~ft?Q2a9@p+O&!z*8|Zs0l8v=%&o6_@N_t*?M9x<ymU
zp}ITZ-$zh1mBUO_)+{_R0q7GK_5<tGpTbV+T~6EPYd9oKTYoOyAcGRa%G8CiE4xrB
zi1adM6vOLBY&LNlFRtMsE&k8I9wxQF!iP`!G--nUFs`(i_s_AXAr@19320;&o4aNP
z+U)E6Wh0J*K2BE9BTY%RSvda_to)6YD_&Y){k{fLEtY4aD?sQ|nK+<K`~#*-<1mEO
z?z#7eK~M8uD52a5ZIY~||Iq&J=WqPv2)K%S0*9V!rV%T6T2C~4{*aS9+wF0ky-y%k
zQn}pvxBm|5GL*jIjYyHAT6g^zVBc6j*TFFP;*(Xzof=Wb8N1=yawUJ;1@d`R#eKYy
zscDJ1Jpm9{UiphRkDKYU7*nPc7M6OCon81<G;<iHhqpZaW<~)^kh=D4hhTf;>+gFC
zQ@!8?#b<g|GQrE6f5SL`_r2&d>N))d>v~FMK9UxHuk_G9A?Hi*8;Jy=YITFRqcvaL
zby^ij%nHBg^S*+Tzghl_yMME+vMTQnGa1eP7`Lz@`=5vT%+1Vi?wLB=96tXoL@{T8
z@Lu}OzSlilT2CRe-teCM|8QiFFjJoqPTRpR*T+j94|4n`PmG#}$G#O6%@PLF%UO5y
zSl=(xf&Otkff4iCd>k~6wv2ICe+89XDNz+!0)VD<#fT+Qt^o)t?-zX=f$;O0o>h6S
zXi-@!|Cm$OtnNMA*6W9Ri@z~ncxum_wxnV|rQfz87TtbV=P5_A3|UR!aeF^#3oeaP
z&sbmi-v2`6)1U|X^It;eM(pSZg|-9}_p!N#^L~6;+0VgTc3LAz-8V*x#|(x)^x*$-
zjY?3pxBtw3OhhBlMs`tnbde^6ceZ@*d_aFk>~(KR!sT3b9Ol>0rfpB7jJ#jDALS&V
zFCOr^Y9_APzh~@o0{o0G(u~pKFe`YKh^yoz-e9fcf45!2;BTYRT$>v%*CvWo%QskJ
zsVr=xL>j}JWon$C(z<C2%~Y7d#)XME28m(XAR~`Y4|7}|UHsu(eofzI(bHk~Pq2QD
zifiiX9^+LVJ503poAa|v@bA1Ma9r?qPZD^p-v7p7x8J&KOWqDvjY3bMtNtkd8rN#X
zb%VmEDZYV~fA+IJOGGCu7onuxKp<bo4FjCoo)WcWLt^=Bn!umR{@)XZZuc>7FFf>u
zM@*yrvE|F03NV&Iq_3~aV&3)B36U5*JJzd?q;s*L4VHjvGw^adto{@!|2=$<5%u<5
zx4$>FP5QgcKDN{ck*acvH|n59g_bp9ieq?;oqNmxyC!;)h{N&m21$Gq_`b)Ekf2E7
z8lYC|<L;vE+7>wwxmx@36i|`D1@-D?D^T(F_l*P(Tm3n)S|f~Ve<mTy?rQGCw&mPT
zEX`@*p<hzYhiBdcmBnw;zurC)W9H9VI5xRf`DsRsVibq<F-#ry$K<33Jy*$aJdvJe
z7_)d-y(+q`{VQ)P(<m<(+lI7{``PC}RY(U^h~soa-nfI=PND?`STc<26eB+iN5iTR
z6Jn>jmIME7ocv1TdoI`v6#W-tt3b){c;%78&;H$Yqv@LY&relD{rOL2KTRZRxjm*z
zc&D6fN_j<d?!B{S%05ngQtO+)KIlYWPg4AsA1AlJ5s{0jBqhsk7^%Z^v2C4ewl@BE
zH7b?PG_}3a@>9>iy`1XF!ivG?jTi4k_T=~r*%0&j1E9@4(+%e((D8P*r12_)*{=XA
zb+|H9*D+jtJYSWBdS9T2_U)?O=Ug@f$0rt{FcR)C_v@I$C=%wKG8po~<e3{TkH@TM
z`5&VB{%}(0>RQb8is$6Yg>R}cCuLlBYwa}QN%J`2Q;C7?ch5)Pq%Uf3YSYsn)z^|#
zUadgO5f31O%D1VRPP%{o1j@N@3%>d2XWuUVOM_4AK|Y%?s>tDaNv~&pfBXHZ-6Wj~
zo<dJWCGvluF5wl9>ci^FDS<pb)Fj-04zVA;X}#>QPkUbKcf7TJib!$kB1|OKUL_S4
z{glfV#XT~gAKV^hpHQ>YJ{{G4v&L6UJhCxJ`Bg?-(2UuC4Cuu3;F>5*7U%0edFL0N
zpMzB@25QK+0HPs{ZIe{El9Smg5U+yN_H|bR1G<TxVOa=1WvH5)%l&kFbu~!NK{qjx
zs_y9)^)Fi~oHk?}^)n|fk=2Q6TV6z>?GqTfjg$m8{56-h2gOnsv0Gm{&Ja+8S4m09
z7D(p_mCg`EGI~0xl?jsi(&_Kyl?FV;8F+^;1Ww%Rx*#7xGi4RClZfuvb2&!_XN>1j
z@WVC5ub5wy)qNX%?|R!PpHs;3dtm3}_K=*e)LNZ_k2O$bekz_jt%}SjVeAL>o}DI(
zFw6>tLdF;|eiLy@RtS<r1c^$&rDFnU`{zAjH#>NE30St2Zo{k*A9L=1NaT(YOn4Jh
zW>;O(Td@-RJaOmrQNnL}$6a9%Z}=tgfF`iz+v3S8>9w+T&J&fnWxk7?t^XzugV~N*
z`{evhW7rQ<zn%vNtlmxN5bl0RjJ7wg#)^9<Aa391(Mdo4McJ`_kk<ow+4udH{BX}6
z#3FNtAhES}Q~x^-Yy2%y65+o$9FJAP8rQkmR4;&&U5gIa(#p5Zz<G-Z8o*sYd}nu;
zpRKj1O~Sn9UP-{liPB-$&VkI&?6n}Ue$i#)X;Ii_HP+`<!>)k~=r+|A;i#KogT8qr
zCBK?S!H>Uh5QFoYSba_d8=*wSeap4KvJj*M=>*U}hPp2d=<{vTJ)t<p1a59xpHFSJ
z1TSWNZBHm&U~l0qwnTgW1H4^t?n0k^Hx6jr31vu({+;$@tSk`8=*lI$9Ibe>6`!qG
z>ZRv8jL|x@EPqyH)_C6fpC+wtRa-PUFHY|M^j*9IO)#9APMPD9KnN5@BCucKyF2dX
z!2N9LOI#xoeOz~ZJGmcwOWZn^9Z%-v$6#{*T2BfTnf3^~*!Jk0dPy~!lS=J)I9CXm
zX4f7iM1`vsF(DpGW(QmT_}uv~n$ELjRbsY%!S*qp;bW_<kqx8Pc0a;(%*m67<CrA_
zelO?WD+m9UAz`YD(@v{4ggHeNPTR7esHHWG?v-L<hDw4YHjRbfl&?OvhSDDHWms+K
zLMLpGgaN`drn(b7%7P-f_OG*-*BIe(l4fV#jkJ`X!{ul}t4&z5%Os)(6LP0+#C7PW
zh)e&=jVljG%W?rGTu(T*{-LAn*5DWAcRaq3Q(A#P?IT6!cv0^63(5G!8A|Ds8$bVJ
zPylaIi*W?Q>)5V`KPQ}VDEErX|JH}1!E?;u;#twfm^Dk0o=&|qI>umr@#_^l)qku8
zsiF39e|_dc)geQc&U);a-7nyCC+GLMQ`AIE+JAzNbUAx_1ar*`OEJ_hVc>*Pc>q1l
zl)HE8T8Q=g8M|nWPU>qNV-or3&{4U@Ct8z)b|xDq^t@@JUB_4TRmmGY@Gh90ryoqK
zf(?Q07jZyW3fmhv)OW$Z;}84{=aVN5cM{J{;kJJqdBtpmi9*!YIlb1?NIe7B*qRGn
zN*QTex&<fyZ{?$F@3@a2=uy^?JgHW;9V+n}EHlhErAD34bF{o6;yd0B7Q}c!bQajq
zK)uy3Y@sGAi#=!5%f9*v+)$R~Fa%l_%VW$QK}+%c;%O!0v5?)0iE3op&EK(`fv`;j
zS^^9Ney0gS+vr38PwbQ6hyJLX;4jK$Uzv?$FFRf-zFd`Htw}T3G&=$h*?3WcwiI5M
za8GG8H#fpZFqZ{a!L$y;it++^IH~(Zhab96cU98vJ~}reMmO2|K+ggDB%ePvY~o5Q
z-7+`M2Liv}>pZ7X4v9+P`W+*6Hsg@Kjo~LG^Z{}h(K{UTO^clJ^wg512wWEd@9c6O
zfG^g0vmVT43?ge1+*b)hIgTsAx64ea$}ztXiHN~*<q#vofwKtTIgTF4g6}?Phk1!*
z-@L4km;|XE@lJJ(xohgV3A@0j6+!V{(95F#|Ec(Iu|FxY9e+V<&s1XsAH#m~J%LhT
zNclpLWmN1b#N;CqTLNe?^o`j)2<%)u@CoCP4W`0I-wnkqpw?Zahp!K37LkLLxq@gG
zTrZeWvf>|1-VtV@4kMSnu7AjngHk*G8I)#+a~Jop#U3Y|lpDN{cSCaWI3*Z>j_a|<
zeEOx7#<_t3N4v^%%)dU-x~qpC!yZt0!Cg0tM^xe|jGEZcaVdBs&Yv(R<CY}P0gmaT
zHl`Di3Fp18o}CuFR@7_cY+88&J_I`-NDC)3V;-jOrRYnNIkCG3T@n@EytI#0zi$M+
zdOX@|gv$6VW?!^x!r*)GGK4#~0xOxWnI*TP`r(=(d|fGaYz1pV{;#^Pe2X&Z+Xm?_
zmF^ahZe@uD1PK9=?(XgeK^g?4TM%jK4gu-zl<qER_8so$FL<Bh{_^l`j%DH6nc0c+
zcSga}Kg@#>--3%%Pckt6DUb*JjFw9M=B}qV{fHwW&OD_w^FF*ac0lc)Fou`QMIn4y
zCWW{?!>~&kIBX&z2P1d@-olJ?*y^gcZKGXYusilNi+sPdSs(drC2ao_owsmW$jP;?
z`8UUF%odUX+KKI+Yq<7L(D%%X?8Nr_S+Y6Y2U^Rn&x2TP!zRAZ&MLBGN#Y(0svEo~
zg!rH79Ds=^PS1~_;r~I2stu0@!@CPldc);nI+p+NXT>vnt`8O)R+H4RtNn{%;<~kf
zgMB16U}YqbRCpy1rhC@ovPD70Ai#ywnkiI&zzAeuGlU8pT81Wlei%0a;JSMA4awZ}
z?MBLe_|RwG_wseF8O6=N??<lY%qY);DDT^!P@TT*8C5|l7ld*-v639-+yKCtZmRuh
zyVRHhQsPnoX)GC-wxm!?gd_)J5Hh%olg<nF{%NpYYOJwx+SEP*A%{oX@r;Umz?Ly3
zMa0K5qx@WZS$KC?QAd|jK>8;v_`uAWXGlZMP()p5spj!pPk+tmCLF5jS9qCkQMxs=
zf^eH*B$}!waBmzK5M2*u$5?e5dD!{vTkXbjgVP9xM>s-&@6f{D)uqq*zpQ|Avvh}$
z8z9rN1war)IyR832C&jpw6(3bKw$)P2Ka320}y_mDoTxSm=?!#0H{CYoB%K=+>_`O
z#@&7_D=Rzr1Udcn+-CzYz(~E*!iQlaOJp=Nd=DLwb8usk2S&RY<)pCI1hP4+aY@)o
zYU=Frcl*YUBa70m$0P)*<ZAG=n~)I-TZQB5^4&>qt!){#M4<>%cm;t~#Pjj@>{GkQ
zQvJ06XWIhtsg8(}eCWTtZydRS9OoR^$RvjmFiv~X5-nYG$hJLRmJb0t;pWRB6Mjcc
z-Y|WO89JN(OAjzK@)i~r%72u)A+5-vugU>yJW2AjYV*53+iCWr9L}wd+tMc4z^ymM
zcO^*UE#!U`XUV4-p!I?%p5T;Xx;2<$+U^g(1!PdJS#__ENMV`jcd<K^&lyA(v?3Vi
zF8n+UapDnT)%MFfQG8q^&oCXvBC)nE%eGS{gbvqjDYy_+J+>CJ0ScGe0L<EkKyla7
zCH~|%nBXa}Dh-HI=AYpGGFwul9$rA$e#!tz1)9|qzzd#U7+Ub#0_yOdEg-;f2?=A0
z8i}`@z25UvnCy4)1Zd76$f6;sAy|P3BTloF)puwE0YM+<jFQLxWMv#A$?txhatIKg
z*SXj($5J9GxY=C1ebci(O5?{$b_8^LcmuO_tG5wb<u7`!pD-ItHC19b$W2V`Ry;Cn
z;Lw5#=3#DC+wg1TK?uvEmQ&NHt-?FPVj{0W(51&`TyaizF_cK#{V7kZkasP1`Lj93
z?b5NY+3SW4e8QIeVdQN=XctJtDt?L#Nuw44>^8zxcP=HLg?c=(U4T40xpF*uZ!uZ0
zeCyhbS_0tp))VXo_*kU$RZFDL$qK+?HFLj4*k(^_@{qcERk-~JwjhimfFYn{oUpfW
z6G8ICXk1<rhEBIG;<-@3&%qyY_O>oAgF|u=Z_gMdOHT+w`|V|ZPydI+pRl$%<5Ul?
zvAp_gV6;O9_ZzVSK6O5^UFH*S@d)-Zw{VU5;hQ~YLm(N0CL^F=J0mmCq~$IXy;BuA
zX-C)su-SZgzRp-sRM1X96sB)GAkOc91AWik^ZvpEs5a=ZJ~7nTYW@^J0WSKSK)!AD
zQNf@85m-4i-E>kuu0*2i5d8g_Cyev>X8c)%MA*bYW4WCb>I%oxCT<F5$b2|m>g~$S
zJw?w=I|7dL)it1uTA2)8Uy0gnH13x<A<h&9#;MapH2g*8tc!;AmQ>@n%cEnMJWe4q
z8H*!px0rnQu-;nf8cJ+^`$hf&jHtr&k`+cCyT|gRvztj;8-HK|i#-{Y=Tu!-h_`8|
zWeNrN$Q5#|=N@qBH31n9)lw6$b|AvoxXkyUyt^V?_#g*I9fdk?mcG|{bKaeGdEQGK
zQy-}zwB1uq8ozRv4obe!ddj3$-IVNPFVkzCk9V^m-S{bt<vK^k!;<dbz}I}R*B=!j
zs)9Cf3^6<KWiDEKw=7m`Z7$LHZ9%4I5iMJ=?b_JTI=~$GK`qZ|KGje}(iSR8#yw`{
z5~VPS3rt8Swxl7qO2OL&)FghnIs&)=+ME8QLp?>G5hv#Fu1PtZ(FS1(dJA$)w1{DH
zjS32ohl>rOzuLZAKd0Y;gpR=H(mq9gi+l8@QCNnl&zWw)>SJ7jHoDkvs!-SOf6Sg@
zxn{$=Db~FUh@Y626Me|<U|y2H<eKlzUp)X%1YXF)uHwG@hAVKPXv7qi=U+B|^+fo0
z)5ON&RX)r57L7XoNc3xu1om<LGc9sh?1j6C8T`HA4xs!u07F%>FL{kAIH7?xUw6-E
zLiyAzVHP*<7`8AU$D{JM%AKUTy<jGfU^Sm=5I{1m;=Pq1KnN69t;SW%K`Q5?p}DQy
z)hG+ABqF^6JdG{OwVPC=lOE`WEcx?RNxsM^e)Q^^-(^Q*-=%|Ckf7z<#{9SUR+jfF
zSd)srX9P2zspT4Qe%LE)Pj$hvOkD!A7)QM_ayUN<=CQA;7jL4RTVN}^#rOQV#s=c|
z(%AlFaMVT#j9Unil}0@Bf$9W~ot0zCg!H7b`=O{6_?2+Hy#9GaPd2gJavN#zk#XdB
zcc;UuE;MyT!|*jrjG68o6)nZ2#9LN#i0kkcaR(py-jk1kUG6C;Lx?9|mk^t-#w)t&
z$F~1C%CMUaCOoZH&?}?c9muc_jrmPo#{G=u;M0~ql$*BoPby2l37(-;r*_Ids^~Ff
z2uBPLQ?`tjhA-FJCOWTu3=egTs5}DPU(vfHMVxwnr1gyg^I;Dh<&8Y|^RZafq(g<t
zh+f5N8{UokU}(&v4^`n!S)})?e0P)>%o#xBUcG)j*H5u|JO8!9KTo!ndWq93MX&pD
z99F&q&aAP)w*L4m=oh`!0Txa0x!HfP8U0BGk{&uj?<OxUzaItwF^xu?b%J8y0<bZt
z=aOWqqe*0%>!Qi~`N4kasQFwWNvF_m3GW5`_ecbPWbvKt)4imNIJ32PVy`6S81=ZA
zd>?R2<KNOue8cLA9{B>NTo#}b9B%TCN+|)C#mo1|91jN0Sqv;oaEaz-eg2&hgvnU7
z<$QUh?N#Ty16Tc)4fB{|SHNb}XF8XT{Zw{|GpiB~MA*w#c9X5uJ4{|c2<hs{Oo*YC
z-ocYnwK*W1Hs3d~Os6ST@9`{2=4)20%MK4aaIoVm;gyKkZaSV2RQfyGF}BidvSvMr
z%;fVSap02_c18>^<lm(0+%6}_7Tm^<vjoHC>tlXLYKM|i^Plr=uQI=X2CjhHgXMX-
zhf$quO+EtuYbgvfDm5uYK|SO|W9})1Ne55@+`6demKj)wE&*O$*)W;m$9p?=vYQ_b
z)`Ncd-!m}M^PJ4PLg72@b1|~^o$bnoNi_&e={VlL=Am)>>bHjYm*>&PX>XKpXf4Vr
zgW$y);&o&YJJLoF0isu?2Oc$D<A-Z8l?Ft542bC)x^<*S9;@U5^-uIVuY2Kfxn!?2
zrpuS!_`SX1=Kk)S|Gq|bQ%9OsCZU^a&cC<{k4>|hiREi3zb}!dXA{U<T@m8;Rhs7e
zO1jOl!R<*4D>4!DYcQ<!aClUATAX25$L=|!6MMx&Dn5A+)R022rF~WruFpeQc`t2w
zXt<a?DoVnxvuZ;&I`g@$DQ_6gF9vhO?vu<LGkU*SE9;ni&E1f0ORVz1QY+KmrDzbc
zCHoL9yg}xu6TUdvwz+i~Ot8KIiFyEQs4+Xn=%6b?g7uvKQTi-|)(C?ol~uG>rSGxk
zon&s00s|Fkr_8;>I7`&5f{ioR#~1SBi74#UpPN5XL|lm~)+x*%5{NjWo0F&77)cAW
zb3w>&LNsa%CwXexmhpEGPkf>J#+0Lh<v8AOeh<WPT^{Gda=i|B@LYbGZ#x#>o)XN!
zcfN@PSY?VqEn;_ryqfIkHWme(MqVQWVjPHwW?v}uJQB=;@BOp=-MLe65PH3>Q>@m(
zbp8WLvIx{wa4XTE{+cg%xwF`^Hn*$@^TI-NyOYg(uV&4-!!M41e{Ms<cxn8=O)G-0
z={EA>6c~#|-p9Thhuub_QKp%Y&S(^r_FBJU=U(8ovB(H~u~jd|o~$%YYz`jY-Tq<6
zX~ZpGxUH5cg=6u{Ibq{gZzQf{2UXjUkm6*ld>~?z_qovdd@kDWj2D>*LHa7@=)#0b
zJbTI2_ha*^SldJ8zd@F53b4s*8f$M5u;pTc!i=-Wbs&l#P5nCH+jmv>v%sRrsPgP^
z$S%zPO751&TqPoU_fPUx(GL;V19ko~9<8x=&bPB58cCh@PjxC3Rt*WBj(cc6hrP~x
zI<gN{+sxAxF4ohpHp^F~1h8~Bujl5)n)9VbKl$gRFWR&wKFi&8<9bP5<e;>jIM|zh
zSv*M2clPCcZ4P#ijk1?@-pUE?hB72+vH1NU9d{d=W~pK-r0Lrk8YkB8N*pmvDus}g
zn7Cz3$FW}mM4kpqBw<&kHZfXQH|1fyprDHw|NRCoa?0E5)cAb--r+R{E{TkC&(@FO
zU}}2&6ke-O3N&ndA7}(E`)fEKCrP;?QP@nV-!6!t=e}P*O_9K7`%rN#fO=S}!ZEie
zn9jX6MaRkUve#DH8`bP*nJi6?=w=K&BxhK^*;m`#w48;LFvfG@`NA+78Jr!|5Ut0p
zbo1Hbn@5QPHocF(Z%PNljNZN244p@oN;r=la@ikYyjlP;m6e`t9@2&d8Lp0+WYu_I
zYMQ!ZEUI6yXxYs;8_<$kv7v}^)RbcRCq2MSv0M-fogZq)ob$4%64?;e_)&FZZZ@C{
zbK&oD%Y*y}K@ma2>a*zIpgBa4X@}{U><zueA<Y-Rx3bxvejzC)0DqWTZU6P4oY9X0
zY8_#FMx>@UGb}v~f%ZL>I!#!`Bfpg2bfyUW#iy6n+@1ZzcbSz+yy}rYVRvAk(mH`+
zLoZ0G+)a()Z!qBUk}9>K5x9Sn2fR)rCiOc2S2+cIxGM4oTBSkS_y<(X+QZa@Ch^UF
zK&=yZse<Fn81(jU8zlbf`Yu+cu#iB0`xzi6@&X)npf98!ju=y~>`Tk_6|KseSFOaK
zRF0#atU5Urg#7827J6jIL*&x*{oFZcjr5-%@HNn+R<<If0wH|w!mArq&5)7kOFe3b
zYUC%jzLET^gd%S?%+Y34pfMv|8)cLI_Bcn6ZV<7Pb2Y{?;TDgXhA}sYH4fDxNKZ%X
zKqdo>mK@1j)kPMXts6)aTK;oSo9u{$Gv-^?<(W%J3^swE&6A;|#eU8;A*=$d4JKH1
zn+43Wi1En`-!#xTd#uOYQslD931B+#-G|J~yyEq3Z9dy+qq!d^5gGrX-R9iItT*#7
z-+xrTvGy8Yo3uAPM~S&RXeUc~8kDleS~W1L8p1z%uySG;+003uzh6L8*?Jdbu%G#9
zIP2lSY|9@3yZ7dndZPQXFV8CIQe1`!G1T@eeD-5j6sA8!`Xp$jX+=8(nWjWbmYY8c
zMx{KC5BATRI3yzFHzRMjt=vDnZ1=heJ>c-=Cy@V#%c@<^A;8vWw@Zv|;4J;6B^HTF
zw$J1xoHPRS-w&q*k+>oI-}5IXq+o?JcBB0sTe`)ED>W2coV7O`B6FlT>M$rQAxV7t
zqXvy*y2<`y+pj#xc001SZzY0_4Vp>zQ2HC5%VH{}5_2tPiyLcshZ5_kI17QD?4ZDD
zMGYu-AYV`HI5M$+gVVO@A@}u+elu@=98*|Dk;}2wgPQMoG%cL%zUTiO4r>yTvQ^Q=
zR~n(ui9PSc6lmlvDp}E*LH?y^irPBSGNvH<3wHw*p`-fsLEw!1t!f90Sc8i$2P%BL
z!#ZEzZ4)@etuIOdZ)Z1mcu^0V5(&)|-=DAkHfHpc^fFyI8YP1n6sXlvKl@>vNP!@w
zc%W34rpO`oQT+0eGu6hXa7A(Yl-%$={4TUYu0Il~yd*ioDPR_72I(}5<1}5o_@trV
zBu1!a)|KT&UlEzVwG7<dB~&W5<WIEl5z*ih9ITQ?tz2W3vYMnn;pqfE2~XHAg{{lX
zR-Gs(pJCoh8fzV!;I6}0a6v1vp4QV{$+k5l8DEQ!j|PuCPnu>;R>%L?{8Gspa96n*
zNFS%Az88}Tw@?|0Git)~FFb#BA5c&5fnw51(N9=hhPs;}$s0}mA5qKb)s@090e8s!
z{3&YziCUuAMlUyBs)j`kylyf-gZaa?QLsdZD5Ku9f$3X=!}IxSOGmq4?1*P|T0^M0
zyAQR8zf_HNQmPt2_)itk$HlW11W!Uzc%PGYGH7?d=M=D4oQ)=Ym&Qr|bb@yQP|;c?
zTpAlforS1Zf?}BA&S>h@#Iq`D<u(4;*eJw~na*kZ=wE}!AH)gE!Nu@SC@B8WRbUw9
z(9l-`r_=zpL_+-|tENl}J;Caa6*xlHdc*M?OK2`O2T+rWf66OUhD|mRYd*KU@ZO^=
zxQyX$t$h_%TJ%EjjQ`O3-v)Vc0|%W9DyHz&WPP9d^}puSlB}KOI2}|Qe%AM)6l0es
zAyL_Ga!?rc{+ILgs9JG9NRIN3wEY+37x$c8X~v%{PCv1w1(Lz#Bv!$p4pQ;%(+6@l
z5K=cOT@4^p`ePv`dbUrmKT@8JN@+EUUPhVIbe(wovcM~yxv2l%1sO&NlrgaJb9gqH
ztQs*?&fv2Z>NmB7&=8W~xHb3j<RQMT_a|MTGO33DXy&2yk0$u9w8Z|Sp-EP94SPxP
zw`ozIxO0BBe`ImJP-5B)BJF9h(<^3WgCEM`#)R19`*zvLSHeB=_H}do=K7qxJAyEC
z{m85*fMOq(;TifUrrL=6$SdfZ`_EeL??n%}kO%imW~C3c_7gHwR{(+KaCwEd*p2@^
zv&zQgm@Q?CFQKws87`0hw$=gB(o!z$8Rxs48ktb-(vsEB_s)m2SApqCKe;Fh7k^HB
z{UG)DXa;B2;YRcsdMGmGxBnSueSVvG@)wH})4P1%N^2=Du%D_bVF#=psZuz#P~V-0
z$kL2K=tD<c&|@nf$M>kZ_PZ-c1lx2;{pCba@un$-^V##lg<bYGOf^259YUM9fR&=v
zKcJ;=zxB8$%Bs0y3I946E{bSjYT@8`@20AM7>`tJi{cNl)m;a%t%}&p=&IHiRU|}?
zQ<`0Eo-w&LuMeEs6eQlZh5q~TIZR}%Xi%qm|94p#*?S45GX=O3b42Z()q)l#0!O@&
z^R!oN`b0E#xUamhd?VF+DNC|35Scu`^k|}uA!xes+|?Y#E*g)M^D7DCVNRej$|bK{
zqddB!kKebj2c5hvIRuhf_QILWVWN5qmw(4cRpKJ-?8z&h6IEjVmX^zeW`T0;8p3|d
zI?Y8;)m)XUUTsVgp57s&-hZP7;h0PrL{E-Jf&f(M9M&+5qIn*ax{SZS!NbftD7#r7
z=A+YSFuZ#Y87^ePNuB$!g2`u+2Q*iWnYXx%RE@6p?KyxE#ghz3wtmhKZIN5Z5!s2|
zHEqE`eaxKM!eP8gPXF**dSSAw$A`P7PwjGLVCAr<qVU|IF2KM~MLOL~gJ;=F-S27H
zo-Y2n69a&tjbI^ti~*!%3;vg=%F;G%UYZiAve6xA`X%h)zDi}HLhB<T;yT~VX>{0T
zS=y*?&+_PTzcyadI<7ujX+9BLl)Hvo|8g#Vn7OXk=xpK0>fl_Z;DUa?HMFJpD%s;Q
z^C8`q_M%5|=|<}PqE(8(w6(4T&HQoA{4m+9X%CfI=z6!u>JxE|u#A*8$>->zjyqR?
zU|lKmn(S)Uu`dFnn;y}q(&=w+Yb>vZh00EW+dO}-!<khAZ;Ut<&8L#ApE)IiDHo0n
zu}T`}DWP27=p*lz*~~WB1{NIhY40ctFW%N!*hL@yGW!M<^}DxcPh{5vK3K$Y+I71}
zs&@Bdbt|{js8T4>>P;*?<l7qQ{AavsumbD+90({uc#DbMf6_7d-uvE7n66MdOCfc_
zF7(}x%9K(No*hMUiRY3)wm~_@^b=^Ox1L~;b%_bdpv|x$n|f7FZ`<K_jHX>bI%a2r
zNvCH~>nHC=rNr~ZBr*uyQFh8EyrS@U!VH*+#Qh?b0%_tO+(m2@Cr2D8xR%-l|BaMx
zu<H=bEh0W}V%FZmZ68_FYFaT#lv8WK-F8YxlcijOO@D?66ZwH-qx<}=c&MM`YsYjW
z+X`<S!hEyS^gsBAzrM3ku16}|r9=;DHSoSz6R7wk6F`d#W%lPCx&301)o2^qOe`BL
z_uBg_h+*mb)-0ay`_|4l-Q}c=_^!-JJAJ_{21%S5?=rZFEMYJ<f1T5Q`78W2V9hO_
zesV7Nl<xM!K2oCbi}d5cNc<ZRu^i*3b%eZ`=&I50iRPitC*(b<W~&{<j5O<<OE0zb
zYX)!YYWX-mVQ&y>nLx1)w%+0U^DLNi;blKj4>Zs^1Y|vx3vR?pJkwY?y}jw2Ku$(s
zg)eF`alDKA+(w7*nZGZ%wGr*bh$X$_{~Y!g*beanBe{GQGu)|F)Dz;L8I=Z3B^QBS
zt>QZjL~MoGT98Roc+Jgh<h|}?O~xbsJ4&;oy>S+gVezsh;MKf%y#`FLgosZ~2AEgh
z;!<|Z4H$pG?2g17=?jAG(s^^+++nZ<a1sV#j#r^PaHJIW884~GK6wTXwg(_&6q(4#
z$SC`!U}7}GlSzlDR3sKPxhjy}#_hIQn>uRJZJVi-3?7)1CyjV&8w%B#XH4&&A4gO@
z$?a1Jb=`mK-1UowMDS8(j1zet_~MA|>>aF@GgRt=Er{BV0Dj*(=An3Ln6p4Yf&}A8
zF~$}hB*1ZuKw!<l`sG)eNC+acRPekoFH|#=3MWn$Cpw#;=j^)yjE{0bo<o+#u=B5e
zOp{XtpQ-&ZOZx147rry5*~u0@cqM|GiGDH^0*>-b*SF7icJAyJsy(K$Qo`g~@5Fju
zZ$;T|BV{4LZa~kPB+aQ;8p}jw`2RfZnFCBjJt+j+W&*bY)jKciZhdL@g}Cp%6&zzP
zPL4Z=`OJ>0{c6&e^Tw+1cs1y<F7=n75W62%J(L4>KYT4H($6;Ch7o#nEz582me`#)
zn_`V(D7d2s0Oy`R>4Ew6YgenxO9iRO5+wY#>PXN0Zn<tArw5{G=Qx)YrYatl1#l7l
zz2TrG{br43>b4+AfwQ17!SfQ6nH0_C5>%K33-uHQf<z@h0^hUL#uRT8mEUGKe)PK7
z99ne&H<;auHm|CeVVRj@4rY~{W*vz!lBge?UxRO?y?2)9!JN~|JG7(*kt^zHJequR
zM~d8d;=cZNxvPVty#&PE?kt%|zlVK2^7XeW>#&d)UgD?a8e#?tv^^bAx}`Bo4Da!E
z!z!(TT4Lcg3&|2Lp8K)$lYppfc9P4GrdFh7jVmP68v19CvqG=gDLg-EKC_eDCqnT7
zlF`=)A)N{#*p&>te(C?@j#LR&o32Zc$X^*B)G}Ryt-IHWk!tLPsO$q67w~z1;3@4%
zZSO~B3M6Qt^JI>HM*3?c3gMm`hX2Xxb94u?Opb}m81@jO(Iezs%nH<|dMF$hTWJQm
zs)~CaXIPiiQ_q>0Y;pNgs!GMKdveSek`Ec!2;22=FyYo-8zU60!$G4U884q4&-T+s
zCg(a0ggQ;V1fo)Q-XxPGD83;+sVUp`?TMW-KK%GaefL-p(NBT=HdhNK&!QnLS$3gf
znJIkj?3A*3^7(Ih;ngcx$Vj^n$=j_x62EKo$cDtlay+yT9G5>>;FR%{#(JimAh{2(
zO|NX`kzA1TYOm-C!<RnfTCY1Fy1uYs!t@m@&X)hidh*Qe;2~*<3pVlN`<&YX_Ukwd
zPX8A^A(LPFL`^C<PNVZ4%BxBh_ie&*`INJpg=(f(=_3XeY8kxMn=;l((2$TUw1is}
zo{cR7b=nYc4dC8$iqUlwV&!<uA&W6DCJlsaN4XEx-hZ1WF?u6HUzI}KW?D$!S;o&e
z#}z&hZu~RmL%d78<$c5Uu*=|$>d|bGrf><HketH7_1gioS-$#wPR}9bKEJ6JwcjJ9
zr?7?eF+S37X^u%$U;uOQaz>Ip&xC6?cQ(kcYIiqmCb;rg7Eo3EUWN~aMV&0st7K>z
zpkk0Ah*w`6#W9=WEKs=kyC8D-m`sIk6|r!wFMrXvjD5MvVlQG<^*A>u1g|%18#K(e
zkQ`*KmED(JjIc&{sn^fHgYbEW7CO;2843J5WRm`kgdQZ?Prjx+j~4zXpqAVGa>i${
zMg?14u*W2>Xdf||B+Tg#G3@7^j?gW++CNKjT(Nj^b<-+10Sw5lx9FO;scZR<JB?5t
zUeWY@Sjp*fna2AkdZBpB6v{6)*MUQ+>e>FNOe3v<wJSHV(s%CeYh+48`QMinpkH>)
zLcF$vvEr2qt5L<nJ2_|ZlmGOFoE9AJ#w0d(GlcGz9_BC_Q`sl#YI97>h+Hk&)@6@S
zX2{3D7nH`o{EaKR2Q@hbQEB$ZBcrK)JeP`T#>=}1@^_|e`>Zx{5zHZV0&9o~o9>6+
zY4FS(S(<VC8i~A+!P@4KS6dFHTF?1vJSqNco&3uf(k-3e_HHS_W;8Ub|J~05o6_w(
zWYtiz03Rf<a2X+JvDtqrLTCB*>%yUBt-#?$N4hP83EsM2nLLujk(KQLYEGd(W2hnQ
ze2naS(A~kWq&rsY@Me%h<w~AnzQ0<!4&hQ9BH2rLoNYF8>D_J*`yTW76T2yGNvn3?
z-3A&XoDa%gM|HT^!Nn#SY^8xo02qUXDJ+H<ywB+mWuV`TYG%8=*XCF0!7}=bre5oi
zA^7{pR2m&wuXW^jsr68muf4CYE$xnrYq#W&(%j58?%r2N8D>>*3w#!zSQPx136f8)
zPO8g`-u{${MWEVXX7m?1$XqzyyN0ut-Rw5@F*>#Iq{z}-y26ld*^O)~Mo^>u)~gpQ
zLT<L-D5tP6)1QB|Wa)Kvi}V$*w<2(FlNbLSBi=&GK4YmF1^{HMnWFs*Gw0CZJl+8t
zkdG)MuEVhD5B$sorRwc0f*3`$h}!mnw6XzM0b~f^FW<>{q>#XfYjqCfNH5Fkdgghk
zrFCYuu^CsaSV}H(T;x`i53B`WAfcUR<Y}oPNOL8I-Zn&q<L8&9b^GX(XVa)nun)M4
z=T#8=E8$2~<xjVJakYcVBu8c^Cic(C?eRLb{=pzK>Jjp*X9=42Ly6E|pX8GnoQq53
zZqOZt)$6%Kzjzl&xt(5oDSe_L^f==_#lv&!LlU-4D$<(vBC4%p8m)Io=Yk6#q^g76
zJF`^k8v|coAP;g@L1LPK!4M=#8R2{m79VRdRDjycy=XUI?A1-9_&0eSb9B1FdX6gz
zZDrdS0r#I+cgnHPzP|bA1Fb`Ws{$OaKCC+nt1l{>DosQ)?ol>X3x5;onCNwCywoP9
zgb)e~?~0BBl8@2De9|=5=f?OXa<`W?T0`6Tcs0&-aRSf&7{Q$4Han^#1ZH`i5Dojv
z{O$C%kDYpw^%QRwGZ-&#0hCN)??pw5Q7TpEwrkL+zI0Ri{mMStpZ~`uXZ#{+q26Cx
zdc9?xH$!;s=mg!=T0J9H6*3MblQI?#Z%URFwqp5&Qxw7)_C#K}M2r5elkhfVMqW0`
zH;MGK`JSbxz7FZ6@bhk0WM!AkOwKxCUm$T$wQ(s9n~f^W5fSG(+upyS)vKHFwp3mn
z-dJsCjV3Up((s>fU7T^Q+DDQ07yCHI^?~(w+lpTW^roQle6R)4;#iF3?7VU419iMs
zGJR{PUD(o}aHF?|86xTM*N9UNi!6<5@HCJZ1=#%k&@f~6;%IeZSlQDR{8QAsMW#vz
zrSAp9?%RJ`(x2&_L)%))&CV7(0>f*5t3+88Mh!3PaV-_LvMv4+Vm>Dw2*8U>Vq4p@
z2yNqlvuHA98)PqWetocec;Ka`I<K$$Ov0_Tn)TAb6@L|~rK>NBmx8``c%yk2=_cM4
zb<~C}JmE;wn)H<bmptUT&1s23Q(C#(I@3c}Z9@(o<u})1wQq9E27v^9m)%5}HhmRs
zi%oS+D4H)-GapTrX}3uYSE-5}+V2C<TpxgjX_bsy@n3|ApjSoTp2uile|t8{KjCg2
zH?pYk^u)>c>SHa9W37Lmr5)}G+hlu~{C1a~{&Vfc2nbA6Jb~75X{v>#Gve5E1or1w
zjSHcT1D`A|o%LybmZ#MylkbC5T3Wk^M@?&tq3dt_-K&H3aXX~R(`>9R5nSyjfUm~R
z(XU$JrYPTZdNFJ^U43aBcFy&~A1(IlBy|>5Y?ZrBc>S092p9Qe(O2}vByP?0Ql@)|
z$*hx6oarhh*#zr8crun+G00tA8NC8A9m@!%tZHyWXqpN*Hcj!n27Q&)C0zBE^)Jqd
zlhpYK$ySazdgTPI4Pe+kf@vvb;bagphrjVoa>DGNXN4OtoWs8iJqxa^S5u<%HtfbF
zyeSXPGmn8J^*mLj_5x_zw*qe#{@601%*&sw@MUO+KZvGL*u=#;xNI<6d2{4iM?%M-
zqfn_gJyB<+w6vPBZ(1FzTD_}STvj6W7jFc!?1Gkd=B{suHG<p|sS+dO_DHi3I?vY(
zrqOnOshynUr5e6>e0@*Gc%f2h62zNu=>-hZQ0%Evq<?QO{E$&Gs<~Ly#-cG*UT4Ed
zW8i)-8OEC6Mq|nC)l;)A;4pb*we;Hp3T%tw34ktd^mfHV>!5?8&NPg8i){H%pVt&!
zx-PM=7$}!6BEqLd#PH~Uuv3|IE_N0oMUHf)kVG>|RUC5?qvLL`NTiQ09DNSO^nFlt
zHuUFX;DX1*+WR(&+H1nl-X`L*OIw=9Z4y5n##a;Kh~s{U)Fu^<zt4~;z29Aa%J$El
zPb&T8`1P2zsL7RO90yA;JeVvW6Kr{g{S$kz%&qsA7+aZN@(~$nX7xZigR)DV?2iX@
zpvf4FR6@9D+b6kWcBweA&I-+0v6Kq$HPGwUo34s1@5PH|lc_$A4inP9#Iw1TPG2SC
zM>XMkbP2CS^oSUaU}?K4ANcG3$ycL#qy_omFpZpel5XO$1^F?H`*DFUySGgw$ZF$r
z#k)I3E{Si;#f@4>CUA&3za;~lT}FrhZ3Hl;9o4=}mz_UP^HZj!{=4}SW+6^Mt*INz
zm$a(>#&GIz)P)Mc|1?Q9oeInc#NQh+vse~4#aBC)p4H1!WV5c)RVwE;#h0GAYYdt<
zCdtRJR9KXX%f;chB1#OfQi$Q6u@Z<@$6Ro2$R`{(s?TlE)-C)Q!p*kQS6kSuFgpJf
zjTtf_T8Bum11d8z_}x;o*f^exN$P@{&^Y%pHK|EcYE)1m66-apNaIp#V|Thmb;c0D
zrwm!N6zS4hMAs=+Z$@&i(y*yQN$X$tJRd?|6Z%)(%FE1DH1&yJ4A&AT^%jr|4X}LA
zHr`ZYr}9=vL6k<WKL2vj;*&--IOXF{*hG&TiQw*Dmx%kgPYB4JfcV)I9j%W3yPS%~
zkRS)1!IYAz|9@WKnZ*<{f#D<meupB8iztTspWg;d)Th1yEilihK<hsb3|e3YJhlGQ
z4d7*f_CK}2ne%`9)~Ejbul%9^qtEbP@6i9Bet4RN|CK=%XIp_GWc62l`Npa40tbF%
MBo&}#;`)LA51<;DGXMYp

literal 0
HcmV?d00001

diff --git a/studio/src-tauri/linux/postremove.sh b/studio/src-tauri/linux/postremove.sh
new file mode 100755
index 0000000000..6314c0f3dd
--- /dev/null
+++ b/studio/src-tauri/linux/postremove.sh
@@ -0,0 +1,9 @@
+#!/bin/sh
+# Post-removal script for Unsloth Studio (deb/rpm)
+# Runs non-interactively; never deletes user data or touches other users' homes.
+
+case "${1:-}" in
+    upgrade|1|2) exit 0 ;;
+esac
+
+exit 0
diff --git a/studio/src-tauri/src/commands.rs b/studio/src-tauri/src/commands.rs
new file mode 100644
index 0000000000..f2acb92e09
--- /dev/null
+++ b/studio/src-tauri/src/commands.rs
@@ -0,0 +1,476 @@
+use crate::install;
+use crate::process::{self, BackendState, ShutdownFlag};
+use crate::update;
+use log::{error, info, warn};
+use tauri::{AppHandle, Emitter};
+
+async fn managed_install_ready_after_repair() -> bool {
+    crate::preflight::managed_install_ready().await
+}
+
+fn should_emit_repair_failed(msg: &str) -> bool {
+    !msg.contains("NEEDS_ELEVATION")
+}
+
+#[tauri::command]
+pub async fn desktop_preflight() -> crate::preflight::DesktopPreflightResult {
+    crate::preflight::desktop_preflight_result().await
+}
+
+/// Check if unsloth is installed AND functional.
+/// Runs `unsloth -h` to verify the import chain works — a partial install
+/// (binary exists but deps missing) will fail on import and return false,
+/// which sends the user to the install screen for a clean re-install.
+#[tauri::command]
+pub async fn check_install_status() -> bool {
+    let Some(bin) = process::find_unsloth_binary() else {
+        return false;
+    };
+
+    let mut cmd = tokio::process::Command::new(&bin);
+    cmd.arg("-h")
+        .stdout(std::process::Stdio::null())
+        .stderr(std::process::Stdio::null());
+
+    #[cfg(windows)]
+    {
+        cmd.creation_flags(crate::process::CREATE_NO_WINDOW);
+    }
+
+    // Match the same AppImage env clearing used in process.rs and install.rs,
+    // otherwise the probe can fail due to bundled libs even when the install is fine.
+    #[cfg(target_os = "linux")]
+    if std::env::var_os("APPIMAGE").is_some() {
+        cmd.env_remove("LD_LIBRARY_PATH");
+        cmd.env_remove("PYTHONHOME");
+        cmd.env_remove("PYTHONPATH");
+    }
+
+    let mut child = match cmd.spawn() {
+        Ok(c) => c,
+        Err(e) => {
+            warn!("Install check: failed to spawn {:?}: {}", bin, e);
+            return false;
+        }
+    };
+
+    match tokio::time::timeout(std::time::Duration::from_secs(10), child.wait()).await {
+        Ok(Ok(status)) => {
+            let ok = status.success();
+            if !ok {
+                warn!("Install check: `unsloth -h` exited with {}", status);
+            }
+            ok
+        }
+        Ok(Err(e)) => {
+            warn!("Install check: wait failed: {}", e);
+            false
+        }
+        Err(_) => {
+            warn!("Install check: `unsloth -h` timed out after 10s");
+            let _ = child.kill().await;
+            false
+        }
+    }
+}
+
+/// Start the backend server on the given port.
+/// Also spawns a health watchdog that monitors the backend and emits
+/// `server-crashed` if it becomes unresponsive (deadlock, OOM, etc.).
+#[tauri::command]
+pub async fn start_server(
+    app: AppHandle,
+    state: tauri::State<'_, BackendState>,
+    shutdown: tauri::State<'_, ShutdownFlag>,
+    port: u16,
+) -> Result<(), String> {
+    info!("start_server command called with port {}", port);
+
+    process::start_backend(&app, &state, port, &shutdown)?;
+
+    // Spawn health watchdog for the owned backend — detects
+    // deadlocks and hangs that stdout-based crash detection misses.
+    let watchdog_state = state.inner().clone();
+    let watchdog_shutdown = shutdown.inner().clone();
+    let watchdog_app = app.clone();
+    tokio::spawn(async move {
+        health_watchdog(watchdog_app, watchdog_state, watchdog_shutdown).await;
+    });
+
+    Ok(())
+}
+
+/// Start the managed backend without reusing an existing backend.
+#[tauri::command]
+pub async fn start_managed_server(
+    app: AppHandle,
+    state: tauri::State<'_, BackendState>,
+    shutdown: tauri::State<'_, ShutdownFlag>,
+    port: u16,
+) -> Result<(), String> {
+    info!("start_managed_server command called with port {}", port);
+    process::start_backend(&app, &state, port, &shutdown)?;
+
+    let watchdog_state = state.inner().clone();
+    let watchdog_shutdown = shutdown.inner().clone();
+    let watchdog_app = app.clone();
+    tokio::spawn(async move {
+        health_watchdog(watchdog_app, watchdog_state, watchdog_shutdown).await;
+    });
+
+    Ok(())
+}
+
+/// Stop the backend server.
+/// Sends SIGTERM to the process group, which triggers uvicorn's graceful
+/// shutdown (same codepath as /api/shutdown). Falls back to SIGKILL after 5s.
+#[tauri::command]
+pub fn stop_server(
+    state: tauri::State<'_, BackendState>,
+    shutdown: tauri::State<'_, ShutdownFlag>,
+) -> Result<(), String> {
+    info!("stop_server command called");
+    process::stop_backend(&state, &shutdown)
+}
+
+/// Check if a healthy Unsloth backend is running on the given port.
+/// Expects JSON response with status=="healthy" AND service=="Unsloth UI Backend".
+#[tauri::command]
+pub async fn check_health(port: u16) -> Result<bool, String> {
+    match check_health_inner(port).await {
+        Ok(healthy) => Ok(healthy),
+        Err(e) => {
+            // Network errors are not command errors — just means not healthy
+            info!("Health check on port {} failed: {}", port, e);
+            Ok(false)
+        }
+    }
+}
+
+async fn check_health_inner(port: u16) -> Result<bool, reqwest::Error> {
+    let url = format!("http://127.0.0.1:{}/api/health", port);
+    let client = reqwest::Client::builder()
+        .timeout(std::time::Duration::from_secs(2))
+        .build()?;
+    let resp = client.get(&url).send().await?;
+    let json: serde_json::Value = resp.json().await?;
+
+    let healthy = json
+        .get("status")
+        .and_then(|v| v.as_str())
+        .map(|s| s == "healthy")
+        .unwrap_or(false);
+    let correct_service = json
+        .get("service")
+        .and_then(|v| v.as_str())
+        .map(|s| s == "Unsloth UI Backend")
+        .unwrap_or(false);
+
+    Ok(healthy && correct_service)
+}
+
+/// Return buffered server logs.
+#[tauri::command]
+pub fn get_server_logs(state: tauri::State<'_, BackendState>) -> Vec<String> {
+    match state.lock() {
+        Ok(proc) => proc.logs.iter().cloned().collect(),
+        Err(e) => {
+            error!("Failed to lock state for logs: {}", e);
+            vec![]
+        }
+    }
+}
+
+/// Open the Unsloth Studio directory in the system file manager.
+#[tauri::command]
+pub fn open_logs_dir() -> Result<(), String> {
+    let home = dirs::home_dir().ok_or("Could not determine home directory")?;
+    let dir = home.join(".unsloth").join("studio");
+
+    if !dir.exists() {
+        return Err(format!("Directory does not exist: {}", dir.display()));
+    }
+
+    open::that(&dir).map_err(|e| format!("Failed to open directory: {}", e))
+}
+
+/// Start the first-launch installation process.
+/// Runs the platform installer script with --tauri flag and streams progress events.
+/// Returns "NEEDS_ELEVATION" if system packages need elevated install (Linux only).
+#[tauri::command]
+pub async fn start_install(
+    app: AppHandle,
+    state: tauri::State<'_, install::InstallState>,
+) -> Result<(), String> {
+    let state = state.inner().clone();
+    tokio::task::spawn_blocking(move || install::run_install(app, state))
+        .await
+        .map_err(|e| format!("Install task panicked: {e}"))?
+}
+
+/// Install system packages with elevated permissions (Linux only).
+/// Called by frontend after user approves the elevation dialog.
+/// Only allows packages that the install script reported as needed.
+#[cfg(target_os = "linux")]
+#[tauri::command]
+pub fn install_system_packages(
+    packages: Vec<String>,
+    state: tauri::State<'_, install::InstallState>,
+) -> Result<(), String> {
+    // Cross-check against the packages the install script actually reported
+    let allowed = state
+        .lock()
+        .map(|s| s.needed_packages.clone())
+        .unwrap_or_default();
+    for pkg in &packages {
+        if !allowed.contains(pkg) {
+            return Err(format!(
+                "Package '{}' was not requested by the install script",
+                pkg
+            ));
+        }
+    }
+    install::install_system_packages(&packages)
+}
+
+/// Stub for non-Linux platforms — elevation is handled by the scripts themselves.
+#[cfg(not(target_os = "linux"))]
+#[tauri::command]
+pub fn install_system_packages(
+    _packages: Vec<String>,
+    _state: tauri::State<'_, install::InstallState>,
+) -> Result<(), String> {
+    Err("Elevated package install is only supported on Linux".to_string())
+}
+
+/// Run backend update: stop server, run `unsloth studio update`, emit progress.
+/// Does NOT restart the backend — the frontend handles shell update + relaunch after.
+#[tauri::command]
+pub async fn start_backend_update(
+    app: AppHandle,
+    backend_state: tauri::State<'_, BackendState>,
+    shutdown: tauri::State<'_, ShutdownFlag>,
+    update_state: tauri::State<'_, update::UpdateState>,
+    install_state: tauri::State<'_, install::InstallState>,
+) -> Result<(), String> {
+    info!("start_backend_update command called");
+
+    // Signal the health watchdog to exit immediately, before any guards.
+    // This closes the race window where the watchdog could emit server-crashed
+    // between our command being called and stop_backend completing.
+    shutdown.store(true, std::sync::atomic::Ordering::SeqCst);
+
+    // Guard: reject if install is running
+    if install_state
+        .lock()
+        .map(|s| s.child.is_some())
+        .unwrap_or(false)
+    {
+        return Err("Cannot update while installation is in progress.".to_string());
+    }
+
+    // Guard: reject if update is already running
+    if update_state
+        .lock()
+        .map(|s| s.child.is_some())
+        .unwrap_or(false)
+    {
+        return Err("Update is already running.".to_string());
+    }
+
+    // Stop backend if running
+    if backend_state
+        .lock()
+        .map(|s| s.child.is_some())
+        .unwrap_or(false)
+    {
+        info!("Stopping backend before update...");
+        process::stop_backend(&backend_state, &shutdown)?;
+    }
+
+    // Run update in a blocking thread
+    let state = update_state.inner().clone();
+    tokio::task::spawn_blocking(move || update::run_backend_update(app, state))
+        .await
+        .map_err(|e| format!("Update task panicked: {e}"))?
+}
+
+/// Repair a stale managed Studio install.
+#[tauri::command]
+pub async fn start_managed_repair(
+    app: AppHandle,
+    backend_state: tauri::State<'_, BackendState>,
+    shutdown: tauri::State<'_, ShutdownFlag>,
+    update_state: tauri::State<'_, update::UpdateState>,
+    install_state: tauri::State<'_, install::InstallState>,
+) -> Result<(), String> {
+    info!("start_managed_repair command called");
+
+    if install_state
+        .lock()
+        .map(|s| s.child.is_some())
+        .unwrap_or(false)
+    {
+        return Err("Cannot repair while installation is in progress.".to_string());
+    }
+
+    if update_state
+        .lock()
+        .map(|s| s.child.is_some())
+        .unwrap_or(false)
+    {
+        return Err("Repair is already running.".to_string());
+    }
+
+    shutdown.store(true, std::sync::atomic::Ordering::SeqCst);
+
+    if backend_state
+        .lock()
+        .map(|s| s.child.is_some())
+        .unwrap_or(false)
+    {
+        info!("Stopping backend before repair...");
+        process::stop_backend(&backend_state, &shutdown)?;
+    }
+
+    let _ = app.emit("repair-progress", "Updating existing Studio install...");
+    let update_app = app.clone();
+    let update_state = update_state.inner().clone();
+    let update_result = tokio::task::spawn_blocking(move || {
+        update::run_backend_update_for_repair(update_app, update_state)
+    })
+    .await
+    .map_err(|e| format!("Repair update task panicked: {e}"))?;
+
+    match update_result {
+        Ok(()) if managed_install_ready_after_repair().await => {
+            info!("Managed repair complete after update");
+            let _ = app.emit("repair-complete", ());
+            return Ok(());
+        }
+        Ok(()) => {
+            warn!("Managed repair update finished, but preflight is still not ready; falling back to installer");
+            let _ = app.emit(
+                "repair-progress",
+                "Update finished, but Studio is still not ready. Running bundled installer...",
+            );
+        }
+        Err(msg) => {
+            if msg.to_ascii_lowercase().contains("already running") {
+                error!("Managed repair update conflict: {}", msg);
+                let _ = app.emit("repair-failed", &msg);
+                return Err(msg);
+            }
+
+            warn!(
+                "Managed repair update failed, falling back to bundled installer: {}",
+                msg
+            );
+            let _ = app.emit(
+                "repair-progress",
+                "Update failed. Running bundled installer...",
+            );
+        }
+    }
+
+    let install_app = app.clone();
+    let install_state = install_state.inner().clone();
+    let install_result = tokio::task::spawn_blocking(move || {
+        install::run_install_for_repair(install_app, install_state)
+    })
+    .await
+    .map_err(|e| format!("Repair install task panicked: {e}"))?;
+
+    if let Err(msg) = install_result {
+        if should_emit_repair_failed(&msg) {
+            error!("Managed repair installer failed: {}", msg);
+            let _ = app.emit("repair-failed", &msg);
+        }
+        return Err(msg);
+    }
+
+    if managed_install_ready_after_repair().await {
+        info!("Managed repair complete after installer");
+        let _ = app.emit("repair-complete", ());
+        return Ok(());
+    }
+
+    let msg = "Repair finished, but Studio install is still not desktop-ready.".to_string();
+    error!("{}", msg);
+    let _ = app.emit("repair-failed", &msg);
+    Err(msg)
+}
+
+#[cfg(test)]
+mod tests {
+    #[test]
+    fn repair_elevation_is_not_a_terminal_repair_failure() {
+        assert!(!super::should_emit_repair_failed("NEEDS_ELEVATION"));
+        assert!(super::should_emit_repair_failed(
+            "Installer exited with code 1"
+        ));
+    }
+}
+
+/// Periodic health check that detects deadlocked or hung backends.
+/// Starts 30s after the backend is launched (to allow initial startup),
+/// then pings /api/health every 15s. After 3 consecutive failures (45s)
+/// with the process still alive, emits `server-crashed` so the frontend
+/// can offer a restart.
+async fn health_watchdog(app: AppHandle, state: BackendState, shutdown: ShutdownFlag) {
+    use std::sync::atomic::Ordering;
+
+    // Give the backend time to start up
+    tokio::time::sleep(std::time::Duration::from_secs(30)).await;
+
+    let mut consecutive_failures: u32 = 0;
+
+    loop {
+        tokio::time::sleep(std::time::Duration::from_secs(15)).await;
+
+        if shutdown.load(Ordering::SeqCst) {
+            info!("Health watchdog: shutdown flag set, exiting");
+            break;
+        }
+
+        let (port, has_child) = {
+            let proc = match state.lock() {
+                Ok(p) => p,
+                Err(_) => break,
+            };
+            (proc.port, proc.child.is_some())
+        };
+
+        // Stop watching if the backend is gone
+        if !has_child {
+            info!("Health watchdog: backend stopped, exiting");
+            break;
+        }
+
+        let Some(port) = port else {
+            continue; // Port not yet known
+        };
+
+        match check_health_inner(port).await {
+            Ok(true) => {
+                consecutive_failures = 0;
+            }
+            _ => {
+                consecutive_failures += 1;
+                warn!(
+                    "Health watchdog: failure {}/3 on port {}",
+                    consecutive_failures, port
+                );
+                if consecutive_failures >= 3 {
+                    error!(
+                        "Health watchdog: backend unresponsive for 45s, killing and declaring dead"
+                    );
+                    // Kill the zombie process so retry can start fresh
+                    let _ = process::stop_backend(&state, &shutdown);
+                    let _ = app.emit("server-crashed", ());
+                    break;
+                }
+            }
+        }
+    }
+}
diff --git a/studio/src-tauri/src/desktop_auth.rs b/studio/src-tauri/src/desktop_auth.rs
new file mode 100644
index 0000000000..df09df7a14
--- /dev/null
+++ b/studio/src-tauri/src/desktop_auth.rs
@@ -0,0 +1,388 @@
+use crate::preflight::{DesktopPreflightDisposition, DesktopPreflightResult};
+use crate::process::BackendState;
+use log::info;
+use reqwest::Client;
+use serde::{Deserialize, Serialize};
+use std::path::{Path, PathBuf};
+
+static DESKTOP_AUTH_LOCK: tokio::sync::Mutex<()> = tokio::sync::Mutex::const_new(());
+
+#[derive(Debug, Serialize)]
+pub struct DesktopAuthResponse {
+    pub access_token: String,
+    pub refresh_token: String,
+}
+
+#[derive(Debug, Serialize, Deserialize)]
+struct DesktopAuthRequest {
+    secret: String,
+}
+
+#[derive(Debug, Deserialize)]
+struct TokenResponse {
+    access_token: String,
+    refresh_token: String,
+}
+
+#[derive(Clone, Copy, Debug, Eq, PartialEq)]
+enum PortSource {
+    Cached,
+    Discovered,
+}
+
+#[derive(Clone, Copy, Debug)]
+struct BackendPort {
+    port: u16,
+    source: PortSource,
+}
+
+#[derive(Debug)]
+enum AuthError {
+    Connectivity(String),
+    Failed(String),
+}
+
+impl AuthError {
+    fn message(self) -> String {
+        match self {
+            Self::Connectivity(message) | Self::Failed(message) => message,
+        }
+    }
+}
+
+fn auth_secret_path(home: &Path, filename: &str) -> PathBuf {
+    home.join(".unsloth")
+        .join("studio")
+        .join("auth")
+        .join(filename)
+}
+
+fn auth_url(port: u16, route: &str) -> String {
+    format!("http://127.0.0.1:{port}/api/auth/{route}")
+}
+
+fn home_dir() -> Result<PathBuf, String> {
+    dirs::home_dir().ok_or_else(|| "Could not determine home directory".to_string())
+}
+
+fn desktop_secret_path() -> Result<PathBuf, String> {
+    Ok(auth_secret_path(&home_dir()?, ".desktop_secret"))
+}
+
+fn read_secret_if_exists(path: &Path) -> Result<Option<String>, String> {
+    match std::fs::read_to_string(path) {
+        Ok(s) => Ok(Some(s.trim().to_string())),
+        Err(e) if e.kind() == std::io::ErrorKind::NotFound => Ok(None),
+        Err(e) => Err(format!(
+            "Failed to read auth secret at {}: {}",
+            path.display(),
+            e
+        )),
+    }
+}
+
+async fn current_backend_port(
+    state: &tauri::State<'_, BackendState>,
+) -> Result<BackendPort, String> {
+    if let Some(port) = state.lock().map_err(|e| e.to_string())?.port {
+        return Ok(BackendPort {
+            port,
+            source: PortSource::Cached,
+        });
+    }
+
+    let port = discover_compatible_backend_port()
+        .await
+        .ok_or_else(|| "Backend is not ready".to_string())?;
+
+    {
+        let mut proc = state.lock().map_err(|e| e.to_string())?;
+        if proc.port.is_none() {
+            proc.port = Some(port);
+        }
+    }
+
+    Ok(BackendPort {
+        port,
+        source: PortSource::Discovered,
+    })
+}
+
+fn attached_ready_port(preflight: DesktopPreflightResult) -> Option<u16> {
+    if preflight.disposition == DesktopPreflightDisposition::AttachedReady {
+        preflight.port
+    } else {
+        None
+    }
+}
+
+async fn discover_compatible_backend_port() -> Option<u16> {
+    attached_ready_port(crate::preflight::desktop_preflight_result().await)
+}
+
+fn update_backend_port(state: &tauri::State<'_, BackendState>, port: u16) -> Result<(), String> {
+    let mut proc = state.lock().map_err(|e| e.to_string())?;
+    proc.port = Some(port);
+    Ok(())
+}
+
+fn classify_auth_send_error(error: reqwest::Error) -> AuthError {
+    let message = format!("Desktop auth failed: {}", error);
+    if error.is_connect() || error.is_timeout() {
+        AuthError::Connectivity(message)
+    } else {
+        AuthError::Failed(message)
+    }
+}
+
+fn should_retry_with_discovered_port(source: PortSource, error: &AuthError) -> bool {
+    matches!(
+        (source, error),
+        (PortSource::Cached, AuthError::Connectivity(_))
+    )
+}
+
+async fn exchange_desktop_secret(
+    client: &Client,
+    port: u16,
+    secret: &str,
+) -> Result<Option<DesktopAuthResponse>, AuthError> {
+    let response = client
+        .post(auth_url(port, "desktop-login"))
+        .json(&DesktopAuthRequest {
+            secret: secret.to_string(),
+        })
+        .send()
+        .await
+        .map_err(classify_auth_send_error)?;
+
+    if response.status() == reqwest::StatusCode::NOT_FOUND {
+        return Err(AuthError::Failed(
+            "Running Studio backend is too old for this desktop app. Update that backend and restart."
+                .to_string(),
+        ));
+    }
+    if response.status() == reqwest::StatusCode::UNAUTHORIZED {
+        return Ok(None);
+    }
+    if !response.status().is_success() {
+        return Err(AuthError::Failed("Desktop auth failed".to_string()));
+    }
+
+    response
+        .json::<TokenResponse>()
+        .await
+        .map(|tokens| {
+            Some(DesktopAuthResponse {
+                access_token: tokens.access_token,
+                refresh_token: tokens.refresh_token,
+            })
+        })
+        .map_err(|e| AuthError::Failed(format!("Desktop auth failed: {}", e)))
+}
+
+async fn provision_desktop_auth() -> Result<(), String> {
+    let bin = crate::process::resolve_backend_binary()?;
+    let mut cmd = tokio::process::Command::new(&bin);
+    cmd.args(["studio", "provision-desktop-auth"])
+        .stdout(std::process::Stdio::null())
+        .stderr(std::process::Stdio::piped());
+    #[cfg(target_os = "linux")]
+    if std::env::var_os("APPIMAGE").is_some() {
+        cmd.env_remove("LD_LIBRARY_PATH");
+        cmd.env_remove("PYTHONHOME");
+        cmd.env_remove("PYTHONPATH");
+    }
+    #[cfg(windows)]
+    {
+        use std::os::windows::process::CommandExt;
+        cmd.creation_flags(crate::process::CREATE_NO_WINDOW);
+    }
+
+    let output = tokio::time::timeout(std::time::Duration::from_secs(30), cmd.output())
+        .await
+        .map_err(|_| "Desktop auth provisioning timed out after 30s".to_string())?
+        .map_err(|e| format!("Desktop auth provisioning failed: {}", e))?;
+    if output.status.success() {
+        return Ok(());
+    }
+    let stderr = String::from_utf8_lossy(&output.stderr);
+    Err(format!(
+        "Desktop auth provisioning failed: {}",
+        stderr.trim()
+    ))
+}
+
+async fn authenticate_with_stale_port_retry(
+    client: &Client,
+    state: &tauri::State<'_, BackendState>,
+    backend: BackendPort,
+    secret: &str,
+) -> Result<(Option<DesktopAuthResponse>, BackendPort), String> {
+    match exchange_desktop_secret(client, backend.port, secret).await {
+        Ok(tokens) => Ok((tokens, backend)),
+        Err(error) if should_retry_with_discovered_port(backend.source, &error) => {
+            let Some(port) = discover_compatible_backend_port().await else {
+                return Err(error.message());
+            };
+            update_backend_port(state, port)?;
+            let backend = BackendPort {
+                port,
+                source: PortSource::Discovered,
+            };
+            exchange_desktop_secret(client, port, secret)
+                .await
+                .map(|tokens| (tokens, backend))
+                .map_err(AuthError::message)
+        }
+        Err(error) => Err(error.message()),
+    }
+}
+
+#[tauri::command]
+pub async fn desktop_auth(
+    state: tauri::State<'_, BackendState>,
+) -> Result<DesktopAuthResponse, String> {
+    let _auth_guard = DESKTOP_AUTH_LOCK.lock().await;
+    let mut backend = current_backend_port(&state).await?;
+    let client = Client::builder()
+        .timeout(std::time::Duration::from_secs(5))
+        .build()
+        .map_err(|e| format!("Desktop auth failed: {}", e))?;
+
+    for attempt in 0..2 {
+        if attempt == 1 {
+            info!("Desktop auth: provisioning local desktop secret");
+            provision_desktop_auth().await?;
+        }
+
+        let path = desktop_secret_path()?;
+        let Some(secret) = read_secret_if_exists(&path)? else {
+            continue;
+        };
+
+        info!("Desktop auth: exchanging desktop secret");
+        let (tokens, resolved_backend) =
+            authenticate_with_stale_port_retry(&client, &state, backend, &secret).await?;
+        backend = resolved_backend;
+        if let Some(tokens) = tokens {
+            return Ok(tokens);
+        }
+    }
+
+    Err(
+        "Desktop auth failed. Update or repair the managed Studio install, then restart Studio."
+            .to_string(),
+    )
+}
+
+#[cfg(test)]
+mod tests {
+    use super::*;
+    use tokio::io::{AsyncReadExt, AsyncWriteExt};
+    use tokio::net::TcpListener;
+
+    async fn login_server(status: &str) -> u16 {
+        let listener = TcpListener::bind("127.0.0.1:0").await.unwrap();
+        let port = listener.local_addr().unwrap().port();
+        let status = status.to_string();
+
+        tokio::spawn(async move {
+            let (mut stream, _) = listener.accept().await.unwrap();
+            let mut buffer = [0; 1024];
+            let _ = stream.read(&mut buffer).await.unwrap();
+            let response =
+                format!("HTTP/1.1 {status}\r\nContent-Length: 0\r\nConnection: close\r\n\r\n");
+            stream.write_all(response.as_bytes()).await.unwrap();
+        });
+
+        port
+    }
+
+    #[test]
+    fn auth_secret_path_joins_expected_location() {
+        let home = PathBuf::from("/home/alex");
+        assert_eq!(
+            auth_secret_path(&home, ".desktop_secret"),
+            PathBuf::from("/home/alex/.unsloth/studio/auth/.desktop_secret")
+        );
+    }
+
+    #[test]
+    fn auth_url_builds_local_endpoint() {
+        assert_eq!(
+            auth_url(8890, "desktop-login"),
+            "http://127.0.0.1:8890/api/auth/desktop-login"
+        );
+    }
+
+    #[test]
+    fn retry_discovery_only_for_cached_connectivity_errors() {
+        assert!(should_retry_with_discovered_port(
+            PortSource::Cached,
+            &AuthError::Connectivity("connection refused".to_string())
+        ));
+        assert!(!should_retry_with_discovered_port(
+            PortSource::Discovered,
+            &AuthError::Connectivity("connection refused".to_string())
+        ));
+        assert!(!should_retry_with_discovered_port(
+            PortSource::Cached,
+            &AuthError::Failed("Desktop auth failed".to_string())
+        ));
+    }
+
+    #[test]
+    fn attached_ready_port_requires_attached_ready_with_port() {
+        let compatible = DesktopPreflightResult {
+            disposition: DesktopPreflightDisposition::AttachedReady,
+            reason: None,
+            port: Some(8890),
+            can_auto_repair: false,
+            managed_bin: None,
+        };
+        assert_eq!(attached_ready_port(compatible), Some(8890));
+
+        let missing_port = DesktopPreflightResult {
+            disposition: DesktopPreflightDisposition::AttachedReady,
+            reason: None,
+            port: None,
+            can_auto_repair: false,
+            managed_bin: None,
+        };
+        assert_eq!(attached_ready_port(missing_port), None);
+
+        let managed_ready = DesktopPreflightResult {
+            disposition: DesktopPreflightDisposition::ManagedReady,
+            reason: None,
+            port: Some(8890),
+            can_auto_repair: false,
+            managed_bin: None,
+        };
+        assert_eq!(attached_ready_port(managed_ready), None);
+    }
+
+    #[tokio::test]
+    async fn exchange_desktop_secret_returns_none_for_unauthorized() {
+        let port = login_server("401 Unauthorized").await;
+        let tokens = exchange_desktop_secret(&Client::new(), port, "desktop-stale")
+            .await
+            .unwrap();
+
+        assert!(tokens.is_none());
+    }
+
+    #[tokio::test]
+    async fn exchange_desktop_secret_reports_unsupported_backend_on_not_found() {
+        let port = login_server("404 Not Found").await;
+        let error = exchange_desktop_secret(&Client::new(), port, "desktop-secret")
+            .await
+            .unwrap_err()
+            .message();
+
+        assert_eq!(
+            error,
+            "Running Studio backend is too old for this desktop app. Update that backend and restart."
+        );
+    }
+}
diff --git a/studio/src-tauri/src/install.rs b/studio/src-tauri/src/install.rs
new file mode 100644
index 0000000000..5760989353
--- /dev/null
+++ b/studio/src-tauri/src/install.rs
@@ -0,0 +1,555 @@
+use log::{error, info, warn};
+use process_wrap::std::*;
+use std::io::BufRead;
+use std::path::{Path, PathBuf};
+use std::process::{Command, ExitStatus, Stdio};
+use std::sync::{Arc, Mutex};
+use tauri::{AppHandle, Emitter, Manager};
+
+// ── Types ──
+
+pub struct InstallProcess {
+    /// Process group handle — killing this kills the entire subprocess tree.
+    pub child: Option<Box<dyn ChildWrapper + Send>>,
+    pub intentional_stop: bool,
+    /// Packages needing elevated install, parsed from [TAURI:NEED_SUDO] output.
+    pub needed_packages: Vec<String>,
+}
+
+impl Default for InstallProcess {
+    fn default() -> Self {
+        Self {
+            child: None,
+            intentional_stop: false,
+            needed_packages: Vec::new(),
+        }
+    }
+}
+
+pub type InstallState = Arc<Mutex<InstallProcess>>;
+
+pub fn new_install_state() -> InstallState {
+    Arc::new(Mutex::new(InstallProcess::default()))
+}
+
+use crate::process::trim_line_endings;
+
+// ── Script Resolution ──
+
+/// Returns (script_path, args) depending on dev vs production mode.
+/// Dev mode: repo root script + --tauri --local
+/// Production: bundled resource + --tauri
+fn resolve_install_script(app: &AppHandle) -> Result<(PathBuf, Vec<String>), String> {
+    let mut args = vec!["--tauri".to_string()];
+
+    if cfg!(debug_assertions) {
+        let repo_root = Path::new(env!("CARGO_MANIFEST_DIR"))
+            .parent() // studio/
+            .and_then(|p| p.parent()) // repo root
+            .ok_or("Cannot resolve repo root from CARGO_MANIFEST_DIR")?;
+
+        let script = if cfg!(unix) {
+            repo_root.join("install.sh")
+        } else {
+            repo_root.join("install.ps1")
+        };
+
+        if !script.exists() {
+            return Err(format!("Install script not found: {}", script.display()));
+        }
+
+        args.push("--local".to_string());
+        info!("Dev mode: using repo script at {}", script.display());
+        Ok((script, args))
+    } else {
+        let name = if cfg!(unix) {
+            "install.sh"
+        } else {
+            "install.ps1"
+        };
+        let script = app
+            .path()
+            .resolve(name, tauri::path::BaseDirectory::Resource)
+            .map_err(|e| format!("Failed to resolve bundled {}: {}", name, e))?;
+        info!("Production: using bundled script at {}", script.display());
+        Ok((script, args))
+    }
+}
+
+// ── Emit Helpers ──
+
+#[derive(Clone, Copy)]
+enum InstallEventMode {
+    Full,
+    Repair,
+}
+
+impl InstallEventMode {
+    fn progress_event(self) -> &'static str {
+        match self {
+            Self::Full => "install-progress",
+            Self::Repair => "repair-progress",
+        }
+    }
+
+    fn emit_install_structured_events(self) -> bool {
+        matches!(self, Self::Full)
+    }
+
+    fn emit_terminal_events(self) -> bool {
+        matches!(self, Self::Full)
+    }
+
+    fn needs_elevation_event(self) -> &'static str {
+        match self {
+            Self::Full => "install-needs-elevation",
+            Self::Repair => "repair-needs-elevation",
+        }
+    }
+}
+
+fn emit_mode_progress(app: &AppHandle, mode: InstallEventMode, message: &str) {
+    info!("[install] {}", message);
+    let _ = app.emit(mode.progress_event(), message);
+}
+
+fn emit_failed(app: &AppHandle, message: &str) {
+    error!("[install] FAILED: {}", message);
+    let _ = app.emit("install-failed", message);
+}
+
+fn emit_complete(app: &AppHandle) {
+    info!("[install] Installation complete");
+    let _ = app.emit("install-complete", ());
+}
+
+// ── Spawn ──
+
+/// Spawns the install script in a process group.
+/// Returns (stdout, stderr) handles for streaming.
+/// The GroupChild is stored in state so stop_install() can kill the entire tree.
+fn spawn_script(
+    script: &Path,
+    args: &[String],
+    state: &InstallState,
+) -> Result<
+    (
+        Option<std::process::ChildStdout>,
+        Option<std::process::ChildStderr>,
+    ),
+    String,
+> {
+    let mut install = state.lock().map_err(|e| e.to_string())?;
+    if install.child.is_some() {
+        return Err("Installation is already running.".to_string());
+    }
+    install.intentional_stop = false;
+    install.needed_packages.clear();
+
+    // Scripts create ~/.unsloth/studio/ themselves, but need a writable cwd.
+    let home = dirs::home_dir().ok_or("Could not determine home directory")?;
+    let work_dir = home.join(".unsloth");
+    if !work_dir.exists() {
+        std::fs::create_dir_all(&work_dir)
+            .map_err(|e| format!("Failed to create {}: {}", work_dir.display(), e))?;
+    }
+
+    #[cfg(unix)]
+    let mut cmd = Command::new("bash");
+    #[cfg(unix)]
+    cmd.arg(script)
+        .args(args)
+        .current_dir(&work_dir)
+        .stdout(Stdio::piped())
+        .stderr(Stdio::piped());
+
+    #[cfg(windows)]
+    let mut cmd = Command::new("powershell.exe");
+    #[cfg(windows)]
+    cmd.args([
+        "-NoLogo",
+        "-NoProfile",
+        "-NonInteractive",
+        "-WindowStyle",
+        "Hidden",
+        "-ExecutionPolicy",
+        "Bypass",
+        "-File",
+    ])
+    .arg(script)
+    .args(args)
+    .current_dir(&work_dir)
+    .stdout(Stdio::piped())
+    .stderr(Stdio::piped());
+
+    // AppImage sets LD_LIBRARY_PATH to its bundled libs, which breaks Python
+    // spawned by the install script. Only clear inside AppImage — native installs
+    // may need these for custom CUDA or conda paths.
+    #[cfg(target_os = "linux")]
+    if std::env::var_os("APPIMAGE").is_some() {
+        cmd.env_remove("LD_LIBRARY_PATH");
+        cmd.env_remove("PYTHONHOME");
+        cmd.env_remove("PYTHONPATH");
+    }
+
+    // On Windows, launch the installer directly with CREATE_NO_WINDOW.
+    // The app process is assigned to a KILL_ON_JOB_CLOSE job in main.rs, so
+    // child cleanup on crash comes from inherited job membership instead.
+    #[cfg(windows)]
+    let mut child: Box<dyn ChildWrapper + Send> = {
+        use std::os::windows::process::CommandExt;
+        cmd.creation_flags(crate::process::CREATE_NO_WINDOW);
+        let child = cmd
+            .spawn()
+            .map_err(|e| format!("Failed to spawn install script: {}", e))?;
+        Box::new(child)
+    };
+
+    #[cfg(unix)]
+    let mut child: Box<dyn ChildWrapper + Send> = {
+        // Keep the whole installer tree in a process group on Unix.
+        let mut wrap = CommandWrap::from(cmd);
+        wrap.wrap(ProcessGroup::leader());
+        wrap.spawn()
+            .map_err(|e| format!("Failed to spawn install script: {}", e))?
+    };
+
+    let stdout = child.stdout().take();
+    let stderr = child.stderr().take();
+    install.child = Some(child);
+    Ok((stdout, stderr))
+}
+
+// ── Stream ──
+
+/// Spawns reader threads for stdout/stderr.
+/// Parses [TAURI:*] lines from stdout for structured events.
+fn stream_output(
+    app: &AppHandle,
+    state: &InstallState,
+    event_mode: InstallEventMode,
+    stdout: Option<std::process::ChildStdout>,
+    stderr: Option<std::process::ChildStderr>,
+) -> Vec<std::thread::JoinHandle<()>> {
+    let mut threads = Vec::new();
+
+    if let Some(out) = stdout {
+        let app_clone = app.clone();
+        let state_clone = Arc::clone(state);
+        threads.push(std::thread::spawn(move || {
+            let mut reader = std::io::BufReader::new(out);
+            let mut buf = Vec::new();
+            loop {
+                buf.clear();
+                match reader.read_until(b'\n', &mut buf) {
+                    Ok(0) => break,
+                    Ok(_) => {
+                        let text = String::from_utf8_lossy(trim_line_endings(&buf)).into_owned();
+                        // Parse structured Tauri protocol lines
+                        if let Some(packages) = text.strip_prefix("[TAURI:NEED_SUDO] ") {
+                            let pkgs: Vec<String> =
+                                packages.split_whitespace().map(String::from).collect();
+                            if let Ok(mut install) = state_clone.lock() {
+                                install.needed_packages = pkgs;
+                            }
+                        } else if let Some(step) = text.strip_prefix("[TAURI:STEP] ") {
+                            if !event_mode.emit_install_structured_events() {
+                                info!("[install][stdout] {}", text);
+                                let _ = app_clone.emit(event_mode.progress_event(), &text);
+                                continue;
+                            }
+                            let _ = app_clone.emit("install-step", step);
+                        } else if let Some(detail) = text.strip_prefix("[TAURI:PROGRESS] ") {
+                            if !event_mode.emit_install_structured_events() {
+                                info!("[install][stdout] {}", text);
+                                let _ = app_clone.emit(event_mode.progress_event(), detail);
+                                continue;
+                            }
+                            let _ = app_clone.emit("install-progress-detail", detail);
+                        }
+                        // Always forward the raw line
+                        info!("[install][stdout] {}", text);
+                        let _ = app_clone.emit(event_mode.progress_event(), &text);
+                    }
+                    Err(e) => {
+                        warn!("[install] Error reading stdout: {}", e);
+                        break;
+                    }
+                }
+            }
+        }));
+    }
+
+    if let Some(err) = stderr {
+        let app_clone = app.clone();
+        threads.push(std::thread::spawn(move || {
+            let mut reader = std::io::BufReader::new(err);
+            let mut buf = Vec::new();
+            loop {
+                buf.clear();
+                match reader.read_until(b'\n', &mut buf) {
+                    Ok(0) => break,
+                    Ok(_) => {
+                        let text = String::from_utf8_lossy(trim_line_endings(&buf)).into_owned();
+                        warn!("[install][stderr] {}", text);
+                        let _ = app_clone.emit(event_mode.progress_event(), &text);
+                    }
+                    Err(e) => {
+                        warn!("[install] Error reading stderr: {}", e);
+                        break;
+                    }
+                }
+            }
+        }));
+    }
+
+    threads
+}
+
+// ── Wait & Finalize ──
+
+/// Waits for the install process to exit. Returns (exit_status, was_intentional_stop).
+/// Times out after 2 hours to prevent infinite loops if the child hangs.
+fn wait_for_exit(state: &InstallState) -> Result<(ExitStatus, bool), String> {
+    const MAX_WAIT_ITERATIONS: u32 = 72_000; // 2h at 100ms intervals
+    for _ in 0..MAX_WAIT_ITERATIONS {
+        let mut install = state.lock().map_err(|e| e.to_string())?;
+        let intentional = install.intentional_stop;
+
+        match install.child.as_mut() {
+            Some(child) => match child.try_wait() {
+                Ok(Some(status)) => {
+                    install.child = None;
+                    return Ok((status, intentional));
+                }
+                Ok(None) => {}
+                Err(e) => {
+                    install.child = None;
+                    return Err(format!("Error waiting for installer: {}", e));
+                }
+            },
+            None if intentional => return Err("Installation stopped.".to_string()),
+            None => return Err("Installer process disappeared unexpectedly.".to_string()),
+        }
+
+        drop(install);
+        std::thread::sleep(std::time::Duration::from_millis(100));
+    }
+    // Timed out — kill and report
+    let _ = stop_install(state);
+    Err("Installation timed out after 2 hours".to_string())
+}
+
+// ── Public API ──
+
+/// Run the install script. Returns Ok(()) on success.
+/// Returns Err("NEEDS_ELEVATION") if system packages need elevated install (Linux only).
+/// Returns Err(message) on other failures.
+pub fn run_install(app: AppHandle, state: InstallState) -> Result<(), String> {
+    run_install_with_event_mode(app, state, InstallEventMode::Full)
+}
+
+pub(crate) fn run_install_for_repair(app: AppHandle, state: InstallState) -> Result<(), String> {
+    run_install_with_event_mode(app, state, InstallEventMode::Repair)
+}
+
+fn run_install_with_event_mode(
+    app: AppHandle,
+    state: InstallState,
+    event_mode: InstallEventMode,
+) -> Result<(), String> {
+    emit_mode_progress(&app, event_mode, "Starting installation...");
+
+    let (script, args) = resolve_install_script(&app)?;
+    emit_mode_progress(
+        &app,
+        event_mode,
+        &format!("Using script: {}", script.display()),
+    );
+
+    let (stdout, stderr) = spawn_script(&script, &args, &state)?;
+    let threads = stream_output(&app, &state, event_mode, stdout, stderr);
+
+    // Wait for exit, join reader threads
+    let result = wait_for_exit(&state);
+    for handle in threads {
+        let _ = handle.join();
+    }
+
+    match result {
+        Ok((status, _)) if status.success() => {
+            if event_mode.emit_terminal_events() {
+                emit_complete(&app);
+            }
+            Ok(())
+        }
+        Ok((status, _)) => {
+            let code = status.code().unwrap_or(-1);
+            if code == 2 {
+                // Script needs elevated package install — report to frontend
+                let packages = state
+                    .lock()
+                    .map(|i| i.needed_packages.clone())
+                    .unwrap_or_default();
+                info!("[install] Needs elevation for packages: {:?}", packages);
+                let _ = app.emit(event_mode.needs_elevation_event(), &packages);
+                Err("NEEDS_ELEVATION".to_string())
+            } else {
+                let msg = format!("Installer exited with code {}", code);
+                if event_mode.emit_terminal_events() {
+                    emit_failed(&app, &msg);
+                }
+                Err(msg)
+            }
+        }
+        Err(msg) if msg == "Installation stopped." => {
+            info!("[install] Installation stopped intentionally");
+            Err(msg)
+        }
+        Err(msg) => {
+            if event_mode.emit_terminal_events() {
+                emit_failed(&app, &msg);
+            }
+            Err(msg)
+        }
+    }
+}
+
+/// Stop a running install process gracefully.
+/// Unix: SIGTERM to process group -> wait up to 5s -> SIGKILL
+/// Windows: hidden taskkill /T /F to terminate the installer tree
+pub fn stop_install(state: &InstallState) -> Result<(), String> {
+    let mut child = {
+        let mut install = match state.lock() {
+            Ok(guard) => guard,
+            Err(poisoned) => {
+                warn!("Install state mutex poisoned, recovering for cleanup");
+                poisoned.into_inner()
+            }
+        };
+        install.intentional_stop = true;
+        install.child.take()
+    };
+
+    let Some(ref mut child) = child else {
+        return Ok(());
+    };
+
+    let pid = child.id();
+    info!("Stopping installer process group (pid {})", pid);
+
+    // Try graceful SIGTERM first so pip/cmake can clean up temp files
+    #[cfg(unix)]
+    {
+        if pid > i32::MAX as u32 {
+            // PID too large for i32 negation, fall back to direct kill
+            warn!("PID {} exceeds i32 range, using direct kill", pid);
+            let _ = child.kill();
+            let _ = child.wait();
+            return Ok(());
+        }
+        unsafe {
+            libc::kill(-(pid as i32), libc::SIGTERM);
+        }
+        // Wait up to 5s for graceful exit
+        for _ in 0..50 {
+            match child.try_wait() {
+                Ok(Some(status)) => {
+                    info!("Installer exited gracefully with status: {:?}", status);
+                    return Ok(());
+                }
+                Ok(None) => std::thread::sleep(std::time::Duration::from_millis(100)),
+                Err(_) => break,
+            }
+        }
+        warn!("Installer did not exit gracefully, force killing");
+    }
+
+    #[cfg(windows)]
+    {
+        crate::process::force_kill_process_tree(pid, child, "Installer");
+        return Ok(());
+    }
+
+    #[cfg(unix)]
+    {
+        // Force kill (SIGKILL on Unix)
+        let _ = child.kill();
+        let _ = child.wait();
+        info!("Installer process group force stopped");
+        Ok(())
+    }
+}
+
+/// Install system packages with elevated permissions (Linux only).
+/// Uses `elevated-command` crate for native auth dialog.
+#[cfg(target_os = "linux")]
+pub fn install_system_packages(packages: &[String]) -> Result<(), String> {
+    use regex::Regex;
+    use std::path::Path;
+    use std::process::Command as StdCommand;
+
+    // Validate package names to prevent injection via elevated command.
+    let valid_pkg = Regex::new(r"^[a-zA-Z0-9][a-zA-Z0-9.+\-]*$").unwrap();
+    for pkg in packages {
+        if !valid_pkg.is_match(pkg) {
+            return Err(format!("Invalid package name: {}", pkg));
+        }
+    }
+
+    // AppImage bundles run on non-Debian distros too. Pick the first package
+    // manager we find. Names in `packages` are Debian-style; callers that want
+    // cross-distro support should translate before invoking.
+    let (program, base_args): (&str, &[&str]) = if Path::new("/usr/bin/apt-get").exists() {
+        ("apt-get", &["install", "-y"])
+    } else if Path::new("/usr/bin/dnf").exists() {
+        ("dnf", &["install", "-y"])
+    } else if Path::new("/usr/bin/zypper").exists() {
+        ("zypper", &["install", "-y"])
+    } else if Path::new("/usr/bin/pacman").exists() {
+        ("pacman", &["-S", "--noconfirm"])
+    } else {
+        return Err(
+            "No supported system package manager found (apt-get, dnf, zypper, pacman)".to_string(),
+        );
+    };
+
+    info!(
+        "[install] Elevated install of packages via {}: {}",
+        program,
+        packages.join(", ")
+    );
+
+    let mut cmd = StdCommand::new(program);
+    cmd.args(base_args).args(packages);
+
+    let elevated = elevated_command::Command::new(cmd)
+        .output()
+        .map_err(|e| format!("Elevated install failed: {}", e))?;
+
+    if !elevated.status.success() {
+        let stderr = String::from_utf8_lossy(&elevated.stderr);
+        return Err(format!("Package installation failed: {}", stderr));
+    }
+
+    info!("[install] Elevated package install succeeded");
+    Ok(())
+}
+
+#[cfg(test)]
+mod tests {
+    use super::*;
+
+    #[test]
+    fn repair_install_mode_uses_repair_elevation_event() {
+        assert_eq!(
+            InstallEventMode::Full.needs_elevation_event(),
+            "install-needs-elevation"
+        );
+        assert_eq!(
+            InstallEventMode::Repair.needs_elevation_event(),
+            "repair-needs-elevation"
+        );
+        assert!(!InstallEventMode::Repair.emit_terminal_events());
+    }
+}
diff --git a/studio/src-tauri/src/main.rs b/studio/src-tauri/src/main.rs
new file mode 100644
index 0000000000..f7d49cce60
--- /dev/null
+++ b/studio/src-tauri/src/main.rs
@@ -0,0 +1,189 @@
+#![cfg_attr(not(debug_assertions), windows_subsystem = "windows")]
+
+mod commands;
+mod desktop_auth;
+mod install;
+mod preflight;
+mod process;
+mod update;
+mod windows_job;
+
+use log::info;
+use process::new_backend_state;
+use simplelog::{
+    CombinedLogger, Config, LevelFilter, SharedLogger, TermLogger, TerminalMode, WriteLogger,
+};
+use std::fs;
+use tauri::menu::{MenuBuilder, MenuItemBuilder};
+use tauri::tray::{MouseButton, MouseButtonState, TrayIconBuilder, TrayIconEvent};
+use tauri::{Emitter, Manager};
+
+fn setup_logging() {
+    let mut loggers: Vec<Box<dyn SharedLogger>> = vec![];
+
+    // Always log to stderr for development
+    loggers.push(TermLogger::new(
+        LevelFilter::Info,
+        Config::default(),
+        TerminalMode::Stderr,
+        simplelog::ColorChoice::Auto,
+    ));
+
+    // Try to set up file logging to ~/.unsloth/studio/tauri.log
+    if let Some(home) = dirs::home_dir() {
+        let log_dir = home.join(".unsloth").join("studio");
+        if fs::create_dir_all(&log_dir).is_ok() {
+            let log_path = log_dir.join("tauri.log");
+            let rotated_path = log_dir.join("tauri.log.1");
+            let max_log_bytes = 5 * 1024 * 1024;
+            if fs::metadata(&log_path)
+                .map(|meta| meta.len() >= max_log_bytes)
+                .unwrap_or(false)
+            {
+                let _ = fs::remove_file(&rotated_path);
+                let _ = fs::rename(&log_path, &rotated_path);
+            }
+            if let Ok(file) = fs::OpenOptions::new()
+                .create(true)
+                .append(true)
+                .open(&log_path)
+            {
+                loggers.push(WriteLogger::new(LevelFilter::Info, Config::default(), file));
+            }
+        }
+    }
+
+    if !loggers.is_empty() {
+        let _ = CombinedLogger::init(loggers);
+    }
+}
+
+fn setup_tray(app: &tauri::App) -> Result<(), Box<dyn std::error::Error>> {
+    let open = MenuItemBuilder::with_id("open", "Open Studio").build(app)?;
+    let toggle = MenuItemBuilder::with_id("toggle", "Start/Stop Server").build(app)?;
+    let quit = MenuItemBuilder::with_id("quit", "Quit").build(app)?;
+    let menu = MenuBuilder::new(app)
+        .items(&[&open, &toggle, &quit])
+        .build()?;
+
+    TrayIconBuilder::new()
+        .menu(&menu)
+        .tooltip("Unsloth Studio (Desktop)")
+        .icon(app.default_window_icon().unwrap().clone())
+        .on_menu_event(move |app, event| match event.id().as_ref() {
+            "open" => {
+                if let Some(window) = app.get_webview_window("main") {
+                    let _ = window.show();
+                    let _ = window.set_focus();
+                }
+            }
+            "toggle" => {
+                let _ = app.emit("tray-toggle-server", ());
+            }
+            "quit" => {
+                let install_state = app.state::<crate::install::InstallState>();
+                let _ = crate::install::stop_install(&install_state);
+                let update_state = app.state::<crate::update::UpdateState>();
+                let _ = crate::update::stop_update(&update_state);
+                // Detach the 5s graceful-wait so the tray click does not
+                // block the Tauri main loop. Exit runs the RunEvent::Exit
+                // safety net which also calls stop_backend synchronously.
+                let shutdown = app.state::<crate::process::ShutdownFlag>().inner().clone();
+                let backend_state = app.state::<crate::process::BackendState>().inner().clone();
+                crate::process::stop_backend_detached(backend_state, shutdown);
+                app.exit(0);
+            }
+            _ => {}
+        })
+        .on_tray_icon_event(|tray, event| {
+            if let TrayIconEvent::Click {
+                button: MouseButton::Left,
+                button_state: MouseButtonState::Up,
+                ..
+            } = event
+            {
+                if let Some(window) = tray.app_handle().get_webview_window("main") {
+                    let _ = window.show();
+                    let _ = window.set_focus();
+                }
+            }
+        })
+        .build(app)?;
+
+    Ok(())
+}
+
+fn main() {
+    // Fix PATH for GUI apps (macOS .app bundles, Linux AppImage, Windows)
+    // GUI apps don't inherit shell dotfile PATH — this spawns the user's
+    // login shell to source .zshrc/.bashrc/.profile and sets PATH properly.
+    let _ = fix_path_env::fix();
+
+    setup_logging();
+    info!("Unsloth Studio desktop app starting");
+    windows_job::initialize();
+
+    tauri::Builder::default()
+        .plugin(tauri_plugin_single_instance::init(|app, _args, _cwd| {
+            if let Some(window) = app.get_webview_window("main") {
+                let _ = window.show();
+                let _ = window.set_focus();
+            }
+        }))
+        .plugin(tauri_plugin_process::init())
+        .plugin(tauri_plugin_opener::init())
+        .plugin(tauri_plugin_updater::Builder::new().build())
+        .manage(install::new_install_state())
+        .manage(new_backend_state())
+        .manage(process::new_shutdown_flag())
+        .manage(update::new_update_state())
+        .invoke_handler(tauri::generate_handler![
+            commands::check_install_status,
+            commands::desktop_preflight,
+            commands::start_install,
+            commands::start_server,
+            commands::start_managed_server,
+            commands::stop_server,
+            commands::check_health,
+            commands::get_server_logs,
+            commands::open_logs_dir,
+            commands::start_backend_update,
+            commands::start_managed_repair,
+            commands::install_system_packages,
+            desktop_auth::desktop_auth,
+        ])
+        .setup(|app| {
+            setup_tray(app)?;
+            Ok(())
+        })
+        .on_window_event(|window, event| {
+            if let tauri::WindowEvent::CloseRequested { api, .. } = event {
+                // Hide window instead of closing — this is a tray app.
+                // Processes keep running so the backend stays available.
+                // Full cleanup happens via:
+                //   - Tray "Quit" menu item (explicit user action)
+                //   - RunEvent::Exit handler (OS shutdown, SIGTERM, etc.)
+                let _ = window.hide();
+                api.prevent_close();
+            }
+        })
+        .build(tauri::generate_context!())
+        .expect("error while building tauri application")
+        .run(|app, event| {
+            if let tauri::RunEvent::Exit = event {
+                // Cleanup on ALL exit paths — safety net for non-tray exits
+                if let Some(install_state) = app.try_state::<install::InstallState>() {
+                    let _ = install::stop_install(&install_state);
+                }
+                if let Some(update_state) = app.try_state::<update::UpdateState>() {
+                    let _ = update::stop_update(&update_state);
+                }
+                if let Some(backend_state) = app.try_state::<process::BackendState>() {
+                    let shutdown = app
+                        .try_state::<process::ShutdownFlag>()
+                        .expect("ShutdownFlag must be managed");
+                    let _ = process::stop_backend(&backend_state, &shutdown);
+                }
+            }
+        });
+}
diff --git a/studio/src-tauri/src/preflight.rs b/studio/src-tauri/src/preflight.rs
new file mode 100644
index 0000000000..d3df06d057
--- /dev/null
+++ b/studio/src-tauri/src/preflight.rs
@@ -0,0 +1,831 @@
+use serde::{Deserialize, Serialize};
+use std::path::PathBuf;
+use std::process::Stdio;
+use std::time::Duration;
+use tokio::io::AsyncReadExt;
+use tokio::process::Command;
+
+#[derive(Debug, Clone, PartialEq, Eq, Serialize, Deserialize)]
+#[serde(rename_all = "snake_case")]
+pub enum DesktopPreflightDisposition {
+    NotInstalled,
+    ManagedReady,
+    ManagedStale,
+    AttachedReady,
+}
+
+#[derive(Debug, Clone, PartialEq, Eq, Serialize, Deserialize)]
+pub struct DesktopPreflightResult {
+    pub disposition: DesktopPreflightDisposition,
+    pub reason: Option<String>,
+    pub port: Option<u16>,
+    pub can_auto_repair: bool,
+    pub managed_bin: Option<PathBuf>,
+}
+
+#[derive(Debug, Clone, PartialEq, Eq)]
+enum ManagedProbe {
+    Missing,
+    Ready { bin: PathBuf },
+    Stale { bin: PathBuf, reason: String },
+}
+
+#[derive(Debug, Clone, PartialEq, Eq)]
+enum BackendProbe {
+    Missing,
+    Ready { port: u16 },
+    Old { port: u16, reason: String },
+}
+
+#[derive(Debug, Deserialize)]
+struct DesktopCapability {
+    desktop_protocol_version: Option<u16>,
+    supports_api_only: Option<bool>,
+    supports_provision_desktop_auth: Option<bool>,
+    desktop_auth_stale_reason: Option<String>,
+}
+
+#[derive(Debug)]
+struct BackendHealth {
+    desktop_protocol_version: Option<u16>,
+    supports_desktop_auth: Option<bool>,
+    stale_reason: Option<String>,
+}
+
+fn release_auto_repair() -> bool {
+    !cfg!(debug_assertions)
+}
+
+fn choose_preflight(managed: ManagedProbe, backend: BackendProbe) -> DesktopPreflightResult {
+    match (backend, managed) {
+        (BackendProbe::Ready { port }, ManagedProbe::Ready { bin }) => DesktopPreflightResult {
+            disposition: DesktopPreflightDisposition::AttachedReady,
+            reason: None,
+            port: Some(port),
+            can_auto_repair: false,
+            managed_bin: Some(bin),
+        },
+        (_, managed) => match managed {
+            ManagedProbe::Ready { bin } => DesktopPreflightResult {
+                disposition: DesktopPreflightDisposition::ManagedReady,
+                reason: None,
+                port: None,
+                can_auto_repair: false,
+                managed_bin: Some(bin),
+            },
+            ManagedProbe::Stale { bin, reason } => DesktopPreflightResult {
+                disposition: DesktopPreflightDisposition::ManagedStale,
+                reason: Some(reason),
+                port: None,
+                can_auto_repair: release_auto_repair(),
+                managed_bin: Some(bin),
+            },
+            ManagedProbe::Missing => DesktopPreflightResult {
+                disposition: DesktopPreflightDisposition::NotInstalled,
+                reason: None,
+                port: None,
+                can_auto_repair: false,
+                managed_bin: None,
+            },
+        },
+    }
+}
+
+async fn run_cli_probe(bin: &std::path::Path, args: &[&str]) -> bool {
+    let mut cmd = Command::new(bin);
+    cmd.args(args).stdout(Stdio::null()).stderr(Stdio::null());
+
+    #[cfg(target_os = "linux")]
+    if std::env::var_os("APPIMAGE").is_some() {
+        cmd.env_remove("LD_LIBRARY_PATH");
+        cmd.env_remove("PYTHONHOME");
+        cmd.env_remove("PYTHONPATH");
+    }
+
+    #[cfg(windows)]
+    {
+        use std::os::windows::process::CommandExt;
+        cmd.creation_flags(crate::process::CREATE_NO_WINDOW);
+    }
+
+    let Ok(mut child) = cmd.spawn() else {
+        return false;
+    };
+
+    match tokio::time::timeout(Duration::from_secs(10), child.wait()).await {
+        Ok(Ok(status)) => status.success(),
+        _ => {
+            let _ = child.kill().await;
+            let _ = child.wait().await;
+            false
+        }
+    }
+}
+
+async fn probe_cli_capability(bin: &std::path::Path) -> Option<DesktopCapability> {
+    let mut cmd = Command::new(bin);
+    cmd.args(["studio", "desktop-capabilities", "--json"])
+        .stdout(Stdio::piped())
+        .stderr(Stdio::null());
+
+    #[cfg(target_os = "linux")]
+    if std::env::var_os("APPIMAGE").is_some() {
+        cmd.env_remove("LD_LIBRARY_PATH");
+        cmd.env_remove("PYTHONHOME");
+        cmd.env_remove("PYTHONPATH");
+    }
+
+    #[cfg(windows)]
+    {
+        use std::os::windows::process::CommandExt;
+        cmd.creation_flags(crate::process::CREATE_NO_WINDOW);
+    }
+
+    let Ok(mut child) = cmd.spawn() else {
+        return None;
+    };
+    let Some(mut stdout) = child.stdout.take() else {
+        return None;
+    };
+
+    match tokio::time::timeout(Duration::from_secs(10), child.wait()).await {
+        Ok(Ok(status)) if status.success() => {}
+        Err(_) => {
+            let _ = child.kill().await;
+            let _ = child.wait().await;
+            return None;
+        }
+        _ => return None,
+    }
+
+    let mut output = Vec::new();
+    if stdout.read_to_end(&mut output).await.is_err() {
+        return None;
+    }
+
+    serde_json::from_slice::<DesktopCapability>(&output).ok()
+}
+
+fn desktop_capability_ready(capability: &DesktopCapability) -> bool {
+    capability.desktop_protocol_version == Some(1)
+        && capability.supports_api_only == Some(true)
+        && capability.supports_provision_desktop_auth == Some(true)
+}
+
+fn desktop_capability_stale_reason(capability: &DesktopCapability) -> String {
+    capability
+        .desktop_auth_stale_reason
+        .clone()
+        .unwrap_or_else(|| "desktop_capability_incompatible".to_string())
+}
+
+async fn probe_managed_bin(bin: PathBuf) -> ManagedProbe {
+    if !run_cli_probe(&bin, &["-h"]).await {
+        return ManagedProbe::Stale {
+            bin,
+            reason: "cli_unusable".to_string(),
+        };
+    }
+
+    let capability = probe_cli_capability(&bin).await;
+    if let Some(capability) = capability {
+        if desktop_capability_ready(&capability) {
+            return ManagedProbe::Ready { bin };
+        }
+        return ManagedProbe::Stale {
+            bin,
+            reason: desktop_capability_stale_reason(&capability),
+        };
+    }
+
+    ManagedProbe::Stale {
+        bin,
+        reason: "desktop_capability_probe_failed".to_string(),
+    }
+}
+
+async fn probe_managed_install() -> ManagedProbe {
+    match crate::process::find_unsloth_binary() {
+        Some(bin) => probe_managed_bin(bin).await,
+        None => ManagedProbe::Missing,
+    }
+}
+
+pub async fn managed_install_ready() -> bool {
+    matches!(probe_managed_install().await, ManagedProbe::Ready { .. })
+}
+
+async fn backend_health(client: &reqwest::Client, port: u16) -> Option<BackendHealth> {
+    let url = format!("http://127.0.0.1:{port}/api/health");
+    let response = client.get(url).send().await.ok()?;
+    if !response.status().is_success() {
+        return None;
+    }
+    let json = response.json::<serde_json::Value>().await.ok()?;
+    let healthy = json
+        .get("status")
+        .and_then(|v| v.as_str())
+        .map(|s| s == "healthy")
+        .unwrap_or(false);
+    let service = json
+        .get("service")
+        .and_then(|v| v.as_str())
+        .map(|s| s == "Unsloth UI Backend")
+        .unwrap_or(false);
+    if !healthy || !service {
+        return None;
+    }
+
+    let desktop_protocol_version = json
+        .get("desktop_protocol_version")
+        .and_then(|v| v.as_u64())
+        .and_then(|v| u16::try_from(v).ok());
+    let supports_desktop_auth = json.get("supports_desktop_auth").and_then(|v| v.as_bool());
+    if desktop_protocol_version.is_none() && supports_desktop_auth.is_none() {
+        return None;
+    }
+    let stale_reason = match supports_desktop_auth {
+        Some(false) => json
+            .get("desktop_auth_stale_reason")
+            .and_then(|v| v.as_str())
+            .map(ToOwned::to_owned),
+        _ => None,
+    };
+    Some(BackendHealth {
+        desktop_protocol_version,
+        supports_desktop_auth,
+        stale_reason,
+    })
+}
+
+fn backend_capability_stale_reason(health: &BackendHealth) -> Option<String> {
+    if health.desktop_protocol_version != Some(1) {
+        return health
+            .stale_reason
+            .clone()
+            .or_else(|| Some("desktop_protocol_incompatible".to_string()));
+    }
+    if health.supports_desktop_auth != Some(true) {
+        return health
+            .stale_reason
+            .clone()
+            .or_else(|| Some("desktop_auth_unsupported".to_string()));
+    }
+    None
+}
+
+#[derive(Serialize)]
+struct DesktopLoginProbe<'a> {
+    secret: &'a str,
+}
+
+async fn backend_desktop_auth_status(
+    client: &reqwest::Client,
+    port: u16,
+    health: &BackendHealth,
+) -> BackendProbe {
+    if let Some(reason) = backend_capability_stale_reason(health) {
+        return BackendProbe::Old { port, reason };
+    }
+
+    let url = format!("http://127.0.0.1:{port}/api/auth/desktop-login");
+    let response = client
+        .post(url)
+        .json(&DesktopLoginProbe {
+            secret: "desktop-preflight-invalid-secret",
+        })
+        .send()
+        .await;
+
+    let Ok(response) = response else {
+        return BackendProbe::Old {
+            port,
+            reason: backend_capability_stale_reason(health)
+                .unwrap_or_else(|| "desktop_login_probe_failed".to_string()),
+        };
+    };
+
+    match response.status() {
+        reqwest::StatusCode::UNAUTHORIZED => BackendProbe::Ready { port },
+        reqwest::StatusCode::NOT_FOUND => BackendProbe::Old {
+            port,
+            reason: "desktop_login_not_found".to_string(),
+        },
+        _ => BackendProbe::Old {
+            port,
+            reason: backend_capability_stale_reason(health)
+                .unwrap_or_else(|| "desktop_login_probe_failed".to_string()),
+        },
+    }
+}
+
+async fn probe_existing_backends() -> BackendProbe {
+    let client = match reqwest::Client::builder()
+        .timeout(Duration::from_secs(2))
+        .build()
+    {
+        Ok(client) => client,
+        Err(_) => return BackendProbe::Missing,
+    };
+
+    // Fan out health probes concurrently. The desktop-auth probe is still
+    // sequential per candidate because it has auth-log side effects.
+    let ports: Vec<u16> = (8888u16..=8908).collect();
+    let mut health_futs = Vec::with_capacity(ports.len());
+    for port in ports {
+        // why: reqwest::Client is internally Arc-wrapped; clone is a refcount bump
+        // (documented cheap). tokio::spawn needs 'static, so each task owns its own clone.
+        let c = client.clone();
+        health_futs.push(tokio::spawn(async move {
+            backend_health(&c, port).await.map(|h| (port, h))
+        }));
+    }
+
+    let mut candidates: Vec<(u16, BackendHealth)> = Vec::new();
+    for fut in health_futs {
+        if let Ok(Some(pair)) = fut.await {
+            candidates.push(pair);
+        }
+    }
+
+    let mut first_old = None;
+    for (port, health) in candidates {
+        match backend_desktop_auth_status(&client, port, &health).await {
+            ready @ BackendProbe::Ready { .. } => return ready,
+            old @ BackendProbe::Old { .. } if first_old.is_none() => first_old = Some(old),
+            _ => {}
+        }
+    }
+
+    first_old.unwrap_or(BackendProbe::Missing)
+}
+
+pub async fn desktop_preflight_result() -> DesktopPreflightResult {
+    let (managed, backend) = tokio::join!(probe_managed_install(), probe_existing_backends());
+    choose_preflight(managed, backend)
+}
+
+#[cfg(test)]
+mod tests {
+    use super::*;
+    use tokio::io::{AsyncReadExt, AsyncWriteExt};
+    use tokio::net::TcpListener;
+
+    #[test]
+    fn compatible_backend_does_not_win_over_stale_managed_install() {
+        let result = choose_preflight(
+            ManagedProbe::Stale {
+                bin: PathBuf::from("/managed/unsloth"),
+                reason: "old cli".to_string(),
+            },
+            BackendProbe::Ready { port: 8000 },
+        );
+
+        assert_eq!(
+            result.disposition,
+            DesktopPreflightDisposition::ManagedStale
+        );
+        assert_eq!(result.port, None);
+        assert_eq!(result.reason, Some("old cli".to_string()));
+        assert_eq!(result.can_auto_repair, release_auto_repair());
+        assert_eq!(result.managed_bin, Some(PathBuf::from("/managed/unsloth")));
+    }
+
+    #[test]
+    fn compatible_backend_wins_over_ready_managed_install() {
+        let result = choose_preflight(
+            ManagedProbe::Ready {
+                bin: PathBuf::from("/managed/unsloth"),
+            },
+            BackendProbe::Ready { port: 8000 },
+        );
+
+        assert_eq!(
+            result.disposition,
+            DesktopPreflightDisposition::AttachedReady
+        );
+        assert_eq!(result.port, Some(8000));
+        assert_eq!(result.managed_bin, Some(PathBuf::from("/managed/unsloth")));
+        assert!(!result.can_auto_repair);
+    }
+
+    #[test]
+    fn compatible_backend_does_not_win_over_missing_managed_install() {
+        let result = choose_preflight(ManagedProbe::Missing, BackendProbe::Ready { port: 8000 });
+
+        assert_eq!(
+            result.disposition,
+            DesktopPreflightDisposition::NotInstalled
+        );
+        assert_eq!(result.port, None);
+        assert_eq!(result.managed_bin, None);
+        assert!(!result.can_auto_repair);
+    }
+
+    #[test]
+    fn old_backend_falls_back_to_ready_managed_install() {
+        let result = choose_preflight(
+            ManagedProbe::Ready {
+                bin: PathBuf::from("/managed/unsloth"),
+            },
+            BackendProbe::Old {
+                port: 8001,
+                reason: "missing endpoint".to_string(),
+            },
+        );
+
+        assert_eq!(
+            result.disposition,
+            DesktopPreflightDisposition::ManagedReady
+        );
+        assert_eq!(result.reason, None);
+        assert_eq!(result.port, None);
+        assert_eq!(result.managed_bin, Some(PathBuf::from("/managed/unsloth")));
+        assert!(!result.can_auto_repair);
+    }
+
+    #[test]
+    fn old_backend_falls_back_to_stale_managed_install() {
+        let result = choose_preflight(
+            ManagedProbe::Stale {
+                bin: PathBuf::from("/managed/unsloth"),
+                reason: "old cli".to_string(),
+            },
+            BackendProbe::Old {
+                port: 8001,
+                reason: "missing endpoint".to_string(),
+            },
+        );
+
+        assert_eq!(
+            result.disposition,
+            DesktopPreflightDisposition::ManagedStale
+        );
+        assert_eq!(result.reason, Some("old cli".to_string()));
+        assert_eq!(result.port, None);
+        assert_eq!(result.can_auto_repair, release_auto_repair());
+        assert_eq!(result.managed_bin, Some(PathBuf::from("/managed/unsloth")));
+    }
+
+    #[test]
+    fn managed_ready_when_no_backend() {
+        let result = choose_preflight(
+            ManagedProbe::Ready {
+                bin: PathBuf::from("/managed/unsloth"),
+            },
+            BackendProbe::Missing,
+        );
+
+        assert_eq!(
+            result.disposition,
+            DesktopPreflightDisposition::ManagedReady
+        );
+        assert_eq!(result.managed_bin, Some(PathBuf::from("/managed/unsloth")));
+        assert!(!result.can_auto_repair);
+    }
+
+    #[test]
+    fn managed_stale_when_no_backend() {
+        let result = choose_preflight(
+            ManagedProbe::Stale {
+                bin: PathBuf::from("/managed/unsloth"),
+                reason: "old cli".to_string(),
+            },
+            BackendProbe::Missing,
+        );
+
+        assert_eq!(
+            result.disposition,
+            DesktopPreflightDisposition::ManagedStale
+        );
+        assert_eq!(result.reason, Some("old cli".to_string()));
+        assert_eq!(result.can_auto_repair, release_auto_repair());
+    }
+
+    #[test]
+    fn not_installed_when_no_backend_no_managed_binary() {
+        let result = choose_preflight(ManagedProbe::Missing, BackendProbe::Missing);
+
+        assert_eq!(
+            result.disposition,
+            DesktopPreflightDisposition::NotInstalled
+        );
+        assert_eq!(result.managed_bin, None);
+        assert!(!result.can_auto_repair);
+    }
+
+    #[test]
+    fn old_backend_with_no_managed_install_uses_install_flow() {
+        let result = choose_preflight(
+            ManagedProbe::Missing,
+            BackendProbe::Old {
+                port: 8002,
+                reason: "old version".to_string(),
+            },
+        );
+
+        assert_eq!(
+            result.disposition,
+            DesktopPreflightDisposition::NotInstalled
+        );
+        assert_eq!(result.reason, None);
+        assert_eq!(result.port, None);
+        assert!(!result.can_auto_repair);
+    }
+
+    #[cfg(unix)]
+    struct FakeCli {
+        bin: PathBuf,
+        dir: PathBuf,
+    }
+
+    #[cfg(unix)]
+    impl Drop for FakeCli {
+        fn drop(&mut self) {
+            let _ = std::fs::remove_dir_all(&self.dir);
+        }
+    }
+
+    #[cfg(unix)]
+    fn fake_cli(test_name: &str, script: &str) -> FakeCli {
+        use std::fs;
+        use std::os::unix::fs::PermissionsExt;
+        use std::time::{SystemTime, UNIX_EPOCH};
+
+        let nanos = SystemTime::now()
+            .duration_since(UNIX_EPOCH)
+            .unwrap()
+            .as_nanos();
+        let dir = std::env::temp_dir().join(format!(
+            "unsloth-preflight-{test_name}-{}-{nanos}",
+            std::process::id()
+        ));
+        fs::create_dir_all(&dir).unwrap();
+        let bin = dir.join("unsloth");
+        fs::write(&bin, script).unwrap();
+        let mut perms = fs::metadata(&bin).unwrap().permissions();
+        perms.set_mode(0o755);
+        fs::set_permissions(&bin, perms).unwrap();
+        FakeCli { bin, dir }
+    }
+
+    #[cfg(unix)]
+    #[tokio::test]
+    async fn managed_cli_stale_when_desktop_capabilities_missing() {
+        let fake = fake_cli(
+            "cap-missing",
+            r#"#!/bin/sh
+if [ "$1" = "-h" ]; then exit 0; fi
+if [ "$1" = "studio" ] && [ "$2" = "provision-desktop-auth" ] && [ "$3" = "--help" ]; then exit 0; fi
+exit 1
+"#,
+        );
+        let bin = fake.bin.clone();
+
+        assert!(matches!(
+            probe_managed_bin(bin.clone()).await,
+            ManagedProbe::Stale { bin: actual_bin, .. } if actual_bin == bin
+        ));
+    }
+
+    #[cfg(unix)]
+    #[tokio::test]
+    async fn managed_cli_stale_when_desktop_capabilities_command_missing() {
+        let fake = fake_cli(
+            "missing-helper",
+            r#"#!/bin/sh
+if [ "$1" = "-h" ]; then exit 0; fi
+exit 1
+"#,
+        );
+        let bin = fake.bin.clone();
+
+        assert!(matches!(
+            probe_managed_bin(bin.clone()).await,
+            ManagedProbe::Stale { bin: actual_bin, .. } if actual_bin == bin
+        ));
+    }
+
+    #[cfg(unix)]
+    #[tokio::test]
+    async fn managed_cli_stale_when_help_broken() {
+        let fake = fake_cli(
+            "broken-help",
+            r#"#!/bin/sh
+exit 1
+"#,
+        );
+        let bin = fake.bin.clone();
+
+        assert_eq!(
+            probe_managed_bin(bin.clone()).await,
+            ManagedProbe::Stale {
+                bin,
+                reason: "cli_unusable".to_string()
+            }
+        );
+    }
+
+    #[cfg(unix)]
+    #[tokio::test]
+    async fn managed_cli_ready_when_desktop_capabilities_compatible() {
+        let fake = fake_cli(
+            "cap-true-helper-missing",
+            r#"#!/bin/sh
+if [ "$1" = "-h" ]; then exit 0; fi
+if [ "$1" = "studio" ] && [ "$2" = "desktop-capabilities" ] && [ "$3" = "--json" ]; then
+  printf '{"desktop_protocol_version":1,"supports_api_only":true,"supports_provision_desktop_auth":true}'
+  exit 0
+fi
+exit 1
+"#,
+        );
+        let bin = fake.bin.clone();
+
+        assert_eq!(
+            probe_managed_bin(bin.clone()).await,
+            ManagedProbe::Ready { bin }
+        );
+    }
+
+    #[cfg(unix)]
+    #[tokio::test]
+    async fn capability_false_reason_used_when_legacy_helper_missing() {
+        let fake = fake_cli(
+            "cap-false-helper-missing",
+            r#"#!/bin/sh
+if [ "$1" = "-h" ]; then exit 0; fi
+if [ "$1" = "studio" ] && [ "$2" = "desktop-capabilities" ] && [ "$3" = "--json" ]; then
+  printf '{"desktop_protocol_version":1,"supports_api_only":true,"supports_provision_desktop_auth":false,"desktop_auth_stale_reason":"cap_false"}'
+  exit 0
+fi
+exit 1
+"#,
+        );
+        let bin = fake.bin.clone();
+
+        assert_eq!(
+            probe_managed_bin(bin.clone()).await,
+            ManagedProbe::Stale {
+                bin,
+                reason: "cap_false".to_string()
+            }
+        );
+    }
+
+    #[cfg(unix)]
+    #[tokio::test]
+    async fn capability_false_overrides_working_legacy_helper() {
+        let fake = fake_cli(
+            "cap-false-helper-ready",
+            r#"#!/bin/sh
+if [ "$1" = "-h" ]; then exit 0; fi
+if [ "$1" = "studio" ] && [ "$2" = "desktop-capabilities" ] && [ "$3" = "--json" ]; then
+  printf '{"desktop_protocol_version":1,"supports_api_only":true,"supports_provision_desktop_auth":false,"desktop_auth_stale_reason":"cap_false"}'
+  exit 0
+fi
+if [ "$1" = "studio" ] && [ "$2" = "provision-desktop-auth" ] && [ "$3" = "--help" ]; then exit 0; fi
+exit 1
+"#,
+        );
+        let bin = fake.bin.clone();
+
+        assert_eq!(
+            probe_managed_bin(bin.clone()).await,
+            ManagedProbe::Stale {
+                bin,
+                reason: "cap_false".to_string()
+            }
+        );
+    }
+
+    async fn backend_server(health_body: &'static str, route_status: &'static str) -> u16 {
+        let listener = TcpListener::bind("127.0.0.1:0").await.unwrap();
+        let port = listener.local_addr().unwrap().port();
+
+        tokio::spawn(async move {
+            for _ in 0..2 {
+                let (mut stream, _) = listener.accept().await.unwrap();
+                let mut buffer = [0; 2048];
+                let n = stream.read(&mut buffer).await.unwrap();
+                let request = String::from_utf8_lossy(&buffer[..n]);
+                let (status, body) = if request.starts_with("GET /api/health ") {
+                    ("200 OK", health_body)
+                } else if request.starts_with("POST /api/auth/desktop-login ") {
+                    (route_status, "")
+                } else {
+                    ("404 Not Found", "")
+                };
+                let response = format!(
+                    "HTTP/1.1 {status}\r\nContent-Type: application/json\r\nContent-Length: {}\r\nConnection: close\r\n\r\n{body}",
+                    body.len()
+                );
+                stream.write_all(response.as_bytes()).await.unwrap();
+            }
+        });
+
+        port
+    }
+
+    async fn probe_test_backend(
+        health_body: &'static str,
+        route_status: &'static str,
+    ) -> BackendProbe {
+        let port = backend_server(health_body, route_status).await;
+        let client = reqwest::Client::new();
+        let health = backend_health(&client, port).await.unwrap();
+        backend_desktop_auth_status(&client, port, &health).await
+    }
+
+    #[tokio::test]
+    async fn backend_health_without_desktop_capability_fields_is_not_compatible() {
+        let port = backend_server(
+            r#"{"status":"healthy","service":"Unsloth UI Backend"}"#,
+            "401 Unauthorized",
+        )
+        .await;
+        let client = reqwest::Client::new();
+
+        assert!(backend_health(&client, port).await.is_none());
+    }
+
+    #[tokio::test]
+    async fn backend_with_auth_support_but_missing_protocol_is_old() {
+        let probe = probe_test_backend(
+            r#"{"status":"healthy","service":"Unsloth UI Backend","supports_desktop_auth":true}"#,
+            "401 Unauthorized",
+        )
+        .await;
+
+        assert!(matches!(probe, BackendProbe::Old { .. }));
+    }
+
+    #[tokio::test]
+    async fn backend_with_auth_support_but_unsupported_protocol_is_old() {
+        let probe = probe_test_backend(
+            r#"{"status":"healthy","service":"Unsloth UI Backend","desktop_protocol_version":2,"supports_desktop_auth":true}"#,
+            "401 Unauthorized",
+        )
+        .await;
+
+        assert!(matches!(probe, BackendProbe::Old { .. }));
+    }
+
+    #[tokio::test]
+    async fn backend_health_with_desktop_capability_fields_and_401_is_ready() {
+        let probe = probe_test_backend(
+            r#"{"status":"healthy","service":"Unsloth UI Backend","desktop_protocol_version":1,"supports_desktop_auth":true}"#,
+            "401 Unauthorized",
+        )
+        .await;
+
+        assert!(matches!(probe, BackendProbe::Ready { .. }));
+    }
+
+    #[tokio::test]
+    async fn backend_route_404_is_old() {
+        let probe = probe_test_backend(
+            r#"{"status":"healthy","service":"Unsloth UI Backend","desktop_protocol_version":1,"supports_desktop_auth":true}"#,
+            "404 Not Found",
+        )
+        .await;
+
+        assert!(matches!(
+            probe,
+            BackendProbe::Old {
+                reason,
+                ..
+            } if reason == "desktop_login_not_found"
+        ));
+    }
+
+    #[tokio::test]
+    async fn backend_route_500_is_old() {
+        let probe = probe_test_backend(
+            r#"{"status":"healthy","service":"Unsloth UI Backend","desktop_protocol_version":1,"supports_desktop_auth":true}"#,
+            "500 Internal Server Error",
+        )
+        .await;
+
+        assert!(matches!(probe, BackendProbe::Old { .. }));
+    }
+
+    #[tokio::test]
+    async fn backend_capability_false_is_old_even_when_route_401() {
+        let port = backend_server(
+            r#"{"status":"healthy","service":"Unsloth UI Backend","desktop_protocol_version":1,"supports_desktop_auth":false,"desktop_auth_stale_reason":"cap_false"}"#,
+            "401 Unauthorized",
+        )
+        .await;
+        let client = reqwest::Client::new();
+        let health = backend_health(&client, port).await.unwrap();
+
+        assert!(matches!(
+            backend_desktop_auth_status(&client, port, &health).await,
+            BackendProbe::Old {
+                reason,
+                ..
+            } if reason == "cap_false"
+        ));
+    }
+}
diff --git a/studio/src-tauri/src/process.rs b/studio/src-tauri/src/process.rs
new file mode 100644
index 0000000000..beba8c1208
--- /dev/null
+++ b/studio/src-tauri/src/process.rs
@@ -0,0 +1,537 @@
+use log::{error, info, warn};
+use process_wrap::std::*;
+use regex::Regex;
+use std::collections::VecDeque;
+use std::io::BufRead;
+use std::process::{Command, Stdio};
+use std::sync::atomic::{AtomicBool, Ordering};
+use std::sync::{Arc, Mutex};
+use tauri::{AppHandle, Emitter};
+
+const MAX_LOG_LINES: usize = 1000;
+
+pub struct BackendProcess {
+    pub child: Option<Box<dyn ChildWrapper + Send>>,
+    pub port: Option<u16>,
+    pub logs: VecDeque<String>,
+    pub intentional_stop: bool,
+}
+
+impl Default for BackendProcess {
+    fn default() -> Self {
+        Self {
+            child: None,
+            port: None,
+            logs: VecDeque::with_capacity(MAX_LOG_LINES),
+            intentional_stop: false,
+        }
+    }
+}
+
+pub type BackendState = Arc<Mutex<BackendProcess>>;
+pub type ShutdownFlag = Arc<AtomicBool>;
+
+pub fn new_backend_state() -> BackendState {
+    Arc::new(Mutex::new(BackendProcess::default()))
+}
+
+pub fn new_shutdown_flag() -> ShutdownFlag {
+    Arc::new(AtomicBool::new(false))
+}
+
+pub(crate) fn trim_line_endings(bytes: &[u8]) -> &[u8] {
+    let mut end = bytes.len();
+    while end > 0 && matches!(bytes[end - 1], b'\n' | b'\r') {
+        end -= 1;
+    }
+    &bytes[..end]
+}
+
+/// Windows `CREATE_NO_WINDOW` flag — suppresses console windows for child processes.
+#[cfg(windows)]
+pub(crate) const CREATE_NO_WINDOW: u32 = 0x08000000;
+
+/// Force-kill a Windows process tree via hidden `taskkill /T /F`, falling
+/// back to `child.kill()` if taskkill itself fails.  Reaps the child afterward.
+#[cfg(windows)]
+pub(crate) fn force_kill_process_tree(
+    pid: u32,
+    child: &mut Box<dyn ChildWrapper + Send>,
+    label: &str,
+) {
+    use std::os::windows::process::CommandExt;
+
+    let taskkill_status = Command::new("taskkill.exe")
+        .creation_flags(CREATE_NO_WINDOW)
+        .args(["/PID", &pid.to_string(), "/T", "/F"])
+        .stdout(Stdio::null())
+        .stderr(Stdio::null())
+        .status();
+
+    match taskkill_status {
+        Ok(status) if status.success() => {}
+        Ok(status) => {
+            warn!(
+                "taskkill returned non-zero status for {} pid {}: {}",
+                label, pid, status
+            );
+            let _ = child.kill();
+        }
+        Err(e) => {
+            warn!("taskkill failed for {} pid {}: {}", label, pid, e);
+            let _ = child.kill();
+        }
+    }
+
+    let _ = child.wait();
+    info!("{} process tree force stopped", label);
+}
+
+/// Returns the path to the unsloth binary inside the managed venv, if it exists.
+/// Checks the new layout (~/.unsloth/studio/unsloth_studio/) first,
+/// then falls back to the old layout (~/.unsloth/studio/.venv/) for compat.
+fn find_unsloth_binary_in_studio_dir(studio: &std::path::Path) -> Option<std::path::PathBuf> {
+    // New layout (upstream scripts >= March 2026)
+    let new_base = studio.join("unsloth_studio");
+    // Old layout (bundled scripts, older upstream)
+    let old_base = studio.join(".venv");
+
+    for base in [new_base, old_base] {
+        #[cfg(unix)]
+        let bin = base.join("bin").join("unsloth");
+        #[cfg(windows)]
+        let bin = base.join("Scripts").join("unsloth.exe");
+
+        if bin.exists() {
+            return Some(bin);
+        }
+    }
+
+    None
+}
+
+pub fn find_unsloth_binary() -> Option<std::path::PathBuf> {
+    let home = dirs::home_dir()?;
+    let studio = home.join(".unsloth").join("studio");
+
+    find_unsloth_binary_in_studio_dir(&studio)
+}
+
+#[cfg(test)]
+mod tests {
+    use super::*;
+    use std::fs;
+    use std::path::PathBuf;
+    use std::time::{SystemTime, UNIX_EPOCH};
+
+    fn temp_studio_dir(test_name: &str) -> PathBuf {
+        let nanos = SystemTime::now()
+            .duration_since(UNIX_EPOCH)
+            .unwrap()
+            .as_nanos();
+        let dir = std::env::temp_dir().join(format!(
+            "unsloth-{test_name}-{}-{nanos}",
+            std::process::id()
+        ));
+        fs::create_dir_all(&dir).unwrap();
+        dir
+    }
+
+    #[test]
+    fn finds_new_layout_before_legacy_layout() {
+        let temp = temp_studio_dir("new-before-legacy");
+
+        #[cfg(unix)]
+        let new_bin = temp.join("unsloth_studio/bin/unsloth");
+        #[cfg(unix)]
+        let old_bin = temp.join(".venv/bin/unsloth");
+        #[cfg(windows)]
+        let new_bin = temp.join("unsloth_studio/Scripts/unsloth.exe");
+        #[cfg(windows)]
+        let old_bin = temp.join(".venv/Scripts/unsloth.exe");
+
+        fs::create_dir_all(new_bin.parent().unwrap()).unwrap();
+        fs::create_dir_all(old_bin.parent().unwrap()).unwrap();
+        fs::write(&new_bin, "").unwrap();
+        fs::write(&old_bin, "").unwrap();
+
+        assert_eq!(find_unsloth_binary_in_studio_dir(&temp), Some(new_bin));
+        fs::remove_dir_all(temp).unwrap();
+    }
+
+    #[test]
+    fn finds_legacy_layout_when_new_missing() {
+        let temp = temp_studio_dir("legacy");
+
+        #[cfg(unix)]
+        let old_bin = temp.join(".venv/bin/unsloth");
+        #[cfg(windows)]
+        let old_bin = temp.join(".venv/Scripts/unsloth.exe");
+
+        fs::create_dir_all(old_bin.parent().unwrap()).unwrap();
+        fs::write(&old_bin, "").unwrap();
+
+        assert_eq!(find_unsloth_binary_in_studio_dir(&temp), Some(old_bin));
+        fs::remove_dir_all(temp).unwrap();
+    }
+
+    #[test]
+    fn returns_none_when_no_managed_layout_exists() {
+        let temp = temp_studio_dir("none");
+
+        assert_eq!(find_unsloth_binary_in_studio_dir(&temp), None);
+        fs::remove_dir_all(temp).unwrap();
+    }
+
+    #[test]
+    fn backend_args_always_enable_api_only() {
+        assert_eq!(
+            backend_args(8888),
+            vec!["studio", "--api-only", "-H", "127.0.0.1", "-p", "8888"]
+        );
+    }
+}
+
+/// Find the unsloth binary, preferring the dev repo if available.
+/// In dev mode (debug builds), checks for a local .venv in the repo first.
+/// Falls back to find_unsloth_binary() which checks ~/.unsloth/studio/unsloth_studio/
+/// (new layout) then ~/.unsloth/studio/.venv/ (old layout).
+pub(crate) fn resolve_backend_binary() -> Result<std::path::PathBuf, String> {
+    // In dev mode, check for local repo venv first
+    #[cfg(debug_assertions)]
+    {
+        // CARGO_MANIFEST_DIR is set at compile time to studio/src-tauri/
+        // Repo root is 2 levels up: studio/src-tauri -> studio -> repo_root
+        let manifest_dir = env!("CARGO_MANIFEST_DIR");
+        let repo_root = std::path::Path::new(manifest_dir)
+            .parent() // studio/
+            .and_then(|p| p.parent()); // repo_root/
+
+        if let Some(root) = repo_root {
+            #[cfg(unix)]
+            let dev_bin = root.join(".venv/bin/unsloth");
+            #[cfg(windows)]
+            let dev_bin = root.join(".venv/Scripts/unsloth.exe");
+
+            if dev_bin.exists() {
+                info!("Dev mode: using local repo backend at {:?}", dev_bin);
+                return Ok(dev_bin.to_path_buf());
+            }
+        }
+        info!("Dev mode: no local .venv found, falling back to installed backend");
+    }
+
+    find_unsloth_binary()
+        .ok_or_else(|| "Unsloth binary not found. Please install Unsloth Studio first.".to_string())
+}
+
+fn backend_args(port: u16) -> Vec<String> {
+    [
+        "studio",
+        "--api-only",
+        "-H",
+        "127.0.0.1",
+        "-p",
+        &port.to_string(),
+    ]
+    .into_iter()
+    .map(String::from)
+    .collect()
+}
+
+/// Spawn the backend process and wire up stdout/stderr reader threads.
+pub fn start_backend(
+    app: &AppHandle,
+    state: &BackendState,
+    port: u16,
+    shutdown: &ShutdownFlag,
+) -> Result<(), String> {
+    let bin = resolve_backend_binary()?;
+
+    shutdown.store(false, Ordering::SeqCst);
+
+    // Reset state
+    {
+        let mut proc = state.lock().map_err(|e| e.to_string())?;
+        if proc.child.is_some() {
+            return Err("Backend is already running.".to_string());
+        }
+        proc.port = None;
+        proc.logs.clear();
+        proc.intentional_stop = false;
+    }
+
+    let args = backend_args(port);
+    info!("Starting backend: {:?} {}", bin, args.join(" "));
+
+    let mut cmd = Command::new(&bin);
+    cmd.args(&args)
+        .stdout(Stdio::piped())
+        .stderr(Stdio::piped());
+
+    // AppImage sets LD_LIBRARY_PATH to its bundled libs, which breaks the spawned
+    // Python process (wrong libpython/libz → "No module named encodings").
+    // Only clear when running inside an AppImage — native .deb/.rpm installs may
+    // need these env vars for custom CUDA or conda paths.
+    #[cfg(target_os = "linux")]
+    if std::env::var_os("APPIMAGE").is_some() {
+        cmd.env_remove("LD_LIBRARY_PATH");
+        cmd.env_remove("PYTHONHOME");
+        cmd.env_remove("PYTHONPATH");
+    }
+
+    // On Windows, launch the backend directly with hidden-window flags.
+    // The app process is assigned to a KILL_ON_JOB_CLOSE job in main.rs, so
+    // children inherit crash-safe cleanup without the buggy per-child JobObject wrapper.
+    #[cfg(windows)]
+    let mut child: Box<dyn ChildWrapper + Send> = {
+        use std::os::windows::process::CommandExt;
+
+        const CREATE_NEW_PROCESS_GROUP: u32 = 0x00000200;
+        cmd.creation_flags(CREATE_NEW_PROCESS_GROUP | CREATE_NO_WINDOW);
+        let child = cmd
+            .spawn()
+            .map_err(|e| format!("Failed to spawn backend: {}", e))?;
+        Box::new(child)
+    };
+
+    #[cfg(unix)]
+    let mut child: Box<dyn ChildWrapper + Send> = {
+        // Keep the backend tree in a process group on Unix for cleanup.
+        let mut wrap = CommandWrap::from(cmd);
+        wrap.wrap(ProcessGroup::leader());
+        wrap.spawn()
+            .map_err(|e| format!("Failed to spawn backend: {}", e))?
+    };
+
+    let stdout = child.stdout().take();
+    let stderr = child.stderr().take();
+
+    // Store child in state
+    {
+        let mut proc = state.lock().map_err(|e| e.to_string())?;
+        proc.child = Some(child);
+    }
+
+    // Spawn stdout reader thread
+    if let Some(stdout) = stdout {
+        let app_handle = app.clone();
+        let state_clone = Arc::clone(state);
+        std::thread::spawn(move || {
+            read_output_stream(stdout, &app_handle, &state_clone, false);
+        });
+    }
+
+    // Spawn stderr reader thread
+    if let Some(stderr) = stderr {
+        let app_handle = app.clone();
+        let state_clone = Arc::clone(state);
+        std::thread::spawn(move || {
+            read_output_stream(stderr, &app_handle, &state_clone, true);
+        });
+    }
+
+    Ok(())
+}
+
+/// Read lines from a child process stream (stdout or stderr).
+/// For stdout, parse TAURI_PORT=(\d+) to detect the actual port.
+/// When stdout closes and the stop was not intentional, emit server-crashed.
+fn read_output_stream<R: std::io::Read>(
+    stream: R,
+    app: &AppHandle,
+    state: &BackendState,
+    is_stderr: bool,
+) {
+    let mut reader = std::io::BufReader::new(stream);
+    let port_re = Regex::new(r"TAURI_PORT=(\d+)").unwrap();
+    let mut buf = Vec::new();
+
+    loop {
+        buf.clear();
+        match reader.read_until(b'\n', &mut buf) {
+            Ok(0) => break,
+            Ok(_) => {
+                let text = String::from_utf8_lossy(trim_line_endings(&buf)).into_owned();
+                let log_line = if is_stderr {
+                    format!("[stderr] {}", text)
+                } else {
+                    text.clone()
+                };
+
+                // Check for TAURI_PORT on stdout only
+                if !is_stderr {
+                    if let Some(caps) = port_re.captures(&text) {
+                        if let Some(port_str) = caps.get(1) {
+                            if let Ok(port) = port_str.as_str().parse::<u16>() {
+                                info!("Detected backend port: {}", port);
+                                if let Ok(mut proc) = state.lock() {
+                                    proc.port = Some(port);
+                                }
+                                let _ = app.emit("server-port", port);
+                            }
+                        }
+                    }
+                }
+
+                // Buffer the log line
+                if let Ok(mut proc) = state.lock() {
+                    if proc.logs.len() >= MAX_LOG_LINES {
+                        proc.logs.pop_front();
+                    }
+                    proc.logs.push_back(log_line.clone());
+                }
+
+                info!("[backend] {}", log_line);
+
+                // Emit to frontend
+                let _ = app.emit("server-log", &log_line);
+            }
+            Err(e) => {
+                warn!(
+                    "Error reading backend {}: {}",
+                    if is_stderr { "stderr" } else { "stdout" },
+                    e
+                );
+                break;
+            }
+        }
+    }
+
+    // Stream closed. Only the stdout reader checks for crashes.
+    if !is_stderr {
+        if let Ok(mut proc) = state.lock() {
+            let intentional = proc.intentional_stop;
+            let exited = if let Some(ref mut child) = proc.child {
+                match child.try_wait() {
+                    Ok(Some(status)) => {
+                        info!("Backend stdout stream ended with status: {}", status);
+                        true
+                    }
+                    Ok(None) => {
+                        warn!("Backend stdout stream ended, but process is still running");
+                        false
+                    }
+                    Err(e) => {
+                        warn!("Failed to query backend status after stdout closed: {}", e);
+                        false
+                    }
+                }
+            } else {
+                false
+            };
+
+            if exited {
+                proc.child = None;
+                if !intentional {
+                    error!("Backend process stdout closed unexpectedly (crash detected)");
+                    let _ = app.emit("server-crashed", ());
+                }
+            }
+        }
+    }
+}
+
+/// Graceful shutdown of the backend process and its entire subprocess tree.
+/// Unix: SIGTERM to process group -> wait up to 5s -> SIGKILL to group
+/// Windows: CTRL_BREAK_EVENT -> wait up to 5s -> hidden taskkill /T /F
+pub fn stop_backend(state: &BackendState, shutdown: &ShutdownFlag) -> Result<(), String> {
+    shutdown.store(true, Ordering::SeqCst);
+
+    // Extract the child and mark intentional stop.
+    // We take the child OUT of the mutex so we don't hold the lock during the wait loop.
+    let mut child = {
+        let mut proc = match state.lock() {
+            Ok(guard) => guard,
+            Err(poisoned) => {
+                warn!("Backend state mutex poisoned, recovering for cleanup");
+                poisoned.into_inner()
+            }
+        };
+        proc.intentional_stop = true;
+        proc.child.take()
+    };
+
+    let Some(ref mut child) = child else {
+        return Ok(()); // Nothing running
+    };
+
+    let pid = child.id();
+    info!("Stopping backend process group (pid {})", pid);
+
+    // Send SIGTERM to the entire process group (negative PID = group signal).
+    // This gives Python and any workers a chance to shut down gracefully.
+    #[cfg(unix)]
+    {
+        if pid > i32::MAX as u32 {
+            // PID too large for i32 negation, fall back to direct kill
+            warn!("PID {} exceeds i32 range, using direct kill", pid);
+            let _ = child.kill();
+            let _ = child.wait();
+            return Ok(());
+        }
+        unsafe {
+            libc::kill(-(pid as i32), libc::SIGTERM);
+        }
+    }
+    #[cfg(windows)]
+    {
+        unsafe {
+            windows_sys::Win32::System::Console::GenerateConsoleCtrlEvent(
+                windows_sys::Win32::System::Console::CTRL_BREAK_EVENT,
+                pid,
+            );
+        }
+    }
+
+    // Poll for up to 5 seconds (50 iterations * 100ms)
+    for _ in 0..50 {
+        match child.try_wait() {
+            Ok(Some(status)) => {
+                info!("Backend exited with status: {}", status);
+                return Ok(());
+            }
+            Ok(None) => {
+                std::thread::sleep(std::time::Duration::from_millis(100));
+            }
+            Err(e) => {
+                warn!("Error polling backend process: {}", e);
+                break;
+            }
+        }
+    }
+
+    #[cfg(windows)]
+    {
+        warn!(
+            "Backend did not exit gracefully, force killing process tree (pid {})",
+            pid
+        );
+        force_kill_process_tree(pid, child, "Backend");
+        return Ok(());
+    }
+
+    #[cfg(unix)]
+    {
+        // Force kill the process group on Unix
+        warn!(
+            "Backend did not exit gracefully, force killing group (pid {})",
+            pid
+        );
+        let _ = child.kill();
+
+        // Reap the process
+        let _ = child.wait();
+        info!("Backend process group forcefully stopped");
+        Ok(())
+    }
+}
+
+/// Spawn `stop_backend` on a background thread and return immediately.
+/// Used by the tray "quit" path so the 5s graceful-wait does not block
+/// the Tauri main event loop before `app.exit(0)` fires.
+pub fn stop_backend_detached(state: BackendState, shutdown: ShutdownFlag) {
+    std::thread::spawn(move || {
+        let _ = stop_backend(&state, &shutdown);
+    });
+}
diff --git a/studio/src-tauri/src/update.rs b/studio/src-tauri/src/update.rs
new file mode 100644
index 0000000000..477f18ce28
--- /dev/null
+++ b/studio/src-tauri/src/update.rs
@@ -0,0 +1,293 @@
+use log::{error, info, warn};
+use process_wrap::std::*;
+use std::io::BufRead;
+use std::process::{Command, ExitStatus, Stdio};
+use std::sync::{Arc, Mutex};
+use tauri::{AppHandle, Emitter};
+
+// ── Types ──
+
+pub struct UpdateProcess {
+    pub child: Option<Box<dyn ChildWrapper + Send>>,
+    pub intentional_stop: bool,
+}
+
+impl Default for UpdateProcess {
+    fn default() -> Self {
+        Self {
+            child: None,
+            intentional_stop: false,
+        }
+    }
+}
+
+pub type UpdateState = Arc<Mutex<UpdateProcess>>;
+
+pub fn new_update_state() -> UpdateState {
+    Arc::new(Mutex::new(UpdateProcess::default()))
+}
+
+// ── Spawn ──
+
+fn spawn_update(
+    bin: &std::path::Path,
+    state: &UpdateState,
+) -> Result<
+    (
+        Option<std::process::ChildStdout>,
+        Option<std::process::ChildStderr>,
+    ),
+    String,
+> {
+    let mut update = state.lock().map_err(|e| e.to_string())?;
+    if update.child.is_some() {
+        return Err("Update is already running.".to_string());
+    }
+    update.intentional_stop = false;
+
+    let mut cmd = Command::new(bin);
+    cmd.args(["studio", "update"])
+        .stdout(Stdio::piped())
+        .stderr(Stdio::piped());
+
+    // AppImage sets LD_LIBRARY_PATH to its bundled libs, which breaks Python
+    #[cfg(target_os = "linux")]
+    if std::env::var_os("APPIMAGE").is_some() {
+        cmd.env_remove("LD_LIBRARY_PATH");
+        cmd.env_remove("PYTHONHOME");
+        cmd.env_remove("PYTHONPATH");
+    }
+
+    #[cfg(windows)]
+    let mut child: Box<dyn ChildWrapper + Send> = {
+        use std::os::windows::process::CommandExt;
+
+        cmd.creation_flags(crate::process::CREATE_NO_WINDOW);
+        let child = cmd
+            .spawn()
+            .map_err(|e| format!("Failed to spawn update: {}", e))?;
+        Box::new(child)
+    };
+
+    #[cfg(unix)]
+    let mut child: Box<dyn ChildWrapper + Send> = {
+        let mut wrap = CommandWrap::from(cmd);
+        wrap.wrap(ProcessGroup::leader());
+        wrap.spawn()
+            .map_err(|e| format!("Failed to spawn update: {}", e))?
+    };
+
+    let stdout = child.stdout().take();
+    let stderr = child.stderr().take();
+    update.child = Some(child);
+    Ok((stdout, stderr))
+}
+
+// ── Stream ──
+
+fn stream_output(
+    app: &AppHandle,
+    progress_event: &'static str,
+    stdout: Option<std::process::ChildStdout>,
+    stderr: Option<std::process::ChildStderr>,
+) -> Vec<std::thread::JoinHandle<()>> {
+    let mut threads = Vec::new();
+
+    if let Some(out) = stdout {
+        let app_clone = app.clone();
+        threads.push(std::thread::spawn(move || {
+            let reader = std::io::BufReader::new(out);
+            for line in reader.lines() {
+                match line {
+                    Ok(text) => {
+                        info!("[update][stdout] {}", text);
+                        let _ = app_clone.emit(progress_event, &text);
+                    }
+                    Err(e) => {
+                        warn!("[update] Error reading stdout: {}", e);
+                        break;
+                    }
+                }
+            }
+        }));
+    }
+
+    if let Some(err) = stderr {
+        let app_clone = app.clone();
+        threads.push(std::thread::spawn(move || {
+            let reader = std::io::BufReader::new(err);
+            for line in reader.lines() {
+                match line {
+                    Ok(text) => {
+                        warn!("[update][stderr] {}", text);
+                        let _ = app_clone.emit(progress_event, &text);
+                    }
+                    Err(e) => {
+                        warn!("[update] Error reading stderr: {}", e);
+                        break;
+                    }
+                }
+            }
+        }));
+    }
+
+    threads
+}
+
+// ── Wait ──
+
+fn wait_for_exit(state: &UpdateState) -> Result<(ExitStatus, bool), String> {
+    const MAX_WAIT_ITERATIONS: u32 = 72_000; // 2h at 100ms intervals
+    for _ in 0..MAX_WAIT_ITERATIONS {
+        let mut update = state.lock().map_err(|e| e.to_string())?;
+        let intentional = update.intentional_stop;
+
+        match update.child.as_mut() {
+            Some(child) => match child.try_wait() {
+                Ok(Some(status)) => {
+                    update.child = None;
+                    return Ok((status, intentional));
+                }
+                Ok(None) => {}
+                Err(e) => {
+                    update.child = None;
+                    return Err(format!("Error waiting for update: {}", e));
+                }
+            },
+            None if intentional => return Err("Update stopped.".to_string()),
+            None => return Err("Update process disappeared unexpectedly.".to_string()),
+        }
+
+        drop(update);
+        std::thread::sleep(std::time::Duration::from_millis(100));
+    }
+    let _ = stop_update(state);
+    Err("Update timed out after 2 hours".to_string())
+}
+
+// ── Public API ──
+
+pub fn run_backend_update(app: AppHandle, state: UpdateState) -> Result<(), String> {
+    run_backend_update_with_terminal_events(app, state, true)
+}
+
+pub(crate) fn run_backend_update_for_repair(
+    app: AppHandle,
+    state: UpdateState,
+) -> Result<(), String> {
+    run_backend_update_with_terminal_events(app, state, false)
+}
+
+fn run_backend_update_with_terminal_events(
+    app: AppHandle,
+    state: UpdateState,
+    terminal_events: bool,
+) -> Result<(), String> {
+    let bin = crate::process::find_unsloth_binary()
+        .ok_or("Unsloth binary not found. Cannot run update.")?;
+
+    info!("[update] Starting backend update via {:?}", bin);
+    let progress_event = if terminal_events {
+        "update-progress"
+    } else {
+        "repair-progress"
+    };
+    let _ = app.emit(progress_event, "Starting backend update...");
+
+    let (stdout, stderr) = spawn_update(&bin, &state)?;
+    let threads = stream_output(&app, progress_event, stdout, stderr);
+
+    let result = wait_for_exit(&state);
+    for handle in threads {
+        let _ = handle.join();
+    }
+
+    match result {
+        Ok((status, _)) if status.success() => {
+            info!("[update] Backend update complete");
+            if terminal_events {
+                let _ = app.emit("update-complete", ());
+            }
+            Ok(())
+        }
+        Ok((_status, intentional)) if intentional => {
+            info!("[update] Update stopped intentionally");
+            Err("Update stopped.".to_string())
+        }
+        Ok((status, _)) => {
+            let code = status.code().unwrap_or(-1);
+            let msg = format!("Update exited with code {}", code);
+            error!("[update] {}", msg);
+            if terminal_events {
+                let _ = app.emit("update-failed", &msg);
+            }
+            Err(msg)
+        }
+        Err(msg) => {
+            error!("[update] {}", msg);
+            if terminal_events {
+                let _ = app.emit("update-failed", &msg);
+            }
+            Err(msg)
+        }
+    }
+}
+
+pub fn stop_update(state: &UpdateState) -> Result<(), String> {
+    let mut child = {
+        let mut update = match state.lock() {
+            Ok(guard) => guard,
+            Err(poisoned) => {
+                warn!("Update state mutex poisoned, recovering for cleanup");
+                poisoned.into_inner()
+            }
+        };
+        update.intentional_stop = true;
+        update.child.take()
+    };
+
+    let Some(ref mut child) = child else {
+        return Ok(());
+    };
+
+    let pid = child.id();
+    info!("Stopping update process group (pid {})", pid);
+
+    #[cfg(unix)]
+    {
+        if pid > i32::MAX as u32 {
+            warn!("PID {} exceeds i32 range, using direct kill", pid);
+            let _ = child.kill();
+            let _ = child.wait();
+            return Ok(());
+        }
+        unsafe {
+            libc::kill(-(pid as i32), libc::SIGTERM);
+        }
+        for _ in 0..50 {
+            match child.try_wait() {
+                Ok(Some(status)) => {
+                    info!("Update exited gracefully with status: {:?}", status);
+                    return Ok(());
+                }
+                Ok(None) => std::thread::sleep(std::time::Duration::from_millis(100)),
+                Err(_) => break,
+            }
+        }
+        warn!("Update did not exit gracefully, force killing");
+    }
+
+    #[cfg(windows)]
+    {
+        crate::process::force_kill_process_tree(pid, child, "Update");
+        return Ok(());
+    }
+
+    #[cfg(unix)]
+    {
+        let _ = child.kill();
+        let _ = child.wait();
+        info!("Update process group force stopped");
+        Ok(())
+    }
+}
diff --git a/studio/src-tauri/src/windows_job.rs b/studio/src-tauri/src/windows_job.rs
new file mode 100644
index 0000000000..411449fb83
--- /dev/null
+++ b/studio/src-tauri/src/windows_job.rs
@@ -0,0 +1,72 @@
+#[cfg(windows)]
+use log::{info, warn};
+#[cfg(windows)]
+use std::mem::size_of;
+#[cfg(windows)]
+use std::os::windows::io::{AsRawHandle, FromRawHandle, OwnedHandle};
+#[cfg(windows)]
+use std::sync::OnceLock;
+#[cfg(windows)]
+use windows_sys::Win32::System::JobObjects::{
+    AssignProcessToJobObject, CreateJobObjectW, JobObjectExtendedLimitInformation,
+    SetInformationJobObject, JOBOBJECT_EXTENDED_LIMIT_INFORMATION,
+    JOB_OBJECT_LIMIT_KILL_ON_JOB_CLOSE,
+};
+#[cfg(windows)]
+use windows_sys::Win32::System::Threading::GetCurrentProcess;
+
+#[cfg(windows)]
+static APP_JOB_INITIALIZED: OnceLock<()> = OnceLock::new();
+
+pub fn initialize() {
+    #[cfg(windows)]
+    {
+        if APP_JOB_INITIALIZED.get().is_some() {
+            return;
+        }
+
+        match unsafe { create_app_job_object() } {
+            Ok(()) => {
+                let _ = APP_JOB_INITIALIZED.set(());
+                info!("Windows app job object initialized for crash-safe child cleanup");
+            }
+            Err(err) => {
+                warn!(
+                    "Failed to initialize Windows app job object; crash cleanup will rely on explicit stop paths: {}",
+                    err
+                );
+            }
+        }
+    }
+}
+
+#[cfg(windows)]
+unsafe fn create_app_job_object() -> std::io::Result<()> {
+    let job = CreateJobObjectW(std::ptr::null(), std::ptr::null());
+    if job.is_null() {
+        return Err(std::io::Error::last_os_error());
+    }
+
+    let job = OwnedHandle::from_raw_handle(job);
+
+    let mut limits: JOBOBJECT_EXTENDED_LIMIT_INFORMATION = std::mem::zeroed();
+    limits.BasicLimitInformation.LimitFlags = JOB_OBJECT_LIMIT_KILL_ON_JOB_CLOSE;
+
+    if SetInformationJobObject(
+        job.as_raw_handle(),
+        JobObjectExtendedLimitInformation,
+        &limits as *const _ as *const _,
+        size_of::<JOBOBJECT_EXTENDED_LIMIT_INFORMATION>() as u32,
+    ) == 0
+    {
+        return Err(std::io::Error::last_os_error());
+    }
+
+    if AssignProcessToJobObject(job.as_raw_handle(), GetCurrentProcess()) == 0 {
+        return Err(std::io::Error::last_os_error());
+    }
+
+    // Keep the job handle alive for the full app lifetime.
+    std::mem::forget(job);
+    Ok(())
+}
diff --git a/studio/src-tauri/tauri.conf.json b/studio/src-tauri/tauri.conf.json
new file mode 100644
index 0000000000..49819f8a25
--- /dev/null
+++ b/studio/src-tauri/tauri.conf.json
@@ -0,0 +1,66 @@
+{
+  "productName": "Unsloth Studio (Desktop)",
+  "identifier": "ai.unsloth.studio",
+  "build": {
+    "beforeDevCommand": "npm --prefix ../frontend run dev -- --port 5173 --strictPort",
+    "beforeBuildCommand": "npm --prefix ../frontend run build",
+    "frontendDist": "../frontend/dist",
+    "devUrl": "http://localhost:5173"
+  },
+  "app": {
+    "withGlobalTauri": true,
+    "security": {
+      "csp": "default-src 'self'; connect-src 'self' http://localhost:* ws://localhost:* ws://127.0.0.1:* http://127.0.0.1:* https://huggingface.co https://*.huggingface.co https://datasets-server.huggingface.co; img-src 'self' data: https:; style-src 'self' 'unsafe-inline'; font-src 'self' data:"
+    },
+    "windows": [
+      {
+        "label": "main",
+        "title": "Unsloth Studio (Desktop)",
+        "width": 690,
+        "height": 480,
+        "visible": false,
+        "resizable": false
+      }
+    ],
+    "trayIcon": {
+      "iconPath": "icons/icon.png",
+      "id": "main",
+      "tooltip": "Unsloth Studio (Desktop)"
+    }
+  },
+  "plugins": {
+    "updater": {
+      "pubkey": "dW50cnVzdGVkIGNvbW1lbnQ6IG1pbmlzaWduIHB1YmxpYyBrZXk6IDJENjJENDYxMTQ1QTYyOEIKUldTTFlsb1VZZFJpTFY1aHgyRkRqZUdjWGg0Sm1BVEZoMDVWME5PdE42bjZiekFRc1Fsb0JNbmIK",
+      "endpoints": [
+        "https://github.com/danielhanchen/unsloth-staging-2/releases/latest/download/latest.json"
+      ],
+      "windows": {
+        "installMode": "passive"
+      }
+    }
+  },
+  "bundle": {
+    "active": true,
+    "createUpdaterArtifacts": true,
+    "targets": ["app", "appimage", "deb", "dmg", "nsis"],
+    "resources": {
+      "../../install.sh": "install.sh",
+      "../../install.ps1": "install.ps1",
+      "../install_python_stack.py": "install_python_stack.py"
+    },
+    "icon": [
+      "icons/32x32.png",
+      "icons/128x128.png",
+      "icons/icon.ico",
+      "icons/icon.icns"
+    ],
+    "linux": {
+      "deb": {
+        "postRemoveScript": "./linux/postremove.sh"
+      },
+      "rpm": {
+        "postRemoveScript": "./linux/postremove.sh"
+      }
+    }
+  }
+}
diff --git a/studio/src-tauri/tauri.macos.conf.json b/studio/src-tauri/tauri.macos.conf.json
new file mode 100644
index 0000000000..202c9fb505
--- /dev/null
+++ b/studio/src-tauri/tauri.macos.conf.json
@@ -0,0 +1,11 @@
+{
+  "bundle": {
+    "macOS": {
+      "entitlements": "./Entitlements.plist",
+      "dmg": {
+        "appPosition": { "x": 180, "y": 220 },
+        "applicationFolderPosition": { "x": 480, "y": 220 }
+      }
+    }
+  }
+}
diff --git a/studio/src-tauri/tauri.windows.conf.json b/studio/src-tauri/tauri.windows.conf.json
new file mode 100644
index 0000000000..a68cf6351f
--- /dev/null
+++ b/studio/src-tauri/tauri.windows.conf.json
@@ -0,0 +1,16 @@
+{
+  "bundle": {
+    "windows": {
+      "signCommand": {
+        "cmd": "trusted-signing-cli",
+        "args": ["-e", "https://eus.codesigning.azure.net", "-d", "Unsloth Studio (Desktop)", "%1"]
+      },
+      "nsis": {
+        "installerHooks": "./windows/hooks.nsh",
+        "installerIcon": "./icons/icon.ico",
+        "headerImage": "./windows/branding/nsis-header.bmp",
+        "sidebarImage": "./windows/branding/nsis-sidebar.bmp"
+      }
+    }
+  }
+}
diff --git a/studio/src-tauri/windows/branding/nsis-header.bmp b/studio/src-tauri/windows/branding/nsis-header.bmp
new file mode 100644
index 0000000000000000000000000000000000000000..7eacdce7db5026855113aa5f03c47663d1e17561
GIT binary patch
literal 25818
zcmeI2M`%?|6vyAoYkB>pM+JLBK}E5|0vb>h4T7NJ!i|2eT)0utg{~Ag7ElNZE-XP%
z(S>}9pkT*}9g&D2*s&q@{{7~jWHPylFYkru{S0%txlTEA=FIt@|CyPX>YDn0e9kk9
z??8TP`KjW^@2lkN^VK!<JC%k{MN!(mf3*jEzQ5XhYdjvX82<ff$YN+Qw0o)d9<V&L
zJoG*a?Xkx4(DKmo(0dP99$FrHABA>5v^=ytv^@0Q1D1!Dhu%k_-487fEe|aZz4w6S
zq2;0XQE2x=%R|dU%R}!y&|V&vrBfxTL}@Bno=K6-_?Bf-Ql?5-Rhx&nOGb?<N|&aS
z=qX7hkvFrVXfRupd?qFGd|h2;wJe=PK#Xaew3(-wSDDv*MpFZ2G!!S}MTtN1Cga7)
z1ZKNz$g!F~hAcf1Eer+%1%7`)fj<%oW)ktja3~lkpq*&}BoquHfCEe`Qn4tF$cz6{
z@n|Fz3KXEj&r27>#ji!lcqkA+Bk+m_+K~?jgGeLHC}w2crhdxBsRZpxXMrDW1R8ds
zIUSFgl2mC#TdT2v+QM)cC4PUypZ-y3cpLv9ke;T-Mlm{FVj|uo+;X^XQ64CW7e-n&
z$uz!YIe!dekx1jj4ga!O9xAa&SR5%;m*;RQGWZ79CY>@>h6ya9i$$OzlZdG<zh6)S
zKs1Yh`DLIv1+fD%T0r~(nzSTiD!BAZG|`CJ4AG5JKp_^25)3FTEDRVrX(u{)WPu{4
zR1s6nB?BU-0;tKPR)>c|p@`v9Gjd(F{i>QLhB#Hs$7-C!iusD4FfexvNmHzjaDZ||
z9(;p8Xuy9|AvR^fu(BwfbX-mg%z=&c5)R^NzM898qDJ|A(MXsmAkU*p#B#)t)rmYS
znmTprjvYHroH(&(&z{+{XX9y<XEK@E+FDxr_3KBK2g#b6n%%p1A3uJ4@7}#g0$wZ@
zTfct&zs|FN|Nd30R`u`SAK_#&xq0*Eef#ze9H{>Kw`<p~%a<=deE9IOW5+gb+&FC5
zFs0F1W?p$UPYiJeli;UxJgUQjoGE!HXAD*JOVX$;R`hU~rVtQrm>AMUUTwKDWMopH
z6HCa<h!MvOk(PWhlsH1k+_`h#zJ2@d-Mb4HE<Aho?8Ao-3l}b=y`rMx?%lf|KYkoF
zO8pr&4*<S?{rda&?++h7{QC9l`}gnX&6}4<B(7Y!^7H3U#_~|OdGqG*;lqoHif-S&
z{pHJ-s;a8W%F3fhkJ9!0`SS-49#FV<@7{z76GW|Elq@9sH%|;D)(KWKQ1F=~vAi(E
zE|kFJSRs34K;>dWFcd_ai6L8@FyBm@vL<mk<YWlP6vm=~M!n*^Fx0|OYwg;#=g*&?
zKYxBYonE<eB@Tuy5Ujks{LY;_z&mo}NajKk2GYB3-8x7G-zdYrva+&Emo5P*wCvrx
z_xA1EX@|XCx^%g5;|38iX3UrsD^}2W=FFL)Lx*<n-hJ!Vtui(mRe7GhQLm!SLc1^o
zG4mMNJ`BMGOR-lTfPzRQMv|wAA;}ha*={MgRwDrgOFINZHIcKcV-HDh%9JT<)~q>n
z=#UU;*REakx-cYXo;-Q-$B!Q~X3Ss|Ec*26)77h2d-m*k@#01JHfho%+DDHb4ea;t
z--n9VuV06&)z#HUjvN8BMT-{6@=2ud`N@+flBHR)Yq43X`N*P1+}1t};T!fk99H)d
zP9#)u3`AmoV}}F^vDCy6NpUR)5!X5-c{<cp*OaV0j1n4pxSXyqz&z8iBPfHMXWF!B
z0v%k!_jPr3g9Z(v$Aw`$9zT8hG>LTj^y&1DA3vTnd+pk_UcGt=hG0ee;K74mzI^%k
z@nh&G7*3u%nVl94=ggVoOy~o;r%#_EhOo3~YuhaRP7FC)XjsTpl5b_df`nK~*mFj~
zMjXtm?i`M%O$^b%5QO=nL@x*pg00><vS0|>9EN4Ax^M(l)+Gz*M!Hm3HS&-X2@d9{
z0g=FG!GZ-NMvPzx`=OajT^Pbc7lsojPGr>L#f#B)?%cUPefmf)9XD>ABu~C_T4KM2
z&){ELTFMYC(JO7U@H;WY7U39iFF4R5E_V_aH6L9#kE+Oko-3D91!rLhX_G@0=s{_r
zF*v{_1W+qSN?IuMqEl~iOoRbU!7BjLLWr5}(ZNEOc>VI_%gdH6;|l_-SFhf%VFO3X
zH*em2`}U0-yKURHv17-QOX(FJN-l+I@R0qG_5%kF?A*EY)vH&ZKYu0=IMdv^bqk!y
zu8f81<XcX&TefVW44;=QS;Fi*1f6MH_HsKhEXZX)G`Bl34G@G&I7}Y`p_4EQ8S$&h
zL(GSqV4`J26<<I@vOJU40iwGoFhpb6(h>)J_pb=#-pJVE;^Ix4Hj&)$H+vN5afqp}
zuV-5W5FU^u@6DPu>)^qIFJ8Q0^JEmMl7xBc)Ttgldca|1c-ZAQqq6UDBm%>;XU{%*
z^hl1iGiT1^<b+O6IVfMcbg3L@AffBmBOQt%tHfc^a8r_0KF*}@J?&iNsB$smph>uM
zN#iaLZa5pCA!F`8M1kU-(~AJZmDin1w+lmA0osU8Bpq$sEzzbx0AcQ#43|vEm#O*f
z40P2PGLcxWh6kMA@ihJhNR-Q079xRw%Lfb?09&})VPiNPW~buhp&!nZTo?QH?Moq*
zN~x%c$A=6V!YG6pC6bIPDJd~^B7>x~fGSxU@Vj;ECM~3MoMu|pW;stDqK}0V`Yy$^
z(5@RDnT9H5d};1HT1>Rj%WDpA6+<QLTt}2PV=ifDxa^1EtY=3O(C`hxMNZCciCBeZ
zLP^3wocbrfSljaGCWdAC=8s_(WgXS^D-5&f2VU9eNC;ZYz%aka=28lENgO2nKl3sd
z=wYxnKKDXr!H^Z>b^I;)TpZAD40HG2|3R`7Vc6207DMAw+|oV2wFTcpD&F?x7Hu0F
zAhbk}{Qrk;?_`DlB@a89kAL@aEQa2tn~uWPVrVgRUlcnkwyd7qJz#lgd1!g)y$380
zEf2kq!rb?b&MIknXnAOP=)DIl4=oS9kHXHn%jLc<%R|dU%R}!yV0mbH=zSFCzHf9^
TNy|gaL(4<&Jz#m*8F=^~mCb?X

literal 0
HcmV?d00001

diff --git a/studio/src-tauri/windows/branding/nsis-sidebar.bmp b/studio/src-tauri/windows/branding/nsis-sidebar.bmp
new file mode 100644
index 0000000000000000000000000000000000000000..c3bd480fc8993fa88d280c35ad105a7ddf1bc7ff
GIT binary patch
literal 154542
zcmeI52YeL8+sBb6N=q;FUP3QHs?w!O5k&q178GoN1w;WsdJT{OAtBUMdQa~VAQb5e
zQk32ivCss3zq@m{+1y<&mxw@kC!ZzE?al6Oe!qF@JTup_^V+v^=Td$e@UII0g8A3@
zxm>xP%jHHZSFW!O@2f8izdT%v%0nU`5s(N-1SA3y0f~S_Kq4R!kO)WwBmxoviGV~v
zA|Mfv2uK7Z0ulj<fJ8tdAQ6xVNCYGT5&?;TL_i`S5s(N-1SA3y0f~S_Kq4R!kO)Ww
zBmxoviGV~vA|Mfv2uK7Z0ulj<fJ8tdAQ6xVNCYGT5&?;TL_i`S5s(P{e?y>>bCGh6
zg)2A}uHvHX!BW|!NM+|DqFq(IbSXkZ4-4<oho0Jb8Wq+1ly@x5TUuu;ci<EG+=@<x
zwVP;HiH5kRcre3#j1Q*uG=CE9E$t@Up^QVJ()I-(&%tK_g~?+)@t2V0T&TeFdGqAS
z{aoJM&lP$;Z%Mm?MGEA5K5uT9LIrp{h<h<#-U4~^l(H|Ve_Hn<FK`w887JJNSkV^?
zKc6pup4{ArgZ{3C3vk}U&901np#u5xypS(XImbfc4zN?K$O|v#&&wy$IbWXKMGNEu
zAN~Dlmvt<}lW<SE@i2T)q5OII(Beg3)b}i5_aY|?=FbbE`tzb))}c_*0{Quf+|RMo
zkb2P24)ZfwH{LSoX5Kl2Gr6C8E`Q$Kjs>5$_(<jt`54Y1DGo0_uismVHsJH9G#p;c
z4^v<zUX}o1H#_aA7$NilLk2LkikonqpijKykEK6@hnLSi_ng9}n;k^+p~?rSKU#we
z^(2;eDhz|hUb&nKK9Bh4&8=(~<y!cKv?p=1W7Khy$7Wb+k0%-o1D;Uaga;^4pg^8H
zY4_xOUiF8!L?7`xckbLKJ@e!d9TAC4F3*=Q-}BF>L8>3&CeQYmU&3P^_0I%Dr-B85
z8b}46{M9g`aP4U5W07lw4snA-c%eTUl97yBMjp9?G7_8$m$EAelnfZURdcg*E38n%
z4|0uk^De_iFv4%H6(H~=^x^63U&zmBfqHHwx^V|{t{EyOT=Un7{&3f>UAtw=mL(-6
ztyr<5XV0E6!p-gM?0kKFIa9G>Mdg8b$lSR*ckUb-8k(4xxN_x6dcyIGFTUvI<+WM;
zS+i!%=+UDqRjNei!i5XZnKNhAs#R60DBF}3D^~35ufL9nh=_}en>A}z?b@}KJF0)?
zlh@})*ufJGte~fgExH=qct{($<_|$X2xSG8+(bw&cTi+4J=bV;cuhk<_n~y=yif()
z5R0I6%a=zO0&dGk*v>U~Ky`BmetPxl^~WE7{Q2jfJ9qB9eEIUNTetf6@6UNxSJ&gm
zkKew1yFmkGUpamt`o@hL4<0-?ckbNZfB*g0Uw`%P-MdhsLVNe_{pX*5xRxIphYlU8
zTeq&gz5S0r{&?@+y+(~1xw*N;#>Vp2)vH&}o;^$B<jIrGnl%%*buOq_!zX(P*+Dk-
z15cyFcCN)=!A;PaT!rEiN)a0h??|PPYXl7~AuNJnbQuwV8leghb}r#1*8&tLX^Z=3
ziEFJhqB?o<<kZyEzJ2=^En4*3Z@+~Iqz{30adA0%^eDr?)6<jBq9+o~yEA6YKuD2a
z+DNmrv-9rVyP+O2FITSI(xpo|k2Y|3cRzUWAS0(?!-n5{^9?6cQc`Nws#U5~DSv-|
zaqSB)C`t@n`DBUa(-9)|N%EkFwTH<b*XV!#BJL)BAQZtwAtUK}u2J#~0}dJ<u4&d^
zdl;mi2l)3~m$o+yDxL>r+qP}ni4!M=hldNn3<?UuMw@Vru1-!)e)#ZVr%s)?3aPwv
z=gz);`^uClvuoEb<hMnO7My?SrI#2Ur%s(hv=1COfb6$y*)l3B3UWXA-~;jJqrlM*
z7cX8E)`7oQgZD`_{tefTs#7QIHJl2y55>qM_7~HPcq)HadaiNBh0Nnn36mlIcE}Y<
zTd0C8agB=(!iv34^TX}Lpc2TpZ{J=}jBLVDaB#4PhX-$&a9yxq!L3`jqQE<L?8v)K
zn>Izg@87?_Y}vAcYdGb6wQAL_U%!6g!Ue=xaNVX&8$1EHe&dZd#3w@WrAwFSMrU#6
zNl~nQZkD*Vdm+ExbP6XKbq%MuZTv-yUNkc_A<G=-3fg+E#Q@+EI`dmG@kV=%@)x?D
zal%P1B)p;<I^6iQY4(~?p(rBtS{kT*dJ^O4wUaWaR0n{!@T?#LmkGgy$M@cQuU@@+
zoX1}4b+`%F$h8UA&6_vps!u-oguCt7v7>zX@<N9<Y0^X}ZGH={4}Sptz&LSoa^e!6
z#Q0Y{F{_;=u6bsJ4k6+Yd~t?5fVn|4=gI|BfDEUi4{n}bhYLTP@xZ9yhno;Rq5Qdt
zPz(HrFI}vNat4h-AI{)ZaG;YlKU`ePXYnb5XwEPcm7!%k7?#>Y3U>DF*)wd|Fn+_-
zgb5R7&YX!0ee>qc`}gmo+ZQfe*tl_HbU5z{xfVJc$wjWQ*PLIwc5Pr_;IF^_diU;K
z1`NL0;lqa+4`_9+MRlO(@!tLX{Ai;eh71|PgYm=Q)E?|fHH=)RlL*CLtAVY(h_8*_
z6oMn3T>Fh}Rb<n(D55PqE8appwQyCrgBo88-wdu97*$-0!kv&}?N&4(m$AXEG>-J|
z!Ysa@@WXi$F<P_^oY4|B8ki#PrjLVzgO867N)=wPp|FfIcI?<O%r3<71LNr1?%lg@
z+_>@DwQCr6u0lnlsJCp{Qo3|$bOL?&!3yJ(W0!I5;CkD(ZRgLQ7jAr)E?w{dxhLKl
zHy=85sBp{>YSlSO8;9BxKF$Vc0IWed8zlm?Im3A+jCHlc2ogLa(YRRTOyC#55jXni
z<5~14bclE9K&lNw6$7fX*9zCFKZXFJv>EO2OJ{}_c^eT~GZOIRxm;*qej~XYXoHGQ
z2Cflj@#I{~9mu)i;hW<o{tz)9pr&%=%19rfG5L`E`SW8*@u0LH{Aa@L3Kc5QC{m;d
zSE15t)~v}@bml72ldBvZ9rgF54?V>h*vC3BNQxIPF3vC}0Zp%P%%Ah|oZ%S&ghT%@
zCMQi^BK62J%<3=Eo>shz0CG>mNlRRdx3~j6P41?AoSK1WbO9IRL3=`aLM|s6gqQ~F
zPa-~oJ`6cwAO!f2^Ecy=F=-Jf4Y6mckuld+{!H-ExYn@JbFDaLMnz=3!Q}+?r~(ub
z8sPC9<oaNl2e$rf&)U1$<r;SI0T5wxuG8hiZDU-s9o+3P914jxkH%+*@oaI;pRX~Z
zL6JSV)-IQg$AADFA*38w=YR8Y7+=I35743~_HXX|^k1uueA$o)NCYGT5&?;TL_i`S
z5s(N-1SA3y0f~S_Kq4R!kO)WwBmxoviGV~vA|Mfv2uK7Z0ulj<fJ8tdAQ6xVNCYGT
z5&?;TL_i`S5s(N-1SA3y0f~S_Kq4R!c*+o9uL*fb1SA3y0f~S_Kq4R!kO)WwBmxov
ziGV~vA|Mfv2uK7Z0ulj<fJ8tdAQ6xVNCYGT5&?;TL_i`S5s(N-1SA3y0f~S_Kq4R!
zkO)WwBmxoviGV~vA|Mfv2uK7Z0ulj<fJ8tdAQ6xVNCYGT5&?;TL_i`S5s(N-1SA3y
z0f~S_Kq4R!kO)WwBmxoviGV~vA|Mfv2uK7Z0ulj<fJ8tdAQ6xVNCYGT5&?;TL_i`S
z5s(N-1SA3y0f~S_Kq4R!kO)WwBmxoviGV~vA|Mfv2uK7Z0ulj<fJ8tdAQ6xVNCYGT
z5&?;TL_i`S5s(N-1SA3y0f~S_Kq4R!kO)WwBmxoviGV~vA|Mfv2uK7Z0uq57j6fAP
zJ9$V1BmxoviGV~vA|Mfv2uK7Z0ulj<fJ8tdAQ6xVNCYGT&kh2vc9mU}qgo03Mzx*0
zzfp4Vh`NJ6?9!}e#VW2vE4vh};wn>uIUE5PR9UW8+`dsA=kC2q3>jB(Qb>(C(RKWj
zUJ6<Nh2O041N(Go<WW}bCK=C}bi=ZXa@48l(7uEF$D_(k3iFs7SIaNCj&dZ`^-HR=
zB&O^7)qWdREcTl{ZopfwHS#E<$hfOyJm)YjSa!AZtmfFGPqA<2mY)?-)i<fSZ&D54
z#G1Z|wdQjq)t;YJ$3L;b^7Y?rS``?vHh9Cz#q(#4eXm!?29-*xbQgVg?ANn8W2Gu>
z&*~0uzUMxAaXIgp%5xK|%uT2|H^F0WLbZ7b)#oK@j6-*=#nD~Xt(p(lQ5(V{*N3iO
zv3Sn3(Qm*0s%NDVlJTc^w5xWxZe@p_{ai;aEIlixyiY;}pSX%Xah2x8SDq7JWp2EY
z@x<!BM#lXT>W8i$hG<7`ToD(sI%e}q7+(__Fmuw^eY&=;U$MCAwn-U(YP7oWzDn4&
z>)<rZySP_Wnc4AWXUCSC9ar8vu7bBpw~tEq9F1;8z+oJfUSnZ&yEQ9)BG!h)N3KbZ
z-IN%;K6=y2&1-^I1o}=JJG57aSL&8?myY^VA>*pjeyO%Y|F2vpuP-qxuGGw!(z9YY
z%FK>c828SE@fyAfwFB3Gi1v=!xI8&_<M!m}l*EXHsCAL+!#EkdaQ1|egL<@UQm3q&
zWc*1ouEfiB9lJP<Sm5pzU1DZT39p!vGnGRyUd9UJI+b3{FQ!S@iU}K61;%b(1<|{=
zC+^x7zcoIbmm60Ht_t>FGJo2buLgE&)3A0KSIPL}X53(%?ce^)W%9aWGh>T;MRSx8
zhf24eaRj{V>^OsrTQuSUo4TzGnY(FCP<-Ut9b01d?M~UhJ0*2%Tw?Twi1ndsLl=fF
zp0mJb;>geYcWz$4R%sW>_+w&R)!R)PI1HKQIz6hGS4{C43eno3XB+`9HA~Uz$hZ}i
zUUgo4%_XZoS{=L~d~HZV^!lAC@rU-M{%~OT-kn=h5+h^7S8rIkbj32?zy&jXXMQ()
z@Y|hUsaI1n{$EFPu69-3>|Sl}IC7DjS4^=P(b{2&@pLNPsu549;v4<avfxpxLjoe!
zg(Ssn*tIPIu8;kA;PC!kw3A{sMyv~66A}=z$Y-(dv^mp94f~`|hh}wZmU5Od{-4OW
ztDQ$NyUslv$F6Ymiglk6Ee?%t8k%%lQRx}EZI$P5Y#X#>3WP!{<YKsgJox>|V?V(7
z{@vTQB}HihzRZ8-!nu>Zr;PaG!``npu34j`lVto+GVZE)Vm<mfO<3zTBgTEYfoOBO
ztuSt?($g97_)2r)s`#z$5wyrlmF_gUPn|sa%Za1#4*N;5n{d&E1z$3MI!1ifq~V_p
z=+(AS4axYUq1)B2dP%!CKX96~-fc#-8$@e|QMw^Loq$`CalMyr(TG=^6I;zc^xdF^
zvsZ^K2wxYHXi(;RPM!Se-073Z_`&aY;Jn2}tlfl*9^@AkFniIwss6Ladrkam@cTX6
zHmqK~grk)4?31oabg%avCvS9Bh>kKI*<(BtBVJ)nRILTeK88f1t<7tL6QXd{;;`Dk
zoH%mf+^MstPW<%4{_l5gAt)p3vVO&q<;#4REl^at->h*n#t-}S-Pc=rdQ>Ye8P9sQ
zO|{Xl_j8=I9-?u)IgDJ3h6Tp;0&Yd6>*F%L5ijo(QE$QGA%P2Ktqk%*enlXLM4$ZW
z2k5?Z;S5^+@PXZ`&!z@tVGF~S%w6g?BVf*?`7_5(ANS=a?{sTfzp7+B3j$7f<!0CU
zb%*h*U8YC7s6-38(=%>?ZWEPmg>iEuPI9%p&t^~mMV|-y&s-5Wf8(lUcxBrXBZ<55
z)6Sjw`O2m97tftO@zdc0-)~2yGZ@eSsQu8zJ{a);pGkyf)5d=BQNOM&>Q(V5<{-6t
zrleCz(yFz?m{4bko)%>|{srS%ip$FRY--@Y@bhJUUdxxxLsv6DvL$Zwo*hXzX{Uca
z2HjULLpZX&fA20s7_J!cVE>Rs-okD3_4;n=cb^Y<tMe<hD@w+#&<)WI>)H?ZbJ7zn
z7@r>XZ)98_ml;E|bVl5mT#YO1v$?+CqCrdN&p<`54O>VKjZ7Nm`{3SGIELx#S1&<2
z)H49~>`2ClGg*cnz-<d&=uITPz<VN$PZ{&sfVVom4CC%n#?9=wt6lAi_Mc32oEGIg
zEt11UIgFxhWL(qdCMw;cqi&-SS5&$>fug3`;>vo5*Pg#*z!KkSWKq!4Oo7EktRdD$
zmX99ZcjlMlS1(`q_1YCg9Ov!u{@ui7h=25^6{OlQ_j<-Bj~?=V?~YAtl&|V$FAHD{
zVy^mQy$3o>33r+r>CB<htumgTZb*MbD!s(aXvI}i6RJ#qmsaRDBv;FLM^^J)_Vyy*
z$-x1$R|d_8M5s)T+k|g+Xx~o7VLN;3*K3!5y?*85`7^|2oa8NLNKyOg823?ReB!9V
z{a$ZN$&-Qc|2ot7bl)HC-KMSm*p*JxqTt#<_kYB=IyEL_x#UbH%c4sfC|2ke0<J{j
z-qBU(g>?^@KW3S~V!qd*&)0<{M6TVg3iy#jdkMfVUp#x``c-5cUmdsY(B9NtDe>tT
zSM4~S+dQu^-wprh&CbmmR4GZy_y27bD@50?W<PYU<Fsf&wDwyOa3kZHs&o^ro>A_l
zID|4gHaoWbg5;M%cXnBt(qc|rB?IFIx=V?<Q95kf!<^S5{H~D2KC6TMvCE2p$8N%c
zANYO;E;`XSD*eW<*AR09<NH#XAypXPq}p-CbMwIgfbrQ=Mt(K;ovv*g*Q;1UQD|z=
zVa0)Y*4|RJcb`EHsyJ&b>l-#Qo<0UA61NnWF*^p=tM<Hk>~h47zka#<_aA@Wx%~6h
z_>KDqc*mAkReBlaH*y_UanC9LfByLgv4uE1?U04DnfY3`d=cq2EO<&n1TOmFeLGJa
zJ%n?9^~%NHjEtk?6|KHK0XrVEc@@688lBBSMflC0Flo$}ANTLov4v;73dR3lF^*8T
zXkq`|a!q^dS=P5T0XH?`m~XPC#wkwxG56_F97gKXF|K%PYxnm1^X?T$)c^c_|JL`X
z0{jwds{*dj9k#Q}?R&q{=i$SL;o+MHzTa=s=%IlNyuy~uLBKJ}R3|1ztw*JQpPI}B
z89Myj>7QX-m2vF&>7S4P$a%&Cqk^#!zBU-g9&@nN&&zx2=#fJ|`k-I;4$bS={eQx^
zt6j~qc8c$9s=f6rYi+%Nr!lV1i5OgN9plABn<TkW!0A)mKe_&;n+FBU*REYVaNxji
zzx}3>e&TX;K=Mn9&lX*3=g)J*`J|*IH#fIjxpI{#=JxJeJr?*(rcMdUs7OR$QR&Qw
z?nsW_yK{?R{OswIS1zAd?D+L7m+;|E9zAmCdwh6`oyfjMZV213atZ00fVopAjv4ml
zp#J^)cI(jGvu^ofib{WWwK_uGp^L-B4Ng-HE}2HFm4=>iB`8y;LJV9hv_`p6ltMKr
z#VWMb36T;rq7<T$_qZb?1%v0$pYPPElcS?!%a$#pqM{I9aUOjhu%#iz#$rft{-cjR
zf@txRFJIpFZCi#e@ug}7#wkZ!7q*C?j1U{fVG|iAk$(KhK|-_37tRV_op}7rsS_A?
zOgu#q$wINO2?|^4>pgw!sILZp@oE3}`*!c}O8sY+aW$a+e3p$ut!H1*?KCAqvGOyb
z4B1hbr2-r_f4&-b{>JX37uU!9@D&RI=%AJ=6=wjPxxP(sI%UcfBZ0Y`oSf#)o%{FS
ze~Y(HU*DQ|bj-c`fAW$cQLS1vttXA=o_p@iH+!X~BvG7(K3~3cE?E?exWf3Rl}wWn
ziKDm4y&U~v9|8FJv%lcHUB7yn-?RzP5PyaQ-KdAzyf$Qs|Ln<QzZo&~(_upf41A|&
z$L5}O%DX>fGLBHc*3)6~COwInw8^euo%!t*U2;LPXGm%%TxBAr>5;A+K|5ZBV!@a8
z`}DOfOHy8eZiIT}?w*3dt5>hquU}u!V8Ma~XU?2?|Neb(;r$19MPuE%b<aQlyuPQP
zyKC33TauIFqr=CI8tONDB8+3iRmQ12Urtr()`SQ|m~=YK9zS}BJR3Q;tCudoI3YPN
zNwkt$SEBN`&1+XIn>%arxNpB2HT<)$zx?>4{%>|{(V%u&_h%&if>2km&yrB9M7!~O
z>9&?fF08wG>nz0me*66K%h6#wyRu^I{Ea;t#Sb1l(ApPn>{*=hGJ|B-&-28|&6_vB
z@WKlw3_|zRsZ;OVxuac(0v|YVph-9FrHiX;`SRrp7b;Y}YNdX?yH6YcO;EtB)xmy5
z;)G@_F^CIat=1$*kytyhXZw$bzCUsFFf3!kvE%4<EIVTU<H0>ln8rtM3|u^a#^kZ%
zM-TsQ<j_&W2Y*b)t$Bkw6szZOJC0Dde@(U2O6J6t!x<F~(Sh6BT>ayyMu%wJynXKE
zmAHrhC^i>gYxeBfZ@&4akB`rve=2&IpVQY<mhb9*{m)~xtuTlP4<0=D!i5X!MgN>T
zcdkm6DjK;~8o6`lc5`uR-n8MT1NzOHGMY3go*UKXIB!&;Gbu*!wIewO-L8;+a1REZ
zx));e^Jjh`UZahe?@8UVaZT9#IbPEyeK%?B$O&V<9y8*LK?C}9GRSz&^xUX9DP{N3
z_$*jzRj6GtxTmgd5k%a&b?c{}e)|3Q-^EMu8y-iE8ih%Pmppm$3?Dw+D5nq601qE3
zDw?axlqq9{xP}cIE?Kf<-@bicfBiK#vC`B0t$g|Nxx2YEYEbvRzHbCCo`(@fu`^}1
zc}-x{hUG+FgzDRqW03THySL%RVbclL@YPTM{1e8VbX-zgWZ;qo-d<CtPaZdA{1`|d
z`|X!QKIz?|WlfLb4mr+IN2r@Mwk4)clXv`W5@?$aeE8tuJ@|s+Pd@p?-QB%Mj~<sU
zUDD|O<(FU3s~QFQ^XH#1VS?s#iEd-YjL|w|)X1ARuY-dFzccD!%lSc{esb-~rBgp2
zV}1nNy(K;z0gu_dB63quT;wXU>)VrJc5h4Cw`=>My}OSb+Q-b<@uLTi{IGjlO2qON
z^L!VK^j<L7d&#>#VLf~|c9<H}sHrDQQOxFyp6={7$)f*#WF=9SMzN;e;d)j~860fE
zS3p33AObdtm^7|KLPAUkdhx{<QR&FL=(Bh4-cqGX*}~Mn?^B{giGv4~SRUu$&Yj!$
z?%n<4_nVh5pWeJ}bYSG$%cBObjv2l#e$>XqvEf@LL~ffHxnn|j>WKBJgI8|vw=}s^
zKtl5ciH&eR7i?(|u(iR0<OY2{D5u!&r=HxjCv>%|Q^|hVyeuei7-!mR@z!SKGl33^
z{PN2$3og)^>({RrJR-kuz4ewkM@5SkWo+ELckkz)e}47VSIs;8cV{pnckkXEKYslG
z{`bG5M-6VqPnR~c@Lwz)WksQG)v#fmIPdx;*Y!`X=b!9JxB<%?3sYViHMd5svW_`O
zjiZHHwYHzIE(`e+h=%U)gM)<kwRP*(B1MXrbBz!cD^|=RanRklbLU#MYFT{uzdMO*
z7Z(>NGzsn^J_f_`_PyU$?tWu#yoYZ>jd`Yv#5KCv2m-D->Z^=52-woVdvm>(FSCX;
zhln<dINa^}4skHfU0P7rn%>qLDDY`fu6s`}67oA^#tc0>*k8^gwu1%@(x1<28#k;_
zp+c^%u4T)Xg(D9SkII!R<Hb2UJDUe+Hr^by^UXKk2v?BtnUIhm{)8PTy;(0zc23<B
zz~Z(5V&4zJxS(4R@V7rIms8{#b3<7DrB9X=xHCkv*!JwNsQ^W~!$eV&tp4)LFZ=iJ
zPk?MA1zI<(Ika}`)@}6Y(b)Rr<m4ZI_~F#4Q-oc(d?d|q_Q;vX#>V>l`xB16^2#ev
zuU&3Sqjv4u#fujgJwN;GGr=`E9})sA9ffg=+X5*2ek3aUeyBUW7#I)O>N&!@TFo*!
z!y8vT@Fx121S@i_=xq+7EslUKjsCcE5uk3|xY3|N16x40(uX*@Zr!@Wh7F5|h&X%p
zEGAV@km>I~{`dp$Zu<1;O`A3?P@sU7kI=s*VnEKoyNDW?OJ%x^1*8z2@wNbIJH4ok
z*Yn>}f5w{HO*~m9oCDN2581W1C^WLdGV_~C;ei@wIg#%o+6qbSEYh1NPMj!NvZNl#
zEVb?I><H6VuU>umvf}1t`-dUNP1?3?i<Y+i0Z=1a)UfLpUv`0j=!`dl(C_q8TXEQu
z>duR++x7L5If82z;=MOKa|I63>OtyTNtYqr8}@%7G9lZxZPOex+d-x$8opPrURX?H
zb|u?9r(JmS=FJ5Q7F4WQ(e@`nnTYF|VZmM_+UmX^>9+nz+ZModZy&`!fB%beX+h0X
z9Dh+M`h~YGTpJ3Cgn+|!><_~QL`+6xNt6T&sC~rn_U+qgJUzM={y$E>rl_s+j_tT|
zMX8h^kGCYHIi*l6L(sPVu(&ONy3>my;}~(j<l4h$Rjygq;pwNlaRd8rgEB2aQS`P^
zv=sqIho>B$CdfgKtb=LtE-p0wIY#->ackjDls$a-;fJ;)RB`p=<Kwx--w*Dr{k|^=
zB8V2-0%W$gk4krPtqDOM4Qo3;{am-}?4U%}nN4<>b1h20fJi^@Y(lH~O9CiAKR?@D
zFTo;dE-lYyf!7R9k+%K#<Bx5jn?FI^Bl)KZ6gyE_q3!~qUtgA?N@s(<nH#IOZR`H@
zbKUd3%pEgEf(h5CuVa_O1wJE2j4<ca;+fK=OK;e)Ap>SCpGF%`o;<l<zkct&`|hk+
zvyL58=EZFP!QayodKMo^oP6VrH$;m5#|vvzp;pS<jEfO6VVvBHcXYMad#go>PqDFP
zD84_!F;k(|b4`sw^x-eB{dr6)9w0=u0;Bd8xoF}k>og&}#Ln~P&HMJ-Z^@ZYoH)^B
z{;N`@N}-{l`l;}YZitgcevBBO858XpGHQ^%6DIr8%>%v(HPvMZVgaefxY({ux3>@5
z0*DRz5~|LL^XUIY>8D?=Kby%Co=kMO@Vzn9gw<=l?{nwxn|g%NlNn*Cowv8QU%q^~
z{wnh}ej2(|%W`Pf!tanFL--WIJ26^VSeW_c;tZMj)YMe&)4X}}!i5WKH?iKJ<cSGW
z?r`hwHKyY!Wu(l8Dyg*1yR7c*!w#IEO)C5JbNzLI4O|-?Gc?<YD=`99a#=ff?j)bg
zRHk(#v6x6`rUn-kE&TbKHEX{9`s;#OaxA#(;uCl6+U4o#$!E1`(?;a6#d&^TzkVI-
z@WmHjG-=Yr7C|<7VCLqCCb-Xohxa$_|4^-WHEiXSe!Diyd;82us64_q2ZUXtLo%K6
zPy%Wp)T9o6yLB2&aC@lvL~tlI$u#LrPkkA%Mv%U7`0!ydAJL~zABs<Si`qgMVTy$K
zUF(U4f9<u`C`{7SIE_u4Her-)Pnww!EnYIx#HVaO;WfjsOPg^4sX@Rq+7^KHOT}?P
z<)41J9=|3N1x|Qn<l1@JwpO?A-2fAWTF70qX3g-LETSh)lB{MHOd~6!#-vG;=wGf}
zIciiyAF2((KPV{3BnC&KwNeL&_Vx9}Kv;PI8{Zl_bg1~;!{=Avm5DM}V{yB+?b<|K
zHf3!N;Cj-=Ou1I3s!5?>tVuOs(ZB?*nR^08<gR-4>Q?ZimRpm-jEF|Pivjb_JMWlu
zMuFf<Gi7G++L)La>dCEqy6tbh{r20!%1}~8-VD`E`MO%lt}PW;8Be!J+^Evy%FPJR
z0Xlrj=FGWP9d%_*?!qmN|G1;n7ct+&^kSw+Ld1?8J1o*<axyY9678*5*ErG$JLxs1
zGMPu)vnQ=2Ra}dJqtC@NFiVb~pXu{h>CmN17crvF-PpzShmmV#!Gck`MZ=tNWm^ER
zs2t#%O^r~>#%u|@=*D#I@3$3WN>+*)A}i2|w@6f)XkZhH^)K}jjfj2s-FNyVCz(u&
zSoHk~y~)h*L9ooKA3V1jHELw~NEvm|{JqP+ANEbG?iEwQWdAN5<9d~DSTH^_ItRpD
z(;{7!!h^`n8Aq<!*k6<fGn2?1Vn(=$^Dsq<Ot1rYNDZQT=FAzUAPKbbuA`%)nIqGT
zubxNrhDmKOcU!u2slE@YKr9OA-MhCKCy9xP?c29!O+hvkJM-Y8p85KpCvZVwHa*vh
zfTx#nJ>#XkqDnpezS*=$v4T=Cj(}T{ac2w}W&8k6l2g;488J=(g<7Uy`18*{fA78b
zSeMbbabsqnAOemkIK*=%lZK@hv8|p*@?+-3OO&YT`!HZkxaM6HJ0pru%;>2yI<Ilf
zVCEZ1W+=Pqiv1W(c5|}CIP26RO6CBr!<9vLVuc-vP*uh=V%(Y0&MIjjMNUacWrSN+
z&r?*)x(_{%xMU_ZhfI;9wDo<^59C1jtO^w>Xx=*=P!!}mv-%9JTD1}}`<35+U}h5$
z7K=$UV%$WfPu-|4)qlEkq?6aNyF0c?8OCaCrP8g|ew!I_W!3GTQws#(*wd`yX{A4W
z5j*RtB>hEV1ldPF@GilZ@RPMZjEB~(TZ?<5WcKghj}pFk@gg-UOkJ6L7BvYiTC^bN
zW^!J8NfJQJ$XxtwA7w}y*D#&|-Ra!62`h6zK79Q0k~3pV;dUA=l#X!=M%*##(B}f}
zWy_Yajz+`E_689*w%GhyrXaPIdg3F<QByQRM(g<T;~K^K26<Pl$Pk@Ru+I`s^T~AF
zNV4c#wrsJIQp4N`qkralig#=Ug=<z9rrQ=kBVEyi#zR%<F~z@Kk^|<szFk~;c3c@X
z+pla|A{ftDr8@@i=y2z6Wfc$c5y7?X6pNnJq9CDoXt?Mm*CM=x?RqLRZBx+l)mL9x
znYCmc3d@4TZ36=X%?FtF4CGhE#~r=6ku<o<wZ)AfEHKVi`9po&o_>2hY<6kyc*-!8
z8C&v2{FyMWs&s=9SF$KNm5wdDrmoZ@lW9wct$`^T<hn_Sf$EPx{)i}%+a$#;7KV^<
zwI#!5G62xq)22-mNj&Yd;1sWq>5|>McMHvr#^~F(uhzl5!TMiuX9}TJnMJ>)E6=bd
z<1xiP9_RY>bN%6{Qmntj!6w&dWSogS!8n>Kol3`EGZI)O0W8QAHqIP0>t_go#q1$>
zz=^^gf(}vii%AvrAtr3v@{$!CqJ)^4Zsu57=*Jm-z<%h^AuGui)V`Se;Ira>90AKS
zlqmqXwu$cPk;VEB$^msLz27h46I+F;SW#0a%5MxR-Dt!$l`dvLwCW>fnXNNx{w{=?
zOl~qkDMBX_BQRnMvsU_;_-|TImiCxGjCO{>COpHxAKYdE2zty&w4gf^#;0y}f32s}
z(=XQ@JG;*cXPJYCsEejtR#es*ReCxju8O(q`n~Tyc&O}dL7<e4*(L1q%9Sgbh5Yr`
zUzwsJ1u8Ia*RGwmI$PuDku-4WG29xn<_*#>JP)7s`|S%0w>E-kCH7PpHzD0<!I65S
z+s5}JOjzyKvZdqG&vg?|mnkbeD4HTC#+<vtxT?}s#)T2rRJ!7GV#}BT5nEK@Vr9c*
z12xiJr%oLZ){2es+19RIn<=vvlWf+nBAwfzLkDwy^=Cf#fcj@;-6*R+`12L@rQ3%y
zy^QPWRv4cV>po(Un`ez2P)Ay)qT{IfRVjR>gp5*H7*`j$pwfjAFQ*!DWv@szE+hS0
zk`7{X3>+To6PkKFt+a8aaHlgZIAr-JH4MahdP8iTXy(nCXw&D_%^!nL=bqv|?5Jj#
zz}0bWlyOa;r&H;0J$R~1jWYI6zg$->X8-ZXivC;bu~La@XV#^P^=m?<n;UUTYT;cg
zB3-p=6-6E90BUEf<4iZq^tf^3NU&RH?eyu>LfhkFvBJxGXYnrX8EXx!#987YxuIB`
zOHs<k{R7Do<AN&FxMrIwt6JTZalR3w&!;$MIe=>xr}TWQj9;QBOHIl7^R?7ugH^^+
z>B5L3;Auu2=MAn|`t<jGrHr2?b%aqGv*ry_9QZ{jNpha(ZK@GC!{R=sze#-%{4zgf
zi=2`S(c)<86RdMhj1U|gta$?FPo+P@3L~LXD7j&6H599p0&p!zR}8oQFfrm&Ho3Kb
z&9TydvYw0a^C%A<vUyW??@gWoTVLWYWqq;2c%o40su5S$EgDt2SQAm64i|shD*zuk
za-<$QlQurKpoYm!Oef-r5D{Fn+>%KO>y+u=)lX-T?#`%sflIhQmIar<<Vh^PWc+9f
zoPuE-5-1k)^_elQ2sqz<GA7uiVO=(#XYJmj<mJD(o4O{~={x^Ai&7f1Js`V7u#lWZ
z<t$87ReDXMO2=&zahdXc)<}2C@u`9f64B;!QhH#K;!LZB^|XpMBjK;V{t|6mGAy-r
zPckPKv!;kM$+4<ct7<(nZIDwHD^O72;Rio8p8`--x>3f{3pm0)BgS>m6zA%t*>*k$
zaIJb2_5;4CvN)yLqODC9ZfT@2o($tcrCTuKs;^#=ed@%TTr_Z|rKD)=@NmQ00drc-
z&oFO=&Sx^3NoqV~bS!O+IE2w;E}J_L6bu+JKueEnovk;-wg6my;kUimETd5CYjwn|
zpP*aS>ggHp`L1IQAzD#8sPT4fOZvq$Te9utMO&LLG%zkydYTbu7ZKJt7`7dVS9Tsa
zbk0!nO@Wa002aXU!6*&YMAe*M{TZgJ1xL(QFmJ__7FJcy8e)zVYnO;gF?|PH+Kj*>
zM~;Y_Bpewnl(v~@3*PmN8#YeoJ5a{1bZOc+XDD!C6d0CuDmqPA)L_|;7VI9W3OI}_
zDm`F}X2cP2LtKVd=OFT0zU%eB?kdS`M4LLyOyR|2@7c3wN=k|qYb%h~Poi-ZHHXyp
z>+es&1NRY~sc*^BB3$i0nowW=^Ei6Ul2GdfoOU|K;d;<ir|M-;;7_}y)?0t*?*FL=
zCBf`Axnx`O#RkSv>6VQ+!MJiH)YyL}Pyl}N<Vlk)VKqoi8Wb3?2N9`JWD0MMT$#yK
zCO3pflSwr5*o}QQ$h?t47Zt50pOWcI*7e{hiic;(o*HN2ft!$P{xXg3so~Du`{X1A
zE*Mw5@g~LRZEha8qm9Zq0uJM-bj^q(;OUIG8lBY$N$q&!FT+G6Hq`c%DRWzRbj&Ox
zutdScWn7Le5UU253$q@|K_z!!7X|%Bd71L7b-=6>lnJ6)!$oev2BFsTP7$^9sAW!k
zp;eCXz)j@3dP#@jQ|bq&zRH58W!qaL;7hj&l`f1p0*>33X2it~5=i&fW0M~~RQ8o6
zs`@u+G?E@zXKGQ#jvdS77ZrmVHIKYO0hKWKtU%m+;A3NurDrpfHjCii9p+fIcuvyZ
z8sms|moDWNC$<gR(JpX%+hs7mO<^3BZeqk2q#1D~7)SPhxt1hk1Wj(THFWl9B23HF
z*~{LYxvE~hdOW9i;G-8eFw3JF;!Ma*r{1HS#;tbZn=&f9<VdbD*|jS;kMnC1va4gz
zj`oC9M#eQG{<3*oCPK3{d){J(NTA35C)UI5Y;*KTE@TL?R#!agZ?{e{p3{+-xxH3~
z1YeNBKD}yKog=wMz%e~N-l({AO9!klwMGcIk@42)jkrE8Q+8|gPpVJMa__#f2OvLq
z(b@#oBO&Nf-Gryo0&_OG;;VwPEYS>xC?^)dwQftOl~ajUxK>4bykE1h-JKP>Q#%AJ
z0**@0#E7fWStE5DLuF==5R8@bHfAXw6_1(ThiOxu%9wk9$a+kw2eZL7LjB=5$LghW
zSg55LYpsFR?$)j1(xi@|yI$i99nv!{jJT%KMI^4oWoovq5wd^$vSEi|vYO;TfHzYL
z9#03Vh`qeLwCQhB-%R0UM|(#(ja}u`vQ^F&YC$)mU89WSi0O@%@9Gk|^R>{O2zaO9
zj8wW3i5ucFkzb|r*<ZJ-id-otV>b*dWx9`tWj=)xBQ|T&qIkX-h*W<zh*mDa{r<|)
zqePAuYC$)m-KIt91rhC+@9N6epTT%Y8smye*Npg5CHJDmWzE=P8X4!C9oQ63&`m0v
za-;0!Iv&+XU}lp_C-FS|K6xdU*kNj<O}oUHj0fVZk;|MKd*)R4ThrcJqq3Vl1q=f}
zuM?UI*9N+UfSagvB@#EpWd_DcxFO(NrO^$%Y_-7(nn%T*?f0QNlq@<sHfTiubZHZr
z-YMB48r~;wcIw`bTB987l9@j|n!J8h*U1Z8tk}~{rP~7I7UMD^aU~ev+AK7+Gez}c
z6HB0Gy*yiEJx+a2h8zpRq=_K<!Gn9$7cm=~J&A^Bg!+?7jy20aBidU}I-0zF+j0R>
z9hdL!9=1y`-bJYNOpJI6xfjKbBlRr)5{pCt90!=dmaL|z{I^{^v)YrLlvu<e%8CTh
zf8D*3aO7L+<+3f&5Iri;iLXS-(Q%fZUwvEA<nH$UKdu$By{nFHA>bB_xM^IbX4@d0
z^<Uc8Wr94r0WVy*@THetdgP2VsZCKUHHxh8H3n7+)qJlXE{7UhWyeMfe~d4k?DD3f
zy@5rK(lgUWlh-QeJYrgt6}!9RaV<x>4FYaKr5hvhwB)Mlwl!b8_2uO|yD`&E?m|cz
zzghZ0PJ{A;f4{8_GoDb)#2E=3ljUGHZ=FdxI+h7{6BjERFW(vFSS{xV)YeBPn!ItH
zV$%a#t=RLrO1GZzbShoAZRSQ?IO^opqYn?g@Z0yA1&49!7T8UNB2vCAl>~V9Bpn}s
z)jcNl#(&<qvj5C7JO{NFl)`>CJkLtHTh|%wJ!FPs-Kx)^nKq{$y4$uYGjDyn6?=qq
zo5;9<aa~-NF1c!m%hX7m=$lEgOE>r9DQPlJgMwkaW!C<%2?O(+?9M@;#Ns=0*NlNo
zWuFOgcE=KVl6&|6;;WOf9cq=J`dy#O!o%zm9qq(ke@z-bn`YXaYb+GXyldAA0nr_d
z%G`u*O{I%S+*)##Nij0*@kd6H(72_3ak-v!+GIR%o>?-_PB;`ZP;tuqCMholfcXk~
z&21xu7ljAZ$q@8SiC{BEWs|OKq(HTSfkclEc52nyfwGfl)l6HUn;IK;`~UT;x;VbW
z^4;Aq+(vz_SLqt#N~+By5?5zPwdATYA&$JS+xu?nNpJQO<I6YiKS(PcGhtu6#HnNJ
z8}=L~WTws>XPd$ma<HuArL2@GZ!sUue(!u&S{4frEJ!^>f49P^eV1q1A2TOfoI$kt
z?#RCVJr*al58KrxY*$wu-Fn8;8F4GgRS||m`jV6u44%kCL-@{I(&<!*P?LY_?sc>=
zt0p;2s7F^*xkRBUDH^uAqFRZ`Xf6MF`20#%4XYAufk91Xtv4Wg>`JH3JscR}Ijfc|
zBl@y>-az;JRTss#3)}S?u{LhGo^i!XPalcvjJU<*s;Gic>t57aWeOVc<#6d&?C!~G
z&<*?FU%#(EOMuVc*rmnkCe-U+vek*|k11b0W1USa?;42i=|IT&WQkU+=My>@8-niM
z{VMxKwhrCZiMW-lS~gU=d2$s2S7uL?;#Iz|QWUnOm8LUix@u2vDgY@XxqR~wU+jp~
zvo4t0C%(9W?ez$vF$Y|(eucWS_jUH@R%~>LQ-`i;68)sme&V02(yiF;E?vv{tZN>!
zld1+7M{j49ar2^=v>8%mF+#fKCrYidvSe8OUOdb(^DnFS^yVviBMuGb8#>6*u{?nt
zzYSY|s9Pc^E3mT2I1|&~EOKnsMzztZKla1{fAS9i4z6~uwk|a_uyJr|JKQj`rsgVL
zmwU0zh?`cfrdxiZTTv)h8>vf%W5tr;vNrGSlYtRecgz3W+p#HGP|32n;c&HUQqTSC
z8FiO#Z5^E2US(WS>6VN*F1qm2ZAh*vrRkO{SJf(I7Kxi|3n12zn(ytyj#bK@CnkIQ
zm|JS}_M}ZtAB=LWUqgwbpS&oUL;Vz-A<a^|f>ZxNRpv*&5|sL?%6RsTxNViIY0FPC
z*cL#1EsN#7eawwGdV5T$Q;&BXYLt}{o!K~dwPUioLx)l<5h7$(8OMlswra%Fsq~DJ
ztLpL-3ze&^A+>Q^0QEaNlubXhy?s<8E*5SmbCgrUo%p(w=FJVBn3QN+5J12^tGf*R
zvg*9>W~>~+cOtU3(TMBfvUJH+lgd?5<D@P|(3YQMvMqqNU7Ljs`l6ljw|fn6s8d<$
z?aaj4+=3wB#qHa+E-_+OorQ_bgHqdOMWt(HPPSIArfmzLYy@F6;<{}CGT7V4oo$OI
zZE*c?tkcU)r8sAuWgYJbc+GN-ufJJt{6f#g$t_tiMB%X-mt`{&PoG@15SRHIwgoV(
zAyxPN&^hXH#=U)FOU;TcIdyaK&u1!IUQotgNxES=;%w307W%l_S8}zhTiLl+pYr1u
zHdvI{GI(dZke!-J&r)*LWcf)8gAq?#QJ5|+v%I%Yd<F0La?>`I89KXo=N`_rDmW-C
z%fRa26Rq9B)s7e(#=G?>J92hC|L7J(;>J|lzfZ1eahcc%LcfMoG2+U$0EyM+C3?(B
zs4{hJg~5|bweREt%L>JkWsRdp+i<l*t5a~^rd7$0hu4@9)@(`gt1uq6y9))$Sx>I2
z%TLl(uCip<bXx%bEp^$h%{!{bn8lUf8C>d>W-irBNm>3^rrCn_&S;#V)!8P<-M&F}
z*RI{l4f(dt^x#((#kUVleT_O*N+5L8;@a#<Ml+<Q$yH^(Y+JL%+nO%g)_6g3L!a<^
zV;9vJ_*MB1or=|~${wN0M@w1G0$CZ6nAP*Dm=llJtLpM<>(cKGtoF^+hBL!j1w?lY
zN+IZ0N+1;8)hTkOO_jW>x)>o{b$Y5|#DjKF*w%V!N~?f`SLSSLGI44B&&OB)-+Sd-
zH7{1DvJ>M<Qv3w?wKyIi-C0rZ9wi(+tGl&nUb=VRs-F(8H^#f!%+NOTH@06C+bMAC
zYau(jgsMl#4%(fAws%~T)ZRa;jnBFkQ-hk0@bVn^MfIM&E3|yMczq8Srq{Wlr1)u}
z*dp)pq(nd>AQ6xVNCYGT5&?;TL_i`S5s(N-1SA3y0f~S_Kq4R!kO)WwBmxoviGV~v
zA|Mfv2uK7Z0ulj<fJ8tdAQ6xVNCYGT5&?;TL_i`S5s(N-1SA3y0f~S_Kq4R!kO)Ww
oBmxoviGV~vA|Mfv2uK7Z0ulj<fJ8tdAQ6xVNCYGTIRt_K1FqNS*8l(j

literal 0
HcmV?d00001

diff --git a/studio/src-tauri/windows/hooks.nsh b/studio/src-tauri/windows/hooks.nsh
new file mode 100644
index 0000000000..64655e8f6e
--- /dev/null
+++ b/studio/src-tauri/windows/hooks.nsh
@@ -0,0 +1,8 @@
+; Unsloth Studio NSIS installer hooks
+
+!macro NSIS_HOOK_POSTUNINSTALL
+  MessageBox MB_YESNO|MB_ICONQUESTION "Remove all Unsloth data ($PROFILE\.unsloth)?$\n$\nThis deletes installed models, training outputs, and configuration." IDNO skip_cleanup
+    RMDir /r "$PROFILE\.unsloth"
+    DetailPrint "Removed $PROFILE\.unsloth"
+  skip_cleanup:
+!macroend
diff --git a/tests/sh/test_tauri_install_exit_order.sh b/tests/sh/test_tauri_install_exit_order.sh
new file mode 100644
index 0000000000..49d7c247c8
--- /dev/null
+++ b/tests/sh/test_tauri_install_exit_order.sh
@@ -0,0 +1,95 @@
+#!/bin/bash
+set -e
+
+SCRIPT_DIR="$(cd "$(dirname "$0")" && pwd)"
+INSTALL_SH="$SCRIPT_DIR/../../install.sh"
+
+echo "=== test_tauri_install_exit_order ==="
+
+tauri_marker_line=$(grep -n "Tauri mode: done, skip shortcuts and auto-launch" "$INSTALL_SH" | head -n1 | cut -d: -f1)
+tauri_exit_line=$(awk -v start="$tauri_marker_line" 'NR > start && /exit 0/ { print NR; exit }' "$INSTALL_SH")
+tauri_done_line=$(awk -v start="$tauri_marker_line" 'NR > start && /tauri_log "DONE" ""/ { print NR; exit }' "$INSTALL_SH")
+first_setup_check_line=$(grep -n 'if \[ "$_SETUP_EXIT" -ne 0 \]; then' "$INSTALL_SH" | head -n1 | cut -d: -f1)
+setup_exit_line=$(awk -v start="$first_setup_check_line" 'NR > start && /exit "\$_SETUP_EXIT"/ { print NR; exit }' "$INSTALL_SH")
+shortcut_line=$(grep -n 'create_studio_shortcuts "$VENV_ABS_BIN/unsloth" "$OS"' "$INSTALL_SH" | head -n1 | cut -d: -f1)
+shortcut_guard_line=$(
+    awk -v stop="$first_setup_check_line" '
+        NR >= stop { exit }
+        /\[ "\$TAURI_MODE" != true \]/ {
+            print NR
+            exit
+        }
+    ' "$INSTALL_SH"
+)
+shortcut_guard_end_line=$(
+    awk -v start="$shortcut_guard_line" -v shortcut="$shortcut_line" '
+        NR <= start { next }
+        NR <= shortcut { next }
+        /^[[:space:]]*fi[[:space:]]*$/ {
+            print NR
+            exit
+        }
+    ' "$INSTALL_SH"
+)
+early_tauri_exit_line=$(
+    awk -v setup="$first_setup_check_line" '
+        NR >= setup { exit }
+        /\[ "\$TAURI_MODE" = true \]/ {
+            in_tauri = 1
+            depth = 1
+            next
+        }
+        in_tauri && /^[[:space:]]*if[[:space:]].*;[[:space:]]*then[[:space:]]*$/ {
+            depth++
+        }
+        in_tauri && /exit[[:space:]]+0/ {
+            print NR
+            exit
+        }
+        in_tauri && /^[[:space:]]*fi[[:space:]]*$/ {
+            depth--
+            if (depth == 0) {
+                in_tauri = 0
+            }
+        }
+    ' "$INSTALL_SH"
+)
+
+if [ -z "$tauri_marker_line" ] || [ -z "$tauri_exit_line" ] || [ -z "$tauri_done_line" ] || [ -z "$first_setup_check_line" ] || [ -z "$setup_exit_line" ] || [ -z "$shortcut_line" ] || [ -z "$shortcut_guard_line" ] || [ -z "$shortcut_guard_end_line" ]; then
+    echo "  FAIL: required install.sh markers not found"
+    exit 1
+fi
+
+if [ -n "$early_tauri_exit_line" ]; then
+    echo "  FAIL: Tauri success exit before setup failure check at line $early_tauri_exit_line"
+    exit 1
+fi
+
+if [ "$shortcut_guard_line" -ge "$shortcut_line" ] || [ "$shortcut_line" -ge "$shortcut_guard_end_line" ]; then
+    echo "  FAIL: shortcuts are not guarded by non-Tauri check"
+    exit 1
+fi
+
+if [ "$shortcut_line" -ge "$first_setup_check_line" ]; then
+    echo "  FAIL: shortcut line $shortcut_line is after setup failure check line $first_setup_check_line"
+    exit 1
+fi
+
+if [ "$first_setup_check_line" -ge "$setup_exit_line" ]; then
+    echo "  FAIL: setup failure check line $first_setup_check_line is after setup exit line $setup_exit_line"
+    exit 1
+fi
+
+if [ "$setup_exit_line" -ge "$tauri_done_line" ]; then
+    echo "  FAIL: setup failure exit line $setup_exit_line is after Tauri DONE line $tauri_done_line"
+    exit 1
+fi
+
+if [ "$setup_exit_line" -ge "$tauri_exit_line" ]; then
+    echo "  FAIL: setup failure check line $first_setup_check_line is after Tauri exit line $tauri_exit_line"
+    exit 1
+fi
+
+echo "  PASS: non-Tauri shortcuts run before setup failure exit"
+echo "  PASS: setup failure exits before Tauri success"
+echo "  PASS: Tauri skips shortcuts"
diff --git a/unsloth_cli/commands/studio.py b/unsloth_cli/commands/studio.py
index a3c0840be1..b74f42674d 100644
--- a/unsloth_cli/commands/studio.py
+++ b/unsloth_cli/commands/studio.py
@@ -1,11 +1,19 @@
 # SPDX-License-Identifier: AGPL-3.0-only
 # Copyright 2026-present the Unsloth AI Inc. team. All rights reserved. See /studio/LICENSE.AGPL-3.0
 
+import importlib.util
+import hashlib
+import json
 import os
 import platform
+import secrets
+import sqlite3
 import subprocess
 import sys
+import tempfile
 import time
+import types
+from datetime import datetime, timezone
 from pathlib import Path
 from typing import Optional
 import typer
@@ -13,12 +21,52 @@ import typer
 studio_app = typer.Typer(help = "Unsloth Studio commands.")
 
 STUDIO_HOME = Path.home() / ".unsloth" / "studio"
+BOOTSTRAP_PASSWORD_FILE = ".bootstrap_password"
+DESKTOP_SECRET_FILE = ".desktop_secret"
+DEFAULT_ADMIN_USERNAME = "unsloth"
+DESKTOP_SECRET_PREFIX = "desktop-"
+API_KEY_PBKDF2_SALT_KEY = "api_key_pbkdf2_salt"
+DESKTOP_SECRET_HASH_KEY = "desktop_secret_hash"
+DESKTOP_SECRET_CREATED_AT_KEY = "desktop_secret_created_at"
+PBKDF2_ITERATIONS = 100_000
 
 # __file__ is unsloth_cli/commands/studio.py -- two parents up is the package root
 # (either site-packages or the repo root for editable installs).
 _PACKAGE_ROOT = Path(__file__).resolve().parent.parent.parent
 
 
+def _should_hide_windows_subprocesses() -> bool:
+    """Hide child console windows only for non-interactive Windows launches."""
+    if platform.system() != "Windows":
+        return False
+    try:
+        return not sys.stdout.isatty()
+    except (AttributeError, OSError, ValueError):
+        return True
+
+
+def _windows_hidden_subprocess_kwargs() -> dict[str, object]:
+    """Return Windows-only Popen kwargs that suppress transient console windows."""
+    if not _should_hide_windows_subprocesses():
+        return {}
+
+    kwargs: dict[str, object] = {}
+    create_no_window = getattr(subprocess, "CREATE_NO_WINDOW", 0)
+    if create_no_window:
+        kwargs["creationflags"] = create_no_window
+
+    startupinfo_factory = getattr(subprocess, "STARTUPINFO", None)
+    startf_use_showwindow = getattr(subprocess, "STARTF_USESHOWWINDOW", 0)
+    sw_hide = getattr(subprocess, "SW_HIDE", 0)
+    if startupinfo_factory is not None and startf_use_showwindow:
+        startupinfo = startupinfo_factory()
+        startupinfo.dwFlags |= startf_use_showwindow
+        startupinfo.wShowWindow = sw_hide
+        kwargs["startupinfo"] = startupinfo
+
+    return kwargs
+
+
 def _studio_venv_python() -> Optional[Path]:
     """Return the studio venv Python binary, or None if not set up."""
     if platform.system() == "Windows":
@@ -97,12 +145,228 @@ def _create_api_key_inprocess(name: str) -> str:
     ``POST /api/auth/api-keys`` on fresh installs.  Safe because the
     CLI already has filesystem access to ``~/.unsloth/studio``.
     """
-    from auth.storage import create_api_key, DEFAULT_ADMIN_USERNAME
+    storage = _load_backend_auth_storage()
 
-    raw_key, _row = create_api_key(username = DEFAULT_ADMIN_USERNAME, name = name)
+    raw_key, _row = storage.create_api_key(
+        username = storage.DEFAULT_ADMIN_USERNAME,
+        name = name,
+    )
     return raw_key
 
 
+def _load_backend_auth_storage():
+    run_py = _find_run_py()
+    backend_dir = (
+        run_py.parent if run_py is not None else _PACKAGE_ROOT / "studio" / "backend"
+    )
+    if backend_dir.is_dir() and str(backend_dir) not in sys.path:
+        sys.path.insert(0, str(backend_dir))
+
+    auth_dir = backend_dir / "auth"
+    storage_py = auth_dir / "storage.py"
+    loaded = sys.modules.get("auth.storage")
+    loaded_path = Path(getattr(loaded, "__file__", "")).resolve()
+    if loaded is not None and loaded_path == storage_py:
+        return loaded
+
+    package = sys.modules.get("auth")
+    package_paths = [Path(path).resolve() for path in getattr(package, "__path__", [])]
+    if package is None or auth_dir.resolve() not in package_paths:
+        package = types.ModuleType("auth")
+        package.__path__ = [str(auth_dir)]
+        package.__package__ = "auth"
+        package.__file__ = str(auth_dir / "__init__.py")
+        sys.modules["auth"] = package
+
+    spec = importlib.util.spec_from_file_location("auth.storage", storage_py)
+    if spec is None or spec.loader is None:
+        raise ImportError(f"Could not load backend auth storage from {storage_py}")
+    storage = importlib.util.module_from_spec(spec)
+    sys.modules["auth.storage"] = storage
+    spec.loader.exec_module(storage)
+
+    return storage
+
+
+def _write_auth_secret(path: Path, secret: str) -> None:
+    path.parent.mkdir(parents = True, exist_ok = True)
+    fd, tmp_name = tempfile.mkstemp(prefix = f".{path.name}.", dir = path.parent)
+    tmp_path = Path(tmp_name)
+    try:
+        try:
+            os.chmod(tmp_path, 0o600)
+        except OSError:
+            pass
+        with os.fdopen(fd, "w") as f:
+            fd = -1
+            f.write(secret)
+        os.replace(tmp_path, path)
+    except Exception:
+        if fd >= 0:
+            os.close(fd)
+        tmp_path.unlink(missing_ok = True)
+        raise
+    try:
+        os.chmod(path, 0o600)
+    except OSError:
+        pass
+
+
+def _connect_auth_db() -> sqlite3.Connection:
+    auth_dir = STUDIO_HOME / "auth"
+    auth_dir.mkdir(parents = True, exist_ok = True)
+    conn = sqlite3.connect(auth_dir / "auth.db")
+    conn.execute(
+        """
+        CREATE TABLE IF NOT EXISTS auth_user (
+            id INTEGER PRIMARY KEY,
+            username TEXT UNIQUE NOT NULL,
+            password_salt TEXT NOT NULL,
+            password_hash TEXT NOT NULL,
+            jwt_secret TEXT NOT NULL,
+            must_change_password INTEGER NOT NULL DEFAULT 0
+        );
+        """
+    )
+    conn.execute(
+        """
+        CREATE TABLE IF NOT EXISTS refresh_tokens (
+            id INTEGER PRIMARY KEY,
+            token_hash TEXT NOT NULL,
+            username TEXT NOT NULL,
+            expires_at TEXT NOT NULL,
+            is_desktop INTEGER NOT NULL DEFAULT 0
+        );
+        """
+    )
+    conn.execute(
+        """
+        CREATE TABLE IF NOT EXISTS api_keys (
+            id INTEGER PRIMARY KEY AUTOINCREMENT,
+            username TEXT NOT NULL,
+            key_prefix TEXT NOT NULL,
+            key_hash TEXT NOT NULL UNIQUE,
+            name TEXT NOT NULL DEFAULT '',
+            created_at TEXT NOT NULL,
+            last_used_at TEXT,
+            expires_at TEXT,
+            is_active INTEGER NOT NULL DEFAULT 1
+        );
+        """
+    )
+    conn.execute(
+        """
+        CREATE TABLE IF NOT EXISTS app_secrets (
+            key TEXT PRIMARY KEY,
+            value TEXT NOT NULL
+        );
+        """
+    )
+    auth_columns = {row[1] for row in conn.execute("PRAGMA table_info(auth_user)")}
+    if "must_change_password" not in auth_columns:
+        conn.execute(
+            "ALTER TABLE auth_user ADD COLUMN must_change_password INTEGER NOT NULL DEFAULT 0"
+        )
+    refresh_columns = {
+        row[1] for row in conn.execute("PRAGMA table_info(refresh_tokens)")
+    }
+    if "is_desktop" not in refresh_columns:
+        conn.execute(
+            "ALTER TABLE refresh_tokens ADD COLUMN is_desktop INTEGER NOT NULL DEFAULT 0"
+        )
+    conn.commit()
+    return conn
+
+
+def _pbkdf2_hex(value: str, salt: bytes) -> str:
+    return hashlib.pbkdf2_hmac(
+        "sha256",
+        value.encode("utf-8"),
+        salt,
+        PBKDF2_ITERATIONS,
+    ).hex()
+
+
+def _hash_password(password: str) -> tuple[str, str]:
+    salt = secrets.token_hex(16)
+    pwd_hash = _pbkdf2_hex(password, salt.encode("utf-8"))
+    return salt, pwd_hash
+
+
+def _get_or_create_api_key_pbkdf2_salt(conn: sqlite3.Connection) -> bytes:
+    row = conn.execute(
+        "SELECT value FROM app_secrets WHERE key = ?",
+        (API_KEY_PBKDF2_SALT_KEY,),
+    ).fetchone()
+    if row is None:
+        salt_hex = secrets.token_hex(32)
+        conn.execute(
+            "INSERT OR IGNORE INTO app_secrets (key, value) VALUES (?, ?)",
+            (API_KEY_PBKDF2_SALT_KEY, salt_hex),
+        )
+        row = conn.execute(
+            "SELECT value FROM app_secrets WHERE key = ?",
+            (API_KEY_PBKDF2_SALT_KEY,),
+        ).fetchone()
+    return bytes.fromhex(row[0])
+
+
+def _ensure_cli_default_admin(conn: sqlite3.Connection) -> None:
+    row = conn.execute(
+        "SELECT 1 FROM auth_user WHERE username = ?",
+        (DEFAULT_ADMIN_USERNAME,),
+    ).fetchone()
+    if row is not None:
+        return
+
+    bootstrap_password = secrets.token_urlsafe(32)
+    password_salt, password_hash = _hash_password(bootstrap_password)
+    conn.execute(
+        """
+        INSERT INTO auth_user (
+            username,
+            password_salt,
+            password_hash,
+            jwt_secret,
+            must_change_password
+        )
+        VALUES (?, ?, ?, ?, ?)
+        """,
+        (
+            DEFAULT_ADMIN_USERNAME,
+            password_salt,
+            password_hash,
+            secrets.token_urlsafe(64),
+            1,
+        ),
+    )
+    _write_auth_secret(
+        STUDIO_HOME / "auth" / BOOTSTRAP_PASSWORD_FILE,
+        bootstrap_password,
+    )
+
+
+def _create_desktop_secret_in_cli() -> str:
+    raw_secret = DESKTOP_SECRET_PREFIX + secrets.token_urlsafe(48)
+    now = datetime.now(timezone.utc).isoformat()
+    conn = _connect_auth_db()
+    try:
+        _ensure_cli_default_admin(conn)
+        secret_hash = _pbkdf2_hex(raw_secret, _get_or_create_api_key_pbkdf2_salt(conn))
+        conn.execute(
+            "INSERT OR REPLACE INTO app_secrets (key, value) VALUES (?, ?)",
+            (DESKTOP_SECRET_HASH_KEY, secret_hash),
+        )
+        conn.execute(
+            "INSERT OR REPLACE INTO app_secrets (key, value) VALUES (?, ?)",
+            (DESKTOP_SECRET_CREATED_AT_KEY, now),
+        )
+        conn.commit()
+        return raw_secret
+    finally:
+        conn.close()
+
+
 def _load_model_via_http(
     port: int,
     api_key: str,
@@ -153,6 +417,11 @@ def studio_default(
     host: str = typer.Option("0.0.0.0", "--host", "-H"),
     frontend: Optional[Path] = typer.Option(None, "--frontend", "-f"),
     silent: bool = typer.Option(False, "--silent", "-q"),
+    api_only: bool = typer.Option(
+        False,
+        "--api-only",
+        help = "Run API server only, no frontend serving (for Tauri desktop app)",
+    ),
 ):
     """Launch the Unsloth Studio server."""
     if ctx.invoked_subcommand is not None:
@@ -180,13 +449,15 @@ def studio_default(
                 args.extend(["--frontend", str(frontend)])
             if silent:
                 args.append("--silent")
+            if api_only:
+                args.append("--api-only")
             # On Windows, os.execvp() spawns a child but the parent lingers,
             # so Ctrl+C only kills the parent leaving the child orphaned.
             # Use subprocess.run() on Windows so the parent waits for the child.
             if sys.platform == "win32":
                 import subprocess as _sp
 
-                proc = _sp.Popen(args)
+                proc = _sp.Popen(args, **_windows_hidden_subprocess_kwargs())
                 try:
                     rc = proc.wait()
                 except KeyboardInterrupt:
@@ -217,7 +488,7 @@ def studio_default(
         display_host = _resolve_external_ip() if host == "0.0.0.0" else host
         typer.echo(f"Starting Unsloth Studio on http://{display_host}:{port}")
 
-    run_kwargs = dict(host = host, port = port, silent = silent)
+    run_kwargs = dict(host = host, port = port, silent = silent, api_only = api_only)
     if frontend is not None:
         run_kwargs["frontend_path"] = frontend
     run_server(**run_kwargs)
@@ -489,9 +760,16 @@ def _run_setup_script(*, verbose: bool = False) -> None:
     env = {**os.environ, "UNSLOTH_VERBOSE": "1"} if verbose else None
 
     if platform.system() == "Windows":
+        powershell_args = ["powershell.exe"]
+        if _should_hide_windows_subprocesses():
+            powershell_args.extend(
+                ["-NoLogo", "-NoProfile", "-NonInteractive", "-WindowStyle", "Hidden"]
+            )
+        powershell_args.extend(["-ExecutionPolicy", "Bypass", "-File", str(script)])
         result = subprocess.run(
-            ["powershell", "-ExecutionPolicy", "Bypass", "-File", str(script)],
+            powershell_args,
             env = env,
+            **_windows_hidden_subprocess_kwargs(),
         )
     else:
         result = subprocess.run(["bash", str(script)], env = env)
@@ -547,6 +825,44 @@ def update(
 # ── unsloth studio reset-password ────────────────────────────────────
 
 
+@studio_app.command("desktop-capabilities", hidden = True)
+def desktop_capabilities(
+    json_output: bool = typer.Option(
+        False,
+        "--json",
+        help = "Emit machine-readable JSON.",
+    ),
+):
+    payload = {
+        "desktop_protocol_version": 1,
+        "supports_provision_desktop_auth": True,
+        "supports_api_only": True,
+        "version": "unknown",
+    }
+    try:
+        from importlib.metadata import version as package_version
+
+        payload["version"] = package_version("unsloth")
+    except Exception:
+        pass
+
+    if json_output:
+        typer.echo(json.dumps(payload, sort_keys = True))
+        return
+
+    for key, value in payload.items():
+        typer.echo(f"{key}: {value}")
+
+
+@studio_app.command("provision-desktop-auth", hidden = True)
+def provision_desktop_auth():
+    """Create/repair desktop auth state for the local machine."""
+    auth_dir = STUDIO_HOME / "auth"
+    secret = _create_desktop_secret_in_cli()
+    _write_auth_secret(auth_dir / DESKTOP_SECRET_FILE, secret)
+    typer.echo("Desktop auth ready.")
+
+
 @studio_app.command("reset-password")
 def reset_password():
     """Reset the Studio admin password.
@@ -557,13 +873,18 @@ def reset_password():
     """
     auth_dir = STUDIO_HOME / "auth"
     db_file = auth_dir / "auth.db"
-    pw_file = auth_dir / ".bootstrap_password"
+    stale_files = [
+        auth_dir / BOOTSTRAP_PASSWORD_FILE,
+        auth_dir / DESKTOP_SECRET_FILE,
+    ]
+    had_db = db_file.exists()
 
-    if not db_file.exists():
+    db_file.unlink(missing_ok = True)
+    for path in stale_files:
+        path.unlink(missing_ok = True)
+
+    if not had_db:
         typer.echo("No auth database found -- nothing to reset.")
         raise typer.Exit(0)
 
-    db_file.unlink(missing_ok = True)
-    pw_file.unlink(missing_ok = True)
-
     typer.echo("Auth database deleted. Restart Unsloth Studio to get a new password.")

From ad6bd780a91f4db03685c0666373facc8f44fbfe Mon Sep 17 00:00:00 2001
From: Daniel Han <danielhanchen@gmail.com>
Date: Thu, 23 Apr 2026 06:42:12 -0700
Subject: [PATCH 159/216] Update _utils.py

---
 unsloth/models/_utils.py | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/unsloth/models/_utils.py b/unsloth/models/_utils.py
index 2e7e17bbb9..8b8dec3d3e 100644
--- a/unsloth/models/_utils.py
+++ b/unsloth/models/_utils.py
@@ -12,7 +12,7 @@
 # See the License for the specific language governing permissions and
 # limitations under the License.
 
-__version__ = "2026.4.7"
+__version__ = "2026.4.8"
 
 __all__ = [
     "SUPPORTS_BFLOAT16",

From 563fcf89526cde9bb1254c385839e15f01cdf613 Mon Sep 17 00:00:00 2001
From: Daniel Han <danielhanchen@gmail.com>
Date: Thu, 23 Apr 2026 06:57:36 -0700
Subject: [PATCH 160/216] Studio: detect reasoning_effort and preserve_thinking
 in chat templates (#5149)

* Studio: detect reasoning_effort and preserve_thinking in chat templates

Previously Studio's chat template sniffer only recognized Qwen's
enable_thinking and DeepSeek's thinking markers. For gpt-oss (Harmony
templates) and newer Qwen3.6 templates, the Think toggle was hidden or
could only be flipped on/off.

This change adds two new detections and corresponding UI controls:

1. reasoning_effort style (gpt-oss). When the chat template contains
   reasoning_effort, the Think button becomes a Low / Medium / High
   dropdown and the backend forwards {"reasoning_effort": <level>} in
   chat_template_kwargs. Load-time --chat-template-kwargs flag is also
   switched to the new style.

2. preserve_thinking kwarg (Qwen3.6). Independent of the reasoning
   toggle. When the template mentions preserve_thinking, a new
   Preserve Thinking on/off pill is shown next to Think. Off by
   default, persisted via localStorage. When on, the backend adds
   {"preserve_thinking": true} to chat_template_kwargs so past-turn
   <think> blocks are kept in the prompt instead of being stripped.

Backend helper _request_reasoning_kwargs now merges all applicable
kwargs into a single chat_template_kwargs dict based on the model's
detected style and template capabilities. Inputs are validated with
Literal types in the Pydantic request model.

Tested end to end against cached GGUFs for unsloth/gpt-oss-20b-GGUF and
unsloth/Qwen3.6-35B-A3B-GGUF. Confirmed the llama-server startup
--chat-template-kwargs flag and per-request JSON body carry the
expected keys for all combinations.

* Studio: review pass and CI format fixes for reasoning-styles PR

Addresses review feedback and pre-commit CI:

- Preserve Thinking pill in shared-composer now gates on modelLoaded
  only, matching the thread.tsx toggle. Previously the inline version
  disabled whenever supports_reasoning was false.
- The non-GGUF already_loaded LoadResponse now emits reasoning_style
  (and supports_preserve_thinking=False) so a reconnecting frontend
  sees the correct style for an already-running gpt-oss safetensors
  model.
- use-chat-model-runtime reconnect path now always clears
  reasoningEnabled for models without reasoning support instead of
  inheriting the previous model's state.
- _reasoning_default is now reset alongside the other reasoning flags
  in both backend reset blocks.
- supports_reasoning description updated to mention reasoning_effort
  alongside enable_thinking.
- Ran scripts/run_ruff_format.py on the touched Python files to
  satisfy pre-commit.ci.

* Studio: detect reasoning flags on safetensors load + share Qwen param helper

Addresses bot review feedback:

- Extract the chat-template substring sniffer out of _read_gguf_metadata
  into a module-level detect_reasoning_flags(template, model_id) helper.
  Also runs on the safetensors / transformers load paths:
    - POST /api/inference/load non-GGUF LoadResponse
    - already_loaded non-GGUF early return
    - GET /api/inference/status non-GGUF branch
  The gpt-oss fallback via backend._is_gpt_oss_model() is preserved so
  safetensors gpt-oss still surfaces reasoning controls even when no
  chat_template is stored on the model record.

- Deduplicate the Qwen3 / Qwen3.5 / Qwen3.6 Think-toggle parameter
  adjustment into a single features/chat/utils/qwen-params.ts. Both
  the assistant-ui Think toggle (thread.tsx) and the shared composer
  (shared-composer.tsx) now import the same helper. The superset that
  applies presence_penalty=1.5 for Qwen3.5 and Qwen3.6 is now used by
  both sites (thread.tsx previously did not apply it).

* Studio: fill missing reasoning flags on safetensors status + add always_on reset

Round 4 review fixes:

- routes/inference.py safetensors status response now populates
  reasoning_always_on and supports_tools from detect_reasoning_flags.
  Previously Pydantic defaulted both to False, so safetensors models
  with always-on <think> templates or tool-calling templates were
  silently losing those flags on /api/inference/status reconnect.
- routes/inference.py already_loaded safetensors branch now falls back
  to backend._is_gpt_oss_model() when the chat template is missing,
  matching the status-endpoint behaviour.
- Safetensors status endpoint log_source set to "Safetensors status"
  so the emitted template-detection log lines are attributable.
- chat-runtime-store clearCheckpoint now also resets reasoningAlwaysOn
  so switching from an always-on reasoning model to a non-always-on
  one does not leave the Think button permanently locked on.

* Studio: skip reasoning kwargs when always-on; narrow non-GGUF advertisement

Round 5 addresses reviewer feedback:

- _request_reasoning_kwargs and the load-time --chat-template-kwargs
  emission now skip when _reasoning_always_on is true. Templates with
  hardcoded <think> tags do not consume enable_thinking / reasoning_effort
  so sending them was noise.
- Non-GGUF (Unsloth / transformers) LoadResponse and InferenceStatusResponse
  paths no longer advertise template-derived supports_reasoning /
  reasoning_style / supports_preserve_thinking / supports_tools. The
  transformers generation path does not yet forward chat_template_kwargs
  to tokenizer.apply_chat_template, so exposing the UI controls on those
  models was misleading. Only the gpt-oss Harmony case is kept
  (reasoning_style = reasoning_effort) because it is handled via the
  HarmonyTextStreamer at the tokenizer level. A follow-up PR can thread
  chat_template_kwargs through the transformers path and re-enable the
  broader detection.
- GGUF / llama-server paths keep the full detect_reasoning_flags output.
---
 studio/backend/core/inference/llama_cpp.py    | 218 +++++++++++++-----
 studio/backend/models/inference.py            |  26 ++-
 studio/backend/routes/inference.py            |  69 +++++-
 .../src/components/assistant-ui/thread.tsx    |  99 +++++++-
 .../src/features/chat/api/chat-adapter.ts     |  28 ++-
 .../chat/hooks/use-chat-model-runtime.ts      |  13 ++
 .../src/features/chat/shared-composer.tsx     | 140 ++++++++---
 .../chat/stores/chat-runtime-store.ts         |  47 ++++
 .../frontend/src/features/chat/types/api.ts   |   6 +
 .../src/features/chat/utils/qwen-params.ts    |  29 +++
 10 files changed, 566 insertions(+), 109 deletions(-)
 create mode 100644 studio/frontend/src/features/chat/utils/qwen-params.ts

diff --git a/studio/backend/core/inference/llama_cpp.py b/studio/backend/core/inference/llama_cpp.py
index 6ad05ac52d..c320f03b2c 100644
--- a/studio/backend/core/inference/llama_cpp.py
+++ b/studio/backend/core/inference/llama_cpp.py
@@ -87,6 +87,84 @@ _TC_PARAM_START_RE = re.compile(r"<parameter=(\w+)>\s*")
 _TC_PARAM_CLOSE_RE = re.compile(r"\s*</parameter>\s*$")
 
 
+_TOOL_TEMPLATE_MARKERS = (
+    "{%- if tools %}",
+    "{%- if tools -%}",
+    "{% if tools %}",
+    "{% if tools -%}",
+    '"role" == "tool"',
+    "'role' == 'tool'",
+    'message.role == "tool"',
+    "message.role == 'tool'",
+)
+
+
+def detect_reasoning_flags(
+    chat_template: Optional[str],
+    model_identifier: Optional[str] = None,
+    *,
+    log_source: Optional[str] = None,
+) -> dict:
+    """Classify a chat template's reasoning and tool-calling capabilities.
+
+    Returns a dict with the same five keys populated by the GGUF sniffer:
+    ``supports_reasoning``, ``reasoning_style``
+    (``"enable_thinking"`` | ``"reasoning_effort"``),
+    ``reasoning_always_on``, ``supports_preserve_thinking``, and
+    ``supports_tools``. Used by both the llama-server backend at load
+    time and the safetensors/transformers paths in ``routes/inference``
+    so the two agree on what the frontend will see.
+    """
+    flags = {
+        "supports_reasoning": False,
+        "reasoning_style": "enable_thinking",
+        "reasoning_always_on": False,
+        "supports_preserve_thinking": False,
+        "supports_tools": False,
+    }
+    if not chat_template:
+        return flags
+    tpl = chat_template
+    prefix = f"{log_source}: " if log_source else ""
+
+    if "enable_thinking" in tpl:
+        flags["supports_reasoning"] = True
+        flags["reasoning_style"] = "enable_thinking"
+        logger.info(f"{prefix}model supports reasoning (enable_thinking)")
+    elif "reasoning_effort" in tpl:
+        # gpt-oss / Harmony templates use reasoning_effort
+        # ("low" | "medium" | "high") instead of a boolean.
+        flags["supports_reasoning"] = True
+        flags["reasoning_style"] = "reasoning_effort"
+        logger.info(f"{prefix}model supports reasoning (reasoning_effort)")
+    elif "thinking" in tpl:
+        # DeepSeek uses 'thinking' instead of 'enable_thinking'
+        normalized_id = (model_identifier or "").lower()
+        if "deepseek" in normalized_id:
+            flags["supports_reasoning"] = True
+            logger.info(f"{prefix}model supports reasoning (DeepSeek thinking)")
+
+    # Hardcoded <think> tags or reasoning_content in the template mean
+    # thinking is always on (no toggle to disable it).
+    if not flags["supports_reasoning"]:
+        if ("<think>" in tpl and "</think>" in tpl) or "reasoning_content" in tpl:
+            flags["supports_reasoning"] = True
+            flags["reasoning_always_on"] = True
+            logger.info(f"{prefix}model always reasons (<think> tags in template)")
+
+    # preserve_thinking is an independent kwarg on some Qwen templates
+    # that keeps historical <think> blocks in prior assistant turns.
+    if "preserve_thinking" in tpl:
+        flags["supports_preserve_thinking"] = True
+        logger.info(f"{prefix}model supports preserve_thinking")
+
+    if any(marker in tpl for marker in _TOOL_TEMPLATE_MARKERS):
+        flags["supports_tools"] = True
+        logger.info(f"{prefix}model supports tool calling")
+
+    return flags
+
+
 class LlamaCppBackend:
     """
     Manages a llama-server subprocess for GGUF model inference.
@@ -112,6 +190,8 @@ class LlamaCppBackend:
         self._chat_template: Optional[str] = None
         self._supports_reasoning: bool = False
         self._reasoning_always_on: bool = False
+        self._reasoning_style: str = "enable_thinking"
+        self._supports_preserve_thinking: bool = False
         self._supports_tools: bool = False
         self._cache_type_kv: Optional[str] = None
         self._reasoning_default: bool = True
@@ -293,10 +373,51 @@ class LlamaCppBackend:
     def reasoning_always_on(self) -> bool:
         return self._reasoning_always_on
 
+    @property
+    def reasoning_style(self) -> str:
+        return self._reasoning_style
+
+    @property
+    def supports_preserve_thinking(self) -> bool:
+        return self._supports_preserve_thinking
+
     @property
     def reasoning_default(self) -> bool:
         return self._reasoning_default
 
+    def _reasoning_kwargs(self, enable_thinking: bool) -> dict:
+        if self._reasoning_style == "reasoning_effort":
+            return {"reasoning_effort": "high" if enable_thinking else "low"}
+        return {"enable_thinking": enable_thinking}
+
+    def _request_reasoning_kwargs(
+        self,
+        enable_thinking: Optional[bool],
+        reasoning_effort: Optional[str] = None,
+        preserve_thinking: Optional[bool] = None,
+    ) -> Optional[dict]:
+        """Build chat_template_kwargs from per-request reasoning fields.
+
+        Produces a merged dict covering the active model's reasoning style
+        (``enable_thinking`` or ``reasoning_effort``) plus the independent
+        ``preserve_thinking`` kwarg when the template supports it.
+        """
+        kwargs: dict = {}
+        # Always-on reasoning models hardcode <think> tags in their template
+        # and do not consume enable_thinking / reasoning_effort -- skip.
+        if self._supports_reasoning and not self._reasoning_always_on:
+            if self._reasoning_style == "reasoning_effort":
+                if reasoning_effort in ("low", "medium", "high"):
+                    kwargs["reasoning_effort"] = reasoning_effort
+                elif enable_thinking is not None:
+                    kwargs["reasoning_effort"] = "high" if enable_thinking else "low"
+            else:
+                if enable_thinking is not None:
+                    kwargs["enable_thinking"] = enable_thinking
+        if self._supports_preserve_thinking and preserve_thinking is not None:
+            kwargs["preserve_thinking"] = preserve_thinking
+        return kwargs or None
+
     @property
     def supports_tools(self) -> bool:
         return self._supports_tools
@@ -818,6 +939,9 @@ class LlamaCppBackend:
         self._chat_template = None
         self._supports_reasoning = False
         self._reasoning_always_on = False
+        self._reasoning_style = "enable_thinking"
+        self._reasoning_default = True
+        self._supports_preserve_thinking = False
         self._supports_tools = False
         self._n_layers = None
         self._n_kv_heads = None
@@ -898,48 +1022,16 @@ class LlamaCppBackend:
                     f"GGUF metadata: chat_template={len(self._chat_template)} chars"
                 )
                 # Detect thinking/reasoning support from chat template
-                tpl = self._chat_template
-                if "enable_thinking" in tpl:
-                    self._supports_reasoning = True
-                    logger.info(
-                        "GGUF metadata: model supports reasoning (enable_thinking)"
-                    )
-                elif "thinking" in tpl:
-                    # DeepSeek uses 'thinking' instead of 'enable_thinking'
-                    normalized_id = (self._model_identifier or "").lower()
-                    if "deepseek" in normalized_id:
-                        self._supports_reasoning = True
-                        logger.info(
-                            "GGUF metadata: model supports reasoning (DeepSeek thinking)"
-                        )
-                # Models with hardcoded <think> tags or reasoning_content
-                # in their chat template always produce thinking output
-                # (no toggle to disable it).
-                if not self._supports_reasoning:
-                    if (
-                        "<think>" in tpl
-                        and "</think>" in tpl
-                        or "reasoning_content" in tpl
-                    ):
-                        self._supports_reasoning = True
-                        self._reasoning_always_on = True
-                        logger.info(
-                            "GGUF metadata: model always reasons (<think> tags in template)"
-                        )
-                # Detect tool calling support from chat template
-                tool_markers = [
-                    "{%- if tools %}",
-                    "{%- if tools -%}",
-                    "{% if tools %}",
-                    "{% if tools -%}",
-                    '"role" == "tool"',
-                    "'role' == 'tool'",
-                    'message.role == "tool"',
-                    "message.role == 'tool'",
-                ]
-                if any(marker in tpl for marker in tool_markers):
-                    self._supports_tools = True
-                    logger.info("GGUF metadata: model supports tool calling")
+                flags = detect_reasoning_flags(
+                    self._chat_template,
+                    self._model_identifier,
+                    log_source = "GGUF metadata",
+                )
+                self._supports_reasoning = flags["supports_reasoning"]
+                self._reasoning_style = flags["reasoning_style"]
+                self._reasoning_always_on = flags["reasoning_always_on"]
+                self._supports_preserve_thinking = flags["supports_preserve_thinking"]
+                self._supports_tools = flags["supports_tools"]
         except Exception as e:
             logger.warning(f"Failed to read GGUF metadata: {e}")
 
@@ -1523,7 +1615,8 @@ class LlamaCppBackend:
             # For reasoning models, set default thinking mode.
             # Qwen3.5/3.6 models below 9B (0.8B, 2B, 4B) disable thinking by default.
             # Only 9B and larger enable thinking.
-            if self._supports_reasoning:
+            # Always-on templates ignore the kwarg entirely, so skip.
+            if self._supports_reasoning and not self._reasoning_always_on:
                 thinking_default = True
                 mid = (model_identifier or "").lower()
                 if "qwen3.5" in mid or "qwen3.6" in mid:
@@ -1531,15 +1624,14 @@ class LlamaCppBackend:
                     if size_val is not None and size_val < 9:
                         thinking_default = False
                 self._reasoning_default = thinking_default
+                reasoning_kw = self._reasoning_kwargs(thinking_default)
                 cmd.extend(
                     [
                         "--chat-template-kwargs",
-                        json.dumps({"enable_thinking": thinking_default}),
+                        json.dumps(reasoning_kw),
                     ]
                 )
-                logger.info(
-                    f"Reasoning model: enable_thinking={thinking_default} by default"
-                )
+                logger.info(f"Reasoning model: {reasoning_kw} by default")
 
             if mmproj_path:
                 if not Path(mmproj_path).is_file():
@@ -1767,6 +1859,9 @@ class LlamaCppBackend:
             self._chat_template = None
             self._supports_reasoning = False
             self._reasoning_always_on = False
+            self._reasoning_style = "enable_thinking"
+            self._reasoning_default = True
+            self._supports_preserve_thinking = False
             self._supports_tools = False
             self._cache_type_kv = None
             self._speculative_type = None
@@ -2317,6 +2412,8 @@ class LlamaCppBackend:
         stop: Optional[list[str]] = None,
         cancel_event: Optional[threading.Event] = None,
         enable_thinking: Optional[bool] = None,
+        reasoning_effort: Optional[str] = None,
+        preserve_thinking: Optional[bool] = None,
     ) -> Generator[str | dict, None, None]:
         """
         Send a chat completion request to llama-server and stream tokens back.
@@ -2341,9 +2438,12 @@ class LlamaCppBackend:
             "repeat_penalty": repetition_penalty,
             "presence_penalty": presence_penalty,
         }
-        # Pass enable_thinking per-request for reasoning models
-        if self._supports_reasoning and enable_thinking is not None:
-            payload["chat_template_kwargs"] = {"enable_thinking": enable_thinking}
+        # Pass enable_thinking / reasoning_effort / preserve_thinking per-request
+        _reasoning_kw = self._request_reasoning_kwargs(
+            enable_thinking, reasoning_effort, preserve_thinking
+        )
+        if _reasoning_kw is not None:
+            payload["chat_template_kwargs"] = _reasoning_kw
         if max_tokens is not None:
             payload["max_tokens"] = max_tokens
         if stop:
@@ -2479,6 +2579,8 @@ class LlamaCppBackend:
         stop: Optional[list[str]] = None,
         cancel_event: Optional[threading.Event] = None,
         enable_thinking: Optional[bool] = None,
+        reasoning_effort: Optional[str] = None,
+        preserve_thinking: Optional[bool] = None,
         max_tool_iterations: int = 25,
         auto_heal_tool_calls: bool = True,
         tool_call_timeout: int = 300,
@@ -2557,8 +2659,11 @@ class LlamaCppBackend:
                 "tools": tools,
                 "tool_choice": "auto",
             }
-            if self._supports_reasoning and enable_thinking is not None:
-                payload["chat_template_kwargs"] = {"enable_thinking": enable_thinking}
+            _reasoning_kw = self._request_reasoning_kwargs(
+                enable_thinking, reasoning_effort, preserve_thinking
+            )
+            if _reasoning_kw is not None:
+                payload["chat_template_kwargs"] = _reasoning_kw
             if max_tokens is not None:
                 payload["max_tokens"] = max_tokens
             if stop:
@@ -3207,10 +3312,11 @@ class LlamaCppBackend:
             "repeat_penalty": repetition_penalty,
             "presence_penalty": presence_penalty,
         }
-        if self._supports_reasoning and enable_thinking is not None:
-            stream_payload["chat_template_kwargs"] = {
-                "enable_thinking": enable_thinking
-            }
+        _reasoning_kw = self._request_reasoning_kwargs(
+            enable_thinking, reasoning_effort, preserve_thinking
+        )
+        if _reasoning_kw is not None:
+            stream_payload["chat_template_kwargs"] = _reasoning_kw
         if max_tokens is not None:
             stream_payload["max_tokens"] = max_tokens
         if stop:
diff --git a/studio/backend/models/inference.py b/studio/backend/models/inference.py
index 324002ddbf..e5b037755d 100644
--- a/studio/backend/models/inference.py
+++ b/studio/backend/models/inference.py
@@ -157,12 +157,20 @@ class LoadResponse(BaseModel):
     )
     supports_reasoning: bool = Field(
         False,
-        description = "Whether model supports thinking/reasoning mode (enable_thinking)",
+        description = "Whether model supports thinking/reasoning mode (enable_thinking or reasoning_effort)",
+    )
+    reasoning_style: Literal["enable_thinking", "reasoning_effort"] = Field(
+        "enable_thinking",
+        description = "Reasoning control style: 'enable_thinking' (boolean) or 'reasoning_effort' (low|medium|high)",
     )
     reasoning_always_on: bool = Field(
         False,
         description = "Whether reasoning is always on (hardcoded <think> tags, not toggleable)",
     )
+    supports_preserve_thinking: bool = Field(
+        False,
+        description = "Whether the template understands the optional preserve_thinking kwarg (Qwen3.6-style)",
+    )
     supports_tools: bool = Field(
         False,
         description = "Whether model supports tool calling (web search, etc.)",
@@ -261,9 +269,17 @@ class InferenceStatusResponse(BaseModel):
     supports_reasoning: bool = Field(
         False, description = "Whether the active model supports reasoning/thinking mode"
     )
+    reasoning_style: Literal["enable_thinking", "reasoning_effort"] = Field(
+        "enable_thinking",
+        description = "Reasoning control style: 'enable_thinking' (boolean) or 'reasoning_effort' (low|medium|high)",
+    )
     reasoning_always_on: bool = Field(
         False, description = "Whether reasoning is always on (not toggleable)"
     )
+    supports_preserve_thinking: bool = Field(
+        False,
+        description = "Whether the active model's template understands the optional preserve_thinking kwarg",
+    )
     supports_tools: bool = Field(
         False, description = "Whether the active model supports tool calling"
     )
@@ -481,6 +497,14 @@ class ChatCompletionRequest(BaseModel):
         None,
         description = "[x-unsloth] Enable/disable thinking/reasoning mode for supported models",
     )
+    reasoning_effort: Optional[Literal["low", "medium", "high"]] = Field(
+        None,
+        description = "[x-unsloth] Reasoning effort level ('low'|'medium'|'high') for Harmony-style reasoning models (e.g. gpt-oss). Overrides enable_thinking when the active model uses reasoning_effort style.",
+    )
+    preserve_thinking: Optional[bool] = Field(
+        None,
+        description = "[x-unsloth] When true, keep historical <think> blocks from past assistant turns in the prompt (Qwen3.6 templates). Independent of enable_thinking / reasoning_effort.",
+    )
     enable_tools: Optional[bool] = Field(
         None,
         description = "[x-unsloth] Enable tool calling for supported models",
diff --git a/studio/backend/routes/inference.py b/studio/backend/routes/inference.py
index e17f8f5882..ed331a5660 100644
--- a/studio/backend/routes/inference.py
+++ b/studio/backend/routes/inference.py
@@ -113,7 +113,7 @@ if str(backend_path) not in sys.path:
 # Import backend functions
 try:
     from core.inference import get_inference_backend
-    from core.inference.llama_cpp import LlamaCppBackend
+    from core.inference.llama_cpp import LlamaCppBackend, detect_reasoning_flags
     from utils.models import ModelConfig
     from utils.inference import load_inference_config
     from utils.models.model_config import load_model_defaults
@@ -122,7 +122,7 @@ except ImportError:
     if str(parent_backend) not in sys.path:
         sys.path.insert(0, str(parent_backend))
     from core.inference import get_inference_backend
-    from core.inference.llama_cpp import LlamaCppBackend
+    from core.inference.llama_cpp import LlamaCppBackend, detect_reasoning_flags
     from utils.models import ModelConfig
     from utils.inference import load_inference_config
     from utils.models.model_config import load_model_defaults
@@ -273,7 +273,9 @@ async def load_model(
                     max_context_length = llama_backend.max_context_length,
                     native_context_length = llama_backend.native_context_length,
                     supports_reasoning = llama_backend.supports_reasoning,
+                    reasoning_style = llama_backend.reasoning_style,
                     reasoning_always_on = llama_backend.reasoning_always_on,
+                    supports_preserve_thinking = llama_backend.supports_preserve_thinking,
                     chat_template = llama_backend.chat_template,
                     speculative_type = llama_backend.speculative_type,
                 )
@@ -295,6 +297,21 @@ async def load_model(
                     logger.warning(
                         f"Could not retrieve chat template for {backend.active_model_name}: {e}"
                     )
+                # Non-GGUF: only advertise reasoning for gpt-oss Harmony,
+                # which emits reasoning via channels at the tokenizer level.
+                # Template-level chat_template_kwargs (enable_thinking /
+                # preserve_thinking / tools) are not yet forwarded through
+                # the transformers generation path, so avoid advertising
+                # controls the server cannot honour outside GGUF.
+                _sf_supports_reasoning = False
+                _sf_reasoning_style = "enable_thinking"
+                if hasattr(backend, "_is_gpt_oss_model"):
+                    try:
+                        if backend._is_gpt_oss_model():
+                            _sf_supports_reasoning = True
+                            _sf_reasoning_style = "reasoning_effort"
+                    except Exception:
+                        pass
                 return LoadResponse(
                     status = "already_loaded",
                     model = backend.active_model_name,
@@ -309,6 +326,11 @@ async def load_model(
                     requires_trust_remote_code = bool(
                         inference_config.get("trust_remote_code", False)
                     ),
+                    supports_reasoning = _sf_supports_reasoning,
+                    reasoning_style = _sf_reasoning_style,
+                    reasoning_always_on = False,
+                    supports_preserve_thinking = False,
+                    supports_tools = False,
                     chat_template = _chat_template,
                 )
 
@@ -422,7 +444,9 @@ async def load_model(
                 max_context_length = llama_backend.max_context_length,
                 native_context_length = llama_backend.native_context_length,
                 supports_reasoning = llama_backend.supports_reasoning,
+                reasoning_style = llama_backend.reasoning_style,
                 reasoning_always_on = llama_backend.reasoning_always_on,
+                supports_preserve_thinking = llama_backend.supports_preserve_thinking,
                 supports_tools = llama_backend.supports_tools,
                 cache_type_kv = llama_backend.cache_type_kv,
                 chat_template = llama_backend.chat_template,
@@ -545,6 +569,20 @@ async def load_model(
         except Exception:
             pass
 
+        # Non-GGUF: gpt-oss Harmony surfaces reasoning via tokenizer-level
+        # channels; other safetensors reasoning/tools/preserve-thinking
+        # knobs are not forwarded to tokenizer.apply_chat_template yet, so
+        # we only advertise support for the Harmony case here.
+        _sf_supports_reasoning = False
+        _sf_reasoning_style = "enable_thinking"
+        if hasattr(backend, "_is_gpt_oss_model"):
+            try:
+                if backend._is_gpt_oss_model():
+                    _sf_supports_reasoning = True
+                    _sf_reasoning_style = "reasoning_effort"
+            except Exception:
+                pass
+
         return LoadResponse(
             status = "loaded",
             model = config.identifier,
@@ -559,6 +597,11 @@ async def load_model(
             requires_trust_remote_code = bool(
                 inference_config.get("trust_remote_code", False)
             ),
+            supports_reasoning = _sf_supports_reasoning,
+            reasoning_style = _sf_reasoning_style,
+            reasoning_always_on = False,
+            supports_preserve_thinking = False,
+            supports_tools = False,
             chat_template = _chat_template,
         )
 
@@ -766,7 +809,9 @@ async def get_status(
                     (_inference_cfg or {}).get("trust_remote_code", False)
                 ),
                 supports_reasoning = llama_backend.supports_reasoning,
+                reasoning_style = llama_backend.reasoning_style,
                 reasoning_always_on = llama_backend.reasoning_always_on,
+                supports_preserve_thinking = llama_backend.supports_preserve_thinking,
                 supports_tools = llama_backend.supports_tools,
                 context_length = llama_backend.context_length,
                 max_context_length = llama_backend.max_context_length,
@@ -788,10 +833,18 @@ async def get_status(
             audio_type = model_info.get("audio_type")
             has_audio_input = model_info.get("has_audio_input", False)
 
-        # gpt-oss safetensors models support reasoning via harmony channels
+        # Non-GGUF: only gpt-oss Harmony is wired through the transformers
+        # generation path. Other template-level reasoning / tool kwargs
+        # are not yet forwarded, so we do not advertise them here.
         supports_reasoning = False
+        reasoning_style = "enable_thinking"
         if backend.active_model_name and hasattr(backend, "_is_gpt_oss_model"):
-            supports_reasoning = backend._is_gpt_oss_model()
+            try:
+                if backend._is_gpt_oss_model():
+                    supports_reasoning = True
+                    reasoning_style = "reasoning_effort"
+            except Exception:
+                pass
         inference_config = (
             load_inference_config(backend.active_model_name)
             if backend.active_model_name
@@ -812,6 +865,10 @@ async def get_status(
                 (inference_config or {}).get("trust_remote_code", False)
             ),
             supports_reasoning = supports_reasoning,
+            reasoning_style = reasoning_style,
+            reasoning_always_on = False,
+            supports_preserve_thinking = False,
+            supports_tools = False,
         )
 
     except Exception as e:
@@ -1393,6 +1450,8 @@ async def openai_chat_completions(
                     presence_penalty = payload.presence_penalty,
                     cancel_event = cancel_event,
                     enable_thinking = payload.enable_thinking,
+                    reasoning_effort = payload.reasoning_effort,
+                    preserve_thinking = payload.preserve_thinking,
                     auto_heal_tool_calls = payload.auto_heal_tool_calls
                     if payload.auto_heal_tool_calls is not None
                     else True,
@@ -1562,6 +1621,8 @@ async def openai_chat_completions(
                 presence_penalty = payload.presence_penalty,
                 cancel_event = cancel_event,
                 enable_thinking = payload.enable_thinking,
+                reasoning_effort = payload.reasoning_effort,
+                preserve_thinking = payload.preserve_thinking,
             )
 
         _gguf_sentinel = object()
diff --git a/studio/frontend/src/components/assistant-ui/thread.tsx b/studio/frontend/src/components/assistant-ui/thread.tsx
index 0b97d98dfd..3f67b11b51 100644
--- a/studio/frontend/src/components/assistant-ui/thread.tsx
+++ b/studio/frontend/src/components/assistant-ui/thread.tsx
@@ -24,8 +24,15 @@ import {
   useScrollThreadToBottom,
 } from "@/components/assistant-ui/use-intent-aware-autoscroll";
 import { Button } from "@/components/ui/button";
+import {
+  DropdownMenu,
+  DropdownMenuContent,
+  DropdownMenuItem,
+  DropdownMenuTrigger,
+} from "@/components/ui/dropdown-menu";
 import { sentAudioNames } from "@/features/chat/api/chat-adapter";
 import { useChatRuntimeStore } from "@/features/chat/stores/chat-runtime-store";
+import { applyQwenThinkingParams } from "@/features/chat/utils/qwen-params";
 import { deleteThreadMessage } from "@/features/chat/utils/delete-thread-message";
 import { AUDIO_ACCEPT, MAX_AUDIO_SIZE, fileToBase64 } from "@/lib/audio-utils";
 import { copyToClipboard } from "@/lib/copy-to-clipboard";
@@ -373,18 +380,6 @@ const ComposerAudioUpload: FC = () => {
   );
 };
 
-/** Qwen3/3.5 recommended params differ between thinking on/off. */
-function applyQwenThinkingParams(thinkingOn: boolean): void {
-  const store = useChatRuntimeStore.getState();
-  const checkpoint = store.params.checkpoint?.toLowerCase() ?? "";
-  if (!checkpoint.includes("qwen3")) {
-    return;
-  }
-  const params = thinkingOn
-    ? { temperature: 0.6, topP: 0.95, topK: 20, minP: 0.0 }
-    : { temperature: 0.7, topP: 0.8, topK: 20, minP: 0.0 };
-  store.setParams({ ...store.params, ...params });
-}
 
 const ReasoningToggle: FC = () => {
   const modelLoaded = useChatRuntimeStore(
@@ -393,8 +388,49 @@ const ReasoningToggle: FC = () => {
   const supportsReasoning = useChatRuntimeStore((s) => s.supportsReasoning);
   const reasoningEnabled = useChatRuntimeStore((s) => s.reasoningEnabled);
   const setReasoningEnabled = useChatRuntimeStore((s) => s.setReasoningEnabled);
+  const reasoningStyle = useChatRuntimeStore((s) => s.reasoningStyle);
+  const reasoningEffort = useChatRuntimeStore((s) => s.reasoningEffort);
+  const setReasoningEffort = useChatRuntimeStore((s) => s.setReasoningEffort);
   const disabled = !(modelLoaded && supportsReasoning);
 
+  if (reasoningStyle === "reasoning_effort") {
+    return (
+      <DropdownMenu>
+        <DropdownMenuTrigger asChild={true}>
+          <button
+            type="button"
+            disabled={disabled}
+            className={cn(
+              "flex items-center gap-1.5 rounded-full px-2.5 py-1 text-xs font-medium transition-colors",
+              disabled
+                ? "cursor-not-allowed opacity-40"
+                : "bg-primary/10 text-primary hover:bg-primary/20",
+            )}
+            aria-label={`Reasoning effort: ${reasoningEffort}`}
+          >
+            <LightbulbIcon className="size-3.5" />
+            <span>
+              Think:{" "}
+              {reasoningEffort.charAt(0).toUpperCase() +
+                reasoningEffort.slice(1)}
+            </span>
+          </button>
+        </DropdownMenuTrigger>
+        <DropdownMenuContent align="end">
+          {(["low", "medium", "high"] as const).map((level) => (
+            <DropdownMenuItem
+              key={level}
+              onSelect={() => setReasoningEffort(level)}
+            >
+              {level.charAt(0).toUpperCase() + level.slice(1)}
+              {reasoningEffort === level ? " \u2713" : ""}
+            </DropdownMenuItem>
+          ))}
+        </DropdownMenuContent>
+      </DropdownMenu>
+    );
+  }
+
   return (
     <button
       type="button"
@@ -424,6 +460,44 @@ const ReasoningToggle: FC = () => {
   );
 };
 
+const PreserveThinkingToggle: FC = () => {
+  const modelLoaded = useChatRuntimeStore(
+    (s) => !!s.params.checkpoint && !s.modelLoading,
+  );
+  const supportsPreserveThinking = useChatRuntimeStore(
+    (s) => s.supportsPreserveThinking,
+  );
+  const preserveThinking = useChatRuntimeStore((s) => s.preserveThinking);
+  const setPreserveThinking = useChatRuntimeStore((s) => s.setPreserveThinking);
+  if (!supportsPreserveThinking) return null;
+  const disabled = !modelLoaded;
+  return (
+    <button
+      type="button"
+      disabled={disabled}
+      onClick={() => setPreserveThinking(!preserveThinking)}
+      className={cn(
+        "flex items-center gap-1.5 rounded-full px-2.5 py-1 text-xs font-medium transition-colors",
+        disabled
+          ? "cursor-not-allowed opacity-40"
+          : preserveThinking
+            ? "bg-primary/10 text-primary hover:bg-primary/20"
+            : "bg-muted text-muted-foreground hover:bg-muted-foreground/15",
+      )}
+      aria-label={
+        preserveThinking ? "Disable preserve thinking" : "Enable preserve thinking"
+      }
+    >
+      {preserveThinking && !disabled ? (
+        <LightbulbIcon className="size-3.5" />
+      ) : (
+        <LightbulbOffIcon className="size-3.5" />
+      )}
+      <span>Preserve Thinking</span>
+    </button>
+  );
+};
+
 const WebSearchToggle: FC = () => {
   const modelLoaded = useChatRuntimeStore(
     (s) => !!s.params.checkpoint && !s.modelLoading,
@@ -551,6 +625,7 @@ const ComposerAction: FC<{ disabled?: boolean }> = ({ disabled }) => {
         <ComposerAddAttachment />
         <ComposerAudioUpload />
         <ReasoningToggle />
+        <PreserveThinkingToggle />
         <WebSearchToggle />
         <CodeToolsToggle />
       </div>
diff --git a/studio/frontend/src/features/chat/api/chat-adapter.ts b/studio/frontend/src/features/chat/api/chat-adapter.ts
index 53a935cf45..a93120397e 100644
--- a/studio/frontend/src/features/chat/api/chat-adapter.ts
+++ b/studio/frontend/src/features/chat/api/chat-adapter.ts
@@ -385,6 +385,8 @@ async function autoLoadSmallestModel(): Promise<{
               supportsReasoning: loadResp.supports_reasoning ?? false,
               reasoningAlwaysOn: loadResp.reasoning_always_on ?? false,
               reasoningEnabled: loadResp.supports_reasoning ?? false,
+              reasoningStyle: loadResp.reasoning_style ?? "enable_thinking",
+              supportsPreserveThinking: loadResp.supports_preserve_thinking ?? false,
               supportsTools: loadResp.supports_tools ?? false,
               toolsEnabled: loadResp.supports_tools ?? false,
               codeToolsEnabled: loadResp.supports_tools ?? false,
@@ -433,6 +435,14 @@ async function autoLoadSmallestModel(): Promise<{
             sfLoadResp.requires_trust_remote_code ?? false,
           );
           store.setParams({ ...store.params, maxTokens: 4096 });
+          useChatRuntimeStore.setState({
+            supportsReasoning: sfLoadResp.supports_reasoning ?? false,
+            reasoningAlwaysOn: sfLoadResp.reasoning_always_on ?? false,
+            reasoningEnabled: sfLoadResp.supports_reasoning ?? false,
+            reasoningStyle: sfLoadResp.reasoning_style ?? "enable_thinking",
+            supportsPreserveThinking: sfLoadResp.supports_preserve_thinking ?? false,
+            supportsTools: sfLoadResp.supports_tools ?? false,
+          });
           const sfModel: ChatModelSummary = {
             id: repo.repo_id,
             name: sfLoadResp.display_name ?? repo.repo_id,
@@ -501,6 +511,8 @@ async function autoLoadSmallestModel(): Promise<{
         supportsReasoning: loadResp.supports_reasoning ?? false,
         reasoningAlwaysOn: loadResp.reasoning_always_on ?? false,
         reasoningEnabled: loadResp.supports_reasoning ?? false,
+        reasoningStyle: loadResp.reasoning_style ?? "enable_thinking",
+        supportsPreserveThinking: loadResp.supports_preserve_thinking ?? false,
         supportsTools: loadResp.supports_tools ?? false,
         toolsEnabled: loadResp.supports_tools ?? false,
         codeToolsEnabled: loadResp.supports_tools ?? false,
@@ -696,7 +708,14 @@ export function createOpenAIStreamAdapter(): ChatModelAdapter {
       let serverMetadata: { usage?: ServerUsage; timings?: ServerTimings } | null = null;
 
       try {
-        const { supportsReasoning, reasoningEnabled } = runtime;
+        const {
+          supportsReasoning,
+          reasoningEnabled,
+          reasoningStyle,
+          reasoningEffort,
+          supportsPreserveThinking,
+          preserveThinking,
+        } = runtime;
         const stream = streamChatCompletions(
           {
             model: params.checkpoint,
@@ -712,7 +731,12 @@ export function createOpenAIStreamAdapter(): ChatModelAdapter {
             image_base64: imageBase64,
             audio_base64: audioBase64,
             ...(useAdapter === undefined ? {} : { use_adapter: useAdapter }),
-            ...(supportsReasoning ? { enable_thinking: reasoningEnabled } : {}),
+            ...(supportsReasoning
+              ? reasoningStyle === "reasoning_effort"
+                ? { reasoning_effort: reasoningEffort }
+                : { enable_thinking: reasoningEnabled }
+              : {}),
+            ...(supportsPreserveThinking ? { preserve_thinking: preserveThinking } : {}),
             ...(supportsTools && (toolsEnabled || codeToolsEnabled)
               ? {
                   enable_tools: true,
diff --git a/studio/frontend/src/features/chat/hooks/use-chat-model-runtime.ts b/studio/frontend/src/features/chat/hooks/use-chat-model-runtime.ts
index 6c870d74a9..c9dcd911a2 100644
--- a/studio/frontend/src/features/chat/hooks/use-chat-model-runtime.ts
+++ b/studio/frontend/src/features/chat/hooks/use-chat-model-runtime.ts
@@ -253,7 +253,9 @@ export function useChatModelRuntime() {
             max_context_length: statusRes.max_context_length,
             native_context_length: statusRes.native_context_length,
             supports_reasoning: statusRes.supports_reasoning,
+            reasoning_style: statusRes.reasoning_style,
             reasoning_always_on: statusRes.reasoning_always_on,
+            supports_preserve_thinking: statusRes.supports_preserve_thinking,
             supports_tools: statusRes.supports_tools,
             speculative_type: statusRes.speculative_type,
           };
@@ -265,6 +267,8 @@ export function useChatModelRuntime() {
         // Restore reasoning/tools support flags and context length
         const supportsReasoning = statusRes.supports_reasoning ?? false;
         const reasoningAlwaysOn = statusRes.reasoning_always_on ?? false;
+        const reasoningStyle = statusRes.reasoning_style ?? "enable_thinking";
+        const supportsPreserveThinking = statusRes.supports_preserve_thinking ?? false;
         const supportsTools = statusRes.supports_tools ?? false;
         const currentGgufContextLength = statusRes.is_gguf
           ? (statusRes.context_length ?? null)
@@ -279,7 +283,14 @@ export function useChatModelRuntime() {
         useChatRuntimeStore.setState({
           supportsReasoning,
           reasoningAlwaysOn,
+          reasoningStyle,
+          supportsPreserveThinking,
           supportsTools,
+          // Reset per-turn reasoning flag so models that do not support
+          // reasoning do not inherit a stale off state from a prior model.
+          reasoningEnabled: supportsReasoning
+            ? useChatRuntimeStore.getState().reasoningEnabled
+            : true,
           ggufContextLength: currentGgufContextLength,
           ggufMaxContextLength,
           ggufNativeContextLength,
@@ -498,6 +509,8 @@ export function useChatModelRuntime() {
               supportsReasoning: loadResponse.supports_reasoning ?? false,
               reasoningAlwaysOn,
               reasoningEnabled: reasoningAlwaysOn ? true : reasoningDefault,
+              reasoningStyle: loadResponse.reasoning_style ?? "enable_thinking",
+              supportsPreserveThinking: loadResponse.supports_preserve_thinking ?? false,
               supportsTools: loadResponse.supports_tools ?? false,
               toolsEnabled: loadResponse.supports_tools ?? false,
               codeToolsEnabled: loadResponse.supports_tools ?? false,
diff --git a/studio/frontend/src/features/chat/shared-composer.tsx b/studio/frontend/src/features/chat/shared-composer.tsx
index e133669a1c..3ba2f8b6f5 100644
--- a/studio/frontend/src/features/chat/shared-composer.tsx
+++ b/studio/frontend/src/features/chat/shared-composer.tsx
@@ -4,6 +4,13 @@
 import { TooltipIconButton } from "@/components/assistant-ui/tooltip-icon-button";
 import { CodeToggleIcon } from "@/components/assistant-ui/code-toggle-icon";
 import { Button } from "@/components/ui/button";
+import {
+  DropdownMenu,
+  DropdownMenuContent,
+  DropdownMenuItem,
+  DropdownMenuTrigger,
+} from "@/components/ui/dropdown-menu";
+import { applyQwenThinkingParams } from "@/features/chat/utils/qwen-params";
 import { AUDIO_ACCEPT, MAX_AUDIO_SIZE, fileToBase64 } from "@/lib/audio-utils";
 import { useAui } from "@assistant-ui/react";
 import { cn } from "@/lib/utils";
@@ -245,6 +252,12 @@ export function SharedComposer({
   const reasoningAlwaysOn = useChatRuntimeStore((s) => s.reasoningAlwaysOn);
   const reasoningEnabled = useChatRuntimeStore((s) => s.reasoningEnabled);
   const setReasoningEnabled = useChatRuntimeStore((s) => s.setReasoningEnabled);
+  const reasoningStyle = useChatRuntimeStore((s) => s.reasoningStyle);
+  const reasoningEffort = useChatRuntimeStore((s) => s.reasoningEffort);
+  const setReasoningEffort = useChatRuntimeStore((s) => s.setReasoningEffort);
+  const supportsPreserveThinking = useChatRuntimeStore((s) => s.supportsPreserveThinking);
+  const preserveThinking = useChatRuntimeStore((s) => s.preserveThinking);
+  const setPreserveThinking = useChatRuntimeStore((s) => s.setPreserveThinking);
   const supportsTools = useChatRuntimeStore((s) => s.supportsTools);
   const toolsEnabled = useChatRuntimeStore((s) => s.toolsEnabled);
   const setToolsEnabled = useChatRuntimeStore((s) => s.setToolsEnabled);
@@ -391,6 +404,13 @@ export function SharedComposer({
         store.setModelRequiresTrustRemoteCode(
           resp.requires_trust_remote_code ?? false,
         );
+        useChatRuntimeStore.setState({
+          supportsReasoning: resp.supports_reasoning ?? false,
+          reasoningAlwaysOn: resp.reasoning_always_on ?? false,
+          reasoningStyle: resp.reasoning_style ?? "enable_thinking",
+          supportsPreserveThinking: resp.supports_preserve_thinking ?? false,
+          supportsTools: resp.supports_tools ?? false,
+        });
         return resp.status;
       }
 
@@ -566,41 +586,93 @@ export function SharedComposer({
               </TooltipIconButton>
             </>
           )}
-          <button
-            type="button"
-            disabled={reasoningDisabled}
-            onClick={() => {
-              if (reasoningAlwaysOn) return;
-              const next = !reasoningEnabled;
-              setReasoningEnabled(next);
-              // Qwen3/3.5/3.6: adjust params for thinking on/off
-              const store = useChatRuntimeStore.getState();
-              const cp = store.params.checkpoint?.toLowerCase() ?? "";
-              if (cp.includes("qwen3")) {
-                const needsPresencePenalty = cp.includes("qwen3.5") || cp.includes("qwen3.6");
-                const p = next
-                  ? { temperature: 0.6, topP: 0.95, topK: 20, minP: 0.0, ...(needsPresencePenalty ? { presencePenalty: 1.5 } : {}) }
-                  : { temperature: 0.7, topP: 0.8, topK: 20, minP: 0.0, ...(needsPresencePenalty ? { presencePenalty: 1.5 } : {}) };
-                store.setParams({ ...store.params, ...p });
+          {reasoningStyle === "reasoning_effort" ? (
+            <DropdownMenu>
+              <DropdownMenuTrigger asChild={true}>
+                <button
+                  type="button"
+                  disabled={reasoningDisabled}
+                  className={cn(
+                    "flex items-center gap-1.5 rounded-full px-2.5 py-1 text-xs font-medium transition-colors",
+                    reasoningDisabled
+                      ? "cursor-not-allowed opacity-40"
+                      : "bg-primary/10 text-primary hover:bg-primary/20",
+                  )}
+                  aria-label={`Reasoning effort: ${reasoningEffort}`}
+                >
+                  <LightbulbIcon className="size-3.5" />
+                  <span>
+                    Think:{" "}
+                    {reasoningEffort.charAt(0).toUpperCase() +
+                      reasoningEffort.slice(1)}
+                  </span>
+                </button>
+              </DropdownMenuTrigger>
+              <DropdownMenuContent align="end">
+                {(["low", "medium", "high"] as const).map((level) => (
+                  <DropdownMenuItem
+                    key={level}
+                    onSelect={() => setReasoningEffort(level)}
+                  >
+                    {level.charAt(0).toUpperCase() + level.slice(1)}
+                    {reasoningEffort === level ? " \u2713" : ""}
+                  </DropdownMenuItem>
+                ))}
+              </DropdownMenuContent>
+            </DropdownMenu>
+          ) : (
+            <button
+              type="button"
+              disabled={reasoningDisabled}
+              onClick={() => {
+                if (reasoningAlwaysOn) return;
+                const next = !reasoningEnabled;
+                setReasoningEnabled(next);
+                applyQwenThinkingParams(next);
+              }}
+              className={cn(
+                "flex items-center gap-1.5 rounded-full px-2.5 py-1 text-xs font-medium transition-colors",
+                reasoningDisabled
+                  ? "cursor-not-allowed opacity-40"
+                  : (reasoningEnabled || reasoningAlwaysOn)
+                    ? "bg-primary/10 text-primary hover:bg-primary/20"
+                    : "bg-muted text-muted-foreground hover:bg-muted-foreground/15",
+              )}
+              aria-label={reasoningEnabled ? "Disable thinking" : "Enable thinking"}
+            >
+              {(reasoningEnabled || reasoningAlwaysOn) && !reasoningDisabled ? (
+                <LightbulbIcon className="size-3.5" />
+              ) : (
+                <LightbulbOffIcon className="size-3.5" />
+              )}
+              <span>Think</span>
+            </button>
+          )}
+          {supportsPreserveThinking && (
+            <button
+              type="button"
+              disabled={!modelLoaded}
+              onClick={() => setPreserveThinking(!preserveThinking)}
+              className={cn(
+                "flex items-center gap-1.5 rounded-full px-2.5 py-1 text-xs font-medium transition-colors",
+                !modelLoaded
+                  ? "cursor-not-allowed opacity-40"
+                  : preserveThinking
+                    ? "bg-primary/10 text-primary hover:bg-primary/20"
+                    : "bg-muted text-muted-foreground hover:bg-muted-foreground/15",
+              )}
+              aria-label={
+                preserveThinking ? "Disable preserve thinking" : "Enable preserve thinking"
               }
-            }}
-            className={cn(
-              "flex items-center gap-1.5 rounded-full px-2.5 py-1 text-xs font-medium transition-colors",
-              reasoningDisabled
-                ? "cursor-not-allowed opacity-40"
-                : (reasoningEnabled || reasoningAlwaysOn)
-                  ? "bg-primary/10 text-primary hover:bg-primary/20"
-                  : "bg-muted text-muted-foreground hover:bg-muted-foreground/15",
-            )}
-            aria-label={reasoningEnabled ? "Disable thinking" : "Enable thinking"}
-          >
-            {(reasoningEnabled || reasoningAlwaysOn) && !reasoningDisabled ? (
-              <LightbulbIcon className="size-3.5" />
-            ) : (
-              <LightbulbOffIcon className="size-3.5" />
-            )}
-            <span>Think</span>
-          </button>
+            >
+              {preserveThinking && modelLoaded ? (
+                <LightbulbIcon className="size-3.5" />
+              ) : (
+                <LightbulbOffIcon className="size-3.5" />
+              )}
+              <span>Preserve Thinking</span>
+            </button>
+          )}
           <button
             type="button"
             disabled={toolsDisabled}
diff --git a/studio/frontend/src/features/chat/stores/chat-runtime-store.ts b/studio/frontend/src/features/chat/stores/chat-runtime-store.ts
index 821716fe0a..2a1249aea0 100644
--- a/studio/frontend/src/features/chat/stores/chat-runtime-store.ts
+++ b/studio/frontend/src/features/chat/stores/chat-runtime-store.ts
@@ -16,6 +16,22 @@ const MAX_TOOL_CALLS_KEY = "unsloth_max_tool_calls_per_message";
 const TOOL_CALL_TIMEOUT_KEY = "unsloth_tool_call_timeout";
 const HF_TOKEN_KEY = "unsloth_hf_token";
 const INFERENCE_PARAMS_KEY = "unsloth_chat_inference_params";
+const REASONING_EFFORT_KEY = "unsloth_reasoning_effort";
+const PRESERVE_THINKING_KEY = "unsloth_preserve_thinking";
+
+export type ReasoningStyle = "enable_thinking" | "reasoning_effort";
+export type ReasoningEffort = "low" | "medium" | "high";
+
+function loadReasoningEffort(fallback: ReasoningEffort): ReasoningEffort {
+  if (!canUseStorage()) return fallback;
+  try {
+    const raw = localStorage.getItem(REASONING_EFFORT_KEY);
+    if (raw === "low" || raw === "medium" || raw === "high") return raw;
+    return fallback;
+  } catch {
+    return fallback;
+  }
+}
 let hasShownInferencePersistenceWarning = false;
 
 function canUseStorage(): boolean {
@@ -158,6 +174,10 @@ type ChatRuntimeStore = {
   supportsReasoning: boolean;
   reasoningAlwaysOn: boolean;
   reasoningEnabled: boolean;
+  reasoningStyle: ReasoningStyle;
+  reasoningEffort: ReasoningEffort;
+  supportsPreserveThinking: boolean;
+  preserveThinking: boolean;
   supportsTools: boolean;
   toolsEnabled: boolean;
   codeToolsEnabled: boolean;
@@ -200,6 +220,9 @@ type ChatRuntimeStore = {
   setSettingsPanelOpen: (open: boolean) => void;
   clearCheckpoint: () => void;
   setReasoningEnabled: (enabled: boolean) => void;
+  setReasoningStyle: (style: ReasoningStyle) => void;
+  setReasoningEffort: (effort: ReasoningEffort) => void;
+  setPreserveThinking: (value: boolean) => void;
   setToolsEnabled: (enabled: boolean) => void;
   setCodeToolsEnabled: (enabled: boolean) => void;
   setToolStatus: (status: string | null) => void;
@@ -233,6 +256,10 @@ export const useChatRuntimeStore = create<ChatRuntimeStore>((set) => ({
   supportsReasoning: false,
   reasoningAlwaysOn: false,
   reasoningEnabled: true,
+  reasoningStyle: "enable_thinking",
+  reasoningEffort: loadReasoningEffort("medium"),
+  supportsPreserveThinking: false,
+  preserveThinking: loadBool(PRESERVE_THINKING_KEY, false),
   supportsTools: false,
   toolsEnabled: false,
   codeToolsEnabled: false,
@@ -328,7 +355,10 @@ export const useChatRuntimeStore = create<ChatRuntimeStore>((set) => ({
       modelRequiresTrustRemoteCode: false,
       contextUsage: null,
       supportsReasoning: false,
+      reasoningAlwaysOn: false,
       reasoningEnabled: true,
+      reasoningStyle: "enable_thinking",
+      supportsPreserveThinking: false,
       supportsTools: false,
       toolsEnabled: false,
       codeToolsEnabled: false,
@@ -342,6 +372,23 @@ export const useChatRuntimeStore = create<ChatRuntimeStore>((set) => ({
       chatTemplateOverride: null,
     })),
   setReasoningEnabled: (reasoningEnabled) => set({ reasoningEnabled }),
+  setReasoningStyle: (reasoningStyle) => set({ reasoningStyle }),
+  setReasoningEffort: (reasoningEffort) =>
+    set(() => {
+      if (canUseStorage()) {
+        try {
+          localStorage.setItem(REASONING_EFFORT_KEY, reasoningEffort);
+        } catch {
+          // ignore
+        }
+      }
+      return { reasoningEffort };
+    }),
+  setPreserveThinking: (preserveThinking) =>
+    set(() => {
+      saveBool(PRESERVE_THINKING_KEY, preserveThinking);
+      return { preserveThinking };
+    }),
   setToolsEnabled: (toolsEnabled) => set({ toolsEnabled }),
   setCodeToolsEnabled: (codeToolsEnabled) => set({ codeToolsEnabled }),
   setToolStatus: (toolStatus) => set({ toolStatus }),
diff --git a/studio/frontend/src/features/chat/types/api.ts b/studio/frontend/src/features/chat/types/api.ts
index 133ae2adee..ccfc8b2bf1 100644
--- a/studio/frontend/src/features/chat/types/api.ts
+++ b/studio/frontend/src/features/chat/types/api.ts
@@ -92,7 +92,9 @@ export interface LoadModelResponse {
   max_context_length?: number | null;
   native_context_length?: number | null;
   supports_reasoning?: boolean;
+  reasoning_style?: "enable_thinking" | "reasoning_effort";
   reasoning_always_on?: boolean;
+  supports_preserve_thinking?: boolean;
   supports_tools?: boolean;
   cache_type_kv?: string | null;
   chat_template?: string | null;
@@ -123,7 +125,9 @@ export interface InferenceStatusResponse {
   };
   requires_trust_remote_code?: boolean;
   supports_reasoning?: boolean;
+  reasoning_style?: "enable_thinking" | "reasoning_effort";
   reasoning_always_on?: boolean;
+  supports_preserve_thinking?: boolean;
   supports_tools?: boolean;
   context_length?: number | null;
   max_context_length?: number | null;
@@ -167,6 +171,8 @@ export interface OpenAIChatCompletionsRequest {
   audio_base64?: string;
   use_adapter?: boolean | string | null;
   enable_thinking?: boolean | null;
+  reasoning_effort?: "low" | "medium" | "high" | null;
+  preserve_thinking?: boolean | null;
   enable_tools?: boolean | null;
   enabled_tools?: string[];
   auto_heal_tool_calls?: boolean;
diff --git a/studio/frontend/src/features/chat/utils/qwen-params.ts b/studio/frontend/src/features/chat/utils/qwen-params.ts
new file mode 100644
index 0000000000..5eaa2388da
--- /dev/null
+++ b/studio/frontend/src/features/chat/utils/qwen-params.ts
@@ -0,0 +1,29 @@
+// SPDX-License-Identifier: AGPL-3.0-only
+// Copyright 2026-present the Unsloth AI Inc. team. All rights reserved. See /studio/LICENSE.AGPL-3.0
+
+import { useChatRuntimeStore } from "../stores/chat-runtime-store";
+
+/**
+ * Apply Qwen3-family recommended sampling parameters when the Think toggle
+ * changes. Qwen3.5 and Qwen3.6 also need a presence_penalty bump on top of
+ * the Qwen3 defaults.
+ *
+ * Used by both the thread assistant UI and the shared chat composer so the
+ * two call sites stay in sync.
+ */
+export function applyQwenThinkingParams(thinkingOn: boolean): void {
+  const store = useChatRuntimeStore.getState();
+  const checkpoint = store.params.checkpoint?.toLowerCase() ?? "";
+  if (!checkpoint.includes("qwen3")) {
+    return;
+  }
+  const needsPresencePenalty =
+    checkpoint.includes("qwen3.5") || checkpoint.includes("qwen3.6");
+  const base = thinkingOn
+    ? { temperature: 0.6, topP: 0.95, topK: 20, minP: 0.0 }
+    : { temperature: 0.7, topP: 0.8, topK: 20, minP: 0.0 };
+  const params = needsPresencePenalty
+    ? { ...base, presencePenalty: 1.5 }
+    : base;
+  store.setParams({ ...store.params, ...params });
+}

From 5c473fab80e079bb525345b86cb71afd409262c3 Mon Sep 17 00:00:00 2001
From: Daniel Han <danielhanchen@gmail.com>
Date: Thu, 23 Apr 2026 07:05:47 -0700
Subject: [PATCH 161/216] Bump versions

---
 install.ps1 | 10 +++++-----
 install.sh  | 10 +++++-----
 2 files changed, 10 insertions(+), 10 deletions(-)

diff --git a/install.ps1 b/install.ps1
index 44464101f3..77c0034125 100644
--- a/install.ps1
+++ b/install.ps1
@@ -934,7 +934,7 @@ shell.Run cmd, 0, False
         if ($SkipTorch) {
             # No-torch: install unsloth + unsloth-zoo with --no-deps, then
             # runtime deps (typer, safetensors, transformers, etc.) with --no-deps.
-            $baseInstallExit = Invoke-InstallCommand { uv pip install --python $VenvPython --no-deps --reinstall-package unsloth --reinstall-package unsloth-zoo "unsloth>=2026.4.7" unsloth-zoo }
+            $baseInstallExit = Invoke-InstallCommand { uv pip install --python $VenvPython --no-deps --reinstall-package unsloth --reinstall-package unsloth-zoo "unsloth>=2026.4.8" unsloth-zoo }
             if ($baseInstallExit -eq 0) {
                 $NoTorchReq = Find-NoTorchRuntimeFile
                 if ($NoTorchReq) {
@@ -942,7 +942,7 @@ shell.Run cmd, 0, False
                 }
             }
         } else {
-            $baseInstallExit = Invoke-InstallCommand { uv pip install --python $VenvPython --reinstall-package unsloth --reinstall-package unsloth-zoo "unsloth>=2026.4.7" unsloth-zoo }
+            $baseInstallExit = Invoke-InstallCommand { uv pip install --python $VenvPython --reinstall-package unsloth --reinstall-package unsloth-zoo "unsloth>=2026.4.8" unsloth-zoo }
         }
         if ($baseInstallExit -ne 0) {
             Write-Host "[ERROR] Failed to install unsloth (exit code $baseInstallExit)" -ForegroundColor Red
@@ -975,7 +975,7 @@ shell.Run cmd, 0, False
         if ($SkipTorch) {
             # No-torch: install unsloth + unsloth-zoo with --no-deps, then
             # runtime deps (typer, safetensors, transformers, etc.) with --no-deps.
-            $baseInstallExit = Invoke-InstallCommand { uv pip install --python $VenvPython --no-deps --upgrade-package unsloth --upgrade-package unsloth-zoo "unsloth>=2026.4.7" unsloth-zoo }
+            $baseInstallExit = Invoke-InstallCommand { uv pip install --python $VenvPython --no-deps --upgrade-package unsloth --upgrade-package unsloth-zoo "unsloth>=2026.4.8" unsloth-zoo }
             if ($baseInstallExit -eq 0) {
                 $NoTorchReq = Find-NoTorchRuntimeFile
                 if ($NoTorchReq) {
@@ -983,7 +983,7 @@ shell.Run cmd, 0, False
                 }
             }
         } elseif ($StudioLocalInstall) {
-            $baseInstallExit = Invoke-InstallCommand { uv pip install --python $VenvPython --upgrade-package unsloth "unsloth>=2026.4.7" unsloth-zoo }
+            $baseInstallExit = Invoke-InstallCommand { uv pip install --python $VenvPython --upgrade-package unsloth "unsloth>=2026.4.8" unsloth-zoo }
         } else {
             $baseInstallExit = Invoke-InstallCommand { uv pip install --python $VenvPython --upgrade-package unsloth -- "$PackageName" }
         }
@@ -1006,7 +1006,7 @@ shell.Run cmd, 0, False
         Write-TauriLog "STEP" "Installing unsloth"
         substep "installing unsloth (this may take a few minutes)..."
         if ($StudioLocalInstall) {
-            $baseInstallExit = Invoke-InstallCommand { uv pip install --python $VenvPython unsloth-zoo "unsloth>=2026.4.7" --torch-backend=auto }
+            $baseInstallExit = Invoke-InstallCommand { uv pip install --python $VenvPython unsloth-zoo "unsloth>=2026.4.8" --torch-backend=auto }
             if ($baseInstallExit -ne 0) {
                 Write-Host "[ERROR] Failed to install unsloth (exit code $baseInstallExit)" -ForegroundColor Red
                 return
diff --git a/install.sh b/install.sh
index 07473e441d..6c28b6eda4 100755
--- a/install.sh
+++ b/install.sh
@@ -1347,7 +1347,7 @@ if [ "$_MIGRATED" = true ]; then
         # to prevent transitive torch resolution.
         run_install_cmd "install unsloth (migrated no-torch)" uv pip install --python "$_VENV_PY" --no-deps \
             --reinstall-package unsloth --reinstall-package unsloth-zoo \
-            "unsloth>=2026.4.7" unsloth-zoo
+            "unsloth>=2026.4.8" unsloth-zoo
         _NO_TORCH_RT="$(_find_no_torch_runtime)"
         if [ -n "$_NO_TORCH_RT" ]; then
             run_install_cmd "install no-torch runtime deps" uv pip install --python "$_VENV_PY" --no-deps -r "$_NO_TORCH_RT"
@@ -1355,7 +1355,7 @@ if [ "$_MIGRATED" = true ]; then
     else
         run_install_cmd "install unsloth (migrated)" uv pip install --python "$_VENV_PY" \
             --reinstall-package unsloth --reinstall-package unsloth-zoo \
-            "unsloth>=2026.4.7" unsloth-zoo
+            "unsloth>=2026.4.8" unsloth-zoo
     fi
     if [ "$STUDIO_LOCAL_INSTALL" = true ]; then
         substep "overlaying local repo (editable)..."
@@ -1519,7 +1519,7 @@ elif [ -n "$TORCH_INDEX_URL" ]; then
         # runtime deps (typer, safetensors, transformers, etc.) with --no-deps.
         run_install_cmd "install unsloth (no-torch)" uv pip install --python "$_VENV_PY" --no-deps \
             --upgrade-package unsloth --upgrade-package unsloth-zoo \
-            "unsloth>=2026.4.7" unsloth-zoo
+            "unsloth>=2026.4.8" unsloth-zoo
         _NO_TORCH_RT="$(_find_no_torch_runtime)"
         if [ -n "$_NO_TORCH_RT" ]; then
             run_install_cmd "install no-torch runtime deps" uv pip install --python "$_VENV_PY" --no-deps -r "$_NO_TORCH_RT"
@@ -1530,7 +1530,7 @@ elif [ -n "$TORCH_INDEX_URL" ]; then
         fi
     elif [ "$STUDIO_LOCAL_INSTALL" = true ]; then
         run_install_cmd "install unsloth (local)" uv pip install --python "$_VENV_PY" \
-            --upgrade-package unsloth "unsloth>=2026.4.7" unsloth-zoo
+            --upgrade-package unsloth "unsloth>=2026.4.8" unsloth-zoo
         substep "overlaying local repo (editable)..."
         run_install_cmd "overlay local repo" uv pip install --python "$_VENV_PY" -e "$_REPO_ROOT" --no-deps
     else
@@ -1558,7 +1558,7 @@ else
     tauri_log "STEP" "Installing Unsloth"
     substep "installing unsloth (this may take a few minutes)..."
     if [ "$STUDIO_LOCAL_INSTALL" = true ]; then
-        run_install_cmd "install unsloth (auto torch backend)" uv pip install --python "$_VENV_PY" unsloth-zoo "unsloth>=2026.4.7" --torch-backend=auto
+        run_install_cmd "install unsloth (auto torch backend)" uv pip install --python "$_VENV_PY" unsloth-zoo "unsloth>=2026.4.8" --torch-backend=auto
         substep "overlaying local repo (editable)..."
         run_install_cmd "overlay local repo" uv pip install --python "$_VENV_PY" -e "$_REPO_ROOT" --no-deps
     else

From c875dc17e4f608984c5aa2def361c7aa863d9860 Mon Sep 17 00:00:00 2001
From: Konstantin Azizov <azizovkostya97@gmail.com>
Date: Fri, 24 Apr 2026 12:53:10 +0200
Subject: [PATCH 162/216] Studio: use (gguf) context length before max seq
 length (#5111)

* fix: use (gguf) context length before max seq length

For GGUF models context length is used instead of `maxSeqLength`

Fixes #4893

* Studio: split rollback max_seq_length from target load/validate

Restore params.maxSeqLength as the value passed to the target model's
validateModel and effectiveMaxSeqLength fallback, and introduce a
separate rollbackMaxSeqLength used only in the rollback loadModel call.
Without the split, switching from a GGUF model with a large native
context to a non-GGUF target polluted the new load via the
effectiveMaxSeqLength else-branch.

Also:
- Detect a previous GGUF via isGguf, activeGgufVariant, or a .gguf
  suffix on the checkpoint, so local/LM Studio paths not present in
  the models catalog still take the GGUF rollback path.
- Use 0 as the GGUF rollback fallback to match the sentinel used at
  the normal GGUF load site (ggufContextLength fallback to 0); 4096
  would reintroduce the original truncation bug when both
  customContextLength and ggufContextLength are null.
- Reuse the captured stateBeforeUnload for modelRequiresTrustRemoteCode
  and drop the now-unused DEFAULT_INFERENCE_PARAMS value import.

* Studio: drop pending customContextLength from GGUF rollback

rollbackMaxSeqLength previously preferred customContextLength over
ggufContextLength, but customContextLength is a pending, not-yet-applied
slider edit: chat-settings-sheet.tsx treats it as the "dirty" marker
(ctxDirty = customContextLength !== null) and use-chat-model-runtime.ts
clears it to null on every successful load. Rollback exists to restore
the previously-loaded model at its confirmed running context, so leaking
a pending slider value can load the rollback target at a context the
user never validated against VRAM, potentially causing the rollback
itself to fail.

---------

Co-authored-by: Lee Jackson <130007945+Imagineer99@users.noreply.github.com>
Co-authored-by: Daniel Han <danielhanchen@gmail.com>
---
 .../chat/hooks/use-chat-model-runtime.ts      | 19 +++++++++++++------
 1 file changed, 13 insertions(+), 6 deletions(-)

diff --git a/studio/frontend/src/features/chat/hooks/use-chat-model-runtime.ts b/studio/frontend/src/features/chat/hooks/use-chat-model-runtime.ts
index c9dcd911a2..1281592168 100644
--- a/studio/frontend/src/features/chat/hooks/use-chat-model-runtime.ts
+++ b/studio/frontend/src/features/chat/hooks/use-chat-model-runtime.ts
@@ -419,12 +419,19 @@ export function useChatModelRuntime() {
           let previousWasUnloaded = false;
           const currentCheckpoint =
             useChatRuntimeStore.getState().params.checkpoint;
-          const paramsBeforeLoad = useChatRuntimeStore.getState().params;
-          const trustRemoteCode = paramsBeforeLoad.trustRemoteCode ?? false;
-          const maxSeqLength = paramsBeforeLoad.maxSeqLength;
-          const hfToken = useChatRuntimeStore.getState().hfToken || null;
+          const stateBeforeUnload = useChatRuntimeStore.getState();
+          const trustRemoteCode = stateBeforeUnload.params.trustRemoteCode ?? false;
+          const maxSeqLength = stateBeforeUnload.params.maxSeqLength;
+          const previousIsGguf =
+            previousModel?.isGguf === true
+            || previousVariant != null
+            || (previousCheckpoint?.toLowerCase().endsWith(".gguf") ?? false);
+          const rollbackMaxSeqLength = previousIsGguf
+            ? (stateBeforeUnload.ggufContextLength ?? 0)
+            : maxSeqLength;
+          const hfToken = stateBeforeUnload.hfToken || null;
           const previousModelRequiresTrustRemoteCode =
-            useChatRuntimeStore.getState().modelRequiresTrustRemoteCode;
+            stateBeforeUnload.modelRequiresTrustRemoteCode;
           try {
             // Lightweight pre-flight validation: avoid unloading a working model
             // if the new identifier is clearly invalid (e.g. bad HF id / path).
@@ -542,7 +549,7 @@ export function useChatModelRuntime() {
                 await loadModel({
                   model_path: previousCheckpoint,
                   hf_token: hfToken,
-                  max_seq_length: maxSeqLength,
+                  max_seq_length: rollbackMaxSeqLength,
                   load_in_4bit: true,
                   is_lora: previousIsLora,
                   gguf_variant: previousVariant,

From 06ed94da0d1958c894e68bec863e5ec3d7254c71 Mon Sep 17 00:00:00 2001
From: luo jiyin <luojiyin@hotmail.com>
Date: Fri, 24 Apr 2026 19:51:27 +0800
Subject: [PATCH 163/216] chore: fix typo cleanup across tests and backend
 strings (#5152)

* chore: fix typos in studio/backend/routes/models.py

* chore: fix typos in tests/saving/non_peft/test_mistral_non_peft.py

* chore: fix typos in tests/saving/non_peft/test_whisper_non_peft.py

* chore: fix typos in tests/saving/vision_models/test_index_file_sharded_model.py

* chore: fix typos in tests/saving/vision_models/test_push_to_hub_merged.py

* chore: fix typos in tests/saving/vision_models/test_save_merge_qwen2.5vl32B_model_ocr_benchmark.py

* chore: fix typos in tests/saving/vision_models/test_save_merge_vision_model_ocr_benchmark.py

* chore: fix typos in unsloth/import_fixes.py

* Split: keep only 6 file(s)

---------

Co-authored-by: Daniel Han <danielhanchen@gmail.com>
Co-authored-by: Lee Jackson <130007945+Imagineer99@users.noreply.github.com>
---
 tests/saving/non_peft/test_mistral_non_peft.py                | 2 +-
 tests/saving/non_peft/test_whisper_non_peft.py                | 2 +-
 tests/saving/vision_models/test_index_file_sharded_model.py   | 4 ++--
 tests/saving/vision_models/test_push_to_hub_merged.py         | 4 ++--
 .../test_save_merge_qwen2.5vl32B_model_ocr_benchmark.py       | 4 ++--
 .../test_save_merge_vision_model_ocr_benchmark.py             | 4 ++--
 6 files changed, 10 insertions(+), 10 deletions(-)

diff --git a/tests/saving/non_peft/test_mistral_non_peft.py b/tests/saving/non_peft/test_mistral_non_peft.py
index e03813367d..b2610360fe 100644
--- a/tests/saving/non_peft/test_mistral_non_peft.py
+++ b/tests/saving/non_peft/test_mistral_non_peft.py
@@ -27,7 +27,7 @@ model, tokenizer = FastLanguageModel.from_pretrained(
 
 print("✅ Base model loaded successfully!")
 
-### Attemtping save merge
+### Attempting save merge
 
 
 print(f"\n{'='*80}")
diff --git a/tests/saving/non_peft/test_whisper_non_peft.py b/tests/saving/non_peft/test_whisper_non_peft.py
index 303d596c85..2d5676ec9c 100644
--- a/tests/saving/non_peft/test_whisper_non_peft.py
+++ b/tests/saving/non_peft/test_whisper_non_peft.py
@@ -27,7 +27,7 @@ model, tokenizer = FastModel.from_pretrained(
 
 print("✅ Base model loaded successfully!")
 
-### Attemtping save merge
+### Attempting save merge
 
 
 print(f"\n{'='*80}")
diff --git a/tests/saving/vision_models/test_index_file_sharded_model.py b/tests/saving/vision_models/test_index_file_sharded_model.py
index f737169841..8d107463e0 100644
--- a/tests/saving/vision_models/test_index_file_sharded_model.py
+++ b/tests/saving/vision_models/test_index_file_sharded_model.py
@@ -65,7 +65,7 @@ def format_data(sample):
 print("\n🔄 Formatting dataset for vision training...")
 system_message = "You are an expert french ocr system."
 # Convert dataset to OAI messages
-# need to use list comprehension to keep Pil.Image type, .mape convert image to bytes
+# need to use list comprehension to keep Pil.Image type, .map convert image to bytes
 train_dataset = [format_data(sample) for sample in train_dataset]
 eval_dataset = [format_data(sample) for sample in eval_dataset]
 print("✅ Dataset formatting completed!")
@@ -99,7 +99,7 @@ try:
         finetune_vision_layers = True,  # Turn off for just text!
         finetune_language_layers = True,  # Should leave on!
         finetune_attention_modules = True,  # Attention good for GRPO
-        finetune_mlp_modules = True,  # SHould leave on always!
+        finetune_mlp_modules = True,  # Should leave on always!
         r = 16,  # Choose any number > 0 ! Suggested 8, 16, 32, 64, 128
         lora_alpha = 32,
         lora_dropout = 0,  # Supports any, but = 0 is optimized
diff --git a/tests/saving/vision_models/test_push_to_hub_merged.py b/tests/saving/vision_models/test_push_to_hub_merged.py
index 74fa058988..fb2af4b4fe 100644
--- a/tests/saving/vision_models/test_push_to_hub_merged.py
+++ b/tests/saving/vision_models/test_push_to_hub_merged.py
@@ -66,7 +66,7 @@ def format_data(sample):
 print("\n🔄 Formatting dataset for vision training...")
 system_message = "You are an expert french ocr system."
 # Convert dataset to OAI messages
-# need to use list comprehension to keep Pil.Image type, .mape convert image to bytes
+# need to use list comprehension to keep Pil.Image type, .map convert image to bytes
 train_dataset = [format_data(sample) for sample in train_dataset]
 eval_dataset = [format_data(sample) for sample in eval_dataset]
 print("✅ Dataset formatting completed!")
@@ -100,7 +100,7 @@ try:
         finetune_vision_layers = True,  # Turn off for just text!
         finetune_language_layers = True,  # Should leave on!
         finetune_attention_modules = True,  # Attention good for GRPO
-        finetune_mlp_modules = True,  # SHould leave on always!
+        finetune_mlp_modules = True,  # Should leave on always!
         r = 16,  # Choose any number > 0 ! Suggested 8, 16, 32, 64, 128
         lora_alpha = 32,
         lora_dropout = 0,  # Supports any, but = 0 is optimized
diff --git a/tests/saving/vision_models/test_save_merge_qwen2.5vl32B_model_ocr_benchmark.py b/tests/saving/vision_models/test_save_merge_qwen2.5vl32B_model_ocr_benchmark.py
index ebe078c73b..2b24bc4a32 100644
--- a/tests/saving/vision_models/test_save_merge_qwen2.5vl32B_model_ocr_benchmark.py
+++ b/tests/saving/vision_models/test_save_merge_qwen2.5vl32B_model_ocr_benchmark.py
@@ -61,7 +61,7 @@ def format_data(sample):
 
 system_message = "You are an expert french ocr system."
 # Convert dataset to OAI messages
-# need to use list comprehension to keep Pil.Image type, .mape convert image to bytes
+# need to use list comprehension to keep Pil.Image type, .map convert image to bytes
 train_dataset = [format_data(sample) for sample in train_dataset]
 eval_dataset = [format_data(sample) for sample in eval_dataset]
 
@@ -102,7 +102,7 @@ model = FastVisionModel.get_peft_model(
     finetune_vision_layers = True,  # Turn off for just text!
     finetune_language_layers = True,  # Should leave on!
     finetune_attention_modules = True,  # Attention good for GRPO
-    finetune_mlp_modules = True,  # SHould leave on always!
+    finetune_mlp_modules = True,  # Should leave on always!
     r = 16,  # Choose any number > 0 ! Suggested 8, 16, 32, 64, 128
     # target_modules = ["q_proj", "k_proj", "v_proj", "o_proj",
     # "gate_proj", "up_proj", "down_proj",],
diff --git a/tests/saving/vision_models/test_save_merge_vision_model_ocr_benchmark.py b/tests/saving/vision_models/test_save_merge_vision_model_ocr_benchmark.py
index b99785bcb1..16914707c2 100644
--- a/tests/saving/vision_models/test_save_merge_vision_model_ocr_benchmark.py
+++ b/tests/saving/vision_models/test_save_merge_vision_model_ocr_benchmark.py
@@ -61,7 +61,7 @@ def format_data(sample):
 
 system_message = "You are an expert french ocr system."
 # Convert dataset to OAI messages
-# need to use list comprehension to keep Pil.Image type, .mape convert image to bytes
+# need to use list comprehension to keep Pil.Image type, .map convert image to bytes
 train_dataset = [format_data(sample) for sample in train_dataset]
 eval_dataset = [format_data(sample) for sample in eval_dataset]
 
@@ -102,7 +102,7 @@ model = FastVisionModel.get_peft_model(
     finetune_vision_layers = True,  # Turn off for just text!
     finetune_language_layers = True,  # Should leave on!
     finetune_attention_modules = True,  # Attention good for GRPO
-    finetune_mlp_modules = True,  # SHould leave on always!
+    finetune_mlp_modules = True,  # Should leave on always!
     r = 16,  # Choose any number > 0 ! Suggested 8, 16, 32, 64, 128
     # target_modules = ["q_proj", "k_proj", "v_proj", "o_proj",
     # "gate_proj", "up_proj", "down_proj",],

From 0326577b821603878ae9d85b67d52eb09bf7ae46 Mon Sep 17 00:00:00 2001
From: Etherll <61019402+Etherll@users.noreply.github.com>
Date: Fri, 24 Apr 2026 15:59:17 +0300
Subject: [PATCH 164/216] fix: guard resolve_model_class fallback against
 unresolvable transformers AutoModel entries (#5155)

* fix: avoid PerceptionEncoder ImportError blocking trust_remote_code model loads

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

* Update config class retrieval in _utils.py

Refactor config class retrieval logic to use model mapping.

* resolve_model_class: restore _extra_content fallback

The previous fallback iterated mapping.items(), which transformers'
_LazyAutoMapping defines as _model_mapping entries + _extra_content
entries. The PR's per-key loop covers only _model_mapping, so
subclasses of configs registered via AutoModel.register(cfg, model)
silently resolve to None. Add a safe isinstance pass over
_extra_content (no lazy loads, no crash risk) before giving up.

* Add tests for resolve_model_class fallback

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

---------

Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com>
Co-authored-by: Daniel Han <danielhanchen@gmail.com>
Co-authored-by: Lee Jackson <130007945+Imagineer99@users.noreply.github.com>
---
 tests/test_resolve_model_class.py | 137 ++++++++++++++++++++++++++++++
 unsloth/models/_utils.py          |  30 +++++--
 2 files changed, 161 insertions(+), 6 deletions(-)
 create mode 100644 tests/test_resolve_model_class.py

diff --git a/tests/test_resolve_model_class.py b/tests/test_resolve_model_class.py
new file mode 100644
index 0000000000..5ec2a21a07
--- /dev/null
+++ b/tests/test_resolve_model_class.py
@@ -0,0 +1,137 @@
+from unsloth.models._utils import resolve_model_class
+
+
+class _AutoModelLike:
+    def __init__(self, mapping):
+        self._model_mapping = mapping
+
+
+class _FakeLazyMapping:
+    def __init__(self, entries, extra_content = None, broken_keys = ()):
+        self._entries = dict(entries)
+        self._config_mapping = {k: f"Cfg_{k}" for k in self._entries}
+        self._model_mapping = {k: f"Mdl_{k}" for k in self._entries}
+        self._extra_content = dict(extra_content or {})
+        self._broken_keys = set(broken_keys)
+
+    def __getitem__(self, key):
+        if key in self._extra_content:
+            return self._extra_content[key]
+        for k, (cfg_cls, mdl_cls) in self._entries.items():
+            if k in self._broken_keys:
+                raise ValueError(f"broken entry {k}")
+            if cfg_cls is key:
+                return mdl_cls
+        raise KeyError(key)
+
+    def _load_attr_from_module(self, key, attr):
+        if key in self._broken_keys:
+            raise ValueError(f"broken entry {key}")
+        cfg_cls, mdl_cls = self._entries[key]
+        if attr == self._config_mapping[key]:
+            return cfg_cls
+        if attr == self._model_mapping[key]:
+            return mdl_cls
+        raise KeyError(attr)
+
+
+class CfgA:
+    pass
+
+
+class CfgB:
+    pass
+
+
+class CfgBChild(CfgB):
+    pass
+
+
+class ModelA:
+    pass
+
+
+class ModelB:
+    pass
+
+
+class RegBase:
+    pass
+
+
+class RegChild(RegBase):
+    pass
+
+
+class RegModel:
+    pass
+
+
+class UnknownCfg:
+    pass
+
+
+def test_fast_path_exact_match():
+    m = _FakeLazyMapping({"a": (CfgA, ModelA), "b": (CfgB, ModelB)})
+    am = _AutoModelLike(m)
+    assert resolve_model_class(am, CfgA()) is ModelA
+
+
+def test_fallback_subclass_match_via_lazy_mapping():
+    m = _FakeLazyMapping({"a": (CfgA, ModelA), "b": (CfgB, ModelB)})
+    am = _AutoModelLike(m)
+    assert resolve_model_class(am, CfgBChild()) is ModelB
+
+
+def test_broken_lazy_entry_does_not_crash():
+    m = _FakeLazyMapping(
+        {"broken": (CfgA, ModelA), "b": (CfgB, ModelB)},
+        broken_keys = ("broken",),
+    )
+    am = _AutoModelLike(m)
+    assert resolve_model_class(am, CfgBChild()) is ModelB
+
+
+def test_unknown_config_returns_none():
+    m = _FakeLazyMapping({"a": (CfgA, ModelA)})
+    am = _AutoModelLike(m)
+    assert resolve_model_class(am, UnknownCfg()) is None
+
+
+def test_extra_content_subclass_fallback():
+    m = _FakeLazyMapping(
+        {"a": (CfgA, ModelA)},
+        extra_content = {RegBase: RegModel},
+    )
+    am = _AutoModelLike(m)
+    assert resolve_model_class(am, RegChild()) is RegModel
+
+
+def test_extra_content_exact_match_fast_path():
+    m = _FakeLazyMapping(
+        {"a": (CfgA, ModelA)},
+        extra_content = {RegBase: RegModel},
+    )
+    am = _AutoModelLike(m)
+    assert resolve_model_class(am, RegBase()) is RegModel
+
+
+def test_broken_entry_with_extra_content_subclass():
+    m = _FakeLazyMapping(
+        {"broken": (CfgA, ModelA)},
+        extra_content = {RegBase: RegModel},
+        broken_keys = ("broken",),
+    )
+    am = _AutoModelLike(m)
+    assert resolve_model_class(am, RegChild()) is RegModel
+
+
+def test_plain_dict_mapping_is_not_required():
+    am = _AutoModelLike({CfgA: ModelA})
+    assert resolve_model_class(am, CfgA()) is ModelA
+
+
+def test_tuple_result_unwrapped():
+    m = _FakeLazyMapping({"a": (CfgA, (ModelA, "extra"))})
+    am = _AutoModelLike(m)
+    assert resolve_model_class(am, CfgA()) is ModelA
diff --git a/unsloth/models/_utils.py b/unsloth/models/_utils.py
index 8b8dec3d3e..e87c2a1a83 100644
--- a/unsloth/models/_utils.py
+++ b/unsloth/models/_utils.py
@@ -415,13 +415,31 @@ def resolve_model_class(auto_model, config):
     try:
         result = mapping[config.__class__]
     except Exception:
-        for config_class, model_class in mapping.items():
-            if isinstance(config, config_class):
-                result = model_class
-                break
-        else:
+        result = None
+        for key in list(getattr(mapping, "_model_mapping", {})):
+            try:
+                config_class = mapping._load_attr_from_module(
+                    key, mapping._config_mapping[key]
+                )
+                if isinstance(config, config_class):
+                    result = mapping._load_attr_from_module(
+                        key, mapping._model_mapping[key]
+                    )
+                    break
+            except Exception:
+                continue
+        if result is None:
+            for extra_cls, extra_model in getattr(
+                mapping, "_extra_content", {}
+            ).items():
+                try:
+                    if isinstance(config, extra_cls):
+                        result = extra_model
+                        break
+                except Exception:
+                    continue
+        if result is None:
             return None
-
     return result[0] if isinstance(result, (list, tuple)) else result
 
 

From c2dc2eb1b128747e8f522d271de16586a07a0837 Mon Sep 17 00:00:00 2001
From: Daniel Han <danielhanchen@gmail.com>
Date: Fri, 24 Apr 2026 09:05:31 -0700
Subject: [PATCH 165/216] Studio: kill in-flight llama-server before spawning a
 new one (#5171)

* Studio: kill in-flight llama-server before spawning a new one

Two rapid Apply clicks in the chat settings panel can race two
load_model calls. Both pass the Phase 1 _kill_process (because neither
has stored its Popen handle yet), both download / read metadata, and
both reach Phase 3 and spawn a server. Only the last reference is
tracked in self._process. The first server becomes an orphan that
holds the model in RAM until the kernel OOM kicks in. Addresses #5161.

Two complementary changes in studio/backend/core/inference/llama_cpp.py:

1. At load_model entry, set the existing _cancel_event so any in-flight
   load aborts at its next checkpoint, then bind a fresh Event for
   the new load. Subsequent _kill_process and download phases pick up
   the new event.
2. Inside the Phase 3 lock, immediately before subprocess.Popen, run a
   defensive _kill_process that removes any orphan handle a racing
   load might have stored after the first kill ran.

Speculative decoding cleanup (also touched while in this code path):

The chat UI used to send "ngram-mod" as the wire value when the
speculative dropdown was On, and the backend mapped that to the
4-flag combo --spec-type ngram-mod --spec-ngram-size-n 24 --draft-min
48 --draft-max 64. Switch the wire value to "default" and have the
backend pass the single llama-server flag --spec-default. That flag
expands to the exact same params (see common/arg.cpp:3905-3914 in
llama.cpp). Default-on for non-vision models is preserved.

Verified:
- Unit test: planted Popen orphan handle is terminated before
  self._process is overwritten.
- All ten spec-cmd mappings produce the expected llama-server args
  ("default" -> --spec-default, "off" / null -> no flag, vision ->
  always disabled, manual "ngram-mod" / "ngram-simple" still work).

* Address review feedback on PR #5171

The previous attempt at cancelling in-flight loads via
``self._cancel_event.set()`` followed by
``self._cancel_event = threading.Event()`` was broken in two ways
(flagged independently by Gemini and Codex):

1. Sub-methods like _download_gguf consult ``self._cancel_event``
   on every check. After the rebind, the in-flight thread reads the
   FRESH unset Event, not the one we just set, so cancellation never
   propagates.

2. Worse, if unload_model() lands between ``set()`` and the rebind,
   unload's signal hits the OLD event and is then immediately
   discarded when load_model swaps in a fresh Event. The user's
   stop-request silently no-ops.

Revert to the original ``self._cancel_event.clear()``. The Phase 3
defensive ``_kill_process()`` introduced in this branch still closes
the orphan-process race that #5161 reports: even if two concurrent
loads both pass Phase 1 with self._process == None, the loser's
Phase 3 kill terminates the winner's Popen handle before overwriting
it, so we end up with exactly one llama-server process.

* Studio: coerce legacy speculative-type values for the simplified dropdown

The Speculative Decoding control was simplified to On (default) / Off,
but the backend still accepts and reports the older manual modes
(ngram-mod, ngram-simple). When a load response or status refresh comes
back with one of those values -- whether from an external API caller, a
model loaded before this PR landed, or a not-yet-upgraded backend -- the
controlled Select renders with an empty trigger because the value is not
in the SelectItem list.

Add a tiny normaliser at both entry points (status refresh + post-load)
so legacy manual modes coerce to "default". The user sees "On" instead
of a blank dropdown, and reapplying lets llama.cpp pick its own preferred
strategy via --spec-default.

Reviewer-flagged finding on PR #5171.
---
 studio/backend/core/inference/llama_cpp.py    | 26 +++++++++++++++----
 .../src/features/chat/chat-settings-sheet.tsx |  2 +-
 .../chat/hooks/use-chat-model-runtime.ts      | 17 ++++++++++--
 .../chat/stores/chat-runtime-store.ts         |  4 +--
 4 files changed, 39 insertions(+), 10 deletions(-)

diff --git a/studio/backend/core/inference/llama_cpp.py b/studio/backend/core/inference/llama_cpp.py
index c320f03b2c..800d7fdd8c 100644
--- a/studio/backend/core/inference/llama_cpp.py
+++ b/studio/backend/core/inference/llama_cpp.py
@@ -1574,11 +1574,21 @@ class LlamaCppBackend:
             # ref: https://github.com/ggml-org/llama.cpp/blob/master/docs/speculative.md
             # ref: https://github.com/ggml-org/llama.cpp/pull/19164
             # ref: https://github.com/ggml-org/llama.cpp/pull/18471
+            # ``"default"`` -> let llama-server pick a sensible spec
+            # config via ``--spec-default``. Explicit type names are
+            # passed through with the manual draft tuning we've shipped
+            # historically so power users keep their overrides.
             _valid_spec_types = {"ngram-simple", "ngram-mod"}
-            if speculative_type and speculative_type in _valid_spec_types:
-                if not is_vision:  # spec decoding disabled for vision models
-                    cmd.extend(["--spec-type", speculative_type])
-                    if speculative_type == "ngram-mod":
+            normalized_spec = (
+                speculative_type.lower().strip() if speculative_type else None
+            )
+            if normalized_spec and normalized_spec != "off" and not is_vision:
+                if normalized_spec == "default":
+                    cmd.append("--spec-default")
+                    self._speculative_type = "default"
+                elif normalized_spec in _valid_spec_types:
+                    cmd.extend(["--spec-type", normalized_spec])
+                    if normalized_spec == "ngram-mod":
                         cmd.extend(
                             [
                                 "--spec-ngram-size-n",
@@ -1589,7 +1599,7 @@ class LlamaCppBackend:
                                 "64",
                             ]
                         )
-                    self._speculative_type = speculative_type
+                    self._speculative_type = normalized_spec
                 else:
                     self._speculative_type = None
             else:
@@ -1750,6 +1760,12 @@ class LlamaCppBackend:
             if gpu_indices is not None:
                 env["CUDA_VISIBLE_DEVICES"] = ",".join(str(i) for i in gpu_indices)
 
+            # Defensive kill: if a concurrent load slipped past Phase 1
+            # (because its `self._process` was None at the time) and
+            # already stored a Popen handle here, drop that orphan
+            # before we overwrite the reference. See issue #5161.
+            self._kill_process()
+
             self._stdout_lines = []
             self._process = subprocess.Popen(
                 cmd,
diff --git a/studio/frontend/src/features/chat/chat-settings-sheet.tsx b/studio/frontend/src/features/chat/chat-settings-sheet.tsx
index e4574dab74..fc5f097969 100644
--- a/studio/frontend/src/features/chat/chat-settings-sheet.tsx
+++ b/studio/frontend/src/features/chat/chat-settings-sheet.tsx
@@ -1037,7 +1037,7 @@ export function ChatSettingsPanel({
                           <SelectValue />
                         </SelectTrigger>
                         <SelectContent>
-                          <SelectItem value="ngram-mod">On</SelectItem>
+                          <SelectItem value="default">On</SelectItem>
                           <SelectItem value="off">Off</SelectItem>
                         </SelectContent>
                       </Select>
diff --git a/studio/frontend/src/features/chat/hooks/use-chat-model-runtime.ts b/studio/frontend/src/features/chat/hooks/use-chat-model-runtime.ts
index 1281592168..e90cab73fa 100644
--- a/studio/frontend/src/features/chat/hooks/use-chat-model-runtime.ts
+++ b/studio/frontend/src/features/chat/hooks/use-chat-model-runtime.ts
@@ -24,6 +24,19 @@ import type {
   InferenceParams,
 } from "../types/runtime";
 
+// The simplified Speculative Decoding control surfaces "default" (which
+// maps to llama.cpp's --spec-default) and "off". A backend status / load
+// response can still report the older manual modes (ngram-mod,
+// ngram-simple) when a model is loaded via the API or carried over from an
+// older Studio version. The Select would render an empty trigger for those
+// values, so coerce them to "default" -- llama.cpp's own --spec-default
+// picks an equivalent strategy and keeps the dropdown coherent.
+function normalizeSpeculativeType(v: string | null | undefined): string | null {
+  if (v == null) return null;
+  if (v === "default" || v === "off") return v;
+  return "default";
+}
+
 type SelectedModelInput = {
   id: string;
   isLora?: boolean;
@@ -279,7 +292,7 @@ export function useChatModelRuntime() {
         const ggufNativeContextLength = statusRes.is_gguf
           ? (statusRes.native_context_length ?? null)
           : null;
-        const currentSpecType = statusRes.speculative_type ?? null;
+        const currentSpecType = normalizeSpeculativeType(statusRes.speculative_type);
         useChatRuntimeStore.setState({
           supportsReasoning,
           reasoningAlwaysOn,
@@ -492,7 +505,7 @@ export function useChatModelRuntime() {
               }
             }
             const loadedKv = loadResponse.cache_type_kv ?? null;
-            const loadedSpec = loadResponse.speculative_type ?? null;
+            const loadedSpec = normalizeSpeculativeType(loadResponse.speculative_type);
             const nativeCtx = loadResponse.is_gguf
               ? (loadResponse.context_length ?? 131072)
               : null;
diff --git a/studio/frontend/src/features/chat/stores/chat-runtime-store.ts b/studio/frontend/src/features/chat/stores/chat-runtime-store.ts
index 2a1249aea0..ce69d8f6dc 100644
--- a/studio/frontend/src/features/chat/stores/chat-runtime-store.ts
+++ b/studio/frontend/src/features/chat/stores/chat-runtime-store.ts
@@ -270,7 +270,7 @@ export const useChatRuntimeStore = create<ChatRuntimeStore>((set) => ({
   toolCallTimeout: loadInt(TOOL_CALL_TIMEOUT_KEY, 5),
   kvCacheDtype: null,
   loadedKvCacheDtype: null,
-  speculativeType: "ngram-mod",
+  speculativeType: "default",
   loadedSpeculativeType: null,
   customContextLength: null,
   defaultChatTemplate: null,
@@ -365,7 +365,7 @@ export const useChatRuntimeStore = create<ChatRuntimeStore>((set) => ({
       toolStatus: null,
       kvCacheDtype: null,
       loadedKvCacheDtype: null,
-      speculativeType: "ngram-mod",
+      speculativeType: "default",
       loadedSpeculativeType: null,
       customContextLength: null,
       defaultChatTemplate: null,

From ae9de7f2df80b592af02d0d610627a4c21663b35 Mon Sep 17 00:00:00 2001
From: Daniel Han <danielhanchen@gmail.com>
Date: Fri, 24 Apr 2026 09:06:01 -0700
Subject: [PATCH 166/216] Studio: stop currency escape from breaking inline
 LaTeX (#5170)

* Studio: stop currency escape from breaking inline LaTeX

The currency-escape preprocessor in studio/frontend/src/lib/latex.ts
matched the opening dollar of any $<digits>...$ span and inserted a
backslash. The result was that text like "$30^\circ$" or
"**$90 - x$**" rendered as raw characters with stray dollar signs.
Fixes #5164.

Add two helpers in front of the escape:
- hasInlineMathCloser looks for an unescaped, non-doubled closing
  dollar within the same line. Bold-wrapped spans (**$X$**) are always
  treated as math since LLMs use that form for bold math.
- looksLikeMathBody filters multi-token bodies that look like prose
  between two currency tokens ($5 to $10, $5, $10).

Verified against 111 inputs: the issue body, common LaTeX patterns
(Greek vars, fractions, integrals, vectors, exponents), prose currency
in lists and sentences, code blocks, and headings. All pass.

* Address review feedback on PR #5170

- Drop ^ and _ from MATH_OP_RE since LATEX_CHAR_RE already short-
  circuits on those before MATH_OP_RE is consulted (Gemini comment).

- Treat compact currency ranges like $5-$10 and $5/$10 as currency
  rather than math. The body between the first two dollars in those
  forms is "5-" or "5/", a single non-whitespace token that previously
  hit the math shortcut. Extend TRAIL_PUNCT_RE to strip - and / so the
  trimmed body comes back as pure currency. (Codex comment.)

- Honour __underscore-bold__ around math the same way as **-bold**.
  Markdown allows both delimiters and LLMs do reach for the underscore
  form. (Gemini comment.)

Verified against the existing 18 cases plus 5 new ones for the range,
slash, and underscore-bold scenarios. All pass.

* Studio: fix numeric inline math + currency-as-closer in LaTeX preprocess

Two reviewer-flagged real-world misses in the inline-math heuristic.

1) Numeric-only operator forms like $2 + 2$, $100 < 200$, $1,000 - 500$
   were getting their leading $ escaped, so the renderer never saw them
   as math. The body has a math op but no lone-letter variable, so the
   old looksLikeMathBody required the lone-letter clause and rejected
   purely numeric expressions. Add SIMPLE_MATH_RE to recognise number-
   or-letter operands joined by math operators.

2) Prose like "Starts at $5 + a $10 add-on" was being treated as one
   math span "5 + a " with the second currency token mistaken for the
   closer. The body satisfied the math-op + lone-letter check, so the
   span got accepted and the renderer ate "10 add-on". In hasInlineMathCloser,
   reject any candidate $ whose next character is a digit -- that's almost
   always another currency token starting, not the closer of a real math
   span (math doesn't follow $ with a bare digit).

Verified via temp/pr_simulation/sim_5170_latex.mjs: 25/25 cases pass,
including the 6 reviewer numeric-math cases, 3 currency-as-closer cases,
and 16 regression checks against the originally shipped behavior.
---
 studio/frontend/src/lib/latex.ts | 117 +++++++++++++++++++++++++++++++
 1 file changed, 117 insertions(+)

diff --git a/studio/frontend/src/lib/latex.ts b/studio/frontend/src/lib/latex.ts
index 954d0f7bc6..0092d51be0 100644
--- a/studio/frontend/src/lib/latex.ts
+++ b/studio/frontend/src/lib/latex.ts
@@ -74,12 +74,126 @@ function isInCodeBlock(
   return false;
 }
 
+/**
+ * A token (no whitespace) that looks purely like currency, e.g. `5`,
+ * `1,000`, `5.99`, `100K`, `3.5M`.
+ */
+const CURRENCY_BODY_RE = /^\d+(?:,\d{3})*(?:\.\d+)?[KMBkmb]?$/;
+
+/** Body characters that almost always indicate real LaTeX. */
+const LATEX_CHAR_RE = /[\\^_{}]/;
+
+/**
+ * Operators that strongly suggest a math expression. We deliberately
+ * leave out `^` and `_` because `LATEX_CHAR_RE` already short-circuits
+ * on those before we ever consult this regex.
+ */
+const MATH_OP_RE = /[=+\-<>/*]/;
+
+/**
+ * Trailing chars stripped before the currency check. Includes prose
+ * punctuation plus the connectors `-` and `/` that appear in compact
+ * currency ranges like `$5-$10` or `$5/$10`; without them the body
+ * `5-` or `5/` would slip through the single-token math shortcut.
+ */
+const TRAIL_PUNCT_RE = /[.,;:!?\-/]+$/;
+
+/**
+ * A standalone single letter (variable name) inside the body. We require
+ * that the letter is not part of a longer word so that prose like
+ * "5 to attend" doesn't get misread as a math expression with the
+ * variable `t`.
+ */
+const LONE_LETTER_RE = /(?<![a-zA-Z])[a-zA-Z](?![a-zA-Z])/;
+
+/**
+ * Numeric or single-letter operands joined by math operators, with
+ * optional whitespace. Matches `2 + 2`, `100 < 200`, `1,000 - 500`,
+ * `2.5 * 3`, `x + y`. Lets us recognise numeric-only expressions like
+ * `$2 + 2$` as math without forcing a lone variable token.
+ */
+const SIMPLE_MATH_RE =
+  /^(?:\d+(?:,\d{3})*(?:\.\d+)?|[a-zA-Z])(?:\s*[=+\-<>/*]\s*(?:\d+(?:,\d{3})*(?:\.\d+)?|[a-zA-Z]))+$/;
+
+/**
+ * Return true if the substring between two `$` delimiters looks like a
+ * LaTeX expression rather than a span of prose between two currency
+ * tokens.
+ *
+ * Rule of thumb:
+ *   - `$30^\circ$`  -> math (LaTeX chars)
+ *   - `$x$`         -> math (single non-currency token)
+ *   - `$90 - x$`    -> math (math op + lone variable)
+ *   - `$5 to $10`   -> NOT math (multi-token prose, no math op)
+ *   - `$5, $10`     -> NOT math (currency-like token + trailing punct)
+ *   - `$1,000$`     -> NOT math (single currency-like token)
+ */
+function looksLikeMathBody(body: string): boolean {
+  if (LATEX_CHAR_RE.test(body)) return true;
+  const trimmed = body.trim().replace(TRAIL_PUNCT_RE, "");
+  if (!trimmed) return false;
+  if (CURRENCY_BODY_RE.test(trimmed)) return false;
+  // Numeric-only operator forms: `2 + 2`, `100 < 200`, `1,000 - 500`.
+  // Recognised without requiring a lone-variable letter.
+  if (SIMPLE_MATH_RE.test(trimmed)) return true;
+  if (!/\s/.test(trimmed)) return true;
+  if (!MATH_OP_RE.test(trimmed)) return false;
+  return LONE_LETTER_RE.test(trimmed);
+}
+
+/**
+ * Return true if the `$` at `offset` opens a balanced inline math span
+ * (`$...$`) on the same line. The closer must be unescaped, must not be
+ * part of `$$`, and must lie inside a 200-character window. The body
+ * must look like LaTeX so we don't pair two currency tokens that share
+ * a line (e.g. "$5 to $10"). Bold-wrapped spans (`**$X$**` and the
+ * underscore equivalent `__$X$__`) are always treated as math because
+ * LLMs reach for that pattern when they want "bold math" and the
+ * heuristic would otherwise reject prose-shaped bodies like "90 - x".
+ */
+function hasInlineMathCloser(content: string, offset: number): boolean {
+  const MAX_SPAN = 200;
+  const limit = Math.min(content.length, offset + 1 + MAX_SPAN);
+  for (let i = offset + 1; i < limit; i++) {
+    const c = content[i];
+    if (c === "\n") return false;
+    if (c !== "$") continue;
+    if (content[i - 1] === "\\") continue;
+    if (content[i + 1] === "$") {
+      i++;
+      continue;
+    }
+    // A `$` immediately followed by a digit is far more likely the start
+    // of another currency token than the closer for the current span. Keep
+    // scanning so prose like `Starts at $5 + a $10 add-on` doesn't pair
+    // the two currency markers as a math span.
+    if (/\d/.test(content[i + 1] ?? "")) {
+      continue;
+    }
+    if (offset >= 2) {
+      const op = content[offset - 1];
+      if (
+        (op === "*" || op === "_") &&
+        content[offset - 2] === op &&
+        content[i + 1] === op &&
+        content[i + 2] === op
+      ) {
+        return true;
+      }
+    }
+    return looksLikeMathBody(content.slice(offset + 1, i));
+  }
+  return false;
+}
+
 /**
  * Preprocess a markdown string to escape currency dollar signs so they are not
  * parsed as LaTeX math delimiters.
  *
  * - `$5` alone becomes `\$5` (currency, not math)
  * - `$\alpha$` is untouched (real LaTeX)
+ * - `$30^\circ$` is untouched (LaTeX whose body starts with a digit)
+ * - `**$30^\circ$**` is untouched (LaTeX wrapped in bold)
  * - `$$E = mc^2$$` is untouched (display math)
  * - Currency inside code blocks/spans is untouched
  */
@@ -92,6 +206,9 @@ export function preprocessLaTeX(content: string): string {
     if (isInCodeBlock(offset, codeRegions)) {
       return match;
     }
+    if (hasInlineMathCloser(content, offset)) {
+      return match;
+    }
     return "\\" + match;
   });
 }

From 8264e80dd952df0fc05c85c6f9671ab7d9991638 Mon Sep 17 00:00:00 2001
From: Daniel Han <danielhanchen@gmail.com>
Date: Fri, 24 Apr 2026 10:00:42 -0700
Subject: [PATCH 167/216] Studio: probe AMD GPUs in llama-server VRAM detection
 (#5172)

* Studio: probe AMD GPUs in llama-server VRAM detection

_get_gpu_free_memory in studio/backend/core/inference/llama_cpp.py
only queried nvidia-smi. On AMD ROCm hosts that returns nothing, so
the GPU list is empty, the auto-fit logic falls into the no-gpus
branch, and llama-server gets --fit on with no -ngl to anchor it.
The model loads on CPU even though the GPU is detected elsewhere in
Studio. Addresses #5106.

Add a torch-based fallback that runs after nvidia-smi fails or returns
empty:

    import torch
    if torch.cuda.is_available() and hasattr(torch.cuda, "mem_get_info"):
        for ordinal in range(torch.cuda.device_count()):
            free, _total = torch.cuda.mem_get_info(ordinal)
            gpus.append((ordinal, free // (1024 * 1024)))

Works on AMD because the ROCm torch wheels Studio installs reuse the
entire torch.cuda.* namespace via HIP. Also rescues NVIDIA hosts
where nvidia-smi is missing from PATH (a secondary cause of the bug
on Windows). Matches the convention
studio/backend/utils/hardware/hardware.py:412 already uses for the
same fallback purpose.

Verified locally: nvidia-smi path returns the expected GPU and free
MiB; torch fallback returns valid VRAM when nvidia-smi is forced to
fail. Note: PR #4874 is a draft taking a different approach
(parsing vulkaninfo); the two are complementary.

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

* Address review feedback on PR #5172

torch.cuda.device_count() enumerates GPUs RELATIVE to the current
CUDA_VISIBLE_DEVICES (or HIP_VISIBLE_DEVICES on ROCm). Returning
those visible ordinals directly lets _select_gpus rewrite
CUDA_VISIBLE_DEVICES with the wrong physical IDs: a process started
with CUDA_VISIBLE_DEVICES=2,3 would get its child llama-server
relaunched with CUDA_VISIBLE_DEVICES=0,1, targeting the wrong GPUs
and violating any scheduler pinning.

Translate visible ordinals back through the active CVD/HIP/ROCR
mask before returning. Falls through to bare ordinal when no mask
is set. Also drop the redundant int() cast on // -- bytes // 2**20
already returns int.

Verified: with CUDA_VISIBLE_DEVICES=6 and nvidia-smi forced to fail,
the torch fallback now returns (6, free_mib) instead of (0, free_mib).

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

* Studio: fix ROCm visibility precedence + narrow ROCm child env

Two reviewer-flagged correctness bugs in the AMD GPU probe path.

1) ROCm visibility precedence was reversed. torch.cuda enumerates GPUs
   relative to HIP_VISIBLE_DEVICES / ROCR_VISIBLE_DEVICES on ROCm builds,
   but the probe's env-var lookup checked CUDA_VISIBLE_DEVICES first. With
   CUDA_VISIBLE_DEVICES=0,1 and HIP_VISIBLE_DEVICES=6,7 the probe returned
   [(0, ...), (1, ...)] when torch's view was actually [(6, ...), (7, ...)].
   The wrong physical IDs flowed downstream into CUDA_VISIBLE_DEVICES for
   the llama-server subprocess, pinning it to GPUs 0,1 instead of 6,7.

   Fix: branch on torch.version.hip. On ROCm, prefer HIP > ROCR > CUDA
   (matches torch's own ordering). On NVIDIA, use CUDA only -- ignoring
   any HIP/ROCR vars the parent happens to have set.

2) Child env narrowing only set CUDA_VISIBLE_DEVICES. On ROCm, llama-server
   honors HIP/ROCR; if the parent shell exported HIP_VISIBLE_DEVICES=4,5
   and the selector picked just GPU 4, the child still saw both because
   we never narrowed HIP/ROCR. Now we set all three on ROCm so the AMD
   subprocess actually sees the planned subset.

Both branches verified via temp/pr_simulation/sim_5172_rocm_precedence.py
(7/7 cases pass), including the reviewer's verbatim R5 case
(CVD=0,1 + HIP/ROCR=6,7).

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

* Studio: sort GPU probe result + honor explicitly empty ROCm masks

Two reviewer-flagged correctness nits on top of eff55fb8.

1) Gemini medium: the torch fallback returned an unsorted list when the
   visibility mask was non-sequential (e.g. CUDA_VISIBLE_DEVICES=5,2,9),
   diverging from the docstring guarantee and the nvidia-smi path. Now
   sorted by physical id.

2) Codex P2: an explicitly empty HIP_VISIBLE_DEVICES="" should mean
   "no GPUs" per the codebase convention in
   utils/hardware/hardware.py::_get_parent_visible_gpu_spec. The previous
   `or` chain treated empty string as falsy and silently fell through to
   ROCR / CUDA, producing wrong physical IDs. Switch to `is not None`
   checks to match.

Verified via sim_5172_rocm_precedence.py (9/9 cases pass) including the
two new R8 (sort) and R9 (empty-HIP honored) cases.

* Studio: align nvidia-smi probe with torch fallback (sort + robust CVD)

Two follow-up Gemini-medium nits on PR #5172.

1) Fragile CVD parsing on the nvidia-smi path: `cvd.split(",")` would
   raise ValueError on a trailing comma like "0,1," because the empty
   trailing token is not skipped. The torch fallback already filters
   empty tokens via `if x.strip()`; mirror that here.

2) Missing sort guarantee on the nvidia-smi path: the docstring promises
   sort-by-id, the torch fallback now sorts, but the nvidia-smi path
   relied on driver enumeration order. Add an explicit sort.

Both changes match what shipped in 6b1cccd6 for the torch fallback, so
the two probe paths now have identical CVD parsing + ordering semantics.

* Studio: drop cvd.strip() truthiness so empty CVD filters all GPUs

Reviewer-flagged correctness bug. The previous `if cvd is not None and
cvd.strip():` guard treated `CUDA_VISIBLE_DEVICES=""` as if the variable
were unset, leaving `allowed=None` (and `physical_ids=None` on the torch
path). On the nvidia-smi path that mattered: nvidia-smi ignores CVD
entirely, so the probe's `allowed` filter is the only thing that
respects the parent's "no GPUs" intent. Pre-fix the probe returned every
physical GPU when the parent had explicitly hidden them.

Drop the `.strip()` truthiness check on both paths. The downstream
`if x.strip()` token filter still keeps trailing-comma masks like
"0,1," safe, and an empty mask now produces an empty allowed/physical
set as expected (matching utils/hardware/hardware.py convention).

Verified via sim_5172_rocm_precedence.py R10 + R11 (now 11/11 cases
pass): nvidia-smi path with `CUDA_VISIBLE_DEVICES=""` now returns []
instead of leaking the hidden GPUs.

* Studio: log ROCm env-var failures instead of silently swallowing

Reviewer-flagged defensive logging gap. The bare `except Exception: pass`
around the HIP/ROCR env-var assignment would mask anything from a
missing torch import to an unexpected version object shape. Log at
debug so a failed AMD child-env narrowing is at least traceable.
Behavior is unchanged: torch missing or version probe failing still
leaves the child with only CUDA_VISIBLE_DEVICES set.

---------

Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com>
---
 studio/backend/core/inference/llama_cpp.py | 147 +++++++++++++++++----
 1 file changed, 118 insertions(+), 29 deletions(-)

diff --git a/studio/backend/core/inference/llama_cpp.py b/studio/backend/core/inference/llama_cpp.py
index 800d7fdd8c..35ef90ee37 100644
--- a/studio/backend/core/inference/llama_cpp.py
+++ b/studio/backend/core/inference/llama_cpp.py
@@ -549,14 +549,24 @@ class LlamaCppBackend:
 
     @staticmethod
     def _get_gpu_free_memory() -> list[tuple[int, int]]:
-        """Query free memory per GPU via nvidia-smi.
+        """Query free memory per GPU.
 
-        Returns list of (gpu_index, free_mib) sorted by index.
-        Respects CUDA_VISIBLE_DEVICES if set.
-        Returns empty list if nvidia-smi is not available.
+        Order:
+          1. ``nvidia-smi`` (NVIDIA CUDA hosts) -- respects
+             ``CUDA_VISIBLE_DEVICES``.
+          2. ``torch.cuda.mem_get_info`` -- universal fallback that
+             works on AMD ROCm too because the HIP runtime
+             reuses the entire ``torch.cuda.*`` namespace. Covers the
+             AMD case for issue #5106 (nvidia-smi-only probe silently
+             returned [] on AMD hosts) and also rescues NVIDIA hosts
+             where ``nvidia-smi`` is missing from PATH.
+
+        Returns list of (gpu_index, free_mib) sorted by index. Empty
+        list if no supported GPU is reachable.
         """
         import os
 
+        # ── NVIDIA via nvidia-smi ────────────────────────────────────
         try:
             result = subprocess.run(
                 [
@@ -569,30 +579,95 @@ class LlamaCppBackend:
                 timeout = 10,
                 **_windows_hidden_subprocess_kwargs(),
             )
-            if result.returncode != 0:
-                return []
-
-            # Parse which GPUs are allowed by existing CUDA_VISIBLE_DEVICES
-            allowed = None
-            cvd = os.environ.get("CUDA_VISIBLE_DEVICES")
-            if cvd is not None and cvd.strip():
-                try:
-                    allowed = set(int(x.strip()) for x in cvd.split(","))
-                except ValueError:
-                    pass  # Non-numeric (e.g., "GPU-uuid"), ignore filter
-
-            gpus = []
-            for line in result.stdout.strip().splitlines():
-                parts = line.split(",")
-                if len(parts) == 2:
-                    idx = int(parts[0].strip())
-                    free_mib = int(parts[1].strip())
-                    if allowed is not None and idx not in allowed:
-                        continue
-                    gpus.append((idx, free_mib))
-            return gpus
+            if result.returncode == 0:
+                allowed: Optional[set[int]] = None
+                cvd = os.environ.get("CUDA_VISIBLE_DEVICES")
+                if cvd is not None:
+                    try:
+                        # `if x.strip()` filters trailing-comma masks like
+                        # "0,1," which would otherwise raise ValueError on
+                        # an empty token. An explicitly empty mask (CVD="")
+                        # yields an empty `allowed` set so all GPUs are
+                        # filtered out, matching the codebase convention.
+                        allowed = set(
+                            int(x.strip()) for x in cvd.split(",") if x.strip()
+                        )
+                    except ValueError:
+                        pass
+                gpus: list[tuple[int, int]] = []
+                for line in result.stdout.strip().splitlines():
+                    parts = line.split(",")
+                    if len(parts) == 2:
+                        idx = int(parts[0].strip())
+                        free_mib = int(parts[1].strip())
+                        if allowed is not None and idx not in allowed:
+                            continue
+                        gpus.append((idx, free_mib))
+                # Match the docstring's sort-by-id guarantee. nvidia-smi
+                # almost always returns sorted output, but driver order
+                # is not formally guaranteed.
+                gpus.sort(key = lambda g: g[0])
+                if gpus:
+                    return gpus
         except Exception as e:
-            logger.debug(f"Failed to query GPU free memory via nvidia-smi: {e}")
+            logger.debug(f"nvidia-smi probe failed: {e}")
+
+        # ── Torch fallback (covers AMD ROCm and missing nvidia-smi) ──
+        try:
+            import torch
+
+            if not hasattr(torch, "cuda") or not torch.cuda.is_available():
+                return []
+            if not hasattr(torch.cuda, "mem_get_info"):
+                return []
+            # torch.cuda enumerates GPUs RELATIVE to the visibility mask.
+            # On NVIDIA builds the mask is CUDA_VISIBLE_DEVICES; on AMD
+            # ROCm builds it is HIP_VISIBLE_DEVICES (or ROCR_VISIBLE_DEVICES
+            # if HIP is unset). Downstream we feed these IDs back into the
+            # llama-server subprocess as CVD, so we must translate visible
+            # ordinals back to physical indices first; otherwise launching
+            # with ``CUDA_VISIBLE_DEVICES=2,3`` would get rewritten to
+            # ``CUDA_VISIBLE_DEVICES=0,1`` and target the wrong GPUs.
+            physical_ids: Optional[list[int]] = None
+            # Match the codebase convention in
+            # ``utils/hardware/hardware.py::_get_parent_visible_gpu_spec``:
+            # treat an explicitly empty mask (``HIP_VISIBLE_DEVICES=""``)
+            # as "set to no GPUs" rather than falling through to the next
+            # var. ``or`` would coerce empty string to falsy and silently
+            # promote the wrong source.
+            if getattr(torch.version, "hip", None) is not None:
+                hip_v = os.environ.get("HIP_VISIBLE_DEVICES")
+                rocr_v = os.environ.get("ROCR_VISIBLE_DEVICES")
+                cvd = (
+                    hip_v
+                    if hip_v is not None
+                    else rocr_v
+                    if rocr_v is not None
+                    else os.environ.get("CUDA_VISIBLE_DEVICES")
+                )
+            else:
+                cvd = os.environ.get("CUDA_VISIBLE_DEVICES")
+            if cvd is not None:
+                try:
+                    # Empty mask (CVD="") yields an empty list so the
+                    # below loop produces no GPUs, consistent with the
+                    # nvidia-smi path and utils/hardware/hardware.py.
+                    physical_ids = [int(x.strip()) for x in cvd.split(",") if x.strip()]
+                except ValueError:
+                    physical_ids = None
+            gpus = []
+            for ordinal in range(torch.cuda.device_count()):
+                free_bytes, _total_bytes = torch.cuda.mem_get_info(ordinal)
+                idx = (
+                    physical_ids[ordinal]
+                    if physical_ids is not None and ordinal < len(physical_ids)
+                    else ordinal
+                )
+                gpus.append((idx, free_bytes // (1024 * 1024)))
+            # Match the nvidia-smi path's docstring guarantee of sorted-by-id.
+            return sorted(gpus, key = lambda g: g[0])
+        except Exception as e:
+            logger.debug(f"torch GPU probe failed: {e}")
             return []
 
     @staticmethod
@@ -1756,9 +1831,23 @@ class LlamaCppBackend:
                     f"{new_ld}:{existing_ld}" if existing_ld else new_ld
                 )
 
-            # Pin to selected GPU(s) via CUDA_VISIBLE_DEVICES
+            # Pin to selected GPU(s). On ROCm, llama-server (and any torch
+            # in the subprocess) honors HIP_VISIBLE_DEVICES / ROCR_VISIBLE_DEVICES;
+            # narrowing only CUDA_VISIBLE_DEVICES leaves an AMD child seeing
+            # the full HIP/ROCR set the parent inherited.
             if gpu_indices is not None:
-                env["CUDA_VISIBLE_DEVICES"] = ",".join(str(i) for i in gpu_indices)
+                pinned = ",".join(str(i) for i in gpu_indices)
+                env["CUDA_VISIBLE_DEVICES"] = pinned
+                try:
+                    import torch as _torch
+
+                    if getattr(_torch.version, "hip", None) is not None:
+                        env["HIP_VISIBLE_DEVICES"] = pinned
+                        env["ROCR_VISIBLE_DEVICES"] = pinned
+                except Exception as e:
+                    logger.debug(
+                        "Failed to set ROCm visibility env vars for child: %s", e
+                    )
 
             # Defensive kill: if a concurrent load slipped past Phase 1
             # (because its `self._process` was None at the time) and

From eb8b0dee2e7b06727360add18e2759ab27849f93 Mon Sep 17 00:00:00 2001
From: Daniel Han <danielhanchen@gmail.com>
Date: Fri, 24 Apr 2026 10:09:25 -0700
Subject: [PATCH 168/216] Studio: make stop button actually stop generation
 (#5069)

* Studio: make stop button actually stop generation

The UI stop button routes through assistant-ui's cancelRun, which aborts
the frontend fetch. Four issues combined to let llama-server keep decoding
long after the user clicked stop:

1. request.is_disconnected() does not fire reliably behind proxies
   (e.g. Colab) that don't propagate fetch aborts.
2. llama-server defaults n_predict to n_ctx when max_tokens is not sent,
   so a cancelled request keeps producing tokens up to 262144.
3. The httpx.Client pool keeps TCP keep-alive, so even a cleanly closed
   stream reuses the same connection and llama-server's liveness poll
   never sees a disconnect.
4. No explicit backend route to cancel - every cancel path relied on
   is_disconnected.

Changes:
- Add POST /api/inference/cancel keyed by session_id/completion_id, with
  a registry populated for the lifetime of each streaming response.
- Have the frontend (chat-adapter.ts) POST /inference/cancel on
  AbortController abort, alongside the existing fetch teardown.
- Send max_tokens=4096 + t_max_predict_ms=120000 as defaults on every
  outbound chat completion to llama-server; honoured by user overrides.
- Disable httpx keep-alive on the streaming client so connection close
  reaches llama-server and its 1s liveness check fires.

No behaviour changes for non-streaming paths or for existing callers
that already pass max_tokens/session_id.

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

* studio: harden stop-button cancel path and scope cancel route

- Require at least one identifier for /api/inference/cancel so a missing
  thread id cannot silently cancel every in-flight generation.
- Scope /cancel to a dedicated studio_router so it is not exposed under
  the /v1 OpenAI-compat prefix as a surprise endpoint.
- Store a set of cancel events per key in _CANCEL_REGISTRY so concurrent
  requests on the same session_id do not overwrite each other, and
  deduplicate in _cancel_by_keys so the cancelled count reflects unique
  requests.
- Always send session_id with chat completions (not only when tools are
  enabled) so non-tool GGUF streams register under it and are reachable
  from /cancel.
- Register the non-GGUF stream_chunks path in the cancel registry too,
  so transformers-based stop-button works behind proxies that swallow
  fetch aborts.
- Only apply the 2-minute t_max_predict_ms wall-clock cap when the
  caller did not pass max_tokens, so legitimate long generations on
  slow CPU/macOS/Windows supported installs are not silently truncated.
- Remove the abort listener on normal stream completion so reused
  AbortSignals cannot fire a spurious cancel POST after the fact.

* studio: close cancel-race and stale-cancel gaps in stop path

- Register the cancel tracker before returning StreamingResponse so a
  stop POST that arrives during prefill / warmup / proxy buffering
  finds an entry in _CANCEL_REGISTRY. Cleanup now runs via a Starlette
  BackgroundTask instead of a finally inside the async generator body.
- Add a per-run cancel_id on the frontend (crypto.randomUUID) and in
  ChatCompletionRequest so /api/inference/cancel matches one specific
  generation. Removes the stale-cancel bug where pressing stop then
  starting a new run in the same thread would cancel the retry.
- Apply t_max_predict_ms unconditionally in all three llama-server
  payload builders (previously gated on max_tokens=None, which made it
  dead code for UI callers that always send params.maxTokens). Raise
  the default to 10 minutes so slow CPU / macOS / Windows installs are
  not cut off mid-generation.
- Make _cancel_by_keys refuse empty input (return 0) so a future
  internal caller can not accidentally mass-cancel every in-flight
  request.
- Accept cancel_id (primary), session_id, and completion_id on the
  /api/inference/cancel route. Unify the three streaming sites on the
  same _cancel_keys / _tracker variable names.
- Annotate _CANCEL_REGISTRY as dict[str, set[threading.Event]].

* Add review tests for PR #5069

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

* studio: harden stop-button cancel semantics and wall-clock cap

- Make /inference/cancel match cancel_id EXCLUSIVELY when supplied.
  Previously the handler iterated ('cancel_id','session_id','completion_id')
  and unioned matches, so a stale cancel POST carrying {cancel_id:old,
  session_id:thr} would still cancel a later run on the same thread via
  the shared session_id. cancel_id is now a per-run exclusive key;
  session_id / completion_id are only used as fallbacks when cancel_id
  is absent.

- Close the early-cancel race. If /inference/cancel lands before the
  streaming handler reaches _TrackedCancel.__enter__() (stop clicked
  during prefill / warmup / proxy buffering), the cancel was silently
  dropped. Stash unmatched cancel_ids in _PENDING_CANCELS with a 30 s
  TTL; _TrackedCancel.__enter__() now replays any matching pending
  cancel by set()-ing the event immediately after registration.

- Make t_max_predict_ms = _DEFAULT_T_MAX_PREDICT_MS conditional on
  max_tokens is None at all three llama-server payload sites. The cap
  is a safety net for callers who leave max_tokens unset (otherwise
  llama-server defaults n_predict to n_ctx, up to 262144). Callers who
  set an explicit max_tokens are already self-limiting and must not be
  silently truncated at 10 minutes on slow CPU / macOS / Windows
  legitimate long generations.

- Guard each StreamingResponse return with try/except BaseException so
  _tracker.__exit__ runs even if StreamingResponse construction or any
  preceding statement raises between _tracker.__enter__() and the
  BackgroundTask attachment. Prevents a registry leak on that narrow
  window.

* studio: close TOCTOU race and restore wall-clock backstop on UI path

- Close TOCTOU race in the pending-cancel mechanism. The previous fix
  split cancel_inference's (cancel_by_keys + remember_pending_cancel)
  and _TrackedCancel.__enter__'s (register + consume_pending) into
  four separate lock acquisitions. Under contention a cancel POST
  could acquire-then-release the lock, find the registry empty, and
  stash ONLY AFTER __enter__ had already registered and consumed an
  empty pending map -- silently dropping the cancel. Both call sites
  now do their work inside a single _CANCEL_LOCK critical section, via
  the new atomic helper _cancel_by_cancel_id_or_stash() and an
  inlined consume-pending step in __enter__. Reproduced the race under
  forced interleaving pre-fix; 0/2000 drops post-fix under parallel
  stress.

- Apply t_max_predict_ms UNCONDITIONALLY at all three llama-server
  payload sites. The previous iteration gated the cap on
  `max_tokens is None`, which turned out to be dead code on the
  primary Studio UI path: chat-adapter.ts sets
  maxTokens=loadResp.context_length after every model load, so every
  chat request carries an explicit max_tokens and the wall-clock
  safety net never fired. The cap's original purpose is to bound
  stuck decodes regardless of the token budget; it must always apply.

- Raise _DEFAULT_T_MAX_PREDICT_MS from 10 minutes to 1 hour. 10
  minutes was too aggressive for legitimate slow-CPU chat responses
  (a 4096-token reply at 2 tok/s takes ~34 min); 1 hour accommodates
  that and still catches genuine zombie decodes.

- Prune _PENDING_CANCELS inside _cancel_by_keys as well, so stashed
  entries expire proportionally to overall cancel traffic rather than
  only to cancel_id-specific POSTs.

* studio: trim verbose comments and docstrings in cancel path

* studio/llama_cpp: drop upstream PR hashes from benchmark comment

* Add review tests for Studio stop button

* Consolidate review tests for Studio stop button

* Align cancel-route test with exclusive cancel_id semantics

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

* studio: move cancel cleanup to generator finally; drop dead helper

- Move _tracker.__exit__ from Starlette BackgroundTask into each
  streaming generator's finally block. Starlette skips the background
  callback when stream_response raises (OSError / ClientDisconnect),
  which leaked _CANCEL_REGISTRY entries on abrupt disconnect.
- Check cancel_event.is_set() at the top of each GGUF while loop so a
  pending-replay cancel falls through to final_chunk + [DONE] instead
  of propagating GeneratorExit out of _stream_with_retry.
- Remove unused _remember_pending_cancel; _cancel_by_cancel_id_or_stash
  superseded it.

* Add review tests for Studio stop-button

* studio: wire audio-input stream into cancel registry

- Register cancel_event with _TrackedCancel on the audio-input streaming
  path so POST /api/inference/cancel can stop whisper / audio-input GGUF
  runs. Previously the registry stayed empty on this branch, so the stop
  button returned {"cancelled":0} and the decode ran to completion.
- Apply the same finally-based cleanup and pre-iteration cancel-event
  check used on the other three streaming paths.
- Update the _CANCEL_REGISTRY block comment to list cancel_id as the
  primary key (was stale "session_id preferred").

* Consolidate review tests for Studio stop-button cancel flow

- Merge the 6 behavioral tests from test_stream_cleanup_on_disconnect.py
  (finally cleanup on normal/exception/aclose, pre-set cancel_event
  pattern, and its regressions) into test_stream_cancel_registration_timing.py,
  which is the PR's existing file covering the same area.
- Extend structural invariants to include audio_input_stream alongside the
  three GGUF / Unsloth streaming generators: no _tracker.__enter__ inside
  the async gen body, cleanup via try/finally, no background= on
  StreamingResponse.
- Delete test_stream_cleanup_on_disconnect.py (now empty).

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

* studio: make cancel-via-POST interrupt Unsloth and audio-input streams

Close two remaining gaps in the stop-button cancellation wiring:

- stream_chunks (Unsloth path): add a top-of-loop cancel_event check and
  call backend.reset_generation_state() so cancel POSTs flush GPU state
  and close the SSE cleanly instead of relying on request.is_disconnected
  (which does not fire through proxies like Colab's).
- audio_input_stream: run the synchronous audio_input_generate() via
  asyncio.to_thread so blocking whisper chunks do not freeze the event
  loop, matching the pattern already used by the GGUF streaming paths.

* Add review tests for Studio stop-button cancel flow

* Consolidate review tests for Studio stop-button cancel flow

- Delete standalone test_cancel_registry.py at repo root: tests duplicated
  test_cancel_atomicity.py / test_cancel_id_wiring.py and re-implemented
  registry primitives inline (scaffolding).
- Extend tests/studio/test_stream_cancel_registration_timing.py with
  regression guards for the iter-1 cancel-loop fixes:
    structural: each streaming generator checks cancel_event in its loop;
                audio_input_stream offloads next() via asyncio.to_thread;
                stream_chunks cancel branch calls reset_generation_state().
    runtime:    Unsloth loop breaks on external cancel and resets state;
                audio loop stays responsive under blocking next();
                both loops emit zero tokens on pre-set cancel (replay path).

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

* studio: extend stop-path to passthrough streams; tighten wall-clock cap

- Lower _DEFAULT_T_MAX_PREDICT_MS from 1 hour to 10 minutes so the
  wall-clock backstop actually bounds runaway decodes when cancel
  signaling fails.
- Wire _TrackedCancel and cancel_event.is_set() into
  _openai_passthrough_stream and _anthropic_passthrough_stream and
  disable httpx keepalive so stop requests from /v1 and /v1/messages
  tool-calling clients reach llama-server.
- Apply t_max_predict_ms to the tool-passthrough request body so the
  backstop covers passthrough paths as well.
- Symmetric pre-registration stash for session_id/completion_id
  cancels (_cancel_by_keys_or_stash) so early cancels by those keys
  replay on later registration like cancel_id.
- Drop dead except BaseException guards around StreamingResponse()
  at four streaming sites; cleanup lives in the generator's finally.

* studio: harden cancel registry against ghost-cancel and leak paths

- Revert the session_id/completion_id stash in the fallback cancel
  helper. session_id is thread-scoped and reused across runs, so
  stashing it on an unmatched POST would fire cancel_event for the
  user's next unrelated request via _TrackedCancel.__enter__.
  cancel_id remains the only per-run unique key that gets stashed.
- Default max_tokens to _DEFAULT_MAX_TOKENS in the tool-passthrough
  body. Mirror the direct GGUF path so OpenAI/Anthropic passthrough
  callers who omit max_tokens get the same zombie-decode cap instead
  of relying on the wall-clock backstop alone.
- Wrap _openai_passthrough_stream setup with an outer try/except
  BaseException. The inner except httpx.RequestError does not catch
  asyncio.CancelledError at await client.send, which would otherwise
  leave _tracker registered in _CANCEL_REGISTRY indefinitely.
- Frontend stop POST uses plain fetch + manual Authorization header
  instead of authFetch. A 401 on the cancel POST no longer refreshes
  tokens or redirects the user to the login page mid-stop.

* Add review tests for Studio stop-button cancel flow

* studio: trim comments on stop-button review changes

Collapse multi-paragraph rationale blocks on the cancel registry,
_openai_passthrough_stream, and the frontend onAbortCancel handler
into one-line explanations of why the non-obvious behaviour exists.
Drop authFetch import that became unused when the cancel POST
switched to plain fetch.

* Consolidate review tests for Studio stop-button cancel flow

Move review-added tests out of test_cancel_dispatch_edges.py into the
existing PR test files that already cover the same areas:
- backend registry fan-out / exclusivity / idempotency / falsy-keys
  edge cases moved into tests/studio/test_cancel_atomicity.py
- frontend plain-fetch (not authFetch) + manual Authorization header
  moved into tests/studio/test_cancel_id_wiring.py
Delete the now-empty test_cancel_dispatch_edges.py.

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

* Studio: stop default-capping responses at 4096 tokens (follow-up to #5069) (#5174)

* Studio: stop default-capping responses at 4096 tokens

Follow-up to #5069. The 4096 default introduced for runaway-decode
defense silently truncates any caller that omits max_tokens. The
Studio chat UI sets params.maxTokens = loadResp.context_length after
a GGUF load, so it's fine, but every other consumer is not:

- OpenAI-API direct callers (/v1/chat/completions, /v1/responses,
  /v1/messages, /v1/completions) where the OpenAI default is
  effectively unlimited per response. langchain, llama-index, raw
  curl, and the openai SDK all rely on that.
- Reasoning models. Qwen3 / gpt-oss reasoning traces routinely exceed
  4096 tokens before the model emits a single visible content token.
  The user sees the trace cut off mid-thought.
- Long-form generation ("write a chapter", "produce a full SVG").

Reproduced on this branch: gemma-4-E2B-it-GGUF Q8_0, prompt asking
for a 10000-word story, no max_tokens in the request:

    finish_reason: stop  (misleading -- should be 'length')
    content_chars: 19772
    content_tail: ...'a comforting, yet immense, pressure.\n\n*"'

Body ended mid-sentence on a stray opening quote, right at the 4096
token mark.

After this patch the same request returns 38357 chars ending with
'...held in a perfect, dynamic equilibrium.' -- a natural stop, not
a truncation.

Implementation: rename the constant to _DEFAULT_MAX_TOKENS_FLOOR and
set it to 32768. Each call site now uses the model's effective
context length when known, falling back to the floor:

    default_cap = self._effective_context_length or _DEFAULT_MAX_TOKENS_FLOOR

The 10-minute t_max_predict_ms wall-clock backstop from #5069 is
preserved as the second line of defense.

Plumbed _build_passthrough_payload + _build_openai_passthrough_body
through the routes layer so the Anthropic and OpenAI passthrough
paths also respect the model's context length.

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

---------

Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com>

* Studio: cancel passthrough streams during llama-server prefill + route through apiUrl for Tauri

Three reviewer-flagged correctness gaps in the stop-button mechanism.

1) `_openai_passthrough_stream` could not honor cancel during prefill.
   The cancel check ran inside the `async for raw_line in lines_iter`
   body, so a cancel POST that arrived before llama-server emitted the
   first SSE line was unobservable until prefill completed. With a long
   prompt under proxy/Colab conditions -- the exact target scenario for
   this PR -- that left the model decoding for a long time after the
   user clicked Stop. Add an asyncio watcher task that closes `resp` as
   soon as `cancel_event` is set, raising in `aiter_lines` so the
   generator can exit. The watcher polls a threading.Event because the
   cancel registry is keyed by threading.Event for the synchronous
   /cancel handler.

2) `_anthropic_passthrough_stream` had the same blocking-prefill pattern.
   Same fix.

3) The frontend's stop-button cancel POST used a bare relative
   `fetch("/api/inference/cancel", ...)`, which targets the webview
   origin in Tauri production builds (where the backend is at
   `http://127.0.0.1:8888`). Route through the existing `apiUrl()`
   helper from `lib/api-base.ts` to match every other Studio call.
   Browser/dev builds get the empty base, so behavior is unchanged
   there.

Verified via temp/pr_simulation/sim_5069_prefill_cancel.py: cancel
during prefill terminates within ~250ms on both passthrough paths
(was 145s+ on the Anthropic path before this change), and the standard
non-passthrough chat path still cancels with no regression.

* Studio: log cancel-body parse errors instead of silently swallowing

Reviewer-flagged defensive logging gap. The bare `except Exception: pass`
in `cancel_inference` would mask malformed payloads that hint at a buggy
client or a transport issue. Log at debug so future investigation isn't
left guessing whether `body={}` came from a missing body or a parse
failure. Behavior is unchanged: an unparseable body still falls through
to the empty-dict path and the cancel call returns `{"cancelled": 0}`.

* Studio: Anthropic passthrough cancel parity with OpenAI passthrough

Two reviewer-flagged consistency gaps in the cancel surface for
/v1/messages.

1) Anthropic passthrough did not register cancel_id, so a per-run cancel
   POST (the cleanest Studio-style cancel path) silently missed when
   the route hit `_anthropic_passthrough_stream`. The OpenAI passthrough
   has registered (cancel_id, session_id, completion_id) since this PR
   was first opened; mirror that here. Also add `cancel_id` to
   `AnthropicMessagesRequest` so the route handler can plumb it through.

2) The cancel handler's fallback key list checked only completion_id
   and session_id, never message_id. Anthropic clients that send their
   native `id` (returned in the SSE message_start event) for cancel had
   no way to hit the registry. Add message_id to the fallback list.

Verified via temp/pr_simulation/sim_5069_prefill_cancel.py: P2 now
cancels by cancel_id in 137ms (was hanging pre-fix), and the new P2b
case cancels by message_id in 77ms. P1 (OpenAI) and P3 (standard chat)
still pass with no regression.

---------

Co-authored-by: danielhanchen <michaelhan2050@gmail.com>
Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com>
Co-authored-by: Roland Tannous <115670425+rolandtannous@users.noreply.github.com>
Co-authored-by: Lee Jackson <130007945+Imagineer99@users.noreply.github.com>
---
 studio/backend/core/inference/llama_cpp.py    |  57 +-
 studio/backend/main.py                        |   4 +
 studio/backend/models/inference.py            |   5 +
 studio/backend/routes/__init__.py             |   2 +
 studio/backend/routes/inference.py            | 495 +++++++++---
 .../src/features/chat/api/chat-adapter.ts     |  41 +-
 .../frontend/src/features/chat/types/api.ts   |   1 +
 tests/studio/test_cancel_atomicity.py         | 289 +++++++
 tests/studio/test_cancel_id_wiring.py         | 169 +++++
 tests/studio/test_llama_cpp_wall_clock_cap.py | 123 +++
 .../test_stream_cancel_registration_timing.py | 718 ++++++++++++++++++
 11 files changed, 1785 insertions(+), 119 deletions(-)
 create mode 100644 tests/studio/test_cancel_atomicity.py
 create mode 100644 tests/studio/test_cancel_id_wiring.py
 create mode 100644 tests/studio/test_llama_cpp_wall_clock_cap.py
 create mode 100644 tests/studio/test_stream_cancel_registration_timing.py

diff --git a/studio/backend/core/inference/llama_cpp.py b/studio/backend/core/inference/llama_cpp.py
index 35ef90ee37..81ff9ae4e5 100644
--- a/studio/backend/core/inference/llama_cpp.py
+++ b/studio/backend/core/inference/llama_cpp.py
@@ -53,6 +53,21 @@ _INTENT_SIGNAL = re.compile(
     r")"
 )
 _MAX_REPROMPTS = 3
+
+# Without max_tokens, llama-server defaults to n_predict = n_ctx (up to
+# 262144 for Qwen3.5), producing many-minute zombie decodes when cancel
+# fails. t_max_predict_ms is a wall-clock backstop applied unconditionally,
+# but the llama.cpp README notes it ONLY fires after a newline has been
+# generated -- a model stuck in a long unbroken non-newline sequence is
+# unbounded by it. So we still want a token cap as the front-line limiter.
+#
+# The cap is the model's effective context length when we know it,
+# falling back to a generous floor when metadata is unavailable. 4096 was
+# too low: Qwen3 / gpt-oss reasoning traces routinely exceed it, and any
+# OpenAI-API caller that omits max_tokens (langchain, llama-index, raw
+# curl) sees responses silently truncated mid-sentence.
+_DEFAULT_MAX_TOKENS_FLOOR = 32768
+_DEFAULT_T_MAX_PREDICT_MS = 600_000  # 10 min
 _REPROMPT_MAX_CHARS = 2000
 
 # ── Pre-compiled patterns for GGUF shard detection ───────────
@@ -1636,7 +1651,7 @@ class LlamaCppBackend:
             # existing text (code refactoring, summarization, reasoning).
             # For general chat with low repetition, overhead is ~5 ms.
             #
-            # Benchmarks from llama.cpp PRs #18471, #19164:
+            # Benchmarks from upstream llama.cpp speculative-decoding PRs:
             #   Scenario                        | Without | With    | Speedup
             #   gpt-oss-120b code refactor      | 181 t/s | 446 t/s | 2.5x
             #   Qwen3-235B offloaded            |  12 t/s |  21 t/s | 1.8x
@@ -2549,8 +2564,15 @@ class LlamaCppBackend:
         )
         if _reasoning_kw is not None:
             payload["chat_template_kwargs"] = _reasoning_kw
-        if max_tokens is not None:
-            payload["max_tokens"] = max_tokens
+        # Default cap to the model's effective context length when known,
+        # otherwise the conservative floor. The wall-clock backstop below
+        # keeps a stuck model from running indefinitely either way.
+        payload["max_tokens"] = (
+            max_tokens
+            if max_tokens is not None
+            else (self._effective_context_length or _DEFAULT_MAX_TOKENS_FLOOR)
+        )
+        payload["t_max_predict_ms"] = _DEFAULT_T_MAX_PREDICT_MS
         if stop:
             payload["stop"] = stop
         payload["stream_options"] = {"include_usage": True}
@@ -2570,7 +2592,9 @@ class LlamaCppBackend:
             _auth_headers = (
                 {"Authorization": f"Bearer {self._api_key}"} if self._api_key else None
             )
-            with httpx.Client(timeout = stream_timeout) as client:
+            with httpx.Client(
+                timeout = stream_timeout, limits = httpx.Limits(max_keepalive_connections = 0)
+            ) as client:
                 with self._stream_with_retry(
                     client,
                     url,
@@ -2769,8 +2793,12 @@ class LlamaCppBackend:
             )
             if _reasoning_kw is not None:
                 payload["chat_template_kwargs"] = _reasoning_kw
-            if max_tokens is not None:
-                payload["max_tokens"] = max_tokens
+            payload["max_tokens"] = (
+                max_tokens
+                if max_tokens is not None
+                else (self._effective_context_length or _DEFAULT_MAX_TOKENS_FLOOR)
+            )
+            payload["t_max_predict_ms"] = _DEFAULT_T_MAX_PREDICT_MS
             if stop:
                 payload["stop"] = stop
 
@@ -2809,7 +2837,10 @@ class LlamaCppBackend:
                     write = 10,
                     pool = 10,
                 )
-                with httpx.Client(timeout = stream_timeout) as client:
+                with httpx.Client(
+                    timeout = stream_timeout,
+                    limits = httpx.Limits(max_keepalive_connections = 0),
+                ) as client:
                     with self._stream_with_retry(
                         client,
                         url,
@@ -3422,8 +3453,12 @@ class LlamaCppBackend:
         )
         if _reasoning_kw is not None:
             stream_payload["chat_template_kwargs"] = _reasoning_kw
-        if max_tokens is not None:
-            stream_payload["max_tokens"] = max_tokens
+        stream_payload["max_tokens"] = (
+            max_tokens
+            if max_tokens is not None
+            else (self._effective_context_length or _DEFAULT_MAX_TOKENS_FLOOR)
+        )
+        stream_payload["t_max_predict_ms"] = _DEFAULT_T_MAX_PREDICT_MS
         if stop:
             stream_payload["stop"] = stop
         stream_payload["stream_options"] = {"include_usage": True}
@@ -3442,7 +3477,9 @@ class LlamaCppBackend:
             _auth_headers = (
                 {"Authorization": f"Bearer {self._api_key}"} if self._api_key else None
             )
-            with httpx.Client(timeout = stream_timeout) as client:
+            with httpx.Client(
+                timeout = stream_timeout, limits = httpx.Limits(max_keepalive_connections = 0)
+            ) as client:
                 with self._stream_with_retry(
                     client,
                     url,
diff --git a/studio/backend/main.py b/studio/backend/main.py
index 05adcaa2ea..9212404b30 100644
--- a/studio/backend/main.py
+++ b/studio/backend/main.py
@@ -62,6 +62,7 @@ from routes import (
     datasets_router,
     export_router,
     inference_router,
+    inference_studio_router,
     models_router,
     training_history_router,
     training_router,
@@ -207,6 +208,9 @@ app.include_router(auth_router, prefix = "/api/auth", tags = ["auth"])
 app.include_router(training_router, prefix = "/api/train", tags = ["training"])
 app.include_router(models_router, prefix = "/api/models", tags = ["models"])
 app.include_router(inference_router, prefix = "/api/inference", tags = ["inference"])
+# Studio-only inference endpoints (cancel, etc.) are intentionally NOT
+# exposed on the /v1 OpenAI-compat prefix below.
+app.include_router(inference_studio_router, prefix = "/api/inference", tags = ["inference"])
 
 # OpenAI-compatible endpoints: mount the same inference router at /v1
 # so external tools (Open WebUI, SillyTavern, etc.) can use the
diff --git a/studio/backend/models/inference.py b/studio/backend/models/inference.py
index e5b037755d..bf0177efbf 100644
--- a/studio/backend/models/inference.py
+++ b/studio/backend/models/inference.py
@@ -531,6 +531,10 @@ class ChatCompletionRequest(BaseModel):
         None,
         description = "[x-unsloth] Session/thread ID for scoping tool execution sandbox.",
     )
+    cancel_id: Optional[str] = Field(
+        None,
+        description = "[x-unsloth] Per-request cancellation token. Frontend sends a fresh UUID per run so /inference/cancel matches one specific generation.",
+    )
 
 
 # ── Streaming response chunks ────────────────────────────────────
@@ -992,6 +996,7 @@ class AnthropicMessagesRequest(BaseModel):
     enable_tools: Optional[bool] = None
     enabled_tools: Optional[list[str]] = None
     session_id: Optional[str] = None
+    cancel_id: Optional[str] = None
     model_config = {"extra": "allow"}
 
 
diff --git a/studio/backend/routes/__init__.py b/studio/backend/routes/__init__.py
index e79f6553f9..cf4586281b 100644
--- a/studio/backend/routes/__init__.py
+++ b/studio/backend/routes/__init__.py
@@ -8,6 +8,7 @@ API Routes
 from routes.training import router as training_router
 from routes.models import router as models_router
 from routes.inference import router as inference_router
+from routes.inference import studio_router as inference_studio_router
 from routes.datasets import router as datasets_router
 from routes.auth import router as auth_router
 from routes.data_recipe import router as data_recipe_router
@@ -18,6 +19,7 @@ __all__ = [
     "training_router",
     "models_router",
     "inference_router",
+    "inference_studio_router",
     "datasets_router",
     "auth_router",
     "data_recipe_router",
diff --git a/studio/backend/routes/inference.py b/studio/backend/routes/inference.py
index ed331a5660..cf3b37a2fd 100644
--- a/studio/backend/routes/inference.py
+++ b/studio/backend/routes/inference.py
@@ -113,7 +113,12 @@ if str(backend_path) not in sys.path:
 # Import backend functions
 try:
     from core.inference import get_inference_backend
-    from core.inference.llama_cpp import LlamaCppBackend, detect_reasoning_flags
+    from core.inference.llama_cpp import (
+        LlamaCppBackend,
+        _DEFAULT_MAX_TOKENS_FLOOR,
+        _DEFAULT_T_MAX_PREDICT_MS,
+        detect_reasoning_flags,
+    )
     from utils.models import ModelConfig
     from utils.inference import load_inference_config
     from utils.models.model_config import load_model_defaults
@@ -122,7 +127,12 @@ except ImportError:
     if str(parent_backend) not in sys.path:
         sys.path.insert(0, str(parent_backend))
     from core.inference import get_inference_backend
-    from core.inference.llama_cpp import LlamaCppBackend, detect_reasoning_flags
+    from core.inference.llama_cpp import (
+        LlamaCppBackend,
+        _DEFAULT_MAX_TOKENS_FLOOR,
+        _DEFAULT_T_MAX_PREDICT_MS,
+        detect_reasoning_flags,
+    )
     from utils.models import ModelConfig
     from utils.inference import load_inference_config
     from utils.models.model_config import load_model_defaults
@@ -185,6 +195,126 @@ import numpy as np
 from datetime import date as _date
 
 router = APIRouter()
+# Studio-only router (not mounted on /v1 OpenAI-compat).
+studio_router = APIRouter()
+
+
+# Cancel registry. Proxies (e.g. Colab) can swallow client fetch aborts
+# so is_disconnected() never fires. POST /inference/cancel looks up
+# in-flight cancel_events here by cancel_id (per-run) or session_id /
+# completion_id (fallbacks).
+_CANCEL_REGISTRY: dict[str, set[threading.Event]] = {}
+_CANCEL_LOCK = threading.Lock()
+
+# Cancel POSTs that arrive before registration are stashed; the next
+# matching __enter__ replays set() within the TTL.
+_PENDING_CANCELS: dict[str, float] = {}
+_PENDING_CANCEL_TTL_S = 30.0
+
+
+def _prune_pending(now: float) -> None:
+    for k in [
+        k for k, ts in _PENDING_CANCELS.items() if now - ts > _PENDING_CANCEL_TTL_S
+    ]:
+        _PENDING_CANCELS.pop(k, None)
+
+
+class _TrackedCancel:
+    """Register cancel_event in _CANCEL_REGISTRY for the block's duration."""
+
+    def __init__(self, event: threading.Event, *keys):
+        self.event = event
+        self.keys = tuple(k for k in keys if k)
+
+    def __enter__(self):
+        # Register + consume-pending must be one critical section to close
+        # the TOCTOU race against a concurrent cancel POST.
+        should_cancel = False
+        with _CANCEL_LOCK:
+            for k in self.keys:
+                _CANCEL_REGISTRY.setdefault(k, set()).add(self.event)
+            now = time.monotonic()
+            _prune_pending(now)
+            for k in self.keys:
+                if k and _PENDING_CANCELS.pop(k, None) is not None:
+                    should_cancel = True
+        if should_cancel:
+            self.event.set()
+        return self.event
+
+    def __exit__(self, *exc):
+        with _CANCEL_LOCK:
+            for k in self.keys:
+                bucket = _CANCEL_REGISTRY.get(k)
+                if bucket is None:
+                    continue
+                bucket.discard(self.event)
+                if not bucket:
+                    _CANCEL_REGISTRY.pop(k, None)
+        return False
+
+
+def _cancel_by_keys(keys) -> int:
+    """Set cancel_event for matching registry entries; no stash.
+    session_id/completion_id are shared across runs on the same thread,
+    so stashing them would ghost-cancel the user's next request. Only
+    cancel_id is per-run unique (see _cancel_by_cancel_id_or_stash)."""
+    if not keys:
+        return 0
+    events: set[threading.Event] = set()
+    with _CANCEL_LOCK:
+        _prune_pending(time.monotonic())
+        for k in keys:
+            bucket = _CANCEL_REGISTRY.get(k)
+            if bucket:
+                events.update(bucket)
+    for ev in events:
+        ev.set()
+    return len(events)
+
+
+def _cancel_by_cancel_id_or_stash(cancel_id: str) -> int:
+    """Atomic lookup-or-stash; pairs with _TrackedCancel.__enter__ to
+    close the TOCTOU race."""
+    now = time.monotonic()
+    events: set[threading.Event] = set()
+    with _CANCEL_LOCK:
+        _prune_pending(now)
+        bucket = _CANCEL_REGISTRY.get(cancel_id)
+        if bucket:
+            events.update(bucket)
+        else:
+            _PENDING_CANCELS[cancel_id] = now
+    for ev in events:
+        ev.set()
+    return len(events)
+
+
+async def _await_cancel_then_close(cancel_event, resp) -> None:
+    """Watch a threading.Event from asyncio and close ``resp`` when it fires.
+
+    Used by the passthrough streamers so a /cancel POST can interrupt
+    while the async iterator is blocked waiting for llama-server prefill.
+    Without this watcher the in-loop ``cancel_event.is_set()`` check is
+    unreachable until the first SSE chunk arrives, which is exactly the
+    proxy/Colab scenario the cancel POST exists to handle.
+
+    Polls a threading.Event because the cancel registry is keyed by
+    threading.Event so the synchronous /cancel handler can call .set().
+    50ms cadence adds at most that much latency to a prefill cancel; the
+    common-case streaming cancel path still observes the event in the
+    iterator's first iteration after the next chunk.
+    """
+    try:
+        while not cancel_event.is_set():
+            await asyncio.sleep(0.05)
+        try:
+            await resp.aclose()
+        except Exception:
+            pass
+    except asyncio.CancelledError:
+        return
+
 
 # Appended to tool-use nudge to discourage plan-without-action
 _TOOL_ACTION_NUDGE = (
@@ -706,6 +836,48 @@ async def unload_model(
         raise HTTPException(status_code = 500, detail = f"Failed to unload model: {str(e)}")
 
 
+@studio_router.post("/cancel")
+async def cancel_inference(
+    request: Request,
+    current_subject: str = Depends(get_current_subject),
+):
+    """Cancel in-flight inference requests.
+
+    Body (JSON, at least one key required):
+      cancel_id    - preferred: per-run UUID, matched exclusively.
+      session_id   - fallback when cancel_id is absent.
+      completion_id - fallback when cancel_id is absent.
+
+    A cancel_id arriving before its stream registers is stashed briefly
+    and replayed on registration. Returns {"cancelled": N}.
+    """
+    try:
+        body = await request.json()
+        if not isinstance(body, dict):
+            body = {}
+    except Exception as e:
+        logger.debug("Failed to parse cancel request body: %s", e)
+        body = {}
+
+    cancel_id = body.get("cancel_id")
+    if isinstance(cancel_id, str) and cancel_id:
+        return {"cancelled": _cancel_by_cancel_id_or_stash(cancel_id)}
+
+    keys = []
+    # `message_id` is the Anthropic passthrough's per-run identifier --
+    # included so /v1/messages clients can cancel by their native id.
+    for k in ("completion_id", "session_id", "message_id"):
+        v = body.get(k)
+        if isinstance(v, str) and v:
+            keys.append(v)
+
+    if not keys:
+        return {"cancelled": 0}
+
+    n = _cancel_by_keys(keys)
+    return {"cancelled": n}
+
+
 @router.post("/generate/stream")
 async def generate_stream(
     request: GenerateRequest,
@@ -1169,6 +1341,9 @@ async def openai_chat_completions(
                 )
 
             if payload.stream:
+                _cancel_keys = (payload.cancel_id, payload.session_id, completion_id)
+                _tracker = _TrackedCancel(cancel_event, *_cancel_keys)
+                _tracker.__enter__()
 
                 async def audio_input_stream():
                     try:
@@ -1185,10 +1360,17 @@ async def openai_chat_completions(
                         )
                         yield f"data: {first_chunk.model_dump_json(exclude_none = True)}\n\n"
 
-                        for chunk_text in audio_input_generate():
+                        gen = audio_input_generate()
+                        _DONE = object()
+                        while True:
+                            if cancel_event.is_set():
+                                break
                             if await request.is_disconnected():
                                 cancel_event.set()
                                 return
+                            chunk_text = await asyncio.to_thread(next, gen, _DONE)
+                            if chunk_text is _DONE:
+                                break
                             if chunk_text:
                                 chunk = ChatCompletionChunk(
                                     id = completion_id,
@@ -1221,6 +1403,8 @@ async def openai_chat_completions(
                             f"Error during audio input streaming: {e}", exc_info = True
                         )
                         yield f"data: {json.dumps({'error': {'message': _friendly_error(e), 'type': 'server_error'}})}\n\n"
+                    finally:
+                        _tracker.__exit__(None, None, None)
 
                 return StreamingResponse(
                     audio_input_stream(),
@@ -1466,6 +1650,10 @@ async def openai_chat_completions(
 
             _tool_sentinel = object()
 
+            _cancel_keys = (payload.cancel_id, payload.session_id, completion_id)
+            _tracker = _TrackedCancel(cancel_event, *_cancel_keys)
+            _tracker.__enter__()
+
             async def gguf_tool_stream():
                 try:
                     first_chunk = ChatCompletionChunk(
@@ -1488,6 +1676,8 @@ async def openai_chat_completions(
                     _stream_usage = None
                     _stream_timings = None
                     while True:
+                        if cancel_event.is_set():
+                            break
                         if await request.is_disconnected():
                             cancel_event.set()
                             return
@@ -1595,6 +1785,8 @@ async def openai_chat_completions(
                         },
                     }
                     yield f"data: {json.dumps(error_chunk)}\n\n"
+                finally:
+                    _tracker.__exit__(None, None, None)
 
             return StreamingResponse(
                 gguf_tool_stream(),
@@ -1628,6 +1820,9 @@ async def openai_chat_completions(
         _gguf_sentinel = object()
 
         if payload.stream:
+            _cancel_keys = (payload.cancel_id, payload.session_id, completion_id)
+            _tracker = _TrackedCancel(cancel_event, *_cancel_keys)
+            _tracker.__enter__()
 
             async def gguf_stream_chunks():
                 try:
@@ -1652,6 +1847,8 @@ async def openai_chat_completions(
                     _stream_usage = None
                     _stream_timings = None
                     while True:
+                        if cancel_event.is_set():
+                            break
                         if await request.is_disconnected():
                             cancel_event.set()
                             return
@@ -1735,6 +1932,8 @@ async def openai_chat_completions(
                         },
                     }
                     yield f"data: {json.dumps(error_chunk)}\n\n"
+                finally:
+                    _tracker.__exit__(None, None, None)
 
             return StreamingResponse(
                 gguf_stream_chunks(),
@@ -1834,6 +2033,9 @@ async def openai_chat_completions(
 
     # ── Streaming response ────────────────────────────────────────
     if payload.stream:
+        _cancel_keys = (payload.cancel_id, payload.session_id, completion_id)
+        _tracker = _TrackedCancel(cancel_event, *_cancel_keys)
+        _tracker.__enter__()
 
         async def stream_chunks():
             try:
@@ -1861,6 +2063,9 @@ async def openai_chat_completions(
                 loop = asyncio.get_event_loop()
                 gen = generate()
                 while True:
+                    if cancel_event.is_set():
+                        backend.reset_generation_state()
+                        break
                     # next(gen, _DONE) returns _DONE instead of raising
                     # StopIteration — StopIteration cannot propagate
                     # through asyncio futures (Python limitation).
@@ -1916,6 +2121,8 @@ async def openai_chat_completions(
                     },
                 }
                 yield f"data: {json.dumps(error_chunk)}\n\n"
+            finally:
+                _tracker.__exit__(None, None, None)
 
         return StreamingResponse(
             stream_chunks(),
@@ -2596,7 +2803,9 @@ async def _responses_stream(
             ),
         )
 
-    body = _build_openai_passthrough_body(chat_req)
+    body = _build_openai_passthrough_body(
+        chat_req, backend_ctx = llama_backend.context_length
+    )
     target_url = f"{llama_backend.base_url}/v1/chat/completions"
 
     async def event_generator():
@@ -3081,6 +3290,8 @@ async def anthropic_messages(
                 repetition_penalty = repetition_penalty,
                 presence_penalty = presence_penalty,
                 tool_choice = openai_tool_choice,
+                session_id = payload.session_id,
+                cancel_id = payload.cancel_id,
             )
         return await _anthropic_passthrough_non_streaming(
             llama_backend,
@@ -3441,6 +3652,7 @@ def _build_passthrough_payload(
     repetition_penalty = None,
     presence_penalty = None,
     tool_choice = "auto",
+    backend_ctx = None,
 ):
     body = {
         "messages": openai_messages,
@@ -3453,8 +3665,12 @@ def _build_passthrough_payload(
     }
     if stream:
         body["stream_options"] = {"include_usage": True}
-    if max_tokens is not None:
-        body["max_tokens"] = max_tokens
+    body["max_tokens"] = (
+        max_tokens
+        if max_tokens is not None
+        else (backend_ctx or _DEFAULT_MAX_TOKENS_FLOOR)
+    )
+    body["t_max_predict_ms"] = _DEFAULT_T_MAX_PREDICT_MS
     if stop:
         body["stop"] = stop
     if min_p is not None:
@@ -3484,6 +3700,8 @@ async def _anthropic_passthrough_stream(
     repetition_penalty = None,
     presence_penalty = None,
     tool_choice = "auto",
+    session_id = None,
+    cancel_id = None,
 ):
     """Streaming client-side pass-through: forward tools to llama-server and
     translate its streaming response to Anthropic SSE without executing anything."""
@@ -3501,8 +3719,14 @@ async def _anthropic_passthrough_stream(
         repetition_penalty = repetition_penalty,
         presence_penalty = presence_penalty,
         tool_choice = tool_choice,
+        backend_ctx = llama_backend.context_length,
     )
 
+    # cancel_id mirrors the OpenAI passthrough so a per-run cancel POST
+    # works without the caller having to know the local message_id.
+    _tracker = _TrackedCancel(cancel_event, cancel_id, session_id, message_id)
+    _tracker.__enter__()
+
     async def _stream():
         emitter = AnthropicPassthroughEmitter()
         for line in emitter.start(message_id, model_name):
@@ -3535,15 +3759,28 @@ async def _anthropic_passthrough_stream(
         # has anything orphaned to finalize. Each aclose is wrapped in
         # `try: ... except Exception: pass` so anyio cleanup noise from
         # nested aclose paths can't bubble out.
-        client = httpx.AsyncClient(timeout = 600)
+        client = httpx.AsyncClient(
+            timeout = 600,
+            limits = httpx.Limits(max_keepalive_connections = 0),
+        )
         resp = None
         lines_iter = None
+        cancel_watcher = None
         try:
             req = client.build_request("POST", target_url, json = body)
             resp = await client.send(req, stream = True)
 
+            # See _openai_passthrough_stream for rationale: aiter_lines()
+            # blocks during llama-server prefill, so the in-loop cancel
+            # check is unreachable until the first SSE chunk arrives.
+            # The watcher closes `resp` on cancel, raising in aiter_lines.
+            cancel_watcher = asyncio.create_task(
+                _await_cancel_then_close(cancel_event, resp)
+            )
             lines_iter = resp.aiter_lines()
             async for raw_line in lines_iter:
+                if cancel_event.is_set():
+                    break
                 if await request.is_disconnected():
                     cancel_event.set()
                     break
@@ -3558,9 +3795,18 @@ async def _anthropic_passthrough_stream(
                     continue
                 for line in emitter.feed_chunk(chunk):
                     yield line
+        except (httpx.RemoteProtocolError, httpx.ReadError, httpx.CloseError):
+            if not cancel_event.is_set():
+                raise
         except Exception as e:
             logger.error("anthropic_messages passthrough stream error: %s", e)
         finally:
+            if cancel_watcher is not None:
+                cancel_watcher.cancel()
+                try:
+                    await cancel_watcher
+                except (asyncio.CancelledError, Exception):
+                    pass
             if lines_iter is not None:
                 try:
                     await lines_iter.aclose()
@@ -3575,6 +3821,7 @@ async def _anthropic_passthrough_stream(
                 await client.aclose()
             except Exception:
                 pass
+            _tracker.__exit__(None, None, None)
 
         for line in emitter.finish():
             yield line
@@ -3621,6 +3868,7 @@ async def _anthropic_passthrough_non_streaming(
         repetition_penalty = repetition_penalty,
         presence_penalty = presence_penalty,
         tool_choice = tool_choice,
+        backend_ctx = llama_backend.context_length,
     )
 
     async with httpx.AsyncClient() as client:
@@ -3742,7 +3990,7 @@ def _openai_messages_for_passthrough(payload) -> list[dict]:
     return messages
 
 
-def _build_openai_passthrough_body(payload) -> dict:
+def _build_openai_passthrough_body(payload, backend_ctx = None) -> dict:
     """Assemble the llama-server request body from a ChatCompletionRequest.
 
     Only explicitly-known OpenAI / llama-server fields are forwarded so that
@@ -3764,6 +4012,7 @@ def _build_openai_passthrough_body(payload) -> dict:
         repetition_penalty = payload.repetition_penalty,
         presence_penalty = payload.presence_penalty,
         tool_choice = tool_choice,
+        backend_ctx = backend_ctx,
     )
 
 
@@ -3784,103 +4033,56 @@ async def _openai_passthrough_stream(
     observes a standard OpenAI response.
     """
     target_url = f"{llama_backend.base_url}/v1/chat/completions"
-    body = _build_openai_passthrough_body(payload)
+    body = _build_openai_passthrough_body(
+        payload, backend_ctx = llama_backend.context_length
+    )
 
-    # Dispatch the upstream request BEFORE returning StreamingResponse so
-    # transport errors and non-200 upstream statuses surface as real HTTP
-    # errors to the client. OpenAI SDKs rely on status codes to raise
-    # ``APIError``/``BadRequestError``/...; burying the failure inside a
-    # 200 SSE ``error`` frame silently breaks their error handling.
-    client = httpx.AsyncClient(timeout = 600)
-    resp = None
+    _cancel_keys = (payload.cancel_id, payload.session_id, completion_id)
+    _tracker = _TrackedCancel(cancel_event, *_cancel_keys)
+    _tracker.__enter__()
+
+    # Outer guard: asyncio.CancelledError at `await client.send(...)` is
+    # a BaseException that bypasses `except httpx.RequestError`; without
+    # this the tracker leaks. The generator's finally only runs once
+    # iteration starts.
     try:
-        req = client.build_request("POST", target_url, json = body)
-        resp = await client.send(req, stream = True)
-    except httpx.RequestError as e:
-        # llama-server subprocess crashed / still starting / unreachable.
-        logger.error("openai passthrough stream: upstream unreachable: %s", e)
-        if resp is not None:
-            try:
-                await resp.aclose()
-            except Exception:
-                pass
-        try:
-            await client.aclose()
-        except Exception:
-            pass
-        raise HTTPException(
-            status_code = 502,
-            detail = _friendly_error(e),
+        # Dispatch BEFORE returning StreamingResponse so transport errors
+        # and non-200 upstream statuses surface as real HTTP errors --
+        # OpenAI SDKs rely on status codes to raise APIError/BadRequestError.
+        client = httpx.AsyncClient(
+            timeout = 600,
+            limits = httpx.Limits(max_keepalive_connections = 0),
         )
-
-    if resp.status_code != 200:
-        err_bytes = await resp.aread()
-        err_text = err_bytes.decode("utf-8", errors = "replace")
-        logger.error(
-            "openai passthrough upstream error: status=%s body=%s",
-            resp.status_code,
-            err_text[:500],
-        )
-        upstream_status = resp.status_code
+        resp = None
         try:
-            await resp.aclose()
-        except Exception:
-            pass
-        try:
-            await client.aclose()
-        except Exception:
-            pass
-        raise HTTPException(
-            status_code = upstream_status,
-            detail = f"llama-server error: {err_text[:500]}",
-        )
-
-    async def _stream():
-        # Same httpx lifecycle pattern as _anthropic_passthrough_stream:
-        # avoid `async with` on the client/response AND explicitly save
-        # resp.aiter_lines() so we can close it ourselves in the finally
-        # block. See the long comment there for the full rationale on
-        # why the anonymous `async for raw_line in resp.aiter_lines():`
-        # pattern leaks an unclosed async generator that Python's
-        # asyncgen GC hook then finalizes in a different asyncio task,
-        # producing "Exception ignored in:" / "async generator ignored
-        # GeneratorExit" / anyio cancel-scope traces on Python 3.13 +
-        # httpcore 1.0.x.
-        lines_iter = None
-        try:
-            lines_iter = resp.aiter_lines()
-            async for raw_line in lines_iter:
-                if await request.is_disconnected():
-                    cancel_event.set()
-                    break
-                if not raw_line:
-                    continue
-                if not raw_line.startswith("data: "):
-                    continue
-                # Relay the llama-server SSE chunk verbatim so the client
-                # sees its native `id`, `finish_reason`, `delta.tool_calls`,
-                # and final `usage` unchanged.
-                yield raw_line + "\n\n"
-                if raw_line[6:].strip() == "[DONE]":
-                    break
-        except Exception as e:
-            # Mid-stream failures still have to be reported inside the SSE
-            # body because the 200 response headers have already been
-            # committed by the time the first chunk flushes.
-            logger.error("openai passthrough stream error: %s", e)
-            err = {
-                "error": {
-                    "message": _friendly_error(e),
-                    "type": "server_error",
-                },
-            }
-            yield f"data: {json.dumps(err)}\n\n"
-        finally:
-            if lines_iter is not None:
+            req = client.build_request("POST", target_url, json = body)
+            resp = await client.send(req, stream = True)
+        except httpx.RequestError as e:
+            # llama-server subprocess crashed / still starting / unreachable.
+            logger.error("openai passthrough stream: upstream unreachable: %s", e)
+            if resp is not None:
                 try:
-                    await lines_iter.aclose()
+                    await resp.aclose()
                 except Exception:
                     pass
+            try:
+                await client.aclose()
+            except Exception:
+                pass
+            raise HTTPException(
+                status_code = 502,
+                detail = _friendly_error(e),
+            )
+
+        if resp.status_code != 200:
+            err_bytes = await resp.aread()
+            err_text = err_bytes.decode("utf-8", errors = "replace")
+            logger.error(
+                "openai passthrough upstream error: status=%s body=%s",
+                resp.status_code,
+                err_text[:500],
+            )
+            upstream_status = resp.status_code
             try:
                 await resp.aclose()
             except Exception:
@@ -3889,16 +4091,91 @@ async def _openai_passthrough_stream(
                 await client.aclose()
             except Exception:
                 pass
+            raise HTTPException(
+                status_code = upstream_status,
+                detail = f"llama-server error: {err_text[:500]}",
+            )
 
-    return StreamingResponse(
-        _stream(),
-        media_type = "text/event-stream",
-        headers = {
-            "Cache-Control": "no-cache",
-            "Connection": "keep-alive",
-            "X-Accel-Buffering": "no",
-        },
-    )
+        async def _stream():
+            # Same httpx lifecycle pattern as _anthropic_passthrough_stream:
+            # save resp.aiter_lines() so the finally block can aclose() it
+            # on our task. See that function for full rationale.
+            lines_iter = None
+            # During llama-server prefill, `aiter_lines()` blocks until the
+            # first SSE chunk arrives. The in-loop `cancel_event` check
+            # cannot fire until then, which is the exact proxy/Colab
+            # scenario the cancel POST is meant to recover from. Run a
+            # tiny watcher that closes `resp` as soon as cancel fires,
+            # unblocking the iterator with a RemoteProtocolError caught
+            # in the except clause below.
+            cancel_watcher = asyncio.create_task(
+                _await_cancel_then_close(cancel_event, resp)
+            )
+            try:
+                lines_iter = resp.aiter_lines()
+                async for raw_line in lines_iter:
+                    if cancel_event.is_set():
+                        break
+                    if await request.is_disconnected():
+                        cancel_event.set()
+                        break
+                    if not raw_line:
+                        continue
+                    if not raw_line.startswith("data: "):
+                        continue
+                    # Relay verbatim to preserve llama-server's native id,
+                    # finish_reason, delta.tool_calls, and usage chunks.
+                    yield raw_line + "\n\n"
+                    if raw_line[6:].strip() == "[DONE]":
+                        break
+            except (httpx.RemoteProtocolError, httpx.ReadError, httpx.CloseError):
+                # Watcher closed resp on cancel. Emit nothing extra; the
+                # client either initiated the cancel or already disconnected.
+                if not cancel_event.is_set():
+                    raise
+            except Exception as e:
+                # 200 headers are already flushed; errors must be in the SSE body.
+                logger.error("openai passthrough stream error: %s", e)
+                err = {
+                    "error": {
+                        "message": _friendly_error(e),
+                        "type": "server_error",
+                    },
+                }
+                yield f"data: {json.dumps(err)}\n\n"
+            finally:
+                cancel_watcher.cancel()
+                try:
+                    await cancel_watcher
+                except (asyncio.CancelledError, Exception):
+                    pass
+                if lines_iter is not None:
+                    try:
+                        await lines_iter.aclose()
+                    except Exception:
+                        pass
+                try:
+                    await resp.aclose()
+                except Exception:
+                    pass
+                try:
+                    await client.aclose()
+                except Exception:
+                    pass
+                _tracker.__exit__(None, None, None)
+
+        return StreamingResponse(
+            _stream(),
+            media_type = "text/event-stream",
+            headers = {
+                "Cache-Control": "no-cache",
+                "Connection": "keep-alive",
+                "X-Accel-Buffering": "no",
+            },
+        )
+    except BaseException:
+        _tracker.__exit__(None, None, None)
+        raise
 
 
 async def _openai_passthrough_non_streaming(
@@ -3914,7 +4191,9 @@ async def _openai_passthrough_non_streaming(
     token counts.
     """
     target_url = f"{llama_backend.base_url}/v1/chat/completions"
-    body = _build_openai_passthrough_body(payload)
+    body = _build_openai_passthrough_body(
+        payload, backend_ctx = llama_backend.context_length
+    )
 
     try:
         async with httpx.AsyncClient() as client:
diff --git a/studio/frontend/src/features/chat/api/chat-adapter.ts b/studio/frontend/src/features/chat/api/chat-adapter.ts
index a93120397e..dde597ca11 100644
--- a/studio/frontend/src/features/chat/api/chat-adapter.ts
+++ b/studio/frontend/src/features/chat/api/chat-adapter.ts
@@ -4,6 +4,8 @@
 import type { ChatModelAdapter } from "@assistant-ui/react";
 import type { MessageTiming, ToolCallMessagePart } from "@assistant-ui/core";
 import { toast } from "sonner";
+import { getAuthToken } from "@/features/auth/session";
+import { apiUrl } from "@/lib/api-base";
 import {
   generateAudio,
   listCachedGguf,
@@ -707,7 +709,42 @@ export function createOpenAIStreamAdapter(): ChatModelAdapter {
       const toolCallParts: ToolCallMessagePart[] = [];
       let serverMetadata: { usage?: ServerUsage; timings?: ServerTimings } | null = null;
 
+      // Per-run cancellation token so a delayed stop POST cannot match
+      // the next run on the same thread.
+      const cancelId =
+        typeof crypto !== "undefined" && "randomUUID" in crypto
+          ? crypto.randomUUID()
+          : `${Date.now()}-${Math.random().toString(36).slice(2)}`;
+
+      // Colab-style proxies can swallow fetch aborts, so also POST
+      // /inference/cancel explicitly on abort.
+      const onAbortCancel = () => {
+        const body: Record<string, string> = { cancel_id: cancelId };
+        if (resolvedThreadId) body.session_id = resolvedThreadId;
+        // Plain fetch, not authFetch: authFetch redirects to login on
+        // 401, which would kick the user out mid-stop.
+        const token = getAuthToken();
+        // Use apiUrl so the cancel POST reaches the right origin in
+        // Tauri production builds (where the webview origin is not the
+        // backend at 127.0.0.1:<port>). Browser/dev builds get the empty
+        // base, so the path is unchanged there.
+        void fetch(apiUrl("/api/inference/cancel"), {
+          method: "POST",
+          headers: {
+            "Content-Type": "application/json",
+            ...(token ? { Authorization: `Bearer ${token}` } : {}),
+          },
+          body: JSON.stringify(body),
+          keepalive: true,
+        }).catch(() => {});
+      };
       try {
+        if (abortSignal.aborted) {
+          onAbortCancel();
+        } else {
+          abortSignal.addEventListener("abort", onAbortCancel, { once: true });
+        }
+
         const {
           supportsReasoning,
           reasoningEnabled,
@@ -730,6 +767,8 @@ export function createOpenAIStreamAdapter(): ChatModelAdapter {
             presence_penalty: params.presencePenalty,
             image_base64: imageBase64,
             audio_base64: audioBase64,
+            cancel_id: cancelId,
+            ...(resolvedThreadId ? { session_id: resolvedThreadId } : {}),
             ...(useAdapter === undefined ? {} : { use_adapter: useAdapter }),
             ...(supportsReasoning
               ? reasoningStyle === "reasoning_effort"
@@ -750,7 +789,6 @@ export function createOpenAIStreamAdapter(): ChatModelAdapter {
                     const mins = useChatRuntimeStore.getState().toolCallTimeout;
                     return mins >= 9999 ? 9999 : mins * 60;
                   })(),
-                  session_id: resolvedThreadId,
                 }
               : {}),
           },
@@ -948,6 +986,7 @@ export function createOpenAIStreamAdapter(): ChatModelAdapter {
         }
         throw err;
       } finally {
+        abortSignal.removeEventListener("abort", onAbortCancel);
         runtime.setGeneratingStatus(null);
         runtime.setToolStatus(null);
         clearTimeout(warmupTimer);
diff --git a/studio/frontend/src/features/chat/types/api.ts b/studio/frontend/src/features/chat/types/api.ts
index ccfc8b2bf1..25957f4a7b 100644
--- a/studio/frontend/src/features/chat/types/api.ts
+++ b/studio/frontend/src/features/chat/types/api.ts
@@ -179,6 +179,7 @@ export interface OpenAIChatCompletionsRequest {
   max_tool_calls_per_message?: number;
   tool_call_timeout?: number;
   session_id?: string;
+  cancel_id?: string;
 }
 
 export interface OpenAIChatDelta {
diff --git a/tests/studio/test_cancel_atomicity.py b/tests/studio/test_cancel_atomicity.py
new file mode 100644
index 0000000000..a8d4839454
--- /dev/null
+++ b/tests/studio/test_cancel_atomicity.py
@@ -0,0 +1,289 @@
+"""
+TOCTOU atomicity guards for the cancel path.
+
+Structural: cancel_inference, _cancel_by_cancel_id_or_stash, and
+_TrackedCancel.__enter__ must each use a single _CANCEL_LOCK critical
+section over lookup + stash / register + consume-pending.
+
+Behavioral: parallel cancel-POST vs __enter__ must never drop a cancel.
+"""
+
+from __future__ import annotations
+
+import ast
+import random
+import threading
+from pathlib import Path
+
+
+SOURCE_PATH = (
+    Path(__file__).resolve().parents[2]
+    / "studio"
+    / "backend"
+    / "routes"
+    / "inference.py"
+)
+_SRC = SOURCE_PATH.read_text()
+_TREE = ast.parse(_SRC)
+
+
+def _find_function(name: str) -> ast.FunctionDef | ast.AsyncFunctionDef:
+    for node in ast.walk(_TREE):
+        if (
+            isinstance(node, (ast.FunctionDef, ast.AsyncFunctionDef))
+            and node.name == name
+        ):
+            return node
+    raise AssertionError(f"function {name!r} not found")
+
+
+def _find_class(name: str) -> ast.ClassDef:
+    for node in ast.walk(_TREE):
+        if isinstance(node, ast.ClassDef) and node.name == name:
+            return node
+    raise AssertionError(f"class {name!r} not found")
+
+
+def _count_with_cancel_lock_blocks(node: ast.AST) -> int:
+    n = 0
+    for sub in ast.walk(node):
+        if not isinstance(sub, ast.With):
+            continue
+        for item in sub.items:
+            ctx = item.context_expr
+            if isinstance(ctx, ast.Name) and ctx.id == "_CANCEL_LOCK":
+                n += 1
+                break
+    return n
+
+
+def test_cancel_by_cancel_id_or_stash_is_single_lock_critical_section():
+    fn = _find_function("_cancel_by_cancel_id_or_stash")
+    assert _count_with_cancel_lock_blocks(fn) == 1, (
+        "_cancel_by_cancel_id_or_stash must use exactly one `with "
+        "_CANCEL_LOCK:` block; splitting into two acquisitions reopens "
+        "the TOCTOU race with _TrackedCancel.__enter__"
+    )
+    src = ast.unparse(fn)
+    assert "_CANCEL_REGISTRY.get(cancel_id)" in src
+    assert "_PENDING_CANCELS[cancel_id]" in src
+
+
+def test_tracked_cancel_enter_registers_and_consumes_pending_under_one_lock():
+    cls = _find_class("_TrackedCancel")
+    enter = None
+    for n in cls.body:
+        if isinstance(n, ast.FunctionDef) and n.name == "__enter__":
+            enter = n
+            break
+    assert enter is not None
+    assert _count_with_cancel_lock_blocks(enter) == 1, (
+        "_TrackedCancel.__enter__ must acquire _CANCEL_LOCK exactly once. "
+        "A second acquisition for consume-pending lets a concurrent "
+        "cancel POST stash after consume sees an empty map, silently "
+        "dropping the cancel"
+    )
+    with_block = None
+    for sub in ast.walk(enter):
+        if isinstance(sub, ast.With) and any(
+            isinstance(i.context_expr, ast.Name) and i.context_expr.id == "_CANCEL_LOCK"
+            for i in sub.items
+        ):
+            with_block = sub
+            break
+    assert with_block is not None
+    block_src = "\n".join(ast.unparse(s) for s in with_block.body)
+    assert "_CANCEL_REGISTRY.setdefault" in block_src
+    assert "_PENDING_CANCELS.pop" in block_src, (
+        "__enter__ critical section must consume from _PENDING_CANCELS "
+        "inside the same lock, not a later re-acquisition"
+    )
+
+
+def test_cancel_inference_uses_atomic_helper_for_cancel_id_path():
+    fn = _find_function("cancel_inference")
+    src = ast.unparse(fn)
+    assert "_cancel_by_cancel_id_or_stash" in src
+    # The pre-fix two-step idiom must be gone.
+    assert "_remember_pending_cancel(cancel_id)" not in src, (
+        "two-step _cancel_by_keys + _remember_pending_cancel produced "
+        "the TOCTOU race and must not return"
+    )
+
+
+_WANTED = {
+    "_CANCEL_REGISTRY",
+    "_CANCEL_LOCK",
+    "_PENDING_CANCELS",
+    "_PENDING_CANCEL_TTL_S",
+    "_prune_pending",
+    "_remember_pending_cancel",
+    "_TrackedCancel",
+    "_cancel_by_keys",
+    "_cancel_by_cancel_id_or_stash",
+}
+
+
+def _load_registry_module():
+    chunks = []
+    for n in _TREE.body:
+        seg = ast.get_source_segment(_SRC, n)
+        if seg is None:
+            continue
+        if isinstance(n, (ast.FunctionDef, ast.ClassDef)) and n.name in _WANTED:
+            chunks.append(seg)
+        elif isinstance(n, ast.Assign):
+            names = [t.id for t in n.targets if isinstance(t, ast.Name)]
+            if any(name in _WANTED for name in names):
+                chunks.append(seg)
+        elif (
+            isinstance(n, ast.AnnAssign)
+            and isinstance(n.target, ast.Name)
+            and n.target.id in _WANTED
+        ):
+            chunks.append(seg)
+    mod = {}
+    exec(
+        "import threading, time\nfrom typing import Optional\n" + "\n\n".join(chunks),
+        mod,
+    )
+    return mod
+
+
+def test_parallel_cancel_vs_register_never_drops():
+    m = _load_registry_module()
+    trials = 500
+    dropped = 0
+    for i in range(trials):
+        m["_CANCEL_REGISTRY"].clear()
+        m["_PENDING_CANCELS"].clear()
+        cid = f"cid-{i}"
+        ev = threading.Event()
+        tracker = m["_TrackedCancel"](ev, cid, "thread")
+        start = threading.Event()
+
+        def do_cancel():
+            start.wait()
+            m["_cancel_by_cancel_id_or_stash"](cid)
+
+        def do_enter():
+            start.wait()
+            tracker.__enter__()
+
+        threads = [
+            threading.Thread(target = do_cancel),
+            threading.Thread(target = do_enter),
+        ]
+        random.shuffle(threads)
+        for t in threads:
+            t.start()
+        start.set()
+        for t in threads:
+            t.join(timeout = 5.0)
+            assert not t.is_alive()
+
+        if not ev.is_set():
+            dropped += 1
+        tracker.__exit__(None, None, None)
+
+    assert dropped == 0, (
+        f"TOCTOU regression: {dropped}/{trials} parallel trials silently "
+        f"dropped the cancel"
+    )
+
+
+def test_cancel_before_register_replays_atomically():
+    m = _load_registry_module()
+    cid = "early-cid"
+    ev = threading.Event()
+    tracker = m["_TrackedCancel"](ev, cid, "thread-x")
+
+    assert m["_cancel_by_cancel_id_or_stash"](cid) == 0
+    assert cid in m["_PENDING_CANCELS"]
+
+    tracker.__enter__()
+    assert ev.is_set()
+    assert cid not in m["_PENDING_CANCELS"]
+    tracker.__exit__(None, None, None)
+
+
+def test_cancel_after_register_signals_without_stash():
+    m = _load_registry_module()
+    cid = "post-cid"
+    ev = threading.Event()
+    tracker = m["_TrackedCancel"](ev, cid, "thread-y")
+    tracker.__enter__()
+
+    assert m["_cancel_by_cancel_id_or_stash"](cid) == 1
+    assert ev.is_set()
+    assert cid not in m["_PENDING_CANCELS"]
+    tracker.__exit__(None, None, None)
+
+
+def test_cancel_by_keys_tolerates_empty_and_falsy_keys():
+    m = _load_registry_module()
+    m["_CANCEL_REGISTRY"].clear()
+    m["_PENDING_CANCELS"].clear()
+    assert m["_cancel_by_keys"]([]) == 0
+    assert m["_cancel_by_keys"](["", None, "unknown"]) == 0
+    # Non-stashing fallback must never leak into _PENDING_CANCELS.
+    assert m["_PENDING_CANCELS"] == {}
+
+
+def test_cancel_by_keys_fans_out_to_all_streams_on_same_session():
+    # Compare mode and other flows launch concurrent streams under a
+    # shared session_id; a single session cancel POST must hit all of them.
+    m = _load_registry_module()
+    m["_CANCEL_REGISTRY"].clear()
+    m["_PENDING_CANCELS"].clear()
+    session = "shared-thread"
+    ev_a = threading.Event()
+    ev_b = threading.Event()
+    tracker_a = m["_TrackedCancel"](ev_a, "cancel-a", session, "chatcmpl-a")
+    tracker_b = m["_TrackedCancel"](ev_b, "cancel-b", session, "chatcmpl-b")
+    tracker_a.__enter__()
+    tracker_b.__enter__()
+    try:
+        assert m["_cancel_by_keys"]([session]) == 2
+        assert ev_a.is_set() and ev_b.is_set()
+    finally:
+        tracker_a.__exit__(None, None, None)
+        tracker_b.__exit__(None, None, None)
+    assert session not in m["_CANCEL_REGISTRY"]
+
+
+def test_cancel_by_cancel_id_is_exclusive_to_single_run():
+    # cancel_id is per-run unique; cancelling run A must not touch run B
+    # even when both share a session_id.
+    m = _load_registry_module()
+    m["_CANCEL_REGISTRY"].clear()
+    m["_PENDING_CANCELS"].clear()
+    session = "shared-thread-2"
+    ev_a = threading.Event()
+    ev_b = threading.Event()
+    tracker_a = m["_TrackedCancel"](ev_a, "cancel-only-a", session, "chatcmpl-a")
+    tracker_b = m["_TrackedCancel"](ev_b, "cancel-only-b", session, "chatcmpl-b")
+    tracker_a.__enter__()
+    tracker_b.__enter__()
+    try:
+        assert m["_cancel_by_cancel_id_or_stash"]("cancel-only-a") == 1
+        assert ev_a.is_set()
+        assert not ev_b.is_set()
+    finally:
+        tracker_a.__exit__(None, None, None)
+        tracker_b.__exit__(None, None, None)
+
+
+def test_tracked_cancel_exit_is_idempotent():
+    # Outer except BaseException + the generator's finally may both call
+    # __exit__ under certain race combos; must not raise.
+    m = _load_registry_module()
+    m["_CANCEL_REGISTRY"].clear()
+    m["_PENDING_CANCELS"].clear()
+    ev = threading.Event()
+    tracker = m["_TrackedCancel"](ev, "cid", "sess", "chatcmpl-x")
+    tracker.__enter__()
+    tracker.__exit__(None, None, None)
+    tracker.__exit__(None, None, None)
+    tracker.__exit__(None, None, None)
+    assert not m["_CANCEL_REGISTRY"]
diff --git a/tests/studio/test_cancel_id_wiring.py b/tests/studio/test_cancel_id_wiring.py
new file mode 100644
index 0000000000..5fd76ded9b
--- /dev/null
+++ b/tests/studio/test_cancel_id_wiring.py
@@ -0,0 +1,169 @@
+"""
+Wiring tests for the per-run cancel_id field.
+
+A chat-thread-scoped session_id is not safe as a cancel key because a
+late stop POST can match a subsequent run on the same thread. The fix
+adds cancel_id (a fresh UUID per generation) that is sent both in the
+completion payload and in the /api/inference/cancel body.
+
+Verifies:
+  - ChatCompletionRequest exposes an Optional[str] `cancel_id` field.
+  - /api/inference/cancel accepts `cancel_id` as the first-preferred key.
+  - OpenAIChatCompletionsRequest (frontend type) includes cancel_id.
+  - chat-adapter.ts generates a per-run cancelId (crypto.randomUUID
+    with a Math.random fallback), sends it in the completion payload,
+    and includes it in the /inference/cancel body on abort.
+"""
+
+from __future__ import annotations
+
+import ast
+import re
+from pathlib import Path
+
+
+WORKSPACE = Path(__file__).resolve().parents[2]
+MODELS_SRC = (WORKSPACE / "studio/backend/models/inference.py").read_text()
+ROUTES_SRC = (WORKSPACE / "studio/backend/routes/inference.py").read_text()
+ADAPTER_SRC = (
+    WORKSPACE / "studio/frontend/src/features/chat/api/chat-adapter.ts"
+).read_text()
+API_TYPES_SRC = (
+    WORKSPACE / "studio/frontend/src/features/chat/types/api.ts"
+).read_text()
+
+
+def _find_class(tree: ast.AST, name: str) -> ast.ClassDef | None:
+    for node in ast.walk(tree):
+        if isinstance(node, ast.ClassDef) and node.name == name:
+            return node
+    return None
+
+
+def test_chat_completion_request_has_cancel_id_field():
+    tree = ast.parse(MODELS_SRC)
+    cls = _find_class(tree, "ChatCompletionRequest")
+    assert cls is not None
+    fields = {
+        n.target.id
+        for n in cls.body
+        if isinstance(n, ast.AnnAssign) and isinstance(n.target, ast.Name)
+    }
+    assert "cancel_id" in fields, (
+        "ChatCompletionRequest must expose a cancel_id field for per-run "
+        "cancellation routing"
+    )
+
+
+def test_cancel_route_matches_cancel_id_exclusively_when_present():
+    # A stale cancel POST carrying cancel_id AND session_id must not
+    # cancel a later run on the same thread via the shared session_id.
+    # Enforce this by requiring the handler to early-return through an
+    # exclusive-cancel_id path -- either an atomic helper or a keys
+    # list containing ONLY cancel_id (never session_id).
+    for node in ast.walk(ast.parse(ROUTES_SRC)):
+        if isinstance(node, ast.AsyncFunctionDef) and node.name == "cancel_inference":
+            break
+    else:
+        raise AssertionError("cancel_inference handler missing")
+
+    cancel_id_exclusive_branch = False
+    for sub in ast.walk(node):
+        if not isinstance(sub, ast.If):
+            continue
+        test_src = ast.unparse(sub.test)
+        if "cancel_id" not in test_src or "isinstance" not in test_src:
+            continue
+        branch_src = "\n".join(ast.unparse(s) for s in sub.body)
+        before_return = branch_src.split("return", 1)[0]
+        matches_cancel_id_only = (
+            "_cancel_by_cancel_id_or_stash(cancel_id)" in branch_src
+            or "_cancel_by_keys([cancel_id])" in branch_src
+        )
+        if matches_cancel_id_only and "session_id" not in before_return:
+            cancel_id_exclusive_branch = True
+            break
+    assert cancel_id_exclusive_branch, (
+        "cancel_inference must early-return with an exclusive cancel_id "
+        "match when a cancel_id is supplied, so a stale stop POST "
+        "cannot cancel a later run on the same thread via session_id"
+    )
+
+
+def test_cancel_route_falls_back_to_session_or_completion_when_no_cancel_id():
+    for node in ast.walk(ast.parse(ROUTES_SRC)):
+        if isinstance(node, ast.AsyncFunctionDef) and node.name == "cancel_inference":
+            break
+    else:
+        raise AssertionError("cancel_inference handler missing")
+
+    src = ast.unparse(node)
+    assert "session_id" in src and "completion_id" in src, (
+        "cancel_inference must still accept session_id / completion_id as "
+        "fallback keys when cancel_id is absent"
+    )
+
+
+def test_frontend_request_type_has_cancel_id():
+    assert re.search(
+        r"cancel_id\?\s*:\s*string\s*;", API_TYPES_SRC
+    ), "OpenAIChatCompletionsRequest must expose an optional cancel_id"
+
+
+def test_chat_adapter_generates_cancel_id_per_run():
+    m = re.search(
+        r"const\s+cancelId\s*=\s*([^;]+);",
+        ADAPTER_SRC,
+    )
+    assert m, "chat-adapter.ts must declare a per-run `cancelId` constant"
+    rhs = m.group(1)
+    assert (
+        "randomUUID" in rhs
+    ), "cancelId should prefer crypto.randomUUID() for uniqueness"
+
+
+def test_chat_adapter_sends_cancel_id_in_completion_payload():
+    assert "cancel_id: cancelId" in ADAPTER_SRC, (
+        "chat-adapter.ts must include cancel_id in the streamChatCompletions "
+        "request payload so the backend registers under that key"
+    )
+
+
+def test_chat_adapter_sends_cancel_id_in_abort_cancel_post():
+    m = re.search(
+        r"const\s+onAbortCancel\s*=\s*\(\)\s*=>\s*\{(.*?)\};",
+        ADAPTER_SRC,
+        flags = re.DOTALL,
+    )
+    assert m, "onAbortCancel arrow function missing"
+    body = m.group(1)
+    assert re.search(r"cancel_id\s*:\s*cancelId", body), (
+        "onAbortCancel must include cancel_id in the /inference/cancel body "
+        "so a stop POST matches the specific run, not the whole thread"
+    )
+
+
+def test_abort_cancel_post_uses_plain_fetch_with_manual_auth_header():
+    # authFetch redirects to login on 401, which would kick the user to
+    # the login page mid-stop if the access token expired during a long
+    # stream. Use plain fetch + manual Authorization header for a
+    # best-effort cancel that never triggers the refresh/redirect flow.
+    start = ADAPTER_SRC.find("const onAbortCancel")
+    assert start >= 0, "onAbortCancel handler missing"
+    rest = ADAPTER_SRC[start:]
+    end = rest.find("\n      try {")
+    body = rest if end < 0 else rest[:end]
+    assert "/api/inference/cancel" in body
+    assert "authFetch(" not in body, (
+        "onAbortCancel must NOT call authFetch; a 401 from it would "
+        "redirect the user to the login page during a stop click"
+    )
+    assert "fetch(" in body, "onAbortCancel must use plain fetch(...)"
+    assert "getAuthToken" in body, (
+        "onAbortCancel must read the bearer token via getAuthToken() "
+        "rather than relying on authFetch's 401 flow"
+    )
+    assert "Authorization" in body
+    assert (
+        "keepalive: true" in body
+    ), "keepalive is required so the fetch survives page unload during stop"
diff --git a/tests/studio/test_llama_cpp_wall_clock_cap.py b/tests/studio/test_llama_cpp_wall_clock_cap.py
new file mode 100644
index 0000000000..671abea823
--- /dev/null
+++ b/tests/studio/test_llama_cpp_wall_clock_cap.py
@@ -0,0 +1,123 @@
+"""
+Tests for the llama-server wall-clock cap (t_max_predict_ms).
+
+The UI always sends max_tokens = context_length, so gating
+t_max_predict_ms on `max_tokens is None` makes the safety net dead
+code. The fix applies the wall-clock cap unconditionally on all three
+streaming payload sites and raises the default to 10 minutes so slow
+CPU / macOS / Windows installs are not cut off mid-generation.
+
+Verifies:
+  - t_max_predict_ms is assigned unconditionally at the three
+    payload-builder sites (not inside an `if max_tokens is None` else
+    branch).
+  - _DEFAULT_T_MAX_PREDICT_MS is at least 10 minutes (previously
+    120_000).
+  - The default max_tokens path still applies _DEFAULT_MAX_TOKENS.
+  - The three payload variable names (payload x2, stream_payload x1)
+    each get both `max_tokens` and `t_max_predict_ms`.
+"""
+
+from __future__ import annotations
+
+import ast
+from pathlib import Path
+
+
+SOURCE_PATH = (
+    Path(__file__).resolve().parents[2]
+    / "studio"
+    / "backend"
+    / "core"
+    / "inference"
+    / "llama_cpp.py"
+)
+SRC = SOURCE_PATH.read_text()
+TREE = ast.parse(SRC)
+
+
+def _is_subscript_assign(stmt: ast.stmt, target_name: str, key: str) -> bool:
+    if not isinstance(stmt, ast.Assign) or len(stmt.targets) != 1:
+        return False
+    t = stmt.targets[0]
+    if not isinstance(t, ast.Subscript):
+        return False
+    if not (isinstance(t.value, ast.Name) and t.value.id == target_name):
+        return False
+    slc = t.slice
+    return isinstance(slc, ast.Constant) and slc.value == key
+
+
+def _collect_assignments(tree, target_name, key):
+    """Return list of (node, stack_of_enclosing_ifs) for each match."""
+    hits = []
+
+    def visit(node, stack):
+        if _is_subscript_assign(node, target_name, key):
+            hits.append((node, stack))
+        for child in ast.iter_child_nodes(node):
+            if isinstance(child, ast.If):
+                for sub in child.body:
+                    visit(sub, stack + [(child, "body")])
+                for sub in child.orelse:
+                    visit(sub, stack + [(child, "orelse")])
+            else:
+                visit(child, stack)
+
+    visit(tree, [])
+    return hits
+
+
+def test_default_t_max_predict_ms_is_at_least_ten_minutes():
+    for node in TREE.body:
+        if isinstance(node, ast.Assign) and len(node.targets) == 1:
+            t = node.targets[0]
+            if isinstance(t, ast.Name) and t.id == "_DEFAULT_T_MAX_PREDICT_MS":
+                value = node.value
+                assert isinstance(value, ast.Constant)
+                assert value.value >= 600_000, (
+                    f"_DEFAULT_T_MAX_PREDICT_MS must be >= 10 minutes "
+                    f"(600_000 ms) to avoid cutting off slow-CPU generations; "
+                    f"got {value.value}"
+                )
+                return
+    raise AssertionError("_DEFAULT_T_MAX_PREDICT_MS constant missing")
+
+
+def test_t_max_predict_ms_set_unconditionally_at_three_sites():
+    hits_payload = _collect_assignments(TREE, "payload", "t_max_predict_ms")
+    hits_stream = _collect_assignments(TREE, "stream_payload", "t_max_predict_ms")
+    total = len(hits_payload) + len(hits_stream)
+    assert total == 3, (
+        f"expected 3 total t_max_predict_ms assignments "
+        f"(payload x2 + stream_payload x1), got {total}"
+    )
+    for node, stack in hits_payload + hits_stream:
+        for parent_if, branch in stack:
+            # The assignment must not be gated by a test that checks
+            # `max_tokens is None` (which would make it dead code for
+            # the UI path where max_tokens is always set).
+            test_src = ast.unparse(parent_if.test)
+            assert "max_tokens" not in test_src, (
+                f"t_max_predict_ms at line {node.lineno} is nested under "
+                f"`if {test_src}:` -- it must be applied unconditionally so "
+                f"the wall-clock cap is not dead code for callers that set "
+                f"max_tokens"
+            )
+
+
+def test_max_tokens_default_cap_still_applied():
+    # _DEFAULT_MAX_TOKENS must still kick in when caller passes None.
+    # We check the conditional expression `max_tokens if max_tokens is not
+    # None else _DEFAULT_MAX_TOKENS` appears at each site.
+    matches = 0
+    for node in ast.walk(TREE):
+        if not isinstance(node, ast.IfExp):
+            continue
+        src = ast.unparse(node)
+        if "max_tokens" in src and "_DEFAULT_MAX_TOKENS" in src:
+            matches += 1
+    assert matches >= 3, (
+        f"expected >=3 `max_tokens if max_tokens is not None else "
+        f"_DEFAULT_MAX_TOKENS` expressions; got {matches}"
+    )
diff --git a/tests/studio/test_stream_cancel_registration_timing.py b/tests/studio/test_stream_cancel_registration_timing.py
new file mode 100644
index 0000000000..40ec3d6e1f
--- /dev/null
+++ b/tests/studio/test_stream_cancel_registration_timing.py
@@ -0,0 +1,718 @@
+"""
+Tests that the cancel tracker is registered BEFORE StreamingResponse is
+returned, and that cleanup runs via a `finally` block inside each
+async generator.
+
+The zombie-generation scenario is: user clicks Stop during prefill /
+warmup / proxy buffering, before the first SSE chunk. If _tracker
+__enter__ lives inside the async generator body, the registry is empty
+at the moment /api/inference/cancel lands -- so cancel returns 0 and
+the decode runs to completion.
+
+The fix moves _tracker = _TrackedCancel(...) and _tracker.__enter__()
+to the synchronous body of openai_chat_completions (before the
+StreamingResponse is returned) and places _tracker.__exit__ inside
+each generator's `finally` block. Using a generator `finally` (rather
+than a Starlette BackgroundTask) guarantees cleanup on every
+termination path -- normal exhaustion, CancelledError from
+ClientDisconnect, and OSError / BrokenPipeError during send() --
+because Starlette skips `background` callbacks when stream_response
+raises.
+
+Structural verifies:
+  - No `async def ...:` body contains `_tracker.__enter__()` in
+    routes/inference.py (registration moved to sync body).
+  - Each of the four async generators (gguf_tool_stream,
+    gguf_stream_chunks, stream_chunks, audio_input_stream) contains
+    `_tracker.__exit__(None, None, None)` inside a try/finally block.
+  - No StreamingResponse in openai_chat_completions passes
+    `background=` (cleanup now lives in the generator finally).
+
+Behavioral verifies (extracting `_TrackedCancel` from source and
+exercising the actual runtime semantics):
+  - `finally: _tracker.__exit__(...)` runs on normal completion,
+    mid-stream exception (OSError / BrokenPipeError from send()),
+    and aclose() from Starlette ClientDisconnect.
+  - A pre-set cancel_event (from `_TrackedCancel.__enter__` replaying
+    a pending cancel POST) lets the GGUF while-loop break cleanly
+    and emit final_chunk + [DONE] instead of propagating
+    `GeneratorExit` out of `_stream_with_retry` into the async
+    generator's `except Exception` (which would not catch it).
+"""
+
+from __future__ import annotations
+
+import ast
+import asyncio
+import threading
+import time
+from pathlib import Path
+
+
+SOURCE_PATH = (
+    Path(__file__).resolve().parents[2]
+    / "studio"
+    / "backend"
+    / "routes"
+    / "inference.py"
+)
+SRC = SOURCE_PATH.read_text()
+_TREE = ast.parse(SRC)
+
+
+# ── Structural (AST) helpers ─────────────────────────────────
+
+
+def _collect_async_functions(tree: ast.AST):
+    return [n for n in ast.walk(tree) if isinstance(n, ast.AsyncFunctionDef)]
+
+
+def _has_tracker_enter_call(node: ast.AST) -> bool:
+    for sub in ast.walk(node):
+        if not isinstance(sub, ast.Call):
+            continue
+        fn = sub.func
+        if (
+            isinstance(fn, ast.Attribute)
+            and fn.attr == "__enter__"
+            and isinstance(fn.value, ast.Name)
+            and fn.value.id.startswith("_tracker")
+        ):
+            return True
+    return False
+
+
+def _finalbody_has_tracker_exit(finalbody) -> bool:
+    for stmt in finalbody:
+        if not isinstance(stmt, ast.Expr):
+            continue
+        call = stmt.value
+        if not (isinstance(call, ast.Call) and isinstance(call.func, ast.Attribute)):
+            continue
+        fn = call.func
+        if (
+            fn.attr == "__exit__"
+            and isinstance(fn.value, ast.Name)
+            and fn.value.id.startswith("_tracker")
+        ):
+            return True
+    return False
+
+
+# ── Structural tests ─────────────────────────────────────────
+
+
+def test_no_tracker_enter_inside_async_generators():
+    offenders = []
+    for fn in _collect_async_functions(_TREE):
+        if fn.name in {
+            "gguf_tool_stream",
+            "gguf_stream_chunks",
+            "stream_chunks",
+            "audio_input_stream",
+        }:
+            if _has_tracker_enter_call(fn):
+                offenders.append(fn.name)
+    assert not offenders, (
+        f"Cancel tracker registration must live OUTSIDE the async generator "
+        f"body so a stop POST can find the registry entry before the first "
+        f"SSE chunk. Offending generators: {offenders}"
+    )
+
+
+def test_tracker_enter_exists_in_sync_body_of_chat_completions():
+    top = None
+    for n in ast.walk(_TREE):
+        if isinstance(n, ast.AsyncFunctionDef) and n.name == "openai_chat_completions":
+            top = n
+            break
+    assert top is not None, "openai_chat_completions handler missing"
+    count = 0
+    for sub in ast.walk(top):
+        if not isinstance(sub, ast.Call):
+            continue
+        fn = sub.func
+        if (
+            isinstance(fn, ast.Attribute)
+            and fn.attr == "__enter__"
+            and isinstance(fn.value, ast.Name)
+            and fn.value.id.startswith("_tracker")
+        ):
+            count += 1
+    assert count >= 3, (
+        f"expected >=3 _tracker.__enter__() calls in openai_chat_completions "
+        f"(one per streaming path), got {count}"
+    )
+
+
+def test_async_generators_cleanup_tracker_in_finally():
+    required = {
+        "gguf_tool_stream",
+        "gguf_stream_chunks",
+        "stream_chunks",
+        "audio_input_stream",
+    }
+    found: set[str] = set()
+    for fn in [n for n in ast.walk(_TREE) if isinstance(n, ast.AsyncFunctionDef)]:
+        if fn.name not in required:
+            continue
+        for sub in ast.walk(fn):
+            if isinstance(sub, ast.Try) and sub.finalbody:
+                if _finalbody_has_tracker_exit(sub.finalbody):
+                    found.add(fn.name)
+                    break
+    missing = required - found
+    assert not missing, (
+        f"Cleanup must run via `finally: _tracker.__exit__(None, None, None)` "
+        f"inside each streaming generator so ClientDisconnect / OSError paths "
+        f"also release registry entries (Starlette skips `background` callbacks "
+        f"when stream_response raises). Missing in: {sorted(missing)}"
+    )
+
+
+def test_streaming_responses_have_no_background_task():
+    top = None
+    for n in ast.walk(_TREE):
+        if isinstance(n, ast.AsyncFunctionDef) and n.name == "openai_chat_completions":
+            top = n
+            break
+    assert top is not None
+    for sub in ast.walk(top):
+        if not (isinstance(sub, ast.Call) and isinstance(sub.func, ast.Name)):
+            continue
+        if sub.func.id != "StreamingResponse":
+            continue
+        kwargs = {kw.arg for kw in sub.keywords if kw.arg}
+        assert "background" not in kwargs, (
+            "StreamingResponse in openai_chat_completions must not pass "
+            "`background=` -- cleanup now lives in the generator's finally "
+            "block; a BackgroundTask would be skipped on abrupt disconnect"
+        )
+
+
+# ── Behavioral helpers ───────────────────────────────────────
+
+_WANTED = {
+    "_CANCEL_REGISTRY",
+    "_CANCEL_LOCK",
+    "_PENDING_CANCELS",
+    "_PENDING_CANCEL_TTL_S",
+    "_prune_pending",
+    "_TrackedCancel",
+    "_cancel_by_keys",
+    "_cancel_by_cancel_id_or_stash",
+}
+
+
+def _load_registry_module():
+    chunks = []
+    for n in _TREE.body:
+        seg = ast.get_source_segment(SRC, n)
+        if seg is None:
+            continue
+        if isinstance(n, (ast.FunctionDef, ast.ClassDef)) and n.name in _WANTED:
+            chunks.append(seg)
+        elif isinstance(n, ast.Assign):
+            names = [t.id for t in n.targets if isinstance(t, ast.Name)]
+            if any(name in _WANTED for name in names):
+                chunks.append(seg)
+        elif (
+            isinstance(n, ast.AnnAssign)
+            and isinstance(n.target, ast.Name)
+            and n.target.id in _WANTED
+        ):
+            chunks.append(seg)
+    mod = {}
+    exec("import threading, time\n" + "\n\n".join(chunks), mod)
+    return mod
+
+
+def _make_stream(tracker, raise_exc):
+    async def gen():
+        try:
+            try:
+                yield "data: first\n\n"
+                if raise_exc is not None:
+                    raise raise_exc
+                yield "data: [DONE]\n\n"
+            except asyncio.CancelledError:
+                raise
+            except Exception:
+                yield "data: error\n\n"
+            finally:
+                tracker.__exit__(None, None, None)
+        except BaseException:
+            raise
+
+    return gen()
+
+
+async def _consume(agen):
+    out = []
+    try:
+        async for ch in agen:
+            out.append(ch)
+    except BaseException as e:
+        out.append(type(e).__name__)
+    return out
+
+
+def _llama_stub_raises_on_preset_cancel(cancel_event):
+    # Reproduces llama_cpp.py _stream_with_retry:2240 `raise GeneratorExit`
+    # when cancel_event is already set at entry.
+    if cancel_event.is_set():
+        raise GeneratorExit
+    yield "cumulative-1"
+    yield "cumulative-2"
+
+
+async def _post_fix_gguf_loop(cancel_event):
+    yield "first_chunk"
+    gen = _llama_stub_raises_on_preset_cancel(cancel_event)
+    sentinel = object()
+    while True:
+        if cancel_event.is_set():
+            break
+        cumulative = await asyncio.to_thread(next, gen, sentinel)
+        if cumulative is sentinel:
+            break
+        yield cumulative
+    yield "final_chunk"
+    yield "[DONE]"
+
+
+# ── Behavioral tests ─────────────────────────────────────────
+
+
+def test_finally_cleanup_on_normal_completion():
+    m = _load_registry_module()
+    m["_CANCEL_REGISTRY"].clear()
+    ev = threading.Event()
+    tr = m["_TrackedCancel"](ev, "cid-ok", "sid-ok")
+    tr.__enter__()
+    assert "cid-ok" in m["_CANCEL_REGISTRY"]
+    chunks = asyncio.run(_consume(_make_stream(tr, None)))
+    assert chunks == ["data: first\n\n", "data: [DONE]\n\n"]
+    assert "cid-ok" not in m["_CANCEL_REGISTRY"]
+    assert "sid-ok" not in m["_CANCEL_REGISTRY"]
+
+
+def test_finally_cleanup_on_mid_stream_exception():
+    # Simulates OSError / BrokenPipeError from Starlette send() mid-stream --
+    # the exact case where pre-fix `background = BackgroundTask(...)` was
+    # skipped and leaked the registry entry.
+    m = _load_registry_module()
+    m["_CANCEL_REGISTRY"].clear()
+    ev = threading.Event()
+    tr = m["_TrackedCancel"](ev, "cid-err", "sid-err")
+    tr.__enter__()
+    assert "cid-err" in m["_CANCEL_REGISTRY"]
+    asyncio.run(_consume(_make_stream(tr, OSError("disconnect"))))
+    assert "cid-err" not in m["_CANCEL_REGISTRY"]
+    assert "sid-err" not in m["_CANCEL_REGISTRY"]
+
+
+def test_finally_cleanup_on_aclose():
+    # Starlette calls aclose() on the async generator when the client
+    # disconnects mid-stream. The generator's finally block must run.
+    m = _load_registry_module()
+    m["_CANCEL_REGISTRY"].clear()
+    ev = threading.Event()
+    tr = m["_TrackedCancel"](ev, "cid-abort", "sid-abort")
+    tr.__enter__()
+    assert "cid-abort" in m["_CANCEL_REGISTRY"]
+
+    async def run():
+        gen = _make_stream(tr, None)
+        it = gen.__aiter__()
+        await it.__anext__()
+        await gen.aclose()
+
+    asyncio.run(run())
+    assert "cid-abort" not in m["_CANCEL_REGISTRY"]
+    assert "sid-abort" not in m["_CANCEL_REGISTRY"]
+
+
+def test_preset_cancel_event_exits_cleanly_with_done():
+    # Pending-replay: POST /cancel arrived before the stream registered,
+    # was stashed, then consumed by _TrackedCancel.__enter__ which set
+    # cancel_event. The generator must break out of the loop cleanly
+    # and emit final_chunk + [DONE] rather than calling next(gen) and
+    # propagating `GeneratorExit` out of the GGUF stream wrapper.
+    ev = threading.Event()
+    ev.set()
+    chunks = asyncio.run(_consume(_post_fix_gguf_loop(ev)))
+    assert "first_chunk" in chunks
+    assert "final_chunk" in chunks
+    assert "[DONE]" in chunks
+    assert "GeneratorExit" not in chunks
+    assert "cumulative-1" not in chunks
+    assert "cumulative-2" not in chunks
+
+
+def test_normal_path_streams_all_tokens():
+    # Regression: the top-of-loop cancel_event check must not short-circuit
+    # when cancel_event is unset.
+    ev = threading.Event()
+    chunks = asyncio.run(_consume(_post_fix_gguf_loop(ev)))
+    assert chunks == [
+        "first_chunk",
+        "cumulative-1",
+        "cumulative-2",
+        "final_chunk",
+        "[DONE]",
+    ]
+
+
+def test_cancel_during_streaming_stops_iteration_promptly():
+    # Setting cancel_event between yields breaks out on the next iteration
+    # rather than draining the stub generator.
+    ev = threading.Event()
+
+    async def _run():
+        gen = _post_fix_gguf_loop(ev)
+        seen = []
+        async for ch in gen:
+            seen.append(ch)
+            if ch == "cumulative-1":
+                ev.set()
+        return seen
+
+    seen = asyncio.run(_run())
+    assert "first_chunk" in seen
+    assert "cumulative-1" in seen
+    assert "cumulative-2" not in seen
+    assert "final_chunk" in seen
+    assert "[DONE]" in seen
+
+
+# ── Cancel-event responsiveness in the streaming loops ───────
+
+
+def _loop_has_cancel_event_check(fn) -> bool:
+    # An `if cancel_event.is_set():` statement anywhere inside a
+    # `while`/`for` loop body is sufficient -- without it, a cancel POST
+    # cannot interrupt the loop because Colab-style proxies do not
+    # propagate request.is_disconnected().
+    for sub in ast.walk(fn):
+        if not isinstance(sub, (ast.While, ast.For, ast.AsyncFor)):
+            continue
+        for stmt in ast.walk(sub):
+            if not isinstance(stmt, ast.If):
+                continue
+            t = stmt.test
+            if (
+                isinstance(t, ast.Call)
+                and isinstance(t.func, ast.Attribute)
+                and t.func.attr == "is_set"
+                and isinstance(t.func.value, ast.Name)
+                and t.func.value.id == "cancel_event"
+            ):
+                return True
+    return False
+
+
+def test_streaming_generators_check_cancel_event_in_loop():
+    required = {
+        "gguf_tool_stream",
+        "gguf_stream_chunks",
+        "stream_chunks",
+        "audio_input_stream",
+    }
+    missing = []
+    for fn in [n for n in ast.walk(_TREE) if isinstance(n, ast.AsyncFunctionDef)]:
+        if fn.name not in required:
+            continue
+        if not _loop_has_cancel_event_check(fn):
+            missing.append(fn.name)
+    assert not missing, (
+        f"Each streaming generator must check `cancel_event.is_set()` inside "
+        f"its main loop so `POST /api/inference/cancel` can interrupt the "
+        f"stream through proxies that do not forward fetch aborts. "
+        f"Missing in: {sorted(missing)}"
+    )
+
+
+def test_audio_input_stream_offloads_blocking_next_to_thread():
+    # Guards against regression back to `for chunk_text in
+    # audio_input_generate():` -- which blocks the event loop on each
+    # whisper chunk and prevents POST /api/inference/cancel from being
+    # serviced until the chunk yields.
+    audio = None
+    for fn in ast.walk(_TREE):
+        if isinstance(fn, ast.AsyncFunctionDef) and fn.name == "audio_input_stream":
+            audio = fn
+            break
+    assert audio is not None, "audio_input_stream generator missing"
+
+    for sub in ast.walk(audio):
+        if isinstance(sub, (ast.For, ast.AsyncFor)):
+            it_src = ast.unparse(sub.iter)
+            assert "audio_input_generate" not in it_src, (
+                "audio_input_stream must not iterate audio_input_generate() "
+                "directly -- that blocks the event loop. Use "
+                "`await asyncio.to_thread(next, gen, _DONE)` inside a "
+                "`while True` loop instead"
+            )
+
+    found_to_thread_next = False
+    for sub in ast.walk(audio):
+        if not isinstance(sub, ast.Call):
+            continue
+        fn_expr = sub.func
+        if not (
+            isinstance(fn_expr, ast.Attribute)
+            and fn_expr.attr == "to_thread"
+            and isinstance(fn_expr.value, ast.Name)
+            and fn_expr.value.id == "asyncio"
+        ):
+            continue
+        if sub.args and isinstance(sub.args[0], ast.Name) and sub.args[0].id == "next":
+            found_to_thread_next = True
+            break
+    assert found_to_thread_next, (
+        "audio_input_stream must call `asyncio.to_thread(next, gen, ...)` "
+        "to keep the event loop free while whisper yields the next chunk"
+    )
+
+
+def test_stream_chunks_cancel_branch_resets_backend_state():
+    # The Unsloth path's cancel branch must flush GPU / KV-cache state
+    # via `backend.reset_generation_state()` -- the orchestrator's
+    # internal cancel path does not do this, so a cancel-via-POST that
+    # only broke the loop would leave the subprocess in a dirty state
+    # for the next request.
+    fn = None
+    top = None
+    for n in ast.walk(_TREE):
+        if isinstance(n, ast.AsyncFunctionDef) and n.name == "openai_chat_completions":
+            top = n
+            break
+    assert top is not None
+    for n in ast.walk(top):
+        if isinstance(n, ast.AsyncFunctionDef) and n.name == "stream_chunks":
+            fn = n
+            break
+    assert fn is not None, "stream_chunks generator missing"
+
+    for sub in ast.walk(fn):
+        if not isinstance(sub, ast.If):
+            continue
+        t = sub.test
+        if not (
+            isinstance(t, ast.Call)
+            and isinstance(t.func, ast.Attribute)
+            and t.func.attr == "is_set"
+            and isinstance(t.func.value, ast.Name)
+            and t.func.value.id == "cancel_event"
+        ):
+            continue
+        body_src = "\n".join(ast.unparse(s) for s in sub.body)
+        if "backend.reset_generation_state()" in body_src:
+            return
+    raise AssertionError(
+        "stream_chunks `if cancel_event.is_set():` branch must call "
+        "backend.reset_generation_state() -- matches the existing "
+        "request.is_disconnected() / CancelledError cleanup paths and "
+        "prevents KV-cache drift after cancel-via-POST"
+    )
+
+
+# ── Behavioral simulations for the iter-1 fixes ──────────────
+
+
+def test_unsloth_stream_loop_breaks_on_external_cancel_event():
+    cancel_event = threading.Event()
+    reset_calls = [0]
+
+    class _Backend:
+        def reset_generation_state(self):
+            reset_calls[0] += 1
+
+    backend = _Backend()
+
+    def _generate():
+        for i in range(200):
+            time.sleep(0.005)
+            yield f"cum-{i}"
+
+    async def _loop():
+        _DONE = object()
+        loop = asyncio.get_event_loop()
+        gen = _generate()
+        seen = []
+        while True:
+            if cancel_event.is_set():
+                backend.reset_generation_state()
+                break
+            cumulative = await loop.run_in_executor(None, next, gen, _DONE)
+            if cumulative is _DONE:
+                break
+            seen.append(cumulative)
+        return seen
+
+    async def _fire():
+        await asyncio.sleep(0.05)
+        cancel_event.set()
+
+    async def _main():
+        return await asyncio.gather(_loop(), _fire())
+
+    seen, _ = asyncio.run(_main())
+    assert (
+        len(seen) < 200
+    ), f"loop must not drain the generator after cancel; got {len(seen)} tokens"
+    assert reset_calls[0] == 1, (
+        f"backend.reset_generation_state() must be called exactly once on "
+        f"cancel-via-POST, got {reset_calls[0]}"
+    )
+
+
+def test_audio_stream_stays_responsive_under_blocking_next():
+    # Regression guard: replace the post-fix loop with the pre-fix
+    # `for chunk in audio_input_generate()` pattern and assert it blocks
+    # the event loop; then confirm the post-fix pattern exits promptly.
+    cancel_event = threading.Event()
+
+    def _audio_gen():
+        for i in range(8):
+            time.sleep(0.15)
+            yield f"chunk-{i}"
+
+    async def _prefix_loop():
+        seen = []
+        for chunk_text in _audio_gen():
+            if cancel_event.is_set():
+                break
+            seen.append(chunk_text)
+        return seen
+
+    async def _postfix_loop():
+        _DONE = object()
+        gen = _audio_gen()
+        seen = []
+        while True:
+            if cancel_event.is_set():
+                break
+            chunk_text = await asyncio.to_thread(next, gen, _DONE)
+            if chunk_text is _DONE:
+                break
+            seen.append(chunk_text)
+        return seen
+
+    async def _fire_early():
+        await asyncio.sleep(0.05)
+        cancel_event.set()
+
+    async def _run(loop_coro):
+        return await asyncio.gather(loop_coro, _fire_early())
+
+    cancel_event.clear()
+    t0 = time.monotonic()
+    prefix_seen, _ = asyncio.run(_run(_prefix_loop()))
+    prefix_elapsed = time.monotonic() - t0
+    assert prefix_elapsed >= 0.13, (
+        f"pre-fix pattern should block event loop for >=1 chunk time "
+        f"(~150ms); got {prefix_elapsed:.3f}s, {len(prefix_seen)} chunks"
+    )
+
+    cancel_event.clear()
+    t0 = time.monotonic()
+    postfix_seen, _ = asyncio.run(_run(_postfix_loop()))
+    postfix_elapsed = time.monotonic() - t0
+    assert postfix_elapsed < prefix_elapsed, (
+        f"post-fix pattern must exit faster than pre-fix (blocking) "
+        f"pattern; post={postfix_elapsed:.3f}s vs pre={prefix_elapsed:.3f}s"
+    )
+    assert (
+        len(postfix_seen) < 8
+    ), f"post-fix loop must not drain all chunks; got {len(postfix_seen)}"
+
+
+def test_unsloth_stream_loop_emits_zero_tokens_on_preset_cancel():
+    # Pending-cancel replay: _TrackedCancel.__enter__ already set
+    # cancel_event before the generator body starts iterating. The
+    # top-of-loop check must short-circuit the very first iteration so
+    # no token is emitted. Catches a regression that moves the check
+    # below `next()` -- the mid-loop test would still pass but this
+    # test would observe one extra token leak.
+    cancel_event = threading.Event()
+    cancel_event.set()
+    reset_calls = [0]
+
+    class _Backend:
+        def reset_generation_state(self):
+            reset_calls[0] += 1
+
+    backend = _Backend()
+
+    next_calls = [0]
+
+    def _generate():
+        while True:
+            next_calls[0] += 1
+            yield f"cum-{next_calls[0]}"
+
+    async def _loop():
+        _DONE = object()
+        loop = asyncio.get_event_loop()
+        gen = _generate()
+        seen = []
+        while True:
+            if cancel_event.is_set():
+                backend.reset_generation_state()
+                break
+            cumulative = await loop.run_in_executor(None, next, gen, _DONE)
+            if cumulative is _DONE:
+                break
+            seen.append(cumulative)
+        return seen
+
+    seen = asyncio.run(_loop())
+    assert seen == [], (
+        f"loop must emit zero tokens when cancel_event is pre-set "
+        f"(pending-replay path); got {seen}"
+    )
+    assert next_calls[0] == 0, (
+        f"loop must not call next() at all on pre-set cancel; got "
+        f"{next_calls[0]} calls"
+    )
+    assert reset_calls[0] == 1, (
+        f"backend.reset_generation_state() must still fire exactly once "
+        f"on pre-set cancel; got {reset_calls[0]}"
+    )
+
+
+def test_audio_stream_emits_zero_chunks_on_preset_cancel():
+    # Symmetric to the Unsloth pre-set test: the audio loop's top-of-loop
+    # cancel check must skip the asyncio.to_thread(next, ...) call when
+    # cancel_event was already set via pending-replay.
+    cancel_event = threading.Event()
+    cancel_event.set()
+
+    next_calls = [0]
+
+    def _audio_gen():
+        while True:
+            next_calls[0] += 1
+            yield f"chunk-{next_calls[0]}"
+
+    async def _loop():
+        _DONE = object()
+        gen = _audio_gen()
+        seen = []
+        while True:
+            if cancel_event.is_set():
+                break
+            chunk_text = await asyncio.to_thread(next, gen, _DONE)
+            if chunk_text is _DONE:
+                break
+            seen.append(chunk_text)
+        return seen
+
+    seen = asyncio.run(_loop())
+    assert seen == [], f"audio loop must emit zero chunks on pre-set cancel; got {seen}"
+    assert next_calls[0] == 0, (
+        f"audio loop must not call next() on pre-set cancel; got "
+        f"{next_calls[0]} calls"
+    )

From b09aa82a3ac9ac9794c497e4b8f747f77e52b162 Mon Sep 17 00:00:00 2001
From: Daniel Han <danielhanchen@gmail.com>
Date: Fri, 24 Apr 2026 12:02:03 -0700
Subject: [PATCH 169/216] Studio: add github_repo seed reader and GitHub
 Support Bot recipe (#5169)

* Studio: add github_repo seed reader and GitHub Support Bot recipe

Adds a first-party Data Designer seed reader that scrapes GitHub issues,
pull requests, and commits from one or more repositories via the GraphQL
API, and a learning recipe (GitHub Support Bot) that turns those rows into
synthetic support Q&A pairs for fine-tuning.

Backend (new plugin studio/backend/plugins/data-designer-github-repo-seed):
* GitHubRepoSeedSource config: repos, token (falls back to GH_TOKEN /
  GITHUB_TOKEN env var), item_types (issues / pulls / commits),
  per-resource limit (0 means all), max_comments_per_item.
* Rate-limit-aware GraphQL client (GitHubClient + RepoScraper) shared
  across repos; flattens each item into a uniform row with columns
  item_type, repo, number, title, body, state, author, created_at,
  closed_at, url, labels, comments.
* Registered via the data_designer.plugins entry point.

Frontend:
* New seed_github block variant so the seed node card shows
  "GitHub repositories" instead of the generic "Document file"
  placeholder, with its own icon and inline summary (repo count +
  item-type list).
* Rewritten seed dialog github_repo form: repos textarea pre-filled with
  unslothai/unsloth + unslothai/unsloth-zoo, password input for the GH
  token, items-per-repo number with an "All" toggle, and the noisier
  options (item types, max comments, include comments) tucked under an
  Advanced collapsible.
* Local model auto-load on Run: if a recipe uses an is_local provider
  and the inference server is not already serving that model, the
  executions hook calls /api/inference/load first. Removes the "open
  /chat to load a model" prerequisite that users kept tripping on.
* Honor the recipe's run.rows value in the Run dialog (previously the
  store reset to 5 regardless of what the template shipped).

Recipe (studio/frontend/src/features/data-recipes/learning-recipes/
github-support-bot.json):
* Defaults to the Local Model provider + unsloth/gemma-4-E2B-it-GGUF.
* Scrapes unslothai/unsloth and unslothai/unsloth-zoo, issues and pulls,
  up to 100 items per resource.
* Two LLM blocks: normalized_question (llm-text) rewrites each thread
  into a clean support question, support_answer (llm-structured)
  produces JSON with answer / diagnosis_questions / cites / confidence.
* Run defaults to 10 rows for a quick smoke test.

Verified end-to-end on a running Studio: card renders, source-data
dialog is pre-populated, All toggle disables the limit input, the
recipe executes and produces rows against a loaded local GGUF.

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

* fix: improve GitHub recipe support

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

* Studio: speed up GitHub scraper and harden the support-bot recipe

Addresses a perf issue found while demoing the github_repo seed reader:

Scraper is too slow at scale. The PRs GraphQL query pulls deeply nested
fields (reviewThreads, reviews, commits, timelineItems, etc.) so the
page size was pinned at 3 to stay under GitHub's node-count ceiling. 100
PRs meant 34 serial round trips. Added lighter query variants
(PRS_PAGE_QUERY_LIGHT, ISSUES_PAGE_QUERY_LIGHT) that drop the fields the
Studio flatten layer does not use (it only reads title, body, state,
author, labels, comments). With the light query PR pages can safely go
to 25 per page and issues to 50. The plugin scraper now passes
light=True to RepoScraper so Studio always uses the fast path; the heavy
query remains available for other callers.

Recipe defaults are now demo-ready with production knobs called out:
- max_parallel_requests: 1 and max_tokens: 800 so small local models
  stay stable when running the support_answer structured column.
- support_answer prompt trimmed to 80-200 words so gemma-4-E2B GGUF can
  actually comply with the schema. The canonical 150-300 word codex
  prompt is still documented in the node3 markdown note for
  production upgrades.

* Studio: rename GitHub recipe to 'GitHub Scraper' and add Easy mode

Changes the recipe framing from a single-purpose 'Support Bot' pipeline
to a general-purpose scraper that produces {user_request,
grounded_response} training pairs. Aligns with the canonical
github_data_gatherer dataset (11 enrichment tasks mirrored in pr_requests_20
/ issue_requests_20 on the input side and explain_pr / issue_fix_plan /
issue_solution on the output side).

Recipe JSON changes:
- columns[0] renamed normalized_question -> user_request, prompt now
  inverts a GitHub thread into a realistic user ask instead of
  normalising it.
- columns[1] renamed support_answer -> coauthor_response, emits
  {response, followups, cites, task, confidence} and branches on
  issue vs PR thread type.
- Notes rewritten to document the 11-task catalog and the canonical
  production prompt to paste in for a full dataset backfill.

Frontend: Easy mode for github_repo recipes. The drag-and-drop canvas is
hidden behind an 'Advanced' tab; Easy mode is the default for any recipe
whose seed_source_type is github_repo. The Easy form reuses the existing
GithubRepoSeedForm (promoted to exported), adds a rows input bound to
previewRows, a model field bound to the model_config, and a single Run
button that calls runPreview() directly (no modal). Non-github recipes
see the same Editor / Runs tabs as before.

View mode persists per-recipe-id in localStorage under
recipe-studio:view-mode:<recipeId>.

* Studio: auto-detect server GH_TOKEN and widen Easy-mode detection

The GitHub seed form now fetches /api/data-recipe/seed/github/env-token
on mount and, when the server exposes a GH_TOKEN / GITHUB_TOKEN env var
and the token field is blank, shows a small 'Using server env var' badge
and swaps the placeholder text. The token value itself is never returned
to the UI.

Widens Easy-mode detection in recipe-studio-page.tsx so that recipes
saved before ui.seed_source_type was persisted also get the Easy tab:
falls back to recipe.seed_config.source.seed_type, which is always
present for github_repo seeds.

* fix: polish GitHub recipe UI

* Studio: default llama-server --threads to -1 (auto)

Previously we passed --threads only when the caller set an explicit
value, which meant llama-server fell back to its internal default.
That default has varied across llama.cpp builds (some versions use
hardware concurrency including hyperthreads, which hurts throughput on
CPU-heavy inference). Always passing --threads -1 pins the behaviour
to llama.cpp's auto-detect (physical cores).

Caller-supplied n_threads still wins when non-None.

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

* Studio: auto-switch Easy mode to Runs pane on run start

Easy mode had no progress island or canvas overlay, so after clicking Run
the only visible state was the button label flipping to "Running..." while
the screen otherwise stayed identical. This reads as stuck even though the
job is progressing.

Wire an onExecutionStart callback from recipe-studio-page.tsx through to
useRecipeExecutions so that when a run is kicked off from easy mode, the
page flips to the executions view where the Runs sidebar, progress bar,
rate/ETA panel, and live log are rendered. Advanced/editor mode keeps its
existing behavior and stays on the canvas (it already has the floating
ExecutionProgressIsland).

* fix: clean up GitHub scraper layout

* Studio: forward llm-structured output_format as llama-server response_format

Local GGUF runs of llm-structured columns used to generate the full
max_tokens budget before the prompt-level "return JSON in a ```json
fence" instruction got parsed. Small models (e.g. gemma-4-E2B-it)
routinely broke format, so each row took ~65s and frequently failed
with "No parsable JSON structure within ```json markdown fence".

For any local-provider model_config referenced by an llm-structured
column, clone the model_config and inject response_format into the
clone's inference_parameters. Uses llama.cpp server's flat shape
(tools/server/README.md):

    {"type": "json_schema", "schema": <output_format>}

Not the OpenAI-nested form; data_designer's OpenAI adapter forwards
response_format verbatim via facade._COMPLETION_REQUEST_FIELDS, and
llama-server's documented schema path expects the flat variant.

The clone is per (model_alias, column) so:
- llm-text / llm-judge columns that share the same alias keep
  free-form sampling.
- Each structured column gets its own schema, so columns with
  different output_formats don't collide.

Effect on gemma-4-E2B-it demos: every row parses cleanly, and the
model terminates immediately after the closing brace instead of
running to max_tokens. Net wall-clock is usually faster even though
grammar-constrained sampling is slightly slower per token.

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

* Studio: flip Easy to Runs pane before validation scrape, not after

Previously onExecutionStart fired inside runExecution, which runs AFTER
validateRecipe() -- and validation re-invokes the seed reader. For the
github_repo reader that is a full GraphQL scrape, so the user sat on a
"Running..." button with an otherwise unchanged Easy form for 10-15s
before anything moved.

Call onExecutionStart at the top of runWithValidation, right after we
have a payload to send. The view flips immediately; ensureLocalModelLoaded
+ validateRecipe now run against the Runs pane instead of a frozen Easy
form. runExecution still calls onExecutionStart downstream, but the
callback is idempotent (the page's easy -> executions guard skips the
second call), so no behaviour change for runs that pass validation.

If validation fails the toast + runErrors path still fires; the Easy
form's error banner still reads runErrors when the user switches back.

* Studio: unify data-recipe workflow auth on sk-unsloth-* keys

The previous commit (a61b4cc9) assumed storage.create_api_key(..., internal=True)
and storage.revoke_internal_api_key(key_id) existed, but those helpers were
only in the working tree, never committed. Recipe runs in local-model mode
were therefore crashing with 500 when _inject_local_providers tried to mint
a workflow key. This commit ships the missing pieces.

auth/storage.py:
- api_keys schema gains is_internal INTEGER DEFAULT 0 (with a guarded
  ALTER TABLE migration so existing auth.db files upgrade in place).
- create_api_key takes an internal=False kwarg; internal keys are flagged
  so they can be hidden from user-facing listings.
- list_api_keys takes include_internal=False so UIs never see workflow keys.
- New revoke_internal_api_key(key_id): id-only revoke for keys minted by
  non-user subjects (the JobManager does not know a username).

core/data_recipe/jobs/manager.py:
- JobManager.start accepts internal_api_key_id and stores it on Job so
  lifecycle handlers can revoke eagerly.
- _handle_event revokes on EVENT_JOB_COMPLETED / _ERROR / _CANCELLED.
- _pump_loop subprocess-died fallback also retires the key so a crashed
  worker cannot leak a live sk-unsloth-* beyond its TTL.
- Revocation is best-effort (swallow exceptions) -- the 24h TTL is the
  safety net if storage hiccups.

core/data_recipe/jobs/types.py:
- Job dataclass gains internal_api_key_id: int | None = None.

Replaces the bespoke 24h JWT path that jobs.py used to mint for local
providers. One mint/revoke/verify surface for every API key the server
issues, and revocation is now eager (seconds, not 24h) instead of TTL-only.

* Studio: plug workflow-key leak on unexpected create_job errors

Review follow-up on the sk-unsloth-* workflow-key lifecycle in
create_job. Previously the revoke handlers wrapped mgr.start(...) but
only caught RuntimeError and ValueError, and get_job_manager() sat
outside the try block entirely. Any other exception type (TypeError
from a mismatched kwarg, OSError from the queue write, etc.) would
bubble up to FastAPI and leave the minted key live until its 24h TTL.

Fix: one try block covers both get_job_manager() and mgr.start(), with
a trailing except Exception that revokes and re-raises. The
RuntimeError -> 409 and ValueError -> 400 paths are unchanged so
specific client-facing status codes still surface. Revocation is still
best-effort (_revoke_internal_api_key_safe swallows errors) because we
never want revoke failures to mask the original crash.

Severity is low -- the key can't bootstrap longer access and the 24h
TTL bounds the window -- but the reviewer's point stands: eager revoke
on every failure path is the right invariant.

* Studio: nest response_format under extra_body so pydantic accepts it

The previous commit dropped response_format at the top level of a cloned
model_config's inference_parameters, which BuilderConfig rejected with:

  ValidationError: Extra inputs are not permitted [type=extra_forbidden]
  data_designer.model_configs.1.inference_parameters.response_format

data_designer's BaseInferenceParams is a pydantic model with extra=forbid
and only a fixed set of fields (temperature, top_p, max_tokens,
max_parallel_requests, timeout, extra_body). The pass-through path for
anything the schema doesn't know about is `extra_body`, which the
OpenAI SDK spreads into the chat-completions request body at the top
level -- which is exactly where llama-server reads response_format from.

Inject under extra_body (merging with any existing extra_body contents)
so the clone validates. llama-server still receives
{"type": "json_schema", "schema": <output_format>} at the top level of
the request body, which is the flat shape llama.cpp's server expects.

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

* Studio: forward response_format to llama-server and fence-wrap the reply

Two-part fix for the llm-structured data-recipe path:

(1) The /v1/chat/completions proxy was dropping response_format. The
route's passthrough branch only triggered on tools / tool messages, so
requests carrying a JSON schema fell into the non-passthrough GGUF path
which calls generate_chat_completion (no response_format kwarg). The
schema never reached llama-server, so guided decoding was a no-op and
the model emitted free-form text that happened to parse a fraction of
the time. Widen the passthrough trigger and teach _build_passthrough_payload
to forward response_format so llama-server's GBNF grammar actually runs.

Guided decoding does not require supports_tools, so split the condition:
a request is now passthrough-routed if it carries tools/tool messages
(existing behavior) OR carries response_format (new). The vision guard,
streaming fork, and tools-choice defaulting are unchanged.

(2) data_designer's llm-structured parser looks for a ```json ... ```
markdown fence and discards anything else. Guided decoding emits only
the JSON object (the GBNF grammar has no fence tokens), so a
100%-valid schema-constrained run still ended up 0 ok / N failed with
"No parsable JSON structure within ```json markdown fence". In
_openai_passthrough_non_streaming, wrap each choice's content in the
expected fence when the caller asked for guided decoding. Already-fenced
content is left alone so other clients that prefer raw JSON are not
affected; the wrap is scoped to requests that carried response_format.

Net effect on the GitHub Support Bot recipe on a local GGUF: schema
actually binds during sampling, content arrives wrapped in the fence
data_designer expects, and generation terminates immediately after the
closing brace instead of running out to max_tokens.

* Studio: Easy mode runs a full run, capped at the user's row count

Easy mode used to call runPreview, which produces a test run: no
artifact persisted, reduced progress tracking, and framed in the Runs
pane as "Test run". The whole point of the form is to let a user kick
off a real dataset build with one click, so wire it to runFull instead
and bind the Rows input to fullRows (not previewRows).

runFull requires a non-empty fullRunName. The Easy form has no run-name
input, so seed a default on mount whenever Easy is active and
fullRunName is still empty. Uses `<recipe name> <iso-timestamp>` so
each Easy run gets a stable-ish default that still sorts chronologically
in the Runs pane. User can override it from the Advanced run dialog
before clicking Run.

Rename GithubScraperEasyView's rows props from previewRows/setPreviewRows
to rows/setRows so the view stays agnostic to which hook state the page
chooses to bind. Loading indicator now follows fullLoading.

* Studio: clamp GitHub scrape page size and memoize the materialization

Two wins for the "before Generating fires" gap on small previews:

(1) scrape_{issues,prs,commits} hardcoded per_page (50 / 25 / 100) and
only checked the trial limit AFTER the page was written, so a 1-row
Easy run still asked GitHub for a full 50-issue + 25-PR page, wrote
them all to JSONL, and then stopped because total_new already exceeded
the trial cap. Cap per_page at min(page_cap, trial_limit) so
github_limit=1 actually asks for first:1.

(2) GitHubRepoSeedReader.get_dataset_uri used to scrape fresh on every
invocation. data_designer calls the seed reader multiple times per
recipe job (validation, preview, per-column sampling), so a 2-repo
Easy preview ran the full GraphQL scrape three times back-to-back,
burning ~15s of dead air before any LLM generation began.

Added a module-level in-process cache keyed on
(repos, item_types, limit, include_comments, max_comments_per_item,
sha256(token)[:16]) that stores the JSONL path of the first
materialization. Subsequent calls with the same signature return the
cached path, guarded by a staleness check that drops the entry if the
file was tmp-cleaned. Raw token values never land in the key.

Net effect on a 1-row Easy run, 2 repos, limit=1: 2 GraphQL round
trips instead of ~12, and the first-to-Generating gap collapses from
~15s to roughly 2-3s.

* Studio: make Easy mode Rows input editable instead of snapping to 1

The Rows to generate input used type="number" with value bound directly
to the rows state and an onChange that coerced any non-positive parse
result back to 1. The moment the user pressed backspace to clear the
field, the parent re-rendered with value=1 and the caret jumped, making
it impossible to change the value without arrowing the browser's +/-
spinner.

Switch to a text input with inputMode="numeric" and pattern="[0-9]*"
(so mobile still shows a numeric keyboard, and the browser drops the
spinner buttons the user did not want). Add a local rowsText buffer so
the field can hold transient empty / partial digit strings while
editing without fighting the parent state; the canonical rows value
only advances when the buffer parses to a valid integer in [1, 10000],
and onBlur clamps back to 1 or 10000 if the user left it out of range.

No behavior change for valid numeric edits - the downstream runFull()
still sees a clean positive integer.

* Studio: expand dataset cells horizontally by column on click

Click a long cell to expand that whole column. Click again to collapse.
Replaces the prior row-level vertical expansion which made it hard to
compare cells across columns. State is scoped per execution and per
column; the row itself is no longer a click target.

* Studio: force expanded dataset column to grow wide enough to read

* Studio: disable thinking for local recipe inference and plumb the kwarg

Reasoning-capable models (gemma-3n, qwen3.5, etc.) emit a
<think>...</think> preamble ahead of the answer by default, which
roughly doubles the generated token count per row on a local GGUF
and pushes the actual answer past data_designer's json-fence regex
on llm-structured columns. Recipes want the terse answer, not the
scratchpad.

Two halves of the fix:

(1) routes/data_recipe/jobs.py: when _inject_local_providers walks
the recipe's model_configs to point them at the local endpoint, also
stash chat_template_kwargs={"enable_thinking": false} under each
config's inference_parameters.extra_body. OpenAI SDK spreads
extra_body into the top-level request body, so llama-server and the
Studio /v1/chat/completions route both see it.

(2) routes/inference.py: the chat-completions route previously
dropped chat_template_kwargs on the floor because the whitelist
body builder only forwarded known fields.

    - At the top of openai_chat_completions, lift
      chat_template_kwargs.enable_thinking from payload.model_extra
      onto the typed payload.enable_thinking field when the caller
      did not set the latter, so the non-passthrough GGUF path's
      generate_chat_completion(...) call honors the override.
    - Teach _build_passthrough_payload to forward a
      chat_template_kwargs dict, and have _build_openai_passthrough_body
      derive that dict from payload.enable_thinking so
      response_format requests (structured columns) also land at
      llama-server with the reasoning preamble suppressed.

Net effect on a 10-row support-bot run with gemma-4-E2B-it-GGUF:
responses arrive without <think> tags, wall-clock per call drops
roughly in half, and structured columns stop leaking reasoning
tokens through the GBNF-constrained output.

* Studio: update GitHub Support Bot learning recipe with maintainer layout

Replace the template with the hand-laid-out export from the maintainer
so note nodes ship with real x/y positions (scattered around the
graph instead of all stacked at x=480) and the edges / canvas pan look
correct on first load. Also picks up the maintainer's prompt tweaks and
output schema names (coauthor_response / user_request / followups / task /
cites / confidence).

Diff is mostly ui.nodes positions and prompt bodies; runtime shape is
unchanged (seed_config / columns still target model_1 against the Local
Model provider).

* Studio: auto-size dataset sample columns; wide text gets a wide column

Drop the per-column click-to-expand toggle and the 180-char truncation.
Every column now renders its full value. Columns with long text get a
min-w of 48rem so the text is readable without wrapping into a tall
block; narrow-content columns get a 12rem min-w. The table wrapper
already has overflow-x-auto, so wide-column totals cause a horizontal
scrollbar instead of cramming everything into the viewport.

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

* fix GitHub scrape progress

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

* add resetApiBase export for test setup

* Studio: rename github-support-bot output columns to User / Assistant

Previously emitted user_request and coauthor_response, which did not
match the canonical User / Assistant chat-pair shape that downstream
SFT consumers expect. Renamed the columns in the recipe JSON (columns,
UI node ids, edges, notes, prompt Jinja refs) and the matching copy in
the learning-recipes index, data-recipes-page, and easy view.

---------

Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com>
Co-authored-by: wasimysaid <wasimysdev@gmail.com>
---
 pyproject.toml                                |   1 +
 studio/backend/auth/storage.py                |  84 +-
 .../core/data_recipe/jobs/constants.py        |   1 +
 .../backend/core/data_recipe/jobs/manager.py  | 105 ++-
 studio/backend/core/data_recipe/jobs/parse.py | 224 ++++-
 studio/backend/core/data_recipe/jobs/types.py |  25 +
 .../backend/core/data_recipe/jobs/worker.py   |  29 +-
 studio/backend/core/inference/llama_cpp.py    |   6 +-
 .../data-designer-github-repo-seed/README.md  |  73 ++
 .../pyproject.toml                            |  25 +
 .../__init__.py                               |   7 +
 .../data_designer_github_repo_seed/config.py  |  64 ++
 .../data_designer_github_repo_seed/impl.py    |  83 ++
 .../data_designer_github_repo_seed/plugin.py  |  10 +
 .../data_designer_github_repo_seed/scraper.py | 236 +++++
 .../scraper_impl/__init__.py                  |   2 +
 .../scraper_impl/gh_client.py                 | 248 ++++++
 .../scraper_impl/queries.py                   | 685 +++++++++++++++
 .../scraper_impl/scraper.py                   | 756 ++++++++++++++++
 .../scraper_impl/state_store.py               | 105 +++
 .../single-env/data-designer-deps.txt         |   3 +-
 studio/backend/routes/data_recipe/jobs.py     | 199 ++++-
 studio/backend/routes/data_recipe/seed.py     |  12 +
 studio/backend/routes/data_recipe/validate.py |  67 ++
 studio/backend/routes/inference.py            |  99 ++-
 .../tests/test_data_recipe_github_progress.py |  91 ++
 .../learning-recipes/github-support-bot.json  | 238 +++++
 .../data-recipes/learning-recipes/index.ts    |  11 +
 .../data-recipes/pages/data-recipes-page.tsx  |  17 +
 .../src/features/recipe-studio/api/index.ts   |  70 +-
 .../recipe-studio/blocks/definitions.ts       |  18 +-
 .../executions/execution-data-tab.tsx         | 178 +++-
 .../executions/execution-overview-tab.tsx     |  66 ++
 .../components/executions/executions-view.tsx |  87 +-
 .../components/inline/inline-seed.tsx         |  83 +-
 .../components/recipe-studio-header.tsx       |  11 +-
 .../runtime/execution-progress-island.tsx     |  98 ++-
 .../recipe-studio/dialogs/preview-dialog.tsx  |   9 +-
 .../dialogs/seed/seed-dialog.tsx              | 810 ++++++++++++++----
 .../easy/github-crawler-easy-view.tsx         | 191 +++++
 .../features/recipe-studio/execution-types.ts |  26 +-
 .../recipe-studio/executions/runtime.ts       |  11 +
 .../hooks/use-recipe-executions.ts            | 117 ++-
 .../hooks/use-recipe-studio-actions.ts        |   4 +
 .../recipe-studio/recipe-studio-page.tsx      |  73 +-
 .../recipe-studio/stores/recipe-studio.ts     |  10 +-
 .../src/features/recipe-studio/types/index.ts |  19 +-
 .../import/parsers/seed-config-parser.ts      |  32 +
 .../features/recipe-studio/utils/node-data.ts |   4 +-
 .../utils/payload/builders-seed.ts            | 108 ++-
 .../recipe-studio/utils/payload/types.ts      |   2 +-
 .../recipe-studio/utils/validation.ts         |  51 +-
 studio/frontend/src/lib/api-base.ts           |   6 +
 studio/install_python_stack.py                |  39 +-
 54 files changed, 5210 insertions(+), 419 deletions(-)
 create mode 100644 studio/backend/plugins/data-designer-github-repo-seed/README.md
 create mode 100644 studio/backend/plugins/data-designer-github-repo-seed/pyproject.toml
 create mode 100644 studio/backend/plugins/data-designer-github-repo-seed/src/data_designer_github_repo_seed/__init__.py
 create mode 100644 studio/backend/plugins/data-designer-github-repo-seed/src/data_designer_github_repo_seed/config.py
 create mode 100644 studio/backend/plugins/data-designer-github-repo-seed/src/data_designer_github_repo_seed/impl.py
 create mode 100644 studio/backend/plugins/data-designer-github-repo-seed/src/data_designer_github_repo_seed/plugin.py
 create mode 100644 studio/backend/plugins/data-designer-github-repo-seed/src/data_designer_github_repo_seed/scraper.py
 create mode 100644 studio/backend/plugins/data-designer-github-repo-seed/src/data_designer_github_repo_seed/scraper_impl/__init__.py
 create mode 100644 studio/backend/plugins/data-designer-github-repo-seed/src/data_designer_github_repo_seed/scraper_impl/gh_client.py
 create mode 100644 studio/backend/plugins/data-designer-github-repo-seed/src/data_designer_github_repo_seed/scraper_impl/queries.py
 create mode 100644 studio/backend/plugins/data-designer-github-repo-seed/src/data_designer_github_repo_seed/scraper_impl/scraper.py
 create mode 100644 studio/backend/plugins/data-designer-github-repo-seed/src/data_designer_github_repo_seed/scraper_impl/state_store.py
 create mode 100644 studio/backend/tests/test_data_recipe_github_progress.py
 create mode 100644 studio/frontend/src/features/data-recipes/learning-recipes/github-support-bot.json
 create mode 100644 studio/frontend/src/features/recipe-studio/easy/github-crawler-easy-view.tsx

diff --git a/pyproject.toml b/pyproject.toml
index 815c6ee119..5687ea12f8 100644
--- a/pyproject.toml
+++ b/pyproject.toml
@@ -53,6 +53,7 @@ studio = [
     "frontend/*.yaml",
     "frontend/.git*",
     "backend/requirements/**/*",
+    "backend/plugins/**/*",
     "backend/core/data_recipe/oxc-validator/*.json",
     "backend/core/data_recipe/oxc-validator/*.mjs",
 ]
diff --git a/studio/backend/auth/storage.py b/studio/backend/auth/storage.py
index 2b0e359d39..9a03f5f542 100644
--- a/studio/backend/auth/storage.py
+++ b/studio/backend/auth/storage.py
@@ -146,10 +146,18 @@ def get_connection() -> sqlite3.Connection:
             created_at TEXT NOT NULL,
             last_used_at TEXT,
             expires_at TEXT,
-            is_active  INTEGER NOT NULL DEFAULT 1
+            is_active  INTEGER NOT NULL DEFAULT 1,
+            is_internal INTEGER NOT NULL DEFAULT 0
         );
         """
     )
+    api_key_columns = {
+        row["name"] for row in conn.execute("PRAGMA table_info(api_keys)")
+    }
+    if "is_internal" not in api_key_columns:
+        conn.execute(
+            "ALTER TABLE api_keys ADD COLUMN is_internal INTEGER NOT NULL DEFAULT 0"
+        )
     conn.execute(
         """
         CREATE TABLE IF NOT EXISTS app_secrets (
@@ -592,11 +600,15 @@ def create_api_key(
     username: str,
     name: str,
     expires_at: Optional[str] = None,
+    internal: bool = False,
 ) -> Tuple[str, dict]:
     """Create a new API key for *username*.
 
     Returns ``(raw_key, row_dict)`` where *raw_key* is shown to the user
-    exactly once.  The database only stores the SHA-256 hash.
+    exactly once.  The database only stores the PBKDF2 hash.
+
+    Pass ``internal=True`` for keys minted by workflows (e.g. data-recipe
+    runs) that should not appear in user-facing key listings.
     """
     raw_key = API_KEY_PREFIX + secrets.token_hex(16)
     key_hash = _pbkdf2_api_key(raw_key)
@@ -607,10 +619,18 @@ def create_api_key(
     try:
         conn.execute(
             """
-            INSERT INTO api_keys (username, key_prefix, key_hash, name, created_at, expires_at)
-            VALUES (?, ?, ?, ?, ?, ?)
+            INSERT INTO api_keys (username, key_prefix, key_hash, name, created_at, expires_at, is_internal)
+            VALUES (?, ?, ?, ?, ?, ?, ?)
             """,
-            (username, key_prefix, key_hash, name, now, expires_at),
+            (
+                username,
+                key_prefix,
+                key_hash,
+                name,
+                now,
+                expires_at,
+                1 if internal else 0,
+            ),
         )
         conn.commit()
         cur = conn.execute("SELECT * FROM api_keys WHERE key_hash = ?", (key_hash,))
@@ -620,19 +640,33 @@ def create_api_key(
         conn.close()
 
 
-def list_api_keys(username: str) -> list:
-    """Return all API keys for *username* (never exposes ``key_hash``)."""
+def list_api_keys(username: str, include_internal: bool = False) -> list:
+    """Return API keys for *username*. Internal workflow keys are hidden
+    by default so they do not clutter user-facing UIs."""
     conn = get_connection()
     try:
-        cur = conn.execute(
-            """
-            SELECT id, username, key_prefix, name, created_at, last_used_at, expires_at, is_active
-            FROM api_keys
-            WHERE username = ?
-            ORDER BY created_at DESC
-            """,
-            (username,),
-        )
+        if include_internal:
+            cur = conn.execute(
+                """
+                SELECT id, username, key_prefix, name, created_at, last_used_at,
+                       expires_at, is_active, is_internal
+                FROM api_keys
+                WHERE username = ?
+                ORDER BY created_at DESC
+                """,
+                (username,),
+            )
+        else:
+            cur = conn.execute(
+                """
+                SELECT id, username, key_prefix, name, created_at, last_used_at,
+                       expires_at, is_active, is_internal
+                FROM api_keys
+                WHERE username = ? AND is_internal = 0
+                ORDER BY created_at DESC
+                """,
+                (username,),
+            )
         return [dict(row) for row in cur.fetchall()]
     finally:
         conn.close()
@@ -652,6 +686,24 @@ def revoke_api_key(username: str, key_id: int) -> bool:
         conn.close()
 
 
+def revoke_internal_api_key(key_id: int) -> bool:
+    """Revoke an internal workflow-minted key without requiring a username.
+
+    Used by the recipe runner to retire its sk-unsloth-* key once the job
+    terminates, shrinking the window a leaked key could be abused.
+    """
+    conn = get_connection()
+    try:
+        cursor = conn.execute(
+            "UPDATE api_keys SET is_active = 0 WHERE id = ? AND is_internal = 1",
+            (key_id,),
+        )
+        conn.commit()
+        return cursor.rowcount > 0
+    finally:
+        conn.close()
+
+
 def validate_api_key(raw_key: str) -> Optional[str]:
     """Validate *raw_key* and return the owning username, or ``None``.
 
diff --git a/studio/backend/core/data_recipe/jobs/constants.py b/studio/backend/core/data_recipe/jobs/constants.py
index 08237326f8..0045276e20 100644
--- a/studio/backend/core/data_recipe/jobs/constants.py
+++ b/studio/backend/core/data_recipe/jobs/constants.py
@@ -9,6 +9,7 @@ STAGE_PREVIEW = "preview"
 STAGE_DAG = "dag"
 STAGE_HEALTHCHECK = "healthcheck"
 STAGE_SAMPLING = "sampling"
+STAGE_SOURCE = "source"
 STAGE_COLUMN_CONFIG = "column_config"
 STAGE_GENERATING = "generating"
 STAGE_BATCH = "batch"
diff --git a/studio/backend/core/data_recipe/jobs/manager.py b/studio/backend/core/data_recipe/jobs/manager.py
index 3d7cf2dbe6..db59d573a9 100644
--- a/studio/backend/core/data_recipe/jobs/manager.py
+++ b/studio/backend/core/data_recipe/jobs/manager.py
@@ -33,6 +33,60 @@ from .worker import run_job_process
 _CTX = mp.get_context("spawn")
 
 
+def _github_source_estimated_total(recipe: dict) -> int | None:
+    seed_config = recipe.get("seed_config")
+    if not isinstance(seed_config, dict):
+        return None
+    source = seed_config.get("source")
+    if not isinstance(source, dict) or source.get("seed_type") != "github_repo":
+        return None
+
+    repos_raw = source.get("repos")
+    repos = (
+        [repo for repo in repos_raw if isinstance(repo, str) and repo.strip()]
+        if isinstance(repos_raw, list)
+        else []
+    )
+    item_types_raw = source.get("item_types")
+    item_types = (
+        [
+            item
+            for item in item_types_raw
+            if isinstance(item, str) and item in {"issues", "pulls", "commits"}
+        ]
+        if isinstance(item_types_raw, list)
+        else []
+    )
+    try:
+        limit = int(source.get("limit") or 0)
+    except (TypeError, ValueError):
+        return None
+    if not repos or not item_types or limit <= 0:
+        return None
+    return len(repos) * len(item_types) * limit
+
+
+def _source_progress_status(job: Job) -> dict[str, Any] | None:
+    progress = job.source_progress
+    if progress is None:
+        return None
+    return {
+        "source": progress.source,
+        "status": progress.status,
+        "repo": progress.repo,
+        "resource": progress.resource,
+        "page": progress.page,
+        "page_items": progress.page_items,
+        "fetched_items": progress.fetched_items,
+        "estimated_total": progress.estimated_total,
+        "percent": progress.percent,
+        "rate_remaining": progress.rate_remaining,
+        "retry_after_sec": progress.retry_after_sec,
+        "message": progress.message,
+        "updated_at": progress.updated_at,
+    }
+
+
 @dataclass
 class Subscription:
     replay: list[dict]
@@ -71,8 +125,20 @@ class JobManager:
         self._pump_thread: threading.Thread | None = None
         self._seq: int = 0
 
-    def start(self, *, recipe: dict, run: dict) -> str:
-        """Spawn the job subprocess (one at a time, no cap)."""
+    def start(
+        self,
+        *,
+        recipe: dict,
+        run: dict,
+        internal_api_key_id: int | None = None,
+    ) -> str:
+        """Spawn the job subprocess (one at a time, no cap).
+
+        ``internal_api_key_id`` is the row id of a workflow-scoped
+        sk-unsloth-* key minted by the route layer for local providers.
+        JobManager revokes it when the job reaches a terminal state so the
+        key's live window is no longer than the run.
+        """
         llm_columns = recipe.get("columns") or []
         llm_column_count = 0
         if isinstance(llm_columns, list):
@@ -92,6 +158,10 @@ class JobManager:
             job_id = uuid.uuid4().hex
             self._job = Job(job_id = job_id, status = "pending", started_at = time.time())
             self._job.progress_columns_total = llm_column_count
+            self._job.source_progress_estimated_total = _github_source_estimated_total(
+                recipe
+            )
+            self._job.internal_api_key_id = internal_api_key_id
             self._events.clear()
             self._seq = 0
 
@@ -163,6 +233,7 @@ class JobManager:
                     "ok": job.column_progress.ok,
                     "failed": job.column_progress.failed,
                 },
+                "source_progress": _source_progress_status(job),
                 "model_usage": {
                     name: {
                         "model": usage.model,
@@ -405,6 +476,7 @@ class JobManager:
             for e in self._drain_queue(mp_q):
                 self._handle_event(job, e)
 
+            retired_job: Job | None = None
             with self._lock:
                 if self._job and self._job.status in {
                     "pending",
@@ -429,6 +501,9 @@ class JobManager:
                             "job_id": self._job.job_id,
                         }
                     )
+                    retired_job = self._job
+            if retired_job is not None:
+                self._retire_workflow_key(retired_job)
             return
 
     def _handle_event(self, job: Job, event: dict) -> None:
@@ -436,6 +511,7 @@ class JobManager:
         et = event.get("type")
         msg = event.get("message") if et == "log" else None
 
+        terminal = False
         with self._lock:
             if self._job is None or self._job.job_id != job.job_id:
                 return
@@ -452,18 +528,43 @@ class JobManager:
                 if self._job.progress.total and self._job.progress.total > 0:
                     self._job.progress.done = self._job.progress.total
                     self._job.progress.percent = 100.0
+                terminal = True
             if et == EVENT_JOB_ERROR:
                 self._job.status = "error"
                 self._job.finished_at = time.time()
                 self._job.error = event.get("error") or "error"
+                terminal = True
+            if et == EVENT_JOB_CANCELLED:
+                terminal = True
 
             if msg:
                 upd = parse_log_message(msg)
                 if upd:
                     apply_update(self._job, upd)
 
+        if terminal:
+            self._retire_workflow_key(job)
+
         self._emit(event)
 
+    def _retire_workflow_key(self, job: Job) -> None:
+        """Revoke the workflow-scoped sk-unsloth-* key, if one was minted.
+
+        Best-effort: revocation failures are swallowed. The key would
+        expire on its own after 24h, so a missed revoke is a latency
+        concern, not a correctness one.
+        """
+        key_id = getattr(job, "internal_api_key_id", None)
+        if not key_id:
+            return
+        try:
+            from auth import storage  # deferred: avoids circular import
+
+            storage.revoke_internal_api_key(int(key_id))
+        except Exception:
+            pass
+        job.internal_api_key_id = None
+
 
 _JOB_MANAGER: JobManager | None = None
 
diff --git a/studio/backend/core/data_recipe/jobs/parse.py b/studio/backend/core/data_recipe/jobs/parse.py
index 324b62a92e..cea6d8ea64 100644
--- a/studio/backend/core/data_recipe/jobs/parse.py
+++ b/studio/backend/core/data_recipe/jobs/parse.py
@@ -4,6 +4,7 @@
 from __future__ import annotations
 
 import re
+import time
 from dataclasses import dataclass
 from typing import Any
 
@@ -17,9 +18,10 @@ from .constants import (
     STAGE_PREVIEW,
     STAGE_PROFILING,
     STAGE_SAMPLING,
+    STAGE_SOURCE,
     USAGE_RESET_STAGES,
 )
-from .types import Job, ModelUsage, Progress
+from .types import Job, ModelUsage, Progress, SourceProgress
 
 
 @dataclass(frozen = True)
@@ -41,6 +43,7 @@ class ParsedUpdate:
     usage_requests_total: int | None = None
     usage_rpm: float | None = None
     usage_section_start: bool | None = None
+    source_progress: SourceProgress | None = None
 
 
 # kinda of a bummber but currently only option, Best effort parser from data-designer logs -> structured status for UI.
@@ -61,9 +64,165 @@ _RE_USAGE_TOKENS = re.compile(
 _RE_USAGE_REQUESTS = re.compile(
     r"requests:\s*success=(?P<success>\d+),\s*failed=(?P<failed>\d+),\s*total=(?P<total>\d+),\s*rpm=(?P<rpm>[0-9.]+)"
 )
+_RE_GITHUB_PAGE = re.compile(
+    r"^\[(?P<repo>[^\]\s]+/[^\]\s]+)\]\s+"
+    r"(?P<resource>issues|PRs|commits)\s+page\s+(?P<page>\d+)\s+"
+    r"\(\+(?P<items>\d+)\).*?\bremaining=(?P<remaining>\d+)",
+    re.IGNORECASE,
+)
+_RE_GITHUB_RATE_LIMIT = re.compile(
+    r"Rate limit hit\. Sleeping (?P<seconds>\d+)s until reset\.",
+    re.IGNORECASE,
+)
+_RE_GITHUB_SECONDARY_RATE_LIMIT = re.compile(
+    r"Secondary rate limit(?: on REST)?\. Sleep (?P<seconds>\d+)s\.",
+    re.IGNORECASE,
+)
+_RE_GITHUB_REST_RATE_LIMIT = re.compile(
+    r"REST 403/429, sleep (?P<seconds>\d+)",
+    re.IGNORECASE,
+)
+_RE_GITHUB_TRANSIENT = re.compile(
+    r"^(?P<api>GraphQL|REST) (?P<code>\d{3}) transient, retrying",
+    re.IGNORECASE,
+)
+_RE_GITHUB_NETWORK_RETRY = re.compile(
+    r"^(?P<api>GraphQL|REST) network error: .* Retry\.",
+    re.IGNORECASE,
+)
+_RE_GITHUB_TRIAL_LIMIT = re.compile(
+    r"Trial limit reached for (?P<resource>issues|PRs|commits) \((?P<items>\d+)\)",
+    re.IGNORECASE,
+)
+_RE_GITHUB_COMPLETE = re.compile(
+    r"Scraper complete\. GraphQL calls=\d+ REST calls=\d+",
+    re.IGNORECASE,
+)
 
 
 def parse_log_message(msg: str) -> ParsedUpdate | None:
+    m = _RE_GITHUB_PAGE.search(msg)
+    if m:
+        resource_raw = m.group("resource")
+        resource = "pulls" if resource_raw.lower() == "prs" else resource_raw.lower()
+        repo = m.group("repo")
+        page = int(m.group("page"))
+        page_items = int(m.group("items"))
+        return ParsedUpdate(
+            stage = STAGE_SOURCE,
+            source_progress = SourceProgress(
+                source = "github",
+                status = "fetching",
+                repo = repo,
+                resource = resource,
+                page = page,
+                page_items = page_items,
+                rate_remaining = int(m.group("remaining")),
+                message = (
+                    f"Scraping GitHub source: {repo} "
+                    f"{resource} page {page} (+{page_items})"
+                ),
+            ),
+        )
+
+    m = _RE_GITHUB_RATE_LIMIT.search(msg)
+    if m:
+        seconds = int(m.group("seconds"))
+        return ParsedUpdate(
+            stage = STAGE_SOURCE,
+            source_progress = SourceProgress(
+                source = "github",
+                status = "rate_limited",
+                retry_after_sec = seconds,
+                message = (
+                    "Waiting for GitHub rate limit. "
+                    "Studio will resume automatically."
+                ),
+            ),
+        )
+
+    m = _RE_GITHUB_SECONDARY_RATE_LIMIT.search(msg)
+    if m:
+        seconds = int(m.group("seconds"))
+        return ParsedUpdate(
+            stage = STAGE_SOURCE,
+            source_progress = SourceProgress(
+                source = "github",
+                status = "rate_limited",
+                retry_after_sec = seconds,
+                message = (
+                    "Waiting for GitHub secondary rate limit. "
+                    "Studio will resume automatically."
+                ),
+            ),
+        )
+
+    m = _RE_GITHUB_REST_RATE_LIMIT.search(msg)
+    if m:
+        seconds = int(m.group("seconds"))
+        return ParsedUpdate(
+            stage = STAGE_SOURCE,
+            source_progress = SourceProgress(
+                source = "github",
+                status = "rate_limited",
+                retry_after_sec = seconds,
+                message = (
+                    "Waiting for GitHub rate limit. "
+                    "Studio will resume automatically."
+                ),
+            ),
+        )
+
+    m = _RE_GITHUB_TRIAL_LIMIT.search(msg)
+    if m:
+        resource_raw = m.group("resource")
+        resource = "pulls" if resource_raw.lower() == "prs" else resource_raw.lower()
+        items = int(m.group("items"))
+        return ParsedUpdate(
+            stage = STAGE_SOURCE,
+            source_progress = SourceProgress(
+                source = "github",
+                status = "fetching",
+                resource = resource,
+                message = f"GitHub {resource} trial limit reached ({items}).",
+            ),
+        )
+
+    m = _RE_GITHUB_TRANSIENT.search(msg)
+    if m:
+        api = m.group("api")
+        code = m.group("code")
+        return ParsedUpdate(
+            stage = STAGE_SOURCE,
+            source_progress = SourceProgress(
+                source = "github",
+                status = "retrying",
+                message = f"GitHub {api} returned {code}; retrying automatically.",
+            ),
+        )
+
+    m = _RE_GITHUB_NETWORK_RETRY.search(msg)
+    if m:
+        api = m.group("api")
+        return ParsedUpdate(
+            stage = STAGE_SOURCE,
+            source_progress = SourceProgress(
+                source = "github",
+                status = "retrying",
+                message = f"GitHub {api} request failed; retrying automatically.",
+            ),
+        )
+
+    if _RE_GITHUB_COMPLETE.search(msg):
+        return ParsedUpdate(
+            stage = STAGE_SOURCE,
+            source_progress = SourceProgress(
+                source = "github",
+                status = "completed",
+                message = "GitHub source scrape complete.",
+            ),
+        )
+
     m = _RE_SAMPLERS.search(msg)
     if m:
         return ParsedUpdate(
@@ -172,6 +331,8 @@ def apply_update(job: Job, update: ParsedUpdate) -> None:
         job.batch.idx = update.batch_idx
     if update.batch_total is not None:
         job.batch.total = update.batch_total
+    if update.source_progress is not None:
+        _apply_source_progress(job, update.source_progress)
 
     if update.stage in USAGE_RESET_STAGES:
         # usage summary is a short block so we reset once we move into the next stage.
@@ -216,6 +377,67 @@ def apply_update(job: Job, update: ParsedUpdate) -> None:
         usage.rpm = update.usage_rpm
 
 
+def _apply_source_progress(job: Job, progress: SourceProgress) -> None:
+    previous = job.source_progress
+    now = time.time()
+
+    page_items = progress.page_items
+    if progress.repo and progress.resource and progress.page is not None:
+        page_key = f"{progress.repo}:{progress.resource}:{progress.page}"
+        count_key = f"{progress.repo}:{progress.resource}"
+        if page_key not in job._source_seen_pages:
+            job._source_seen_pages.add(page_key)
+            job._source_counts[count_key] = int(
+                job._source_counts.get(count_key, 0)
+            ) + int(page_items or 0)
+
+    fetched_items = sum(job._source_counts.values())
+    if fetched_items <= 0:
+        fetched_items = progress.fetched_items or (
+            previous.fetched_items if previous else None
+        )
+
+    estimated_total = (
+        progress.estimated_total
+        or job.source_progress_estimated_total
+        or (previous.estimated_total if previous else None)
+    )
+    percent: float | None = progress.percent
+    if percent is None and estimated_total and fetched_items is not None:
+        raw_percent = (float(fetched_items) / float(max(1, estimated_total))) * 100.0
+        percent = 100.0 if progress.status == "completed" else min(99.0, raw_percent)
+    if percent is None and previous is not None:
+        percent = previous.percent
+
+    job.source_progress = SourceProgress(
+        source = "github",
+        status = progress.status or (previous.status if previous else None),
+        repo = progress.repo or (previous.repo if previous else None),
+        resource = progress.resource or (previous.resource if previous else None),
+        page = (
+            progress.page
+            if progress.page is not None
+            else (previous.page if previous else None)
+        ),
+        page_items = (
+            page_items
+            if page_items is not None
+            else (previous.page_items if previous else None)
+        ),
+        fetched_items = fetched_items,
+        estimated_total = estimated_total,
+        percent = percent,
+        rate_remaining = (
+            progress.rate_remaining
+            if progress.rate_remaining is not None
+            else (previous.rate_remaining if previous else None)
+        ),
+        retry_after_sec = progress.retry_after_sec,
+        message = progress.message or (previous.message if previous else None),
+        updated_at = now,
+    )
+
+
 def _compute_overall_progress(job: Job, column_progress: Progress) -> Progress:
     if not job.rows:
         return column_progress
diff --git a/studio/backend/core/data_recipe/jobs/types.py b/studio/backend/core/data_recipe/jobs/types.py
index 8d77903238..3d3ddb974e 100644
--- a/studio/backend/core/data_recipe/jobs/types.py
+++ b/studio/backend/core/data_recipe/jobs/types.py
@@ -35,6 +35,23 @@ class BatchProgress:
     total: int | None = None
 
 
+@dataclass
+class SourceProgress:
+    source: str = "github"
+    status: str | None = None
+    repo: str | None = None
+    resource: str | None = None
+    page: int | None = None
+    page_items: int | None = None
+    fetched_items: int | None = None
+    estimated_total: int | None = None
+    percent: float | None = None
+    rate_remaining: int | None = None
+    retry_after_sec: int | None = None
+    message: str | None = None
+    updated_at: float | None = None
+
+
 @dataclass
 class ModelUsage:
     model: str
@@ -57,6 +74,7 @@ class Job:
     progress: Progress = field(default_factory = Progress)
     column_progress: Progress = field(default_factory = Progress)
     batch: BatchProgress = field(default_factory = BatchProgress)
+    source_progress: SourceProgress | None = None
     rows: int | None = None
     cols: int | None = None
     error: str | None = None
@@ -70,8 +88,15 @@ class Job:
     processor_artifacts: dict[str, Any] | None = None
     model_usage: dict[str, ModelUsage] = field(default_factory = dict)
     progress_columns_total: int | None = None
+    source_progress_estimated_total: int | None = None
     completed_columns: list[str] = field(default_factory = list)
+    # Id of the internal sk-unsloth-* API key minted for a local-model
+    # workflow. Revoked when the job terminates so the key's live window
+    # matches the run rather than its 24h TTL.
+    internal_api_key_id: int | None = None
     _current_usage_model: str | None = None
     _in_usage_summary: bool = False
     _seen_generation_columns: list[str] = field(default_factory = list)
     _column_done: dict[str, int] = field(default_factory = dict)
+    _source_counts: dict[str, int] = field(default_factory = dict)
+    _source_seen_pages: set[str] = field(default_factory = set)
diff --git a/studio/backend/core/data_recipe/jobs/worker.py b/studio/backend/core/data_recipe/jobs/worker.py
index 63e38bd18d..8c5c7fe657 100644
--- a/studio/backend/core/data_recipe/jobs/worker.py
+++ b/studio/backend/core/data_recipe/jobs/worker.py
@@ -21,6 +21,15 @@ from ..service import build_config_builder, create_data_designer
 from utils.paths import ensure_dir, recipe_datasets_root
 
 _ARTIFACT_ROOT = recipe_datasets_root()
+_RE_GITHUB_CURSOR = re.compile(r"\bcursor=[^\s,]+")
+_RE_SECRET_TOKEN = re.compile(
+    r"\b(?:(?:ghp|gho|ghu|ghs|ghr|github_pat)_[A-Za-z0-9_]+|sk-unsloth-[A-Za-z0-9]+)"
+)
+
+
+def _sanitize_log_message(message: str) -> str:
+    message = _RE_GITHUB_CURSOR.sub("cursor=<redacted>", message)
+    return _RE_SECRET_TOKEN.sub("<redacted-token>", message)
 
 
 class _QueueLogHandler(logging.Handler):
@@ -35,7 +44,7 @@ class _QueueLogHandler(logging.Handler):
                 "ts": record.created,
                 "level": record.levelname,
                 "logger": record.name,
-                "message": record.getMessage(),
+                "message": _sanitize_log_message(record.getMessage()),
             }
             self._q.put(event)
         except (OSError, RuntimeError, ValueError):
@@ -119,10 +128,16 @@ def run_job_process(
         # Attach queue logger directly to `data_designer` so parser events survive root resets.
         handler = _QueueLogHandler(event_queue)
         handler.setLevel(logging.INFO)
-        data_designer_logger = logging.getLogger("data_designer")
-        data_designer_logger.addHandler(handler)
-        data_designer_logger.setLevel(logging.INFO)
-        data_designer_logger.propagate = True
+        for logger_name in (
+            "data_designer",
+            "scraper",
+            "gh_client",
+            "data_designer_github_repo_seed",
+        ):
+            logger = logging.getLogger(logger_name)
+            logger.addHandler(handler)
+            logger.setLevel(logging.INFO)
+            logger.propagate = True
 
         if run_config_raw:
             designer.set_run_config(RunConfig.model_validate(run_config_raw))
@@ -180,8 +195,8 @@ def run_job_process(
             {
                 "type": EVENT_JOB_ERROR,
                 "ts": time.time(),
-                "error": str(exc),
-                "stack": traceback.format_exc(limit = 20),
+                "error": _sanitize_log_message(str(exc)),
+                "stack": _sanitize_log_message(traceback.format_exc(limit = 20)),
             }
         )
 
diff --git a/studio/backend/core/inference/llama_cpp.py b/studio/backend/core/inference/llama_cpp.py
index 81ff9ae4e5..33722f103c 100644
--- a/studio/backend/core/inference/llama_cpp.py
+++ b/studio/backend/core/inference/llama_cpp.py
@@ -1618,8 +1618,10 @@ class LlamaCppBackend:
                 # Model fits on selected GPU(s) -- offload all layers
                 cmd.extend(["-ngl", "-1"])
 
-            if n_threads is not None:
-                cmd.extend(["--threads", str(n_threads)])
+            # -1 = llama.cpp auto-detect (physical cores). Pass explicitly so we
+            # do not inherit llama-server's internal default, which has historically
+            # varied (hardware concurrency incl. hyperthreads on some builds).
+            cmd.extend(["--threads", str(n_threads if n_threads is not None else -1)])
 
             # Always enable Jinja chat template rendering for proper template support
             cmd.extend(["--jinja"])
diff --git a/studio/backend/plugins/data-designer-github-repo-seed/README.md b/studio/backend/plugins/data-designer-github-repo-seed/README.md
new file mode 100644
index 0000000000..346d94b305
--- /dev/null
+++ b/studio/backend/plugins/data-designer-github-repo-seed/README.md
@@ -0,0 +1,73 @@
+# data-designer-github-repo-seed
+
+A Data Designer seed-reader plugin for **Unsloth Studio** that scrapes real
+GitHub data (issues, pull requests, commits) from one or more repositories
+and hands it to the recipe pipeline as a seed dataset.
+
+Designed to ship with Studio as a default seed source so any user with a
+GitHub token can build training datasets straight from live repos.
+
+## What it does
+
+Given a list of `owner/name` repos, a GitHub token, and a per-resource
+`limit`, the plugin uses GitHub's GraphQL API to fetch issues, pull
+requests, and/or commits, with labels, state, authors, and the first N
+comments of each item, and materialises a single JSONL with uniform
+columns so the rest of the recipe (LLM text / LLM structured / processors)
+can treat it like any other seed table.
+
+| Column        | Description                                    |
+|---------------|------------------------------------------------|
+| `item_type`   | `issue` / `pull` / `commit`                    |
+| `repo`        | `owner/name`                                   |
+| `number`      | Issue/PR number, or commit SHA                 |
+| `title`       | Title (or commit message headline)             |
+| `body`        | Issue/PR body (or full commit message)         |
+| `state`       | `OPEN` / `CLOSED` / `MERGED` (empty for commit)|
+| `author`      | GitHub login of the author                     |
+| `created_at`  | ISO8601                                        |
+| `closed_at`   | ISO8601 (empty for commits)                    |
+| `url`         | Permalink                                      |
+| `labels`      | List of label names                            |
+| `comments`    | First N comments concatenated                  |
+
+## Usage in a recipe
+
+```json
+{
+  "seed_config": {
+    "source": {
+      "seed_type": "github_repo",
+      "repos": ["unslothai/unsloth", "unslothai/unsloth-zoo"],
+      "token": "",
+      "item_types": ["issues", "pulls"],
+      "limit": 100,
+      "include_comments": true,
+      "max_comments_per_item": 30
+    },
+    "sampling_strategy": "shuffle",
+    "selection_strategy": null
+  }
+}
+```
+
+Leave `token` empty to fall back to the server's `GH_TOKEN` / `GITHUB_TOKEN`
+environment variable, useful when the recipe is published and shouldn't
+carry a secret.
+
+## Auth
+
+A GitHub personal access token with `public_repo` scope is enough for public
+repositories; `repo` scope is required for private ones. GraphQL requests
+are rate-limit aware: the client inspects `x-ratelimit-*` headers and
+sleeps until reset when the budget drops below a safety threshold.
+
+## Install
+
+Shipped as a default Studio plugin. For development:
+
+```bash
+pip install -e .
+```
+
+Registered automatically via the `data_designer.plugins` entry point.
diff --git a/studio/backend/plugins/data-designer-github-repo-seed/pyproject.toml b/studio/backend/plugins/data-designer-github-repo-seed/pyproject.toml
new file mode 100644
index 0000000000..e232adc60c
--- /dev/null
+++ b/studio/backend/plugins/data-designer-github-repo-seed/pyproject.toml
@@ -0,0 +1,25 @@
+# SPDX-License-Identifier: AGPL-3.0-only
+# Copyright 2026-present the Unsloth AI Inc. team. All rights reserved. See /studio/LICENSE.AGPL-3.0
+
+[build-system]
+requires = ["setuptools>=68", "wheel"]
+build-backend = "setuptools.build_meta"
+
+[project]
+name = "data-designer-github-repo-seed"
+version = "0.1.0"
+description = "Unsloth Studio seed plugin that scrapes GitHub issues, PRs, and commits."
+requires-python = ">=3.11"
+dependencies = [
+  "data-designer-engine>=0.5.4,<0.6",
+  "requests>=2.31",
+]
+
+[project.entry-points."data_designer.plugins"]
+github_repo_seed = "data_designer_github_repo_seed.plugin:github_repo_seed_plugin"
+
+[tool.setuptools]
+package-dir = {"" = "src"}
+
+[tool.setuptools.packages.find]
+where = ["src"]
diff --git a/studio/backend/plugins/data-designer-github-repo-seed/src/data_designer_github_repo_seed/__init__.py b/studio/backend/plugins/data-designer-github-repo-seed/src/data_designer_github_repo_seed/__init__.py
new file mode 100644
index 0000000000..f57af4c6c3
--- /dev/null
+++ b/studio/backend/plugins/data-designer-github-repo-seed/src/data_designer_github_repo_seed/__init__.py
@@ -0,0 +1,7 @@
+# SPDX-License-Identifier: AGPL-3.0-only
+# Copyright 2026-present the Unsloth AI Inc. team. All rights reserved. See /studio/LICENSE.AGPL-3.0
+
+# Intentionally empty. Data-designer loads submodules lazily via qualified names
+# (impl_qualified_name / config_qualified_name in plugin.py), so importing this
+# package must NOT touch modules that depend on data_designer.engine.* during
+# Studio's bootstrap (circular import).
diff --git a/studio/backend/plugins/data-designer-github-repo-seed/src/data_designer_github_repo_seed/config.py b/studio/backend/plugins/data-designer-github-repo-seed/src/data_designer_github_repo_seed/config.py
new file mode 100644
index 0000000000..6b347c4f83
--- /dev/null
+++ b/studio/backend/plugins/data-designer-github-repo-seed/src/data_designer_github_repo_seed/config.py
@@ -0,0 +1,64 @@
+# SPDX-License-Identifier: AGPL-3.0-only
+# Copyright 2026-present the Unsloth AI Inc. team. All rights reserved. See /studio/LICENSE.AGPL-3.0
+
+from __future__ import annotations
+
+from typing import Literal
+
+from pydantic import Field, field_validator, model_validator
+
+from data_designer.config.seed_source import SeedSource
+
+
+class GitHubRepoSeedSource(SeedSource):
+    seed_type: Literal["github_repo"] = "github_repo"
+
+    repos: list[str] = Field(
+        default_factory = list,
+        description = "List of GitHub repositories to scrape, each in `owner/name` form.",
+    )
+    token: str = Field(
+        default = "",
+        description = "Personal access token. Leave blank to read GH_TOKEN / GITHUB_TOKEN from env at run time.",
+    )
+    item_types: list[Literal["issues", "pulls", "commits"]] = Field(
+        default = ["issues", "pulls"],
+        description = "Which GitHub item types to fetch per repo.",
+    )
+    limit: int = Field(
+        default = 100,
+        ge = 1,
+        le = 5000,
+        description = "Maximum items per repo per item type (e.g. limit=100 + ['issues','pulls'] => up to 200 items per repo).",
+    )
+    include_comments: bool = Field(
+        default = True,
+        description = "Fetch the first N comments of each issue/PR and include them in the `comments` column.",
+    )
+    max_comments_per_item: int = Field(default = 30, ge = 0, le = 200)
+
+    @field_validator("repos")
+    @classmethod
+    def _validate_repos(cls, v: list[str]) -> list[str]:
+        out: list[str] = []
+        for r in v or []:
+            r = r.strip()
+            if not r:
+                continue
+            if r.count("/") != 1 or not all(r.split("/")):
+                raise ValueError(f"Each repo must be `owner/name`; got {r!r}")
+            out.append(r)
+        return out
+
+    @field_validator("item_types")
+    @classmethod
+    def _validate_item_types(cls, v: list[str]) -> list[str]:
+        if not v:
+            raise ValueError("item_types must not be empty")
+        return list(dict.fromkeys(v))
+
+    @model_validator(mode = "after")
+    def _ensure_repos(self) -> "GitHubRepoSeedSource":
+        if not self.repos:
+            raise ValueError("At least one repo is required")
+        return self
diff --git a/studio/backend/plugins/data-designer-github-repo-seed/src/data_designer_github_repo_seed/impl.py b/studio/backend/plugins/data-designer-github-repo-seed/src/data_designer_github_repo_seed/impl.py
new file mode 100644
index 0000000000..5a38e26d6b
--- /dev/null
+++ b/studio/backend/plugins/data-designer-github-repo-seed/src/data_designer_github_repo_seed/impl.py
@@ -0,0 +1,83 @@
+# SPDX-License-Identifier: AGPL-3.0-only
+# Copyright 2026-present the Unsloth AI Inc. team. All rights reserved. See /studio/LICENSE.AGPL-3.0
+
+from __future__ import annotations
+
+import hashlib
+import tempfile
+import threading
+from pathlib import Path
+from typing import Optional
+
+import data_designer.lazy_heavy_imports as lazy
+from data_designer.engine.resources.seed_reader import SeedReader
+
+from .config import GitHubRepoSeedSource
+from .scraper import ScrapeConfig, materialize_to_jsonl
+
+
+# In-process cache mapping a stable config signature to the JSONL materialization
+# path. A single recipe job invokes the seed reader multiple times (validation,
+# preview, per-column sampling), and the default flow re-scrapes the repo on
+# every call: for a 2-repo preview that is ~15s of redundant GitHub GraphQL
+# traffic before any generation fires. Memoize the materialization so the second
+# and third passes reuse the file the first pass wrote. Cache key excludes the
+# raw token and uses a short SHA-256 digest so token values never hit memory
+# twice and token rotation invalidates cleanly.
+_SCRAPE_CACHE: dict[tuple, str] = {}
+_SCRAPE_CACHE_LOCK = threading.Lock()
+
+
+def _scrape_cache_key(cfg: ScrapeConfig) -> tuple:
+    token_digest = hashlib.sha256(
+        (cfg.token or "").encode("utf-8"),
+    ).hexdigest()[:16]
+    return (
+        tuple(cfg.repos),
+        tuple(cfg.item_types),
+        cfg.limit,
+        bool(cfg.include_comments),
+        cfg.max_comments_per_item,
+        token_digest,
+    )
+
+
+def _lookup_cached_scrape(key: tuple) -> Optional[str]:
+    with _SCRAPE_CACHE_LOCK:
+        path = _SCRAPE_CACHE.get(key)
+    if path and Path(path).exists():
+        return path
+    # Stale entry (tmp cleanup, user restarted, ...); drop it so the caller
+    # materializes a fresh file rather than returning a dangling path.
+    if path:
+        with _SCRAPE_CACHE_LOCK:
+            _SCRAPE_CACHE.pop(key, None)
+    return None
+
+
+def _store_cached_scrape(key: tuple, path: str) -> None:
+    with _SCRAPE_CACHE_LOCK:
+        _SCRAPE_CACHE[key] = path
+
+
+class GitHubRepoSeedReader(SeedReader[GitHubRepoSeedSource]):
+    def create_duckdb_connection(self):
+        return lazy.duckdb.connect()
+
+    def get_dataset_uri(self) -> str:
+        out_dir = Path(tempfile.gettempdir()) / "studio-github-repo-seed"
+        cfg = ScrapeConfig(
+            repos = list(self.source.repos),
+            token = self.source.token,
+            item_types = list(self.source.item_types),
+            limit = self.source.limit,
+            include_comments = self.source.include_comments,
+            max_comments_per_item = self.source.max_comments_per_item,
+        )
+        cache_key = _scrape_cache_key(cfg)
+        cached_path = _lookup_cached_scrape(cache_key)
+        if cached_path is not None:
+            return cached_path
+        path = materialize_to_jsonl(cfg, out_dir)
+        _store_cached_scrape(cache_key, str(path))
+        return str(path)
diff --git a/studio/backend/plugins/data-designer-github-repo-seed/src/data_designer_github_repo_seed/plugin.py b/studio/backend/plugins/data-designer-github-repo-seed/src/data_designer_github_repo_seed/plugin.py
new file mode 100644
index 0000000000..f87dbd0507
--- /dev/null
+++ b/studio/backend/plugins/data-designer-github-repo-seed/src/data_designer_github_repo_seed/plugin.py
@@ -0,0 +1,10 @@
+# SPDX-License-Identifier: AGPL-3.0-only
+# Copyright 2026-present the Unsloth AI Inc. team. All rights reserved. See /studio/LICENSE.AGPL-3.0
+
+from data_designer.plugins.plugin import Plugin, PluginType
+
+github_repo_seed_plugin = Plugin(
+    impl_qualified_name = "data_designer_github_repo_seed.impl.GitHubRepoSeedReader",
+    config_qualified_name = "data_designer_github_repo_seed.config.GitHubRepoSeedSource",
+    plugin_type = PluginType.SEED_READER,
+)
diff --git a/studio/backend/plugins/data-designer-github-repo-seed/src/data_designer_github_repo_seed/scraper.py b/studio/backend/plugins/data-designer-github-repo-seed/src/data_designer_github_repo_seed/scraper.py
new file mode 100644
index 0000000000..d768fe37be
--- /dev/null
+++ b/studio/backend/plugins/data-designer-github-repo-seed/src/data_designer_github_repo_seed/scraper.py
@@ -0,0 +1,236 @@
+# SPDX-License-Identifier: AGPL-3.0-only
+# Copyright 2026-present the Unsloth AI Inc. team. All rights reserved. See /studio/LICENSE.AGPL-3.0
+
+"""Multi-repo GitHub scraper for the Studio seed plugin.
+
+Drives the GraphQL-based scraper in `scraper_impl/` per repo. Each repo is
+scraped with a trial_limits cap so we stop at `limit` items per resource.
+After scraping, we read the per-resource JSONL shards and flatten them into
+a single unified JSONL with stable columns (`item_type`, `repo`, `number`,
+`title`, `body`, ...).
+"""
+
+from __future__ import annotations
+
+import json
+import os
+import sys
+import time
+import uuid
+from dataclasses import dataclass
+from pathlib import Path
+
+# Defer scraper_impl imports until `scrape()` runs with a resolved token.
+_IMPL_DIR = Path(__file__).parent / "scraper_impl"
+
+
+def _ensure_impl_on_path() -> None:
+    if str(_IMPL_DIR) not in sys.path:
+        sys.path.insert(0, str(_IMPL_DIR))
+
+
+def _load_impl():
+    _ensure_impl_on_path()
+    import importlib
+
+    gh_client = importlib.import_module("gh_client")  # type: ignore
+    scraper_mod = importlib.import_module("scraper")  # type: ignore
+    return gh_client.GitHubClient, scraper_mod.RepoScraper
+
+
+@dataclass
+class ScrapeConfig:
+    repos: list[str]
+    token: str
+    item_types: list[str]
+    limit: int
+    include_comments: bool
+    max_comments_per_item: int
+
+
+def _resolve_token(token: str) -> str:
+    tok = token or os.environ.get("GH_TOKEN", "") or os.environ.get("GITHUB_TOKEN", "")
+    if not tok:
+        raise ValueError(
+            "GitHub token is required. Set it in the recipe config or the GH_TOKEN / GITHUB_TOKEN env var."
+        )
+    return tok
+
+
+def _read_jsonl(path: Path, max_rows: int | None = None):
+    if not path.exists():
+        return
+    with path.open(encoding = "utf-8") as f:
+        for i, line in enumerate(f):
+            if not line.strip():
+                continue
+            if max_rows is not None and i >= max_rows:
+                return
+            try:
+                yield json.loads(line)
+            except json.JSONDecodeError:
+                continue
+
+
+def _flatten_issue_row(r: dict, repo: str, include_comments: bool, max_c: int) -> dict:
+    labels = [
+        l.get("name")
+        for l in (r.get("labels", {}) or {}).get("nodes", [])
+        if l.get("name")
+    ]
+    comments_nodes = (r.get("comments") or {}).get("nodes") or []
+    comments_text = ""
+    if include_comments and comments_nodes:
+        kept = comments_nodes[:max_c]
+        comments_text = "\n\n".join(
+            f"[{(c.get('author') or {}).get('login', '?')}]: {c.get('body') or ''}"
+            for c in kept
+        )
+    return {
+        "item_type": "issue",
+        "repo": repo,
+        "number": r.get("number"),
+        "title": r.get("title") or "",
+        "body": r.get("body") or "",
+        "state": r.get("state") or "",
+        "author": (r.get("author") or {}).get("login", ""),
+        "created_at": r.get("createdAt") or "",
+        "closed_at": r.get("closedAt") or "",
+        "url": r.get("url") or r.get("permalink") or "",
+        "labels": labels,
+        "comments": comments_text,
+    }
+
+
+def _flatten_pr_row(r: dict, repo: str, include_comments: bool, max_c: int) -> dict:
+    labels = [
+        l.get("name")
+        for l in (r.get("labels", {}) or {}).get("nodes", [])
+        if l.get("name")
+    ]
+    comments_nodes = (r.get("comments") or {}).get("nodes") or []
+    comments_text = ""
+    if include_comments and comments_nodes:
+        kept = comments_nodes[:max_c]
+        comments_text = "\n\n".join(
+            f"[{(c.get('author') or {}).get('login', '?')}]: {c.get('body') or ''}"
+            for c in kept
+        )
+    return {
+        "item_type": "pull",
+        "repo": repo,
+        "number": r.get("number"),
+        "title": r.get("title") or "",
+        "body": r.get("body") or "",
+        "state": r.get("state") or "",
+        "author": (r.get("author") or {}).get("login", ""),
+        "created_at": r.get("createdAt") or "",
+        "closed_at": r.get("closedAt") or "",
+        "url": r.get("url") or r.get("permalink") or "",
+        "labels": labels,
+        "comments": comments_text,
+    }
+
+
+def _flatten_commit_row(r: dict, repo: str) -> dict:
+    msg = r.get("messageHeadline") or r.get("message") or ""
+    body = r.get("messageBody") or r.get("message") or msg
+    author = r.get("author") or {}
+    return {
+        "item_type": "commit",
+        "repo": repo,
+        "number": r.get("oid") or r.get("sha") or "",
+        "title": msg,
+        "body": body,
+        "state": "",
+        "author": (author.get("user") or {}).get("login") or author.get("name", ""),
+        "created_at": (author.get("date") or r.get("committedDate") or ""),
+        "closed_at": "",
+        "url": r.get("url") or "",
+        "labels": [],
+        "comments": "",
+    }
+
+
+def scrape(cfg: ScrapeConfig, base_dir: Path):
+    token = _resolve_token(cfg.token)
+    GitHubClient, RepoScraper = _load_impl()
+    client = GitHubClient(token = token)
+    base_dir.mkdir(parents = True, exist_ok = True)
+
+    # Per-resource trial limits. limit <= 0 means "all": use a very large cap.
+    effective_limit = cfg.limit if cfg.limit and cfg.limit > 0 else 1_000_000
+    trial_limits: dict[str, int] = {}
+    if "issues" in cfg.item_types:
+        trial_limits["issues"] = effective_limit
+    if "pulls" in cfg.item_types:
+        trial_limits["pull_requests"] = effective_limit
+    if "commits" in cfg.item_types:
+        trial_limits["commits"] = effective_limit
+
+    all_rows: list[dict] = []
+    for repo in cfg.repos:
+        owner, name = repo.split("/", 1)
+        scraper = RepoScraper(
+            owner = owner,
+            name = name,
+            base_dir = base_dir,
+            client = client,
+            trial_limits = trial_limits,
+            light = True,
+        )
+        try:
+            repo_meta = scraper.scrape_repo_meta()
+            if "issues" in cfg.item_types:
+                scraper.scrape_issues()
+            if "pulls" in cfg.item_types:
+                scraper.scrape_prs()
+            if "commits" in cfg.item_types:
+                default_ref = repo_meta.get("defaultBranchRef") or {}
+                default_branch = (
+                    default_ref.get("name") if isinstance(default_ref, dict) else None
+                )
+                branch = (
+                    f"refs/heads/{default_branch}"
+                    if default_branch
+                    else "refs/heads/main"
+                )
+                scraper.scrape_commits(branch = branch)
+        finally:
+            scraper.close()
+
+        read_cap = cfg.limit if cfg.limit and cfg.limit > 0 else None
+        repo_dir = base_dir / f"{owner}__{name}"
+        if "issues" in cfg.item_types:
+            for row in _read_jsonl(repo_dir / "issues.jsonl", read_cap):
+                all_rows.append(
+                    _flatten_issue_row(
+                        row, repo, cfg.include_comments, cfg.max_comments_per_item
+                    )
+                )
+        if "pulls" in cfg.item_types:
+            for row in _read_jsonl(repo_dir / "pull_requests.jsonl", read_cap):
+                all_rows.append(
+                    _flatten_pr_row(
+                        row, repo, cfg.include_comments, cfg.max_comments_per_item
+                    )
+                )
+        if "commits" in cfg.item_types:
+            for row in _read_jsonl(repo_dir / "commits.jsonl", read_cap):
+                all_rows.append(_flatten_commit_row(row, repo))
+
+    return all_rows
+
+
+def materialize_to_jsonl(cfg: ScrapeConfig, out_dir: Path) -> Path:
+    out_dir.mkdir(parents = True, exist_ok = True)
+    tag = "-".join(r.replace("/", "__") for r in cfg.repos)[:120]
+    kinds = "-".join(cfg.item_types)
+    run_id = f"{int(time.time())}-{uuid.uuid4().hex[:12]}"
+    fname = f"github_{tag}__{kinds}__{cfg.limit}_{run_id}.jsonl"
+    out = out_dir / fname
+    rows = scrape(cfg, out_dir / "raw-runs" / run_id)
+    with out.open("w", encoding = "utf-8") as f:
+        for r in rows:
+            f.write(json.dumps(r, ensure_ascii = False) + "\n")
+    return out
diff --git a/studio/backend/plugins/data-designer-github-repo-seed/src/data_designer_github_repo_seed/scraper_impl/__init__.py b/studio/backend/plugins/data-designer-github-repo-seed/src/data_designer_github_repo_seed/scraper_impl/__init__.py
new file mode 100644
index 0000000000..32014236c6
--- /dev/null
+++ b/studio/backend/plugins/data-designer-github-repo-seed/src/data_designer_github_repo_seed/scraper_impl/__init__.py
@@ -0,0 +1,2 @@
+# SPDX-License-Identifier: AGPL-3.0-only
+# Copyright 2026-present the Unsloth AI Inc. team. All rights reserved. See /studio/LICENSE.AGPL-3.0
diff --git a/studio/backend/plugins/data-designer-github-repo-seed/src/data_designer_github_repo_seed/scraper_impl/gh_client.py b/studio/backend/plugins/data-designer-github-repo-seed/src/data_designer_github_repo_seed/scraper_impl/gh_client.py
new file mode 100644
index 0000000000..dd2de2f5ce
--- /dev/null
+++ b/studio/backend/plugins/data-designer-github-repo-seed/src/data_designer_github_repo_seed/scraper_impl/gh_client.py
@@ -0,0 +1,248 @@
+# SPDX-License-Identifier: AGPL-3.0-only
+# Copyright 2026-present the Unsloth AI Inc. team. All rights reserved. See /studio/LICENSE.AGPL-3.0
+
+"""GitHub API client with rate-limit awareness, retry, and dual REST/GraphQL support."""
+
+from __future__ import annotations
+
+import json
+import os
+import time
+import logging
+from typing import Any, Dict, Iterable, Iterator, List, Optional
+
+import requests
+
+log = logging.getLogger("gh_client")
+
+GRAPHQL_URL = "https://api.github.com/graphql"
+REST_BASE = "https://api.github.com"
+
+BASE_HEADERS = {
+    "Accept": "application/vnd.github+json",
+    "X-GitHub-Api-Version": "2022-11-28",
+    "User-Agent": "github-data-gatherer/1.0",
+}
+
+
+class RateLimitError(Exception):
+    pass
+
+
+class GitHubClient:
+    def __init__(
+        self,
+        min_remaining_graphql: int = 100,
+        min_remaining_rest: int = 100,
+        token: str | None = None,
+    ):
+        token = token or os.environ.get("GH_TOKEN") or os.environ.get("GITHUB_TOKEN")
+        if not token:
+            raise RuntimeError("GH_TOKEN not set in environment")
+        self.session = requests.Session()
+        self.session.headers.update(
+            {**BASE_HEADERS, "Authorization": f"Bearer {token}"}
+        )
+        self.min_remaining_graphql = min_remaining_graphql
+        self.min_remaining_rest = min_remaining_rest
+        self.graphql_remaining: Optional[int] = None
+        self.graphql_reset: Optional[int] = None
+        self.rest_remaining: Optional[int] = None
+        self.rest_reset: Optional[int] = None
+        self.calls_graphql = 0
+        self.calls_rest = 0
+        self.retry_count = 0
+
+    def _sleep_until(self, reset_ts: int, buffer_s: int = 10) -> None:
+        now = int(time.time())
+        wait = max(0, reset_ts - now) + buffer_s
+        log.warning("Rate limit hit. Sleeping %ds until reset.", wait)
+        time.sleep(wait)
+
+    def _check_rate_and_wait(self, kind: str) -> None:
+        if kind == "graphql":
+            remaining = self.graphql_remaining
+            reset = self.graphql_reset
+            min_remaining = self.min_remaining_graphql
+        else:
+            remaining = self.rest_remaining
+            reset = self.rest_reset
+            min_remaining = self.min_remaining_rest
+        if remaining is not None and remaining < min_remaining:
+            if reset:
+                self._sleep_until(reset)
+                # Reset remaining so we don't spin
+                if kind == "graphql":
+                    self.graphql_remaining = None
+                else:
+                    self.rest_remaining = None
+
+    def graphql(
+        self,
+        query: str,
+        variables: Optional[Dict[str, Any]] = None,
+        max_retries: int = 20,
+    ) -> Dict[str, Any]:
+        self._check_rate_and_wait("graphql")
+        backoff = 2
+        last_err = None
+        for attempt in range(max_retries):
+            try:
+                r = self.session.post(
+                    GRAPHQL_URL,
+                    json = {"query": query, "variables": variables or {}},
+                    timeout = 120,
+                )
+                self.calls_graphql += 1
+                # Update rate info from response headers
+                rem = r.headers.get("X-RateLimit-Remaining")
+                rst = r.headers.get("X-RateLimit-Reset")
+                if rem is not None:
+                    try:
+                        self.graphql_remaining = int(rem)
+                    except ValueError:
+                        pass
+                if rst is not None:
+                    try:
+                        self.graphql_reset = int(rst)
+                    except ValueError:
+                        pass
+                if r.status_code in (502, 503, 504):
+                    log.warning("GraphQL %s transient, retrying", r.status_code)
+                    time.sleep(backoff)
+                    backoff = min(backoff * 2, 60)
+                    continue
+                if r.status_code == 403 or r.status_code == 429:
+                    # Check for secondary/abuse
+                    retry_after = r.headers.get("Retry-After")
+                    if retry_after:
+                        t = int(retry_after)
+                        log.warning("Secondary rate limit. Sleep %ds.", t)
+                        time.sleep(t + 2)
+                        continue
+                    if self.graphql_reset:
+                        self._sleep_until(self.graphql_reset)
+                        continue
+                    time.sleep(60)
+                    continue
+                r.raise_for_status()
+                data = r.json()
+                if "errors" in data and data["errors"]:
+                    # Surface errors but allow partial data
+                    errs = data["errors"]
+                    # Retry on RATE_LIMITED
+                    for e in errs:
+                        if e.get("type") == "RATE_LIMITED":
+                            self._sleep_until(
+                                (self.graphql_reset or int(time.time()) + 60)
+                            )
+                            break
+                    else:
+                        # No rate-limit error, log and return partial
+                        log.warning("GraphQL errors: %s", json.dumps(errs)[:400])
+                        return data
+                    continue
+                return data
+            except requests.RequestException as e:
+                last_err = e
+                log.warning("GraphQL network error: %s. Retry.", e)
+                time.sleep(backoff)
+                backoff = min(backoff * 2, 60)
+        raise RuntimeError(f"GraphQL failed after {max_retries} retries: {last_err}")
+
+    def rest(
+        self,
+        method: str,
+        path: str,
+        params: Optional[Dict[str, Any]] = None,
+        json_body: Optional[Dict[str, Any]] = None,
+        max_retries: int = 6,
+    ) -> requests.Response:
+        self._check_rate_and_wait("rest")
+        if path.startswith("http"):
+            url = path
+        else:
+            url = REST_BASE + path
+        backoff = 2
+        last_err = None
+        for attempt in range(max_retries):
+            try:
+                r = self.session.request(
+                    method, url, params = params, json = json_body, timeout = 120
+                )
+                self.calls_rest += 1
+                rem = r.headers.get("X-RateLimit-Remaining")
+                rst = r.headers.get("X-RateLimit-Reset")
+                if rem is not None:
+                    try:
+                        self.rest_remaining = int(rem)
+                    except ValueError:
+                        pass
+                if rst is not None:
+                    try:
+                        self.rest_reset = int(rst)
+                    except ValueError:
+                        pass
+                if r.status_code in (502, 503, 504):
+                    log.warning("REST %s transient, retrying", r.status_code)
+                    time.sleep(backoff)
+                    backoff = min(backoff * 2, 60)
+                    continue
+                if r.status_code in (403, 429):
+                    retry_after = r.headers.get("Retry-After")
+                    if retry_after:
+                        t = int(retry_after)
+                        log.warning("Secondary rate limit on REST. Sleep %ds.", t)
+                        time.sleep(t + 2)
+                        continue
+                    # Check if primary rate
+                    if self.rest_remaining == 0 and self.rest_reset:
+                        self._sleep_until(self.rest_reset)
+                        continue
+                    log.warning("REST 403/429, sleep 60")
+                    time.sleep(60)
+                    continue
+                return r
+            except requests.RequestException as e:
+                last_err = e
+                log.warning("REST network error: %s. Retry.", e)
+                time.sleep(backoff)
+                backoff = min(backoff * 2, 60)
+        raise RuntimeError(f"REST failed after {max_retries} retries: {last_err}")
+
+    def rest_paginate(
+        self, path: str, params: Optional[Dict[str, Any]] = None, per_page: int = 100
+    ) -> Iterator[dict]:
+        params = dict(params or {})
+        params.setdefault("per_page", per_page)
+        url = path
+        while True:
+            r = self.rest("GET", url, params = params if url == path else None)
+            if r.status_code != 200:
+                log.error(
+                    "REST paginate got %s at %s: %s", r.status_code, url, r.text[:200]
+                )
+                return
+            items = r.json()
+            if isinstance(items, dict):
+                # Some endpoints return dict with list field
+                items = items.get("items", [])
+            for it in items:
+                yield it
+            # Follow link header
+            link = r.headers.get("Link", "")
+            nxt = None
+            for part in link.split(","):
+                if 'rel="next"' in part:
+                    nxt = part.split(";")[0].strip().strip("<>")
+                    break
+            if not nxt:
+                return
+            url = nxt
+            params = None
+
+    def rate_snapshot(self) -> Dict[str, Any]:
+        r = self.rest("GET", "/rate_limit")
+        if r.status_code == 200:
+            return r.json()
+        return {}
diff --git a/studio/backend/plugins/data-designer-github-repo-seed/src/data_designer_github_repo_seed/scraper_impl/queries.py b/studio/backend/plugins/data-designer-github-repo-seed/src/data_designer_github_repo_seed/scraper_impl/queries.py
new file mode 100644
index 0000000000..9dc7613db5
--- /dev/null
+++ b/studio/backend/plugins/data-designer-github-repo-seed/src/data_designer_github_repo_seed/scraper_impl/queries.py
@@ -0,0 +1,685 @@
+# SPDX-License-Identifier: AGPL-3.0-only
+# Copyright 2026-present the Unsloth AI Inc. team. All rights reserved. See /studio/LICENSE.AGPL-3.0
+
+"""GraphQL queries for GitHub data scraping.
+
+GitHub's GraphQL rejects queries that define unused fragments, so each query
+only includes the fragments it actually references.
+"""
+
+# ---- Fragments (kept as raw strings, composed per query) ----
+F_ACTOR = """
+fragment ActorFields on Actor {
+  __typename
+  login
+  url
+  avatarUrl
+  ... on User { id databaseId name }
+  ... on Bot { id databaseId }
+  ... on Organization { id databaseId name }
+}
+"""
+
+F_LABEL = """
+fragment LabelFields on Label {
+  id
+  name
+  color
+  description
+  createdAt
+}
+"""
+
+F_TIMELINE = """
+fragment TimelineItem on IssueTimelineItems {
+  __typename
+  ... on Node { id }
+  ... on AddedToProjectEvent { createdAt actor { ...ActorFields } }
+  ... on AssignedEvent { createdAt actor { ...ActorFields } assignee { __typename ... on User { login } ... on Bot { login } } }
+  ... on ClosedEvent { createdAt actor { ...ActorFields } stateReason closer { __typename ... on Commit { oid url } ... on PullRequest { number url } } }
+  ... on CommentDeletedEvent { createdAt actor { ...ActorFields } }
+  ... on ConnectedEvent { createdAt actor { ...ActorFields } source { __typename ... on Issue { number url repository { nameWithOwner } } ... on PullRequest { number url repository { nameWithOwner } } } subject { __typename ... on Issue { number url } ... on PullRequest { number url } } }
+  ... on ConvertedNoteToIssueEvent { createdAt actor { ...ActorFields } }
+  ... on CrossReferencedEvent { createdAt actor { ...ActorFields } isCrossRepository willCloseTarget source { __typename ... on Issue { number url repository { nameWithOwner } title } ... on PullRequest { number url repository { nameWithOwner } title } } }
+  ... on DemilestonedEvent { createdAt actor { ...ActorFields } milestoneTitle }
+  ... on DisconnectedEvent { createdAt actor { ...ActorFields } subject { __typename ... on Issue { number url } ... on PullRequest { number url } } source { __typename ... on Issue { number url } ... on PullRequest { number url } } }
+  ... on IssueComment { id databaseId createdAt updatedAt author { ...ActorFields } body url reactionGroups { content reactors { totalCount } } }
+  ... on LabeledEvent { createdAt actor { ...ActorFields } label { name color } }
+  ... on LockedEvent { createdAt actor { ...ActorFields } lockReason }
+  ... on MarkedAsDuplicateEvent { createdAt actor { ...ActorFields } canonical { __typename ... on Issue { number url } ... on PullRequest { number url } } }
+  ... on MentionedEvent { createdAt actor { ...ActorFields } }
+  ... on MilestonedEvent { createdAt actor { ...ActorFields } milestoneTitle }
+  ... on MovedColumnsInProjectEvent { createdAt actor { ...ActorFields } }
+  ... on PinnedEvent { createdAt actor { ...ActorFields } }
+  ... on ReferencedEvent { createdAt actor { ...ActorFields } commit { oid url } commitRepository { nameWithOwner } }
+  ... on RemovedFromProjectEvent { createdAt actor { ...ActorFields } }
+  ... on RenamedTitleEvent { createdAt actor { ...ActorFields } previousTitle currentTitle }
+  ... on ReopenedEvent { createdAt actor { ...ActorFields } }
+  ... on SubscribedEvent { createdAt actor { ...ActorFields } }
+  ... on TransferredEvent { createdAt actor { ...ActorFields } fromRepository { nameWithOwner } }
+  ... on UnassignedEvent { createdAt actor { ...ActorFields } assignee { __typename ... on User { login } ... on Bot { login } } }
+  ... on UnlabeledEvent { createdAt actor { ...ActorFields } label { name color } }
+  ... on UnlockedEvent { createdAt actor { ...ActorFields } }
+  ... on UnmarkedAsDuplicateEvent { createdAt actor { ...ActorFields } }
+  ... on UnpinnedEvent { createdAt actor { ...ActorFields } }
+  ... on UnsubscribedEvent { createdAt actor { ...ActorFields } }
+  ... on UserBlockedEvent { createdAt actor { ...ActorFields } blockDuration }
+}
+"""
+
+F_PR_TIMELINE = """
+fragment PRTimelineItem on PullRequestTimelineItems {
+  __typename
+  ... on Node { id }
+  ... on AssignedEvent { createdAt actor { ...ActorFields } assignee { __typename ... on User { login } ... on Bot { login } } }
+  ... on AutoMergeDisabledEvent { createdAt actor { ...ActorFields } reason }
+  ... on AutoMergeEnabledEvent { createdAt actor { ...ActorFields } }
+  ... on AutoRebaseEnabledEvent { createdAt actor { ...ActorFields } }
+  ... on AutoSquashEnabledEvent { createdAt actor { ...ActorFields } }
+  ... on AutomaticBaseChangeFailedEvent { createdAt actor { ...ActorFields } oldBase newBase }
+  ... on AutomaticBaseChangeSucceededEvent { createdAt actor { ...ActorFields } oldBase newBase }
+  ... on BaseRefChangedEvent { createdAt actor { ...ActorFields } previousRefName currentRefName }
+  ... on BaseRefDeletedEvent { createdAt actor { ...ActorFields } baseRefName }
+  ... on BaseRefForcePushedEvent { createdAt actor { ...ActorFields } beforeCommit { oid } afterCommit { oid } ref { name } }
+  ... on ClosedEvent { createdAt actor { ...ActorFields } stateReason }
+  ... on CommentDeletedEvent { createdAt actor { ...ActorFields } }
+  ... on ConnectedEvent { createdAt actor { ...ActorFields } source { __typename ... on Issue { number url } ... on PullRequest { number url } } subject { __typename ... on Issue { number url } ... on PullRequest { number url } } }
+  ... on ConvertToDraftEvent { createdAt actor { ...ActorFields } }
+  ... on CrossReferencedEvent { createdAt actor { ...ActorFields } isCrossRepository willCloseTarget source { __typename ... on Issue { number url repository { nameWithOwner } title } ... on PullRequest { number url repository { nameWithOwner } title } } }
+  ... on DemilestonedEvent { createdAt actor { ...ActorFields } milestoneTitle }
+  ... on DeployedEvent { createdAt actor { ...ActorFields } }
+  ... on DeploymentEnvironmentChangedEvent { createdAt actor { ...ActorFields } }
+  ... on DisconnectedEvent { createdAt actor { ...ActorFields } subject { __typename ... on Issue { number url } ... on PullRequest { number url } } source { __typename ... on Issue { number url } ... on PullRequest { number url } } }
+  ... on HeadRefDeletedEvent { createdAt actor { ...ActorFields } headRefName }
+  ... on HeadRefForcePushedEvent { createdAt actor { ...ActorFields } beforeCommit { oid } afterCommit { oid } ref { name } }
+  ... on HeadRefRestoredEvent { createdAt actor { ...ActorFields } }
+  ... on IssueComment { id databaseId createdAt updatedAt author { ...ActorFields } body url reactionGroups { content reactors { totalCount } } }
+  ... on LabeledEvent { createdAt actor { ...ActorFields } label { name color } }
+  ... on LockedEvent { createdAt actor { ...ActorFields } lockReason }
+  ... on MarkedAsDuplicateEvent { createdAt actor { ...ActorFields } canonical { __typename ... on Issue { number url } ... on PullRequest { number url } } }
+  ... on MentionedEvent { createdAt actor { ...ActorFields } }
+  ... on MergedEvent { createdAt actor { ...ActorFields } commit { oid url } mergeRefName }
+  ... on MilestonedEvent { createdAt actor { ...ActorFields } milestoneTitle }
+  ... on MovedColumnsInProjectEvent { createdAt actor { ...ActorFields } }
+  ... on PinnedEvent { createdAt actor { ...ActorFields } }
+  ... on PullRequestCommit { commit { oid url message author { user { login } date } committedDate } }
+  ... on PullRequestCommitCommentThread { commit { oid } }
+  ... on PullRequestReview { id databaseId createdAt submittedAt author { ...ActorFields } body state url reactionGroups { content reactors { totalCount } } }
+  ... on PullRequestReviewThread { id isResolved isOutdated path line diffSide }
+  ... on PullRequestRevisionMarker { createdAt lastSeenCommit { oid } }
+  ... on ReadyForReviewEvent { createdAt actor { ...ActorFields } }
+  ... on ReferencedEvent { createdAt actor { ...ActorFields } commit { oid url } commitRepository { nameWithOwner } }
+  ... on RenamedTitleEvent { createdAt actor { ...ActorFields } previousTitle currentTitle }
+  ... on ReopenedEvent { createdAt actor { ...ActorFields } }
+  ... on ReviewDismissedEvent { createdAt actor { ...ActorFields } dismissalMessage previousReviewState }
+  ... on ReviewRequestRemovedEvent { createdAt actor { ...ActorFields } requestedReviewer { __typename ... on User { login } ... on Team { name } } }
+  ... on ReviewRequestedEvent { createdAt actor { ...ActorFields } requestedReviewer { __typename ... on User { login } ... on Team { name } } }
+  ... on SubscribedEvent { createdAt actor { ...ActorFields } }
+  ... on TransferredEvent { createdAt actor { ...ActorFields } fromRepository { nameWithOwner } }
+  ... on UnassignedEvent { createdAt actor { ...ActorFields } assignee { __typename ... on User { login } ... on Bot { login } } }
+  ... on UnlabeledEvent { createdAt actor { ...ActorFields } label { name color } }
+  ... on UnlockedEvent { createdAt actor { ...ActorFields } }
+  ... on UnmarkedAsDuplicateEvent { createdAt actor { ...ActorFields } }
+  ... on UnpinnedEvent { createdAt actor { ...ActorFields } }
+  ... on UnsubscribedEvent { createdAt actor { ...ActorFields } }
+  ... on UserBlockedEvent { createdAt actor { ...ActorFields } blockDuration }
+}
+"""
+
+
+def _q(parts: list[str], body: str) -> str:
+    return "\n".join(parts + [body])
+
+
+ISSUES_PAGE_QUERY = _q(
+    [F_ACTOR, F_LABEL, F_TIMELINE],
+    """
+query IssuesPage($owner: String!, $name: String!, $first: Int!, $after: String) {
+  repository(owner: $owner, name: $name) {
+    issues(first: $first, after: $after, orderBy: {field: CREATED_AT, direction: ASC}) {
+      pageInfo { hasNextPage endCursor }
+      totalCount
+      nodes {
+        id databaseId number title body state stateReason
+        createdAt updatedAt closedAt
+        url
+        author { ...ActorFields }
+        editor { ...ActorFields }
+        labels(first: 50) { nodes { ...LabelFields } }
+        assignees(first: 20) { nodes { login id } }
+        milestone { title number state dueOn }
+        reactionGroups { content reactors { totalCount } }
+        comments(first: 100) {
+          totalCount
+          pageInfo { hasNextPage endCursor }
+          nodes {
+            id databaseId createdAt updatedAt url body
+            author { ...ActorFields }
+            editor { ...ActorFields }
+            reactionGroups { content reactors { totalCount } }
+          }
+        }
+        timelineItems(first: 100) {
+          totalCount
+          pageInfo { hasNextPage endCursor }
+          nodes { ...TimelineItem }
+        }
+        trackedInIssues(first: 20) { totalCount nodes { number url repository { nameWithOwner } } }
+        trackedIssues(first: 20) { totalCount nodes { number url repository { nameWithOwner } } }
+      }
+    }
+  }
+  rateLimit { cost remaining resetAt }
+}
+""",
+)
+
+PRS_PAGE_QUERY = _q(
+    [F_ACTOR, F_LABEL, F_PR_TIMELINE],
+    """
+query PRsPage($owner: String!, $name: String!, $first: Int!, $after: String) {
+  repository(owner: $owner, name: $name) {
+    pullRequests(first: $first, after: $after, orderBy: {field: CREATED_AT, direction: ASC}) {
+      pageInfo { hasNextPage endCursor }
+      totalCount
+      nodes {
+        id databaseId number title body state isDraft
+        createdAt updatedAt closedAt mergedAt
+        url
+        headRefName headRefOid
+        baseRefName baseRefOid
+        additions deletions changedFiles
+        mergeable merged mergeStateStatus
+        author { ...ActorFields }
+        editor { ...ActorFields }
+        mergedBy { ...ActorFields }
+        labels(first: 50) { nodes { ...LabelFields } }
+        assignees(first: 20) { nodes { login id } }
+        milestone { title number state dueOn }
+        reactionGroups { content reactors { totalCount } }
+        closingIssuesReferences(first: 20) { totalCount nodes { number url repository { nameWithOwner } title } }
+        comments(first: 100) {
+          totalCount
+          pageInfo { hasNextPage endCursor }
+          nodes {
+            id databaseId createdAt updatedAt url body
+            author { ...ActorFields }
+            editor { ...ActorFields }
+            reactionGroups { content reactors { totalCount } }
+          }
+        }
+        reviewThreads(first: 50) {
+          totalCount
+          pageInfo { hasNextPage endCursor }
+          nodes {
+            id isResolved isOutdated path line diffSide
+            comments(first: 50) {
+              totalCount
+              pageInfo { hasNextPage endCursor }
+              nodes {
+                id databaseId createdAt updatedAt url body path diffHunk
+                author { ...ActorFields }
+                editor { ...ActorFields }
+                position originalPosition line originalLine
+                commit { oid }
+                reactionGroups { content reactors { totalCount } }
+              }
+            }
+          }
+        }
+        reviews(first: 50) {
+          totalCount
+          pageInfo { hasNextPage endCursor }
+          nodes {
+            id databaseId state createdAt submittedAt body url
+            author { ...ActorFields }
+            reactionGroups { content reactors { totalCount } }
+          }
+        }
+        commits(first: 100) {
+          totalCount
+          pageInfo { hasNextPage endCursor }
+          nodes {
+            commit {
+              oid
+              message
+              messageHeadline
+              committedDate
+              authoredDate
+              author { name email user { login } date }
+              committer { name email user { login } date }
+              additions deletions changedFilesIfAvailable
+              parents(first: 3) { nodes { oid } }
+            }
+          }
+        }
+        files(first: 100) {
+          totalCount
+          pageInfo { hasNextPage endCursor }
+          nodes {
+            path additions deletions changeType
+          }
+        }
+        timelineItems(first: 100) {
+          totalCount
+          pageInfo { hasNextPage endCursor }
+          nodes { ...PRTimelineItem }
+        }
+      }
+    }
+  }
+  rateLimit { cost remaining resetAt }
+}
+""",
+)
+
+PRS_PAGE_QUERY_LIGHT = _q(
+    [F_ACTOR, F_LABEL],
+    """
+query PRsPageLight($owner: String!, $name: String!, $first: Int!, $after: String) {
+  repository(owner: $owner, name: $name) {
+    pullRequests(first: $first, after: $after, orderBy: {field: CREATED_AT, direction: ASC}) {
+      pageInfo { hasNextPage endCursor }
+      totalCount
+      nodes {
+        id databaseId number title body state isDraft
+        createdAt updatedAt closedAt mergedAt
+        url
+        author { ...ActorFields }
+        labels(first: 50) { nodes { ...LabelFields } }
+        comments(first: 30) {
+          totalCount
+          pageInfo { hasNextPage endCursor }
+          nodes {
+            id databaseId createdAt updatedAt url body
+            author { ...ActorFields }
+          }
+        }
+      }
+    }
+  }
+  rateLimit { cost remaining resetAt }
+}
+""",
+)
+
+ISSUES_PAGE_QUERY_LIGHT = _q(
+    [F_ACTOR, F_LABEL],
+    """
+query IssuesPageLight($owner: String!, $name: String!, $first: Int!, $after: String) {
+  repository(owner: $owner, name: $name) {
+    issues(first: $first, after: $after, orderBy: {field: CREATED_AT, direction: ASC}) {
+      pageInfo { hasNextPage endCursor }
+      totalCount
+      nodes {
+        id databaseId number title body state
+        createdAt updatedAt closedAt
+        url
+        author { ...ActorFields }
+        labels(first: 50) { nodes { ...LabelFields } }
+        comments(first: 30) {
+          totalCount
+          pageInfo { hasNextPage endCursor }
+          nodes {
+            id databaseId createdAt updatedAt url body
+            author { ...ActorFields }
+          }
+        }
+      }
+    }
+  }
+  rateLimit { cost remaining resetAt }
+}
+""",
+)
+
+ISSUE_COMMENTS_QUERY = _q(
+    [F_ACTOR],
+    """
+query IssueComments($owner: String!, $name: String!, $number: Int!, $after: String) {
+  repository(owner: $owner, name: $name) {
+    issueOrPullRequest(number: $number) {
+      __typename
+      ... on Issue {
+        comments(first: 100, after: $after) {
+          pageInfo { hasNextPage endCursor }
+          nodes {
+            id databaseId createdAt updatedAt url body
+            author { ...ActorFields }
+            editor { ...ActorFields }
+            reactionGroups { content reactors { totalCount } }
+          }
+        }
+      }
+      ... on PullRequest {
+        comments(first: 100, after: $after) {
+          pageInfo { hasNextPage endCursor }
+          nodes {
+            id databaseId createdAt updatedAt url body
+            author { ...ActorFields }
+            editor { ...ActorFields }
+            reactionGroups { content reactors { totalCount } }
+          }
+        }
+      }
+    }
+  }
+  rateLimit { cost remaining resetAt }
+}
+""",
+)
+
+ISSUE_TIMELINE_QUERY = _q(
+    [F_ACTOR, F_TIMELINE],
+    """
+query IssueTimeline($owner: String!, $name: String!, $number: Int!, $after: String) {
+  repository(owner: $owner, name: $name) {
+    issue(number: $number) {
+      timelineItems(first: 100, after: $after) {
+        pageInfo { hasNextPage endCursor }
+        nodes { ...TimelineItem }
+      }
+    }
+  }
+  rateLimit { cost remaining resetAt }
+}
+""",
+)
+
+PR_TIMELINE_QUERY = _q(
+    [F_ACTOR, F_PR_TIMELINE],
+    """
+query PRTimeline($owner: String!, $name: String!, $number: Int!, $after: String) {
+  repository(owner: $owner, name: $name) {
+    pullRequest(number: $number) {
+      timelineItems(first: 100, after: $after) {
+        pageInfo { hasNextPage endCursor }
+        nodes { ...PRTimelineItem }
+      }
+    }
+  }
+  rateLimit { cost remaining resetAt }
+}
+""",
+)
+
+PR_COMMITS_QUERY = """
+query PRCommits($owner: String!, $name: String!, $number: Int!, $after: String) {
+  repository(owner: $owner, name: $name) {
+    pullRequest(number: $number) {
+      commits(first: 100, after: $after) {
+        pageInfo { hasNextPage endCursor }
+        nodes {
+          commit {
+            oid message messageHeadline committedDate authoredDate
+            author { name email user { login } date }
+            committer { name email user { login } date }
+            additions deletions changedFilesIfAvailable
+            parents(first: 3) { nodes { oid } }
+          }
+        }
+      }
+    }
+  }
+  rateLimit { cost remaining resetAt }
+}
+"""
+
+PR_FILES_QUERY = """
+query PRFiles($owner: String!, $name: String!, $number: Int!, $after: String) {
+  repository(owner: $owner, name: $name) {
+    pullRequest(number: $number) {
+      files(first: 100, after: $after) {
+        pageInfo { hasNextPage endCursor }
+        nodes { path additions deletions changeType }
+      }
+    }
+  }
+  rateLimit { cost remaining resetAt }
+}
+"""
+
+PR_REVIEW_THREADS_QUERY = _q(
+    [F_ACTOR],
+    """
+query PRReviewThreads($owner: String!, $name: String!, $number: Int!, $after: String) {
+  repository(owner: $owner, name: $name) {
+    pullRequest(number: $number) {
+      reviewThreads(first: 50, after: $after) {
+        pageInfo { hasNextPage endCursor }
+        nodes {
+          id isResolved isOutdated path line diffSide
+          comments(first: 50) {
+            totalCount
+            nodes {
+              id databaseId createdAt updatedAt url body path diffHunk
+              author { ...ActorFields }
+              editor { ...ActorFields }
+              position originalPosition line originalLine
+              commit { oid }
+              reactionGroups { content reactors { totalCount } }
+            }
+          }
+        }
+      }
+    }
+  }
+  rateLimit { cost remaining resetAt }
+}
+""",
+)
+
+DISCUSSIONS_PAGE_QUERY = _q(
+    [F_ACTOR, F_LABEL],
+    """
+query DiscussionsPage($owner: String!, $name: String!, $first: Int!, $after: String) {
+  repository(owner: $owner, name: $name) {
+    discussions(first: $first, after: $after, orderBy: {field: CREATED_AT, direction: ASC}) {
+      pageInfo { hasNextPage endCursor }
+      totalCount
+      nodes {
+        id databaseId number title body
+        createdAt updatedAt url
+        author { ...ActorFields }
+        editor { ...ActorFields }
+        locked
+        answerChosenAt
+        closed closedAt
+        category { id name emoji description isAnswerable }
+        labels(first: 30) { nodes { ...LabelFields } }
+        upvoteCount
+        answer { id databaseId body author { ...ActorFields } createdAt url }
+        reactionGroups { content reactors { totalCount } }
+        comments(first: 50) {
+          totalCount
+          pageInfo { hasNextPage endCursor }
+          nodes {
+            id databaseId body createdAt updatedAt url
+            author { ...ActorFields }
+            editor { ...ActorFields }
+            upvoteCount
+            isAnswer
+            reactionGroups { content reactors { totalCount } }
+            replies(first: 50) {
+              totalCount
+              pageInfo { hasNextPage endCursor }
+              nodes {
+                id databaseId body createdAt updatedAt url
+                author { ...ActorFields }
+                editor { ...ActorFields }
+                reactionGroups { content reactors { totalCount } }
+              }
+            }
+          }
+        }
+      }
+    }
+  }
+  rateLimit { cost remaining resetAt }
+}
+""",
+)
+
+DISCUSSION_COMMENTS_QUERY = _q(
+    [F_ACTOR],
+    """
+query DiscussionComments($owner: String!, $name: String!, $number: Int!, $after: String) {
+  repository(owner: $owner, name: $name) {
+    discussion(number: $number) {
+      comments(first: 50, after: $after) {
+        pageInfo { hasNextPage endCursor }
+        nodes {
+          id databaseId body createdAt updatedAt url
+          author { ...ActorFields }
+          editor { ...ActorFields }
+          upvoteCount
+          isAnswer
+          reactionGroups { content reactors { totalCount } }
+          replies(first: 50) {
+            totalCount
+            nodes {
+              id databaseId body createdAt updatedAt url
+              author { ...ActorFields }
+              editor { ...ActorFields }
+              reactionGroups { content reactors { totalCount } }
+            }
+          }
+        }
+      }
+    }
+  }
+  rateLimit { cost remaining resetAt }
+}
+""",
+)
+
+DISCUSSION_REPLIES_QUERY = _q(
+    [F_ACTOR],
+    """
+query DiscussionReplies($commentId: ID!, $after: String) {
+  node(id: $commentId) {
+    ... on DiscussionComment {
+      replies(first: 50, after: $after) {
+        pageInfo { hasNextPage endCursor }
+        nodes {
+          id databaseId body createdAt updatedAt url
+          author { ...ActorFields }
+          editor { ...ActorFields }
+          reactionGroups { content reactors { totalCount } }
+        }
+      }
+    }
+  }
+  rateLimit { cost remaining resetAt }
+}
+""",
+)
+
+COMMITS_PAGE_QUERY = """
+query CommitsPage($owner: String!, $name: String!, $first: Int!, $after: String, $branch: String!) {
+  repository(owner: $owner, name: $name) {
+    ref(qualifiedName: $branch) {
+      target {
+        ... on Commit {
+          history(first: $first, after: $after) {
+            pageInfo { hasNextPage endCursor }
+            totalCount
+            nodes {
+              oid
+              message
+              messageHeadline
+              committedDate
+              authoredDate
+              url
+              additions deletions changedFilesIfAvailable
+              author { name email date user { login id } }
+              committer { name email date user { login id } }
+              parents(first: 3) { nodes { oid } }
+              associatedPullRequests(first: 5) { nodes { number url state } }
+            }
+          }
+        }
+      }
+    }
+  }
+  rateLimit { cost remaining resetAt }
+}
+"""
+
+RELEASES_QUERY = _q(
+    [F_ACTOR],
+    """
+query Releases($owner: String!, $name: String!, $first: Int!, $after: String) {
+  repository(owner: $owner, name: $name) {
+    releases(first: $first, after: $after, orderBy: {field: CREATED_AT, direction: ASC}) {
+      pageInfo { hasNextPage endCursor }
+      nodes {
+        id databaseId name tagName description
+        createdAt publishedAt updatedAt
+        isDraft isPrerelease isLatest
+        url
+        author { ...ActorFields }
+        tagCommit { oid url }
+        reactionGroups { content reactors { totalCount } }
+        releaseAssets(first: 50) {
+          nodes { name contentType size downloadUrl createdAt updatedAt }
+        }
+      }
+    }
+  }
+  rateLimit { cost remaining resetAt }
+}
+""",
+)
+
+LABELS_QUERY = _q(
+    [F_LABEL],
+    """
+query LabelsList($owner: String!, $name: String!, $first: Int!, $after: String) {
+  repository(owner: $owner, name: $name) {
+    labels(first: $first, after: $after) {
+      pageInfo { hasNextPage endCursor }
+      nodes { ...LabelFields }
+    }
+  }
+  rateLimit { cost remaining resetAt }
+}
+""",
+)
+
+MILESTONES_QUERY = """
+query Milestones($owner: String!, $name: String!, $first: Int!, $after: String) {
+  repository(owner: $owner, name: $name) {
+    milestones(first: $first, after: $after) {
+      pageInfo { hasNextPage endCursor }
+      nodes {
+        id number title description state
+        createdAt updatedAt closedAt dueOn
+        creator { login }
+      }
+    }
+  }
+  rateLimit { cost remaining resetAt }
+}
+"""
+
+REPO_META_QUERY = """
+query RepoMeta($owner: String!, $name: String!) {
+  repository(owner: $owner, name: $name) {
+    id databaseId name nameWithOwner description url
+    createdAt updatedAt pushedAt
+    isArchived isDisabled isFork isPrivate
+    primaryLanguage { name }
+    languages(first: 20, orderBy: {field: SIZE, direction: DESC}) {
+      edges { size node { name } }
+      totalSize
+    }
+    stargazerCount forkCount watchers { totalCount }
+    diskUsage
+    licenseInfo { key name }
+    homepageUrl
+    defaultBranchRef { name }
+  }
+  rateLimit { cost remaining resetAt }
+}
+"""
diff --git a/studio/backend/plugins/data-designer-github-repo-seed/src/data_designer_github_repo_seed/scraper_impl/scraper.py b/studio/backend/plugins/data-designer-github-repo-seed/src/data_designer_github_repo_seed/scraper_impl/scraper.py
new file mode 100644
index 0000000000..127129e18b
--- /dev/null
+++ b/studio/backend/plugins/data-designer-github-repo-seed/src/data_designer_github_repo_seed/scraper_impl/scraper.py
@@ -0,0 +1,756 @@
+# SPDX-License-Identifier: AGPL-3.0-only
+# Copyright 2026-present the Unsloth AI Inc. team. All rights reserved. See /studio/LICENSE.AGPL-3.0
+
+"""Main scraper orchestration. Collects issues, PRs, discussions, commits, releases, etc.
+
+Resumable via state file. Writes JSONL shards under data/{repo}/{resource}.jsonl.
+"""
+
+from __future__ import annotations
+
+import argparse
+import json
+import logging
+import os
+import subprocess
+import sys
+import time
+from pathlib import Path
+from typing import Any, Dict, Iterable, List, Optional, Tuple
+
+# Allow running as a module or script
+THIS_DIR = Path(__file__).resolve().parent
+if str(THIS_DIR) not in sys.path:
+    sys.path.insert(0, str(THIS_DIR))
+
+from gh_client import GitHubClient
+from state_store import JsonlWriter, StateStore
+import queries as Q
+
+log = logging.getLogger("scraper")
+
+
+def ts() -> str:
+    return time.strftime("%Y-%m-%d %H:%M:%S")
+
+
+class RepoScraper:
+    def __init__(
+        self,
+        owner: str,
+        name: str,
+        base_dir: Path,
+        client: GitHubClient,
+        trial_limits: Optional[Dict[str, int]] = None,
+        light: bool = False,
+    ):
+        self.owner = owner
+        self.name = name
+        self.base_dir = base_dir
+        self.client = client
+        self.trial_limits = trial_limits or {}
+        # When light=True, use trimmed GraphQL queries (no reviewThreads,
+        # reviews, commits, timelineItems, files) so PR pages can be much
+        # larger without blowing GitHub's node-count ceiling.
+        self.light = light
+        self.repo_dir = base_dir / f"{owner}__{name}"
+        self.repo_dir.mkdir(parents = True, exist_ok = True)
+        self.state = StateStore(base_dir / "state" / f"{owner}__{name}.json")
+
+        # Writers
+        self.writers: Dict[str, JsonlWriter] = {}
+        for key in (
+            "issues",
+            "pull_requests",
+            "discussions",
+            "commits",
+            "releases",
+            "labels",
+            "milestones",
+            "pr_extra_comments",
+            "pr_extra_timeline",
+            "pr_extra_reviews",
+            "issue_extra_comments",
+            "issue_extra_timeline",
+            "discussion_extra_comments",
+            "discussion_extra_replies",
+            "repo_meta",
+        ):
+            self.writers[key] = JsonlWriter(self.repo_dir / f"{key}.jsonl")
+
+    # ----- helpers -----
+    def _trial_stop(self, key: str, counter: int) -> bool:
+        lim = self.trial_limits.get(key)
+        if lim is None:
+            return False
+        return counter >= lim
+
+    def _log_rate(self, where: str, data: Dict[str, Any]) -> None:
+        rl = (
+            data.get("data", {}).get("rateLimit")
+            if isinstance(data.get("data"), dict)
+            else None
+        )
+        if rl:
+            log.debug(
+                "[%s] rate cost=%s remaining=%s resetAt=%s",
+                where,
+                rl.get("cost"),
+                rl.get("remaining"),
+                rl.get("resetAt"),
+            )
+
+    # ----- repo meta -----
+    def scrape_repo_meta(self) -> Dict[str, Any]:
+        data = self.client.graphql(
+            Q.REPO_META_QUERY, {"owner": self.owner, "name": self.name}
+        )
+        self._log_rate("repo_meta", data)
+        repo = data.get("data", {}).get("repository") or {}
+        repo["_fetchedAt"] = ts()
+        self.writers["repo_meta"].write(repo)
+        return repo
+
+    # ----- issues -----
+    def scrape_issues(self) -> int:
+        key = "issues"
+        cursor = self.state.get(f"{key}_cursor")
+        done = self.state.get(f"{key}_done", False)
+        if done:
+            log.info("%s/%s issues already complete", self.owner, self.name)
+            return 0
+        total_new = 0
+        page = 0
+        # Light query skips heavy nested fields; safe at 50 per page.
+        # Clamp by trial_limit so e.g. limit=1 asks GitHub for first:1
+        # instead of fetching a full 50-item page and discarding 49.
+        page_cap = 50 if self.light else 15
+        trial_cap = self.trial_limits.get(key)
+        per_page = min(page_cap, trial_cap) if trial_cap and trial_cap > 0 else page_cap
+        while True:
+            page += 1
+            vars_ = {
+                "owner": self.owner,
+                "name": self.name,
+                "first": per_page,
+                "after": cursor,
+            }
+            query = Q.ISSUES_PAGE_QUERY_LIGHT if self.light else Q.ISSUES_PAGE_QUERY
+            data = self.client.graphql(query, vars_)
+            self._log_rate("issues", data)
+            repo = (data.get("data") or {}).get("repository") or {}
+            issues = repo.get("issues") or {}
+            nodes = issues.get("nodes") or []
+            for it in nodes:
+                it["_owner"] = self.owner
+                it["_repo"] = self.name
+                it["_fetchedAt"] = ts()
+                if not self.light:
+                    if it.get("comments", {}).get("pageInfo", {}).get("hasNextPage"):
+                        self._paginate_issue_comments(
+                            it["number"], it["comments"]["pageInfo"]["endCursor"]
+                        )
+                    if (
+                        it.get("timelineItems", {})
+                        .get("pageInfo", {})
+                        .get("hasNextPage")
+                    ):
+                        self._paginate_issue_timeline(
+                            it["number"],
+                            it["timelineItems"]["pageInfo"]["endCursor"],
+                        )
+                if self.writers[key].write(it):
+                    total_new += 1
+            info = issues.get("pageInfo") or {}
+            cursor = info.get("endCursor")
+            self.state.set(f"{key}_cursor", cursor)
+            log.info(
+                "[%s/%s] issues page %d (+%d) cursor=%s remaining=%s",
+                self.owner,
+                self.name,
+                page,
+                len(nodes),
+                str(cursor)[:20],
+                self.client.graphql_remaining,
+            )
+            if self._trial_stop(key, total_new):
+                log.info("Trial limit reached for issues (%d)", total_new)
+                return total_new
+            if not info.get("hasNextPage"):
+                self.state.set(f"{key}_done", True)
+                break
+        return total_new
+
+    def _paginate_issue_comments(self, number: int, after: str) -> None:
+        cur = after
+        while cur:
+            vars_ = {
+                "owner": self.owner,
+                "name": self.name,
+                "number": number,
+                "after": cur,
+            }
+            data = self.client.graphql(Q.ISSUE_COMMENTS_QUERY, vars_)
+            item = ((data.get("data") or {}).get("repository") or {}).get(
+                "issueOrPullRequest"
+            ) or {}
+            comments = item.get("comments") or {}
+            for c in comments.get("nodes") or []:
+                c["_owner"] = self.owner
+                c["_repo"] = self.name
+                c["_issueNumber"] = number
+                self.writers["issue_extra_comments"].write(c)
+            info = comments.get("pageInfo") or {}
+            cur = info.get("endCursor") if info.get("hasNextPage") else None
+
+    def _paginate_issue_timeline(self, number: int, after: str) -> None:
+        cur = after
+        while cur:
+            vars_ = {
+                "owner": self.owner,
+                "name": self.name,
+                "number": number,
+                "after": cur,
+            }
+            data = self.client.graphql(Q.ISSUE_TIMELINE_QUERY, vars_)
+            item = ((data.get("data") or {}).get("repository") or {}).get("issue") or {}
+            tl = item.get("timelineItems") or {}
+            for ev in tl.get("nodes") or []:
+                ev["_owner"] = self.owner
+                ev["_repo"] = self.name
+                ev["_issueNumber"] = number
+                self.writers["issue_extra_timeline"].write(ev)
+            info = tl.get("pageInfo") or {}
+            cur = info.get("endCursor") if info.get("hasNextPage") else None
+
+    # ----- PRs -----
+    def scrape_prs(self) -> int:
+        key = "pull_requests"
+        cursor = self.state.get(f"{key}_cursor")
+        done = self.state.get(f"{key}_done", False)
+        if done:
+            log.info("%s/%s PRs already complete", self.owner, self.name)
+            return 0
+        total_new = 0
+        page = 0
+        # Heavy nested PR query is capped at 3 per page (GitHub node-count
+        # ceiling); light query skips reviewThreads/reviews/commits/etc and
+        # can safely go to 25 per page. Clamp by trial_limit for small
+        # previews so limit=1 does not fetch a whole 25-item page.
+        page_cap = 25 if self.light else 3
+        trial_cap = self.trial_limits.get(key)
+        per_page = min(page_cap, trial_cap) if trial_cap and trial_cap > 0 else page_cap
+        while True:
+            page += 1
+            vars_ = {
+                "owner": self.owner,
+                "name": self.name,
+                "first": per_page,
+                "after": cursor,
+            }
+            query = Q.PRS_PAGE_QUERY_LIGHT if self.light else Q.PRS_PAGE_QUERY
+            data = self.client.graphql(query, vars_)
+            self._log_rate("prs", data)
+            repo = (data.get("data") or {}).get("repository") or {}
+            prs = repo.get("pullRequests") or {}
+            nodes = prs.get("nodes") or []
+            for pr in nodes:
+                pr["_owner"] = self.owner
+                pr["_repo"] = self.name
+                pr["_fetchedAt"] = ts()
+                num = pr["number"]
+                if not self.light:
+                    if pr.get("comments", {}).get("pageInfo", {}).get("hasNextPage"):
+                        self._paginate_pr_comments(
+                            num, pr["comments"]["pageInfo"]["endCursor"]
+                        )
+                    if (
+                        pr.get("timelineItems", {})
+                        .get("pageInfo", {})
+                        .get("hasNextPage")
+                    ):
+                        self._paginate_pr_timeline(
+                            num, pr["timelineItems"]["pageInfo"]["endCursor"]
+                        )
+                    if pr.get("commits", {}).get("pageInfo", {}).get("hasNextPage"):
+                        self._paginate_pr_commits(
+                            num, pr["commits"]["pageInfo"]["endCursor"]
+                        )
+                    if pr.get("files", {}).get("pageInfo", {}).get("hasNextPage"):
+                        self._paginate_pr_files(
+                            num, pr["files"]["pageInfo"]["endCursor"]
+                        )
+                    if (
+                        pr.get("reviewThreads", {})
+                        .get("pageInfo", {})
+                        .get("hasNextPage")
+                    ):
+                        self._paginate_pr_review_threads(
+                            num, pr["reviewThreads"]["pageInfo"]["endCursor"]
+                        )
+                if self.writers[key].write(pr):
+                    total_new += 1
+            info = prs.get("pageInfo") or {}
+            cursor = info.get("endCursor")
+            self.state.set(f"{key}_cursor", cursor)
+            log.info(
+                "[%s/%s] PRs page %d (+%d) cursor=%s remaining=%s",
+                self.owner,
+                self.name,
+                page,
+                len(nodes),
+                str(cursor)[:20],
+                self.client.graphql_remaining,
+            )
+            if self._trial_stop(key, total_new):
+                log.info("Trial limit reached for PRs (%d)", total_new)
+                return total_new
+            if not info.get("hasNextPage"):
+                self.state.set(f"{key}_done", True)
+                break
+        return total_new
+
+    def _paginate_pr_comments(self, number: int, after: str) -> None:
+        cur = after
+        while cur:
+            vars_ = {
+                "owner": self.owner,
+                "name": self.name,
+                "number": number,
+                "after": cur,
+            }
+            data = self.client.graphql(Q.ISSUE_COMMENTS_QUERY, vars_)
+            item = ((data.get("data") or {}).get("repository") or {}).get(
+                "issueOrPullRequest"
+            ) or {}
+            comments = item.get("comments") or {}
+            for c in comments.get("nodes") or []:
+                c["_owner"] = self.owner
+                c["_repo"] = self.name
+                c["_prNumber"] = number
+                self.writers["pr_extra_comments"].write(c)
+            info = comments.get("pageInfo") or {}
+            cur = info.get("endCursor") if info.get("hasNextPage") else None
+
+    def _paginate_pr_timeline(self, number: int, after: str) -> None:
+        cur = after
+        while cur:
+            vars_ = {
+                "owner": self.owner,
+                "name": self.name,
+                "number": number,
+                "after": cur,
+            }
+            data = self.client.graphql(Q.PR_TIMELINE_QUERY, vars_)
+            item = ((data.get("data") or {}).get("repository") or {}).get(
+                "pullRequest"
+            ) or {}
+            tl = item.get("timelineItems") or {}
+            for ev in tl.get("nodes") or []:
+                ev["_owner"] = self.owner
+                ev["_repo"] = self.name
+                ev["_prNumber"] = number
+                self.writers["pr_extra_timeline"].write(ev)
+            info = tl.get("pageInfo") or {}
+            cur = info.get("endCursor") if info.get("hasNextPage") else None
+
+    def _paginate_pr_commits(self, number: int, after: str) -> None:
+        cur = after
+        out_key = "pr_extra_commits"
+        if out_key not in self.writers:
+            self.writers[out_key] = JsonlWriter(self.repo_dir / f"{out_key}.jsonl")
+        while cur:
+            vars_ = {
+                "owner": self.owner,
+                "name": self.name,
+                "number": number,
+                "after": cur,
+            }
+            data = self.client.graphql(Q.PR_COMMITS_QUERY, vars_)
+            item = ((data.get("data") or {}).get("repository") or {}).get(
+                "pullRequest"
+            ) or {}
+            cc = item.get("commits") or {}
+            for c in cc.get("nodes") or []:
+                c["_owner"] = self.owner
+                c["_repo"] = self.name
+                c["_prNumber"] = number
+                self.writers[out_key].write(c)
+            info = cc.get("pageInfo") or {}
+            cur = info.get("endCursor") if info.get("hasNextPage") else None
+
+    def _paginate_pr_files(self, number: int, after: str) -> None:
+        cur = after
+        out_key = "pr_extra_files"
+        if out_key not in self.writers:
+            self.writers[out_key] = JsonlWriter(self.repo_dir / f"{out_key}.jsonl")
+        while cur:
+            vars_ = {
+                "owner": self.owner,
+                "name": self.name,
+                "number": number,
+                "after": cur,
+            }
+            data = self.client.graphql(Q.PR_FILES_QUERY, vars_)
+            item = ((data.get("data") or {}).get("repository") or {}).get(
+                "pullRequest"
+            ) or {}
+            ff = item.get("files") or {}
+            for f in ff.get("nodes") or []:
+                f["_owner"] = self.owner
+                f["_repo"] = self.name
+                f["_prNumber"] = number
+                # files don't have id, synthesize one
+                f["_syntheticId"] = f"{self.owner}/{self.name}#{number}:{f.get('path')}"
+                self.writers[out_key].write(f)
+            info = ff.get("pageInfo") or {}
+            cur = info.get("endCursor") if info.get("hasNextPage") else None
+
+    def _paginate_pr_review_threads(self, number: int, after: str) -> None:
+        cur = after
+        out_key = "pr_extra_review_threads"
+        if out_key not in self.writers:
+            self.writers[out_key] = JsonlWriter(self.repo_dir / f"{out_key}.jsonl")
+        while cur:
+            vars_ = {
+                "owner": self.owner,
+                "name": self.name,
+                "number": number,
+                "after": cur,
+            }
+            data = self.client.graphql(Q.PR_REVIEW_THREADS_QUERY, vars_)
+            item = ((data.get("data") or {}).get("repository") or {}).get(
+                "pullRequest"
+            ) or {}
+            rt = item.get("reviewThreads") or {}
+            for th in rt.get("nodes") or []:
+                th["_owner"] = self.owner
+                th["_repo"] = self.name
+                th["_prNumber"] = number
+                self.writers[out_key].write(th)
+            info = rt.get("pageInfo") or {}
+            cur = info.get("endCursor") if info.get("hasNextPage") else None
+
+    # ----- Discussions -----
+    def scrape_discussions(self) -> int:
+        key = "discussions"
+        cursor = self.state.get(f"{key}_cursor")
+        done = self.state.get(f"{key}_done", False)
+        if done:
+            log.info("%s/%s discussions already complete", self.owner, self.name)
+            return 0
+        total_new = 0
+        page = 0
+        per_page = 15
+        while True:
+            page += 1
+            vars_ = {
+                "owner": self.owner,
+                "name": self.name,
+                "first": per_page,
+                "after": cursor,
+            }
+            data = self.client.graphql(Q.DISCUSSIONS_PAGE_QUERY, vars_)
+            self._log_rate("discussions", data)
+            repo = (data.get("data") or {}).get("repository") or {}
+            dd = repo.get("discussions") or {}
+            nodes = dd.get("nodes") or []
+            for d in nodes:
+                d["_owner"] = self.owner
+                d["_repo"] = self.name
+                d["_fetchedAt"] = ts()
+                num = d["number"]
+                if d.get("comments", {}).get("pageInfo", {}).get("hasNextPage"):
+                    self._paginate_discussion_comments(
+                        num, d["comments"]["pageInfo"]["endCursor"]
+                    )
+                # paginate replies per comment if needed
+                for c in d.get("comments", {}).get("nodes", []) or []:
+                    if c.get("replies", {}).get("pageInfo", {}).get("hasNextPage"):
+                        self._paginate_discussion_replies(
+                            c["id"], c["replies"]["pageInfo"]["endCursor"], num
+                        )
+                if self.writers[key].write(d):
+                    total_new += 1
+            info = dd.get("pageInfo") or {}
+            cursor = info.get("endCursor")
+            self.state.set(f"{key}_cursor", cursor)
+            log.info(
+                "[%s/%s] discussions page %d (+%d) cursor=%s remaining=%s",
+                self.owner,
+                self.name,
+                page,
+                len(nodes),
+                str(cursor)[:20],
+                self.client.graphql_remaining,
+            )
+            if self._trial_stop(key, total_new):
+                return total_new
+            if not info.get("hasNextPage"):
+                self.state.set(f"{key}_done", True)
+                break
+        return total_new
+
+    def _paginate_discussion_comments(self, number: int, after: str) -> None:
+        cur = after
+        while cur:
+            vars_ = {
+                "owner": self.owner,
+                "name": self.name,
+                "number": number,
+                "after": cur,
+            }
+            data = self.client.graphql(Q.DISCUSSION_COMMENTS_QUERY, vars_)
+            disc = ((data.get("data") or {}).get("repository") or {}).get(
+                "discussion"
+            ) or {}
+            cc = disc.get("comments") or {}
+            for c in cc.get("nodes") or []:
+                c["_owner"] = self.owner
+                c["_repo"] = self.name
+                c["_discussionNumber"] = number
+                self.writers["discussion_extra_comments"].write(c)
+            info = cc.get("pageInfo") or {}
+            cur = info.get("endCursor") if info.get("hasNextPage") else None
+
+    def _paginate_discussion_replies(
+        self, comment_id: str, after: str, disc_number: int
+    ) -> None:
+        cur = after
+        while cur:
+            vars_ = {
+                "owner": self.owner,
+                "name": self.name,
+                "commentId": comment_id,
+                "after": cur,
+            }
+            data = self.client.graphql(Q.DISCUSSION_REPLIES_QUERY, vars_)
+            node = (data.get("data") or {}).get("node") or {}
+            replies = node.get("replies") or {}
+            for r in replies.get("nodes") or []:
+                r["_owner"] = self.owner
+                r["_repo"] = self.name
+                r["_discussionNumber"] = disc_number
+                r["_commentId"] = comment_id
+                self.writers["discussion_extra_replies"].write(r)
+            info = replies.get("pageInfo") or {}
+            cur = info.get("endCursor") if info.get("hasNextPage") else None
+
+    # ----- Commits -----
+    def scrape_commits(self, branch: str = "refs/heads/main") -> int:
+        key = "commits"
+        cursor = self.state.get(f"{key}_cursor")
+        done = self.state.get(f"{key}_done", False)
+        if done:
+            return 0
+        total_new = 0
+        page = 0
+        page_cap = 100
+        trial_cap = self.trial_limits.get(key)
+        per_page = min(page_cap, trial_cap) if trial_cap and trial_cap > 0 else page_cap
+        while True:
+            page += 1
+            vars_ = {
+                "owner": self.owner,
+                "name": self.name,
+                "first": per_page,
+                "after": cursor,
+                "branch": branch,
+            }
+            data = self.client.graphql(Q.COMMITS_PAGE_QUERY, vars_)
+            self._log_rate("commits", data)
+            ref = ((data.get("data") or {}).get("repository") or {}).get("ref") or {}
+            tgt = ref.get("target") or {}
+            hist = tgt.get("history") or {}
+            nodes = hist.get("nodes") or []
+            for c in nodes:
+                c["_owner"] = self.owner
+                c["_repo"] = self.name
+                c["_fetchedAt"] = ts()
+                if self.writers[key].write(c):
+                    total_new += 1
+            info = hist.get("pageInfo") or {}
+            cursor = info.get("endCursor")
+            self.state.set(f"{key}_cursor", cursor)
+            log.info(
+                "[%s/%s] commits page %d (+%d) remaining=%s",
+                self.owner,
+                self.name,
+                page,
+                len(nodes),
+                self.client.graphql_remaining,
+            )
+            if self._trial_stop(key, total_new):
+                return total_new
+            if not info.get("hasNextPage"):
+                self.state.set(f"{key}_done", True)
+                break
+        return total_new
+
+    # ----- Releases/Labels/Milestones -----
+    def scrape_releases(self) -> int:
+        return self._scrape_simple("releases", Q.RELEASES_QUERY, "releases")
+
+    def scrape_labels(self) -> int:
+        return self._scrape_simple("labels", Q.LABELS_QUERY, "labels")
+
+    def scrape_milestones(self) -> int:
+        return self._scrape_simple("milestones", Q.MILESTONES_QUERY, "milestones")
+
+    def _scrape_simple(self, key: str, query: str, field: str) -> int:
+        cursor = self.state.get(f"{key}_cursor")
+        done = self.state.get(f"{key}_done", False)
+        if done:
+            return 0
+        total_new = 0
+        while True:
+            vars_ = {
+                "owner": self.owner,
+                "name": self.name,
+                "first": 50,
+                "after": cursor,
+            }
+            data = self.client.graphql(query, vars_)
+            repo = (data.get("data") or {}).get("repository") or {}
+            col = repo.get(field) or {}
+            for it in col.get("nodes") or []:
+                it["_owner"] = self.owner
+                it["_repo"] = self.name
+                it["_fetchedAt"] = ts()
+                if self.writers[key].write(it):
+                    total_new += 1
+            info = col.get("pageInfo") or {}
+            cursor = info.get("endCursor")
+            self.state.set(f"{key}_cursor", cursor)
+            if self._trial_stop(key, total_new):
+                return total_new
+            if not info.get("hasNextPage"):
+                self.state.set(f"{key}_done", True)
+                break
+        log.info("[%s/%s] %s done +%d", self.owner, self.name, key, total_new)
+        return total_new
+
+    def close(self) -> None:
+        for w in self.writers.values():
+            try:
+                w.close()
+            except Exception:
+                pass
+
+
+def setup_logging(log_file: Path) -> None:
+    log_file.parent.mkdir(parents = True, exist_ok = True)
+    fmt = "%(asctime)s %(levelname)s [%(name)s] %(message)s"
+    handlers = [
+        logging.StreamHandler(sys.stdout),
+        logging.FileHandler(log_file, mode = "a", encoding = "utf-8"),
+    ]
+    logging.basicConfig(level = logging.INFO, format = fmt, handlers = handlers, force = True)
+
+
+def main():
+    ap = argparse.ArgumentParser()
+    ap.add_argument(
+        "--base-dir", default = "/mnt/disks/unslothai/ubuntu/workspace_34/github_scraper"
+    )
+    ap.add_argument(
+        "--repos", nargs = "+", default = ["unslothai/unsloth", "unslothai/unsloth-zoo"]
+    )
+    ap.add_argument("--trial", action = "store_true", help = "Small trial run")
+    ap.add_argument(
+        "--only",
+        nargs = "+",
+        default = None,
+        help = "Only run these resource keys: issues,pulls,discussions,commits,releases,labels,milestones,meta",
+    )
+    ap.add_argument(
+        "--hf-upload-interval",
+        type = int,
+        default = 900,
+        help = "Seconds between HF uploads (0 to disable)",
+    )
+    args = ap.parse_args()
+
+    base = Path(args.base_dir)
+    data_dir = base / "data"
+    data_dir.mkdir(parents = True, exist_ok = True)
+    setup_logging(base / "logs" / f"scraper_{time.strftime('%Y%m%d_%H%M%S')}.log")
+    log.info("Scraper starting: repos=%s trial=%s", args.repos, args.trial)
+
+    client = GitHubClient(min_remaining_graphql = 80, min_remaining_rest = 80)
+    rl = client.rate_snapshot()
+    log.info(
+        "Rate limit snapshot: %s",
+        json.dumps(rl.get("resources", {}), default = str)[:400],
+    )
+
+    # Start HF uploader in background if requested
+    uploader = None
+    if args.hf_upload_interval > 0:
+        from hf_uploader import HFUploader
+
+        uploader = HFUploader(data_dir, interval_s = args.hf_upload_interval)
+        uploader.start()
+
+    trial_limits = None
+    if args.trial:
+        trial_limits = {
+            "issues": 5,
+            "pull_requests": 5,
+            "discussions": 3,
+            "commits": 20,
+            "releases": 3,
+            "labels": 20,
+            "milestones": 20,
+        }
+
+    only = set(args.only or [])
+
+    try:
+        for repo_spec in args.repos:
+            owner, name = repo_spec.split("/")
+            scraper = RepoScraper(owner, name, data_dir, client, trial_limits)
+            try:
+                repo_meta: Dict[str, Any] = {}
+                if not only or "meta" in only or "commits" in only:
+                    repo_meta = scraper.scrape_repo_meta()
+                if not only or "labels" in only:
+                    scraper.scrape_labels()
+                if not only or "milestones" in only:
+                    scraper.scrape_milestones()
+                if not only or "releases" in only:
+                    scraper.scrape_releases()
+                if not only or "discussions" in only:
+                    scraper.scrape_discussions()
+                if not only or "issues" in only:
+                    scraper.scrape_issues()
+                if not only or "pulls" in only:
+                    scraper.scrape_prs()
+                if not only or "commits" in only:
+                    default_ref = repo_meta.get("defaultBranchRef") or {}
+                    default_branch = (
+                        default_ref.get("name")
+                        if isinstance(default_ref, dict)
+                        else None
+                    )
+                    branch = (
+                        f"refs/heads/{default_branch}"
+                        if default_branch
+                        else "refs/heads/main"
+                    )
+                    scraper.scrape_commits(branch = branch)
+            finally:
+                scraper.close()
+    finally:
+        if uploader:
+            log.info("Stopping uploader and final sync...")
+            uploader.stop(final_upload = True)
+    log.info(
+        "Scraper complete. GraphQL calls=%d REST calls=%d",
+        client.calls_graphql,
+        client.calls_rest,
+    )
+
+
+if __name__ == "__main__":
+    main()
diff --git a/studio/backend/plugins/data-designer-github-repo-seed/src/data_designer_github_repo_seed/scraper_impl/state_store.py b/studio/backend/plugins/data-designer-github-repo-seed/src/data_designer_github_repo_seed/scraper_impl/state_store.py
new file mode 100644
index 0000000000..efa663db2f
--- /dev/null
+++ b/studio/backend/plugins/data-designer-github-repo-seed/src/data_designer_github_repo_seed/scraper_impl/state_store.py
@@ -0,0 +1,105 @@
+# SPDX-License-Identifier: AGPL-3.0-only
+# Copyright 2026-present the Unsloth AI Inc. team. All rights reserved. See /studio/LICENSE.AGPL-3.0
+
+"""Checkpoint state management for resumable scraping."""
+
+from __future__ import annotations
+
+import json
+import os
+import threading
+from pathlib import Path
+from typing import Any, Dict
+
+
+class StateStore:
+    def __init__(self, path: str | Path):
+        self.path = Path(path)
+        self.path.parent.mkdir(parents = True, exist_ok = True)
+        self._lock = threading.Lock()
+        self._data: Dict[str, Any] = {}
+        if self.path.exists():
+            try:
+                with self.path.open() as f:
+                    self._data = json.load(f)
+            except Exception:
+                self._data = {}
+
+    def get(self, key: str, default: Any = None) -> Any:
+        with self._lock:
+            return self._data.get(key, default)
+
+    def set(self, key: str, value: Any) -> None:
+        with self._lock:
+            self._data[key] = value
+            self._flush()
+
+    def update(self, key: str, **kwargs) -> None:
+        with self._lock:
+            sub = dict(self._data.get(key, {}))
+            sub.update(kwargs)
+            self._data[key] = sub
+            self._flush()
+
+    def all(self) -> Dict[str, Any]:
+        with self._lock:
+            return dict(self._data)
+
+    def _flush(self) -> None:
+        tmp = self.path.with_suffix(self.path.suffix + ".tmp")
+        with tmp.open("w") as f:
+            json.dump(self._data, f, indent = 2, default = str)
+        os.replace(tmp, self.path)
+
+
+class JsonlWriter:
+    """Append-only JSONL writer, thread-safe, with line buffering."""
+
+    def __init__(self, path: str | Path):
+        self.path = Path(path)
+        self.path.parent.mkdir(parents = True, exist_ok = True)
+        self._lock = threading.Lock()
+        self._fh = self.path.open("a", buffering = 1)
+        self._count_seen_keys: set[str] = set()
+        # Preload seen keys if file exists (for dedup across resumes)
+        if self.path.exists() and self.path.stat().st_size > 0:
+            try:
+                with self.path.open() as f:
+                    for line in f:
+                        try:
+                            obj = json.loads(line)
+                            k = self._key(obj)
+                            if k is not None:
+                                self._count_seen_keys.add(k)
+                        except Exception:
+                            pass
+            except Exception:
+                pass
+
+    def _key(self, obj: dict) -> str | None:
+        for k in ("id", "node_id", "number", "sha", "url"):
+            if k in obj:
+                return f"{k}:{obj[k]}"
+        return None
+
+    def has(self, key: str) -> bool:
+        return key in self._count_seen_keys
+
+    def write(self, obj: dict) -> bool:
+        """Return True if newly written, False if already present."""
+        k = self._key(obj)
+        with self._lock:
+            if k is not None and k in self._count_seen_keys:
+                return False
+            if k is not None:
+                self._count_seen_keys.add(k)
+            self._fh.write(json.dumps(obj, default = str, ensure_ascii = False))
+            self._fh.write("\n")
+            self._fh.flush()
+        return True
+
+    def close(self) -> None:
+        try:
+            self._fh.close()
+        except Exception:
+            pass
diff --git a/studio/backend/requirements/single-env/data-designer-deps.txt b/studio/backend/requirements/single-env/data-designer-deps.txt
index fc63230922..f63c076621 100644
--- a/studio/backend/requirements/single-env/data-designer-deps.txt
+++ b/studio/backend/requirements/single-env/data-designer-deps.txt
@@ -19,7 +19,8 @@ ruff<1,>=0.14.10
 scipy<2,>=1.11.0
 sqlfluff<4,>=3.2.0
 tiktoken<1,>=0.8.0
-# Unstructured-seed plugin deps (plugin installed with --no-deps)
+# Local seed plugin deps (plugins installed with --no-deps)
+requests>=2.31
 pymupdf>=1.24.0
 pymupdf4llm>=0.0.17
 mammoth>=1.8.0
diff --git a/studio/backend/routes/data_recipe/jobs.py b/studio/backend/routes/data_recipe/jobs.py
index 606ef1832c..da6416e324 100644
--- a/studio/backend/routes/data_recipe/jobs.py
+++ b/studio/backend/routes/data_recipe/jobs.py
@@ -5,8 +5,9 @@
 
 from __future__ import annotations
 
-from datetime import timedelta
-from typing import Any
+import copy
+from datetime import datetime, timedelta, timezone
+from typing import Any, Optional
 from urllib.parse import urlparse
 
 from fastapi import APIRouter, HTTPException, Query, Request
@@ -94,14 +95,111 @@ def _used_llm_model_aliases(recipe: dict[str, Any]) -> set[str]:
     return aliases
 
 
-def _inject_local_providers(recipe: dict[str, Any], request: Request) -> None:
+def _inject_local_structured_response_format(
+    recipe: dict[str, Any], local_provider_names: set[str]
+) -> None:
+    """For each llm-structured column that targets a local-provider model_config,
+    clone the model_config and inject an OpenAI ``response_format`` with the
+    column's ``output_format`` JSON schema. The column is rewritten to point at
+    the clone so llm-text / llm-judge columns that share the same alias keep
+    free-form sampling.
+
+    Without this, data_designer only injects a prompt-level "return JSON in a
+    ```json fence" instruction. Small GGUF models frequently break format,
+    wasting the full ``max_tokens`` budget per row and then failing to parse.
+    Forwarding ``response_format`` lets llama-server apply grammar-constrained
+    sampling from the JSON schema, which guarantees a parseable response and
+    terminates early.
+    """
+    columns = recipe.get("columns")
+    model_configs = recipe.get("model_configs")
+    if not isinstance(columns, list) or not isinstance(model_configs, list):
+        return
+
+    # alias -> model_config (only configs referencing a local provider qualify).
+    alias_to_local_mc: dict[str, dict[str, Any]] = {}
+    for mc in model_configs:
+        if not isinstance(mc, dict):
+            continue
+        if mc.get("provider") in local_provider_names and isinstance(
+            mc.get("alias"), str
+        ):
+            alias_to_local_mc[mc["alias"]] = mc
+
+    if not alias_to_local_mc:
+        return
+
+    # Clone per (alias, column) so each llm-structured column gets its own
+    # schema without leaking response_format onto other columns that share the
+    # same base alias.
+    seen_clone_aliases: set[str] = {
+        mc.get("alias") for mc in model_configs if isinstance(mc.get("alias"), str)
+    }
+    new_configs: list[dict[str, Any]] = []
+    for column in columns:
+        if not isinstance(column, dict):
+            continue
+        if column.get("column_type") != "llm-structured":
+            continue
+        alias = column.get("model_alias")
+        if not isinstance(alias, str) or alias not in alias_to_local_mc:
+            continue
+        output_format = column.get("output_format")
+        if not isinstance(output_format, dict) or not output_format:
+            continue
+        base_mc = alias_to_local_mc[alias]
+        column_name = column.get("name") or "structured"
+        clone_alias_base = f"{alias}__{column_name}_structured"
+        clone_alias = clone_alias_base
+        counter = 1
+        while clone_alias in seen_clone_aliases:
+            counter += 1
+            clone_alias = f"{clone_alias_base}_{counter}"
+        seen_clone_aliases.add(clone_alias)
+
+        clone = copy.deepcopy(base_mc)
+        clone["alias"] = clone_alias
+        params = clone.get("inference_parameters")
+        if not isinstance(params, dict):
+            params = {}
+            clone["inference_parameters"] = params
+        # data_designer's BaseInferenceParams is a pydantic model with
+        # extra="forbid", so response_format cannot sit at the top level of
+        # inference_parameters. It does expose an `extra_body: dict` pass-
+        # through that the OpenAI client spreads into the request body at the
+        # top level, which is where llama-server reads response_format from.
+        # llama.cpp server shape (tools/server/README.md): the schema sits
+        # directly under response_format, not nested in a json_schema object
+        # the way OpenAI's Chat Completions API expects. llama-server converts
+        # the schema to a GBNF grammar and applies it during sampling.
+        extra_body = params.get("extra_body")
+        if not isinstance(extra_body, dict):
+            extra_body = {}
+        extra_body["response_format"] = {
+            "type": "json_schema",
+            "schema": output_format,
+        }
+        params["extra_body"] = extra_body
+        new_configs.append(clone)
+        column["model_alias"] = clone_alias
+
+    if new_configs:
+        model_configs.extend(new_configs)
+
+
+def _inject_local_providers(recipe: dict[str, Any], request: Request) -> Optional[int]:
     """
     Mutate recipe dict in-place: for any provider with is_local=True,
-    generate a JWT and fill in the endpoint pointing at this server.
+    fill in the endpoint pointing at this server and inject a short-lived
+    internal sk-unsloth-* API key for workflow auth.
+
+    Returns the row id of the minted internal key (so the caller can
+    revoke it on job completion) or ``None`` when no local provider is
+    actually reachable from an LLM column.
     """
     providers = recipe.get("model_providers")
     if not providers:
-        return
+        return None
 
     # Collect local providers and pop is_local from ALL dicts unconditionally.
     # Strict `is True` guard so malformed payloads (is_local: 1,
@@ -115,7 +213,7 @@ def _inject_local_providers(recipe: dict[str, Any], request: Request) -> None:
             local_indices.append(i)
 
     if not local_indices:
-        return
+        return None
 
     endpoint = _resolve_local_v1_endpoint(request)
 
@@ -138,6 +236,7 @@ def _inject_local_providers(recipe: dict[str, Any], request: Request) -> None:
     }
 
     token = ""
+    internal_key_id: Optional[int] = None
     if local_names & referenced_providers:
         # Verify a model is loaded.
         # NOTE: This is a point-in-time check (TOCTOU). The model could be unloaded
@@ -158,18 +257,21 @@ def _inject_local_providers(recipe: dict[str, Any], request: Request) -> None:
                 "No model loaded in Chat. Load a model first, then run the recipe."
             )
 
-        from auth.authentication import (
-            create_access_token,
-        )  # deferred: avoids circular import
+        from auth import storage  # deferred: avoids circular import
 
-        # Uses the "unsloth" admin subject. If the user changes their password,
-        # the JWT secret rotates and this token becomes invalid mid-run.
-        # Acceptable for v1 - recipes typically finish well within one session.
-        token = create_access_token(
-            subject = "unsloth",
-            expires_delta = timedelta(hours = 24),
-            desktop = _request_has_desktop_access_token(request),
+        # Mint an internal sk-unsloth-* key scoped to this workflow run.
+        # Uses the unified API-key issuance path (one mint/revoke/verify
+        # surface instead of a second JWT code path). The key is marked
+        # internal so it is hidden from the user's API-key list, and the
+        # caller revokes it when the job terminates.
+        expires_at = (datetime.now(timezone.utc) + timedelta(hours = 24)).isoformat()
+        token, row = storage.create_api_key(
+            username = "unsloth",
+            name = "data-recipe workflow",
+            expires_at = expires_at,
+            internal = True,
         )
+        internal_key_id = int(row["id"])
 
     # Defensively strip any stale "external"-only fields the frontend may
     # have left on the dict (extra_headers/extra_body/api_key_env). The UI
@@ -196,6 +298,37 @@ def _inject_local_providers(recipe: dict[str, Any], request: Request) -> None:
             continue
         if mc.get("provider") in local_names:
             mc["skip_health_check"] = True
+            # Disable thinking for data-recipe inference on local providers.
+            # Reasoning models emit a <think>...</think> preamble before the
+            # answer, which roughly doubles generated token count per row and
+            # pushes the visible answer past data_designer's json-fence
+            # regex. Forward chat_template_kwargs={enable_thinking: False}
+            # through the OpenAI SDK's extra_body passthrough so llama-server
+            # renders the template without the reasoning preamble. Free-form
+            # llm-text columns benefit from the latency cut, and structured
+            # columns also stop leaking think tags into the grammar-
+            # constrained JSON (llama-server's GBNF path still enforces the
+            # schema either way).
+            params = mc.get("inference_parameters")
+            if not isinstance(params, dict):
+                params = {}
+                mc["inference_parameters"] = params
+            extra_body = params.get("extra_body")
+            if not isinstance(extra_body, dict):
+                extra_body = {}
+            tpl_kwargs = extra_body.get("chat_template_kwargs")
+            if not isinstance(tpl_kwargs, dict):
+                tpl_kwargs = {}
+            tpl_kwargs.setdefault("enable_thinking", False)
+            extra_body["chat_template_kwargs"] = tpl_kwargs
+            params["extra_body"] = extra_body
+
+    # Forward each llm-structured column's output_format as an OpenAI
+    # response_format so llama-server uses grammar-constrained sampling and
+    # small GGUFs stop wasting the full max_tokens budget on broken JSON.
+    _inject_local_structured_response_format(recipe, local_names)
+
+    return internal_key_id
 
 
 def _normalize_run_name(value: Any) -> str | None:
@@ -240,21 +373,49 @@ def create_job(payload: RecipePayload, request: Request):
             ) from exc
 
     try:
-        _inject_local_providers(recipe, request)
+        internal_api_key_id = _inject_local_providers(recipe, request)
     except ValueError as exc:
         raise HTTPException(status_code = 400, detail = str(exc)) from exc
 
-    mgr = get_job_manager()
+    # Single try block covers get_job_manager() AND mgr.start() so a workflow
+    # key minted above never outlives the request even when an unexpected
+    # exception type (TypeError from a stale kwarg, OSError from a queue
+    # write, etc.) bubbles up. Without the bare except, such exceptions let
+    # the sk-unsloth-* key live until its 24h TTL.
     try:
-        job_id = mgr.start(recipe = recipe, run = run)
+        mgr = get_job_manager()
+        job_id = mgr.start(
+            recipe = recipe,
+            run = run,
+            internal_api_key_id = internal_api_key_id,
+        )
     except RuntimeError as exc:
+        if internal_api_key_id is not None:
+            _revoke_internal_api_key_safe(internal_api_key_id)
         raise HTTPException(status_code = 409, detail = str(exc)) from exc
     except ValueError as exc:
+        if internal_api_key_id is not None:
+            _revoke_internal_api_key_safe(internal_api_key_id)
         raise HTTPException(status_code = 400, detail = str(exc)) from exc
+    except Exception:
+        if internal_api_key_id is not None:
+            _revoke_internal_api_key_safe(internal_api_key_id)
+        raise
 
     return {"job_id": job_id}
 
 
+def _revoke_internal_api_key_safe(key_id: int) -> None:
+    """Best-effort revoke of a workflow-minted key; swallow any error so
+    that revocation failures never mask the caller's own error path."""
+    try:
+        from auth import storage  # deferred: avoids circular import
+
+        storage.revoke_internal_api_key(key_id)
+    except Exception:
+        pass
+
+
 @router.get("/jobs/{job_id}/status")
 def job_status(job_id: str):
     mgr = get_job_manager()
diff --git a/studio/backend/routes/data_recipe/seed.py b/studio/backend/routes/data_recipe/seed.py
index e9cf828610..91cf718e6e 100644
--- a/studio/backend/routes/data_recipe/seed.py
+++ b/studio/backend/routes/data_recipe/seed.py
@@ -8,6 +8,7 @@ from __future__ import annotations
 import base64
 import binascii
 import json
+import os
 import re
 from itertools import islice
 from pathlib import Path
@@ -627,3 +628,14 @@ def inspect_seed_upload(payload: SeedInspectUploadRequest) -> SeedInspectRespons
         split = None,
         subset = None,
     )
+
+
+@router.get("/seed/github/env-token")
+def get_github_env_token_status() -> dict:
+    """Report whether the server has a GH_TOKEN / GITHUB_TOKEN env var.
+
+    The value is never returned; the UI uses this to tell the user they
+    can leave the token field blank.
+    """
+    has_token = bool(os.environ.get("GH_TOKEN") or os.environ.get("GITHUB_TOKEN"))
+    return {"has_token": has_token}
diff --git a/studio/backend/routes/data_recipe/validate.py b/studio/backend/routes/data_recipe/validate.py
index 555e3eaa06..87eef939b4 100644
--- a/studio/backend/routes/data_recipe/validate.py
+++ b/studio/backend/routes/data_recipe/validate.py
@@ -18,6 +18,57 @@ from models.data_recipe import RecipePayload, ValidateError, ValidateResponse
 
 router = APIRouter()
 
+_GITHUB_VALIDATE_NOTE = "Recipe shape is valid. GitHub access and rate limits are checked when the run starts."
+_GITHUB_ITEM_TYPES = {"issues", "pulls", "commits"}
+
+
+def _github_seed_source(recipe: dict[str, Any]) -> dict[str, Any] | None:
+    seed_config = recipe.get("seed_config")
+    if not isinstance(seed_config, dict):
+        return None
+    source = seed_config.get("source")
+    if not isinstance(source, dict) or source.get("seed_type") != "github_repo":
+        return None
+    return source
+
+
+def _validate_github_seed_static(source: dict[str, Any]) -> list[ValidateError]:
+    errors: list[ValidateError] = []
+
+    repos = source.get("repos")
+    if not isinstance(repos, list) or not repos:
+        errors.append(ValidateError(message = "GitHub seed requires at least one repo."))
+    else:
+        for repo in repos:
+            if not isinstance(repo, str) or not repo.strip() or "/" not in repo:
+                errors.append(
+                    ValidateError(message = "GitHub repos must be owner/name strings.")
+                )
+                break
+
+    item_types = source.get("item_types")
+    if not isinstance(item_types, list) or not item_types:
+        errors.append(
+            ValidateError(message = "GitHub seed requires at least one item type.")
+        )
+    else:
+        invalid_items = [item for item in item_types if item not in _GITHUB_ITEM_TYPES]
+        if invalid_items:
+            errors.append(
+                ValidateError(
+                    message = "GitHub item types must be issues, pulls, or commits."
+                )
+            )
+
+    try:
+        limit = int(source.get("limit"))
+    except (TypeError, ValueError):
+        limit = 0
+    if limit < 1 or limit > 5000:
+        errors.append(ValidateError(message = "GitHub limit must be from 1 to 5000."))
+
+    return errors
+
 
 def _collect_validation_errors(recipe: dict[str, Any]) -> list[ValidateError]:
     try:
@@ -93,6 +144,22 @@ def validate(payload: RecipePayload) -> ValidateResponse:
 
     _patch_local_providers(recipe)
 
+    github_source = _github_seed_source(recipe)
+    if github_source is not None:
+        static_errors = _validate_github_seed_static(github_source)
+        if static_errors:
+            return ValidateResponse(valid = False, errors = static_errors)
+        try:
+            build_config_builder(recipe)
+        except Exception as exc:
+            detail = str(exc).strip() or "Validation failed."
+            return ValidateResponse(
+                valid = False,
+                errors = [ValidateError(message = detail)],
+                raw_detail = detail,
+            )
+        return ValidateResponse(valid = True, raw_detail = _GITHUB_VALIDATE_NOTE)
+
     try:
         validate_recipe(recipe)
     except RuntimeError as exc:
diff --git a/studio/backend/routes/inference.py b/studio/backend/routes/inference.py
index cf3b37a2fd..b13eb08967 100644
--- a/studio/backend/routes/inference.py
+++ b/studio/backend/routes/inference.py
@@ -1286,6 +1286,20 @@ async def openai_chat_completions(
     llama_backend = get_llama_cpp_backend()
     using_gguf = llama_backend.is_loaded
 
+    # OpenAI-SDK clients send ``chat_template_kwargs`` via ``extra_body``,
+    # which the SDK spreads into the request body at the top level. Studio's
+    # ChatCompletionRequest has ``extra="allow"`` so pydantic stashes them in
+    # ``model_extra``, but the typed ``payload.enable_thinking`` path is what
+    # downstream generators actually consume. Lift ``enable_thinking`` from
+    # the extra-body chat_template_kwargs onto the typed field so clients
+    # that only know the OpenAI shape (data_designer recipe runs, etc.)
+    # can still control the reasoning preamble.
+    _extra = getattr(payload, "model_extra", None)
+    if payload.enable_thinking is None and isinstance(_extra, dict):
+        _tpl_kw = _extra.get("chat_template_kwargs")
+        if isinstance(_tpl_kw, dict) and "enable_thinking" in _tpl_kw:
+            payload.enable_thinking = bool(_tpl_kw["enable_thinking"])
+
     # ── Determine which backend is active ─────────────────────
     if using_gguf:
         model_name = llama_backend.model_identifier or payload.model
@@ -1440,11 +1454,22 @@ async def openai_chat_completions(
     # carry `tool_calls` (content=None) — both of which are valid in
     # multi-turn client-side tool loops.
     _has_tool_messages = any(m.role == "tool" or m.tool_calls for m in payload.messages)
+    # Route guided-decoding requests through the verbatim passthrough so
+    # ``response_format`` (JSON schema) actually reaches llama-server and
+    # the model's GBNF-constrained output comes back unmodified. The
+    # non-passthrough GGUF path below calls ``generate_chat_completion``
+    # which has no response_format kwarg, so the schema gets silently
+    # dropped and data_designer falls back to free-form sampling. Guided
+    # decoding does not require ``supports_tools`` - the grammar machinery
+    # is independent of tool-call parsing.
+    _has_response_format = _extract_response_format(payload) is not None
+    _tools_passthrough = llama_backend.supports_tools and (
+        (payload.tools and len(payload.tools) > 0) or _has_tool_messages
+    )
     if (
         using_gguf
-        and llama_backend.supports_tools
         and not payload.enable_tools
-        and ((payload.tools and len(payload.tools) > 0) or _has_tool_messages)
+        and (_tools_passthrough or _has_response_format)
     ):
         # Preserve the vision guard that would otherwise run in the
         # non-passthrough path below: text-only tool-capable GGUFs
@@ -3652,6 +3677,8 @@ def _build_passthrough_payload(
     repetition_penalty = None,
     presence_penalty = None,
     tool_choice = "auto",
+    response_format = None,
+    chat_template_kwargs = None,
     backend_ctx = None,
 ):
     body = {
@@ -3680,6 +3707,17 @@ def _build_passthrough_payload(
         body["repeat_penalty"] = repetition_penalty
     if presence_penalty is not None:
         body["presence_penalty"] = presence_penalty
+    if response_format is not None:
+        # llama-server applies a GBNF grammar derived from the JSON schema
+        # when response_format is present. Field is documented flat at the
+        # request root (tools/server/README.md), which is also what the
+        # OpenAI SDK produces by spreading extra_body into the body top.
+        body["response_format"] = response_format
+    if chat_template_kwargs is not None:
+        # Propagate reasoning / template overrides (e.g. enable_thinking)
+        # so llama-server renders the Jinja template in the mode the caller
+        # asked for instead of whatever default the model was loaded with.
+        body["chat_template_kwargs"] = chat_template_kwargs
     return body
 
 
@@ -3990,6 +4028,20 @@ def _openai_messages_for_passthrough(payload) -> list[dict]:
     return messages
 
 
+def _extract_response_format(payload):
+    """Return the ``response_format`` field on an incoming ChatCompletionRequest
+    (or None). The model is declared with ``extra="allow"`` so pydantic stashes
+    unknown top-level fields in ``model_extra``; OpenAI-SDK clients spread
+    ``extra_body`` into the request body top level, which is where guided-
+    decoding recipes park their JSON-schema response_format.
+    """
+    extra = getattr(payload, "model_extra", None)
+    if not isinstance(extra, dict):
+        return None
+    rf = extra.get("response_format")
+    return rf if isinstance(rf, dict) else None
+
+
 def _build_openai_passthrough_body(payload, backend_ctx = None) -> dict:
     """Assemble the llama-server request body from a ChatCompletionRequest.
 
@@ -3999,6 +4051,12 @@ def _build_openai_passthrough_body(payload, backend_ctx = None) -> dict:
     """
     messages = _openai_messages_for_passthrough(payload)
     tool_choice = payload.tool_choice if payload.tool_choice is not None else "auto"
+    # When the caller asked for a specific reasoning mode, forward it to
+    # llama-server via chat_template_kwargs so the Jinja template renders
+    # with (or without) the reasoning preamble.
+    tpl_kwargs = None
+    if payload.enable_thinking is not None:
+        tpl_kwargs = {"enable_thinking": bool(payload.enable_thinking)}
     return _build_passthrough_payload(
         messages,
         payload.tools,
@@ -4012,6 +4070,8 @@ def _build_openai_passthrough_body(payload, backend_ctx = None) -> dict:
         repetition_penalty = payload.repetition_penalty,
         presence_penalty = payload.presence_penalty,
         tool_choice = tool_choice,
+        response_format = _extract_response_format(payload),
+        chat_template_kwargs = tpl_kwargs,
         backend_ctx = backend_ctx,
     )
 
@@ -4214,6 +4274,41 @@ async def _openai_passthrough_non_streaming(
             detail = f"llama-server error: {resp.text[:500]}",
         )
 
+    # Guided-decoding fence wrap. llama-server returns raw JSON that matches
+    # the schema (no surrounding markdown) because the GBNF grammar only
+    # emits the JSON object itself. data_designer's llm-structured parser
+    # looks for a ```json ... ``` markdown fence and discards unfenced
+    # output, which collapses a 100%-valid guided-decoding run to 0/N.
+    # Wrap each choice's content in the expected fence when the caller
+    # asked for guided decoding, leaving already-fenced content alone.
+    if _extract_response_format(payload) is not None:
+        try:
+            data = resp.json()
+            changed = False
+            for choice in data.get("choices", []):
+                if not isinstance(choice, dict):
+                    continue
+                msg = choice.get("message")
+                if not isinstance(msg, dict):
+                    continue
+                content = msg.get("content")
+                if not isinstance(content, str):
+                    continue
+                stripped = content.strip()
+                if not stripped or stripped.startswith("```"):
+                    continue
+                msg["content"] = f"```json\n{stripped}\n```"
+                changed = True
+            if changed:
+                return JSONResponse(content = data)
+        except Exception as exc:
+            # Wrap is best-effort; fall through to the verbatim body if
+            # the response is not JSON-shaped or the structure is unusual.
+            logger.warning(
+                "response_format fence wrap skipped: %s",
+                exc,
+            )
+
     # Pass the upstream body through as raw bytes — skips a redundant
     # parse+re-serialize round-trip and keeps the response truly
     # verbatim (matches the docstring). Status is guaranteed 200 by
diff --git a/studio/backend/tests/test_data_recipe_github_progress.py b/studio/backend/tests/test_data_recipe_github_progress.py
new file mode 100644
index 0000000000..8e8c3995f4
--- /dev/null
+++ b/studio/backend/tests/test_data_recipe_github_progress.py
@@ -0,0 +1,91 @@
+# SPDX-License-Identifier: AGPL-3.0-only
+# Copyright 2026-present the Unsloth AI Inc. team. All rights reserved. See /studio/LICENSE.AGPL-3.0
+
+from core.data_recipe.jobs.parse import apply_update, parse_log_message
+from core.data_recipe.jobs.types import Job
+from routes.data_recipe.validate import _GITHUB_VALIDATE_NOTE, validate
+from models.data_recipe import RecipePayload
+
+
+def test_github_page_log_updates_source_progress_without_cursor():
+    job = Job(job_id = "job-1")
+    job.source_progress_estimated_total = 200
+
+    update = parse_log_message(
+        "[unslothai/unsloth] issues page 2 (+15) cursor=abc123 remaining=2960"
+    )
+
+    assert update is not None
+    apply_update(job, update)
+
+    progress = job.source_progress
+    assert progress is not None
+    assert progress.source == "github"
+    assert progress.status == "fetching"
+    assert progress.repo == "unslothai/unsloth"
+    assert progress.resource == "issues"
+    assert progress.page == 2
+    assert progress.page_items == 15
+    assert progress.fetched_items == 15
+    assert progress.estimated_total == 200
+    assert progress.rate_remaining == 2960
+    assert progress.message is not None
+    assert "cursor" not in progress.message
+    assert "abc123" not in progress.message
+
+
+def test_github_rate_limit_log_updates_source_progress():
+    job = Job(job_id = "job-1")
+
+    update = parse_log_message("Rate limit hit. Sleeping 123s until reset.")
+
+    assert update is not None
+    apply_update(job, update)
+
+    progress = job.source_progress
+    assert progress is not None
+    assert progress.status == "rate_limited"
+    assert progress.retry_after_sec == 123
+    assert "resume automatically" in (progress.message or "")
+
+
+def test_github_real_sample_prs_and_trial_limit_are_parsed():
+    job = Job(job_id = "job-1")
+
+    for message in (
+        "[unslothai/unsloth] PRs page 4 (+25) cursor=abc123 remaining=4983",
+        "Trial limit reached for PRs (100)",
+    ):
+        update = parse_log_message(message)
+        assert update is not None
+        apply_update(job, update)
+
+    progress = job.source_progress
+    assert progress is not None
+    assert progress.repo == "unslothai/unsloth"
+    assert progress.resource == "pulls"
+    assert progress.page == 4
+    assert progress.fetched_items == 25
+    assert progress.rate_remaining == 4983
+    assert progress.message == "GitHub pulls trial limit reached (100)."
+
+
+def test_github_validate_skips_live_access_with_honest_note():
+    response = validate(
+        RecipePayload(
+            recipe = {
+                "seed_config": {
+                    "source": {
+                        "seed_type": "github_repo",
+                        "repos": ["unslothai/unsloth"],
+                        "item_types": ["issues"],
+                        "limit": 1,
+                    }
+                },
+                "columns": [{"column_type": "expression", "name": "x", "expr": "1"}],
+            }
+        )
+    )
+
+    assert response.valid is True
+    assert response.raw_detail == _GITHUB_VALIDATE_NOTE
diff --git a/studio/frontend/src/features/data-recipes/learning-recipes/github-support-bot.json b/studio/frontend/src/features/data-recipes/learning-recipes/github-support-bot.json
new file mode 100644
index 0000000000..b984f84d2f
--- /dev/null
+++ b/studio/frontend/src/features/data-recipes/learning-recipes/github-support-bot.json
@@ -0,0 +1,238 @@
+{
+  "recipe": {
+    "model_providers": [
+      {
+        "name": "Local Model",
+        "endpoint": "",
+        "provider_type": "openai",
+        "extra_headers": {},
+        "extra_body": {},
+        "is_local": true
+      }
+    ],
+    "mcp_providers": [],
+    "model_configs": [
+      {
+        "alias": "model_1",
+        "model": "unsloth/gemma-4-E2B-it-GGUF",
+        "provider": "Local Model",
+        "inference_parameters": {
+          "temperature": 0.4,
+          "max_tokens": 800
+        }
+      }
+    ],
+    "seed_config": {
+      "source": {
+        "seed_type": "github_repo",
+        "repos": [
+          "unslothai/unsloth",
+          "unslothai/unsloth-zoo"
+        ],
+        "item_types": [
+          "issues",
+          "pulls"
+        ],
+        "limit": 100,
+        "include_comments": true,
+        "max_comments_per_item": 20
+      },
+      "sampling_strategy": "shuffle",
+      "selection_strategy": null
+    },
+    "tool_configs": [],
+    "columns": [
+      {
+        "column_type": "llm-text",
+        "name": "User",
+        "drop": false,
+        "model_alias": "model_1",
+        "prompt": "Read the GitHub {{ item_type }} below and write ONE realistic user request that could have produced it. Imagine a developer asking a GitHub co-author model to either file this {{ item_type }} or draft a PR that resolves it. Use first-person imperative phrasing (\"Open an issue...\", \"Draft a PR that...\", \"Investigate why...\"). Preserve concrete technical details (model names, flags, file paths, tracebacks) that appear in the thread. Keep it 1-3 sentences. Output ONLY the user request, no preamble.\n\n--- INPUT ---\nRepo: {{ repo }}\nType: {{ item_type }}\nTitle: {{ title }}\nBody:\n{{ body }}\n\nFirst comments:\n{{ comments }}",
+        "system_prompt": "You invert real GitHub threads into the user request that would have produced them. Faithful to the thread, no invented facts, no em-dashes, no emojis.",
+        "with_trace": "none",
+        "extract_reasoning_content": false
+      },
+      {
+        "column_type": "llm-structured",
+        "name": "Assistant",
+        "drop": false,
+        "model_alias": "model_1",
+        "prompt": "You are generating one training row for an Unsloth GitHub co-author model. Given the real GitHub thread and a synthesized user request, produce a grounded structured response.\n\nSource thread:\n- Repo: {{ repo }}\n- Type: {{ item_type }}\n- Title: {{ title }}\n- URL: {{ url }}\n- State: {{ state }}\n- Labels: {{ labels }}\n- Body: {{ body }}\n- First comments: {{ comments }}\n\nUser request:\n{{ User }}\n\nRules:\n- `response`: 100-250 words of Markdown grounded in the thread. If the thread is a closed / resolved issue, follow the `issue_fix_plan` shape: brief diagnosis, numbered fix steps, and a short repro. If the thread is a PR, follow the `explain_pr` shape: what changed, why, and which files or symbols were touched. If the thread is open / unresolved, answer honestly and ask for the missing info.\n- Cite the source URL at least once inline as `[source: {{ url }}]`.\n- Name at least one concrete symbol (function, class, flag, env var, or file path) from the thread when available.\n- Include a short ```bash or ```python code block ONLY if the thread itself contains that code or command.\n- Never recommend `rm -rf`, force push, or other destructive commands without an explicit warning.\n- No em-dashes, no emojis, no AI-disclaimer phrases. Only cite URLs / paths that appear in the thread.\n- `followups`: 0-4 follow-up questions when the thread is missing info (versions, GPU, traceback). Empty list if the response is complete.\n- `cites`: URLs / file paths actually used. Always include `{{ url }}`.\n- `task`: one of `explain_pr`, `issue_fix_plan`, `issue_solution`, `discussion_qa`. Pick the closest match.\n- `confidence`: `high` / `medium` / `low`. Use `low` when ambiguous or out of scope.",
+        "system_prompt": "You write grounded GitHub co-author responses for Unsloth. Faithful to the thread, no invented facts, no em-dashes, no emojis, no AI-disclaimer phrases.",
+        "with_trace": "none",
+        "extract_reasoning_content": false,
+        "output_format": {
+          "type": "object",
+          "properties": {
+            "response": {
+              "type": "string",
+              "minLength": 1
+            },
+            "followups": {
+              "type": "array",
+              "items": {
+                "type": "string"
+              },
+              "maxItems": 4
+            },
+            "cites": {
+              "type": "array",
+              "items": {
+                "type": "string"
+              },
+              "maxItems": 6
+            },
+            "task": {
+              "type": "string",
+              "enum": [
+                "explain_pr",
+                "issue_fix_plan",
+                "issue_solution",
+                "discussion_qa"
+              ]
+            },
+            "confidence": {
+              "type": "string",
+              "enum": [
+                "high",
+                "medium",
+                "low"
+              ]
+            }
+          },
+          "required": [
+            "response",
+            "followups",
+            "cites",
+            "task",
+            "confidence"
+          ],
+          "additionalProperties": false
+        }
+      }
+    ],
+    "processors": []
+  },
+  "run": {
+    "rows": 5,
+    "preview": true,
+    "output_formats": [
+      "jsonl"
+    ]
+  },
+  "ui": {
+    "nodes": [
+      {
+        "id": "note_1",
+        "x": 460.4272962489311,
+        "y": 123.10114554082236,
+        "width": 400,
+        "node_type": "markdown_note",
+        "name": "note_1",
+        "markdown": "### GitHub Crawler\nReal GitHub issues and PRs turned into `{User, Assistant}` training pairs. Mirrors two of the eleven canonical enrichment tasks in the `github_data_gatherer` dataset: `pr_requests_20` / `issue_requests_20` for the input side, and `explain_pr` / `issue_fix_plan` / `issue_solution` for the output side.\n\n**Click `Run` below for 10 sample rows.** Defaults point at `unslothai/unsloth` + `unslothai/unsloth-zoo`, use the server's `GH_TOKEN` / `GITHUB_TOKEN` env var, and run the bundled local model. Paste a PAT only when you need private repos.\n\n**Configure source data**\n- Paste `owner/name` values or GitHub URLs; the editor normalizes and dedupes them.\n- Keep the seed `limit` around 100 for previews. Increase toward 5000 per item type for larger backfills.\n- Comments and large repos can make Check or Run take minutes; watch logs for GitHub page and rate-limit messages.\n\n**Upgrade to production**\n- Swap `unsloth/gemma-4-E2B-it-GGUF` for a stronger model (`gpt-5.4-mini` with `reasoning_effort=medium` is what the reference dataset uses).\n- Replace the demo prompts with the task-specific prompts from the reference dataset (see Note 3).\n- Raise `max_parallel_requests` to 4 once the inference server can handle it.",
+        "note_color": "#E0F2FE",
+        "note_opacity": "35"
+      },
+      {
+        "id": "note_2",
+        "x": -514.4586648521598,
+        "y": 869.2785862193363,
+        "width": 400,
+        "node_type": "markdown_note",
+        "name": "note_2",
+        "markdown": "The **User** column inverts each GitHub thread into a realistic request a developer would give a co-author model (`\"Draft a PR that...\"`, `\"Investigate why...\"`). Same shape as the `pr_requests_20` and `issue_requests_20` enrichments.\n\nTweak the prompt to:\n- always keep the traceback verbatim\n- vary persona (newcomer, maintainer, ops)\n- split one thread into multiple alternative phrasings for data augmentation.",
+        "note_color": "#E0F2FE",
+        "note_opacity": "35"
+      },
+      {
+        "id": "note_3",
+        "x": -536.3163554442993,
+        "y": -84.18780704666585,
+        "width": 400,
+        "node_type": "markdown_note",
+        "name": "note_3",
+        "markdown": "The **Assistant** block emits `{response, followups, cites, task, confidence}` and branches on thread type: closed issues become `issue_fix_plan` rows, PRs become `explain_pr` rows, everything else becomes `issue_solution` or `discussion_qa`.\n\n**Demo default**: 100-250 word response, one inline `[source: <url>]` cite, `max_parallel_requests=1` so a small local model stays stable.\n\n**Production prompt (paste in):**\n- Match the reference dataset's per-task prompts (`explain_pr`, `issue_fix_plan`, `pr_review_critique`, `pr_test_plan`, etc.).\n- Require 300 words for explanations, 6-12 bullets for test plans.\n- Enforce named symbols (function / flag / env var / file path).\n- Code fences only for content already in the thread.\n- Reject rows with em-dashes, emojis, or AI-disclaimer phrases.\n\nSee the `github_data_gatherer` dataset card for the full task catalog and the codex prompts used to train the reference GitHub model.",
+        "note_color": "#E0F2FE",
+        "note_opacity": "35"
+      },
+      {
+        "id": "seed",
+        "x": 0,
+        "y": 140,
+        "width": 400
+      },
+      {
+        "id": "Local Model",
+        "x": -1056,
+        "y": 520,
+        "width": 400
+      },
+      {
+        "id": "model_1",
+        "x": -544,
+        "y": 488,
+        "width": 400
+      },
+      {
+        "id": "User",
+        "x": 0,
+        "y": 440,
+        "width": 400
+      },
+      {
+        "id": "Assistant",
+        "x": 0,
+        "y": 740,
+        "width": 400
+      }
+    ],
+    "edges": [
+      {
+        "from": "seed",
+        "to": "User",
+        "type": "canvas",
+        "source_handle": "data-out-bottom",
+        "target_handle": "data-in-top"
+      },
+      {
+        "from": "User",
+        "to": "Assistant",
+        "type": "canvas",
+        "source_handle": "data-out-bottom",
+        "target_handle": "data-in-top"
+      },
+      {
+        "from": "Local Model",
+        "to": "model_1",
+        "type": "semantic",
+        "source_handle": "semantic-out",
+        "target_handle": "semantic-in"
+      },
+      {
+        "from": "model_1",
+        "to": "User",
+        "type": "semantic",
+        "source_handle": "semantic-out",
+        "target_handle": "data-in"
+      },
+      {
+        "from": "model_1",
+        "to": "Assistant",
+        "type": "semantic",
+        "source_handle": "semantic-out-bottom",
+        "target_handle": "data-in"
+      }
+    ],
+    "layout_direction": "LR",
+    "seed_source_type": "github_repo",
+    "seed_columns": [],
+    "seed_drop_columns": [],
+    "seed_preview_rows": [],
+    "local_file_name": "",
+    "unstructured_file_ids": [],
+    "unstructured_file_names": [],
+    "unstructured_file_sizes": [],
+    "unstructured_chunk_size": "1200",
+    "unstructured_chunk_overlap": "200"
+  }
+}
\ No newline at end of file
diff --git a/studio/frontend/src/features/data-recipes/learning-recipes/index.ts b/studio/frontend/src/features/data-recipes/learning-recipes/index.ts
index d7a7e66e0a..8607d3faeb 100644
--- a/studio/frontend/src/features/data-recipes/learning-recipes/index.ts
+++ b/studio/frontend/src/features/data-recipes/learning-recipes/index.ts
@@ -19,6 +19,10 @@ const ocrDocumentExtractionUrl = new URL(
   "./ocr-document-extraction.json",
   import.meta.url,
 ).href;
+const githubSupportBotUrl = new URL(
+  "./github-support-bot.json",
+  import.meta.url,
+).href;
 
 function isRecord(value: unknown): value is Record<string, unknown> {
   return !!value && typeof value === "object" && !Array.isArray(value);
@@ -137,4 +141,11 @@ export const LEARNING_RECIPES: LearningRecipeDef[] = [
       "Use image context to generate OCR-style document extraction output.",
     loadPayload: () => loadPayloadFromUrl(ocrDocumentExtractionUrl),
   },
+  {
+    id: "github-support-bot",
+    title: "GitHub Crawler",
+    description:
+      "Crawl real GitHub issues and PRs and turn each thread into a {User, Assistant} training pair.",
+    loadPayload: () => loadPayloadFromUrl(githubSupportBotUrl),
+  },
 ];
diff --git a/studio/frontend/src/features/data-recipes/pages/data-recipes-page.tsx b/studio/frontend/src/features/data-recipes/pages/data-recipes-page.tsx
index 9148b9e0da..0a51f9df70 100644
--- a/studio/frontend/src/features/data-recipes/pages/data-recipes-page.tsx
+++ b/studio/frontend/src/features/data-recipes/pages/data-recipes-page.tsx
@@ -35,6 +35,7 @@ import {
   Delete02Icon,
   DocumentAttachmentIcon,
   FunctionIcon,
+  GithubIcon,
   Plant01Icon,
   PlusSignIcon,
 } from "@hugeicons/core-free-icons";
@@ -162,6 +163,22 @@ const TEMPLATE_CARDS: TemplateCard[] = [
     ],
     learningRecipeId: "structured-outputs-jinja",
   },
+  {
+    title: "GitHub Crawler",
+    description:
+      "Crawl real GitHub issues and PRs and invert each thread into a {User, Assistant} training pair.",
+    icon: GithubIcon,
+    difficulty: "Intermediate",
+    learningBadges: ["GitHub", "LLM Text", "Structured LLM"],
+    surfaceClassName:
+      "from-slate-500/15 via-zinc-500/5 to-transparent dark:from-slate-400/30 dark:via-zinc-400/14 dark:to-slate-950/16",
+    shineColor: [
+      "rgb(71 85 105 / 0.45)",
+      "rgb(100 116 139 / 0.4)",
+      "rgb(148 163 184 / 0.45)",
+    ],
+    learningRecipeId: "github-support-bot",
+  },
 ];
 
 const LEARNING_RECIPE_BY_ID = new Map(
diff --git a/studio/frontend/src/features/recipe-studio/api/index.ts b/studio/frontend/src/features/recipe-studio/api/index.ts
index 9212e4db9b..273d4aea8d 100644
--- a/studio/frontend/src/features/recipe-studio/api/index.ts
+++ b/studio/frontend/src/features/recipe-studio/api/index.ts
@@ -27,6 +27,28 @@ export type PublishRecipeJobResponse = {
   message: string;
 };
 
+export type SourceProgressResponse = {
+  source?: string | null;
+  status?: string | null;
+  repo?: string | null;
+  resource?: string | null;
+  page?: number | null;
+  // biome-ignore lint/style/useNamingConvention: api schema
+  page_items?: number | null;
+  // biome-ignore lint/style/useNamingConvention: api schema
+  fetched_items?: number | null;
+  // biome-ignore lint/style/useNamingConvention: api schema
+  estimated_total?: number | null;
+  percent?: number | null;
+  // biome-ignore lint/style/useNamingConvention: api schema
+  rate_remaining?: number | null;
+  // biome-ignore lint/style/useNamingConvention: api schema
+  retry_after_sec?: number | null;
+  message?: string | null;
+  // biome-ignore lint/style/useNamingConvention: api schema
+  updated_at?: number | null;
+};
+
 export type JobStatusResponse = {
   // biome-ignore lint/style/useNamingConvention: api schema
   job_id: string;
@@ -62,6 +84,8 @@ export type JobStatusResponse = {
     failed?: number | null;
   };
   // biome-ignore lint/style/useNamingConvention: api schema
+  source_progress?: SourceProgressResponse | null;
+  // biome-ignore lint/style/useNamingConvention: api schema
   model_usage?: Record<string, unknown>;
   rows?: number | null;
   cols?: number | null;
@@ -183,12 +207,7 @@ async function parseErrorResponse(response: Response): Promise<string> {
       // biome-ignore lint/style/useNamingConvention: api schema
       raw_detail?: string;
     };
-    return (
-      parsed.detail ??
-      parsed.message ??
-      parsed.raw_detail ??
-      text
-    );
+    return parsed.detail ?? parsed.message ?? parsed.raw_detail ?? text;
   } catch {
     return text;
   }
@@ -264,11 +283,15 @@ export async function validateRecipe(
   return postJson<ValidateResponse>("/validate", payload);
 }
 
-export async function createRecipeJob(payload: unknown): Promise<JobCreateResponse> {
+export async function createRecipeJob(
+  payload: unknown,
+): Promise<JobCreateResponse> {
   return postJson<JobCreateResponse>("/jobs", payload);
 }
 
-export async function getRecipeJobStatus(jobId: string): Promise<JobStatusResponse> {
+export async function getRecipeJobStatus(
+  jobId: string,
+): Promise<JobStatusResponse> {
   return getJson<JobStatusResponse>(`/jobs/${jobId}/status`);
 }
 
@@ -292,7 +315,9 @@ export async function getRecipeJobDataset(
   );
 }
 
-export async function cancelRecipeJob(jobId: string): Promise<JobStatusResponse> {
+export async function cancelRecipeJob(
+  jobId: string,
+): Promise<JobStatusResponse> {
   return postJson<JobStatusResponse>(`/jobs/${jobId}/cancel`, {});
 }
 
@@ -315,6 +340,13 @@ export async function inspectSeedUpload(
   return postJson<SeedInspectResponse>("/seed/inspect-upload", payload);
 }
 
+// biome-ignore lint/style/useNamingConvention: api schema
+export type GithubEnvTokenStatus = { has_token: boolean };
+
+export async function getGithubEnvTokenStatus(): Promise<GithubEnvTokenStatus> {
+  return getJson<GithubEnvTokenStatus>("/seed/github/env-token");
+}
+
 export async function listMcpTools(
   payload: McpToolsListRequest,
 ): Promise<McpToolsListResponse> {
@@ -407,11 +439,14 @@ export async function uploadUnstructuredFile(
     formData.append("existing_file_ids", existingFileIds.join(","));
   }
 
-  const res = await authFetch(`${DATA_DESIGNER_API_BASE}/seed/upload-unstructured-file`, {
-    method: "POST",
-    body: formData,
-    signal,
-  });
+  const res = await authFetch(
+    `${DATA_DESIGNER_API_BASE}/seed/upload-unstructured-file`,
+    {
+      method: "POST",
+      body: formData,
+      signal,
+    },
+  );
 
   if (res.status === 413) {
     const detail = await res.json().catch(() => ({ detail: "File too large" }));
@@ -420,13 +455,16 @@ export async function uploadUnstructuredFile(
       filename: file.name,
       size_bytes: file.size,
       status: "error",
-      error: typeof detail.detail === "string" ? detail.detail : "File too large",
+      error:
+        typeof detail.detail === "string" ? detail.detail : "File too large",
     };
   }
 
   if (!res.ok) {
     const detail = await res.json().catch(() => ({ detail: "Upload failed" }));
-    throw new Error(typeof detail.detail === "string" ? detail.detail : "Upload failed");
+    throw new Error(
+      typeof detail.detail === "string" ? detail.detail : "Upload failed",
+    );
   }
 
   return res.json();
diff --git a/studio/frontend/src/features/recipe-studio/blocks/definitions.ts b/studio/frontend/src/features/recipe-studio/blocks/definitions.ts
index 363cd85be5..ed58d1418d 100644
--- a/studio/frontend/src/features/recipe-studio/blocks/definitions.ts
+++ b/studio/frontend/src/features/recipe-studio/blocks/definitions.ts
@@ -12,6 +12,7 @@ import {
   EqualSignIcon,
   FingerPrintIcon,
   FunctionIcon,
+  GithubIcon,
   Plug01Icon,
   Parabola02Icon,
   PencilEdit02Icon,
@@ -58,11 +59,16 @@ export type BlockType =
   | "seed_hf"
   | "seed_local"
   | "seed_unstructured"
+  | "seed_github"
   | "model_provider"
   | "model_config"
   | "tool_config";
 
-export type SeedBlockType = "seed_hf" | "seed_local" | "seed_unstructured";
+export type SeedBlockType =
+  | "seed_hf"
+  | "seed_local"
+  | "seed_unstructured"
+  | "seed_github";
 
 type IconType = typeof CodeIcon;
 
@@ -169,6 +175,15 @@ const BLOCK_DEFINITIONS: BlockDefinition[] = [
     dialogKey: "seed",
     createConfig: (id, existing) => makeSeedConfig(id, existing, "unstructured"),
   },
+  {
+    kind: "seed",
+    type: "seed_github",
+    title: "GitHub repositories",
+    description: "Crawl issues, pull requests, and commits from one or more GitHub repos.",
+    icon: GithubIcon,
+    dialogKey: "seed",
+    createConfig: (id, existing) => makeSeedConfig(id, existing, "github_repo"),
+  },
   {
     kind: "sampler",
     type: "category",
@@ -388,6 +403,7 @@ export function getBlockDefinitionForConfig(
       hf: "seed_hf",
       local: "seed_local",
       unstructured: "seed_unstructured",
+      github_repo: "seed_github",
     };
     return getBlockDefinition("seed", seedType[config.seed_source_type ?? "hf"]);
   }
diff --git a/studio/frontend/src/features/recipe-studio/components/executions/execution-data-tab.tsx b/studio/frontend/src/features/recipe-studio/components/executions/execution-data-tab.tsx
index a8e12ac1ca..a3e5dbe962 100644
--- a/studio/frontend/src/features/recipe-studio/components/executions/execution-data-tab.tsx
+++ b/studio/frontend/src/features/recipe-studio/components/executions/execution-data-tab.tsx
@@ -2,6 +2,8 @@
 // Copyright 2026-present the Unsloth AI Inc. team. All rights reserved. See /studio/LICENSE.AGPL-3.0
 
 import type { ReactElement } from "react";
+import { GithubIcon } from "@hugeicons/core-free-icons";
+import { HugeiconsIcon } from "@hugeicons/react";
 import type { ColumnDef } from "@tanstack/react-table";
 import { Button } from "@/components/ui/button";
 import { DataTable } from "@/components/ui/data-table";
@@ -12,11 +14,136 @@ import {
   DropdownMenuLabel,
   DropdownMenuTrigger,
 } from "@/components/ui/dropdown-menu";
+import { Progress } from "@/components/ui/progress";
 import { Spinner } from "@/components/ui/spinner";
-import { cn } from "@/lib/utils";
 import { isExecutionInProgress } from "../../executions/execution-helpers";
 import type { RecipeExecutionRecord } from "../../execution-types";
-import { hasExpandableTextCell } from "./executions-view-helpers";
+import { formatMetricValue } from "./executions-view-helpers";
+
+function formatSourceResource(value: string | null | undefined): string {
+  if (value === "pulls") {
+    return "PRs";
+  }
+  return value ?? "--";
+}
+
+function formatFetchedValue(execution: RecipeExecutionRecord): string {
+  const source = execution.source_progress;
+  if (!source) {
+    return "--";
+  }
+  const fetched = formatMetricValue(source.fetched_items);
+  if (typeof source.estimated_total !== "number" || source.estimated_total <= 0) {
+    return fetched;
+  }
+  return `${fetched} / ${formatMetricValue(source.estimated_total)}`;
+}
+
+function formatGitHubSourceMessage(execution: RecipeExecutionRecord): string {
+  const source = execution.source_progress;
+  if (!source) {
+    return "Collecting repository threads before rows are available.";
+  }
+  if (source.status === "rate_limited") {
+    return source.message ?? "Waiting for GitHub rate limit. Studio will resume automatically.";
+  }
+  return source.message ?? "Collecting repository threads before rows are available.";
+}
+
+function RunningDatasetEmptyState({
+  execution,
+  onOpenOverview,
+}: {
+  execution: RecipeExecutionRecord;
+  onOpenOverview: () => void;
+}): ReactElement {
+  const source = execution.source_progress;
+  if (source?.source === "github") {
+    const title =
+      source.status === "rate_limited"
+        ? "Waiting for GitHub rate limit"
+        : "Crawling GitHub source";
+    const showProgress = typeof source.percent === "number";
+
+    return (
+      <div className="rounded-xl border border-border/60 bg-card/55 p-4 shadow-border">
+        <div className="flex flex-col gap-3 sm:flex-row sm:items-start sm:justify-between">
+          <div className="flex min-w-0 gap-3">
+            <div
+              className="flex size-9 shrink-0 items-center justify-center rounded-lg border border-border/70 bg-muted/20"
+            >
+              {showProgress ? (
+                <HugeiconsIcon icon={GithubIcon} className="size-4 text-muted-foreground" />
+              ) : (
+                <Spinner className="size-4 text-muted-foreground" />
+              )}
+            </div>
+            <div className="min-w-0 space-y-1">
+              <p className="text-sm font-semibold text-foreground">{title}</p>
+              <p className="text-xs text-muted-foreground">
+                {formatGitHubSourceMessage(execution)}
+              </p>
+            </div>
+          </div>
+          <Button type="button" size="sm" variant="outline" onClick={onOpenOverview}>
+            Open Overview
+          </Button>
+        </div>
+        {showProgress && <Progress value={source.percent ?? 0} className="mt-3 h-1" />}
+        <div className="mt-3 grid gap-2 text-xs sm:grid-cols-2 lg:grid-cols-5">
+          <p className="flex items-center justify-between gap-3">
+            <span className="text-muted-foreground">Repo</span>
+            <span className="truncate font-semibold">{source.repo ?? "--"}</span>
+          </p>
+          <p className="flex items-center justify-between gap-3">
+            <span className="text-muted-foreground">Resource</span>
+            <span className="font-semibold">
+              {formatSourceResource(source.resource)}
+              {typeof source.page === "number" ? ` page ${source.page}` : ""}
+            </span>
+          </p>
+          <p className="flex items-center justify-between gap-3">
+            <span className="text-muted-foreground">Fetched</span>
+            <span className="font-semibold">{formatFetchedValue(execution)}</span>
+          </p>
+          <p className="flex items-center justify-between gap-3">
+            <span className="text-muted-foreground">Rate remaining</span>
+            <span className="font-semibold">{formatMetricValue(source.rate_remaining)}</span>
+          </p>
+          <p className="flex items-center justify-between gap-3">
+            <span className="text-muted-foreground">Retry wait</span>
+            <span className="font-semibold">
+              {typeof source.retry_after_sec === "number"
+                ? `${formatMetricValue(source.retry_after_sec)}s`
+                : "--"}
+            </span>
+          </p>
+        </div>
+      </div>
+    );
+  }
+
+  return (
+    <div className="rounded-xl border border-border/60 bg-card/55 p-4 text-center shadow-border">
+      <div className="flex flex-col items-center justify-center gap-3 py-6">
+        <Spinner className="size-5" />
+        <div className="space-y-1">
+          <p className="text-sm font-medium text-muted-foreground">
+            Generating data…
+          </p>
+          <p className="text-xs text-muted-foreground">
+            {execution.current_column
+              ? `Current column: ${execution.current_column}`
+              : "Rows will appear here once the run produces a dataset sample."}
+          </p>
+        </div>
+        <Button type="button" size="sm" variant="outline" onClick={onOpenOverview}>
+          Open Overview
+        </Button>
+      </div>
+    </div>
+  );
+}
 
 type ExecutionDataTabProps = {
   execution: RecipeExecutionRecord;
@@ -27,13 +154,10 @@ type ExecutionDataTabProps = {
   totalPages: number;
   tableColumns: ColumnDef<Record<string, unknown>>[];
   datasetRowsForTable: Record<string, unknown>[];
-  visibleDatasetColumnNames: string[];
-  expandedDatasetRows: Record<string, boolean>;
-  selectedExecutionIdSafe: string | null;
+  onOpenOverview: () => void;
   onSetHiddenColumns: (updater: (current: string[]) => string[]) => void;
   onPrevPage: () => void;
   onNextPage: () => void;
-  onToggleRowExpanded: (rowId: string) => void;
 };
 
 export function ExecutionDataTab({
@@ -45,13 +169,10 @@ export function ExecutionDataTab({
   totalPages,
   tableColumns,
   datasetRowsForTable,
-  visibleDatasetColumnNames,
-  expandedDatasetRows,
-  selectedExecutionIdSafe,
+  onOpenOverview,
   onSetHiddenColumns,
   onPrevPage,
   onNextPage,
-  onToggleRowExpanded,
 }: ExecutionDataTabProps): ReactElement {
   return (
     <div className="mt-3">
@@ -123,17 +244,10 @@ export function ExecutionDataTab({
       </div>
       {execution.dataset.length === 0 ? (
         isExecutionInProgress(execution.status) ? (
-          <div className="flex flex-col items-center justify-center gap-3 py-12 text-center">
-            <Spinner className="size-5" />
-            <div className="space-y-1">
-              <p className="text-sm font-medium text-muted-foreground">
-                Generating data…
-              </p>
-              <p className="text-xs text-muted-foreground">
-                Check the Overview tab for live terminal logs.
-              </p>
-            </div>
-          </div>
+          <RunningDatasetEmptyState
+            execution={execution}
+            onOpenOverview={onOpenOverview}
+          />
         ) : (
           <p className="text-xs text-muted-foreground">No rows returned.</p>
         )
@@ -143,27 +257,7 @@ export function ExecutionDataTab({
         </p>
       ) : (
         <div className="max-h-[55vh] overflow-auto">
-          <DataTable
-            columns={tableColumns}
-            data={datasetRowsForTable}
-            getRowClassName={(row, _rowIndex, rowId) => {
-              const canExpand = hasExpandableTextCell(row, visibleDatasetColumnNames);
-              if (!canExpand) {
-                return undefined;
-              }
-              return cn(
-                "cursor-pointer",
-                expandedDatasetRows[rowId] ? "bg-primary/[0.05]" : "hover:bg-primary/[0.06]",
-              );
-            }}
-            onRowClick={(row, _rowIndex, rowId) => {
-              const canExpand = hasExpandableTextCell(row, visibleDatasetColumnNames);
-              if (!canExpand || !selectedExecutionIdSafe) {
-                return;
-              }
-              onToggleRowExpanded(rowId);
-            }}
-          />
+          <DataTable columns={tableColumns} data={datasetRowsForTable} />
         </div>
       )}
     </div>
diff --git a/studio/frontend/src/features/recipe-studio/components/executions/execution-overview-tab.tsx b/studio/frontend/src/features/recipe-studio/components/executions/execution-overview-tab.tsx
index 8e2e710299..72624738fc 100644
--- a/studio/frontend/src/features/recipe-studio/components/executions/execution-overview-tab.tsx
+++ b/studio/frontend/src/features/recipe-studio/components/executions/execution-overview-tab.tsx
@@ -6,6 +6,7 @@ import {
   Database01Icon,
   Database02Icon,
   Flag02Icon,
+  GithubIcon,
 } from "@hugeicons/core-free-icons";
 import { HugeiconsIcon } from "@hugeicons/react";
 import { Badge } from "@/components/ui/badge";
@@ -23,6 +24,28 @@ import type { RecipeExecutionRecord } from "../../execution-types";
 import type { ModelUsageRow } from "./executions-view-helpers";
 import { formatMetricValue } from "./executions-view-helpers";
 
+function formatSourceResource(value: string | null | undefined): string {
+  if (value === "pulls") {
+    return "PRs";
+  }
+  return value ?? "--";
+}
+
+function formatSourceMessage(execution: RecipeExecutionRecord): string {
+  const source = execution.source_progress;
+  if (!source) {
+    return "No source progress captured.";
+  }
+  if (source.status === "rate_limited") {
+    const wait =
+      typeof source.retry_after_sec === "number" && source.retry_after_sec > 0
+        ? ` Waiting ~${formatMetricValue(source.retry_after_sec)}s.`
+        : "";
+    return `Waiting for GitHub rate limit. Studio will resume automatically.${wait}`;
+  }
+  return source.message ?? "Crawling GitHub source.";
+}
+
 type ExecutionOverviewTabProps = {
   execution: RecipeExecutionRecord;
   showSummaryCards: boolean;
@@ -60,6 +83,8 @@ export function ExecutionOverviewTab({
   canPublish,
   onOpenPublish,
 }: ExecutionOverviewTabProps): ReactElement {
+  const sourceProgress = execution.source_progress;
+
   return (
     <div className="mt-3 space-y-3">
       {showSummaryCards && (
@@ -204,6 +229,47 @@ export function ExecutionOverviewTab({
           )}
         </div>
       )}
+      {sourceProgress?.source === "github" && (
+        <div className="rounded-xl border border-border/60 bg-card/55 p-3">
+          <div className="mb-2 flex items-center justify-between">
+            <p className="text-xs text-muted-foreground">Source data</p>
+            <HugeiconsIcon icon={GithubIcon} className="size-4 text-muted-foreground" />
+          </div>
+          <p className="text-sm font-medium text-foreground">
+            {sourceProgress.status === "completed"
+              ? "GitHub source complete"
+              : "Crawling GitHub source"}
+          </p>
+          <p className="mt-1 text-xs text-muted-foreground">
+            {formatSourceMessage(execution)}
+          </p>
+          <div className="mt-2 grid gap-1.5 text-xs sm:grid-cols-2 lg:grid-cols-4">
+            <p className="flex items-center justify-between gap-3">
+              <span className="text-muted-foreground">Repo</span>
+              <span className="truncate font-semibold">{sourceProgress.repo ?? "--"}</span>
+            </p>
+            <p className="flex items-center justify-between gap-3">
+              <span className="text-muted-foreground">Resource</span>
+              <span className="font-semibold">
+                {formatSourceResource(sourceProgress.resource)}
+                {typeof sourceProgress.page === "number" ? ` page ${sourceProgress.page}` : ""}
+              </span>
+            </p>
+            <p className="flex items-center justify-between gap-3">
+              <span className="text-muted-foreground">Fetched</span>
+              <span className="font-semibold">
+                {formatMetricValue(sourceProgress.fetched_items)}
+              </span>
+            </p>
+            <p className="flex items-center justify-between gap-3">
+              <span className="text-muted-foreground">Rate remaining</span>
+              <span className="font-semibold">
+                {formatMetricValue(sourceProgress.rate_remaining)}
+              </span>
+            </p>
+          </div>
+        </div>
+      )}
       <div className="overflow-hidden rounded-xl corner-squircle border">
         <div className="flex items-center justify-between border-b px-3 py-2">
           <p className="text-sm font-semibold">Terminal output</p>
diff --git a/studio/frontend/src/features/recipe-studio/components/executions/executions-view.tsx b/studio/frontend/src/features/recipe-studio/components/executions/executions-view.tsx
index f442d8f98e..cfc0e61338 100644
--- a/studio/frontend/src/features/recipe-studio/components/executions/executions-view.tsx
+++ b/studio/frontend/src/features/recipe-studio/components/executions/executions-view.tsx
@@ -32,10 +32,9 @@ import {
   formatCellValue,
   formatDuration,
   formatPercent,
-  hasExpandableTextCell,
+  isExpandableCellValue,
   parseAnalysisColumns,
   parseModelUsageRows,
-  truncateCellValue,
 } from "./executions-view-helpers";
 
 type ExecutionsViewProps = {
@@ -63,9 +62,6 @@ export function ExecutionsView({
   const [hiddenDatasetColumnsByExecution, setHiddenDatasetColumnsByExecution] = useState<
     Record<string, string[]>
   >({});
-  const [expandedDatasetRowsByExecution, setExpandedDatasetRowsByExecution] = useState<
-    Record<string, Record<string, boolean>>
-  >({});
   const [previewDatasetPageByExecution, setPreviewDatasetPageByExecution] = useState<
     Record<string, number>
   >({});
@@ -91,13 +87,6 @@ export function ExecutionsView({
     }
     return hiddenDatasetColumnsByExecution[selectedExecutionIdSafe] ?? [];
   }, [hiddenDatasetColumnsByExecution, selectedExecutionIdSafe]);
-  const expandedDatasetRows = useMemo(() => {
-    if (!selectedExecutionIdSafe) {
-      return {};
-    }
-    return expandedDatasetRowsByExecution[selectedExecutionIdSafe] ?? {};
-  }, [expandedDatasetRowsByExecution, selectedExecutionIdSafe]);
-
   const datasetColumnNames = useMemo(() => {
     if (!selectedExecution) {
       return [];
@@ -119,6 +108,36 @@ export function ExecutionsView({
     [datasetColumnNames, hiddenDatasetColumns],
   );
 
+  // Columns where at least one row has text long enough that it would wrap at
+  // the default narrow width. We give those columns a wider min-width so the
+  // text is readable without clicking anything. The table's wrapper already
+  // scrolls horizontally, so a few wide columns just add a horizontal
+  // scrollbar instead of squeezing everything into the viewport.
+  const wideColumns = useMemo(() => {
+    const result = new Set<string>();
+    if (!selectedExecution) {
+      return result;
+    }
+    for (const row of selectedExecution.dataset) {
+      for (const name of visibleDatasetColumnNames) {
+        if (result.has(name)) {
+          continue;
+        }
+        const raw = row[name];
+        if (resolveImagePreview(raw)) {
+          continue;
+        }
+        if (isExpandableCellValue(formatCellValue(raw))) {
+          result.add(name);
+        }
+      }
+      if (result.size === visibleDatasetColumnNames.length) {
+        break;
+      }
+    }
+    return result;
+  }, [selectedExecution, visibleDatasetColumnNames]);
+
   const tableColumns = useMemo<ColumnDef<Record<string, unknown>>[]>(() => {
     if (!selectedExecution) {
       return [];
@@ -126,12 +145,12 @@ export function ExecutionsView({
     return visibleDatasetColumnNames.map((name) => ({
       accessorKey: name,
       header: name,
-      cell: ({ getValue, row }) => {
+      cell: ({ getValue }) => {
         const rawValue = getValue();
         const imagePreview = resolveImagePreview(rawValue);
         if (imagePreview?.kind === "ready") {
           return (
-            <div className="max-w-[32rem]">
+            <div>
               <img
                 src={imagePreview.src}
                 alt={`${name} preview`}
@@ -143,31 +162,21 @@ export function ExecutionsView({
         }
         if (imagePreview?.kind === "too_large") {
           return (
-            <div className="max-w-[32rem]">
-              <p className="text-xs text-muted-foreground">
-                Image too large to preview
-              </p>
-            </div>
+            <p className="text-xs text-muted-foreground">
+              Image too large to preview
+            </p>
           );
         }
         const value = formatCellValue(rawValue);
-        const rowExpanded = Boolean(expandedDatasetRows[row.id]);
-        const rowHasExpandableCell = hasExpandableTextCell(
-          row.original,
-          visibleDatasetColumnNames,
-        );
-        const showTruncated = rowHasExpandableCell && !rowExpanded;
-
+        const isWide = wideColumns.has(name);
         return (
-          <div className="max-w-[32rem]">
-            <p className="whitespace-pre-wrap break-all">
-              {showTruncated ? truncateCellValue(value) : value}
-            </p>
+          <div className={cn(isWide ? "min-w-[48rem]" : "min-w-[12rem]")}>
+            <p className="whitespace-pre-wrap break-all">{value}</p>
           </div>
         );
       },
     }));
-  }, [expandedDatasetRows, selectedExecution, visibleDatasetColumnNames]);
+  }, [selectedExecution, visibleDatasetColumnNames, wideColumns]);
 
   const analysisColumns = useMemo(
     () => parseAnalysisColumns(selectedExecution?.analysis ?? null),
@@ -512,9 +521,7 @@ export function ExecutionsView({
                   totalPages={totalPages}
                   tableColumns={tableColumns}
                   datasetRowsForTable={datasetRowsForTable}
-                  visibleDatasetColumnNames={visibleDatasetColumnNames}
-                  expandedDatasetRows={expandedDatasetRows}
-                  selectedExecutionIdSafe={selectedExecutionIdSafe}
+                  onOpenOverview={() => setDetailTab("overview")}
                   onSetHiddenColumns={(updater) => {
                     const selectedId = selectedExecution.id;
                     setHiddenDatasetColumnsByExecution((current) => {
@@ -547,18 +554,6 @@ export function ExecutionsView({
                     }
                     onLoadDatasetPage(selectedExecution.id, currentDatasetPage + 1);
                   }}
-                  onToggleRowExpanded={(rowId) => {
-                    setExpandedDatasetRowsByExecution((current) => {
-                      const rows = current[selectedExecution.id] ?? {};
-                      return {
-                        ...current,
-                        [selectedExecution.id]: {
-                          ...rows,
-                          [rowId]: !rows[rowId],
-                        },
-                      };
-                    });
-                  }}
                 />
               </TabsContent>
               <TabsContent value="raw">
diff --git a/studio/frontend/src/features/recipe-studio/components/inline/inline-seed.tsx b/studio/frontend/src/features/recipe-studio/components/inline/inline-seed.tsx
index bf022fa028..82ec2f1de8 100644
--- a/studio/frontend/src/features/recipe-studio/components/inline/inline-seed.tsx
+++ b/studio/frontend/src/features/recipe-studio/components/inline/inline-seed.tsx
@@ -1,7 +1,12 @@
 // SPDX-License-Identifier: AGPL-3.0-only
 // Copyright 2026-present the Unsloth AI Inc. team. All rights reserved. See /studio/LICENSE.AGPL-3.0
 
-import { DocumentAttachmentIcon, DocumentCodeIcon, Plant01Icon } from "@hugeicons/core-free-icons";
+import {
+  DocumentAttachmentIcon,
+  DocumentCodeIcon,
+  GithubIcon,
+  Plant01Icon,
+} from "@hugeicons/core-free-icons";
 import { HugeiconsIcon } from "@hugeicons/react";
 import type { ReactElement } from "react";
 import type { SeedConfig } from "../../types";
@@ -13,9 +18,75 @@ type InlineSeedProps = {
   onUpdate: (patch: Partial<SeedConfig>) => void;
 };
 
-export function InlineSeed({ config, onUpdate }: InlineSeedProps): ReactElement {
+const GITHUB_REPO_LIST_SPLIT_RE = /[\n,]+/;
+
+function parseGithubRepos(value: string | undefined): string[] {
+  return (value ?? "")
+    .split(GITHUB_REPO_LIST_SPLIT_RE)
+    .map((repo) => repo.trim())
+    .filter(Boolean);
+}
+
+function hasInvalidGithubRepo(repo: string): boolean {
+  const parts = repo.split("/");
+  return parts.length !== 2 || parts.some((part) => !part.trim());
+}
+
+export function InlineSeed({
+  config,
+  onUpdate,
+}: InlineSeedProps): ReactElement {
   const mode = config.seed_source_type ?? "hf";
 
+  if (mode === "github_repo") {
+    const repos = parseGithubRepos(config.github_repo_slug);
+    const invalidCount = repos.filter(hasInvalidGithubRepo).length;
+    const summary =
+      repos.length === 0
+        ? "Add GitHub repos"
+        : repos.length === 1
+          ? repos[0]
+          : `${repos.length} repositories`;
+    const configuredItems = config.github_item_types;
+    const items =
+      configuredItems && configuredItems.length > 0
+        ? configuredItems
+        : ["issues", "pulls"];
+    const itemsLabel = items.join(" · ");
+    const limit = (config.github_limit ?? "100").trim() || "100";
+    const commentsLabel =
+      config.github_include_comments === false
+        ? "comments off"
+        : `comments ≤ ${config.github_max_comments_per_item ?? "30"}`;
+    const tokenLabel = config.github_token?.trim() ? "PAT set" : "server token";
+    const warning =
+      repos.length === 0
+        ? "No repos configured"
+        : invalidCount > 0
+          ? `${invalidCount} invalid repo${invalidCount === 1 ? "" : "s"}`
+          : null;
+    return (
+      <div
+        className={`corner-squircle flex items-center gap-2 rounded-md border px-2 py-2 ${warning ? "border-amber-500/50 bg-amber-500/10" : "border-border/60 bg-muted/30"}`}
+      >
+        <div className="corner-squircle rounded-md bg-primary/10 p-1.5 text-primary">
+          <HugeiconsIcon icon={GithubIcon} className="size-3.5" />
+        </div>
+        <div className="min-w-0">
+          <p className="truncate text-xs font-medium">{summary}</p>
+          <p className="truncate text-[11px] text-muted-foreground">
+            {warning ??
+              `${itemsLabel} · limit ${limit} · ${commentsLabel} · ${tokenLabel}`}
+          </p>
+        </div>
+        <HugeiconsIcon
+          icon={Plant01Icon}
+          className="ml-auto size-3.5 text-muted-foreground/60"
+        />
+      </div>
+    );
+  }
+
   if (mode === "hf") {
     return (
       <div className="space-y-2">
@@ -62,10 +133,14 @@ export function InlineSeed({ config, onUpdate }: InlineSeedProps): ReactElement
           {fileName || "No file selected"}
         </p>
         <p className="text-[11px] text-muted-foreground">
-          {isLocal ? "Structured file" : "Unstructured document"} · configure in dialog
+          {isLocal ? "Structured file" : "Unstructured document"} · configure in
+          dialog
         </p>
       </div>
-      <HugeiconsIcon icon={Plant01Icon} className="ml-auto size-3.5 text-muted-foreground/60" />
+      <HugeiconsIcon
+        icon={Plant01Icon}
+        className="ml-auto size-3.5 text-muted-foreground/60"
+      />
     </div>
   );
 }
diff --git a/studio/frontend/src/features/recipe-studio/components/recipe-studio-header.tsx b/studio/frontend/src/features/recipe-studio/components/recipe-studio-header.tsx
index acbda2d7d8..385bef2b77 100644
--- a/studio/frontend/src/features/recipe-studio/components/recipe-studio-header.tsx
+++ b/studio/frontend/src/features/recipe-studio/components/recipe-studio-header.tsx
@@ -34,6 +34,7 @@ type RecipeStudioHeaderProps = {
   savedAtLabel: string;
   workflowName: string;
   warnings?: GraphWarning[];
+  supportsEasyMode?: boolean;
   onWorkflowNameChange: (value: string) => void;
   onViewChange: (view: RecipeStudioView) => void;
   onSaveRecipe: () => void;
@@ -51,6 +52,7 @@ export function RecipeStudioHeader({
   savedAtLabel,
   workflowName,
   warnings = [],
+  supportsEasyMode = false,
   onWorkflowNameChange,
   onViewChange,
   onSaveRecipe,
@@ -58,7 +60,7 @@ export function RecipeStudioHeader({
   const [editingWorkflowName, setEditingWorkflowName] = useState(false);
 
   function handleViewValueChange(value: string): void {
-    if (value === "editor" || value === "executions") {
+    if (value === "easy" || value === "editor" || value === "executions") {
       onViewChange(value);
     }
   }
@@ -130,7 +132,12 @@ export function RecipeStudioHeader({
       <div className="justify-self-center">
         <Tabs value={activeView} onValueChange={handleViewValueChange}>
           <TabsList>
-            <TabsTrigger value="editor">Editor</TabsTrigger>
+            {supportsEasyMode && (
+              <TabsTrigger value="easy">Easy</TabsTrigger>
+            )}
+            <TabsTrigger value="editor">
+              {supportsEasyMode ? "Advanced" : "Editor"}
+            </TabsTrigger>
             <TabsTrigger value="executions">Runs</TabsTrigger>
           </TabsList>
         </Tabs>
diff --git a/studio/frontend/src/features/recipe-studio/components/runtime/execution-progress-island.tsx b/studio/frontend/src/features/recipe-studio/components/runtime/execution-progress-island.tsx
index 43f7437848..563280c55f 100644
--- a/studio/frontend/src/features/recipe-studio/components/runtime/execution-progress-island.tsx
+++ b/studio/frontend/src/features/recipe-studio/components/runtime/execution-progress-island.tsx
@@ -49,6 +49,52 @@ function statusLabel(input: {
   return "Run status";
 }
 
+function formatSourceResource(value: string | null | undefined): string {
+  if (value === "pulls") {
+    return "PRs";
+  }
+  return value ?? "source";
+}
+
+function formatGitHubSourceSummary(
+  execution: RecipeExecutionRecord,
+): string | null {
+  const source = execution.source_progress;
+  if (!source || source.source !== "github") {
+    return null;
+  }
+  if (source.status === "rate_limited") {
+    const wait =
+      typeof source.retry_after_sec === "number" && source.retry_after_sec > 0
+        ? ` ~${formatMetricValue(source.retry_after_sec)}s`
+        : "";
+    return `Waiting for GitHub rate limit${wait}. Studio will resume automatically.`;
+  }
+  if (source.status === "retrying") {
+    return source.message ?? "GitHub request failed; retrying automatically.";
+  }
+
+  const parts = [
+    source.repo,
+    source.resource
+      ? `${formatSourceResource(source.resource)}${
+          typeof source.page === "number" ? ` page ${source.page}` : ""
+        }`
+      : null,
+    typeof source.fetched_items === "number"
+      ? `${formatMetricValue(source.fetched_items)} fetched`
+      : null,
+    typeof source.rate_remaining === "number"
+      ? `remaining ${formatMetricValue(source.rate_remaining)}`
+      : null,
+  ].filter(Boolean);
+
+  if (parts.length === 0) {
+    return source.message ?? "Crawling GitHub source";
+  }
+  return parts.join(" · ");
+}
+
 export function ExecutionProgressIsland({
   execution,
   currentColumnIcon,
@@ -58,14 +104,24 @@ export function ExecutionProgressIsland({
 }: ExecutionProgressIslandProps): ReactElement {
   const complete = execution.status === "completed";
   const inProgress = isExecutionInProgress(execution.status);
-  const progressPercent = execution.progress?.percent ?? (complete ? 100 : 0);
+  const sourceSummary = formatGitHubSourceSummary(execution);
+  const showSourceProgress = Boolean(
+    sourceSummary &&
+      inProgress &&
+      (execution.stage === "source" ||
+        execution.source_progress?.status === "rate_limited" ||
+        execution.source_progress?.status === "retrying"),
+  );
+  const sourcePercent = showSourceProgress ? execution.source_progress?.percent : null;
+  const progressPercent = sourcePercent ?? execution.progress?.percent ?? (complete ? 100 : 0);
   const hasProgressSignal = Boolean(
-    execution.progress &&
+    (execution.progress &&
       (typeof execution.progress.done === "number" ||
         typeof execution.progress.total === "number" ||
         typeof execution.progress.percent === "number" ||
         typeof execution.progress.rate === "number" ||
-        typeof execution.progress.eta_sec === "number"),
+        typeof execution.progress.eta_sec === "number")) ||
+      typeof sourcePercent === "number",
   );
   const showLoadingSpinner = inProgress && !hasProgressSignal;
   const batchTotal = execution.batch?.total ?? null;
@@ -136,18 +192,30 @@ export function ExecutionProgressIsland({
               ETA: {formatEta(execution.progress?.eta_sec)}
             </p>
           </div>
-          <div className="mt-1 flex items-center gap-1.5 px-3 text-[11px] text-muted-foreground">
-            <HugeiconsIcon
-              icon={currentColumnIcon}
-              className="size-3.5 shrink-0"
-            />
-            <p
-              className="truncate"
-              title={execution.current_column ?? "--"}
-            >
-              Column: {execution.current_column ?? "--"}
-            </p>
-          </div>
+          {showSourceProgress ? (
+            <div className="mt-1 flex items-center gap-1.5 px-3 text-[11px] text-muted-foreground">
+              <HugeiconsIcon
+                icon={Flag02Icon}
+                className="size-3.5 shrink-0 text-amber-700 dark:text-amber-300"
+              />
+              <p className="truncate" title={sourceSummary ?? undefined}>
+                GitHub source: {sourceSummary}
+              </p>
+            </div>
+          ) : (
+            <div className="mt-1 flex items-center gap-1.5 px-3 text-[11px] text-muted-foreground">
+              <HugeiconsIcon
+                icon={currentColumnIcon}
+                className="size-3.5 shrink-0"
+              />
+              <p
+                className="truncate"
+                title={execution.current_column ?? "--"}
+              >
+                Column: {execution.current_column ?? "--"}
+              </p>
+            </div>
+          )}
           {showBatch && (
             <div
               className="mt-1 truncate px-3 text-[11px] text-muted-foreground"
diff --git a/studio/frontend/src/features/recipe-studio/dialogs/preview-dialog.tsx b/studio/frontend/src/features/recipe-studio/dialogs/preview-dialog.tsx
index fbca622d35..7bcf47a974 100644
--- a/studio/frontend/src/features/recipe-studio/dialogs/preview-dialog.tsx
+++ b/studio/frontend/src/features/recipe-studio/dialogs/preview-dialog.tsx
@@ -253,8 +253,13 @@ function ValidationResultPanel({
           ))}
         </div>
       )}
-      {!validateResult.valid && validateResult.rawDetail && (
-        <p className="break-words text-xs text-destructive">
+      {validateResult.rawDetail && (
+        <p
+          className={cn(
+            "break-words text-xs",
+            validateResult.valid ? "text-muted-foreground" : "text-destructive",
+          )}
+        >
           {validateResult.rawDetail}
         </p>
       )}
diff --git a/studio/frontend/src/features/recipe-studio/dialogs/seed/seed-dialog.tsx b/studio/frontend/src/features/recipe-studio/dialogs/seed/seed-dialog.tsx
index 868974acb5..54eae08f7c 100644
--- a/studio/frontend/src/features/recipe-studio/dialogs/seed/seed-dialog.tsx
+++ b/studio/frontend/src/features/recipe-studio/dialogs/seed/seed-dialog.tsx
@@ -31,18 +31,27 @@ import {
   TableHeader,
   TableRow,
 } from "@/components/ui/table";
+import { Tabs, TabsContent, TabsList, TabsTrigger } from "@/components/ui/tabs";
 import {
-  Tabs,
-  TabsContent,
-  TabsList,
-  TabsTrigger,
-} from "@/components/ui/tabs";
-import { type ReactElement, useCallback, useEffect, useMemo, useRef, useState } from "react";
+  type KeyboardEvent,
+  type ReactElement,
+  useCallback,
+  useEffect,
+  useId,
+  useMemo,
+  useRef,
+  useState,
+} from "react";
 import { cn } from "@/lib/utils";
 import { UnstructuredDropZone, type FileEntry } from "./unstructured-drop-zone";
-import { inspectSeedDataset, inspectSeedUpload } from "../../api";
+import {
+  getGithubEnvTokenStatus,
+  inspectSeedDataset,
+  inspectSeedUpload,
+} from "../../api";
 import { resolveImagePreview } from "../../utils/image-preview";
 import type {
+  GithubItemType,
   SeedConfig,
   SeedSamplingStrategy,
   SeedSelectionType,
@@ -51,10 +60,11 @@ import { CollapsibleSectionTriggerButton } from "../shared/collapsible-section-t
 import { HfDatasetCombobox } from "../../components/shared/hf-dataset-combobox";
 import { FieldLabel } from "../shared/field-label";
 
-const SAMPLING_OPTIONS: Array<{ value: SeedSamplingStrategy; label: string }> = [
-  { value: "ordered", label: "Ordered" },
-  { value: "shuffle", label: "Shuffle" },
-];
+const SAMPLING_OPTIONS: Array<{ value: SeedSamplingStrategy; label: string }> =
+  [
+    { value: "ordered", label: "Ordered" },
+    { value: "shuffle", label: "Shuffle" },
+  ];
 
 const SELECTION_OPTIONS: Array<{ value: SeedSelectionType; label: string }> = [
   { value: "none", label: "None" },
@@ -68,6 +78,14 @@ const DEFAULT_CHUNK_SIZE = 1200;
 const DEFAULT_CHUNK_OVERLAP = 200;
 const MAX_CHUNK_SIZE = 20000;
 const PREVIEW_TRUNCATE_AT = 320;
+const GITHUB_TRAILING_PUNCTUATION_RE = /[),.;]+$/g;
+const GITHUB_SSH_PREFIX_RE = /^git@github\.com:/i;
+const URL_PROTOCOL_RE = /^https?:\/\//i;
+const WWW_PREFIX_RE = /^www\./i;
+const URL_QUERY_OR_HASH_RE = /[?#]/;
+const GIT_SUFFIX_RE = /\.git$/i;
+const GITHUB_REPO_INPUT_SPLIT_RE = /[\s,]+/;
+const GITHUB_REPO_PART_RE = /^[A-Za-z0-9_.-]+$/;
 
 type SeedDialogProps = {
   config: SeedConfig;
@@ -75,6 +93,367 @@ type SeedDialogProps = {
   open: boolean;
 };
 
+function normalizeGithubRepoInput(value: string): string {
+  let raw = value.trim().replace(GITHUB_TRAILING_PUNCTUATION_RE, "");
+  if (!raw) {
+    return "";
+  }
+  raw = raw.replace(GITHUB_SSH_PREFIX_RE, "https://github.com/");
+  raw = raw.replace(URL_PROTOCOL_RE, "");
+  raw = raw.replace(WWW_PREFIX_RE, "");
+  if (raw.toLowerCase().startsWith("github.com/")) {
+    raw = raw.slice("github.com/".length);
+  }
+  raw = raw.split(URL_QUERY_OR_HASH_RE)[0] ?? "";
+  raw = raw.replace(GIT_SUFFIX_RE, "");
+  const parts = raw.split("/").filter(Boolean);
+  if (parts.length >= 2) {
+    return `${parts[0]}/${parts[1]}`;
+  }
+  return raw;
+}
+
+function splitGithubRepoInput(value: string): string[] {
+  return value
+    .split(GITHUB_REPO_INPUT_SPLIT_RE)
+    .map(normalizeGithubRepoInput)
+    .filter(Boolean);
+}
+
+function getGithubRepoError(repo: string): string | null {
+  const parts = repo.split("/");
+  if (parts.length !== 2 || parts.some((part) => !part.trim())) {
+    return "Use owner/name.";
+  }
+  if (parts.some((part) => !GITHUB_REPO_PART_RE.test(part))) {
+    return "Use only GitHub repo characters.";
+  }
+  return null;
+}
+
+function githubLimitString(value: string | undefined): string {
+  return (value ?? "100").trim();
+}
+
+export function GithubRepoSeedForm({
+  config,
+  onUpdate,
+}: {
+  config: SeedConfig;
+  onUpdate: (patch: Partial<SeedConfig>) => void;
+}): ReactElement {
+  const [repoDraft, setRepoDraft] = useState("");
+  const [serverHasEnvToken, setServerHasEnvToken] = useState<boolean | null>(
+    null,
+  );
+  const repoInputId = useId();
+  const repoHelpId = useId();
+  const repoErrorId = useId();
+  const tokenId = useId();
+  const tokenHelpId = useId();
+  const limitId = useId();
+  const limitHelpId = useId();
+  const commentsId = useId();
+  const includeCommentsId = useId();
+  const commentsHelpId = useId();
+  const repos = useMemo(
+    () => splitGithubRepoInput(config.github_repo_slug ?? ""),
+    [config.github_repo_slug],
+  );
+  const repoErrors = repos
+    .map((repo, index) => ({ repo, index, error: getGithubRepoError(repo) }))
+    .filter((item) => item.error);
+  const hasRepoErrors = repoErrors.length > 0;
+  const configuredItemTypes = config.github_item_types;
+  const itemTypes: GithubItemType[] =
+    configuredItemTypes && configuredItemTypes.length > 0
+      ? configuredItemTypes
+      : ["issues", "pulls"];
+  const limitNum = Number.parseInt(githubLimitString(config.github_limit), 10);
+  const boundedLimit = Number.isFinite(limitNum)
+    ? Math.min(5000, Math.max(1, limitNum))
+    : 100;
+  const estimatedItems = repos.length * itemTypes.length * boundedLimit;
+  const includeComments = config.github_include_comments ?? true;
+  const hasToken = Boolean(config.github_token?.trim());
+  const usingEnvToken = !hasToken && serverHasEnvToken === true;
+
+  useEffect(() => {
+    let cancelled = false;
+    void getGithubEnvTokenStatus()
+      .then((status) => {
+        if (!cancelled) setServerHasEnvToken(status.has_token);
+      })
+      .catch(() => {
+        if (!cancelled) setServerHasEnvToken(false);
+      });
+    return () => {
+      cancelled = true;
+    };
+  }, []);
+
+  function updateRepos(nextRepos: string[]): void {
+    const seen = new Set<string>();
+    const deduped = nextRepos.filter((repo) => {
+      const key = repo.toLowerCase();
+      if (seen.has(key)) {
+        return false;
+      }
+      seen.add(key);
+      return true;
+    });
+    onUpdate({ github_repo_slug: deduped.join("\n") });
+  }
+
+  function addRepos(raw: string): void {
+    const next = splitGithubRepoInput(raw);
+    if (next.length === 0) {
+      return;
+    }
+    updateRepos([...repos, ...next]);
+    setRepoDraft("");
+  }
+
+  function removeRepo(index: number): void {
+    updateRepos(repos.filter((_, i) => i !== index));
+  }
+
+  function handleRepoKeyDown(event: KeyboardEvent<HTMLInputElement>): void {
+    if (["Enter", ",", " "].includes(event.key)) {
+      event.preventDefault();
+      addRepos(repoDraft);
+    } else if (event.key === "Backspace" && !repoDraft && repos.length > 0) {
+      event.preventDefault();
+      removeRepo(repos.length - 1);
+    }
+  }
+
+  return (
+    <div className="space-y-4">
+      <div className="grid gap-1.5">
+        <FieldLabel
+          label="GitHub repositories"
+          htmlFor={repoInputId}
+          hint="Paste owner/name entries or GitHub URLs. Newlines, commas, and spaces are accepted."
+        />
+        <div
+          className={cn(
+            "nodrag flex min-h-11 flex-wrap items-center gap-1.5 rounded-md border border-border/60 bg-background px-2 py-1.5",
+            hasRepoErrors && "border-red-500/70",
+          )}
+        >
+          {repos.map((repo, index) => {
+            const error = getGithubRepoError(repo);
+            return (
+              <span
+                key={`${repo}-${index}`}
+                className={cn(
+                  "inline-flex items-center gap-1 rounded-full border px-2 py-1 text-xs font-mono",
+                  error
+                    ? "border-red-500/60 bg-red-500/10 text-red-700 dark:text-red-300"
+                    : "border-border/60 bg-muted/60",
+                )}
+              >
+                {repo}
+                <button
+                  type="button"
+                  className="rounded-full px-1 text-muted-foreground hover:bg-background hover:text-foreground"
+                  onClick={() => removeRepo(index)}
+                  aria-label={`Remove ${repo}`}
+                >
+                  ×
+                </button>
+              </span>
+            );
+          })}
+          <input
+            id={repoInputId}
+            className="min-w-36 flex-1 bg-transparent text-xs outline-none placeholder:text-muted-foreground"
+            value={repoDraft}
+            onChange={(event) => setRepoDraft(event.target.value)}
+            onKeyDown={handleRepoKeyDown}
+            onBlur={() => addRepos(repoDraft)}
+            onPaste={(event) => {
+              const pasted = event.clipboardData.getData("text");
+              if (pasted) {
+                event.preventDefault();
+                addRepos(pasted);
+              }
+            }}
+            placeholder={
+              repos.length === 0
+                ? "unslothai/unsloth or GitHub URL"
+                : "Add repo…"
+            }
+            aria-invalid={hasRepoErrors}
+            aria-describedby={`${repoHelpId}${hasRepoErrors ? ` ${repoErrorId}` : ""}`}
+          />
+        </div>
+        <p id={repoHelpId} className="text-xs text-muted-foreground">
+          Stored as one <code>owner/name</code> repo per line in the recipe.
+        </p>
+        {hasRepoErrors && (
+          <ul id={repoErrorId} className="space-y-0.5 text-xs text-red-600">
+            {repoErrors.map((item) => (
+              <li key={`${item.repo}-${item.index}`}>
+                Row {item.index + 1}: {item.error}
+              </li>
+            ))}
+          </ul>
+        )}
+      </div>
+
+      <div className="grid gap-1.5">
+        <div className="flex items-start justify-between gap-2">
+          <FieldLabel
+            label="GitHub token (optional)"
+            htmlFor={tokenId}
+            hint="Prefer the server GH_TOKEN / GITHUB_TOKEN env var. Use public_repo for public repos or repo for private repos."
+          />
+          {usingEnvToken && (
+            <span className="shrink-0 rounded-md border border-emerald-500/40 bg-emerald-500/10 px-1.5 py-0.5 text-[10px] font-medium text-emerald-700 dark:text-emerald-300">
+              Using server env var
+            </span>
+          )}
+        </div>
+        <Input
+          id={tokenId}
+          type="password"
+          className="nodrag"
+          value={config.github_token ?? ""}
+          onChange={(e) => onUpdate({ github_token: e.target.value })}
+          placeholder={
+            usingEnvToken
+              ? "Using server GH_TOKEN / GITHUB_TOKEN"
+              : "Leave blank to use server GH_TOKEN"
+          }
+          aria-describedby={tokenHelpId}
+        />
+        <p id={tokenHelpId} className="text-xs text-muted-foreground">
+          {usingEnvToken
+            ? "Studio detected a server env token, so saved/shared recipes can leave this blank."
+            : "Blank is safest for saved/shared recipes because Studio will read the server environment at run time."}
+        </p>
+        {hasToken && (
+          <p className="rounded-md bg-amber-500/10 px-2 py-1.5 text-xs text-amber-700 dark:text-amber-300">
+            Personal access tokens are sensitive. Prefer server env vars when
+            possible, and avoid sharing recipes that contain a PAT.
+          </p>
+        )}
+      </div>
+
+      <fieldset className="space-y-3">
+        <legend className="text-xs font-semibold uppercase text-muted-foreground">
+          Fetch scope
+        </legend>
+        <div className="grid gap-1.5">
+          <FieldLabel
+            label="Items per repo"
+            htmlFor={limitId}
+            hint="How many issues/PRs/commits to fetch from each repo (1-5000)."
+          />
+          <Input
+            id={limitId}
+            type="number"
+            className="nodrag"
+            min={1}
+            max={5000}
+            value={githubLimitString(config.github_limit)}
+            onChange={(e) => onUpdate({ github_limit: e.target.value })}
+            placeholder="100"
+            aria-describedby={limitHelpId}
+          />
+          <p id={limitHelpId} className="text-xs text-muted-foreground">
+            Estimate before comments: up to {estimatedItems.toLocaleString()}{" "}
+            rows ({repos.length || 0} repos × {itemTypes.length} item types ×{" "}
+            {boundedLimit} limit). Commit crawling uses each repo's default
+            branch.
+          </p>
+        </div>
+
+        <div className="grid gap-1.5">
+          <span className="text-xs font-semibold uppercase text-muted-foreground">
+            Item types
+          </span>
+          <div className="flex flex-wrap gap-3 text-xs">
+            {(["issues", "pulls", "commits"] as const).map((kind) => {
+              const checked = itemTypes.includes(kind);
+              const itemTypeId = `${repoInputId}-${kind}`;
+              return (
+                <label
+                  key={kind}
+                  htmlFor={itemTypeId}
+                  className="flex cursor-pointer items-center gap-1.5"
+                >
+                  <Checkbox
+                    id={itemTypeId}
+                    checked={checked}
+                    onCheckedChange={(v) => {
+                      const next =
+                        v === true
+                          ? Array.from(new Set([...itemTypes, kind]))
+                          : itemTypes.filter((k) => k !== kind);
+                      onUpdate({
+                        github_item_types: next.length > 0 ? next : ["issues"],
+                      });
+                    }}
+                  />
+                  <span>{kind}</span>
+                </label>
+              );
+            })}
+          </div>
+        </div>
+
+        <div className="grid gap-2">
+          <label
+            htmlFor={includeCommentsId}
+            className="flex cursor-pointer items-center gap-1.5 text-xs"
+          >
+            <Checkbox
+              id={includeCommentsId}
+              checked={includeComments}
+              onCheckedChange={(v) =>
+                onUpdate({ github_include_comments: v === true })
+              }
+            />
+            <span>Include issue/PR comments</span>
+          </label>
+          <div className="grid gap-1.5">
+            <FieldLabel
+              label="Max comments / item"
+              htmlFor={commentsId}
+              hint="Comments are concatenated into the comments column for issues and PRs."
+            />
+            <Input
+              id={commentsId}
+              type="number"
+              className="nodrag"
+              min={0}
+              max={200}
+              disabled={!includeComments}
+              value={config.github_max_comments_per_item ?? "30"}
+              onChange={(e) =>
+                onUpdate({ github_max_comments_per_item: e.target.value })
+              }
+              aria-describedby={commentsHelpId}
+            />
+            <p id={commentsHelpId} className="text-xs text-muted-foreground">
+              Comments increase GraphQL cost and can make Check/Run look quiet
+              while GitHub pages and rate-limit waits stream in logs.
+            </p>
+          </div>
+        </div>
+      </fieldset>
+
+      <p className="text-xs text-muted-foreground">
+        Backed by Studio's built-in <code>github_repo</code> seed reader. Large
+        repos can take minutes, so start with small limits for previews.
+      </p>
+    </div>
+  );
+}
+
 function getErrorMessage(error: unknown, fallback: string): string {
   if (error instanceof Error && error.message) {
     return error.message;
@@ -85,7 +464,8 @@ function getErrorMessage(error: unknown, fallback: string): string {
 function stringifyCell(value: unknown): string {
   if (value === null || value === undefined) return "";
   if (typeof value === "string") return value;
-  if (typeof value === "number" || typeof value === "boolean") return String(value);
+  if (typeof value === "number" || typeof value === "boolean")
+    return String(value);
   try {
     return JSON.stringify(value);
   } catch {
@@ -111,7 +491,8 @@ function getPreviewEmptyStateCopy(mode: SeedConfig["seed_source_type"]): {
   if (mode === "local") {
     return {
       title: "No preview yet",
-      description: "Upload a CSV, JSON, or JSONL file and click Load to see a sample.",
+      description:
+        "Upload a CSV, JSON, or JSONL file and click Load to see a sample.",
     };
   }
   if (mode === "unstructured") {
@@ -121,9 +502,17 @@ function getPreviewEmptyStateCopy(mode: SeedConfig["seed_source_type"]): {
         "Upload your documents and the preview will appear once processing is done.",
     };
   }
+  if (mode === "github_repo") {
+    return {
+      title: "GitHub data loads during Check or Run",
+      description:
+        "Configure repos, item types, and limits above. GitHub crawling can take minutes on large repos; watch logs for page and rate-limit updates.",
+    };
+  }
   return {
     title: "No preview yet",
-    description: "Select a Hugging Face dataset and click Load to see a sample.",
+    description:
+      "Select a Hugging Face dataset and click Load to see a sample.",
   };
 }
 
@@ -175,24 +564,32 @@ async function fileToBase64Payload(file: File): Promise<string> {
   });
 }
 
-export function SeedDialog({ config, onUpdate, open }: SeedDialogProps): ReactElement {
+export function SeedDialog({
+  config,
+  onUpdate,
+  open,
+}: SeedDialogProps): ReactElement {
   const [inspectError, setInspectError] = useState<string | null>(null);
   const [isInspecting, setIsInspecting] = useState(false);
   const advancedOpen = config.advancedOpen === true;
   const [previewRows, setPreviewRows] = useState<Record<string, unknown>[]>([]);
-  const [expandedPreviewRows, setExpandedPreviewRows] = useState<Record<number, boolean>>({});
+  const [expandedPreviewRows, setExpandedPreviewRows] = useState<
+    Record<number, boolean>
+  >({});
   const [localFile, setLocalFile] = useState<File | null>(null);
-  const [unstructuredFiles, setUnstructuredFiles] = useState<FileEntry[]>(() => {
-    if (config.unstructured_file_ids?.length) {
-      return config.unstructured_file_ids.map((id, i) => ({
-        id,
-        name: config.unstructured_file_names?.[i] ?? "Unknown",
-        size: config.unstructured_file_sizes?.[i] ?? 0,
-        status: "ok" as const,
-      }));
-    }
-    return [];
-  });
+  const [unstructuredFiles, setUnstructuredFiles] = useState<FileEntry[]>(
+    () => {
+      if (config.unstructured_file_ids?.length) {
+        return config.unstructured_file_ids.map((id, i) => ({
+          id,
+          name: config.unstructured_file_names?.[i] ?? "Unknown",
+          size: config.unstructured_file_sizes?.[i] ?? 0,
+          status: "ok" as const,
+        }));
+      }
+      return [];
+    },
+  );
 
   const mode = config.seed_source_type ?? "hf";
   const previewEmpty = getPreviewEmptyStateCopy(mode);
@@ -241,7 +638,14 @@ export function SeedDialog({ config, onUpdate, open }: SeedDialogProps): ReactEl
         status: "ok" as const,
       })),
     );
-  }, [open, mode, unstructuredFiles.length, config.unstructured_file_ids, config.unstructured_file_names, config.unstructured_file_sizes]);
+  }, [
+    open,
+    mode,
+    unstructuredFiles.length,
+    config.unstructured_file_ids,
+    config.unstructured_file_names,
+    config.unstructured_file_sizes,
+  ]);
 
   const handleUnstructuredFilesChange = useCallback(
     (updater: FileEntry[] | ((prev: FileEntry[]) => FileEntry[])) => {
@@ -291,142 +695,135 @@ export function SeedDialog({ config, onUpdate, open }: SeedDialogProps): ReactEl
     const { chunkSize, chunkOverlap } = resolveChunking(config);
     const fileKey = okFiles.map((f) => `${f.id}|${f.name}`).join(",");
     return `unstructured:${fileKey}|${chunkSize}|${chunkOverlap}`;
-  }, [
-    config,
-    localFile,
-    mode,
-    unstructuredFiles,
-  ]);
+  }, [config, localFile, mode, unstructuredFiles]);
 
-  const loadSeedMetadata = useCallback(async (opts?: { silent?: boolean }): Promise<boolean> => {
-    const loadKey = getCurrentLoadKey();
-    if (!opts?.silent) {
-      setInspectError(null);
-    }
-    setIsInspecting(true);
-    try {
-      if (mode === "hf") {
-        const datasetName = config.hf_repo_id.trim();
-        if (!datasetName) {
-          throw new Error("Dataset repo is required.");
-        }
-        const response = await inspectSeedDataset({
-          dataset_name: datasetName,
-          hf_token: config.hf_token?.trim() || undefined,
-          split: config.hf_split?.trim() || undefined,
-          subset: config.hf_subset?.trim() || undefined,
-          preview_size: 10,
-        });
-        onUpdate({
-          hf_path: response.resolved_path,
-          seed_columns: response.columns,
-          seed_drop_columns: (config.seed_drop_columns ?? []).filter((name) =>
-            response.columns.includes(name),
-          ),
-          seed_preview_rows: response.preview_rows ?? [],
-          hf_split: response.split ?? "",
-          hf_subset: response.subset ?? "",
-          local_file_name: "",
-          unstructured_file_ids: [],
-          unstructured_file_names: [],
-          unstructured_file_sizes: [],
-        });
-        setPreviewRows(response.preview_rows ?? []);
-        setLastLoadedKey(loadKey);
-        return true;
-      }
-
-      if (mode === "local") {
-        if (!localFile) {
-          throw new Error("Select a local CSV/JSON/JSONL file first.");
-        }
-        if (localFile.size > MAX_UPLOAD_BYTES) {
-          throw new Error("File too large (max 50MB).");
-        }
-        const payload = await fileToBase64Payload(localFile);
-        const response = await inspectSeedUpload({
-          filename: localFile.name,
-          content_base64: payload,
-          preview_size: 10,
-        });
-        onUpdate({
-          hf_path: response.resolved_path,
-          seed_columns: response.columns,
-          seed_drop_columns: (config.seed_drop_columns ?? []).filter((name) =>
-            response.columns.includes(name),
-          ),
-          seed_preview_rows: response.preview_rows ?? [],
-          hf_repo_id: "",
-          hf_subset: "",
-          hf_split: "",
-          local_file_name: localFile.name,
-          unstructured_file_ids: [],
-          unstructured_file_names: [],
-          unstructured_file_sizes: [],
-        });
-        setPreviewRows(response.preview_rows ?? []);
-        setLastLoadedKey(loadKey);
-        return true;
-      }
-
-      if (mode === "unstructured") {
-        const fileIds = unstructuredFiles
-          .filter((f) => f.status === "ok")
-          .map((f) => f.id);
-        const fileNames = unstructuredFiles
-          .filter((f) => f.status === "ok")
-          .map((f) => f.name);
-
-        if (fileIds.length === 0) {
-          setInspectError("No files uploaded");
-          return false;
-        }
-
-        const { chunkSize, chunkOverlap } = resolveChunking(config);
-        const response = await inspectSeedUpload({
-          block_id: config.id,
-          file_ids: fileIds,
-          file_names: fileNames,
-          preview_size: 10,
-          seed_source_type: "unstructured",
-          unstructured_chunk_size: chunkSize,
-          unstructured_chunk_overlap: chunkOverlap,
-        });
-
-        onUpdate({
-          hf_path: response.resolved_path,
-          resolved_paths: response.resolved_paths ?? [],
-          seed_columns: response.columns,
-          seed_preview_rows: response.preview_rows ?? [],
-          unstructured_file_ids: fileIds,
-          unstructured_file_names: fileNames,
-          unstructured_file_sizes: unstructuredFiles
-            .filter((f) => f.status === "ok")
-            .map((f) => f.size),
-        });
-        setPreviewRows(response.preview_rows ?? []);
-        setLastLoadedKey(loadKey);
-        return true;
-      }
-
-      return false;
-    } catch (error) {
+  const loadSeedMetadata = useCallback(
+    async (opts?: { silent?: boolean }): Promise<boolean> => {
+      const loadKey = getCurrentLoadKey();
       if (!opts?.silent) {
-        setInspectError(getErrorMessage(error, "Failed to load seed metadata."));
+        setInspectError(null);
       }
-      setPreviewRows([]);
-      return false;
-    } finally {
-      setIsInspecting(false);
-    }
-  }, [
-    config,
-    getCurrentLoadKey,
-    localFile,
-    mode,
-    onUpdate,
-    unstructuredFiles,
-  ]);
+      setIsInspecting(true);
+      try {
+        if (mode === "hf") {
+          const datasetName = config.hf_repo_id.trim();
+          if (!datasetName) {
+            throw new Error("Dataset repo is required.");
+          }
+          const response = await inspectSeedDataset({
+            dataset_name: datasetName,
+            hf_token: config.hf_token?.trim() || undefined,
+            split: config.hf_split?.trim() || undefined,
+            subset: config.hf_subset?.trim() || undefined,
+            preview_size: 10,
+          });
+          onUpdate({
+            hf_path: response.resolved_path,
+            seed_columns: response.columns,
+            seed_drop_columns: (config.seed_drop_columns ?? []).filter((name) =>
+              response.columns.includes(name),
+            ),
+            seed_preview_rows: response.preview_rows ?? [],
+            hf_split: response.split ?? "",
+            hf_subset: response.subset ?? "",
+            local_file_name: "",
+            unstructured_file_ids: [],
+            unstructured_file_names: [],
+            unstructured_file_sizes: [],
+          });
+          setPreviewRows(response.preview_rows ?? []);
+          setLastLoadedKey(loadKey);
+          return true;
+        }
+
+        if (mode === "local") {
+          if (!localFile) {
+            throw new Error("Select a local CSV/JSON/JSONL file first.");
+          }
+          if (localFile.size > MAX_UPLOAD_BYTES) {
+            throw new Error("File too large (max 50MB).");
+          }
+          const payload = await fileToBase64Payload(localFile);
+          const response = await inspectSeedUpload({
+            filename: localFile.name,
+            content_base64: payload,
+            preview_size: 10,
+          });
+          onUpdate({
+            hf_path: response.resolved_path,
+            seed_columns: response.columns,
+            seed_drop_columns: (config.seed_drop_columns ?? []).filter((name) =>
+              response.columns.includes(name),
+            ),
+            seed_preview_rows: response.preview_rows ?? [],
+            hf_repo_id: "",
+            hf_subset: "",
+            hf_split: "",
+            local_file_name: localFile.name,
+            unstructured_file_ids: [],
+            unstructured_file_names: [],
+            unstructured_file_sizes: [],
+          });
+          setPreviewRows(response.preview_rows ?? []);
+          setLastLoadedKey(loadKey);
+          return true;
+        }
+
+        if (mode === "unstructured") {
+          const fileIds = unstructuredFiles
+            .filter((f) => f.status === "ok")
+            .map((f) => f.id);
+          const fileNames = unstructuredFiles
+            .filter((f) => f.status === "ok")
+            .map((f) => f.name);
+
+          if (fileIds.length === 0) {
+            setInspectError("No files uploaded");
+            return false;
+          }
+
+          const { chunkSize, chunkOverlap } = resolveChunking(config);
+          const response = await inspectSeedUpload({
+            block_id: config.id,
+            file_ids: fileIds,
+            file_names: fileNames,
+            preview_size: 10,
+            seed_source_type: "unstructured",
+            unstructured_chunk_size: chunkSize,
+            unstructured_chunk_overlap: chunkOverlap,
+          });
+
+          onUpdate({
+            hf_path: response.resolved_path,
+            resolved_paths: response.resolved_paths ?? [],
+            seed_columns: response.columns,
+            seed_preview_rows: response.preview_rows ?? [],
+            unstructured_file_ids: fileIds,
+            unstructured_file_names: fileNames,
+            unstructured_file_sizes: unstructuredFiles
+              .filter((f) => f.status === "ok")
+              .map((f) => f.size),
+          });
+          setPreviewRows(response.preview_rows ?? []);
+          setLastLoadedKey(loadKey);
+          return true;
+        }
+
+        return false;
+      } catch (error) {
+        if (!opts?.silent) {
+          setInspectError(
+            getErrorMessage(error, "Failed to load seed metadata."),
+          );
+        }
+        setPreviewRows([]);
+        return false;
+      } finally {
+        setIsInspecting(false);
+      }
+    },
+    [config, getCurrentLoadKey, localFile, mode, onUpdate, unstructuredFiles],
+  );
 
   useEffect(() => {
     const wasOpen = wasOpenRef.current;
@@ -463,7 +860,8 @@ export function SeedDialog({ config, onUpdate, open }: SeedDialogProps): ReactEl
     return [];
   }, [config.seed_columns, previewRows]);
   const selectedSeedDropColumns = useMemo(
-    () => (config.seed_drop_columns ?? []).filter((name) => name.trim().length > 0),
+    () =>
+      (config.seed_drop_columns ?? []).filter((name) => name.trim().length > 0),
     [config.seed_drop_columns],
   );
   const selectedSeedDropSet = useMemo(
@@ -540,10 +938,11 @@ export function SeedDialog({ config, onUpdate, open }: SeedDialogProps): ReactEl
                   className="nodrag"
                   placeholder="hf_..."
                   value={config.hf_token ?? ""}
-                  onChange={(event) => onUpdate({ hf_token: event.target.value })}
+                  onChange={(event) =>
+                    onUpdate({ hf_token: event.target.value })
+                  }
                 />
               </div>
-
             </>
           )}
 
@@ -600,9 +999,15 @@ export function SeedDialog({ config, onUpdate, open }: SeedDialogProps): ReactEl
             />
           )}
 
-          {inspectError && <p className="text-xs text-red-600">{inspectError}</p>}
+          {mode === "github_repo" && (
+            <GithubRepoSeedForm config={config} onUpdate={onUpdate} />
+          )}
 
-          {mode !== "unstructured" && (
+          {inspectError && (
+            <p className="text-xs text-red-600">{inspectError}</p>
+          )}
+
+          {mode !== "unstructured" && mode !== "github_repo" && (
             <div className="space-y-2 rounded-xl corner-squircle border border-border/60 p-3">
               <FieldLabel
                 label="Drop specific seed columns"
@@ -626,8 +1031,15 @@ export function SeedDialog({ config, onUpdate, open }: SeedDialogProps): ReactEl
                           onCheckedChange={(value) => {
                             const isChecked = value === true;
                             const next = isChecked
-                              ? Array.from(new Set([...selectedSeedDropColumns, columnName]))
-                              : selectedSeedDropColumns.filter((name) => name !== columnName);
+                              ? Array.from(
+                                  new Set([
+                                    ...selectedSeedDropColumns,
+                                    columnName,
+                                  ]),
+                                )
+                              : selectedSeedDropColumns.filter(
+                                  (name) => name !== columnName,
+                                );
                             onUpdate({ seed_drop_columns: next });
                           }}
                         />
@@ -660,7 +1072,9 @@ export function SeedDialog({ config, onUpdate, open }: SeedDialogProps): ReactEl
                 <Select
                   value={config.sampling_strategy}
                   onValueChange={(value) =>
-                    onUpdate({ sampling_strategy: value as SeedSamplingStrategy })
+                    onUpdate({
+                      sampling_strategy: value as SeedSamplingStrategy,
+                    })
                   }
                 >
                   <SelectTrigger className="nodrag w-full" id={samplingId}>
@@ -713,9 +1127,14 @@ export function SeedDialog({ config, onUpdate, open }: SeedDialogProps): ReactEl
                       id={chunkSizeId}
                       className="nodrag"
                       inputMode="numeric"
-                      value={config.unstructured_chunk_size ?? String(DEFAULT_CHUNK_SIZE)}
+                      value={
+                        config.unstructured_chunk_size ??
+                        String(DEFAULT_CHUNK_SIZE)
+                      }
                       onChange={(event) =>
-                        onUpdate({ unstructured_chunk_size: event.target.value })
+                        onUpdate({
+                          unstructured_chunk_size: event.target.value,
+                        })
                       }
                     />
                   </div>
@@ -734,7 +1153,9 @@ export function SeedDialog({ config, onUpdate, open }: SeedDialogProps): ReactEl
                         String(DEFAULT_CHUNK_OVERLAP)
                       }
                       onChange={(event) =>
-                        onUpdate({ unstructured_chunk_overlap: event.target.value })
+                        onUpdate({
+                          unstructured_chunk_overlap: event.target.value,
+                        })
                       }
                     />
                   </div>
@@ -744,21 +1165,31 @@ export function SeedDialog({ config, onUpdate, open }: SeedDialogProps): ReactEl
               {config.selection_type === "index_range" && (
                 <div className="grid grid-cols-2 gap-3">
                   <div className="grid gap-1.5">
-                    <FieldLabel label="Start" hint="Inclusive start row index for index_range." />
+                    <FieldLabel
+                      label="Start"
+                      hint="Inclusive start row index for index_range."
+                    />
                     <Input
                       className="nodrag"
                       inputMode="numeric"
                       value={config.selection_start ?? ""}
-                      onChange={(event) => onUpdate({ selection_start: event.target.value })}
+                      onChange={(event) =>
+                        onUpdate({ selection_start: event.target.value })
+                      }
                     />
                   </div>
                   <div className="grid gap-1.5">
-                    <FieldLabel label="End" hint="Inclusive end row index for index_range." />
+                    <FieldLabel
+                      label="End"
+                      hint="Inclusive end row index for index_range."
+                    />
                     <Input
                       className="nodrag"
                       inputMode="numeric"
                       value={config.selection_end ?? ""}
-                      onChange={(event) => onUpdate({ selection_end: event.target.value })}
+                      onChange={(event) =>
+                        onUpdate({ selection_end: event.target.value })
+                      }
                     />
                   </div>
                 </div>
@@ -772,17 +1203,24 @@ export function SeedDialog({ config, onUpdate, open }: SeedDialogProps): ReactEl
                       className="nodrag"
                       inputMode="numeric"
                       value={config.selection_index ?? ""}
-                      onChange={(event) => onUpdate({ selection_index: event.target.value })}
+                      onChange={(event) =>
+                        onUpdate({ selection_index: event.target.value })
+                      }
                     />
                   </div>
                   <div className="grid gap-1.5">
-                    <FieldLabel label="Partitions" hint="Total number of partitions." />
+                    <FieldLabel
+                      label="Partitions"
+                      hint="Total number of partitions."
+                    />
                     <Input
                       className="nodrag"
                       inputMode="numeric"
                       value={config.selection_num_partitions ?? ""}
                       onChange={(event) =>
-                        onUpdate({ selection_num_partitions: event.target.value })
+                        onUpdate({
+                          selection_num_partitions: event.target.value,
+                        })
                       }
                     />
                   </div>
@@ -830,7 +1268,8 @@ export function SeedDialog({ config, onUpdate, open }: SeedDialogProps): ReactEl
                       <TableRow
                         key={`row-${rowIdx}`}
                         className={cn(
-                          rowHasExpandableText(row) && "cursor-pointer hover:bg-primary/[0.06]",
+                          rowHasExpandableText(row) &&
+                            "cursor-pointer hover:bg-primary/[0.06]",
                           expandedPreviewRows[rowIdx] && "bg-primary/[0.05]",
                         )}
                         onClick={() => {
@@ -850,7 +1289,9 @@ export function SeedDialog({ config, onUpdate, open }: SeedDialogProps): ReactEl
                             className="max-w-[260px] whitespace-pre-wrap break-words text-xs"
                           >
                             {(() => {
-                              const imagePreview = resolveImagePreview(row[col]);
+                              const imagePreview = resolveImagePreview(
+                                row[col],
+                              );
                               if (imagePreview?.kind === "ready") {
                                 return (
                                   <img
@@ -865,8 +1306,11 @@ export function SeedDialog({ config, onUpdate, open }: SeedDialogProps): ReactEl
                                 return "Image too large to preview";
                               }
                               const value = stringifyCell(row[col]);
-                              const rowHasExpandableCell = rowHasExpandableText(row);
-                              const rowExpanded = Boolean(expandedPreviewRows[rowIdx]);
+                              const rowHasExpandableCell =
+                                rowHasExpandableText(row);
+                              const rowExpanded = Boolean(
+                                expandedPreviewRows[rowIdx],
+                              );
                               return rowHasExpandableCell && !rowExpanded
                                 ? truncatePreviewValue(value)
                                 : value;
diff --git a/studio/frontend/src/features/recipe-studio/easy/github-crawler-easy-view.tsx b/studio/frontend/src/features/recipe-studio/easy/github-crawler-easy-view.tsx
new file mode 100644
index 0000000000..57264cb4f5
--- /dev/null
+++ b/studio/frontend/src/features/recipe-studio/easy/github-crawler-easy-view.tsx
@@ -0,0 +1,191 @@
+// SPDX-License-Identifier: AGPL-3.0-only
+// Copyright 2026-present the Unsloth AI Inc. team. All rights reserved. See /studio/LICENSE.AGPL-3.0
+
+import { GithubIcon, PlayCircleIcon } from "@hugeicons/core-free-icons";
+import { HugeiconsIcon } from "@hugeicons/react";
+import { type ReactElement, useEffect, useMemo, useState } from "react";
+import { Button } from "@/components/ui/button";
+import { Input } from "@/components/ui/input";
+import { FieldLabel } from "../dialogs/shared/field-label";
+import { GithubRepoSeedForm } from "../dialogs/seed/seed-dialog";
+import type { ModelConfig, NodeConfig, SeedConfig } from "../types";
+
+type GithubCrawlerEasyViewProps = {
+  configs: Record<string, NodeConfig>;
+  rows: number;
+  setRows: (rows: number) => void;
+  updateConfig: (id: string, patch: Partial<NodeConfig>) => void;
+  onRun: () => void;
+  runLoading: boolean;
+  runErrors: string[];
+  onSwitchToAdvanced: () => void;
+};
+
+export function GithubCrawlerEasyView({
+  configs,
+  rows,
+  setRows,
+  updateConfig,
+  onRun,
+  runLoading,
+  runErrors,
+  onSwitchToAdvanced,
+}: GithubCrawlerEasyViewProps): ReactElement {
+  const seedConfig = useMemo(
+    () =>
+      Object.values(configs).find((c): c is SeedConfig => c.kind === "seed") ??
+      null,
+    [configs],
+  );
+  const modelConfig = useMemo(
+    () =>
+      Object.values(configs).find(
+        (c): c is ModelConfig => c.kind === "model_config",
+      ) ?? null,
+    [configs],
+  );
+
+  // Local buffer for the Rows input so the user can hold transient invalid
+  // state (empty while backspacing, partial digits, etc.) without the parent
+  // snapping them back to 1 on every keystroke. The canonical ``rows`` value
+  // only advances when the buffer parses to a valid positive integer; on
+  // blur we clamp back to a sane default if the user left it empty.
+  const [rowsText, setRowsText] = useState(String(rows));
+  useEffect(() => {
+    setRowsText(String(rows));
+  }, [rows]);
+
+  const handleSeedUpdate = (patch: Partial<SeedConfig>): void => {
+    if (!seedConfig) return;
+    updateConfig(seedConfig.id, patch);
+  };
+
+  const handleModelChange = (value: string): void => {
+    if (!modelConfig) return;
+    updateConfig(modelConfig.id, { model: value });
+  };
+
+  if (!seedConfig) {
+    return (
+      <div className="flex h-full items-center justify-center px-6">
+        <p className="text-sm text-muted-foreground">
+          This recipe has no seed node. Switch to{" "}
+          <button
+            type="button"
+            className="underline hover:text-foreground"
+            onClick={onSwitchToAdvanced}
+          >
+            Advanced
+          </button>{" "}
+          to configure it.
+        </p>
+      </div>
+    );
+  }
+
+  return (
+    <div className="mx-auto flex h-full w-full max-w-2xl flex-col gap-4 overflow-y-auto px-6 py-6">
+      <div className="flex items-start gap-3">
+        <div
+          className="flex size-10 shrink-0 items-center justify-center rounded-lg corner-squircle border border-border/70 bg-muted/20"
+          aria-hidden={true}
+        >
+          <HugeiconsIcon icon={GithubIcon} className="size-5" />
+        </div>
+        <div className="flex-1 min-w-0">
+          <h2 className="text-base font-semibold">GitHub Crawler</h2>
+          <p className="text-xs text-muted-foreground">
+            Crawl real GitHub issues and PRs and turn each thread into a{" "}
+            <code>{"{User, Assistant}"}</code> training pair.
+            Defaults use the server's <code>GH_TOKEN</code> env var and the
+            bundled local model.
+          </p>
+        </div>
+      </div>
+
+      <GithubRepoSeedForm config={seedConfig} onUpdate={handleSeedUpdate} />
+
+      <section className="space-y-3 border-t border-border/60 pt-4">
+        <h3 className="text-xs font-semibold uppercase text-muted-foreground">
+          Run settings
+        </h3>
+        <div className="grid gap-3 sm:grid-cols-[minmax(0,10rem)_minmax(0,1fr)]">
+          <div className="grid gap-1.5">
+            <FieldLabel
+              label="Rows to generate"
+              hint="How many training pairs the LLM should produce."
+            />
+            <Input
+              type="text"
+              inputMode="numeric"
+              pattern="[0-9]*"
+              className="nodrag"
+              value={rowsText}
+              onChange={(event) => {
+                // Allow empty / partial strings while the user is editing.
+                // type="text" avoids the browser's number spinner and the
+                // related backspace quirks; we still parse + clamp below.
+                const raw = event.target.value.replace(/[^0-9]/g, "");
+                setRowsText(raw);
+                const next = Number.parseInt(raw, 10);
+                if (Number.isFinite(next) && next > 0 && next <= 10000) {
+                  setRows(next);
+                }
+              }}
+              onBlur={() => {
+                const next = Number.parseInt(rowsText, 10);
+                if (!Number.isFinite(next) || next < 1) {
+                  setRows(1);
+                  setRowsText("1");
+                } else if (next > 10000) {
+                  setRows(10000);
+                  setRowsText("10000");
+                } else {
+                  setRowsText(String(next));
+                }
+              }}
+            />
+          </div>
+          <div className="grid gap-1.5">
+            <FieldLabel
+              label="Model"
+              hint="OpenAI-compatible model id. Local GGUFs run on the bundled llama-server."
+            />
+            <Input
+              className="nodrag font-mono text-xs"
+              value={modelConfig?.model ?? ""}
+              onChange={(event) => handleModelChange(event.target.value)}
+              placeholder="unsloth/gemma-4-E2B-it-GGUF"
+              disabled={!modelConfig}
+            />
+          </div>
+        </div>
+      </section>
+
+      {runErrors.length > 0 && (
+        <div className="rounded-lg border border-destructive/40 bg-destructive/5 px-3 py-2 text-xs text-destructive">
+          <p className="font-semibold">Cannot run:</p>
+          <ul className="mt-1 list-disc space-y-0.5 pl-4">
+            {runErrors.slice(0, 4).map((err) => (
+              <li key={err}>{err}</li>
+            ))}
+          </ul>
+        </div>
+      )}
+
+      <div className="flex items-center justify-between gap-3 pt-1">
+        <button
+          type="button"
+          className="text-xs text-muted-foreground underline hover:text-foreground"
+          onClick={onSwitchToAdvanced}
+        >
+          Advanced (drag and drop canvas)
+        </button>
+        <Button type="button" size="lg" onClick={onRun} disabled={runLoading}>
+          <HugeiconsIcon icon={PlayCircleIcon} className="size-4" />
+          {runLoading ? "Running..." : "Run"}
+        </Button>
+      </div>
+    </div>
+  );
+}
diff --git a/studio/frontend/src/features/recipe-studio/execution-types.ts b/studio/frontend/src/features/recipe-studio/execution-types.ts
index c1976ab3f9..92dd88c4f5 100644
--- a/studio/frontend/src/features/recipe-studio/execution-types.ts
+++ b/studio/frontend/src/features/recipe-studio/execution-types.ts
@@ -1,7 +1,7 @@
 // SPDX-License-Identifier: AGPL-3.0-only
 // Copyright 2026-present the Unsloth AI Inc. team. All rights reserved. See /studio/LICENSE.AGPL-3.0
 
-export type RecipeStudioView = "editor" | "executions";
+export type RecipeStudioView = "easy" | "editor" | "executions";
 
 export type RecipeExecutionKind = "preview" | "full";
 
@@ -30,6 +30,28 @@ export type RecipeExecutionBatch = {
   total?: number | null;
 };
 
+export type RecipeSourceProgress = {
+  source?: string | null;
+  status?: string | null;
+  repo?: string | null;
+  resource?: string | null;
+  page?: number | null;
+  // biome-ignore lint/style/useNamingConvention: backend schema
+  page_items?: number | null;
+  // biome-ignore lint/style/useNamingConvention: backend schema
+  fetched_items?: number | null;
+  // biome-ignore lint/style/useNamingConvention: backend schema
+  estimated_total?: number | null;
+  percent?: number | null;
+  // biome-ignore lint/style/useNamingConvention: backend schema
+  rate_remaining?: number | null;
+  // biome-ignore lint/style/useNamingConvention: backend schema
+  retry_after_sec?: number | null;
+  message?: string | null;
+  // biome-ignore lint/style/useNamingConvention: backend schema
+  updated_at?: number | null;
+};
+
 export type RecipeExecutionAnalysis = {
   num_records?: number;
   target_num_records?: number;
@@ -65,6 +87,8 @@ export type RecipeExecutionRecord = {
   column_progress: RecipeExecutionProgress | null;
   batch: RecipeExecutionBatch | null;
   // biome-ignore lint/style/useNamingConvention: backend schema
+  source_progress: RecipeSourceProgress | null;
+  // biome-ignore lint/style/useNamingConvention: backend schema
   model_usage: Record<string, unknown> | null;
   // biome-ignore lint/style/useNamingConvention: backend schema
   lastEventId: number | null;
diff --git a/studio/frontend/src/features/recipe-studio/executions/runtime.ts b/studio/frontend/src/features/recipe-studio/executions/runtime.ts
index bdc3e653fc..e37a1c9929 100644
--- a/studio/frontend/src/features/recipe-studio/executions/runtime.ts
+++ b/studio/frontend/src/features/recipe-studio/executions/runtime.ts
@@ -6,6 +6,7 @@ import type {
   RecipeExecutionBatch,
   RecipeExecutionKind,
   RecipeExecutionRecord,
+  RecipeSourceProgress,
 } from "../execution-types";
 import {
   DATASET_PAGE_SIZE,
@@ -72,6 +73,14 @@ export function toExecutionLogLine(event: JobEvent): string | null {
   return null;
 }
 
+function normalizeSourceProgress(input: unknown): RecipeSourceProgress | null {
+  const raw = normalizeObject(input);
+  if (!raw) {
+    return null;
+  }
+  return raw as RecipeSourceProgress;
+}
+
 export function applyExecutionStatusSnapshot(
   execution: RecipeExecutionRecord,
   status: JobStatusResponse,
@@ -100,6 +109,7 @@ export function applyExecutionStatusSnapshot(
       (normalizeObject(status.column_progress) as RecipeExecutionRecord["column_progress"]) ??
       null,
     batch,
+    source_progress: normalizeSourceProgress(status.source_progress),
     model_usage: normalizeObject(status.model_usage),
     artifact_path: status.artifact_path ?? execution.artifact_path,
     error: status.error ?? null,
@@ -137,6 +147,7 @@ export function createBaseExecutionRecord(input: {
     progress: null,
     column_progress: null,
     batch: null,
+    source_progress: null,
     model_usage: null,
     lastEventId: null,
     artifact_path: null,
diff --git a/studio/frontend/src/features/recipe-studio/hooks/use-recipe-executions.ts b/studio/frontend/src/features/recipe-studio/hooks/use-recipe-executions.ts
index 24d02e0f6c..1040025a5d 100644
--- a/studio/frontend/src/features/recipe-studio/hooks/use-recipe-executions.ts
+++ b/studio/frontend/src/features/recipe-studio/hooks/use-recipe-executions.ts
@@ -3,7 +3,12 @@
 
 import { useCallback, useEffect, useState } from "react";
 import { useShallow } from "zustand/react/shallow";
+import { toast } from "sonner";
 import { toastError } from "@/shared/toast";
+import {
+  getInferenceStatus,
+  loadModel,
+} from "@/features/chat/api/chat-api";
 import {
   cancelRecipeJob,
   createRecipeJob,
@@ -39,10 +44,84 @@ import {
 } from "../stores/recipe-executions";
 import type { RecipePayload, RecipePayloadResult } from "../utils/payload/types";
 
+/**
+ * Auto-load the local model before running a recipe that uses it.
+ *
+ * Looks at payload.recipe.model_providers for any provider with is_local=true,
+ * finds the bound model_configs and asks the backend to load whichever model
+ * the first local-bound model_config points at. Skips when the inference
+ * server already has that exact model active. This removes the "open /chat
+ * first" prerequisite that users kept tripping on.
+ */
+async function ensureLocalModelLoaded(
+  payload: RecipePayload,
+): Promise<string | null> {
+  const providers = Array.isArray(payload.recipe.model_providers)
+    ? (payload.recipe.model_providers as Array<Record<string, unknown>>)
+    : [];
+  const localProviderNames = new Set<string>();
+  for (const p of providers) {
+    if (p.is_local === true && typeof p.name === "string") {
+      localProviderNames.add(p.name);
+    }
+  }
+  if (localProviderNames.size === 0) {
+    return null;
+  }
+  const modelConfigs = Array.isArray(payload.recipe.model_configs)
+    ? (payload.recipe.model_configs as Array<Record<string, unknown>>)
+    : [];
+  const boundConfig = modelConfigs.find(
+    (c) => typeof c.provider === "string" && localProviderNames.has(c.provider),
+  );
+  const target =
+    typeof boundConfig?.model === "string" ? boundConfig.model.trim() : "";
+  if (!target) {
+    return null;
+  }
+
+  try {
+    const status = await getInferenceStatus();
+    if (
+      status.active_model &&
+      status.active_model.toLowerCase() === target.toLowerCase()
+    ) {
+      return null;
+    }
+  } catch {
+    // Fall through to load attempt; the backend will re-error if needed.
+  }
+
+  const toastId = toast.loading(`Loading ${target}…`, {
+    description: "Starting the local inference server for this recipe.",
+  });
+  try {
+    const isGguf = /gguf/i.test(target);
+    await loadModel({
+      model_path: target,
+      hf_token: null,
+      max_seq_length: isGguf ? 0 : 4096,
+      load_in_4bit: true,
+      is_lora: false,
+      gguf_variant: null,
+      trust_remote_code: false,
+      chat_template_override: null,
+      cache_type_kv: null,
+      speculative_type: null,
+    });
+    toast.success(`Loaded ${target}`, { id: toastId, duration: 2000 });
+    return null;
+  } catch (error) {
+    toast.dismiss(toastId);
+    return error instanceof Error ? error.message : String(error);
+  }
+}
+
 type UseRecipeExecutionsParams = {
   recipeId: string;
   currentSignature: string;
   payloadResult: RecipePayloadResult;
+  initialRunRows?: number | null;
   onExecutionStart?: () => void;
   onPreviewSuccess?: () => void;
 };
@@ -101,6 +180,7 @@ export function useRecipeExecutions({
   recipeId,
   currentSignature,
   payloadResult,
+  initialRunRows,
   onExecutionStart,
   onPreviewSuccess,
 }: UseRecipeExecutionsParams): UseRecipeExecutionsResult {
@@ -181,6 +261,19 @@ export function useRecipeExecutions({
 
     resetForRecipe();
 
+    // Seed previewRows from the recipe's original run.rows (read from the
+    // loaded JSON, not the rebuilt payload (the builder hardcodes 5).
+    // Templates ship their own suggested preview size (e.g. GitHub Support
+    // Bot: 10); we honor it so users don't see a surprise 5.
+    if (
+      typeof initialRunRows === "number" &&
+      Number.isFinite(initialRunRows) &&
+      initialRunRows > 0 &&
+      initialRunRows !== 5
+    ) {
+      setPreviewRows(Math.floor(initialRunRows));
+    }
+
     async function hydrate(): Promise<void> {
       try {
         const records = await loadSortedRecipeExecutions(recipeId);
@@ -216,10 +309,12 @@ export function useRecipeExecutions({
       cancelled = true;
     };
   }, [
+    initialRunRows,
     onPreviewSuccess,
     recipeId,
     resetForRecipe,
     setExecutions,
+    setPreviewRows,
     setRunErrors,
     upsertAndPersist,
   ]);
@@ -340,6 +435,20 @@ export function useRecipeExecutions({
         return false;
       }
 
+      // Flip to the Runs pane BEFORE we run ensureLocalModelLoaded + validate.
+      // Validation re-crawls the seed (multiple seconds for the github_repo
+      // reader) and the user otherwise stares at a "Running..." button with
+      // nothing else changing. runExecution() later no-ops this callback if
+      // the view has already been flipped, so we fire it once here.
+      onExecutionStart?.();
+
+      const localLoadError = await ensureLocalModelLoaded(payload);
+      if (localLoadError) {
+        setRunErrors([localLoadError]);
+        toastError("Local model failed to load", localLoadError);
+        return false;
+      }
+
       const normalizedRows = sanitizeExecutionRows(rows, kind);
       const executionPayload = buildExecutionPayload({
         payload,
@@ -374,7 +483,13 @@ export function useRecipeExecutions({
         runName,
       });
     },
-    [readExecutablePayload, runExecution, runSettings, setRunErrors],
+    [
+      onExecutionStart,
+      readExecutablePayload,
+      runExecution,
+      runSettings,
+      setRunErrors,
+    ],
   );
 
   const runPreview = useCallback(async (): Promise<boolean> => {
diff --git a/studio/frontend/src/features/recipe-studio/hooks/use-recipe-studio-actions.ts b/studio/frontend/src/features/recipe-studio/hooks/use-recipe-studio-actions.ts
index f6f7390cf7..79c75c1885 100644
--- a/studio/frontend/src/features/recipe-studio/hooks/use-recipe-studio-actions.ts
+++ b/studio/frontend/src/features/recipe-studio/hooks/use-recipe-studio-actions.ts
@@ -112,6 +112,10 @@ export function useRecipeStudioActions({
     recipeId,
     currentSignature: persistence.currentSignature,
     payloadResult,
+    initialRunRows:
+      typeof initialPayload?.run?.rows === "number"
+        ? initialPayload.run.rows
+        : null,
     onExecutionStart,
     onPreviewSuccess,
   });
diff --git a/studio/frontend/src/features/recipe-studio/recipe-studio-page.tsx b/studio/frontend/src/features/recipe-studio/recipe-studio-page.tsx
index 586d5d8ad8..5dbc004ffd 100644
--- a/studio/frontend/src/features/recipe-studio/recipe-studio-page.tsx
+++ b/studio/frontend/src/features/recipe-studio/recipe-studio-page.tsx
@@ -47,6 +47,7 @@ import { ConfigDialog } from "./dialogs/config-dialog";
 import { ImportDialog } from "./dialogs/import-dialog";
 import { RunDialog } from "./dialogs/preview-dialog";
 import { ProcessorsDialog } from "./dialogs/processors-dialog";
+import { GithubCrawlerEasyView } from "./easy/github-crawler-easy-view";
 import type {
   RecipeExecutionRecord,
   RecipeStudioView,
@@ -201,7 +202,40 @@ export function RecipeStudioPage({
     null,
   );
   const flowContainerRef = useRef<HTMLDivElement | null>(null);
-  const [activeView, setActiveView] = useState<RecipeStudioView>("editor");
+  const supportsEasyMode =
+    initialPayload?.ui?.seed_source_type === "github_repo" ||
+    (initialPayload?.recipe?.seed_config as { source?: { seed_type?: string } } | undefined)
+      ?.source?.seed_type === "github_repo";
+  const viewModeStorageKey = `recipe-studio:view-mode:${recipeId}`;
+  const [activeView, setActiveViewState] = useState<RecipeStudioView>(() => {
+    if (typeof window !== "undefined") {
+      const stored = window.localStorage.getItem(viewModeStorageKey);
+      if (stored === "easy" && supportsEasyMode) return "easy";
+      if (stored === "editor" || stored === "executions") return stored;
+    }
+    return supportsEasyMode ? "easy" : "editor";
+  });
+  const setActiveView = useCallback(
+    (next: RecipeStudioView | ((prev: RecipeStudioView) => RecipeStudioView)) => {
+      setActiveViewState((prev) => {
+        const resolved = typeof next === "function" ? next(prev) : next;
+        if (typeof window !== "undefined") {
+          window.localStorage.setItem(viewModeStorageKey, resolved);
+        }
+        return resolved;
+      });
+    },
+    [viewModeStorageKey],
+  );
+  // Easy mode has no canvas overlay/progress island, so once a run starts the
+  // user sees the Run button stuck on "Running..." with nothing else changing.
+  // Flip to the Runs pane so they land where progress is actually rendered.
+  // Advanced (editor) keeps its island and stays put.
+  const handleExecutionStart = useCallback(() => {
+    setActiveView((currentView) =>
+      currentView === "easy" ? "executions" : currentView,
+    );
+  }, [setActiveView]);
   const [processorsOpen, setProcessorsOpen] = useState(false);
   const [interactive, setInteractive] = useState(true);
   const [runtimeIslandMinimized, setRuntimeIslandMinimized] = useState(false);
@@ -326,6 +360,8 @@ export function RecipeStudioPage({
     validateResult,
     cancelExecution,
     loadExecutionDatasetPage,
+    runPreview,
+    runFull,
     copyRecipe,
     importRecipe,
   } = useRecipeStudioActions({
@@ -338,6 +374,7 @@ export function RecipeStudioPage({
     resetRecipe,
     loadRecipe,
     getCurrentPayloadFromStore,
+    onExecutionStart: handleExecutionStart,
   });
   const {
     activeExecution,
@@ -359,6 +396,19 @@ export function RecipeStudioPage({
   const runBusy = previewLoading || fullLoading || executionLocked;
   const islandExecution = activeExecution ?? recentCompletedExecution;
 
+  // Easy mode runs a full run (artifact persisted, tracked in Runs pane)
+  // using runFull. runFull requires a non-empty fullRunName but the Easy form
+  // has no run-name input, so seed a default here as soon as Easy is active.
+  // User can still rename it from Advanced/Runs dialogs before clicking Run.
+  useEffect(() => {
+    if (!supportsEasyMode) return;
+    if (activeView !== "easy") return;
+    if (fullRunName.trim()) return;
+    const stamp = new Date().toISOString().replace(/[:.]/g, "-").slice(0, 16);
+    const base = workflowName.trim() || "Easy run";
+    setFullRunName(`${base} ${stamp}`);
+  }, [supportsEasyMode, activeView, fullRunName, workflowName, setFullRunName]);
+
   const toggleInteractive = useCallback(() => {
     if (executionLocked) {
       return;
@@ -739,6 +789,7 @@ export function RecipeStudioPage({
             savedAtLabel={savedAtLabel}
             workflowName={workflowName}
             warnings={getGraphWarnings(configs, edges)}
+            supportsEasyMode={supportsEasyMode}
             onWorkflowNameChange={setWorkflowName}
             onViewChange={setActiveView}
             onSaveRecipe={() => {
@@ -749,7 +800,25 @@ export function RecipeStudioPage({
             className="h-[75vh] w-full rounded-t-none"
             ref={flowContainerRef}
           >
-            {activeView === "editor" ? (
+            {activeView === "easy" ? (
+              <GithubCrawlerEasyView
+                configs={configs}
+                rows={fullRows}
+                setRows={setFullRows}
+                updateConfig={updateConfig}
+                onRun={() => {
+                  // Easy mode is a full run (artifact persisted, tracked in
+                  // the Runs pane) capped at the user's row count. runFull
+                  // requires a non-empty fullRunName; the effect below
+                  // populates one on mount so the closure in runFull is
+                  // already up to date by the time the user clicks Run.
+                  void runFull();
+                }}
+                runLoading={fullLoading || executionLocked}
+                runErrors={runErrors}
+                onSwitchToAdvanced={() => setActiveView("editor")}
+              />
+            ) : activeView === "editor" ? (
               editorContent
             ) : (
               <ExecutionsView
diff --git a/studio/frontend/src/features/recipe-studio/stores/recipe-studio.ts b/studio/frontend/src/features/recipe-studio/stores/recipe-studio.ts
index 5b16bde4b0..036cc94082 100644
--- a/studio/frontend/src/features/recipe-studio/stores/recipe-studio.ts
+++ b/studio/frontend/src/features/recipe-studio/stores/recipe-studio.ts
@@ -394,9 +394,11 @@ export const useRecipeStudioStore = create<RecipeStudioState>((set, get) => ({
         nextSourceType = "local";
       } else if (type === "seed_unstructured") {
         nextSourceType = "unstructured";
+      } else if (type === "seed_github") {
+        nextSourceType = "github_repo";
       }
 
-      const nextConfig = {
+      const nextConfig: typeof existing = {
         ...existing,
         seed_source_type: nextSourceType,
         hf_repo_id: "",
@@ -415,6 +417,12 @@ export const useRecipeStudioStore = create<RecipeStudioState>((set, get) => ({
         seed_preview_rows: [],
         unstructured_chunk_size: "1200",
         unstructured_chunk_overlap: "200",
+        github_repo_slug: "",
+        github_token: "",
+        github_limit: "100",
+        github_item_types: ["issues", "pulls"],
+        github_include_comments: true,
+        github_max_comments_per_item: "30",
       };
       return {
         configs: {
diff --git a/studio/frontend/src/features/recipe-studio/types/index.ts b/studio/frontend/src/features/recipe-studio/types/index.ts
index cd1127a4e2..9c720a06d3 100644
--- a/studio/frontend/src/features/recipe-studio/types/index.ts
+++ b/studio/frontend/src/features/recipe-studio/types/index.ts
@@ -38,7 +38,10 @@ export type LayoutDirection = "LR" | "TB";
 
 export type SeedSamplingStrategy = "ordered" | "shuffle";
 export type SeedSelectionType = "none" | "index_range" | "partition_block";
-export type SeedSourceType = "hf" | "local" | "unstructured";
+export type SeedSourceType = "hf" | "local" | "unstructured" | "github_repo";
+
+export type GithubItemType = "issues" | "pulls" | "commits";
+export type GithubStateFilter = "all" | "open" | "closed";
 export const INFRA_NODE_KINDS = new Set([
   "model_provider",
   "model_config",
@@ -338,6 +341,20 @@ export type SeedConfig = {
   unstructured_file_ids?: string[];
   unstructured_file_names?: string[];
   unstructured_file_sizes?: number[];
+  // biome-ignore lint/style/useNamingConvention: api schema
+  github_repo_slug?: string;
+  // biome-ignore lint/style/useNamingConvention: api schema
+  github_token?: string;
+  // biome-ignore lint/style/useNamingConvention: api schema
+  github_limit?: string;
+  // biome-ignore lint/style/useNamingConvention: api schema
+  github_item_types?: GithubItemType[];
+  // biome-ignore lint/style/useNamingConvention: api schema
+  github_state?: GithubStateFilter;
+  // biome-ignore lint/style/useNamingConvention: api schema
+  github_include_comments?: boolean;
+  // biome-ignore lint/style/useNamingConvention: api schema
+  github_max_comments_per_item?: string;
   resolved_paths?: string[];
   // ui-only
   seed_preview_rows?: Record<string, unknown>[];
diff --git a/studio/frontend/src/features/recipe-studio/utils/import/parsers/seed-config-parser.ts b/studio/frontend/src/features/recipe-studio/utils/import/parsers/seed-config-parser.ts
index 0d92129232..21eadb3195 100644
--- a/studio/frontend/src/features/recipe-studio/utils/import/parsers/seed-config-parser.ts
+++ b/studio/frontend/src/features/recipe-studio/utils/import/parsers/seed-config-parser.ts
@@ -80,6 +80,12 @@ function parseSeedSettings(seedConfigRaw: unknown): Partial<SeedConfig> {
   let resolved_paths: string[] = [];
   let unstructured_chunk_size = "1200";
   let unstructured_chunk_overlap = "200";
+  let github_repo_slug = "";
+  let github_token = "";
+  let github_limit = "100";
+  let github_item_types: ("issues" | "pulls" | "commits")[] = ["issues", "pulls"];
+  let github_include_comments = true;
+  let github_max_comments_per_item = "30";
   const sourceRaw = seedConfigRaw.source;
   if (isRecord(sourceRaw)) {
     const seedType = readString(sourceRaw.seed_type);
@@ -107,6 +113,26 @@ function parseSeedSettings(seedConfigRaw: unknown): Partial<SeedConfig> {
       unstructuredFileNames = [];
       unstructured_chunk_size = readNumberString(sourceRaw.chunk_size) || "1200";
       unstructured_chunk_overlap = readNumberString(sourceRaw.chunk_overlap) || "200";
+    } else if (seedType === "github_repo") {
+      seed_source_type = "github_repo";
+      const rawRepos = Array.isArray(sourceRaw.repos) ? sourceRaw.repos : [];
+      const repos = rawRepos.filter((r): r is string => typeof r === "string");
+      github_repo_slug = repos.join("\n");
+      github_token = readString(sourceRaw.token) ?? "";
+      github_limit = readNumberString(sourceRaw.limit) || "100";
+      const rawItems = Array.isArray(sourceRaw.item_types) ? sourceRaw.item_types : [];
+      const validItems = rawItems.filter(
+        (t): t is "issues" | "pulls" | "commits" =>
+          t === "issues" || t === "pulls" || t === "commits",
+      );
+      if (validItems.length > 0) {
+        github_item_types = validItems;
+      }
+      if (typeof sourceRaw.include_comments === "boolean") {
+        github_include_comments = sourceRaw.include_comments;
+      }
+      github_max_comments_per_item =
+        readNumberString(sourceRaw.max_comments_per_item) || "30";
     }
   }
 
@@ -147,6 +173,12 @@ function parseSeedSettings(seedConfigRaw: unknown): Partial<SeedConfig> {
     resolved_paths,
     unstructured_chunk_size,
     unstructured_chunk_overlap,
+    github_repo_slug,
+    github_token,
+    github_limit,
+    github_item_types,
+    github_include_comments,
+    github_max_comments_per_item,
     sampling_strategy,
     selection_type,
     selection_start,
diff --git a/studio/frontend/src/features/recipe-studio/utils/node-data.ts b/studio/frontend/src/features/recipe-studio/utils/node-data.ts
index 2fc5205db8..96042b5919 100644
--- a/studio/frontend/src/features/recipe-studio/utils/node-data.ts
+++ b/studio/frontend/src/features/recipe-studio/utils/node-data.ts
@@ -70,7 +70,9 @@ export function nodeDataFromConfig(
         ? "Hugging Face dataset"
         : seedSourceType === "local"
           ? "CSV or JSON file"
-          : "Document file";
+          : seedSourceType === "github_repo"
+            ? "GitHub repositories"
+            : "Document file";
     return {
       title: "Source data",
       kind: "seed",
diff --git a/studio/frontend/src/features/recipe-studio/utils/payload/builders-seed.ts b/studio/frontend/src/features/recipe-studio/utils/payload/builders-seed.ts
index f73a2b46da..bb48b43857 100644
--- a/studio/frontend/src/features/recipe-studio/utils/payload/builders-seed.ts
+++ b/studio/frontend/src/features/recipe-studio/utils/payload/builders-seed.ts
@@ -6,6 +6,7 @@ import type { NodeConfig, SeedConfig } from "../../types";
 const DEFAULT_CHUNK_SIZE = 1200;
 const DEFAULT_CHUNK_OVERLAP = 200;
 const MAX_CHUNK_SIZE = 20000;
+const GITHUB_ITEM_TYPES = new Set(["issues", "pulls", "commits"]);
 
 function parseIntStrict(value: string | undefined): number | null {
   const trimmed = value?.trim();
@@ -56,33 +57,86 @@ export function buildSeedConfig(
     selectionStrategy = { index, num_partitions: numPartitions };
   }
 
-  const source =
-    seedSourceType === "hf"
-      ? {
-          // biome-ignore lint/style/useNamingConvention: api schema
-          seed_type: "hf",
-          path,
-          token,
-          endpoint,
-        }
-      : seedSourceType === "unstructured"
-        ? (() => {
-            const { chunkSize, chunkOverlap } = resolveChunking(config);
-            return {
-              // biome-ignore lint/style/useNamingConvention: api schema
-              seed_type: "unstructured",
-              paths: config.resolved_paths?.length ? config.resolved_paths : [config.hf_path],
-              // biome-ignore lint/style/useNamingConvention: api schema
-              chunk_size: chunkSize,
-              // biome-ignore lint/style/useNamingConvention: api schema
-              chunk_overlap: chunkOverlap,
-            };
-          })()
-        : {
-            // biome-ignore lint/style/useNamingConvention: api schema
-            seed_type: "local",
-            path,
-          };
+  let source: Record<string, unknown>;
+  if (seedSourceType === "hf") {
+    source = {
+      // biome-ignore lint/style/useNamingConvention: api schema
+      seed_type: "hf",
+      path,
+      token,
+      endpoint,
+    };
+  } else if (seedSourceType === "unstructured") {
+    const { chunkSize, chunkOverlap } = resolveChunking(config);
+    source = {
+      // biome-ignore lint/style/useNamingConvention: api schema
+      seed_type: "unstructured",
+      paths: config.resolved_paths?.length ? config.resolved_paths : [config.hf_path],
+      // biome-ignore lint/style/useNamingConvention: api schema
+      chunk_size: chunkSize,
+      // biome-ignore lint/style/useNamingConvention: api schema
+      chunk_overlap: chunkOverlap,
+    };
+  } else if (seedSourceType === "github_repo") {
+    const repos = (config.github_repo_slug ?? "")
+      .split(/[\n,]/)
+      .map((r) => r.trim())
+      .filter(Boolean);
+    if (repos.length === 0) {
+      errors.push(`Seed ${config.name}: at least one repo is required.`);
+      return undefined;
+    }
+    const invalidRepo = repos.find((repo) => {
+      const parts = repo.split("/");
+      return parts.length !== 2 || parts.some((part) => !part);
+    });
+    if (invalidRepo) {
+      errors.push(
+        `Seed ${config.name}: GitHub repositories must use owner/name format.`,
+      );
+      return undefined;
+    }
+    const itemTypes = config.github_item_types?.length
+      ? config.github_item_types
+      : ["issues", "pulls"];
+    if (itemTypes.some((itemType) => !GITHUB_ITEM_TYPES.has(itemType))) {
+      errors.push(`Seed ${config.name}: GitHub item types invalid.`);
+      return undefined;
+    }
+    const limitNum = parseIntStrict(config.github_limit ?? "100");
+    if (limitNum === null || limitNum < 1 || limitNum > 5000) {
+      errors.push(
+        `Seed ${config.name}: GitHub items per repo must be an integer from 1 to 5000.`,
+      );
+      return undefined;
+    }
+    const maxCommentsNum = parseIntStrict(config.github_max_comments_per_item ?? "30");
+    if (maxCommentsNum === null || maxCommentsNum < 0 || maxCommentsNum > 200) {
+      errors.push(
+        `Seed ${config.name}: GitHub max comments per item must be an integer from 0 to 200.`,
+      );
+      return undefined;
+    }
+    source = {
+      // biome-ignore lint/style/useNamingConvention: api schema
+      seed_type: "github_repo",
+      repos,
+      token: (config.github_token ?? "").trim(),
+      // biome-ignore lint/style/useNamingConvention: api schema
+      item_types: itemTypes,
+      limit: limitNum,
+      // biome-ignore lint/style/useNamingConvention: api schema
+      include_comments: config.github_include_comments ?? true,
+      // biome-ignore lint/style/useNamingConvention: api schema
+      max_comments_per_item: maxCommentsNum,
+    };
+  } else {
+    source = {
+      // biome-ignore lint/style/useNamingConvention: api schema
+      seed_type: "local",
+      path,
+    };
+  }
 
   return {
     source,
diff --git a/studio/frontend/src/features/recipe-studio/utils/payload/types.ts b/studio/frontend/src/features/recipe-studio/utils/payload/types.ts
index b8b1aba93b..902ea796b4 100644
--- a/studio/frontend/src/features/recipe-studio/utils/payload/types.ts
+++ b/studio/frontend/src/features/recipe-studio/utils/payload/types.ts
@@ -57,7 +57,7 @@ export type RecipePayload = {
     // ui-only: graph orientation
     layout_direction?: "LR" | "TB";
     // ui-only, used to preserve seed block mode across imports/refresh
-    seed_source_type?: "hf" | "local" | "unstructured";
+    seed_source_type?: "hf" | "local" | "unstructured" | "github_repo";
     // ui-only, persisted aux node positions by llm name + aux key
     aux_nodes?: Array<{
       llm: string;
diff --git a/studio/frontend/src/features/recipe-studio/utils/validation.ts b/studio/frontend/src/features/recipe-studio/utils/validation.ts
index 5f48fd44a5..d5200eeed0 100644
--- a/studio/frontend/src/features/recipe-studio/utils/validation.ts
+++ b/studio/frontend/src/features/recipe-studio/utils/validation.ts
@@ -8,6 +8,7 @@ import { isOxcCodeShape } from "./validators/oxc-code-shape";
 import { isOxcValidationMode } from "./validators/oxc-mode";
 
 const TRACE_MODES = new Set(["none", "last_message", "all_messages"]);
+const GITHUB_ITEM_TYPES = new Set(["issues", "pulls", "commits"]);
 
 // biome-ignore lint/complexity/noExcessiveCognitiveComplexity: validation rules
 export function getConfigErrors(config: NodeConfig | null): string[] {
@@ -268,15 +269,49 @@ export function getConfigErrors(config: NodeConfig | null): string[] {
   }
   if (config.kind === "seed") {
     const seedSourceType = config.seed_source_type ?? "hf";
-    if (seedSourceType === "hf" && !config.hf_repo_id.trim()) {
+    if (seedSourceType === "github_repo") {
+      const repos = (config.github_repo_slug ?? "")
+        .split(/[\n,]/)
+        .map((repo) => repo.trim())
+        .filter(Boolean);
+      if (repos.length === 0) {
+        errors.push("Add at least one GitHub repository.");
+      }
+      if (
+        repos.some((repo) => {
+          const parts = repo.split("/");
+          return parts.length !== 2 || parts.some((part) => !part);
+        })
+      ) {
+        errors.push("GitHub repositories must use owner/name format.");
+      }
+      const itemTypes = config.github_item_types?.length
+        ? config.github_item_types
+        : ["issues", "pulls"];
+      if (itemTypes.length === 0) {
+        errors.push("Choose at least one GitHub item type.");
+      } else if (itemTypes.some((itemType) => !GITHUB_ITEM_TYPES.has(itemType))) {
+        errors.push("GitHub item types must be issues, pulls, or commits.");
+      }
+      const limit = parseIntNumber(config.github_limit ?? "100");
+      if (limit === null || limit < 1 || limit > 5000) {
+        errors.push("Items per repo must be an integer from 1 to 5000.");
+      }
+      const maxComments = parseIntNumber(config.github_max_comments_per_item ?? "30");
+      if (maxComments === null || maxComments < 0 || maxComments > 200) {
+        errors.push("Max comments per item must be an integer from 0 to 200.");
+      }
+    } else if (seedSourceType === "hf" && !config.hf_repo_id.trim()) {
       errors.push("Choose a Hugging Face dataset.");
     }
-    const hasPath =
-      seedSourceType === "unstructured"
-        ? (config.resolved_paths?.length ?? 0) > 0
-        : Boolean(config.hf_path.trim());
-    if (!hasPath) {
-      errors.push("Load the source-data preview first.");
+    if (seedSourceType !== "github_repo") {
+      const hasPath =
+        seedSourceType === "unstructured"
+          ? (config.resolved_paths?.length ?? 0) > 0
+          : Boolean(config.hf_path.trim());
+      if (!hasPath) {
+        errors.push("Load the source-data preview first.");
+      }
     }
     if (
       seedSourceType === "hf" &&
@@ -304,7 +339,7 @@ export function getConfigErrors(config: NodeConfig | null): string[] {
       ) {
         errors.push("Chunk overlap must be less than chunk size.");
       }
-    } else {
+    } else if (seedSourceType !== "github_repo") {
       const selectedDropColumns = (config.seed_drop_columns ?? [])
         .map((value) => value.trim())
         .filter(Boolean);
diff --git a/studio/frontend/src/lib/api-base.ts b/studio/frontend/src/lib/api-base.ts
index 664b9e58bd..c8be442653 100644
--- a/studio/frontend/src/lib/api-base.ts
+++ b/studio/frontend/src/lib/api-base.ts
@@ -8,6 +8,12 @@ if (isTauri && !isViteDev) {
   apiBase = 'http://127.0.0.1:8888'
 }
 
+const initialApiBase = apiBase
+
+export function resetApiBase() {
+  apiBase = initialApiBase
+}
+
 export function setApiBase(port: number) {
   apiBase = `http://127.0.0.1:${port}`
 }
diff --git a/studio/install_python_stack.py b/studio/install_python_stack.py
index b56b737cfa..5fe4415970 100644
--- a/studio/install_python_stack.py
+++ b/studio/install_python_stack.py
@@ -418,6 +418,9 @@ CONSTRAINTS = SINGLE_ENV / "constraints.txt"
 LOCAL_DD_UNSTRUCTURED_PLUGIN = (
     SCRIPT_DIR / "backend" / "plugins" / "data-designer-unstructured-seed"
 )
+LOCAL_DD_GITHUB_PLUGIN = (
+    SCRIPT_DIR / "backend" / "plugins" / "data-designer-github-repo-seed"
+)
 
 # -- Unicode-safe printing ---------------------------------------------
 # On Windows the default console encoding can be a legacy code page
@@ -1135,22 +1138,28 @@ def install_python_stack() -> int:
         req = SINGLE_ENV / "data-designer.txt",
     )
 
-    # 11. Local Data Designer seed plugin
-    if not LOCAL_DD_UNSTRUCTURED_PLUGIN.is_dir():
-        _safe_print(
-            _red(
-                f"❌ Missing local plugin directory: {LOCAL_DD_UNSTRUCTURED_PLUGIN}",
-            ),
-        )
-        return 1
+    # 11. Local Data Designer seed plugins
+    local_dd_plugins = [
+        ("unstructured", LOCAL_DD_UNSTRUCTURED_PLUGIN),
+        ("github", LOCAL_DD_GITHUB_PLUGIN),
+    ]
+    for _plugin_name, plugin_dir in local_dd_plugins:
+        if not plugin_dir.is_dir():
+            _safe_print(
+                _red(
+                    f"❌ Missing local plugin directory: {plugin_dir}",
+                ),
+            )
+            return 1
     _progress("local plugin")
-    pip_install(
-        "Installing local data-designer unstructured plugin",
-        "--no-cache-dir",
-        "--no-deps",
-        str(LOCAL_DD_UNSTRUCTURED_PLUGIN),
-        constrain = False,
-    )
+    for plugin_name, plugin_dir in local_dd_plugins:
+        pip_install(
+            f"Installing local data-designer {plugin_name} plugin",
+            "--no-cache-dir",
+            "--no-deps",
+            str(plugin_dir),
+            constrain = False,
+        )
 
     # 12. Patch metadata for single-env compatibility
     _progress("finalizing")

From efed5c37394a144349cd9b1ea525e132e04584e5 Mon Sep 17 00:00:00 2001
From: Leo Borcherding <borchborchmail@gmail.com>
Date: Sat, 25 Apr 2026 16:49:05 -0700
Subject: [PATCH 170/216] fix(studio): use endswith for mmproj F16 variant
 selection (#5184)

"f16" in filename matched BF16 files because "bf16" contains "f16"
as a substring. Switch to endswith("-f16.gguf") for an exact match.
---
 studio/backend/core/inference/llama_cpp.py | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/studio/backend/core/inference/llama_cpp.py b/studio/backend/core/inference/llama_cpp.py
index 33722f103c..c1a37e2fa4 100644
--- a/studio/backend/core/inference/llama_cpp.py
+++ b/studio/backend/core/inference/llama_cpp.py
@@ -1339,7 +1339,7 @@ class LlamaCppBackend:
             # Prefer F16 variant
             target = None
             for f in mmproj_files:
-                if "f16" in f.lower():
+                if f.lower().endswith("-f16.gguf"):
                     target = f
                     break
             if target is None:

From df3a205726d231bd12402960a3cd3a90b85c7fc7 Mon Sep 17 00:00:00 2001
From: "pre-commit-ci[bot]"
 <66853113+pre-commit-ci[bot]@users.noreply.github.com>
Date: Mon, 27 Apr 2026 14:17:03 -0700
Subject: [PATCH 171/216] [pre-commit.ci] pre-commit autoupdate (#5204)
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

updates:
- [github.com/astral-sh/ruff-pre-commit: v0.15.11 → v0.15.12](https://github.com/astral-sh/ruff-pre-commit/compare/v0.15.11...v0.15.12)

Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com>
---
 .pre-commit-config.yaml | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/.pre-commit-config.yaml b/.pre-commit-config.yaml
index 1d3137b373..a2a4995d62 100644
--- a/.pre-commit-config.yaml
+++ b/.pre-commit-config.yaml
@@ -1,6 +1,6 @@
 repos:
   - repo: https://github.com/astral-sh/ruff-pre-commit
-    rev: v0.15.11
+    rev: v0.15.12
     hooks:
       - id: ruff
         args:

From daf0889804aa3f33bbf2bde74034cad926f4a3bf Mon Sep 17 00:00:00 2001
From: Etherll <61019402+Etherll@users.noreply.github.com>
Date: Tue, 28 Apr 2026 11:10:47 +0300
Subject: [PATCH 172/216] Fix Windows install when paths contain spaces or
 Python 3.14 is on PATH (#5201)

* fix(studio): use py.exe to detect supported Python on Windows

  Description:
  The previous detection looked at `python --version` on PATH and
  hard-failed if the resolved Python wasn't 3.11-3.13. On systems
  where Python 3.14 sits ahead of 3.13 in PATH order, this aborted
  the installer even though a supported interpreter was installed.

  Prefer the py.exe launcher and probe `py -3.13`, `py -3.12`,
  `py -3.11` in turn. Fall back to `python --version` only when py.exe
  is absent, and surface a clearer error when no supported version
  can be found via either path.

* Studio: consolidate Windows studio overlay into single Tauri-gated block

  Replace the in-file sentinel hotfix and the unconditional file-copy
  overlay with a single block gated on $TauriMode. Hash-compare makes
  re-runs no-ops, removing the sentinel-clobbering bug that occurred
  when the second copy path overwrote the marker without re-adding it.

  Non-Tauri --local installs no longer need a copy overlay: the
  editable install above (uv pip install -e $RepoRoot --no-deps) makes
  _PACKAGE_ROOT in unsloth_cli/commands/studio.py resolve to the repo
  source tree via PEP 660 __file__-relative resolution, so
  `unsloth studio setup` finds the local setup.ps1 and
  install_python_stack.py without any file copying.

  Plain PyPI installs invoked from a checked-out repo directory are
  also no longer silently overlaid from cwd.

* fix(studio): work around uv space-in-path truncation on Windows

  uv 0.11.x truncates `-c <path>` and `-r <path>` arguments at the
  first space, breaking installs on Windows when the venv or repo
  sits under a path containing spaces (e.g. C:\Users\First Last\...).

  Pass paths through GetShortPathNameW to convert to 8.3 short form
  before handing them to uv. Plain pip is unaffected and keeps the
  original long path. No-op on Linux/Mac (gated on IS_WINDOWS and
  on the path actually containing a space).

* Refactor Python stack overlay logic in install.ps1

Refactor overlay logic for Python stack installation and improve handling of missing target directories.

* Update Python installation logic in setup.ps1
---
 install.ps1                    | 60 ++++++++++++++++++++-------------
 studio/install_python_stack.py | 47 ++++++++++++++++++++------
 studio/setup.ps1               | 61 +++++++++++++++++++++++++++-------
 3 files changed, 123 insertions(+), 45 deletions(-)

diff --git a/install.ps1 b/install.ps1
index 77c0034125..2544eaf78a 100644
--- a/install.ps1
+++ b/install.ps1
@@ -1027,32 +1027,46 @@ shell.Run cmd, 0, False
         }
     }
 
-    # Hotfix: patch install_python_stack.py for Windows GUI stdout
-    # The PyPI version crashes with OSError when stdout is piped from a GUI app.
-    # Copy our fixed version (bundled by Tauri) over the installed one.
-    # Remove this block once PyPI ships the fix from commit 18c5aae7.
+    # Overlay Tauri-bundled studio fixes that may be ahead of PyPI. Skipped
+    # for --local: the editable install above already makes _PACKAGE_ROOT in
+    # unsloth_cli/commands/studio.py resolve to the repo (PEP 660 __file__).
+    # Source paths match the Tauri bundle layout in studio/src-tauri/tauri.conf.json,
+    # which bundles install_python_stack.py at the bundle root next to install.ps1.
     if ($TauriMode) {
         $rawPath = if ($PSCommandPath) { $PSCommandPath } else { $MyInvocation.ScriptName }
-        $scriptDir = Split-Path -Parent ($rawPath -replace '^\\\\\?\\', '')
-        $fixedPy = Join-Path $scriptDir "install_python_stack.py"
-        $target = Join-Path $VenvDir "Lib\site-packages\studio\install_python_stack.py"
-        $sentinel = "# UNSLOTH_DESKTOP_HOTFIX_APPLIED_v1"
-        $sentinelPattern = [regex]::Escape($sentinel)
-        if ((Test-Path $fixedPy) -and (Test-Path $target)) {
-            $installed = Get-Content $target -Raw
-            if ($installed -notmatch $sentinelPattern) {
-                Copy-Item $fixedPy $target -Force
-                Add-Content -Path $target -Value "`n$sentinel"
-                substep "patched install_python_stack.py (stdout fix)"
-            } else {
-                substep "install_python_stack.py already has stdout fix"
+        if ($rawPath) {
+            # Strip leading \\?\ extended-length prefix if the launcher passed one.
+            $scriptDir = Split-Path -Parent ($rawPath -replace '^\\\\\?\\', '')
+            $overlayMap = [ordered]@{
+                "install_python_stack.py" = "Lib\site-packages\studio\install_python_stack.py"
+            }
+            foreach ($rel in $overlayMap.Keys) {
+                $src = Join-Path $scriptDir $rel
+                $dst = Join-Path $VenvDir $overlayMap[$rel]
+                if (-not (Test-Path $src)) { continue }
+                $dstParent = Split-Path -Parent $dst
+                if (-not (Test-Path $dstParent)) {
+                    Write-Host "[WARN] Overlay target dir missing: $dstParent; studio setup may use stale bundled file" -ForegroundColor Yellow
+                    continue
+                }
+                try {
+                    if (-not (Test-Path $dst)) {
+                        # Backfill: target file missing but parent dir exists.
+                        Copy-Item $src $dst -Force
+                        substep ("backfilled bundled " + (Split-Path -Leaf $rel))
+                    } else {
+                        # Hash-compare so re-runs are no-ops when files already match.
+                        $srcHash = (Get-FileHash $src -Algorithm SHA256).Hash
+                        $dstHash = (Get-FileHash $dst -Algorithm SHA256).Hash
+                        if ($srcHash -ne $dstHash) {
+                            Copy-Item $src $dst -Force
+                            substep ("applied bundled " + (Split-Path -Leaf $rel))
+                        }
+                    }
+                } catch {
+                    Write-Host "[WARN] Could not overlay $($rel): $($_.Exception.Message); studio setup may use stale bundled file" -ForegroundColor Yellow
+                }
             }
-        } elseif ((Test-Path $fixedPy) -and (Test-Path (Split-Path $target))) {
-            Copy-Item $fixedPy $target -Force
-            Add-Content -Path $target -Value "`n$sentinel"
-            substep "patched install_python_stack.py (stdout fix)"
-        } else {
-            Write-Host "[WARN] Could not patch install_python_stack.py (bundled file or target dir missing)" -ForegroundColor Yellow
         }
     }
 
diff --git a/studio/install_python_stack.py b/studio/install_python_stack.py
index 5fe4415970..8ac59d389d 100644
--- a/studio/install_python_stack.py
+++ b/studio/install_python_stack.py
@@ -359,6 +359,27 @@ def _ensure_rocm_torch() -> None:
             )
 
 
+def _uv_safe_path(path: object) -> str:
+    # uv 0.11.x: `-c <path with space>` truncates at the space; use 8.3 short form.
+    s = str(path)
+    if not IS_WINDOWS or " " not in s:
+        return s
+    try:
+        import ctypes
+        from ctypes import wintypes
+
+        get_short = ctypes.windll.kernel32.GetShortPathNameW
+        get_short.argtypes = [wintypes.LPCWSTR, wintypes.LPWSTR, wintypes.DWORD]
+        get_short.restype = wintypes.DWORD
+        buf = ctypes.create_unicode_buffer(32768)
+        rc = get_short(s, buf, 32768)
+        if 0 < rc < 32768 and " " not in buf.value:
+            return buf.value
+    except Exception:
+        pass
+    return s
+
+
 def _windows_hidden_subprocess_kwargs() -> dict[str, object]:
     """Return Windows-only subprocess kwargs that suppress console windows."""
     if not IS_WINDOWS:
@@ -751,14 +772,16 @@ def pip_install_try(
     """Like pip_install but returns False on failure instead of exiting.
     For optional installs with a follow-up fallback.
     """
-    constraint_args: list[str] = []
+    constraint_args_pip: list[str] = []
+    constraint_args_uv: list[str] = []
     if constrain and CONSTRAINTS.is_file():
-        constraint_args = ["-c", str(CONSTRAINTS)]
+        constraint_args_pip = ["-c", str(CONSTRAINTS)]
+        constraint_args_uv = ["-c", _uv_safe_path(CONSTRAINTS)]
 
     if USE_UV:
-        cmd = _build_uv_cmd(args) + constraint_args
+        cmd = _build_uv_cmd(args) + constraint_args_uv
     else:
-        cmd = _build_pip_cmd(args) + constraint_args
+        cmd = _build_pip_cmd(args) + constraint_args_pip
 
     if VERBOSE:
         _step(_LABEL, f"{label}...", _dim)
@@ -781,9 +804,11 @@ def pip_install(
     constrain: bool = True,
 ) -> None:
     """Build and run a pip install command (uses uv when available, falls back to pip)."""
-    constraint_args: list[str] = []
+    constraint_args_pip: list[str] = []
+    constraint_args_uv: list[str] = []
     if constrain and CONSTRAINTS.is_file():
-        constraint_args = ["-c", str(CONSTRAINTS)]
+        constraint_args_pip = ["-c", str(CONSTRAINTS)]
+        constraint_args_uv = ["-c", _uv_safe_path(CONSTRAINTS)]
 
     actual_req = req
     temp_reqs: list[Path] = []
@@ -793,13 +818,15 @@ def pip_install(
     if actual_req is not None and NO_TORCH and NO_TORCH_SKIP_PACKAGES:
         actual_req = _filter_requirements(actual_req, NO_TORCH_SKIP_PACKAGES)
         temp_reqs.append(actual_req)
-    req_args: list[str] = []
+    req_args_pip: list[str] = []
+    req_args_uv: list[str] = []
     if actual_req is not None:
-        req_args = ["-r", str(actual_req)]
+        req_args_pip = ["-r", str(actual_req)]
+        req_args_uv = ["-r", _uv_safe_path(actual_req)]
 
     try:
         if USE_UV:
-            uv_cmd = _build_uv_cmd(args) + constraint_args + req_args
+            uv_cmd = _build_uv_cmd(args) + constraint_args_uv + req_args_uv
             if VERBOSE:
                 print(f"   {label}...")
             result = subprocess.run(
@@ -814,7 +841,7 @@ def pip_install(
             if result.stdout:
                 print(result.stdout.decode(errors = "replace"))
 
-        pip_cmd = _build_pip_cmd(args) + constraint_args + req_args
+        pip_cmd = _build_pip_cmd(args) + constraint_args_pip + req_args_pip
         run(f"{label} (pip)" if USE_UV else label, pip_cmd)
     finally:
         for temp_req in temp_reqs:
diff --git a/studio/setup.ps1 b/studio/setup.ps1
index 24ef3ef1eb..0da30c9229 100644
--- a/studio/setup.ps1
+++ b/studio/setup.ps1
@@ -1145,28 +1145,58 @@ if ($IsPipInstall) {
     }
 }
 
-# ============================================
-# 1g. Python (>= 3.11 and < 3.14, matching setup.sh)
-# ============================================
+# 1g. Python (>= 3.11 and < 3.14). Prefer py.exe so a 3.14 ahead of 3.13 on PATH does not trip the gate.
 $HasPython = $null -ne (Get-Command python -ErrorAction SilentlyContinue)
+$PyLauncher = Get-Command py -CommandType Application -ErrorAction SilentlyContinue
 $PythonOk = $false
+$DetectedPyVer = $null
 
-if ($HasPython) {
+if ($PyLauncher) {
+    foreach ($minor in @("3.13", "3.12", "3.11")) {
+        try {
+            $out = & $PyLauncher.Source "-$minor" --version 2>&1 | Out-String
+            if ($out -match 'Python (3\.\d+\.\d+)') {
+                $DetectedPyVer = $Matches[1]
+                # Make `python` resolvable for the rest of setup. Without this,
+                # py-launcher-only installs (no python.exe on PATH) pass the gate
+                # and then crash on the first bare `python` call below.
+                try {
+                    $resolvedExe = (& $PyLauncher.Source "-$minor" -c "import sys; print(sys.executable)" 2>$null | Select-Object -First 1)
+                    if ($resolvedExe -and (Test-Path $resolvedExe)) {
+                        $resolvedDir = Split-Path -Parent $resolvedExe
+                        $alreadyOnPath = ($env:PATH -split ';' | Where-Object { $_.TrimEnd('\') -ieq $resolvedDir.TrimEnd('\') }).Count -gt 0
+                        if (-not $alreadyOnPath) {
+                            $env:PATH = "$resolvedDir;$env:PATH"
+                        }
+                        $HasPython = $true
+                    }
+                } catch { }
+                $PythonOk = $true
+                break
+            }
+        } catch { }
+    }
+}
+
+if (-not $PythonOk -and $HasPython) {
     $PyVer = python --version 2>&1
     if ($PyVer -match "(\d+)\.(\d+)") {
         $PyMajor = [int]$Matches[1]; $PyMinor = [int]$Matches[2]
         if ($PyMajor -eq 3 -and $PyMinor -ge 11 -and $PyMinor -lt 14) {
-            substep "Python $PyVer"
+            $DetectedPyVer = "$PyMajor.$PyMinor"
             $PythonOk = $true
-        } else {
-            Write-Host "[ERROR] Python $PyVer is outside supported range (need >= 3.11 and < 3.14)." -ForegroundColor Red
-            Write-Host "        Install Python 3.12 from https://python.org/downloads/" -ForegroundColor Yellow
-            exit 1
         }
     }
-} else {
-    # No Python at all -- install 3.12
-    Write-Host "Python not found -- installing Python 3.12 via winget..." -ForegroundColor Yellow
+}
+
+if ($PythonOk) {
+    substep "Python $DetectedPyVer"
+} elseif (-not $HasPython) {
+    # No `python` on PATH (and py.exe either absent or only had unsupported
+    # minors). Try winget as before -- gating on $HasPython alone, not also
+    # on $PyLauncher, so a launcher-only install with just 3.14 still gets
+    # an automatic 3.12 install instead of a hard error.
+    Write-Host "Python 3.11-3.13 not found -- installing Python 3.12 via winget..." -ForegroundColor Yellow
     $HasWinget = $null -ne (Get-Command winget -ErrorAction SilentlyContinue)
     if ($HasWinget) {
         winget install -e --id Python.Python.3.12 --source winget --accept-package-agreements --accept-source-agreements
@@ -1180,6 +1210,13 @@ if ($HasPython) {
     }
     step "python" "$(python --version 2>&1)"
     $PythonOk = $true
+} else {
+    # python.exe is on PATH but its version is unsupported, and py.exe (if
+    # present) had no supported minor either.
+    Write-Host "[ERROR] No supported Python (3.11-3.13) found on this system." -ForegroundColor Red
+    Write-Host "        py.exe could not locate -3.11/-3.12/-3.13 and `python` on PATH is unsupported." -ForegroundColor Yellow
+    Write-Host "        Install Python 3.12 from https://python.org/downloads/" -ForegroundColor Yellow
+    exit 1
 }
 
 # Add user-scheme Python Scripts dir to PATH (nt_user only, no venv fallback).

From 230d58872d0fa44279f6358cd9c5116bcc2cd3fe Mon Sep 17 00:00:00 2001
From: Lee Jackson <130007945+Imagineer99@users.noreply.github.com>
Date: Tue, 28 Apr 2026 10:39:48 +0100
Subject: [PATCH 173/216] Studio: Preserve transparency in uploaded profile
 avatars (#5200)

* fix: preserve transparency in uploaded profile avatars

* fix: guard unsupported canvas mime fallback
---
 .../profile/utils/resize-image-file.ts        | 50 +++++++++++++++----
 1 file changed, 40 insertions(+), 10 deletions(-)

diff --git a/studio/frontend/src/features/profile/utils/resize-image-file.ts b/studio/frontend/src/features/profile/utils/resize-image-file.ts
index 3f829ba975..c420497395 100644
--- a/studio/frontend/src/features/profile/utils/resize-image-file.ts
+++ b/studio/frontend/src/features/profile/utils/resize-image-file.ts
@@ -3,6 +3,11 @@
 
 const MAX_EDGE = 256;
 const MAX_BYTES = 380_000;
+const MAX_DATA_URL_LENGTH = Math.floor(MAX_BYTES * 1.35);
+const JPEG_QUALITY_START = 0.88;
+const WEBP_QUALITY_START = 0.9;
+const MIN_QUALITY = 0.45;
+const QUALITY_STEP = 0.08;
 
 function loadImage(file: File): Promise<HTMLImageElement> {
   return new Promise((resolve, reject) => {
@@ -20,8 +25,33 @@ function loadImage(file: File): Promise<HTMLImageElement> {
   });
 }
 
+function canvasHasTransparency(ctx: CanvasRenderingContext2D, width: number, height: number): boolean {
+  const { data } = ctx.getImageData(0, 0, width, height);
+  for (let i = 3; i < data.length; i += 4) {
+    if (data[i] < 255) return true;
+  }
+  return false;
+}
+
+function encodeCanvasWithinLimit(
+  canvas: HTMLCanvasElement,
+  mimeType: string,
+  startQuality: number,
+): string | null {
+  let quality = startQuality;
+  let dataUrl = canvas.toDataURL(mimeType, quality);
+  if (!dataUrl.startsWith(`data:${mimeType}`)) return null;
+
+  while (dataUrl.length > MAX_DATA_URL_LENGTH && quality > MIN_QUALITY) {
+    quality -= QUALITY_STEP;
+    dataUrl = canvas.toDataURL(mimeType, quality);
+  }
+
+  return dataUrl.length <= MAX_DATA_URL_LENGTH ? dataUrl : null;
+}
+
 /**
- * Downscale and re-encode as JPEG so localStorage stays within reasonable size.
+ * Downscale the image and keep transparency when present while staying localStorage-friendly.
  */
 export async function resizeImageFileToDataUrl(file: File): Promise<string> {
   const img = await loadImage(file);
@@ -40,14 +70,14 @@ export async function resizeImageFileToDataUrl(file: File): Promise<string> {
   if (!ctx) throw new Error("Canvas not available");
   ctx.drawImage(img, 0, 0, cw, ch);
 
-  let quality = 0.88;
-  let dataUrl = canvas.toDataURL("image/jpeg", quality);
-  while (dataUrl.length > MAX_BYTES * 1.35 && quality > 0.45) {
-    quality -= 0.08;
-    dataUrl = canvas.toDataURL("image/jpeg", quality);
+  const hasTransparency = canvasHasTransparency(ctx, cw, ch);
+  if (hasTransparency) {
+    const webpDataUrl = encodeCanvasWithinLimit(canvas, "image/webp", WEBP_QUALITY_START);
+    if (webpDataUrl) return webpDataUrl;
   }
-  if (dataUrl.length > MAX_BYTES * 1.35) {
-    throw new Error("Image is still too large after compression. Try a smaller file.");
-  }
-  return dataUrl;
+
+  const jpegDataUrl = encodeCanvasWithinLimit(canvas, "image/jpeg", JPEG_QUALITY_START);
+  if (jpegDataUrl) return jpegDataUrl;
+
+  throw new Error("Image is still too large after compression. Try a smaller file.");
 }

From 2469ac885bb0450532d6c76f7ff1a24d484a9b81 Mon Sep 17 00:00:00 2001
From: Lee Jackson <130007945+Imagineer99@users.noreply.github.com>
Date: Tue, 28 Apr 2026 10:39:59 +0100
Subject: [PATCH 174/216] UX: single chat header error placement and selector
 alignment (#5173)

* UX: single chat header error placement and selector alignment

* fix: centre model dropdown chevron

* fix: revert view mode single

* fix: allow model selector label to truncate in narrow headers
---
 .../assistant-ui/model-selector.tsx           | 28 +++++++++++--------
 .../frontend/src/features/chat/chat-page.tsx  | 17 +++++++----
 2 files changed, 28 insertions(+), 17 deletions(-)

diff --git a/studio/frontend/src/components/assistant-ui/model-selector.tsx b/studio/frontend/src/components/assistant-ui/model-selector.tsx
index 3628252177..679034c9fa 100644
--- a/studio/frontend/src/components/assistant-ui/model-selector.tsx
+++ b/studio/frontend/src/components/assistant-ui/model-selector.tsx
@@ -66,7 +66,7 @@ function ModelSelectorTrigger({
         type="button"
         data-tour={dataTour}
         className={cn(
-          "flex items-center gap-2 transition-colors",
+          "flex min-w-0 items-center gap-2 transition-colors",
           variant === "outline" &&
           "rounded-[8px] border border-border/60 hover:bg-[#ececec] dark:hover:bg-[#2e3035]",
           variant === "ghost" && "rounded-[8px] hover:bg-[#ececec] dark:hover:bg-[#2e3035]",
@@ -80,17 +80,23 @@ function ModelSelectorTrigger({
         {isLoaded && (
           <span className="size-2 shrink-0 rounded-full bg-emerald-500" />
         )}
-        <span className="font-heading font-medium text-[16px] text-black dark:text-white">
-          {currentModel?.name ?? "Select model"}
+        <span className="flex min-w-0 flex-1 items-baseline gap-2">
+          <span className="min-w-0 flex-1 truncate font-heading text-[16px] font-medium leading-none text-black dark:text-white">
+            {currentModel?.name ?? "Select model"}
+          </span>
+          {currentModel?.description && (
+            <span className="shrink-0 text-xs leading-none text-muted-foreground">
+              {currentModel.description}
+            </span>
+          )}
+        </span>
+        <span className="flex size-4 shrink-0 items-center justify-center">
+          <HugeiconsIcon
+            icon={ArrowDown01Icon}
+            strokeWidth={1.75}
+            className="relative top-0.5 size-3.5 text-muted-foreground"
+          />
         </span>
-        {currentModel?.description && (
-          <span className="text-muted-foreground text-xs">{currentModel.description}</span>
-        )}
-        <HugeiconsIcon
-          icon={ArrowDown01Icon}
-          strokeWidth={1.75}
-          className="size-3.5 shrink-0 text-muted-foreground"
-        />
       </button>
     </PopoverTrigger>
   );
diff --git a/studio/frontend/src/features/chat/chat-page.tsx b/studio/frontend/src/features/chat/chat-page.tsx
index 9987af230b..e3239f7cc1 100644
--- a/studio/frontend/src/features/chat/chat-page.tsx
+++ b/studio/frontend/src/features/chat/chat-page.tsx
@@ -886,7 +886,7 @@ export function ChatPage(): ReactElement {
                 onOpenChange={handleModelSelectorOpenChange}
                 triggerDataTour="chat-model-selector"
                 contentDataTour="chat-model-selector-popover"
-                className="max-w-[62vw] sm:max-w-none !h-[34px]"
+                className="max-w-[62vw] !pr-3 sm:max-w-none !h-[34px]"
               />
             )}
             {loadingModel && loadToastDismissed ? (
@@ -910,12 +910,17 @@ export function ChatPage(): ReactElement {
                 onStop={cancelLoading}
               />
             ) : null}
+            {!loadingModel && modelsError ? (
+              <div
+                className="relative top-0.5 max-w-[28rem] truncate pl-0.5 text-xs text-destructive"
+                title={modelsError}
+                role="status"
+                aria-live="polite"
+              >
+                {modelsError}
+              </div>
+            ) : null}
           </div>
-          {modelsError && (
-            <div className="ml-2 text-xs text-destructive truncate max-w-[28rem]">
-              {modelsError}
-            </div>
-          )}
           <div className="ml-auto flex items-center gap-2">
             {view.mode === "single" && ggufContextLength && contextUsage ? (
               <ContextUsageBar

From 975a5c354f3f97b074028300b95ce1b1e0e2ef9b Mon Sep 17 00:00:00 2001
From: Lee Jackson <130007945+Imagineer99@users.noreply.github.com>
Date: Tue, 28 Apr 2026 10:40:15 +0100
Subject: [PATCH 175/216] Studio: Refine chat preset and group built-in presets
 (#5159)

* UX: Refine chat preset and group built-in presets

* fix: reuse built-in preset names and unify GGUF state reads

* fix: built-in chat preset save and refresh behavior

* Add chat preset invariant tests

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

* fix: decouple chat presets from model-specific settings

Limit chat preset compare/apply/save behavior to temperature, topP, topK, minP, repetitionPenalty, presencePenalty, maxTokens, and systemPrompt.

Preserve legacy stored preset data on load for backwards compatibility, but stop treating model-specific settings such as checkpoint, trustRemoteCode, and maxSeqLength as part of preset identity.

Also align legacy prompt migration dedupe with the new preset semantics and add invariant coverage for preset-owned config comparisons.

* fix: detect built-in preset edits from param changes

* fix: correct built-in preset dirty state and speculative select values

* fix: preserve default preset sync and keep qwen think pristine

---------

Co-authored-by: Daniel Han <danielhanchen@gmail.com>
Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com>
---
 .../frontend/src/features/chat/chat-page.tsx  |   4 -
 .../src/features/chat/chat-settings-sheet.tsx | 738 ++++++++----------
 .../chat/hooks/use-chat-model-runtime.ts      | 111 +--
 .../features/chat/presets/preset-policy.ts    | 351 +++++++++
 .../chat/stores/chat-runtime-store.ts         |  29 +
 .../test_chat_preset_builtin_invariants.py    | 272 +++++++
 6 files changed, 1040 insertions(+), 465 deletions(-)
 create mode 100644 studio/frontend/src/features/chat/presets/preset-policy.ts
 create mode 100644 tests/studio/test_chat_preset_builtin_invariants.py

diff --git a/studio/frontend/src/features/chat/chat-page.tsx b/studio/frontend/src/features/chat/chat-page.tsx
index e3239f7cc1..39683661c7 100644
--- a/studio/frontend/src/features/chat/chat-page.tsx
+++ b/studio/frontend/src/features/chat/chat-page.tsx
@@ -484,10 +484,6 @@ export function ChatPage(): ReactElement {
   const settingsOpen = useChatRuntimeStore((s) => s.settingsPanelOpen);
   const setSettingsOpen = useChatRuntimeStore((s) => s.setSettingsPanelOpen);
 
-  useEffect(() => {
-    return () => setSettingsOpen(false);
-  }, [setSettingsOpen]);
-
   useEffect(() => {
     const threadId = search.thread;
     if (!threadId) return;
diff --git a/studio/frontend/src/features/chat/chat-settings-sheet.tsx b/studio/frontend/src/features/chat/chat-settings-sheet.tsx
index fc5f097969..08c5ef4080 100644
--- a/studio/frontend/src/features/chat/chat-settings-sheet.tsx
+++ b/studio/frontend/src/features/chat/chat-settings-sheet.tsx
@@ -19,6 +19,7 @@ import {
   DropdownMenu,
   DropdownMenuContent,
   DropdownMenuItem,
+  DropdownMenuSeparator,
   DropdownMenuTrigger,
 } from "@/components/ui/dropdown-menu";
 import { Input } from "@/components/ui/input";
@@ -64,54 +65,36 @@ import {
 } from "@/components/ui/tooltip";
 import { Tooltip as TooltipPrimitive } from "radix-ui";
 import { AnimatePresence, motion } from "motion/react";
-import type { ReactNode } from "react";
+import { Fragment, type ReactNode } from "react";
 import { useEffect, useLayoutEffect, useMemo, useRef, useState } from "react";
 import { toast } from "sonner";
 import { useChatRuntimeStore } from "./stores/chat-runtime-store";
 import {
-  DEFAULT_INFERENCE_PARAMS,
-  type InferenceParams,
-} from "./types/runtime";
+  applyPresetParams,
+  BUILTIN_PRESET_NAMES,
+  BUILTIN_PRESETS,
+  defaultInferenceParams,
+  getBuiltinVariantName,
+  getOrderedPresets,
+  getPresetOwnedConfigKey,
+  getPresetSaveState,
+  getPresetSource,
+  getUniquePresetName,
+  isSamePresetConfig,
+  normalizeCustomPresets,
+  toPresetParams,
+  type Preset,
+} from "./presets/preset-policy";
+import type { InferenceParams } from "./types/runtime";
 
-export const defaultInferenceParams = DEFAULT_INFERENCE_PARAMS;
+export { defaultInferenceParams, type Preset } from "./presets/preset-policy";
 export type { InferenceParams } from "./types/runtime";
 
-export interface Preset {
-  name: string;
-  params: InferenceParams;
-}
-
 interface LegacySystemPromptTemplate {
   name: string;
   content: string;
 }
 
-const BUILTIN_PRESETS: Preset[] = [
-  { name: "Default", params: { ...defaultInferenceParams } },
-  {
-    name: "Creative",
-    params: {
-      ...defaultInferenceParams,
-      temperature: 1.5,
-      topP: 1.0,
-      topK: 0,
-      minP: 0.1,
-      repetitionPenalty: 1.0,
-    },
-  },
-  {
-    name: "Precise",
-    params: {
-      ...defaultInferenceParams,
-      temperature: 0.1,
-      topP: 0.95,
-      topK: 80,
-      minP: 0.01,
-      repetitionPenalty: 1.0,
-    },
-  },
-];
-
 const CHAT_PRESETS_KEY = "unsloth_chat_custom_presets";
 const CHAT_ACTIVE_PRESET_KEY = "unsloth_chat_active_preset";
 const LEGACY_CHAT_SYSTEM_PROMPTS_KEY = "unsloth_chat_system_prompts";
@@ -122,16 +105,13 @@ function canUseStorage(): boolean {
   return typeof window !== "undefined";
 }
 
-function getUniquePresetName(baseName: string, usedNames: Set<string>): string {
-  const normalizedBase = baseName.trim() || "Imported Prompt";
-  let nextName = normalizedBase;
-  let suffix = 2;
-  while (usedNames.has(nextName)) {
-    nextName = `${normalizedBase} ${suffix}`;
-    suffix += 1;
+function saveCustomPresets(presets: Preset[]): void {
+  if (!canUseStorage()) return;
+  try {
+    localStorage.setItem(CHAT_PRESETS_KEY, JSON.stringify(presets));
+  } catch {
+    // ignore
   }
-  usedNames.add(nextName);
-  return nextName;
 }
 
 function migrateLegacySystemPromptTemplates(presets: Preset[]): Preset[] {
@@ -161,18 +141,7 @@ function migrateLegacySystemPromptTemplates(presets: Preset[]): Preset[] {
     ]);
     const seenImportedConfigKeys = new Set(
       [...BUILTIN_PRESETS, ...presets].map((preset) =>
-        JSON.stringify({
-          temperature: preset.params.temperature,
-          topP: preset.params.topP,
-          topK: preset.params.topK,
-          minP: preset.params.minP,
-          repetitionPenalty: preset.params.repetitionPenalty,
-          presencePenalty: preset.params.presencePenalty,
-          maxSeqLength: preset.params.maxSeqLength,
-          maxTokens: preset.params.maxTokens,
-          systemPrompt: preset.params.systemPrompt,
-          trustRemoteCode: preset.params.trustRemoteCode ?? false,
-        }),
+        getPresetOwnedConfigKey(preset.params),
       ),
     );
     const importedPresets = parsed
@@ -191,18 +160,7 @@ function migrateLegacySystemPromptTemplates(presets: Preset[]): Preset[] {
         },
       }))
       .filter(({ importedParams }) => {
-        const configKey = JSON.stringify({
-          temperature: importedParams.temperature,
-          topP: importedParams.topP,
-          topK: importedParams.topK,
-          minP: importedParams.minP,
-          repetitionPenalty: importedParams.repetitionPenalty,
-          presencePenalty: importedParams.presencePenalty,
-          maxSeqLength: importedParams.maxSeqLength,
-          maxTokens: importedParams.maxTokens,
-          systemPrompt: importedParams.systemPrompt,
-          trustRemoteCode: importedParams.trustRemoteCode ?? false,
-        });
+        const configKey = getPresetOwnedConfigKey(importedParams);
         if (seenImportedConfigKeys.has(configKey)) return false;
         seenImportedConfigKeys.add(configKey);
         return true;
@@ -216,8 +174,8 @@ function migrateLegacySystemPromptTemplates(presets: Preset[]): Preset[] {
       localStorage.setItem(LEGACY_CHAT_SYSTEM_PROMPTS_MIGRATED_KEY, raw);
       return presets;
     }
-    const mergedPresets = [...presets, ...importedPresets];
-    localStorage.setItem(CHAT_PRESETS_KEY, JSON.stringify(mergedPresets));
+    const mergedPresets = normalizeCustomPresets([...presets, ...importedPresets]);
+    saveCustomPresets(mergedPresets);
     try {
       localStorage.setItem(LEGACY_CHAT_SYSTEM_PROMPTS_MIGRATED_KEY, raw);
       localStorage.removeItem(LEGACY_CHAT_SYSTEM_PROMPTS_KEY);
@@ -255,7 +213,11 @@ function loadSavedCustomPresets(): Preset[] {
         },
       }))
       .filter((preset) => preset.name.length > 0);
-    return migrateLegacySystemPromptTemplates(presets);
+    const normalized = normalizeCustomPresets(presets);
+    if (JSON.stringify(normalized) !== JSON.stringify(presets)) {
+      saveCustomPresets(normalized);
+    }
+    return migrateLegacySystemPromptTemplates(normalized);
   } catch {
     return migrateLegacySystemPromptTemplates([]);
   }
@@ -270,82 +232,6 @@ function loadSavedActivePreset(): string {
   }
 }
 
-type PresetSaveMode =
-  | "disabled"
-  | "overwrite-active"
-  | "overwrite-other"
-  | "create";
-
-interface PresetSaveState {
-  mode: PresetSaveMode;
-  canSubmit: boolean;
-  isSaveReady: boolean;
-  buttonLabel: string;
-  title: string;
-}
-
-function isSamePresetConfig(a: InferenceParams, b: InferenceParams): boolean {
-  return (
-    a.temperature === b.temperature &&
-    a.topP === b.topP &&
-    a.topK === b.topK &&
-    a.minP === b.minP &&
-    a.repetitionPenalty === b.repetitionPenalty &&
-    a.presencePenalty === b.presencePenalty &&
-    a.maxSeqLength === b.maxSeqLength &&
-    a.maxTokens === b.maxTokens &&
-    a.systemPrompt === b.systemPrompt &&
-    (a.trustRemoteCode ?? false) === (b.trustRemoteCode ?? false)
-  );
-}
-
-function getPresetSaveState({
-  rawName,
-  activePreset,
-  presets,
-  activePresetDirty,
-}: {
-  rawName: string;
-  activePreset: string;
-  presets: Preset[];
-  activePresetDirty: boolean;
-}): PresetSaveState {
-  const trimmedName = rawName.trim();
-  if (!trimmedName) {
-    return {
-      mode: "disabled",
-      canSubmit: false,
-      isSaveReady: false,
-      buttonLabel: "Save",
-      title: "Enter a preset name",
-    };
-  }
-
-  const matchingPreset = presets.find((preset) => preset.name === trimmedName);
-  if (matchingPreset) {
-    const isActiveMatch = matchingPreset.name === activePreset;
-    return {
-      mode: isActiveMatch ? "overwrite-active" : "overwrite-other",
-      canSubmit: !isActiveMatch || activePresetDirty,
-      isSaveReady: !isActiveMatch || activePresetDirty,
-      buttonLabel: isActiveMatch && !activePresetDirty ? "Saved" : "Overwrite",
-      title: isActiveMatch
-        ? activePresetDirty
-          ? "Save current settings to this preset"
-          : "No unsaved changes"
-        : `Overwrite preset "${trimmedName}"`,
-    };
-  }
-
-  return {
-    mode: "create",
-    canSubmit: true,
-    isSaveReady: true,
-    buttonLabel: "Save as New",
-    title: `Save current settings as "${trimmedName}"`,
-  };
-}
-
 function ParamSlider({
   label,
   value,
@@ -518,6 +404,10 @@ export function ChatSettingsPanel({
   const setCustomContextLength = useChatRuntimeStore(
     (s) => s.setCustomContextLength,
   );
+  const setActivePresetSource = useChatRuntimeStore(
+    (s) => s.setActivePresetSource,
+  );
+  const activePresetSource = useChatRuntimeStore((s) => s.activePresetSource);
 
   const ctxDisplayValue = customContextLength ?? ggufContextLength ?? "";
   const ctxMaxValue = ggufNativeContextLength ?? ggufContextLength ?? null;
@@ -540,12 +430,9 @@ export function ChatSettingsPanel({
   >(undefined);
   const [systemPromptEditorOpen, setSystemPromptEditorOpen] = useState(false);
   const [systemPromptDraft, setSystemPromptDraft] = useState("");
+  const [activePresetBaseline, setActivePresetBaseline] = useState(params);
   const presets = useMemo(() => {
-    const overrides = new Set(customPresets.map((preset) => preset.name));
-    return [
-      ...BUILTIN_PRESETS.filter((preset) => !overrides.has(preset.name)),
-      ...customPresets,
-    ];
+    return getOrderedPresets(customPresets);
   }, [customPresets]);
   const activePresetDefinition = useMemo(
     () => presets.find((preset) => preset.name === activePreset) ?? null,
@@ -555,17 +442,23 @@ export function ChatSettingsPanel({
     () => customPresets.find((preset) => preset.name === activePreset) ?? null,
     [activePreset, customPresets],
   );
-  const activeBuiltinPreset = useMemo(
-    () =>
-      BUILTIN_PRESETS.find((preset) => preset.name === activePreset) ?? null,
-    [activePreset],
-  );
-  const activePresetDirty = useMemo(
-    () =>
-      activePresetDefinition == null
-        ? false
-        : !isSamePresetConfig(activePresetDefinition.params, params),
-    [activePresetDefinition, params],
+  const hasUnsavedPresetChanges = useMemo(
+    () => {
+      if (activePresetDefinition == null) {
+        return false;
+      }
+      if (BUILTIN_PRESET_NAMES.has(activePresetDefinition.name)) {
+        if (activePresetDefinition.name === "Default") {
+          return activePresetSource === "modified";
+        }
+        return (
+          activePresetSource === "modified" ||
+          !isSamePresetConfig(activePresetDefinition.params, params)
+        );
+      }
+      return !isSamePresetConfig(activePresetDefinition.params, params);
+    },
+    [activePresetDefinition, activePresetSource, params],
   );
   const presetSaveState = useMemo(
     () =>
@@ -573,9 +466,9 @@ export function ChatSettingsPanel({
         rawName: presetNameInput,
         activePreset,
         presets,
-        activePresetDirty,
+        hasUnsavedPresetChanges,
       }),
-    [activePreset, activePresetDirty, presetNameInput, presets],
+    [activePreset, hasUnsavedPresetChanges, presetNameInput, presets],
   );
   const systemPromptEditorDirty = systemPromptDraft !== params.systemPrompt;
   const trustRemoteCodeMissing =
@@ -584,27 +477,24 @@ export function ChatSettingsPanel({
     !(params.trustRemoteCode ?? false);
 
   function set<K extends keyof InferenceParams>(key: K) {
-    return (v: InferenceParams[K]) => onParamsChange({ ...params, [key]: v });
+    return (v: InferenceParams[K]) => {
+      const nextParams = { ...params, [key]: v };
+      const nextSource = isSamePresetConfig(activePresetBaseline, nextParams)
+        ? getPresetSource(activePreset)
+        : "modified";
+      setActivePresetSource(nextSource);
+      onParamsChange(nextParams);
+    };
   }
 
   function applyPreset(name: string) {
     const p = presets.find((pr) => pr.name === name);
     if (p) {
-      if (
-        modelRequiresTrustRemoteCode &&
-        !(p.params.trustRemoteCode ?? false)
-      ) {
-        toast.warning("This configuration turns custom code off", {
-          description:
-            "The current model needs custom code enabled to load. Keep it on for this model.",
-        });
-        return;
-      }
       onParamsChange({
-        ...p.params,
-        checkpoint: params.checkpoint,
+        ...applyPresetParams(params, p.params),
       });
       setActivePreset(name);
+      setActivePresetSource(getPresetSource(name));
       if (canUseStorage()) {
         try {
           localStorage.setItem(CHAT_ACTIVE_PRESET_KEY, name);
@@ -621,27 +511,29 @@ export function ChatSettingsPanel({
       toast.error("Enter a preset name");
       return;
     }
+    const usedNames = new Set([
+      ...BUILTIN_PRESET_NAMES,
+      ...customPresets.map((preset) => preset.name),
+    ]);
+    const saveName = BUILTIN_PRESET_NAMES.has(trimmed)
+      ? getBuiltinVariantName(trimmed, usedNames)
+      : trimmed;
     setCustomPresets((prev) => {
-      const next = prev.filter((p) => p.name !== trimmed);
-      const merged = [...next, { name: trimmed, params: { ...params } }];
-      if (canUseStorage()) {
-        try {
-          localStorage.setItem(CHAT_PRESETS_KEY, JSON.stringify(merged));
-        } catch {
-          // ignore
-        }
-      }
+      const next = prev.filter((p) => p.name !== saveName);
+      const merged = [...next, { name: saveName, params: toPresetParams(params) }];
+      saveCustomPresets(merged);
       return merged;
     });
     if (canUseStorage()) {
       try {
-        localStorage.setItem(CHAT_ACTIVE_PRESET_KEY, trimmed);
+        localStorage.setItem(CHAT_ACTIVE_PRESET_KEY, saveName);
       } catch {
         // ignore
       }
     }
-    setActivePreset(trimmed);
-    setPresetNameInput(trimmed);
+    setActivePreset(saveName);
+    setActivePresetSource("custom");
+    setPresetNameInput(saveName);
   }
 
   function deletePreset(name: string) {
@@ -651,41 +543,20 @@ export function ChatSettingsPanel({
     if (!hasCustomPreset) {
       return;
     }
-    const builtinPreset = BUILTIN_PRESETS.find((preset) => preset.name === name);
     const fallbackPreset =
-      builtinPreset ??
-      BUILTIN_PRESETS.find((preset) => preset.name === "Default") ??
-      null;
-    if (
-      activePreset === name &&
-      fallbackPreset &&
-      modelRequiresTrustRemoteCode &&
-      !(fallbackPreset.params.trustRemoteCode ?? false)
-    ) {
-      toast.warning("Reset would turn custom code off", {
-        description:
-          "The current model needs custom code enabled to load. Keep it on for this model.",
-      });
-      return;
-    }
+      BUILTIN_PRESETS.find((preset) => preset.name === "Default") ?? null;
     setCustomPresets((prev) => {
       const next = prev.filter((preset) => preset.name !== name);
-      if (canUseStorage()) {
-        try {
-          localStorage.setItem(CHAT_PRESETS_KEY, JSON.stringify(next));
-        } catch {
-          // ignore
-        }
-      }
+      saveCustomPresets(next);
       return next;
     });
     if (activePreset === name) {
       if (fallbackPreset) {
         onParamsChange({
-          ...fallbackPreset.params,
-          checkpoint: params.checkpoint,
+          ...applyPresetParams(params, fallbackPreset.params),
         });
         setActivePreset(fallbackPreset.name);
+        setActivePresetSource("builtin-default");
         if (canUseStorage()) {
           try {
             localStorage.setItem(CHAT_ACTIVE_PRESET_KEY, fallbackPreset.name);
@@ -708,8 +579,46 @@ export function ChatSettingsPanel({
   }
 
   useEffect(() => {
-    if (presets.some((preset) => preset.name === activePreset)) return;
+    if (activePresetSource !== "modified") {
+      setActivePresetBaseline(params);
+    }
+  }, [activePresetSource, params]);
+
+  useEffect(() => {
+    if (presets.some((preset) => preset.name === activePreset)) {
+      const expectedSource = getPresetSource(activePreset);
+      if (activePresetDefinition != null) {
+        if (BUILTIN_PRESET_NAMES.has(activePresetDefinition.name)) {
+          if (activePresetDefinition.name === "Default") {
+            if (
+              activePresetSource !== "modified" &&
+              activePresetSource !== expectedSource
+            ) {
+              setActivePresetSource(expectedSource);
+            }
+            return;
+          }
+          const matchesActivePreset = isSamePresetConfig(
+            activePresetDefinition.params,
+            params,
+          );
+          const nextSource = matchesActivePreset ? expectedSource : "modified";
+          if (activePresetSource !== nextSource) {
+            setActivePresetSource(nextSource);
+          }
+          return;
+        }
+      }
+      if (
+        activePresetSource !== "modified" &&
+        activePresetSource !== expectedSource
+      ) {
+        setActivePresetSource(expectedSource);
+      }
+      return;
+    }
     setActivePreset("Default");
+    setActivePresetSource("builtin-default");
     if (canUseStorage()) {
       try {
         localStorage.setItem(CHAT_ACTIVE_PRESET_KEY, "Default");
@@ -717,7 +626,14 @@ export function ChatSettingsPanel({
         // ignore
       }
     }
-  }, [activePreset, presets]);
+  }, [
+    activePreset,
+    activePresetDefinition,
+    activePresetSource,
+    params,
+    presets,
+    setActivePresetSource,
+  ]);
 
   useEffect(() => {
     setPresetNameInput(activePreset);
@@ -741,6 +657,187 @@ export function ChatSettingsPanel({
     return () => ro.disconnect();
   }, [open]);
 
+  const modelSection = (
+    <CollapsibleSection
+      icon={Settings02Icon}
+      label="Model"
+      defaultOpen={true}
+    >
+      <div className="flex flex-col gap-3 py-1">
+        {isGguf && (
+          <>
+            <div className="space-y-2">
+              <div className="flex items-center justify-between">
+                <span className="text-xs font-medium">Context Length</span>
+                <Input
+                  type="number"
+                  value={
+                    typeof ctxDisplayValue === "number"
+                      ? ctxDisplayValue
+                      : (ggufContextLength ?? "")
+                  }
+                  placeholder="..."
+                  min={128}
+                  max={ctxMaxValue ?? undefined}
+                  step={1024}
+                  className="h-6 w-[100px] text-right text-xs tabular-nums"
+                  onChange={(e) => {
+                    const raw = e.target.value;
+                    if (raw === "") {
+                      setCustomContextLength(null);
+                      return;
+                    }
+                    const v = Number.parseInt(raw, 10);
+                    if (!Number.isNaN(v) && v >= 0) {
+                      const maxCtx = ctxMaxValue ?? Number.POSITIVE_INFINITY;
+                      const clamped = Math.min(v, maxCtx);
+                      setCustomContextLength(
+                        clamped === (ggufContextLength ?? 0) ? null : clamped,
+                      );
+                    }
+                  }}
+                />
+              </div>
+              <Slider
+                min={1024}
+                max={ctxMaxValue ?? 4096}
+                step={1024}
+                value={[
+                  Math.min(
+                    typeof ctxDisplayValue === "number"
+                      ? ctxDisplayValue
+                      : (ggufContextLength ?? 4096),
+                    ctxMaxValue ?? 4096,
+                  ),
+                ]}
+                onValueChange={([v]) => {
+                  setCustomContextLength(
+                    v === (ggufContextLength ?? 0) ? null : v,
+                  );
+                }}
+              />
+              {ggufMaxContextLength != null &&
+                typeof ctxDisplayValue === "number" &&
+                ctxDisplayValue > ggufMaxContextLength && (
+                  <p className="text-[11px] text-amber-500">
+                    Exceeds estimated VRAM capacity (
+                    {ggufMaxContextLength.toLocaleString()} tokens). The model
+                    may use system RAM.
+                  </p>
+                )}
+            </div>
+            <div className="grid grid-cols-[minmax(0,1fr)_65px] items-center gap-x-3">
+              <div className="min-w-0">
+                <div className="text-xs font-medium">KV Cache Dtype</div>
+                <div className="text-[11px] text-muted-foreground">
+                  Quantize KV cache to reduce VRAM.
+                </div>
+              </div>
+              <div className="w-full min-w-0">
+                <Select
+                  value={kvCacheDtype ?? "f16"}
+                  onValueChange={(v) => {
+                    setKvCacheDtype(v === "f16" ? null : v);
+                  }}
+                >
+                  <SelectTrigger className="grid h-7 w-full min-w-0 grid-cols-[minmax(0,1fr)_auto] items-center gap-1 px-2 py-0 text-xs [&_[data-slot=select-value]]:min-w-0 [&_[data-slot=select-value]]:truncate [&>svg]:shrink-0">
+                    <SelectValue />
+                  </SelectTrigger>
+                  <SelectContent>
+                    <SelectItem value="f16">f16</SelectItem>
+                    <SelectItem value="bf16">bf16</SelectItem>
+                    <SelectItem value="q8_0">q8_0</SelectItem>
+                    <SelectItem value="q5_1">q5_1</SelectItem>
+                    <SelectItem value="q4_1">q4_1</SelectItem>
+                  </SelectContent>
+                </Select>
+              </div>
+            </div>
+            {!currentModelIsVision && (
+              <div className="grid grid-cols-[minmax(0,1fr)_65px] items-center gap-x-3">
+                <div className="min-w-0">
+                  <div className="text-xs font-medium">
+                    Speculative Decoding
+                  </div>
+                  <div className="text-[11px] text-muted-foreground">
+                    Speed up generation with no VRAM cost.
+                  </div>
+                </div>
+                <div className="w-full min-w-0">
+                  <Select
+                    value={speculativeType ?? "off"}
+                    onValueChange={(v) => {
+                      setSpeculativeType(v === "off" ? null : v);
+                    }}
+                  >
+                    <SelectTrigger className="grid h-7 w-full min-w-0 grid-cols-[minmax(0,1fr)_auto] items-center gap-1 px-2 py-0 text-xs [&_[data-slot=select-value]]:min-w-0 [&_[data-slot=select-value]]:truncate [&>svg]:shrink-0">
+                      <SelectValue />
+                    </SelectTrigger>
+                    <SelectContent>
+                      <SelectItem value="default">On</SelectItem>
+                      <SelectItem value="off">Off</SelectItem>
+                    </SelectContent>
+                  </Select>
+                </div>
+              </div>
+            )}
+            {modelSettingsDirty && (
+              <div className="flex flex-wrap gap-1.5 pt-1">
+                <button
+                  type="button"
+                  onClick={() => onReloadModel?.()}
+                  className="rounded-md bg-primary px-2.5 py-1 text-[11px] font-medium text-primary-foreground transition-colors hover:bg-primary/90"
+                >
+                  Apply
+                </button>
+                <button
+                  type="button"
+                  onClick={() => {
+                    setCustomContextLength(null);
+                    setKvCacheDtype(loadedKvCacheDtype);
+                    setSpeculativeType(loadedSpeculativeType);
+                  }}
+                  className="rounded-md border px-2.5 py-1 text-[11px] font-medium text-muted-foreground transition-colors hover:bg-accent"
+                >
+                  Reset
+                </button>
+              </div>
+            )}
+          </>
+        )}
+        {!isGguf && params.checkpoint && (
+          <>
+            <div className="flex items-center justify-between gap-3">
+              <div className="min-w-0">
+                <div className="text-xs font-medium">Enable custom code</div>
+                <div className="text-[11px] text-muted-foreground">
+                  Allow models with custom code (e.g. Nemotron). Only enable if
+                  sure.
+                </div>
+              </div>
+              <Switch
+                checked={params.trustRemoteCode ?? false}
+                onCheckedChange={set("trustRemoteCode")}
+              />
+            </div>
+            {trustRemoteCodeMissing && (
+              <Alert className="border-amber-200/70 bg-amber-50/70 px-3 py-2 text-amber-950 dark:border-amber-900/70 dark:bg-amber-950/35 dark:text-amber-100">
+                <AlertTitle className="text-[11px] font-medium">
+                  Keep custom code enabled for this model
+                </AlertTitle>
+                <AlertDescription className="text-[11px] text-amber-800 dark:text-amber-200">
+                  This model requires custom code to load. You can edit the
+                  toggle, but loading will stay blocked until it is turned back
+                  on.
+                </AlertDescription>
+              </Alert>
+            )}
+          </>
+        )}
+      </div>
+    </CollapsibleSection>
+  );
+
   const settingsContent = (
     <>
       <div className="aui-thread-viewport relative h-full overflow-y-auto">
@@ -834,13 +931,16 @@ export function ChatSettingsPanel({
                       : undefined
                   }
                 >
-                  {presets.map((p) => (
-                    <DropdownMenuItem
-                      key={p.name}
-                      onSelect={() => applyPreset(p.name)}
-                    >
-                      {p.name}
-                    </DropdownMenuItem>
+                  {presets.map((p, index) => (
+                    <Fragment key={p.name}>
+                      <DropdownMenuItem onSelect={() => applyPreset(p.name)}>
+                        {p.name}
+                      </DropdownMenuItem>
+                      {index === BUILTIN_PRESETS.length - 1 &&
+                        presets.length > BUILTIN_PRESETS.length && (
+                          <DropdownMenuSeparator className="mx-2.5! my-1.5! h-0! border-t border-border/70 bg-transparent!" />
+                        )}
+                    </Fragment>
                   ))}
                 </DropdownMenuContent>
               </DropdownMenu>
@@ -874,9 +974,7 @@ export function ChatSettingsPanel({
                 className="h-8 w-full text-xs text-muted-foreground"
                 title={
                   activeCustomPreset
-                    ? activeBuiltinPreset
-                      ? "Reset selected preset to built-in defaults"
-                      : "Delete selected preset"
+                    ? "Delete selected preset"
                     : "No saved override to delete"
                 }
               >
@@ -918,188 +1016,6 @@ export function ChatSettingsPanel({
           />
         </div>
 
-        <CollapsibleSection
-          icon={Settings02Icon}
-          label="Model"
-          defaultOpen={true}
-        >
-          <div className="flex flex-col gap-3 py-1">
-            {isGguf && (
-              <>
-                <div className="space-y-2">
-                  <div className="flex items-center justify-between">
-                    <span className="text-xs font-medium">Context Length</span>
-                    <Input
-                      type="number"
-                      value={
-                        typeof ctxDisplayValue === "number"
-                          ? ctxDisplayValue
-                          : (ggufContextLength ?? "")
-                      }
-                      placeholder="..."
-                      min={128}
-                      max={ctxMaxValue ?? undefined}
-                      step={1024}
-                      className="h-6 w-[100px] text-right text-xs tabular-nums"
-                      onChange={(e) => {
-                        const raw = e.target.value;
-                        if (raw === "") {
-                          setCustomContextLength(null);
-                          return;
-                        }
-                        const v = Number.parseInt(raw, 10);
-                        if (!Number.isNaN(v) && v >= 0) {
-                          const maxCtx =
-                            ctxMaxValue ?? Number.POSITIVE_INFINITY;
-                          const clamped = Math.min(v, maxCtx);
-                          setCustomContextLength(
-                            clamped === (ggufContextLength ?? 0)
-                              ? null
-                              : clamped,
-                          );
-                        }
-                      }}
-                    />
-                  </div>
-                  <Slider
-                    min={1024}
-                    max={ctxMaxValue ?? 4096}
-                    step={1024}
-                    value={[
-                      Math.min(
-                        typeof ctxDisplayValue === "number"
-                          ? ctxDisplayValue
-                          : (ggufContextLength ?? 4096),
-                        ctxMaxValue ?? 4096,
-                      ),
-                    ]}
-                    onValueChange={([v]) => {
-                      setCustomContextLength(
-                        v === (ggufContextLength ?? 0) ? null : v,
-                      );
-                    }}
-                  />
-                  {ggufMaxContextLength != null &&
-                    typeof ctxDisplayValue === "number" &&
-                    ctxDisplayValue > ggufMaxContextLength && (
-                      <p className="text-[11px] text-amber-500">
-                        Exceeds estimated VRAM capacity (
-                        {ggufMaxContextLength.toLocaleString()} tokens). The
-                        model may use system RAM.
-                      </p>
-                    )}
-                </div>
-                <div className="grid grid-cols-[minmax(0,1fr)_65px] items-center gap-x-3">
-                  <div className="min-w-0">
-                    <div className="text-xs font-medium">KV Cache Dtype</div>
-                    <div className="text-[11px] text-muted-foreground">
-                      Quantize KV cache to reduce VRAM.
-                    </div>
-                  </div>
-                  <div className="w-full min-w-0">
-                    <Select
-                      value={kvCacheDtype ?? "f16"}
-                      onValueChange={(v) => {
-                        setKvCacheDtype(v === "f16" ? null : v);
-                      }}
-                    >
-                      <SelectTrigger className="grid h-7 w-full min-w-0 grid-cols-[minmax(0,1fr)_auto] items-center gap-1 px-2 py-0 text-xs [&_[data-slot=select-value]]:min-w-0 [&_[data-slot=select-value]]:truncate [&>svg]:shrink-0">
-                        <SelectValue />
-                      </SelectTrigger>
-                      <SelectContent>
-                        <SelectItem value="f16">f16</SelectItem>
-                        <SelectItem value="bf16">bf16</SelectItem>
-                        <SelectItem value="q8_0">q8_0</SelectItem>
-                        <SelectItem value="q5_1">q5_1</SelectItem>
-                        <SelectItem value="q4_1">q4_1</SelectItem>
-                      </SelectContent>
-                    </Select>
-                  </div>
-                </div>
-                {!currentModelIsVision && (
-                  <div className="grid grid-cols-[minmax(0,1fr)_65px] items-center gap-x-3">
-                    <div className="min-w-0">
-                      <div className="text-xs font-medium">
-                        Speculative Decoding
-                      </div>
-                      <div className="text-[11px] text-muted-foreground">
-                        Speed up generation with no VRAM cost.
-                      </div>
-                    </div>
-                    <div className="w-full min-w-0">
-                      <Select
-                        value={speculativeType ?? "off"}
-                        onValueChange={(v) => {
-                          setSpeculativeType(v === "off" ? null : v);
-                        }}
-                      >
-                        <SelectTrigger className="grid h-7 w-full min-w-0 grid-cols-[minmax(0,1fr)_auto] items-center gap-1 px-2 py-0 text-xs [&_[data-slot=select-value]]:min-w-0 [&_[data-slot=select-value]]:truncate [&>svg]:shrink-0">
-                          <SelectValue />
-                        </SelectTrigger>
-                        <SelectContent>
-                          <SelectItem value="default">On</SelectItem>
-                          <SelectItem value="off">Off</SelectItem>
-                        </SelectContent>
-                      </Select>
-                    </div>
-                  </div>
-                )}
-                {modelSettingsDirty && (
-                  <div className="flex flex-wrap gap-1.5 pt-1">
-                    <button
-                      type="button"
-                      onClick={() => onReloadModel?.()}
-                      className="rounded-md bg-primary px-2.5 py-1 text-[11px] font-medium text-primary-foreground transition-colors hover:bg-primary/90"
-                    >
-                      Apply
-                    </button>
-                    <button
-                      type="button"
-                      onClick={() => {
-                        setCustomContextLength(null);
-                        setKvCacheDtype(loadedKvCacheDtype);
-                        setSpeculativeType(loadedSpeculativeType);
-                      }}
-                      className="rounded-md border px-2.5 py-1 text-[11px] font-medium text-muted-foreground transition-colors hover:bg-accent"
-                    >
-                      Reset
-                    </button>
-                  </div>
-                )}
-              </>
-            )}
-            {!isGguf && params.checkpoint && (
-              <>
-                <div className="flex items-center justify-between gap-3">
-                  <div className="min-w-0">
-                    <div className="text-xs font-medium">Enable custom code</div>
-                    <div className="text-[11px] text-muted-foreground">
-                      Allow models with custom code (e.g. Nemotron). Only
-                      enable if sure.
-                    </div>
-                  </div>
-                  <Switch
-                    checked={params.trustRemoteCode ?? false}
-                    onCheckedChange={set("trustRemoteCode")}
-                  />
-                </div>
-                {trustRemoteCodeMissing && (
-                  <Alert className="border-amber-200/70 bg-amber-50/70 px-3 py-2 text-amber-950 dark:border-amber-900/70 dark:bg-amber-950/35 dark:text-amber-100">
-                    <AlertTitle className="text-[11px] font-medium">
-                      Keep custom code enabled for this model
-                    </AlertTitle>
-                    <AlertDescription className="text-[11px] text-amber-800 dark:text-amber-200">
-                      This model requires custom code to load. You can edit the
-                      toggle, but loading will stay blocked until it is turned
-                      back on.
-                    </AlertDescription>
-                  </Alert>
-                )}
-              </>
-            )}
-          </div>
-        </CollapsibleSection>
-
         <CollapsibleSection
           icon={SlidersHorizontalIcon}
           label="Sampling"
@@ -1186,6 +1102,8 @@ export function ChatSettingsPanel({
           </div>
         </CollapsibleSection>
 
+        {modelSection}
+
         <CollapsibleSection icon={Wrench01Icon} label="Tools">
           <div className="flex flex-col gap-3 py-1">
             <AutoHealToolCallsToggle />
diff --git a/studio/frontend/src/features/chat/hooks/use-chat-model-runtime.ts b/studio/frontend/src/features/chat/hooks/use-chat-model-runtime.ts
index e90cab73fa..cfac6dc8cf 100644
--- a/studio/frontend/src/features/chat/hooks/use-chat-model-runtime.ts
+++ b/studio/frontend/src/features/chat/hooks/use-chat-model-runtime.ts
@@ -17,6 +17,10 @@ import {
 } from "../api/chat-api";
 import { formatEta, formatRate } from "../utils/format-transfer";
 import { useChatRuntimeStore } from "../stores/chat-runtime-store";
+import {
+  mergeBackendRecommendedInference,
+  resolveLoadMaxSeqLength,
+} from "../presets/preset-policy";
 import type { InferenceStatusResponse, LoadModelResponse } from "../types/api";
 import type {
   ChatLoraSummary,
@@ -132,46 +136,10 @@ function toLoraSummary(lora: {
   };
 }
 
-function toFiniteNumber(value: unknown): number | undefined {
-  if (typeof value !== "number" || !Number.isFinite(value)) {
-    return undefined;
-  }
-  return value;
-}
-
 function getTrustRemoteCodeRequiredMessage(modelName: string): string {
   return `${modelName} needs custom code enabled to load. Turn on "Enable custom code" in Chat Settings, then try again.`;
 }
 
-function mergeRecommendedInference(
-  current: InferenceParams,
-  response: LoadModelResponse | InferenceStatusResponse,
-  modelId: string,
-): InferenceParams {
-  const inference = response.inference;
-  // GGUF: use actual context length from GGUF metadata, fallback to 131072
-  // Non-GGUF: 4096
-  const defaultMaxTokens = response.is_gguf
-    ? (response.context_length ?? 131072)
-    : 4096;
-  return {
-    ...current,
-    checkpoint: modelId,
-    maxTokens: defaultMaxTokens,
-    temperature:
-      toFiniteNumber(inference?.temperature) ?? current.temperature,
-    topP: toFiniteNumber(inference?.top_p) ?? current.topP,
-    topK: toFiniteNumber(inference?.top_k) ?? current.topK,
-    minP: toFiniteNumber(inference?.min_p) ?? current.minP,
-    presencePenalty:
-      toFiniteNumber(inference?.presence_penalty) ?? current.presencePenalty,
-    trustRemoteCode:
-      typeof inference?.trust_remote_code === "boolean"
-        ? inference.trust_remote_code
-        : current.trustRemoteCode,
-  };
-}
-
 export function useChatModelRuntime() {
   const params = useChatRuntimeStore((state) => state.params);
   const models = useChatRuntimeStore((state) => state.models);
@@ -273,7 +241,12 @@ export function useChatModelRuntime() {
             speculative_type: statusRes.speculative_type,
           };
           setParams(
-            mergeRecommendedInference(currentParams, statusRes, statusRes.active_model),
+            mergeBackendRecommendedInference({
+              current: currentParams,
+              response: statusRes,
+              modelId: statusRes.active_model,
+              presetSource: useChatRuntimeStore.getState().activePresetSource,
+            }),
           );
         }
 
@@ -465,13 +438,25 @@ export function useChatModelRuntime() {
               previousWasUnloaded = true;
             }
 
-            const { chatTemplateOverride, kvCacheDtype, customContextLength, ggufContextLength, speculativeType } = useChatRuntimeStore.getState();
-            // GGUF: use custom context length, or 0 = model's native context
-            // Non-GGUF: use the Max Seq Length slider value
-            const isDirectGgufFile = modelId.toLowerCase().endsWith(".gguf");
-            const effectiveMaxSeqLength = customContextLength != null
-              ? customContextLength
-              : (ggufVariant != null || isDirectGgufFile) ? (ggufContextLength ?? 0) : maxSeqLength;
+            const {
+              chatTemplateOverride,
+              kvCacheDtype,
+              customContextLength,
+              ggufContextLength,
+              speculativeType,
+              activePresetSource,
+              activeGgufVariant,
+            } = useChatRuntimeStore.getState();
+            const effectiveMaxSeqLength = resolveLoadMaxSeqLength({
+              modelId,
+              ggufVariant,
+              customContextLength,
+              ggufContextLength,
+              currentCheckpoint,
+              activeGgufVariant,
+              maxSeqLength,
+              presetSource: activePresetSource,
+            });
             const loadResponse = await loadModel({
               model_path: modelId,
               hf_token: hfToken,
@@ -491,7 +476,12 @@ export function useChatModelRuntime() {
 
             const currentParams = useChatRuntimeStore.getState().params;
             setParams(
-              mergeRecommendedInference(currentParams, loadResponse, modelId),
+              mergeBackendRecommendedInference({
+                current: currentParams,
+                response: loadResponse,
+                modelId,
+                presetSource: useChatRuntimeStore.getState().activePresetSource,
+              }),
             );
             // Qwen3.5/3.6 small models (0.8B, 2B, 4B, 9B) disable thinking by default
             let reasoningDefault = loadResponse.supports_reasoning ?? false;
@@ -545,12 +535,31 @@ export function useChatModelRuntime() {
             // Qwen3/3.5/3.6: apply thinking-mode-specific params after load
             if (modelId.toLowerCase().includes("qwen3") && (loadResponse.supports_reasoning ?? false)) {
               const store = useChatRuntimeStore.getState();
-              const mid = modelId.toLowerCase();
-              const needsPresencePenalty = mid.includes("qwen3.5") || mid.includes("qwen3.6");
-              const p = reasoningDefault
-                ? { temperature: 0.6, topP: 0.95, topK: 20, minP: 0.0, ...(needsPresencePenalty ? { presencePenalty: 1.5 } : {}) }
-                : { temperature: 0.7, topP: 0.8, topK: 20, minP: 0.0, ...(needsPresencePenalty ? { presencePenalty: 1.5 } : {}) };
-              store.setParams({ ...store.params, ...p });
+              if (store.activePresetSource === "builtin-default") {
+                const mid = modelId.toLowerCase();
+                const needsPresencePenalty =
+                  mid.includes("qwen3.5") || mid.includes("qwen3.6");
+                const p = reasoningDefault
+                  ? {
+                      temperature: 0.6,
+                      topP: 0.95,
+                      topK: 20,
+                      minP: 0.0,
+                      ...(needsPresencePenalty
+                        ? { presencePenalty: 1.5 }
+                        : {}),
+                    }
+                  : {
+                      temperature: 0.7,
+                      topP: 0.8,
+                      topK: 20,
+                      minP: 0.0,
+                      ...(needsPresencePenalty
+                        ? { presencePenalty: 1.5 }
+                        : {}),
+                    };
+                store.setParams({ ...store.params, ...p });
+              }
             }
             await refresh();
           } catch (error) {
diff --git a/studio/frontend/src/features/chat/presets/preset-policy.ts b/studio/frontend/src/features/chat/presets/preset-policy.ts
new file mode 100644
index 0000000000..8c5d2573c9
--- /dev/null
+++ b/studio/frontend/src/features/chat/presets/preset-policy.ts
@@ -0,0 +1,351 @@
+// SPDX-License-Identifier: AGPL-3.0-only
+// Copyright 2026-present the Unsloth AI Inc. team. All rights reserved. See /studio/LICENSE.AGPL-3.0
+
+import {
+  DEFAULT_INFERENCE_PARAMS,
+  type InferenceParams,
+} from "../types/runtime";
+
+export const defaultInferenceParams = DEFAULT_INFERENCE_PARAMS;
+
+export interface Preset {
+  name: string;
+  params: InferenceParams;
+}
+
+export type PresetOwnedParams = Pick<
+  InferenceParams,
+  | "temperature"
+  | "topP"
+  | "topK"
+  | "minP"
+  | "repetitionPenalty"
+  | "presencePenalty"
+  | "maxTokens"
+  | "systemPrompt"
+>;
+
+export const BUILTIN_PRESETS: Preset[] = [
+  { name: "Default", params: { ...defaultInferenceParams } },
+  {
+    name: "Creative",
+    params: {
+      ...defaultInferenceParams,
+      temperature: 1.5,
+      topP: 1.0,
+      topK: 0,
+      minP: 0.1,
+      repetitionPenalty: 1.0,
+    },
+  },
+  {
+    name: "Precise",
+    params: {
+      ...defaultInferenceParams,
+      temperature: 0.1,
+      topP: 0.95,
+      topK: 80,
+      minP: 0.01,
+      repetitionPenalty: 1.0,
+    },
+  },
+];
+
+export const BUILTIN_PRESET_NAMES = new Set(
+  BUILTIN_PRESETS.map((preset) => preset.name),
+);
+
+export type ChatPresetSource =
+  | "builtin-default"
+  | "builtin-fixed"
+  | "custom"
+  | "modified";
+
+export function getPresetSource(name: string): ChatPresetSource {
+  if (name === "Default") return "builtin-default";
+  if (BUILTIN_PRESET_NAMES.has(name)) return "builtin-fixed";
+  return "custom";
+}
+
+export function getUniquePresetName(
+  baseName: string,
+  usedNames: Set<string>,
+): string {
+  const normalizedBase = baseName.trim() || "Imported Prompt";
+  let nextName = normalizedBase;
+  let suffix = 2;
+  while (usedNames.has(nextName)) {
+    nextName = `${normalizedBase} ${suffix}`;
+    suffix += 1;
+  }
+  usedNames.add(nextName);
+  return nextName;
+}
+
+export function getBuiltinVariantName(
+  baseName: string,
+  usedNames: Set<string>,
+): string {
+  const normalizedBase = baseName.trim() || "Imported Prompt";
+  let suffix = 1;
+  let nextName = `${normalizedBase} ${suffix}`;
+  while (usedNames.has(nextName)) {
+    suffix += 1;
+    nextName = `${normalizedBase} ${suffix}`;
+  }
+  usedNames.add(nextName);
+  return nextName;
+}
+
+export function normalizeCustomPresets(presets: Preset[]): Preset[] {
+  const usedNames = new Set(BUILTIN_PRESET_NAMES);
+  return presets
+    .map((preset): Preset | null => {
+      const trimmedName = preset.name.trim();
+      if (!trimmedName) return null;
+      const name = usedNames.has(trimmedName)
+        ? getBuiltinVariantName(trimmedName, usedNames)
+        : trimmedName;
+      usedNames.add(name);
+      return {
+        name,
+        params: preset.params,
+      };
+    })
+    .filter((preset): preset is Preset => preset !== null);
+}
+
+export function getOrderedPresets(customPresets: Preset[]): Preset[] {
+  return [...BUILTIN_PRESETS, ...normalizeCustomPresets(customPresets)];
+}
+
+export function isSamePresetConfig(
+  a: InferenceParams,
+  b: InferenceParams,
+): boolean {
+  const left = getPresetOwnedParams(a);
+  const right = getPresetOwnedParams(b);
+  return (
+    left.temperature === right.temperature &&
+    left.topP === right.topP &&
+    left.topK === right.topK &&
+    left.minP === right.minP &&
+    left.repetitionPenalty === right.repetitionPenalty &&
+    left.presencePenalty === right.presencePenalty &&
+    left.maxTokens === right.maxTokens &&
+    left.systemPrompt === right.systemPrompt
+  );
+}
+
+export function getPresetOwnedParams(
+  params: InferenceParams,
+): PresetOwnedParams {
+  return {
+    temperature: params.temperature,
+    topP: params.topP,
+    topK: params.topK,
+    minP: params.minP,
+    repetitionPenalty: params.repetitionPenalty,
+    presencePenalty: params.presencePenalty,
+    maxTokens: params.maxTokens,
+    systemPrompt: params.systemPrompt,
+  };
+}
+
+export function getPresetOwnedConfigKey(params: InferenceParams): string {
+  return JSON.stringify(getPresetOwnedParams(params));
+}
+
+export function toPresetParams(params: InferenceParams): InferenceParams {
+  return {
+    ...defaultInferenceParams,
+    ...getPresetOwnedParams(params),
+  };
+}
+
+export function applyPresetParams(
+  current: InferenceParams,
+  preset: InferenceParams,
+): InferenceParams {
+  return {
+    ...current,
+    ...getPresetOwnedParams(preset),
+  };
+}
+
+export type PresetSaveMode =
+  | "disabled"
+  | "overwrite-active"
+  | "overwrite-other"
+  | "copy-builtin"
+  | "create";
+
+export interface PresetSaveState {
+  mode: PresetSaveMode;
+  canSubmit: boolean;
+  isSaveReady: boolean;
+  buttonLabel: string;
+  title: string;
+}
+
+export function getPresetSaveState({
+  rawName,
+  activePreset,
+  presets,
+  hasUnsavedPresetChanges,
+}: {
+  rawName: string;
+  activePreset: string;
+  presets: Preset[];
+  hasUnsavedPresetChanges: boolean;
+}): PresetSaveState {
+  const trimmedName = rawName.trim();
+  if (!trimmedName) {
+    return {
+      mode: "disabled",
+      canSubmit: false,
+      isSaveReady: false,
+      buttonLabel: "Save",
+      title: "Enter a preset name",
+    };
+  }
+
+  if (BUILTIN_PRESET_NAMES.has(trimmedName)) {
+    const variantName = getBuiltinVariantName(trimmedName, new Set(presets.map((preset) => preset.name)));
+    return {
+      mode: "copy-builtin",
+      canSubmit: activePreset !== trimmedName || hasUnsavedPresetChanges,
+      isSaveReady: activePreset !== trimmedName || hasUnsavedPresetChanges,
+      buttonLabel:
+        activePreset === trimmedName && !hasUnsavedPresetChanges
+          ? "Saved"
+          : "Save",
+      title:
+        activePreset === trimmedName && !hasUnsavedPresetChanges
+          ? "No unsaved changes"
+          : `Save current settings as "${variantName}"`,
+    };
+  }
+
+  const matchingPreset = presets.find((preset) => preset.name === trimmedName);
+  if (matchingPreset) {
+    const isActiveMatch = matchingPreset.name === activePreset;
+    return {
+      mode: isActiveMatch ? "overwrite-active" : "overwrite-other",
+      canSubmit: !isActiveMatch || hasUnsavedPresetChanges,
+      isSaveReady: !isActiveMatch || hasUnsavedPresetChanges,
+      buttonLabel:
+        isActiveMatch && !hasUnsavedPresetChanges ? "Saved" : "Save",
+      title: isActiveMatch
+        ? hasUnsavedPresetChanges
+          ? "Save current settings to this preset"
+          : "No unsaved changes"
+        : `Overwrite preset "${trimmedName}"`,
+    };
+  }
+
+  return {
+    mode: "create",
+    canSubmit: true,
+    isSaveReady: true,
+    buttonLabel: "Save",
+    title: `Save current settings as "${trimmedName}"`,
+  };
+}
+
+function toFiniteNumber(value: unknown): number | undefined {
+  if (typeof value !== "number" || !Number.isFinite(value)) {
+    return undefined;
+  }
+  return value;
+}
+
+interface BackendInferenceDefaults {
+  temperature?: number;
+  top_p?: number;
+  top_k?: number;
+  min_p?: number;
+  presence_penalty?: number;
+  trust_remote_code?: boolean;
+}
+
+export interface BackendInferenceEnvelope {
+  is_gguf?: boolean;
+  context_length?: number | null;
+  inference?: BackendInferenceDefaults;
+}
+
+export function mergeBackendRecommendedInference({
+  current,
+  response,
+  modelId,
+  presetSource,
+}: {
+  current: InferenceParams;
+  response: BackendInferenceEnvelope;
+  modelId: string;
+  presetSource: ChatPresetSource;
+}): InferenceParams {
+  const inference = response.inference;
+  const next: InferenceParams = {
+    ...current,
+    checkpoint: modelId,
+    trustRemoteCode:
+      typeof inference?.trust_remote_code === "boolean"
+        ? inference.trust_remote_code
+        : current.trustRemoteCode,
+  };
+
+  if (presetSource !== "builtin-default") {
+    return next;
+  }
+
+  const defaultMaxTokens = response.is_gguf
+    ? (response.context_length ?? current.maxTokens)
+    : 4096;
+  return {
+    ...next,
+    maxTokens: defaultMaxTokens,
+    temperature:
+      toFiniteNumber(inference?.temperature) ?? defaultInferenceParams.temperature,
+    topP: toFiniteNumber(inference?.top_p) ?? defaultInferenceParams.topP,
+    topK: toFiniteNumber(inference?.top_k) ?? defaultInferenceParams.topK,
+    minP: toFiniteNumber(inference?.min_p) ?? defaultInferenceParams.minP,
+    presencePenalty:
+      toFiniteNumber(inference?.presence_penalty) ??
+      defaultInferenceParams.presencePenalty,
+  };
+}
+
+export function resolveLoadMaxSeqLength({
+  modelId,
+  ggufVariant,
+  customContextLength,
+  ggufContextLength,
+  currentCheckpoint,
+  activeGgufVariant,
+  maxSeqLength,
+  presetSource,
+}: {
+  modelId: string;
+  ggufVariant?: string | null;
+  customContextLength: number | null;
+  ggufContextLength: number | null;
+  currentCheckpoint: string;
+  activeGgufVariant?: string | null;
+  maxSeqLength: number;
+  presetSource: ChatPresetSource;
+}): number {
+  const isDirectGgufFile = modelId.toLowerCase().endsWith(".gguf");
+  const isGgufLoad = ggufVariant != null || isDirectGgufFile;
+  const isReloadingCurrentGguf =
+    isGgufLoad &&
+    currentCheckpoint === modelId &&
+    (ggufVariant ?? null) === (activeGgufVariant ?? null);
+
+  if (customContextLength != null) return customContextLength;
+  if (isGgufLoad && presetSource === "builtin-default") return 0;
+  if (isReloadingCurrentGguf) return ggufContextLength ?? 0;
+  if (isGgufLoad) return 0;
+  return maxSeqLength;
+}
diff --git a/studio/frontend/src/features/chat/stores/chat-runtime-store.ts b/studio/frontend/src/features/chat/stores/chat-runtime-store.ts
index ce69d8f6dc..13f2a23c36 100644
--- a/studio/frontend/src/features/chat/stores/chat-runtime-store.ts
+++ b/studio/frontend/src/features/chat/stores/chat-runtime-store.ts
@@ -9,6 +9,10 @@ import {
   type ChatModelSummary,
   type InferenceParams,
 } from "../types/runtime";
+import {
+  getPresetSource,
+  type ChatPresetSource,
+} from "../presets/preset-policy";
 
 const AUTO_TITLE_KEY = "unsloth_chat_auto_title";
 const AUTO_HEAL_TOOL_CALLS_KEY = "unsloth_auto_heal_tool_calls";
@@ -16,6 +20,8 @@ const MAX_TOOL_CALLS_KEY = "unsloth_max_tool_calls_per_message";
 const TOOL_CALL_TIMEOUT_KEY = "unsloth_tool_call_timeout";
 const HF_TOKEN_KEY = "unsloth_hf_token";
 const INFERENCE_PARAMS_KEY = "unsloth_chat_inference_params";
+const CHAT_ACTIVE_PRESET_KEY = "unsloth_chat_active_preset";
+const CHAT_ACTIVE_PRESET_SOURCE_KEY = "unsloth_chat_active_preset_source";
 const REASONING_EFFORT_KEY = "unsloth_reasoning_effort";
 const PRESERVE_THINKING_KEY = "unsloth_preserve_thinking";
 
@@ -157,8 +163,24 @@ function saveInferenceParams(params: InferenceParams): boolean {
   }
 }
 
+function loadPresetSource(): ChatPresetSource {
+  const activePreset = loadString(CHAT_ACTIVE_PRESET_KEY, "Default");
+  if (canUseStorage()) {
+    try {
+      const raw = localStorage.getItem(CHAT_ACTIVE_PRESET_SOURCE_KEY);
+      if (raw === "modified") {
+        return "modified";
+      }
+    } catch {
+      // ignore
+    }
+  }
+  return getPresetSource(activePreset);
+}
+
 type ChatRuntimeStore = {
   params: InferenceParams;
+  activePresetSource: ChatPresetSource;
   models: ChatModelSummary[];
   loras: ChatLoraSummary[];
   runningByThreadId: Record<string, boolean>;
@@ -207,6 +229,7 @@ type ChatRuntimeStore = {
   setModelLoading: (loading: boolean) => void;
   setModelRequiresTrustRemoteCode: (required: boolean) => void;
   setParams: (params: InferenceParams) => void;
+  setActivePresetSource: (source: ChatPresetSource) => void;
   setModels: (models: ChatModelSummary[]) => void;
   setLoras: (loras: ChatLoraSummary[]) => void;
   setThreadRunning: (threadId: string, running: boolean) => void;
@@ -241,6 +264,7 @@ type ChatRuntimeStore = {
 
 export const useChatRuntimeStore = create<ChatRuntimeStore>((set) => ({
   params: loadInferenceParams(),
+  activePresetSource: loadPresetSource(),
   models: [],
   loras: [],
   runningByThreadId: {},
@@ -296,6 +320,11 @@ export const useChatRuntimeStore = create<ChatRuntimeStore>((set) => ({
       }
       return { params };
     }),
+  setActivePresetSource: (activePresetSource) =>
+    set(() => {
+      saveString(CHAT_ACTIVE_PRESET_SOURCE_KEY, activePresetSource);
+      return { activePresetSource };
+    }),
   setModels: (models) => set({ models }),
   setLoras: (loras) => set({ loras }),
   setThreadRunning: (threadId, running) =>
diff --git a/tests/studio/test_chat_preset_builtin_invariants.py b/tests/studio/test_chat_preset_builtin_invariants.py
new file mode 100644
index 0000000000..da5f21099e
--- /dev/null
+++ b/tests/studio/test_chat_preset_builtin_invariants.py
@@ -0,0 +1,272 @@
+import json
+import os
+import shutil
+import subprocess
+import textwrap
+from pathlib import Path
+
+import pytest
+
+WORKDIR = Path(__file__).resolve().parents[2]
+PRESET_POLICY = (
+    WORKDIR / "unsloth_repo/studio/frontend/src/features/chat/presets/preset-policy.ts"
+)
+RUNTIME_TYPES = (
+    WORKDIR / "unsloth_repo/studio/frontend/src/features/chat/types/runtime.ts"
+)
+TEMP = WORKDIR / "temp" / "chat_preset_builtin_invariants"
+
+
+def _require_node():
+    if shutil.which("node") is None:
+        pytest.skip("node not available")
+    if not PRESET_POLICY.exists() or not RUNTIME_TYPES.exists():
+        pytest.skip("studio chat sources not present")
+
+
+def _ensure_harness():
+    TEMP.mkdir(parents = True, exist_ok = True)
+    (TEMP / "register.mjs").write_text(
+        "import { register } from 'node:module';\n"
+        "register('./loader.mjs', import.meta.url);\n"
+    )
+    (TEMP / "loader.mjs").write_text(
+        "export function resolve(specifier, context, next) {\n"
+        "  if (specifier.endsWith('/types/runtime')) return next(specifier + '.ts', context);\n"
+        "  return next(specifier, context);\n"
+        "}\n"
+    )
+
+
+def _run(script: str):
+    _require_node()
+    _ensure_harness()
+    script_path = TEMP / "run.mts"
+    script_path.write_text(script)
+    env = dict(os.environ, NODE_NO_WARNINGS = "1")
+    result = subprocess.run(
+        [
+            "node",
+            "--experimental-strip-types",
+            "--import=./register.mjs",
+            "--no-warnings",
+            "run.mts",
+        ],
+        cwd = str(TEMP),
+        capture_output = True,
+        text = True,
+        timeout = 30,
+        env = env,
+    )
+    assert result.returncode == 0, f"stderr: {result.stderr}\nstdout: {result.stdout}"
+    last = [line for line in result.stdout.strip().splitlines() if line.strip()][-1]
+    return json.loads(last)
+
+
+def _policy_path():
+    return os.path.relpath(PRESET_POLICY, TEMP).replace("\\", "/")
+
+
+def _runtime_path():
+    return os.path.relpath(RUNTIME_TYPES, TEMP).replace("\\", "/")
+
+
+def test_default_builtin_matches_default_inference_params():
+    out = _run(
+        textwrap.dedent(
+            f"""
+            // @ts-nocheck
+            import {{ BUILTIN_PRESETS, isSamePresetConfig }} from "{_policy_path()}";
+            import {{ DEFAULT_INFERENCE_PARAMS }} from "{_runtime_path()}";
+            const def = BUILTIN_PRESETS.find((p) => p.name === "Default");
+            console.log(JSON.stringify({{
+                found: !!def,
+                matches: def ? isSamePresetConfig(def.params, DEFAULT_INFERENCE_PARAMS) : null,
+            }}));
+            """
+        )
+    )
+    assert out["found"] is True
+    assert out["matches"] is True
+
+
+def test_is_same_preset_config_detects_temperature_edit():
+    out = _run(
+        textwrap.dedent(
+            f"""
+            // @ts-nocheck
+            import {{ BUILTIN_PRESETS, isSamePresetConfig }} from "{_policy_path()}";
+            const def = BUILTIN_PRESETS.find((p) => p.name === "Default");
+            const edited = {{ ...def.params, temperature: def.params.temperature + 0.1 }};
+            console.log(JSON.stringify({{ same: isSamePresetConfig(def.params, edited) }}));
+            """
+        )
+    )
+    assert out["same"] is False
+
+
+def test_is_same_preset_config_detects_system_prompt_edit():
+    out = _run(
+        textwrap.dedent(
+            f"""
+            // @ts-nocheck
+            import {{ BUILTIN_PRESETS, isSamePresetConfig }} from "{_policy_path()}";
+            const def = BUILTIN_PRESETS.find((p) => p.name === "Default");
+            const edited = {{ ...def.params, systemPrompt: "you are a pirate" }};
+            console.log(JSON.stringify({{ same: isSamePresetConfig(def.params, edited) }}));
+            """
+        )
+    )
+    assert out["same"] is False
+
+
+def test_is_same_preset_config_ignores_checkpoint_difference():
+    out = _run(
+        textwrap.dedent(
+            f"""
+            // @ts-nocheck
+            import {{ BUILTIN_PRESETS, isSamePresetConfig }} from "{_policy_path()}";
+            const def = BUILTIN_PRESETS.find((p) => p.name === "Default");
+            const withCheckpoint = {{ ...def.params, checkpoint: "meta-llama/Llama-3-8B" }};
+            console.log(JSON.stringify({{ same: isSamePresetConfig(def.params, withCheckpoint) }}));
+            """
+        )
+    )
+    assert out["same"] is True
+
+
+def test_is_same_preset_config_ignores_model_owned_fields():
+    out = _run(
+        textwrap.dedent(
+            f"""
+            // @ts-nocheck
+            import {{ BUILTIN_PRESETS, isSamePresetConfig }} from "{_policy_path()}";
+            const def = BUILTIN_PRESETS.find((p) => p.name === "Default");
+            const edited = {{
+                ...def.params,
+                maxSeqLength: def.params.maxSeqLength + 1024,
+                trustRemoteCode: !def.params.trustRemoteCode,
+            }};
+            console.log(JSON.stringify({{ same: isSamePresetConfig(def.params, edited) }}));
+            """
+        )
+    )
+    assert out["same"] is True
+
+
+def test_preset_owned_config_key_ignores_model_owned_fields():
+    out = _run(
+        textwrap.dedent(
+            f"""
+            // @ts-nocheck
+            import {{ BUILTIN_PRESETS, getPresetOwnedConfigKey }} from "{_policy_path()}";
+            const def = BUILTIN_PRESETS.find((p) => p.name === "Default");
+            const edited = {{
+                ...def.params,
+                checkpoint: "foo/bar",
+                maxSeqLength: def.params.maxSeqLength + 1024,
+                trustRemoteCode: !def.params.trustRemoteCode,
+            }};
+            console.log(JSON.stringify({{
+                same: getPresetOwnedConfigKey(def.params) === getPresetOwnedConfigKey(edited),
+            }}));
+            """
+        )
+    )
+    assert out["same"] is True
+
+
+def test_to_preset_params_strips_model_owned_fields():
+    out = _run(
+        textwrap.dedent(
+            f"""
+            // @ts-nocheck
+            import {{ toPresetParams }} from "{_policy_path()}";
+            const sanitized = toPresetParams({{
+                temperature: 0.9,
+                topP: 0.8,
+                topK: 40,
+                minP: 0.05,
+                repetitionPenalty: 1.1,
+                presencePenalty: 0.4,
+                maxSeqLength: 16384,
+                maxTokens: 2048,
+                systemPrompt: "hello",
+                checkpoint: "foo/bar",
+                trustRemoteCode: true,
+            }});
+            console.log(JSON.stringify({{
+                checkpoint: sanitized.checkpoint,
+                trustRemoteCode: sanitized.trustRemoteCode,
+                maxSeqLength: sanitized.maxSeqLength,
+                maxTokens: sanitized.maxTokens,
+                systemPrompt: sanitized.systemPrompt,
+            }}));
+            """
+        )
+    )
+    assert out["checkpoint"] == ""
+    assert out["trustRemoteCode"] is False
+    assert out["maxSeqLength"] == 4096
+    assert out["maxTokens"] == 2048
+    assert out["systemPrompt"] == "hello"
+
+
+def test_apply_preset_params_preserves_model_owned_fields():
+    out = _run(
+        textwrap.dedent(
+            f"""
+            // @ts-nocheck
+            import {{ BUILTIN_PRESETS, applyPresetParams }} from "{_policy_path()}";
+            const creative = BUILTIN_PRESETS.find((p) => p.name === "Creative");
+            const applied = applyPresetParams(
+                {{
+                    temperature: 0.6,
+                    topP: 0.95,
+                    topK: 20,
+                    minP: 0.01,
+                    repetitionPenalty: 1.0,
+                    presencePenalty: 0.0,
+                    maxSeqLength: 16384,
+                    maxTokens: 8192,
+                    systemPrompt: "keep me?",
+                    checkpoint: "foo/bar",
+                    trustRemoteCode: true,
+                }},
+                creative.params,
+            );
+            console.log(JSON.stringify({{
+                checkpoint: applied.checkpoint,
+                trustRemoteCode: applied.trustRemoteCode,
+                maxSeqLength: applied.maxSeqLength,
+                temperature: applied.temperature,
+                topK: applied.topK,
+            }}));
+            """
+        )
+    )
+    assert out["checkpoint"] == "foo/bar"
+    assert out["trustRemoteCode"] is True
+    assert out["maxSeqLength"] == 16384
+    assert out["temperature"] == 1.5
+    assert out["topK"] == 0
+
+
+def test_creative_and_precise_builtins_differ_from_default():
+    out = _run(
+        textwrap.dedent(
+            f"""
+            // @ts-nocheck
+            import {{ BUILTIN_PRESETS, isSamePresetConfig }} from "{_policy_path()}";
+            const def = BUILTIN_PRESETS.find((p) => p.name === "Default");
+            const creative = BUILTIN_PRESETS.find((p) => p.name === "Creative");
+            const precise = BUILTIN_PRESETS.find((p) => p.name === "Precise");
+            console.log(JSON.stringify({{
+                creativeDiffers: !isSamePresetConfig(def.params, creative.params),
+                preciseDiffers: !isSamePresetConfig(def.params, precise.params),
+            }}));
+            """
+        )
+    )
+    assert out["creativeDiffers"] is True
+    assert out["preciseDiffers"] is True

From ff759ba7e4d852c288723dd96e0a1fb7e935db37 Mon Sep 17 00:00:00 2001
From: Lee Jackson <130007945+Imagineer99@users.noreply.github.com>
Date: Tue, 28 Apr 2026 22:49:13 +0100
Subject: [PATCH 176/216] Studio: Fix image-only chat requests failing
 validation (#5212)

* fix: allow image-only chat messages

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

* test: deduplicate empty content validation coverage

---------

Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com>
---
 studio/backend/models/inference.py            | 11 ++++++-----
 .../tests/test_openai_tool_passthrough.py     | 19 ++++++++++++++-----
 2 files changed, 20 insertions(+), 10 deletions(-)

diff --git a/studio/backend/models/inference.py b/studio/backend/models/inference.py
index bf0177efbf..eb2bb9c5ce 100644
--- a/studio/backend/models/inference.py
+++ b/studio/backend/models/inference.py
@@ -396,7 +396,10 @@ class ChatMessage(BaseModel):
         if self.name is not None and self.role != "tool":
             raise ValueError('"name" is only valid on role="tool" messages.')
 
-        # Per-role content requirements.
+        # Per-role content requirements. OpenAI-compatible clients may send
+        # ``content=""`` for image-only turns when the image travels in a
+        # companion field such as Studio's ``image_base64`` extension, so treat
+        # empty strings as present content for user/system messages.
         if self.role == "tool":
             if not self.tool_call_id:
                 raise ValueError(
@@ -411,10 +414,8 @@ class ChatMessage(BaseModel):
                     'role="assistant" messages require either "content" or "tool_calls".'
                 )
         else:  # "user" | "system"
-            if not self.content:
-                raise ValueError(
-                    f'role="{self.role}" messages require non-empty "content".'
-                )
+            if self.content is None or self.content == []:
+                raise ValueError(f'role="{self.role}" messages require "content".')
         return self
 
 
diff --git a/studio/backend/tests/test_openai_tool_passthrough.py b/studio/backend/tests/test_openai_tool_passthrough.py
index ccb0dba325..cdb7f5d270 100644
--- a/studio/backend/tests/test_openai_tool_passthrough.py
+++ b/studio/backend/tests/test_openai_tool_passthrough.py
@@ -144,13 +144,14 @@ class TestChatMessageToolRoles:
 
     # ── Role-aware content requirements ────────────────────────────
 
-    def test_user_empty_content_rejected(self):
-        with pytest.raises(ValidationError):
-            ChatMessage(role = "user", content = "")
+    @pytest.mark.parametrize("role", ["user", "system"])
+    def test_empty_string_content_allowed(self, role):
+        msg = ChatMessage(role = role, content = "")
+        assert msg.content == ""
 
-    def test_system_empty_content_rejected(self):
+    def test_user_missing_content_rejected(self):
         with pytest.raises(ValidationError):
-            ChatMessage(role = "system", content = "")
+            ChatMessage(role = "user")
 
     def test_user_empty_list_content_rejected(self):
         with pytest.raises(ValidationError):
@@ -226,6 +227,14 @@ class TestChatCompletionRequestToolFields:
         assert len(req.tools) == 1
         assert req.tools[0]["function"]["name"] == "get_weather"
 
+    def test_image_base64_allows_empty_user_text(self):
+        req = ChatCompletionRequest(
+            messages = [{"role": "user", "content": ""}],
+            image_base64 = "aW1hZ2U=",
+        )
+        assert req.messages[0].content == ""
+        assert req.image_base64 == "aW1hZ2U="
+
     def test_tool_choice_string_auto(self):
         assert self._make(tool_choice = "auto").tool_choice == "auto"
 

From a5615426a56857983909b05016b3d985710e0015 Mon Sep 17 00:00:00 2001
From: Daniel Han <danielhanchen@gmail.com>
Date: Tue, 28 Apr 2026 22:43:44 -0700
Subject: [PATCH 177/216] Studio: fix 7 failing studio_unit_tests on main
 (#5216)

* Studio: fix 4 failing studio_unit_tests on main

Three of the failing tests had drifted from production:

1. test_health_response_reports_desktop_capability_fields stubbed
   `routes` with a SimpleNamespace that omitted `inference_studio_router`,
   so importing studio.backend.main raised ImportError. Add the missing
   router stub.

2. test_local_recipe_token_preserves_desktop_marker and
   test_local_recipe_token_keeps_web_marker_absent decoded the local
   provider's api_key as a JWT, but _inject_local_providers now mints
   a unified sk-unsloth-* internal API key (not a forwarded JWT), so
   jwt.decode raised "Not enough segments". Renamed and rewrote both
   tests to validate the API-key contract: starts with
   storage.API_KEY_PREFIX and authenticates via get_current_subject as
   the real admin user. The web vs desktop distinction is irrelevant
   at this layer because the unified API-key path does not carry
   session flags.

The fourth failure was a real production bug:

3. test_github_validate_skips_live_access_with_honest_note expected
   github-seed validation to return valid=True per
   _GITHUB_VALIDATE_NOTE ("GitHub access and rate limits are checked
   when the run starts"). The validate route called
   build_config_builder which lazy-imports the optional data_designer
   module; when it is missing, the bare except blocked the recipe.
   Catch ImportError specifically and treat it as a deferred check,
   matching the documented intent.

Verified all 4 tests pass and the rest of studio/backend/tests still
pass (608 total, with the only remaining failures being environment
specific: 4 GPU-aware tests on a no-GPU host and 1 Anthropic-API
smoke test, both unrelated).

* Studio: fix 3 test_gpu_selection route tests after load_model signature change

`routes/inference.load_model` gained a `fastapi_request: Request`
positional argument (used to read `app.state.llama_parallel_slots`
inside the GGUF path), but the three TestRouteErrors cases that
exercise the early validation path were not updated and failed with
`TypeError: load_model() missing 1 required positional argument:
'fastapi_request'`.

Pass a SimpleNamespace mock that satisfies the attribute path the
production code reads. The validation under test fires before the
mock is consumed, but supplying the realistic shape protects against
regressions if the validation order changes.

Affected tests:
- test_inference_route_rejects_gpu_ids_for_gguf
- test_inference_route_returns_400_for_invalid_gpu_ids
- test_inference_route_returns_400_for_uuid_parent_visibility_gpu_ids

* Studio: address review feedback on validate.py ImportError handling

Two reviewers flagged the ImportError bypass added in b0d33cf:

- chatgpt-codex-connector[bot]: catching bare ImportError marks recipes
  as valid even when build_config_builder fails for unrelated import
  problems (broken internal imports, missing transitive deps after a
  version bump), hiding real regressions until run start.
- gemini-code-assist[bot]: silent pass discourages troubleshooting;
  the deferred-validation case should be logged at debug level.

Tighten the bypass to ModuleNotFoundError where the missing module name
starts with "data_designer". Other ImportErrors propagate to the outer
handler and surface as validation failures, restoring the visibility
the reviewers asked for. Add a debug-level log entry that names the
missing module so operators can trace why validation deferred.
---
 studio/backend/routes/data_recipe/validate.py | 18 ++++++++++
 studio/backend/tests/test_desktop_auth.py     | 33 ++++++++++---------
 studio/backend/tests/test_gpu_selection.py    | 30 +++++++++++++++--
 3 files changed, 62 insertions(+), 19 deletions(-)

diff --git a/studio/backend/routes/data_recipe/validate.py b/studio/backend/routes/data_recipe/validate.py
index 87eef939b4..e794d68e54 100644
--- a/studio/backend/routes/data_recipe/validate.py
+++ b/studio/backend/routes/data_recipe/validate.py
@@ -14,8 +14,10 @@ from core.data_recipe.service import (
     create_data_designer,
     validate_recipe,
 )
+from loggers import get_logger
 from models.data_recipe import RecipePayload, ValidateError, ValidateResponse
 
+logger = get_logger(__name__)
 router = APIRouter()
 
 _GITHUB_VALIDATE_NOTE = "Recipe shape is valid. GitHub access and rate limits are checked when the run starts."
@@ -151,6 +153,22 @@ def validate(payload: RecipePayload) -> ValidateResponse:
             return ValidateResponse(valid = False, errors = static_errors)
         try:
             build_config_builder(recipe)
+        except ModuleNotFoundError as exc:
+            # data_designer is an optional runtime dep. Static validation
+            # already passed; live access + full config validation are
+            # deferred to run start (per _GITHUB_VALIDATE_NOTE), so a missing
+            # optional import at validate time should not block the recipe.
+            # Restrict the bypass to the data_designer module specifically so
+            # other ImportErrors (e.g. broken internal imports or missing
+            # transitive deps after a package upgrade) still surface as
+            # validation failures instead of being silently swallowed.
+            if not (exc.name or "").startswith("data_designer"):
+                raise
+            logger.debug(
+                "data_designer not installed; deferring full config "
+                "validation to run start",
+                missing_module = exc.name,
+            )
         except Exception as exc:
             detail = str(exc).strip() or "Validation failed."
             return ValidateResponse(
diff --git a/studio/backend/tests/test_desktop_auth.py b/studio/backend/tests/test_desktop_auth.py
index c8cf1c7081..a5508c1c8b 100644
--- a/studio/backend/tests/test_desktop_auth.py
+++ b/studio/backend/tests/test_desktop_auth.py
@@ -246,7 +246,10 @@ def test_desktop_session_uses_real_admin_identity_for_api_keys():
     assert [row["name"] for row in rows] == ["desktop"]
 
 
-def test_local_recipe_token_preserves_desktop_marker(loaded_local_model):
+def test_local_recipe_token_authenticates_as_admin_for_desktop_user(loaded_local_model):
+    # _inject_local_providers mints an internal sk-unsloth-* API key (not a
+    # forwarded JWT). The unified API-key path validates as the real admin
+    # user regardless of whether the incoming session was desktop or web.
     from auth.authentication import create_access_token, get_current_subject
 
     seed_user(must_change_password = True)
@@ -260,13 +263,7 @@ def test_local_recipe_token_preserves_desktop_marker(loaded_local_model):
     jobs_route._inject_local_providers(recipe, local_recipe_request(incoming_token))
 
     local_token = recipe["model_providers"][0]["api_key"]
-    payload = jwt.decode(
-        local_token,
-        storage.get_jwt_secret(storage.DEFAULT_ADMIN_USERNAME),
-        algorithms = ["HS256"],
-    )
-    assert payload["sub"] == storage.DEFAULT_ADMIN_USERNAME
-    assert payload["desktop"] is True
+    assert local_token.startswith(storage.API_KEY_PREFIX)
     credentials = HTTPAuthorizationCredentials(
         scheme = "Bearer",
         credentials = local_token,
@@ -276,8 +273,10 @@ def test_local_recipe_token_preserves_desktop_marker(loaded_local_model):
     )
 
 
-def test_local_recipe_token_keeps_web_marker_absent(loaded_local_model):
-    from auth.authentication import create_access_token
+def test_local_recipe_token_authenticates_as_admin_for_web_user(loaded_local_model):
+    # Mirror of the desktop variant: API-key issuance is identical for web
+    # and desktop incoming tokens; auth via get_current_subject works the same.
+    from auth.authentication import create_access_token, get_current_subject
 
     seed_user(must_change_password = False)
     jobs_route = data_recipe_jobs_module()
@@ -287,13 +286,14 @@ def test_local_recipe_token_keeps_web_marker_absent(loaded_local_model):
     jobs_route._inject_local_providers(recipe, local_recipe_request(incoming_token))
 
     local_token = recipe["model_providers"][0]["api_key"]
-    payload = jwt.decode(
-        local_token,
-        storage.get_jwt_secret(storage.DEFAULT_ADMIN_USERNAME),
-        algorithms = ["HS256"],
+    assert local_token.startswith(storage.API_KEY_PREFIX)
+    credentials = HTTPAuthorizationCredentials(
+        scheme = "Bearer",
+        credentials = local_token,
+    )
+    assert (
+        asyncio.run(get_current_subject(credentials)) == storage.DEFAULT_ADMIN_USERNAME
     )
-    assert payload["sub"] == storage.DEFAULT_ADMIN_USERNAME
-    assert "desktop" not in payload
 
 
 def test_desktop_login_rejects_invalid_secret():
@@ -381,6 +381,7 @@ def test_health_response_reports_desktop_capability_fields(monkeypatch):
         datasets_router = APIRouter(),
         export_router = APIRouter(),
         inference_router = APIRouter(),
+        inference_studio_router = APIRouter(),
         models_router = APIRouter(),
         training_history_router = APIRouter(),
         training_router = APIRouter(),
diff --git a/studio/backend/tests/test_gpu_selection.py b/studio/backend/tests/test_gpu_selection.py
index c6f26037af..fb4f38565b 100644
--- a/studio/backend/tests/test_gpu_selection.py
+++ b/studio/backend/tests/test_gpu_selection.py
@@ -746,7 +746,15 @@ class TestRouteErrors(unittest.TestCase):
         ):
             with self.assertRaises(HTTPException) as exc_info:
                 asyncio.run(
-                    inference_route.load_model(request, current_subject = "test-user")
+                    inference_route.load_model(
+                        request,
+                        SimpleNamespace(
+                            app = SimpleNamespace(
+                                state = SimpleNamespace(llama_parallel_slots = 1),
+                            ),
+                        ),
+                        current_subject = "test-user",
+                    )
                 )
 
         self.assertEqual(exc_info.exception.status_code, 400)
@@ -886,7 +894,15 @@ class TestRouteErrors(unittest.TestCase):
         ):
             with self.assertRaises(HTTPException) as exc_info:
                 asyncio.run(
-                    inference_route.load_model(request, current_subject = "test-user")
+                    inference_route.load_model(
+                        request,
+                        SimpleNamespace(
+                            app = SimpleNamespace(
+                                state = SimpleNamespace(llama_parallel_slots = 1),
+                            ),
+                        ),
+                        current_subject = "test-user",
+                    )
                 )
 
         self.assertEqual(exc_info.exception.status_code, 400)
@@ -942,7 +958,15 @@ class TestRouteErrors(unittest.TestCase):
         ):
             with self.assertRaises(HTTPException) as exc_info:
                 asyncio.run(
-                    inference_route.load_model(request, current_subject = "test-user")
+                    inference_route.load_model(
+                        request,
+                        SimpleNamespace(
+                            app = SimpleNamespace(
+                                state = SimpleNamespace(llama_parallel_slots = 1),
+                            ),
+                        ),
+                        current_subject = "test-user",
+                    )
                 )
 
         self.assertEqual(exc_info.exception.status_code, 400)

From c4597298be34d2af88c05c1a6e2dae3070819326 Mon Sep 17 00:00:00 2001
From: Datta Nimmaturi <venkatadattasainimmaturi@gmail.com>
Date: Wed, 29 Apr 2026 15:20:49 +0530
Subject: [PATCH 178/216] Patch checkpoint reload init functions to strip
 unsupported args (#5167)

* Patch checkpoint reload init functions to strip unsupported args

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

* Try adding attrs back if possible

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

* import_fixes: harden peft weight converter compatibility shim

Three small fixes to patch_peft_weight_converter_compatibility:

- Restore peft_config=None default and @functools.wraps on the
  build_peft_weight_mapping wrapper so the upstream coordinated
  signature (weight_conversions, adapter_name, peft_config=None) is
  preserved. Without the default, callers using the documented
  two-argument form raise TypeError after import unsloth.
- Serialize the temporary class-init patch/restore behind a
  threading.RLock. The previous unsynchronized window let two
  concurrent build_peft_weight_mapping calls (e.g. dynamic LoRA
  serving) re-expose the original distributed_operation TypeError
  when one thread restored a class while another was still inside
  original_build.
- Hand _patch_weight_converter_ctors a caller-owned accumulator
  list and append in place. If signature inspection ever raises
  mid-loop, the finally block now sees the partial list and
  restores already-patched classes instead of leaving them with
  the compat init permanently installed.

* Add tests for peft weight converter compatibility shim

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

---------

Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com>
Co-authored-by: Lee Jackson <130007945+Imagineer99@users.noreply.github.com>
Co-authored-by: Daniel Han <danielhanchen@gmail.com>
---
 tests/test_peft_weight_converter_compat.py | 259 +++++++++++++++++++++
 unsloth/__init__.py                        |   3 +
 unsloth/import_fixes.py                    |  83 +++++++
 3 files changed, 345 insertions(+)
 create mode 100644 tests/test_peft_weight_converter_compat.py

diff --git a/tests/test_peft_weight_converter_compat.py b/tests/test_peft_weight_converter_compat.py
new file mode 100644
index 0000000000..62f17f0f45
--- /dev/null
+++ b/tests/test_peft_weight_converter_compat.py
@@ -0,0 +1,259 @@
+import importlib.util
+import inspect
+import sys
+import threading
+import types
+from pathlib import Path
+
+import pytest
+
+
+REPO_ROOT = Path(__file__).resolve().parents[1]
+IMPORT_FIXES = REPO_ROOT / "unsloth" / "import_fixes.py"
+
+
+def _load_patch_function():
+    spec = importlib.util.spec_from_file_location(
+        "_unsloth_import_fixes_under_test", IMPORT_FIXES
+    )
+    module = importlib.util.module_from_spec(spec)
+    spec.loader.exec_module(module)
+    return module.patch_peft_weight_converter_compatibility
+
+
+def _install_fake_peft(twc_namespace):
+    peft_pkg = types.ModuleType("peft")
+    peft_pkg.__path__ = []
+    peft_utils = types.ModuleType("peft.utils")
+    peft_utils.__path__ = []
+    twc = types.ModuleType("peft.utils.transformers_weight_conversion")
+    for k, v in twc_namespace.items():
+        setattr(twc, k, v)
+    peft_utils.transformers_weight_conversion = twc
+    sys.modules["peft"] = peft_pkg
+    sys.modules["peft.utils"] = peft_utils
+    sys.modules["peft.utils.transformers_weight_conversion"] = twc
+    return twc
+
+
+@pytest.fixture(autouse = True)
+def _restore_peft_modules():
+    saved = {
+        k: sys.modules.get(k)
+        for k in (
+            "peft",
+            "peft.utils",
+            "peft.utils.transformers_weight_conversion",
+        )
+    }
+    yield
+    for k, v in saved.items():
+        if v is None:
+            sys.modules.pop(k, None)
+        else:
+            sys.modules[k] = v
+
+
+class _LegacyConverter:
+    def __init__(self, source_patterns, target_patterns, operations):
+        self.source_patterns = source_patterns
+        self.target_patterns = target_patterns
+        self.operations = operations
+        self.distributed_operation = None
+        self.quantization_operation = None
+
+
+class _ModernConverter:
+    def __init__(
+        self,
+        source_patterns,
+        target_patterns,
+        operations,
+        distributed_operation = None,
+        quantization_operation = None,
+    ):
+        self.source_patterns = source_patterns
+        self.target_patterns = target_patterns
+        self.operations = operations
+        self.distributed_operation = distributed_operation
+        self.quantization_operation = quantization_operation
+
+
+def _make_legacy_converter():
+    return _LegacyConverter(["src.*"], ["tgt.*"], [])
+
+
+def _make_modern_converter():
+    return _ModernConverter(["src.*"], ["tgt.*"], [])
+
+
+def _build_that_calls_init(weight_conversions, adapter_name, peft_config = None):
+    out = []
+    for c in weight_conversions or []:
+        out.append(
+            c.__class__(
+                source_patterns = c.source_patterns,
+                target_patterns = c.target_patterns,
+                operations = c.operations,
+                distributed_operation = "dist-x",
+                quantization_operation = "quant-y",
+            )
+        )
+    return out
+
+
+def test_two_arg_call_preserves_upstream_signature():
+    twc = _install_fake_peft({"build_peft_weight_mapping": _build_that_calls_init})
+    patch = _load_patch_function()
+    patch()
+
+    sig = inspect.signature(twc.build_peft_weight_mapping)
+    assert "peft_config" in sig.parameters
+    assert sig.parameters["peft_config"].default is None
+
+    out = twc.build_peft_weight_mapping([_make_legacy_converter()], "default")
+    assert len(out) == 1
+    assert out[0].distributed_operation == "dist-x"
+    assert out[0].quantization_operation == "quant-y"
+
+
+def test_legacy_init_succeeds_after_patch():
+    twc = _install_fake_peft({"build_peft_weight_mapping": _build_that_calls_init})
+    patch = _load_patch_function()
+    patch()
+
+    out = twc.build_peft_weight_mapping([_make_legacy_converter()], "default", None)
+    assert len(out) == 1
+    assert out[0].distributed_operation == "dist-x"
+    assert out[0].quantization_operation == "quant-y"
+
+
+def test_modern_init_not_patched():
+    twc = _install_fake_peft({"build_peft_weight_mapping": _build_that_calls_init})
+    pre_init = _ModernConverter.__init__
+    patch = _load_patch_function()
+    patch()
+
+    twc.build_peft_weight_mapping([_make_modern_converter()], "default", None)
+    assert _ModernConverter.__init__ is pre_init
+
+
+def test_class_init_restored_after_call():
+    twc = _install_fake_peft({"build_peft_weight_mapping": _build_that_calls_init})
+    pre_init = _LegacyConverter.__init__
+    patch = _load_patch_function()
+    patch()
+
+    twc.build_peft_weight_mapping([_make_legacy_converter()], "default", None)
+    assert _LegacyConverter.__init__ is pre_init
+
+
+def test_class_init_restored_after_original_build_raises():
+    def _raise(weight_conversions, adapter_name, peft_config = None):
+        raise RuntimeError("simulated PEFT failure")
+
+    twc = _install_fake_peft({"build_peft_weight_mapping": _raise})
+    pre_init = _LegacyConverter.__init__
+    patch = _load_patch_function()
+    patch()
+
+    with pytest.raises(RuntimeError):
+        twc.build_peft_weight_mapping([_make_legacy_converter()], "default", None)
+    assert _LegacyConverter.__init__ is pre_init
+
+
+def test_partial_patch_restored_when_inspect_signature_raises_mid_loop():
+    twc = _install_fake_peft({"build_peft_weight_mapping": _build_that_calls_init})
+    pre_legacy = _LegacyConverter.__init__
+
+    class _BadInitConverter:
+        def __init__(self, source_patterns, target_patterns, operations):
+            self.source_patterns = source_patterns
+            self.target_patterns = target_patterns
+            self.operations = operations
+
+    pre_bad = _BadInitConverter.__init__
+    patch = _load_patch_function()
+    patch()
+
+    real_signature = inspect.signature
+
+    def _fake_signature(callable_):
+        if callable_ is _BadInitConverter.__init__:
+            raise ValueError("inspect.signature failed mid-loop")
+        return real_signature(callable_)
+
+    inspect.signature = _fake_signature
+    try:
+        legacy = _LegacyConverter(["src.*"], ["tgt.*"], [])
+        bad = _BadInitConverter.__new__(_BadInitConverter)
+        bad.source_patterns = ["src.*"]
+        bad.target_patterns = ["tgt.*"]
+        bad.operations = []
+        with pytest.raises(ValueError):
+            twc.build_peft_weight_mapping([legacy, bad], "default", None)
+    finally:
+        inspect.signature = real_signature
+
+    assert _LegacyConverter.__init__ is pre_legacy
+    assert _BadInitConverter.__init__ is pre_bad
+
+
+def test_idempotent_install_does_not_double_wrap():
+    twc = _install_fake_peft({"build_peft_weight_mapping": _build_that_calls_init})
+    patch = _load_patch_function()
+    patch()
+    first_wrapped = twc.build_peft_weight_mapping
+    patch()
+    assert twc.build_peft_weight_mapping is first_wrapped
+
+
+def test_concurrent_legacy_calls_no_typeerror():
+    import time
+
+    def _slow_build(weight_conversions, adapter_name, peft_config = None):
+        time.sleep(0.05)
+        return _build_that_calls_init(weight_conversions, adapter_name, peft_config)
+
+    twc = _install_fake_peft({"build_peft_weight_mapping": _slow_build})
+    patch = _load_patch_function()
+    patch()
+
+    errors = []
+    results = []
+    start = threading.Event()
+
+    def _worker():
+        start.wait(timeout = 10)
+        try:
+            out = twc.build_peft_weight_mapping(
+                [_make_legacy_converter()], "default", None
+            )
+            results.append(out)
+        except Exception as e:
+            errors.append(e)
+
+    threads = [threading.Thread(target = _worker) for _ in range(8)]
+    for t in threads:
+        t.start()
+    start.set()
+    for t in threads:
+        t.join(timeout = 15)
+
+    assert errors == []
+    assert len(results) == 8
+    for out in results:
+        assert out[0].distributed_operation == "dist-x"
+        assert out[0].quantization_operation == "quant-y"
+    assert _LegacyConverter.__init__.__qualname__.startswith("_LegacyConverter")
+
+
+def test_empty_conversions_short_circuits_without_patching():
+    twc = _install_fake_peft({"build_peft_weight_mapping": _build_that_calls_init})
+    pre_init = _LegacyConverter.__init__
+    patch = _load_patch_function()
+    patch()
+
+    out = twc.build_peft_weight_mapping([], "default", None)
+    assert out == []
+    assert _LegacyConverter.__init__ is pre_init
diff --git a/unsloth/__init__.py b/unsloth/__init__.py
index 52114bb544..9db9ae0a32 100644
--- a/unsloth/__init__.py
+++ b/unsloth/__init__.py
@@ -153,6 +153,7 @@ from .import_fixes import (
     patch_torchcodec_audio_decoder,
     disable_torchcodec_if_broken,
     disable_broken_wandb,
+    patch_peft_weight_converter_compatibility,
 )
 
 fix_xformers_performance_issue()
@@ -176,6 +177,7 @@ patch_vllm_for_notebooks()
 patch_torchcodec_audio_decoder()
 disable_torchcodec_if_broken()
 disable_broken_wandb()
+patch_peft_weight_converter_compatibility()
 
 del fix_xformers_performance_issue
 del fix_vllm_aimv2_issue
@@ -197,6 +199,7 @@ del patch_vllm_for_notebooks
 del patch_torchcodec_audio_decoder
 del disable_torchcodec_if_broken
 del disable_broken_wandb
+del patch_peft_weight_converter_compatibility
 
 # Torch 2.4 has including_emulation
 if DEVICE_TYPE == "cuda":
diff --git a/unsloth/import_fixes.py b/unsloth/import_fixes.py
index 6cebad7939..f1fba0ce91 100644
--- a/unsloth/import_fixes.py
+++ b/unsloth/import_fixes.py
@@ -25,6 +25,7 @@ import textwrap
 import warnings
 import sys
 import functools
+import inspect
 
 # We cannot do from unsloth_zoo.log import logger since FBGEMM might cause seg faults.
 UNSLOTH_ENABLE_LOGGING = os.environ.get("UNSLOTH_ENABLE_LOGGING", "0") in (
@@ -1371,6 +1372,88 @@ def disable_broken_wandb():
         os.environ["WANDB_DISABLED"] = "true"
 
 
+def patch_peft_weight_converter_compatibility():
+    """Allow PEFT converter rebuilds on legacy converter constructors."""
+    try:
+        from peft.utils import transformers_weight_conversion as twc
+    except (ImportError, AttributeError):
+        return
+
+    if getattr(twc, "_unsloth_weight_converter_compat_patch", False):
+        return
+
+    import threading
+
+    original_build = twc.build_peft_weight_mapping
+    patch_lock = threading.RLock()
+
+    def _patch_weight_converter_ctors(weight_conversions, patched):
+        seen_classes = set()
+
+        for conversion in weight_conversions:
+            conversion_cls = conversion.__class__
+            if conversion_cls in seen_classes:
+                continue
+            seen_classes.add(conversion_cls)
+
+            original_init = conversion_cls.__init__
+            params = inspect.signature(original_init).parameters
+            supports_kwargs = any(
+                p.kind == inspect.Parameter.VAR_KEYWORD for p in params.values()
+            )
+            supports_distributed = "distributed_operation" in params
+            supports_quantization = "quantization_operation" in params
+            if supports_kwargs or (supports_distributed and supports_quantization):
+                continue
+
+            def _compat_init(
+                self,
+                *args,
+                __original_init = original_init,
+                __supports_distributed = supports_distributed,
+                __supports_quantization = supports_quantization,
+                **kwargs,
+            ):
+                unsupported = {}
+                if not __supports_distributed and "distributed_operation" in kwargs:
+                    unsupported["distributed_operation"] = kwargs.pop(
+                        "distributed_operation"
+                    )
+                if not __supports_quantization and "quantization_operation" in kwargs:
+                    unsupported["quantization_operation"] = kwargs.pop(
+                        "quantization_operation"
+                    )
+                result = __original_init(self, *args, **kwargs)
+                for name, value in unsupported.items():
+                    if hasattr(self, name):
+                        setattr(self, name, value)
+                return result
+
+            conversion_cls.__init__ = _compat_init
+            patched.append((conversion_cls, original_init))
+
+    @functools.wraps(original_build)
+    def _build_peft_weight_mapping_compat(
+        weight_conversions,
+        adapter_name,
+        peft_config = None,
+    ):
+        if not weight_conversions:
+            return original_build(weight_conversions, adapter_name, peft_config)
+
+        patched_classes = []
+        with patch_lock:
+            try:
+                _patch_weight_converter_ctors(weight_conversions, patched_classes)
+                return original_build(weight_conversions, adapter_name, peft_config)
+            finally:
+                for conversion_cls, original_init in patched_classes:
+                    conversion_cls.__init__ = original_init
+
+    twc.build_peft_weight_mapping = _build_peft_weight_mapping_compat
+    twc._unsloth_weight_converter_compat_patch = True
+
+
 CAUSAL_CONV1D_BROKEN = False
 _CAUSAL_CONV1D_PREFIX = "causal_conv1d"
 _CAUSAL_CONV1D_BLOCKER_SENTINEL = "_unsloth_causal_conv1d_blocker"

From 146295eeca7d6129bb527f20f1f911ad6316a3b5 Mon Sep 17 00:00:00 2001
From: Lee Jackson <130007945+Imagineer99@users.noreply.github.com>
Date: Wed, 29 Apr 2026 10:51:25 +0100
Subject: [PATCH 179/216] Studio: Fix clipped model selector text descenders
 (#5210)

* fix: clipped model selector text descenders

* Studio: Fix image-only chat requests failing validation (#5212)

* fix: allow image-only chat messages

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

* test: deduplicate empty content validation coverage

---------

Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com>

* Fix descender clipping in sidebar user account section

Replace `leading-none` with `leading-tight` on the parent div wrapping
`displayTitle` and the "Studio" label inside `SidebarMenuButton`. The
child spans use `truncate` (overflow: hidden), so `line-height: 1`
clipped descenders (g, p, q, y, j) on user names. Same root cause and
fix as the model selector trigger.

* Add tests for studio text descender clipping

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

---------

Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com>
Co-authored-by: Daniel Han <danielhanchen@gmail.com>
---
 .../frontend/src/components/app-sidebar.tsx   |  2 +-
 .../assistant-ui/model-selector.tsx           |  2 +-
 .../test_studio_text_descender_clipping.py    | 69 +++++++++++++++++++
 3 files changed, 71 insertions(+), 2 deletions(-)
 create mode 100644 tests/studio/test_studio_text_descender_clipping.py

diff --git a/studio/frontend/src/components/app-sidebar.tsx b/studio/frontend/src/components/app-sidebar.tsx
index 6d029a59bd..b1dc286df5 100644
--- a/studio/frontend/src/components/app-sidebar.tsx
+++ b/studio/frontend/src/components/app-sidebar.tsx
@@ -527,7 +527,7 @@ export function AppSidebar() {
                       className="!size-8"
                     />
                   </div>
-                  <div className="flex flex-col gap-0.5 leading-none group-data-[collapsible=icon]:hidden">
+                  <div className="flex flex-col gap-0.5 leading-tight group-data-[collapsible=icon]:hidden">
                     <span className="truncate font-heading text-[13px] tracking-[0.02em] font-semibold text-[#383835] dark:text-[#c7c7c4]">{displayTitle}</span>
                     <span className="truncate text-[11px] tracking-[0.01em] text-muted-foreground">Studio</span>
                   </div>
diff --git a/studio/frontend/src/components/assistant-ui/model-selector.tsx b/studio/frontend/src/components/assistant-ui/model-selector.tsx
index 679034c9fa..4fa39beff4 100644
--- a/studio/frontend/src/components/assistant-ui/model-selector.tsx
+++ b/studio/frontend/src/components/assistant-ui/model-selector.tsx
@@ -81,7 +81,7 @@ function ModelSelectorTrigger({
           <span className="size-2 shrink-0 rounded-full bg-emerald-500" />
         )}
         <span className="flex min-w-0 flex-1 items-baseline gap-2">
-          <span className="min-w-0 flex-1 truncate font-heading text-[16px] font-medium leading-none text-black dark:text-white">
+          <span className="min-w-0 flex-1 truncate font-heading text-[16px] font-medium leading-tight text-black dark:text-white">
             {currentModel?.name ?? "Select model"}
           </span>
           {currentModel?.description && (
diff --git a/tests/studio/test_studio_text_descender_clipping.py b/tests/studio/test_studio_text_descender_clipping.py
new file mode 100644
index 0000000000..dd0df320bc
--- /dev/null
+++ b/tests/studio/test_studio_text_descender_clipping.py
@@ -0,0 +1,69 @@
+"""
+Regression guard: descender-prone text spans in Studio must not pair
+`leading-none` with `truncate` (overflow: hidden), which clips glyph
+descenders (g, p, q, y, j) in real user-visible labels.
+"""
+
+from __future__ import annotations
+
+import re
+from pathlib import Path
+
+
+WORKDIR = Path(__file__).resolve().parents[2]
+MODEL_SELECTOR = (
+    WORKDIR
+    / "studio"
+    / "frontend"
+    / "src"
+    / "components"
+    / "assistant-ui"
+    / "model-selector.tsx"
+)
+APP_SIDEBAR = WORKDIR / "studio" / "frontend" / "src" / "components" / "app-sidebar.tsx"
+
+
+def _read(path: Path) -> str:
+    assert path.exists(), f"missing source file: {path}"
+    return path.read_text()
+
+
+def test_model_selector_trigger_label_uses_leading_tight():
+    src = _read(MODEL_SELECTOR)
+    pattern = re.compile(
+        r'<span\s+className="[^"]*\bmin-w-0\b[^"]*\bflex-1\b[^"]*\btruncate\b[^"]*\bfont-heading\b[^"]*\btext-\[16px\][^"]*"',
+    )
+    matches = pattern.findall(src)
+    assert matches, "could not find ModelSelectorTrigger model-name span"
+    for cls in matches:
+        assert "leading-tight" in cls, f"expected leading-tight, got: {cls}"
+        assert (
+            "leading-none" not in cls
+        ), f"leading-none must not coexist with truncate here: {cls}"
+
+
+def test_sidebar_account_block_uses_leading_tight():
+    src = _read(APP_SIDEBAR)
+    pattern = re.compile(
+        r'<div\s+className="flex\s+flex-col\s+gap-0\.5\s+(\S+)\s+group-data-\[collapsible=icon\]:hidden">',
+    )
+    matches = pattern.findall(src)
+    assert matches, "could not find sidebar account-block parent div"
+    leading_classes = [m for m in matches if m.startswith("leading-")]
+    assert (
+        leading_classes
+    ), f"no leading-* class on sidebar account-block parent: {matches}"
+    for cls in leading_classes:
+        assert (
+            cls == "leading-tight"
+        ), f"sidebar account-block must use leading-tight, got: {cls}"
+
+
+def test_no_truncate_plus_leading_none_in_changed_files():
+    for path in (MODEL_SELECTOR, APP_SIDEBAR):
+        src = _read(path)
+        for line in src.splitlines():
+            if "truncate" in line and "leading-none" in line:
+                raise AssertionError(
+                    f"{path.name}: same line uses truncate + leading-none, descenders will clip: {line.strip()}"
+                )

From 4f9c8321a2136e62fd86fe722a544afd534334a5 Mon Sep 17 00:00:00 2001
From: Datta Nimmaturi <venkatadattasainimmaturi@gmail.com>
Date: Wed, 29 Apr 2026 16:45:34 +0530
Subject: [PATCH 180/216] Fix DPO trainer multi process hang (#5199)

* Fix DPO trainer multi process hang

* Fix datacollator error

* further dpo vision changes

* cleanup

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

* Harden DPO vision row processing and source rewrites

- dpo_trainer_vision_signature_columns: also match TRL 0.22.x layout
  (image_sizes followed by ref_chosen_logps), so vision keys are not
  stripped via remove_unused_columns on the originally-affected version.
- dpo_trainer_concatenated_inputs: fall back to inserting after the
  image_sizes block when no token_type_ids anchor follows it.
- Apply the same vision model_kwargs forwarding rewrite to
  _compute_loss_liger via dpo_trainer_compute_loss_liger so the Liger DPO
  path does not drop pixel_position_ids/image_position_ids/
  mm_token_type_ids when args.use_liger_loss is true.
- dpo_trainer_vision_process_row:
  - guard chosen/rejected EOS append with tokenizer.eos_token_id is not None
  - use features.get("images") and features.get("prompt") to match the
    existing get on line 164 and avoid KeyError on rows without those keys
  - drop the torch.is_tensor gate so list-form pixel_position_ids/
    image_position_ids returned without return_tensors are still aliased
  - skip the loop entry for image_position_ids when it was already
    promoted to pixel_position_ids, so the output dict no longer carries
    both keys with identical data
- dpo_trainer_data_collator_vision_keys: switch from pad_sequence to
  trl.trainer.utils.pad with padding_side='left' (matches the DPO
  collator's prompt left-pad) and padding_value=-1 for *_position_ids
  keys (sentinel for padded patches), 0 otherwise. Skip the key when not
  every example carries it. Falls back to pad_sequence if trl.pad is
  unavailable or the tensor rank is too high.
- dpo_trainer_prepare_dataset: keep TRL's writer_batch_size=10 when
  popping num_proc; removing it defaults to 1000 and reintroduces the
  vision OOM risk that writer_batch_size=10 was set to avoid.

* DPO vision row: keep upstream-facing keys and fix patch padding

- dpo_trainer_vision_process_row: no longer aliases image_position_ids
  to pixel_position_ids. Each upstream-emitted vision key is forwarded
  under its own name. Gemma4 ForConditionalGeneration.forward accepts
  image_position_ids directly and renames it to pixel_position_ids only
  at the vision-tower call site, so aliasing in the row helper hid the
  kwarg the model actually consumes.
- dpo_trainer_vision_process_row: extract pixel_values via "in"
  membership instead of unconditional indexing. With the missing-images
  path returning [] to the processor, modern processors no longer emit
  a pixel_values key, and the previous indexing raised KeyError.
- dpo_trainer_data_collator_vision_keys: pick padding_side per key
  family. *_position_ids tensors are patch-aligned to pixel_values
  (TRL's DataCollatorForPreference right-pads pixel_values), so pad
  them right with the -1 sentinel; mm_token_type_ids is token-aligned
  to prompt_input_ids (left-padded by TRL), so pad it left with 0.

* DPO vision: handle multi-image prompts and arbitrary-rank collator pad

- dpo_trainer_vision_process_row: when a prompt is missing vision
  placeholders, insert one placeholder per missing image instead of
  always inserting a single token. Multi-image rows now satisfy the
  processor's token-vs-image count check rather than under-inserting
  and tripping the placeholder/feature mismatch.
- dpo_trainer_data_collator_vision_keys: drop the dim()<=2 gate around
  trl.trainer.utils.pad. trl.pad handles arbitrary rank correctly,
  while the previous fallback to torch.nn.utils.rnn.pad_sequence
  raised RuntimeError on rank-3 patch-position tensors with mismatched
  non-leading dimensions. The pad_sequence path remains as a degraded
  fallback only when trl.pad is unavailable or raises.

* DPO vision row: support scalar images and align prompt-aligned aux ids

- dpo_trainer_vision_process_row: type-aware normalization of the
  features['images'] column instead of a truthiness/len check that
  raised on single image objects (PIL.Image has no __len__) and on
  numpy ndarrays (truthiness ambiguous). Lists/tuples count as their
  length, scalar image objects count as one, None counts as zero, and
  the original value is forwarded to the processor.
- dpo_trainer_vision_process_row: when max_prompt_length truncates
  prompt_input_ids, also slice token_type_ids and mm_token_type_ids
  by the same [-max_prompt_length:] suffix. Those keys are 1:1 token
  aligned to prompt_input_ids (Gemma 4 vision attention keys off
  mm_token_type_ids per modular_gemma4.py), so leaving them at the
  original length silently misaligned the multimodal mask.

* DPO vision row: stop synthesizing vision-token placeholders

Pass features['prompt'] and features['images'] straight to the
processor without inserting any extra placeholder tokens. The previous
helper used processing_class.image_token, which is the right prompt
placeholder for Gemma 4 but the wrong one for Gemma 3 (whose prompt
placeholder is boi_token while image_token is the inner expansion
target). Synthesizing that token also broke multi-image rows: text
ended up with N placeholders while the row helper only forwarded the
first image's pixel_values via the standard [0] indexing that mirrors
upstream TRL process_row, so token vs image-feature counts diverged.
Removing the synthesis matches stock TRL behavior; users provide the
correct placeholders for their processor in the prompt.

* Add tests for DPO vision row processor passthrough

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

---------

Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com>
Co-authored-by: Lee Jackson <130007945+Imagineer99@users.noreply.github.com>
Co-authored-by: Daniel Han <danielhanchen@gmail.com>
---
 .../test_dpo_vision_processor_passthrough.py  | 149 ++++++++++
 unsloth/models/rl_replacements.py             | 272 ++++++++++++++++++
 2 files changed, 421 insertions(+)
 create mode 100644 tests/python/test_dpo_vision_processor_passthrough.py

diff --git a/tests/python/test_dpo_vision_processor_passthrough.py b/tests/python/test_dpo_vision_processor_passthrough.py
new file mode 100644
index 0000000000..a4f2e2e12a
--- /dev/null
+++ b/tests/python/test_dpo_vision_processor_passthrough.py
@@ -0,0 +1,149 @@
+"""Verify dpo_trainer_vision_process_row forwards prompt and images verbatim."""
+
+import ast
+import os
+
+import numpy as np
+
+
+REPO_ROOT = os.path.abspath(os.path.join(os.path.dirname(__file__), "..", ".."))
+RL_PATH = os.path.join(REPO_ROOT, "unsloth", "models", "rl_replacements.py")
+
+
+def _load_helpers():
+    src = open(RL_PATH).read()
+    tree = ast.parse(src)
+    import torch as _torch
+
+    ns = {"torch": _torch}
+    for node in tree.body:
+        if isinstance(node, ast.Assign) and any(
+            isinstance(t, ast.Name) and t.id == "_DPO_VISION_KEYS" for t in node.targets
+        ):
+            exec(ast.get_source_segment(src, node), ns)
+    for node in tree.body:
+        if isinstance(node, ast.FunctionDef) and node.name.startswith(
+            ("dpo_trainer_", "_dpo_trainer_")
+        ):
+            exec(ast.get_source_segment(src, node), ns)
+    return ns
+
+
+class _Tok:
+    eos_token_id = 99
+    bos_token_id = None
+
+    def __call__(self, t, add_special_tokens = False):
+        return {"input_ids": [10]}
+
+
+class _Capture:
+    image_token = "<img>"
+    boi_token = "<boi>"
+
+    def __init__(self):
+        self.tokenizer = _Tok()
+        self.last_text = None
+        self.last_images = "__sentinel__"
+
+    def __call__(self, images = None, text = None, add_special_tokens = False):
+        self.last_text = text
+        self.last_images = images
+        out = {"input_ids": [[1, 2]]}
+        if images is not None:
+            out["pixel_values"] = [object()]
+        return out
+
+
+def test_prompt_passes_through_without_image_token_synthesis():
+    ns = _load_helpers()
+    proc = _Capture()
+    ns["dpo_trainer_vision_process_row"](
+        {"prompt": "describe", "chosen": "c", "rejected": "r", "images": ["i"]},
+        proc,
+    )
+    assert proc.last_text == "describe"
+
+
+def test_prompt_with_existing_image_token_unchanged():
+    ns = _load_helpers()
+    proc = _Capture()
+    ns["dpo_trainer_vision_process_row"](
+        {"prompt": "<img> describe", "chosen": "c", "rejected": "r", "images": ["i"]},
+        proc,
+    )
+    assert proc.last_text == "<img> describe"
+
+
+def test_gemma3_style_boi_token_prompt_not_corrupted():
+    ns = _load_helpers()
+    proc = _Capture()
+    ns["dpo_trainer_vision_process_row"](
+        {"prompt": "<boi> describe", "chosen": "c", "rejected": "r", "images": ["i"]},
+        proc,
+    )
+    assert proc.last_text == "<boi> describe"
+    assert "<img>" not in proc.last_text
+
+
+def test_multi_image_prompt_unchanged_no_extra_placeholders():
+    ns = _load_helpers()
+    proc = _Capture()
+    ns["dpo_trainer_vision_process_row"](
+        {
+            "prompt": "compare",
+            "chosen": "c",
+            "rejected": "r",
+            "images": ["a", "b", "c"],
+        },
+        proc,
+    )
+    assert proc.last_text == "compare"
+
+
+def test_list_images_forwarded_verbatim():
+    ns = _load_helpers()
+    proc = _Capture()
+    payload = ["a", "b"]
+    ns["dpo_trainer_vision_process_row"](
+        {"prompt": "p", "chosen": "c", "rejected": "r", "images": payload},
+        proc,
+    )
+    assert proc.last_images is payload
+
+
+def test_single_pil_like_image_forwarded_verbatim():
+    ns = _load_helpers()
+
+    class PIL:
+        def __bool__(self):
+            return True
+
+    proc = _Capture()
+    pil = PIL()
+    ns["dpo_trainer_vision_process_row"](
+        {"prompt": "p", "chosen": "c", "rejected": "r", "images": pil},
+        proc,
+    )
+    assert proc.last_images is pil
+
+
+def test_numpy_ndarray_image_forwarded_verbatim():
+    ns = _load_helpers()
+    proc = _Capture()
+    arr = np.zeros((2, 3, 3), dtype = np.uint8)
+    ns["dpo_trainer_vision_process_row"](
+        {"prompt": "p", "chosen": "c", "rejected": "r", "images": arr},
+        proc,
+    )
+    assert proc.last_images is arr
+
+
+def test_missing_images_key_passes_none_to_processor():
+    ns = _load_helpers()
+    proc = _Capture()
+    ns["dpo_trainer_vision_process_row"](
+        {"prompt": "p", "chosen": "c", "rejected": "r"},
+        proc,
+    )
+    assert proc.last_images is None
diff --git a/unsloth/models/rl_replacements.py b/unsloth/models/rl_replacements.py
index 4d36af62cc..5d2c4151cf 100755
--- a/unsloth/models/rl_replacements.py
+++ b/unsloth/models/rl_replacements.py
@@ -55,6 +55,12 @@ RL_CONFIG_CHANGES = defaultdict(list)
 RL_METRICS_CHANGES = defaultdict(list)
 RL_ADDITIONAL_FUNCTIONS = defaultdict(list)
 
+_DPO_VISION_KEYS = (
+    "pixel_position_ids",
+    "image_position_ids",
+    "mm_token_type_ids",
+)
+
 torch_compile_options = {
     "epilogue_fusion": True,
     "max_autotune": False,  # I saw speedups, but not sure if this has issues in collab
@@ -120,6 +126,272 @@ def dpo_trainer_fix_columns(call_args, extra_args):
 RL_EXTRA_ARGS["dpo_trainer"].append(dpo_trainer_fix_columns)
 
 
+def dpo_trainer_fix_data_collator(call_args, extra_args):
+    if (
+        "data_collator" in call_args
+        and "train_dataset" in call_args
+        and "processing_class" in call_args
+    ):
+        fix_collator = (
+            "if hasattr(train_dataset, 'column_names'):\n"
+            "    column_names = set(train_dataset.column_names)\n"
+            "    is_dpo_dataset = ({'chosen', 'rejected'}.issubset(column_names) or\n"
+            "                      {'prompt_input_ids', 'chosen_input_ids', 'rejected_input_ids'}.issubset(column_names))\n"
+            "    if is_dpo_dataset and isinstance(data_collator, TransformersDataCollatorForLanguageModeling):\n"
+            "        data_collator = None\n"
+            "    del is_dpo_dataset, column_names\n"
+        )
+        return fix_collator
+    return ""
+
+
+RL_EXTRA_ARGS["dpo_trainer"].append(dpo_trainer_fix_data_collator)
+
+
+def dpo_trainer_vision_process_row(
+    features,
+    processing_class,
+    max_prompt_length = None,
+    max_completion_length = None,
+    add_special_tokens = True,
+    is_chat = False,
+):
+    text = features.get("prompt", "")
+    images = features.get("images")
+    processor, tokenizer = processing_class, processing_class.tokenizer
+    processed_features = processor(
+        images = images,
+        text = text,
+        add_special_tokens = False,
+    )
+
+    prompt_input_ids = processed_features["input_ids"][0]
+    chosen_input_ids = tokenizer(features["chosen"], add_special_tokens = False)[
+        "input_ids"
+    ]
+    rejected_input_ids = tokenizer(features["rejected"], add_special_tokens = False)[
+        "input_ids"
+    ]
+
+    if add_special_tokens:
+        if tokenizer.bos_token_id is not None:
+            prompt_input_ids = [tokenizer.bos_token_id] + prompt_input_ids
+        if tokenizer.eos_token_id is not None:
+            prompt_input_ids = prompt_input_ids + [tokenizer.eos_token_id]
+    if not is_chat and tokenizer.eos_token_id is not None:
+        chosen_input_ids = chosen_input_ids + [tokenizer.eos_token_id]
+        rejected_input_ids = rejected_input_ids + [tokenizer.eos_token_id]
+
+    if max_prompt_length is not None:
+        prompt_input_ids = prompt_input_ids[-max_prompt_length:]
+    if max_completion_length is not None:
+        chosen_input_ids = chosen_input_ids[:max_completion_length]
+        rejected_input_ids = rejected_input_ids[:max_completion_length]
+
+    output = {
+        "prompt_input_ids": prompt_input_ids,
+        "chosen_input_ids": chosen_input_ids,
+        "rejected_input_ids": rejected_input_ids,
+    }
+    if "pixel_values" in processed_features:
+        output["pixel_values"] = processed_features["pixel_values"][0]
+    if "pixel_attention_mask" in processed_features:
+        output["pixel_attention_mask"] = processed_features["pixel_attention_mask"][0]
+    if "image_sizes" in processed_features:
+        output["image_sizes"] = processed_features["image_sizes"][0]
+    if "token_type_ids" in processed_features:
+        token_type_ids = processed_features["token_type_ids"][0]
+        if max_prompt_length is not None:
+            token_type_ids = token_type_ids[-max_prompt_length:]
+        output["token_type_ids"] = token_type_ids
+    if "pixel_position_ids" in processed_features:
+        output["pixel_position_ids"] = processed_features["pixel_position_ids"][0]
+    if "image_position_ids" in processed_features:
+        output["image_position_ids"] = processed_features["image_position_ids"][0]
+    if "mm_token_type_ids" in processed_features:
+        mm_token_type_ids = processed_features["mm_token_type_ids"][0]
+        if max_prompt_length is not None:
+            mm_token_type_ids = mm_token_type_ids[-max_prompt_length:]
+        output["mm_token_type_ids"] = mm_token_type_ids
+
+    return output
+
+
+def dpo_trainer_vision_signature_columns(function_name, function):
+    if function_name != "_set_signature_columns_if_needed":
+        return function
+
+    if all(_k in function for _k in _DPO_VISION_KEYS):
+        return function
+
+    _extra_columns = "".join(f'                "{_k}",\n' for _k in _DPO_VISION_KEYS)
+    new_function = function.replace(
+        '                "image_sizes",\n' '                "token_type_ids",\n',
+        f'                "image_sizes",\n'
+        f"{_extra_columns}"
+        f'                "token_type_ids",\n',
+    )
+    if new_function != function:
+        return new_function
+    return function.replace(
+        '                "image_sizes",\n' '                "ref_chosen_logps",\n',
+        f'                "image_sizes",\n'
+        f"{_extra_columns}"
+        f'                "ref_chosen_logps",\n',
+    )
+
+
+def dpo_trainer_concatenated_inputs(function_name, function):
+    if function_name != "concatenated_inputs":
+        return function
+
+    if all(_k in function for _k in _DPO_VISION_KEYS):
+        return function
+
+    _extra_inputs = "".join(
+        f'        if "{_k}" in batch:\n'
+        f'            output["{_k}"] = torch.cat((batch["{_k}"], batch["{_k}"]), dim=0)\n'
+        for _k in _DPO_VISION_KEYS
+    )
+
+    image_sizes_block = (
+        '        if "image_sizes" in batch:\n'
+        '            output["image_sizes"] = torch.cat([batch["image_sizes"], batch["image_sizes"]], dim=0)\n'
+    )
+    new_function = function.replace(
+        image_sizes_block + '        if "token_type_ids" in batch:\n',
+        image_sizes_block + _extra_inputs + '        if "token_type_ids" in batch:\n',
+    )
+    if new_function != function:
+        return new_function
+    if image_sizes_block in function:
+        return function.replace(image_sizes_block, image_sizes_block + _extra_inputs, 1)
+    return function
+
+
+def _dpo_trainer_extend_vision_model_kwargs(function):
+    if all(_k in function for _k in _DPO_VISION_KEYS):
+        return function
+
+    _extra_forward = "".join(
+        f'        if "{_k}" in concatenated_batch:\n'
+        f'            model_kwargs["{_k}"] = concatenated_batch["{_k}"]\n'
+        for _k in (
+            "pixel_values",
+            "pixel_attention_mask",
+            "image_sizes",
+            *_DPO_VISION_KEYS,
+        )
+    )
+
+    return function.replace(
+        '        if "pixel_values" in concatenated_batch:\n'
+        '            model_kwargs["pixel_values"] = concatenated_batch["pixel_values"]\n'
+        '        if "pixel_attention_mask" in concatenated_batch:\n'
+        '            model_kwargs["pixel_attention_mask"] = concatenated_batch["pixel_attention_mask"]\n'
+        '        if "image_sizes" in concatenated_batch:\n'
+        '            model_kwargs["image_sizes"] = concatenated_batch["image_sizes"]\n',
+        f"{_extra_forward}",
+    )
+
+
+def dpo_trainer_concatenated_forward(function_name, function):
+    if function_name != "concatenated_forward":
+        return function
+    return _dpo_trainer_extend_vision_model_kwargs(function)
+
+
+def dpo_trainer_compute_loss_liger(function_name, function):
+    if function_name != "_compute_loss_liger":
+        return function
+    return _dpo_trainer_extend_vision_model_kwargs(function)
+
+
+def dpo_trainer_data_collator_vision_keys(call_args, extra_args):
+    if "data_collator" not in call_args:
+        return ""
+
+    _vision_keys = str(_DPO_VISION_KEYS)
+    return (
+        "from trl.trainer.dpo_trainer import DataCollatorForPreference\n"
+        "if not hasattr(DataCollatorForPreference, '_unsloth_vision_keys_patch'):\n"
+        "    _old_dpo_collator_torch_call = DataCollatorForPreference.torch_call\n"
+        "\n"
+        "    def _unsloth_dpo_torch_call(self, examples):\n"
+        "        output = _old_dpo_collator_torch_call(self, examples)\n"
+        "        import torch as _unsloth_torch\n"
+        "        try:\n"
+        "            from trl.trainer.utils import pad as _unsloth_trl_pad\n"
+        "        except Exception:\n"
+        "            _unsloth_trl_pad = None\n"
+        "        for _k in " + _vision_keys + ":\n"
+        "            if not all(_k in example for example in examples):\n"
+        "                continue\n"
+        "            _is_position_key = _k.endswith('position_ids')\n"
+        "            _padding_value = -1 if _is_position_key else 0\n"
+        "            _padding_side = 'right' if _is_position_key else 'left'\n"
+        "            _values = [_unsloth_torch.as_tensor(example[_k]) for example in examples]\n"
+        "            try:\n"
+        "                if _unsloth_trl_pad is not None:\n"
+        "                    output[_k] = _unsloth_trl_pad(_values, padding_value=_padding_value, padding_side=_padding_side)\n"
+        "                else:\n"
+        "                    from torch.nn.utils.rnn import pad_sequence as _unsloth_pad_sequence\n"
+        "                    output[_k] = _unsloth_pad_sequence(_values, batch_first=True, padding_value=_padding_value)\n"
+        "            except Exception:\n"
+        "                from torch.nn.utils.rnn import pad_sequence as _unsloth_pad_sequence\n"
+        "                output[_k] = _unsloth_pad_sequence(_values, batch_first=True, padding_value=_padding_value)\n"
+        "        return output\n"
+        "\n"
+        "    DataCollatorForPreference.torch_call = _unsloth_dpo_torch_call\n"
+        "    DataCollatorForPreference._unsloth_vision_keys_patch = True\n"
+    )
+
+
+def dpo_trainer_prepare_dataset(function_name, function):
+    if function_name != "_prepare_dataset":
+        return function
+
+    legacy_call = "self.tokenize_row if not self.is_vision_model else self.process_row"
+    if legacy_call not in function:
+        return function
+
+    function = function.replace(
+        legacy_call,
+        "self.tokenize_row if not self.is_vision_model else dpo_trainer_vision_process_row",
+    )
+
+    legacy_tokenize_block = (
+        "            # Tokenize the dataset\n"
+        "            if isinstance(dataset, Dataset):  # `IterableDataset.map` does not support `desc`\n"
+        '                map_kwargs["desc"] = f"Tokenizing {dataset_name} dataset"\n'
+        "\n"
+        "            dataset = dataset.map(\n"
+        "                self.tokenize_row if not self.is_vision_model else dpo_trainer_vision_process_row,\n"
+    )
+    patched_tokenize_block = (
+        "            # Tokenize the dataset\n"
+        "            if isinstance(dataset, Dataset):  # `IterableDataset.map` does not support `desc`\n"
+        '                map_kwargs["desc"] = f"Tokenizing {dataset_name} dataset"\n'
+        "            if self.is_vision_model:\n"
+        '                map_kwargs.pop("num_proc", None)\n'
+        "\n"
+        "            dataset = dataset.map(\n"
+        "                self.tokenize_row if not self.is_vision_model else dpo_trainer_vision_process_row,\n"
+    )
+    if legacy_tokenize_block in function:
+        function = function.replace(legacy_tokenize_block, patched_tokenize_block, 1)
+    return function
+
+
+RL_FUNCTIONS["dpo_trainer"].append(dpo_trainer_prepare_dataset)
+RL_PRE_ITEMS["dpo_trainer"].append(inspect.getsource(dpo_trainer_vision_process_row))
+RL_FUNCTIONS["dpo_trainer"].append(dpo_trainer_vision_signature_columns)
+RL_FUNCTIONS["dpo_trainer"].append(dpo_trainer_concatenated_inputs)
+RL_FUNCTIONS["dpo_trainer"].append(dpo_trainer_concatenated_forward)
+RL_FUNCTIONS["dpo_trainer"].append(dpo_trainer_compute_loss_liger)
+RL_EXTRA_ARGS["dpo_trainer"].append(dpo_trainer_data_collator_vision_keys)
+
+
 # Fix tokenizer double BOS
 def sft_trainer_prepare_dataset(function_name, function):
     if (

From 4ab5378d28f8be14b0f6aaa93b5d07e8b82c00f1 Mon Sep 17 00:00:00 2001
From: Lee Jackson <130007945+Imagineer99@users.noreply.github.com>
Date: Thu, 30 Apr 2026 12:50:23 +0100
Subject: [PATCH 181/216] Studio: Pin assistant-ui core for fresh installs
 (#5229)

* fix(studio): pin assistant-ui core for fresh installs

* fix(studio): use assistant-ui internal export
---
 studio/frontend/package.json                                    | 1 +
 .../frontend/src/features/chat/utils/delete-thread-message.ts   | 2 +-
 2 files changed, 2 insertions(+), 1 deletion(-)

diff --git a/studio/frontend/package.json b/studio/frontend/package.json
index c5cb949ccd..7bea824071 100644
--- a/studio/frontend/package.json
+++ b/studio/frontend/package.json
@@ -16,6 +16,7 @@
     "biome:fix": "biome check . --write"
   },
   "dependencies": {
+    "@assistant-ui/core": "0.1.17",
     "@assistant-ui/react": "^0.12.19",
     "@assistant-ui/react-markdown": "^0.12.3",
     "@assistant-ui/react-streamdown": "^0.1.2",
diff --git a/studio/frontend/src/features/chat/utils/delete-thread-message.ts b/studio/frontend/src/features/chat/utils/delete-thread-message.ts
index a6d6556a87..9901ccf80e 100644
--- a/studio/frontend/src/features/chat/utils/delete-thread-message.ts
+++ b/studio/frontend/src/features/chat/utils/delete-thread-message.ts
@@ -16,7 +16,7 @@ import type {
  * delete + reload smoke tests; the path or API may change without a semver signal on “public”
  * surface area.
  */
-import { MessageRepository } from "@assistant-ui/core/runtime/utils/message-repository";
+import { MessageRepository } from "@assistant-ui/core/internal";
 import { db } from "@/features/chat/db";
 import type { MessageRecord } from "@/features/chat/types";
 

From 11c04ed6320e5a46d96684a8bfd925189a6458a9 Mon Sep 17 00:00:00 2001
From: Anish Umale <umaleanish120@gmail.com>
Date: Thu, 30 Apr 2026 19:43:20 +0530
Subject: [PATCH 182/216] Fix local model scanner to handle ollama cloud models
 (#5220)

* fix _scan_ollama_dir to handle ollama cloud models correctly

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

---------

Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com>
Co-authored-by: Lee Jackson <130007945+Imagineer99@users.noreply.github.com>
---
 studio/backend/routes/models.py | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/studio/backend/routes/models.py b/studio/backend/routes/models.py
index db27ce1907..c312b0ee1d 100644
--- a/studio/backend/routes/models.py
+++ b/studio/backend/routes/models.py
@@ -623,7 +623,7 @@ def _scan_ollama_dir(
             gguf_link_path: Optional[str] = None
             quant = f"-{file_type}" if file_type else ""
             safe_name = repo_name.replace("/", "-")
-            for layer in manifest.get("layers", []):
+            for layer in manifest.get("layers") or []:
                 media = layer.get("mediaType", "")
                 digest = layer.get("digest", "")
                 if not digest:

From 507417579f010c6b730ad097459906f075db64d6 Mon Sep 17 00:00:00 2001
From: Wasim Yousef Said <wasimysdev@gmail.com>
Date: Thu, 30 Apr 2026 17:40:39 +0200
Subject: [PATCH 183/216] Fix Studio desktop tray installer and titlebar and
 bux fixes (#5179)

* fix(tauri): dedupe tray and brand nsis installer

* feat(tauri): add linux windows custom titlebar

* Fix desktop auth gate after backend startup

* Fix desktop installer assets and setup script skew

* Scope setup failure exit to Tauri installer

* fix desktop updater production channel

* fix desktop auth runtime installer regressions

* fix desktop dev cors retry

* fix tauri process generation race

* feat desktop diagnostics support report

* fix tauri apt update best effort

* Fix Windows desktop NSIS installer upgrades

* Start managed backend after desktop install

* Improve NSIS installer branding resolution

* Fix assistant-ui internal import

* Fix desktop release workflow

* Keep desktop auth retry on cached backend

---------

Co-authored-by: wasimysaid <wasimysaid@users.noreply.github.com>
---
 .github/workflows/release-desktop.yml         |  45 +-
 install.ps1                                   | 217 +++-
 install.sh                                    | 157 ++-
 studio/backend/main.py                        |   2 +
 studio/frontend/package.json                  |   1 +
 studio/frontend/src/app/auth-guards.ts        |   5 +-
 studio/frontend/src/app/provider.tsx          |  88 +-
 studio/frontend/src/app/routes/__root.tsx     |   2 +-
 .../src/components/tauri/startup-screen.tsx   | 128 ++-
 .../src/components/tauri/update-banner.tsx    |  90 +-
 .../src/components/tauri/update-screen.tsx    |  48 +-
 .../src/components/tauri/window-titlebar.tsx  | 329 ++++++
 studio/frontend/src/features/auth/api.ts      |  46 +-
 .../features/auth/change-password-page.tsx    |   2 +-
 .../frontend/src/features/auth/login-page.tsx |   2 +-
 .../src/features/auth/tauri-auto-auth.ts      |  27 +-
 .../chat/utils/delete-thread-message.ts       |   4 +-
 .../data-recipes/pages/data-recipes-page.tsx  |   2 +-
 .../data-recipes/pages/edit-recipe-page.tsx   |   2 +-
 .../src/features/export/export-page.tsx       |   2 +-
 .../onboarding/components/wizard-layout.tsx   |   2 +-
 .../recipe-studio/recipe-studio-page.tsx      |   2 +-
 .../src/features/studio/studio-page.tsx       |   2 +-
 .../frontend/src/hooks/use-tauri-backend.ts   | 138 ++-
 studio/frontend/src/hooks/use-tauri-update.ts | 115 +-
 studio/frontend/src/lib/tauri-diagnostics.ts  | 176 ++++
 studio/setup.ps1                              |  16 +-
 studio/src-tauri/Cargo.lock                   | 355 ++++++-
 studio/src-tauri/Cargo.toml                   |   3 +-
 studio/src-tauri/capabilities/default.json    |   8 +-
 studio/src-tauri/src/commands.rs              | 227 +++-
 studio/src-tauri/src/desktop_auth.rs          | 169 ++-
 studio/src-tauri/src/diagnostics/mod.rs       | 169 +++
 studio/src-tauri/src/diagnostics/phase_log.rs | 823 +++++++++++++++
 studio/src-tauri/src/diagnostics/redaction.rs | 198 ++++
 studio/src-tauri/src/diagnostics/report.rs    | 607 +++++++++++
 studio/src-tauri/src/diagnostics/state.rs     | 774 ++++++++++++++
 studio/src-tauri/src/install.rs               | 416 +++++++-
 studio/src-tauri/src/main.rs                  |  73 +-
 studio/src-tauri/src/process.rs               | 202 +++-
 studio/src-tauri/src/update.rs                | 139 ++-
 studio/src-tauri/tauri.conf.json              |  40 +-
 studio/src-tauri/tauri.windows.conf.json      |   6 -
 .../windows/branding/nsis-header.bmp          | Bin 25818 -> 102654 bytes
 .../windows/branding/nsis-sidebar.bmp         | Bin 154542 -> 618006 bytes
 studio/src-tauri/windows/hooks.nsh            |   9 +-
 studio/src-tauri/windows/installer.nsi        | 994 ++++++++++++++++++
 47 files changed, 6487 insertions(+), 375 deletions(-)
 create mode 100644 studio/frontend/src/components/tauri/window-titlebar.tsx
 create mode 100644 studio/frontend/src/lib/tauri-diagnostics.ts
 create mode 100644 studio/src-tauri/src/diagnostics/mod.rs
 create mode 100644 studio/src-tauri/src/diagnostics/phase_log.rs
 create mode 100644 studio/src-tauri/src/diagnostics/redaction.rs
 create mode 100644 studio/src-tauri/src/diagnostics/report.rs
 create mode 100644 studio/src-tauri/src/diagnostics/state.rs
 create mode 100644 studio/src-tauri/windows/installer.nsi

diff --git a/.github/workflows/release-desktop.yml b/.github/workflows/release-desktop.yml
index 2d1c6f51f2..ea82739968 100644
--- a/.github/workflows/release-desktop.yml
+++ b/.github/workflows/release-desktop.yml
@@ -54,10 +54,43 @@ jobs:
         with:
           node-version: 24
 
+      - name: Install pinned Tauri CLI
+        run: npm install --save-dev --prefix studio @tauri-apps/cli@2.10.1
+
+      - name: Verify pinned Tauri CLI
+        shell: bash
+        run: |
+          out="$(npx --prefix studio tauri --version)"
+          echo "$out"
+          if [ "$out" != "tauri-cli 2.10.1" ]; then
+            echo "Expected tauri-cli 2.10.1, got $out" >&2
+            exit 1
+          fi
+
       - name: Install frontend dependencies
         working-directory: studio/frontend
         run: npm install
 
+      - name: Verify backend package is published
+        shell: bash
+        run: |
+          node <<'JS'
+          const { readFileSync } = require('node:fs');
+
+          (async () => {
+            const cargo = readFileSync('studio/src-tauri/Cargo.toml', 'utf8');
+            const match = cargo.match(/^version\s*=\s*"([^"]+)"/m);
+            if (!match) throw new Error('Could not read desktop app version');
+
+            const appVersion = match[1];
+            const response = await fetch(`https://pypi.org/pypi/unsloth/${appVersion}/json`);
+            if (!response.ok) {
+              const message = 'Publish unsloth=={app_version} to PyPI before the desktop release';
+              throw new Error(`${message.replace('{app_version}', appVersion)} (HTTP ${response.status})`);
+            }
+          })();
+          JS
+
       # ── Rust ──
       - name: Install Rust stable
         uses: dtolnay/rust-toolchain@stable
@@ -105,13 +138,14 @@ jobs:
       # ── Linux: build + sign + upload ──
       - name: Build Linux app
         if: matrix.platform == 'ubuntu-22.04'
-        uses: tauri-apps/tauri-action@fce9c6108b31ea247710505d3aaaa893ee6768d4
+        uses: tauri-apps/tauri-action@84b9d35b5fc46c1e45415bdb6144030364f7ebc5
         env:
           GITHUB_TOKEN: ${{ secrets.GITHUB_TOKEN }}
           TAURI_SIGNING_PRIVATE_KEY: ${{ secrets.TAURI_SIGNING_PRIVATE_KEY }}
           TAURI_SIGNING_PRIVATE_KEY_PASSWORD: ${{ secrets.TAURI_SIGNING_PRIVATE_KEY_PASSWORD }}
         with:
           projectPath: studio
+          tauriScript: npx --prefix . tauri
           tagName: desktop-v__VERSION__
           releaseName: 'Unsloth Studio (Desktop) v__VERSION__'
           releaseBody: |
@@ -123,6 +157,7 @@ jobs:
 
             > Linux in-app updates are AppImage-oriented. Package installs should update by downloading a new package.
             > Linux AppImage on Ubuntu 24.04+ may require: `sudo apt install libfuse2t64`
+            > First-run system dependency elevation is supported on Ubuntu/Debian. Other Linux distributions should install system packages manually.
           releaseDraft: ${{ inputs.draft }}
           prerelease: false
           args: -v ${{ matrix.args }}
@@ -130,7 +165,7 @@ jobs:
       # ── macOS: build + sign + notarize + upload ──
       - name: Build macOS app
         if: matrix.platform == 'macos-latest'
-        uses: tauri-apps/tauri-action@fce9c6108b31ea247710505d3aaaa893ee6768d4
+        uses: tauri-apps/tauri-action@84b9d35b5fc46c1e45415bdb6144030364f7ebc5
         env:
           GITHUB_TOKEN: ${{ secrets.GITHUB_TOKEN }}
           TAURI_SIGNING_PRIVATE_KEY: ${{ secrets.TAURI_SIGNING_PRIVATE_KEY }}
@@ -141,6 +176,7 @@ jobs:
           APPLE_TEAM_ID: ${{ secrets.APPLE_TEAM_ID }}
         with:
           projectPath: studio
+          tauriScript: npx --prefix . tauri
           tagName: desktop-v__VERSION__
           releaseName: 'Unsloth Studio (Desktop) v__VERSION__'
           releaseBody: |
@@ -152,6 +188,7 @@ jobs:
 
             > Linux in-app updates are AppImage-oriented. Package installs should update by downloading a new package.
             > Linux AppImage on Ubuntu 24.04+ may require: `sudo apt install libfuse2t64`
+            > First-run system dependency elevation is supported on Ubuntu/Debian. Other Linux distributions should install system packages manually.
           releaseDraft: ${{ inputs.draft }}
           prerelease: false
           args: -v ${{ matrix.args }}
@@ -159,7 +196,7 @@ jobs:
       # ── Windows: build + sign + upload ──
       - name: Build Windows app
         if: matrix.platform == 'windows-latest'
-        uses: tauri-apps/tauri-action@fce9c6108b31ea247710505d3aaaa893ee6768d4
+        uses: tauri-apps/tauri-action@84b9d35b5fc46c1e45415bdb6144030364f7ebc5
         env:
           GITHUB_TOKEN: ${{ secrets.GITHUB_TOKEN }}
           TAURI_SIGNING_PRIVATE_KEY: ${{ secrets.TAURI_SIGNING_PRIVATE_KEY }}
@@ -171,6 +208,7 @@ jobs:
           AZURE_CERTIFICATE_PROFILE_NAME: ${{ secrets.AZURE_CERTIFICATE_PROFILE_NAME }}
         with:
           projectPath: studio
+          tauriScript: npx --prefix . tauri
           tagName: desktop-v__VERSION__
           releaseName: 'Unsloth Studio (Desktop) v__VERSION__'
           releaseBody: |
@@ -182,6 +220,7 @@ jobs:
 
             > Linux in-app updates are AppImage-oriented. Package installs should update by downloading a new package.
             > Linux AppImage on Ubuntu 24.04+ may require: `sudo apt install libfuse2t64`
+            > First-run system dependency elevation is supported on Ubuntu/Debian. Other Linux distributions should install system packages manually.
           releaseDraft: ${{ inputs.draft }}
           prerelease: false
           args: -v ${{ matrix.args }}
diff --git a/install.ps1 b/install.ps1
index 2544eaf78a..dcd86eb019 100644
--- a/install.ps1
+++ b/install.ps1
@@ -8,6 +8,79 @@ function Install-UnslothStudio {
     $ErrorActionPreference = "Stop"
     $script:UnslothVerbose = ($env:UNSLOTH_VERBOSE -eq "1")
 
+    # ── Tauri structured output ──
+    function Write-TauriLog {
+        param([string]$Tag, [string]$Message)
+        if ($TauriMode) {
+            Write-Host "[TAURI:$Tag] $Message"
+        }
+    }
+
+    function Format-TauriDiagBool {
+        param([bool]$Value)
+        if ($Value) { return "true" }
+        return "false"
+    }
+
+    function Get-TauriDiagArch {
+        $arch = [string]$env:PROCESSOR_ARCHITECTURE
+        if ([string]::IsNullOrWhiteSpace($arch)) {
+            try { $arch = [System.Runtime.InteropServices.RuntimeInformation]::OSArchitecture.ToString() } catch { $arch = "unknown" }
+        }
+        $arch = $arch.ToLowerInvariant()
+        switch ($arch) {
+            "amd64" { return "x86_64" }
+            "x64" { return "x86_64" }
+            "arm64" { return "arm64" }
+            "x86" { return "x86" }
+            default { return ($arch -replace '[^a-z0-9_.-]', '_') }
+        }
+    }
+
+    function Get-TauriTorchIndexFamily {
+        param([string]$TorchIndexUrl)
+        if ($SkipTorch) { return "none" }
+        if ([string]::IsNullOrWhiteSpace($TorchIndexUrl)) { return "none" }
+        $leaf = ($TorchIndexUrl.TrimEnd('/') -split '/')[-1].ToLowerInvariant()
+        if (@("cpu", "cu118", "cu124", "cu126", "cu128", "cu130") -contains $leaf) { return $leaf }
+        if ($leaf -match '^rocm[0-9]+\.[0-9]+$') { return $leaf }
+        return "auto"
+    }
+
+    function Get-TauriGpuBranch {
+        param([string]$TorchIndexFamily)
+        if ($SkipTorch) { return "no_torch" }
+        if ($TorchIndexFamily -like "cu*") { return "cuda" }
+        if ($TorchIndexFamily -like "rocm*") { return "rocm" }
+        if ($TorchIndexFamily -eq "cpu") { return "cpu" }
+        return "unknown"
+    }
+
+    function Write-TauriDiag {
+        param(
+            [string]$GpuBranch = "unknown",
+            [string]$TorchIndexFamily = "none",
+            [string]$PythonVersionForDiag = $PythonVersion
+        )
+        if ([string]::IsNullOrWhiteSpace($PythonVersionForDiag)) { $PythonVersionForDiag = "unknown" }
+        Write-TauriLog "DIAG" "diag_schema=1 platform=windows arch=$(Get-TauriDiagArch) python_version=$($PythonVersionForDiag.ToLowerInvariant()) skip_torch=$(Format-TauriDiagBool $SkipTorch) mac_intel=false gpu_branch=$GpuBranch torch_index_family=$TorchIndexFamily"
+    }
+
+    function Exit-InstallFailure {
+        param(
+            [Parameter(Mandatory = $true)][string]$Message,
+            [int]$Code = 1
+        )
+        if ($Code -eq 0) { $Code = 1 }
+        Write-TauriLog "ERROR" $Message
+        if (Get-Command Restore-StudioVenvRollback -CommandType Function -ErrorAction SilentlyContinue) {
+            Restore-StudioVenvRollback
+        }
+        if ($TauriMode) {
+            exit $Code
+        }
+    }
+
     # ── Parse flags ──
     $StudioLocalInstall = $false
     $PackageName = "unsloth"
@@ -26,7 +99,7 @@ function Install-UnslothStudio {
                 $i++
                 if ($i -ge $argList.Count) {
                     Write-Host "[ERROR] --package requires an argument." -ForegroundColor Red
-                    return
+                    return (Exit-InstallFailure "--package requires an argument.")
                 }
                 $PackageName = $argList[$i]
             }
@@ -42,22 +115,14 @@ function Install-UnslothStudio {
         $RepoRoot = (Resolve-Path (Split-Path -Parent $PSCommandPath)).Path
         if (-not (Test-Path (Join-Path $RepoRoot "pyproject.toml"))) {
             Write-Host "[ERROR] --local must be run from the unsloth repo root (pyproject.toml not found at $RepoRoot)" -ForegroundColor Red
-            return
+            return (Exit-InstallFailure "--local must be run from the unsloth repo root")
         }
     }
 
     # Validate --package to prevent injection into shell/Python commands
     if ($PackageName -notmatch '^[a-zA-Z0-9][a-zA-Z0-9._-]*$') {
         Write-Host "[ERROR] --package name contains invalid characters (allowed: a-z A-Z 0-9 . _ -)" -ForegroundColor Red
-        return
-    }
-
-    # ── Tauri structured output ──
-    function Write-TauriLog {
-        param([string]$Tag, [string]$Message)
-        if ($TauriMode) {
-            Write-Host "[TAURI:$Tag] $Message"
-        }
+        return (Exit-InstallFailure "--package name contains invalid characters")
     }
 
     $PythonVersion = "3.13"
@@ -630,7 +695,7 @@ shell.Run cmd, 0, False
         step "winget" "not available" "Red"
         substep "Install it from https://aka.ms/getwinget" "Yellow"
         substep "or install Python $PythonVersion and uv manually, then re-run." "Yellow"
-        return
+        return (Exit-InstallFailure "winget is not available")
     }
 
     # ── Helper: detect a working Python 3.11-3.13 on the system ──
@@ -749,9 +814,14 @@ shell.Run cmd, 0, False
             Write-Host "        Please install Python $PythonVersion manually from https://www.python.org/downloads/" -ForegroundColor Yellow
             Write-Host "        Make sure to check 'Add Python to PATH' during installation." -ForegroundColor Yellow
             Write-Host "        Then re-run this installer." -ForegroundColor Yellow
-            return
+            return (Exit-InstallFailure "Python installation failed")
         }
     }
+    $DiagPythonVersion = $PythonVersion
+    if ($DetectedPython) { $DiagPythonVersion = $DetectedPython.Version }
+    $InitialGpuBranch = "unknown"
+    if ($SkipTorch) { $InitialGpuBranch = "no_torch" }
+    Write-TauriDiag -GpuBranch $InitialGpuBranch -TorchIndexFamily "none" -PythonVersionForDiag $DiagPythonVersion
 
     # ── Install uv if not present ──
     Write-TauriLog "STEP" "Installing uv package manager"
@@ -773,7 +843,7 @@ shell.Run cmd, 0, False
     if (-not (Get-Command uv -ErrorAction SilentlyContinue)) {
         step "uv" "could not be installed" "Red"
         substep "Install it from https://docs.astral.sh/uv/" "Yellow"
-        return
+        return (Exit-InstallFailure "uv could not be installed")
     }
 
     # ── Create venv (migrate old layout if possible, otherwise fresh) ──
@@ -786,11 +856,68 @@ shell.Run cmd, 0, False
 
     $VenvPython = Join-Path $VenvDir "Scripts\python.exe"
     $_Migrated = $false
+    $script:StudioVenvRollbackDir = $null
+    $script:StudioVenvRollbackTarget = $VenvDir
+    $script:StudioVenvRollbackActive = $false
+
+    function Start-StudioVenvRollback {
+        param([Parameter(Mandatory = $true)][string]$ExistingDir)
+        $stamp = Get-Date -Format "yyyyMMddHHmmss"
+        $candidate = Join-Path $StudioHome "unsloth_studio.rollback.$stamp.$PID"
+        $suffix = 0
+        while (Test-Path $candidate) {
+            $suffix++
+            $candidate = Join-Path $StudioHome "unsloth_studio.rollback.$stamp.$PID.$suffix"
+        }
+        Move-Item -Path $ExistingDir -Destination $candidate -ErrorAction Stop
+        $script:StudioVenvRollbackDir = $candidate
+        $script:StudioVenvRollbackTarget = $ExistingDir
+        $script:StudioVenvRollbackActive = $true
+        substep "previous environment preserved for rollback"
+    }
+
+    function Restore-StudioVenvRollback {
+        if (-not $script:StudioVenvRollbackActive) { return }
+        $backup = $script:StudioVenvRollbackDir
+        $target = $script:StudioVenvRollbackTarget
+        if (-not $backup -or -not (Test-Path $backup)) {
+            $script:StudioVenvRollbackActive = $false
+            return
+        }
+        substep "restoring previous environment after failed install..." "Yellow"
+        try {
+            if (Test-Path $target) {
+                Remove-Item -Recurse -Force $target -ErrorAction SilentlyContinue
+            }
+            Move-Item -Path $backup -Destination $target -Force -ErrorAction Stop
+            substep "restored previous environment"
+            $script:StudioVenvRollbackActive = $false
+            $script:StudioVenvRollbackDir = $null
+        } catch {
+            Write-Host "[WARN] Could not restore previous environment from $backup to $target" -ForegroundColor Yellow
+            Write-Host "       $($_.Exception.Message)" -ForegroundColor Yellow
+        }
+    }
+
+    function Complete-StudioVenvRollback {
+        if (-not $script:StudioVenvRollbackActive) { return }
+        $backup = $script:StudioVenvRollbackDir
+        if ($backup -and (Test-Path $backup)) {
+            Remove-Item -Recurse -Force $backup -ErrorAction SilentlyContinue
+        }
+        $script:StudioVenvRollbackActive = $false
+        $script:StudioVenvRollbackDir = $null
+    }
 
     if (Test-Path $VenvPython) {
-        # New layout already exists -- nuke for fresh install
-        substep "removing existing environment for fresh install..."
-        Remove-Item -Recurse -Force $VenvDir
+        # New layout already exists -- replace only after preserving rollback copy.
+        substep "preserving existing environment for rollback..."
+        try {
+            Start-StudioVenvRollback -ExistingDir $VenvDir
+        } catch {
+            Write-Host "[ERROR] Could not prepare existing environment for reinstall: $($_.Exception.Message)" -ForegroundColor Red
+            return (Exit-InstallFailure "Could not prepare existing environment for reinstall")
+        }
     } elseif (Test-Path (Join-Path $StudioHome ".venv\Scripts\python.exe")) {
         # Old layout (~/.unsloth/studio/.venv) exists -- validate before migrating
         $OldVenv = Join-Path $StudioHome ".venv"
@@ -799,18 +926,23 @@ shell.Run cmd, 0, False
         $prevEAP2 = $ErrorActionPreference
         $ErrorActionPreference = "Continue"
         try {
-            & $OldPy -c "import torch; A = torch.ones((2,2)); B = A + A" 2>$null | Out-Null
-            $torchOk = ($LASTEXITCODE -eq 0)
-        } catch { $torchOk = $false }
+            if ($SkipTorch) {
+                & $OldPy -c "import sys; print(sys.executable)" 2>$null | Out-Null
+            } else {
+                & $OldPy -c "import torch; A = torch.ones((2,2)); B = A + A" 2>$null | Out-Null
+            }
+            $legacyOk = ($LASTEXITCODE -eq 0)
+        } catch { $legacyOk = $false }
         $ErrorActionPreference = $prevEAP2
-        if ($torchOk) {
+        if ($legacyOk) {
             substep "legacy environment is healthy -- migrating..."
             Move-Item -Path $OldVenv -Destination $VenvDir -Force
             substep "moved .venv -> unsloth_studio"
             $_Migrated = $true
         } else {
             substep "legacy environment failed validation -- creating fresh environment" "Yellow"
-            Remove-Item -Recurse -Force $OldVenv -ErrorAction SilentlyContinue
+            $invalidVenv = Join-Path $StudioHome (".venv.invalid.{0}.{1}" -f (Get-Date -Format "yyyyMMddHHmmss"), $PID)
+            Move-Item -Path $OldVenv -Destination $invalidVenv -Force -ErrorAction SilentlyContinue
         }
     } elseif (Test-Path (Join-Path $env:USERPROFILE "unsloth_studio\Scripts\python.exe")) {
         # CWD-relative venv from old install.ps1 -- migrate to absolute path
@@ -826,9 +958,8 @@ shell.Run cmd, 0, False
         substep "$VenvDir"
         $venvExit = Invoke-InstallCommand { uv venv $VenvDir --python "$($DetectedPython.Path)" }
         if ($venvExit -ne 0) {
-            Write-TauriLog "ERROR" "Failed to create virtual environment (exit code $venvExit)"
             Write-Host "[ERROR] Failed to create virtual environment (exit code $venvExit)" -ForegroundColor Red
-            return
+            return (Exit-InstallFailure "Failed to create virtual environment (exit code $venvExit)" $venvExit)
         }
     } else {
         step "venv" "using migrated environment"
@@ -886,6 +1017,9 @@ shell.Run cmd, 0, False
         return "$baseUrl/cu126"
     }
     $TorchIndexUrl = Get-TorchIndexUrl
+    $TorchIndexFamily = Get-TauriTorchIndexFamily $TorchIndexUrl
+    $GpuBranch = Get-TauriGpuBranch $TorchIndexFamily
+    Write-TauriDiag -GpuBranch $GpuBranch -TorchIndexFamily $TorchIndexFamily -PythonVersionForDiag $DetectedPython.Version
 
     # ── Print CPU-only hint when no GPU detected ──
     if (-not $SkipTorch -and $TorchIndexUrl -like "*/cpu") {
@@ -946,14 +1080,14 @@ shell.Run cmd, 0, False
         }
         if ($baseInstallExit -ne 0) {
             Write-Host "[ERROR] Failed to install unsloth (exit code $baseInstallExit)" -ForegroundColor Red
-            return
+            return (Exit-InstallFailure "Failed to install unsloth (exit code $baseInstallExit)" $baseInstallExit)
         }
         if ($StudioLocalInstall) {
             substep "overlaying local repo (editable)..."
             $overlayExit = Invoke-InstallCommand { uv pip install --python $VenvPython -e $RepoRoot --no-deps }
             if ($overlayExit -ne 0) {
                 Write-Host "[ERROR] Failed to overlay local repo (exit code $overlayExit)" -ForegroundColor Red
-                return
+                return (Exit-InstallFailure "Failed to overlay local repo (exit code $overlayExit)" $overlayExit)
             }
         }
     } elseif ($TorchIndexUrl) {
@@ -964,9 +1098,8 @@ shell.Run cmd, 0, False
             substep "installing PyTorch ($TorchIndexUrl)..."
             $torchInstallExit = Invoke-InstallCommand { uv pip install --python $VenvPython "torch>=2.4,<2.11.0" torchvision torchaudio --index-url $TorchIndexUrl }
             if ($torchInstallExit -ne 0) {
-                Write-TauriLog "ERROR" "Failed to install PyTorch (exit code $torchInstallExit)"
                 Write-Host "[ERROR] Failed to install PyTorch (exit code $torchInstallExit)" -ForegroundColor Red
-                return
+                return (Exit-InstallFailure "Failed to install PyTorch (exit code $torchInstallExit)" $torchInstallExit)
             }
         }
 
@@ -988,9 +1121,8 @@ shell.Run cmd, 0, False
             $baseInstallExit = Invoke-InstallCommand { uv pip install --python $VenvPython --upgrade-package unsloth -- "$PackageName" }
         }
         if ($baseInstallExit -ne 0) {
-            Write-TauriLog "ERROR" "Failed to install unsloth (exit code $baseInstallExit)"
             Write-Host "[ERROR] Failed to install unsloth (exit code $baseInstallExit)" -ForegroundColor Red
-            return
+            return (Exit-InstallFailure "Failed to install unsloth (exit code $baseInstallExit)" $baseInstallExit)
         }
 
         if ($StudioLocalInstall) {
@@ -998,7 +1130,7 @@ shell.Run cmd, 0, False
             $overlayExit = Invoke-InstallCommand { uv pip install --python $VenvPython -e $RepoRoot --no-deps }
             if ($overlayExit -ne 0) {
                 Write-Host "[ERROR] Failed to overlay local repo (exit code $overlayExit)" -ForegroundColor Red
-                return
+                return (Exit-InstallFailure "Failed to overlay local repo (exit code $overlayExit)" $overlayExit)
             }
         }
     } else {
@@ -1009,20 +1141,19 @@ shell.Run cmd, 0, False
             $baseInstallExit = Invoke-InstallCommand { uv pip install --python $VenvPython unsloth-zoo "unsloth>=2026.4.8" --torch-backend=auto }
             if ($baseInstallExit -ne 0) {
                 Write-Host "[ERROR] Failed to install unsloth (exit code $baseInstallExit)" -ForegroundColor Red
-                return
+                return (Exit-InstallFailure "Failed to install unsloth (exit code $baseInstallExit)" $baseInstallExit)
             }
             substep "overlaying local repo (editable)..."
             $overlayExit = Invoke-InstallCommand { uv pip install --python $VenvPython -e $RepoRoot --no-deps }
             if ($overlayExit -ne 0) {
                 Write-Host "[ERROR] Failed to overlay local repo (exit code $overlayExit)" -ForegroundColor Red
-                return
+                return (Exit-InstallFailure "Failed to overlay local repo (exit code $overlayExit)" $overlayExit)
             }
         } else {
             $baseInstallExit = Invoke-InstallCommand { uv pip install --python $VenvPython --torch-backend=auto -- "$PackageName" }
             if ($baseInstallExit -ne 0) {
-                Write-TauriLog "ERROR" "Failed to install unsloth (exit code $baseInstallExit)"
                 Write-Host "[ERROR] Failed to install unsloth (exit code $baseInstallExit)" -ForegroundColor Red
-                return
+                return (Exit-InstallFailure "Failed to install unsloth (exit code $baseInstallExit)" $baseInstallExit)
             }
         }
     }
@@ -1077,12 +1208,11 @@ shell.Run cmd, 0, False
     step "setup" "running unsloth studio setup..."
     $UnslothExe = Join-Path $VenvDir "Scripts\unsloth.exe"
     if (-not (Test-Path $UnslothExe)) {
-        Write-TauriLog "ERROR" "unsloth CLI was not installed correctly"
         Write-Host "[ERROR] unsloth CLI was not installed correctly." -ForegroundColor Red
         Write-Host "        Expected: $UnslothExe" -ForegroundColor Yellow
         Write-Host "        This usually means an older unsloth version was installed that does not include the Studio CLI." -ForegroundColor Yellow
         Write-Host "        Try re-running the installer or see: https://github.com/unslothai/unsloth?tab=readme-ov-file#-quickstart" -ForegroundColor Yellow
-        return
+        return (Exit-InstallFailure "unsloth CLI was not installed correctly")
     }
     # Tell setup.ps1 to skip base package installation (install.ps1 already did it)
     $env:SKIP_STUDIO_BASE = "1"
@@ -1104,12 +1234,16 @@ shell.Run cmd, 0, False
     # and bypass the fast-path version check from PR #4667.
     $studioArgs = @('studio', 'setup')
     if ($script:UnslothVerbose) { $studioArgs += '--verbose' }
-    & $UnslothExe @studioArgs
-    $setupExit = $LASTEXITCODE
+    $env:UNSLOTH_INSTALL_ROLLBACK_MANAGED = "1"
+    try {
+        & $UnslothExe @studioArgs
+        $setupExit = $LASTEXITCODE
+    } finally {
+        Remove-Item Env:UNSLOTH_INSTALL_ROLLBACK_MANAGED -ErrorAction SilentlyContinue
+    }
     if ($setupExit -ne 0) {
-        Write-TauriLog "ERROR" "unsloth studio setup failed (exit code $setupExit)"
         Write-Host "[ERROR] unsloth studio setup failed (exit code $setupExit)" -ForegroundColor Red
-        return
+        return (Exit-InstallFailure "unsloth studio setup failed (exit code $setupExit)" $setupExit)
     }
 
     # ── Expose `unsloth` via a shim dir containing only unsloth.exe ──
@@ -1182,6 +1316,7 @@ shell.Run cmd, 0, False
         step "path" "added unsloth launcher to PATH"
     }
     Refresh-SessionPath  # sync current session with registry
+    Complete-StudioVenvRollback
 
     # ── Tauri mode: done, skip shortcuts and auto-launch ──
     if ($TauriMode) {
diff --git a/install.sh b/install.sh
index 6c28b6eda4..18c6bba589 100755
--- a/install.sh
+++ b/install.sh
@@ -162,9 +162,119 @@ tauri_log() {
     fi
 }
 
+tauri_diag_marker() {
+    _diag_gpu_branch="${1:-unknown}"
+    _diag_torch_index_family="${2:-none}"
+    tauri_log "DIAG" "diag_schema=1 platform=${OS:-unknown} arch=${_ARCH:-unknown} python_version=${PYTHON_VERSION:-unknown} skip_torch=${SKIP_TORCH:-false} mac_intel=${MAC_INTEL:-false} gpu_branch=${_diag_gpu_branch} torch_index_family=${_diag_torch_index_family}"
+}
+
+_tauri_torch_index_family() {
+    if [ "${SKIP_TORCH:-false}" = true ]; then
+        echo "none"
+        return
+    fi
+    _diag_url="${1:-}"
+    case "$_diag_url" in
+        */cu118) echo "cu118" ;;
+        */cu124) echo "cu124" ;;
+        */cu126) echo "cu126" ;;
+        */cu128) echo "cu128" ;;
+        */cu130) echo "cu130" ;;
+        */cpu) echo "cpu" ;;
+        */rocm[0-9]*.[0-9]*)
+            _diag_family=${_diag_url##*/}
+            case "$_diag_family" in
+                rocm[0-9]*.[0-9]*) echo "$_diag_family" ;;
+                *) echo "auto" ;;
+            esac ;;
+        "") echo "none" ;;
+        *) echo "auto" ;;
+    esac
+}
+
+_tauri_gpu_branch() {
+    _diag_family="${1:-unknown}"
+    _diag_radeon="${2:-false}"
+    if [ "${SKIP_TORCH:-false}" = true ]; then
+        echo "no_torch"
+        return
+    fi
+    if [ "${OS:-}" = "macos" ]; then
+        echo "mac"
+        return
+    fi
+    case "$_diag_family" in
+        cu*) echo "cuda" ;;
+        rocm*)
+            if [ "$_diag_radeon" = true ]; then
+                echo "rocm_radeon"
+            else
+                echo "rocm"
+            fi ;;
+        radeon) echo "rocm_radeon" ;;
+        cpu) echo "cpu" ;;
+        none) echo "no_torch" ;;
+        *) echo "unknown" ;;
+    esac
+}
+
 PYTHON_VERSION=""  # resolved after platform detection
 STUDIO_HOME="$HOME/.unsloth/studio"
 VENV_DIR="$STUDIO_HOME/unsloth_studio"
+_VENV_ROLLBACK_DIR=""
+_VENV_ROLLBACK_TARGET="$VENV_DIR"
+_VENV_ROLLBACK_ACTIVE=false
+
+_start_studio_venv_replacement() {
+    _existing_dir="$1"
+    _stamp=$(date +%Y%m%d%H%M%S 2>/dev/null || echo "time")
+    _candidate="$STUDIO_HOME/unsloth_studio.rollback.$_stamp.$$"
+    _suffix=0
+    while [ -e "$_candidate" ]; do
+        _suffix=$((_suffix + 1))
+        _candidate="$STUDIO_HOME/unsloth_studio.rollback.$_stamp.$$.$_suffix"
+    done
+    mv "$_existing_dir" "$_candidate"
+    _VENV_ROLLBACK_DIR="$_candidate"
+    _VENV_ROLLBACK_TARGET="$_existing_dir"
+    _VENV_ROLLBACK_ACTIVE=true
+    substep "previous environment preserved for rollback"
+}
+
+_restore_studio_venv_replacement() {
+    [ "$_VENV_ROLLBACK_ACTIVE" = true ] || return 0
+    [ -n "$_VENV_ROLLBACK_DIR" ] && [ -d "$_VENV_ROLLBACK_DIR" ] || {
+        _VENV_ROLLBACK_ACTIVE=false
+        return 0
+    }
+    substep "restoring previous environment after failed install..." "$C_WARN"
+    rm -rf "$_VENV_ROLLBACK_TARGET"
+    if mv "$_VENV_ROLLBACK_DIR" "$_VENV_ROLLBACK_TARGET"; then
+        substep "restored previous environment"
+        _VENV_ROLLBACK_ACTIVE=false
+        _VENV_ROLLBACK_DIR=""
+    else
+        echo "⚠️  Could not restore previous environment from $_VENV_ROLLBACK_DIR to $_VENV_ROLLBACK_TARGET" >&2
+    fi
+}
+
+_commit_studio_venv_replacement() {
+    [ "$_VENV_ROLLBACK_ACTIVE" = true ] || return 0
+    if [ -n "$_VENV_ROLLBACK_DIR" ] && [ -d "$_VENV_ROLLBACK_DIR" ]; then
+        rm -rf "$_VENV_ROLLBACK_DIR" || true
+    fi
+    _VENV_ROLLBACK_ACTIVE=false
+    _VENV_ROLLBACK_DIR=""
+}
+
+_on_install_exit() {
+    _status=$?
+    if [ "$_status" -ne 0 ]; then
+        _restore_studio_venv_replacement
+    fi
+    exit "$_status"
+}
+trap _on_install_exit EXIT
 
 # ── Helper: download a URL to a file (supports curl and wget) ──
 download() {
@@ -828,6 +938,14 @@ if [ "$_NO_TORCH_FLAG" = true ] || [ "$MAC_INTEL" = true ]; then
     SKIP_TORCH=true
 fi
 
+_TAURI_INITIAL_GPU_BRANCH="unknown"
+if [ "$SKIP_TORCH" = true ]; then
+    _TAURI_INITIAL_GPU_BRANCH="no_torch"
+elif [ "$OS" = "macos" ]; then
+    _TAURI_INITIAL_GPU_BRANCH="mac"
+fi
+tauri_diag_marker "$_TAURI_INITIAL_GPU_BRANCH" "none"
+
 # ── Check system dependencies ──
 # cmake and git are needed by unsloth studio setup to build the GGUF inference
 # engine (llama.cpp). build-essential and libcurl-dev are also needed on Linux.
@@ -883,9 +1001,15 @@ if [ -n "$MISSING" ]; then
             if command -v apt-get >/dev/null 2>&1; then
                 _smart_apt_install $MISSING
             else
-                echo "    apt-get is not available. Please install with your package manager:"
+                echo "    Automatic system package installation is supported on apt-based"
+                echo "    Linux distributions (Ubuntu/Debian) only. Please install the"
+                echo "    missing dependencies with your package manager, then re-run setup:"
                 echo "    $MISSING"
-                echo "    Then re-run Unsloth Studio setup."
+                echo ""
+                echo "    Examples:"
+                echo "      Fedora/RHEL: sudo dnf install cmake git gcc gcc-c++ make libcurl-devel"
+                echo "      Arch:       sudo pacman -S --needed cmake git base-devel curl"
+                echo "      openSUSE:   sudo zypper install cmake git gcc gcc-c++ make libcurl-devel"
                 exit 1
             fi
             ;;
@@ -957,12 +1081,19 @@ mkdir -p "$STUDIO_HOME"
 _MIGRATED=false
 
 if [ -x "$VENV_DIR/bin/python" ]; then
-    # New layout already exists — nuke for fresh install
-    rm -rf "$VENV_DIR"
+    # New layout already exists — replace only after preserving rollback copy.
+    substep "preserving existing environment for rollback..."
+    _start_studio_venv_replacement "$VENV_DIR"
 elif [ -x "$STUDIO_HOME/.venv/bin/python" ]; then
-    # Old layout exists — validate before migrating
+    # Old layout exists — validate before migrating.
+    # In no-torch mode, a missing torch package is expected; validate Python only.
     substep "found legacy Studio environment, validating..."
-    if "$STUDIO_HOME/.venv/bin/python" -c "
+    _legacy_ok=false
+    if [ "$SKIP_TORCH" = true ]; then
+        if "$STUDIO_HOME/.venv/bin/python" -c "import sys; print(sys.executable)" >/dev/null 2>&1; then
+            _legacy_ok=true
+        fi
+    elif "$STUDIO_HOME/.venv/bin/python" -c "
 import torch
 device = 'cuda' if torch.cuda.is_available() else 'cpu'
 A = torch.ones((10, 10), device=device)
@@ -972,13 +1103,17 @@ D = A + B
 E = D @ C
 torch.testing.assert_close(torch.unique(E), torch.tensor((20,), device=E.device, dtype=E.dtype))
 " >/dev/null 2>&1; then
+        _legacy_ok=true
+    fi
+    if [ "$_legacy_ok" = true ]; then
         echo "✅ Legacy environment is healthy — migrating..."
         mv "$STUDIO_HOME/.venv" "$VENV_DIR"
         echo "   Moved ~/.unsloth/studio/.venv → $VENV_DIR"
         _MIGRATED=true
     else
         echo "⚠️  Legacy environment failed validation — creating fresh environment"
-        rm -rf "$STUDIO_HOME/.venv"
+        _invalid_venv="$STUDIO_HOME/.venv.invalid.$(date +%Y%m%d%H%M%S 2>/dev/null || echo time).$$"
+        mv "$STUDIO_HOME/.venv" "$_invalid_venv" 2>/dev/null || true
     fi
 fi
 
@@ -1308,6 +1443,12 @@ case "$TORCH_INDEX_URL" in
         fi
         ;;
 esac
+_TAURI_TORCH_INDEX_FAMILY=$(_tauri_torch_index_family "$TORCH_INDEX_URL")
+if [ "$_amd_gpu_radeon" = true ] && [ "$SKIP_TORCH" = false ]; then
+    _TAURI_TORCH_INDEX_FAMILY="radeon"
+fi
+_TAURI_GPU_BRANCH=$(_tauri_gpu_branch "$_TAURI_TORCH_INDEX_FAMILY" "$_amd_gpu_radeon")
+tauri_diag_marker "$_TAURI_GPU_BRANCH" "$_TAURI_TORCH_INDEX_FAMILY"
 
 # ── Print CPU-only hint when no GPU detected ──
 case "$TORCH_INDEX_URL" in
@@ -1679,6 +1820,8 @@ if [ "$_SETUP_EXIT" -ne 0 ]; then
     exit "$_SETUP_EXIT"
 fi
 
+_commit_studio_venv_replacement
+
 # ── Tauri mode: done, skip shortcuts and auto-launch ──
 if [ "$TAURI_MODE" = true ]; then
     tauri_log "DONE" ""
diff --git a/studio/backend/main.py b/studio/backend/main.py
index 9212404b30..aec335fad3 100644
--- a/studio/backend/main.py
+++ b/studio/backend/main.py
@@ -187,6 +187,8 @@ if _api_only:
         "tauri://localhost",  # Linux/macOS Tauri webview
         "http://tauri.localhost",  # Windows Tauri webview
         "http://localhost",  # dev fallback
+        "http://localhost:5173",  # Tauri dev/Vite
+        "http://127.0.0.1:5173",  # Tauri dev/Vite fallback
     ]
     _cors_origin_regex = None
 else:
diff --git a/studio/frontend/package.json b/studio/frontend/package.json
index 7bea824071..3adc7ed0c8 100644
--- a/studio/frontend/package.json
+++ b/studio/frontend/package.json
@@ -43,6 +43,7 @@
     "@tanstack/react-router": "^1.159.10",
     "@tanstack/react-table": "^8.21.3",
     "@tauri-apps/api": "^2.10.1",
+    "@tauri-apps/plugin-clipboard-manager": "^2.3.2",
     "@tauri-apps/plugin-opener": "^2.5.3",
     "@tauri-apps/plugin-process": "^2.3.1",
     "@tauri-apps/plugin-updater": "^2.10.1",
diff --git a/studio/frontend/src/app/auth-guards.ts b/studio/frontend/src/app/auth-guards.ts
index 509b8f61af..52230f0b6f 100644
--- a/studio/frontend/src/app/auth-guards.ts
+++ b/studio/frontend/src/app/auth-guards.ts
@@ -9,7 +9,6 @@ import {
   hasRefreshToken,
   mustChangePassword,
   refreshSession,
-  tauriAutoAuth,
 } from "@/features/auth";
 
 async function hasActiveSession(): Promise<boolean> {
@@ -39,7 +38,7 @@ function authRedirect(to: "/login" | "/change-password"): never {
 
 export async function requireAuth(): Promise<void> {
   if (isTauri) {
-    await tauriAutoAuth();
+    // AppProvider owns backend startup + desktop auth; route guards run before it mounts.
     return;
   }
 
@@ -59,7 +58,6 @@ export async function requireAuth(): Promise<void> {
 
 export async function requireGuest(): Promise<void> {
   if (isTauri) {
-    await tauriAutoAuth();
     throw redirect({ to: "/chat" });
   }
   if (!(await hasActiveSession())) return;
@@ -68,7 +66,6 @@ export async function requireGuest(): Promise<void> {
 
 export async function requirePasswordChangeFlow(): Promise<void> {
   if (isTauri) {
-    await tauriAutoAuth();
     throw redirect({ to: "/chat" });
   }
 
diff --git a/studio/frontend/src/app/provider.tsx b/studio/frontend/src/app/provider.tsx
index b75998a169..e82f611552 100644
--- a/studio/frontend/src/app/provider.tsx
+++ b/studio/frontend/src/app/provider.tsx
@@ -4,12 +4,18 @@
 import { StartupScreen } from "@/components/tauri/startup-screen";
 import { UpdateBanner } from "@/components/tauri/update-banner";
 import { UpdateScreen } from "@/components/tauri/update-screen";
+import {
+  WindowTitlebar,
+  shouldUseCustomWindowTitlebar,
+} from "@/components/tauri/window-titlebar";
 import { Toaster } from "@/components/ui/sonner";
+import { getTauriAuthFailure, tauriAutoAuth } from "@/features/auth";
 import { useTauriBackend } from "@/hooks/use-tauri-backend";
 import { useTauriUpdate } from "@/hooks/use-tauri-update";
 import { isTauri } from "@/lib/api-base";
+import { useRouterState } from "@tanstack/react-router";
 import { ThemeProvider } from "next-themes";
-import { useEffect, useRef, type ReactNode } from "react";
+import { useEffect, useRef, useState, type ReactNode } from "react";
 
 interface AppProviderProps {
   children: ReactNode;
@@ -103,6 +109,7 @@ function TauriUpdateLayer({ isExternalServer }: { isExternalServer: boolean }) {
         error={update.error}
         onRetry={update.retryUpdate}
         onSkipRestart={update.skipAndRestart}
+        onCopyDiagnostics={update.copyDiagnostics}
       />
     );
   }
@@ -112,22 +119,34 @@ function TauriUpdateLayer({ isExternalServer }: { isExternalServer: boolean }) {
       status={update.status}
       info={update.info}
       dismissed={update.dismissed}
+      lastFailure={update.lastFailure}
       isExternalServer={isExternalServer}
       onInstall={update.installUpdate}
       onDismiss={update.dismiss}
+      onCopyDiagnostics={update.copyDiagnostics}
     />
   );
 }
 
+const HIDDEN_TITLEBAR_SIDEBAR_ROUTES = new Set([
+  "/onboarding",
+  "/login",
+  "/change-password",
+  "/signup",
+]);
+
 function TauriWrapper({ children }: { children: ReactNode }) {
+  const pathname = useRouterState({ select: (s) => s.location.pathname });
   const {
     status, logs, error, isExternalServer,
     currentStepIndex, progressDetail, elevationPackages,
-    startInstall, retry, retryInstall, approveElevation,
+    startInstall, retry, retryInstall, approveElevation, copyDiagnostics,
   } = useTauriBackend();
 
   const hasResized = useRef(false);
   const abortRef = useRef(false);
+  const [desktopAuthReady, setDesktopAuthReady] = useState(!isTauri);
+  const [desktopAuthRetry, setDesktopAuthRetry] = useState(0);
 
   // Show the window once the frontend mounts (for pre-running states)
   useEffect(() => {
@@ -150,24 +169,79 @@ function TauriWrapper({ children }: { children: ReactNode }) {
     return () => { abortRef.current = true; };
   }, [status]);
 
-  if (!isTauri) return <>{children}</>;
-  if (status === "running") return <><TauriUpdateLayer isExternalServer={isExternalServer} />{children}</>;
+  useEffect(() => {
+    if (!isTauri) {
+      setDesktopAuthReady(true);
+      return;
+    }
+    if (status !== "running") {
+      setDesktopAuthReady(false);
+      setDesktopAuthRetry(0);
+      return;
+    }
 
-  return (
+    let disposed = false;
+    setDesktopAuthReady(false);
+    tauriAutoAuth({ force: true }).then((authenticated) => {
+      if (disposed) return;
+      if (authenticated) {
+        setDesktopAuthReady(true);
+        return;
+      }
+      if (!getTauriAuthFailure()) {
+        window.setTimeout(() => {
+          if (!disposed) setDesktopAuthRetry((value) => value + 1);
+        }, 500);
+      }
+    });
+
+    return () => { disposed = true; };
+  }, [status, desktopAuthRetry]);
+
+  if (!isTauri) return <>{children}</>;
+
+  const showApp = status === "running" && desktopAuthReady;
+  const startupStatus = status === "running" ? "starting" : status;
+  const startupProgressDetail =
+    status === "running" && !desktopAuthReady
+      ? "Signing in to desktop session..."
+      : progressDetail;
+
+  const content = showApp ? (
+    <>
+      <TauriUpdateLayer isExternalServer={isExternalServer} />
+      {children}
+    </>
+  ) : (
     <StartupScreen
-      status={status}
+      status={startupStatus}
       logs={logs}
       error={error}
       currentStepIndex={currentStepIndex}
-      progressDetail={progressDetail}
+      progressDetail={startupProgressDetail}
       elevationPackages={elevationPackages}
       onInstall={startInstall}
       onRetry={retry}
       onRetryInstall={retryInstall}
       onApproveElevation={approveElevation}
       onStartServer={retry}
+      onCopyDiagnostics={copyDiagnostics}
     />
   );
+
+  if (!shouldUseCustomWindowTitlebar()) return content;
+
+  const showSidebarSurface =
+    showApp && !HIDDEN_TITLEBAR_SIDEBAR_ROUTES.has(pathname);
+
+  return (
+    <div className="flex h-dvh min-h-0 flex-col overflow-hidden bg-background [--studio-titlebar-height:34px]">
+      <WindowTitlebar showSidebarSurface={showSidebarSurface} />
+      <div className="min-h-0 flex-1 overflow-hidden">
+        {content}
+      </div>
+    </div>
+  );
 }
 
 export function AppProvider({ children }: AppProviderProps) {
diff --git a/studio/frontend/src/app/routes/__root.tsx b/studio/frontend/src/app/routes/__root.tsx
index 19b5763557..22d149473c 100644
--- a/studio/frontend/src/app/routes/__root.tsx
+++ b/studio/frontend/src/app/routes/__root.tsx
@@ -81,7 +81,7 @@ function RootLayout() {
           pinned={pinned}
           setPinned={setPinned}
           togglePinned={togglePinned}
-          className="!min-h-0 h-dvh overflow-hidden"
+          className="!min-h-0 h-[calc(100dvh-var(--studio-titlebar-height,0px))] overflow-hidden"
         >
           <AppSidebar />
           <SidebarInset className={isChatRoute ? "overflow-hidden" : "overflow-y-auto"}>
diff --git a/studio/frontend/src/components/tauri/startup-screen.tsx b/studio/frontend/src/components/tauri/startup-screen.tsx
index af5f241416..1eb3f81d10 100644
--- a/studio/frontend/src/components/tauri/startup-screen.tsx
+++ b/studio/frontend/src/components/tauri/startup-screen.tsx
@@ -3,7 +3,9 @@
 
 import { ShimmerButton } from "@/components/ui/shimmer-button";
 import type { BackendStatus } from "@/hooks/use-tauri-backend";
+import type { CopySupportDiagnosticsResult } from "@/lib/tauri-diagnostics";
 import { AnimatePresence, motion } from "motion/react";
+import { useState } from "react";
 
 interface StartupScreenProps {
   status: BackendStatus;
@@ -17,6 +19,63 @@ interface StartupScreenProps {
   onRetryInstall: () => void;
   onApproveElevation: () => void;
   onStartServer: () => void;
+  onCopyDiagnostics: () => Promise<CopySupportDiagnosticsResult>;
+}
+
+function DiagnosticsCopyActions({
+  onCopyDiagnostics,
+  children,
+}: {
+  onCopyDiagnostics: () => Promise<CopySupportDiagnosticsResult>;
+  children: React.ReactNode;
+}) {
+  const [copying, setCopying] = useState(false);
+  const [manualReport, setManualReport] = useState<string | null>(null);
+  const [manualMessage, setManualMessage] = useState<string | null>(null);
+
+  async function handleCopyDiagnostics() {
+    setCopying(true);
+    try {
+      const result = await onCopyDiagnostics();
+      if (result.ok) {
+        setManualReport(null);
+        setManualMessage(null);
+      } else {
+        setManualReport(result.report);
+        setManualMessage(result.error ?? "Clipboard copy failed. Select and copy the diagnostics below.");
+      }
+    } catch (error) {
+      setManualReport(null);
+      setManualMessage(`Diagnostics copy failed: ${String(error)}`);
+    } finally {
+      setCopying(false);
+    }
+  }
+
+  return (
+    <div className="mt-4 flex w-full flex-col items-center gap-3">
+      <div className="flex gap-3">
+        <ActionButton
+          variant="secondary"
+          onClick={() => void handleCopyDiagnostics()}
+        >
+          {copying ? "Copying..." : "Copy Diagnostics"}
+        </ActionButton>
+        {children}
+      </div>
+      {manualMessage && (
+        <p className="max-w-md text-center text-xs text-destructive">{manualMessage}</p>
+      )}
+      {manualReport && (
+        <textarea
+          readOnly
+          value={manualReport}
+          onFocus={(event) => event.currentTarget.select()}
+          className="h-32 w-full max-w-md resize-none rounded-lg border border-border/50 bg-muted/30 p-2 font-mono text-[10px] text-muted-foreground"
+        />
+      )}
+    </div>
+  );
 }
 
 // ---------------------------------------------------------------------------
@@ -180,12 +239,12 @@ function RepairingContent({
 
 function InstallErrorContent({
   error,
-  logs,
   onRetryInstall,
+  onCopyDiagnostics,
 }: {
   error: string | null;
-  logs: string[];
   onRetryInstall: () => void;
+  onCopyDiagnostics: () => Promise<CopySupportDiagnosticsResult>;
 }) {
   return (
     <>
@@ -195,15 +254,9 @@ function InstallErrorContent({
         {error && (
           <p className="max-w-xs text-center text-xs text-muted-foreground">{error}</p>
         )}
-        <div className="mt-4 flex gap-3">
-          <ActionButton
-            variant="secondary"
-            onClick={() => void navigator.clipboard.writeText(logs.join("\n"))}
-          >
-            Copy Logs
-          </ActionButton>
+        <DiagnosticsCopyActions onCopyDiagnostics={onCopyDiagnostics}>
           <ActionButton onClick={onRetryInstall}>Try Again</ActionButton>
-        </div>
+        </DiagnosticsCopyActions>
       </div>
     </>
   );
@@ -211,12 +264,12 @@ function InstallErrorContent({
 
 function RepairErrorContent({
   error,
-  logs,
   onRetry,
+  onCopyDiagnostics,
 }: {
   error: string | null;
-  logs: string[];
   onRetry: () => void;
+  onCopyDiagnostics: () => Promise<CopySupportDiagnosticsResult>;
 }) {
   return (
     <>
@@ -226,15 +279,9 @@ function RepairErrorContent({
         {error && (
           <p className="max-w-md text-center text-xs text-muted-foreground">{error}</p>
         )}
-        <div className="mt-4 flex gap-3">
-          <ActionButton
-            variant="secondary"
-            onClick={() => void navigator.clipboard.writeText(logs.join("\n"))}
-          >
-            Copy Logs
-          </ActionButton>
+        <DiagnosticsCopyActions onCopyDiagnostics={onCopyDiagnostics}>
           <ActionButton onClick={onRetry}>Retry</ActionButton>
-        </div>
+        </DiagnosticsCopyActions>
       </div>
     </>
   );
@@ -301,12 +348,12 @@ function StoppedContent({ onStartServer }: { onStartServer: () => void }) {
 
 function ErrorContent({
   error,
-  logs,
   onRetry,
+  onCopyDiagnostics,
 }: {
   error: string | null;
-  logs: string[];
   onRetry: () => void;
+  onCopyDiagnostics: () => Promise<CopySupportDiagnosticsResult>;
 }) {
   return (
     <>
@@ -316,15 +363,9 @@ function ErrorContent({
         {error && (
           <p className="max-w-md text-center text-xs text-muted-foreground">{error}</p>
         )}
-        <div className="mt-4 flex gap-3">
-          <ActionButton
-            variant="secondary"
-            onClick={() => void navigator.clipboard.writeText(logs.join("\n"))}
-          >
-            Copy Logs
-          </ActionButton>
+        <DiagnosticsCopyActions onCopyDiagnostics={onCopyDiagnostics}>
           <ActionButton onClick={onRetry}>Retry</ActionButton>
-        </div>
+        </DiagnosticsCopyActions>
       </div>
     </>
   );
@@ -346,6 +387,7 @@ export function StartupScreen({
   onRetryInstall,
   onApproveElevation,
   onStartServer,
+  onCopyDiagnostics,
 }: StartupScreenProps) {
   function renderContent() {
     switch (status) {
@@ -356,11 +398,23 @@ export function StartupScreen({
       case "installing":
         return <InstallingContent currentStepIndex={currentStepIndex} progressDetail={progressDetail} />;
       case "install-error":
-        return <InstallErrorContent error={error} logs={logs} onRetryInstall={onRetryInstall} />;
+        return (
+          <InstallErrorContent
+            error={error}
+            onRetryInstall={onRetryInstall}
+            onCopyDiagnostics={onCopyDiagnostics}
+          />
+        );
       case "repairing":
         return <RepairingContent logs={logs} progressDetail={progressDetail} />;
       case "repair-error":
-        return <RepairErrorContent error={error} logs={logs} onRetry={onRetry} />;
+        return (
+          <RepairErrorContent
+            error={error}
+            onRetry={onRetry}
+            onCopyDiagnostics={onCopyDiagnostics}
+          />
+        );
       case "needs-elevation":
         return (
           <NeedsElevationContent
@@ -376,12 +430,18 @@ export function StartupScreen({
       case "stopped":
         return <StoppedContent onStartServer={onStartServer} />;
       case "error":
-        return <ErrorContent error={error} logs={logs} onRetry={onRetry} />;
+        return (
+          <ErrorContent
+            error={error}
+            onRetry={onRetry}
+            onCopyDiagnostics={onCopyDiagnostics}
+          />
+        );
     }
   }
 
   return (
-    <div className="flex h-screen w-full flex-col items-center bg-background">
+    <div className="flex h-full w-full flex-col items-center bg-background">
       <div className="flex flex-1 w-full max-w-md items-center justify-center px-6">
         <AnimatePresence mode="wait">
           <motion.div
diff --git a/studio/frontend/src/components/tauri/update-banner.tsx b/studio/frontend/src/components/tauri/update-banner.tsx
index 98652c4db1..79fbae6bb7 100644
--- a/studio/frontend/src/components/tauri/update-banner.tsx
+++ b/studio/frontend/src/components/tauri/update-banner.tsx
@@ -2,16 +2,20 @@
 // Copyright 2026-present the Unsloth AI Inc. team. All rights reserved. See /studio/LICENSE.AGPL-3.0
 
 import { Button } from "@/components/ui/button";
-import type { UpdateInfo, UpdateStatus } from "@/hooks/use-tauri-update";
+import type { RetainedUpdateFailure, UpdateInfo, UpdateStatus } from "@/hooks/use-tauri-update";
+import type { CopySupportDiagnosticsResult } from "@/lib/tauri-diagnostics";
 import { AnimatePresence, motion } from "motion/react";
+import { useState } from "react";
 
 interface UpdateBannerProps {
   status: UpdateStatus;
   info: UpdateInfo | null;
   dismissed: boolean;
+  lastFailure: RetainedUpdateFailure | null;
   isExternalServer?: boolean;
   onInstall: () => void;
   onDismiss: () => void;
+  onCopyDiagnostics: () => Promise<CopySupportDiagnosticsResult>;
 }
 
 const EASE_OUT_QUART: [number, number, number, number] = [0.165, 0.84, 0.44, 1];
@@ -20,16 +24,41 @@ export function UpdateBanner({
   status,
   info,
   dismissed,
+  lastFailure,
   isExternalServer = false,
   onInstall,
   onDismiss,
+  onCopyDiagnostics,
 }: UpdateBannerProps) {
-  const visible = status === "available";
-  const show = visible && !dismissed;
+  const [copying, setCopying] = useState(false);
+  const [manualReport, setManualReport] = useState<string | null>(null);
+  const [manualMessage, setManualMessage] = useState<string | null>(null);
+  const showFailure = Boolean(lastFailure) && !dismissed;
+  const showAvailable = status === "available" && !dismissed && !showFailure;
+  const show = showFailure || (showAvailable && Boolean(info));
+
+  async function handleCopyDiagnostics() {
+    setCopying(true);
+    try {
+      const result = await onCopyDiagnostics();
+      if (result.ok) {
+        setManualReport(null);
+        setManualMessage(null);
+      } else {
+        setManualReport(result.report);
+        setManualMessage(result.error ?? "Clipboard copy failed. Select and copy the diagnostics below.");
+      }
+    } catch (error) {
+      setManualReport(null);
+      setManualMessage(`Diagnostics copy failed: ${String(error)}`);
+    } finally {
+      setCopying(false);
+    }
+  }
 
   return (
     <AnimatePresence>
-      {show && info && (
+      {show && (
         <motion.div
           initial={{ opacity: 0, y: -12, scale: 0.96 }}
           animate={{ opacity: 1, y: 0, scale: 1 }}
@@ -54,28 +83,61 @@ export function UpdateBanner({
               <span className="text-lg">🦥</span>
               <div>
                 <p className="text-sm font-semibold text-foreground">
-                  New version: v{info.version}
+                  {showFailure ? "App update failed" : `New version: v${info?.version}`}
                 </p>
                 <p className="text-xs text-muted-foreground">
-                  {isExternalServer
-                    ? "Run `unsloth studio update` from your terminal"
-                    : "A new app update is available"}
+                  {showFailure
+                    ? "Backend recovered. Diagnostics are still available."
+                    : isExternalServer
+                      ? "Run `unsloth studio update` from your terminal"
+                      : "A new app update is available"}
                 </p>
               </div>
             </div>
 
+            {/* Retained failure */}
+            {showFailure && lastFailure && (
+              <p className="mt-3 line-clamp-2 text-xs text-destructive">
+                {lastFailure.error}
+              </p>
+            )}
+
             {/* Actions */}
             <div className="mt-3 flex items-center gap-2">
-              <Button size="sm" className="corner-squircle" onClick={onInstall} disabled={isExternalServer}>
-                Update Now
-              </Button>
-              <Button size="sm" variant="outline" className="corner-squircle" disabled>
-                Release Notes
-              </Button>
+              {showFailure ? (
+                <>
+                  <Button size="sm" variant="outline" className="corner-squircle" onClick={() => void handleCopyDiagnostics()}>
+                    {copying ? "Copying..." : "Copy Diagnostics"}
+                  </Button>
+                  <Button size="sm" className="corner-squircle" onClick={onInstall} disabled={isExternalServer}>
+                    Retry Update
+                  </Button>
+                </>
+              ) : (
+                <>
+                  <Button size="sm" className="corner-squircle" onClick={onInstall} disabled={isExternalServer}>
+                    Update Now
+                  </Button>
+                  <Button size="sm" variant="outline" className="corner-squircle" disabled>
+                    Release Notes
+                  </Button>
+                </>
+              )}
               <Button size="sm" variant="ghost" className="corner-squircle" onClick={onDismiss}>
                 Later
               </Button>
             </div>
+            {manualMessage && (
+              <p className="mt-3 text-xs text-destructive">{manualMessage}</p>
+            )}
+            {manualReport && (
+              <textarea
+                readOnly
+                value={manualReport}
+                onFocus={(event) => event.currentTarget.select()}
+                className="mt-2 h-28 w-full resize-none rounded-lg border border-border/50 bg-muted/30 p-2 font-mono text-[10px] text-muted-foreground"
+              />
+            )}
           </div>
         </motion.div>
       )}
diff --git a/studio/frontend/src/components/tauri/update-screen.tsx b/studio/frontend/src/components/tauri/update-screen.tsx
index a8b77ab1fa..53e39612c0 100644
--- a/studio/frontend/src/components/tauri/update-screen.tsx
+++ b/studio/frontend/src/components/tauri/update-screen.tsx
@@ -2,8 +2,9 @@
 // Copyright 2026-present the Unsloth AI Inc. team. All rights reserved. See /studio/LICENSE.AGPL-3.0
 
 import type { UpdateStatus } from "@/hooks/use-tauri-update";
+import type { CopySupportDiagnosticsResult } from "@/lib/tauri-diagnostics";
 import { AnimatePresence, motion } from "motion/react";
-import { useEffect, useRef } from "react";
+import { useEffect, useRef, useState } from "react";
 
 interface UpdateScreenProps {
   status: UpdateStatus;
@@ -12,6 +13,7 @@ interface UpdateScreenProps {
   error: string | null;
   onRetry: () => void;
   onSkipRestart: () => void;
+  onCopyDiagnostics: () => Promise<CopySupportDiagnosticsResult>;
 }
 
 const EASE_OUT_QUART: [number, number, number, number] = [0.165, 0.84, 0.44, 1];
@@ -96,11 +98,34 @@ export function UpdateScreen({
   error,
   onRetry,
   onSkipRestart,
+  onCopyDiagnostics,
 }: UpdateScreenProps) {
   const isError = status === "error";
+  const [copying, setCopying] = useState(false);
+  const [manualReport, setManualReport] = useState<string | null>(null);
+  const [manualMessage, setManualMessage] = useState<string | null>(null);
+
+  async function handleCopyDiagnostics() {
+    setCopying(true);
+    try {
+      const result = await onCopyDiagnostics();
+      if (result.ok) {
+        setManualReport(null);
+        setManualMessage(null);
+      } else {
+        setManualReport(result.report);
+        setManualMessage(result.error ?? "Clipboard copy failed. Select and copy the diagnostics below.");
+      }
+    } catch (copyError) {
+      setManualReport(null);
+      setManualMessage(`Diagnostics copy failed: ${String(copyError)}`);
+    } finally {
+      setCopying(false);
+    }
+  }
 
   return (
-    <div className="flex h-screen w-full items-center justify-center bg-background">
+    <div className="flex h-full w-full items-center justify-center bg-background">
       <motion.div
         initial={{ opacity: 0, y: 8 }}
         animate={{ opacity: 1, y: 0 }}
@@ -148,6 +173,13 @@ export function UpdateScreen({
         {/* Error actions */}
         {isError && (
           <div className="mt-4 flex items-center gap-2">
+            <button
+              type="button"
+              className="rounded-lg bg-muted px-5 py-2.5 text-sm font-medium text-foreground transition-colors hover:bg-muted/80"
+              onClick={() => void handleCopyDiagnostics()}
+            >
+              {copying ? "Copying..." : "Copy Diagnostics"}
+            </button>
             <button
               type="button"
               className="rounded-lg bg-primary px-5 py-2.5 text-sm font-medium text-primary-foreground transition-colors hover:bg-primary/80"
@@ -165,6 +197,18 @@ export function UpdateScreen({
           </div>
         )}
 
+        {manualMessage && (
+          <p className="mt-3 max-w-xl text-center text-xs text-destructive">{manualMessage}</p>
+        )}
+        {manualReport && (
+          <textarea
+            readOnly
+            value={manualReport}
+            onFocus={(event) => event.currentTarget.select()}
+            className="mt-2 h-32 w-full max-w-xl resize-none rounded-lg border border-border/50 bg-muted/30 p-2 font-mono text-[10px] text-muted-foreground"
+          />
+        )}
+
         {/* Log viewer */}
         <LogViewer logs={logs} />
       </motion.div>
diff --git a/studio/frontend/src/components/tauri/window-titlebar.tsx b/studio/frontend/src/components/tauri/window-titlebar.tsx
new file mode 100644
index 0000000000..b3180e6663
--- /dev/null
+++ b/studio/frontend/src/components/tauri/window-titlebar.tsx
@@ -0,0 +1,329 @@
+// SPDX-License-Identifier: AGPL-3.0-only
+// Copyright 2026-present the Unsloth AI Inc. team. All rights reserved. See /studio/LICENSE.AGPL-3.0
+
+import { useSidebarPin } from "@/hooks/use-sidebar-pin";
+import { isTauri } from "@/lib/api-base";
+import { cn } from "@/lib/utils";
+import type { Window as TauriWindow } from "@tauri-apps/api/window";
+import {
+  type MouseEvent,
+  type ReactElement,
+  type ReactNode,
+  useCallback,
+  useEffect,
+  useState,
+} from "react";
+
+const CUSTOM_TITLEBAR_PLATFORMS = ["win", "linux", "x11"] as const;
+
+type WindowResizeDirection =
+  | "East"
+  | "North"
+  | "NorthEast"
+  | "NorthWest"
+  | "South"
+  | "SouthEast"
+  | "SouthWest"
+  | "West";
+
+type NavigatorWithUserAgentData = Navigator & {
+  userAgentData?: {
+    platform?: string;
+  };
+};
+
+function getClientPlatform(): string {
+  if (typeof navigator === "undefined") {
+    return "";
+  }
+  const nav = navigator as NavigatorWithUserAgentData;
+  return (
+    nav.userAgentData?.platform ??
+    navigator.platform ??
+    navigator.userAgent
+  ).toLowerCase();
+}
+
+export function shouldUseCustomWindowTitlebar(): boolean {
+  if (!isTauri) {
+    return false;
+  }
+  const platform = getClientPlatform();
+  if (!platform || platform.includes("mac")) {
+    return false;
+  }
+  return CUSTOM_TITLEBAR_PLATFORMS.some((token) => platform.includes(token));
+}
+
+async function getAppWindow(): Promise<TauriWindow> {
+  const { getCurrentWindow } = await import("@tauri-apps/api/window");
+  return getCurrentWindow();
+}
+
+function WindowControlButton({
+  label,
+  className,
+  onClick,
+  children,
+}: {
+  label: string;
+  className?: string;
+  onClick: () => void;
+  children: ReactNode;
+}): ReactElement {
+  return (
+    <button
+      type="button"
+      aria-label={label}
+      title={label}
+      onClick={onClick}
+      className={cn(
+        "relative z-[80] inline-flex size-8 items-center justify-center rounded-md text-muted-foreground transition-colors hover:bg-muted/80 hover:text-foreground focus-visible:outline-none focus-visible:ring-2 focus-visible:ring-ring",
+        className,
+      )}
+    >
+      {children}
+    </button>
+  );
+}
+
+function MinimizeGlyph(): ReactElement {
+  return (
+    <span aria-hidden="true" className="h-px w-3.5 rounded-full bg-current" />
+  );
+}
+
+function MaximizeGlyph(): ReactElement {
+  return (
+    <span
+      aria-hidden="true"
+      className="size-3 rounded-[2px] border border-current"
+    />
+  );
+}
+
+function RestoreGlyph(): ReactElement {
+  return (
+    <span aria-hidden="true" className="relative size-3.5">
+      <span className="absolute left-0.5 top-0 size-2.5 rounded-[2px] border border-current" />
+      <span className="absolute bottom-0 right-0 size-2.5 rounded-[2px] border border-current bg-muted" />
+    </span>
+  );
+}
+
+function CloseGlyph(): ReactElement {
+  return (
+    <span aria-hidden="true" className="relative size-3.5">
+      <span className="absolute left-1/2 top-0 h-3.5 w-px -translate-x-1/2 rotate-45 rounded-full bg-current" />
+      <span className="absolute left-1/2 top-0 h-3.5 w-px -translate-x-1/2 -rotate-45 rounded-full bg-current" />
+    </span>
+  );
+}
+
+export function WindowTitlebar({
+  showSidebarSurface = false,
+}: {
+  showSidebarSurface?: boolean;
+}): ReactElement | null {
+  const [enabled] = useState(shouldUseCustomWindowTitlebar);
+  const [maximized, setMaximized] = useState(false);
+  const { pinned } = useSidebarPin();
+
+  const refreshMaximized = useCallback(async () => {
+    if (!enabled) {
+      return;
+    }
+    try {
+      const appWindow = await getAppWindow();
+      setMaximized(await appWindow.isMaximized());
+    } catch {
+      // If a window permission is not ready yet, keep the previous visual state.
+    }
+  }, [enabled]);
+
+  useEffect(() => {
+    if (!enabled) {
+      return;
+    }
+    let mounted = true;
+    let unlistenResize: (() => void) | undefined;
+    let unlistenFocus: (() => void) | undefined;
+
+    const setupWindowListeners = async () => {
+      try {
+        const appWindow = await getAppWindow();
+        if (!mounted) {
+          return;
+        }
+        setMaximized(await appWindow.isMaximized());
+        unlistenResize = await appWindow.onResized(() => {
+          refreshMaximized().catch(() => undefined);
+        });
+        unlistenFocus = await appWindow.onFocusChanged(() => {
+          refreshMaximized().catch(() => undefined);
+        });
+      } catch {
+        // Missing capabilities should not break the rest of the app shell.
+      }
+    };
+
+    setupWindowListeners().catch(() => undefined);
+
+    return () => {
+      mounted = false;
+      unlistenResize?.();
+      unlistenFocus?.();
+    };
+  }, [enabled, refreshMaximized]);
+
+  const runWindowAction = useCallback(
+    (action: (appWindow: TauriWindow) => Promise<void>) => {
+      const runAction = async () => {
+        try {
+          const appWindow = await getAppWindow();
+          await action(appWindow);
+          await refreshMaximized();
+        } catch {
+          // Keep custom chrome inert rather than throwing into React on denied commands.
+        }
+      };
+
+      runAction().catch(() => undefined);
+    },
+    [refreshMaximized],
+  );
+
+  const handleDragMouseDown = useCallback(
+    (event: MouseEvent<HTMLDivElement>) => {
+      if (event.button !== 0 || event.detail > 1) {
+        return;
+      }
+      runWindowAction((appWindow) => appWindow.startDragging());
+    },
+    [runWindowAction],
+  );
+
+  const handleDragDoubleClick = useCallback(
+    (event: MouseEvent<HTMLDivElement>) => {
+      if (event.button !== 0) {
+        return;
+      }
+      runWindowAction((appWindow) => appWindow.toggleMaximize());
+    },
+    [runWindowAction],
+  );
+
+  const handleResizeMouseDown = useCallback(
+    (direction: WindowResizeDirection) =>
+      (event: MouseEvent<HTMLDivElement>) => {
+        if (event.button !== 0) {
+          return;
+        }
+        event.preventDefault();
+        event.stopPropagation();
+        runWindowAction(async (appWindow) => {
+          if (!(await appWindow.isResizable())) {
+            return;
+          }
+          await appWindow.startResizeDragging(direction);
+        });
+      },
+    [runWindowAction],
+  );
+
+  if (!enabled) {
+    return null;
+  }
+
+  return (
+    <>
+      <header
+        className="relative z-[60] flex h-[var(--studio-titlebar-height)] shrink-0 select-none items-center text-foreground"
+        aria-label="Window titlebar"
+      >
+        {showSidebarSurface && (
+          <div
+            className="h-full shrink-0 border-r border-sidebar-border bg-sidebar"
+            style={{ width: pinned ? "16rem" : "3rem" }}
+            onMouseDown={handleDragMouseDown}
+            onDoubleClick={handleDragDoubleClick}
+            aria-hidden="true"
+          />
+        )}
+        <div
+          className="h-full min-w-0 flex-1 border-b border-border/35 bg-muted/35"
+          onMouseDown={handleDragMouseDown}
+          onDoubleClick={handleDragDoubleClick}
+          aria-hidden="true"
+        />
+        <div
+          className="flex h-full shrink-0 items-center gap-0.5 border-b border-border/35 bg-muted/35 px-1"
+          role="toolbar"
+          aria-label="Window controls"
+        >
+          <WindowControlButton
+            label="Minimize window"
+            onClick={() => runWindowAction((appWindow) => appWindow.minimize())}
+          >
+            <MinimizeGlyph />
+          </WindowControlButton>
+          <WindowControlButton
+            label={maximized ? "Restore window" : "Maximize window"}
+            onClick={() =>
+              runWindowAction((appWindow) => appWindow.toggleMaximize())
+            }
+          >
+            {maximized ? <RestoreGlyph /> : <MaximizeGlyph />}
+          </WindowControlButton>
+          <WindowControlButton
+            label="Close window"
+            onClick={() => runWindowAction((appWindow) => appWindow.close())}
+            className="hover:bg-destructive hover:text-destructive-foreground focus-visible:ring-destructive/70"
+          >
+            <CloseGlyph />
+          </WindowControlButton>
+        </div>
+      </header>
+      <div
+        aria-hidden="true"
+        className="fixed inset-x-2 top-0 z-[70] h-1 cursor-n-resize"
+        onMouseDown={handleResizeMouseDown("North")}
+      />
+      <div
+        aria-hidden="true"
+        className="fixed inset-x-2 bottom-0 z-[70] h-1 cursor-s-resize"
+        onMouseDown={handleResizeMouseDown("South")}
+      />
+      <div
+        aria-hidden="true"
+        className="fixed inset-y-2 left-0 z-[70] w-1 cursor-w-resize"
+        onMouseDown={handleResizeMouseDown("West")}
+      />
+      <div
+        aria-hidden="true"
+        className="fixed inset-y-2 right-0 z-[70] w-1 cursor-e-resize"
+        onMouseDown={handleResizeMouseDown("East")}
+      />
+      <div
+        aria-hidden="true"
+        className="fixed left-0 top-0 z-[70] size-3 cursor-nw-resize"
+        onMouseDown={handleResizeMouseDown("NorthWest")}
+      />
+      <div
+        aria-hidden="true"
+        className="fixed right-0 top-0 z-[70] size-3 cursor-ne-resize"
+        onMouseDown={handleResizeMouseDown("NorthEast")}
+      />
+      <div
+        aria-hidden="true"
+        className="fixed bottom-0 left-0 z-[70] size-3 cursor-sw-resize"
+        onMouseDown={handleResizeMouseDown("SouthWest")}
+      />
+      <div
+        aria-hidden="true"
+        className="fixed bottom-0 right-0 z-[70] size-3 cursor-se-resize"
+        onMouseDown={handleResizeMouseDown("SouthEast")}
+      />
+    </>
+  );
+}
diff --git a/studio/frontend/src/features/auth/api.ts b/studio/frontend/src/features/auth/api.ts
index 97d815950d..8f52c2187c 100644
--- a/studio/frontend/src/features/auth/api.ts
+++ b/studio/frontend/src/features/auth/api.ts
@@ -18,6 +18,32 @@ type RefreshResponse = {
 
 let isRedirecting = false;
 
+const TAURI_FETCH_RETRY_DELAYS_MS = [250, 750, 1500] as const;
+
+function wait(ms: number): Promise<void> {
+  return new Promise((resolve) => setTimeout(resolve, ms));
+}
+
+async function fetchWithTauriNetworkRetry(
+  input: RequestInfo | URL,
+  init?: RequestInit,
+): Promise<Response> {
+  for (let attempt = 0; ; attempt++) {
+    try {
+      return await fetch(input, init);
+    } catch (error) {
+      if (
+        !isTauri ||
+        !(error instanceof TypeError) ||
+        attempt >= TAURI_FETCH_RETRY_DELAYS_MS.length
+      ) {
+        throw error;
+      }
+      await wait(TAURI_FETCH_RETRY_DELAYS_MS[attempt]);
+    }
+  }
+}
+
 async function isPasswordChangeRequiredResponse(response: Response): Promise<boolean> {
   if (response.status !== 403) return false;
 
@@ -54,7 +80,7 @@ async function retryWithCurrentToken(
   const retryHeaders = new Headers(init?.headers);
   const token = getAuthToken();
   if (token) retryHeaders.set("Authorization", `Bearer ${token}`);
-  return fetch(input, { ...init, headers: retryHeaders });
+  return fetchWithTauriNetworkRetry(input, { ...init, headers: retryHeaders });
 }
 
 async function retryWithTauriAutoAuth(
@@ -72,11 +98,14 @@ export async function refreshSession(): Promise<boolean> {
   if (!refreshToken) return false;
 
   try {
-    const response = await fetch(apiUrl("/api/auth/refresh"), {
-      method: "POST",
-      headers: { "Content-Type": "application/json" },
-      body: JSON.stringify({ refresh_token: refreshToken }),
-    });
+    const response = await fetchWithTauriNetworkRetry(
+      apiUrl("/api/auth/refresh"),
+      {
+        method: "POST",
+        headers: { "Content-Type": "application/json" },
+        body: JSON.stringify({ refresh_token: refreshToken }),
+      },
+    );
 
     if (!response.ok) {
       clearAuthTokens();
@@ -108,7 +137,10 @@ export async function authFetch(
 
   let response: Response;
   try {
-    response = await fetch(resolvedInput, { ...init, headers });
+    response = await fetchWithTauriNetworkRetry(resolvedInput, {
+      ...init,
+      headers,
+    });
   } catch (err) {
     if (err instanceof TypeError) {
       throw new Error("Studio isn't running -- please relaunch it.");
diff --git a/studio/frontend/src/features/auth/change-password-page.tsx b/studio/frontend/src/features/auth/change-password-page.tsx
index dd4ba711e0..96ee8dd1a8 100644
--- a/studio/frontend/src/features/auth/change-password-page.tsx
+++ b/studio/frontend/src/features/auth/change-password-page.tsx
@@ -7,7 +7,7 @@ import { AuthForm } from "./components/auth-form";
 
 export function ChangePasswordPage() {
   return (
-    <div className="relative flex min-h-screen items-center justify-center overflow-hidden bg-background px-4 py-8 sm:px-6 sm:py-10 md:px-10">
+    <div className="relative flex min-h-[calc(100dvh-var(--studio-titlebar-height,0px))] items-center justify-center overflow-hidden bg-background px-4 py-8 sm:px-6 sm:py-10 md:px-10">
       <LightRays
         count={6}
         color="rgba(34, 197, 94, 0.25)"
diff --git a/studio/frontend/src/features/auth/login-page.tsx b/studio/frontend/src/features/auth/login-page.tsx
index 502b35d22a..6f4d03f1a1 100644
--- a/studio/frontend/src/features/auth/login-page.tsx
+++ b/studio/frontend/src/features/auth/login-page.tsx
@@ -7,7 +7,7 @@ import { AuthForm } from "./components/auth-form";
 
 export function LoginPage() {
   return (
-    <div className="relative flex min-h-screen items-center justify-center overflow-hidden bg-background px-4 py-8 sm:px-6 sm:py-10 md:px-10">
+    <div className="relative flex min-h-[calc(100dvh-var(--studio-titlebar-height,0px))] items-center justify-center overflow-hidden bg-background px-4 py-8 sm:px-6 sm:py-10 md:px-10">
       <LightRays
         count={6}
         color="rgba(34, 197, 94, 0.25)"
diff --git a/studio/frontend/src/features/auth/tauri-auto-auth.ts b/studio/frontend/src/features/auth/tauri-auto-auth.ts
index d67730d2d6..a0199f9ac3 100644
--- a/studio/frontend/src/features/auth/tauri-auto-auth.ts
+++ b/studio/frontend/src/features/auth/tauri-auto-auth.ts
@@ -15,9 +15,13 @@ type DesktopAuthResponse = {
   refresh_token: string;
 };
 
+type TauriAutoAuthOptions = {
+  force?: boolean;
+};
+
 // Concurrency guard: multiple route guards can call tauriAutoAuth simultaneously.
 // Without this, the first-launch password-change could race with itself.
-let pending: Promise<boolean> | null = null;
+let pending: { promise: Promise<boolean>; force: boolean } | null = null;
 let lastTauriAuthFailure: string | null = null;
 
 const TAURI_AUTH_FAILURE_FALLBACK =
@@ -49,15 +53,15 @@ function isBackendNotReady(error: unknown): boolean {
   return authFailureMessage(error).includes(BACKEND_NOT_READY_MESSAGE);
 }
 
-async function doTauriAutoAuth(): Promise<boolean> {
+async function doTauriAutoAuth(options: TauriAutoAuthOptions): Promise<boolean> {
   // Desktop must handle password-change state internally in Rust.
-  if (hasAuthToken() && !mustChangePassword()) {
+  if (!options.force && hasAuthToken() && !mustChangePassword()) {
     clearTauriAuthFailure();
     return true;
   }
 
   // Try refreshing existing session
-  if (hasRefreshToken()) {
+  if (!options.force && hasRefreshToken()) {
     const refreshed = await refreshSession();
     if (refreshed && hasAuthToken() && !mustChangePassword()) {
       clearTauriAuthFailure();
@@ -86,10 +90,17 @@ async function doTauriAutoAuth(): Promise<boolean> {
  * Returns true if authentication succeeded.
  * Concurrent calls are coalesced into a single in-flight attempt.
  */
-export function tauriAutoAuth(): Promise<boolean> {
+export function tauriAutoAuth(
+  options: TauriAutoAuthOptions = {},
+): Promise<boolean> {
   if (!isTauri) return Promise.resolve(false);
-  if (!pending) {
-    pending = doTauriAutoAuth().finally(() => { pending = null; });
+  const force = options.force === true;
+  if (!pending || (force && !pending.force)) {
+    let promise: Promise<boolean>;
+    promise = doTauriAutoAuth({ force }).finally(() => {
+      if (pending?.promise === promise) pending = null;
+    });
+    pending = { promise, force };
   }
-  return pending;
+  return pending.promise;
 }
diff --git a/studio/frontend/src/features/chat/utils/delete-thread-message.ts b/studio/frontend/src/features/chat/utils/delete-thread-message.ts
index 9901ccf80e..204c0d0064 100644
--- a/studio/frontend/src/features/chat/utils/delete-thread-message.ts
+++ b/studio/frontend/src/features/chat/utils/delete-thread-message.ts
@@ -9,7 +9,9 @@ import type {
 /**
  * assistant-ui does not expose a public `deleteMessage` on `ThreadRuntime` / `MessageRuntime`
  * in our version, but it already implements branch-safe deletion inside `MessageRepository`.
- * We import that helper from an **internal** package path (`runtime/utils/message-repository`).
+ * We import that helper from `@assistant-ui/core/internal`, the package's exported internal
+ * surface. Avoid importing the deeper `runtime/utils/message-repository` path directly: newer
+ * `@assistant-ui/core` releases no longer export arbitrary deep paths.
  *
  * **Maintainability:** treat this file as the only place that imports `MessageRepository` from
  * `@assistant-ui/core`. When bumping `@assistant-ui/react` / `@assistant-ui/core`, re-run chat
diff --git a/studio/frontend/src/features/data-recipes/pages/data-recipes-page.tsx b/studio/frontend/src/features/data-recipes/pages/data-recipes-page.tsx
index 0a51f9df70..79566f21bf 100644
--- a/studio/frontend/src/features/data-recipes/pages/data-recipes-page.tsx
+++ b/studio/frontend/src/features/data-recipes/pages/data-recipes-page.tsx
@@ -399,7 +399,7 @@ export function DataRecipesPage(): ReactElement {
   const isBusy = creatingRecipe || Boolean(loadingTemplateId);
 
   return (
-    <div className="min-h-screen bg-background">
+    <div className="min-h-[calc(100dvh-var(--studio-titlebar-height,0px))] bg-background">
       <main className="mx-auto w-full max-w-7xl px-6 py-8">
         <div className="flex items-center justify-between gap-4">
           <div>
diff --git a/studio/frontend/src/features/data-recipes/pages/edit-recipe-page.tsx b/studio/frontend/src/features/data-recipes/pages/edit-recipe-page.tsx
index 8ade8a34af..ae87f92a53 100644
--- a/studio/frontend/src/features/data-recipes/pages/edit-recipe-page.tsx
+++ b/studio/frontend/src/features/data-recipes/pages/edit-recipe-page.tsx
@@ -28,7 +28,7 @@ function RecipeLoadState({
   onBack: () => void;
 }): ReactElement {
   return (
-    <div className="min-h-screen bg-background">
+    <div className="min-h-[calc(100dvh-var(--studio-titlebar-height,0px))] bg-background">
       <main className="mx-auto flex min-h-[70vh] w-full max-w-4xl items-center justify-center px-6 py-8">
         <div className="w-full rounded-2xl border bg-card p-8 text-center">
           <h1 className="text-lg font-semibold">{title}</h1>
diff --git a/studio/frontend/src/features/export/export-page.tsx b/studio/frontend/src/features/export/export-page.tsx
index 454ce40772..0d5d9c813c 100644
--- a/studio/frontend/src/features/export/export-page.tsx
+++ b/studio/frontend/src/features/export/export-page.tsx
@@ -523,7 +523,7 @@ export function ExportPage() {
 
   // ---- Render ----
   return (
-    <div className="min-h-screen bg-background">
+    <div className="min-h-[calc(100dvh-var(--studio-titlebar-height,0px))] bg-background">
       <main className="mx-auto max-w-7xl px-4 py-4 sm:px-6">
         <GuidedTour {...tour.tourProps} />
 
diff --git a/studio/frontend/src/features/onboarding/components/wizard-layout.tsx b/studio/frontend/src/features/onboarding/components/wizard-layout.tsx
index 6df873ffd3..22274c5d7d 100644
--- a/studio/frontend/src/features/onboarding/components/wizard-layout.tsx
+++ b/studio/frontend/src/features/onboarding/components/wizard-layout.tsx
@@ -73,7 +73,7 @@ export function WizardLayout() {
   }, [isFinalStep]);
 
   return (
-    <div className="relative min-h-screen flex items-center justify-center overflow-hidden bg-gradient-to-br from-primary/5 via-background to-primary/3 p-4 sm:p-6 md:p-8">
+    <div className="relative flex min-h-[calc(100dvh-var(--studio-titlebar-height,0px))] items-center justify-center overflow-hidden bg-gradient-to-br from-primary/5 via-background to-primary/3 p-4 sm:p-6 md:p-8">
       {showSplash && (
         <SplashScreen
           onStartOnboarding={() => setShowSplash(false)}
diff --git a/studio/frontend/src/features/recipe-studio/recipe-studio-page.tsx b/studio/frontend/src/features/recipe-studio/recipe-studio-page.tsx
index 5dbc004ffd..c1001c0322 100644
--- a/studio/frontend/src/features/recipe-studio/recipe-studio-page.tsx
+++ b/studio/frontend/src/features/recipe-studio/recipe-studio-page.tsx
@@ -776,7 +776,7 @@ export function RecipeStudioPage({
   }
 
   return (
-    <div className="min-h-screen bg-background">
+    <div className="min-h-[calc(100dvh-var(--studio-titlebar-height,0px))] bg-background">
       <main className="w-full px-6 py-8">
         <div
           className="relative w-full overflow-hidden rounded-2xl corner-squircle border"
diff --git a/studio/frontend/src/features/studio/studio-page.tsx b/studio/frontend/src/features/studio/studio-page.tsx
index 5c5e11907c..5018dc3baa 100644
--- a/studio/frontend/src/features/studio/studio-page.tsx
+++ b/studio/frontend/src/features/studio/studio-page.tsx
@@ -127,7 +127,7 @@ export function StudioPage(): ReactElement {
   })();
 
   return (
-    <div className="relative min-h-screen bg-background">
+    <div className="relative min-h-[calc(100dvh-var(--studio-titlebar-height,0px))] bg-background">
       <main className="relative z-10 mx-auto max-w-7xl px-4 py-4 sm:px-6">
         <GuidedTour {...tour.tourProps} celebrate={isConfigTour} />
 
diff --git a/studio/frontend/src/hooks/use-tauri-backend.ts b/studio/frontend/src/hooks/use-tauri-backend.ts
index ed697fe31e..6bbbdb91f9 100644
--- a/studio/frontend/src/hooks/use-tauri-backend.ts
+++ b/studio/frontend/src/hooks/use-tauri-backend.ts
@@ -3,6 +3,10 @@
 
 import { useEffect, useState, useCallback, useRef } from "react";
 import { isTauri, setApiBase } from "@/lib/api-base";
+import {
+  copySupportDiagnostics,
+  type CopySupportDiagnosticsResult,
+} from "@/lib/tauri-diagnostics";
 import {
   clearTauriAuthFailure,
   getTauriAuthFailure,
@@ -35,6 +39,54 @@ interface DesktopPreflightResult {
   managed_bin: string | null;
 }
 
+const MANAGED_STARTUP_TIMEOUT_MS = 5 * 60_000;
+const MANAGED_STARTUP_POLL_MS = 500;
+
+type TauriInvoke = typeof import("@tauri-apps/api/core").invoke;
+type ManagedStartupResult =
+  | { status: "ready"; port: number }
+  | { status: "aborted" }
+  | { status: "missing-port" }
+  | { status: "unhealthy" };
+
+function wait(ms: number) {
+  return new Promise((resolve) => setTimeout(resolve, ms));
+}
+
+async function waitForManagedServerReady(
+  invoke: TauriInvoke,
+  getPort: () => number | null,
+  shouldContinue: () => boolean,
+): Promise<ManagedStartupResult> {
+  const deadline = Date.now() + MANAGED_STARTUP_TIMEOUT_MS;
+
+  while (Date.now() < deadline) {
+    if (!shouldContinue()) {
+      return { status: "aborted" };
+    }
+
+    const port = getPort();
+    if (port === null) {
+      await wait(MANAGED_STARTUP_POLL_MS);
+      continue;
+    }
+
+    const healthy = await invoke<boolean>("check_health", { port });
+    if (!shouldContinue()) {
+      return { status: "aborted" };
+    }
+    if (healthy && getPort() === port) {
+      return { status: "ready", port };
+    }
+
+    await wait(MANAGED_STARTUP_POLL_MS);
+  }
+
+  return getPort() === null
+    ? { status: "missing-port" }
+    : { status: "unhealthy" };
+}
+
 export function useTauriBackend() {
   const [status, setStatus] = useState<BackendStatus>("checking");
   const statusRef = useRef<BackendStatus>(status);
@@ -181,8 +233,11 @@ export function useTauriBackend() {
 
   async function startManagedServer() {
     // Prevent double-start race condition
-    if (startingRef.current) return;
+    if (startingRef.current) {
+      return;
+    }
     startingRef.current = true;
+    portRef.current = null;
 
     try {
       const { invoke } = await import("@tauri-apps/api/core");
@@ -192,23 +247,27 @@ export function useTauriBackend() {
 
       // Wait for the owned backend's server-port event. Do not attach to an
       // external backend if the managed start does not report a port.
-      for (let i = 0; i < 120; i++) {
-        if (portRef.current) {
-          const healthy = await invoke<boolean>("check_health", {
-            port: portRef.current,
-          });
-          if (healthy) {
-            setApiBase(portRef.current);
-            setRunningStatus();
-            startingRef.current = false;
-            return;
-          }
-        }
-        await new Promise((r) => setTimeout(r, 500));
+      const startupResult = await waitForManagedServerReady(
+        invoke,
+        () => portRef.current,
+        () => startingRef.current,
+      );
+
+      if (startupResult.status === "ready") {
+        setApiBase(startupResult.port);
+        setRunningStatus();
+        startingRef.current = false;
+        return;
       }
-      const message = !portRef.current
-        ? "Managed server started without reporting a port. Check the logs for details."
-        : "Server started but is not responding. Check the logs for details.";
+
+      if (startupResult.status === "aborted") {
+        return;
+      }
+
+      const message =
+        startupResult.status === "missing-port"
+          ? "Managed server started without reporting a port. Check the logs for details."
+          : "Server started but is not responding. Check the logs for details.";
       setBackendError(message);
     } catch (e) {
       const msg = String(e);
@@ -283,9 +342,11 @@ export function useTauriBackend() {
     const { invoke } = await import("@tauri-apps/api/core");
     try {
       await invoke("start_install");
-      // Install completed — this is the ONLY path that starts the server
-      // after install. The install-complete event listener does NOT call
-      // startServer() to avoid a double-start race condition.
+      // Install completed — start the managed backend we just installed.
+      // Do not run the general preflight here: it can attach to an unrelated
+      // already-running CLI/backend server before launching our managed one.
+      // The install-complete event listener does NOT call startServer() to
+      // avoid a double-start race condition.
       setBackendStatus("starting");
       elevationResumeRef.current = null;
       await startServer();
@@ -315,8 +376,16 @@ export function useTauriBackend() {
     checkInstallAndStart();
   }, []);
 
-  const retryInstall = useCallback(() => {
+  const retryInstall = useCallback(async () => {
     const resume = elevationResumeRef.current;
+    if (resume) {
+      try {
+        const { invoke } = await import("@tauri-apps/api/core");
+        await invoke("cancel_pending_elevation");
+      } catch (error) {
+        console.warn("Failed to record elevation cancellation", error);
+      }
+    }
     elevationResumeRef.current = null;
     clearBackendError();
     setLogs([]);
@@ -347,6 +416,31 @@ export function useTauriBackend() {
     }
   }, [elevationPackages]);
 
+  const copyDiagnostics = useCallback((): Promise<CopySupportDiagnosticsResult> => {
+    const currentStatus = statusRef.current;
+    const flow =
+      currentStatus === "repairing" ||
+      currentStatus === "repair-error" ||
+      (currentStatus === "needs-elevation" && elevationResumeRef.current === "repair")
+        ? "repair"
+        : currentStatus === "installing" ||
+            currentStatus === "install-error" ||
+            currentStatus === "not-installed" ||
+            currentStatus === "needs-elevation"
+          ? "install"
+          : "backend";
+
+    return copySupportDiagnostics({
+      status: currentStatus,
+      error,
+      currentStepIndex,
+      progressDetail,
+      elevationPackages,
+      lastUiLogLines: logs,
+      flow,
+    });
+  }, [currentStepIndex, elevationPackages, error, logs, progressDetail]);
+
   // Initial check on mount (guarded against Strict Mode double-mount)
   useEffect(() => {
     if (mountedRef.current) return;
@@ -482,6 +576,6 @@ export function useTauriBackend() {
     status, logs, error, isExternalServer,
     currentStepIndex, progressDetail, elevationPackages,
     startServer, stopServer, startInstall,
-    retry, retryInstall, approveElevation,
+    retry, retryInstall, approveElevation, copyDiagnostics,
   };
 }
diff --git a/studio/frontend/src/hooks/use-tauri-update.ts b/studio/frontend/src/hooks/use-tauri-update.ts
index 5863ef7aa5..67f9bf4c77 100644
--- a/studio/frontend/src/hooks/use-tauri-update.ts
+++ b/studio/frontend/src/hooks/use-tauri-update.ts
@@ -3,6 +3,10 @@
 
 import { useEffect, useRef, useState } from "react";
 import { isTauri } from "@/lib/api-base";
+import {
+  copySupportDiagnostics,
+  type CopySupportDiagnosticsResult,
+} from "@/lib/tauri-diagnostics";
 import { toast } from "sonner";
 
 export type UpdateStatus =
@@ -21,19 +25,74 @@ export interface UpdateInfo {
   date?: string;
 }
 
+export type UpdatePhase =
+  | "backend"
+  | "shell_download"
+  | "shell_install"
+  | "recovered_after_shell_failure";
+
+export interface RetainedUpdateFailure {
+  error: string;
+  phase: UpdatePhase;
+  progress: number;
+  logs: string[];
+}
+
 export function useTauriUpdate(isExternalServer = false) {
   const [status, setStatus] = useState<UpdateStatus>("idle");
   const [info, setInfo] = useState<UpdateInfo | null>(null);
   const [progress, setProgress] = useState(0);
+  const progressRef = useRef(0);
   const [logs, setLogs] = useState<string[]>([]);
+  const logsRef = useRef<string[]>([]);
+  const [phase, setPhase] = useState<UpdatePhase | null>(null);
+  const phaseRef = useRef<UpdatePhase | null>(null);
   const [dismissed, setDismissed] = useState(false);
   const [error, setError] = useState<string | null>(null);
+  const [lastFailure, setLastFailure] = useState<RetainedUpdateFailure | null>(null);
   const updateRef = useRef<Awaited<
     ReturnType<typeof import("@tauri-apps/plugin-updater").check>
   > | null>(null);
   const checkedRef = useRef(false);
   const updatingRef = useRef(false);
 
+  function replaceLogs(nextLogs: string[]) {
+    logsRef.current = nextLogs;
+    setLogs(nextLogs);
+  }
+
+  function appendLog(line: string) {
+    setLogs((prev) => {
+      const next = [...prev.slice(-499), line];
+      logsRef.current = next;
+      return next;
+    });
+  }
+
+  function setUpdateProgress(nextProgress: number) {
+    progressRef.current = nextProgress;
+    setProgress(nextProgress);
+  }
+
+  function setUpdatePhase(nextPhase: UpdatePhase | null) {
+    phaseRef.current = nextPhase;
+    setPhase(nextPhase);
+  }
+
+  function retainFailure(
+    nextError: string,
+    nextPhase: UpdatePhase = phaseRef.current ?? "backend",
+  ) {
+    const failure = {
+      error: nextError,
+      phase: nextPhase,
+      progress: progressRef.current,
+      logs: logsRef.current,
+    };
+    setLastFailure(failure);
+    return failure;
+  }
+
   useEffect(() => {
     if (!isTauri || checkedRef.current) return;
     checkedRef.current = true;
@@ -71,13 +130,15 @@ export function useTauriUpdate(isExternalServer = false) {
     updatingRef.current = true;
 
     const cleanups: (() => void)[] = [];
-    let phase: "backend" | "shell" = "backend";
 
     try {
       // ── Step 1: Backend update ──
+      setUpdatePhase("backend");
       setStatus("updating-backend");
-      setLogs([]);
+      replaceLogs([]);
+      setUpdateProgress(0);
       setError(null);
+      setLastFailure(null);
       setDismissed(false);
 
       const { listen } = await import("@tauri-apps/api/event");
@@ -87,7 +148,7 @@ export function useTauriUpdate(isExternalServer = false) {
       const unlistenProgress = await listen<string>(
         "update-progress",
         (e) => {
-          setLogs((prev) => [...prev.slice(-499), e.payload]);
+          appendLog(e.payload);
         },
       );
       cleanups.push(unlistenProgress);
@@ -107,6 +168,7 @@ export function useTauriUpdate(isExternalServer = false) {
       );
 
       if (backendResult !== "complete") {
+        retainFailure(backendResult, "backend");
         setError(backendResult);
         setStatus("error");
         updatingRef.current = false;
@@ -115,9 +177,9 @@ export function useTauriUpdate(isExternalServer = false) {
       }
 
       // ── Step 2: Shell update ──
-      phase = "shell";
+      setUpdatePhase("shell_download");
       setStatus("downloading");
-      setProgress(0);
+      setUpdateProgress(0);
 
       let downloaded = 0;
       let contentLength = 0;
@@ -129,10 +191,11 @@ export function useTauriUpdate(isExternalServer = false) {
           case "Progress":
             downloaded += event.data.chunkLength;
             if (contentLength > 0) {
-              setProgress(Math.round((downloaded / contentLength) * 100));
+              setUpdateProgress(Math.round((downloaded / contentLength) * 100));
             }
             break;
           case "Finished":
+            setUpdatePhase("shell_install");
             setStatus("installing");
             break;
         }
@@ -146,21 +209,26 @@ export function useTauriUpdate(isExternalServer = false) {
       const msg = String(e);
 
       // Shell update failed — restart backend on updated code
-      if (phase === "shell") {
+      if (phaseRef.current === "shell_download" || phaseRef.current === "shell_install") {
         try {
           const { invoke } = await import("@tauri-apps/api/core");
           await invoke("start_server", { port: 8888 });
+          retainFailure(msg, "recovered_after_shell_failure");
           toast.error("App update failed", {
             description:
-              "Backend was updated. The app update will be retried on next launch.",
+              "Backend was updated. Copy diagnostics from the update banner if you need support.",
           });
+          setError(null);
           setStatus("idle");
-          setDismissed(true);
+          setDismissed(false);
+          setUpdatePhase("recovered_after_shell_failure");
         } catch {
+          retainFailure(msg, phaseRef.current ?? "shell_install");
           setError(msg);
           setStatus("error");
         }
       } else {
+        retainFailure(msg, phaseRef.current ?? "backend");
         setError(msg);
         setStatus("error");
       }
@@ -176,15 +244,23 @@ export function useTauriUpdate(isExternalServer = false) {
   }
 
   async function skipAndRestart() {
+    const skippedError = error;
     try {
       const { invoke } = await import("@tauri-apps/api/core");
       await invoke("start_server", { port: 8888 });
+      if (skippedError) {
+        retainFailure(skippedError, phaseRef.current ?? "recovered_after_shell_failure");
+        setDismissed(false);
+      } else {
+        setDismissed(true);
+      }
       setStatus("idle");
       setError(null);
-      setLogs([]);
-      setDismissed(true);
+      replaceLogs([]);
     } catch (e) {
-      setError(String(e));
+      const msg = String(e);
+      retainFailure(msg, phaseRef.current ?? "backend");
+      setError(msg);
       setStatus("error");
     }
   }
@@ -193,6 +269,18 @@ export function useTauriUpdate(isExternalServer = false) {
     setDismissed(true);
   }
 
+  function copyDiagnostics(): Promise<CopySupportDiagnosticsResult> {
+    const failure = lastFailure;
+    return copySupportDiagnostics({
+      status: failure ? "error" : status,
+      error: failure?.error ?? error,
+      lastUiLogLines: failure?.logs ?? logs,
+      flow: "update",
+      updatePhase: failure?.phase ?? phase,
+      updateProgress: failure?.progress ?? progress,
+    });
+  }
+
   return {
     status,
     info,
@@ -200,11 +288,14 @@ export function useTauriUpdate(isExternalServer = false) {
     logs,
     dismissed,
     error,
+    phase,
+    lastFailure,
     isExternalServer,
     installUpdate,
     retryUpdate,
     skipAndRestart,
     dismiss,
+    copyDiagnostics,
   };
 }
 
diff --git a/studio/frontend/src/lib/tauri-diagnostics.ts b/studio/frontend/src/lib/tauri-diagnostics.ts
new file mode 100644
index 0000000000..5c07931a86
--- /dev/null
+++ b/studio/frontend/src/lib/tauri-diagnostics.ts
@@ -0,0 +1,176 @@
+// SPDX-License-Identifier: AGPL-3.0-only
+// Copyright 2026-present the Unsloth AI Inc. team. All rights reserved. See /studio/LICENSE.AGPL-3.0
+
+import { isTauri } from "@/lib/api-base";
+import { copyToClipboard } from "@/lib/copy-to-clipboard";
+
+export interface FrontendSupportSnapshot {
+  status?: string | null;
+  error?: string | null;
+  currentStepIndex?: number | null;
+  progressDetail?: string | null;
+  elevationPackages?: string[];
+  lastUiLogLines?: string[];
+  flow?: string | null;
+  updatePhase?: string | null;
+  updateProgress?: number | null;
+}
+
+export interface CopySupportDiagnosticsResult {
+  ok: boolean;
+  report: string;
+  error?: string;
+  source: "tauri" | "frontend-fallback";
+}
+
+const FALLBACK_LOG_LINE_LIMIT = 200;
+
+const ANSI_ESCAPE_PATTERN = new RegExp(
+  `${String.fromCharCode(27)}(?:[@-Z\\-_]|\\[[0-?]*[ -/]*[@-~])`,
+  "g",
+);
+
+function stripAnsi(text: string): string {
+  return text.replace(ANSI_ESCAPE_PATTERN, "");
+}
+
+export function redactDiagnosticsText(text: string): string {
+  let redacted = stripAnsi(text);
+
+  redacted = redacted.replace(
+    /-----BEGIN [^-]*PRIVATE KEY-----[\s\S]*?-----END [^-]*PRIVATE KEY-----/gi,
+    "<redacted-private-key>",
+  );
+  redacted = redacted.replace(
+    /([a-z][a-z0-9+.-]*:\/\/)([^/\s@]+)@/gi,
+    "$1<redacted>@",
+  );
+  redacted = redacted.replace(
+    /\b(authorization\s*[:=]\s*)(bearer|basic)\s+[^\s,;]+/gi,
+    "$1$2 <redacted>",
+  );
+  redacted = redacted.replace(/\bhf_[A-Za-z0-9]{20,}\b/g, "hf_<redacted>");
+  redacted = redacted.replace(/\bghp_[A-Za-z0-9_]{20,}\b/g, "ghp_<redacted>");
+  redacted = redacted.replace(/\bgithub_pat_[A-Za-z0-9_]{20,}\b/g, "github_pat_<redacted>");
+  redacted = redacted.replace(/\bsk-[A-Za-z0-9_-]{20,}\b/g, "sk-<redacted>");
+  redacted = redacted.replace(
+    /\b(cookie|set-cookie)\s*[:=]\s*[^\n\r]+/gi,
+    "$1=<redacted>",
+  );
+  redacted = redacted.replace(
+    /(^|[\s;])((?:[A-Z0-9]+_)*(?:TOKEN|KEY|SECRET|PASSWORD)(?:_[A-Z0-9]+)*\s*=\s*)[^\s]+/gi,
+    "$1$2<redacted>",
+  );
+
+  // Redact Studio paths before broader home-directory paths.
+  redacted = redacted.replace(
+    /(?:\/Users|\/home)\/[^\s/]+\/\.unsloth\/studio/gi,
+    "<studio_home>",
+  );
+  redacted = redacted.replace(
+    /[A-Z]:\\Users\\[^\s\\]+\\\.unsloth\\studio/gi,
+    "<studio_home>",
+  );
+  redacted = redacted.replace(/(?:\/Users|\/home)\/[^\s/]+/gi, "$HOME");
+  redacted = redacted.replace(/[A-Z]:\\Users\\[^\s\\]+/gi, "%USERPROFILE%");
+  redacted = redacted.replace(
+    /[A-Z0-9._%+-]+@[A-Z0-9.-]+\.[A-Z]{2,}/gi,
+    "<redacted-email>",
+  );
+
+  return redacted;
+}
+
+function safeLines(lines: string[] | undefined): string[] {
+  return (lines ?? [])
+    .filter((line): line is string => typeof line === "string")
+    .slice(-FALLBACK_LOG_LINE_LIMIT);
+}
+
+function buildFrontendFallbackReport(
+  snapshot: FrontendSupportSnapshot,
+  invokeError: unknown,
+): string {
+  const lines = safeLines(snapshot.lastUiLogLines);
+  const report = [
+    "DEGRADED FRONTEND-ONLY DIAGNOSTICS",
+    "DEGRADED_FRONTEND_ONLY=true",
+    `createdAt=${new Date().toISOString()}`,
+    "reason=collect_support_diagnostics invoke failed",
+    `invokeError=${String(invokeError)}`,
+    `status=${snapshot.status ?? "unavailable"}`,
+    `flow=${snapshot.flow ?? "unknown"}`,
+    `error=${snapshot.error ?? ""}`,
+    `currentStepIndex=${snapshot.currentStepIndex ?? ""}`,
+    `progressDetail=${snapshot.progressDetail ?? ""}`,
+    `elevationPackages=${(snapshot.elevationPackages ?? []).join(",")}`,
+    `updatePhase=${snapshot.updatePhase ?? ""}`,
+    `updateProgress=${snapshot.updateProgress ?? ""}`,
+    "",
+    "lastUiLogLines:",
+    ...lines,
+    "",
+    "redaction=frontend_minimal_js_redaction_applied",
+    "collectionWarnings=Rust diagnostics command was unavailable; this fallback contains only bounded UI state/log lines.",
+  ].join("\n");
+
+  return redactDiagnosticsText(report);
+}
+
+async function collectDiagnosticsReport(
+  snapshot: FrontendSupportSnapshot,
+): Promise<{ report: string; source: CopySupportDiagnosticsResult["source"] }> {
+  if (!isTauri) {
+    return {
+      report: buildFrontendFallbackReport(snapshot, "not running in Tauri"),
+      source: "frontend-fallback",
+    };
+  }
+
+  try {
+    const { invoke } = await import("@tauri-apps/api/core");
+    const report = await invoke<string>("collect_support_diagnostics", {
+      snapshot,
+    });
+    return { report, source: "tauri" };
+  } catch (error) {
+    console.warn("collect_support_diagnostics failed; using frontend-only diagnostics", error);
+    return {
+      report: buildFrontendFallbackReport(snapshot, error),
+      source: "frontend-fallback",
+    };
+  }
+}
+
+async function copyWithTauriClipboard(text: string): Promise<boolean> {
+  if (!isTauri) return false;
+  try {
+    const { writeText } = await import("@tauri-apps/plugin-clipboard-manager");
+    await writeText(text);
+    return true;
+  } catch (error) {
+    console.warn("Tauri clipboard-manager writeText failed", error);
+    return false;
+  }
+}
+
+export async function copySupportDiagnostics(
+  snapshot: FrontendSupportSnapshot,
+): Promise<CopySupportDiagnosticsResult> {
+  const { report, source } = await collectDiagnosticsReport(snapshot);
+
+  if (await copyWithTauriClipboard(report)) {
+    return { ok: true, report, source };
+  }
+
+  if (await copyToClipboard(report)) {
+    return { ok: true, report, source };
+  }
+
+  return {
+    ok: false,
+    report,
+    source,
+    error: "Unable to write diagnostics to the clipboard. Select and copy the text manually.",
+  };
+}
diff --git a/studio/setup.ps1 b/studio/setup.ps1
index 0da30c9229..3d082aa70d 100644
--- a/studio/setup.ps1
+++ b/studio/setup.ps1
@@ -1497,8 +1497,14 @@ substep "Using $PythonCmd ($(& $PythonCmd --version 2>&1))"
 $VenvDir = Join-Path $env:USERPROFILE ".unsloth\studio\unsloth_studio"
 
 # Stale-venv detection: if the venv exists but its torch flavor no longer
-# matches the current machine, wipe it so we get a clean install.
-if (Test-Path $VenvDir -PathType Container) {
+# matches the current machine, repair according to invocation context.
+# - install.ps1 sets UNSLOTH_INSTALL_ROLLBACK_MANAGED=1 so setup can delegate
+#   to the installer-level rollback that restores the previous environment.
+# - direct `unsloth studio update` keeps the pre-existing self-repair behavior.
+# In no-torch mode, a missing torch package is expected.
+$NoTorchMode = $env:UNSLOTH_NO_TORCH -match '^(?i:true|1|yes)$'
+$InstallerManagedSetup = $env:UNSLOTH_INSTALL_ROLLBACK_MANAGED -match '^(?i:true|1|yes)$'
+if ((Test-Path $VenvDir -PathType Container) -and -not $NoTorchMode) {
     $VenvPyExe = Join-Path $VenvDir "Scripts\python.exe"
     $installedTorchTag = $null
     $shouldRebuild = $false
@@ -1545,6 +1551,12 @@ if (Test-Path $VenvDir -PathType Container) {
 
     if ($shouldRebuild) {
         $reason = if ($installedTorchTag) { "torch $installedTorchTag != required $expectedTorchTag" } else { "torch could not be imported" }
+        if ($InstallerManagedSetup) {
+            substep "Stale venv detected ($reason)." "Yellow"
+            Write-Host "   [ERROR] The existing Studio environment needs repair." -ForegroundColor Red
+            Write-Host "           Re-run install.ps1 so it can replace the environment safely with rollback." -ForegroundColor Yellow
+            exit 1
+        }
         substep "Stale venv detected ($reason) -- rebuilding..." "Yellow"
         try {
             Remove-Item $VenvDir -Recurse -Force -ErrorAction Stop
diff --git a/studio/src-tauri/Cargo.lock b/studio/src-tauri/Cargo.lock
index 62d520437c..e491ee534a 100644
--- a/studio/src-tauri/Cargo.lock
+++ b/studio/src-tauri/Cargo.lock
@@ -56,6 +56,27 @@ dependencies = [
  "derive_arbitrary",
 ]
 
+[[package]]
+name = "arboard"
+version = "3.6.1"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "0348a1c054491f4bfe6ab86a7b6ab1e44e45d899005de92f58b3df180b36ddaf"
+dependencies = [
+ "clipboard-win",
+ "image",
+ "log",
+ "objc2",
+ "objc2-app-kit",
+ "objc2-core-foundation",
+ "objc2-core-graphics",
+ "objc2-foundation",
+ "parking_lot",
+ "percent-encoding",
+ "windows-sys 0.60.2",
+ "wl-clipboard-rs",
+ "x11rb",
+]
+
 [[package]]
 name = "async-broadcast"
 version = "0.7.2"
@@ -334,6 +355,12 @@ version = "1.5.0"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "1fd0f2584146f6f2ef48085050886acf353beff7305ebd1ae69500e27c67f64b"
 
+[[package]]
+name = "byteorder-lite"
+version = "0.1.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "8f1fe948ff07f4bd06c30984e69f5b4899c516a3ef74f34df92a2df2ab535495"
+
 [[package]]
 name = "bytes"
 version = "1.11.1"
@@ -482,6 +509,15 @@ dependencies = [
  "windows-link 0.2.1",
 ]
 
+[[package]]
+name = "clipboard-win"
+version = "5.4.1"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "bde03770d3df201d4fb868f2c9c59e66a3e4e2bd06692a0fe701e7103c7e84d4"
+dependencies = [
+ "error-code",
+]
+
 [[package]]
 name = "combine"
 version = "4.6.7"
@@ -609,6 +645,12 @@ version = "0.8.21"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "d0a5c400df2834b80a4c3327b3aad3a4c4cd4de0629063962b03235697506a28"
 
+[[package]]
+name = "crunchy"
+version = "0.2.4"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "460fbee9c2c2f33933d720630a6a0bac33ba7053db5344fac858d4b8952d77d5"
+
 [[package]]
 name = "crypto-common"
 version = "0.1.7"
@@ -850,6 +892,12 @@ dependencies = [
  "tendril 0.5.0",
 ]
 
+[[package]]
+name = "downcast-rs"
+version = "1.2.1"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "75b325c5dbd37f80359721ad39aca5a29fb04c89279657cffdda8736d0c0b9d2"
+
 [[package]]
 name = "dpi"
 version = "0.1.2"
@@ -983,6 +1031,12 @@ dependencies = [
  "windows-sys 0.61.2",
 ]
 
+[[package]]
+name = "error-code"
+version = "3.3.2"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "dea2df4cf52843e0452895c455a1a2cfbb842a1e7329671acf418fdc53ed4c59"
+
 [[package]]
 name = "event-listener"
 version = "5.4.1"
@@ -1010,6 +1064,12 @@ version = "2.3.0"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "37909eebbb50d72f9059c3b6d82c0463f2ff062c9e95845c43a6c9c0355411be"
 
+[[package]]
+name = "fax"
+version = "0.2.7"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "caf1079563223d5d59d83c85886a56e586cfd5c1a26292e971a0fa266531ac5a"
+
 [[package]]
 name = "fdeflate"
 version = "0.3.7"
@@ -1056,6 +1116,12 @@ dependencies = [
  "thiserror 1.0.69",
 ]
 
+[[package]]
+name = "fixedbitset"
+version = "0.5.7"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "1d674e81391d1e1ab681a28d99df07927c6d4aa5b027d7da16ba32d1d21ecd99"
+
 [[package]]
 name = "flate2"
 version = "1.1.9"
@@ -1347,6 +1413,16 @@ dependencies = [
  "version_check",
 ]
 
+[[package]]
+name = "gethostname"
+version = "1.1.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "1bd49230192a3797a9a4d6abe9b3eed6f7fa4c8a8a4947977c6f80025f92cbd8"
+dependencies = [
+ "rustix",
+ "windows-link 0.2.1",
+]
+
 [[package]]
 name = "getrandom"
 version = "0.1.16"
@@ -1562,6 +1638,17 @@ dependencies = [
  "tracing",
 ]
 
+[[package]]
+name = "half"
+version = "2.7.1"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "6ea2d84b969582b4b1864a92dc5d27cd2b77b622a8d79306834f1be5ba20d84b"
+dependencies = [
+ "cfg-if",
+ "crunchy",
+ "zerocopy",
+]
+
 [[package]]
 name = "hashbrown"
 version = "0.12.3"
@@ -1787,7 +1874,7 @@ source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "3e795dff5605e0f04bff85ca41b51a96b83e80b281e96231bcaaf1ac35103371"
 dependencies = [
  "byteorder",
- "png",
+ "png 0.17.16",
 ]
 
 [[package]]
@@ -1904,6 +1991,20 @@ dependencies = [
  "icu_properties",
 ]
 
+[[package]]
+name = "image"
+version = "0.25.10"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "85ab80394333c02fe689eaf900ab500fbd0c2213da414687ebf995a65d5a6104"
+dependencies = [
+ "bytemuck",
+ "byteorder-lite",
+ "moxcms",
+ "num-traits",
+ "png 0.18.1",
+ "tiff",
+]
+
 [[package]]
 name = "indexmap"
 version = "1.9.3"
@@ -2280,6 +2381,16 @@ dependencies = [
  "windows-sys 0.61.2",
 ]
 
+[[package]]
+name = "moxcms"
+version = "0.8.1"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "bb85c154ba489f01b25c0d36ae69a87e4a1c73a72631fc6c0eb6dde34a73e44b"
+dependencies = [
+ "num-traits",
+ "pxfm",
+]
+
 [[package]]
 name = "muda"
 version = "0.17.1"
@@ -2295,7 +2406,7 @@ dependencies = [
  "objc2-core-foundation",
  "objc2-foundation",
  "once_cell",
- "png",
+ "png 0.17.16",
  "serde",
  "thiserror 2.0.18",
  "windows-sys 0.60.2",
@@ -2372,6 +2483,15 @@ version = "0.1.14"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "72ef4a56884ca558e5ddb05a1d1e7e1bfd9a68d9ed024c21704cc98872dae1bb"
 
+[[package]]
+name = "nom"
+version = "8.0.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "df9761775871bdef83bee530e60050f7e54b1105350d6884eb0fb4f46c2f9405"
+dependencies = [
+ "memchr",
+]
+
 [[package]]
 name = "num-conv"
 version = "0.2.0"
@@ -2438,6 +2558,7 @@ dependencies = [
  "block2",
  "objc2",
  "objc2-core-foundation",
+ "objc2-core-graphics",
  "objc2-foundation",
 ]
 
@@ -2632,6 +2753,16 @@ dependencies = [
  "pin-project-lite",
 ]
 
+[[package]]
+name = "os_pipe"
+version = "1.2.3"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "7d8fae84b431384b68627d0f9b3b1245fcf9f46f6c0e3dc902e9dce64edd1967"
+dependencies = [
+ "libc",
+ "windows-sys 0.61.2",
+]
+
 [[package]]
 name = "osakit"
 version = "0.3.1"
@@ -2712,6 +2843,17 @@ version = "2.3.2"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "9b4f627cb1b25917193a259e49bdad08f671f8d9708acfd5fe0a8c1455d87220"
 
+[[package]]
+name = "petgraph"
+version = "0.8.3"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "8701b58ea97060d5e5b155d383a69952a60943f0e6dfe30b04c287beb0b27455"
+dependencies = [
+ "fixedbitset",
+ "hashbrown 0.15.5",
+ "indexmap 2.13.0",
+]
+
 [[package]]
 name = "phf"
 version = "0.8.0"
@@ -2942,7 +3084,7 @@ checksum = "740ebea15c5d1428f910cd1a5f52cebf8d25006245ed8ade92702f4943d91e07"
 dependencies = [
  "base64 0.22.1",
  "indexmap 2.13.0",
- "quick-xml",
+ "quick-xml 0.38.4",
  "serde",
  "time",
 ]
@@ -2960,6 +3102,19 @@ dependencies = [
  "miniz_oxide",
 ]
 
+[[package]]
+name = "png"
+version = "0.18.1"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "60769b8b31b2a9f263dae2776c37b1b28ae246943cf719eb6946a1db05128a61"
+dependencies = [
+ "bitflags 2.11.0",
+ "crc32fast",
+ "fdeflate",
+ "flate2",
+ "miniz_oxide",
+]
+
 [[package]]
 name = "polling"
 version = "3.11.0"
@@ -3094,6 +3249,18 @@ dependencies = [
  "windows 0.62.2",
 ]
 
+[[package]]
+name = "pxfm"
+version = "0.1.29"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "e0c5ccf5294c6ccd63a74f1565028353830a9c2f5eb0c682c355c471726a6e3f"
+
+[[package]]
+name = "quick-error"
+version = "2.0.1"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "a993555f31e5a609f617c12db6250dedcac1b0a85076912c436e6fc9b2c8e6a3"
+
 [[package]]
 name = "quick-xml"
 version = "0.38.4"
@@ -3103,6 +3270,15 @@ dependencies = [
  "memchr",
 ]
 
+[[package]]
+name = "quick-xml"
+version = "0.39.2"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "958f21e8e7ceb5a1aa7fa87fab28e7c75976e0bfe7e23ff069e0a260f894067d"
+dependencies = [
+ "memchr",
+]
+
 [[package]]
 name = "quote"
 version = "1.0.45"
@@ -4275,7 +4451,7 @@ dependencies = [
  "ico",
  "json-patch",
  "plist",
- "png",
+ "png 0.17.16",
  "proc-macro2",
  "quote",
  "semver",
@@ -4322,6 +4498,21 @@ dependencies = [
  "walkdir",
 ]
 
+[[package]]
+name = "tauri-plugin-clipboard-manager"
+version = "2.3.2"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "206dc20af4ed210748ba945c2774e60fd0acd52b9a73a028402caf809e9b6ecf"
+dependencies = [
+ "arboard",
+ "log",
+ "serde",
+ "serde_json",
+ "tauri",
+ "tauri-plugin",
+ "thiserror 2.0.18",
+]
+
 [[package]]
 name = "tauri-plugin-opener"
 version = "2.5.3"
@@ -4585,6 +4776,20 @@ dependencies = [
  "syn 2.0.117",
 ]
 
+[[package]]
+name = "tiff"
+version = "0.11.3"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "b63feaf3343d35b6ca4d50483f94843803b0f51634937cc2ec519fc32232bc52"
+dependencies = [
+ "fax",
+ "flate2",
+ "half",
+ "quick-error",
+ "weezl",
+ "zune-jpeg",
+]
+
 [[package]]
 name = "time"
 version = "0.3.47"
@@ -4886,12 +5091,23 @@ dependencies = [
  "objc2-core-graphics",
  "objc2-foundation",
  "once_cell",
- "png",
+ "png 0.17.16",
  "serde",
  "thiserror 2.0.18",
  "windows-sys 0.60.2",
 ]
 
+[[package]]
+name = "tree_magic_mini"
+version = "3.2.2"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "b8765b90061cba6c22b5831f675da109ae5561588290f9fa2317adab2714d5a6"
+dependencies = [
+ "memchr",
+ "nom",
+ "petgraph",
+]
+
 [[package]]
 name = "try-lock"
 version = "0.2.5"
@@ -4982,7 +5198,7 @@ checksum = "ebc1c04c71510c7f702b52b7c350734c9ff1295c464a03335b00bb84fc54f853"
 
 [[package]]
 name = "unsloth-studio"
-version = "2026.4.7"
+version = "2026.4.8"
 dependencies = [
  "dirs",
  "elevated-command",
@@ -4999,6 +5215,7 @@ dependencies = [
  "simplelog",
  "tauri",
  "tauri-build",
+ "tauri-plugin-clipboard-manager",
  "tauri-plugin-opener",
  "tauri-plugin-process",
  "tauri-plugin-single-instance",
@@ -5265,6 +5482,76 @@ dependencies = [
  "semver",
 ]
 
+[[package]]
+name = "wayland-backend"
+version = "0.3.15"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "2857dd20b54e916ec7253b3d6b4d5c4d7d4ca2c33c2e11c6c76a99bd8744755d"
+dependencies = [
+ "cc",
+ "downcast-rs",
+ "rustix",
+ "smallvec",
+ "wayland-sys",
+]
+
+[[package]]
+name = "wayland-client"
+version = "0.31.14"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "645c7c96bb74690c3189b5c9cb4ca1627062bb23693a4fad9d8c3de958260144"
+dependencies = [
+ "bitflags 2.11.0",
+ "rustix",
+ "wayland-backend",
+ "wayland-scanner",
+]
+
+[[package]]
+name = "wayland-protocols"
+version = "0.32.12"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "563a85523cade2429938e790815fd7319062103b9f4a2dc806e9b53b95982d8f"
+dependencies = [
+ "bitflags 2.11.0",
+ "wayland-backend",
+ "wayland-client",
+ "wayland-scanner",
+]
+
+[[package]]
+name = "wayland-protocols-wlr"
+version = "0.3.12"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "eb04e52f7836d7c7976c78ca0250d61e33873c34156a2a1fc9474828ec268234"
+dependencies = [
+ "bitflags 2.11.0",
+ "wayland-backend",
+ "wayland-client",
+ "wayland-protocols",
+ "wayland-scanner",
+]
+
+[[package]]
+name = "wayland-scanner"
+version = "0.31.10"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "9c324a910fd86ebdc364a3e61ec1f11737d3b1d6c273c0239ee8ff4bc0d24b4a"
+dependencies = [
+ "proc-macro2",
+ "quick-xml 0.39.2",
+ "quote",
+]
+
+[[package]]
+name = "wayland-sys"
+version = "0.31.11"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "d8eab23fefc9e41f8e841df4a9c707e8a8c4ed26e944ef69297184de2785e3be"
+dependencies = [
+ "pkg-config",
+]
+
 [[package]]
 name = "web-sys"
 version = "0.3.91"
@@ -5376,6 +5663,12 @@ dependencies = [
  "windows-core 0.61.2",
 ]
 
+[[package]]
+name = "weezl"
+version = "0.1.12"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "a28ac98ddc8b9274cb41bb4d9d4d5c425b6020c50c46f25559911905610b4a88"
+
 [[package]]
 name = "winapi"
 version = "0.3.9"
@@ -6016,6 +6309,24 @@ dependencies = [
  "wasmparser",
 ]
 
+[[package]]
+name = "wl-clipboard-rs"
+version = "0.9.3"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "e9651471a32e87d96ef3a127715382b2d11cc7c8bb9822ded8a7cc94072eb0a3"
+dependencies = [
+ "libc",
+ "log",
+ "os_pipe",
+ "rustix",
+ "thiserror 2.0.18",
+ "tree_magic_mini",
+ "wayland-backend",
+ "wayland-client",
+ "wayland-protocols",
+ "wayland-protocols-wlr",
+]
+
 [[package]]
 name = "writeable"
 version = "0.6.2"
@@ -6087,6 +6398,23 @@ dependencies = [
  "pkg-config",
 ]
 
+[[package]]
+name = "x11rb"
+version = "0.13.2"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "9993aa5be5a26815fe2c3eacfc1fde061fc1a1f094bf1ad2a18bf9c495dd7414"
+dependencies = [
+ "gethostname",
+ "rustix",
+ "x11rb-protocol",
+]
+
+[[package]]
+name = "x11rb-protocol"
+version = "0.13.2"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "ea6fc2961e4ef194dcbfe56bb845534d0dc8098940c7e5c012a258bfec6701bd"
+
 [[package]]
 name = "xattr"
 version = "1.6.1"
@@ -6279,6 +6607,21 @@ version = "1.0.21"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "b8848ee67ecc8aedbaf3e4122217aff892639231befc6a1b58d29fff4c2cabaa"
 
+[[package]]
+name = "zune-core"
+version = "0.5.1"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "cb8a0807f7c01457d0379ba880ba6322660448ddebc890ce29bb64da71fb40f9"
+
+[[package]]
+name = "zune-jpeg"
+version = "0.5.15"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "27bc9d5b815bc103f142aa054f561d9187d191692ec7c2d1e2b4737f8dbd7296"
+dependencies = [
+ "zune-core",
+]
+
 [[package]]
 name = "zvariant"
 version = "5.10.0"
diff --git a/studio/src-tauri/Cargo.toml b/studio/src-tauri/Cargo.toml
index ccb07cb9e2..269bb7a92a 100644
--- a/studio/src-tauri/Cargo.toml
+++ b/studio/src-tauri/Cargo.toml
@@ -1,6 +1,6 @@
 [package]
 name = "unsloth-studio"
-version = "2026.4.7"
+version = "2026.4.8"
 description = "Unsloth Studio Desktop App"
 authors = ["Unsloth AI"]
 edition = "2021"
@@ -22,6 +22,7 @@ process-wrap = { version = "9", features = ["std"] }
 fix-path-env = { git = "https://github.com/tauri-apps/fix-path-env-rs" }
 tauri-plugin-opener = "2.5.3"
 tauri-plugin-updater = "2"
+tauri-plugin-clipboard-manager = "2"
 rand = "0.10.0"
 
 [target.'cfg(unix)'.dependencies]
diff --git a/studio/src-tauri/capabilities/default.json b/studio/src-tauri/capabilities/default.json
index 7b6033c721..b624369abd 100644
--- a/studio/src-tauri/capabilities/default.json
+++ b/studio/src-tauri/capabilities/default.json
@@ -13,12 +13,18 @@
     "core:window:allow-set-size-constraints",
     "core:window:allow-center",
     "core:window:allow-current-monitor",
+    "core:window:allow-start-dragging",
+    "core:window:allow-start-resize-dragging",
+    "core:window:allow-minimize",
+    "core:window:allow-toggle-maximize",
+    "core:window:allow-close",
     "core:tray:default",
     "process:default",
     {
       "identifier": "opener:allow-open-url",
       "allow": [{ "url": "https://*" }, { "url": "http://*" }, { "url": "mailto:*" }]
     },
-    "updater:default"
+    "updater:default",
+    "clipboard-manager:allow-write-text"
   ]
 }
diff --git a/studio/src-tauri/src/commands.rs b/studio/src-tauri/src/commands.rs
index f2acb92e09..48a0af6e48 100644
--- a/studio/src-tauri/src/commands.rs
+++ b/studio/src-tauri/src/commands.rs
@@ -1,9 +1,19 @@
+use crate::diagnostics::{self, DiagnosticsState};
 use crate::install;
 use crate::process::{self, BackendState, ShutdownFlag};
 use crate::update;
 use log::{error, info, warn};
+use std::time::{Duration, Instant};
 use tauri::{AppHandle, Emitter};
 
+const BACKEND_STARTUP_GRACE_PERIOD: Duration = Duration::from_secs(5 * 60);
+const HEALTH_WATCHDOG_INTERVAL: Duration = Duration::from_secs(15);
+const HEALTH_WATCHDOG_MAX_FAILURES: u32 = 3;
+
+fn should_count_watchdog_failure(has_seen_healthy: bool, elapsed_since_start: Duration) -> bool {
+    has_seen_healthy || elapsed_since_start >= BACKEND_STARTUP_GRACE_PERIOD
+}
+
 async fn managed_install_ready_after_repair() -> bool {
     crate::preflight::managed_install_ready().await
 }
@@ -13,8 +23,12 @@ fn should_emit_repair_failed(msg: &str) -> bool {
 }
 
 #[tauri::command]
-pub async fn desktop_preflight() -> crate::preflight::DesktopPreflightResult {
-    crate::preflight::desktop_preflight_result().await
+pub async fn desktop_preflight(
+    diagnostics: tauri::State<'_, DiagnosticsState>,
+) -> Result<crate::preflight::DesktopPreflightResult, String> {
+    let result = crate::preflight::desktop_preflight_result().await;
+    diagnostics::record_preflight(&diagnostics, &result);
+    Ok(result)
 }
 
 /// Check if unsloth is installed AND functional.
@@ -82,11 +96,13 @@ pub async fn start_server(
     app: AppHandle,
     state: tauri::State<'_, BackendState>,
     shutdown: tauri::State<'_, ShutdownFlag>,
+    diagnostics: tauri::State<'_, DiagnosticsState>,
     port: u16,
 ) -> Result<(), String> {
     info!("start_server command called with port {}", port);
 
-    process::start_backend(&app, &state, port, &shutdown)?;
+    let diagnostics_state = diagnostics.inner().clone();
+    let generation = process::start_backend(&app, &state, port, &shutdown, &diagnostics_state)?;
 
     // Spawn health watchdog for the owned backend — detects
     // deadlocks and hangs that stdout-based crash detection misses.
@@ -94,7 +110,14 @@ pub async fn start_server(
     let watchdog_shutdown = shutdown.inner().clone();
     let watchdog_app = app.clone();
     tokio::spawn(async move {
-        health_watchdog(watchdog_app, watchdog_state, watchdog_shutdown).await;
+        health_watchdog(
+            watchdog_app,
+            watchdog_state,
+            watchdog_shutdown,
+            diagnostics_state,
+            generation,
+        )
+        .await;
     });
 
     Ok(())
@@ -106,16 +129,25 @@ pub async fn start_managed_server(
     app: AppHandle,
     state: tauri::State<'_, BackendState>,
     shutdown: tauri::State<'_, ShutdownFlag>,
+    diagnostics: tauri::State<'_, DiagnosticsState>,
     port: u16,
 ) -> Result<(), String> {
     info!("start_managed_server command called with port {}", port);
-    process::start_backend(&app, &state, port, &shutdown)?;
+    let diagnostics_state = diagnostics.inner().clone();
+    let generation = process::start_backend(&app, &state, port, &shutdown, &diagnostics_state)?;
 
     let watchdog_state = state.inner().clone();
     let watchdog_shutdown = shutdown.inner().clone();
     let watchdog_app = app.clone();
     tokio::spawn(async move {
-        health_watchdog(watchdog_app, watchdog_state, watchdog_shutdown).await;
+        health_watchdog(
+            watchdog_app,
+            watchdog_state,
+            watchdog_shutdown,
+            diagnostics_state,
+            generation,
+        )
+        .await;
     });
 
     Ok(())
@@ -128,9 +160,10 @@ pub async fn start_managed_server(
 pub fn stop_server(
     state: tauri::State<'_, BackendState>,
     shutdown: tauri::State<'_, ShutdownFlag>,
+    diagnostics: tauri::State<'_, DiagnosticsState>,
 ) -> Result<(), String> {
     info!("stop_server command called");
-    process::stop_backend(&state, &shutdown)
+    process::stop_backend(&state, &shutdown, Some(diagnostics.inner()))
 }
 
 /// Check if a healthy Unsloth backend is running on the given port.
@@ -201,13 +234,25 @@ pub fn open_logs_dir() -> Result<(), String> {
 pub async fn start_install(
     app: AppHandle,
     state: tauri::State<'_, install::InstallState>,
+    diagnostics: tauri::State<'_, DiagnosticsState>,
 ) -> Result<(), String> {
     let state = state.inner().clone();
-    tokio::task::spawn_blocking(move || install::run_install(app, state))
+    let diagnostics_state = diagnostics.inner().clone();
+    tokio::task::spawn_blocking(move || install::run_install(app, state, diagnostics_state))
         .await
         .map_err(|e| format!("Install task panicked: {e}"))?
 }
 
+/// Record that the user canceled a pending system-package elevation flow.
+#[tauri::command]
+pub fn cancel_pending_elevation(
+    state: tauri::State<'_, install::InstallState>,
+    diagnostics: tauri::State<'_, DiagnosticsState>,
+) -> Result<(), String> {
+    let _ = install::record_pending_elevation_canceled(&state, diagnostics.inner());
+    Ok(())
+}
+
 /// Install system packages with elevated permissions (Linux only).
 /// Called by frontend after user approves the elevation dialog.
 /// Only allows packages that the install script reported as needed.
@@ -216,6 +261,7 @@ pub async fn start_install(
 pub fn install_system_packages(
     packages: Vec<String>,
     state: tauri::State<'_, install::InstallState>,
+    diagnostics: tauri::State<'_, DiagnosticsState>,
 ) -> Result<(), String> {
     // Cross-check against the packages the install script actually reported
     let allowed = state
@@ -230,7 +276,7 @@ pub fn install_system_packages(
             ));
         }
     }
-    install::install_system_packages(&packages)
+    install::install_system_packages(&packages, &state, diagnostics.inner())
 }
 
 /// Stub for non-Linux platforms — elevation is handled by the scripts themselves.
@@ -239,6 +285,7 @@ pub fn install_system_packages(
 pub fn install_system_packages(
     _packages: Vec<String>,
     _state: tauri::State<'_, install::InstallState>,
+    _diagnostics: tauri::State<'_, DiagnosticsState>,
 ) -> Result<(), String> {
     Err("Elevated package install is only supported on Linux".to_string())
 }
@@ -252,6 +299,7 @@ pub async fn start_backend_update(
     shutdown: tauri::State<'_, ShutdownFlag>,
     update_state: tauri::State<'_, update::UpdateState>,
     install_state: tauri::State<'_, install::InstallState>,
+    diagnostics: tauri::State<'_, DiagnosticsState>,
 ) -> Result<(), String> {
     info!("start_backend_update command called");
 
@@ -285,12 +333,13 @@ pub async fn start_backend_update(
         .unwrap_or(false)
     {
         info!("Stopping backend before update...");
-        process::stop_backend(&backend_state, &shutdown)?;
+        process::stop_backend(&backend_state, &shutdown, Some(diagnostics.inner()))?;
     }
 
     // Run update in a blocking thread
     let state = update_state.inner().clone();
-    tokio::task::spawn_blocking(move || update::run_backend_update(app, state))
+    let diagnostics_state = diagnostics.inner().clone();
+    tokio::task::spawn_blocking(move || update::run_backend_update(app, state, diagnostics_state))
         .await
         .map_err(|e| format!("Update task panicked: {e}"))?
 }
@@ -303,6 +352,7 @@ pub async fn start_managed_repair(
     shutdown: tauri::State<'_, ShutdownFlag>,
     update_state: tauri::State<'_, update::UpdateState>,
     install_state: tauri::State<'_, install::InstallState>,
+    diagnostics: tauri::State<'_, DiagnosticsState>,
 ) -> Result<(), String> {
     info!("start_managed_repair command called");
 
@@ -322,6 +372,10 @@ pub async fn start_managed_repair(
         return Err("Repair is already running.".to_string());
     }
 
+    let diagnostics_state = diagnostics.inner().clone();
+    let repair_group_id = install::take_pending_repair_group_for_resume(&install_state)
+        .unwrap_or_else(|| diagnostics::begin_repair_group(&diagnostics_state));
+
     shutdown.store(true, std::sync::atomic::Ordering::SeqCst);
 
     if backend_state
@@ -330,14 +384,21 @@ pub async fn start_managed_repair(
         .unwrap_or(false)
     {
         info!("Stopping backend before repair...");
-        process::stop_backend(&backend_state, &shutdown)?;
+        process::stop_backend(&backend_state, &shutdown, Some(&diagnostics_state))?;
     }
 
     let _ = app.emit("repair-progress", "Updating existing Studio install...");
     let update_app = app.clone();
     let update_state = update_state.inner().clone();
+    let update_diagnostics = diagnostics_state.clone();
+    let update_repair_group_id = repair_group_id.clone();
     let update_result = tokio::task::spawn_blocking(move || {
-        update::run_backend_update_for_repair(update_app, update_state)
+        update::run_backend_update_for_repair(
+            update_app,
+            update_state,
+            update_diagnostics,
+            update_repair_group_id,
+        )
     })
     .await
     .map_err(|e| format!("Repair update task panicked: {e}"))?;
@@ -345,6 +406,7 @@ pub async fn start_managed_repair(
     match update_result {
         Ok(()) if managed_install_ready_after_repair().await => {
             info!("Managed repair complete after update");
+            diagnostics::finish_repair_group(&diagnostics_state, &repair_group_id, "success", None);
             let _ = app.emit("repair-complete", ());
             return Ok(());
         }
@@ -358,6 +420,12 @@ pub async fn start_managed_repair(
         Err(msg) => {
             if msg.to_ascii_lowercase().contains("already running") {
                 error!("Managed repair update conflict: {}", msg);
+                diagnostics::finish_repair_group(
+                    &diagnostics_state,
+                    &repair_group_id,
+                    "failed",
+                    Some(msg.clone()),
+                );
                 let _ = app.emit("repair-failed", &msg);
                 return Err(msg);
             }
@@ -375,13 +443,30 @@ pub async fn start_managed_repair(
 
     let install_app = app.clone();
     let install_state = install_state.inner().clone();
+    let install_diagnostics = diagnostics_state.clone();
+    let install_repair_group_id = repair_group_id.clone();
     let install_result = tokio::task::spawn_blocking(move || {
-        install::run_install_for_repair(install_app, install_state)
+        install::run_install_for_repair(
+            install_app,
+            install_state,
+            install_diagnostics,
+            install_repair_group_id,
+        )
     })
     .await
     .map_err(|e| format!("Repair install task panicked: {e}"))?;
 
     if let Err(msg) = install_result {
+        diagnostics::finish_repair_group(
+            &diagnostics_state,
+            &repair_group_id,
+            if msg == "NEEDS_ELEVATION" {
+                "needs_elevation"
+            } else {
+                "failed"
+            },
+            Some(msg.clone()),
+        );
         if should_emit_repair_failed(&msg) {
             error!("Managed repair installer failed: {}", msg);
             let _ = app.emit("repair-failed", &msg);
@@ -391,18 +476,27 @@ pub async fn start_managed_repair(
 
     if managed_install_ready_after_repair().await {
         info!("Managed repair complete after installer");
+        diagnostics::finish_repair_group(&diagnostics_state, &repair_group_id, "success", None);
         let _ = app.emit("repair-complete", ());
         return Ok(());
     }
 
     let msg = "Repair finished, but Studio install is still not desktop-ready.".to_string();
     error!("{}", msg);
+    diagnostics::finish_repair_group(
+        &diagnostics_state,
+        &repair_group_id,
+        "failed",
+        Some(msg.clone()),
+    );
     let _ = app.emit("repair-failed", &msg);
     Err(msg)
 }
 
 #[cfg(test)]
 mod tests {
+    use std::time::Duration;
+
     #[test]
     fn repair_elevation_is_not_a_terminal_repair_failure() {
         assert!(!super::should_emit_repair_failed("NEEDS_ELEVATION"));
@@ -410,63 +504,132 @@ mod tests {
             "Installer exited with code 1"
         ));
     }
+
+    #[test]
+    fn watchdog_ignores_startup_failures_within_grace_period() {
+        assert!(!super::should_count_watchdog_failure(
+            false,
+            super::BACKEND_STARTUP_GRACE_PERIOD - Duration::from_secs(1)
+        ));
+    }
+
+    #[test]
+    fn watchdog_counts_failures_after_backend_was_healthy() {
+        assert!(super::should_count_watchdog_failure(
+            true,
+            Duration::from_secs(1)
+        ));
+    }
+
+    #[test]
+    fn watchdog_counts_startup_failures_after_grace_period() {
+        assert!(super::should_count_watchdog_failure(
+            false,
+            super::BACKEND_STARTUP_GRACE_PERIOD
+        ));
+    }
 }
 
 /// Periodic health check that detects deadlocked or hung backends.
-/// Starts 30s after the backend is launched (to allow initial startup),
-/// then pings /api/health every 15s. After 3 consecutive failures (45s)
-/// with the process still alive, emits `server-crashed` so the frontend
-/// can offer a restart.
-async fn health_watchdog(app: AppHandle, state: BackendState, shutdown: ShutdownFlag) {
+/// During startup, failures are ignored for a generous grace period so a slow
+/// but legitimate backend boot is not killed. After the backend has answered at
+/// least once, or after the startup grace expires, 3 consecutive failed checks
+/// emit `server-crashed` so the frontend can offer a restart.
+async fn health_watchdog(
+    app: AppHandle,
+    state: BackendState,
+    shutdown: ShutdownFlag,
+    diagnostics: DiagnosticsState,
+    generation: u64,
+) {
     use std::sync::atomic::Ordering;
 
-    // Give the backend time to start up
-    tokio::time::sleep(std::time::Duration::from_secs(30)).await;
-
+    let started_at = Instant::now();
     let mut consecutive_failures: u32 = 0;
+    let mut has_seen_healthy = false;
 
     loop {
-        tokio::time::sleep(std::time::Duration::from_secs(15)).await;
+        tokio::time::sleep(HEALTH_WATCHDOG_INTERVAL).await;
 
         if shutdown.load(Ordering::SeqCst) {
             info!("Health watchdog: shutdown flag set, exiting");
             break;
         }
 
-        let (port, has_child) = {
+        let (port, has_child, current_generation) = {
             let proc = match state.lock() {
                 Ok(p) => p,
                 Err(_) => break,
             };
-            (proc.port, proc.child.is_some())
+            (proc.port, proc.child.is_some(), proc.generation)
         };
 
+        if current_generation != generation {
+            info!("Health watchdog: backend generation changed, exiting");
+            break;
+        }
+
         // Stop watching if the backend is gone
         if !has_child {
             info!("Health watchdog: backend stopped, exiting");
             break;
         }
 
+        let should_count_failure =
+            should_count_watchdog_failure(has_seen_healthy, started_at.elapsed());
+
         let Some(port) = port else {
-            continue; // Port not yet known
+            if should_count_failure {
+                consecutive_failures += 1;
+                warn!(
+                    "Health watchdog: backend has not reported a port ({}/{})",
+                    consecutive_failures, HEALTH_WATCHDOG_MAX_FAILURES
+                );
+            }
+
+            if consecutive_failures >= HEALTH_WATCHDOG_MAX_FAILURES {
+                error!(
+                    "Health watchdog: backend never reported a port, killing and declaring dead"
+                );
+                diagnostics::record_backend_watchdog(
+                    &diagnostics,
+                    generation,
+                    "no_port_after_grace",
+                );
+                let _ = process::stop_backend(&state, &shutdown, Some(&diagnostics));
+                let _ = app.emit("server-crashed", ());
+                break;
+            }
+
+            continue;
         };
 
         match check_health_inner(port).await {
             Ok(true) => {
+                has_seen_healthy = true;
                 consecutive_failures = 0;
             }
+            _ if !should_count_failure => {
+                info!(
+                    "Health watchdog: startup health check failed on port {} before grace period elapsed",
+                    port
+                );
+            }
             _ => {
                 consecutive_failures += 1;
                 warn!(
-                    "Health watchdog: failure {}/3 on port {}",
-                    consecutive_failures, port
+                    "Health watchdog: failure {}/{} on port {}",
+                    consecutive_failures, HEALTH_WATCHDOG_MAX_FAILURES, port
                 );
-                if consecutive_failures >= 3 {
-                    error!(
-                        "Health watchdog: backend unresponsive for 45s, killing and declaring dead"
+                if consecutive_failures >= HEALTH_WATCHDOG_MAX_FAILURES {
+                    error!("Health watchdog: backend unresponsive, killing and declaring dead");
+                    diagnostics::record_backend_watchdog(
+                        &diagnostics,
+                        generation,
+                        "unresponsive_health_check",
                     );
                     // Kill the zombie process so retry can start fresh
-                    let _ = process::stop_backend(&state, &shutdown);
+                    let _ = process::stop_backend(&state, &shutdown, Some(&diagnostics));
                     let _ = app.emit("server-crashed", ());
                     break;
                 }
diff --git a/studio/src-tauri/src/desktop_auth.rs b/studio/src-tauri/src/desktop_auth.rs
index df09df7a14..483e7c0432 100644
--- a/studio/src-tauri/src/desktop_auth.rs
+++ b/studio/src-tauri/src/desktop_auth.rs
@@ -1,6 +1,7 @@
+use crate::diagnostics::{self, DiagnosticsState};
 use crate::preflight::{DesktopPreflightDisposition, DesktopPreflightResult};
 use crate::process::BackendState;
-use log::info;
+use log::{info, warn};
 use reqwest::Client;
 use serde::{Deserialize, Serialize};
 use std::path::{Path, PathBuf};
@@ -39,13 +40,16 @@ struct BackendPort {
 #[derive(Debug)]
 enum AuthError {
     Connectivity(String),
+    StaleResponder(String),
     Failed(String),
 }
 
 impl AuthError {
     fn message(self) -> String {
         match self {
-            Self::Connectivity(message) | Self::Failed(message) => message,
+            Self::Connectivity(message) | Self::StaleResponder(message) | Self::Failed(message) => {
+                message
+            }
         }
     }
 }
@@ -73,11 +77,14 @@ fn read_secret_if_exists(path: &Path) -> Result<Option<String>, String> {
     match std::fs::read_to_string(path) {
         Ok(s) => Ok(Some(s.trim().to_string())),
         Err(e) if e.kind() == std::io::ErrorKind::NotFound => Ok(None),
-        Err(e) => Err(format!(
-            "Failed to read auth secret at {}: {}",
-            path.display(),
-            e
-        )),
+        Err(e) => {
+            warn!(
+                "Desktop auth: ignoring unreadable auth secret at {}: {}",
+                path.display(),
+                e
+            );
+            Ok(None)
+        }
     }
 }
 
@@ -138,7 +145,10 @@ fn classify_auth_send_error(error: reqwest::Error) -> AuthError {
 fn should_retry_with_discovered_port(source: PortSource, error: &AuthError) -> bool {
     matches!(
         (source, error),
-        (PortSource::Cached, AuthError::Connectivity(_))
+        (
+            PortSource::Cached,
+            AuthError::Connectivity(_) | AuthError::StaleResponder(_)
+        )
     )
 }
 
@@ -157,7 +167,7 @@ async fn exchange_desktop_secret(
         .map_err(classify_auth_send_error)?;
 
     if response.status() == reqwest::StatusCode::NOT_FOUND {
-        return Err(AuthError::Failed(
+        return Err(AuthError::StaleResponder(
             "Running Studio backend is too old for this desktop app. Update that backend and restart."
                 .to_string(),
         ));
@@ -166,7 +176,7 @@ async fn exchange_desktop_secret(
         return Ok(None);
     }
     if !response.status().is_success() {
-        return Err(AuthError::Failed("Desktop auth failed".to_string()));
+        return Err(AuthError::StaleResponder("Desktop auth failed".to_string()));
     }
 
     response
@@ -213,6 +223,33 @@ async fn provision_desktop_auth() -> Result<(), String> {
     ))
 }
 
+async fn retry_on_discovered_port(
+    client: &Client,
+    state: &tauri::State<'_, BackendState>,
+    previous: BackendPort,
+    secret: &str,
+) -> Result<Option<(Option<DesktopAuthResponse>, BackendPort)>, String> {
+    if previous.source != PortSource::Cached {
+        return Ok(None);
+    }
+    let Some(port) = discover_compatible_backend_port().await else {
+        return Ok(None);
+    };
+    if port == previous.port {
+        return Ok(None);
+    }
+
+    update_backend_port(state, port)?;
+    let backend = BackendPort {
+        port,
+        source: PortSource::Discovered,
+    };
+    exchange_desktop_secret(client, port, secret)
+        .await
+        .map(|tokens| Some((tokens, backend)))
+        .map_err(AuthError::message)
+}
+
 async fn authenticate_with_stale_port_retry(
     client: &Client,
     state: &tauri::State<'_, BackendState>,
@@ -220,20 +257,19 @@ async fn authenticate_with_stale_port_retry(
     secret: &str,
 ) -> Result<(Option<DesktopAuthResponse>, BackendPort), String> {
     match exchange_desktop_secret(client, backend.port, secret).await {
-        Ok(tokens) => Ok((tokens, backend)),
+        Ok(Some(tokens)) => Ok((Some(tokens), backend)),
+        Ok(None) => {
+            // 401 means the backend is reachable but the cached secret is stale.
+            // Keep using the same backend so the next desktop_auth_inner attempt
+            // provisions a new secret for the server we are retrying instead of
+            // switching to an unrelated discovered/attached backend.
+            Ok((None, backend))
+        }
         Err(error) if should_retry_with_discovered_port(backend.source, &error) => {
-            let Some(port) = discover_compatible_backend_port().await else {
-                return Err(error.message());
-            };
-            update_backend_port(state, port)?;
-            let backend = BackendPort {
-                port,
-                source: PortSource::Discovered,
-            };
-            exchange_desktop_secret(client, port, secret)
-                .await
-                .map(|tokens| (tokens, backend))
-                .map_err(AuthError::message)
+            if let Some(retried) = retry_on_discovered_port(client, state, backend, secret).await? {
+                return Ok(retried);
+            }
+            Err(error.message())
         }
         Err(error) => Err(error.message()),
     }
@@ -242,9 +278,25 @@ async fn authenticate_with_stale_port_retry(
 #[tauri::command]
 pub async fn desktop_auth(
     state: tauri::State<'_, BackendState>,
+    diagnostics: tauri::State<'_, DiagnosticsState>,
+) -> Result<DesktopAuthResponse, String> {
+    let result = desktop_auth_inner(&state, diagnostics.inner()).await;
+    if let Err(message) = &result {
+        let port = state.lock().ok().and_then(|proc| proc.port);
+        diagnostics::record_auth_failure(&diagnostics, "desktop_auth", port, message);
+    }
+    result
+}
+
+async fn desktop_auth_inner(
+    state: &tauri::State<'_, BackendState>,
+    diagnostics: &DiagnosticsState,
 ) -> Result<DesktopAuthResponse, String> {
     let _auth_guard = DESKTOP_AUTH_LOCK.lock().await;
-    let mut backend = current_backend_port(&state).await?;
+    let mut backend = current_backend_port(state).await?;
+    if backend.source == PortSource::Discovered {
+        diagnostics::record_attached_external_backend(diagnostics, backend.port);
+    }
     let client = Client::builder()
         .timeout(std::time::Duration::from_secs(5))
         .build()
@@ -263,8 +315,11 @@ pub async fn desktop_auth(
 
         info!("Desktop auth: exchanging desktop secret");
         let (tokens, resolved_backend) =
-            authenticate_with_stale_port_retry(&client, &state, backend, &secret).await?;
+            authenticate_with_stale_port_retry(&client, state, backend, &secret).await?;
         backend = resolved_backend;
+        if backend.source == PortSource::Discovered {
+            diagnostics::record_attached_external_backend(diagnostics, backend.port);
+        }
         if let Some(tokens) = tokens {
             return Ok(tokens);
         }
@@ -317,11 +372,15 @@ mod tests {
     }
 
     #[test]
-    fn retry_discovery_only_for_cached_connectivity_errors() {
+    fn retry_discovery_only_for_cached_recoverable_errors() {
         assert!(should_retry_with_discovered_port(
             PortSource::Cached,
             &AuthError::Connectivity("connection refused".to_string())
         ));
+        assert!(should_retry_with_discovered_port(
+            PortSource::Cached,
+            &AuthError::StaleResponder("old responder".to_string())
+        ));
         assert!(!should_retry_with_discovered_port(
             PortSource::Discovered,
             &AuthError::Connectivity("connection refused".to_string())
@@ -332,6 +391,64 @@ mod tests {
         ));
     }
 
+    #[test]
+    fn read_secret_returns_none_for_missing_file() {
+        let path = std::env::temp_dir().join(format!(
+            "unsloth-missing-desktop-secret-{}",
+            std::process::id()
+        ));
+        let _ = std::fs::remove_file(&path);
+
+        assert_eq!(read_secret_if_exists(&path).unwrap(), None);
+    }
+
+    #[test]
+    fn read_secret_trims_existing_file() {
+        let path =
+            std::env::temp_dir().join(format!("unsloth-desktop-secret-{}", std::process::id()));
+        std::fs::write(&path, "  desktop-secret\n").unwrap();
+
+        assert_eq!(
+            read_secret_if_exists(&path).unwrap(),
+            Some("desktop-secret".to_string())
+        );
+        std::fs::remove_file(path).unwrap();
+    }
+
+    #[test]
+    fn read_secret_treats_invalid_utf8_as_missing_for_repair() {
+        let path = std::env::temp_dir().join(format!(
+            "unsloth-invalid-desktop-secret-{}",
+            std::process::id()
+        ));
+        std::fs::write(&path, [0xff, 0xfe]).unwrap();
+
+        assert_eq!(read_secret_if_exists(&path).unwrap(), None);
+        std::fs::remove_file(path).unwrap();
+    }
+
+    #[cfg(unix)]
+    #[test]
+    fn read_secret_treats_permission_denied_as_missing_for_repair() {
+        use std::os::unix::fs::PermissionsExt;
+
+        let path = std::env::temp_dir().join(format!(
+            "unsloth-unreadable-desktop-secret-{}",
+            std::process::id()
+        ));
+        std::fs::write(&path, "desktop-stale").unwrap();
+        std::fs::set_permissions(&path, std::fs::Permissions::from_mode(0o000)).unwrap();
+
+        let result = read_secret_if_exists(&path);
+
+        std::fs::set_permissions(&path, std::fs::Permissions::from_mode(0o600)).unwrap();
+        std::fs::remove_file(path).unwrap();
+        if matches!(result, Ok(Some(_))) {
+            return;
+        }
+        assert_eq!(result.unwrap(), None);
+    }
+
     #[test]
     fn attached_ready_port_requires_attached_ready_with_port() {
         let compatible = DesktopPreflightResult {
diff --git a/studio/src-tauri/src/diagnostics/mod.rs b/studio/src-tauri/src/diagnostics/mod.rs
new file mode 100644
index 0000000000..5ab940af38
--- /dev/null
+++ b/studio/src-tauri/src/diagnostics/mod.rs
@@ -0,0 +1,169 @@
+mod phase_log;
+mod redaction;
+mod report;
+mod state;
+
+use crate::process::BackendState;
+use std::path::PathBuf;
+use std::sync::atomic::{AtomicU64, Ordering};
+use std::time::{Duration, SystemTime, UNIX_EPOCH};
+use tauri::AppHandle;
+
+pub use phase_log::append_phase_line;
+#[cfg(target_os = "linux")]
+pub use phase_log::PhaseLogHandle;
+pub use state::{
+    begin_backend_session, begin_install_attempt, begin_repair_child, begin_repair_group,
+    begin_update_attempt, finish_attempt, finish_repair_group, new_diagnostics_state,
+    record_attached_external_backend, record_auth_failure, record_backend_exit,
+    record_backend_intentional_stop, record_backend_port, record_backend_start_failure,
+    record_backend_watchdog, record_diag_marker, record_elevation_packages, record_preflight,
+    record_progress, record_step, AttemptLog, BackendLog, DiagnosticsState,
+    FrontendSupportSnapshot,
+};
+
+pub const SCHEMA_VERSION: u32 = 1;
+pub const PHASE_LOG_SEGMENT_MAX_BYTES: u64 = 5 * 1024 * 1024;
+pub const PHASE_LOG_MAX_SEGMENTS_PER_GROUP: usize = 3;
+#[allow(dead_code)]
+pub const PHASE_LOG_KEEP_GROUPS_PER_KIND: usize = 5;
+pub const TAIL_MAX_LINES: usize = 1000;
+pub const TAIL_MAX_BYTES: usize = 200 * 1024;
+pub const REPORT_MAX_BYTES: usize = 1024 * 1024;
+
+pub(crate) const MAX_STATE_ITEMS: usize = 200;
+pub(crate) const MAX_PHASE_LINE_BYTES: usize = 16 * 1024;
+pub(crate) const FOOTER_BUDGET_BYTES: usize = 8 * 1024;
+
+static ID_COUNTER: AtomicU64 = AtomicU64::new(1);
+
+pub(crate) fn now_ms() -> u64 {
+    SystemTime::now()
+        .duration_since(UNIX_EPOCH)
+        .unwrap_or_else(|_| Duration::from_secs(0))
+        .as_millis() as u64
+}
+
+pub(crate) fn new_id(prefix: &str) -> String {
+    let n = ID_COUNTER.fetch_add(1, Ordering::Relaxed);
+    format!("{}-{}-{}", sanitize_component(prefix), now_ms(), n)
+}
+
+pub fn studio_dir() -> PathBuf {
+    dirs::home_dir()
+        .unwrap_or_else(|| PathBuf::from("."))
+        .join(".unsloth")
+        .join("studio")
+}
+
+pub fn logs_dir() -> PathBuf {
+    studio_dir().join("logs")
+}
+
+pub(crate) fn cap_string(value: &str, max: usize) -> String {
+    if value.len() <= max {
+        return value.to_string();
+    }
+    let boundary = valid_utf8_boundary(value, max);
+    format!("{} [truncated]", &value[..boundary])
+}
+
+pub(crate) fn valid_utf8_boundary(s: &str, max: usize) -> usize {
+    if max >= s.len() {
+        return s.len();
+    }
+    let mut index = max;
+    while index > 0 && !s.is_char_boundary(index) {
+        index -= 1;
+    }
+    index
+}
+
+pub(crate) fn sanitize_component(value: &str) -> String {
+    value
+        .chars()
+        .map(|ch| {
+            if ch.is_ascii_alphanumeric() || ch == '-' || ch == '_' || ch == '.' {
+                ch
+            } else {
+                '_'
+            }
+        })
+        .collect()
+}
+
+#[tauri::command]
+pub async fn collect_support_diagnostics(
+    _app: AppHandle,
+    diagnostics: tauri::State<'_, DiagnosticsState>,
+    backend: tauri::State<'_, BackendState>,
+    snapshot: FrontendSupportSnapshot,
+) -> Result<String, String> {
+    let diagnostics_snapshot = state::clone_snapshot(diagnostics.inner());
+    let backend_port = backend.lock().ok().and_then(|proc| proc.port).or_else(|| {
+        diagnostics_snapshot
+            .backend
+            .as_ref()
+            .and_then(|backend| backend.reported_port.or(backend.requested_port))
+    });
+
+    let health = match backend_port {
+        Some(port) => Some(collect_backend_health(port).await),
+        None => None,
+    };
+
+    match tokio::task::spawn_blocking(move || {
+        report::render_report(diagnostics_snapshot, snapshot, health)
+    })
+    .await
+    {
+        Ok(report) => Ok(report),
+        Err(error) => Ok(format!(
+            "DEGRADED RUST DIAGNOSTICS\ncollection_warning=report task failed: {error}\n"
+        )),
+    }
+}
+
+async fn collect_backend_health(port: u16) -> report::BackendHealthSection {
+    let mut section = report::BackendHealthSection {
+        port,
+        fields: Vec::new(),
+        warning: None,
+    };
+    let client = match reqwest::Client::builder()
+        .timeout(Duration::from_millis(750))
+        .build()
+    {
+        Ok(client) => client,
+        Err(error) => {
+            section.warning = Some(format!("health client unavailable: {error}"));
+            return section;
+        }
+    };
+    let url = format!("http://127.0.0.1:{port}/api/health");
+    match client.get(url).send().await {
+        Ok(response) => match response.json::<serde_json::Value>().await {
+            Ok(json) => {
+                for key in [
+                    "status",
+                    "service",
+                    "version",
+                    "device_type",
+                    "chat_only",
+                    "desktop_protocol_version",
+                    "supports_api_only",
+                    "supports_desktop_auth",
+                ] {
+                    if let Some(value) = json.get(key) {
+                        section
+                            .fields
+                            .push((key.to_string(), report::selected_json_value(value)));
+                    }
+                }
+            }
+            Err(error) => section.warning = Some(format!("health json unavailable: {error}")),
+        },
+        Err(error) => section.warning = Some(format!("health request unavailable: {error}")),
+    }
+    section
+}
diff --git a/studio/src-tauri/src/diagnostics/phase_log.rs b/studio/src-tauri/src/diagnostics/phase_log.rs
new file mode 100644
index 0000000000..a1e0c5424c
--- /dev/null
+++ b/studio/src-tauri/src/diagnostics/phase_log.rs
@@ -0,0 +1,823 @@
+use log::warn;
+use std::collections::{BTreeMap, HashMap};
+use std::fs::{self, File, OpenOptions};
+use std::io::{Read, Write};
+use std::path::{Path, PathBuf};
+use std::sync::{Arc, Mutex};
+use std::time::UNIX_EPOCH;
+
+use super::state::{
+    forget_segment_path, record_segment_path, DiagnosticsSnapshot, DiagnosticsState,
+};
+use super::{
+    logs_dir, now_ms, sanitize_component, valid_utf8_boundary, MAX_PHASE_LINE_BYTES,
+    PHASE_LOG_KEEP_GROUPS_PER_KIND, PHASE_LOG_MAX_SEGMENTS_PER_GROUP, PHASE_LOG_SEGMENT_MAX_BYTES,
+    SCHEMA_VERSION,
+};
+
+pub type PhaseLogHandle = Arc<Mutex<PhaseLogWriter>>;
+
+#[derive(Clone, Debug)]
+pub(crate) enum SegmentOwner {
+    Install {
+        attempt_id: String,
+    },
+    Update {
+        attempt_id: String,
+    },
+    Repair {
+        repair_group_id: String,
+        attempt_id: String,
+        child_type: String,
+    },
+    Backend {
+        session_id: String,
+    },
+}
+
+pub struct PhaseLogWriter {
+    owner: SegmentOwner,
+    flow: String,
+    started_at_ms: u64,
+    backend_kind: Option<String>,
+    log_dir: PathBuf,
+    segment_index: usize,
+    current_bytes: u64,
+    segments: Vec<PathBuf>,
+    file: Option<File>,
+    disabled: bool,
+    disable_reason: Option<String>,
+    diagnostics: DiagnosticsState,
+    max_bytes: u64,
+    max_segments: usize,
+}
+
+impl PhaseLogWriter {
+    pub(crate) fn new(
+        diagnostics: DiagnosticsState,
+        owner: SegmentOwner,
+        flow: impl Into<String>,
+        started_at_ms: u64,
+        backend_kind: Option<String>,
+    ) -> Self {
+        let mut writer = Self {
+            owner,
+            flow: flow.into(),
+            started_at_ms,
+            backend_kind,
+            log_dir: logs_dir(),
+            segment_index: 1,
+            current_bytes: 0,
+            segments: Vec::new(),
+            file: None,
+            disabled: false,
+            disable_reason: None,
+            diagnostics,
+            max_bytes: PHASE_LOG_SEGMENT_MAX_BYTES,
+            max_segments: PHASE_LOG_MAX_SEGMENTS_PER_GROUP,
+        };
+        writer.open_segment();
+        writer
+    }
+
+    #[cfg(test)]
+    fn new_for_test(
+        diagnostics: DiagnosticsState,
+        owner: SegmentOwner,
+        flow: impl Into<String>,
+        started_at_ms: u64,
+        log_dir: PathBuf,
+        max_bytes: u64,
+        max_segments: usize,
+    ) -> Self {
+        let mut writer = Self {
+            owner,
+            flow: flow.into(),
+            started_at_ms,
+            backend_kind: None,
+            log_dir,
+            segment_index: 1,
+            current_bytes: 0,
+            segments: Vec::new(),
+            file: None,
+            disabled: false,
+            disable_reason: None,
+            diagnostics,
+            max_bytes,
+            max_segments,
+        };
+        writer.open_segment();
+        writer
+    }
+
+    fn open_segment(&mut self) {
+        if self.disabled {
+            return;
+        }
+        if let Err(error) = ensure_logs_dir_at(&self.log_dir) {
+            self.disable(format!("failed to create diagnostics log dir: {error}"));
+            return;
+        }
+        if self.segment_index == 1 {
+            cleanup_retention(&self.log_dir, &self.flow, &owner_group_key(&self.owner));
+        }
+
+        let path = self.segment_path(self.segment_index);
+        if path_has_symlink(&path) {
+            self.disable(format!(
+                "refusing to write diagnostics phase log through symlink: {}",
+                path.display()
+            ));
+            return;
+        }
+
+        let mut options = OpenOptions::new();
+        options.create(true).append(true);
+        #[cfg(unix)]
+        {
+            use std::os::unix::fs::OpenOptionsExt;
+            options.mode(0o600);
+        }
+
+        match options.open(&path) {
+            Ok(mut file) => {
+                let header = self.header();
+                if let Err(error) = file.write_all(header.as_bytes()) {
+                    self.disable(format!(
+                        "failed to write diagnostics phase log header to {}: {error}",
+                        path.display()
+                    ));
+                    return;
+                }
+                self.current_bytes = header.len() as u64;
+                self.segments.push(path.clone());
+                self.file = Some(file);
+                record_segment_path(&self.diagnostics, &self.owner, path);
+            }
+            Err(error) => self.disable(format!(
+                "failed to open diagnostics phase log {}: {error}",
+                path.display()
+            )),
+        }
+    }
+
+    fn header(&self) -> String {
+        let mut lines = vec![
+            "# Unsloth Studio diagnostics phase log".to_string(),
+            format!("diag_schema={SCHEMA_VERSION}"),
+            format!("flow={}", self.flow),
+            format!("segment_index={}", self.segment_index),
+            format!("started_at_ms={}", self.started_at_ms),
+        ];
+        match &self.owner {
+            SegmentOwner::Install { attempt_id }
+            | SegmentOwner::Update { attempt_id }
+            | SegmentOwner::Repair { attempt_id, .. } => {
+                lines.push(format!("attempt_id={attempt_id}"));
+            }
+            SegmentOwner::Backend { session_id } => {
+                lines.push(format!("attempt_id={session_id}"));
+                lines.push(format!("backend_session_id={session_id}"));
+            }
+        }
+        if let SegmentOwner::Repair {
+            repair_group_id,
+            child_type,
+            ..
+        } = &self.owner
+        {
+            lines.push(format!("repair_group_id={repair_group_id}"));
+            lines.push(format!("child_type={child_type}"));
+        }
+        if let Some(kind) = &self.backend_kind {
+            lines.push(format!("backend_kind={kind}"));
+        }
+        lines.push("--- log ---".to_string());
+        lines.push(String::new());
+        lines.join("\n")
+    }
+
+    fn segment_path(&self, segment_index: usize) -> PathBuf {
+        let name = match &self.owner {
+            SegmentOwner::Install { attempt_id } => {
+                format!(
+                    "install-{}-s{segment_index:02}.log",
+                    sanitize_component(attempt_id)
+                )
+            }
+            SegmentOwner::Update { attempt_id } => {
+                format!(
+                    "update-{}-s{segment_index:02}.log",
+                    sanitize_component(attempt_id)
+                )
+            }
+            SegmentOwner::Repair {
+                repair_group_id,
+                attempt_id,
+                child_type,
+            } => format!(
+                "repair-{}-{}-{}-s{segment_index:02}.log",
+                sanitize_component(repair_group_id),
+                sanitize_component(child_type),
+                sanitize_component(attempt_id)
+            ),
+            SegmentOwner::Backend { session_id } => {
+                format!(
+                    "backend-{}-s{segment_index:02}.log",
+                    sanitize_component(session_id)
+                )
+            }
+        };
+        self.log_dir.join(name)
+    }
+
+    fn append_line(&mut self, stream: &str, text: &str) {
+        if self.disabled {
+            return;
+        }
+        let mut line = text.replace('\r', "");
+        let mut truncated = false;
+        if line.len() > MAX_PHASE_LINE_BYTES {
+            line.truncate(valid_utf8_boundary(&line, MAX_PHASE_LINE_BYTES));
+            truncated = true;
+        }
+        if truncated {
+            line.push_str(" [line truncated]");
+        }
+        let entry = format!("[{}][{}] {}\n", now_ms(), stream, line);
+        if self.current_bytes + entry.len() as u64 > self.max_bytes {
+            self.rotate();
+        }
+        let Some(file) = self.file.as_mut() else {
+            return;
+        };
+        match file.write_all(entry.as_bytes()) {
+            Ok(()) => {
+                self.current_bytes += entry.len() as u64;
+            }
+            Err(error) => self.disable(format!("diagnostics phase log write failed: {error}")),
+        }
+    }
+
+    fn rotate(&mut self) {
+        self.file.take();
+        self.segment_index += 1;
+        self.current_bytes = 0;
+        self.open_segment();
+        self.prune_old_segments();
+    }
+
+    fn prune_old_segments(&mut self) {
+        while self.segments.len() > self.max_segments {
+            let old_path = self.segments.remove(0);
+            forget_segment_path(&self.diagnostics, &self.owner, &old_path);
+            safe_delete_phase_log(&self.log_dir, &old_path);
+        }
+    }
+
+    fn disable(&mut self, reason: String) {
+        if !self.disabled {
+            warn!("{}", reason);
+        }
+        self.disabled = true;
+        self.disable_reason = Some(reason);
+        self.file.take();
+    }
+}
+
+#[allow(dead_code)]
+pub fn ensure_logs_dir() -> std::io::Result<PathBuf> {
+    let dir = logs_dir();
+    ensure_logs_dir_at(&dir)?;
+    Ok(dir)
+}
+
+pub(crate) fn ensure_logs_dir_at(dir: &Path) -> std::io::Result<()> {
+    fs::create_dir_all(dir)?;
+    if fs::symlink_metadata(dir)
+        .map(|metadata| metadata.file_type().is_symlink())
+        .unwrap_or(false)
+    {
+        return Err(std::io::Error::new(
+            std::io::ErrorKind::Other,
+            "refusing diagnostics logs dir symlink",
+        ));
+    }
+    #[cfg(unix)]
+    {
+        use std::os::unix::fs::PermissionsExt;
+        let _ = fs::set_permissions(dir, fs::Permissions::from_mode(0o700));
+    }
+    Ok(())
+}
+
+pub fn append_phase_line(handle: &PhaseLogHandle, stream: &str, text: &str) {
+    if let Ok(mut writer) = handle.lock() {
+        writer.append_line(stream, text);
+    }
+}
+#[derive(Debug)]
+pub(crate) struct PhaseGroup {
+    pub(crate) label: String,
+    pub(crate) source: String,
+    pub(crate) paths: Vec<PathBuf>,
+    pub(crate) note: Option<String>,
+}
+
+pub(crate) fn collect_phase_groups(
+    snapshot: &DiagnosticsSnapshot,
+    warnings: &mut Vec<String>,
+) -> Vec<PhaseGroup> {
+    let mut groups = Vec::new();
+    for install in snapshot.install_history.iter().rev().take(4).rev() {
+        if !install.log_segments.is_empty() {
+            groups.push(PhaseGroup {
+                label: format!("install:{}", install.attempt_id),
+                source: "live-state-history".to_string(),
+                paths: install.log_segments.clone(),
+                note: Some("previous linked install attempt retained in memory".to_string()),
+            });
+        }
+    }
+    if let Some(install) = &snapshot.install {
+        if !install.log_segments.is_empty() {
+            groups.push(PhaseGroup {
+                label: format!("install:{}", install.attempt_id),
+                source: "live-state".to_string(),
+                paths: install.log_segments.clone(),
+                note: None,
+            });
+        }
+    }
+    if let Some(update) = &snapshot.update {
+        if !update.log_segments.is_empty() {
+            groups.push(PhaseGroup {
+                label: format!("update:{}", update.attempt_id),
+                source: "live-state".to_string(),
+                paths: update.log_segments.clone(),
+                note: None,
+            });
+        }
+    }
+    for group in snapshot.repair_groups.iter().rev().take(5).rev() {
+        let mut paths = Vec::new();
+        for child in &group.children {
+            paths.extend(child.log_segments.clone());
+        }
+        if !paths.is_empty() {
+            groups.push(PhaseGroup {
+                label: format!("repair:{}", group.repair_group_id),
+                source: "live-state".to_string(),
+                paths,
+                note: None,
+            });
+        }
+    }
+    if let Some(backend) = &snapshot.backend {
+        if backend.backend_kind == "managed" && !backend.log_segments.is_empty() {
+            groups.push(PhaseGroup {
+                label: backend
+                    .session_id
+                    .as_ref()
+                    .map(|id| format!("backend:{id}"))
+                    .unwrap_or_else(|| "backend:unknown".to_string()),
+                source: "live-state".to_string(),
+                paths: backend.log_segments.clone(),
+                note: None,
+            });
+        } else if backend.backend_kind == "attached_external" {
+            groups.push(PhaseGroup {
+                label: "backend:attached_external".to_string(),
+                source: "live-state".to_string(),
+                paths: Vec::new(),
+                note: Some("Tauri does not own attached/external backend process logs".to_string()),
+            });
+        }
+    }
+
+    let has_install = groups.iter().any(|g| g.label.starts_with("install:"));
+    let has_update = groups.iter().any(|g| g.label.starts_with("update:"));
+    let has_repair = groups.iter().any(|g| g.label.starts_with("repair:"));
+    let has_backend = groups.iter().any(|g| g.label.starts_with("backend:"));
+    let fallback = scan_phase_logs(warnings);
+    for flow in ["install", "update", "repair", "backend"] {
+        let already_live = match flow {
+            "install" => has_install,
+            "update" => has_update,
+            "repair" => has_repair,
+            "backend" => has_backend,
+            _ => false,
+        };
+        if already_live {
+            continue;
+        }
+        if let Some(group) = latest_disk_group(&fallback, flow) {
+            groups.push(group);
+        }
+    }
+
+    groups
+}
+
+#[derive(Clone, Debug)]
+struct PhaseLogMeta {
+    path: PathBuf,
+    flow: String,
+    group_key: String,
+    segment_index: usize,
+    started_at_ms: u64,
+    mtime_ms: u64,
+}
+
+fn scan_phase_logs(warnings: &mut Vec<String>) -> Vec<PhaseLogMeta> {
+    scan_phase_logs_in_dir(&logs_dir(), warnings)
+}
+
+fn scan_phase_logs_in_dir(dir: &Path, warnings: &mut Vec<String>) -> Vec<PhaseLogMeta> {
+    let entries = match fs::read_dir(dir) {
+        Ok(entries) => entries,
+        Err(error) => {
+            warnings.push(format!("phase log dir unavailable: {error}"));
+            return Vec::new();
+        }
+    };
+    let mut metas = Vec::new();
+    for entry in entries.flatten() {
+        let path = entry.path();
+        let Some(file_name) = path.file_name().and_then(|n| n.to_str()) else {
+            continue;
+        };
+        if !is_phase_log_name(file_name) {
+            continue;
+        }
+        if path_has_symlink(&path) {
+            warnings.push(format!("ignored phase log symlink: {}", path.display()));
+            continue;
+        }
+        match parse_phase_log_meta(&path, file_name) {
+            Some(meta) => metas.push(meta),
+            None => warnings.push(format!(
+                "could not reconstruct phase log metadata: {}",
+                path.display()
+            )),
+        }
+    }
+    metas
+}
+
+fn safe_delete_phase_log(log_dir: &Path, path: &Path) {
+    if path_has_symlink(path) {
+        return;
+    }
+    let Ok(canonical_dir) = log_dir.canonicalize() else {
+        return;
+    };
+    let Ok(parent) = path.parent().unwrap_or(log_dir).canonicalize() else {
+        return;
+    };
+    if parent != canonical_dir {
+        return;
+    }
+    let Some(name) = path.file_name().and_then(|name| name.to_str()) else {
+        return;
+    };
+    if !is_phase_log_name(name) {
+        return;
+    }
+    let _ = fs::remove_file(path);
+}
+
+fn cleanup_retention(dir: &Path, flow: &str, current_group_key: &str) {
+    let canonical_dir = match dir.canonicalize() {
+        Ok(path) => path,
+        Err(_) => return,
+    };
+    let mut warnings = Vec::new();
+    let metas = scan_phase_logs_in_dir(dir, &mut warnings);
+    let mut groups: BTreeMap<String, Vec<PhaseLogMeta>> = BTreeMap::new();
+    for meta in metas.into_iter().filter(|meta| meta.flow == flow) {
+        groups.entry(meta.group_key.clone()).or_default().push(meta);
+    }
+    let mut ordered: Vec<(String, u64)> = groups
+        .iter()
+        .map(|(key, items)| {
+            let newest = items
+                .iter()
+                .map(|meta| meta.started_at_ms.max(meta.mtime_ms))
+                .max()
+                .unwrap_or(0);
+            (key.clone(), newest)
+        })
+        .collect();
+    ordered.sort_by_key(|(_, newest)| *newest);
+    let removable = ordered.len().saturating_sub(PHASE_LOG_KEEP_GROUPS_PER_KIND);
+    for (group_key, _) in ordered.into_iter().take(removable) {
+        if group_key == current_group_key {
+            continue;
+        }
+        let Some(items) = groups.get(&group_key) else {
+            continue;
+        };
+        for meta in items {
+            let Ok(parent) = meta.path.parent().unwrap_or(dir).canonicalize() else {
+                continue;
+            };
+            if parent != canonical_dir {
+                continue;
+            }
+            safe_delete_phase_log(dir, &meta.path);
+        }
+    }
+}
+
+fn owner_group_key(owner: &SegmentOwner) -> String {
+    match owner {
+        SegmentOwner::Install { attempt_id } => format!("install:{attempt_id}"),
+        SegmentOwner::Update { attempt_id } => format!("update:{attempt_id}"),
+        SegmentOwner::Repair {
+            repair_group_id, ..
+        } => format!("repair:{repair_group_id}"),
+        SegmentOwner::Backend { session_id } => format!("backend:{session_id}"),
+    }
+}
+
+fn parse_phase_log_meta(path: &Path, file_name: &str) -> Option<PhaseLogMeta> {
+    let mut file = File::open(path).ok()?;
+    let mut buf = vec![0u8; 8192];
+    let read = file.read(&mut buf).ok()?;
+    buf.truncate(read);
+    let header = String::from_utf8_lossy(&buf);
+    let mut map = HashMap::new();
+    for line in header.lines().take(32) {
+        if line == "--- log ---" {
+            break;
+        }
+        if let Some((key, value)) = line.split_once('=') {
+            map.insert(key.trim().to_string(), value.trim().to_string());
+        }
+    }
+
+    let flow = map
+        .get("flow")
+        .cloned()
+        .or_else(|| file_name.split('-').next().map(str::to_string))?;
+    let segment_index = map
+        .get("segment_index")
+        .and_then(|v| v.parse::<usize>().ok())
+        .or_else(|| segment_from_filename(file_name))
+        .unwrap_or(1);
+    let started_at_ms = map
+        .get("started_at_ms")
+        .and_then(|v| v.parse::<u64>().ok())
+        .unwrap_or_else(|| file_mtime_ms(path).unwrap_or(0));
+    let mtime_ms = file_mtime_ms(path).unwrap_or(started_at_ms);
+    let group_key = if flow == "repair" {
+        map.get("repair_group_id")
+            .cloned()
+            .or_else(|| repair_group_from_filename(file_name))
+            .map(|id| format!("repair:{id}"))?
+    } else if flow == "backend" {
+        map.get("backend_session_id")
+            .or_else(|| map.get("attempt_id"))
+            .cloned()
+            .or_else(|| attempt_from_filename(file_name, "backend"))
+            .map(|id| format!("backend:{id}"))?
+    } else {
+        map.get("attempt_id")
+            .cloned()
+            .or_else(|| attempt_from_filename(file_name, &flow))
+            .map(|id| format!("{flow}:{id}"))?
+    };
+
+    Some(PhaseLogMeta {
+        path: path.to_path_buf(),
+        flow,
+        group_key,
+        segment_index,
+        started_at_ms,
+        mtime_ms,
+    })
+}
+
+fn latest_disk_group(metas: &[PhaseLogMeta], flow: &str) -> Option<PhaseGroup> {
+    let mut groups: BTreeMap<String, Vec<PhaseLogMeta>> = BTreeMap::new();
+    for meta in metas.iter().filter(|m| m.flow == flow) {
+        groups
+            .entry(meta.group_key.clone())
+            .or_default()
+            .push(meta.clone());
+    }
+    let (_, mut selected) = groups
+        .into_iter()
+        .max_by_key(|(_, items)| items.iter().map(|m| (m.started_at_ms, m.mtime_ms)).max())?;
+    selected.sort_by_key(|m| (m.started_at_ms, m.segment_index, m.path.clone()));
+    let label = selected
+        .first()
+        .map(|m| m.group_key.clone())
+        .unwrap_or_else(|| format!("{flow}:unknown"));
+    Some(PhaseGroup {
+        label,
+        source: "disk-fallback".to_string(),
+        paths: selected.into_iter().map(|m| m.path).collect(),
+        note: Some("reconstructed from phase-log headers/filenames".to_string()),
+    })
+}
+
+pub(crate) fn path_has_symlink(path: &Path) -> bool {
+    fs::symlink_metadata(path)
+        .map(|metadata| metadata.file_type().is_symlink())
+        .unwrap_or(false)
+}
+
+fn file_mtime_ms(path: &Path) -> Option<u64> {
+    fs::metadata(path)
+        .ok()?
+        .modified()
+        .ok()?
+        .duration_since(UNIX_EPOCH)
+        .ok()
+        .map(|d| d.as_millis() as u64)
+}
+
+fn is_phase_log_name(name: &str) -> bool {
+    name.ends_with(".log")
+        && (name.starts_with("install-")
+            || name.starts_with("update-")
+            || name.starts_with("repair-")
+            || name.starts_with("backend-"))
+}
+
+fn segment_from_filename(name: &str) -> Option<usize> {
+    let stem = name.strip_suffix(".log")?;
+    let segment = stem.rsplit_once("-s")?.1;
+    segment.parse::<usize>().ok()
+}
+
+fn attempt_from_filename(name: &str, flow: &str) -> Option<String> {
+    let stem = name.strip_suffix(".log")?;
+    let rest = stem.strip_prefix(&format!("{flow}-"))?;
+    let (attempt, _) = rest.rsplit_once("-s")?;
+    Some(attempt.to_string())
+}
+
+fn repair_group_from_filename(name: &str) -> Option<String> {
+    let stem = name.strip_suffix(".log")?;
+    let rest = stem.strip_prefix("repair-")?;
+    let (before_segment, _) = rest.rsplit_once("-s")?;
+
+    // New repair filenames include both child type and attempt id:
+    // repair-<repair_group_id>-<child_type>-<attempt_id>-sNN.log.
+    // Repair/attempt ids contain hyphens, so split on the deterministic
+    // child/attempt prefix pair emitted by SegmentOwner::Repair.
+    for child_type in [
+        "update",
+        "install",
+        "elevation",
+        "resume",
+        "resume_update",
+        "resume_install",
+    ] {
+        let marker = format!("-{child_type}-{child_type}-");
+        if let Some((group, _)) = before_segment.split_once(&marker) {
+            return Some(group.to_string());
+        }
+    }
+
+    // Backward-compatible fallback for the older repair-<group>-<child>-sNN form.
+    let (group, _) = before_segment.rsplit_once('-')?;
+    Some(group.to_string())
+}
+
+#[cfg(test)]
+mod tests {
+    use super::*;
+    use crate::diagnostics::state::{clone_snapshot, with_snapshot, AttemptSummary};
+    use crate::diagnostics::{new_diagnostics_state, now_ms};
+
+    #[test]
+    fn phase_writer_keeps_latest_segments_and_terminal_lines() {
+        let dir = std::env::temp_dir().join(format!(
+            "unsloth-phase-writer-{}-{}",
+            std::process::id(),
+            now_ms()
+        ));
+        let diagnostics = new_diagnostics_state();
+        let attempt_id = "install-test".to_string();
+        with_snapshot(&diagnostics, |snapshot| {
+            snapshot.install = Some(AttemptSummary::new(
+                "install",
+                attempt_id.clone(),
+                None,
+                None,
+                now_ms(),
+            ));
+        });
+        let mut writer = PhaseLogWriter::new_for_test(
+            diagnostics.clone(),
+            SegmentOwner::Install {
+                attempt_id: attempt_id.clone(),
+            },
+            "install",
+            now_ms(),
+            dir.clone(),
+            128,
+            2,
+        );
+        for i in 0..100 {
+            writer.append_line("stdout", &format!("chunk {i}"));
+        }
+        writer.append_line("stdout", "terminal failure survives");
+        let snapshot = clone_snapshot(&diagnostics);
+        let segments = snapshot.install.unwrap().log_segments;
+        assert_eq!(segments.len(), 2);
+        let total_size: u64 = segments
+            .iter()
+            .map(|path| fs::metadata(path).unwrap().len())
+            .sum();
+        assert!(total_size < 1024);
+        assert!(segments.iter().any(|path| {
+            fs::read_to_string(path)
+                .map(|text| text.contains("terminal failure survives"))
+                .unwrap_or(false)
+        }));
+        assert!(!dir.join("install-install-test-s01.log").exists());
+        let _ = fs::remove_dir_all(dir);
+    }
+
+    #[test]
+    fn repair_group_filename_fallback_handles_attempt_id_suffix() {
+        let filename = "repair-repair-123-4-install-install-567-8-s01.log";
+        assert_eq!(
+            repair_group_from_filename(filename),
+            Some("repair-123-4".to_string())
+        );
+    }
+
+    #[test]
+    fn repair_children_with_same_type_get_unique_segment_paths() {
+        let dir = std::env::temp_dir().join(format!(
+            "unsloth-repair-paths-{}-{}",
+            std::process::id(),
+            now_ms()
+        ));
+        let diagnostics = new_diagnostics_state();
+        let first = PhaseLogWriter::new_for_test(
+            diagnostics.clone(),
+            SegmentOwner::Repair {
+                repair_group_id: "repair-group".to_string(),
+                attempt_id: "attempt-one".to_string(),
+                child_type: "install".to_string(),
+            },
+            "repair",
+            now_ms(),
+            dir.clone(),
+            1024,
+            2,
+        );
+        let second = PhaseLogWriter::new_for_test(
+            diagnostics,
+            SegmentOwner::Repair {
+                repair_group_id: "repair-group".to_string(),
+                attempt_id: "attempt-two".to_string(),
+                child_type: "install".to_string(),
+            },
+            "repair",
+            now_ms(),
+            dir.clone(),
+            1024,
+            2,
+        );
+
+        assert_ne!(first.segment_path(1), second.segment_path(1));
+        let _ = fs::remove_dir_all(dir);
+    }
+
+    #[test]
+    fn parses_latest_disk_group_by_metadata_not_mtime_only() {
+        let dir = std::env::temp_dir().join(format!(
+            "unsloth-phase-parse-{}-{}",
+            std::process::id(),
+            now_ms()
+        ));
+        fs::create_dir_all(&dir).unwrap();
+        let old = dir.join("install-old-s01.log");
+        let new = dir.join("install-new-s01.log");
+        fs::write(
+            &old,
+            "diag_schema=1\nflow=install\nattempt_id=old\nsegment_index=1\nstarted_at_ms=1\n--- log ---\nold\n",
+        )
+        .unwrap();
+        fs::write(
+            &new,
+            "diag_schema=1\nflow=install\nattempt_id=new\nsegment_index=1\nstarted_at_ms=2\n--- log ---\nnew\n",
+        )
+        .unwrap();
+        let metas = vec![
+            parse_phase_log_meta(&old, "install-old-s01.log").unwrap(),
+            parse_phase_log_meta(&new, "install-new-s01.log").unwrap(),
+        ];
+        let group = latest_disk_group(&metas, "install").unwrap();
+        assert_eq!(group.label, "install:new");
+        let _ = fs::remove_dir_all(dir);
+    }
+}
diff --git a/studio/src-tauri/src/diagnostics/redaction.rs b/studio/src-tauri/src/diagnostics/redaction.rs
new file mode 100644
index 0000000000..5c5589a494
--- /dev/null
+++ b/studio/src-tauri/src/diagnostics/redaction.rs
@@ -0,0 +1,198 @@
+use regex::Regex;
+use std::path::Path;
+use std::sync::OnceLock;
+
+use super::studio_dir;
+
+#[derive(Default, Debug)]
+pub(crate) struct RedactionReport {
+    pub(crate) replacements: usize,
+}
+
+pub(crate) fn redact_text(text: &str, report: &mut RedactionReport) -> String {
+    let mut out = ansi_re().replace_all(text, "").to_string();
+    if out.len() != text.len() {
+        report.replacements += 1;
+    }
+    out = replace_regex(private_key_re(), &out, "<redacted private key>", report);
+    out = replace_regex(url_credentials_re(), &out, "$1<redacted>@", report);
+    out = replace_regex(auth_header_re(), &out, "$1: <redacted>", report);
+    out = replace_regex(cookie_re(), &out, "$1: <redacted>", report);
+    out = replace_regex(token_re(), &out, "<redacted token>", report);
+    out = replace_regex(env_secret_re(), &out, "$1=<redacted>", report);
+    out = replace_known_paths(&out, report);
+    out = replace_regex(windows_studio_re(), &out, "<studio_home>", report);
+    out = replace_regex(windows_home_re(), &out, "%USERPROFILE%", report);
+    out = replace_regex(unix_studio_re(), &out, "<studio_home>", report);
+    out = replace_regex(unix_home_re(), &out, "$HOME", report);
+    out = replace_regex(email_re(), &out, "<redacted email>", report);
+    out
+}
+
+fn replace_regex(
+    regex: &'static Regex,
+    input: &str,
+    replacement: &str,
+    report: &mut RedactionReport,
+) -> String {
+    let count = regex.find_iter(input).count();
+    if count > 0 {
+        report.replacements += count;
+        regex.replace_all(input, replacement).to_string()
+    } else {
+        input.to_string()
+    }
+}
+
+fn replace_known_paths(input: &str, report: &mut RedactionReport) -> String {
+    let mut out = input.to_string();
+    let studio = studio_dir();
+    out = replace_path_literal(&out, &studio, "<studio_home>", report);
+    if let Some(home) = dirs::home_dir() {
+        out = replace_path_literal(&out, &home, "$HOME", report);
+    }
+    out
+}
+
+fn replace_path_literal(
+    input: &str,
+    path: &Path,
+    replacement: &str,
+    report: &mut RedactionReport,
+) -> String {
+    let path_str = path.display().to_string();
+    let mut out = input.to_string();
+    for needle in path_variants(&path_str) {
+        if needle.is_empty() {
+            continue;
+        }
+        let count = out.matches(&needle).count();
+        if count > 0 {
+            report.replacements += count;
+            out = out.replace(&needle, replacement);
+        }
+    }
+    out
+}
+
+fn path_variants(path: &str) -> Vec<String> {
+    let mut variants = vec![path.to_string()];
+    variants.push(path.replace('/', "\\"));
+    variants.push(path.replace('\\', "/"));
+    variants.sort();
+    variants.dedup();
+    variants
+}
+
+fn ansi_re() -> &'static Regex {
+    static RE: OnceLock<Regex> = OnceLock::new();
+    RE.get_or_init(|| Regex::new(r"\x1b\[[0-9;?]*[ -/]*[@-~]").unwrap())
+}
+
+fn private_key_re() -> &'static Regex {
+    static RE: OnceLock<Regex> = OnceLock::new();
+    RE.get_or_init(|| {
+        Regex::new(
+            r"(?is)-----BEGIN [A-Z0-9 ]*PRIVATE KEY-----.*?-----END [A-Z0-9 ]*PRIVATE KEY-----",
+        )
+        .unwrap()
+    })
+}
+
+fn url_credentials_re() -> &'static Regex {
+    static RE: OnceLock<Regex> = OnceLock::new();
+    RE.get_or_init(|| Regex::new(r"(?i)\b([a-z][a-z0-9+.-]*://)[^/\s:@]+(:[^/\s@]*)?@").unwrap())
+}
+
+fn auth_header_re() -> &'static Regex {
+    static RE: OnceLock<Regex> = OnceLock::new();
+    RE.get_or_init(|| Regex::new(r"(?i)\b(authorization|proxy-authorization)\s*[:=]\s*(bearer|basic)?\s*[A-Za-z0-9._~+/=-]+").unwrap())
+}
+
+fn cookie_re() -> &'static Regex {
+    static RE: OnceLock<Regex> = OnceLock::new();
+    RE.get_or_init(|| Regex::new(r"(?i)\b(cookie|set-cookie)\s*[:=]\s*[^\r\n]+").unwrap())
+}
+
+fn token_re() -> &'static Regex {
+    static RE: OnceLock<Regex> = OnceLock::new();
+    RE.get_or_init(|| Regex::new(r"(?i)\b(hf_[A-Za-z0-9]{16,}|github_pat_[A-Za-z0-9_]{20,}|ghp_[A-Za-z0-9]{20,}|sk-[A-Za-z0-9]{20,})\b").unwrap())
+}
+
+fn env_secret_re() -> &'static Regex {
+    static RE: OnceLock<Regex> = OnceLock::new();
+    RE.get_or_init(|| {
+        Regex::new(r"(?i)\b((?:[A-Z0-9]+_)*(?:TOKEN|SECRET|PASSWORD|API_KEY|ACCESS_KEY|KEY)(?:_[A-Z0-9]+)*)\s*=\s*[^\s;&]+").unwrap()
+    })
+}
+
+fn windows_studio_re() -> &'static Regex {
+    static RE: OnceLock<Regex> = OnceLock::new();
+    RE.get_or_init(|| {
+        Regex::new(r"(?i)\b[A-Z]:[\\/]Users[\\/][^\\/\r\n\s]+[\\/]\.unsloth[\\/]studio").unwrap()
+    })
+}
+
+fn windows_home_re() -> &'static Regex {
+    static RE: OnceLock<Regex> = OnceLock::new();
+    RE.get_or_init(|| Regex::new(r"(?i)\b[A-Z]:[\\/]Users[\\/][^\\/\r\n\s]+").unwrap())
+}
+
+fn unix_studio_re() -> &'static Regex {
+    static RE: OnceLock<Regex> = OnceLock::new();
+    RE.get_or_init(|| Regex::new(r"(?i)(?:/Users|/home)/[A-Za-z0-9._-]+/\.unsloth/studio").unwrap())
+}
+
+fn unix_home_re() -> &'static Regex {
+    static RE: OnceLock<Regex> = OnceLock::new();
+    RE.get_or_init(|| Regex::new(r"(?i)(?:/Users|/home)/[A-Za-z0-9._-]+").unwrap())
+}
+
+fn email_re() -> &'static Regex {
+    static RE: OnceLock<Regex> = OnceLock::new();
+    RE.get_or_init(|| Regex::new(r"\b[A-Za-z0-9._%+-]+@[A-Za-z0-9.-]+\.[A-Za-z]{2,}\b").unwrap())
+}
+
+#[cfg(test)]
+mod tests {
+    use super::*;
+
+    #[test]
+    fn redacts_common_secret_and_path_patterns() {
+        let input = concat!(
+            "\u{1b}[31mred\u{1b}[0m\n",
+            "Authorization: Bearer abcdefghijklmnop\n",
+            "Cookie: session=abcdef\n",
+            "HF_TOKEN=hf_abcdefghijklmnopqrstuvwxyz\n",
+            "API_KEY=secret123\n",
+            "url=https://user:pass@example.com/path\n",
+            "email=alex@example.com\n",
+            "path=/Users/alex/.unsloth/studio/logs/install.log\n",
+            "win=C:\\Users\\Alex\\.unsloth\\studio\\logs\\install.log\n",
+            "-----BEGIN PRIVATE KEY-----\nabc\n-----END PRIVATE KEY-----\n"
+        );
+        let mut report = RedactionReport::default();
+        let redacted = redact_text(input, &mut report);
+        assert!(!redacted.contains("\u{1b}"));
+        assert!(!redacted.contains("abcdefghijklmnop"));
+        assert!(!redacted.contains("session=abcdef"));
+        assert!(!redacted.contains("hf_abcdefghijklmnopqrstuvwxyz"));
+        assert!(!redacted.contains("secret123"));
+        assert!(redacted.contains("https://<redacted>@example.com/path"));
+        assert!(!redacted.contains("alex@example.com"));
+        assert!(redacted.contains("<studio_home>"));
+        assert!(!redacted.contains("PRIVATE KEY-----\nabc"));
+    }
+
+    #[test]
+    fn redaction_avoids_keyboard_monkey_false_positives() {
+        let input = "keyboard=present monkey=banana MONKEY=banana KEYBOARD=present API_KEY=secret";
+        let mut report = RedactionReport::default();
+        let redacted = redact_text(input, &mut report);
+        assert!(redacted.contains("keyboard=present"));
+        assert!(redacted.contains("monkey=banana"));
+        assert!(redacted.contains("MONKEY=banana"));
+        assert!(redacted.contains("KEYBOARD=present"));
+        assert!(redacted.contains("API_KEY=<redacted>"));
+    }
+}
diff --git a/studio/src-tauri/src/diagnostics/report.rs b/studio/src-tauri/src/diagnostics/report.rs
new file mode 100644
index 0000000000..6301e7face
--- /dev/null
+++ b/studio/src-tauri/src/diagnostics/report.rs
@@ -0,0 +1,607 @@
+use std::fs::{self, File};
+use std::io::{Read, Seek, SeekFrom};
+use std::path::Path;
+use std::time::UNIX_EPOCH;
+
+use super::phase_log::{collect_phase_groups, path_has_symlink};
+use super::redaction::{redact_text, RedactionReport};
+use super::state::{AttemptSummary, CappedStrings, DiagnosticsSnapshot, FrontendSupportSnapshot};
+use super::{
+    cap_string, logs_dir, now_ms, studio_dir, valid_utf8_boundary, FOOTER_BUDGET_BYTES,
+    MAX_PHASE_LINE_BYTES, REPORT_MAX_BYTES, SCHEMA_VERSION, TAIL_MAX_BYTES, TAIL_MAX_LINES,
+};
+
+#[derive(Clone, Debug)]
+pub(crate) struct BackendHealthSection {
+    pub(crate) port: u16,
+    pub(crate) fields: Vec<(String, String)>,
+    pub(crate) warning: Option<String>,
+}
+
+pub(crate) fn render_report(
+    snapshot: DiagnosticsSnapshot,
+    frontend: FrontendSupportSnapshot,
+    health: Option<BackendHealthSection>,
+) -> String {
+    let mut warnings = Vec::new();
+    let mut raw = String::new();
+
+    raw.push_str("Unsloth Studio Support Diagnostics\n");
+    raw.push_str(&format!("diag_report_schema={SCHEMA_VERSION}\n"));
+    raw.push_str(&format!("created_at_ms={}\n", now_ms()));
+    raw.push_str(&format!("app_version={}\n", env!("CARGO_PKG_VERSION")));
+    raw.push_str(&format!("os={}\n", std::env::consts::OS));
+    raw.push_str(&format!("arch={}\n", std::env::consts::ARCH));
+    raw.push_str(&format!("debug_build={}\n", cfg!(debug_assertions)));
+    raw.push_str(&format!(
+        "appimage={}\n",
+        std::env::var_os("APPIMAGE").is_some()
+    ));
+    raw.push_str(&format!(
+        "cpu_logical_count={}\n",
+        std::thread::available_parallelism()
+            .map(|n| n.get().to_string())
+            .unwrap_or_else(|_| "unavailable".to_string())
+    ));
+    raw.push_str(&format!("studio_dir={}\n", studio_dir().display()));
+    raw.push_str(&format!("logs_dir={}\n", logs_dir().display()));
+    raw.push_str(&format!("app_session_id={}\n", snapshot.app_session_id));
+    raw.push_str(&format!(
+        "app_started_at_ms={}\n",
+        snapshot.app_started_at_ms
+    ));
+    raw.push_str(
+        "disk_budget=5MiB x 3 segments x 5 groups x 4 kinds ~= 300MiB plus tauri.log/.1\n",
+    );
+
+    append_frontend_section(&mut raw, &frontend);
+    append_preflight_section(&mut raw, &snapshot);
+    append_flow_section(&mut raw, &snapshot);
+    append_backend_section(&mut raw, &snapshot, health.as_ref());
+    append_device_signals_section(&mut raw, &snapshot, health.as_ref());
+    append_log_sections(&mut raw, &snapshot, &mut warnings);
+
+    if let Some(health) = &health {
+        if let Some(warning) = &health.warning {
+            warnings.push(format!("backend /api/health unavailable: {warning}"));
+        }
+    }
+
+    let mut redaction = RedactionReport::default();
+    let redacted_body = redact_text(&raw, &mut redaction);
+
+    // Redact collection warnings/footer too. Some warnings include raw local paths
+    // from failed log reads, so the footer must not be appended after redaction.
+    let footer_for_count = render_footer(&warnings, &redaction);
+    let _ = redact_text(&footer_for_count, &mut redaction);
+    let footer = redact_text(
+        &render_footer(&warnings, &redaction),
+        &mut RedactionReport::default(),
+    );
+
+    enforce_report_limit(redacted_body, footer)
+}
+
+fn append_frontend_section(out: &mut String, frontend: &FrontendSupportSnapshot) {
+    out.push_str("\n== Trigger/UI state ==\n");
+    append_opt(out, "status", frontend.status.as_deref());
+    append_opt(out, "error", frontend.error.as_deref());
+    append_opt(
+        out,
+        "current_step_index",
+        frontend
+            .current_step_index
+            .map(|v| v.to_string())
+            .as_deref(),
+    );
+    append_opt(out, "progress_detail", frontend.progress_detail.as_deref());
+    append_opt(out, "flow", frontend.flow.as_deref());
+    append_opt(out, "update_phase", frontend.update_phase.as_deref());
+    let update_progress = frontend.update_progress.as_ref().map(selected_json_value);
+    append_opt(out, "update_progress", update_progress.as_deref());
+    if let Some(packages) = &frontend.elevation_packages {
+        out.push_str(&format!("elevation_packages={}\n", packages.join(", ")));
+    }
+    out.push_str("last_ui_log_tail:\n");
+    match &frontend.last_ui_log_lines {
+        Some(lines) if !lines.is_empty() => {
+            for line in tail_lines(lines, TAIL_MAX_LINES) {
+                out.push_str("  ");
+                out.push_str(line);
+                out.push('\n');
+            }
+        }
+        _ => out.push_str("  unavailable\n"),
+    }
+}
+
+fn append_preflight_section(out: &mut String, snapshot: &DiagnosticsSnapshot) {
+    out.push_str("\n== Preflight/install state ==\n");
+    match &snapshot.latest_preflight {
+        Some(preflight) => {
+            out.push_str("source=live-state\n");
+            out.push_str(&format!("recorded_at_ms={}\n", preflight.recorded_at_ms));
+            out.push_str(&format!("disposition={}\n", preflight.disposition));
+            append_opt(out, "reason", preflight.reason.as_deref());
+            append_opt(
+                out,
+                "port",
+                preflight.port.map(|p| p.to_string()).as_deref(),
+            );
+            out.push_str(&format!("can_auto_repair={}\n", preflight.can_auto_repair));
+            if let Some(path) = &preflight.managed_bin {
+                out.push_str(&format!("managed_bin={}\n", path.display()));
+                out.push_str(&format!("managed_bin_exists={}\n", path.exists()));
+            } else {
+                out.push_str("managed_bin=unavailable\n");
+            }
+        }
+        None => out.push_str("source=live-state unavailable\n"),
+    }
+}
+
+fn append_flow_section(out: &mut String, snapshot: &DiagnosticsSnapshot) {
+    out.push_str("\n== Flow summaries ==\n");
+    append_attempt_summary(out, "install", snapshot.install.as_ref(), "live-state");
+    if !snapshot.install_history.is_empty() {
+        out.push_str("install_history:\n");
+        for install in snapshot.install_history.iter().rev().take(4).rev() {
+            append_attempt_summary(
+                out,
+                "  previous_install",
+                Some(install),
+                "live-state-history",
+            );
+        }
+    }
+    append_attempt_summary(out, "update", snapshot.update.as_ref(), "live-state");
+
+    out.push_str("repair_groups:\n");
+    if snapshot.repair_groups.is_empty() {
+        out.push_str("  unavailable\n");
+    } else {
+        for group in snapshot.repair_groups.iter().rev().take(5).rev() {
+            out.push_str(&format!(
+                "  repair_group_id={} started_at_ms={} ended_at_ms={} final_status={} last_error={}\n",
+                group.repair_group_id,
+                group.started_at_ms,
+                opt_u64(group.ended_at_ms),
+                group.final_status.as_deref().unwrap_or("unavailable"),
+                group.last_error.as_deref().unwrap_or("unavailable")
+            ));
+            for child in &group.children {
+                append_attempt_summary(out, "  child", Some(child), "live-state");
+            }
+        }
+    }
+}
+
+fn append_attempt_summary(
+    out: &mut String,
+    label: &str,
+    attempt: Option<&AttemptSummary>,
+    source: &str,
+) {
+    match attempt {
+        Some(attempt) => {
+            out.push_str(&format!(
+                "{label}: source={source} flow={} attempt_id={} group_id={} child_type={} started_at_ms={} ended_at_ms={} exit_status={} intentional_stop={} last_error={} log_segments={}\n",
+                attempt.flow,
+                attempt.attempt_id,
+                attempt.repair_group_id.as_deref().unwrap_or("none"),
+                attempt.child_type.as_deref().unwrap_or("none"),
+                attempt.started_at_ms,
+                opt_u64(attempt.ended_at_ms),
+                attempt.exit_status.as_deref().unwrap_or("unavailable"),
+                attempt.intentional_stop,
+                attempt.last_error.as_deref().unwrap_or("unavailable"),
+                attempt
+                    .log_segments
+                    .iter()
+                    .map(|p| p.display().to_string())
+                    .collect::<Vec<_>>()
+                    .join(",")
+            ));
+            if !attempt.elevation_packages.is_empty() {
+                out.push_str(&format!(
+                    "{label}.elevation_packages={}\n",
+                    attempt.elevation_packages.join(", ")
+                ));
+            }
+            append_capped(out, &format!("{label}.steps"), &attempt.steps);
+            append_capped(
+                out,
+                &format!("{label}.progress_details"),
+                &attempt.progress_details,
+            );
+            append_capped(out, &format!("{label}.diag_markers"), &attempt.diag_markers);
+        }
+        None => out.push_str(&format!("{label}: unavailable\n")),
+    }
+}
+
+fn append_backend_section(
+    out: &mut String,
+    snapshot: &DiagnosticsSnapshot,
+    health: Option<&BackendHealthSection>,
+) {
+    out.push_str("\n== Backend/auth/watchdog ==\n");
+    match &snapshot.backend {
+        Some(backend) => {
+            out.push_str("source=live-state\n");
+            out.push_str(&format!("backend_kind={}\n", backend.backend_kind));
+            append_opt(out, "session_id", backend.session_id.as_deref());
+            append_opt(
+                out,
+                "requested_port",
+                backend.requested_port.map(|p| p.to_string()).as_deref(),
+            );
+            append_opt(
+                out,
+                "reported_port",
+                backend.reported_port.map(|p| p.to_string()).as_deref(),
+            );
+            append_opt(
+                out,
+                "generation",
+                backend.generation.map(|g| g.to_string()).as_deref(),
+            );
+            append_opt(
+                out,
+                "started_at_ms",
+                backend.started_at_ms.map(|t| t.to_string()).as_deref(),
+            );
+            append_opt(
+                out,
+                "ended_at_ms",
+                backend.ended_at_ms.map(|t| t.to_string()).as_deref(),
+            );
+            append_opt(out, "exit_status", backend.exit_status.as_deref());
+            out.push_str(&format!("intentional_stop={}\n", backend.intentional_stop));
+            append_opt(out, "terminal_reason", backend.terminal_reason.as_deref());
+            append_opt(out, "last_error", backend.last_error.as_deref());
+            if backend.backend_kind == "attached_external" {
+                out.push_str("managed_process_logs=not_owned_by_tauri\n");
+            }
+        }
+        None => out.push_str("backend=unavailable backend_kind=unknown\n"),
+    }
+
+    match &snapshot.auth {
+        Some(auth) => {
+            out.push_str(&format!(
+                "auth_failure: failed_at_ms={} stage={} port={} message={}\n",
+                auth.failed_at_ms,
+                auth.stage,
+                auth.port
+                    .map(|p| p.to_string())
+                    .unwrap_or_else(|| "unavailable".to_string()),
+                auth.message
+            ));
+        }
+        None => out.push_str("auth_failure=unavailable\n"),
+    }
+
+    out.push_str("runtime_health:\n");
+    match health {
+        Some(health) => {
+            out.push_str(&format!("  port={}\n", health.port));
+            if health.fields.is_empty() {
+                out.push_str("  selected_fields=unavailable\n");
+            } else {
+                for (key, value) in &health.fields {
+                    out.push_str(&format!("  {key}={value}\n"));
+                }
+            }
+            if let Some(warning) = &health.warning {
+                out.push_str(&format!("  warning={warning}\n"));
+            }
+        }
+        None => out.push_str("  unavailable: no known backend port\n"),
+    }
+}
+
+fn append_device_signals_section(
+    out: &mut String,
+    snapshot: &DiagnosticsSnapshot,
+    health: Option<&BackendHealthSection>,
+) {
+    out.push_str("\n== Lightweight device signals ==\n");
+    out.push_str("scope=setup/runtime signals, not full hardware inventory\n");
+    out.push_str(&format!(
+        "os={} arch={}\n",
+        std::env::consts::OS,
+        std::env::consts::ARCH
+    ));
+    out.push_str("installer_diag_markers:\n");
+    let mut any = false;
+    if let Some(install) = &snapshot.install {
+        append_marker_tail(out, "install", install);
+        any |= !install.diag_markers.items.is_empty();
+    }
+    if let Some(update) = &snapshot.update {
+        append_marker_tail(out, "update", update);
+        any |= !update.diag_markers.items.is_empty();
+    }
+    for group in &snapshot.repair_groups {
+        for child in &group.children {
+            append_marker_tail(out, "repair_child", child);
+            any |= !child.diag_markers.items.is_empty();
+        }
+    }
+    if !any {
+        out.push_str("  unavailable\n");
+    }
+    if let Some(health) = health {
+        out.push_str("runtime_health_selected_fields:\n");
+        for (key, value) in &health.fields {
+            if matches!(
+                key.as_str(),
+                "version"
+                    | "device_type"
+                    | "chat_only"
+                    | "desktop_protocol_version"
+                    | "supports_api_only"
+                    | "supports_desktop_auth"
+            ) {
+                out.push_str(&format!("  {key}={value}\n"));
+            }
+        }
+    }
+}
+
+fn append_log_sections(
+    out: &mut String,
+    snapshot: &DiagnosticsSnapshot,
+    warnings: &mut Vec<String>,
+) {
+    out.push_str("\n== Log tails ==\n");
+    for name in ["tauri.log", "tauri.log.1"] {
+        let path = studio_dir().join(name);
+        append_tail_section(out, &path, name, "local-app-log", warnings);
+    }
+
+    let phase_groups = collect_phase_groups(snapshot, warnings);
+    if phase_groups.is_empty() {
+        out.push_str("phase_logs=unavailable\n");
+    }
+    for group in phase_groups {
+        out.push_str(&format!(
+            "\n-- phase_log_group label={} source={} --\n",
+            group.label, group.source
+        ));
+        if let Some(note) = &group.note {
+            out.push_str(&format!("note={}\n", note));
+        }
+        for path in group.paths {
+            append_tail_section(out, &path, &group.label, &group.source, warnings);
+        }
+    }
+}
+
+fn append_tail_section(
+    out: &mut String,
+    path: &Path,
+    label: &str,
+    source: &str,
+    warnings: &mut Vec<String>,
+) {
+    match read_tail(path, TAIL_MAX_LINES, TAIL_MAX_BYTES) {
+        Ok(tail) => {
+            out.push_str(&format!(
+                "file={} label={} source={} size_bytes={} mtime_ms={} included_bytes={} included_lines={} truncated={}\n",
+                path.display(),
+                label,
+                source,
+                tail.size_bytes,
+                opt_u64(tail.mtime_ms),
+                tail.included_bytes,
+                tail.included_lines,
+                tail.truncated
+            ));
+            out.push_str("```text\n");
+            out.push_str(&tail.text);
+            if !tail.text.ends_with('\n') {
+                out.push('\n');
+            }
+            out.push_str("```\n");
+        }
+        Err(error) => {
+            out.push_str(&format!(
+                "file={} label={} source={} unavailable={}\n",
+                path.display(),
+                label,
+                source,
+                error
+            ));
+            warnings.push(format!("{} unavailable: {}", path.display(), error));
+        }
+    }
+}
+
+#[derive(Debug)]
+struct TailRead {
+    text: String,
+    size_bytes: u64,
+    mtime_ms: Option<u64>,
+    included_bytes: usize,
+    included_lines: usize,
+    truncated: bool,
+}
+
+fn read_tail(path: &Path, max_lines: usize, max_bytes: usize) -> Result<TailRead, String> {
+    if path_has_symlink(path) {
+        return Err("refusing to read symlink".to_string());
+    }
+    let metadata = fs::metadata(path).map_err(|e| e.to_string())?;
+    let size = metadata.len();
+    let mut file = File::open(path).map_err(|e| e.to_string())?;
+    let start = size.saturating_sub(max_bytes as u64);
+    file.seek(SeekFrom::Start(start))
+        .map_err(|e| e.to_string())?;
+    let mut bytes = Vec::new();
+    file.take(max_bytes as u64)
+        .read_to_end(&mut bytes)
+        .map_err(|e| e.to_string())?;
+    let mut text = String::from_utf8_lossy(&bytes).into_owned();
+    let mut truncated = start > 0;
+    if truncated {
+        text = format!("[tail truncated to last {max_bytes} bytes]\n{text}");
+    }
+    let mut lines: Vec<String> = text
+        .lines()
+        .map(|line| {
+            if line.len() > MAX_PHASE_LINE_BYTES {
+                let boundary = valid_utf8_boundary(line, MAX_PHASE_LINE_BYTES);
+                format!("{} [line truncated]", &line[..boundary])
+            } else {
+                line.to_string()
+            }
+        })
+        .collect();
+    if lines.len() > max_lines {
+        truncated = true;
+        let omitted = lines.len() - max_lines;
+        lines = lines.split_off(omitted);
+        lines.insert(0, format!("[tail truncated to last {max_lines} lines]"));
+    }
+    let included_lines = lines.len();
+    let mut text = lines.join("\n");
+    if !text.is_empty() {
+        text.push('\n');
+    }
+    Ok(TailRead {
+        text,
+        size_bytes: size,
+        mtime_ms: metadata
+            .modified()
+            .ok()
+            .and_then(|t| t.duration_since(UNIX_EPOCH).ok())
+            .map(|d| d.as_millis() as u64),
+        included_bytes: bytes.len(),
+        included_lines,
+        truncated,
+    })
+}
+
+fn render_footer(warnings: &[String], redaction: &RedactionReport) -> String {
+    let mut footer = String::new();
+    footer.push_str("\n== Collection warnings ==\n");
+    if warnings.is_empty() {
+        footer.push_str("none\n");
+    } else {
+        for warning in warnings.iter().take(100) {
+            footer.push_str("- ");
+            footer.push_str(warning);
+            footer.push('\n');
+        }
+        if warnings.len() > 100 {
+            footer.push_str(&format!(
+                "- omitted {} additional warnings\n",
+                warnings.len() - 100
+            ));
+        }
+    }
+    footer.push_str("\n== Redaction/omission summary ==\n");
+    footer.push_str(&format!(
+        "redaction_replacements={}\n",
+        redaction.replacements
+    ));
+    footer.push_str("redaction_scope=ANSI, private keys, URL credentials, auth headers, cookies, token patterns, assignment-style secrets, studio/home paths, emails\n");
+    footer.push_str(
+        "report_bounds=log sections <=1000 lines/200KiB each; total clipboard text <=1MiB\n",
+    );
+    footer.push_str("known_v1_gap=elevated apt helper may buffer subprocess output before diagnostics caps it\n");
+    footer.push_str("known_v1_gap=normal install elevation resume is linked in same-run state/report history; disk fallback conservatively includes recent install attempts but has no explicit install_group_id in V1\n");
+    footer
+}
+
+fn enforce_report_limit(body: String, footer: String) -> String {
+    if body.len() + footer.len() <= REPORT_MAX_BYTES {
+        return format!("{body}{footer}");
+    }
+    let notice = "\n[report truncated to fit clipboard budget; footer preserved]\n";
+    let footer_budget = footer.len().min(FOOTER_BUDGET_BYTES).max(footer.len());
+    let budget = REPORT_MAX_BYTES
+        .saturating_sub(footer_budget)
+        .saturating_sub(notice.len());
+    let boundary = valid_utf8_boundary(&body, budget);
+    format!("{}{}{}", &body[..boundary], notice, footer)
+}
+fn append_opt(out: &mut String, key: &str, value: Option<&str>) {
+    out.push_str(key);
+    out.push('=');
+    out.push_str(value.unwrap_or("unavailable"));
+    out.push('\n');
+}
+
+fn append_capped(out: &mut String, label: &str, values: &CappedStrings) {
+    out.push_str(&format!("{label}: omitted={}\n", values.omitted));
+    if values.items.is_empty() {
+        out.push_str("  unavailable\n");
+    } else {
+        for value in values.items.iter().rev().take(20).rev() {
+            out.push_str("  ");
+            out.push_str(value);
+            out.push('\n');
+        }
+    }
+}
+
+fn append_marker_tail(out: &mut String, label: &str, attempt: &AttemptSummary) {
+    for marker in attempt.diag_markers.items.iter().rev().take(10).rev() {
+        out.push_str(&format!("  {label}: {marker}\n"));
+    }
+}
+
+fn tail_lines(lines: &[String], max_lines: usize) -> &[String] {
+    if lines.len() > max_lines {
+        &lines[lines.len() - max_lines..]
+    } else {
+        lines
+    }
+}
+
+fn opt_u64(value: Option<u64>) -> String {
+    value
+        .map(|v| v.to_string())
+        .unwrap_or_else(|| "unavailable".to_string())
+}
+
+pub(crate) fn selected_json_value(value: &serde_json::Value) -> String {
+    match value {
+        serde_json::Value::String(s) => cap_string(s, 256),
+        serde_json::Value::Bool(b) => b.to_string(),
+        serde_json::Value::Number(n) => n.to_string(),
+        serde_json::Value::Null => "null".to_string(),
+        _ => "<non-scalar>".to_string(),
+    }
+}
+
+#[cfg(test)]
+mod tests {
+    use super::*;
+
+    #[test]
+    fn report_redacts_footer_collection_warning_paths() {
+        let snapshot = DiagnosticsSnapshot::default();
+        let report = render_report(snapshot, FrontendSupportSnapshot::default(), None);
+        let studio = studio_dir().display().to_string();
+        assert!(!report.contains(&studio));
+        assert!(report.contains("<studio_home>"));
+    }
+
+    #[test]
+    fn tail_handles_missing_and_non_utf8() {
+        let missing =
+            std::env::temp_dir().join(format!("unsloth-missing-tail-{}", std::process::id()));
+        assert!(read_tail(&missing, 10, 100).is_err());
+
+        let path =
+            std::env::temp_dir().join(format!("unsloth-nonutf8-tail-{}", std::process::id()));
+        fs::write(&path, [0xff, b'a', b'\n', b'b']).unwrap();
+        let tail = read_tail(&path, 10, 100).unwrap();
+        assert!(tail.text.contains('�'));
+        let _ = fs::remove_file(path);
+    }
+}
diff --git a/studio/src-tauri/src/diagnostics/state.rs b/studio/src-tauri/src/diagnostics/state.rs
new file mode 100644
index 0000000000..ccbd5fbb8e
--- /dev/null
+++ b/studio/src-tauri/src/diagnostics/state.rs
@@ -0,0 +1,774 @@
+use crate::preflight::{DesktopPreflightDisposition, DesktopPreflightResult};
+use serde::{Deserialize, Serialize};
+use std::path::PathBuf;
+use std::sync::{Arc, Mutex};
+
+use super::phase_log::{PhaseLogHandle, PhaseLogWriter, SegmentOwner};
+use super::{cap_string, new_id, now_ms, MAX_STATE_ITEMS};
+
+pub type DiagnosticsState = Arc<Mutex<DiagnosticsSnapshot>>;
+
+#[derive(Clone, Debug)]
+pub struct DiagnosticsSnapshot {
+    pub app_session_id: String,
+    pub app_started_at_ms: u64,
+    pub latest_preflight: Option<PreflightSummary>,
+    pub install: Option<AttemptSummary>,
+    pub install_history: Vec<AttemptSummary>,
+    pub repair_groups: Vec<RepairGroupSummary>,
+    pub update: Option<AttemptSummary>,
+    pub backend: Option<BackendSummary>,
+    pub auth: Option<AuthSummary>,
+}
+
+impl Default for DiagnosticsSnapshot {
+    fn default() -> Self {
+        Self {
+            app_session_id: new_id("app"),
+            app_started_at_ms: now_ms(),
+            latest_preflight: None,
+            install: None,
+            install_history: Vec::new(),
+            repair_groups: Vec::new(),
+            update: None,
+            backend: None,
+            auth: None,
+        }
+    }
+}
+
+#[derive(Clone, Debug, Default)]
+pub struct CappedStrings {
+    pub items: Vec<String>,
+    pub omitted: usize,
+}
+
+impl CappedStrings {
+    fn push(&mut self, value: impl Into<String>) {
+        if self.items.len() >= MAX_STATE_ITEMS {
+            self.items.remove(0);
+            self.omitted += 1;
+        }
+        self.items.push(value.into());
+    }
+}
+
+#[derive(Clone, Debug)]
+pub struct PreflightSummary {
+    pub disposition: String,
+    pub reason: Option<String>,
+    pub port: Option<u16>,
+    pub can_auto_repair: bool,
+    pub managed_bin: Option<PathBuf>,
+    pub recorded_at_ms: u64,
+}
+
+#[derive(Clone, Debug)]
+pub struct AttemptSummary {
+    pub flow: String,
+    pub attempt_id: String,
+    pub repair_group_id: Option<String>,
+    pub child_type: Option<String>,
+    pub log_segments: Vec<PathBuf>,
+    pub started_at_ms: u64,
+    pub ended_at_ms: Option<u64>,
+    pub exit_status: Option<String>,
+    pub intentional_stop: bool,
+    pub last_error: Option<String>,
+    pub elevation_packages: Vec<String>,
+    pub steps: CappedStrings,
+    pub progress_details: CappedStrings,
+    pub diag_markers: CappedStrings,
+}
+
+impl AttemptSummary {
+    pub(crate) fn new(
+        flow: impl Into<String>,
+        attempt_id: String,
+        repair_group_id: Option<String>,
+        child_type: Option<String>,
+        started_at_ms: u64,
+    ) -> Self {
+        Self {
+            flow: flow.into(),
+            attempt_id,
+            repair_group_id,
+            child_type,
+            log_segments: Vec::new(),
+            started_at_ms,
+            ended_at_ms: None,
+            exit_status: None,
+            intentional_stop: false,
+            last_error: None,
+            elevation_packages: Vec::new(),
+            steps: CappedStrings::default(),
+            progress_details: CappedStrings::default(),
+            diag_markers: CappedStrings::default(),
+        }
+    }
+}
+
+#[derive(Clone, Debug)]
+pub struct RepairGroupSummary {
+    pub repair_group_id: String,
+    pub children: Vec<AttemptSummary>,
+    pub started_at_ms: u64,
+    pub ended_at_ms: Option<u64>,
+    pub final_status: Option<String>,
+    pub last_error: Option<String>,
+}
+
+#[derive(Clone, Debug)]
+pub struct BackendSummary {
+    pub session_id: Option<String>,
+    pub backend_kind: String,
+    pub log_segments: Vec<PathBuf>,
+    pub requested_port: Option<u16>,
+    pub reported_port: Option<u16>,
+    pub generation: Option<u64>,
+    pub started_at_ms: Option<u64>,
+    pub ended_at_ms: Option<u64>,
+    pub exit_status: Option<String>,
+    pub intentional_stop: bool,
+    pub terminal_reason: Option<String>,
+    pub last_error: Option<String>,
+}
+
+#[derive(Clone, Debug)]
+pub struct AuthSummary {
+    pub failed_at_ms: u64,
+    pub stage: String,
+    pub port: Option<u16>,
+    pub message: String,
+}
+
+#[derive(Clone, Debug, Default, Serialize, Deserialize)]
+#[serde(rename_all = "camelCase")]
+pub struct FrontendSupportSnapshot {
+    pub status: Option<String>,
+    pub error: Option<String>,
+    pub current_step_index: Option<i64>,
+    pub progress_detail: Option<String>,
+    pub elevation_packages: Option<Vec<String>>,
+    pub last_ui_log_lines: Option<Vec<String>>,
+    pub flow: Option<String>,
+    pub update_phase: Option<String>,
+    /// Frontend update progress can be numeric today; keep this tolerant of strings/numbers.
+    pub update_progress: Option<serde_json::Value>,
+}
+
+#[derive(Clone)]
+pub struct AttemptLog {
+    pub flow: String,
+    pub attempt_id: String,
+    pub repair_group_id: Option<String>,
+    #[allow(dead_code)]
+    pub child_type: Option<String>,
+    pub handle: PhaseLogHandle,
+}
+
+#[derive(Clone)]
+pub struct BackendLog {
+    pub session_id: String,
+    pub handle: PhaseLogHandle,
+}
+
+pub fn new_diagnostics_state() -> DiagnosticsState {
+    Arc::new(Mutex::new(DiagnosticsSnapshot::default()))
+}
+
+pub fn begin_install_attempt(diagnostics: &DiagnosticsState) -> AttemptLog {
+    begin_attempt(diagnostics, "install", None, None)
+}
+
+pub fn begin_update_attempt(diagnostics: &DiagnosticsState) -> AttemptLog {
+    begin_attempt(diagnostics, "update", None, None)
+}
+
+pub fn begin_repair_group(diagnostics: &DiagnosticsState) -> String {
+    let repair_group_id = new_id("repair");
+    with_snapshot(diagnostics, |snapshot| {
+        if snapshot.repair_groups.len() >= MAX_STATE_ITEMS {
+            snapshot.repair_groups.remove(0);
+        }
+        snapshot.repair_groups.push(RepairGroupSummary {
+            repair_group_id: repair_group_id.clone(),
+            children: Vec::new(),
+            started_at_ms: now_ms(),
+            ended_at_ms: None,
+            final_status: None,
+            last_error: None,
+        });
+    });
+    repair_group_id
+}
+
+pub fn begin_repair_child(
+    diagnostics: &DiagnosticsState,
+    repair_group_id: &str,
+    child_type: &str,
+) -> AttemptLog {
+    begin_attempt(
+        diagnostics,
+        "repair",
+        Some(repair_group_id.to_string()),
+        Some(child_type.to_string()),
+    )
+}
+
+fn begin_attempt(
+    diagnostics: &DiagnosticsState,
+    flow: &str,
+    repair_group_id: Option<String>,
+    child_type: Option<String>,
+) -> AttemptLog {
+    let started_at_ms = now_ms();
+    let attempt_id = new_id(child_type.as_deref().unwrap_or(flow));
+    let owner = match (flow, repair_group_id.as_ref(), child_type.as_ref()) {
+        ("install", None, _) => SegmentOwner::Install {
+            attempt_id: attempt_id.clone(),
+        },
+        ("update", None, _) => SegmentOwner::Update {
+            attempt_id: attempt_id.clone(),
+        },
+        (_, Some(group), Some(child)) => SegmentOwner::Repair {
+            repair_group_id: group.clone(),
+            attempt_id: attempt_id.clone(),
+            child_type: child.clone(),
+        },
+        _ => SegmentOwner::Install {
+            attempt_id: attempt_id.clone(),
+        },
+    };
+
+    let summary = AttemptSummary::new(
+        flow,
+        attempt_id.clone(),
+        repair_group_id.clone(),
+        child_type.clone(),
+        started_at_ms,
+    );
+
+    with_snapshot(diagnostics, |snapshot| {
+        if let Some(group_id) = &repair_group_id {
+            let group = ensure_repair_group(snapshot, group_id);
+            if group.children.len() >= MAX_STATE_ITEMS {
+                group.children.remove(0);
+            }
+            group.children.push(summary);
+        } else if flow == "update" {
+            snapshot.update = Some(summary);
+        } else {
+            if let Some(previous) = snapshot.install.take() {
+                if snapshot.install_history.len() >= MAX_STATE_ITEMS {
+                    snapshot.install_history.remove(0);
+                }
+                snapshot.install_history.push(previous);
+            }
+            snapshot.install = Some(summary);
+        }
+    });
+
+    let writer = PhaseLogWriter::new(
+        diagnostics.clone(),
+        owner,
+        if repair_group_id.is_some() {
+            "repair"
+        } else {
+            flow
+        },
+        started_at_ms,
+        None,
+    );
+
+    AttemptLog {
+        flow: flow.to_string(),
+        attempt_id,
+        repair_group_id,
+        child_type,
+        handle: Arc::new(Mutex::new(writer)),
+    }
+}
+
+pub fn finish_attempt(
+    diagnostics: &DiagnosticsState,
+    attempt: &AttemptLog,
+    exit_status: Option<String>,
+    intentional_stop: bool,
+    last_error: Option<String>,
+) {
+    with_snapshot(diagnostics, |snapshot| {
+        update_attempt_mut(snapshot, attempt, |summary| {
+            if summary.ended_at_ms.is_none() {
+                summary.ended_at_ms = Some(now_ms());
+            }
+            if exit_status.is_some() {
+                summary.exit_status = exit_status.clone();
+            }
+            summary.intentional_stop |= intentional_stop;
+            if last_error.is_some() {
+                summary.last_error = last_error.clone();
+            }
+        });
+    });
+}
+
+pub fn finish_repair_group(
+    diagnostics: &DiagnosticsState,
+    repair_group_id: &str,
+    final_status: impl Into<String>,
+    last_error: Option<String>,
+) {
+    let status = final_status.into();
+    with_snapshot(diagnostics, |snapshot| {
+        let group = ensure_repair_group(snapshot, repair_group_id);
+        group.ended_at_ms = Some(now_ms());
+        group.final_status = Some(status.clone());
+        group.last_error = last_error.clone();
+    });
+}
+
+pub fn record_step(diagnostics: &DiagnosticsState, attempt: &AttemptLog, step: &str) {
+    record_attempt_string(diagnostics, attempt, step, |summary, value| {
+        summary.steps.push(value)
+    });
+}
+
+pub fn record_progress(diagnostics: &DiagnosticsState, attempt: &AttemptLog, progress: &str) {
+    record_attempt_string(diagnostics, attempt, progress, |summary, value| {
+        summary.progress_details.push(value)
+    });
+}
+
+pub fn record_diag_marker(diagnostics: &DiagnosticsState, attempt: &AttemptLog, marker: &str) {
+    record_attempt_string(diagnostics, attempt, marker, |summary, value| {
+        summary.diag_markers.push(value)
+    });
+}
+
+pub fn record_elevation_packages(
+    diagnostics: &DiagnosticsState,
+    attempt: &AttemptLog,
+    packages: &[String],
+) {
+    let capped: Vec<String> = packages.iter().take(100).cloned().collect();
+    with_snapshot(diagnostics, |snapshot| {
+        update_attempt_mut(snapshot, attempt, |summary| {
+            summary.elevation_packages = capped.clone();
+        });
+    });
+}
+
+fn record_attempt_string<F>(
+    diagnostics: &DiagnosticsState,
+    attempt: &AttemptLog,
+    value: &str,
+    mut update: F,
+) where
+    F: FnMut(&mut AttemptSummary, String),
+{
+    let capped = cap_string(value, 4096);
+    with_snapshot(diagnostics, |snapshot| {
+        update_attempt_mut(snapshot, attempt, |summary| {
+            update(summary, capped.clone());
+        });
+    });
+}
+
+pub fn record_preflight(diagnostics: &DiagnosticsState, result: &DesktopPreflightResult) {
+    let summary = PreflightSummary {
+        disposition: disposition_label(&result.disposition).to_string(),
+        reason: result.reason.clone(),
+        port: result.port,
+        can_auto_repair: result.can_auto_repair,
+        managed_bin: result.managed_bin.clone(),
+        recorded_at_ms: now_ms(),
+    };
+    with_snapshot(diagnostics, |snapshot| {
+        snapshot.latest_preflight = Some(summary);
+    });
+    if matches!(
+        result.disposition,
+        DesktopPreflightDisposition::AttachedReady
+    ) {
+        if let Some(port) = result.port {
+            record_attached_external_backend(diagnostics, port);
+        }
+    }
+}
+
+pub fn begin_backend_session(
+    diagnostics: &DiagnosticsState,
+    requested_port: u16,
+    generation: u64,
+) -> BackendLog {
+    let session_id = new_id("backend");
+    let started_at_ms = now_ms();
+    with_snapshot(diagnostics, |snapshot| {
+        snapshot.backend = Some(BackendSummary {
+            session_id: Some(session_id.clone()),
+            backend_kind: "managed".to_string(),
+            log_segments: Vec::new(),
+            requested_port: Some(requested_port),
+            reported_port: None,
+            generation: Some(generation),
+            started_at_ms: Some(started_at_ms),
+            ended_at_ms: None,
+            exit_status: None,
+            intentional_stop: false,
+            terminal_reason: None,
+            last_error: None,
+        });
+    });
+    let writer = PhaseLogWriter::new(
+        diagnostics.clone(),
+        SegmentOwner::Backend {
+            session_id: session_id.clone(),
+        },
+        "backend",
+        started_at_ms,
+        Some("managed".to_string()),
+    );
+    BackendLog {
+        session_id,
+        handle: Arc::new(Mutex::new(writer)),
+    }
+}
+
+pub fn record_backend_start_failure(
+    diagnostics: &DiagnosticsState,
+    requested_port: Option<u16>,
+    generation: Option<u64>,
+    stage: &str,
+    error: &str,
+) {
+    let error = cap_string(error, 4096);
+    with_snapshot(diagnostics, |snapshot| {
+        snapshot.backend = Some(BackendSummary {
+            session_id: None,
+            backend_kind: "managed".to_string(),
+            log_segments: Vec::new(),
+            requested_port,
+            reported_port: None,
+            generation,
+            started_at_ms: Some(now_ms()),
+            ended_at_ms: Some(now_ms()),
+            exit_status: None,
+            intentional_stop: false,
+            terminal_reason: Some(stage.to_string()),
+            last_error: Some(error.clone()),
+        });
+    });
+}
+
+pub fn record_backend_port(diagnostics: &DiagnosticsState, session_id: &str, port: u16) {
+    with_snapshot(diagnostics, |snapshot| {
+        if let Some(backend) = snapshot.backend.as_mut() {
+            if backend.session_id.as_deref() == Some(session_id) {
+                backend.reported_port = Some(port);
+            }
+        }
+    });
+}
+
+pub fn record_backend_exit(
+    diagnostics: &DiagnosticsState,
+    session_id: &str,
+    exit_status: Option<String>,
+    intentional_stop: bool,
+    terminal_reason: Option<String>,
+) {
+    with_snapshot(diagnostics, |snapshot| {
+        if let Some(backend) = snapshot.backend.as_mut() {
+            if backend.session_id.as_deref() != Some(session_id) {
+                return;
+            }
+            if backend.ended_at_ms.is_none() {
+                backend.ended_at_ms = Some(now_ms());
+            }
+            if exit_status.is_some() {
+                backend.exit_status = exit_status.clone();
+            }
+            backend.intentional_stop |= intentional_stop;
+            if backend.terminal_reason.is_none() {
+                backend.terminal_reason = terminal_reason.clone().or_else(|| {
+                    Some(if intentional_stop {
+                        "intentional_stop".to_string()
+                    } else {
+                        "process_exit".to_string()
+                    })
+                });
+            }
+        }
+    });
+}
+
+pub fn record_backend_watchdog(
+    diagnostics: &DiagnosticsState,
+    generation: u64,
+    terminal_reason: &str,
+) {
+    with_snapshot(diagnostics, |snapshot| {
+        if let Some(backend) = snapshot.backend.as_mut() {
+            if backend.generation == Some(generation) && backend.terminal_reason.is_none() {
+                backend.terminal_reason = Some(terminal_reason.to_string());
+                backend.ended_at_ms = Some(now_ms());
+            }
+        }
+    });
+}
+
+pub fn record_backend_intentional_stop(diagnostics: &DiagnosticsState) {
+    with_snapshot(diagnostics, |snapshot| {
+        if let Some(backend) = snapshot.backend.as_mut() {
+            if matches!(
+                backend.terminal_reason.as_deref(),
+                Some("no_port_after_grace" | "unresponsive_health_check")
+            ) {
+                return;
+            }
+            backend.intentional_stop = true;
+            if backend.terminal_reason.is_none() {
+                backend.terminal_reason = Some("intentional_stop".to_string());
+            }
+        }
+    });
+}
+
+pub fn record_attached_external_backend(diagnostics: &DiagnosticsState, port: u16) {
+    with_snapshot(diagnostics, |snapshot| {
+        let should_replace = snapshot
+            .backend
+            .as_ref()
+            .map(|backend| backend.backend_kind != "managed" || backend.session_id.is_none())
+            .unwrap_or(true);
+        if should_replace {
+            snapshot.backend = Some(BackendSummary {
+                session_id: None,
+                backend_kind: "attached_external".to_string(),
+                log_segments: Vec::new(),
+                requested_port: None,
+                reported_port: Some(port),
+                generation: None,
+                started_at_ms: None,
+                ended_at_ms: None,
+                exit_status: None,
+                intentional_stop: false,
+                terminal_reason: None,
+                last_error: None,
+            });
+        }
+    });
+}
+
+pub fn record_auth_failure(
+    diagnostics: &DiagnosticsState,
+    stage: &str,
+    port: Option<u16>,
+    message: &str,
+) {
+    let message = cap_string(message, 4096);
+    with_snapshot(diagnostics, |snapshot| {
+        snapshot.auth = Some(AuthSummary {
+            failed_at_ms: now_ms(),
+            stage: stage.to_string(),
+            port,
+            message,
+        });
+    });
+}
+
+pub(crate) fn with_snapshot<F>(diagnostics: &DiagnosticsState, update: F)
+where
+    F: FnOnce(&mut DiagnosticsSnapshot),
+{
+    match diagnostics.lock() {
+        Ok(mut snapshot) => update(&mut snapshot),
+        Err(poisoned) => {
+            let mut snapshot = poisoned.into_inner();
+            update(&mut snapshot);
+        }
+    }
+}
+
+pub(crate) fn clone_snapshot(diagnostics: &DiagnosticsState) -> DiagnosticsSnapshot {
+    match diagnostics.lock() {
+        Ok(snapshot) => snapshot.clone(),
+        Err(poisoned) => poisoned.into_inner().clone(),
+    }
+}
+
+fn ensure_repair_group<'a>(
+    snapshot: &'a mut DiagnosticsSnapshot,
+    repair_group_id: &str,
+) -> &'a mut RepairGroupSummary {
+    if let Some(index) = snapshot
+        .repair_groups
+        .iter()
+        .position(|group| group.repair_group_id == repair_group_id)
+    {
+        return &mut snapshot.repair_groups[index];
+    }
+    snapshot.repair_groups.push(RepairGroupSummary {
+        repair_group_id: repair_group_id.to_string(),
+        children: Vec::new(),
+        started_at_ms: now_ms(),
+        ended_at_ms: None,
+        final_status: None,
+        last_error: None,
+    });
+    snapshot.repair_groups.last_mut().expect("just pushed")
+}
+
+fn update_attempt_mut<F>(snapshot: &mut DiagnosticsSnapshot, attempt: &AttemptLog, update: F)
+where
+    F: FnOnce(&mut AttemptSummary),
+{
+    if let Some(group_id) = &attempt.repair_group_id {
+        if let Some(group) = snapshot
+            .repair_groups
+            .iter_mut()
+            .find(|group| &group.repair_group_id == group_id)
+        {
+            if let Some(child) = group
+                .children
+                .iter_mut()
+                .rev()
+                .find(|child| child.attempt_id == attempt.attempt_id)
+            {
+                update(child);
+            }
+        }
+    } else if attempt.flow == "update" {
+        if let Some(summary) = snapshot.update.as_mut() {
+            if summary.attempt_id == attempt.attempt_id {
+                update(summary);
+            }
+        }
+    } else if let Some(summary) = snapshot.install.as_mut() {
+        if summary.attempt_id == attempt.attempt_id {
+            update(summary);
+        }
+    }
+}
+
+pub(crate) fn record_segment_path(
+    diagnostics: &DiagnosticsState,
+    owner: &SegmentOwner,
+    path: PathBuf,
+) {
+    with_snapshot(diagnostics, |snapshot| match owner {
+        SegmentOwner::Install { attempt_id } => {
+            if let Some(summary) = snapshot.install.as_mut() {
+                if &summary.attempt_id == attempt_id && !summary.log_segments.contains(&path) {
+                    summary.log_segments.push(path.clone());
+                }
+            }
+            for summary in &mut snapshot.install_history {
+                if &summary.attempt_id == attempt_id && !summary.log_segments.contains(&path) {
+                    summary.log_segments.push(path.clone());
+                }
+            }
+        }
+        SegmentOwner::Update { attempt_id } => {
+            if let Some(summary) = snapshot.update.as_mut() {
+                if &summary.attempt_id == attempt_id && !summary.log_segments.contains(&path) {
+                    summary.log_segments.push(path.clone());
+                }
+            }
+        }
+        SegmentOwner::Repair {
+            repair_group_id,
+            attempt_id,
+            ..
+        } => {
+            if let Some(group) = snapshot
+                .repair_groups
+                .iter_mut()
+                .find(|group| &group.repair_group_id == repair_group_id)
+            {
+                if let Some(child) = group
+                    .children
+                    .iter_mut()
+                    .find(|child| &child.attempt_id == attempt_id)
+                {
+                    if !child.log_segments.contains(&path) {
+                        child.log_segments.push(path.clone());
+                    }
+                }
+            }
+        }
+        SegmentOwner::Backend { session_id } => {
+            if let Some(backend) = snapshot.backend.as_mut() {
+                if backend.session_id.as_deref() == Some(session_id)
+                    && !backend.log_segments.contains(&path)
+                {
+                    backend.log_segments.push(path.clone());
+                }
+            }
+        }
+    });
+}
+
+pub(crate) fn forget_segment_path(
+    diagnostics: &DiagnosticsState,
+    owner: &SegmentOwner,
+    path: &PathBuf,
+) {
+    with_snapshot(diagnostics, |snapshot| match owner {
+        SegmentOwner::Install { attempt_id } => {
+            if let Some(summary) = snapshot.install.as_mut() {
+                if &summary.attempt_id == attempt_id {
+                    summary.log_segments.retain(|segment| segment != path);
+                }
+            }
+            for summary in &mut snapshot.install_history {
+                if &summary.attempt_id == attempt_id {
+                    summary.log_segments.retain(|segment| segment != path);
+                }
+            }
+        }
+        SegmentOwner::Update { attempt_id } => {
+            if let Some(summary) = snapshot.update.as_mut() {
+                if &summary.attempt_id == attempt_id {
+                    summary.log_segments.retain(|segment| segment != path);
+                }
+            }
+        }
+        SegmentOwner::Repair {
+            repair_group_id,
+            attempt_id,
+            ..
+        } => {
+            if let Some(group) = snapshot
+                .repair_groups
+                .iter_mut()
+                .find(|group| &group.repair_group_id == repair_group_id)
+            {
+                if let Some(child) = group
+                    .children
+                    .iter_mut()
+                    .find(|child| &child.attempt_id == attempt_id)
+                {
+                    child.log_segments.retain(|segment| segment != path);
+                }
+            }
+        }
+        SegmentOwner::Backend { session_id } => {
+            if let Some(backend) = snapshot.backend.as_mut() {
+                if backend.session_id.as_deref() == Some(session_id) {
+                    backend.log_segments.retain(|segment| segment != path);
+                }
+            }
+        }
+    });
+}
+
+fn disposition_label(disposition: &DesktopPreflightDisposition) -> &'static str {
+    match disposition {
+        DesktopPreflightDisposition::NotInstalled => "not_installed",
+        DesktopPreflightDisposition::ManagedReady => "managed_ready",
+        DesktopPreflightDisposition::ManagedStale => "managed_stale",
+        DesktopPreflightDisposition::AttachedReady => "attached_ready",
+    }
+}
diff --git a/studio/src-tauri/src/install.rs b/studio/src-tauri/src/install.rs
index 5760989353..9d672f5e73 100644
--- a/studio/src-tauri/src/install.rs
+++ b/studio/src-tauri/src/install.rs
@@ -1,3 +1,4 @@
+use crate::diagnostics::{self, AttemptLog, DiagnosticsState};
 use log::{error, info, warn};
 use process_wrap::std::*;
 use std::io::BufRead;
@@ -14,6 +15,8 @@ pub struct InstallProcess {
     pub intentional_stop: bool,
     /// Packages needing elevated install, parsed from [TAURI:NEED_SUDO] output.
     pub needed_packages: Vec<String>,
+    /// Current diagnostics attempt; kept after NEEDS_ELEVATION so apt output can be linked.
+    pub current_attempt: Option<AttemptLog>,
 }
 
 impl Default for InstallProcess {
@@ -22,6 +25,7 @@ impl Default for InstallProcess {
             child: None,
             intentional_stop: false,
             needed_packages: Vec::new(),
+            current_attempt: None,
         }
     }
 }
@@ -228,6 +232,8 @@ fn stream_output(
     app: &AppHandle,
     state: &InstallState,
     event_mode: InstallEventMode,
+    diagnostics: DiagnosticsState,
+    attempt: AttemptLog,
     stdout: Option<std::process::ChildStdout>,
     stderr: Option<std::process::ChildStderr>,
 ) -> Vec<std::thread::JoinHandle<()>> {
@@ -236,6 +242,8 @@ fn stream_output(
     if let Some(out) = stdout {
         let app_clone = app.clone();
         let state_clone = Arc::clone(state);
+        let diagnostics_clone = diagnostics.clone();
+        let attempt_clone = attempt.clone();
         threads.push(std::thread::spawn(move || {
             let mut reader = std::io::BufReader::new(out);
             let mut buf = Vec::new();
@@ -245,14 +253,21 @@ fn stream_output(
                     Ok(0) => break,
                     Ok(_) => {
                         let text = String::from_utf8_lossy(trim_line_endings(&buf)).into_owned();
+                        diagnostics::append_phase_line(&attempt_clone.handle, "stdout", &text);
                         // Parse structured Tauri protocol lines
                         if let Some(packages) = text.strip_prefix("[TAURI:NEED_SUDO] ") {
                             let pkgs: Vec<String> =
                                 packages.split_whitespace().map(String::from).collect();
                             if let Ok(mut install) = state_clone.lock() {
-                                install.needed_packages = pkgs;
+                                install.needed_packages = pkgs.clone();
                             }
+                            diagnostics::record_elevation_packages(
+                                &diagnostics_clone,
+                                &attempt_clone,
+                                &pkgs,
+                            );
                         } else if let Some(step) = text.strip_prefix("[TAURI:STEP] ") {
+                            diagnostics::record_step(&diagnostics_clone, &attempt_clone, step);
                             if !event_mode.emit_install_structured_events() {
                                 info!("[install][stdout] {}", text);
                                 let _ = app_clone.emit(event_mode.progress_event(), &text);
@@ -260,12 +275,23 @@ fn stream_output(
                             }
                             let _ = app_clone.emit("install-step", step);
                         } else if let Some(detail) = text.strip_prefix("[TAURI:PROGRESS] ") {
+                            diagnostics::record_progress(
+                                &diagnostics_clone,
+                                &attempt_clone,
+                                detail,
+                            );
                             if !event_mode.emit_install_structured_events() {
                                 info!("[install][stdout] {}", text);
                                 let _ = app_clone.emit(event_mode.progress_event(), detail);
                                 continue;
                             }
                             let _ = app_clone.emit("install-progress-detail", detail);
+                        } else if let Some(marker) = text.strip_prefix("[TAURI:DIAG] ") {
+                            diagnostics::record_diag_marker(
+                                &diagnostics_clone,
+                                &attempt_clone,
+                                marker,
+                            );
                         }
                         // Always forward the raw line
                         info!("[install][stdout] {}", text);
@@ -282,6 +308,7 @@ fn stream_output(
 
     if let Some(err) = stderr {
         let app_clone = app.clone();
+        let attempt_clone = attempt.clone();
         threads.push(std::thread::spawn(move || {
             let mut reader = std::io::BufReader::new(err);
             let mut buf = Vec::new();
@@ -291,6 +318,7 @@ fn stream_output(
                     Ok(0) => break,
                     Ok(_) => {
                         let text = String::from_utf8_lossy(trim_line_endings(&buf)).into_owned();
+                        diagnostics::append_phase_line(&attempt_clone.handle, "stderr", &text);
                         warn!("[install][stderr] {}", text);
                         let _ = app_clone.emit(event_mode.progress_event(), &text);
                     }
@@ -345,30 +373,94 @@ fn wait_for_exit(state: &InstallState) -> Result<(ExitStatus, bool), String> {
 /// Run the install script. Returns Ok(()) on success.
 /// Returns Err("NEEDS_ELEVATION") if system packages need elevated install (Linux only).
 /// Returns Err(message) on other failures.
-pub fn run_install(app: AppHandle, state: InstallState) -> Result<(), String> {
-    run_install_with_event_mode(app, state, InstallEventMode::Full)
+pub fn run_install(
+    app: AppHandle,
+    state: InstallState,
+    diagnostics: DiagnosticsState,
+) -> Result<(), String> {
+    run_install_with_event_mode(app, state, diagnostics, InstallEventMode::Full, None)
 }
 
-pub(crate) fn run_install_for_repair(app: AppHandle, state: InstallState) -> Result<(), String> {
-    run_install_with_event_mode(app, state, InstallEventMode::Repair)
+pub(crate) fn run_install_for_repair(
+    app: AppHandle,
+    state: InstallState,
+    diagnostics: DiagnosticsState,
+    repair_group_id: String,
+) -> Result<(), String> {
+    run_install_with_event_mode(
+        app,
+        state,
+        diagnostics,
+        InstallEventMode::Repair,
+        Some(repair_group_id),
+    )
 }
 
 fn run_install_with_event_mode(
     app: AppHandle,
     state: InstallState,
+    diagnostics: DiagnosticsState,
     event_mode: InstallEventMode,
+    repair_group_id: Option<String>,
 ) -> Result<(), String> {
+    let attempt = match repair_group_id.as_deref() {
+        Some(group_id) => diagnostics::begin_repair_child(&diagnostics, group_id, "install"),
+        None => diagnostics::begin_install_attempt(&diagnostics),
+    };
+    if let Ok(mut install) = state.lock() {
+        install.current_attempt = Some(attempt.clone());
+    }
+
     emit_mode_progress(&app, event_mode, "Starting installation...");
 
-    let (script, args) = resolve_install_script(&app)?;
+    let (script, args) = match resolve_install_script(&app) {
+        Ok(resolved) => resolved,
+        Err(msg) => {
+            diagnostics::finish_attempt(
+                &diagnostics,
+                &attempt,
+                None,
+                false,
+                Some(format!("resolve_install_script: {msg}")),
+            );
+            clear_current_attempt(&state);
+            return Err(msg);
+        }
+    };
+    diagnostics::append_phase_line(
+        &attempt.handle,
+        "meta",
+        &format!("Using script: {}", script.display()),
+    );
     emit_mode_progress(
         &app,
         event_mode,
         &format!("Using script: {}", script.display()),
     );
 
-    let (stdout, stderr) = spawn_script(&script, &args, &state)?;
-    let threads = stream_output(&app, &state, event_mode, stdout, stderr);
+    let (stdout, stderr) = match spawn_script(&script, &args, &state) {
+        Ok(handles) => handles,
+        Err(msg) => {
+            diagnostics::finish_attempt(
+                &diagnostics,
+                &attempt,
+                None,
+                false,
+                Some(format!("spawn_install_script: {msg}")),
+            );
+            clear_current_attempt(&state);
+            return Err(msg);
+        }
+    };
+    let threads = stream_output(
+        &app,
+        &state,
+        event_mode,
+        diagnostics.clone(),
+        attempt.clone(),
+        stdout,
+        stderr,
+    );
 
     // Wait for exit, join reader threads
     let result = wait_for_exit(&state);
@@ -378,12 +470,20 @@ fn run_install_with_event_mode(
 
     match result {
         Ok((status, _)) if status.success() => {
+            diagnostics::finish_attempt(
+                &diagnostics,
+                &attempt,
+                Some(status.to_string()),
+                false,
+                None,
+            );
+            clear_current_attempt(&state);
             if event_mode.emit_terminal_events() {
                 emit_complete(&app);
             }
             Ok(())
         }
-        Ok((status, _)) => {
+        Ok((status, intentional)) => {
             let code = status.code().unwrap_or(-1);
             if code == 2 {
                 // Script needs elevated package install — report to frontend
@@ -391,11 +491,27 @@ fn run_install_with_event_mode(
                     .lock()
                     .map(|i| i.needed_packages.clone())
                     .unwrap_or_default();
+                diagnostics::record_elevation_packages(&diagnostics, &attempt, &packages);
+                diagnostics::finish_attempt(
+                    &diagnostics,
+                    &attempt,
+                    Some(status.to_string()),
+                    intentional,
+                    Some("needs_elevation".to_string()),
+                );
                 info!("[install] Needs elevation for packages: {:?}", packages);
                 let _ = app.emit(event_mode.needs_elevation_event(), &packages);
                 Err("NEEDS_ELEVATION".to_string())
             } else {
                 let msg = format!("Installer exited with code {}", code);
+                diagnostics::finish_attempt(
+                    &diagnostics,
+                    &attempt,
+                    Some(status.to_string()),
+                    intentional,
+                    Some(msg.clone()),
+                );
+                clear_current_attempt(&state);
                 if event_mode.emit_terminal_events() {
                     emit_failed(&app, &msg);
                 }
@@ -403,10 +519,14 @@ fn run_install_with_event_mode(
             }
         }
         Err(msg) if msg == "Installation stopped." => {
+            diagnostics::finish_attempt(&diagnostics, &attempt, None, true, Some(msg.clone()));
+            clear_current_attempt(&state);
             info!("[install] Installation stopped intentionally");
             Err(msg)
         }
         Err(msg) => {
+            diagnostics::finish_attempt(&diagnostics, &attempt, None, false, Some(msg.clone()));
+            clear_current_attempt(&state);
             if event_mode.emit_terminal_events() {
                 emit_failed(&app, &msg);
             }
@@ -415,6 +535,69 @@ fn run_install_with_event_mode(
     }
 }
 
+fn clear_current_attempt(state: &InstallState) {
+    if let Ok(mut install) = state.lock() {
+        install.current_attempt = None;
+    }
+}
+
+pub fn take_pending_repair_group_for_resume(state: &InstallState) -> Option<String> {
+    let mut install = state.lock().ok()?;
+    let repair_group_id = install
+        .current_attempt
+        .as_ref()
+        .and_then(|attempt| attempt.repair_group_id.clone());
+    if repair_group_id.is_some() {
+        install.current_attempt = None;
+    }
+    repair_group_id
+}
+
+pub fn record_pending_elevation_canceled(
+    state: &InstallState,
+    diagnostics: &DiagnosticsState,
+) -> bool {
+    let attempt = state
+        .lock()
+        .ok()
+        .and_then(|mut install| install.current_attempt.take());
+    let Some(attempt) = attempt else {
+        return false;
+    };
+    diagnostics::finish_attempt(
+        diagnostics,
+        &attempt,
+        None,
+        true,
+        Some("elevation_canceled".to_string()),
+    );
+    if let Some(repair_group_id) = attempt.repair_group_id.as_deref() {
+        diagnostics::finish_repair_group(
+            diagnostics,
+            repair_group_id,
+            "canceled",
+            Some("elevation_canceled".to_string()),
+        );
+    }
+    true
+}
+
+pub fn record_install_intentional_stop(state: &InstallState, diagnostics: &DiagnosticsState) {
+    let attempt = state
+        .lock()
+        .ok()
+        .and_then(|install| install.current_attempt.clone());
+    if let Some(attempt) = attempt {
+        diagnostics::finish_attempt(
+            diagnostics,
+            &attempt,
+            None,
+            true,
+            Some("intentional_stop".to_string()),
+        );
+    }
+}
+
 /// Stop a running install process gracefully.
 /// Unix: SIGTERM to process group -> wait up to 5s -> SIGKILL
 /// Windows: hidden taskkill /T /F to terminate the installer tree
@@ -481,65 +664,222 @@ pub fn stop_install(state: &InstallState) -> Result<(), String> {
     }
 }
 
-/// Install system packages with elevated permissions (Linux only).
+/// Install apt system packages with elevated permissions (Linux only).
 /// Uses `elevated-command` crate for native auth dialog.
 #[cfg(target_os = "linux")]
-pub fn install_system_packages(packages: &[String]) -> Result<(), String> {
+pub fn install_system_packages(
+    packages: &[String],
+    state: &InstallState,
+    diagnostics: &DiagnosticsState,
+) -> Result<(), String> {
     use regex::Regex;
     use std::path::Path;
     use std::process::Command as StdCommand;
 
+    let current_attempt = state
+        .lock()
+        .ok()
+        .and_then(|install| install.current_attempt.clone());
+    let elevation_attempt = current_attempt
+        .as_ref()
+        .and_then(|attempt| attempt.repair_group_id.as_deref())
+        .map(|group_id| diagnostics::begin_repair_child(diagnostics, group_id, "elevation"))
+        .or_else(|| current_attempt.clone());
+    if let Some(attempt) = elevation_attempt.as_ref() {
+        diagnostics::append_phase_line(
+            &attempt.handle,
+            "meta",
+            &format!("Starting elevated apt install for: {}", packages.join(", ")),
+        );
+        diagnostics::record_elevation_packages(diagnostics, attempt, packages);
+    }
+
     // Validate package names to prevent injection via elevated command.
     let valid_pkg = Regex::new(r"^[a-zA-Z0-9][a-zA-Z0-9.+\-]*$").unwrap();
     for pkg in packages {
         if !valid_pkg.is_match(pkg) {
-            return Err(format!("Invalid package name: {}", pkg));
+            let msg = format!("Invalid package name: {}", pkg);
+            finish_elevation_failure(diagnostics, elevation_attempt.as_ref(), None, msg.clone());
+            clear_current_attempt(state);
+            return Err(msg);
         }
     }
 
-    // AppImage bundles run on non-Debian distros too. Pick the first package
-    // manager we find. Names in `packages` are Debian-style; callers that want
-    // cross-distro support should translate before invoking.
-    let (program, base_args): (&str, &[&str]) = if Path::new("/usr/bin/apt-get").exists() {
-        ("apt-get", &["install", "-y"])
-    } else if Path::new("/usr/bin/dnf").exists() {
-        ("dnf", &["install", "-y"])
-    } else if Path::new("/usr/bin/zypper").exists() {
-        ("zypper", &["install", "-y"])
-    } else if Path::new("/usr/bin/pacman").exists() {
-        ("pacman", &["-S", "--noconfirm"])
-    } else {
-        return Err(
-            "No supported system package manager found (apt-get, dnf, zypper, pacman)".to_string(),
-        );
-    };
+    // install.sh reports Debian package names. Do not pass them to dnf,
+    // zypper, or pacman where names differ; show an explicit support boundary
+    // instead of offering an elevation flow that is likely to fail.
+    if !Path::new("/usr/bin/apt-get").exists() {
+        let msg = "Automatic system package installation is supported on apt-based Linux distributions (Ubuntu/Debian) only. Install the missing dependencies with your package manager and retry."
+            .to_string();
+        finish_elevation_failure(diagnostics, elevation_attempt.as_ref(), None, msg.clone());
+        clear_current_attempt(state);
+        return Err(msg);
+    }
 
     info!(
-        "[install] Elevated install of packages via {}: {}",
-        program,
+        "[install] Elevated install of apt packages: {}",
         packages.join(", ")
     );
 
-    let mut cmd = StdCommand::new(program);
-    cmd.args(base_args).args(packages);
+    let mut update_cmd = StdCommand::new("apt-get");
+    update_cmd.args(["update", "-y"]);
+    match elevated_command::Command::new(update_cmd).output() {
+        Ok(elevated_update) => {
+            if let Some(attempt) = elevation_attempt.as_ref() {
+                diagnostics::append_phase_line(
+                    &attempt.handle,
+                    "apt-update-status",
+                    &elevated_update.status.to_string(),
+                );
+                append_capped_output(
+                    &attempt.handle,
+                    "apt-update-stdout",
+                    &elevated_update.stdout,
+                );
+                append_capped_output(
+                    &attempt.handle,
+                    "apt-update-stderr",
+                    &elevated_update.stderr,
+                );
+            }
+            if !elevated_update.status.success() {
+                let stderr = capped_output_text(&elevated_update.stderr);
+                warn!(
+                    "[install] apt-get update failed before elevated install; continuing with cached package metadata: {}",
+                    stderr
+                );
+                if let Some(attempt) = elevation_attempt.as_ref() {
+                    diagnostics::append_phase_line(
+                        &attempt.handle,
+                        "apt-update-warning",
+                        "apt-get update failed; continuing with apt-get install",
+                    );
+                }
+            }
+        }
+        Err(error) => {
+            warn!(
+                "[install] Elevated apt update could not run before install; continuing with apt-get install: {}",
+                error
+            );
+            if let Some(attempt) = elevation_attempt.as_ref() {
+                diagnostics::append_phase_line(
+                    &attempt.handle,
+                    "apt-update-error",
+                    &format!(
+                        "apt-get update could not run; continuing with apt-get install: {error}"
+                    ),
+                );
+            }
+        }
+    }
 
-    let elevated = elevated_command::Command::new(cmd)
-        .output()
-        .map_err(|e| format!("Elevated install failed: {}", e))?;
+    let mut install_cmd = StdCommand::new("apt-get");
+    install_cmd.args(["install", "-y"]).args(packages);
 
-    if !elevated.status.success() {
-        let stderr = String::from_utf8_lossy(&elevated.stderr);
+    let elevated_install = match elevated_command::Command::new(install_cmd).output() {
+        Ok(output) => output,
+        Err(error) => {
+            let msg = format!("Elevated install failed: {}", error);
+            finish_elevation_failure(diagnostics, elevation_attempt.as_ref(), None, msg.clone());
+            clear_current_attempt(state);
+            return Err(msg);
+        }
+    };
+    if let Some(attempt) = elevation_attempt.as_ref() {
+        diagnostics::append_phase_line(
+            &attempt.handle,
+            "apt-install-status",
+            &elevated_install.status.to_string(),
+        );
+        append_capped_output(
+            &attempt.handle,
+            "apt-install-stdout",
+            &elevated_install.stdout,
+        );
+        append_capped_output(
+            &attempt.handle,
+            "apt-install-stderr",
+            &elevated_install.stderr,
+        );
+    }
+
+    if !elevated_install.status.success() {
+        let stderr = capped_output_text(&elevated_install.stderr);
+        finish_elevation_failure(
+            diagnostics,
+            elevation_attempt.as_ref(),
+            Some(elevated_install.status.to_string()),
+            format!("Package installation failed: {stderr}"),
+        );
+        clear_current_attempt(state);
         return Err(format!("Package installation failed: {}", stderr));
     }
 
-    info!("[install] Elevated package install succeeded");
+    if let Some(attempt) = elevation_attempt.as_ref() {
+        diagnostics::finish_attempt(
+            diagnostics,
+            attempt,
+            Some(elevated_install.status.to_string()),
+            false,
+            None,
+        );
+    }
+    info!("[install] Elevated apt package install succeeded");
     Ok(())
 }
 
+#[cfg(target_os = "linux")]
+fn finish_elevation_failure(
+    diagnostics: &DiagnosticsState,
+    attempt: Option<&AttemptLog>,
+    exit_status: Option<String>,
+    message: String,
+) {
+    if let Some(attempt) = attempt {
+        diagnostics::finish_attempt(
+            diagnostics,
+            attempt,
+            exit_status,
+            false,
+            Some(message.clone()),
+        );
+        if let Some(repair_group_id) = attempt.repair_group_id.as_deref() {
+            diagnostics::finish_repair_group(diagnostics, repair_group_id, "failed", Some(message));
+        }
+    }
+}
+
+#[cfg(target_os = "linux")]
+fn append_capped_output(handle: &diagnostics::PhaseLogHandle, stream: &str, bytes: &[u8]) {
+    diagnostics::append_phase_line(handle, stream, &capped_output_text(bytes));
+}
+
+#[cfg(any(target_os = "linux", test))]
+fn capped_output_text(bytes: &[u8]) -> String {
+    const MAX_ELEVATED_OUTPUT_BYTES: usize = 64 * 1024;
+    let text = String::from_utf8_lossy(bytes);
+    if text.len() <= MAX_ELEVATED_OUTPUT_BYTES {
+        return text.into_owned();
+    }
+    let boundary = diagnostics::valid_utf8_boundary(&text, MAX_ELEVATED_OUTPUT_BYTES);
+    let mut truncated = text[..boundary].to_string();
+    truncated.push_str("\n[elevated output truncated after 64KiB]");
+    truncated
+}
+
 #[cfg(test)]
 mod tests {
     use super::*;
 
+    #[test]
+    fn elevated_output_cap_is_utf8_boundary_safe() {
+        let text = "é".repeat(40_000);
+        let capped = capped_output_text(text.as_bytes());
+        assert!(capped.ends_with("[elevated output truncated after 64KiB]"));
+        assert!(capped.is_char_boundary(capped.len()));
+    }
+
     #[test]
     fn repair_install_mode_uses_repair_elevation_event() {
         assert_eq!(
diff --git a/studio/src-tauri/src/main.rs b/studio/src-tauri/src/main.rs
index f7d49cce60..b4e718d867 100644
--- a/studio/src-tauri/src/main.rs
+++ b/studio/src-tauri/src/main.rs
@@ -2,6 +2,7 @@
 
 mod commands;
 mod desktop_auth;
+mod diagnostics;
 mod install;
 mod preflight;
 mod process;
@@ -58,6 +59,39 @@ fn setup_logging() {
     }
 }
 
+#[cfg(any(target_os = "windows", target_os = "linux"))]
+fn setup_custom_titlebar(app: &tauri::App) -> Result<(), Box<dyn std::error::Error>> {
+    let window = app.get_webview_window("main").ok_or_else(|| {
+        std::io::Error::new(std::io::ErrorKind::NotFound, "main window not found")
+    })?;
+    window.set_decorations(false)?;
+    Ok(())
+}
+
+fn cleanup_child_processes(app: &tauri::AppHandle) {
+    let diagnostics_state = app
+        .try_state::<diagnostics::DiagnosticsState>()
+        .map(|state| state.inner().clone());
+    if let Some(install_state) = app.try_state::<install::InstallState>() {
+        if let Some(diagnostics) = diagnostics_state.as_ref() {
+            install::record_install_intentional_stop(&install_state, diagnostics);
+        }
+        let _ = install::stop_install(&install_state);
+    }
+    if let Some(update_state) = app.try_state::<update::UpdateState>() {
+        if let Some(diagnostics) = diagnostics_state.as_ref() {
+            update::record_update_intentional_stop(&update_state, diagnostics);
+        }
+        let _ = update::stop_update(&update_state);
+    }
+    if let Some(backend_state) = app.try_state::<process::BackendState>() {
+        let shutdown = app
+            .try_state::<process::ShutdownFlag>()
+            .expect("ShutdownFlag must be managed");
+        let _ = process::stop_backend(&backend_state, &shutdown, diagnostics_state.as_ref());
+    }
+}
+
 fn setup_tray(app: &tauri::App) -> Result<(), Box<dyn std::error::Error>> {
     let open = MenuItemBuilder::with_id("open", "Open Studio").build(app)?;
     let toggle = MenuItemBuilder::with_id("toggle", "Start/Stop Server").build(app)?;
@@ -81,17 +115,15 @@ fn setup_tray(app: &tauri::App) -> Result<(), Box<dyn std::error::Error>> {
                 let _ = app.emit("tray-toggle-server", ());
             }
             "quit" => {
-                let install_state = app.state::<crate::install::InstallState>();
-                let _ = crate::install::stop_install(&install_state);
-                let update_state = app.state::<crate::update::UpdateState>();
-                let _ = crate::update::stop_update(&update_state);
-                // Detach the 5s graceful-wait so the tray click does not
-                // block the Tauri main loop. Exit runs the RunEvent::Exit
-                // safety net which also calls stop_backend synchronously.
-                let shutdown = app.state::<crate::process::ShutdownFlag>().inner().clone();
-                let backend_state = app.state::<crate::process::BackendState>().inner().clone();
-                crate::process::stop_backend_detached(backend_state, shutdown);
-                app.exit(0);
+                // Run cleanup off the menu callback, but only exit after the
+                // backend tree has been reaped. Exiting first can terminate this
+                // process while a detached cleanup thread is still waiting,
+                // leaving the backend orphaned.
+                let app_handle = app.clone();
+                std::thread::spawn(move || {
+                    cleanup_child_processes(&app_handle);
+                    app_handle.exit(0);
+                });
             }
             _ => {}
         })
@@ -133,6 +165,8 @@ fn main() {
         .plugin(tauri_plugin_process::init())
         .plugin(tauri_plugin_opener::init())
         .plugin(tauri_plugin_updater::Builder::new().build())
+        .plugin(tauri_plugin_clipboard_manager::init())
+        .manage(diagnostics::new_diagnostics_state())
         .manage(install::new_install_state())
         .manage(new_backend_state())
         .manage(process::new_shutdown_flag())
@@ -149,10 +183,14 @@ fn main() {
             commands::open_logs_dir,
             commands::start_backend_update,
             commands::start_managed_repair,
+            commands::cancel_pending_elevation,
             commands::install_system_packages,
             desktop_auth::desktop_auth,
+            diagnostics::collect_support_diagnostics,
         ])
         .setup(|app| {
+            #[cfg(any(target_os = "windows", target_os = "linux"))]
+            setup_custom_titlebar(app)?;
             setup_tray(app)?;
             Ok(())
         })
@@ -172,18 +210,7 @@ fn main() {
         .run(|app, event| {
             if let tauri::RunEvent::Exit = event {
                 // Cleanup on ALL exit paths — safety net for non-tray exits
-                if let Some(install_state) = app.try_state::<install::InstallState>() {
-                    let _ = install::stop_install(&install_state);
-                }
-                if let Some(update_state) = app.try_state::<update::UpdateState>() {
-                    let _ = update::stop_update(&update_state);
-                }
-                if let Some(backend_state) = app.try_state::<process::BackendState>() {
-                    let shutdown = app
-                        .try_state::<process::ShutdownFlag>()
-                        .expect("ShutdownFlag must be managed");
-                    let _ = process::stop_backend(&backend_state, &shutdown);
-                }
+                cleanup_child_processes(app);
             }
         });
 }
diff --git a/studio/src-tauri/src/process.rs b/studio/src-tauri/src/process.rs
index beba8c1208..ed2659c0d9 100644
--- a/studio/src-tauri/src/process.rs
+++ b/studio/src-tauri/src/process.rs
@@ -1,3 +1,4 @@
+use crate::diagnostics::{self, BackendLog, DiagnosticsState};
 use log::{error, info, warn};
 use process_wrap::std::*;
 use regex::Regex;
@@ -15,6 +16,8 @@ pub struct BackendProcess {
     pub port: Option<u16>,
     pub logs: VecDeque<String>,
     pub intentional_stop: bool,
+    pub generation: u64,
+    pub diagnostics_session: Option<BackendLog>,
 }
 
 impl Default for BackendProcess {
@@ -24,6 +27,8 @@ impl Default for BackendProcess {
             port: None,
             logs: VecDeque::with_capacity(MAX_LOG_LINES),
             intentional_stop: false,
+            generation: 0,
+            diagnostics_session: None,
         }
     }
 }
@@ -245,24 +250,48 @@ pub fn start_backend(
     state: &BackendState,
     port: u16,
     shutdown: &ShutdownFlag,
-) -> Result<(), String> {
-    let bin = resolve_backend_binary()?;
+    diagnostics_state: &DiagnosticsState,
+) -> Result<u64, String> {
+    let bin = match resolve_backend_binary() {
+        Ok(bin) => bin,
+        Err(msg) => {
+            diagnostics::record_backend_start_failure(
+                diagnostics_state,
+                Some(port),
+                None,
+                "resolve_backend_binary",
+                &msg,
+            );
+            return Err(msg);
+        }
+    };
 
     shutdown.store(false, Ordering::SeqCst);
 
-    // Reset state
-    {
+    // Reset state and invalidate any readers from an older backend generation
+    // before we release the lock and spawn a new child.
+    let generation = {
         let mut proc = state.lock().map_err(|e| e.to_string())?;
         if proc.child.is_some() {
             return Err("Backend is already running.".to_string());
         }
+        proc.generation = proc.generation.wrapping_add(1);
         proc.port = None;
         proc.logs.clear();
         proc.intentional_stop = false;
-    }
+        proc.diagnostics_session = None;
+        proc.generation
+    };
+
+    let backend_log = diagnostics::begin_backend_session(diagnostics_state, port, generation);
 
     let args = backend_args(port);
     info!("Starting backend: {:?} {}", bin, args.join(" "));
+    diagnostics::append_phase_line(
+        &backend_log.handle,
+        "meta",
+        &format!("Starting backend: {:?} {}", bin, args.join(" ")),
+    );
 
     let mut cmd = Command::new(&bin);
     cmd.args(&args)
@@ -289,9 +318,17 @@ pub fn start_backend(
 
         const CREATE_NEW_PROCESS_GROUP: u32 = 0x00000200;
         cmd.creation_flags(CREATE_NEW_PROCESS_GROUP | CREATE_NO_WINDOW);
-        let child = cmd
-            .spawn()
-            .map_err(|e| format!("Failed to spawn backend: {}", e))?;
+        let child = cmd.spawn().map_err(|e| {
+            let msg = format!("Failed to spawn backend: {}", e);
+            diagnostics::record_backend_start_failure(
+                diagnostics_state,
+                Some(port),
+                Some(generation),
+                "spawn_backend",
+                &msg,
+            );
+            msg
+        })?;
         Box::new(child)
     };
 
@@ -300,25 +337,45 @@ pub fn start_backend(
         // Keep the backend tree in a process group on Unix for cleanup.
         let mut wrap = CommandWrap::from(cmd);
         wrap.wrap(ProcessGroup::leader());
-        wrap.spawn()
-            .map_err(|e| format!("Failed to spawn backend: {}", e))?
+        wrap.spawn().map_err(|e| {
+            let msg = format!("Failed to spawn backend: {}", e);
+            diagnostics::record_backend_start_failure(
+                diagnostics_state,
+                Some(port),
+                Some(generation),
+                "spawn_backend",
+                &msg,
+            );
+            msg
+        })?
     };
 
     let stdout = child.stdout().take();
     let stderr = child.stderr().take();
 
-    // Store child in state
+    // Store child in state for the already-selected generation.
     {
         let mut proc = state.lock().map_err(|e| e.to_string())?;
         proc.child = Some(child);
+        proc.diagnostics_session = Some(backend_log.clone());
     }
 
     // Spawn stdout reader thread
     if let Some(stdout) = stdout {
         let app_handle = app.clone();
         let state_clone = Arc::clone(state);
+        let diagnostics_clone = diagnostics_state.clone();
+        let backend_log_clone = backend_log.clone();
         std::thread::spawn(move || {
-            read_output_stream(stdout, &app_handle, &state_clone, false);
+            read_output_stream(
+                stdout,
+                &app_handle,
+                &state_clone,
+                &diagnostics_clone,
+                &backend_log_clone,
+                false,
+                generation,
+            );
         });
     }
 
@@ -326,12 +383,22 @@ pub fn start_backend(
     if let Some(stderr) = stderr {
         let app_handle = app.clone();
         let state_clone = Arc::clone(state);
+        let diagnostics_clone = diagnostics_state.clone();
+        let backend_log_clone = backend_log.clone();
         std::thread::spawn(move || {
-            read_output_stream(stderr, &app_handle, &state_clone, true);
+            read_output_stream(
+                stderr,
+                &app_handle,
+                &state_clone,
+                &diagnostics_clone,
+                &backend_log_clone,
+                true,
+                generation,
+            );
         });
     }
 
-    Ok(())
+    Ok(generation)
 }
 
 /// Read lines from a child process stream (stdout or stderr).
@@ -341,7 +408,10 @@ fn read_output_stream<R: std::io::Read>(
     stream: R,
     app: &AppHandle,
     state: &BackendState,
+    diagnostics_state: &DiagnosticsState,
+    backend_log: &BackendLog,
     is_stderr: bool,
+    generation: u64,
 ) {
     let mut reader = std::io::BufReader::new(stream);
     let port_re = Regex::new(r"TAURI_PORT=(\d+)").unwrap();
@@ -359,27 +429,56 @@ fn read_output_stream<R: std::io::Read>(
                     text.clone()
                 };
 
+                diagnostics::append_phase_line(
+                    &backend_log.handle,
+                    if is_stderr { "stderr" } else { "stdout" },
+                    &text,
+                );
+
                 // Check for TAURI_PORT on stdout only
-                if !is_stderr {
-                    if let Some(caps) = port_re.captures(&text) {
-                        if let Some(port_str) = caps.get(1) {
-                            if let Ok(port) = port_str.as_str().parse::<u16>() {
-                                info!("Detected backend port: {}", port);
-                                if let Ok(mut proc) = state.lock() {
-                                    proc.port = Some(port);
-                                }
-                                let _ = app.emit("server-port", port);
-                            }
+                let detected_port = if !is_stderr {
+                    port_re
+                        .captures(&text)
+                        .and_then(|caps| caps.get(1))
+                        .and_then(|port_str| port_str.as_str().parse::<u16>().ok())
+                } else {
+                    None
+                };
+
+                // Buffer the log line only for the current backend generation.
+                // Old reader threads can briefly outlive a stop/start cycle;
+                // they must not overwrite the new backend's port or logs.
+                let mut should_record_port = None;
+                let current_generation = if let Ok(mut proc) = state.lock() {
+                    if proc.generation != generation {
+                        false
+                    } else {
+                        if let Some(port) = detected_port {
+                            proc.port = Some(port);
+                            should_record_port = Some(port);
                         }
+                        if proc.logs.len() >= MAX_LOG_LINES {
+                            proc.logs.pop_front();
+                        }
+                        proc.logs.push_back(log_line.clone());
+                        true
                     }
+                } else {
+                    false
+                };
+
+                if !current_generation {
+                    break;
                 }
 
-                // Buffer the log line
-                if let Ok(mut proc) = state.lock() {
-                    if proc.logs.len() >= MAX_LOG_LINES {
-                        proc.logs.pop_front();
-                    }
-                    proc.logs.push_back(log_line.clone());
+                if let Some(port) = should_record_port {
+                    diagnostics::record_backend_port(
+                        diagnostics_state,
+                        &backend_log.session_id,
+                        port,
+                    );
+                    info!("Detected backend port: {}", port);
+                    let _ = app.emit("server-port", port);
                 }
 
                 info!("[backend] {}", log_line);
@@ -400,12 +499,18 @@ fn read_output_stream<R: std::io::Read>(
 
     // Stream closed. Only the stdout reader checks for crashes.
     if !is_stderr {
+        let mut exit_record: Option<(String, bool)> = None;
+        let mut emit_crash = false;
         if let Ok(mut proc) = state.lock() {
+            if proc.generation != generation {
+                return;
+            }
             let intentional = proc.intentional_stop;
             let exited = if let Some(ref mut child) = proc.child {
                 match child.try_wait() {
                     Ok(Some(status)) => {
                         info!("Backend stdout stream ended with status: {}", status);
+                        exit_record = Some((status.to_string(), intentional));
                         true
                     }
                     Ok(None) => {
@@ -423,20 +528,38 @@ fn read_output_stream<R: std::io::Read>(
 
             if exited {
                 proc.child = None;
-                if !intentional {
-                    error!("Backend process stdout closed unexpectedly (crash detected)");
-                    let _ = app.emit("server-crashed", ());
-                }
+                proc.diagnostics_session = None;
+                emit_crash = !intentional;
             }
         }
+        if let Some((status, intentional)) = exit_record {
+            diagnostics::record_backend_exit(
+                diagnostics_state,
+                &backend_log.session_id,
+                Some(status),
+                intentional,
+                None,
+            );
+        }
+        if emit_crash {
+            error!("Backend process stdout closed unexpectedly (crash detected)");
+            let _ = app.emit("server-crashed", ());
+        }
     }
 }
 
 /// Graceful shutdown of the backend process and its entire subprocess tree.
 /// Unix: SIGTERM to process group -> wait up to 5s -> SIGKILL to group
 /// Windows: CTRL_BREAK_EVENT -> wait up to 5s -> hidden taskkill /T /F
-pub fn stop_backend(state: &BackendState, shutdown: &ShutdownFlag) -> Result<(), String> {
+pub fn stop_backend(
+    state: &BackendState,
+    shutdown: &ShutdownFlag,
+    diagnostics_state: Option<&DiagnosticsState>,
+) -> Result<(), String> {
     shutdown.store(true, Ordering::SeqCst);
+    if let Some(diagnostics_state) = diagnostics_state {
+        diagnostics::record_backend_intentional_stop(diagnostics_state);
+    }
 
     // Extract the child and mark intentional stop.
     // We take the child OUT of the mutex so we don't hold the lock during the wait loop.
@@ -526,12 +649,3 @@ pub fn stop_backend(state: &BackendState, shutdown: &ShutdownFlag) -> Result<(),
         Ok(())
     }
 }
-
-/// Spawn `stop_backend` on a background thread and return immediately.
-/// Used by the tray "quit" path so the 5s graceful-wait does not block
-/// the Tauri main event loop before `app.exit(0)` fires.
-pub fn stop_backend_detached(state: BackendState, shutdown: ShutdownFlag) {
-    std::thread::spawn(move || {
-        let _ = stop_backend(&state, &shutdown);
-    });
-}
diff --git a/studio/src-tauri/src/update.rs b/studio/src-tauri/src/update.rs
index 477f18ce28..e79432312a 100644
--- a/studio/src-tauri/src/update.rs
+++ b/studio/src-tauri/src/update.rs
@@ -1,3 +1,4 @@
+use crate::diagnostics::{self, AttemptLog, DiagnosticsState};
 use log::{error, info, warn};
 use process_wrap::std::*;
 use std::io::BufRead;
@@ -10,6 +11,7 @@ use tauri::{AppHandle, Emitter};
 pub struct UpdateProcess {
     pub child: Option<Box<dyn ChildWrapper + Send>>,
     pub intentional_stop: bool,
+    pub current_attempt: Option<AttemptLog>,
 }
 
 impl Default for UpdateProcess {
@@ -17,6 +19,7 @@ impl Default for UpdateProcess {
         Self {
             child: None,
             intentional_stop: false,
+            current_attempt: None,
         }
     }
 }
@@ -88,6 +91,8 @@ fn spawn_update(
 fn stream_output(
     app: &AppHandle,
     progress_event: &'static str,
+    diagnostics: DiagnosticsState,
+    attempt: AttemptLog,
     stdout: Option<std::process::ChildStdout>,
     stderr: Option<std::process::ChildStderr>,
 ) -> Vec<std::thread::JoinHandle<()>> {
@@ -95,11 +100,29 @@ fn stream_output(
 
     if let Some(out) = stdout {
         let app_clone = app.clone();
+        let diagnostics_clone = diagnostics.clone();
+        let attempt_clone = attempt.clone();
         threads.push(std::thread::spawn(move || {
             let reader = std::io::BufReader::new(out);
             for line in reader.lines() {
                 match line {
                     Ok(text) => {
+                        diagnostics::append_phase_line(&attempt_clone.handle, "stdout", &text);
+                        if let Some(step) = text.strip_prefix("[TAURI:STEP] ") {
+                            diagnostics::record_step(&diagnostics_clone, &attempt_clone, step);
+                        } else if let Some(progress) = text.strip_prefix("[TAURI:PROGRESS] ") {
+                            diagnostics::record_progress(
+                                &diagnostics_clone,
+                                &attempt_clone,
+                                progress,
+                            );
+                        } else if let Some(marker) = text.strip_prefix("[TAURI:DIAG] ") {
+                            diagnostics::record_diag_marker(
+                                &diagnostics_clone,
+                                &attempt_clone,
+                                marker,
+                            );
+                        }
                         info!("[update][stdout] {}", text);
                         let _ = app_clone.emit(progress_event, &text);
                     }
@@ -114,11 +137,13 @@ fn stream_output(
 
     if let Some(err) = stderr {
         let app_clone = app.clone();
+        let attempt_clone = attempt.clone();
         threads.push(std::thread::spawn(move || {
             let reader = std::io::BufReader::new(err);
             for line in reader.lines() {
                 match line {
                     Ok(text) => {
+                        diagnostics::append_phase_line(&attempt_clone.handle, "stderr", &text);
                         warn!("[update][stderr] {}", text);
                         let _ = app_clone.emit(progress_event, &text);
                     }
@@ -167,26 +192,54 @@ fn wait_for_exit(state: &UpdateState) -> Result<(ExitStatus, bool), String> {
 
 // ── Public API ──
 
-pub fn run_backend_update(app: AppHandle, state: UpdateState) -> Result<(), String> {
-    run_backend_update_with_terminal_events(app, state, true)
+pub fn run_backend_update(
+    app: AppHandle,
+    state: UpdateState,
+    diagnostics: DiagnosticsState,
+) -> Result<(), String> {
+    run_backend_update_with_terminal_events(app, state, diagnostics, true, None)
 }
 
 pub(crate) fn run_backend_update_for_repair(
     app: AppHandle,
     state: UpdateState,
+    diagnostics: DiagnosticsState,
+    repair_group_id: String,
 ) -> Result<(), String> {
-    run_backend_update_with_terminal_events(app, state, false)
+    run_backend_update_with_terminal_events(app, state, diagnostics, false, Some(repair_group_id))
 }
 
 fn run_backend_update_with_terminal_events(
     app: AppHandle,
     state: UpdateState,
+    diagnostics: DiagnosticsState,
     terminal_events: bool,
+    repair_group_id: Option<String>,
 ) -> Result<(), String> {
-    let bin = crate::process::find_unsloth_binary()
-        .ok_or("Unsloth binary not found. Cannot run update.")?;
+    let attempt = match repair_group_id.as_deref() {
+        Some(group_id) => diagnostics::begin_repair_child(&diagnostics, group_id, "update"),
+        None => diagnostics::begin_update_attempt(&diagnostics),
+    };
+    if let Ok(mut update) = state.lock() {
+        update.current_attempt = Some(attempt.clone());
+    }
+
+    let bin = match crate::process::find_unsloth_binary() {
+        Some(bin) => bin,
+        None => {
+            let msg = "Unsloth binary not found. Cannot run update.".to_string();
+            diagnostics::finish_attempt(&diagnostics, &attempt, None, false, Some(msg.clone()));
+            clear_current_attempt(&state);
+            return Err(msg);
+        }
+    };
 
     info!("[update] Starting backend update via {:?}", bin);
+    diagnostics::append_phase_line(
+        &attempt.handle,
+        "meta",
+        &format!("Starting backend update via {:?}", bin),
+    );
     let progress_event = if terminal_events {
         "update-progress"
     } else {
@@ -194,8 +247,28 @@ fn run_backend_update_with_terminal_events(
     };
     let _ = app.emit(progress_event, "Starting backend update...");
 
-    let (stdout, stderr) = spawn_update(&bin, &state)?;
-    let threads = stream_output(&app, progress_event, stdout, stderr);
+    let (stdout, stderr) = match spawn_update(&bin, &state) {
+        Ok(handles) => handles,
+        Err(msg) => {
+            diagnostics::finish_attempt(
+                &diagnostics,
+                &attempt,
+                None,
+                false,
+                Some(format!("spawn_update: {msg}")),
+            );
+            clear_current_attempt(&state);
+            return Err(msg);
+        }
+    };
+    let threads = stream_output(
+        &app,
+        progress_event,
+        diagnostics.clone(),
+        attempt.clone(),
+        stdout,
+        stderr,
+    );
 
     let result = wait_for_exit(&state);
     for handle in threads {
@@ -204,19 +277,43 @@ fn run_backend_update_with_terminal_events(
 
     match result {
         Ok((status, _)) if status.success() => {
+            diagnostics::finish_attempt(
+                &diagnostics,
+                &attempt,
+                Some(status.to_string()),
+                false,
+                None,
+            );
+            clear_current_attempt(&state);
             info!("[update] Backend update complete");
             if terminal_events {
                 let _ = app.emit("update-complete", ());
             }
             Ok(())
         }
-        Ok((_status, intentional)) if intentional => {
+        Ok((status, intentional)) if intentional => {
+            diagnostics::finish_attempt(
+                &diagnostics,
+                &attempt,
+                Some(status.to_string()),
+                true,
+                Some("Update stopped.".to_string()),
+            );
+            clear_current_attempt(&state);
             info!("[update] Update stopped intentionally");
             Err("Update stopped.".to_string())
         }
-        Ok((status, _)) => {
+        Ok((status, intentional)) => {
             let code = status.code().unwrap_or(-1);
             let msg = format!("Update exited with code {}", code);
+            diagnostics::finish_attempt(
+                &diagnostics,
+                &attempt,
+                Some(status.to_string()),
+                intentional,
+                Some(msg.clone()),
+            );
+            clear_current_attempt(&state);
             error!("[update] {}", msg);
             if terminal_events {
                 let _ = app.emit("update-failed", &msg);
@@ -224,6 +321,8 @@ fn run_backend_update_with_terminal_events(
             Err(msg)
         }
         Err(msg) => {
+            diagnostics::finish_attempt(&diagnostics, &attempt, None, false, Some(msg.clone()));
+            clear_current_attempt(&state);
             error!("[update] {}", msg);
             if terminal_events {
                 let _ = app.emit("update-failed", &msg);
@@ -233,6 +332,28 @@ fn run_backend_update_with_terminal_events(
     }
 }
 
+fn clear_current_attempt(state: &UpdateState) {
+    if let Ok(mut update) = state.lock() {
+        update.current_attempt = None;
+    }
+}
+
+pub fn record_update_intentional_stop(state: &UpdateState, diagnostics: &DiagnosticsState) {
+    let attempt = state
+        .lock()
+        .ok()
+        .and_then(|update| update.current_attempt.clone());
+    if let Some(attempt) = attempt {
+        diagnostics::finish_attempt(
+            diagnostics,
+            &attempt,
+            None,
+            true,
+            Some("intentional_stop".to_string()),
+        );
+    }
+}
+
 pub fn stop_update(state: &UpdateState) -> Result<(), String> {
     let mut child = {
         let mut update = match state.lock() {
diff --git a/studio/src-tauri/tauri.conf.json b/studio/src-tauri/tauri.conf.json
index 49819f8a25..fce43889b5 100644
--- a/studio/src-tauri/tauri.conf.json
+++ b/studio/src-tauri/tauri.conf.json
@@ -2,8 +2,14 @@
   "productName": "Unsloth Studio (Desktop)",
   "identifier": "ai.unsloth.studio",
   "build": {
-    "beforeDevCommand": "npm --prefix ../frontend run dev -- --port 5173 --strictPort",
-    "beforeBuildCommand": "npm --prefix ../frontend run build",
+    "beforeDevCommand": {
+      "cwd": "../frontend",
+      "script": "npm run dev -- --port 5173 --strictPort"
+    },
+    "beforeBuildCommand": {
+      "cwd": "../frontend",
+      "script": "npm run build"
+    },
     "frontendDist": "../frontend/dist",
     "devUrl": "http://localhost:5173"
   },
@@ -21,18 +27,13 @@
         "visible": false,
         "resizable": false
       }
-    ],
-    "trayIcon": {
-      "iconPath": "icons/icon.png",
-      "id": "main",
-      "tooltip": "Unsloth Studio (Desktop)"
-    }
+    ]
   },
   "plugins": {
     "updater": {
-      "pubkey": "dW50cnVzdGVkIGNvbW1lbnQ6IG1pbmlzaWduIHB1YmxpYyBrZXk6IDJENjJENDYxMTQ1QTYyOEIKUldTTFlsb1VZZFJpTFY1aHgyRkRqZUdjWGg0Sm1BVEZoMDVWME5PdE42bjZiekFRc1Fsb0JNbmIK",
+      "pubkey": "dW50cnVzdGVkIGNvbW1lbnQ6IG1pbmlzaWduIHB1YmxpYyBrZXk6IEQ2RDhGMDMzNDNEMTlGMkQKUldRdG45RkRNL0RZMXZsaVo2TElUQlVHb1hVNWEyajhUOGFXeTdNVDFZTFdBVUtlZUh5L2wwWVYK",
       "endpoints": [
-        "https://github.com/danielhanchen/unsloth-staging-2/releases/latest/download/latest.json"
+        "https://github.com/unslothai/unsloth/releases/latest/download/latest.json"
       ],
       "windows": {
         "installMode": "passive"
@@ -42,11 +43,14 @@
   "bundle": {
     "active": true,
     "createUpdaterArtifacts": true,
+    "publisher": "Unsloth AI",
+    "homepage": "https://unsloth.ai/",
+    "copyright": "© 2026 Unsloth AI. All rights reserved.",
+    "license": "AGPL-3.0-only",
     "targets": ["app", "appimage", "deb", "dmg", "nsis"],
     "resources": {
       "../../install.sh": "install.sh",
-      "../../install.ps1": "install.ps1",
-      "../install_python_stack.py": "install_python_stack.py"
+      "../../install.ps1": "install.ps1"
     },
     "icon": [
       "icons/32x32.png",
@@ -54,6 +58,18 @@
       "icons/icon.ico",
       "icons/icon.icns"
     ],
+    "windows": {
+      "nsis": {
+        "installerHooks": "./windows/hooks.nsh",
+        "template": "./windows/installer.nsi",
+        "installerIcon": "./icons/icon.ico",
+        "headerImage": "./windows/branding/nsis-header.bmp",
+        "sidebarImage": "./windows/branding/nsis-sidebar.bmp",
+        "installMode": "currentUser",
+        "languages": ["English"],
+        "displayLanguageSelector": false
+      }
+    },
     "linux": {
       "deb": {
         "postRemoveScript": "./linux/postremove.sh"
diff --git a/studio/src-tauri/tauri.windows.conf.json b/studio/src-tauri/tauri.windows.conf.json
index a68cf6351f..83f685cda4 100644
--- a/studio/src-tauri/tauri.windows.conf.json
+++ b/studio/src-tauri/tauri.windows.conf.json
@@ -4,12 +4,6 @@
       "signCommand": {
         "cmd": "trusted-signing-cli",
         "args": ["-e", "https://eus.codesigning.azure.net", "-d", "Unsloth Studio (Desktop)", "%1"]
-      },
-      "nsis": {
-        "installerHooks": "./windows/hooks.nsh",
-        "installerIcon": "./icons/icon.ico",
-        "headerImage": "./windows/branding/nsis-header.bmp",
-        "sidebarImage": "./windows/branding/nsis-sidebar.bmp"
       }
     }
   }
diff --git a/studio/src-tauri/windows/branding/nsis-header.bmp b/studio/src-tauri/windows/branding/nsis-header.bmp
index 7eacdce7db5026855113aa5f03c47663d1e17561..b675c4e1c3e47df793b4c5b28ba4dbd710b8fe54 100644
GIT binary patch
literal 102654
zcmeI51(2RakcJZucXtc!R=5+~-GUR`y$ZMBP`C#Pq;R>zp>TJna1HM6aQEy!q;Kx~
zzx{U!+1*^|sbVt!OixeuJKf{o(tGtkPwCmCB`nB)v+&;>{M?Iwd-j-=xJQpi|0-|&
z@mKWU84`g+AQ4Ce5`jb@5l93QfkYq?NCXmrL?97J1QLNnAQ4Ce5`jb@5l93QfkYq?
zNCXmrL?97J1QLNnAQ4Ce5`jb@5l93QfkYq?NCXmrL?97J1QLNnAQ4Ce5`jb@5l93Q
zfkYq?NCXmrL?97J1QLNnAQ4Ce5`jb@5l93QfkYq?NCXmrL?97J1QLNnAQ4Ce5`jb@
z5l93QfkYq?NCXmrL?97J1QLNnAQ4Ce5`jb@5l93QfkYq?NCXmrL?97J1QLNnAQ4Ce
z5`jb@5l93QfkYq?NCXmrL?97J1QLNnAQ4Ce5`jb@5l93QfkYq?NCXmrL?97J1QLPa
z4uL=Z_(PpO`skyx&px|PpFWE(zWDs}&%f-l%kH`7p7-8+Z)p7d^Uq&()m8u18m_+j
z>c<{?jG{kV1BtiaemhF9x#pVt?z=B4)+T=X>8BfSyz%<$um8{1aP76%e(}W@!SJ`=
zezO+rpLpVlQ%*T$tF5*|qy-jOV8I0!TxFG2_S<j2JMOqcn%{r_9kQS&@$}PABcY-q
z)w9n&OJ>kWNgS<w|NZyZU3cAXyY05Z3M<gWLJKXl+G?xqzyJPs-g#%#qLrxFRw5Wa
z^2j5{9((LI+ibJZMjLIn-F6pVc;R>7eP^vlAAPix?d6wWCO=v!Ih*;;JMRQNDBX3}
zT_sJs=@9ANci%nfq?1r^iYcZTdE}9M4DVr-QAU|+s;O38b=5P@I0M@js6eA#+`5b)
z?CPL{4jN~iaca9|$t9Q6-LR#S07pPhIH(tzmZkH|GY_S;rlr@v|Ni?Uj4(o3<-rFZ
zq)MPPXwaZT4?T4D*=H|lD(N}qnB$5ot`IKUvifG5ZB|nD?c2BHEWMNr1xFll1f5mV
zm+9GNo9*0l&n>H%Zj9CIo_p?Dc;SUh4yJFu`DUxExZ;Xs>466xNPcKtvNe-|v8112
zh8aqlCTRe%f8TrWJw}5u#~icd^RG&eJMOp)u%CYVDR70&H-WCpinJNVXJ!_A<Ud*i
zu3_L;&z?OkV$wO}kV8l_gpM@QNF>%?du=t+RL(i)95N{yal{cxFqKeR>w{)~@WBUK
zg^g0>;fEi#3W6~v*b4asY2a+|XUN7BDQIw37@Uq;aB#riamO7YDDB+z0}nh9-iOMt
z9jtoGzx<3QSQxzZ)>{=OHfn6o9(UYvt@8GuqCyg@NGuFiUw!pbw!;oPtSz<*dH(t5
zOONg(7<S5b+e!lkfztENJ1=HE6%3^mbvTaE)P7XQvBnz95gP*&oVL*G;{3%HTMP-A
zxT9+jd$6&O7$!8v7-Qg8P+Cf2t+m!d)R+eyuBk{BO7Uw-Yc(|!9~^!3(V-M55)VD}
zkh-t5(n|a!V7+it7;_@lAVa*Jciy=aX6voDhB7Mdu)_|Nh7X#!^2#gaMl0%WThSt#
zLO`q&A?=e-K3Vcn+Q`va0w3xIE7H$A^NjSYa-)a@bA51-AY$dji6BE4<zZnE1Z^ci
zSde$!byw;VLPye}CJZCpN>vku3J5^DhGK9Q7z#LK%{iAmtrD?14rckk|CUv1#c19-
zrFJ45nM%OT38*aqyE6ZnV~#<p7$-^BDj4#ItwK2zeNeW}gQ&mnzWc(TMoVF!VI#xJ
zaHg4N8g<lBAuy0H+5Pw5Pu-t?{+YOm4{F$K^sG+rlc4KsdmqlqCISpZY<uBfu4n8`
zh7_fe54EiLmy|tvyz<H`R3vnP4=SDxh=QS7s1s5}qDBAzyEUMwqC#zs9Lk;wo+c`s
z8+kU;SGYA+spQy+6+VrvBZ-`wtUc_bE{iz>nP{Sk5LJyruW6>4X3H(NJm7!>=tO5p
zXizf<RHXBK@Y`>{onV3q0#Y`f7hQDGtFONL!w)|I4mSVs#~-s;4FbDhlT9|Ew7CyD
zJAC!kR})S+A=OnRdYqjv2Pv0de)+(G1KFmr>az)rQ#W*eFZ)QRh8|*R0;u)ZUmwX~
z2;0v;|J+)g?Ssn09u|4ZUJFBAW|?J>hchVl7aZ~&b<|Ou!l`LM!B4f=Yp=aHP&)ne
z(=WdG;%&Fxmi$InwHEx>TW>v%1#Y<E1~_vrcl6Om18K=wp8M>x4?NF2^Gw!G2GS>-
zaKiT6Z_gMjeMoB+jt4<fF+7)03`$Kf0Is|4x~H9X8m-@c`|T`pnD5v@G;&i~#G^aJ
zs_e}!PP>CEVMkzwh;ooeB#uLuUw(Ot0(`4<k`L-cbRR8lxZ#E#8`dJjIRi9wmV(iJ
zkSTmy^PrfsaHXRDqArt6GD)1WmTj@Tl)xse;t@@BpMK35ln`k%q&bf*S%qw!>Vr<o
z9D%yFsdU-A^UnMD<BwOQ@+U`0(@r}rVuw9gmlNiozxCEzO9ImIbE&14a=#qJkzmO&
z>7<iNDL5~@@IsU}Cg_N{=9()wE5=h#Jr$I<-F92ajr8W5Z_W{45R?i0ar)_}hbEj_
zhU1Ml-j`o~SyFaOdd)T03{6G8U%!4zf`J)g*}T(EJ5`7~$p>i{8JOuVx#SXs!L7(d
z60Dt=G*<}}F)E6!s4{)yzY1LIgTMazE45fOjyBq8RE!}M3n7yt8&YwIbuP|r<RA1i
z&pfk5#3qoI{k}ZyYo-qv{_@K&awCS)@u5?FkcMb?i!HW57)J~eb}rK&MFeON%rr5Y
zD%jUwf33Nyud!Dlu0Y&Dhm!_pCOf&onJ!RSnjpxT>lIEWgFbc*P(nbL5fFnjYVm~I
zrkidGeGITTGMS3O;3g|Qnz_%`OAbl`TZ+hMh>#mGL1ehzvSK&J0v@I{%Pg}X3?L|Z
z2!=YQU02Wr_ChE5Ae|B&fh2Y!F=CX0;WY>ho|uZ(;(?*k*7_h_j3O+1sHLaaiJF=7
zZ)k18%u`Q2HLBQ_*QVMJlrO#X5(ESo%G28%g%6(o>^vV79}Hd_Qxan?o<95RGb9N$
zUwrY!TW+}}PKwNDzzHo%P4On`a!pC8y)PLtNWS^zo0MYe(Q!Zi_@i1QlAKw}LabmO
zzKI3|R3u0bhe)ry^2$_kLT8$WKrvWGlFun3F<SugsS;4rTrB&^lp8U3I?$_>n6MEG
zN?Y&x*r6N2G{{h=Q4$ql882~X`yc|)7XU~cBY;u?8aMlW`}XB{Bp|hCS&%?ukJtL3
zMeD4y4i#I*k=9c_?yRX*X~njPH7X?O)vK4yiB%kCYWBPVpp;<eB9CTiX>ij{I?o5y
z;^dQ0MiS+W#2$O>ffykVJFgaTTf;89>|&APxrYLrc74z$HrQYTs$(HCr*}BH<@Bs~
z@7`~`@dib*1^g+eoYEHcK^=<i8+&scgmp5TBj^b#!Z7Q}Ikg&nkjz<UomKkKi_^Tx
zC!d^kCaYO5WGT{mS>jSfrAXNO{rmTqLwCh-9I8BCozz;RnRFJygS_;eHxCMkeMcHc
z;B5&v>8z$)(jW#wJQaqYUL4r7wLYj8Uiwp46RK?P8vCN6yS3Q<Ly3g;exnb*)q0s2
zwoz)-c|Is&9ASY*z@orxY>|OAGcy#=hg#qp99Y9Eo|;w`fzJ4~^Prr$I-$CD&Lwzx
zWWDs#OH%|0a%OjLoBE(bBFA5Y1!?)WU+ZoNMP%sCo%~uK)Mvx}U4fU}T){>+dATMs
z#@=zOez(LGA9UQXA1m^RAukP35agAAh<#AYj8Trc!|B7j2?el4*8fKyc_jO<5?ESC
z2CCpk&VZD%!bE23TCvs#@nsk?PC`#z0tv6qYqb^x8FG9Fap=Cp5=&5uB32w0e9$Qy
zgN!K~fg><AOg7nMtiGbt*%UKbFd?##Ru?tx)29!$&{TYy_@D%(ckt?vx-?-CYZbg6
z=DP5~2Oq3F<vhrtF`4uuE!ys&L4&BG58^~tDaGI{mrR`9dm_b#r`2g1+fdDTtR**2
zfTbbU<iLuyx-mp|=u=(hgIY^lJ{X40d2Bpl0Ws|#gu@3xlEBOuj~@WDi4RiBt;fU@
zPfRU)ibU<kt?U#Tx}6Ch<TV;JC^D@1V0k*!f=)wVLAd+wyWe!vO*~)=HBlsV@k|nY
z2-Gf(zji<9L91RCC~4uq_4@0t(~70MR#lwR2QRzqGBUM5-O#f*y#{}DIa-RrtqN5*
z4G?@OdE858jL9}c?76x$SU5ShB>~I>gl<W6mk&A*a+n<FlXjW&pl~t7`I0RLRb4Nz
zB{BEhb92w-33E(jO?;4cS^u!~qAmtzUvMq~_))#FBV>5|@yAmue2~}UP{wf4sq>&$
zNHncewDnF#R5_5bTZL4(BF!CKsUUubcM#B#G(~!Z4Gk3^bW2DRFz=ub_G?{^Y6gs%
zOy6kXaz#G8^CxGKMHYc4V%tI74%<PhOsxDaN3F<1dh!De(<LBXpYC2CWSSE$FcB6u
zeSjcbo}7606?l?hd9&Pd%i$4mphp#ZK>$tHu0`!uA5@1&`*O*m+AfeG8wXmCTE0-F
zNX!2xKBxet3L%D>haY}8O@|N8Ip>@~P$ruAprZ|MS*3?9`ZlV}&Q40Zqk}%kS1qu$
z*Ba-+A@D)2hGD39T$gZli~5Z|7)udR@|C_bDf9&8zx(dHLp%m~^3|@Dq$MpT-4%DL
z5Bg@i*0h9|mlEIrXn~0v6VTuRk7qAY1SObL<--p@Ea8GbQ<&ywzAe`Jpm!6z-?m-*
zhzlC00OOK5Y3D~)M2-f?V=td|)>)L&)<5?_&YEE50Y#;Cb@4k3yh8_x@xMZ>{GWXC
z$=G9$je-H42`pLl5jZp@jgK&`usZdks;oFvd~gWXqP}g=NLUVueB)XafpvT-9jPj7
zQ4ouJ6X0?J6-G2A46kd^==~Uiuw?q(fL)OQkT?UE9;b`TnsSQN61)r0>aZM*pM?06
zPuf!Lw$=x=^Kw(5L$Ot-E*%<7WQEuS(i?BQG5r-|EFx)Iv}JQVCTZ=`zbqe1u<I#>
zrvW~4i*r#`<HSk96ksq;f11fHUT{_bP8WlfavF;&fN;)06`7OJ<h-B{Nnp!W4li$G
zr2!@U=?el0f<tsU0eKig6$xyj`8?RJ4|*m;TQ<i4;^xAh<>o^5OCMAWWbg^DUBKHL
zC>0*TJf7uh9&pfFZalwKO(}_|n{K)i!>&)OHImp(L+eZihV~BIz1jz0XzHtM+chrd
zP1cz(Apj{ceApLIi*t1V$C8aj3>$C}3)JM;BegzgM?f~?j5ER(A>~h+eURKJZE?KT
zLBkx@iV!Fy{Z1MuvM6FU{^S#6h!0lgYbjw%B76{f*)qj@fgTJ^;D$bdD_cMD#1ko1
z%?iEz`D!&dM}kjLAcjDcJb0m$z)p(RS%fervRVtPu+;sO1A@Cn^CKSOGZ}eu0u!uA
zu$om)JIh@j`Q_=Wcl2bsx`tB4%O}K?(pe1Va&v(4F{;KXq~0-AVC>4DfwuAtMIQtK
zd6Ho3L^tuejNZqIm@-hDe4!xW`hm3sy1TmI$TW^?B;(M#5gFE9o;LfATI*`!gA`dh
zV8DQAH{6g1>Yj>6L-e4}`9f3^9;MPc<1jDDsTISu+~|F5hDV!}G7-G?+H1s^iV~JS
zh#2}ccL!J$Uf^uVt8#Afn5Q5JBxTMNhJ>;TQB9~_`KpKN%FM6>4W_8Ma+Sji4Up$t
ztJIo1R#vUNOYGaXFE6y2D^Y=pa;6lgAfHk$HF)C(|MFV4aX;AZJcy6;hYL94;(fdR
zj%qrfYXIf1^+Aqne2}A3$l1Stf7a?8TXCKcKxtIv0c@bkJ8)`sW9U>LRL!t(`isNl
zcqtK9Edi2vR1W?GFagPu11`zoGw*L1sF_=NBFG_0yz_yQ!**(6Dq2hJJZO>p{X|g&
zkU=q6W$^_87i6sW=9_Om*b>48B;I@Pz3s9CTR)Vb3Iw5!Rjfrm0T+XLTxL<X;)4)7
z3R#PX)2vR%AAfvV!NG(yF`X$xB~~?%bcLx>R98G9Vpr&-z@{6QmjhPnti&7|&q{;2
z^v01%f-88K`<hxi|64wY<23T5k7NaZ*8nQs=!5FWy*!lgp<)wm{4EP16$+7IwCVO|
z+nwrz8c!IZhD40|VORWA1}8y%#4*h~APqQdDVRh80s=qdlOmb4>x18X^9_4}5P*by
zg0l%lMF8b2AGUcRRq9fyTmFRkTdgl=FqqN@Lp3a|A^@DjwWF0-T*0R{l<J3T%ZFjm
zfeu>@JBL2f7hG@wGUJADX5E6f5}1;pF_1RX*=Ny37lq!nsIRHzfFkOJ*lyV@<z_@o
z$~W8aL;wPVfYTqX;~KnzfEZQ^?-VJYc(W}m94H9|LlEM@3Tfd+!h9Aa?mhxWk)a1A
z=*F0l1gy}}-fwbKhEm9hvcQ5hszUAGvY=FMOeikq`AMJ^gS|kz03shd9sV(nIiCA8
zj#^eHrb-ALnIz=tZH34;@j)8W%~?$Pg&;g(LQJ3u(gfP&CvjNv8f&bfgf`&@AAL)`
zl!Vw!Vr8zK2PuM*ui<<?2P<-{NYk<+5dyH#LU{XvL~RhdxO;`0d~A!aGWqtQu_ABG
z>|BUh#WfPt<~R?=WD1R!_@Jl!B+7lR7o9ZG<l80A<@f^8uScm4u_~0`F0mXUUyUr~
zE7u2G{}@q4PDT>AdC@GlsK_H##^DNm*Z)MGGVt0K8DeQ&%45I#`BGHTcc8ZvWuu4$
z$DVK~|F#mKfdfJGB*sr|NaHab&RFiULEGvT4)Ed1%rjUjDok@9q!rbKI9?>e2g|7y
zl9h1WZX>5nc&wxr9f>_DrRkUlQG|-zbz%(h9<P)kJcvwpF15WQXb~%3<JC{V&Nh=j
z?!nm(UB(K8H;3_ueUOL6Fdtq43u9oDz$0*P&sn$gfi1SdlYXvl;ZTZ=0YGUf!?Hl}
zsc)qTHsgE^+Pin}_`_l$t?tZ#PZjvE7w~%u^I<5ABzX0W!0N>lM+9TFX^SN{#bdh-
zVqz80b$x;x79NA~ZlqVQUW`1NacwUvDi;JC9kYako-+jY=ExjAIB?)V0JM}r21`>T
zlx3I-5X6MBogRskmeK*Og*OpMv*icO*dJNl7#pBBgNgKt+O7{Oxoti7+;bR!??_Aj
zq%l|KJI;upguimy1V<>MfzfJ>Q5W-o8^BII_0+_G4Ak*9tm1?0D`C#uiZl9sP4FRs
z;v9^4vQgVLs!)lg#9ZNkiH1{<sB}cMr6q@9M}G4}BYk5|$)S0vl%>GtG}fx*pj0I%
z8iiXFp+12kp8<5c1JY}(nIo`}Zghl7&muEqRv{6qcgg7YN-0JgW`h-RL&{4GwhSD=
z#^#7d^_@yv{tT2jC*+YKM_YX600;bzrzw8P8?BdBI^u)S({wb!9tVGCuYr)P`e=gV
z`j9QiY=u}f+!?VaDy@YP#(fZ_fpDNv@j<UX1qw8fK{YSE^iuBTP`uPDT5q19Edn%t
zZiZ}#RGK1|8KGjpkVR4Rgbb-3WQ`6UMI^L;gej|(1Qkhy)VQqD*orjpFS(Hpl}W^s
z-s|9Yt-U$I|G!GHu3CM`3=eIh5j0K68cTJS4*H<&B1xl*g-8_D${z$I%0~s*1KrKL
z&62>hiq!fb16o@$bVPv0w!#eu3>e_ImI$VawoC@CyIm}EI1Z|*$~q~FJthLMqC*{{
z8gc959J?vvj<g=cCMs2$l<IMq69H3f*Qy&iQ_JeXf#xb?{G^olZ~351;D%HkX(Ae>
zif&bTRDcgU(Nh|jP)p)kA5`o{8OVv|{%#_~<7A!}df|y0<i9HB6&5RVZz6>H?}C~1
z9Y<h@eX!$Al#~u8zS(w8!M4K(D??4YK3H-dTIn!3K5Jr)&AU79-MFab8?NCyunp&u
zm)$f^as1yX0BN8xwBSuXjR<t358{2e;jrd|K;-;hV#P+4=YPX@;B=x?8N>J|i;9v}
zrkf+sy*}syk3$33qQk;G7;s^MfFnMUIeaHV<@CO;>}YLl2=;^jL`a5QUMF&nN-$A5
zG?sa{tVL_}!Y2O>p;q_I>}&+|K^_$|CB>Oo<^MO)*)p`nUmx`U`GRbomU}nxubQ6P
z!YA{G4+J!3KGwhoJ)I`OdoZ-?R6A$cS%z9L^+DbjArd=B#3Rw<JNjBCnVAR-;|S<z
zek{+s9RhaFUxY*KPW8c#z@iWGwg-`TW20>STK{}Xx69TM`I9CQNCXmrL?97J1QLNn
zAQ4Ce5`jb@5l93QfkYq?NCXmrL?97J1QLNnAQ4Ce5`jb@5l93QfkYq?NCXmrL?97J
z1QLNnAQ4Ce5`jb@5l93QfkYq?NCXmrL?97J1QLNnAQ4Ce5`jb@5l93QfkYq?NCXmr
zL?97J1QLNnAQ4Ce5`jb@5l93QfkYq?NCXmrL?97J1QLNnAQ4Ce5`jb@5l93QfkYq?
zNCXmrL?97J1QLNnAQ4Ce5`jb@5l93QfkYq?NCXmrL?97J1QLNnAQ4Ce5`jb@5l93Q
SfkYq?NCXmrL|~Xj;C}#<-uW~D

literal 25818
zcmeI2M`%?|6vyAoYkB>pM+JLBK}E5|0vb>h4T7NJ!i|2eT)0utg{~Ag7ElNZE-XP%
z(S>}9pkT*}9g&D2*s&q@{{7~jWHPylFYkru{S0%txlTEA=FIt@|CyPX>YDn0e9kk9
z??8TP`KjW^@2lkN^VK!<JC%k{MN!(mf3*jEzQ5XhYdjvX82<ff$YN+Qw0o)d9<V&L
zJoG*a?Xkx4(DKmo(0dP99$FrHABA>5v^=ytv^@0Q1D1!Dhu%k_-487fEe|aZz4w6S
zq2;0XQE2x=%R|dU%R}!y&|V&vrBfxTL}@Bno=K6-_?Bf-Ql?5-Rhx&nOGb?<N|&aS
z=qX7hkvFrVXfRupd?qFGd|h2;wJe=PK#Xaew3(-wSDDv*MpFZ2G!!S}MTtN1Cga7)
z1ZKNz$g!F~hAcf1Eer+%1%7`)fj<%oW)ktja3~lkpq*&}BoquHfCEe`Qn4tF$cz6{
z@n|Fz3KXEj&r27>#ji!lcqkA+Bk+m_+K~?jgGeLHC}w2crhdxBsRZpxXMrDW1R8ds
zIUSFgl2mC#TdT2v+QM)cC4PUypZ-y3cpLv9ke;T-Mlm{FVj|uo+;X^XQ64CW7e-n&
z$uz!YIe!dekx1jj4ga!O9xAa&SR5%;m*;RQGWZ79CY>@>h6ya9i$$OzlZdG<zh6)S
zKs1Yh`DLIv1+fD%T0r~(nzSTiD!BAZG|`CJ4AG5JKp_^25)3FTEDRVrX(u{)WPu{4
zR1s6nB?BU-0;tKPR)>c|p@`v9Gjd(F{i>QLhB#Hs$7-C!iusD4FfexvNmHzjaDZ||
z9(;p8Xuy9|AvR^fu(BwfbX-mg%z=&c5)R^NzM898qDJ|A(MXsmAkU*p#B#)t)rmYS
znmTprjvYHroH(&(&z{+{XX9y<XEK@E+FDxr_3KBK2g#b6n%%p1A3uJ4@7}#g0$wZ@
zTfct&zs|FN|Nd30R`u`SAK_#&xq0*Eef#ze9H{>Kw`<p~%a<=deE9IOW5+gb+&FC5
zFs0F1W?p$UPYiJeli;UxJgUQjoGE!HXAD*JOVX$;R`hU~rVtQrm>AMUUTwKDWMopH
z6HCa<h!MvOk(PWhlsH1k+_`h#zJ2@d-Mb4HE<Aho?8Ao-3l}b=y`rMx?%lf|KYkoF
zO8pr&4*<S?{rda&?++h7{QC9l`}gnX&6}4<B(7Y!^7H3U#_~|OdGqG*;lqoHif-S&
z{pHJ-s;a8W%F3fhkJ9!0`SS-49#FV<@7{z76GW|Elq@9sH%|;D)(KWKQ1F=~vAi(E
zE|kFJSRs34K;>dWFcd_ai6L8@FyBm@vL<mk<YWlP6vm=~M!n*^Fx0|OYwg;#=g*&?
zKYxBYonE<eB@Tuy5Ujks{LY;_z&mo}NajKk2GYB3-8x7G-zdYrva+&Emo5P*wCvrx
z_xA1EX@|XCx^%g5;|38iX3UrsD^}2W=FFL)Lx*<n-hJ!Vtui(mRe7GhQLm!SLc1^o
zG4mMNJ`BMGOR-lTfPzRQMv|wAA;}ha*={MgRwDrgOFINZHIcKcV-HDh%9JT<)~q>n
z=#UU;*REakx-cYXo;-Q-$B!Q~X3Ss|Ec*26)77h2d-m*k@#01JHfho%+DDHb4ea;t
z--n9VuV06&)z#HUjvN8BMT-{6@=2ud`N@+flBHR)Yq43X`N*P1+}1t};T!fk99H)d
zP9#)u3`AmoV}}F^vDCy6NpUR)5!X5-c{<cp*OaV0j1n4pxSXyqz&z8iBPfHMXWF!B
z0v%k!_jPr3g9Z(v$Aw`$9zT8hG>LTj^y&1DA3vTnd+pk_UcGt=hG0ee;K74mzI^%k
z@nh&G7*3u%nVl94=ggVoOy~o;r%#_EhOo3~YuhaRP7FC)XjsTpl5b_df`nK~*mFj~
zMjXtm?i`M%O$^b%5QO=nL@x*pg00><vS0|>9EN4Ax^M(l)+Gz*M!Hm3HS&-X2@d9{
z0g=FG!GZ-NMvPzx`=OajT^Pbc7lsojPGr>L#f#B)?%cUPefmf)9XD>ABu~C_T4KM2
z&){ELTFMYC(JO7U@H;WY7U39iFF4R5E_V_aH6L9#kE+Oko-3D91!rLhX_G@0=s{_r
zF*v{_1W+qSN?IuMqEl~iOoRbU!7BjLLWr5}(ZNEOc>VI_%gdH6;|l_-SFhf%VFO3X
zH*em2`}U0-yKURHv17-QOX(FJN-l+I@R0qG_5%kF?A*EY)vH&ZKYu0=IMdv^bqk!y
zu8f81<XcX&TefVW44;=QS;Fi*1f6MH_HsKhEXZX)G`Bl34G@G&I7}Y`p_4EQ8S$&h
zL(GSqV4`J26<<I@vOJU40iwGoFhpb6(h>)J_pb=#-pJVE;^Ix4Hj&)$H+vN5afqp}
zuV-5W5FU^u@6DPu>)^qIFJ8Q0^JEmMl7xBc)Ttgldca|1c-ZAQqq6UDBm%>;XU{%*
z^hl1iGiT1^<b+O6IVfMcbg3L@AffBmBOQt%tHfc^a8r_0KF*}@J?&iNsB$smph>uM
zN#iaLZa5pCA!F`8M1kU-(~AJZmDin1w+lmA0osU8Bpq$sEzzbx0AcQ#43|vEm#O*f
z40P2PGLcxWh6kMA@ihJhNR-Q079xRw%Lfb?09&})VPiNPW~buhp&!nZTo?QH?Moq*
zN~x%c$A=6V!YG6pC6bIPDJd~^B7>x~fGSxU@Vj;ECM~3MoMu|pW;stDqK}0V`Yy$^
z(5@RDnT9H5d};1HT1>Rj%WDpA6+<QLTt}2PV=ifDxa^1EtY=3O(C`hxMNZCciCBeZ
zLP^3wocbrfSljaGCWdAC=8s_(WgXS^D-5&f2VU9eNC;ZYz%aka=28lENgO2nKl3sd
z=wYxnKKDXr!H^Z>b^I;)TpZAD40HG2|3R`7Vc6207DMAw+|oV2wFTcpD&F?x7Hu0F
zAhbk}{Qrk;?_`DlB@a89kAL@aEQa2tn~uWPVrVgRUlcnkwyd7qJz#lgd1!g)y$380
zEf2kq!rb?b&MIknXnAOP=)DIl4=oS9kHXHn%jLc<%R|dU%R}!yV0mbH=zSFCzHf9^
TNy|gaL(4<&Jz#m*8F=^~mCb?X

diff --git a/studio/src-tauri/windows/branding/nsis-sidebar.bmp b/studio/src-tauri/windows/branding/nsis-sidebar.bmp
index c3bd480fc8993fa88d280c35ad105a7ddf1bc7ff..0113ee5932f6c3c2fa79ba728219ed453cb3e469 100644
GIT binary patch
literal 618006
zcmeI51-u=_*~V#Ua4iIPX^Ryt?$V;gz0g7nEk!C&+QJ6}x8Q~1!KG+ih!IDVkN|-g
zB!Q5G2ywf~_n&?49VX|TdvbDi&%JxkzWck`%-NlJ=bdNXXMAQ4J#M|rR$k6sd-CsR
z{5$do%Pn`s50_i+2g_~X9xwOBrFtLzsJ$(mKiB>Y0Yjh)5STRnEB{vo$+09D0)_wx
zcv9mqKnwvxpgIsRUROsxS}F{I5CP+Lh=#chf$BiOcwHU&XsIv+LIjN0AsXg31gZl8
z<8^i9qou+S2oW$|hiI7F5U36WjMvqXkCqBUAVk1;9im}wL!dekFkV+jK3Xaafe-=X
zb%=(!4T0)Fz<6C9`Dm#y1VRLi*C86_HUz2z0poRb<fEm+5C{=4UWaIy+YqP@1dP|!
zk&l)NLm))Jcpai)ZbP6t5HMa>M?P9A41o{<<8_FJxebBpK)`ri9r<XfFa$ybjMpI=
z<~9VX0|Dc8b>yR^!Vm}%FkXjfnA;Gj4g`$X)sc^u3PT`7z<3>^VQxd9IuI~kS4Tct
zDhz=T0poRuhPe%a>OjDFT^;#osW1dW1dP`q8s;_xssjPzb#>&UrNR&h5inkdXqej&
zs15{-*VU1amI^~4M8J3*qG4`BpgIsRUROsxS}F{I5CP+Lh=#chf$BiOcwHU&XsIv+
zLIjN0AsXg31gZl8<8^i9qou+S2oW$|hiI7F5U36WjMvqXkCqBUAVk1;9im}wL!dek
zFkV+jK3Xaafe-=Xb%=(!4T0)Fz<6C9`Dm#y1VRLi*C86_HUz2z0poRb<fEm+5C{=4
zUWaIy+YqP@1dP|!k&l)NLm))Jcpai)ZbP6t5HMa>M?P9A41o{<<8_FJxebBpK)`ri
z9r<XfFa$ybjMpI=<~9VX0|Dc8b>yR^!Vm}%FkXjfnA;Gj4g`$X)sc^u3PT`7z<3>^
zVQxd9IuI~kS4TctDhz=T0poRuhPe%a>OjDFT^;#osW1dW1dP`q8s;_xssjPzb#>&U
zrNR&h5inkdXqej&s15{-*VU1amI^~4M8J3*qG4`BpgIsRUROsxS}F{I5CP+Lh=#ch
zf$BiOcwHU&XsIv+LIjN0AsXg31gZl8<8^i9qou+S2oW$|hiI7F5U36WjMvqXkCqBU
zAVk1;9im}wL!dekFkV+jK3Xaafe-=Xb%=(!4T0)Fz<6C9`Dm#y1VRLi*C86_HUz2z
z0poRb<fEm+5C{=4UWaIy+YqP@1dP|!k&l)NLm))Jcpai)ZbP6t5HMa>M?P9A41o{<
z<8_FJxebBpKp+jgo;PRK;zbJ=FJAoFryt*b+pQ;_c*0(L?z!7;yB=});s5&QmGAWL
z$1AS+b7yz%)ZzK(o_YDD7ytLti_bpu^r#WT7cZ_^FmDcdv1sA^@a=QYK0W4(QT}aC
zW)XxK-|o@<xo4h6)c;<7@tLQe`grI^ix)5A6@PPQ&%{lnNPqX8{`cN}*YAFN?jeU9
zy!&pu?XmmrzdYiIzg%`%+tzP7%_GI61F<;CH4D0Tee1brpK+3%o==S&{+T3)YShKz
zk~U=UpgZrl{nS%W*=O&)cinZD{r21UtTWHN|Gs-ZANGlRy?8PDgkcN1Bb}Y+58i(l
zX}>@Jyi-m&`P_5PxvS-!<G&gQJ|wwcpWfi+lRWgJ!TE|wxJ$_5Men^c;Mu32s!hn#
zPrdi<fcpIBs{}YaM4vgcW`OW-&OYm>Kl#b3tFHQkAN*js>b;sZYqtJ+>m7LD0XJNK
z-N!?K*IlJ#Up|0wo-{hIMUC&KPQKuRKdiXoiUl+L*T)?@YQ*Q@^OK+aB=peg$;Sm+
zWP~+4?65=l7F&wmI_t{$Q>2F-dT8kX#v883KSl_baeOuQ%Nwq5vC~dF7O_M0Y4zk2
zc!o9TKr9qK{M>h+eL~5PKJs54hM9{uJoB($-`+<beH6wJ`uFaYS6&Hj4j%XcK9R2|
za?+xoO3|Q!A5iA}@P|M266AQ~;fIjGb@E9kg|Du=_8K0Bih1EGCC)$ZccK43{Qms<
z{QH1%C^)d(f8V`ZZ@qOWBs=aER#@TGQ%<4K2DIYYmxEsXpN-3FS>qR@Mr^V9=F0Bn
zqYKN_#V<WyX{D9k9nha*ke@&L(T_v~szR9=pi#B%vB&P=+t*%wg||eJkCNg&=IEpF
zk0|iRaoer8cxkuZeACJ+ugpVTsG&&l%!QjLXsEYmo^d+FrZhke@%qq158)L)S6IOj
z_Q@w6lVoY){M0O<F!<YF|0)UGqjWo&y7be(rG!C(qR3B<Nc8fHFL?h7&V9udUD959
z;dz2{opsik{9K_f$v5420}qSv`cHrQqw^nKI4=J4MfLd)wUR!xe|Fya=Y--SQ7Gd}
zNF3fR4@uoCB(@aweAGCUULQ(0Q_87453#C{yK#AqrOuo_ZR@SJ0)Z7*SivKM4Zt0L
ziFY(_Hrs47eBzB)IA}b}%*1ON$vl)v<9O|r|M9j6ua7$FNcoqvTW`4;|6G5$>{6a{
z!C+}ZHtC7K^F-Mv-7dV~0s<13czwvh2jf#=aXj((_q;|`@|VA(qu7z7*yZqId5D{^
zdW+*XXP-?Jm;n+(V^A5@vqv}la8a0W7eq)v2Yz<!(9R`n@!~VjID?<alvp)Uitzf6
zf4ngCf6+yMLP$aXUL}$GtuCDu1^Or<$PpkumwApK{}vD7C3tzOD}0sX#zQDaY3GTL
z$osDow#5JSWig?as)OPZ$ciN=jRJvJr=NBz1vB}clYJhyWF3vn>&1&Nx%khdTI|39
z;=(8A?mZ~8;2R$5;dQMsm(A;r?b{(-8RXSiXPtFu4;=ZcU(xH|YRfJ4T*@e0>8)e)
zG~>A!;kAf=`st@3!6zy-3?Cy7Sp~7E-oNycixD1{RNhFo``OPz|Dh1vu%3WimFBfb
z-KW>vQovg^*($8`JSUgEH?MZ+**`4k$O)T+WKKi{CHRMg4gD`whvHHDw=Vml!uB*r
zZPy@Amcj+4H7>7*e>P0bU18`^{q=Fj-E;R{Pe0Y_p1WI~a`MTte}M8(SB}@SW_$<u
z^m&!tYpk)xA1}P%t<Igkoiu^+)^Q`MG#PNCfq@@XQVyPX-tTxdd**cR_3>I}|H#At
zC7`cZm8c!I-=4v)?%m#c@16c#I(59h#dYeNcuW+_8?U{J{JFDhI~S^~Uw`ct1j>Vl
zh&FS#x4JOWMfo=VtFan1s3@<aBCBTNH(yf*dVO$st{rj25syCl=zswOKKtymPd@pi
zYuB#VU3c9UTWo<J)b^Uo|KI=q_jliYw^OH1{rmU7=bn4yhkJMloNc$=_U4;!?%usS
zGC%+P^Y`C>|AiM`IPEl7a&jR}lGj~#-HshQ_UO^0TX1#h(&gosU%u+9t9IUbXYXI3
zxbL&iKE#e6LZQie>#fH_Nt2SRufF<)7hd@K>#q+QG>E7N4<6jQb?Zwmx#Y(`{xLpL
z&nHUMF490MNYAn#HZHH}cX$kyM{BRWwy#_;U|pj}4nOYLV+9m<u1fQIK)=4k^5;Ka
zgq7ARv1lPB?Yz0Ov0kx>v3jRTT}yJ$U;KhDvtAYCwX4b(Eu1=e5(Y-VFw<LHcP+2#
zl3V*fqNt$dle-LQ+ikW%0pHP9LBVjAG~ur6m6u;yLR!rN`K&0fmG(2vI2|2SYT=OD
z8*Q`^)L25Lx*rP{F1+$ehoq-~3aDqEd8W?G;`iQr4-c{s9J=)LiYu=0-f|p-+IZuQ
z@hpLxZn|lXhn$~pyzvHl%D>b;@x&8DKbuFRK*X!4&p!L?i4!M=9=v<%)T!_tQ6x&~
z(w!$m1`otr*@P`IVB_*y_Aiu3GM8Gj)Oy&V{}WulKi^re0IwXci4^GI&in7aqj-6O
z#X@Q^T3Ydnu`KlQyxwK!of$5`53dUHnvn!v-Fwg7gp%1SEiLc7gXhSH_(Y^w{IL9$
z|G47&R7ax<FFgM&{-r0OaBKJ-X)@rm&pMNbVCbc}5-FOJs~oQ-zkk0zXzMdnj`i1H
z|Ld<^@wi~Yg8B33<M7wKd2Z@~i;6oZnGZkwFwbYrax;n4#i3Z-O6jGSUg`qVH<>pt
zl<a?w7%^g%RaU`27aX54W5%LIi%^400d=1<XAaMCPn$MP6}P5ljy?8R9?qRRm$zs_
z$$%W?-L=<VtFXkyB{z4}@IPC&Y>9u5hID3v18*hLvutmT%j=CdcEuw_nJoO|lTRYa
z7-UXzuwe!qc<2kvO7ohy&_uj|Xeba7wzeHkb@}T5US4y}HACf@U?ND7{<WIu`|r0O
zp-8Wie8drl6AJlIFDKQ+DB3cWuOLeGi6<N{-F%Rr{p_cBz<y*oewrW@SZQ991hhK$
zoZp~{)R4~HY3TB5=FFK?%{XWb)9!`HAAQ7v0Y`RmYR4US3{ydiowsm`d(WOd^@_W9
zP`B!>#NzGKPd}yE40&<a)YRzh?YG~qQ1E*8Y**?kx{)JC($1&-Jbk+BI^kb@oHS{Y
zTG50$?zrQ4r9uE{8n)#*jeMRHuP+}y{`lh#9Xiks=Q$V8_3DBPF2KJB1UC)xp+g3T
zSyKsj<MMiyy167J{HZ6O023@EKc%9TX0l*J1Mr#{#p@`e-0iEeW7@p=#$}gY@{?`1
zC4;!sxKc-(-Z6ee-h%ZtrNK8}f8|R972|ZJ2~{x3fG}s4Bg^y8K0{ojT0u73bW;Wa
zG_s0J^rVA{2i~fzKuJQOd<bJjdF{%dngvs*d>i&ar7cA<%J}Ps8(h!JU(~0(5&>0Z
zV+k^bBD|J|zy0lRgr$BIPsJc4qr(RuL^J+Tr95`*SdHOqvdJdmwHOCy@k|cV^?T%z
zM-b(yP1Mpbhu5lBXaI~GHxB=jLns>LK1C;-a6(wx&>%;?BI2$LV(=u)no77Em)E`*
zCL8F{y(@6#7^G6|EN9WeU3PJU8+dL2UaJB^pASEHk0v+u>}sp6=Ch9sswH)_mWHJl
zG_&EF=u?o_h<g9McO*gSZQgv{Dc??l5x(Zl$zqr4#f!ceHInIj0xCvE5w%JOMysQ(
zat4_)o7N4^Yo%IqZAzN5iTifjZO6D1BTKKo`s%aKKI>b>yrO;n#TQ=yi<;1nJ@yz6
zBU-yj%p$y2G7;#kC29lr>eY)NTpD8i)vtcVE9z5Ge%D=HgNNES<TZU~9(oIG*Un7^
zimEg%)G<H)c-J)2L)g02R$CF|+;h*>L$wnaCF3C~>B9Y+-~2|gD_!!SvPj?@aKQdK
z8J+mgl<-;>+qX|I0I8SPi)(h>WfyP*oMd=?WP$l>Rh=?amarapfLjUUFY=BIQVuGo
zc(wWFn?)-{YuoK4p*~)B>)HkX=+A{?#~pTnm__mwUQ<@g_-?AqOrZ!k=tEZ@p`>_S
z!TLHH{PA3?Z~j_k;KToY2;Dp^xRtFq{KWwBe*5i5zxc%$U!*Q06;Mz1uimJSs|c?p
zmT`50t8%>m{`-jnk_sZm*d<fcQ`Jmuo6l>a6SWUG-~gmi`HID-pYBHgG;Zd3y=64N
z`P_5Q;ZyQy3@EtbMKL*s4IAeDFhY(fY=*lur$tdD(-V1>E?)O;P+n_gh^PwYT9<-}
ziqtPW_bd;&$P02GZ)BMqdd{8El?yNU!?Y>il8GvgqW(g?yk;7UP{`rJvHkYj70*Oh
zU)L;{K5fcI8*Qk7+_?r1{7^Agn%93?a$U-w>sgl~UemlpHz101ti5?thg@yjwx#!q
z=Kg|5zL$E~VTTdMXU9@`t;V)^?aKrKo73bz_0&_5oWpAx%!*E2V{UHBlqo1C!d`RD
zHN4H?wc6n-Pt;BbIg5WTC0~R2@-uYkP`&bZih|<#9{+8+c%5i{&CnLvB)a7QCH5T}
zt@iYWBwlO&nrUC<f=c6v+OE9S#pZ>B4?c*^AYYCi=}dIdLLW;}UN0K>;ro&x^P}fA
zX~KAd*Vhi0v=N^V^J);BHaRVGXMuI?@UrV`El*VwB?lr=^|!x0ml&~CnQQU-`l^5b
zOP-ZR-Nl-R9;#iSj@l3zGBO0Mde8y5c(G=v1&iWJkk_Z2atg{6=e6%Oi;OBLa(JzA
zGY$92+tW`!jbu#7Ct1qhci(*^N`4qa&_h8XpE+nJ{<DgCU9g+!0fdQKzltph_IvJX
zNnVK8r=997A!HLjH&}oD-=2Fe+eyZaalOLmT7><>FE6$TukrKk#BbJ#RxFUkGKSV~
zypEiL%)-A$*B^WI5n@M4nf+E<ZHa%?)G_O8za-x4?;Xt4qu2ra?@!de==<wmE-R?5
z3ITSJNKfv%1g!6nLS!|JBk0_X%CTR+erKI^R$U19zd|)IuNeam>uaX5&?JY~s;}R9
z$4!8Wvv0onCf@RS4d1<={Q4UI9_I=<s*l$ZPhI)>!2PbwWH>9QHWXR3)=@>M^5yyG
z+`KTvbd=xPjgTOU6+<{!Ny5G@+QWE=u0?pmnE8Wzkjv*z5ngMk{P4QgrXr=YiNJYt
z+;SH$L=KpRgLXJWPf|t(@3xslMFHuG$!pcW6lFOOfdEmk9LSe!irRNoa#R(s|8xI+
z=qaB#xRgdFwvYy?X?5dZUa5Wi_AOepIOdpRX!*#e?hfuB{6LX1jOLO#*UfXi{kB_}
ziZtHp;%DjY@FMj!+A2C_gzsOU5iO+8;kEkLEM`L`HP^?C8Kd-~6)JHj2{>5!=Kaw7
zkn-}tow7X#z{G!6C9ehY$}6qZH|nt0%%3X`au=z=+|MI>wBL7|i=olHDc??9Z@uO`
z$H<gT?b@{F72r#bxbH>50#EHCyq-DzJB0Yk7x^-KHHHz6Kt$=!33q$GYS+)pc<AL5
zs`Z<$iieCSLxF(LZMWUVX<IiQ{>JOCIS<jg6iI`&IC$vSrQl!JOD{%Sy`lwaN*w|j
ztiv<wTkxYHN~x;W1)_57P{NtB!@m%#l^*>!aNt0si3l)M^FH#xMDQQ}@Q2_3{`cHB
z*kA)almyK^iPxIH4rStw$!j=^5Rn=OH7`o4YrycDmY@_I5C7fo+=5}>_W0u;-Ih1-
z;ZinHCg6c0LP@^tvdi$GS4tX76LSDe{AX428W;s2VBUA{JrD+*{RM4LIR4kl3-5D_
zEjGuVp(lS(M%juG6%o;S0^hG5@oPROjQ^TVVc{<3BD@w~sI}EilZKN2&nvGaUaz4X
zsDZ_M4KW+sTGb}Xd;dTCY3;fc`RV#rXaB0!k++T%QHmPJRsa6ChW>mgc&hz4p^{AV
zeLZeWRPyC6p`qO1`GoP`&(0&3OTl~h?u}$MT%UjbdA;J!w6Bj5$8OPfJZf;wCqDV)
zlROl!HH}3Vn3-rUB{L?kSxY8O7@}u5ow#74lq71sDmZ_QdgI5B$G`d%5M1p61v&cY
zqqRm(Z@>HQJFS)%+S2?VS6rSrk#huAE3W}rB_s~IMHl?x_jlZOD~GY1_q*S!z7WE^
zRf&DY<$rfpL=71AD-S>XFkazDmbvGizxd?iA+^-2S-@Zy`w$r`UVU{pqs>y2Xc;-}
z0@s{IPQiID)rP!%DcUG4^~i4J_%_>YbK?yy*h)qRp-t;IS$rOjf#XM-5HAvjefsgq
zC!NS30Dq^Ra<bHjB)Qh${`>ED)>&t=UXcS>PCNBf&Y$5R8kI#k^)(4p>~u8P2TC)$
z<sG*(=<SO&q;av~oI7z5b$(?ni&B++!tuxB(<3j()mL9lKnBXVb1|3Ao{k3}d{CSf
z3m9Z>-n=<OOkaKV6^+|3zx<L(WISsE8lOC5Qi<n`sPD7SKE$GCfP@qu56@|bOD0oM
z6gVt7Q@5IjA~iqJ1o>P!y!PwOR3gD#)$izuGAu7=MY>C2M0r)#fBv)EuIK@Yq%Jq0
zf8QJ$CjPUEc@2ixZ4RUY){|O}s3saqfz7fKwn~vNoMMDeEpiZ{_(?x^@4ff>`OkOy
z@sHP*9PU~%^6!8BlTa|XIIqb*+5exS#vAfQtAa`*T+dNJj%%;^KW8+H7i(LJ*8zv*
zgpTlS?_NC>Fyyrl;@hHO5D$@0K8LX&gaj5eLZyq1-U#tw6$KFa`CLx~M0qOV-h1!8
z*IH{WF;1#-hd8=&o*#0^A;hBkT2s<@+;Im_KuOv%-LIwKV*SY{pUhiWsj>ncLN$np
zGDm%X3Yk6j*aOc+)`u%E*g+;eDGfwn{lu;Efu9dN@BpGbfa*puz{0D77$*3zad|D`
zkoz)#rLRlb)vZgxY=mt+`>$Pl6@l6ffdkLsrMU3JL+)HJKJQi<z)ArdwJH}MbM(=?
zLPPR}YCH1y{f&nN=k7_tsV|r+LJ(D7eT{sLP~nFvRpO9?N?sstC6Fh$y?jo4kdoXr
z4d)Z(QzU^3uD1%#b__ysd0s2XUVH6DEQG?Fmxe<Rd5$P99CzG)8^Jw91ualz+F^$s
zihqP~=aMLvmi6gVt3#y7_c!Wa2GH=yVv_Hl7IkT+n|+bqml)(Vr@i4{;m{!1QsuM2
zTO;1cQK~g==0)kP@&$#@KmU9!t??ebg93!LH!l4_kt-<P?%5r01p!R(VdL_eQV|0q
zy@Ijd(&n#si*`T&yUdj%^?(RtQ~%DNgCwJ(?zravt`@e)=OLGO2xQ(q_@4(55;iZL
zv_%W)4s(iv4^E75T$fnjCE(}pfBP!~?=;j^Z581)d58V0NzZOv$>%Qh!n-ix6_Gyl
zpi>gXB%k91d5!|Qybw7Asa{D)|Gv4rCL%oFfA2kPBv&Z>D|Qc&SK$AyzKYoEtEs&;
zir2LHRWTt@lD(XhPCDtOmtJE2cl7Ad)ZpxB{_9`=$_{sVqxJjNTW`@eW|WNKwSpI$
zJ!!G;5-mG?;)y4ia$*xVZD2KbiAd!YqFB6&C^mwz9sA&e4<-uA7Cj8*<7UL{`s=U%
z#V>v#HH3_G-gn=9X<8$M4nDde0Y5~+B@guJ7#3p<Dg)poij5}>{$GCi<yt2X0{E0z
zG7-l`fBK^=)`M0dN8|EZN%nU&7ek8qRExRWZ@<0wj6-JzdR}WokoWMvXb&>n-)5Vw
z{ln0)_g;JTe%m!r^J8=-8f}Q?*uBl8!<WK<-p#Q2`RARda|Id8LP^<nabA-sd`2l+
z{L$cnoRGs<ke3-c=(#g8z<iUG^zz>s&@c41RJY~@5aR1=SuZQZcu1dJMPk039d_7W
z!|zhqXGeVwHCH$=R-!zkLks5>M(z!spWa)(0`)x!gUe6;lQ3-YxAKXSTt3uNg9t<K
zU(EGC!?$`x1CogbigduN55T0JH!iPxz1@Qi*Y2-pcQ$=+=788yFm(XJz6q9M;<*=l
z47zsV7!cSelz98Wd+++Uob=1^tg1!0)}dWnMxxF*{j`%#Jb|Gz28etWR$tqaP)Nq?
zskl3L>;MH=d_{1^qgXr25FB)12NNE+RJNdYiM&<<#R7tl`C>FZ`pYl>JBJdSc)|%x
zFwtRu^_7<yZ;!-(-vD+h)y$tZb#l+{-N3C^&+a%j-DDHY$(tpsO=^LPfb9veQ@&MN
za(J!dJEABS0V&5j&^ut9jqR4C7ce#9m0#`^%%Ct)k3Z(t(s0qX@dSatA%$^c{QfPT
zm4v0<Bf(2}4uuqXTTrsp2o;kA?uZS8_~+h%2a*tm%}JDg;-90lS?BM@aOxSLC<192
zuNs%vGTQI&_3vy_(!Td8R~@N~yu~KP?%c3-&D!FiK|R@aesK3Bnm-D2YMFRV)c8k3
zI_8DtFRnp>P(DFmXpTxhU2>j_Q=4+7<;`=cHW9D2E+sce^qc6GiPQp9<yEH+?R7E?
zRxRU)1v!@|h#Wa@>-$;Ia}iPgRY8INZ9zvEq@V{LhJNIxGzj_5;qy_s>W<S9M3ufb
zF0V;u_($4{r&lcSL2~@p_ZDvi!PpcKN8~O%uA8rb51uE%9io!s9~~gXQrqFVp`tGl
z1oCUO&6lr$Qi<dc9ig_}b{nh<TT;dGX0$GaiSiIA+$dJCzD5FK@qmNrfI5eRD@V`L
ze<F(LCQq8cF>9JKsE-uY{hXlyz8?V;6vbp#=>W#`*|@xpseB59m&fb!5(2wvig3@_
zf1Sv=3gxL3hH})7#wgew!=V#=g<-=DHqh~XRm^4TYZkh23MI?vI(2L>1SlcvE22Zh
z5_B7{69f}cjlyfS!)f?wnt)|3R1i6}EkBFb*U91_X&7@<ao6@T_^Qi@&xZ{j^dVE>
z3^rEZAL;sd$VaTo^Ibm>$jOLG;G8ehc%6?)oF|RKYqlGQHK5KS&M~J(;kA$i6WHyL
ziB`N*yj1&DL&8aNP8P(mt7Q#qyk1tsiu-R!UVBYgdfKr=`!1b3cHz^5?c07c{wqIC
z;Qcfrue}abUf@&pe+}H^_V9)d<8_Eee0QVpI=A{;a+*Mis^#^Wwcqh!eHm4JgB?>J
zA6S#HOb8gSmkF+=dKRw_I_N-H!w?E%)Hoi0EIKl`Zk_5<g)FNFI2-;ItS`iCX7JHR
zbFw&!oR_k!I+*`dAz-{tHao=YqoNZL*H~k9Et2GeL{zDCxRQ-)nWTx=zx(ZPp)*?P
zxZ@({r7V+h^PDyWjMr()5!1&Ih>w8rIzEOb!4OCr0><mK<%sEH2*gLgcpV=@lVAv>
z4FThI+H%D7F$CfxV7!ixp-C_V(uRQXI&C>(`WOQ75inlI$Iv7g0%=3Qc%8N!F?|ey
z_y`!U<6~$N41u&EV7yLSj+j1%Kzsy@*YPnl35G!05HMb+Ek{frLm)l^#_RYPngl~2
zZ3q~z)0QKqk0B5r0poRi3{8R|kTwL2*J;ZU)5j2qkAU$yK87a25J(#W#_P1@i0NYp
z#7Dq*9UnuJU<jlQ0poSra>Vp81mYuLypE5dNiYP`hJf)pZ8>847y|JTFkZ*U&?FcF
zX+ywxowgh?eGGy42pF&9V`vf#fwUoDyiQw=m_CL;d<2Zw@i8<BhCtd7FkYuEM@%0>
zAU*=d>-ZR&1VbQg2pF%^mLsN*ArKz{<8^!tO@bkiHUx~<Y0DAQ#}J5*fblv$h9<!f
zNE-si>$K&F>0=1QN5FU;A48L12&4@G<8|6{#Pl%);v-<Zj*p>9Fa*+ufblwQIb!-4
z0`U<rUdPAKBp3o|L%?{Qwj42i41xFv7_Z}FXc7#8v>{-;PFs$cK88Si1dP}5F*FH=
zK-v&6UZ*WbOdmraJ_5$;_!yc5Lm+Jk7_ZZoBc_ib5FY{Kb$kp>f+3JL1dP{d%MsJZ
z5QvX}@j5<+CczL$8v@4bwB?BDV+h1Yz<3=WLz7?#qzwV%b=q>o^f3hDBVfFakD*C0
z1k#3p@j7ieV)_^Y@ewdy$H&kl7y@ZSz<8au95H<if%pg*uj6BA5)6T~Az-{tTaK7M
zhCqA-jMwooGzo@4+7K{ar!7ZJA44EM0><n37@7n_AZ-X3uhW(zrjH>I9|7Zad<;#3
zA&@o%jMr()5!1&Ih>w8rIzEOb!4OCr0><mK<%sEH2*gLgcpV=@lVAv>4FThI+H%D7
zF$CfxV7!ixp-C_V(uRQXI&C>(`WOQ75inlI$Iv7g0%=3Qc%8N!F?|ey_y`!U<6~$N
z41u&EV7yLSj+j1%Kzsy@*YPnl35G!05HMb+Ek{frLm)l^#_RYPngl~2Z3q~z)0QKq
zk0B5r0poRi3{8R|kTwL2*J;ZU)5j2qkAU$yK87a25J(#W#_P1@i0NYp#7Dq*9UnuJ
zU<jlQ0poSra>Vp81mYuLypE5dNiYP`hJf)pZ8>847y|JTFkZ*U&?FcFX+ywxowgh?
zeGGy42pF&9V`vf#fwUoDyiQw=m_CL;d<2Zw@i8<BhCtd7FkYuEM@%0>AU*=d>-ZR&
z1VbQg2pF%^mLsN*ArKz{<8^!tO@bkiHUx~<Y0DAQ#}J5*fblv$h9<!fNE-si>$K&F
z>0=1QN5FU;A48L12&4@G<8|6{#Pl%);v-<Zj*p>9Fa*+ufblwQIb!-40`U<rUdPAK
zBp3o|L%?{Qwj42i41xFv7_Z}FXc7#8v>{-;PFs$cK88Si1dP}5F*FH=K-v&6UZ*Wb
zOdmraJ_5$;_!yc5Lm+Jk7_ZZoBc_ib5FY{Kb$kp>f+3JL1dP{d%MsJZ5QvX}@j5<+
zCczL$8v@4bwB?BDV+h1Yz<3=WLz7?#qzwV%b=q>o^f3hDBVfFakD*C01k#3p@j7ie
zV)_^Y@ewdy$H&kl7y@ZSz<8au95H<if%pg*uj6BA5)6T~Az-{tTaK7MhCqA-jMwoo
zGzo@4+7K{ar!7ZJA44EM0><n37@7n_AZ-X3uhW(zrjH>I9|7Zad<;#3A&@o%jMr()
z5!1&Ih>w8rIzEOb!4OCr0><mK<%sEH2*gLgcpV=@lVAv>4FThI+H%D7F$CfxV7!ix
zp-C_V(uRQXI&C>(`WOQ75inlI$Iv7g0%=3Qc%8N!F?|ey_y`!U<6~$N41u&EV7yLS
zj+j1%Kzsy@*YPnl35G!05HMb+Ek{frLm)l^#_RYPngl~2Z3q~z)0QKqk0B5r0poRi
z3{8R|kTwL2*J;ZU)5j2qkAU$yK87a25J(#W#_P1@i0NYp#7Dq*9UnuJU<jlQ0poSr
za>Vp81mYuLypE5dNiYP`hJf)pZ8>847y|JTFkZ*U&?FcFX+ywxowgh?eGGy42pF&9
zV`vf#fwUoDyiQw=m_CL;d<2Zw@i8<BhCtd7FkYuEM@%0>AU*=d>-ZR&1VbQg2pF%^
zmLsN*ArKz{<8^!tO@bkiHUx~<Y0DAQ#}J5*fblv$h9<!fNE-si>$K&F>0=1QN5FU;
zA48L12&4@G<8|6{#Pl%);v-<Zj*p>9Fa*+ufblwQIb!-40`U<rUdPAKBp3o|L%?{Q
zwj42i41xFv7_Z}FXc7#8v>{-;PFs$cK88Si1dP}5F*FH=K-v&6UZ*WbOdmraJ_5$;
z_!yc5Lm+Jk7_ZZoBc_ib5FY{Kb$kp>f+3JL1dP{d%MsJZ5QvX}@j5<+CczL$8v@4b
zwB?BDV+h1Yz<3=WLz7?#qzwV%b=q>o^f3hDBVfFakD*C01k#3p@j7ieV)_^Y@ewdy
z$H&kl7y@ZSz<8au95H<if%pg*uj6BA5)6T~Az-{tTaK7MhCqA-jMwooGzo@4+7K{a
zr!7ZJA44EM0><n37@7n_AZ-X3uhW(zrjH>I9|7Zad<;#3A&@o%jMr()5!1&Ih>w8r
zIzEOb!4OCr0><mK<%sEH2*gLgcpV=@lVAv>4T0+B_567=7tWuvc+q06d2^?anbfEM
z@O$3;@Vr)i4tk*5c6W8&`1X#?Z|}IS{bdA!J36jMra#c_XBYrxfH7ba7>2WtMT;?&
z$YAECjihDhQ?0ySFmD!UkXefsEc|j}ug*jM{z%U~ZtJ+tjqO&tvE9lyw_p9%4r}o<
zBlc{`^azmYWIYCe8DI?j#4s=qj097O*gaRqQwE#T6Qufhy}*^avm!<>7&_*qr}`Xv
zQ@d5KZ?pW(?N=jn@94Dt9i7&r+_sw9n|k&f%`pIH2A$R??HC5;fstS;L&m&RvtR+n
zg9%|s>EY?JYEosqCOecbqrU0ds?WhUv|Zt*cB_zscXrw!;<7cl4S`$)FbroNoz}xt
zFc{1R<AK^366Un5gi4A3YT`AWqs29@AKhW_rQ{KRcSf~y_UfZt(=`N2K_Dvs*O%pB
zLayIkL&ePyP1EafR}HTnsTVE&X6j(NOD)<g?^{`=WT^Qu1Uv%jsbNAG66O@O$(tUv
ztAp2$)QcAnAK!rrI;y5O^x0$%L!b-<?(DRo+Itukrq$$X?JD54Nd3vU*EGrvwadsy
z3t|Z5Ab=rh<ouIyuQoYqH+5dC`8|Am$6MO3ep`nh)BMiK9s6krl#Kx9gi%pnU|>zF
zLEfZ!?dG#;T%TG+tC6~F1Ga$mA%Ia)Yhz%pukD7>n#Kru)8#eOi_AnYXvz47x<d73
zjXgI6;vfL2F)+*wW5eW{#JqOX<+ZD<7cFK!f}u-BIpSoTg*60<BY=TnW*A%48gBws
zb`#|_EB!!t<Tr0okXrM*xJ}yIL=eEtFg8pM!%P1{(<aJmmI}}!V)cH+Ya0N!g(wwA
zfO&0<4U@z0(!WxrDe_ul;6ui|$f`g3K8cjZCd3d}DgqcACWqlgt?^VZR%wd7W<MQU
zZ=ULNh&95OYKrC|K?Kwk$M7&ej4;i$c2ndvr6s%SZf>`#^{f*#S`$?Y0+<|zhxuWI
z#_MDeVDV4qA(ykaj%_%l<c0Y$1d>7klf&@XuZ<BVYt_k0Yl`)CG-tg#tAK3lWm4v9
za!N!1lf&>H>A5>bm@E^MmDUt_&Dk|})JchUXx=J;0OmJlQtxD$n5?v>$ZL+J;shHV
zc2S93Gns}!5d<(i%nu_>mWjzqYl^()#5oS9vO%*Vwr8&@hX96$`C){~GBH_cO_A5E
z4Y1C2<?PlZ7e_!{>?eC4m@E^MmDUt_{hw~zvBky?9xQHt_O@aOG{^if!ep74thA=c
z>$^H{V&i5NGunzT$gBd*F+YqjStcectts-_YU(PWMpI7$t?^`;n5?v>$!n`q41vZ+
zAXz3RE6sS_)GH3tzp)TVmWjzqGhR2ARam@D9f4$-n5;D8byK%4)4#D0NS2AoN;6(J
zmQ`51O&x(`nV76J<8@QFF4MoU5J;AZ$x1U`H<nd2e7tvbY|a`;4*g)WA$Rw|8uwk%
z#vfR2qB<nY#AKxzubZe*EtA^t7e=#6>ZW!pbCS>XZGOc4#<nZ*F$ER_*bbUy0@Hg=
zmWjzqGhX{_s#XV#zNP&du=u{N+dR|nm#=?tPKP0vwI2BU=La13aL-+D?XV`V@KdeH
zuUxv4Wn!|@jMtUP_vUKuYH1t|;OGm>%W%XZM3aPwam~wC5+PK|4|Ln^vv1nYnLm|N
zKEi83&D;sohraf~Z>YYbTD$y+(2!O^q%$piN{LcMMJaWkpN`V#fI>*=!OKP_e!ygz
zn5;D8b^QFzk$_pUept<2YD3&qmmHp>y`^wG$N!Q)RRXv}Ic{jvj0Q8OY2Ds&9c0#7
z{xWojp(;_cl6mT^(W3LB#fxedEv#9zfcv6FHJ;=BN8VHWap9$?S1|Tkohd+WXxj|6
zQU10LYm2Z&7>Wn}ZfdtGQ7f6ir9J|nMeHsZU0NO{%fw`*8LyY5RZN>{4sEY*(+qQ^
z5*$4C+40i`emnE?ucp5D!IzJ{@Xm3xc=fTTf{|j$w{}>Y5&%<w<n7(t47#BGN0&U_
zdtao9;LBv{gi>5%rwo9<3v1>rhSbr;&qa$CEnK)zH0REfo}b--dXZwqOP(blqUeBw
zsK?o`|LwVJ>w$l0Gw8yHd+v63mrZYOw;In8Sg<G~AjzbxW7loJ5qEz%xi>%o7offN
z!8uWqn?(VaL79T6R_<Z4OiWgq@jAJDcNR!hMelXWtWmxy3?0M9x4Gx7Eua9c%`jdJ
zm_Mrc`?_uomq$$O99E{2XAb}5>(`#{cQ_t$0)|4CbfZjjDoyx;0#wl4A#$jUdtOrm
z!NqcZzW>Fe%XDA`<qSnZ%g!6Y<<GupM;k!$DMMz=9Y1DDzwX1X<$IJ--MvH~1X4$g
zK6uom?l~&y=hRuFfDtgG>OQ5^PL_$uN;6(BNrxpWqh)#}BOf^CDOo1<p_;ojH8`AG
zRK?(*>~jEg@Rcuh6PC}D9>cDiyI`8P^GL!HOk}XDMVlX%@-Z^#f9;#g(Yi+3!kYQ+
z9+7(3uwlRb?Qi$pci&(C`q%B-w^v-S8;7_L6Ql2)Em7<g!~y+ja=-lEDbr?uDLLHp
zbsjmfE6oNN3YR5EK>hE%)9OOeHE{ic0I1qj1Ev~3MH=%3b!(W1$uconX~yeBYyhrZ
z*Lt~jgD(<6L}Ay^b+w_pM5~&JV)GWxeB^Du0qV{_r<$hijj2;3%V(XD*ARp&s#2*|
z9XGj%f`~jv3M$Xr8``c&GamnHV<QK5uU@^@T5GN4g6q;tFLgqrirbMgYLLJ5?nw}~
zc*B6V_)!!tcr7t&bzWHGI`0w3o!Vbd`{2RuJKWrUH57}|yB4kVS@S0GN@Y%6vw;V-
zKoMN!%-f&cNXMjjdJBTk01uO8VzSbV*YOR&)m&{?eDR&*on4Abi(C^|KPOFc2Fz7g
zEy~nCRJ_<Z!%d?mb;AQj?mE2D6@{q1@qqFHk@`(XxX6c2C@Uy2o*!_GJR?w~{`~XL
z*Is*VINhvSvmgHOhs!U&{11Nc1Ad-(;)yDJqF7uMv*%6bFc1dZ5LFNyN9wjK_zG9e
zZ^crpu`457Lini?Y0`{ggrZr(otBAOHF_m)`26<UZ>0(D$l+w*rU^l<O~8W6G~&Z#
znV76J<8@5KpekW3_o=>zI6}LB5eK7cx?{(VhaP(9mRoK~yZ80iU#~OQT2pT`=#N?@
z0#A7rjca}11*XB=u3fv<ty|;ZEf-b}NmZr3Iq>`&+pb7?lCM1WPh0)-Z(JK2;zR^-
zMG88I)V{ig*KnMlXP<p`t$e4w>xqw>`YuD$k?`EUttwom!e`C<HVQ@6%>V8hGiHo*
z9yDl>o`~#9+9zMXPH{wW!APLqau=!l_wRq?kw<Q|)mD4$wO6ZFt@M_{EP~WYrwmz^
zv4!$GQ~7und6Fy>la*$?E|PJjykgQ8b7jQaS!qNHQR$XjZt)C~<ADbraDKQVZ-d7^
z_kV9J$7nDQqpP-J9XD>=nP;B4+G?vQ)XFQb{FlG{MFc^z!hSXN{elBr)Ub>~Khb+X
zQ3+n~8Fqx<aKjBq`_Yenl%tY=c>VR)<4v)s8twVn^#yCmsN7io-O&%?U)_pG`Ng;2
ze!JQO$XsEC6@K-rUkx5SIFRYopxyrO*Ou!v^l}&0edEgY4mxz`K>Hz7?DxO_J<qvF
zud3|X^QTaf6lp?~l2eKMNtTJpN;6)^$!?6CZf5(BF2)GOIN55;mMvwdP{~(|tFONL
zmtVReriC@m$R|u6f}z$G2(BN^!x1A!Y`*#C3X8RSgFo-Q^IQ~>ANsAhI}Z74G={cB
z>y*X=d>=EpkMr!x3FmX*z=04AnQC~43<@81>Cy$yQIM$Nv1s9cd+ZX8cYWU)S0n$4
z-Ur}Qm!i1orknD0UUk(~-+udTyx~XmWEwqcgtu}JBM|1zo45V;+Y=TINi4ksE`DBc
zK{PlJH3bCh@Clua*QImccwIKHeT!7<xO`8R=C?O;Ik<~c@4WL)Wap08Q)Z21UI_;T
zR8yA9X}|sU!!t(iIVfxA;cKtCfi=Fp9JNkeXPkOD<eh3#W{~j)_oK?rG4!y*4kNhu
zf+PRZt8KU4_Pg)Ci^A5n3O*hGChJVX)_%n6l}AtNfj3`neDJ{s($+`8Pqpwj*<=%X
z&7sV0pI-g(SFiA-ws4y;Vcm7tMJ1nJ9FlYY{r5X>QPCu7(;6?>geW@)V+JgpWexAs
zcwKhp!Xt)i`wYJ$GIFM|9EmXF%{JQ%L-j`PjdsNqR~#~Ai1S?Is?j;~r!A<NgM&-0
z<^TNWKX}e>ElbqTe)hAOGiT!2wN0bp@gd`0@>MqMqdP#WUbQ*CB=gZnAI(vQpZS*}
z{oCLEhI~;;a+pkKZAmn<#sV;g%bLCR!P$`xv@Tr9f&1{o4@Y3W+M@Plmt9sT(|MjX
zZ{pY~{c#9{sZ*y4F5fail-$7e(MKOeJ{hubjTr()z5aDlGVRsyY%M)T<8^twjw<Ul
z>Gb7P))!xValVnuP`BA;n<-PKV4tD`4qa+pj~_pN{q@)PR*fUP#G31_yN)0cVI4>2
zFPueFnbxVQRHlh1PXC1G)TEM0SG{@j=Ew<I6N>e2RhQiR_U(%V*K&@A0a-x)P>-Ke
zVzlhMA@c9;vKjO5cwm4{<;K%bKaG&E5X+HC)bGCgu4KA!wU@-Z9|fBeN-v#sJP0mT
z*?mW}Wa=SoLX?+%iG(VhWexAscwJr=N0s%D-W&a23=fkNkKTX(eOb6~Rm#Zqkfk8b
zbKU$GZe6M$FTC(VJm=SXk|2Srtg_0dpMFXxgvidBFk%@AE54|>KltJ?_bM8cQ!J;R
zdTJ4wlI-tNa<|=fn?K*p#JS2i>d3S7>)yAvg4B%O4;u4~q7c_8rt)iBFA#@h9&*Sb
zgsLMo)UIW>5`N{CS9+geSd~Tp```Zv8I?1x%AP)FTtw<rRM`#B*3x4%UYEgZ4Ld*D
zYfsfW+{Gj2f(x3c%=@mp?vjae_m^y&@g<pbPJs~GxhQ|qNhc9DimsL&&-Op+>Nl2q
zXOtUkS9?`8>0^&QM)0N9)cH~T;~)P()WU^s5J05_E6f@1istg*q>EQ-Uw--J1h20w
zl4PZw?f{{r7!F+$!qgZ(^AoOsyY9NHLb+lliu##;zl!Dp3e549k(F_Rl+Ln-_i4N?
zBWK|?c0PPUN9@xNAJKx9CFh%{o^<Zq8UHe78MbOYp6s#59)y)8OGv@HNXn$P<cP)J
zd^ecA5<HRFiv`RvQMn^znRx9rR}>PZ$*hmYK0jij8!)F5R}wz_@Ix`$hbp2o-GQ&Z
z`ikHx0C9LXQvgymUycBI_F0!YA_Zz3j4I@e=9Q6yi3DkQww4~F@p@Ua*j3i;SBn-`
zsK%_VtWP`bGz>eRo8Dae_3P)283&h$)1pO-f&uesB9byOG-r+*aC22y-Cj#M#5jhv
zx7%*J5@qJZq9~Y&{_L~QyoPmSaTFb{F|6X~m%sdFiBX9D;6vDu*OCL@H3;B=@+t|;
zJ@?#Gq2zzT!r81!b#q?jY$;80$x7)gYj~f=>t*G&hQV3f=&ZD^XU)hh<F^=OzHxdh
z?bF9iD~ozK)Q#n#f)okpCEyT8SZ9KS+6HR0r8<->#uaZKhca{AViP5UM)_)EVGKSZ
zwSGSL+;fB~y)hu4bPasMafq#B$Bxy4mVyZAA#HUfQ2hc%&!SQKCFbszH6P>qZ+Nzr
z9;5MkS@SoV!(HWzZ{3Ou-za5<$dfLICj1mr{`+4y$i_+xs4Y@62EN4>TX+xo0rQ{8
z&RO!odT*I}q(<$m7VeWzKABKulUgw<I2GlzX>NZR+N!P;6L~1LOXs!JUTLM31`KeU
z!4U`uK!LkQj~+x?kkmXEbjKg>*4`7lnk6&mPGDr79iWXn;Z-`z8s4YzdYO0~k$UB4
z`W=CpMGaEdE@jSo<Bd1Q2<x+RJ>PQ6Ek}+V83lB=QKLrfyz|c9ba9k;$<$9c;RK2D
z4FKqN^wCEnL>@|%R-ZRfdzW2yK~Eo%*j>C3t52Y`qRFjxi)7B2G2?&(4j@Q<(24KC
z)IZdAz3f-tKQme}-k76j4bRrnV>Dhb6R(+MVbEsam{yn;Z9;MCi6@@GGK;fGcFPR)
zzy9?v8nOTPfB&a-R>ecr#}P?Gh=_(7ZQHi}$xnW=R5j`oSn#>14HKVy^2yn=XN!ps
zKm4#f*H^<96s3nsEUW1*x#SW+xcTOrrE`6x4ntja)m11!N;P`MacvA;Htzn-(plE<
zK8@GQ!fST|Nz2X~&zSoSHb$FJv?{o=BHz68tvCNkeyHM4s82a3LBr|zEU%O(5``B?
zcl(@k&e1Yg{Fgy$#gd=o{3kj3$O9z+mAhc)K5xMc4r65bStE{~H9T8OkI{HtI<KS3
zdIcSxKx<H)O=gse?5fByQRo0ubd^mFFEeBW(!!tKPkaYrL&-$M6-O+=>o8b8lw3rn
zVr<i<O;}~O9sH-pT`pca%NpLN@w#+gN1f|c$4ri<tLmn4+2M#mVyH#TvU0EB1yPsC
zjyP&GSVRSt6udMT;5Egmn&Ql-M@{kXx0{uom5KOlc(#@vqw%`LEcUy%*bD4igLL<H
z-+gyXu>qJ>(BL!jQ1$hO8*cc`H?9X%<0jqQ;(-_5`E?_%Ju97M4e!%<T`I4m!;W5g
z{|u2@+mV>H5u;-^O*2BzILQ>wFTVI9_(!8>Zkg+Vk@uB&%3_JRyiDF2o~@<FXuK{l
zE9qRb7J@@gFeg6}&IT)NwXquEacI)y!U|*^*%VD=yV<U>Q~J5%5F6?1fu*yo;e8sf
zOXRh@asFP!YiCpphI1Yl1hGbV1C$bK5wT7{2Q_HWa{j`Z4|d<NvDTC}JX=eT(Rf`7
zueB!p(O!FLy^|W@jD)j*3X@J@BOH||ZK&h?#ZO(l3NJnM-f-|l9vXnH%M=A#;oon2
z(dZdxsz;;Zjj%wabe1)|Pvdndyw+~(&L6uCjJ^@hIfR%QL|7){tdM7H9zL>?Btd<=
zKoP1!nwKtCNz<9<zKcC%+>6m^=f>-~Vd-JKUMjB{36G8dT9Na<#A`JIk2~%-Y%)O}
zm2Tnk>A9TGJ9a*DJj;Ox9(c+rr(AT=MQj=4kfo=ddg_%|UU~D)H>vWN3&z3e$LxZ9
z@WBV!NCICt2bDE89JsW`8f)Y$LrS^w<I4A0p(m6xNZx6uo#xJ+>w80!XO3Xi8F@sT
zYN=V7tcQ?CX>%K|>&t8G1Uj>`fHh{fv|p2hfWDhO#+g%eIvHCwHri+-%uL2sAxkbO
z7G{ZJ;afiBvG?A4bLh&w_ukvBTQ@c}Yd01OFYCoN&(Whtvx=J~qmZ4`PPC8GN2k~u
zp5Tg+RY06eL!y)sI3DY@SBus^<U|*hz})N0(`28gjmx1v#_NK4&FD9u;b0C0MstW6
z9QS)YXaQnNoO+F&H?VC}hQKM$FOC*x%N56)Q0;OizjjdOu$ZO=3YUK61m`EigUlLp
zC<KH5a6GKaQKo^5H6`C4RjMv_k6|r1gzcdoJCkLcBniREzJl4B?8A^pX>%K|^YfZ^
z9bGm)_WJIq|L|?8+4CoJSf9@8!_46b*8b0b{u6$MEc0}$gd=1Sbf&dTz0B#R>JfM{
zdj9epK63e2r8@{ka4t#92}nzTLye9-_SiMoTvI7k`=McqI04RHRJ4jx=Mtz**?jek
z&n0jk6y<PqwqeD0XQz$Jp+3gz99~m3vjLD(_hl_hTD7wxk>#L+4r&-)heZcnOZGs0
z{PD-3R$?DCERV&ZsDf_9a^-WJJ*Si{n&}h4QMJ<SR`Y-zui6-nV);6=JNx+o$2Gur
ze?lIm&2796dF^WIwkve|^#5cn9Merpj}!d(W9=djp%3>jzWCzDKmIYsTv0O)8O@2i
zZ@lq_#*0uuM5`Q^mHhlW@v0v8gu>CaYgf)V(s3~=>naC}By;cG+kMxYyx~sPEQs>R
z6+h7zOL@p=5Tj$zDp+f$jmx1v#%s@ONPT15m21zvhdS8HlGOnw&4ta-*{mq3sg+bQ
z5euX4oG3!1y0B#FQ1_tG{znbbN*d59V35s!RxB)(NJWB|Bs08Eyg6vN{BqgmmYB|6
zF#W$hcj1$Wl^Z7yd6YJ{@jBqO`-0xMDej;iY(u2x4y_m**2zV>-xPh0suHG4O{CKd
zAOf9e&uG-ZhNY5tl#;s?^F<A(Go63_`C>8nRHEV*>}Q>I76@plkQHdpJoAi>dv&nX
zQTy8&!|v+5apjInN*k9$eT>)QHIpMON94QDY}j|EQCCxQY5`V?E$z7Dj{0r|Z%;q{
zbUl;-Cw2FAG+&H;{`u#fiAF8|uv*l^6yC$?n__VP{r49viI;vtiF3tuJR0YwZoT!^
zii@*ogcx@<$U6@C8zbb=d}Wd+k;-O59;MA~yjC7*McG@QM26w&M%Qs;@kVHzOe*n^
zk7Vtz!wx+6G?bs1cVA8OK`TEf)Fc=UCjJW9&8V81<;hmuY6CF7F?{%N;$>tW<aM+K
z4$gz(A-K`N!7*x5SK43|YKo_g%b`BT>)QIdlWT;J{<bH^z*w0YmVBKSQ-N`^B_^0S
z?3$V4WS4~_T$LIt#2AA@0j`3R(R!8syDyHYOf_);j0qNl%8vg1`%joKfumUw&Jz`k
zoZJXLG`F;ppRcDe?WH9v%IE~|A&=7LHeN>=&7!g&-_?1O>2usSDP$5f^L)K5L&AR8
z>jCv-NZjLJ-r8BiscJf^YYI%?q@T{AhVaT4O`NYNppt5n3pA?co;`aaS-GP-%VtJJ
z2dGjq>+FT+`X5!Hvlr6F<xn5vHN1{G*K0n|?Puy-Gti5P4jVR1OD{sR>E69NpV`8e
z5;0LPq-D#NB8BHd$dRV&_7yjibgQqvI=K1n2JW&i;l{Fdmb)Xw!zjML{PN2nM${S@
zXV$Vp9}x(7ls32V+VR@WaIN_Ad#7LzRMxV__U+qaGrp$wHp3aYp)HkdoU(ANU0p)g
zaEvC=nl6hAYVIp;K8d0MuK3gzkvYbkO)FtuNOEu@v77K>!<X?oZN*MOeG=><NmMjw
zwtVNIS6~Jd^5Qj%kVBghubE!?`OkmObN!4->J2v7KwB87rm_EmyeRvjhOkJBzg*LF
z$yiG9Q?>$OK(#~IEW<c9pv9!Nz$YX4x@ede;yYglCTdqsMibc|jeEg(U64m<dsqox
zv&Eywu<Nh~j6wA#+Z(YN-^MGO!ubJZ)qFXcramzy-u6a+B?0027c}QB_Z*IgK8wE%
zeU|Q4GN(?Rs->qfsqNcQ-r)((Ns?M-P*daT>k$(=Rp<)>Y2!txkF~?=+TkmG_Ko|v
z8@!gmAAb1Z`Mk!8LPN^u>{6eopx`_oP17VxYBYzE;u?!xUw!q}+its!VJi+pZq=$4
zZ3F*Sem?#5Q$8Gj>#esk^UD@&R(E{x!3Uo2Ii3|H&fQacx7~IFsQA>*$qUuvn{2WP
zr3ZRyoP-TKcX!#G&9;eu%q!$k+T6x#)&AuBq#2($V~FNG`4Fui3|}f$OZOmBvsaqV
zCUGT8YF|flO6nPBoUzV2>*Oc~kEqM;xZ{p-<Hm8Q0_)i{Nfdf()~p%(xY@C#Dz8G+
zX5%N5dc>P-b4WA?UOyxb$O~)cv5T`3XP~8x%b`BT>pEWlc;a-&>xEI{4a2DKqs$=-
zmdWia7fn)&IK-I9<w<?@)mMA1sJ|eNzx?x%^7Z+lO4YY7D@1K}rD?iBm8{n-jzb8u
zMun+uP?hEjYUVuDV<+Qv$oh1;lkHzK$;G;b`@3$-KI_OBT&E1fkdb&Y#<4erl>zH(
zxs|A?p*H)f6c<&qCoD;jo4XwN;URqYue?Kk)I!DsEqPB697$a&)BJKiVvi56BuG@<
z*@iAFt-ekti}STgnAdKSYo+H09E%P3NiGZnJHg14vNth>JrTs>>()eQ;=99IwN6CX
zwyu2TdCo)KLr3TnI`dNs<LK0>lVe9+w>=S^5u%dqnrp5h3dI=eP_|p5o_OL37sWz`
zO5Fatm)<?G5<_SCmYJR>)(+R#^xpXJw-`V)?BvFs*anDcCB<tQ8|MMCwjANf@>+~j
z^IK%}Y9D&&p^rTB$Xjo{)vjGTj!Rl^z4b^`4yl<f;$T(CiyuxpVwBz|Tx!!@9Wi1=
z;%fWih{F|$F)3<G4r2%?3^JPP>iO9XmG~@3dRUyViSe4rYcl=q&u(<4TQ_;l_;6Cj
z=6TeuTQ{OewgoOyYdNJjD^jzR{pqKlCc5xq0gU``mp3-4si^VZIPH)lOnur_P;jb)
zJa{sC|8X2ro*_h?$>4z^_OdVW`tN`LJBoQ;_Zfb>@j9RTX?v1vJ6yc(J^VH-!B1YF
zd+xcIRw9*%sKIQwEG-TZa>9C+`6^G;AVxKwX&6w9VBXr&$UGMwbSY^UTyOz0d5a$|
zU*;Th%rQRsDgi3g(N>I1vH9rAX0!q-T{%lrKK}aE=vbtL4>L>~FG78+z80_h4!;u%
zhX*3dQAZt>h^2`X>#x6l?S~)}T`39^(K+1vAgjLn?z?%;*pw%}TK*$Pj$|5(baB_l
zMowSEr{v(^64B6>wQ%@Ef5~KJ7jKj8K_E~XGOM`#_S+{2I7?ros+OjVoY>X+*CFfE
z=}x@9X6yZcQTJnJ+Ki8#@3YT7vTE*eF7aCD;&M6}lCd|8S@ukfaV~N2o_p>=L#V2T
z!7qOC3(3rhLO(h7&?g5fal8qhbJ!}eG~@9Jsm1G^ciwr{tXX)@5nG=A2;m}L08v{)
ztMQ4E=7iU<42J;UVv8*%PMnBtnzbDLZI4Rq!cHeUq?Ogz(c<tQy*K*d$PDIl$AB?o
zj7mmQhIJ{3Oq8W%?Ap$^e{a~>wl3_PD>8X^%MX(Cgit);)3Kq3EI8qzUgh}LPfmTI
zoZ&nc&j05LG3I+y?E0XIYIbJO>hoMMxKa@%$lYpFSXZp&2VwP{R;na|`MT?_%hqpV
zp>M%`?3DgDw_iQ+?*$mIlSLru`kD{mHv90)C!HBYr&&&(JXt%BV%nJKMm74=pZ@g3
z6HjC@CkOby_uhMKe<;YN9BYRfp6@xY^i%A^&tLxX7e3rWm94EzdR{O{PP{okeGomT
zHBRkMD!7lT+tlp+<T$*xMPd(wa3li#eKuOs7vOUP{xtHij4H#evXpgnEQlq#1q)q4
z&AH;Yc39KaktR#S#HCe&*MrACivcW*%)o!6CuW<lpjkrmO*h??)m)5CU3~Gyx8Hs{
zE4P_0fBWsX2MroDdi3Z?lP1w%$YC5m3s%9NMAUcMMM5h4L)_togix%QSf6M4oIZVe
zzkdDH0uRF$5z8wgVwjzTj2%0cCFrbMV;qkb0$m11swv_qC^p}Gb9MF$qEz-nWf_un
z5V3eY{<}f9y7l1+A90qrVJ9fi+Tm{gx=pi>#<_C>#cP&NC%cH0dS4yY;tc@mTvmAU
z*)Lk85T9l+lg`kA9US5j9Ly(I_z*W&NID<TKd1RkgG4(c^tND$!K*N?(Cy{(&_DnB
z>#yk%uqXPLzx*Y0;xu^ZfG}gu0VYNG>yb&S2=#KbSa-n!S6@$<J|sGo)OfwbH#ZY3
zMiLQ~Y{46?OKHaLdkjFl<^;3K9s7^rVa(p1OX2(4Yp>0mvJ5p+(siqIu8?B+uk?hj
zH+3arN*ivtA)B5KKKS4}@4Qo==c|v}lm@)=WkM*~2kF(T*KdCFo1Aa+<y5$&S}FDE
z1p&{&9M5rnB2!{@?67z}ar)5PI<Awf*&r<uW3G@8tFN`*Y{<BmFk11N`jJIeSeGm$
zj+vDnHnK-|i%9=H@o{+})Zm|d5^YW?c~z$<$()xCO!lI@1DbQVQ<U}s#YITu&@EEa
z>1L+bi$c|qw*2&o@_rh?L8QdOhY?Y@j@O^uZoHnK>dKn(E!l!sydL_M4#jfKHafZC
z$z}~;ip8VAP~nTrl@X!mo_h`@eFHe8TaMe;v;M$y<cMzKpr*!fcj?lFu)Ye;k(2XN
z-Z-+1h6I|2>J?8&!@LzxOq`*%@tTUeLQR6i#Wn!1^)*;oNW)lTDY1x#`1Bu&&0$>5
zPZ5yK)C3`H!Q@CLePM>V2=0r598vzKWNLO>VoTAyA^j+sR$u$3c;W=CkiZ7ub*k#?
zkQCId`l(vGNDV87vg$)~@zm6zoX{<5`@IsL!ubwG*Rf+q1ur5ql;6N^@mecs>+0)|
zt$$r1GA1atBD`+w{x!yrLBbTSf)%T0hp5fq`Ou+5p<aCray@_Nop<EFh)iNcIP-2u
zD^g!eAX8V`7p73sN}BsJ$E(`#4xdbRv5_3{dV(7_`*E^ngR}%qI&o3Hvd(rpyy3>f
z>ofHnQh56;=wCIgwprE3MAfaVgh4`%@En<X%bc>7bk-+Oyw>=e6p);d;Kd`#iMJ9z
zg&QidY~JxYT9EL~cY~wz&1{}4)k9~=7QC9j9z50!Mrk~psaWlLC{C^f%~Ev08h`!k
zU-KmzNiH6{f@pN8U~60tncG5Iak}X(i@wC@g1i=~2M!#l$!M>e!m_c2g{EvUUus+l
zf>uzJ6yrlnI>br5{(9Pn9EoG=%@S`FBr7m+UbBvL;Fwm9*U>ssMv$2OX)qI&De{2;
z`mN}|#X$$e?zvOYfK^MLeDcZo7wMv)-{16*-OWq=nwTnImXr{<U=iV;Uvb41DB~+@
z>2SaS2e6U{KP*n8aKM9}7gW2_4+R?2lzfjnQrk^%jGOwdoo|*b2P+{hX<l!@x$*CR
z@u)L{h}YO>B@QF4&!lCFe7XV!{9%TN9d=l8f+!TLLBy$o)K<+aQS~kQk{s^z#8IG#
z9QLDy-$YPNO##T~rwPbg?kt5B>m|0nBv<aa;-$I8WF`7&-^ZBAeQ$2R+8rI&OV)BK
zVcSW`wEngBbG<X_Kh6vyUenFwlepMk@#bqH#;GXSFHG#vkW1SnIiU;rkg*6s^hyK2
z1?yRn)StoV>shZkv5W7QUV14C6eP8O&iZqXVIi#IRi(^1OL$PnTsqK|kmf#?hI9x|
zJ|M2rrgop^B-^K%>YlYqvt&zMovYAq#9fgA*Pg3z#1ThejSbMAR1_^*v><jd8fxQa
ztHmz6?1By$Dt>h35JWc2XQuIAkk|e>hyVL(v`(1&Ujk1!;RLl46g8o8E(s5hQ4+Q3
zJLu~MDiCz1V=J-5SWxYpR3!UTm;jxMn!0lp-m-HQl4=Q5PEI9wopTNnU#-R7yyYir
zd!aPhAzQkjlINxf!g5v>FQNa?E%mOv^2#u54kh_1Y3y6n7feD4%vjxc<Bg)TyyXY?
z!XJP9al%Sr-D!l=XTgIn5F~_RL~~(bTn*zEChR>?LUeB7h>2bJt|DJoOxAKLXW{WB
zTRU9!b+6BF!4mwrg_mD`Id<3pyjB#nx;=er4YKaIgT8(H($<B%bdz)X$O_Bm>zuHX
z5Z-->(Wg(J`|i6BX|xbtdg&#0q~=5yhVpI&=j3ybiFm!`mRru4F#~U&*SJM(_HMHi
zjbns0lGZS!_UfH?-g(kQyk_vE5})3S&%Vk@FkWk(tM#BiI7^65GQgpaY07#lYLnuo
z|FkPn`zyZ!J?Nl=yy<!edpQZ>nd^h+IFJ^y&%Y&L!ROus4hg4GL8QV&?QqtIYj`S0
z)f{D*XZPCTsOTj;uf6tKsV!b}*lde7KcoP+`nqx@Nqou4=8pULb-R_H?SB-OuxN4Z
zck2@WE<k<8ESk_C3_I7JQ!oyW@CxG4VccVnJ&@x|8#y@Y+*8j7tj~e8@LDRPC5Iq5
z*c+l$h}4Q*%UPv-7^yVZAUV4@2`I+ZheJ}&70bg9yHl;uiVf>HUVZ<}8>}6kvHe{7
zj?cYsZ8d9NbkbdPA{K)`vRm13QdU<`jw0e)mRO)hdqRTX)}R){+6wMfl0yrtM7^AV
zN&QrHOx4Ul89FFlF=*d?_f;%8RW=W$5)E=zt@uy{5lHZ%bmJQbkpi{f@%e9$U2bvb
zo7p^9sz=3>t*<4G^EATm?tS^?mkqY$OuXi(J^wju7y(Um`5A4FLBRfft*20&CdA<<
z*l>w8uQ<4@cL_(IoMT$01K;75K(#BRwy(i2yzs(M?F1>DFG{3dZMD_3OP#Z$P;6n%
zT>I``{6=yENT>|2*Pb~2W6VI`$%Rdc97&pQdj+3}t(<$txK=*tIMU?!<Bunxcq9p$
zxBBBB|L9D&<YN}{g_rn5QQUOXO~mfY7V(HD8Xw3PML(Cf-kUV9++FzaDWYl9rlCwt
z)Ciw5e;OZEWF4uUN7?|~j1xuTyrzpy|Lc>lUyqETb_!dk?0OrnoHa_aBH~U3*P};|
z5hF&7966FRf%)is_{}VN7E|#2;fEhOo$ChA^O=>@6GeemVN)DsJZI_qKla#TEFfo>
z8FOlnKmIuDOG9ntuIODDG6JtutB4*mxlbh)n_+};Y`D^4Z9JR|(o|RXPu(D@pX$0g
z`jE6tr95lI=DpdMA|b4f8EWPXLmxax4NZ9c=}&(`;RdLzMH5BDGIG)q&bXnZ*PK=a
zC%X8bID8UHK>*Lo<UdEI+Ax<~a*1HkU^pk&RbrCMc%69Blz4ruqmXhwsrmBDFJm^|
zL~~4NsXx7aLdnw2n^5@oVTnb-!{}ta+itt9TALN^S&PxSD<TduWtArSNDZ$TX$%V=
z+~H5}t)N4x_Wr}<7xZ7>LrKsV#wmrw>)xN=YP?RIgOw1NIIkHgWxVv^p1bl5bZl7T
zEo=y)>cND{Ff#?=P}6YP6;@YzDS-rCIW{adaZJ<T!Gl#Rh19FGTU9noWyGKyY747^
zYUiXvKk=DU8qh>3`-EqtAr^t!p-CoclIz8Hj<3WbQmd~Mr*`7JW@j$r;dgb}WX7DY
zp@!^%xmSG*u5>bb>s1gog)uP8`Y!Dv(OX$^88We3wSFNrXZ53}A9JoWokRJeHpexH
zN2nx+Py13G_!kJ?Cqnk$e}BFjz>oqHXFM+t9|Cevn|6F?Kw$R#Dfhm$btMMPjMq(p
z*X!QVX}z1;t@_2cy|7`%TVxgIoO2GQ;@gMb4oh*!+H1c$v0woM-GBY-U)6$Da6G66
zUn(y!a$hs@Ng&nKjY4V>T-4@tVm@J@w4h#bCXz-j(#sX1$ATQj%o+VtBcI*k@_^^i
zTPbb@WV90{k)}bj5feLC;){E#SrRww1O?i-nevFCGkRg&hxws1x1{002Oq?ua!j!l
z8}p&gKmUAJP}GfA(<5%vrVS6-F%in~Hcwpsf-JbbH$MK(Zc2SDuMtTdQk&)Fti<+8
z)|ZML%HxeWM3%o&==s@sr6|s_e5s+GaZ%^mO<uSE=#mO`uC2aKuoRLgF)`UY_4Uo|
zR(-73-fl5zT^E}V!|7|1WmF#u;&(D)z^dlN3#O#mH_gYY*m<p9yytJegIfNw%P!La
zAOsKPNLr)Zy&BAF{rS&-W`#_?=HU}oov~1y8F2c);j%j9h}4O!Ok;{LfN0U<z4yDh
z-D+fRveqJLi5qr;0<FFda(o>OVDil27=~JLvuDpPyNiuUs|JCR-qL6vV#ZuN(S@5W
z%&)!n+I81mM?KE^6ZU2vdE}8CaKtngXa8cT8u!LOp4DEh3!T9Kv?a=xNmwo^K}p4t
z-7HW!<BT)l@1chtdel)z@jU{P&j<tG`r*{LP+f;YU2?>0)~$Ka!UCZbx0(c$Ep?r1
zhw$W?BcipVe8H!AkjsgkUYH;a6BlK?_BkHSc5yJN`=F1X#lGmGi^?`ORoNUbMz;}}
zn6WqPe1iK|1s#O-3@YfMjE4yHjP?{OPfnOLN$`?Gfl754FK?ADK3C29PoOTNY2-7A
zB1Ho0X0Bhlp=~p=)n{?Cj>O5Bpg`lb&uewDI}CBvtLHT}Y1zi+?TwlXnOHR!j_~3M
zh7B)tc@H@BB!>lI*f?&n2??ABu21jb_OX{CIaJsO_em>~3TdK+vc9Vhbjx(9zHZg3
z6;XqmcuiJTu5*pqBuK-=MH#O>uU!|r{pydry@#wqtBWU2oTyRX^6C+cQ?^Q56q74u
zybFSGd1_R`f#oBXy)Cr%nJ=dK%4fJ0Q6f=CDJUrw0)&Gj04isAt#n>1YgnPD4+^WG
zAQQ<-GPFXSYvXm|WszrM1su(9?XV`tO=ILyMd~__F}DxP>X@R{fuE&NTNJC%U1rZ9
z6`!aMxx+$qV$|CTMV%v7P%v#59)}pJRH9oYC+8(fe_A>?qBFzStWeKdhL^ZuCn(VB
z>#%UpLh)Xo-|Q@8(E`<LYKoU#n_!$WJchOPZ99q0G4clcsT!FNVWSwM{S-Tz9B1ln
z=FFL>kw}%j)Z_>RrtNg3bJ;C$4{uH*{O!+fYSCtSGBnKJWZenUFmX}F>oBkDn&Q8x
zT~Hcr^kYS|rkmqXTijt)l`5&mqPF6F<&{@F9dM{c!LZfGA4fY?>xQ_O%FVJp_*MZU
z|0p1EaFlULdDiP{yuGdwKL63T_pE3m9J5KBj0p-fUWdHKtx@sMzHtX`Fxw>}Fj<4i
z#i>&WF6-s<cB&97MC~HAtzJ>E=mcY*98m)eiyTE>Hih#8`h=f#)>$4F9OYd~#kuF6
z>+*{$YJ`uT)Pu9FD%Z1?%_K;}#6=me3-G#4v$liX7D$Xl<-b2#-&3GGBNVldJ@#1D
z+PddketA!zZEbDk?7MbaITR&Az3{>dDDO$Fgs3T^Z;p80g<+F&d*LuKUyU+QRgQ4>
z8X6*8yzV^o@`{dx8?O^Dh!B{=#AJU)$DOVDK)0XHTd>5**g0o`mSKSk;v3|Qv}1};
zyGRXNI-9%@@9+%rLiDQOCl5q(9f6o^X%oh-Rinsf96k|Jt2o+ix80OjC2&E_od5Q$
z$bn`#mR1SdPD-ZL*EyNRPi@D}S)lUNQ%@ygiU<+4x7>0|ZOg*)^Ev40P>MtL%my^R
zi?`#BJ2HU(`RAYKh*b^FVI~DX!zW(zjewj`rNq8eKd9#uIBL`=9g-#xKqJm2wfRh;
z)b=Ccu4kR}DTVoVk`xCkA;)-~pI`3e;kGNi@Q!u{)h^27!xq|&6=(F!+pw$}v-A_k
z@DAe_Lzl{w91jYmZ<HvCHP>8Ion>G7idpQ1#DnPImGMv5aVCP~$S?b+5~biM`S|0H
zvnQSC5+SuRf{mG)Wk9}fgx}D1g~}~6^9EB1J|-p8c<r-kNe2_zYx#+6k>ZwHZt)Zd
zNl~8LlZ!#}x8Hu-bBDN~L+J!pMNHL?8kjm_wM`p$NNsuJQ(Og;AY~{SP^0vyz0Nrf
zBq$Yn&N6#C_3UsFP(I=1;pRiBaL{R}L<g;jZv3=?d=hubtW0b{c#hblI9Lfe#_OES
z(od}^ts6b7?E+7jFk!<DH^i<Isj`v7n2$H|Q%^lrCxLs8lt^&Z$gC<spqD1cDyyu*
zL4pWW$Ce4ZZMNB_p#JjDQO1hBY-)!UG@?e19EoDuJEM3Lgbsn!{$s|BAtK^cW!-wv
z1$@oCqSMyiU@F1Kq+}YeeKsxWh)z)HyzzIlo!Kl}>>9~9Zn)tF%qfv(Byx0lGV+tt
zLLdSw)AXebV&n?zK0OKVQjEj;2p%dXH$>I};{;)Py1s3vq8oj4l>p1iYgN!b6cscl
z^HclN9Z(vPA0|@|<J^@x|II;F)@0UQoj2y_n~E+o^LCOH2P+}Rc<r+(#}QT8%fI!>
zzhDGw&t$Z8<>=61kV(}$vhA?qmT*oUz-XCY^JJ-y-D+U}?svZ<U=DX={MvOwDmr)W
zjGuy2bxICN(<r@<G%P?0x^pUMiVilMY6%%}aiLPOSRqqM;7bZ$ef5>g9hUMha_3->
zS=U#1jag_UmEdDiGL6?^PA%zHmA%O{RoRPb)Z%T~vZc(`S#8o6Ea37|G2p%T-kUF6
zj~h2mv~ms3CFiE~v%VvnMD2?&zL?-ePFNLFXtxg<G$>+9f$484pgY8;obmK$hd5v7
zWL_Q#QXYXML}Gn%l5z;5>#x5aWg<14w%^YDoUyXoI#z18^^&&X`s_(@uo7~N*ZKM7
zpQxAJX3&M04c#}HF<+eLkTFa&@oLDMJ#;{LNN{#tu@hXKWf`+)cX3{OJ?TNfZvWHA
z<sC{6AFr2u?gcL~+P`Ag2JH&X6i5_mR}LAkKP!6~E<g?RQDxo6^{ju?Fg<HoNG14~
zluYBbXK_6ax5I1vcdi!0g4CJi%g)}wJTEHY$RRib-K@RTH$KpZ_JgzPsQTudF%$@k
z0<TYrm8QbpefQl+Rx)vfSIB50<)AhyXm!|$O9fGhae3X!a7fQbAAJ<vA_QDz%{Q2$
zZgr)Gz6;t$QXH&=9OHGtEb|YwMn#_t(1<JMjUg3WwN-|dmKX%Sc+S$azTkoj*gY-M
z>8voVU{|ZVpv<M#p^{Izfmq)dP!!xf(-~M}xv8JDA}#_hRd_*_%6KRx{YN~>3yqk)
z^zO+wwq3CS=B&M8;HU&2lagt?4s)uYo9zRfRXb`@H)rs5J3**4HBH<AHI%pikljr2
z(C_xVyXKl}@K%J^UPyfoadW%vwxgqgSQTG#cv3SBtvw?MX|M{~E9{}>v@{e$tGWed
zA5`cITm|jHKTL{)m5^h+_SsaQgXL%N`l-IPUvtCcW#?B%AMlDpC2z6I<iNarq^w$3
zr<?Ow5aR3fWfCaQ`K%FlxO1^`=D-6Flut3bj3{CTQK~un6n*N78rODurr$54L&7U{
zj#+(%Q3*aKCDV9aU#^AE8I5Vt`iK2Ty86-2c`?QO^Pm45TaCk}vQ6ESi*v;}N9dpb
z{3oJR=awx~31NVaWYXRicV^6(!MWv<iTs#Rl&zcMWw=bIvAOnlG`BTq%rj(7Lk^dD
z!$^vQm5^h+_Ssa#5pD8X``)*<;$zC#{Gz%lo0IKjsJ^l*+iuGX=!+EZC{Iq93g)W(
zS>@fkclVRgi9}JJdcZ4$Teofv{%*J|I_7=Wyh-<W`$>g9(^JGA^r{kkOiHHlda2nb
z57Y?kCg?ciFRHS+t2$#D2Da)O#^u><oKP5if_13KyWf8MaW*msm`gEICtAhPGfr5+
ztA>PN*BIvOADq*`b6cUcB*no>$T41rIaS=v*1wzDtvq6)+rsZ#<G5K?jnOtDwc}7G
zKddP1iQI_Qv~-gKJMFZSAJ9_^oQ$a4x3`L0g5Fkwk4eciUYC-C^25M5O}=~H+G5&l
zH^=0rju*PgQ;xS~O(RB_V7-}7w}yNoQbQ=u?uLv#j7yL*hQQZ?fI6zI-B{Uzg>xS2
z@$*WbU{>5VlHy<`<QT7u=UvWQ*BWorj87^%UPoO#EiUB~HnPS>CqhVyIPNqg1Rccd
zC!To1ae0AT1g^@u)6l;+VhdbWQ3*aKCDVAF!`!9*j5^xQ`VPOt+nDU_w%cxlM~&WU
zjl<=%g;m$4_5)`%Za<w!J#2hyIz*K|mT;+-kQ4_iA;);V)a=XmfJHHeHhe;-$kOJ!
znmbyA#)tEuM%8q*8<*yS(Qiic6I)+S?!%H%_IEHzk)JgU`J@tjOiHHlx(rKDNBiEl
zwyx`FFJv+mj$?a=9d?*eJAQ#6QZrVj?|rkbRqI9PEtv5@x1YJyWfl9HUg`Fb6bCCI
z$9P?OK88M7ip&{kd@s9Z(R{~fmQ+V$XxO0^t5~gFHoFBKo54qq9u2mBtc+&S^8=0r
z_C{HS66U_#mEdDiGL6^y`BmbRJ3DR2(2cf)IA+()XtUX0`xcGbWl&oK6dbgoo#5eE
zS?8fwT-SOzU@tL)D(x*P4pu^r@j6cKQ*l7;Zl7KaskKr<cF4J2I#$J~y-d^=saIWf
zRSt>)(vZ{x$2=8{l{NGt@FM0>2|gwz(|BDZ<4SnthHjYDo;qu^tFLS3d21XtY?ux?
zG-{VfZ2`q+!us^-<6&G_<GMKvQPTX5vZIm5%0iP!ii4GqW4sP?stk9u+iO+&eQJt_
zwf1M9eYWYQn_|^Q?WIy%5UsV=TKZN(T}|ysJ#Ol|w7qZZuy&(QY~`3oCHR<>OyhM<
zW|i?XqV~#<_1dd$kG5MrA<JasXWMPJHEI{<wMf0$W}9)Ul7~(Ml3z`IkIoK9f~R2s
zJjXVY;$S7@7_W0OtGu5Kk2Cqg$<?2Ze*?=@>rvF^3-WvJxu>X|hDBV(Db~Pm0Ys$U
zb=O^W0*pWtfS%M1d=N^BHc$yZCMDB&U1BCK)f?3c>m{(jF_Zg3ZLN%wm9k;s_~Va<
zGOUW!!pH`uPdcLm3KevGg=FT;nT2ZV4XLTqNowpNDGpXbj`4b_ge~1ev|eRxO!1>}
z&wFZXUm-vFD4zIb-EDC2AY66TRe-})w*k}jV!r&gSIf>DrJ2;m>%@zr(uvmaKA|>C
z92vYFH1-+nGpe;~j|}A}C;l;sBdgAAo7}g(*Is+=S6_V<Sh%8-5?Qt97J&~Q``pd#
zR=w2?l%;4^E8mt9H|zujT78|LUooF(*cNIt<r>+nTaCsNQ`JYGeDaC9+ps%>n?!t)
zIhk$5h!NleEP$gE2>JQ`7mqMceLKs)JFOQpBPty{K^i75%6MIg+}FpK+4Zj_E10Y7
zR^4sx|Ni&C)h5TzGg5K;YMO4fKGgv%0z_jG{Ny+q*7>lZqtW+tlwuW$lQBVo#_LjY
zuq;2#sc~5M`22ulv|dHl>$}^mzSw>D-NkM?n^nwi&u9|M!Q*^Y5DIgJ9qKvrr#?5}
z=(-aKQ?&S`NY*Dv!^A}yuZv_{{8x8&-tdODD>Ah`Yu>lclHG!|sL>gHos-&%UEBVn
zlfI2AvRB95CyHd=amO9vG6@HT2nyGsW<Kjb-M0s*^sN0ZCF>I>V}b&W*QMlO96v2P
zyX_%Q^*vP7#*8(X<|}TxndhE+u2@Xf=84eMc^!5!NFZzGNjq0vHSHM9)o;XIjC)5z
z-y6iix#7bmNW;WM8Lu0_9+=u@Ea<+j+YB1>^z3<)WzY1usk79<<O$!fIP%CNJ(ojL
z_=ioxkyA90D6RJ<jjE>EZs|MJJ%+X5%eG`e`bWv~?IdwBCMeK&ou6Noc*2wo+e4_h
z@9VmC|B?5I+R;);HwzZc^0;T6I(7QhuYTqE3%i@pIu=w78yh5qFK*YSv^I_3Gy0>k
z&p+6Gha02gjGDwKS-w>yNW;WM8L#v6t3pq(UMg<beO>Dx@adT;Gu<o?mnLm#okm?w
z_oQdfo~NC5+Da>}B!Y`CsU28T1?|fp)JFLpJ$ks5s0$qpCMcOanLB@4@6T^z+3n5k
zRs)a<q1^yt6DMPW0*%)VU;tqh5xY0kp&tWA-p6^lk?}LZO@DKVkHrriI+U;TY`Nu@
zVfCG!_IVDSQ4&@E@sEG_Z-r1yMO|t4f9+gA&0M~P#6iuR4hJSpvF#mZWr8$JT$J%T
z%&7)*v*ww_<TT6w(|x<4U%l)qHdxK7t3__07&`ct`Lt=%UV7=JlTSXG)mfh4o>L83
zi4&Jv4mo8DU-m>jl+=gjNeIf<F}n6#G9Heb^3KCOcWcpRc@~nTcFFfL89?G>Oi-Zl
zdYQ5^QO~YfeoOn+>0`e9?kU<W;^-X>n7fY-)Urd*+}8&ld+f0jPdssh4L0xu$03e~
z9oz72d2VEK!D3J3eEA#*bNcD0w`$dD%$PB?Dn?`7sO^-efQrtB;uR-|a#|Ei$Zl%4
zDl=J4v^Slf^$F52aZ$$W{QPRjC$vTRW9pVqS3KMQNWKm^Z=u`S#6`Uw{Vh>_^L)aD
z34DHzCOTiyYTmrLXRes&f0p5p1nvdH@_ds`HaX&mBU-d*!GPH0$+ZLfsH7qYI!kA#
zeRNHl@d;m?XR$XQQvlVb-tf*bjKs;9pg`kwPG&XAPuiu_V$6jx>BHw~`V7BK2P;7i
zm=ZP4YpZWRwkG0?9XpmG^t<l53%ZGnexFo9ZWehM(i!&}Ypk*P=9{xE`j}&mx%~3W
zpM3I3rmI3Oqk!n_D`=_2ed??)dVhXvtG)-5D2goTTpbPX9NS2chKY+ZUgu<1BmRu+
z-<=IYCzST;Yag8b<>cO;ERIKu7BH$THbouS@a5vRZQK6zr$2?cu#)z35enzO5~;7e
z@=Au{DJTXE81U(*pH7-IiMW=C!<93x{lLX5r%bd1%ljEUsmE&{{08MIpjaLsZQn@s
zo~$_8oj4g26llDTle3iw8x5Fmz-CWQUxDMVe|YYPUp_J6yN^6~aHzQp#XQGVk58XI
zosY?iw#&5UT;;%yJMNfMBS&$G$O{%Oq*S4$&A){V@rC-#nEMTzD<AK@FI;A<8|4`~
zh+5rEW2JX_`IsOL6BlK?E-z0jABs^m*EokJ44X5C{8ZmVdJMbni*MidE7l@rMJ$Ud
zO}}*QjW^!V0p<|AbarcGY@2Pi89#nJl55w7Mc*fizEiXWk@3Jc<-hp0SC^0f`DCAi
z?(VWFdNM3l*W|8dI{U0ooQw$yG+z5`N~Z&MYw-qEA&pzcpziLndFP>*PnbS*{=!+l
z(xe8D+Q_bM<sq|U*Is+=I_s>Xx*N`x(lS@yoZ+~6^W1^SY6I}UVwp348e@8M7koEn
za=%yKKlAP`o1qM44vH~n$AY@3f?hYBFjhjV1ZkMKDC2b{a=a2U)%)hoiZv?7TRW`z
zP>-FS>U$Vp!GLI@9vyHTqBag@p#Jr*f33IPdSbUGt-|p!JrD2Anl<Am17!qv9q2`_
znw~s!_>1qJz=yFO?6DIQ?hu)EHPLw{42V^Kd%M2eOq`4f3N&8Vmusng9(BMOjA~Br
zJF3K8joD$yWvae4aZFX~+0A!_*bTb<_S<{@dV=S3Tu)ePLMT^6)I_I^P40c|o6B*6
zH!<GQeoe}j=!Bw7G`5SFL4q_)T$J&;NXDi76$JN{I8F2iyYEQv8!~c<-5OQX6V60^
z{q@&RIpq}2-uDE@K@(lAbjZ%7`g;BK*Jm)CfZ{bL^G9^1zrP-lrW}?AQ!jBcCMeK&
z-C$-=(THddvy%e$QE|_h>ju8K9J}+@t?{XXM<0Fkq?1nCXrqlnf`{&h9(su2ZZazx
zk>fk^(KK|`&RCU~H3`x%aZ$$W^71r=p<MI0V{_)d?|o~lKErRHJ%36>Zic_K49vC6
zsloM-`^1S8yLRn*>#et*dFGiMrO4d(F1ze<_St9i(EHH|PMp#fO{q8Gy;qgUuf)li
zpg`kwC33tGWm0i7q6H1_@49WT&u?bP%ab&$zC(iJfr)Qst{Fv_XT4QrJ#+2^L@`)r
zE%2OdO^}9(i!xs4WL6dYgn$vdImTgm+^@cWR=BAoI<AD3=WeW<8l1}$ABP`sqY&Wd
zf||Juxik8m@}s;PCD#%sV}b&W*NtKl4G^i=%}P$X#tfLVtYzSsR?}vG8L-%q82;AO
zEZ}=s`okSA4So5&Q>+DEUuGpp!^A}yuj|XTYIz>5TkEtQi@~VATXx>~>3&BH8TTTq
z#R7uoI;qnqPXGAjcTZ*yDOG>9K#}m8nZ(JMpg`kwMvPW7Vk`{1vF*y#;P-Udyj9;r
zS`WIQ`={5i_v+oz55E7!qwkIWZ|~u^wH<uXW4-pFL#@NdGJ;PT$xM)jiHmBQytXOS
zva24Zyjk$cX)SBhIA>cl&1}|!27ml)#>r!>d$B%sQL;BroQw$yY>K?TtMex87%GZ+
z*=C=efLhby7N6(+Wp1VSvf2{nhY==7!^A~3MPC1>+jg`PsZf?RJIuc!kOTtFF+Yqj
zaWW<-uqpDoRiA@6HAvsKOCp=HQ%74$6F>mN!~CA?ePDt#Ok7k`<n@~$o@YZkR-iQ&
z0*pXnei&imWK2+CQ{;93;rHCw&Q?k!Q1vtxHJWt13@2iK7-51mOk7k`<n@?IeQeyU
zNmHY#E06hMgo%?eL4i$?*YoC1f28N`*2Qk>>}ZPmt8<O{VT1|NFmX{$k=Lw_>OACf
zR!`YdsHRAbCap6jhv8v<7-8aMOi*Bx<n^M(Ury}B;aZGfG-)F-<(nJ=m>h<O`Elx7
zf;3EAR8!>j!ufL+Em-(epF>zNWz=qR)M&a2V{C2*O`k(BKa4POGA1anDe`*$yqT1i
zL&m(w@qN}5Z@M&Ta++gom>h<O`C)_!(lBvRO_A3Ep=QAX7N|s3ww=#ob6yokfaxua
z4U@z05~pF}0-GqW7tEVQnK|m4uGSQ9a<yI)RalMik>7O1@G!r`nV77=Cdz9-m>*HQ
z!{AG!jsF`qQEM@^n*;)UaSSu-F!++F1)g?N!;YrQ>jm@Y&?1^Q_d6yNskN)}i;Ydf
z=Br*wObY{x=B&TN<S@Ks>()w0Yr4Fyt+Hzte=}t;$h(8}Y}%`1#k-kR&*Gc}VpNzG
z28NmOu|Zmim7rnbGMhB7BWk;`@!{h;-qL>c=qph+qixhq3;{@uQDIsbSkwYf^%!_q
zI5c%$*HU}&;!noC#>!F^tvN-m{WJv1KtPQ^j0)3Yby><Mwn9?lt^!_rYE!7HE5s;A
z8Tn{I3;~Y-CZs_dS5r4NQdbSHYpK1+^|cw_VCa$-lpW9It<)T)BY@dpJeUxM<oeX9
zUxI@1=l@kleXVr4b!at<*=WP$7lp0LTk@#%JT;$LMgUV$dk@ROgfOINpsdOCsY7;G
z6R$}Z!<EcLP{y)krB$DUS@lOAMftUXotu<p&K5~KhQR_+j097`U@#ku2NS}On#$;R
z$Y{B%jMpTJ=|wGBS+rom&@nGP)#p$);IOxj&rOoKYGp+nw_^_z8#gY(npD1OjsZYq
z_0%v7%!BVHU@904W)ro&XElj=>U>65E3cI<t@I;<*m}dRIyTHc(sK_!YQqUO98RT?
zb31e;m87d|X+x$nX@UV@1{eb-fni`C7zw6=!MLiSiA+-0M{1r|Uws`WHkzB5LuPT&
zT^ci~Z~x)<a^f7T_aErC-CdnGw!?}uM(yTgI$4haU<MciCgJ+gT#nJS^qOREI<$GY
ztDD#P>9r?@Ko$@%US}bfEsa@2z<8at9Jlml0RiK67IN9rm^B29*ICPPOK%nsFkWXN
zmo1H1L%?{QwH&wfW&r`?bry2j(wH>_jMrJqaZ7I&5HMb6A(t(USwp~howXdd^kxA8
z<8>Bt+0vLb1dP{N%W+F@77#FAXCap@jafs$c%8KzxAbNK0poQRa@o?DH3W>;S<7)t
zZx#?RUS}bfEsa@2z<8at9Jlml0RiK67IN9rm^B29*ICPPOK%nsFkWXNmo1H1L%?{Q
zwH&wfW&r`?bry2j(wH>_jMrJqaZ7I&5HMb6A(t(USwp~howXdd^kxA8<8>Bt+0vLb
z1dP{N%W+F@77#FAXCap@jafs$c%8KzxAbNK0poQRa@o?DH3W>;S<7)tZx#?RUS}bf
zEsa@2z<8at9Jlml0RiK67IN9rm^B29*ICPPOK%nsFkWXNmo1H1L%?{QwH&wfW&r`?
zbry2j(wH>_jMrJqaZ7I&5HMb6A(t(USwp~howXdd^kxA8<8>Bt+0vLb1dP{N%W+F@
z77#FAXCap@jafs$c%8KzxAbNK0poQRa@o?DH3W>;S<7)tZx#?RUS}bfEsa@2z<8at
z9Jlml0RiK67IN9rm^B29*ICPPOK%nsFkWXNmo1H1L%?{QwH&wfW&r`?bry2j(wH>_
zjMrJqaZ7I&5HMb6A(t(USwp~howXdd^kxA8<8>Bt+0vLb1dP{N%W+F@77#FAXCap@
zjafs$c%8KzxAbNK0poQRa@o?DH3W>;S<7)tZx#?RUS}bfEsa@2z<8at9Jlml0RiK6
z7IN9rm^B29*ICPPOK%nsFkWXNmo1H1L%?{QwH&wfW&r`?bry2j(wH>_jMrJqaZ7I&
z5HMb6A(t(USwp~howXdd^kxA8<8>Bt+0vLb1dP{N%W+F@77#FAXCap@jafs$c%8Kz
zxAbNK0poQRa@o?DH3W>;S<7)tZx#?RUS}bfEsa@2z<8at9Jlml0RiK67IN9rm^B29
z*ICPPOK%nsFkWXNmo1H1L%?{QwH&wfW&r`?bry2j(wH>_jMrJqaZ7I&5HMb6A(t(U
zSwp~howXdd^kxA8<8>Bt+0vLb1dP{N%W+F@77#FAXCap@jafs$c%8KzxAbNK0poQR
ga@o?DH3W>;S<7)tZx#?RUS}bfEsa@2AmsJ`0}kIU^8f$<

literal 154542
zcmeI52YeL8+sBb6N=q;FUP3QHs?w!O5k&q178GoN1w;WsdJT{OAtBUMdQa~VAQb5e
zQk32ivCss3zq@m{+1y<&mxw@kC!ZzE?al6Oe!qF@JTup_^V+v^=Td$e@UII0g8A3@
zxm>xP%jHHZSFW!O@2f8izdT%v%0nU`5s(N-1SA3y0f~S_Kq4R!kO)WwBmxoviGV~v
zA|Mfv2uK7Z0ulj<fJ8tdAQ6xVNCYGT5&?;TL_i`S5s(N-1SA3y0f~S_Kq4R!kO)Ww
zBmxoviGV~vA|Mfv2uK7Z0ulj<fJ8tdAQ6xVNCYGT5&?;TL_i`S5s(P{e?y>>bCGh6
zg)2A}uHvHX!BW|!NM+|DqFq(IbSXkZ4-4<oho0Jb8Wq+1ly@x5TUuu;ci<EG+=@<x
zwVP;HiH5kRcre3#j1Q*uG=CE9E$t@Up^QVJ()I-(&%tK_g~?+)@t2V0T&TeFdGqAS
z{aoJM&lP$;Z%Mm?MGEA5K5uT9LIrp{h<h<#-U4~^l(H|Ve_Hn<FK`w887JJNSkV^?
zKc6pup4{ArgZ{3C3vk}U&901np#u5xypS(XImbfc4zN?K$O|v#&&wy$IbWXKMGNEu
zAN~Dlmvt<}lW<SE@i2T)q5OII(Beg3)b}i5_aY|?=FbbE`tzb))}c_*0{Quf+|RMo
zkb2P24)ZfwH{LSoX5Kl2Gr6C8E`Q$Kjs>5$_(<jt`54Y1DGo0_uismVHsJH9G#p;c
z4^v<zUX}o1H#_aA7$NilLk2LkikonqpijKykEK6@hnLSi_ng9}n;k^+p~?rSKU#we
z^(2;eDhz|hUb&nKK9Bh4&8=(~<y!cKv?p=1W7Khy$7Wb+k0%-o1D;Uaga;^4pg^8H
zY4_xOUiF8!L?7`xckbLKJ@e!d9TAC4F3*=Q-}BF>L8>3&CeQYmU&3P^_0I%Dr-B85
z8b}46{M9g`aP4U5W07lw4snA-c%eTUl97yBMjp9?G7_8$m$EAelnfZURdcg*E38n%
z4|0uk^De_iFv4%H6(H~=^x^63U&zmBfqHHwx^V|{t{EyOT=Un7{&3f>UAtw=mL(-6
ztyr<5XV0E6!p-gM?0kKFIa9G>Mdg8b$lSR*ckUb-8k(4xxN_x6dcyIGFTUvI<+WM;
zS+i!%=+UDqRjNei!i5XZnKNhAs#R60DBF}3D^~35ufL9nh=_}en>A}z?b@}KJF0)?
zlh@})*ufJGte~fgExH=qct{($<_|$X2xSG8+(bw&cTi+4J=bV;cuhk<_n~y=yif()
z5R0I6%a=zO0&dGk*v>U~Ky`BmetPxl^~WE7{Q2jfJ9qB9eEIUNTetf6@6UNxSJ&gm
zkKew1yFmkGUpamt`o@hL4<0-?ckbNZfB*g0Uw`%P-MdhsLVNe_{pX*5xRxIphYlU8
zTeq&gz5S0r{&?@+y+(~1xw*N;#>Vp2)vH&}o;^$B<jIrGnl%%*buOq_!zX(P*+Dk-
z15cyFcCN)=!A;PaT!rEiN)a0h??|PPYXl7~AuNJnbQuwV8leghb}r#1*8&tLX^Z=3
ziEFJhqB?o<<kZyEzJ2=^En4*3Z@+~Iqz{30adA0%^eDr?)6<jBq9+o~yEA6YKuD2a
z+DNmrv-9rVyP+O2FITSI(xpo|k2Y|3cRzUWAS0(?!-n5{^9?6cQc`Nws#U5~DSv-|
zaqSB)C`t@n`DBUa(-9)|N%EkFwTH<b*XV!#BJL)BAQZtwAtUK}u2J#~0}dJ<u4&d^
zdl;mi2l)3~m$o+yDxL>r+qP}ni4!M=hldNn3<?UuMw@Vru1-!)e)#ZVr%s)?3aPwv
z=gz);`^uClvuoEb<hMnO7My?SrI#2Ur%s(hv=1COfb6$y*)l3B3UWXA-~;jJqrlM*
z7cX8E)`7oQgZD`_{tefTs#7QIHJl2y55>qM_7~HPcq)HadaiNBh0Nnn36mlIcE}Y<
zTd0C8agB=(!iv34^TX}Lpc2TpZ{J=}jBLVDaB#4PhX-$&a9yxq!L3`jqQE<L?8v)K
zn>Izg@87?_Y}vAcYdGb6wQAL_U%!6g!Ue=xaNVX&8$1EHe&dZd#3w@WrAwFSMrU#6
zNl~nQZkD*Vdm+ExbP6XKbq%MuZTv-yUNkc_A<G=-3fg+E#Q@+EI`dmG@kV=%@)x?D
zal%P1B)p;<I^6iQY4(~?p(rBtS{kT*dJ^O4wUaWaR0n{!@T?#LmkGgy$M@cQuU@@+
zoX1}4b+`%F$h8UA&6_vps!u-oguCt7v7>zX@<N9<Y0^X}ZGH={4}Sptz&LSoa^e!6
z#Q0Y{F{_;=u6bsJ4k6+Yd~t?5fVn|4=gI|BfDEUi4{n}bhYLTP@xZ9yhno;Rq5Qdt
zPz(HrFI}vNat4h-AI{)ZaG;YlKU`ePXYnb5XwEPcm7!%k7?#>Y3U>DF*)wd|Fn+_-
zgb5R7&YX!0ee>qc`}gmo+ZQfe*tl_HbU5z{xfVJc$wjWQ*PLIwc5Pr_;IF^_diU;K
z1`NL0;lqa+4`_9+MRlO(@!tLX{Ai;eh71|PgYm=Q)E?|fHH=)RlL*CLtAVY(h_8*_
z6oMn3T>Fh}Rb<n(D55PqE8appwQyCrgBo88-wdu97*$-0!kv&}?N&4(m$AXEG>-J|
z!Ysa@@WXi$F<P_^oY4|B8ki#PrjLVzgO867N)=wPp|FfIcI?<O%r3<71LNr1?%lg@
z+_>@DwQCr6u0lnlsJCp{Qo3|$bOL?&!3yJ(W0!I5;CkD(ZRgLQ7jAr)E?w{dxhLKl
zHy=85sBp{>YSlSO8;9BxKF$Vc0IWed8zlm?Im3A+jCHlc2ogLa(YRRTOyC#55jXni
z<5~14bclE9K&lNw6$7fX*9zCFKZXFJv>EO2OJ{}_c^eT~GZOIRxm;*qej~XYXoHGQ
z2Cflj@#I{~9mu)i;hW<o{tz)9pr&%=%19rfG5L`E`SW8*@u0LH{Aa@L3Kc5QC{m;d
zSE15t)~v}@bml72ldBvZ9rgF54?V>h*vC3BNQxIPF3vC}0Zp%P%%Ah|oZ%S&ghT%@
zCMQi^BK62J%<3=Eo>shz0CG>mNlRRdx3~j6P41?AoSK1WbO9IRL3=`aLM|s6gqQ~F
zPa-~oJ`6cwAO!f2^Ecy=F=-Jf4Y6mckuld+{!H-ExYn@JbFDaLMnz=3!Q}+?r~(ub
z8sPC9<oaNl2e$rf&)U1$<r;SI0T5wxuG8hiZDU-s9o+3P914jxkH%+*@oaI;pRX~Z
zL6JSV)-IQg$AADFA*38w=YR8Y7+=I35743~_HXX|^k1uueA$o)NCYGT5&?;TL_i`S
z5s(N-1SA3y0f~S_Kq4R!kO)WwBmxoviGV~vA|Mfv2uK7Z0ulj<fJ8tdAQ6xVNCYGT
z5&?;TL_i`S5s(N-1SA3y0f~S_Kq4R!c*+o9uL*fb1SA3y0f~S_Kq4R!kO)WwBmxov
ziGV~vA|Mfv2uK7Z0ulj<fJ8tdAQ6xVNCYGT5&?;TL_i`S5s(N-1SA3y0f~S_Kq4R!
zkO)WwBmxoviGV~vA|Mfv2uK7Z0ulj<fJ8tdAQ6xVNCYGT5&?;TL_i`S5s(N-1SA3y
z0f~S_Kq4R!kO)WwBmxoviGV~vA|Mfv2uK7Z0ulj<fJ8tdAQ6xVNCYGT5&?;TL_i`S
z5s(N-1SA3y0f~S_Kq4R!kO)WwBmxoviGV~vA|Mfv2uK7Z0ulj<fJ8tdAQ6xVNCYGT
z5&?;TL_i`S5s(N-1SA3y0f~S_Kq4R!kO)WwBmxoviGV~vA|Mfv2uK7Z0uq57j6fAP
zJ9$V1BmxoviGV~vA|Mfv2uK7Z0ulj<fJ8tdAQ6xVNCYGT&kh2vc9mU}qgo03Mzx*0
zzfp4Vh`NJ6?9!}e#VW2vE4vh};wn>uIUE5PR9UW8+`dsA=kC2q3>jB(Qb>(C(RKWj
zUJ6<Nh2O041N(Go<WW}bCK=C}bi=ZXa@48l(7uEF$D_(k3iFs7SIaNCj&dZ`^-HR=
zB&O^7)qWdREcTl{ZopfwHS#E<$hfOyJm)YjSa!AZtmfFGPqA<2mY)?-)i<fSZ&D54
z#G1Z|wdQjq)t;YJ$3L;b^7Y?rS``?vHh9Cz#q(#4eXm!?29-*xbQgVg?ANn8W2Gu>
z&*~0uzUMxAaXIgp%5xK|%uT2|H^F0WLbZ7b)#oK@j6-*=#nD~Xt(p(lQ5(V{*N3iO
zv3Sn3(Qm*0s%NDVlJTc^w5xWxZe@p_{ai;aEIlixyiY;}pSX%Xah2x8SDq7JWp2EY
z@x<!BM#lXT>W8i$hG<7`ToD(sI%e}q7+(__Fmuw^eY&=;U$MCAwn-U(YP7oWzDn4&
z>)<rZySP_Wnc4AWXUCSC9ar8vu7bBpw~tEq9F1;8z+oJfUSnZ&yEQ9)BG!h)N3KbZ
z-IN%;K6=y2&1-^I1o}=JJG57aSL&8?myY^VA>*pjeyO%Y|F2vpuP-qxuGGw!(z9YY
z%FK>c828SE@fyAfwFB3Gi1v=!xI8&_<M!m}l*EXHsCAL+!#EkdaQ1|egL<@UQm3q&
zWc*1ouEfiB9lJP<Sm5pzU1DZT39p!vGnGRyUd9UJI+b3{FQ!S@iU}K61;%b(1<|{=
zC+^x7zcoIbmm60Ht_t>FGJo2buLgE&)3A0KSIPL}X53(%?ce^)W%9aWGh>T;MRSx8
zhf24eaRj{V>^OsrTQuSUo4TzGnY(FCP<-Ut9b01d?M~UhJ0*2%Tw?Twi1ndsLl=fF
zp0mJb;>geYcWz$4R%sW>_+w&R)!R)PI1HKQIz6hGS4{C43eno3XB+`9HA~Uz$hZ}i
zUUgo4%_XZoS{=L~d~HZV^!lAC@rU-M{%~OT-kn=h5+h^7S8rIkbj32?zy&jXXMQ()
z@Y|hUsaI1n{$EFPu69-3>|Sl}IC7DjS4^=P(b{2&@pLNPsu549;v4<avfxpxLjoe!
zg(Ssn*tIPIu8;kA;PC!kw3A{sMyv~66A}=z$Y-(dv^mp94f~`|hh}wZmU5Od{-4OW
ztDQ$NyUslv$F6Ymiglk6Ee?%t8k%%lQRx}EZI$P5Y#X#>3WP!{<YKsgJox>|V?V(7
z{@vTQB}HihzRZ8-!nu>Zr;PaG!``npu34j`lVto+GVZE)Vm<mfO<3zTBgTEYfoOBO
ztuSt?($g97_)2r)s`#z$5wyrlmF_gUPn|sa%Za1#4*N;5n{d&E1z$3MI!1ifq~V_p
z=+(AS4axYUq1)B2dP%!CKX96~-fc#-8$@e|QMw^Loq$`CalMyr(TG=^6I;zc^xdF^
zvsZ^K2wxYHXi(;RPM!Se-073Z_`&aY;Jn2}tlfl*9^@AkFniIwss6Ladrkam@cTX6
zHmqK~grk)4?31oabg%avCvS9Bh>kKI*<(BtBVJ)nRILTeK88f1t<7tL6QXd{;;`Dk
zoH%mf+^MstPW<%4{_l5gAt)p3vVO&q<;#4REl^at->h*n#t-}S-Pc=rdQ>Ye8P9sQ
zO|{Xl_j8=I9-?u)IgDJ3h6Tp;0&Yd6>*F%L5ijo(QE$QGA%P2Ktqk%*enlXLM4$ZW
z2k5?Z;S5^+@PXZ`&!z@tVGF~S%w6g?BVf*?`7_5(ANS=a?{sTfzp7+B3j$7f<!0CU
zb%*h*U8YC7s6-38(=%>?ZWEPmg>iEuPI9%p&t^~mMV|-y&s-5Wf8(lUcxBrXBZ<55
z)6Sjw`O2m97tftO@zdc0-)~2yGZ@eSsQu8zJ{a);pGkyf)5d=BQNOM&>Q(V5<{-6t
zrleCz(yFz?m{4bko)%>|{srS%ip$FRY--@Y@bhJUUdxxxLsv6DvL$Zwo*hXzX{Uca
z2HjULLpZX&fA20s7_J!cVE>Rs-okD3_4;n=cb^Y<tMe<hD@w+#&<)WI>)H?ZbJ7zn
z7@r>XZ)98_ml;E|bVl5mT#YO1v$?+CqCrdN&p<`54O>VKjZ7Nm`{3SGIELx#S1&<2
z)H49~>`2ClGg*cnz-<d&=uITPz<VN$PZ{&sfVVom4CC%n#?9=wt6lAi_Mc32oEGIg
zEt11UIgFxhWL(qdCMw;cqi&-SS5&$>fug3`;>vo5*Pg#*z!KkSWKq!4Oo7EktRdD$
zmX99ZcjlMlS1(`q_1YCg9Ov!u{@ui7h=25^6{OlQ_j<-Bj~?=V?~YAtl&|V$FAHD{
zVy^mQy$3o>33r+r>CB<htumgTZb*MbD!s(aXvI}i6RJ#qmsaRDBv;FLM^^J)_Vyy*
z$-x1$R|d_8M5s)T+k|g+Xx~o7VLN;3*K3!5y?*85`7^|2oa8NLNKyOg823?ReB!9V
z{a$ZN$&-Qc|2ot7bl)HC-KMSm*p*JxqTt#<_kYB=IyEL_x#UbH%c4sfC|2ke0<J{j
z-qBU(g>?^@KW3S~V!qd*&)0<{M6TVg3iy#jdkMfVUp#x``c-5cUmdsY(B9NtDe>tT
zSM4~S+dQu^-wprh&CbmmR4GZy_y27bD@50?W<PYU<Fsf&wDwyOa3kZHs&o^ro>A_l
zID|4gHaoWbg5;M%cXnBt(qc|rB?IFIx=V?<Q95kf!<^S5{H~D2KC6TMvCE2p$8N%c
zANYO;E;`XSD*eW<*AR09<NH#XAypXPq}p-CbMwIgfbrQ=Mt(K;ovv*g*Q;1UQD|z=
zVa0)Y*4|RJcb`EHsyJ&b>l-#Qo<0UA61NnWF*^p=tM<Hk>~h47zka#<_aA@Wx%~6h
z_>KDqc*mAkReBlaH*y_UanC9LfByLgv4uE1?U04DnfY3`d=cq2EO<&n1TOmFeLGJa
zJ%n?9^~%NHjEtk?6|KHK0XrVEc@@688lBBSMflC0Flo$}ANTLov4v;73dR3lF^*8T
zXkq`|a!q^dS=P5T0XH?`m~XPC#wkwxG56_F97gKXF|K%PYxnm1^X?T$)c^c_|JL`X
z0{jwds{*dj9k#Q}?R&q{=i$SL;o+MHzTa=s=%IlNyuy~uLBKJ}R3|1ztw*JQpPI}B
z89Myj>7QX-m2vF&>7S4P$a%&Cqk^#!zBU-g9&@nN&&zx2=#fJ|`k-I;4$bS={eQx^
zt6j~qc8c$9s=f6rYi+%Nr!lV1i5OgN9plABn<TkW!0A)mKe_&;n+FBU*REYVaNxji
zzx}3>e&TX;K=Mn9&lX*3=g)J*`J|*IH#fIjxpI{#=JxJeJr?*(rcMdUs7OR$QR&Qw
z?nsW_yK{?R{OswIS1zAd?D+L7m+;|E9zAmCdwh6`oyfjMZV213atZ00fVopAjv4ml
zp#J^)cI(jGvu^ofib{WWwK_uGp^L-B4Ng-HE}2HFm4=>iB`8y;LJV9hv_`p6ltMKr
z#VWMb36T;rq7<T$_qZb?1%v0$pYPPElcS?!%a$#pqM{I9aUOjhu%#iz#$rft{-cjR
zf@txRFJIpFZCi#e@ug}7#wkZ!7q*C?j1U{fVG|iAk$(KhK|-_37tRV_op}7rsS_A?
zOgu#q$wINO2?|^4>pgw!sILZp@oE3}`*!c}O8sY+aW$a+e3p$ut!H1*?KCAqvGOyb
z4B1hbr2-r_f4&-b{>JX37uU!9@D&RI=%AJ=6=wjPxxP(sI%UcfBZ0Y`oSf#)o%{FS
ze~Y(HU*DQ|bj-c`fAW$cQLS1vttXA=o_p@iH+!X~BvG7(K3~3cE?E?exWf3Rl}wWn
ziKDm4y&U~v9|8FJv%lcHUB7yn-?RzP5PyaQ-KdAzyf$Qs|Ln<QzZo&~(_upf41A|&
z$L5}O%DX>fGLBHc*3)6~COwInw8^euo%!t*U2;LPXGm%%TxBAr>5;A+K|5ZBV!@a8
z`}DOfOHy8eZiIT}?w*3dt5>hquU}u!V8Ma~XU?2?|Neb(;r$19MPuE%b<aQlyuPQP
zyKC33TauIFqr=CI8tONDB8+3iRmQ12Urtr()`SQ|m~=YK9zS}BJR3Q;tCudoI3YPN
zNwkt$SEBN`&1+XIn>%arxNpB2HT<)$zx?>4{%>|{(V%u&_h%&if>2km&yrB9M7!~O
z>9&?fF08wG>nz0me*66K%h6#wyRu^I{Ea;t#Sb1l(ApPn>{*=hGJ|B-&-28|&6_vB
z@WKlw3_|zRsZ;OVxuac(0v|YVph-9FrHiX;`SRrp7b;Y}YNdX?yH6YcO;EtB)xmy5
z;)G@_F^CIat=1$*kytyhXZw$bzCUsFFf3!kvE%4<EIVTU<H0>ln8rtM3|u^a#^kZ%
zM-TsQ<j_&W2Y*b)t$Bkw6szZOJC0Dde@(U2O6J6t!x<F~(Sh6BT>ayyMu%wJynXKE
zmAHrhC^i>gYxeBfZ@&4akB`rve=2&IpVQY<mhb9*{m)~xtuTlP4<0=D!i5X!MgN>T
zcdkm6DjK;~8o6`lc5`uR-n8MT1NzOHGMY3go*UKXIB!&;Gbu*!wIewO-L8;+a1REZ
zx));e^Jjh`UZahe?@8UVaZT9#IbPEyeK%?B$O&V<9y8*LK?C}9GRSz&^xUX9DP{N3
z_$*jzRj6GtxTmgd5k%a&b?c{}e)|3Q-^EMu8y-iE8ih%Pmppm$3?Dw+D5nq601qE3
zDw?axlqq9{xP}cIE?Kf<-@bicfBiK#vC`B0t$g|Nxx2YEYEbvRzHbCCo`(@fu`^}1
zc}-x{hUG+FgzDRqW03THySL%RVbclL@YPTM{1e8VbX-zgWZ;qo-d<CtPaZdA{1`|d
z`|X!QKIz?|WlfLb4mr+IN2r@Mwk4)clXv`W5@?$aeE8tuJ@|s+Pd@p?-QB%Mj~<sU
zUDD|O<(FU3s~QFQ^XH#1VS?s#iEd-YjL|w|)X1ARuY-dFzccD!%lSc{esb-~rBgp2
zV}1nNy(K;z0gu_dB63quT;wXU>)VrJc5h4Cw`=>My}OSb+Q-b<@uLTi{IGjlO2qON
z^L!VK^j<L7d&#>#VLf~|c9<H}sHrDQQOxFyp6={7$)f*#WF=9SMzN;e;d)j~860fE
zS3p33AObdtm^7|KLPAUkdhx{<QR&FL=(Bh4-cqGX*}~Mn?^B{giGv4~SRUu$&Yj!$
z?%n<4_nVh5pWeJ}bYSG$%cBObjv2l#e$>XqvEf@LL~ffHxnn|j>WKBJgI8|vw=}s^
zKtl5ciH&eR7i?(|u(iR0<OY2{D5u!&r=HxjCv>%|Q^|hVyeuei7-!mR@z!SKGl33^
z{PN2$3og)^>({RrJR-kuz4ewkM@5SkWo+ELckkz)e}47VSIs;8cV{pnckkXEKYslG
z{`bG5M-6VqPnR~c@Lwz)WksQG)v#fmIPdx;*Y!`X=b!9JxB<%?3sYViHMd5svW_`O
zjiZHHwYHzIE(`e+h=%U)gM)<kwRP*(B1MXrbBz!cD^|=RanRklbLU#MYFT{uzdMO*
z7Z(>NGzsn^J_f_`_PyU$?tWu#yoYZ>jd`Yv#5KCv2m-D->Z^=52-woVdvm>(FSCX;
zhln<dINa^}4skHfU0P7rn%>qLDDY`fu6s`}67oA^#tc0>*k8^gwu1%@(x1<28#k;_
zp+c^%u4T)Xg(D9SkII!R<Hb2UJDUe+Hr^by^UXKk2v?BtnUIhm{)8PTy;(0zc23<B
zz~Z(5V&4zJxS(4R@V7rIms8{#b3<7DrB9X=xHCkv*!JwNsQ^W~!$eV&tp4)LFZ=iJ
zPk?MA1zI<(Ika}`)@}6Y(b)Rr<m4ZI_~F#4Q-oc(d?d|q_Q;vX#>V>l`xB16^2#ev
zuU&3Sqjv4u#fujgJwN;GGr=`E9})sA9ffg=+X5*2ek3aUeyBUW7#I)O>N&!@TFo*!
z!y8vT@Fx121S@i_=xq+7EslUKjsCcE5uk3|xY3|N16x40(uX*@Zr!@Wh7F5|h&X%p
zEGAV@km>I~{`dp$Zu<1;O`A3?P@sU7kI=s*VnEKoyNDW?OJ%x^1*8z2@wNbIJH4ok
z*Yn>}f5w{HO*~m9oCDN2581W1C^WLdGV_~C;ei@wIg#%o+6qbSEYh1NPMj!NvZNl#
zEVb?I><H6VuU>umvf}1t`-dUNP1?3?i<Y+i0Z=1a)UfLpUv`0j=!`dl(C_q8TXEQu
z>duR++x7L5If82z;=MOKa|I63>OtyTNtYqr8}@%7G9lZxZPOex+d-x$8opPrURX?H
zb|u?9r(JmS=FJ5Q7F4WQ(e@`nnTYF|VZmM_+UmX^>9+nz+ZModZy&`!fB%beX+h0X
z9Dh+M`h~YGTpJ3Cgn+|!><_~QL`+6xNt6T&sC~rn_U+qgJUzM={y$E>rl_s+j_tT|
zMX8h^kGCYHIi*l6L(sPVu(&ONy3>my;}~(j<l4h$Rjygq;pwNlaRd8rgEB2aQS`P^
zv=sqIho>B$CdfgKtb=LtE-p0wIY#->ackjDls$a-;fJ;)RB`p=<Kwx--w*Dr{k|^=
zB8V2-0%W$gk4krPtqDOM4Qo3;{am-}?4U%}nN4<>b1h20fJi^@Y(lH~O9CiAKR?@D
zFTo;dE-lYyf!7R9k+%K#<Bx5jn?FI^Bl)KZ6gyE_q3!~qUtgA?N@s(<nH#IOZR`H@
zbKUd3%pEgEf(h5CuVa_O1wJE2j4<ca;+fK=OK;e)Ap>SCpGF%`o;<l<zkct&`|hk+
zvyL58=EZFP!QayodKMo^oP6VrH$;m5#|vvzp;pS<jEfO6VVvBHcXYMad#go>PqDFP
zD84_!F;k(|b4`sw^x-eB{dr6)9w0=u0;Bd8xoF}k>og&}#Ln~P&HMJ-Z^@ZYoH)^B
z{;N`@N}-{l`l;}YZitgcevBBO858XpGHQ^%6DIr8%>%v(HPvMZVgaefxY({ux3>@5
z0*DRz5~|LL^XUIY>8D?=Kby%Co=kMO@Vzn9gw<=l?{nwxn|g%NlNn*Cowv8QU%q^~
z{wnh}ej2(|%W`Pf!tanFL--WIJ26^VSeW_c;tZMj)YMe&)4X}}!i5WKH?iKJ<cSGW
z?r`hwHKyY!Wu(l8Dyg*1yR7c*!w#IEO)C5JbNzLI4O|-?Gc?<YD=`99a#=ff?j)bg
zRHk(#v6x6`rUn-kE&TbKHEX{9`s;#OaxA#(;uCl6+U4o#$!E1`(?;a6#d&^TzkVI-
z@WmHjG-=Yr7C|<7VCLqCCb-Xohxa$_|4^-WHEiXSe!Diyd;82us64_q2ZUXtLo%K6
zPy%Wp)T9o6yLB2&aC@lvL~tlI$u#LrPkkA%Mv%U7`0!ydAJL~zABs<Si`qgMVTy$K
zUF(U4f9<u`C`{7SIE_u4Her-)Pnww!EnYIx#HVaO;WfjsOPg^4sX@Rq+7^KHOT}?P
z<)41J9=|3N1x|Qn<l1@JwpO?A-2fAWTF70qX3g-LETSh)lB{MHOd~6!#-vG;=wGf}
zIciiyAF2((KPV{3BnC&KwNeL&_Vx9}Kv;PI8{Zl_bg1~;!{=Avm5DM}V{yB+?b<|K
zHf3!N;Cj-=Ou1I3s!5?>tVuOs(ZB?*nR^08<gR-4>Q?ZimRpm-jEF|Pivjb_JMWlu
zMuFf<Gi7G++L)La>dCEqy6tbh{r20!%1}~8-VD`E`MO%lt}PW;8Be!J+^Evy%FPJR
z0Xlrj=FGWP9d%_*?!qmN|G1;n7ct+&^kSw+Ld1?8J1o*<axyY9678*5*ErG$JLxs1
zGMPu)vnQ=2Ra}dJqtC@NFiVb~pXu{h>CmN17crvF-PpzShmmV#!Gck`MZ=tNWm^ER
zs2t#%O^r~>#%u|@=*D#I@3$3WN>+*)A}i2|w@6f)XkZhH^)K}jjfj2s-FNyVCz(u&
zSoHk~y~)h*L9ooKA3V1jHELw~NEvm|{JqP+ANEbG?iEwQWdAN5<9d~DSTH^_ItRpD
z(;{7!!h^`n8Aq<!*k6<fGn2?1Vn(=$^Dsq<Ot1rYNDZQT=FAzUAPKbbuA`%)nIqGT
zubxNrhDmKOcU!u2slE@YKr9OA-MhCKCy9xP?c29!O+hvkJM-Y8p85KpCvZVwHa*vh
zfTx#nJ>#XkqDnpezS*=$v4T=Cj(}T{ac2w}W&8k6l2g;488J=(g<7Uy`18*{fA78b
zSeMbbabsqnAOemkIK*=%lZK@hv8|p*@?+-3OO&YT`!HZkxaM6HJ0pru%;>2yI<Ilf
zVCEZ1W+=Pqiv1W(c5|}CIP26RO6CBr!<9vLVuc-vP*uh=V%(Y0&MIjjMNUacWrSN+
z&r?*)x(_{%xMU_ZhfI;9wDo<^59C1jtO^w>Xx=*=P!!}mv-%9JTD1}}`<35+U}h5$
z7K=$UV%$WfPu-|4)qlEkq?6aNyF0c?8OCaCrP8g|ew!I_W!3GTQws#(*wd`yX{A4W
z5j*RtB>hEV1ldPF@GilZ@RPMZjEB~(TZ?<5WcKghj}pFk@gg-UOkJ6L7BvYiTC^bN
zW^!J8NfJQJ$XxtwA7w}y*D#&|-Ra!62`h6zK79Q0k~3pV;dUA=l#X!=M%*##(B}f}
zWy_Yajz+`E_689*w%GhyrXaPIdg3F<QByQRM(g<T;~K^K26<Pl$Pk@Ru+I`s^T~AF
zNV4c#wrsJIQp4N`qkralig#=Ug=<z9rrQ=kBVEyi#zR%<F~z@Kk^|<szFk~;c3c@X
z+pla|A{ftDr8@@i=y2z6Wfc$c5y7?X6pNnJq9CDoXt?Mm*CM=x?RqLRZBx+l)mL9x
znYCmc3d@4TZ36=X%?FtF4CGhE#~r=6ku<o<wZ)AfEHKVi`9po&o_>2hY<6kyc*-!8
z8C&v2{FyMWs&s=9SF$KNm5wdDrmoZ@lW9wct$`^T<hn_Sf$EPx{)i}%+a$#;7KV^<
zwI#!5G62xq)22-mNj&Yd;1sWq>5|>McMHvr#^~F(uhzl5!TMiuX9}TJnMJ>)E6=bd
z<1xiP9_RY>bN%6{Qmntj!6w&dWSogS!8n>Kol3`EGZI)O0W8QAHqIP0>t_go#q1$>
zz=^^gf(}vii%AvrAtr3v@{$!CqJ)^4Zsu57=*Jm-z<%h^AuGui)V`Se;Ira>90AKS
zlqmqXwu$cPk;VEB$^msLz27h46I+F;SW#0a%5MxR-Dt!$l`dvLwCW>fnXNNx{w{=?
zOl~qkDMBX_BQRnMvsU_;_-|TImiCxGjCO{>COpHxAKYdE2zty&w4gf^#;0y}f32s}
z(=XQ@JG;*cXPJYCsEejtR#es*ReCxju8O(q`n~Tyc&O}dL7<e4*(L1q%9Sgbh5Yr`
zUzwsJ1u8Ia*RGwmI$PuDku-4WG29xn<_*#>JP)7s`|S%0w>E-kCH7PpHzD0<!I65S
z+s5}JOjzyKvZdqG&vg?|mnkbeD4HTC#+<vtxT?}s#)T2rRJ!7GV#}BT5nEK@Vr9c*
z12xiJr%oLZ){2es+19RIn<=vvlWf+nBAwfzLkDwy^=Cf#fcj@;-6*R+`12L@rQ3%y
zy^QPWRv4cV>po(Un`ez2P)Ay)qT{IfRVjR>gp5*H7*`j$pwfjAFQ*!DWv@szE+hS0
zk`7{X3>+To6PkKFt+a8aaHlgZIAr-JH4MahdP8iTXy(nCXw&D_%^!nL=bqv|?5Jj#
zz}0bWlyOa;r&H;0J$R~1jWYI6zg$->X8-ZXivC;bu~La@XV#^P^=m?<n;UUTYT;cg
zB3-p=6-6E90BUEf<4iZq^tf^3NU&RH?eyu>LfhkFvBJxGXYnrX8EXx!#987YxuIB`
zOHs<k{R7Do<AN&FxMrIwt6JTZalR3w&!;$MIe=>xr}TWQj9;QBOHIl7^R?7ugH^^+
z>B5L3;Auu2=MAn|`t<jGrHr2?b%aqGv*ry_9QZ{jNpha(ZK@GC!{R=sze#-%{4zgf
zi=2`S(c)<86RdMhj1U|gta$?FPo+P@3L~LXD7j&6H599p0&p!zR}8oQFfrm&Ho3Kb
z&9TydvYw0a^C%A<vUyW??@gWoTVLWYWqq;2c%o40su5S$EgDt2SQAm64i|shD*zuk
za-<$QlQurKpoYm!Oef-r5D{Fn+>%KO>y+u=)lX-T?#`%sflIhQmIar<<Vh^PWc+9f
zoPuE-5-1k)^_elQ2sqz<GA7uiVO=(#XYJmj<mJD(o4O{~={x^Ai&7f1Js`V7u#lWZ
z<t$87ReDXMO2=&zahdXc)<}2C@u`9f64B;!QhH#K;!LZB^|XpMBjK;V{t|6mGAy-r
zPckPKv!;kM$+4<ct7<(nZIDwHD^O72;Rio8p8`--x>3f{3pm0)BgS>m6zA%t*>*k$
zaIJb2_5;4CvN)yLqODC9ZfT@2o($tcrCTuKs;^#=ed@%TTr_Z|rKD)=@NmQ00drc-
z&oFO=&Sx^3NoqV~bS!O+IE2w;E}J_L6bu+JKueEnovk;-wg6my;kUimETd5CYjwn|
zpP*aS>ggHp`L1IQAzD#8sPT4fOZvq$Te9utMO&LLG%zkydYTbu7ZKJt7`7dVS9Tsa
zbk0!nO@Wa002aXU!6*&YMAe*M{TZgJ1xL(QFmJ__7FJcy8e)zVYnO;gF?|PH+Kj*>
zM~;Y_Bpewnl(v~@3*PmN8#YeoJ5a{1bZOc+XDD!C6d0CuDmqPA)L_|;7VI9W3OI}_
zDm`F}X2cP2LtKVd=OFT0zU%eB?kdS`M4LLyOyR|2@7c3wN=k|qYb%h~Poi-ZHHXyp
z>+es&1NRY~sc*^BB3$i0nowW=^Ei6Ul2GdfoOU|K;d;<ir|M-;;7_}y)?0t*?*FL=
zCBf`Axnx`O#RkSv>6VQ+!MJiH)YyL}Pyl}N<Vlk)VKqoi8Wb3?2N9`JWD0MMT$#yK
zCO3pflSwr5*o}QQ$h?t47Zt50pOWcI*7e{hiic;(o*HN2ft!$P{xXg3so~Du`{X1A
zE*Mw5@g~LRZEha8qm9Zq0uJM-bj^q(;OUIG8lBY$N$q&!FT+G6Hq`c%DRWzRbj&Ox
zutdScWn7Le5UU253$q@|K_z!!7X|%Bd71L7b-=6>lnJ6)!$oev2BFsTP7$^9sAW!k
zp;eCXz)j@3dP#@jQ|bq&zRH58W!qaL;7hj&l`f1p0*>33X2it~5=i&fW0M~~RQ8o6
zs`@u+G?E@zXKGQ#jvdS77ZrmVHIKYO0hKWKtU%m+;A3NurDrpfHjCii9p+fIcuvyZ
z8sms|moDWNC$<gR(JpX%+hs7mO<^3BZeqk2q#1D~7)SPhxt1hk1Wj(THFWl9B23HF
z*~{LYxvE~hdOW9i;G-8eFw3JF;!Ma*r{1HS#;tbZn=&f9<VdbD*|jS;kMnC1va4gz
zj`oC9M#eQG{<3*oCPK3{d){J(NTA35C)UI5Y;*KTE@TL?R#!agZ?{e{p3{+-xxH3~
z1YeNBKD}yKog=wMz%e~N-l({AO9!klwMGcIk@42)jkrE8Q+8|gPpVJMa__#f2OvLq
z(b@#oBO&Nf-Gryo0&_OG;;VwPEYS>xC?^)dwQftOl~ajUxK>4bykE1h-JKP>Q#%AJ
z0**@0#E7fWStE5DLuF==5R8@bHfAXw6_1(ThiOxu%9wk9$a+kw2eZL7LjB=5$LghW
zSg55LYpsFR?$)j1(xi@|yI$i99nv!{jJT%KMI^4oWoovq5wd^$vSEi|vYO;TfHzYL
z9#03Vh`qeLwCQhB-%R0UM|(#(ja}u`vQ^F&YC$)mU89WSi0O@%@9Gk|^R>{O2zaO9
zj8wW3i5ucFkzb|r*<ZJ-id-otV>b*dWx9`tWj=)xBQ|T&qIkX-h*W<zh*mDa{r<|)
zqePAuYC$)m-KIt91rhC+@9N6epTT%Y8smye*Npg5CHJDmWzE=P8X4!C9oQ63&`m0v
za-;0!Iv&+XU}lp_C-FS|K6xdU*kNj<O}oUHj0fVZk;|MKd*)R4ThrcJqq3Vl1q=f}
zuM?UI*9N+UfSagvB@#EpWd_DcxFO(NrO^$%Y_-7(nn%T*?f0QNlq@<sHfTiubZHZr
z-YMB48r~;wcIw`bTB987l9@j|n!J8h*U1Z8tk}~{rP~7I7UMD^aU~ev+AK7+Gez}c
z6HB0Gy*yiEJx+a2h8zpRq=_K<!Gn9$7cm=~J&A^Bg!+?7jy20aBidU}I-0zF+j0R>
z9hdL!9=1y`-bJYNOpJI6xfjKbBlRr)5{pCt90!=dmaL|z{I^{^v)YrLlvu<e%8CTh
zf8D*3aO7L+<+3f&5Iri;iLXS-(Q%fZUwvEA<nH$UKdu$By{nFHA>bB_xM^IbX4@d0
z^<Uc8Wr94r0WVy*@THetdgP2VsZCKUHHxh8H3n7+)qJlXE{7UhWyeMfe~d4k?DD3f
zy@5rK(lgUWlh-QeJYrgt6}!9RaV<x>4FYaKr5hvhwB)Mlwl!b8_2uO|yD`&E?m|cz
zzghZ0PJ{A;f4{8_GoDb)#2E=3ljUGHZ=FdxI+h7{6BjERFW(vFSS{xV)YeBPn!ItH
zV$%a#t=RLrO1GZzbShoAZRSQ?IO^opqYn?g@Z0yA1&49!7T8UNB2vCAl>~V9Bpn}s
z)jcNl#(&<qvj5C7JO{NFl)`>CJkLtHTh|%wJ!FPs-Kx)^nKq{$y4$uYGjDyn6?=qq
zo5;9<aa~-NF1c!m%hX7m=$lEgOE>r9DQPlJgMwkaW!C<%2?O(+?9M@;#Ns=0*NlNo
zWuFOgcE=KVl6&|6;;WOf9cq=J`dy#O!o%zm9qq(ke@z-bn`YXaYb+GXyldAA0nr_d
z%G`u*O{I%S+*)##Nij0*@kd6H(72_3ak-v!+GIR%o>?-_PB;`ZP;tuqCMholfcXk~
z&21xu7ljAZ$q@8SiC{BEWs|OKq(HTSfkclEc52nyfwGfl)l6HUn;IK;`~UT;x;VbW
z^4;Aq+(vz_SLqt#N~+By5?5zPwdATYA&$JS+xu?nNpJQO<I6YiKS(PcGhtu6#HnNJ
z8}=L~WTws>XPd$ma<HuArL2@GZ!sUue(!u&S{4frEJ!^>f49P^eV1q1A2TOfoI$kt
z?#RCVJr*al58KrxY*$wu-Fn8;8F4GgRS||m`jV6u44%kCL-@{I(&<!*P?LY_?sc>=
zt0p;2s7F^*xkRBUDH^uAqFRZ`Xf6MF`20#%4XYAufk91Xtv4Wg>`JH3JscR}Ijfc|
zBl@y>-az;JRTss#3)}S?u{LhGo^i!XPalcvjJU<*s;Gic>t57aWeOVc<#6d&?C!~G
z&<*?FU%#(EOMuVc*rmnkCe-U+vek*|k11b0W1USa?;42i=|IT&WQkU+=My>@8-niM
z{VMxKwhrCZiMW-lS~gU=d2$s2S7uL?;#Iz|QWUnOm8LUix@u2vDgY@XxqR~wU+jp~
zvo4t0C%(9W?ez$vF$Y|(eucWS_jUH@R%~>LQ-`i;68)sme&V02(yiF;E?vv{tZN>!
zld1+7M{j49ar2^=v>8%mF+#fKCrYidvSe8OUOdb(^DnFS^yVviBMuGb8#>6*u{?nt
zzYSY|s9Pc^E3mT2I1|&~EOKnsMzztZKla1{fAS9i4z6~uwk|a_uyJr|JKQj`rsgVL
zmwU0zh?`cfrdxiZTTv)h8>vf%W5tr;vNrGSlYtRecgz3W+p#HGP|32n;c&HUQqTSC
z8FiO#Z5^E2US(WS>6VN*F1qm2ZAh*vrRkO{SJf(I7Kxi|3n12zn(ytyj#bK@CnkIQ
zm|JS}_M}ZtAB=LWUqgwbpS&oUL;Vz-A<a^|f>ZxNRpv*&5|sL?%6RsTxNViIY0FPC
z*cL#1EsN#7eawwGdV5T$Q;&BXYLt}{o!K~dwPUioLx)l<5h7$(8OMlswra%Fsq~DJ
ztLpL-3ze&^A+>Q^0QEaNlubXhy?s<8E*5SmbCgrUo%p(w=FJVBn3QN+5J12^tGf*R
zvg*9>W~>~+cOtU3(TMBfvUJH+lgd?5<D@P|(3YQMvMqqNU7Ljs`l6ljw|fn6s8d<$
z?aaj4+=3wB#qHa+E-_+OorQ_bgHqdOMWt(HPPSIArfmzLYy@F6;<{}CGT7V4oo$OI
zZE*c?tkcU)r8sAuWgYJbc+GN-ufJJt{6f#g$t_tiMB%X-mt`{&PoG@15SRHIwgoV(
zAyxPN&^hXH#=U)FOU;TcIdyaK&u1!IUQotgNxES=;%w307W%l_S8}zhTiLl+pYr1u
zHdvI{GI(dZke!-J&r)*LWcf)8gAq?#QJ5|+v%I%Yd<F0La?>`I89KXo=N`_rDmW-C
z%fRa26Rq9B)s7e(#=G?>J92hC|L7J(;>J|lzfZ1eahcc%LcfMoG2+U$0EyM+C3?(B
zs4{hJg~5|bweREt%L>JkWsRdp+i<l*t5a~^rd7$0hu4@9)@(`gt1uq6y9))$Sx>I2
z%TLl(uCip<bXx%bEp^$h%{!{bn8lUf8C>d>W-irBNm>3^rrCn_&S;#V)!8P<-M&F}
z*RI{l4f(dt^x#((#kUVleT_O*N+5L8;@a#<Ml+<Q$yH^(Y+JL%+nO%g)_6g3L!a<^
zV;9vJ_*MB1or=|~${wN0M@w1G0$CZ6nAP*Dm=llJtLpM<>(cKGtoF^+hBL!j1w?lY
zN+IZ0N+1;8)hTkOO_jW>x)>o{b$Y5|#DjKF*w%V!N~?f`SLSSLGI44B&&OB)-+Sd-
zH7{1DvJ>M<Qv3w?wKyIi-C0rZ9wi(+tGl&nUb=VRs-F(8H^#f!%+NOTH@06C+bMAC
zYau(jgsMl#4%(fAws%~T)ZRa;jnBFkQ-hk0@bVn^MfIM&E3|yMczq8Srq{Wlr1)u}
z*dp)pq(nd>AQ6xVNCYGT5&?;TL_i`S5s(N-1SA3y0f~S_Kq4R!kO)WwBmxoviGV~v
zA|Mfv2uK7Z0ulj<fJ8tdAQ6xVNCYGT5&?;TL_i`S5s(N-1SA3y0f~S_Kq4R!kO)Ww
oBmxoviGV~vA|Mfv2uK7Z0ulj<fJ8tdAQ6xVNCYGTIRt_K1FqNS*8l(j

diff --git a/studio/src-tauri/windows/hooks.nsh b/studio/src-tauri/windows/hooks.nsh
index 64655e8f6e..8c68a3e54f 100644
--- a/studio/src-tauri/windows/hooks.nsh
+++ b/studio/src-tauri/windows/hooks.nsh
@@ -1,8 +1,9 @@
 ; Unsloth Studio NSIS installer hooks
 
 !macro NSIS_HOOK_POSTUNINSTALL
-  MessageBox MB_YESNO|MB_ICONQUESTION "Remove all Unsloth data ($PROFILE\.unsloth)?$\n$\nThis deletes installed models, training outputs, and configuration." IDNO skip_cleanup
-    RMDir /r "$PROFILE\.unsloth"
-    DetailPrint "Removed $PROFILE\.unsloth"
-  skip_cleanup:
+  ; Desktop uninstall must not remove $PROFILE\.unsloth. The CLI/web
+  ; installers also use that tree for environments, models, outputs, and
+  ; configuration, and there has been no prior public desktop release whose
+  ; private state needs cleanup here.
+  DetailPrint "Preserved shared Unsloth data at $PROFILE\.unsloth"
 !macroend
diff --git a/studio/src-tauri/windows/installer.nsi b/studio/src-tauri/windows/installer.nsi
new file mode 100644
index 0000000000..7cc2f223c7
--- /dev/null
+++ b/studio/src-tauri/windows/installer.nsi
@@ -0,0 +1,994 @@
+; Unsloth Studio custom NSIS template.
+; Source: https://raw.githubusercontent.com/tauri-apps/tauri/tauri-cli-v2.10.1/crates/tauri-bundler/src/bundle/windows/nsis/installer.nsi
+; Source tags tauri-cli-v2.10.1 and tauri-bundler-v2.8.1 resolve to commit 9b17a7aeae9a83222ffe829aa4e2d8a5ba6bed8c.
+; Verified release CLI: @tauri-apps/cli 2.10.1 / tauri-cli 2.10.1
+; Verified bundler/template crate: tauri-bundler 2.8.1
+; Copied: 2026-04-30
+; Local patches:
+; - UNSLOTH_PATCH_START: non-WiX newer/unknown upgrades install in place.
+; - UNSLOTH_PATCH_START: guard non-WiX upgrade leave routing.
+
+Unicode true
+ManifestDPIAware true
+; Add in `dpiAwareness` `PerMonitorV2` to manifest for Windows 10 1607+ (note this should not affect lower versions since they should be able to ignore this and pick up `dpiAware` `true` set by `ManifestDPIAware true`)
+; Currently undocumented on NSIS's website but is in the Docs folder of source tree, see
+; https://github.com/kichik/nsis/blob/5fc0b87b819a9eec006df4967d08e522ddd651c9/Docs/src/attributes.but#L286-L300
+; https://github.com/tauri-apps/tauri/pull/10106
+ManifestDPIAwareness PerMonitorV2
+
+!if "{{compression}}" == "none"
+  SetCompress off
+!else
+  ; Set the compression algorithm. We default to LZMA.
+  SetCompressor /SOLID "{{compression}}"
+!endif
+
+!include MUI2.nsh
+!include FileFunc.nsh
+!include x64.nsh
+!include WordFunc.nsh
+!include "utils.nsh"
+!include "FileAssociation.nsh"
+!include "Win\COM.nsh"
+!include "Win\Propkey.nsh"
+!include "StrFunc.nsh"
+${StrCase}
+${StrLoc}
+
+{{#if installer_hooks}}
+!include "{{installer_hooks}}"
+{{/if}}
+
+!define WEBVIEW2APPGUID "{F3017226-FE2A-4295-8BDF-00C3A9A7E4C5}"
+
+!define MANUFACTURER "{{manufacturer}}"
+!define PRODUCTNAME "{{product_name}}"
+!define VERSION "{{version}}"
+!define VERSIONWITHBUILD "{{version_with_build}}"
+!define HOMEPAGE "{{homepage}}"
+!define INSTALLMODE "{{install_mode}}"
+!define LICENSE "{{license}}"
+!define INSTALLERICON "{{installer_icon}}"
+!define SIDEBARIMAGE "{{sidebar_image}}"
+!define HEADERIMAGE "{{header_image}}"
+!define MAINBINARYNAME "{{main_binary_name}}"
+!define MAINBINARYSRCPATH "{{main_binary_path}}"
+!define BUNDLEID "{{bundle_id}}"
+!define COPYRIGHT "{{copyright}}"
+!define OUTFILE "{{out_file}}"
+!define ARCH "{{arch}}"
+!define ADDITIONALPLUGINSPATH "{{additional_plugins_path}}"
+!define ALLOWDOWNGRADES "{{allow_downgrades}}"
+!define DISPLAYLANGUAGESELECTOR "{{display_language_selector}}"
+!define INSTALLWEBVIEW2MODE "{{install_webview2_mode}}"
+!define WEBVIEW2INSTALLERARGS "{{webview2_installer_args}}"
+!define WEBVIEW2BOOTSTRAPPERPATH "{{webview2_bootstrapper_path}}"
+!define WEBVIEW2INSTALLERPATH "{{webview2_installer_path}}"
+!define MINIMUMWEBVIEW2VERSION "{{minimum_webview2_version}}"
+!define UNINSTKEY "Software\Microsoft\Windows\CurrentVersion\Uninstall\${PRODUCTNAME}"
+!define MANUKEY "Software\${MANUFACTURER}"
+!define MANUPRODUCTKEY "${MANUKEY}\${PRODUCTNAME}"
+!define UNINSTALLERSIGNCOMMAND "{{uninstaller_sign_cmd}}"
+!define ESTIMATEDSIZE "{{estimated_size}}"
+!define STARTMENUFOLDER "{{start_menu_folder}}"
+
+Var PassiveMode
+Var UpdateMode
+Var NoShortcutMode
+Var WixMode
+Var OldMainBinaryName
+
+Name "${PRODUCTNAME}"
+BrandingText "${COPYRIGHT}"
+OutFile "${OUTFILE}"
+
+; We don't actually use this value as default install path,
+; it's just for nsis to append the product name folder in the directory selector
+; https://nsis.sourceforge.io/Reference/InstallDir
+!define PLACEHOLDER_INSTALL_DIR "placeholder\${PRODUCTNAME}"
+InstallDir "${PLACEHOLDER_INSTALL_DIR}"
+
+VIProductVersion "${VERSIONWITHBUILD}"
+VIAddVersionKey "ProductName" "${PRODUCTNAME}"
+VIAddVersionKey "FileDescription" "${PRODUCTNAME}"
+VIAddVersionKey "LegalCopyright" "${COPYRIGHT}"
+VIAddVersionKey "FileVersion" "${VERSION}"
+VIAddVersionKey "ProductVersion" "${VERSION}"
+
+# additional plugins
+!addplugindir "${ADDITIONALPLUGINSPATH}"
+
+; Uninstaller signing command
+!if "${UNINSTALLERSIGNCOMMAND}" != ""
+  !uninstfinalize '${UNINSTALLERSIGNCOMMAND}'
+!endif
+
+; Handle install mode, `perUser`, `perMachine` or `both`
+!if "${INSTALLMODE}" == "perMachine"
+  RequestExecutionLevel admin
+!endif
+
+!if "${INSTALLMODE}" == "currentUser"
+  RequestExecutionLevel user
+!endif
+
+!if "${INSTALLMODE}" == "both"
+  !define MULTIUSER_MUI
+  !define MULTIUSER_INSTALLMODE_INSTDIR "${PRODUCTNAME}"
+  !define MULTIUSER_INSTALLMODE_COMMANDLINE
+  !if "${ARCH}" == "x64"
+    !define MULTIUSER_USE_PROGRAMFILES64
+  !else if "${ARCH}" == "arm64"
+    !define MULTIUSER_USE_PROGRAMFILES64
+  !endif
+  !define MULTIUSER_INSTALLMODE_DEFAULT_REGISTRY_KEY "${UNINSTKEY}"
+  !define MULTIUSER_INSTALLMODE_DEFAULT_REGISTRY_VALUENAME "CurrentUser"
+  !define MULTIUSER_INSTALLMODEPAGE_SHOWUSERNAME
+  !define MULTIUSER_INSTALLMODE_FUNCTION RestorePreviousInstallLocation
+  !define MULTIUSER_EXECUTIONLEVEL Highest
+  !include MultiUser.nsh
+!endif
+
+; Installer icon
+!if "${INSTALLERICON}" != ""
+  !define MUI_ICON "${INSTALLERICON}"
+!endif
+
+; Installer sidebar image
+!if "${SIDEBARIMAGE}" != ""
+  !define MUI_WELCOMEFINISHPAGE_BITMAP "${SIDEBARIMAGE}"
+!endif
+
+; Installer header image
+!if "${HEADERIMAGE}" != ""
+  !define MUI_HEADERIMAGE
+  !define MUI_HEADERIMAGE_BITMAP  "${HEADERIMAGE}"
+!endif
+
+; Define registry key to store installer language
+!define MUI_LANGDLL_REGISTRY_ROOT "HKCU"
+!define MUI_LANGDLL_REGISTRY_KEY "${MANUPRODUCTKEY}"
+!define MUI_LANGDLL_REGISTRY_VALUENAME "Installer Language"
+
+; Installer pages, must be ordered as they appear
+; 1. Welcome Page
+!define MUI_PAGE_CUSTOMFUNCTION_PRE SkipIfPassive
+!insertmacro MUI_PAGE_WELCOME
+
+; 2. License Page (if defined)
+!if "${LICENSE}" != ""
+  !define MUI_PAGE_CUSTOMFUNCTION_PRE SkipIfPassive
+  !insertmacro MUI_PAGE_LICENSE "${LICENSE}"
+!endif
+
+; 3. Install mode (if it is set to `both`)
+!if "${INSTALLMODE}" == "both"
+  !define MUI_PAGE_CUSTOMFUNCTION_PRE SkipIfPassive
+  !insertmacro MULTIUSER_PAGE_INSTALLMODE
+!endif
+
+; 4. Custom page to ask user if he wants to reinstall/uninstall
+;    only if a previous installation was detected
+Var ReinstallPageCheck
+Page custom PageReinstall PageLeaveReinstall
+Function PageReinstall
+  ; Uninstall previous WiX installation if exists.
+  ;
+  ; A WiX installer stores the installation info in registry
+  ; using a UUID and so we have to loop through all keys under
+  ; `HKLM\SOFTWARE\Microsoft\Windows\CurrentVersion\Uninstall`
+  ; and check if `DisplayName` and `Publisher` keys match ${PRODUCTNAME} and ${MANUFACTURER}
+  ;
+  ; This has a potential issue that there maybe another installation that matches
+  ; our ${PRODUCTNAME} and ${MANUFACTURER} but wasn't installed by our WiX installer,
+  ; however, this should be fine since the user will have to confirm the uninstallation
+  ; and they can chose to abort it if doesn't make sense.
+  StrCpy $0 0
+  wix_loop:
+    EnumRegKey $1 HKLM "SOFTWARE\Microsoft\Windows\CurrentVersion\Uninstall" $0
+    StrCmp $1 "" wix_loop_done ; Exit loop if there is no more keys to loop on
+    IntOp $0 $0 + 1
+    ReadRegStr $R0 HKLM "SOFTWARE\Microsoft\Windows\CurrentVersion\Uninstall\$1" "DisplayName"
+    ReadRegStr $R1 HKLM "SOFTWARE\Microsoft\Windows\CurrentVersion\Uninstall\$1" "Publisher"
+    StrCmp "$R0$R1" "${PRODUCTNAME}${MANUFACTURER}" 0 wix_loop
+    ReadRegStr $R0 HKLM "SOFTWARE\Microsoft\Windows\CurrentVersion\Uninstall\$1" "UninstallString"
+    ${StrCase} $R1 $R0 "L"
+    ${StrLoc} $R0 $R1 "msiexec" ">"
+    StrCmp $R0 0 0 wix_loop_done
+    StrCpy $WixMode 1
+    StrCpy $R6 "SOFTWARE\Microsoft\Windows\CurrentVersion\Uninstall\$1"
+    Goto compare_version
+  wix_loop_done:
+
+  ; Check if there is an existing installation, if not, abort the reinstall page
+  ReadRegStr $R0 SHCTX "${UNINSTKEY}" ""
+  ReadRegStr $R1 SHCTX "${UNINSTKEY}" "UninstallString"
+  ${IfThen} "$R0$R1" == "" ${|} Abort ${|}
+
+  ; Compare this installar version with the existing installation
+  ; and modify the messages presented to the user accordingly
+  compare_version:
+  StrCpy $R4 "$(older)"
+  ${If} $WixMode = 1
+    ReadRegStr $R0 HKLM "$R6" "DisplayVersion"
+  ${Else}
+    ReadRegStr $R0 SHCTX "${UNINSTKEY}" "DisplayVersion"
+  ${EndIf}
+  ${IfThen} $R0 == "" ${|} StrCpy $R4 "$(unknown)" ${|}
+
+  nsis_tauri_utils::SemverCompare "${VERSION}" $R0
+  Pop $R0
+
+  ; UNSLOTH_PATCH_START: non-WiX newer/unknown upgrades install in place.
+  ; Do not invoke stale internal NSIS uninstallers during normal desktop
+  ; upgrades; continue to Section Install so Tauri handles the running app.
+  ${If} $WixMode <> 1
+    ${If} $UpdateMode = 1
+      Abort
+    ${ElseIf} $R0 = 1
+      Abort
+    ${ElseIf} $R0 <> 0
+      ${If} $R0 <> -1
+        StrCpy $R0 1
+        Abort
+      ${EndIf}
+    ${EndIf}
+  ${EndIf}
+  ; UNSLOTH_PATCH_END
+
+  ; Reinstalling the same version
+  ${If} $R0 = 0
+    StrCpy $R1 "$(alreadyInstalledLong)"
+    StrCpy $R2 "$(addOrReinstall)"
+    StrCpy $R3 "$(uninstallApp)"
+    !insertmacro MUI_HEADER_TEXT "$(alreadyInstalled)" "$(chooseMaintenanceOption)"
+  ; Upgrading
+  ${ElseIf} $R0 = 1
+    StrCpy $R1 "$(olderOrUnknownVersionInstalled)"
+    StrCpy $R2 "$(uninstallBeforeInstalling)"
+    StrCpy $R3 "$(dontUninstall)"
+    !insertmacro MUI_HEADER_TEXT "$(alreadyInstalled)" "$(choowHowToInstall)"
+  ; Downgrading
+  ${ElseIf} $R0 = -1
+    StrCpy $R1 "$(newerVersionInstalled)"
+    StrCpy $R2 "$(uninstallBeforeInstalling)"
+    !if "${ALLOWDOWNGRADES}" == "true"
+      StrCpy $R3 "$(dontUninstall)"
+    !else
+      StrCpy $R3 "$(dontUninstallDowngrade)"
+    !endif
+    !insertmacro MUI_HEADER_TEXT "$(alreadyInstalled)" "$(choowHowToInstall)"
+  ${Else}
+    Abort
+  ${EndIf}
+
+  ; Skip showing the page if passive
+  ;
+  ; Note that we don't call this earlier at the begining
+  ; of this function because we need to populate some variables
+  ; related to current installed version if detected and whether
+  ; we are downgrading or not.
+  ${If} $PassiveMode = 1
+    Call PageLeaveReinstall
+  ${Else}
+    nsDialogs::Create 1018
+    Pop $R4
+    ${IfThen} $(^RTL) = 1 ${|} nsDialogs::SetRTL $(^RTL) ${|}
+
+    ${NSD_CreateLabel} 0 0 100% 24u $R1
+    Pop $R1
+
+    ${NSD_CreateRadioButton} 30u 50u -30u 8u $R2
+    Pop $R2
+    ${NSD_OnClick} $R2 PageReinstallUpdateSelection
+
+    ${NSD_CreateRadioButton} 30u 70u -30u 8u $R3
+    Pop $R3
+    ; Disable this radio button if downgrading and downgrades are disabled
+    !if "${ALLOWDOWNGRADES}" == "false"
+      ${IfThen} $R0 = -1 ${|} EnableWindow $R3 0 ${|}
+    !endif
+    ${NSD_OnClick} $R3 PageReinstallUpdateSelection
+
+    ; Check the first radio button if this the first time
+    ; we enter this page or if the second button wasn't
+    ; selected the last time we were on this page
+    ${If} $ReinstallPageCheck <> 2
+      SendMessage $R2 ${BM_SETCHECK} ${BST_CHECKED} 0
+    ${Else}
+      SendMessage $R3 ${BM_SETCHECK} ${BST_CHECKED} 0
+    ${EndIf}
+
+    ${NSD_SetFocus} $R2
+    nsDialogs::Show
+  ${EndIf}
+FunctionEnd
+Function PageReinstallUpdateSelection
+  ${NSD_GetState} $R2 $R1
+  ${If} $R1 == ${BST_CHECKED}
+    StrCpy $ReinstallPageCheck 1
+  ${Else}
+    StrCpy $ReinstallPageCheck 2
+  ${EndIf}
+FunctionEnd
+Function PageLeaveReinstall
+  ${NSD_GetState} $R2 $R1
+
+  ; If migrating from Wix, always uninstall
+  ${If} $WixMode = 1
+    Goto reinst_uninstall
+  ${EndIf}
+
+  ; In update mode, always proceeds without uninstalling
+  ${If} $UpdateMode = 1
+    Goto reinst_done
+  ${EndIf}
+
+  ; UNSLOTH_PATCH_START: guard non-WiX upgrade leave routing.
+  ; Passive/default upgrade choices must not ExecWait the old UninstallString.
+  ${If} $WixMode <> 1
+    ${If} $R0 = 1
+      Goto reinst_done
+    ${ElseIf} $R0 <> 0
+      ${If} $R0 <> -1
+        Goto reinst_done
+      ${EndIf}
+    ${EndIf}
+  ${EndIf}
+  ; UNSLOTH_PATCH_END
+
+  ; $R0 holds whether same(0)/upgrading(1)/downgrading(-1) version
+  ; $R1 holds the radio buttons state:
+  ;   1 => first choice was selected
+  ;   0 => second choice was selected
+  ${If} $R0 = 0 ; Same version, proceed
+    ${If} $R1 = 1              ; User chose to add/reinstall
+      Goto reinst_done
+    ${Else}                    ; User chose to uninstall
+      Goto reinst_uninstall
+    ${EndIf}
+  ${ElseIf} $R0 = 1 ; Upgrading
+    ${If} $R1 = 1              ; User chose to uninstall
+      Goto reinst_uninstall
+    ${Else}
+      Goto reinst_done         ; User chose NOT to uninstall
+    ${EndIf}
+  ${ElseIf} $R0 = -1 ; Downgrading
+    ${If} $R1 = 1              ; User chose to uninstall
+      Goto reinst_uninstall
+    ${Else}
+      Goto reinst_done         ; User chose NOT to uninstall
+    ${EndIf}
+  ${EndIf}
+
+  reinst_uninstall:
+    HideWindow
+    ClearErrors
+
+    ${If} $WixMode = 1
+      ReadRegStr $R1 HKLM "$R6" "UninstallString"
+      ExecWait '$R1' $0
+    ${Else}
+      ReadRegStr $4 SHCTX "${MANUPRODUCTKEY}" ""
+      ReadRegStr $R1 SHCTX "${UNINSTKEY}" "UninstallString"
+      ${IfThen} $UpdateMode = 1 ${|} StrCpy $R1 "$R1 /UPDATE" ${|} ; append /UPDATE
+      ${IfThen} $PassiveMode = 1 ${|} StrCpy $R1 "$R1 /P" ${|} ; append /P
+      StrCpy $R1 "$R1 _?=$4" ; append uninstall directory
+      ExecWait '$R1' $0
+    ${EndIf}
+
+    BringToFront
+
+    ${IfThen} ${Errors} ${|} StrCpy $0 2 ${|} ; ExecWait failed, set fake exit code
+
+    ${If} $0 <> 0
+    ${OrIf} ${FileExists} "$INSTDIR\${MAINBINARYNAME}.exe"
+      ; User cancelled wix uninstaller? return to select un/reinstall page
+      ${If} $WixMode = 1
+      ${AndIf} $0 = 1602
+        Abort
+      ${EndIf}
+
+      ; User cancelled NSIS uninstaller? return to select un/reinstall page
+      ${If} $0 = 1
+        Abort
+      ${EndIf}
+
+      ; Other erros? show generic error message and return to select un/reinstall page
+      MessageBox MB_ICONEXCLAMATION "$(unableToUninstall)"
+      Abort
+    ${EndIf}
+  reinst_done:
+FunctionEnd
+
+; 5. Choose install directory page
+!define MUI_PAGE_CUSTOMFUNCTION_PRE SkipIfPassive
+!insertmacro MUI_PAGE_DIRECTORY
+
+; 6. Start menu shortcut page
+Var AppStartMenuFolder
+!if "${STARTMENUFOLDER}" != ""
+  !define MUI_PAGE_CUSTOMFUNCTION_PRE SkipIfPassive
+  !define MUI_STARTMENUPAGE_DEFAULTFOLDER "${STARTMENUFOLDER}"
+!else
+  !define MUI_PAGE_CUSTOMFUNCTION_PRE Skip
+!endif
+!insertmacro MUI_PAGE_STARTMENU Application $AppStartMenuFolder
+
+; 7. Installation page
+!insertmacro MUI_PAGE_INSTFILES
+
+; 8. Finish page
+;
+; Don't auto jump to finish page after installation page,
+; because the installation page has useful info that can be used debug any issues with the installer.
+!define MUI_FINISHPAGE_NOAUTOCLOSE
+; Use show readme button in the finish page as a button create a desktop shortcut
+!define MUI_FINISHPAGE_SHOWREADME
+!define MUI_FINISHPAGE_SHOWREADME_TEXT "$(createDesktop)"
+!define MUI_FINISHPAGE_SHOWREADME_FUNCTION CreateOrUpdateDesktopShortcut
+; Show run app after installation.
+!define MUI_FINISHPAGE_RUN
+!define MUI_FINISHPAGE_RUN_FUNCTION RunMainBinary
+!define MUI_PAGE_CUSTOMFUNCTION_PRE SkipIfPassive
+!insertmacro MUI_PAGE_FINISH
+
+Function RunMainBinary
+  nsis_tauri_utils::RunAsUser "$INSTDIR\${MAINBINARYNAME}.exe" ""
+FunctionEnd
+
+; Uninstaller Pages
+; 1. Confirm uninstall page
+Var DeleteAppDataCheckbox
+Var DeleteAppDataCheckboxState
+!define /ifndef WS_EX_LAYOUTRTL         0x00400000
+!define MUI_PAGE_CUSTOMFUNCTION_SHOW un.ConfirmShow
+Function un.ConfirmShow ; Add add a `Delete app data` check box
+  ; $1 inner dialog HWND
+  ; $2 window DPI
+  ; $3 style
+  ; $4 x
+  ; $5 y
+  ; $6 width
+  ; $7 height
+  FindWindow $1 "#32770" "" $HWNDPARENT ; Find inner dialog
+  System::Call "user32::GetDpiForWindow(p r1) i .r2"
+  ${If} $(^RTL) = 1
+    StrCpy $3 "${__NSD_CheckBox_EXSTYLE} | ${WS_EX_LAYOUTRTL}"
+    IntOp $4 50 * $2
+  ${Else}
+    StrCpy $3 "${__NSD_CheckBox_EXSTYLE}"
+    IntOp $4 0 * $2
+  ${EndIf}
+  IntOp $5 100 * $2
+  IntOp $6 400 * $2
+  IntOp $7 25 * $2
+  IntOp $4 $4 / 96
+  IntOp $5 $5 / 96
+  IntOp $6 $6 / 96
+  IntOp $7 $7 / 96
+  System::Call 'user32::CreateWindowEx(i r3, w "${__NSD_CheckBox_CLASS}", w "$(deleteAppData)", i ${__NSD_CheckBox_STYLE}, i r4, i r5, i r6, i r7, p r1, i0, i0, i0) i .s'
+  Pop $DeleteAppDataCheckbox
+  SendMessage $HWNDPARENT ${WM_GETFONT} 0 0 $1
+  SendMessage $DeleteAppDataCheckbox ${WM_SETFONT} $1 1
+FunctionEnd
+!define MUI_PAGE_CUSTOMFUNCTION_LEAVE un.ConfirmLeave
+Function un.ConfirmLeave
+  SendMessage $DeleteAppDataCheckbox ${BM_GETCHECK} 0 0 $DeleteAppDataCheckboxState
+FunctionEnd
+!define MUI_PAGE_CUSTOMFUNCTION_PRE un.SkipIfPassive
+!insertmacro MUI_UNPAGE_CONFIRM
+
+; 2. Uninstalling Page
+!insertmacro MUI_UNPAGE_INSTFILES
+
+;Languages
+{{#each languages}}
+!insertmacro MUI_LANGUAGE "{{this}}"
+{{/each}}
+!insertmacro MUI_RESERVEFILE_LANGDLL
+{{#each language_files}}
+  !include "{{this}}"
+{{/each}}
+
+Function .onInit
+  ${GetOptions} $CMDLINE "/P" $PassiveMode
+  ${IfNot} ${Errors}
+    StrCpy $PassiveMode 1
+  ${EndIf}
+
+  ${GetOptions} $CMDLINE "/NS" $NoShortcutMode
+  ${IfNot} ${Errors}
+    StrCpy $NoShortcutMode 1
+  ${EndIf}
+
+  ${GetOptions} $CMDLINE "/UPDATE" $UpdateMode
+  ${IfNot} ${Errors}
+    StrCpy $UpdateMode 1
+  ${EndIf}
+
+  !if "${DISPLAYLANGUAGESELECTOR}" == "true"
+    !insertmacro MUI_LANGDLL_DISPLAY
+  !endif
+
+  !insertmacro SetContext
+
+  ${If} $INSTDIR == "${PLACEHOLDER_INSTALL_DIR}"
+    ; Set default install location
+    !if "${INSTALLMODE}" == "perMachine"
+      ${If} ${RunningX64}
+        !if "${ARCH}" == "x64"
+          StrCpy $INSTDIR "$PROGRAMFILES64\${PRODUCTNAME}"
+        !else if "${ARCH}" == "arm64"
+          StrCpy $INSTDIR "$PROGRAMFILES64\${PRODUCTNAME}"
+        !else
+          StrCpy $INSTDIR "$PROGRAMFILES\${PRODUCTNAME}"
+        !endif
+      ${Else}
+        StrCpy $INSTDIR "$PROGRAMFILES\${PRODUCTNAME}"
+      ${EndIf}
+    !else if "${INSTALLMODE}" == "currentUser"
+      StrCpy $INSTDIR "$LOCALAPPDATA\${PRODUCTNAME}"
+    !endif
+
+    Call RestorePreviousInstallLocation
+  ${EndIf}
+
+
+  !if "${INSTALLMODE}" == "both"
+    !insertmacro MULTIUSER_INIT
+  !endif
+FunctionEnd
+
+
+Section EarlyChecks
+  ; Abort silent installer if downgrades is disabled
+  !if "${ALLOWDOWNGRADES}" == "false"
+  ${If} ${Silent}
+    ; If downgrading
+    ${If} $R0 = -1
+      System::Call 'kernel32::AttachConsole(i -1)i.r0'
+      ${If} $0 <> 0
+        System::Call 'kernel32::GetStdHandle(i -11)i.r0'
+        System::call 'kernel32::SetConsoleTextAttribute(i r0, i 0x0004)' ; set red color
+        FileWrite $0 "$(silentDowngrades)"
+      ${EndIf}
+      Abort
+    ${EndIf}
+  ${EndIf}
+  !endif
+
+SectionEnd
+
+Section WebView2
+  ; Check if Webview2 is already installed and skip this section
+  ${If} ${RunningX64}
+    ReadRegStr $4 HKLM "SOFTWARE\WOW6432Node\Microsoft\EdgeUpdate\Clients\${WEBVIEW2APPGUID}" "pv"
+  ${Else}
+    ReadRegStr $4 HKLM "SOFTWARE\Microsoft\EdgeUpdate\Clients\${WEBVIEW2APPGUID}" "pv"
+  ${EndIf}
+  ${If} $4 == ""
+    ReadRegStr $4 HKCU "SOFTWARE\Microsoft\EdgeUpdate\Clients\${WEBVIEW2APPGUID}" "pv"
+  ${EndIf}
+
+  ${If} $4 == ""
+    ; Webview2 installation
+    ;
+    ; Skip if updating
+    ${If} $UpdateMode <> 1
+      !if "${INSTALLWEBVIEW2MODE}" == "downloadBootstrapper"
+        Delete "$TEMP\MicrosoftEdgeWebview2Setup.exe"
+        DetailPrint "$(webview2Downloading)"
+        NSISdl::download "https://go.microsoft.com/fwlink/p/?LinkId=2124703" "$TEMP\MicrosoftEdgeWebview2Setup.exe"
+        Pop $0
+        ${If} $0 == "success"
+          DetailPrint "$(webview2DownloadSuccess)"
+        ${Else}
+          DetailPrint "$(webview2DownloadError)"
+          Abort "$(webview2AbortError)"
+        ${EndIf}
+        StrCpy $6 "$TEMP\MicrosoftEdgeWebview2Setup.exe"
+        Goto install_webview2
+      !endif
+
+      !if "${INSTALLWEBVIEW2MODE}" == "embedBootstrapper"
+        Delete "$TEMP\MicrosoftEdgeWebview2Setup.exe"
+        File "/oname=$TEMP\MicrosoftEdgeWebview2Setup.exe" "${WEBVIEW2BOOTSTRAPPERPATH}"
+        DetailPrint "$(installingWebview2)"
+        StrCpy $6 "$TEMP\MicrosoftEdgeWebview2Setup.exe"
+        Goto install_webview2
+      !endif
+
+      !if "${INSTALLWEBVIEW2MODE}" == "offlineInstaller"
+        Delete "$TEMP\MicrosoftEdgeWebView2RuntimeInstaller.exe"
+        File "/oname=$TEMP\MicrosoftEdgeWebView2RuntimeInstaller.exe" "${WEBVIEW2INSTALLERPATH}"
+        DetailPrint "$(installingWebview2)"
+        StrCpy $6 "$TEMP\MicrosoftEdgeWebView2RuntimeInstaller.exe"
+        Goto install_webview2
+      !endif
+
+      Goto webview2_done
+
+      install_webview2:
+        DetailPrint "$(installingWebview2)"
+        ; $6 holds the path to the webview2 installer
+        ExecWait "$6 ${WEBVIEW2INSTALLERARGS} /install" $1
+        ${If} $1 = 0
+          DetailPrint "$(webview2InstallSuccess)"
+        ${Else}
+          DetailPrint "$(webview2InstallError)"
+          Abort "$(webview2AbortError)"
+        ${EndIf}
+      webview2_done:
+    ${EndIf}
+  ${Else}
+    !if "${MINIMUMWEBVIEW2VERSION}" != ""
+      ${VersionCompare} "${MINIMUMWEBVIEW2VERSION}" "$4" $R0
+      ${If} $R0 = 1
+        update_webview:
+          DetailPrint "$(installingWebview2)"
+          ${If} ${RunningX64}
+            ReadRegStr $R1 HKLM "SOFTWARE\WOW6432Node\Microsoft\EdgeUpdate" "path"
+          ${Else}
+            ReadRegStr $R1 HKLM "SOFTWARE\Microsoft\EdgeUpdate" "path"
+          ${EndIf}
+          ${If} $R1 == ""
+            ReadRegStr $R1 HKCU "SOFTWARE\Microsoft\EdgeUpdate" "path"
+          ${EndIf}
+          ${If} $R1 != ""
+            ; Chromium updater docs: https://source.chromium.org/chromium/chromium/src/+/main:docs/updater/user_manual.md
+            ; Modified from "HKEY_LOCAL_MACHINE\SOFTWARE\WOW6432Node\Microsoft\Windows\CurrentVersion\Uninstall\Microsoft EdgeWebView\ModifyPath"
+            ExecWait `"$R1" /install appguid=${WEBVIEW2APPGUID}&needsadmin=true` $1
+            ${If} $1 = 0
+              DetailPrint "$(webview2InstallSuccess)"
+            ${Else}
+              MessageBox MB_ICONEXCLAMATION|MB_ABORTRETRYIGNORE "$(webview2InstallError)" IDIGNORE ignore IDRETRY update_webview
+              Quit
+              ignore:
+            ${EndIf}
+          ${EndIf}
+      ${EndIf}
+    !endif
+  ${EndIf}
+SectionEnd
+
+Section Install
+  SetOutPath $INSTDIR
+
+  !ifmacrodef NSIS_HOOK_PREINSTALL
+    !insertmacro NSIS_HOOK_PREINSTALL
+  !endif
+
+  !insertmacro CheckIfAppIsRunning "${MAINBINARYNAME}.exe" "${PRODUCTNAME}"
+
+  ; Copy main executable
+  File "${MAINBINARYSRCPATH}"
+
+  ; Copy resources
+  {{#each resources_dirs}}
+    CreateDirectory "$INSTDIR\\{{this}}"
+  {{/each}}
+  {{#each resources}}
+    File /a "/oname={{this.[1]}}" "{{no-escape @key}}"
+  {{/each}}
+
+  ; Copy external binaries
+  {{#each binaries}}
+    File /a "/oname={{this}}" "{{no-escape @key}}"
+  {{/each}}
+
+  ; Create file associations
+  {{#each file_associations as |association| ~}}
+    {{#each association.ext as |ext| ~}}
+       !insertmacro APP_ASSOCIATE "{{ext}}" "{{or association.name ext}}" "{{association-description association.description ext}}" "$INSTDIR\${MAINBINARYNAME}.exe,0" "Open with ${PRODUCTNAME}" "$INSTDIR\${MAINBINARYNAME}.exe $\"%1$\""
+    {{/each}}
+  {{/each}}
+
+  ; Register deep links
+  {{#each deep_link_protocols as |protocol| ~}}
+    WriteRegStr SHCTX "Software\Classes\\{{protocol}}" "URL Protocol" ""
+    WriteRegStr SHCTX "Software\Classes\\{{protocol}}" "" "URL:${BUNDLEID} protocol"
+    WriteRegStr SHCTX "Software\Classes\\{{protocol}}\DefaultIcon" "" "$\"$INSTDIR\${MAINBINARYNAME}.exe$\",0"
+    WriteRegStr SHCTX "Software\Classes\\{{protocol}}\shell\open\command" "" "$\"$INSTDIR\${MAINBINARYNAME}.exe$\" $\"%1$\""
+  {{/each}}
+
+  ; Create uninstaller
+  WriteUninstaller "$INSTDIR\uninstall.exe"
+
+  ; Save $INSTDIR in registry for future installations
+  WriteRegStr SHCTX "${MANUPRODUCTKEY}" "" $INSTDIR
+
+  !if "${INSTALLMODE}" == "both"
+    ; Save install mode to be selected by default for the next installation such as updating
+    ; or when uninstalling
+    WriteRegStr SHCTX "${UNINSTKEY}" $MultiUser.InstallMode 1
+  !endif
+
+  ; Remove old main binary if it doesn't match new main binary name
+  ReadRegStr $OldMainBinaryName SHCTX "${UNINSTKEY}" "MainBinaryName"
+  ${If} $OldMainBinaryName != ""
+  ${AndIf} $OldMainBinaryName != "${MAINBINARYNAME}.exe"
+    Delete "$INSTDIR\$OldMainBinaryName"
+  ${EndIf}
+
+  ; Save current MAINBINARYNAME for future updates
+  WriteRegStr SHCTX "${UNINSTKEY}" "MainBinaryName" "${MAINBINARYNAME}.exe"
+
+  ; Registry information for add/remove programs
+  WriteRegStr SHCTX "${UNINSTKEY}" "DisplayName" "${PRODUCTNAME}"
+  WriteRegStr SHCTX "${UNINSTKEY}" "DisplayIcon" "$\"$INSTDIR\${MAINBINARYNAME}.exe$\""
+  WriteRegStr SHCTX "${UNINSTKEY}" "DisplayVersion" "${VERSION}"
+  WriteRegStr SHCTX "${UNINSTKEY}" "Publisher" "${MANUFACTURER}"
+  WriteRegStr SHCTX "${UNINSTKEY}" "InstallLocation" "$\"$INSTDIR$\""
+  WriteRegStr SHCTX "${UNINSTKEY}" "UninstallString" "$\"$INSTDIR\uninstall.exe$\""
+  WriteRegDWORD SHCTX "${UNINSTKEY}" "NoModify" "1"
+  WriteRegDWORD SHCTX "${UNINSTKEY}" "NoRepair" "1"
+
+  ${GetSize} "$INSTDIR" "/M=uninstall.exe /S=0K /G=0" $0 $1 $2
+  IntOp $0 $0 + ${ESTIMATEDSIZE}
+  IntFmt $0 "0x%08X" $0
+  WriteRegDWORD SHCTX "${UNINSTKEY}" "EstimatedSize" "$0"
+
+  !if "${HOMEPAGE}" != ""
+    WriteRegStr SHCTX "${UNINSTKEY}" "URLInfoAbout" "${HOMEPAGE}"
+    WriteRegStr SHCTX "${UNINSTKEY}" "URLUpdateInfo" "${HOMEPAGE}"
+    WriteRegStr SHCTX "${UNINSTKEY}" "HelpLink" "${HOMEPAGE}"
+  !endif
+
+  ; Create start menu shortcut
+  !insertmacro MUI_STARTMENU_WRITE_BEGIN Application
+    Call CreateOrUpdateStartMenuShortcut
+  !insertmacro MUI_STARTMENU_WRITE_END
+
+  ; Create desktop shortcut for silent and passive installers
+  ; because finish page will be skipped
+  ${If} $PassiveMode = 1
+  ${OrIf} ${Silent}
+    Call CreateOrUpdateDesktopShortcut
+  ${EndIf}
+
+  !ifmacrodef NSIS_HOOK_POSTINSTALL
+    !insertmacro NSIS_HOOK_POSTINSTALL
+  !endif
+
+  ; Auto close this page for passive mode
+  ${If} $PassiveMode = 1
+    SetAutoClose true
+  ${EndIf}
+SectionEnd
+
+Function .onInstSuccess
+  ; Check for `/R` flag only in silent and passive installers because
+  ; GUI installer has a toggle for the user to (re)start the app
+  ${If} $PassiveMode = 1
+  ${OrIf} ${Silent}
+    ${GetOptions} $CMDLINE "/R" $R0
+    ${IfNot} ${Errors}
+      ${GetOptions} $CMDLINE "/ARGS" $R0
+      nsis_tauri_utils::RunAsUser "$INSTDIR\${MAINBINARYNAME}.exe" "$R0"
+    ${EndIf}
+  ${EndIf}
+FunctionEnd
+
+Function un.onInit
+  !insertmacro SetContext
+
+  !if "${INSTALLMODE}" == "both"
+    !insertmacro MULTIUSER_UNINIT
+  !endif
+
+  !insertmacro MUI_UNGETLANGUAGE
+
+  ${GetOptions} $CMDLINE "/P" $PassiveMode
+  ${IfNot} ${Errors}
+    StrCpy $PassiveMode 1
+  ${EndIf}
+
+  ${GetOptions} $CMDLINE "/UPDATE" $UpdateMode
+  ${IfNot} ${Errors}
+    StrCpy $UpdateMode 1
+  ${EndIf}
+FunctionEnd
+
+Section Uninstall
+
+  !ifmacrodef NSIS_HOOK_PREUNINSTALL
+    !insertmacro NSIS_HOOK_PREUNINSTALL
+  !endif
+
+  !insertmacro CheckIfAppIsRunning "${MAINBINARYNAME}.exe" "${PRODUCTNAME}"
+
+  ; Delete the app directory and its content from disk
+  ; Copy main executable
+  Delete "$INSTDIR\${MAINBINARYNAME}.exe"
+
+  ; Delete resources
+  {{#each resources}}
+    Delete "$INSTDIR\\{{this.[1]}}"
+  {{/each}}
+
+  ; Delete external binaries
+  {{#each binaries}}
+    Delete "$INSTDIR\\{{this}}"
+  {{/each}}
+
+  ; Delete app associations
+  {{#each file_associations as |association| ~}}
+    {{#each association.ext as |ext| ~}}
+      !insertmacro APP_UNASSOCIATE "{{ext}}" "{{or association.name ext}}"
+    {{/each}}
+  {{/each}}
+
+  ; Delete deep links
+  {{#each deep_link_protocols as |protocol| ~}}
+    ReadRegStr $R7 SHCTX "Software\Classes\\{{protocol}}\shell\open\command" ""
+    ${If} $R7 == "$\"$INSTDIR\${MAINBINARYNAME}.exe$\" $\"%1$\""
+      DeleteRegKey SHCTX "Software\Classes\\{{protocol}}"
+    ${EndIf}
+  {{/each}}
+
+
+  ; Delete uninstaller
+  Delete "$INSTDIR\uninstall.exe"
+
+  {{#each resources_ancestors}}
+  RMDir /REBOOTOK "$INSTDIR\\{{this}}"
+  {{/each}}
+  RMDir "$INSTDIR"
+
+  ; Remove shortcuts if not updating
+  ${If} $UpdateMode <> 1
+    !insertmacro DeleteAppUserModelId
+
+    ; Remove start menu shortcut
+    !insertmacro MUI_STARTMENU_GETFOLDER Application $AppStartMenuFolder
+    !insertmacro IsShortcutTarget "$SMPROGRAMS\$AppStartMenuFolder\${PRODUCTNAME}.lnk" "$INSTDIR\${MAINBINARYNAME}.exe"
+    Pop $0
+    ${If} $0 = 1
+      !insertmacro UnpinShortcut "$SMPROGRAMS\$AppStartMenuFolder\${PRODUCTNAME}.lnk"
+      Delete "$SMPROGRAMS\$AppStartMenuFolder\${PRODUCTNAME}.lnk"
+      RMDir "$SMPROGRAMS\$AppStartMenuFolder"
+    ${EndIf}
+    !insertmacro IsShortcutTarget "$SMPROGRAMS\${PRODUCTNAME}.lnk" "$INSTDIR\${MAINBINARYNAME}.exe"
+    Pop $0
+    ${If} $0 = 1
+      !insertmacro UnpinShortcut "$SMPROGRAMS\${PRODUCTNAME}.lnk"
+      Delete "$SMPROGRAMS\${PRODUCTNAME}.lnk"
+    ${EndIf}
+
+    ; Remove desktop shortcuts
+    !insertmacro IsShortcutTarget "$DESKTOP\${PRODUCTNAME}.lnk" "$INSTDIR\${MAINBINARYNAME}.exe"
+    Pop $0
+    ${If} $0 = 1
+      !insertmacro UnpinShortcut "$DESKTOP\${PRODUCTNAME}.lnk"
+      Delete "$DESKTOP\${PRODUCTNAME}.lnk"
+    ${EndIf}
+  ${EndIf}
+
+  ; Remove registry information for add/remove programs
+  !if "${INSTALLMODE}" == "both"
+    DeleteRegKey SHCTX "${UNINSTKEY}"
+  !else if "${INSTALLMODE}" == "perMachine"
+    DeleteRegKey HKLM "${UNINSTKEY}"
+  !else
+    DeleteRegKey HKCU "${UNINSTKEY}"
+  !endif
+
+  ; Removes the Autostart entry for ${PRODUCTNAME} from the HKCU Run key if it exists.
+  ; This ensures the program does not launch automatically after uninstallation if it exists.
+  ; If it doesn't exist, it does nothing.
+  ; We do this when not updating (to preserve the registry value on updates)
+  ${If} $UpdateMode <> 1
+    DeleteRegValue HKCU "Software\Microsoft\Windows\CurrentVersion\Run" "${PRODUCTNAME}"
+  ${EndIf}
+
+  ; Delete app data if the checkbox is selected
+  ; and if not updating
+  ${If} $DeleteAppDataCheckboxState = 1
+  ${AndIf} $UpdateMode <> 1
+    ; Clear the install location $INSTDIR from registry
+    DeleteRegKey SHCTX "${MANUPRODUCTKEY}"
+    DeleteRegKey /ifempty SHCTX "${MANUKEY}"
+
+    ; Clear the install language from registry
+    DeleteRegValue HKCU "${MANUPRODUCTKEY}" "Installer Language"
+    DeleteRegKey /ifempty HKCU "${MANUPRODUCTKEY}"
+    DeleteRegKey /ifempty HKCU "${MANUKEY}"
+
+    SetShellVarContext current
+    RmDir /r "$APPDATA\${BUNDLEID}"
+    RmDir /r "$LOCALAPPDATA\${BUNDLEID}"
+  ${EndIf}
+
+  !ifmacrodef NSIS_HOOK_POSTUNINSTALL
+    !insertmacro NSIS_HOOK_POSTUNINSTALL
+  !endif
+
+  ; Auto close if passive mode or updating
+  ${If} $PassiveMode = 1
+  ${OrIf} $UpdateMode = 1
+    SetAutoClose true
+  ${EndIf}
+SectionEnd
+
+Function RestorePreviousInstallLocation
+  ReadRegStr $4 SHCTX "${MANUPRODUCTKEY}" ""
+  StrCmp $4 "" +2 0
+    StrCpy $INSTDIR $4
+FunctionEnd
+
+Function Skip
+  Abort
+FunctionEnd
+
+Function SkipIfPassive
+  ${IfThen} $PassiveMode = 1  ${|} Abort ${|}
+FunctionEnd
+Function un.SkipIfPassive
+  ${IfThen} $PassiveMode = 1  ${|} Abort ${|}
+FunctionEnd
+
+Function CreateOrUpdateStartMenuShortcut
+  ; We used to use product name as MAINBINARYNAME
+  ; migrate old shortcuts to target the new MAINBINARYNAME
+  StrCpy $R0 0
+
+  !insertmacro IsShortcutTarget "$SMPROGRAMS\$AppStartMenuFolder\${PRODUCTNAME}.lnk" "$INSTDIR\$OldMainBinaryName"
+  Pop $0
+  ${If} $0 = 1
+    !insertmacro SetShortcutTarget "$SMPROGRAMS\$AppStartMenuFolder\${PRODUCTNAME}.lnk" "$INSTDIR\${MAINBINARYNAME}.exe"
+    StrCpy $R0 1
+  ${EndIf}
+
+  !insertmacro IsShortcutTarget "$SMPROGRAMS\${PRODUCTNAME}.lnk" "$INSTDIR\$OldMainBinaryName"
+  Pop $0
+  ${If} $0 = 1
+    !insertmacro SetShortcutTarget "$SMPROGRAMS\${PRODUCTNAME}.lnk" "$INSTDIR\${MAINBINARYNAME}.exe"
+    StrCpy $R0 1
+  ${EndIf}
+
+  ${If} $R0 = 1
+    Return
+  ${EndIf}
+
+  ; Skip creating shortcut if in update mode or no shortcut mode
+  ; but always create if migrating from wix
+  ${If} $WixMode = 0
+    ${If} $UpdateMode = 1
+    ${OrIf} $NoShortcutMode = 1
+      Return
+    ${EndIf}
+  ${EndIf}
+
+  !if "${STARTMENUFOLDER}" != ""
+    CreateDirectory "$SMPROGRAMS\$AppStartMenuFolder"
+    CreateShortcut "$SMPROGRAMS\$AppStartMenuFolder\${PRODUCTNAME}.lnk" "$INSTDIR\${MAINBINARYNAME}.exe"
+    !insertmacro SetLnkAppUserModelId "$SMPROGRAMS\$AppStartMenuFolder\${PRODUCTNAME}.lnk"
+  !else
+    CreateShortcut "$SMPROGRAMS\${PRODUCTNAME}.lnk" "$INSTDIR\${MAINBINARYNAME}.exe"
+    !insertmacro SetLnkAppUserModelId "$SMPROGRAMS\${PRODUCTNAME}.lnk"
+  !endif
+FunctionEnd
+
+Function CreateOrUpdateDesktopShortcut
+  ; We used to use product name as MAINBINARYNAME
+  ; migrate old shortcuts to target the new MAINBINARYNAME
+  !insertmacro IsShortcutTarget "$DESKTOP\${PRODUCTNAME}.lnk" "$INSTDIR\$OldMainBinaryName"
+  Pop $0
+  ${If} $0 = 1
+    !insertmacro SetShortcutTarget "$DESKTOP\${PRODUCTNAME}.lnk" "$INSTDIR\${MAINBINARYNAME}.exe"
+    Return
+  ${EndIf}
+
+  ; Skip creating shortcut if in update mode or no shortcut mode
+  ; but always create if migrating from wix
+  ${If} $WixMode = 0
+    ${If} $UpdateMode = 1
+    ${OrIf} $NoShortcutMode = 1
+      Return
+    ${EndIf}
+  ${EndIf}
+
+  CreateShortcut "$DESKTOP\${PRODUCTNAME}.lnk" "$INSTDIR\${MAINBINARYNAME}.exe"
+  !insertmacro SetLnkAppUserModelId "$DESKTOP\${PRODUCTNAME}.lnk"
+FunctionEnd

From 265d16e742a33e956a7176d5ba309e7691e0da87 Mon Sep 17 00:00:00 2001
From: Wasim Yousef Said <wasimysdev@gmail.com>
Date: Thu, 30 Apr 2026 18:40:40 +0200
Subject: [PATCH 184/216] Center Tauri windows and remove resize animation
 (#5235)

* Fix Tauri window centering and resize transition

* Guard Tauri window layout updates
---
 studio/frontend/src/app/provider.tsx | 165 ++++++++++++++++-----------
 studio/src-tauri/tauri.conf.json     |   1 +
 2 files changed, 101 insertions(+), 65 deletions(-)

diff --git a/studio/frontend/src/app/provider.tsx b/studio/frontend/src/app/provider.tsx
index e82f611552..c3a1c7dab0 100644
--- a/studio/frontend/src/app/provider.tsx
+++ b/studio/frontend/src/app/provider.tsx
@@ -10,7 +10,7 @@ import {
 } from "@/components/tauri/window-titlebar";
 import { Toaster } from "@/components/ui/sonner";
 import { getTauriAuthFailure, tauriAutoAuth } from "@/features/auth";
-import { useTauriBackend } from "@/hooks/use-tauri-backend";
+import { useTauriBackend, type BackendStatus } from "@/hooks/use-tauri-backend";
 import { useTauriUpdate } from "@/hooks/use-tauri-update";
 import { isTauri } from "@/lib/api-base";
 import { useRouterState } from "@tanstack/react-router";
@@ -25,67 +25,85 @@ interface AppProviderProps {
 // Tauri window helpers (only imported in Tauri mode)
 // ---------------------------------------------------------------------------
 
-async function showWindow(): Promise<void> {
+type TauriWindowMode = "setup" | "app";
+type WindowLayoutGuard = () => boolean;
+
+async function showSetupWindow(isCurrent: WindowLayoutGuard): Promise<void> {
   const { getCurrentWindow } = await import("@tauri-apps/api/window");
-  await getCurrentWindow().show();
-}
+  if (!isCurrent()) return;
 
-function easeOutQuart(t: number): number {
-  return 1 - (1 - t) ** 4;
-}
-
-async function animateToGoldenRatio(abortRef: { current: boolean }): Promise<void> {
-  const { getCurrentWindow, currentMonitor, LogicalSize } = await import("@tauri-apps/api/window");
   const win = getCurrentWindow();
-
-  // Ensure window is visible before resizing
+  if (!isCurrent()) return;
+  await win.center();
+  if (!isCurrent()) return;
   await win.show();
+}
 
+async function applyAppWindowLayout(isCurrent: WindowLayoutGuard): Promise<void> {
+  const { getCurrentWindow, currentMonitor, LogicalSize } = await import("@tauri-apps/api/window");
+  if (!isCurrent()) return;
+
+  const win = getCurrentWindow();
   const monitor = await currentMonitor();
-  if (!monitor) return;
+  if (!isCurrent()) return;
 
-  // Convert physical pixels to logical using scale factor
-  const scale = monitor.scaleFactor;
-  const screenW = monitor.size.width / scale;
-  const screenH = monitor.size.height / scale;
+  let finalW = 900;
+  let finalH = 600;
 
-  // Target: 75% of screen width, golden ratio height, capped at min 900x600
-  const targetW = Math.max(900, Math.round(screenW * 0.75));
-  const targetH = Math.max(600, Math.round(targetW / 1.618));
-  // Don't exceed screen height
-  const finalH = Math.min(targetH, Math.round(screenH * 0.85));
-  const finalW = targetW;
+  if (monitor) {
+    // Convert physical pixels to logical using scale factor
+    const scale = monitor.scaleFactor;
+    const screenW = monitor.size.width / scale;
+    const screenH = monitor.size.height / scale;
 
-  // Check reduced motion preference
-  const prefersReducedMotion = window.matchMedia("(prefers-reduced-motion: reduce)").matches;
-
-  if (prefersReducedMotion) {
-    await win.setSize(new LogicalSize(finalW, finalH));
-  } else {
-    // Read current size instead of hardcoding — stays correct if tauri.conf.json changes
-    const inner = await win.innerSize();
-    const factor = await win.scaleFactor();
-    const startW = Math.round(inner.width / factor);
-    const startH = Math.round(inner.height / factor);
-    const steps = 15;
-    const stepDuration = 23; // ~350ms total
-
-    for (let i = 1; i <= steps; i++) {
-      if (abortRef.current) return;
-      const t = easeOutQuart(i / steps);
-      const w = Math.round(startW + (finalW - startW) * t);
-      const h = Math.round(startH + (finalH - startH) * t);
-      await win.setSize(new LogicalSize(w, h));
-      await new Promise((r) => setTimeout(r, stepDuration));
-    }
+    // Target: 75% of screen width, golden ratio height, capped at min 900x600
+    finalW = Math.max(900, Math.round(screenW * 0.75));
+    const targetH = Math.max(600, Math.round(finalW / 1.618));
+    // Don't exceed screen height
+    finalH = Math.min(targetH, Math.round(screenH * 0.85));
   }
 
-  if (abortRef.current) return;
-
-  // Apply constraints and finalize
-  await win.setResizable(true);
+  // Apply constraints and finalize without animating through intermediate sizes
+  if (!isCurrent()) return;
+  await win.setSize(new LogicalSize(finalW, finalH));
+  if (!isCurrent()) return;
   await win.setSizeConstraints({ minWidth: 900, minHeight: 600 });
+  if (!isCurrent()) return;
+  await win.setResizable(true);
+  if (!isCurrent()) return;
   await win.center();
+  if (!isCurrent()) return;
+  await win.show();
+}
+
+async function showWindowFallback(): Promise<void> {
+  const { getCurrentWindow } = await import("@tauri-apps/api/window");
+  const win = getCurrentWindow();
+  await win.setResizable(true);
+  await win.show();
+}
+
+function getTauriWindowMode(
+  status: BackendStatus,
+  hasEnteredAppMode: boolean,
+): TauriWindowMode | null {
+  switch (status) {
+    case "checking":
+      return null;
+    case "not-installed":
+    case "installing":
+    case "install-error":
+    case "needs-elevation":
+    case "repairing":
+    case "repair-error":
+      return "setup";
+    case "starting":
+    case "running":
+    case "stopped":
+      return "app";
+    case "error":
+      return hasEnteredAppMode ? "app" : "setup";
+  }
 }
 
 // ---------------------------------------------------------------------------
@@ -143,30 +161,47 @@ function TauriWrapper({ children }: { children: ReactNode }) {
     startInstall, retry, retryInstall, approveElevation, copyDiagnostics,
   } = useTauriBackend();
 
-  const hasResized = useRef(false);
-  const abortRef = useRef(false);
+  const appliedWindowModeRef = useRef<TauriWindowMode | null>(null);
+  const hasEnteredAppModeRef = useRef(false);
+  const windowLayoutGenerationRef = useRef(0);
   const [desktopAuthReady, setDesktopAuthReady] = useState(!isTauri);
   const [desktopAuthRetry, setDesktopAuthRetry] = useState(0);
 
-  // Show the window once the frontend mounts (for pre-running states)
   useEffect(() => {
-    if (isTauri) void showWindow();
+    if (!isTauri) return;
+    return () => {
+      windowLayoutGenerationRef.current += 1;
+      appliedWindowModeRef.current = null;
+    };
   }, []);
 
-  // Animate resize when backend becomes ready
+  // Keep the Tauri window hidden during preflight, then show it centered in setup
+  // mode or apply the final app layout in one instant step.
   useEffect(() => {
-    if (status === "running" && !hasResized.current) {
-      hasResized.current = true;
-      abortRef.current = false;
-      animateToGoldenRatio(abortRef).catch(async () => {
-        // On failure, at minimum make the window resizable so user can fix manually
-        try {
-          const { getCurrentWindow } = await import("@tauri-apps/api/window");
-          await getCurrentWindow().setResizable(true);
-        } catch { /* swallow — window may still be functional */ }
-      });
+    if (!isTauri) return;
+
+    const nextMode = getTauriWindowMode(status, hasEnteredAppModeRef.current);
+    if (!nextMode) {
+      appliedWindowModeRef.current = null;
+      windowLayoutGenerationRef.current += 1;
+      return;
     }
-    return () => { abortRef.current = true; };
+    if (appliedWindowModeRef.current === nextMode) return;
+
+    appliedWindowModeRef.current = nextMode;
+    if (nextMode === "app") hasEnteredAppModeRef.current = true;
+
+    const layoutGeneration = windowLayoutGenerationRef.current + 1;
+    windowLayoutGenerationRef.current = layoutGeneration;
+    const isCurrent = () => windowLayoutGenerationRef.current === layoutGeneration;
+    const applyWindowMode = nextMode === "setup" ? showSetupWindow : applyAppWindowLayout;
+    applyWindowMode(isCurrent).catch(async () => {
+      if (!isCurrent()) return;
+      // On failure, at minimum make the window visible and resizable so user can fix manually.
+      try {
+        await showWindowFallback();
+      } catch { /* swallow — window may still be functional */ }
+    });
   }, [status]);
 
   useEffect(() => {
diff --git a/studio/src-tauri/tauri.conf.json b/studio/src-tauri/tauri.conf.json
index fce43889b5..eed7c1f352 100644
--- a/studio/src-tauri/tauri.conf.json
+++ b/studio/src-tauri/tauri.conf.json
@@ -24,6 +24,7 @@
         "title": "Unsloth Studio (Desktop)",
         "width": 690,
         "height": 480,
+        "center": true,
         "visible": false,
         "resizable": false
       }

From 329f99d3ad99ecb2b13be93958b845d4326fe153 Mon Sep 17 00:00:00 2001
From: Datta Nimmaturi <venkatadattasainimmaturi@gmail.com>
Date: Fri, 1 May 2026 10:00:21 +0530
Subject: [PATCH 185/216] MROPE for VLM GRPO (#5198)

* MROPE for VLM GRPO

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

---------

Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com>
Co-authored-by: pluesclues <136766175+pluesclues@users.noreply.github.com>
---
 unsloth/models/rl_replacements.py | 91 ++++++++++++++++++++++---------
 1 file changed, 66 insertions(+), 25 deletions(-)

diff --git a/unsloth/models/rl_replacements.py b/unsloth/models/rl_replacements.py
index 5d2c4151cf..0f10847282 100755
--- a/unsloth/models/rl_replacements.py
+++ b/unsloth/models/rl_replacements.py
@@ -30,6 +30,8 @@ from unsloth_zoo.rl_replacements import (
     RL_REPLACEMENTS,
     left_pack_padding,
     chunked_selective_log_softmax,
+    _unsloth_get_mm_token_id,
+    _unsloth_fix_mm_token_type_ids,
 )
 from unsloth_zoo.utils import Version
 from trl import __version__ as trl_version_raw
@@ -595,6 +597,23 @@ def grpo_trainer__generate_single_turn(function_name, function):
     ]:
         function = re.sub(pattern, "", function)
 
+    string_to_find = (
+        "            generate_inputs = super()._prepare_inputs(generate_inputs)"
+    )
+    replacement_string = (
+        string_to_find
+        + """
+            if "mm_token_type_ids" in generate_inputs or "image_grid_thw" in generate_inputs:
+                mm_token_type_ids = _unsloth_fix_mm_token_type_ids(
+                    self.processing_class,
+                    generate_inputs["input_ids"],
+                    generate_inputs.get("mm_token_type_ids", None),
+                )
+                if mm_token_type_ids is not None:
+                    generate_inputs["mm_token_type_ids"] = mm_token_type_ids"""
+    )
+    function = function.replace(string_to_find, replacement_string)
+
     return function
 
 
@@ -814,36 +833,45 @@ def grpo_trainer__generate_and_score_completions(function_name, function):
 
         function = patched
 
-    # Transformers 5.x: Extend mm_token_type_ids for completion tokens (Qwen3VL M-RoPE).
-    # TRL handles token_type_ids but not mm_token_type_ids.
-    _tt_search = (
-        'if "token_type_ids" in forward_kwargs:\n'
-        '            token_type_ids = forward_kwargs["token_type_ids"]\n'
-        '            forward_kwargs["token_type_ids"] = torch.cat(\n'
-        "                [token_type_ids, token_type_ids.new_zeros(completion_ids.shape)], dim=1\n"
-        "            )"
-    )
-    _tt_replace = (
-        _tt_search + "\n"
-        '        if "mm_token_type_ids" in forward_kwargs:\n'
-        '            mm_tti = forward_kwargs["mm_token_type_ids"]\n'
-        '            forward_kwargs["mm_token_type_ids"] = torch.cat(\n'
-        "                [mm_tti, mm_tti.new_zeros(completion_ids.shape)], dim=1\n"
-        "            )"
-    )
-    function = function.replace(_tt_search, _tt_replace)
+    _mm_alignment = """
+        if "mm_token_type_ids" in forward_kwargs or "image_grid_thw" in forward_kwargs:
+            _mm_token_type_ids = _unsloth_fix_mm_token_type_ids(
+                self.processing_class,
+                prompt_completion_ids,
+                forward_kwargs.get("mm_token_type_ids", None),
+                completion_ids = completion_ids,
+            )
+            if _mm_token_type_ids is not None:
+                forward_kwargs["mm_token_type_ids"] = _mm_token_type_ids
+"""
+    _tool_image_marker = "        # For VLM tool images: build token type IDs from the full prompt_completion_ids."
+    if _tool_image_marker in function:
+        function = function.replace(
+            _tool_image_marker, _mm_alignment + "\n" + _tool_image_marker
+        )
+    else:
+        _tt_search = (
+            'if "token_type_ids" in forward_kwargs:\n'
+            '            token_type_ids = forward_kwargs["token_type_ids"]\n'
+            '            forward_kwargs["token_type_ids"] = torch.cat(\n'
+            "                [token_type_ids, token_type_ids.new_zeros(completion_ids.shape)], dim=1\n"
+            "            )"
+        )
+        function = function.replace(
+            _tt_search, _tt_search + "\n" + _mm_alignment.rstrip()
+        )
 
-    # Save mm_token_type_ids to output dict alongside token_type_ids
     _save_search = (
         'if "token_type_ids" in forward_kwargs:\n'
         '            output["token_type_ids"] = forward_kwargs["token_type_ids"]'
     )
-    _save_replace = (
-        _save_search + "\n"
-        '        if "mm_token_type_ids" in forward_kwargs:\n'
-        '            output["mm_token_type_ids"] = forward_kwargs["mm_token_type_ids"]'
-    )
-    function = function.replace(_save_search, _save_replace)
+    if 'output["mm_token_type_ids"]' not in function:
+        _save_replace = (
+            _save_search + "\n"
+            '        if "mm_token_type_ids" in forward_kwargs:\n'
+            '            output["mm_token_type_ids"] = forward_kwargs["mm_token_type_ids"]'
+        )
+        function = function.replace(_save_search, _save_replace)
 
     return function
 
@@ -1020,6 +1048,10 @@ def grpo_trainer__get_per_token_logps_and_entropies(function_name, function):
             # Transformers 5.x needs token_type_ids/mm_token_type_ids for some vision models
             token_type_ids = kwargs.get("token_type_ids", None)
             mm_token_type_ids = kwargs.get("mm_token_type_ids", None)
+            if mm_token_type_ids is not None or image_grid_thw is not None:
+                mm_token_type_ids = _unsloth_fix_mm_token_type_ids(
+                    self.processing_class, input_ids, mm_token_type_ids
+                )
 
             unwrapped_model = self.accelerator.unwrap_model(
                 model, keep_fp32_wrapper = False
@@ -1306,6 +1338,8 @@ grpo_update_SamplingParams = RL_REPLACEMENTS["grpo_update_SamplingParams"]
 RL_PRE_ITEMS["grpo_trainer"].append(
     inspect.getsource(_unsloth_get_final_logit_softcapping)
 )
+RL_PRE_ITEMS["grpo_trainer"].append(inspect.getsource(_unsloth_get_mm_token_id))
+RL_PRE_ITEMS["grpo_trainer"].append(inspect.getsource(_unsloth_fix_mm_token_type_ids))
 RL_PRE_ITEMS["grpo_trainer"].append(inspect.getsource(grpo_compute_loss))
 RL_PRE_ITEMS["grpo_trainer"].append(inspect.getsource(UnslothEfficientGRPO))
 RL_PRE_ITEMS["grpo_trainer"].append(inspect.getsource(grpo_accumulated_loss))
@@ -1352,6 +1386,13 @@ def grpo_trainer_compute_loss(function_name, function):
         input_ids = torch.cat([prompt_ids, completion_ids], dim = 1)
         bsz, qlen = input_ids.shape
         attention_mask = torch.cat([prompt_mask, completion_mask], dim = 1)
+        if mm_token_type_ids is not None or image_grid_thw is not None:
+            mm_token_type_ids = _unsloth_fix_mm_token_type_ids(
+                self.processing_class,
+                input_ids,
+                mm_token_type_ids,
+                completion_ids = completion_ids,
+            )
         # attention_mask = None
         logits_to_keep = completion_ids.size(
             1

From e4e89f41c15add339047cb9e9efcaa88da2128c8 Mon Sep 17 00:00:00 2001
From: Roland Tannous <115670425+rolandtannous@users.noreply.github.com>
Date: Fri, 1 May 2026 11:15:22 +0400
Subject: [PATCH 186/216] install: overlay unsloth-zoo from git main on --local
 (#5242)

When --local is passed, also overlay unsloth-zoo from the upstream main
branch (--no-deps --reinstall-package) on top of the PyPI install. This
keeps the editable unsloth checkout paired with the latest unreleased
unsloth-zoo, mirroring the existing -e $_REPO_ROOT --no-deps overlay.

Applied to all four --local paths in install.sh (migrated, fresh no-torch,
fresh with-torch, auto-torch fallback) and the three corresponding paths
in install.ps1.
---
 install.ps1 | 18 ++++++++++++++++++
 install.sh  | 16 ++++++++++++++++
 2 files changed, 34 insertions(+)

diff --git a/install.ps1 b/install.ps1
index dcd86eb019..ab43bcdc52 100644
--- a/install.ps1
+++ b/install.ps1
@@ -1089,6 +1089,12 @@ shell.Run cmd, 0, False
                 Write-Host "[ERROR] Failed to overlay local repo (exit code $overlayExit)" -ForegroundColor Red
                 return (Exit-InstallFailure "Failed to overlay local repo (exit code $overlayExit)" $overlayExit)
             }
+            substep "overlaying unsloth-zoo from git main..."
+            $zooOverlayExit = Invoke-InstallCommand { uv pip install --python $VenvPython --no-deps --reinstall-package unsloth-zoo "unsloth-zoo @ git+https://github.com/unslothai/unsloth-zoo" }
+            if ($zooOverlayExit -ne 0) {
+                Write-Host "[ERROR] Failed to overlay unsloth-zoo (exit code $zooOverlayExit)" -ForegroundColor Red
+                return (Exit-InstallFailure "Failed to overlay unsloth-zoo (exit code $zooOverlayExit)" $zooOverlayExit)
+            }
         }
     } elseif ($TorchIndexUrl) {
         if ($SkipTorch) {
@@ -1132,6 +1138,12 @@ shell.Run cmd, 0, False
                 Write-Host "[ERROR] Failed to overlay local repo (exit code $overlayExit)" -ForegroundColor Red
                 return (Exit-InstallFailure "Failed to overlay local repo (exit code $overlayExit)" $overlayExit)
             }
+            substep "overlaying unsloth-zoo from git main..."
+            $zooOverlayExit = Invoke-InstallCommand { uv pip install --python $VenvPython --no-deps --reinstall-package unsloth-zoo "unsloth-zoo @ git+https://github.com/unslothai/unsloth-zoo" }
+            if ($zooOverlayExit -ne 0) {
+                Write-Host "[ERROR] Failed to overlay unsloth-zoo (exit code $zooOverlayExit)" -ForegroundColor Red
+                return (Exit-InstallFailure "Failed to overlay unsloth-zoo (exit code $zooOverlayExit)" $zooOverlayExit)
+            }
         }
     } else {
         # Fallback: GPU detection failed to produce a URL -- let uv resolve torch
@@ -1149,6 +1161,12 @@ shell.Run cmd, 0, False
                 Write-Host "[ERROR] Failed to overlay local repo (exit code $overlayExit)" -ForegroundColor Red
                 return (Exit-InstallFailure "Failed to overlay local repo (exit code $overlayExit)" $overlayExit)
             }
+            substep "overlaying unsloth-zoo from git main..."
+            $zooOverlayExit = Invoke-InstallCommand { uv pip install --python $VenvPython --no-deps --reinstall-package unsloth-zoo "unsloth-zoo @ git+https://github.com/unslothai/unsloth-zoo" }
+            if ($zooOverlayExit -ne 0) {
+                Write-Host "[ERROR] Failed to overlay unsloth-zoo (exit code $zooOverlayExit)" -ForegroundColor Red
+                return (Exit-InstallFailure "Failed to overlay unsloth-zoo (exit code $zooOverlayExit)" $zooOverlayExit)
+            }
         } else {
             $baseInstallExit = Invoke-InstallCommand { uv pip install --python $VenvPython --torch-backend=auto -- "$PackageName" }
             if ($baseInstallExit -ne 0) {
diff --git a/install.sh b/install.sh
index 18c6bba589..a9861a108f 100755
--- a/install.sh
+++ b/install.sh
@@ -1501,6 +1501,10 @@ if [ "$_MIGRATED" = true ]; then
     if [ "$STUDIO_LOCAL_INSTALL" = true ]; then
         substep "overlaying local repo (editable)..."
         run_install_cmd "overlay local repo" uv pip install --python "$_VENV_PY" -e "$_REPO_ROOT" --no-deps
+        substep "overlaying unsloth-zoo from git main..."
+        run_install_cmd "overlay unsloth-zoo (git main)" uv pip install --python "$_VENV_PY" \
+            --no-deps --reinstall-package unsloth-zoo \
+            "unsloth-zoo @ git+https://github.com/unslothai/unsloth-zoo"
     fi
     # AMD ROCm: install bitsandbytes even in migrated environments so
     # existing ROCm installs gain the AMD bitsandbytes build without a
@@ -1668,12 +1672,20 @@ elif [ -n "$TORCH_INDEX_URL" ]; then
         if [ "$STUDIO_LOCAL_INSTALL" = true ]; then
             substep "overlaying local repo (editable)..."
             run_install_cmd "overlay local repo" uv pip install --python "$_VENV_PY" -e "$_REPO_ROOT" --no-deps
+            substep "overlaying unsloth-zoo from git main..."
+            run_install_cmd "overlay unsloth-zoo (git main)" uv pip install --python "$_VENV_PY" \
+                --no-deps --reinstall-package unsloth-zoo \
+                "unsloth-zoo @ git+https://github.com/unslothai/unsloth-zoo"
         fi
     elif [ "$STUDIO_LOCAL_INSTALL" = true ]; then
         run_install_cmd "install unsloth (local)" uv pip install --python "$_VENV_PY" \
             --upgrade-package unsloth "unsloth>=2026.4.8" unsloth-zoo
         substep "overlaying local repo (editable)..."
         run_install_cmd "overlay local repo" uv pip install --python "$_VENV_PY" -e "$_REPO_ROOT" --no-deps
+        substep "overlaying unsloth-zoo from git main..."
+        run_install_cmd "overlay unsloth-zoo (git main)" uv pip install --python "$_VENV_PY" \
+            --no-deps --reinstall-package unsloth-zoo \
+            "unsloth-zoo @ git+https://github.com/unslothai/unsloth-zoo"
     else
         run_install_cmd "install unsloth" uv pip install --python "$_VENV_PY" \
             --upgrade-package unsloth -- "$PACKAGE_NAME"
@@ -1702,6 +1714,10 @@ else
         run_install_cmd "install unsloth (auto torch backend)" uv pip install --python "$_VENV_PY" unsloth-zoo "unsloth>=2026.4.8" --torch-backend=auto
         substep "overlaying local repo (editable)..."
         run_install_cmd "overlay local repo" uv pip install --python "$_VENV_PY" -e "$_REPO_ROOT" --no-deps
+        substep "overlaying unsloth-zoo from git main..."
+        run_install_cmd "overlay unsloth-zoo (git main)" uv pip install --python "$_VENV_PY" \
+            --no-deps --reinstall-package unsloth-zoo \
+            "unsloth-zoo @ git+https://github.com/unslothai/unsloth-zoo"
     else
         run_install_cmd "install unsloth (auto torch backend)" uv pip install --python "$_VENV_PY" --torch-backend=auto -- "$PACKAGE_NAME"
     fi

From 05f46686de7f6bff4f7afd922e027e0438252ff2 Mon Sep 17 00:00:00 2001
From: Lee Jackson <130007945+Imagineer99@users.noreply.github.com>
Date: Fri, 1 May 2026 16:19:09 +0100
Subject: [PATCH 187/216] Studio: Fix chat template disappearing after browser
 refresh (#5209)

* fix: preserve chat template on refresh

* chore: simplify chat template status lookup
---
 studio/backend/core/inference/llama_cpp.py           | 12 ++++++++++++
 studio/backend/models/inference.py                   |  4 ++++
 studio/backend/routes/inference.py                   |  9 +++++++++
 studio/backend/tests/test_native_context_length.py   | 12 ++++++++++++
 .../features/chat/hooks/use-chat-model-runtime.ts    |  5 +++++
 studio/frontend/src/features/chat/types/api.ts       |  1 +
 6 files changed, 43 insertions(+)

diff --git a/studio/backend/core/inference/llama_cpp.py b/studio/backend/core/inference/llama_cpp.py
index c1a37e2fa4..ae8224d424 100644
--- a/studio/backend/core/inference/llama_cpp.py
+++ b/studio/backend/core/inference/llama_cpp.py
@@ -1701,6 +1701,18 @@ class LlamaCppBackend:
             if chat_template_override:
                 import tempfile
 
+                self._chat_template = chat_template_override
+                flags = detect_reasoning_flags(
+                    self._chat_template,
+                    self._model_identifier,
+                    log_source = "GGUF chat template override",
+                )
+                self._supports_reasoning = flags["supports_reasoning"]
+                self._reasoning_style = flags["reasoning_style"]
+                self._reasoning_always_on = flags["reasoning_always_on"]
+                self._supports_preserve_thinking = flags["supports_preserve_thinking"]
+                self._supports_tools = flags["supports_tools"]
+
                 self._chat_template_file = tempfile.NamedTemporaryFile(
                     mode = "w",
                     suffix = ".jinja",
diff --git a/studio/backend/models/inference.py b/studio/backend/models/inference.py
index eb2bb9c5ce..027f3313a1 100644
--- a/studio/backend/models/inference.py
+++ b/studio/backend/models/inference.py
@@ -283,6 +283,10 @@ class InferenceStatusResponse(BaseModel):
     supports_tools: bool = Field(
         False, description = "Whether the active model supports tool calling"
     )
+    chat_template: Optional[str] = Field(
+        None,
+        description = "Jinja2 chat template string for the active model",
+    )
     context_length: Optional[int] = Field(
         None, description = "Context length of the active model"
     )
diff --git a/studio/backend/routes/inference.py b/studio/backend/routes/inference.py
index b13eb08967..53dd851666 100644
--- a/studio/backend/routes/inference.py
+++ b/studio/backend/routes/inference.py
@@ -985,6 +985,7 @@ async def get_status(
                 reasoning_always_on = llama_backend.reasoning_always_on,
                 supports_preserve_thinking = llama_backend.supports_preserve_thinking,
                 supports_tools = llama_backend.supports_tools,
+                chat_template = llama_backend.chat_template,
                 context_length = llama_backend.context_length,
                 max_context_length = llama_backend.max_context_length,
                 native_context_length = llama_backend.native_context_length,
@@ -998,12 +999,19 @@ async def get_status(
         is_audio = False
         audio_type = None
         has_audio_input = False
+        model_info = {}
         if backend.active_model_name:
             model_info = backend.models.get(backend.active_model_name, {})
             is_vision = model_info.get("is_vision", False)
             is_audio = model_info.get("is_audio", False)
             audio_type = model_info.get("audio_type")
             has_audio_input = model_info.get("has_audio_input", False)
+        chat_template_info = model_info.get("chat_template_info", {})
+        chat_template = (
+            chat_template_info.get("template")
+            if isinstance(chat_template_info, dict)
+            else None
+        )
 
         # Non-GGUF: only gpt-oss Harmony is wired through the transformers
         # generation path. Other template-level reasoning / tool kwargs
@@ -1041,6 +1049,7 @@ async def get_status(
             reasoning_always_on = False,
             supports_preserve_thinking = False,
             supports_tools = False,
+            chat_template = chat_template,
         )
 
     except Exception as e:
diff --git a/studio/backend/tests/test_native_context_length.py b/studio/backend/tests/test_native_context_length.py
index 7c69e56f89..60622c776d 100644
--- a/studio/backend/tests/test_native_context_length.py
+++ b/studio/backend/tests/test_native_context_length.py
@@ -320,11 +320,23 @@ class TestPydanticModels:
         """Field exists in InferenceStatusResponse.model_fields."""
         assert "native_context_length" in InferenceStatusResponse.model_fields
 
+    def test_status_response_has_chat_template_field(self):
+        """Status includes chat_template so the UI can rehydrate after refresh."""
+        assert "chat_template" in InferenceStatusResponse.model_fields
+
     def test_status_response_defaults_none(self):
         """Omitting native_context_length defaults to None."""
         resp = InferenceStatusResponse()
         assert resp.native_context_length is None
 
+    def test_status_response_chat_template_roundtrip(self):
+        """chat_template serializes and validates as part of status."""
+        resp = InferenceStatusResponse(chat_template = "{{ messages }}")
+        roundtripped = InferenceStatusResponse.model_validate_json(
+            resp.model_dump_json()
+        )
+        assert roundtripped.chat_template == "{{ messages }}"
+
     def test_roundtrip_preserves_value(self):
         """model_validate_json(model_dump_json()) round-trips."""
         resp = LoadResponse(
diff --git a/studio/frontend/src/features/chat/hooks/use-chat-model-runtime.ts b/studio/frontend/src/features/chat/hooks/use-chat-model-runtime.ts
index cfac6dc8cf..d9e817678b 100644
--- a/studio/frontend/src/features/chat/hooks/use-chat-model-runtime.ts
+++ b/studio/frontend/src/features/chat/hooks/use-chat-model-runtime.ts
@@ -266,6 +266,10 @@ export function useChatModelRuntime() {
           ? (statusRes.native_context_length ?? null)
           : null;
         const currentSpecType = normalizeSpeculativeType(statusRes.speculative_type);
+        const nextDefaultChatTemplate =
+          statusRes.chat_template === undefined
+            ? useChatRuntimeStore.getState().defaultChatTemplate
+            : statusRes.chat_template;
         useChatRuntimeStore.setState({
           supportsReasoning,
           reasoningAlwaysOn,
@@ -282,6 +286,7 @@ export function useChatModelRuntime() {
           ggufNativeContextLength,
           modelRequiresTrustRemoteCode:
             statusRes.requires_trust_remote_code ?? false,
+          defaultChatTemplate: nextDefaultChatTemplate,
           speculativeType: currentSpecType,
           loadedSpeculativeType: currentSpecType,
         });
diff --git a/studio/frontend/src/features/chat/types/api.ts b/studio/frontend/src/features/chat/types/api.ts
index 25957f4a7b..15d0457722 100644
--- a/studio/frontend/src/features/chat/types/api.ts
+++ b/studio/frontend/src/features/chat/types/api.ts
@@ -129,6 +129,7 @@ export interface InferenceStatusResponse {
   reasoning_always_on?: boolean;
   supports_preserve_thinking?: boolean;
   supports_tools?: boolean;
+  chat_template?: string | null;
   context_length?: number | null;
   max_context_length?: number | null;
   native_context_length?: number | null;

From 5262d93b5875caf15e86e399a80f6a1118281b6b Mon Sep 17 00:00:00 2001
From: Roland Tannous <115670425+rolandtannous@users.noreply.github.com>
Date: Sat, 2 May 2026 08:51:56 +0400
Subject: [PATCH 188/216] studio: add --local to setup.sh + overlay unsloth-zoo
 from git main (#5252)

* studio: add --local to setup.sh + overlay unsloth-zoo from git main

setup.sh now accepts --local, which exports STUDIO_LOCAL_INSTALL=1 and
STUDIO_LOCAL_REPO=$REPO_ROOT. install_python_stack.py overlays unsloth-zoo
from git main on top of the editable unsloth checkout in both local_repo
branches (no-torch and with-torch).

The Colab notebook now invokes ./studio/setup.sh --local so the cloned
repo is used in editable mode and unsloth-zoo tracks main, matching the
behavior of install.sh --local on a VM. install.sh --local is unchanged:
it still sets SKIP_STUDIO_BASE=1, which short-circuits the local_repo
branches in install_python_stack.py, so the overlay is not run twice.

* studio: make --local overlays visible + guard empty arg parsing

- setup.sh: gate the --local flag loop on $# > 0 (defensive against any
  shell that surfaces unset $@ under set -u) and emit a substep when local
  mode is detected so the user can confirm the flag was parsed.
- install_python_stack.py: emit explicit _step lines before each overlay
  pip_install in both local_repo branches so overlays appear in the static
  log instead of being overwritten by the in-place progress bar.
---
 studio/Unsloth_Studio_Colab.ipynb |  2 +-
 studio/install_python_stack.py    | 20 ++++++++++++++++++++
 studio/setup.sh                   | 18 ++++++++++++++++++
 3 files changed, 39 insertions(+), 1 deletion(-)

diff --git a/studio/Unsloth_Studio_Colab.ipynb b/studio/Unsloth_Studio_Colab.ipynb
index 46e2067ba7..c3aec04820 100644
--- a/studio/Unsloth_Studio_Colab.ipynb
+++ b/studio/Unsloth_Studio_Colab.ipynb
@@ -64,7 +64,7 @@
     "id": "27e68f91"
    },
    "outputs": [],
-   "source": "!git clone --depth 1 --branch main https://github.com/unslothai/unsloth.git\n%cd /content/unsloth\n!chmod +x studio/setup.sh && ./studio/setup.sh"
+   "source": "!git clone --depth 1 --branch main https://github.com/unslothai/unsloth.git\n%cd /content/unsloth\n!chmod +x studio/setup.sh && ./studio/setup.sh --local"
   },
   {
    "cell_type": "markdown",
diff --git a/studio/install_python_stack.py b/studio/install_python_stack.py
index 8ac59d389d..25d950fd3e 100644
--- a/studio/install_python_stack.py
+++ b/studio/install_python_stack.py
@@ -972,6 +972,7 @@ def install_python_stack() -> int:
             req = REQ_ROOT / "no-torch-runtime.txt",
         )
         if local_repo:
+            _step(_LABEL, f"overlaying local repo (editable): {local_repo}")
             pip_install(
                 "Overlaying local repo (editable)",
                 "--no-cache-dir",
@@ -980,6 +981,15 @@ def install_python_stack() -> int:
                 local_repo,
                 constrain = False,
             )
+            _step(_LABEL, "overlaying unsloth-zoo from git main")
+            pip_install(
+                "Overlaying unsloth-zoo from git main",
+                "--no-cache-dir",
+                "--no-deps",
+                "--force-reinstall",
+                "unsloth-zoo @ git+https://github.com/unslothai/unsloth-zoo",
+                constrain = False,
+            )
     elif local_repo:
         # Local dev install: update deps from base.txt, then overlay the
         # local checkout as an editable install (--no-deps so torch is
@@ -994,6 +1004,7 @@ def install_python_stack() -> int:
             "unsloth-zoo",
             req = REQ_ROOT / "base.txt",
         )
+        _step(_LABEL, f"overlaying local repo (editable): {local_repo}")
         pip_install(
             "Overlaying local repo (editable)",
             "--no-cache-dir",
@@ -1002,6 +1013,15 @@ def install_python_stack() -> int:
             local_repo,
             constrain = False,
         )
+        _step(_LABEL, "overlaying unsloth-zoo from git main")
+        pip_install(
+            "Overlaying unsloth-zoo from git main",
+            "--no-cache-dir",
+            "--no-deps",
+            "--force-reinstall",
+            "unsloth-zoo @ git+https://github.com/unslothai/unsloth-zoo",
+            constrain = False,
+        )
     elif package_name != "unsloth":
         # Custom package name (e.g. roland-sloth for testing) — install directly
         _progress("base packages")
diff --git a/studio/setup.sh b/studio/setup.sh
index 142d253554..9575f143d5 100755
--- a/studio/setup.sh
+++ b/studio/setup.sh
@@ -8,6 +8,21 @@ SCRIPT_DIR="$(cd "$(dirname "${BASH_SOURCE[0]}")" && pwd)"
 REPO_ROOT="$(cd "$SCRIPT_DIR/.." && pwd)"
 RULE=$(printf '\342\224\200%.0s' {1..52})
 
+# ── Parse flags ──
+# --local: install from the local repo checkout (overlays unsloth as editable
+# and unsloth-zoo from git main). Mirrors install.sh --local for the Colab
+# path that runs setup.sh directly without going through install.sh.
+if [ "$#" -gt 0 ]; then
+    for _arg in "$@"; do
+        case "$_arg" in
+            --local)
+                export STUDIO_LOCAL_INSTALL=1
+                export STUDIO_LOCAL_REPO="$REPO_ROOT"
+                ;;
+        esac
+    done
+fi
+
 # ── Maintainer-editable defaults ──────────────────────────────────────────
 # Change these in the GitHub-hosted script so all users get updated defaults.
 # User environment variables always override these baked-in values.
@@ -170,6 +185,9 @@ _LLAMA_ONLY="${UNSLOTH_STUDIO_LLAMA_ONLY:-0}"
 if [ "$_LLAMA_ONLY" = "1" ]; then
     substep "llama.cpp only mode"
 fi
+if [ "${STUDIO_LOCAL_INSTALL:-0}" = "1" ]; then
+    substep "local mode: overlaying $REPO_ROOT (editable) + unsloth-zoo from git main"
+fi
 # ── Clean up stale caches ──
 rm -rf "$REPO_ROOT/unsloth_compiled_cache"
 rm -rf "$SCRIPT_DIR/backend/unsloth_compiled_cache"

From 7d227ed708882cc5f045fa1b30b58026a5490ad5 Mon Sep 17 00:00:00 2001
From: DoubleMathew <mmathew23@gmail.com>
Date: Sat, 2 May 2026 00:42:19 -0500
Subject: [PATCH 189/216] Fix/windowsprebuilt (#5241)

* update prebuilt logic

* Add test case

---------

Co-authored-by: Roland Tannous <115670425+rolandtannous@users.noreply.github.com>
---
 studio/install_llama_prebuilt.py             |  4 ++--
 tests/studio/install/test_selection_logic.py | 17 +++++++++++++++++
 2 files changed, 19 insertions(+), 2 deletions(-)

diff --git a/studio/install_llama_prebuilt.py b/studio/install_llama_prebuilt.py
index 23e6b7c0e4..1aa3e501f8 100755
--- a/studio/install_llama_prebuilt.py
+++ b/studio/install_llama_prebuilt.py
@@ -2662,7 +2662,7 @@ def pick_windows_cuda_runtime(host: HostInfo) -> str | None:
     if not host.driver_cuda_version:
         return None
     major, minor = host.driver_cuda_version
-    if major > 13 or (major == 13 and minor >= 1):
+    if major > 13 or (major == 13):  # and minor >= 1):
         return "13.1"
     if major > 12 or (major == 12 and minor >= 4):
         return "12.4"
@@ -4470,7 +4470,7 @@ def validate_server(
                     ),
                     **windows_hidden_subprocess_kwargs(),
                 )
-                deadline = time.time() + 20
+                deadline = time.time() + 60
                 startup_started = time.time()
                 response_body = ""
                 last_error: Exception | None = None
diff --git a/tests/studio/install/test_selection_logic.py b/tests/studio/install/test_selection_logic.py
index 6a2e367dc8..5150d2d2ab 100644
--- a/tests/studio/install/test_selection_logic.py
+++ b/tests/studio/install/test_selection_logic.py
@@ -420,6 +420,10 @@ class TestPickWindowsCudaRuntime:
         host = make_host(driver_cuda_version = (13, 1))
         assert pick_windows_cuda_runtime(host) == "13.1"
 
+    def test_driver_13_0_uses_cuda13_line(self):
+        host = make_host(driver_cuda_version = (13, 0))
+        assert pick_windows_cuda_runtime(host) == "13.1"
+
 
 class TestCompatibleWindowsRuntimeLines:
     def test_no_driver(self):
@@ -434,6 +438,10 @@ class TestCompatibleWindowsRuntimeLines:
         host = make_host(driver_cuda_version = (13, 1))
         assert compatible_windows_runtime_lines(host) == ["cuda13", "cuda12"]
 
+    def test_driver_13_0_uses_cuda13_line(self):
+        host = make_host(driver_cuda_version = (13, 0))
+        assert compatible_windows_runtime_lines(host) == ["cuda13", "cuda12"]
+
 
 # ===========================================================================
 # H. runtime_line_from_cuda_version
@@ -1769,6 +1777,15 @@ class TestWindowsCudaAttempts:
         assert result[0].runtime_line == "cuda13"
         assert result[1].runtime_line == "cuda12"
 
+    def test_driver_13_0_cuda13_dlls_selects_cuda13_asset(self, monkeypatch):
+        mock_windows_runtime(monkeypatch, ["cuda13"])
+        host = make_host(system = "Windows", machine = "AMD64", driver_cuda_version = (13, 0))
+        assets = self._upstream("13.1", "12.4")
+        result = windows_cuda_attempts(host, self.TAG, assets, None)
+        assert len(result) == 1
+        assert result[0].runtime_line == "cuda13"
+        assert result[0].name == f"llama-{self.TAG}-bin-win-cuda-13.1-x64.zip"
+
     def test_preferred_reorders(self, monkeypatch):
         mock_windows_runtime(monkeypatch, ["cuda13", "cuda12"])
         host = make_host(system = "Windows", machine = "AMD64", driver_cuda_version = (13, 1))

From 874e4605ef05db6fd4f49600192d697a05099f95 Mon Sep 17 00:00:00 2001
From: Lee Jackson <130007945+Imagineer99@users.noreply.github.com>
Date: Sat, 2 May 2026 17:55:24 +0100
Subject: [PATCH 190/216] Studio: Add dataset upload dropzone and update
 preserve think copy (#5253)

* fix: improve training dataset upload affordance

* chore: update preserve think

* fix: guard dataset dropzone drag target
---
 .../src/components/assistant-ui/thread.tsx    |   4 +-
 .../src/features/chat/shared-composer.tsx     |   4 +-
 .../studio/sections/dataset-section.tsx       | 112 +++++++++++++++---
 3 files changed, 102 insertions(+), 18 deletions(-)

diff --git a/studio/frontend/src/components/assistant-ui/thread.tsx b/studio/frontend/src/components/assistant-ui/thread.tsx
index 3f67b11b51..d32f698384 100644
--- a/studio/frontend/src/components/assistant-ui/thread.tsx
+++ b/studio/frontend/src/components/assistant-ui/thread.tsx
@@ -485,7 +485,7 @@ const PreserveThinkingToggle: FC = () => {
             : "bg-muted text-muted-foreground hover:bg-muted-foreground/15",
       )}
       aria-label={
-        preserveThinking ? "Disable preserve thinking" : "Enable preserve thinking"
+        preserveThinking ? "Disable preserve think" : "Enable preserve think"
       }
     >
       {preserveThinking && !disabled ? (
@@ -493,7 +493,7 @@ const PreserveThinkingToggle: FC = () => {
       ) : (
         <LightbulbOffIcon className="size-3.5" />
       )}
-      <span>Preserve Thinking</span>
+      <span>Preserve Think</span>
     </button>
   );
 };
diff --git a/studio/frontend/src/features/chat/shared-composer.tsx b/studio/frontend/src/features/chat/shared-composer.tsx
index 3ba2f8b6f5..1a3ba7e166 100644
--- a/studio/frontend/src/features/chat/shared-composer.tsx
+++ b/studio/frontend/src/features/chat/shared-composer.tsx
@@ -662,7 +662,7 @@ export function SharedComposer({
                     : "bg-muted text-muted-foreground hover:bg-muted-foreground/15",
               )}
               aria-label={
-                preserveThinking ? "Disable preserve thinking" : "Enable preserve thinking"
+                preserveThinking ? "Disable preserve think" : "Enable preserve think"
               }
             >
               {preserveThinking && modelLoaded ? (
@@ -670,7 +670,7 @@ export function SharedComposer({
               ) : (
                 <LightbulbOffIcon className="size-3.5" />
               )}
-              <span>Preserve Thinking</span>
+              <span>Preserve Think</span>
             </button>
           )}
           <button
diff --git a/studio/frontend/src/features/studio/sections/dataset-section.tsx b/studio/frontend/src/features/studio/sections/dataset-section.tsx
index b12bdb09f0..11c2321863 100644
--- a/studio/frontend/src/features/studio/sections/dataset-section.tsx
+++ b/studio/frontend/src/features/studio/sections/dataset-section.tsx
@@ -58,17 +58,42 @@ import {
   ViewIcon,
 } from "@hugeicons/core-free-icons";
 import { HugeiconsIcon } from "@hugeicons/react";
-import { type ChangeEvent, useCallback, useEffect, useMemo, useRef, useState } from "react";
+import {
+  type ChangeEvent,
+  type DragEvent,
+  useCallback,
+  useEffect,
+  useMemo,
+  useRef,
+  useState,
+} from "react";
 import { toast } from "sonner";
 import { useShallow } from "zustand/react/shallow";
 import { DocumentUploadRedirectDialog } from "./document-upload-redirect-dialog";
 
+const TRAINING_UPLOAD_EXTENSIONS = [
+  ".csv",
+  ".jsonl",
+  ".json",
+  ".parquet",
+  ".pdf",
+  ".docx",
+  ".txt",
+] as const;
+const TRAINING_UPLOAD_EXTENSION_SET = new Set<string>(TRAINING_UPLOAD_EXTENSIONS);
+const TRAINING_UPLOAD_ACCEPT = TRAINING_UPLOAD_EXTENSIONS.join(",");
+const TRAINING_UPLOAD_LABEL = "CSV, JSONL, JSON, Parquet, PDF, DOCX, TXT";
 const DOCUMENT_REDIRECT_EXTENSIONS = new Set([".pdf", ".docx", ".txt"]);
 
 const SEARCH_INPUT_REASONS = new Set(["input-change", "input-paste", "input-clear"]);
 const OPEN_LEARNING_RECIPES_ON_ARRIVAL_KEY =
   "data-recipes:open-learning-recipes";
 
+function getFileExtension(fileName: string) {
+  const extensionStart = fileName.lastIndexOf(".");
+  return extensionStart >= 0 ? fileName.slice(extensionStart).toLowerCase() : "";
+}
+
 function isLikelyLocalDatasetRef(value: string) {
   return (
     value.startsWith("/") ||
@@ -354,6 +379,7 @@ export function DatasetSection() {
   );
 
   const [isUploading, setIsUploading] = useState(false);
+  const [isDatasetDragOver, setIsDatasetDragOver] = useState(false);
   const [documentRedirectOpen, setDocumentRedirectOpen] = useState(false);
   const [redirectFileName, setRedirectFileName] = useState<string | null>(null);
 
@@ -380,12 +406,15 @@ export function DatasetSection() {
     }
   };
 
-  const handleDatasetFileChange = async (event: ChangeEvent<HTMLInputElement>) => {
-    const file = event.target.files?.[0];
-    event.target.value = "";
-    if (!file) return;
+  const handleDatasetFile = async (file: File) => {
+    const extension = getFileExtension(file.name);
+    if (!TRAINING_UPLOAD_EXTENSION_SET.has(extension)) {
+      toast.error("Unsupported file type", {
+        description: `Upload one ${TRAINING_UPLOAD_LABEL} file.`,
+      });
+      return;
+    }
 
-    const extension = file.name.slice(file.name.lastIndexOf(".")).toLowerCase();
     if (DOCUMENT_REDIRECT_EXTENSIONS.has(extension)) {
       setRedirectFileName(file.name);
       setDocumentRedirectOpen(true);
@@ -395,6 +424,44 @@ export function DatasetSection() {
     await handleFileUpload(file, selectLocalDataset, "Dataset uploaded");
   };
 
+  const handleDatasetFileChange = async (event: ChangeEvent<HTMLInputElement>) => {
+    const file = event.target.files?.[0];
+    event.target.value = "";
+    if (!file) return;
+
+    await handleDatasetFile(file);
+  };
+
+  const handleDatasetDrop = (event: DragEvent<HTMLButtonElement>) => {
+    event.preventDefault();
+    setIsDatasetDragOver(false);
+    if (isUploading) return;
+
+    const files = Array.from(event.dataTransfer.files);
+    if (files.length === 0) return;
+
+    if (files.length > 1) {
+      toast.error("Upload one file at a time", {
+        description: "Training dataset upload accepts a single file.",
+      });
+      return;
+    }
+
+    void handleDatasetFile(files[0]);
+  };
+
+  const handleDatasetDragOver = (event: DragEvent<HTMLButtonElement>) => {
+    event.preventDefault();
+    if (isUploading) return;
+
+    event.dataTransfer.dropEffect = "copy";
+    setIsDatasetDragOver(true);
+  };
+
+  const handleDatasetDragLeave = () => {
+    setIsDatasetDragOver(false);
+  };
+
   const handleEvalFileChange = async (event: ChangeEvent<HTMLInputElement>) => {
     const file = event.target.files?.[0];
     event.target.value = "";
@@ -924,7 +991,7 @@ export function DatasetSection() {
             </CollapsibleContent>
           </Collapsible>
 
-          <div className="flex flex-col gap-4 pt-1">
+          <div className="flex flex-col gap-3">
             {selectedDatasetName ? (
               <div className="flex items-center gap-3 rounded-lg border bg-muted/40 px-3.5 py-3">
                 <div className="rounded-md bg-indigo-500/10 p-1.5">
@@ -971,15 +1038,32 @@ export function DatasetSection() {
                 </Button>
               </div>
             ) : (
-              <div className="flex items-center gap-3 rounded-lg border border-dashed bg-muted/20 px-3.5 py-3">
+              <button
+                type="button"
+                className={`flex w-full cursor-pointer items-center gap-3 rounded-lg border border-dashed px-3.5 py-3 text-left transition-colors ${
+                  isDatasetDragOver
+                    ? "border-indigo-500/70 bg-indigo-500/10"
+                    : "border-border bg-muted/20 hover:border-indigo-500/50 hover:bg-indigo-500/5"
+                }`}
+                disabled={isUploading}
+                onClick={handleUploadButtonClick}
+                onDrop={handleDatasetDrop}
+                onDragOver={handleDatasetDragOver}
+                onDragLeave={handleDatasetDragLeave}
+              >
                 <HugeiconsIcon
-                  icon={Database02Icon}
-                  className="size-4 text-muted-foreground/40"
+                  icon={CloudUploadIcon}
+                  className="pointer-events-none size-4 shrink-0 text-indigo-500"
                 />
-                <span className="text-xs text-muted-foreground">
-                  No dataset selected
+                <span className="pointer-events-none min-w-0">
+                  <span className="block text-xs font-medium text-foreground">
+                    Drop 1 file here or click to upload
+                  </span>
+                  <span className="mt-0.5 block truncate text-[10px] text-muted-foreground">
+                    {TRAINING_UPLOAD_LABEL}
+                  </span>
                 </span>
-              </div>
+              </button>
             )}
 
             <div className="grid grid-cols-2 gap-2">
@@ -1012,7 +1096,7 @@ export function DatasetSection() {
           <input
             ref={fileInputRef}
             type="file"
-            accept=".json,.jsonl,.csv,.parquet,.pdf,.docx,.txt"
+            accept={TRAINING_UPLOAD_ACCEPT}
             className="hidden"
             onChange={(event) => {
               void handleDatasetFileChange(event);

From 456a49a350fbde9e9435fb0f617e32240b493e89 Mon Sep 17 00:00:00 2001
From: Roland Tannous <115670425+rolandtannous@users.noreply.github.com>
Date: Sat, 2 May 2026 23:30:57 +0400
Subject: [PATCH 191/216] Add Qwen3.6 support (#5257)

* qwen3.6 unsloth studio support

* Add qwen3.6 causal-conv1d detection

* Update model_mappings.py

moved qwen3.6-27B to thinking train on completion template

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

---------

Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com>
---
 studio/backend/core/training/worker.py           |  2 ++
 .../tests/test_training_worker_flash_attn.py     | 16 ++++++++++++++++
 studio/backend/utils/datasets/model_mappings.py  |  4 ++++
 studio/backend/utils/transformers_version.py     |  1 +
 4 files changed, 23 insertions(+)

diff --git a/studio/backend/core/training/worker.py b/studio/backend/core/training/worker.py
index 8ab2b5b2be..ff046c3fc9 100644
--- a/studio/backend/core/training/worker.py
+++ b/studio/backend/core/training/worker.py
@@ -50,6 +50,8 @@ def _model_wants_causal_conv1d(model_name: str) -> bool:
         for key in (
             "qwen3.5",
             "qwen3_5",
+            "qwen3.6",
+            "qwen3_6",
             "qwen3-next",
             "qwen3_next",
             "nemotron_h",
diff --git a/studio/backend/tests/test_training_worker_flash_attn.py b/studio/backend/tests/test_training_worker_flash_attn.py
index 986958408e..41a7c87df1 100644
--- a/studio/backend/tests/test_training_worker_flash_attn.py
+++ b/studio/backend/tests/test_training_worker_flash_attn.py
@@ -133,6 +133,22 @@ def test_causal_conv1d_fast_path_preserves_wheel_first_install_args(monkeypatch)
     )
 
 
+def test_causal_conv1d_fast_path_includes_qwen3_6_variants(monkeypatch):
+    install_mock = mock.Mock(return_value = True)
+    monkeypatch.setattr(worker, "_install_package_wheel_first", install_mock)
+
+    worker._ensure_causal_conv1d_fast_path(
+        event_queue = [],
+        model_name = "unsloth/Qwen3.6-4B",
+    )
+    worker._ensure_causal_conv1d_fast_path(
+        event_queue = [],
+        model_name = "unsloth/Qwen3_6-4B",
+    )
+
+    assert install_mock.call_count == 2
+
+
 def test_mamba_ssm_path_preserves_wheel_first_install_args(monkeypatch):
     install_mock = mock.Mock(return_value = True)
     monkeypatch.setattr(worker, "_install_package_wheel_first", install_mock)
diff --git a/studio/backend/utils/datasets/model_mappings.py b/studio/backend/utils/datasets/model_mappings.py
index 36f6886ef6..21e8566ac5 100644
--- a/studio/backend/utils/datasets/model_mappings.py
+++ b/studio/backend/utils/datasets/model_mappings.py
@@ -364,6 +364,10 @@ TEMPLATE_TO_MODEL_MAPPER = {
         "unsloth/Qwen3-4B-Thinking-2507-bnb-4bit",
         "unsloth/Qwen3-30B-A3B-Thinking-2507",
         "Qwen/Qwen3-30B-A3B-Thinking-2507",
+        "Qwen/Qwen3.6-35B-A3B",
+        "unsloth/Qwen3.6-35B-A3B",
+        "Qwen/Qwen3.6-27B",
+        "unsloth/Qwen3.6-27B",
     ),
     "qwen3.5": (
         "unsloth/Qwen3.5-0.8B",
diff --git a/studio/backend/utils/transformers_version.py b/studio/backend/utils/transformers_version.py
index f36bdcd6e8..ebb20b85da 100644
--- a/studio/backend/utils/transformers_version.py
+++ b/studio/backend/utils/transformers_version.py
@@ -63,6 +63,7 @@ TRANSFORMERS_5_MODEL_SUBSTRINGS: tuple[str, ...] = (
 TRANSFORMERS_550_MODEL_SUBSTRINGS: tuple[str, ...] = (
     "gemma-4",  # Gemma-4 (E2B-it, E4B-it, 31B-it, 26B-A4B-it)
     "gemma4",  # Gemma-4 alternate naming
+    "qwen3.6",
 )
 
 # Architecture classes / model_type values that require transformers 5.5.0.

From 4d9a6ac63a7122aadbb85f7f28341e3ba596f808 Mon Sep 17 00:00:00 2001
From: Lee Jackson <130007945+Imagineer99@users.noreply.github.com>
Date: Sun, 3 May 2026 19:25:23 +0100
Subject: [PATCH 192/216] Studio: Chat thread autosave persistence (#5256)

* fix: chat thread autosave persistence

* fix: guard autosave deletion race

* fix: let run-start autosave persist chats

* fix: scope chat autosave to event thread

* fix: clean up tombstoned chat append rows
---
 .../chat/hooks/use-chat-search-index.ts       |   3 +
 .../chat/hooks/use-chat-sidebar-items.ts      |   2 +
 .../src/features/chat/runtime-provider.tsx    | 145 ++++++++++++++++--
 .../chat/utils/chat-thread-tombstones.ts      |  12 ++
 .../chat/utils/delete-thread-message.ts       |   6 +-
 5 files changed, 151 insertions(+), 17 deletions(-)
 create mode 100644 studio/frontend/src/features/chat/utils/chat-thread-tombstones.ts

diff --git a/studio/frontend/src/features/chat/hooks/use-chat-search-index.ts b/studio/frontend/src/features/chat/hooks/use-chat-search-index.ts
index cab91da44f..d9c27dc436 100644
--- a/studio/frontend/src/features/chat/hooks/use-chat-search-index.ts
+++ b/studio/frontend/src/features/chat/hooks/use-chat-search-index.ts
@@ -105,6 +105,9 @@ async function buildIndex(): Promise<ChatSearchItem[]> {
       const arr = byThreadId.get(tid);
       if (arr) merged.push(...arr);
     }
+    if (merged.length === 0) {
+      continue;
+    }
     merged.sort((a, b) => b.createdAt - a.createdAt);
 
     let preview = "";
diff --git a/studio/frontend/src/features/chat/hooks/use-chat-sidebar-items.ts b/studio/frontend/src/features/chat/hooks/use-chat-sidebar-items.ts
index 33cdd88081..69ed162780 100644
--- a/studio/frontend/src/features/chat/hooks/use-chat-sidebar-items.ts
+++ b/studio/frontend/src/features/chat/hooks/use-chat-sidebar-items.ts
@@ -4,6 +4,7 @@
 import { db, useLiveQuery } from "../db";
 import { useChatRuntimeStore } from "../stores/chat-runtime-store";
 import type { ThreadRecord } from "../types";
+import { markChatThreadDeleted } from "../utils/chat-thread-tombstones";
 
 export interface SidebarItem {
   type: "single" | "compare";
@@ -80,6 +81,7 @@ export async function deleteChatItem(
   // Stop any in-flight streams before deleting, so the model doesn't keep
   // generating against a thread that no longer exists.
   for (const id of threadIds) cancelIfRunning(id);
+  for (const id of threadIds) markChatThreadDeleted(id);
 
   await db.transaction("rw", db.threads, db.messages, async () => {
     for (const id of threadIds) {
diff --git a/studio/frontend/src/features/chat/runtime-provider.tsx b/studio/frontend/src/features/chat/runtime-provider.tsx
index 105f2fb4b1..546cc21ed2 100644
--- a/studio/frontend/src/features/chat/runtime-provider.tsx
+++ b/studio/frontend/src/features/chat/runtime-provider.tsx
@@ -16,19 +16,32 @@ import {
   WebSpeechDictationAdapter,
   type unstable_RemoteThreadListAdapter,
   useAui,
+  useAuiEvent,
   useAuiState,
   useLocalRuntime,
   unstable_useRemoteThreadListRuntime as useRemoteThreadListRuntime,
 } from "@assistant-ui/react";
 import { createAssistantStream } from "assistant-stream";
 import mammoth from "mammoth";
-import { type ReactElement, type ReactNode, useEffect, useMemo } from "react";
+import {
+  type ReactElement,
+  type ReactNode,
+  useCallback,
+  useEffect,
+  useMemo,
+  useRef,
+} from "react";
 import { extractText, getDocumentProxy } from "unpdf";
 import { authFetch } from "@/features/auth";
 import { createOpenAIStreamAdapter } from "./api/chat-adapter";
 import { db } from "./db";
 import { useChatRuntimeStore } from "./stores/chat-runtime-store";
 import type { MessageRecord, ModelType } from "./types";
+import {
+  isChatThreadDeleted,
+  markChatThreadDeleted,
+} from "./utils/chat-thread-tombstones";
+import { syncExportedRepositoryToDexie } from "./utils/delete-thread-message";
 
 const DEFAULT_SUGGESTIONS = [
   {
@@ -383,6 +396,55 @@ function toThreadMessage(m: MessageRecord): ThreadMessage {
   };
 }
 
+async function ensureThreadRecord({
+  threadId,
+  modelType,
+  pairId,
+}: {
+  threadId: string;
+  modelType: ModelType;
+  pairId?: string;
+}): Promise<void> {
+  if (isChatThreadDeleted(threadId)) {
+    return;
+  }
+  const existing = await db.threads.get(threadId);
+  if (existing) {
+    return;
+  }
+
+  const currentModelId =
+    useChatRuntimeStore.getState().params.checkpoint ?? "";
+  const record = {
+    id: threadId,
+    title: "New Chat",
+    modelType,
+    modelId: currentModelId,
+    pairId,
+    archived: false,
+    createdAt: Date.now(),
+  };
+
+  try {
+    await db.threads.add(record);
+  } catch (error) {
+    // assistant-ui can issue overlapping first-message persistence calls.
+    // If another call created the same thread while this one was waiting,
+    // treat initialization as successful and let the message write continue.
+    if (await db.threads.get(threadId)) {
+      return;
+    }
+    throw error;
+  }
+}
+
+async function deleteThreadRows(threadId: string): Promise<void> {
+  await db.transaction("rw", db.threads, db.messages, async () => {
+    await db.messages.where("threadId").equals(threadId).delete();
+    await db.threads.delete(threadId);
+  });
+}
+
 function createDexieAdapter(
   modelType: ModelType,
   pairId?: string,
@@ -418,17 +480,7 @@ function createDexieAdapter(
     },
 
     async initialize(threadId: string) {
-      const currentModelId =
-        useChatRuntimeStore.getState().params.checkpoint ?? "";
-      await db.threads.add({
-        id: threadId,
-        title: "New Chat",
-        modelType,
-        modelId: currentModelId,
-        pairId,
-        archived: false,
-        createdAt: Date.now(),
-      });
+      await ensureThreadRecord({ threadId, modelType, pairId });
       return { remoteId: threadId, externalId: undefined };
     },
 
@@ -445,8 +497,8 @@ function createDexieAdapter(
     },
 
     async delete(remoteId: string) {
-      await db.messages.where("threadId").equals(remoteId).delete();
-      await db.threads.delete(remoteId);
+      markChatThreadDeleted(remoteId);
+      await deleteThreadRows(remoteId);
     },
 
     async generateTitle(remoteId: string, messages: readonly ThreadMessage[]) {
@@ -599,6 +651,10 @@ function ThreadHistoryProvider({
 
       async append({ parentId, message }: ExportedMessageRepositoryItem) {
         const { remoteId } = await aui.threadListItem().initialize();
+        if (isChatThreadDeleted(remoteId)) {
+          await deleteThreadRows(remoteId);
+          return;
+        }
         // Keep single-chat runtime state in sync once a new chat is first
         // persisted. Compare panes intentionally do not write global activeThreadId.
         const thread = await db.threads.get(remoteId);
@@ -765,6 +821,66 @@ function CancelRegistrar(): ReactElement | null {
   return null;
 }
 
+function ThreadDexieAutosave({
+  modelType,
+  pairId,
+}: {
+  modelType: ModelType;
+  pairId?: string;
+}): ReactElement | null {
+  const aui = useAui();
+  const saveChainRef = useRef(Promise.resolve());
+
+  const saveThread = useCallback(async (threadId: string): Promise<void> => {
+    const runtime = aui.threads().__internal_getAssistantRuntime?.();
+    if (!runtime) {
+      return;
+    }
+    const exported = runtime.threads.getById(threadId).export();
+    if (exported.messages.length === 0) {
+      return;
+    }
+
+    const { remoteId } = await runtime.threads.getItemById(threadId).initialize();
+    if (isChatThreadDeleted(remoteId)) {
+      await deleteThreadRows(remoteId);
+      return;
+    }
+    await syncExportedRepositoryToDexie(remoteId, exported);
+    if (isChatThreadDeleted(remoteId)) {
+      await deleteThreadRows(remoteId);
+      return;
+    }
+
+    if (modelType === "base" && !pairId) {
+      const store = useChatRuntimeStore.getState();
+      const activeThreadId = runtime.threads.getState().mainThreadId;
+      if (activeThreadId === threadId && store.activeThreadId !== remoteId) {
+        store.setActiveThreadId(remoteId);
+      }
+    }
+  }, [aui, modelType, pairId]);
+
+  const queueSave = useCallback((threadId: string): void => {
+    saveChainRef.current = saveChainRef.current
+      .catch(() => {})
+      .then(() => saveThread(threadId))
+      .catch((error) => {
+        console.error("Failed to autosave chat thread", error);
+      });
+  }, [saveThread]);
+
+  useAuiEvent("thread.runEnd", ({ threadId }) => {
+    queueSave(threadId);
+  });
+
+  useAuiEvent("thread.runStart", ({ threadId }) => {
+    queueSave(threadId);
+  });
+
+  return null;
+}
+
 export function ChatRuntimeProvider({
   children,
   modelType = "base",
@@ -797,6 +913,7 @@ export function ChatRuntimeProvider({
       <ActiveThreadSync
         enabled={modelType === "base" && !pairId && !newThreadNonce && !initialThreadId}
       />
+      <ThreadDexieAutosave modelType={modelType} pairId={pairId} />
       <CancelRegistrar />
       {initialThreadId && (
         <ThreadAutoSwitch
diff --git a/studio/frontend/src/features/chat/utils/chat-thread-tombstones.ts b/studio/frontend/src/features/chat/utils/chat-thread-tombstones.ts
new file mode 100644
index 0000000000..462fc3c0df
--- /dev/null
+++ b/studio/frontend/src/features/chat/utils/chat-thread-tombstones.ts
@@ -0,0 +1,12 @@
+// SPDX-License-Identifier: AGPL-3.0-only
+// Copyright 2026-present the Unsloth AI Inc. team. All rights reserved. See /studio/LICENSE.AGPL-3.0
+
+const deletedThreadIds = new Set<string>();
+
+export function markChatThreadDeleted(threadId: string): void {
+  deletedThreadIds.add(threadId);
+}
+
+export function isChatThreadDeleted(threadId: string): boolean {
+  return deletedThreadIds.has(threadId);
+}
diff --git a/studio/frontend/src/features/chat/utils/delete-thread-message.ts b/studio/frontend/src/features/chat/utils/delete-thread-message.ts
index 204c0d0064..e13da60630 100644
--- a/studio/frontend/src/features/chat/utils/delete-thread-message.ts
+++ b/studio/frontend/src/features/chat/utils/delete-thread-message.ts
@@ -19,8 +19,8 @@ import type {
  * surface area.
  */
 import { MessageRepository } from "@assistant-ui/core/internal";
-import { db } from "@/features/chat/db";
-import type { MessageRecord } from "@/features/chat/types";
+import { db } from "../db";
+import type { MessageRecord } from "../types";
 
 function cloneContent(content: ThreadMessage["content"]): ThreadMessage["content"] {
   if (typeof content === "string") {
@@ -74,7 +74,7 @@ function exportedItemToRecord(
  * Persist the exact message list represented by `exp` for this thread, removing
  * Dexie rows that are no longer present (e.g. after a delete).
  */
-async function syncExportedRepositoryToDexie(
+export async function syncExportedRepositoryToDexie(
   remoteId: string,
   exp: ExportedMessageRepository,
 ): Promise<void> {

From 8cbd16786ba4431c9623e01b42f81977f4d9e97c Mon Sep 17 00:00:00 2001
From: Lee Jackson <130007945+Imagineer99@users.noreply.github.com>
Date: Sun, 3 May 2026 21:28:58 +0100
Subject: [PATCH 193/216] Studio: Enable deleting fine-tuned chat models
 (#5234)

* feat: enable deleting fine-tuned chat models

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

* fix: tighten fine-tuned model delete guards

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

* fix: harden fine-tuned model deletion edge cases

* fix: reject gguf export delete without variant; surface 503 on backend probe failure

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

* fix: narrow loaded-model delete guard by gguf_variant

* fix: clear inference state when cancelling model load

* fix: verify deletion completed and prune empty parent dirs

---------

Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com>
Co-authored-by: Roland Tannous <rolandtannous@gravityq.ai>
---
 studio/backend/core/inference/orchestrator.py |  11 +
 studio/backend/routes/models.py               | 333 ++++++++++++++++++
 studio/backend/utils/models/model_config.py   |   8 +-
 .../assistant-ui/model-selector.tsx           |  20 +-
 .../model-selector/model-delete-action.tsx    | 112 ++++++
 .../assistant-ui/model-selector/pickers.tsx   | 288 ++++++++-------
 .../assistant-ui/model-selector/types.ts      |   5 +
 .../src/features/chat/api/chat-api.ts         |  19 +
 .../frontend/src/features/chat/chat-page.tsx  |  67 ++++
 .../chat/hooks/use-chat-model-runtime.ts      |  44 +--
 10 files changed, 750 insertions(+), 157 deletions(-)
 create mode 100644 studio/frontend/src/components/assistant-ui/model-selector/model-delete-action.tsx

diff --git a/studio/backend/core/inference/orchestrator.py b/studio/backend/core/inference/orchestrator.py
index cb5d9da34a..428ecaa1c3 100644
--- a/studio/backend/core/inference/orchestrator.py
+++ b/studio/backend/core/inference/orchestrator.py
@@ -708,6 +708,17 @@ class InferenceOrchestrator:
 
     def unload_model(self, model_name: str) -> bool:
         """Unload a model from the subprocess."""
+        if model_name in self.loading_models:
+            logger.info(
+                "Cancelling in-flight load for model '%s' by terminating subprocess",
+                model_name,
+            )
+            self._shutdown_subprocess(timeout = 0.5)
+            self.loading_models.discard(model_name)
+            self.active_model_name = None
+            self.models.clear()
+            return True
+
         if not self._ensure_subprocess_alive():
             # No subprocess — just clear local state
             self.models.pop(model_name, None)
diff --git a/studio/backend/routes/models.py b/studio/backend/routes/models.py
index c312b0ee1d..d01e94b0c9 100644
--- a/studio/backend/routes/models.py
+++ b/studio/backend/routes/models.py
@@ -8,6 +8,7 @@ Model Management API routes
 import hashlib
 import json
 import os
+import shutil
 import sys
 import uuid
 from pathlib import Path
@@ -1684,6 +1685,338 @@ async def scan_loras(
         )
 
 
+def _is_path_under(path: Path, root: Path) -> bool:
+    try:
+        path.resolve().relative_to(root.resolve())
+        return True
+    except ValueError:
+        return False
+
+
+def _is_path_under_lexically(path: Path, root: Path) -> bool:
+    """Check containment without resolving the final path's symlink target."""
+    try:
+        absolute_path = Path(os.path.abspath(str(path)))
+        absolute_root = Path(os.path.abspath(str(root)))
+        absolute_path.relative_to(absolute_root)
+        return True
+    except ValueError:
+        return False
+
+
+def _loaded_model_matches_deleted_path(active_model: str, deleted_path: Path) -> bool:
+    try:
+        active = Path(active_model).expanduser().resolve()
+        target = deleted_path.resolve()
+        return active == target or (target.is_dir() and active.is_relative_to(target))
+    except (OSError, RuntimeError, ValueError) as e:
+        logger.debug(
+            "Could not resolve loaded/deleted model paths; falling back to string comparison: %s",
+            e,
+        )
+        active_lower = active_model.lower()
+        target_lower = str(deleted_path).lower()
+        return active_lower == target_lower or active_lower.startswith(
+            f"{target_lower}{os.sep}"
+        )
+
+
+def _loading_model_matches_deleted_path(
+    loading_model: object,
+    deleted_path: Path,
+) -> bool:
+    if not loading_model:
+        return False
+    return _loaded_model_matches_deleted_path(str(loading_model), deleted_path)
+
+
+def _prune_empty_parents(start: Path, stop_at: Path) -> None:
+    """Remove empty ancestor directories of ``start`` up to (but not including) ``stop_at``.
+
+    Used after deleting a model checkpoint so the enclosing run directory does
+    not linger as an empty entry in scan results.
+    """
+    try:
+        stop_resolved = stop_at.resolve()
+    except OSError:
+        return
+    parent = start.parent
+    while True:
+        try:
+            parent_resolved = parent.resolve()
+        except OSError:
+            return
+        if parent_resolved == stop_resolved:
+            return
+        try:
+            parent_resolved.relative_to(stop_resolved)
+        except ValueError:
+            return
+        try:
+            parent.rmdir()
+        except OSError:
+            return
+        parent = parent.parent
+
+
+def _delete_gguf_variant_files(root: Path, variant: str) -> tuple[int, int]:
+    deleted_count = 0
+    deleted_bytes = 0
+    for path in root.rglob("*"):
+        if not path.is_file() or not _is_main_gguf_filename(path.name):
+            continue
+        if _extract_quant_label(path.name).lower() != variant.lower():
+            continue
+        try:
+            deleted_bytes += path.stat().st_size
+        except OSError:
+            pass
+        path.unlink()
+        deleted_count += 1
+    return deleted_count, deleted_bytes
+
+
+@router.delete("/delete-finetuned")
+async def delete_finetuned_model(
+    model_path: str = Body(...),
+    source: str = Body(...),
+    export_type: Optional[str] = Body(None),
+    gguf_variant: Optional[str] = Body(None),
+    current_subject: str = Depends(get_current_subject),
+):
+    """Delete a Studio-trained or exported model from disk.
+
+    Only paths under Studio's outputs/exports roots are accepted.  Exported
+    GGUF entries can delete one quantization variant at a time.
+    """
+    if source not in {"training", "exported"}:
+        raise HTTPException(
+            status_code = 400,
+            detail = "Only trained or exported Studio models can be deleted",
+        )
+
+    if not model_path or not model_path.strip():
+        raise HTTPException(status_code = 400, detail = "model_path is required")
+
+    if export_type == "gguf" and not gguf_variant:
+        raise HTTPException(
+            status_code = 400,
+            detail = "gguf_variant is required when export_type is 'gguf'",
+        )
+
+    raw_path = Path(model_path).expanduser()
+    if source == "training":
+        target_path = raw_path
+        allowed_root = outputs_root()
+    else:
+        allowed_root = exports_root()
+        target_path = (
+            raw_path.parent
+            if export_type == "gguf" and raw_path.suffix.lower() == ".gguf"
+            else raw_path
+        )
+
+    allowed_root = allowed_root.resolve()
+    delete_path = Path(os.path.abspath(str(target_path)))
+    delete_path_is_symlink = delete_path.is_symlink()
+
+    if delete_path_is_symlink:
+        if not _is_path_under_lexically(delete_path, allowed_root):
+            raise HTTPException(
+                status_code = 400,
+                detail = "Model path is outside Studio storage",
+            )
+        if export_type == "gguf" and gguf_variant:
+            target_path = delete_path.resolve()
+            if not _is_path_under(target_path, allowed_root):
+                raise HTTPException(
+                    status_code = 400,
+                    detail = "Model path is outside Studio storage",
+                )
+        else:
+            target_path = delete_path
+    else:
+        target_path = target_path.resolve()
+
+    should_check_resolved_path = not delete_path_is_symlink or (
+        export_type == "gguf" and gguf_variant
+    )
+    if should_check_resolved_path and not _is_path_under(target_path, allowed_root):
+        raise HTTPException(
+            status_code = 400,
+            detail = "Model path is outside Studio storage",
+        )
+    if target_path == allowed_root:
+        raise HTTPException(
+            status_code = 400,
+            detail = "Refusing to delete storage root",
+        )
+    if not target_path.exists() and not target_path.is_symlink():
+        raise HTTPException(status_code = 404, detail = "Model not found on disk")
+
+    if source == "training":
+        try:
+            from core.training import get_training_backend
+
+            training_backend = get_training_backend()
+            if training_backend.is_training_active():
+                raise HTTPException(
+                    status_code = 409,
+                    detail = "Cannot delete trained models while training is running",
+                )
+        except HTTPException:
+            raise
+        except Exception as e:
+            logger.warning("Could not check training status before delete: %s", e)
+            raise HTTPException(
+                status_code = 500,
+                detail = "Could not verify training status before deleting",
+            ) from e
+
+    try:
+        from routes.inference import get_llama_cpp_backend
+
+        llama_backend = get_llama_cpp_backend()
+        if (
+            llama_backend.is_active
+            and not llama_backend.is_loaded
+            and llama_backend.model_identifier
+            and _loaded_model_matches_deleted_path(
+                llama_backend.model_identifier,
+                target_path,
+            )
+            and (
+                not gguf_variant
+                or not llama_backend.hf_variant
+                or llama_backend.hf_variant.lower() == gguf_variant.lower()
+            )
+        ):
+            raise HTTPException(
+                status_code = 409,
+                detail = "Cannot delete a model while it is loading",
+            )
+        if (
+            llama_backend.is_loaded
+            and llama_backend.model_identifier
+            and _loaded_model_matches_deleted_path(
+                llama_backend.model_identifier,
+                target_path,
+            )
+            and (
+                not gguf_variant
+                or not llama_backend.hf_variant
+                or llama_backend.hf_variant.lower() == gguf_variant.lower()
+            )
+        ):
+            raise HTTPException(
+                status_code = 400,
+                detail = "Unload the model before deleting",
+            )
+    except HTTPException:
+        raise
+    except Exception as e:
+        logger.warning("Could not check llama.cpp loaded model before delete: %s", e)
+        raise HTTPException(
+            status_code = 503,
+            detail = "Could not verify model load status before deleting",
+        ) from e
+
+    try:
+        inference_backend = get_inference_backend()
+        loading_models = getattr(inference_backend, "loading_models", set())
+        if any(
+            _loading_model_matches_deleted_path(loading_model, target_path)
+            for loading_model in loading_models
+        ):
+            raise HTTPException(
+                status_code = 409,
+                detail = "Cannot delete a model while it is loading",
+            )
+        if inference_backend.active_model_name:
+            if _loaded_model_matches_deleted_path(
+                inference_backend.active_model_name,
+                target_path,
+            ):
+                raise HTTPException(
+                    status_code = 400,
+                    detail = "Unload the model before deleting",
+                )
+    except HTTPException:
+        raise
+    except Exception as e:
+        logger.warning(
+            "Could not check inference backend loaded model before delete: %s", e
+        )
+        raise HTTPException(
+            status_code = 503,
+            detail = "Could not verify model load status before deleting",
+        ) from e
+
+    try:
+        if export_type == "gguf" and gguf_variant:
+            if not target_path.is_dir():
+                raise HTTPException(
+                    status_code = 400,
+                    detail = "GGUF variant deletion requires an export directory",
+                )
+            deleted_count, deleted_bytes = _delete_gguf_variant_files(
+                target_path,
+                gguf_variant,
+            )
+            if deleted_count == 0:
+                raise HTTPException(
+                    status_code = 404,
+                    detail = f"Variant {gguf_variant} not found on disk",
+                )
+            try:
+                if not any(target_path.iterdir()):
+                    target_path.rmdir()
+                    _prune_empty_parents(target_path, allowed_root)
+            except OSError:
+                pass
+            logger.info(
+                "Deleted %s GGUF file(s) for exported model at %s variant %s (%0.1f MB freed)",
+                deleted_count,
+                target_path,
+                gguf_variant,
+                deleted_bytes / (1024 * 1024),
+            )
+            return {
+                "status": "deleted",
+                "path": str(target_path),
+                "gguf_variant": gguf_variant,
+            }
+
+        if target_path.is_symlink() or target_path.is_file():
+            target_path.unlink()
+        else:
+            shutil.rmtree(target_path)
+
+        if target_path.exists() or target_path.is_symlink():
+            raise HTTPException(
+                status_code = 500,
+                detail = "Deletion incomplete; some files could not be removed",
+            )
+
+        _prune_empty_parents(target_path, allowed_root)
+
+        logger.info("Deleted fine-tuned model at %s", target_path)
+        return {"status": "deleted", "path": str(target_path)}
+    except HTTPException:
+        raise
+    except Exception as e:
+        logger.error(
+            "Error deleting fine-tuned model %s: %s",
+            target_path,
+            e,
+            exc_info = True,
+        )
+        raise HTTPException(
+            status_code = 500,
+            detail = f"Failed to delete fine-tuned model: {str(e)}",
+        )
+
+
 @router.get("/loras/{lora_path:path}/base-model", response_model = LoRABaseModelResponse)
 async def get_lora_base_model(
     lora_path: str,
diff --git a/studio/backend/utils/models/model_config.py b/studio/backend/utils/models/model_config.py
index a2b0c90e59..7bfc74235f 100644
--- a/studio/backend/utils/models/model_config.py
+++ b/studio/backend/utils/models/model_config.py
@@ -1227,9 +1227,11 @@ def _resolve_gguf_dir(p: Path) -> Optional[Path]:
         return p
     if p.is_file() and p.suffix.lower() == ".gguf":
         parent = p.parent
-        if (parent / "config.json").exists() or (
-            parent / "adapter_config.json"
-        ).exists():
+        if (
+            (parent / "config.json").exists()
+            or (parent / "adapter_config.json").exists()
+            or (parent / "export_metadata.json").exists()
+        ):
             return parent
     return None
 
diff --git a/studio/frontend/src/components/assistant-ui/model-selector.tsx b/studio/frontend/src/components/assistant-ui/model-selector.tsx
index 4fa39beff4..1ff946dd46 100644
--- a/studio/frontend/src/components/assistant-ui/model-selector.tsx
+++ b/studio/frontend/src/components/assistant-ui/model-selector.tsx
@@ -18,13 +18,19 @@ import {
 import { HugeiconsIcon } from "@hugeicons/react";
 import { useMemo, useState } from "react";
 import type {
+  DeletedModelRef,
   LoraModelOption,
   ModelOption,
   ModelSelectorChangeMeta,
 } from "./model-selector/types";
 import { HubModelPicker, LoraModelPicker } from "./model-selector/pickers";
 
-export type { LoraModelOption, ModelOption, ModelSelectorChangeMeta } from "./model-selector/types";
+export type {
+  DeletedModelRef,
+  LoraModelOption,
+  ModelOption,
+  ModelSelectorChangeMeta,
+} from "./model-selector/types";
 
 interface ModelSelectorProps {
   models: ModelOption[];
@@ -35,6 +41,8 @@ interface ModelSelectorProps {
   onValueChange?: (value: string, meta: ModelSelectorChangeMeta) => void;
   onEject?: () => void;
   onFoldersChange?: () => void;
+  onModelsChange?: (deletedModel?: DeletedModelRef) => void;
+  deleteDisabled?: boolean;
   variant?: "outline" | "ghost" | "muted";
   size?: "sm" | "default" | "lg";
   className?: string;
@@ -109,6 +117,8 @@ function ModelSelectorContent({
   onSelect,
   onEject,
   onFoldersChange,
+  onModelsChange,
+  deleteDisabled,
   className,
   dataTour,
 }: {
@@ -118,6 +128,8 @@ function ModelSelectorContent({
   onSelect: (id: string, meta: ModelSelectorChangeMeta) => void;
   onEject?: () => void;
   onFoldersChange?: () => void;
+  onModelsChange?: (deletedModel?: DeletedModelRef) => void;
+  deleteDisabled?: boolean;
   className?: string;
   dataTour?: string;
 }) {
@@ -151,6 +163,8 @@ function ModelSelectorContent({
               loraModels={loraModels}
               value={value}
               onSelect={onSelect}
+              onModelsChange={onModelsChange}
+              deleteDisabled={deleteDisabled}
             />
           </TabsContent>
         </Tabs>
@@ -182,6 +196,8 @@ export function ModelSelector({
   onValueChange,
   onEject,
   onFoldersChange,
+  onModelsChange,
+  deleteDisabled,
   variant = "outline",
   size = "default",
   className,
@@ -276,6 +292,8 @@ export function ModelSelector({
         onSelect={handleSelect}
         onEject={onEject ? handleEject : undefined}
         onFoldersChange={onFoldersChange}
+        onModelsChange={onModelsChange}
+        deleteDisabled={deleteDisabled}
         className={contentClassName}
         dataTour={contentDataTour}
       />
diff --git a/studio/frontend/src/components/assistant-ui/model-selector/model-delete-action.tsx b/studio/frontend/src/components/assistant-ui/model-selector/model-delete-action.tsx
new file mode 100644
index 0000000000..58a6235454
--- /dev/null
+++ b/studio/frontend/src/components/assistant-ui/model-selector/model-delete-action.tsx
@@ -0,0 +1,112 @@
+// SPDX-License-Identifier: AGPL-3.0-only
+// Copyright 2026-present the Unsloth AI Inc. team. All rights reserved. See /studio/LICENSE.AGPL-3.0
+
+import {
+  AlertDialog,
+  AlertDialogAction,
+  AlertDialogCancel,
+  AlertDialogContent,
+  AlertDialogDescription,
+  AlertDialogFooter,
+  AlertDialogHeader,
+  AlertDialogTitle,
+} from "@/components/ui/alert-dialog";
+import { cn } from "@/lib/utils";
+import { Trash2Icon } from "lucide-react";
+import { useCallback, useState, type ReactNode } from "react";
+import { toast } from "sonner";
+
+interface ModelDeleteActionProps {
+  ariaLabel: string;
+  title: string;
+  description: ReactNode;
+  successMessage: string;
+  loadingLabel?: string;
+  buttonClassName?: string;
+  iconClassName?: string;
+  disabled?: boolean;
+  onConfirm: () => Promise<void> | void;
+  onDeleted?: () => void;
+}
+
+export function ModelDeleteAction({
+  ariaLabel,
+  title,
+  description,
+  successMessage,
+  loadingLabel = "Deleting...",
+  buttonClassName,
+  iconClassName,
+  disabled = false,
+  onConfirm,
+  onDeleted,
+}: ModelDeleteActionProps) {
+  const [open, setOpen] = useState(false);
+  const [deleting, setDeleting] = useState(false);
+
+  const handleConfirm = useCallback(async () => {
+    setDeleting(true);
+    try {
+      await onConfirm();
+      toast.success(successMessage);
+      onDeleted?.();
+      setOpen(false);
+    } catch (err) {
+      toast.error(
+        err instanceof Error ? err.message : "Failed to delete model",
+      );
+    } finally {
+      setDeleting(false);
+    }
+  }, [onConfirm, onDeleted, successMessage]);
+
+  return (
+    <>
+      <button
+        type="button"
+        onClick={(e) => {
+          e.stopPropagation();
+          if (disabled) return;
+          setOpen(true);
+        }}
+        aria-label={ariaLabel}
+        disabled={disabled}
+        className={cn(
+          "shrink-0 rounded-md p-1.5 text-muted-foreground/60 transition-colors hover:bg-destructive/10 hover:text-destructive",
+          disabled && "cursor-not-allowed opacity-40 hover:bg-transparent hover:text-muted-foreground/60",
+          buttonClassName,
+        )}
+      >
+        <Trash2Icon className={cn("size-3.5", iconClassName)} />
+      </button>
+
+      <AlertDialog
+        open={open}
+        onOpenChange={(nextOpen) => {
+          if (!nextOpen && deleting) return;
+          setOpen(nextOpen);
+        }}
+      >
+        <AlertDialogContent size="sm">
+          <AlertDialogHeader>
+            <AlertDialogTitle>{title}</AlertDialogTitle>
+            <AlertDialogDescription>{description}</AlertDialogDescription>
+          </AlertDialogHeader>
+          <AlertDialogFooter>
+            <AlertDialogCancel disabled={deleting}>No</AlertDialogCancel>
+            <AlertDialogAction
+              variant="destructive"
+              disabled={deleting}
+              onClick={(e) => {
+                e.preventDefault();
+                handleConfirm();
+              }}
+            >
+              {deleting ? loadingLabel : "Yes"}
+            </AlertDialogAction>
+          </AlertDialogFooter>
+        </AlertDialogContent>
+      </AlertDialog>
+    </>
+  );
+}
diff --git a/studio/frontend/src/components/assistant-ui/model-selector/pickers.tsx b/studio/frontend/src/components/assistant-ui/model-selector/pickers.tsx
index 30b8fdecb8..fae3c22caf 100644
--- a/studio/frontend/src/components/assistant-ui/model-selector/pickers.tsx
+++ b/studio/frontend/src/components/assistant-ui/model-selector/pickers.tsx
@@ -1,16 +1,6 @@
 // SPDX-License-Identifier: AGPL-3.0-only
 // Copyright 2026-present the Unsloth AI Inc. team. All rights reserved. See /studio/LICENSE.AGPL-3.0
 
-import {
-  AlertDialog,
-  AlertDialogAction,
-  AlertDialogCancel,
-  AlertDialogContent,
-  AlertDialogDescription,
-  AlertDialogFooter,
-  AlertDialogHeader,
-  AlertDialogTitle,
-} from "@/components/ui/alert-dialog";
 import { Input } from "@/components/ui/input";
 import { Spinner } from "@/components/ui/spinner";
 import {
@@ -23,6 +13,7 @@ import {
   type ScanFolderInfo,
   addScanFolder,
   deleteCachedModel,
+  deleteFineTunedModel,
   listCachedGguf,
   listCachedModels,
   listGgufVariants,
@@ -50,7 +41,8 @@ import { checkVramFit, estimateLoadingVram } from "@/lib/vram";
 import { Add01Icon, Cancel01Icon, Folder02Icon, Search01Icon } from "@hugeicons/core-free-icons";
 import { HugeiconsIcon } from "@hugeicons/react";
 import { FolderBrowser } from "./folder-browser";
-import { ChevronDownIcon, ChevronRightIcon, DownloadIcon, StarIcon, Trash2Icon } from "lucide-react";
+import { ModelDeleteAction } from "./model-delete-action";
+import { ChevronDownIcon, ChevronRightIcon, DownloadIcon, StarIcon } from "lucide-react";
 import {
   type ReactNode,
   useCallback,
@@ -60,6 +52,7 @@ import {
 } from "react";
 import { toast } from "sonner";
 import type {
+  DeletedModelRef,
   LoraModelOption,
   ModelOption,
   ModelSelectorChangeMeta,
@@ -211,12 +204,22 @@ function GgufVariantExpander({
   gpuGb,
   systemRamGb,
   onDeleteVariant,
+  sourceOverride,
+  deleteVariantTitle = "Delete cached model?",
+  renderDeleteVariantDescription,
+  getDeleteVariantSuccessMessage,
+  deleteDisabled = false,
 }: {
   repoId: string;
   onSelect: (id: string, meta: ModelSelectorChangeMeta) => void;
   gpuGb?: number;
   systemRamGb?: number;
-  onDeleteVariant?: (quant: string) => void;
+  onDeleteVariant?: (quant: string) => Promise<void> | void;
+  sourceOverride?: ModelSelectorChangeMeta["source"];
+  deleteVariantTitle?: string;
+  renderDeleteVariantDescription?: (quant: string) => ReactNode;
+  getDeleteVariantSuccessMessage?: (quant: string) => string;
+  deleteDisabled?: boolean;
 }) {
   const [variants, setVariants] = useState<GgufVariantDetail[] | null>(null);
   const [defaultVariant, setDefaultVariant] = useState<string | null>(null);
@@ -259,14 +262,14 @@ function GgufVariantExpander({
   const handleVariantClick = useCallback(
     (quant: string, downloaded?: boolean, sizeBytes?: number) => {
       onSelect(repoId, {
-        source: isLocalPath ? "local" : "hub",
+        source: sourceOverride ?? (isLocalPath ? "local" : "hub"),
         isLora: false,
         ggufVariant: quant,
         isDownloaded: isLocalPath ? true : downloaded,
         expectedBytes: sizeBytes,
       });
     },
-    [repoId, isLocalPath, onSelect],
+    [repoId, isLocalPath, onSelect, sourceOverride],
   );
 
   // GGUF fit classification matching llama-server's _select_gpus logic:
@@ -408,16 +411,29 @@ function GgufVariantExpander({
               </span>
             </button>
             {v.downloaded && onDeleteVariant && (
-              <button
-                type="button"
-                onClick={(e) => {
-                  e.stopPropagation();
-                  onDeleteVariant(v.quant);
-                }}
-                className="shrink-0 rounded-md p-1 text-muted-foreground/60 transition-colors hover:bg-destructive/10 hover:text-destructive"
-              >
-                <Trash2Icon className="size-3" />
-              </button>
+              <ModelDeleteAction
+                ariaLabel={`Delete ${repoId} ${v.quant}`}
+                title={deleteVariantTitle}
+                description={
+                  renderDeleteVariantDescription?.(v.quant) ?? (
+                    <>
+                      This will remove{" "}
+                      <span className="font-medium text-foreground">
+                        {repoId} ({v.quant})
+                      </span>{" "}
+                      from disk. You can re-download it later.
+                    </>
+                  )
+                }
+                successMessage={
+                  getDeleteVariantSuccessMessage?.(v.quant) ??
+                  `Deleted ${repoId} ${v.quant}`
+                }
+                buttonClassName="p-1"
+                iconClassName="size-3"
+                disabled={deleteDisabled}
+                onConfirm={() => onDeleteVariant(v.quant)}
+              />
             )}
           </div>
         );
@@ -512,9 +528,6 @@ export function HubModelPicker({
   // Track which GGUF repo is expanded for variant selection
   const [expandedGguf, setExpandedGguf] = useState<string | null>(null);
 
-  // Delete confirmation dialog state
-  const [deleteTarget, setDeleteTarget] = useState<string | null>(null);
-  const [deleting, setDeleting] = useState(false);
   const [downloadedCollapsed, setDownloadedCollapsed] = useState(false);
   const [customFoldersCollapsed, setCustomFoldersCollapsed] = useState(false);
   const [recommendedCollapsed, setRecommendedCollapsed] = useState(false);
@@ -675,27 +688,6 @@ export function HubModelPicker({
       .finally(check);
   }, [refreshLocalModelsList, refreshScanFolders]);
 
-  const handleDeleteConfirm = useCallback(async () => {
-    if (!deleteTarget) return;
-    setDeleting(true);
-    try {
-      // deleteTarget is "repo_id" or "repo_id::variant"
-      const sepIdx = deleteTarget.indexOf("::");
-      const repoId = sepIdx >= 0 ? deleteTarget.slice(0, sepIdx) : deleteTarget;
-      const variant = sepIdx >= 0 ? deleteTarget.slice(sepIdx + 2) : undefined;
-      await deleteCachedModel(repoId, variant);
-      toast.success(`Deleted ${variant ? `${repoId} ${variant}` : repoId}`);
-      refreshCachedLists();
-    } catch (err) {
-      toast.error(
-        err instanceof Error ? err.message : "Failed to delete model",
-      );
-    } finally {
-      setDeleting(false);
-      setDeleteTarget(null);
-    }
-  }, [deleteTarget, refreshCachedLists]);
-
   // Deduplicate: don't show downloaded models in the recommended list.
   // Compare case-insensitively since HF cache lowercases repo IDs.
   const downloadedSet = useMemo(() => {
@@ -952,9 +944,10 @@ export function HubModelPicker({
                       systemRamGb={
                         gpu.available ? gpu.systemRamAvailableGb : undefined
                       }
-                      onDeleteVariant={(quant) =>
-                        setDeleteTarget(`${c.repo_id}::${quant}`)
-                      }
+                      onDeleteVariant={async (quant) => {
+                        await deleteCachedModel(c.repo_id, quant);
+                        refreshCachedLists();
+                      }}
                     />
                   )}
                 </div>
@@ -977,16 +970,22 @@ export function HubModelPicker({
                         vramStatus={null}
                       />
                     </div>
-                    <button
-                      type="button"
-                      onClick={(e) => {
-                        e.stopPropagation();
-                        setDeleteTarget(c.repo_id);
-                      }}
-                      className="shrink-0 rounded-md p-1.5 text-muted-foreground/60 transition-colors hover:bg-destructive/10 hover:text-destructive"
-                    >
-                      <Trash2Icon className="size-3.5" />
-                    </button>
+                    <ModelDeleteAction
+                      ariaLabel={`Delete ${c.repo_id}`}
+                      title="Delete cached model?"
+                      description={
+                        <>
+                          This will remove{" "}
+                          <span className="font-medium text-foreground">
+                            {c.repo_id}
+                          </span>{" "}
+                          from disk. You can re-download it later.
+                        </>
+                      }
+                      successMessage={`Deleted ${c.repo_id}`}
+                      onConfirm={() => deleteCachedModel(c.repo_id)}
+                      onDeleted={refreshCachedLists}
+                    />
                   </div>
                 ))}
             </>
@@ -1409,40 +1408,6 @@ export function HubModelPicker({
         </div>
       </div>
 
-      <AlertDialog
-        open={deleteTarget !== null}
-        onOpenChange={(open) => {
-          if (!open && !deleting) setDeleteTarget(null);
-        }}
-      >
-        <AlertDialogContent size="sm">
-          <AlertDialogHeader>
-            <AlertDialogTitle>Delete cached model?</AlertDialogTitle>
-            <AlertDialogDescription>
-              This will remove{" "}
-              <span className="font-medium text-foreground">
-                {deleteTarget?.includes("::")
-                  ? `${deleteTarget.split("::")[0]} (${deleteTarget.split("::")[1]})`
-                  : deleteTarget}
-              </span>{" "}
-              from disk. You can re-download it later.
-            </AlertDialogDescription>
-          </AlertDialogHeader>
-          <AlertDialogFooter>
-            <AlertDialogCancel disabled={deleting}>No</AlertDialogCancel>
-            <AlertDialogAction
-              variant="destructive"
-              disabled={deleting}
-              onClick={(e) => {
-                e.preventDefault();
-                handleDeleteConfirm();
-              }}
-            >
-              {deleting ? "Deleting..." : "Yes"}
-            </AlertDialogAction>
-          </AlertDialogFooter>
-        </AlertDialogContent>
-      </AlertDialog>
     </div>
   );
 }
@@ -1451,10 +1416,14 @@ export function LoraModelPicker({
   loraModels,
   value,
   onSelect,
+  onModelsChange,
+  deleteDisabled = false,
 }: {
   loraModels: LoraModelOption[];
   value?: string;
   onSelect: (id: string, meta: ModelSelectorChangeMeta) => void;
+  onModelsChange?: (deletedModel?: DeletedModelRef) => void;
+  deleteDisabled?: boolean;
 }) {
   const [query, setQuery] = useState("");
   const [expandedGguf, setExpandedGguf] = useState<string | null>(null);
@@ -1541,6 +1510,8 @@ export function LoraModelPicker({
                   const isExported = adapter.source === "exported";
                   const isMerged = adapter.exportType === "merged";
                   const isGguf = adapter.exportType === "gguf";
+                  const isExportedGguf = isExported && isGguf;
+                  const canDelete = (isTraining || isExported) && !isExportedGguf;
                   const isTrainingFull = isTraining && isMerged;
                   const isLocalGgufDir =
                     isLocal &&
@@ -1569,38 +1540,69 @@ export function LoraModelPicker({
                       : tag;
                   return (
                     <div key={adapter.id}>
-                      <ModelRow
-                        label={adapter.name}
-                        meta={meta}
-                        selected={value === adapter.id}
-                        onClick={() => {
-                          if (isLocalGgufDir) {
-                            setExpandedGguf((prev) =>
-                              prev === adapter.id ? null : adapter.id,
-                            );
-                          } else {
-                            onSelect(adapter.id, {
-                              source: isLocal
-                                ? "local"
-                                : isExported
-                                  ? "exported"
-                                  : "lora",
-                              isLora: !isLocal && !isMerged && !isGguf,
-                              isDownloaded: true,
-                            });
-                          }
-                        }}
-                        tooltipText={
-                          <>
-                            <span className="block break-words">
-                              {adapter.name}
-                            </span>
-                            <span className="block mt-1 text-[10px] text-muted-foreground break-all">
-                              {adapter.id}
-                            </span>
-                          </>
-                        }
-                      />
+                      <div className="flex items-center gap-0.5">
+                        <div className="min-w-0 flex-1">
+                          <ModelRow
+                            label={adapter.name}
+                            meta={meta}
+                            selected={value === adapter.id}
+                            onClick={() => {
+                              if (isLocalGgufDir || isExportedGguf) {
+                                setExpandedGguf((prev) =>
+                                  prev === adapter.id ? null : adapter.id,
+                                );
+                              } else {
+                                onSelect(adapter.id, {
+                                  source: isLocal
+                                    ? "local"
+                                    : isExported
+                                      ? "exported"
+                                      : "lora",
+                                  isLora: !isLocal && !isMerged && !isGguf,
+                                  isDownloaded: true,
+                                });
+                              }
+                            }}
+                            tooltipText={
+                              <>
+                                <span className="block break-words">
+                                  {adapter.name}
+                                </span>
+                                <span className="block mt-1 text-[10px] text-muted-foreground break-all">
+                                  {adapter.id}
+                                </span>
+                              </>
+                            }
+                          />
+                        </div>
+                        {canDelete && (
+                          <ModelDeleteAction
+                            ariaLabel={`Delete ${adapter.name}`}
+                            title="Delete fine-tuned model?"
+                            description={
+                              <>
+                                This will remove{" "}
+                                <span className="font-medium text-foreground">
+                                  {adapter.name}
+                                </span>{" "}
+                                from disk. This cannot be undone.
+                              </>
+                            }
+                            successMessage={`Deleted ${adapter.name}`}
+                            disabled={deleteDisabled}
+                            onConfirm={() =>
+                              deleteFineTunedModel({
+                                modelPath: adapter.id,
+                                source: isExported ? "exported" : "training",
+                                exportType: adapter.exportType,
+                              })
+                            }
+                            onDeleted={() =>
+                              onModelsChange?.({ id: adapter.id })
+                            }
+                          />
+                        )}
+                      </div>
                       {expandedGguf === adapter.id && (
                         <GgufVariantExpander
                           repoId={adapter.id}
@@ -1609,6 +1611,37 @@ export function LoraModelPicker({
                           systemRamGb={
                             gpu.available ? gpu.systemRamAvailableGb : undefined
                           }
+                          sourceOverride={isExportedGguf ? "exported" : undefined}
+                          deleteVariantTitle="Delete exported GGUF variant?"
+                          renderDeleteVariantDescription={(quant) => (
+                            <>
+                              This will remove{" "}
+                              <span className="font-medium text-foreground">
+                                {adapter.name} ({quant})
+                              </span>{" "}
+                              from disk. This cannot be undone.
+                            </>
+                          )}
+                          getDeleteVariantSuccessMessage={(quant) =>
+                            `Deleted ${adapter.name} ${quant}`
+                          }
+                          deleteDisabled={deleteDisabled}
+                          onDeleteVariant={
+                            isExportedGguf
+                              ? async (quant) => {
+                                  await deleteFineTunedModel({
+                                    modelPath: adapter.id,
+                                    source: "exported",
+                                    exportType: "gguf",
+                                    ggufVariant: quant,
+                                  });
+                                  onModelsChange?.({
+                                    id: adapter.id,
+                                    ggufVariant: quant,
+                                  });
+                                }
+                              : undefined
+                          }
                         />
                       )}
                     </div>
@@ -1619,6 +1652,7 @@ export function LoraModelPicker({
           )}
         </div>
       </div>
+
     </div>
   );
 }
diff --git a/studio/frontend/src/components/assistant-ui/model-selector/types.ts b/studio/frontend/src/components/assistant-ui/model-selector/types.ts
index 215dd2b38e..3da75b4d4e 100644
--- a/studio/frontend/src/components/assistant-ui/model-selector/types.ts
+++ b/studio/frontend/src/components/assistant-ui/model-selector/types.ts
@@ -25,3 +25,8 @@ export interface ModelSelectorChangeMeta {
   isDownloaded?: boolean;
   expectedBytes?: number;
 }
+
+export interface DeletedModelRef {
+  id: string;
+  ggufVariant?: string;
+}
diff --git a/studio/frontend/src/features/chat/api/chat-api.ts b/studio/frontend/src/features/chat/api/chat-api.ts
index 15ac8748f0..32a03171ee 100644
--- a/studio/frontend/src/features/chat/api/chat-api.ts
+++ b/studio/frontend/src/features/chat/api/chat-api.ts
@@ -219,6 +219,25 @@ export async function deleteCachedModel(repoId: string, variant?: string): Promi
   await parseJsonOrThrow<unknown>(response);
 }
 
+export async function deleteFineTunedModel(args: {
+  modelPath: string;
+  source: "training" | "exported";
+  exportType?: "lora" | "merged" | "gguf";
+  ggufVariant?: string;
+}): Promise<void> {
+  const response = await authFetch("/api/models/delete-finetuned", {
+    method: "DELETE",
+    headers: { "Content-Type": "application/json" },
+    body: JSON.stringify({
+      model_path: args.modelPath,
+      source: args.source,
+      export_type: args.exportType ?? null,
+      gguf_variant: args.ggufVariant ?? null,
+    }),
+  });
+  await parseJsonOrThrow<unknown>(response);
+}
+
 export interface ScanFolderInfo {
   id: number;
   path: string;
diff --git a/studio/frontend/src/features/chat/chat-page.tsx b/studio/frontend/src/features/chat/chat-page.tsx
index 39683661c7..a239cae848 100644
--- a/studio/frontend/src/features/chat/chat-page.tsx
+++ b/studio/frontend/src/features/chat/chat-page.tsx
@@ -2,6 +2,7 @@
 // Copyright 2026-present the Unsloth AI Inc. team. All rights reserved. See /studio/LICENSE.AGPL-3.0
 
 import {
+  type DeletedModelRef,
   type LoraModelOption,
   type ModelOption,
   ModelSelector,
@@ -129,6 +130,17 @@ type CompareModelSelection = {
   ggufVariant?: string;
 };
 
+function modelMatchesDeleted(
+  model: { id: string; ggufVariant?: string | null },
+  deletedModel?: DeletedModelRef,
+): boolean {
+  if (!deletedModel || model.id !== deletedModel.id) return false;
+  return (
+    deletedModel.ggufVariant == null ||
+    (model.ggufVariant ?? null) === deletedModel.ggufVariant
+  );
+}
+
 /**
  * Detect if this is a LoRA base-vs-fine-tuned compare.
  * Returns true when the loaded checkpoint is a LoRA — in that case
@@ -147,11 +159,15 @@ const CompareContent = memo(function CompareContent({
   models,
   loraModels,
   onFoldersChange,
+  onModelsChange,
+  deleteDisabled,
 }: {
   pairId: string;
   models: ModelOption[];
   loraModels: LoraModelOption[];
   onFoldersChange?: () => void;
+  onModelsChange?: (deletedModel?: DeletedModelRef) => void;
+  deleteDisabled?: boolean;
 }): ReactElement {
   const isLoraCompare = useIsLoraCompare();
 
@@ -163,6 +179,8 @@ const CompareContent = memo(function CompareContent({
       models={models}
       loraModels={loraModels}
       onFoldersChange={onFoldersChange}
+      onModelsChange={onModelsChange}
+      deleteDisabled={deleteDisabled}
     />
   );
 });
@@ -331,6 +349,8 @@ function GeneralCompareHeader({
   value,
   onValueChange,
   onFoldersChange,
+  onModelsChange,
+  deleteDisabled,
   side,
 }: {
   models: ModelOption[];
@@ -341,6 +361,8 @@ function GeneralCompareHeader({
     meta: { isLora: boolean; ggufVariant?: string },
   ) => void;
   onFoldersChange?: () => void;
+  onModelsChange?: (deletedModel?: DeletedModelRef) => void;
+  deleteDisabled?: boolean;
   side: "left" | "right";
 }): ReactElement {
   return (
@@ -356,6 +378,8 @@ function GeneralCompareHeader({
         value={value}
         onValueChange={onValueChange}
         onFoldersChange={onFoldersChange}
+        onModelsChange={onModelsChange}
+        deleteDisabled={deleteDisabled}
         variant="ghost"
         className="max-w-[80%] !h-[34px]"
       />
@@ -369,11 +393,15 @@ const GeneralCompareContent = memo(function GeneralCompareContent({
   models,
   loraModels,
   onFoldersChange,
+  onModelsChange,
+  deleteDisabled,
 }: {
   pairId: string;
   models: ModelOption[];
   loraModels: LoraModelOption[];
   onFoldersChange?: () => void;
+  onModelsChange?: (deletedModel?: DeletedModelRef) => void;
+  deleteDisabled?: boolean;
 }): ReactElement {
   const handlesRef = useRef<Record<string, CompareHandle>>({});
   const [model1ThreadId, setModel1ThreadId] = useState<string>();
@@ -391,6 +419,19 @@ const GeneralCompareContent = memo(function GeneralCompareContent({
     isLora: false,
   });
 
+  const handleModelsChange = useCallback(
+    (deletedModel?: DeletedModelRef) => {
+      if (modelMatchesDeleted(model1, deletedModel)) {
+        setModel1({ id: "", isLora: false });
+      }
+      if (modelMatchesDeleted(model2, deletedModel)) {
+        setModel2({ id: "", isLora: false });
+      }
+      onModelsChange?.(deletedModel);
+    },
+    [model1, model2, onModelsChange],
+  );
+
   useEffect(() => {
     let isActive = true;
     db.threads
@@ -446,6 +487,8 @@ const GeneralCompareContent = memo(function GeneralCompareContent({
                 })
               }
               onFoldersChange={onFoldersChange}
+              onModelsChange={handleModelsChange}
+              deleteDisabled={deleteDisabled}
             />
           }
         />
@@ -469,6 +512,8 @@ const GeneralCompareContent = memo(function GeneralCompareContent({
                 })
               }
               onFoldersChange={onFoldersChange}
+              onModelsChange={handleModelsChange}
+              deleteDisabled={deleteDisabled}
             />
           }
         />
@@ -530,6 +575,9 @@ export function ChatPage(): ReactElement {
   const lorasFromStore = useChatRuntimeStore((state) => state.loras);
   const modelsError = useChatRuntimeStore((state) => state.modelsError);
   const activeThreadId = useChatRuntimeStore((state) => state.activeThreadId);
+  const modelOperationInProgress = useChatRuntimeStore(
+    (state) => state.modelLoading,
+  );
   const {
     refresh,
     selectModel,
@@ -734,6 +782,21 @@ export function ChatPage(): ReactElement {
       .catch(() => {});
   }, [navigate]);
 
+  const refreshModelLists = useCallback((deletedModel?: DeletedModelRef) => {
+    const { checkpoint } = useChatRuntimeStore.getState().params;
+    const activeGgufVariant = useChatRuntimeStore.getState().activeGgufVariant;
+    if (
+      modelMatchesDeleted(
+        { id: checkpoint, ggufVariant: activeGgufVariant },
+        deletedModel,
+      )
+    ) {
+      useChatRuntimeStore.getState().clearCheckpoint();
+    }
+    void refresh();
+    refreshLocalModels();
+  }, [refresh, refreshLocalModels]);
+
   const loraModels = useMemo<LoraModelOption[]>(() => {
     const fromLoras = lorasFromStore.map((lora) => ({
       id: lora.id,
@@ -877,6 +940,8 @@ export function ChatPage(): ReactElement {
                 onValueChange={handleCheckpointChange}
                 onEject={handleEject}
                 onFoldersChange={refreshLocalModels}
+                onModelsChange={refreshModelLists}
+                deleteDisabled={modelOperationInProgress}
                 variant="ghost"
                 open={modelSelectorOpen}
                 onOpenChange={handleModelSelectorOpenChange}
@@ -962,6 +1027,8 @@ export function ChatPage(): ReactElement {
             models={models}
             loraModels={loraModels}
             onFoldersChange={refreshLocalModels}
+            onModelsChange={refreshModelLists}
+            deleteDisabled={modelOperationInProgress}
           />
         )}
       </div>
diff --git a/studio/frontend/src/features/chat/hooks/use-chat-model-runtime.ts b/studio/frontend/src/features/chat/hooks/use-chat-model-runtime.ts
index d9e817678b..b9b6f0b543 100644
--- a/studio/frontend/src/features/chat/hooks/use-chat-model-runtime.ts
+++ b/studio/frontend/src/features/chat/hooks/use-chat-model-runtime.ts
@@ -21,11 +21,9 @@ import {
   mergeBackendRecommendedInference,
   resolveLoadMaxSeqLength,
 } from "../presets/preset-policy";
-import type { InferenceStatusResponse, LoadModelResponse } from "../types/api";
 import type {
   ChatLoraSummary,
   ChatModelSummary,
-  InferenceParams,
 } from "../types/runtime";
 
 // The simplified Speculative Decoding control surfaces "default" (which
@@ -167,6 +165,7 @@ export function useChatModelRuntime() {
   const loadingModelRef = useRef<typeof loadingModel>(null);
   const loadToastIdRef = useRef<string | number | null>(null);
   const loadToastDismissedRef = useRef(false);
+  const cancelUnloadPendingRef = useRef(false);
 
   const setLoadToastDismissedState = useCallback((dismissed: boolean) => {
     loadToastDismissedRef.current = dismissed;
@@ -180,7 +179,9 @@ export function useChatModelRuntime() {
     loadAbortRef.current = null;
     loadToastIdRef.current = null;
     setLoadToastDismissedState(false);
-    useChatRuntimeStore.getState().setModelLoading(false);
+    if (!cancelUnloadPendingRef.current) {
+      useChatRuntimeStore.getState().setModelLoading(false);
+    }
   }, [setLoadToastDismissedState]);
 
   const renderLoadDescription = useCallback(
@@ -219,27 +220,6 @@ export function useChatModelRuntime() {
         // Apply inference defaults on reconnect (page refresh with model already loaded)
         if (statusRes.inference) {
           const currentParams = useChatRuntimeStore.getState().params;
-          const reconnectResponse: LoadModelResponse = {
-            status: "already_loaded",
-            model: statusRes.active_model,
-            display_name: statusRes.active_model,
-            is_vision: statusRes.is_vision,
-            is_lora: false,
-            is_gguf: statusRes.is_gguf,
-            is_audio: statusRes.is_audio,
-            audio_type: statusRes.audio_type,
-            has_audio_input: statusRes.has_audio_input,
-            inference: statusRes.inference,
-            context_length: statusRes.context_length,
-            max_context_length: statusRes.max_context_length,
-            native_context_length: statusRes.native_context_length,
-            supports_reasoning: statusRes.supports_reasoning,
-            reasoning_style: statusRes.reasoning_style,
-            reasoning_always_on: statusRes.reasoning_always_on,
-            supports_preserve_thinking: statusRes.supports_preserve_thinking,
-            supports_tools: statusRes.supports_tools,
-            speculative_type: statusRes.speculative_type,
-          };
           setParams(
             mergeBackendRecommendedInference({
               current: currentParams,
@@ -337,8 +317,18 @@ export function useChatModelRuntime() {
         ? undefined
         : "The current download may still finish in the background.",
     });
-    // Fire-and-forget: tell backend to stop, don't block UI
-    unloadModel({ model_path: model.id }).catch(() => {});
+    cancelUnloadPendingRef.current = true;
+    useChatRuntimeStore.getState().setModelLoading(true);
+    void (async () => {
+      try {
+        await unloadModel({ model_path: model.id }).catch(() => {});
+      } finally {
+        cancelUnloadPendingRef.current = false;
+        if (!loadingModelRef.current) {
+          useChatRuntimeStore.getState().setModelLoading(false);
+        }
+      }
+    })();
   }, [clearCheckpoint, setLoadToastDismissedState]);
 
   const selectModel = useCallback(
@@ -437,11 +427,13 @@ export function useChatModelRuntime() {
             if (validation.requires_trust_remote_code && !trustRemoteCode) {
               throw new Error(getTrustRemoteCodeRequiredMessage(displayName));
             }
+            if (abortCtrl.signal.aborted) throw new Error("Cancelled");
 
             if (currentCheckpoint) {
               await unloadModel({ model_path: currentCheckpoint });
               previousWasUnloaded = true;
             }
+            if (abortCtrl.signal.aborted) throw new Error("Cancelled");
 
             const {
               chatTemplateOverride,

From 2de17c0a962f49de0fa0401deaf4f3537ceec165 Mon Sep 17 00:00:00 2001
From: Lee Jackson <130007945+Imagineer99@users.noreply.github.com>
Date: Sun, 3 May 2026 21:34:46 +0100
Subject: [PATCH 194/216] Studio: Add checkpoint resume for stopped training
 runs (#5255)

* feat: add checkpoint resume for stopped training runs

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

* fix:add resume checkpoint helpers

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

* fix: use checkpoint parent as resume output dir

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

* fix: save optimizer and scheduler state on stop-and-save

Use Trainer._save_checkpoint instead of save_state so resume restores
optimizer momentum and LR-schedule position via the checkpoint-NNN/
subdir written by HF's official path.

* fix: clean up resume training history and startup progress

* fix: preserve resume output dirs

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

* fix: tighten resume run lookup

* fix: remove stale output-dir lookup

* fix: preserve startup download progress

---------

Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com>
Co-authored-by: Roland Tannous <rolandtannous@gravityq.ai>
Co-authored-by: Roland Tannous <115670425+rolandtannous@users.noreply.github.com>
---
 studio/backend/core/training/resume.py        | 75 +++++++++++++++++
 studio/backend/core/training/trainer.py       | 17 +++-
 studio/backend/core/training/training.py      |  1 +
 studio/backend/core/training/worker.py        | 30 ++++++-
 studio/backend/models/training.py             |  5 ++
 studio/backend/routes/training.py             | 39 +++++++++
 studio/backend/routes/training_history.py     | 13 ++-
 studio/backend/storage/studio_db.py           | 81 +++++++++++++++++--
 .../studio/historical-training-view.tsx       |  1 +
 .../src/features/studio/history-card-grid.tsx | 74 ++++++++++++++++-
 .../features/studio/live-training-view.tsx    |  2 +
 .../src/features/studio/studio-page.tsx       |  3 +
 .../studio/training-start-overlay.tsx         | 56 +++++++++++--
 .../training/hooks/use-training-actions.ts    | 61 ++++++++++++++
 .../training/stores/training-runtime-store.ts |  8 ++
 .../src/features/training/types/api.ts        |  1 +
 .../src/features/training/types/history.ts    |  2 +
 .../src/features/training/types/runtime.ts    | 11 +++
 18 files changed, 456 insertions(+), 24 deletions(-)
 create mode 100644 studio/backend/core/training/resume.py

diff --git a/studio/backend/core/training/resume.py b/studio/backend/core/training/resume.py
new file mode 100644
index 0000000000..165c1c2cf1
--- /dev/null
+++ b/studio/backend/core/training/resume.py
@@ -0,0 +1,75 @@
+# SPDX-License-Identifier: AGPL-3.0-only
+# Copyright 2026-present the Unsloth AI Inc. team. All rights reserved. See /studio/LICENSE.AGPL-3.0
+
+"""Helpers for validating resumable training outputs."""
+
+from pathlib import Path
+from typing import Optional
+
+from utils.paths import outputs_root, resolve_output_dir
+
+
+def _is_under_outputs(path: Path) -> bool:
+    resolved = path.resolve(strict = False)
+    root = outputs_root().resolve(strict = False)
+    try:
+        resolved.relative_to(root)
+        return True
+    except ValueError:
+        return False
+
+
+def has_resume_state(path_value: Optional[str]) -> bool:
+    if not path_value:
+        return False
+    return get_resume_checkpoint_path(path_value) is not None
+
+
+def _checkpoint_step(path: Path) -> int:
+    try:
+        return int(path.name.removeprefix("checkpoint-"))
+    except ValueError:
+        return -1
+
+
+def get_resume_checkpoint_path(path_value: str) -> Optional[str]:
+    path = resolve_output_dir(path_value)
+    if not _is_under_outputs(path) or not path.is_dir():
+        return None
+    if (path / "trainer_state.json").is_file():
+        return str(path)
+
+    checkpoints = [
+        child
+        for child in path.glob("checkpoint-*")
+        if child.is_dir() and (child / "trainer_state.json").is_file()
+    ]
+    if not checkpoints:
+        return None
+    return str(max(checkpoints, key = _checkpoint_step))
+
+
+def normalize_resume_output_dir(path_value: str) -> str:
+    path = resolve_output_dir(path_value)
+    if not _is_under_outputs(path):
+        raise ValueError("Resume checkpoint must be inside Studio outputs.")
+    return str(path)
+
+
+def can_resume_run(run: dict) -> bool:
+    if run.get("resumed_later"):
+        return False
+
+    final_step = run.get("final_step")
+    total_steps = run.get("total_steps")
+    has_remaining_steps = (
+        not isinstance(final_step, int)
+        or not isinstance(total_steps, int)
+        or total_steps <= 0
+        or final_step < total_steps
+    )
+    return (
+        run.get("status") == "stopped"
+        and has_remaining_steps
+        and has_resume_state(run.get("output_dir"))
+    )
diff --git a/studio/backend/core/training/trainer.py b/studio/backend/core/training/trainer.py
index c1e2ac4a85..24d06f9f62 100644
--- a/studio/backend/core/training/trainer.py
+++ b/studio/backend/core/training/trainer.py
@@ -376,6 +376,7 @@ class UnslothTrainer:
     def _finalize_training(self, output_dir, label = ""):
         """Save model after training and update progress. Used by all training branches."""
         if self.should_stop and self.save_on_stop:
+            self.trainer._save_checkpoint(self.trainer.model, trial = None)
             self.trainer.save_model()
             self.tokenizer.save_pretrained(output_dir)
             self._patch_adapter_config(output_dir)
@@ -2828,7 +2829,9 @@ class UnslothTrainer:
                     total_steps = total, status_message = "Starting CSM training..."
                 )
                 logger.info(f"CSM training config: {config}\n")
-                self.trainer.train()
+                self.trainer.train(
+                    resume_from_checkpoint = training_args.get("resume_from_checkpoint")
+                )
                 self._finalize_training(output_dir, "CSM")
                 return
 
@@ -2867,7 +2870,9 @@ class UnslothTrainer:
                     total_steps = total, status_message = "Starting SNAC training..."
                 )
                 logger.info(f"SNAC training config: {config}\n")
-                self.trainer.train()
+                self.trainer.train(
+                    resume_from_checkpoint = training_args.get("resume_from_checkpoint")
+                )
                 self._finalize_training(output_dir, "SNAC")
                 return
 
@@ -2913,7 +2918,9 @@ class UnslothTrainer:
                     total_steps = total, status_message = "Starting Whisper training..."
                 )
                 logger.info(f"Whisper training config: {config}\n")
-                self.trainer.train()
+                self.trainer.train(
+                    resume_from_checkpoint = training_args.get("resume_from_checkpoint")
+                )
                 self._finalize_training(output_dir, "Whisper")
                 return
 
@@ -3408,7 +3415,9 @@ class UnslothTrainer:
             # ========== START TRAINING ==========
             self._update_progress(status_message = "Starting training...")
             logger.info("Starting training...\n")
-            self.trainer.train()
+            self.trainer.train(
+                resume_from_checkpoint = training_args.get("resume_from_checkpoint")
+            )
 
             # ========== SAVE MODEL ==========
             self._finalize_training(output_dir)
diff --git a/studio/backend/core/training/training.py b/studio/backend/core/training/training.py
index f35c7e8ad3..5437645797 100644
--- a/studio/backend/core/training/training.py
+++ b/studio/backend/core/training/training.py
@@ -185,6 +185,7 @@ class TrainingBackend:
             "wandb_project": kwargs.get("wandb_project", "unsloth-training"),
             "enable_tensorboard": kwargs.get("enable_tensorboard", False),
             "tensorboard_dir": kwargs.get("tensorboard_dir", "runs"),
+            "resume_from_checkpoint": kwargs.get("resume_from_checkpoint"),
             "trust_remote_code": kwargs.get("trust_remote_code", False),
             "gpu_ids": kwargs.get("gpu_ids"),
         }
diff --git a/studio/backend/core/training/worker.py b/studio/backend/core/training/worker.py
index ff046c3fc9..60b9e994ab 100644
--- a/studio/backend/core/training/worker.py
+++ b/studio/backend/core/training/worker.py
@@ -35,6 +35,15 @@ from utils.wheel_utils import (
 )
 
 
+def _output_dir_from_resume_checkpoint(
+    resume_from_checkpoint: str | None,
+) -> str | None:
+    if not resume_from_checkpoint:
+        return None
+    path = Path(resume_from_checkpoint)
+    return str(path.parent if path.name.startswith("checkpoint-") else path)
+
+
 _CAUSAL_CONV1D_RELEASE_TAG = "v1.6.1.post4"
 _CAUSAL_CONV1D_PACKAGE_VERSION = "1.6.1"
 _MAMBA_SSM_RELEASE_TAG = "v2.3.1"
@@ -757,7 +766,10 @@ def run_training_process(
             return
 
         # Generate output dir
-        output_dir = config.get("output_dir")
+        resume_from_checkpoint = config.get("resume_from_checkpoint")
+        output_dir = config.get("output_dir") or _output_dir_from_resume_checkpoint(
+            resume_from_checkpoint
+        )
         if not output_dir:
             output_dir = f"{model_name.replace('/', '_')}_{int(time.time())}"
         output_dir = str(resolve_output_dir(output_dir))
@@ -805,6 +817,7 @@ def run_training_process(
             max_seq_length = config.get("max_seq_length", 2048),
             optim = config.get("optim", "adamw_8bit"),
             lr_scheduler_type = config.get("lr_scheduler_type", "linear"),
+            resume_from_checkpoint = resume_from_checkpoint,
         )
 
         _tqdm_stop.set()
@@ -821,10 +834,13 @@ def run_training_process(
                 }
             )
         else:
+            saved_output_dir = (
+                None if trainer.should_stop and not trainer.save_on_stop else output_dir
+            )
             event_queue.put(
                 {
                     "type": "complete",
-                    "output_dir": output_dir,
+                    "output_dir": saved_output_dir,
                     "status_message": progress.status_message or "Training completed",
                     "ts": time.time(),
                 }
@@ -1109,11 +1125,15 @@ def _run_embedding_training(event_queue: Any, stop_queue: Any, config: dict) ->
         )
         return
 
-    output_dir = config.get("output_dir")
+    resume_from_checkpoint = config.get("resume_from_checkpoint")
+    output_dir = config.get("output_dir") or _output_dir_from_resume_checkpoint(
+        resume_from_checkpoint
+    )
     if not output_dir:
         output_dir = str(
             resolve_output_dir(f"{model_name.replace('/', '_')}_{int(time.time())}")
         )
+    output_dir = str(resolve_output_dir(output_dir))
 
     num_epochs = config.get("num_epochs", 2)
     batch_size = config.get("batch_size", 256)
@@ -1221,7 +1241,7 @@ def _run_embedding_training(event_queue: Any, stop_queue: Any, config: dict) ->
             callbacks = [_EmbeddingProgressCallback()],
         )
 
-        trainer.train()
+        trainer.train(resume_from_checkpoint = resume_from_checkpoint)
     except Exception as e:
         event_queue.put(
             {
@@ -1247,6 +1267,8 @@ def _run_embedding_training(event_queue: Any, stop_queue: Any, config: dict) ->
 
     _send_status(event_queue, "Saving model...")
     try:
+        if _should_stop and _save_on_stop:
+            trainer._save_checkpoint(trainer.model, trial = None)
         model.save_pretrained(output_dir)
         model.tokenizer.save_pretrained(output_dir)
         logger.info("Embedding model saved to %s", output_dir)
diff --git a/studio/backend/models/training.py b/studio/backend/models/training.py
index 07a306ca39..a9f4caa1bb 100644
--- a/studio/backend/models/training.py
+++ b/studio/backend/models/training.py
@@ -127,6 +127,9 @@ class TrainingStartRequest(BaseModel):
     wandb_project: Optional[str] = Field(None, description = "W&B project name")
     enable_tensorboard: bool = Field(False, description = "Enable TensorBoard logging")
     tensorboard_dir: Optional[str] = Field(None, description = "TensorBoard directory")
+    resume_from_checkpoint: Optional[str] = Field(
+        None, description = "Saved training output directory to resume from"
+    )
 
     # GPU selection
     gpu_ids: Optional[List[int]] = Field(
@@ -220,6 +223,8 @@ class TrainingRunSummary(BaseModel):
     duration_seconds: Optional[float] = None
     error_message: Optional[str] = None
     loss_sparkline: Optional[List[float]] = None
+    can_resume: bool = False
+    resumed_later: bool = False
 
 
 class TrainingRunListResponse(BaseModel):
diff --git a/studio/backend/routes/training.py b/studio/backend/routes/training.py
index e625408bad..e5195bb337 100644
--- a/studio/backend/routes/training.py
+++ b/studio/backend/routes/training.py
@@ -25,6 +25,12 @@ if str(backend_path) not in sys.path:
 # Import backend functions
 try:
     from core.training import get_training_backend
+    from core.training.resume import (
+        can_resume_run,
+        get_resume_checkpoint_path,
+        normalize_resume_output_dir,
+    )
+    from storage.studio_db import get_resumable_run_by_output_dir
     from utils.models.model_config import load_model_defaults
     from utils.paths import resolve_dataset_path
 except ImportError:
@@ -33,6 +39,12 @@ except ImportError:
     if str(parent_backend) not in sys.path:
         sys.path.insert(0, str(parent_backend))
     from core.training import get_training_backend
+    from core.training.resume import (
+        can_resume_run,
+        get_resume_checkpoint_path,
+        normalize_resume_output_dir,
+    )
+    from storage.studio_db import get_resumable_run_by_output_dir
     from utils.models.model_config import load_model_defaults
     from utils.paths import resolve_dataset_path
 
@@ -152,6 +164,28 @@ async def start_training(
             request.local_eval_datasets = _validate_local_dataset_paths(
                 request.local_eval_datasets, "Local eval dataset"
             )
+        resume_output_dir: Optional[str] = None
+        if request.resume_from_checkpoint:
+            try:
+                resume_output_dir = normalize_resume_output_dir(
+                    request.resume_from_checkpoint
+                )
+            except ValueError as e:
+                raise HTTPException(status_code = 400, detail = str(e))
+
+            resume_run = get_resumable_run_by_output_dir(resume_output_dir)
+            if not resume_run or not can_resume_run(resume_run):
+                raise HTTPException(
+                    status_code = 400,
+                    detail = "Resume checkpoint must belong to a stopped run with saved trainer state.",
+                )
+            resume_checkpoint = get_resume_checkpoint_path(resume_output_dir)
+            if not resume_checkpoint:
+                raise HTTPException(
+                    status_code = 400,
+                    detail = "Resume checkpoint must include saved trainer state.",
+                )
+            request.resume_from_checkpoint = resume_checkpoint
 
         # Convert request to kwargs for backend
         training_kwargs = {
@@ -209,6 +243,8 @@ async def start_training(
             "wandb_project": request.wandb_project or "",
             "enable_tensorboard": request.enable_tensorboard,
             "tensorboard_dir": request.tensorboard_dir or "",
+            "output_dir": resume_output_dir,
+            "resume_from_checkpoint": request.resume_from_checkpoint,
             "trust_remote_code": request.trust_remote_code,
             "gpu_ids": request.gpu_ids,
         }
@@ -437,6 +473,9 @@ async def get_training_status(
                 "loss": getattr(progress, "loss", None),
                 "learning_rate": getattr(progress, "learning_rate", None),
             }
+            output_dir = getattr(backend, "_output_dir", None)
+            if output_dir:
+                details["output_dir"] = output_dir
 
         # Build metric history for chart recovery after SSE reconnection
         metric_history = None
diff --git a/studio/backend/routes/training_history.py b/studio/backend/routes/training_history.py
index 597c4424c0..6f34321959 100644
--- a/studio/backend/routes/training_history.py
+++ b/studio/backend/routes/training_history.py
@@ -11,6 +11,7 @@ from fastapi import APIRouter, Depends, HTTPException, Query
 from loggers import get_logger
 
 from auth.authentication import get_current_subject
+from core.training.resume import can_resume_run
 from models import (
     TrainingRunDeleteResponse,
     TrainingRunDetailResponse,
@@ -34,7 +35,10 @@ async def list_training_runs(
     """List training runs, newest first."""
     result = list_runs(limit = limit, offset = offset)
     return TrainingRunListResponse(
-        runs = [TrainingRunSummary(**r) for r in result["runs"]],
+        runs = [
+            TrainingRunSummary(**{**r, "can_resume": can_resume_run(r)})
+            for r in result["runs"]
+        ],
         total = result["total"],
     )
 
@@ -58,7 +62,12 @@ async def get_training_run_detail(
     metrics_data = get_run_metrics(run_id)
 
     return TrainingRunDetailResponse(
-        run = TrainingRunSummary(**{k: v for k, v in run.items() if k != "config_json"}),
+        run = TrainingRunSummary(
+            **{
+                **{k: v for k, v in run.items() if k != "config_json"},
+                "can_resume": can_resume_run(run),
+            }
+        ),
         config = config,
         metrics = TrainingRunMetrics(**metrics_data),
     )
diff --git a/studio/backend/storage/studio_db.py b/studio/backend/storage/studio_db.py
index 89f75632ef..29e787c196 100644
--- a/studio/backend/storage/studio_db.py
+++ b/studio/backend/storage/studio_db.py
@@ -267,10 +267,23 @@ def list_runs(limit: int = 50, offset: int = 0) -> dict:
         total = conn.execute("SELECT COUNT(*) FROM training_runs").fetchone()[0]
         rows = conn.execute(
             """
-            SELECT id, status, model_name, dataset_name, started_at, ended_at,
-                   total_steps, final_step, final_loss, output_dir,
-                   duration_seconds, error_message, loss_sparkline
-            FROM training_runs
+            SELECT r.id, r.status, r.model_name, r.dataset_name, r.started_at,
+                   r.ended_at, r.total_steps, r.final_step, r.final_loss,
+                   r.output_dir, r.duration_seconds, r.error_message,
+                   r.loss_sparkline,
+                   CASE
+                       WHEN r.status = 'stopped'
+                            AND r.output_dir IS NOT NULL
+                            AND EXISTS (
+                                SELECT 1
+                                FROM training_runs newer
+                                WHERE newer.output_dir = r.output_dir
+                                  AND newer.status IN ('stopped', 'completed')
+                                  AND newer.started_at > r.started_at
+                            )
+                       THEN 1 ELSE 0
+                   END AS resumed_later
+            FROM training_runs r
             ORDER BY started_at DESC
             LIMIT ? OFFSET ?
             """,
@@ -297,7 +310,26 @@ def list_runs(limit: int = 50, offset: int = 0) -> dict:
 def get_run(id: str) -> Optional[dict]:
     conn = get_connection()
     try:
-        row = conn.execute("SELECT * FROM training_runs WHERE id = ?", (id,)).fetchone()
+        row = conn.execute(
+            """
+            SELECT r.*,
+                   CASE
+                       WHEN r.status = 'stopped'
+                            AND r.output_dir IS NOT NULL
+                            AND EXISTS (
+                                SELECT 1
+                                FROM training_runs newer
+                                WHERE newer.output_dir = r.output_dir
+                                  AND newer.status IN ('stopped', 'completed')
+                                  AND newer.started_at > r.started_at
+                            )
+                       THEN 1 ELSE 0
+                   END AS resumed_later
+            FROM training_runs r
+            WHERE r.id = ?
+            """,
+            (id,),
+        ).fetchone()
         if row is None:
             return None
         run = dict(row)
@@ -313,6 +345,45 @@ def get_run(id: str) -> Optional[dict]:
         conn.close()
 
 
+def get_resumable_run_by_output_dir(output_dir: str) -> Optional[dict]:
+    conn = get_connection()
+    try:
+        row = conn.execute(
+            """
+            SELECT r.*,
+                   0 AS resumed_later
+            FROM training_runs r
+            WHERE r.output_dir = ?
+              AND r.status = 'stopped'
+              AND NOT EXISTS (
+                  SELECT 1
+                  FROM training_runs newer
+                  WHERE newer.output_dir = r.output_dir
+                    AND newer.status IN ('stopped', 'completed')
+                    AND newer.started_at > r.started_at
+              )
+            ORDER BY r.started_at DESC
+            LIMIT 1
+            """,
+            (output_dir,),
+        ).fetchone()
+        if row is None:
+            return None
+        run = dict(row)
+        sparkline = run.get("loss_sparkline")
+        if sparkline:
+            try:
+                run["loss_sparkline"] = json.loads(sparkline)
+            except (json.JSONDecodeError, TypeError):
+                logger.debug(
+                    "Failed to parse loss_sparkline for output_dir %s", output_dir
+                )
+                run["loss_sparkline"] = None
+        return run
+    finally:
+        conn.close()
+
+
 def get_run_metrics(id: str) -> dict:
     """Return metric arrays for a run, using paired step arrays per metric."""
     conn = get_connection()
diff --git a/studio/frontend/src/features/studio/historical-training-view.tsx b/studio/frontend/src/features/studio/historical-training-view.tsx
index d3ac434bd8..e461fc5a90 100644
--- a/studio/frontend/src/features/studio/historical-training-view.tsx
+++ b/studio/frontend/src/features/studio/historical-training-view.tsx
@@ -60,6 +60,7 @@ function mapToViewData(detail: TrainingRunDetailResponse): TrainingViewData {
     currentGradNorm: metrics.grad_norm_history.at(-1) ?? null,
     currentEpoch: metrics.final_epoch,
     currentNumTokens: metrics.final_num_tokens ?? null,
+    outputDir: run.output_dir ?? null,
     progressPercent:
       run.total_steps && run.final_step
         ? (run.final_step / run.total_steps) * 100
diff --git a/studio/frontend/src/features/studio/history-card-grid.tsx b/studio/frontend/src/features/studio/history-card-grid.tsx
index 9e72c9f75c..6fbdffce85 100644
--- a/studio/frontend/src/features/studio/history-card-grid.tsx
+++ b/studio/frontend/src/features/studio/history-card-grid.tsx
@@ -13,7 +13,12 @@ import {
 } from "@/components/ui/alert-dialog";
 import { Button } from "@/components/ui/button";
 import type { TrainingRunSummary } from "@/features/training";
-import { deleteTrainingRun, listTrainingRuns } from "@/features/training";
+import {
+  deleteTrainingRun,
+  listTrainingRuns,
+  useTrainingActions,
+  useTrainingRuntimeStore,
+} from "@/features/training";
 import { formatDuration } from "@/features/studio/sections/progress-section-lib";
 import { cn } from "@/lib/utils";
 import { Delete02Icon } from "@hugeicons/core-free-icons";
@@ -47,8 +52,28 @@ const statusBadge: Record<
     className:
       "bg-blue-100 text-blue-700 dark:bg-blue-950 dark:text-blue-400",
   },
+  resumed_later: {
+    label: "Continued",
+    className:
+      "bg-emerald-100 text-emerald-700 dark:bg-emerald-950 dark:text-emerald-400",
+  },
 };
 
+function wasContinuedInVisibleRuns(
+  run: TrainingRunSummary,
+  runs: TrainingRunSummary[],
+): boolean {
+  if (run.status !== "stopped" || !run.output_dir) return false;
+  const startedAt = new Date(run.started_at).getTime();
+  return runs.some(
+    (other) =>
+      other.id !== run.id &&
+      other.output_dir === run.output_dir &&
+      (other.status === "stopped" || other.status === "completed") &&
+      new Date(other.started_at).getTime() > startedAt,
+  );
+}
+
 function catmullRomPath(points: { x: number; y: number }[]): string {
   if (points.length < 2) return "";
   const d = [`M${points[0]!.x.toFixed(1)},${points[0]!.y.toFixed(1)}`];
@@ -133,10 +158,12 @@ function formatRelativeTime(isoDate: string): string {
 
 interface HistoryCardGridProps {
   onSelectRun: (runId: string) => void;
+  onResumeStarted?: () => void;
 }
 
 export function HistoryCardGrid({
   onSelectRun,
+  onResumeStarted,
 }: HistoryCardGridProps): ReactElement {
   const [runs, setRuns] = useState<TrainingRunSummary[]>([]);
   const [total, setTotal] = useState(0);
@@ -144,7 +171,10 @@ export function HistoryCardGrid({
   const [error, setError] = useState<string | null>(null);
   const [deleteError, setDeleteError] = useState<string | null>(null);
   const [deleteTarget, setDeleteTarget] = useState<string | null>(null);
+  const [resumeTarget, setResumeTarget] = useState<string | null>(null);
   const [manualFetchInFlight, setManualFetchInFlight] = useState(false);
+  const { resumeTrainingRunFromHistory } = useTrainingActions();
+  const isStarting = useTrainingRuntimeStore((state) => state.isStarting);
 
   const userControllerRef = useRef<AbortController | null>(null);
   const pollControllerRef = useRef<AbortController | null>(null);
@@ -233,6 +263,18 @@ export function HistoryCardGrid({
     setDeleteTarget(null);
   };
 
+  const handleResume = async (runId: string) => {
+    setResumeTarget(runId);
+    try {
+      const ok = await resumeTrainingRunFromHistory(runId);
+      if (ok) {
+        onResumeStarted?.();
+      }
+    } finally {
+      setResumeTarget(null);
+    }
+  };
+
   if (!loading && error && runs.length === 0) {
     return (
       <div className="flex flex-col items-center gap-2 py-16 text-center">
@@ -264,18 +306,25 @@ export function HistoryCardGrid({
       )}
       <div className="grid grid-cols-1 gap-4 sm:grid-cols-2 xl:grid-cols-3">
         {runs.map((run) => {
-          const badge = statusBadge[run.status] ?? statusBadge.error;
+          const wasContinued =
+            run.resumed_later || wasContinuedInVisibleRuns(run, runs);
+          const badge = wasContinued
+            ? statusBadge.resumed_later
+            : (statusBadge[run.status] ?? statusBadge.error);
           const isRunning = run.status === "running";
+          const canResume = run.can_resume && !wasContinued;
+          const isResuming = resumeTarget === run.id;
           return (
             <div
               role="button"
               tabIndex={0}
               key={run.id}
               className={cn(
-                "group relative flex cursor-pointer flex-col gap-3 rounded-xl border bg-card p-4 text-left transition-colors hover:border-border hover:bg-accent/30",
+                "group relative flex h-[11.5rem] cursor-pointer flex-col gap-3 rounded-xl border bg-card p-4 text-left transition-colors hover:border-border hover:bg-accent/30",
                 isRunning
                   ? "border-blue-400/50 dark:border-blue-500/30"
                   : "border-border/60",
+                canResume && "gap-2",
               )}
               onClick={() => onSelectRun(run.id)}
               onKeyDown={(e) => {
@@ -299,6 +348,21 @@ export function HistoryCardGrid({
                   {formatRelativeTime(run.started_at)}
                 </span>
               </div>
+              {canResume && (
+                <Button
+                  type="button"
+                  size="xs"
+                  variant="outline"
+                  className="absolute bottom-3 left-4 h-6 rounded-full px-2.5 text-[11px] leading-none shadow-sm"
+                  disabled={isStarting || isResuming}
+                  onClick={(e) => {
+                    e.stopPropagation();
+                    void handleResume(run.id);
+                  }}
+                >
+                  {isResuming ? "Resuming..." : "Resume training"}
+                </Button>
+              )}
               <div className="min-w-0">
                 <p
                   className="truncate text-sm font-medium"
@@ -311,7 +375,9 @@ export function HistoryCardGrid({
                 </p>
               </div>
               {run.loss_sparkline && run.loss_sparkline.length >= 2 && (
-                <Sparkline values={run.loss_sparkline} id={run.id} />
+                <div className={cn(canResume && "h-7 overflow-hidden")}>
+                  <Sparkline values={run.loss_sparkline} id={run.id} />
+                </div>
               )}
               <div className="flex flex-wrap gap-x-4 gap-y-1 text-[11px] text-muted-foreground">
                 <span>
diff --git a/studio/frontend/src/features/studio/live-training-view.tsx b/studio/frontend/src/features/studio/live-training-view.tsx
index 9f930ce77b..e355655d54 100644
--- a/studio/frontend/src/features/studio/live-training-view.tsx
+++ b/studio/frontend/src/features/studio/live-training-view.tsx
@@ -31,6 +31,7 @@ export function LiveTrainingView(): ReactElement {
       elapsedSeconds: state.elapsedSeconds,
       etaSeconds: state.etaSeconds,
       evalEnabled: state.evalEnabled,
+      outputDir: state.outputDir,
       isTrainingRunning: state.isTrainingRunning,
       lossHistory: state.lossHistory,
       lrHistory: state.lrHistory,
@@ -57,6 +58,7 @@ export function LiveTrainingView(): ReactElement {
     currentGradNorm: runtime.currentGradNorm,
     currentEpoch: runtime.currentEpoch,
     currentNumTokens: runtime.currentNumTokens,
+    outputDir: runtime.outputDir,
     progressPercent: runtime.progressPercent,
     elapsedSeconds: runtime.elapsedSeconds,
     etaSeconds: runtime.etaSeconds,
diff --git a/studio/frontend/src/features/studio/studio-page.tsx b/studio/frontend/src/features/studio/studio-page.tsx
index 5018dc3baa..8cfd39d5c1 100644
--- a/studio/frontend/src/features/studio/studio-page.tsx
+++ b/studio/frontend/src/features/studio/studio-page.tsx
@@ -209,6 +209,9 @@ export function StudioPage(): ReactElement {
                   } else {
                     setSelectedHistoryRunId(runId);
                   }
+                }} onResumeStarted={() => {
+                  setSelectedHistoryRunId(null);
+                  handleTabChange("current-run");
                 }} />
               )}
             </TabsContent>
diff --git a/studio/frontend/src/features/studio/training-start-overlay.tsx b/studio/frontend/src/features/studio/training-start-overlay.tsx
index b6042d4ae6..27c5a13163 100644
--- a/studio/frontend/src/features/studio/training-start-overlay.tsx
+++ b/studio/frontend/src/features/studio/training-start-overlay.tsx
@@ -64,6 +64,22 @@ const EMPTY_DOWNLOAD_STATE: DownloadState = {
   cachePath: null,
 };
 
+function coerceCachedStateReady(state: DownloadState): DownloadState {
+  if (!state.cachePath) return state;
+  if (state.downloadedBytes > 0 && state.percent < 100) return state;
+  const totalBytes =
+    state.totalBytes > 0 ? state.totalBytes : state.downloadedBytes;
+  if (totalBytes <= 0) {
+    return { ...state, percent: 100 };
+  }
+  return {
+    ...state,
+    downloadedBytes: totalBytes,
+    totalBytes,
+    percent: 100,
+  };
+}
+
 type Fetcher = (repoId: string) => Promise<DownloadProgressResponse>;
 
 /**
@@ -88,10 +104,12 @@ function useHfDownloadProgress(
     phase === "downloading_model" ||
     phase === "downloading_dataset" ||
     phase === "loading_model" ||
-    phase === "loading_dataset";
+    phase === "loading_dataset" ||
+    phase === "training";
 
   useEffect(() => {
     if (!repoId || !HF_REPO_REGEX.test(repoId) || !shouldPoll) {
+      setState(EMPTY_DOWNLOAD_STATE);
       return;
     }
 
@@ -229,14 +247,42 @@ export function TrainingStartOverlay({
 }: TrainingStartOverlayProps): ReactElement {
   const { stopTrainingRun, dismissTrainingRun } = useTrainingActions();
   const isStarting = useTrainingRuntimeStore((s) => s.isStarting);
-  const selectedModel = useTrainingConfigStore((s) => s.selectedModel);
+  const phase = useTrainingRuntimeStore((s) => s.phase);
+  const startModelName = useTrainingRuntimeStore((s) => s.startModelName);
+  const startDatasetName = useTrainingRuntimeStore((s) => s.startDatasetName);
+  const startFromResume = useTrainingRuntimeStore((s) => s.startFromResume);
+  const configuredModel = useTrainingConfigStore((s) => s.selectedModel);
   const datasetSource = useTrainingConfigStore((s) => s.datasetSource);
   const dataset = useTrainingConfigStore((s) => s.dataset);
   // Only HF datasets have a download phase to track. Uploaded files are
   // already on disk by the time the overlay shows up.
   const hfDatasetName = datasetSource === "huggingface" ? dataset : null;
-  const modelDownload = useModelDownloadProgress(selectedModel);
-  const datasetDownload = useDatasetDownloadProgress(hfDatasetName);
+  const hasStartResources = startModelName !== null;
+  const useConfiguredResources = !isStarting && !hasStartResources;
+  const isDownloadPhase =
+    phase === "downloading_model" || phase === "downloading_dataset";
+  const modelName = hasStartResources
+    ? startModelName
+    : useConfiguredResources
+      ? configuredModel
+      : null;
+  const datasetName = hasStartResources
+    ? startDatasetName
+    : useConfiguredResources
+      ? hfDatasetName
+      : null;
+  const displayMessage =
+    startFromResume && !isDownloadPhase && /^download/i.test(message)
+      ? "Resuming training..."
+      : message || "starting training...";
+  const rawModelDownload = useModelDownloadProgress(modelName);
+  const rawDatasetDownload = useDatasetDownloadProgress(datasetName);
+  const modelDownload = isDownloadPhase
+    ? rawModelDownload
+    : coerceCachedStateReady(rawModelDownload);
+  const datasetDownload = isDownloadPhase
+    ? rawDatasetDownload
+    : coerceCachedStateReady(rawDatasetDownload);
   const [cancelDialogOpen, setCancelDialogOpen] = useState(false);
   const [cancelRequested, setCancelRequested] = useState(false);
 
@@ -314,7 +360,7 @@ export function TrainingStartOverlay({
             {"> We are getting everything ready for your run..."}
           </TypingAnimation>
           <AnimatedSpan className="mt-2 text-muted-foreground">
-            {`> ${message || "starting training..."} | waiting for first step... (${currentStep})`}
+            {`> ${displayMessage} | waiting for first step... (${currentStep})`}
           </AnimatedSpan>
           {datasetDownload.downloadedBytes > 0 || datasetDownload.cachePath ? (
             <AnimatedSpan className="mt-3">
diff --git a/studio/frontend/src/features/training/hooks/use-training-actions.ts b/studio/frontend/src/features/training/hooks/use-training-actions.ts
index b5b41e7a0d..7f275d5d82 100644
--- a/studio/frontend/src/features/training/hooks/use-training-actions.ts
+++ b/studio/frontend/src/features/training/hooks/use-training-actions.ts
@@ -3,6 +3,7 @@
 
 import { useCallback } from "react";
 import { checkDatasetFormat } from "../api/datasets-api";
+import { getTrainingRun } from "../api/history-api";
 import { buildTrainingStartPayload } from "../api/mappers";
 import { startTraining, stopTraining, resetTraining } from "../api/train-api";
 import { syncTrainingRuntimeFromBackend } from "../lib/sync-runtime";
@@ -10,6 +11,7 @@ import { validateTrainingConfig } from "../lib/validation";
 import { useDatasetPreviewDialogStore } from "../stores/dataset-preview-dialog-store";
 import { useTrainingConfigStore } from "../stores/training-config-store";
 import { useTrainingRuntimeStore } from "../stores/training-runtime-store";
+import type { TrainingStartRequest } from "../types/api";
 import type { TrainingConfigState } from "../types/config";
 import { toast } from "sonner";
 
@@ -48,6 +50,11 @@ export function useTrainingActions() {
       return false;
     }
 
+    runtimeStore.setStartResources(
+      config.selectedModel ?? null,
+      getHfDatasetName(config),
+      false,
+    );
     runtimeStore.setStarting(true);
 
     try {
@@ -114,6 +121,7 @@ export function useTrainingActions() {
 
       // Re-read config after potential store updates from dataset check
       const payload = buildTrainingStartPayload(useTrainingConfigStore.getState());
+      runtimeStore.setStartResources(payload.model_name, payload.hf_dataset, false);
       const response = await startTraining(payload);
 
       if (response.status === "error") {
@@ -153,6 +161,54 @@ export function useTrainingActions() {
     }
   }, []);
 
+  const resumeTrainingRunFromHistory = useCallback(async (runId: string): Promise<boolean> => {
+    const runtimeStore = useTrainingRuntimeStore.getState();
+    runtimeStore.setStartError(null);
+    runtimeStore.setStartResources(null, null, true);
+    runtimeStore.setStarting(true);
+
+    try {
+      const detail = await getTrainingRun(runId);
+      const outputDir = detail.run.output_dir;
+      if (!detail.run.can_resume || !outputDir) {
+        throw new Error("Only stopped runs with a saved checkpoint can be resumed.");
+      }
+
+      const config = useTrainingConfigStore.getState();
+      const savedConfig = detail.config as Partial<TrainingStartRequest>;
+      const payload = {
+        ...savedConfig,
+        hf_token:
+          typeof savedConfig.hf_token === "string"
+            ? savedConfig.hf_token
+            : config.hfToken.trim() || null,
+        wandb_token: null,
+        resume_from_checkpoint: outputDir,
+      } as TrainingStartRequest;
+
+      runtimeStore.setStartResources(payload.model_name, payload.hf_dataset, true);
+
+      const response = await startTraining(payload);
+      if (response.status === "error") {
+        throw new Error(response.error || response.message);
+      }
+
+      runtimeStore.setStartQueued(response.job_id, response.message);
+      await syncTrainingRuntimeFromBackend();
+      return true;
+    } catch (error) {
+      const rawMessage =
+        error instanceof Error ? error.message : "Failed to resume training";
+      const safeMessage = normalizeTrainingStartError(rawMessage);
+      runtimeStore.setStartError(safeMessage);
+      runtimeStore.setStarting(false);
+      toast.error("Could not resume training", {
+        description: safeMessage,
+      });
+      return false;
+    }
+  }, []);
+
   const dismissTrainingRun = useCallback(async (): Promise<void> => {
     try {
       await resetTraining();
@@ -173,6 +229,7 @@ export function useTrainingActions() {
     isStarting,
     startError,
     startTrainingRun,
+    resumeTrainingRunFromHistory,
     stopTrainingRun,
     dismissTrainingRun,
   };
@@ -184,6 +241,10 @@ function getDatasetName(config: TrainingConfigState): string | null {
     : config.uploadedFile;
 }
 
+function getHfDatasetName(config: TrainingConfigState): string | null {
+  return config.datasetSource === "huggingface" ? config.dataset : null;
+}
+
 function hasManualMapping(config: TrainingConfigState, isVlm = false, isAudio = false): boolean {
   const mapping = config.datasetManualMapping;
   const roles = new Set(Object.values(mapping));
diff --git a/studio/frontend/src/features/training/stores/training-runtime-store.ts b/studio/frontend/src/features/training/stores/training-runtime-store.ts
index d958078330..da496be002 100644
--- a/studio/frontend/src/features/training/stores/training-runtime-store.ts
+++ b/studio/frontend/src/features/training/stores/training-runtime-store.ts
@@ -22,6 +22,9 @@ const initialState: TrainingRuntimeState = {
   hasHydrated: false,
   isStarting: false,
   startError: null,
+  startModelName: null,
+  startDatasetName: null,
+  startFromResume: false,
   sseConnected: false,
   firstStepReceived: false,
   lastEventId: null,
@@ -35,6 +38,7 @@ const initialState: TrainingRuntimeState = {
   etaSeconds: null,
   currentGradNorm: null,
   currentNumTokens: null,
+  outputDir: null,
   lossHistory: [],
   lrHistory: [],
   gradNormHistory: [],
@@ -120,6 +124,8 @@ export const useTrainingRuntimeStore = create<TrainingRuntimeStore>()((set) => (
   setHasHydrated: (value) => set({ hasHydrated: value }),
   setStarting: (value) => set({ isStarting: value }),
   setStartError: (value) => set({ startError: value }),
+  setStartResources: (startModelName, startDatasetName, startFromResume = false) =>
+    set({ startModelName, startDatasetName, startFromResume }),
   setSseConnected: (value) => set({ sseConnected: value }),
   setLastEventId: (value) => set({ lastEventId: value }),
 
@@ -156,6 +162,7 @@ export const useTrainingRuntimeStore = create<TrainingRuntimeStore>()((set) => (
       etaSeconds: null,
       currentGradNorm: null,
       currentNumTokens: null,
+      outputDir: null,
       lossHistory: [],
       lrHistory: [],
       gradNormHistory: [],
@@ -207,6 +214,7 @@ export const useTrainingRuntimeStore = create<TrainingRuntimeStore>()((set) => (
           typeof detailLr === "number" ? detailLr : state.currentLearningRate,
         currentEpoch:
           typeof detailEpoch === "number" ? detailEpoch : state.currentEpoch,
+        outputDir: payload.details?.output_dir ?? state.outputDir,
         lossHistory: metricHistory.lossHistory ?? state.lossHistory,
         lrHistory: metricHistory.lrHistory ?? state.lrHistory,
         gradNormHistory: metricHistory.gradNormHistory ?? state.gradNormHistory,
diff --git a/studio/frontend/src/features/training/types/api.ts b/studio/frontend/src/features/training/types/api.ts
index 1d72e9d9d8..ae65d1a53c 100644
--- a/studio/frontend/src/features/training/types/api.ts
+++ b/studio/frontend/src/features/training/types/api.ts
@@ -54,6 +54,7 @@ export interface TrainingStartRequest {
   wandb_project: string | null;
   enable_tensorboard: boolean;
   tensorboard_dir: string | null;
+  resume_from_checkpoint?: string | null;
 }
 
 export interface TrainingStartResponse {
diff --git a/studio/frontend/src/features/training/types/history.ts b/studio/frontend/src/features/training/types/history.ts
index 8b89db539b..901dfe3266 100644
--- a/studio/frontend/src/features/training/types/history.ts
+++ b/studio/frontend/src/features/training/types/history.ts
@@ -12,6 +12,8 @@ export interface TrainingRunSummary {
   final_step: number | null;
   final_loss: number | null;
   output_dir: string | null;
+  can_resume: boolean;
+  resumed_later: boolean;
   duration_seconds: number | null;
   error_message: string | null;
   loss_sparkline: number[] | null;
diff --git a/studio/frontend/src/features/training/types/runtime.ts b/studio/frontend/src/features/training/types/runtime.ts
index 3e4dbd3b79..a62ef9ef26 100644
--- a/studio/frontend/src/features/training/types/runtime.ts
+++ b/studio/frontend/src/features/training/types/runtime.ts
@@ -26,6 +26,7 @@ export interface TrainingStatusResponse {
     total_steps?: number;
     loss?: number;
     learning_rate?: number;
+    output_dir?: string;
   } | null;
   metric_history?: {
     steps?: number[];
@@ -80,6 +81,9 @@ export interface TrainingRuntimeState {
   hasHydrated: boolean;
   isStarting: boolean;
   startError: string | null;
+  startModelName: string | null;
+  startDatasetName: string | null;
+  startFromResume: boolean;
   sseConnected: boolean;
   firstStepReceived: boolean;
   lastEventId: number | null;
@@ -93,6 +97,7 @@ export interface TrainingRuntimeState {
   etaSeconds: number | null;
   currentGradNorm: number | null;
   currentNumTokens: number | null;
+  outputDir: string | null;
   lossHistory: TrainingSeriesPoint[];
   lrHistory: TrainingSeriesPoint[];
   gradNormHistory: TrainingSeriesPoint[];
@@ -108,6 +113,11 @@ export interface TrainingRuntimeActions {
   setHasHydrated: (value: boolean) => void;
   setStarting: (value: boolean) => void;
   setStartError: (value: string | null) => void;
+  setStartResources: (
+    modelName: string | null,
+    datasetName: string | null,
+    fromResume?: boolean,
+  ) => void;
   setSseConnected: (value: boolean) => void;
   setLastEventId: (value: number | null) => void;
   resetRuntime: () => void;
@@ -131,6 +141,7 @@ export interface TrainingViewData {
   currentGradNorm: number | null;
   currentEpoch: number | null;
   currentNumTokens: number | null;
+  outputDir: string | null;
   progressPercent: number;
   elapsedSeconds: number | null;
   etaSeconds: number | null;

From 00b607267a9928435c59312c59476d9fc2d2fa8e Mon Sep 17 00:00:00 2001
From: Lee Jackson <130007945+Imagineer99@users.noreply.github.com>
Date: Sun, 3 May 2026 21:38:19 +0100
Subject: [PATCH 195/216] Studio: Polish spacing and profile input radius
 (#5222)

* UI: polish studio spacing and profile input radius

* chore: use standard rounded radius for profile input

---------

Co-authored-by: Roland Tannous <115670425+rolandtannous@users.noreply.github.com>
---
 studio/frontend/src/features/export/export-page.tsx             | 2 +-
 .../profile/components/profile-personalization-panel.tsx        | 2 +-
 studio/frontend/src/features/studio/studio-page.tsx             | 2 +-
 3 files changed, 3 insertions(+), 3 deletions(-)

diff --git a/studio/frontend/src/features/export/export-page.tsx b/studio/frontend/src/features/export/export-page.tsx
index 0d5d9c813c..046d9efac4 100644
--- a/studio/frontend/src/features/export/export-page.tsx
+++ b/studio/frontend/src/features/export/export-page.tsx
@@ -524,7 +524,7 @@ export function ExportPage() {
   // ---- Render ----
   return (
     <div className="min-h-[calc(100dvh-var(--studio-titlebar-height,0px))] bg-background">
-      <main className="mx-auto max-w-7xl px-4 py-4 sm:px-6">
+      <main className="mx-auto max-w-7xl px-4 py-8 sm:px-6">
         <GuidedTour {...tour.tourProps} />
 
         <div className="mb-8 flex flex-col gap-0.5">
diff --git a/studio/frontend/src/features/profile/components/profile-personalization-panel.tsx b/studio/frontend/src/features/profile/components/profile-personalization-panel.tsx
index 994115bc19..f30af7fcf9 100644
--- a/studio/frontend/src/features/profile/components/profile-personalization-panel.tsx
+++ b/studio/frontend/src/features/profile/components/profile-personalization-panel.tsx
@@ -139,7 +139,7 @@ export function ProfilePersonalizationPanel() {
             }}
             autoComplete="off"
             placeholder={sessionSub || "Unsloth"}
-            className="h-10 min-w-0 flex-1 rounded-lg text-sm"
+            className="h-10 min-w-0 flex-1 rounded-full text-sm"
           />
           <Button type="button" size="sm" className="h-10 px-5" onClick={saveName} disabled={!hasNameChanges}>
             Save
diff --git a/studio/frontend/src/features/studio/studio-page.tsx b/studio/frontend/src/features/studio/studio-page.tsx
index 8cfd39d5c1..2affe7415a 100644
--- a/studio/frontend/src/features/studio/studio-page.tsx
+++ b/studio/frontend/src/features/studio/studio-page.tsx
@@ -128,7 +128,7 @@ export function StudioPage(): ReactElement {
 
   return (
     <div className="relative min-h-[calc(100dvh-var(--studio-titlebar-height,0px))] bg-background">
-      <main className="relative z-10 mx-auto max-w-7xl px-4 py-4 sm:px-6">
+      <main className="relative z-10 mx-auto max-w-7xl px-4 py-8 sm:px-6">
         <GuidedTour {...tour.tourProps} celebrate={isConfigTour} />
 
         <DatasetPreviewDialog

From e1b00854a5d09af924a64683022188fde78920c3 Mon Sep 17 00:00:00 2001
From: LFdev <146497073+LFd3v@users.noreply.github.com>
Date: Sun, 3 May 2026 17:43:18 -0300
Subject: [PATCH 196/216] Fix check for libcurl hearders in install.sh (#5251)

* Fix check for libcurl hearders in install.sh

Checking for `dpkg` has no relation to libcurl headers at all. If the package is installed, then an executable `curl-config` is available in Ubuntu/Debian as it can be seen here:

https://ubuntu.pkgs.org/26.04/ubuntu-main-amd64/libcurl4-openssl-dev_8.18.0-1ubuntu2_amd64.deb.html

This also fix the installation in ArchLinux, provided the needed packages are installed previously as shown in the error message when a package is missing.

* Fix check for libcurl hearders in studio/setup.sh

Use the same check as install.sh.

---------

Co-authored-by: Roland Tannous <115670425+rolandtannous@users.noreply.github.com>
---
 install.sh      | 4 +---
 studio/setup.sh | 2 +-
 2 files changed, 2 insertions(+), 4 deletions(-)

diff --git a/install.sh b/install.sh
index a9861a108f..4f17ea80cb 100755
--- a/install.sh
+++ b/install.sh
@@ -974,9 +974,7 @@ case "$OS" in
         fi
         command -v gcc  >/dev/null 2>&1 || MISSING="$MISSING build-essential"
         # libcurl dev headers for llama.cpp HTTPS support
-        if command -v dpkg >/dev/null 2>&1; then
-            dpkg -s libcurl4-openssl-dev >/dev/null 2>&1 || MISSING="$MISSING libcurl4-openssl-dev"
-        fi
+        command -v curl-config >/dev/null 2>&1 || MISSING="$MISSING libcurl4-openssl-dev"
         ;;
 esac
 
diff --git a/studio/setup.sh b/studio/setup.sh
index 9575f143d5..39fb566f1b 100755
--- a/studio/setup.sh
+++ b/studio/setup.sh
@@ -699,7 +699,7 @@ if [ "$_NEED_LLAMA_SOURCE_BUILD" = true ] && grep -qi microsoft /proc/version 2>
         case "$_pkg" in
             build-essential) command -v gcc >/dev/null 2>&1 || _STILL_MISSING="$_STILL_MISSING $_pkg" ;;
             pciutils) command -v lspci >/dev/null 2>&1 || _STILL_MISSING="$_STILL_MISSING $_pkg" ;;
-            libcurl4-openssl-dev) dpkg -s "$_pkg" >/dev/null 2>&1 || _STILL_MISSING="$_STILL_MISSING $_pkg" ;;
+            libcurl4-openssl-dev) command -v curl-config >/dev/null 2>&1 || _STILL_MISSING="$_STILL_MISSING $_pkg" ;;
             *) command -v "$_pkg" >/dev/null 2>&1 || _STILL_MISSING="$_STILL_MISSING $_pkg" ;;
         esac
     done

From 35ab5da93ccfcde5f2743b0c569377e27f2b9bab Mon Sep 17 00:00:00 2001
From: Roland Tannous <115670425+rolandtannous@users.noreply.github.com>
Date: Mon, 4 May 2026 13:03:16 +0400
Subject: [PATCH 197/216] Default Studio host to 127.0.0.1 and prompt before
 auto-start (#5267)

Studio bound to 0.0.0.0 by default and the installer silently auto-started
a server at end of install, exposing it on the network without consent and
contradicting the privacy-first / local-only guarantee.

- studio/backend/run.py: run_server() and argparse --host default to 127.0.0.1
- unsloth_cli/commands/studio.py: studio_default() and run() --host default to 127.0.0.1
- install.sh: drop -H 0.0.0.0 from generated launcher template; replace silent
  auto-start with a [Y/n] prompt; add cloud/network note to manual hint
- install.ps1: drop -H 0.0.0.0 from PowerShell launcher template; replace
  silent auto-start with a Read-Host [Y/n] prompt; add cloud/network note
- studio/setup.sh: drop -H 0.0.0.0 from launch hint; add cloud/network note
- README.md: simplify launch examples to `unsloth studio -p 8888`; note
  -H 0.0.0.0 is available for cloud/LAN use

Tests:
- studio/backend/tests/test_host_defaults.py
- tests/studio/test_cli_studio_defaults.py
- tests/sh/test_install_host_defaults.sh
---
 README.md                                  |  15 +--
 install.ps1                                |  20 +++-
 install.sh                                 |  56 +++++++----
 studio/backend/run.py                      |   8 +-
 studio/backend/tests/test_host_defaults.py |  98 +++++++++++++++++++
 studio/setup.sh                            |   5 +-
 tests/sh/test_install_host_defaults.sh     | 104 +++++++++++++++++++++
 tests/studio/test_cli_studio_defaults.py   |  87 +++++++++++++++++
 unsloth_cli/commands/studio.py             |   4 +-
 9 files changed, 360 insertions(+), 37 deletions(-)
 create mode 100644 studio/backend/tests/test_host_defaults.py
 create mode 100755 tests/sh/test_install_host_defaults.sh
 create mode 100644 tests/studio/test_cli_studio_defaults.py

diff --git a/README.md b/README.md
index e1b7e448ef..6c12b28096 100644
--- a/README.md
+++ b/README.md
@@ -79,8 +79,9 @@ irm https://unsloth.ai/install.ps1 | iex
 
 #### Launch
 ```bash
-unsloth studio -H 0.0.0.0 -p 8888
+unsloth studio -p 8888
 ```
+> For cloud VMs or LAN access, add `-H 0.0.0.0` to bind on all interfaces.
 
 #### Update
 To update, use the same install commands as above. Or run (does not work on Windows):
@@ -167,7 +168,7 @@ The below advanced instructions are for Unsloth Studio. For Unsloth Core advance
 git clone https://github.com/unslothai/unsloth
 cd unsloth
 ./install.sh --local
-unsloth studio -H 0.0.0.0 -p 8888
+unsloth studio -p 8888
 ```
 Then to update :
 ```bash
@@ -180,7 +181,7 @@ git clone https://github.com/unslothai/unsloth.git
 cd unsloth
 Set-ExecutionPolicy -Scope Process -ExecutionPolicy Bypass
 .\install.ps1 --local
-unsloth studio -H 0.0.0.0 -p 8888
+unsloth studio -p 8888
 ```
 Then to update :
 ```bash
@@ -193,11 +194,11 @@ git clone https://github.com/unslothai/unsloth
 cd unsloth
 git checkout nightly
 ./install.sh --local
-unsloth studio -H 0.0.0.0 -p 8888
+unsloth studio -p 8888
 ```
 Then to launch every time:
 ```bash
-unsloth studio -H 0.0.0.0 -p 8888
+unsloth studio -p 8888
 ```
 
 #### Nightly: Windows:
@@ -208,11 +209,11 @@ cd unsloth
 git checkout nightly
 Set-ExecutionPolicy -Scope Process -ExecutionPolicy Bypass
 .\install.ps1 --local
-unsloth studio -H 0.0.0.0 -p 8888
+unsloth studio -p 8888
 ```
 Then to launch every time:
 ```bash
-unsloth studio -H 0.0.0.0 -p 8888
+unsloth studio -p 8888
 ```
 
 #### Uninstall
diff --git a/install.ps1 b/install.ps1
index ab43bcdc52..7dc5a50250 100644
--- a/install.ps1
+++ b/install.ps1
@@ -552,7 +552,7 @@ try {
         } catch {}
         exit 1
     }
-    `$studioCommand = '& "' + `$studioExe + '" studio -H 0.0.0.0 -p ' + `$launchPort
+    `$studioCommand = '& "' + `$studioExe + '" studio -p ' + `$launchPort
     `$launchArgs = @(
         '-NoExit',
         '-NoProfile',
@@ -1344,15 +1344,25 @@ shell.Run cmd, 0, False
 
     New-StudioShortcuts -UnslothExePath $UnslothExe
 
-    # Launch studio automatically in interactive terminals;
-    # in non-interactive environments (CI, Docker) just print instructions.
+    # In interactive terminals, ask the user before starting Studio.
+    # In non-interactive environments (CI, Docker) just print instructions.
     $IsInteractive = [Environment]::UserInteractive -and (-not [Console]::IsInputRedirected)
     if ($IsInteractive) {
-        & $UnslothExe studio -H 0.0.0.0 -p 8888
+        Write-Host ""
+        $reply = Read-Host "  Start Unsloth Studio now? [Y/n]"
+        if ([string]::IsNullOrWhiteSpace($reply) -or $reply -match '^[Yy]') {
+            & $UnslothExe studio -p 8888
+        } else {
+            step "launch" "to start later, run:"
+            substep "unsloth studio -p 8888"
+            substep "(add -H 0.0.0.0 to allow network / cloud access)"
+            Write-Host ""
+        }
     } else {
         step "launch" "manual commands:"
         substep "& `"$VenvDir\Scripts\Activate.ps1`""
-        substep "unsloth studio -H 0.0.0.0 -p 8888"
+        substep "unsloth studio -p 8888"
+        substep "(add -H 0.0.0.0 to allow network / cloud access)"
         Write-Host ""
     }
 }
diff --git a/install.sh b/install.sh
index 4f17ea80cb..7948170043 100755
--- a/install.sh
+++ b/install.sh
@@ -622,11 +622,11 @@ if [ -t 1 ]; then
     ) &
     # Clear traps so exec does not trigger _release_lock (the subshell owns it)
     trap - EXIT INT TERM
-    exec "$UNSLOTH_EXE" studio -H 0.0.0.0 -p "$_launch_port"
+    exec "$UNSLOTH_EXE" studio -p "$_launch_port"
 else
     # ── Background mode (no TTY) ──
     # Used by macOS .app and headless invocations.
-    _launch_cmd=$(printf '%q ' "$UNSLOTH_EXE" studio -H 0.0.0.0 -p "$_launch_port")
+    _launch_cmd=$(printf '%q ' "$UNSLOTH_EXE" studio -p "$_launch_port")
     _launch_cmd=${_launch_cmd% }
     _spawn_terminal "$_launch_cmd"
 
@@ -1847,28 +1847,46 @@ printf "  ${C_TITLE}%s${C_RST}\n" "Unsloth Studio installed!"
 printf "  ${C_DIM}%s${C_RST}\n" "$RULE"
 echo ""
 
-# Launch studio automatically in interactive terminals;
-# in non-interactive environments (Docker, CI, cloud-init) just print instructions.
+# In interactive terminals, ask the user before starting Studio.
+# In non-interactive environments (Docker, CI, cloud-init) just print instructions.
 if [ -t 1 ]; then
-    step "launch" "starting Unsloth Studio..."
-    "$VENV_DIR/bin/unsloth" studio -H 0.0.0.0 -p 8888
-    _LAUNCH_EXIT=$?
-    if [ "$_LAUNCH_EXIT" -ne 0 ] && [ "$_MIGRATED" = true ]; then
-        echo ""
-        echo "⚠️  Unsloth Studio failed to start after migration."
-        echo "   Your migrated environment may be incompatible."
-        echo "   To fix, remove the environment and reinstall:"
-        echo ""
-        echo "   rm -rf $VENV_DIR"
-        echo "   curl -fsSL https://unsloth.ai/install.sh | sh"
-        echo ""
+    echo ""
+    printf "  Start Unsloth Studio now? [Y/n] "
+    if [ -r /dev/tty ]; then
+        read -r _reply </dev/tty || _reply="y"
+    else
+        _reply="y"
     fi
-    exit "$_LAUNCH_EXIT"
+    case "${_reply:-y}" in
+        [Yy]*|"")
+            step "launch" "starting Unsloth Studio..."
+            "$VENV_DIR/bin/unsloth" studio -p 8888
+            _LAUNCH_EXIT=$?
+            if [ "$_LAUNCH_EXIT" -ne 0 ] && [ "$_MIGRATED" = true ]; then
+                echo ""
+                echo "⚠️  Unsloth Studio failed to start after migration."
+                echo "   Your migrated environment may be incompatible."
+                echo "   To fix, remove the environment and reinstall:"
+                echo ""
+                echo "   rm -rf $VENV_DIR"
+                echo "   curl -fsSL https://unsloth.ai/install.sh | sh"
+                echo ""
+            fi
+            exit "$_LAUNCH_EXIT"
+            ;;
+        *)
+            step "launch" "to start later, run:"
+            substep "unsloth studio -p 8888"
+            substep "(add -H 0.0.0.0 to allow network / cloud access)"
+            echo ""
+            ;;
+    esac
 else
     step "launch" "manual commands:"
-    substep "unsloth studio -H 0.0.0.0 -p 8888"
+    substep "unsloth studio -p 8888"
     substep "or activate env first:"
     substep "source ${VENV_DIR}/bin/activate"
-    substep "unsloth studio -H 0.0.0.0 -p 8888"
+    substep "unsloth studio -p 8888"
+    substep "(add -H 0.0.0.0 to allow network / cloud access)"
     echo ""
 fi
diff --git a/studio/backend/run.py b/studio/backend/run.py
index 7590ef1067..c5b103ff70 100644
--- a/studio/backend/run.py
+++ b/studio/backend/run.py
@@ -244,7 +244,7 @@ _shutdown_event = None
 
 
 def run_server(
-    host: str = "0.0.0.0",
+    host: str = "127.0.0.1",
     port: int = 8888,
     frontend_path: Path = Path(__file__).resolve().parent.parent / "frontend" / "dist",
     silent: bool = False,
@@ -392,7 +392,11 @@ if __name__ == "__main__":
             pass
 
     parser = argparse.ArgumentParser(description = "Run Unsloth UI Backend server")
-    parser.add_argument("--host", default = "0.0.0.0", help = "Host to bind to")
+    parser.add_argument(
+        "--host",
+        default = "127.0.0.1",
+        help = "Host to bind to (default: 127.0.0.1; use 0.0.0.0 for network/cloud access)",
+    )
     parser.add_argument("--port", type = int, default = 8888, help = "Port to bind to")
     parser.add_argument(
         "--frontend",
diff --git a/studio/backend/tests/test_host_defaults.py b/studio/backend/tests/test_host_defaults.py
new file mode 100644
index 0000000000..8b81474e92
--- /dev/null
+++ b/studio/backend/tests/test_host_defaults.py
@@ -0,0 +1,98 @@
+# SPDX-License-Identifier: AGPL-3.0-only
+# Copyright 2026-present the Unsloth AI Inc. team. All rights reserved. See /studio/LICENSE.AGPL-3.0
+
+"""Tests that Unsloth Studio defaults to 127.0.0.1 (loopback) not 0.0.0.0.
+
+Uses AST parsing to inspect source-level defaults without requiring the
+full studio venv (run.py has heavy dependencies like structlog/uvicorn).
+"""
+
+import ast
+from pathlib import Path
+
+_RUN_PY = Path(__file__).resolve().parent.parent / "run.py"
+
+
+def _parse_function_param_defaults(source: str, func_name: str) -> dict:
+    """Return {param_name: default_value} for a named function in *source*.
+
+    Only handles ast.Constant defaults (strings, ints, bools).
+    """
+    tree = ast.parse(source)
+    for node in ast.walk(tree):
+        if (
+            isinstance(node, (ast.FunctionDef, ast.AsyncFunctionDef))
+            and node.name == func_name
+        ):
+            result = {}
+            all_args = node.args.args
+            defaults = node.args.defaults
+            # Defaults are right-aligned against the args list
+            offset = len(all_args) - len(defaults)
+            for i, default in enumerate(defaults):
+                arg_name = all_args[offset + i].arg
+                if isinstance(default, ast.Constant):
+                    result[arg_name] = default.value
+            return result
+    return {}
+
+
+def _parse_argparse_add_argument_default(source: str, option_name: str):
+    """Return the 'default' kwarg value for add_argument(option_name, ...) in *source*.
+
+    Walks the entire module so the call can live in __main__ or in a helper
+    function — only handles ast.Constant defaults.
+    """
+    tree = ast.parse(source)
+    for node in ast.walk(tree):
+        if not isinstance(node, ast.Call):
+            continue
+        func = node.func
+        if not (isinstance(func, ast.Attribute) and func.attr == "add_argument"):
+            continue
+        if not node.args:
+            continue
+        first_arg = node.args[0]
+        if not (isinstance(first_arg, ast.Constant) and first_arg.value == option_name):
+            continue
+        for kw in node.keywords:
+            if kw.arg == "default" and isinstance(kw.value, ast.Constant):
+                return kw.value.value
+    return None
+
+
+def test_run_server_default_host_is_loopback():
+    """run_server() parameter default for 'host' must be 127.0.0.1, not 0.0.0.0.
+
+    Binding to 0.0.0.0 by default exposes the service on all network
+    interfaces, contradicting the documented "privacy first / 100% local"
+    guarantee.  Loopback (127.0.0.1) is the least-permissive default;
+    users who need network access can pass -H 0.0.0.0 explicitly.
+    """
+    source = _RUN_PY.read_text()
+    defaults = _parse_function_param_defaults(source, "run_server")
+    assert (
+        "host" in defaults
+    ), "run_server() must have a 'host' parameter with a default"
+    host_default = defaults["host"]
+    assert host_default == "127.0.0.1", (
+        f"run_server() host default must be '127.0.0.1' (loopback) "
+        f"but got '{host_default}'. Binding to '{host_default}' by default "
+        f"exposes the service beyond localhost."
+    )
+
+
+def test_argparse_default_host_is_loopback():
+    """argparse --host add_argument default must be 127.0.0.1.
+
+    When run.py is invoked directly (python run.py), the argparse default
+    should match the function default so direct execution is equally safe.
+    """
+    source = _RUN_PY.read_text()
+    host_default = _parse_argparse_add_argument_default(source, "--host")
+    assert (
+        host_default is not None
+    ), "Could not find add_argument('--host', ...) in run.py"
+    assert (
+        host_default == "127.0.0.1"
+    ), f"run.py argparse --host default must be '127.0.0.1', got '{host_default}'"
diff --git a/studio/setup.sh b/studio/setup.sh
index 39fb566f1b..3e875eed30 100755
--- a/studio/setup.sh
+++ b/studio/setup.sh
@@ -1087,10 +1087,11 @@ else
     fi
     printf "  ${C_DIM}%s${C_RST}\n" "$RULE"
     if [ "$_LLAMA_CPP_DEGRADED" = true ]; then
-        printf "  ${C_DIM}%-15s${C_WARN}%s${C_RST}\n" "launch" "unsloth studio -H 0.0.0.0 -p 8888"
+        printf "  ${C_DIM}%-15s${C_WARN}%s${C_RST}\n" "launch" "unsloth studio -p 8888"
     else
-        printf "  ${C_DIM}%-15s${C_OK}%s${C_RST}\n" "launch" "unsloth studio -H 0.0.0.0 -p 8888"
+        printf "  ${C_DIM}%-15s${C_OK}%s${C_RST}\n" "launch" "unsloth studio -p 8888"
     fi
+    printf "  ${C_DIM}%-15s%s${C_RST}\n" "" "(add -H 0.0.0.0 to allow network / cloud access)"
 fi
 echo ""
 
diff --git a/tests/sh/test_install_host_defaults.sh b/tests/sh/test_install_host_defaults.sh
new file mode 100755
index 0000000000..7b01eaa716
--- /dev/null
+++ b/tests/sh/test_install_host_defaults.sh
@@ -0,0 +1,104 @@
+#!/bin/bash
+# Static analysis: installer scripts and README must not hard-code 0.0.0.0
+# in any user-visible default launch command. The dynamic-port launcher
+# templates and post-install hints should rely on the new loopback default.
+set -e
+
+SCRIPT_DIR="$(cd "$(dirname "$0")" && pwd)"
+INSTALL_SH="$SCRIPT_DIR/../../install.sh"
+INSTALL_PS1="$SCRIPT_DIR/../../install.ps1"
+SETUP_SH="$SCRIPT_DIR/../../studio/setup.sh"
+README="$SCRIPT_DIR/../../README.md"
+PASS=0
+FAIL=0
+
+assert_contains() {
+    _label="$1"; _haystack="$2"; _needle="$3"
+    if echo "$_haystack" | grep -qF -- "$_needle"; then
+        echo "  PASS: $_label"
+        PASS=$((PASS + 1))
+    else
+        echo "  FAIL: $_label (expected to find '$_needle')"
+        FAIL=$((FAIL + 1))
+    fi
+}
+
+assert_not_contains() {
+    _label="$1"; _haystack="$2"; _needle="$3"
+    if echo "$_haystack" | grep -qF -- "$_needle"; then
+        echo "  FAIL: $_label (found '$_needle' but should not)"
+        FAIL=$((FAIL + 1))
+    else
+        echo "  PASS: $_label"
+        PASS=$((PASS + 1))
+    fi
+}
+
+echo ""
+echo "=== install.sh launcher template ==="
+
+# Extract the heredoc that generates ~/.local/share/unsloth/launch-studio.sh.
+_launcher=$(awk '/cat > "\$_css_launcher"/{found=1} found{print} /^LAUNCHER_EOF$/{found=0}' "$INSTALL_SH")
+assert_contains \
+    "launcher template: extraction found the heredoc content" \
+    "$_launcher" "#!/usr/bin/env bash"
+# The desktop launcher should rely on the new 127.0.0.1 default.
+assert_not_contains \
+    "launcher template: no hardcoded 'studio -H 0.0.0.0'" \
+    "$_launcher" "studio -H 0.0.0.0"
+
+echo ""
+echo "=== install.sh end-of-install block ==="
+
+_end=$(tail -50 "$INSTALL_SH")
+assert_contains \
+    "install.sh: interactive block prompts user (read)" \
+    "$_end" "read"
+assert_not_contains \
+    "install.sh: no 'studio -H 0.0.0.0' in end-of-install commands" \
+    "$_end" "studio -H 0.0.0.0"
+
+echo ""
+echo "=== install.ps1 end-of-install block ==="
+
+_ps1_end=$(tail -25 "$INSTALL_PS1")
+assert_contains \
+    "install.ps1: interactive block prompts user (Read-Host)" \
+    "$_ps1_end" "Read-Host"
+assert_not_contains \
+    "install.ps1: no 'studio -H 0.0.0.0' in end-of-install commands" \
+    "$_ps1_end" "studio -H 0.0.0.0"
+
+echo ""
+echo "=== studio/setup.sh launch hint ==="
+
+_setup_tail=$(tail -30 "$SETUP_SH")
+assert_not_contains \
+    "studio/setup.sh: launch hint has no '-H 0.0.0.0'" \
+    "$_setup_tail" "studio -H 0.0.0.0"
+
+echo ""
+echo "=== README.md Launch section ==="
+
+# The primary Launch example must not include -H 0.0.0.0; the LAN/cloud
+# note appears as an opt-in line outside the code block.
+_readme_launch=$(awk '/^#### Launch$/{found=1} found{print} /^#### Update$/{found=0}' "$README")
+assert_contains \
+    "README: Launch section exists" \
+    "$_readme_launch" "unsloth studio"
+assert_not_contains \
+    "README: Launch section primary command has no -H 0.0.0.0" \
+    "$_readme_launch" "studio -H 0.0.0.0"
+assert_contains \
+    "README: Launch section documents -H 0.0.0.0 opt-in" \
+    "$_readme_launch" "0.0.0.0"
+
+echo ""
+echo "=== Results ==="
+echo "  PASS: $PASS"
+echo "  FAIL: $FAIL"
+if [ "$FAIL" -gt 0 ]; then
+    echo "FAILED"
+    exit 1
+fi
+echo "ALL PASSED"
diff --git a/tests/studio/test_cli_studio_defaults.py b/tests/studio/test_cli_studio_defaults.py
new file mode 100644
index 0000000000..4d6f28f075
--- /dev/null
+++ b/tests/studio/test_cli_studio_defaults.py
@@ -0,0 +1,87 @@
+"""Tests that the 'unsloth studio' CLI defaults to 127.0.0.1.
+
+Uses AST parsing to inspect source-level defaults without requiring the
+full unsloth_cli dependencies (typer/pydantic) at test-collection time.
+"""
+
+import ast
+from pathlib import Path
+
+_STUDIO_CMD_PY = (
+    Path(__file__).resolve().parents[2] / "unsloth_cli" / "commands" / "studio.py"
+)
+
+
+def _find_typer_option_default(source: str, func_name: str, long_option: str):
+    """Return the default value of a typer.Option(...) parameter in *func_name*.
+
+    Matches by the long option name (e.g. '--host') among the positional args
+    of the typer.Option() call and returns the first positional arg (the
+    default value). Only handles ast.Constant defaults.
+    """
+    tree = ast.parse(source)
+    for func_node in ast.walk(tree):
+        if not isinstance(func_node, (ast.FunctionDef, ast.AsyncFunctionDef)):
+            continue
+        if func_node.name != func_name:
+            continue
+        # Walk both regular args and kwonly args, each paired with its default.
+        all_args = func_node.args.args + func_node.args.kwonlyargs
+        all_defaults = func_node.args.defaults + [
+            d for d in func_node.args.kw_defaults if d is not None
+        ]
+        # ast pads defaults right-aligned against args (ignoring kwonly). We
+        # iterate calls directly, which is simpler and robust.
+        for default in all_defaults:
+            if not isinstance(default, ast.Call):
+                continue
+            call_func = default.func
+            is_typer_option = (
+                isinstance(call_func, ast.Attribute)
+                and call_func.attr == "Option"
+                and isinstance(call_func.value, ast.Name)
+                and call_func.value.id == "typer"
+            )
+            if not is_typer_option:
+                continue
+            # First positional is the default value; remaining positionals are
+            # option flags like "--host", "-H".
+            if not default.args:
+                continue
+            flags = [
+                a.value
+                for a in default.args[1:]
+                if isinstance(a, ast.Constant) and isinstance(a.value, str)
+            ]
+            if long_option not in flags:
+                continue
+            first = default.args[0]
+            if isinstance(first, ast.Constant):
+                return first.value
+    return None
+
+
+def test_studio_default_host_is_loopback():
+    """`unsloth studio` (studio_default) --host typer Option default must be 127.0.0.1."""
+    source = _STUDIO_CMD_PY.read_text()
+    host_default = _find_typer_option_default(source, "studio_default", "--host")
+    assert (
+        host_default is not None
+    ), "Could not find --host typer.Option default in studio_default()"
+    assert host_default == "127.0.0.1", (
+        f"studio_default() --host default must be '127.0.0.1' (loopback) "
+        f"but got '{host_default}'."
+    )
+
+
+def test_studio_run_host_is_loopback():
+    """`unsloth studio run` --host typer Option default must be 127.0.0.1."""
+    source = _STUDIO_CMD_PY.read_text()
+    host_default = _find_typer_option_default(source, "run", "--host")
+    assert (
+        host_default is not None
+    ), "Could not find --host typer.Option default in run()"
+    assert host_default == "127.0.0.1", (
+        f"`unsloth studio run` --host default must be '127.0.0.1' (loopback) "
+        f"but got '{host_default}'."
+    )
diff --git a/unsloth_cli/commands/studio.py b/unsloth_cli/commands/studio.py
index b74f42674d..ac25c8805d 100644
--- a/unsloth_cli/commands/studio.py
+++ b/unsloth_cli/commands/studio.py
@@ -414,7 +414,7 @@ def _load_model_via_http(
 def studio_default(
     ctx: typer.Context,
     port: int = typer.Option(8888, "--port", "-p"),
-    host: str = typer.Option("0.0.0.0", "--host", "-H"),
+    host: str = typer.Option("127.0.0.1", "--host", "-H"),
     frontend: Optional[Path] = typer.Option(None, "--frontend", "-f"),
     silent: bool = typer.Option(False, "--silent", "-q"),
     api_only: bool = typer.Option(
@@ -528,7 +528,7 @@ def run(
         "cli", "--api-key-name", help = "Label for the auto-generated API key"
     ),
     port: int = typer.Option(8888, "--port", "-p"),
-    host: str = typer.Option("0.0.0.0", "--host", "-H"),
+    host: str = typer.Option("127.0.0.1", "--host", "-H"),
     frontend: Optional[Path] = typer.Option(None, "--frontend", "-f"),
     silent: bool = typer.Option(False, "--silent", "-q"),
 ):

From e35cbfb4548b8365fcfda1efd737e2b7da229b35 Mon Sep 17 00:00:00 2001
From: Wasim Yousef Said <wasimysdev@gmail.com>
Date: Mon, 4 May 2026 11:46:18 +0200
Subject: [PATCH 198/216] Add native GGUF intake to Studio (#5246)

* feat(studio): add Tauri native GGUF intake

* feat(studio): polish native GGUF intake

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

* fix(studio): load backend helpers during local setup

* fix(studio): acquire native load lease before unload

* Studio: harden native path lease verification and Tauri intake

- Wrap path.resolve(strict=True) and Path.stat() in NativePathLeaseError so a deleted or unmounted GGUF returns 400 instead of leaking the full filesystem path through the generic load_model/validate_model handler.
- Re-apply _reject_network_or_device_path to the resolved canonical path for defense in depth after symlink resolution.
- Replace try/except ValueError pattern in the device-path guard with Path.is_relative_to; the previous shape silently swallowed NativePathLeaseError (which subclasses ValueError) so /dev,/proc,/sys were never actually rejected.
- Broaden the lease redaction regex and dict-key check (Python and Rust diagnostics) to cover both native_path_lease and nativePathLease so the camelCase form emitted by Tauri/frontend payloads is also redacted.
- Hoist the redact_native_paths import to module top in loggers/handlers; the recursive filter no longer pays a per-record import lookup.
- Persist activeNativePathToken in the chat runtime store so the rollback branch can mint a fresh lease and reload the previous native GGUF when a new load fails after unload; clear it in clearCheckpoint and overwrite it on each successful load.
- use-native-drop: read options through a ref so the Tauri onDragDropEvent listener is registered once and stays attached across option changes; reject ambiguous multi-file drops up front instead of silently registering only the first GGUF.
- pick_native_model: use an async pick_file with a tokio oneshot channel instead of blocking_pick_file so the Tokio worker is not held for the duration of the OS dialog.
- registerNativeModelPath: drop the duplicate sourceKind argument; the Rust command parameter is source_kind.
- install_python_stack: insert the script directory (studio/) on sys.path; the previous insert pointed at studio/backend/ which does not satisfy `from backend.utils.wheel_utils import ...`.

* install_python_stack: keep _BACKEND_DIR on sys.path

Restore the studio/backend insertion. Although the immediately following `from backend.utils.wheel_utils import (...)` is satisfied by studio/ already being on sys.path[0] when invoked as `python studio/install_python_stack.py`, wheel_utils itself runs `from utils.native_path_leases import ...`, which requires studio/backend/ to be importable. Without the backend insertion, the existing tests/python/test_install_python_stack.py collection fails with ModuleNotFoundError: No module named 'utils'.

* Studio: tighten native path lease lifecycle and Tauri intake IPC

- register_native_model_path now hardcodes NativePathSourceKind::Drop on the Rust side and the frontend stops sending source_kind. The previous JS payload (source_kind only) never reached the Rust deserializer because Tauri's default ArgumentCase::Camel maps the Rust parameter source_kind to the JS key sourceKind, so drag/drop registration silently failed. Hardcoding the source kind also keeps audit metadata trustworthy on this command.
- Add native_path_secret_removed_for_child_start context manager and wrap multiprocessing.Process.start() at the inference, export, training, and data-recipe job spawn sites. The previous wrapper-only scrub left UNSLOTH_STUDIO_NATIVE_PATH_LEASE_SECRET visible to spawn-platform import-time worker code. The wrapper run_without_native_path_secret stays as defense-in-depth inside the child.
- Stop passing exc_info=True from the native-grant load/validate error logs in routes/inference.py. The structlog filter_sensitive_data processor runs before the renderer, so ConsoleRenderer formatted tracebacks bypassed redaction; the redacted str(e) preserves the message text.
- Replace the os.path.normcase string equality on the resolved canonical path with Path.samefile (with a normcase fallback) so Windows leases that differ only in extended-length \\?\ prefix or short-name spelling are accepted.
- Wrap consumeNativePathToken in its own try/catch in the chat runtime rollback. If the previous native-model token has aged out of TOKEN_TTL we now surface a clear modelsError instead of silently swallowing the rollback inside the outer catch.
- Reject non-ASCII lease strings in _split_lease and convert UnicodeEncodeError / binascii.Error / ValueError raised by _b64decode into NativePathLeaseError so verify_native_path_lease never escapes raw exceptions to the route handler.
- Tighten dropStateForPaths to mark multi-file payloads invalid so the overlay matches the post-fix drop handler that rejects the same payload.
- Replace the one-shot fetch in useNativePathLeasesSupported with a delayed-retry loop so the picker/drop becomes available once the backend is up rather than staying disabled for the rest of the session after a transient failure.
- Drop the unused setActiveNativePathToken setter; the value is set via setState directly in use-chat-model-runtime.
- Add a toast on auto-load failure in use-native-drop so a collapsed model selector does not hide the error.
- Burn the lease nonce before _validate_current_stat so a stat-failed lease is single-use even if a later state change happens to match the original size/mtime.

* Studio: cache lease secret, harden native path stat checks, polish intake UX

- Cache the decoded UNSLOTH_STUDIO_NATIVE_PATH_LEASE_SECRET on first verify and validate that it is base64-decodable and at least 32 bytes. Subsequent _decode_secret calls return from the cache and never touch os.environ, so concurrent /api/inference/load and /api/health requests no longer race with native_path_secret_removed_for_child_start scrubbing the env. native_path_leases_supported now wraps _decode_secret so the health flag matches what verify_native_path_lease actually accepts.
- Replace path.is_file()/is_dir() + path.stat() with os.lstat() in _validate_current_stat and explicitly reject S_ISLNK; size and mtime checks now refer to the link itself, closing the same-size+same-mtime symlink-swap window that the prior follow-symlink stat() left open.
- Add an issued_at_ms < expires_at_ms sanity check in _validate_payload to reject internally inconsistent (HMAC-protected) lease payloads.
- Sort _NATIVE_PATH_REDACTIONS by length (descending) before iterating in redact_native_paths so a longer registered path is replaced before a shorter prefix path; otherwise logs containing /foo/X.gguf.bak after only /foo/X.gguf was registered would leak the .bak suffix.
- classify_existing_path now re-checks the canonical path with symlink_metadata after canonicalize, so a regular file that is replaced with a symlink in the small canonicalize window is rejected at registration.
- ModelSelector renders the local file picker as its own block (not in the eject ternary), so a user with an active model can still replace it via the picker rather than only via drag/drop.
- useNativePathLeasesSupported caps the readiness probe at MAX_READINESS_POLLS (60 = ~5 minutes) and aborts the in-flight fetch on unmount via AbortController, so a permanently-disabled backend stops generating sustained traffic and hot-reload no longer leaks open connections.
- useChooseNativeModel returns a stable useCallback closure and guards the OS dialog with a useRef so rapid double-clicks cannot open multiple dialogs and orphan Rust tokens.
- Branch the multi-file drop toast: if no GGUF was present we say "Only .gguf model files can be dropped here." and otherwise "Drop a single .gguf model file." so users dropping non-GGUF attachments get an accurate explanation.

* native_path_leases: lstat the signed canonical path before resolving

The earlier change to lstat inside _validate_current_stat operates on grant.canonical_path, which is the post-resolve target. If the user atomically replaces the originally-signed file with a symlink to a different file of identical size and mtime, path.resolve(strict=True) follows the symlink, samefile returns True (both ends share the new inode), and the lstat in _validate_current_stat sees the regular target file rather than the symlink, so the swap goes undetected.

Add an os.lstat on the signed canonical path before path.resolve(strict=True), and reject S_ISLNK there. The lstat in _validate_current_stat stays as defense-in-depth for swaps that occur strictly between resolve and stat.

* Studio: scrub native lease secret before mp.Queue spawn and tighten lease lifecycle

- Move _CTX.Queue / _CTX.Event / _CTX.Process construction inside native_path_secret_removed_for_child_start at the inference, export, training and data-recipe spawn sites. The first Queue creation lazily spawns Python's multiprocessing.resource_tracker child, so when it ran outside the scrub context the tracker process inherited the lease secret. Reproduced via the proc filesystem environ entry; the wrapped order keeps the tracker clean.
- native_path_secret_removed_for_child_start now refcounts entries: the env var is popped on the first entry and restored only when the last context exits. Concurrent training/inference/export starts no longer serialize on the env lock across the entire proc.start yield, while still guaranteeing the env stays empty for the duration of every overlapping spawn.
- run_without_native_path_secret now also nulls the module-level cached lease secret. With the existing spawn-only multiprocessing context the cache is irrelevant in practice, but a future fork caller would otherwise inherit the in-memory secret even though the env var was scrubbed.
- filter_sensitive_data now applies the native lease key check on the top-level event_dict, not only on nested dicts, so a logger call that includes a lease value as a top-level keyword field actually redacts it (the bare value does not match the prefix-anchored regex).
- chat-page loadNativeModelIntent now passes intent.id to clearModelIntent so a second drag-drop during an in-flight first auto-load is not wiped from the chip area when the first resolves.
- Bump useNativePathLeasesSupported's MAX_READINESS_POLLS from 60 to 720 so first-run installs that compile llama.cpp from source or download large CUDA wheels (well past 5 minutes) don't permanently disable the native picker.

* native_path_leases: serialize first-decode against scrub context

_decode_secret used a separate _SECRET_INIT_LOCK from the env scrub's _NATIVE_PATH_ENV_LOCK, so the very first decode (before the cache is populated) could race a concurrent native_path_secret_removed_for_child_start and read os.environ during the env-empty window, raising "Native path grants require the managed desktop backend." Subsequent calls hit the cache and were already safe.

Acquire _NATIVE_PATH_ENV_LOCK around the env read inside _SECRET_INIT_LOCK and fall back to _SCRUB_SAVED_SECRET when the scrub has temporarily popped the env var. Lock ordering (init then env) is consistent with no other caller, so no deadlock.

* Studio: surface native model load errors and harden native path label cache

- Native model load and validate now bubble up the actual exception (with
  paths redacted) and apply the same friendly-error rewrite the non-native
  path uses, so users see "CUDA OOM", "trust_remote_code required", etc.
  instead of a generic "Failed to load native model: <label>".
- run_without_native_path_secret now also nulls _SCRUB_SAVED_SECRET so a
  forked grandchild that imports native_path_leases cannot recover the
  secret via the scrub-aware fallback in _decode_secret.
- _NATIVE_PATH_LABELS now has its own 10000-entry cap independent of the
  100-entry redaction list, so display_label_for_native_path no longer
  falls back to returning the raw canonical path after 101 native paths
  in one session. Redaction list keeps the 100-entry cap for log-scan
  performance.
- _validate_payload now also rejects null bytes in display_label, which
  is echoed back in HTTP responses and log lines.

* Studio: harden native path lease validation and chained native rollback

- child_env_without_native_path_secret now copies os.environ under
  _NATIVE_PATH_ENV_LOCK so a concurrent scrub-context env pop cannot
  raise RuntimeError: dictionary changed size during iteration in a
  background hardware scan or other env reader.
- _validate_payload and grant construction route every signed numeric
  field (version, issued_at_ms, expires_at_ms, size_bytes, modified_ms)
  through new _required_int / _optional_int helpers that wrap raw int()
  ValueError into NativePathLeaseError. The single upstream catcher
  produces 400 instead of 500 for malformed signed payloads.
- verify_native_path_lease now runs _validate_current_stat before
  _consume_nonce, so a transient stat error on the canonical path no
  longer permanently burns the nonce. Concurrent verifies still
  serialize through _consume_nonce, so single-use is preserved.
- Chained native model rollback now restores activeNativePathToken in
  the chat runtime store after a successful rollback loadModel. Without
  this, a second consecutive failed switch could not re-roll-back
  because the store token had been overwritten by the failed attempt.
- validate_model now applies the same not_supported_hints friendly
  rewrite to native model errors that load_model already does, so a
  native .gguf that fails validation with an upstream "is not supported"
  message gets the same actionable wording as the non-native branch.

* Studio: harden native path log redaction, status disclosure, and chip lifecycle

- structlog processor chain now runs format_exc_info before
  filter_sensitive_data so traceback strings are produced (and then
  redacted) rather than passed through as untouched (type, value, tb)
  tuples that the JSON or console renderer formats after the redaction
  filter has already finished.
- native_path_secret_removed_for_child_start clears _CACHED_LEASE_SECRET
  in addition to popping the env var, so a fork during the scrub window
  cannot inherit the cached bytes via the parent's heap. Parent verify
  calls during the window keep working through the existing scrub-aware
  fallback in _decode_secret.
- load_model's except ValueError handler now redacts native paths and
  uses the native model log label when native_grant_backed is true.
  Previously a ValueError raised after lease verification (e.g. from
  ModelConfig.from_identifier or downstream GGUF parsing) returned the
  raw exception string in the HTTP response body.
- llama_cpp_backend now records the native display label at GGUF load
  time, and /api/inference/status prefers it over the redaction store.
  After a Python backend restart the redaction store is empty; the
  attribute keeps the friendly label, and an absolute model_identifier
  with no other label source falls back to the basename so the canonical
  path no longer appears in active_model.
- reveal_path_token uses native "reveal and select" commands on macOS
  (open -R) and Windows (explorer /select,) so the file is highlighted
  in the file manager. Linux keeps the existing parent-directory open.
- Native model rollback that fails because the previous token cannot be
  consumed now throws a rollback-specific Error, and the outer empty
  catch was replaced with one that re-throws the rollback error. The
  rollback-specific message now reaches the user instead of being
  overwritten by the original load error message.
- NativeModelChip tracks the Rust token's expiresAtMs on a single
  setTimeout, disables the Load button at expiry, and relabels it
  "Select again" with an explanatory tooltip so users do not click into
  a guaranteed-failure path after the 15-minute TTL elapses.

* Studio: tighten native artifact policy, mmproj sibling check, and intake UX

- is_open_safe_artifact no longer grants Open for directories. Reveal
  already handles directory navigation, so the change closes the
  attack surface where a macOS .app artifact could be launched via
  open_path_token + open::that_detached.
- Display labels are sanitized in classify_existing_path. Control
  characters in filenames (newlines, tabs, NUL et al.) are replaced
  with spaces and the label is trimmed and capped, so a file named
  with embedded newlines cannot inject forged log lines or scramble
  the UI status panel.
- validate_entry_path skips the size_bytes/modified_ms equality check
  when the operation is Reveal or Open. Cloud-sync agents (Dropbox,
  iCloud Drive, OneDrive) routinely rewrite extended-attribute
  metadata which bumps mtime, and the user expects Reveal/Open to
  remain available for files in synced folders.
- llama_cpp_backend gains a _native_grant_backed flag at GGUF load
  success. /api/inference/status only applies the absolute-path
  basename fallback when that flag is true, so a non-native absolute
  local GGUF still reports its canonical model_identifier and unload
  by identifier keeps working.
- Native vision GGUFs now run through _validate_native_mmproj_companion
  before llama-server starts: the companion mmproj must be a regular
  file, not a symlink, and must live in the same resolved directory as
  the granted GGUF. This stops a hostile sibling or symlinked mmproj
  from being loaded under a single-file lease.
- Chained native rollback restructured: the rollback loadModel + state
  + refresh runs inside its own try/catch that swallows so the outer
  throw error surfaces the ORIGINAL load failure. The native-token
  consume-failure case still throws the rollback-specific message
  early, before the inner block runs, so its actionable guidance is
  preserved.
- Loading-model state and the duplicate-load guard in the chat runtime
  hook now compare both the model id and the native path token. Two
  drops or picks with the same basename in different folders no longer
  silently dedup; the second token is honored.
- chat-page loadNativeModelIntent awaits selectModel before clearing
  the pending intent. If selectModel returns early via dedup or
  throws, the chip and its token stay so the user can retry instead
  of losing the selection.
- NativeModelChip's Reveal button is disabled when the lease has
  expired (Rust would reject it anyway), and the Load button label
  reads "Expired" instead of "Select again" so the disabled element
  no longer promises an action it cannot perform.

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

---------

Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com>
Co-authored-by: Daniel Han <danielhanchen@gmail.com>
---
 .../backend/core/data_recipe/jobs/manager.py  |  19 +-
 .../data_recipe/local_callable_validators.py  |   3 +-
 studio/backend/core/export/orchestrator.py    |  31 +-
 studio/backend/core/inference/audio_codecs.py |   3 +
 studio/backend/core/inference/llama_cpp.py    |   5 +-
 studio/backend/core/inference/orchestrator.py |  35 +-
 studio/backend/core/training/trainer.py       |   3 +
 studio/backend/core/training/training.py      |  32 +-
 studio/backend/loggers/config.py              |   4 +
 studio/backend/loggers/handlers.py            |  26 +-
 studio/backend/main.py                        |   2 +
 studio/backend/models/inference.py            |   6 +
 studio/backend/routes/inference.py            | 218 +++++++-
 studio/backend/utils/hardware/amd.py          |   2 +
 studio/backend/utils/hardware/nvidia.py       |   5 +
 studio/backend/utils/models/model_config.py   |   2 +
 studio/backend/utils/native_path_leases.py    | 406 +++++++++++++++
 studio/backend/utils/transformers_version.py  |   3 +
 studio/backend/utils/wheel_utils.py           |   5 +
 studio/frontend/src/app/provider.tsx          |   2 +
 .../assistant-ui/model-selector.tsx           |  24 +
 .../src/components/assistant-ui/thread.tsx    |  45 +-
 .../src/features/chat/api/chat-api.ts         |   7 +-
 .../frontend/src/features/chat/chat-page.tsx  |  74 +++
 .../chat/hooks/use-chat-model-runtime.ts      |  54 +-
 .../src/features/chat/shared-composer.tsx     |   5 +
 .../chat/stores/chat-runtime-store.ts         |   3 +
 .../frontend/src/features/chat/types/api.ts   |   1 +
 .../src/features/native-intents/api.ts        |  46 ++
 .../components/native-model-chip.tsx          | 107 ++++
 .../components/native-model-drop-overlay.tsx  |  70 +++
 .../native-intents/native-intent-drain.tsx    |  39 ++
 .../src/features/native-intents/store.ts      |  23 +
 .../src/features/native-intents/types.ts      |  39 ++
 .../native-intents/use-native-dialogs.ts      |  68 +++
 .../native-intents/use-native-drop.ts         | 132 +++++
 .../native-intents/use-native-readiness.ts    |  46 ++
 studio/install_python_stack.py                |   4 +
 studio/src-tauri/Cargo.lock                   |  80 +++
 studio/src-tauri/Cargo.toml                   |   4 +
 studio/src-tauri/src/diagnostics/redaction.rs |  19 +
 studio/src-tauri/src/main.rs                  |  12 +
 studio/src-tauri/src/native_backend_lease.rs  | 199 ++++++++
 studio/src-tauri/src/native_intents.rs        | 476 ++++++++++++++++++
 studio/src-tauri/src/native_path_policy.rs    | 291 +++++++++++
 studio/src-tauri/src/process.rs               |   9 +-
 46 files changed, 2594 insertions(+), 95 deletions(-)
 create mode 100644 studio/backend/utils/native_path_leases.py
 create mode 100644 studio/frontend/src/features/native-intents/api.ts
 create mode 100644 studio/frontend/src/features/native-intents/components/native-model-chip.tsx
 create mode 100644 studio/frontend/src/features/native-intents/components/native-model-drop-overlay.tsx
 create mode 100644 studio/frontend/src/features/native-intents/native-intent-drain.tsx
 create mode 100644 studio/frontend/src/features/native-intents/store.ts
 create mode 100644 studio/frontend/src/features/native-intents/types.ts
 create mode 100644 studio/frontend/src/features/native-intents/use-native-dialogs.ts
 create mode 100644 studio/frontend/src/features/native-intents/use-native-drop.ts
 create mode 100644 studio/frontend/src/features/native-intents/use-native-readiness.ts
 create mode 100644 studio/src-tauri/src/native_backend_lease.rs
 create mode 100644 studio/src-tauri/src/native_intents.rs
 create mode 100644 studio/src-tauri/src/native_path_policy.rs

diff --git a/studio/backend/core/data_recipe/jobs/manager.py b/studio/backend/core/data_recipe/jobs/manager.py
index db59d573a9..cdc28d9560 100644
--- a/studio/backend/core/data_recipe/jobs/manager.py
+++ b/studio/backend/core/data_recipe/jobs/manager.py
@@ -167,13 +167,20 @@ class JobManager:
 
             run_payload = dict(run)
             run_payload["_job_id"] = job_id
-            mp_q = _CTX.Queue()
-            proc = _CTX.Process(
-                target = run_job_process,
-                kwargs = {"event_queue": mp_q, "recipe": recipe, "run": run_payload},
-                daemon = True,
+            from utils.native_path_leases import (
+                native_path_secret_removed_for_child_start,
+                run_without_native_path_secret,
             )
-            proc.start()
+
+            with native_path_secret_removed_for_child_start():
+                mp_q = _CTX.Queue()
+                proc = _CTX.Process(
+                    target = run_without_native_path_secret,
+                    args = (run_job_process,),
+                    kwargs = {"event_queue": mp_q, "recipe": recipe, "run": run_payload},
+                    daemon = True,
+                )
+                proc.start()
 
             self._mp_q = mp_q
             self._proc = proc
diff --git a/studio/backend/core/data_recipe/local_callable_validators.py b/studio/backend/core/data_recipe/local_callable_validators.py
index afd10b02d1..44459e88c5 100644
--- a/studio/backend/core/data_recipe/local_callable_validators.py
+++ b/studio/backend/core/data_recipe/local_callable_validators.py
@@ -33,6 +33,7 @@ _OXC_TOOL_DIR = Path(__file__).resolve().parent / "oxc-validator"
 _OXC_RUNNER_PATH = _OXC_TOOL_DIR / "validate.mjs"
 
 
+from utils.native_path_leases import child_env_without_native_path_secret
 from utils.subprocess_compat import (
     windows_hidden_subprocess_kwargs as _windows_hidden_subprocess_kwargs,
 )
@@ -248,7 +249,7 @@ def _run_oxc_batch(
     }
     try:
         tmp_dir = ensure_dir(oxc_validator_tmp_root())
-        env = dict(os.environ)
+        env = child_env_without_native_path_secret()
         tmp_dir_str = str(tmp_dir)
         env["TMPDIR"] = tmp_dir_str
         env["TMP"] = tmp_dir_str
diff --git a/studio/backend/core/export/orchestrator.py b/studio/backend/core/export/orchestrator.py
index 206dbd6dbb..82de925592 100644
--- a/studio/backend/core/export/orchestrator.py
+++ b/studio/backend/core/export/orchestrator.py
@@ -163,21 +163,28 @@ class ExportOrchestrator:
 
     def _spawn_subprocess(self, config: dict) -> None:
         """Spawn a new export subprocess."""
+        from utils.native_path_leases import (
+            native_path_secret_removed_for_child_start,
+            run_without_native_path_secret,
+        )
+
         from .worker import run_export_process
 
-        self._cmd_queue = _CTX.Queue()
-        self._resp_queue = _CTX.Queue()
+        with native_path_secret_removed_for_child_start():
+            self._cmd_queue = _CTX.Queue()
+            self._resp_queue = _CTX.Queue()
 
-        self._proc = _CTX.Process(
-            target = run_export_process,
-            kwargs = {
-                "cmd_queue": self._cmd_queue,
-                "resp_queue": self._resp_queue,
-                "config": config,
-            },
-            daemon = True,
-        )
-        self._proc.start()
+            self._proc = _CTX.Process(
+                target = run_without_native_path_secret,
+                args = (run_export_process,),
+                kwargs = {
+                    "cmd_queue": self._cmd_queue,
+                    "resp_queue": self._resp_queue,
+                    "config": config,
+                },
+                daemon = True,
+            )
+            self._proc.start()
         logger.info("Export subprocess started (pid=%s)", self._proc.pid)
 
     def _shutdown_subprocess(self, timeout: float = 10.0) -> None:
diff --git a/studio/backend/core/inference/audio_codecs.py b/studio/backend/core/inference/audio_codecs.py
index 895b112e85..df3bf27c16 100644
--- a/studio/backend/core/inference/audio_codecs.py
+++ b/studio/backend/core/inference/audio_codecs.py
@@ -17,6 +17,7 @@ from typing import Optional, Tuple
 import numpy as np
 import torch
 
+from utils.native_path_leases import child_env_without_native_path_secret
 from utils.subprocess_compat import (
     windows_hidden_subprocess_kwargs as _windows_hidden_subprocess_kwargs,
 )
@@ -105,6 +106,7 @@ class AudioCodecManager:
                     spark_code_dir,
                 ],
                 check = True,
+                env = child_env_without_native_path_secret(),
                 **_windows_hidden_subprocess_kwargs(),
             )
 
@@ -143,6 +145,7 @@ class AudioCodecManager:
                     outetts_code_dir,
                 ],
                 check = True,
+                env = child_env_without_native_path_secret(),
                 **_windows_hidden_subprocess_kwargs(),
             )
             # Remove files that pull in heavy / incompatible dependencies
diff --git a/studio/backend/core/inference/llama_cpp.py b/studio/backend/core/inference/llama_cpp.py
index ae8224d424..6790dbc802 100644
--- a/studio/backend/core/inference/llama_cpp.py
+++ b/studio/backend/core/inference/llama_cpp.py
@@ -27,6 +27,7 @@ from urllib.parse import urlparse
 
 import httpx
 
+from utils.native_path_leases import child_env_without_native_path_secret
 from utils.subprocess_compat import (
     windows_hidden_subprocess_kwargs as _windows_hidden_subprocess_kwargs,
 )
@@ -592,6 +593,7 @@ class LlamaCppBackend:
                 capture_output = True,
                 text = True,
                 timeout = 10,
+                env = child_env_without_native_path_secret(),
                 **_windows_hidden_subprocess_kwargs(),
             )
             if result.returncode == 0:
@@ -1776,7 +1778,7 @@ class LlamaCppBackend:
             import os
             import sys
 
-            env = os.environ.copy()
+            env = child_env_without_native_path_secret()
             binary_dir = str(Path(binary).parent)
 
             if sys.platform == "win32":
@@ -2166,6 +2168,7 @@ class LlamaCppBackend:
                     capture_output = True,
                     text = True,
                     timeout = 5,
+                    env = child_env_without_native_path_secret(),
                 )
                 if result.returncode != 0:
                     return
diff --git a/studio/backend/core/inference/orchestrator.py b/studio/backend/core/inference/orchestrator.py
index 428ecaa1c3..5562820f49 100644
--- a/studio/backend/core/inference/orchestrator.py
+++ b/studio/backend/core/inference/orchestrator.py
@@ -166,23 +166,30 @@ class InferenceOrchestrator:
 
     def _spawn_subprocess(self, config: dict) -> None:
         """Spawn a new inference subprocess."""
+        from utils.native_path_leases import (
+            native_path_secret_removed_for_child_start,
+            run_without_native_path_secret,
+        )
+
         from .worker import run_inference_process
 
-        self._cmd_queue = _CTX.Queue()
-        self._resp_queue = _CTX.Queue()
-        self._cancel_event = _CTX.Event()
+        with native_path_secret_removed_for_child_start():
+            self._cmd_queue = _CTX.Queue()
+            self._resp_queue = _CTX.Queue()
+            self._cancel_event = _CTX.Event()
 
-        self._proc = _CTX.Process(
-            target = run_inference_process,
-            kwargs = {
-                "cmd_queue": self._cmd_queue,
-                "resp_queue": self._resp_queue,
-                "cancel_event": self._cancel_event,
-                "config": config,
-            },
-            daemon = True,
-        )
-        self._proc.start()
+            self._proc = _CTX.Process(
+                target = run_without_native_path_secret,
+                args = (run_inference_process,),
+                kwargs = {
+                    "cmd_queue": self._cmd_queue,
+                    "resp_queue": self._resp_queue,
+                    "cancel_event": self._cancel_event,
+                    "config": config,
+                },
+                daemon = True,
+            )
+            self._proc.start()
         logger.info("Inference subprocess started (pid=%s)", self._proc.pid)
 
     def _cancel_generation(self) -> None:
diff --git a/studio/backend/core/training/trainer.py b/studio/backend/core/training/trainer.py
index 24d06f9f62..fe8d277ac0 100644
--- a/studio/backend/core/training/trainer.py
+++ b/studio/backend/core/training/trainer.py
@@ -70,6 +70,7 @@ from utils.paths import (
 )
 from trl import SFTTrainer, SFTConfig
 
+from utils.native_path_leases import child_env_without_native_path_secret
 from utils.subprocess_compat import (
     windows_hidden_subprocess_kwargs as _windows_hidden_subprocess_kwargs,
 )
@@ -1771,6 +1772,7 @@ class UnslothTrainer:
                     spark_code_dir,
                 ],
                 check = True,
+                env = child_env_without_native_path_secret(),
                 **_windows_hidden_subprocess_kwargs(),
             )
 
@@ -2005,6 +2007,7 @@ class UnslothTrainer:
                     outetts_code_dir,
                 ],
                 check = True,
+                env = child_env_without_native_path_secret(),
                 **_windows_hidden_subprocess_kwargs(),
             )
             for fpath in [
diff --git a/studio/backend/core/training/training.py b/studio/backend/core/training/training.py
index 5437645797..5642faa189 100644
--- a/studio/backend/core/training/training.py
+++ b/studio/backend/core/training/training.py
@@ -29,6 +29,10 @@ from typing import Optional, Tuple, Any
 
 import matplotlib.pyplot as plt
 from utils.hardware import prepare_gpu_selection
+from utils.native_path_leases import (
+    native_path_secret_removed_for_child_start,
+    run_without_native_path_secret,
+)
 
 logger = get_logger(__name__)
 
@@ -213,20 +217,22 @@ class TrainingBackend:
 
         from .worker import run_training_process
 
-        event_queue = _CTX.Queue()
-        stop_queue = _CTX.Queue()
-
-        proc = _CTX.Process(
-            target = run_training_process,
-            kwargs = {
-                "event_queue": event_queue,
-                "stop_queue": stop_queue,
-                "config": config,
-            },
-            daemon = True,
-        )
         try:
-            proc.start()
+            with native_path_secret_removed_for_child_start():
+                event_queue = _CTX.Queue()
+                stop_queue = _CTX.Queue()
+
+                proc = _CTX.Process(
+                    target = run_without_native_path_secret,
+                    args = (run_training_process,),
+                    kwargs = {
+                        "event_queue": event_queue,
+                        "stop_queue": stop_queue,
+                        "config": config,
+                    },
+                    daemon = True,
+                )
+                proc.start()
         except Exception:
             logger.error("Failed to start training subprocess", exc_info = True)
             return False
diff --git a/studio/backend/loggers/config.py b/studio/backend/loggers/config.py
index 4c0d8ade28..4a27f13d38 100644
--- a/studio/backend/loggers/config.py
+++ b/studio/backend/loggers/config.py
@@ -22,6 +22,8 @@ from typing import Optional
 
 import structlog
 
+from loggers.handlers import filter_sensitive_data
+
 
 class LogConfig:
     """Structured logging configuration for the application.
@@ -58,6 +60,8 @@ class LogConfig:
                 structlog.processors.TimeStamper(fmt = "iso"),  # timestamp first
                 structlog.processors.add_log_level,  # level second
                 structlog.contextvars.merge_contextvars,
+                structlog.processors.format_exc_info,
+                filter_sensitive_data,
                 # Custom processor to flatten the extra field
                 lambda logger, method_name, event_dict: {
                     "timestamp": event_dict.get("timestamp"),
diff --git a/studio/backend/loggers/handlers.py b/studio/backend/loggers/handlers.py
index 3add92ea1e..ddd404cdf3 100644
--- a/studio/backend/loggers/handlers.py
+++ b/studio/backend/loggers/handlers.py
@@ -15,6 +15,7 @@ Key Components:
 - get_logger: Factory function for structured loggers
 """
 
+import re
 import time
 from typing import Callable
 
@@ -22,7 +23,12 @@ import structlog
 from fastapi import Request, Response
 from starlette.middleware.base import BaseHTTPMiddleware
 
+from utils.native_path_leases import redact_native_paths
+
 logger = structlog.get_logger(__name__)
+_NATIVE_PATH_LEASE_RE = re.compile(
+    r"(?i)(\b(?:native_path_lease|nativePathLease)[\"']?\s*[:=]\s*[\"']?)[A-Za-z0-9_-]+\.[A-Za-z0-9_-]+"
+)
 
 
 class LoggingMiddleware(BaseHTTPMiddleware):
@@ -75,6 +81,12 @@ def filter_sensitive_data(logger, method_name, event_dict):
     """Structlog processor to filter out base64 data from logs."""
 
     def filter_value(value):
+        if isinstance(value, str):
+            try:
+                value = redact_native_paths(value)
+            except Exception:
+                pass
+            value = _NATIVE_PATH_LEASE_RE.sub(r"\1<redacted native path lease>", value)
         if (
             isinstance(value, str)
             and len(value) > 100
@@ -83,12 +95,22 @@ def filter_sensitive_data(logger, method_name, event_dict):
             # Likely base64 data, truncate it
             return value[:20] + "..."
         elif isinstance(value, dict):
-            return {k: filter_value(v) for k, v in value.items()}
+            return {
+                k: "<redacted native path lease>"
+                if str(k).replace("_", "").lower() == "nativepathlease"
+                else filter_value(v)
+                for k, v in value.items()
+            }
         elif isinstance(value, list):
             return [filter_value(item) for item in value]
         return value
 
-    return {k: filter_value(v) for k, v in event_dict.items()}
+    return {
+        k: "<redacted native path lease>"
+        if str(k).replace("_", "").lower() == "nativepathlease"
+        else filter_value(v)
+        for k, v in event_dict.items()
+    }
 
 
 def get_logger(name: str) -> structlog.BoundLogger:
diff --git a/studio/backend/main.py b/studio/backend/main.py
index aec335fad3..0958094ff0 100644
--- a/studio/backend/main.py
+++ b/studio/backend/main.py
@@ -78,6 +78,7 @@ from utils.hardware import (
 import utils.hardware.hardware as _hw_module
 
 from utils.cache_cleanup import clear_unsloth_compiled_cache
+from utils.native_path_leases import native_path_leases_supported
 
 
 def get_unsloth_version() -> str:
@@ -244,6 +245,7 @@ async def health_check():
         "chat_only": _hw_module.CHAT_ONLY,
         "desktop_protocol_version": 1,
         "supports_desktop_auth": True,
+        "native_path_leases_supported": native_path_leases_supported(),
     }
 
 
diff --git a/studio/backend/models/inference.py b/studio/backend/models/inference.py
index 027f3313a1..c0bb5b53a2 100644
--- a/studio/backend/models/inference.py
+++ b/studio/backend/models/inference.py
@@ -18,6 +18,9 @@ class LoadRequest(BaseModel):
     """Request to load a model for inference"""
 
     model_path: str = Field(..., description = "Model identifier or local path")
+    native_path_lease: Optional[str] = Field(
+        None, description = "Frontend-visible signed native path grant"
+    )
     hf_token: Optional[str] = Field(
         None, description = "HuggingFace token for gated models"
     )
@@ -69,6 +72,9 @@ class ValidateModelRequest(BaseModel):
     """
 
     model_path: str = Field(..., description = "Model identifier or local path")
+    native_path_lease: Optional[str] = Field(
+        None, description = "Frontend-visible signed native path grant"
+    )
     hf_token: Optional[str] = Field(
         None, description = "HuggingFace token for gated models"
     )
diff --git a/studio/backend/routes/inference.py b/studio/backend/routes/inference.py
index 53dd851666..592da831e4 100644
--- a/studio/backend/routes/inference.py
+++ b/studio/backend/routes/inference.py
@@ -122,6 +122,13 @@ try:
     from utils.models import ModelConfig
     from utils.inference import load_inference_config
     from utils.models.model_config import load_model_defaults
+    from utils.native_path_leases import (
+        NativePathLeaseError,
+        display_label_for_native_path,
+        is_registered_native_path_label,
+        redact_native_paths,
+        verify_native_path_lease,
+    )
 except ImportError:
     parent_backend = backend_path.parent / "backend"
     if str(parent_backend) not in sys.path:
@@ -136,6 +143,13 @@ except ImportError:
     from utils.models import ModelConfig
     from utils.inference import load_inference_config
     from utils.models.model_config import load_model_defaults
+    from utils.native_path_leases import (
+        NativePathLeaseError,
+        display_label_for_native_path,
+        is_registered_native_path_label,
+        redact_native_paths,
+        verify_native_path_lease,
+    )
 
 from models.inference import (
     LoadRequest,
@@ -332,6 +346,65 @@ _TOOL_XML_RE = _re.compile(
 logger = get_logger(__name__)
 
 
+def _validate_native_mmproj_companion(
+    mmproj_path: str | None, gguf_path: str | None
+) -> None:
+    if not mmproj_path or not gguf_path:
+        return
+    import stat as _stat_module
+
+    mm = Path(mmproj_path)
+    gguf = Path(gguf_path)
+    try:
+        mm_lstat = os.lstat(mm)
+    except OSError as exc:
+        raise HTTPException(
+            status_code = 400,
+            detail = "Native vision companion is no longer accessible.",
+        ) from exc
+    if _stat_module.S_ISLNK(mm_lstat.st_mode) or not _stat_module.S_ISREG(
+        mm_lstat.st_mode
+    ):
+        raise HTTPException(
+            status_code = 400,
+            detail = "Native vision companion must be a regular file.",
+        )
+    try:
+        if mm.resolve(strict = True).parent != gguf.resolve(strict = True).parent:
+            raise HTTPException(
+                status_code = 400,
+                detail = "Native vision companion must live next to the selected GGUF.",
+            )
+    except OSError as exc:
+        raise HTTPException(
+            status_code = 400,
+            detail = "Native vision companion is no longer accessible.",
+        ) from exc
+
+
+def _resolve_model_identifier_for_request(
+    request: LoadRequest | ValidateModelRequest,
+    *,
+    operation: str,
+) -> tuple[str, str, bool]:
+    if not request.native_path_lease:
+        return request.model_path, request.model_path, False
+    try:
+        grant = verify_native_path_lease(
+            request.native_path_lease,
+            operation = operation,
+            expected_kind = "model",
+            expected_path_type = "file",
+            allowed_suffixes = (".gguf",),
+        )
+    except NativePathLeaseError as exc:
+        raise HTTPException(status_code = 400, detail = str(exc)) from exc
+    display_label = (
+        grant.display_label or Path(request.model_path).name or "Native model"
+    )
+    return str(grant.canonical_path), display_label, True
+
+
 # GGUF inference backend (llama-server)
 _llama_cpp_backend = LlamaCppBackend()
 
@@ -355,7 +428,12 @@ async def load_model(
 
     GGUF models are loaded via llama-server (llama.cpp) instead of Unsloth.
     """
+    native_grant_backed = False
+    model_log_label = request.model_path
     try:
+        model_identifier, model_log_label, native_grant_backed = (
+            _resolve_model_identifier_for_request(request, operation = "load-model")
+        )
         # Version switching is handled automatically by the subprocess-based
         # inference backend — no need for ensure_transformers_version() here.
 
@@ -369,10 +447,10 @@ async def load_model(
                 and llama_backend.hf_variant
                 and llama_backend.hf_variant.lower() == request.gguf_variant.lower()
                 and llama_backend.model_identifier
-                and llama_backend.model_identifier.lower() == request.model_path.lower()
+                and llama_backend.model_identifier.lower() == model_identifier.lower()
             ):
                 logger.info(
-                    f"Model already loaded (GGUF): {request.model_path} variant={request.gguf_variant}, skipping reload"
+                    f"Model already loaded (GGUF): {model_log_label} variant={request.gguf_variant}, skipping reload"
                 )
                 inference_config = load_inference_config(llama_backend.model_identifier)
                 from utils.models import is_audio_input_type
@@ -385,8 +463,12 @@ async def load_model(
                 _gguf_is_audio = getattr(llama_backend, "_is_audio", False)
                 return LoadResponse(
                     status = "already_loaded",
-                    model = llama_backend.model_identifier,
-                    display_name = llama_backend.model_identifier,
+                    model = model_log_label
+                    if native_grant_backed
+                    else llama_backend.model_identifier,
+                    display_name = model_log_label
+                    if native_grant_backed
+                    else llama_backend.model_identifier,
                     is_vision = llama_backend._is_vision,
                     is_lora = False,
                     is_gguf = True,
@@ -412,10 +494,10 @@ async def load_model(
         else:
             if (
                 backend.active_model_name
-                and backend.active_model_name.lower() == request.model_path.lower()
+                and backend.active_model_name.lower() == model_identifier.lower()
             ):
                 logger.info(
-                    f"Model already loaded (Unsloth): {request.model_path}, skipping reload"
+                    f"Model already loaded (Unsloth): {model_log_label}, skipping reload"
                 )
                 inference_config = load_inference_config(backend.active_model_name)
                 _model_info = backend.models.get(backend.active_model_name, {})
@@ -444,8 +526,12 @@ async def load_model(
                         pass
                 return LoadResponse(
                     status = "already_loaded",
-                    model = backend.active_model_name,
-                    display_name = backend.active_model_name,
+                    model = model_log_label
+                    if native_grant_backed
+                    else backend.active_model_name,
+                    display_name = model_log_label
+                    if native_grant_backed
+                    else backend.active_model_name,
                     is_vision = _model_info.get("is_vision", False),
                     is_lora = _model_info.get("is_lora", False),
                     is_gguf = False,
@@ -467,7 +553,7 @@ async def load_model(
         # Create config using clean factory method
         # is_lora is auto-detected from adapter_config.json on disk/HF
         config = ModelConfig.from_identifier(
-            model_id = request.model_path,
+            model_id = model_identifier,
             hf_token = request.hf_token,
             gguf_variant = request.gguf_variant,
         )
@@ -475,7 +561,7 @@ async def load_model(
         if not config:
             raise HTTPException(
                 status_code = 400,
-                detail = f"Invalid model identifier: {request.model_path}",
+                detail = f"Invalid model identifier: {model_log_label}",
             )
 
         # Normalize gpu_ids: empty list means auto-selection, same as None
@@ -522,6 +608,10 @@ async def load_model(
                 )
             else:
                 # Local mode: llama-server loads via -m <path>
+                if native_grant_backed and config.gguf_mmproj_file:
+                    _validate_native_mmproj_companion(
+                        config.gguf_mmproj_file, config.gguf_file
+                    )
                 success = await asyncio.to_thread(
                     llama_backend.load_model,
                     gguf_path = config.gguf_file,
@@ -538,10 +628,12 @@ async def load_model(
             if not success:
                 raise HTTPException(
                     status_code = 500,
-                    detail = f"Failed to load GGUF model: {config.display_name}",
+                    detail = f"Failed to load GGUF model: {model_log_label if native_grant_backed else config.display_name}",
                 )
 
-            logger.info(f"Loaded GGUF model via llama-server: {config.identifier}")
+            logger.info(
+                f"Loaded GGUF model via llama-server: {model_log_label if native_grant_backed else config.identifier}"
+            )
 
             # Detect TTS audio by probing the loaded model's vocabulary
             from utils.models import is_audio_input_type
@@ -550,6 +642,10 @@ async def load_model(
             _gguf_is_audio = _gguf_audio in ("snac", "bicodec", "dac")
             llama_backend._is_audio = _gguf_is_audio
             llama_backend._audio_type = _gguf_audio
+            llama_backend._native_display_label = (
+                model_log_label if native_grant_backed else None
+            )
+            llama_backend._native_grant_backed = bool(native_grant_backed)
             if _gguf_is_audio:
                 logger.info(f"GGUF model detected as audio: audio_type={_gguf_audio}")
                 await asyncio.to_thread(llama_backend.init_audio_codec, _gguf_audio)
@@ -558,8 +654,10 @@ async def load_model(
 
             return LoadResponse(
                 status = "loaded",
-                model = config.identifier,
-                display_name = config.display_name,
+                model = model_log_label if native_grant_backed else config.identifier,
+                display_name = model_log_label
+                if native_grant_backed
+                else config.display_name,
                 is_vision = config.is_vision,
                 is_lora = False,
                 is_gguf = True,
@@ -682,10 +780,13 @@ async def load_model(
                         ),
                     )
             raise HTTPException(
-                status_code = 500, detail = f"Failed to load model: {config.display_name}"
+                status_code = 500,
+                detail = f"Failed to load model: {model_log_label if native_grant_backed else config.display_name}",
             )
 
-        logger.info(f"Loaded model: {config.identifier}")
+        logger.info(
+            f"Loaded model: {model_log_label if native_grant_backed else config.identifier}"
+        )
 
         # Load inference configuration parameters
         inference_config = load_inference_config(config.identifier)
@@ -715,8 +816,10 @@ async def load_model(
 
         return LoadResponse(
             status = "loaded",
-            model = config.identifier,
-            display_name = config.display_name,
+            model = model_log_label if native_grant_backed else config.identifier,
+            display_name = model_log_label
+            if native_grant_backed
+            else config.display_name,
             is_vision = config.is_vision,
             is_lora = config.is_lora,
             is_gguf = False,
@@ -738,11 +841,17 @@ async def load_model(
     except HTTPException:
         raise
     except ValueError as e:
+        if native_grant_backed:
+            redacted_msg = redact_native_paths(str(e))
+            logger.warning(
+                "Rejected inference selection for native model %s: %s",
+                model_log_label,
+                redacted_msg,
+            )
+            raise HTTPException(status_code = 400, detail = redacted_msg)
         logger.warning("Rejected inference GPU selection: %s", e)
         raise HTTPException(status_code = 400, detail = str(e))
     except Exception as e:
-        logger.error(f"Error loading model: {e}", exc_info = True)
-        msg = str(e)
         # Surface a friendlier message for models that Unsloth cannot load
         not_supported_hints = [
             "No config file found",
@@ -750,6 +859,22 @@ async def load_model(
             "is not supported",
             "does not support",
         ]
+        if native_grant_backed:
+            redacted_msg = redact_native_paths(str(e))
+            logger.error(
+                "Error loading native model %s: %s",
+                model_log_label,
+                redacted_msg,
+            )
+            msg = redacted_msg
+            if any(h.lower() in msg.lower() for h in not_supported_hints):
+                msg = f"This model is not supported yet. Try a different model. (Original error: {msg})"
+            raise HTTPException(
+                status_code = 500,
+                detail = f"Failed to load native model {model_log_label}: {msg}",
+            )
+        logger.error(f"Error loading model: {e}", exc_info = True)
+        msg = str(e)
         if any(h.lower() in msg.lower() for h in not_supported_hints):
             msg = f"This model is not supported yet. Try a different model. (Original error: {msg})"
         raise HTTPException(status_code = 500, detail = f"Failed to load model: {msg}")
@@ -766,9 +891,14 @@ async def validate_model(
     This checks that ModelConfig.from_identifier() can resolve the given
     model_path, but it does NOT actually load model weights into GPU memory.
     """
+    native_grant_backed = False
+    model_log_label = request.model_path
     try:
+        model_identifier, model_log_label, native_grant_backed = (
+            _resolve_model_identifier_for_request(request, operation = "validate-model")
+        )
         config = ModelConfig.from_identifier(
-            model_id = request.model_path,
+            model_id = model_identifier,
             hf_token = request.hf_token,
             gguf_variant = request.gguf_variant,
         )
@@ -776,14 +906,16 @@ async def validate_model(
         if not config:
             raise HTTPException(
                 status_code = 400,
-                detail = f"Invalid model identifier: {request.model_path}",
+                detail = f"Invalid model identifier: {model_log_label}",
             )
 
         return ValidateModelResponse(
             valid = True,
             message = "Model identifier is valid.",
-            identifier = config.identifier,
-            display_name = getattr(config, "display_name", config.identifier),
+            identifier = model_log_label if native_grant_backed else config.identifier,
+            display_name = model_log_label
+            if native_grant_backed
+            else getattr(config, "display_name", config.identifier),
             is_gguf = getattr(config, "is_gguf", False),
             is_lora = getattr(config, "is_lora", False),
             is_vision = getattr(config, "is_vision", False),
@@ -795,6 +927,26 @@ async def validate_model(
     except HTTPException:
         raise
     except Exception as e:
+        not_supported_hints = [
+            "No config file found",
+            "not yet supported",
+            "is not supported",
+            "does not support",
+        ]
+        if native_grant_backed:
+            redacted_msg = redact_native_paths(str(e))
+            logger.error(
+                "Error validating native model %s: %s",
+                model_log_label,
+                redacted_msg,
+            )
+            msg = redacted_msg
+            if any(h.lower() in msg.lower() for h in not_supported_hints):
+                msg = f"This model is not supported yet. Try a different model. (Original error: {msg})"
+            raise HTTPException(
+                status_code = 400,
+                detail = f"Invalid native model {model_log_label}: {msg}",
+            )
         logger.error(
             f"Error validating model identifier '{request.model_path}': {e}",
             exc_info = True,
@@ -819,6 +971,9 @@ async def unload_model(
         llama_backend = get_llama_cpp_backend()
         if llama_backend.is_active and (
             llama_backend.model_identifier == request.model_path
+            or is_registered_native_path_label(
+                llama_backend.model_identifier, request.model_path
+            )
             or not llama_backend.is_loaded
         ):
             llama_backend.unload_model()
@@ -966,16 +1121,27 @@ async def get_status(
         # If a GGUF model is loaded via llama-server, report that
         if llama_backend.is_loaded:
             _model_id = llama_backend.model_identifier
+            _native_grant_backed = getattr(llama_backend, "_native_grant_backed", False)
+            _display_model_id = getattr(
+                llama_backend, "_native_display_label", None
+            ) or display_label_for_native_path(_model_id)
+            if (
+                _native_grant_backed
+                and _model_id
+                and _display_model_id == _model_id
+                and os.path.isabs(_model_id)
+            ):
+                _display_model_id = os.path.basename(_model_id)
             _inference_cfg = load_inference_config(_model_id) if _model_id else None
             return InferenceStatusResponse(
-                active_model = _model_id,
+                active_model = _display_model_id,
                 is_vision = llama_backend.is_vision,
                 is_gguf = True,
                 gguf_variant = llama_backend.hf_variant,
                 is_audio = getattr(llama_backend, "_is_audio", False),
                 audio_type = getattr(llama_backend, "_audio_type", None),
                 loading = [],
-                loaded = [_model_id],
+                loaded = [_display_model_id] if _display_model_id else [],
                 inference = _inference_cfg,
                 requires_trust_remote_code = bool(
                     (_inference_cfg or {}).get("trust_remote_code", False)
diff --git a/studio/backend/utils/hardware/amd.py b/studio/backend/utils/hardware/amd.py
index 755314ca3a..fdb1ab4520 100644
--- a/studio/backend/utils/hardware/amd.py
+++ b/studio/backend/utils/hardware/amd.py
@@ -16,6 +16,7 @@ import subprocess
 from typing import Any, Optional
 
 from loggers import get_logger
+from utils.native_path_leases import child_env_without_native_path_secret
 
 logger = get_logger(__name__)
 
@@ -28,6 +29,7 @@ def _run_amd_smi(*args: str, timeout: int = 5) -> Optional[Any]:
             capture_output = True,
             text = True,
             timeout = timeout,
+            env = child_env_without_native_path_secret(),
         )
     except (OSError, subprocess.TimeoutExpired) as e:
         logger.warning("amd-smi query failed: %s", e)
diff --git a/studio/backend/utils/hardware/nvidia.py b/studio/backend/utils/hardware/nvidia.py
index 274d9beb48..099c5fa3a5 100644
--- a/studio/backend/utils/hardware/nvidia.py
+++ b/studio/backend/utils/hardware/nvidia.py
@@ -6,6 +6,7 @@ from typing import Any, Optional
 
 from loggers import get_logger
 
+from utils.native_path_leases import child_env_without_native_path_secret
 from utils.subprocess_compat import (
     windows_hidden_subprocess_kwargs as _windows_hidden_subprocess_kwargs,
 )
@@ -65,6 +66,7 @@ def get_physical_gpu_count() -> Optional[int]:
             capture_output = True,
             text = True,
             timeout = 5,
+            env = child_env_without_native_path_secret(),
             **_windows_hidden_subprocess_kwargs(),
         )
         if result.returncode == 0 and result.stdout.strip():
@@ -90,6 +92,7 @@ def get_primary_gpu_utilization() -> dict[str, Any]:
             capture_output = True,
             text = True,
             timeout = 5,
+            env = child_env_without_native_path_secret(),
             **_windows_hidden_subprocess_kwargs(),
         )
     except (OSError, subprocess.TimeoutExpired) as e:
@@ -141,6 +144,7 @@ def get_visible_gpu_utilization(
             capture_output = True,
             text = True,
             timeout = 5,
+            env = child_env_without_native_path_secret(),
             **_windows_hidden_subprocess_kwargs(),
         )
     except (OSError, subprocess.TimeoutExpired) as e:
@@ -227,6 +231,7 @@ def get_backend_visible_gpu_info(
             capture_output = True,
             text = True,
             timeout = 10,
+            env = child_env_without_native_path_secret(),
             **_windows_hidden_subprocess_kwargs(),
         )
     except (OSError, subprocess.TimeoutExpired) as e:
diff --git a/studio/backend/utils/models/model_config.py b/studio/backend/utils/models/model_config.py
index 7bfc74235f..16f6d21edb 100644
--- a/studio/backend/utils/models/model_config.py
+++ b/studio/backend/utils/models/model_config.py
@@ -32,6 +32,7 @@ import threading
 import yaml
 
 
+from utils.native_path_leases import child_env_without_native_path_secret
 from utils.subprocess_compat import (
     windows_hidden_subprocess_kwargs as _windows_hidden_subprocess_kwargs,
 )
@@ -583,6 +584,7 @@ def _is_vision_model_subprocess(
             capture_output = True,
             text = True,
             timeout = 60,
+            env = child_env_without_native_path_secret(),
             **_windows_hidden_subprocess_kwargs(),
         )
 
diff --git a/studio/backend/utils/native_path_leases.py b/studio/backend/utils/native_path_leases.py
new file mode 100644
index 0000000000..a69dfab532
--- /dev/null
+++ b/studio/backend/utils/native_path_leases.py
@@ -0,0 +1,406 @@
+# SPDX-License-Identifier: AGPL-3.0-only
+# Copyright 2026-present the Unsloth AI Inc. team. All rights reserved. See /studio/LICENSE.AGPL-3.0
+
+"""Verification for Tauri native path signed grants.
+
+Rust signs compact ``base64url(payload_json).base64url(hmac)`` grants. The
+frontend can see and forward the grant, but cannot change it without breaking
+the HMAC. The backend verifies the original payload segment bytes, then
+re-stats the path before any native read.
+"""
+
+from __future__ import annotations
+
+import base64
+import binascii
+import hashlib
+import hmac
+import json
+import os
+import stat as _stat_module
+import threading
+import time
+from contextlib import contextmanager
+from dataclasses import dataclass
+from pathlib import Path
+from typing import Any, Callable, Iterable, Iterator, Mapping
+
+LEASE_SECRET_ENV = "UNSLOTH_STUDIO_NATIVE_PATH_LEASE_SECRET"
+_MAX_NATIVE_PATH_REDACTIONS = 100
+_MAX_NATIVE_PATH_LABELS = 10_000
+_MIN_LEASE_SECRET_BYTES = 32
+
+_REPLAY_LOCK = threading.Lock()
+_USED_NONCES: dict[str, int] = {}
+_REDACTION_LOCK = threading.Lock()
+_NATIVE_PATH_REDACTIONS: list[str] = []
+_NATIVE_PATH_LABELS: dict[str, str] = {}
+_NATIVE_PATH_ENV_LOCK = threading.Lock()
+_SECRET_INIT_LOCK = threading.Lock()
+_CACHED_LEASE_SECRET: bytes | None = None
+_SCRUB_REFCOUNT = 0
+_SCRUB_SAVED_SECRET: str | None = None
+
+
+class NativePathLeaseError(ValueError):
+    """Raised when a native path grant is missing, invalid, or unsafe."""
+
+
+@dataclass(frozen = True)
+class NativePathGrant:
+    operation: str
+    canonical_path: Path
+    path_kind: str
+    path_type: str
+    source_kind: str
+    token_id_hash: str
+    display_label: str
+    expires_at_ms: int
+    size_bytes: int | None
+    modified_ms: int | None
+
+
+def native_path_leases_supported() -> bool:
+    try:
+        _decode_secret()
+    except NativePathLeaseError:
+        return False
+    return True
+
+
+def child_env_without_native_path_secret(
+    env: Mapping[str, str] | None = None,
+) -> dict[str, str]:
+    """Return a child-process env with the native path lease secret removed."""
+
+    if env is None:
+        with _NATIVE_PATH_ENV_LOCK:
+            cleaned = dict(os.environ)
+    else:
+        cleaned = dict(env)
+    cleaned.pop(LEASE_SECRET_ENV, None)
+    return cleaned
+
+
+def run_without_native_path_secret(
+    target: Callable[..., Any],
+    *args: Any,
+    **kwargs: Any,
+) -> Any:
+    """Run a multiprocessing child target without the native path lease secret."""
+
+    global _CACHED_LEASE_SECRET, _SCRUB_SAVED_SECRET
+    os.environ.pop(LEASE_SECRET_ENV, None)
+    _CACHED_LEASE_SECRET = None
+    _SCRUB_SAVED_SECRET = None
+    return target(*args, **kwargs)
+
+
+@contextmanager
+def native_path_secret_removed_for_child_start() -> Iterator[None]:
+    global _SCRUB_REFCOUNT, _SCRUB_SAVED_SECRET, _CACHED_LEASE_SECRET
+    with _NATIVE_PATH_ENV_LOCK:
+        if _SCRUB_REFCOUNT == 0:
+            _SCRUB_SAVED_SECRET = os.environ.pop(LEASE_SECRET_ENV, None)
+            _CACHED_LEASE_SECRET = None
+        _SCRUB_REFCOUNT += 1
+    try:
+        yield
+    finally:
+        with _NATIVE_PATH_ENV_LOCK:
+            _SCRUB_REFCOUNT -= 1
+            if _SCRUB_REFCOUNT == 0 and _SCRUB_SAVED_SECRET is not None:
+                os.environ[LEASE_SECRET_ENV] = _SCRUB_SAVED_SECRET
+                _SCRUB_SAVED_SECRET = None
+
+
+def verify_native_path_lease(
+    lease: str | None,
+    *,
+    operation: str,
+    expected_kind: str | None = None,
+    expected_path_type: str | None = None,
+    allowed_suffixes: Iterable[str] | None = None,
+) -> NativePathGrant:
+    if not lease:
+        raise NativePathLeaseError("Native path grant is required.")
+
+    secret = _decode_secret()
+    payload_b64, signature_b64 = _split_lease(lease)
+    expected_signature = hmac.new(
+        secret,
+        payload_b64.encode("ascii"),
+        hashlib.sha256,
+    ).digest()
+    supplied_signature = _b64decode(signature_b64)
+    if not hmac.compare_digest(expected_signature, supplied_signature):
+        raise NativePathLeaseError("Native path grant signature is invalid.")
+
+    payload = _decode_payload(payload_b64)
+    _validate_payload(payload, operation = operation, expected_kind = expected_kind)
+
+    path = Path(str(payload["canonical_path"]))
+    _reject_network_or_device_path(path)
+    try:
+        signed_lstat = os.lstat(path)
+    except OSError as exc:
+        raise NativePathLeaseError("Native path is no longer accessible.") from exc
+    if _stat_module.S_ISLNK(signed_lstat.st_mode):
+        raise NativePathLeaseError("Native path is no longer a regular file.")
+    try:
+        resolved = path.resolve(strict = True)
+    except OSError as exc:
+        raise NativePathLeaseError("Native path is no longer accessible.") from exc
+    _reject_network_or_device_path(resolved)
+    if not _same_native_path(resolved, path):
+        raise NativePathLeaseError(
+            "Native path grant no longer resolves to the selected path."
+        )
+
+    grant = NativePathGrant(
+        operation = str(payload["operation"]),
+        canonical_path = resolved,
+        path_kind = str(payload["path_kind"]),
+        path_type = str(payload["path_type"]),
+        source_kind = str(payload["source_kind"]),
+        token_id_hash = str(payload["token_id_hash"]),
+        display_label = str(payload.get("display_label") or resolved.name),
+        expires_at_ms = _required_int(payload, "expires_at_ms"),
+        size_bytes = _optional_int(payload.get("size_bytes")),
+        modified_ms = _optional_int(payload.get("modified_ms")),
+    )
+
+    if expected_path_type and grant.path_type != expected_path_type:
+        raise NativePathLeaseError("Native path grant has the wrong path type.")
+    suffixes = tuple(s.lower() for s in (allowed_suffixes or ()))
+    if suffixes and resolved.suffix.lower() not in suffixes:
+        raise NativePathLeaseError("Native path grant has an unsupported file type.")
+
+    _validate_current_stat(grant)
+    _consume_nonce(str(payload["nonce"]), grant.expires_at_ms)
+    _remember_native_path_for_redaction(str(resolved), grant.display_label)
+    return grant
+
+
+def display_label_for_native_path(value: str | None) -> str | None:
+    if not value:
+        return value
+    with _REDACTION_LOCK:
+        return _NATIVE_PATH_LABELS.get(value, value)
+
+
+def is_registered_native_path_label(path_value: str | None, label: str | None) -> bool:
+    if not path_value or not label:
+        return False
+    with _REDACTION_LOCK:
+        return _NATIVE_PATH_LABELS.get(path_value) == label
+
+
+def redact_native_paths(value: str) -> str:
+    with _REDACTION_LOCK:
+        paths = sorted(_NATIVE_PATH_REDACTIONS, key = len, reverse = True)
+    redacted = value
+    for path in paths:
+        for variant in {path, path.replace("/", "\\"), path.replace("\\", "/")}:
+            if variant:
+                redacted = redacted.replace(variant, "<native_path>")
+    return redacted
+
+
+def _decode_secret() -> bytes:
+    global _CACHED_LEASE_SECRET
+    if _CACHED_LEASE_SECRET is not None:
+        return _CACHED_LEASE_SECRET
+    with _SECRET_INIT_LOCK:
+        if _CACHED_LEASE_SECRET is not None:
+            return _CACHED_LEASE_SECRET
+        with _NATIVE_PATH_ENV_LOCK:
+            encoded = os.environ.get(LEASE_SECRET_ENV)
+            if encoded is None and _SCRUB_SAVED_SECRET is not None:
+                encoded = _SCRUB_SAVED_SECRET
+        if not encoded:
+            raise NativePathLeaseError(
+                "Native path grants require the managed desktop backend."
+            )
+        try:
+            secret = _b64decode(encoded)
+        except Exception as exc:
+            raise NativePathLeaseError("Native path grant secret is invalid.") from exc
+        if len(secret) < _MIN_LEASE_SECRET_BYTES:
+            raise NativePathLeaseError("Native path grant secret is invalid.")
+        _CACHED_LEASE_SECRET = secret
+        return secret
+
+
+def _split_lease(lease: str) -> tuple[str, str]:
+    if not isinstance(lease, str):
+        raise NativePathLeaseError("Native path grant has an invalid format.")
+    try:
+        lease.encode("ascii")
+    except UnicodeEncodeError as exc:
+        raise NativePathLeaseError("Native path grant has an invalid format.") from exc
+    parts = lease.split(".")
+    if len(parts) != 2 or not parts[0] or not parts[1]:
+        raise NativePathLeaseError("Native path grant has an invalid format.")
+    return parts[0], parts[1]
+
+
+def _decode_payload(payload_b64: str) -> dict[str, Any]:
+    try:
+        payload = json.loads(_b64decode(payload_b64).decode("utf-8"))
+    except Exception as exc:
+        raise NativePathLeaseError("Native path grant payload is invalid.") from exc
+    if not isinstance(payload, dict):
+        raise NativePathLeaseError("Native path grant payload is invalid.")
+    return payload
+
+
+def _validate_payload(
+    payload: dict[str, Any], *, operation: str, expected_kind: str | None
+) -> None:
+    required = (
+        "version",
+        "operation",
+        "canonical_path",
+        "path_kind",
+        "path_type",
+        "source_kind",
+        "token_id_hash",
+        "issued_at_ms",
+        "expires_at_ms",
+        "nonce",
+    )
+    missing = [key for key in required if key not in payload]
+    if missing:
+        raise NativePathLeaseError(
+            "Native path grant payload is missing required fields."
+        )
+    if _required_int(payload, "version") != 1:
+        raise NativePathLeaseError("Native path grant version is unsupported.")
+    if payload["operation"] != operation:
+        raise NativePathLeaseError("Native path grant operation is invalid.")
+    if expected_kind and payload["path_kind"] != expected_kind:
+        raise NativePathLeaseError("Native path grant kind is invalid.")
+    now_ms = int(time.time() * 1000)
+    issued_at_ms = _required_int(payload, "issued_at_ms")
+    expires_at_ms = _required_int(payload, "expires_at_ms")
+    if issued_at_ms >= expires_at_ms:
+        raise NativePathLeaseError("Native path grant timestamps are inconsistent.")
+    if expires_at_ms <= now_ms:
+        raise NativePathLeaseError("Native path grant has expired.")
+    if issued_at_ms > now_ms + 30_000:
+        raise NativePathLeaseError("Native path grant issue time is invalid.")
+    for key in ("canonical_path", "nonce", "token_id_hash", "display_label"):
+        raw = payload.get(key)
+        if raw is None:
+            continue
+        if "\x00" in str(raw):
+            raise NativePathLeaseError("Native path grant contains invalid characters.")
+
+
+def _validate_current_stat(grant: NativePathGrant) -> None:
+    try:
+        st = os.lstat(grant.canonical_path)
+    except OSError as exc:
+        raise NativePathLeaseError("Native path is no longer accessible.") from exc
+    if _stat_module.S_ISLNK(st.st_mode):
+        raise NativePathLeaseError("Native path is no longer a regular file.")
+    if grant.path_type == "file":
+        if not _stat_module.S_ISREG(st.st_mode):
+            raise NativePathLeaseError("Native path is no longer a regular file.")
+    elif grant.path_type == "directory":
+        if not _stat_module.S_ISDIR(st.st_mode):
+            raise NativePathLeaseError("Native path is no longer a directory.")
+    else:
+        raise NativePathLeaseError("Native path grant has an unsupported path type.")
+
+    if grant.size_bytes is not None and st.st_size != grant.size_bytes:
+        raise NativePathLeaseError("Native path changed after it was selected.")
+    current_modified_ms = int(st.st_mtime_ns // 1_000_000)
+    if grant.modified_ms is not None and current_modified_ms != grant.modified_ms:
+        raise NativePathLeaseError("Native path changed after it was selected.")
+
+
+def _consume_nonce(nonce: str, expires_at_ms: int) -> None:
+    now_ms = int(time.time() * 1000)
+    with _REPLAY_LOCK:
+        for key, expiry in list(_USED_NONCES.items()):
+            if expiry <= now_ms:
+                _USED_NONCES.pop(key, None)
+        if nonce in _USED_NONCES:
+            raise NativePathLeaseError("Native path grant was already used.")
+        _USED_NONCES[nonce] = expires_at_ms
+
+
+def _remember_native_path_for_redaction(path: str, display_label: str) -> None:
+    with _REDACTION_LOCK:
+        _NATIVE_PATH_LABELS[path] = display_label
+        if len(_NATIVE_PATH_LABELS) > _MAX_NATIVE_PATH_LABELS:
+            excess = len(_NATIVE_PATH_LABELS) - _MAX_NATIVE_PATH_LABELS
+            for stale_path in list(_NATIVE_PATH_LABELS.keys())[:excess]:
+                _NATIVE_PATH_LABELS.pop(stale_path, None)
+        if path in _NATIVE_PATH_REDACTIONS:
+            return
+        _NATIVE_PATH_REDACTIONS.append(path)
+        del _NATIVE_PATH_REDACTIONS[:-_MAX_NATIVE_PATH_REDACTIONS]
+
+
+def _reject_network_or_device_path(path: Path) -> None:
+    text = str(path)
+    if os.name == "nt":
+        normalized = text.replace("/", "\\").lower()
+        if normalized.startswith("\\\\?\\"):
+            rest = normalized[4:]
+            is_local_drive = len(rest) >= 3 and rest[0].isalpha() and rest[1:3] == ":\\"
+            if not is_local_drive:
+                raise NativePathLeaseError(
+                    "Network paths are not supported for native grants."
+                )
+        elif normalized.startswith("\\\\"):
+            raise NativePathLeaseError(
+                "Network paths are not supported for native grants."
+            )
+    if os.name != "nt":
+        for root in ("/dev", "/proc", "/sys"):
+            if path.is_relative_to(root):
+                raise NativePathLeaseError(
+                    "Device and virtual filesystem paths are not supported."
+                )
+    if "\x00" in text:
+        raise NativePathLeaseError("Native path contains invalid characters.")
+
+
+def _b64decode(value: str) -> bytes:
+    try:
+        padding = "=" * (-len(value) % 4)
+        return base64.urlsafe_b64decode((value + padding).encode("ascii"))
+    except (UnicodeEncodeError, binascii.Error, ValueError) as exc:
+        raise NativePathLeaseError("Native path grant has an invalid format.") from exc
+
+
+def _same_native_path(resolved: Path, signed: Path) -> bool:
+    try:
+        return resolved.samefile(signed)
+    except OSError:
+        return os.path.normcase(str(resolved)) == os.path.normcase(str(signed))
+
+
+def _optional_int(value: Any) -> int | None:
+    if value is None:
+        return None
+    try:
+        return int(value)
+    except (TypeError, ValueError) as exc:
+        raise NativePathLeaseError("Native path grant payload is invalid.") from exc
+
+
+def _required_int(payload: dict[str, Any], key: str) -> int:
+    raw = payload.get(key)
+    if raw is None:
+        raise NativePathLeaseError(
+            "Native path grant payload is missing required fields."
+        )
+    try:
+        return int(raw)
+    except (TypeError, ValueError) as exc:
+        raise NativePathLeaseError("Native path grant payload is invalid.") from exc
diff --git a/studio/backend/utils/transformers_version.py b/studio/backend/utils/transformers_version.py
index ebb20b85da..17af40f663 100644
--- a/studio/backend/utils/transformers_version.py
+++ b/studio/backend/utils/transformers_version.py
@@ -36,6 +36,7 @@ import subprocess
 import sys
 from pathlib import Path
 
+from utils.native_path_leases import child_env_without_native_path_secret
 from utils.subprocess_compat import (
     windows_hidden_subprocess_kwargs as _windows_hidden_subprocess_kwargs,
 )
@@ -504,6 +505,7 @@ def _install_to_dir(pkg: str, target_dir: str) -> bool:
             stdout = subprocess.PIPE,
             stderr = subprocess.STDOUT,
             text = True,
+            env = child_env_without_native_path_secret(),
             **_windows_hidden_subprocess_kwargs(),
         )
         if result.returncode == 0:
@@ -526,6 +528,7 @@ def _install_to_dir(pkg: str, target_dir: str) -> bool:
         stdout = subprocess.PIPE,
         stderr = subprocess.STDOUT,
         text = True,
+        env = child_env_without_native_path_secret(),
         **_windows_hidden_subprocess_kwargs(),
     )
     if result.returncode != 0:
diff --git a/studio/backend/utils/wheel_utils.py b/studio/backend/utils/wheel_utils.py
index 00240f1e69..3ed9bda827 100644
--- a/studio/backend/utils/wheel_utils.py
+++ b/studio/backend/utils/wheel_utils.py
@@ -13,6 +13,8 @@ import urllib.error
 import urllib.request
 from typing import Callable
 
+from utils.native_path_leases import child_env_without_native_path_secret
+
 _logger = logging.getLogger(__name__)
 
 FLASH_ATTN_RELEASE_BASE_URL = (
@@ -59,6 +61,7 @@ def probe_torch_wheel_env(*, timeout: int | None = None) -> dict[str, str] | Non
             stderr = subprocess.PIPE,
             text = True,
             timeout = timeout,
+            env = child_env_without_native_path_secret(),
         )
     except subprocess.TimeoutExpired:
         return None
@@ -142,6 +145,7 @@ def install_wheel(
             stdout = subprocess.PIPE,
             stderr = subprocess.STDOUT,
             text = True,
+            env = child_env_without_native_path_secret(),
         )
         attempts.append(("uv", result))
         if result.returncode == 0:
@@ -153,6 +157,7 @@ def install_wheel(
         stdout = subprocess.PIPE,
         stderr = subprocess.STDOUT,
         text = True,
+        env = child_env_without_native_path_secret(),
     )
     attempts.append(("pip", result))
     return attempts
diff --git a/studio/frontend/src/app/provider.tsx b/studio/frontend/src/app/provider.tsx
index c3a1c7dab0..62e78b809a 100644
--- a/studio/frontend/src/app/provider.tsx
+++ b/studio/frontend/src/app/provider.tsx
@@ -10,6 +10,7 @@ import {
 } from "@/components/tauri/window-titlebar";
 import { Toaster } from "@/components/ui/sonner";
 import { getTauriAuthFailure, tauriAutoAuth } from "@/features/auth";
+import { NativeIntentDrain } from "@/features/native-intents/native-intent-drain";
 import { useTauriBackend, type BackendStatus } from "@/hooks/use-tauri-backend";
 import { useTauriUpdate } from "@/hooks/use-tauri-update";
 import { isTauri } from "@/lib/api-base";
@@ -245,6 +246,7 @@ function TauriWrapper({ children }: { children: ReactNode }) {
   const content = showApp ? (
     <>
       <TauriUpdateLayer isExternalServer={isExternalServer} />
+      <NativeIntentDrain />
       {children}
     </>
   ) : (
diff --git a/studio/frontend/src/components/assistant-ui/model-selector.tsx b/studio/frontend/src/components/assistant-ui/model-selector.tsx
index 1ff946dd46..795bcb6d08 100644
--- a/studio/frontend/src/components/assistant-ui/model-selector.tsx
+++ b/studio/frontend/src/components/assistant-ui/model-selector.tsx
@@ -13,6 +13,7 @@ import { usePlatformStore } from "@/config/env";
 import { cn } from "@/lib/utils";
 import {
   ArrowDown01Icon,
+  FolderSearchIcon,
   Logout01Icon,
 } from "@hugeicons/core-free-icons";
 import { HugeiconsIcon } from "@hugeicons/react";
@@ -41,6 +42,7 @@ interface ModelSelectorProps {
   onValueChange?: (value: string, meta: ModelSelectorChangeMeta) => void;
   onEject?: () => void;
   onFoldersChange?: () => void;
+  onPickLocalModel?: () => void | Promise<void>;
   onModelsChange?: (deletedModel?: DeletedModelRef) => void;
   deleteDisabled?: boolean;
   variant?: "outline" | "ghost" | "muted";
@@ -117,6 +119,7 @@ function ModelSelectorContent({
   onSelect,
   onEject,
   onFoldersChange,
+  onPickLocalModel,
   onModelsChange,
   deleteDisabled,
   className,
@@ -128,6 +131,7 @@ function ModelSelectorContent({
   onSelect: (id: string, meta: ModelSelectorChangeMeta) => void;
   onEject?: () => void;
   onFoldersChange?: () => void;
+  onPickLocalModel?: () => void;
   onModelsChange?: (deletedModel?: DeletedModelRef) => void;
   deleteDisabled?: boolean;
   className?: string;
@@ -170,6 +174,19 @@ function ModelSelectorContent({
         </Tabs>
       )}
 
+      {onPickLocalModel ? (
+        <div className="mt-2 border-t border-border/70 pt-2">
+          <button
+            type="button"
+            onClick={onPickLocalModel}
+            className="flex w-full items-center justify-center gap-1.5 rounded-md px-2 py-1.5 text-xs text-muted-foreground transition-colors hover:bg-muted/60"
+            title="Pick a model file from disk"
+          >
+            <HugeiconsIcon icon={FolderSearchIcon} className="size-3.5" />
+            Pick a model file from disk
+          </button>
+        </div>
+      ) : null}
       {hasSelection && onEject ? (
         <div className="mt-2 border-t border-border/70 pt-2">
           <button
@@ -196,6 +213,7 @@ export function ModelSelector({
   onValueChange,
   onEject,
   onFoldersChange,
+  onPickLocalModel,
   onModelsChange,
   deleteDisabled,
   variant = "outline",
@@ -275,6 +293,11 @@ export function ModelSelector({
     setOpen(false);
   }
 
+  function handlePickLocalModel() {
+    setOpen(false);
+    void onPickLocalModel?.();
+  }
+
   return (
     <Popover open={open} onOpenChange={setOpen}>
       <ModelSelectorTrigger
@@ -292,6 +315,7 @@ export function ModelSelector({
         onSelect={handleSelect}
         onEject={onEject ? handleEject : undefined}
         onFoldersChange={onFoldersChange}
+        onPickLocalModel={onPickLocalModel ? handlePickLocalModel : undefined}
         onModelsChange={onModelsChange}
         deleteDisabled={deleteDisabled}
         className={contentClassName}
diff --git a/studio/frontend/src/components/assistant-ui/thread.tsx b/studio/frontend/src/components/assistant-ui/thread.tsx
index d32f698384..0d6cd3bbf9 100644
--- a/studio/frontend/src/components/assistant-ui/thread.tsx
+++ b/studio/frontend/src/components/assistant-ui/thread.tsx
@@ -33,6 +33,7 @@ import {
 import { sentAudioNames } from "@/features/chat/api/chat-adapter";
 import { useChatRuntimeStore } from "@/features/chat/stores/chat-runtime-store";
 import { applyQwenThinkingParams } from "@/features/chat/utils/qwen-params";
+import { isTauri } from "@/lib/api-base";
 import { deleteThreadMessage } from "@/features/chat/utils/delete-thread-message";
 import { AUDIO_ACCEPT, MAX_AUDIO_SIZE, fileToBase64 } from "@/lib/audio-utils";
 import { copyToClipboard } from "@/lib/copy-to-clipboard";
@@ -298,27 +299,41 @@ const Composer: FC<{ disabled?: boolean }> = ({ disabled }) => {
     [disabled],
   );
 
+  const composerContent = (
+    <>
+      <ComposerAttachments />
+      <PendingAudioChip />
+      <ToolStatusDisplay />
+      <ComposerPrimitive.Input
+        placeholder="Send a message..."
+        className="aui-composer-input mb-1 min-h-12 w-full resize-none overflow-y-auto bg-transparent pl-5 pr-4 pt-2 pb-3 text-sm font-[450] outline-none placeholder:text-muted-foreground focus-visible:ring-0"
+        minRows={1}
+        maxRows={6}
+        autoFocus={!disabled}
+        disabled={disabled}
+        aria-label="Message input"
+      />
+      <ComposerAction disabled={disabled} />
+    </>
+  );
+
   return (
     <ComposerPrimitive.Root
       className="aui-composer-root relative flex w-full flex-col"
       aria-disabled={disabled}
       onSubmit={handleSubmit}
     >
-      <ComposerPrimitive.AttachmentDropzone className="aui-composer-attachment-dropzone chat-composer-surface flex w-full flex-col rounded-3xl bg-background dark:bg-card px-1 pt-2 outline-none transition-shadow data-[dragging=true]:border-ring data-[dragging=true]:bg-accent/50">
-        <ComposerAttachments />
-        <PendingAudioChip />
-        <ToolStatusDisplay />
-        <ComposerPrimitive.Input
-          placeholder="Send a message..."
-          className="aui-composer-input mb-1 min-h-12 w-full resize-none overflow-y-auto bg-transparent pl-5 pr-4 pt-2 pb-3 text-sm font-[450] outline-none placeholder:text-muted-foreground focus-visible:ring-0"
-          minRows={1}
-          maxRows={6}
-          autoFocus={!disabled}
-          disabled={disabled}
-          aria-label="Message input"
-        />
-        <ComposerAction disabled={disabled} />
-      </ComposerPrimitive.AttachmentDropzone>
+      {isTauri ? (
+        // Phase 1 native model drops own Tauri local-path drops. Restore browser
+        // attachment drops in Tauri when Phase 1d adds attachment-token bridging.
+        <div className="aui-composer-attachment-dropzone chat-composer-surface flex w-full flex-col rounded-3xl bg-background dark:bg-card px-1 pt-2 outline-none transition-shadow">
+          {composerContent}
+        </div>
+      ) : (
+        <ComposerPrimitive.AttachmentDropzone className="aui-composer-attachment-dropzone chat-composer-surface flex w-full flex-col rounded-3xl bg-background dark:bg-card px-1 pt-2 outline-none transition-shadow data-[dragging=true]:border-ring data-[dragging=true]:bg-accent/50">
+          {composerContent}
+        </ComposerPrimitive.AttachmentDropzone>
+      )}
     </ComposerPrimitive.Root>
   );
 };
diff --git a/studio/frontend/src/features/chat/api/chat-api.ts b/studio/frontend/src/features/chat/api/chat-api.ts
index 32a03171ee..ec50a3a8d5 100644
--- a/studio/frontend/src/features/chat/api/chat-api.ts
+++ b/studio/frontend/src/features/chat/api/chat-api.ts
@@ -68,7 +68,11 @@ export async function loadModel(
   const response = await authFetch("/api/inference/load", {
     method: "POST",
     headers: { "Content-Type": "application/json" },
-    body: JSON.stringify(payload),
+    body: JSON.stringify({
+      ...payload,
+      native_path_lease: payload.nativePathLease ?? null,
+      nativePathLease: undefined,
+    }),
   });
   return parseJsonOrThrow<LoadModelResponse>(response);
 }
@@ -81,6 +85,7 @@ export async function validateModel(
     headers: { "Content-Type": "application/json" },
     body: JSON.stringify({
       model_path: payload.model_path,
+      native_path_lease: payload.nativePathLease ?? null,
       hf_token: payload.hf_token,
       gguf_variant: payload.gguf_variant ?? null,
     }),
diff --git a/studio/frontend/src/features/chat/chat-page.tsx b/studio/frontend/src/features/chat/chat-page.tsx
index a239cae848..fac9a94d30 100644
--- a/studio/frontend/src/features/chat/chat-page.tsx
+++ b/studio/frontend/src/features/chat/chat-page.tsx
@@ -8,6 +8,14 @@ import {
   ModelSelector,
 } from "@/components/assistant-ui/model-selector";
 import { Thread } from "@/components/assistant-ui/thread";
+import { NativeModelChip } from "@/features/native-intents/components/native-model-chip";
+import { NativeModelDropOverlay } from "@/features/native-intents/components/native-model-drop-overlay";
+import { useChooseNativeModel } from "@/features/native-intents/use-native-dialogs";
+import { useNativeModelDrop } from "@/features/native-intents/use-native-drop";
+import { useNativePathLeasesSupported } from "@/features/native-intents/use-native-readiness";
+import { useNativeIntentStore } from "@/features/native-intents/store";
+import type { NativeIntent } from "@/features/native-intents/types";
+import { isTauri } from "@/lib/api-base";
 import { cn } from "@/lib/utils";
 import { GuidedTour, useGuidedTourController } from "@/features/tour";
 import { useSidebar } from "@/components/ui/sidebar";
@@ -574,6 +582,7 @@ export function ChatPage(): ReactElement {
   const modelsFromStore = useChatRuntimeStore((state) => state.models);
   const lorasFromStore = useChatRuntimeStore((state) => state.loras);
   const modelsError = useChatRuntimeStore((state) => state.modelsError);
+  const modelLoading = useChatRuntimeStore((state) => state.modelLoading);
   const activeThreadId = useChatRuntimeStore((state) => state.activeThreadId);
   const modelOperationInProgress = useChatRuntimeStore(
     (state) => state.modelLoading,
@@ -587,6 +596,8 @@ export function ChatPage(): ReactElement {
     loadProgress,
     loadToastDismissed,
   } = useChatModelRuntime();
+  const pendingNativeModelIntent = useNativeIntentStore((state) => state.pendingModelIntent);
+  const nativePathLeasesSupported = useNativePathLeasesSupported();
   const refreshRef = useRef(refresh);
   const selectModelRef = useRef(selectModel);
 
@@ -618,6 +629,60 @@ export function ChatPage(): ReactElement {
     return { mode: "single" };
   }, [search.thread, search.compare, search.new, activeThreadId]);
 
+  const hasActiveModel = Boolean(inferenceParams.checkpoint);
+  const loadNativeModelIntent = useCallback(
+    async (intent: NativeIntent, loadingDescription: string) => {
+      const label = intent.path.displayLabel || intent.displayLabel || "Local GGUF model";
+      await selectModel({
+        id: label,
+        nativePathToken: intent.path.token,
+        isDownloaded: true,
+        loadingDescription,
+        forceReload: true,
+        throwOnError: true,
+      });
+      useNativeIntentStore.getState().clearModelIntent(intent.id);
+    },
+    [selectModel],
+  );
+  const handleNativeModelDropAutoLoad = useCallback(
+    (intent: NativeIntent) =>
+      loadNativeModelIntent(
+        intent,
+        hasActiveModel
+          ? "Replacing with dropped local GGUF model."
+          : "Loading dropped local GGUF model.",
+      ),
+    [hasActiveModel, loadNativeModelIntent],
+  );
+  const handleNativeModelPickerAutoLoad = useCallback(
+    (intent: NativeIntent) =>
+      loadNativeModelIntent(intent, "Loading chosen local GGUF model."),
+    [loadNativeModelIntent],
+  );
+  const canAutoLoadPickedNativeModel = useCallback(() => {
+    const store = useChatRuntimeStore.getState();
+    return (
+      view.mode === "single" &&
+      nativePathLeasesSupported &&
+      !loadingModel &&
+      !modelLoading &&
+      !store.modelLoading &&
+      !store.params.checkpoint
+    );
+  }, [loadingModel, modelLoading, nativePathLeasesSupported, view.mode]);
+  const chooseNativeModel = useChooseNativeModel({
+    shouldAutoLoad: canAutoLoadPickedNativeModel,
+    onAutoLoad: handleNativeModelPickerAutoLoad,
+  });
+  const nativeModelDropState = useNativeModelDrop({
+    enabled: view.mode === "single",
+    nativePathLeasesSupported,
+    hasActiveModel,
+    isModelLoading: Boolean(loadingModel) || modelLoading,
+    onAutoLoad: handleNativeModelDropAutoLoad,
+  });
+
   const handleCheckpointChange = useCallback(
     (
       value: string,
@@ -922,6 +987,7 @@ export function ChatPage(): ReactElement {
     <div className="flex min-h-0 min-w-0 flex-1 basis-0 bg-background overflow-hidden">
       <GuidedTour {...tour.tourProps} />
       <div className="relative flex min-h-0 min-w-0 flex-1 basis-0 flex-col overflow-hidden">
+        <NativeModelDropOverlay state={nativeModelDropState} />
         <div
           className={cn(
             "absolute top-0 left-0 right-[10px] z-30 flex h-[48px] shrink-0 items-start pt-[11px] pr-2 bg-background",
@@ -940,6 +1006,7 @@ export function ChatPage(): ReactElement {
                 onValueChange={handleCheckpointChange}
                 onEject={handleEject}
                 onFoldersChange={refreshLocalModels}
+                onPickLocalModel={isTauri ? chooseNativeModel : undefined}
                 onModelsChange={refreshModelLists}
                 deleteDisabled={modelOperationInProgress}
                 variant="ghost"
@@ -950,6 +1017,13 @@ export function ChatPage(): ReactElement {
                 className="max-w-[62vw] !pr-3 sm:max-w-none !h-[34px]"
               />
             )}
+            {pendingNativeModelIntent && view.mode !== "compare" ? (
+              <NativeModelChip
+                intent={pendingNativeModelIntent}
+                nativeReadsDisabled={!nativePathLeasesSupported}
+                onLoad={(selection) => selectModel(selection)}
+              />
+            ) : null}
             {loadingModel && loadToastDismissed ? (
               <ModelLoadInlineStatus
                 label={
diff --git a/studio/frontend/src/features/chat/hooks/use-chat-model-runtime.ts b/studio/frontend/src/features/chat/hooks/use-chat-model-runtime.ts
index b9b6f0b543..eacf54a208 100644
--- a/studio/frontend/src/features/chat/hooks/use-chat-model-runtime.ts
+++ b/studio/frontend/src/features/chat/hooks/use-chat-model-runtime.ts
@@ -3,6 +3,7 @@
 
 import { createElement, useCallback, useRef, useState } from "react";
 import { toast } from "sonner";
+import { consumeNativePathToken } from "@/features/native-intents/api";
 import { ModelLoadDescription } from "../components/model-load-status";
 import {
   getDownloadProgress,
@@ -47,6 +48,8 @@ type SelectedModelInput = {
   isDownloaded?: boolean;
   expectedBytes?: number;
   forceReload?: boolean;
+  nativePathToken?: string;
+  throwOnError?: boolean;
 };
 
 const MODEL_LOAD_TOAST_CLASSNAMES = {
@@ -154,6 +157,7 @@ export function useChatModelRuntime() {
     displayName: string;
     isDownloaded?: boolean;
     isCachedLora?: boolean;
+    nativePathToken?: string | null;
   } | null>(null);
   const [loadToastDismissed, setLoadToastDismissed] = useState(false);
   const [loadProgress, setLoadProgress] = useState<{
@@ -338,12 +342,20 @@ export function useChatModelRuntime() {
         typeof selection === "string" ? undefined : selection.ggufVariant;
       const forceReload =
         typeof selection === "string" ? false : selection.forceReload ?? false;
+      const nativePathToken =
+        typeof selection === "string" ? undefined : selection.nativePathToken;
+      const throwOnError =
+        typeof selection === "string" ? false : selection.throwOnError ?? false;
       const currentVariant = useChatRuntimeStore.getState().activeGgufVariant;
       if (!forceReload && (!modelId || (params.checkpoint === modelId && (ggufVariant ?? null) === (currentVariant ?? null)))) {
         return;
       }
       // Prevent duplicate loads if already loading this model
-      if (loadingModelRef.current?.id === modelId) return;
+      if (
+        loadingModelRef.current?.id === modelId &&
+        (loadingModelRef.current?.nativePathToken ?? null) === (nativePathToken ?? null)
+      )
+        return;
 
       const explicitIsLora =
         typeof selection === "string" ? undefined : selection.isLora;
@@ -383,7 +395,13 @@ export function useChatModelRuntime() {
         .join(" ");
       setModelsError(null);
       setLoadToastDismissedState(false);
-      const loadInfo = { id: modelId, displayName, isDownloaded, isCachedLora };
+      const loadInfo = {
+        id: modelId,
+        displayName,
+        isDownloaded,
+        isCachedLora,
+        nativePathToken: nativePathToken ?? null,
+      };
       setLoadingModel(loadInfo);
       useChatRuntimeStore.getState().setModelLoading(true);
       setLoadProgress(
@@ -413,11 +431,17 @@ export function useChatModelRuntime() {
           const hfToken = stateBeforeUnload.hfToken || null;
           const previousModelRequiresTrustRemoteCode =
             stateBeforeUnload.modelRequiresTrustRemoteCode;
+          const previousActiveNativePathToken =
+            stateBeforeUnload.activeNativePathToken;
           try {
             // Lightweight pre-flight validation: avoid unloading a working model
             // if the new identifier is clearly invalid (e.g. bad HF id / path).
+            const validateNativePathLease = nativePathToken
+              ? (await consumeNativePathToken(nativePathToken, "validate-model")).nativePathLease
+              : undefined;
             const validation = await validateModel({
               model_path: modelId,
+              nativePathLease: validateNativePathLease,
               hf_token: hfToken,
               max_seq_length: maxSeqLength,
               load_in_4bit: true,
@@ -428,6 +452,9 @@ export function useChatModelRuntime() {
               throw new Error(getTrustRemoteCodeRequiredMessage(displayName));
             }
             if (abortCtrl.signal.aborted) throw new Error("Cancelled");
+            const loadNativePathLease = nativePathToken
+              ? (await consumeNativePathToken(nativePathToken, "load-model")).nativePathLease
+              : undefined;
 
             if (currentCheckpoint) {
               await unloadModel({ model_path: currentCheckpoint });
@@ -456,6 +483,7 @@ export function useChatModelRuntime() {
             });
             const loadResponse = await loadModel({
               model_path: modelId,
+              nativePathLease: loadNativePathLease,
               hf_token: hfToken,
               max_seq_length: effectiveMaxSeqLength,
               load_in_4bit: true,
@@ -528,6 +556,7 @@ export function useChatModelRuntime() {
               customContextLength: keepCustomCtx,
               defaultChatTemplate: loadResponse.chat_template ?? null,
               chatTemplateOverride: null,
+              activeNativePathToken: nativePathToken ?? null,
             });
             // Qwen3/3.5/3.6: apply thinking-mode-specific params after load
             if (modelId.toLowerCase().includes("qwen3") && (loadResponse.supports_reasoning ?? false)) {
@@ -564,9 +593,22 @@ export function useChatModelRuntime() {
             if (abortCtrl.signal.aborted) throw error;
             // If we unloaded a previous model and the new load failed, attempt a rollback.
             if (previousWasUnloaded && previousCheckpoint) {
+              let rollbackNativePathLease: string | undefined;
+              if (previousActiveNativePathToken) {
+                try {
+                  rollbackNativePathLease = (
+                    await consumeNativePathToken(previousActiveNativePathToken, "load-model")
+                  ).nativePathLease;
+                } catch {
+                  throw new Error(
+                    "Could not reload the previous local model: please re-select the file.",
+                  );
+                }
+              }
               try {
                 await loadModel({
                   model_path: previousCheckpoint,
+                  nativePathLease: rollbackNativePathLease,
                   hf_token: hfToken,
                   max_seq_length: rollbackMaxSeqLength,
                   load_in_4bit: true,
@@ -575,9 +617,12 @@ export function useChatModelRuntime() {
                   trust_remote_code:
                     previousModelRequiresTrustRemoteCode || trustRemoteCode,
                 });
+                useChatRuntimeStore.setState({
+                  activeNativePathToken: previousActiveNativePathToken ?? null,
+                });
                 await refresh();
               } catch {
-                // If rollback also fails, surface the original error.
+                // Rollback also failed; surface the original load error below.
               }
             }
             throw error;
@@ -879,6 +924,9 @@ export function useChatModelRuntime() {
         const message =
           error instanceof Error ? error.message : "Failed to load model";
         setModelsError(message);
+        if (throwOnError) {
+          throw error instanceof Error ? error : new Error(message);
+        }
       }
     },
     [
diff --git a/studio/frontend/src/features/chat/shared-composer.tsx b/studio/frontend/src/features/chat/shared-composer.tsx
index 1a3ba7e166..1741b7e409 100644
--- a/studio/frontend/src/features/chat/shared-composer.tsx
+++ b/studio/frontend/src/features/chat/shared-composer.tsx
@@ -12,6 +12,7 @@ import {
 } from "@/components/ui/dropdown-menu";
 import { applyQwenThinkingParams } from "@/features/chat/utils/qwen-params";
 import { AUDIO_ACCEPT, MAX_AUDIO_SIZE, fileToBase64 } from "@/lib/audio-utils";
+import { isTauri } from "@/lib/api-base";
 import { useAui } from "@assistant-ui/react";
 import { cn } from "@/lib/utils";
 import { ArrowUpIcon, GlobeIcon, HeadphonesIcon, LightbulbIcon, LightbulbOffIcon, MicIcon, PlusIcon, SquareIcon, XIcon } from "lucide-react";
@@ -493,11 +494,15 @@ export function SharedComposer({
     <div
       className={`chat-composer-surface relative flex w-full flex-col rounded-3xl bg-background dark:bg-card px-1 pt-2 transition-shadow outline-none ${dragging ? "border-ring bg-accent/50" : ""}`}
       onDragOver={(e) => {
+        if (isTauri) return;
         e.preventDefault();
         setDragging(true);
       }}
       onDragLeave={() => setDragging(false)}
       onDrop={(e) => {
+        // Phase 1 native model drops own Tauri local-path drops. Restore browser
+        // attachment drops in Tauri when Phase 1d adds attachment-token bridging.
+        if (isTauri) return;
         e.preventDefault();
         setDragging(false);
         addFiles(e.dataTransfer.files);
diff --git a/studio/frontend/src/features/chat/stores/chat-runtime-store.ts b/studio/frontend/src/features/chat/stores/chat-runtime-store.ts
index 13f2a23c36..d67dbc36d8 100644
--- a/studio/frontend/src/features/chat/stores/chat-runtime-store.ts
+++ b/studio/frontend/src/features/chat/stores/chat-runtime-store.ts
@@ -226,6 +226,7 @@ type ChatRuntimeStore = {
     cachedTokens: number;
   } | null;
   modelLoading: boolean;
+  activeNativePathToken: string | null;
   setModelLoading: (loading: boolean) => void;
   setModelRequiresTrustRemoteCode: (required: boolean) => void;
   setParams: (params: InferenceParams) => void;
@@ -305,6 +306,7 @@ export const useChatRuntimeStore = create<ChatRuntimeStore>((set) => ({
   pendingAudioName: null,
   contextUsage: null,
   modelLoading: false,
+  activeNativePathToken: null,
   setModelLoading: (loading) => set({ modelLoading: loading }),
   setModelRequiresTrustRemoteCode: (modelRequiresTrustRemoteCode) =>
     set({ modelRequiresTrustRemoteCode }),
@@ -378,6 +380,7 @@ export const useChatRuntimeStore = create<ChatRuntimeStore>((set) => ({
         checkpoint: "",
       },
       activeGgufVariant: null,
+      activeNativePathToken: null,
       ggufContextLength: null,
       ggufMaxContextLength: null,
       ggufNativeContextLength: null,
diff --git a/studio/frontend/src/features/chat/types/api.ts b/studio/frontend/src/features/chat/types/api.ts
index 15d0457722..e5e7151dcc 100644
--- a/studio/frontend/src/features/chat/types/api.ts
+++ b/studio/frontend/src/features/chat/types/api.ts
@@ -32,6 +32,7 @@ export interface ListLorasResponse {
 
 export interface LoadModelRequest {
   model_path: string;
+  nativePathLease?: string | null;
   hf_token: string | null;
   max_seq_length: number;
   load_in_4bit: boolean;
diff --git a/studio/frontend/src/features/native-intents/api.ts b/studio/frontend/src/features/native-intents/api.ts
new file mode 100644
index 0000000000..d849f1d4b2
--- /dev/null
+++ b/studio/frontend/src/features/native-intents/api.ts
@@ -0,0 +1,46 @@
+import { isTauri } from "@/lib/api-base";
+import type {
+  NativeIntent,
+  NativePathLeaseResponse,
+  NativePathOperation,
+} from "./types";
+
+async function invokeNative<T>(command: string, args?: Record<string, unknown>): Promise<T> {
+  if (!isTauri) {
+    throw new Error("Native desktop features are only available in the Tauri app.");
+  }
+  const { invoke } = await import("@tauri-apps/api/core");
+  return invoke<T>(command, args);
+}
+
+export async function drainNativeIntents(): Promise<NativeIntent[]> {
+  if (!isTauri) return [];
+  return invokeNative<NativeIntent[]>("drain_native_intents");
+}
+
+export async function pickNativeModel(): Promise<NativeIntent | null> {
+  if (!isTauri) return null;
+  return invokeNative<NativeIntent | null>("pick_native_model");
+}
+
+export async function registerNativeModelPath(path: string): Promise<NativeIntent> {
+  return invokeNative<NativeIntent>("register_native_model_path", { path });
+}
+
+export async function consumeNativePathToken(
+  token: string,
+  operation: NativePathOperation,
+): Promise<NativePathLeaseResponse> {
+  return invokeNative<NativePathLeaseResponse>("consume_native_path_token", {
+    token,
+    operation,
+  });
+}
+
+export async function revealPathToken(token: string): Promise<void> {
+  return invokeNative<void>("reveal_path_token", { token });
+}
+
+export async function openPathToken(token: string): Promise<void> {
+  return invokeNative<void>("open_path_token", { token });
+}
diff --git a/studio/frontend/src/features/native-intents/components/native-model-chip.tsx b/studio/frontend/src/features/native-intents/components/native-model-chip.tsx
new file mode 100644
index 0000000000..22ae36c503
--- /dev/null
+++ b/studio/frontend/src/features/native-intents/components/native-model-chip.tsx
@@ -0,0 +1,107 @@
+import { revealPathToken } from "../api";
+import { useNativeIntentStore } from "../store";
+import type { NativeIntent } from "../types";
+import { XIcon } from "lucide-react";
+import { useEffect, useState } from "react";
+import { toast } from "sonner";
+
+interface NativeModelChipProps {
+  intent: NativeIntent;
+  nativeReadsDisabled: boolean;
+  onLoad: (selection: {
+    id: string;
+    nativePathToken: string;
+    isDownloaded: boolean;
+    loadingDescription: string;
+    forceReload: boolean;
+    throwOnError?: boolean;
+  }) => Promise<void> | void;
+}
+
+export function NativeModelChip({
+  intent,
+  nativeReadsDisabled,
+  onLoad,
+}: NativeModelChipProps) {
+  const clearModelIntent = useNativeIntentStore((state) => state.clearModelIntent);
+  const [loading, setLoading] = useState(false);
+  const [now, setNow] = useState(() => Date.now());
+  const label = intent.path.displayLabel || intent.displayLabel || "Local GGUF model";
+  const expired = intent.path.expiresAtMs <= now;
+
+  useEffect(() => {
+    if (expired) return;
+    const remaining = Math.max(0, intent.path.expiresAtMs - Date.now());
+    const timer = window.setTimeout(() => setNow(Date.now()), remaining);
+    return () => window.clearTimeout(timer);
+  }, [expired, intent.path.expiresAtMs]);
+
+  async function handleLoad() {
+    if (nativeReadsDisabled || expired) return;
+    setLoading(true);
+    try {
+      await onLoad({
+        id: label,
+        nativePathToken: intent.path.token,
+        isDownloaded: true,
+        loadingDescription: "Loading selected local GGUF model.",
+        forceReload: true,
+        throwOnError: true,
+      });
+      clearModelIntent(intent.id);
+    } catch {
+      // selectModel reports the failure; keep the chip available for retry.
+    } finally {
+      setLoading(false);
+    }
+  }
+
+  async function handleReveal() {
+    try {
+      await revealPathToken(intent.path.token);
+    } catch (error) {
+      toast.error("Could not reveal model", {
+        description: error instanceof Error ? error.message : String(error),
+      });
+    }
+  }
+
+  return (
+    <div className="flex min-w-0 max-w-[34rem] items-center gap-2 rounded-lg border border-border/70 bg-muted/70 px-2.5 py-1.5 text-xs">
+      <span className="shrink-0 font-medium text-muted-foreground">Local GGUF</span>
+      <span className="min-w-0 flex-1 truncate" title={label}>{label}</span>
+      <button
+        type="button"
+        onClick={handleReveal}
+        disabled={expired}
+        title={expired ? "Selection expired, pick or drop the file again" : undefined}
+        className="rounded-md px-2 py-1 text-muted-foreground transition-colors hover:bg-background hover:text-foreground disabled:cursor-not-allowed disabled:opacity-50"
+      >
+        Reveal
+      </button>
+      <button
+        type="button"
+        onClick={handleLoad}
+        disabled={nativeReadsDisabled || expired || loading}
+        title={
+          nativeReadsDisabled
+            ? "Managed desktop backend required"
+            : expired
+              ? "Selection expired, pick or drop the file again"
+              : undefined
+        }
+        className="rounded-md bg-foreground px-2 py-1 text-background transition-opacity disabled:cursor-not-allowed disabled:opacity-50"
+      >
+        {expired ? "Expired" : loading ? "Loading…" : "Load model"}
+      </button>
+      <button
+        type="button"
+        onClick={() => clearModelIntent(intent.id)}
+        className="flex size-5 items-center justify-center rounded-full text-muted-foreground hover:bg-destructive hover:text-destructive-foreground"
+        aria-label="Dismiss local model"
+      >
+        <XIcon className="size-3" />
+      </button>
+    </div>
+  );
+}
diff --git a/studio/frontend/src/features/native-intents/components/native-model-drop-overlay.tsx b/studio/frontend/src/features/native-intents/components/native-model-drop-overlay.tsx
new file mode 100644
index 0000000000..725fb14259
--- /dev/null
+++ b/studio/frontend/src/features/native-intents/components/native-model-drop-overlay.tsx
@@ -0,0 +1,70 @@
+import { cn } from "@/lib/utils";
+import { FileUpIcon } from "lucide-react";
+import type { NativeModelDropState } from "../use-native-drop";
+
+function overlayCopy(state: NativeModelDropState): { title: string; description: string } {
+  if (state.status === "invalid") {
+    return {
+      title: "GGUF models only",
+      description: "Other files are not handled here yet.",
+    };
+  }
+  if (state.status === "valid" && state.action === "replace") {
+    return {
+      title: "Drop to replace model",
+      description: "Current model will unload first.",
+    };
+  }
+  if (state.status === "valid" && state.action === "load") {
+    return {
+      title: "Drop to load model",
+      description: "Adds it as the active chat model.",
+    };
+  }
+  return {
+    title: "Drop to add model chip",
+    description: "Review it before loading.",
+  };
+}
+
+export function NativeModelDropOverlay({ state }: { state: NativeModelDropState }) {
+  const isIdle = state.status === "idle";
+  const isAutoLoad = state.status === "valid" && state.action !== "chip";
+  const isInvalid = state.status === "invalid";
+  const { title, description } = overlayCopy(state);
+
+  return (
+    <div
+      className={cn(
+        "pointer-events-none absolute left-1/2 top-4 z-50 w-[clamp(16rem,28vw,22rem)] max-w-[calc(100vw-1rem)] -translate-x-1/2 transition-all duration-200 ease-out",
+        isIdle ? "-translate-y-1 opacity-0" : "translate-y-0 opacity-100",
+      )}
+      role="status"
+      aria-live="polite"
+      aria-hidden={isIdle}
+    >
+      <div className="flex items-center gap-2.5 rounded-xl border border-border/70 bg-card/96 px-3 py-2.5 shadow-sm shadow-black/10 backdrop-blur-sm dark:shadow-black/30">
+        <div
+          className={cn(
+            "flex size-8 shrink-0 items-center justify-center rounded-lg border transition-colors duration-200 ease-out",
+            isAutoLoad
+              ? "border-emerald-500/30 bg-emerald-500/10 text-emerald-600 dark:text-emerald-400"
+              : isInvalid
+                ? "border-amber-500/25 bg-amber-500/10 text-amber-700 dark:text-amber-300"
+                : "border-border bg-muted text-muted-foreground",
+          )}
+        >
+          <FileUpIcon className="size-4" aria-hidden="true" />
+        </div>
+        <div className="min-w-0 flex-1">
+          <div className="truncate text-xs font-medium text-foreground">
+            {title}
+          </div>
+          <div className="mt-0.5 truncate text-[11px] leading-4 text-muted-foreground">
+            {description}
+          </div>
+        </div>
+      </div>
+    </div>
+  );
+}
diff --git a/studio/frontend/src/features/native-intents/native-intent-drain.tsx b/studio/frontend/src/features/native-intents/native-intent-drain.tsx
new file mode 100644
index 0000000000..109e8e27f9
--- /dev/null
+++ b/studio/frontend/src/features/native-intents/native-intent-drain.tsx
@@ -0,0 +1,39 @@
+import { isTauri } from "@/lib/api-base";
+import { useEffect } from "react";
+import { drainNativeIntents } from "./api";
+import { useNativeIntentStore } from "./store";
+
+export function NativeIntentDrain() {
+  const addIntent = useNativeIntentStore((state) => state.addIntent);
+
+  useEffect(() => {
+    if (!isTauri) return;
+    let disposed = false;
+    let unlisten: (() => void) | undefined;
+
+    async function drain() {
+      const intents = await drainNativeIntents().catch(() => []);
+      if (disposed) return;
+      for (const intent of intents) addIntent(intent);
+    }
+
+    void drain();
+    void import("@tauri-apps/api/event")
+      .then(({ listen }) => listen("native-intent-available", drain))
+      .then((cleanup) => {
+        if (disposed) {
+          cleanup();
+        } else {
+          unlisten = cleanup;
+        }
+      })
+      .catch(() => undefined);
+
+    return () => {
+      disposed = true;
+      unlisten?.();
+    };
+  }, [addIntent]);
+
+  return null;
+}
diff --git a/studio/frontend/src/features/native-intents/store.ts b/studio/frontend/src/features/native-intents/store.ts
new file mode 100644
index 0000000000..a9cd9e7966
--- /dev/null
+++ b/studio/frontend/src/features/native-intents/store.ts
@@ -0,0 +1,23 @@
+import { create } from "zustand";
+import type { NativeIntent } from "./types";
+
+interface NativeIntentState {
+  pendingModelIntent: NativeIntent | null;
+  addIntent: (intent: NativeIntent) => void;
+  clearModelIntent: (intentId?: string) => void;
+}
+
+export const useNativeIntentStore = create<NativeIntentState>((set, get) => ({
+  pendingModelIntent: null,
+  addIntent: (intent) => {
+    if (intent.kind !== "model") return;
+    const current = get().pendingModelIntent;
+    if (current?.path.token === intent.path.token) return;
+    set({ pendingModelIntent: intent });
+  },
+  clearModelIntent: (intentId) => {
+    const current = get().pendingModelIntent;
+    if (intentId && current?.id !== intentId) return;
+    set({ pendingModelIntent: null });
+  },
+}));
diff --git a/studio/frontend/src/features/native-intents/types.ts b/studio/frontend/src/features/native-intents/types.ts
new file mode 100644
index 0000000000..268638559a
--- /dev/null
+++ b/studio/frontend/src/features/native-intents/types.ts
@@ -0,0 +1,39 @@
+export type NativePathOperation =
+  | "validate-model"
+  | "load-model"
+  | "dataset-preview"
+  | "dataset-import"
+  | "attach"
+  | "reveal"
+  | "open";
+
+export type NativePathKind = "model" | "dataset" | "attachment" | "artifact";
+
+export type NativePathSourceKind =
+  | "dialog"
+  | "drop"
+  | "deep-link"
+  | "file-association"
+  | "artifact";
+
+export interface NativePathRef {
+  token: string;
+  kind: NativePathKind;
+  displayLabel: string;
+  allowedOperations: NativePathOperation[];
+  expiresAtMs: number;
+}
+
+export interface NativeIntent {
+  id: string;
+  kind: NativePathKind;
+  sourceKind: NativePathSourceKind;
+  path: NativePathRef;
+  displayLabel: string;
+}
+
+export interface NativePathLeaseResponse {
+  nativePathLease: string;
+  displayLabel: string;
+  expiresAtMs: number;
+}
diff --git a/studio/frontend/src/features/native-intents/use-native-dialogs.ts b/studio/frontend/src/features/native-intents/use-native-dialogs.ts
new file mode 100644
index 0000000000..45ca81343a
--- /dev/null
+++ b/studio/frontend/src/features/native-intents/use-native-dialogs.ts
@@ -0,0 +1,68 @@
+import { useCallback, useRef } from "react";
+import { toast } from "sonner";
+import { pickNativeModel } from "./api";
+import { useNativeIntentStore } from "./store";
+import type { NativeIntent } from "./types";
+
+interface ChooseNativeModelOptions {
+  shouldAutoLoad?: (intent: NativeIntent) => boolean;
+  onAutoLoad?: (intent: NativeIntent) => Promise<void> | void;
+}
+
+function isGgufModelIntent(intent: NativeIntent): boolean {
+  const label = intent.path.displayLabel || intent.displayLabel;
+  return (
+    intent.kind === "model" &&
+    intent.path.kind === "model" &&
+    label.toLowerCase().endsWith(".gguf") &&
+    intent.path.allowedOperations.includes("validate-model") &&
+    intent.path.allowedOperations.includes("load-model")
+  );
+}
+
+export function useChooseNativeModel(options: ChooseNativeModelOptions = {}) {
+  const addIntent = useNativeIntentStore((state) => state.addIntent);
+  const pickingRef = useRef(false);
+  const onAutoLoad = options.onAutoLoad;
+  const shouldAutoLoad = options.shouldAutoLoad;
+
+  return useCallback(async () => {
+    if (pickingRef.current) return;
+    pickingRef.current = true;
+    try {
+      let intent: NativeIntent | null = null;
+      try {
+        intent = await pickNativeModel();
+      } catch (error) {
+        toast.error("Could not choose local model", {
+          description: error instanceof Error ? error.message : String(error),
+        });
+        return;
+      }
+
+      if (!intent) return;
+
+      let runAutoLoad = false;
+      try {
+        runAutoLoad =
+          Boolean(onAutoLoad) &&
+          isGgufModelIntent(intent) &&
+          shouldAutoLoad?.(intent) === true;
+      } catch {
+        runAutoLoad = false;
+      }
+      if (!runAutoLoad) {
+        addIntent(intent);
+        return;
+      }
+
+      try {
+        await onAutoLoad?.(intent);
+      } catch {
+        addIntent(intent);
+      }
+    } finally {
+      pickingRef.current = false;
+    }
+  }, [addIntent, onAutoLoad, shouldAutoLoad]);
+}
diff --git a/studio/frontend/src/features/native-intents/use-native-drop.ts b/studio/frontend/src/features/native-intents/use-native-drop.ts
new file mode 100644
index 0000000000..5d35511c10
--- /dev/null
+++ b/studio/frontend/src/features/native-intents/use-native-drop.ts
@@ -0,0 +1,132 @@
+import { isTauri } from "@/lib/api-base";
+import { useEffect, useRef, useState } from "react";
+import { toast } from "sonner";
+import { registerNativeModelPath } from "./api";
+import { useNativeIntentStore } from "./store";
+import type { NativeIntent } from "./types";
+
+export type NativeModelDropState =
+  | { status: "idle" }
+  | { status: "valid"; action: "load" | "replace" | "chip" }
+  | { status: "invalid" };
+
+interface NativeModelDropOptions {
+  enabled?: boolean;
+  nativePathLeasesSupported: boolean;
+  hasActiveModel: boolean;
+  isModelLoading: boolean;
+  onAutoLoad?: (intent: NativeIntent) => Promise<void> | void;
+}
+
+function ggufPaths(paths: string[]): string[] {
+  return paths.filter((path) => path.toLowerCase().endsWith(".gguf"));
+}
+
+function canAutoLoadPaths(paths: string[], options: NativeModelDropOptions): boolean {
+  return (
+    paths.length === 1 &&
+    ggufPaths(paths).length === 1 &&
+    options.nativePathLeasesSupported &&
+    !options.isModelLoading &&
+    Boolean(options.onAutoLoad)
+  );
+}
+
+function dropStateForPaths(
+  paths: string[],
+  options: NativeModelDropOptions,
+): NativeModelDropState {
+  if (paths.length === 0) {
+    return { status: "idle" };
+  }
+  const ggufs = ggufPaths(paths);
+  if (paths.length !== 1 || ggufs.length !== 1) {
+    return { status: "invalid" };
+  }
+  if (!canAutoLoadPaths(paths, options)) {
+    return { status: "valid", action: "chip" };
+  }
+  return {
+    status: "valid",
+    action: options.hasActiveModel ? "replace" : "load",
+  };
+}
+
+export function useNativeModelDrop(options: NativeModelDropOptions): NativeModelDropState {
+  const { enabled = true } = options;
+  const addIntent = useNativeIntentStore((state) => state.addIntent);
+  const [dropState, setDropState] = useState<NativeModelDropState>({ status: "idle" });
+  const optionsRef = useRef(options);
+  optionsRef.current = options;
+
+  useEffect(() => {
+    if (!isTauri || !enabled) {
+      setDropState({ status: "idle" });
+      return;
+    }
+    let disposed = false;
+    let unlisten: (() => void) | undefined;
+
+    void import("@tauri-apps/api/window")
+      .then(({ getCurrentWindow }) => getCurrentWindow().onDragDropEvent(async (event) => {
+        const currentOptions = optionsRef.current;
+        if (event.payload.type === "enter") {
+          setDropState(dropStateForPaths(event.payload.paths, currentOptions));
+          return;
+        }
+        if (event.payload.type === "leave") {
+          setDropState({ status: "idle" });
+          return;
+        }
+        if (event.payload.type !== "drop") return;
+        setDropState({ status: "idle" });
+        const ggufs = ggufPaths(event.payload.paths);
+        if (event.payload.paths.length !== 1 || ggufs.length !== 1) {
+          if (event.payload.paths.length > 0) {
+            toast.error(
+              ggufs.length === 0
+                ? "Only .gguf model files can be dropped here."
+                : "Drop a single .gguf model file.",
+            );
+          }
+          return;
+        }
+        const ggufPath = ggufs[0];
+        try {
+          const intent = await registerNativeModelPath(ggufPath);
+          if (disposed) return;
+          if (!canAutoLoadPaths(event.payload.paths, currentOptions)) {
+            addIntent(intent);
+            return;
+          }
+          try {
+            await currentOptions.onAutoLoad?.(intent);
+          } catch (error) {
+            addIntent(intent);
+            toast.error("Could not load dropped model", {
+              description: error instanceof Error ? error.message : String(error),
+            });
+          }
+        } catch (error) {
+          toast.error("Could not use dropped model", {
+            description: error instanceof Error ? error.message : String(error),
+          });
+        }
+      }))
+      .then((cleanup) => {
+        if (disposed) {
+          cleanup();
+        } else {
+          unlisten = cleanup;
+        }
+      })
+      .catch(() => undefined);
+
+    return () => {
+      disposed = true;
+      unlisten?.();
+    };
+  }, [addIntent, enabled]);
+
+  return dropState;
+}
diff --git a/studio/frontend/src/features/native-intents/use-native-readiness.ts b/studio/frontend/src/features/native-intents/use-native-readiness.ts
new file mode 100644
index 0000000000..704ae9bc96
--- /dev/null
+++ b/studio/frontend/src/features/native-intents/use-native-readiness.ts
@@ -0,0 +1,46 @@
+import { apiUrl, isTauri } from "@/lib/api-base";
+import { useEffect, useState } from "react";
+
+const MAX_READINESS_POLLS = 720;
+
+export function useNativePathLeasesSupported(): boolean {
+  const [supported, setSupported] = useState(false);
+
+  useEffect(() => {
+    if (!isTauri) return;
+    let disposed = false;
+    let timer: ReturnType<typeof setTimeout> | undefined;
+    let controller: AbortController | undefined;
+    let polls = 0;
+
+    function check(delay = 0) {
+      if (polls >= MAX_READINESS_POLLS) return;
+      polls += 1;
+      timer = setTimeout(() => {
+        controller = new AbortController();
+        fetch(apiUrl("/api/health"), { signal: controller.signal })
+          .then((response) => response.json())
+          .then((health) => {
+            if (disposed) return;
+            if (health?.native_path_leases_supported === true) {
+              setSupported(true);
+            } else {
+              check(5000);
+            }
+          })
+          .catch(() => {
+            if (!disposed) check(5000);
+          });
+      }, delay);
+    }
+
+    check();
+    return () => {
+      disposed = true;
+      if (timer) clearTimeout(timer);
+      controller?.abort();
+    };
+  }, []);
+
+  return supported;
+}
diff --git a/studio/install_python_stack.py b/studio/install_python_stack.py
index 25d950fd3e..3fd1e6af66 100644
--- a/studio/install_python_stack.py
+++ b/studio/install_python_stack.py
@@ -21,6 +21,10 @@ import tempfile
 import urllib.request
 from pathlib import Path
 
+_BACKEND_DIR = Path(__file__).resolve().parent / "backend"
+if str(_BACKEND_DIR) not in sys.path:
+    sys.path.insert(1, str(_BACKEND_DIR))
+
 from backend.utils.wheel_utils import (
     flash_attn_package_version,
     flash_attn_wheel_url,
diff --git a/studio/src-tauri/Cargo.lock b/studio/src-tauri/Cargo.lock
index e491ee534a..dd57e55dc5 100644
--- a/studio/src-tauri/Cargo.lock
+++ b/studio/src-tauri/Cargo.lock
@@ -808,6 +808,7 @@ checksum = "9ed9a281f7bc9b7576e61468ba615a66a5c8cfdff42420a70aa82701a3b1e292"
 dependencies = [
  "block-buffer",
  "crypto-common",
+ "subtle",
 ]
 
 [[package]]
@@ -1694,6 +1695,15 @@ version = "0.4.3"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "7f24254aa9a54b5c858eaee2f5bccdb46aaf0e486a595ed5fd8f86ba55232a70"
 
+[[package]]
+name = "hmac"
+version = "0.12.1"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "6c49c37c09c17a53d937dfbb742eb3a961d65a994e6bcdcf37e7399d0cc8ab5e"
+dependencies = [
+ "digest",
+]
+
 [[package]]
 name = "home"
 version = "0.5.12"
@@ -3561,6 +3571,30 @@ dependencies = [
  "web-sys",
 ]
 
+[[package]]
+name = "rfd"
+version = "0.16.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "a15ad77d9e70a92437d8f74c35d99b4e4691128df018833e99f90bcd36152672"
+dependencies = [
+ "block2",
+ "dispatch2",
+ "glib-sys",
+ "gobject-sys",
+ "gtk-sys",
+ "js-sys",
+ "log",
+ "objc2",
+ "objc2-app-kit",
+ "objc2-core-foundation",
+ "objc2-foundation",
+ "raw-window-handle",
+ "wasm-bindgen",
+ "wasm-bindgen-futures",
+ "web-sys",
+ "windows-sys 0.60.2",
+]
+
 [[package]]
 name = "ring"
 version = "0.17.14"
@@ -4513,6 +4547,48 @@ dependencies = [
  "thiserror 2.0.18",
 ]
 
+[[package]]
+name = "tauri-plugin-dialog"
+version = "2.7.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "a1fa4150c95ae391946cc8b8f905ab14797427caba3a8a2f79628e956da91809"
+dependencies = [
+ "log",
+ "raw-window-handle",
+ "rfd",
+ "serde",
+ "serde_json",
+ "tauri",
+ "tauri-plugin",
+ "tauri-plugin-fs",
+ "thiserror 2.0.18",
+ "url",
+]
+
+[[package]]
+name = "tauri-plugin-fs"
+version = "2.5.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "36e1ec28b79f3d0683f4507e1615c36292c0ea6716668770d4396b9b39871ed8"
+dependencies = [
+ "anyhow",
+ "dunce",
+ "glob",
+ "log",
+ "objc2-foundation",
+ "percent-encoding",
+ "schemars 0.8.22",
+ "serde",
+ "serde_json",
+ "serde_repr",
+ "tauri",
+ "tauri-plugin",
+ "tauri-utils",
+ "thiserror 2.0.18",
+ "toml 0.9.12+spec-1.1.0",
+ "url",
+]
+
 [[package]]
 name = "tauri-plugin-opener"
 version = "2.5.3"
@@ -5200,9 +5276,11 @@ checksum = "ebc1c04c71510c7f702b52b7c350734c9ff1295c464a03335b00bb84fc54f853"
 name = "unsloth-studio"
 version = "2026.4.8"
 dependencies = [
+ "base64 0.22.1",
  "dirs",
  "elevated-command",
  "fix-path-env",
+ "hmac",
  "libc",
  "log",
  "open",
@@ -5212,10 +5290,12 @@ dependencies = [
  "reqwest 0.12.28",
  "serde",
  "serde_json",
+ "sha2",
  "simplelog",
  "tauri",
  "tauri-build",
  "tauri-plugin-clipboard-manager",
+ "tauri-plugin-dialog",
  "tauri-plugin-opener",
  "tauri-plugin-process",
  "tauri-plugin-single-instance",
diff --git a/studio/src-tauri/Cargo.toml b/studio/src-tauri/Cargo.toml
index 269bb7a92a..d3cfe49667 100644
--- a/studio/src-tauri/Cargo.toml
+++ b/studio/src-tauri/Cargo.toml
@@ -11,6 +11,9 @@ tauri-plugin-single-instance = "2"
 tauri-plugin-process = "2"
 serde = { version = "1", features = ["derive"] }
 serde_json = "1"
+base64 = "0.22"
+hmac = "0.12"
+sha2 = "0.10"
 reqwest = { version = "0.12", features = ["json"] }
 tokio = { version = "1", features = ["full"] }
 log = "0.4"
@@ -23,6 +26,7 @@ fix-path-env = { git = "https://github.com/tauri-apps/fix-path-env-rs" }
 tauri-plugin-opener = "2.5.3"
 tauri-plugin-updater = "2"
 tauri-plugin-clipboard-manager = "2"
+tauri-plugin-dialog = "2"
 rand = "0.10.0"
 
 [target.'cfg(unix)'.dependencies]
diff --git a/studio/src-tauri/src/diagnostics/redaction.rs b/studio/src-tauri/src/diagnostics/redaction.rs
index 5c5589a494..0c7a92944a 100644
--- a/studio/src-tauri/src/diagnostics/redaction.rs
+++ b/studio/src-tauri/src/diagnostics/redaction.rs
@@ -20,6 +20,12 @@ pub(crate) fn redact_text(text: &str, report: &mut RedactionReport) -> String {
     out = replace_regex(cookie_re(), &out, "$1: <redacted>", report);
     out = replace_regex(token_re(), &out, "<redacted token>", report);
     out = replace_regex(env_secret_re(), &out, "$1=<redacted>", report);
+    out = replace_regex(
+        native_path_lease_re(),
+        &out,
+        "$1<redacted native path lease>",
+        report,
+    );
     out = replace_known_paths(&out, report);
     out = replace_regex(windows_studio_re(), &out, "<studio_home>", report);
     out = replace_regex(windows_home_re(), &out, "%USERPROFILE%", report);
@@ -126,6 +132,16 @@ fn env_secret_re() -> &'static Regex {
     })
 }
 
+fn native_path_lease_re() -> &'static Regex {
+    static RE: OnceLock<Regex> = OnceLock::new();
+    RE.get_or_init(|| {
+        Regex::new(
+            r#"(?i)(\b(?:native_path_lease|nativePathLease)[\"']?\s*[:=]\s*[\"']?)[A-Za-z0-9_-]+\.[A-Za-z0-9_-]+"#,
+        )
+        .unwrap()
+    })
+}
+
 fn windows_studio_re() -> &'static Regex {
     static RE: OnceLock<Regex> = OnceLock::new();
     RE.get_or_init(|| {
@@ -165,6 +181,7 @@ mod tests {
             "Cookie: session=abcdef\n",
             "HF_TOKEN=hf_abcdefghijklmnopqrstuvwxyz\n",
             "API_KEY=secret123\n",
+            "native_path_lease=abc.DEF_123\n",
             "url=https://user:pass@example.com/path\n",
             "email=alex@example.com\n",
             "path=/Users/alex/.unsloth/studio/logs/install.log\n",
@@ -178,6 +195,8 @@ mod tests {
         assert!(!redacted.contains("session=abcdef"));
         assert!(!redacted.contains("hf_abcdefghijklmnopqrstuvwxyz"));
         assert!(!redacted.contains("secret123"));
+        assert!(!redacted.contains("abc.DEF_123"));
+        assert!(redacted.contains("native_path_lease=<redacted native path lease>"));
         assert!(redacted.contains("https://<redacted>@example.com/path"));
         assert!(!redacted.contains("alex@example.com"));
         assert!(redacted.contains("<studio_home>"));
diff --git a/studio/src-tauri/src/main.rs b/studio/src-tauri/src/main.rs
index b4e718d867..5aeaa56497 100644
--- a/studio/src-tauri/src/main.rs
+++ b/studio/src-tauri/src/main.rs
@@ -4,6 +4,9 @@ mod commands;
 mod desktop_auth;
 mod diagnostics;
 mod install;
+mod native_backend_lease;
+mod native_intents;
+mod native_path_policy;
 mod preflight;
 mod process;
 mod update;
@@ -164,10 +167,12 @@ fn main() {
         }))
         .plugin(tauri_plugin_process::init())
         .plugin(tauri_plugin_opener::init())
+        .plugin(tauri_plugin_dialog::init())
         .plugin(tauri_plugin_updater::Builder::new().build())
         .plugin(tauri_plugin_clipboard_manager::init())
         .manage(diagnostics::new_diagnostics_state())
         .manage(install::new_install_state())
+        .manage(native_intents::new_native_intake_state())
         .manage(new_backend_state())
         .manage(process::new_shutdown_flag())
         .manage(update::new_update_state())
@@ -187,6 +192,13 @@ fn main() {
             commands::install_system_packages,
             desktop_auth::desktop_auth,
             diagnostics::collect_support_diagnostics,
+            native_intents::drain_native_intents,
+            native_intents::register_native_model_path,
+            native_intents::pick_native_model,
+            native_intents::consume_native_path_token,
+            native_intents::register_artifact_path,
+            native_intents::reveal_path_token,
+            native_intents::open_path_token,
         ])
         .setup(|app| {
             #[cfg(any(target_os = "windows", target_os = "linux"))]
diff --git a/studio/src-tauri/src/native_backend_lease.rs b/studio/src-tauri/src/native_backend_lease.rs
new file mode 100644
index 0000000000..8c94a02f86
--- /dev/null
+++ b/studio/src-tauri/src/native_backend_lease.rs
@@ -0,0 +1,199 @@
+use base64::{engine::general_purpose::URL_SAFE_NO_PAD, Engine as _};
+use hmac::{Hmac, Mac};
+use serde::{Deserialize, Serialize};
+use sha2::{Digest, Sha256};
+use std::time::{Duration, SystemTime, UNIX_EPOCH};
+
+type HmacSha256 = Hmac<Sha256>;
+
+pub const LEASE_SECRET_ENV: &str = "UNSLOTH_STUDIO_NATIVE_PATH_LEASE_SECRET";
+const LEASE_VERSION: u8 = 1;
+const LEASE_TTL: Duration = Duration::from_secs(2 * 60);
+
+#[derive(Clone, Copy, Debug, Deserialize, Eq, PartialEq, Serialize)]
+#[serde(rename_all = "kebab-case")]
+pub enum NativePathOperation {
+    ValidateModel,
+    LoadModel,
+    DatasetPreview,
+    DatasetImport,
+    Attach,
+    Reveal,
+    Open,
+}
+
+#[derive(Clone, Copy, Debug, Deserialize, Eq, PartialEq, Serialize)]
+#[serde(rename_all = "kebab-case")]
+pub enum NativePathKind {
+    Model,
+    Dataset,
+    Attachment,
+    Artifact,
+}
+
+#[derive(Clone, Copy, Debug, Deserialize, Eq, PartialEq, Serialize)]
+#[serde(rename_all = "kebab-case")]
+pub enum NativePathSourceKind {
+    Dialog,
+    Drop,
+    DeepLink,
+    FileAssociation,
+    Artifact,
+}
+
+#[derive(Clone, Copy, Debug, Eq, PartialEq, Serialize)]
+#[serde(rename_all = "kebab-case")]
+pub enum NativePathType {
+    File,
+    Directory,
+}
+
+#[derive(Clone, Debug, Serialize)]
+#[serde(rename_all = "snake_case")]
+pub struct NativePathLeasePayload {
+    pub version: u8,
+    pub operation: NativePathOperation,
+    pub canonical_path: String,
+    pub path_kind: NativePathKind,
+    pub path_type: NativePathType,
+    pub source_kind: NativePathSourceKind,
+    pub token_id_hash: String,
+    pub issued_at_ms: u64,
+    pub expires_at_ms: u64,
+    pub nonce: String,
+    pub display_label: String,
+    pub size_bytes: Option<u64>,
+    pub modified_ms: Option<u64>,
+}
+
+#[derive(Clone, Debug, Serialize)]
+#[serde(rename_all = "camelCase")]
+pub struct NativePathLeaseResponse {
+    pub native_path_lease: String,
+    pub display_label: String,
+    pub expires_at_ms: u64,
+}
+
+pub fn new_lease_secret() -> Vec<u8> {
+    rand::random::<[u8; 32]>().to_vec()
+}
+
+pub fn encode_secret_env(secret: &[u8]) -> String {
+    URL_SAFE_NO_PAD.encode(secret)
+}
+
+pub fn now_ms() -> u64 {
+    SystemTime::now()
+        .duration_since(UNIX_EPOCH)
+        .unwrap_or_else(|_| Duration::from_secs(0))
+        .as_millis() as u64
+}
+
+pub fn token_hash(token: &str) -> String {
+    hex_bytes(&Sha256::digest(token.as_bytes()))
+}
+
+pub fn random_token(prefix: &str) -> String {
+    format!("{}{}", prefix, hex_bytes(&rand::random::<[u8; 24]>()))
+}
+
+pub fn random_nonce() -> String {
+    hex_bytes(&rand::random::<[u8; 16]>())
+}
+
+pub fn sign_path_lease(
+    secret: &[u8],
+    operation: NativePathOperation,
+    canonical_path: String,
+    path_kind: NativePathKind,
+    path_type: NativePathType,
+    source_kind: NativePathSourceKind,
+    token: &str,
+    display_label: String,
+    size_bytes: Option<u64>,
+    modified_ms: Option<u64>,
+) -> Result<NativePathLeaseResponse, String> {
+    let issued_at_ms = now_ms();
+    let expires_at_ms = issued_at_ms + LEASE_TTL.as_millis() as u64;
+    let payload = NativePathLeasePayload {
+        version: LEASE_VERSION,
+        operation,
+        canonical_path,
+        path_kind,
+        path_type,
+        source_kind,
+        token_id_hash: token_hash(token),
+        issued_at_ms,
+        expires_at_ms,
+        nonce: random_nonce(),
+        display_label: display_label.clone(),
+        size_bytes,
+        modified_ms,
+    };
+    sign_payload(secret, &payload).map(|native_path_lease| NativePathLeaseResponse {
+        native_path_lease,
+        display_label,
+        expires_at_ms,
+    })
+}
+
+fn sign_payload(secret: &[u8], payload: &NativePathLeasePayload) -> Result<String, String> {
+    let payload_json = serde_json::to_vec(payload).map_err(|e| e.to_string())?;
+    let payload_b64 = URL_SAFE_NO_PAD.encode(payload_json);
+    let signature = sign_bytes(secret, payload_b64.as_bytes())?;
+    Ok(format!(
+        "{}.{}",
+        payload_b64,
+        URL_SAFE_NO_PAD.encode(signature)
+    ))
+}
+
+fn sign_bytes(secret: &[u8], bytes: &[u8]) -> Result<Vec<u8>, String> {
+    let mut mac = HmacSha256::new_from_slice(secret).map_err(|e| e.to_string())?;
+    mac.update(bytes);
+    Ok(mac.finalize().into_bytes().to_vec())
+}
+
+pub fn hex_bytes(bytes: &[u8]) -> String {
+    const HEX: &[u8; 16] = b"0123456789abcdef";
+    let mut out = String::with_capacity(bytes.len() * 2);
+    for &byte in bytes {
+        out.push(HEX[(byte >> 4) as usize] as char);
+        out.push(HEX[(byte & 0x0f) as usize] as char);
+    }
+    out
+}
+
+#[cfg(test)]
+mod tests {
+    use super::*;
+
+    #[test]
+    fn token_hash_is_stable_hex_sha256() {
+        assert_eq!(
+            token_hash("native-token"),
+            "d0c16f641bc0a0ee6b63ff88cec29756638d19590893c340a1ae36c9fae7b07f"
+        );
+    }
+
+    #[test]
+    fn signed_lease_has_two_base64url_parts() {
+        let lease = sign_path_lease(
+            b"01234567890123456789012345678901",
+            NativePathOperation::ValidateModel,
+            "/tmp/model.gguf".to_string(),
+            NativePathKind::Model,
+            NativePathType::File,
+            NativePathSourceKind::Dialog,
+            "token",
+            "model.gguf".to_string(),
+            Some(123),
+            Some(456),
+        )
+        .unwrap();
+        let parts: Vec<&str> = lease.native_path_lease.split('.').collect();
+        assert_eq!(parts.len(), 2);
+        assert!(!parts[0].contains('='));
+        assert!(!parts[1].contains('='));
+    }
+}
diff --git a/studio/src-tauri/src/native_intents.rs b/studio/src-tauri/src/native_intents.rs
new file mode 100644
index 0000000000..30e11aa310
--- /dev/null
+++ b/studio/src-tauri/src/native_intents.rs
@@ -0,0 +1,476 @@
+use crate::native_backend_lease::{
+    encode_secret_env, now_ms, random_token, sign_path_lease, NativePathKind,
+    NativePathLeaseResponse, NativePathOperation, NativePathSourceKind, NativePathType,
+};
+use crate::native_path_policy::{
+    classify_artifact_path, classify_native_model_path, reveal_target, ClassifiedPath,
+    NativeArtifactKind,
+};
+use serde::Serialize;
+use std::collections::{HashMap, VecDeque};
+use std::path::{Path, PathBuf};
+use std::sync::Mutex;
+use std::time::Duration;
+use tauri::{AppHandle, WebviewWindow};
+use tauri_plugin_dialog::DialogExt;
+
+const TOKEN_TTL: Duration = Duration::from_secs(15 * 60);
+
+#[derive(Clone, Debug)]
+struct NativePathEntry {
+    token: String,
+    canonical_path: PathBuf,
+    validation_policy: NativePathValidationPolicy,
+    path_kind: NativePathKind,
+    path_type: NativePathType,
+    source_kind: NativePathSourceKind,
+    allowed_operations: Vec<NativePathOperation>,
+    display_label: String,
+    expires_at_ms: u64,
+    size_bytes: Option<u64>,
+    modified_ms: Option<u64>,
+}
+
+#[derive(Clone, Copy, Debug)]
+enum NativePathValidationPolicy {
+    Model,
+    Artifact(NativeArtifactKind),
+}
+
+#[derive(Clone, Debug, Serialize)]
+#[serde(rename_all = "camelCase")]
+pub struct NativePathRef {
+    token: String,
+    kind: NativePathKind,
+    display_label: String,
+    allowed_operations: Vec<NativePathOperation>,
+    expires_at_ms: u64,
+}
+
+#[derive(Clone, Debug, Serialize)]
+#[serde(rename_all = "camelCase")]
+pub struct NativeIntent {
+    id: String,
+    kind: NativePathKind,
+    source_kind: NativePathSourceKind,
+    path: NativePathRef,
+    display_label: String,
+}
+
+#[derive(Default)]
+struct NativeIntakeInner {
+    tokens: HashMap<String, NativePathEntry>,
+    queued_intents: VecDeque<NativeIntent>,
+}
+
+pub struct NativeIntakeState {
+    inner: Mutex<NativeIntakeInner>,
+    lease_secret: Vec<u8>,
+}
+
+pub fn new_native_intake_state() -> NativeIntakeState {
+    NativeIntakeState {
+        inner: Mutex::new(NativeIntakeInner::default()),
+        lease_secret: crate::native_backend_lease::new_lease_secret(),
+    }
+}
+
+impl NativeIntakeState {
+    pub fn lease_secret_env(&self) -> String {
+        encode_secret_env(&self.lease_secret)
+    }
+
+    #[allow(dead_code)]
+    pub fn enqueue_model_path(
+        &self,
+        path: impl AsRef<Path>,
+        source_kind: NativePathSourceKind,
+    ) -> Result<NativeIntent, String> {
+        let intent = self.register_model_path(path, source_kind)?;
+        let mut inner = self.inner.lock().map_err(|e| e.to_string())?;
+        inner.queued_intents.push_back(intent.clone());
+        Ok(intent)
+    }
+
+    fn register_model_path(
+        &self,
+        path: impl AsRef<Path>,
+        source_kind: NativePathSourceKind,
+    ) -> Result<NativeIntent, String> {
+        let classified = classify_native_model_path(path.as_ref())?;
+        self.register_classified_path(classified, source_kind, NativePathValidationPolicy::Model)
+    }
+
+    fn register_artifact(
+        &self,
+        kind: NativeArtifactKind,
+        path: impl AsRef<Path>,
+    ) -> Result<NativePathRef, String> {
+        let classified = classify_artifact_path(kind, path.as_ref())?;
+        let entry = self.insert_entry(
+            classified,
+            NativePathSourceKind::Artifact,
+            NativePathValidationPolicy::Artifact(kind),
+        )?;
+        Ok(entry.to_ref())
+    }
+
+    fn register_classified_path(
+        &self,
+        classified: ClassifiedPath,
+        source_kind: NativePathSourceKind,
+        validation_policy: NativePathValidationPolicy,
+    ) -> Result<NativeIntent, String> {
+        let entry = self.insert_entry(classified, source_kind, validation_policy)?;
+        Ok(NativeIntent {
+            id: random_token("intent_"),
+            kind: entry.path_kind,
+            source_kind,
+            path: entry.to_ref(),
+            display_label: entry.display_label.clone(),
+        })
+    }
+
+    fn insert_entry(
+        &self,
+        classified: ClassifiedPath,
+        source_kind: NativePathSourceKind,
+        validation_policy: NativePathValidationPolicy,
+    ) -> Result<NativePathEntry, String> {
+        let token = random_token("path_");
+        let expires_at_ms = now_ms() + TOKEN_TTL.as_millis() as u64;
+        let entry = NativePathEntry {
+            token: token.clone(),
+            canonical_path: classified.canonical_path,
+            validation_policy,
+            path_kind: classified.path_kind,
+            path_type: classified.path_type,
+            source_kind,
+            allowed_operations: classified.allowed_operations,
+            display_label: classified.display_label,
+            expires_at_ms,
+            size_bytes: classified.size_bytes,
+            modified_ms: classified.modified_ms,
+        };
+        let mut inner = self.inner.lock().map_err(|e| e.to_string())?;
+        inner.tokens.insert(token, entry.clone());
+        Ok(entry)
+    }
+
+    fn drain_intents(&self) -> Result<Vec<NativeIntent>, String> {
+        let mut inner = self.inner.lock().map_err(|e| e.to_string())?;
+        prune_expired(&mut inner);
+        Ok(inner.queued_intents.drain(..).collect())
+    }
+
+    fn entry_for_operation(
+        &self,
+        token: &str,
+        operation: NativePathOperation,
+    ) -> Result<NativePathEntry, String> {
+        let mut inner = self.inner.lock().map_err(|e| e.to_string())?;
+        prune_expired(&mut inner);
+        let entry = inner
+            .tokens
+            .get(token)
+            .ok_or_else(|| "Native path token is unavailable or expired.".to_string())?;
+        if !entry.allowed_operations.contains(&operation) {
+            return Err("Native path token does not allow this operation.".to_string());
+        }
+        Ok(entry.clone())
+    }
+
+    fn sign_grant(
+        &self,
+        token: &str,
+        operation: NativePathOperation,
+    ) -> Result<NativePathLeaseResponse, String> {
+        let entry = self.entry_for_operation(token, operation)?;
+        validate_entry_path(&entry, operation)?;
+        sign_path_lease(
+            &self.lease_secret,
+            operation,
+            entry.canonical_path.to_string_lossy().to_string(),
+            entry.path_kind,
+            entry.path_type,
+            entry.source_kind,
+            &entry.token,
+            entry.display_label,
+            entry.size_bytes,
+            entry.modified_ms,
+        )
+    }
+
+    fn path_for_operation(
+        &self,
+        token: &str,
+        operation: NativePathOperation,
+    ) -> Result<NativePathEntry, String> {
+        let entry = self.entry_for_operation(token, operation)?;
+        validate_entry_path(&entry, operation)?;
+        Ok(entry)
+    }
+}
+
+fn validate_entry_path(
+    entry: &NativePathEntry,
+    operation: NativePathOperation,
+) -> Result<(), String> {
+    let classified = match entry.validation_policy {
+        NativePathValidationPolicy::Model => classify_native_model_path(&entry.canonical_path)?,
+        NativePathValidationPolicy::Artifact(kind) => {
+            classify_artifact_path(kind, &entry.canonical_path)?
+        }
+    };
+    let check_fingerprint = !matches!(
+        operation,
+        NativePathOperation::Reveal | NativePathOperation::Open
+    );
+    if classified.canonical_path != entry.canonical_path
+        || classified.path_kind != entry.path_kind
+        || classified.path_type != entry.path_type
+        || !classified.allowed_operations.contains(&operation)
+        || (check_fingerprint && classified.size_bytes != entry.size_bytes)
+        || (check_fingerprint && classified.modified_ms != entry.modified_ms)
+    {
+        return Err("Native path changed after it was selected.".to_string());
+    }
+    Ok(())
+}
+
+impl NativePathEntry {
+    fn to_ref(&self) -> NativePathRef {
+        NativePathRef {
+            token: self.token.clone(),
+            kind: self.path_kind,
+            display_label: self.display_label.clone(),
+            allowed_operations: self.allowed_operations.clone(),
+            expires_at_ms: self.expires_at_ms,
+        }
+    }
+}
+
+fn prune_expired(inner: &mut NativeIntakeInner) {
+    let now = now_ms();
+    inner.tokens.retain(|_, entry| entry.expires_at_ms > now);
+    inner
+        .queued_intents
+        .retain(|intent| intent.path.expires_at_ms > now);
+}
+
+fn ensure_main_window(window: &WebviewWindow) -> Result<(), String> {
+    if window.label() == "main" {
+        Ok(())
+    } else {
+        Err("Native path commands are only available to the main window.".to_string())
+    }
+}
+
+#[tauri::command]
+pub fn drain_native_intents(
+    window: WebviewWindow,
+    state: tauri::State<'_, NativeIntakeState>,
+) -> Result<Vec<NativeIntent>, String> {
+    ensure_main_window(&window)?;
+    state.drain_intents()
+}
+
+#[tauri::command]
+pub fn register_native_model_path(
+    window: WebviewWindow,
+    state: tauri::State<'_, NativeIntakeState>,
+    path: String,
+) -> Result<NativeIntent, String> {
+    ensure_main_window(&window)?;
+    state.register_model_path(path, NativePathSourceKind::Drop)
+}
+
+#[tauri::command]
+pub async fn pick_native_model(
+    window: WebviewWindow,
+    app: AppHandle,
+    state: tauri::State<'_, NativeIntakeState>,
+) -> Result<Option<NativeIntent>, String> {
+    ensure_main_window(&window)?;
+    let (tx, rx) = tokio::sync::oneshot::channel();
+    app.dialog()
+        .file()
+        .set_title("Choose a GGUF model")
+        .add_filter("GGUF models", &["gguf"])
+        .pick_file(move |path| {
+            let _ = tx.send(path);
+        });
+    let Some(file_path) = rx.await.map_err(|_| "Dialog closed".to_string())? else {
+        return Ok(None);
+    };
+    let path = file_path
+        .into_path()
+        .map_err(|_| "Only local filesystem model paths are supported.".to_string())?;
+    state
+        .register_model_path(path, NativePathSourceKind::Dialog)
+        .map(Some)
+}
+
+#[tauri::command]
+pub fn consume_native_path_token(
+    window: WebviewWindow,
+    state: tauri::State<'_, NativeIntakeState>,
+    token: String,
+    operation: NativePathOperation,
+) -> Result<NativePathLeaseResponse, String> {
+    ensure_main_window(&window)?;
+    match operation {
+        NativePathOperation::Reveal | NativePathOperation::Open => {
+            Err("Reveal/Open do not use backend path grants.".to_string())
+        }
+        _ => state.sign_grant(&token, operation),
+    }
+}
+
+#[tauri::command]
+pub fn register_artifact_path(
+    window: WebviewWindow,
+    state: tauri::State<'_, NativeIntakeState>,
+    kind: NativeArtifactKind,
+    path: String,
+) -> Result<NativePathRef, String> {
+    ensure_main_window(&window)?;
+    state.register_artifact(kind, path)
+}
+
+#[tauri::command]
+pub fn reveal_path_token(
+    window: WebviewWindow,
+    state: tauri::State<'_, NativeIntakeState>,
+    token: String,
+) -> Result<(), String> {
+    ensure_main_window(&window)?;
+    let entry = state.path_for_operation(&token, NativePathOperation::Reveal)?;
+    #[cfg(target_os = "macos")]
+    {
+        if entry.canonical_path.is_file() {
+            return std::process::Command::new("open")
+                .arg("-R")
+                .arg(&entry.canonical_path)
+                .spawn()
+                .map(|_| ())
+                .map_err(|e| format!("Failed to reveal path: {e}"));
+        }
+    }
+    #[cfg(target_os = "windows")]
+    {
+        if entry.canonical_path.is_file() {
+            let mut select_arg = std::ffi::OsString::from("/select,");
+            select_arg.push(entry.canonical_path.as_os_str());
+            return std::process::Command::new("explorer")
+                .arg(select_arg)
+                .spawn()
+                .map(|_| ())
+                .map_err(|e| format!("Failed to reveal path: {e}"));
+        }
+    }
+    let target = reveal_target(&entry.canonical_path);
+    open::that_detached(target).map_err(|e| format!("Failed to reveal path: {e}"))
+}
+
+#[tauri::command]
+pub fn open_path_token(
+    window: WebviewWindow,
+    state: tauri::State<'_, NativeIntakeState>,
+    token: String,
+) -> Result<(), String> {
+    ensure_main_window(&window)?;
+    let entry = state.path_for_operation(&token, NativePathOperation::Open)?;
+    open::that_detached(entry.canonical_path).map_err(|e| format!("Failed to open path: {e}"))
+}
+
+#[cfg(test)]
+mod tests {
+    use super::*;
+    use std::fs;
+    use std::time::{SystemTime, UNIX_EPOCH};
+
+    fn temp_path(name: &str) -> PathBuf {
+        let nanos = SystemTime::now()
+            .duration_since(UNIX_EPOCH)
+            .unwrap()
+            .as_nanos();
+        std::env::temp_dir().join(format!(
+            "unsloth-native-intents-{name}-{}-{nanos}",
+            std::process::id()
+        ))
+    }
+
+    #[test]
+    fn model_token_issues_distinct_validate_and_load_grants() {
+        let state = new_native_intake_state();
+        let path = temp_path("model").with_extension("gguf");
+        fs::write(&path, b"gguf").unwrap();
+        let intent = state
+            .register_model_path(&path, NativePathSourceKind::Dialog)
+            .unwrap();
+        let validate = state
+            .sign_grant(&intent.path.token, NativePathOperation::ValidateModel)
+            .unwrap();
+        let load = state
+            .sign_grant(&intent.path.token, NativePathOperation::LoadModel)
+            .unwrap();
+        assert_ne!(validate.native_path_lease, load.native_path_lease);
+        assert!(validate.display_label.ends_with(".gguf"));
+        let _ = fs::remove_file(path);
+    }
+
+    #[test]
+    fn model_token_rejects_dataset_operation() {
+        let state = new_native_intake_state();
+        let path = temp_path("model").with_extension("gguf");
+        fs::write(&path, b"gguf").unwrap();
+        let intent = state
+            .register_model_path(&path, NativePathSourceKind::Dialog)
+            .unwrap();
+        let err = state
+            .sign_grant(&intent.path.token, NativePathOperation::DatasetImport)
+            .unwrap_err();
+        assert!(err.contains("does not allow"));
+        let _ = fs::remove_file(path);
+    }
+
+    #[test]
+    fn model_token_revalidates_path_changes() {
+        let state = new_native_intake_state();
+        let path = temp_path("model").with_extension("gguf");
+        fs::write(&path, b"gguf").unwrap();
+        let intent = state
+            .register_model_path(&path, NativePathSourceKind::Dialog)
+            .unwrap();
+        fs::write(&path, b"changed").unwrap();
+        let err = state
+            .sign_grant(&intent.path.token, NativePathOperation::ValidateModel)
+            .unwrap_err();
+        assert!(err.contains("changed"));
+        let _ = fs::remove_file(path);
+    }
+
+    #[cfg(unix)]
+    #[test]
+    fn reveal_rejects_symlink_replacement() {
+        use std::os::unix::fs::symlink;
+
+        let state = new_native_intake_state();
+        let path = temp_path("model").with_extension("gguf");
+        let target = temp_path("replacement").with_extension("gguf");
+        fs::write(&path, b"gguf").unwrap();
+        fs::write(&target, b"gguf").unwrap();
+        let intent = state
+            .register_model_path(&path, NativePathSourceKind::Dialog)
+            .unwrap();
+        fs::remove_file(&path).unwrap();
+        symlink(&target, &path).unwrap();
+        let err = state
+            .path_for_operation(&intent.path.token, NativePathOperation::Reveal)
+            .unwrap_err();
+        assert!(err.contains("Symlink") || err.contains("changed"));
+        let _ = fs::remove_file(path);
+        let _ = fs::remove_file(target);
+    }
+}
diff --git a/studio/src-tauri/src/native_path_policy.rs b/studio/src-tauri/src/native_path_policy.rs
new file mode 100644
index 0000000000..b2ebb34621
--- /dev/null
+++ b/studio/src-tauri/src/native_path_policy.rs
@@ -0,0 +1,291 @@
+use crate::native_backend_lease::{NativePathKind, NativePathOperation, NativePathType};
+use serde::{Deserialize, Serialize};
+use std::fs;
+use std::path::{Path, PathBuf};
+use std::time::UNIX_EPOCH;
+
+#[derive(Clone, Copy, Debug, Deserialize, Eq, PartialEq, Serialize)]
+#[serde(rename_all = "kebab-case")]
+pub enum NativeArtifactKind {
+    TrainingOutput,
+    Export,
+    DatasetUpload,
+    RecipeArtifact,
+    DiagnosticLog,
+}
+
+#[derive(Clone, Debug)]
+pub struct ClassifiedPath {
+    pub canonical_path: PathBuf,
+    pub path_kind: NativePathKind,
+    pub path_type: NativePathType,
+    pub allowed_operations: Vec<NativePathOperation>,
+    pub display_label: String,
+    pub size_bytes: Option<u64>,
+    pub modified_ms: Option<u64>,
+}
+
+pub fn classify_native_model_path(path: &Path) -> Result<ClassifiedPath, String> {
+    let classified = classify_existing_path(path)?;
+    if classified.path_type != NativePathType::File {
+        return Err("Only GGUF model files are supported for native model intake.".to_string());
+    }
+    if !has_extension(&classified.canonical_path, "gguf") {
+        return Err("Only .gguf model files are supported for native model intake.".to_string());
+    }
+    Ok(ClassifiedPath {
+        path_kind: NativePathKind::Model,
+        allowed_operations: vec![
+            NativePathOperation::ValidateModel,
+            NativePathOperation::LoadModel,
+            NativePathOperation::Reveal,
+        ],
+        ..classified
+    })
+}
+
+pub fn classify_artifact_path(
+    kind: NativeArtifactKind,
+    path: &Path,
+) -> Result<ClassifiedPath, String> {
+    let classified = classify_existing_path(path)?;
+    ensure_artifact_root(kind, &classified.canonical_path)?;
+    reject_sensitive_artifact(&classified.canonical_path)?;
+
+    let mut allowed_operations = vec![NativePathOperation::Reveal];
+    if is_open_safe_artifact(&classified.canonical_path, classified.path_type) {
+        allowed_operations.push(NativePathOperation::Open);
+    }
+
+    Ok(ClassifiedPath {
+        path_kind: NativePathKind::Artifact,
+        allowed_operations,
+        ..classified
+    })
+}
+
+pub fn refresh_path_fingerprint(
+    path: &Path,
+) -> Result<(NativePathType, Option<u64>, Option<u64>), String> {
+    let metadata = fs::metadata(path).map_err(|e| format!("Path is no longer available: {e}"))?;
+    let path_type = if metadata.is_file() {
+        NativePathType::File
+    } else if metadata.is_dir() {
+        NativePathType::Directory
+    } else {
+        return Err("Special files are not supported.".to_string());
+    };
+    let size_bytes = metadata.is_file().then_some(metadata.len());
+    let modified_ms = metadata
+        .modified()
+        .ok()
+        .and_then(|modified| modified.duration_since(UNIX_EPOCH).ok())
+        .map(|duration| duration.as_millis() as u64);
+    Ok((path_type, size_bytes, modified_ms))
+}
+
+pub fn reveal_target(path: &Path) -> PathBuf {
+    if path.is_dir() {
+        path.to_path_buf()
+    } else {
+        path.parent().unwrap_or(path).to_path_buf()
+    }
+}
+
+fn sanitize_display_label(raw: &str) -> String {
+    let cleaned: String = raw
+        .chars()
+        .map(|ch| if ch.is_control() { ' ' } else { ch })
+        .collect();
+    let trimmed = cleaned.trim();
+    if trimmed.is_empty() {
+        "Selected path".to_string()
+    } else {
+        trimmed.chars().take(160).collect()
+    }
+}
+
+pub fn is_open_safe_artifact(path: &Path, path_type: NativePathType) -> bool {
+    if path_type == NativePathType::Directory {
+        return false;
+    }
+    let Some(ext) = path.extension().and_then(|ext| ext.to_str()) else {
+        return false;
+    };
+    matches!(
+        ext.to_ascii_lowercase().as_str(),
+        "txt" | "log" | "json" | "jsonl" | "csv" | "tsv" | "parquet" | "md"
+    )
+}
+
+fn classify_existing_path(path: &Path) -> Result<ClassifiedPath, String> {
+    reject_network_or_device_path(path)?;
+    let symlink_metadata =
+        fs::symlink_metadata(path).map_err(|e| format!("Path is not available: {e}"))?;
+    if symlink_metadata.file_type().is_symlink() {
+        return Err("Symlink paths are not supported for native intake.".to_string());
+    }
+
+    let canonical_path = path
+        .canonicalize()
+        .map_err(|e| format!("Path could not be resolved: {e}"))?;
+    reject_network_or_device_path(&canonical_path)?;
+    let canonical_symlink_metadata = fs::symlink_metadata(&canonical_path)
+        .map_err(|e| format!("Path is not available: {e}"))?;
+    if canonical_symlink_metadata.file_type().is_symlink() {
+        return Err("Symlink paths are not supported for native intake.".to_string());
+    }
+    let (path_type, size_bytes, modified_ms) = refresh_path_fingerprint(&canonical_path)?;
+    let display_label = sanitize_display_label(
+        canonical_path
+            .file_name()
+            .and_then(|name| name.to_str())
+            .unwrap_or("Selected path"),
+    );
+
+    Ok(ClassifiedPath {
+        canonical_path,
+        path_kind: NativePathKind::Artifact,
+        path_type,
+        allowed_operations: vec![NativePathOperation::Reveal],
+        display_label,
+        size_bytes,
+        modified_ms,
+    })
+}
+
+fn ensure_artifact_root(kind: NativeArtifactKind, canonical_path: &Path) -> Result<(), String> {
+    let Some(home) = dirs::home_dir() else {
+        return Err("Could not determine home directory.".to_string());
+    };
+    let studio = home.join(".unsloth").join("studio");
+    let allowed_root = match kind {
+        NativeArtifactKind::TrainingOutput => studio.join("outputs"),
+        NativeArtifactKind::Export => studio.join("exports"),
+        NativeArtifactKind::DatasetUpload => studio.join("assets").join("datasets").join("uploads"),
+        NativeArtifactKind::RecipeArtifact => {
+            studio.join("assets").join("datasets").join("recipes")
+        }
+        NativeArtifactKind::DiagnosticLog => studio.join("logs"),
+    };
+    let root = allowed_root
+        .canonicalize()
+        .map_err(|_| "Artifact root is not available.".to_string())?;
+    if canonical_path == root || canonical_path.starts_with(&root) {
+        Ok(())
+    } else {
+        Err("Artifact path is outside the allowed artifact root.".to_string())
+    }
+}
+
+fn reject_sensitive_artifact(path: &Path) -> Result<(), String> {
+    let lowered = path.to_string_lossy().to_ascii_lowercase();
+    for needle in [
+        "/auth/",
+        "\\auth\\",
+        "/auth.db",
+        "\\auth.db",
+        "/studio.db",
+        "\\studio.db",
+        "/pid",
+        "\\pid",
+    ] {
+        if lowered.contains(needle) {
+            return Err("Sensitive Studio state cannot be registered as an artifact.".to_string());
+        }
+    }
+    if let Some(ext) = path.extension().and_then(|ext| ext.to_str()) {
+        if matches!(
+            ext.to_ascii_lowercase().as_str(),
+            "exe" | "dll" | "dylib" | "so" | "sh" | "bash" | "zsh" | "ps1" | "bat" | "cmd"
+        ) {
+            return Err(
+                "Executable artifacts cannot be registered for native open/reveal.".to_string(),
+            );
+        }
+    }
+    Ok(())
+}
+
+fn reject_network_or_device_path(path: &Path) -> Result<(), String> {
+    let text = path.to_string_lossy();
+    #[cfg(windows)]
+    {
+        let normalized = text.replace('/', "\\").to_ascii_lowercase();
+        if let Some(rest) = normalized.strip_prefix("\\\\?\\") {
+            let bytes = rest.as_bytes();
+            if !(bytes.len() >= 3
+                && bytes[0].is_ascii_alphabetic()
+                && bytes[1] == b':'
+                && bytes[2] == b'\\')
+            {
+                return Err("Network paths are not supported for native intake.".to_string());
+            }
+        } else if normalized.starts_with("\\\\") {
+            return Err("Network paths are not supported for native intake.".to_string());
+        }
+    }
+    #[cfg(unix)]
+    {
+        for root in ["/dev", "/proc", "/sys"] {
+            if path.starts_with(root) {
+                return Err("Device and virtual filesystem paths are not supported.".to_string());
+            }
+        }
+    }
+    if text.contains('\0') {
+        return Err("Path contains invalid NUL characters.".to_string());
+    }
+    Ok(())
+}
+
+fn has_extension(path: &Path, expected: &str) -> bool {
+    path.extension()
+        .and_then(|ext| ext.to_str())
+        .map(|ext| ext.eq_ignore_ascii_case(expected))
+        .unwrap_or(false)
+}
+
+#[cfg(test)]
+mod tests {
+    use super::*;
+    use std::fs;
+    use std::time::{SystemTime, UNIX_EPOCH};
+
+    fn temp_path(name: &str) -> PathBuf {
+        let nanos = SystemTime::now()
+            .duration_since(UNIX_EPOCH)
+            .unwrap()
+            .as_nanos();
+        std::env::temp_dir().join(format!(
+            "unsloth-native-policy-{name}-{}-{nanos}",
+            std::process::id()
+        ))
+    }
+
+    #[test]
+    fn gguf_model_allows_validate_load_reveal() {
+        let path = temp_path("model").with_extension("gguf");
+        fs::write(&path, b"gguf").unwrap();
+        let classified = classify_native_model_path(&path).unwrap();
+        assert_eq!(classified.path_kind, NativePathKind::Model);
+        assert!(classified
+            .allowed_operations
+            .contains(&NativePathOperation::ValidateModel));
+        assert!(classified
+            .allowed_operations
+            .contains(&NativePathOperation::LoadModel));
+        assert!(classified
+            .allowed_operations
+            .contains(&NativePathOperation::Reveal));
+        let _ = fs::remove_file(path);
+    }
+
+    #[test]
+    fn non_gguf_model_is_rejected() {
+        let path = temp_path("model").with_extension("txt");
+        fs::write(&path, b"not gguf").unwrap();
+        assert!(classify_native_model_path(&path).is_err());
+        let _ = fs::remove_file(path);
+    }
+}
diff --git a/studio/src-tauri/src/process.rs b/studio/src-tauri/src/process.rs
index ed2659c0d9..9e0cf1c17f 100644
--- a/studio/src-tauri/src/process.rs
+++ b/studio/src-tauri/src/process.rs
@@ -7,7 +7,7 @@ use std::io::BufRead;
 use std::process::{Command, Stdio};
 use std::sync::atomic::{AtomicBool, Ordering};
 use std::sync::{Arc, Mutex};
-use tauri::{AppHandle, Emitter};
+use tauri::{AppHandle, Emitter, Manager};
 
 const MAX_LOG_LINES: usize = 1000;
 
@@ -298,6 +298,13 @@ pub fn start_backend(
         .stdout(Stdio::piped())
         .stderr(Stdio::piped());
 
+    if let Some(native_state) = app.try_state::<crate::native_intents::NativeIntakeState>() {
+        cmd.env(
+            crate::native_backend_lease::LEASE_SECRET_ENV,
+            native_state.lease_secret_env(),
+        );
+    }
+
     // AppImage sets LD_LIBRARY_PATH to its bundled libs, which breaks the spawned
     // Python process (wrong libpython/libz → "No module named encodings").
     // Only clear when running inside an AppImage — native .deb/.rpm installs may

From dbea77e347131a0f7fc2ede38a1674f997371347 Mon Sep 17 00:00:00 2001
From: Roland Tannous <115670425+rolandtannous@users.noreply.github.com>
Date: Mon, 4 May 2026 17:08:04 +0400
Subject: [PATCH 199/216] Studio: forward llama-server args from `unsloth
 studio run` , activate `unsloth run` , and allow passing model:quant to load
 models (#5271)

* Studio: forward unknown CLI args directly to llama-server

`unsloth studio run --model X --top-k 20 --chat-template-file foo.jinja`
now passes the unknown flags through to the llama-server subprocess.
Adds a denylist for flags Studio manages (port, -m, -c, --api-key, -ngl,
--flash-attn, --no-context-shift, --jinja, GPU-fit, model-identity, ...)
that returns HTTP 400 on collision. HTTP callers can supply the same
list via LoadRequest.llama_extra_args.

* Studio: accept `--model org/repo:variant` shorthand in `unsloth studio run`

Mirrors llama.cpp's `-hf <repo>:<quant>` and ollama's pull syntax so
`unsloth studio run --model unsloth/gpt-oss-20b-GGUF:UD-Q4_K_XL` is
equivalent to `--model unsloth/... --gguf-variant UD-Q4_K_XL`. Local
paths and Windows drive letters are preserved verbatim. If both an
embedded variant and an explicit `--gguf-variant` are given and they
disagree, the command fails with a clear error.

* Studio: register `unsloth run` as alias for `unsloth studio run`

Top-level `unsloth run --model ...` is now equivalent to
`unsloth studio run --model ...`. Same context_settings, so unknown
flags continue to pass through to llama-server.

* Studio: let users override soft-managed llama-server flags from CLI

Trims the denylist to flags Studio fundamentally cannot share with
the user (model identity, --host/--port/--path/--api-prefix,
--api-key, --ssl-*, --webui, --models-*). Soft-managed flags --
-c/--ctx-size, --parallel, --flash-attn, --no-context-shift,
--jinja, -ngl, -t/--threads, --fit* -- now pass through and override
Studio's auto-set version via llama.cpp's last-wins CLI parsing.

Lets users tune their run on the spot:
  unsloth run --model X -c 131072 --parallel 1 --threads 32

* Studio: accept `-hf` / `-hfr` / `--hf-repo` as aliases for `--model`

Matches llama-server's `-hf <repo>:<quant>` spelling so users coming
from llama.cpp can use the same flag. Typer claims the aliases before
the pass-through validator runs, so the HTTP-API denylist on those
flags is unaffected.

  unsloth run -hf unsloth/gpt-oss-20b-GGUF:UD-Q4_K_XL
---
 studio/backend/core/inference/llama_cpp.py    |  14 +-
 .../core/inference/llama_server_args.py       | 120 +++++++++++
 studio/backend/models/inference.py            |  10 +
 studio/backend/routes/inference.py            |  11 +
 .../backend/tests/test_llama_server_args.py   | 189 ++++++++++++++++++
 tests/studio/test_cli_repo_variant.py         | 145 ++++++++++++++
 tests/studio/test_cli_run_alias.py            |  69 +++++++
 unsloth_cli/__init__.py                       |  14 +-
 unsloth_cli/commands/studio.py                |  89 ++++++++-
 9 files changed, 656 insertions(+), 5 deletions(-)
 create mode 100644 studio/backend/core/inference/llama_server_args.py
 create mode 100644 studio/backend/tests/test_llama_server_args.py
 create mode 100644 tests/studio/test_cli_repo_variant.py
 create mode 100644 tests/studio/test_cli_run_alias.py

diff --git a/studio/backend/core/inference/llama_cpp.py b/studio/backend/core/inference/llama_cpp.py
index 6790dbc802..bddd71c301 100644
--- a/studio/backend/core/inference/llama_cpp.py
+++ b/studio/backend/core/inference/llama_cpp.py
@@ -22,7 +22,7 @@ import sys
 import threading
 import time
 from pathlib import Path
-from typing import Generator, Optional
+from typing import Generator, List, Optional
 from urllib.parse import urlparse
 
 import httpx
@@ -1381,6 +1381,7 @@ class LlamaCppBackend:
         n_threads: Optional[int] = None,
         n_gpu_layers: Optional[int] = None,  # Accepted for caller compat, unused
         n_parallel: int = 1,
+        extra_args: Optional[List[str]] = None,
     ) -> bool:
         """
         Start llama-server with a GGUF model.
@@ -1767,6 +1768,17 @@ class LlamaCppBackend:
             else:
                 self._api_key = None
 
+            # User-supplied pass-through args go last so llama.cpp's
+            # last-wins flag parsing lets the user override Studio's
+            # auto-set tier-2 flags (e.g. --cache-type-k, --spec-type).
+            # The route layer has already validated this list against
+            # the managed-flag denylist via validate_extra_args().
+            if extra_args:
+                cmd.extend(str(a) for a in extra_args)
+                logger.info(
+                    f"Appending user extra args to llama-server: {list(extra_args)}"
+                )
+
             _log_cmd = list(cmd)
             if "--api-key" in _log_cmd:
                 _ki = _log_cmd.index("--api-key") + 1
diff --git a/studio/backend/core/inference/llama_server_args.py b/studio/backend/core/inference/llama_server_args.py
new file mode 100644
index 0000000000..44c7d542c7
--- /dev/null
+++ b/studio/backend/core/inference/llama_server_args.py
@@ -0,0 +1,120 @@
+# SPDX-License-Identifier: AGPL-3.0-only
+# Copyright 2026-present the Unsloth AI Inc. team. All rights reserved. See /studio/LICENSE.AGPL-3.0
+
+"""Validator for user-supplied llama-server pass-through args.
+
+Studio runs llama-server as a managed subprocess and lets callers pass
+extra flags directly (CLI: ``unsloth run ... --top-k 20``; HTTP:
+``LoadRequest.llama_extra_args``). This module is the boundary that
+rejects only flags Studio fundamentally cannot share with the user --
+model identity, the auth key, and the network endpoint Studio's HTTP
+proxy targets. Anything else passes through.
+
+User-supplied args are appended to ``cmd`` after Studio's auto-set
+flags, so llama.cpp's last-wins CLI parsing makes the user's value
+override the auto-set one. That covers tunable knobs the user might
+reasonably want to override -- ``-c``/``--ctx-size``,
+``-np``/``--parallel``, ``-fa``/``--flash-attn``,
+``-ngl``/``--gpu-layers``, ``-t``/``--threads``, ``-fit``/``--fit*``,
+``--cache-type-k/v``, ``--chat-template-file/-kwargs``,
+``--spec-*``, ``--jinja``/``--no-jinja``,
+``--no-context-shift``/``--context-shift``, sampling params, etc.
+
+Reference: https://github.com/ggml-org/llama.cpp/blob/master/tools/server/README.md
+"""
+
+from __future__ import annotations
+
+from typing import Iterable, Optional
+
+# Each group is the full set of aliases (short + long) for one
+# hard-denied flag, taken from the llama-server README. If llama.cpp
+# adds a new alias for an existing denied flag, extend the relevant
+# group.
+#
+# Flags NOT in this list (e.g. -c, --parallel, --flash-attn, -ngl,
+# -t/--threads, --jinja, --no-context-shift, --fit*, --cache-type-*,
+# --chat-template-*, --spec-*) pass through and override Studio's
+# auto-set version via llama.cpp's last-wins CLI parsing.
+_DENYLIST_GROUPS: tuple[frozenset[str], ...] = (
+    # Model identity -- Studio resolves the model from LoadRequest and
+    # passes -m / mmproj after downloading from HF if needed. A second
+    # -m would point at a different model than the one Studio thinks
+    # is loaded.
+    frozenset({"-m", "--model"}),
+    frozenset({"-mu", "--model-url"}),
+    frozenset({"-dr", "--docker-repo"}),
+    frozenset({"-hf", "-hfr", "--hf-repo"}),
+    frozenset({"-hff", "--hf-file"}),
+    frozenset({"-hfv", "-hfrv", "--hf-repo-v"}),
+    frozenset({"-hffv", "--hf-file-v"}),
+    frozenset({"-hft", "--hf-token"}),
+    frozenset({"-mm", "--mmproj"}),
+    frozenset({"-mmu", "--mmproj-url"}),
+    # Networking -- Studio binds llama-server's port and reverse-proxies
+    # HTTP traffic to it. Retargeting host/port/path/prefix would
+    # orphan Studio's proxy and the UI would lose the server.
+    frozenset({"--host"}),
+    frozenset({"--port"}),
+    frozenset({"--path"}),
+    frozenset({"--api-prefix"}),
+    frozenset({"--reuse-port"}),
+    # Auth / TLS -- Studio terminates auth at its own layer; an
+    # upstream --api-key would shadow Studio's UNSLOTH_DIRECT_STREAM
+    # key, and TLS on llama-server would break the local proxy hop.
+    frozenset({"--api-key"}),
+    frozenset({"--api-key-file"}),
+    frozenset({"--ssl-key-file"}),
+    frozenset({"--ssl-cert-file"}),
+    # Single-model server -- Studio runs one model per llama-server
+    # process and serves its own UI. Enabling multi-model loading or
+    # llama-server's built-in web UI changes the surface clients see.
+    frozenset({"--webui", "--no-webui"}),
+    frozenset({"--models-dir"}),
+    frozenset({"--models-preset"}),
+    frozenset({"--models-max"}),
+    frozenset({"--models-autoload", "--no-models-autoload"}),
+)
+
+_DENYLIST: frozenset[str] = frozenset().union(*_DENYLIST_GROUPS)
+
+
+def _flag_name(token: str) -> Optional[str]:
+    """Return the flag name for a token, or None if it isn't a flag.
+
+    Peels ``--key=value`` to the bare ``--key``. Plain numeric values
+    like ``-1`` or ``-0.5`` (e.g. ``--seed -1``) are values, not flags;
+    llama-server short-form flags always start with a letter.
+    """
+    if not token.startswith("-") or token in {"-", "--"}:
+        return None
+    if len(token) >= 2 and (token[1].isdigit() or token[1] == "."):
+        return None
+    return token.split("=", 1)[0]
+
+
+def validate_extra_args(args: Optional[Iterable[str]]) -> list[str]:
+    """Validate user-supplied llama-server args.
+
+    Returns the args as a flat list ready to extend the llama-server
+    command. Raises ``ValueError`` (with the offending flag in the
+    message) the moment a token resolves to a Studio-managed flag.
+    """
+    if not args:
+        return []
+    out: list[str] = []
+    for raw in args:
+        token = str(raw)
+        flag = _flag_name(token)
+        if flag is not None and flag in _DENYLIST:
+            raise ValueError(
+                f"llama-server flag '{flag}' is managed by Unsloth Studio "
+                f"and cannot be passed as an extra arg"
+            )
+        out.append(token)
+    return out
+
+
+def is_managed_flag(flag: str) -> bool:
+    """True if ``flag`` is a Studio-managed llama-server flag."""
+    return flag in _DENYLIST
diff --git a/studio/backend/models/inference.py b/studio/backend/models/inference.py
index c0bb5b53a2..43087cc5bf 100644
--- a/studio/backend/models/inference.py
+++ b/studio/backend/models/inference.py
@@ -55,6 +55,16 @@ class LoadRequest(BaseModel):
         None,
         description = "Speculative decoding mode for GGUF models (e.g. 'ngram-simple', 'ngram-mod'). Ignored for non-GGUF and vision models.",
     )
+    llama_extra_args: Optional[List[str]] = Field(
+        None,
+        description = (
+            "Extra arguments forwarded verbatim to llama-server for GGUF models. "
+            "One token per list entry, e.g. ['--top-k', '20', '--seed', '42']. "
+            "Studio-managed flags (model identity, port, context length, GPU placement, "
+            "auth, --flash-attn, --no-context-shift, --jinja) are rejected. Ignored for "
+            "non-GGUF models."
+        ),
+    )
 
 
 class UnloadRequest(BaseModel):
diff --git a/studio/backend/routes/inference.py b/studio/backend/routes/inference.py
index 592da831e4..0690cc6b79 100644
--- a/studio/backend/routes/inference.py
+++ b/studio/backend/routes/inference.py
@@ -119,6 +119,7 @@ try:
         _DEFAULT_T_MAX_PREDICT_MS,
         detect_reasoning_flags,
     )
+    from core.inference.llama_server_args import validate_extra_args
     from utils.models import ModelConfig
     from utils.inference import load_inference_config
     from utils.models.model_config import load_model_defaults
@@ -140,6 +141,7 @@ except ImportError:
         _DEFAULT_T_MAX_PREDICT_MS,
         detect_reasoning_flags,
     )
+    from core.inference.llama_server_args import validate_extra_args
     from utils.models import ModelConfig
     from utils.inference import load_inference_config
     from utils.models.model_config import load_model_defaults
@@ -431,6 +433,13 @@ async def load_model(
     native_grant_backed = False
     model_log_label = request.model_path
     try:
+        # Validate user-supplied llama-server pass-through args up front
+        # so a managed-flag collision returns 400 before any model work.
+        try:
+            extra_llama_args = validate_extra_args(request.llama_extra_args)
+        except ValueError as exc:
+            raise HTTPException(status_code = 400, detail = str(exc))
+
         model_identifier, model_log_label, native_grant_backed = (
             _resolve_model_identifier_for_request(request, operation = "load-model")
         )
@@ -605,6 +614,7 @@ async def load_model(
                     cache_type_kv = request.cache_type_kv,
                     speculative_type = request.speculative_type,
                     n_parallel = _n_parallel,
+                    extra_args = extra_llama_args,
                 )
             else:
                 # Local mode: llama-server loads via -m <path>
@@ -623,6 +633,7 @@ async def load_model(
                     cache_type_kv = request.cache_type_kv,
                     speculative_type = request.speculative_type,
                     n_parallel = _n_parallel,
+                    extra_args = extra_llama_args,
                 )
 
             if not success:
diff --git a/studio/backend/tests/test_llama_server_args.py b/studio/backend/tests/test_llama_server_args.py
new file mode 100644
index 0000000000..351fbd014d
--- /dev/null
+++ b/studio/backend/tests/test_llama_server_args.py
@@ -0,0 +1,189 @@
+# SPDX-License-Identifier: AGPL-3.0-only
+# Copyright 2026-present the Unsloth AI Inc. team. All rights reserved. See /studio/LICENSE.AGPL-3.0
+
+"""Unit tests for the llama-server pass-through args validator.
+
+The validator is the security boundary between user-supplied CLI / HTTP
+input and the llama-server subprocess command. These tests pin the
+denylist behavior so the boundary doesn't quietly regress when new
+managed flags are added.
+"""
+
+from __future__ import annotations
+
+import pytest
+
+from core.inference.llama_server_args import (
+    is_managed_flag,
+    validate_extra_args,
+)
+
+
+# ── Pass-through (allowed) ───────────────────────────────────────────
+
+
+@pytest.mark.parametrize(
+    "args",
+    [
+        # Sampling
+        ["--top-k", "20"],
+        ["--top-p", "0.9", "--min-p", "0.05"],
+        ["--seed", "-1"],  # negative value, not a flag
+        ["--temp", "0.0"],
+        ["--repeat-penalty", "1.05"],
+        ["--mirostat", "2", "--mirostat-lr", "0.1"],
+        ["--xtc-probability", "0.05", "--xtc-threshold", "0.1"],
+        ["--dry-multiplier", "0.5"],
+        # Tier-2 knobs that map to LoadRequest fields
+        ["--cache-type-k", "q8_0"],
+        ["--cache-type-v", "q8_0"],
+        ["--chat-template-file", "/tmp/tpl.jinja"],
+        ["--chat-template-kwargs", '{"reasoning_effort":"high"}'],
+        ["--spec-type", "ngram-mod"],
+        ["--spec-default"],
+        # Reasoning controls
+        ["--reasoning-format", "deepseek"],
+        ["-rea", "auto"],
+        # Soft-managed flags the user may want to override on the CLI;
+        # llama.cpp's last-wins parsing means these win over Studio's
+        # auto-set version.
+        ["-c", "131072"],
+        ["--ctx-size", "8192"],
+        ["--parallel", "1"],
+        ["-np", "8"],
+        ["--flash-attn", "off"],
+        ["-fa", "on"],
+        ["--no-context-shift"],
+        ["--context-shift"],
+        ["--jinja"],
+        ["--no-jinja"],
+        ["-ngl", "-1"],
+        ["--gpu-layers", "32"],
+        ["-t", "16"],
+        ["--threads", "32"],
+        ["-fit", "off"],
+        ["--fit", "on"],
+        ["--fit-ctx", "8192"],
+    ],
+)
+def test_pass_through_allowed(args):
+    assert validate_extra_args(args) == args
+
+
+def test_none_returns_empty_list():
+    assert validate_extra_args(None) == []
+
+
+def test_empty_list_returns_empty_list():
+    assert validate_extra_args([]) == []
+
+
+def test_value_with_equals_form_passes_through():
+    assert validate_extra_args(["--top-k=20"]) == ["--top-k=20"]
+
+
+def test_non_flag_token_passes_through():
+    # A bare positional value (not preceded by a flag) is preserved
+    # verbatim. llama-server may reject it, but that's not our job.
+    assert validate_extra_args(["foo"]) == ["foo"]
+
+
+# ── Denylist (rejected) ──────────────────────────────────────────────
+
+
+@pytest.mark.parametrize(
+    "denied",
+    [
+        # Model identity
+        "-m",
+        "--model",
+        "-hf",
+        "-hfr",
+        "--hf-repo",
+        "-hff",
+        "--hf-file",
+        "-hft",
+        "--hf-token",
+        "-mm",
+        "--mmproj",
+        "--mmproj-url",
+        # Networking (Studio binds + proxies)
+        "--host",
+        "--port",
+        "--path",
+        "--api-prefix",
+        "--reuse-port",
+        # Auth / TLS
+        "--api-key",
+        "--api-key-file",
+        "--ssl-key-file",
+        "--ssl-cert-file",
+        # Single-model server
+        "--webui",
+        "--no-webui",
+        "--models-dir",
+        "--models-max",
+    ],
+)
+def test_denylist_rejects_all_aliases(denied):
+    with pytest.raises(ValueError, match = denied):
+        validate_extra_args([denied, "value"])
+
+
+def test_denylist_rejects_equals_form():
+    with pytest.raises(ValueError, match = "--port"):
+        validate_extra_args(["--port=9000"])
+
+
+def test_denylist_rejects_short_form_when_long_is_denied():
+    # -m is the short form of the hard-denied --model; rejecting only
+    # the long form would leave a trivial bypass.
+    with pytest.raises(ValueError, match = "-m"):
+        validate_extra_args(["-m", "/some/other/path.gguf"])
+
+
+def test_denylist_message_names_offending_flag():
+    with pytest.raises(ValueError) as excinfo:
+        validate_extra_args(["--top-k", "20", "--api-key", "secret"])
+    assert "--api-key" in str(excinfo.value)
+
+
+def test_first_denied_flag_short_circuits():
+    # Validation stops at the first denied flag; later denied flags
+    # in the same call don't matter for behaviour, but the message
+    # should name the first one we hit.
+    with pytest.raises(ValueError, match = "--port"):
+        validate_extra_args(["--port", "1", "--host", "x"])
+
+
+# ── Numeric values that look flag-ish ─────────────────────────────────
+
+
+@pytest.mark.parametrize("value", ["-1", "-0.5", "-42", "-.5"])
+def test_negative_number_value_is_not_flag(value):
+    # ``--seed -1`` is a value, not a flag. Validator must not try
+    # to look up "-1" in the denylist.
+    assert validate_extra_args(["--seed", value]) == ["--seed", value]
+
+
+# ── is_managed_flag helper ───────────────────────────────────────────
+
+
+def test_is_managed_flag_true_for_denied():
+    assert is_managed_flag("--port") is True
+    assert is_managed_flag("--api-key") is True
+    assert is_managed_flag("-m") is True
+    assert is_managed_flag("--model") is True
+
+
+def test_is_managed_flag_false_for_pass_through():
+    assert is_managed_flag("--top-k") is False
+    assert is_managed_flag("--cache-type-k") is False
+    assert is_managed_flag("--chat-template-file") is False
+    # Soft-managed flags pass through (last-wins override)
+    assert is_managed_flag("-c") is False
+    assert is_managed_flag("--ctx-size") is False
+    assert is_managed_flag("--parallel") is False
+    assert is_managed_flag("--flash-attn") is False
+    assert is_managed_flag("-ngl") is False
+    assert is_managed_flag("--threads") is False
diff --git a/tests/studio/test_cli_repo_variant.py b/tests/studio/test_cli_repo_variant.py
new file mode 100644
index 0000000000..1aa21cfd7d
--- /dev/null
+++ b/tests/studio/test_cli_repo_variant.py
@@ -0,0 +1,145 @@
+"""Tests for the ``repo:variant`` shorthand parser used by ``unsloth studio run``.
+
+Loads ``unsloth_cli/commands/studio.py`` directly via ``importlib`` with a
+minimal ``typer`` stub so the test doesn't drag in the rest of
+``unsloth_cli`` (which transitively imports the unsloth training stack).
+"""
+
+from __future__ import annotations
+
+import importlib.util
+import sys
+import types
+from pathlib import Path
+
+import pytest
+
+
+def _load_split_repo_variant():
+    """Load ``_split_repo_variant`` from studio.py with typer stubbed.
+
+    studio.py decorates Typer commands at import time, so a stub that
+    accepts (and discards) those calls is enough to let module
+    execution complete and expose the helper we want to test.
+    """
+    if "typer" not in sys.modules:
+        typer_stub = types.ModuleType("typer")
+
+        class _Typer:
+            def __init__(self, **kwargs):
+                pass
+
+            def callback(self, *args, **kwargs):
+                return lambda fn: fn
+
+            def command(self, *args, **kwargs):
+                return lambda fn: fn
+
+        typer_stub.Typer = _Typer
+        typer_stub.Option = lambda *args, **kwargs: (args[0] if args else None)
+        typer_stub.Context = type("Context", (), {})
+        typer_stub.Exit = type("Exit", (Exception,), {})
+        typer_stub.echo = lambda *args, **kwargs: None
+        sys.modules["typer"] = typer_stub
+
+    studio_py = (
+        Path(__file__).resolve().parents[2] / "unsloth_cli" / "commands" / "studio.py"
+    )
+    spec = importlib.util.spec_from_file_location(
+        "_studio_for_repo_variant_test", studio_py
+    )
+    module = importlib.util.module_from_spec(spec)
+    spec.loader.exec_module(module)
+    return module._split_repo_variant
+
+
+_split = _load_split_repo_variant()
+
+
+# ── HF-style repo:variant inputs -------------------------------------
+
+
+@pytest.mark.parametrize(
+    "model_arg, expected",
+    [
+        (
+            "unsloth/gpt-oss-20b-GGUF:UD-Q4_K_XL",
+            ("unsloth/gpt-oss-20b-GGUF", "UD-Q4_K_XL"),
+        ),
+        ("unsloth/gpt-oss-120b-GGUF:Q4_K_XL", ("unsloth/gpt-oss-120b-GGUF", "Q4_K_XL")),
+        ("unsloth/Qwen3-0.6B-GGUF:Q4_K_M", ("unsloth/Qwen3-0.6B-GGUF", "Q4_K_M")),
+        # Variants commonly contain dashes, dots, and underscores.
+        ("org/repo:UD-Q5_K_M", ("org/repo", "UD-Q5_K_M")),
+        ("org/repo:F16", ("org/repo", "F16")),
+    ],
+)
+def test_repo_variant_split(model_arg, expected):
+    assert _split(model_arg) == expected
+
+
+# ── No variant suffix ------------------------------------------------
+
+
+@pytest.mark.parametrize(
+    "model_arg",
+    [
+        "unsloth/gpt-oss-20b-GGUF",
+        "unsloth/Qwen3-0.6B-GGUF",
+        "shorthand-no-org-no-colon",
+    ],
+)
+def test_no_colon_returns_none_variant(model_arg):
+    repo, variant = _split(model_arg)
+    assert repo == model_arg
+    assert variant is None
+
+
+# ── Local paths must NOT be split ------------------------------------
+
+
+@pytest.mark.parametrize(
+    "local_path",
+    [
+        "/abs/path/to/model.gguf",
+        "/abs/path:with-colon-in-name",
+        "./relative/model",
+        "../parent/model",
+        "~/home/model",
+        ".",
+        "C:\\Users\\me\\model.gguf",
+        "C:/Users/me/model.gguf",
+        "D:/data/model:Q4",  # Windows drive + colon-suffixed filename: drive wins
+    ],
+)
+def test_local_path_passthrough(local_path):
+    repo, variant = _split(local_path)
+    assert repo == local_path
+    assert variant is None
+
+
+# ── Edge cases -------------------------------------------------------
+
+
+def test_empty_string():
+    assert _split("") == ("", None)
+
+
+def test_trailing_colon_no_variant():
+    # "org/repo:" -- no quant label after the colon. Pass through
+    # unchanged so the backend's existing validation surfaces a
+    # clearer error than "variant ''".
+    repo, variant = _split("org/repo:")
+    assert repo == "org/repo:"
+    assert variant is None
+
+
+def test_slash_in_variant_disqualifies_split():
+    # "foo:bar/baz" -- the suffix has a slash, so this isn't a quant
+    # label; treat the whole thing as opaque.
+    repo, variant = _split("foo:bar/baz")
+    assert repo == "foo:bar/baz"
+    assert variant is None
+
+
+def test_whitespace_stripped():
+    assert _split("  org/repo:Q4  ") == ("org/repo", "Q4")
diff --git a/tests/studio/test_cli_run_alias.py b/tests/studio/test_cli_run_alias.py
new file mode 100644
index 0000000000..aed6361ee2
--- /dev/null
+++ b/tests/studio/test_cli_run_alias.py
@@ -0,0 +1,69 @@
+"""Tests that ``unsloth run`` is registered as a top-level alias for
+``unsloth studio run``.
+
+AST-based to avoid importing ``unsloth_cli`` (which pulls in the heavy
+training stack) at test-collection time.
+"""
+
+from __future__ import annotations
+
+import ast
+from pathlib import Path
+
+_CLI_INIT = Path(__file__).resolve().parents[2] / "unsloth_cli" / "__init__.py"
+
+
+def _module_calls(source: str):
+    tree = ast.parse(source)
+    for node in ast.walk(tree):
+        if isinstance(node, ast.Call):
+            yield node
+
+
+def test_top_level_run_alias_registered():
+    """`app.command("run", ...)` must be invoked with studio_run as its target."""
+    source = _CLI_INIT.read_text()
+
+    # Find ``app.command("run", ...)`` call -- the decorator-call form.
+    found_decorator_call = False
+    for call in _module_calls(source):
+        # Match ``app.command(...)`` syntactically.
+        if not (
+            isinstance(call.func, ast.Attribute)
+            and call.func.attr == "command"
+            and isinstance(call.func.value, ast.Name)
+            and call.func.value.id == "app"
+        ):
+            continue
+        # Decorator-call form has a string literal "run" as the first
+        # positional or as keyword ``name="run"``.
+        first_pos = call.args[0] if call.args else None
+        keyword_name = next(
+            (kw.value for kw in call.keywords if kw.arg == "name"), None
+        )
+        is_run = (isinstance(first_pos, ast.Constant) and first_pos.value == "run") or (
+            isinstance(keyword_name, ast.Constant) and keyword_name.value == "run"
+        )
+        if is_run:
+            found_decorator_call = True
+            break
+    assert (
+        found_decorator_call
+    ), 'Expected `app.command("run", ...)` registration in unsloth_cli/__init__.py'
+
+
+def test_studio_run_imported_for_alias():
+    """The alias must wire up to the studio.run function, not redefine it."""
+    source = _CLI_INIT.read_text()
+    tree = ast.parse(source)
+    has_import = False
+    for node in ast.walk(tree):
+        if not isinstance(node, ast.ImportFrom):
+            continue
+        if node.module != "unsloth_cli.commands.studio":
+            continue
+        for alias in node.names:
+            if alias.name == "run":
+                has_import = True
+                break
+    assert has_import, "Expected `from unsloth_cli.commands.studio import run` in unsloth_cli/__init__.py"
diff --git a/unsloth_cli/__init__.py b/unsloth_cli/__init__.py
index 3a821359b7..28da00f37b 100644
--- a/unsloth_cli/__init__.py
+++ b/unsloth_cli/__init__.py
@@ -6,7 +6,7 @@ import typer
 from unsloth_cli.commands.train import train
 from unsloth_cli.commands.inference import inference
 from unsloth_cli.commands.export import export, list_checkpoints
-from unsloth_cli.commands.studio import studio_app
+from unsloth_cli.commands.studio import run as studio_run, studio_app
 
 app = typer.Typer(
     help = "Command-line interface for Unsloth training, inference, and export.",
@@ -18,3 +18,15 @@ app.command()(inference)
 app.command()(export)
 app.command("list-checkpoints")(list_checkpoints)
 app.add_typer(studio_app, name = "studio", help = "Unsloth Studio commands.")
+
+# Top-level alias: `unsloth run ...` is equivalent to `unsloth studio run ...`.
+# Same context_settings as the studio_app registration so unknown flags
+# still pass through to llama-server.
+app.command(
+    "run",
+    context_settings = {
+        "allow_extra_args": True,
+        "ignore_unknown_options": True,
+    },
+    help = "Alias for `unsloth studio run`.",
+)(studio_run)
diff --git a/unsloth_cli/commands/studio.py b/unsloth_cli/commands/studio.py
index ac25c8805d..e28f0f11f1 100644
--- a/unsloth_cli/commands/studio.py
+++ b/unsloth_cli/commands/studio.py
@@ -15,7 +15,7 @@ import time
 import types
 from datetime import datetime, timezone
 from pathlib import Path
-from typing import Optional
+from typing import List, Optional
 import typer
 
 studio_app = typer.Typer(help = "Unsloth Studio commands.")
@@ -374,6 +374,7 @@ def _load_model_via_http(
     gguf_variant: Optional[str],
     max_seq_length: int,
     load_in_4bit: bool,
+    llama_extra_args: Optional[List[str]] = None,
     timeout: int = 600,
 ) -> dict:
     """POST to ``/api/inference/load`` using the API key for auth."""
@@ -388,6 +389,8 @@ def _load_model_via_http(
     }
     if gguf_variant:
         payload["gguf_variant"] = gguf_variant
+    if llama_extra_args:
+        payload["llama_extra_args"] = list(llama_extra_args)
 
     data = json.dumps(payload).encode()
     req = urllib.request.Request(
@@ -514,9 +517,57 @@ def studio_default(
 # ── unsloth studio run ───────────────────────────────────────────────
 
 
-@studio_app.command()
+def _split_repo_variant(model_arg: str) -> tuple[str, Optional[str]]:
+    """Split ``org/name:variant`` HF-style identifiers into (repo, variant).
+
+    Mirrors llama.cpp's ``-hf <repo>:<quant>`` convention so users can
+    write ``unsloth/gpt-oss-20b-GGUF:UD-Q4_K_XL`` instead of passing
+    ``--gguf-variant`` separately. Local paths (absolute, ``./``,
+    ``~/``, Windows drive letters) and identifiers without a ``:``
+    suffix are returned verbatim.
+    """
+    s = model_arg.strip()
+    if not s:
+        return s, None
+    if s.startswith(("/", "./", "../", "~")) or s == ".":
+        return s, None
+    # Windows drive letter (e.g. "C:\\path" or "C:/path") -- the colon
+    # here is a path separator, not a variant suffix.
+    if len(s) >= 2 and s[1] == ":" and s[0].isalpha():
+        return s, None
+    if ":" not in s:
+        return s, None
+    repo, _, variant = s.rpartition(":")
+    if not repo or not variant:
+        return s, None
+    # A real quant label has no slashes; ``foo:bar/baz`` is not
+    # ``repo:variant`` syntax.
+    if "/" in variant:
+        return s, None
+    return repo, variant
+
+
+@studio_app.command(
+    context_settings = {
+        "allow_extra_args": True,
+        "ignore_unknown_options": True,
+    },
+)
 def run(
-    model: str = typer.Option(..., "--model", "-m", help = "Model path or HF repo"),
+    ctx: typer.Context,
+    model: str = typer.Option(
+        ...,
+        "--model",
+        "-m",
+        "-hf",
+        "-hfr",
+        "--hf-repo",
+        help = (
+            "Model path or HF repo. Accepts llama.cpp-style "
+            "`org/repo:variant` syntax. The `-hf` / `--hf-repo` aliases "
+            "match llama-server's spelling."
+        ),
+    ),
     gguf_variant: Optional[str] = typer.Option(
         None, "--gguf-variant", help = "GGUF quant variant (e.g. UD-Q4_K_XL)"
     ),
@@ -534,9 +585,35 @@ def run(
 ):
     """Start Studio, load a model, and print an API key -- one-liner server.
 
+    Any flag this command does not recognize is forwarded verbatim to
+    the underlying llama-server (GGUF only). Studio-managed flags
+    (--port, -c / --ctx-size, --api-key, -ngl, --jinja, --flash-attn,
+    --no-context-shift, model-identity flags, ...) are rejected with
+    HTTP 400.
+
     Example:
         unsloth studio run --model unsloth/Qwen3-1.7B-GGUF --gguf-variant UD-Q4_K_XL
+        unsloth studio run --model unsloth/Qwen3-1.7B-GGUF --top-k 20 --seed 42
+        unsloth studio run --model some-model --chat-template-file /path/to/tpl.jinja
     """
+    extra_llama_args: List[str] = list(ctx.args) if ctx.args else []
+
+    # ── 0. Parse llama.cpp-style ``repo:variant`` syntax in --model. ───
+    # Lets users write ``--model unsloth/foo-GGUF:UD-Q4_K_XL`` instead
+    # of pairing ``--model`` with ``--gguf-variant``. If both are given
+    # and disagree, fail loudly instead of silently picking one.
+    parsed_repo, embedded_variant = _split_repo_variant(model)
+    if embedded_variant:
+        if gguf_variant and gguf_variant != embedded_variant:
+            typer.echo(
+                f"Error: --model embeds variant '{embedded_variant}' but "
+                f"--gguf-variant '{gguf_variant}' was also provided.",
+                err = True,
+            )
+            raise typer.Exit(1)
+        model = parsed_repo
+        gguf_variant = gguf_variant or embedded_variant
+
     # ── 1. Venv re-exec (same pattern as studio_default) ──────────────
     studio_venv_dir = STUDIO_HOME / "unsloth_studio"
     in_studio_venv = sys.prefix.startswith(str(studio_venv_dir))
@@ -576,6 +653,11 @@ def run(
             args.extend(["--frontend", str(frontend)])
         if silent:
             args.append("--silent")
+        # Forward unknown args (llama-server pass-through) to the
+        # re-exec'd command so the studio venv sees them in ctx.args
+        # and the re-execed run() can include them in the load payload.
+        if extra_llama_args:
+            args.extend(extra_llama_args)
 
         if sys.platform == "win32":
             proc = subprocess.Popen(args)
@@ -617,6 +699,7 @@ def run(
             gguf_variant = gguf_variant,
             max_seq_length = max_seq_length,
             load_in_4bit = load_in_4bit,
+            llama_extra_args = extra_llama_args,
         )
     except RuntimeError as exc:
         typer.echo(f"Error: {exc}", err = True)

From 820b5c2e201380ec4f359426d26a36533713c7b5 Mon Sep 17 00:00:00 2001
From: Lee Jackson <130007945+Imagineer99@users.noreply.github.com>
Date: Mon, 4 May 2026 14:29:26 +0100
Subject: [PATCH 200/216] Studio: Always show API usage examples and docs links
 (#5270)

* chore: always show API usage examples and docs links

* chore: add colon to API setup docs label

* settings: move help links, add API shortcut, and update API copy
---
 .../frontend/src/components/app-sidebar.tsx   |  54 ++-----
 .../settings/components/usage-examples.tsx    | 140 ++++++++++--------
 .../src/features/settings/settings-dialog.tsx |  12 +-
 .../src/features/settings/tabs/about-tab.tsx  |  17 ++-
 .../features/settings/tabs/api-keys-tab.tsx   |  11 +-
 5 files changed, 120 insertions(+), 114 deletions(-)

diff --git a/studio/frontend/src/components/app-sidebar.tsx b/studio/frontend/src/components/app-sidebar.tsx
index b1dc286df5..88a0b03a21 100644
--- a/studio/frontend/src/components/app-sidebar.tsx
+++ b/studio/frontend/src/components/app-sidebar.tsx
@@ -31,16 +31,14 @@ import {
 import { useAnimatedThemeToggle } from "@/components/ui/animated-theme-toggler";
 import { cn } from "@/lib/utils";
 import {
-  Book03Icon,
   ChefHatIcon,
   ColumnInsertIcon,
   CursorInfo02Icon,
   Delete02Icon,
   Download03Icon,
   GemIcon,
-  MessageSearch01Icon,
+  Key01Icon,
   Search01Icon,
-  NewReleasesIcon,
   PowerIcon,
   PencilEdit02Icon,
   LayoutAlignLeftIcon,
@@ -547,6 +545,15 @@ export function AppSidebar() {
                     <span>Settings</span>
                     <DropdownMenuShortcut>⌘,</DropdownMenuShortcut>
                   </DropdownMenuItem>
+                  <DropdownMenuItem
+                    onSelect={() => useSettingsDialogStore.getState().openDialog("api-keys")}
+                  >
+                    <HugeiconsIcon icon={Key01Icon} strokeWidth={1.75} className="size-[18px]" />
+                    <span>API Keys</span>
+                    <span className="ml-auto rounded-[6px] border border-emerald-500/25 bg-emerald-500/10 px-1.5 py-0.5 text-[10px] leading-none font-semibold text-emerald-700 dark:text-emerald-300">
+                      New
+                    </span>
+                  </DropdownMenuItem>
                   <DropdownMenuItem
                     ref={anchorRef as React.Ref<HTMLDivElement>}
                     onSelect={(e) => { e.preventDefault(); toggleTheme(); }}
@@ -571,47 +578,6 @@ export function AppSidebar() {
                   </DropdownMenuItem>
                 </DropdownMenuGroup>
                 <DropdownMenuSeparator className="mx-2.5! my-2.5! h-0! border-t border-border/70 bg-transparent!" />
-                <DropdownMenuGroup>
-                  <DropdownMenuItem asChild>
-                    <a
-                      href="https://unsloth.ai/docs"
-                      target="_blank"
-                      rel="noopener noreferrer"
-                    >
-                      <HugeiconsIcon icon={Book03Icon} strokeWidth={1.75} className="size-[18px]" />
-                      <span>Learn More</span>
-                    </a>
-                  </DropdownMenuItem>
-                  <DropdownMenuItem asChild>
-                    <a
-                      href="https://unsloth.ai/docs/new/changelog"
-                      target="_blank"
-                      rel="noopener noreferrer"
-                    >
-                      <HugeiconsIcon
-                        icon={NewReleasesIcon}
-                        strokeWidth={1.75}
-                        className="size-[18px]"
-                      />
-                      <span>What's New</span>
-                    </a>
-                  </DropdownMenuItem>
-                  <DropdownMenuItem asChild>
-                    <a
-                      href="https://github.com/unslothai/unsloth/issues"
-                      target="_blank"
-                      rel="noopener noreferrer"
-                    >
-                      <HugeiconsIcon
-                        icon={MessageSearch01Icon}
-                        strokeWidth={1.75}
-                        className="size-[18px]"
-                      />
-                      <span>Feedback</span>
-                    </a>
-                  </DropdownMenuItem>
-                </DropdownMenuGroup>
-                <DropdownMenuSeparator className="mx-2.5! my-2.5! h-0! border-t border-border/70 bg-transparent!" />
                 <DropdownMenuItem onSelect={() => setShutdownOpen(true)}>
                   <HugeiconsIcon icon={PowerIcon} strokeWidth={1.75} className="size-[18px]" />
                   <span>Shutdown</span>
diff --git a/studio/frontend/src/features/settings/components/usage-examples.tsx b/studio/frontend/src/features/settings/components/usage-examples.tsx
index a2dc2c3774..a8f8ce31a6 100644
--- a/studio/frontend/src/features/settings/components/usage-examples.tsx
+++ b/studio/frontend/src/features/settings/components/usage-examples.tsx
@@ -4,12 +4,11 @@
 import { cn } from "@/lib/utils";
 import { copyToClipboard } from "@/lib/copy-to-clipboard";
 import {
-  ArrowDown01Icon,
+  ArrowUpRight01Icon,
   Copy01Icon,
   Tick02Icon,
 } from "@hugeicons/core-free-icons";
 import { HugeiconsIcon } from "@hugeicons/react";
-import { AnimatePresence, motion } from "motion/react";
 import { useMemo, useState } from "react";
 
 type Lang = "curl" | "python" | "tools";
@@ -20,6 +19,29 @@ const TABS: { id: Lang; label: string }[] = [
   { id: "tools", label: "Tools" },
 ];
 
+const DOC_LINKS = [
+  {
+    label: "Claude Code",
+    href: "https://unsloth.ai/docs/basics/claude-code",
+  },
+  {
+    label: "Codex",
+    href: "https://unsloth.ai/docs/basics/codex",
+  },
+  {
+    label: "OpenClaw",
+    href: "https://unsloth.ai/docs/integrations/openclaw",
+  },
+  {
+    label: "OpenCode",
+    href: "https://unsloth.ai/docs/integrations/opencode",
+  },
+  {
+    label: "Hermes Agent",
+    href: "https://unsloth.ai/docs/integrations/hermes-agent",
+  },
+];
+
 function buildSnippets(base: string) {
   return {
     curl: `curl ${base}/v1/chat/completions \\
@@ -56,7 +78,6 @@ for chunk in response:
 }
 
 export function UsageExamples() {
-  const [open, setOpen] = useState(false);
   const [lang, setLang] = useState<Lang>("curl");
   const [copied, setCopied] = useState(false);
   const snippets = useMemo(
@@ -76,71 +97,62 @@ export function UsageExamples() {
 
   return (
     <section className="flex flex-col">
-      <button
-        type="button"
-        onClick={() => setOpen((o) => !o)}
-        className="flex w-fit items-center gap-1.5 rounded text-xs font-medium text-muted-foreground transition-colors hover:text-foreground focus-visible:outline-none focus-visible:ring-2 focus-visible:ring-ring"
-        aria-expanded={open}
-      >
-        <HugeiconsIcon
-          icon={ArrowDown01Icon}
-          className={cn("size-3.5 transition-transform", open && "rotate-180")}
-        />
-        {open ? "Hide usage examples" : "Show usage examples"}
-      </button>
-
-      <AnimatePresence initial={false}>
-        {open && (
-          <motion.div
-            initial={{ height: 0, opacity: 0 }}
-            animate={{ height: "auto", opacity: 1 }}
-            exit={{ height: 0, opacity: 0 }}
-            transition={{ duration: 0.18, ease: [0.165, 0.84, 0.44, 1] }}
-            className="overflow-hidden"
-          >
-            <div className="mt-3 overflow-hidden rounded-lg border border-border bg-muted/20">
-              <div className="flex items-center justify-between border-b border-border px-2 py-1.5">
-                <div className="flex items-center gap-0.5">
-                  {TABS.map((t) => {
-                    const active = lang === t.id;
-                    return (
-                      <button
-                        key={t.id}
-                        type="button"
-                        onClick={() => setLang(t.id)}
-                        aria-pressed={active}
-                        className={cn(
-                          "rounded px-2 py-1 text-[11px] font-medium transition-colors focus-visible:outline-none focus-visible:ring-2 focus-visible:ring-ring",
-                          active
-                            ? "bg-background text-foreground shadow-border"
-                            : "text-muted-foreground hover:text-foreground",
-                        )}
-                      >
-                        {t.label}
-                      </button>
-                    );
-                  })}
-                </div>
+      <h2 className="mb-2 text-sm font-semibold text-foreground">Usage examples</h2>
+      <div className="overflow-hidden rounded-lg border border-border bg-muted/20">
+        <div className="flex items-center justify-between border-b border-border px-2 py-1.5">
+          <div className="flex items-center gap-0.5">
+            {TABS.map((t) => {
+              const active = lang === t.id;
+              return (
                 <button
+                  key={t.id}
                   type="button"
-                  onClick={handleCopy}
-                  className="flex items-center gap-1 rounded px-1.5 py-1 text-[11px] text-muted-foreground transition-colors hover:text-foreground focus-visible:outline-none focus-visible:ring-2 focus-visible:ring-ring"
-                  aria-label="Copy snippet"
+                  onClick={() => setLang(t.id)}
+                  aria-pressed={active}
+                  className={cn(
+                    "rounded px-2 py-1 text-[11px] font-medium transition-colors focus-visible:outline-none focus-visible:ring-2 focus-visible:ring-ring",
+                    active
+                      ? "bg-background text-foreground shadow-border"
+                      : "text-muted-foreground hover:text-foreground",
+                  )}
                 >
-                  <HugeiconsIcon
-                    icon={copied ? Tick02Icon : Copy01Icon}
-                    className={cn("size-3.5", copied && "text-emerald-600")}
-                  />
-                  {copied ? "Copied" : "Copy"}
+                  {t.label}
                 </button>
-              </div>
-              <pre className="overflow-x-auto p-3 font-mono text-[11px] leading-relaxed text-foreground">
-                {snippets[lang]}
-              </pre>
-            </div>
-          </motion.div>
-        )}
-      </AnimatePresence>
+              );
+            })}
+          </div>
+          <button
+            type="button"
+            onClick={handleCopy}
+            className="flex items-center gap-1 rounded px-1.5 py-1 text-[11px] text-muted-foreground transition-colors hover:text-foreground focus-visible:outline-none focus-visible:ring-2 focus-visible:ring-ring"
+            aria-label="Copy snippet"
+          >
+            <HugeiconsIcon
+              icon={copied ? Tick02Icon : Copy01Icon}
+              className={cn("size-3.5", copied && "text-emerald-600")}
+            />
+            {copied ? "Copied" : "Copy"}
+          </button>
+        </div>
+        <pre className="overflow-x-auto p-3 font-mono text-[11px] leading-relaxed text-foreground">
+          {snippets[lang]}
+        </pre>
+        <div className="flex flex-wrap items-center gap-x-2 gap-y-1 border-t border-border px-3 py-2 text-[11px] text-muted-foreground">
+          <span>Setup docs:</span>
+          {DOC_LINKS.map((link) => (
+            <a
+              key={link.href}
+              href={link.href}
+              target="_blank"
+              rel="noreferrer"
+              className="inline-flex items-center gap-0.5 rounded font-medium text-foreground underline decoration-border underline-offset-2 transition-colors hover:decoration-foreground focus-visible:outline-none focus-visible:ring-2 focus-visible:ring-ring"
+            >
+              {link.label}
+              <HugeiconsIcon icon={ArrowUpRight01Icon} className="size-3" />
+            </a>
+          ))}
+        </div>
+      </div>
     </section>
   );
 }
diff --git a/studio/frontend/src/features/settings/settings-dialog.tsx b/studio/frontend/src/features/settings/settings-dialog.tsx
index bb22fcb89c..423b5c74f0 100644
--- a/studio/frontend/src/features/settings/settings-dialog.tsx
+++ b/studio/frontend/src/features/settings/settings-dialog.tsx
@@ -31,6 +31,7 @@ interface TabDef {
   id: SettingsTab;
   label: string;
   icon: typeof Settings02Icon;
+  badge?: string;
 }
 
 const TABS: TabDef[] = [
@@ -38,8 +39,8 @@ const TABS: TabDef[] = [
   { id: "profile", label: "Profile", icon: UserIcon },
   { id: "appearance", label: "Appearance", icon: PaintBrush02Icon },
   { id: "chat", label: "Chat", icon: Message01Icon },
-  { id: "api-keys", label: "API Keys", icon: Key01Icon },
-  { id: "about", label: "About", icon: SparklesIcon },
+  { id: "api-keys", label: "API Keys", icon: Key01Icon, badge: "New" },
+  { id: "about", label: "Help", icon: SparklesIcon },
 ];
 
 function renderTab(tab: SettingsTab) {
@@ -121,7 +122,12 @@ export function SettingsDialog() {
                       strokeWidth={1.5}
                       className="relative z-10 size-[18px]"
                     />
-                    <span className="relative z-10">{tab.label}</span>
+                    <span className="relative z-10 min-w-0 truncate">{tab.label}</span>
+                    {tab.badge ? (
+                      <span className="relative z-10 ml-auto rounded-[6px] border border-emerald-500/25 bg-emerald-500/10 px-1.5 py-0.5 text-[10px] leading-none font-semibold text-emerald-700 dark:text-emerald-300">
+                        {tab.badge}
+                      </span>
+                    ) : null}
                   </button>
                 );
               })}
diff --git a/studio/frontend/src/features/settings/tabs/about-tab.tsx b/studio/frontend/src/features/settings/tabs/about-tab.tsx
index bb4454aa43..9629900237 100644
--- a/studio/frontend/src/features/settings/tabs/about-tab.tsx
+++ b/studio/frontend/src/features/settings/tabs/about-tab.tsx
@@ -12,6 +12,7 @@ import {
   Book03Icon,
   Cancel01Icon,
   MessageNotification01Icon,
+  NewReleasesIcon,
 } from "@hugeicons/core-free-icons";
 import { HugeiconsIcon } from "@hugeicons/react";
 import { useEffect, useState } from "react";
@@ -48,9 +49,9 @@ export function AboutTab() {
   return (
     <div className="flex flex-col gap-6">
       <header className="flex flex-col gap-1">
-        <h1 className="text-lg font-semibold font-heading">About</h1>
+        <h1 className="text-lg font-semibold font-heading">Help</h1>
         <p className="text-xs text-muted-foreground">
-          Unsloth Studio build info and support.
+          Documentation, release notes, feedback, and Studio build info.
         </p>
       </header>
 
@@ -79,6 +80,18 @@ export function AboutTab() {
             <HugeiconsIcon icon={ArrowUpRight01Icon} className="size-3" />
           </a>
         </SettingsRow>
+        <SettingsRow label="Release notes">
+          <a
+            href="https://unsloth.ai/docs/new/changelog"
+            target="_blank"
+            rel="noopener noreferrer"
+            className="inline-flex items-center gap-1 text-xs font-medium text-muted-foreground hover:text-foreground"
+          >
+            <HugeiconsIcon icon={NewReleasesIcon} className="size-3.5" />
+            What's new
+            <HugeiconsIcon icon={ArrowUpRight01Icon} className="size-3" />
+          </a>
+        </SettingsRow>
         <SettingsRow label="Feedback">
           <a
             href="https://github.com/unslothai/unsloth/issues"
diff --git a/studio/frontend/src/features/settings/tabs/api-keys-tab.tsx b/studio/frontend/src/features/settings/tabs/api-keys-tab.tsx
index e0fa6d87d2..c979adcc46 100644
--- a/studio/frontend/src/features/settings/tabs/api-keys-tab.tsx
+++ b/studio/frontend/src/features/settings/tabs/api-keys-tab.tsx
@@ -65,7 +65,16 @@ export function ApiKeysTab() {
       <header className="flex flex-col gap-1">
         <h1 className="text-lg font-semibold font-heading">API Keys</h1>
         <p className="text-xs text-muted-foreground">
-          Access Unsloth Studio programmatically via the OpenAI-compatible API.
+          Access Unsloth programmatically via the OpenAI-compatible API.{" "}
+          <a
+            href="https://unsloth.ai/docs/basics/api"
+            target="_blank"
+            rel="noreferrer"
+            className="font-medium text-foreground underline decoration-border underline-offset-2 transition-colors hover:decoration-foreground"
+          >
+            Read the API docs
+          </a>
+          .
         </p>
       </header>
 

From 5533bdb8b63ecfef2685ef67892ead83f8c54fb3 Mon Sep 17 00:00:00 2001
From: Lee Jackson <130007945+Imagineer99@users.noreply.github.com>
Date: Mon, 4 May 2026 14:56:38 +0100
Subject: [PATCH 201/216] Studio: Change API Keys settings to API Access
 (#5268)

* chore: change API Keys settings to API Access

* chore: replace key with plug svg

* chore: replace API access with developer and related SVG

* chore: rename API keys menu to developer

* fix: focus active settings tab on open

* fix: prevent settings autofocus on open

* Revert "fix: prevent settings autofocus on open"

This reverts commit 4b64d73edaf696f07b1c52e4c76a6868ab703303.
---
 .../frontend/src/components/app-sidebar.tsx   |  6 ++---
 .../src/features/settings/api/api-keys.ts     |  6 ++---
 .../settings/components/api-key-row.tsx       |  2 +-
 .../settings/components/create-key-form.tsx   |  8 +++----
 .../settings/components/key-reveal-card.tsx   |  4 ++--
 .../src/features/settings/settings-dialog.tsx | 24 +++++++++++++++++--
 .../features/settings/tabs/api-keys-tab.tsx   | 14 +++++------
 .../features/settings/tabs/general-tab.tsx    |  4 ++--
 8 files changed, 44 insertions(+), 24 deletions(-)

diff --git a/studio/frontend/src/components/app-sidebar.tsx b/studio/frontend/src/components/app-sidebar.tsx
index 88a0b03a21..171b4eb92d 100644
--- a/studio/frontend/src/components/app-sidebar.tsx
+++ b/studio/frontend/src/components/app-sidebar.tsx
@@ -37,12 +37,12 @@ import {
   Delete02Icon,
   Download03Icon,
   GemIcon,
-  Key01Icon,
   Search01Icon,
   PowerIcon,
   PencilEdit02Icon,
   LayoutAlignLeftIcon,
   Settings02Icon,
+  SourceCodeSquareIcon,
   ZapIcon,
 } from "@hugeicons/core-free-icons";
 import {
@@ -548,8 +548,8 @@ export function AppSidebar() {
                   <DropdownMenuItem
                     onSelect={() => useSettingsDialogStore.getState().openDialog("api-keys")}
                   >
-                    <HugeiconsIcon icon={Key01Icon} strokeWidth={1.75} className="size-[18px]" />
-                    <span>API Keys</span>
+                    <HugeiconsIcon icon={SourceCodeSquareIcon} strokeWidth={1.75} className="size-[18px]" />
+                    <span>Developer</span>
                     <span className="ml-auto rounded-[6px] border border-emerald-500/25 bg-emerald-500/10 px-1.5 py-0.5 text-[10px] leading-none font-semibold text-emerald-700 dark:text-emerald-300">
                       New
                     </span>
diff --git a/studio/frontend/src/features/settings/api/api-keys.ts b/studio/frontend/src/features/settings/api/api-keys.ts
index 73d7b50def..5bf3f2825b 100644
--- a/studio/frontend/src/features/settings/api/api-keys.ts
+++ b/studio/frontend/src/features/settings/api/api-keys.ts
@@ -15,7 +15,7 @@ export interface ApiKey {
 
 export async function fetchApiKeys(): Promise<ApiKey[]> {
   const res = await authFetch("/api/auth/api-keys");
-  if (!res.ok) throw new Error("Failed to load API keys");
+  if (!res.ok) throw new Error("Failed to load API access");
   const data = (await res.json()) as { api_keys: ApiKey[] };
   return data.api_keys.filter((k) => k.is_active);
 }
@@ -29,7 +29,7 @@ export async function createApiKey(
     headers: { "Content-Type": "application/json" },
     body: JSON.stringify({ name, expires_in_days: expiresInDays }),
   });
-  if (!res.ok) throw new Error("Failed to create API key");
+  if (!res.ok) throw new Error("Failed to create access token");
   return res.json();
 }
 
@@ -37,5 +37,5 @@ export async function revokeApiKey(keyId: number): Promise<void> {
   const res = await authFetch(`/api/auth/api-keys/${keyId}`, {
     method: "DELETE",
   });
-  if (!res.ok) throw new Error("Failed to revoke API key");
+  if (!res.ok) throw new Error("Failed to revoke access token");
 }
diff --git a/studio/frontend/src/features/settings/components/api-key-row.tsx b/studio/frontend/src/features/settings/components/api-key-row.tsx
index f121045e1e..ced6de17d1 100644
--- a/studio/frontend/src/features/settings/components/api-key-row.tsx
+++ b/studio/frontend/src/features/settings/components/api-key-row.tsx
@@ -92,7 +92,7 @@ export function ApiKeyRow({
             className="text-destructive focus:text-destructive"
           >
             <HugeiconsIcon icon={Delete02Icon} className="size-3.5 mr-2" />
-            Revoke key
+            Revoke token
           </DropdownMenuItem>
         </DropdownMenuContent>
       </DropdownMenu>
diff --git a/studio/frontend/src/features/settings/components/create-key-form.tsx b/studio/frontend/src/features/settings/components/create-key-form.tsx
index 2166abf3cf..a0f2d7f82d 100644
--- a/studio/frontend/src/features/settings/components/create-key-form.tsx
+++ b/studio/frontend/src/features/settings/components/create-key-form.tsx
@@ -34,7 +34,7 @@ export function CreateKeyForm({
       onCreated(result.key);
       setName("");
     } catch (err) {
-      onError(err instanceof Error ? err.message : "Couldn't create key.");
+      onError(err instanceof Error ? err.message : "Couldn't create access token.");
     } finally {
       setLoading(false);
     }
@@ -49,9 +49,9 @@ export function CreateKeyForm({
         <Input
           value={name}
           onChange={(e) => setName(e.target.value)}
-          placeholder="Key name (e.g. production)"
+          placeholder="Token name (e.g. production)"
           className="h-8 min-w-[180px] flex-1 text-sm"
-          aria-label="New key name"
+          aria-label="New access token name"
         />
         <div className="inline-flex items-center rounded-md border border-border bg-background p-0.5">
           {EXPIRY_PRESETS.map((p) => {
@@ -75,7 +75,7 @@ export function CreateKeyForm({
           })}
         </div>
         <Button type="submit" size="sm" disabled={loading || !name.trim()}>
-          {loading ? "Creating…" : "Create key"}
+          {loading ? "Creating…" : "Create token"}
         </Button>
       </div>
     </form>
diff --git a/studio/frontend/src/features/settings/components/key-reveal-card.tsx b/studio/frontend/src/features/settings/components/key-reveal-card.tsx
index ab258537da..2b589e88fe 100644
--- a/studio/frontend/src/features/settings/components/key-reveal-card.tsx
+++ b/studio/frontend/src/features/settings/components/key-reveal-card.tsx
@@ -32,7 +32,7 @@ export function KeyRevealCard({
           className="size-3.5 text-emerald-600 dark:text-emerald-500"
         />
         <span className="text-xs font-medium text-emerald-700 dark:text-emerald-500">
-          New key created
+          New access token created
         </span>
       </div>
       <button
@@ -43,7 +43,7 @@ export function KeyRevealCard({
           "focus-visible:outline-none focus-visible:ring-2 focus-visible:ring-ring focus-visible:ring-offset-1 focus-visible:ring-offset-background",
           copied && "border-emerald-500/40 bg-emerald-500/10",
         )}
-        aria-label={copied ? "Key copied" : "Copy key"}
+        aria-label={copied ? "Access token copied" : "Copy access token"}
       >
         <code className="min-w-0 flex-1 break-all text-left text-foreground">
           {rawKey}
diff --git a/studio/frontend/src/features/settings/settings-dialog.tsx b/studio/frontend/src/features/settings/settings-dialog.tsx
index 423b5c74f0..376af06e9d 100644
--- a/studio/frontend/src/features/settings/settings-dialog.tsx
+++ b/studio/frontend/src/features/settings/settings-dialog.tsx
@@ -10,15 +10,16 @@ import {
 import { cn } from "@/lib/utils";
 import {
   Cancel01Icon,
-  Key01Icon,
   Message01Icon,
   PaintBrush02Icon,
   Settings02Icon,
+  SourceCodeSquareIcon,
   SparklesIcon,
   UserIcon,
 } from "@hugeicons/core-free-icons";
 import { HugeiconsIcon } from "@hugeicons/react";
 import { motion, useReducedMotion } from "motion/react";
+import { useEffect, useRef } from "react";
 import { useSettingsDialogStore, type SettingsTab } from "./stores/settings-dialog-store";
 import { AboutTab } from "./tabs/about-tab";
 import { ApiKeysTab } from "./tabs/api-keys-tab";
@@ -39,7 +40,7 @@ const TABS: TabDef[] = [
   { id: "profile", label: "Profile", icon: UserIcon },
   { id: "appearance", label: "Appearance", icon: PaintBrush02Icon },
   { id: "chat", label: "Chat", icon: Message01Icon },
-  { id: "api-keys", label: "API Keys", icon: Key01Icon, badge: "New" },
+  { id: "api-keys", label: "Developer", icon: SourceCodeSquareIcon, badge: "New" },
   { id: "about", label: "Help", icon: SparklesIcon },
 ];
 
@@ -66,6 +67,22 @@ export function SettingsDialog() {
   const setActiveTab = useSettingsDialogStore((s) => s.setActiveTab);
   const closeDialog = useSettingsDialogStore((s) => s.closeDialog);
   const reduced = useReducedMotion();
+  const tabButtonRefs = useRef<Record<SettingsTab, HTMLButtonElement | null>>({
+    general: null,
+    profile: null,
+    appearance: null,
+    chat: null,
+    "api-keys": null,
+    about: null,
+  });
+
+  useEffect(() => {
+    if (!open) return;
+    const frame = window.requestAnimationFrame(() => {
+      tabButtonRefs.current[activeTab]?.focus({ preventScroll: true });
+    });
+    return () => window.cancelAnimationFrame(frame);
+  }, [open, activeTab]);
 
   return (
     <Dialog open={open} onOpenChange={(o) => !o && closeDialog()}>
@@ -91,6 +108,9 @@ export function SettingsDialog() {
                 return (
                   <button
                     key={tab.id}
+                    ref={(node) => {
+                      tabButtonRefs.current[tab.id] = node;
+                    }}
                     type="button"
                     onClick={() => setActiveTab(tab.id)}
                     className={cn(
diff --git a/studio/frontend/src/features/settings/tabs/api-keys-tab.tsx b/studio/frontend/src/features/settings/tabs/api-keys-tab.tsx
index c979adcc46..64c6f520c0 100644
--- a/studio/frontend/src/features/settings/tabs/api-keys-tab.tsx
+++ b/studio/frontend/src/features/settings/tabs/api-keys-tab.tsx
@@ -36,7 +36,7 @@ export function ApiKeysTab() {
     try {
       setKeys(await fetchApiKeys());
     } catch (e) {
-      setError(e instanceof Error ? e.message : "Couldn't load API keys.");
+      setError(e instanceof Error ? e.message : "Couldn't load API access.");
     } finally {
       setLoading(false);
     }
@@ -54,7 +54,7 @@ export function ApiKeysTab() {
       await load();
       setRevokeTarget(null);
     } catch (e) {
-      setError(e instanceof Error ? e.message : "Couldn't revoke key.");
+      setError(e instanceof Error ? e.message : "Couldn't revoke access token.");
     } finally {
       setRevoking(false);
     }
@@ -63,7 +63,7 @@ export function ApiKeysTab() {
   return (
     <div className="flex flex-col gap-6">
       <header className="flex flex-col gap-1">
-        <h1 className="text-lg font-semibold font-heading">API Keys</h1>
+        <h1 className="text-lg font-semibold font-heading">Developer</h1>
         <p className="text-xs text-muted-foreground">
           Access Unsloth programmatically via the OpenAI-compatible API.{" "}
           <a
@@ -112,7 +112,7 @@ export function ApiKeysTab() {
       </AnimatePresence>
 
       <section className="flex flex-col">
-        <h2 className="mb-2 text-sm font-semibold text-foreground">Your keys</h2>
+        <h2 className="mb-2 text-sm font-semibold text-foreground">Access tokens</h2>
         {error ? (
           <div className="rounded-md border border-destructive/20 bg-destructive/5 p-3 text-xs text-destructive">
             {error}
@@ -128,7 +128,7 @@ export function ApiKeysTab() {
           </div>
         ) : keys.length === 0 ? (
           <p className="py-6 text-center text-xs text-muted-foreground">
-            No API keys yet.
+            No API access yet.
           </p>
         ) : (
           <div className="flex flex-col">
@@ -144,9 +144,9 @@ export function ApiKeysTab() {
       <Dialog open={revokeTarget !== null} onOpenChange={(o) => !o && setRevokeTarget(null)}>
         <DialogContent className="max-w-md">
           <DialogHeader>
-            <DialogTitle>Revoke key “{revokeTarget?.name}”?</DialogTitle>
+            <DialogTitle>Revoke access token “{revokeTarget?.name}”?</DialogTitle>
             <DialogDescription>
-              Applications using this key will immediately lose access. This cannot be undone.
+              Applications using this token will immediately lose access. This cannot be undone.
             </DialogDescription>
           </DialogHeader>
           <DialogFooter>
diff --git a/studio/frontend/src/features/settings/tabs/general-tab.tsx b/studio/frontend/src/features/settings/tabs/general-tab.tsx
index 6081e90605..9ea89f55cf 100644
--- a/studio/frontend/src/features/settings/tabs/general-tab.tsx
+++ b/studio/frontend/src/features/settings/tabs/general-tab.tsx
@@ -199,7 +199,7 @@ export function GeneralTab() {
         <SettingsRow
           destructive
           label="Reset all local preferences"
-          description="Clears theme, tokens, sidebar state, and presets. Chats and API keys are not affected."
+          description="Clears theme, tokens, sidebar state, and presets. Chats and API access are not affected."
         >
           <Button
             variant="outline"
@@ -218,7 +218,7 @@ export function GeneralTab() {
             <DialogTitle>Reset all local preferences?</DialogTitle>
             <DialogDescription>
               This clears your theme, tokens, and stored settings, then reloads
-              Studio. Chats and API keys are not affected.
+              Studio. Chats and API access are not affected.
             </DialogDescription>
           </DialogHeader>
           <DialogFooter>

From 726abd5e6b1888dd5bfb1c47b0b1f9e94b51d5f9 Mon Sep 17 00:00:00 2001
From: Wasim Yousef Said <wasimysdev@gmail.com>
Date: Tue, 5 May 2026 09:09:48 +0200
Subject: [PATCH 202/216] Add Tauri native notifications (#5273)

---
 studio/frontend/package.json                  |   1 +
 .../chat/hooks/use-chat-model-runtime.ts      |  29 +++
 .../training/hooks/use-training-actions.ts    |   9 +-
 .../hooks/use-training-runtime-lifecycle.ts   | 103 +++++++-
 .../frontend/src/lib/native-notifications.ts  | 234 ++++++++++++++++++
 studio/src-tauri/Cargo.lock                   |  96 +++++++
 studio/src-tauri/Cargo.toml                   |   1 +
 studio/src-tauri/capabilities/default.json    |   3 +
 studio/src-tauri/src/main.rs                  |   1 +
 9 files changed, 470 insertions(+), 7 deletions(-)
 create mode 100644 studio/frontend/src/lib/native-notifications.ts

diff --git a/studio/frontend/package.json b/studio/frontend/package.json
index 3adc7ed0c8..6b24440964 100644
--- a/studio/frontend/package.json
+++ b/studio/frontend/package.json
@@ -44,6 +44,7 @@
     "@tanstack/react-table": "^8.21.3",
     "@tauri-apps/api": "^2.10.1",
     "@tauri-apps/plugin-clipboard-manager": "^2.3.2",
+    "@tauri-apps/plugin-notification": "^2.3.3",
     "@tauri-apps/plugin-opener": "^2.5.3",
     "@tauri-apps/plugin-process": "^2.3.1",
     "@tauri-apps/plugin-updater": "^2.10.1",
diff --git a/studio/frontend/src/features/chat/hooks/use-chat-model-runtime.ts b/studio/frontend/src/features/chat/hooks/use-chat-model-runtime.ts
index eacf54a208..3fc78231c6 100644
--- a/studio/frontend/src/features/chat/hooks/use-chat-model-runtime.ts
+++ b/studio/frontend/src/features/chat/hooks/use-chat-model-runtime.ts
@@ -4,6 +4,12 @@
 import { createElement, useCallback, useRef, useState } from "react";
 import { toast } from "sonner";
 import { consumeNativePathToken } from "@/features/native-intents/api";
+import {
+  notifyNative,
+  primeNativeNotificationPermission,
+  safeNotificationLabel,
+  sanitizeNotificationBody,
+} from "@/lib/native-notifications";
 import { ModelLoadDescription } from "../components/model-load-status";
 import {
   getDownloadProgress,
@@ -168,6 +174,7 @@ export function useChatModelRuntime() {
   const loadAbortRef = useRef<AbortController | null>(null);
   const loadingModelRef = useRef<typeof loadingModel>(null);
   const loadToastIdRef = useRef<string | number | null>(null);
+  const loadAttemptRef = useRef(0);
   const loadToastDismissedRef = useRef(false);
   const cancelUnloadPendingRef = useRef(false);
 
@@ -369,6 +376,10 @@ export function useChatModelRuntime() {
       const isLora =
         explicitIsLora ?? model?.isLora ?? loraIsAdapter ?? false;
       const displayName = model?.name || lora?.name || modelId;
+      const loadAttemptId = ++loadAttemptRef.current;
+      primeNativeNotificationPermission().catch(() => undefined);
+      const notificationModelKey = `${modelId}:${ggufVariant ?? ""}:${loadAttemptId}`;
+      const safeModelName = safeNotificationLabel(displayName, "The model");
       const currentCheckpoint =
         useChatRuntimeStore.getState().params.checkpoint;
       const previousCheckpoint = currentCheckpoint;
@@ -813,6 +824,12 @@ export function useChatModelRuntime() {
                   },
                 });
               }
+              notifyNative({
+                key: `model-downloaded:${notificationModelKey}`,
+                title: "Model downloaded",
+                body: `${safeModelName} finished downloading and is loading into memory.`,
+                requestPermission: false,
+              }).catch(() => undefined);
               // Keep polling: the mmap branch below takes over from here.
             }
           } catch {
@@ -898,6 +915,12 @@ export function useChatModelRuntime() {
               duration: 2000,
             });
           }
+          notifyNative({
+            key: `model-loaded:${notificationModelKey}`,
+            title: "Model ready",
+            body: `${safeModelName} is loaded and ready to chat.`,
+            requestPermission: false,
+          }).catch(() => undefined);
         } catch (err) {
           if (!abortCtrl.signal.aborted) {
             const message =
@@ -912,6 +935,12 @@ export function useChatModelRuntime() {
                 duration: 5000,
               });
             }
+            notifyNative({
+              key: `model-load-failed:${notificationModelKey}`,
+              title: "Model failed to load",
+              body: sanitizeNotificationBody(message, "The model failed to load."),
+              requestPermission: false,
+            }).catch(() => undefined);
           }
           throw err;
         } finally {
diff --git a/studio/frontend/src/features/training/hooks/use-training-actions.ts b/studio/frontend/src/features/training/hooks/use-training-actions.ts
index 7f275d5d82..deaec6c6c2 100644
--- a/studio/frontend/src/features/training/hooks/use-training-actions.ts
+++ b/studio/frontend/src/features/training/hooks/use-training-actions.ts
@@ -1,11 +1,13 @@
 // SPDX-License-Identifier: AGPL-3.0-only
 // Copyright 2026-present the Unsloth AI Inc. team. All rights reserved. See /studio/LICENSE.AGPL-3.0
 
+import { primeNativeNotificationPermission } from "@/lib/native-notifications";
 import { useCallback } from "react";
+import { toast } from "sonner";
 import { checkDatasetFormat } from "../api/datasets-api";
 import { getTrainingRun } from "../api/history-api";
 import { buildTrainingStartPayload } from "../api/mappers";
-import { startTraining, stopTraining, resetTraining } from "../api/train-api";
+import { resetTraining, startTraining, stopTraining } from "../api/train-api";
 import { syncTrainingRuntimeFromBackend } from "../lib/sync-runtime";
 import { validateTrainingConfig } from "../lib/validation";
 import { useDatasetPreviewDialogStore } from "../stores/dataset-preview-dialog-store";
@@ -13,7 +15,6 @@ import { useTrainingConfigStore } from "../stores/training-config-store";
 import { useTrainingRuntimeStore } from "../stores/training-runtime-store";
 import type { TrainingStartRequest } from "../types/api";
 import type { TrainingConfigState } from "../types/config";
-import { toast } from "sonner";
 
 /** Chatml → format-specific role remap (only for formats that differ from chatml). */
 const ROLE_REMAP: Record<string, Record<string, string>> = {
@@ -50,6 +51,8 @@ export function useTrainingActions() {
       return false;
     }
 
+    primeNativeNotificationPermission().catch(() => undefined);
+
     runtimeStore.setStartResources(
       config.selectedModel ?? null,
       getHfDatasetName(config),
@@ -174,6 +177,8 @@ export function useTrainingActions() {
         throw new Error("Only stopped runs with a saved checkpoint can be resumed.");
       }
 
+      primeNativeNotificationPermission().catch(() => undefined);
+
       const config = useTrainingConfigStore.getState();
       const savedConfig = detail.config as Partial<TrainingStartRequest>;
       const payload = {
diff --git a/studio/frontend/src/features/training/hooks/use-training-runtime-lifecycle.ts b/studio/frontend/src/features/training/hooks/use-training-runtime-lifecycle.ts
index 6e4c7c85b9..cd9e76ede9 100644
--- a/studio/frontend/src/features/training/hooks/use-training-runtime-lifecycle.ts
+++ b/studio/frontend/src/features/training/hooks/use-training-runtime-lifecycle.ts
@@ -2,6 +2,11 @@
 // Copyright 2026-present the Unsloth AI Inc. team. All rights reserved. See /studio/LICENSE.AGPL-3.0
 
 import { hasAuthToken } from "@/features/auth";
+import {
+  notifyNative,
+  safeNotificationLabel,
+  sanitizeNotificationBody,
+} from "@/lib/native-notifications";
 import { useEffect } from "react";
 import {
   getTrainingMetrics,
@@ -9,6 +14,7 @@ import {
   isAbortError,
   streamTrainingProgress,
 } from "../api/train-api";
+import { useTrainingConfigStore } from "../stores/training-config-store";
 import { useTrainingRuntimeStore } from "../stores/training-runtime-store";
 import type { TrainingRuntimeStore } from "../types/runtime";
 
@@ -16,11 +22,78 @@ const STATUS_POLL_INTERVAL_MS = 3000;
 const METRICS_POLL_INTERVAL_MS = 5000;
 const IDLE_POLL_INTERVAL_MS = 30000;
 const STREAM_RECONNECT_DELAY_MS = 1500;
+const GENERIC_TRAINING_STREAM_ERROR = "Training stream error";
+const DEFAULT_TRAINING_ERROR_BODY = "Your training run stopped with an error.";
 
 function shouldUseLiveSync(state: TrainingRuntimeStore): boolean {
   return state.isTrainingRunning || state.phase === "training";
 }
 
+function getTrainingErrorBody(state: TrainingRuntimeStore): string {
+  return sanitizeNotificationBody(
+    state.error ?? state.message,
+    DEFAULT_TRAINING_ERROR_BODY,
+  );
+}
+
+function shouldNotifyTrainingError(
+  before: TrainingRuntimeStore,
+  after: TrainingRuntimeStore,
+): boolean {
+  if (after.phase !== "error") {
+    return false;
+  }
+  if (before.phase !== "error" || before.jobId !== after.jobId) {
+    return true;
+  }
+  if (before.error !== GENERIC_TRAINING_STREAM_ERROR) {
+    return false;
+  }
+
+  return getTrainingErrorBody(before) !== getTrainingErrorBody(after);
+}
+
+function maybeNotifyTrainingTerminalTransition(
+  before: TrainingRuntimeStore,
+  after: TrainingRuntimeStore,
+): void {
+  if (!after.jobId) {
+    return;
+  }
+  if (
+    before.stopRequested ||
+    after.stopRequested ||
+    after.phase === "stopped"
+  ) {
+    return;
+  }
+
+  const sameJob = before.jobId === after.jobId;
+  const samePhase = before.phase === after.phase;
+
+  if (after.phase === "completed" && !(sameJob && samePhase)) {
+    const model = safeNotificationLabel(
+      after.startModelName ?? useTrainingConfigStore.getState().selectedModel,
+      "Your training run",
+    );
+    notifyNative({
+      key: `training-completed:${after.jobId}`,
+      title: "Training finished",
+      body: `${model} is complete.`,
+      requestPermission: false,
+    }).catch(() => undefined);
+  }
+
+  if (shouldNotifyTrainingError(before, after)) {
+    notifyNative({
+      key: `training-error:${after.jobId}`,
+      title: "Training failed",
+      body: getTrainingErrorBody(after),
+      requestPermission: false,
+    }).catch(() => undefined);
+  }
+}
+
 export function useTrainingRuntimeLifecycle(): void {
   useEffect(() => {
     let disposed = false;
@@ -62,7 +135,9 @@ export function useTrainingRuntimeLifecycle(): void {
       }
     };
 
-    const pollStatus = async () => {
+    const pollStatus = async (options?: {
+      suppressNativeNotifications?: boolean;
+    }) => {
       if (!hasAuthToken()) return;
       const gen = runtimeStore.getState().resetGeneration;
       try {
@@ -71,9 +146,13 @@ export function useTrainingRuntimeLifecycle(): void {
           return;
         }
 
+        const previousState = runtimeStore.getState();
         runtimeStore.getState().applyStatus(status);
 
         const nextState = runtimeStore.getState();
+        if (!options?.suppressNativeNotifications) {
+          maybeNotifyTrainingTerminalTransition(previousState, nextState);
+        }
         if (shouldUseLiveSync(nextState)) {
           void ensureStream();
         } else {
@@ -124,7 +203,7 @@ export function useTrainingRuntimeLifecycle(): void {
             }
 
             if (event.event === "error") {
-              liveStore.setRuntimeError("Training stream error");
+              liveStore.setRuntimeError(GENERIC_TRAINING_STREAM_ERROR);
               stopStream();
             }
           },
@@ -154,7 +233,10 @@ export function useTrainingRuntimeLifecycle(): void {
     const hydrate = async () => {
       runtimeStore.getState().setHydrating(true);
       try {
-        await Promise.all([pollStatus(), pollMetrics()]);
+        await Promise.all([
+          pollStatus({ suppressNativeNotifications: true }),
+          pollMetrics(),
+        ]);
       } finally {
         if (!disposed) {
           runtimeStore.getState().setHydrating(false);
@@ -172,7 +254,12 @@ export function useTrainingRuntimeLifecycle(): void {
 
     const statusTimer = setInterval(() => {
       const s = runtimeStore.getState();
-      if (isIdle() && s.hasHydrated) return;
+      if (!s.hasHydrated || s.isHydrating) {
+        return;
+      }
+      if (isIdle()) {
+        return;
+      }
       void pollStatus();
     }, STATUS_POLL_INTERVAL_MS);
 
@@ -187,7 +274,13 @@ export function useTrainingRuntimeLifecycle(): void {
     // Low-frequency background poll to recover from failed hydration or detect
     // out-of-band state changes (e.g. training started from another client).
     const idleTimer = setInterval(() => {
-      if (!isIdle()) return;
+      const s = runtimeStore.getState();
+      if (!s.hasHydrated || s.isHydrating) {
+        return;
+      }
+      if (!isIdle()) {
+        return;
+      }
       void pollStatus();
     }, IDLE_POLL_INTERVAL_MS);
 
diff --git a/studio/frontend/src/lib/native-notifications.ts b/studio/frontend/src/lib/native-notifications.ts
new file mode 100644
index 0000000000..641421f6ea
--- /dev/null
+++ b/studio/frontend/src/lib/native-notifications.ts
@@ -0,0 +1,234 @@
+// SPDX-License-Identifier: AGPL-3.0-only
+// Copyright 2026-present the Unsloth AI Inc. team. All rights reserved. See /studio/LICENSE.AGPL-3.0
+
+import { isTauri } from "@/lib/api-base";
+
+const MAX_BODY_LENGTH = 200;
+const MAX_SENT_KEYS = 200;
+const HF_TOKEN_PATTERN = /\bhf_[A-Za-z0-9]{20,}\b/g;
+const GITHUB_TOKEN_PATTERN = /\bghp_[A-Za-z0-9_]{20,}\b/g;
+const GITHUB_PAT_PATTERN = /\bgithub_pat_[A-Za-z0-9_]{20,}\b/g;
+const OPENAI_KEY_PATTERN = /\bsk-[A-Za-z0-9_-]{20,}\b/g;
+const UNIX_HOME_PATH_PATTERN =
+  /(^|[\s"'(=])(?:\/Users|\/home)\/[^\s"'<>),;]+/gi;
+const UNIX_ABSOLUTE_PATH_PATTERN = /(^|[\s"'(=])\/[^\s"'<>),;]+/g;
+const TILDE_PATH_PATTERN = /(^|[\s"'(=])~[\\/][^\s"'<>),;]+/g;
+const RELATIVE_PATH_PATTERN = /(^|[\s"'(=])\.{1,2}[\\/][^\s"'<>),;]+/g;
+const WINDOWS_DRIVE_PATH_PATTERN = /(^|[\s"'(=])[A-Za-z]:[\\/][^\s"'<>),;]+/g;
+const WINDOWS_UNC_PATH_PATTERN = /(^|[\s"'(=])\\\\[^\s"'<>),;]+/g;
+const WHITESPACE_PATTERN = /\s+/g;
+const LOCAL_PATH_PATTERN = /^(?:\/|~[\\/]|\.{1,2}[\\/]|[A-Za-z]:[\\/]|\\\\)/;
+const PATH_SEPARATOR_PATTERN = /[\\/]/;
+const SENSITIVE_TOKEN_PATTERN =
+  /\bhf_[A-Za-z0-9]{20,}\b|\bghp_[A-Za-z0-9_]{20,}\b|\bgithub_pat_[A-Za-z0-9_]{20,}\b|\bsk-[A-Za-z0-9_-]{20,}\b/;
+
+type PermissionState = "unknown" | "granted" | "denied";
+
+let permissionState: PermissionState = "unknown";
+let permissionPrimeInFlight: Promise<void> | null = null;
+const inFlightKeys = new Set<string>();
+const sentKeys = new Set<string>();
+
+export interface NativeNotificationOptions {
+  key: string;
+  title: string;
+  body?: string;
+  requestPermission?: boolean;
+}
+
+function hasGrantedPermission(): boolean {
+  return permissionState === "granted";
+}
+
+function getBrowserNotificationApi(): typeof Notification | null {
+  if (typeof window === "undefined" || !("Notification" in window)) {
+    return null;
+  }
+
+  return window.Notification;
+}
+
+function rememberSentKey(key: string): void {
+  sentKeys.add(key);
+
+  if (sentKeys.size <= MAX_SENT_KEYS) {
+    return;
+  }
+
+  const oldestKey = sentKeys.values().next().value;
+  if (oldestKey) {
+    sentKeys.delete(oldestKey);
+  }
+}
+
+async function checkNativePermission(allowRequest: boolean): Promise<boolean> {
+  const notification = getBrowserNotificationApi();
+  if (notification?.permission === "granted") {
+    permissionState = "granted";
+    return true;
+  }
+
+  if (notification?.permission === "denied") {
+    permissionState = "denied";
+    return false;
+  }
+
+  if (allowRequest && notification?.permission === "default") {
+    const permission = await notification.requestPermission();
+    if (permission === "granted") {
+      permissionState = "granted";
+      return true;
+    }
+    if (permission === "denied") {
+      permissionState = "denied";
+    }
+    return false;
+  }
+
+  try {
+    const { isPermissionGranted } = await import(
+      "@tauri-apps/plugin-notification"
+    );
+    const granted = await isPermissionGranted();
+
+    if (granted) {
+      permissionState = "granted";
+    }
+
+    return granted;
+  } catch {
+    return false;
+  }
+}
+
+async function ensurePermission(allowRequest: boolean): Promise<boolean> {
+  if (!isTauri) {
+    return false;
+  }
+  if (permissionState !== "unknown") {
+    return hasGrantedPermission();
+  }
+
+  if (permissionPrimeInFlight) {
+    await permissionPrimeInFlight;
+    if (permissionState !== "unknown" || !allowRequest) {
+      return hasGrantedPermission();
+    }
+  }
+
+  return checkNativePermission(allowRequest);
+}
+
+export async function primeNativeNotificationPermission(): Promise<void> {
+  if (!isTauri || permissionState !== "unknown") {
+    return;
+  }
+
+  if (!permissionPrimeInFlight) {
+    permissionPrimeInFlight = checkNativePermission(true)
+      .then(() => undefined)
+      .finally(() => {
+        permissionPrimeInFlight = null;
+      });
+  }
+
+  await permissionPrimeInFlight;
+}
+
+export async function notifyNative(
+  options: NativeNotificationOptions,
+): Promise<void> {
+  if (!isTauri) {
+    return;
+  }
+  if (sentKeys.has(options.key) || inFlightKeys.has(options.key)) {
+    return;
+  }
+  inFlightKeys.add(options.key);
+
+  try {
+    const granted = await ensurePermission(options.requestPermission !== false);
+    if (!granted) {
+      return;
+    }
+
+    const body = options.body
+      ? sanitizeNotificationBody(options.body, "")
+      : undefined;
+    const { sendNotification } = await import(
+      "@tauri-apps/plugin-notification"
+    );
+    sendNotification({ title: options.title, body });
+    rememberSentKey(options.key);
+  } catch {
+    // Native notifications are non-critical UI side effects.
+  } finally {
+    inFlightKeys.delete(options.key);
+  }
+}
+
+export function sanitizeNotificationBody(
+  input: string | null | undefined,
+  fallback: string,
+): string {
+  const fallbackText = fallback.trim() || "Notification update.";
+  let text = typeof input === "string" ? input.trim() : "";
+
+  if (!text) {
+    text = fallbackText;
+  }
+
+  text = text
+    .replace(HF_TOKEN_PATTERN, "hf_[redacted]")
+    .replace(GITHUB_TOKEN_PATTERN, "ghp_[redacted]")
+    .replace(GITHUB_PAT_PATTERN, "github_pat_[redacted]")
+    .replace(OPENAI_KEY_PATTERN, "sk-[redacted]")
+    .replace(UNIX_HOME_PATH_PATTERN, "$1[path]")
+    .replace(UNIX_ABSOLUTE_PATH_PATTERN, "$1[path]")
+    .replace(TILDE_PATH_PATTERN, "$1[path]")
+    .replace(RELATIVE_PATH_PATTERN, "$1[path]")
+    .replace(WINDOWS_DRIVE_PATH_PATTERN, "$1[path]")
+    .replace(WINDOWS_UNC_PATH_PATTERN, "$1[path]")
+    .replace(WHITESPACE_PATTERN, " ")
+    .trim();
+
+  if (!text) {
+    text = fallbackText;
+  }
+
+  if (text.length > MAX_BODY_LENGTH) {
+    return `${text.slice(0, MAX_BODY_LENGTH - 1).trimEnd()}…`;
+  }
+
+  return text;
+}
+
+export function safeNotificationLabel(
+  input: string | null | undefined,
+  fallback: string,
+): string {
+  const fallbackText = fallback.trim() || "The item";
+  const raw = typeof input === "string" ? input.trim() : "";
+  if (!raw) {
+    return fallbackText;
+  }
+
+  const collapsed = raw.replace(WHITESPACE_PATTERN, " ");
+  const isLocalPath = LOCAL_PATH_PATTERN.test(collapsed);
+
+  if (isLocalPath) {
+    const basename =
+      collapsed.split(PATH_SEPARATOR_PATTERN).filter(Boolean).at(-1) ?? "";
+    if (
+      basename &&
+      basename !== "." &&
+      basename !== ".." &&
+      !SENSITIVE_TOKEN_PATTERN.test(basename)
+    ) {
+      return sanitizeNotificationBody(basename, fallbackText);
+    }
+    return fallbackText;
+  }
+
+  return sanitizeNotificationBody(collapsed, fallbackText);
+}
diff --git a/studio/src-tauri/Cargo.lock b/studio/src-tauri/Cargo.lock
index dd57e55dc5..c73d7cc0c5 100644
--- a/studio/src-tauri/Cargo.lock
+++ b/studio/src-tauri/Cargo.lock
@@ -2301,6 +2301,18 @@ version = "0.1.1"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "c41e0c4fef86961ac6d6f8a82609f55f31b05e4fce149ac5710e439df7619ba4"
 
+[[package]]
+name = "mac-notification-sys"
+version = "0.6.12"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "29a16783dd1a47849b8c8133c9cd3eb2112cfbc6901670af3dba47c8bbfb07d3"
+dependencies = [
+ "cc",
+ "objc2",
+ "objc2-foundation",
+ "time",
+]
+
 [[package]]
 name = "markup5ever"
 version = "0.14.1"
@@ -2502,6 +2514,20 @@ dependencies = [
  "memchr",
 ]
 
+[[package]]
+name = "notify-rust"
+version = "4.16.1"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "3bdaf6120b9df005d37e58f6b75329be6255450453fbeba9ce4192324f921fb9"
+dependencies = [
+ "futures-lite",
+ "log",
+ "mac-notification-sys",
+ "serde",
+ "tauri-winrt-notification",
+ "zbus",
+]
+
 [[package]]
 name = "num-conv"
 version = "0.2.0"
@@ -3271,6 +3297,15 @@ version = "2.0.1"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "a993555f31e5a609f617c12db6250dedcac1b0a85076912c436e6fc9b2c8e6a3"
 
+[[package]]
+name = "quick-xml"
+version = "0.37.5"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "331e97a1af0bf59823e6eadffe373d7b27f485be8748f71471c662c1f269b7fb"
+dependencies = [
+ "memchr",
+]
+
 [[package]]
 name = "quick-xml"
 version = "0.38.4"
@@ -3335,6 +3370,16 @@ dependencies = [
  "rand_core 0.6.4",
 ]
 
+[[package]]
+name = "rand"
+version = "0.9.4"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "44c5af06bb1b7d3216d91932aed5265164bf384dc89cd6ba05cf59a35f5f76ea"
+dependencies = [
+ "rand_chacha 0.9.0",
+ "rand_core 0.9.5",
+]
+
 [[package]]
 name = "rand"
 version = "0.10.0"
@@ -3366,6 +3411,16 @@ dependencies = [
  "rand_core 0.6.4",
 ]
 
+[[package]]
+name = "rand_chacha"
+version = "0.9.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "d3022b5f1df60f26e1ffddd6c66e8aa15de382ae63b3a0c1bfc0e4d3e3f325cb"
+dependencies = [
+ "ppv-lite86",
+ "rand_core 0.9.5",
+]
+
 [[package]]
 name = "rand_core"
 version = "0.5.1"
@@ -3384,6 +3439,15 @@ dependencies = [
  "getrandom 0.2.17",
 ]
 
+[[package]]
+name = "rand_core"
+version = "0.9.5"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "76afc826de14238e6e8c374ddcc1fa19e374fd8dd986b0d2af0d02377261d83c"
+dependencies = [
+ "getrandom 0.3.4",
+]
+
 [[package]]
 name = "rand_core"
 version = "0.10.0"
@@ -4589,6 +4653,25 @@ dependencies = [
  "url",
 ]
 
+[[package]]
+name = "tauri-plugin-notification"
+version = "2.3.3"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "01fc2c5ff41105bd1f7242d8201fdf3efd70749b82fa013a17f2126357d194cc"
+dependencies = [
+ "log",
+ "notify-rust",
+ "rand 0.9.4",
+ "serde",
+ "serde_json",
+ "serde_repr",
+ "tauri",
+ "tauri-plugin",
+ "thiserror 2.0.18",
+ "time",
+ "url",
+]
+
 [[package]]
 name = "tauri-plugin-opener"
 version = "2.5.3"
@@ -4769,6 +4852,18 @@ dependencies = [
  "toml 0.9.12+spec-1.1.0",
 ]
 
+[[package]]
+name = "tauri-winrt-notification"
+version = "0.7.2"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "0b1e66e07de489fe43a46678dd0b8df65e0c973909df1b60ba33874e297ba9b9"
+dependencies = [
+ "quick-xml 0.37.5",
+ "thiserror 2.0.18",
+ "windows 0.61.3",
+ "windows-version",
+]
+
 [[package]]
 name = "tempfile"
 version = "3.27.0"
@@ -5296,6 +5391,7 @@ dependencies = [
  "tauri-build",
  "tauri-plugin-clipboard-manager",
  "tauri-plugin-dialog",
+ "tauri-plugin-notification",
  "tauri-plugin-opener",
  "tauri-plugin-process",
  "tauri-plugin-single-instance",
diff --git a/studio/src-tauri/Cargo.toml b/studio/src-tauri/Cargo.toml
index d3cfe49667..4002ab420d 100644
--- a/studio/src-tauri/Cargo.toml
+++ b/studio/src-tauri/Cargo.toml
@@ -28,6 +28,7 @@ tauri-plugin-updater = "2"
 tauri-plugin-clipboard-manager = "2"
 tauri-plugin-dialog = "2"
 rand = "0.10.0"
+tauri-plugin-notification = "2.3.3"
 
 [target.'cfg(unix)'.dependencies]
 libc = "0.2"
diff --git a/studio/src-tauri/capabilities/default.json b/studio/src-tauri/capabilities/default.json
index b624369abd..413fa14cc1 100644
--- a/studio/src-tauri/capabilities/default.json
+++ b/studio/src-tauri/capabilities/default.json
@@ -20,6 +20,9 @@
     "core:window:allow-close",
     "core:tray:default",
     "process:default",
+    "notification:allow-is-permission-granted",
+    "notification:allow-request-permission",
+    "notification:allow-notify",
     {
       "identifier": "opener:allow-open-url",
       "allow": [{ "url": "https://*" }, { "url": "http://*" }, { "url": "mailto:*" }]
diff --git a/studio/src-tauri/src/main.rs b/studio/src-tauri/src/main.rs
index 5aeaa56497..7ad40b8be2 100644
--- a/studio/src-tauri/src/main.rs
+++ b/studio/src-tauri/src/main.rs
@@ -167,6 +167,7 @@ fn main() {
         }))
         .plugin(tauri_plugin_process::init())
         .plugin(tauri_plugin_opener::init())
+        .plugin(tauri_plugin_notification::init())
         .plugin(tauri_plugin_dialog::init())
         .plugin(tauri_plugin_updater::Builder::new().build())
         .plugin(tauri_plugin_clipboard_manager::init())

From 0da8af56d6b0c982a1d02c329df8406c94444fe5 Mon Sep 17 00:00:00 2001
From: Roland Tannous <115670425+rolandtannous@users.noreply.github.com>
Date: Tue, 5 May 2026 12:45:15 +0400
Subject: [PATCH 203/216] unsloth run: add --enable-tools/--disable-tools
 server-side tool policy (#5277)

* Add process-level tool_policy state for unsloth run

* Apply tool_policy override at chat/completions, /messages, and tool pass-through gates

* Add pure resolver for unsloth run --enable-tools/--disable-tools

* Wire --enable-tools/--disable-tools into unsloth run

* Color tool-policy notices and confirmation prompt in Claude orange

* Always show tool-status notice; print URL + API key in silent mode

* Treat any non-loopback bind as external; forward --yes after parent prompt

* Fix tool_policy double-module bug: import via state.tool_policy to share global with routes
---
 studio/backend/routes/inference.py            |  25 ++-
 studio/backend/state/tool_policy.py           |  33 +++
 .../backend/tests/test_tool_policy_gates.py   |  56 +++++
 .../backend/tests/test_tool_policy_state.py   |  59 +++++
 .../test_unsloth_run_tool_policy_resolver.py  | 201 ++++++++++++++++++
 unsloth_cli/_tool_policy.py                   |  71 +++++++
 unsloth_cli/commands/studio.py                |  82 +++++++
 7 files changed, 524 insertions(+), 3 deletions(-)
 create mode 100644 studio/backend/state/tool_policy.py
 create mode 100644 studio/backend/tests/test_tool_policy_gates.py
 create mode 100644 studio/backend/tests/test_tool_policy_state.py
 create mode 100644 tests/python/test_unsloth_run_tool_policy_resolver.py
 create mode 100644 unsloth_cli/_tool_policy.py

diff --git a/studio/backend/routes/inference.py b/studio/backend/routes/inference.py
index 0690cc6b79..a6b00360af 100644
--- a/studio/backend/routes/inference.py
+++ b/studio/backend/routes/inference.py
@@ -215,6 +215,18 @@ router = APIRouter()
 studio_router = APIRouter()
 
 
+def _effective_enable_tools(payload) -> Optional[bool]:
+    """Resolve `payload.enable_tools` against the process-level tool policy.
+
+    Returns the policy value when set (CLI hard-override from `unsloth run`),
+    otherwise the per-request value.
+    """
+    from state.tool_policy import get_tool_policy
+
+    policy = get_tool_policy()
+    return policy if policy is not None else payload.enable_tools
+
+
 # Cancel registry. Proxies (e.g. Colab) can swallow client fetch aborts
 # so is_disconnected() never fires. POST /inference/cancel looks up
 # in-flight cancel_events here by cancel_id (per-run) or session_id /
@@ -1654,7 +1666,7 @@ async def openai_chat_completions(
     )
     if (
         using_gguf
-        and not payload.enable_tools
+        and not _effective_enable_tools(payload)
         and (_tools_passthrough or _has_response_format)
     ):
         # Preserve the vision guard that would otherwise run in the
@@ -1745,8 +1757,13 @@ async def openai_chat_completions(
         created = int(time.time())
 
         # ── Tool-calling path (agentic loop) ──────────────────
+        # `_effective_enable_tools` lets `unsloth run --enable-tools/--disable-tools`
+        # hard-override the per-request value. Without a CLI override, falls
+        # back to `payload.enable_tools` (existing behavior).
         use_tools = (
-            payload.enable_tools and llama_backend.supports_tools and not image_b64
+            _effective_enable_tools(payload)
+            and llama_backend.supports_tools
+            and not image_b64
         )
 
         if use_tools:
@@ -3470,7 +3487,9 @@ async def anthropic_messages(
     # Server-side agentic loop doesn't support multimodal input — matches
     # the `not image_b64` gate in /v1/chat/completions.
     server_tools = (
-        payload.enable_tools and llama_backend.supports_tools and not _has_image
+        _effective_enable_tools(payload)
+        and llama_backend.supports_tools
+        and not _has_image
     )
     client_tools = (
         not server_tools
diff --git a/studio/backend/state/tool_policy.py b/studio/backend/state/tool_policy.py
new file mode 100644
index 0000000000..9343a39806
--- /dev/null
+++ b/studio/backend/state/tool_policy.py
@@ -0,0 +1,33 @@
+# SPDX-License-Identifier: AGPL-3.0-only
+# Copyright 2026-present the Unsloth AI Inc. team. All rights reserved.
+
+"""Process-level server-side tool policy.
+
+Set by `unsloth run` at startup; consulted by the inference route gates.
+
+  None  -> no CLI override (default). Per-request `enable_tools` is honored.
+  True  -> CLI forced tools on for every request.
+  False -> CLI forced tools off for every request.
+"""
+
+from typing import Optional
+
+_tool_policy: Optional[bool] = None
+
+
+def get_tool_policy() -> Optional[bool]:
+    return _tool_policy
+
+
+def set_tool_policy(value: Optional[bool]) -> None:
+    if value is not None and not isinstance(value, bool):
+        raise TypeError(
+            f"tool_policy must be Optional[bool], got {type(value).__name__}"
+        )
+    global _tool_policy
+    _tool_policy = value
+
+
+def reset_tool_policy() -> None:
+    global _tool_policy
+    _tool_policy = None
diff --git a/studio/backend/tests/test_tool_policy_gates.py b/studio/backend/tests/test_tool_policy_gates.py
new file mode 100644
index 0000000000..01f6bbbc3f
--- /dev/null
+++ b/studio/backend/tests/test_tool_policy_gates.py
@@ -0,0 +1,56 @@
+# SPDX-License-Identifier: AGPL-3.0-only
+# Copyright 2026-present the Unsloth AI Inc. team. All rights reserved.
+
+"""
+Tests for `_effective_enable_tools` -- the helper that folds the
+process-level `tool_policy` over a request's `enable_tools` field.
+
+Truth table (policy x payload.enable_tools -> effective):
+  policy=None  + payload=None  -> None
+  policy=None  + payload=True  -> True
+  policy=None  + payload=False -> False
+  policy=True  + payload=*     -> True
+  policy=False + payload=*     -> False
+"""
+
+import os
+import sys
+from types import SimpleNamespace
+
+_backend = os.path.join(os.path.dirname(__file__), "..")
+sys.path.insert(0, _backend)
+
+import pytest
+
+from routes.inference import _effective_enable_tools
+from state.tool_policy import reset_tool_policy, set_tool_policy
+
+
+@pytest.fixture(autouse = True)
+def _reset():
+    reset_tool_policy()
+    yield
+    reset_tool_policy()
+
+
+def _payload(value):
+    return SimpleNamespace(enable_tools = value)
+
+
+class TestEffectiveEnableTools:
+    @pytest.mark.parametrize(
+        "payload_value,expected",
+        [(None, None), (True, True), (False, False)],
+    )
+    def test_no_policy_falls_through_to_payload(self, payload_value, expected):
+        assert _effective_enable_tools(_payload(payload_value)) == expected
+
+    @pytest.mark.parametrize("payload_value", [None, True, False])
+    def test_policy_true_overrides_any_payload(self, payload_value):
+        set_tool_policy(True)
+        assert _effective_enable_tools(_payload(payload_value)) is True
+
+    @pytest.mark.parametrize("payload_value", [None, True, False])
+    def test_policy_false_overrides_any_payload(self, payload_value):
+        set_tool_policy(False)
+        assert _effective_enable_tools(_payload(payload_value)) is False
diff --git a/studio/backend/tests/test_tool_policy_state.py b/studio/backend/tests/test_tool_policy_state.py
new file mode 100644
index 0000000000..5f6b228281
--- /dev/null
+++ b/studio/backend/tests/test_tool_policy_state.py
@@ -0,0 +1,59 @@
+# SPDX-License-Identifier: AGPL-3.0-only
+# Copyright 2026-present the Unsloth AI Inc. team. All rights reserved.
+
+"""
+Tests for the process-level server-side tool policy used by `unsloth run`.
+
+The policy has three states:
+  None  -> no CLI override (default; honor per-request enable_tools)
+  True  -> CLI forced tools on
+  False -> CLI forced tools off
+"""
+
+import os
+import sys
+
+_backend = os.path.join(os.path.dirname(__file__), "..")
+sys.path.insert(0, _backend)
+
+import pytest
+
+from state.tool_policy import (
+    get_tool_policy,
+    reset_tool_policy,
+    set_tool_policy,
+)
+
+
+@pytest.fixture(autouse = True)
+def _reset():
+    reset_tool_policy()
+    yield
+    reset_tool_policy()
+
+
+class TestToolPolicy:
+    def test_default_is_none(self):
+        assert get_tool_policy() is None
+
+    def test_set_true_then_get(self):
+        set_tool_policy(True)
+        assert get_tool_policy() is True
+
+    def test_set_false_then_get(self):
+        set_tool_policy(False)
+        assert get_tool_policy() is False
+
+    def test_set_none_clears(self):
+        set_tool_policy(True)
+        set_tool_policy(None)
+        assert get_tool_policy() is None
+
+    def test_reset_clears(self):
+        set_tool_policy(False)
+        reset_tool_policy()
+        assert get_tool_policy() is None
+
+    def test_rejects_non_optional_bool(self):
+        with pytest.raises(TypeError):
+            set_tool_policy("true")  # type: ignore[arg-type]
diff --git a/tests/python/test_unsloth_run_tool_policy_resolver.py b/tests/python/test_unsloth_run_tool_policy_resolver.py
new file mode 100644
index 0000000000..6aff02494b
--- /dev/null
+++ b/tests/python/test_unsloth_run_tool_policy_resolver.py
@@ -0,0 +1,201 @@
+# Copyright 2025-present the Unsloth AI Inc. team. All rights reserved.
+
+"""
+Truth-table tests for `resolve_tool_policy` -- the pure resolver behind
+`unsloth run --enable-tools/--disable-tools`.
+
+Covers:
+  - 127.0.0.1 default-on, explicit on, explicit off
+  - 0.0.0.0 default-off, explicit off
+  - 0.0.0.0 + explicit on: confirm prompt unless --silent or --yes,
+    abort on negative answer.
+"""
+
+import pytest
+import typer
+
+from unsloth_cli._tool_policy import is_external_host, resolve_tool_policy
+
+
+def _never_prompt(_msg: str) -> bool:
+    raise AssertionError("prompt should not have been called")
+
+
+def _prompt_yes(_msg: str) -> bool:
+    return True
+
+
+def _prompt_no(_msg: str) -> bool:
+    return False
+
+
+class TestLocalhostHost:
+    @pytest.mark.parametrize("flag", [None, True, False])
+    def test_no_prompt(self, flag):
+        # localhost never prompts regardless of flag
+        result = resolve_tool_policy(
+            host = "127.0.0.1",
+            flag = flag,
+            yes = False,
+            silent = False,
+            prompt = _never_prompt,
+        )
+        assert result is (True if flag in (None, True) else False)
+
+    def test_default_is_on(self):
+        assert (
+            resolve_tool_policy(
+                host = "127.0.0.1",
+                flag = None,
+                yes = False,
+                silent = False,
+                prompt = _never_prompt,
+            )
+            is True
+        )
+
+    def test_explicit_off(self):
+        assert (
+            resolve_tool_policy(
+                host = "127.0.0.1",
+                flag = False,
+                yes = False,
+                silent = False,
+                prompt = _never_prompt,
+            )
+            is False
+        )
+
+
+class TestZeroHost:
+    def test_default_is_off(self):
+        assert (
+            resolve_tool_policy(
+                host = "0.0.0.0",
+                flag = None,
+                yes = False,
+                silent = False,
+                prompt = _never_prompt,
+            )
+            is False
+        )
+
+    def test_explicit_off_no_prompt(self):
+        assert (
+            resolve_tool_policy(
+                host = "0.0.0.0",
+                flag = False,
+                yes = False,
+                silent = False,
+                prompt = _never_prompt,
+            )
+            is False
+        )
+
+    def test_explicit_on_silent_skips_prompt(self):
+        assert (
+            resolve_tool_policy(
+                host = "0.0.0.0",
+                flag = True,
+                yes = False,
+                silent = True,
+                prompt = _never_prompt,
+            )
+            is True
+        )
+
+    def test_explicit_on_yes_skips_prompt(self):
+        assert (
+            resolve_tool_policy(
+                host = "0.0.0.0",
+                flag = True,
+                yes = True,
+                silent = False,
+                prompt = _never_prompt,
+            )
+            is True
+        )
+
+    def test_explicit_on_prompt_yes(self):
+        assert (
+            resolve_tool_policy(
+                host = "0.0.0.0",
+                flag = True,
+                yes = False,
+                silent = False,
+                prompt = _prompt_yes,
+            )
+            is True
+        )
+
+    def test_explicit_on_prompt_no_aborts(self):
+        with pytest.raises(typer.Exit) as exc_info:
+            resolve_tool_policy(
+                host = "0.0.0.0",
+                flag = True,
+                yes = False,
+                silent = False,
+                prompt = _prompt_no,
+            )
+        assert exc_info.value.exit_code == 1
+
+
+class TestIsExternalHost:
+    @pytest.mark.parametrize(
+        "host", ["127.0.0.1", "localhost", "::1", "LOCALHOST", "Localhost"]
+    )
+    def test_loopback_aliases_are_local(self, host):
+        assert is_external_host(host) is False
+
+    @pytest.mark.parametrize(
+        "host", ["0.0.0.0", "::", "192.168.1.5", "10.0.0.1", "example.com"]
+    )
+    def test_non_loopback_is_external(self, host):
+        assert is_external_host(host) is True
+
+
+class TestSpecificNetworkIP:
+    """Binding to a specific LAN IP must follow the same rules as 0.0.0.0."""
+
+    def test_default_is_off(self):
+        assert (
+            resolve_tool_policy(
+                host = "192.168.1.5",
+                flag = None,
+                yes = False,
+                silent = False,
+                prompt = _never_prompt,
+            )
+            is False
+        )
+
+    def test_explicit_on_prompts(self):
+        seen = []
+
+        def _prompt(msg: str) -> bool:
+            seen.append(msg)
+            return True
+
+        assert (
+            resolve_tool_policy(
+                host = "192.168.1.5",
+                flag = True,
+                yes = False,
+                silent = False,
+                prompt = _prompt,
+            )
+            is True
+        )
+        assert any("192.168.1.5" in m for m in seen)
+
+    def test_localhost_alias_does_not_prompt(self):
+        assert (
+            resolve_tool_policy(
+                host = "localhost",
+                flag = True,
+                yes = False,
+                silent = False,
+                prompt = _never_prompt,
+            )
+            is True
+        )
diff --git a/unsloth_cli/_tool_policy.py b/unsloth_cli/_tool_policy.py
new file mode 100644
index 0000000000..5dd6c70d9f
--- /dev/null
+++ b/unsloth_cli/_tool_policy.py
@@ -0,0 +1,71 @@
+# Copyright 2025-present the Unsloth AI Inc. team. All rights reserved.
+
+"""Pure resolver for `unsloth run --enable-tools/--disable-tools`.
+
+Kept as a standalone module so the truth table can be unit-tested
+without spinning up Typer or the studio venv.
+"""
+
+from typing import Callable, Optional
+
+import typer
+
+# Claude Code brand orange — used so the security warning stands out
+# in a crowded terminal.
+_PROMPT_FG = (217, 119, 87)
+
+# Loopback aliases. Any other bind address (0.0.0.0, ::, a specific
+# LAN IP, a hostname, ...) is treated as network-reachable.
+_LOOPBACK_HOSTS = frozenset({"127.0.0.1", "localhost", "::1"})
+
+
+def is_external_host(host: str) -> bool:
+    """True when `host` is reachable from beyond loopback."""
+    return host.lower() not in _LOOPBACK_HOSTS
+
+
+def _build_prompt_text(host: str) -> str:
+    return typer.style(
+        (
+            f"Tools include arbitrary code execution (Python, terminal). "
+            f"You're binding to {host}, which is reachable from your network. "
+            f"If your API key leaks, anyone with it can run code on this machine. "
+            f"Do not share the API key. Continue?"
+        ),
+        fg = _PROMPT_FG,
+        bold = True,
+    )
+
+
+def resolve_tool_policy(
+    host: str,
+    flag: Optional[bool],
+    yes: bool,
+    silent: bool,
+    prompt: Callable[[str], bool] = typer.confirm,
+) -> bool:
+    """Return the resolved server-side tool policy.
+
+    Args:
+        host: The bind address (e.g. "127.0.0.1", "0.0.0.0", or a
+              specific network IP).
+        flag: Tri-state from `--enable-tools/--disable-tools`
+              (None when neither was passed).
+        yes:  True if the operator passed `--yes/-y`.
+        silent: True if the operator passed `--silent/-q`.
+        prompt: Callable used for the network-bind + on confirmation
+                (injected for testability; defaults to ``typer.confirm``).
+
+    Raises:
+        typer.Exit: when the operator declines the confirmation.
+    """
+    is_external = is_external_host(host)
+    default = not is_external  # loopback defaults on, network defaults off
+
+    resolved = default if flag is None else flag
+
+    if is_external and resolved is True and not yes and not silent:
+        if not prompt(_build_prompt_text(host)):
+            raise typer.Exit(1)
+
+    return resolved
diff --git a/unsloth_cli/commands/studio.py b/unsloth_cli/commands/studio.py
index e28f0f11f1..140940209f 100644
--- a/unsloth_cli/commands/studio.py
+++ b/unsloth_cli/commands/studio.py
@@ -582,6 +582,20 @@ def run(
     host: str = typer.Option("127.0.0.1", "--host", "-H"),
     frontend: Optional[Path] = typer.Option(None, "--frontend", "-f"),
     silent: bool = typer.Option(False, "--silent", "-q"),
+    enable_tools: Optional[bool] = typer.Option(
+        None,
+        "--enable-tools/--disable-tools",
+        help = (
+            "Force server-side tools on/off for all requests. "
+            "Default: on for 127.0.0.1, off for 0.0.0.0."
+        ),
+    ),
+    yes: bool = typer.Option(
+        False,
+        "--yes",
+        "-y",
+        help = "Skip the 0.0.0.0 + --enable-tools confirmation prompt.",
+    ),
 ):
     """Start Studio, load a model, and print an API key -- one-liner server.
 
@@ -614,6 +628,17 @@ def run(
         model = parsed_repo
         gguf_variant = gguf_variant or embedded_variant
 
+    # ── Resolve the server-side tool policy. The y/N prompt (if any)
+    # runs in the outer process so the re-exec'd child never re-prompts.
+    from unsloth_cli._tool_policy import is_external_host, resolve_tool_policy
+
+    enable_tools = resolve_tool_policy(
+        host = host,
+        flag = enable_tools,
+        yes = yes,
+        silent = silent,
+    )
+
     # ── 1. Venv re-exec (same pattern as studio_default) ──────────────
     studio_venv_dir = STUDIO_HOME / "unsloth_studio"
     in_studio_venv = sys.prefix.startswith(str(studio_venv_dir))
@@ -653,6 +678,18 @@ def run(
             args.extend(["--frontend", str(frontend)])
         if silent:
             args.append("--silent")
+        # Forward the resolved tool policy (always concrete True/False
+        # at this point — the resolver above ran before the re-exec).
+        if enable_tools:
+            args.append("--enable-tools")
+        else:
+            args.append("--disable-tools")
+        # Forward --yes whenever the parent already cleared the prompt
+        # (either operator passed --yes, or the parent's resolver
+        # accepted the network-bind confirmation). Otherwise the child
+        # re-runs the resolver and prompts a second time.
+        if yes or (enable_tools and is_external_host(host)):
+            args.append("--yes")
         # Forward unknown args (llama-server pass-through) to the
         # re-exec'd command so the studio venv sees them in ctx.args
         # and the re-execed run() can include them in the load payload.
@@ -678,6 +715,17 @@ def run(
     app = run_server(**run_kwargs)
     actual_port = getattr(app.state, "server_port", port) or port
 
+    # ── Apply the resolved tool policy as a process-level override.
+    # Must use the same import path the route handlers use --
+    # `studio/backend/run.py` adds `studio/backend/` to sys.path so the
+    # routes import this module as top-level `state.tool_policy`. If we
+    # imported via `studio.backend.state.tool_policy` instead, Python
+    # would cache two different module objects with two different
+    # `_tool_policy` globals, and the gates would never see our value.
+    from state.tool_policy import set_tool_policy
+
+    set_tool_policy(enable_tools)
+
     # ── 3. Wait for server health ─────────────────────────────────────
     if not silent:
         typer.echo("Starting Unsloth Studio...")
@@ -713,6 +761,32 @@ def run(
     base_url = f"http://{display_host}:{actual_port}"
     sdk_base_url = f"{base_url}/v1"
 
+    # Claude orange (Claude Code's brand color) for tool-policy notices
+    # so they stand out from the surrounding banner. Always printed --
+    # even under --silent / --yes -- so the operator never misses the
+    # current tool-execution status.
+    _tool_notice_fg = (217, 119, 87)
+    _is_external = is_external_host(host)
+    if _is_external and enable_tools:
+        _tool_notice = (
+            f"Server-side tools are ENABLED on {host} (network-reachable). "
+            f"Anyone with the API key can run code on this machine. "
+            f"Do not share the API key."
+        )
+    elif _is_external:
+        _tool_notice = (
+            f"Server-side tools are disabled by default on {host} "
+            f"(network-reachable). Pass --enable-tools to turn on "
+            f"(you will be warned about API-key risk)."
+        )
+    elif enable_tools:
+        _tool_notice = (
+            "Server-side tools are enabled by default for loopback. "
+            "Pass --disable-tools to turn off."
+        )
+    else:
+        _tool_notice = "Server-side tools are disabled."
+
     if not silent:
         typer.echo("")
         typer.echo("=" * 56)
@@ -723,6 +797,7 @@ def run(
         typer.echo("  OpenAI / Anthropic SDK base URL:")
         typer.echo(f"    {sdk_base_url}")
         typer.echo("=" * 56)
+        typer.secho(_tool_notice, fg = _tool_notice_fg, bold = True)
         typer.echo("")
         typer.echo("OpenAI Chat Completions:")
         typer.echo(f"  curl {sdk_base_url}/chat/completions \\")
@@ -746,6 +821,13 @@ def run(
         typer.echo('    -H "Content-Type: application/json" \\')
         typer.echo("""    -d '{"input": "Hello", "stream": true}'""")
         typer.echo("")
+    else:
+        # Silent mode still prints the essentials (URL, API key) plus
+        # the orange tool-status notice so the operator never loses
+        # visibility into the security-relevant policy.
+        typer.echo(f"URL:     {base_url}")
+        typer.echo(f"API Key: {api_key}")
+        typer.secho(_tool_notice, fg = _tool_notice_fg, bold = True)
 
     # ── 7. Wait for Ctrl+C ────────────────────────────────────────────
     from studio.backend.run import _shutdown_event, _graceful_shutdown, _server

From 2ef98d382e0e38603b18854cc7b2451859bdfd3c Mon Sep 17 00:00:00 2001
From: Ricardo-M-L <69202550+Ricardo-M-L@users.noreply.github.com>
Date: Tue, 5 May 2026 18:48:04 +0800
Subject: [PATCH 204/216] fix: use % 8 instead of // 8 in FP8 weight shape
 check (#5243)

* fix: preserve bf16 GGUF file when explicitly requested in quantization list

When users request multiple quantization methods including the base format
(e.g., ["q4_k_m", "bf16"]), the bf16 GGUF serves as both the intermediate
conversion and a user-requested output. The cleanup step unconditionally
deleted this file, losing the explicitly requested bf16 output.

Only delete the intermediate base GGUF when the user did not request it.

Fixes #4932

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

* fix: keep reverse() outside conditional deletion to preserve VLM ordering

Address review feedback: the reverse() call must always execute when
quants_created is True to maintain correct [text_model, mmproj] ordering
for VLMs. Only the file deletion should be conditional on whether the
user requested the base format.

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Signed-off-by: Ricardo-M-L <ricardoporsche001@icloud.com>

* fix: move preserved base GGUF away from list boundaries for correct example commands

Address review from @Datta0: when the base format (e.g. bf16) is kept
in all_saved_locations, it could end up at [-1], causing the VLM example
command to use bf16 as --mmproj instead of the actual projector file.

Move the preserved base file to index 1 (after the primary quantized
model, before mmproj) so [0] and [-1] remain correct for example
commands.

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Signed-off-by: Ricardo-M-L <ricardoporsche001@icloud.com>

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

* fix: use % 8 instead of // 8 in FP8 weight shape check

weight.shape[X] // 8 != 0 is True for any non-zero dimension, causing
incorrect fallback to dequantization for small weights. Using % 8
correctly checks non-divisibility: weights not divisible by 8 should
dequantize, while those with % 8 == 0 stay on the fast kernel path.

* Preserve sharded base GGUF files during cleanup

convert_to_gguf can return multiple base text shards plus an mmproj
entry when llama.cpp splits the output. The previous cleanup only
removed/repositioned base_gguf=initial_files[0]:

- when the base format is NOT in quantization_method, sibling shards
  were left both in all_saved_locations and on disk as orphans
- when the base IS preserved, the reverse + insert(1, base_gguf)
  step left a sibling base shard at all_saved_locations[-1] for VLMs,
  so the example llama-mtmd-cli command ended up with --mmproj
  pointing at a text shard instead of the projector

Treat every initial file whose basename does not contain "-mmproj"
as part of the base set, then remove/unlink or reposition all of
them together. Drop the redundant frozenset() construction at both
call sites and the dead `base_gguf in all_saved_locations` clause
in the reorder guard.

* Apply bias in FP8 dequant fallback and dedupe full-precision flag

unsloth/kernels/fp8.py:
  FbgemmFp8Linear_matmul.forward had a dequant fallback that called
  torch_matmul without adding bias. The fast row-wise branch and the
  block FP8 branch both apply `output = output + bias if bias is not
  None else output` immediately after the matmul; the fallback now
  matches. This silently dropped bias for any FP8 layer routed to the
  fallback (Qwen 2.5 VL gate/up_proj 3420x1280, transposed-weight
  backward dispatch, and the small-shape cases newly routed here by
  the recent `% 8` divisibility fix).

unsloth/save.py:
  preserved_base inside the cleanup block and want_full_precision below
  it computed the identical expression `first_conversion in
  quantization_method`. Hoist want_full_precision above the cleanup
  block, reuse it for the not-preserved deletion and the preserved
  reposition, and assign True directly in the GPT-OSS branch.

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

---------

Signed-off-by: Ricardo-M-L <ricardoporsche001@icloud.com>
Co-authored-by: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com>
Co-authored-by: Daniel Han <danielhanchen@gmail.com>
Co-authored-by: Lee Jackson <130007945+Imagineer99@users.noreply.github.com>
---
 unsloth/kernels/fp8.py |  3 ++-
 unsloth/save.py        | 26 ++++++++++++++++++++------
 2 files changed, 22 insertions(+), 7 deletions(-)

diff --git a/unsloth/kernels/fp8.py b/unsloth/kernels/fp8.py
index a57f4ffb64..f704d4e3c4 100644
--- a/unsloth/kernels/fp8.py
+++ b/unsloth/kernels/fp8.py
@@ -432,13 +432,14 @@ class FbgemmFp8Linear_matmul(torch.autograd.Function):
         elif (
             weight.shape[0] != weight_scale.shape[0]
             and weight.shape[1] == weight_scale.shape[0]
-        ) or (weight.shape[0] // 8 != 0 or weight.shape[1] // 8 != 0):
+        ) or (weight.shape[0] % 8 != 0 or weight.shape[1] % 8 != 0):
             # Either the weight/scale is transposed or its shape is not divisible by 8. Both cases, dequantizing is the preferred way.
             # The transpose case is generally noticed in backward pass when we do dY@W instead of @W.T as we do for forward.
             # The shape case, I noticed to happen in MLP of Qwen 2.5 VL 7B where the gate proj is of shape (3420, 1280) and 3420/8=427.5
 
             W_deq = weight_dequant(weight, weight_scale).T
             output = torch_matmul(x, W_deq)
+            output = output + bias if bias is not None else output
             del W_deq
         else:
             raise ValueError(
diff --git a/unsloth/save.py b/unsloth/save.py
index df3df13a04..472eb14933 100644
--- a/unsloth/save.py
+++ b/unsloth/save.py
@@ -1552,19 +1552,33 @@ def save_to_gguf(
                             f"Error: {e}"
                         )
         print("Unsloth: Model files cleanup...")
+        want_full_precision = first_conversion in quantization_method
         if quants_created:
-            all_saved_locations.remove(base_gguf)
-            Path(base_gguf).unlink(missing_ok = True)
+            # convert_to_gguf may return multiple base shards plus an mmproj entry,
+            # so treat every initial file that is not an mmproj as part of the base set.
+            base_files = [
+                f for f in initial_files if "-mmproj" not in os.path.basename(f).lower()
+            ]
+            if not want_full_precision:
+                for f in base_files:
+                    if f in all_saved_locations:
+                        all_saved_locations.remove(f)
+                    Path(f).unlink(missing_ok = True)
 
             # flip the list to get [text_model, mmproj] order. for text models stays the same.
             all_saved_locations.reverse()
+
+            # When the base format is preserved, move base files (incl. shards) away from
+            # list boundaries so example commands ([0]=model, [-1]=mmproj) stay correct.
+            if want_full_precision and len(all_saved_locations) > len(base_files) + 1:
+                for f in base_files:
+                    if f in all_saved_locations:
+                        all_saved_locations.remove(f)
+                for i, f in enumerate(base_files):
+                    all_saved_locations.insert(1 + i, f)
     else:
         print("Unsloth: GPT-OSS model - skipping additional quantizations")
-
-    if is_gpt_oss:
         want_full_precision = True
-    else:
-        want_full_precision = first_conversion in frozenset(quantization_method)
 
     print(f"Unsloth: All GGUF conversions completed successfully!")
     print(f"Generated files: {all_saved_locations}")

From b39f4b282ac58d8818ee366a7dfde498bd6c4fcc Mon Sep 17 00:00:00 2001
From: DoubleMathew <mmathew23@gmail.com>
Date: Tue, 5 May 2026 06:03:28 -0500
Subject: [PATCH 205/216] Pin Studio GGUF export to llama.cpp's local convert
 script (#5275)

* Pin Studio GGUF export to local llama.cpp convert script

setdefault UNSLOTH_LLAMA_CPP_SCRIPTS_DIR=LLAMA_CPP_DEFAULT_DIR before
save_pretrained_gguf so the convert_hf_to_gguf.py used at conversion
time matches the pinned llama-quantize binary and gguf-py installed
under ~/.unsloth/llama.cpp. Without this, the script is pulled from
upstream master and can drift past the binary's gguf API, causing
intermittent export failures.

setdefault preserves any explicit user override; validation of the
path lives in unsloth_zoo's _resolve_local_convert_script (warns and
falls back to network on a bad value).

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

* Scrub .github/workflows for staging push (matches staging base)

* Pin GGUF convert script for hub-only export path

Hoist the UNSLOTH_LLAMA_CPP_SCRIPTS_DIR setdefault and the
unsloth_zoo.llama_cpp import out of the if save_directory: block so
push_to_hub_gguf also runs with the pin. The worker passes
save_directory="" for hub-only exports, which previously skipped the
local branch and left the convert script fetched from master.

* Trim GGUF convert script pin rationale comment

Collapse 7 lines of rationale into 3 lines stating the load-bearing
facts: pin matches llama-quantize binary, set before both branches
because hub-only export has empty save_directory.

* Sync .github/workflows with upstream author branch

* Scrub .github/workflows for staging push (matches staging base)

* Warn when unsloth_zoo is too old to honor UNSLOTH_LLAMA_CPP_SCRIPTS_DIR

Studio's GGUF export sets UNSLOTH_LLAMA_CPP_SCRIPTS_DIR before
save_pretrained_gguf and push_to_hub_gguf so unsloth_zoo can prefer the
local pinned convert_hf_to_gguf.py. The resolver only exists in the
companion unsloth_zoo change; on older zoo builds permitted by the
current dependency floor, the env var is silently ignored and the
converter is still downloaded from llama.cpp master.

Probe for the resolver and emit a one-time warning so operators know the
pin is inactive and can upgrade unsloth_zoo.

* Combine the GGUF script-pin imports into one guarded block and warn once

Both LLAMA_CPP_DEFAULT_DIR and the resolver probe come from
unsloth_zoo.llama_cpp; older zoo wheels (e.g. 2026.1.4) lack
LLAMA_CPP_DEFAULT_DIR, so the previous unguarded import could crash the
GGUF export path on environments installed with --no-deps or a manually
pinned zoo. Move the constant import alongside the resolver probe inside
a single try/except ImportError so a missing symbol degrades to the
warning instead of a hard crash, matching the graceful-degradation
intent the probe was added for.

The compatibility warning previously fired on every export call because
'from X import Y' re-raises ImportError on every invocation when Y is
absent. Gate emission on a module-level flag so operators see it once
per process instead of once per export.

* Add Studio GGUF export script-pin test coverage

Consolidate tests for the UNSLOTH_LLAMA_CPP_SCRIPTS_DIR env-var pin in
ExportBackend.export_gguf into a single behavior-named module:

- AST-asserts the module-level _LLAMA_CPP_SCRIPTS_WARNING_EMITTED flag,
  the merged try-block importing both LLAMA_CPP_DEFAULT_DIR and
  _resolve_local_convert_script, and the warn-once gate inside the
  ImportError handler.
- Behaviorally verifies setdefault preserves explicit user overrides,
  assigns the default when unset, fires the compatibility warning at
  most once across multiple export calls, and degrades to a warning
  (without setting the env var) when LLAMA_CPP_DEFAULT_DIR itself is
  missing on an older unsloth_zoo.

* Sync .github/workflows with upstream author branch

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

---------

Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com>
Co-authored-by: Daniel Han <danielhanchen@gmail.com>
Co-authored-by: Lee Jackson <130007945+Imagineer99@users.noreply.github.com>
---
 studio/backend/core/export/export.py          |  27 ++
 .../test_studio_gguf_export_script_pin.py     | 231 ++++++++++++++++++
 2 files changed, 258 insertions(+)
 create mode 100644 tests/studio/test_studio_gguf_export_script_pin.py

diff --git a/studio/backend/core/export/export.py b/studio/backend/core/export/export.py
index d8f2e8fa37..6fee5a38f7 100644
--- a/studio/backend/core/export/export.py
+++ b/studio/backend/core/export/export.py
@@ -28,6 +28,8 @@ from core.inference import get_inference_backend
 
 logger = get_logger(__name__)
 
+_LLAMA_CPP_SCRIPTS_WARNING_EMITTED = False
+
 
 def _is_wsl():
     """Detect if running under Windows Subsystem for Linux."""
@@ -529,6 +531,31 @@ class ExportBackend:
             # Convert quantization method to lowercase for unsloth
             quant_method = quantization_method.lower()
 
+            # Pin convert_hf_to_gguf.py to the same llama.cpp ref as the
+            # llama-quantize binary (Studio installs at a tagged ref via
+            # setup.sh) so it can't drift past the pinned binary's gguf API.
+            # Set before both branches; hub-only export has save_directory == "".
+            global _LLAMA_CPP_SCRIPTS_WARNING_EMITTED
+            try:
+                from unsloth_zoo.llama_cpp import (
+                    LLAMA_CPP_DEFAULT_DIR,
+                    _resolve_local_convert_script,  # noqa: F401
+                )
+
+                os.environ.setdefault(
+                    "UNSLOTH_LLAMA_CPP_SCRIPTS_DIR", LLAMA_CPP_DEFAULT_DIR
+                )
+            except ImportError:
+                if not _LLAMA_CPP_SCRIPTS_WARNING_EMITTED:
+                    logger.warning(
+                        "Unsloth: installed unsloth_zoo does not honor "
+                        "UNSLOTH_LLAMA_CPP_SCRIPTS_DIR; convert_hf_to_gguf.py will "
+                        "still be downloaded from llama.cpp master and may drift "
+                        "past the pinned llama-quantize binary. Upgrade unsloth_zoo "
+                        "to activate the local script pin."
+                    )
+                    _LLAMA_CPP_SCRIPTS_WARNING_EMITTED = True
+
             # Save locally if requested
             if save_directory:
                 save_directory = str(resolve_export_dir(save_directory))
diff --git a/tests/studio/test_studio_gguf_export_script_pin.py b/tests/studio/test_studio_gguf_export_script_pin.py
new file mode 100644
index 0000000000..12aaaafff4
--- /dev/null
+++ b/tests/studio/test_studio_gguf_export_script_pin.py
@@ -0,0 +1,231 @@
+"""Tests for Studio GGUF export pinning convert_hf_to_gguf.py via
+UNSLOTH_LLAMA_CPP_SCRIPTS_DIR with graceful fallback when unsloth_zoo
+lacks the local-script resolver.
+
+Verifies:
+  - export.py imports LLAMA_CPP_DEFAULT_DIR and _resolve_local_convert_script
+    from unsloth_zoo.llama_cpp inside a single try/except ImportError so a
+    zoo missing either symbol degrades to a warning instead of crashing.
+  - os.environ.setdefault("UNSLOTH_LLAMA_CPP_SCRIPTS_DIR", LLAMA_CPP_DEFAULT_DIR)
+    is called inside the try; setdefault preserves explicit user overrides
+    and assigns the default when unset.
+  - The compatibility warning is gated on a module-level flag so it fires
+    once per process rather than on every export call.
+"""
+
+from __future__ import annotations
+
+import ast
+import os
+import sys
+import types
+from pathlib import Path
+
+
+SOURCE_PATH = (
+    Path(__file__).resolve().parents[2]
+    / "studio"
+    / "backend"
+    / "core"
+    / "export"
+    / "export.py"
+)
+SRC = SOURCE_PATH.read_text()
+TREE = ast.parse(SRC)
+
+
+def _module_level_assignments(tree: ast.Module):
+    for node in tree.body:
+        if isinstance(node, ast.Assign):
+            for target in node.targets:
+                if isinstance(target, ast.Name):
+                    yield target.id, node.value
+
+
+def _find_pin_try(tree: ast.AST):
+    for node in ast.walk(tree):
+        if not isinstance(node, ast.Try):
+            continue
+        for stmt in node.body:
+            if (
+                isinstance(stmt, ast.ImportFrom)
+                and stmt.module == "unsloth_zoo.llama_cpp"
+                and any(
+                    alias.name == "_resolve_local_convert_script"
+                    for alias in stmt.names
+                )
+            ):
+                return node
+    return None
+
+
+def test_warning_flag_defined_at_module_scope():
+    flags = {
+        name: value
+        for name, value in _module_level_assignments(TREE)
+        if name == "_LLAMA_CPP_SCRIPTS_WARNING_EMITTED"
+    }
+    assert flags, "expected module-level _LLAMA_CPP_SCRIPTS_WARNING_EMITTED flag"
+    init = flags["_LLAMA_CPP_SCRIPTS_WARNING_EMITTED"]
+    assert isinstance(init, ast.Constant) and init.value is False
+
+
+def test_constant_and_resolver_imported_in_same_try():
+    try_node = _find_pin_try(TREE)
+    assert try_node is not None
+    imported = []
+    for stmt in try_node.body:
+        if isinstance(stmt, ast.ImportFrom) and stmt.module == "unsloth_zoo.llama_cpp":
+            imported.extend(alias.name for alias in stmt.names)
+    assert "LLAMA_CPP_DEFAULT_DIR" in imported
+    assert "_resolve_local_convert_script" in imported
+
+
+def test_setdefault_inside_try_block():
+    try_node = _find_pin_try(TREE)
+    assert try_node is not None
+    setdefault_calls = []
+    for stmt in try_node.body:
+        for node in ast.walk(stmt):
+            if (
+                isinstance(node, ast.Call)
+                and isinstance(node.func, ast.Attribute)
+                and node.func.attr == "setdefault"
+                and isinstance(node.func.value, ast.Attribute)
+                and node.func.value.attr == "environ"
+                and node.args
+                and isinstance(node.args[0], ast.Constant)
+                and node.args[0].value == "UNSLOTH_LLAMA_CPP_SCRIPTS_DIR"
+            ):
+                setdefault_calls.append(node)
+    assert setdefault_calls
+    second = setdefault_calls[0].args[1]
+    assert isinstance(second, ast.Name) and second.id == "LLAMA_CPP_DEFAULT_DIR"
+
+
+def test_warning_handler_gated_on_module_flag():
+    try_node = _find_pin_try(TREE)
+    assert try_node is not None
+    handlers = [
+        h
+        for h in try_node.handlers
+        if isinstance(h.type, ast.Name) and h.type.id == "ImportError"
+    ]
+    assert handlers
+    handler = handlers[0]
+    flag_reads = []
+    flag_writes = []
+    warning_calls = []
+    for node in ast.walk(ast.Module(body = handler.body, type_ignores = [])):
+        if (
+            isinstance(node, ast.Name)
+            and node.id == "_LLAMA_CPP_SCRIPTS_WARNING_EMITTED"
+        ):
+            if isinstance(node.ctx, ast.Load):
+                flag_reads.append(node)
+            elif isinstance(node.ctx, ast.Store):
+                flag_writes.append(node)
+        elif (
+            isinstance(node, ast.Call)
+            and isinstance(node.func, ast.Attribute)
+            and node.func.attr == "warning"
+        ):
+            warning_calls.append(node)
+    assert flag_reads
+    assert flag_writes
+    assert warning_calls
+    msg = ast.dump(warning_calls[0])
+    assert "UNSLOTH_LLAMA_CPP_SCRIPTS_DIR" in msg
+    assert "unsloth_zoo" in msg
+
+
+def test_default_dir_is_string_for_setdefault_compat():
+    from unsloth_zoo.llama_cpp import LLAMA_CPP_DEFAULT_DIR
+
+    assert isinstance(LLAMA_CPP_DEFAULT_DIR, str)
+
+
+def test_setdefault_preserves_explicit_user_override(monkeypatch):
+    monkeypatch.setenv("UNSLOTH_LLAMA_CPP_SCRIPTS_DIR", "/explicit/override")
+    from unsloth_zoo.llama_cpp import LLAMA_CPP_DEFAULT_DIR
+
+    os.environ.setdefault("UNSLOTH_LLAMA_CPP_SCRIPTS_DIR", LLAMA_CPP_DEFAULT_DIR)
+    assert os.environ["UNSLOTH_LLAMA_CPP_SCRIPTS_DIR"] == "/explicit/override"
+
+
+def test_setdefault_assigns_default_when_unset(monkeypatch):
+    monkeypatch.delenv("UNSLOTH_LLAMA_CPP_SCRIPTS_DIR", raising = False)
+    from unsloth_zoo.llama_cpp import LLAMA_CPP_DEFAULT_DIR
+
+    os.environ.setdefault("UNSLOTH_LLAMA_CPP_SCRIPTS_DIR", LLAMA_CPP_DEFAULT_DIR)
+    assert os.environ["UNSLOTH_LLAMA_CPP_SCRIPTS_DIR"] == LLAMA_CPP_DEFAULT_DIR
+
+
+def _simulate_pin_block(emit_records, set_value):
+    fake = types.ModuleType("unsloth_zoo.llama_cpp")
+    if set_value is not None:
+        fake.LLAMA_CPP_DEFAULT_DIR = set_value
+    sys.modules["unsloth_zoo.llama_cpp"] = fake
+
+    state = {"emitted": False}
+
+    def run_once():
+        try:
+            from unsloth_zoo.llama_cpp import (
+                LLAMA_CPP_DEFAULT_DIR,
+                _resolve_local_convert_script,  # noqa: F401
+            )
+
+            os.environ.setdefault(
+                "UNSLOTH_LLAMA_CPP_SCRIPTS_DIR", LLAMA_CPP_DEFAULT_DIR
+            )
+        except ImportError:
+            if not state["emitted"]:
+                emit_records.append("warned")
+                state["emitted"] = True
+
+    return run_once
+
+
+def test_warning_fires_at_most_once_across_calls(monkeypatch):
+    monkeypatch.delenv("UNSLOTH_LLAMA_CPP_SCRIPTS_DIR", raising = False)
+    emits = []
+    runner = _simulate_pin_block(emits, set_value = "/fake/default")
+    runner()
+    runner()
+    runner()
+    assert emits == ["warned"]
+
+
+def test_missing_default_dir_degrades_to_warning(monkeypatch):
+    monkeypatch.delenv("UNSLOTH_LLAMA_CPP_SCRIPTS_DIR", raising = False)
+    emits = []
+    runner = _simulate_pin_block(emits, set_value = None)
+    runner()
+    assert emits == ["warned"]
+    assert "UNSLOTH_LLAMA_CPP_SCRIPTS_DIR" not in os.environ
+
+
+def test_no_warning_when_both_symbols_present(monkeypatch):
+    monkeypatch.delenv("UNSLOTH_LLAMA_CPP_SCRIPTS_DIR", raising = False)
+    fake = types.ModuleType("unsloth_zoo.llama_cpp")
+    fake.LLAMA_CPP_DEFAULT_DIR = "/fake/dir"
+    fake._resolve_local_convert_script = lambda: None
+    monkeypatch.setitem(sys.modules, "unsloth_zoo.llama_cpp", fake)
+
+    emits = []
+    state = {"emitted": False}
+    try:
+        from unsloth_zoo.llama_cpp import (
+            LLAMA_CPP_DEFAULT_DIR,
+            _resolve_local_convert_script,  # noqa: F401
+        )
+
+        os.environ.setdefault("UNSLOTH_LLAMA_CPP_SCRIPTS_DIR", LLAMA_CPP_DEFAULT_DIR)
+    except ImportError:
+        if not state["emitted"]:
+            emits.append("warned")
+            state["emitted"] = True
+
+    assert emits == []
+    assert os.environ.get("UNSLOTH_LLAMA_CPP_SCRIPTS_DIR") == "/fake/dir"

From 6b13cab74651528b4e57e2b86e0803cf2a0e21c6 Mon Sep 17 00:00:00 2001
From: Datta Nimmaturi <venkatadattasainimmaturi@gmail.com>
Date: Tue, 5 May 2026 16:36:46 +0530
Subject: [PATCH 206/216] fix KVCache estimates for gemma4 style sliding window
 models (#5225)

* fix KVCache estimates for gemma4 style sliding window models

Signed-off-by: Datta Nimmaturi <venkatadattasainimmaturi@gmail.com>

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

* studio: add per-arch SWA pattern fallback + n_kv_heads mirror for PR #5225

The pattern-aware SWA estimator added in this PR only fires when the
GGUF carries `<arch>.attention.sliding_window_pattern`. Today's
Gemma-2 / Gemma-3 / Gemma-3n / gpt-oss / Phi-3 GGUFs ship
`attention.sliding_window` but not the pattern field (llama.cpp's
converter strips it), so the new branch is bypassed and we fall back to
the legacy 1/4-global heuristic on the most popular SWA arches in our
catalogue (gemma3 alone has 6+ variants in the unsloth/* top 30 by
downloads, plus gpt-oss-20b/120b).

Two additions on top of this PR:

1. `_SWA_PATTERN_DEFAULTS_BY_ARCH` table keyed by GGUF arch name. When
   the GGUF reports a sliding window but no pattern, we synthesise the
   pattern from the architecture's canonical period (gemma2=2,
   gemma3=6, gemma3n=5, gpt_oss=2, phi3=1, cohere2=4). Periods sourced
   from a survey of the top 150 unsloth/* HF configs against
   `text_config.layer_types` and `Gemma*Config.sliding_window_pattern`.

2. Mirror `_n_kv_heads_by_layer` into the scalar `_n_kv_heads` (using
   max as a conservative upper bound) when the head_count_kv array is
   read. Without this, any non-SWA estimator path (GQA, legacy) on a
   Gemma-4-style model falls through to `n_heads`, which can be many
   times larger than the real per-layer KV head count. Also let
   `_can_estimate_kv` accept the array directly as belt-and-suspenders.

End-to-end check on `unsloth/gemma-3-270m-it-Q4_K_M.gguf` (18 layers,
sliding_window=512, no pattern field): the parser now resolves the
pattern to period=6 (3 global, 15 SWA), matching the actual
Gemma3TextConfig default. KV estimate at 32k context drops from
141 MB (legacy 1/4) to 108 MB (per-layer), a 23% reduction that
directly translates into more headroom for `_fit_context_to_vram` and
fewer cases where the slider lands on the 4096 floor.

Tests: extended `test_kv_cache_estimation.py` with
`TestArchSwaPatternDefaults` covering the six tabled arches, an
unknown-arch negative, explicit-pattern precedence, and a
no-sliding-window negative; updated `test_array_fields_parsed` to
reflect the new mirror semantics; updated
`test_end_to_end_synthetic_swa` to use the period=6 expectation. All
102 tests in the kv-cache / context-fit / max-context suites pass.

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

* studio: tighten arch SWA table to verified non-regressing entries

Audit of every unsloth/* HF model (1334 repos, all config.json fetched
in scripts/survey_all_unsloth.py) plus end-to-end checks against five
real GGUFs (gemma-3-270m, gemma-3-1b, qwen2.5-0.5b, phi-3.5-mini,
falcon-h1-0.5b, granite-4.1-8b) confirms:

* Pure-GQA arches (llama, qwen3, mistral3, glm4, llama4, ...) and the
  qwen2 family with use_sliding_window=False all reach Path 4 GQA
  cleanly. The llama.cpp converter strips `attention.sliding_window`
  for qwen2/qwen2_vl/qwen2_5_vl when use_sliding_window=False, so the
  SWA path never fires for them. Verified on Qwen2.5-0.5B-GGUF: no
  sliding_window field in metadata.
* MLA arches (deepseek_v3/v32/v4, glm4_moe_lite, glm_moe_dsa, kimi_k25)
  emit `kv_lora_rank` -> Path 1 fires correctly.
* Hybrid Mamba/Attn arches that emit both ssm.* and
  full_attention_interval (qwen3_5, qwen3_5_moe, qwen3_next) -> Path 2
  fires correctly.

Two table changes:

1. Drop the `phi3` entry. Phi-3 GGUFs emit
   `phi3.attention.sliding_window=262144` but never emit
   `attention.key_length`/`value_length`, so the SWA path is gated
   off and the estimator falls to the legacy formula. The huge
   sliding_window also means SWA layers and global layers cache
   identical numbers of tokens at any practical context, so a fallback
   would be a no-op anyway. The previous `phi3: 1` entry was also
   semantically wrong: period=1 with the (i+1)%N!=0 rule produces
   all-global, not the all-SWA you'd want for Phi-3.
2. Document the audit findings in the table comment, including the
   two arches we deliberately skip (phi3, qwen2*) and the one
   architecture family that is not a regression vs. main but is also
   not yet optimal (mistral v0.1/v0.2 all-SWA every-layer cannot be
   expressed with the period sentinel).

Tests: added `test_non_swa_arch_uses_full_attention_path` parametrized
over llama / qwen2 / qwen3 / mistral / mistral3 / glm4 / llama4 to
pin the invariant that pure-GQA arches never receive a synthetic SWA
pattern. Removed phi3 from the parametrize list of
`test_arch_default_pattern_applied`. All 108 tests pass.

Known separately tracked (not addressed here): falcon-h1 GGUFs ship
ssm.* + key_length but no full_attention_interval, so the hybrid
path 2 cannot fire and the estimator falls to GQA path 4, which
counts every block as an attention layer. Affects 8 unsloth/* repos
(~500 downloads). Same gap exists for granitemoehybrid-class GGUFs.
Worth a follow-up that adds either a HYBRID_ATTENTION_INTERVAL_BY_ARCH
table or a tensor-name probe.

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

* studio: make SWA pattern resolver dynamic so new models work without code changes

The static `_SWA_PATTERN_DEFAULTS_BY_ARCH` dict only covered
architectures we knew about at PR-merge time. New SWA models would
need a code change here every time a new arch shipped, which doesn't
scale. Replaced with a 4-tier resolver so any newly-released model
that lands on Hugging Face with a normal `config.json` is covered
automatically:

  Tier 0 (parser)  -- explicit GGUF metadata if the converter emits it
                      (BOOL array or scalar period). Already supported.
  Tier 1 (cache)   -- $UNSLOTH_STUDIO_HOME/swa_cache.json. Populated by
                      previous Tier 3 fetches. Survives restarts.
  Tier 2 (bootstrap) -- `_BOOTSTRAP_SWA_DEFAULTS` shipped with Studio.
                      Same five entries as the old static table
                      (gemma2/3/3n/gpt_oss/cohere2). Lets fully-offline
                      installs keep working for popular SWA arches
                      with zero network.
  Tier 3 (HF fetch)  -- pulls `config.json` from the GGUF's source HF
                      repo and reads `sliding_window_pattern` (int) or
                      `text_config.layer_types` (string array). Result
                      is cached to Tier 1 so subsequent loads are
                      offline-fast. Disabled by
                      `UNSLOTH_STUDIO_OFFLINE=1`. Network errors and
                      missing repos fall through silently.
  Tier 4 (caller)  -- legacy 1/4-global SWA estimate (unchanged).

The GGUF parser now also extracts a handful of `general.*` keys
(`source.huggingface.repository`, `source.url`, `source.repo_url`,
`base_model.0.repo_url`, `base_model.0.organization` + `.name`,
`organization` + `basename`) so the resolver has source-repo
candidates to try.

End-to-end smoke against a brand-new arch (`never_seen_before_arch`,
not in the bootstrap dict) pointing at `google/gemma-3-1b-it`:
resolver fetched the HF config, derived period=6, materialised the
26-layer mask with 4 global layers (indices 5/11/17/23), and wrote
`{"never_seen_before_arch": 6}` to the on-disk cache. Next load hits
Tier 1 with no network.

Tests: added `TestDynamicSwaResolver` with 10 tests covering each
tier (period derivation, aperiodic mask handling, URL parsing,
bootstrap precedence, cache precedence, HF fetch + persistence,
candidate fallback, offline env knob, network failure). All 118
tests in the kv-cache / context-fit / max-context suites pass.

The `_SWA_PATTERN_DEFAULTS_BY_ARCH` name was retired in favour of
`_BOOTSTRAP_SWA_DEFAULTS` to make the tier semantics explicit.

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

* studio: add Tier 2.5 transformers introspection to SWA resolver

Slots a new tier between bootstrap and HF fetch that asks the
locally-installed `transformers` package directly. Two strategies, in
order, both offline-friendly:

  a. Default-instantiate the matching `Config` class via
     `CONFIG_MAPPING[arch]()` and read `sliding_window_pattern` /
     `text_config.layer_types`. Drills into `text_config` for
     multimodal wrappers.
  b. `inspect.getsource(cfg_class)` regex-parse for
     `sliding_window_pattern: int = N` defaults. Catches configs
     whose constructor raises (missing required args), or where the
     default is bound only in the __init__ signature. Walks
     `cfg_class.sub_configs["text_config"]` too so multimodal wrappers
     that delegate to a TextConfig still get inspected.

Resolver chain is now 5 tiers: GGUF metadata, on-disk cache,
bootstrap defaults, transformers introspection, HF Hub fetch, legacy
fallback. Tier 2.5 results are persisted to the same on-disk cache as
Tier 3 so subsequent loads skip the import overhead.

`_arch_aliases` normalises hyphen vs underscore variants (`falcon-h1`
vs `falcon_h1`) since GGUF and HF disagree for a handful of arches.

Cross-version verification (probe at `temp/swa_probe/`):

```
arch       transformers 4.57.6      transformers 5.7.0
gemma3     6                        6
gemma2     2                        2
cohere2    4                        4
gpt_oss    2                        2
gemma3n    5                        5
gemma4     ARCH-MISSING             6        <- new arch picked up automatically
falcon_h1  None                     [True]*32 <- per-layer mask used verbatim
phi3       None                     None
mistral    None                     None
qwen2      1 (all-global)           1
llama      None                     None
deepseek_v3 None                    None
```

The `gemma4` and `falcon_h1` rows are the headline: a brand-new arch
that lands in transformers (gemma4 is 5.x-only) is supported by the
resolver the moment a user upgrades the package, with zero edits to
this file. Same applies to any future arch with a `Config` class.

Tests: added `TestTransformersIntrospection` with 6 cases covering
arch-alias normalisation, real-arch resolution against the live
transformers, inspect.getsource fallback when default-init raises,
graceful behaviour when transformers is unavailable, unknown-arch
returns None, and Tier-2.5-before-Tier-3 ordering. Also adjusted the
existing Tier 3 failure test to mock Tier 2.5 out so it specifically
exercises the network-failure path. All 124 tests in the kv-cache /
context-fit / max-context suites pass.

Updated the module-level resolver comment from "4-tier" to "5-tier"
to document the new tier.

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

* studio: consolidate verbose comments and docstrings in SWA resolver

Net -345 lines: -210 in llama_cpp.py, -357/+111 in
test_kv_cache_estimation.py. No behaviour change; only comments,
docstrings, and one tests-only `_SWA_FIELDS` helper to remove the
copy-pasted GGUF metadata dict from each resolver test.

Code changes only delete or shorten:
  * 5-tier resolver header collapsed from a 38-line block diagram to
    a 9-line summary; the rest is the function bodies.
  * Bootstrap dict per-arch comments collapsed to one-line `Config`
    references.
  * `_swa_cache_path`, `_save_swa_cache`, `_period_from_layer_types`,
    `_arch_aliases`, `_swa_entry_from_config_obj`,
    `_resolve_swa_pattern`, `_resolve_swa_entry_from_transformers`
    docstrings stripped to one line or removed when the body is
    self-evident.
  * Tier-by-tier inline comments inside `_resolve_swa_pattern` removed
    (function body reads top-to-bottom in tier order).
  * Path-3 SWA estimator comment shortened from a 12-line tier
    breakdown to 3 lines.
  * Parser fallback comment block (originally explained the resolver
    in-line) trimmed to two lines pointing at the resolver.
  * `_can_estimate_kv` legacy-clause comment shortened to one line.
  * GGUF `general.*` WANTED block comment shortened to one line.

Test changes:
  * Per-test docstrings dropped where the test name and body already
    explain intent.
  * Class-level docstrings reduced to one line.
  * Common GGUF field dict factored to module-level `_SWA_FIELDS`.
  * Multi-line URL/list assertions collapsed to one-liners.

All 124 tests pass.

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

* studio: account for SWA cache double-buffering in path 3 estimate

Cross-check against llama.cpp ground truth (running llama-server with
--parallel 1 and reading the `llama_kv_cache: size = X MiB ( N cells,
M layers, ... )` log lines) showed the SWA path under-counted by
~20% on Gemma-3-shaped models:

  GGUF                                pred MiB  actual MiB  ratio
  gemma-3-270m-it-Q4_K_M               31.50    39.00       0.81
  gemma-3-1b-it-Q2_K                   43.00    54.00       0.80

Root cause: llama.cpp double-buffers the SWA cache so it can keep the
current and next windows during the shift, allocating
`2 * sliding_window` cells per SWA layer (capped at n_ctx). My formula
was using `min(n_ctx, sliding_window)` instead of
`min(n_ctx, 2 * sliding_window)`. Verified directly:

  llama_kv_cache_iswa: creating SWA KV cache, size = 1024 cells
  llama_kv_cache: size = 15.00 MiB (1024 cells, 15 layers, ...)

with `gemma3.attention.sliding_window = 512` -> 2 * 512 = 1024 cells.

Fix: introduce `swa_cells = min(n_ctx, 2 * swa)` in path 3 and use
that for both the per-layer-pattern branch and the legacy
1/4-global fallback.

Re-run after the fix:

  GGUF                                pred MiB  actual MiB  ratio
  gemma-3-270m-it-Q4_K_M               39.00    39.00       1.000
  gemma-3-1b-it-Q2_K                   54.00    54.00       1.000
  qwen2.5-0.5b-instruct-q4_k_m         96.00    96.00       1.000
  Phi-3.5-mini-instruct-Q4_K_M       3072.00  3072.00       1.000
  Falcon-H1-0.5B-Instruct-Q4_K_M      144.00   144.00       1.000
  granite-4.1-8b-Q3_K_M              1280.00  1280.00       1.000

All 5 paths now match llama.cpp's actual allocation exactly under
single-sequence inference (Studio's default).

Tests: updated `test_gemma3`, `test_gpt_oss`,
`test_gemma4_per_layer_swa_metadata`, `test_ctx_smaller_than_window`,
`test_odd_layer_count`, and `test_end_to_end_synthetic_swa` to use
the doubled SWA cell count. All 124 tests in the kv-cache /
context-fit / max-context suites pass.

* studio: tolerate truncated GGUF input so resolver fallback still runs

Wraps each iteration of the GGUF KV-pair loop in a try/except that
breaks out cleanly on `struct.error` or `UnicodeDecodeError`, instead
of letting the outer try eat the exception and skip the SWA resolver
fallback at the end.

The motivating use case is reading the GGUF metadata via an HF Hub
HTTP byte-range fetch. The first ~128 KiB of a typical GGUF contains
all the metadata we need (arch, block_count, attention.*, sliding
window, ssm, MLA fields, plus the tokenizer config) -- but for models
with large tokenizer vocabs (Gemma 3 has 262144 tokens) the tokenizer
arrays spill past the 128 KiB boundary. The truncation used to bubble
out as `unpack requires a buffer of 8 bytes`, abandoning the resolver
fallback and leaving us with no SWA pattern (so the SWA path fell
through to the legacy 1/4 estimate).

Verified end to end against `unsloth/gemma-3-1b-it-GGUF`:

  Range-fetch first 128 KiB of `gemma-3-1b-it-Q2_K.gguf` over HTTP
  (HTTP 206 Partial Content), parse:

    arch              = gemma3
    block_count       = 26
    attention.sliding_window = 512
    sliding_window_pattern   = set (4 global) <- via Tier 2 bootstrap
    KV @ ctx=8192            = 54.00 MiB     <- matches llama.cpp
                                              ground truth

This means Studio can preview KV-cache requirements (and therefore
auto-context fit) for any HF GGUF without downloading the weights.

All 124 tests pass.

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

* studio: thread llama-server KV flags through the estimator

Adds keyword-only knobs to _estimate_kv_cache_bytes and _fit_context_to_vram
that mirror the llama-server CLI options that change KV memory:

  --swa-full         (swa_full)        SWA layers cache the full n_ctx
                                       instead of 2 * sliding_window cells
  --parallel N       (n_parallel)      number of server slots
  --kv-unified       (kv_unified)      single shared KV buffer; when off,
                                       multiplies KV by n_parallel
  --ctx-checkpoints  (ctx_checkpoints) per-slot SWA snapshots, each one
                                       sliding-window of state per SWA layer
  --kv-offload       (kv_on_gpu)       when off, KV lives in CPU RAM and is
                                       not subtracted from the VRAM budget

Defaults preserve the previous behavior (swa_full=False, n_parallel=1,
kv_unified=True, ctx_checkpoints=0, kv_on_gpu=True) so existing call sites
are unaffected. All five paths (MLA, hybrid, SWA pattern, SWA fallback,
GQA, legacy) now apply the per-slot replication factor; the SWA paths
also honor swa_full and add the checkpoint term when applicable.

Tests: TestServerFlags (17 cases) covers every flag, the no-op cases, the
swa_full + ctx_checkpoints interaction, slot multiplication on each path,
and the kv_on_gpu shortcut in _fit_context_to_vram.

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

* studio: account for shared_kv_layers (Gemma 3n / Gemma 4)

Gemma 3n and Gemma 4 set <arch>.attention.shared_kv_layers in the GGUF
metadata (convert_hf_to_gguf.py lines 7578 and 7712). The trailing N
layers of the model reuse KV from earlier layers and don't allocate
their own cache, so n_layers in the per-layer formulas overcounted by
exactly that many blocks. For google/gemma-3n-E4B-it (35 layers, 15
shared), this puts ~43% of the KV estimate back on the table.

Changes:

- _read_gguf_metadata parses <arch>.attention.shared_kv_layers into
  self._shared_kv_layers; init / unload / reparse all reset it.
- _estimate_kv_cache_bytes computes n_layers_kv = max(1, n_layers -
  shared_kv_layers) and substitutes it for n_layers in:
    Path 1 (MLA), Path 3 (SWA pattern loop bound and the no-pattern
    fallback), Path 4 (GQA), Path 5 (legacy). Path 2 (hybrid) keeps
    n_layers since hybrid + shared_kv combined isn't a thing today and
    the semantics would need to specify which attention layers are
    shared.
- max(1, ...) floor protects against pathological GGUFs where shared
  >= n_layers.
- Composes naturally with --swa-full, --kv-unified / --parallel,
  --ctx-checkpoints, and the per-layer SWA pattern from the dynamic
  resolver. When the field is unset (every other arch) the math is
  byte-identical to before.

Tests: TestSharedKVLayers (13 cases) covers each path's drop, the
no-op-when-unset case, the floor at one layer, composition with the
server-flag knobs, and lifecycle reset. test_end_to_end_synthetic_shared_kv_round_trip
exercises the full GGUF parse -> estimate path on a synthetic gemma3n_text
blob. Existing TestLifecycle tests extended to cover the new field.

Full suite: 132 passing.

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

* studio: only stub httpx in tests when the real lib is missing

The unit suite stubs httpx unconditionally so tests can run on a minimal
Python install. Surfaced during a fresh-venv simulation: when the stub
is installed via setdefault on a system that DOES have httpx,
huggingface_hub.errors fails to import HTTPError / Response at module
load time, which the transformers introspection tier swallows via its
bare except. Result: TestTransformersIntrospection passes in venvs
where httpx happened to be imported first (workspace) and silently
fails in venvs where it doesn't (fresh uv venv).

Switch to "only stub when real lib unavailable", and round out the stub
with HTTPError, RequestError, and Response so any test environment
without httpx still gets a complete enough surface for huggingface_hub
to import.

* studio: per-layer-type --parallel N memory accounting for SWA

Empirical verification against llama-server (see
workspace_5/temp/sim_pr5225/probe_parallel_full_matrix.py and
verify_parallel_matches_server.py) showed the prior whole-cache
slot_factor multiplication in _estimate_kv_cache_bytes was wrong for
n_parallel > 1. The actual rule, verified bit-exact across the full
(parallel x ctx) grid for both SWA and pure-GQA models:

  * non-SWA layers: total cells = n_ctx, partitioned across slots
                    (per-slot ctx = n_ctx / parallel). Total memory
                    is CONSTANT in n_parallel.
  * SWA layers:     per-slot cells = 2 * sliding_window (clamped at
                    n_ctx and at per_slot_ctx when ctx is split among
                    many slots). Total memory grows LINEARLY in
                    n_parallel.
  * --kv-unified:   no measurable difference to total memory; both
                    modes yield the same byte total in measured cases.
                    Retained as accepted-but-ignored kwarg for API
                    forward-compat.

Closed form (Path 3 with per-layer pattern):
    total_kv = sum_global_layers(n_ctx * n_kv * (k+v) * bpe)
             + parallel * sum_swa_layers(
                   min(2*sliding_window, n_ctx, n_ctx//parallel)
                   * n_kv_layer * (k_swa + v_swa) * bpe
               )
    + parallel * checkpoint_extra_per_slot   (when ctx_checkpoints > 0)

Changes to _estimate_kv_cache_bytes:
- Path 3 (SWA pattern): accumulate global_bytes and swa_bytes_per_slot
  separately; final result = global_bytes + slots * (swa_bps + cp_bps).
- Path 3 (no-pattern fallback): same split using the 1/4-global heuristic.
- Paths 1 / 2 / 4 / 5: drop the slot_factor multiplication. Non-SWA
  caches don't scale with --parallel.
- swa_full=True: SWA cells = per_slot_ctx (was n_ctx), so slots
  cancels out and total stays constant. Matches llama-server's
  --swa-full --parallel N output exactly.

Production wiring fix in start():
- Seven internal calls to _estimate_kv_cache_bytes / _fit_context_to_vram
  used the default n_parallel=1, even though load_model accepts the
  caller's n_parallel value (forwarded to llama-server via --parallel
  on the command line). Pass n_parallel through all seven so VRAM
  budgeting is correct when an operator sets parallel slots above 1.
  Studio's default ships at 1 so production today is unaffected;
  this completes the wiring for operators who tune it.

Tests:
- TestParallelSWAScaling (10 new cases): closed-form invariants per
  path, swa_full + parallel collapse, kv_unified no-op proof,
  per-slot SWA cell clamping, and the empirical Gemma-3 270m formula
  (24 + parallel * 15 MiB at ctx=8192) baked from the verifier.
- TestServerFlags: rewrote 4 assertions and renamed 2 to reflect the
  per-layer rule; non-SWA paths now correctly assert constancy.
- TestSharedKVLayers::test_composes_with_n_parallel: rewrote to assert
  only the SWA portion of the unshared layers scales.

Backward compatibility: at n_parallel=1 the output is bit-identical to
before this change (verified across 120,960 sweep combinations and the
141-test suite in both workspace and fresh-uv-venv environments).
Verifier output at --parallel in {1,2,4,8} x ctx in {4096,8192,16384}
shows ratio 1.000 against llama-server for both SWA and pure-GQA
models (24/24 cells exact match).

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

* studio: accept new estimator kwargs in load-time test stubs

`test_llama_cpp_context_fit.py` and `test_llama_cpp_max_context_threshold.py`
patch `_estimate_kv_cache_bytes` with constant per-token stubs and then
call the real `_fit_context_to_vram`. After 29dcf96e threaded the
llama-server flag kwargs (`swa_full`, `n_parallel`, `kv_unified`,
`ctx_checkpoints`) through `_fit_context_to_vram`, the production method
forwards them to the stubbed estimator and the old positional-only stubs
raise `TypeError`.

These two suites exercise the load-time fit decision and the max-context
threshold property with a constant per-token KV cost; SWA / parallel-slot
accounting is intentionally out of scope, so the stubs absorb the new
kwargs and ignore them. No production change.

Restores both files to fully passing: 15/15 in `test_llama_cpp_context_fit`
and 8/8 in `test_llama_cpp_max_context_threshold`. Combined with the
existing 141/141 in `test_kv_cache_estimation`, the three KV-cache test
modules are 164/164 green.

---------

Signed-off-by: Datta Nimmaturi <venkatadattasainimmaturi@gmail.com>
Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com>
Co-authored-by: Lee Jackson <130007945+Imagineer99@users.noreply.github.com>
Co-authored-by: Roland Tannous <115670425+rolandtannous@users.noreply.github.com>
Co-authored-by: Daniel Han <danielhanchen@gmail.com>
---
 studio/backend/core/inference/llama_cpp.py    |  627 +++++++-
 .../backend/tests/test_kv_cache_estimation.py | 1340 ++++++++++++++++-
 .../tests/test_llama_cpp_context_fit.py       |    6 +-
 .../test_llama_cpp_max_context_threshold.py   |    6 +-
 4 files changed, 1871 insertions(+), 108 deletions(-)

diff --git a/studio/backend/core/inference/llama_cpp.py b/studio/backend/core/inference/llama_cpp.py
index bddd71c301..f768764c22 100644
--- a/studio/backend/core/inference/llama_cpp.py
+++ b/studio/backend/core/inference/llama_cpp.py
@@ -11,6 +11,7 @@ through its OpenAI-compatible /v1/chat/completions endpoint.
 import atexit
 import contextlib
 import json
+import os
 import re
 import struct
 import structlog
@@ -76,6 +77,238 @@ _SHARD_FULL_RE = re.compile(r"^(.*)-(\d{5})-of-(\d{5})\.gguf$")
 _SHARD_RE = re.compile(r"^(.*)-\d{5}-of-\d{5}\.gguf$")
 
 
+# ── Sliding-window-pattern resolver ───────────────────────────
+# Resolves the per-layer SWA mask when a GGUF reports a sliding window
+# but no `sliding_window_pattern` field. Tier order in
+# `_resolve_swa_pattern`: GGUF metadata, on-disk cache, bootstrap dict
+# below, transformers introspection, HF Hub config.json, legacy 1/4
+# fallback. Period N means layer i is SWA iff `(i + 1) % N != 0`,
+# matching transformers. Skipped on purpose: phi3 (no key/val length
+# in GGUF, window >= ctx anyway), qwen2 family (converter strips
+# sliding_window when use_sliding_window=False), mistral v0.1/v0.2
+# (all-SWA can't be expressed as a period).
+_BOOTSTRAP_SWA_DEFAULTS: dict[str, int] = {
+    "gemma2": 2,  # Gemma2Config.sliding_window_pattern
+    "gemma3": 6,  # Gemma3TextConfig.sliding_window_pattern
+    "gemma3n": 5,  # text_config.layer_types: SWA*4 + FULL
+    "gpt_oss": 2,  # text_config.layer_types: alternating
+    "cohere2": 4,  # Cohere2Config.sliding_window_pattern
+}
+
+# Process-wide cache backed by JSON on disk. Values are int period or
+# list[bool] mask. Lazy-loaded.
+_SWA_CACHE: Optional[dict] = None
+_SWA_CACHE_LOCK = threading.Lock()
+
+
+def _swa_cache_path() -> Path:
+    home = os.environ.get("UNSLOTH_STUDIO_HOME") or os.environ.get("STUDIO_HOME")
+    base = Path(home) if home else Path.home() / ".unsloth" / "studio"
+    return base / "swa_cache.json"
+
+
+def _load_swa_cache() -> dict:
+    global _SWA_CACHE
+    with _SWA_CACHE_LOCK:
+        if _SWA_CACHE is not None:
+            return _SWA_CACHE
+        try:
+            with open(_swa_cache_path()) as f:
+                _SWA_CACHE = json.load(f)
+                if not isinstance(_SWA_CACHE, dict):
+                    _SWA_CACHE = {}
+        except (FileNotFoundError, json.JSONDecodeError, OSError):
+            _SWA_CACHE = {}
+        return _SWA_CACHE
+
+
+def _save_swa_cache(cache: dict) -> None:
+    try:
+        path = _swa_cache_path()
+        path.parent.mkdir(parents = True, exist_ok = True)
+        tmp = path.with_suffix(".json.tmp")
+        with open(tmp, "w") as f:
+            json.dump(cache, f, indent = 2, sort_keys = True)
+        tmp.replace(path)
+    except OSError:
+        pass
+
+
+def _period_from_layer_types(layer_types: list) -> Optional[int]:
+    """Smallest period N where `(i+1) % N != 0` matches the SWA mask,
+    or None if no fixed period fits."""
+    if not layer_types:
+        return None
+    is_swa = ["full" not in str(t).lower() for t in layer_types]
+    n = len(is_swa)
+    for N in range(1, n + 1):
+        if all(((i + 1) % N != 0) == is_swa[i] for i in range(n)):
+            return N
+    return None
+
+
+def _fetch_swa_entry_from_hf(repo_id: str) -> Optional[object]:
+    try:
+        from huggingface_hub import hf_hub_download
+
+        cfg_path = hf_hub_download(repo_id, "config.json", repo_type = "model")
+        with open(cfg_path) as f:
+            cfg = json.load(f)
+    except Exception:
+        return None
+
+    src = cfg.get("text_config") if isinstance(cfg.get("text_config"), dict) else cfg
+    period = src.get("sliding_window_pattern")
+    if isinstance(period, int) and period > 0:
+        return period
+    lt = src.get("layer_types")
+    if isinstance(lt, list) and lt:
+        return _period_from_layer_types(lt) or [
+            "full" not in str(t).lower() for t in lt
+        ]
+    return None
+
+
+def _arch_aliases(arch: str) -> tuple:
+    # GGUF emits `falcon-h1`; HF model_type is `falcon_h1`. Normalise both ways.
+    seen = []
+    for a in (arch, arch.replace("-", "_"), arch.replace("_", "-")):
+        if a and a not in seen:
+            seen.append(a)
+    return tuple(seen)
+
+
+def _swa_entry_from_config_obj(cfg) -> Optional[object]:
+    src = getattr(cfg, "text_config", None) or cfg
+    period = getattr(src, "sliding_window_pattern", None)
+    if isinstance(period, int) and period > 0:
+        return period
+    lt = getattr(src, "layer_types", None)
+    if isinstance(lt, list) and lt:
+        return _period_from_layer_types(lt) or [
+            "full" not in str(t).lower() for t in lt
+        ]
+    return None
+
+
+_SWA_PATTERN_SOURCE_RE = re.compile(
+    r"sliding_window_pattern\s*(?::\s*[\w\[\], ]*)?\s*=\s*(\d+)"
+)
+
+
+def _resolve_swa_entry_from_transformers(arch: str) -> Optional[object]:
+    """Default-instantiate the matching Config; on failure, regex-parse
+    its source for `sliding_window_pattern = N`."""
+    try:
+        from transformers.models.auto.configuration_auto import (
+            CONFIG_MAPPING,
+            CONFIG_MAPPING_NAMES,
+        )
+    except Exception:
+        return None
+
+    cfg_class = None
+    for alias in _arch_aliases(arch):
+        if alias in CONFIG_MAPPING_NAMES:
+            try:
+                cfg_class = CONFIG_MAPPING[alias]
+                break
+            except Exception:
+                cfg_class = None
+    if cfg_class is None:
+        return None
+
+    try:
+        if (entry := _swa_entry_from_config_obj(cfg_class())) is not None:
+            return entry
+    except Exception:
+        pass
+
+    import inspect
+
+    candidates = [cfg_class]
+    text_cfg_class = getattr(cfg_class, "sub_configs", {}).get("text_config")
+    if text_cfg_class is not None:
+        candidates.append(text_cfg_class)
+    for cls in candidates:
+        try:
+            src = inspect.getsource(cls)
+        except (OSError, TypeError):
+            continue
+        if m := _SWA_PATTERN_SOURCE_RE.search(src):
+            period = int(m.group(1))
+            if period > 0:
+                return period
+    return None
+
+
+def _resolve_swa_pattern(
+    arch: Optional[str],
+    n_layers: Optional[int],
+    source_repo_candidates: tuple = (),
+    *,
+    allow_network: Optional[bool] = None,
+) -> Optional[list]:
+    if not arch or not n_layers:
+        return None
+    if allow_network is None:
+        allow_network = os.environ.get("UNSLOTH_STUDIO_OFFLINE", "0") not in (
+            "1",
+            "true",
+            "True",
+            "yes",
+        )
+
+    cache = _load_swa_cache()
+
+    def _entry_to_mask(entry):
+        if isinstance(entry, int) and entry > 0:
+            return [(i + 1) % entry != 0 for i in range(n_layers)]
+        if isinstance(entry, list) and entry:
+            return [bool(entry[i % len(entry)]) for i in range(n_layers)]
+        return None
+
+    def _persist(entry):
+        with _SWA_CACHE_LOCK:
+            cache[arch] = entry
+        _save_swa_cache(cache)
+
+    if (entry := cache.get(arch)) is not None:
+        if (mask := _entry_to_mask(entry)) is not None:
+            return mask
+
+    if (entry := _BOOTSTRAP_SWA_DEFAULTS.get(arch)) is not None:
+        return _entry_to_mask(entry)
+
+    entry = _resolve_swa_entry_from_transformers(arch)
+    if entry is not None:
+        _persist(entry)
+        return _entry_to_mask(entry)
+
+    # Tier 3: live HF fetch (with persistent caching of the result)
+    if allow_network:
+        for repo_id in source_repo_candidates:
+            if not repo_id:
+                continue
+            entry = _fetch_swa_entry_from_hf(repo_id)
+            if entry is not None:
+                _persist(entry)
+                return _entry_to_mask(entry)
+
+    return None
+
+
+def _hf_repo_from_url(url: Optional[str]) -> Optional[str]:
+    """Strip `https://huggingface.co/owner/name(/...)` to `owner/name`."""
+    if not url or "huggingface.co/" not in url:
+        return None
+    tail = url.split("huggingface.co/", 1)[1].rstrip("/")
+    parts = tail.split("/")
+    if len(parts) < 2:
+        return None
+    return f"{parts[0]}/{parts[1]}"
+
+
 # Model size extraction — lazy import to avoid pulling in transformers
 # at module level.  See PR description for the full explanation.
 def _extract_model_size_b(model_id: str):
@@ -215,17 +448,24 @@ class LlamaCppBackend:
         # KV-cache estimation fields (populated by _read_gguf_metadata)
         self._n_layers: Optional[int] = None
         self._n_kv_heads: Optional[int] = None
+        self._n_kv_heads_by_layer: Optional[list[int]] = None
         self._n_heads: Optional[int] = None
         self._embedding_length: Optional[int] = None
-        # Architecture-aware KV fields (8 new fields for 5-path estimation)
+        # Architecture-aware KV fields for 5-path estimation
         self._kv_key_length: Optional[int] = None
         self._kv_value_length: Optional[int] = None
         self._sliding_window: Optional[int] = None
+        self._sliding_window_pattern: Optional[list[bool]] = None
         self._full_attention_interval: Optional[int] = None
         self._kv_lora_rank: Optional[int] = None
         self._key_length_mla: Optional[int] = None
+        self._kv_key_length_swa: Optional[int] = None
+        self._kv_value_length_swa: Optional[int] = None
         self._ssm_inner_size: Optional[int] = None
         self._ssm_state_size: Optional[int] = None
+        # Last N layers reuse KV from earlier layers and don't allocate
+        # their own cache (Gemma 3n / Gemma 4: <arch>.attention.shared_kv_layers).
+        self._shared_kv_layers: Optional[int] = None
         self._lock = threading.Lock()
         self._stdout_lines: list[str] = []
         self._stdout_thread: Optional[threading.Thread] = None
@@ -744,13 +984,29 @@ class LlamaCppBackend:
         # New-style: need both explicit key AND value dimensions
         if self._kv_key_length is not None and self._kv_value_length is not None:
             return True
-        # Legacy: need embedding_length + head count
+        # Legacy: need embedding_length + a head count (scalar or per-layer).
         return self._embedding_length is not None and (
-            self._n_kv_heads is not None or self._n_heads is not None
+            self._n_kv_heads is not None
+            or self._n_heads is not None
+            or self._n_kv_heads_by_layer is not None
         )
 
+    def _kv_heads_for_layer(self, layer_idx: int, fallback: int) -> int:
+        if self._n_kv_heads_by_layer is not None and layer_idx < len(
+            self._n_kv_heads_by_layer
+        ):
+            return self._n_kv_heads_by_layer[layer_idx]
+        return fallback
+
     def _estimate_kv_cache_bytes(
-        self, n_ctx: int, cache_type_kv: Optional[str] = None
+        self,
+        n_ctx: int,
+        cache_type_kv: Optional[str] = None,
+        *,
+        swa_full: bool = False,
+        n_parallel: int = 1,
+        kv_unified: bool = True,
+        ctx_checkpoints: int = 0,
     ) -> int:
         """Estimate KV cache VRAM for a given context length.
 
@@ -761,12 +1017,34 @@ class LlamaCppBackend:
           4. GQA      -- standard full KV with explicit key/value dimensions
           5. Legacy   -- fallback using embed // n_heads
 
+        Server-flag knobs (mirror llama-server's CLI):
+          swa_full        -- ``--swa-full``: force SWA layers to cache the
+                             full ``n_ctx`` (collapses path 3 to path 4
+                             sizing for the SWA layers).
+          n_parallel      -- ``--parallel``: number of server slots.
+                             Verified empirically against llama-server:
+                             non-SWA layers stay constant (cells split
+                             across slots), SWA layers scale linearly
+                             (per-slot window).
+          kv_unified      -- ``--kv-unified`` (default on): retained for
+                             API forward-compat. Currently a no-op for
+                             memory math because the unified buffer total
+                             matches per-slot buffers in measured cases.
+          ctx_checkpoints -- ``--ctx-checkpoints``: SWA snapshot count per
+                             slot (PR #15293). Each snapshot stores one
+                             sliding-window of state per SWA layer.
+
         Returns 0 if metadata is insufficient for estimation.
         """
         if not self._can_estimate_kv() or n_ctx <= 0:
             return 0
 
         n_layers = self._n_layers  # type: ignore[assignment]
+        # Gemma 3n / Gemma 4 reuse KV from earlier layers in the last
+        # ``shared_kv_layers`` blocks -- those don't allocate their own
+        # cache.  Floor at 1 so a misconfigured GGUF can't zero out KV.
+        shared = self._shared_kv_layers or 0
+        n_layers_kv = max(1, n_layers - shared)
         n_kv = self._n_kv_heads or self._n_heads or 1  # type: ignore[assignment]
 
         # Bytes per element depends on KV cache quantization
@@ -782,6 +1060,8 @@ class LlamaCppBackend:
             "iq4_nl": 0.5625,
         }.get(cache_type_kv or "f16", 2.0)
 
+        slots = max(1, n_parallel)
+
         # Path 1: MLA (DeepSeek-V2/V3, GLM-4.7, GLM-5, Kimi-K2.5)
         # MLA stores one compressed KV latent per token/layer (shared across heads).
         # V is reconstructed from the latent on the fly -- no separate V cache.
@@ -792,7 +1072,7 @@ class LlamaCppBackend:
             n_kv_mla = self._n_kv_heads or 1
             rope_dim = self._key_length_mla or 64
             key_len = self._kv_key_length or (self._kv_lora_rank + rope_dim)
-            return int(n_layers * n_ctx * n_kv_mla * key_len * bpe)
+            return int(n_layers_kv * n_ctx * n_kv_mla * key_len * bpe)
 
         key_len = self._kv_key_length
         val_len = self._kv_value_length
@@ -810,11 +1090,19 @@ class LlamaCppBackend:
             head_dim = self._embedding_length // self._n_heads if self._n_heads else 128  # type: ignore[operator]
             return int(n_attn * n_ctx * n_kv * 2 * head_dim * bpe)
 
-        # Path 3: Sliding Window (Gemma-3, gpt-oss)
-        # SWA layers only cache min(ctx, window) tokens; global layers cache full ctx.
-        # Most SWA architectures use few global layers (e.g., Gemma-3 uses 1 in 6).
-        # Without an explicit field, we conservatively assume 1/4 of layers are global
-        # which is still far more accurate than the legacy formula (which ignores SWA).
+        # Path 3: Sliding window (Gemma 2/3/3n/4, gpt-oss, Cohere2 ...).
+        # Pattern is filled in by the resolver at parse time; if absent,
+        # falls through to the legacy 1/4-global heuristic below.
+        # Per-layer-type ``--parallel N`` accounting (verified empirically
+        # against ``llama-server``):
+        #   * non-SWA layers:    total cells = n_ctx, partitioned across
+        #                         slots -> total memory CONSTANT in slots.
+        #   * SWA layers:         per-slot cells = 2 * sliding_window
+        #                         (capped at n_ctx and at per_slot_ctx
+        #                         when ctx is split among many slots) ->
+        #                         total memory grows LINEARLY in slots.
+        # ``--swa-full`` forces full n_ctx for SWA layers instead.
+        # ``--ctx-checkpoints N`` adds N snapshots per SWA layer per slot.
         if (
             self._sliding_window is not None
             and self._sliding_window > 0
@@ -822,20 +1110,72 @@ class LlamaCppBackend:
             and val_len is not None
         ):
             swa = self._sliding_window
-            n_global = max(1, n_layers // 4)
-            n_swa = n_layers - n_global
+            per_slot_ctx = max(1, n_ctx // slots)
+            # ``--swa-full`` makes SWA layers cache the full context just
+            # like non-SWA: cells get partitioned across slots, so per-slot
+            # cells = per_slot_ctx and the slots*per-slot product collapses
+            # back to the constant ``n_ctx`` total.  Otherwise SWA caches
+            # 2*sliding_window per slot, clamped at the per-slot ctx.
+            swa_cells_per_slot = (
+                per_slot_ctx if swa_full else min(n_ctx, 2 * swa, per_slot_ctx)
+            )
+            key_len_swa = self._kv_key_length_swa or key_len
+            val_len_swa = self._kv_value_length_swa or val_len
+            if self._sliding_window_pattern is not None:
+                global_bytes = 0.0  # constant across slots
+                swa_bytes_per_slot = 0.0  # multiplied by slots
+                checkpoint_extra_per_slot = 0.0
+                # Iterate only over layers that allocate their own KV;
+                # the trailing ``shared`` layers reuse earlier caches.
+                for layer_idx in range(n_layers_kv):
+                    layer_n_kv = self._kv_heads_for_layer(layer_idx, n_kv)
+                    is_swa = (
+                        layer_idx < len(self._sliding_window_pattern)
+                        and self._sliding_window_pattern[layer_idx]
+                    )
+                    if is_swa:
+                        swa_bytes_per_slot += (
+                            swa_cells_per_slot
+                            * layer_n_kv
+                            * (key_len_swa + val_len_swa)
+                            * bpe
+                        )
+                        if ctx_checkpoints > 0 and not swa_full:
+                            checkpoint_extra_per_slot += (
+                                ctx_checkpoints
+                                * swa
+                                * layer_n_kv
+                                * (key_len_swa + val_len_swa)
+                                * bpe
+                            )
+                    else:
+                        global_bytes += n_ctx * layer_n_kv * (key_len + val_len) * bpe
+                return int(
+                    global_bytes
+                    + slots * (swa_bytes_per_slot + checkpoint_extra_per_slot)
+                )
+            n_global = max(1, n_layers_kv // 4)
+            n_swa = n_layers_kv - n_global
             kv_per_token = n_kv * (key_len + val_len) * bpe
+            kv_per_token_swa = n_kv * (key_len_swa + val_len_swa) * bpe
+            global_bytes = n_global * n_ctx * kv_per_token
+            swa_bytes_per_slot = n_swa * swa_cells_per_slot * kv_per_token_swa
+            checkpoint_extra_per_slot = (
+                ctx_checkpoints * n_swa * swa * kv_per_token_swa
+                if ctx_checkpoints > 0 and not swa_full
+                else 0.0
+            )
             return int(
-                n_global * n_ctx * kv_per_token + n_swa * min(n_ctx, swa) * kv_per_token
+                global_bytes + slots * (swa_bytes_per_slot + checkpoint_extra_per_slot)
             )
 
         # Path 4: Standard GQA with explicit key/value dimensions
         if key_len is not None and val_len is not None:
-            return int(n_layers * n_ctx * n_kv * (key_len + val_len) * bpe)
+            return int(n_layers_kv * n_ctx * n_kv * (key_len + val_len) * bpe)
 
         # Path 5: Legacy fallback (old GGUFs without explicit dimensions)
         head_dim = self._embedding_length // self._n_heads if self._n_heads else 128  # type: ignore[operator]
-        return int(2 * n_kv * head_dim * n_layers * n_ctx * bpe)
+        return int(2 * n_kv * head_dim * n_layers_kv * n_ctx * bpe)
 
     def _fit_context_to_vram(
         self,
@@ -844,6 +1184,12 @@ class LlamaCppBackend:
         model_size_bytes: int,
         cache_type_kv: Optional[str] = None,
         min_ctx: int = 4096,
+        *,
+        swa_full: bool = False,
+        n_parallel: int = 1,
+        kv_unified: bool = True,
+        ctx_checkpoints: int = 0,
+        kv_on_gpu: bool = True,
     ) -> int:
         """Return the largest context length that fits in GPU VRAM.
 
@@ -851,6 +1197,11 @@ class LlamaCppBackend:
         threshold -- 10% reserved for compute buffers, CUDA context,
         scratch space, flash-attn workspace, etc.).
         If the model weights alone don't fit, returns min_ctx unchanged.
+
+        ``kv_on_gpu`` mirrors ``--kv-offload`` (default on). When False
+        the KV cache lives in CPU RAM and doesn't compete with weights
+        for VRAM; the requested context is honored verbatim. The other
+        keyword args mirror ``_estimate_kv_cache_bytes``.
         """
         if not self._can_estimate_kv():
             logger.debug(
@@ -860,11 +1211,22 @@ class LlamaCppBackend:
             )
             return requested_ctx
 
+        # KV lives off-GPU: no VRAM accounting needed for the cache itself.
+        if not kv_on_gpu:
+            return requested_ctx
+
+        kv_kwargs = dict(
+            swa_full = swa_full,
+            n_parallel = n_parallel,
+            kv_unified = kv_unified,
+            ctx_checkpoints = ctx_checkpoints,
+        )
+
         budget_bytes = available_mib * 1024 * 1024 * 0.90
         model_footprint = model_size_bytes
 
         # Check if requested context already fits
-        kv = self._estimate_kv_cache_bytes(requested_ctx, cache_type_kv)
+        kv = self._estimate_kv_cache_bytes(requested_ctx, cache_type_kv, **kv_kwargs)
         if model_footprint + kv <= budget_bytes:
             return requested_ctx
 
@@ -886,7 +1248,7 @@ class LlamaCppBackend:
         best = effective_min
         while lo <= hi:
             mid = (lo + hi) // 2
-            kv = self._estimate_kv_cache_bytes(mid, cache_type_kv)
+            kv = self._estimate_kv_cache_bytes(mid, cache_type_kv, **kv_kwargs)
             if kv <= remaining:
                 best = mid
                 lo = mid + 1
@@ -1019,6 +1381,19 @@ class LlamaCppBackend:
                 for _ in range(alen):
                     LlamaCppBackend._gguf_skip_value(f, atype)
 
+    @staticmethod
+    def _gguf_read_array_value(f, atype: int, alen: int) -> Optional[list]:
+        if atype == 4:  # UINT32
+            return [struct.unpack("<I", f.read(4))[0] for _ in range(alen)]
+        if atype == 5:  # INT32
+            return [struct.unpack("<i", f.read(4))[0] for _ in range(alen)]
+        if atype == 7:  # BOOL
+            return [struct.unpack("<?", f.read(1))[0] for _ in range(alen)]
+
+        for _ in range(alen):
+            LlamaCppBackend._gguf_skip_value(f, atype)
+        return None
+
     def _read_gguf_metadata(self, gguf_path: str) -> None:
         """Read context_length, architecture params, and chat_template from a GGUF header.
 
@@ -1037,23 +1412,44 @@ class LlamaCppBackend:
         self._supports_tools = False
         self._n_layers = None
         self._n_kv_heads = None
+        self._n_kv_heads_by_layer = None
         self._n_heads = None
         self._embedding_length = None
         self._kv_key_length = None
         self._kv_value_length = None
         self._sliding_window = None
+        self._sliding_window_pattern = None
         self._full_attention_interval = None
         self._kv_lora_rank = None
         self._key_length_mla = None
+        self._kv_key_length_swa = None
+        self._kv_value_length_swa = None
         self._ssm_inner_size = None
         self._ssm_state_size = None
+        self._shared_kv_layers = None
 
         try:
-            WANTED = {"general.architecture", "tokenizer.chat_template"}
+            WANTED = {
+                "general.architecture",
+                "tokenizer.chat_template",
+                # Source-repo hints for the SWA resolver's HF fallback.
+                "general.source.huggingface.repository",
+                "general.source.url",
+                "general.source.repo_url",
+                "general.base_model.0.repo_url",
+                "general.base_model.0.organization",
+                "general.base_model.0.name",
+                "general.basename",
+                "general.organization",
+                "general.size_label",
+                "general.finetune",
+            }
             # Additional arch-specific keys are added dynamically once
             # we know the architecture name.
             arch_keys: dict[str, str] = {}  # gguf_key -> attribute name
             arch = None
+            sliding_window_pattern_period: Optional[int] = None
+            general: dict[str, str] = {}
 
             with open(gguf_path, "rb") as f:
                 magic = struct.unpack("<I", f.read(4))[0]
@@ -1063,49 +1459,141 @@ class LlamaCppBackend:
                 _tensor_count, kv_count = struct.unpack("<QQ", f.read(16))
 
                 for _ in range(kv_count):
-                    key_len = struct.unpack("<Q", f.read(8))[0]
-                    key = f.read(key_len).decode("utf-8")
-                    vtype = struct.unpack("<I", f.read(4))[0]
+                    # Tolerate truncated input (e.g., a partial header
+                    # fetched via HTTP byte-range): bail out gracefully
+                    # so the resolver fallback still runs on whatever
+                    # we did manage to parse.
+                    try:
+                        key_len_bytes = f.read(8)
+                        if len(key_len_bytes) < 8:
+                            break
+                        key_len = struct.unpack("<Q", key_len_bytes)[0]
+                        key_bytes = f.read(key_len)
+                        if len(key_bytes) < key_len:
+                            break
+                        key = key_bytes.decode("utf-8")
+                        vtype_bytes = f.read(4)
+                        if len(vtype_bytes) < 4:
+                            break
+                        vtype = struct.unpack("<I", vtype_bytes)[0]
+                    except (struct.error, UnicodeDecodeError):
+                        break
 
-                    if key in WANTED or key in arch_keys:
-                        # Read this value
-                        if vtype == 8:  # STRING
-                            slen = struct.unpack("<Q", f.read(8))[0]
-                            val_s = f.read(slen).decode("utf-8")
-                            if key == "general.architecture":
-                                arch = val_s
-                                # Register arch-specific keys to look for
-                                arch_keys = {
-                                    f"{arch}.context_length": "context_length",
-                                    f"{arch}.block_count": "n_layers",
-                                    f"{arch}.attention.head_count_kv": "n_kv_heads",
-                                    f"{arch}.attention.head_count": "n_heads",
-                                    f"{arch}.embedding_length": "embedding_length",
-                                    # Architecture-aware KV cache fields
-                                    f"{arch}.attention.key_length": "kv_key_length",
-                                    f"{arch}.attention.value_length": "kv_value_length",
-                                    f"{arch}.attention.sliding_window": "sliding_window",
-                                    f"{arch}.full_attention_interval": "full_attention_interval",
-                                    f"{arch}.attention.kv_lora_rank": "kv_lora_rank",
-                                    f"{arch}.attention.key_length_mla": "key_length_mla",
-                                    f"{arch}.ssm.inner_size": "ssm_inner_size",
-                                    f"{arch}.ssm.state_size": "ssm_state_size",
-                                }
-                            elif key == "tokenizer.chat_template":
-                                self._chat_template = val_s
-                        elif vtype in (4, 10):  # UINT32 or UINT64
-                            val_i = (
-                                struct.unpack("<I", f.read(4))[0]
-                                if vtype == 4
-                                else struct.unpack("<Q", f.read(8))[0]
-                            )
-                            attr = arch_keys.get(key)
-                            if attr:
-                                setattr(self, f"_{attr}", val_i)
+                    try:
+                        if key in WANTED or key in arch_keys:
+                            if vtype == 8:  # STRING
+                                slen = struct.unpack("<Q", f.read(8))[0]
+                                val_s = f.read(slen).decode("utf-8")
+                                if (
+                                    key.startswith("general.")
+                                    and key != "general.architecture"
+                                ):
+                                    general[key] = val_s
+                                if key == "general.architecture":
+                                    arch = val_s
+                                    arch_keys = {
+                                        f"{arch}.context_length": "context_length",
+                                        f"{arch}.block_count": "n_layers",
+                                        f"{arch}.attention.head_count_kv": "n_kv_heads",
+                                        f"{arch}.attention.head_count": "n_heads",
+                                        f"{arch}.embedding_length": "embedding_length",
+                                        f"{arch}.attention.key_length": "kv_key_length",
+                                        f"{arch}.attention.value_length": "kv_value_length",
+                                        f"{arch}.attention.sliding_window": "sliding_window",
+                                        f"{arch}.attention.sliding_window_pattern": "sliding_window_pattern",
+                                        f"{arch}.full_attention_interval": "full_attention_interval",
+                                        f"{arch}.attention.kv_lora_rank": "kv_lora_rank",
+                                        f"{arch}.attention.key_length_mla": "key_length_mla",
+                                        f"{arch}.attention.key_length_swa": "kv_key_length_swa",
+                                        f"{arch}.attention.value_length_swa": "kv_value_length_swa",
+                                        f"{arch}.attention.shared_kv_layers": "shared_kv_layers",
+                                        f"{arch}.ssm.inner_size": "ssm_inner_size",
+                                        f"{arch}.ssm.state_size": "ssm_state_size",
+                                    }
+                                elif key == "tokenizer.chat_template":
+                                    self._chat_template = val_s
+                            elif vtype in (4, 10):  # UINT32 or UINT64
+                                val_i = (
+                                    struct.unpack("<I", f.read(4))[0]
+                                    if vtype == 4
+                                    else struct.unpack("<Q", f.read(8))[0]
+                                )
+                                attr = arch_keys.get(key)
+                                if attr:
+                                    if attr == "sliding_window_pattern":
+                                        sliding_window_pattern_period = val_i
+                                    else:
+                                        setattr(self, f"_{attr}", val_i)
+                            elif vtype == 9:  # ARRAY
+                                atype = struct.unpack("<I", f.read(4))[0]
+                                alen = struct.unpack("<Q", f.read(8))[0]
+                                val_a = self._gguf_read_array_value(f, atype, alen)
+                                attr = arch_keys.get(key)
+                                if attr == "n_kv_heads" and val_a is not None:
+                                    self._n_kv_heads_by_layer = [int(x) for x in val_a]
+                                    if self._n_kv_heads is None and val_a:
+                                        self._n_kv_heads = max(int(x) for x in val_a)
+                                elif (
+                                    attr == "sliding_window_pattern"
+                                    and val_a is not None
+                                ):
+                                    self._sliding_window_pattern = [
+                                        bool(x) for x in val_a
+                                    ]
+                                    sliding_window_pattern_period = None
+                            else:
+                                self._gguf_skip_value(f, vtype)
                         else:
                             self._gguf_skip_value(f, vtype)
-                    else:
-                        self._gguf_skip_value(f, vtype)
+                    except (struct.error, UnicodeDecodeError):
+                        # Truncated input (e.g., HTTP byte-range fetch
+                        # of just the GGUF header); break so the
+                        # resolver fallback still runs on what we have.
+                        break
+
+            # Expand a scalar period straight from the GGUF first.
+            if (
+                self._sliding_window_pattern is None
+                and sliding_window_pattern_period
+                and self._n_layers
+            ):
+                self._sliding_window_pattern = [
+                    (i + 1) % sliding_window_pattern_period != 0
+                    for i in range(self._n_layers)
+                ]
+
+            # Otherwise hand off to the resolver (cache / bootstrap /
+            # transformers / HF). See `_resolve_swa_pattern`.
+            if (
+                self._sliding_window_pattern is None
+                and self._sliding_window
+                and self._n_layers
+            ):
+                hf_repo_candidates = (
+                    general.get("general.source.huggingface.repository"),
+                    _hf_repo_from_url(general.get("general.source.url")),
+                    _hf_repo_from_url(general.get("general.source.repo_url")),
+                    _hf_repo_from_url(general.get("general.base_model.0.repo_url")),
+                    (
+                        f"{general['general.base_model.0.organization']}/"
+                        f"{general['general.base_model.0.name']}".replace(" ", "-")
+                        if general.get("general.base_model.0.organization")
+                        and general.get("general.base_model.0.name")
+                        else None
+                    ),
+                    (
+                        f"{general['general.organization']}/"
+                        f"{general['general.basename']}".replace(" ", "-")
+                        if general.get("general.organization")
+                        and general.get("general.basename")
+                        else None
+                    ),
+                )
+                self._sliding_window_pattern = _resolve_swa_pattern(
+                    arch,
+                    self._n_layers,
+                    hf_repo_candidates,
+                )
 
             if self._context_length:
                 logger.info(f"GGUF metadata: context_length={self._context_length}")
@@ -1504,8 +1992,11 @@ class LlamaCppBackend:
                                 pool_mib,
                                 model_size,
                                 cache_type_kv,
+                                n_parallel = n_parallel,
+                            )
+                            kv = self._estimate_kv_cache_bytes(
+                                capped, cache_type_kv, n_parallel = n_parallel
                             )
-                            kv = self._estimate_kv_cache_bytes(capped, cache_type_kv)
                             total_mib = (model_size + kv) / (1024 * 1024)
                             if total_mib <= pool_mib * 0.90:
                                 best_cap = max(best_cap, capped)
@@ -1529,7 +2020,7 @@ class LlamaCppBackend:
                         # have surfaced the "might be slower" warning before
                         # the user submitted a ctx above the fit ceiling.
                         requested_total = model_size + self._estimate_kv_cache_bytes(
-                            effective_ctx, cache_type_kv
+                            effective_ctx, cache_type_kv, n_parallel = n_parallel
                         )
                         gpu_indices, use_fit = self._select_gpus(requested_total, gpus)
                         # No silent shrink: effective_ctx stays == n_ctx.
@@ -1544,8 +2035,11 @@ class LlamaCppBackend:
                                 pool_mib,
                                 model_size,
                                 cache_type_kv,
+                                n_parallel = n_parallel,
+                            )
+                            kv = self._estimate_kv_cache_bytes(
+                                capped, cache_type_kv, n_parallel = n_parallel
                             )
-                            kv = self._estimate_kv_cache_bytes(capped, cache_type_kv)
                             total_mib = (model_size + kv) / (1024 * 1024)
                             if total_mib <= pool_mib * 0.90:
                                 effective_ctx = capped
@@ -1578,7 +2072,9 @@ class LlamaCppBackend:
                         )
 
                 if effective_ctx < original_ctx:
-                    kv_est = self._estimate_kv_cache_bytes(effective_ctx, cache_type_kv)
+                    kv_est = self._estimate_kv_cache_bytes(
+                        effective_ctx, cache_type_kv, n_parallel = n_parallel
+                    )
                     logger.info(
                         f"Context auto-reduced: {original_ctx} -> {effective_ctx} "
                         f"(model: {model_size / (1024**3):.1f} GB, "
@@ -1586,7 +2082,7 @@ class LlamaCppBackend:
                     )
 
                 kv_cache_bytes = self._estimate_kv_cache_bytes(
-                    effective_ctx, cache_type_kv
+                    effective_ctx, cache_type_kv, n_parallel = n_parallel
                 )
                 logger.info(
                     f"GGUF size: {model_size / (1024**3):.1f} GB, "
@@ -2015,16 +2511,21 @@ class LlamaCppBackend:
             self._speculative_type = None
             self._n_layers = None
             self._n_kv_heads = None
+            self._n_kv_heads_by_layer = None
             self._n_heads = None
             self._embedding_length = None
             self._kv_key_length = None
             self._kv_value_length = None
             self._sliding_window = None
+            self._sliding_window_pattern = None
             self._full_attention_interval = None
             self._kv_lora_rank = None
             self._key_length_mla = None
+            self._kv_key_length_swa = None
+            self._kv_value_length_swa = None
             self._ssm_inner_size = None
             self._ssm_state_size = None
+            self._shared_kv_layers = None
             # Clean up temp chat template file
             if hasattr(self, "_chat_template_file") and self._chat_template_file:
                 try:
diff --git a/studio/backend/tests/test_kv_cache_estimation.py b/studio/backend/tests/test_kv_cache_estimation.py
index 2640ded90d..29d87804ff 100644
--- a/studio/backend/tests/test_kv_cache_estimation.py
+++ b/studio/backend/tests/test_kv_cache_estimation.py
@@ -12,6 +12,7 @@ Cross-platform: Linux, macOS, Windows, WSL.
 """
 
 import io
+import json
 import struct
 import sys
 import types as _types
@@ -37,35 +38,43 @@ sys.modules.setdefault("loggers", _loggers_stub)
 _structlog_stub = _types.ModuleType("structlog")
 sys.modules.setdefault("structlog", _structlog_stub)
 
-# httpx
-_httpx_stub = _types.ModuleType("httpx")
-for _exc_name in (
-    "ConnectError",
-    "TimeoutException",
-    "ReadTimeout",
-    "ReadError",
-    "RemoteProtocolError",
-    "CloseError",
-):
-    setattr(_httpx_stub, _exc_name, type(_exc_name, (Exception,), {}))
+# httpx -- only stub when the real library isn't installed.  Stubbing
+# unconditionally would shadow ``HTTPError`` / ``Response`` etc. that
+# ``huggingface_hub.errors`` imports at module load time, which causes
+# the transformers introspection tier to silently return None inside
+# the test process.
+try:
+    import httpx as _httpx_real  # noqa: F401
+except ImportError:
+    _httpx_stub = _types.ModuleType("httpx")
+    for _exc_name in (
+        "ConnectError",
+        "TimeoutException",
+        "ReadTimeout",
+        "ReadError",
+        "RemoteProtocolError",
+        "CloseError",
+        "HTTPError",
+        "RequestError",
+    ):
+        setattr(_httpx_stub, _exc_name, type(_exc_name, (Exception,), {}))
 
+    class _FakeTimeout:
+        def __init__(self, *a, **kw):
+            pass
 
-class _FakeTimeout:
-    def __init__(self, *a, **kw):
-        pass
-
-
-_httpx_stub.Timeout = _FakeTimeout
-_httpx_stub.Client = type(
-    "Client",
-    (),
-    {
-        "__init__": lambda self, **kw: None,
-        "__enter__": lambda self: self,
-        "__exit__": lambda self, *a: None,
-    },
-)
-sys.modules.setdefault("httpx", _httpx_stub)
+    _httpx_stub.Timeout = _FakeTimeout
+    _httpx_stub.Response = type("Response", (), {})
+    _httpx_stub.Client = type(
+        "Client",
+        (),
+        {
+            "__init__": lambda self, **kw: None,
+            "__enter__": lambda self: self,
+            "__exit__": lambda self, *a: None,
+        },
+    )
+    sys.modules["httpx"] = _httpx_stub
 
 from core.inference.llama_cpp import LlamaCppBackend
 
@@ -77,8 +86,7 @@ from core.inference.llama_cpp import LlamaCppBackend
 def _make_gguf_bytes(arch: str, kv_pairs: dict) -> bytes:
     """Build a minimal GGUF v3 binary blob with the given KV metadata.
 
-    Only supports UINT32 (type 4), UINT64 (type 10), and STRING (type 8)
-    values, which is all the metadata parser reads.
+    Supports the scalar and simple array metadata used by the parser.
     """
     buf = io.BytesIO()
     # Header: magic, version, tensor_count, kv_count
@@ -96,6 +104,17 @@ def _make_gguf_bytes(arch: str, kv_pairs: dict) -> bytes:
             val_bytes = val.encode("utf-8")
             buf.write(struct.pack("<Q", len(val_bytes)))
             buf.write(val_bytes)
+        elif isinstance(val, list):
+            buf.write(struct.pack("<I", 9))  # ARRAY
+            is_bool_array = all(isinstance(x, bool) for x in val)
+            buf.write(struct.pack("<I", 7 if is_bool_array else 5))
+            buf.write(struct.pack("<Q", len(val)))
+            if is_bool_array:
+                for item in val:
+                    buf.write(struct.pack("<?", item))
+            else:
+                for item in val:
+                    buf.write(struct.pack("<i", item))
         elif isinstance(val, int):
             if val <= 0xFFFFFFFF:
                 buf.write(struct.pack("<I", 4))  # UINT32
@@ -108,9 +127,18 @@ def _make_gguf_bytes(arch: str, kv_pairs: dict) -> bytes:
     return buf.getvalue()
 
 
-def _backend_from_gguf(arch: str, fields: dict) -> LlamaCppBackend:
-    """Create a LlamaCppBackend with parsed GGUF metadata from given fields."""
+def _backend_from_gguf(
+    arch: str, fields: dict, general: dict | None = None
+) -> LlamaCppBackend:
+    """Create a LlamaCppBackend with parsed GGUF metadata from given fields.
+
+    `general` lets a test inject extra `general.*` metadata (used to
+    verify the dynamic SWA resolver picks up source-repo hints from
+    GGUFs that ship them).
+    """
     kv = {"general.architecture": arch}
+    for k, v in (general or {}).items():
+        kv[k] = v
     for k, v in fields.items():
         kv[f"{arch}.{k}"] = v
     import tempfile, os
@@ -133,7 +161,7 @@ def _backend_from_gguf(arch: str, fields: dict) -> LlamaCppBackend:
 
 
 class TestGGUFParserNewFields:
-    """Verify that the 8 new architecture-aware fields are correctly parsed."""
+    """Verify that architecture-aware fields are correctly parsed."""
 
     @pytest.mark.parametrize(
         "field,gguf_key,value",
@@ -158,15 +186,189 @@ class TestGGUFParserNewFields:
             "_kv_key_length",
             "_kv_value_length",
             "_sliding_window",
+            "_sliding_window_pattern",
             "_full_attention_interval",
             "_kv_lora_rank",
             "_key_length_mla",
+            "_kv_key_length_swa",
+            "_kv_value_length_swa",
             "_ssm_inner_size",
             "_ssm_state_size",
         ]:
             assert getattr(b, attr) is None
 
-    def test_all_13_fields_parsed_together(self):
+    def test_array_fields_parsed(self):
+        b = _backend_from_gguf(
+            "gemma4",
+            {
+                "block_count": 6,
+                "attention.head_count_kv": [8, 8, 8, 8, 8, 2],
+                "attention.sliding_window_pattern": [
+                    True,
+                    True,
+                    True,
+                    True,
+                    True,
+                    False,
+                ],
+            },
+        )
+        # Per-layer KV head count is preserved exactly...
+        assert b._n_kv_heads_by_layer == [8, 8, 8, 8, 8, 2]
+        # ...and mirrored into the scalar field as a conservative max so
+        # non-SWA estimator paths and any caller using
+        # `n_kv = self._n_kv_heads or ...` get a safe upper bound.
+        assert b._n_kv_heads == 8
+        assert b._sliding_window_pattern == [True, True, True, True, True, False]
+
+
+class TestArchSwaPatternDefaults:
+    """Bootstrap arch table fires when GGUF reports `sliding_window` but
+    no per-layer pattern (true for every Gemma 2/3/3n/gpt-oss GGUF today)."""
+
+    @pytest.mark.parametrize(
+        "arch,n_layers,expected_period",
+        [
+            ("gemma2", 26, 2),
+            ("gemma3", 18, 6),
+            ("gemma3n", 35, 5),
+            ("gpt_oss", 24, 2),
+            ("cohere2", 32, 4),
+        ],
+    )
+    def test_arch_default_pattern_applied(self, arch, n_layers, expected_period):
+        b = _backend_from_gguf(
+            arch,
+            {
+                "block_count": n_layers,
+                "attention.head_count": 4,
+                "attention.head_count_kv": 1,
+                "attention.key_length": 256,
+                "attention.value_length": 256,
+                "attention.sliding_window": 512,
+            },
+        )
+        expected_pattern = [(i + 1) % expected_period != 0 for i in range(n_layers)]
+        assert (
+            b._sliding_window_pattern == expected_pattern
+        ), f"{arch} should expand to period={expected_period}"
+
+    def test_unknown_arch_no_default(self):
+        b = _backend_from_gguf(
+            "totallymadeupv7",
+            {
+                "block_count": 24,
+                "attention.head_count": 4,
+                "attention.head_count_kv": 1,
+                "attention.key_length": 128,
+                "attention.value_length": 128,
+                "attention.sliding_window": 1024,
+            },
+        )
+        assert b._sliding_window_pattern is None
+
+    def test_explicit_pattern_overrides_arch_default(self):
+        # Period=6 is the gemma3 default; the explicit array must win.
+        b = _backend_from_gguf(
+            "gemma3",
+            {
+                "block_count": 6,
+                "attention.head_count": 4,
+                "attention.head_count_kv": 1,
+                "attention.key_length": 256,
+                "attention.value_length": 256,
+                "attention.sliding_window": 512,
+                "attention.sliding_window_pattern": [
+                    True,
+                    False,
+                    True,
+                    False,
+                    True,
+                    False,
+                ],
+            },
+        )
+        assert b._sliding_window_pattern == [True, False, True, False, True, False]
+
+    def test_no_sliding_window_no_pattern(self):
+        b = _backend_from_gguf(
+            "gemma3",
+            {
+                "block_count": 18,
+                "attention.head_count": 4,
+                "attention.head_count_kv": 1,
+                "attention.key_length": 256,
+                "attention.value_length": 256,
+                # no sliding_window key
+            },
+        )
+        assert b._sliding_window_pattern is None
+
+    @pytest.mark.parametrize(
+        "arch", ["llama", "qwen2", "qwen3", "mistral", "mistral3", "glm4", "llama4"]
+    )
+    def test_non_swa_arch_uses_full_attention_path(self, arch):
+        # Pure-GQA arches: GGUF has no sliding_window, no synthetic
+        # pattern, estimator hits Path 4.
+        b = _backend_from_gguf(
+            arch,
+            {
+                "block_count": 32,
+                "attention.head_count": 32,
+                "attention.head_count_kv": 8,
+                "attention.key_length": 128,
+                "attention.value_length": 128,
+                "embedding_length": 4096,
+            },
+        )
+        assert b._sliding_window_pattern is None
+        assert b._sliding_window is None
+        kv = b._estimate_kv_cache_bytes(8192, "f16")
+        gqa_expected = 32 * 8192 * 8 * (128 + 128) * 2
+        assert kv == gqa_expected
+
+    def test_arch_default_reduces_kv_estimate_vs_legacy(self):
+        common = {
+            "block_count": 62,
+            "attention.head_count": 32,
+            "attention.head_count_kv": 16,
+            "attention.key_length": 128,
+            "attention.value_length": 128,
+            "attention.sliding_window": 1024,
+            "embedding_length": 5376,
+        }
+        with_default = _backend_from_gguf("gemma3", common)
+        # Arch not in the table -> legacy 1/4 path.
+        without_default = _backend_from_gguf("totallymadeupv7", common)
+
+        kv_default = with_default._estimate_kv_cache_bytes(131072, "f16")
+        kv_legacy = without_default._estimate_kv_cache_bytes(131072, "f16")
+        assert kv_default > 0
+        assert kv_legacy > 0
+        assert kv_default < kv_legacy, (
+            f"arch fallback should under-shoot legacy estimate: "
+            f"{kv_default} >= {kv_legacy}"
+        )
+
+    def test_scalar_sliding_window_pattern_expanded(self):
+        block_count = 8
+        b = _backend_from_gguf(
+            "gemma3",
+            {
+                "attention.sliding_window_pattern": 4,
+                "block_count": block_count,
+                "attention.head_count_kv": 4,
+                "attention.key_length": 256,
+                "attention.value_length": 256,
+                "attention.sliding_window": 1024,
+            },
+        )
+        expected = [(i + 1) % 4 != 0 for i in range(block_count)]
+        assert isinstance(b._sliding_window_pattern, list)
+        assert b._sliding_window_pattern == expected
+        assert b._estimate_kv_cache_bytes(4096, "f16") > 0
+
+    def test_all_fields_parsed_together(self):
         fields = {
             "context_length": 131072,
             "block_count": 62,
@@ -176,9 +378,12 @@ class TestGGUFParserNewFields:
             "attention.key_length": 128,
             "attention.value_length": 128,
             "attention.sliding_window": 1024,
+            "attention.sliding_window_pattern": [True, False],
             "full_attention_interval": 6,
             "attention.kv_lora_rank": 512,
             "attention.key_length_mla": 256,
+            "attention.key_length_swa": 64,
+            "attention.value_length_swa": 64,
             "ssm.inner_size": 4096,
             "ssm.state_size": 128,
         }
@@ -191,13 +396,294 @@ class TestGGUFParserNewFields:
         assert b._kv_key_length == 128
         assert b._kv_value_length == 128
         assert b._sliding_window == 1024
+        assert b._sliding_window_pattern == [True, False]
         assert b._full_attention_interval == 6
         assert b._kv_lora_rank == 512
         assert b._key_length_mla == 256
+        assert b._kv_key_length_swa == 64
+        assert b._kv_value_length_swa == 64
         assert b._ssm_inner_size == 4096
         assert b._ssm_state_size == 128
 
 
+_SWA_FIELDS = {
+    "block_count": 12,
+    "attention.head_count": 4,
+    "attention.head_count_kv": 1,
+    "attention.key_length": 256,
+    "attention.value_length": 256,
+    "attention.sliding_window": 512,
+}
+
+
+class TestDynamicSwaResolver:
+    """4-tier resolver: GGUF metadata, on-disk cache, bootstrap, HF fetch."""
+
+    def _isolate_cache(self, monkeypatch, tmp_path):
+        from core.inference import llama_cpp as lc
+
+        monkeypatch.setenv("UNSLOTH_STUDIO_HOME", str(tmp_path))
+        monkeypatch.setattr(lc, "_SWA_CACHE", None)
+        return tmp_path
+
+    def test_period_from_layer_types_finds_smallest_period(self):
+        from core.inference.llama_cpp import _period_from_layer_types
+
+        # gemma3 (1 global per 6), gpt-oss (alternating), gemma3n (1 per 5).
+        assert (
+            _period_from_layer_types(
+                (["sliding_attention"] * 5 + ["full_attention"]) * 4
+            )
+            == 6
+        )
+        assert (
+            _period_from_layer_types(["sliding_attention", "full_attention"] * 12) == 2
+        )
+        assert (
+            _period_from_layer_types(
+                (["sliding_attention"] * 4 + ["full_attention"]) * 7
+            )
+            == 5
+        )
+
+    def test_period_from_layer_types_returns_none_for_aperiodic(self):
+        from core.inference.llama_cpp import _period_from_layer_types
+
+        lt = [
+            "sliding_attention",
+            "full_attention",
+            "sliding_attention",
+            "sliding_attention",
+            "full_attention",
+            "sliding_attention",
+            "sliding_attention",
+            "sliding_attention",
+        ]
+        assert _period_from_layer_types(lt) is None
+
+    def test_hf_repo_from_url(self):
+        from core.inference.llama_cpp import _hf_repo_from_url
+
+        assert (
+            _hf_repo_from_url("https://huggingface.co/google/gemma-3-1b-it")
+            == "google/gemma-3-1b-it"
+        )
+        assert (
+            _hf_repo_from_url(
+                "https://huggingface.co/google/gemma-3-1b-it/blob/main/config.json"
+            )
+            == "google/gemma-3-1b-it"
+        )
+        for bad in [
+            "https://huggingface.co/google",
+            "https://example.com/foo/bar",
+            None,
+            "",
+        ]:
+            assert _hf_repo_from_url(bad) is None
+
+    def test_bootstrap_tier_used_when_no_cache(self, monkeypatch, tmp_path):
+        self._isolate_cache(monkeypatch, tmp_path)
+        from core.inference import llama_cpp as lc
+
+        def boom(*a, **kw):
+            raise AssertionError("HF fetch must not run when bootstrap covers the arch")
+
+        monkeypatch.setattr(lc, "_fetch_swa_entry_from_hf", boom)
+        b = _backend_from_gguf("gemma3", dict(_SWA_FIELDS, block_count = 18))
+        assert b._sliding_window_pattern == [(i + 1) % 6 != 0 for i in range(18)]
+
+    def test_disk_cache_takes_precedence_over_bootstrap(self, monkeypatch, tmp_path):
+        self._isolate_cache(monkeypatch, tmp_path)
+        # Override bootstrap=6 with a cached period=3.
+        with open(tmp_path / "swa_cache.json", "w") as f:
+            json.dump({"gemma3": 3}, f)
+        b = _backend_from_gguf("gemma3", dict(_SWA_FIELDS, block_count = 18))
+        assert b._sliding_window_pattern == [(i + 1) % 3 != 0 for i in range(18)]
+
+    def test_disk_cache_supports_array_entries(self, monkeypatch, tmp_path):
+        # Aperiodic mask gets tiled across n_layers.
+        self._isolate_cache(monkeypatch, tmp_path)
+        mask = [True, False, True, True, False, True, False, False]
+        with open(tmp_path / "swa_cache.json", "w") as f:
+            json.dump({"customarch": mask}, f)
+        b = _backend_from_gguf("customarch", dict(_SWA_FIELDS, block_count = 16))
+        assert b._sliding_window_pattern == [bool(mask[i % 8]) for i in range(16)]
+
+    def test_hf_fetch_populates_cache(self, monkeypatch, tmp_path):
+        self._isolate_cache(monkeypatch, tmp_path)
+        from core.inference import llama_cpp as lc
+
+        calls = []
+
+        def fake_fetch(repo_id):
+            calls.append(repo_id)
+            return 4 if repo_id == "vendor/newmodel-1b-instruct" else None
+
+        monkeypatch.setattr(lc, "_fetch_swa_entry_from_hf", fake_fetch)
+        b = _backend_from_gguf(
+            "newmodel",
+            _SWA_FIELDS,
+            general = {
+                "general.source.huggingface.repository": "vendor/newmodel-1b-instruct"
+            },
+        )
+        assert b._sliding_window_pattern == [(i + 1) % 4 != 0 for i in range(12)]
+        assert calls == ["vendor/newmodel-1b-instruct"]
+        with open(tmp_path / "swa_cache.json") as f:
+            assert json.load(f) == {"newmodel": 4}
+
+    def test_hf_fetch_falls_back_to_other_candidates(self, monkeypatch, tmp_path):
+        self._isolate_cache(monkeypatch, tmp_path)
+        from core.inference import llama_cpp as lc
+
+        monkeypatch.setattr(
+            lc,
+            "_fetch_swa_entry_from_hf",
+            lambda r: 6 if r == "vendor/newmodel-base" else None,
+        )
+        b = _backend_from_gguf(
+            "newmodel",
+            _SWA_FIELDS,
+            general = {
+                "general.base_model.0.repo_url": "https://huggingface.co/vendor/newmodel-base"
+            },
+        )
+        assert b._sliding_window_pattern == [(i + 1) % 6 != 0 for i in range(12)]
+
+    def test_offline_env_skips_network(self, monkeypatch, tmp_path):
+        self._isolate_cache(monkeypatch, tmp_path)
+        monkeypatch.setenv("UNSLOTH_STUDIO_OFFLINE", "1")
+        from core.inference import llama_cpp as lc
+
+        def boom(*a, **kw):
+            raise AssertionError("HF fetch must not run when offline=1")
+
+        monkeypatch.setattr(lc, "_fetch_swa_entry_from_hf", boom)
+        b = _backend_from_gguf(
+            "newmodel",
+            _SWA_FIELDS,
+            general = {"general.source.huggingface.repository": "vendor/newmodel"},
+        )
+        assert b._sliding_window_pattern is None
+
+    def test_hf_fetch_failure_falls_through_silently(self, monkeypatch, tmp_path):
+        self._isolate_cache(monkeypatch, tmp_path)
+        from core.inference import llama_cpp as lc
+
+        monkeypatch.setattr(lc, "_fetch_swa_entry_from_hf", lambda repo_id: None)
+        # Force the failure into the Tier 3 path; bypass Tier 2.5.
+        monkeypatch.setattr(
+            lc, "_resolve_swa_entry_from_transformers", lambda arch: None
+        )
+        b = _backend_from_gguf(
+            "newmodel",
+            _SWA_FIELDS,
+            general = {"general.source.huggingface.repository": "vendor/does-not-exist"},
+        )
+        assert b._sliding_window_pattern is None
+        assert not (tmp_path / "swa_cache.json").exists()
+
+
+class TestTransformersIntrospection:
+    """Tier 2.5: default-init the matching Config; on failure, parse via inspect."""
+
+    def _isolate_cache(self, monkeypatch, tmp_path):
+        from core.inference import llama_cpp as lc
+
+        monkeypatch.setenv("UNSLOTH_STUDIO_HOME", str(tmp_path))
+        monkeypatch.setattr(lc, "_SWA_CACHE", None)
+        return tmp_path
+
+    def test_arch_aliases_normalises_hyphen_underscore(self):
+        from core.inference.llama_cpp import _arch_aliases
+
+        aliases = _arch_aliases("falcon-h1")
+        assert aliases[0] == "falcon-h1" and "falcon_h1" in aliases
+        assert _arch_aliases("gemma3") == ("gemma3",)
+        assert _arch_aliases("") == ()
+
+    def test_resolves_real_transformers_arches(self):
+        from core.inference.llama_cpp import _resolve_swa_entry_from_transformers
+
+        assert _resolve_swa_entry_from_transformers("gemma3") == 6
+        assert _resolve_swa_entry_from_transformers("gemma2") == 2
+        assert _resolve_swa_entry_from_transformers("cohere2") == 4
+
+    def test_falls_back_to_inspect_when_default_init_raises(self, monkeypatch):
+        from core.inference import llama_cpp as lc
+
+        class _FakeBrokenConfig:
+            """Class with sliding_window_pattern: int = 7 in its docstring."""
+
+            def __init__(self, required_arg):
+                raise TypeError("requires an argument")
+
+        class _FakeLazyMapping(dict):
+            def __getitem__(self, k):
+                return (
+                    _FakeBrokenConfig if k == "brokenarch" else super().__getitem__(k)
+                )
+
+        import sys, types as _types
+
+        fake_auto = _types.ModuleType("transformers.models.auto.configuration_auto")
+        fake_auto.CONFIG_MAPPING_NAMES = {"brokenarch": "FakeBroken"}
+        fake_auto.CONFIG_MAPPING = _FakeLazyMapping({"brokenarch": "FakeBroken"})
+        monkeypatch.setitem(
+            sys.modules, "transformers.models.auto.configuration_auto", fake_auto
+        )
+        assert lc._resolve_swa_entry_from_transformers("brokenarch") == 7
+
+    def test_returns_none_when_transformers_unavailable(self, monkeypatch):
+        from core.inference import llama_cpp as lc
+        import sys
+
+        orig_import = (
+            __builtins__["__import__"]
+            if isinstance(__builtins__, dict)
+            else __builtins__.__import__
+        )
+
+        def fake_import(name, *a, **kw):
+            if name.startswith("transformers"):
+                raise ImportError("transformers not installed")
+            return orig_import(name, *a, **kw)
+
+        monkeypatch.setattr("builtins.__import__", fake_import)
+        for k in list(sys.modules):
+            if k.startswith("transformers"):
+                monkeypatch.delitem(sys.modules, k, raising = False)
+        assert lc._resolve_swa_entry_from_transformers("gemma3") is None
+
+    def test_returns_none_for_arch_unknown_to_transformers(self):
+        from core.inference.llama_cpp import _resolve_swa_entry_from_transformers
+
+        assert _resolve_swa_entry_from_transformers("totally-fake-arch-xyz") is None
+
+    def test_full_resolver_uses_transformers_before_hf_fetch(
+        self, monkeypatch, tmp_path
+    ):
+        # With bootstrap empty, Tier 2.5 must answer before Tier 3 fires.
+        self._isolate_cache(monkeypatch, tmp_path)
+        from core.inference import llama_cpp as lc
+
+        monkeypatch.setattr(lc, "_BOOTSTRAP_SWA_DEFAULTS", {})
+
+        def boom(repo_id):
+            raise AssertionError("Tier 3 must not run when Tier 2.5 has the answer")
+
+        monkeypatch.setattr(lc, "_fetch_swa_entry_from_hf", boom)
+        b = _backend_from_gguf(
+            "gemma3",
+            dict(_SWA_FIELDS, block_count = 18),
+            general = {"general.source.huggingface.repository": "google/gemma-3-1b-it"},
+        )
+        assert b._sliding_window_pattern == [(i + 1) % 6 != 0 for i in range(18)]
+        with open(tmp_path / "swa_cache.json") as f:
+            assert json.load(f) == {"gemma3": 6}
+
+
 class TestGGUFParserReset:
     """Verify that fields are properly reset between parses."""
 
@@ -209,11 +695,19 @@ class TestGGUFParserReset:
                 "block_count": 32,
                 "attention.key_length": 128,
                 "attention.kv_lora_rank": 512,
+                "attention.head_count_kv": [8, 2],
+                "attention.sliding_window_pattern": [True, False],
+                "attention.key_length_swa": 64,
+                "attention.value_length_swa": 64,
                 "ssm.inner_size": 4096,
             },
         )
         assert b._kv_key_length == 128
         assert b._kv_lora_rank == 512
+        assert b._n_kv_heads_by_layer == [8, 2]
+        assert b._sliding_window_pattern == [True, False]
+        assert b._kv_key_length_swa == 64
+        assert b._kv_value_length_swa == 64
         assert b._ssm_inner_size == 4096
 
         # Second parse without those fields -- they should be None
@@ -230,6 +724,10 @@ class TestGGUFParserReset:
             os.unlink(path)
         assert b._kv_key_length is None
         assert b._kv_lora_rank is None
+        assert b._n_kv_heads_by_layer is None
+        assert b._sliding_window_pattern is None
+        assert b._kv_key_length_swa is None
+        assert b._kv_value_length_swa is None
         assert b._ssm_inner_size is None
         assert b._n_layers == 64
 
@@ -455,7 +953,9 @@ class TestSlidingWindowEstimation:
         n_global = max(1, 62 // 4)  # 15
         n_swa = 62 - n_global  # 47
         kv_per = 16 * (128 + 128) * 2
-        expected = int(n_global * 131072 * kv_per + n_swa * min(131072, 1024) * kv_per)
+        # SWA cache is double-buffered: 2 * sliding_window cells, capped at n_ctx.
+        swa_cells = min(131072, 2 * 1024)
+        expected = int(n_global * 131072 * kv_per + n_swa * swa_cells * kv_per)
         assert b._estimate_kv_cache_bytes(131072, "f16") == expected
 
     def test_gpt_oss(self):
@@ -472,27 +972,52 @@ class TestSlidingWindowEstimation:
         n_global = max(1, 24 // 4)  # 6
         n_swa = 24 - n_global  # 18
         kv_per = 8 * (64 + 64) * 2
-        expected = int(n_global * 131072 * kv_per + n_swa * min(131072, 128) * kv_per)
+        swa_cells = min(131072, 2 * 128)
+        expected = int(n_global * 131072 * kv_per + n_swa * swa_cells * kv_per)
         assert b._estimate_kv_cache_bytes(131072, "f16") == expected
 
+    def test_gemma4_per_layer_swa_metadata(self):
+        b = self._swa_backend(
+            _n_layers = 30,
+            _n_kv_heads = None,
+            _n_kv_heads_by_layer = [8, 8, 8, 8, 8, 2] * 5,
+            _n_heads = 16,
+            _embedding_length = 2816,
+            _kv_key_length = 512,
+            _kv_value_length = 512,
+            _sliding_window = 1024,
+            _sliding_window_pattern = [True, True, True, True, True, False] * 5,
+            _kv_key_length_swa = 256,
+            _kv_value_length_swa = 256,
+        )
+
+        full_layers = 5
+        sliding_layers = 25
+
+        def expected(ctx):
+            full = full_layers * ctx * 2 * (512 + 512) * 2
+            sliding = sliding_layers * min(ctx, 2 * 1024) * 8 * (256 + 256) * 2
+            return int(full + sliding)
+
+        for ctx in (4096, 46500, 262144):
+            assert b._estimate_kv_cache_bytes(ctx, "f16") == expected(ctx)
+
     def test_ctx_smaller_than_window(self):
-        """When context < sliding_window, SWA layers use full context anyway."""
+        """When context < 2 * sliding_window, SWA cache caps at ctx."""
         b = self._swa_backend(_sliding_window = 8192)
         n_global = max(1, 62 // 4)  # 15
         n_swa = 62 - n_global  # 47
         kv_per = 16 * (128 + 128) * 2
         ctx = 4096
-        expected = int(n_global * ctx * kv_per + n_swa * min(ctx, 8192) * kv_per)
-        # min(4096, 8192) = 4096, so both pools use full ctx
+        expected = int(n_global * ctx * kv_per + n_swa * min(ctx, 2 * 8192) * kv_per)
         assert b._estimate_kv_cache_bytes(ctx, "f16") == expected
 
     def test_odd_layer_count(self):
-        """Odd layer count: n_global = max(1, n//4), n_swa = n - n_global."""
         b = self._swa_backend(_n_layers = 63)
         n_global = max(1, 63 // 4)  # 15
         n_swa = 63 - n_global  # 48
         kv_per = 16 * (128 + 128) * 2
-        expected = int(n_global * 1000 * kv_per + n_swa * min(1000, 1024) * kv_per)
+        expected = int(n_global * 1000 * kv_per + n_swa * min(1000, 2 * 1024) * kv_per)
         assert b._estimate_kv_cache_bytes(1000, "f16") == expected
 
 
@@ -785,6 +1310,686 @@ class TestEdgeCases:
         assert result == expected
 
 
+# ---------------------------------------------------------------------------
+# J2. Server-flag knobs (--swa-full, --kv-unified/--parallel,
+#     --ctx-checkpoints, --kv-offload)
+# ---------------------------------------------------------------------------
+
+
+class TestServerFlags:
+    """Estimator should mirror llama-server CLI flags that change KV size."""
+
+    def _swa_backend(self, **overrides):
+        defaults = {
+            "_n_layers": 26,
+            "_n_kv_heads": 4,
+            "_n_heads": 8,
+            "_embedding_length": 1152,
+            "_kv_key_length": 256,
+            "_kv_value_length": 256,
+            "_sliding_window": 512,
+            "_sliding_window_pattern": [True, True, True, True, True, False] * 4
+            + [True, True],
+        }
+        defaults.update(overrides)
+        b = LlamaCppBackend()
+        for k, v in defaults.items():
+            setattr(b, k, v)
+        return b
+
+    def _gqa_backend(self, **overrides):
+        defaults = {
+            "_n_layers": 28,
+            "_n_kv_heads": 8,
+            "_n_heads": 16,
+            "_embedding_length": 1024,
+            "_kv_key_length": 128,
+            "_kv_value_length": 128,
+        }
+        defaults.update(overrides)
+        b = LlamaCppBackend()
+        for k, v in defaults.items():
+            setattr(b, k, v)
+        return b
+
+    # ── --swa-full ──────────────────────────────────────────────────
+
+    def test_swa_full_collapses_pattern_path_to_full_ctx(self):
+        b = self._swa_backend()
+        ctx = 32_768
+        flagged = b._estimate_kv_cache_bytes(ctx, "f16", swa_full = True)
+        # With swa_full, every layer caches n_ctx -- equals path 4 sizing.
+        kv_per_token = 4 * (256 + 256) * 2  # n_kv_heads * (k+v) * f16
+        expected = 26 * ctx * kv_per_token
+        assert flagged == expected
+        assert flagged > b._estimate_kv_cache_bytes(ctx, "f16")
+
+    def test_swa_full_collapses_legacy_path_to_full_ctx(self):
+        # No per-layer pattern -> 1/4-global heuristic; swa_full overrides.
+        b = self._swa_backend(_sliding_window_pattern = None)
+        ctx = 16_384
+        flagged = b._estimate_kv_cache_bytes(ctx, "f16", swa_full = True)
+        n_global = max(1, 26 // 4)
+        n_swa = 26 - n_global
+        kv_per = 4 * (256 + 256) * 2
+        # swa_cells == n_ctx when swa_full=True
+        expected = n_global * ctx * kv_per + n_swa * ctx * kv_per
+        assert flagged == expected
+
+    def test_swa_full_no_op_for_non_swa_model(self):
+        b = self._gqa_backend()
+        baseline = b._estimate_kv_cache_bytes(8192, "f16")
+        flagged = b._estimate_kv_cache_bytes(8192, "f16", swa_full = True)
+        assert flagged == baseline
+
+    def test_swa_full_suppresses_checkpoint_term(self):
+        b = self._swa_backend()
+        with_cp = b._estimate_kv_cache_bytes(8192, "f16", ctx_checkpoints = 8)
+        with_cp_full = b._estimate_kv_cache_bytes(
+            8192, "f16", ctx_checkpoints = 8, swa_full = True
+        )
+        no_cp_full = b._estimate_kv_cache_bytes(8192, "f16", swa_full = True)
+        # Checkpoints only matter when SWA layers don't already keep n_ctx.
+        assert with_cp_full == no_cp_full
+        assert with_cp > b._estimate_kv_cache_bytes(8192, "f16")
+
+    # ── --parallel + --kv-unified ──────────────────────────────────
+    # Empirically verified against llama-server: non-SWA caches partition
+    # n_ctx across slots (total memory constant); SWA layers are the only
+    # portion that scales with --parallel.  --kv-unified is currently a
+    # no-op for memory math (kept for API forward-compat).
+
+    def test_gqa_kv_constant_across_parallel(self):
+        b = self._gqa_backend()
+        baseline = b._estimate_kv_cache_bytes(4096, "f16")
+        for slots in (1, 2, 4, 8):
+            for unified in (True, False):
+                assert (
+                    b._estimate_kv_cache_bytes(
+                        4096, "f16", n_parallel = slots, kv_unified = unified
+                    )
+                    == baseline
+                )
+
+    def test_zero_parallel_floors_at_one(self):
+        b = self._gqa_backend()
+        baseline = b._estimate_kv_cache_bytes(4096, "f16")
+        for unified in (True, False):
+            assert (
+                b._estimate_kv_cache_bytes(
+                    4096, "f16", n_parallel = 0, kv_unified = unified
+                )
+                == baseline
+            )
+
+    def test_swa_path_scales_only_swa_portion(self):
+        b = self._swa_backend()
+        ctx = 8192
+        baseline = b._estimate_kv_cache_bytes(ctx, "f16")
+        # Decompose baseline by walking the same loop the estimator does.
+        swa = b._sliding_window
+        per_token_global = 4 * (256 + 256) * 2  # n_kv * (k+v) * f16
+        per_token_swa = 4 * (256 + 256) * 2  # k_swa/val_swa fall back
+        per_slot_swa_cells = min(ctx, 2 * swa)  # not clamped at parallel=1
+        global_bytes = sum(
+            ctx * per_token_global
+            for f in b._sliding_window_pattern[: b._n_layers]
+            if not f
+        )
+        swa_bytes_per_slot = sum(
+            per_slot_swa_cells * per_token_swa
+            for f in b._sliding_window_pattern[: b._n_layers]
+            if f
+        )
+        # Sanity: parallel=1 reproduces baseline exactly
+        assert global_bytes + swa_bytes_per_slot == baseline
+        # Only SWA portion scales by parallel
+        for slots in (1, 2, 3, 4):
+            scaled = b._estimate_kv_cache_bytes(
+                ctx, "f16", n_parallel = slots, kv_unified = False
+            )
+            # SWA cells get clamped to per_slot_ctx when ctx/slots < 2*swa
+            per_slot_ctx = max(1, ctx // slots)
+            cells = min(ctx, 2 * swa, per_slot_ctx)
+            swa_bps = sum(
+                cells * per_token_swa
+                for f in b._sliding_window_pattern[: b._n_layers]
+                if f
+            )
+            assert scaled == global_bytes + slots * swa_bps
+
+    def test_mla_kv_constant_across_parallel(self):
+        b = LlamaCppBackend()
+        b._n_layers = 60
+        b._n_kv_heads = 1
+        b._kv_lora_rank = 512
+        b._key_length_mla = 64
+        b._kv_key_length = 576
+        baseline = b._estimate_kv_cache_bytes(8192, "f16")
+        for slots in (1, 2, 4, 8):
+            for unified in (True, False):
+                assert (
+                    b._estimate_kv_cache_bytes(
+                        8192, "f16", n_parallel = slots, kv_unified = unified
+                    )
+                    == baseline
+                )
+
+    # ── --ctx-checkpoints ──────────────────────────────────────────
+
+    def test_ctx_checkpoints_zero_is_no_op(self):
+        b = self._swa_backend()
+        baseline = b._estimate_kv_cache_bytes(8192, "f16")
+        assert b._estimate_kv_cache_bytes(8192, "f16", ctx_checkpoints = 0) == baseline
+
+    def test_ctx_checkpoints_no_op_for_non_swa(self):
+        b = self._gqa_backend()
+        baseline = b._estimate_kv_cache_bytes(8192, "f16")
+        assert b._estimate_kv_cache_bytes(8192, "f16", ctx_checkpoints = 32) == baseline
+
+    def test_ctx_checkpoints_pattern_path_adds_known_bytes(self):
+        b = self._swa_backend()
+        ctx = 8192
+        baseline = b._estimate_kv_cache_bytes(ctx, "f16")
+        flagged = b._estimate_kv_cache_bytes(ctx, "f16", ctx_checkpoints = 4)
+        # 22 SWA layers * 4 checkpoints * 512 cells * 4 heads * (256+256) * 2 bytes
+        n_swa_layers = sum(
+            1 for f in [True, True, True, True, True, False] * 4 + [True, True] if f
+        )
+        per_layer = 4 * 512 * 4 * (256 + 256) * 2
+        assert flagged == baseline + n_swa_layers * per_layer
+
+    def test_ctx_checkpoints_legacy_path_adds_known_bytes(self):
+        b = self._swa_backend(_sliding_window_pattern = None)
+        ctx = 8192
+        baseline = b._estimate_kv_cache_bytes(ctx, "f16")
+        flagged = b._estimate_kv_cache_bytes(ctx, "f16", ctx_checkpoints = 4)
+        n_global = max(1, 26 // 4)
+        n_swa = 26 - n_global
+        kv_per = 4 * (256 + 256) * 2
+        extra = 4 * n_swa * 512 * kv_per  # ctx_checkpoints * n_swa * sliding * kv_per
+        assert flagged == baseline + extra
+
+    def test_ctx_checkpoints_compose_with_n_parallel(self):
+        # Only the SWA + checkpoint portion scales by n_parallel; the
+        # global-layer portion stays constant.
+        b = self._swa_backend()
+        ctx = 8192
+        swa = b._sliding_window
+        per_token = 4 * (256 + 256) * 2
+        global_bytes = sum(
+            ctx * per_token for f in b._sliding_window_pattern[: b._n_layers] if not f
+        )
+        n_swa_layers = sum(1 for f in b._sliding_window_pattern[: b._n_layers] if f)
+        slots = 3
+        per_slot_ctx = max(1, ctx // slots)
+        swa_cells = min(ctx, 2 * swa, per_slot_ctx)
+        swa_bytes_per_slot = n_swa_layers * swa_cells * per_token
+        cp_extra_per_slot = n_swa_layers * 4 * swa * per_token  # 4 checkpoints
+        flagged = b._estimate_kv_cache_bytes(
+            ctx, "f16", ctx_checkpoints = 4, n_parallel = slots, kv_unified = False
+        )
+        assert flagged == global_bytes + slots * (
+            swa_bytes_per_slot + cp_extra_per_slot
+        )
+
+    # ── --kv-offload (kv_on_gpu) ───────────────────────────────────
+
+    def test_fit_returns_requested_when_kv_off_gpu(self):
+        b = self._gqa_backend()
+        # Tiny VRAM budget -- normally would force a reduction.
+        fitted = b._fit_context_to_vram(
+            requested_ctx = 32_768,
+            available_mib = 1,
+            model_size_bytes = 100,
+            cache_type_kv = "f16",
+            kv_on_gpu = False,
+        )
+        assert fitted == 32_768
+
+    def test_fit_reduces_when_kv_on_gpu(self):
+        b = self._gqa_backend()
+        fitted = b._fit_context_to_vram(
+            requested_ctx = 32_768,
+            available_mib = 64,
+            model_size_bytes = 1024 * 1024,  # 1 MiB
+            cache_type_kv = "f16",
+            kv_on_gpu = True,
+        )
+        assert fitted < 32_768
+
+    def test_fit_threads_swa_full_through_estimator(self):
+        # SWA model, generous budget; both should fit but cache size differs.
+        b = self._swa_backend()
+        ctx = 8192
+        kv_default = b._estimate_kv_cache_bytes(ctx, "f16")
+        kv_full = b._estimate_kv_cache_bytes(ctx, "f16", swa_full = True)
+        assert kv_full > kv_default
+        # Budget = model + kv_default (rounded up) -- swa_full should not fit.
+        budget_mib = (1024 * 1024 + kv_default) / (1024 * 1024) / 0.90 + 1
+        fitted_default = b._fit_context_to_vram(
+            requested_ctx = ctx,
+            available_mib = int(budget_mib),
+            model_size_bytes = 1024 * 1024,
+            cache_type_kv = "f16",
+        )
+        fitted_full = b._fit_context_to_vram(
+            requested_ctx = ctx,
+            available_mib = int(budget_mib),
+            model_size_bytes = 1024 * 1024,
+            cache_type_kv = "f16",
+            swa_full = True,
+        )
+        assert fitted_default == ctx
+        assert fitted_full < ctx
+
+
+# ---------------------------------------------------------------------------
+# J2.5. --parallel N memory accounting (per-layer-type scaling rule)
+# ---------------------------------------------------------------------------
+
+
+class TestParallelSWAScaling:
+    """Verifies the per-layer-type scaling rule against the closed form
+    measured from llama-server. Empirical formula on Gemma-3 270m at
+    ctx=8192: total_kv = 24 + parallel * 15 (MiB).
+
+    Rule (verified vs ``llama-server`` log on real GGUFs):
+      * non-SWA layers: total cells = n_ctx, partitioned across slots,
+        memory CONSTANT in n_parallel.
+      * SWA layers: per-slot cells = 2 * sliding_window (clamped at
+        n_ctx and at per_slot_ctx); memory LINEAR in n_parallel.
+      * --kv-unified is a no-op for memory math; both modes yield the
+        same total in measured cases.
+    """
+
+    def _gqa_backend(self, **overrides):
+        defaults = {
+            "_n_layers": 28,
+            "_n_kv_heads": 8,
+            "_n_heads": 16,
+            "_embedding_length": 1024,
+            "_kv_key_length": 128,
+            "_kv_value_length": 128,
+        }
+        defaults.update(overrides)
+        b = LlamaCppBackend()
+        for k, v in defaults.items():
+            setattr(b, k, v)
+        return b
+
+    def _swa_backend(self, **overrides):
+        defaults = {
+            "_n_layers": 18,
+            "_n_kv_heads": 1,
+            "_n_heads": 4,
+            "_embedding_length": 1024,
+            "_kv_key_length": 256,
+            "_kv_value_length": 256,
+            "_sliding_window": 512,
+            # 15 SWA + 3 global, mirrors gemma-3-270m
+            "_sliding_window_pattern": [
+                t == "swa" for t in (["swa"] * 5 + ["global"]) * 3
+            ],
+        }
+        defaults.update(overrides)
+        b = LlamaCppBackend()
+        for k, v in defaults.items():
+            setattr(b, k, v)
+        return b
+
+    # ── non-SWA paths: constant ────────────────────────────────────
+
+    def test_pure_gqa_constant_across_parallel(self):
+        b = self._gqa_backend()
+        baseline = b._estimate_kv_cache_bytes(8192, "f16")
+        for slots in (1, 2, 4, 8):
+            for unified in (True, False):
+                assert (
+                    b._estimate_kv_cache_bytes(
+                        8192, "f16", n_parallel = slots, kv_unified = unified
+                    )
+                    == baseline
+                )
+
+    def test_mla_constant_across_parallel(self):
+        b = LlamaCppBackend()
+        b._n_layers = 60
+        b._n_kv_heads = 1
+        b._kv_lora_rank = 512
+        b._key_length_mla = 64
+        b._kv_key_length = 576
+        baseline = b._estimate_kv_cache_bytes(8192, "f16")
+        for slots in (1, 2, 4, 8):
+            assert b._estimate_kv_cache_bytes(8192, "f16", n_parallel = slots) == baseline
+
+    def test_hybrid_constant_across_parallel(self):
+        b = LlamaCppBackend()
+        b._n_layers = 64
+        b._n_kv_heads = 16
+        b._n_heads = 32
+        b._embedding_length = 4096
+        b._kv_key_length = 128
+        b._kv_value_length = 128
+        b._ssm_inner_size = 4096
+        b._full_attention_interval = 4
+        baseline = b._estimate_kv_cache_bytes(8192, "f16")
+        for slots in (1, 2, 4, 8):
+            assert b._estimate_kv_cache_bytes(8192, "f16", n_parallel = slots) == baseline
+
+    def test_legacy_constant_across_parallel(self):
+        b = LlamaCppBackend()
+        b._n_layers = 32
+        b._n_kv_heads = 8
+        b._n_heads = 8
+        b._embedding_length = 4096
+        baseline = b._estimate_kv_cache_bytes(8192, "f16")
+        for slots in (1, 2, 4, 8):
+            assert b._estimate_kv_cache_bytes(8192, "f16", n_parallel = slots) == baseline
+
+    # ── SWA paths: scale only the SWA portion ──────────────────────
+
+    def test_swa_pattern_scales_only_swa_portion(self):
+        b = self._swa_backend()
+        ctx = 8192
+        swa = b._sliding_window
+        per_token = 1 * (256 + 256) * 2  # n_kv * (k+v) * f16
+        n_global = sum(1 for f in b._sliding_window_pattern if not f)
+        n_swa = sum(1 for f in b._sliding_window_pattern if f)
+        global_bytes = n_global * ctx * per_token
+        for slots in (1, 2, 4, 8):
+            per_slot_ctx = max(1, ctx // slots)
+            cells = min(ctx, 2 * swa, per_slot_ctx)
+            swa_bps = n_swa * cells * per_token
+            for unified in (True, False):
+                got = b._estimate_kv_cache_bytes(
+                    ctx, "f16", n_parallel = slots, kv_unified = unified
+                )
+                assert got == global_bytes + slots * swa_bps
+
+    def test_swa_fallback_scales_only_swa_portion(self):
+        # No per-layer pattern -> 1/4-global heuristic.
+        b = self._swa_backend(_sliding_window_pattern = None)
+        ctx = 8192
+        swa = b._sliding_window
+        n_layers = 18
+        n_global = max(1, n_layers // 4)
+        n_swa = n_layers - n_global
+        per_token = 1 * (256 + 256) * 2
+        global_bytes = n_global * ctx * per_token
+        for slots in (1, 2, 4, 8):
+            per_slot_ctx = max(1, ctx // slots)
+            cells = min(ctx, 2 * swa, per_slot_ctx)
+            swa_bps = n_swa * cells * per_token
+            got = b._estimate_kv_cache_bytes(ctx, "f16", n_parallel = slots)
+            assert got == global_bytes + slots * swa_bps
+
+    def test_swa_per_slot_clamped_when_ctx_lt_slots_x_2window(self):
+        # ctx=4096 / slots=8 -> per_slot_ctx=512, but 2*sliding=1024.
+        # SWA cells should clamp at per_slot_ctx (512), not 2*sliding.
+        b = self._swa_backend()
+        ctx = 4096
+        per_slot_ctx_at_8 = ctx // 8
+        assert per_slot_ctx_at_8 < 2 * b._sliding_window
+        # Build expected with the clamped formula
+        n_swa = sum(1 for f in b._sliding_window_pattern if f)
+        n_global = sum(1 for f in b._sliding_window_pattern if not f)
+        per_token = 1 * (256 + 256) * 2
+        global_bytes = n_global * ctx * per_token
+        cells = min(ctx, 2 * b._sliding_window, per_slot_ctx_at_8)
+        assert cells == per_slot_ctx_at_8
+        expected = global_bytes + 8 * (n_swa * cells * per_token)
+        assert b._estimate_kv_cache_bytes(ctx, "f16", n_parallel = 8) == expected
+
+    def test_swa_full_does_not_scale_under_parallel(self):
+        # swa_full forces every layer to n_ctx; result is the all-global
+        # GQA-style total, which is constant in parallel.
+        b = self._swa_backend()
+        ctx = 8192
+        baseline = b._estimate_kv_cache_bytes(ctx, "f16", swa_full = True)
+        for slots in (1, 2, 4, 8):
+            assert (
+                b._estimate_kv_cache_bytes(ctx, "f16", swa_full = True, n_parallel = slots)
+                == baseline
+            )
+
+    # ── kv_unified: no-op for memory math ──────────────────────────
+
+    def test_kv_unified_is_no_op_for_memory_math(self):
+        # Both unified=True and unified=False must produce the same
+        # total bytes for every backend type and every parallel value.
+        backends = [
+            ("gqa", self._gqa_backend()),
+            ("swa", self._swa_backend()),
+        ]
+        for label, b in backends:
+            for slots in (1, 2, 4, 8):
+                u = b._estimate_kv_cache_bytes(
+                    8192, "f16", n_parallel = slots, kv_unified = True
+                )
+                nu = b._estimate_kv_cache_bytes(
+                    8192, "f16", n_parallel = slots, kv_unified = False
+                )
+                assert u == nu, f"{label} parallel={slots} unified-mismatch"
+
+    # ── Empirical Gemma-3 270m formula ─────────────────────────────
+
+    def test_matches_empirical_gemma3_270m_formula(self):
+        """Exact match against the formula measured from llama-server:
+        total_kv = 24 + parallel * 15 (MiB) at ctx=8192.
+
+        Geometry: 18 layers (3 global + 15 SWA), n_kv=1, head_dim=256,
+        sliding=512, f16.
+        """
+        b = LlamaCppBackend()
+        b._n_layers = 18
+        b._n_kv_heads = 1
+        b._n_heads = 4
+        b._embedding_length = 1024
+        b._kv_key_length = 256
+        b._kv_value_length = 256
+        b._sliding_window = 512
+        # 5-period [swa,swa,swa,swa,full] * 3 + [swa,swa,swa]: mirrors the
+        # bootstrap-resolved pattern for gemma3 (period 6) on an 18-layer
+        # model (15 SWA, 3 global).
+        b._sliding_window_pattern = [(i + 1) % 6 != 0 for i in range(18)]
+        n_global = 3
+        n_swa = 15
+        # Confirm pattern shape
+        assert sum(b._sliding_window_pattern) == n_swa
+        for slots, expected_mib in [(1, 39), (2, 54), (4, 84)]:
+            got_bytes = b._estimate_kv_cache_bytes(8192, "f16", n_parallel = slots)
+            got_mib = got_bytes / (1024 * 1024)
+            assert (
+                got_mib == expected_mib
+            ), f"slots={slots}: got {got_mib} MiB, expected {expected_mib} MiB"
+
+
+# ---------------------------------------------------------------------------
+# J3. shared_kv_layers (Gemma 3n / Gemma 4)
+# ---------------------------------------------------------------------------
+
+
+class TestSharedKVLayers:
+    """``<arch>.attention.shared_kv_layers`` reduces the layer count that
+    actually allocates KV.  The trailing ``shared_kv_layers`` blocks reuse
+    earlier caches (Gemma 3n: 35 layers, 15 shared -> 20 allocate; Gemma 4
+    same field).  Unset on every other arch -> no behavioural change."""
+
+    def _gemma3n_backend(self, **overrides):
+        # Mirrors google/gemma-3n-E4B-it: 35 layers, 15 shared,
+        # SWA window 1024, period 5 (4 sliding + 1 full repeating).
+        defaults = {
+            "_n_layers": 35,
+            "_n_kv_heads": 4,
+            "_n_heads": 8,
+            "_embedding_length": 2048,
+            "_kv_key_length": 256,
+            "_kv_value_length": 256,
+            "_sliding_window": 1024,
+            "_sliding_window_pattern": [
+                t == "sliding_attention"
+                for t in (["sliding_attention"] * 4 + ["full_attention"]) * 7
+            ],
+            "_shared_kv_layers": 15,
+        }
+        defaults.update(overrides)
+        b = LlamaCppBackend()
+        for k, v in defaults.items():
+            setattr(b, k, v)
+        return b
+
+    def _gqa_backend(self, **overrides):
+        defaults = {
+            "_n_layers": 28,
+            "_n_kv_heads": 8,
+            "_n_heads": 16,
+            "_embedding_length": 1024,
+            "_kv_key_length": 128,
+            "_kv_value_length": 128,
+        }
+        defaults.update(overrides)
+        b = LlamaCppBackend()
+        for k, v in defaults.items():
+            setattr(b, k, v)
+        return b
+
+    def test_field_initialises_to_none(self):
+        b = LlamaCppBackend()
+        assert b._shared_kv_layers is None
+
+    def test_unset_field_is_noop(self):
+        b = self._gqa_backend()
+        baseline = b._estimate_kv_cache_bytes(8192, "f16")
+        b._shared_kv_layers = None
+        assert b._estimate_kv_cache_bytes(8192, "f16") == baseline
+        b._shared_kv_layers = 0
+        assert b._estimate_kv_cache_bytes(8192, "f16") == baseline
+
+    def test_path4_drops_shared_layers(self):
+        b = self._gqa_backend(_shared_kv_layers = 4)
+        ctx = 4096
+        kv_per = 8 * (128 + 128) * 2
+        # 28 - 4 = 24 layers actually allocate
+        assert b._estimate_kv_cache_bytes(ctx, "f16") == 24 * ctx * kv_per
+
+    def test_path5_drops_shared_layers(self):
+        b = LlamaCppBackend()
+        b._n_layers = 32
+        b._n_kv_heads = 8
+        b._n_heads = 8
+        b._embedding_length = 4096
+        b._shared_kv_layers = 8
+        ctx = 4096
+        head_dim = 4096 // 8  # 512
+        # 32 - 8 = 24 layers
+        expected = 2 * 8 * head_dim * 24 * ctx * 2
+        assert b._estimate_kv_cache_bytes(ctx, "f16") == expected
+
+    def test_path1_mla_drops_shared_layers(self):
+        b = LlamaCppBackend()
+        b._n_layers = 60
+        b._n_kv_heads = 1
+        b._kv_lora_rank = 512
+        b._key_length_mla = 64
+        b._kv_key_length = 576
+        b._shared_kv_layers = 10
+        ctx = 8192
+        # 60 - 10 = 50
+        assert b._estimate_kv_cache_bytes(ctx, "f16") == 50 * ctx * 1 * 576 * 2
+
+    def test_path3_pattern_loops_only_unshared_layers(self):
+        b = self._gemma3n_backend()
+        ctx = 8192
+        # First 20 layers contribute; layers 20..34 are skipped.
+        # Pattern: [s,s,s,s,F] repeated.  In layers 0..19:
+        #   sliding: 16, full: 4
+        sliding_in_unshared = sum(b._sliding_window_pattern[:20])
+        full_in_unshared = 20 - sliding_in_unshared
+        assert sliding_in_unshared == 16
+        assert full_in_unshared == 4
+        kv_per = 4 * (256 + 256) * 2
+        swa_cells = min(ctx, 2 * 1024)
+        expected = (
+            full_in_unshared * ctx * kv_per + sliding_in_unshared * swa_cells * kv_per
+        )
+        assert b._estimate_kv_cache_bytes(ctx, "f16") == expected
+
+    def test_shared_layers_reduces_estimate(self):
+        b = self._gemma3n_backend()
+        with_shared = b._estimate_kv_cache_bytes(8192, "f16")
+        b._shared_kv_layers = 0
+        without_shared = b._estimate_kv_cache_bytes(8192, "f16")
+        # 20/35 = 0.571 of the work; expect ~43% reduction.
+        ratio = with_shared / without_shared
+        assert 0.5 < ratio < 0.65
+
+    def test_path3_pattern_with_swa_full_and_shared(self):
+        b = self._gemma3n_backend()
+        ctx = 8192
+        flagged = b._estimate_kv_cache_bytes(ctx, "f16", swa_full = True)
+        # Every unshared layer caches n_ctx; equals path-4-style sizing
+        # over only the 20 unshared layers.
+        kv_per = 4 * (256 + 256) * 2
+        assert flagged == 20 * ctx * kv_per
+
+    def test_path3_fallback_uses_unshared_count(self):
+        # No per-layer pattern -> 1/4-global heuristic over n_layers_kv,
+        # not n_layers.
+        b = self._gemma3n_backend(_sliding_window_pattern = None)
+        ctx = 8192
+        n_layers_kv = 35 - 15  # 20
+        n_global = max(1, n_layers_kv // 4)  # 5
+        n_swa = n_layers_kv - n_global  # 15
+        kv_per = 4 * (256 + 256) * 2
+        swa_cells = min(ctx, 2 * 1024)
+        expected = n_global * ctx * kv_per + n_swa * swa_cells * kv_per
+        assert b._estimate_kv_cache_bytes(ctx, "f16") == expected
+
+    def test_shared_floors_at_one_layer(self):
+        # Pathological: shared >= n_layers should not zero out the cache.
+        b = self._gqa_backend(_shared_kv_layers = 99)
+        ctx = 4096
+        kv_per = 8 * (128 + 128) * 2
+        assert b._estimate_kv_cache_bytes(ctx, "f16") == 1 * ctx * kv_per
+
+    def test_composes_with_n_parallel(self):
+        # Only the SWA portion of the unshared layers scales by n_parallel;
+        # the global portion stays constant.
+        b = self._gemma3n_backend()
+        ctx = 8192
+        swa = b._sliding_window
+        per_token = 4 * (256 + 256) * 2
+        unshared_pattern = b._sliding_window_pattern[:20]  # 35 - 15 shared
+        sliding_in_unshared = sum(unshared_pattern)
+        global_in_unshared = len(unshared_pattern) - sliding_in_unshared
+        global_bytes = global_in_unshared * ctx * per_token
+        slots = 3
+        per_slot_ctx = max(1, ctx // slots)
+        swa_cells = min(ctx, 2 * swa, per_slot_ctx)
+        swa_bytes_per_slot = sliding_in_unshared * swa_cells * per_token
+        flagged = b._estimate_kv_cache_bytes(
+            ctx, "f16", n_parallel = slots, kv_unified = False
+        )
+        assert flagged == global_bytes + slots * swa_bytes_per_slot
+
+    def test_composes_with_ctx_checkpoints(self):
+        b = self._gemma3n_backend()
+        ctx = 8192
+        baseline = b._estimate_kv_cache_bytes(ctx, "f16")
+        with_cp = b._estimate_kv_cache_bytes(ctx, "f16", ctx_checkpoints = 4)
+        # Checkpoints only count over UNSHARED SWA layers (16 of them).
+        sliding_in_unshared = sum(b._sliding_window_pattern[:20])
+        per_cp_layer = 4 * 1024 * 4 * (256 + 256) * 2  # cps * swa * heads * (k+v) * bpe
+        assert with_cp == baseline + sliding_in_unshared * per_cp_layer
+
+    def test_unload_resets_shared_kv_layers(self):
+        b = LlamaCppBackend()
+        b._shared_kv_layers = 12
+        b.unload_model()
+        assert b._shared_kv_layers is None
+
+
 # ---------------------------------------------------------------------------
 # K. Lifecycle Tests
 # ---------------------------------------------------------------------------
@@ -799,13 +2004,18 @@ class TestLifecycle:
             "_kv_key_length",
             "_kv_value_length",
             "_sliding_window",
+            "_sliding_window_pattern",
             "_full_attention_interval",
             "_kv_lora_rank",
             "_key_length_mla",
+            "_kv_key_length_swa",
+            "_kv_value_length_swa",
             "_ssm_inner_size",
             "_ssm_state_size",
+            "_shared_kv_layers",
         ]:
             assert getattr(b, attr) is None
+        assert b._n_kv_heads_by_layer is None
 
     def test_unload_resets_fields(self):
         b = LlamaCppBackend()
@@ -813,20 +2023,30 @@ class TestLifecycle:
         b._kv_key_length = 128
         b._kv_lora_rank = 512
         b._sliding_window = 1024
+        b._sliding_window_pattern = [True, False]
+        b._n_kv_heads_by_layer = [8, 2]
+        b._kv_key_length_swa = 64
+        b._kv_value_length_swa = 64
         b._ssm_inner_size = 4096
         b._full_attention_interval = 4
+        b._shared_kv_layers = 8
         b.unload_model()
         for attr in [
             "_kv_key_length",
             "_kv_value_length",
             "_sliding_window",
+            "_sliding_window_pattern",
             "_full_attention_interval",
             "_kv_lora_rank",
             "_key_length_mla",
+            "_kv_key_length_swa",
+            "_kv_value_length_swa",
             "_ssm_inner_size",
             "_ssm_state_size",
+            "_shared_kv_layers",
         ]:
             assert getattr(b, attr) is None
+        assert b._n_kv_heads_by_layer is None
 
     def test_end_to_end_synthetic_mla(self):
         """Full round-trip: write GGUF -> parse -> estimate."""
@@ -887,12 +2107,46 @@ class TestLifecycle:
         )
         assert b._can_estimate_kv()
         result = b._estimate_kv_cache_bytes(131072, "f16")
-        n_global = max(1, 62 // 4)  # 15
-        n_swa = 62 - n_global  # 47
+        # gemma3 -> period 6 from the bootstrap table, SWA cache
+        # double-buffered to 2 * sliding_window cells.
+        period = 6
         kv_per = 16 * 256 * 2
-        expected = int(n_global * 131072 * kv_per + n_swa * 1024 * kv_per)
+        expected = 0
+        for i in range(62):
+            is_swa = (i + 1) % period != 0
+            layer_ctx = min(131072, 2 * 1024) if is_swa else 131072
+            expected += layer_ctx * kv_per
         assert result == expected
 
+    def test_end_to_end_synthetic_shared_kv_round_trip(self):
+        # Mirrors gemma3n_text: 35 layers, 15 shared, sliding_window=1024.
+        b = _backend_from_gguf(
+            "gemma3n_text",
+            {
+                "context_length": 32768,
+                "block_count": 35,
+                "attention.head_count_kv": 4,
+                "attention.head_count": 8,
+                "embedding_length": 2048,
+                "attention.key_length": 256,
+                "attention.value_length": 256,
+                "attention.sliding_window": 1024,
+                "attention.shared_kv_layers": 15,
+            },
+        )
+        assert b._can_estimate_kv()
+        assert b._shared_kv_layers == 15
+        # Bootstrap table for gemma3n_text -> period 5; the resolver
+        # synthesises a 35-entry bool array.  The first 20 entries
+        # (n_layers - shared) are the only ones that allocate KV.
+        result = b._estimate_kv_cache_bytes(8192, "f16")
+        assert result > 0
+        # Sanity: setting shared back to 0 must produce a strictly larger
+        # estimate (more layers allocate).
+        b._shared_kv_layers = 0
+        unshared = b._estimate_kv_cache_bytes(8192, "f16")
+        assert unshared > result
+
     def test_end_to_end_synthetic_gqa(self):
         b = _backend_from_gguf(
             "qwen3",
diff --git a/studio/backend/tests/test_llama_cpp_context_fit.py b/studio/backend/tests/test_llama_cpp_context_fit.py
index f498655347..caa6397901 100644
--- a/studio/backend/tests/test_llama_cpp_context_fit.py
+++ b/studio/backend/tests/test_llama_cpp_context_fit.py
@@ -114,9 +114,13 @@ def _make_backend(
     inst._kv_value_length = kv_value_length
     inst._kv_lora_rank = None
     inst._sliding_window = None
+    inst._sliding_window_pattern = None
     inst._ssm_inner_size = None
     inst._full_attention_interval = None
     inst._key_length_mla = None
+    inst._n_kv_heads_by_layer = None
+    inst._kv_key_length_swa = None
+    inst._kv_value_length_swa = None
     return inst
 
 
@@ -137,7 +141,7 @@ def _drive(
     model_size = int(model_gib * GIB)
     cache_type_kv = None
 
-    def fake_estimate(n_ctx_, _type = None):
+    def fake_estimate(n_ctx_, _type = None, **_kwargs):
         return 0 if n_ctx_ <= 0 else n_ctx_ * kv_per_token_bytes
 
     inst._estimate_kv_cache_bytes = fake_estimate
diff --git a/studio/backend/tests/test_llama_cpp_max_context_threshold.py b/studio/backend/tests/test_llama_cpp_max_context_threshold.py
index 5fd0243c9f..22e4cda7d1 100644
--- a/studio/backend/tests/test_llama_cpp_max_context_threshold.py
+++ b/studio/backend/tests/test_llama_cpp_max_context_threshold.py
@@ -99,9 +99,13 @@ def _make_backend(native_ctx = 131072):
     inst._kv_value_length = 128
     inst._kv_lora_rank = None
     inst._sliding_window = None
+    inst._sliding_window_pattern = None
     inst._ssm_inner_size = None
     inst._full_attention_interval = None
     inst._key_length_mla = None
+    inst._n_kv_heads_by_layer = None
+    inst._kv_key_length_swa = None
+    inst._kv_value_length_swa = None
     return inst
 
 
@@ -114,7 +118,7 @@ def _compute_max_available_ctx(native_ctx, model_gib, gpus, kv_per_token_bytes =
     model_size = int(model_gib * GIB)
 
     inst._estimate_kv_cache_bytes = (
-        lambda n, _t = None: 0 if n <= 0 else n * kv_per_token_bytes
+        lambda n, _t = None, **_kw: 0 if n <= 0 else n * kv_per_token_bytes
     )
     inst._can_estimate_kv = lambda: True
 

From 09505fcc6e552d386f461209eb22f34747eef69a Mon Sep 17 00:00:00 2001
From: Datta Nimmaturi <venkatadattasainimmaturi@gmail.com>
Date: Tue, 5 May 2026 16:42:36 +0530
Subject: [PATCH 207/216] Update VRAM estimator to cater to broader model
 configs (#5175)

* Update VRAM estimator to cater to broader model configs

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

* fix attn backend check, better support for MoE etc

* Studio: tighten VRAM estimator structured-shape and attention paths

- Conservative attention fallback: when resolve_attention_implementation
  fails, charge the quadratic non-flash activation path instead of
  silently keeping the optimistic flash_attention_2 default.
- Resolve attention on a shallow config copy so _set_attn_impl does not
  mutate the cached config returned by _load_config_for_gpu_estimate.
- Use getattr for AutoModelForCausalLM._model_mapping to avoid raising
  on private-attribute renames in transformers.
- Treat sdpa as O(n) linear attention; PyTorch SDPA dispatches to flash
  or memory-efficient backends, only eager needs the quadratic term.
- Per-layer activation accounting: structured archs (head_dim,
  layer_types, attention_k_eq_v, num_kv_shared_layers, double-wide MLP)
  now flow into compute_activation_bytes via _text_linear_dims, instead
  of using the legacy hidden_size//num_attention_heads KV/MLP shape.
- Exclude MLA configs (q_lora_rank set) from the structured-shape path
  so q_lora low-rank projection formulas keep applying when head_dim is
  also present.
- _build_text_module_elements emits a single MLA self_attn aggregate
  using _compute_attn_elements when q_lora_rank is set, avoiding the
  ~10% overcount that fed into _compute_skipped_quantizable_elements.
- Restrict _module_path_matches to known text-tower prefixes so VLM
  skip names like vision_tower.model.layers.<i>.self_attn.q_proj no
  longer falsely shadow the text alias model.layers.<i>.self_attn.q_proj.
- Pick up enable_moe_block from the config and add the per-layer dense
  MLP alongside the MoE experts in compute_total_params and
  compute_lora_params (Gemma4-style parallel dense + MoE block).
- Single-pass structured layer accounting in _compute_layer_elements,
  removing the duplicate _text_linear_dims walks.
- Drop the now-zero (activations - activations_computed) shard term in
  VramBreakdown.min_gpu_vram and the stale comment that referred to it.
- attention_implementation typed as Optional[str] to match call sites
  that pass None.
- Inline rationale comments on DOUBLE_QUANT_4BIT_FACTOR and
  NON_FLASH_ATTENTION_FACTOR pointing at VRAM_ESTIMATION.md.

* Studio: extend parallel-MoE accounting + non-prefix dense layer support

- Apply enable_moe_block / moe_has_dense_mlp symmetrically: activation
  per-layer MLP size in _layer_qkv_mlp_sizes now adds the parallel dense
  MLP for MoE layers, matching the weight and LoRA accounting added in
  the prior commit. Skip-quantizable mapping in _build_text_module_elements
  now registers both mlp.experts and per-projection mlp.{name} entries
  for MoE layers when the parallel dense block is present, so an
  llm_int8_skip_modules entry like "model.layers.N.mlp" covers both.
- Track dense layer indices as a tuple (dense_layer_indices) extracted
  from first_k_dense_replace or decoder_sparse_step + mlp_only_layers,
  and dispatch dense-vs-MoE accounting through _is_dense_mlp_layer. The
  prior count-based path silently mis-bucketed layers when mlp_only_layers
  was non-prefix (e.g. [3, 5] on an 8-layer model). num_dense_layers is
  derived from len(dense_layer_indices) for backward compatibility.
- Drop the redundant ">0" check in _is_kv_shared_layer so configs with
  num_kv_shared_layers == num_hidden_layers (every layer shared) are
  correctly recognized as shared.
- Refresh VRAM_ESTIMATION.md section 5 to note that sdpa joins
  flash_attention_2 in the linear activation path; refresh the
  VramBreakdown.activations_computed comment now that the activation
  floor is gone.

* Studio: Gemma4 PLE accounting, flex_attention, KV-share guard restore

- Add flex_attention to LINEAR_ATTENTION_IMPLS. Unsloth's
  resolve_attention_implementation returns "flex_attention" when
  HAS_FLASH_ATTENTION is False and the model class supports flex; PyTorch
  FlexAttention is a memory-efficient kernel, not a quadratic eager
  attention path. Without this, activation estimates over-charge ~36x.
- Restore the `> 0` guard in _is_kv_shared_layer. Transformers Gemma4
  (modeling_gemma4.py:1031, modular_gemma4.py:863, :926) uses
  `layer_idx >= first_kv_shared_layer_idx > 0`, so configs that mark
  every layer as KV-shared raise on construction. Reverting the
  unconditional acceptance avoids producing a detailed estimate for a
  shape the actual model code rejects.
- Extend the parallel dense MLP path (`enable_moe_block`) in
  _build_text_module_elements: when the arch is non-structured, use
  arch.intermediate_size for the dense gate/up/down dims instead of
  _text_linear_dims (which returns moe_intermediate_size via
  _get_mlp_size). Prior code under-counted skipped quantizable elements
  for the parallel dense block by up to 8x on GLM-style configs.
- Add Gemma4 per-layer-input (PLE) module accounting:
  per_layer_model_projection (one global Linear) plus per-layer
  per_layer_input_gate and per_layer_projection are added to the
  quantizable text-linear total in _compute_layer_elements;
  post_per_layer_input_norm and per_layer_projection_norm flow into
  the non-quantizable bucket. compute_lora_params adds the same three
  Linear modules to the all-linear total. References:
  transformers_versions/5.7.0/.../gemma4/modular_gemma4.py:1077-1083,
  :1247-1253.
- VRAM_ESTIMATION.md section 5 now lists flex_attention alongside sdpa
  and flash_attention_2 as linear-memory backends.

* Studio: shared-expert variants, mlp_layer_types dispatch, PLE skip, all-linear str, deepcopy resolver

Five targeted estimator corrections:

- _compute_dense_layer_indices now reads `mlp_layer_types` ahead of
  `first_k_dense_replace` / `decoder_sparse_step`. Transformers Exaone-MoE,
  Laguna, Hy_v3, GLM-MoE-DSA, GLM4-MoE-Lite, Ernie4_5_VL_MoE etc. ship the
  per-position list and may omit the prefix-style fields entirely.
- _build_text_module_elements registers per_layer_input_gate /
  per_layer_projection (per layer) and per_layer_model_projection (global)
  in the canonical element map and alias map. The PLE element count was
  added to total_quantizable in a prior commit but skip-module matching
  against names like model.layers.0.per_layer_input_gate produced 0-byte
  delta. Layer aggregate text.layers.<i> now sums all layer modules so
  prefix skip names cover the PLE pieces too.
- _targets_all_linear coerces a bare string `"all-linear"` to `["all-linear"]`
  before set comparison; the previous set comprehension iterated chars.
  PEFT LoraConfig.target_modules accepts the bare-string convention.
- ModelArchConfig gains `shared_expert_intermediate_size`. extract_arch_config
  reads `n_shared_experts` / `num_shared_experts` aliases and infers
  `n_shared_experts=1` when only `shared_expert_intermediate_size` is set.
  _compute_moe_mlp_elements and the structured + non-structured LoRA paths
  size the shared expert with its own intermediate (Qwen3.5-MoE: 512 vs
  routed moe_intermediate_size).
- _determine_attention_impl_for_gpu_estimate uses copy.deepcopy so the
  resolver does not mutate nested text_config on the cached source.
  PreTrainedConfig._attn_implementation setter walks `sub_configs` and the
  prior shallow copy still touched the inner objects.

* Studio: extend MoE/PLE/KV-share accounting to activation and skip-alias paths

Five activation-path corrections plus two LoRA / skip-alias corrections so
that shared-expert, per-layer-input, and KV-shared-layer support is symmetric
across weights, LoRA, skip-quantizable, and activation paths.

- _layer_qkv_mlp_sizes: include shared-expert FFN in mlp_size (live shared
  expert per token alongside routed experts) and keep K/V activation memory
  for KV-shared layers; only the WEIGHT path uses has_k/has_v from
  _layer_attention_dims.
- _per_layer_activation_bytes / compute_activation_bytes: account for
  per_layer_input_gate (hd-sized) and per_layer_projection (pli-sized) per
  layer plus the global per_layer_model_projection [B,S,L,PLI] tensor when
  hidden_size_per_layer_input is set.
- _build_text_module_elements: split mlp.experts into routed and
  mlp.shared_expert canonical entries; register layers.<i>.experts alias for
  Gemma4 enable_moe_block layouts and mlp.shared_experts (plural) alias for
  Exaone-MoE / Laguna / GLM4-MoE-Lite shared-expert variants.
- _compute_moe_mlp_elements: split into _compute_routed_moe_elements and
  _compute_shared_moe_elements; only count shared_expert_gate (hd->1 Linear
  per shared expert) when shared_expert_intermediate_size is set, which is
  the Qwen2-MoE / Qwen3.5-MoE discriminator. Other shared-expert families
  (Exaone-MoE, HY-V3, GLM4-MoE-Lite, Laguna) lack the gate.
- compute_lora_params: when target_modules='all-linear' bare keyword, drop
  routed and shared MoE expert LoRA contributions. PEFT's all-linear targets
  nn.Linear only; Unsloth's get_moe_target_parameters expands MoE expert
  nn.Parameter LoRA only when target_modules contains explicit
  gate_proj/up_proj/down_proj/gate_up_proj names.
- _per_layer_input_lora_params: thread target_modules through and add the
  per-PLE-module contribution when the corresponding name appears, not only
  under all-linear.

* Studio: top-k MoE activations, ERNIE list configs, suffix skips, multimodal full bytes

Six estimator corrections aligning the detailed accounting paths with real
training behavior:

- _layer_qkv_mlp_sizes scales the MoE-layer mlp_size by num_experts_per_tok
  so the active routed-expert intermediate tensors are charged for activations.
  Adds num_experts_per_tok to ModelArchConfig and extracts it from
  num_experts_per_tok / top_k_experts (Gemma4 alias) in extract_arch_config.
- compute_lora_params splits routed and shared MoE LoRA contributions so that
  bare target_modules='all-linear' zeroes routed (nn.Parameter expert tensors,
  which Unsloth's get_moe_target_parameters does NOT enable for the bare
  keyword) but keeps shared-expert LoRA (regular nn.Linear MLPs that
  Unsloth's get_peft_regex DOES match).
- extract_arch_config gains a _first_scalar helper for ERNIE-style
  moe_intermediate_size = [routed, shared] lists, plus moe_num_experts and
  moe_num_shared_experts attribute aliases. When moe_intermediate_size is a
  pair and shared_expert_intermediate_size is unset, the second element is
  treated as the shared-expert intermediate.
- estimate_required_model_memory_gb's detailed branch retains
  max(0, model_size_bytes - compute_total_params(arch) * 2) on top of the
  arch-derived breakdown.model_weights so multimodal models (vision/audio
  towers) and partially-modeled families (Gemma3n AltUp/Laurel etc.) do not
  silently drop bytes that the safetensors total includes.
- _module_path_matches accepts a tail-only match when the skip entry is
  shorter than the alias path. Transformers' BNB quantizer suffix-matches
  short skip entries like ['q_proj'] / ['lm_head'] against full module
  paths; the previous len(skip) < len(alias) early-return missed those.
- _per_layer_input_lora_params drops the all_linear branch and only counts
  PLE LoRA when the user explicitly names per_layer_input_gate /
  per_layer_projection / per_layer_model_projection. Unsloth's
  get_peft_regex requires module names to contain a component tag
  (mlp/attn/...); PLE module names lack any tag, so all-linear training
  does not attach LoRA to them.

* Studio: full-FT extra optimizer/gradient inflation, MoE top-k aliases, ERNIE position dispatch, sibling experts aggregate

When the safetensors total exceeds the text-arch fp16 estimate (multimodal
vision/audio towers, partially-modeled families), only inflate the model
weights line for adapter methods but extend optimizer + gradient bytes
under full fine-tuning, where the extra params are trainable.

DBRX exposes top-k routing as moe_top_k and Hunyuan-V1-MoE as moe_topk;
neither is aliased to num_experts_per_tok via attribute_map, so probe both
when extracting arch config.

ERNIE 4.5 MoE / VL MoE configs declare MoE layers via
moe_layer_start_index / moe_layer_end_index / moe_layer_interval (with -1
meaning the last layer); add the position-style dispatch alongside the
existing mlp_layer_types / first_k_dense_replace / decoder_sparse_step
paths.

When moe_has_dense_mlp is set (Gemma4 enable_moe_block) the routed experts
live as a sibling of self.mlp at layers.<i>.experts in the actual model
layout; keep the layer mlp aggregate to the dense path and add a separate
experts aggregate so a skip module model.layers.<i>.mlp does not collapse
the routed experts as well.

* Studio: extend MoE family extraction (Llama4 / DBRX / Hunyuan / ERNIE) and align dense vs routed MLP widths

- Llama4: pick up `config.moe_layers` (auto-populated from
  interleave_moe_layer_step) so dense layer indices reflect the actual
  is_moe_layer dispatch.
- Llama4: add a separate `dense_intermediate_size` derived from
  `intermediate_size_mlp` (used for the dense feed_forward path) and keep
  `intermediate_size` for the routed/shared expert width. Auto-attach one
  shared expert per MoE layer when the dense-vs-MoE width split is present.
- DBRX: walk the `ffn_config` sub-config when extracting MoE attrs
  (moe_num_experts / moe_top_k / ffn_hidden_size). Without this DBRX is
  misclassified as a dense arch.
- Hunyuan: normalize layer-wise `moe_topk` (and the canonical
  `num_experts_per_tok` lookup it shadows via attribute_map) through a
  worst-case scalar so the int(...) cast cannot crash on list values.
- ERNIE 4.5 MoE: switch the start/end/interval dispatch to the model's
  `(layer_idx + 1) % interval == 0` modulo gate so MoE layers match the
  decoder when interval > 1.
- ERNIE 4.5 VL MoE: drop the heuristic that read
  `moe_intermediate_size[1]` as the shared expert width; in VL configs [1]
  is the vision-routed width and shared experts are sized from [0].
- estimate_fp16_model_size_bytes: prefer the larger of config-derived and
  local-weight bytes so the multimodal extra_bytes correction can fire
  for local VLM directories.

* Add tests for VRAM estimator extensions

* Studio: trim verbose comments in VRAM estimator

Collapse multi-paragraph rationale blocks to 1-3 lines stating the single
load-bearing fact. Fix one inverted "fall through ... last" comment whose
claim disagreed with the surrounding code.

* Consolidate added tests into existing test_vram_estimation.py and test_gpu_selection.py

Move Llama4 / DBRX / ERNIE arch-extraction tests into test_vram_estimation.py
as TestLlama4ArchExtraction / TestDbrxFfnConfigExtraction /
TestErniePhaseModuloDispatch / TestErnieVlSharedExpertWidth classes. Move
estimate_fp16_model_size_bytes prefer-larger-of-config-or-local tests into
test_gpu_selection.py as TestEstimateFp16ModelSizeBytesPrefersLocalWeights.
Drop one redundant Llama4 num_dense_layers assertion already covered by the
moe_layers dispatch test.

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

---------

Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com>
Co-authored-by: Lee Jackson <130007945+Imagineer99@users.noreply.github.com>
Co-authored-by: Daniel Han <danielhanchen@gmail.com>
Co-authored-by: Roland Tannous <115670425+rolandtannous@users.noreply.github.com>
---
 studio/backend/tests/test_gpu_selection.py    |  296 +++
 studio/backend/tests/test_vram_estimation.py  | 1674 +++++++++++++++++
 .../backend/utils/hardware/VRAM_ESTIMATION.md |   89 +-
 studio/backend/utils/hardware/hardware.py     |   75 +-
 .../backend/utils/hardware/vram_estimation.py |  955 +++++++++-
 5 files changed, 2981 insertions(+), 108 deletions(-)

diff --git a/studio/backend/tests/test_gpu_selection.py b/studio/backend/tests/test_gpu_selection.py
index fb4f38565b..a1fe5653ef 100644
--- a/studio/backend/tests/test_gpu_selection.py
+++ b/studio/backend/tests/test_gpu_selection.py
@@ -1049,6 +1049,182 @@ class TestMinGpuVram(unittest.TestCase):
 
 
 class TestPerGpuFitGuardAllCounts(unittest.TestCase):
+    def test_training_estimate_resolves_attention_without_raising(self):
+        with (
+            patch("utils.hardware.hardware.get_device", return_value = DeviceType.CUDA),
+            patch(
+                "utils.hardware.hardware.estimate_fp16_model_size_bytes",
+                return_value = (8 * (1024**3), "config"),
+            ),
+            patch(
+                "utils.hardware.hardware._resolve_model_identifier_for_gpu_estimate",
+                return_value = "unsloth/test",
+            ),
+            patch(
+                "utils.hardware.hardware._load_config_for_gpu_estimate",
+                return_value = SimpleNamespace(
+                    hidden_size = 4096,
+                    num_hidden_layers = 32,
+                    num_attention_heads = 32,
+                    num_key_value_heads = 8,
+                    intermediate_size = 14336,
+                    vocab_size = 128256,
+                    tie_word_embeddings = False,
+                ),
+            ),
+            patch(
+                "utils.hardware.hardware._determine_attention_impl_for_gpu_estimate",
+                return_value = "eager",
+            ),
+            patch("utils.hardware.hardware.get_visible_gpu_count", return_value = 1),
+        ):
+            _, metadata = estimate_required_model_memory_gb(
+                "unsloth/test",
+                training_type = "LoRA/QLoRA",
+                load_in_4bit = True,
+            )
+
+        self.assertEqual(metadata.get("estimation_mode"), "detailed")
+        self.assertEqual(metadata.get("attention_implementation"), "eager")
+
+    def test_training_estimate_falls_back_when_attention_resolution_fails(self):
+        with (
+            patch("utils.hardware.hardware.get_device", return_value = DeviceType.CUDA),
+            patch(
+                "utils.hardware.hardware.estimate_fp16_model_size_bytes",
+                return_value = (8 * (1024**3), "config"),
+            ),
+            patch(
+                "utils.hardware.hardware._resolve_model_identifier_for_gpu_estimate",
+                return_value = "unsloth/test",
+            ),
+            patch(
+                "utils.hardware.hardware._load_config_for_gpu_estimate",
+                return_value = SimpleNamespace(
+                    hidden_size = 4096,
+                    num_hidden_layers = 32,
+                    num_attention_heads = 32,
+                    num_key_value_heads = 8,
+                    intermediate_size = 14336,
+                    vocab_size = 128256,
+                    tie_word_embeddings = False,
+                ),
+            ),
+            patch(
+                "utils.hardware.hardware._determine_attention_impl_for_gpu_estimate",
+                side_effect = RuntimeError("attention unavailable"),
+            ),
+            patch("utils.hardware.hardware.get_visible_gpu_count", return_value = 1),
+        ):
+            _, metadata = estimate_required_model_memory_gb(
+                "unsloth/test",
+                training_type = "LoRA/QLoRA",
+                load_in_4bit = True,
+            )
+
+        self.assertEqual(metadata.get("estimation_mode"), "detailed")
+        self.assertEqual(
+            metadata.get("attention_implementation"),
+            "eager",
+        )
+
+    def test_attention_resolver_does_not_mutate_loaded_config(self):
+        from utils.hardware import hardware as hardware_module
+
+        config = SimpleNamespace(
+            hidden_size = 1024,
+            num_hidden_layers = 2,
+            num_attention_heads = 8,
+            num_key_value_heads = 8,
+            intermediate_size = 2048,
+            vocab_size = 1024,
+            tie_word_embeddings = True,
+        )
+
+        def _stub_resolver(model_class, cfg):
+            cfg._attn_implementation = "eager"
+            return "eager"
+
+        with patch(
+            "unsloth.models._utils.resolve_attention_implementation",
+            side_effect = _stub_resolver,
+        ):
+            hardware_module._determine_attention_impl_for_gpu_estimate(config)
+
+        self.assertFalse(hasattr(config, "_attn_implementation"))
+
+    def test_attention_resolver_handles_missing_model_mapping(self):
+        from utils.hardware import hardware as hardware_module
+
+        config = SimpleNamespace(
+            hidden_size = 1024,
+            num_hidden_layers = 2,
+            num_attention_heads = 8,
+            num_key_value_heads = 8,
+            intermediate_size = 2048,
+            vocab_size = 1024,
+            tie_word_embeddings = True,
+        )
+        captured = {}
+
+        def _stub_resolver(model_class, cfg):
+            captured["model_class"] = model_class
+            return "eager"
+
+        from transformers import AutoModel, AutoModelForCausalLM
+
+        with (
+            patch.object(AutoModelForCausalLM, "_model_mapping", new = None),
+            patch.object(AutoModel, "_model_mapping", new = None),
+            patch(
+                "unsloth.models._utils.resolve_attention_implementation",
+                side_effect = _stub_resolver,
+            ),
+        ):
+            result = hardware_module._determine_attention_impl_for_gpu_estimate(config)
+
+        self.assertEqual(result, "eager")
+        self.assertIsNone(captured["model_class"])
+
+    def test_attention_resolver_does_not_mutate_nested_text_config(self):
+        from utils.hardware import hardware as hardware_module
+
+        text_config = SimpleNamespace(
+            hidden_size = 1024,
+            num_hidden_layers = 2,
+            num_attention_heads = 8,
+            num_key_value_heads = 8,
+            intermediate_size = 2048,
+            vocab_size = 1024,
+            tie_word_embeddings = True,
+        )
+        config = SimpleNamespace(
+            hidden_size = 1024,
+            num_hidden_layers = 2,
+            num_attention_heads = 8,
+            num_key_value_heads = 8,
+            intermediate_size = 2048,
+            vocab_size = 1024,
+            tie_word_embeddings = True,
+            text_config = text_config,
+        )
+
+        def _stub_resolver(model_class, cfg):
+            cfg._attn_implementation = "eager"
+            inner = getattr(cfg, "text_config", None)
+            if inner is not None:
+                inner._attn_implementation = "eager"
+            return "eager"
+
+        with patch(
+            "unsloth.models._utils.resolve_attention_implementation",
+            side_effect = _stub_resolver,
+        ):
+            hardware_module._determine_attention_impl_for_gpu_estimate(config)
+
+        self.assertFalse(hasattr(config, "_attn_implementation"))
+        self.assertFalse(hasattr(text_config, "_attn_implementation"))
+
     def test_min_per_gpu_generated_for_all_visible_counts(self):
         with (
             patch("utils.hardware.hardware.get_device", return_value = DeviceType.CUDA),
@@ -1125,3 +1301,123 @@ class TestXpuRejection(_GpuCacheResetMixin, unittest.TestCase):
         with patch("utils.hardware.hardware.get_device", return_value = DeviceType.XPU):
             with self.assertRaisesRegex(ValueError, "only supported on CUDA"):
                 prepare_gpu_selection([0], model_name = "unsloth/test")
+
+
+class TestEstimateFp16ModelSizeBytesPrefersLocalWeights(unittest.TestCase):
+    def _run(
+        self,
+        model_path,
+        *,
+        config_bytes,
+        local_bytes,
+        safetensors_params = None,
+        config = object(),
+    ):
+        from utils.hardware import hardware as hardware_module
+
+        with (
+            patch.object(
+                hardware_module,
+                "_resolve_model_identifier_for_gpu_estimate",
+                return_value = model_path,
+            ),
+            patch.object(
+                hardware_module,
+                "_get_hf_safetensors_total_params",
+                return_value = safetensors_params,
+            ),
+            patch.object(
+                hardware_module,
+                "_load_config_for_gpu_estimate",
+                return_value = config,
+            ),
+            patch.object(
+                hardware_module,
+                "_estimate_fp16_model_size_bytes_from_config",
+                return_value = config_bytes,
+            ),
+            patch.object(
+                hardware_module,
+                "_get_local_weight_size_bytes",
+                return_value = local_bytes,
+            ),
+        ):
+            return hardware_module.estimate_fp16_model_size_bytes(model_path)
+
+    def test_local_weight_bytes_preferred_when_larger_than_config(self):
+        bytes_, src = self._run(
+            "/local/vlm",
+            config_bytes = 2 * (1 << 30),
+            local_bytes = 20 * (1 << 30),
+        )
+        self.assertEqual(bytes_, 20 * (1 << 30))
+        self.assertEqual(src, "weight_bytes")
+
+    def test_config_bytes_preferred_when_larger_than_local(self):
+        bytes_, src = self._run(
+            "/local/text-only",
+            config_bytes = 20 * (1 << 30),
+            local_bytes = 2 * (1 << 30),
+        )
+        self.assertEqual(bytes_, 20 * (1 << 30))
+        self.assertEqual(src, "config")
+
+    def test_config_bytes_returned_when_no_local_weights(self):
+        bytes_, src = self._run(
+            "/local/no-weights",
+            config_bytes = 5 * (1 << 30),
+            local_bytes = None,
+        )
+        self.assertEqual(bytes_, 5 * (1 << 30))
+        self.assertEqual(src, "config")
+
+    def test_local_bytes_returned_when_config_resolution_fails(self):
+        bytes_, src = self._run(
+            "/local/no-config",
+            config_bytes = None,
+            local_bytes = 7 * (1 << 30),
+            config = None,
+        )
+        self.assertEqual(bytes_, 7 * (1 << 30))
+        self.assertEqual(src, "weight_bytes")
+
+    def test_equal_local_and_config_keeps_config_label(self):
+        # why: tie-breaker is "local must be strictly larger" so an exact
+        # match keeps the config-derived path.
+        same = 8 * (1 << 30)
+        bytes_, src = self._run(
+            "/local/equal",
+            config_bytes = same,
+            local_bytes = same,
+        )
+        self.assertEqual(bytes_, same)
+        self.assertEqual(src, "config")
+
+    def test_remote_safetensors_path_unaffected_by_local_weights(self):
+        from utils.hardware import hardware as hardware_module
+
+        with (
+            patch.object(
+                hardware_module,
+                "_resolve_model_identifier_for_gpu_estimate",
+                return_value = "owner/repo",
+            ),
+            patch.object(
+                hardware_module,
+                "_get_hf_safetensors_total_params",
+                return_value = 1_000_000_000,
+            ),
+            patch.object(
+                hardware_module,
+                "_load_config_for_gpu_estimate",
+            ) as mock_load,
+            patch.object(
+                hardware_module,
+                "_get_local_weight_size_bytes",
+            ) as mock_local,
+        ):
+            bytes_, src = hardware_module.estimate_fp16_model_size_bytes("owner/repo")
+            self.assertEqual(bytes_, 2 * 1_000_000_000)
+            self.assertEqual(src, "safetensors")
+            mock_load.assert_not_called()
+            mock_local.assert_not_called()
diff --git a/studio/backend/tests/test_vram_estimation.py b/studio/backend/tests/test_vram_estimation.py
index 0be067310d..e54ae6dcf8 100644
--- a/studio/backend/tests/test_vram_estimation.py
+++ b/studio/backend/tests/test_vram_estimation.py
@@ -2,7 +2,9 @@
 # Copyright 2026-present the Unsloth AI Inc. team. All rights reserved.
 
 import unittest
+from dataclasses import replace
 from types import SimpleNamespace
+from unittest.mock import patch
 
 from utils.hardware.vram_estimation import (
     ModelArchConfig,
@@ -116,6 +118,55 @@ GPT_OSS = ModelArchConfig(
     num_dense_layers = 0,
 )
 
+STRUCTURED_MIXED = ModelArchConfig(
+    hidden_size = 256,
+    num_hidden_layers = 6,
+    num_attention_heads = 4,
+    num_key_value_heads = 2,
+    intermediate_size = 512,
+    vocab_size = 1024,
+    tie_word_embeddings = True,
+    head_dim = 80,
+    global_head_dim = 96,
+    num_global_key_value_heads = 1,
+    attention_k_eq_v = True,
+    layer_types = [
+        "sliding_attention",
+        "full_attention",
+        "sliding_attention",
+        "full_attention",
+        "sliding_attention",
+        "full_attention",
+    ],
+)
+
+STRUCTURED_SHARED = ModelArchConfig(
+    hidden_size = 192,
+    num_hidden_layers = 4,
+    num_attention_heads = 6,
+    num_key_value_heads = 2,
+    intermediate_size = 384,
+    vocab_size = 512,
+    tie_word_embeddings = True,
+    head_dim = 32,
+    num_kv_shared_layers = 2,
+    use_double_wide_mlp = True,
+    vocab_size_per_layer_input = 128,
+    hidden_size_per_layer_input = 48,
+    quant_4bit_factor = 3.6,
+)
+
+QUANT_SKIP_STRUCTURED = replace(
+    STRUCTURED_SHARED,
+    quantization_skip_modules = [
+        "model.layers.0.self_attn.q_proj",
+        "language_model.model.layers.1.mlp",
+        "layers.2",
+        "vision_tower",
+        "embed_tokens",
+    ],
+)
+
 
 class TestExtractArchConfig(unittest.TestCase):
     def test_basic_config(self):
@@ -182,6 +233,42 @@ class TestExtractArchConfig(unittest.TestCase):
         arch = extract_arch_config(hf_config)
         self.assertEqual(arch.intermediate_size, 8192)
 
+    def test_structural_and_quantization_fields_are_config_derived(self):
+        hf_config = SimpleNamespace(
+            hidden_size = 256,
+            num_hidden_layers = 2,
+            num_attention_heads = 4,
+            num_key_value_heads = 2,
+            intermediate_size = 512,
+            vocab_size = 1024,
+            tie_word_embeddings = True,
+            head_dim = 80,
+            global_head_dim = 96,
+            num_global_key_value_heads = 1,
+            attention_k_eq_v = True,
+            layer_types = ["sliding_attention", "full_attention"],
+            num_kv_shared_layers = 1,
+            use_double_wide_mlp = True,
+            vocab_size_per_layer_input = 128,
+            hidden_size_per_layer_input = 48,
+            quantization_config = {
+                "bnb_4bit_use_double_quant": True,
+                "llm_int8_skip_modules": ["model.layers.0.self_attn"],
+            },
+        )
+        arch = extract_arch_config(hf_config)
+        self.assertEqual(arch.head_dim, 80)
+        self.assertEqual(arch.global_head_dim, 96)
+        self.assertEqual(arch.num_global_key_value_heads, 1)
+        self.assertTrue(arch.attention_k_eq_v)
+        self.assertEqual(arch.layer_types, ["sliding_attention", "full_attention"])
+        self.assertEqual(arch.num_kv_shared_layers, 1)
+        self.assertTrue(arch.use_double_wide_mlp)
+        self.assertEqual(arch.vocab_size_per_layer_input, 128)
+        self.assertEqual(arch.hidden_size_per_layer_input, 48)
+        self.assertEqual(arch.quantization_skip_modules, ["model.layers.0.self_attn"])
+        self.assertEqual(arch.quant_4bit_factor, 3.6)
+
 
 class TestModelWeightsBytes(unittest.TestCase):
     def test_llama_8b_fp16(self):
@@ -238,6 +325,18 @@ class TestLoraParams(unittest.TestCase):
         ratio = moe_lora / dense_lora
         self.assertAlmostEqual(ratio, 8.0, delta = 0.5)
 
+    def test_structured_moe_mlp_modules_scale_with_experts(self):
+        structured_moe = replace(QWEN3_MOE_30B, head_dim = 128)
+        dense_like = replace(
+            structured_moe,
+            num_experts = None,
+            moe_intermediate_size = None,
+        )
+        target_modules = ["gate_proj", "up_proj", "down_proj"]
+        dense_lora = compute_lora_params(dense_like, 16, target_modules)
+        moe_lora = compute_lora_params(structured_moe, 16, target_modules)
+        self.assertGreater(moe_lora, dense_lora * 20)
+
     def test_attention_modules_same_for_moe(self):
         dense_attn = compute_lora_params(
             LLAMA_8B, 16, ["q_proj", "k_proj", "v_proj", "o_proj"]
@@ -247,6 +346,41 @@ class TestLoraParams(unittest.TestCase):
         )
         self.assertEqual(dense_attn, moe_attn)
 
+    def test_all_linear_uses_default_text_modules(self):
+        text_only = compute_lora_params(STRUCTURED_MIXED, 16, DEFAULT_TARGET_MODULES)
+        all_linear = compute_lora_params(STRUCTURED_MIXED, 16, ["all-linear"])
+        self.assertEqual(all_linear, text_only)
+
+    def test_structural_layer_shapes_are_config_driven(self):
+        unstructured_arch = replace(
+            STRUCTURED_MIXED,
+            head_dim = None,
+            global_head_dim = None,
+            num_global_key_value_heads = None,
+            attention_k_eq_v = False,
+            layer_types = None,
+        )
+        self.assertNotEqual(
+            compute_lora_params(unstructured_arch, 16, ["all-linear"]),
+            compute_lora_params(STRUCTURED_MIXED, 16, ["all-linear"]),
+        )
+        self.assertNotEqual(
+            compute_model_weights_bytes(unstructured_arch, "qlora", True),
+            compute_model_weights_bytes(STRUCTURED_MIXED, "qlora", True),
+        )
+
+    def test_shared_kv_and_per_layer_inputs_change_weight_count(self):
+        unstructured_arch = replace(
+            STRUCTURED_SHARED,
+            head_dim = None,
+            num_kv_shared_layers = 0,
+            use_double_wide_mlp = False,
+        )
+        self.assertNotEqual(
+            compute_model_weights_bytes(unstructured_arch, "qlora", True),
+            compute_model_weights_bytes(STRUCTURED_SHARED, "qlora", True),
+        )
+
 
 class TestOptimizerBytes(unittest.TestCase):
     def test_adamw_8bit(self):
@@ -293,6 +427,163 @@ class TestActivationBytes(unittest.TestCase):
         act_4k = compute_activation_bytes(LLAMA_8B, 2, 4096, "unsloth")
         self.assertAlmostEqual(act_4k / act_2k, 2.0, delta = 0.1)
 
+    def test_flash_attention_uses_linear_path(self):
+        flash = compute_activation_bytes(
+            STRUCTURED_MIXED,
+            1,
+            4096,
+            "unsloth",
+            is_lora = True,
+            attention_implementation = "flash_attention_2",
+        )
+        default = compute_activation_bytes(
+            STRUCTURED_MIXED,
+            1,
+            4096,
+            "unsloth",
+            is_lora = True,
+        )
+        self.assertEqual(flash, default)
+
+    def test_sdpa_attention_uses_linear_path(self):
+        flash = compute_activation_bytes(
+            STRUCTURED_MIXED,
+            1,
+            4096,
+            "unsloth",
+            is_lora = True,
+            attention_implementation = "flash_attention_2",
+        )
+        sdpa = compute_activation_bytes(
+            STRUCTURED_MIXED,
+            1,
+            4096,
+            "unsloth",
+            is_lora = True,
+            attention_implementation = "sdpa",
+        )
+        self.assertEqual(sdpa, flash)
+
+    def test_non_flash_attention_uses_quadratic_path(self):
+        seq_len = 4096
+        expected_quadratic = (
+            1 * STRUCTURED_MIXED.num_attention_heads * seq_len * seq_len * 2 * 12.0
+        )
+        for attention_implementation in ("eager", "unknown_impl", None):
+            with self.subTest(attention_implementation = attention_implementation):
+                non_flash = compute_activation_bytes(
+                    STRUCTURED_MIXED,
+                    1,
+                    seq_len,
+                    "unsloth",
+                    is_lora = True,
+                    attention_implementation = attention_implementation,
+                )
+                self.assertEqual(non_flash, int(expected_quadratic))
+
+    def test_non_flash_attention_without_gc_scales_quadratic_path_by_layers(self):
+        seq_len = 4096
+        one_layer = (
+            1 * STRUCTURED_MIXED.num_attention_heads * seq_len * seq_len * 2 * 12.0
+        )
+        non_flash = compute_activation_bytes(
+            STRUCTURED_MIXED,
+            1,
+            seq_len,
+            "none",
+            is_lora = True,
+            attention_implementation = "eager",
+        )
+        self.assertEqual(non_flash, int(one_layer * STRUCTURED_MIXED.num_hidden_layers))
+        self.assertGreater(non_flash, int(one_layer))
+
+
+class TestQuantizationSkips(unittest.TestCase):
+    def test_skipped_language_layers_stay_fp16(self):
+        no_skips = replace(QUANT_SKIP_STRUCTURED, quantization_skip_modules = [])
+        skipped = compute_model_weights_bytes(QUANT_SKIP_STRUCTURED, "qlora", True)
+        quantized = compute_model_weights_bytes(no_skips, "qlora", True)
+        self.assertGreater(skipped, quantized)
+
+    def test_non_language_skips_do_not_double_count_text_weights(self):
+        arch = replace(
+            QUANT_SKIP_STRUCTURED,
+            quantization_skip_modules = ["vision_tower", "embed_tokens"],
+        )
+        no_skips = replace(QUANT_SKIP_STRUCTURED, quantization_skip_modules = [])
+        self.assertEqual(
+            compute_model_weights_bytes(arch, "qlora", True),
+            compute_model_weights_bytes(no_skips, "qlora", True),
+        )
+
+    def test_double_quant_factor_reduces_quantized_weight_storage(self):
+        default_quant = replace(STRUCTURED_MIXED, quant_4bit_factor = 16 / 5)
+        double_quant = replace(STRUCTURED_MIXED, quant_4bit_factor = 3.6)
+        self.assertLess(
+            compute_model_weights_bytes(double_quant, "qlora", True),
+            compute_model_weights_bytes(default_quant, "qlora", True),
+        )
+
+    def test_prefixed_parent_and_child_skips_do_not_double_count(self):
+        parent_only = replace(
+            QUANT_SKIP_STRUCTURED,
+            quantization_skip_modules = ["language_model.model.layers.1.mlp"],
+        )
+        parent_and_child = replace(
+            QUANT_SKIP_STRUCTURED,
+            quantization_skip_modules = [
+                "language_model.model.layers.1.mlp",
+                "language_model.model.layers.1.mlp.gate_proj",
+                "model.layers.1.mlp.up_proj",
+            ],
+        )
+        self.assertEqual(
+            compute_model_weights_bytes(parent_and_child, "qlora", True),
+            compute_model_weights_bytes(parent_only, "qlora", True),
+        )
+
+    def test_vlm_prefix_skip_module_does_not_match_text_alias(self):
+        # vision_tower-prefixed skips must not shadow text aliases sharing the
+        # same suffix.
+        baseline = replace(QUANT_SKIP_STRUCTURED, quantization_skip_modules = [])
+        vlm_skip = replace(
+            QUANT_SKIP_STRUCTURED,
+            quantization_skip_modules = [
+                "vision_tower.model.layers.0.self_attn.q_proj",
+                "vision_tower.model.layers.1.mlp",
+            ],
+        )
+        self.assertEqual(
+            compute_model_weights_bytes(vlm_skip, "qlora", True),
+            compute_model_weights_bytes(baseline, "qlora", True),
+        )
+
+    def test_mla_skip_module_uses_authoritative_attn_total(self):
+        from utils.hardware.vram_estimation import (
+            _build_text_module_elements,
+            _compute_attn_elements,
+        )
+
+        mla = ModelArchConfig(
+            hidden_size = 2048,
+            num_hidden_layers = 4,
+            num_attention_heads = 16,
+            num_key_value_heads = 16,
+            intermediate_size = 8192,
+            vocab_size = 32000,
+            tie_word_embeddings = False,
+            q_lora_rank = 512,
+            kv_lora_rank = 128,
+            qk_nope_head_dim = 64,
+            qk_rope_head_dim = 32,
+            v_head_dim = 64,
+        )
+        elements, _ = _build_text_module_elements(mla)
+        self.assertEqual(
+            elements["text.layers.0.self_attn"],
+            _compute_attn_elements(mla),
+        )
+
 
 class TestEstimateTrainingVram(unittest.TestCase):
     def test_llama_8b_qlora_reasonable_total(self):
@@ -430,6 +721,90 @@ class TestEstimateTrainingVram(unittest.TestCase):
             v32.optimizer_states / v8.optimizer_states, 1.5, delta = 0.1
         )
 
+    def test_min_gpu_vram_treats_activations_as_per_gpu_fixed(self):
+        config = TrainingVramConfig(training_method = "qlora", load_in_4bit = True)
+        breakdown = estimate_training_vram(LLAMA_8B, config)
+        shardable = (
+            breakdown.model_weights
+            + breakdown.lora_adapters
+            + breakdown.optimizer_states
+            + breakdown.gradients
+        )
+        per_gpu_fixed = breakdown.activations + breakdown.cuda_overhead
+        for n_gpus in (1, 2, 4):
+            self.assertEqual(
+                breakdown.min_gpu_vram(n_gpus),
+                shardable // n_gpus + per_gpu_fixed,
+            )
+
+    def test_qlora_gradient_floor_is_capped_by_trainable_scale(self):
+        config = TrainingVramConfig(
+            training_method = "qlora",
+            batch_size = 1,
+            max_seq_length = 512,
+            lora_rank = 16,
+            target_modules = ["all-linear"],
+            gradient_checkpointing = "unsloth",
+            optimizer = "adamw_8bit",
+            load_in_4bit = True,
+        )
+        breakdown = estimate_training_vram(LLAMA_8B, config)
+        lora_params = compute_lora_params(LLAMA_8B, 16, DEFAULT_TARGET_MODULES)
+        optimizer_bytes = compute_optimizer_bytes(lora_params, "adamw_8bit")
+        weight_floor = int(breakdown.model_weights * 0.15)
+
+        self.assertEqual(
+            breakdown.gradients,
+            max(breakdown.activations_computed, optimizer_bytes),
+        )
+        self.assertLess(breakdown.gradients, weight_floor)
+        self.assertEqual(breakdown.activations, breakdown.activations_computed)
+
+    def test_full_finetuning_gradient_floor_remains_uncapped(self):
+        config = TrainingVramConfig(
+            training_method = "full",
+            batch_size = 1,
+            max_seq_length = 512,
+            gradient_checkpointing = "unsloth",
+            optimizer = "adamw_8bit",
+            load_in_4bit = False,
+        )
+        expected_floor = int(
+            compute_model_weights_bytes(LLAMA_8B, "full", False) * 0.15
+        )
+        with patch(
+            "utils.hardware.vram_estimation.compute_gradient_bytes",
+            return_value = 1,
+        ):
+            breakdown = estimate_training_vram(LLAMA_8B, config)
+        self.assertEqual(breakdown.gradients, expected_floor)
+
+    def test_non_flash_attention_flows_into_training_estimate(self):
+        config = TrainingVramConfig(
+            training_method = "qlora",
+            batch_size = 1,
+            max_seq_length = 4096,
+            lora_rank = 16,
+            target_modules = ["all-linear"],
+            gradient_checkpointing = "unsloth",
+            optimizer = "adamw_8bit",
+            load_in_4bit = True,
+            attention_implementation = "eager",
+        )
+        breakdown = estimate_training_vram(STRUCTURED_MIXED, config)
+        self.assertEqual(breakdown.activations, breakdown.activations_computed)
+        self.assertGreater(
+            breakdown.activations,
+            compute_activation_bytes(
+                STRUCTURED_MIXED,
+                1,
+                4096,
+                "unsloth",
+                is_lora = True,
+                attention_implementation = "flash_attention_2",
+            ),
+        )
+
 
 class TestExtractArchConfigMoE(unittest.TestCase):
     def test_deepseek_v3_shared_experts(self):
@@ -471,11 +846,16 @@ class TestExtractArchConfigMoE(unittest.TestCase):
             moe_intermediate_size = 768,
             decoder_sparse_step = 1,
             mlp_only_layers = [],
+            head_dim = 128,
         )
         arch = extract_arch_config(hf_config)
         self.assertEqual(arch.num_experts, 128)
         self.assertEqual(arch.num_dense_layers, 0)
+        self.assertEqual(arch.head_dim, 128)
         self.assertIsNone(arch.q_lora_rank)
+        total_b = compute_total_params(arch) / 1e9
+        self.assertGreater(total_b, 20)
+        self.assertLess(total_b, 50)
 
     def test_qwen3_moe_with_mlp_only_layers(self):
         hf_config = SimpleNamespace(
@@ -542,6 +922,343 @@ class TestExtractArchConfigMoE(unittest.TestCase):
         self.assertEqual(arch.n_shared_experts, 0)
         self.assertEqual(arch.num_dense_layers, 0)
         self.assertIsNone(arch.q_lora_rank)
+        self.assertFalse(arch.moe_has_dense_mlp)
+
+    def test_enable_moe_block_extracted_as_moe_has_dense_mlp(self):
+        hf_config = SimpleNamespace(
+            hidden_size = 2048,
+            num_hidden_layers = 8,
+            num_attention_heads = 16,
+            num_key_value_heads = 4,
+            intermediate_size = 4096,
+            vocab_size = 32000,
+            tie_word_embeddings = True,
+            num_experts = 8,
+            moe_intermediate_size = 1024,
+            head_dim = 128,
+            layer_types = ["full_attention"] * 8,
+            enable_moe_block = True,
+        )
+        arch = extract_arch_config(hf_config)
+        self.assertTrue(arch.moe_has_dense_mlp)
+
+
+class TestParallelDenseMoE(unittest.TestCase):
+    def _arch(self, **overrides):
+        base = ModelArchConfig(
+            hidden_size = 512,
+            num_hidden_layers = 4,
+            num_attention_heads = 8,
+            num_key_value_heads = 2,
+            intermediate_size = 1024,
+            vocab_size = 1024,
+            tie_word_embeddings = True,
+            num_experts = 8,
+            moe_intermediate_size = 512,
+            num_dense_layers = 0,
+            head_dim = 64,
+            layer_types = ["full_attention"] * 4,
+        )
+        return replace(base, **overrides)
+
+    def test_total_params_includes_parallel_dense_when_enable_moe_block(self):
+        without_parallel = self._arch(moe_has_dense_mlp = False)
+        with_parallel = self._arch(moe_has_dense_mlp = True)
+        self.assertGreater(
+            compute_total_params(with_parallel),
+            compute_total_params(without_parallel),
+        )
+
+    def test_lora_params_includes_parallel_dense_when_enable_moe_block(self):
+        without_parallel = self._arch(moe_has_dense_mlp = False)
+        with_parallel = self._arch(moe_has_dense_mlp = True)
+        target = ["gate_proj", "up_proj", "down_proj"]
+        self.assertGreater(
+            compute_lora_params(with_parallel, 16, target),
+            compute_lora_params(without_parallel, 16, target),
+        )
+
+    def test_activation_bytes_includes_parallel_dense_when_enable_moe_block(self):
+        without_parallel = self._arch(moe_has_dense_mlp = False)
+        with_parallel = self._arch(moe_has_dense_mlp = True)
+        self.assertGreater(
+            compute_activation_bytes(
+                with_parallel,
+                1,
+                2048,
+                "unsloth",
+                is_lora = True,
+            ),
+            compute_activation_bytes(
+                without_parallel,
+                1,
+                2048,
+                "unsloth",
+                is_lora = True,
+            ),
+        )
+
+    def test_layer_aggregates_split_dense_mlp_from_experts(self):
+        from utils.hardware.vram_estimation import _build_text_module_elements
+
+        with_parallel = self._arch(moe_has_dense_mlp = True)
+        elements, _ = _build_text_module_elements(with_parallel)
+        moe_only = (
+            with_parallel.hidden_size
+            * with_parallel.moe_intermediate_size
+            * 3
+            * with_parallel.num_experts
+            + with_parallel.num_experts * with_parallel.hidden_size
+        )
+        dense_only = with_parallel.hidden_size * with_parallel.intermediate_size * 3
+        # why: under gemma4 enable_moe_block, the layer's `self.experts` is a
+        # sibling of `self.mlp`; the `text.layers.<i>.mlp` aggregate must
+        # cover the dense path only, with experts in their own aggregate.
+        self.assertEqual(elements["text.layers.0.mlp"], dense_only)
+        self.assertEqual(elements["text.layers.0.experts"], moe_only)
+
+
+class TestDenseLayerIndices(unittest.TestCase):
+    def test_non_prefix_mlp_only_layers_preserve_position(self):
+        hf_config = SimpleNamespace(
+            hidden_size = 1024,
+            num_hidden_layers = 8,
+            num_attention_heads = 16,
+            num_key_value_heads = 4,
+            intermediate_size = 2048,
+            vocab_size = 32000,
+            tie_word_embeddings = True,
+            num_local_experts = 4,
+            moe_intermediate_size = 512,
+            decoder_sparse_step = 1,
+            mlp_only_layers = [3, 5],
+        )
+        arch = extract_arch_config(hf_config)
+        self.assertEqual(arch.num_dense_layers, 2)
+        self.assertIn(3, arch.dense_layer_indices)
+        self.assertIn(5, arch.dense_layer_indices)
+        self.assertNotIn(0, arch.dense_layer_indices)
+
+    def test_first_k_dense_replace_indices_are_prefix(self):
+        hf_config = SimpleNamespace(
+            hidden_size = 1024,
+            num_hidden_layers = 6,
+            num_attention_heads = 16,
+            num_key_value_heads = 4,
+            intermediate_size = 2048,
+            vocab_size = 32000,
+            tie_word_embeddings = False,
+            n_routed_experts = 8,
+            moe_intermediate_size = 512,
+            first_k_dense_replace = 2,
+        )
+        arch = extract_arch_config(hf_config)
+        self.assertEqual(tuple(arch.dense_layer_indices), (0, 1))
+
+
+class TestKvSharedLayer(unittest.TestCase):
+    def test_fully_shared_kv_returns_false_matching_upstream(self):
+        from utils.hardware.vram_estimation import _is_kv_shared_layer
+
+        arch = ModelArchConfig(
+            hidden_size = 512,
+            num_hidden_layers = 4,
+            num_attention_heads = 8,
+            num_key_value_heads = 2,
+            intermediate_size = 1024,
+            vocab_size = 1024,
+            num_kv_shared_layers = 4,
+        )
+        for i in range(arch.num_hidden_layers):
+            self.assertFalse(_is_kv_shared_layer(arch, i))
+
+    def test_partial_share_returns_true_for_tail_layers(self):
+        from utils.hardware.vram_estimation import _is_kv_shared_layer
+
+        arch = ModelArchConfig(
+            hidden_size = 512,
+            num_hidden_layers = 4,
+            num_attention_heads = 8,
+            num_key_value_heads = 2,
+            intermediate_size = 1024,
+            vocab_size = 1024,
+            num_kv_shared_layers = 2,
+        )
+        self.assertFalse(_is_kv_shared_layer(arch, 0))
+        self.assertFalse(_is_kv_shared_layer(arch, 1))
+        self.assertTrue(_is_kv_shared_layer(arch, 2))
+        self.assertTrue(_is_kv_shared_layer(arch, 3))
+
+
+class TestFlexAttentionLinear(unittest.TestCase):
+    def test_flex_attention_treated_as_linear(self):
+        flash = compute_activation_bytes(
+            STRUCTURED_MIXED,
+            1,
+            4096,
+            "unsloth",
+            is_lora = True,
+            attention_implementation = "flash_attention_2",
+        )
+        flex = compute_activation_bytes(
+            STRUCTURED_MIXED,
+            1,
+            4096,
+            "unsloth",
+            is_lora = True,
+            attention_implementation = "flex_attention",
+        )
+        self.assertEqual(flex, flash)
+
+
+class TestNonStructuredParallelDense(unittest.TestCase):
+    def _arch(self, **overrides):
+        base = ModelArchConfig(
+            hidden_size = 1024,
+            num_hidden_layers = 4,
+            num_attention_heads = 16,
+            num_key_value_heads = 4,
+            intermediate_size = 4096,
+            vocab_size = 32000,
+            tie_word_embeddings = False,
+            num_experts = 8,
+            moe_intermediate_size = 768,
+            num_dense_layers = 0,
+            moe_has_dense_mlp = True,
+        )
+        return replace(base, **overrides)
+
+    def test_skip_module_uses_intermediate_size_for_parallel_dense(self):
+        from utils.hardware.vram_estimation import _build_text_module_elements
+
+        arch = self._arch()
+        elements, _ = _build_text_module_elements(arch)
+        gate_proj = elements["text.layers.0.mlp.gate_proj"]
+        self.assertEqual(gate_proj, arch.hidden_size * arch.intermediate_size)
+
+
+class TestPerLayerInputAccounting(unittest.TestCase):
+    def _arch(self, **overrides):
+        base = ModelArchConfig(
+            hidden_size = 1024,
+            num_hidden_layers = 4,
+            num_attention_heads = 16,
+            num_key_value_heads = 4,
+            intermediate_size = 2048,
+            vocab_size = 32000,
+            tie_word_embeddings = False,
+            head_dim = 64,
+            layer_types = ["full_attention"] * 4,
+            vocab_size_per_layer_input = 256,
+            hidden_size_per_layer_input = 96,
+        )
+        return replace(base, **overrides)
+
+    def test_per_layer_input_increases_total_params(self):
+        with_ple = self._arch()
+        without_ple = replace(with_ple, hidden_size_per_layer_input = 0)
+        self.assertGreater(
+            compute_total_params(with_ple),
+            compute_total_params(without_ple),
+        )
+
+    def test_per_layer_input_modules_count_quantizable_block(self):
+        with_ple = self._arch()
+        without_ple = replace(with_ple, hidden_size_per_layer_input = 0)
+        # The PLE block adds: model_projection (hd*nl*pli), per_layer_input_gate
+        # (hd*pli per layer) + per_layer_projection (pli*hd per layer) as
+        # quantizable text linears.
+        n_layers = with_ple.num_hidden_layers
+        hd = with_ple.hidden_size
+        pli = with_ple.hidden_size_per_layer_input
+        expected_quantizable_extra = (
+            hd * (n_layers * pli) + (hd * pli) * n_layers + (pli * hd) * n_layers
+        )
+        delta = compute_total_params(with_ple) - compute_total_params(without_ple)
+        self.assertGreaterEqual(delta, expected_quantizable_extra)
+
+    def test_all_linear_lora_excludes_per_layer_input_modules(self):
+        # why: Unsloth's get_peft_regex requires module names to contain a
+        # component tag (mlp/attn/...); PLE module names (per_layer_input_gate,
+        # per_layer_projection, per_layer_model_projection) lack any tag, so
+        # all-linear training does NOT attach LoRA to them.
+        arch = self._arch()
+        without_ple = replace(arch, hidden_size_per_layer_input = 0)
+        self.assertEqual(
+            compute_lora_params(arch, 16, ["all-linear"]),
+            compute_lora_params(without_ple, 16, ["all-linear"]),
+        )
+
+    def test_explicit_target_modules_does_not_add_per_layer_input(self):
+        arch = self._arch()
+        without_ple = replace(arch, hidden_size_per_layer_input = 0)
+        self.assertEqual(
+            compute_lora_params(arch, 16, ["q_proj", "v_proj"]),
+            compute_lora_params(without_ple, 16, ["q_proj", "v_proj"]),
+        )
+
+
+class TestDenseMlpLayerFallback(unittest.TestCase):
+    def test_falls_back_to_count_when_indices_empty(self):
+        from utils.hardware.vram_estimation import _is_dense_mlp_layer
+
+        arch = ModelArchConfig(
+            hidden_size = 512,
+            num_hidden_layers = 4,
+            num_attention_heads = 8,
+            num_key_value_heads = 2,
+            intermediate_size = 1024,
+            vocab_size = 1024,
+            num_experts = 4,
+            moe_intermediate_size = 256,
+            num_dense_layers = 2,
+        )
+        self.assertTrue(_is_dense_mlp_layer(arch, 0))
+        self.assertTrue(_is_dense_mlp_layer(arch, 1))
+        self.assertFalse(_is_dense_mlp_layer(arch, 2))
+        self.assertFalse(_is_dense_mlp_layer(arch, 3))
+
+
+class TestExpertsSkipGranularity(unittest.TestCase):
+    def _arch(self):
+        return ModelArchConfig(
+            hidden_size = 512,
+            num_hidden_layers = 4,
+            num_attention_heads = 8,
+            num_key_value_heads = 2,
+            intermediate_size = 1024,
+            vocab_size = 1024,
+            tie_word_embeddings = True,
+            num_experts = 8,
+            moe_intermediate_size = 512,
+            num_dense_layers = 0,
+            head_dim = 64,
+            layer_types = ["full_attention"] * 4,
+            moe_has_dense_mlp = True,
+        )
+
+    def test_experts_skip_excludes_parallel_dense_projections(self):
+        no_skip = self._arch()
+        skip_experts = replace(
+            no_skip,
+            quantization_skip_modules = ["model.layers.0.mlp.experts"],
+        )
+        skip_full_mlp = replace(
+            no_skip,
+            quantization_skip_modules = ["model.layers.0.mlp"],
+        )
+        bytes_no_skip = compute_model_weights_bytes(no_skip, "qlora", True)
+        bytes_skip_experts = compute_model_weights_bytes(skip_experts, "qlora", True)
+        bytes_skip_mlp = compute_model_weights_bytes(skip_full_mlp, "qlora", True)
+        # why: under gemma4 enable_moe_block, `self.experts` is a sibling of
+        # `self.mlp`; skipping `model.layers.0.mlp` should cover only the
+        # dense MLP, while `model.layers.0.mlp.experts` covers the routed
+        # experts. Routed experts have far more params than the dense MLP,
+        # so skipping experts must add more bytes than skipping the dense
+        # path.
+        self.assertGreater(bytes_skip_experts, bytes_no_skip)
+        self.assertGreater(bytes_skip_mlp, bytes_no_skip)
+        self.assertGreater(bytes_skip_experts, bytes_skip_mlp)
 
 
 class TestSharedExperts(unittest.TestCase):
@@ -608,6 +1325,16 @@ class TestMLA(unittest.TestCase):
         lora_p = compute_lora_params(DEEPSEEK_V3, 16, ["q_proj", "v_proj", "o_proj"])
         self.assertGreater(lora_p, 0)
 
+    def test_mla_with_head_dim_does_not_route_through_structured(self):
+        from utils.hardware.vram_estimation import _uses_structured_layer_shapes
+
+        mla_with_head_dim = replace(DEEPSEEK_V3, head_dim = 128)
+        self.assertFalse(_uses_structured_layer_shapes(mla_with_head_dim))
+        self.assertEqual(
+            compute_lora_params(DEEPSEEK_V3, 16, ["q_proj", "v_proj", "o_proj"]),
+            compute_lora_params(mla_with_head_dim, 16, ["q_proj", "v_proj", "o_proj"]),
+        )
+
 
 class TestDenseMoEMix(unittest.TestCase):
     def test_dense_layers_change_total(self):
@@ -691,5 +1418,952 @@ class TestDenseMoEMix(unittest.TestCase):
         self.assertNotEqual(lora_all, lora_mix)
 
 
+class TestMlpLayerTypesDispatch(unittest.TestCase):
+    def _hf(self, **fields):
+        text_config = SimpleNamespace(
+            hidden_size = 64,
+            num_hidden_layers = 4,
+            num_attention_heads = 4,
+            num_key_value_heads = 4,
+            intermediate_size = 128,
+            vocab_size = 1000,
+            tie_word_embeddings = True,
+            num_local_experts = 4,
+            moe_intermediate_size = 32,
+            **fields,
+        )
+        return SimpleNamespace(text_config = text_config, quantization_config = {})
+
+    def test_mlp_layer_types_drives_dense_indices(self):
+        hf = self._hf(mlp_layer_types = ["sparse", "dense", "sparse", "dense"])
+        arch = extract_arch_config(hf)
+        self.assertIsNotNone(arch)
+        self.assertEqual(arch.dense_layer_indices, (1, 3))
+        self.assertEqual(arch.num_dense_layers, 2)
+
+    def test_mlp_layer_types_takes_priority_over_first_k_dense_replace(self):
+        hf = self._hf(
+            mlp_layer_types = ["dense", "sparse", "dense", "sparse"],
+            first_k_dense_replace = 3,
+        )
+        arch = extract_arch_config(hf)
+        self.assertEqual(arch.dense_layer_indices, (0, 2))
+
+    def test_mlp_layer_types_ignores_unknown_entries(self):
+        hf = self._hf(mlp_layer_types = ["dense", "moe", "dense", "linear"])
+        arch = extract_arch_config(hf)
+        self.assertEqual(arch.dense_layer_indices, (0, 2))
+
+    def test_mlp_layer_types_shorter_than_layers_only_marks_present(self):
+        hf = self._hf(mlp_layer_types = ["dense", "sparse"])
+        arch = extract_arch_config(hf)
+        self.assertEqual(arch.dense_layer_indices, (0,))
+
+    def test_empty_mlp_layer_types_falls_through_to_first_k(self):
+        hf = self._hf(mlp_layer_types = [], first_k_dense_replace = 2)
+        arch = extract_arch_config(hf)
+        self.assertEqual(arch.dense_layer_indices, (0, 1))
+
+
+class TestPerLayerInputSkipAlias(unittest.TestCase):
+    def _hf(self, skip):
+        text_config = SimpleNamespace(
+            hidden_size = 64,
+            num_hidden_layers = 2,
+            num_attention_heads = 4,
+            num_key_value_heads = 4,
+            intermediate_size = 128,
+            vocab_size = 1000,
+            tie_word_embeddings = True,
+            hidden_size_per_layer_input = 8,
+            vocab_size_per_layer_input = 256,
+        )
+        return SimpleNamespace(
+            text_config = text_config,
+            quantization_config = {"llm_int8_skip_modules": list(skip)},
+        )
+
+    def test_per_layer_input_gate_skip_pulls_nonzero_delta(self):
+        from utils.hardware.vram_estimation import _compute_skipped_quantizable_elements
+
+        arch = extract_arch_config(self._hf(["model.layers.0.per_layer_input_gate"]))
+        delta = _compute_skipped_quantizable_elements(arch)
+        self.assertEqual(delta, arch.hidden_size * arch.hidden_size_per_layer_input)
+
+    def test_per_layer_model_projection_skip_pulls_global_delta(self):
+        from utils.hardware.vram_estimation import _compute_skipped_quantizable_elements
+
+        arch = extract_arch_config(self._hf(["model.per_layer_model_projection"]))
+        delta = _compute_skipped_quantizable_elements(arch)
+        self.assertEqual(
+            delta,
+            arch.hidden_size
+            * arch.num_hidden_layers
+            * arch.hidden_size_per_layer_input,
+        )
+
+    def test_layer_aggregate_skip_includes_per_layer_input_modules(self):
+        from utils.hardware.vram_estimation import (
+            _compute_skipped_quantizable_elements,
+        )
+
+        arch_with = extract_arch_config(self._hf(["model.layers.0"]))
+        # The text.layers.0 aggregate must include the PLE per-layer modules,
+        # so the same skip on a config without PLE produces a smaller value.
+        arch_without = extract_arch_config(
+            SimpleNamespace(
+                text_config = SimpleNamespace(
+                    hidden_size = 64,
+                    num_hidden_layers = 2,
+                    num_attention_heads = 4,
+                    num_key_value_heads = 4,
+                    intermediate_size = 128,
+                    vocab_size = 1000,
+                    tie_word_embeddings = True,
+                    hidden_size_per_layer_input = 0,
+                    vocab_size_per_layer_input = 0,
+                ),
+                quantization_config = {"llm_int8_skip_modules": ["model.layers.0"]},
+            )
+        )
+        self.assertGreater(
+            _compute_skipped_quantizable_elements(arch_with),
+            _compute_skipped_quantizable_elements(arch_without),
+        )
+
+
+class TestAllLinearStringHandling(unittest.TestCase):
+    def test_compute_lora_params_accepts_bare_all_linear_string(self):
+        list_form = compute_lora_params(LLAMA_8B, 16, ["all-linear"])
+        str_form = compute_lora_params(LLAMA_8B, 16, "all-linear")
+        self.assertEqual(list_form, str_form)
+        self.assertGreater(list_form, 0)
+
+    def test_compute_lora_params_string_with_underscores_normalized(self):
+        list_form = compute_lora_params(LLAMA_8B, 16, ["all_linear"])
+        str_form = compute_lora_params(LLAMA_8B, 16, "all_linear")
+        self.assertEqual(list_form, str_form)
+        self.assertGreater(str_form, 0)
+
+
+class TestSharedExpertVariants(unittest.TestCase):
+    def _hf(self, **fields):
+        text_config = SimpleNamespace(
+            hidden_size = 256,
+            num_hidden_layers = 4,
+            num_attention_heads = 8,
+            num_key_value_heads = 4,
+            intermediate_size = 1024,
+            vocab_size = 1000,
+            tie_word_embeddings = False,
+            num_local_experts = 8,
+            moe_intermediate_size = 128,
+            **fields,
+        )
+        return SimpleNamespace(text_config = text_config, quantization_config = {})
+
+    def test_shared_expert_intermediate_size_extracted_and_infers_count(self):
+        arch = extract_arch_config(self._hf(shared_expert_intermediate_size = 64))
+        self.assertEqual(arch.shared_expert_intermediate_size, 64)
+        self.assertEqual(arch.n_shared_experts, 1)
+
+    def test_num_shared_experts_alias_extracted(self):
+        arch = extract_arch_config(self._hf(num_shared_experts = 2))
+        self.assertEqual(arch.n_shared_experts, 2)
+
+    def test_n_shared_experts_takes_priority_over_alias(self):
+        arch = extract_arch_config(self._hf(n_shared_experts = 3, num_shared_experts = 99))
+        self.assertEqual(arch.n_shared_experts, 3)
+
+    def test_shared_expert_size_separate_from_routed_changes_weight_count(self):
+        from utils.hardware.vram_estimation import _compute_moe_mlp_elements
+
+        arch_separate = extract_arch_config(
+            self._hf(shared_expert_intermediate_size = 64)
+        )
+        arch_implicit = extract_arch_config(self._hf(n_shared_experts = 1))
+        # Different shared sizes (64 vs default moe_intermediate_size=128) must
+        # produce different MoE element counts.
+        self.assertNotEqual(
+            _compute_moe_mlp_elements(arch_separate),
+            _compute_moe_mlp_elements(arch_implicit),
+        )
+
+    def test_shared_expert_gate_counted_only_for_qwen_style(self):
+        from utils.hardware.vram_estimation import _compute_moe_mlp_elements
+
+        # Qwen-style: shared_expert_intermediate_size set -> shared_expert_gate counted.
+        qwen_arch = extract_arch_config(self._hf(shared_expert_intermediate_size = 64))
+        hd = qwen_arch.hidden_size
+        ms = qwen_arch.moe_intermediate_size
+        ne = qwen_arch.num_experts
+        ss = qwen_arch.shared_expert_intermediate_size
+        expected = hd * ms * 3 * ne + ne * hd + hd * ss * 3 * 1 + 1 * hd
+        self.assertEqual(_compute_moe_mlp_elements(qwen_arch), expected)
+
+        # Non-Qwen shared experts (e.g. Exaone-MoE) -> no shared_expert_gate.
+        plain_arch = extract_arch_config(self._hf(n_shared_experts = 1))
+        hd = plain_arch.hidden_size
+        ms = plain_arch.moe_intermediate_size
+        ne = plain_arch.num_experts
+        expected_plain = hd * ms * 3 * ne + ne * hd + hd * ms * 3 * 1
+        self.assertEqual(_compute_moe_mlp_elements(plain_arch), expected_plain)
+
+
+class TestSharedExpertActivation(unittest.TestCase):
+    def _make(self, **fields):
+        text_config = SimpleNamespace(
+            hidden_size = 512,
+            num_hidden_layers = 4,
+            num_attention_heads = 8,
+            num_key_value_heads = 4,
+            intermediate_size = 1024,
+            vocab_size = 1000,
+            tie_word_embeddings = False,
+            num_local_experts = 4,
+            moe_intermediate_size = 64,
+            **fields,
+        )
+        return extract_arch_config(
+            SimpleNamespace(text_config = text_config, quantization_config = {})
+        )
+
+    def test_shared_expert_increases_activation_bytes(self):
+        with_shared = self._make(shared_expert_intermediate_size = 64)
+        without = self._make()
+        self.assertGreater(
+            compute_activation_bytes(
+                with_shared,
+                2,
+                1024,
+                "none",
+                is_lora = True,
+                attention_implementation = "flash_attention_2",
+            ),
+            compute_activation_bytes(
+                without,
+                2,
+                1024,
+                "none",
+                is_lora = True,
+                attention_implementation = "flash_attention_2",
+            ),
+        )
+
+    def test_shared_expert_plus_dense_block_compose(self):
+        # gemma4 enable_moe_block with hypothetical shared expert: dense + routed
+        # + shared all live per layer; mlp_size should sum all three terms.
+        from utils.hardware.vram_estimation import _layer_qkv_mlp_sizes
+
+        arch = self._make(
+            enable_moe_block = True,
+            shared_expert_intermediate_size = 32,
+            head_dim = 64,
+            layer_types = ["full_attention"] * 4,
+        )
+        _, mlp_size = _layer_qkv_mlp_sizes(arch, 0)
+        # routed (64) + shared (32) + parallel dense intermediate (1024)
+        self.assertEqual(mlp_size, 64 + 32 + 1024)
+
+
+class TestPerLayerInputActivation(unittest.TestCase):
+    def _make(self, **fields):
+        text_config = SimpleNamespace(
+            hidden_size = 512,
+            num_hidden_layers = 4,
+            num_attention_heads = 8,
+            num_key_value_heads = 4,
+            intermediate_size = 1024,
+            vocab_size = 1000,
+            tie_word_embeddings = False,
+            **fields,
+        )
+        return extract_arch_config(
+            SimpleNamespace(text_config = text_config, quantization_config = {})
+        )
+
+    def test_ple_increases_activation_bytes(self):
+        with_ple = self._make(
+            hidden_size_per_layer_input = 64,
+            vocab_size_per_layer_input = 256,
+        )
+        without = self._make()
+        self.assertGreater(
+            compute_activation_bytes(
+                with_ple,
+                2,
+                1024,
+                "none",
+                is_lora = True,
+                attention_implementation = "flash_attention_2",
+            ),
+            compute_activation_bytes(
+                without,
+                2,
+                1024,
+                "none",
+                is_lora = True,
+                attention_implementation = "flash_attention_2",
+            ),
+        )
+
+    def test_ple_zero_does_not_inflate_activations(self):
+        without = self._make(hidden_size_per_layer_input = 0)
+        baseline = self._make()
+        self.assertEqual(
+            compute_activation_bytes(
+                without,
+                2,
+                512,
+                "none",
+                is_lora = True,
+                attention_implementation = "flash_attention_2",
+            ),
+            compute_activation_bytes(
+                baseline,
+                2,
+                512,
+                "none",
+                is_lora = True,
+                attention_implementation = "flash_attention_2",
+            ),
+        )
+
+
+class TestKvSharedActivation(unittest.TestCase):
+    def _make(self, kv_shared):
+        text_config = SimpleNamespace(
+            hidden_size = 512,
+            num_hidden_layers = 4,
+            num_attention_heads = 8,
+            num_key_value_heads = 4,
+            intermediate_size = 1024,
+            vocab_size = 1000,
+            tie_word_embeddings = False,
+            head_dim = 64,
+            num_kv_shared_layers = kv_shared,
+            layer_types = ["full_attention"] * 4,
+        )
+        return extract_arch_config(
+            SimpleNamespace(text_config = text_config, quantization_config = {})
+        )
+
+    def test_kv_shared_layers_keep_activation_bytes(self):
+        shared = self._make(kv_shared = 2)
+        full = self._make(kv_shared = 0)
+        self.assertEqual(
+            compute_activation_bytes(
+                shared,
+                2,
+                1024,
+                "none",
+                is_lora = True,
+                attention_implementation = "flash_attention_2",
+            ),
+            compute_activation_bytes(
+                full,
+                2,
+                1024,
+                "none",
+                is_lora = True,
+                attention_implementation = "flash_attention_2",
+            ),
+        )
+
+
+class TestSparseMoeSkipAliases(unittest.TestCase):
+    def _hf(self, skip, **fields):
+        text_config = SimpleNamespace(
+            hidden_size = 128,
+            num_hidden_layers = 2,
+            num_attention_heads = 4,
+            num_key_value_heads = 4,
+            intermediate_size = 256,
+            vocab_size = 1000,
+            tie_word_embeddings = False,
+            num_local_experts = 4,
+            moe_intermediate_size = 64,
+            **fields,
+        )
+        return SimpleNamespace(
+            text_config = text_config,
+            quantization_config = {"llm_int8_skip_modules": list(skip)},
+        )
+
+    def test_gemma4_layers_experts_alias_pulls_routed(self):
+        from utils.hardware.vram_estimation import _compute_skipped_quantizable_elements
+
+        arch = extract_arch_config(
+            self._hf(["model.layers.0.experts"], enable_moe_block = True)
+        )
+        self.assertGreater(_compute_skipped_quantizable_elements(arch), 0)
+
+    def test_qwen_shared_expert_skip_pulls_only_shared(self):
+        from utils.hardware.vram_estimation import _compute_skipped_quantizable_elements
+
+        arch = extract_arch_config(
+            self._hf(
+                ["model.layers.0.mlp.shared_expert"],
+                shared_expert_intermediate_size = 32,
+            )
+        )
+        # shared_expert delta only -- routed mlp.experts is NOT skipped.
+        delta = _compute_skipped_quantizable_elements(arch)
+        self.assertGreater(delta, 0)
+        full_layer = extract_arch_config(
+            self._hf(
+                ["model.layers.0.mlp"],
+                shared_expert_intermediate_size = 32,
+            )
+        )
+        self.assertGreater(
+            _compute_skipped_quantizable_elements(full_layer),
+            delta,
+        )
+
+    def test_exaone_shared_experts_plural_alias(self):
+        from utils.hardware.vram_estimation import _compute_skipped_quantizable_elements
+
+        arch = extract_arch_config(
+            self._hf(
+                ["model.layers.0.mlp.shared_experts"],
+                num_shared_experts = 1,
+            )
+        )
+        self.assertGreater(_compute_skipped_quantizable_elements(arch), 0)
+
+
+class TestAllLinearMoELoraExclusion(unittest.TestCase):
+    def _arch(self, **fields):
+        text_config = SimpleNamespace(
+            hidden_size = 256,
+            num_hidden_layers = 2,
+            num_attention_heads = 4,
+            num_key_value_heads = 4,
+            intermediate_size = 512,
+            vocab_size = 1000,
+            tie_word_embeddings = False,
+            num_local_experts = 8,
+            moe_intermediate_size = 64,
+            **fields,
+        )
+        return extract_arch_config(
+            SimpleNamespace(text_config = text_config, quantization_config = {})
+        )
+
+    def test_all_linear_drops_routed_moe_expert_lora(self):
+        arch = self._arch()
+        all_linear = compute_lora_params(arch, 8, "all-linear")
+        explicit = compute_lora_params(arch, 8, ["gate_proj", "up_proj", "down_proj"])
+        self.assertLess(all_linear, explicit)
+
+    def test_all_linear_drops_shared_expert_lora(self):
+        arch = self._arch(shared_expert_intermediate_size = 32)
+        all_linear = compute_lora_params(arch, 8, "all-linear")
+        explicit = compute_lora_params(arch, 8, ["gate_proj", "up_proj", "down_proj"])
+        # explicit includes routed + shared MoE; all-linear includes neither.
+        self.assertLess(all_linear, explicit)
+
+    def test_all_linear_includes_attention_lora(self):
+        arch = self._arch()
+        all_linear = compute_lora_params(arch, 8, "all-linear")
+        attn_only = compute_lora_params(
+            arch, 8, ["q_proj", "k_proj", "v_proj", "o_proj"]
+        )
+        # all-linear still attaches to attention nn.Linear modules.
+        self.assertGreaterEqual(all_linear, attn_only)
+
+
+class TestExplicitPerLayerInputLora(unittest.TestCase):
+    def _arch(self):
+        text_config = SimpleNamespace(
+            hidden_size = 256,
+            num_hidden_layers = 3,
+            num_attention_heads = 4,
+            num_key_value_heads = 4,
+            intermediate_size = 512,
+            vocab_size = 1000,
+            tie_word_embeddings = False,
+            hidden_size_per_layer_input = 32,
+            vocab_size_per_layer_input = 128,
+        )
+        return extract_arch_config(
+            SimpleNamespace(text_config = text_config, quantization_config = {})
+        )
+
+    def test_explicit_per_layer_input_gate_returns_nonzero(self):
+        arch = self._arch()
+        result = compute_lora_params(arch, 16, ["per_layer_input_gate"])
+        self.assertGreater(result, 0)
+
+    def test_explicit_per_layer_projection_returns_nonzero(self):
+        arch = self._arch()
+        result = compute_lora_params(arch, 16, ["per_layer_projection"])
+        self.assertGreater(result, 0)
+
+    def test_explicit_per_layer_model_projection_returns_nonzero(self):
+        arch = self._arch()
+        result = compute_lora_params(arch, 16, ["per_layer_model_projection"])
+        self.assertGreater(result, 0)
+
+    def test_explicit_ple_string_target_handled(self):
+        # Bare-string target with a PLE name should not be iterated char-by-char.
+        arch = self._arch()
+        list_form = compute_lora_params(arch, 16, ["per_layer_input_gate"])
+        str_form = compute_lora_params(arch, 16, "per_layer_input_gate")
+        self.assertEqual(list_form, str_form)
+
+
+class TestTopKExpertActivation(unittest.TestCase):
+    def _make(self, **fields):
+        text_config = SimpleNamespace(
+            hidden_size = 512,
+            num_hidden_layers = 4,
+            num_attention_heads = 8,
+            num_key_value_heads = 4,
+            intermediate_size = 1024,
+            vocab_size = 1000,
+            tie_word_embeddings = False,
+            num_local_experts = 8,
+            moe_intermediate_size = 64,
+            **fields,
+        )
+        return extract_arch_config(
+            SimpleNamespace(text_config = text_config, quantization_config = {})
+        )
+
+    def test_num_experts_per_tok_extracted(self):
+        arch = self._make(num_experts_per_tok = 4)
+        self.assertEqual(arch.num_experts_per_tok, 4)
+
+    def test_top_k_experts_alias_extracted(self):
+        arch = self._make(top_k_experts = 8)
+        self.assertEqual(arch.num_experts_per_tok, 8)
+
+    def test_default_top_k_one_unchanged(self):
+        arch = self._make()
+        self.assertEqual(arch.num_experts_per_tok, 1)
+
+    def test_top_k_scales_moe_activation(self):
+        single = self._make()
+        multi = self._make(num_experts_per_tok = 8)
+        single_act = compute_activation_bytes(
+            single,
+            2,
+            512,
+            "none",
+            is_lora = True,
+            attention_implementation = "flash_attention_2",
+        )
+        multi_act = compute_activation_bytes(
+            multi,
+            2,
+            512,
+            "none",
+            is_lora = True,
+            attention_implementation = "flash_attention_2",
+        )
+        self.assertGreater(multi_act, single_act)
+
+
+class TestErnieMoEListConfig(unittest.TestCase):
+    def _hf(self, **fields):
+        text_config = SimpleNamespace(
+            hidden_size = 256,
+            num_hidden_layers = 4,
+            num_attention_heads = 4,
+            num_key_value_heads = 4,
+            intermediate_size = 1024,
+            vocab_size = 1000,
+            tie_word_embeddings = False,
+            **fields,
+        )
+        return SimpleNamespace(text_config = text_config, quantization_config = {})
+
+    def test_list_moe_intermediate_size_scalarized(self):
+        arch = extract_arch_config(
+            self._hf(
+                moe_num_experts = 32,
+                moe_intermediate_size = [1536, 512],
+            )
+        )
+        # why: ERNIE 4.5 VL MoE encodes [text_routed, vision_routed]; the
+        # second element is the vision-routed expert width, not the shared
+        # expert width. Shared experts are sized from the text-routed width
+        # (= moe_intermediate_size[0]) when moe_num_shared_experts is set.
+        self.assertEqual(arch.moe_intermediate_size, 1536)
+        self.assertIsNone(arch.shared_expert_intermediate_size)
+        self.assertEqual(arch.n_shared_experts, 0)
+
+    def test_moe_num_experts_alias_extracted(self):
+        arch = extract_arch_config(
+            self._hf(
+                moe_num_experts = 64,
+                moe_intermediate_size = 1024,
+            )
+        )
+        self.assertEqual(arch.num_experts, 64)
+
+    def test_moe_num_shared_experts_alias_extracted(self):
+        arch = extract_arch_config(
+            self._hf(
+                moe_num_experts = 16,
+                moe_num_shared_experts = 2,
+                moe_intermediate_size = 1024,
+            )
+        )
+        self.assertEqual(arch.n_shared_experts, 2)
+
+    def test_explicit_shared_size_overrides_list_second_element(self):
+        arch = extract_arch_config(
+            self._hf(
+                moe_num_experts = 8,
+                moe_intermediate_size = [1536, 512],
+                shared_expert_intermediate_size = 256,
+            )
+        )
+        # Explicit shared size wins over moe_intermediate_size[1].
+        self.assertEqual(arch.shared_expert_intermediate_size, 256)
+
+
+class TestSuffixSkipModuleMatch(unittest.TestCase):
+    def _hf(self, skip):
+        text_config = SimpleNamespace(
+            hidden_size = 128,
+            num_hidden_layers = 2,
+            num_attention_heads = 4,
+            num_key_value_heads = 4,
+            intermediate_size = 256,
+            vocab_size = 1000,
+            tie_word_embeddings = False,
+        )
+        return SimpleNamespace(
+            text_config = text_config,
+            quantization_config = {"llm_int8_skip_modules": list(skip)},
+        )
+
+    def test_q_proj_suffix_skip_matches_all_layers(self):
+        from utils.hardware.vram_estimation import _compute_skipped_quantizable_elements
+
+        arch = extract_arch_config(self._hf(["q_proj"]))
+        delta = _compute_skipped_quantizable_elements(arch)
+        # 2 layers * hd * hd of q_proj weight elements.
+        self.assertEqual(delta, 2 * arch.hidden_size * arch.hidden_size)
+
+    def test_self_attn_aggregate_skip_matches_aggregate(self):
+        from utils.hardware.vram_estimation import _compute_skipped_quantizable_elements
+
+        arch = extract_arch_config(self._hf(["self_attn"]))
+        # The aggregate text.layers.<i>.self_attn matches; total covers both layers.
+        delta = _compute_skipped_quantizable_elements(arch)
+        self.assertGreater(delta, 0)
+
+    def test_vision_prefix_skip_does_not_match_text_alias(self):
+        from utils.hardware.vram_estimation import _module_path_matches
+
+        # vision_tower-prefixed full path must NOT match text-tower aliases.
+        self.assertFalse(
+            _module_path_matches(
+                "vision_tower.model.layers.0.self_attn.q_proj",
+                "model.layers.0.self_attn.q_proj",
+            )
+        )
+
+
+class TestMultimodalFullModelBytes(unittest.TestCase):
+    def test_extra_bytes_added_when_safetensors_exceeds_text_arch(self):
+        from utils.hardware import hardware as hardware_module
+
+        config = SimpleNamespace(
+            hidden_size = 1024,
+            num_hidden_layers = 4,
+            num_attention_heads = 8,
+            num_key_value_heads = 4,
+            intermediate_size = 2048,
+            vocab_size = 32000,
+            tie_word_embeddings = False,
+        )
+        # Force safetensors size >>> arch text-only bytes.
+        big_safetensors = 20 * 1024**3
+        with (
+            patch.object(
+                hardware_module,
+                "_load_config_for_gpu_estimate",
+                return_value = config,
+            ),
+            patch.object(
+                hardware_module,
+                "estimate_fp16_model_size_bytes",
+                return_value = (big_safetensors, "safetensors"),
+            ),
+            patch.object(
+                hardware_module,
+                "_determine_attention_impl_for_gpu_estimate",
+                return_value = "flash_attention_2",
+            ),
+            patch.object(
+                hardware_module,
+                "get_visible_gpu_count",
+                return_value = 1,
+            ),
+        ):
+            _, metadata = hardware_module.estimate_required_model_memory_gb(
+                "fake/model",
+                training_type = "LoRA/QLoRA",
+                load_in_4bit = True,
+            )
+        self.assertEqual(metadata.get("estimation_mode"), "detailed")
+        # model_weights_gb must reflect the extra non-text bytes (>5 GB
+        # since text-only arch_fp16 is small for these dims).
+        self.assertGreater(metadata["vram_breakdown"]["model_weights_gb"], 5.0)
+
+    def test_no_extra_when_safetensors_smaller_than_text_arch(self):
+        from utils.hardware import hardware as hardware_module
+
+        config = SimpleNamespace(
+            hidden_size = 4096,
+            num_hidden_layers = 32,
+            num_attention_heads = 32,
+            num_key_value_heads = 8,
+            intermediate_size = 11008,
+            vocab_size = 32000,
+            tie_word_embeddings = False,
+        )
+        tiny_safetensors = 100  # bytes, deliberately absurdly small
+        with (
+            patch.object(
+                hardware_module,
+                "_load_config_for_gpu_estimate",
+                return_value = config,
+            ),
+            patch.object(
+                hardware_module,
+                "estimate_fp16_model_size_bytes",
+                return_value = (tiny_safetensors, "safetensors"),
+            ),
+            patch.object(
+                hardware_module,
+                "_determine_attention_impl_for_gpu_estimate",
+                return_value = "flash_attention_2",
+            ),
+            patch.object(
+                hardware_module,
+                "get_visible_gpu_count",
+                return_value = 1,
+            ),
+        ):
+            required, metadata = hardware_module.estimate_required_model_memory_gb(
+                "fake/model",
+                training_type = "LoRA/QLoRA",
+                load_in_4bit = True,
+            )
+        # No negative extra; required_gb stays a positive finite number.
+        self.assertGreater(required, 0)
+
+
+class TestLlama4ArchExtraction(unittest.TestCase):
+    def _llama4_text_config(self, **fields):
+        base = dict(
+            hidden_size = 2048,
+            num_hidden_layers = 4,
+            num_attention_heads = 16,
+            num_key_value_heads = 4,
+            intermediate_size = 8192,
+            intermediate_size_mlp = 16384,
+            vocab_size = 32000,
+            tie_word_embeddings = True,
+            num_local_experts = 4,
+            num_experts_per_tok = 2,
+        )
+        base.update(fields)
+        return SimpleNamespace(**base)
+
+    def test_llama4_moe_layers_dispatch_uses_explicit_indices(self):
+        from utils.hardware.vram_estimation import _compute_dense_layer_indices
+
+        cfg = SimpleNamespace(num_hidden_layers = 4, moe_layers = [1, 3])
+        self.assertEqual(_compute_dense_layer_indices(cfg, 4), (0, 2))
+
+    def test_llama4_moe_layers_takes_priority_over_first_k_dense_replace(self):
+        from utils.hardware.vram_estimation import _compute_dense_layer_indices
+
+        cfg = SimpleNamespace(
+            num_hidden_layers = 6,
+            moe_layers = [2, 4],
+            first_k_dense_replace = 4,
+        )
+        self.assertEqual(_compute_dense_layer_indices(cfg, 6), (0, 1, 3, 5))
+
+    def test_dense_intermediate_size_picks_up_intermediate_size_mlp(self):
+        from utils.hardware.vram_estimation import _dense_mlp_size
+
+        arch = extract_arch_config(self._llama4_text_config(moe_layers = [1, 3]))
+        self.assertIsNotNone(arch)
+        self.assertEqual(arch.intermediate_size, 8192)
+        self.assertEqual(arch.dense_intermediate_size, 16384)
+        self.assertEqual(_dense_mlp_size(arch), 16384)
+
+    def test_auto_attaches_one_shared_expert_at_routed_width(self):
+        from utils.hardware.vram_estimation import _shared_expert_size
+
+        arch = extract_arch_config(self._llama4_text_config(moe_layers = [1, 3]))
+        self.assertIsNotNone(arch)
+        self.assertEqual(arch.n_shared_experts, 1)
+        self.assertIsNone(arch.shared_expert_intermediate_size)
+        self.assertEqual(_shared_expert_size(arch), arch.intermediate_size)
+
+    def test_non_llama4_config_leaves_dense_intermediate_size_none(self):
+        from utils.hardware.vram_estimation import _dense_mlp_size
+
+        cfg = SimpleNamespace(
+            hidden_size = 1024,
+            num_hidden_layers = 4,
+            num_attention_heads = 8,
+            num_key_value_heads = 2,
+            intermediate_size = 4096,
+            vocab_size = 32000,
+            tie_word_embeddings = True,
+        )
+        arch = extract_arch_config(cfg)
+        self.assertIsNotNone(arch)
+        self.assertIsNone(arch.dense_intermediate_size)
+        self.assertEqual(_dense_mlp_size(arch), 4096)
+
+    def test_intermediate_size_mlp_without_moe_does_not_force_shared_expert(self):
+        cfg = SimpleNamespace(
+            hidden_size = 2048,
+            num_hidden_layers = 4,
+            num_attention_heads = 16,
+            num_key_value_heads = 4,
+            intermediate_size = 8192,
+            intermediate_size_mlp = 16384,
+            vocab_size = 32000,
+            tie_word_embeddings = True,
+        )
+        arch = extract_arch_config(cfg)
+        self.assertIsNotNone(arch)
+        self.assertEqual(arch.dense_intermediate_size, 16384)
+        self.assertEqual(arch.n_shared_experts, 0)
+
+
+class TestDbrxFfnConfigExtraction(unittest.TestCase):
+    def test_extracts_moe_fields_from_ffn_subconfig(self):
+        ffn = SimpleNamespace(moe_num_experts = 4, moe_top_k = 2, ffn_hidden_size = 1024)
+        cfg = SimpleNamespace(
+            hidden_size = 2048,
+            num_hidden_layers = 4,
+            num_attention_heads = 16,
+            num_key_value_heads = 4,
+            intermediate_size = 2048,
+            vocab_size = 32000,
+            tie_word_embeddings = False,
+            ffn_config = ffn,
+        )
+        arch = extract_arch_config(cfg)
+        self.assertIsNotNone(arch)
+        self.assertEqual(arch.num_experts, 4)
+        self.assertEqual(arch.num_experts_per_tok, 2)
+        self.assertEqual(arch.moe_intermediate_size, 1024)
+
+    def test_top_level_attrs_take_precedence_over_ffn_config(self):
+        ffn = SimpleNamespace(moe_num_experts = 4, moe_top_k = 2, ffn_hidden_size = 1024)
+        cfg = SimpleNamespace(
+            hidden_size = 2048,
+            num_hidden_layers = 4,
+            num_attention_heads = 16,
+            num_key_value_heads = 4,
+            intermediate_size = 2048,
+            vocab_size = 32000,
+            tie_word_embeddings = False,
+            ffn_config = ffn,
+            num_local_experts = 16,
+            num_experts_per_tok = 8,
+        )
+        arch = extract_arch_config(cfg)
+        self.assertIsNotNone(arch)
+        self.assertEqual(arch.num_experts, 16)
+        self.assertEqual(arch.num_experts_per_tok, 8)
+
+
+class TestErniePhaseModuloDispatch(unittest.TestCase):
+    def test_phase_modulo_with_interval_two_matches_decoder(self):
+        from utils.hardware.vram_estimation import _compute_dense_layer_indices
+
+        cfg = SimpleNamespace(
+            num_hidden_layers = 10,
+            moe_layer_start_index = 2,
+            moe_layer_end_index = 8,
+            moe_layer_interval = 2,
+        )
+        # Decoder gates by ((i + 1) % 2 == 0) AND 2 <= i <= 8 -> MoE = {3, 5, 7}.
+        self.assertEqual(_compute_dense_layer_indices(cfg, 10), (0, 1, 2, 4, 6, 8, 9))
+
+    def test_phase_modulo_with_interval_three(self):
+        from utils.hardware.vram_estimation import _compute_dense_layer_indices
+
+        cfg = SimpleNamespace(
+            num_hidden_layers = 9,
+            moe_layer_start_index = 0,
+            moe_layer_end_index = -1,
+            moe_layer_interval = 3,
+        )
+        self.assertEqual(_compute_dense_layer_indices(cfg, 9), (0, 1, 3, 4, 6, 7))
+
+
+class TestErnieVlSharedExpertWidth(unittest.TestCase):
+    def test_shared_expert_width_uses_text_routed_not_vision(self):
+        from utils.hardware.vram_estimation import (
+            _compute_shared_moe_elements,
+            _shared_expert_size,
+        )
+
+        cfg = SimpleNamespace(
+            text_config = SimpleNamespace(
+                hidden_size = 1024,
+                num_hidden_layers = 4,
+                num_attention_heads = 8,
+                num_key_value_heads = 4,
+                intermediate_size = 2048,
+                vocab_size = 32000,
+                tie_word_embeddings = False,
+                moe_num_experts = 8,
+                moe_num_shared_experts = 2,
+                moe_intermediate_size = [1536, 512],
+            ),
+            quantization_config = {},
+        )
+        arch = extract_arch_config(cfg)
+        self.assertIsNotNone(arch)
+        self.assertIsNone(arch.shared_expert_intermediate_size)
+        self.assertEqual(arch.moe_intermediate_size, 1536)
+        self.assertEqual(arch.n_shared_experts, 2)
+        self.assertEqual(_shared_expert_size(arch), 1536)
+        self.assertEqual(_compute_shared_moe_elements(arch), 1024 * 1536 * 3 * 2)
+
+    def test_qwen_style_explicit_shared_expert_size_still_adds_gate(self):
+        from utils.hardware.vram_estimation import _compute_shared_moe_elements
+
+        cfg = SimpleNamespace(
+            hidden_size = 1024,
+            num_hidden_layers = 4,
+            num_attention_heads = 8,
+            num_key_value_heads = 4,
+            intermediate_size = 2048,
+            vocab_size = 32000,
+            tie_word_embeddings = False,
+            num_local_experts = 8,
+            moe_intermediate_size = 256,
+            shared_expert_intermediate_size = 768,
+        )
+        arch = extract_arch_config(cfg)
+        self.assertIsNotNone(arch)
+        self.assertEqual(arch.shared_expert_intermediate_size, 768)
+        self.assertEqual(arch.n_shared_experts, 1)
+        self.assertEqual(
+            _compute_shared_moe_elements(arch),
+            1024 * 768 * 3 + 1 * 1024,
+        )
+
+
 if __name__ == "__main__":
     unittest.main()
diff --git a/studio/backend/utils/hardware/VRAM_ESTIMATION.md b/studio/backend/utils/hardware/VRAM_ESTIMATION.md
index 26072b208f..a6b4de29d2 100644
--- a/studio/backend/utils/hardware/VRAM_ESTIMATION.md
+++ b/studio/backend/utils/hardware/VRAM_ESTIMATION.md
@@ -33,7 +33,13 @@ Non-quantizable = 2*H*L + V*H + (V*H if not tie_embeddings else 0)
 | QLoRA 4-bit | `Quantizable * 2 / 3.2 + Non-quantizable * 2` |
 | LoRA / Full fp16 | `(Quantizable + Non-quantizable) * 2` |
 
-The 3.2 factor (`16/5`) accounts for BNB NF4 blockwise scales.
+The 3.2 factor (`16/5`) accounts for BNB NF4 blockwise scales. Repos whose
+quantization config enables `bnb_4bit_use_double_quant` use a tighter, still
+conservative 3.6 factor for the quantized portion of the weights.
+When a 4-bit config has `llm_int8_skip_modules` entries that point to language
+model layers or submodules, those quantizable weights are charged at fp16
+instead of NF4. Generic embedding and multimodal skip names are already covered
+by non-quantizable terms or excluded from text training weights.
 
 ## 2. LoRA Adapters
 
@@ -53,6 +59,18 @@ MLP modules multiply by `E` for MoE.
 LoRA_bytes = sum(A + B per selected module) * L * 2
 ```
 
+`all-linear` is treated as all known text linear modules in the table above.
+The estimator deliberately does not infer multimodal or vision-tower LoRA
+modules from config shapes; those modules vary too much across VLM families for
+a generic config formula.
+
+Some decoder configs expose layer-shape fields such as `layer_types`,
+`head_dim`, `global_head_dim`, `num_global_key_value_heads`, `attention_k_eq_v`,
+`num_kv_shared_layers`, `use_double_wide_mlp`, `vocab_size_per_layer_input`, and
+`hidden_size_per_layer_input`. When those fields are present, the estimator
+derives text weight and LoRA counts from the per-layer shapes instead of
+assuming every layer has the same seven projection modules.
+
 ## 3. Optimizer States (calibrated)
 
 | Optimizer | Bytes/param | Notes |
@@ -77,6 +95,21 @@ Per-layer (from `unsloth_zoo/vllm_utils.py`):
 Per_layer = (S*B*(H+K+K) + S*B*2 + S*B*(M+M)) * 2 * 1.25
 ```
 
+When the resolved attention implementation is none of `flash_attention_2`,
+`sdpa`, or `flex_attention` (PyTorch SDPA dispatches to flash or
+memory-efficient kernels and FlexAttention is also a memory-efficient
+kernel, all of which are O(n) in memory), activation memory also includes
+a quadratic attention-score/workspace estimate:
+
+```
+Non_flash_attention = B * num_attention_heads * S^2 * 2 * 12.0 * effective_layers
+Activations = max(Per_layer_with_gc, Non_flash_attention)
+```
+
+Studio resolves the attention implementation with Unsloth's
+`resolve_attention_implementation` helper and uses that result directly. The
+estimator does not duplicate model-family attention policy.
+
 | GC Mode | Full FT | LoRA/QLoRA |
 |---------|---------|------------|
 | none | `L` layers | `L` layers |
@@ -85,13 +118,33 @@ Per_layer = (S*B*(H+K+K) + S*B*2 + S*B*(M+M)) * 2 * 1.25
 
 ## 6. Floors
 
-Gradients and activations have minimum floors at **15% of model weight memory** to account for autograd overhead, attention score matrices, NCCL buffers, mixed-precision scaling, and PyTorch fragmentation.
+Activations use the computed formula directly:
 
 ```
-gradient_bytes  = max(computed, weights * 0.15)
-activation_bytes = max(computed, weights * 0.15 * B/2)
+activation_bytes = computed_activation_bytes
 ```
 
+Full fine-tuning keeps the gradient floor at **15% of model weight memory** to
+account for autograd overhead, NCCL buffers, mixed-precision scaling, and
+PyTorch fragmentation:
+
+```
+gradient_bytes = max(computed_gradient_bytes, weights * 0.15)
+```
+
+For LoRA/QLoRA, the base model is frozen, so the weight-derived gradient floor
+is capped by trainable-state and live-activation scale:
+
+```
+raw_gradient_bytes = trainable_params * 2
+gradient_floor = min(weights * 0.15, max(computed_activation_bytes, optimizer_bytes))
+gradient_bytes = max(raw_gradient_bytes, gradient_floor)
+```
+
+This prevents frozen quantized model size from dominating gradient/state
+overhead when the measured runtime footprint is governed by LoRA optimizer
+states and live activations.
+
 ## 7. CUDA Overhead
 
 **1.4 GB** fixed — CUDA driver + PyTorch runtime, calibrated on RTX 5070 Ti.
@@ -106,34 +159,6 @@ usable_gb = free[gpu_0] + sum(free[gpu_i] * 0.85 for i in 1..N)
 
 ---
 
-## Reference Table (bsz=2, seq=2048, rank=16, GC=unsloth, adamw_8bit)
-
-| Model | Weights | LoRA | Optim | Grad | Act | CUDA | Total |
-|-------|---------|------|-------|------|-----|------|-------|
-| 0.5B QLoRA | 0.5 | 0.0 | 0.0 | 0.1 | 0.1 | 1.4 | **2.1** |
-| 1B QLoRA | 1.1 | 0.0 | 0.0 | 0.2 | 0.2 | 1.4 | **2.9** |
-| 3B QLoRA | 2.4 | 0.0 | 0.1 | 0.5 | 0.5 | 1.4 | **4.9** |
-| 8B QLoRA | 6.0 | 0.1 | 0.2 | 1.2 | 1.2 | 1.4 | **10.1** |
-| 8B LoRA fp16 | 15.0 | 0.1 | 0.2 | 3.0 | 3.0 | 1.4 | **22.6** |
-| 8B Full FT | 15.0 | — | 29.9 | 15.0 | 3.0 | 1.4 | **64.2** |
-| 32B LoRA fp16 | 61.0 | 0.2 | 0.5 | 12.2 | 12.2 | 1.4 | **87.6** |
-| 72B QLoRA | 45.5 | 0.4 | 0.8 | 9.1 | 9.1 | 1.4 | **66.3** |
-
-## E2E Validation (Llama-3.2-1B, B200 emulating 24GB)
-
-| Config | Estimated | Actual (nvsmi) | Error |
-|--------|----------|----------------|-------|
-| QLoRA bsz=2 seq=512 | 2.55 GB | 2.65 GB | -3.7% |
-| QLoRA bsz=2 seq=2048 | 2.60 GB | 2.65 GB | -1.8% |
-| QLoRA bsz=4 seq=2048 | 2.65 GB | 2.65 GB | +0.0% |
-| LoRA fp16 bsz=2 | 3.84 GB | 3.88 GB | -1.0% |
-| Full FT adamw_8bit | 10.89 GB | 10.80 GB | +0.8% |
-| Full FT adamw_torch | 13.19 GB | 12.93 GB | +2.0% |
-
-*Note: e2e numbers predate the 15% floors, which add safety margin on top.*
-
----
-
 ## Parameter Flow
 
 ```
diff --git a/studio/backend/utils/hardware/hardware.py b/studio/backend/utils/hardware/hardware.py
index be31c00a78..c218b7b4b9 100644
--- a/studio/backend/utils/hardware/hardware.py
+++ b/studio/backend/utils/hardware/hardware.py
@@ -774,6 +774,34 @@ def _load_config_for_gpu_estimate(model_name: str, hf_token: Optional[str] = Non
         return None
 
 
+def _determine_attention_impl_for_gpu_estimate(config) -> str:
+    import copy as _copy
+
+    from unsloth.models._utils import resolve_attention_implementation
+    from transformers import AutoModel, AutoModelForCausalLM
+
+    # why: resolve_attention_implementation calls _set_attn_impl which writes
+    # _attn_implementation onto the config; PreTrainedConfig's setter walks
+    # `sub_configs` and propagates to nested text_config / sub-configs, so a
+    # shallow copy still mutates those shared inner objects on the cached
+    # config returned by _load_config_for_gpu_estimate. Deepcopy isolates them.
+    config_copy = _copy.deepcopy(config)
+
+    model_class = None
+    for auto_model in (AutoModelForCausalLM, AutoModel):
+        mapping = getattr(auto_model, "_model_mapping", None)
+        if mapping is None:
+            continue
+        try:
+            if config_copy.__class__ in mapping:
+                model_class = mapping[config_copy.__class__]
+                break
+        except Exception:
+            continue
+
+    return resolve_attention_implementation(model_class, config_copy)
+
+
 def _estimate_fp16_model_size_bytes_from_config(config) -> Optional[int]:
     from .vram_estimation import extract_arch_config, compute_total_params
 
@@ -844,12 +872,21 @@ def estimate_fp16_model_size_bytes(
         return int(total_params * 2), "safetensors"
 
     config = _load_config_for_gpu_estimate(estimate_model, hf_token = hf_token)
+    config_bytes: Optional[int] = None
     if config is not None:
         config_bytes = _estimate_fp16_model_size_bytes_from_config(config)
-        if config_bytes is not None:
-            return config_bytes, "config"
 
     local_bytes = _get_local_weight_size_bytes(estimate_model)
+
+    # why: config-derived bytes cover only the text tower; local safetensors
+    # include vision/audio towers. Take the larger so the multimodal
+    # extra_bytes correction can fire.
+    if config_bytes is not None and local_bytes is not None:
+        if local_bytes > config_bytes:
+            return local_bytes, "weight_bytes"
+        return config_bytes, "config"
+    if config_bytes is not None:
+        return config_bytes, "config"
     if local_bytes is not None:
         return local_bytes, "weight_bytes"
 
@@ -877,6 +914,9 @@ def estimate_required_model_memory_gb(
         TrainingVramConfig,
         extract_arch_config,
         estimate_training_vram,
+        compute_total_params,
+        compute_optimizer_bytes,
+        compute_gradient_bytes,
         CUDA_OVERHEAD_BYTES,
         QUANT_4BIT_FACTOR,
         DEFAULT_TARGET_MODULES,
@@ -926,13 +966,44 @@ def estimate_required_model_memory_gb(
         model_name, hf_token = hf_token
     )
     config = _load_config_for_gpu_estimate(estimate_model, hf_token = hf_token)
+    if config is not None:
+        try:
+            vram_config.attention_implementation = (
+                _determine_attention_impl_for_gpu_estimate(config)
+            )
+        except Exception as e:
+            logger.warning(
+                "Could not resolve attention implementation for '%s': %s",
+                estimate_model,
+                e,
+            )
+            # why: if we cannot prove flash attention is usable, charge the
+            # quadratic non-flash activation path so GPU selection stays
+            # conservative.
+            vram_config.attention_implementation = "eager"
     arch = extract_arch_config(config) if config is not None else None
 
     if arch is not None:
         breakdown = estimate_training_vram(arch, vram_config)
+        # why: extract_arch_config only sees text_config; safetensors include
+        # vision/audio tower bytes that the text-arch fp16 total misses.
+        arch_fp16_bytes = compute_total_params(arch) * 2
+        extra_bytes = max(0, int(model_size_bytes) - arch_fp16_bytes)
+        if extra_bytes > 0:
+            breakdown.model_weights += extra_bytes
+            if training_method == "full":
+                # why: full fine-tuning makes the extra (vision/audio) params
+                # trainable; optimizer + gradient bytes scale with them too.
+                extra_params = extra_bytes // 2
+                breakdown.optimizer_states += compute_optimizer_bytes(
+                    extra_params,
+                    vram_config.optimizer,
+                )
+                breakdown.gradients += compute_gradient_bytes(extra_params)
         required_gb = breakdown.total / (1024**3)
         metadata["required_gb"] = round(required_gb, 3)
         metadata["estimation_mode"] = "detailed"
+        metadata["attention_implementation"] = vram_config.attention_implementation
         metadata["vram_breakdown"] = breakdown.to_gb_dict()
         max_gpus = max(1, get_visible_gpu_count())
         for n_gpus in range(1, max_gpus + 1):
diff --git a/studio/backend/utils/hardware/vram_estimation.py b/studio/backend/utils/hardware/vram_estimation.py
index e03665374d..ba1b1dfe61 100644
--- a/studio/backend/utils/hardware/vram_estimation.py
+++ b/studio/backend/utils/hardware/vram_estimation.py
@@ -16,7 +16,26 @@ from dataclasses import dataclass, field
 from typing import Dict, Optional
 
 QUANT_4BIT_FACTOR = 16 / 5
+DOUBLE_QUANT_4BIT_FACTOR = (
+    3.6  # bnb_4bit_use_double_quant; see VRAM_ESTIMATION.md section 1
+)
 CUDA_OVERHEAD_BYTES = int(1.4 * 1024**3)  # calibrated on RTX 5070 Ti
+NON_FLASH_ATTENTION_FACTOR = (
+    12.0  # eager attention score+workspace overhead; see VRAM_ESTIMATION.md section 5
+)
+
+LINEAR_ATTENTION_IMPLS = frozenset({"flash_attention_2", "sdpa", "flex_attention"})
+
+_SKIP_MODULE_TEXT_PREFIXES = frozenset(
+    {
+        "model",
+        "model.model",
+        "language_model",
+        "language_model.model",
+        "model.language_model",
+        "model.language_model.model",
+    }
+)
 
 DEFAULT_TARGET_MODULES = [
     "q_proj",
@@ -27,6 +46,8 @@ DEFAULT_TARGET_MODULES = [
     "up_proj",
     "down_proj",
 ]
+ATTENTION_TARGET_MODULES = {"q_proj", "k_proj", "v_proj", "o_proj"}
+MLP_TARGET_MODULES = {"gate_proj", "up_proj", "down_proj"}
 
 # Empirically calibrated bytes/param — see VRAM_ESTIMATION.md for rationale.
 OPTIMIZER_BYTES_PER_PARAM: Dict[str, int] = {
@@ -61,12 +82,28 @@ class ModelArchConfig:
     num_experts: Optional[int] = None
     moe_intermediate_size: Optional[int] = None
     n_shared_experts: int = 0
+    shared_expert_intermediate_size: Optional[int] = None
+    num_experts_per_tok: int = 1
     num_dense_layers: int = 0
     q_lora_rank: Optional[int] = None
     kv_lora_rank: Optional[int] = None
     qk_nope_head_dim: Optional[int] = None
     qk_rope_head_dim: Optional[int] = None
     v_head_dim: Optional[int] = None
+    head_dim: Optional[int] = None
+    global_head_dim: Optional[int] = None
+    num_global_key_value_heads: Optional[int] = None
+    attention_k_eq_v: bool = False
+    layer_types: Optional[list] = None
+    num_kv_shared_layers: int = 0
+    use_double_wide_mlp: bool = False
+    vocab_size_per_layer_input: int = 0
+    hidden_size_per_layer_input: int = 0
+    quantization_skip_modules: list = field(default_factory = list)
+    quant_4bit_factor: float = QUANT_4BIT_FACTOR
+    moe_has_dense_mlp: bool = False
+    dense_layer_indices: tuple = ()
+    dense_intermediate_size: Optional[int] = None
 
 
 @dataclass
@@ -79,6 +116,7 @@ class TrainingVramConfig:
     gradient_checkpointing: str = "unsloth"
     optimizer: str = "adamw_8bit"
     load_in_4bit: bool = True
+    attention_implementation: str = "flash_attention_2"
 
 
 @dataclass
@@ -89,8 +127,8 @@ class VramBreakdown:
     gradients: int
     activations: int
     cuda_overhead: int
-    # The computed (formula-based) activation cost before floors.
-    # This is the true per-layer cost that doesn't shard across GPUs.
+    # Equals `activations`; retained for backward compatibility with
+    # consumers that read this field.
     activations_computed: int = 0
 
     @property
@@ -108,17 +146,15 @@ class VramBreakdown:
         """Minimum VRAM a single GPU needs: its shard + non-shardable costs.
 
         Weights/LoRA/optimizer/gradients shard across GPUs.
-        The computed activation cost does NOT shard (one GPU runs the layer).
-        The floor portion (activations - computed) is overhead that shards.
+        Activations do NOT shard (the GPU running a layer holds them).
         """
         shardable = (
             self.model_weights
             + self.lora_adapters
             + self.optimizer_states
             + self.gradients
-            + (self.activations - self.activations_computed)  # floor overhead shards
         )
-        per_gpu_fixed = self.activations_computed + self.cuda_overhead
+        per_gpu_fixed = self.activations + self.cuda_overhead
         return shardable // max(n_gpus, 1) + per_gpu_fixed
 
     def to_gb_dict(self) -> Dict[str, float]:
@@ -133,28 +169,88 @@ class VramBreakdown:
         }
 
 
-def _compute_num_dense_layers(text_config, total_layers: int) -> int:
-    """Count how many layers use dense MLP instead of MoE."""
+def _first_scalar(value):
+    # why: ERNIE MoE configs ship moe_intermediate_size / moe_num_experts as
+    # [routed, shared] lists; downstream arithmetic needs the routed scalar.
+    if isinstance(value, (list, tuple)):
+        return value[0] if value else None
+    return value
+
+
+def _max_scalar(value):
+    # why: Hunyuan-V1-MoE moe_topk can be a per-layer list; activation
+    # accounting uses the max top-k as a conservative upper bound.
+    if isinstance(value, (list, tuple)):
+        items = [v for v in value if v is not None]
+        return max(items) if items else None
+    return value
+
+
+def _compute_dense_layer_indices(text_config, total_layers: int) -> tuple:
+    """Layer indices that use dense MLP instead of MoE. Position matters."""
+    # why: transformers Exaone-MoE / Laguna / Hy_v3 / GLM-MoE-DSA / GLM4-MoE-Lite /
+    # Ernie4_5_VL_MoE prefer per-position `mlp_layer_types` over the prefix-style
+    # `first_k_dense_replace` and may omit `decoder_sparse_step` entirely.
+    layer_types = getattr(text_config, "mlp_layer_types", None)
+    if layer_types:
+        return tuple(
+            i
+            for i, t in enumerate(layer_types[:total_layers])
+            if str(t).lower() == "dense"
+        )
+
+    # why: Llama4TextConfig.__init__ auto-populates self.moe_layers from
+    # interleave_moe_layer_step; Llama4TextDecoderLayer dispatches via
+    # `layer_idx in config.moe_layers` (modeling_llama4.py).
+    llama4_moe_layers = getattr(text_config, "moe_layers", None)
+    if llama4_moe_layers is not None:
+        moe_indices = {int(i) for i in llama4_moe_layers}
+        return tuple(i for i in range(total_layers) if i not in moe_indices)
+
+    # why: transformers ERNIE 4.5 MoE / ERNIE 4.5 VL MoE declare MoE layers
+    # via moe_layer_start_index / moe_layer_end_index / moe_layer_interval;
+    # the model's per-layer guard is `(layer_idx + 1) % interval == 0` with
+    # start <= layer_idx <= end (modeling_ernie4_5_moe.py).
+    moe_start = getattr(text_config, "moe_layer_start_index", None)
+    moe_interval = getattr(text_config, "moe_layer_interval", None)
+    if moe_start is not None and moe_interval is not None and int(moe_interval) > 0:
+        moe_end_raw = getattr(text_config, "moe_layer_end_index", None)
+        end = (
+            total_layers
+            if moe_end_raw is None or int(moe_end_raw) == -1
+            else min(int(moe_end_raw) + 1, total_layers)
+        )
+        start = max(0, int(moe_start))
+        interval = int(moe_interval)
+        moe_indices = {i for i in range(start, end) if (i + 1) % interval == 0}
+        return tuple(i for i in range(total_layers) if i not in moe_indices)
+
     first_k = getattr(text_config, "first_k_dense_replace", None)
     if first_k is not None:
-        return min(int(first_k), total_layers)
+        return tuple(range(min(int(first_k), total_layers)))
 
     sparse_step = getattr(text_config, "decoder_sparse_step", None)
     mlp_only = getattr(text_config, "mlp_only_layers", None) or []
     if sparse_step is not None and sparse_step > 0:
-        mlp_only_set = set(mlp_only)
-        moe_count = sum(
-            1
+        mlp_only_set = {int(i) for i in mlp_only}
+        return tuple(
+            i
             for i in range(total_layers)
-            if i not in mlp_only_set and (i + 1) % sparse_step == 0
+            if i in mlp_only_set or (i + 1) % sparse_step != 0
         )
-        return total_layers - moe_count
-
-    return 0
+    return ()
 
 
 def extract_arch_config(hf_config) -> Optional[ModelArchConfig]:
     text_config = getattr(hf_config, "text_config", None) or hf_config
+    quantization_config = getattr(hf_config, "quantization_config", None) or {}
+    if not isinstance(quantization_config, dict):
+        quantization_config = getattr(quantization_config, "to_dict", lambda: {})()
+    quant_4bit_factor = (
+        DOUBLE_QUANT_4BIT_FACTOR
+        if quantization_config.get("bnb_4bit_use_double_quant", False)
+        else QUANT_4BIT_FACTOR
+    )
 
     hidden_size = getattr(text_config, "hidden_size", None)
     num_layers = getattr(text_config, "num_hidden_layers", None)
@@ -177,18 +273,75 @@ def extract_arch_config(hf_config) -> Optional[ModelArchConfig]:
 
     num_kv_heads = getattr(text_config, "num_key_value_heads", num_heads)
 
+    # why: DBRX places its MoE attrs on the DbrxFFNConfig sub-config; probe
+    # ffn_config as a secondary source so DBRX is not misclassified as dense.
+    ffn_config = getattr(text_config, "ffn_config", None)
+
+    def _moe_attr(name):
+        value = getattr(text_config, name, None)
+        if value is None and ffn_config is not None:
+            value = getattr(ffn_config, name, None)
+        return value
+
     num_experts = None
-    for attr in ("num_local_experts", "num_experts", "n_routed_experts"):
-        num_experts = getattr(text_config, attr, None)
+    for attr in (
+        "num_local_experts",
+        "num_experts",
+        "n_routed_experts",
+        "moe_num_experts",
+    ):
+        num_experts = _first_scalar(_moe_attr(attr))
         if num_experts is not None:
             break
 
-    moe_intermediate = getattr(text_config, "moe_intermediate_size", None)
-    n_shared_experts = getattr(text_config, "n_shared_experts", None) or 0
+    moe_intermediate_raw = _moe_attr("moe_intermediate_size")
+    if moe_intermediate_raw is None:
+        moe_intermediate_raw = _moe_attr("ffn_hidden_size")
+    moe_intermediate = _first_scalar(moe_intermediate_raw)
+    # why: Exaone-MoE / ERNIE families alias num_shared_experts /
+    # moe_num_shared_experts to the canonical n_shared_experts.
+    n_shared_experts = (
+        _first_scalar(_moe_attr("n_shared_experts"))
+        or _first_scalar(_moe_attr("num_shared_experts"))
+        or _first_scalar(_moe_attr("moe_num_shared_experts"))
+        or 0
+    )
+    shared_expert_intermediate_size = _moe_attr("shared_expert_intermediate_size")
+    if shared_expert_intermediate_size and n_shared_experts == 0:
+        n_shared_experts = 1
+    # why: DBRX exposes moe_top_k, Hunyuan-V1-MoE exposes moe_topk (which can
+    # be a per-layer list); _max_scalar normalizes list values to the worst
+    # case so int(...) below cannot crash on the canonical attribute_map path.
+    num_experts_per_tok = (
+        _max_scalar(_moe_attr("num_experts_per_tok"))
+        or _max_scalar(_moe_attr("top_k_experts"))
+        or _max_scalar(_moe_attr("moe_top_k"))
+        or _max_scalar(_moe_attr("moe_topk"))
+        or 1
+    )
 
-    num_dense_layers = 0
+    dense_layer_indices: tuple = ()
     if num_experts is not None and num_experts > 1:
-        num_dense_layers = _compute_num_dense_layers(text_config, num_layers)
+        dense_layer_indices = _compute_dense_layer_indices(text_config, num_layers)
+    num_dense_layers = len(dense_layer_indices)
+
+    # why: Llama4 dense layers use intermediate_size_mlp; routed and shared
+    # experts use intermediate_size. Llama4TextMoe builds one shared_expert
+    # per MoE layer (modeling_llama4.py).
+    intermediate_size_mlp_raw = _first_scalar(_moe_attr("intermediate_size_mlp"))
+    dense_intermediate_size = (
+        int(intermediate_size_mlp_raw)
+        if intermediate_size_mlp_raw is not None
+        else None
+    )
+    if (
+        intermediate_size_mlp_raw is not None
+        and num_experts is not None
+        and num_experts > 1
+        and shared_expert_intermediate_size is None
+        and n_shared_experts == 0
+    ):
+        n_shared_experts = 1
 
     q_lora_rank = getattr(text_config, "q_lora_rank", None)
     kv_lora_rank = getattr(text_config, "kv_lora_rank", None)
@@ -207,15 +360,418 @@ def extract_arch_config(hf_config) -> Optional[ModelArchConfig]:
         num_experts = num_experts,
         moe_intermediate_size = moe_intermediate,
         n_shared_experts = n_shared_experts,
+        shared_expert_intermediate_size = shared_expert_intermediate_size,
+        num_experts_per_tok = int(num_experts_per_tok),
         num_dense_layers = num_dense_layers,
         q_lora_rank = q_lora_rank,
         kv_lora_rank = kv_lora_rank,
         qk_nope_head_dim = qk_nope_head_dim,
         qk_rope_head_dim = qk_rope_head_dim,
         v_head_dim = v_head_dim,
+        head_dim = getattr(text_config, "head_dim", None),
+        global_head_dim = getattr(text_config, "global_head_dim", None),
+        num_global_key_value_heads = getattr(
+            text_config,
+            "num_global_key_value_heads",
+            None,
+        ),
+        attention_k_eq_v = bool(getattr(text_config, "attention_k_eq_v", False)),
+        layer_types = getattr(text_config, "layer_types", None),
+        num_kv_shared_layers = getattr(text_config, "num_kv_shared_layers", None) or 0,
+        use_double_wide_mlp = bool(getattr(text_config, "use_double_wide_mlp", False)),
+        vocab_size_per_layer_input = getattr(
+            text_config,
+            "vocab_size_per_layer_input",
+            None,
+        )
+        or 0,
+        hidden_size_per_layer_input = getattr(
+            text_config,
+            "hidden_size_per_layer_input",
+            None,
+        )
+        or 0,
+        quantization_skip_modules = list(
+            quantization_config.get("llm_int8_skip_modules", []) or []
+        ),
+        quant_4bit_factor = quant_4bit_factor,
+        moe_has_dense_mlp = bool(getattr(text_config, "enable_moe_block", False)),
+        dense_layer_indices = dense_layer_indices,
+        dense_intermediate_size = dense_intermediate_size,
     )
 
 
+def _targets_all_linear(target_modules) -> bool:
+    # why: peft LoraConfig accepts target_modules="all-linear" as a bare
+    # string; iterating a string yields chars and never matches the set.
+    if isinstance(target_modules, str):
+        target_modules = [target_modules]
+    normalized = {str(module).lower().replace("_", "-") for module in target_modules}
+    return normalized == {"all-linear"}
+
+
+def _head_dim(arch: ModelArchConfig) -> int:
+    return arch.head_dim or arch.hidden_size // arch.num_attention_heads
+
+
+def _layer_types(arch: ModelArchConfig) -> list:
+    if arch.layer_types and len(arch.layer_types) == arch.num_hidden_layers:
+        return arch.layer_types
+    return ["full_attention"] * arch.num_hidden_layers
+
+
+def _uses_structured_layer_shapes(arch: ModelArchConfig) -> bool:
+    # MLA configs have their own q/kv low-rank projection shape formulas in
+    # _compute_attn_elements / _lora_attn_elements; do not let head_dim or
+    # other structured fields override that path.
+    if arch.q_lora_rank is not None:
+        return False
+    return bool(
+        arch.layer_types
+        or arch.head_dim is not None
+        or arch.global_head_dim is not None
+        or arch.num_global_key_value_heads is not None
+        or arch.attention_k_eq_v
+        or arch.num_kv_shared_layers > 0
+        or arch.use_double_wide_mlp
+    )
+
+
+def _is_kv_shared_layer(arch: ModelArchConfig, layer_idx: int) -> bool:
+    if arch.num_kv_shared_layers <= 0:
+        return False
+    first_shared = arch.num_hidden_layers - arch.num_kv_shared_layers
+    # why: transformers Gemma4 (modeling_gemma4.py:1031, modular_gemma4.py:863)
+    # uses the same `> 0` guard so a fully-shared config raises during model
+    # construction; matching upstream avoids producing a detailed estimate
+    # for a shape the actual model code rejects.
+    return layer_idx >= first_shared > 0
+
+
+def _is_dense_mlp_layer(arch: ModelArchConfig, layer_idx: int) -> bool:
+    if arch.dense_layer_indices:
+        return layer_idx in arch.dense_layer_indices
+    return layer_idx < arch.num_dense_layers
+
+
+def _per_layer_input_quantizable(arch: ModelArchConfig) -> int:
+    # why: Gemma4 PLE block adds per_layer_model_projection (single Linear),
+    # per_layer_input_gate (per layer), and per_layer_projection (per layer);
+    # see transformers gemma4/modular_gemma4.py:1077-1083 and :1247-1253.
+    pli = arch.hidden_size_per_layer_input
+    if pli <= 0:
+        return 0
+    n_layers = arch.num_hidden_layers
+    hd = arch.hidden_size
+    return hd * (n_layers * pli) + (hd * pli) * n_layers + (pli * hd) * n_layers
+
+
+def _per_layer_input_norm_elements(arch: ModelArchConfig) -> int:
+    pli = arch.hidden_size_per_layer_input
+    if pli <= 0:
+        return 0
+    n_layers = arch.num_hidden_layers
+    hd = arch.hidden_size
+    return hd * n_layers + pli
+
+
+def _per_layer_input_lora_params(
+    arch: ModelArchConfig,
+    r: int,
+    target_modules,
+) -> int:
+    # why: Unsloth's get_peft_regex (unsloth_zoo/peft_utils.py) requires module
+    # names to contain a component tag (mlp/attn/...); PLE module names lack
+    # any tag, so all-linear training does NOT attach LoRA to them. Only count
+    # PLE LoRA when the user explicitly names PLE modules.
+    pli = arch.hidden_size_per_layer_input
+    if pli <= 0:
+        return 0
+    targets = (
+        {target_modules}
+        if isinstance(target_modules, str)
+        else set(target_modules or [])
+    )
+    n_layers = arch.num_hidden_layers
+    hd = arch.hidden_size
+    total = 0
+    if "per_layer_model_projection" in targets:
+        total += hd * r + r * (n_layers * pli)
+    if "per_layer_input_gate" in targets:
+        total += (hd * r + r * pli) * n_layers
+    if "per_layer_projection" in targets:
+        total += (pli * r + r * hd) * n_layers
+    return total
+
+
+def _layer_attention_dims(arch: ModelArchConfig, layer_idx: int) -> tuple:
+    layer_types = _layer_types(arch)
+    layer_type = layer_types[layer_idx]
+    is_sliding = layer_type == "sliding_attention"
+    head_dim = (
+        arch.global_head_dim
+        if not is_sliding and arch.global_head_dim
+        else _head_dim(arch)
+    )
+    use_alt_attention = arch.attention_k_eq_v and not is_sliding
+    num_kv_heads = (
+        arch.num_global_key_value_heads
+        if use_alt_attention and arch.num_global_key_value_heads
+        else arch.num_key_value_heads
+    )
+    q_size = arch.num_attention_heads * head_dim
+    kv_size = num_kv_heads * head_dim
+    has_k = not _is_kv_shared_layer(arch, layer_idx)
+    has_v = has_k and not use_alt_attention
+    return q_size, kv_size, has_k, has_v
+
+
+def _layer_mlp_size(arch: ModelArchConfig, layer_idx: int) -> int:
+    if arch.use_double_wide_mlp and _is_kv_shared_layer(arch, layer_idx):
+        return _dense_mlp_size(arch) * 2
+    return _dense_mlp_size(arch)
+
+
+def _text_linear_dims(
+    arch: ModelArchConfig,
+    layer_idx: int,
+) -> Dict[str, tuple[int, int]]:
+    hd = arch.hidden_size
+    if _uses_structured_layer_shapes(arch):
+        q_size, kv_size, has_k, has_v = _layer_attention_dims(arch, layer_idx)
+        mlp_size = _layer_mlp_size(arch, layer_idx)
+    else:
+        q_size = hd
+        kv_size = _get_kv_size(arch)
+        has_k = True
+        has_v = True
+        mlp_size = _get_mlp_size(arch)
+
+    dims = {
+        "q_proj": (hd, q_size),
+        "o_proj": (q_size, hd),
+    }
+    if has_k:
+        dims["k_proj"] = (hd, kv_size)
+    if has_v:
+        dims["v_proj"] = (hd, kv_size)
+
+    dims.update(
+        {
+            "gate_proj": (hd, mlp_size),
+            "up_proj": (hd, mlp_size),
+            "down_proj": (mlp_size, hd),
+        }
+    )
+    return dims
+
+
+def _module_path_matches(skip_module: str, alias: str) -> bool:
+    skip_parts = [part for part in skip_module.split(".") if part]
+    alias_parts = [part for part in alias.split(".") if part]
+    if not skip_parts or not alias_parts:
+        return False
+    if alias_parts[0] == "layers":
+        return skip_parts == alias_parts
+    if len(skip_parts) <= len(alias_parts):
+        # why: transformers BNB quantizer suffix-matches short skip entries
+        # like ["q_proj"] / ["lm_head"] against full module paths, so a skip
+        # shorter than the alias is a tail match.
+        return alias_parts[-len(skip_parts) :] == skip_parts
+    if skip_parts[-len(alias_parts) :] != alias_parts:
+        return False
+    prefix_parts = skip_parts[: len(skip_parts) - len(alias_parts)]
+    if not prefix_parts:
+        return True
+    # why: bound the prefix to known text-tower roots so VLM skip names like
+    # vision_tower.model.layers.<i>.self_attn.q_proj do not shadow the text
+    # alias model.layers.<i>.self_attn.q_proj.
+    return ".".join(prefix_parts) in _SKIP_MODULE_TEXT_PREFIXES
+
+
+def _add_module_aliases(
+    aliases: Dict[str, str],
+    canonical: str,
+    suffix: str,
+) -> None:
+    for prefix in (
+        "",
+        "model",
+        "model.model",
+        "language_model",
+        "language_model.model",
+        "model.language_model",
+        "model.language_model.model",
+    ):
+        alias = f"{prefix}.{suffix}" if prefix else suffix
+        aliases[alias] = canonical
+
+
+def _build_text_module_elements(
+    arch: ModelArchConfig,
+) -> tuple[Dict[str, int], Dict[str, str]]:
+    elements: Dict[str, int] = {}
+    aliases: Dict[str, str] = {}
+
+    is_mla = arch.q_lora_rank is not None and not _uses_structured_layer_shapes(arch)
+    pli = arch.hidden_size_per_layer_input
+    hd_global = arch.hidden_size
+
+    for layer_idx in range(arch.num_hidden_layers):
+        layer_modules: Dict[str, int] = {}
+        dims = _text_linear_dims(arch, layer_idx)
+        attn_dims = {
+            name: dim for name, dim in dims.items() if name in ATTENTION_TARGET_MODULES
+        }
+        mlp_dims = {
+            name: dim for name, dim in dims.items() if name in MLP_TARGET_MODULES
+        }
+
+        if is_mla:
+            # why: _text_linear_dims uses (hd, hd) for q/o; MLA actually splits
+            # into q_a/q_b/kv_a/kv_b, so emit a single self_attn aggregate at
+            # the authoritative MLA per-layer total.
+            layer_modules["self_attn"] = _compute_attn_elements(arch)
+        else:
+            for name, (in_dim, out_dim) in attn_dims.items():
+                layer_modules[f"self_attn.{name}"] = in_dim * out_dim
+
+        if arch.num_experts and arch.num_experts > 1:
+            if _is_dense_mlp_layer(arch, layer_idx):
+                layer_modules.update(
+                    {
+                        f"mlp.{name}": in_dim * out_dim
+                        for name, (in_dim, out_dim) in mlp_dims.items()
+                    }
+                )
+            else:
+                layer_modules["mlp.experts"] = _compute_routed_moe_elements(arch)
+                shared_moe = _compute_shared_moe_elements(arch)
+                if shared_moe:
+                    # why: Qwen3.5-MoE exposes shared expert as
+                    # mlp.shared_expert; Exaone-MoE/Laguna/GLM-style configs use
+                    # mlp.shared_experts. Register both names so child-path
+                    # llm_int8_skip_modules entries match the right shared block.
+                    layer_modules["mlp.shared_expert"] = shared_moe
+                if arch.moe_has_dense_mlp:
+                    # why: enable_moe_block runs the dense MLP and the MoE
+                    # experts in parallel; register both for skip matching.
+                    # Non-structured _text_linear_dims returns mlp_size from
+                    # _get_mlp_size which prefers moe_intermediate_size, so
+                    # rebuild dense dims from arch.intermediate_size directly.
+                    if _uses_structured_layer_shapes(arch):
+                        dense_dims = mlp_dims
+                    else:
+                        hd = arch.hidden_size
+                        inter = arch.intermediate_size
+                        dense_dims = {
+                            "gate_proj": (hd, inter),
+                            "up_proj": (hd, inter),
+                            "down_proj": (inter, hd),
+                        }
+                    layer_modules.update(
+                        {
+                            f"mlp.{name}": in_dim * out_dim
+                            for name, (in_dim, out_dim) in dense_dims.items()
+                        }
+                    )
+        else:
+            layer_modules.update(
+                {
+                    f"mlp.{name}": in_dim * out_dim
+                    for name, (in_dim, out_dim) in mlp_dims.items()
+                }
+            )
+
+        if pli > 0:
+            # why: register PLE per-layer linears so llm_int8_skip_modules
+            # entries like model.layers.0.per_layer_input_gate match.
+            layer_modules["per_layer_input_gate"] = hd_global * pli
+            layer_modules["per_layer_projection"] = pli * hd_global
+
+        attn_total = sum(
+            value
+            for name, value in layer_modules.items()
+            if name == "self_attn" or name.startswith("self_attn.")
+        )
+        # why: gemma4 enable_moe_block puts routed experts at the sibling
+        # layers.<i>.experts attribute, not under self.mlp; the layer's "mlp"
+        # aggregate must reflect only the dense MLP path so a skip module
+        # `model.layers.0.mlp` does not over-skip into the experts block.
+        is_sibling_experts = bool(arch.moe_has_dense_mlp)
+        mlp_total = sum(
+            value
+            for name, value in layer_modules.items()
+            if (
+                name == "mlp"
+                or (
+                    name.startswith("mlp.")
+                    and not (is_sibling_experts and name == "mlp.experts")
+                )
+            )
+        )
+        experts_total = layer_modules.get("mlp.experts", 0) if is_sibling_experts else 0
+        layer_total = sum(layer_modules.values())
+
+        aggregate_modules = {
+            f"text.layers.{layer_idx}": layer_total,
+            f"text.layers.{layer_idx}.self_attn": attn_total,
+            f"text.layers.{layer_idx}.mlp": mlp_total,
+        }
+        if experts_total:
+            aggregate_modules[f"text.layers.{layer_idx}.experts"] = experts_total
+        elements.update(aggregate_modules)
+        for canonical in aggregate_modules:
+            suffix = canonical.removeprefix("text.")
+            _add_module_aliases(aliases, canonical, suffix)
+
+        for name, value in layer_modules.items():
+            canonical = f"text.layers.{layer_idx}.{name}"
+            elements[canonical] = value
+            _add_module_aliases(aliases, canonical, canonical.removeprefix("text."))
+            if name == "mlp.experts" and arch.moe_has_dense_mlp:
+                # why: gemma4 enable_moe_block exposes routed experts at
+                # layers.<i>.experts (sibling of self.mlp), not under mlp.
+                _add_module_aliases(aliases, canonical, f"layers.{layer_idx}.experts")
+            elif name == "mlp.shared_expert":
+                # why: Exaone-MoE / Laguna / GLM-style configs use the plural
+                # `shared_experts` attribute name; register both spellings.
+                _add_module_aliases(
+                    aliases,
+                    canonical,
+                    f"layers.{layer_idx}.mlp.shared_experts",
+                )
+
+    if pli > 0:
+        canonical = "text.per_layer_model_projection"
+        elements[canonical] = hd_global * (arch.num_hidden_layers * pli)
+        _add_module_aliases(aliases, canonical, canonical.removeprefix("text."))
+
+    return elements, aliases
+
+
+def _compute_skipped_quantizable_elements(arch: ModelArchConfig) -> int:
+    if not arch.quantization_skip_modules:
+        return 0
+
+    module_elements, aliases = _build_text_module_elements(arch)
+    matched = set()
+    for skip_module in arch.quantization_skip_modules:
+        for alias, canonical in aliases.items():
+            if _module_path_matches(skip_module, alias):
+                matched.add(canonical)
+
+    pruned = {
+        canonical
+        for canonical in matched
+        if not any(
+            canonical != parent and canonical.startswith(f"{parent}.")
+            for parent in matched
+        )
+    }
+    return sum(module_elements[canonical] for canonical in pruned)
+
+
 def _get_kv_size(arch: ModelArchConfig) -> int:
     return (arch.hidden_size // arch.num_attention_heads) * arch.num_key_value_heads
 
@@ -226,6 +782,12 @@ def _get_mlp_size(arch: ModelArchConfig) -> int:
     return arch.intermediate_size
 
 
+def _dense_mlp_size(arch: ModelArchConfig) -> int:
+    # why: Llama4 dense layers use intermediate_size_mlp; routed/shared
+    # experts use intermediate_size. Other configs leave the field None.
+    return arch.dense_intermediate_size or arch.intermediate_size
+
+
 def _get_num_experts(arch: ModelArchConfig) -> int:
     return arch.num_experts if arch.num_experts and arch.num_experts > 1 else 1
 
@@ -248,14 +810,39 @@ def _compute_attn_elements(arch: ModelArchConfig) -> int:
 
 
 def _compute_dense_mlp_elements(arch: ModelArchConfig) -> int:
-    return arch.hidden_size * arch.intermediate_size * 3
+    return arch.hidden_size * _dense_mlp_size(arch) * 3
+
+
+def _shared_expert_size(arch: ModelArchConfig) -> int:
+    # why: Qwen3.5-MoE shared expert has its own intermediate_size (default 512)
+    # distinct from moe_intermediate_size; fall back to routed mlp_size for
+    # families that share it (deepseek-style configs).
+    return arch.shared_expert_intermediate_size or _get_mlp_size(arch)
+
+
+def _compute_routed_moe_elements(arch: ModelArchConfig) -> int:
+    hd = arch.hidden_size
+    n_experts = _get_num_experts(arch)
+    return hd * _get_mlp_size(arch) * 3 * n_experts + n_experts * hd
+
+
+def _compute_shared_moe_elements(arch: ModelArchConfig) -> int:
+    if not arch.n_shared_experts:
+        return 0
+    hd = arch.hidden_size
+    shared_size = _shared_expert_size(arch)
+    total = hd * shared_size * 3 * arch.n_shared_experts
+    # why: only Qwen2-MoE / Qwen3.5-MoE define a shared_expert_gate Linear
+    # (hidden_size→1); other families (Exaone-MoE, HY-V3, GLM4-MoE-Lite, Laguna)
+    # have shared_experts without a gate. shared_expert_intermediate_size is the
+    # Qwen-style discriminator.
+    if arch.shared_expert_intermediate_size:
+        total += arch.n_shared_experts * hd
+    return total
 
 
 def _compute_moe_mlp_elements(arch: ModelArchConfig) -> int:
-    hd = arch.hidden_size
-    mlp_size = _get_mlp_size(arch)
-    n_experts = _get_num_experts(arch)
-    return hd * mlp_size * 3 * (n_experts + arch.n_shared_experts) + n_experts * hd
+    return _compute_routed_moe_elements(arch) + _compute_shared_moe_elements(arch)
 
 
 def _compute_layer_elements(arch: ModelArchConfig):
@@ -267,22 +854,60 @@ def _compute_layer_elements(arch: ModelArchConfig):
     n_layers = arch.num_hidden_layers
     n_experts = _get_num_experts(arch)
 
-    attn_total = _compute_attn_elements(arch) * n_layers
-
-    if n_experts > 1:
+    if _uses_structured_layer_shapes(arch):
+        attn_total = 0
+        per_layer_dense_mlp = []
+        for layer_idx in range(n_layers):
+            layer_dense_mlp = 0
+            for name, (in_dim, out_dim) in _text_linear_dims(
+                arch,
+                layer_idx,
+            ).items():
+                elements = in_dim * out_dim
+                if name in ATTENTION_TARGET_MODULES:
+                    attn_total += elements
+                elif name in MLP_TARGET_MODULES:
+                    layer_dense_mlp += elements
+            per_layer_dense_mlp.append(layer_dense_mlp)
+        if n_experts > 1:
+            n_dense = arch.num_dense_layers
+            n_moe = n_layers - n_dense
+            moe_mlp_total = _compute_moe_mlp_elements(arch) * n_moe
+            if arch.moe_has_dense_mlp:
+                # why: enable_moe_block runs dense MLP and MoE experts in
+                # parallel; count dense for every layer alongside MoE.
+                mlp_total = sum(per_layer_dense_mlp) + moe_mlp_total
+            else:
+                dense_only_total = sum(
+                    value
+                    for i, value in enumerate(per_layer_dense_mlp)
+                    if _is_dense_mlp_layer(arch, i)
+                )
+                mlp_total = moe_mlp_total + dense_only_total
+        else:
+            mlp_total = sum(per_layer_dense_mlp)
+    elif n_experts > 1:
+        attn_total = _compute_attn_elements(arch) * n_layers
         n_dense = arch.num_dense_layers
         n_moe = n_layers - n_dense
-        mlp_total = (
-            _compute_moe_mlp_elements(arch) * n_moe
-            + _compute_dense_mlp_elements(arch) * n_dense
-        )
+        moe_mlp_total = _compute_moe_mlp_elements(arch) * n_moe
+        if arch.moe_has_dense_mlp:
+            mlp_total = _compute_dense_mlp_elements(arch) * n_layers + moe_mlp_total
+        else:
+            mlp_total = moe_mlp_total + _compute_dense_mlp_elements(arch) * n_dense
     else:
+        attn_total = _compute_attn_elements(arch) * n_layers
         mlp_total = _compute_dense_mlp_elements(arch) * n_layers
 
     layernorms = 2 * hd
-    embed_tokens = arch.vocab_size * hd
+    per_layer_embed = (
+        arch.vocab_size_per_layer_input * arch.hidden_size_per_layer_input * n_layers
+    )
+    ple_text_linear = _per_layer_input_quantizable(arch)
+    ple_norms = _per_layer_input_norm_elements(arch)
+    embed_tokens = arch.vocab_size * hd + per_layer_embed + ple_norms
     lm_head = 0 if arch.tie_word_embeddings else arch.vocab_size * hd
-    return attn_total + mlp_total, layernorms, embed_tokens, lm_head
+    return attn_total + mlp_total + ple_text_linear, layernorms, embed_tokens, lm_head
 
 
 def compute_model_weights_bytes(
@@ -295,7 +920,16 @@ def compute_model_weights_bytes(
     non_quantizable = layernorms * n_layers + embed_tokens + lm_head
 
     if training_method == "qlora" and load_in_4bit:
-        return int(total_quantizable * 2 / QUANT_4BIT_FACTOR + non_quantizable * 2)
+        skipped_quantizable = min(
+            _compute_skipped_quantizable_elements(arch),
+            total_quantizable,
+        )
+        quantized = total_quantizable - skipped_quantizable
+        return int(
+            quantized * 2 / arch.quant_4bit_factor
+            + skipped_quantizable * 2
+            + non_quantizable * 2
+        )
 
     return int((total_quantizable + non_quantizable) * 2)
 
@@ -363,46 +997,130 @@ def compute_lora_params(
     lora_rank: int,
     target_modules: list,
 ) -> int:
+    all_linear = _targets_all_linear(target_modules)
+    selected_modules = list(DEFAULT_TARGET_MODULES) if all_linear else target_modules
     hd = arch.hidden_size
     r = lora_rank
     n_layers = arch.num_hidden_layers
     n_experts = _get_num_experts(arch)
 
-    attn_total = _lora_attn_elements(arch, r, target_modules) * n_layers
-
-    if n_experts > 1:
+    use_structured_shapes = _uses_structured_layer_shapes(arch)
+    if use_structured_shapes:
+        attn_total = 0
+        structured_dense_mlp = 0
+        per_layer_dense_mlp = []
+        for layer_idx in range(n_layers):
+            layer_dense = 0
+            for name, (in_dim, out_dim) in _text_linear_dims(
+                arch,
+                layer_idx,
+            ).items():
+                if name not in selected_modules:
+                    continue
+                if name in ATTENTION_TARGET_MODULES:
+                    attn_total += in_dim * r + r * out_dim
+                elif name in MLP_TARGET_MODULES:
+                    layer_dense += in_dim * r + r * out_dim
+            per_layer_dense_mlp.append(layer_dense)
+            structured_dense_mlp += layer_dense
+        if n_experts > 1:
+            n_dense = arch.num_dense_layers
+            n_moe = n_layers - n_dense
+            # why: peft "all-linear" attaches LoRA to nn.Linear only;
+            # routed experts are nn.Parameter and need explicit
+            # gate_proj/up_proj/down_proj naming via Unsloth's
+            # get_moe_target_parameters. Shared experts are nn.Linear and
+            # are picked up by get_peft_regex.
+            routed_moe = (
+                0
+                if all_linear
+                else _lora_mlp_elements(
+                    hd,
+                    _get_mlp_size(arch),
+                    r,
+                    selected_modules,
+                    n_experts,
+                )
+            )
+            shared_moe = _lora_mlp_elements(
+                hd,
+                _shared_expert_size(arch),
+                r,
+                selected_modules,
+                arch.n_shared_experts,
+            )
+            moe_mlp = routed_moe + shared_moe
+            if arch.moe_has_dense_mlp:
+                # why: parallel dense MLP coexists with MoE on every layer.
+                mlp_total = structured_dense_mlp + moe_mlp * n_moe
+            else:
+                dense_only = sum(
+                    value
+                    for i, value in enumerate(per_layer_dense_mlp)
+                    if _is_dense_mlp_layer(arch, i)
+                )
+                mlp_total = moe_mlp * n_moe + dense_only
+        else:
+            mlp_total = structured_dense_mlp
+        return (
+            attn_total
+            + mlp_total
+            + _per_layer_input_lora_params(arch, r, target_modules)
+        )
+    elif n_experts > 1:
+        attn_total = _lora_attn_elements(arch, r, selected_modules) * n_layers
         n_dense = arch.num_dense_layers
         n_moe = n_layers - n_dense
-        # Include shared experts alongside routed experts
-        moe_expert_mult = n_experts + arch.n_shared_experts
-        moe_mlp = _lora_mlp_elements(
-            hd,
-            _get_mlp_size(arch),
-            r,
-            target_modules,
-            moe_expert_mult,
+        # why: routed and shared experts may use different intermediate sizes
+        # (Qwen3.5-MoE: routed mlp_size != shared_expert_intermediate_size).
+        # See structured branch for the all-linear exclusion rationale; only
+        # routed (nn.Parameter) experts are excluded under all-linear.
+        routed_moe = (
+            0
+            if all_linear
+            else _lora_mlp_elements(
+                hd,
+                _get_mlp_size(arch),
+                r,
+                selected_modules,
+                n_experts,
+            )
         )
+        shared_moe = _lora_mlp_elements(
+            hd,
+            _shared_expert_size(arch),
+            r,
+            selected_modules,
+            arch.n_shared_experts,
+        )
+        moe_mlp = routed_moe + shared_moe
         dense_mlp = _lora_mlp_elements(
             hd,
-            arch.intermediate_size,
+            _dense_mlp_size(arch),
             r,
-            target_modules,
+            selected_modules,
             1,
         )
-        mlp_total = moe_mlp * n_moe + dense_mlp * n_dense
+        if arch.moe_has_dense_mlp:
+            mlp_total = moe_mlp * n_moe + dense_mlp * n_layers
+        else:
+            mlp_total = moe_mlp * n_moe + dense_mlp * n_dense
     else:
+        attn_total = _lora_attn_elements(arch, r, selected_modules) * n_layers
         mlp_total = (
             _lora_mlp_elements(
                 hd,
-                arch.intermediate_size,
+                _dense_mlp_size(arch),
                 r,
-                target_modules,
+                selected_modules,
                 1,
             )
             * n_layers
         )
 
-    return attn_total + mlp_total
+    return (
+        attn_total + mlp_total + _per_layer_input_lora_params(arch, r, target_modules)
+    )
 
 
 def compute_lora_adapter_bytes(lora_params: int) -> int:
@@ -419,26 +1137,88 @@ def compute_gradient_bytes(trainable_params: int) -> int:
     return trainable_params * 2
 
 
+def _is_linear_attention(attention_implementation: Optional[str]) -> bool:
+    # why: PyTorch SDPA dispatches to flash/memory-efficient O(n) backends; only
+    # eager (and other non-flash impls) need the quadratic correction.
+    return attention_implementation in LINEAR_ATTENTION_IMPLS
+
+
+def _compute_non_flash_attention_bytes(
+    arch: ModelArchConfig,
+    batch_size: int,
+    seq_len: int,
+    effective_layers: float,
+) -> int:
+    score_elements = batch_size * arch.num_attention_heads * seq_len * seq_len
+    return int(score_elements * 2 * NON_FLASH_ATTENTION_FACTOR * effective_layers)
+
+
+def _layer_qkv_mlp_sizes(arch: ModelArchConfig, layer_idx: int) -> tuple:
+    n_experts = _get_num_experts(arch)
+    is_moe_layer = n_experts > 1 and not _is_dense_mlp_layer(arch, layer_idx)
+    if _uses_structured_layer_shapes(arch):
+        q_size, kv_size, _has_k, _has_v = _layer_attention_dims(arch, layer_idx)
+        # why: KV-shared layers (Gemma4/Gemma3n) drop k_proj/v_proj WEIGHTS but
+        # the donor layer's K/V tensors stay alive across the shared range, so
+        # activation memory still pays for kv_size; only the weight path uses
+        # has_k/has_v.
+        layer_type = _layer_types(arch)[layer_idx]
+        use_alt_attention = arch.attention_k_eq_v and layer_type != "sliding_attention"
+        kv_count = 1 if use_alt_attention else 2
+        qkv_size = q_size + kv_size * kv_count
+        if is_moe_layer:
+            # why: each token routes through `num_experts_per_tok` experts; their
+            # gate/up/down intermediates are all live during MLP forward.
+            mlp_size = _get_mlp_size(arch) * arch.num_experts_per_tok
+            if arch.n_shared_experts:
+                mlp_size += _shared_expert_size(arch) * arch.n_shared_experts
+            if arch.moe_has_dense_mlp:
+                mlp_size += _layer_mlp_size(arch, layer_idx)
+        else:
+            mlp_size = _layer_mlp_size(arch, layer_idx)
+        return qkv_size, mlp_size
+    kv_size = _get_kv_size(arch)
+    if is_moe_layer:
+        mlp_size = _get_mlp_size(arch) * arch.num_experts_per_tok
+        if arch.n_shared_experts:
+            mlp_size += _shared_expert_size(arch) * arch.n_shared_experts
+        if arch.moe_has_dense_mlp:
+            mlp_size += arch.intermediate_size
+    else:
+        mlp_size = _get_mlp_size(arch)
+    return arch.hidden_size + kv_size + kv_size, mlp_size
+
+
+def _per_layer_activation_bytes(
+    arch: ModelArchConfig,
+    layer_idx: int,
+    batch_size: int,
+    seq_len: int,
+) -> int:
+    qkv_size, mlp_size = _layer_qkv_mlp_sizes(arch, layer_idx)
+    activation_qkv = seq_len * batch_size * qkv_size
+    residual_memory = (seq_len * batch_size) * 2
+    activation_mlp = seq_len * batch_size * (mlp_size + mlp_size)
+    # why: per_layer_input_gate (hd-sized) and per_layer_projection (pli-sized)
+    # outputs materialize once per decoder layer when hidden_size_per_layer_input
+    # is set; see gemma4/modular_gemma4.py:1141-1145.
+    pli = arch.hidden_size_per_layer_input
+    activation_ple = seq_len * batch_size * (arch.hidden_size + pli) if pli > 0 else 0
+    return int(
+        (activation_qkv + residual_memory + activation_mlp + activation_ple) * 2 * 1.25
+    )
+
+
 def compute_activation_bytes(
     arch: ModelArchConfig,
     batch_size: int,
     seq_len: int,
     gradient_checkpointing: str,
     is_lora: bool = False,
+    attention_implementation: Optional[str] = "flash_attention_2",
 ) -> int:
-    hd = arch.hidden_size
-    kv_size = _get_kv_size(arch)
-    mlp_size = _get_mlp_size(arch)
-    bsz = batch_size
     n_layers = arch.num_hidden_layers
 
-    activation_qkv = seq_len * bsz * (hd + kv_size + kv_size)
-    residual_memory = (seq_len * bsz) * 2
-    activation_mlp = seq_len * bsz * (mlp_size + mlp_size)
-
-    per_layer_bytes = (activation_qkv + residual_memory + activation_mlp) * 2
-    per_layer_bytes = int(per_layer_bytes * 1.25)
-
     gc_key = gradient_checkpointing.lower()
     gc_entry = GC_LAYER_MULTIPLIERS.get(gc_key, (None, None))
     full_ft_mult, lora_mult = gc_entry
@@ -446,10 +1226,35 @@ def compute_activation_bytes(
 
     if gc_multiplier is None:
         effective_layers = n_layers
+        linear_bytes = sum(
+            _per_layer_activation_bytes(arch, i, batch_size, seq_len)
+            for i in range(n_layers)
+        )
     else:
         effective_layers = gc_multiplier
+        max_layer_bytes = max(
+            _per_layer_activation_bytes(arch, i, batch_size, seq_len)
+            for i in range(n_layers)
+        )
+        linear_bytes = int(max_layer_bytes * effective_layers)
 
-    return int(per_layer_bytes * effective_layers)
+    # why: gemma4 per_layer_model_projection runs once outside the per-decoder
+    # loop and materializes a [B, S, L, PLI] tensor; see modular_gemma4.py:1247.
+    pli = arch.hidden_size_per_layer_input
+    if pli > 0:
+        linear_bytes += int(seq_len * batch_size * n_layers * pli * 2 * 1.25)
+
+    if _is_linear_attention(attention_implementation):
+        return linear_bytes
+    return max(
+        linear_bytes,
+        _compute_non_flash_attention_bytes(
+            arch,
+            batch_size,
+            seq_len,
+            effective_layers,
+        ),
+    )
 
 
 def estimate_training_vram(
@@ -474,21 +1279,23 @@ def estimate_training_vram(
 
     trainable_params = lora_params if is_lora else compute_total_params(arch)
     optimizer_bytes = compute_optimizer_bytes(trainable_params, config.optimizer)
-    gradient_bytes = max(
-        compute_gradient_bytes(trainable_params),
-        int(model_weights * 0.15),
-    )
     activations_computed = compute_activation_bytes(
         arch,
         config.batch_size,
         config.max_seq_length,
         config.gradient_checkpointing,
         is_lora = is_lora,
+        attention_implementation = config.attention_implementation,
     )
-    activation_bytes = max(
-        activations_computed,
-        int(model_weights * 0.15 * (config.batch_size / 2)),
-    )
+    raw_gradient_bytes = compute_gradient_bytes(trainable_params)
+    gradient_floor = int(model_weights * 0.15)
+    if is_lora:
+        gradient_floor = min(
+            gradient_floor,
+            max(activations_computed, optimizer_bytes),
+        )
+    gradient_bytes = max(raw_gradient_bytes, gradient_floor)
+    activation_bytes = activations_computed
 
     return VramBreakdown(
         model_weights = model_weights,

From 680d43a488362f9efa380d27176a220ad25d3cc2 Mon Sep 17 00:00:00 2001
From: Etherll <61019402+Etherll@users.noreply.github.com>
Date: Tue, 5 May 2026 14:15:54 +0300
Subject: [PATCH 208/216] Fix FastSentenceTransformer loading with newer
 sentence-transformers (#5259)

* Fix FastSentenceTransformer compatibility with sentence-transformers 5.4

* Support varied Transformer init signatures

Detect Transformer.__init__ parameters and build init kwargs accordingly so trust_remote_code and other args are passed using the correct names. Instead of unconditionally using model_args/config_args, the code now inspects the constructor to decide between model_kwargs/config_kwargs vs model_args/config_args and also sets processor_kwargs or tokenizer_args when present. Initializes Transformer with constructed transformer_kwargs (including max_seq_length) to improve compatibility with different Transformer implementations.

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

* Harden SentenceTransformer path and module checks

* Scrub .github/workflows for staging push (matches staging base)

* Guard auto_model write in FastSentenceTransformer._apply_torch_compile

On sentence-transformers >=5.4 Transformer.auto_model is a read-only
@property backed by self.model, so a direct assignment raises
AttributeError. The two get_peft_model paths already guard the write
with isinstance(getattr(type(...), "auto_model", None), property);
the auto-compile path missed the same guard, which broke the default
trainer path whenever max_steps >= _compile_threshold.

* Add tests for FastSentenceTransformer property guards

* Tighten FastSentenceTransformer redirect lifecycle tests

Drop a duplicate assertion-less case, remove dead AST extraction helper,
and trim unused imports. The remaining six tests cover substitution on
match, restoration on constructor exception, passthrough for unrelated
names, pathlib.Path normalisation, trailing slash handling, and the
no-identifier guard.

* Sync .github/workflows with upstream author branch

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

* Avoid sharing trust_remote_code kwargs dict across constructor buckets

In FastSentenceTransformer._create_transformer_module, the same
trust_remote_code_kwargs dict was being assigned to model_kwargs,
config_kwargs, and processor_kwargs (or model_args / config_args /
tokenizer_args) on the Transformer constructor. transformers'
from_pretrained code paths (configuration_utils, auto_factory,
processing_auto, etc.) call kwargs.pop("trust_remote_code", ...) on
the dict they receive, which would drain the shared object and silently
strip trust_remote_code from the other buckets. Pass an independent
copy to each bucket so subsequent buckets and any pass-through
auxiliary loads still see trust_remote_code.

* Wire do_lower_case and return_dict through Transformer init for ST 5.4

In FastSentenceTransformer._create_transformer_module:

- When Transformer.__init__ accepts do_lower_case (ST 5.4+), pass
  the unsloth tokenizer's do_lower_case as a constructor kwarg. The
  existing post-init attribute assignment alone is too late: ST 5.4's
  __init__ uses do_lower_case to install a Lowercase normalizer on
  tokenizer.backend_tokenizer.normalizer, which is not re-applied if
  we only set the attribute after construction. The post-init line
  is preserved untouched for older ST versions.

- Add return_dict to the manually completed model_forward_params set
  so wrapped models with forward(*args, **kwargs) signatures keep ST's
  forced dict-like output safety net. ST 5.4's own __init__ unions the
  forward signature with the same set plus return_dict; the previous
  override silently dropped it.

* Preserve flash-attention forward keys when wrapping ST 5.4 Transformer

Sentence-transformers 5.4's Transformer.__init__ calls
_can_flatten_inputs() during construction, which augments
self.model_forward_params with cu_seq_lens_q, cu_seq_lens_k,
max_length_q, max_length_k, seq_idx whenever feature-extraction with
text modality, the torch backend, flash-attention 2, and varlen
flash-attn support are all available. The post-init override of
transformer_module.model_forward_params used to replace the attribute
outright, silently dropping those keys so ST's preprocess() filter
stripped flash-attn kwargs before reaching model.forward.

Snapshot the constructor-populated set first, leave the existing
overwrite intact for the forward-signature plus tokenizer keys, and
union the snapshot back in so flash-attn forwarding keeps working on
ST 5.4. For older sentence-transformers releases the attribute is
absent and getattr returns an empty set, leaving behavior unchanged.

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

---------

Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com>
Co-authored-by: Daniel Han <danielhanchen@gmail.com>
---
 ...sentence_transformer_redirect_lifecycle.py | 218 ++++++++++++++++++
 unsloth/models/sentence_transformer.py        | 187 +++++++++++++--
 2 files changed, 381 insertions(+), 24 deletions(-)
 create mode 100644 tests/python/test_fast_sentence_transformer_redirect_lifecycle.py

diff --git a/tests/python/test_fast_sentence_transformer_redirect_lifecycle.py b/tests/python/test_fast_sentence_transformer_redirect_lifecycle.py
new file mode 100644
index 0000000000..ff9b91ec23
--- /dev/null
+++ b/tests/python/test_fast_sentence_transformer_redirect_lifecycle.py
@@ -0,0 +1,218 @@
+"""FastSentenceTransformer constructor-redirect lifecycle:
+- AutoModel/AutoProcessor/AutoTokenizer.from_pretrained are restored even
+  when the Transformer constructor raises (try/finally invariant).
+- The closure that decides whether to substitute the pre-loaded objects
+  (`is_requested_model_name`) handles HF repo IDs, local paths, trailing
+  slashes, pathlib.Path objects, and missing identifiers correctly.
+"""
+
+from __future__ import annotations
+
+import os
+import pathlib
+import sys
+import types
+
+
+class _FakeAuto:
+    def __init__(self, name):
+        self.name = name
+        self.from_pretrained = self._original
+
+    def _original(self, *args, **kwargs):
+        return ("orig", self.name, args, kwargs)
+
+
+class _RecordingTransformerOk:
+    last_calls = None
+
+    def __init__(self, model_name, **kwargs):
+        from transformers import AutoModel, AutoProcessor, AutoTokenizer
+
+        type(self).last_calls = {
+            "model": AutoModel.from_pretrained(model_name),
+            "processor": AutoProcessor.from_pretrained(model_name),
+            "tokenizer": AutoTokenizer.from_pretrained(model_name),
+        }
+
+
+class _RaisingTransformer:
+    def __init__(self, *a, **kw):
+        from transformers import AutoModel
+
+        AutoModel.from_pretrained(a[0] if a else kw.get("model_name_or_path"))
+        raise RuntimeError("simulated init failure")
+
+
+def _build_driver(transformer_class):
+    transformers_mod = types.ModuleType("transformers")
+    transformers_mod.AutoModel = _FakeAuto("AutoModel")
+    transformers_mod.AutoProcessor = _FakeAuto("AutoProcessor")
+    transformers_mod.AutoTokenizer = _FakeAuto("AutoTokenizer")
+    sys.modules["transformers"] = transformers_mod
+
+    st_root = types.ModuleType("sentence_transformers")
+    st_models = types.ModuleType("sentence_transformers.models")
+    st_models.Transformer = transformer_class
+    sys.modules["sentence_transformers"] = st_root
+    sys.modules["sentence_transformers.models"] = st_models
+
+    captured = {"calls": None}
+
+    def driver(model_name, model, tokenizer):
+        from transformers import AutoModel, AutoProcessor, AutoTokenizer
+        from sentence_transformers.models import Transformer
+
+        def is_requested_model_name(args, kwargs):
+            requested = None
+            if args:
+                requested = args[0]
+            else:
+                requested = kwargs.get("pretrained_model_name_or_path")
+                if requested is None:
+                    requested = kwargs.get("model_name_or_path")
+            if requested is None:
+                return False
+            try:
+                requested = os.fspath(requested)
+                expected = os.fspath(model_name)
+            except (TypeError, ValueError):
+                return False
+            if requested == expected:
+                return True
+            try:
+                if os.path.exists(requested) or os.path.exists(expected):
+                    return os.path.abspath(requested) == os.path.abspath(expected)
+            except (OSError, TypeError, ValueError):
+                pass
+            return False
+
+        original_model = AutoModel.from_pretrained
+        original_processor = AutoProcessor.from_pretrained
+        original_tokenizer = AutoTokenizer.from_pretrained
+
+        def return_existing_model(*a, **kw):
+            return model if is_requested_model_name(a, kw) else original_model(*a, **kw)
+
+        def return_existing_tokenizer(*a, **kw):
+            return (
+                tokenizer
+                if is_requested_model_name(a, kw)
+                else original_tokenizer(*a, **kw)
+            )
+
+        def return_existing_processor(*a, **kw):
+            return (
+                tokenizer
+                if is_requested_model_name(a, kw)
+                else original_processor(*a, **kw)
+            )
+
+        try:
+            AutoModel.from_pretrained = return_existing_model
+            AutoProcessor.from_pretrained = return_existing_processor
+            AutoTokenizer.from_pretrained = return_existing_tokenizer
+            t = Transformer(model_name)
+            captured["calls"] = getattr(type(t), "last_calls", None)
+            return t
+        finally:
+            AutoModel.from_pretrained = original_model
+            AutoProcessor.from_pretrained = original_processor
+            AutoTokenizer.from_pretrained = original_tokenizer
+
+    return driver, transformers_mod, captured
+
+
+def test_redirect_substitutes_preloaded_objects_on_match():
+    driver, _mod, captured = _build_driver(_RecordingTransformerOk)
+    sentinel_model = object()
+    sentinel_tok = object()
+    driver("sentence-transformers/all-MiniLM-L6-v2", sentinel_model, sentinel_tok)
+    calls = captured["calls"]
+    assert calls["model"] is sentinel_model
+    assert calls["processor"] is sentinel_tok
+    assert calls["tokenizer"] is sentinel_tok
+
+
+def test_redirect_restored_on_constructor_exception():
+    driver, transformers_mod, _ = _build_driver(_RaisingTransformer)
+    pre_model = transformers_mod.AutoModel.from_pretrained
+    pre_processor = transformers_mod.AutoProcessor.from_pretrained
+    pre_tokenizer = transformers_mod.AutoTokenizer.from_pretrained
+
+    try:
+        driver("model-id", object(), object())
+    except RuntimeError:
+        pass
+
+    assert transformers_mod.AutoModel.from_pretrained is pre_model
+    assert transformers_mod.AutoProcessor.from_pretrained is pre_processor
+    assert transformers_mod.AutoTokenizer.from_pretrained is pre_tokenizer
+
+
+def test_redirect_passes_through_for_other_model_names():
+    class _OtherNameTransformer:
+        captured = None
+
+        def __init__(self, model_name, **kw):
+            from transformers import AutoModel
+
+            type(self).captured = AutoModel.from_pretrained("some-other/aux-model")
+
+    driver, *_ = _build_driver(_OtherNameTransformer)
+    sentinel = object()
+    driver("primary/model-id", sentinel, object())
+    assert _OtherNameTransformer.captured is not sentinel
+    assert isinstance(_OtherNameTransformer.captured, tuple)
+    assert _OtherNameTransformer.captured[0] == "orig"
+
+
+def test_is_requested_model_name_handles_pathlib_path(tmp_path):
+    target = tmp_path / "model_dir"
+    target.mkdir()
+
+    class _PathTransformer:
+        last_calls = None
+
+        def __init__(self, model_name, **kw):
+            from transformers import AutoModel
+
+            type(self).last_calls = AutoModel.from_pretrained(pathlib.Path(model_name))
+
+    driver, *_ = _build_driver(_PathTransformer)
+    sentinel_model = object()
+    driver(str(target), sentinel_model, object())
+    assert _PathTransformer.last_calls is sentinel_model
+
+
+def test_is_requested_model_name_trailing_slash_local_path(tmp_path):
+    target = tmp_path / "model_dir"
+    target.mkdir()
+
+    class _SlashTransformer:
+        last_calls = None
+
+        def __init__(self, model_name, **kw):
+            from transformers import AutoModel
+
+            type(self).last_calls = AutoModel.from_pretrained(str(target) + "/")
+
+    driver, *_ = _build_driver(_SlashTransformer)
+    sentinel_model = object()
+    driver(str(target), sentinel_model, object())
+    assert _SlashTransformer.last_calls is sentinel_model
+
+
+def test_is_requested_model_name_returns_false_when_no_identifier():
+    captured = {"args": None}
+
+    class _NoNameTransformer:
+        def __init__(self, model_name, **kw):
+            from transformers import AutoModel
+
+            captured["args"] = AutoModel.from_pretrained(some_other_kwarg = "x")
+
+    driver, *_ = _build_driver(_NoNameTransformer)
+    driver("primary/model-id", object(), object())
+    assert isinstance(captured["args"], tuple)
+    assert captured["args"][0] == "orig"
diff --git a/unsloth/models/sentence_transformer.py b/unsloth/models/sentence_transformer.py
index 541875a3fc..52ff268314 100644
--- a/unsloth/models/sentence_transformer.py
+++ b/unsloth/models/sentence_transformer.py
@@ -23,6 +23,7 @@ from ._utils import (
 import inspect
 import json
 import os
+import threading
 import types
 from huggingface_hub import hf_hub_download
 from typing import Optional
@@ -42,6 +43,9 @@ import contextlib
 import shutil
 
 
+_CREATE_TRANSFORMER_MODULE_LOCK = threading.RLock()
+
+
 def _save_pretrained_torchao(
     self,
     save_directory,
@@ -1012,37 +1016,128 @@ class FastSentenceTransformer(FastModel):
         from sentence_transformers.models import Transformer
 
         # prevents sentence-transformers from loading the model a second time, thanks Etherl
-        original_from_pretrained = AutoModel.from_pretrained
+        # Also redirect AutoProcessor / AutoTokenizer so the Transformer.__init__
+        # picks up our pre-fixed tokenizer. On sentence-transformers >=5.4 the
+        # `tokenizer` attribute became a read-only @property backed by `self.processor`,
+        # so a post-init assignment raises AttributeError; redirecting the constructor's
+        # AutoProcessor.from_pretrained call sets self.processor correctly and keeps
+        # downstream state (input_formatter) consistent.
+        from transformers import AutoProcessor, AutoTokenizer
 
-        def return_existing_model(*args, **kwargs):
-            return model
+        def is_requested_model_name(args, kwargs):
+            requested = None
+            if args:
+                requested = args[0]
+            else:
+                requested = kwargs.get("pretrained_model_name_or_path")
+                if requested is None:
+                    requested = kwargs.get("model_name_or_path")
+            if requested is None:
+                return False
 
-        try:
-            # Temporarily redirect AutoModel loading to return our pre-loaded model
-            AutoModel.from_pretrained = return_existing_model
+            try:
+                requested = os.fspath(requested)
+                expected = os.fspath(model_name)
+            except (TypeError, ValueError) as exception:
+                logging.debug(
+                    "Unsloth: Could not normalize SentenceTransformer model path: %s",
+                    exception,
+                )
+                return False
+            if requested == expected:
+                return True
 
-            # Initialize Transformer
-            transformer_module = Transformer(
-                model_name,
-                max_seq_length = max_seq_length,
-                model_args = {"trust_remote_code": trust_remote_code},
-                config_args = {"trust_remote_code": trust_remote_code},
-            )
-        finally:
-            # Restore original functionality immediately
-            AutoModel.from_pretrained = original_from_pretrained
+            try:
+                if os.path.exists(requested) or os.path.exists(expected):
+                    return os.path.abspath(requested) == os.path.abspath(expected)
+            except (OSError, TypeError, ValueError) as exception:
+                logging.debug(
+                    "Unsloth: Could not compare SentenceTransformer model paths: %s",
+                    exception,
+                )
+            return False
 
-        transformer_module.tokenizer = tokenizer
+        with _CREATE_TRANSFORMER_MODULE_LOCK:
+            original_model_from_pretrained = AutoModel.from_pretrained
+            original_processor_from_pretrained = AutoProcessor.from_pretrained
+            original_tokenizer_from_pretrained = AutoTokenizer.from_pretrained
+
+            def return_existing_model(*args, **kwargs):
+                if is_requested_model_name(args, kwargs):
+                    return model
+                return original_model_from_pretrained(*args, **kwargs)
+
+            def return_existing_tokenizer(*args, **kwargs):
+                if is_requested_model_name(args, kwargs):
+                    return tokenizer
+                return original_tokenizer_from_pretrained(*args, **kwargs)
+
+            def return_existing_processor(*args, **kwargs):
+                if is_requested_model_name(args, kwargs):
+                    return tokenizer
+                return original_processor_from_pretrained(*args, **kwargs)
+
+            try:
+                # Temporarily redirect Auto* loading to return our pre-loaded objects
+                AutoModel.from_pretrained = return_existing_model
+                AutoProcessor.from_pretrained = return_existing_processor
+                AutoTokenizer.from_pretrained = return_existing_tokenizer
+
+                transformer_init_params = inspect.signature(
+                    Transformer.__init__
+                ).parameters
+                trust_remote_code_kwargs = {"trust_remote_code": trust_remote_code}
+                do_lower_case = getattr(tokenizer, "do_lower_case", False)
+                transformer_kwargs = {"max_seq_length": max_seq_length}
+                if "do_lower_case" in transformer_init_params:
+                    transformer_kwargs["do_lower_case"] = do_lower_case
+                if "model_kwargs" in transformer_init_params:
+                    transformer_kwargs["model_kwargs"] = trust_remote_code_kwargs.copy()
+                    transformer_kwargs["config_kwargs"] = (
+                        trust_remote_code_kwargs.copy()
+                    )
+                else:
+                    transformer_kwargs["model_args"] = trust_remote_code_kwargs.copy()
+                    transformer_kwargs["config_args"] = trust_remote_code_kwargs.copy()
+                if "processor_kwargs" in transformer_init_params:
+                    transformer_kwargs["processor_kwargs"] = (
+                        trust_remote_code_kwargs.copy()
+                    )
+                elif "tokenizer_args" in transformer_init_params:
+                    transformer_kwargs["tokenizer_args"] = (
+                        trust_remote_code_kwargs.copy()
+                    )
+
+                # Initialize Transformer
+                transformer_module = Transformer(model_name, **transformer_kwargs)
+            finally:
+                # Restore original functionality immediately
+                AutoModel.from_pretrained = original_model_from_pretrained
+                AutoProcessor.from_pretrained = original_processor_from_pretrained
+                AutoTokenizer.from_pretrained = original_tokenizer_from_pretrained
+
+        # On sentence-transformers >=5.4 `tokenizer` is a read-only property backed
+        # by `self.processor` (already wired via the redirect above). On older
+        # versions it's a regular attribute and the explicit assignment is required.
+        if not isinstance(
+            getattr(type(transformer_module), "tokenizer", None), property
+        ):
+            transformer_module.tokenizer = tokenizer
         transformer_module.do_lower_case = getattr(tokenizer, "do_lower_case", False)
 
         # sentence-transformers only passes along known keys to model.forward
+        preinit_model_forward_params = getattr(
+            transformer_module, "model_forward_params", set()
+        )
         model_forward_params = list(inspect.signature(model.forward).parameters)
         transformer_module.model_forward_params = set(model_forward_params) | {
             "input_ids",
             "attention_mask",
             "token_type_ids",
             "inputs_embeds",
+            "return_dict",
         }
+        transformer_module.model_forward_params |= preinit_model_forward_params
 
         # determine max_seq_length if not provided
         if max_seq_length is None:
@@ -1056,7 +1151,11 @@ class FastSentenceTransformer(FastModel):
                 max_seq_length = 512
 
         transformer_module.max_seq_length = max_seq_length
-        transformer_module.config_keys = ["max_seq_length", "do_lower_case"]
+        config_keys = list(getattr(transformer_module, "config_keys", []) or [])
+        for config_key in ("max_seq_length", "do_lower_case"):
+            if config_key not in config_keys:
+                config_keys.append(config_key)
+        transformer_module.config_keys = config_keys
         transformer_module.save_in_root = True
 
         if hasattr(model, "config"):
@@ -1064,6 +1163,29 @@ class FastSentenceTransformer(FastModel):
 
         return transformer_module
 
+    @staticmethod
+    def _is_transformer_module_ref(class_ref):
+        if class_ref in {
+            "sentence_transformers.models.Transformer",
+            "sentence_transformers.models.transformer.Transformer",
+            "sentence_transformers.base.modules.transformer.Transformer",
+        }:
+            return True
+
+        try:
+            from sentence_transformers.models import Transformer
+            from sentence_transformers.util import import_from_string
+
+            module_class = import_from_string(class_ref)
+            return module_class is Transformer
+        except (ImportError, AttributeError, TypeError, ValueError) as exception:
+            logging.debug(
+                "Unsloth: Could not resolve SentenceTransformer module ref %r: %s",
+                class_ref,
+                exception,
+            )
+            return False
+
     @staticmethod
     def _load_modules(
         model_name,
@@ -1096,7 +1218,7 @@ class FastSentenceTransformer(FastModel):
                     "name", str(module_config.get("idx", len(modules)))
                 )
 
-                if class_ref == "sentence_transformers.models.Transformer":
+                if FastSentenceTransformer._is_transformer_module_ref(class_ref):
                     transformer_module = (
                         FastSentenceTransformer._create_transformer_module(
                             model_name,
@@ -1299,7 +1421,10 @@ class FastSentenceTransformer(FastModel):
         if hasattr(model, "__getitem__"):
             inner_model = model[0].auto_model
             compiled = torch.compile(inner_model, mode = mode)
-            model[0].auto_model = compiled
+            if isinstance(getattr(type(model[0]), "auto_model", None), property):
+                model[0].model = compiled
+            else:
+                model[0].auto_model = compiled
             # Fix for accelerate unwrap_model bug:
             # When SentenceTransformer contains a compiled inner model,
             # accelerate checks has_compiled_regions() which returns True,
@@ -1933,8 +2058,15 @@ class FastSentenceTransformer(FastModel):
                         "Unsloth: Re-enabling torch.compile since gradient checkpointing is not supported"
                     )
 
-                # Re-assign the peft model back to the transformer module
-                transformer_module.auto_model = peft_model
+                # Re-assign the peft model back to the transformer module.
+                # On sentence-transformers >=5.4 `auto_model` is a read-only property
+                # backed by `self.model`, so write to the backing attribute there.
+                if isinstance(
+                    getattr(type(transformer_module), "auto_model", None), property
+                ):
+                    transformer_module.model = peft_model
+                else:
+                    transformer_module.auto_model = peft_model
 
                 # Store compile info for auto-compile at trainer time
                 # torch.compile is deferred until training starts so we can check max_steps
@@ -1980,8 +2112,15 @@ class FastSentenceTransformer(FastModel):
                 **kwargs,
             )
 
-            # re-assign the peft model back to the transformer module
-            transformer_module.auto_model = peft_model
+            # re-assign the peft model back to the transformer module.
+            # On sentence-transformers >=5.4 `auto_model` is a read-only property
+            # backed by `self.model`, so write to the backing attribute there.
+            if isinstance(
+                getattr(type(transformer_module), "auto_model", None), property
+            ):
+                transformer_module.model = peft_model
+            else:
+                transformer_module.auto_model = peft_model
             return model
         else:
             return FastModel.get_peft_model(

From 19f305238e6126abd7809f61a93453d4adb2fda1 Mon Sep 17 00:00:00 2001
From: Lee Jackson <130007945+Imagineer99@users.noreply.github.com>
Date: Tue, 5 May 2026 12:19:41 +0100
Subject: [PATCH 209/216] Studio: Preserve chat history during autosave (#5278)

* fix: chat recents reopening after new chat

* fix: optimize chat delete pruning query
---
 .../chat/utils/delete-thread-message.ts       | 23 +++++++++++--------
 1 file changed, 13 insertions(+), 10 deletions(-)

diff --git a/studio/frontend/src/features/chat/utils/delete-thread-message.ts b/studio/frontend/src/features/chat/utils/delete-thread-message.ts
index e13da60630..d410c37a34 100644
--- a/studio/frontend/src/features/chat/utils/delete-thread-message.ts
+++ b/studio/frontend/src/features/chat/utils/delete-thread-message.ts
@@ -71,21 +71,24 @@ function exportedItemToRecord(
 }
 
 /**
- * Persist the exact message list represented by `exp` for this thread, removing
- * Dexie rows that are no longer present (e.g. after a delete).
+ * Persist exported messages, pruning only for explicit delete flows.
  */
 export async function syncExportedRepositoryToDexie(
   remoteId: string,
   exp: ExportedMessageRepository,
+  options: { pruneMissing?: boolean } = {},
 ): Promise<void> {
   await db.transaction("rw", db.messages, async () => {
-    const keepIds = new Set(exp.messages.map((x) => x.message.id));
-    const existing = await db.messages.where("threadId").equals(remoteId).toArray();
-    const idsToDelete = existing
-      .filter((m) => !keepIds.has(m.id))
-      .map((m) => m.id);
-    if (idsToDelete.length > 0) {
-      await db.messages.bulkDelete(idsToDelete);
+    if (options.pruneMissing) {
+      const keepIds = new Set(exp.messages.map((x) => x.message.id));
+      const existingIds = await db.messages
+        .where("threadId")
+        .equals(remoteId)
+        .primaryKeys();
+      const idsToDelete = existingIds.filter((id) => !keepIds.has(String(id)));
+      if (idsToDelete.length > 0) {
+        await db.messages.bulkDelete(idsToDelete);
+      }
     }
     await db.messages.bulkPut(
       exp.messages.map(({ message, parentId }) =>
@@ -115,7 +118,7 @@ export async function deleteThreadMessage(args: {
   repo.deleteMessage(messageId);
   const next = repo.export();
   if (remoteId) {
-    await syncExportedRepositoryToDexie(remoteId, next);
+    await syncExportedRepositoryToDexie(remoteId, next, { pruneMissing: true });
   }
   thread.import(next);
 }

From be874c72e606e9326f096ab62684b5b502ce9fd6 Mon Sep 17 00:00:00 2001
From: Daniel Han <danielhanchen@gmail.com>
Date: Tue, 5 May 2026 05:06:30 -0700
Subject: [PATCH 210/216] Update pyproject.toml

---
 pyproject.toml | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/pyproject.toml b/pyproject.toml
index 5687ea12f8..c2f884e192 100644
--- a/pyproject.toml
+++ b/pyproject.toml
@@ -89,7 +89,7 @@ huggingfacenotorch = [
 ]
 huggingface = [
     "unsloth[huggingfacenotorch]",
-    "unsloth_zoo>=2026.4.8",
+    "unsloth_zoo>=2026.5.1",
     "torchvision",
     "unsloth[triton]",
 ]
@@ -579,7 +579,7 @@ colab-ampere-torch220 = [
     "flash-attn>=2.6.3 ; ('linux' in sys_platform)",
 ]
 colab-new = [
-    "unsloth_zoo>=2026.4.8",
+    "unsloth_zoo>=2026.5.1",
     "packaging",
     "tyro",
     "transformers>=4.51.3,!=4.52.0,!=4.52.1,!=4.52.2,!=4.52.3,!=4.53.0,!=4.54.0,!=4.55.0,!=4.55.1,!=4.57.0,!=4.57.4,!=4.57.5,!=5.0.0,!=5.1.0,<=5.5.0",

From 2fba3b6d9d094fca432e895dc4be91c822f38677 Mon Sep 17 00:00:00 2001
From: Daniel Han <danielhanchen@gmail.com>
Date: Tue, 5 May 2026 05:07:06 -0700
Subject: [PATCH 211/216] Update _utils.py

---
 unsloth/models/_utils.py | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/unsloth/models/_utils.py b/unsloth/models/_utils.py
index e87c2a1a83..1120e1941d 100644
--- a/unsloth/models/_utils.py
+++ b/unsloth/models/_utils.py
@@ -12,7 +12,7 @@
 # See the License for the specific language governing permissions and
 # limitations under the License.
 
-__version__ = "2026.4.8"
+__version__ = "2026.5.1"
 
 __all__ = [
     "SUPPORTS_BFLOAT16",

From d741cc928be7bde731fd84ec07756393352e7a97 Mon Sep 17 00:00:00 2001
From: Lee Jackson <130007945+Imagineer99@users.noreply.github.com>
Date: Tue, 5 May 2026 13:11:52 +0100
Subject: [PATCH 212/216] fix: developer to api (#5281)

---
 studio/frontend/src/components/app-sidebar.tsx              | 6 +++---
 studio/frontend/src/features/settings/settings-dialog.tsx   | 4 ++--
 studio/frontend/src/features/settings/tabs/api-keys-tab.tsx | 2 +-
 3 files changed, 6 insertions(+), 6 deletions(-)

diff --git a/studio/frontend/src/components/app-sidebar.tsx b/studio/frontend/src/components/app-sidebar.tsx
index 171b4eb92d..433e04b3e5 100644
--- a/studio/frontend/src/components/app-sidebar.tsx
+++ b/studio/frontend/src/components/app-sidebar.tsx
@@ -37,12 +37,12 @@ import {
   Delete02Icon,
   Download03Icon,
   GemIcon,
+  Globe02Icon,
   Search01Icon,
   PowerIcon,
   PencilEdit02Icon,
   LayoutAlignLeftIcon,
   Settings02Icon,
-  SourceCodeSquareIcon,
   ZapIcon,
 } from "@hugeicons/core-free-icons";
 import {
@@ -548,8 +548,8 @@ export function AppSidebar() {
                   <DropdownMenuItem
                     onSelect={() => useSettingsDialogStore.getState().openDialog("api-keys")}
                   >
-                    <HugeiconsIcon icon={SourceCodeSquareIcon} strokeWidth={1.75} className="size-[18px]" />
-                    <span>Developer</span>
+                    <HugeiconsIcon icon={Globe02Icon} strokeWidth={1.75} className="size-[18px]" />
+                    <span>API</span>
                     <span className="ml-auto rounded-[6px] border border-emerald-500/25 bg-emerald-500/10 px-1.5 py-0.5 text-[10px] leading-none font-semibold text-emerald-700 dark:text-emerald-300">
                       New
                     </span>
diff --git a/studio/frontend/src/features/settings/settings-dialog.tsx b/studio/frontend/src/features/settings/settings-dialog.tsx
index 376af06e9d..d33069edbb 100644
--- a/studio/frontend/src/features/settings/settings-dialog.tsx
+++ b/studio/frontend/src/features/settings/settings-dialog.tsx
@@ -10,10 +10,10 @@ import {
 import { cn } from "@/lib/utils";
 import {
   Cancel01Icon,
+  Globe02Icon,
   Message01Icon,
   PaintBrush02Icon,
   Settings02Icon,
-  SourceCodeSquareIcon,
   SparklesIcon,
   UserIcon,
 } from "@hugeicons/core-free-icons";
@@ -40,7 +40,7 @@ const TABS: TabDef[] = [
   { id: "profile", label: "Profile", icon: UserIcon },
   { id: "appearance", label: "Appearance", icon: PaintBrush02Icon },
   { id: "chat", label: "Chat", icon: Message01Icon },
-  { id: "api-keys", label: "Developer", icon: SourceCodeSquareIcon, badge: "New" },
+  { id: "api-keys", label: "API", icon: Globe02Icon, badge: "New" },
   { id: "about", label: "Help", icon: SparklesIcon },
 ];
 
diff --git a/studio/frontend/src/features/settings/tabs/api-keys-tab.tsx b/studio/frontend/src/features/settings/tabs/api-keys-tab.tsx
index 64c6f520c0..ac9ec40543 100644
--- a/studio/frontend/src/features/settings/tabs/api-keys-tab.tsx
+++ b/studio/frontend/src/features/settings/tabs/api-keys-tab.tsx
@@ -63,7 +63,7 @@ export function ApiKeysTab() {
   return (
     <div className="flex flex-col gap-6">
       <header className="flex flex-col gap-1">
-        <h1 className="text-lg font-semibold font-heading">Developer</h1>
+        <h1 className="text-lg font-semibold font-heading">API</h1>
         <p className="text-xs text-muted-foreground">
           Access Unsloth programmatically via the OpenAI-compatible API.{" "}
           <a

From d8a0bebbc0d6d547109c2cbab40fa1e626e195b5 Mon Sep 17 00:00:00 2001
From: Lee Jackson <130007945+Imagineer99@users.noreply.github.com>
Date: Tue, 5 May 2026 13:19:56 +0100
Subject: [PATCH 213/216] Studio: help svg replacement and Unsloth sidebar text
 (#5282)

* fix: developer to api

* fix: help svg and Unsloth text

---------

Co-authored-by: Roland Tannous <115670425+rolandtannous@users.noreply.github.com>
---
 studio/frontend/src/components/app-sidebar.tsx | 9 ++++++++-
 1 file changed, 8 insertions(+), 1 deletion(-)

diff --git a/studio/frontend/src/components/app-sidebar.tsx b/studio/frontend/src/components/app-sidebar.tsx
index 433e04b3e5..edcd5120eb 100644
--- a/studio/frontend/src/components/app-sidebar.tsx
+++ b/studio/frontend/src/components/app-sidebar.tsx
@@ -42,6 +42,7 @@ import {
   PowerIcon,
   PencilEdit02Icon,
   LayoutAlignLeftIcon,
+  HelpCircleIcon,
   Settings02Icon,
   ZapIcon,
 } from "@hugeicons/core-free-icons";
@@ -527,7 +528,7 @@ export function AppSidebar() {
                   </div>
                   <div className="flex flex-col gap-0.5 leading-tight group-data-[collapsible=icon]:hidden">
                     <span className="truncate font-heading text-[13px] tracking-[0.02em] font-semibold text-[#383835] dark:text-[#c7c7c4]">{displayTitle}</span>
-                    <span className="truncate text-[11px] tracking-[0.01em] text-muted-foreground">Studio</span>
+                    <span className="truncate text-[11px] tracking-[0.01em] text-muted-foreground">Unsloth</span>
                   </div>
                   <ChevronsUpDown strokeWidth={1.25} className="ml-auto size-4 text-muted-foreground group-data-[collapsible=icon]:hidden" />
                 </SidebarMenuButton>
@@ -578,6 +579,12 @@ export function AppSidebar() {
                   </DropdownMenuItem>
                 </DropdownMenuGroup>
                 <DropdownMenuSeparator className="mx-2.5! my-2.5! h-0! border-t border-border/70 bg-transparent!" />
+                <DropdownMenuItem
+                  onSelect={() => useSettingsDialogStore.getState().openDialog("about")}
+                >
+                  <HugeiconsIcon icon={HelpCircleIcon} strokeWidth={1.75} className="size-[18px]" />
+                  <span>Help</span>
+                </DropdownMenuItem>
                 <DropdownMenuItem onSelect={() => setShutdownOpen(true)}>
                   <HugeiconsIcon icon={PowerIcon} strokeWidth={1.75} className="size-[18px]" />
                   <span>Shutdown</span>

From 832f48c41a6bea8835eac6009cb50acd31bced0c Mon Sep 17 00:00:00 2001
From: Lee Jackson <130007945+Imagineer99@users.noreply.github.com>
Date: Tue, 5 May 2026 13:22:52 +0100
Subject: [PATCH 214/216] Chore/help svg (#5283)

* fix: developer to api

* fix: help svg and Unsloth text

* svg fix

---------

Co-authored-by: Roland Tannous <115670425+rolandtannous@users.noreply.github.com>
---
 studio/frontend/src/features/settings/settings-dialog.tsx | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/studio/frontend/src/features/settings/settings-dialog.tsx b/studio/frontend/src/features/settings/settings-dialog.tsx
index d33069edbb..63c0a9d388 100644
--- a/studio/frontend/src/features/settings/settings-dialog.tsx
+++ b/studio/frontend/src/features/settings/settings-dialog.tsx
@@ -11,10 +11,10 @@ import { cn } from "@/lib/utils";
 import {
   Cancel01Icon,
   Globe02Icon,
+  HelpCircleIcon,
   Message01Icon,
   PaintBrush02Icon,
   Settings02Icon,
-  SparklesIcon,
   UserIcon,
 } from "@hugeicons/core-free-icons";
 import { HugeiconsIcon } from "@hugeicons/react";
@@ -41,7 +41,7 @@ const TABS: TabDef[] = [
   { id: "appearance", label: "Appearance", icon: PaintBrush02Icon },
   { id: "chat", label: "Chat", icon: Message01Icon },
   { id: "api-keys", label: "API", icon: Globe02Icon, badge: "New" },
-  { id: "about", label: "Help", icon: SparklesIcon },
+  { id: "about", label: "Help", icon: HelpCircleIcon },
 ];
 
 function renderTab(tab: SettingsTab) {

From d5918d706ee4741a927de2eebce7b370fe6e640e Mon Sep 17 00:00:00 2001
From: Daniel Han <danielhanchen@gmail.com>
Date: Tue, 5 May 2026 05:38:22 -0700
Subject: [PATCH 215/216] Bump installer floor to 2026.5.1 (#5284)

---
 install.ps1 | 10 +++++-----
 install.sh  | 10 +++++-----
 2 files changed, 10 insertions(+), 10 deletions(-)

diff --git a/install.ps1 b/install.ps1
index 7dc5a50250..a02cef0f6a 100644
--- a/install.ps1
+++ b/install.ps1
@@ -1068,7 +1068,7 @@ shell.Run cmd, 0, False
         if ($SkipTorch) {
             # No-torch: install unsloth + unsloth-zoo with --no-deps, then
             # runtime deps (typer, safetensors, transformers, etc.) with --no-deps.
-            $baseInstallExit = Invoke-InstallCommand { uv pip install --python $VenvPython --no-deps --reinstall-package unsloth --reinstall-package unsloth-zoo "unsloth>=2026.4.8" unsloth-zoo }
+            $baseInstallExit = Invoke-InstallCommand { uv pip install --python $VenvPython --no-deps --reinstall-package unsloth --reinstall-package unsloth-zoo "unsloth>=2026.5.1" unsloth-zoo }
             if ($baseInstallExit -eq 0) {
                 $NoTorchReq = Find-NoTorchRuntimeFile
                 if ($NoTorchReq) {
@@ -1076,7 +1076,7 @@ shell.Run cmd, 0, False
                 }
             }
         } else {
-            $baseInstallExit = Invoke-InstallCommand { uv pip install --python $VenvPython --reinstall-package unsloth --reinstall-package unsloth-zoo "unsloth>=2026.4.8" unsloth-zoo }
+            $baseInstallExit = Invoke-InstallCommand { uv pip install --python $VenvPython --reinstall-package unsloth --reinstall-package unsloth-zoo "unsloth>=2026.5.1" unsloth-zoo }
         }
         if ($baseInstallExit -ne 0) {
             Write-Host "[ERROR] Failed to install unsloth (exit code $baseInstallExit)" -ForegroundColor Red
@@ -1114,7 +1114,7 @@ shell.Run cmd, 0, False
         if ($SkipTorch) {
             # No-torch: install unsloth + unsloth-zoo with --no-deps, then
             # runtime deps (typer, safetensors, transformers, etc.) with --no-deps.
-            $baseInstallExit = Invoke-InstallCommand { uv pip install --python $VenvPython --no-deps --upgrade-package unsloth --upgrade-package unsloth-zoo "unsloth>=2026.4.8" unsloth-zoo }
+            $baseInstallExit = Invoke-InstallCommand { uv pip install --python $VenvPython --no-deps --upgrade-package unsloth --upgrade-package unsloth-zoo "unsloth>=2026.5.1" unsloth-zoo }
             if ($baseInstallExit -eq 0) {
                 $NoTorchReq = Find-NoTorchRuntimeFile
                 if ($NoTorchReq) {
@@ -1122,7 +1122,7 @@ shell.Run cmd, 0, False
                 }
             }
         } elseif ($StudioLocalInstall) {
-            $baseInstallExit = Invoke-InstallCommand { uv pip install --python $VenvPython --upgrade-package unsloth "unsloth>=2026.4.8" unsloth-zoo }
+            $baseInstallExit = Invoke-InstallCommand { uv pip install --python $VenvPython --upgrade-package unsloth "unsloth>=2026.5.1" unsloth-zoo }
         } else {
             $baseInstallExit = Invoke-InstallCommand { uv pip install --python $VenvPython --upgrade-package unsloth -- "$PackageName" }
         }
@@ -1150,7 +1150,7 @@ shell.Run cmd, 0, False
         Write-TauriLog "STEP" "Installing unsloth"
         substep "installing unsloth (this may take a few minutes)..."
         if ($StudioLocalInstall) {
-            $baseInstallExit = Invoke-InstallCommand { uv pip install --python $VenvPython unsloth-zoo "unsloth>=2026.4.8" --torch-backend=auto }
+            $baseInstallExit = Invoke-InstallCommand { uv pip install --python $VenvPython unsloth-zoo "unsloth>=2026.5.1" --torch-backend=auto }
             if ($baseInstallExit -ne 0) {
                 Write-Host "[ERROR] Failed to install unsloth (exit code $baseInstallExit)" -ForegroundColor Red
                 return (Exit-InstallFailure "Failed to install unsloth (exit code $baseInstallExit)" $baseInstallExit)
diff --git a/install.sh b/install.sh
index 7948170043..fc8bd27fb8 100755
--- a/install.sh
+++ b/install.sh
@@ -1486,7 +1486,7 @@ if [ "$_MIGRATED" = true ]; then
         # to prevent transitive torch resolution.
         run_install_cmd "install unsloth (migrated no-torch)" uv pip install --python "$_VENV_PY" --no-deps \
             --reinstall-package unsloth --reinstall-package unsloth-zoo \
-            "unsloth>=2026.4.8" unsloth-zoo
+            "unsloth>=2026.5.1" unsloth-zoo
         _NO_TORCH_RT="$(_find_no_torch_runtime)"
         if [ -n "$_NO_TORCH_RT" ]; then
             run_install_cmd "install no-torch runtime deps" uv pip install --python "$_VENV_PY" --no-deps -r "$_NO_TORCH_RT"
@@ -1494,7 +1494,7 @@ if [ "$_MIGRATED" = true ]; then
     else
         run_install_cmd "install unsloth (migrated)" uv pip install --python "$_VENV_PY" \
             --reinstall-package unsloth --reinstall-package unsloth-zoo \
-            "unsloth>=2026.4.8" unsloth-zoo
+            "unsloth>=2026.5.1" unsloth-zoo
     fi
     if [ "$STUDIO_LOCAL_INSTALL" = true ]; then
         substep "overlaying local repo (editable)..."
@@ -1662,7 +1662,7 @@ elif [ -n "$TORCH_INDEX_URL" ]; then
         # runtime deps (typer, safetensors, transformers, etc.) with --no-deps.
         run_install_cmd "install unsloth (no-torch)" uv pip install --python "$_VENV_PY" --no-deps \
             --upgrade-package unsloth --upgrade-package unsloth-zoo \
-            "unsloth>=2026.4.8" unsloth-zoo
+            "unsloth>=2026.5.1" unsloth-zoo
         _NO_TORCH_RT="$(_find_no_torch_runtime)"
         if [ -n "$_NO_TORCH_RT" ]; then
             run_install_cmd "install no-torch runtime deps" uv pip install --python "$_VENV_PY" --no-deps -r "$_NO_TORCH_RT"
@@ -1677,7 +1677,7 @@ elif [ -n "$TORCH_INDEX_URL" ]; then
         fi
     elif [ "$STUDIO_LOCAL_INSTALL" = true ]; then
         run_install_cmd "install unsloth (local)" uv pip install --python "$_VENV_PY" \
-            --upgrade-package unsloth "unsloth>=2026.4.8" unsloth-zoo
+            --upgrade-package unsloth "unsloth>=2026.5.1" unsloth-zoo
         substep "overlaying local repo (editable)..."
         run_install_cmd "overlay local repo" uv pip install --python "$_VENV_PY" -e "$_REPO_ROOT" --no-deps
         substep "overlaying unsloth-zoo from git main..."
@@ -1709,7 +1709,7 @@ else
     tauri_log "STEP" "Installing Unsloth"
     substep "installing unsloth (this may take a few minutes)..."
     if [ "$STUDIO_LOCAL_INSTALL" = true ]; then
-        run_install_cmd "install unsloth (auto torch backend)" uv pip install --python "$_VENV_PY" unsloth-zoo "unsloth>=2026.4.8" --torch-backend=auto
+        run_install_cmd "install unsloth (auto torch backend)" uv pip install --python "$_VENV_PY" unsloth-zoo "unsloth>=2026.5.1" --torch-backend=auto
         substep "overlaying local repo (editable)..."
         run_install_cmd "overlay local repo" uv pip install --python "$_VENV_PY" -e "$_REPO_ROOT" --no-deps
         substep "overlaying unsloth-zoo from git main..."

From 9bab3b955ec3be719178aa43cd69a51ff479d0af Mon Sep 17 00:00:00 2001
From: Michael Han <107991372+shimmyshimmer@users.noreply.github.com>
Date: Tue, 5 May 2026 06:13:35 -0700
Subject: [PATCH 216/216] Add API Inference endpoint

---
 README.md | 5 +++--
 1 file changed, 3 insertions(+), 2 deletions(-)

diff --git a/README.md b/README.md
index 6c12b28096..a654518d14 100644
--- a/README.md
+++ b/README.md
@@ -43,7 +43,8 @@ Unsloth Studio (Beta) lets you run and train text, [audio](https://unsloth.ai/do
 * **Export models**: [Save or export](https://unsloth.ai/docs/new/studio/export) models to GGUF, 16-bit safetensors and other formats.
 * **Tool calling**: Support for [self-healing tool calling](https://unsloth.ai/docs/new/studio/chat#auto-healing-tool-calling) and web search
 * **[Code execution](https://unsloth.ai/docs/new/studio/chat#code-execution)**: lets LLMs test code in Claude artifacts and sandbox environments
-* [Auto-tune inference parameters](https://unsloth.ai/docs/new/studio/chat#auto-parameter-tuning) and customize chat templates.
+* **[API inference endpoint](https://unsloth.ai/docs/basics/api)**: Deploy and run local LLMs in Claude Code, Codex tools with Unsloth
+* [Auto set inference settings](https://unsloth.ai/docs/new/studio/chat#auto-parameter-tuning) and customize chat templates.
 * We work directly with teams behind [gpt-oss](https://docs.unsloth.ai/new/gpt-oss-how-to-run-and-fine-tune#unsloth-fixes-for-gpt-oss), [Qwen3](https://www.reddit.com/r/LocalLLaMA/comments/1kaodxu/qwen3_unsloth_dynamic_ggufs_128k_context_bug_fixes/), [Llama 4](https://github.com/ggml-org/llama.cpp/pull/12889), [Mistral](models/tutorials/devstral-how-to-run-and-fine-tune.md), [Gemma 1-3](https://news.ycombinator.com/item?id=39671146), and [Phi-4](https://unsloth.ai/blog/phi4), where we’ve fixed bugs that improve model accuracy.
 * Upload images, audio, PDFs, code, DOCX and more file types to chat with.
 ### Training
@@ -149,6 +150,7 @@ Read our [guide](https://unsloth.ai/docs/get-started/fine-tuning-llms-guide). Ad
 - See detailed documentation for Unsloth [here](https://unsloth.ai/docs)
 
 ## 🦥 Unsloth News
+- **API inference endpoint**: Deploy and run local LLMs in Claude Code, Codex tools. [Guide](https://unsloth.ai/docs/basics/api)
 - **Qwen3.6**: Qwen3.6-35B-A3B can now be trained and run in Unsloth Studio. [Blog](https://unsloth.ai/docs/models/qwen3.6)
 - **Gemma 4**: Run and train Google’s new models directly in Unsloth. [Blog](https://unsloth.ai/docs/models/gemma-4)
 - **Introducing Unsloth Studio**: our new web UI for running and training LLMs. [Blog](https://unsloth.ai/docs/new/studio)
@@ -159,7 +161,6 @@ Read our [guide](https://unsloth.ai/docs/get-started/fine-tuning-llms-guide). Ad
 - New RoPE & MLP **Triton Kernels** & **Padding Free + Packing**: 3x faster training & 30% less VRAM. [Blog](https://unsloth.ai/docs/new/3x-faster-training-packing)
 - **500K Context**: Training a 20B model with >500K context is now possible on an 80GB GPU. [Blog](https://unsloth.ai/docs/blog/500k-context-length-fine-tuning)
 - **FP8 & Vision RL**: You can now do FP8 & VLM GRPO on consumer GPUs. [FP8 Blog](https://unsloth.ai/docs/get-started/reinforcement-learning-rl-guide/fp8-reinforcement-learning) • [Vision RL](https://unsloth.ai/docs/get-started/reinforcement-learning-rl-guide/vision-reinforcement-learning-vlm-rl)
-- **gpt-oss** by OpenAI: Read our [RL blog](https://unsloth.ai/docs/models/gpt-oss-how-to-run-and-fine-tune/gpt-oss-reinforcement-learning), [Flex Attention](https://unsloth.ai/docs/models/gpt-oss-how-to-run-and-fine-tune/long-context-gpt-oss-training) blog and [Guide](https://unsloth.ai/docs/models/gpt-oss-how-to-run-and-fine-tune).
 
 ## 📥 Advanced Installation
 The below advanced instructions are for Unsloth Studio. For Unsloth Core advanced installation, [view our docs](https://unsloth.ai/docs/get-started/install/pip-install#advanced-pip-installation).